{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.18463791267305643, "eval_steps": 500, "global_step": 30048, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 2.2172949002217297e-08, "loss": 4.7251, "step": 1 }, { "epoch": 0.0, "learning_rate": 4.4345898004434594e-08, "loss": 4.8129, "step": 2 }, { "epoch": 0.0, "learning_rate": 6.651884700665188e-08, "loss": 4.8132, "step": 3 }, { "epoch": 0.0, "learning_rate": 8.869179600886919e-08, "loss": 4.7702, "step": 4 }, { "epoch": 0.0, "learning_rate": 1.1086474501108649e-07, "loss": 4.6731, "step": 5 }, { "epoch": 0.0, "learning_rate": 1.3303769401330377e-07, "loss": 4.878, "step": 6 }, { "epoch": 0.0, "learning_rate": 1.5521064301552109e-07, "loss": 4.815, "step": 7 }, { "epoch": 0.0, "learning_rate": 1.7738359201773838e-07, "loss": 4.6106, "step": 8 }, { "epoch": 0.0, "learning_rate": 1.9955654101995567e-07, "loss": 4.781, "step": 9 }, { "epoch": 0.0, "learning_rate": 2.2172949002217298e-07, "loss": 4.8475, "step": 10 }, { "epoch": 0.0, "learning_rate": 2.439024390243903e-07, "loss": 4.7899, "step": 11 }, { "epoch": 0.0, "learning_rate": 2.6607538802660754e-07, "loss": 4.7501, "step": 12 }, { "epoch": 0.0, "learning_rate": 2.8824833702882486e-07, "loss": 4.722, "step": 13 }, { "epoch": 0.0, "learning_rate": 3.1042128603104217e-07, "loss": 4.8346, "step": 14 }, { "epoch": 0.0, "learning_rate": 3.3259423503325944e-07, "loss": 4.6773, "step": 15 }, { "epoch": 0.0, "learning_rate": 3.5476718403547675e-07, "loss": 4.527, "step": 16 }, { "epoch": 0.0, "learning_rate": 3.7694013303769407e-07, "loss": 4.6716, "step": 17 }, { "epoch": 0.0, "learning_rate": 3.9911308203991133e-07, "loss": 4.7305, "step": 18 }, { "epoch": 0.0, "learning_rate": 4.2128603104212865e-07, "loss": 4.648, "step": 19 }, { "epoch": 0.0, "learning_rate": 4.4345898004434597e-07, "loss": 4.4968, "step": 20 }, { "epoch": 0.0, "learning_rate": 4.6563192904656323e-07, "loss": 4.4882, "step": 21 }, { "epoch": 0.0, "learning_rate": 4.878048780487805e-07, "loss": 4.4307, "step": 22 }, { "epoch": 0.0, "learning_rate": 5.099778270509978e-07, "loss": 4.4932, "step": 23 }, { "epoch": 0.0, "learning_rate": 5.321507760532151e-07, "loss": 4.3287, "step": 24 }, { "epoch": 0.0, "learning_rate": 5.543237250554324e-07, "loss": 4.3799, "step": 25 }, { "epoch": 0.0, "learning_rate": 5.764966740576497e-07, "loss": 4.4621, "step": 26 }, { "epoch": 0.0, "learning_rate": 5.98669623059867e-07, "loss": 4.1453, "step": 27 }, { "epoch": 0.0, "learning_rate": 6.208425720620843e-07, "loss": 3.7498, "step": 28 }, { "epoch": 0.0, "learning_rate": 6.430155210643016e-07, "loss": 3.9323, "step": 29 }, { "epoch": 0.0, "learning_rate": 6.651884700665189e-07, "loss": 3.7856, "step": 30 }, { "epoch": 0.0, "learning_rate": 6.873614190687362e-07, "loss": 3.8264, "step": 31 }, { "epoch": 0.0, "learning_rate": 7.095343680709535e-07, "loss": 3.7112, "step": 32 }, { "epoch": 0.0, "learning_rate": 7.317073170731707e-07, "loss": 3.7861, "step": 33 }, { "epoch": 0.0, "learning_rate": 7.538802660753881e-07, "loss": 3.7512, "step": 34 }, { "epoch": 0.0, "learning_rate": 7.760532150776054e-07, "loss": 3.6641, "step": 35 }, { "epoch": 0.0, "learning_rate": 7.982261640798227e-07, "loss": 3.5889, "step": 36 }, { "epoch": 0.0, "learning_rate": 8.203991130820401e-07, "loss": 3.6463, "step": 37 }, { "epoch": 0.0, "learning_rate": 8.425720620842573e-07, "loss": 3.5956, "step": 38 }, { "epoch": 0.0, "learning_rate": 8.647450110864745e-07, "loss": 3.5397, "step": 39 }, { "epoch": 0.0, "learning_rate": 8.869179600886919e-07, "loss": 3.3729, "step": 40 }, { "epoch": 0.0, "learning_rate": 9.090909090909091e-07, "loss": 3.4597, "step": 41 }, { "epoch": 0.0, "learning_rate": 9.312638580931265e-07, "loss": 3.4685, "step": 42 }, { "epoch": 0.0, "learning_rate": 9.534368070953437e-07, "loss": 3.3783, "step": 43 }, { "epoch": 0.0, "learning_rate": 9.75609756097561e-07, "loss": 3.347, "step": 44 }, { "epoch": 0.0, "learning_rate": 9.977827050997784e-07, "loss": 3.4088, "step": 45 }, { "epoch": 0.0, "learning_rate": 1.0199556541019955e-06, "loss": 3.35, "step": 46 }, { "epoch": 0.0, "learning_rate": 1.042128603104213e-06, "loss": 3.3289, "step": 47 }, { "epoch": 0.0, "learning_rate": 1.0643015521064302e-06, "loss": 3.2273, "step": 48 }, { "epoch": 0.0, "learning_rate": 1.0864745011086475e-06, "loss": 3.3801, "step": 49 }, { "epoch": 0.0, "learning_rate": 1.1086474501108648e-06, "loss": 3.3223, "step": 50 }, { "epoch": 0.0, "learning_rate": 1.130820399113082e-06, "loss": 3.2038, "step": 51 }, { "epoch": 0.0, "learning_rate": 1.1529933481152994e-06, "loss": 3.1646, "step": 52 }, { "epoch": 0.0, "learning_rate": 1.1751662971175167e-06, "loss": 3.2898, "step": 53 }, { "epoch": 0.0, "learning_rate": 1.197339246119734e-06, "loss": 3.3017, "step": 54 }, { "epoch": 0.0, "learning_rate": 1.2195121951219514e-06, "loss": 3.2258, "step": 55 }, { "epoch": 0.0, "learning_rate": 1.2416851441241687e-06, "loss": 3.1261, "step": 56 }, { "epoch": 0.0, "learning_rate": 1.2638580931263858e-06, "loss": 3.1941, "step": 57 }, { "epoch": 0.0, "learning_rate": 1.2860310421286031e-06, "loss": 3.1958, "step": 58 }, { "epoch": 0.0, "learning_rate": 1.3082039911308206e-06, "loss": 3.1682, "step": 59 }, { "epoch": 0.0, "learning_rate": 1.3303769401330377e-06, "loss": 3.082, "step": 60 }, { "epoch": 0.0, "learning_rate": 1.352549889135255e-06, "loss": 3.0914, "step": 61 }, { "epoch": 0.0, "learning_rate": 1.3747228381374724e-06, "loss": 3.1407, "step": 62 }, { "epoch": 0.0, "learning_rate": 1.3968957871396897e-06, "loss": 3.0865, "step": 63 }, { "epoch": 0.0, "learning_rate": 1.419068736141907e-06, "loss": 3.0432, "step": 64 }, { "epoch": 0.0, "learning_rate": 1.4412416851441243e-06, "loss": 3.1017, "step": 65 }, { "epoch": 0.0, "learning_rate": 1.4634146341463414e-06, "loss": 3.0568, "step": 66 }, { "epoch": 0.0, "learning_rate": 1.485587583148559e-06, "loss": 3.0132, "step": 67 }, { "epoch": 0.0, "learning_rate": 1.5077605321507763e-06, "loss": 2.9624, "step": 68 }, { "epoch": 0.0, "learning_rate": 1.5299334811529934e-06, "loss": 3.0492, "step": 69 }, { "epoch": 0.0, "learning_rate": 1.5521064301552107e-06, "loss": 3.0296, "step": 70 }, { "epoch": 0.0, "learning_rate": 1.5742793791574282e-06, "loss": 3.0014, "step": 71 }, { "epoch": 0.0, "learning_rate": 1.5964523281596453e-06, "loss": 2.9632, "step": 72 }, { "epoch": 0.0, "learning_rate": 1.6186252771618627e-06, "loss": 3.053, "step": 73 }, { "epoch": 0.0, "learning_rate": 1.6407982261640802e-06, "loss": 2.9771, "step": 74 }, { "epoch": 0.0, "learning_rate": 1.6629711751662973e-06, "loss": 2.9162, "step": 75 }, { "epoch": 0.0, "learning_rate": 1.6851441241685146e-06, "loss": 2.8951, "step": 76 }, { "epoch": 0.0, "learning_rate": 1.707317073170732e-06, "loss": 2.9821, "step": 77 }, { "epoch": 0.0, "learning_rate": 1.729490022172949e-06, "loss": 2.9855, "step": 78 }, { "epoch": 0.0, "learning_rate": 1.7516629711751666e-06, "loss": 2.94, "step": 79 }, { "epoch": 0.0, "learning_rate": 1.7738359201773839e-06, "loss": 2.8282, "step": 80 }, { "epoch": 0.0, "learning_rate": 1.796008869179601e-06, "loss": 2.936, "step": 81 }, { "epoch": 0.0, "learning_rate": 1.8181818181818183e-06, "loss": 2.85, "step": 82 }, { "epoch": 0.0, "learning_rate": 1.8403547671840354e-06, "loss": 2.8375, "step": 83 }, { "epoch": 0.0, "learning_rate": 1.862527716186253e-06, "loss": 2.8279, "step": 84 }, { "epoch": 0.0, "learning_rate": 1.8847006651884702e-06, "loss": 2.8696, "step": 85 }, { "epoch": 0.0, "learning_rate": 1.9068736141906873e-06, "loss": 2.8714, "step": 86 }, { "epoch": 0.0, "learning_rate": 1.9290465631929047e-06, "loss": 2.7816, "step": 87 }, { "epoch": 0.0, "learning_rate": 1.951219512195122e-06, "loss": 2.7118, "step": 88 }, { "epoch": 0.0, "learning_rate": 1.9733924611973393e-06, "loss": 2.8171, "step": 89 }, { "epoch": 0.0, "learning_rate": 1.995565410199557e-06, "loss": 2.8288, "step": 90 }, { "epoch": 0.0, "learning_rate": 2.017738359201774e-06, "loss": 2.7812, "step": 91 }, { "epoch": 0.0, "learning_rate": 2.039911308203991e-06, "loss": 2.7061, "step": 92 }, { "epoch": 0.0, "learning_rate": 2.0620842572062086e-06, "loss": 2.7904, "step": 93 }, { "epoch": 0.0, "learning_rate": 2.084257206208426e-06, "loss": 2.814, "step": 94 }, { "epoch": 0.0, "learning_rate": 2.106430155210643e-06, "loss": 2.778, "step": 95 }, { "epoch": 0.0, "learning_rate": 2.1286031042128603e-06, "loss": 2.6624, "step": 96 }, { "epoch": 0.0, "learning_rate": 2.150776053215078e-06, "loss": 2.7932, "step": 97 }, { "epoch": 0.0, "learning_rate": 2.172949002217295e-06, "loss": 2.7222, "step": 98 }, { "epoch": 0.0, "learning_rate": 2.1951219512195125e-06, "loss": 2.6812, "step": 99 }, { "epoch": 0.0, "learning_rate": 2.2172949002217296e-06, "loss": 2.6107, "step": 100 }, { "epoch": 0.0, "learning_rate": 2.2394678492239467e-06, "loss": 2.7249, "step": 101 }, { "epoch": 0.0, "learning_rate": 2.261640798226164e-06, "loss": 2.6828, "step": 102 }, { "epoch": 0.0, "learning_rate": 2.2838137472283817e-06, "loss": 2.6304, "step": 103 }, { "epoch": 0.0, "learning_rate": 2.305986696230599e-06, "loss": 2.6085, "step": 104 }, { "epoch": 0.0, "learning_rate": 2.3281596452328164e-06, "loss": 2.6667, "step": 105 }, { "epoch": 0.0, "learning_rate": 2.3503325942350335e-06, "loss": 2.6393, "step": 106 }, { "epoch": 0.0, "learning_rate": 2.3725055432372506e-06, "loss": 2.5721, "step": 107 }, { "epoch": 0.0, "learning_rate": 2.394678492239468e-06, "loss": 2.5187, "step": 108 }, { "epoch": 0.0, "learning_rate": 2.4168514412416856e-06, "loss": 2.6876, "step": 109 }, { "epoch": 0.0, "learning_rate": 2.4390243902439027e-06, "loss": 2.545, "step": 110 }, { "epoch": 0.0, "learning_rate": 2.46119733924612e-06, "loss": 2.5268, "step": 111 }, { "epoch": 0.0, "learning_rate": 2.4833702882483374e-06, "loss": 2.5219, "step": 112 }, { "epoch": 0.0, "learning_rate": 2.5055432372505545e-06, "loss": 2.6036, "step": 113 }, { "epoch": 0.0, "learning_rate": 2.5277161862527716e-06, "loss": 2.4997, "step": 114 }, { "epoch": 0.0, "learning_rate": 2.549889135254989e-06, "loss": 2.4829, "step": 115 }, { "epoch": 0.0, "learning_rate": 2.5720620842572062e-06, "loss": 2.4623, "step": 116 }, { "epoch": 0.0, "learning_rate": 2.5942350332594233e-06, "loss": 2.5281, "step": 117 }, { "epoch": 0.0, "learning_rate": 2.6164079822616413e-06, "loss": 2.4791, "step": 118 }, { "epoch": 0.0, "learning_rate": 2.6385809312638584e-06, "loss": 2.4106, "step": 119 }, { "epoch": 0.0, "learning_rate": 2.6607538802660755e-06, "loss": 2.3917, "step": 120 }, { "epoch": 0.0, "learning_rate": 2.682926829268293e-06, "loss": 2.481, "step": 121 }, { "epoch": 0.0, "learning_rate": 2.70509977827051e-06, "loss": 2.4423, "step": 122 }, { "epoch": 0.0, "learning_rate": 2.7272727272727272e-06, "loss": 2.411, "step": 123 }, { "epoch": 0.0, "learning_rate": 2.7494456762749448e-06, "loss": 2.3414, "step": 124 }, { "epoch": 0.0, "learning_rate": 2.7716186252771623e-06, "loss": 2.3875, "step": 125 }, { "epoch": 0.0, "learning_rate": 2.7937915742793794e-06, "loss": 2.3591, "step": 126 }, { "epoch": 0.0, "learning_rate": 2.815964523281597e-06, "loss": 2.3479, "step": 127 }, { "epoch": 0.0, "learning_rate": 2.838137472283814e-06, "loss": 2.3183, "step": 128 }, { "epoch": 0.0, "learning_rate": 2.860310421286031e-06, "loss": 2.3633, "step": 129 }, { "epoch": 0.0, "learning_rate": 2.8824833702882487e-06, "loss": 2.3137, "step": 130 }, { "epoch": 0.0, "learning_rate": 2.9046563192904658e-06, "loss": 2.2869, "step": 131 }, { "epoch": 0.0, "learning_rate": 2.926829268292683e-06, "loss": 2.2459, "step": 132 }, { "epoch": 0.0, "learning_rate": 2.949002217294901e-06, "loss": 2.3333, "step": 133 }, { "epoch": 0.0, "learning_rate": 2.971175166297118e-06, "loss": 2.2386, "step": 134 }, { "epoch": 0.0, "learning_rate": 2.993348115299335e-06, "loss": 2.2004, "step": 135 }, { "epoch": 0.0, "learning_rate": 3.0155210643015526e-06, "loss": 2.1729, "step": 136 }, { "epoch": 0.0, "learning_rate": 3.0376940133037697e-06, "loss": 2.2764, "step": 137 }, { "epoch": 0.0, "learning_rate": 3.0598669623059868e-06, "loss": 2.2337, "step": 138 }, { "epoch": 0.0, "learning_rate": 3.0820399113082043e-06, "loss": 2.2229, "step": 139 }, { "epoch": 0.0, "learning_rate": 3.1042128603104214e-06, "loss": 2.1709, "step": 140 }, { "epoch": 0.0, "learning_rate": 3.1263858093126385e-06, "loss": 2.2662, "step": 141 }, { "epoch": 0.0, "learning_rate": 3.1485587583148565e-06, "loss": 2.1708, "step": 142 }, { "epoch": 0.0, "learning_rate": 3.1707317073170736e-06, "loss": 2.1625, "step": 143 }, { "epoch": 0.0, "learning_rate": 3.1929046563192907e-06, "loss": 2.1325, "step": 144 }, { "epoch": 0.0, "learning_rate": 3.215077605321508e-06, "loss": 2.1897, "step": 145 }, { "epoch": 0.0, "learning_rate": 3.2372505543237253e-06, "loss": 2.1749, "step": 146 }, { "epoch": 0.0, "learning_rate": 3.2594235033259424e-06, "loss": 2.0634, "step": 147 }, { "epoch": 0.0, "learning_rate": 3.2815964523281604e-06, "loss": 2.0798, "step": 148 }, { "epoch": 0.0, "learning_rate": 3.3037694013303775e-06, "loss": 2.1638, "step": 149 }, { "epoch": 0.0, "learning_rate": 3.3259423503325946e-06, "loss": 2.1078, "step": 150 }, { "epoch": 0.01, "learning_rate": 3.348115299334812e-06, "loss": 2.1106, "step": 151 }, { "epoch": 0.01, "learning_rate": 3.370288248337029e-06, "loss": 2.0791, "step": 152 }, { "epoch": 0.01, "learning_rate": 3.3924611973392463e-06, "loss": 2.1428, "step": 153 }, { "epoch": 0.01, "learning_rate": 3.414634146341464e-06, "loss": 2.0592, "step": 154 }, { "epoch": 0.01, "learning_rate": 3.436807095343681e-06, "loss": 2.0472, "step": 155 }, { "epoch": 0.01, "learning_rate": 3.458980044345898e-06, "loss": 2.0195, "step": 156 }, { "epoch": 0.01, "learning_rate": 3.481152993348116e-06, "loss": 2.0708, "step": 157 }, { "epoch": 0.01, "learning_rate": 3.503325942350333e-06, "loss": 2.0676, "step": 158 }, { "epoch": 0.01, "learning_rate": 3.52549889135255e-06, "loss": 2.0305, "step": 159 }, { "epoch": 0.01, "learning_rate": 3.5476718403547677e-06, "loss": 2.0057, "step": 160 }, { "epoch": 0.01, "learning_rate": 3.569844789356985e-06, "loss": 2.1341, "step": 161 }, { "epoch": 0.01, "learning_rate": 3.592017738359202e-06, "loss": 2.0039, "step": 162 }, { "epoch": 0.01, "learning_rate": 3.614190687361419e-06, "loss": 1.9887, "step": 163 }, { "epoch": 0.01, "learning_rate": 3.6363636363636366e-06, "loss": 2.0071, "step": 164 }, { "epoch": 0.01, "learning_rate": 3.6585365853658537e-06, "loss": 2.0671, "step": 165 }, { "epoch": 0.01, "learning_rate": 3.680709534368071e-06, "loss": 2.0313, "step": 166 }, { "epoch": 0.01, "learning_rate": 3.7028824833702887e-06, "loss": 1.9517, "step": 167 }, { "epoch": 0.01, "learning_rate": 3.725055432372506e-06, "loss": 1.978, "step": 168 }, { "epoch": 0.01, "learning_rate": 3.747228381374723e-06, "loss": 2.0459, "step": 169 }, { "epoch": 0.01, "learning_rate": 3.7694013303769405e-06, "loss": 2.0095, "step": 170 }, { "epoch": 0.01, "learning_rate": 3.7915742793791576e-06, "loss": 2.0244, "step": 171 }, { "epoch": 0.01, "learning_rate": 3.8137472283813747e-06, "loss": 1.9299, "step": 172 }, { "epoch": 0.01, "learning_rate": 3.835920177383592e-06, "loss": 2.0012, "step": 173 }, { "epoch": 0.01, "learning_rate": 3.858093126385809e-06, "loss": 1.9922, "step": 174 }, { "epoch": 0.01, "learning_rate": 3.8802660753880264e-06, "loss": 1.9779, "step": 175 }, { "epoch": 0.01, "learning_rate": 3.902439024390244e-06, "loss": 1.9723, "step": 176 }, { "epoch": 0.01, "learning_rate": 3.9246119733924615e-06, "loss": 1.9945, "step": 177 }, { "epoch": 0.01, "learning_rate": 3.946784922394679e-06, "loss": 1.9707, "step": 178 }, { "epoch": 0.01, "learning_rate": 3.9689578713968965e-06, "loss": 1.9346, "step": 179 }, { "epoch": 0.01, "learning_rate": 3.991130820399114e-06, "loss": 1.8959, "step": 180 }, { "epoch": 0.01, "learning_rate": 4.013303769401331e-06, "loss": 2.0167, "step": 181 }, { "epoch": 0.01, "learning_rate": 4.035476718403548e-06, "loss": 1.9537, "step": 182 }, { "epoch": 0.01, "learning_rate": 4.057649667405765e-06, "loss": 1.9528, "step": 183 }, { "epoch": 0.01, "learning_rate": 4.079822616407982e-06, "loss": 1.9644, "step": 184 }, { "epoch": 0.01, "learning_rate": 4.1019955654102e-06, "loss": 2.0372, "step": 185 }, { "epoch": 0.01, "learning_rate": 4.124168514412417e-06, "loss": 1.9752, "step": 186 }, { "epoch": 0.01, "learning_rate": 4.146341463414634e-06, "loss": 1.8928, "step": 187 }, { "epoch": 0.01, "learning_rate": 4.168514412416852e-06, "loss": 1.923, "step": 188 }, { "epoch": 0.01, "learning_rate": 4.190687361419069e-06, "loss": 1.9501, "step": 189 }, { "epoch": 0.01, "learning_rate": 4.212860310421286e-06, "loss": 1.9584, "step": 190 }, { "epoch": 0.01, "learning_rate": 4.2350332594235035e-06, "loss": 1.9371, "step": 191 }, { "epoch": 0.01, "learning_rate": 4.257206208425721e-06, "loss": 1.8721, "step": 192 }, { "epoch": 0.01, "learning_rate": 4.279379157427938e-06, "loss": 1.9387, "step": 193 }, { "epoch": 0.01, "learning_rate": 4.301552106430156e-06, "loss": 1.9214, "step": 194 }, { "epoch": 0.01, "learning_rate": 4.323725055432373e-06, "loss": 1.9402, "step": 195 }, { "epoch": 0.01, "learning_rate": 4.34589800443459e-06, "loss": 1.9352, "step": 196 }, { "epoch": 0.01, "learning_rate": 4.368070953436808e-06, "loss": 1.943, "step": 197 }, { "epoch": 0.01, "learning_rate": 4.390243902439025e-06, "loss": 1.8933, "step": 198 }, { "epoch": 0.01, "learning_rate": 4.412416851441242e-06, "loss": 1.8605, "step": 199 }, { "epoch": 0.01, "learning_rate": 4.434589800443459e-06, "loss": 1.8524, "step": 200 }, { "epoch": 0.01, "learning_rate": 4.456762749445676e-06, "loss": 1.9615, "step": 201 }, { "epoch": 0.01, "learning_rate": 4.478935698447893e-06, "loss": 1.9478, "step": 202 }, { "epoch": 0.01, "learning_rate": 4.501108647450111e-06, "loss": 1.9058, "step": 203 }, { "epoch": 0.01, "learning_rate": 4.523281596452328e-06, "loss": 1.8686, "step": 204 }, { "epoch": 0.01, "learning_rate": 4.5454545454545455e-06, "loss": 1.9283, "step": 205 }, { "epoch": 0.01, "learning_rate": 4.5676274944567635e-06, "loss": 1.9139, "step": 206 }, { "epoch": 0.01, "learning_rate": 4.5898004434589806e-06, "loss": 1.8858, "step": 207 }, { "epoch": 0.01, "learning_rate": 4.611973392461198e-06, "loss": 1.9015, "step": 208 }, { "epoch": 0.01, "learning_rate": 4.634146341463416e-06, "loss": 1.9741, "step": 209 }, { "epoch": 0.01, "learning_rate": 4.656319290465633e-06, "loss": 1.931, "step": 210 }, { "epoch": 0.01, "learning_rate": 4.67849223946785e-06, "loss": 1.8865, "step": 211 }, { "epoch": 0.01, "learning_rate": 4.700665188470067e-06, "loss": 1.862, "step": 212 }, { "epoch": 0.01, "learning_rate": 4.722838137472284e-06, "loss": 1.9276, "step": 213 }, { "epoch": 0.01, "learning_rate": 4.745011086474501e-06, "loss": 1.8846, "step": 214 }, { "epoch": 0.01, "learning_rate": 4.767184035476719e-06, "loss": 1.8763, "step": 215 }, { "epoch": 0.01, "learning_rate": 4.789356984478936e-06, "loss": 1.8287, "step": 216 }, { "epoch": 0.01, "learning_rate": 4.811529933481153e-06, "loss": 1.9244, "step": 217 }, { "epoch": 0.01, "learning_rate": 4.833702882483371e-06, "loss": 1.8979, "step": 218 }, { "epoch": 0.01, "learning_rate": 4.855875831485588e-06, "loss": 1.8603, "step": 219 }, { "epoch": 0.01, "learning_rate": 4.8780487804878055e-06, "loss": 1.8694, "step": 220 }, { "epoch": 0.01, "learning_rate": 4.900221729490023e-06, "loss": 1.8916, "step": 221 }, { "epoch": 0.01, "learning_rate": 4.92239467849224e-06, "loss": 1.8825, "step": 222 }, { "epoch": 0.01, "learning_rate": 4.944567627494457e-06, "loss": 1.791, "step": 223 }, { "epoch": 0.01, "learning_rate": 4.966740576496675e-06, "loss": 1.8823, "step": 224 }, { "epoch": 0.01, "learning_rate": 4.988913525498892e-06, "loss": 1.916, "step": 225 }, { "epoch": 0.01, "learning_rate": 5.011086474501109e-06, "loss": 1.8894, "step": 226 }, { "epoch": 0.01, "learning_rate": 5.033259423503326e-06, "loss": 1.8962, "step": 227 }, { "epoch": 0.01, "learning_rate": 5.055432372505543e-06, "loss": 1.8231, "step": 228 }, { "epoch": 0.01, "learning_rate": 5.077605321507761e-06, "loss": 1.8637, "step": 229 }, { "epoch": 0.01, "learning_rate": 5.099778270509978e-06, "loss": 1.8361, "step": 230 }, { "epoch": 0.01, "learning_rate": 5.121951219512195e-06, "loss": 1.8458, "step": 231 }, { "epoch": 0.01, "learning_rate": 5.1441241685144124e-06, "loss": 1.8247, "step": 232 }, { "epoch": 0.01, "learning_rate": 5.1662971175166295e-06, "loss": 1.8908, "step": 233 }, { "epoch": 0.01, "learning_rate": 5.188470066518847e-06, "loss": 1.9312, "step": 234 }, { "epoch": 0.01, "learning_rate": 5.2106430155210654e-06, "loss": 1.8309, "step": 235 }, { "epoch": 0.01, "learning_rate": 5.2328159645232826e-06, "loss": 1.8141, "step": 236 }, { "epoch": 0.01, "learning_rate": 5.2549889135255e-06, "loss": 1.8425, "step": 237 }, { "epoch": 0.01, "learning_rate": 5.277161862527717e-06, "loss": 1.8285, "step": 238 }, { "epoch": 0.01, "learning_rate": 5.299334811529934e-06, "loss": 1.8257, "step": 239 }, { "epoch": 0.01, "learning_rate": 5.321507760532151e-06, "loss": 1.8092, "step": 240 }, { "epoch": 0.01, "learning_rate": 5.343680709534369e-06, "loss": 1.8307, "step": 241 }, { "epoch": 0.01, "learning_rate": 5.365853658536586e-06, "loss": 1.8315, "step": 242 }, { "epoch": 0.01, "learning_rate": 5.388026607538803e-06, "loss": 1.8247, "step": 243 }, { "epoch": 0.01, "learning_rate": 5.41019955654102e-06, "loss": 1.8069, "step": 244 }, { "epoch": 0.01, "learning_rate": 5.432372505543237e-06, "loss": 1.8131, "step": 245 }, { "epoch": 0.01, "learning_rate": 5.4545454545454545e-06, "loss": 1.8299, "step": 246 }, { "epoch": 0.01, "learning_rate": 5.476718403547672e-06, "loss": 1.8013, "step": 247 }, { "epoch": 0.01, "learning_rate": 5.4988913525498895e-06, "loss": 1.7932, "step": 248 }, { "epoch": 0.01, "learning_rate": 5.5210643015521075e-06, "loss": 1.8621, "step": 249 }, { "epoch": 0.01, "learning_rate": 5.5432372505543246e-06, "loss": 1.8195, "step": 250 }, { "epoch": 0.01, "learning_rate": 5.565410199556542e-06, "loss": 1.7842, "step": 251 }, { "epoch": 0.01, "learning_rate": 5.587583148558759e-06, "loss": 1.7685, "step": 252 }, { "epoch": 0.01, "learning_rate": 5.609756097560977e-06, "loss": 1.8246, "step": 253 }, { "epoch": 0.01, "learning_rate": 5.631929046563194e-06, "loss": 1.79, "step": 254 }, { "epoch": 0.01, "learning_rate": 5.654101995565411e-06, "loss": 1.7674, "step": 255 }, { "epoch": 0.01, "learning_rate": 5.676274944567628e-06, "loss": 1.7869, "step": 256 }, { "epoch": 0.01, "learning_rate": 5.698447893569845e-06, "loss": 1.8739, "step": 257 }, { "epoch": 0.01, "learning_rate": 5.720620842572062e-06, "loss": 1.8323, "step": 258 }, { "epoch": 0.01, "learning_rate": 5.742793791574279e-06, "loss": 1.7666, "step": 259 }, { "epoch": 0.01, "learning_rate": 5.764966740576497e-06, "loss": 1.7833, "step": 260 }, { "epoch": 0.01, "learning_rate": 5.787139689578714e-06, "loss": 1.8694, "step": 261 }, { "epoch": 0.01, "learning_rate": 5.8093126385809315e-06, "loss": 1.8036, "step": 262 }, { "epoch": 0.01, "learning_rate": 5.831485587583149e-06, "loss": 1.789, "step": 263 }, { "epoch": 0.01, "learning_rate": 5.853658536585366e-06, "loss": 1.7743, "step": 264 }, { "epoch": 0.01, "learning_rate": 5.875831485587583e-06, "loss": 1.8249, "step": 265 }, { "epoch": 0.01, "learning_rate": 5.898004434589802e-06, "loss": 1.8293, "step": 266 }, { "epoch": 0.01, "learning_rate": 5.920177383592019e-06, "loss": 1.7748, "step": 267 }, { "epoch": 0.01, "learning_rate": 5.942350332594236e-06, "loss": 1.7291, "step": 268 }, { "epoch": 0.01, "learning_rate": 5.964523281596453e-06, "loss": 1.7609, "step": 269 }, { "epoch": 0.01, "learning_rate": 5.98669623059867e-06, "loss": 1.8185, "step": 270 }, { "epoch": 0.01, "learning_rate": 6.008869179600887e-06, "loss": 1.8122, "step": 271 }, { "epoch": 0.01, "learning_rate": 6.031042128603105e-06, "loss": 1.7486, "step": 272 }, { "epoch": 0.01, "learning_rate": 6.053215077605322e-06, "loss": 1.8058, "step": 273 }, { "epoch": 0.01, "learning_rate": 6.075388026607539e-06, "loss": 1.7903, "step": 274 }, { "epoch": 0.01, "learning_rate": 6.0975609756097564e-06, "loss": 1.7836, "step": 275 }, { "epoch": 0.01, "learning_rate": 6.1197339246119735e-06, "loss": 1.7591, "step": 276 }, { "epoch": 0.01, "learning_rate": 6.141906873614191e-06, "loss": 1.8202, "step": 277 }, { "epoch": 0.01, "learning_rate": 6.164079822616409e-06, "loss": 1.766, "step": 278 }, { "epoch": 0.01, "learning_rate": 6.186252771618626e-06, "loss": 1.7766, "step": 279 }, { "epoch": 0.01, "learning_rate": 6.208425720620843e-06, "loss": 1.7421, "step": 280 }, { "epoch": 0.01, "learning_rate": 6.23059866962306e-06, "loss": 1.7714, "step": 281 }, { "epoch": 0.01, "learning_rate": 6.252771618625277e-06, "loss": 1.8139, "step": 282 }, { "epoch": 0.01, "learning_rate": 6.274944567627494e-06, "loss": 1.7449, "step": 283 }, { "epoch": 0.01, "learning_rate": 6.297117516629713e-06, "loss": 1.79, "step": 284 }, { "epoch": 0.01, "learning_rate": 6.31929046563193e-06, "loss": 1.8125, "step": 285 }, { "epoch": 0.01, "learning_rate": 6.341463414634147e-06, "loss": 1.7874, "step": 286 }, { "epoch": 0.01, "learning_rate": 6.363636363636364e-06, "loss": 1.7209, "step": 287 }, { "epoch": 0.01, "learning_rate": 6.385809312638581e-06, "loss": 1.7742, "step": 288 }, { "epoch": 0.01, "learning_rate": 6.4079822616407984e-06, "loss": 1.8017, "step": 289 }, { "epoch": 0.01, "learning_rate": 6.430155210643016e-06, "loss": 1.7915, "step": 290 }, { "epoch": 0.01, "learning_rate": 6.4523281596452335e-06, "loss": 1.747, "step": 291 }, { "epoch": 0.01, "learning_rate": 6.474501108647451e-06, "loss": 1.7805, "step": 292 }, { "epoch": 0.01, "learning_rate": 6.496674057649668e-06, "loss": 1.7899, "step": 293 }, { "epoch": 0.01, "learning_rate": 6.518847006651885e-06, "loss": 1.8054, "step": 294 }, { "epoch": 0.01, "learning_rate": 6.541019955654102e-06, "loss": 1.757, "step": 295 }, { "epoch": 0.01, "learning_rate": 6.563192904656321e-06, "loss": 1.78, "step": 296 }, { "epoch": 0.01, "learning_rate": 6.585365853658538e-06, "loss": 1.744, "step": 297 }, { "epoch": 0.01, "learning_rate": 6.607538802660755e-06, "loss": 1.7765, "step": 298 }, { "epoch": 0.01, "learning_rate": 6.629711751662972e-06, "loss": 1.7671, "step": 299 }, { "epoch": 0.01, "learning_rate": 6.651884700665189e-06, "loss": 1.7312, "step": 300 }, { "epoch": 0.01, "learning_rate": 6.674057649667406e-06, "loss": 1.8076, "step": 301 }, { "epoch": 0.01, "learning_rate": 6.696230598669624e-06, "loss": 1.7574, "step": 302 }, { "epoch": 0.01, "learning_rate": 6.718403547671841e-06, "loss": 1.7043, "step": 303 }, { "epoch": 0.01, "learning_rate": 6.740576496674058e-06, "loss": 1.7048, "step": 304 }, { "epoch": 0.01, "learning_rate": 6.7627494456762755e-06, "loss": 1.7641, "step": 305 }, { "epoch": 0.01, "learning_rate": 6.784922394678493e-06, "loss": 1.6943, "step": 306 }, { "epoch": 0.01, "learning_rate": 6.80709534368071e-06, "loss": 1.688, "step": 307 }, { "epoch": 0.01, "learning_rate": 6.829268292682928e-06, "loss": 1.7406, "step": 308 }, { "epoch": 0.01, "learning_rate": 6.851441241685145e-06, "loss": 1.7811, "step": 309 }, { "epoch": 0.01, "learning_rate": 6.873614190687362e-06, "loss": 1.7439, "step": 310 }, { "epoch": 0.01, "learning_rate": 6.895787139689579e-06, "loss": 1.7193, "step": 311 }, { "epoch": 0.01, "learning_rate": 6.917960088691796e-06, "loss": 1.6927, "step": 312 }, { "epoch": 0.01, "learning_rate": 6.940133037694013e-06, "loss": 1.7993, "step": 313 }, { "epoch": 0.01, "learning_rate": 6.962305986696232e-06, "loss": 1.7057, "step": 314 }, { "epoch": 0.01, "learning_rate": 6.984478935698449e-06, "loss": 1.7104, "step": 315 }, { "epoch": 0.01, "learning_rate": 7.006651884700666e-06, "loss": 1.6829, "step": 316 }, { "epoch": 0.01, "learning_rate": 7.028824833702883e-06, "loss": 1.8051, "step": 317 }, { "epoch": 0.01, "learning_rate": 7.0509977827051e-06, "loss": 1.7402, "step": 318 }, { "epoch": 0.01, "learning_rate": 7.0731707317073175e-06, "loss": 1.7034, "step": 319 }, { "epoch": 0.01, "learning_rate": 7.0953436807095355e-06, "loss": 1.7201, "step": 320 }, { "epoch": 0.01, "learning_rate": 7.117516629711753e-06, "loss": 1.7434, "step": 321 }, { "epoch": 0.01, "learning_rate": 7.13968957871397e-06, "loss": 1.6782, "step": 322 }, { "epoch": 0.01, "learning_rate": 7.161862527716187e-06, "loss": 1.7131, "step": 323 }, { "epoch": 0.01, "learning_rate": 7.184035476718404e-06, "loss": 1.7023, "step": 324 }, { "epoch": 0.01, "learning_rate": 7.206208425720621e-06, "loss": 1.7669, "step": 325 }, { "epoch": 0.01, "learning_rate": 7.228381374722838e-06, "loss": 1.7076, "step": 326 }, { "epoch": 0.01, "learning_rate": 7.250554323725056e-06, "loss": 1.6749, "step": 327 }, { "epoch": 0.01, "learning_rate": 7.272727272727273e-06, "loss": 1.6841, "step": 328 }, { "epoch": 0.01, "learning_rate": 7.29490022172949e-06, "loss": 1.7499, "step": 329 }, { "epoch": 0.01, "learning_rate": 7.317073170731707e-06, "loss": 1.7285, "step": 330 }, { "epoch": 0.01, "learning_rate": 7.3392461197339245e-06, "loss": 1.688, "step": 331 }, { "epoch": 0.01, "learning_rate": 7.361419068736142e-06, "loss": 1.6768, "step": 332 }, { "epoch": 0.01, "learning_rate": 7.38359201773836e-06, "loss": 1.6887, "step": 333 }, { "epoch": 0.01, "learning_rate": 7.4057649667405775e-06, "loss": 1.6898, "step": 334 }, { "epoch": 0.01, "learning_rate": 7.427937915742795e-06, "loss": 1.6629, "step": 335 }, { "epoch": 0.01, "learning_rate": 7.450110864745012e-06, "loss": 1.6602, "step": 336 }, { "epoch": 0.01, "learning_rate": 7.472283813747229e-06, "loss": 1.7002, "step": 337 }, { "epoch": 0.01, "learning_rate": 7.494456762749446e-06, "loss": 1.7089, "step": 338 }, { "epoch": 0.01, "learning_rate": 7.516629711751664e-06, "loss": 1.6801, "step": 339 }, { "epoch": 0.01, "learning_rate": 7.538802660753881e-06, "loss": 1.6216, "step": 340 }, { "epoch": 0.01, "learning_rate": 7.560975609756098e-06, "loss": 1.7316, "step": 341 }, { "epoch": 0.01, "learning_rate": 7.583148558758315e-06, "loss": 1.6414, "step": 342 }, { "epoch": 0.01, "learning_rate": 7.605321507760532e-06, "loss": 1.6189, "step": 343 }, { "epoch": 0.01, "learning_rate": 7.627494456762749e-06, "loss": 1.614, "step": 344 }, { "epoch": 0.01, "learning_rate": 7.649667405764967e-06, "loss": 1.673, "step": 345 }, { "epoch": 0.01, "learning_rate": 7.671840354767184e-06, "loss": 1.6518, "step": 346 }, { "epoch": 0.01, "learning_rate": 7.694013303769402e-06, "loss": 1.6573, "step": 347 }, { "epoch": 0.01, "learning_rate": 7.716186252771619e-06, "loss": 1.6309, "step": 348 }, { "epoch": 0.01, "learning_rate": 7.738359201773836e-06, "loss": 1.6771, "step": 349 }, { "epoch": 0.01, "learning_rate": 7.760532150776053e-06, "loss": 1.7001, "step": 350 }, { "epoch": 0.01, "learning_rate": 7.782705099778272e-06, "loss": 1.6304, "step": 351 }, { "epoch": 0.01, "learning_rate": 7.804878048780489e-06, "loss": 1.5644, "step": 352 }, { "epoch": 0.01, "learning_rate": 7.827050997782706e-06, "loss": 1.641, "step": 353 }, { "epoch": 0.01, "learning_rate": 7.849223946784923e-06, "loss": 1.616, "step": 354 }, { "epoch": 0.01, "learning_rate": 7.87139689578714e-06, "loss": 1.5923, "step": 355 }, { "epoch": 0.01, "learning_rate": 7.893569844789357e-06, "loss": 1.5888, "step": 356 }, { "epoch": 0.01, "learning_rate": 7.915742793791576e-06, "loss": 1.6297, "step": 357 }, { "epoch": 0.01, "learning_rate": 7.937915742793793e-06, "loss": 1.6001, "step": 358 }, { "epoch": 0.01, "learning_rate": 7.96008869179601e-06, "loss": 1.6149, "step": 359 }, { "epoch": 0.01, "learning_rate": 7.982261640798227e-06, "loss": 1.5597, "step": 360 }, { "epoch": 0.01, "learning_rate": 8.004434589800444e-06, "loss": 1.5891, "step": 361 }, { "epoch": 0.01, "learning_rate": 8.026607538802662e-06, "loss": 1.6033, "step": 362 }, { "epoch": 0.01, "learning_rate": 8.048780487804879e-06, "loss": 1.5361, "step": 363 }, { "epoch": 0.01, "learning_rate": 8.070953436807096e-06, "loss": 1.5294, "step": 364 }, { "epoch": 0.01, "learning_rate": 8.093126385809313e-06, "loss": 1.5732, "step": 365 }, { "epoch": 0.01, "learning_rate": 8.11529933481153e-06, "loss": 1.5349, "step": 366 }, { "epoch": 0.01, "learning_rate": 8.137472283813747e-06, "loss": 1.489, "step": 367 }, { "epoch": 0.01, "learning_rate": 8.159645232815964e-06, "loss": 1.4676, "step": 368 }, { "epoch": 0.01, "learning_rate": 8.181818181818183e-06, "loss": 1.5458, "step": 369 }, { "epoch": 0.01, "learning_rate": 8.2039911308204e-06, "loss": 1.538, "step": 370 }, { "epoch": 0.01, "learning_rate": 8.226164079822617e-06, "loss": 1.4359, "step": 371 }, { "epoch": 0.01, "learning_rate": 8.248337028824834e-06, "loss": 1.4407, "step": 372 }, { "epoch": 0.01, "learning_rate": 8.270509977827051e-06, "loss": 1.4821, "step": 373 }, { "epoch": 0.01, "learning_rate": 8.292682926829268e-06, "loss": 1.4271, "step": 374 }, { "epoch": 0.01, "learning_rate": 8.314855875831487e-06, "loss": 1.4411, "step": 375 }, { "epoch": 0.01, "learning_rate": 8.337028824833704e-06, "loss": 1.409, "step": 376 }, { "epoch": 0.01, "learning_rate": 8.359201773835921e-06, "loss": 1.4713, "step": 377 }, { "epoch": 0.01, "learning_rate": 8.381374722838139e-06, "loss": 1.4466, "step": 378 }, { "epoch": 0.01, "learning_rate": 8.403547671840356e-06, "loss": 1.4216, "step": 379 }, { "epoch": 0.01, "learning_rate": 8.425720620842573e-06, "loss": 1.35, "step": 380 }, { "epoch": 0.01, "learning_rate": 8.44789356984479e-06, "loss": 1.4128, "step": 381 }, { "epoch": 0.01, "learning_rate": 8.470066518847007e-06, "loss": 1.4169, "step": 382 }, { "epoch": 0.01, "learning_rate": 8.492239467849224e-06, "loss": 1.3924, "step": 383 }, { "epoch": 0.01, "learning_rate": 8.514412416851441e-06, "loss": 1.3481, "step": 384 }, { "epoch": 0.01, "learning_rate": 8.536585365853658e-06, "loss": 1.4094, "step": 385 }, { "epoch": 0.01, "learning_rate": 8.558758314855875e-06, "loss": 1.3407, "step": 386 }, { "epoch": 0.01, "learning_rate": 8.580931263858093e-06, "loss": 1.3904, "step": 387 }, { "epoch": 0.01, "learning_rate": 8.603104212860311e-06, "loss": 1.3218, "step": 388 }, { "epoch": 0.01, "learning_rate": 8.625277161862528e-06, "loss": 1.4045, "step": 389 }, { "epoch": 0.01, "learning_rate": 8.647450110864746e-06, "loss": 1.3692, "step": 390 }, { "epoch": 0.01, "learning_rate": 8.669623059866963e-06, "loss": 1.3403, "step": 391 }, { "epoch": 0.01, "learning_rate": 8.69179600886918e-06, "loss": 1.3429, "step": 392 }, { "epoch": 0.01, "learning_rate": 8.713968957871397e-06, "loss": 1.3595, "step": 393 }, { "epoch": 0.01, "learning_rate": 8.736141906873616e-06, "loss": 1.3653, "step": 394 }, { "epoch": 0.01, "learning_rate": 8.758314855875833e-06, "loss": 1.3923, "step": 395 }, { "epoch": 0.01, "learning_rate": 8.78048780487805e-06, "loss": 1.2663, "step": 396 }, { "epoch": 0.01, "learning_rate": 8.802660753880267e-06, "loss": 1.3694, "step": 397 }, { "epoch": 0.01, "learning_rate": 8.824833702882484e-06, "loss": 1.3371, "step": 398 }, { "epoch": 0.01, "learning_rate": 8.847006651884701e-06, "loss": 1.3473, "step": 399 }, { "epoch": 0.01, "learning_rate": 8.869179600886918e-06, "loss": 1.2608, "step": 400 }, { "epoch": 0.01, "learning_rate": 8.891352549889135e-06, "loss": 1.3146, "step": 401 }, { "epoch": 0.01, "learning_rate": 8.913525498891353e-06, "loss": 1.3129, "step": 402 }, { "epoch": 0.01, "learning_rate": 8.93569844789357e-06, "loss": 1.3076, "step": 403 }, { "epoch": 0.01, "learning_rate": 8.957871396895787e-06, "loss": 1.2514, "step": 404 }, { "epoch": 0.01, "learning_rate": 8.980044345898006e-06, "loss": 1.3649, "step": 405 }, { "epoch": 0.01, "learning_rate": 9.002217294900223e-06, "loss": 1.3163, "step": 406 }, { "epoch": 0.01, "learning_rate": 9.02439024390244e-06, "loss": 1.3609, "step": 407 }, { "epoch": 0.01, "learning_rate": 9.046563192904657e-06, "loss": 1.271, "step": 408 }, { "epoch": 0.01, "learning_rate": 9.068736141906874e-06, "loss": 1.3124, "step": 409 }, { "epoch": 0.01, "learning_rate": 9.090909090909091e-06, "loss": 1.2641, "step": 410 }, { "epoch": 0.01, "learning_rate": 9.113082039911308e-06, "loss": 1.3294, "step": 411 }, { "epoch": 0.01, "learning_rate": 9.135254988913527e-06, "loss": 1.2809, "step": 412 }, { "epoch": 0.01, "learning_rate": 9.157427937915744e-06, "loss": 1.3148, "step": 413 }, { "epoch": 0.01, "learning_rate": 9.179600886917961e-06, "loss": 1.2986, "step": 414 }, { "epoch": 0.01, "learning_rate": 9.201773835920178e-06, "loss": 1.2697, "step": 415 }, { "epoch": 0.01, "learning_rate": 9.223946784922395e-06, "loss": 1.2763, "step": 416 }, { "epoch": 0.01, "learning_rate": 9.246119733924612e-06, "loss": 1.3022, "step": 417 }, { "epoch": 0.01, "learning_rate": 9.268292682926831e-06, "loss": 1.2414, "step": 418 }, { "epoch": 0.01, "learning_rate": 9.290465631929048e-06, "loss": 1.2704, "step": 419 }, { "epoch": 0.01, "learning_rate": 9.312638580931265e-06, "loss": 1.2551, "step": 420 }, { "epoch": 0.01, "learning_rate": 9.334811529933483e-06, "loss": 1.316, "step": 421 }, { "epoch": 0.01, "learning_rate": 9.3569844789357e-06, "loss": 1.2669, "step": 422 }, { "epoch": 0.01, "learning_rate": 9.379157427937917e-06, "loss": 1.294, "step": 423 }, { "epoch": 0.01, "learning_rate": 9.401330376940134e-06, "loss": 1.2464, "step": 424 }, { "epoch": 0.01, "learning_rate": 9.423503325942351e-06, "loss": 1.2734, "step": 425 }, { "epoch": 0.01, "learning_rate": 9.445676274944568e-06, "loss": 1.2825, "step": 426 }, { "epoch": 0.01, "learning_rate": 9.467849223946785e-06, "loss": 1.2754, "step": 427 }, { "epoch": 0.01, "learning_rate": 9.490022172949002e-06, "loss": 1.2231, "step": 428 }, { "epoch": 0.01, "learning_rate": 9.51219512195122e-06, "loss": 1.2815, "step": 429 }, { "epoch": 0.01, "learning_rate": 9.534368070953438e-06, "loss": 1.2714, "step": 430 }, { "epoch": 0.01, "learning_rate": 9.556541019955655e-06, "loss": 1.2607, "step": 431 }, { "epoch": 0.01, "learning_rate": 9.578713968957872e-06, "loss": 1.2245, "step": 432 }, { "epoch": 0.01, "learning_rate": 9.60088691796009e-06, "loss": 1.2634, "step": 433 }, { "epoch": 0.01, "learning_rate": 9.623059866962307e-06, "loss": 1.2514, "step": 434 }, { "epoch": 0.01, "learning_rate": 9.645232815964524e-06, "loss": 1.2664, "step": 435 }, { "epoch": 0.01, "learning_rate": 9.667405764966743e-06, "loss": 1.2496, "step": 436 }, { "epoch": 0.01, "learning_rate": 9.68957871396896e-06, "loss": 1.2497, "step": 437 }, { "epoch": 0.01, "learning_rate": 9.711751662971177e-06, "loss": 1.2594, "step": 438 }, { "epoch": 0.01, "learning_rate": 9.733924611973394e-06, "loss": 1.1997, "step": 439 }, { "epoch": 0.01, "learning_rate": 9.756097560975611e-06, "loss": 1.2047, "step": 440 }, { "epoch": 0.01, "learning_rate": 9.778270509977828e-06, "loss": 1.2574, "step": 441 }, { "epoch": 0.01, "learning_rate": 9.800443458980045e-06, "loss": 1.302, "step": 442 }, { "epoch": 0.01, "learning_rate": 9.822616407982262e-06, "loss": 1.2386, "step": 443 }, { "epoch": 0.01, "learning_rate": 9.84478935698448e-06, "loss": 1.1973, "step": 444 }, { "epoch": 0.01, "learning_rate": 9.866962305986696e-06, "loss": 1.2467, "step": 445 }, { "epoch": 0.01, "learning_rate": 9.889135254988914e-06, "loss": 1.2976, "step": 446 }, { "epoch": 0.01, "learning_rate": 9.91130820399113e-06, "loss": 1.2681, "step": 447 }, { "epoch": 0.01, "learning_rate": 9.93348115299335e-06, "loss": 1.1859, "step": 448 }, { "epoch": 0.01, "learning_rate": 9.955654101995567e-06, "loss": 1.2714, "step": 449 }, { "epoch": 0.01, "learning_rate": 9.977827050997784e-06, "loss": 1.1792, "step": 450 }, { "epoch": 0.02, "learning_rate": 1e-05, "loss": 1.247, "step": 451 }, { "epoch": 0.02, "learning_rate": 1.0022172949002218e-05, "loss": 1.1643, "step": 452 }, { "epoch": 0.02, "learning_rate": 1.0044345898004435e-05, "loss": 1.252, "step": 453 }, { "epoch": 0.02, "learning_rate": 1.0066518847006652e-05, "loss": 1.1823, "step": 454 }, { "epoch": 0.02, "learning_rate": 1.008869179600887e-05, "loss": 1.2487, "step": 455 }, { "epoch": 0.02, "learning_rate": 1.0110864745011086e-05, "loss": 1.2146, "step": 456 }, { "epoch": 0.02, "learning_rate": 1.0133037694013303e-05, "loss": 1.2263, "step": 457 }, { "epoch": 0.02, "learning_rate": 1.0155210643015522e-05, "loss": 1.2612, "step": 458 }, { "epoch": 0.02, "learning_rate": 1.017738359201774e-05, "loss": 1.257, "step": 459 }, { "epoch": 0.02, "learning_rate": 1.0199556541019956e-05, "loss": 1.1801, "step": 460 }, { "epoch": 0.02, "learning_rate": 1.0221729490022174e-05, "loss": 1.2172, "step": 461 }, { "epoch": 0.02, "learning_rate": 1.024390243902439e-05, "loss": 1.2377, "step": 462 }, { "epoch": 0.02, "learning_rate": 1.0266075388026608e-05, "loss": 1.2324, "step": 463 }, { "epoch": 0.02, "learning_rate": 1.0288248337028825e-05, "loss": 1.1677, "step": 464 }, { "epoch": 0.02, "learning_rate": 1.0310421286031042e-05, "loss": 1.2483, "step": 465 }, { "epoch": 0.02, "learning_rate": 1.0332594235033259e-05, "loss": 1.2033, "step": 466 }, { "epoch": 0.02, "learning_rate": 1.0354767184035476e-05, "loss": 1.246, "step": 467 }, { "epoch": 0.02, "learning_rate": 1.0376940133037693e-05, "loss": 1.1854, "step": 468 }, { "epoch": 0.02, "learning_rate": 1.039911308203991e-05, "loss": 1.2264, "step": 469 }, { "epoch": 0.02, "learning_rate": 1.0421286031042131e-05, "loss": 1.2846, "step": 470 }, { "epoch": 0.02, "learning_rate": 1.0443458980044348e-05, "loss": 1.2129, "step": 471 }, { "epoch": 0.02, "learning_rate": 1.0465631929046565e-05, "loss": 1.1679, "step": 472 }, { "epoch": 0.02, "learning_rate": 1.0487804878048782e-05, "loss": 1.2556, "step": 473 }, { "epoch": 0.02, "learning_rate": 1.0509977827051e-05, "loss": 1.2305, "step": 474 }, { "epoch": 0.02, "learning_rate": 1.0532150776053216e-05, "loss": 1.254, "step": 475 }, { "epoch": 0.02, "learning_rate": 1.0554323725055434e-05, "loss": 1.1752, "step": 476 }, { "epoch": 0.02, "learning_rate": 1.057649667405765e-05, "loss": 1.2016, "step": 477 }, { "epoch": 0.02, "learning_rate": 1.0598669623059868e-05, "loss": 1.1908, "step": 478 }, { "epoch": 0.02, "learning_rate": 1.0620842572062085e-05, "loss": 1.2339, "step": 479 }, { "epoch": 0.02, "learning_rate": 1.0643015521064302e-05, "loss": 1.1298, "step": 480 }, { "epoch": 0.02, "learning_rate": 1.0665188470066519e-05, "loss": 1.2367, "step": 481 }, { "epoch": 0.02, "learning_rate": 1.0687361419068738e-05, "loss": 1.2002, "step": 482 }, { "epoch": 0.02, "learning_rate": 1.0709534368070955e-05, "loss": 1.2318, "step": 483 }, { "epoch": 0.02, "learning_rate": 1.0731707317073172e-05, "loss": 1.1605, "step": 484 }, { "epoch": 0.02, "learning_rate": 1.075388026607539e-05, "loss": 1.2443, "step": 485 }, { "epoch": 0.02, "learning_rate": 1.0776053215077606e-05, "loss": 1.2039, "step": 486 }, { "epoch": 0.02, "learning_rate": 1.0798226164079823e-05, "loss": 1.2088, "step": 487 }, { "epoch": 0.02, "learning_rate": 1.082039911308204e-05, "loss": 1.1637, "step": 488 }, { "epoch": 0.02, "learning_rate": 1.0842572062084258e-05, "loss": 1.2488, "step": 489 }, { "epoch": 0.02, "learning_rate": 1.0864745011086475e-05, "loss": 1.2379, "step": 490 }, { "epoch": 0.02, "learning_rate": 1.0886917960088692e-05, "loss": 1.2024, "step": 491 }, { "epoch": 0.02, "learning_rate": 1.0909090909090909e-05, "loss": 1.1792, "step": 492 }, { "epoch": 0.02, "learning_rate": 1.0931263858093126e-05, "loss": 1.2416, "step": 493 }, { "epoch": 0.02, "learning_rate": 1.0953436807095345e-05, "loss": 1.2341, "step": 494 }, { "epoch": 0.02, "learning_rate": 1.0975609756097562e-05, "loss": 1.2084, "step": 495 }, { "epoch": 0.02, "learning_rate": 1.0997782705099779e-05, "loss": 1.1777, "step": 496 }, { "epoch": 0.02, "learning_rate": 1.1019955654101996e-05, "loss": 1.221, "step": 497 }, { "epoch": 0.02, "learning_rate": 1.1042128603104215e-05, "loss": 1.2421, "step": 498 }, { "epoch": 0.02, "learning_rate": 1.1064301552106432e-05, "loss": 1.2549, "step": 499 }, { "epoch": 0.02, "learning_rate": 1.1086474501108649e-05, "loss": 1.1827, "step": 500 }, { "epoch": 0.02, "learning_rate": 1.1108647450110866e-05, "loss": 1.2093, "step": 501 }, { "epoch": 0.02, "learning_rate": 1.1130820399113083e-05, "loss": 1.2008, "step": 502 }, { "epoch": 0.02, "learning_rate": 1.11529933481153e-05, "loss": 1.2338, "step": 503 }, { "epoch": 0.02, "learning_rate": 1.1175166297117518e-05, "loss": 1.1421, "step": 504 }, { "epoch": 0.02, "learning_rate": 1.1197339246119735e-05, "loss": 1.2296, "step": 505 }, { "epoch": 0.02, "learning_rate": 1.1219512195121953e-05, "loss": 1.1752, "step": 506 }, { "epoch": 0.02, "learning_rate": 1.124168514412417e-05, "loss": 1.2046, "step": 507 }, { "epoch": 0.02, "learning_rate": 1.1263858093126388e-05, "loss": 1.1804, "step": 508 }, { "epoch": 0.02, "learning_rate": 1.1286031042128605e-05, "loss": 1.2357, "step": 509 }, { "epoch": 0.02, "learning_rate": 1.1308203991130822e-05, "loss": 1.18, "step": 510 }, { "epoch": 0.02, "learning_rate": 1.1330376940133039e-05, "loss": 1.2501, "step": 511 }, { "epoch": 0.02, "learning_rate": 1.1352549889135256e-05, "loss": 1.1496, "step": 512 }, { "epoch": 0.02, "learning_rate": 1.1374722838137473e-05, "loss": 1.2453, "step": 513 }, { "epoch": 0.02, "learning_rate": 1.139689578713969e-05, "loss": 1.2015, "step": 514 }, { "epoch": 0.02, "learning_rate": 1.1419068736141907e-05, "loss": 1.2039, "step": 515 }, { "epoch": 0.02, "learning_rate": 1.1441241685144125e-05, "loss": 1.1816, "step": 516 }, { "epoch": 0.02, "learning_rate": 1.1463414634146342e-05, "loss": 1.1822, "step": 517 }, { "epoch": 0.02, "learning_rate": 1.1485587583148559e-05, "loss": 1.1972, "step": 518 }, { "epoch": 0.02, "learning_rate": 1.1507760532150778e-05, "loss": 1.2312, "step": 519 }, { "epoch": 0.02, "learning_rate": 1.1529933481152995e-05, "loss": 1.1242, "step": 520 }, { "epoch": 0.02, "learning_rate": 1.1552106430155212e-05, "loss": 1.2307, "step": 521 }, { "epoch": 0.02, "learning_rate": 1.1574279379157429e-05, "loss": 1.2329, "step": 522 }, { "epoch": 0.02, "learning_rate": 1.1596452328159646e-05, "loss": 1.2124, "step": 523 }, { "epoch": 0.02, "learning_rate": 1.1618625277161863e-05, "loss": 1.1729, "step": 524 }, { "epoch": 0.02, "learning_rate": 1.164079822616408e-05, "loss": 1.202, "step": 525 }, { "epoch": 0.02, "learning_rate": 1.1662971175166297e-05, "loss": 1.1936, "step": 526 }, { "epoch": 0.02, "learning_rate": 1.1685144124168514e-05, "loss": 1.2083, "step": 527 }, { "epoch": 0.02, "learning_rate": 1.1707317073170731e-05, "loss": 1.1586, "step": 528 }, { "epoch": 0.02, "learning_rate": 1.1729490022172949e-05, "loss": 1.2016, "step": 529 }, { "epoch": 0.02, "learning_rate": 1.1751662971175166e-05, "loss": 1.1842, "step": 530 }, { "epoch": 0.02, "learning_rate": 1.1773835920177386e-05, "loss": 1.2107, "step": 531 }, { "epoch": 0.02, "learning_rate": 1.1796008869179603e-05, "loss": 1.1758, "step": 532 }, { "epoch": 0.02, "learning_rate": 1.181818181818182e-05, "loss": 1.206, "step": 533 }, { "epoch": 0.02, "learning_rate": 1.1840354767184037e-05, "loss": 1.1681, "step": 534 }, { "epoch": 0.02, "learning_rate": 1.1862527716186255e-05, "loss": 1.1934, "step": 535 }, { "epoch": 0.02, "learning_rate": 1.1884700665188472e-05, "loss": 1.1517, "step": 536 }, { "epoch": 0.02, "learning_rate": 1.1906873614190689e-05, "loss": 1.173, "step": 537 }, { "epoch": 0.02, "learning_rate": 1.1929046563192906e-05, "loss": 1.1942, "step": 538 }, { "epoch": 0.02, "learning_rate": 1.1951219512195123e-05, "loss": 1.1528, "step": 539 }, { "epoch": 0.02, "learning_rate": 1.197339246119734e-05, "loss": 1.1783, "step": 540 }, { "epoch": 0.02, "learning_rate": 1.1995565410199557e-05, "loss": 1.1462, "step": 541 }, { "epoch": 0.02, "learning_rate": 1.2017738359201774e-05, "loss": 1.1841, "step": 542 }, { "epoch": 0.02, "learning_rate": 1.2039911308203993e-05, "loss": 1.1681, "step": 543 }, { "epoch": 0.02, "learning_rate": 1.206208425720621e-05, "loss": 1.1787, "step": 544 }, { "epoch": 0.02, "learning_rate": 1.2084257206208427e-05, "loss": 1.1955, "step": 545 }, { "epoch": 0.02, "learning_rate": 1.2106430155210644e-05, "loss": 1.1901, "step": 546 }, { "epoch": 0.02, "learning_rate": 1.2128603104212862e-05, "loss": 1.2014, "step": 547 }, { "epoch": 0.02, "learning_rate": 1.2150776053215079e-05, "loss": 1.1799, "step": 548 }, { "epoch": 0.02, "learning_rate": 1.2172949002217296e-05, "loss": 1.1945, "step": 549 }, { "epoch": 0.02, "learning_rate": 1.2195121951219513e-05, "loss": 1.1823, "step": 550 }, { "epoch": 0.02, "learning_rate": 1.221729490022173e-05, "loss": 1.1374, "step": 551 }, { "epoch": 0.02, "learning_rate": 1.2239467849223947e-05, "loss": 1.1608, "step": 552 }, { "epoch": 0.02, "learning_rate": 1.2261640798226164e-05, "loss": 1.1873, "step": 553 }, { "epoch": 0.02, "learning_rate": 1.2283813747228381e-05, "loss": 1.1765, "step": 554 }, { "epoch": 0.02, "learning_rate": 1.23059866962306e-05, "loss": 1.2298, "step": 555 }, { "epoch": 0.02, "learning_rate": 1.2328159645232817e-05, "loss": 1.1413, "step": 556 }, { "epoch": 0.02, "learning_rate": 1.2350332594235034e-05, "loss": 1.1774, "step": 557 }, { "epoch": 0.02, "learning_rate": 1.2372505543237251e-05, "loss": 1.1656, "step": 558 }, { "epoch": 0.02, "learning_rate": 1.2394678492239469e-05, "loss": 1.1722, "step": 559 }, { "epoch": 0.02, "learning_rate": 1.2416851441241686e-05, "loss": 1.147, "step": 560 }, { "epoch": 0.02, "learning_rate": 1.2439024390243903e-05, "loss": 1.1815, "step": 561 }, { "epoch": 0.02, "learning_rate": 1.246119733924612e-05, "loss": 1.181, "step": 562 }, { "epoch": 0.02, "learning_rate": 1.2483370288248337e-05, "loss": 1.1291, "step": 563 }, { "epoch": 0.02, "learning_rate": 1.2505543237250554e-05, "loss": 1.1248, "step": 564 }, { "epoch": 0.02, "learning_rate": 1.2527716186252771e-05, "loss": 1.1845, "step": 565 }, { "epoch": 0.02, "learning_rate": 1.2549889135254988e-05, "loss": 1.1557, "step": 566 }, { "epoch": 0.02, "learning_rate": 1.2572062084257209e-05, "loss": 1.1823, "step": 567 }, { "epoch": 0.02, "learning_rate": 1.2594235033259426e-05, "loss": 1.1562, "step": 568 }, { "epoch": 0.02, "learning_rate": 1.2616407982261643e-05, "loss": 1.198, "step": 569 }, { "epoch": 0.02, "learning_rate": 1.263858093126386e-05, "loss": 1.1738, "step": 570 }, { "epoch": 0.02, "learning_rate": 1.2660753880266077e-05, "loss": 1.1438, "step": 571 }, { "epoch": 0.02, "learning_rate": 1.2682926829268294e-05, "loss": 1.084, "step": 572 }, { "epoch": 0.02, "learning_rate": 1.2705099778270511e-05, "loss": 1.1644, "step": 573 }, { "epoch": 0.02, "learning_rate": 1.2727272727272728e-05, "loss": 1.1757, "step": 574 }, { "epoch": 0.02, "learning_rate": 1.2749445676274946e-05, "loss": 1.1887, "step": 575 }, { "epoch": 0.02, "learning_rate": 1.2771618625277163e-05, "loss": 1.1459, "step": 576 }, { "epoch": 0.02, "learning_rate": 1.279379157427938e-05, "loss": 1.1593, "step": 577 }, { "epoch": 0.02, "learning_rate": 1.2815964523281597e-05, "loss": 1.1741, "step": 578 }, { "epoch": 0.02, "learning_rate": 1.2838137472283816e-05, "loss": 1.1864, "step": 579 }, { "epoch": 0.02, "learning_rate": 1.2860310421286033e-05, "loss": 1.1371, "step": 580 }, { "epoch": 0.02, "learning_rate": 1.288248337028825e-05, "loss": 1.1615, "step": 581 }, { "epoch": 0.02, "learning_rate": 1.2904656319290467e-05, "loss": 1.1441, "step": 582 }, { "epoch": 0.02, "learning_rate": 1.2926829268292684e-05, "loss": 1.1393, "step": 583 }, { "epoch": 0.02, "learning_rate": 1.2949002217294901e-05, "loss": 1.1479, "step": 584 }, { "epoch": 0.02, "learning_rate": 1.2971175166297118e-05, "loss": 1.1804, "step": 585 }, { "epoch": 0.02, "learning_rate": 1.2993348115299335e-05, "loss": 1.1467, "step": 586 }, { "epoch": 0.02, "learning_rate": 1.3015521064301553e-05, "loss": 1.1713, "step": 587 }, { "epoch": 0.02, "learning_rate": 1.303769401330377e-05, "loss": 1.1725, "step": 588 }, { "epoch": 0.02, "learning_rate": 1.3059866962305987e-05, "loss": 1.2157, "step": 589 }, { "epoch": 0.02, "learning_rate": 1.3082039911308204e-05, "loss": 1.1818, "step": 590 }, { "epoch": 0.02, "learning_rate": 1.3104212860310421e-05, "loss": 1.1081, "step": 591 }, { "epoch": 0.02, "learning_rate": 1.3126385809312641e-05, "loss": 1.132, "step": 592 }, { "epoch": 0.02, "learning_rate": 1.3148558758314859e-05, "loss": 1.1686, "step": 593 }, { "epoch": 0.02, "learning_rate": 1.3170731707317076e-05, "loss": 1.1496, "step": 594 }, { "epoch": 0.02, "learning_rate": 1.3192904656319293e-05, "loss": 1.152, "step": 595 }, { "epoch": 0.02, "learning_rate": 1.321507760532151e-05, "loss": 1.1317, "step": 596 }, { "epoch": 0.02, "learning_rate": 1.3237250554323727e-05, "loss": 1.1826, "step": 597 }, { "epoch": 0.02, "learning_rate": 1.3259423503325944e-05, "loss": 1.1753, "step": 598 }, { "epoch": 0.02, "learning_rate": 1.3281596452328161e-05, "loss": 1.1751, "step": 599 }, { "epoch": 0.02, "learning_rate": 1.3303769401330378e-05, "loss": 1.1209, "step": 600 }, { "epoch": 0.02, "learning_rate": 1.3325942350332595e-05, "loss": 1.2016, "step": 601 }, { "epoch": 0.02, "learning_rate": 1.3348115299334812e-05, "loss": 1.1606, "step": 602 }, { "epoch": 0.02, "learning_rate": 1.337028824833703e-05, "loss": 1.1717, "step": 603 }, { "epoch": 0.02, "learning_rate": 1.3392461197339248e-05, "loss": 1.1419, "step": 604 }, { "epoch": 0.02, "learning_rate": 1.3414634146341466e-05, "loss": 1.1936, "step": 605 }, { "epoch": 0.02, "learning_rate": 1.3436807095343683e-05, "loss": 1.1841, "step": 606 }, { "epoch": 0.02, "learning_rate": 1.34589800443459e-05, "loss": 1.1708, "step": 607 }, { "epoch": 0.02, "learning_rate": 1.3481152993348117e-05, "loss": 1.1048, "step": 608 }, { "epoch": 0.02, "learning_rate": 1.3503325942350334e-05, "loss": 1.2195, "step": 609 }, { "epoch": 0.02, "learning_rate": 1.3525498891352551e-05, "loss": 1.1431, "step": 610 }, { "epoch": 0.02, "learning_rate": 1.3547671840354768e-05, "loss": 1.1888, "step": 611 }, { "epoch": 0.02, "learning_rate": 1.3569844789356985e-05, "loss": 1.1266, "step": 612 }, { "epoch": 0.02, "learning_rate": 1.3592017738359202e-05, "loss": 1.182, "step": 613 }, { "epoch": 0.02, "learning_rate": 1.361419068736142e-05, "loss": 1.1662, "step": 614 }, { "epoch": 0.02, "learning_rate": 1.3636363636363637e-05, "loss": 1.1943, "step": 615 }, { "epoch": 0.02, "learning_rate": 1.3658536585365855e-05, "loss": 1.0887, "step": 616 }, { "epoch": 0.02, "learning_rate": 1.3680709534368072e-05, "loss": 1.1878, "step": 617 }, { "epoch": 0.02, "learning_rate": 1.370288248337029e-05, "loss": 1.18, "step": 618 }, { "epoch": 0.02, "learning_rate": 1.3725055432372507e-05, "loss": 1.1883, "step": 619 }, { "epoch": 0.02, "learning_rate": 1.3747228381374724e-05, "loss": 1.1606, "step": 620 }, { "epoch": 0.02, "learning_rate": 1.3769401330376941e-05, "loss": 1.1512, "step": 621 }, { "epoch": 0.02, "learning_rate": 1.3791574279379158e-05, "loss": 1.1886, "step": 622 }, { "epoch": 0.02, "learning_rate": 1.3813747228381375e-05, "loss": 1.1598, "step": 623 }, { "epoch": 0.02, "learning_rate": 1.3835920177383592e-05, "loss": 1.1478, "step": 624 }, { "epoch": 0.02, "learning_rate": 1.385809312638581e-05, "loss": 1.1638, "step": 625 }, { "epoch": 0.02, "learning_rate": 1.3880266075388026e-05, "loss": 1.188, "step": 626 }, { "epoch": 0.02, "learning_rate": 1.3902439024390244e-05, "loss": 1.188, "step": 627 }, { "epoch": 0.02, "learning_rate": 1.3924611973392464e-05, "loss": 1.1144, "step": 628 }, { "epoch": 0.02, "learning_rate": 1.3946784922394681e-05, "loss": 1.1288, "step": 629 }, { "epoch": 0.02, "learning_rate": 1.3968957871396898e-05, "loss": 1.0777, "step": 630 }, { "epoch": 0.02, "learning_rate": 1.3991130820399115e-05, "loss": 1.1428, "step": 631 }, { "epoch": 0.02, "learning_rate": 1.4013303769401332e-05, "loss": 1.1539, "step": 632 }, { "epoch": 0.02, "learning_rate": 1.403547671840355e-05, "loss": 1.1202, "step": 633 }, { "epoch": 0.02, "learning_rate": 1.4057649667405767e-05, "loss": 1.1715, "step": 634 }, { "epoch": 0.02, "learning_rate": 1.4079822616407984e-05, "loss": 1.1473, "step": 635 }, { "epoch": 0.02, "learning_rate": 1.41019955654102e-05, "loss": 1.173, "step": 636 }, { "epoch": 0.02, "learning_rate": 1.4124168514412418e-05, "loss": 1.1699, "step": 637 }, { "epoch": 0.02, "learning_rate": 1.4146341463414635e-05, "loss": 1.1524, "step": 638 }, { "epoch": 0.02, "learning_rate": 1.4168514412416852e-05, "loss": 1.161, "step": 639 }, { "epoch": 0.02, "learning_rate": 1.4190687361419071e-05, "loss": 1.1941, "step": 640 }, { "epoch": 0.02, "learning_rate": 1.4212860310421288e-05, "loss": 1.182, "step": 641 }, { "epoch": 0.02, "learning_rate": 1.4235033259423505e-05, "loss": 1.142, "step": 642 }, { "epoch": 0.02, "learning_rate": 1.4257206208425722e-05, "loss": 1.1723, "step": 643 }, { "epoch": 0.02, "learning_rate": 1.427937915742794e-05, "loss": 1.1297, "step": 644 }, { "epoch": 0.02, "learning_rate": 1.4301552106430156e-05, "loss": 1.1792, "step": 645 }, { "epoch": 0.02, "learning_rate": 1.4323725055432374e-05, "loss": 1.1521, "step": 646 }, { "epoch": 0.02, "learning_rate": 1.434589800443459e-05, "loss": 1.1359, "step": 647 }, { "epoch": 0.02, "learning_rate": 1.4368070953436808e-05, "loss": 1.1525, "step": 648 }, { "epoch": 0.02, "learning_rate": 1.4390243902439025e-05, "loss": 1.1321, "step": 649 }, { "epoch": 0.02, "learning_rate": 1.4412416851441242e-05, "loss": 1.1805, "step": 650 }, { "epoch": 0.02, "learning_rate": 1.4434589800443459e-05, "loss": 1.1635, "step": 651 }, { "epoch": 0.02, "learning_rate": 1.4456762749445676e-05, "loss": 1.1444, "step": 652 }, { "epoch": 0.02, "learning_rate": 1.4478935698447895e-05, "loss": 1.1253, "step": 653 }, { "epoch": 0.02, "learning_rate": 1.4501108647450112e-05, "loss": 1.1323, "step": 654 }, { "epoch": 0.02, "learning_rate": 1.452328159645233e-05, "loss": 1.1557, "step": 655 }, { "epoch": 0.02, "learning_rate": 1.4545454545454546e-05, "loss": 1.1444, "step": 656 }, { "epoch": 0.02, "learning_rate": 1.4567627494456763e-05, "loss": 1.1785, "step": 657 }, { "epoch": 0.02, "learning_rate": 1.458980044345898e-05, "loss": 1.1899, "step": 658 }, { "epoch": 0.02, "learning_rate": 1.4611973392461198e-05, "loss": 1.1609, "step": 659 }, { "epoch": 0.02, "learning_rate": 1.4634146341463415e-05, "loss": 1.0693, "step": 660 }, { "epoch": 0.02, "learning_rate": 1.4656319290465632e-05, "loss": 1.1602, "step": 661 }, { "epoch": 0.02, "learning_rate": 1.4678492239467849e-05, "loss": 1.1566, "step": 662 }, { "epoch": 0.02, "learning_rate": 1.4700665188470066e-05, "loss": 1.1581, "step": 663 }, { "epoch": 0.02, "learning_rate": 1.4722838137472283e-05, "loss": 1.0885, "step": 664 }, { "epoch": 0.02, "learning_rate": 1.4745011086474504e-05, "loss": 1.1433, "step": 665 }, { "epoch": 0.02, "learning_rate": 1.476718403547672e-05, "loss": 1.1706, "step": 666 }, { "epoch": 0.02, "learning_rate": 1.4789356984478938e-05, "loss": 1.1542, "step": 667 }, { "epoch": 0.02, "learning_rate": 1.4811529933481155e-05, "loss": 1.1375, "step": 668 }, { "epoch": 0.02, "learning_rate": 1.4833702882483372e-05, "loss": 1.1747, "step": 669 }, { "epoch": 0.02, "learning_rate": 1.485587583148559e-05, "loss": 1.1397, "step": 670 }, { "epoch": 0.02, "learning_rate": 1.4878048780487806e-05, "loss": 1.1372, "step": 671 }, { "epoch": 0.02, "learning_rate": 1.4900221729490023e-05, "loss": 1.1314, "step": 672 }, { "epoch": 0.02, "learning_rate": 1.492239467849224e-05, "loss": 1.1427, "step": 673 }, { "epoch": 0.02, "learning_rate": 1.4944567627494458e-05, "loss": 1.1674, "step": 674 }, { "epoch": 0.02, "learning_rate": 1.4966740576496675e-05, "loss": 1.1594, "step": 675 }, { "epoch": 0.02, "learning_rate": 1.4988913525498892e-05, "loss": 1.1022, "step": 676 }, { "epoch": 0.02, "learning_rate": 1.501108647450111e-05, "loss": 1.1814, "step": 677 }, { "epoch": 0.02, "learning_rate": 1.5033259423503328e-05, "loss": 1.1693, "step": 678 }, { "epoch": 0.02, "learning_rate": 1.5055432372505545e-05, "loss": 1.1386, "step": 679 }, { "epoch": 0.02, "learning_rate": 1.5077605321507762e-05, "loss": 1.1555, "step": 680 }, { "epoch": 0.02, "learning_rate": 1.5099778270509979e-05, "loss": 1.1555, "step": 681 }, { "epoch": 0.02, "learning_rate": 1.5121951219512196e-05, "loss": 1.1659, "step": 682 }, { "epoch": 0.02, "learning_rate": 1.5144124168514413e-05, "loss": 1.1601, "step": 683 }, { "epoch": 0.02, "learning_rate": 1.516629711751663e-05, "loss": 1.1312, "step": 684 }, { "epoch": 0.02, "learning_rate": 1.5188470066518847e-05, "loss": 1.1552, "step": 685 }, { "epoch": 0.02, "learning_rate": 1.5210643015521065e-05, "loss": 1.1716, "step": 686 }, { "epoch": 0.02, "learning_rate": 1.5232815964523282e-05, "loss": 1.1985, "step": 687 }, { "epoch": 0.02, "learning_rate": 1.5254988913525499e-05, "loss": 1.1714, "step": 688 }, { "epoch": 0.02, "learning_rate": 1.5277161862527718e-05, "loss": 1.1221, "step": 689 }, { "epoch": 0.02, "learning_rate": 1.5299334811529935e-05, "loss": 1.1496, "step": 690 }, { "epoch": 0.02, "learning_rate": 1.5321507760532152e-05, "loss": 1.1478, "step": 691 }, { "epoch": 0.02, "learning_rate": 1.534368070953437e-05, "loss": 1.1043, "step": 692 }, { "epoch": 0.02, "learning_rate": 1.5365853658536586e-05, "loss": 1.137, "step": 693 }, { "epoch": 0.02, "learning_rate": 1.5388026607538803e-05, "loss": 1.1106, "step": 694 }, { "epoch": 0.02, "learning_rate": 1.541019955654102e-05, "loss": 1.1327, "step": 695 }, { "epoch": 0.02, "learning_rate": 1.5432372505543237e-05, "loss": 1.1627, "step": 696 }, { "epoch": 0.02, "learning_rate": 1.5454545454545454e-05, "loss": 1.1649, "step": 697 }, { "epoch": 0.02, "learning_rate": 1.547671840354767e-05, "loss": 1.1562, "step": 698 }, { "epoch": 0.02, "learning_rate": 1.549889135254989e-05, "loss": 1.1463, "step": 699 }, { "epoch": 0.02, "learning_rate": 1.5521064301552106e-05, "loss": 1.0791, "step": 700 }, { "epoch": 0.02, "learning_rate": 1.5543237250554326e-05, "loss": 1.1499, "step": 701 }, { "epoch": 0.02, "learning_rate": 1.5565410199556543e-05, "loss": 1.1537, "step": 702 }, { "epoch": 0.02, "learning_rate": 1.558758314855876e-05, "loss": 1.1426, "step": 703 }, { "epoch": 0.02, "learning_rate": 1.5609756097560978e-05, "loss": 1.0657, "step": 704 }, { "epoch": 0.02, "learning_rate": 1.5631929046563195e-05, "loss": 1.1398, "step": 705 }, { "epoch": 0.02, "learning_rate": 1.5654101995565412e-05, "loss": 1.121, "step": 706 }, { "epoch": 0.02, "learning_rate": 1.567627494456763e-05, "loss": 1.1146, "step": 707 }, { "epoch": 0.02, "learning_rate": 1.5698447893569846e-05, "loss": 1.1093, "step": 708 }, { "epoch": 0.02, "learning_rate": 1.5720620842572063e-05, "loss": 1.1501, "step": 709 }, { "epoch": 0.02, "learning_rate": 1.574279379157428e-05, "loss": 1.1424, "step": 710 }, { "epoch": 0.02, "learning_rate": 1.5764966740576497e-05, "loss": 1.1312, "step": 711 }, { "epoch": 0.02, "learning_rate": 1.5787139689578714e-05, "loss": 1.1329, "step": 712 }, { "epoch": 0.02, "learning_rate": 1.580931263858093e-05, "loss": 1.1661, "step": 713 }, { "epoch": 0.02, "learning_rate": 1.5831485587583152e-05, "loss": 1.14, "step": 714 }, { "epoch": 0.02, "learning_rate": 1.585365853658537e-05, "loss": 1.1232, "step": 715 }, { "epoch": 0.02, "learning_rate": 1.5875831485587586e-05, "loss": 1.0751, "step": 716 }, { "epoch": 0.02, "learning_rate": 1.5898004434589803e-05, "loss": 1.1287, "step": 717 }, { "epoch": 0.02, "learning_rate": 1.592017738359202e-05, "loss": 1.1779, "step": 718 }, { "epoch": 0.02, "learning_rate": 1.5942350332594238e-05, "loss": 1.1462, "step": 719 }, { "epoch": 0.02, "learning_rate": 1.5964523281596455e-05, "loss": 1.1319, "step": 720 }, { "epoch": 0.02, "learning_rate": 1.5986696230598672e-05, "loss": 1.1591, "step": 721 }, { "epoch": 0.02, "learning_rate": 1.600886917960089e-05, "loss": 1.1407, "step": 722 }, { "epoch": 0.02, "learning_rate": 1.6031042128603106e-05, "loss": 1.1633, "step": 723 }, { "epoch": 0.02, "learning_rate": 1.6053215077605323e-05, "loss": 1.0853, "step": 724 }, { "epoch": 0.02, "learning_rate": 1.607538802660754e-05, "loss": 1.181, "step": 725 }, { "epoch": 0.02, "learning_rate": 1.6097560975609757e-05, "loss": 1.1071, "step": 726 }, { "epoch": 0.02, "learning_rate": 1.6119733924611974e-05, "loss": 1.159, "step": 727 }, { "epoch": 0.02, "learning_rate": 1.614190687361419e-05, "loss": 1.0798, "step": 728 }, { "epoch": 0.02, "learning_rate": 1.616407982261641e-05, "loss": 1.1261, "step": 729 }, { "epoch": 0.02, "learning_rate": 1.6186252771618626e-05, "loss": 1.1441, "step": 730 }, { "epoch": 0.02, "learning_rate": 1.6208425720620843e-05, "loss": 1.1422, "step": 731 }, { "epoch": 0.02, "learning_rate": 1.623059866962306e-05, "loss": 1.1214, "step": 732 }, { "epoch": 0.02, "learning_rate": 1.6252771618625277e-05, "loss": 1.1344, "step": 733 }, { "epoch": 0.02, "learning_rate": 1.6274944567627494e-05, "loss": 1.0935, "step": 734 }, { "epoch": 0.02, "learning_rate": 1.629711751662971e-05, "loss": 1.1497, "step": 735 }, { "epoch": 0.02, "learning_rate": 1.6319290465631928e-05, "loss": 1.081, "step": 736 }, { "epoch": 0.02, "learning_rate": 1.6341463414634145e-05, "loss": 1.1504, "step": 737 }, { "epoch": 0.02, "learning_rate": 1.6363636363636366e-05, "loss": 1.1531, "step": 738 }, { "epoch": 0.02, "learning_rate": 1.6385809312638583e-05, "loss": 1.1488, "step": 739 }, { "epoch": 0.02, "learning_rate": 1.64079822616408e-05, "loss": 1.1185, "step": 740 }, { "epoch": 0.02, "learning_rate": 1.6430155210643017e-05, "loss": 1.1391, "step": 741 }, { "epoch": 0.02, "learning_rate": 1.6452328159645234e-05, "loss": 1.1364, "step": 742 }, { "epoch": 0.02, "learning_rate": 1.647450110864745e-05, "loss": 1.1548, "step": 743 }, { "epoch": 0.02, "learning_rate": 1.649667405764967e-05, "loss": 1.1305, "step": 744 }, { "epoch": 0.02, "learning_rate": 1.6518847006651886e-05, "loss": 1.1848, "step": 745 }, { "epoch": 0.02, "learning_rate": 1.6541019955654103e-05, "loss": 1.1098, "step": 746 }, { "epoch": 0.02, "learning_rate": 1.656319290465632e-05, "loss": 1.1337, "step": 747 }, { "epoch": 0.02, "learning_rate": 1.6585365853658537e-05, "loss": 1.1084, "step": 748 }, { "epoch": 0.02, "learning_rate": 1.6607538802660754e-05, "loss": 1.1774, "step": 749 }, { "epoch": 0.02, "learning_rate": 1.6629711751662975e-05, "loss": 1.173, "step": 750 }, { "epoch": 0.02, "learning_rate": 1.665188470066519e-05, "loss": 1.1362, "step": 751 }, { "epoch": 0.03, "learning_rate": 1.667405764966741e-05, "loss": 1.0775, "step": 752 }, { "epoch": 0.03, "learning_rate": 1.6696230598669626e-05, "loss": 1.1397, "step": 753 }, { "epoch": 0.03, "learning_rate": 1.6718403547671843e-05, "loss": 1.122, "step": 754 }, { "epoch": 0.03, "learning_rate": 1.674057649667406e-05, "loss": 1.0989, "step": 755 }, { "epoch": 0.03, "learning_rate": 1.6762749445676277e-05, "loss": 1.125, "step": 756 }, { "epoch": 0.03, "learning_rate": 1.6784922394678494e-05, "loss": 1.1605, "step": 757 }, { "epoch": 0.03, "learning_rate": 1.680709534368071e-05, "loss": 1.067, "step": 758 }, { "epoch": 0.03, "learning_rate": 1.682926829268293e-05, "loss": 1.1261, "step": 759 }, { "epoch": 0.03, "learning_rate": 1.6851441241685146e-05, "loss": 1.0985, "step": 760 }, { "epoch": 0.03, "learning_rate": 1.6873614190687363e-05, "loss": 1.1644, "step": 761 }, { "epoch": 0.03, "learning_rate": 1.689578713968958e-05, "loss": 1.1079, "step": 762 }, { "epoch": 0.03, "learning_rate": 1.6917960088691797e-05, "loss": 1.1031, "step": 763 }, { "epoch": 0.03, "learning_rate": 1.6940133037694014e-05, "loss": 1.1249, "step": 764 }, { "epoch": 0.03, "learning_rate": 1.696230598669623e-05, "loss": 1.152, "step": 765 }, { "epoch": 0.03, "learning_rate": 1.6984478935698448e-05, "loss": 1.1294, "step": 766 }, { "epoch": 0.03, "learning_rate": 1.7006651884700665e-05, "loss": 1.1825, "step": 767 }, { "epoch": 0.03, "learning_rate": 1.7028824833702882e-05, "loss": 1.1299, "step": 768 }, { "epoch": 0.03, "learning_rate": 1.70509977827051e-05, "loss": 1.1567, "step": 769 }, { "epoch": 0.03, "learning_rate": 1.7073170731707317e-05, "loss": 1.1272, "step": 770 }, { "epoch": 0.03, "learning_rate": 1.7095343680709534e-05, "loss": 1.1595, "step": 771 }, { "epoch": 0.03, "learning_rate": 1.711751662971175e-05, "loss": 1.1321, "step": 772 }, { "epoch": 0.03, "learning_rate": 1.7139689578713968e-05, "loss": 1.1476, "step": 773 }, { "epoch": 0.03, "learning_rate": 1.7161862527716185e-05, "loss": 1.1425, "step": 774 }, { "epoch": 0.03, "learning_rate": 1.7184035476718406e-05, "loss": 1.1488, "step": 775 }, { "epoch": 0.03, "learning_rate": 1.7206208425720623e-05, "loss": 1.1132, "step": 776 }, { "epoch": 0.03, "learning_rate": 1.722838137472284e-05, "loss": 1.1379, "step": 777 }, { "epoch": 0.03, "learning_rate": 1.7250554323725057e-05, "loss": 1.1532, "step": 778 }, { "epoch": 0.03, "learning_rate": 1.7272727272727274e-05, "loss": 1.1414, "step": 779 }, { "epoch": 0.03, "learning_rate": 1.729490022172949e-05, "loss": 1.158, "step": 780 }, { "epoch": 0.03, "learning_rate": 1.7317073170731708e-05, "loss": 1.1204, "step": 781 }, { "epoch": 0.03, "learning_rate": 1.7339246119733925e-05, "loss": 1.1275, "step": 782 }, { "epoch": 0.03, "learning_rate": 1.7361419068736142e-05, "loss": 1.1634, "step": 783 }, { "epoch": 0.03, "learning_rate": 1.738359201773836e-05, "loss": 1.1099, "step": 784 }, { "epoch": 0.03, "learning_rate": 1.7405764966740577e-05, "loss": 1.1478, "step": 785 }, { "epoch": 0.03, "learning_rate": 1.7427937915742794e-05, "loss": 1.1291, "step": 786 }, { "epoch": 0.03, "learning_rate": 1.7450110864745014e-05, "loss": 1.1449, "step": 787 }, { "epoch": 0.03, "learning_rate": 1.747228381374723e-05, "loss": 1.0803, "step": 788 }, { "epoch": 0.03, "learning_rate": 1.749445676274945e-05, "loss": 1.1551, "step": 789 }, { "epoch": 0.03, "learning_rate": 1.7516629711751666e-05, "loss": 1.1358, "step": 790 }, { "epoch": 0.03, "learning_rate": 1.7538802660753883e-05, "loss": 1.0959, "step": 791 }, { "epoch": 0.03, "learning_rate": 1.75609756097561e-05, "loss": 1.0714, "step": 792 }, { "epoch": 0.03, "learning_rate": 1.7583148558758317e-05, "loss": 1.1495, "step": 793 }, { "epoch": 0.03, "learning_rate": 1.7605321507760534e-05, "loss": 1.1378, "step": 794 }, { "epoch": 0.03, "learning_rate": 1.762749445676275e-05, "loss": 1.1699, "step": 795 }, { "epoch": 0.03, "learning_rate": 1.7649667405764968e-05, "loss": 1.0983, "step": 796 }, { "epoch": 0.03, "learning_rate": 1.7671840354767185e-05, "loss": 1.1323, "step": 797 }, { "epoch": 0.03, "learning_rate": 1.7694013303769402e-05, "loss": 1.1108, "step": 798 }, { "epoch": 0.03, "learning_rate": 1.771618625277162e-05, "loss": 1.1132, "step": 799 }, { "epoch": 0.03, "learning_rate": 1.7738359201773837e-05, "loss": 1.0887, "step": 800 }, { "epoch": 0.03, "learning_rate": 1.7760532150776054e-05, "loss": 1.1353, "step": 801 }, { "epoch": 0.03, "learning_rate": 1.778270509977827e-05, "loss": 1.146, "step": 802 }, { "epoch": 0.03, "learning_rate": 1.7804878048780488e-05, "loss": 1.1806, "step": 803 }, { "epoch": 0.03, "learning_rate": 1.7827050997782705e-05, "loss": 1.0936, "step": 804 }, { "epoch": 0.03, "learning_rate": 1.7849223946784922e-05, "loss": 1.1447, "step": 805 }, { "epoch": 0.03, "learning_rate": 1.787139689578714e-05, "loss": 1.1823, "step": 806 }, { "epoch": 0.03, "learning_rate": 1.7893569844789356e-05, "loss": 1.1507, "step": 807 }, { "epoch": 0.03, "learning_rate": 1.7915742793791573e-05, "loss": 1.1027, "step": 808 }, { "epoch": 0.03, "learning_rate": 1.793791574279379e-05, "loss": 1.139, "step": 809 }, { "epoch": 0.03, "learning_rate": 1.796008869179601e-05, "loss": 1.1196, "step": 810 }, { "epoch": 0.03, "learning_rate": 1.7982261640798228e-05, "loss": 1.1698, "step": 811 }, { "epoch": 0.03, "learning_rate": 1.8004434589800445e-05, "loss": 1.0976, "step": 812 }, { "epoch": 0.03, "learning_rate": 1.8026607538802662e-05, "loss": 1.1403, "step": 813 }, { "epoch": 0.03, "learning_rate": 1.804878048780488e-05, "loss": 1.1176, "step": 814 }, { "epoch": 0.03, "learning_rate": 1.8070953436807097e-05, "loss": 1.1464, "step": 815 }, { "epoch": 0.03, "learning_rate": 1.8093126385809314e-05, "loss": 1.0792, "step": 816 }, { "epoch": 0.03, "learning_rate": 1.811529933481153e-05, "loss": 1.146, "step": 817 }, { "epoch": 0.03, "learning_rate": 1.8137472283813748e-05, "loss": 1.1469, "step": 818 }, { "epoch": 0.03, "learning_rate": 1.8159645232815965e-05, "loss": 1.1367, "step": 819 }, { "epoch": 0.03, "learning_rate": 1.8181818181818182e-05, "loss": 1.1158, "step": 820 }, { "epoch": 0.03, "learning_rate": 1.82039911308204e-05, "loss": 1.1332, "step": 821 }, { "epoch": 0.03, "learning_rate": 1.8226164079822616e-05, "loss": 1.1247, "step": 822 }, { "epoch": 0.03, "learning_rate": 1.8248337028824837e-05, "loss": 1.1527, "step": 823 }, { "epoch": 0.03, "learning_rate": 1.8270509977827054e-05, "loss": 1.0742, "step": 824 }, { "epoch": 0.03, "learning_rate": 1.829268292682927e-05, "loss": 1.174, "step": 825 }, { "epoch": 0.03, "learning_rate": 1.8314855875831488e-05, "loss": 1.1107, "step": 826 }, { "epoch": 0.03, "learning_rate": 1.8337028824833705e-05, "loss": 1.122, "step": 827 }, { "epoch": 0.03, "learning_rate": 1.8359201773835922e-05, "loss": 1.0789, "step": 828 }, { "epoch": 0.03, "learning_rate": 1.838137472283814e-05, "loss": 1.1448, "step": 829 }, { "epoch": 0.03, "learning_rate": 1.8403547671840357e-05, "loss": 1.1179, "step": 830 }, { "epoch": 0.03, "learning_rate": 1.8425720620842574e-05, "loss": 1.1228, "step": 831 }, { "epoch": 0.03, "learning_rate": 1.844789356984479e-05, "loss": 1.0905, "step": 832 }, { "epoch": 0.03, "learning_rate": 1.8470066518847008e-05, "loss": 1.1241, "step": 833 }, { "epoch": 0.03, "learning_rate": 1.8492239467849225e-05, "loss": 1.1061, "step": 834 }, { "epoch": 0.03, "learning_rate": 1.8514412416851445e-05, "loss": 1.1079, "step": 835 }, { "epoch": 0.03, "learning_rate": 1.8536585365853663e-05, "loss": 1.0779, "step": 836 }, { "epoch": 0.03, "learning_rate": 1.855875831485588e-05, "loss": 1.1355, "step": 837 }, { "epoch": 0.03, "learning_rate": 1.8580931263858097e-05, "loss": 1.1529, "step": 838 }, { "epoch": 0.03, "learning_rate": 1.8603104212860314e-05, "loss": 1.149, "step": 839 }, { "epoch": 0.03, "learning_rate": 1.862527716186253e-05, "loss": 1.1337, "step": 840 }, { "epoch": 0.03, "learning_rate": 1.8647450110864748e-05, "loss": 1.1478, "step": 841 }, { "epoch": 0.03, "learning_rate": 1.8669623059866965e-05, "loss": 1.1742, "step": 842 }, { "epoch": 0.03, "learning_rate": 1.8691796008869182e-05, "loss": 1.1461, "step": 843 }, { "epoch": 0.03, "learning_rate": 1.87139689578714e-05, "loss": 1.1274, "step": 844 }, { "epoch": 0.03, "learning_rate": 1.8736141906873616e-05, "loss": 1.1438, "step": 845 }, { "epoch": 0.03, "learning_rate": 1.8758314855875834e-05, "loss": 1.1226, "step": 846 }, { "epoch": 0.03, "learning_rate": 1.878048780487805e-05, "loss": 1.1491, "step": 847 }, { "epoch": 0.03, "learning_rate": 1.8802660753880268e-05, "loss": 1.0956, "step": 848 }, { "epoch": 0.03, "learning_rate": 1.8824833702882485e-05, "loss": 1.1444, "step": 849 }, { "epoch": 0.03, "learning_rate": 1.8847006651884702e-05, "loss": 1.1019, "step": 850 }, { "epoch": 0.03, "learning_rate": 1.886917960088692e-05, "loss": 1.1511, "step": 851 }, { "epoch": 0.03, "learning_rate": 1.8891352549889136e-05, "loss": 1.0902, "step": 852 }, { "epoch": 0.03, "learning_rate": 1.8913525498891353e-05, "loss": 1.1497, "step": 853 }, { "epoch": 0.03, "learning_rate": 1.893569844789357e-05, "loss": 1.1282, "step": 854 }, { "epoch": 0.03, "learning_rate": 1.8957871396895788e-05, "loss": 1.1297, "step": 855 }, { "epoch": 0.03, "learning_rate": 1.8980044345898005e-05, "loss": 1.1037, "step": 856 }, { "epoch": 0.03, "learning_rate": 1.9002217294900222e-05, "loss": 1.1268, "step": 857 }, { "epoch": 0.03, "learning_rate": 1.902439024390244e-05, "loss": 1.1476, "step": 858 }, { "epoch": 0.03, "learning_rate": 1.9046563192904656e-05, "loss": 1.1278, "step": 859 }, { "epoch": 0.03, "learning_rate": 1.9068736141906876e-05, "loss": 1.0773, "step": 860 }, { "epoch": 0.03, "learning_rate": 1.9090909090909094e-05, "loss": 1.1691, "step": 861 }, { "epoch": 0.03, "learning_rate": 1.911308203991131e-05, "loss": 1.117, "step": 862 }, { "epoch": 0.03, "learning_rate": 1.9135254988913528e-05, "loss": 1.1528, "step": 863 }, { "epoch": 0.03, "learning_rate": 1.9157427937915745e-05, "loss": 1.123, "step": 864 }, { "epoch": 0.03, "learning_rate": 1.9179600886917962e-05, "loss": 1.1156, "step": 865 }, { "epoch": 0.03, "learning_rate": 1.920177383592018e-05, "loss": 1.1011, "step": 866 }, { "epoch": 0.03, "learning_rate": 1.9223946784922396e-05, "loss": 1.1549, "step": 867 }, { "epoch": 0.03, "learning_rate": 1.9246119733924613e-05, "loss": 1.0897, "step": 868 }, { "epoch": 0.03, "learning_rate": 1.926829268292683e-05, "loss": 1.115, "step": 869 }, { "epoch": 0.03, "learning_rate": 1.9290465631929047e-05, "loss": 1.0832, "step": 870 }, { "epoch": 0.03, "learning_rate": 1.9312638580931265e-05, "loss": 1.1224, "step": 871 }, { "epoch": 0.03, "learning_rate": 1.9334811529933485e-05, "loss": 1.1336, "step": 872 }, { "epoch": 0.03, "learning_rate": 1.9356984478935702e-05, "loss": 1.1336, "step": 873 }, { "epoch": 0.03, "learning_rate": 1.937915742793792e-05, "loss": 1.1256, "step": 874 }, { "epoch": 0.03, "learning_rate": 1.9401330376940136e-05, "loss": 1.1551, "step": 875 }, { "epoch": 0.03, "learning_rate": 1.9423503325942354e-05, "loss": 1.0896, "step": 876 }, { "epoch": 0.03, "learning_rate": 1.944567627494457e-05, "loss": 1.169, "step": 877 }, { "epoch": 0.03, "learning_rate": 1.9467849223946788e-05, "loss": 1.1195, "step": 878 }, { "epoch": 0.03, "learning_rate": 1.9490022172949005e-05, "loss": 1.1365, "step": 879 }, { "epoch": 0.03, "learning_rate": 1.9512195121951222e-05, "loss": 1.0986, "step": 880 }, { "epoch": 0.03, "learning_rate": 1.953436807095344e-05, "loss": 1.1663, "step": 881 }, { "epoch": 0.03, "learning_rate": 1.9556541019955656e-05, "loss": 1.1423, "step": 882 }, { "epoch": 0.03, "learning_rate": 1.9578713968957873e-05, "loss": 1.1382, "step": 883 }, { "epoch": 0.03, "learning_rate": 1.960088691796009e-05, "loss": 1.1228, "step": 884 }, { "epoch": 0.03, "learning_rate": 1.9623059866962307e-05, "loss": 1.1605, "step": 885 }, { "epoch": 0.03, "learning_rate": 1.9645232815964525e-05, "loss": 1.1472, "step": 886 }, { "epoch": 0.03, "learning_rate": 1.966740576496674e-05, "loss": 1.1799, "step": 887 }, { "epoch": 0.03, "learning_rate": 1.968957871396896e-05, "loss": 1.1282, "step": 888 }, { "epoch": 0.03, "learning_rate": 1.9711751662971176e-05, "loss": 1.1083, "step": 889 }, { "epoch": 0.03, "learning_rate": 1.9733924611973393e-05, "loss": 1.1469, "step": 890 }, { "epoch": 0.03, "learning_rate": 1.975609756097561e-05, "loss": 1.1321, "step": 891 }, { "epoch": 0.03, "learning_rate": 1.9778270509977827e-05, "loss": 1.1178, "step": 892 }, { "epoch": 0.03, "learning_rate": 1.9800443458980044e-05, "loss": 1.0849, "step": 893 }, { "epoch": 0.03, "learning_rate": 1.982261640798226e-05, "loss": 1.1219, "step": 894 }, { "epoch": 0.03, "learning_rate": 1.984478935698448e-05, "loss": 1.1475, "step": 895 }, { "epoch": 0.03, "learning_rate": 1.98669623059867e-05, "loss": 1.1185, "step": 896 }, { "epoch": 0.03, "learning_rate": 1.9889135254988916e-05, "loss": 1.1175, "step": 897 }, { "epoch": 0.03, "learning_rate": 1.9911308203991133e-05, "loss": 1.1405, "step": 898 }, { "epoch": 0.03, "learning_rate": 1.993348115299335e-05, "loss": 1.0891, "step": 899 }, { "epoch": 0.03, "learning_rate": 1.9955654101995567e-05, "loss": 1.1247, "step": 900 }, { "epoch": 0.03, "learning_rate": 1.9977827050997785e-05, "loss": 1.096, "step": 901 }, { "epoch": 0.03, "learning_rate": 2e-05, "loss": 1.0932, "step": 902 }, { "epoch": 0.03, "learning_rate": 1.9999999941908602e-05, "loss": 1.103, "step": 903 }, { "epoch": 0.03, "learning_rate": 1.9999999767634412e-05, "loss": 1.0772, "step": 904 }, { "epoch": 0.03, "learning_rate": 1.9999999477177426e-05, "loss": 1.1504, "step": 905 }, { "epoch": 0.03, "learning_rate": 1.9999999070537647e-05, "loss": 1.0941, "step": 906 }, { "epoch": 0.03, "learning_rate": 1.9999998547715085e-05, "loss": 1.1309, "step": 907 }, { "epoch": 0.03, "learning_rate": 1.9999997908709746e-05, "loss": 1.0786, "step": 908 }, { "epoch": 0.03, "learning_rate": 1.9999997153521632e-05, "loss": 1.1196, "step": 909 }, { "epoch": 0.03, "learning_rate": 1.9999996282150755e-05, "loss": 1.1348, "step": 910 }, { "epoch": 0.03, "learning_rate": 1.999999529459713e-05, "loss": 1.1092, "step": 911 }, { "epoch": 0.03, "learning_rate": 1.9999994190860757e-05, "loss": 1.0711, "step": 912 }, { "epoch": 0.03, "learning_rate": 1.999999297094166e-05, "loss": 1.1215, "step": 913 }, { "epoch": 0.03, "learning_rate": 1.9999991634839847e-05, "loss": 1.1694, "step": 914 }, { "epoch": 0.03, "learning_rate": 1.9999990182555337e-05, "loss": 1.12, "step": 915 }, { "epoch": 0.03, "learning_rate": 1.9999988614088142e-05, "loss": 1.0929, "step": 916 }, { "epoch": 0.03, "learning_rate": 1.9999986929438286e-05, "loss": 1.1338, "step": 917 }, { "epoch": 0.03, "learning_rate": 1.9999985128605782e-05, "loss": 1.1313, "step": 918 }, { "epoch": 0.03, "learning_rate": 1.9999983211590658e-05, "loss": 1.1462, "step": 919 }, { "epoch": 0.03, "learning_rate": 1.9999981178392934e-05, "loss": 1.1115, "step": 920 }, { "epoch": 0.03, "learning_rate": 1.999997902901263e-05, "loss": 1.0978, "step": 921 }, { "epoch": 0.03, "learning_rate": 1.9999976763449775e-05, "loss": 1.1145, "step": 922 }, { "epoch": 0.03, "learning_rate": 1.9999974381704394e-05, "loss": 1.1454, "step": 923 }, { "epoch": 0.03, "learning_rate": 1.9999971883776514e-05, "loss": 1.1348, "step": 924 }, { "epoch": 0.03, "learning_rate": 1.9999969269666166e-05, "loss": 1.1183, "step": 925 }, { "epoch": 0.03, "learning_rate": 1.9999966539373378e-05, "loss": 1.1327, "step": 926 }, { "epoch": 0.03, "learning_rate": 1.9999963692898183e-05, "loss": 1.1357, "step": 927 }, { "epoch": 0.03, "learning_rate": 1.9999960730240615e-05, "loss": 1.1078, "step": 928 }, { "epoch": 0.03, "learning_rate": 1.9999957651400704e-05, "loss": 1.145, "step": 929 }, { "epoch": 0.03, "learning_rate": 1.999995445637849e-05, "loss": 1.1173, "step": 930 }, { "epoch": 0.03, "learning_rate": 1.999995114517401e-05, "loss": 1.1291, "step": 931 }, { "epoch": 0.03, "learning_rate": 1.9999947717787303e-05, "loss": 1.1127, "step": 932 }, { "epoch": 0.03, "learning_rate": 1.9999944174218406e-05, "loss": 1.1316, "step": 933 }, { "epoch": 0.03, "learning_rate": 1.999994051446736e-05, "loss": 1.1079, "step": 934 }, { "epoch": 0.03, "learning_rate": 1.999993673853421e-05, "loss": 1.1105, "step": 935 }, { "epoch": 0.03, "learning_rate": 1.9999932846419e-05, "loss": 1.1262, "step": 936 }, { "epoch": 0.03, "learning_rate": 1.9999928838121776e-05, "loss": 1.1188, "step": 937 }, { "epoch": 0.03, "learning_rate": 1.999992471364258e-05, "loss": 1.1214, "step": 938 }, { "epoch": 0.03, "learning_rate": 1.9999920472981462e-05, "loss": 1.1151, "step": 939 }, { "epoch": 0.03, "learning_rate": 1.9999916116138475e-05, "loss": 1.0386, "step": 940 }, { "epoch": 0.03, "learning_rate": 1.9999911643113666e-05, "loss": 1.1539, "step": 941 }, { "epoch": 0.03, "learning_rate": 1.9999907053907085e-05, "loss": 1.142, "step": 942 }, { "epoch": 0.03, "learning_rate": 1.999990234851879e-05, "loss": 1.1228, "step": 943 }, { "epoch": 0.03, "learning_rate": 1.9999897526948832e-05, "loss": 1.1172, "step": 944 }, { "epoch": 0.03, "learning_rate": 1.999989258919727e-05, "loss": 1.1521, "step": 945 }, { "epoch": 0.03, "learning_rate": 1.9999887535264157e-05, "loss": 1.1337, "step": 946 }, { "epoch": 0.03, "learning_rate": 1.999988236514956e-05, "loss": 1.1087, "step": 947 }, { "epoch": 0.03, "learning_rate": 1.9999877078853528e-05, "loss": 1.0975, "step": 948 }, { "epoch": 0.03, "learning_rate": 1.9999871676376132e-05, "loss": 1.1756, "step": 949 }, { "epoch": 0.03, "learning_rate": 1.9999866157717428e-05, "loss": 1.0963, "step": 950 }, { "epoch": 0.03, "learning_rate": 1.999986052287749e-05, "loss": 1.1089, "step": 951 }, { "epoch": 0.03, "learning_rate": 1.9999854771856366e-05, "loss": 1.1093, "step": 952 }, { "epoch": 0.03, "learning_rate": 1.999984890465414e-05, "loss": 1.1247, "step": 953 }, { "epoch": 0.03, "learning_rate": 1.999984292127087e-05, "loss": 1.096, "step": 954 }, { "epoch": 0.03, "learning_rate": 1.999983682170663e-05, "loss": 1.1355, "step": 955 }, { "epoch": 0.03, "learning_rate": 1.9999830605961495e-05, "loss": 1.0852, "step": 956 }, { "epoch": 0.03, "learning_rate": 1.9999824274035525e-05, "loss": 1.1967, "step": 957 }, { "epoch": 0.03, "learning_rate": 1.9999817825928804e-05, "loss": 1.1134, "step": 958 }, { "epoch": 0.03, "learning_rate": 1.99998112616414e-05, "loss": 1.1161, "step": 959 }, { "epoch": 0.03, "learning_rate": 1.99998045811734e-05, "loss": 1.1031, "step": 960 }, { "epoch": 0.03, "learning_rate": 1.9999797784524866e-05, "loss": 1.1286, "step": 961 }, { "epoch": 0.03, "learning_rate": 1.999979087169589e-05, "loss": 1.0892, "step": 962 }, { "epoch": 0.03, "learning_rate": 1.9999783842686548e-05, "loss": 1.1482, "step": 963 }, { "epoch": 0.03, "learning_rate": 1.9999776697496923e-05, "loss": 1.059, "step": 964 }, { "epoch": 0.03, "learning_rate": 1.9999769436127093e-05, "loss": 1.1013, "step": 965 }, { "epoch": 0.03, "learning_rate": 1.9999762058577147e-05, "loss": 1.0965, "step": 966 }, { "epoch": 0.03, "learning_rate": 1.999975456484717e-05, "loss": 1.1252, "step": 967 }, { "epoch": 0.03, "learning_rate": 1.9999746954937244e-05, "loss": 1.1139, "step": 968 }, { "epoch": 0.03, "learning_rate": 1.9999739228847466e-05, "loss": 1.1268, "step": 969 }, { "epoch": 0.03, "learning_rate": 1.999973138657792e-05, "loss": 1.1461, "step": 970 }, { "epoch": 0.03, "learning_rate": 1.9999723428128704e-05, "loss": 1.1136, "step": 971 }, { "epoch": 0.03, "learning_rate": 1.99997153534999e-05, "loss": 1.0557, "step": 972 }, { "epoch": 0.03, "learning_rate": 1.9999707162691607e-05, "loss": 1.1603, "step": 973 }, { "epoch": 0.03, "learning_rate": 1.9999698855703924e-05, "loss": 1.1304, "step": 974 }, { "epoch": 0.03, "learning_rate": 1.999969043253694e-05, "loss": 1.0797, "step": 975 }, { "epoch": 0.03, "learning_rate": 1.999968189319076e-05, "loss": 1.08, "step": 976 }, { "epoch": 0.03, "learning_rate": 1.999967323766548e-05, "loss": 1.1251, "step": 977 }, { "epoch": 0.03, "learning_rate": 1.9999664465961195e-05, "loss": 1.0975, "step": 978 }, { "epoch": 0.03, "learning_rate": 1.9999655578078016e-05, "loss": 1.1648, "step": 979 }, { "epoch": 0.03, "learning_rate": 1.9999646574016044e-05, "loss": 1.0861, "step": 980 }, { "epoch": 0.03, "learning_rate": 1.999963745377538e-05, "loss": 1.0697, "step": 981 }, { "epoch": 0.03, "learning_rate": 1.999962821735613e-05, "loss": 1.0835, "step": 982 }, { "epoch": 0.03, "learning_rate": 1.999961886475841e-05, "loss": 1.1213, "step": 983 }, { "epoch": 0.03, "learning_rate": 1.9999609395982314e-05, "loss": 1.0837, "step": 984 }, { "epoch": 0.03, "learning_rate": 1.9999599811027967e-05, "loss": 1.1172, "step": 985 }, { "epoch": 0.03, "learning_rate": 1.9999590109895467e-05, "loss": 1.1329, "step": 986 }, { "epoch": 0.03, "learning_rate": 1.999958029258494e-05, "loss": 1.1911, "step": 987 }, { "epoch": 0.03, "learning_rate": 1.999957035909649e-05, "loss": 1.0619, "step": 988 }, { "epoch": 0.03, "learning_rate": 1.9999560309430234e-05, "loss": 1.1458, "step": 989 }, { "epoch": 0.03, "learning_rate": 1.9999550143586295e-05, "loss": 1.1356, "step": 990 }, { "epoch": 0.03, "learning_rate": 1.9999539861564784e-05, "loss": 1.0852, "step": 991 }, { "epoch": 0.03, "learning_rate": 1.999952946336582e-05, "loss": 1.0835, "step": 992 }, { "epoch": 0.03, "learning_rate": 1.999951894898953e-05, "loss": 1.0919, "step": 993 }, { "epoch": 0.03, "learning_rate": 1.9999508318436035e-05, "loss": 1.0651, "step": 994 }, { "epoch": 0.03, "learning_rate": 1.9999497571705456e-05, "loss": 1.126, "step": 995 }, { "epoch": 0.03, "learning_rate": 1.999948670879791e-05, "loss": 1.0745, "step": 996 }, { "epoch": 0.03, "learning_rate": 1.9999475729713543e-05, "loss": 1.1174, "step": 997 }, { "epoch": 0.03, "learning_rate": 1.9999464634452467e-05, "loss": 1.1426, "step": 998 }, { "epoch": 0.03, "learning_rate": 1.9999453423014815e-05, "loss": 1.0947, "step": 999 }, { "epoch": 0.03, "learning_rate": 1.9999442095400713e-05, "loss": 1.0944, "step": 1000 }, { "epoch": 0.03, "learning_rate": 1.9999430651610302e-05, "loss": 1.1152, "step": 1001 }, { "epoch": 0.03, "learning_rate": 1.999941909164371e-05, "loss": 1.1254, "step": 1002 }, { "epoch": 0.03, "learning_rate": 1.9999407415501068e-05, "loss": 1.1069, "step": 1003 }, { "epoch": 0.03, "learning_rate": 1.9999395623182515e-05, "loss": 1.0402, "step": 1004 }, { "epoch": 0.03, "learning_rate": 1.9999383714688187e-05, "loss": 1.1565, "step": 1005 }, { "epoch": 0.03, "learning_rate": 1.9999371690018227e-05, "loss": 1.1674, "step": 1006 }, { "epoch": 0.03, "learning_rate": 1.999935954917277e-05, "loss": 1.106, "step": 1007 }, { "epoch": 0.03, "learning_rate": 1.9999347292151955e-05, "loss": 1.0915, "step": 1008 }, { "epoch": 0.03, "learning_rate": 1.999933491895593e-05, "loss": 1.1184, "step": 1009 }, { "epoch": 0.03, "learning_rate": 1.9999322429584835e-05, "loss": 1.1082, "step": 1010 }, { "epoch": 0.03, "learning_rate": 1.9999309824038816e-05, "loss": 1.0975, "step": 1011 }, { "epoch": 0.03, "learning_rate": 1.999929710231802e-05, "loss": 1.1112, "step": 1012 }, { "epoch": 0.03, "learning_rate": 1.9999284264422595e-05, "loss": 1.0914, "step": 1013 }, { "epoch": 0.03, "learning_rate": 1.999927131035269e-05, "loss": 1.1159, "step": 1014 }, { "epoch": 0.03, "learning_rate": 1.9999258240108455e-05, "loss": 1.1106, "step": 1015 }, { "epoch": 0.03, "learning_rate": 1.999924505369004e-05, "loss": 1.0782, "step": 1016 }, { "epoch": 0.03, "learning_rate": 1.99992317510976e-05, "loss": 1.1306, "step": 1017 }, { "epoch": 0.03, "learning_rate": 1.9999218332331292e-05, "loss": 1.0985, "step": 1018 }, { "epoch": 0.03, "learning_rate": 1.999920479739127e-05, "loss": 1.1108, "step": 1019 }, { "epoch": 0.03, "learning_rate": 1.999919114627769e-05, "loss": 1.1022, "step": 1020 }, { "epoch": 0.03, "learning_rate": 1.999917737899071e-05, "loss": 1.1238, "step": 1021 }, { "epoch": 0.03, "learning_rate": 1.9999163495530495e-05, "loss": 1.1322, "step": 1022 }, { "epoch": 0.03, "learning_rate": 1.9999149495897198e-05, "loss": 1.1052, "step": 1023 }, { "epoch": 0.03, "learning_rate": 1.999913538009099e-05, "loss": 1.0832, "step": 1024 }, { "epoch": 0.03, "learning_rate": 1.999912114811203e-05, "loss": 1.1385, "step": 1025 }, { "epoch": 0.03, "learning_rate": 1.9999106799960483e-05, "loss": 1.1195, "step": 1026 }, { "epoch": 0.03, "learning_rate": 1.999909233563652e-05, "loss": 1.0825, "step": 1027 }, { "epoch": 0.03, "learning_rate": 1.9999077755140305e-05, "loss": 1.127, "step": 1028 }, { "epoch": 0.03, "learning_rate": 1.999906305847201e-05, "loss": 1.1388, "step": 1029 }, { "epoch": 0.03, "learning_rate": 1.9999048245631803e-05, "loss": 1.08, "step": 1030 }, { "epoch": 0.03, "learning_rate": 1.999903331661986e-05, "loss": 1.1319, "step": 1031 }, { "epoch": 0.03, "learning_rate": 1.9999018271436352e-05, "loss": 1.0388, "step": 1032 }, { "epoch": 0.03, "learning_rate": 1.9999003110081453e-05, "loss": 1.1259, "step": 1033 }, { "epoch": 0.03, "learning_rate": 1.999898783255534e-05, "loss": 1.1248, "step": 1034 }, { "epoch": 0.03, "learning_rate": 1.999897243885819e-05, "loss": 1.1481, "step": 1035 }, { "epoch": 0.03, "learning_rate": 1.9998956928990184e-05, "loss": 1.0388, "step": 1036 }, { "epoch": 0.03, "learning_rate": 1.99989413029515e-05, "loss": 1.1242, "step": 1037 }, { "epoch": 0.03, "learning_rate": 1.999892556074232e-05, "loss": 1.0864, "step": 1038 }, { "epoch": 0.03, "learning_rate": 1.9998909702362827e-05, "loss": 1.1238, "step": 1039 }, { "epoch": 0.03, "learning_rate": 1.9998893727813206e-05, "loss": 1.0756, "step": 1040 }, { "epoch": 0.03, "learning_rate": 1.9998877637093642e-05, "loss": 1.1145, "step": 1041 }, { "epoch": 0.03, "learning_rate": 1.9998861430204323e-05, "loss": 1.1484, "step": 1042 }, { "epoch": 0.03, "learning_rate": 1.9998845107145435e-05, "loss": 1.118, "step": 1043 }, { "epoch": 0.03, "learning_rate": 1.9998828667917167e-05, "loss": 1.0885, "step": 1044 }, { "epoch": 0.03, "learning_rate": 1.9998812112519713e-05, "loss": 1.1029, "step": 1045 }, { "epoch": 0.03, "learning_rate": 1.9998795440953266e-05, "loss": 1.1198, "step": 1046 }, { "epoch": 0.03, "learning_rate": 1.9998778653218015e-05, "loss": 1.1713, "step": 1047 }, { "epoch": 0.03, "learning_rate": 1.999876174931416e-05, "loss": 1.0891, "step": 1048 }, { "epoch": 0.03, "learning_rate": 1.9998744729241895e-05, "loss": 1.0597, "step": 1049 }, { "epoch": 0.03, "learning_rate": 1.999872759300142e-05, "loss": 1.1107, "step": 1050 }, { "epoch": 0.03, "learning_rate": 1.999871034059293e-05, "loss": 1.0819, "step": 1051 }, { "epoch": 0.04, "learning_rate": 1.999869297201663e-05, "loss": 1.0811, "step": 1052 }, { "epoch": 0.04, "learning_rate": 1.9998675487272715e-05, "loss": 1.1149, "step": 1053 }, { "epoch": 0.04, "learning_rate": 1.9998657886361393e-05, "loss": 1.1447, "step": 1054 }, { "epoch": 0.04, "learning_rate": 1.999864016928287e-05, "loss": 1.0432, "step": 1055 }, { "epoch": 0.04, "learning_rate": 1.999862233603735e-05, "loss": 1.0742, "step": 1056 }, { "epoch": 0.04, "learning_rate": 1.999860438662504e-05, "loss": 1.1188, "step": 1057 }, { "epoch": 0.04, "learning_rate": 1.9998586321046148e-05, "loss": 1.1087, "step": 1058 }, { "epoch": 0.04, "learning_rate": 1.9998568139300888e-05, "loss": 1.1706, "step": 1059 }, { "epoch": 0.04, "learning_rate": 1.9998549841389464e-05, "loss": 1.0703, "step": 1060 }, { "epoch": 0.04, "learning_rate": 1.9998531427312092e-05, "loss": 1.0986, "step": 1061 }, { "epoch": 0.04, "learning_rate": 1.9998512897068987e-05, "loss": 1.0697, "step": 1062 }, { "epoch": 0.04, "learning_rate": 1.999849425066036e-05, "loss": 1.1016, "step": 1063 }, { "epoch": 0.04, "learning_rate": 1.999847548808644e-05, "loss": 1.0482, "step": 1064 }, { "epoch": 0.04, "learning_rate": 1.999845660934743e-05, "loss": 1.0949, "step": 1065 }, { "epoch": 0.04, "learning_rate": 1.9998437614443554e-05, "loss": 1.1398, "step": 1066 }, { "epoch": 0.04, "learning_rate": 1.999841850337504e-05, "loss": 1.1044, "step": 1067 }, { "epoch": 0.04, "learning_rate": 1.99983992761421e-05, "loss": 1.0675, "step": 1068 }, { "epoch": 0.04, "learning_rate": 1.9998379932744964e-05, "loss": 1.0887, "step": 1069 }, { "epoch": 0.04, "learning_rate": 1.9998360473183852e-05, "loss": 1.0885, "step": 1070 }, { "epoch": 0.04, "learning_rate": 1.9998340897458994e-05, "loss": 1.112, "step": 1071 }, { "epoch": 0.04, "learning_rate": 1.9998321205570616e-05, "loss": 1.1084, "step": 1072 }, { "epoch": 0.04, "learning_rate": 1.9998301397518946e-05, "loss": 1.1099, "step": 1073 }, { "epoch": 0.04, "learning_rate": 1.9998281473304216e-05, "loss": 1.1195, "step": 1074 }, { "epoch": 0.04, "learning_rate": 1.9998261432926655e-05, "loss": 1.1038, "step": 1075 }, { "epoch": 0.04, "learning_rate": 1.9998241276386498e-05, "loss": 1.061, "step": 1076 }, { "epoch": 0.04, "learning_rate": 1.999822100368398e-05, "loss": 1.0954, "step": 1077 }, { "epoch": 0.04, "learning_rate": 1.999820061481933e-05, "loss": 1.1016, "step": 1078 }, { "epoch": 0.04, "learning_rate": 1.9998180109792793e-05, "loss": 1.0914, "step": 1079 }, { "epoch": 0.04, "learning_rate": 1.9998159488604602e-05, "loss": 1.106, "step": 1080 }, { "epoch": 0.04, "learning_rate": 1.9998138751255e-05, "loss": 1.1268, "step": 1081 }, { "epoch": 0.04, "learning_rate": 1.9998117897744228e-05, "loss": 1.0965, "step": 1082 }, { "epoch": 0.04, "learning_rate": 1.9998096928072526e-05, "loss": 1.0701, "step": 1083 }, { "epoch": 0.04, "learning_rate": 1.9998075842240136e-05, "loss": 1.0464, "step": 1084 }, { "epoch": 0.04, "learning_rate": 1.9998054640247307e-05, "loss": 1.1335, "step": 1085 }, { "epoch": 0.04, "learning_rate": 1.9998033322094284e-05, "loss": 1.1002, "step": 1086 }, { "epoch": 0.04, "learning_rate": 1.9998011887781314e-05, "loss": 1.0905, "step": 1087 }, { "epoch": 0.04, "learning_rate": 1.999799033730865e-05, "loss": 1.0631, "step": 1088 }, { "epoch": 0.04, "learning_rate": 1.9997968670676534e-05, "loss": 1.1405, "step": 1089 }, { "epoch": 0.04, "learning_rate": 1.9997946887885224e-05, "loss": 1.0945, "step": 1090 }, { "epoch": 0.04, "learning_rate": 1.999792498893497e-05, "loss": 1.0709, "step": 1091 }, { "epoch": 0.04, "learning_rate": 1.999790297382603e-05, "loss": 1.0727, "step": 1092 }, { "epoch": 0.04, "learning_rate": 1.9997880842558657e-05, "loss": 1.0959, "step": 1093 }, { "epoch": 0.04, "learning_rate": 1.999785859513311e-05, "loss": 1.0949, "step": 1094 }, { "epoch": 0.04, "learning_rate": 1.9997836231549645e-05, "loss": 1.0797, "step": 1095 }, { "epoch": 0.04, "learning_rate": 1.9997813751808523e-05, "loss": 1.0656, "step": 1096 }, { "epoch": 0.04, "learning_rate": 1.999779115591001e-05, "loss": 1.125, "step": 1097 }, { "epoch": 0.04, "learning_rate": 1.999776844385436e-05, "loss": 1.0941, "step": 1098 }, { "epoch": 0.04, "learning_rate": 1.9997745615641842e-05, "loss": 1.0987, "step": 1099 }, { "epoch": 0.04, "learning_rate": 1.9997722671272716e-05, "loss": 1.0353, "step": 1100 }, { "epoch": 0.04, "learning_rate": 1.999769961074726e-05, "loss": 1.1216, "step": 1101 }, { "epoch": 0.04, "learning_rate": 1.9997676434065728e-05, "loss": 1.1016, "step": 1102 }, { "epoch": 0.04, "learning_rate": 1.99976531412284e-05, "loss": 1.1274, "step": 1103 }, { "epoch": 0.04, "learning_rate": 1.999762973223554e-05, "loss": 1.0311, "step": 1104 }, { "epoch": 0.04, "learning_rate": 1.9997606207087422e-05, "loss": 1.1244, "step": 1105 }, { "epoch": 0.04, "learning_rate": 1.9997582565784324e-05, "loss": 1.0637, "step": 1106 }, { "epoch": 0.04, "learning_rate": 1.999755880832651e-05, "loss": 1.115, "step": 1107 }, { "epoch": 0.04, "learning_rate": 1.9997534934714267e-05, "loss": 1.0931, "step": 1108 }, { "epoch": 0.04, "learning_rate": 1.9997510944947866e-05, "loss": 1.1623, "step": 1109 }, { "epoch": 0.04, "learning_rate": 1.999748683902759e-05, "loss": 1.0813, "step": 1110 }, { "epoch": 0.04, "learning_rate": 1.9997462616953712e-05, "loss": 1.1172, "step": 1111 }, { "epoch": 0.04, "learning_rate": 1.9997438278726525e-05, "loss": 1.0826, "step": 1112 }, { "epoch": 0.04, "learning_rate": 1.99974138243463e-05, "loss": 1.1241, "step": 1113 }, { "epoch": 0.04, "learning_rate": 1.9997389253813324e-05, "loss": 1.0956, "step": 1114 }, { "epoch": 0.04, "learning_rate": 1.999736456712789e-05, "loss": 1.0949, "step": 1115 }, { "epoch": 0.04, "learning_rate": 1.9997339764290275e-05, "loss": 1.081, "step": 1116 }, { "epoch": 0.04, "learning_rate": 1.9997314845300772e-05, "loss": 1.1138, "step": 1117 }, { "epoch": 0.04, "learning_rate": 1.999728981015967e-05, "loss": 1.0858, "step": 1118 }, { "epoch": 0.04, "learning_rate": 1.9997264658867262e-05, "loss": 1.0977, "step": 1119 }, { "epoch": 0.04, "learning_rate": 1.9997239391423834e-05, "loss": 1.0961, "step": 1120 }, { "epoch": 0.04, "learning_rate": 1.9997214007829685e-05, "loss": 1.0886, "step": 1121 }, { "epoch": 0.04, "learning_rate": 1.999718850808511e-05, "loss": 1.1215, "step": 1122 }, { "epoch": 0.04, "learning_rate": 1.9997162892190403e-05, "loss": 1.1319, "step": 1123 }, { "epoch": 0.04, "learning_rate": 1.999713716014586e-05, "loss": 1.0422, "step": 1124 }, { "epoch": 0.04, "learning_rate": 1.9997111311951784e-05, "loss": 1.0831, "step": 1125 }, { "epoch": 0.04, "learning_rate": 1.9997085347608475e-05, "loss": 1.0576, "step": 1126 }, { "epoch": 0.04, "learning_rate": 1.999705926711623e-05, "loss": 1.1, "step": 1127 }, { "epoch": 0.04, "learning_rate": 1.999703307047536e-05, "loss": 1.066, "step": 1128 }, { "epoch": 0.04, "learning_rate": 1.9997006757686162e-05, "loss": 1.1243, "step": 1129 }, { "epoch": 0.04, "learning_rate": 1.9996980328748944e-05, "loss": 1.0728, "step": 1130 }, { "epoch": 0.04, "learning_rate": 1.9996953783664014e-05, "loss": 1.0854, "step": 1131 }, { "epoch": 0.04, "learning_rate": 1.9996927122431676e-05, "loss": 1.0601, "step": 1132 }, { "epoch": 0.04, "learning_rate": 1.999690034505225e-05, "loss": 1.1102, "step": 1133 }, { "epoch": 0.04, "learning_rate": 1.9996873451526038e-05, "loss": 1.1169, "step": 1134 }, { "epoch": 0.04, "learning_rate": 1.9996846441853352e-05, "loss": 1.1082, "step": 1135 }, { "epoch": 0.04, "learning_rate": 1.999681931603451e-05, "loss": 1.0808, "step": 1136 }, { "epoch": 0.04, "learning_rate": 1.999679207406983e-05, "loss": 1.0618, "step": 1137 }, { "epoch": 0.04, "learning_rate": 1.999676471595962e-05, "loss": 1.0655, "step": 1138 }, { "epoch": 0.04, "learning_rate": 1.9996737241704208e-05, "loss": 1.059, "step": 1139 }, { "epoch": 0.04, "learning_rate": 1.9996709651303903e-05, "loss": 1.1139, "step": 1140 }, { "epoch": 0.04, "learning_rate": 1.999668194475903e-05, "loss": 1.1062, "step": 1141 }, { "epoch": 0.04, "learning_rate": 1.999665412206991e-05, "loss": 1.1121, "step": 1142 }, { "epoch": 0.04, "learning_rate": 1.9996626183236874e-05, "loss": 1.0751, "step": 1143 }, { "epoch": 0.04, "learning_rate": 1.9996598128260235e-05, "loss": 1.0842, "step": 1144 }, { "epoch": 0.04, "learning_rate": 1.9996569957140325e-05, "loss": 1.1022, "step": 1145 }, { "epoch": 0.04, "learning_rate": 1.999654166987747e-05, "loss": 1.1055, "step": 1146 }, { "epoch": 0.04, "learning_rate": 1.9996513266471998e-05, "loss": 1.081, "step": 1147 }, { "epoch": 0.04, "learning_rate": 1.9996484746924238e-05, "loss": 1.035, "step": 1148 }, { "epoch": 0.04, "learning_rate": 1.999645611123453e-05, "loss": 1.1547, "step": 1149 }, { "epoch": 0.04, "learning_rate": 1.9996427359403192e-05, "loss": 1.0678, "step": 1150 }, { "epoch": 0.04, "learning_rate": 1.999639849143057e-05, "loss": 1.1061, "step": 1151 }, { "epoch": 0.04, "learning_rate": 1.9996369507316996e-05, "loss": 1.0448, "step": 1152 }, { "epoch": 0.04, "learning_rate": 1.9996340407062804e-05, "loss": 1.101, "step": 1153 }, { "epoch": 0.04, "learning_rate": 1.9996311190668333e-05, "loss": 1.0705, "step": 1154 }, { "epoch": 0.04, "learning_rate": 1.999628185813393e-05, "loss": 1.09, "step": 1155 }, { "epoch": 0.04, "learning_rate": 1.9996252409459924e-05, "loss": 1.0825, "step": 1156 }, { "epoch": 0.04, "learning_rate": 1.9996222844646663e-05, "loss": 1.0967, "step": 1157 }, { "epoch": 0.04, "learning_rate": 1.999619316369449e-05, "loss": 1.1095, "step": 1158 }, { "epoch": 0.04, "learning_rate": 1.999616336660375e-05, "loss": 1.0608, "step": 1159 }, { "epoch": 0.04, "learning_rate": 1.999613345337479e-05, "loss": 1.0651, "step": 1160 }, { "epoch": 0.04, "learning_rate": 1.9996103424007954e-05, "loss": 1.1246, "step": 1161 }, { "epoch": 0.04, "learning_rate": 1.9996073278503596e-05, "loss": 1.0789, "step": 1162 }, { "epoch": 0.04, "learning_rate": 1.9996043016862065e-05, "loss": 1.1499, "step": 1163 }, { "epoch": 0.04, "learning_rate": 1.999601263908371e-05, "loss": 1.0569, "step": 1164 }, { "epoch": 0.04, "learning_rate": 1.9995982145168884e-05, "loss": 1.0992, "step": 1165 }, { "epoch": 0.04, "learning_rate": 1.9995951535117942e-05, "loss": 1.0878, "step": 1166 }, { "epoch": 0.04, "learning_rate": 1.999592080893124e-05, "loss": 1.0863, "step": 1167 }, { "epoch": 0.04, "learning_rate": 1.999588996660914e-05, "loss": 1.0782, "step": 1168 }, { "epoch": 0.04, "learning_rate": 1.999585900815199e-05, "loss": 1.1242, "step": 1169 }, { "epoch": 0.04, "learning_rate": 1.9995827933560158e-05, "loss": 1.0892, "step": 1170 }, { "epoch": 0.04, "learning_rate": 1.9995796742834e-05, "loss": 1.0728, "step": 1171 }, { "epoch": 0.04, "learning_rate": 1.9995765435973886e-05, "loss": 1.0848, "step": 1172 }, { "epoch": 0.04, "learning_rate": 1.999573401298017e-05, "loss": 1.1162, "step": 1173 }, { "epoch": 0.04, "learning_rate": 1.9995702473853224e-05, "loss": 1.1342, "step": 1174 }, { "epoch": 0.04, "learning_rate": 1.9995670818593406e-05, "loss": 1.1345, "step": 1175 }, { "epoch": 0.04, "learning_rate": 1.9995639047201093e-05, "loss": 1.0455, "step": 1176 }, { "epoch": 0.04, "learning_rate": 1.9995607159676656e-05, "loss": 1.1553, "step": 1177 }, { "epoch": 0.04, "learning_rate": 1.9995575156020454e-05, "loss": 1.085, "step": 1178 }, { "epoch": 0.04, "learning_rate": 1.9995543036232864e-05, "loss": 1.0776, "step": 1179 }, { "epoch": 0.04, "learning_rate": 1.9995510800314265e-05, "loss": 1.0745, "step": 1180 }, { "epoch": 0.04, "learning_rate": 1.9995478448265023e-05, "loss": 1.1079, "step": 1181 }, { "epoch": 0.04, "learning_rate": 1.999544598008552e-05, "loss": 1.1151, "step": 1182 }, { "epoch": 0.04, "learning_rate": 1.999541339577613e-05, "loss": 1.0864, "step": 1183 }, { "epoch": 0.04, "learning_rate": 1.999538069533723e-05, "loss": 1.1115, "step": 1184 }, { "epoch": 0.04, "learning_rate": 1.999534787876921e-05, "loss": 1.1222, "step": 1185 }, { "epoch": 0.04, "learning_rate": 1.9995314946072437e-05, "loss": 1.0596, "step": 1186 }, { "epoch": 0.04, "learning_rate": 1.99952818972473e-05, "loss": 1.1163, "step": 1187 }, { "epoch": 0.04, "learning_rate": 1.999524873229419e-05, "loss": 1.075, "step": 1188 }, { "epoch": 0.04, "learning_rate": 1.9995215451213478e-05, "loss": 1.0637, "step": 1189 }, { "epoch": 0.04, "learning_rate": 1.999518205400556e-05, "loss": 1.1142, "step": 1190 }, { "epoch": 0.04, "learning_rate": 1.9995148540670823e-05, "loss": 1.1073, "step": 1191 }, { "epoch": 0.04, "learning_rate": 1.9995114911209655e-05, "loss": 1.0545, "step": 1192 }, { "epoch": 0.04, "learning_rate": 1.999508116562245e-05, "loss": 1.0974, "step": 1193 }, { "epoch": 0.04, "learning_rate": 1.9995047303909593e-05, "loss": 1.1476, "step": 1194 }, { "epoch": 0.04, "learning_rate": 1.9995013326071486e-05, "loss": 1.0988, "step": 1195 }, { "epoch": 0.04, "learning_rate": 1.999497923210852e-05, "loss": 1.0874, "step": 1196 }, { "epoch": 0.04, "learning_rate": 1.9994945022021085e-05, "loss": 1.0809, "step": 1197 }, { "epoch": 0.04, "learning_rate": 1.9994910695809588e-05, "loss": 1.1092, "step": 1198 }, { "epoch": 0.04, "learning_rate": 1.999487625347442e-05, "loss": 1.076, "step": 1199 }, { "epoch": 0.04, "learning_rate": 1.999484169501599e-05, "loss": 1.085, "step": 1200 }, { "epoch": 0.04, "learning_rate": 1.9994807020434693e-05, "loss": 1.0914, "step": 1201 }, { "epoch": 0.04, "learning_rate": 1.9994772229730938e-05, "loss": 1.0872, "step": 1202 }, { "epoch": 0.04, "learning_rate": 1.999473732290512e-05, "loss": 1.1108, "step": 1203 }, { "epoch": 0.04, "learning_rate": 1.9994702299957645e-05, "loss": 1.0738, "step": 1204 }, { "epoch": 0.04, "learning_rate": 1.999466716088893e-05, "loss": 1.0981, "step": 1205 }, { "epoch": 0.04, "learning_rate": 1.999463190569938e-05, "loss": 1.0941, "step": 1206 }, { "epoch": 0.04, "learning_rate": 1.9994596534389396e-05, "loss": 1.0858, "step": 1207 }, { "epoch": 0.04, "learning_rate": 1.9994561046959395e-05, "loss": 1.1124, "step": 1208 }, { "epoch": 0.04, "learning_rate": 1.9994525443409793e-05, "loss": 1.0852, "step": 1209 }, { "epoch": 0.04, "learning_rate": 1.9994489723740996e-05, "loss": 1.075, "step": 1210 }, { "epoch": 0.04, "learning_rate": 1.9994453887953425e-05, "loss": 1.0765, "step": 1211 }, { "epoch": 0.04, "learning_rate": 1.9994417936047497e-05, "loss": 1.042, "step": 1212 }, { "epoch": 0.04, "learning_rate": 1.9994381868023624e-05, "loss": 1.0905, "step": 1213 }, { "epoch": 0.04, "learning_rate": 1.9994345683882227e-05, "loss": 1.0978, "step": 1214 }, { "epoch": 0.04, "learning_rate": 1.999430938362373e-05, "loss": 1.128, "step": 1215 }, { "epoch": 0.04, "learning_rate": 1.999427296724855e-05, "loss": 1.0459, "step": 1216 }, { "epoch": 0.04, "learning_rate": 1.9994236434757113e-05, "loss": 1.0871, "step": 1217 }, { "epoch": 0.04, "learning_rate": 1.9994199786149845e-05, "loss": 1.0895, "step": 1218 }, { "epoch": 0.04, "learning_rate": 1.9994163021427167e-05, "loss": 1.0972, "step": 1219 }, { "epoch": 0.04, "learning_rate": 1.999412614058951e-05, "loss": 1.0178, "step": 1220 }, { "epoch": 0.04, "learning_rate": 1.9994089143637298e-05, "loss": 1.0618, "step": 1221 }, { "epoch": 0.04, "learning_rate": 1.999405203057097e-05, "loss": 1.0546, "step": 1222 }, { "epoch": 0.04, "learning_rate": 1.9994014801390943e-05, "loss": 1.1154, "step": 1223 }, { "epoch": 0.04, "learning_rate": 1.9993977456097663e-05, "loss": 1.0715, "step": 1224 }, { "epoch": 0.04, "learning_rate": 1.9993939994691558e-05, "loss": 1.1356, "step": 1225 }, { "epoch": 0.04, "learning_rate": 1.9993902417173065e-05, "loss": 1.0798, "step": 1226 }, { "epoch": 0.04, "learning_rate": 1.9993864723542615e-05, "loss": 1.0657, "step": 1227 }, { "epoch": 0.04, "learning_rate": 1.9993826913800654e-05, "loss": 1.0786, "step": 1228 }, { "epoch": 0.04, "learning_rate": 1.9993788987947617e-05, "loss": 1.1097, "step": 1229 }, { "epoch": 0.04, "learning_rate": 1.9993750945983944e-05, "loss": 1.086, "step": 1230 }, { "epoch": 0.04, "learning_rate": 1.9993712787910082e-05, "loss": 1.0983, "step": 1231 }, { "epoch": 0.04, "learning_rate": 1.9993674513726465e-05, "loss": 1.0575, "step": 1232 }, { "epoch": 0.04, "learning_rate": 1.9993636123433547e-05, "loss": 1.129, "step": 1233 }, { "epoch": 0.04, "learning_rate": 1.9993597617031772e-05, "loss": 1.1105, "step": 1234 }, { "epoch": 0.04, "learning_rate": 1.9993558994521583e-05, "loss": 1.0475, "step": 1235 }, { "epoch": 0.04, "learning_rate": 1.9993520255903428e-05, "loss": 1.0094, "step": 1236 }, { "epoch": 0.04, "learning_rate": 1.9993481401177764e-05, "loss": 1.0728, "step": 1237 }, { "epoch": 0.04, "learning_rate": 1.999344243034504e-05, "loss": 1.0679, "step": 1238 }, { "epoch": 0.04, "learning_rate": 1.9993403343405706e-05, "loss": 1.049, "step": 1239 }, { "epoch": 0.04, "learning_rate": 1.999336414036022e-05, "loss": 1.0422, "step": 1240 }, { "epoch": 0.04, "learning_rate": 1.9993324821209033e-05, "loss": 1.0943, "step": 1241 }, { "epoch": 0.04, "learning_rate": 1.9993285385952605e-05, "loss": 1.0601, "step": 1242 }, { "epoch": 0.04, "learning_rate": 1.9993245834591392e-05, "loss": 1.0578, "step": 1243 }, { "epoch": 0.04, "learning_rate": 1.9993206167125855e-05, "loss": 1.0277, "step": 1244 }, { "epoch": 0.04, "learning_rate": 1.9993166383556458e-05, "loss": 1.1173, "step": 1245 }, { "epoch": 0.04, "learning_rate": 1.9993126483883656e-05, "loss": 1.1048, "step": 1246 }, { "epoch": 0.04, "learning_rate": 1.999308646810792e-05, "loss": 1.1134, "step": 1247 }, { "epoch": 0.04, "learning_rate": 1.9993046336229707e-05, "loss": 1.0767, "step": 1248 }, { "epoch": 0.04, "learning_rate": 1.999300608824949e-05, "loss": 1.1096, "step": 1249 }, { "epoch": 0.04, "learning_rate": 1.9992965724167736e-05, "loss": 1.0979, "step": 1250 }, { "epoch": 0.04, "learning_rate": 1.9992925243984908e-05, "loss": 1.0696, "step": 1251 }, { "epoch": 0.04, "learning_rate": 1.9992884647701483e-05, "loss": 1.0549, "step": 1252 }, { "epoch": 0.04, "learning_rate": 1.9992843935317928e-05, "loss": 1.1029, "step": 1253 }, { "epoch": 0.04, "learning_rate": 1.9992803106834718e-05, "loss": 1.0925, "step": 1254 }, { "epoch": 0.04, "learning_rate": 1.999276216225233e-05, "loss": 1.1111, "step": 1255 }, { "epoch": 0.04, "learning_rate": 1.9992721101571238e-05, "loss": 1.0515, "step": 1256 }, { "epoch": 0.04, "learning_rate": 1.9992679924791915e-05, "loss": 1.1115, "step": 1257 }, { "epoch": 0.04, "learning_rate": 1.9992638631914845e-05, "loss": 1.0969, "step": 1258 }, { "epoch": 0.04, "learning_rate": 1.9992597222940506e-05, "loss": 1.103, "step": 1259 }, { "epoch": 0.04, "learning_rate": 1.9992555697869377e-05, "loss": 1.0733, "step": 1260 }, { "epoch": 0.04, "learning_rate": 1.999251405670194e-05, "loss": 1.1012, "step": 1261 }, { "epoch": 0.04, "learning_rate": 1.9992472299438685e-05, "loss": 1.1269, "step": 1262 }, { "epoch": 0.04, "learning_rate": 1.999243042608009e-05, "loss": 1.0637, "step": 1263 }, { "epoch": 0.04, "learning_rate": 1.9992388436626645e-05, "loss": 1.0071, "step": 1264 }, { "epoch": 0.04, "learning_rate": 1.999234633107884e-05, "loss": 1.0631, "step": 1265 }, { "epoch": 0.04, "learning_rate": 1.9992304109437157e-05, "loss": 1.068, "step": 1266 }, { "epoch": 0.04, "learning_rate": 1.9992261771702094e-05, "loss": 1.0868, "step": 1267 }, { "epoch": 0.04, "learning_rate": 1.9992219317874138e-05, "loss": 1.0979, "step": 1268 }, { "epoch": 0.04, "learning_rate": 1.9992176747953788e-05, "loss": 1.0938, "step": 1269 }, { "epoch": 0.04, "learning_rate": 1.9992134061941528e-05, "loss": 1.0711, "step": 1270 }, { "epoch": 0.04, "learning_rate": 1.9992091259837867e-05, "loss": 1.0898, "step": 1271 }, { "epoch": 0.04, "learning_rate": 1.999204834164329e-05, "loss": 1.0518, "step": 1272 }, { "epoch": 0.04, "learning_rate": 1.9992005307358307e-05, "loss": 1.1169, "step": 1273 }, { "epoch": 0.04, "learning_rate": 1.9991962156983412e-05, "loss": 1.0979, "step": 1274 }, { "epoch": 0.04, "learning_rate": 1.9991918890519104e-05, "loss": 1.0651, "step": 1275 }, { "epoch": 0.04, "learning_rate": 1.999187550796589e-05, "loss": 1.019, "step": 1276 }, { "epoch": 0.04, "learning_rate": 1.9991832009324272e-05, "loss": 1.1007, "step": 1277 }, { "epoch": 0.04, "learning_rate": 1.999178839459476e-05, "loss": 1.1158, "step": 1278 }, { "epoch": 0.04, "learning_rate": 1.999174466377785e-05, "loss": 1.0755, "step": 1279 }, { "epoch": 0.04, "learning_rate": 1.9991700816874062e-05, "loss": 1.0495, "step": 1280 }, { "epoch": 0.04, "learning_rate": 1.99916568538839e-05, "loss": 1.0795, "step": 1281 }, { "epoch": 0.04, "learning_rate": 1.999161277480787e-05, "loss": 1.0683, "step": 1282 }, { "epoch": 0.04, "learning_rate": 1.9991568579646495e-05, "loss": 1.0812, "step": 1283 }, { "epoch": 0.04, "learning_rate": 1.9991524268400278e-05, "loss": 1.0478, "step": 1284 }, { "epoch": 0.04, "learning_rate": 1.999147984106974e-05, "loss": 1.0843, "step": 1285 }, { "epoch": 0.04, "learning_rate": 1.9991435297655396e-05, "loss": 1.0919, "step": 1286 }, { "epoch": 0.04, "learning_rate": 1.9991390638157762e-05, "loss": 1.105, "step": 1287 }, { "epoch": 0.04, "learning_rate": 1.999134586257736e-05, "loss": 1.0457, "step": 1288 }, { "epoch": 0.04, "learning_rate": 1.9991300970914707e-05, "loss": 1.0659, "step": 1289 }, { "epoch": 0.04, "learning_rate": 1.9991255963170326e-05, "loss": 1.1352, "step": 1290 }, { "epoch": 0.04, "learning_rate": 1.9991210839344737e-05, "loss": 1.105, "step": 1291 }, { "epoch": 0.04, "learning_rate": 1.9991165599438475e-05, "loss": 1.0484, "step": 1292 }, { "epoch": 0.04, "learning_rate": 1.999112024345205e-05, "loss": 1.1037, "step": 1293 }, { "epoch": 0.04, "learning_rate": 1.9991074771385998e-05, "loss": 1.072, "step": 1294 }, { "epoch": 0.04, "learning_rate": 1.9991029183240848e-05, "loss": 1.0997, "step": 1295 }, { "epoch": 0.04, "learning_rate": 1.9990983479017124e-05, "loss": 1.0416, "step": 1296 }, { "epoch": 0.04, "learning_rate": 1.9990937658715365e-05, "loss": 1.1016, "step": 1297 }, { "epoch": 0.04, "learning_rate": 1.9990891722336095e-05, "loss": 1.0763, "step": 1298 }, { "epoch": 0.04, "learning_rate": 1.999084566987985e-05, "loss": 1.0761, "step": 1299 }, { "epoch": 0.04, "learning_rate": 1.999079950134717e-05, "loss": 1.0329, "step": 1300 }, { "epoch": 0.04, "learning_rate": 1.9990753216738587e-05, "loss": 1.0717, "step": 1301 }, { "epoch": 0.04, "learning_rate": 1.999070681605464e-05, "loss": 1.1079, "step": 1302 }, { "epoch": 0.04, "learning_rate": 1.999066029929587e-05, "loss": 1.0422, "step": 1303 }, { "epoch": 0.04, "learning_rate": 1.999061366646281e-05, "loss": 1.0384, "step": 1304 }, { "epoch": 0.04, "learning_rate": 1.9990566917556007e-05, "loss": 1.1168, "step": 1305 }, { "epoch": 0.04, "learning_rate": 1.9990520052576012e-05, "loss": 1.138, "step": 1306 }, { "epoch": 0.04, "learning_rate": 1.9990473071523354e-05, "loss": 1.071, "step": 1307 }, { "epoch": 0.04, "learning_rate": 1.999042597439859e-05, "loss": 1.0398, "step": 1308 }, { "epoch": 0.04, "learning_rate": 1.9990378761202263e-05, "loss": 1.1044, "step": 1309 }, { "epoch": 0.04, "learning_rate": 1.9990331431934923e-05, "loss": 1.0848, "step": 1310 }, { "epoch": 0.04, "learning_rate": 1.9990283986597117e-05, "loss": 1.0505, "step": 1311 }, { "epoch": 0.04, "learning_rate": 1.99902364251894e-05, "loss": 1.0344, "step": 1312 }, { "epoch": 0.04, "learning_rate": 1.999018874771232e-05, "loss": 1.0793, "step": 1313 }, { "epoch": 0.04, "learning_rate": 1.999014095416644e-05, "loss": 1.0791, "step": 1314 }, { "epoch": 0.04, "learning_rate": 1.9990093044552304e-05, "loss": 1.1309, "step": 1315 }, { "epoch": 0.04, "learning_rate": 1.9990045018870474e-05, "loss": 1.0529, "step": 1316 }, { "epoch": 0.04, "learning_rate": 1.9989996877121508e-05, "loss": 1.1073, "step": 1317 }, { "epoch": 0.04, "learning_rate": 1.998994861930597e-05, "loss": 1.0783, "step": 1318 }, { "epoch": 0.04, "learning_rate": 1.998990024542441e-05, "loss": 1.0585, "step": 1319 }, { "epoch": 0.04, "learning_rate": 1.99898517554774e-05, "loss": 1.0366, "step": 1320 }, { "epoch": 0.04, "learning_rate": 1.9989803149465494e-05, "loss": 1.1031, "step": 1321 }, { "epoch": 0.04, "learning_rate": 1.9989754427389265e-05, "loss": 1.1016, "step": 1322 }, { "epoch": 0.04, "learning_rate": 1.9989705589249278e-05, "loss": 1.1225, "step": 1323 }, { "epoch": 0.04, "learning_rate": 1.9989656635046092e-05, "loss": 1.0347, "step": 1324 }, { "epoch": 0.04, "learning_rate": 1.9989607564780287e-05, "loss": 1.0974, "step": 1325 }, { "epoch": 0.04, "learning_rate": 1.9989558378452428e-05, "loss": 1.0794, "step": 1326 }, { "epoch": 0.04, "learning_rate": 1.9989509076063085e-05, "loss": 1.0953, "step": 1327 }, { "epoch": 0.04, "learning_rate": 1.9989459657612836e-05, "loss": 1.0552, "step": 1328 }, { "epoch": 0.04, "learning_rate": 1.9989410123102245e-05, "loss": 1.1197, "step": 1329 }, { "epoch": 0.04, "learning_rate": 1.99893604725319e-05, "loss": 1.0863, "step": 1330 }, { "epoch": 0.04, "learning_rate": 1.998931070590237e-05, "loss": 1.0901, "step": 1331 }, { "epoch": 0.04, "learning_rate": 1.9989260823214238e-05, "loss": 1.0397, "step": 1332 }, { "epoch": 0.04, "learning_rate": 1.9989210824468077e-05, "loss": 1.0816, "step": 1333 }, { "epoch": 0.04, "learning_rate": 1.998916070966447e-05, "loss": 1.1059, "step": 1334 }, { "epoch": 0.04, "learning_rate": 1.9989110478804004e-05, "loss": 1.1222, "step": 1335 }, { "epoch": 0.04, "learning_rate": 1.998906013188726e-05, "loss": 1.0696, "step": 1336 }, { "epoch": 0.04, "learning_rate": 1.998900966891482e-05, "loss": 1.0853, "step": 1337 }, { "epoch": 0.04, "learning_rate": 1.9988959089887274e-05, "loss": 1.114, "step": 1338 }, { "epoch": 0.04, "learning_rate": 1.9988908394805207e-05, "loss": 1.1007, "step": 1339 }, { "epoch": 0.04, "learning_rate": 1.998885758366921e-05, "loss": 1.1202, "step": 1340 }, { "epoch": 0.04, "learning_rate": 1.998880665647988e-05, "loss": 1.0927, "step": 1341 }, { "epoch": 0.04, "learning_rate": 1.998875561323779e-05, "loss": 1.0635, "step": 1342 }, { "epoch": 0.04, "learning_rate": 1.9988704453943552e-05, "loss": 1.0592, "step": 1343 }, { "epoch": 0.04, "learning_rate": 1.998865317859775e-05, "loss": 1.0438, "step": 1344 }, { "epoch": 0.04, "learning_rate": 1.9988601787200982e-05, "loss": 1.0814, "step": 1345 }, { "epoch": 0.04, "learning_rate": 1.9988550279753848e-05, "loss": 1.0979, "step": 1346 }, { "epoch": 0.04, "learning_rate": 1.9988498656256938e-05, "loss": 1.1022, "step": 1347 }, { "epoch": 0.04, "learning_rate": 1.9988446916710863e-05, "loss": 1.013, "step": 1348 }, { "epoch": 0.04, "learning_rate": 1.998839506111622e-05, "loss": 1.0838, "step": 1349 }, { "epoch": 0.04, "learning_rate": 1.998834308947361e-05, "loss": 1.1023, "step": 1350 }, { "epoch": 0.04, "learning_rate": 1.9988291001783633e-05, "loss": 1.0884, "step": 1351 }, { "epoch": 0.04, "learning_rate": 1.99882387980469e-05, "loss": 1.0502, "step": 1352 }, { "epoch": 0.05, "learning_rate": 1.998818647826402e-05, "loss": 1.0522, "step": 1353 }, { "epoch": 0.05, "learning_rate": 1.998813404243559e-05, "loss": 1.0548, "step": 1354 }, { "epoch": 0.05, "learning_rate": 1.9988081490562232e-05, "loss": 1.1002, "step": 1355 }, { "epoch": 0.05, "learning_rate": 1.998802882264455e-05, "loss": 1.0846, "step": 1356 }, { "epoch": 0.05, "learning_rate": 1.998797603868315e-05, "loss": 1.0803, "step": 1357 }, { "epoch": 0.05, "learning_rate": 1.998792313867866e-05, "loss": 1.0848, "step": 1358 }, { "epoch": 0.05, "learning_rate": 1.9987870122631682e-05, "loss": 1.0774, "step": 1359 }, { "epoch": 0.05, "learning_rate": 1.998781699054284e-05, "loss": 1.0308, "step": 1360 }, { "epoch": 0.05, "learning_rate": 1.9987763742412744e-05, "loss": 1.081, "step": 1361 }, { "epoch": 0.05, "learning_rate": 1.9987710378242016e-05, "loss": 1.0948, "step": 1362 }, { "epoch": 0.05, "learning_rate": 1.9987656898031278e-05, "loss": 1.0524, "step": 1363 }, { "epoch": 0.05, "learning_rate": 1.998760330178115e-05, "loss": 1.0913, "step": 1364 }, { "epoch": 0.05, "learning_rate": 1.9987549589492253e-05, "loss": 1.0833, "step": 1365 }, { "epoch": 0.05, "learning_rate": 1.998749576116521e-05, "loss": 1.0898, "step": 1366 }, { "epoch": 0.05, "learning_rate": 1.9987441816800654e-05, "loss": 1.0665, "step": 1367 }, { "epoch": 0.05, "learning_rate": 1.9987387756399204e-05, "loss": 1.0741, "step": 1368 }, { "epoch": 0.05, "learning_rate": 1.9987333579961488e-05, "loss": 1.066, "step": 1369 }, { "epoch": 0.05, "learning_rate": 1.9987279287488143e-05, "loss": 1.0291, "step": 1370 }, { "epoch": 0.05, "learning_rate": 1.998722487897979e-05, "loss": 1.1238, "step": 1371 }, { "epoch": 0.05, "learning_rate": 1.998717035443707e-05, "loss": 1.01, "step": 1372 }, { "epoch": 0.05, "learning_rate": 1.998711571386061e-05, "loss": 1.0381, "step": 1373 }, { "epoch": 0.05, "learning_rate": 1.9987060957251047e-05, "loss": 1.0752, "step": 1374 }, { "epoch": 0.05, "learning_rate": 1.9987006084609017e-05, "loss": 1.075, "step": 1375 }, { "epoch": 0.05, "learning_rate": 1.9986951095935158e-05, "loss": 1.0407, "step": 1376 }, { "epoch": 0.05, "learning_rate": 1.998689599123011e-05, "loss": 1.0961, "step": 1377 }, { "epoch": 0.05, "learning_rate": 1.9986840770494514e-05, "loss": 1.0366, "step": 1378 }, { "epoch": 0.05, "learning_rate": 1.9986785433729006e-05, "loss": 1.1065, "step": 1379 }, { "epoch": 0.05, "learning_rate": 1.9986729980934235e-05, "loss": 1.0455, "step": 1380 }, { "epoch": 0.05, "learning_rate": 1.998667441211084e-05, "loss": 1.0567, "step": 1381 }, { "epoch": 0.05, "learning_rate": 1.998661872725947e-05, "loss": 1.0627, "step": 1382 }, { "epoch": 0.05, "learning_rate": 1.9986562926380773e-05, "loss": 1.1105, "step": 1383 }, { "epoch": 0.05, "learning_rate": 1.9986507009475397e-05, "loss": 1.0171, "step": 1384 }, { "epoch": 0.05, "learning_rate": 1.9986450976543986e-05, "loss": 1.0827, "step": 1385 }, { "epoch": 0.05, "learning_rate": 1.99863948275872e-05, "loss": 1.061, "step": 1386 }, { "epoch": 0.05, "learning_rate": 1.9986338562605686e-05, "loss": 1.081, "step": 1387 }, { "epoch": 0.05, "learning_rate": 1.9986282181600092e-05, "loss": 1.0513, "step": 1388 }, { "epoch": 0.05, "learning_rate": 1.9986225684571086e-05, "loss": 1.0804, "step": 1389 }, { "epoch": 0.05, "learning_rate": 1.9986169071519315e-05, "loss": 1.0996, "step": 1390 }, { "epoch": 0.05, "learning_rate": 1.9986112342445444e-05, "loss": 1.0435, "step": 1391 }, { "epoch": 0.05, "learning_rate": 1.998605549735012e-05, "loss": 1.0834, "step": 1392 }, { "epoch": 0.05, "learning_rate": 1.998599853623402e-05, "loss": 1.0443, "step": 1393 }, { "epoch": 0.05, "learning_rate": 1.9985941459097793e-05, "loss": 1.094, "step": 1394 }, { "epoch": 0.05, "learning_rate": 1.9985884265942105e-05, "loss": 1.0482, "step": 1395 }, { "epoch": 0.05, "learning_rate": 1.998582695676762e-05, "loss": 1.057, "step": 1396 }, { "epoch": 0.05, "learning_rate": 1.998576953157501e-05, "loss": 1.0891, "step": 1397 }, { "epoch": 0.05, "learning_rate": 1.9985711990364933e-05, "loss": 1.085, "step": 1398 }, { "epoch": 0.05, "learning_rate": 1.9985654333138063e-05, "loss": 1.0892, "step": 1399 }, { "epoch": 0.05, "learning_rate": 1.998559655989507e-05, "loss": 1.0799, "step": 1400 }, { "epoch": 0.05, "learning_rate": 1.9985538670636626e-05, "loss": 1.0674, "step": 1401 }, { "epoch": 0.05, "learning_rate": 1.99854806653634e-05, "loss": 1.0609, "step": 1402 }, { "epoch": 0.05, "learning_rate": 1.9985422544076066e-05, "loss": 1.0441, "step": 1403 }, { "epoch": 0.05, "learning_rate": 1.9985364306775305e-05, "loss": 1.0788, "step": 1404 }, { "epoch": 0.05, "learning_rate": 1.9985305953461784e-05, "loss": 1.0932, "step": 1405 }, { "epoch": 0.05, "learning_rate": 1.998524748413619e-05, "loss": 1.0668, "step": 1406 }, { "epoch": 0.05, "learning_rate": 1.99851888987992e-05, "loss": 1.0593, "step": 1407 }, { "epoch": 0.05, "learning_rate": 1.9985130197451493e-05, "loss": 1.0565, "step": 1408 }, { "epoch": 0.05, "learning_rate": 1.9985071380093746e-05, "loss": 1.0946, "step": 1409 }, { "epoch": 0.05, "learning_rate": 1.9985012446726656e-05, "loss": 1.0894, "step": 1410 }, { "epoch": 0.05, "learning_rate": 1.9984953397350895e-05, "loss": 1.0415, "step": 1411 }, { "epoch": 0.05, "learning_rate": 1.9984894231967155e-05, "loss": 1.0791, "step": 1412 }, { "epoch": 0.05, "learning_rate": 1.9984834950576123e-05, "loss": 1.1196, "step": 1413 }, { "epoch": 0.05, "learning_rate": 1.9984775553178485e-05, "loss": 1.034, "step": 1414 }, { "epoch": 0.05, "learning_rate": 1.998471603977493e-05, "loss": 1.0978, "step": 1415 }, { "epoch": 0.05, "learning_rate": 1.9984656410366157e-05, "loss": 1.0318, "step": 1416 }, { "epoch": 0.05, "learning_rate": 1.9984596664952852e-05, "loss": 1.0908, "step": 1417 }, { "epoch": 0.05, "learning_rate": 1.998453680353571e-05, "loss": 1.0931, "step": 1418 }, { "epoch": 0.05, "learning_rate": 1.998447682611543e-05, "loss": 1.0872, "step": 1419 }, { "epoch": 0.05, "learning_rate": 1.998441673269271e-05, "loss": 1.0427, "step": 1420 }, { "epoch": 0.05, "learning_rate": 1.998435652326824e-05, "loss": 1.0893, "step": 1421 }, { "epoch": 0.05, "learning_rate": 1.9984296197842722e-05, "loss": 1.0993, "step": 1422 }, { "epoch": 0.05, "learning_rate": 1.9984235756416864e-05, "loss": 1.0819, "step": 1423 }, { "epoch": 0.05, "learning_rate": 1.9984175198991362e-05, "loss": 1.0442, "step": 1424 }, { "epoch": 0.05, "learning_rate": 1.998411452556692e-05, "loss": 1.0918, "step": 1425 }, { "epoch": 0.05, "learning_rate": 1.9984053736144244e-05, "loss": 1.1116, "step": 1426 }, { "epoch": 0.05, "learning_rate": 1.998399283072404e-05, "loss": 1.0893, "step": 1427 }, { "epoch": 0.05, "learning_rate": 1.9983931809307016e-05, "loss": 1.0649, "step": 1428 }, { "epoch": 0.05, "learning_rate": 1.9983870671893884e-05, "loss": 1.1258, "step": 1429 }, { "epoch": 0.05, "learning_rate": 1.998380941848535e-05, "loss": 1.1357, "step": 1430 }, { "epoch": 0.05, "learning_rate": 1.9983748049082125e-05, "loss": 1.083, "step": 1431 }, { "epoch": 0.05, "learning_rate": 1.998368656368492e-05, "loss": 1.0307, "step": 1432 }, { "epoch": 0.05, "learning_rate": 1.998362496229446e-05, "loss": 1.0679, "step": 1433 }, { "epoch": 0.05, "learning_rate": 1.9983563244911447e-05, "loss": 1.0887, "step": 1434 }, { "epoch": 0.05, "learning_rate": 1.9983501411536608e-05, "loss": 1.0531, "step": 1435 }, { "epoch": 0.05, "learning_rate": 1.998343946217066e-05, "loss": 1.0494, "step": 1436 }, { "epoch": 0.05, "learning_rate": 1.998337739681432e-05, "loss": 1.0736, "step": 1437 }, { "epoch": 0.05, "learning_rate": 1.998331521546831e-05, "loss": 1.062, "step": 1438 }, { "epoch": 0.05, "learning_rate": 1.9983252918133348e-05, "loss": 1.0586, "step": 1439 }, { "epoch": 0.05, "learning_rate": 1.9983190504810167e-05, "loss": 1.008, "step": 1440 }, { "epoch": 0.05, "learning_rate": 1.998312797549949e-05, "loss": 1.0859, "step": 1441 }, { "epoch": 0.05, "learning_rate": 1.9983065330202035e-05, "loss": 1.0702, "step": 1442 }, { "epoch": 0.05, "learning_rate": 1.9983002568918536e-05, "loss": 1.0922, "step": 1443 }, { "epoch": 0.05, "learning_rate": 1.9982939691649725e-05, "loss": 1.0406, "step": 1444 }, { "epoch": 0.05, "learning_rate": 1.998287669839633e-05, "loss": 1.0713, "step": 1445 }, { "epoch": 0.05, "learning_rate": 1.998281358915908e-05, "loss": 1.0745, "step": 1446 }, { "epoch": 0.05, "learning_rate": 1.998275036393871e-05, "loss": 1.0606, "step": 1447 }, { "epoch": 0.05, "learning_rate": 1.998268702273596e-05, "loss": 1.0266, "step": 1448 }, { "epoch": 0.05, "learning_rate": 1.9982623565551553e-05, "loss": 1.0679, "step": 1449 }, { "epoch": 0.05, "learning_rate": 1.998255999238624e-05, "loss": 1.0683, "step": 1450 }, { "epoch": 0.05, "learning_rate": 1.9982496303240753e-05, "loss": 1.0842, "step": 1451 }, { "epoch": 0.05, "learning_rate": 1.9982432498115832e-05, "loss": 1.0418, "step": 1452 }, { "epoch": 0.05, "learning_rate": 1.998236857701222e-05, "loss": 1.0819, "step": 1453 }, { "epoch": 0.05, "learning_rate": 1.9982304539930657e-05, "loss": 1.1131, "step": 1454 }, { "epoch": 0.05, "learning_rate": 1.9982240386871893e-05, "loss": 1.0288, "step": 1455 }, { "epoch": 0.05, "learning_rate": 1.9982176117836663e-05, "loss": 1.0463, "step": 1456 }, { "epoch": 0.05, "learning_rate": 1.9982111732825723e-05, "loss": 1.1356, "step": 1457 }, { "epoch": 0.05, "learning_rate": 1.9982047231839817e-05, "loss": 1.0645, "step": 1458 }, { "epoch": 0.05, "learning_rate": 1.99819826148797e-05, "loss": 1.1031, "step": 1459 }, { "epoch": 0.05, "learning_rate": 1.9981917881946113e-05, "loss": 1.0563, "step": 1460 }, { "epoch": 0.05, "learning_rate": 1.998185303303981e-05, "loss": 1.0989, "step": 1461 }, { "epoch": 0.05, "learning_rate": 1.9981788068161557e-05, "loss": 1.0741, "step": 1462 }, { "epoch": 0.05, "learning_rate": 1.9981722987312092e-05, "loss": 1.112, "step": 1463 }, { "epoch": 0.05, "learning_rate": 1.9981657790492178e-05, "loss": 0.9983, "step": 1464 }, { "epoch": 0.05, "learning_rate": 1.9981592477702577e-05, "loss": 1.0328, "step": 1465 }, { "epoch": 0.05, "learning_rate": 1.9981527048944043e-05, "loss": 1.0616, "step": 1466 }, { "epoch": 0.05, "learning_rate": 1.9981461504217336e-05, "loss": 1.0857, "step": 1467 }, { "epoch": 0.05, "learning_rate": 1.998139584352322e-05, "loss": 1.0408, "step": 1468 }, { "epoch": 0.05, "learning_rate": 1.9981330066862453e-05, "loss": 1.0794, "step": 1469 }, { "epoch": 0.05, "learning_rate": 1.9981264174235805e-05, "loss": 1.0921, "step": 1470 }, { "epoch": 0.05, "learning_rate": 1.9981198165644038e-05, "loss": 1.0987, "step": 1471 }, { "epoch": 0.05, "learning_rate": 1.998113204108792e-05, "loss": 1.0378, "step": 1472 }, { "epoch": 0.05, "learning_rate": 1.9981065800568218e-05, "loss": 1.0854, "step": 1473 }, { "epoch": 0.05, "learning_rate": 1.9980999444085704e-05, "loss": 1.0938, "step": 1474 }, { "epoch": 0.05, "learning_rate": 1.9980932971641148e-05, "loss": 1.1216, "step": 1475 }, { "epoch": 0.05, "learning_rate": 1.998086638323532e-05, "loss": 1.0893, "step": 1476 }, { "epoch": 0.05, "learning_rate": 1.9980799678869e-05, "loss": 1.0927, "step": 1477 }, { "epoch": 0.05, "learning_rate": 1.9980732858542957e-05, "loss": 1.0713, "step": 1478 }, { "epoch": 0.05, "learning_rate": 1.9980665922257968e-05, "loss": 1.0931, "step": 1479 }, { "epoch": 0.05, "learning_rate": 1.9980598870014812e-05, "loss": 1.033, "step": 1480 }, { "epoch": 0.05, "learning_rate": 1.998053170181427e-05, "loss": 1.0689, "step": 1481 }, { "epoch": 0.05, "learning_rate": 1.9980464417657118e-05, "loss": 1.0957, "step": 1482 }, { "epoch": 0.05, "learning_rate": 1.9980397017544137e-05, "loss": 1.1088, "step": 1483 }, { "epoch": 0.05, "learning_rate": 1.9980329501476117e-05, "loss": 1.0477, "step": 1484 }, { "epoch": 0.05, "learning_rate": 1.9980261869453837e-05, "loss": 1.1036, "step": 1485 }, { "epoch": 0.05, "learning_rate": 1.9980194121478083e-05, "loss": 1.0634, "step": 1486 }, { "epoch": 0.05, "learning_rate": 1.9980126257549644e-05, "loss": 1.0974, "step": 1487 }, { "epoch": 0.05, "learning_rate": 1.998005827766931e-05, "loss": 1.0619, "step": 1488 }, { "epoch": 0.05, "learning_rate": 1.9979990181837866e-05, "loss": 1.0756, "step": 1489 }, { "epoch": 0.05, "learning_rate": 1.9979921970056106e-05, "loss": 1.0981, "step": 1490 }, { "epoch": 0.05, "learning_rate": 1.9979853642324823e-05, "loss": 1.0643, "step": 1491 }, { "epoch": 0.05, "learning_rate": 1.997978519864481e-05, "loss": 1.046, "step": 1492 }, { "epoch": 0.05, "learning_rate": 1.997971663901686e-05, "loss": 1.0595, "step": 1493 }, { "epoch": 0.05, "learning_rate": 1.9979647963441776e-05, "loss": 1.0951, "step": 1494 }, { "epoch": 0.05, "learning_rate": 1.9979579171920346e-05, "loss": 1.0656, "step": 1495 }, { "epoch": 0.05, "learning_rate": 1.9979510264453378e-05, "loss": 1.0238, "step": 1496 }, { "epoch": 0.05, "learning_rate": 1.997944124104167e-05, "loss": 1.085, "step": 1497 }, { "epoch": 0.05, "learning_rate": 1.9979372101686025e-05, "loss": 1.0944, "step": 1498 }, { "epoch": 0.05, "learning_rate": 1.997930284638724e-05, "loss": 1.0628, "step": 1499 }, { "epoch": 0.05, "learning_rate": 1.9979233475146127e-05, "loss": 1.0822, "step": 1500 }, { "epoch": 0.05, "learning_rate": 1.997916398796349e-05, "loss": 1.0986, "step": 1501 }, { "epoch": 0.05, "learning_rate": 1.9979094384840133e-05, "loss": 1.0665, "step": 1502 }, { "epoch": 0.05, "learning_rate": 1.997902466577687e-05, "loss": 1.1054, "step": 1503 }, { "epoch": 0.05, "learning_rate": 1.9978954830774506e-05, "loss": 1.0189, "step": 1504 }, { "epoch": 0.05, "learning_rate": 1.9978884879833857e-05, "loss": 1.0979, "step": 1505 }, { "epoch": 0.05, "learning_rate": 1.9978814812955732e-05, "loss": 1.1235, "step": 1506 }, { "epoch": 0.05, "learning_rate": 1.9978744630140948e-05, "loss": 1.0667, "step": 1507 }, { "epoch": 0.05, "learning_rate": 1.9978674331390315e-05, "loss": 1.0504, "step": 1508 }, { "epoch": 0.05, "learning_rate": 1.9978603916704657e-05, "loss": 1.0898, "step": 1509 }, { "epoch": 0.05, "learning_rate": 1.997853338608479e-05, "loss": 1.1061, "step": 1510 }, { "epoch": 0.05, "learning_rate": 1.9978462739531528e-05, "loss": 1.0674, "step": 1511 }, { "epoch": 0.05, "learning_rate": 1.9978391977045698e-05, "loss": 1.0267, "step": 1512 }, { "epoch": 0.05, "learning_rate": 1.997832109862812e-05, "loss": 1.0897, "step": 1513 }, { "epoch": 0.05, "learning_rate": 1.9978250104279617e-05, "loss": 1.0622, "step": 1514 }, { "epoch": 0.05, "learning_rate": 1.997817899400101e-05, "loss": 1.0845, "step": 1515 }, { "epoch": 0.05, "learning_rate": 1.9978107767793137e-05, "loss": 1.0779, "step": 1516 }, { "epoch": 0.05, "learning_rate": 1.9978036425656813e-05, "loss": 1.0426, "step": 1517 }, { "epoch": 0.05, "learning_rate": 1.9977964967592877e-05, "loss": 1.0647, "step": 1518 }, { "epoch": 0.05, "learning_rate": 1.997789339360215e-05, "loss": 1.0676, "step": 1519 }, { "epoch": 0.05, "learning_rate": 1.997782170368547e-05, "loss": 1.0496, "step": 1520 }, { "epoch": 0.05, "learning_rate": 1.9977749897843666e-05, "loss": 1.0949, "step": 1521 }, { "epoch": 0.05, "learning_rate": 1.9977677976077574e-05, "loss": 1.0238, "step": 1522 }, { "epoch": 0.05, "learning_rate": 1.9977605938388033e-05, "loss": 1.1, "step": 1523 }, { "epoch": 0.05, "learning_rate": 1.997753378477587e-05, "loss": 1.0461, "step": 1524 }, { "epoch": 0.05, "learning_rate": 1.9977461515241932e-05, "loss": 1.0494, "step": 1525 }, { "epoch": 0.05, "learning_rate": 1.997738912978706e-05, "loss": 1.0644, "step": 1526 }, { "epoch": 0.05, "learning_rate": 1.997731662841209e-05, "loss": 1.052, "step": 1527 }, { "epoch": 0.05, "learning_rate": 1.9977244011117866e-05, "loss": 1.0316, "step": 1528 }, { "epoch": 0.05, "learning_rate": 1.9977171277905228e-05, "loss": 1.0134, "step": 1529 }, { "epoch": 0.05, "learning_rate": 1.9977098428775027e-05, "loss": 1.0274, "step": 1530 }, { "epoch": 0.05, "learning_rate": 1.9977025463728106e-05, "loss": 1.057, "step": 1531 }, { "epoch": 0.05, "learning_rate": 1.9976952382765313e-05, "loss": 1.0455, "step": 1532 }, { "epoch": 0.05, "learning_rate": 1.99768791858875e-05, "loss": 1.0883, "step": 1533 }, { "epoch": 0.05, "learning_rate": 1.9976805873095513e-05, "loss": 1.0653, "step": 1534 }, { "epoch": 0.05, "learning_rate": 1.9976732444390207e-05, "loss": 1.0961, "step": 1535 }, { "epoch": 0.05, "learning_rate": 1.9976658899772434e-05, "loss": 1.0709, "step": 1536 }, { "epoch": 0.05, "learning_rate": 1.997658523924305e-05, "loss": 1.093, "step": 1537 }, { "epoch": 0.05, "learning_rate": 1.9976511462802906e-05, "loss": 1.0535, "step": 1538 }, { "epoch": 0.05, "learning_rate": 1.9976437570452863e-05, "loss": 1.0594, "step": 1539 }, { "epoch": 0.05, "learning_rate": 1.9976363562193778e-05, "loss": 1.0338, "step": 1540 }, { "epoch": 0.05, "learning_rate": 1.9976289438026514e-05, "loss": 1.065, "step": 1541 }, { "epoch": 0.05, "learning_rate": 1.997621519795193e-05, "loss": 1.0761, "step": 1542 }, { "epoch": 0.05, "learning_rate": 1.9976140841970885e-05, "loss": 1.0907, "step": 1543 }, { "epoch": 0.05, "learning_rate": 1.997606637008425e-05, "loss": 1.0573, "step": 1544 }, { "epoch": 0.05, "learning_rate": 1.9975991782292886e-05, "loss": 1.0818, "step": 1545 }, { "epoch": 0.05, "learning_rate": 1.997591707859766e-05, "loss": 1.0605, "step": 1546 }, { "epoch": 0.05, "learning_rate": 1.997584225899944e-05, "loss": 1.0672, "step": 1547 }, { "epoch": 0.05, "learning_rate": 1.9975767323499097e-05, "loss": 1.0378, "step": 1548 }, { "epoch": 0.05, "learning_rate": 1.9975692272097497e-05, "loss": 1.0521, "step": 1549 }, { "epoch": 0.05, "learning_rate": 1.9975617104795515e-05, "loss": 1.0339, "step": 1550 }, { "epoch": 0.05, "learning_rate": 1.9975541821594028e-05, "loss": 1.0589, "step": 1551 }, { "epoch": 0.05, "learning_rate": 1.9975466422493904e-05, "loss": 1.0273, "step": 1552 }, { "epoch": 0.05, "learning_rate": 1.997539090749602e-05, "loss": 1.0529, "step": 1553 }, { "epoch": 0.05, "learning_rate": 1.9975315276601258e-05, "loss": 1.0347, "step": 1554 }, { "epoch": 0.05, "learning_rate": 1.9975239529810494e-05, "loss": 1.0377, "step": 1555 }, { "epoch": 0.05, "learning_rate": 1.997516366712461e-05, "loss": 1.0414, "step": 1556 }, { "epoch": 0.05, "learning_rate": 1.9975087688544483e-05, "loss": 1.1072, "step": 1557 }, { "epoch": 0.05, "learning_rate": 1.9975011594070997e-05, "loss": 1.0499, "step": 1558 }, { "epoch": 0.05, "learning_rate": 1.9974935383705037e-05, "loss": 1.0307, "step": 1559 }, { "epoch": 0.05, "learning_rate": 1.9974859057447494e-05, "loss": 1.0712, "step": 1560 }, { "epoch": 0.05, "learning_rate": 1.9974782615299246e-05, "loss": 1.08, "step": 1561 }, { "epoch": 0.05, "learning_rate": 1.9974706057261182e-05, "loss": 1.0215, "step": 1562 }, { "epoch": 0.05, "learning_rate": 1.9974629383334196e-05, "loss": 1.0533, "step": 1563 }, { "epoch": 0.05, "learning_rate": 1.997455259351918e-05, "loss": 1.0341, "step": 1564 }, { "epoch": 0.05, "learning_rate": 1.997447568781702e-05, "loss": 1.0708, "step": 1565 }, { "epoch": 0.05, "learning_rate": 1.9974398666228612e-05, "loss": 1.0976, "step": 1566 }, { "epoch": 0.05, "learning_rate": 1.9974321528754853e-05, "loss": 1.0728, "step": 1567 }, { "epoch": 0.05, "learning_rate": 1.997424427539664e-05, "loss": 1.016, "step": 1568 }, { "epoch": 0.05, "learning_rate": 1.997416690615486e-05, "loss": 1.0693, "step": 1569 }, { "epoch": 0.05, "learning_rate": 1.9974089421030428e-05, "loss": 1.0662, "step": 1570 }, { "epoch": 0.05, "learning_rate": 1.9974011820024233e-05, "loss": 1.0877, "step": 1571 }, { "epoch": 0.05, "learning_rate": 1.997393410313718e-05, "loss": 1.0459, "step": 1572 }, { "epoch": 0.05, "learning_rate": 1.9973856270370174e-05, "loss": 1.0838, "step": 1573 }, { "epoch": 0.05, "learning_rate": 1.9973778321724115e-05, "loss": 1.0847, "step": 1574 }, { "epoch": 0.05, "learning_rate": 1.9973700257199913e-05, "loss": 1.0657, "step": 1575 }, { "epoch": 0.05, "learning_rate": 1.997362207679847e-05, "loss": 1.0377, "step": 1576 }, { "epoch": 0.05, "learning_rate": 1.99735437805207e-05, "loss": 1.0884, "step": 1577 }, { "epoch": 0.05, "learning_rate": 1.9973465368367506e-05, "loss": 1.0418, "step": 1578 }, { "epoch": 0.05, "learning_rate": 1.9973386840339807e-05, "loss": 1.0368, "step": 1579 }, { "epoch": 0.05, "learning_rate": 1.9973308196438507e-05, "loss": 1.0447, "step": 1580 }, { "epoch": 0.05, "learning_rate": 1.997322943666453e-05, "loss": 1.0538, "step": 1581 }, { "epoch": 0.05, "learning_rate": 1.9973150561018778e-05, "loss": 1.0781, "step": 1582 }, { "epoch": 0.05, "learning_rate": 1.9973071569502176e-05, "loss": 1.0791, "step": 1583 }, { "epoch": 0.05, "learning_rate": 1.997299246211564e-05, "loss": 1.0358, "step": 1584 }, { "epoch": 0.05, "learning_rate": 1.997291323886009e-05, "loss": 1.1033, "step": 1585 }, { "epoch": 0.05, "learning_rate": 1.9972833899736445e-05, "loss": 1.0623, "step": 1586 }, { "epoch": 0.05, "learning_rate": 1.997275444474563e-05, "loss": 1.0584, "step": 1587 }, { "epoch": 0.05, "learning_rate": 1.9972674873888566e-05, "loss": 1.0283, "step": 1588 }, { "epoch": 0.05, "learning_rate": 1.9972595187166172e-05, "loss": 1.0553, "step": 1589 }, { "epoch": 0.05, "learning_rate": 1.9972515384579382e-05, "loss": 1.069, "step": 1590 }, { "epoch": 0.05, "learning_rate": 1.9972435466129117e-05, "loss": 1.0947, "step": 1591 }, { "epoch": 0.05, "learning_rate": 1.997235543181631e-05, "loss": 1.0253, "step": 1592 }, { "epoch": 0.05, "learning_rate": 1.997227528164189e-05, "loss": 1.0425, "step": 1593 }, { "epoch": 0.05, "learning_rate": 1.9972195015606787e-05, "loss": 1.0803, "step": 1594 }, { "epoch": 0.05, "learning_rate": 1.9972114633711938e-05, "loss": 1.0366, "step": 1595 }, { "epoch": 0.05, "learning_rate": 1.997203413595827e-05, "loss": 1.0311, "step": 1596 }, { "epoch": 0.05, "learning_rate": 1.997195352234672e-05, "loss": 1.0707, "step": 1597 }, { "epoch": 0.05, "learning_rate": 1.9971872792878226e-05, "loss": 1.0681, "step": 1598 }, { "epoch": 0.05, "learning_rate": 1.9971791947553725e-05, "loss": 1.0847, "step": 1599 }, { "epoch": 0.05, "learning_rate": 1.997171098637416e-05, "loss": 1.0514, "step": 1600 }, { "epoch": 0.05, "learning_rate": 1.9971629909340464e-05, "loss": 1.0774, "step": 1601 }, { "epoch": 0.05, "learning_rate": 1.9971548716453587e-05, "loss": 1.0977, "step": 1602 }, { "epoch": 0.05, "learning_rate": 1.9971467407714467e-05, "loss": 1.1023, "step": 1603 }, { "epoch": 0.05, "learning_rate": 1.9971385983124053e-05, "loss": 1.0371, "step": 1604 }, { "epoch": 0.05, "learning_rate": 1.9971304442683288e-05, "loss": 1.0723, "step": 1605 }, { "epoch": 0.05, "learning_rate": 1.997122278639312e-05, "loss": 1.0407, "step": 1606 }, { "epoch": 0.05, "learning_rate": 1.99711410142545e-05, "loss": 1.093, "step": 1607 }, { "epoch": 0.05, "learning_rate": 1.997105912626837e-05, "loss": 1.0585, "step": 1608 }, { "epoch": 0.05, "learning_rate": 1.997097712243569e-05, "loss": 1.0357, "step": 1609 }, { "epoch": 0.05, "learning_rate": 1.9970895002757413e-05, "loss": 1.0252, "step": 1610 }, { "epoch": 0.05, "learning_rate": 1.9970812767234483e-05, "loss": 1.0516, "step": 1611 }, { "epoch": 0.05, "learning_rate": 1.9970730415867868e-05, "loss": 1.0221, "step": 1612 }, { "epoch": 0.05, "learning_rate": 1.997064794865852e-05, "loss": 1.0886, "step": 1613 }, { "epoch": 0.05, "learning_rate": 1.997056536560739e-05, "loss": 1.0259, "step": 1614 }, { "epoch": 0.05, "learning_rate": 1.997048266671545e-05, "loss": 1.0727, "step": 1615 }, { "epoch": 0.05, "learning_rate": 1.9970399851983653e-05, "loss": 1.054, "step": 1616 }, { "epoch": 0.05, "learning_rate": 1.997031692141296e-05, "loss": 1.0601, "step": 1617 }, { "epoch": 0.05, "learning_rate": 1.9970233875004344e-05, "loss": 1.0443, "step": 1618 }, { "epoch": 0.05, "learning_rate": 1.9970150712758757e-05, "loss": 1.0318, "step": 1619 }, { "epoch": 0.05, "learning_rate": 1.9970067434677173e-05, "loss": 1.0847, "step": 1620 }, { "epoch": 0.05, "learning_rate": 1.996998404076056e-05, "loss": 1.0644, "step": 1621 }, { "epoch": 0.05, "learning_rate": 1.996990053100988e-05, "loss": 1.0582, "step": 1622 }, { "epoch": 0.05, "learning_rate": 1.996981690542611e-05, "loss": 1.07, "step": 1623 }, { "epoch": 0.05, "learning_rate": 1.9969733164010224e-05, "loss": 1.0495, "step": 1624 }, { "epoch": 0.05, "learning_rate": 1.9969649306763187e-05, "loss": 1.0989, "step": 1625 }, { "epoch": 0.05, "learning_rate": 1.9969565333685977e-05, "loss": 1.0793, "step": 1626 }, { "epoch": 0.05, "learning_rate": 1.9969481244779566e-05, "loss": 1.0374, "step": 1627 }, { "epoch": 0.05, "learning_rate": 1.996939704004494e-05, "loss": 1.0583, "step": 1628 }, { "epoch": 0.05, "learning_rate": 1.996931271948307e-05, "loss": 1.0645, "step": 1629 }, { "epoch": 0.05, "learning_rate": 1.9969228283094937e-05, "loss": 1.0739, "step": 1630 }, { "epoch": 0.05, "learning_rate": 1.996914373088152e-05, "loss": 1.0548, "step": 1631 }, { "epoch": 0.05, "learning_rate": 1.9969059062843806e-05, "loss": 1.0445, "step": 1632 }, { "epoch": 0.05, "learning_rate": 1.9968974278982777e-05, "loss": 1.0765, "step": 1633 }, { "epoch": 0.05, "learning_rate": 1.996888937929942e-05, "loss": 1.0726, "step": 1634 }, { "epoch": 0.05, "learning_rate": 1.9968804363794713e-05, "loss": 1.0631, "step": 1635 }, { "epoch": 0.05, "learning_rate": 1.9968719232469652e-05, "loss": 1.0199, "step": 1636 }, { "epoch": 0.05, "learning_rate": 1.9968633985325226e-05, "loss": 1.0508, "step": 1637 }, { "epoch": 0.05, "learning_rate": 1.996854862236242e-05, "loss": 1.0476, "step": 1638 }, { "epoch": 0.05, "learning_rate": 1.996846314358223e-05, "loss": 1.028, "step": 1639 }, { "epoch": 0.05, "learning_rate": 1.996837754898565e-05, "loss": 1.0197, "step": 1640 }, { "epoch": 0.05, "learning_rate": 1.9968291838573672e-05, "loss": 1.0486, "step": 1641 }, { "epoch": 0.05, "learning_rate": 1.996820601234729e-05, "loss": 1.0909, "step": 1642 }, { "epoch": 0.05, "learning_rate": 1.9968120070307503e-05, "loss": 1.0367, "step": 1643 }, { "epoch": 0.05, "learning_rate": 1.996803401245531e-05, "loss": 1.0757, "step": 1644 }, { "epoch": 0.05, "learning_rate": 1.9967947838791714e-05, "loss": 1.082, "step": 1645 }, { "epoch": 0.05, "learning_rate": 1.996786154931771e-05, "loss": 1.0383, "step": 1646 }, { "epoch": 0.05, "learning_rate": 1.9967775144034304e-05, "loss": 1.0539, "step": 1647 }, { "epoch": 0.05, "learning_rate": 1.99676886229425e-05, "loss": 1.0359, "step": 1648 }, { "epoch": 0.05, "learning_rate": 1.9967601986043303e-05, "loss": 1.055, "step": 1649 }, { "epoch": 0.05, "learning_rate": 1.996751523333772e-05, "loss": 1.0267, "step": 1650 }, { "epoch": 0.05, "learning_rate": 1.9967428364826755e-05, "loss": 1.066, "step": 1651 }, { "epoch": 0.05, "learning_rate": 1.996734138051142e-05, "loss": 1.052, "step": 1652 }, { "epoch": 0.06, "learning_rate": 1.9967254280392732e-05, "loss": 1.0895, "step": 1653 }, { "epoch": 0.06, "learning_rate": 1.996716706447169e-05, "loss": 1.0704, "step": 1654 }, { "epoch": 0.06, "learning_rate": 1.9967079732749317e-05, "loss": 1.0319, "step": 1655 }, { "epoch": 0.06, "learning_rate": 1.9966992285226627e-05, "loss": 1.0309, "step": 1656 }, { "epoch": 0.06, "learning_rate": 1.996690472190463e-05, "loss": 1.0422, "step": 1657 }, { "epoch": 0.06, "learning_rate": 1.996681704278435e-05, "loss": 1.0692, "step": 1658 }, { "epoch": 0.06, "learning_rate": 1.99667292478668e-05, "loss": 1.055, "step": 1659 }, { "epoch": 0.06, "learning_rate": 1.9966641337153006e-05, "loss": 0.9863, "step": 1660 }, { "epoch": 0.06, "learning_rate": 1.9966553310643985e-05, "loss": 1.0579, "step": 1661 }, { "epoch": 0.06, "learning_rate": 1.996646516834076e-05, "loss": 1.063, "step": 1662 }, { "epoch": 0.06, "learning_rate": 1.996637691024436e-05, "loss": 1.0513, "step": 1663 }, { "epoch": 0.06, "learning_rate": 1.9966288536355803e-05, "loss": 1.0415, "step": 1664 }, { "epoch": 0.06, "learning_rate": 1.996620004667612e-05, "loss": 1.0912, "step": 1665 }, { "epoch": 0.06, "learning_rate": 1.996611144120634e-05, "loss": 1.0598, "step": 1666 }, { "epoch": 0.06, "learning_rate": 1.9966022719947483e-05, "loss": 1.0375, "step": 1667 }, { "epoch": 0.06, "learning_rate": 1.9965933882900596e-05, "loss": 1.0513, "step": 1668 }, { "epoch": 0.06, "learning_rate": 1.99658449300667e-05, "loss": 1.0803, "step": 1669 }, { "epoch": 0.06, "learning_rate": 1.996575586144683e-05, "loss": 1.0131, "step": 1670 }, { "epoch": 0.06, "learning_rate": 1.9965666677042025e-05, "loss": 1.0831, "step": 1671 }, { "epoch": 0.06, "learning_rate": 1.9965577376853315e-05, "loss": 1.0549, "step": 1672 }, { "epoch": 0.06, "learning_rate": 1.9965487960881743e-05, "loss": 1.0558, "step": 1673 }, { "epoch": 0.06, "learning_rate": 1.9965398429128343e-05, "loss": 1.0807, "step": 1674 }, { "epoch": 0.06, "learning_rate": 1.996530878159416e-05, "loss": 1.0837, "step": 1675 }, { "epoch": 0.06, "learning_rate": 1.996521901828023e-05, "loss": 1.0267, "step": 1676 }, { "epoch": 0.06, "learning_rate": 1.9965129139187604e-05, "loss": 1.0648, "step": 1677 }, { "epoch": 0.06, "learning_rate": 1.996503914431732e-05, "loss": 1.0705, "step": 1678 }, { "epoch": 0.06, "learning_rate": 1.9964949033670424e-05, "loss": 1.0377, "step": 1679 }, { "epoch": 0.06, "learning_rate": 1.9964858807247963e-05, "loss": 1.0793, "step": 1680 }, { "epoch": 0.06, "learning_rate": 1.9964768465050986e-05, "loss": 1.0826, "step": 1681 }, { "epoch": 0.06, "learning_rate": 1.9964678007080544e-05, "loss": 1.0409, "step": 1682 }, { "epoch": 0.06, "learning_rate": 1.9964587433337685e-05, "loss": 1.1011, "step": 1683 }, { "epoch": 0.06, "learning_rate": 1.9964496743823464e-05, "loss": 1.0384, "step": 1684 }, { "epoch": 0.06, "learning_rate": 1.9964405938538935e-05, "loss": 1.0669, "step": 1685 }, { "epoch": 0.06, "learning_rate": 1.996431501748515e-05, "loss": 1.0478, "step": 1686 }, { "epoch": 0.06, "learning_rate": 1.996422398066317e-05, "loss": 1.0379, "step": 1687 }, { "epoch": 0.06, "learning_rate": 1.9964132828074048e-05, "loss": 1.0774, "step": 1688 }, { "epoch": 0.06, "learning_rate": 1.9964041559718846e-05, "loss": 1.0504, "step": 1689 }, { "epoch": 0.06, "learning_rate": 1.996395017559862e-05, "loss": 1.0705, "step": 1690 }, { "epoch": 0.06, "learning_rate": 1.9963858675714437e-05, "loss": 1.0395, "step": 1691 }, { "epoch": 0.06, "learning_rate": 1.996376706006736e-05, "loss": 1.0146, "step": 1692 }, { "epoch": 0.06, "learning_rate": 1.9963675328658448e-05, "loss": 1.1027, "step": 1693 }, { "epoch": 0.06, "learning_rate": 1.9963583481488774e-05, "loss": 1.0545, "step": 1694 }, { "epoch": 0.06, "learning_rate": 1.9963491518559396e-05, "loss": 1.0559, "step": 1695 }, { "epoch": 0.06, "learning_rate": 1.9963399439871392e-05, "loss": 1.0303, "step": 1696 }, { "epoch": 0.06, "learning_rate": 1.9963307245425826e-05, "loss": 1.0577, "step": 1697 }, { "epoch": 0.06, "learning_rate": 1.996321493522377e-05, "loss": 1.0487, "step": 1698 }, { "epoch": 0.06, "learning_rate": 1.9963122509266297e-05, "loss": 1.0455, "step": 1699 }, { "epoch": 0.06, "learning_rate": 1.996302996755448e-05, "loss": 1.0194, "step": 1700 }, { "epoch": 0.06, "learning_rate": 1.99629373100894e-05, "loss": 1.0679, "step": 1701 }, { "epoch": 0.06, "learning_rate": 1.9962844536872127e-05, "loss": 1.0806, "step": 1702 }, { "epoch": 0.06, "learning_rate": 1.996275164790374e-05, "loss": 1.0572, "step": 1703 }, { "epoch": 0.06, "learning_rate": 1.9962658643185317e-05, "loss": 1.0205, "step": 1704 }, { "epoch": 0.06, "learning_rate": 1.9962565522717942e-05, "loss": 1.0559, "step": 1705 }, { "epoch": 0.06, "learning_rate": 1.996247228650269e-05, "loss": 1.0289, "step": 1706 }, { "epoch": 0.06, "learning_rate": 1.9962378934540657e-05, "loss": 1.0408, "step": 1707 }, { "epoch": 0.06, "learning_rate": 1.9962285466832914e-05, "loss": 0.9928, "step": 1708 }, { "epoch": 0.06, "learning_rate": 1.996219188338056e-05, "loss": 1.0832, "step": 1709 }, { "epoch": 0.06, "learning_rate": 1.996209818418467e-05, "loss": 1.0785, "step": 1710 }, { "epoch": 0.06, "learning_rate": 1.9962004369246338e-05, "loss": 1.0414, "step": 1711 }, { "epoch": 0.06, "learning_rate": 1.9961910438566657e-05, "loss": 1.0459, "step": 1712 }, { "epoch": 0.06, "learning_rate": 1.996181639214671e-05, "loss": 1.0435, "step": 1713 }, { "epoch": 0.06, "learning_rate": 1.99617222299876e-05, "loss": 1.0623, "step": 1714 }, { "epoch": 0.06, "learning_rate": 1.9961627952090416e-05, "loss": 1.0835, "step": 1715 }, { "epoch": 0.06, "learning_rate": 1.9961533558456248e-05, "loss": 1.0627, "step": 1716 }, { "epoch": 0.06, "learning_rate": 1.99614390490862e-05, "loss": 1.0927, "step": 1717 }, { "epoch": 0.06, "learning_rate": 1.996134442398137e-05, "loss": 1.1253, "step": 1718 }, { "epoch": 0.06, "learning_rate": 1.9961249683142855e-05, "loss": 1.0831, "step": 1719 }, { "epoch": 0.06, "learning_rate": 1.9961154826571754e-05, "loss": 1.0345, "step": 1720 }, { "epoch": 0.06, "learning_rate": 1.9961059854269172e-05, "loss": 1.1089, "step": 1721 }, { "epoch": 0.06, "learning_rate": 1.996096476623621e-05, "loss": 1.0432, "step": 1722 }, { "epoch": 0.06, "learning_rate": 1.996086956247398e-05, "loss": 1.0741, "step": 1723 }, { "epoch": 0.06, "learning_rate": 1.9960774242983577e-05, "loss": 1.0311, "step": 1724 }, { "epoch": 0.06, "learning_rate": 1.9960678807766113e-05, "loss": 1.0718, "step": 1725 }, { "epoch": 0.06, "learning_rate": 1.99605832568227e-05, "loss": 1.0743, "step": 1726 }, { "epoch": 0.06, "learning_rate": 1.9960487590154445e-05, "loss": 1.0671, "step": 1727 }, { "epoch": 0.06, "learning_rate": 1.9960391807762462e-05, "loss": 1.0573, "step": 1728 }, { "epoch": 0.06, "learning_rate": 1.996029590964786e-05, "loss": 1.0808, "step": 1729 }, { "epoch": 0.06, "learning_rate": 1.9960199895811755e-05, "loss": 1.0438, "step": 1730 }, { "epoch": 0.06, "learning_rate": 1.9960103766255264e-05, "loss": 1.0459, "step": 1731 }, { "epoch": 0.06, "learning_rate": 1.9960007520979504e-05, "loss": 1.0388, "step": 1732 }, { "epoch": 0.06, "learning_rate": 1.995991115998559e-05, "loss": 1.0982, "step": 1733 }, { "epoch": 0.06, "learning_rate": 1.995981468327464e-05, "loss": 1.0746, "step": 1734 }, { "epoch": 0.06, "learning_rate": 1.9959718090847784e-05, "loss": 1.0601, "step": 1735 }, { "epoch": 0.06, "learning_rate": 1.9959621382706133e-05, "loss": 1.0138, "step": 1736 }, { "epoch": 0.06, "learning_rate": 1.9959524558850824e-05, "loss": 1.0762, "step": 1737 }, { "epoch": 0.06, "learning_rate": 1.9959427619282963e-05, "loss": 1.1076, "step": 1738 }, { "epoch": 0.06, "learning_rate": 1.9959330564003697e-05, "loss": 1.0441, "step": 1739 }, { "epoch": 0.06, "learning_rate": 1.995923339301414e-05, "loss": 1.0486, "step": 1740 }, { "epoch": 0.06, "learning_rate": 1.9959136106315424e-05, "loss": 1.0677, "step": 1741 }, { "epoch": 0.06, "learning_rate": 1.9959038703908683e-05, "loss": 1.0637, "step": 1742 }, { "epoch": 0.06, "learning_rate": 1.9958941185795042e-05, "loss": 1.0499, "step": 1743 }, { "epoch": 0.06, "learning_rate": 1.995884355197564e-05, "loss": 1.0674, "step": 1744 }, { "epoch": 0.06, "learning_rate": 1.9958745802451607e-05, "loss": 1.0996, "step": 1745 }, { "epoch": 0.06, "learning_rate": 1.9958647937224085e-05, "loss": 1.0649, "step": 1746 }, { "epoch": 0.06, "learning_rate": 1.9958549956294203e-05, "loss": 1.0586, "step": 1747 }, { "epoch": 0.06, "learning_rate": 1.9958451859663104e-05, "loss": 1.0214, "step": 1748 }, { "epoch": 0.06, "learning_rate": 1.9958353647331926e-05, "loss": 1.0541, "step": 1749 }, { "epoch": 0.06, "learning_rate": 1.9958255319301814e-05, "loss": 1.0346, "step": 1750 }, { "epoch": 0.06, "learning_rate": 1.9958156875573906e-05, "loss": 1.0715, "step": 1751 }, { "epoch": 0.06, "learning_rate": 1.995805831614935e-05, "loss": 1.0204, "step": 1752 }, { "epoch": 0.06, "learning_rate": 1.9957959641029282e-05, "loss": 1.0403, "step": 1753 }, { "epoch": 0.06, "learning_rate": 1.995786085021486e-05, "loss": 1.0648, "step": 1754 }, { "epoch": 0.06, "learning_rate": 1.995776194370722e-05, "loss": 1.0523, "step": 1755 }, { "epoch": 0.06, "learning_rate": 1.9957662921507523e-05, "loss": 1.0276, "step": 1756 }, { "epoch": 0.06, "learning_rate": 1.995756378361691e-05, "loss": 1.0754, "step": 1757 }, { "epoch": 0.06, "learning_rate": 1.9957464530036543e-05, "loss": 1.0061, "step": 1758 }, { "epoch": 0.06, "learning_rate": 1.9957365160767566e-05, "loss": 1.0188, "step": 1759 }, { "epoch": 0.06, "learning_rate": 1.9957265675811133e-05, "loss": 1.0232, "step": 1760 }, { "epoch": 0.06, "learning_rate": 1.9957166075168407e-05, "loss": 1.0656, "step": 1761 }, { "epoch": 0.06, "learning_rate": 1.9957066358840544e-05, "loss": 1.0376, "step": 1762 }, { "epoch": 0.06, "learning_rate": 1.9956966526828695e-05, "loss": 1.0645, "step": 1763 }, { "epoch": 0.06, "learning_rate": 1.9956866579134026e-05, "loss": 1.0367, "step": 1764 }, { "epoch": 0.06, "learning_rate": 1.9956766515757696e-05, "loss": 1.0556, "step": 1765 }, { "epoch": 0.06, "learning_rate": 1.995666633670087e-05, "loss": 1.0491, "step": 1766 }, { "epoch": 0.06, "learning_rate": 1.995656604196471e-05, "loss": 1.0877, "step": 1767 }, { "epoch": 0.06, "learning_rate": 1.9956465631550387e-05, "loss": 1.0266, "step": 1768 }, { "epoch": 0.06, "learning_rate": 1.9956365105459056e-05, "loss": 1.0531, "step": 1769 }, { "epoch": 0.06, "learning_rate": 1.9956264463691896e-05, "loss": 1.0524, "step": 1770 }, { "epoch": 0.06, "learning_rate": 1.995616370625007e-05, "loss": 1.1179, "step": 1771 }, { "epoch": 0.06, "learning_rate": 1.9956062833134746e-05, "loss": 0.9834, "step": 1772 }, { "epoch": 0.06, "learning_rate": 1.9955961844347107e-05, "loss": 1.0499, "step": 1773 }, { "epoch": 0.06, "learning_rate": 1.9955860739888315e-05, "loss": 1.0426, "step": 1774 }, { "epoch": 0.06, "learning_rate": 1.9955759519759553e-05, "loss": 1.0504, "step": 1775 }, { "epoch": 0.06, "learning_rate": 1.995565818396199e-05, "loss": 1.0508, "step": 1776 }, { "epoch": 0.06, "learning_rate": 1.9955556732496805e-05, "loss": 1.0721, "step": 1777 }, { "epoch": 0.06, "learning_rate": 1.9955455165365184e-05, "loss": 1.0702, "step": 1778 }, { "epoch": 0.06, "learning_rate": 1.9955353482568297e-05, "loss": 1.0522, "step": 1779 }, { "epoch": 0.06, "learning_rate": 1.9955251684107333e-05, "loss": 1.016, "step": 1780 }, { "epoch": 0.06, "learning_rate": 1.9955149769983468e-05, "loss": 1.0904, "step": 1781 }, { "epoch": 0.06, "learning_rate": 1.995504774019789e-05, "loss": 1.0791, "step": 1782 }, { "epoch": 0.06, "learning_rate": 1.9954945594751788e-05, "loss": 1.0319, "step": 1783 }, { "epoch": 0.06, "learning_rate": 1.9954843333646337e-05, "loss": 1.0327, "step": 1784 }, { "epoch": 0.06, "learning_rate": 1.9954740956882743e-05, "loss": 1.0596, "step": 1785 }, { "epoch": 0.06, "learning_rate": 1.9954638464462176e-05, "loss": 1.0627, "step": 1786 }, { "epoch": 0.06, "learning_rate": 1.995453585638584e-05, "loss": 1.08, "step": 1787 }, { "epoch": 0.06, "learning_rate": 1.995443313265492e-05, "loss": 1.0177, "step": 1788 }, { "epoch": 0.06, "learning_rate": 1.9954330293270618e-05, "loss": 1.0822, "step": 1789 }, { "epoch": 0.06, "learning_rate": 1.995422733823412e-05, "loss": 1.0578, "step": 1790 }, { "epoch": 0.06, "learning_rate": 1.9954124267546627e-05, "loss": 1.0309, "step": 1791 }, { "epoch": 0.06, "learning_rate": 1.9954021081209334e-05, "loss": 0.9919, "step": 1792 }, { "epoch": 0.06, "learning_rate": 1.9953917779223442e-05, "loss": 1.0162, "step": 1793 }, { "epoch": 0.06, "learning_rate": 1.995381436159015e-05, "loss": 1.0303, "step": 1794 }, { "epoch": 0.06, "learning_rate": 1.995371082831066e-05, "loss": 1.0535, "step": 1795 }, { "epoch": 0.06, "learning_rate": 1.995360717938617e-05, "loss": 1.0328, "step": 1796 }, { "epoch": 0.06, "learning_rate": 1.9953503414817895e-05, "loss": 1.0654, "step": 1797 }, { "epoch": 0.06, "learning_rate": 1.995339953460703e-05, "loss": 1.0753, "step": 1798 }, { "epoch": 0.06, "learning_rate": 1.9953295538754788e-05, "loss": 1.0733, "step": 1799 }, { "epoch": 0.06, "learning_rate": 1.9953191427262375e-05, "loss": 1.0449, "step": 1800 }, { "epoch": 0.06, "learning_rate": 1.9953087200131e-05, "loss": 1.1248, "step": 1801 }, { "epoch": 0.06, "learning_rate": 1.9952982857361878e-05, "loss": 1.0706, "step": 1802 }, { "epoch": 0.06, "learning_rate": 1.9952878398956214e-05, "loss": 1.0631, "step": 1803 }, { "epoch": 0.06, "learning_rate": 1.995277382491523e-05, "loss": 1.0455, "step": 1804 }, { "epoch": 0.06, "learning_rate": 1.9952669135240135e-05, "loss": 1.0717, "step": 1805 }, { "epoch": 0.06, "learning_rate": 1.9952564329932148e-05, "loss": 1.0831, "step": 1806 }, { "epoch": 0.06, "learning_rate": 1.9952459408992485e-05, "loss": 1.1095, "step": 1807 }, { "epoch": 0.06, "learning_rate": 1.995235437242237e-05, "loss": 1.0546, "step": 1808 }, { "epoch": 0.06, "learning_rate": 1.9952249220223012e-05, "loss": 1.0505, "step": 1809 }, { "epoch": 0.06, "learning_rate": 1.9952143952395646e-05, "loss": 1.0794, "step": 1810 }, { "epoch": 0.06, "learning_rate": 1.9952038568941485e-05, "loss": 1.1017, "step": 1811 }, { "epoch": 0.06, "learning_rate": 1.9951933069861758e-05, "loss": 1.0834, "step": 1812 }, { "epoch": 0.06, "learning_rate": 1.995182745515769e-05, "loss": 1.0459, "step": 1813 }, { "epoch": 0.06, "learning_rate": 1.995172172483051e-05, "loss": 1.0505, "step": 1814 }, { "epoch": 0.06, "learning_rate": 1.9951615878881445e-05, "loss": 1.0856, "step": 1815 }, { "epoch": 0.06, "learning_rate": 1.9951509917311723e-05, "loss": 1.0301, "step": 1816 }, { "epoch": 0.06, "learning_rate": 1.9951403840122573e-05, "loss": 1.0798, "step": 1817 }, { "epoch": 0.06, "learning_rate": 1.9951297647315235e-05, "loss": 1.0448, "step": 1818 }, { "epoch": 0.06, "learning_rate": 1.9951191338890935e-05, "loss": 1.099, "step": 1819 }, { "epoch": 0.06, "learning_rate": 1.9951084914850917e-05, "loss": 1.0287, "step": 1820 }, { "epoch": 0.06, "learning_rate": 1.9950978375196408e-05, "loss": 1.0917, "step": 1821 }, { "epoch": 0.06, "learning_rate": 1.995087171992865e-05, "loss": 1.0441, "step": 1822 }, { "epoch": 0.06, "learning_rate": 1.9950764949048884e-05, "loss": 1.0348, "step": 1823 }, { "epoch": 0.06, "learning_rate": 1.9950658062558347e-05, "loss": 1.047, "step": 1824 }, { "epoch": 0.06, "learning_rate": 1.9950551060458282e-05, "loss": 1.1024, "step": 1825 }, { "epoch": 0.06, "learning_rate": 1.9950443942749936e-05, "loss": 1.0533, "step": 1826 }, { "epoch": 0.06, "learning_rate": 1.9950336709434545e-05, "loss": 1.1039, "step": 1827 }, { "epoch": 0.06, "learning_rate": 1.9950229360513364e-05, "loss": 1.0303, "step": 1828 }, { "epoch": 0.06, "learning_rate": 1.9950121895987635e-05, "loss": 1.0613, "step": 1829 }, { "epoch": 0.06, "learning_rate": 1.995001431585861e-05, "loss": 1.0637, "step": 1830 }, { "epoch": 0.06, "learning_rate": 1.9949906620127532e-05, "loss": 1.05, "step": 1831 }, { "epoch": 0.06, "learning_rate": 1.994979880879566e-05, "loss": 1.0303, "step": 1832 }, { "epoch": 0.06, "learning_rate": 1.9949690881864242e-05, "loss": 1.0355, "step": 1833 }, { "epoch": 0.06, "learning_rate": 1.9949582839334533e-05, "loss": 1.049, "step": 1834 }, { "epoch": 0.06, "learning_rate": 1.994947468120779e-05, "loss": 1.0516, "step": 1835 }, { "epoch": 0.06, "learning_rate": 1.994936640748527e-05, "loss": 1.0382, "step": 1836 }, { "epoch": 0.06, "learning_rate": 1.9949258018168227e-05, "loss": 1.0635, "step": 1837 }, { "epoch": 0.06, "learning_rate": 1.9949149513257923e-05, "loss": 1.0385, "step": 1838 }, { "epoch": 0.06, "learning_rate": 1.9949040892755614e-05, "loss": 1.0392, "step": 1839 }, { "epoch": 0.06, "learning_rate": 1.994893215666257e-05, "loss": 1.0481, "step": 1840 }, { "epoch": 0.06, "learning_rate": 1.994882330498005e-05, "loss": 1.0876, "step": 1841 }, { "epoch": 0.06, "learning_rate": 1.994871433770932e-05, "loss": 1.0471, "step": 1842 }, { "epoch": 0.06, "learning_rate": 1.9948605254851646e-05, "loss": 1.0487, "step": 1843 }, { "epoch": 0.06, "learning_rate": 1.9948496056408294e-05, "loss": 1.0603, "step": 1844 }, { "epoch": 0.06, "learning_rate": 1.994838674238053e-05, "loss": 1.0552, "step": 1845 }, { "epoch": 0.06, "learning_rate": 1.994827731276963e-05, "loss": 1.0843, "step": 1846 }, { "epoch": 0.06, "learning_rate": 1.9948167767576865e-05, "loss": 1.1015, "step": 1847 }, { "epoch": 0.06, "learning_rate": 1.99480581068035e-05, "loss": 1.0065, "step": 1848 }, { "epoch": 0.06, "learning_rate": 1.994794833045082e-05, "loss": 1.0774, "step": 1849 }, { "epoch": 0.06, "learning_rate": 1.994783843852009e-05, "loss": 1.0678, "step": 1850 }, { "epoch": 0.06, "learning_rate": 1.9947728431012595e-05, "loss": 1.0857, "step": 1851 }, { "epoch": 0.06, "learning_rate": 1.9947618307929607e-05, "loss": 1.0371, "step": 1852 }, { "epoch": 0.06, "learning_rate": 1.994750806927241e-05, "loss": 1.1079, "step": 1853 }, { "epoch": 0.06, "learning_rate": 1.9947397715042283e-05, "loss": 1.0606, "step": 1854 }, { "epoch": 0.06, "learning_rate": 1.9947287245240506e-05, "loss": 1.0623, "step": 1855 }, { "epoch": 0.06, "learning_rate": 1.9947176659868365e-05, "loss": 1.0272, "step": 1856 }, { "epoch": 0.06, "learning_rate": 1.9947065958927147e-05, "loss": 1.1057, "step": 1857 }, { "epoch": 0.06, "learning_rate": 1.9946955142418135e-05, "loss": 1.0313, "step": 1858 }, { "epoch": 0.06, "learning_rate": 1.994684421034261e-05, "loss": 1.0668, "step": 1859 }, { "epoch": 0.06, "learning_rate": 1.9946733162701877e-05, "loss": 1.0385, "step": 1860 }, { "epoch": 0.06, "learning_rate": 1.9946621999497215e-05, "loss": 1.036, "step": 1861 }, { "epoch": 0.06, "learning_rate": 1.9946510720729913e-05, "loss": 1.0343, "step": 1862 }, { "epoch": 0.06, "learning_rate": 1.9946399326401273e-05, "loss": 1.0368, "step": 1863 }, { "epoch": 0.06, "learning_rate": 1.9946287816512582e-05, "loss": 1.0145, "step": 1864 }, { "epoch": 0.06, "learning_rate": 1.994617619106514e-05, "loss": 1.0639, "step": 1865 }, { "epoch": 0.06, "learning_rate": 1.9946064450060243e-05, "loss": 1.025, "step": 1866 }, { "epoch": 0.06, "learning_rate": 1.9945952593499185e-05, "loss": 1.0936, "step": 1867 }, { "epoch": 0.06, "learning_rate": 1.994584062138327e-05, "loss": 1.0123, "step": 1868 }, { "epoch": 0.06, "learning_rate": 1.9945728533713798e-05, "loss": 1.07, "step": 1869 }, { "epoch": 0.06, "learning_rate": 1.994561633049207e-05, "loss": 1.0364, "step": 1870 }, { "epoch": 0.06, "learning_rate": 1.9945504011719394e-05, "loss": 1.034, "step": 1871 }, { "epoch": 0.06, "learning_rate": 1.9945391577397067e-05, "loss": 1.0414, "step": 1872 }, { "epoch": 0.06, "learning_rate": 1.9945279027526405e-05, "loss": 1.0631, "step": 1873 }, { "epoch": 0.06, "learning_rate": 1.9945166362108706e-05, "loss": 1.0508, "step": 1874 }, { "epoch": 0.06, "learning_rate": 1.9945053581145285e-05, "loss": 1.0684, "step": 1875 }, { "epoch": 0.06, "learning_rate": 1.994494068463745e-05, "loss": 1.0249, "step": 1876 }, { "epoch": 0.06, "learning_rate": 1.9944827672586518e-05, "loss": 1.0781, "step": 1877 }, { "epoch": 0.06, "learning_rate": 1.9944714544993792e-05, "loss": 1.0712, "step": 1878 }, { "epoch": 0.06, "learning_rate": 1.9944601301860596e-05, "loss": 1.0616, "step": 1879 }, { "epoch": 0.06, "learning_rate": 1.994448794318824e-05, "loss": 1.0068, "step": 1880 }, { "epoch": 0.06, "learning_rate": 1.9944374468978044e-05, "loss": 1.0761, "step": 1881 }, { "epoch": 0.06, "learning_rate": 1.994426087923132e-05, "loss": 1.0577, "step": 1882 }, { "epoch": 0.06, "learning_rate": 1.9944147173949396e-05, "loss": 1.0325, "step": 1883 }, { "epoch": 0.06, "learning_rate": 1.9944033353133593e-05, "loss": 1.02, "step": 1884 }, { "epoch": 0.06, "learning_rate": 1.9943919416785224e-05, "loss": 1.0819, "step": 1885 }, { "epoch": 0.06, "learning_rate": 1.9943805364905623e-05, "loss": 1.0581, "step": 1886 }, { "epoch": 0.06, "learning_rate": 1.994369119749611e-05, "loss": 1.0238, "step": 1887 }, { "epoch": 0.06, "learning_rate": 1.994357691455801e-05, "loss": 1.0588, "step": 1888 }, { "epoch": 0.06, "learning_rate": 1.9943462516092654e-05, "loss": 1.08, "step": 1889 }, { "epoch": 0.06, "learning_rate": 1.9943348002101374e-05, "loss": 1.0642, "step": 1890 }, { "epoch": 0.06, "learning_rate": 1.9943233372585493e-05, "loss": 1.0301, "step": 1891 }, { "epoch": 0.06, "learning_rate": 1.9943118627546346e-05, "loss": 1.0441, "step": 1892 }, { "epoch": 0.06, "learning_rate": 1.994300376698527e-05, "loss": 1.0802, "step": 1893 }, { "epoch": 0.06, "learning_rate": 1.9942888790903593e-05, "loss": 1.0717, "step": 1894 }, { "epoch": 0.06, "learning_rate": 1.9942773699302657e-05, "loss": 1.064, "step": 1895 }, { "epoch": 0.06, "learning_rate": 1.9942658492183795e-05, "loss": 1.0038, "step": 1896 }, { "epoch": 0.06, "learning_rate": 1.9942543169548347e-05, "loss": 1.0486, "step": 1897 }, { "epoch": 0.06, "learning_rate": 1.994242773139765e-05, "loss": 1.0452, "step": 1898 }, { "epoch": 0.06, "learning_rate": 1.994231217773305e-05, "loss": 1.1145, "step": 1899 }, { "epoch": 0.06, "learning_rate": 1.9942196508555887e-05, "loss": 1.0609, "step": 1900 }, { "epoch": 0.06, "learning_rate": 1.9942080723867505e-05, "loss": 1.0716, "step": 1901 }, { "epoch": 0.06, "learning_rate": 1.994196482366925e-05, "loss": 1.0907, "step": 1902 }, { "epoch": 0.06, "learning_rate": 1.9941848807962466e-05, "loss": 1.0832, "step": 1903 }, { "epoch": 0.06, "learning_rate": 1.9941732676748503e-05, "loss": 1.0425, "step": 1904 }, { "epoch": 0.06, "learning_rate": 1.9941616430028713e-05, "loss": 1.095, "step": 1905 }, { "epoch": 0.06, "learning_rate": 1.9941500067804444e-05, "loss": 1.055, "step": 1906 }, { "epoch": 0.06, "learning_rate": 1.9941383590077045e-05, "loss": 1.0969, "step": 1907 }, { "epoch": 0.06, "learning_rate": 1.994126699684787e-05, "loss": 1.039, "step": 1908 }, { "epoch": 0.06, "learning_rate": 1.9941150288118275e-05, "loss": 1.0781, "step": 1909 }, { "epoch": 0.06, "learning_rate": 1.994103346388962e-05, "loss": 1.037, "step": 1910 }, { "epoch": 0.06, "learning_rate": 1.994091652416326e-05, "loss": 1.0196, "step": 1911 }, { "epoch": 0.06, "learning_rate": 1.994079946894055e-05, "loss": 1.038, "step": 1912 }, { "epoch": 0.06, "learning_rate": 1.994068229822285e-05, "loss": 1.0508, "step": 1913 }, { "epoch": 0.06, "learning_rate": 1.9940565012011528e-05, "loss": 1.0438, "step": 1914 }, { "epoch": 0.06, "learning_rate": 1.9940447610307938e-05, "loss": 1.0667, "step": 1915 }, { "epoch": 0.06, "learning_rate": 1.9940330093113453e-05, "loss": 1.0376, "step": 1916 }, { "epoch": 0.06, "learning_rate": 1.994021246042943e-05, "loss": 1.0676, "step": 1917 }, { "epoch": 0.06, "learning_rate": 1.994009471225724e-05, "loss": 1.0368, "step": 1918 }, { "epoch": 0.06, "learning_rate": 1.9939976848598254e-05, "loss": 1.0965, "step": 1919 }, { "epoch": 0.06, "learning_rate": 1.9939858869453832e-05, "loss": 1.0373, "step": 1920 }, { "epoch": 0.06, "learning_rate": 1.9939740774825355e-05, "loss": 1.0983, "step": 1921 }, { "epoch": 0.06, "learning_rate": 1.9939622564714187e-05, "loss": 1.0752, "step": 1922 }, { "epoch": 0.06, "learning_rate": 1.9939504239121706e-05, "loss": 1.0366, "step": 1923 }, { "epoch": 0.06, "learning_rate": 1.9939385798049286e-05, "loss": 1.0306, "step": 1924 }, { "epoch": 0.06, "learning_rate": 1.9939267241498303e-05, "loss": 1.0512, "step": 1925 }, { "epoch": 0.06, "learning_rate": 1.9939148569470132e-05, "loss": 1.036, "step": 1926 }, { "epoch": 0.06, "learning_rate": 1.9939029781966156e-05, "loss": 1.0302, "step": 1927 }, { "epoch": 0.06, "learning_rate": 1.9938910878987754e-05, "loss": 1.0308, "step": 1928 }, { "epoch": 0.06, "learning_rate": 1.9938791860536308e-05, "loss": 1.0561, "step": 1929 }, { "epoch": 0.06, "learning_rate": 1.9938672726613193e-05, "loss": 1.0581, "step": 1930 }, { "epoch": 0.06, "learning_rate": 1.9938553477219803e-05, "loss": 1.0676, "step": 1931 }, { "epoch": 0.06, "learning_rate": 1.993843411235752e-05, "loss": 1.0461, "step": 1932 }, { "epoch": 0.06, "learning_rate": 1.9938314632027732e-05, "loss": 1.0326, "step": 1933 }, { "epoch": 0.06, "learning_rate": 1.9938195036231825e-05, "loss": 1.075, "step": 1934 }, { "epoch": 0.06, "learning_rate": 1.993807532497119e-05, "loss": 1.0574, "step": 1935 }, { "epoch": 0.06, "learning_rate": 1.9937955498247214e-05, "loss": 1.042, "step": 1936 }, { "epoch": 0.06, "learning_rate": 1.9937835556061294e-05, "loss": 1.0595, "step": 1937 }, { "epoch": 0.06, "learning_rate": 1.993771549841482e-05, "loss": 1.0508, "step": 1938 }, { "epoch": 0.06, "learning_rate": 1.9937595325309193e-05, "loss": 1.0627, "step": 1939 }, { "epoch": 0.06, "learning_rate": 1.9937475036745802e-05, "loss": 1.0342, "step": 1940 }, { "epoch": 0.06, "learning_rate": 1.993735463272605e-05, "loss": 1.0535, "step": 1941 }, { "epoch": 0.06, "learning_rate": 1.993723411325133e-05, "loss": 1.0898, "step": 1942 }, { "epoch": 0.06, "learning_rate": 1.9937113478323047e-05, "loss": 1.0578, "step": 1943 }, { "epoch": 0.06, "learning_rate": 1.9936992727942602e-05, "loss": 1.0455, "step": 1944 }, { "epoch": 0.06, "learning_rate": 1.9936871862111397e-05, "loss": 1.0899, "step": 1945 }, { "epoch": 0.06, "learning_rate": 1.9936750880830838e-05, "loss": 1.0406, "step": 1946 }, { "epoch": 0.06, "learning_rate": 1.9936629784102327e-05, "loss": 1.024, "step": 1947 }, { "epoch": 0.06, "learning_rate": 1.993650857192727e-05, "loss": 1.0359, "step": 1948 }, { "epoch": 0.06, "learning_rate": 1.9936387244307083e-05, "loss": 1.0886, "step": 1949 }, { "epoch": 0.06, "learning_rate": 1.9936265801243168e-05, "loss": 1.0555, "step": 1950 }, { "epoch": 0.06, "learning_rate": 1.993614424273694e-05, "loss": 1.0654, "step": 1951 }, { "epoch": 0.06, "learning_rate": 1.993602256878981e-05, "loss": 1.0421, "step": 1952 }, { "epoch": 0.06, "learning_rate": 1.993590077940319e-05, "loss": 1.063, "step": 1953 }, { "epoch": 0.07, "learning_rate": 1.99357788745785e-05, "loss": 1.0561, "step": 1954 }, { "epoch": 0.07, "learning_rate": 1.993565685431715e-05, "loss": 1.0786, "step": 1955 }, { "epoch": 0.07, "learning_rate": 1.9935534718620563e-05, "loss": 1.0554, "step": 1956 }, { "epoch": 0.07, "learning_rate": 1.993541246749015e-05, "loss": 1.0578, "step": 1957 }, { "epoch": 0.07, "learning_rate": 1.993529010092734e-05, "loss": 1.0474, "step": 1958 }, { "epoch": 0.07, "learning_rate": 1.9935167618933556e-05, "loss": 1.0714, "step": 1959 }, { "epoch": 0.07, "learning_rate": 1.993504502151021e-05, "loss": 1.0281, "step": 1960 }, { "epoch": 0.07, "learning_rate": 1.993492230865874e-05, "loss": 1.1004, "step": 1961 }, { "epoch": 0.07, "learning_rate": 1.993479948038056e-05, "loss": 1.0652, "step": 1962 }, { "epoch": 0.07, "learning_rate": 1.9934676536677102e-05, "loss": 1.1024, "step": 1963 }, { "epoch": 0.07, "learning_rate": 1.9934553477549795e-05, "loss": 1.0472, "step": 1964 }, { "epoch": 0.07, "learning_rate": 1.9934430303000072e-05, "loss": 1.0488, "step": 1965 }, { "epoch": 0.07, "learning_rate": 1.9934307013029355e-05, "loss": 1.0753, "step": 1966 }, { "epoch": 0.07, "learning_rate": 1.9934183607639085e-05, "loss": 1.0296, "step": 1967 }, { "epoch": 0.07, "learning_rate": 1.993406008683069e-05, "loss": 1.0716, "step": 1968 }, { "epoch": 0.07, "learning_rate": 1.9933936450605605e-05, "loss": 1.0428, "step": 1969 }, { "epoch": 0.07, "learning_rate": 1.9933812698965274e-05, "loss": 1.0672, "step": 1970 }, { "epoch": 0.07, "learning_rate": 1.9933688831911128e-05, "loss": 1.0714, "step": 1971 }, { "epoch": 0.07, "learning_rate": 1.9933564849444602e-05, "loss": 1.0439, "step": 1972 }, { "epoch": 0.07, "learning_rate": 1.9933440751567148e-05, "loss": 1.084, "step": 1973 }, { "epoch": 0.07, "learning_rate": 1.99333165382802e-05, "loss": 1.0475, "step": 1974 }, { "epoch": 0.07, "learning_rate": 1.9933192209585207e-05, "loss": 1.0391, "step": 1975 }, { "epoch": 0.07, "learning_rate": 1.9933067765483604e-05, "loss": 1.0581, "step": 1976 }, { "epoch": 0.07, "learning_rate": 1.9932943205976845e-05, "loss": 1.1036, "step": 1977 }, { "epoch": 0.07, "learning_rate": 1.9932818531066377e-05, "loss": 1.0943, "step": 1978 }, { "epoch": 0.07, "learning_rate": 1.993269374075364e-05, "loss": 1.0394, "step": 1979 }, { "epoch": 0.07, "learning_rate": 1.9932568835040094e-05, "loss": 1.0326, "step": 1980 }, { "epoch": 0.07, "learning_rate": 1.9932443813927186e-05, "loss": 1.0714, "step": 1981 }, { "epoch": 0.07, "learning_rate": 1.9932318677416366e-05, "loss": 1.0298, "step": 1982 }, { "epoch": 0.07, "learning_rate": 1.9932193425509094e-05, "loss": 1.0271, "step": 1983 }, { "epoch": 0.07, "learning_rate": 1.993206805820682e-05, "loss": 0.9751, "step": 1984 }, { "epoch": 0.07, "learning_rate": 1.9931942575511003e-05, "loss": 1.0798, "step": 1985 }, { "epoch": 0.07, "learning_rate": 1.99318169774231e-05, "loss": 1.0661, "step": 1986 }, { "epoch": 0.07, "learning_rate": 1.993169126394457e-05, "loss": 1.0525, "step": 1987 }, { "epoch": 0.07, "learning_rate": 1.9931565435076872e-05, "loss": 1.0373, "step": 1988 }, { "epoch": 0.07, "learning_rate": 1.9931439490821475e-05, "loss": 1.0927, "step": 1989 }, { "epoch": 0.07, "learning_rate": 1.9931313431179834e-05, "loss": 1.0547, "step": 1990 }, { "epoch": 0.07, "learning_rate": 1.9931187256153417e-05, "loss": 1.1091, "step": 1991 }, { "epoch": 0.07, "learning_rate": 1.993106096574369e-05, "loss": 1.0294, "step": 1992 }, { "epoch": 0.07, "learning_rate": 1.9930934559952118e-05, "loss": 1.0847, "step": 1993 }, { "epoch": 0.07, "learning_rate": 1.9930808038780174e-05, "loss": 0.9987, "step": 1994 }, { "epoch": 0.07, "learning_rate": 1.9930681402229326e-05, "loss": 1.059, "step": 1995 }, { "epoch": 0.07, "learning_rate": 1.9930554650301044e-05, "loss": 0.9787, "step": 1996 }, { "epoch": 0.07, "learning_rate": 1.99304277829968e-05, "loss": 1.0551, "step": 1997 }, { "epoch": 0.07, "learning_rate": 1.993030080031807e-05, "loss": 1.0655, "step": 1998 }, { "epoch": 0.07, "learning_rate": 1.993017370226633e-05, "loss": 1.0449, "step": 1999 }, { "epoch": 0.07, "learning_rate": 1.9930046488843056e-05, "loss": 1.0567, "step": 2000 }, { "epoch": 0.07, "learning_rate": 1.9929919160049724e-05, "loss": 1.0767, "step": 2001 }, { "epoch": 0.07, "learning_rate": 1.992979171588781e-05, "loss": 1.0764, "step": 2002 }, { "epoch": 0.07, "learning_rate": 1.9929664156358807e-05, "loss": 1.0901, "step": 2003 }, { "epoch": 0.07, "learning_rate": 1.9929536481464188e-05, "loss": 1.027, "step": 2004 }, { "epoch": 0.07, "learning_rate": 1.9929408691205436e-05, "loss": 1.0609, "step": 2005 }, { "epoch": 0.07, "learning_rate": 1.992928078558404e-05, "loss": 1.0845, "step": 2006 }, { "epoch": 0.07, "learning_rate": 1.992915276460148e-05, "loss": 1.0456, "step": 2007 }, { "epoch": 0.07, "learning_rate": 1.992902462825925e-05, "loss": 1.0042, "step": 2008 }, { "epoch": 0.07, "learning_rate": 1.992889637655883e-05, "loss": 1.0463, "step": 2009 }, { "epoch": 0.07, "learning_rate": 1.9928768009501725e-05, "loss": 1.0409, "step": 2010 }, { "epoch": 0.07, "learning_rate": 1.992863952708941e-05, "loss": 1.0711, "step": 2011 }, { "epoch": 0.07, "learning_rate": 1.992851092932339e-05, "loss": 1.0718, "step": 2012 }, { "epoch": 0.07, "learning_rate": 1.992838221620515e-05, "loss": 1.0326, "step": 2013 }, { "epoch": 0.07, "learning_rate": 1.992825338773619e-05, "loss": 1.0573, "step": 2014 }, { "epoch": 0.07, "learning_rate": 1.992812444391801e-05, "loss": 1.0497, "step": 2015 }, { "epoch": 0.07, "learning_rate": 1.99279953847521e-05, "loss": 1.0472, "step": 2016 }, { "epoch": 0.07, "learning_rate": 1.9927866210239967e-05, "loss": 1.0635, "step": 2017 }, { "epoch": 0.07, "learning_rate": 1.9927736920383104e-05, "loss": 1.0506, "step": 2018 }, { "epoch": 0.07, "learning_rate": 1.992760751518302e-05, "loss": 1.0143, "step": 2019 }, { "epoch": 0.07, "learning_rate": 1.9927477994641217e-05, "loss": 1.0484, "step": 2020 }, { "epoch": 0.07, "learning_rate": 1.9927348358759198e-05, "loss": 1.0516, "step": 2021 }, { "epoch": 0.07, "learning_rate": 1.992721860753847e-05, "loss": 1.0089, "step": 2022 }, { "epoch": 0.07, "learning_rate": 1.992708874098054e-05, "loss": 1.0564, "step": 2023 }, { "epoch": 0.07, "learning_rate": 1.9926958759086918e-05, "loss": 1.0212, "step": 2024 }, { "epoch": 0.07, "learning_rate": 1.9926828661859114e-05, "loss": 1.0546, "step": 2025 }, { "epoch": 0.07, "learning_rate": 1.992669844929864e-05, "loss": 1.0149, "step": 2026 }, { "epoch": 0.07, "learning_rate": 1.9926568121407007e-05, "loss": 1.0687, "step": 2027 }, { "epoch": 0.07, "learning_rate": 1.992643767818573e-05, "loss": 0.9974, "step": 2028 }, { "epoch": 0.07, "learning_rate": 1.9926307119636324e-05, "loss": 1.0353, "step": 2029 }, { "epoch": 0.07, "learning_rate": 1.9926176445760308e-05, "loss": 1.052, "step": 2030 }, { "epoch": 0.07, "learning_rate": 1.9926045656559197e-05, "loss": 1.0513, "step": 2031 }, { "epoch": 0.07, "learning_rate": 1.9925914752034513e-05, "loss": 1.0366, "step": 2032 }, { "epoch": 0.07, "learning_rate": 1.9925783732187777e-05, "loss": 1.0715, "step": 2033 }, { "epoch": 0.07, "learning_rate": 1.9925652597020507e-05, "loss": 1.0297, "step": 2034 }, { "epoch": 0.07, "learning_rate": 1.9925521346534235e-05, "loss": 1.0815, "step": 2035 }, { "epoch": 0.07, "learning_rate": 1.9925389980730475e-05, "loss": 1.0261, "step": 2036 }, { "epoch": 0.07, "learning_rate": 1.9925258499610765e-05, "loss": 1.0411, "step": 2037 }, { "epoch": 0.07, "learning_rate": 1.9925126903176623e-05, "loss": 1.0322, "step": 2038 }, { "epoch": 0.07, "learning_rate": 1.992499519142958e-05, "loss": 1.0451, "step": 2039 }, { "epoch": 0.07, "learning_rate": 1.9924863364371173e-05, "loss": 1.0494, "step": 2040 }, { "epoch": 0.07, "learning_rate": 1.9924731422002925e-05, "loss": 1.0261, "step": 2041 }, { "epoch": 0.07, "learning_rate": 1.992459936432637e-05, "loss": 0.9903, "step": 2042 }, { "epoch": 0.07, "learning_rate": 1.9924467191343047e-05, "loss": 1.0269, "step": 2043 }, { "epoch": 0.07, "learning_rate": 1.9924334903054495e-05, "loss": 1.0, "step": 2044 }, { "epoch": 0.07, "learning_rate": 1.992420249946224e-05, "loss": 1.0524, "step": 2045 }, { "epoch": 0.07, "learning_rate": 1.9924069980567823e-05, "loss": 1.0631, "step": 2046 }, { "epoch": 0.07, "learning_rate": 1.992393734637279e-05, "loss": 1.0618, "step": 2047 }, { "epoch": 0.07, "learning_rate": 1.9923804596878678e-05, "loss": 1.0186, "step": 2048 }, { "epoch": 0.07, "learning_rate": 1.9923671732087027e-05, "loss": 1.045, "step": 2049 }, { "epoch": 0.07, "learning_rate": 1.9923538751999385e-05, "loss": 1.056, "step": 2050 }, { "epoch": 0.07, "learning_rate": 1.9923405656617295e-05, "loss": 1.0333, "step": 2051 }, { "epoch": 0.07, "learning_rate": 1.9923272445942304e-05, "loss": 1.009, "step": 2052 }, { "epoch": 0.07, "learning_rate": 1.9923139119975955e-05, "loss": 1.0306, "step": 2053 }, { "epoch": 0.07, "learning_rate": 1.9923005678719805e-05, "loss": 1.0491, "step": 2054 }, { "epoch": 0.07, "learning_rate": 1.9922872122175404e-05, "loss": 1.0668, "step": 2055 }, { "epoch": 0.07, "learning_rate": 1.9922738450344293e-05, "loss": 1.0326, "step": 2056 }, { "epoch": 0.07, "learning_rate": 1.9922604663228037e-05, "loss": 1.0479, "step": 2057 }, { "epoch": 0.07, "learning_rate": 1.9922470760828184e-05, "loss": 1.0379, "step": 2058 }, { "epoch": 0.07, "learning_rate": 1.9922336743146293e-05, "loss": 1.0352, "step": 2059 }, { "epoch": 0.07, "learning_rate": 1.9922202610183917e-05, "loss": 1.046, "step": 2060 }, { "epoch": 0.07, "learning_rate": 1.992206836194262e-05, "loss": 1.0933, "step": 2061 }, { "epoch": 0.07, "learning_rate": 1.9921933998423956e-05, "loss": 1.0655, "step": 2062 }, { "epoch": 0.07, "learning_rate": 1.992179951962949e-05, "loss": 1.017, "step": 2063 }, { "epoch": 0.07, "learning_rate": 1.992166492556078e-05, "loss": 1.0299, "step": 2064 }, { "epoch": 0.07, "learning_rate": 1.9921530216219394e-05, "loss": 1.0673, "step": 2065 }, { "epoch": 0.07, "learning_rate": 1.9921395391606898e-05, "loss": 1.0618, "step": 2066 }, { "epoch": 0.07, "learning_rate": 1.9921260451724852e-05, "loss": 1.0684, "step": 2067 }, { "epoch": 0.07, "learning_rate": 1.9921125396574835e-05, "loss": 1.0302, "step": 2068 }, { "epoch": 0.07, "learning_rate": 1.9920990226158404e-05, "loss": 1.0713, "step": 2069 }, { "epoch": 0.07, "learning_rate": 1.9920854940477136e-05, "loss": 1.0416, "step": 2070 }, { "epoch": 0.07, "learning_rate": 1.99207195395326e-05, "loss": 1.0138, "step": 2071 }, { "epoch": 0.07, "learning_rate": 1.9920584023326372e-05, "loss": 1.0092, "step": 2072 }, { "epoch": 0.07, "learning_rate": 1.9920448391860025e-05, "loss": 1.0467, "step": 2073 }, { "epoch": 0.07, "learning_rate": 1.9920312645135135e-05, "loss": 1.0459, "step": 2074 }, { "epoch": 0.07, "learning_rate": 1.9920176783153277e-05, "loss": 1.0367, "step": 2075 }, { "epoch": 0.07, "learning_rate": 1.9920040805916035e-05, "loss": 1.0346, "step": 2076 }, { "epoch": 0.07, "learning_rate": 1.9919904713424985e-05, "loss": 1.0222, "step": 2077 }, { "epoch": 0.07, "learning_rate": 1.9919768505681706e-05, "loss": 1.0768, "step": 2078 }, { "epoch": 0.07, "learning_rate": 1.9919632182687784e-05, "loss": 1.0098, "step": 2079 }, { "epoch": 0.07, "learning_rate": 1.99194957444448e-05, "loss": 1.0208, "step": 2080 }, { "epoch": 0.07, "learning_rate": 1.9919359190954344e-05, "loss": 1.0759, "step": 2081 }, { "epoch": 0.07, "learning_rate": 1.9919222522217998e-05, "loss": 0.9814, "step": 2082 }, { "epoch": 0.07, "learning_rate": 1.9919085738237348e-05, "loss": 1.0378, "step": 2083 }, { "epoch": 0.07, "learning_rate": 1.9918948839013993e-05, "loss": 1.0548, "step": 2084 }, { "epoch": 0.07, "learning_rate": 1.9918811824549515e-05, "loss": 1.02, "step": 2085 }, { "epoch": 0.07, "learning_rate": 1.9918674694845506e-05, "loss": 1.0799, "step": 2086 }, { "epoch": 0.07, "learning_rate": 1.991853744990356e-05, "loss": 1.0476, "step": 2087 }, { "epoch": 0.07, "learning_rate": 1.9918400089725277e-05, "loss": 0.9944, "step": 2088 }, { "epoch": 0.07, "learning_rate": 1.9918262614312246e-05, "loss": 1.0526, "step": 2089 }, { "epoch": 0.07, "learning_rate": 1.9918125023666063e-05, "loss": 1.0539, "step": 2090 }, { "epoch": 0.07, "learning_rate": 1.9917987317788333e-05, "loss": 1.0406, "step": 2091 }, { "epoch": 0.07, "learning_rate": 1.9917849496680657e-05, "loss": 0.9972, "step": 2092 }, { "epoch": 0.07, "learning_rate": 1.9917711560344632e-05, "loss": 1.0781, "step": 2093 }, { "epoch": 0.07, "learning_rate": 1.9917573508781856e-05, "loss": 1.0366, "step": 2094 }, { "epoch": 0.07, "learning_rate": 1.9917435341993943e-05, "loss": 1.0563, "step": 2095 }, { "epoch": 0.07, "learning_rate": 1.991729705998249e-05, "loss": 1.0087, "step": 2096 }, { "epoch": 0.07, "learning_rate": 1.9917158662749106e-05, "loss": 1.0278, "step": 2097 }, { "epoch": 0.07, "learning_rate": 1.9917020150295406e-05, "loss": 1.058, "step": 2098 }, { "epoch": 0.07, "learning_rate": 1.9916881522622988e-05, "loss": 1.0763, "step": 2099 }, { "epoch": 0.07, "learning_rate": 1.991674277973347e-05, "loss": 0.9908, "step": 2100 }, { "epoch": 0.07, "learning_rate": 1.991660392162846e-05, "loss": 1.0521, "step": 2101 }, { "epoch": 0.07, "learning_rate": 1.991646494830957e-05, "loss": 1.0592, "step": 2102 }, { "epoch": 0.07, "learning_rate": 1.9916325859778424e-05, "loss": 1.0559, "step": 2103 }, { "epoch": 0.07, "learning_rate": 1.9916186656036626e-05, "loss": 1.0693, "step": 2104 }, { "epoch": 0.07, "learning_rate": 1.9916047337085803e-05, "loss": 1.0184, "step": 2105 }, { "epoch": 0.07, "learning_rate": 1.9915907902927567e-05, "loss": 1.0617, "step": 2106 }, { "epoch": 0.07, "learning_rate": 1.9915768353563542e-05, "loss": 1.0724, "step": 2107 }, { "epoch": 0.07, "learning_rate": 1.991562868899535e-05, "loss": 1.0255, "step": 2108 }, { "epoch": 0.07, "learning_rate": 1.9915488909224607e-05, "loss": 1.0851, "step": 2109 }, { "epoch": 0.07, "learning_rate": 1.991534901425294e-05, "loss": 1.0606, "step": 2110 }, { "epoch": 0.07, "learning_rate": 1.991520900408198e-05, "loss": 1.1078, "step": 2111 }, { "epoch": 0.07, "learning_rate": 1.991506887871335e-05, "loss": 1.036, "step": 2112 }, { "epoch": 0.07, "learning_rate": 1.9914928638148674e-05, "loss": 1.0472, "step": 2113 }, { "epoch": 0.07, "learning_rate": 1.991478828238959e-05, "loss": 1.0312, "step": 2114 }, { "epoch": 0.07, "learning_rate": 1.991464781143772e-05, "loss": 1.0369, "step": 2115 }, { "epoch": 0.07, "learning_rate": 1.9914507225294697e-05, "loss": 1.0023, "step": 2116 }, { "epoch": 0.07, "learning_rate": 1.991436652396216e-05, "loss": 1.0891, "step": 2117 }, { "epoch": 0.07, "learning_rate": 1.991422570744174e-05, "loss": 1.0795, "step": 2118 }, { "epoch": 0.07, "learning_rate": 1.9914084775735073e-05, "loss": 1.0641, "step": 2119 }, { "epoch": 0.07, "learning_rate": 1.99139437288438e-05, "loss": 1.0392, "step": 2120 }, { "epoch": 0.07, "learning_rate": 1.9913802566769553e-05, "loss": 1.0323, "step": 2121 }, { "epoch": 0.07, "learning_rate": 1.991366128951398e-05, "loss": 1.0688, "step": 2122 }, { "epoch": 0.07, "learning_rate": 1.9913519897078717e-05, "loss": 1.0008, "step": 2123 }, { "epoch": 0.07, "learning_rate": 1.9913378389465405e-05, "loss": 0.9967, "step": 2124 }, { "epoch": 0.07, "learning_rate": 1.9913236766675693e-05, "loss": 1.0363, "step": 2125 }, { "epoch": 0.07, "learning_rate": 1.9913095028711226e-05, "loss": 1.0199, "step": 2126 }, { "epoch": 0.07, "learning_rate": 1.991295317557365e-05, "loss": 1.0636, "step": 2127 }, { "epoch": 0.07, "learning_rate": 1.991281120726461e-05, "loss": 1.0272, "step": 2128 }, { "epoch": 0.07, "learning_rate": 1.9912669123785756e-05, "loss": 1.0525, "step": 2129 }, { "epoch": 0.07, "learning_rate": 1.9912526925138744e-05, "loss": 1.0497, "step": 2130 }, { "epoch": 0.07, "learning_rate": 1.9912384611325222e-05, "loss": 1.0128, "step": 2131 }, { "epoch": 0.07, "learning_rate": 1.991224218234684e-05, "loss": 1.0625, "step": 2132 }, { "epoch": 0.07, "learning_rate": 1.991209963820526e-05, "loss": 1.0543, "step": 2133 }, { "epoch": 0.07, "learning_rate": 1.9911956978902135e-05, "loss": 1.0593, "step": 2134 }, { "epoch": 0.07, "learning_rate": 1.9911814204439123e-05, "loss": 1.1041, "step": 2135 }, { "epoch": 0.07, "learning_rate": 1.9911671314817883e-05, "loss": 1.0254, "step": 2136 }, { "epoch": 0.07, "learning_rate": 1.9911528310040073e-05, "loss": 1.0385, "step": 2137 }, { "epoch": 0.07, "learning_rate": 1.9911385190107356e-05, "loss": 0.9933, "step": 2138 }, { "epoch": 0.07, "learning_rate": 1.9911241955021393e-05, "loss": 1.0264, "step": 2139 }, { "epoch": 0.07, "learning_rate": 1.9911098604783853e-05, "loss": 1.0243, "step": 2140 }, { "epoch": 0.07, "learning_rate": 1.9910955139396395e-05, "loss": 1.0759, "step": 2141 }, { "epoch": 0.07, "learning_rate": 1.9910811558860693e-05, "loss": 1.0291, "step": 2142 }, { "epoch": 0.07, "learning_rate": 1.9910667863178407e-05, "loss": 1.0491, "step": 2143 }, { "epoch": 0.07, "learning_rate": 1.9910524052351214e-05, "loss": 1.0346, "step": 2144 }, { "epoch": 0.07, "learning_rate": 1.991038012638078e-05, "loss": 1.0997, "step": 2145 }, { "epoch": 0.07, "learning_rate": 1.991023608526878e-05, "loss": 1.0516, "step": 2146 }, { "epoch": 0.07, "learning_rate": 1.9910091929016883e-05, "loss": 1.0794, "step": 2147 }, { "epoch": 0.07, "learning_rate": 1.990994765762677e-05, "loss": 1.0247, "step": 2148 }, { "epoch": 0.07, "learning_rate": 1.9909803271100113e-05, "loss": 1.0543, "step": 2149 }, { "epoch": 0.07, "learning_rate": 1.9909658769438592e-05, "loss": 1.0534, "step": 2150 }, { "epoch": 0.07, "learning_rate": 1.9909514152643883e-05, "loss": 1.0673, "step": 2151 }, { "epoch": 0.07, "learning_rate": 1.990936942071767e-05, "loss": 1.0295, "step": 2152 }, { "epoch": 0.07, "learning_rate": 1.9909224573661634e-05, "loss": 1.0116, "step": 2153 }, { "epoch": 0.07, "learning_rate": 1.990907961147745e-05, "loss": 1.019, "step": 2154 }, { "epoch": 0.07, "learning_rate": 1.990893453416681e-05, "loss": 1.0535, "step": 2155 }, { "epoch": 0.07, "learning_rate": 1.9908789341731402e-05, "loss": 0.9999, "step": 2156 }, { "epoch": 0.07, "learning_rate": 1.990864403417291e-05, "loss": 1.0455, "step": 2157 }, { "epoch": 0.07, "learning_rate": 1.9908498611493016e-05, "loss": 1.0588, "step": 2158 }, { "epoch": 0.07, "learning_rate": 1.9908353073693417e-05, "loss": 1.047, "step": 2159 }, { "epoch": 0.07, "learning_rate": 1.9908207420775803e-05, "loss": 1.02, "step": 2160 }, { "epoch": 0.07, "learning_rate": 1.990806165274186e-05, "loss": 1.0578, "step": 2161 }, { "epoch": 0.07, "learning_rate": 1.9907915769593293e-05, "loss": 1.0553, "step": 2162 }, { "epoch": 0.07, "learning_rate": 1.9907769771331784e-05, "loss": 1.0745, "step": 2163 }, { "epoch": 0.07, "learning_rate": 1.990762365795904e-05, "loss": 1.0537, "step": 2164 }, { "epoch": 0.07, "learning_rate": 1.9907477429476752e-05, "loss": 1.0772, "step": 2165 }, { "epoch": 0.07, "learning_rate": 1.9907331085886623e-05, "loss": 1.083, "step": 2166 }, { "epoch": 0.07, "learning_rate": 1.990718462719035e-05, "loss": 1.1182, "step": 2167 }, { "epoch": 0.07, "learning_rate": 1.9907038053389637e-05, "loss": 1.0618, "step": 2168 }, { "epoch": 0.07, "learning_rate": 1.9906891364486185e-05, "loss": 1.1021, "step": 2169 }, { "epoch": 0.07, "learning_rate": 1.99067445604817e-05, "loss": 1.0368, "step": 2170 }, { "epoch": 0.07, "learning_rate": 1.9906597641377883e-05, "loss": 1.0466, "step": 2171 }, { "epoch": 0.07, "learning_rate": 1.990645060717645e-05, "loss": 0.9968, "step": 2172 }, { "epoch": 0.07, "learning_rate": 1.99063034578791e-05, "loss": 1.0748, "step": 2173 }, { "epoch": 0.07, "learning_rate": 1.9906156193487545e-05, "loss": 1.0217, "step": 2174 }, { "epoch": 0.07, "learning_rate": 1.9906008814003505e-05, "loss": 1.0665, "step": 2175 }, { "epoch": 0.07, "learning_rate": 1.990586131942868e-05, "loss": 0.9821, "step": 2176 }, { "epoch": 0.07, "learning_rate": 1.9905713709764787e-05, "loss": 1.0702, "step": 2177 }, { "epoch": 0.07, "learning_rate": 1.9905565985013546e-05, "loss": 1.056, "step": 2178 }, { "epoch": 0.07, "learning_rate": 1.9905418145176667e-05, "loss": 1.0085, "step": 2179 }, { "epoch": 0.07, "learning_rate": 1.9905270190255874e-05, "loss": 1.0526, "step": 2180 }, { "epoch": 0.07, "learning_rate": 1.9905122120252878e-05, "loss": 1.0473, "step": 2181 }, { "epoch": 0.07, "learning_rate": 1.9904973935169408e-05, "loss": 1.0295, "step": 2182 }, { "epoch": 0.07, "learning_rate": 1.9904825635007178e-05, "loss": 1.0563, "step": 2183 }, { "epoch": 0.07, "learning_rate": 1.9904677219767916e-05, "loss": 1.0678, "step": 2184 }, { "epoch": 0.07, "learning_rate": 1.9904528689453346e-05, "loss": 1.038, "step": 2185 }, { "epoch": 0.07, "learning_rate": 1.990438004406519e-05, "loss": 1.0553, "step": 2186 }, { "epoch": 0.07, "learning_rate": 1.9904231283605177e-05, "loss": 1.0758, "step": 2187 }, { "epoch": 0.07, "learning_rate": 1.9904082408075038e-05, "loss": 1.0753, "step": 2188 }, { "epoch": 0.07, "learning_rate": 1.99039334174765e-05, "loss": 1.0581, "step": 2189 }, { "epoch": 0.07, "learning_rate": 1.990378431181129e-05, "loss": 1.0871, "step": 2190 }, { "epoch": 0.07, "learning_rate": 1.990363509108115e-05, "loss": 1.0213, "step": 2191 }, { "epoch": 0.07, "learning_rate": 1.990348575528781e-05, "loss": 1.0459, "step": 2192 }, { "epoch": 0.07, "learning_rate": 1.9903336304432997e-05, "loss": 1.0566, "step": 2193 }, { "epoch": 0.07, "learning_rate": 1.990318673851846e-05, "loss": 1.0489, "step": 2194 }, { "epoch": 0.07, "learning_rate": 1.9903037057545928e-05, "loss": 1.0395, "step": 2195 }, { "epoch": 0.07, "learning_rate": 1.990288726151714e-05, "loss": 0.9977, "step": 2196 }, { "epoch": 0.07, "learning_rate": 1.990273735043384e-05, "loss": 1.0512, "step": 2197 }, { "epoch": 0.07, "learning_rate": 1.990258732429777e-05, "loss": 1.0243, "step": 2198 }, { "epoch": 0.07, "learning_rate": 1.9902437183110673e-05, "loss": 1.0559, "step": 2199 }, { "epoch": 0.07, "learning_rate": 1.990228692687429e-05, "loss": 1.0372, "step": 2200 }, { "epoch": 0.07, "learning_rate": 1.990213655559037e-05, "loss": 1.0783, "step": 2201 }, { "epoch": 0.07, "learning_rate": 1.9901986069260656e-05, "loss": 1.0183, "step": 2202 }, { "epoch": 0.07, "learning_rate": 1.99018354678869e-05, "loss": 1.0919, "step": 2203 }, { "epoch": 0.07, "learning_rate": 1.9901684751470855e-05, "loss": 1.091, "step": 2204 }, { "epoch": 0.07, "learning_rate": 1.9901533920014264e-05, "loss": 1.0349, "step": 2205 }, { "epoch": 0.07, "learning_rate": 1.9901382973518886e-05, "loss": 1.0711, "step": 2206 }, { "epoch": 0.07, "learning_rate": 1.990123191198647e-05, "loss": 1.0408, "step": 2207 }, { "epoch": 0.07, "learning_rate": 1.9901080735418776e-05, "loss": 0.9876, "step": 2208 }, { "epoch": 0.07, "learning_rate": 1.9900929443817553e-05, "loss": 1.0161, "step": 2209 }, { "epoch": 0.07, "learning_rate": 1.9900778037184568e-05, "loss": 1.0554, "step": 2210 }, { "epoch": 0.07, "learning_rate": 1.9900626515521573e-05, "loss": 1.0123, "step": 2211 }, { "epoch": 0.07, "learning_rate": 1.9900474878830333e-05, "loss": 1.0061, "step": 2212 }, { "epoch": 0.07, "learning_rate": 1.9900323127112608e-05, "loss": 1.0416, "step": 2213 }, { "epoch": 0.07, "learning_rate": 1.990017126037016e-05, "loss": 1.0792, "step": 2214 }, { "epoch": 0.07, "learning_rate": 1.9900019278604753e-05, "loss": 1.0407, "step": 2215 }, { "epoch": 0.07, "learning_rate": 1.9899867181818152e-05, "loss": 1.0388, "step": 2216 }, { "epoch": 0.07, "learning_rate": 1.989971497001213e-05, "loss": 1.1011, "step": 2217 }, { "epoch": 0.07, "learning_rate": 1.989956264318845e-05, "loss": 1.0397, "step": 2218 }, { "epoch": 0.07, "learning_rate": 1.989941020134888e-05, "loss": 1.0044, "step": 2219 }, { "epoch": 0.07, "learning_rate": 1.98992576444952e-05, "loss": 1.0477, "step": 2220 }, { "epoch": 0.07, "learning_rate": 1.989910497262917e-05, "loss": 1.0411, "step": 2221 }, { "epoch": 0.07, "learning_rate": 1.989895218575258e-05, "loss": 1.042, "step": 2222 }, { "epoch": 0.07, "learning_rate": 1.989879928386719e-05, "loss": 1.0494, "step": 2223 }, { "epoch": 0.07, "learning_rate": 1.9898646266974785e-05, "loss": 1.0037, "step": 2224 }, { "epoch": 0.07, "learning_rate": 1.9898493135077137e-05, "loss": 1.039, "step": 2225 }, { "epoch": 0.07, "learning_rate": 1.9898339888176027e-05, "loss": 1.0554, "step": 2226 }, { "epoch": 0.07, "learning_rate": 1.989818652627324e-05, "loss": 1.0442, "step": 2227 }, { "epoch": 0.07, "learning_rate": 1.9898033049370554e-05, "loss": 1.0053, "step": 2228 }, { "epoch": 0.07, "learning_rate": 1.989787945746975e-05, "loss": 1.0237, "step": 2229 }, { "epoch": 0.07, "learning_rate": 1.9897725750572618e-05, "loss": 1.0252, "step": 2230 }, { "epoch": 0.07, "learning_rate": 1.9897571928680938e-05, "loss": 1.0812, "step": 2231 }, { "epoch": 0.07, "learning_rate": 1.9897417991796502e-05, "loss": 1.0333, "step": 2232 }, { "epoch": 0.07, "learning_rate": 1.9897263939921096e-05, "loss": 1.0664, "step": 2233 }, { "epoch": 0.07, "learning_rate": 1.9897109773056513e-05, "loss": 1.004, "step": 2234 }, { "epoch": 0.07, "learning_rate": 1.9896955491204536e-05, "loss": 1.0105, "step": 2235 }, { "epoch": 0.07, "learning_rate": 1.9896801094366967e-05, "loss": 1.0384, "step": 2236 }, { "epoch": 0.07, "learning_rate": 1.9896646582545593e-05, "loss": 1.0538, "step": 2237 }, { "epoch": 0.07, "learning_rate": 1.9896491955742213e-05, "loss": 1.0138, "step": 2238 }, { "epoch": 0.07, "learning_rate": 1.9896337213958624e-05, "loss": 1.0155, "step": 2239 }, { "epoch": 0.07, "learning_rate": 1.9896182357196618e-05, "loss": 1.0029, "step": 2240 }, { "epoch": 0.07, "learning_rate": 1.9896027385458004e-05, "loss": 1.057, "step": 2241 }, { "epoch": 0.07, "learning_rate": 1.989587229874457e-05, "loss": 1.0197, "step": 2242 }, { "epoch": 0.07, "learning_rate": 1.989571709705813e-05, "loss": 1.0752, "step": 2243 }, { "epoch": 0.07, "learning_rate": 1.989556178040048e-05, "loss": 1.0377, "step": 2244 }, { "epoch": 0.07, "learning_rate": 1.9895406348773428e-05, "loss": 1.0144, "step": 2245 }, { "epoch": 0.07, "learning_rate": 1.9895250802178777e-05, "loss": 1.054, "step": 2246 }, { "epoch": 0.07, "learning_rate": 1.9895095140618337e-05, "loss": 1.0447, "step": 2247 }, { "epoch": 0.07, "learning_rate": 1.9894939364093916e-05, "loss": 1.0229, "step": 2248 }, { "epoch": 0.07, "learning_rate": 1.9894783472607317e-05, "loss": 1.0423, "step": 2249 }, { "epoch": 0.07, "learning_rate": 1.9894627466160362e-05, "loss": 1.0565, "step": 2250 }, { "epoch": 0.07, "learning_rate": 1.989447134475486e-05, "loss": 1.0488, "step": 2251 }, { "epoch": 0.07, "learning_rate": 1.989431510839262e-05, "loss": 1.0315, "step": 2252 }, { "epoch": 0.07, "learning_rate": 1.9894158757075458e-05, "loss": 1.0563, "step": 2253 }, { "epoch": 0.08, "learning_rate": 1.9894002290805196e-05, "loss": 1.0264, "step": 2254 }, { "epoch": 0.08, "learning_rate": 1.989384570958365e-05, "loss": 1.1031, "step": 2255 }, { "epoch": 0.08, "learning_rate": 1.9893689013412637e-05, "loss": 1.0102, "step": 2256 }, { "epoch": 0.08, "learning_rate": 1.9893532202293978e-05, "loss": 1.0878, "step": 2257 }, { "epoch": 0.08, "learning_rate": 1.98933752762295e-05, "loss": 1.0522, "step": 2258 }, { "epoch": 0.08, "learning_rate": 1.9893218235221016e-05, "loss": 1.076, "step": 2259 }, { "epoch": 0.08, "learning_rate": 1.989306107927036e-05, "loss": 1.0364, "step": 2260 }, { "epoch": 0.08, "learning_rate": 1.9892903808379356e-05, "loss": 1.057, "step": 2261 }, { "epoch": 0.08, "learning_rate": 1.9892746422549825e-05, "loss": 1.0479, "step": 2262 }, { "epoch": 0.08, "learning_rate": 1.9892588921783602e-05, "loss": 1.068, "step": 2263 }, { "epoch": 0.08, "learning_rate": 1.9892431306082516e-05, "loss": 1.0293, "step": 2264 }, { "epoch": 0.08, "learning_rate": 1.9892273575448396e-05, "loss": 1.0403, "step": 2265 }, { "epoch": 0.08, "learning_rate": 1.9892115729883076e-05, "loss": 1.0453, "step": 2266 }, { "epoch": 0.08, "learning_rate": 1.989195776938839e-05, "loss": 1.0315, "step": 2267 }, { "epoch": 0.08, "learning_rate": 1.989179969396617e-05, "loss": 1.0559, "step": 2268 }, { "epoch": 0.08, "learning_rate": 1.989164150361826e-05, "loss": 1.0374, "step": 2269 }, { "epoch": 0.08, "learning_rate": 1.9891483198346488e-05, "loss": 1.0497, "step": 2270 }, { "epoch": 0.08, "learning_rate": 1.9891324778152702e-05, "loss": 1.0564, "step": 2271 }, { "epoch": 0.08, "learning_rate": 1.9891166243038737e-05, "loss": 1.0431, "step": 2272 }, { "epoch": 0.08, "learning_rate": 1.989100759300644e-05, "loss": 1.0126, "step": 2273 }, { "epoch": 0.08, "learning_rate": 1.9890848828057648e-05, "loss": 1.0617, "step": 2274 }, { "epoch": 0.08, "learning_rate": 1.9890689948194213e-05, "loss": 1.0132, "step": 2275 }, { "epoch": 0.08, "learning_rate": 1.9890530953417976e-05, "loss": 0.9793, "step": 2276 }, { "epoch": 0.08, "learning_rate": 1.9890371843730782e-05, "loss": 1.0634, "step": 2277 }, { "epoch": 0.08, "learning_rate": 1.989021261913448e-05, "loss": 1.0209, "step": 2278 }, { "epoch": 0.08, "learning_rate": 1.989005327963093e-05, "loss": 1.0443, "step": 2279 }, { "epoch": 0.08, "learning_rate": 1.9889893825221972e-05, "loss": 1.0379, "step": 2280 }, { "epoch": 0.08, "learning_rate": 1.9889734255909463e-05, "loss": 1.0371, "step": 2281 }, { "epoch": 0.08, "learning_rate": 1.9889574571695254e-05, "loss": 1.0581, "step": 2282 }, { "epoch": 0.08, "learning_rate": 1.9889414772581202e-05, "loss": 1.0452, "step": 2283 }, { "epoch": 0.08, "learning_rate": 1.9889254858569167e-05, "loss": 1.0368, "step": 2284 }, { "epoch": 0.08, "learning_rate": 1.9889094829661006e-05, "loss": 1.0218, "step": 2285 }, { "epoch": 0.08, "learning_rate": 1.988893468585857e-05, "loss": 1.0362, "step": 2286 }, { "epoch": 0.08, "learning_rate": 1.988877442716373e-05, "loss": 1.0323, "step": 2287 }, { "epoch": 0.08, "learning_rate": 1.9888614053578347e-05, "loss": 1.0238, "step": 2288 }, { "epoch": 0.08, "learning_rate": 1.9888453565104274e-05, "loss": 0.9976, "step": 2289 }, { "epoch": 0.08, "learning_rate": 1.9888292961743386e-05, "loss": 1.0193, "step": 2290 }, { "epoch": 0.08, "learning_rate": 1.9888132243497545e-05, "loss": 1.0291, "step": 2291 }, { "epoch": 0.08, "learning_rate": 1.988797141036862e-05, "loss": 1.026, "step": 2292 }, { "epoch": 0.08, "learning_rate": 1.9887810462358478e-05, "loss": 1.0357, "step": 2293 }, { "epoch": 0.08, "learning_rate": 1.9887649399468988e-05, "loss": 1.0572, "step": 2294 }, { "epoch": 0.08, "learning_rate": 1.9887488221702024e-05, "loss": 1.0287, "step": 2295 }, { "epoch": 0.08, "learning_rate": 1.9887326929059458e-05, "loss": 1.0193, "step": 2296 }, { "epoch": 0.08, "learning_rate": 1.9887165521543162e-05, "loss": 1.0961, "step": 2297 }, { "epoch": 0.08, "learning_rate": 1.9887003999155012e-05, "loss": 1.0211, "step": 2298 }, { "epoch": 0.08, "learning_rate": 1.9886842361896886e-05, "loss": 1.0068, "step": 2299 }, { "epoch": 0.08, "learning_rate": 1.9886680609770663e-05, "loss": 1.0385, "step": 2300 }, { "epoch": 0.08, "learning_rate": 1.9886518742778217e-05, "loss": 1.0559, "step": 2301 }, { "epoch": 0.08, "learning_rate": 1.9886356760921433e-05, "loss": 1.0129, "step": 2302 }, { "epoch": 0.08, "learning_rate": 1.9886194664202193e-05, "loss": 1.1186, "step": 2303 }, { "epoch": 0.08, "learning_rate": 1.9886032452622375e-05, "loss": 0.9805, "step": 2304 }, { "epoch": 0.08, "learning_rate": 1.9885870126183874e-05, "loss": 1.0314, "step": 2305 }, { "epoch": 0.08, "learning_rate": 1.9885707684888565e-05, "loss": 1.0354, "step": 2306 }, { "epoch": 0.08, "learning_rate": 1.9885545128738343e-05, "loss": 1.0802, "step": 2307 }, { "epoch": 0.08, "learning_rate": 1.9885382457735095e-05, "loss": 1.034, "step": 2308 }, { "epoch": 0.08, "learning_rate": 1.9885219671880706e-05, "loss": 1.0483, "step": 2309 }, { "epoch": 0.08, "learning_rate": 1.9885056771177074e-05, "loss": 1.0204, "step": 2310 }, { "epoch": 0.08, "learning_rate": 1.9884893755626086e-05, "loss": 1.0495, "step": 2311 }, { "epoch": 0.08, "learning_rate": 1.9884730625229637e-05, "loss": 0.9778, "step": 2312 }, { "epoch": 0.08, "learning_rate": 1.9884567379989628e-05, "loss": 1.1313, "step": 2313 }, { "epoch": 0.08, "learning_rate": 1.988440401990795e-05, "loss": 1.0312, "step": 2314 }, { "epoch": 0.08, "learning_rate": 1.9884240544986502e-05, "loss": 1.0731, "step": 2315 }, { "epoch": 0.08, "learning_rate": 1.9884076955227182e-05, "loss": 1.0237, "step": 2316 }, { "epoch": 0.08, "learning_rate": 1.9883913250631896e-05, "loss": 1.0218, "step": 2317 }, { "epoch": 0.08, "learning_rate": 1.988374943120254e-05, "loss": 1.0594, "step": 2318 }, { "epoch": 0.08, "learning_rate": 1.988358549694102e-05, "loss": 1.0403, "step": 2319 }, { "epoch": 0.08, "learning_rate": 1.988342144784924e-05, "loss": 1.031, "step": 2320 }, { "epoch": 0.08, "learning_rate": 1.988325728392911e-05, "loss": 1.0353, "step": 2321 }, { "epoch": 0.08, "learning_rate": 1.9883093005182527e-05, "loss": 1.0529, "step": 2322 }, { "epoch": 0.08, "learning_rate": 1.9882928611611407e-05, "loss": 1.0383, "step": 2323 }, { "epoch": 0.08, "learning_rate": 1.9882764103217663e-05, "loss": 1.0209, "step": 2324 }, { "epoch": 0.08, "learning_rate": 1.98825994800032e-05, "loss": 1.0503, "step": 2325 }, { "epoch": 0.08, "learning_rate": 1.9882434741969936e-05, "loss": 1.0793, "step": 2326 }, { "epoch": 0.08, "learning_rate": 1.9882269889119778e-05, "loss": 1.0662, "step": 2327 }, { "epoch": 0.08, "learning_rate": 1.9882104921454646e-05, "loss": 0.9943, "step": 2328 }, { "epoch": 0.08, "learning_rate": 1.9881939838976458e-05, "loss": 1.0389, "step": 2329 }, { "epoch": 0.08, "learning_rate": 1.9881774641687128e-05, "loss": 1.0859, "step": 2330 }, { "epoch": 0.08, "learning_rate": 1.988160932958858e-05, "loss": 1.0486, "step": 2331 }, { "epoch": 0.08, "learning_rate": 1.988144390268273e-05, "loss": 1.0188, "step": 2332 }, { "epoch": 0.08, "learning_rate": 1.9881278360971502e-05, "loss": 1.0715, "step": 2333 }, { "epoch": 0.08, "learning_rate": 1.988111270445682e-05, "loss": 1.0185, "step": 2334 }, { "epoch": 0.08, "learning_rate": 1.9880946933140606e-05, "loss": 1.0416, "step": 2335 }, { "epoch": 0.08, "learning_rate": 1.9880781047024788e-05, "loss": 1.0271, "step": 2336 }, { "epoch": 0.08, "learning_rate": 1.9880615046111296e-05, "loss": 1.0339, "step": 2337 }, { "epoch": 0.08, "learning_rate": 1.988044893040205e-05, "loss": 1.0922, "step": 2338 }, { "epoch": 0.08, "learning_rate": 1.9880282699898992e-05, "loss": 1.0251, "step": 2339 }, { "epoch": 0.08, "learning_rate": 1.9880116354604045e-05, "loss": 0.9996, "step": 2340 }, { "epoch": 0.08, "learning_rate": 1.9879949894519142e-05, "loss": 1.0454, "step": 2341 }, { "epoch": 0.08, "learning_rate": 1.987978331964622e-05, "loss": 1.0473, "step": 2342 }, { "epoch": 0.08, "learning_rate": 1.987961662998721e-05, "loss": 1.0538, "step": 2343 }, { "epoch": 0.08, "learning_rate": 1.9879449825544056e-05, "loss": 1.0511, "step": 2344 }, { "epoch": 0.08, "learning_rate": 1.987928290631869e-05, "loss": 1.0585, "step": 2345 }, { "epoch": 0.08, "learning_rate": 1.9879115872313054e-05, "loss": 1.0652, "step": 2346 }, { "epoch": 0.08, "learning_rate": 1.9878948723529084e-05, "loss": 1.0721, "step": 2347 }, { "epoch": 0.08, "learning_rate": 1.987878145996873e-05, "loss": 1.0293, "step": 2348 }, { "epoch": 0.08, "learning_rate": 1.987861408163393e-05, "loss": 1.0426, "step": 2349 }, { "epoch": 0.08, "learning_rate": 1.987844658852663e-05, "loss": 1.022, "step": 2350 }, { "epoch": 0.08, "learning_rate": 1.987827898064877e-05, "loss": 1.0626, "step": 2351 }, { "epoch": 0.08, "learning_rate": 1.987811125800231e-05, "loss": 1.0207, "step": 2352 }, { "epoch": 0.08, "learning_rate": 1.9877943420589188e-05, "loss": 1.0295, "step": 2353 }, { "epoch": 0.08, "learning_rate": 1.987777546841136e-05, "loss": 1.0869, "step": 2354 }, { "epoch": 0.08, "learning_rate": 1.987760740147077e-05, "loss": 1.0398, "step": 2355 }, { "epoch": 0.08, "learning_rate": 1.987743921976938e-05, "loss": 1.0364, "step": 2356 }, { "epoch": 0.08, "learning_rate": 1.9877270923309136e-05, "loss": 1.073, "step": 2357 }, { "epoch": 0.08, "learning_rate": 1.9877102512092e-05, "loss": 1.0147, "step": 2358 }, { "epoch": 0.08, "learning_rate": 1.987693398611992e-05, "loss": 1.0263, "step": 2359 }, { "epoch": 0.08, "learning_rate": 1.9876765345394866e-05, "loss": 1.062, "step": 2360 }, { "epoch": 0.08, "learning_rate": 1.9876596589918787e-05, "loss": 1.064, "step": 2361 }, { "epoch": 0.08, "learning_rate": 1.9876427719693648e-05, "loss": 1.0731, "step": 2362 }, { "epoch": 0.08, "learning_rate": 1.987625873472141e-05, "loss": 1.05, "step": 2363 }, { "epoch": 0.08, "learning_rate": 1.9876089635004037e-05, "loss": 1.0025, "step": 2364 }, { "epoch": 0.08, "learning_rate": 1.987592042054349e-05, "loss": 1.0701, "step": 2365 }, { "epoch": 0.08, "learning_rate": 1.9875751091341742e-05, "loss": 1.0574, "step": 2366 }, { "epoch": 0.08, "learning_rate": 1.9875581647400757e-05, "loss": 1.0296, "step": 2367 }, { "epoch": 0.08, "learning_rate": 1.9875412088722502e-05, "loss": 1.0044, "step": 2368 }, { "epoch": 0.08, "learning_rate": 1.987524241530895e-05, "loss": 1.0465, "step": 2369 }, { "epoch": 0.08, "learning_rate": 1.9875072627162065e-05, "loss": 1.0226, "step": 2370 }, { "epoch": 0.08, "learning_rate": 1.987490272428383e-05, "loss": 1.0398, "step": 2371 }, { "epoch": 0.08, "learning_rate": 1.9874732706676216e-05, "loss": 0.98, "step": 2372 }, { "epoch": 0.08, "learning_rate": 1.987456257434119e-05, "loss": 1.0255, "step": 2373 }, { "epoch": 0.08, "learning_rate": 1.987439232728074e-05, "loss": 1.0557, "step": 2374 }, { "epoch": 0.08, "learning_rate": 1.987422196549684e-05, "loss": 1.0828, "step": 2375 }, { "epoch": 0.08, "learning_rate": 1.9874051488991462e-05, "loss": 1.0108, "step": 2376 }, { "epoch": 0.08, "learning_rate": 1.9873880897766597e-05, "loss": 1.0628, "step": 2377 }, { "epoch": 0.08, "learning_rate": 1.9873710191824224e-05, "loss": 1.0256, "step": 2378 }, { "epoch": 0.08, "learning_rate": 1.9873539371166323e-05, "loss": 1.0954, "step": 2379 }, { "epoch": 0.08, "learning_rate": 1.987336843579488e-05, "loss": 1.02, "step": 2380 }, { "epoch": 0.08, "learning_rate": 1.9873197385711885e-05, "loss": 1.0667, "step": 2381 }, { "epoch": 0.08, "learning_rate": 1.9873026220919324e-05, "loss": 1.056, "step": 2382 }, { "epoch": 0.08, "learning_rate": 1.987285494141918e-05, "loss": 1.0325, "step": 2383 }, { "epoch": 0.08, "learning_rate": 1.9872683547213446e-05, "loss": 1.0477, "step": 2384 }, { "epoch": 0.08, "learning_rate": 1.9872512038304117e-05, "loss": 1.0504, "step": 2385 }, { "epoch": 0.08, "learning_rate": 1.9872340414693184e-05, "loss": 1.0671, "step": 2386 }, { "epoch": 0.08, "learning_rate": 1.9872168676382634e-05, "loss": 1.0428, "step": 2387 }, { "epoch": 0.08, "learning_rate": 1.9871996823374475e-05, "loss": 1.053, "step": 2388 }, { "epoch": 0.08, "learning_rate": 1.9871824855670692e-05, "loss": 1.0687, "step": 2389 }, { "epoch": 0.08, "learning_rate": 1.9871652773273288e-05, "loss": 1.0687, "step": 2390 }, { "epoch": 0.08, "learning_rate": 1.9871480576184262e-05, "loss": 1.0369, "step": 2391 }, { "epoch": 0.08, "learning_rate": 1.9871308264405616e-05, "loss": 1.047, "step": 2392 }, { "epoch": 0.08, "learning_rate": 1.987113583793935e-05, "loss": 1.0673, "step": 2393 }, { "epoch": 0.08, "learning_rate": 1.987096329678747e-05, "loss": 1.0284, "step": 2394 }, { "epoch": 0.08, "learning_rate": 1.9870790640951975e-05, "loss": 1.0512, "step": 2395 }, { "epoch": 0.08, "learning_rate": 1.9870617870434878e-05, "loss": 1.0196, "step": 2396 }, { "epoch": 0.08, "learning_rate": 1.9870444985238176e-05, "loss": 1.1178, "step": 2397 }, { "epoch": 0.08, "learning_rate": 1.9870271985363893e-05, "loss": 1.0538, "step": 2398 }, { "epoch": 0.08, "learning_rate": 1.9870098870814023e-05, "loss": 1.0461, "step": 2399 }, { "epoch": 0.08, "learning_rate": 1.9869925641590587e-05, "loss": 1.0382, "step": 2400 }, { "epoch": 0.08, "learning_rate": 1.9869752297695597e-05, "loss": 1.056, "step": 2401 }, { "epoch": 0.08, "learning_rate": 1.9869578839131065e-05, "loss": 1.0238, "step": 2402 }, { "epoch": 0.08, "learning_rate": 1.9869405265899004e-05, "loss": 1.0072, "step": 2403 }, { "epoch": 0.08, "learning_rate": 1.9869231578001437e-05, "loss": 1.0116, "step": 2404 }, { "epoch": 0.08, "learning_rate": 1.9869057775440376e-05, "loss": 1.0389, "step": 2405 }, { "epoch": 0.08, "learning_rate": 1.986888385821784e-05, "loss": 1.0205, "step": 2406 }, { "epoch": 0.08, "learning_rate": 1.9868709826335856e-05, "loss": 1.0665, "step": 2407 }, { "epoch": 0.08, "learning_rate": 1.986853567979644e-05, "loss": 1.0014, "step": 2408 }, { "epoch": 0.08, "learning_rate": 1.9868361418601617e-05, "loss": 1.0191, "step": 2409 }, { "epoch": 0.08, "learning_rate": 1.9868187042753414e-05, "loss": 1.0553, "step": 2410 }, { "epoch": 0.08, "learning_rate": 1.9868012552253853e-05, "loss": 1.0682, "step": 2411 }, { "epoch": 0.08, "learning_rate": 1.986783794710496e-05, "loss": 1.0404, "step": 2412 }, { "epoch": 0.08, "learning_rate": 1.986766322730877e-05, "loss": 1.0652, "step": 2413 }, { "epoch": 0.08, "learning_rate": 1.986748839286731e-05, "loss": 1.0433, "step": 2414 }, { "epoch": 0.08, "learning_rate": 1.9867313443782612e-05, "loss": 1.0433, "step": 2415 }, { "epoch": 0.08, "learning_rate": 1.9867138380056706e-05, "loss": 1.0082, "step": 2416 }, { "epoch": 0.08, "learning_rate": 1.9866963201691626e-05, "loss": 1.0748, "step": 2417 }, { "epoch": 0.08, "learning_rate": 1.9866787908689413e-05, "loss": 1.0329, "step": 2418 }, { "epoch": 0.08, "learning_rate": 1.9866612501052094e-05, "loss": 1.0417, "step": 2419 }, { "epoch": 0.08, "learning_rate": 1.9866436978781714e-05, "loss": 1.0062, "step": 2420 }, { "epoch": 0.08, "learning_rate": 1.9866261341880312e-05, "loss": 1.0997, "step": 2421 }, { "epoch": 0.08, "learning_rate": 1.9866085590349928e-05, "loss": 1.0382, "step": 2422 }, { "epoch": 0.08, "learning_rate": 1.98659097241926e-05, "loss": 1.0236, "step": 2423 }, { "epoch": 0.08, "learning_rate": 1.9865733743410374e-05, "loss": 1.0434, "step": 2424 }, { "epoch": 0.08, "learning_rate": 1.9865557648005298e-05, "loss": 1.0414, "step": 2425 }, { "epoch": 0.08, "learning_rate": 1.986538143797941e-05, "loss": 1.0236, "step": 2426 }, { "epoch": 0.08, "learning_rate": 1.9865205113334768e-05, "loss": 1.0417, "step": 2427 }, { "epoch": 0.08, "learning_rate": 1.986502867407341e-05, "loss": 1.0576, "step": 2428 }, { "epoch": 0.08, "learning_rate": 1.9864852120197393e-05, "loss": 1.0425, "step": 2429 }, { "epoch": 0.08, "learning_rate": 1.9864675451708767e-05, "loss": 1.0324, "step": 2430 }, { "epoch": 0.08, "learning_rate": 1.986449866860958e-05, "loss": 1.0156, "step": 2431 }, { "epoch": 0.08, "learning_rate": 1.986432177090189e-05, "loss": 0.9885, "step": 2432 }, { "epoch": 0.08, "learning_rate": 1.9864144758587755e-05, "loss": 1.0581, "step": 2433 }, { "epoch": 0.08, "learning_rate": 1.9863967631669228e-05, "loss": 1.0417, "step": 2434 }, { "epoch": 0.08, "learning_rate": 1.9863790390148363e-05, "loss": 1.0736, "step": 2435 }, { "epoch": 0.08, "learning_rate": 1.9863613034027224e-05, "loss": 1.0006, "step": 2436 }, { "epoch": 0.08, "learning_rate": 1.9863435563307873e-05, "loss": 1.0648, "step": 2437 }, { "epoch": 0.08, "learning_rate": 1.9863257977992368e-05, "loss": 1.0085, "step": 2438 }, { "epoch": 0.08, "learning_rate": 1.9863080278082778e-05, "loss": 1.0341, "step": 2439 }, { "epoch": 0.08, "learning_rate": 1.986290246358116e-05, "loss": 1.0013, "step": 2440 }, { "epoch": 0.08, "learning_rate": 1.9862724534489585e-05, "loss": 1.0536, "step": 2441 }, { "epoch": 0.08, "learning_rate": 1.9862546490810115e-05, "loss": 1.015, "step": 2442 }, { "epoch": 0.08, "learning_rate": 1.9862368332544827e-05, "loss": 1.0571, "step": 2443 }, { "epoch": 0.08, "learning_rate": 1.9862190059695786e-05, "loss": 1.0635, "step": 2444 }, { "epoch": 0.08, "learning_rate": 1.986201167226506e-05, "loss": 1.0586, "step": 2445 }, { "epoch": 0.08, "learning_rate": 1.9861833170254727e-05, "loss": 1.0229, "step": 2446 }, { "epoch": 0.08, "learning_rate": 1.9861654553666858e-05, "loss": 1.044, "step": 2447 }, { "epoch": 0.08, "learning_rate": 1.9861475822503528e-05, "loss": 0.9877, "step": 2448 }, { "epoch": 0.08, "learning_rate": 1.986129697676682e-05, "loss": 1.062, "step": 2449 }, { "epoch": 0.08, "learning_rate": 1.98611180164588e-05, "loss": 1.0487, "step": 2450 }, { "epoch": 0.08, "learning_rate": 1.9860938941581555e-05, "loss": 1.0624, "step": 2451 }, { "epoch": 0.08, "learning_rate": 1.9860759752137165e-05, "loss": 0.9898, "step": 2452 }, { "epoch": 0.08, "learning_rate": 1.9860580448127713e-05, "loss": 1.074, "step": 2453 }, { "epoch": 0.08, "learning_rate": 1.986040102955528e-05, "loss": 1.0188, "step": 2454 }, { "epoch": 0.08, "learning_rate": 1.986022149642195e-05, "loss": 1.0574, "step": 2455 }, { "epoch": 0.08, "learning_rate": 1.986004184872981e-05, "loss": 1.0782, "step": 2456 }, { "epoch": 0.08, "learning_rate": 1.9859862086480945e-05, "loss": 1.0551, "step": 2457 }, { "epoch": 0.08, "learning_rate": 1.985968220967745e-05, "loss": 1.0707, "step": 2458 }, { "epoch": 0.08, "learning_rate": 1.9859502218321408e-05, "loss": 1.0405, "step": 2459 }, { "epoch": 0.08, "learning_rate": 1.985932211241491e-05, "loss": 1.0385, "step": 2460 }, { "epoch": 0.08, "learning_rate": 1.9859141891960056e-05, "loss": 1.067, "step": 2461 }, { "epoch": 0.08, "learning_rate": 1.985896155695893e-05, "loss": 1.0436, "step": 2462 }, { "epoch": 0.08, "learning_rate": 1.9858781107413636e-05, "loss": 1.0408, "step": 2463 }, { "epoch": 0.08, "learning_rate": 1.9858600543326262e-05, "loss": 1.0002, "step": 2464 }, { "epoch": 0.08, "learning_rate": 1.9858419864698912e-05, "loss": 1.0418, "step": 2465 }, { "epoch": 0.08, "learning_rate": 1.9858239071533683e-05, "loss": 1.0236, "step": 2466 }, { "epoch": 0.08, "learning_rate": 1.985805816383268e-05, "loss": 1.0402, "step": 2467 }, { "epoch": 0.08, "learning_rate": 1.9857877141597995e-05, "loss": 0.9782, "step": 2468 }, { "epoch": 0.08, "learning_rate": 1.9857696004831744e-05, "loss": 1.041, "step": 2469 }, { "epoch": 0.08, "learning_rate": 1.985751475353602e-05, "loss": 1.0048, "step": 2470 }, { "epoch": 0.08, "learning_rate": 1.9857333387712934e-05, "loss": 1.0709, "step": 2471 }, { "epoch": 0.08, "learning_rate": 1.9857151907364593e-05, "loss": 1.0185, "step": 2472 }, { "epoch": 0.08, "learning_rate": 1.9856970312493105e-05, "loss": 1.1126, "step": 2473 }, { "epoch": 0.08, "learning_rate": 1.985678860310058e-05, "loss": 1.0312, "step": 2474 }, { "epoch": 0.08, "learning_rate": 1.985660677918913e-05, "loss": 1.0347, "step": 2475 }, { "epoch": 0.08, "learning_rate": 1.9856424840760866e-05, "loss": 1.0456, "step": 2476 }, { "epoch": 0.08, "learning_rate": 1.98562427878179e-05, "loss": 1.0332, "step": 2477 }, { "epoch": 0.08, "learning_rate": 1.985606062036235e-05, "loss": 1.0563, "step": 2478 }, { "epoch": 0.08, "learning_rate": 1.9855878338396338e-05, "loss": 1.0271, "step": 2479 }, { "epoch": 0.08, "learning_rate": 1.985569594192197e-05, "loss": 1.0123, "step": 2480 }, { "epoch": 0.08, "learning_rate": 1.985551343094137e-05, "loss": 1.0427, "step": 2481 }, { "epoch": 0.08, "learning_rate": 1.9855330805456662e-05, "loss": 1.0679, "step": 2482 }, { "epoch": 0.08, "learning_rate": 1.9855148065469965e-05, "loss": 1.0642, "step": 2483 }, { "epoch": 0.08, "learning_rate": 1.98549652109834e-05, "loss": 1.0215, "step": 2484 }, { "epoch": 0.08, "learning_rate": 1.9854782241999098e-05, "loss": 1.0203, "step": 2485 }, { "epoch": 0.08, "learning_rate": 1.9854599158519175e-05, "loss": 1.0503, "step": 2486 }, { "epoch": 0.08, "learning_rate": 1.985441596054577e-05, "loss": 1.0458, "step": 2487 }, { "epoch": 0.08, "learning_rate": 1.9854232648081e-05, "loss": 1.0233, "step": 2488 }, { "epoch": 0.08, "learning_rate": 1.9854049221127e-05, "loss": 1.0487, "step": 2489 }, { "epoch": 0.08, "learning_rate": 1.9853865679685906e-05, "loss": 1.0366, "step": 2490 }, { "epoch": 0.08, "learning_rate": 1.985368202375984e-05, "loss": 1.041, "step": 2491 }, { "epoch": 0.08, "learning_rate": 1.9853498253350946e-05, "loss": 1.0062, "step": 2492 }, { "epoch": 0.08, "learning_rate": 1.9853314368461348e-05, "loss": 1.0769, "step": 2493 }, { "epoch": 0.08, "learning_rate": 1.985313036909319e-05, "loss": 1.0445, "step": 2494 }, { "epoch": 0.08, "learning_rate": 1.985294625524861e-05, "loss": 1.0614, "step": 2495 }, { "epoch": 0.08, "learning_rate": 1.985276202692975e-05, "loss": 1.0113, "step": 2496 }, { "epoch": 0.08, "learning_rate": 1.9852577684138737e-05, "loss": 1.0547, "step": 2497 }, { "epoch": 0.08, "learning_rate": 1.985239322687773e-05, "loss": 1.0469, "step": 2498 }, { "epoch": 0.08, "learning_rate": 1.985220865514886e-05, "loss": 1.081, "step": 2499 }, { "epoch": 0.08, "learning_rate": 1.9852023968954275e-05, "loss": 1.0003, "step": 2500 }, { "epoch": 0.08, "learning_rate": 1.985183916829612e-05, "loss": 1.0785, "step": 2501 }, { "epoch": 0.08, "learning_rate": 1.9851654253176545e-05, "loss": 1.0359, "step": 2502 }, { "epoch": 0.08, "learning_rate": 1.9851469223597694e-05, "loss": 1.0231, "step": 2503 }, { "epoch": 0.08, "learning_rate": 1.985128407956172e-05, "loss": 1.0198, "step": 2504 }, { "epoch": 0.08, "learning_rate": 1.9851098821070776e-05, "loss": 1.0905, "step": 2505 }, { "epoch": 0.08, "learning_rate": 1.9850913448127008e-05, "loss": 1.0045, "step": 2506 }, { "epoch": 0.08, "learning_rate": 1.9850727960732575e-05, "loss": 1.04, "step": 2507 }, { "epoch": 0.08, "learning_rate": 1.985054235888963e-05, "loss": 1.0367, "step": 2508 }, { "epoch": 0.08, "learning_rate": 1.985035664260033e-05, "loss": 1.0267, "step": 2509 }, { "epoch": 0.08, "learning_rate": 1.9850170811866833e-05, "loss": 1.0452, "step": 2510 }, { "epoch": 0.08, "learning_rate": 1.9849984866691295e-05, "loss": 1.0377, "step": 2511 }, { "epoch": 0.08, "learning_rate": 1.984979880707588e-05, "loss": 1.0014, "step": 2512 }, { "epoch": 0.08, "learning_rate": 1.9849612633022748e-05, "loss": 1.0231, "step": 2513 }, { "epoch": 0.08, "learning_rate": 1.984942634453406e-05, "loss": 1.0537, "step": 2514 }, { "epoch": 0.08, "learning_rate": 1.984923994161199e-05, "loss": 1.0533, "step": 2515 }, { "epoch": 0.08, "learning_rate": 1.9849053424258692e-05, "loss": 1.0409, "step": 2516 }, { "epoch": 0.08, "learning_rate": 1.9848866792476337e-05, "loss": 1.0232, "step": 2517 }, { "epoch": 0.08, "learning_rate": 1.9848680046267095e-05, "loss": 1.0242, "step": 2518 }, { "epoch": 0.08, "learning_rate": 1.9848493185633134e-05, "loss": 1.0355, "step": 2519 }, { "epoch": 0.08, "learning_rate": 1.9848306210576623e-05, "loss": 1.0162, "step": 2520 }, { "epoch": 0.08, "learning_rate": 1.984811912109974e-05, "loss": 1.071, "step": 2521 }, { "epoch": 0.08, "learning_rate": 1.9847931917204658e-05, "loss": 1.0122, "step": 2522 }, { "epoch": 0.08, "learning_rate": 1.9847744598893544e-05, "loss": 1.0504, "step": 2523 }, { "epoch": 0.08, "learning_rate": 1.9847557166168584e-05, "loss": 1.0042, "step": 2524 }, { "epoch": 0.08, "learning_rate": 1.984736961903195e-05, "loss": 1.0241, "step": 2525 }, { "epoch": 0.08, "learning_rate": 1.984718195748582e-05, "loss": 1.0477, "step": 2526 }, { "epoch": 0.08, "learning_rate": 1.984699418153238e-05, "loss": 1.037, "step": 2527 }, { "epoch": 0.08, "learning_rate": 1.9846806291173807e-05, "loss": 1.0384, "step": 2528 }, { "epoch": 0.08, "learning_rate": 1.9846618286412282e-05, "loss": 1.0496, "step": 2529 }, { "epoch": 0.08, "learning_rate": 1.984643016725e-05, "loss": 1.0548, "step": 2530 }, { "epoch": 0.08, "learning_rate": 1.9846241933689133e-05, "loss": 1.0729, "step": 2531 }, { "epoch": 0.08, "learning_rate": 1.9846053585731876e-05, "loss": 0.9952, "step": 2532 }, { "epoch": 0.08, "learning_rate": 1.984586512338042e-05, "loss": 1.0824, "step": 2533 }, { "epoch": 0.08, "learning_rate": 1.9845676546636944e-05, "loss": 1.0145, "step": 2534 }, { "epoch": 0.08, "learning_rate": 1.9845487855503645e-05, "loss": 1.0337, "step": 2535 }, { "epoch": 0.08, "learning_rate": 1.9845299049982718e-05, "loss": 1.0518, "step": 2536 }, { "epoch": 0.08, "learning_rate": 1.9845110130076354e-05, "loss": 1.0843, "step": 2537 }, { "epoch": 0.08, "learning_rate": 1.9844921095786746e-05, "loss": 1.0157, "step": 2538 }, { "epoch": 0.08, "learning_rate": 1.9844731947116095e-05, "loss": 1.0272, "step": 2539 }, { "epoch": 0.08, "learning_rate": 1.984454268406659e-05, "loss": 1.0394, "step": 2540 }, { "epoch": 0.08, "learning_rate": 1.9844353306640442e-05, "loss": 1.0708, "step": 2541 }, { "epoch": 0.08, "learning_rate": 1.984416381483984e-05, "loss": 1.0202, "step": 2542 }, { "epoch": 0.08, "learning_rate": 1.9843974208666994e-05, "loss": 1.0432, "step": 2543 }, { "epoch": 0.08, "learning_rate": 1.9843784488124102e-05, "loss": 0.9677, "step": 2544 }, { "epoch": 0.08, "learning_rate": 1.984359465321337e-05, "loss": 1.0488, "step": 2545 }, { "epoch": 0.08, "learning_rate": 1.9843404703937e-05, "loss": 1.0158, "step": 2546 }, { "epoch": 0.08, "learning_rate": 1.9843214640297208e-05, "loss": 1.0618, "step": 2547 }, { "epoch": 0.08, "learning_rate": 1.9843024462296192e-05, "loss": 1.0244, "step": 2548 }, { "epoch": 0.08, "learning_rate": 1.9842834169936163e-05, "loss": 1.0349, "step": 2549 }, { "epoch": 0.08, "learning_rate": 1.984264376321934e-05, "loss": 1.0416, "step": 2550 }, { "epoch": 0.08, "learning_rate": 1.9842453242147926e-05, "loss": 1.0213, "step": 2551 }, { "epoch": 0.08, "learning_rate": 1.984226260672414e-05, "loss": 0.9846, "step": 2552 }, { "epoch": 0.08, "learning_rate": 1.9842071856950197e-05, "loss": 1.0524, "step": 2553 }, { "epoch": 0.08, "learning_rate": 1.9841880992828306e-05, "loss": 1.0021, "step": 2554 }, { "epoch": 0.09, "learning_rate": 1.9841690014360694e-05, "loss": 1.0366, "step": 2555 }, { "epoch": 0.09, "learning_rate": 1.9841498921549574e-05, "loss": 0.9932, "step": 2556 }, { "epoch": 0.09, "learning_rate": 1.9841307714397168e-05, "loss": 1.0668, "step": 2557 }, { "epoch": 0.09, "learning_rate": 1.9841116392905698e-05, "loss": 1.0358, "step": 2558 }, { "epoch": 0.09, "learning_rate": 1.9840924957077385e-05, "loss": 1.0725, "step": 2559 }, { "epoch": 0.09, "learning_rate": 1.9840733406914453e-05, "loss": 1.0046, "step": 2560 }, { "epoch": 0.09, "learning_rate": 1.9840541742419137e-05, "loss": 1.0316, "step": 2561 }, { "epoch": 0.09, "learning_rate": 1.9840349963593647e-05, "loss": 1.0469, "step": 2562 }, { "epoch": 0.09, "learning_rate": 1.9840158070440224e-05, "loss": 1.0499, "step": 2563 }, { "epoch": 0.09, "learning_rate": 1.983996606296109e-05, "loss": 1.0395, "step": 2564 }, { "epoch": 0.09, "learning_rate": 1.983977394115848e-05, "loss": 1.0806, "step": 2565 }, { "epoch": 0.09, "learning_rate": 1.9839581705034626e-05, "loss": 1.051, "step": 2566 }, { "epoch": 0.09, "learning_rate": 1.9839389354591758e-05, "loss": 1.0301, "step": 2567 }, { "epoch": 0.09, "learning_rate": 1.9839196889832116e-05, "loss": 0.9747, "step": 2568 }, { "epoch": 0.09, "learning_rate": 1.9839004310757935e-05, "loss": 1.0618, "step": 2569 }, { "epoch": 0.09, "learning_rate": 1.983881161737145e-05, "loss": 1.0361, "step": 2570 }, { "epoch": 0.09, "learning_rate": 1.9838618809674897e-05, "loss": 1.0288, "step": 2571 }, { "epoch": 0.09, "learning_rate": 1.9838425887670523e-05, "loss": 1.0201, "step": 2572 }, { "epoch": 0.09, "learning_rate": 1.9838232851360567e-05, "loss": 1.086, "step": 2573 }, { "epoch": 0.09, "learning_rate": 1.983803970074727e-05, "loss": 1.0549, "step": 2574 }, { "epoch": 0.09, "learning_rate": 1.9837846435832876e-05, "loss": 1.0376, "step": 2575 }, { "epoch": 0.09, "learning_rate": 1.983765305661963e-05, "loss": 1.0203, "step": 2576 }, { "epoch": 0.09, "learning_rate": 1.9837459563109786e-05, "loss": 1.0596, "step": 2577 }, { "epoch": 0.09, "learning_rate": 1.983726595530558e-05, "loss": 1.014, "step": 2578 }, { "epoch": 0.09, "learning_rate": 1.9837072233209272e-05, "loss": 1.0464, "step": 2579 }, { "epoch": 0.09, "learning_rate": 1.9836878396823104e-05, "loss": 1.016, "step": 2580 }, { "epoch": 0.09, "learning_rate": 1.983668444614934e-05, "loss": 1.0883, "step": 2581 }, { "epoch": 0.09, "learning_rate": 1.9836490381190215e-05, "loss": 1.0418, "step": 2582 }, { "epoch": 0.09, "learning_rate": 1.9836296201948004e-05, "loss": 1.0769, "step": 2583 }, { "epoch": 0.09, "learning_rate": 1.9836101908424944e-05, "loss": 1.0188, "step": 2584 }, { "epoch": 0.09, "learning_rate": 1.983590750062331e-05, "loss": 1.0674, "step": 2585 }, { "epoch": 0.09, "learning_rate": 1.983571297854535e-05, "loss": 1.0816, "step": 2586 }, { "epoch": 0.09, "learning_rate": 1.9835518342193328e-05, "loss": 1.0472, "step": 2587 }, { "epoch": 0.09, "learning_rate": 1.98353235915695e-05, "loss": 1.0134, "step": 2588 }, { "epoch": 0.09, "learning_rate": 1.9835128726676138e-05, "loss": 1.0956, "step": 2589 }, { "epoch": 0.09, "learning_rate": 1.9834933747515496e-05, "loss": 0.995, "step": 2590 }, { "epoch": 0.09, "learning_rate": 1.9834738654089842e-05, "loss": 1.0292, "step": 2591 }, { "epoch": 0.09, "learning_rate": 1.983454344640145e-05, "loss": 1.0557, "step": 2592 }, { "epoch": 0.09, "learning_rate": 1.983434812445258e-05, "loss": 1.1052, "step": 2593 }, { "epoch": 0.09, "learning_rate": 1.9834152688245504e-05, "loss": 1.0411, "step": 2594 }, { "epoch": 0.09, "learning_rate": 1.9833957137782495e-05, "loss": 1.0275, "step": 2595 }, { "epoch": 0.09, "learning_rate": 1.9833761473065817e-05, "loss": 1.0779, "step": 2596 }, { "epoch": 0.09, "learning_rate": 1.983356569409775e-05, "loss": 1.0459, "step": 2597 }, { "epoch": 0.09, "learning_rate": 1.9833369800880572e-05, "loss": 1.0385, "step": 2598 }, { "epoch": 0.09, "learning_rate": 1.9833173793416552e-05, "loss": 1.0588, "step": 2599 }, { "epoch": 0.09, "learning_rate": 1.9832977671707968e-05, "loss": 1.0348, "step": 2600 }, { "epoch": 0.09, "learning_rate": 1.98327814357571e-05, "loss": 1.0209, "step": 2601 }, { "epoch": 0.09, "learning_rate": 1.9832585085566228e-05, "loss": 1.0271, "step": 2602 }, { "epoch": 0.09, "learning_rate": 1.9832388621137633e-05, "loss": 1.0722, "step": 2603 }, { "epoch": 0.09, "learning_rate": 1.98321920424736e-05, "loss": 1.0178, "step": 2604 }, { "epoch": 0.09, "learning_rate": 1.983199534957641e-05, "loss": 1.0425, "step": 2605 }, { "epoch": 0.09, "learning_rate": 1.983179854244835e-05, "loss": 1.0235, "step": 2606 }, { "epoch": 0.09, "learning_rate": 1.98316016210917e-05, "loss": 1.0803, "step": 2607 }, { "epoch": 0.09, "learning_rate": 1.983140458550876e-05, "loss": 1.0373, "step": 2608 }, { "epoch": 0.09, "learning_rate": 1.9831207435701808e-05, "loss": 1.0138, "step": 2609 }, { "epoch": 0.09, "learning_rate": 1.9831010171673146e-05, "loss": 1.0343, "step": 2610 }, { "epoch": 0.09, "learning_rate": 1.983081279342505e-05, "loss": 1.0289, "step": 2611 }, { "epoch": 0.09, "learning_rate": 1.983061530095983e-05, "loss": 1.0291, "step": 2612 }, { "epoch": 0.09, "learning_rate": 1.9830417694279766e-05, "loss": 1.078, "step": 2613 }, { "epoch": 0.09, "learning_rate": 1.9830219973387167e-05, "loss": 1.0528, "step": 2614 }, { "epoch": 0.09, "learning_rate": 1.983002213828432e-05, "loss": 1.0513, "step": 2615 }, { "epoch": 0.09, "learning_rate": 1.982982418897353e-05, "loss": 1.045, "step": 2616 }, { "epoch": 0.09, "learning_rate": 1.982962612545709e-05, "loss": 1.0655, "step": 2617 }, { "epoch": 0.09, "learning_rate": 1.982942794773731e-05, "loss": 1.0633, "step": 2618 }, { "epoch": 0.09, "learning_rate": 1.9829229655816485e-05, "loss": 1.0926, "step": 2619 }, { "epoch": 0.09, "learning_rate": 1.9829031249696918e-05, "loss": 0.9643, "step": 2620 }, { "epoch": 0.09, "learning_rate": 1.9828832729380925e-05, "loss": 1.0417, "step": 2621 }, { "epoch": 0.09, "learning_rate": 1.98286340948708e-05, "loss": 1.0315, "step": 2622 }, { "epoch": 0.09, "learning_rate": 1.982843534616886e-05, "loss": 1.0542, "step": 2623 }, { "epoch": 0.09, "learning_rate": 1.982823648327741e-05, "loss": 1.0216, "step": 2624 }, { "epoch": 0.09, "learning_rate": 1.9828037506198753e-05, "loss": 1.0421, "step": 2625 }, { "epoch": 0.09, "learning_rate": 1.9827838414935217e-05, "loss": 1.0416, "step": 2626 }, { "epoch": 0.09, "learning_rate": 1.9827639209489103e-05, "loss": 1.0179, "step": 2627 }, { "epoch": 0.09, "learning_rate": 1.982743988986273e-05, "loss": 1.0547, "step": 2628 }, { "epoch": 0.09, "learning_rate": 1.982724045605841e-05, "loss": 1.0586, "step": 2629 }, { "epoch": 0.09, "learning_rate": 1.9827040908078464e-05, "loss": 1.0329, "step": 2630 }, { "epoch": 0.09, "learning_rate": 1.982684124592521e-05, "loss": 1.0582, "step": 2631 }, { "epoch": 0.09, "learning_rate": 1.9826641469600968e-05, "loss": 1.0179, "step": 2632 }, { "epoch": 0.09, "learning_rate": 1.9826441579108057e-05, "loss": 1.0688, "step": 2633 }, { "epoch": 0.09, "learning_rate": 1.98262415744488e-05, "loss": 1.1096, "step": 2634 }, { "epoch": 0.09, "learning_rate": 1.982604145562552e-05, "loss": 1.0312, "step": 2635 }, { "epoch": 0.09, "learning_rate": 1.982584122264054e-05, "loss": 0.9847, "step": 2636 }, { "epoch": 0.09, "learning_rate": 1.9825640875496193e-05, "loss": 1.0137, "step": 2637 }, { "epoch": 0.09, "learning_rate": 1.9825440414194803e-05, "loss": 1.0476, "step": 2638 }, { "epoch": 0.09, "learning_rate": 1.98252398387387e-05, "loss": 0.9848, "step": 2639 }, { "epoch": 0.09, "learning_rate": 1.9825039149130214e-05, "loss": 1.0, "step": 2640 }, { "epoch": 0.09, "learning_rate": 1.9824838345371675e-05, "loss": 1.0165, "step": 2641 }, { "epoch": 0.09, "learning_rate": 1.9824637427465414e-05, "loss": 1.0549, "step": 2642 }, { "epoch": 0.09, "learning_rate": 1.982443639541377e-05, "loss": 1.0858, "step": 2643 }, { "epoch": 0.09, "learning_rate": 1.982423524921908e-05, "loss": 0.9867, "step": 2644 }, { "epoch": 0.09, "learning_rate": 1.9824033988883677e-05, "loss": 1.0614, "step": 2645 }, { "epoch": 0.09, "learning_rate": 1.98238326144099e-05, "loss": 1.0216, "step": 2646 }, { "epoch": 0.09, "learning_rate": 1.9823631125800086e-05, "loss": 1.0192, "step": 2647 }, { "epoch": 0.09, "learning_rate": 1.982342952305658e-05, "loss": 1.0084, "step": 2648 }, { "epoch": 0.09, "learning_rate": 1.9823227806181725e-05, "loss": 1.0455, "step": 2649 }, { "epoch": 0.09, "learning_rate": 1.9823025975177863e-05, "loss": 1.059, "step": 2650 }, { "epoch": 0.09, "learning_rate": 1.9822824030047337e-05, "loss": 1.0269, "step": 2651 }, { "epoch": 0.09, "learning_rate": 1.9822621970792494e-05, "loss": 0.9977, "step": 2652 }, { "epoch": 0.09, "learning_rate": 1.9822419797415686e-05, "loss": 1.0601, "step": 2653 }, { "epoch": 0.09, "learning_rate": 1.9822217509919256e-05, "loss": 0.9944, "step": 2654 }, { "epoch": 0.09, "learning_rate": 1.9822015108305558e-05, "loss": 1.0851, "step": 2655 }, { "epoch": 0.09, "learning_rate": 1.982181259257694e-05, "loss": 1.0305, "step": 2656 }, { "epoch": 0.09, "learning_rate": 1.982160996273576e-05, "loss": 1.052, "step": 2657 }, { "epoch": 0.09, "learning_rate": 1.9821407218784364e-05, "loss": 1.056, "step": 2658 }, { "epoch": 0.09, "learning_rate": 1.982120436072512e-05, "loss": 1.0673, "step": 2659 }, { "epoch": 0.09, "learning_rate": 1.9821001388560374e-05, "loss": 1.0128, "step": 2660 }, { "epoch": 0.09, "learning_rate": 1.982079830229249e-05, "loss": 1.0034, "step": 2661 }, { "epoch": 0.09, "learning_rate": 1.9820595101923824e-05, "loss": 1.066, "step": 2662 }, { "epoch": 0.09, "learning_rate": 1.9820391787456735e-05, "loss": 1.059, "step": 2663 }, { "epoch": 0.09, "learning_rate": 1.9820188358893593e-05, "loss": 0.9757, "step": 2664 }, { "epoch": 0.09, "learning_rate": 1.9819984816236753e-05, "loss": 1.0373, "step": 2665 }, { "epoch": 0.09, "learning_rate": 1.9819781159488587e-05, "loss": 1.0325, "step": 2666 }, { "epoch": 0.09, "learning_rate": 1.9819577388651455e-05, "loss": 1.0592, "step": 2667 }, { "epoch": 0.09, "learning_rate": 1.981937350372773e-05, "loss": 1.0111, "step": 2668 }, { "epoch": 0.09, "learning_rate": 1.9819169504719778e-05, "loss": 1.0725, "step": 2669 }, { "epoch": 0.09, "learning_rate": 1.981896539162997e-05, "loss": 1.0183, "step": 2670 }, { "epoch": 0.09, "learning_rate": 1.9818761164460673e-05, "loss": 1.0355, "step": 2671 }, { "epoch": 0.09, "learning_rate": 1.981855682321427e-05, "loss": 1.0145, "step": 2672 }, { "epoch": 0.09, "learning_rate": 1.981835236789312e-05, "loss": 1.0444, "step": 2673 }, { "epoch": 0.09, "learning_rate": 1.981814779849961e-05, "loss": 1.0498, "step": 2674 }, { "epoch": 0.09, "learning_rate": 1.9817943115036117e-05, "loss": 1.067, "step": 2675 }, { "epoch": 0.09, "learning_rate": 1.9817738317505014e-05, "loss": 1.0078, "step": 2676 }, { "epoch": 0.09, "learning_rate": 1.9817533405908683e-05, "loss": 1.0611, "step": 2677 }, { "epoch": 0.09, "learning_rate": 1.98173283802495e-05, "loss": 1.0342, "step": 2678 }, { "epoch": 0.09, "learning_rate": 1.9817123240529856e-05, "loss": 1.0181, "step": 2679 }, { "epoch": 0.09, "learning_rate": 1.9816917986752126e-05, "loss": 1.0069, "step": 2680 }, { "epoch": 0.09, "learning_rate": 1.98167126189187e-05, "loss": 1.0483, "step": 2681 }, { "epoch": 0.09, "learning_rate": 1.9816507137031962e-05, "loss": 1.0399, "step": 2682 }, { "epoch": 0.09, "learning_rate": 1.9816301541094298e-05, "loss": 1.0354, "step": 2683 }, { "epoch": 0.09, "learning_rate": 1.98160958311081e-05, "loss": 0.9924, "step": 2684 }, { "epoch": 0.09, "learning_rate": 1.9815890007075755e-05, "loss": 1.0492, "step": 2685 }, { "epoch": 0.09, "learning_rate": 1.9815684068999653e-05, "loss": 1.0788, "step": 2686 }, { "epoch": 0.09, "learning_rate": 1.981547801688219e-05, "loss": 1.0618, "step": 2687 }, { "epoch": 0.09, "learning_rate": 1.981527185072576e-05, "loss": 1.0292, "step": 2688 }, { "epoch": 0.09, "learning_rate": 1.9815065570532757e-05, "loss": 1.0094, "step": 2689 }, { "epoch": 0.09, "learning_rate": 1.9814859176305578e-05, "loss": 1.0314, "step": 2690 }, { "epoch": 0.09, "learning_rate": 1.9814652668046622e-05, "loss": 1.1147, "step": 2691 }, { "epoch": 0.09, "learning_rate": 1.9814446045758286e-05, "loss": 1.014, "step": 2692 }, { "epoch": 0.09, "learning_rate": 1.9814239309442973e-05, "loss": 1.0376, "step": 2693 }, { "epoch": 0.09, "learning_rate": 1.981403245910308e-05, "loss": 1.0178, "step": 2694 }, { "epoch": 0.09, "learning_rate": 1.9813825494741016e-05, "loss": 1.0467, "step": 2695 }, { "epoch": 0.09, "learning_rate": 1.9813618416359186e-05, "loss": 1.0031, "step": 2696 }, { "epoch": 0.09, "learning_rate": 1.981341122395999e-05, "loss": 1.0534, "step": 2697 }, { "epoch": 0.09, "learning_rate": 1.9813203917545838e-05, "loss": 1.0114, "step": 2698 }, { "epoch": 0.09, "learning_rate": 1.981299649711914e-05, "loss": 1.0223, "step": 2699 }, { "epoch": 0.09, "learning_rate": 1.9812788962682308e-05, "loss": 0.9969, "step": 2700 }, { "epoch": 0.09, "learning_rate": 1.9812581314237747e-05, "loss": 0.9864, "step": 2701 }, { "epoch": 0.09, "learning_rate": 1.9812373551787877e-05, "loss": 1.0488, "step": 2702 }, { "epoch": 0.09, "learning_rate": 1.98121656753351e-05, "loss": 1.0547, "step": 2703 }, { "epoch": 0.09, "learning_rate": 1.9811957684881844e-05, "loss": 1.0444, "step": 2704 }, { "epoch": 0.09, "learning_rate": 1.981174958043052e-05, "loss": 1.0446, "step": 2705 }, { "epoch": 0.09, "learning_rate": 1.9811541361983545e-05, "loss": 1.0559, "step": 2706 }, { "epoch": 0.09, "learning_rate": 1.981133302954334e-05, "loss": 1.0106, "step": 2707 }, { "epoch": 0.09, "learning_rate": 1.981112458311232e-05, "loss": 1.0089, "step": 2708 }, { "epoch": 0.09, "learning_rate": 1.9810916022692918e-05, "loss": 1.0624, "step": 2709 }, { "epoch": 0.09, "learning_rate": 1.9810707348287546e-05, "loss": 1.006, "step": 2710 }, { "epoch": 0.09, "learning_rate": 1.9810498559898635e-05, "loss": 1.052, "step": 2711 }, { "epoch": 0.09, "learning_rate": 1.9810289657528606e-05, "loss": 1.0034, "step": 2712 }, { "epoch": 0.09, "learning_rate": 1.9810080641179893e-05, "loss": 1.0555, "step": 2713 }, { "epoch": 0.09, "learning_rate": 1.9809871510854917e-05, "loss": 1.0663, "step": 2714 }, { "epoch": 0.09, "learning_rate": 1.9809662266556112e-05, "loss": 1.06, "step": 2715 }, { "epoch": 0.09, "learning_rate": 1.9809452908285907e-05, "loss": 1.0451, "step": 2716 }, { "epoch": 0.09, "learning_rate": 1.9809243436046738e-05, "loss": 1.0855, "step": 2717 }, { "epoch": 0.09, "learning_rate": 1.9809033849841034e-05, "loss": 1.0772, "step": 2718 }, { "epoch": 0.09, "learning_rate": 1.980882414967123e-05, "loss": 1.0403, "step": 2719 }, { "epoch": 0.09, "learning_rate": 1.980861433553977e-05, "loss": 1.0047, "step": 2720 }, { "epoch": 0.09, "learning_rate": 1.9808404407449085e-05, "loss": 1.0656, "step": 2721 }, { "epoch": 0.09, "learning_rate": 1.9808194365401612e-05, "loss": 1.0342, "step": 2722 }, { "epoch": 0.09, "learning_rate": 1.9807984209399796e-05, "loss": 1.0854, "step": 2723 }, { "epoch": 0.09, "learning_rate": 1.9807773939446078e-05, "loss": 1.015, "step": 2724 }, { "epoch": 0.09, "learning_rate": 1.9807563555542898e-05, "loss": 1.0595, "step": 2725 }, { "epoch": 0.09, "learning_rate": 1.9807353057692706e-05, "loss": 1.0238, "step": 2726 }, { "epoch": 0.09, "learning_rate": 1.980714244589794e-05, "loss": 1.0487, "step": 2727 }, { "epoch": 0.09, "learning_rate": 1.9806931720161052e-05, "loss": 0.9896, "step": 2728 }, { "epoch": 0.09, "learning_rate": 1.980672088048449e-05, "loss": 1.045, "step": 2729 }, { "epoch": 0.09, "learning_rate": 1.9806509926870704e-05, "loss": 1.0429, "step": 2730 }, { "epoch": 0.09, "learning_rate": 1.9806298859322143e-05, "loss": 1.0478, "step": 2731 }, { "epoch": 0.09, "learning_rate": 1.9806087677841258e-05, "loss": 0.9799, "step": 2732 }, { "epoch": 0.09, "learning_rate": 1.9805876382430508e-05, "loss": 1.0175, "step": 2733 }, { "epoch": 0.09, "learning_rate": 1.980566497309234e-05, "loss": 1.0237, "step": 2734 }, { "epoch": 0.09, "learning_rate": 1.9805453449829217e-05, "loss": 1.0095, "step": 2735 }, { "epoch": 0.09, "learning_rate": 1.9805241812643596e-05, "loss": 0.9888, "step": 2736 }, { "epoch": 0.09, "learning_rate": 1.9805030061537933e-05, "loss": 1.027, "step": 2737 }, { "epoch": 0.09, "learning_rate": 1.980481819651469e-05, "loss": 1.1067, "step": 2738 }, { "epoch": 0.09, "learning_rate": 1.9804606217576323e-05, "loss": 1.0163, "step": 2739 }, { "epoch": 0.09, "learning_rate": 1.9804394124725303e-05, "loss": 1.0225, "step": 2740 }, { "epoch": 0.09, "learning_rate": 1.980418191796409e-05, "loss": 1.0521, "step": 2741 }, { "epoch": 0.09, "learning_rate": 1.9803969597295153e-05, "loss": 1.037, "step": 2742 }, { "epoch": 0.09, "learning_rate": 1.9803757162720952e-05, "loss": 1.0111, "step": 2743 }, { "epoch": 0.09, "learning_rate": 1.9803544614243965e-05, "loss": 0.9758, "step": 2744 }, { "epoch": 0.09, "learning_rate": 1.980333195186665e-05, "loss": 1.0582, "step": 2745 }, { "epoch": 0.09, "learning_rate": 1.9803119175591484e-05, "loss": 1.0147, "step": 2746 }, { "epoch": 0.09, "learning_rate": 1.980290628542094e-05, "loss": 1.0959, "step": 2747 }, { "epoch": 0.09, "learning_rate": 1.9802693281357488e-05, "loss": 0.9798, "step": 2748 }, { "epoch": 0.09, "learning_rate": 1.9802480163403607e-05, "loss": 1.0185, "step": 2749 }, { "epoch": 0.09, "learning_rate": 1.980226693156177e-05, "loss": 1.0102, "step": 2750 }, { "epoch": 0.09, "learning_rate": 1.9802053585834454e-05, "loss": 1.0311, "step": 2751 }, { "epoch": 0.09, "learning_rate": 1.980184012622414e-05, "loss": 0.9986, "step": 2752 }, { "epoch": 0.09, "learning_rate": 1.9801626552733303e-05, "loss": 1.0401, "step": 2753 }, { "epoch": 0.09, "learning_rate": 1.9801412865364434e-05, "loss": 1.0506, "step": 2754 }, { "epoch": 0.09, "learning_rate": 1.9801199064120008e-05, "loss": 1.048, "step": 2755 }, { "epoch": 0.09, "learning_rate": 1.980098514900251e-05, "loss": 0.9828, "step": 2756 }, { "epoch": 0.09, "learning_rate": 1.9800771120014427e-05, "loss": 1.032, "step": 2757 }, { "epoch": 0.09, "learning_rate": 1.9800556977158243e-05, "loss": 1.0014, "step": 2758 }, { "epoch": 0.09, "learning_rate": 1.980034272043645e-05, "loss": 1.0693, "step": 2759 }, { "epoch": 0.09, "learning_rate": 1.9800128349851533e-05, "loss": 1.0179, "step": 2760 }, { "epoch": 0.09, "learning_rate": 1.9799913865405985e-05, "loss": 1.0521, "step": 2761 }, { "epoch": 0.09, "learning_rate": 1.97996992671023e-05, "loss": 1.0147, "step": 2762 }, { "epoch": 0.09, "learning_rate": 1.9799484554942965e-05, "loss": 1.1059, "step": 2763 }, { "epoch": 0.09, "learning_rate": 1.9799269728930483e-05, "loss": 0.9707, "step": 2764 }, { "epoch": 0.09, "learning_rate": 1.979905478906734e-05, "loss": 1.0672, "step": 2765 }, { "epoch": 0.09, "learning_rate": 1.9798839735356045e-05, "loss": 1.0427, "step": 2766 }, { "epoch": 0.09, "learning_rate": 1.9798624567799086e-05, "loss": 1.0781, "step": 2767 }, { "epoch": 0.09, "learning_rate": 1.9798409286398968e-05, "loss": 0.9843, "step": 2768 }, { "epoch": 0.09, "learning_rate": 1.9798193891158194e-05, "loss": 1.0409, "step": 2769 }, { "epoch": 0.09, "learning_rate": 1.9797978382079258e-05, "loss": 1.037, "step": 2770 }, { "epoch": 0.09, "learning_rate": 1.9797762759164675e-05, "loss": 1.045, "step": 2771 }, { "epoch": 0.09, "learning_rate": 1.979754702241694e-05, "loss": 1.0127, "step": 2772 }, { "epoch": 0.09, "learning_rate": 1.9797331171838566e-05, "loss": 1.077, "step": 2773 }, { "epoch": 0.09, "learning_rate": 1.9797115207432062e-05, "loss": 1.021, "step": 2774 }, { "epoch": 0.09, "learning_rate": 1.9796899129199932e-05, "loss": 1.0263, "step": 2775 }, { "epoch": 0.09, "learning_rate": 1.9796682937144687e-05, "loss": 1.0183, "step": 2776 }, { "epoch": 0.09, "learning_rate": 1.9796466631268844e-05, "loss": 1.0284, "step": 2777 }, { "epoch": 0.09, "learning_rate": 1.979625021157491e-05, "loss": 1.0582, "step": 2778 }, { "epoch": 0.09, "learning_rate": 1.9796033678065404e-05, "loss": 1.0704, "step": 2779 }, { "epoch": 0.09, "learning_rate": 1.9795817030742838e-05, "loss": 1.0644, "step": 2780 }, { "epoch": 0.09, "learning_rate": 1.979560026960973e-05, "loss": 1.0574, "step": 2781 }, { "epoch": 0.09, "learning_rate": 1.9795383394668602e-05, "loss": 1.0397, "step": 2782 }, { "epoch": 0.09, "learning_rate": 1.979516640592197e-05, "loss": 1.0106, "step": 2783 }, { "epoch": 0.09, "learning_rate": 1.9794949303372355e-05, "loss": 0.9664, "step": 2784 }, { "epoch": 0.09, "learning_rate": 1.979473208702228e-05, "loss": 1.0439, "step": 2785 }, { "epoch": 0.09, "learning_rate": 1.979451475687427e-05, "loss": 0.9754, "step": 2786 }, { "epoch": 0.09, "learning_rate": 1.979429731293085e-05, "loss": 1.0428, "step": 2787 }, { "epoch": 0.09, "learning_rate": 1.9794079755194545e-05, "loss": 0.9822, "step": 2788 }, { "epoch": 0.09, "learning_rate": 1.979386208366788e-05, "loss": 1.032, "step": 2789 }, { "epoch": 0.09, "learning_rate": 1.979364429835339e-05, "loss": 1.0256, "step": 2790 }, { "epoch": 0.09, "learning_rate": 1.97934263992536e-05, "loss": 1.0113, "step": 2791 }, { "epoch": 0.09, "learning_rate": 1.9793208386371047e-05, "loss": 1.0183, "step": 2792 }, { "epoch": 0.09, "learning_rate": 1.9792990259708258e-05, "loss": 1.0379, "step": 2793 }, { "epoch": 0.09, "learning_rate": 1.9792772019267768e-05, "loss": 1.0478, "step": 2794 }, { "epoch": 0.09, "learning_rate": 1.979255366505212e-05, "loss": 1.0098, "step": 2795 }, { "epoch": 0.09, "learning_rate": 1.9792335197063844e-05, "loss": 0.9752, "step": 2796 }, { "epoch": 0.09, "learning_rate": 1.9792116615305475e-05, "loss": 1.0492, "step": 2797 }, { "epoch": 0.09, "learning_rate": 1.9791897919779567e-05, "loss": 0.9813, "step": 2798 }, { "epoch": 0.09, "learning_rate": 1.9791679110488643e-05, "loss": 1.0211, "step": 2799 }, { "epoch": 0.09, "learning_rate": 1.9791460187435258e-05, "loss": 1.0152, "step": 2800 }, { "epoch": 0.09, "learning_rate": 1.9791241150621943e-05, "loss": 1.0418, "step": 2801 }, { "epoch": 0.09, "learning_rate": 1.979102200005126e-05, "loss": 1.028, "step": 2802 }, { "epoch": 0.09, "learning_rate": 1.979080273572574e-05, "loss": 1.0274, "step": 2803 }, { "epoch": 0.09, "learning_rate": 1.9790583357647936e-05, "loss": 1.0367, "step": 2804 }, { "epoch": 0.09, "learning_rate": 1.97903638658204e-05, "loss": 1.0587, "step": 2805 }, { "epoch": 0.09, "learning_rate": 1.979014426024568e-05, "loss": 1.0135, "step": 2806 }, { "epoch": 0.09, "learning_rate": 1.9789924540926323e-05, "loss": 1.0147, "step": 2807 }, { "epoch": 0.09, "learning_rate": 1.978970470786489e-05, "loss": 0.9727, "step": 2808 }, { "epoch": 0.09, "learning_rate": 1.9789484761063925e-05, "loss": 1.0522, "step": 2809 }, { "epoch": 0.09, "learning_rate": 1.9789264700525994e-05, "loss": 1.0337, "step": 2810 }, { "epoch": 0.09, "learning_rate": 1.9789044526253645e-05, "loss": 1.0113, "step": 2811 }, { "epoch": 0.09, "learning_rate": 1.978882423824944e-05, "loss": 1.018, "step": 2812 }, { "epoch": 0.09, "learning_rate": 1.9788603836515943e-05, "loss": 1.0445, "step": 2813 }, { "epoch": 0.09, "learning_rate": 1.9788383321055704e-05, "loss": 1.0064, "step": 2814 }, { "epoch": 0.09, "learning_rate": 1.9788162691871295e-05, "loss": 1.0547, "step": 2815 }, { "epoch": 0.09, "learning_rate": 1.978794194896527e-05, "loss": 1.0307, "step": 2816 }, { "epoch": 0.09, "learning_rate": 1.9787721092340202e-05, "loss": 1.0308, "step": 2817 }, { "epoch": 0.09, "learning_rate": 1.9787500121998657e-05, "loss": 1.0216, "step": 2818 }, { "epoch": 0.09, "learning_rate": 1.9787279037943192e-05, "loss": 1.039, "step": 2819 }, { "epoch": 0.09, "learning_rate": 1.9787057840176386e-05, "loss": 1.0132, "step": 2820 }, { "epoch": 0.09, "learning_rate": 1.9786836528700805e-05, "loss": 1.0213, "step": 2821 }, { "epoch": 0.09, "learning_rate": 1.978661510351902e-05, "loss": 1.0008, "step": 2822 }, { "epoch": 0.09, "learning_rate": 1.9786393564633604e-05, "loss": 1.0274, "step": 2823 }, { "epoch": 0.09, "learning_rate": 1.9786171912047135e-05, "loss": 1.0123, "step": 2824 }, { "epoch": 0.09, "learning_rate": 1.9785950145762183e-05, "loss": 1.0293, "step": 2825 }, { "epoch": 0.09, "learning_rate": 1.9785728265781327e-05, "loss": 1.0656, "step": 2826 }, { "epoch": 0.09, "learning_rate": 1.978550627210714e-05, "loss": 1.0267, "step": 2827 }, { "epoch": 0.09, "learning_rate": 1.9785284164742207e-05, "loss": 1.0375, "step": 2828 }, { "epoch": 0.09, "learning_rate": 1.9785061943689108e-05, "loss": 1.0223, "step": 2829 }, { "epoch": 0.09, "learning_rate": 1.978483960895042e-05, "loss": 1.029, "step": 2830 }, { "epoch": 0.09, "learning_rate": 1.9784617160528736e-05, "loss": 1.0584, "step": 2831 }, { "epoch": 0.09, "learning_rate": 1.9784394598426626e-05, "loss": 0.9812, "step": 2832 }, { "epoch": 0.09, "learning_rate": 1.9784171922646687e-05, "loss": 1.0451, "step": 2833 }, { "epoch": 0.09, "learning_rate": 1.9783949133191506e-05, "loss": 1.0351, "step": 2834 }, { "epoch": 0.09, "learning_rate": 1.9783726230063665e-05, "loss": 1.0404, "step": 2835 }, { "epoch": 0.09, "learning_rate": 1.978350321326576e-05, "loss": 0.9911, "step": 2836 }, { "epoch": 0.09, "learning_rate": 1.9783280082800377e-05, "loss": 1.0413, "step": 2837 }, { "epoch": 0.09, "learning_rate": 1.978305683867011e-05, "loss": 1.0507, "step": 2838 }, { "epoch": 0.09, "learning_rate": 1.9782833480877555e-05, "loss": 1.0307, "step": 2839 }, { "epoch": 0.09, "learning_rate": 1.9782610009425305e-05, "loss": 1.0198, "step": 2840 }, { "epoch": 0.09, "learning_rate": 1.9782386424315955e-05, "loss": 1.0247, "step": 2841 }, { "epoch": 0.09, "learning_rate": 1.9782162725552107e-05, "loss": 1.0242, "step": 2842 }, { "epoch": 0.09, "learning_rate": 1.9781938913136352e-05, "loss": 1.0305, "step": 2843 }, { "epoch": 0.09, "learning_rate": 1.97817149870713e-05, "loss": 1.0048, "step": 2844 }, { "epoch": 0.09, "learning_rate": 1.978149094735955e-05, "loss": 1.0602, "step": 2845 }, { "epoch": 0.09, "learning_rate": 1.97812667940037e-05, "loss": 1.0084, "step": 2846 }, { "epoch": 0.09, "learning_rate": 1.9781042527006357e-05, "loss": 1.0756, "step": 2847 }, { "epoch": 0.09, "learning_rate": 1.978081814637013e-05, "loss": 0.9993, "step": 2848 }, { "epoch": 0.09, "learning_rate": 1.978059365209762e-05, "loss": 1.0331, "step": 2849 }, { "epoch": 0.09, "learning_rate": 1.9780369044191443e-05, "loss": 1.0228, "step": 2850 }, { "epoch": 0.09, "learning_rate": 1.97801443226542e-05, "loss": 1.0201, "step": 2851 }, { "epoch": 0.09, "learning_rate": 1.9779919487488507e-05, "loss": 1.0038, "step": 2852 }, { "epoch": 0.09, "learning_rate": 1.9779694538696973e-05, "loss": 1.0251, "step": 2853 }, { "epoch": 0.09, "learning_rate": 1.9779469476282216e-05, "loss": 1.024, "step": 2854 }, { "epoch": 0.1, "learning_rate": 1.9779244300246846e-05, "loss": 1.0303, "step": 2855 }, { "epoch": 0.1, "learning_rate": 1.9779019010593485e-05, "loss": 1.0122, "step": 2856 }, { "epoch": 0.1, "learning_rate": 1.9778793607324743e-05, "loss": 1.0041, "step": 2857 }, { "epoch": 0.1, "learning_rate": 1.9778568090443247e-05, "loss": 1.0287, "step": 2858 }, { "epoch": 0.1, "learning_rate": 1.977834245995161e-05, "loss": 1.0411, "step": 2859 }, { "epoch": 0.1, "learning_rate": 1.9778116715852457e-05, "loss": 1.0026, "step": 2860 }, { "epoch": 0.1, "learning_rate": 1.9777890858148407e-05, "loss": 1.0291, "step": 2861 }, { "epoch": 0.1, "learning_rate": 1.9777664886842093e-05, "loss": 1.0277, "step": 2862 }, { "epoch": 0.1, "learning_rate": 1.977743880193613e-05, "loss": 1.029, "step": 2863 }, { "epoch": 0.1, "learning_rate": 1.977721260343315e-05, "loss": 0.9766, "step": 2864 }, { "epoch": 0.1, "learning_rate": 1.9776986291335782e-05, "loss": 1.0574, "step": 2865 }, { "epoch": 0.1, "learning_rate": 1.977675986564665e-05, "loss": 1.0473, "step": 2866 }, { "epoch": 0.1, "learning_rate": 1.977653332636839e-05, "loss": 1.0452, "step": 2867 }, { "epoch": 0.1, "learning_rate": 1.977630667350363e-05, "loss": 0.9898, "step": 2868 }, { "epoch": 0.1, "learning_rate": 1.977607990705501e-05, "loss": 1.0337, "step": 2869 }, { "epoch": 0.1, "learning_rate": 1.9775853027025162e-05, "loss": 1.0486, "step": 2870 }, { "epoch": 0.1, "learning_rate": 1.9775626033416712e-05, "loss": 1.0424, "step": 2871 }, { "epoch": 0.1, "learning_rate": 1.977539892623231e-05, "loss": 1.0168, "step": 2872 }, { "epoch": 0.1, "learning_rate": 1.9775171705474593e-05, "loss": 1.0324, "step": 2873 }, { "epoch": 0.1, "learning_rate": 1.9774944371146195e-05, "loss": 1.0275, "step": 2874 }, { "epoch": 0.1, "learning_rate": 1.977471692324976e-05, "loss": 1.0685, "step": 2875 }, { "epoch": 0.1, "learning_rate": 1.9774489361787933e-05, "loss": 1.025, "step": 2876 }, { "epoch": 0.1, "learning_rate": 1.977426168676335e-05, "loss": 1.0309, "step": 2877 }, { "epoch": 0.1, "learning_rate": 1.9774033898178668e-05, "loss": 1.0529, "step": 2878 }, { "epoch": 0.1, "learning_rate": 1.9773805996036523e-05, "loss": 1.0218, "step": 2879 }, { "epoch": 0.1, "learning_rate": 1.9773577980339572e-05, "loss": 1.0288, "step": 2880 }, { "epoch": 0.1, "learning_rate": 1.9773349851090455e-05, "loss": 1.0427, "step": 2881 }, { "epoch": 0.1, "learning_rate": 1.977312160829183e-05, "loss": 1.0706, "step": 2882 }, { "epoch": 0.1, "learning_rate": 1.9772893251946344e-05, "loss": 1.039, "step": 2883 }, { "epoch": 0.1, "learning_rate": 1.9772664782056653e-05, "loss": 1.0079, "step": 2884 }, { "epoch": 0.1, "learning_rate": 1.9772436198625407e-05, "loss": 1.0543, "step": 2885 }, { "epoch": 0.1, "learning_rate": 1.9772207501655266e-05, "loss": 1.0078, "step": 2886 }, { "epoch": 0.1, "learning_rate": 1.9771978691148886e-05, "loss": 1.0456, "step": 2887 }, { "epoch": 0.1, "learning_rate": 1.9771749767108924e-05, "loss": 1.0333, "step": 2888 }, { "epoch": 0.1, "learning_rate": 1.9771520729538046e-05, "loss": 1.0163, "step": 2889 }, { "epoch": 0.1, "learning_rate": 1.97712915784389e-05, "loss": 1.0467, "step": 2890 }, { "epoch": 0.1, "learning_rate": 1.9771062313814164e-05, "loss": 1.0422, "step": 2891 }, { "epoch": 0.1, "learning_rate": 1.9770832935666493e-05, "loss": 0.9986, "step": 2892 }, { "epoch": 0.1, "learning_rate": 1.9770603443998552e-05, "loss": 1.0833, "step": 2893 }, { "epoch": 0.1, "learning_rate": 1.9770373838813007e-05, "loss": 1.0264, "step": 2894 }, { "epoch": 0.1, "learning_rate": 1.9770144120112527e-05, "loss": 1.0293, "step": 2895 }, { "epoch": 0.1, "learning_rate": 1.976991428789978e-05, "loss": 0.9852, "step": 2896 }, { "epoch": 0.1, "learning_rate": 1.976968434217744e-05, "loss": 1.0302, "step": 2897 }, { "epoch": 0.1, "learning_rate": 1.9769454282948173e-05, "loss": 1.0407, "step": 2898 }, { "epoch": 0.1, "learning_rate": 1.9769224110214656e-05, "loss": 1.016, "step": 2899 }, { "epoch": 0.1, "learning_rate": 1.976899382397956e-05, "loss": 1.015, "step": 2900 }, { "epoch": 0.1, "learning_rate": 1.9768763424245564e-05, "loss": 1.0334, "step": 2901 }, { "epoch": 0.1, "learning_rate": 1.9768532911015342e-05, "loss": 1.0201, "step": 2902 }, { "epoch": 0.1, "learning_rate": 1.9768302284291576e-05, "loss": 0.9974, "step": 2903 }, { "epoch": 0.1, "learning_rate": 1.9768071544076942e-05, "loss": 0.9793, "step": 2904 }, { "epoch": 0.1, "learning_rate": 1.976784069037412e-05, "loss": 1.0568, "step": 2905 }, { "epoch": 0.1, "learning_rate": 1.9767609723185795e-05, "loss": 0.9831, "step": 2906 }, { "epoch": 0.1, "learning_rate": 1.976737864251465e-05, "loss": 1.0545, "step": 2907 }, { "epoch": 0.1, "learning_rate": 1.9767147448363366e-05, "loss": 1.0483, "step": 2908 }, { "epoch": 0.1, "learning_rate": 1.9766916140734635e-05, "loss": 1.032, "step": 2909 }, { "epoch": 0.1, "learning_rate": 1.9766684719631143e-05, "loss": 1.0285, "step": 2910 }, { "epoch": 0.1, "learning_rate": 1.9766453185055573e-05, "loss": 1.0436, "step": 2911 }, { "epoch": 0.1, "learning_rate": 1.976622153701062e-05, "loss": 1.0009, "step": 2912 }, { "epoch": 0.1, "learning_rate": 1.9765989775498975e-05, "loss": 1.0084, "step": 2913 }, { "epoch": 0.1, "learning_rate": 1.976575790052333e-05, "loss": 1.0365, "step": 2914 }, { "epoch": 0.1, "learning_rate": 1.976552591208638e-05, "loss": 1.0625, "step": 2915 }, { "epoch": 0.1, "learning_rate": 1.976529381019082e-05, "loss": 1.0103, "step": 2916 }, { "epoch": 0.1, "learning_rate": 1.9765061594839345e-05, "loss": 1.0378, "step": 2917 }, { "epoch": 0.1, "learning_rate": 1.976482926603465e-05, "loss": 1.0264, "step": 2918 }, { "epoch": 0.1, "learning_rate": 1.9764596823779446e-05, "loss": 1.0775, "step": 2919 }, { "epoch": 0.1, "learning_rate": 1.9764364268076422e-05, "loss": 0.9746, "step": 2920 }, { "epoch": 0.1, "learning_rate": 1.9764131598928284e-05, "loss": 1.0343, "step": 2921 }, { "epoch": 0.1, "learning_rate": 1.9763898816337734e-05, "loss": 1.047, "step": 2922 }, { "epoch": 0.1, "learning_rate": 1.9763665920307483e-05, "loss": 1.0426, "step": 2923 }, { "epoch": 0.1, "learning_rate": 1.9763432910840222e-05, "loss": 1.0083, "step": 2924 }, { "epoch": 0.1, "learning_rate": 1.9763199787938675e-05, "loss": 1.042, "step": 2925 }, { "epoch": 0.1, "learning_rate": 1.976296655160554e-05, "loss": 0.9996, "step": 2926 }, { "epoch": 0.1, "learning_rate": 1.9762733201843533e-05, "loss": 0.9994, "step": 2927 }, { "epoch": 0.1, "learning_rate": 1.976249973865536e-05, "loss": 0.9777, "step": 2928 }, { "epoch": 0.1, "learning_rate": 1.9762266162043734e-05, "loss": 1.0645, "step": 2929 }, { "epoch": 0.1, "learning_rate": 1.9762032472011372e-05, "loss": 1.0069, "step": 2930 }, { "epoch": 0.1, "learning_rate": 1.9761798668560987e-05, "loss": 1.061, "step": 2931 }, { "epoch": 0.1, "learning_rate": 1.9761564751695295e-05, "loss": 0.9991, "step": 2932 }, { "epoch": 0.1, "learning_rate": 1.9761330721417015e-05, "loss": 1.0085, "step": 2933 }, { "epoch": 0.1, "learning_rate": 1.9761096577728863e-05, "loss": 1.0388, "step": 2934 }, { "epoch": 0.1, "learning_rate": 1.9760862320633563e-05, "loss": 0.9736, "step": 2935 }, { "epoch": 0.1, "learning_rate": 1.976062795013384e-05, "loss": 1.0102, "step": 2936 }, { "epoch": 0.1, "learning_rate": 1.9760393466232405e-05, "loss": 1.0543, "step": 2937 }, { "epoch": 0.1, "learning_rate": 1.9760158868931994e-05, "loss": 1.0388, "step": 2938 }, { "epoch": 0.1, "learning_rate": 1.975992415823533e-05, "loss": 1.0195, "step": 2939 }, { "epoch": 0.1, "learning_rate": 1.9759689334145133e-05, "loss": 1.0129, "step": 2940 }, { "epoch": 0.1, "learning_rate": 1.975945439666414e-05, "loss": 0.9883, "step": 2941 }, { "epoch": 0.1, "learning_rate": 1.975921934579508e-05, "loss": 1.0251, "step": 2942 }, { "epoch": 0.1, "learning_rate": 1.9758984181540673e-05, "loss": 1.014, "step": 2943 }, { "epoch": 0.1, "learning_rate": 1.9758748903903667e-05, "loss": 1.0218, "step": 2944 }, { "epoch": 0.1, "learning_rate": 1.9758513512886784e-05, "loss": 0.9928, "step": 2945 }, { "epoch": 0.1, "learning_rate": 1.9758278008492762e-05, "loss": 1.0036, "step": 2946 }, { "epoch": 0.1, "learning_rate": 1.9758042390724336e-05, "loss": 1.0524, "step": 2947 }, { "epoch": 0.1, "learning_rate": 1.975780665958425e-05, "loss": 0.9969, "step": 2948 }, { "epoch": 0.1, "learning_rate": 1.9757570815075234e-05, "loss": 0.987, "step": 2949 }, { "epoch": 0.1, "learning_rate": 1.9757334857200034e-05, "loss": 1.0514, "step": 2950 }, { "epoch": 0.1, "learning_rate": 1.975709878596139e-05, "loss": 1.0149, "step": 2951 }, { "epoch": 0.1, "learning_rate": 1.9756862601362042e-05, "loss": 1.0037, "step": 2952 }, { "epoch": 0.1, "learning_rate": 1.975662630340474e-05, "loss": 1.0048, "step": 2953 }, { "epoch": 0.1, "learning_rate": 1.975638989209222e-05, "loss": 1.0105, "step": 2954 }, { "epoch": 0.1, "learning_rate": 1.9756153367427236e-05, "loss": 1.0203, "step": 2955 }, { "epoch": 0.1, "learning_rate": 1.9755916729412538e-05, "loss": 0.9805, "step": 2956 }, { "epoch": 0.1, "learning_rate": 1.975567997805087e-05, "loss": 1.0435, "step": 2957 }, { "epoch": 0.1, "learning_rate": 1.9755443113344984e-05, "loss": 1.028, "step": 2958 }, { "epoch": 0.1, "learning_rate": 1.975520613529763e-05, "loss": 1.0627, "step": 2959 }, { "epoch": 0.1, "learning_rate": 1.975496904391157e-05, "loss": 0.9983, "step": 2960 }, { "epoch": 0.1, "learning_rate": 1.9754731839189545e-05, "loss": 1.0298, "step": 2961 }, { "epoch": 0.1, "learning_rate": 1.975449452113432e-05, "loss": 1.058, "step": 2962 }, { "epoch": 0.1, "learning_rate": 1.975425708974865e-05, "loss": 1.0058, "step": 2963 }, { "epoch": 0.1, "learning_rate": 1.9754019545035296e-05, "loss": 0.9891, "step": 2964 }, { "epoch": 0.1, "learning_rate": 1.9753781886997014e-05, "loss": 1.0295, "step": 2965 }, { "epoch": 0.1, "learning_rate": 1.975354411563657e-05, "loss": 1.0504, "step": 2966 }, { "epoch": 0.1, "learning_rate": 1.975330623095672e-05, "loss": 1.0091, "step": 2967 }, { "epoch": 0.1, "learning_rate": 1.9753068232960232e-05, "loss": 0.9792, "step": 2968 }, { "epoch": 0.1, "learning_rate": 1.975283012164987e-05, "loss": 1.0168, "step": 2969 }, { "epoch": 0.1, "learning_rate": 1.9752591897028403e-05, "loss": 1.039, "step": 2970 }, { "epoch": 0.1, "learning_rate": 1.975235355909859e-05, "loss": 1.0445, "step": 2971 }, { "epoch": 0.1, "learning_rate": 1.9752115107863213e-05, "loss": 0.995, "step": 2972 }, { "epoch": 0.1, "learning_rate": 1.9751876543325036e-05, "loss": 1.0076, "step": 2973 }, { "epoch": 0.1, "learning_rate": 1.9751637865486824e-05, "loss": 1.0088, "step": 2974 }, { "epoch": 0.1, "learning_rate": 1.9751399074351366e-05, "loss": 1.0175, "step": 2975 }, { "epoch": 0.1, "learning_rate": 1.9751160169921418e-05, "loss": 0.9943, "step": 2976 }, { "epoch": 0.1, "learning_rate": 1.9750921152199774e-05, "loss": 1.0176, "step": 2977 }, { "epoch": 0.1, "learning_rate": 1.9750682021189195e-05, "loss": 0.9667, "step": 2978 }, { "epoch": 0.1, "learning_rate": 1.9750442776892468e-05, "loss": 0.9983, "step": 2979 }, { "epoch": 0.1, "learning_rate": 1.975020341931237e-05, "loss": 1.0191, "step": 2980 }, { "epoch": 0.1, "learning_rate": 1.9749963948451683e-05, "loss": 1.0491, "step": 2981 }, { "epoch": 0.1, "learning_rate": 1.974972436431319e-05, "loss": 1.0346, "step": 2982 }, { "epoch": 0.1, "learning_rate": 1.9749484666899672e-05, "loss": 1.007, "step": 2983 }, { "epoch": 0.1, "learning_rate": 1.9749244856213915e-05, "loss": 1.0284, "step": 2984 }, { "epoch": 0.1, "learning_rate": 1.9749004932258703e-05, "loss": 1.0262, "step": 2985 }, { "epoch": 0.1, "learning_rate": 1.9748764895036833e-05, "loss": 1.0043, "step": 2986 }, { "epoch": 0.1, "learning_rate": 1.974852474455108e-05, "loss": 1.058, "step": 2987 }, { "epoch": 0.1, "learning_rate": 1.9748284480804244e-05, "loss": 1.0041, "step": 2988 }, { "epoch": 0.1, "learning_rate": 1.9748044103799112e-05, "loss": 1.068, "step": 2989 }, { "epoch": 0.1, "learning_rate": 1.9747803613538475e-05, "loss": 0.9992, "step": 2990 }, { "epoch": 0.1, "learning_rate": 1.9747563010025136e-05, "loss": 1.0421, "step": 2991 }, { "epoch": 0.1, "learning_rate": 1.974732229326188e-05, "loss": 0.9633, "step": 2992 }, { "epoch": 0.1, "learning_rate": 1.974708146325151e-05, "loss": 1.0501, "step": 2993 }, { "epoch": 0.1, "learning_rate": 1.9746840519996824e-05, "loss": 1.0309, "step": 2994 }, { "epoch": 0.1, "learning_rate": 1.974659946350062e-05, "loss": 1.0334, "step": 2995 }, { "epoch": 0.1, "learning_rate": 1.9746358293765696e-05, "loss": 1.0134, "step": 2996 }, { "epoch": 0.1, "learning_rate": 1.9746117010794853e-05, "loss": 1.0327, "step": 2997 }, { "epoch": 0.1, "learning_rate": 1.9745875614590903e-05, "loss": 1.0457, "step": 2998 }, { "epoch": 0.1, "learning_rate": 1.9745634105156642e-05, "loss": 0.9725, "step": 2999 }, { "epoch": 0.1, "learning_rate": 1.974539248249488e-05, "loss": 1.0144, "step": 3000 }, { "epoch": 0.1, "learning_rate": 1.9745150746608425e-05, "loss": 1.0091, "step": 3001 }, { "epoch": 0.1, "learning_rate": 1.9744908897500085e-05, "loss": 1.0283, "step": 3002 }, { "epoch": 0.1, "learning_rate": 1.9744666935172663e-05, "loss": 1.0247, "step": 3003 }, { "epoch": 0.1, "learning_rate": 1.974442485962898e-05, "loss": 0.9964, "step": 3004 }, { "epoch": 0.1, "learning_rate": 1.9744182670871843e-05, "loss": 0.9894, "step": 3005 }, { "epoch": 0.1, "learning_rate": 1.9743940368904067e-05, "loss": 1.0273, "step": 3006 }, { "epoch": 0.1, "learning_rate": 1.974369795372847e-05, "loss": 1.0242, "step": 3007 }, { "epoch": 0.1, "learning_rate": 1.9743455425347862e-05, "loss": 1.0266, "step": 3008 }, { "epoch": 0.1, "learning_rate": 1.9743212783765065e-05, "loss": 1.0568, "step": 3009 }, { "epoch": 0.1, "learning_rate": 1.97429700289829e-05, "loss": 0.9657, "step": 3010 }, { "epoch": 0.1, "learning_rate": 1.9742727161004186e-05, "loss": 1.0304, "step": 3011 }, { "epoch": 0.1, "learning_rate": 1.974248417983174e-05, "loss": 0.9955, "step": 3012 }, { "epoch": 0.1, "learning_rate": 1.9742241085468393e-05, "loss": 1.0463, "step": 3013 }, { "epoch": 0.1, "learning_rate": 1.974199787791696e-05, "loss": 1.0232, "step": 3014 }, { "epoch": 0.1, "learning_rate": 1.9741754557180277e-05, "loss": 1.0437, "step": 3015 }, { "epoch": 0.1, "learning_rate": 1.9741511123261167e-05, "loss": 1.0003, "step": 3016 }, { "epoch": 0.1, "learning_rate": 1.974126757616245e-05, "loss": 0.9818, "step": 3017 }, { "epoch": 0.1, "learning_rate": 1.9741023915886964e-05, "loss": 1.0622, "step": 3018 }, { "epoch": 0.1, "learning_rate": 1.9740780142437543e-05, "loss": 1.0292, "step": 3019 }, { "epoch": 0.1, "learning_rate": 1.974053625581701e-05, "loss": 1.0167, "step": 3020 }, { "epoch": 0.1, "learning_rate": 1.974029225602821e-05, "loss": 1.0303, "step": 3021 }, { "epoch": 0.1, "learning_rate": 1.9740048143073967e-05, "loss": 1.0229, "step": 3022 }, { "epoch": 0.1, "learning_rate": 1.973980391695712e-05, "loss": 1.0235, "step": 3023 }, { "epoch": 0.1, "learning_rate": 1.973955957768051e-05, "loss": 0.9922, "step": 3024 }, { "epoch": 0.1, "learning_rate": 1.973931512524697e-05, "loss": 1.0004, "step": 3025 }, { "epoch": 0.1, "learning_rate": 1.9739070559659347e-05, "loss": 1.0221, "step": 3026 }, { "epoch": 0.1, "learning_rate": 1.973882588092048e-05, "loss": 1.0321, "step": 3027 }, { "epoch": 0.1, "learning_rate": 1.973858108903321e-05, "loss": 1.042, "step": 3028 }, { "epoch": 0.1, "learning_rate": 1.9738336184000383e-05, "loss": 1.0066, "step": 3029 }, { "epoch": 0.1, "learning_rate": 1.9738091165824838e-05, "loss": 1.0394, "step": 3030 }, { "epoch": 0.1, "learning_rate": 1.9737846034509432e-05, "loss": 1.0185, "step": 3031 }, { "epoch": 0.1, "learning_rate": 1.9737600790057007e-05, "loss": 1.0152, "step": 3032 }, { "epoch": 0.1, "learning_rate": 1.9737355432470417e-05, "loss": 1.0596, "step": 3033 }, { "epoch": 0.1, "learning_rate": 1.9737109961752507e-05, "loss": 0.994, "step": 3034 }, { "epoch": 0.1, "learning_rate": 1.9736864377906127e-05, "loss": 1.0376, "step": 3035 }, { "epoch": 0.1, "learning_rate": 1.9736618680934136e-05, "loss": 1.0187, "step": 3036 }, { "epoch": 0.1, "learning_rate": 1.9736372870839387e-05, "loss": 1.0465, "step": 3037 }, { "epoch": 0.1, "learning_rate": 1.9736126947624737e-05, "loss": 1.0193, "step": 3038 }, { "epoch": 0.1, "learning_rate": 1.973588091129304e-05, "loss": 1.0208, "step": 3039 }, { "epoch": 0.1, "learning_rate": 1.9735634761847156e-05, "loss": 0.9978, "step": 3040 }, { "epoch": 0.1, "learning_rate": 1.973538849928995e-05, "loss": 1.0372, "step": 3041 }, { "epoch": 0.1, "learning_rate": 1.9735142123624274e-05, "loss": 0.9859, "step": 3042 }, { "epoch": 0.1, "learning_rate": 1.9734895634852996e-05, "loss": 1.0349, "step": 3043 }, { "epoch": 0.1, "learning_rate": 1.973464903297898e-05, "loss": 1.0208, "step": 3044 }, { "epoch": 0.1, "learning_rate": 1.9734402318005087e-05, "loss": 1.0216, "step": 3045 }, { "epoch": 0.1, "learning_rate": 1.9734155489934186e-05, "loss": 0.9952, "step": 3046 }, { "epoch": 0.1, "learning_rate": 1.9733908548769146e-05, "loss": 1.0146, "step": 3047 }, { "epoch": 0.1, "learning_rate": 1.9733661494512838e-05, "loss": 1.0014, "step": 3048 }, { "epoch": 0.1, "learning_rate": 1.9733414327168124e-05, "loss": 1.0454, "step": 3049 }, { "epoch": 0.1, "learning_rate": 1.9733167046737885e-05, "loss": 1.002, "step": 3050 }, { "epoch": 0.1, "learning_rate": 1.9732919653224986e-05, "loss": 1.0274, "step": 3051 }, { "epoch": 0.1, "learning_rate": 1.9732672146632308e-05, "loss": 1.0127, "step": 3052 }, { "epoch": 0.1, "learning_rate": 1.9732424526962724e-05, "loss": 1.044, "step": 3053 }, { "epoch": 0.1, "learning_rate": 1.973217679421911e-05, "loss": 1.0183, "step": 3054 }, { "epoch": 0.1, "learning_rate": 1.9731928948404343e-05, "loss": 1.0088, "step": 3055 }, { "epoch": 0.1, "learning_rate": 1.9731680989521305e-05, "loss": 0.992, "step": 3056 }, { "epoch": 0.1, "learning_rate": 1.973143291757288e-05, "loss": 1.0296, "step": 3057 }, { "epoch": 0.1, "learning_rate": 1.973118473256194e-05, "loss": 1.0476, "step": 3058 }, { "epoch": 0.1, "learning_rate": 1.973093643449138e-05, "loss": 0.988, "step": 3059 }, { "epoch": 0.1, "learning_rate": 1.9730688023364077e-05, "loss": 1.0109, "step": 3060 }, { "epoch": 0.1, "learning_rate": 1.9730439499182923e-05, "loss": 1.0259, "step": 3061 }, { "epoch": 0.1, "learning_rate": 1.97301908619508e-05, "loss": 1.0096, "step": 3062 }, { "epoch": 0.1, "learning_rate": 1.9729942111670598e-05, "loss": 1.0114, "step": 3063 }, { "epoch": 0.1, "learning_rate": 1.972969324834521e-05, "loss": 0.9891, "step": 3064 }, { "epoch": 0.1, "learning_rate": 1.9729444271977524e-05, "loss": 1.0441, "step": 3065 }, { "epoch": 0.1, "learning_rate": 1.9729195182570434e-05, "loss": 1.0394, "step": 3066 }, { "epoch": 0.1, "learning_rate": 1.9728945980126838e-05, "loss": 1.0188, "step": 3067 }, { "epoch": 0.1, "learning_rate": 1.9728696664649622e-05, "loss": 0.9942, "step": 3068 }, { "epoch": 0.1, "learning_rate": 1.972844723614169e-05, "loss": 1.0421, "step": 3069 }, { "epoch": 0.1, "learning_rate": 1.972819769460594e-05, "loss": 0.9966, "step": 3070 }, { "epoch": 0.1, "learning_rate": 1.9727948040045267e-05, "loss": 0.9945, "step": 3071 }, { "epoch": 0.1, "learning_rate": 1.9727698272462573e-05, "loss": 1.0088, "step": 3072 }, { "epoch": 0.1, "learning_rate": 1.9727448391860763e-05, "loss": 1.0086, "step": 3073 }, { "epoch": 0.1, "learning_rate": 1.9727198398242736e-05, "loss": 1.0086, "step": 3074 }, { "epoch": 0.1, "learning_rate": 1.9726948291611398e-05, "loss": 0.9911, "step": 3075 }, { "epoch": 0.1, "learning_rate": 1.9726698071969655e-05, "loss": 1.0206, "step": 3076 }, { "epoch": 0.1, "learning_rate": 1.9726447739320416e-05, "loss": 1.0165, "step": 3077 }, { "epoch": 0.1, "learning_rate": 1.9726197293666586e-05, "loss": 1.0275, "step": 3078 }, { "epoch": 0.1, "learning_rate": 1.9725946735011078e-05, "loss": 0.9963, "step": 3079 }, { "epoch": 0.1, "learning_rate": 1.97256960633568e-05, "loss": 0.9795, "step": 3080 }, { "epoch": 0.1, "learning_rate": 1.9725445278706666e-05, "loss": 1.0561, "step": 3081 }, { "epoch": 0.1, "learning_rate": 1.972519438106359e-05, "loss": 0.9982, "step": 3082 }, { "epoch": 0.1, "learning_rate": 1.972494337043049e-05, "loss": 1.0181, "step": 3083 }, { "epoch": 0.1, "learning_rate": 1.9724692246810272e-05, "loss": 0.9493, "step": 3084 }, { "epoch": 0.1, "learning_rate": 1.9724441010205865e-05, "loss": 0.9951, "step": 3085 }, { "epoch": 0.1, "learning_rate": 1.972418966062018e-05, "loss": 1.0201, "step": 3086 }, { "epoch": 0.1, "learning_rate": 1.9723938198056144e-05, "loss": 1.0184, "step": 3087 }, { "epoch": 0.1, "learning_rate": 1.9723686622516674e-05, "loss": 0.9874, "step": 3088 }, { "epoch": 0.1, "learning_rate": 1.9723434934004692e-05, "loss": 1.0464, "step": 3089 }, { "epoch": 0.1, "learning_rate": 1.9723183132523127e-05, "loss": 0.9857, "step": 3090 }, { "epoch": 0.1, "learning_rate": 1.9722931218074902e-05, "loss": 1.0154, "step": 3091 }, { "epoch": 0.1, "learning_rate": 1.972267919066294e-05, "loss": 0.9792, "step": 3092 }, { "epoch": 0.1, "learning_rate": 1.9722427050290176e-05, "loss": 1.0067, "step": 3093 }, { "epoch": 0.1, "learning_rate": 1.9722174796959535e-05, "loss": 1.066, "step": 3094 }, { "epoch": 0.1, "learning_rate": 1.972192243067395e-05, "loss": 1.0118, "step": 3095 }, { "epoch": 0.1, "learning_rate": 1.972166995143635e-05, "loss": 0.9864, "step": 3096 }, { "epoch": 0.1, "learning_rate": 1.9721417359249672e-05, "loss": 1.0245, "step": 3097 }, { "epoch": 0.1, "learning_rate": 1.9721164654116848e-05, "loss": 1.0256, "step": 3098 }, { "epoch": 0.1, "learning_rate": 1.9720911836040817e-05, "loss": 1.0059, "step": 3099 }, { "epoch": 0.1, "learning_rate": 1.9720658905024513e-05, "loss": 0.9505, "step": 3100 }, { "epoch": 0.1, "learning_rate": 1.9720405861070874e-05, "loss": 1.023, "step": 3101 }, { "epoch": 0.1, "learning_rate": 1.9720152704182843e-05, "loss": 0.9739, "step": 3102 }, { "epoch": 0.1, "learning_rate": 1.9719899434363362e-05, "loss": 1.0456, "step": 3103 }, { "epoch": 0.1, "learning_rate": 1.9719646051615368e-05, "loss": 1.0195, "step": 3104 }, { "epoch": 0.1, "learning_rate": 1.9719392555941812e-05, "loss": 1.0236, "step": 3105 }, { "epoch": 0.1, "learning_rate": 1.9719138947345632e-05, "loss": 0.9819, "step": 3106 }, { "epoch": 0.1, "learning_rate": 1.9718885225829783e-05, "loss": 1.018, "step": 3107 }, { "epoch": 0.1, "learning_rate": 1.9718631391397207e-05, "loss": 1.0024, "step": 3108 }, { "epoch": 0.1, "learning_rate": 1.971837744405085e-05, "loss": 1.0167, "step": 3109 }, { "epoch": 0.1, "learning_rate": 1.9718123383793674e-05, "loss": 0.994, "step": 3110 }, { "epoch": 0.1, "learning_rate": 1.9717869210628615e-05, "loss": 1.0172, "step": 3111 }, { "epoch": 0.1, "learning_rate": 1.9717614924558638e-05, "loss": 1.0127, "step": 3112 }, { "epoch": 0.1, "learning_rate": 1.9717360525586694e-05, "loss": 1.016, "step": 3113 }, { "epoch": 0.1, "learning_rate": 1.9717106013715738e-05, "loss": 1.0238, "step": 3114 }, { "epoch": 0.1, "learning_rate": 1.971685138894873e-05, "loss": 0.9918, "step": 3115 }, { "epoch": 0.1, "learning_rate": 1.9716596651288623e-05, "loss": 1.0152, "step": 3116 }, { "epoch": 0.1, "learning_rate": 1.971634180073838e-05, "loss": 1.0102, "step": 3117 }, { "epoch": 0.1, "learning_rate": 1.971608683730096e-05, "loss": 1.0097, "step": 3118 }, { "epoch": 0.1, "learning_rate": 1.9715831760979326e-05, "loss": 0.9737, "step": 3119 }, { "epoch": 0.1, "learning_rate": 1.9715576571776448e-05, "loss": 1.0306, "step": 3120 }, { "epoch": 0.1, "learning_rate": 1.971532126969528e-05, "loss": 1.033, "step": 3121 }, { "epoch": 0.1, "learning_rate": 1.9715065854738794e-05, "loss": 1.0216, "step": 3122 }, { "epoch": 0.1, "learning_rate": 1.9714810326909954e-05, "loss": 1.0333, "step": 3123 }, { "epoch": 0.1, "learning_rate": 1.9714554686211735e-05, "loss": 1.0457, "step": 3124 }, { "epoch": 0.1, "learning_rate": 1.97142989326471e-05, "loss": 0.9943, "step": 3125 }, { "epoch": 0.1, "learning_rate": 1.971404306621903e-05, "loss": 1.0319, "step": 3126 }, { "epoch": 0.1, "learning_rate": 1.9713787086930485e-05, "loss": 1.0002, "step": 3127 }, { "epoch": 0.1, "learning_rate": 1.971353099478445e-05, "loss": 1.0193, "step": 3128 }, { "epoch": 0.1, "learning_rate": 1.9713274789783893e-05, "loss": 1.0013, "step": 3129 }, { "epoch": 0.1, "learning_rate": 1.9713018471931794e-05, "loss": 1.0068, "step": 3130 }, { "epoch": 0.1, "learning_rate": 1.971276204123113e-05, "loss": 1.0432, "step": 3131 }, { "epoch": 0.1, "learning_rate": 1.9712505497684883e-05, "loss": 0.9872, "step": 3132 }, { "epoch": 0.1, "learning_rate": 1.9712248841296032e-05, "loss": 1.0206, "step": 3133 }, { "epoch": 0.1, "learning_rate": 1.9711992072067557e-05, "loss": 1.0237, "step": 3134 }, { "epoch": 0.1, "learning_rate": 1.971173519000244e-05, "loss": 1.0495, "step": 3135 }, { "epoch": 0.1, "learning_rate": 1.9711478195103674e-05, "loss": 0.9674, "step": 3136 }, { "epoch": 0.1, "learning_rate": 1.9711221087374236e-05, "loss": 1.047, "step": 3137 }, { "epoch": 0.1, "learning_rate": 1.9710963866817113e-05, "loss": 1.0146, "step": 3138 }, { "epoch": 0.1, "learning_rate": 1.9710706533435303e-05, "loss": 1.0288, "step": 3139 }, { "epoch": 0.1, "learning_rate": 1.9710449087231784e-05, "loss": 0.9888, "step": 3140 }, { "epoch": 0.1, "learning_rate": 1.9710191528209558e-05, "loss": 1.0346, "step": 3141 }, { "epoch": 0.1, "learning_rate": 1.9709933856371607e-05, "loss": 0.9911, "step": 3142 }, { "epoch": 0.1, "learning_rate": 1.970967607172093e-05, "loss": 1.0129, "step": 3143 }, { "epoch": 0.1, "learning_rate": 1.9709418174260523e-05, "loss": 0.9736, "step": 3144 }, { "epoch": 0.1, "learning_rate": 1.970916016399338e-05, "loss": 0.9936, "step": 3145 }, { "epoch": 0.1, "learning_rate": 1.97089020409225e-05, "loss": 1.0191, "step": 3146 }, { "epoch": 0.1, "learning_rate": 1.970864380505088e-05, "loss": 1.0732, "step": 3147 }, { "epoch": 0.1, "learning_rate": 1.9708385456381523e-05, "loss": 1.0042, "step": 3148 }, { "epoch": 0.1, "learning_rate": 1.970812699491743e-05, "loss": 1.0346, "step": 3149 }, { "epoch": 0.1, "learning_rate": 1.9707868420661604e-05, "loss": 1.0034, "step": 3150 }, { "epoch": 0.1, "learning_rate": 1.9707609733617047e-05, "loss": 0.9702, "step": 3151 }, { "epoch": 0.1, "learning_rate": 1.9707350933786763e-05, "loss": 0.9984, "step": 3152 }, { "epoch": 0.1, "learning_rate": 1.9707092021173765e-05, "loss": 1.0034, "step": 3153 }, { "epoch": 0.1, "learning_rate": 1.9706832995781057e-05, "loss": 1.031, "step": 3154 }, { "epoch": 0.1, "learning_rate": 1.970657385761165e-05, "loss": 1.0502, "step": 3155 }, { "epoch": 0.11, "learning_rate": 1.9706314606668554e-05, "loss": 0.975, "step": 3156 }, { "epoch": 0.11, "learning_rate": 1.9706055242954777e-05, "loss": 1.0403, "step": 3157 }, { "epoch": 0.11, "learning_rate": 1.9705795766473344e-05, "loss": 0.9782, "step": 3158 }, { "epoch": 0.11, "learning_rate": 1.9705536177227256e-05, "loss": 1.0667, "step": 3159 }, { "epoch": 0.11, "learning_rate": 1.9705276475219538e-05, "loss": 0.9935, "step": 3160 }, { "epoch": 0.11, "learning_rate": 1.97050166604532e-05, "loss": 0.9701, "step": 3161 }, { "epoch": 0.11, "learning_rate": 1.9704756732931268e-05, "loss": 1.0201, "step": 3162 }, { "epoch": 0.11, "learning_rate": 1.970449669265676e-05, "loss": 1.0273, "step": 3163 }, { "epoch": 0.11, "learning_rate": 1.9704236539632695e-05, "loss": 1.018, "step": 3164 }, { "epoch": 0.11, "learning_rate": 1.9703976273862097e-05, "loss": 1.0123, "step": 3165 }, { "epoch": 0.11, "learning_rate": 1.970371589534799e-05, "loss": 0.9931, "step": 3166 }, { "epoch": 0.11, "learning_rate": 1.9703455404093397e-05, "loss": 1.0252, "step": 3167 }, { "epoch": 0.11, "learning_rate": 1.9703194800101344e-05, "loss": 1.0495, "step": 3168 }, { "epoch": 0.11, "learning_rate": 1.9702934083374866e-05, "loss": 1.0216, "step": 3169 }, { "epoch": 0.11, "learning_rate": 1.970267325391698e-05, "loss": 0.9668, "step": 3170 }, { "epoch": 0.11, "learning_rate": 1.970241231173073e-05, "loss": 1.036, "step": 3171 }, { "epoch": 0.11, "learning_rate": 1.9702151256819138e-05, "loss": 0.9716, "step": 3172 }, { "epoch": 0.11, "learning_rate": 1.9701890089185242e-05, "loss": 1.0469, "step": 3173 }, { "epoch": 0.11, "learning_rate": 1.970162880883207e-05, "loss": 1.0139, "step": 3174 }, { "epoch": 0.11, "learning_rate": 1.9701367415762663e-05, "loss": 1.0626, "step": 3175 }, { "epoch": 0.11, "learning_rate": 1.970110590998006e-05, "loss": 0.9676, "step": 3176 }, { "epoch": 0.11, "learning_rate": 1.9700844291487297e-05, "loss": 1.0458, "step": 3177 }, { "epoch": 0.11, "learning_rate": 1.970058256028741e-05, "loss": 1.0042, "step": 3178 }, { "epoch": 0.11, "learning_rate": 1.970032071638344e-05, "loss": 1.0141, "step": 3179 }, { "epoch": 0.11, "learning_rate": 1.9700058759778433e-05, "loss": 1.0432, "step": 3180 }, { "epoch": 0.11, "learning_rate": 1.9699796690475434e-05, "loss": 1.0078, "step": 3181 }, { "epoch": 0.11, "learning_rate": 1.9699534508477483e-05, "loss": 1.0014, "step": 3182 }, { "epoch": 0.11, "learning_rate": 1.969927221378763e-05, "loss": 1.0292, "step": 3183 }, { "epoch": 0.11, "learning_rate": 1.969900980640892e-05, "loss": 1.0365, "step": 3184 }, { "epoch": 0.11, "learning_rate": 1.96987472863444e-05, "loss": 1.0323, "step": 3185 }, { "epoch": 0.11, "learning_rate": 1.969848465359712e-05, "loss": 1.0085, "step": 3186 }, { "epoch": 0.11, "learning_rate": 1.969822190817014e-05, "loss": 1.0279, "step": 3187 }, { "epoch": 0.11, "learning_rate": 1.9697959050066503e-05, "loss": 1.0051, "step": 3188 }, { "epoch": 0.11, "learning_rate": 1.9697696079289265e-05, "loss": 1.022, "step": 3189 }, { "epoch": 0.11, "learning_rate": 1.9697432995841484e-05, "loss": 1.0006, "step": 3190 }, { "epoch": 0.11, "learning_rate": 1.9697169799726215e-05, "loss": 1.0185, "step": 3191 }, { "epoch": 0.11, "learning_rate": 1.9696906490946514e-05, "loss": 0.978, "step": 3192 }, { "epoch": 0.11, "learning_rate": 1.9696643069505445e-05, "loss": 1.031, "step": 3193 }, { "epoch": 0.11, "learning_rate": 1.9696379535406066e-05, "loss": 1.0138, "step": 3194 }, { "epoch": 0.11, "learning_rate": 1.9696115888651437e-05, "loss": 0.9691, "step": 3195 }, { "epoch": 0.11, "learning_rate": 1.969585212924462e-05, "loss": 1.026, "step": 3196 }, { "epoch": 0.11, "learning_rate": 1.969558825718869e-05, "loss": 1.0235, "step": 3197 }, { "epoch": 0.11, "learning_rate": 1.9695324272486694e-05, "loss": 1.0046, "step": 3198 }, { "epoch": 0.11, "learning_rate": 1.9695060175141715e-05, "loss": 0.9756, "step": 3199 }, { "epoch": 0.11, "learning_rate": 1.9694795965156818e-05, "loss": 0.976, "step": 3200 }, { "epoch": 0.11, "learning_rate": 1.969453164253507e-05, "loss": 1.0023, "step": 3201 }, { "epoch": 0.11, "learning_rate": 1.9694267207279542e-05, "loss": 0.9684, "step": 3202 }, { "epoch": 0.11, "learning_rate": 1.9694002659393306e-05, "loss": 1.0616, "step": 3203 }, { "epoch": 0.11, "learning_rate": 1.9693737998879438e-05, "loss": 0.9933, "step": 3204 }, { "epoch": 0.11, "learning_rate": 1.969347322574101e-05, "loss": 1.015, "step": 3205 }, { "epoch": 0.11, "learning_rate": 1.96932083399811e-05, "loss": 1.0049, "step": 3206 }, { "epoch": 0.11, "learning_rate": 1.9692943341602787e-05, "loss": 1.0258, "step": 3207 }, { "epoch": 0.11, "learning_rate": 1.9692678230609147e-05, "loss": 0.979, "step": 3208 }, { "epoch": 0.11, "learning_rate": 1.9692413007003262e-05, "loss": 1.0383, "step": 3209 }, { "epoch": 0.11, "learning_rate": 1.969214767078821e-05, "loss": 1.0234, "step": 3210 }, { "epoch": 0.11, "learning_rate": 1.969188222196708e-05, "loss": 1.028, "step": 3211 }, { "epoch": 0.11, "learning_rate": 1.969161666054295e-05, "loss": 0.934, "step": 3212 }, { "epoch": 0.11, "learning_rate": 1.969135098651891e-05, "loss": 1.0112, "step": 3213 }, { "epoch": 0.11, "learning_rate": 1.9691085199898045e-05, "loss": 1.0125, "step": 3214 }, { "epoch": 0.11, "learning_rate": 1.969081930068344e-05, "loss": 1.0495, "step": 3215 }, { "epoch": 0.11, "learning_rate": 1.9690553288878187e-05, "loss": 0.9631, "step": 3216 }, { "epoch": 0.11, "learning_rate": 1.9690287164485377e-05, "loss": 1.0401, "step": 3217 }, { "epoch": 0.11, "learning_rate": 1.96900209275081e-05, "loss": 0.9506, "step": 3218 }, { "epoch": 0.11, "learning_rate": 1.9689754577949455e-05, "loss": 1.0279, "step": 3219 }, { "epoch": 0.11, "learning_rate": 1.9689488115812527e-05, "loss": 0.991, "step": 3220 }, { "epoch": 0.11, "learning_rate": 1.9689221541100417e-05, "loss": 0.9876, "step": 3221 }, { "epoch": 0.11, "learning_rate": 1.9688954853816226e-05, "loss": 1.0171, "step": 3222 }, { "epoch": 0.11, "learning_rate": 1.9688688053963047e-05, "loss": 1.0163, "step": 3223 }, { "epoch": 0.11, "learning_rate": 1.968842114154398e-05, "loss": 1.0055, "step": 3224 }, { "epoch": 0.11, "learning_rate": 1.9688154116562125e-05, "loss": 0.9735, "step": 3225 }, { "epoch": 0.11, "learning_rate": 1.9687886979020593e-05, "loss": 1.042, "step": 3226 }, { "epoch": 0.11, "learning_rate": 1.9687619728922476e-05, "loss": 1.0529, "step": 3227 }, { "epoch": 0.11, "learning_rate": 1.9687352366270886e-05, "loss": 0.9925, "step": 3228 }, { "epoch": 0.11, "learning_rate": 1.968708489106893e-05, "loss": 0.9971, "step": 3229 }, { "epoch": 0.11, "learning_rate": 1.968681730331971e-05, "loss": 0.9657, "step": 3230 }, { "epoch": 0.11, "learning_rate": 1.968654960302634e-05, "loss": 1.0313, "step": 3231 }, { "epoch": 0.11, "learning_rate": 1.9686281790191927e-05, "loss": 1.0104, "step": 3232 }, { "epoch": 0.11, "learning_rate": 1.9686013864819582e-05, "loss": 1.0023, "step": 3233 }, { "epoch": 0.11, "learning_rate": 1.9685745826912425e-05, "loss": 1.0081, "step": 3234 }, { "epoch": 0.11, "learning_rate": 1.968547767647356e-05, "loss": 1.0208, "step": 3235 }, { "epoch": 0.11, "learning_rate": 1.9685209413506107e-05, "loss": 0.9916, "step": 3236 }, { "epoch": 0.11, "learning_rate": 1.9684941038013186e-05, "loss": 1.0188, "step": 3237 }, { "epoch": 0.11, "learning_rate": 1.968467254999791e-05, "loss": 1.0041, "step": 3238 }, { "epoch": 0.11, "learning_rate": 1.9684403949463404e-05, "loss": 1.0615, "step": 3239 }, { "epoch": 0.11, "learning_rate": 1.9684135236412783e-05, "loss": 0.9977, "step": 3240 }, { "epoch": 0.11, "learning_rate": 1.968386641084917e-05, "loss": 1.0493, "step": 3241 }, { "epoch": 0.11, "learning_rate": 1.9683597472775692e-05, "loss": 0.9923, "step": 3242 }, { "epoch": 0.11, "learning_rate": 1.968332842219547e-05, "loss": 1.0052, "step": 3243 }, { "epoch": 0.11, "learning_rate": 1.968305925911163e-05, "loss": 1.0154, "step": 3244 }, { "epoch": 0.11, "learning_rate": 1.96827899835273e-05, "loss": 1.0811, "step": 3245 }, { "epoch": 0.11, "learning_rate": 1.968252059544561e-05, "loss": 1.0295, "step": 3246 }, { "epoch": 0.11, "learning_rate": 1.9682251094869688e-05, "loss": 1.0227, "step": 3247 }, { "epoch": 0.11, "learning_rate": 1.9681981481802664e-05, "loss": 1.0006, "step": 3248 }, { "epoch": 0.11, "learning_rate": 1.9681711756247673e-05, "loss": 1.0205, "step": 3249 }, { "epoch": 0.11, "learning_rate": 1.9681441918207848e-05, "loss": 1.042, "step": 3250 }, { "epoch": 0.11, "learning_rate": 1.9681171967686324e-05, "loss": 0.9992, "step": 3251 }, { "epoch": 0.11, "learning_rate": 1.9680901904686237e-05, "loss": 1.002, "step": 3252 }, { "epoch": 0.11, "learning_rate": 1.9680631729210723e-05, "loss": 0.9849, "step": 3253 }, { "epoch": 0.11, "learning_rate": 1.9680361441262924e-05, "loss": 1.0231, "step": 3254 }, { "epoch": 0.11, "learning_rate": 1.9680091040845982e-05, "loss": 0.965, "step": 3255 }, { "epoch": 0.11, "learning_rate": 1.9679820527963035e-05, "loss": 0.9761, "step": 3256 }, { "epoch": 0.11, "learning_rate": 1.9679549902617222e-05, "loss": 0.993, "step": 3257 }, { "epoch": 0.11, "learning_rate": 1.96792791648117e-05, "loss": 1.0112, "step": 3258 }, { "epoch": 0.11, "learning_rate": 1.96790083145496e-05, "loss": 1.0259, "step": 3259 }, { "epoch": 0.11, "learning_rate": 1.9678737351834077e-05, "loss": 0.9881, "step": 3260 }, { "epoch": 0.11, "learning_rate": 1.9678466276668274e-05, "loss": 1.0278, "step": 3261 }, { "epoch": 0.11, "learning_rate": 1.9678195089055347e-05, "loss": 1.0346, "step": 3262 }, { "epoch": 0.11, "learning_rate": 1.9677923788998444e-05, "loss": 1.0252, "step": 3263 }, { "epoch": 0.11, "learning_rate": 1.9677652376500716e-05, "loss": 0.9758, "step": 3264 }, { "epoch": 0.11, "learning_rate": 1.9677380851565317e-05, "loss": 1.0493, "step": 3265 }, { "epoch": 0.11, "learning_rate": 1.9677109214195398e-05, "loss": 1.0583, "step": 3266 }, { "epoch": 0.11, "learning_rate": 1.9676837464394123e-05, "loss": 1.0175, "step": 3267 }, { "epoch": 0.11, "learning_rate": 1.9676565602164645e-05, "loss": 0.9862, "step": 3268 }, { "epoch": 0.11, "learning_rate": 1.9676293627510116e-05, "loss": 1.0067, "step": 3269 }, { "epoch": 0.11, "learning_rate": 1.967602154043371e-05, "loss": 1.0368, "step": 3270 }, { "epoch": 0.11, "learning_rate": 1.9675749340938574e-05, "loss": 0.9605, "step": 3271 }, { "epoch": 0.11, "learning_rate": 1.967547702902788e-05, "loss": 0.9881, "step": 3272 }, { "epoch": 0.11, "learning_rate": 1.967520460470479e-05, "loss": 1.0307, "step": 3273 }, { "epoch": 0.11, "learning_rate": 1.9674932067972464e-05, "loss": 0.997, "step": 3274 }, { "epoch": 0.11, "learning_rate": 1.9674659418834073e-05, "loss": 1.043, "step": 3275 }, { "epoch": 0.11, "learning_rate": 1.9674386657292784e-05, "loss": 1.0256, "step": 3276 }, { "epoch": 0.11, "learning_rate": 1.9674113783351764e-05, "loss": 1.009, "step": 3277 }, { "epoch": 0.11, "learning_rate": 1.967384079701419e-05, "loss": 1.0182, "step": 3278 }, { "epoch": 0.11, "learning_rate": 1.9673567698283225e-05, "loss": 1.0293, "step": 3279 }, { "epoch": 0.11, "learning_rate": 1.9673294487162045e-05, "loss": 1.046, "step": 3280 }, { "epoch": 0.11, "learning_rate": 1.967302116365383e-05, "loss": 1.0475, "step": 3281 }, { "epoch": 0.11, "learning_rate": 1.9672747727761747e-05, "loss": 1.0112, "step": 3282 }, { "epoch": 0.11, "learning_rate": 1.967247417948898e-05, "loss": 1.0561, "step": 3283 }, { "epoch": 0.11, "learning_rate": 1.9672200518838696e-05, "loss": 1.011, "step": 3284 }, { "epoch": 0.11, "learning_rate": 1.967192674581409e-05, "loss": 1.0232, "step": 3285 }, { "epoch": 0.11, "learning_rate": 1.967165286041833e-05, "loss": 0.9708, "step": 3286 }, { "epoch": 0.11, "learning_rate": 1.9671378862654606e-05, "loss": 1.0374, "step": 3287 }, { "epoch": 0.11, "learning_rate": 1.9671104752526093e-05, "loss": 0.952, "step": 3288 }, { "epoch": 0.11, "learning_rate": 1.9670830530035986e-05, "loss": 1.0056, "step": 3289 }, { "epoch": 0.11, "learning_rate": 1.9670556195187464e-05, "loss": 1.0078, "step": 3290 }, { "epoch": 0.11, "learning_rate": 1.9670281747983718e-05, "loss": 1.0381, "step": 3291 }, { "epoch": 0.11, "learning_rate": 1.967000718842793e-05, "loss": 0.9838, "step": 3292 }, { "epoch": 0.11, "learning_rate": 1.96697325165233e-05, "loss": 1.0425, "step": 3293 }, { "epoch": 0.11, "learning_rate": 1.966945773227301e-05, "loss": 1.0917, "step": 3294 }, { "epoch": 0.11, "learning_rate": 1.966918283568026e-05, "loss": 1.0629, "step": 3295 }, { "epoch": 0.11, "learning_rate": 1.966890782674824e-05, "loss": 1.0174, "step": 3296 }, { "epoch": 0.11, "learning_rate": 1.9668632705480143e-05, "loss": 1.037, "step": 3297 }, { "epoch": 0.11, "learning_rate": 1.966835747187917e-05, "loss": 0.9964, "step": 3298 }, { "epoch": 0.11, "learning_rate": 1.9668082125948517e-05, "loss": 1.0416, "step": 3299 }, { "epoch": 0.11, "learning_rate": 1.966780666769138e-05, "loss": 1.0332, "step": 3300 }, { "epoch": 0.11, "learning_rate": 1.9667531097110964e-05, "loss": 1.0375, "step": 3301 }, { "epoch": 0.11, "learning_rate": 1.9667255414210466e-05, "loss": 0.9808, "step": 3302 }, { "epoch": 0.11, "learning_rate": 1.9666979618993096e-05, "loss": 1.0352, "step": 3303 }, { "epoch": 0.11, "learning_rate": 1.9666703711462052e-05, "loss": 0.9556, "step": 3304 }, { "epoch": 0.11, "learning_rate": 1.966642769162054e-05, "loss": 1.0089, "step": 3305 }, { "epoch": 0.11, "learning_rate": 1.9666151559471768e-05, "loss": 1.0006, "step": 3306 }, { "epoch": 0.11, "learning_rate": 1.966587531501895e-05, "loss": 1.0809, "step": 3307 }, { "epoch": 0.11, "learning_rate": 1.9665598958265285e-05, "loss": 1.0065, "step": 3308 }, { "epoch": 0.11, "learning_rate": 1.966532248921399e-05, "loss": 1.0451, "step": 3309 }, { "epoch": 0.11, "learning_rate": 1.9665045907868272e-05, "loss": 0.9833, "step": 3310 }, { "epoch": 0.11, "learning_rate": 1.9664769214231353e-05, "loss": 0.9822, "step": 3311 }, { "epoch": 0.11, "learning_rate": 1.966449240830644e-05, "loss": 1.0054, "step": 3312 }, { "epoch": 0.11, "learning_rate": 1.9664215490096755e-05, "loss": 1.0063, "step": 3313 }, { "epoch": 0.11, "learning_rate": 1.9663938459605512e-05, "loss": 1.0272, "step": 3314 }, { "epoch": 0.11, "learning_rate": 1.966366131683593e-05, "loss": 1.0034, "step": 3315 }, { "epoch": 0.11, "learning_rate": 1.9663384061791228e-05, "loss": 0.9974, "step": 3316 }, { "epoch": 0.11, "learning_rate": 1.9663106694474624e-05, "loss": 1.0812, "step": 3317 }, { "epoch": 0.11, "learning_rate": 1.9662829214889352e-05, "loss": 1.0108, "step": 3318 }, { "epoch": 0.11, "learning_rate": 1.9662551623038625e-05, "loss": 1.0629, "step": 3319 }, { "epoch": 0.11, "learning_rate": 1.966227391892567e-05, "loss": 0.9891, "step": 3320 }, { "epoch": 0.11, "learning_rate": 1.9661996102553716e-05, "loss": 1.0482, "step": 3321 }, { "epoch": 0.11, "learning_rate": 1.9661718173925992e-05, "loss": 0.9727, "step": 3322 }, { "epoch": 0.11, "learning_rate": 1.9661440133045722e-05, "loss": 0.9802, "step": 3323 }, { "epoch": 0.11, "learning_rate": 1.966116197991614e-05, "loss": 0.9847, "step": 3324 }, { "epoch": 0.11, "learning_rate": 1.9660883714540477e-05, "loss": 1.0501, "step": 3325 }, { "epoch": 0.11, "learning_rate": 1.9660605336921968e-05, "loss": 1.0187, "step": 3326 }, { "epoch": 0.11, "learning_rate": 1.9660326847063844e-05, "loss": 0.9996, "step": 3327 }, { "epoch": 0.11, "learning_rate": 1.9660048244969338e-05, "loss": 1.0473, "step": 3328 }, { "epoch": 0.11, "learning_rate": 1.9659769530641696e-05, "loss": 1.0171, "step": 3329 }, { "epoch": 0.11, "learning_rate": 1.9659490704084147e-05, "loss": 0.9917, "step": 3330 }, { "epoch": 0.11, "learning_rate": 1.9659211765299936e-05, "loss": 0.9928, "step": 3331 }, { "epoch": 0.11, "learning_rate": 1.96589327142923e-05, "loss": 1.0322, "step": 3332 }, { "epoch": 0.11, "learning_rate": 1.9658653551064484e-05, "loss": 1.0205, "step": 3333 }, { "epoch": 0.11, "learning_rate": 1.965837427561973e-05, "loss": 1.0622, "step": 3334 }, { "epoch": 0.11, "learning_rate": 1.9658094887961285e-05, "loss": 0.9986, "step": 3335 }, { "epoch": 0.11, "learning_rate": 1.9657815388092395e-05, "loss": 1.0318, "step": 3336 }, { "epoch": 0.11, "learning_rate": 1.96575357760163e-05, "loss": 1.0142, "step": 3337 }, { "epoch": 0.11, "learning_rate": 1.9657256051736256e-05, "loss": 0.9947, "step": 3338 }, { "epoch": 0.11, "learning_rate": 1.9656976215255513e-05, "loss": 0.9804, "step": 3339 }, { "epoch": 0.11, "learning_rate": 1.9656696266577318e-05, "loss": 1.0043, "step": 3340 }, { "epoch": 0.11, "learning_rate": 1.9656416205704927e-05, "loss": 1.0739, "step": 3341 }, { "epoch": 0.11, "learning_rate": 1.9656136032641592e-05, "loss": 1.0033, "step": 3342 }, { "epoch": 0.11, "learning_rate": 1.965585574739057e-05, "loss": 0.9878, "step": 3343 }, { "epoch": 0.11, "learning_rate": 1.9655575349955115e-05, "loss": 0.9945, "step": 3344 }, { "epoch": 0.11, "learning_rate": 1.9655294840338485e-05, "loss": 1.0448, "step": 3345 }, { "epoch": 0.11, "learning_rate": 1.965501421854394e-05, "loss": 1.0312, "step": 3346 }, { "epoch": 0.11, "learning_rate": 1.9654733484574742e-05, "loss": 1.0174, "step": 3347 }, { "epoch": 0.11, "learning_rate": 1.9654452638434152e-05, "loss": 0.9892, "step": 3348 }, { "epoch": 0.11, "learning_rate": 1.9654171680125427e-05, "loss": 1.0085, "step": 3349 }, { "epoch": 0.11, "learning_rate": 1.965389060965184e-05, "loss": 0.989, "step": 3350 }, { "epoch": 0.11, "learning_rate": 1.9653609427016653e-05, "loss": 1.0041, "step": 3351 }, { "epoch": 0.11, "learning_rate": 1.9653328132223127e-05, "loss": 1.0079, "step": 3352 }, { "epoch": 0.11, "learning_rate": 1.9653046725274542e-05, "loss": 1.0212, "step": 3353 }, { "epoch": 0.11, "learning_rate": 1.965276520617416e-05, "loss": 1.0117, "step": 3354 }, { "epoch": 0.11, "learning_rate": 1.965248357492525e-05, "loss": 1.0348, "step": 3355 }, { "epoch": 0.11, "learning_rate": 1.965220183153109e-05, "loss": 0.9905, "step": 3356 }, { "epoch": 0.11, "learning_rate": 1.965191997599495e-05, "loss": 1.0218, "step": 3357 }, { "epoch": 0.11, "learning_rate": 1.9651638008320104e-05, "loss": 1.029, "step": 3358 }, { "epoch": 0.11, "learning_rate": 1.965135592850983e-05, "loss": 1.0184, "step": 3359 }, { "epoch": 0.11, "learning_rate": 1.9651073736567405e-05, "loss": 0.9564, "step": 3360 }, { "epoch": 0.11, "learning_rate": 1.9650791432496104e-05, "loss": 1.0373, "step": 3361 }, { "epoch": 0.11, "learning_rate": 1.965050901629921e-05, "loss": 1.0315, "step": 3362 }, { "epoch": 0.11, "learning_rate": 1.9650226487980006e-05, "loss": 1.0368, "step": 3363 }, { "epoch": 0.11, "learning_rate": 1.964994384754177e-05, "loss": 0.9644, "step": 3364 }, { "epoch": 0.11, "learning_rate": 1.9649661094987792e-05, "loss": 1.0629, "step": 3365 }, { "epoch": 0.11, "learning_rate": 1.964937823032135e-05, "loss": 1.0247, "step": 3366 }, { "epoch": 0.11, "learning_rate": 1.9649095253545734e-05, "loss": 0.9872, "step": 3367 }, { "epoch": 0.11, "learning_rate": 1.9648812164664232e-05, "loss": 1.0061, "step": 3368 }, { "epoch": 0.11, "learning_rate": 1.9648528963680135e-05, "loss": 1.0215, "step": 3369 }, { "epoch": 0.11, "learning_rate": 1.9648245650596726e-05, "loss": 0.9887, "step": 3370 }, { "epoch": 0.11, "learning_rate": 1.964796222541731e-05, "loss": 1.0452, "step": 3371 }, { "epoch": 0.11, "learning_rate": 1.9647678688145163e-05, "loss": 0.9695, "step": 3372 }, { "epoch": 0.11, "learning_rate": 1.9647395038783586e-05, "loss": 1.0539, "step": 3373 }, { "epoch": 0.11, "learning_rate": 1.964711127733588e-05, "loss": 0.9966, "step": 3374 }, { "epoch": 0.11, "learning_rate": 1.964682740380534e-05, "loss": 0.9984, "step": 3375 }, { "epoch": 0.11, "learning_rate": 1.964654341819526e-05, "loss": 0.9598, "step": 3376 }, { "epoch": 0.11, "learning_rate": 1.964625932050894e-05, "loss": 1.0587, "step": 3377 }, { "epoch": 0.11, "learning_rate": 1.9645975110749683e-05, "loss": 1.0809, "step": 3378 }, { "epoch": 0.11, "learning_rate": 1.964569078892079e-05, "loss": 1.0161, "step": 3379 }, { "epoch": 0.11, "learning_rate": 1.9645406355025565e-05, "loss": 0.9725, "step": 3380 }, { "epoch": 0.11, "learning_rate": 1.964512180906731e-05, "loss": 1.0194, "step": 3381 }, { "epoch": 0.11, "learning_rate": 1.9644837151049336e-05, "loss": 1.029, "step": 3382 }, { "epoch": 0.11, "learning_rate": 1.9644552380974945e-05, "loss": 1.0523, "step": 3383 }, { "epoch": 0.11, "learning_rate": 1.964426749884745e-05, "loss": 0.9844, "step": 3384 }, { "epoch": 0.11, "learning_rate": 1.964398250467016e-05, "loss": 1.0314, "step": 3385 }, { "epoch": 0.11, "learning_rate": 1.964369739844638e-05, "loss": 0.9921, "step": 3386 }, { "epoch": 0.11, "learning_rate": 1.964341218017943e-05, "loss": 0.9758, "step": 3387 }, { "epoch": 0.11, "learning_rate": 1.964312684987262e-05, "loss": 1.0075, "step": 3388 }, { "epoch": 0.11, "learning_rate": 1.9642841407529265e-05, "loss": 1.0256, "step": 3389 }, { "epoch": 0.11, "learning_rate": 1.9642555853152687e-05, "loss": 1.0223, "step": 3390 }, { "epoch": 0.11, "learning_rate": 1.9642270186746195e-05, "loss": 1.0226, "step": 3391 }, { "epoch": 0.11, "learning_rate": 1.9641984408313112e-05, "loss": 1.0205, "step": 3392 }, { "epoch": 0.11, "learning_rate": 1.964169851785676e-05, "loss": 1.0352, "step": 3393 }, { "epoch": 0.11, "learning_rate": 1.9641412515380457e-05, "loss": 1.0105, "step": 3394 }, { "epoch": 0.11, "learning_rate": 1.964112640088753e-05, "loss": 1.0005, "step": 3395 }, { "epoch": 0.11, "learning_rate": 1.9640840174381298e-05, "loss": 1.0179, "step": 3396 }, { "epoch": 0.11, "learning_rate": 1.964055383586509e-05, "loss": 1.0563, "step": 3397 }, { "epoch": 0.11, "learning_rate": 1.9640267385342232e-05, "loss": 1.0081, "step": 3398 }, { "epoch": 0.11, "learning_rate": 1.9639980822816055e-05, "loss": 1.0391, "step": 3399 }, { "epoch": 0.11, "learning_rate": 1.9639694148289883e-05, "loss": 0.9798, "step": 3400 }, { "epoch": 0.11, "learning_rate": 1.963940736176705e-05, "loss": 0.9934, "step": 3401 }, { "epoch": 0.11, "learning_rate": 1.9639120463250885e-05, "loss": 1.0453, "step": 3402 }, { "epoch": 0.11, "learning_rate": 1.9638833452744727e-05, "loss": 1.0217, "step": 3403 }, { "epoch": 0.11, "learning_rate": 1.9638546330251903e-05, "loss": 0.9487, "step": 3404 }, { "epoch": 0.11, "learning_rate": 1.9638259095775756e-05, "loss": 1.0203, "step": 3405 }, { "epoch": 0.11, "learning_rate": 1.9637971749319614e-05, "loss": 0.9696, "step": 3406 }, { "epoch": 0.11, "learning_rate": 1.9637684290886828e-05, "loss": 1.0203, "step": 3407 }, { "epoch": 0.11, "learning_rate": 1.9637396720480728e-05, "loss": 1.0314, "step": 3408 }, { "epoch": 0.11, "learning_rate": 1.963710903810466e-05, "loss": 1.0214, "step": 3409 }, { "epoch": 0.11, "learning_rate": 1.9636821243761966e-05, "loss": 0.9793, "step": 3410 }, { "epoch": 0.11, "learning_rate": 1.9636533337455986e-05, "loss": 1.0432, "step": 3411 }, { "epoch": 0.11, "learning_rate": 1.9636245319190065e-05, "loss": 1.0235, "step": 3412 }, { "epoch": 0.11, "learning_rate": 1.9635957188967557e-05, "loss": 1.0205, "step": 3413 }, { "epoch": 0.11, "learning_rate": 1.96356689467918e-05, "loss": 1.0106, "step": 3414 }, { "epoch": 0.11, "learning_rate": 1.9635380592666148e-05, "loss": 1.0029, "step": 3415 }, { "epoch": 0.11, "learning_rate": 1.9635092126593952e-05, "loss": 0.9724, "step": 3416 }, { "epoch": 0.11, "learning_rate": 1.963480354857856e-05, "loss": 1.04, "step": 3417 }, { "epoch": 0.11, "learning_rate": 1.9634514858623327e-05, "loss": 1.0047, "step": 3418 }, { "epoch": 0.11, "learning_rate": 1.9634226056731607e-05, "loss": 1.0192, "step": 3419 }, { "epoch": 0.11, "learning_rate": 1.963393714290675e-05, "loss": 0.9959, "step": 3420 }, { "epoch": 0.11, "learning_rate": 1.9633648117152123e-05, "loss": 1.0091, "step": 3421 }, { "epoch": 0.11, "learning_rate": 1.9633358979471078e-05, "loss": 1.0073, "step": 3422 }, { "epoch": 0.11, "learning_rate": 1.9633069729866975e-05, "loss": 1.0243, "step": 3423 }, { "epoch": 0.11, "learning_rate": 1.9632780368343175e-05, "loss": 0.9376, "step": 3424 }, { "epoch": 0.11, "learning_rate": 1.9632490894903036e-05, "loss": 1.0237, "step": 3425 }, { "epoch": 0.11, "learning_rate": 1.9632201309549926e-05, "loss": 0.9809, "step": 3426 }, { "epoch": 0.11, "learning_rate": 1.963191161228721e-05, "loss": 1.0069, "step": 3427 }, { "epoch": 0.11, "learning_rate": 1.963162180311825e-05, "loss": 0.9651, "step": 3428 }, { "epoch": 0.11, "learning_rate": 1.9631331882046417e-05, "loss": 1.0372, "step": 3429 }, { "epoch": 0.11, "learning_rate": 1.9631041849075074e-05, "loss": 0.9555, "step": 3430 }, { "epoch": 0.11, "learning_rate": 1.9630751704207595e-05, "loss": 1.0003, "step": 3431 }, { "epoch": 0.11, "learning_rate": 1.9630461447447354e-05, "loss": 0.972, "step": 3432 }, { "epoch": 0.11, "learning_rate": 1.9630171078797714e-05, "loss": 0.999, "step": 3433 }, { "epoch": 0.11, "learning_rate": 1.9629880598262056e-05, "loss": 1.0365, "step": 3434 }, { "epoch": 0.11, "learning_rate": 1.962959000584375e-05, "loss": 1.0276, "step": 3435 }, { "epoch": 0.11, "learning_rate": 1.9629299301546178e-05, "loss": 0.971, "step": 3436 }, { "epoch": 0.11, "learning_rate": 1.9629008485372715e-05, "loss": 1.0277, "step": 3437 }, { "epoch": 0.11, "learning_rate": 1.9628717557326734e-05, "loss": 1.0047, "step": 3438 }, { "epoch": 0.11, "learning_rate": 1.9628426517411625e-05, "loss": 1.0131, "step": 3439 }, { "epoch": 0.11, "learning_rate": 1.9628135365630766e-05, "loss": 0.984, "step": 3440 }, { "epoch": 0.11, "learning_rate": 1.9627844101987533e-05, "loss": 1.0228, "step": 3441 }, { "epoch": 0.11, "learning_rate": 1.962755272648532e-05, "loss": 1.0184, "step": 3442 }, { "epoch": 0.11, "learning_rate": 1.9627261239127505e-05, "loss": 1.0292, "step": 3443 }, { "epoch": 0.11, "learning_rate": 1.9626969639917476e-05, "loss": 0.9798, "step": 3444 }, { "epoch": 0.11, "learning_rate": 1.9626677928858624e-05, "loss": 1.0229, "step": 3445 }, { "epoch": 0.11, "learning_rate": 1.9626386105954336e-05, "loss": 0.9632, "step": 3446 }, { "epoch": 0.11, "learning_rate": 1.9626094171208003e-05, "loss": 1.0142, "step": 3447 }, { "epoch": 0.11, "learning_rate": 1.9625802124623015e-05, "loss": 1.0227, "step": 3448 }, { "epoch": 0.11, "learning_rate": 1.9625509966202766e-05, "loss": 1.0109, "step": 3449 }, { "epoch": 0.11, "learning_rate": 1.9625217695950655e-05, "loss": 1.0406, "step": 3450 }, { "epoch": 0.11, "learning_rate": 1.962492531387007e-05, "loss": 1.0414, "step": 3451 }, { "epoch": 0.11, "learning_rate": 1.962463281996441e-05, "loss": 1.0186, "step": 3452 }, { "epoch": 0.11, "learning_rate": 1.962434021423708e-05, "loss": 1.0118, "step": 3453 }, { "epoch": 0.11, "learning_rate": 1.9624047496691473e-05, "loss": 1.0189, "step": 3454 }, { "epoch": 0.11, "learning_rate": 1.962375466733099e-05, "loss": 1.0182, "step": 3455 }, { "epoch": 0.12, "learning_rate": 1.962346172615903e-05, "loss": 1.0013, "step": 3456 }, { "epoch": 0.12, "learning_rate": 1.9623168673179006e-05, "loss": 1.0603, "step": 3457 }, { "epoch": 0.12, "learning_rate": 1.9622875508394318e-05, "loss": 1.0328, "step": 3458 }, { "epoch": 0.12, "learning_rate": 1.9622582231808372e-05, "loss": 1.0003, "step": 3459 }, { "epoch": 0.12, "learning_rate": 1.9622288843424574e-05, "loss": 0.9804, "step": 3460 }, { "epoch": 0.12, "learning_rate": 1.9621995343246335e-05, "loss": 1.017, "step": 3461 }, { "epoch": 0.12, "learning_rate": 1.962170173127706e-05, "loss": 0.9954, "step": 3462 }, { "epoch": 0.12, "learning_rate": 1.9621408007520165e-05, "loss": 1.0447, "step": 3463 }, { "epoch": 0.12, "learning_rate": 1.962111417197907e-05, "loss": 1.0202, "step": 3464 }, { "epoch": 0.12, "learning_rate": 1.962082022465717e-05, "loss": 1.0017, "step": 3465 }, { "epoch": 0.12, "learning_rate": 1.9620526165557893e-05, "loss": 0.9949, "step": 3466 }, { "epoch": 0.12, "learning_rate": 1.9620231994684655e-05, "loss": 1.012, "step": 3467 }, { "epoch": 0.12, "learning_rate": 1.9619937712040872e-05, "loss": 0.9751, "step": 3468 }, { "epoch": 0.12, "learning_rate": 1.9619643317629964e-05, "loss": 1.0015, "step": 3469 }, { "epoch": 0.12, "learning_rate": 1.961934881145535e-05, "loss": 0.9774, "step": 3470 }, { "epoch": 0.12, "learning_rate": 1.9619054193520446e-05, "loss": 1.002, "step": 3471 }, { "epoch": 0.12, "learning_rate": 1.9618759463828686e-05, "loss": 0.9882, "step": 3472 }, { "epoch": 0.12, "learning_rate": 1.961846462238349e-05, "loss": 1.0226, "step": 3473 }, { "epoch": 0.12, "learning_rate": 1.961816966918828e-05, "loss": 1.0396, "step": 3474 }, { "epoch": 0.12, "learning_rate": 1.9617874604246487e-05, "loss": 1.0075, "step": 3475 }, { "epoch": 0.12, "learning_rate": 1.9617579427561536e-05, "loss": 0.9843, "step": 3476 }, { "epoch": 0.12, "learning_rate": 1.961728413913686e-05, "loss": 1.0224, "step": 3477 }, { "epoch": 0.12, "learning_rate": 1.961698873897589e-05, "loss": 1.0345, "step": 3478 }, { "epoch": 0.12, "learning_rate": 1.961669322708205e-05, "loss": 1.0162, "step": 3479 }, { "epoch": 0.12, "learning_rate": 1.9616397603458783e-05, "loss": 1.0133, "step": 3480 }, { "epoch": 0.12, "learning_rate": 1.9616101868109522e-05, "loss": 1.0938, "step": 3481 }, { "epoch": 0.12, "learning_rate": 1.96158060210377e-05, "loss": 0.9996, "step": 3482 }, { "epoch": 0.12, "learning_rate": 1.961551006224675e-05, "loss": 0.9876, "step": 3483 }, { "epoch": 0.12, "learning_rate": 1.9615213991740117e-05, "loss": 0.9602, "step": 3484 }, { "epoch": 0.12, "learning_rate": 1.961491780952124e-05, "loss": 1.0427, "step": 3485 }, { "epoch": 0.12, "learning_rate": 1.9614621515593564e-05, "loss": 1.0251, "step": 3486 }, { "epoch": 0.12, "learning_rate": 1.9614325109960524e-05, "loss": 0.9927, "step": 3487 }, { "epoch": 0.12, "learning_rate": 1.9614028592625564e-05, "loss": 0.9924, "step": 3488 }, { "epoch": 0.12, "learning_rate": 1.9613731963592134e-05, "loss": 0.9968, "step": 3489 }, { "epoch": 0.12, "learning_rate": 1.961343522286368e-05, "loss": 1.0121, "step": 3490 }, { "epoch": 0.12, "learning_rate": 1.9613138370443647e-05, "loss": 1.0405, "step": 3491 }, { "epoch": 0.12, "learning_rate": 1.961284140633548e-05, "loss": 1.0057, "step": 3492 }, { "epoch": 0.12, "learning_rate": 1.961254433054264e-05, "loss": 1.014, "step": 3493 }, { "epoch": 0.12, "learning_rate": 1.961224714306857e-05, "loss": 1.0262, "step": 3494 }, { "epoch": 0.12, "learning_rate": 1.9611949843916725e-05, "loss": 1.0224, "step": 3495 }, { "epoch": 0.12, "learning_rate": 1.9611652433090557e-05, "loss": 0.9702, "step": 3496 }, { "epoch": 0.12, "learning_rate": 1.9611354910593525e-05, "loss": 1.0222, "step": 3497 }, { "epoch": 0.12, "learning_rate": 1.9611057276429085e-05, "loss": 1.009, "step": 3498 }, { "epoch": 0.12, "learning_rate": 1.9610759530600693e-05, "loss": 1.0285, "step": 3499 }, { "epoch": 0.12, "learning_rate": 1.9610461673111812e-05, "loss": 0.9865, "step": 3500 }, { "epoch": 0.12, "learning_rate": 1.9610163703965897e-05, "loss": 1.0482, "step": 3501 }, { "epoch": 0.12, "learning_rate": 1.9609865623166416e-05, "loss": 0.9989, "step": 3502 }, { "epoch": 0.12, "learning_rate": 1.960956743071683e-05, "loss": 1.0455, "step": 3503 }, { "epoch": 0.12, "learning_rate": 1.96092691266206e-05, "loss": 1.0242, "step": 3504 }, { "epoch": 0.12, "learning_rate": 1.9608970710881196e-05, "loss": 1.0113, "step": 3505 }, { "epoch": 0.12, "learning_rate": 1.9608672183502083e-05, "loss": 1.0226, "step": 3506 }, { "epoch": 0.12, "learning_rate": 1.960837354448673e-05, "loss": 1.002, "step": 3507 }, { "epoch": 0.12, "learning_rate": 1.9608074793838606e-05, "loss": 1.0013, "step": 3508 }, { "epoch": 0.12, "learning_rate": 1.9607775931561184e-05, "loss": 1.0125, "step": 3509 }, { "epoch": 0.12, "learning_rate": 1.9607476957657933e-05, "loss": 0.9818, "step": 3510 }, { "epoch": 0.12, "learning_rate": 1.960717787213233e-05, "loss": 1.0166, "step": 3511 }, { "epoch": 0.12, "learning_rate": 1.9606878674987847e-05, "loss": 0.9689, "step": 3512 }, { "epoch": 0.12, "learning_rate": 1.960657936622796e-05, "loss": 1.0278, "step": 3513 }, { "epoch": 0.12, "learning_rate": 1.9606279945856154e-05, "loss": 1.0006, "step": 3514 }, { "epoch": 0.12, "learning_rate": 1.9605980413875897e-05, "loss": 1.0106, "step": 3515 }, { "epoch": 0.12, "learning_rate": 1.9605680770290674e-05, "loss": 0.9774, "step": 3516 }, { "epoch": 0.12, "learning_rate": 1.9605381015103968e-05, "loss": 1.0012, "step": 3517 }, { "epoch": 0.12, "learning_rate": 1.9605081148319258e-05, "loss": 0.9512, "step": 3518 }, { "epoch": 0.12, "learning_rate": 1.960478116994003e-05, "loss": 1.0402, "step": 3519 }, { "epoch": 0.12, "learning_rate": 1.960448107996977e-05, "loss": 1.0062, "step": 3520 }, { "epoch": 0.12, "learning_rate": 1.960418087841196e-05, "loss": 1.0119, "step": 3521 }, { "epoch": 0.12, "learning_rate": 1.9603880565270096e-05, "loss": 0.9653, "step": 3522 }, { "epoch": 0.12, "learning_rate": 1.9603580140547657e-05, "loss": 1.0345, "step": 3523 }, { "epoch": 0.12, "learning_rate": 1.9603279604248146e-05, "loss": 1.0235, "step": 3524 }, { "epoch": 0.12, "learning_rate": 1.960297895637504e-05, "loss": 1.021, "step": 3525 }, { "epoch": 0.12, "learning_rate": 1.960267819693184e-05, "loss": 1.022, "step": 3526 }, { "epoch": 0.12, "learning_rate": 1.9602377325922047e-05, "loss": 0.9726, "step": 3527 }, { "epoch": 0.12, "learning_rate": 1.9602076343349143e-05, "loss": 1.0227, "step": 3528 }, { "epoch": 0.12, "learning_rate": 1.9601775249216633e-05, "loss": 1.0277, "step": 3529 }, { "epoch": 0.12, "learning_rate": 1.9601474043528014e-05, "loss": 1.0003, "step": 3530 }, { "epoch": 0.12, "learning_rate": 1.9601172726286782e-05, "loss": 0.9936, "step": 3531 }, { "epoch": 0.12, "learning_rate": 1.9600871297496445e-05, "loss": 1.0324, "step": 3532 }, { "epoch": 0.12, "learning_rate": 1.96005697571605e-05, "loss": 1.0511, "step": 3533 }, { "epoch": 0.12, "learning_rate": 1.960026810528245e-05, "loss": 0.9797, "step": 3534 }, { "epoch": 0.12, "learning_rate": 1.95999663418658e-05, "loss": 1.0435, "step": 3535 }, { "epoch": 0.12, "learning_rate": 1.959966446691406e-05, "loss": 0.9668, "step": 3536 }, { "epoch": 0.12, "learning_rate": 1.959936248043073e-05, "loss": 1.0083, "step": 3537 }, { "epoch": 0.12, "learning_rate": 1.9599060382419327e-05, "loss": 0.988, "step": 3538 }, { "epoch": 0.12, "learning_rate": 1.9598758172883353e-05, "loss": 1.0091, "step": 3539 }, { "epoch": 0.12, "learning_rate": 1.9598455851826323e-05, "loss": 0.9715, "step": 3540 }, { "epoch": 0.12, "learning_rate": 1.959815341925175e-05, "loss": 0.9942, "step": 3541 }, { "epoch": 0.12, "learning_rate": 1.959785087516315e-05, "loss": 0.9814, "step": 3542 }, { "epoch": 0.12, "learning_rate": 1.959754821956403e-05, "loss": 0.984, "step": 3543 }, { "epoch": 0.12, "learning_rate": 1.9597245452457915e-05, "loss": 0.9807, "step": 3544 }, { "epoch": 0.12, "learning_rate": 1.9596942573848315e-05, "loss": 1.0479, "step": 3545 }, { "epoch": 0.12, "learning_rate": 1.9596639583738757e-05, "loss": 1.0113, "step": 3546 }, { "epoch": 0.12, "learning_rate": 1.9596336482132755e-05, "loss": 1.012, "step": 3547 }, { "epoch": 0.12, "learning_rate": 1.9596033269033833e-05, "loss": 0.9797, "step": 3548 }, { "epoch": 0.12, "learning_rate": 1.9595729944445513e-05, "loss": 1.0387, "step": 3549 }, { "epoch": 0.12, "learning_rate": 1.9595426508371323e-05, "loss": 0.9853, "step": 3550 }, { "epoch": 0.12, "learning_rate": 1.959512296081478e-05, "loss": 1.0114, "step": 3551 }, { "epoch": 0.12, "learning_rate": 1.9594819301779417e-05, "loss": 0.9681, "step": 3552 }, { "epoch": 0.12, "learning_rate": 1.959451553126876e-05, "loss": 0.9975, "step": 3553 }, { "epoch": 0.12, "learning_rate": 1.9594211649286343e-05, "loss": 1.024, "step": 3554 }, { "epoch": 0.12, "learning_rate": 1.9593907655835688e-05, "loss": 1.0153, "step": 3555 }, { "epoch": 0.12, "learning_rate": 1.9593603550920335e-05, "loss": 0.9981, "step": 3556 }, { "epoch": 0.12, "learning_rate": 1.959329933454381e-05, "loss": 0.9975, "step": 3557 }, { "epoch": 0.12, "learning_rate": 1.9592995006709653e-05, "loss": 1.0191, "step": 3558 }, { "epoch": 0.12, "learning_rate": 1.9592690567421397e-05, "loss": 0.9771, "step": 3559 }, { "epoch": 0.12, "learning_rate": 1.9592386016682582e-05, "loss": 0.9781, "step": 3560 }, { "epoch": 0.12, "learning_rate": 1.9592081354496742e-05, "loss": 0.9803, "step": 3561 }, { "epoch": 0.12, "learning_rate": 1.9591776580867418e-05, "loss": 1.009, "step": 3562 }, { "epoch": 0.12, "learning_rate": 1.9591471695798153e-05, "loss": 1.0392, "step": 3563 }, { "epoch": 0.12, "learning_rate": 1.9591166699292488e-05, "loss": 1.0151, "step": 3564 }, { "epoch": 0.12, "learning_rate": 1.9590861591353967e-05, "loss": 1.0218, "step": 3565 }, { "epoch": 0.12, "learning_rate": 1.9590556371986135e-05, "loss": 1.0103, "step": 3566 }, { "epoch": 0.12, "learning_rate": 1.9590251041192537e-05, "loss": 1.0241, "step": 3567 }, { "epoch": 0.12, "learning_rate": 1.958994559897672e-05, "loss": 1.0111, "step": 3568 }, { "epoch": 0.12, "learning_rate": 1.9589640045342232e-05, "loss": 1.0073, "step": 3569 }, { "epoch": 0.12, "learning_rate": 1.9589334380292628e-05, "loss": 0.995, "step": 3570 }, { "epoch": 0.12, "learning_rate": 1.9589028603831453e-05, "loss": 1.0079, "step": 3571 }, { "epoch": 0.12, "learning_rate": 1.9588722715962267e-05, "loss": 0.9163, "step": 3572 }, { "epoch": 0.12, "learning_rate": 1.9588416716688618e-05, "loss": 1.013, "step": 3573 }, { "epoch": 0.12, "learning_rate": 1.9588110606014063e-05, "loss": 1.0127, "step": 3574 }, { "epoch": 0.12, "learning_rate": 1.9587804383942156e-05, "loss": 0.9813, "step": 3575 }, { "epoch": 0.12, "learning_rate": 1.9587498050476458e-05, "loss": 0.986, "step": 3576 }, { "epoch": 0.12, "learning_rate": 1.9587191605620527e-05, "loss": 0.9769, "step": 3577 }, { "epoch": 0.12, "learning_rate": 1.958688504937792e-05, "loss": 1.0265, "step": 3578 }, { "epoch": 0.12, "learning_rate": 1.9586578381752206e-05, "loss": 1.0059, "step": 3579 }, { "epoch": 0.12, "learning_rate": 1.9586271602746944e-05, "loss": 0.9958, "step": 3580 }, { "epoch": 0.12, "learning_rate": 1.9585964712365696e-05, "loss": 0.991, "step": 3581 }, { "epoch": 0.12, "learning_rate": 1.958565771061203e-05, "loss": 0.95, "step": 3582 }, { "epoch": 0.12, "learning_rate": 1.9585350597489516e-05, "loss": 0.976, "step": 3583 }, { "epoch": 0.12, "learning_rate": 1.9585043373001716e-05, "loss": 0.9924, "step": 3584 }, { "epoch": 0.12, "learning_rate": 1.95847360371522e-05, "loss": 1.027, "step": 3585 }, { "epoch": 0.12, "learning_rate": 1.9584428589944548e-05, "loss": 0.9973, "step": 3586 }, { "epoch": 0.12, "learning_rate": 1.9584121031382315e-05, "loss": 0.9926, "step": 3587 }, { "epoch": 0.12, "learning_rate": 1.958381336146909e-05, "loss": 0.9605, "step": 3588 }, { "epoch": 0.12, "learning_rate": 1.958350558020844e-05, "loss": 0.999, "step": 3589 }, { "epoch": 0.12, "learning_rate": 1.9583197687603946e-05, "loss": 1.0107, "step": 3590 }, { "epoch": 0.12, "learning_rate": 1.958288968365918e-05, "loss": 1.0287, "step": 3591 }, { "epoch": 0.12, "learning_rate": 1.9582581568377718e-05, "loss": 0.9814, "step": 3592 }, { "epoch": 0.12, "learning_rate": 1.9582273341763147e-05, "loss": 0.9571, "step": 3593 }, { "epoch": 0.12, "learning_rate": 1.9581965003819045e-05, "loss": 1.0005, "step": 3594 }, { "epoch": 0.12, "learning_rate": 1.9581656554548997e-05, "loss": 1.0222, "step": 3595 }, { "epoch": 0.12, "learning_rate": 1.9581347993956582e-05, "loss": 1.0107, "step": 3596 }, { "epoch": 0.12, "learning_rate": 1.9581039322045387e-05, "loss": 1.0163, "step": 3597 }, { "epoch": 0.12, "learning_rate": 1.9580730538819e-05, "loss": 1.0303, "step": 3598 }, { "epoch": 0.12, "learning_rate": 1.9580421644281006e-05, "loss": 0.9752, "step": 3599 }, { "epoch": 0.12, "learning_rate": 1.9580112638434992e-05, "loss": 0.9807, "step": 3600 }, { "epoch": 0.12, "learning_rate": 1.9579803521284554e-05, "loss": 0.9995, "step": 3601 }, { "epoch": 0.12, "learning_rate": 1.9579494292833278e-05, "loss": 1.0028, "step": 3602 }, { "epoch": 0.12, "learning_rate": 1.957918495308476e-05, "loss": 1.0245, "step": 3603 }, { "epoch": 0.12, "learning_rate": 1.9578875502042592e-05, "loss": 0.9774, "step": 3604 }, { "epoch": 0.12, "learning_rate": 1.957856593971037e-05, "loss": 1.022, "step": 3605 }, { "epoch": 0.12, "learning_rate": 1.9578256266091692e-05, "loss": 1.0295, "step": 3606 }, { "epoch": 0.12, "learning_rate": 1.9577946481190154e-05, "loss": 0.9601, "step": 3607 }, { "epoch": 0.12, "learning_rate": 1.9577636585009355e-05, "loss": 0.9475, "step": 3608 }, { "epoch": 0.12, "learning_rate": 1.95773265775529e-05, "loss": 0.996, "step": 3609 }, { "epoch": 0.12, "learning_rate": 1.957701645882438e-05, "loss": 1.0164, "step": 3610 }, { "epoch": 0.12, "learning_rate": 1.957670622882741e-05, "loss": 1.0079, "step": 3611 }, { "epoch": 0.12, "learning_rate": 1.957639588756559e-05, "loss": 1.0124, "step": 3612 }, { "epoch": 0.12, "learning_rate": 1.957608543504252e-05, "loss": 0.9894, "step": 3613 }, { "epoch": 0.12, "learning_rate": 1.957577487126182e-05, "loss": 0.9763, "step": 3614 }, { "epoch": 0.12, "learning_rate": 1.9575464196227082e-05, "loss": 0.9847, "step": 3615 }, { "epoch": 0.12, "learning_rate": 1.957515340994193e-05, "loss": 1.0203, "step": 3616 }, { "epoch": 0.12, "learning_rate": 1.9574842512409965e-05, "loss": 0.9779, "step": 3617 }, { "epoch": 0.12, "learning_rate": 1.95745315036348e-05, "loss": 0.9753, "step": 3618 }, { "epoch": 0.12, "learning_rate": 1.9574220383620054e-05, "loss": 1.0081, "step": 3619 }, { "epoch": 0.12, "learning_rate": 1.9573909152369342e-05, "loss": 0.9877, "step": 3620 }, { "epoch": 0.12, "learning_rate": 1.957359780988627e-05, "loss": 0.9877, "step": 3621 }, { "epoch": 0.12, "learning_rate": 1.9573286356174467e-05, "loss": 0.9922, "step": 3622 }, { "epoch": 0.12, "learning_rate": 1.9572974791237548e-05, "loss": 0.986, "step": 3623 }, { "epoch": 0.12, "learning_rate": 1.9572663115079125e-05, "loss": 0.9839, "step": 3624 }, { "epoch": 0.12, "learning_rate": 1.957235132770283e-05, "loss": 1.0033, "step": 3625 }, { "epoch": 0.12, "learning_rate": 1.9572039429112277e-05, "loss": 1.0282, "step": 3626 }, { "epoch": 0.12, "learning_rate": 1.9571727419311096e-05, "loss": 1.0334, "step": 3627 }, { "epoch": 0.12, "learning_rate": 1.9571415298302908e-05, "loss": 0.9958, "step": 3628 }, { "epoch": 0.12, "learning_rate": 1.9571103066091343e-05, "loss": 1.01, "step": 3629 }, { "epoch": 0.12, "learning_rate": 1.9570790722680025e-05, "loss": 0.9819, "step": 3630 }, { "epoch": 0.12, "learning_rate": 1.9570478268072587e-05, "loss": 0.9617, "step": 3631 }, { "epoch": 0.12, "learning_rate": 1.9570165702272654e-05, "loss": 0.9928, "step": 3632 }, { "epoch": 0.12, "learning_rate": 1.956985302528386e-05, "loss": 1.0254, "step": 3633 }, { "epoch": 0.12, "learning_rate": 1.9569540237109837e-05, "loss": 0.9611, "step": 3634 }, { "epoch": 0.12, "learning_rate": 1.9569227337754225e-05, "loss": 0.9872, "step": 3635 }, { "epoch": 0.12, "learning_rate": 1.956891432722065e-05, "loss": 0.9912, "step": 3636 }, { "epoch": 0.12, "learning_rate": 1.956860120551275e-05, "loss": 1.024, "step": 3637 }, { "epoch": 0.12, "learning_rate": 1.956828797263417e-05, "loss": 0.9487, "step": 3638 }, { "epoch": 0.12, "learning_rate": 1.9567974628588543e-05, "loss": 1.024, "step": 3639 }, { "epoch": 0.12, "learning_rate": 1.9567661173379515e-05, "loss": 0.9952, "step": 3640 }, { "epoch": 0.12, "learning_rate": 1.9567347607010722e-05, "loss": 0.9887, "step": 3641 }, { "epoch": 0.12, "learning_rate": 1.956703392948581e-05, "loss": 0.9841, "step": 3642 }, { "epoch": 0.12, "learning_rate": 1.956672014080842e-05, "loss": 1.0332, "step": 3643 }, { "epoch": 0.12, "learning_rate": 1.9566406240982204e-05, "loss": 0.9652, "step": 3644 }, { "epoch": 0.12, "learning_rate": 1.9566092230010808e-05, "loss": 0.9983, "step": 3645 }, { "epoch": 0.12, "learning_rate": 1.9565778107897874e-05, "loss": 1.0124, "step": 3646 }, { "epoch": 0.12, "learning_rate": 1.956546387464705e-05, "loss": 1.0027, "step": 3647 }, { "epoch": 0.12, "learning_rate": 1.9565149530262003e-05, "loss": 0.9845, "step": 3648 }, { "epoch": 0.12, "learning_rate": 1.9564835074746367e-05, "loss": 0.9953, "step": 3649 }, { "epoch": 0.12, "learning_rate": 1.9564520508103806e-05, "loss": 1.0042, "step": 3650 }, { "epoch": 0.12, "learning_rate": 1.956420583033797e-05, "loss": 0.983, "step": 3651 }, { "epoch": 0.12, "learning_rate": 1.9563891041452515e-05, "loss": 0.984, "step": 3652 }, { "epoch": 0.12, "learning_rate": 1.9563576141451102e-05, "loss": 1.0099, "step": 3653 }, { "epoch": 0.12, "learning_rate": 1.9563261130337387e-05, "loss": 0.9584, "step": 3654 }, { "epoch": 0.12, "learning_rate": 1.956294600811503e-05, "loss": 1.0192, "step": 3655 }, { "epoch": 0.12, "learning_rate": 1.9562630774787695e-05, "loss": 0.9362, "step": 3656 }, { "epoch": 0.12, "learning_rate": 1.9562315430359038e-05, "loss": 1.0256, "step": 3657 }, { "epoch": 0.12, "learning_rate": 1.9561999974832727e-05, "loss": 1.0349, "step": 3658 }, { "epoch": 0.12, "learning_rate": 1.956168440821243e-05, "loss": 0.9941, "step": 3659 }, { "epoch": 0.12, "learning_rate": 1.9561368730501808e-05, "loss": 0.9839, "step": 3660 }, { "epoch": 0.12, "learning_rate": 1.956105294170453e-05, "loss": 1.0144, "step": 3661 }, { "epoch": 0.12, "learning_rate": 1.9560737041824266e-05, "loss": 1.0005, "step": 3662 }, { "epoch": 0.12, "learning_rate": 1.956042103086469e-05, "loss": 1.024, "step": 3663 }, { "epoch": 0.12, "learning_rate": 1.9560104908829466e-05, "loss": 0.9785, "step": 3664 }, { "epoch": 0.12, "learning_rate": 1.9559788675722268e-05, "loss": 1.0052, "step": 3665 }, { "epoch": 0.12, "learning_rate": 1.9559472331546776e-05, "loss": 0.9946, "step": 3666 }, { "epoch": 0.12, "learning_rate": 1.9559155876306658e-05, "loss": 1.0313, "step": 3667 }, { "epoch": 0.12, "learning_rate": 1.95588393100056e-05, "loss": 1.0073, "step": 3668 }, { "epoch": 0.12, "learning_rate": 1.9558522632647268e-05, "loss": 1.0037, "step": 3669 }, { "epoch": 0.12, "learning_rate": 1.955820584423535e-05, "loss": 1.0089, "step": 3670 }, { "epoch": 0.12, "learning_rate": 1.9557888944773524e-05, "loss": 1.0233, "step": 3671 }, { "epoch": 0.12, "learning_rate": 1.955757193426547e-05, "loss": 0.9923, "step": 3672 }, { "epoch": 0.12, "learning_rate": 1.9557254812714873e-05, "loss": 0.9772, "step": 3673 }, { "epoch": 0.12, "learning_rate": 1.9556937580125422e-05, "loss": 0.988, "step": 3674 }, { "epoch": 0.12, "learning_rate": 1.9556620236500794e-05, "loss": 1.0141, "step": 3675 }, { "epoch": 0.12, "learning_rate": 1.955630278184468e-05, "loss": 0.9924, "step": 3676 }, { "epoch": 0.12, "learning_rate": 1.955598521616077e-05, "loss": 1.0175, "step": 3677 }, { "epoch": 0.12, "learning_rate": 1.9555667539452753e-05, "loss": 0.9882, "step": 3678 }, { "epoch": 0.12, "learning_rate": 1.9555349751724315e-05, "loss": 0.9885, "step": 3679 }, { "epoch": 0.12, "learning_rate": 1.9555031852979156e-05, "loss": 0.9675, "step": 3680 }, { "epoch": 0.12, "learning_rate": 1.9554713843220963e-05, "loss": 1.0247, "step": 3681 }, { "epoch": 0.12, "learning_rate": 1.9554395722453435e-05, "loss": 1.0437, "step": 3682 }, { "epoch": 0.12, "learning_rate": 1.955407749068026e-05, "loss": 1.0122, "step": 3683 }, { "epoch": 0.12, "learning_rate": 1.9553759147905146e-05, "loss": 1.0141, "step": 3684 }, { "epoch": 0.12, "learning_rate": 1.955344069413179e-05, "loss": 1.0179, "step": 3685 }, { "epoch": 0.12, "learning_rate": 1.9553122129363884e-05, "loss": 1.0323, "step": 3686 }, { "epoch": 0.12, "learning_rate": 1.955280345360514e-05, "loss": 1.0151, "step": 3687 }, { "epoch": 0.12, "learning_rate": 1.9552484666859247e-05, "loss": 1.004, "step": 3688 }, { "epoch": 0.12, "learning_rate": 1.9552165769129917e-05, "loss": 1.058, "step": 3689 }, { "epoch": 0.12, "learning_rate": 1.955184676042086e-05, "loss": 1.0169, "step": 3690 }, { "epoch": 0.12, "learning_rate": 1.9551527640735774e-05, "loss": 1.029, "step": 3691 }, { "epoch": 0.12, "learning_rate": 1.955120841007837e-05, "loss": 0.9704, "step": 3692 }, { "epoch": 0.12, "learning_rate": 1.9550889068452354e-05, "loss": 1.0094, "step": 3693 }, { "epoch": 0.12, "learning_rate": 1.9550569615861438e-05, "loss": 1.0338, "step": 3694 }, { "epoch": 0.12, "learning_rate": 1.9550250052309334e-05, "loss": 0.9979, "step": 3695 }, { "epoch": 0.12, "learning_rate": 1.9549930377799757e-05, "loss": 1.01, "step": 3696 }, { "epoch": 0.12, "learning_rate": 1.9549610592336416e-05, "loss": 0.9704, "step": 3697 }, { "epoch": 0.12, "learning_rate": 1.954929069592303e-05, "loss": 0.9906, "step": 3698 }, { "epoch": 0.12, "learning_rate": 1.9548970688563315e-05, "loss": 1.0455, "step": 3699 }, { "epoch": 0.12, "learning_rate": 1.954865057026099e-05, "loss": 1.0108, "step": 3700 }, { "epoch": 0.12, "learning_rate": 1.9548330341019774e-05, "loss": 1.0324, "step": 3701 }, { "epoch": 0.12, "learning_rate": 1.954801000084338e-05, "loss": 0.9848, "step": 3702 }, { "epoch": 0.12, "learning_rate": 1.9547689549735542e-05, "loss": 1.0391, "step": 3703 }, { "epoch": 0.12, "learning_rate": 1.9547368987699974e-05, "loss": 1.0003, "step": 3704 }, { "epoch": 0.12, "learning_rate": 1.9547048314740403e-05, "loss": 1.015, "step": 3705 }, { "epoch": 0.12, "learning_rate": 1.954672753086056e-05, "loss": 1.015, "step": 3706 }, { "epoch": 0.12, "learning_rate": 1.9546406636064165e-05, "loss": 1.0311, "step": 3707 }, { "epoch": 0.12, "learning_rate": 1.9546085630354948e-05, "loss": 0.9641, "step": 3708 }, { "epoch": 0.12, "learning_rate": 1.954576451373664e-05, "loss": 0.9929, "step": 3709 }, { "epoch": 0.12, "learning_rate": 1.9545443286212967e-05, "loss": 1.0118, "step": 3710 }, { "epoch": 0.12, "learning_rate": 1.954512194778767e-05, "loss": 0.9776, "step": 3711 }, { "epoch": 0.12, "learning_rate": 1.9544800498464476e-05, "loss": 0.967, "step": 3712 }, { "epoch": 0.12, "learning_rate": 1.9544478938247122e-05, "loss": 1.03, "step": 3713 }, { "epoch": 0.12, "learning_rate": 1.9544157267139342e-05, "loss": 1.0128, "step": 3714 }, { "epoch": 0.12, "learning_rate": 1.9543835485144873e-05, "loss": 1.008, "step": 3715 }, { "epoch": 0.12, "learning_rate": 1.9543513592267452e-05, "loss": 0.9973, "step": 3716 }, { "epoch": 0.12, "learning_rate": 1.954319158851083e-05, "loss": 1.0195, "step": 3717 }, { "epoch": 0.12, "learning_rate": 1.9542869473878733e-05, "loss": 0.9978, "step": 3718 }, { "epoch": 0.12, "learning_rate": 1.9542547248374912e-05, "loss": 1.0191, "step": 3719 }, { "epoch": 0.12, "learning_rate": 1.9542224912003108e-05, "loss": 0.9876, "step": 3720 }, { "epoch": 0.12, "learning_rate": 1.9541902464767072e-05, "loss": 1.0331, "step": 3721 }, { "epoch": 0.12, "learning_rate": 1.954157990667054e-05, "loss": 1.0023, "step": 3722 }, { "epoch": 0.12, "learning_rate": 1.9541257237717267e-05, "loss": 1.0391, "step": 3723 }, { "epoch": 0.12, "learning_rate": 1.9540934457911e-05, "loss": 0.9729, "step": 3724 }, { "epoch": 0.12, "learning_rate": 1.954061156725549e-05, "loss": 0.9909, "step": 3725 }, { "epoch": 0.12, "learning_rate": 1.9540288565754482e-05, "loss": 0.9792, "step": 3726 }, { "epoch": 0.12, "learning_rate": 1.953996545341174e-05, "loss": 1.0167, "step": 3727 }, { "epoch": 0.12, "learning_rate": 1.9539642230231006e-05, "loss": 1.0217, "step": 3728 }, { "epoch": 0.12, "learning_rate": 1.9539318896216044e-05, "loss": 1.0067, "step": 3729 }, { "epoch": 0.12, "learning_rate": 1.953899545137061e-05, "loss": 1.0254, "step": 3730 }, { "epoch": 0.12, "learning_rate": 1.9538671895698457e-05, "loss": 1.0199, "step": 3731 }, { "epoch": 0.12, "learning_rate": 1.9538348229203348e-05, "loss": 0.9775, "step": 3732 }, { "epoch": 0.12, "learning_rate": 1.9538024451889042e-05, "loss": 1.0089, "step": 3733 }, { "epoch": 0.12, "learning_rate": 1.9537700563759303e-05, "loss": 0.9941, "step": 3734 }, { "epoch": 0.12, "learning_rate": 1.9537376564817888e-05, "loss": 1.0195, "step": 3735 }, { "epoch": 0.12, "learning_rate": 1.953705245506857e-05, "loss": 0.9642, "step": 3736 }, { "epoch": 0.12, "learning_rate": 1.9536728234515107e-05, "loss": 1.0291, "step": 3737 }, { "epoch": 0.12, "learning_rate": 1.953640390316127e-05, "loss": 1.0087, "step": 3738 }, { "epoch": 0.12, "learning_rate": 1.9536079461010828e-05, "loss": 0.9794, "step": 3739 }, { "epoch": 0.12, "learning_rate": 1.9535754908067544e-05, "loss": 0.9617, "step": 3740 }, { "epoch": 0.12, "learning_rate": 1.9535430244335197e-05, "loss": 1.0021, "step": 3741 }, { "epoch": 0.12, "learning_rate": 1.9535105469817553e-05, "loss": 1.0117, "step": 3742 }, { "epoch": 0.12, "learning_rate": 1.953478058451839e-05, "loss": 0.9879, "step": 3743 }, { "epoch": 0.12, "learning_rate": 1.953445558844148e-05, "loss": 0.9962, "step": 3744 }, { "epoch": 0.12, "learning_rate": 1.9534130481590593e-05, "loss": 0.9932, "step": 3745 }, { "epoch": 0.12, "learning_rate": 1.953380526396952e-05, "loss": 1.0403, "step": 3746 }, { "epoch": 0.12, "learning_rate": 1.953347993558203e-05, "loss": 1.0304, "step": 3747 }, { "epoch": 0.12, "learning_rate": 1.9533154496431904e-05, "loss": 1.0039, "step": 3748 }, { "epoch": 0.12, "learning_rate": 1.953282894652292e-05, "loss": 1.0364, "step": 3749 }, { "epoch": 0.12, "learning_rate": 1.953250328585887e-05, "loss": 1.0055, "step": 3750 }, { "epoch": 0.12, "learning_rate": 1.9532177514443525e-05, "loss": 1.0522, "step": 3751 }, { "epoch": 0.12, "learning_rate": 1.953185163228068e-05, "loss": 0.9687, "step": 3752 }, { "epoch": 0.12, "learning_rate": 1.953152563937412e-05, "loss": 0.9916, "step": 3753 }, { "epoch": 0.12, "learning_rate": 1.953119953572763e-05, "loss": 1.0235, "step": 3754 }, { "epoch": 0.12, "learning_rate": 1.953087332134499e-05, "loss": 0.9981, "step": 3755 }, { "epoch": 0.12, "learning_rate": 1.953054699623001e-05, "loss": 0.9981, "step": 3756 }, { "epoch": 0.13, "learning_rate": 1.9530220560386467e-05, "loss": 1.0032, "step": 3757 }, { "epoch": 0.13, "learning_rate": 1.952989401381815e-05, "loss": 1.0218, "step": 3758 }, { "epoch": 0.13, "learning_rate": 1.952956735652887e-05, "loss": 0.9409, "step": 3759 }, { "epoch": 0.13, "learning_rate": 1.9529240588522404e-05, "loss": 0.9682, "step": 3760 }, { "epoch": 0.13, "learning_rate": 1.9528913709802562e-05, "loss": 1.0272, "step": 3761 }, { "epoch": 0.13, "learning_rate": 1.9528586720373136e-05, "loss": 1.0224, "step": 3762 }, { "epoch": 0.13, "learning_rate": 1.9528259620237923e-05, "loss": 1.0263, "step": 3763 }, { "epoch": 0.13, "learning_rate": 1.952793240940072e-05, "loss": 0.9568, "step": 3764 }, { "epoch": 0.13, "learning_rate": 1.9527605087865343e-05, "loss": 1.02, "step": 3765 }, { "epoch": 0.13, "learning_rate": 1.9527277655635582e-05, "loss": 0.9931, "step": 3766 }, { "epoch": 0.13, "learning_rate": 1.9526950112715248e-05, "loss": 0.9838, "step": 3767 }, { "epoch": 0.13, "learning_rate": 1.952662245910814e-05, "loss": 0.9849, "step": 3768 }, { "epoch": 0.13, "learning_rate": 1.9526294694818074e-05, "loss": 1.0005, "step": 3769 }, { "epoch": 0.13, "learning_rate": 1.952596681984885e-05, "loss": 1.0852, "step": 3770 }, { "epoch": 0.13, "learning_rate": 1.952563883420428e-05, "loss": 1.0282, "step": 3771 }, { "epoch": 0.13, "learning_rate": 1.9525310737888173e-05, "loss": 1.0107, "step": 3772 }, { "epoch": 0.13, "learning_rate": 1.9524982530904345e-05, "loss": 0.9997, "step": 3773 }, { "epoch": 0.13, "learning_rate": 1.9524654213256608e-05, "loss": 1.0169, "step": 3774 }, { "epoch": 0.13, "learning_rate": 1.952432578494877e-05, "loss": 0.9916, "step": 3775 }, { "epoch": 0.13, "learning_rate": 1.9523997245984656e-05, "loss": 0.9799, "step": 3776 }, { "epoch": 0.13, "learning_rate": 1.9523668596368078e-05, "loss": 1.0252, "step": 3777 }, { "epoch": 0.13, "learning_rate": 1.9523339836102858e-05, "loss": 1.0097, "step": 3778 }, { "epoch": 0.13, "learning_rate": 1.9523010965192807e-05, "loss": 1.0002, "step": 3779 }, { "epoch": 0.13, "learning_rate": 1.9522681983641757e-05, "loss": 1.0124, "step": 3780 }, { "epoch": 0.13, "learning_rate": 1.9522352891453523e-05, "loss": 0.9881, "step": 3781 }, { "epoch": 0.13, "learning_rate": 1.952202368863193e-05, "loss": 1.0011, "step": 3782 }, { "epoch": 0.13, "learning_rate": 1.9521694375180806e-05, "loss": 1.0206, "step": 3783 }, { "epoch": 0.13, "learning_rate": 1.9521364951103973e-05, "loss": 1.0027, "step": 3784 }, { "epoch": 0.13, "learning_rate": 1.952103541640526e-05, "loss": 1.015, "step": 3785 }, { "epoch": 0.13, "learning_rate": 1.9520705771088496e-05, "loss": 1.0067, "step": 3786 }, { "epoch": 0.13, "learning_rate": 1.952037601515751e-05, "loss": 1.0225, "step": 3787 }, { "epoch": 0.13, "learning_rate": 1.9520046148616134e-05, "loss": 0.9721, "step": 3788 }, { "epoch": 0.13, "learning_rate": 1.95197161714682e-05, "loss": 1.0388, "step": 3789 }, { "epoch": 0.13, "learning_rate": 1.951938608371754e-05, "loss": 0.9872, "step": 3790 }, { "epoch": 0.13, "learning_rate": 1.9519055885367993e-05, "loss": 0.9828, "step": 3791 }, { "epoch": 0.13, "learning_rate": 1.951872557642339e-05, "loss": 0.9804, "step": 3792 }, { "epoch": 0.13, "learning_rate": 1.9518395156887574e-05, "loss": 1.053, "step": 3793 }, { "epoch": 0.13, "learning_rate": 1.9518064626764384e-05, "loss": 1.0047, "step": 3794 }, { "epoch": 0.13, "learning_rate": 1.9517733986057656e-05, "loss": 1.0143, "step": 3795 }, { "epoch": 0.13, "learning_rate": 1.951740323477123e-05, "loss": 1.0377, "step": 3796 }, { "epoch": 0.13, "learning_rate": 1.951707237290896e-05, "loss": 1.0115, "step": 3797 }, { "epoch": 0.13, "learning_rate": 1.9516741400474675e-05, "loss": 0.9962, "step": 3798 }, { "epoch": 0.13, "learning_rate": 1.951641031747223e-05, "loss": 1.0033, "step": 3799 }, { "epoch": 0.13, "learning_rate": 1.951607912390547e-05, "loss": 1.0194, "step": 3800 }, { "epoch": 0.13, "learning_rate": 1.9515747819778238e-05, "loss": 0.9893, "step": 3801 }, { "epoch": 0.13, "learning_rate": 1.951541640509439e-05, "loss": 0.9735, "step": 3802 }, { "epoch": 0.13, "learning_rate": 1.9515084879857776e-05, "loss": 1.0, "step": 3803 }, { "epoch": 0.13, "learning_rate": 1.9514753244072242e-05, "loss": 0.9837, "step": 3804 }, { "epoch": 0.13, "learning_rate": 1.9514421497741647e-05, "loss": 1.0382, "step": 3805 }, { "epoch": 0.13, "learning_rate": 1.951408964086984e-05, "loss": 0.9764, "step": 3806 }, { "epoch": 0.13, "learning_rate": 1.9513757673460682e-05, "loss": 1.05, "step": 3807 }, { "epoch": 0.13, "learning_rate": 1.9513425595518027e-05, "loss": 0.9804, "step": 3808 }, { "epoch": 0.13, "learning_rate": 1.9513093407045734e-05, "loss": 1.0424, "step": 3809 }, { "epoch": 0.13, "learning_rate": 1.951276110804766e-05, "loss": 1.0407, "step": 3810 }, { "epoch": 0.13, "learning_rate": 1.9512428698527668e-05, "loss": 1.0185, "step": 3811 }, { "epoch": 0.13, "learning_rate": 1.951209617848962e-05, "loss": 0.97, "step": 3812 }, { "epoch": 0.13, "learning_rate": 1.9511763547937378e-05, "loss": 0.9863, "step": 3813 }, { "epoch": 0.13, "learning_rate": 1.951143080687481e-05, "loss": 1.0202, "step": 3814 }, { "epoch": 0.13, "learning_rate": 1.9511097955305776e-05, "loss": 1.008, "step": 3815 }, { "epoch": 0.13, "learning_rate": 1.951076499323415e-05, "loss": 1.0005, "step": 3816 }, { "epoch": 0.13, "learning_rate": 1.9510431920663797e-05, "loss": 1.049, "step": 3817 }, { "epoch": 0.13, "learning_rate": 1.9510098737598584e-05, "loss": 1.0173, "step": 3818 }, { "epoch": 0.13, "learning_rate": 1.9509765444042388e-05, "loss": 0.9886, "step": 3819 }, { "epoch": 0.13, "learning_rate": 1.9509432039999075e-05, "loss": 0.9826, "step": 3820 }, { "epoch": 0.13, "learning_rate": 1.9509098525472522e-05, "loss": 0.9829, "step": 3821 }, { "epoch": 0.13, "learning_rate": 1.9508764900466608e-05, "loss": 0.9776, "step": 3822 }, { "epoch": 0.13, "learning_rate": 1.9508431164985202e-05, "loss": 1.0218, "step": 3823 }, { "epoch": 0.13, "learning_rate": 1.9508097319032184e-05, "loss": 0.9856, "step": 3824 }, { "epoch": 0.13, "learning_rate": 1.9507763362611432e-05, "loss": 1.0461, "step": 3825 }, { "epoch": 0.13, "learning_rate": 1.9507429295726827e-05, "loss": 1.037, "step": 3826 }, { "epoch": 0.13, "learning_rate": 1.950709511838225e-05, "loss": 1.0077, "step": 3827 }, { "epoch": 0.13, "learning_rate": 1.9506760830581587e-05, "loss": 0.9869, "step": 3828 }, { "epoch": 0.13, "learning_rate": 1.9506426432328714e-05, "loss": 1.0397, "step": 3829 }, { "epoch": 0.13, "learning_rate": 1.9506091923627523e-05, "loss": 0.9844, "step": 3830 }, { "epoch": 0.13, "learning_rate": 1.9505757304481898e-05, "loss": 1.0225, "step": 3831 }, { "epoch": 0.13, "learning_rate": 1.9505422574895723e-05, "loss": 1.032, "step": 3832 }, { "epoch": 0.13, "learning_rate": 1.95050877348729e-05, "loss": 1.0466, "step": 3833 }, { "epoch": 0.13, "learning_rate": 1.9504752784417302e-05, "loss": 1.0061, "step": 3834 }, { "epoch": 0.13, "learning_rate": 1.9504417723532832e-05, "loss": 1.0516, "step": 3835 }, { "epoch": 0.13, "learning_rate": 1.9504082552223377e-05, "loss": 0.9975, "step": 3836 }, { "epoch": 0.13, "learning_rate": 1.9503747270492836e-05, "loss": 1.0201, "step": 3837 }, { "epoch": 0.13, "learning_rate": 1.9503411878345103e-05, "loss": 1.0072, "step": 3838 }, { "epoch": 0.13, "learning_rate": 1.950307637578407e-05, "loss": 1.0278, "step": 3839 }, { "epoch": 0.13, "learning_rate": 1.950274076281364e-05, "loss": 1.0027, "step": 3840 }, { "epoch": 0.13, "learning_rate": 1.9502405039437714e-05, "loss": 0.9927, "step": 3841 }, { "epoch": 0.13, "learning_rate": 1.9502069205660185e-05, "loss": 0.9666, "step": 3842 }, { "epoch": 0.13, "learning_rate": 1.950173326148496e-05, "loss": 1.0043, "step": 3843 }, { "epoch": 0.13, "learning_rate": 1.950139720691594e-05, "loss": 0.9972, "step": 3844 }, { "epoch": 0.13, "learning_rate": 1.9501061041957034e-05, "loss": 1.0332, "step": 3845 }, { "epoch": 0.13, "learning_rate": 1.9500724766612143e-05, "loss": 1.0113, "step": 3846 }, { "epoch": 0.13, "learning_rate": 1.9500388380885174e-05, "loss": 1.0216, "step": 3847 }, { "epoch": 0.13, "learning_rate": 1.950005188478004e-05, "loss": 1.011, "step": 3848 }, { "epoch": 0.13, "learning_rate": 1.9499715278300644e-05, "loss": 1.0204, "step": 3849 }, { "epoch": 0.13, "learning_rate": 1.9499378561450898e-05, "loss": 1.0132, "step": 3850 }, { "epoch": 0.13, "learning_rate": 1.949904173423472e-05, "loss": 1.0224, "step": 3851 }, { "epoch": 0.13, "learning_rate": 1.949870479665602e-05, "loss": 0.9665, "step": 3852 }, { "epoch": 0.13, "learning_rate": 1.9498367748718706e-05, "loss": 1.0546, "step": 3853 }, { "epoch": 0.13, "learning_rate": 1.9498030590426702e-05, "loss": 0.9967, "step": 3854 }, { "epoch": 0.13, "learning_rate": 1.9497693321783926e-05, "loss": 1.0166, "step": 3855 }, { "epoch": 0.13, "learning_rate": 1.949735594279429e-05, "loss": 0.9758, "step": 3856 }, { "epoch": 0.13, "learning_rate": 1.949701845346172e-05, "loss": 1.0289, "step": 3857 }, { "epoch": 0.13, "learning_rate": 1.9496680853790133e-05, "loss": 0.9497, "step": 3858 }, { "epoch": 0.13, "learning_rate": 1.9496343143783452e-05, "loss": 1.0073, "step": 3859 }, { "epoch": 0.13, "learning_rate": 1.94960053234456e-05, "loss": 1.004, "step": 3860 }, { "epoch": 0.13, "learning_rate": 1.9495667392780504e-05, "loss": 0.9623, "step": 3861 }, { "epoch": 0.13, "learning_rate": 1.949532935179209e-05, "loss": 1.0323, "step": 3862 }, { "epoch": 0.13, "learning_rate": 1.9494991200484287e-05, "loss": 1.0159, "step": 3863 }, { "epoch": 0.13, "learning_rate": 1.949465293886102e-05, "loss": 0.9806, "step": 3864 }, { "epoch": 0.13, "learning_rate": 1.9494314566926218e-05, "loss": 0.983, "step": 3865 }, { "epoch": 0.13, "learning_rate": 1.9493976084683814e-05, "loss": 1.002, "step": 3866 }, { "epoch": 0.13, "learning_rate": 1.9493637492137744e-05, "loss": 1.037, "step": 3867 }, { "epoch": 0.13, "learning_rate": 1.9493298789291937e-05, "loss": 0.9644, "step": 3868 }, { "epoch": 0.13, "learning_rate": 1.9492959976150332e-05, "loss": 1.0391, "step": 3869 }, { "epoch": 0.13, "learning_rate": 1.9492621052716865e-05, "loss": 1.0044, "step": 3870 }, { "epoch": 0.13, "learning_rate": 1.9492282018995473e-05, "loss": 1.0397, "step": 3871 }, { "epoch": 0.13, "learning_rate": 1.949194287499009e-05, "loss": 0.9631, "step": 3872 }, { "epoch": 0.13, "learning_rate": 1.9491603620704663e-05, "loss": 0.9697, "step": 3873 }, { "epoch": 0.13, "learning_rate": 1.949126425614313e-05, "loss": 1.0048, "step": 3874 }, { "epoch": 0.13, "learning_rate": 1.949092478130944e-05, "loss": 1.0346, "step": 3875 }, { "epoch": 0.13, "learning_rate": 1.9490585196207527e-05, "loss": 0.9837, "step": 3876 }, { "epoch": 0.13, "learning_rate": 1.9490245500841344e-05, "loss": 0.9725, "step": 3877 }, { "epoch": 0.13, "learning_rate": 1.9489905695214836e-05, "loss": 0.9899, "step": 3878 }, { "epoch": 0.13, "learning_rate": 1.9489565779331946e-05, "loss": 1.0204, "step": 3879 }, { "epoch": 0.13, "learning_rate": 1.9489225753196633e-05, "loss": 0.958, "step": 3880 }, { "epoch": 0.13, "learning_rate": 1.948888561681284e-05, "loss": 1.0117, "step": 3881 }, { "epoch": 0.13, "learning_rate": 1.948854537018452e-05, "loss": 0.9954, "step": 3882 }, { "epoch": 0.13, "learning_rate": 1.9488205013315627e-05, "loss": 1.0056, "step": 3883 }, { "epoch": 0.13, "learning_rate": 1.9487864546210113e-05, "loss": 1.0321, "step": 3884 }, { "epoch": 0.13, "learning_rate": 1.948752396887194e-05, "loss": 1.0103, "step": 3885 }, { "epoch": 0.13, "learning_rate": 1.9487183281305058e-05, "loss": 0.9917, "step": 3886 }, { "epoch": 0.13, "learning_rate": 1.948684248351343e-05, "loss": 0.967, "step": 3887 }, { "epoch": 0.13, "learning_rate": 1.9486501575501015e-05, "loss": 0.9327, "step": 3888 }, { "epoch": 0.13, "learning_rate": 1.948616055727177e-05, "loss": 1.0133, "step": 3889 }, { "epoch": 0.13, "learning_rate": 1.9485819428829655e-05, "loss": 0.9694, "step": 3890 }, { "epoch": 0.13, "learning_rate": 1.9485478190178644e-05, "loss": 1.0484, "step": 3891 }, { "epoch": 0.13, "learning_rate": 1.9485136841322694e-05, "loss": 0.9886, "step": 3892 }, { "epoch": 0.13, "learning_rate": 1.9484795382265772e-05, "loss": 1.0767, "step": 3893 }, { "epoch": 0.13, "learning_rate": 1.948445381301184e-05, "loss": 0.9671, "step": 3894 }, { "epoch": 0.13, "learning_rate": 1.948411213356488e-05, "loss": 1.0188, "step": 3895 }, { "epoch": 0.13, "learning_rate": 1.9483770343928847e-05, "loss": 0.9804, "step": 3896 }, { "epoch": 0.13, "learning_rate": 1.9483428444107723e-05, "loss": 1.0127, "step": 3897 }, { "epoch": 0.13, "learning_rate": 1.948308643410547e-05, "loss": 0.9732, "step": 3898 }, { "epoch": 0.13, "learning_rate": 1.9482744313926073e-05, "loss": 0.9953, "step": 3899 }, { "epoch": 0.13, "learning_rate": 1.94824020835735e-05, "loss": 0.9915, "step": 3900 }, { "epoch": 0.13, "learning_rate": 1.948205974305173e-05, "loss": 1.0567, "step": 3901 }, { "epoch": 0.13, "learning_rate": 1.9481717292364735e-05, "loss": 0.9946, "step": 3902 }, { "epoch": 0.13, "learning_rate": 1.9481374731516496e-05, "loss": 0.9978, "step": 3903 }, { "epoch": 0.13, "learning_rate": 1.9481032060511e-05, "loss": 1.0145, "step": 3904 }, { "epoch": 0.13, "learning_rate": 1.948068927935222e-05, "loss": 1.015, "step": 3905 }, { "epoch": 0.13, "learning_rate": 1.948034638804414e-05, "loss": 1.0115, "step": 3906 }, { "epoch": 0.13, "learning_rate": 1.9480003386590748e-05, "loss": 1.0453, "step": 3907 }, { "epoch": 0.13, "learning_rate": 1.947966027499602e-05, "loss": 1.0145, "step": 3908 }, { "epoch": 0.13, "learning_rate": 1.9479317053263954e-05, "loss": 1.0233, "step": 3909 }, { "epoch": 0.13, "learning_rate": 1.947897372139853e-05, "loss": 1.0516, "step": 3910 }, { "epoch": 0.13, "learning_rate": 1.9478630279403737e-05, "loss": 1.0077, "step": 3911 }, { "epoch": 0.13, "learning_rate": 1.947828672728357e-05, "loss": 1.0093, "step": 3912 }, { "epoch": 0.13, "learning_rate": 1.9477943065042016e-05, "loss": 0.9878, "step": 3913 }, { "epoch": 0.13, "learning_rate": 1.9477599292683075e-05, "loss": 0.9709, "step": 3914 }, { "epoch": 0.13, "learning_rate": 1.947725541021073e-05, "loss": 1.0174, "step": 3915 }, { "epoch": 0.13, "learning_rate": 1.9476911417628985e-05, "loss": 0.9856, "step": 3916 }, { "epoch": 0.13, "learning_rate": 1.9476567314941828e-05, "loss": 1.0079, "step": 3917 }, { "epoch": 0.13, "learning_rate": 1.9476223102153266e-05, "loss": 0.9856, "step": 3918 }, { "epoch": 0.13, "learning_rate": 1.9475878779267293e-05, "loss": 0.979, "step": 3919 }, { "epoch": 0.13, "learning_rate": 1.9475534346287912e-05, "loss": 1.0021, "step": 3920 }, { "epoch": 0.13, "learning_rate": 1.9475189803219124e-05, "loss": 0.9946, "step": 3921 }, { "epoch": 0.13, "learning_rate": 1.947484515006493e-05, "loss": 0.9831, "step": 3922 }, { "epoch": 0.13, "learning_rate": 1.9474500386829337e-05, "loss": 0.9907, "step": 3923 }, { "epoch": 0.13, "learning_rate": 1.947415551351635e-05, "loss": 1.0001, "step": 3924 }, { "epoch": 0.13, "learning_rate": 1.9473810530129973e-05, "loss": 1.0208, "step": 3925 }, { "epoch": 0.13, "learning_rate": 1.947346543667422e-05, "loss": 1.0158, "step": 3926 }, { "epoch": 0.13, "learning_rate": 1.9473120233153093e-05, "loss": 1.0162, "step": 3927 }, { "epoch": 0.13, "learning_rate": 1.9472774919570606e-05, "loss": 0.9881, "step": 3928 }, { "epoch": 0.13, "learning_rate": 1.9472429495930777e-05, "loss": 1.0256, "step": 3929 }, { "epoch": 0.13, "learning_rate": 1.947208396223761e-05, "loss": 1.0009, "step": 3930 }, { "epoch": 0.13, "learning_rate": 1.9471738318495124e-05, "loss": 1.0205, "step": 3931 }, { "epoch": 0.13, "learning_rate": 1.947139256470733e-05, "loss": 1.0018, "step": 3932 }, { "epoch": 0.13, "learning_rate": 1.9471046700878252e-05, "loss": 1.0155, "step": 3933 }, { "epoch": 0.13, "learning_rate": 1.9470700727011905e-05, "loss": 0.9557, "step": 3934 }, { "epoch": 0.13, "learning_rate": 1.947035464311231e-05, "loss": 1.0345, "step": 3935 }, { "epoch": 0.13, "learning_rate": 1.947000844918349e-05, "loss": 0.9735, "step": 3936 }, { "epoch": 0.13, "learning_rate": 1.9469662145229457e-05, "loss": 1.0014, "step": 3937 }, { "epoch": 0.13, "learning_rate": 1.9469315731254244e-05, "loss": 0.978, "step": 3938 }, { "epoch": 0.13, "learning_rate": 1.9468969207261876e-05, "loss": 1.0195, "step": 3939 }, { "epoch": 0.13, "learning_rate": 1.9468622573256373e-05, "loss": 1.0144, "step": 3940 }, { "epoch": 0.13, "learning_rate": 1.9468275829241765e-05, "loss": 0.997, "step": 3941 }, { "epoch": 0.13, "learning_rate": 1.9467928975222087e-05, "loss": 0.9865, "step": 3942 }, { "epoch": 0.13, "learning_rate": 1.9467582011201357e-05, "loss": 1.0332, "step": 3943 }, { "epoch": 0.13, "learning_rate": 1.946723493718361e-05, "loss": 0.9918, "step": 3944 }, { "epoch": 0.13, "learning_rate": 1.9466887753172887e-05, "loss": 1.0131, "step": 3945 }, { "epoch": 0.13, "learning_rate": 1.9466540459173213e-05, "loss": 1.0373, "step": 3946 }, { "epoch": 0.13, "learning_rate": 1.9466193055188622e-05, "loss": 0.9846, "step": 3947 }, { "epoch": 0.13, "learning_rate": 1.9465845541223156e-05, "loss": 0.9676, "step": 3948 }, { "epoch": 0.13, "learning_rate": 1.9465497917280853e-05, "loss": 1.0294, "step": 3949 }, { "epoch": 0.13, "learning_rate": 1.946515018336574e-05, "loss": 1.0021, "step": 3950 }, { "epoch": 0.13, "learning_rate": 1.9464802339481873e-05, "loss": 0.9979, "step": 3951 }, { "epoch": 0.13, "learning_rate": 1.946445438563328e-05, "loss": 0.9913, "step": 3952 }, { "epoch": 0.13, "learning_rate": 1.9464106321824017e-05, "loss": 1.0202, "step": 3953 }, { "epoch": 0.13, "learning_rate": 1.9463758148058117e-05, "loss": 1.0335, "step": 3954 }, { "epoch": 0.13, "learning_rate": 1.9463409864339628e-05, "loss": 0.9954, "step": 3955 }, { "epoch": 0.13, "learning_rate": 1.94630614706726e-05, "loss": 0.9747, "step": 3956 }, { "epoch": 0.13, "learning_rate": 1.9462712967061075e-05, "loss": 0.9934, "step": 3957 }, { "epoch": 0.13, "learning_rate": 1.9462364353509106e-05, "loss": 0.9625, "step": 3958 }, { "epoch": 0.13, "learning_rate": 1.9462015630020743e-05, "loss": 0.9842, "step": 3959 }, { "epoch": 0.13, "learning_rate": 1.9461666796600034e-05, "loss": 1.0113, "step": 3960 }, { "epoch": 0.13, "learning_rate": 1.946131785325104e-05, "loss": 0.9834, "step": 3961 }, { "epoch": 0.13, "learning_rate": 1.9460968799977805e-05, "loss": 0.9939, "step": 3962 }, { "epoch": 0.13, "learning_rate": 1.946061963678439e-05, "loss": 1.0189, "step": 3963 }, { "epoch": 0.13, "learning_rate": 1.946027036367485e-05, "loss": 0.9877, "step": 3964 }, { "epoch": 0.13, "learning_rate": 1.9459920980653246e-05, "loss": 1.0318, "step": 3965 }, { "epoch": 0.13, "learning_rate": 1.9459571487723634e-05, "loss": 0.9815, "step": 3966 }, { "epoch": 0.13, "learning_rate": 1.9459221884890078e-05, "loss": 1.0061, "step": 3967 }, { "epoch": 0.13, "learning_rate": 1.9458872172156632e-05, "loss": 0.9745, "step": 3968 }, { "epoch": 0.13, "learning_rate": 1.945852234952737e-05, "loss": 0.9897, "step": 3969 }, { "epoch": 0.13, "learning_rate": 1.9458172417006347e-05, "loss": 1.0277, "step": 3970 }, { "epoch": 0.13, "learning_rate": 1.9457822374597636e-05, "loss": 0.9788, "step": 3971 }, { "epoch": 0.13, "learning_rate": 1.9457472222305296e-05, "loss": 0.9726, "step": 3972 }, { "epoch": 0.13, "learning_rate": 1.9457121960133402e-05, "loss": 1.0356, "step": 3973 }, { "epoch": 0.13, "learning_rate": 1.945677158808602e-05, "loss": 1.0186, "step": 3974 }, { "epoch": 0.13, "learning_rate": 1.945642110616722e-05, "loss": 1.0198, "step": 3975 }, { "epoch": 0.13, "learning_rate": 1.945607051438108e-05, "loss": 0.9906, "step": 3976 }, { "epoch": 0.13, "learning_rate": 1.9455719812731666e-05, "loss": 0.9808, "step": 3977 }, { "epoch": 0.13, "learning_rate": 1.9455369001223053e-05, "loss": 0.9461, "step": 3978 }, { "epoch": 0.13, "learning_rate": 1.9455018079859322e-05, "loss": 1.0155, "step": 3979 }, { "epoch": 0.13, "learning_rate": 1.9454667048644548e-05, "loss": 0.9363, "step": 3980 }, { "epoch": 0.13, "learning_rate": 1.9454315907582808e-05, "loss": 1.0145, "step": 3981 }, { "epoch": 0.13, "learning_rate": 1.9453964656678185e-05, "loss": 1.0032, "step": 3982 }, { "epoch": 0.13, "learning_rate": 1.9453613295934753e-05, "loss": 1.0329, "step": 3983 }, { "epoch": 0.13, "learning_rate": 1.94532618253566e-05, "loss": 0.9776, "step": 3984 }, { "epoch": 0.13, "learning_rate": 1.9452910244947805e-05, "loss": 1.0092, "step": 3985 }, { "epoch": 0.13, "learning_rate": 1.945255855471246e-05, "loss": 1.0415, "step": 3986 }, { "epoch": 0.13, "learning_rate": 1.9452206754654646e-05, "loss": 0.9769, "step": 3987 }, { "epoch": 0.13, "learning_rate": 1.945185484477845e-05, "loss": 0.9828, "step": 3988 }, { "epoch": 0.13, "learning_rate": 1.9451502825087964e-05, "loss": 1.0031, "step": 3989 }, { "epoch": 0.13, "learning_rate": 1.9451150695587272e-05, "loss": 1.0094, "step": 3990 }, { "epoch": 0.13, "learning_rate": 1.9450798456280472e-05, "loss": 0.9602, "step": 3991 }, { "epoch": 0.13, "learning_rate": 1.945044610717165e-05, "loss": 0.9575, "step": 3992 }, { "epoch": 0.13, "learning_rate": 1.9450093648264902e-05, "loss": 1.0158, "step": 3993 }, { "epoch": 0.13, "learning_rate": 1.944974107956433e-05, "loss": 1.0118, "step": 3994 }, { "epoch": 0.13, "learning_rate": 1.9449388401074015e-05, "loss": 0.9948, "step": 3995 }, { "epoch": 0.13, "learning_rate": 1.944903561279807e-05, "loss": 0.9535, "step": 3996 }, { "epoch": 0.13, "learning_rate": 1.9448682714740587e-05, "loss": 0.9619, "step": 3997 }, { "epoch": 0.13, "learning_rate": 1.9448329706905665e-05, "loss": 0.9983, "step": 3998 }, { "epoch": 0.13, "learning_rate": 1.9447976589297405e-05, "loss": 0.9715, "step": 3999 }, { "epoch": 0.13, "learning_rate": 1.9447623361919915e-05, "loss": 0.9999, "step": 4000 }, { "epoch": 0.13, "learning_rate": 1.9447270024777293e-05, "loss": 0.9876, "step": 4001 }, { "epoch": 0.13, "learning_rate": 1.9446916577873645e-05, "loss": 1.0125, "step": 4002 }, { "epoch": 0.13, "learning_rate": 1.944656302121308e-05, "loss": 0.9612, "step": 4003 }, { "epoch": 0.13, "learning_rate": 1.9446209354799704e-05, "loss": 0.98, "step": 4004 }, { "epoch": 0.13, "learning_rate": 1.944585557863763e-05, "loss": 0.9932, "step": 4005 }, { "epoch": 0.13, "learning_rate": 1.9445501692730963e-05, "loss": 1.0149, "step": 4006 }, { "epoch": 0.13, "learning_rate": 1.9445147697083816e-05, "loss": 0.9631, "step": 4007 }, { "epoch": 0.13, "learning_rate": 1.94447935917003e-05, "loss": 0.9655, "step": 4008 }, { "epoch": 0.13, "learning_rate": 1.9444439376584533e-05, "loss": 1.0388, "step": 4009 }, { "epoch": 0.13, "learning_rate": 1.944408505174063e-05, "loss": 0.9651, "step": 4010 }, { "epoch": 0.13, "learning_rate": 1.9443730617172705e-05, "loss": 0.9733, "step": 4011 }, { "epoch": 0.13, "learning_rate": 1.9443376072884876e-05, "loss": 0.9856, "step": 4012 }, { "epoch": 0.13, "learning_rate": 1.9443021418881266e-05, "loss": 1.0017, "step": 4013 }, { "epoch": 0.13, "learning_rate": 1.944266665516599e-05, "loss": 0.9826, "step": 4014 }, { "epoch": 0.13, "learning_rate": 1.9442311781743175e-05, "loss": 0.9768, "step": 4015 }, { "epoch": 0.13, "learning_rate": 1.944195679861694e-05, "loss": 0.979, "step": 4016 }, { "epoch": 0.13, "learning_rate": 1.9441601705791413e-05, "loss": 1.0157, "step": 4017 }, { "epoch": 0.13, "learning_rate": 1.9441246503270716e-05, "loss": 0.9827, "step": 4018 }, { "epoch": 0.13, "learning_rate": 1.944089119105898e-05, "loss": 0.9741, "step": 4019 }, { "epoch": 0.13, "learning_rate": 1.9440535769160326e-05, "loss": 1.0358, "step": 4020 }, { "epoch": 0.13, "learning_rate": 1.944018023757889e-05, "loss": 1.0141, "step": 4021 }, { "epoch": 0.13, "learning_rate": 1.94398245963188e-05, "loss": 1.023, "step": 4022 }, { "epoch": 0.13, "learning_rate": 1.943946884538419e-05, "loss": 1.0217, "step": 4023 }, { "epoch": 0.13, "learning_rate": 1.9439112984779188e-05, "loss": 0.9781, "step": 4024 }, { "epoch": 0.13, "learning_rate": 1.9438757014507936e-05, "loss": 0.9921, "step": 4025 }, { "epoch": 0.13, "learning_rate": 1.9438400934574565e-05, "loss": 0.9857, "step": 4026 }, { "epoch": 0.13, "learning_rate": 1.9438044744983214e-05, "loss": 1.0543, "step": 4027 }, { "epoch": 0.13, "learning_rate": 1.9437688445738016e-05, "loss": 0.994, "step": 4028 }, { "epoch": 0.13, "learning_rate": 1.943733203684312e-05, "loss": 0.9943, "step": 4029 }, { "epoch": 0.13, "learning_rate": 1.9436975518302657e-05, "loss": 0.9808, "step": 4030 }, { "epoch": 0.13, "learning_rate": 1.9436618890120777e-05, "loss": 1.0306, "step": 4031 }, { "epoch": 0.13, "learning_rate": 1.9436262152301618e-05, "loss": 1.007, "step": 4032 }, { "epoch": 0.13, "learning_rate": 1.943590530484933e-05, "loss": 1.0166, "step": 4033 }, { "epoch": 0.13, "learning_rate": 1.9435548347768054e-05, "loss": 1.0571, "step": 4034 }, { "epoch": 0.13, "learning_rate": 1.943519128106194e-05, "loss": 1.0029, "step": 4035 }, { "epoch": 0.13, "learning_rate": 1.9434834104735135e-05, "loss": 0.9729, "step": 4036 }, { "epoch": 0.13, "learning_rate": 1.943447681879179e-05, "loss": 1.0204, "step": 4037 }, { "epoch": 0.13, "learning_rate": 1.9434119423236054e-05, "loss": 0.9844, "step": 4038 }, { "epoch": 0.13, "learning_rate": 1.9433761918072083e-05, "loss": 1.0319, "step": 4039 }, { "epoch": 0.13, "learning_rate": 1.9433404303304027e-05, "loss": 0.968, "step": 4040 }, { "epoch": 0.13, "learning_rate": 1.9433046578936043e-05, "loss": 0.9992, "step": 4041 }, { "epoch": 0.13, "learning_rate": 1.9432688744972282e-05, "loss": 0.9707, "step": 4042 }, { "epoch": 0.13, "learning_rate": 1.943233080141691e-05, "loss": 0.991, "step": 4043 }, { "epoch": 0.13, "learning_rate": 1.9431972748274082e-05, "loss": 0.9925, "step": 4044 }, { "epoch": 0.13, "learning_rate": 1.9431614585547954e-05, "loss": 1.0065, "step": 4045 }, { "epoch": 0.13, "learning_rate": 1.9431256313242692e-05, "loss": 1.0235, "step": 4046 }, { "epoch": 0.13, "learning_rate": 1.943089793136246e-05, "loss": 0.9804, "step": 4047 }, { "epoch": 0.13, "learning_rate": 1.9430539439911416e-05, "loss": 0.9743, "step": 4048 }, { "epoch": 0.13, "learning_rate": 1.9430180838893727e-05, "loss": 0.997, "step": 4049 }, { "epoch": 0.13, "learning_rate": 1.9429822128313564e-05, "loss": 0.9651, "step": 4050 }, { "epoch": 0.13, "learning_rate": 1.942946330817509e-05, "loss": 1.0329, "step": 4051 }, { "epoch": 0.13, "learning_rate": 1.9429104378482474e-05, "loss": 0.9815, "step": 4052 }, { "epoch": 0.13, "learning_rate": 1.9428745339239886e-05, "loss": 0.9697, "step": 4053 }, { "epoch": 0.13, "learning_rate": 1.94283861904515e-05, "loss": 0.9929, "step": 4054 }, { "epoch": 0.13, "learning_rate": 1.9428026932121485e-05, "loss": 0.9893, "step": 4055 }, { "epoch": 0.13, "learning_rate": 1.9427667564254017e-05, "loss": 0.9784, "step": 4056 }, { "epoch": 0.14, "learning_rate": 1.9427308086853272e-05, "loss": 0.9972, "step": 4057 }, { "epoch": 0.14, "learning_rate": 1.942694849992343e-05, "loss": 0.9767, "step": 4058 }, { "epoch": 0.14, "learning_rate": 1.942658880346866e-05, "loss": 0.9906, "step": 4059 }, { "epoch": 0.14, "learning_rate": 1.9426228997493145e-05, "loss": 0.9838, "step": 4060 }, { "epoch": 0.14, "learning_rate": 1.9425869082001068e-05, "loss": 1.0104, "step": 4061 }, { "epoch": 0.14, "learning_rate": 1.942550905699661e-05, "loss": 0.9874, "step": 4062 }, { "epoch": 0.14, "learning_rate": 1.9425148922483948e-05, "loss": 1.028, "step": 4063 }, { "epoch": 0.14, "learning_rate": 1.9424788678467273e-05, "loss": 0.9976, "step": 4064 }, { "epoch": 0.14, "learning_rate": 1.942442832495077e-05, "loss": 1.006, "step": 4065 }, { "epoch": 0.14, "learning_rate": 1.942406786193862e-05, "loss": 0.9991, "step": 4066 }, { "epoch": 0.14, "learning_rate": 1.9423707289435018e-05, "loss": 0.9981, "step": 4067 }, { "epoch": 0.14, "learning_rate": 1.9423346607444145e-05, "loss": 0.9477, "step": 4068 }, { "epoch": 0.14, "learning_rate": 1.9422985815970202e-05, "loss": 0.9966, "step": 4069 }, { "epoch": 0.14, "learning_rate": 1.942262491501737e-05, "loss": 0.9936, "step": 4070 }, { "epoch": 0.14, "learning_rate": 1.942226390458985e-05, "loss": 0.9799, "step": 4071 }, { "epoch": 0.14, "learning_rate": 1.942190278469183e-05, "loss": 0.9607, "step": 4072 }, { "epoch": 0.14, "learning_rate": 1.9421541555327515e-05, "loss": 0.9938, "step": 4073 }, { "epoch": 0.14, "learning_rate": 1.942118021650109e-05, "loss": 0.993, "step": 4074 }, { "epoch": 0.14, "learning_rate": 1.9420818768216763e-05, "loss": 1.0378, "step": 4075 }, { "epoch": 0.14, "learning_rate": 1.942045721047873e-05, "loss": 0.9706, "step": 4076 }, { "epoch": 0.14, "learning_rate": 1.9420095543291188e-05, "loss": 0.9949, "step": 4077 }, { "epoch": 0.14, "learning_rate": 1.9419733766658347e-05, "loss": 1.0275, "step": 4078 }, { "epoch": 0.14, "learning_rate": 1.94193718805844e-05, "loss": 0.9869, "step": 4079 }, { "epoch": 0.14, "learning_rate": 1.941900988507356e-05, "loss": 1.0072, "step": 4080 }, { "epoch": 0.14, "learning_rate": 1.941864778013003e-05, "loss": 0.9855, "step": 4081 }, { "epoch": 0.14, "learning_rate": 1.9418285565758014e-05, "loss": 0.9362, "step": 4082 }, { "epoch": 0.14, "learning_rate": 1.9417923241961723e-05, "loss": 1.0224, "step": 4083 }, { "epoch": 0.14, "learning_rate": 1.941756080874537e-05, "loss": 0.9903, "step": 4084 }, { "epoch": 0.14, "learning_rate": 1.9417198266113156e-05, "loss": 0.9818, "step": 4085 }, { "epoch": 0.14, "learning_rate": 1.9416835614069304e-05, "loss": 1.0014, "step": 4086 }, { "epoch": 0.14, "learning_rate": 1.941647285261802e-05, "loss": 1.0189, "step": 4087 }, { "epoch": 0.14, "learning_rate": 1.9416109981763526e-05, "loss": 1.0104, "step": 4088 }, { "epoch": 0.14, "learning_rate": 1.9415747001510032e-05, "loss": 1.0176, "step": 4089 }, { "epoch": 0.14, "learning_rate": 1.9415383911861756e-05, "loss": 0.9763, "step": 4090 }, { "epoch": 0.14, "learning_rate": 1.9415020712822916e-05, "loss": 0.9833, "step": 4091 }, { "epoch": 0.14, "learning_rate": 1.9414657404397735e-05, "loss": 0.9671, "step": 4092 }, { "epoch": 0.14, "learning_rate": 1.941429398659043e-05, "loss": 0.9946, "step": 4093 }, { "epoch": 0.14, "learning_rate": 1.941393045940523e-05, "loss": 0.9971, "step": 4094 }, { "epoch": 0.14, "learning_rate": 1.9413566822846345e-05, "loss": 1.016, "step": 4095 }, { "epoch": 0.14, "learning_rate": 1.9413203076918016e-05, "loss": 0.988, "step": 4096 }, { "epoch": 0.14, "learning_rate": 1.941283922162446e-05, "loss": 0.9955, "step": 4097 }, { "epoch": 0.14, "learning_rate": 1.9412475256969904e-05, "loss": 1.0112, "step": 4098 }, { "epoch": 0.14, "learning_rate": 1.9412111182958578e-05, "loss": 1.0026, "step": 4099 }, { "epoch": 0.14, "learning_rate": 1.9411746999594716e-05, "loss": 0.9685, "step": 4100 }, { "epoch": 0.14, "learning_rate": 1.9411382706882543e-05, "loss": 1.0216, "step": 4101 }, { "epoch": 0.14, "learning_rate": 1.9411018304826297e-05, "loss": 0.9733, "step": 4102 }, { "epoch": 0.14, "learning_rate": 1.9410653793430207e-05, "loss": 1.0237, "step": 4103 }, { "epoch": 0.14, "learning_rate": 1.941028917269851e-05, "loss": 1.0073, "step": 4104 }, { "epoch": 0.14, "learning_rate": 1.9409924442635445e-05, "loss": 0.9487, "step": 4105 }, { "epoch": 0.14, "learning_rate": 1.9409559603245246e-05, "loss": 0.9924, "step": 4106 }, { "epoch": 0.14, "learning_rate": 1.940919465453215e-05, "loss": 1.0086, "step": 4107 }, { "epoch": 0.14, "learning_rate": 1.9408829596500404e-05, "loss": 1.0006, "step": 4108 }, { "epoch": 0.14, "learning_rate": 1.940846442915424e-05, "loss": 0.9663, "step": 4109 }, { "epoch": 0.14, "learning_rate": 1.9408099152497906e-05, "loss": 0.9726, "step": 4110 }, { "epoch": 0.14, "learning_rate": 1.940773376653565e-05, "loss": 1.0074, "step": 4111 }, { "epoch": 0.14, "learning_rate": 1.940736827127171e-05, "loss": 1.0198, "step": 4112 }, { "epoch": 0.14, "learning_rate": 1.9407002666710334e-05, "loss": 0.9983, "step": 4113 }, { "epoch": 0.14, "learning_rate": 1.9406636952855772e-05, "loss": 0.9855, "step": 4114 }, { "epoch": 0.14, "learning_rate": 1.9406271129712272e-05, "loss": 0.9733, "step": 4115 }, { "epoch": 0.14, "learning_rate": 1.9405905197284084e-05, "loss": 0.9922, "step": 4116 }, { "epoch": 0.14, "learning_rate": 1.940553915557546e-05, "loss": 1.0265, "step": 4117 }, { "epoch": 0.14, "learning_rate": 1.9405173004590652e-05, "loss": 1.0148, "step": 4118 }, { "epoch": 0.14, "learning_rate": 1.9404806744333913e-05, "loss": 0.9862, "step": 4119 }, { "epoch": 0.14, "learning_rate": 1.94044403748095e-05, "loss": 0.9916, "step": 4120 }, { "epoch": 0.14, "learning_rate": 1.9404073896021673e-05, "loss": 0.9623, "step": 4121 }, { "epoch": 0.14, "learning_rate": 1.940370730797468e-05, "loss": 1.0274, "step": 4122 }, { "epoch": 0.14, "learning_rate": 1.9403340610672788e-05, "loss": 1.004, "step": 4123 }, { "epoch": 0.14, "learning_rate": 1.940297380412026e-05, "loss": 0.9786, "step": 4124 }, { "epoch": 0.14, "learning_rate": 1.9402606888321347e-05, "loss": 1.0388, "step": 4125 }, { "epoch": 0.14, "learning_rate": 1.9402239863280322e-05, "loss": 0.9943, "step": 4126 }, { "epoch": 0.14, "learning_rate": 1.9401872729001442e-05, "loss": 1.002, "step": 4127 }, { "epoch": 0.14, "learning_rate": 1.940150548548898e-05, "loss": 0.9564, "step": 4128 }, { "epoch": 0.14, "learning_rate": 1.9401138132747197e-05, "loss": 1.0, "step": 4129 }, { "epoch": 0.14, "learning_rate": 1.9400770670780363e-05, "loss": 0.9843, "step": 4130 }, { "epoch": 0.14, "learning_rate": 1.9400403099592744e-05, "loss": 1.0165, "step": 4131 }, { "epoch": 0.14, "learning_rate": 1.940003541918862e-05, "loss": 0.9847, "step": 4132 }, { "epoch": 0.14, "learning_rate": 1.939966762957225e-05, "loss": 0.996, "step": 4133 }, { "epoch": 0.14, "learning_rate": 1.9399299730747917e-05, "loss": 0.964, "step": 4134 }, { "epoch": 0.14, "learning_rate": 1.9398931722719888e-05, "loss": 1.0328, "step": 4135 }, { "epoch": 0.14, "learning_rate": 1.9398563605492446e-05, "loss": 0.9619, "step": 4136 }, { "epoch": 0.14, "learning_rate": 1.9398195379069863e-05, "loss": 0.9841, "step": 4137 }, { "epoch": 0.14, "learning_rate": 1.9397827043456415e-05, "loss": 0.9554, "step": 4138 }, { "epoch": 0.14, "learning_rate": 1.939745859865639e-05, "loss": 0.9937, "step": 4139 }, { "epoch": 0.14, "learning_rate": 1.939709004467406e-05, "loss": 1.0078, "step": 4140 }, { "epoch": 0.14, "learning_rate": 1.9396721381513713e-05, "loss": 1.0242, "step": 4141 }, { "epoch": 0.14, "learning_rate": 1.9396352609179626e-05, "loss": 1.0263, "step": 4142 }, { "epoch": 0.14, "learning_rate": 1.939598372767609e-05, "loss": 1.0022, "step": 4143 }, { "epoch": 0.14, "learning_rate": 1.9395614737007385e-05, "loss": 0.981, "step": 4144 }, { "epoch": 0.14, "learning_rate": 1.9395245637177807e-05, "loss": 1.002, "step": 4145 }, { "epoch": 0.14, "learning_rate": 1.9394876428191632e-05, "loss": 1.0189, "step": 4146 }, { "epoch": 0.14, "learning_rate": 1.939450711005316e-05, "loss": 0.983, "step": 4147 }, { "epoch": 0.14, "learning_rate": 1.9394137682766675e-05, "loss": 0.9952, "step": 4148 }, { "epoch": 0.14, "learning_rate": 1.9393768146336475e-05, "loss": 0.9739, "step": 4149 }, { "epoch": 0.14, "learning_rate": 1.939339850076685e-05, "loss": 0.9977, "step": 4150 }, { "epoch": 0.14, "learning_rate": 1.939302874606209e-05, "loss": 0.9254, "step": 4151 }, { "epoch": 0.14, "learning_rate": 1.93926588822265e-05, "loss": 0.9867, "step": 4152 }, { "epoch": 0.14, "learning_rate": 1.939228890926437e-05, "loss": 0.9554, "step": 4153 }, { "epoch": 0.14, "learning_rate": 1.9391918827180002e-05, "loss": 0.9622, "step": 4154 }, { "epoch": 0.14, "learning_rate": 1.9391548635977697e-05, "loss": 0.9783, "step": 4155 }, { "epoch": 0.14, "learning_rate": 1.9391178335661754e-05, "loss": 0.9669, "step": 4156 }, { "epoch": 0.14, "learning_rate": 1.9390807926236474e-05, "loss": 1.0166, "step": 4157 }, { "epoch": 0.14, "learning_rate": 1.939043740770616e-05, "loss": 0.9941, "step": 4158 }, { "epoch": 0.14, "learning_rate": 1.939006678007512e-05, "loss": 0.9603, "step": 4159 }, { "epoch": 0.14, "learning_rate": 1.938969604334766e-05, "loss": 0.9548, "step": 4160 }, { "epoch": 0.14, "learning_rate": 1.9389325197528086e-05, "loss": 1.0094, "step": 4161 }, { "epoch": 0.14, "learning_rate": 1.9388954242620707e-05, "loss": 0.9939, "step": 4162 }, { "epoch": 0.14, "learning_rate": 1.938858317862983e-05, "loss": 0.9923, "step": 4163 }, { "epoch": 0.14, "learning_rate": 1.938821200555977e-05, "loss": 0.9757, "step": 4164 }, { "epoch": 0.14, "learning_rate": 1.938784072341484e-05, "loss": 1.0266, "step": 4165 }, { "epoch": 0.14, "learning_rate": 1.9387469332199348e-05, "loss": 1.0074, "step": 4166 }, { "epoch": 0.14, "learning_rate": 1.9387097831917615e-05, "loss": 1.0131, "step": 4167 }, { "epoch": 0.14, "learning_rate": 1.9386726222573953e-05, "loss": 0.9748, "step": 4168 }, { "epoch": 0.14, "learning_rate": 1.938635450417268e-05, "loss": 0.9796, "step": 4169 }, { "epoch": 0.14, "learning_rate": 1.938598267671812e-05, "loss": 1.0123, "step": 4170 }, { "epoch": 0.14, "learning_rate": 1.9385610740214588e-05, "loss": 1.0249, "step": 4171 }, { "epoch": 0.14, "learning_rate": 1.9385238694666403e-05, "loss": 0.9625, "step": 4172 }, { "epoch": 0.14, "learning_rate": 1.9384866540077892e-05, "loss": 1.0014, "step": 4173 }, { "epoch": 0.14, "learning_rate": 1.938449427645338e-05, "loss": 0.9589, "step": 4174 }, { "epoch": 0.14, "learning_rate": 1.938412190379719e-05, "loss": 1.0038, "step": 4175 }, { "epoch": 0.14, "learning_rate": 1.938374942211364e-05, "loss": 1.0025, "step": 4176 }, { "epoch": 0.14, "learning_rate": 1.9383376831407076e-05, "loss": 1.011, "step": 4177 }, { "epoch": 0.14, "learning_rate": 1.938300413168181e-05, "loss": 0.9715, "step": 4178 }, { "epoch": 0.14, "learning_rate": 1.938263132294218e-05, "loss": 0.988, "step": 4179 }, { "epoch": 0.14, "learning_rate": 1.938225840519252e-05, "loss": 0.963, "step": 4180 }, { "epoch": 0.14, "learning_rate": 1.9381885378437154e-05, "loss": 1.0306, "step": 4181 }, { "epoch": 0.14, "learning_rate": 1.938151224268042e-05, "loss": 1.0021, "step": 4182 }, { "epoch": 0.14, "learning_rate": 1.9381138997926653e-05, "loss": 0.9889, "step": 4183 }, { "epoch": 0.14, "learning_rate": 1.938076564418019e-05, "loss": 0.9626, "step": 4184 }, { "epoch": 0.14, "learning_rate": 1.9380392181445374e-05, "loss": 0.9902, "step": 4185 }, { "epoch": 0.14, "learning_rate": 1.9380018609726537e-05, "loss": 1.0028, "step": 4186 }, { "epoch": 0.14, "learning_rate": 1.9379644929028017e-05, "loss": 1.0297, "step": 4187 }, { "epoch": 0.14, "learning_rate": 1.9379271139354162e-05, "loss": 0.9957, "step": 4188 }, { "epoch": 0.14, "learning_rate": 1.9378897240709315e-05, "loss": 0.9847, "step": 4189 }, { "epoch": 0.14, "learning_rate": 1.9378523233097816e-05, "loss": 1.0068, "step": 4190 }, { "epoch": 0.14, "learning_rate": 1.937814911652401e-05, "loss": 0.9967, "step": 4191 }, { "epoch": 0.14, "learning_rate": 1.9377774890992248e-05, "loss": 0.9849, "step": 4192 }, { "epoch": 0.14, "learning_rate": 1.9377400556506872e-05, "loss": 0.9728, "step": 4193 }, { "epoch": 0.14, "learning_rate": 1.9377026113072236e-05, "loss": 1.0008, "step": 4194 }, { "epoch": 0.14, "learning_rate": 1.9376651560692693e-05, "loss": 1.0063, "step": 4195 }, { "epoch": 0.14, "learning_rate": 1.9376276899372584e-05, "loss": 0.9348, "step": 4196 }, { "epoch": 0.14, "learning_rate": 1.9375902129116272e-05, "loss": 0.9903, "step": 4197 }, { "epoch": 0.14, "learning_rate": 1.9375527249928104e-05, "loss": 0.9776, "step": 4198 }, { "epoch": 0.14, "learning_rate": 1.9375152261812445e-05, "loss": 0.9879, "step": 4199 }, { "epoch": 0.14, "learning_rate": 1.937477716477364e-05, "loss": 0.974, "step": 4200 }, { "epoch": 0.14, "learning_rate": 1.9374401958816056e-05, "loss": 1.0125, "step": 4201 }, { "epoch": 0.14, "learning_rate": 1.9374026643944048e-05, "loss": 1.0068, "step": 4202 }, { "epoch": 0.14, "learning_rate": 1.9373651220161976e-05, "loss": 0.9859, "step": 4203 }, { "epoch": 0.14, "learning_rate": 1.9373275687474208e-05, "loss": 0.977, "step": 4204 }, { "epoch": 0.14, "learning_rate": 1.9372900045885098e-05, "loss": 0.9865, "step": 4205 }, { "epoch": 0.14, "learning_rate": 1.9372524295399014e-05, "loss": 0.9688, "step": 4206 }, { "epoch": 0.14, "learning_rate": 1.9372148436020325e-05, "loss": 1.0185, "step": 4207 }, { "epoch": 0.14, "learning_rate": 1.9371772467753393e-05, "loss": 0.9336, "step": 4208 }, { "epoch": 0.14, "learning_rate": 1.9371396390602587e-05, "loss": 0.9989, "step": 4209 }, { "epoch": 0.14, "learning_rate": 1.937102020457228e-05, "loss": 0.9932, "step": 4210 }, { "epoch": 0.14, "learning_rate": 1.937064390966684e-05, "loss": 0.988, "step": 4211 }, { "epoch": 0.14, "learning_rate": 1.9370267505890637e-05, "loss": 0.9656, "step": 4212 }, { "epoch": 0.14, "learning_rate": 1.9369890993248047e-05, "loss": 1.0072, "step": 4213 }, { "epoch": 0.14, "learning_rate": 1.9369514371743445e-05, "loss": 0.946, "step": 4214 }, { "epoch": 0.14, "learning_rate": 1.9369137641381207e-05, "loss": 1.0002, "step": 4215 }, { "epoch": 0.14, "learning_rate": 1.9368760802165704e-05, "loss": 1.0023, "step": 4216 }, { "epoch": 0.14, "learning_rate": 1.936838385410132e-05, "loss": 1.0068, "step": 4217 }, { "epoch": 0.14, "learning_rate": 1.9368006797192432e-05, "loss": 1.0161, "step": 4218 }, { "epoch": 0.14, "learning_rate": 1.9367629631443425e-05, "loss": 1.0071, "step": 4219 }, { "epoch": 0.14, "learning_rate": 1.9367252356858676e-05, "loss": 0.9837, "step": 4220 }, { "epoch": 0.14, "learning_rate": 1.936687497344257e-05, "loss": 1.0095, "step": 4221 }, { "epoch": 0.14, "learning_rate": 1.936649748119949e-05, "loss": 0.996, "step": 4222 }, { "epoch": 0.14, "learning_rate": 1.9366119880133825e-05, "loss": 1.0238, "step": 4223 }, { "epoch": 0.14, "learning_rate": 1.9365742170249958e-05, "loss": 0.9658, "step": 4224 }, { "epoch": 0.14, "learning_rate": 1.9365364351552284e-05, "loss": 0.9969, "step": 4225 }, { "epoch": 0.14, "learning_rate": 1.9364986424045185e-05, "loss": 0.9984, "step": 4226 }, { "epoch": 0.14, "learning_rate": 1.9364608387733058e-05, "loss": 0.9678, "step": 4227 }, { "epoch": 0.14, "learning_rate": 1.936423024262029e-05, "loss": 0.9762, "step": 4228 }, { "epoch": 0.14, "learning_rate": 1.9363851988711278e-05, "loss": 0.9734, "step": 4229 }, { "epoch": 0.14, "learning_rate": 1.9363473626010414e-05, "loss": 0.9914, "step": 4230 }, { "epoch": 0.14, "learning_rate": 1.9363095154522097e-05, "loss": 1.0079, "step": 4231 }, { "epoch": 0.14, "learning_rate": 1.9362716574250724e-05, "loss": 1.0012, "step": 4232 }, { "epoch": 0.14, "learning_rate": 1.936233788520069e-05, "loss": 1.0089, "step": 4233 }, { "epoch": 0.14, "learning_rate": 1.9361959087376395e-05, "loss": 0.9599, "step": 4234 }, { "epoch": 0.14, "learning_rate": 1.9361580180782245e-05, "loss": 0.9977, "step": 4235 }, { "epoch": 0.14, "learning_rate": 1.9361201165422637e-05, "loss": 0.9708, "step": 4236 }, { "epoch": 0.14, "learning_rate": 1.9360822041301977e-05, "loss": 1.0333, "step": 4237 }, { "epoch": 0.14, "learning_rate": 1.9360442808424667e-05, "loss": 0.9629, "step": 4238 }, { "epoch": 0.14, "learning_rate": 1.9360063466795117e-05, "loss": 1.0405, "step": 4239 }, { "epoch": 0.14, "learning_rate": 1.9359684016417735e-05, "loss": 0.973, "step": 4240 }, { "epoch": 0.14, "learning_rate": 1.9359304457296924e-05, "loss": 0.9841, "step": 4241 }, { "epoch": 0.14, "learning_rate": 1.93589247894371e-05, "loss": 0.9618, "step": 4242 }, { "epoch": 0.14, "learning_rate": 1.935854501284267e-05, "loss": 1.0098, "step": 4243 }, { "epoch": 0.14, "learning_rate": 1.9358165127518043e-05, "loss": 0.9657, "step": 4244 }, { "epoch": 0.14, "learning_rate": 1.9357785133467644e-05, "loss": 1.0692, "step": 4245 }, { "epoch": 0.14, "learning_rate": 1.9357405030695873e-05, "loss": 1.0054, "step": 4246 }, { "epoch": 0.14, "learning_rate": 1.935702481920716e-05, "loss": 1.0072, "step": 4247 }, { "epoch": 0.14, "learning_rate": 1.9356644499005917e-05, "loss": 0.9748, "step": 4248 }, { "epoch": 0.14, "learning_rate": 1.935626407009656e-05, "loss": 1.0092, "step": 4249 }, { "epoch": 0.14, "learning_rate": 1.9355883532483507e-05, "loss": 0.9632, "step": 4250 }, { "epoch": 0.14, "learning_rate": 1.935550288617119e-05, "loss": 1.0055, "step": 4251 }, { "epoch": 0.14, "learning_rate": 1.935512213116402e-05, "loss": 0.9799, "step": 4252 }, { "epoch": 0.14, "learning_rate": 1.9354741267466427e-05, "loss": 0.9868, "step": 4253 }, { "epoch": 0.14, "learning_rate": 1.9354360295082836e-05, "loss": 0.9996, "step": 4254 }, { "epoch": 0.14, "learning_rate": 1.935397921401767e-05, "loss": 0.9812, "step": 4255 }, { "epoch": 0.14, "learning_rate": 1.9353598024275356e-05, "loss": 0.9948, "step": 4256 }, { "epoch": 0.14, "learning_rate": 1.9353216725860327e-05, "loss": 1.0145, "step": 4257 }, { "epoch": 0.14, "learning_rate": 1.9352835318777012e-05, "loss": 1.0164, "step": 4258 }, { "epoch": 0.14, "learning_rate": 1.935245380302984e-05, "loss": 0.9595, "step": 4259 }, { "epoch": 0.14, "learning_rate": 1.9352072178623244e-05, "loss": 0.9786, "step": 4260 }, { "epoch": 0.14, "learning_rate": 1.935169044556166e-05, "loss": 0.9834, "step": 4261 }, { "epoch": 0.14, "learning_rate": 1.9351308603849525e-05, "loss": 0.9599, "step": 4262 }, { "epoch": 0.14, "learning_rate": 1.9350926653491268e-05, "loss": 1.0285, "step": 4263 }, { "epoch": 0.14, "learning_rate": 1.935054459449133e-05, "loss": 0.978, "step": 4264 }, { "epoch": 0.14, "learning_rate": 1.9350162426854152e-05, "loss": 0.9769, "step": 4265 }, { "epoch": 0.14, "learning_rate": 1.934978015058417e-05, "loss": 0.9877, "step": 4266 }, { "epoch": 0.14, "learning_rate": 1.9349397765685827e-05, "loss": 0.9982, "step": 4267 }, { "epoch": 0.14, "learning_rate": 1.9349015272163573e-05, "loss": 0.9491, "step": 4268 }, { "epoch": 0.14, "learning_rate": 1.9348632670021842e-05, "loss": 1.0092, "step": 4269 }, { "epoch": 0.14, "learning_rate": 1.9348249959265083e-05, "loss": 0.9581, "step": 4270 }, { "epoch": 0.14, "learning_rate": 1.934786713989774e-05, "loss": 1.03, "step": 4271 }, { "epoch": 0.14, "learning_rate": 1.9347484211924266e-05, "loss": 0.9649, "step": 4272 }, { "epoch": 0.14, "learning_rate": 1.9347101175349104e-05, "loss": 1.035, "step": 4273 }, { "epoch": 0.14, "learning_rate": 1.934671803017671e-05, "loss": 1.0072, "step": 4274 }, { "epoch": 0.14, "learning_rate": 1.9346334776411532e-05, "loss": 0.9801, "step": 4275 }, { "epoch": 0.14, "learning_rate": 1.9345951414058023e-05, "loss": 0.993, "step": 4276 }, { "epoch": 0.14, "learning_rate": 1.9345567943120635e-05, "loss": 0.997, "step": 4277 }, { "epoch": 0.14, "learning_rate": 1.934518436360383e-05, "loss": 1.0031, "step": 4278 }, { "epoch": 0.14, "learning_rate": 1.934480067551206e-05, "loss": 0.9547, "step": 4279 }, { "epoch": 0.14, "learning_rate": 1.9344416878849776e-05, "loss": 0.9814, "step": 4280 }, { "epoch": 0.14, "learning_rate": 1.934403297362145e-05, "loss": 1.0022, "step": 4281 }, { "epoch": 0.14, "learning_rate": 1.9343648959831536e-05, "loss": 1.0188, "step": 4282 }, { "epoch": 0.14, "learning_rate": 1.9343264837484494e-05, "loss": 0.9947, "step": 4283 }, { "epoch": 0.14, "learning_rate": 1.934288060658479e-05, "loss": 0.9419, "step": 4284 }, { "epoch": 0.14, "learning_rate": 1.9342496267136884e-05, "loss": 1.0107, "step": 4285 }, { "epoch": 0.14, "learning_rate": 1.9342111819145247e-05, "loss": 0.9943, "step": 4286 }, { "epoch": 0.14, "learning_rate": 1.9341727262614344e-05, "loss": 1.0116, "step": 4287 }, { "epoch": 0.14, "learning_rate": 1.934134259754864e-05, "loss": 0.9558, "step": 4288 }, { "epoch": 0.14, "learning_rate": 1.934095782395261e-05, "loss": 0.9725, "step": 4289 }, { "epoch": 0.14, "learning_rate": 1.9340572941830716e-05, "loss": 1.0203, "step": 4290 }, { "epoch": 0.14, "learning_rate": 1.934018795118743e-05, "loss": 0.9684, "step": 4291 }, { "epoch": 0.14, "learning_rate": 1.9339802852027233e-05, "loss": 0.9864, "step": 4292 }, { "epoch": 0.14, "learning_rate": 1.9339417644354595e-05, "loss": 1.0252, "step": 4293 }, { "epoch": 0.14, "learning_rate": 1.933903232817399e-05, "loss": 1.0103, "step": 4294 }, { "epoch": 0.14, "learning_rate": 1.9338646903489903e-05, "loss": 1.0032, "step": 4295 }, { "epoch": 0.14, "learning_rate": 1.93382613703068e-05, "loss": 0.988, "step": 4296 }, { "epoch": 0.14, "learning_rate": 1.9337875728629164e-05, "loss": 0.9834, "step": 4297 }, { "epoch": 0.14, "learning_rate": 1.933748997846148e-05, "loss": 0.9434, "step": 4298 }, { "epoch": 0.14, "learning_rate": 1.9337104119808222e-05, "loss": 1.0065, "step": 4299 }, { "epoch": 0.14, "learning_rate": 1.9336718152673884e-05, "loss": 0.9847, "step": 4300 }, { "epoch": 0.14, "learning_rate": 1.933633207706294e-05, "loss": 0.9631, "step": 4301 }, { "epoch": 0.14, "learning_rate": 1.933594589297988e-05, "loss": 0.9919, "step": 4302 }, { "epoch": 0.14, "learning_rate": 1.9335559600429187e-05, "loss": 0.9991, "step": 4303 }, { "epoch": 0.14, "learning_rate": 1.9335173199415358e-05, "loss": 0.9616, "step": 4304 }, { "epoch": 0.14, "learning_rate": 1.9334786689942876e-05, "loss": 1.0008, "step": 4305 }, { "epoch": 0.14, "learning_rate": 1.9334400072016228e-05, "loss": 0.9917, "step": 4306 }, { "epoch": 0.14, "learning_rate": 1.9334013345639913e-05, "loss": 0.9929, "step": 4307 }, { "epoch": 0.14, "learning_rate": 1.9333626510818423e-05, "loss": 0.9497, "step": 4308 }, { "epoch": 0.14, "learning_rate": 1.933323956755625e-05, "loss": 0.9562, "step": 4309 }, { "epoch": 0.14, "learning_rate": 1.9332852515857886e-05, "loss": 1.0039, "step": 4310 }, { "epoch": 0.14, "learning_rate": 1.9332465355727833e-05, "loss": 0.9758, "step": 4311 }, { "epoch": 0.14, "learning_rate": 1.933207808717059e-05, "loss": 0.9927, "step": 4312 }, { "epoch": 0.14, "learning_rate": 1.9331690710190654e-05, "loss": 0.9799, "step": 4313 }, { "epoch": 0.14, "learning_rate": 1.9331303224792526e-05, "loss": 0.9692, "step": 4314 }, { "epoch": 0.14, "learning_rate": 1.933091563098071e-05, "loss": 0.9319, "step": 4315 }, { "epoch": 0.14, "learning_rate": 1.9330527928759705e-05, "loss": 0.9523, "step": 4316 }, { "epoch": 0.14, "learning_rate": 1.9330140118134017e-05, "loss": 0.9726, "step": 4317 }, { "epoch": 0.14, "learning_rate": 1.9329752199108153e-05, "loss": 0.9902, "step": 4318 }, { "epoch": 0.14, "learning_rate": 1.9329364171686622e-05, "loss": 1.013, "step": 4319 }, { "epoch": 0.14, "learning_rate": 1.9328976035873927e-05, "loss": 0.9803, "step": 4320 }, { "epoch": 0.14, "learning_rate": 1.932858779167458e-05, "loss": 0.9734, "step": 4321 }, { "epoch": 0.14, "learning_rate": 1.9328199439093094e-05, "loss": 0.9717, "step": 4322 }, { "epoch": 0.14, "learning_rate": 1.9327810978133975e-05, "loss": 0.9872, "step": 4323 }, { "epoch": 0.14, "learning_rate": 1.9327422408801744e-05, "loss": 0.9406, "step": 4324 }, { "epoch": 0.14, "learning_rate": 1.932703373110091e-05, "loss": 0.992, "step": 4325 }, { "epoch": 0.14, "learning_rate": 1.9326644945035987e-05, "loss": 1.034, "step": 4326 }, { "epoch": 0.14, "learning_rate": 1.93262560506115e-05, "loss": 0.9803, "step": 4327 }, { "epoch": 0.14, "learning_rate": 1.932586704783196e-05, "loss": 0.9392, "step": 4328 }, { "epoch": 0.14, "learning_rate": 1.9325477936701888e-05, "loss": 0.9807, "step": 4329 }, { "epoch": 0.14, "learning_rate": 1.9325088717225808e-05, "loss": 0.9638, "step": 4330 }, { "epoch": 0.14, "learning_rate": 1.932469938940824e-05, "loss": 1.0192, "step": 4331 }, { "epoch": 0.14, "learning_rate": 1.9324309953253706e-05, "loss": 0.9492, "step": 4332 }, { "epoch": 0.14, "learning_rate": 1.9323920408766734e-05, "loss": 1.0054, "step": 4333 }, { "epoch": 0.14, "learning_rate": 1.9323530755951848e-05, "loss": 0.9913, "step": 4334 }, { "epoch": 0.14, "learning_rate": 1.9323140994813573e-05, "loss": 1.0067, "step": 4335 }, { "epoch": 0.14, "learning_rate": 1.9322751125356438e-05, "loss": 0.979, "step": 4336 }, { "epoch": 0.14, "learning_rate": 1.9322361147584978e-05, "loss": 0.981, "step": 4337 }, { "epoch": 0.14, "learning_rate": 1.9321971061503716e-05, "loss": 0.9975, "step": 4338 }, { "epoch": 0.14, "learning_rate": 1.9321580867117186e-05, "loss": 0.9891, "step": 4339 }, { "epoch": 0.14, "learning_rate": 1.932119056442993e-05, "loss": 0.9388, "step": 4340 }, { "epoch": 0.14, "learning_rate": 1.932080015344647e-05, "loss": 1.0246, "step": 4341 }, { "epoch": 0.14, "learning_rate": 1.932040963417135e-05, "loss": 0.9754, "step": 4342 }, { "epoch": 0.14, "learning_rate": 1.9320019006609103e-05, "loss": 1.0084, "step": 4343 }, { "epoch": 0.14, "learning_rate": 1.931962827076427e-05, "loss": 0.9899, "step": 4344 }, { "epoch": 0.14, "learning_rate": 1.931923742664139e-05, "loss": 0.9595, "step": 4345 }, { "epoch": 0.14, "learning_rate": 1.9318846474245007e-05, "loss": 1.0377, "step": 4346 }, { "epoch": 0.14, "learning_rate": 1.9318455413579657e-05, "loss": 0.9949, "step": 4347 }, { "epoch": 0.14, "learning_rate": 1.9318064244649887e-05, "loss": 0.9329, "step": 4348 }, { "epoch": 0.14, "learning_rate": 1.9317672967460243e-05, "loss": 0.974, "step": 4349 }, { "epoch": 0.14, "learning_rate": 1.9317281582015264e-05, "loss": 1.012, "step": 4350 }, { "epoch": 0.14, "learning_rate": 1.931689008831951e-05, "loss": 0.9984, "step": 4351 }, { "epoch": 0.14, "learning_rate": 1.9316498486377514e-05, "loss": 0.9926, "step": 4352 }, { "epoch": 0.14, "learning_rate": 1.931610677619384e-05, "loss": 1.0211, "step": 4353 }, { "epoch": 0.14, "learning_rate": 1.931571495777303e-05, "loss": 0.9664, "step": 4354 }, { "epoch": 0.14, "learning_rate": 1.931532303111964e-05, "loss": 1.0043, "step": 4355 }, { "epoch": 0.14, "learning_rate": 1.931493099623822e-05, "loss": 0.9655, "step": 4356 }, { "epoch": 0.15, "learning_rate": 1.931453885313333e-05, "loss": 0.9293, "step": 4357 }, { "epoch": 0.15, "learning_rate": 1.931414660180952e-05, "loss": 1.0099, "step": 4358 }, { "epoch": 0.15, "learning_rate": 1.9313754242271354e-05, "loss": 0.9967, "step": 4359 }, { "epoch": 0.15, "learning_rate": 1.9313361774523387e-05, "loss": 0.9598, "step": 4360 }, { "epoch": 0.15, "learning_rate": 1.9312969198570176e-05, "loss": 0.9513, "step": 4361 }, { "epoch": 0.15, "learning_rate": 1.931257651441629e-05, "loss": 0.9857, "step": 4362 }, { "epoch": 0.15, "learning_rate": 1.931218372206628e-05, "loss": 0.9545, "step": 4363 }, { "epoch": 0.15, "learning_rate": 1.931179082152472e-05, "loss": 0.9547, "step": 4364 }, { "epoch": 0.15, "learning_rate": 1.9311397812796168e-05, "loss": 0.9747, "step": 4365 }, { "epoch": 0.15, "learning_rate": 1.9311004695885193e-05, "loss": 0.9932, "step": 4366 }, { "epoch": 0.15, "learning_rate": 1.9310611470796363e-05, "loss": 1.0042, "step": 4367 }, { "epoch": 0.15, "learning_rate": 1.9310218137534247e-05, "loss": 0.9173, "step": 4368 }, { "epoch": 0.15, "learning_rate": 1.930982469610341e-05, "loss": 0.9853, "step": 4369 }, { "epoch": 0.15, "learning_rate": 1.9309431146508433e-05, "loss": 0.9447, "step": 4370 }, { "epoch": 0.15, "learning_rate": 1.9309037488753876e-05, "loss": 1.0012, "step": 4371 }, { "epoch": 0.15, "learning_rate": 1.930864372284432e-05, "loss": 0.9262, "step": 4372 }, { "epoch": 0.15, "learning_rate": 1.930824984878434e-05, "loss": 0.9554, "step": 4373 }, { "epoch": 0.15, "learning_rate": 1.9307855866578513e-05, "loss": 0.9834, "step": 4374 }, { "epoch": 0.15, "learning_rate": 1.930746177623141e-05, "loss": 0.9745, "step": 4375 }, { "epoch": 0.15, "learning_rate": 1.930706757774761e-05, "loss": 0.9853, "step": 4376 }, { "epoch": 0.15, "learning_rate": 1.9306673271131705e-05, "loss": 0.9856, "step": 4377 }, { "epoch": 0.15, "learning_rate": 1.9306278856388262e-05, "loss": 1.0127, "step": 4378 }, { "epoch": 0.15, "learning_rate": 1.9305884333521873e-05, "loss": 0.9647, "step": 4379 }, { "epoch": 0.15, "learning_rate": 1.9305489702537116e-05, "loss": 0.8956, "step": 4380 }, { "epoch": 0.15, "learning_rate": 1.9305094963438574e-05, "loss": 0.9639, "step": 4381 }, { "epoch": 0.15, "learning_rate": 1.9304700116230842e-05, "loss": 1.013, "step": 4382 }, { "epoch": 0.15, "learning_rate": 1.93043051609185e-05, "loss": 1.023, "step": 4383 }, { "epoch": 0.15, "learning_rate": 1.9303910097506144e-05, "loss": 0.9702, "step": 4384 }, { "epoch": 0.15, "learning_rate": 1.9303514925998353e-05, "loss": 0.9849, "step": 4385 }, { "epoch": 0.15, "learning_rate": 1.930311964639973e-05, "loss": 0.948, "step": 4386 }, { "epoch": 0.15, "learning_rate": 1.9302724258714856e-05, "loss": 0.9894, "step": 4387 }, { "epoch": 0.15, "learning_rate": 1.9302328762948335e-05, "loss": 0.9449, "step": 4388 }, { "epoch": 0.15, "learning_rate": 1.9301933159104754e-05, "loss": 0.9992, "step": 4389 }, { "epoch": 0.15, "learning_rate": 1.9301537447188713e-05, "loss": 1.011, "step": 4390 }, { "epoch": 0.15, "learning_rate": 1.9301141627204813e-05, "loss": 0.9757, "step": 4391 }, { "epoch": 0.15, "learning_rate": 1.9300745699157647e-05, "loss": 0.9672, "step": 4392 }, { "epoch": 0.15, "learning_rate": 1.9300349663051817e-05, "loss": 0.9984, "step": 4393 }, { "epoch": 0.15, "learning_rate": 1.9299953518891922e-05, "loss": 0.9772, "step": 4394 }, { "epoch": 0.15, "learning_rate": 1.929955726668257e-05, "loss": 1.0358, "step": 4395 }, { "epoch": 0.15, "learning_rate": 1.929916090642836e-05, "loss": 0.9993, "step": 4396 }, { "epoch": 0.15, "learning_rate": 1.9298764438133898e-05, "loss": 0.998, "step": 4397 }, { "epoch": 0.15, "learning_rate": 1.929836786180379e-05, "loss": 0.9981, "step": 4398 }, { "epoch": 0.15, "learning_rate": 1.9297971177442646e-05, "loss": 1.0216, "step": 4399 }, { "epoch": 0.15, "learning_rate": 1.9297574385055074e-05, "loss": 0.9345, "step": 4400 }, { "epoch": 0.15, "learning_rate": 1.9297177484645682e-05, "loss": 1.0197, "step": 4401 }, { "epoch": 0.15, "learning_rate": 1.9296780476219084e-05, "loss": 1.0077, "step": 4402 }, { "epoch": 0.15, "learning_rate": 1.9296383359779888e-05, "loss": 0.9894, "step": 4403 }, { "epoch": 0.15, "learning_rate": 1.9295986135332712e-05, "loss": 0.9557, "step": 4404 }, { "epoch": 0.15, "learning_rate": 1.929558880288217e-05, "loss": 0.9308, "step": 4405 }, { "epoch": 0.15, "learning_rate": 1.929519136243288e-05, "loss": 0.9787, "step": 4406 }, { "epoch": 0.15, "learning_rate": 1.9294793813989458e-05, "loss": 0.9772, "step": 4407 }, { "epoch": 0.15, "learning_rate": 1.929439615755652e-05, "loss": 0.9601, "step": 4408 }, { "epoch": 0.15, "learning_rate": 1.9293998393138687e-05, "loss": 0.9566, "step": 4409 }, { "epoch": 0.15, "learning_rate": 1.9293600520740587e-05, "loss": 0.9971, "step": 4410 }, { "epoch": 0.15, "learning_rate": 1.9293202540366833e-05, "loss": 0.9995, "step": 4411 }, { "epoch": 0.15, "learning_rate": 1.9292804452022057e-05, "loss": 0.9995, "step": 4412 }, { "epoch": 0.15, "learning_rate": 1.929240625571088e-05, "loss": 1.0046, "step": 4413 }, { "epoch": 0.15, "learning_rate": 1.9292007951437927e-05, "loss": 0.9611, "step": 4414 }, { "epoch": 0.15, "learning_rate": 1.929160953920783e-05, "loss": 0.9764, "step": 4415 }, { "epoch": 0.15, "learning_rate": 1.9291211019025216e-05, "loss": 1.0182, "step": 4416 }, { "epoch": 0.15, "learning_rate": 1.9290812390894712e-05, "loss": 1.0092, "step": 4417 }, { "epoch": 0.15, "learning_rate": 1.929041365482095e-05, "loss": 0.9799, "step": 4418 }, { "epoch": 0.15, "learning_rate": 1.9290014810808572e-05, "loss": 0.9898, "step": 4419 }, { "epoch": 0.15, "learning_rate": 1.92896158588622e-05, "loss": 0.9605, "step": 4420 }, { "epoch": 0.15, "learning_rate": 1.928921679898647e-05, "loss": 1.0193, "step": 4421 }, { "epoch": 0.15, "learning_rate": 1.9288817631186025e-05, "loss": 0.984, "step": 4422 }, { "epoch": 0.15, "learning_rate": 1.92884183554655e-05, "loss": 0.9769, "step": 4423 }, { "epoch": 0.15, "learning_rate": 1.9288018971829535e-05, "loss": 0.9905, "step": 4424 }, { "epoch": 0.15, "learning_rate": 1.9287619480282765e-05, "loss": 1.0052, "step": 4425 }, { "epoch": 0.15, "learning_rate": 1.928721988082984e-05, "loss": 0.9748, "step": 4426 }, { "epoch": 0.15, "learning_rate": 1.9286820173475395e-05, "loss": 0.9661, "step": 4427 }, { "epoch": 0.15, "learning_rate": 1.9286420358224074e-05, "loss": 0.9513, "step": 4428 }, { "epoch": 0.15, "learning_rate": 1.928602043508053e-05, "loss": 0.9944, "step": 4429 }, { "epoch": 0.15, "learning_rate": 1.92856204040494e-05, "loss": 0.9714, "step": 4430 }, { "epoch": 0.15, "learning_rate": 1.9285220265135337e-05, "loss": 1.0109, "step": 4431 }, { "epoch": 0.15, "learning_rate": 1.928482001834299e-05, "loss": 1.0234, "step": 4432 }, { "epoch": 0.15, "learning_rate": 1.9284419663677006e-05, "loss": 1.0118, "step": 4433 }, { "epoch": 0.15, "learning_rate": 1.928401920114204e-05, "loss": 0.9823, "step": 4434 }, { "epoch": 0.15, "learning_rate": 1.928361863074274e-05, "loss": 0.9623, "step": 4435 }, { "epoch": 0.15, "learning_rate": 1.9283217952483767e-05, "loss": 0.9699, "step": 4436 }, { "epoch": 0.15, "learning_rate": 1.928281716636977e-05, "loss": 0.9705, "step": 4437 }, { "epoch": 0.15, "learning_rate": 1.928241627240541e-05, "loss": 0.9972, "step": 4438 }, { "epoch": 0.15, "learning_rate": 1.928201527059534e-05, "loss": 0.9664, "step": 4439 }, { "epoch": 0.15, "learning_rate": 1.9281614160944226e-05, "loss": 0.9769, "step": 4440 }, { "epoch": 0.15, "learning_rate": 1.9281212943456715e-05, "loss": 1.0309, "step": 4441 }, { "epoch": 0.15, "learning_rate": 1.9280811618137486e-05, "loss": 0.9935, "step": 4442 }, { "epoch": 0.15, "learning_rate": 1.9280410184991186e-05, "loss": 0.9915, "step": 4443 }, { "epoch": 0.15, "learning_rate": 1.928000864402249e-05, "loss": 0.9693, "step": 4444 }, { "epoch": 0.15, "learning_rate": 1.9279606995236056e-05, "loss": 0.9625, "step": 4445 }, { "epoch": 0.15, "learning_rate": 1.9279205238636554e-05, "loss": 1.0001, "step": 4446 }, { "epoch": 0.15, "learning_rate": 1.927880337422865e-05, "loss": 0.9467, "step": 4447 }, { "epoch": 0.15, "learning_rate": 1.9278401402017017e-05, "loss": 0.96, "step": 4448 }, { "epoch": 0.15, "learning_rate": 1.927799932200632e-05, "loss": 1.0418, "step": 4449 }, { "epoch": 0.15, "learning_rate": 1.9277597134201232e-05, "loss": 0.9662, "step": 4450 }, { "epoch": 0.15, "learning_rate": 1.927719483860643e-05, "loss": 0.9894, "step": 4451 }, { "epoch": 0.15, "learning_rate": 1.9276792435226582e-05, "loss": 0.9643, "step": 4452 }, { "epoch": 0.15, "learning_rate": 1.9276389924066366e-05, "loss": 0.9889, "step": 4453 }, { "epoch": 0.15, "learning_rate": 1.927598730513046e-05, "loss": 1.0021, "step": 4454 }, { "epoch": 0.15, "learning_rate": 1.9275584578423535e-05, "loss": 0.9937, "step": 4455 }, { "epoch": 0.15, "learning_rate": 1.927518174395028e-05, "loss": 0.9842, "step": 4456 }, { "epoch": 0.15, "learning_rate": 1.9274778801715367e-05, "loss": 0.9672, "step": 4457 }, { "epoch": 0.15, "learning_rate": 1.9274375751723483e-05, "loss": 0.9771, "step": 4458 }, { "epoch": 0.15, "learning_rate": 1.9273972593979305e-05, "loss": 0.9995, "step": 4459 }, { "epoch": 0.15, "learning_rate": 1.9273569328487522e-05, "loss": 0.9827, "step": 4460 }, { "epoch": 0.15, "learning_rate": 1.927316595525282e-05, "loss": 0.9809, "step": 4461 }, { "epoch": 0.15, "learning_rate": 1.927276247427988e-05, "loss": 0.9939, "step": 4462 }, { "epoch": 0.15, "learning_rate": 1.9272358885573394e-05, "loss": 0.9611, "step": 4463 }, { "epoch": 0.15, "learning_rate": 1.927195518913805e-05, "loss": 0.9774, "step": 4464 }, { "epoch": 0.15, "learning_rate": 1.927155138497854e-05, "loss": 1.019, "step": 4465 }, { "epoch": 0.15, "learning_rate": 1.927114747309955e-05, "loss": 0.9494, "step": 4466 }, { "epoch": 0.15, "learning_rate": 1.927074345350578e-05, "loss": 1.0247, "step": 4467 }, { "epoch": 0.15, "learning_rate": 1.927033932620192e-05, "loss": 0.9532, "step": 4468 }, { "epoch": 0.15, "learning_rate": 1.9269935091192664e-05, "loss": 1.028, "step": 4469 }, { "epoch": 0.15, "learning_rate": 1.926953074848271e-05, "loss": 1.0133, "step": 4470 }, { "epoch": 0.15, "learning_rate": 1.926912629807676e-05, "loss": 1.0066, "step": 4471 }, { "epoch": 0.15, "learning_rate": 1.9268721739979504e-05, "loss": 0.9289, "step": 4472 }, { "epoch": 0.15, "learning_rate": 1.9268317074195647e-05, "loss": 0.9969, "step": 4473 }, { "epoch": 0.15, "learning_rate": 1.9267912300729895e-05, "loss": 0.986, "step": 4474 }, { "epoch": 0.15, "learning_rate": 1.9267507419586943e-05, "loss": 0.9653, "step": 4475 }, { "epoch": 0.15, "learning_rate": 1.92671024307715e-05, "loss": 0.9717, "step": 4476 }, { "epoch": 0.15, "learning_rate": 1.9266697334288273e-05, "loss": 0.9977, "step": 4477 }, { "epoch": 0.15, "learning_rate": 1.926629213014196e-05, "loss": 0.9966, "step": 4478 }, { "epoch": 0.15, "learning_rate": 1.9265886818337277e-05, "loss": 1.0058, "step": 4479 }, { "epoch": 0.15, "learning_rate": 1.926548139887893e-05, "loss": 0.9501, "step": 4480 }, { "epoch": 0.15, "learning_rate": 1.9265075871771627e-05, "loss": 0.97, "step": 4481 }, { "epoch": 0.15, "learning_rate": 1.9264670237020086e-05, "loss": 0.9968, "step": 4482 }, { "epoch": 0.15, "learning_rate": 1.9264264494629016e-05, "loss": 0.953, "step": 4483 }, { "epoch": 0.15, "learning_rate": 1.9263858644603127e-05, "loss": 0.9894, "step": 4484 }, { "epoch": 0.15, "learning_rate": 1.926345268694714e-05, "loss": 0.9868, "step": 4485 }, { "epoch": 0.15, "learning_rate": 1.926304662166577e-05, "loss": 0.9878, "step": 4486 }, { "epoch": 0.15, "learning_rate": 1.926264044876373e-05, "loss": 1.0014, "step": 4487 }, { "epoch": 0.15, "learning_rate": 1.9262234168245747e-05, "loss": 0.9523, "step": 4488 }, { "epoch": 0.15, "learning_rate": 1.926182778011654e-05, "loss": 0.9971, "step": 4489 }, { "epoch": 0.15, "learning_rate": 1.9261421284380825e-05, "loss": 0.9731, "step": 4490 }, { "epoch": 0.15, "learning_rate": 1.9261014681043332e-05, "loss": 1.0266, "step": 4491 }, { "epoch": 0.15, "learning_rate": 1.9260607970108775e-05, "loss": 0.9467, "step": 4492 }, { "epoch": 0.15, "learning_rate": 1.926020115158189e-05, "loss": 0.9521, "step": 4493 }, { "epoch": 0.15, "learning_rate": 1.92597942254674e-05, "loss": 0.9424, "step": 4494 }, { "epoch": 0.15, "learning_rate": 1.9259387191770026e-05, "loss": 0.9664, "step": 4495 }, { "epoch": 0.15, "learning_rate": 1.9258980050494505e-05, "loss": 0.9844, "step": 4496 }, { "epoch": 0.15, "learning_rate": 1.9258572801645567e-05, "loss": 1.005, "step": 4497 }, { "epoch": 0.15, "learning_rate": 1.925816544522794e-05, "loss": 0.9514, "step": 4498 }, { "epoch": 0.15, "learning_rate": 1.9257757981246357e-05, "loss": 0.9779, "step": 4499 }, { "epoch": 0.15, "learning_rate": 1.9257350409705557e-05, "loss": 0.9635, "step": 4500 }, { "epoch": 0.15, "learning_rate": 1.9256942730610268e-05, "loss": 0.9968, "step": 4501 }, { "epoch": 0.15, "learning_rate": 1.925653494396523e-05, "loss": 0.985, "step": 4502 }, { "epoch": 0.15, "learning_rate": 1.9256127049775184e-05, "loss": 1.0224, "step": 4503 }, { "epoch": 0.15, "learning_rate": 1.9255719048044862e-05, "loss": 0.9836, "step": 4504 }, { "epoch": 0.15, "learning_rate": 1.9255310938779012e-05, "loss": 0.9671, "step": 4505 }, { "epoch": 0.15, "learning_rate": 1.925490272198237e-05, "loss": 0.981, "step": 4506 }, { "epoch": 0.15, "learning_rate": 1.9254494397659682e-05, "loss": 0.9494, "step": 4507 }, { "epoch": 0.15, "learning_rate": 1.925408596581569e-05, "loss": 0.9614, "step": 4508 }, { "epoch": 0.15, "learning_rate": 1.9253677426455138e-05, "loss": 0.9861, "step": 4509 }, { "epoch": 0.15, "learning_rate": 1.9253268779582775e-05, "loss": 0.9741, "step": 4510 }, { "epoch": 0.15, "learning_rate": 1.9252860025203347e-05, "loss": 0.9792, "step": 4511 }, { "epoch": 0.15, "learning_rate": 1.9252451163321605e-05, "loss": 0.9596, "step": 4512 }, { "epoch": 0.15, "learning_rate": 1.92520421939423e-05, "loss": 1.0265, "step": 4513 }, { "epoch": 0.15, "learning_rate": 1.925163311707018e-05, "loss": 0.9521, "step": 4514 }, { "epoch": 0.15, "learning_rate": 1.925122393271e-05, "loss": 0.9832, "step": 4515 }, { "epoch": 0.15, "learning_rate": 1.9250814640866514e-05, "loss": 0.9909, "step": 4516 }, { "epoch": 0.15, "learning_rate": 1.9250405241544477e-05, "loss": 0.9878, "step": 4517 }, { "epoch": 0.15, "learning_rate": 1.9249995734748647e-05, "loss": 0.965, "step": 4518 }, { "epoch": 0.15, "learning_rate": 1.924958612048378e-05, "loss": 0.9464, "step": 4519 }, { "epoch": 0.15, "learning_rate": 1.924917639875463e-05, "loss": 0.9481, "step": 4520 }, { "epoch": 0.15, "learning_rate": 1.924876656956597e-05, "loss": 1.009, "step": 4521 }, { "epoch": 0.15, "learning_rate": 1.924835663292255e-05, "loss": 0.993, "step": 4522 }, { "epoch": 0.15, "learning_rate": 1.9247946588829138e-05, "loss": 0.9483, "step": 4523 }, { "epoch": 0.15, "learning_rate": 1.9247536437290497e-05, "loss": 0.932, "step": 4524 }, { "epoch": 0.15, "learning_rate": 1.9247126178311388e-05, "loss": 1.0144, "step": 4525 }, { "epoch": 0.15, "learning_rate": 1.9246715811896585e-05, "loss": 1.0183, "step": 4526 }, { "epoch": 0.15, "learning_rate": 1.9246305338050852e-05, "loss": 0.9781, "step": 4527 }, { "epoch": 0.15, "learning_rate": 1.924589475677896e-05, "loss": 0.9767, "step": 4528 }, { "epoch": 0.15, "learning_rate": 1.9245484068085677e-05, "loss": 1.0123, "step": 4529 }, { "epoch": 0.15, "learning_rate": 1.924507327197577e-05, "loss": 0.988, "step": 4530 }, { "epoch": 0.15, "learning_rate": 1.9244662368454025e-05, "loss": 0.9628, "step": 4531 }, { "epoch": 0.15, "learning_rate": 1.92442513575252e-05, "loss": 0.9634, "step": 4532 }, { "epoch": 0.15, "learning_rate": 1.9243840239194084e-05, "loss": 1.0089, "step": 4533 }, { "epoch": 0.15, "learning_rate": 1.9243429013465444e-05, "loss": 0.9833, "step": 4534 }, { "epoch": 0.15, "learning_rate": 1.924301768034406e-05, "loss": 1.003, "step": 4535 }, { "epoch": 0.15, "learning_rate": 1.9242606239834717e-05, "loss": 1.0007, "step": 4536 }, { "epoch": 0.15, "learning_rate": 1.9242194691942187e-05, "loss": 0.9955, "step": 4537 }, { "epoch": 0.15, "learning_rate": 1.9241783036671258e-05, "loss": 0.9656, "step": 4538 }, { "epoch": 0.15, "learning_rate": 1.9241371274026708e-05, "loss": 0.9989, "step": 4539 }, { "epoch": 0.15, "learning_rate": 1.924095940401332e-05, "loss": 1.0083, "step": 4540 }, { "epoch": 0.15, "learning_rate": 1.9240547426635884e-05, "loss": 0.9996, "step": 4541 }, { "epoch": 0.15, "learning_rate": 1.9240135341899186e-05, "loss": 0.9673, "step": 4542 }, { "epoch": 0.15, "learning_rate": 1.923972314980801e-05, "loss": 0.9861, "step": 4543 }, { "epoch": 0.15, "learning_rate": 1.9239310850367148e-05, "loss": 0.9801, "step": 4544 }, { "epoch": 0.15, "learning_rate": 1.9238898443581387e-05, "loss": 1.0054, "step": 4545 }, { "epoch": 0.15, "learning_rate": 1.9238485929455522e-05, "loss": 0.9977, "step": 4546 }, { "epoch": 0.15, "learning_rate": 1.9238073307994347e-05, "loss": 0.968, "step": 4547 }, { "epoch": 0.15, "learning_rate": 1.9237660579202648e-05, "loss": 0.9662, "step": 4548 }, { "epoch": 0.15, "learning_rate": 1.923724774308523e-05, "loss": 1.0024, "step": 4549 }, { "epoch": 0.15, "learning_rate": 1.9236834799646883e-05, "loss": 0.9799, "step": 4550 }, { "epoch": 0.15, "learning_rate": 1.923642174889241e-05, "loss": 0.9622, "step": 4551 }, { "epoch": 0.15, "learning_rate": 1.92360085908266e-05, "loss": 0.9866, "step": 4552 }, { "epoch": 0.15, "learning_rate": 1.9235595325454265e-05, "loss": 0.9795, "step": 4553 }, { "epoch": 0.15, "learning_rate": 1.9235181952780203e-05, "loss": 0.9699, "step": 4554 }, { "epoch": 0.15, "learning_rate": 1.923476847280921e-05, "loss": 0.9837, "step": 4555 }, { "epoch": 0.15, "learning_rate": 1.9234354885546096e-05, "loss": 0.9779, "step": 4556 }, { "epoch": 0.15, "learning_rate": 1.9233941190995665e-05, "loss": 0.9799, "step": 4557 }, { "epoch": 0.15, "learning_rate": 1.9233527389162723e-05, "loss": 0.9728, "step": 4558 }, { "epoch": 0.15, "learning_rate": 1.923311348005208e-05, "loss": 0.9584, "step": 4559 }, { "epoch": 0.15, "learning_rate": 1.9232699463668543e-05, "loss": 0.9561, "step": 4560 }, { "epoch": 0.15, "learning_rate": 1.9232285340016922e-05, "loss": 0.9805, "step": 4561 }, { "epoch": 0.15, "learning_rate": 1.9231871109102027e-05, "loss": 0.9543, "step": 4562 }, { "epoch": 0.15, "learning_rate": 1.9231456770928673e-05, "loss": 0.98, "step": 4563 }, { "epoch": 0.15, "learning_rate": 1.923104232550167e-05, "loss": 0.9959, "step": 4564 }, { "epoch": 0.15, "learning_rate": 1.9230627772825843e-05, "loss": 0.9733, "step": 4565 }, { "epoch": 0.15, "learning_rate": 1.9230213112906e-05, "loss": 0.9879, "step": 4566 }, { "epoch": 0.15, "learning_rate": 1.9229798345746955e-05, "loss": 1.0112, "step": 4567 }, { "epoch": 0.15, "learning_rate": 1.9229383471353536e-05, "loss": 0.9783, "step": 4568 }, { "epoch": 0.15, "learning_rate": 1.9228968489730558e-05, "loss": 0.9589, "step": 4569 }, { "epoch": 0.15, "learning_rate": 1.9228553400882844e-05, "loss": 0.96, "step": 4570 }, { "epoch": 0.15, "learning_rate": 1.922813820481522e-05, "loss": 1.0128, "step": 4571 }, { "epoch": 0.15, "learning_rate": 1.9227722901532503e-05, "loss": 0.9638, "step": 4572 }, { "epoch": 0.15, "learning_rate": 1.9227307491039518e-05, "loss": 0.9984, "step": 4573 }, { "epoch": 0.15, "learning_rate": 1.92268919733411e-05, "loss": 1.0087, "step": 4574 }, { "epoch": 0.15, "learning_rate": 1.922647634844207e-05, "loss": 1.0003, "step": 4575 }, { "epoch": 0.15, "learning_rate": 1.9226060616347254e-05, "loss": 0.9677, "step": 4576 }, { "epoch": 0.15, "learning_rate": 1.9225644777061488e-05, "loss": 0.9756, "step": 4577 }, { "epoch": 0.15, "learning_rate": 1.9225228830589605e-05, "loss": 0.9857, "step": 4578 }, { "epoch": 0.15, "learning_rate": 1.9224812776936433e-05, "loss": 0.9658, "step": 4579 }, { "epoch": 0.15, "learning_rate": 1.9224396616106803e-05, "loss": 0.9807, "step": 4580 }, { "epoch": 0.15, "learning_rate": 1.9223980348105557e-05, "loss": 0.9545, "step": 4581 }, { "epoch": 0.15, "learning_rate": 1.922356397293753e-05, "loss": 0.9479, "step": 4582 }, { "epoch": 0.15, "learning_rate": 1.9223147490607555e-05, "loss": 1.0172, "step": 4583 }, { "epoch": 0.15, "learning_rate": 1.9222730901120473e-05, "loss": 0.9752, "step": 4584 }, { "epoch": 0.15, "learning_rate": 1.9222314204481125e-05, "loss": 0.9638, "step": 4585 }, { "epoch": 0.15, "learning_rate": 1.9221897400694355e-05, "loss": 0.9866, "step": 4586 }, { "epoch": 0.15, "learning_rate": 1.9221480489764998e-05, "loss": 0.9724, "step": 4587 }, { "epoch": 0.15, "learning_rate": 1.9221063471697907e-05, "loss": 0.9458, "step": 4588 }, { "epoch": 0.15, "learning_rate": 1.922064634649792e-05, "loss": 1.0022, "step": 4589 }, { "epoch": 0.15, "learning_rate": 1.9220229114169886e-05, "loss": 0.9981, "step": 4590 }, { "epoch": 0.15, "learning_rate": 1.921981177471865e-05, "loss": 0.9741, "step": 4591 }, { "epoch": 0.15, "learning_rate": 1.921939432814907e-05, "loss": 1.0016, "step": 4592 }, { "epoch": 0.15, "learning_rate": 1.921897677446598e-05, "loss": 0.9705, "step": 4593 }, { "epoch": 0.15, "learning_rate": 1.9218559113674244e-05, "loss": 0.9675, "step": 4594 }, { "epoch": 0.15, "learning_rate": 1.921814134577871e-05, "loss": 0.9622, "step": 4595 }, { "epoch": 0.15, "learning_rate": 1.9217723470784233e-05, "loss": 0.9999, "step": 4596 }, { "epoch": 0.15, "learning_rate": 1.921730548869567e-05, "loss": 1.0057, "step": 4597 }, { "epoch": 0.15, "learning_rate": 1.9216887399517872e-05, "loss": 0.9877, "step": 4598 }, { "epoch": 0.15, "learning_rate": 1.92164692032557e-05, "loss": 0.9453, "step": 4599 }, { "epoch": 0.15, "learning_rate": 1.921605089991401e-05, "loss": 0.9839, "step": 4600 }, { "epoch": 0.15, "learning_rate": 1.9215632489497667e-05, "loss": 1.0108, "step": 4601 }, { "epoch": 0.15, "learning_rate": 1.9215213972011524e-05, "loss": 0.9617, "step": 4602 }, { "epoch": 0.15, "learning_rate": 1.9214795347460454e-05, "loss": 0.9231, "step": 4603 }, { "epoch": 0.15, "learning_rate": 1.9214376615849312e-05, "loss": 1.0179, "step": 4604 }, { "epoch": 0.15, "learning_rate": 1.9213957777182967e-05, "loss": 0.9644, "step": 4605 }, { "epoch": 0.15, "learning_rate": 1.9213538831466286e-05, "loss": 0.9699, "step": 4606 }, { "epoch": 0.15, "learning_rate": 1.921311977870413e-05, "loss": 0.988, "step": 4607 }, { "epoch": 0.15, "learning_rate": 1.9212700618901377e-05, "loss": 0.9753, "step": 4608 }, { "epoch": 0.15, "learning_rate": 1.9212281352062892e-05, "loss": 0.9644, "step": 4609 }, { "epoch": 0.15, "learning_rate": 1.9211861978193544e-05, "loss": 0.9991, "step": 4610 }, { "epoch": 0.15, "learning_rate": 1.921144249729821e-05, "loss": 0.9949, "step": 4611 }, { "epoch": 0.15, "learning_rate": 1.921102290938176e-05, "loss": 0.9356, "step": 4612 }, { "epoch": 0.15, "learning_rate": 1.921060321444907e-05, "loss": 1.0038, "step": 4613 }, { "epoch": 0.15, "learning_rate": 1.921018341250502e-05, "loss": 0.9734, "step": 4614 }, { "epoch": 0.15, "learning_rate": 1.920976350355448e-05, "loss": 1.0061, "step": 4615 }, { "epoch": 0.15, "learning_rate": 1.9209343487602333e-05, "loss": 1.0206, "step": 4616 }, { "epoch": 0.15, "learning_rate": 1.9208923364653462e-05, "loss": 0.9464, "step": 4617 }, { "epoch": 0.15, "learning_rate": 1.9208503134712742e-05, "loss": 0.95, "step": 4618 }, { "epoch": 0.15, "learning_rate": 1.9208082797785057e-05, "loss": 0.9866, "step": 4619 }, { "epoch": 0.15, "learning_rate": 1.920766235387529e-05, "loss": 0.9897, "step": 4620 }, { "epoch": 0.15, "learning_rate": 1.9207241802988332e-05, "loss": 0.9643, "step": 4621 }, { "epoch": 0.15, "learning_rate": 1.920682114512906e-05, "loss": 0.9877, "step": 4622 }, { "epoch": 0.15, "learning_rate": 1.920640038030237e-05, "loss": 0.9697, "step": 4623 }, { "epoch": 0.15, "learning_rate": 1.9205979508513145e-05, "loss": 0.9909, "step": 4624 }, { "epoch": 0.15, "learning_rate": 1.9205558529766274e-05, "loss": 1.002, "step": 4625 }, { "epoch": 0.15, "learning_rate": 1.920513744406665e-05, "loss": 0.9943, "step": 4626 }, { "epoch": 0.15, "learning_rate": 1.9204716251419165e-05, "loss": 1.0032, "step": 4627 }, { "epoch": 0.15, "learning_rate": 1.9204294951828716e-05, "loss": 0.9811, "step": 4628 }, { "epoch": 0.15, "learning_rate": 1.9203873545300195e-05, "loss": 0.9651, "step": 4629 }, { "epoch": 0.15, "learning_rate": 1.9203452031838497e-05, "loss": 0.9766, "step": 4630 }, { "epoch": 0.15, "learning_rate": 1.920303041144852e-05, "loss": 1.0216, "step": 4631 }, { "epoch": 0.15, "learning_rate": 1.920260868413516e-05, "loss": 1.005, "step": 4632 }, { "epoch": 0.15, "learning_rate": 1.920218684990332e-05, "loss": 0.9825, "step": 4633 }, { "epoch": 0.15, "learning_rate": 1.92017649087579e-05, "loss": 0.9784, "step": 4634 }, { "epoch": 0.15, "learning_rate": 1.9201342860703805e-05, "loss": 0.9888, "step": 4635 }, { "epoch": 0.15, "learning_rate": 1.9200920705745933e-05, "loss": 1.014, "step": 4636 }, { "epoch": 0.15, "learning_rate": 1.9200498443889196e-05, "loss": 1.0333, "step": 4637 }, { "epoch": 0.15, "learning_rate": 1.920007607513849e-05, "loss": 0.9648, "step": 4638 }, { "epoch": 0.15, "learning_rate": 1.9199653599498732e-05, "loss": 0.9671, "step": 4639 }, { "epoch": 0.15, "learning_rate": 1.9199231016974824e-05, "loss": 0.9312, "step": 4640 }, { "epoch": 0.15, "learning_rate": 1.9198808327571678e-05, "loss": 1.0039, "step": 4641 }, { "epoch": 0.15, "learning_rate": 1.9198385531294204e-05, "loss": 0.9855, "step": 4642 }, { "epoch": 0.15, "learning_rate": 1.919796262814732e-05, "loss": 0.9251, "step": 4643 }, { "epoch": 0.15, "learning_rate": 1.919753961813593e-05, "loss": 0.9689, "step": 4644 }, { "epoch": 0.15, "learning_rate": 1.9197116501264954e-05, "loss": 0.9976, "step": 4645 }, { "epoch": 0.15, "learning_rate": 1.9196693277539306e-05, "loss": 0.9343, "step": 4646 }, { "epoch": 0.15, "learning_rate": 1.9196269946963905e-05, "loss": 0.9961, "step": 4647 }, { "epoch": 0.15, "learning_rate": 1.919584650954367e-05, "loss": 0.9741, "step": 4648 }, { "epoch": 0.15, "learning_rate": 1.9195422965283517e-05, "loss": 0.9827, "step": 4649 }, { "epoch": 0.15, "learning_rate": 1.9194999314188368e-05, "loss": 1.0236, "step": 4650 }, { "epoch": 0.15, "learning_rate": 1.9194575556263148e-05, "loss": 0.9985, "step": 4651 }, { "epoch": 0.15, "learning_rate": 1.9194151691512778e-05, "loss": 0.9666, "step": 4652 }, { "epoch": 0.15, "learning_rate": 1.9193727719942185e-05, "loss": 0.9997, "step": 4653 }, { "epoch": 0.15, "learning_rate": 1.919330364155629e-05, "loss": 1.01, "step": 4654 }, { "epoch": 0.15, "learning_rate": 1.9192879456360023e-05, "loss": 0.9546, "step": 4655 }, { "epoch": 0.15, "learning_rate": 1.9192455164358314e-05, "loss": 0.9681, "step": 4656 }, { "epoch": 0.15, "learning_rate": 1.919203076555609e-05, "loss": 0.9843, "step": 4657 }, { "epoch": 0.16, "learning_rate": 1.9191606259958278e-05, "loss": 0.9957, "step": 4658 }, { "epoch": 0.16, "learning_rate": 1.9191181647569822e-05, "loss": 1.0099, "step": 4659 }, { "epoch": 0.16, "learning_rate": 1.9190756928395645e-05, "loss": 0.9776, "step": 4660 }, { "epoch": 0.16, "learning_rate": 1.9190332102440682e-05, "loss": 1.0018, "step": 4661 }, { "epoch": 0.16, "learning_rate": 1.918990716970987e-05, "loss": 1.0047, "step": 4662 }, { "epoch": 0.16, "learning_rate": 1.918948213020815e-05, "loss": 1.0041, "step": 4663 }, { "epoch": 0.16, "learning_rate": 1.9189056983940457e-05, "loss": 0.9445, "step": 4664 }, { "epoch": 0.16, "learning_rate": 1.918863173091173e-05, "loss": 1.0048, "step": 4665 }, { "epoch": 0.16, "learning_rate": 1.9188206371126913e-05, "loss": 0.9598, "step": 4666 }, { "epoch": 0.16, "learning_rate": 1.918778090459094e-05, "loss": 0.9704, "step": 4667 }, { "epoch": 0.16, "learning_rate": 1.9187355331308764e-05, "loss": 0.9931, "step": 4668 }, { "epoch": 0.16, "learning_rate": 1.9186929651285324e-05, "loss": 0.9815, "step": 4669 }, { "epoch": 0.16, "learning_rate": 1.9186503864525565e-05, "loss": 0.9918, "step": 4670 }, { "epoch": 0.16, "learning_rate": 1.9186077971034437e-05, "loss": 0.9763, "step": 4671 }, { "epoch": 0.16, "learning_rate": 1.9185651970816886e-05, "loss": 0.9464, "step": 4672 }, { "epoch": 0.16, "learning_rate": 1.9185225863877863e-05, "loss": 0.9986, "step": 4673 }, { "epoch": 0.16, "learning_rate": 1.9184799650222318e-05, "loss": 1.0105, "step": 4674 }, { "epoch": 0.16, "learning_rate": 1.91843733298552e-05, "loss": 1.0263, "step": 4675 }, { "epoch": 0.16, "learning_rate": 1.9183946902781468e-05, "loss": 0.9441, "step": 4676 }, { "epoch": 0.16, "learning_rate": 1.9183520369006073e-05, "loss": 1.0053, "step": 4677 }, { "epoch": 0.16, "learning_rate": 1.9183093728533966e-05, "loss": 0.9872, "step": 4678 }, { "epoch": 0.16, "learning_rate": 1.918266698137011e-05, "loss": 1.0008, "step": 4679 }, { "epoch": 0.16, "learning_rate": 1.9182240127519465e-05, "loss": 0.9502, "step": 4680 }, { "epoch": 0.16, "learning_rate": 1.9181813166986988e-05, "loss": 0.9891, "step": 4681 }, { "epoch": 0.16, "learning_rate": 1.9181386099777636e-05, "loss": 0.9752, "step": 4682 }, { "epoch": 0.16, "learning_rate": 1.918095892589637e-05, "loss": 1.0076, "step": 4683 }, { "epoch": 0.16, "learning_rate": 1.9180531645348157e-05, "loss": 0.9528, "step": 4684 }, { "epoch": 0.16, "learning_rate": 1.918010425813796e-05, "loss": 0.9816, "step": 4685 }, { "epoch": 0.16, "learning_rate": 1.917967676427075e-05, "loss": 0.9714, "step": 4686 }, { "epoch": 0.16, "learning_rate": 1.9179249163751484e-05, "loss": 0.9805, "step": 4687 }, { "epoch": 0.16, "learning_rate": 1.917882145658514e-05, "loss": 0.9951, "step": 4688 }, { "epoch": 0.16, "learning_rate": 1.9178393642776675e-05, "loss": 1.0034, "step": 4689 }, { "epoch": 0.16, "learning_rate": 1.9177965722331072e-05, "loss": 0.9577, "step": 4690 }, { "epoch": 0.16, "learning_rate": 1.9177537695253295e-05, "loss": 0.9671, "step": 4691 }, { "epoch": 0.16, "learning_rate": 1.917710956154832e-05, "loss": 0.965, "step": 4692 }, { "epoch": 0.16, "learning_rate": 1.9176681321221115e-05, "loss": 0.9522, "step": 4693 }, { "epoch": 0.16, "learning_rate": 1.917625297427667e-05, "loss": 0.9768, "step": 4694 }, { "epoch": 0.16, "learning_rate": 1.9175824520719944e-05, "loss": 0.9826, "step": 4695 }, { "epoch": 0.16, "learning_rate": 1.9175395960555927e-05, "loss": 0.9924, "step": 4696 }, { "epoch": 0.16, "learning_rate": 1.917496729378959e-05, "loss": 0.9799, "step": 4697 }, { "epoch": 0.16, "learning_rate": 1.917453852042592e-05, "loss": 1.0265, "step": 4698 }, { "epoch": 0.16, "learning_rate": 1.91741096404699e-05, "loss": 0.9513, "step": 4699 }, { "epoch": 0.16, "learning_rate": 1.917368065392651e-05, "loss": 0.9657, "step": 4700 }, { "epoch": 0.16, "learning_rate": 1.9173251560800725e-05, "loss": 0.9726, "step": 4701 }, { "epoch": 0.16, "learning_rate": 1.9172822361097542e-05, "loss": 0.9938, "step": 4702 }, { "epoch": 0.16, "learning_rate": 1.9172393054821948e-05, "loss": 1.0008, "step": 4703 }, { "epoch": 0.16, "learning_rate": 1.9171963641978923e-05, "loss": 0.928, "step": 4704 }, { "epoch": 0.16, "learning_rate": 1.917153412257346e-05, "loss": 0.9733, "step": 4705 }, { "epoch": 0.16, "learning_rate": 1.917110449661055e-05, "loss": 0.9484, "step": 4706 }, { "epoch": 0.16, "learning_rate": 1.9170674764095186e-05, "loss": 0.9598, "step": 4707 }, { "epoch": 0.16, "learning_rate": 1.9170244925032354e-05, "loss": 0.9758, "step": 4708 }, { "epoch": 0.16, "learning_rate": 1.9169814979427055e-05, "loss": 0.9667, "step": 4709 }, { "epoch": 0.16, "learning_rate": 1.916938492728428e-05, "loss": 0.9414, "step": 4710 }, { "epoch": 0.16, "learning_rate": 1.9168954768609033e-05, "loss": 0.9771, "step": 4711 }, { "epoch": 0.16, "learning_rate": 1.9168524503406297e-05, "loss": 0.9724, "step": 4712 }, { "epoch": 0.16, "learning_rate": 1.9168094131681084e-05, "loss": 0.9765, "step": 4713 }, { "epoch": 0.16, "learning_rate": 1.9167663653438392e-05, "loss": 1.0029, "step": 4714 }, { "epoch": 0.16, "learning_rate": 1.9167233068683217e-05, "loss": 1.0509, "step": 4715 }, { "epoch": 0.16, "learning_rate": 1.9166802377420568e-05, "loss": 0.9529, "step": 4716 }, { "epoch": 0.16, "learning_rate": 1.9166371579655445e-05, "loss": 1.0002, "step": 4717 }, { "epoch": 0.16, "learning_rate": 1.9165940675392852e-05, "loss": 1.0094, "step": 4718 }, { "epoch": 0.16, "learning_rate": 1.9165509664637798e-05, "loss": 0.9688, "step": 4719 }, { "epoch": 0.16, "learning_rate": 1.9165078547395293e-05, "loss": 0.9351, "step": 4720 }, { "epoch": 0.16, "learning_rate": 1.916464732367034e-05, "loss": 0.9837, "step": 4721 }, { "epoch": 0.16, "learning_rate": 1.9164215993467952e-05, "loss": 1.0208, "step": 4722 }, { "epoch": 0.16, "learning_rate": 1.9163784556793144e-05, "loss": 1.0053, "step": 4723 }, { "epoch": 0.16, "learning_rate": 1.9163353013650917e-05, "loss": 0.9364, "step": 4724 }, { "epoch": 0.16, "learning_rate": 1.9162921364046297e-05, "loss": 0.9971, "step": 4725 }, { "epoch": 0.16, "learning_rate": 1.91624896079843e-05, "loss": 0.9679, "step": 4726 }, { "epoch": 0.16, "learning_rate": 1.9162057745469932e-05, "loss": 1.0145, "step": 4727 }, { "epoch": 0.16, "learning_rate": 1.9161625776508216e-05, "loss": 0.9389, "step": 4728 }, { "epoch": 0.16, "learning_rate": 1.916119370110417e-05, "loss": 0.9626, "step": 4729 }, { "epoch": 0.16, "learning_rate": 1.916076151926281e-05, "loss": 0.9782, "step": 4730 }, { "epoch": 0.16, "learning_rate": 1.9160329230989164e-05, "loss": 0.963, "step": 4731 }, { "epoch": 0.16, "learning_rate": 1.9159896836288255e-05, "loss": 0.9672, "step": 4732 }, { "epoch": 0.16, "learning_rate": 1.91594643351651e-05, "loss": 1.0199, "step": 4733 }, { "epoch": 0.16, "learning_rate": 1.915903172762473e-05, "loss": 0.9441, "step": 4734 }, { "epoch": 0.16, "learning_rate": 1.9158599013672166e-05, "loss": 0.9693, "step": 4735 }, { "epoch": 0.16, "learning_rate": 1.9158166193312437e-05, "loss": 0.9855, "step": 4736 }, { "epoch": 0.16, "learning_rate": 1.9157733266550577e-05, "loss": 0.987, "step": 4737 }, { "epoch": 0.16, "learning_rate": 1.9157300233391608e-05, "loss": 0.9424, "step": 4738 }, { "epoch": 0.16, "learning_rate": 1.9156867093840567e-05, "loss": 1.0092, "step": 4739 }, { "epoch": 0.16, "learning_rate": 1.915643384790248e-05, "loss": 0.9643, "step": 4740 }, { "epoch": 0.16, "learning_rate": 1.9156000495582387e-05, "loss": 0.9445, "step": 4741 }, { "epoch": 0.16, "learning_rate": 1.915556703688532e-05, "loss": 0.9678, "step": 4742 }, { "epoch": 0.16, "learning_rate": 1.9155133471816314e-05, "loss": 0.9766, "step": 4743 }, { "epoch": 0.16, "learning_rate": 1.915469980038041e-05, "loss": 0.9515, "step": 4744 }, { "epoch": 0.16, "learning_rate": 1.915426602258264e-05, "loss": 1.0443, "step": 4745 }, { "epoch": 0.16, "learning_rate": 1.9153832138428054e-05, "loss": 0.9489, "step": 4746 }, { "epoch": 0.16, "learning_rate": 1.915339814792168e-05, "loss": 1.0282, "step": 4747 }, { "epoch": 0.16, "learning_rate": 1.915296405106857e-05, "loss": 0.9813, "step": 4748 }, { "epoch": 0.16, "learning_rate": 1.9152529847873766e-05, "loss": 0.9794, "step": 4749 }, { "epoch": 0.16, "learning_rate": 1.915209553834231e-05, "loss": 0.9696, "step": 4750 }, { "epoch": 0.16, "learning_rate": 1.915166112247925e-05, "loss": 0.9659, "step": 4751 }, { "epoch": 0.16, "learning_rate": 1.915122660028963e-05, "loss": 0.9804, "step": 4752 }, { "epoch": 0.16, "learning_rate": 1.9150791971778503e-05, "loss": 0.981, "step": 4753 }, { "epoch": 0.16, "learning_rate": 1.9150357236950916e-05, "loss": 0.971, "step": 4754 }, { "epoch": 0.16, "learning_rate": 1.9149922395811918e-05, "loss": 0.9997, "step": 4755 }, { "epoch": 0.16, "learning_rate": 1.9149487448366565e-05, "loss": 0.9818, "step": 4756 }, { "epoch": 0.16, "learning_rate": 1.9149052394619905e-05, "loss": 0.9832, "step": 4757 }, { "epoch": 0.16, "learning_rate": 1.9148617234577004e-05, "loss": 0.9796, "step": 4758 }, { "epoch": 0.16, "learning_rate": 1.9148181968242903e-05, "loss": 1.017, "step": 4759 }, { "epoch": 0.16, "learning_rate": 1.9147746595622668e-05, "loss": 0.9486, "step": 4760 }, { "epoch": 0.16, "learning_rate": 1.9147311116721356e-05, "loss": 1.0173, "step": 4761 }, { "epoch": 0.16, "learning_rate": 1.9146875531544026e-05, "loss": 0.9694, "step": 4762 }, { "epoch": 0.16, "learning_rate": 1.9146439840095737e-05, "loss": 1.0023, "step": 4763 }, { "epoch": 0.16, "learning_rate": 1.9146004042381555e-05, "loss": 0.9936, "step": 4764 }, { "epoch": 0.16, "learning_rate": 1.914556813840654e-05, "loss": 0.9916, "step": 4765 }, { "epoch": 0.16, "learning_rate": 1.9145132128175755e-05, "loss": 0.9633, "step": 4766 }, { "epoch": 0.16, "learning_rate": 1.914469601169427e-05, "loss": 0.9832, "step": 4767 }, { "epoch": 0.16, "learning_rate": 1.914425978896715e-05, "loss": 0.9911, "step": 4768 }, { "epoch": 0.16, "learning_rate": 1.9143823459999464e-05, "loss": 0.9861, "step": 4769 }, { "epoch": 0.16, "learning_rate": 1.9143387024796278e-05, "loss": 1.0034, "step": 4770 }, { "epoch": 0.16, "learning_rate": 1.9142950483362668e-05, "loss": 0.9645, "step": 4771 }, { "epoch": 0.16, "learning_rate": 1.91425138357037e-05, "loss": 0.9791, "step": 4772 }, { "epoch": 0.16, "learning_rate": 1.914207708182445e-05, "loss": 0.9959, "step": 4773 }, { "epoch": 0.16, "learning_rate": 1.9141640221729996e-05, "loss": 0.9671, "step": 4774 }, { "epoch": 0.16, "learning_rate": 1.9141203255425408e-05, "loss": 0.963, "step": 4775 }, { "epoch": 0.16, "learning_rate": 1.914076618291576e-05, "loss": 0.9574, "step": 4776 }, { "epoch": 0.16, "learning_rate": 1.9140329004206138e-05, "loss": 0.9922, "step": 4777 }, { "epoch": 0.16, "learning_rate": 1.913989171930162e-05, "loss": 0.9534, "step": 4778 }, { "epoch": 0.16, "learning_rate": 1.9139454328207282e-05, "loss": 0.9664, "step": 4779 }, { "epoch": 0.16, "learning_rate": 1.9139016830928213e-05, "loss": 0.9938, "step": 4780 }, { "epoch": 0.16, "learning_rate": 1.9138579227469484e-05, "loss": 1.0269, "step": 4781 }, { "epoch": 0.16, "learning_rate": 1.9138141517836192e-05, "loss": 0.9944, "step": 4782 }, { "epoch": 0.16, "learning_rate": 1.9137703702033415e-05, "loss": 1.0043, "step": 4783 }, { "epoch": 0.16, "learning_rate": 1.913726578006624e-05, "loss": 0.9236, "step": 4784 }, { "epoch": 0.16, "learning_rate": 1.913682775193976e-05, "loss": 0.9692, "step": 4785 }, { "epoch": 0.16, "learning_rate": 1.9136389617659056e-05, "loss": 0.9856, "step": 4786 }, { "epoch": 0.16, "learning_rate": 1.9135951377229224e-05, "loss": 0.9732, "step": 4787 }, { "epoch": 0.16, "learning_rate": 1.9135513030655358e-05, "loss": 0.9466, "step": 4788 }, { "epoch": 0.16, "learning_rate": 1.9135074577942545e-05, "loss": 0.9502, "step": 4789 }, { "epoch": 0.16, "learning_rate": 1.9134636019095883e-05, "loss": 1.012, "step": 4790 }, { "epoch": 0.16, "learning_rate": 1.9134197354120463e-05, "loss": 0.9647, "step": 4791 }, { "epoch": 0.16, "learning_rate": 1.913375858302138e-05, "loss": 0.9485, "step": 4792 }, { "epoch": 0.16, "learning_rate": 1.9133319705803743e-05, "loss": 1.0023, "step": 4793 }, { "epoch": 0.16, "learning_rate": 1.9132880722472646e-05, "loss": 0.9652, "step": 4794 }, { "epoch": 0.16, "learning_rate": 1.9132441633033182e-05, "loss": 1.032, "step": 4795 }, { "epoch": 0.16, "learning_rate": 1.913200243749046e-05, "loss": 0.9434, "step": 4796 }, { "epoch": 0.16, "learning_rate": 1.9131563135849577e-05, "loss": 0.9975, "step": 4797 }, { "epoch": 0.16, "learning_rate": 1.9131123728115644e-05, "loss": 0.9657, "step": 4798 }, { "epoch": 0.16, "learning_rate": 1.913068421429376e-05, "loss": 0.988, "step": 4799 }, { "epoch": 0.16, "learning_rate": 1.9130244594389032e-05, "loss": 0.9446, "step": 4800 }, { "epoch": 0.16, "learning_rate": 1.9129804868406575e-05, "loss": 1.0199, "step": 4801 }, { "epoch": 0.16, "learning_rate": 1.912936503635149e-05, "loss": 0.9407, "step": 4802 }, { "epoch": 0.16, "learning_rate": 1.9128925098228887e-05, "loss": 1.017, "step": 4803 }, { "epoch": 0.16, "learning_rate": 1.912848505404388e-05, "loss": 0.9909, "step": 4804 }, { "epoch": 0.16, "learning_rate": 1.9128044903801588e-05, "loss": 1.0371, "step": 4805 }, { "epoch": 0.16, "learning_rate": 1.912760464750711e-05, "loss": 0.9733, "step": 4806 }, { "epoch": 0.16, "learning_rate": 1.9127164285165575e-05, "loss": 1.0048, "step": 4807 }, { "epoch": 0.16, "learning_rate": 1.9126723816782088e-05, "loss": 0.9752, "step": 4808 }, { "epoch": 0.16, "learning_rate": 1.9126283242361775e-05, "loss": 0.9939, "step": 4809 }, { "epoch": 0.16, "learning_rate": 1.9125842561909752e-05, "loss": 1.0067, "step": 4810 }, { "epoch": 0.16, "learning_rate": 1.9125401775431136e-05, "loss": 0.9983, "step": 4811 }, { "epoch": 0.16, "learning_rate": 1.9124960882931055e-05, "loss": 0.9267, "step": 4812 }, { "epoch": 0.16, "learning_rate": 1.9124519884414624e-05, "loss": 0.9936, "step": 4813 }, { "epoch": 0.16, "learning_rate": 1.912407877988697e-05, "loss": 0.999, "step": 4814 }, { "epoch": 0.16, "learning_rate": 1.9123637569353218e-05, "loss": 1.0016, "step": 4815 }, { "epoch": 0.16, "learning_rate": 1.9123196252818494e-05, "loss": 0.9714, "step": 4816 }, { "epoch": 0.16, "learning_rate": 1.9122754830287923e-05, "loss": 1.0008, "step": 4817 }, { "epoch": 0.16, "learning_rate": 1.9122313301766638e-05, "loss": 0.9631, "step": 4818 }, { "epoch": 0.16, "learning_rate": 1.9121871667259767e-05, "loss": 0.9519, "step": 4819 }, { "epoch": 0.16, "learning_rate": 1.9121429926772438e-05, "loss": 0.9581, "step": 4820 }, { "epoch": 0.16, "learning_rate": 1.9120988080309785e-05, "loss": 0.9615, "step": 4821 }, { "epoch": 0.16, "learning_rate": 1.9120546127876944e-05, "loss": 0.965, "step": 4822 }, { "epoch": 0.16, "learning_rate": 1.9120104069479047e-05, "loss": 0.9847, "step": 4823 }, { "epoch": 0.16, "learning_rate": 1.9119661905121236e-05, "loss": 0.9769, "step": 4824 }, { "epoch": 0.16, "learning_rate": 1.9119219634808636e-05, "loss": 0.9565, "step": 4825 }, { "epoch": 0.16, "learning_rate": 1.9118777258546396e-05, "loss": 1.0218, "step": 4826 }, { "epoch": 0.16, "learning_rate": 1.9118334776339655e-05, "loss": 0.9805, "step": 4827 }, { "epoch": 0.16, "learning_rate": 1.9117892188193547e-05, "loss": 0.9807, "step": 4828 }, { "epoch": 0.16, "learning_rate": 1.911744949411322e-05, "loss": 0.9627, "step": 4829 }, { "epoch": 0.16, "learning_rate": 1.911700669410381e-05, "loss": 0.9452, "step": 4830 }, { "epoch": 0.16, "learning_rate": 1.9116563788170474e-05, "loss": 1.023, "step": 4831 }, { "epoch": 0.16, "learning_rate": 1.9116120776318354e-05, "loss": 0.9113, "step": 4832 }, { "epoch": 0.16, "learning_rate": 1.9115677658552586e-05, "loss": 0.9701, "step": 4833 }, { "epoch": 0.16, "learning_rate": 1.9115234434878325e-05, "loss": 0.9658, "step": 4834 }, { "epoch": 0.16, "learning_rate": 1.9114791105300728e-05, "loss": 0.991, "step": 4835 }, { "epoch": 0.16, "learning_rate": 1.9114347669824937e-05, "loss": 0.973, "step": 4836 }, { "epoch": 0.16, "learning_rate": 1.9113904128456105e-05, "loss": 0.9779, "step": 4837 }, { "epoch": 0.16, "learning_rate": 1.911346048119939e-05, "loss": 0.9141, "step": 4838 }, { "epoch": 0.16, "learning_rate": 1.911301672805994e-05, "loss": 0.9947, "step": 4839 }, { "epoch": 0.16, "learning_rate": 1.9112572869042916e-05, "loss": 0.9375, "step": 4840 }, { "epoch": 0.16, "learning_rate": 1.911212890415347e-05, "loss": 1.0089, "step": 4841 }, { "epoch": 0.16, "learning_rate": 1.9111684833396765e-05, "loss": 0.9765, "step": 4842 }, { "epoch": 0.16, "learning_rate": 1.9111240656777957e-05, "loss": 0.9819, "step": 4843 }, { "epoch": 0.16, "learning_rate": 1.911079637430221e-05, "loss": 0.9869, "step": 4844 }, { "epoch": 0.16, "learning_rate": 1.9110351985974677e-05, "loss": 0.9604, "step": 4845 }, { "epoch": 0.16, "learning_rate": 1.9109907491800533e-05, "loss": 0.9648, "step": 4846 }, { "epoch": 0.16, "learning_rate": 1.9109462891784933e-05, "loss": 1.0168, "step": 4847 }, { "epoch": 0.16, "learning_rate": 1.9109018185933048e-05, "loss": 0.9729, "step": 4848 }, { "epoch": 0.16, "learning_rate": 1.910857337425004e-05, "loss": 1.0141, "step": 4849 }, { "epoch": 0.16, "learning_rate": 1.9108128456741084e-05, "loss": 1.0059, "step": 4850 }, { "epoch": 0.16, "learning_rate": 1.9107683433411343e-05, "loss": 0.9793, "step": 4851 }, { "epoch": 0.16, "learning_rate": 1.910723830426599e-05, "loss": 0.9906, "step": 4852 }, { "epoch": 0.16, "learning_rate": 1.9106793069310194e-05, "loss": 1.0169, "step": 4853 }, { "epoch": 0.16, "learning_rate": 1.9106347728549134e-05, "loss": 0.9631, "step": 4854 }, { "epoch": 0.16, "learning_rate": 1.910590228198798e-05, "loss": 1.0598, "step": 4855 }, { "epoch": 0.16, "learning_rate": 1.9105456729631902e-05, "loss": 0.9466, "step": 4856 }, { "epoch": 0.16, "learning_rate": 1.9105011071486085e-05, "loss": 1.046, "step": 4857 }, { "epoch": 0.16, "learning_rate": 1.9104565307555702e-05, "loss": 0.9679, "step": 4858 }, { "epoch": 0.16, "learning_rate": 1.9104119437845938e-05, "loss": 0.9562, "step": 4859 }, { "epoch": 0.16, "learning_rate": 1.910367346236196e-05, "loss": 1.0259, "step": 4860 }, { "epoch": 0.16, "learning_rate": 1.9103227381108967e-05, "loss": 1.0104, "step": 4861 }, { "epoch": 0.16, "learning_rate": 1.910278119409213e-05, "loss": 0.9675, "step": 4862 }, { "epoch": 0.16, "learning_rate": 1.9102334901316634e-05, "loss": 1.0029, "step": 4863 }, { "epoch": 0.16, "learning_rate": 1.910188850278767e-05, "loss": 0.9738, "step": 4864 }, { "epoch": 0.16, "learning_rate": 1.9101441998510416e-05, "loss": 0.9598, "step": 4865 }, { "epoch": 0.16, "learning_rate": 1.9100995388490067e-05, "loss": 1.0018, "step": 4866 }, { "epoch": 0.16, "learning_rate": 1.9100548672731808e-05, "loss": 1.0114, "step": 4867 }, { "epoch": 0.16, "learning_rate": 1.910010185124083e-05, "loss": 0.9314, "step": 4868 }, { "epoch": 0.16, "learning_rate": 1.9099654924022327e-05, "loss": 1.0201, "step": 4869 }, { "epoch": 0.16, "learning_rate": 1.9099207891081484e-05, "loss": 0.9738, "step": 4870 }, { "epoch": 0.16, "learning_rate": 1.9098760752423505e-05, "loss": 1.034, "step": 4871 }, { "epoch": 0.16, "learning_rate": 1.9098313508053576e-05, "loss": 0.9949, "step": 4872 }, { "epoch": 0.16, "learning_rate": 1.9097866157976895e-05, "loss": 0.9769, "step": 4873 }, { "epoch": 0.16, "learning_rate": 1.9097418702198665e-05, "loss": 1.0157, "step": 4874 }, { "epoch": 0.16, "learning_rate": 1.909697114072408e-05, "loss": 0.9575, "step": 4875 }, { "epoch": 0.16, "learning_rate": 1.909652347355834e-05, "loss": 0.9705, "step": 4876 }, { "epoch": 0.16, "learning_rate": 1.9096075700706647e-05, "loss": 0.9217, "step": 4877 }, { "epoch": 0.16, "learning_rate": 1.9095627822174202e-05, "loss": 0.9807, "step": 4878 }, { "epoch": 0.16, "learning_rate": 1.9095179837966214e-05, "loss": 1.0123, "step": 4879 }, { "epoch": 0.16, "learning_rate": 1.9094731748087878e-05, "loss": 0.9636, "step": 4880 }, { "epoch": 0.16, "learning_rate": 1.9094283552544407e-05, "loss": 0.9811, "step": 4881 }, { "epoch": 0.16, "learning_rate": 1.909383525134101e-05, "loss": 0.9418, "step": 4882 }, { "epoch": 0.16, "learning_rate": 1.9093386844482894e-05, "loss": 0.9882, "step": 4883 }, { "epoch": 0.16, "learning_rate": 1.909293833197526e-05, "loss": 0.9379, "step": 4884 }, { "epoch": 0.16, "learning_rate": 1.9092489713823332e-05, "loss": 1.0167, "step": 4885 }, { "epoch": 0.16, "learning_rate": 1.9092040990032314e-05, "loss": 1.0216, "step": 4886 }, { "epoch": 0.16, "learning_rate": 1.909159216060742e-05, "loss": 0.968, "step": 4887 }, { "epoch": 0.16, "learning_rate": 1.9091143225553868e-05, "loss": 0.9524, "step": 4888 }, { "epoch": 0.16, "learning_rate": 1.9090694184876874e-05, "loss": 0.967, "step": 4889 }, { "epoch": 0.16, "learning_rate": 1.909024503858165e-05, "loss": 0.9671, "step": 4890 }, { "epoch": 0.16, "learning_rate": 1.908979578667342e-05, "loss": 0.9668, "step": 4891 }, { "epoch": 0.16, "learning_rate": 1.9089346429157397e-05, "loss": 0.9866, "step": 4892 }, { "epoch": 0.16, "learning_rate": 1.908889696603881e-05, "loss": 1.0162, "step": 4893 }, { "epoch": 0.16, "learning_rate": 1.9088447397322875e-05, "loss": 0.9703, "step": 4894 }, { "epoch": 0.16, "learning_rate": 1.9087997723014817e-05, "loss": 0.9929, "step": 4895 }, { "epoch": 0.16, "learning_rate": 1.908754794311986e-05, "loss": 0.9726, "step": 4896 }, { "epoch": 0.16, "learning_rate": 1.9087098057643235e-05, "loss": 1.0197, "step": 4897 }, { "epoch": 0.16, "learning_rate": 1.908664806659016e-05, "loss": 0.9863, "step": 4898 }, { "epoch": 0.16, "learning_rate": 1.9086197969965867e-05, "loss": 0.9878, "step": 4899 }, { "epoch": 0.16, "learning_rate": 1.9085747767775584e-05, "loss": 0.9902, "step": 4900 }, { "epoch": 0.16, "learning_rate": 1.9085297460024543e-05, "loss": 1.0177, "step": 4901 }, { "epoch": 0.16, "learning_rate": 1.908484704671798e-05, "loss": 0.9434, "step": 4902 }, { "epoch": 0.16, "learning_rate": 1.908439652786112e-05, "loss": 0.9932, "step": 4903 }, { "epoch": 0.16, "learning_rate": 1.90839459034592e-05, "loss": 0.9518, "step": 4904 }, { "epoch": 0.16, "learning_rate": 1.9083495173517465e-05, "loss": 0.9964, "step": 4905 }, { "epoch": 0.16, "learning_rate": 1.9083044338041136e-05, "loss": 0.995, "step": 4906 }, { "epoch": 0.16, "learning_rate": 1.908259339703546e-05, "loss": 0.9794, "step": 4907 }, { "epoch": 0.16, "learning_rate": 1.9082142350505672e-05, "loss": 0.9752, "step": 4908 }, { "epoch": 0.16, "learning_rate": 1.9081691198457017e-05, "loss": 1.0175, "step": 4909 }, { "epoch": 0.16, "learning_rate": 1.9081239940894734e-05, "loss": 0.933, "step": 4910 }, { "epoch": 0.16, "learning_rate": 1.9080788577824066e-05, "loss": 0.9998, "step": 4911 }, { "epoch": 0.16, "learning_rate": 1.9080337109250257e-05, "loss": 0.9615, "step": 4912 }, { "epoch": 0.16, "learning_rate": 1.9079885535178556e-05, "loss": 0.9433, "step": 4913 }, { "epoch": 0.16, "learning_rate": 1.9079433855614203e-05, "loss": 0.9848, "step": 4914 }, { "epoch": 0.16, "learning_rate": 1.9078982070562448e-05, "loss": 0.9463, "step": 4915 }, { "epoch": 0.16, "learning_rate": 1.9078530180028544e-05, "loss": 0.9793, "step": 4916 }, { "epoch": 0.16, "learning_rate": 1.9078078184017738e-05, "loss": 0.9692, "step": 4917 }, { "epoch": 0.16, "learning_rate": 1.9077626082535276e-05, "loss": 0.9569, "step": 4918 }, { "epoch": 0.16, "learning_rate": 1.907717387558642e-05, "loss": 0.9821, "step": 4919 }, { "epoch": 0.16, "learning_rate": 1.907672156317642e-05, "loss": 0.9561, "step": 4920 }, { "epoch": 0.16, "learning_rate": 1.9076269145310532e-05, "loss": 0.9721, "step": 4921 }, { "epoch": 0.16, "learning_rate": 1.9075816621994008e-05, "loss": 0.9758, "step": 4922 }, { "epoch": 0.16, "learning_rate": 1.907536399323211e-05, "loss": 0.9644, "step": 4923 }, { "epoch": 0.16, "learning_rate": 1.9074911259030098e-05, "loss": 0.9688, "step": 4924 }, { "epoch": 0.16, "learning_rate": 1.907445841939323e-05, "loss": 0.9785, "step": 4925 }, { "epoch": 0.16, "learning_rate": 1.9074005474326762e-05, "loss": 0.9932, "step": 4926 }, { "epoch": 0.16, "learning_rate": 1.9073552423835963e-05, "loss": 0.98, "step": 4927 }, { "epoch": 0.16, "learning_rate": 1.90730992679261e-05, "loss": 0.9728, "step": 4928 }, { "epoch": 0.16, "learning_rate": 1.9072646006602425e-05, "loss": 1.0305, "step": 4929 }, { "epoch": 0.16, "learning_rate": 1.9072192639870218e-05, "loss": 0.9451, "step": 4930 }, { "epoch": 0.16, "learning_rate": 1.907173916773474e-05, "loss": 1.0144, "step": 4931 }, { "epoch": 0.16, "learning_rate": 1.9071285590201255e-05, "loss": 1.0156, "step": 4932 }, { "epoch": 0.16, "learning_rate": 1.9070831907275044e-05, "loss": 1.0076, "step": 4933 }, { "epoch": 0.16, "learning_rate": 1.9070378118961365e-05, "loss": 0.954, "step": 4934 }, { "epoch": 0.16, "learning_rate": 1.9069924225265503e-05, "loss": 0.9703, "step": 4935 }, { "epoch": 0.16, "learning_rate": 1.9069470226192724e-05, "loss": 0.991, "step": 4936 }, { "epoch": 0.16, "learning_rate": 1.90690161217483e-05, "loss": 0.9909, "step": 4937 }, { "epoch": 0.16, "learning_rate": 1.9068561911937517e-05, "loss": 0.9721, "step": 4938 }, { "epoch": 0.16, "learning_rate": 1.9068107596765642e-05, "loss": 0.9717, "step": 4939 }, { "epoch": 0.16, "learning_rate": 1.906765317623796e-05, "loss": 0.9187, "step": 4940 }, { "epoch": 0.16, "learning_rate": 1.9067198650359744e-05, "loss": 1.0195, "step": 4941 }, { "epoch": 0.16, "learning_rate": 1.9066744019136282e-05, "loss": 0.9368, "step": 4942 }, { "epoch": 0.16, "learning_rate": 1.9066289282572855e-05, "loss": 0.9834, "step": 4943 }, { "epoch": 0.16, "learning_rate": 1.906583444067474e-05, "loss": 1.0228, "step": 4944 }, { "epoch": 0.16, "learning_rate": 1.9065379493447227e-05, "loss": 0.9787, "step": 4945 }, { "epoch": 0.16, "learning_rate": 1.90649244408956e-05, "loss": 0.9575, "step": 4946 }, { "epoch": 0.16, "learning_rate": 1.9064469283025148e-05, "loss": 0.9653, "step": 4947 }, { "epoch": 0.16, "learning_rate": 1.9064014019841155e-05, "loss": 0.9777, "step": 4948 }, { "epoch": 0.16, "learning_rate": 1.9063558651348916e-05, "loss": 0.9938, "step": 4949 }, { "epoch": 0.16, "learning_rate": 1.9063103177553714e-05, "loss": 0.982, "step": 4950 }, { "epoch": 0.16, "learning_rate": 1.906264759846085e-05, "loss": 1.0087, "step": 4951 }, { "epoch": 0.16, "learning_rate": 1.9062191914075606e-05, "loss": 0.963, "step": 4952 }, { "epoch": 0.16, "learning_rate": 1.906173612440329e-05, "loss": 0.9881, "step": 4953 }, { "epoch": 0.16, "learning_rate": 1.9061280229449186e-05, "loss": 0.9987, "step": 4954 }, { "epoch": 0.16, "learning_rate": 1.9060824229218596e-05, "loss": 1.0063, "step": 4955 }, { "epoch": 0.16, "learning_rate": 1.9060368123716816e-05, "loss": 0.9285, "step": 4956 }, { "epoch": 0.16, "learning_rate": 1.9059911912949146e-05, "loss": 0.9331, "step": 4957 }, { "epoch": 0.17, "learning_rate": 1.9059455596920885e-05, "loss": 0.9623, "step": 4958 }, { "epoch": 0.17, "learning_rate": 1.905899917563734e-05, "loss": 0.9941, "step": 4959 }, { "epoch": 0.17, "learning_rate": 1.9058542649103803e-05, "loss": 0.9593, "step": 4960 }, { "epoch": 0.17, "learning_rate": 1.905808601732559e-05, "loss": 0.9963, "step": 4961 }, { "epoch": 0.17, "learning_rate": 1.9057629280307998e-05, "loss": 0.9957, "step": 4962 }, { "epoch": 0.17, "learning_rate": 1.9057172438056337e-05, "loss": 1.0014, "step": 4963 }, { "epoch": 0.17, "learning_rate": 1.9056715490575917e-05, "loss": 0.9719, "step": 4964 }, { "epoch": 0.17, "learning_rate": 1.9056258437872043e-05, "loss": 0.9543, "step": 4965 }, { "epoch": 0.17, "learning_rate": 1.9055801279950024e-05, "loss": 0.973, "step": 4966 }, { "epoch": 0.17, "learning_rate": 1.9055344016815177e-05, "loss": 1.0212, "step": 4967 }, { "epoch": 0.17, "learning_rate": 1.9054886648472806e-05, "loss": 0.9682, "step": 4968 }, { "epoch": 0.17, "learning_rate": 1.9054429174928236e-05, "loss": 1.0037, "step": 4969 }, { "epoch": 0.17, "learning_rate": 1.9053971596186774e-05, "loss": 0.9733, "step": 4970 }, { "epoch": 0.17, "learning_rate": 1.9053513912253737e-05, "loss": 0.988, "step": 4971 }, { "epoch": 0.17, "learning_rate": 1.905305612313445e-05, "loss": 0.9474, "step": 4972 }, { "epoch": 0.17, "learning_rate": 1.9052598228834217e-05, "loss": 1.0326, "step": 4973 }, { "epoch": 0.17, "learning_rate": 1.9052140229358375e-05, "loss": 0.967, "step": 4974 }, { "epoch": 0.17, "learning_rate": 1.905168212471223e-05, "loss": 0.9439, "step": 4975 }, { "epoch": 0.17, "learning_rate": 1.9051223914901114e-05, "loss": 0.9622, "step": 4976 }, { "epoch": 0.17, "learning_rate": 1.9050765599930348e-05, "loss": 0.985, "step": 4977 }, { "epoch": 0.17, "learning_rate": 1.9050307179805252e-05, "loss": 0.9822, "step": 4978 }, { "epoch": 0.17, "learning_rate": 1.9049848654531164e-05, "loss": 0.9868, "step": 4979 }, { "epoch": 0.17, "learning_rate": 1.9049390024113402e-05, "loss": 0.9568, "step": 4980 }, { "epoch": 0.17, "learning_rate": 1.9048931288557296e-05, "loss": 0.9955, "step": 4981 }, { "epoch": 0.17, "learning_rate": 1.9048472447868175e-05, "loss": 0.99, "step": 4982 }, { "epoch": 0.17, "learning_rate": 1.9048013502051373e-05, "loss": 0.9464, "step": 4983 }, { "epoch": 0.17, "learning_rate": 1.904755445111222e-05, "loss": 0.9634, "step": 4984 }, { "epoch": 0.17, "learning_rate": 1.904709529505605e-05, "loss": 0.9712, "step": 4985 }, { "epoch": 0.17, "learning_rate": 1.9046636033888192e-05, "loss": 0.9771, "step": 4986 }, { "epoch": 0.17, "learning_rate": 1.9046176667613993e-05, "loss": 0.9318, "step": 4987 }, { "epoch": 0.17, "learning_rate": 1.9045717196238783e-05, "loss": 0.9648, "step": 4988 }, { "epoch": 0.17, "learning_rate": 1.90452576197679e-05, "loss": 1.0, "step": 4989 }, { "epoch": 0.17, "learning_rate": 1.9044797938206685e-05, "loss": 0.9723, "step": 4990 }, { "epoch": 0.17, "learning_rate": 1.904433815156048e-05, "loss": 1.0033, "step": 4991 }, { "epoch": 0.17, "learning_rate": 1.9043878259834626e-05, "loss": 0.9597, "step": 4992 }, { "epoch": 0.17, "learning_rate": 1.9043418263034464e-05, "loss": 1.0101, "step": 4993 }, { "epoch": 0.17, "learning_rate": 1.904295816116534e-05, "loss": 0.9355, "step": 4994 }, { "epoch": 0.17, "learning_rate": 1.9042497954232598e-05, "loss": 0.9908, "step": 4995 }, { "epoch": 0.17, "learning_rate": 1.904203764224159e-05, "loss": 0.9722, "step": 4996 }, { "epoch": 0.17, "learning_rate": 1.904157722519766e-05, "loss": 0.9919, "step": 4997 }, { "epoch": 0.17, "learning_rate": 1.9041116703106156e-05, "loss": 0.996, "step": 4998 }, { "epoch": 0.17, "learning_rate": 1.9040656075972433e-05, "loss": 0.9767, "step": 4999 }, { "epoch": 0.17, "learning_rate": 1.9040195343801838e-05, "loss": 0.9704, "step": 5000 }, { "epoch": 0.17, "learning_rate": 1.9039734506599726e-05, "loss": 1.0673, "step": 5001 }, { "epoch": 0.17, "learning_rate": 1.903927356437145e-05, "loss": 0.9698, "step": 5002 }, { "epoch": 0.17, "learning_rate": 1.9038812517122367e-05, "loss": 1.0493, "step": 5003 }, { "epoch": 0.17, "learning_rate": 1.9038351364857836e-05, "loss": 0.9536, "step": 5004 }, { "epoch": 0.17, "learning_rate": 1.903789010758321e-05, "loss": 0.9744, "step": 5005 }, { "epoch": 0.17, "learning_rate": 1.9037428745303845e-05, "loss": 0.9743, "step": 5006 }, { "epoch": 0.17, "learning_rate": 1.903696727802511e-05, "loss": 1.0011, "step": 5007 }, { "epoch": 0.17, "learning_rate": 1.9036505705752365e-05, "loss": 0.9716, "step": 5008 }, { "epoch": 0.17, "learning_rate": 1.9036044028490965e-05, "loss": 0.9544, "step": 5009 }, { "epoch": 0.17, "learning_rate": 1.9035582246246283e-05, "loss": 1.0198, "step": 5010 }, { "epoch": 0.17, "learning_rate": 1.9035120359023674e-05, "loss": 1.0433, "step": 5011 }, { "epoch": 0.17, "learning_rate": 1.9034658366828517e-05, "loss": 0.9364, "step": 5012 }, { "epoch": 0.17, "learning_rate": 1.903419626966617e-05, "loss": 1.0208, "step": 5013 }, { "epoch": 0.17, "learning_rate": 1.9033734067542007e-05, "loss": 1.0121, "step": 5014 }, { "epoch": 0.17, "learning_rate": 1.9033271760461393e-05, "loss": 1.0385, "step": 5015 }, { "epoch": 0.17, "learning_rate": 1.9032809348429703e-05, "loss": 1.0191, "step": 5016 }, { "epoch": 0.17, "learning_rate": 1.903234683145231e-05, "loss": 1.0246, "step": 5017 }, { "epoch": 0.17, "learning_rate": 1.9031884209534585e-05, "loss": 0.9583, "step": 5018 }, { "epoch": 0.17, "learning_rate": 1.90314214826819e-05, "loss": 0.9986, "step": 5019 }, { "epoch": 0.17, "learning_rate": 1.9030958650899642e-05, "loss": 0.9398, "step": 5020 }, { "epoch": 0.17, "learning_rate": 1.9030495714193177e-05, "loss": 0.987, "step": 5021 }, { "epoch": 0.17, "learning_rate": 1.9030032672567887e-05, "loss": 0.977, "step": 5022 }, { "epoch": 0.17, "learning_rate": 1.9029569526029158e-05, "loss": 0.965, "step": 5023 }, { "epoch": 0.17, "learning_rate": 1.902910627458236e-05, "loss": 0.9452, "step": 5024 }, { "epoch": 0.17, "learning_rate": 1.9028642918232884e-05, "loss": 0.967, "step": 5025 }, { "epoch": 0.17, "learning_rate": 1.9028179456986108e-05, "loss": 0.9936, "step": 5026 }, { "epoch": 0.17, "learning_rate": 1.902771589084742e-05, "loss": 0.9746, "step": 5027 }, { "epoch": 0.17, "learning_rate": 1.9027252219822206e-05, "loss": 0.9674, "step": 5028 }, { "epoch": 0.17, "learning_rate": 1.902678844391585e-05, "loss": 1.0369, "step": 5029 }, { "epoch": 0.17, "learning_rate": 1.902632456313374e-05, "loss": 0.9868, "step": 5030 }, { "epoch": 0.17, "learning_rate": 1.902586057748127e-05, "loss": 0.9881, "step": 5031 }, { "epoch": 0.17, "learning_rate": 1.9025396486963827e-05, "loss": 0.9604, "step": 5032 }, { "epoch": 0.17, "learning_rate": 1.9024932291586805e-05, "loss": 0.9562, "step": 5033 }, { "epoch": 0.17, "learning_rate": 1.9024467991355595e-05, "loss": 0.971, "step": 5034 }, { "epoch": 0.17, "learning_rate": 1.9024003586275594e-05, "loss": 0.966, "step": 5035 }, { "epoch": 0.17, "learning_rate": 1.90235390763522e-05, "loss": 0.9726, "step": 5036 }, { "epoch": 0.17, "learning_rate": 1.9023074461590797e-05, "loss": 1.0005, "step": 5037 }, { "epoch": 0.17, "learning_rate": 1.9022609741996797e-05, "loss": 0.9295, "step": 5038 }, { "epoch": 0.17, "learning_rate": 1.9022144917575595e-05, "loss": 0.9767, "step": 5039 }, { "epoch": 0.17, "learning_rate": 1.902167998833259e-05, "loss": 1.0009, "step": 5040 }, { "epoch": 0.17, "learning_rate": 1.9021214954273185e-05, "loss": 0.98, "step": 5041 }, { "epoch": 0.17, "learning_rate": 1.902074981540278e-05, "loss": 0.9256, "step": 5042 }, { "epoch": 0.17, "learning_rate": 1.9020284571726783e-05, "loss": 0.9655, "step": 5043 }, { "epoch": 0.17, "learning_rate": 1.9019819223250596e-05, "loss": 0.9378, "step": 5044 }, { "epoch": 0.17, "learning_rate": 1.901935376997963e-05, "loss": 1.0022, "step": 5045 }, { "epoch": 0.17, "learning_rate": 1.901888821191929e-05, "loss": 0.9409, "step": 5046 }, { "epoch": 0.17, "learning_rate": 1.9018422549074987e-05, "loss": 0.9854, "step": 5047 }, { "epoch": 0.17, "learning_rate": 1.9017956781452126e-05, "loss": 0.932, "step": 5048 }, { "epoch": 0.17, "learning_rate": 1.901749090905612e-05, "loss": 0.9895, "step": 5049 }, { "epoch": 0.17, "learning_rate": 1.9017024931892387e-05, "loss": 0.9595, "step": 5050 }, { "epoch": 0.17, "learning_rate": 1.9016558849966333e-05, "loss": 0.9734, "step": 5051 }, { "epoch": 0.17, "learning_rate": 1.9016092663283383e-05, "loss": 0.9688, "step": 5052 }, { "epoch": 0.17, "learning_rate": 1.9015626371848944e-05, "loss": 1.0311, "step": 5053 }, { "epoch": 0.17, "learning_rate": 1.9015159975668438e-05, "loss": 0.9704, "step": 5054 }, { "epoch": 0.17, "learning_rate": 1.901469347474728e-05, "loss": 0.9556, "step": 5055 }, { "epoch": 0.17, "learning_rate": 1.9014226869090894e-05, "loss": 0.9587, "step": 5056 }, { "epoch": 0.17, "learning_rate": 1.9013760158704702e-05, "loss": 0.9953, "step": 5057 }, { "epoch": 0.17, "learning_rate": 1.9013293343594125e-05, "loss": 0.9744, "step": 5058 }, { "epoch": 0.17, "learning_rate": 1.901282642376458e-05, "loss": 0.9542, "step": 5059 }, { "epoch": 0.17, "learning_rate": 1.9012359399221504e-05, "loss": 0.9794, "step": 5060 }, { "epoch": 0.17, "learning_rate": 1.9011892269970313e-05, "loss": 0.9935, "step": 5061 }, { "epoch": 0.17, "learning_rate": 1.901142503601644e-05, "loss": 0.9783, "step": 5062 }, { "epoch": 0.17, "learning_rate": 1.901095769736531e-05, "loss": 0.9759, "step": 5063 }, { "epoch": 0.17, "learning_rate": 1.9010490254022355e-05, "loss": 0.968, "step": 5064 }, { "epoch": 0.17, "learning_rate": 1.9010022705993003e-05, "loss": 0.9673, "step": 5065 }, { "epoch": 0.17, "learning_rate": 1.9009555053282687e-05, "loss": 0.9807, "step": 5066 }, { "epoch": 0.17, "learning_rate": 1.9009087295896847e-05, "loss": 0.9983, "step": 5067 }, { "epoch": 0.17, "learning_rate": 1.9008619433840906e-05, "loss": 0.9553, "step": 5068 }, { "epoch": 0.17, "learning_rate": 1.9008151467120308e-05, "loss": 1.0089, "step": 5069 }, { "epoch": 0.17, "learning_rate": 1.900768339574049e-05, "loss": 0.9805, "step": 5070 }, { "epoch": 0.17, "learning_rate": 1.9007215219706883e-05, "loss": 1.0234, "step": 5071 }, { "epoch": 0.17, "learning_rate": 1.9006746939024935e-05, "loss": 0.9891, "step": 5072 }, { "epoch": 0.17, "learning_rate": 1.900627855370008e-05, "loss": 0.9639, "step": 5073 }, { "epoch": 0.17, "learning_rate": 1.9005810063737763e-05, "loss": 0.9676, "step": 5074 }, { "epoch": 0.17, "learning_rate": 1.900534146914343e-05, "loss": 0.9699, "step": 5075 }, { "epoch": 0.17, "learning_rate": 1.9004872769922515e-05, "loss": 0.984, "step": 5076 }, { "epoch": 0.17, "learning_rate": 1.9004403966080476e-05, "loss": 1.0008, "step": 5077 }, { "epoch": 0.17, "learning_rate": 1.9003935057622752e-05, "loss": 0.9677, "step": 5078 }, { "epoch": 0.17, "learning_rate": 1.9003466044554796e-05, "loss": 1.0099, "step": 5079 }, { "epoch": 0.17, "learning_rate": 1.9002996926882052e-05, "loss": 0.9616, "step": 5080 }, { "epoch": 0.17, "learning_rate": 1.9002527704609968e-05, "loss": 0.993, "step": 5081 }, { "epoch": 0.17, "learning_rate": 1.900205837774401e-05, "loss": 0.9843, "step": 5082 }, { "epoch": 0.17, "learning_rate": 1.900158894628961e-05, "loss": 0.9883, "step": 5083 }, { "epoch": 0.17, "learning_rate": 1.900111941025224e-05, "loss": 0.9538, "step": 5084 }, { "epoch": 0.17, "learning_rate": 1.9000649769637344e-05, "loss": 0.9788, "step": 5085 }, { "epoch": 0.17, "learning_rate": 1.9000180024450385e-05, "loss": 0.9806, "step": 5086 }, { "epoch": 0.17, "learning_rate": 1.8999710174696816e-05, "loss": 1.004, "step": 5087 }, { "epoch": 0.17, "learning_rate": 1.89992402203821e-05, "loss": 0.9476, "step": 5088 }, { "epoch": 0.17, "learning_rate": 1.8998770161511693e-05, "loss": 0.9714, "step": 5089 }, { "epoch": 0.17, "learning_rate": 1.899829999809106e-05, "loss": 0.9545, "step": 5090 }, { "epoch": 0.17, "learning_rate": 1.8997829730125662e-05, "loss": 0.9662, "step": 5091 }, { "epoch": 0.17, "learning_rate": 1.8997359357620963e-05, "loss": 0.9609, "step": 5092 }, { "epoch": 0.17, "learning_rate": 1.8996888880582426e-05, "loss": 0.9793, "step": 5093 }, { "epoch": 0.17, "learning_rate": 1.8996418299015515e-05, "loss": 0.9703, "step": 5094 }, { "epoch": 0.17, "learning_rate": 1.8995947612925706e-05, "loss": 0.994, "step": 5095 }, { "epoch": 0.17, "learning_rate": 1.899547682231846e-05, "loss": 0.937, "step": 5096 }, { "epoch": 0.17, "learning_rate": 1.8995005927199246e-05, "loss": 0.981, "step": 5097 }, { "epoch": 0.17, "learning_rate": 1.899453492757354e-05, "loss": 0.9736, "step": 5098 }, { "epoch": 0.17, "learning_rate": 1.8994063823446817e-05, "loss": 0.9861, "step": 5099 }, { "epoch": 0.17, "learning_rate": 1.8993592614824542e-05, "loss": 0.9777, "step": 5100 }, { "epoch": 0.17, "learning_rate": 1.8993121301712194e-05, "loss": 1.0096, "step": 5101 }, { "epoch": 0.17, "learning_rate": 1.899264988411525e-05, "loss": 0.9621, "step": 5102 }, { "epoch": 0.17, "learning_rate": 1.899217836203918e-05, "loss": 1.0013, "step": 5103 }, { "epoch": 0.17, "learning_rate": 1.8991706735489472e-05, "loss": 1.0021, "step": 5104 }, { "epoch": 0.17, "learning_rate": 1.89912350044716e-05, "loss": 0.9272, "step": 5105 }, { "epoch": 0.17, "learning_rate": 1.8990763168991045e-05, "loss": 0.9853, "step": 5106 }, { "epoch": 0.17, "learning_rate": 1.8990291229053293e-05, "loss": 0.9318, "step": 5107 }, { "epoch": 0.17, "learning_rate": 1.898981918466382e-05, "loss": 0.9781, "step": 5108 }, { "epoch": 0.17, "learning_rate": 1.8989347035828115e-05, "loss": 1.0079, "step": 5109 }, { "epoch": 0.17, "learning_rate": 1.8988874782551662e-05, "loss": 0.9903, "step": 5110 }, { "epoch": 0.17, "learning_rate": 1.898840242483995e-05, "loss": 0.945, "step": 5111 }, { "epoch": 0.17, "learning_rate": 1.8987929962698465e-05, "loss": 0.9736, "step": 5112 }, { "epoch": 0.17, "learning_rate": 1.8987457396132698e-05, "loss": 0.9583, "step": 5113 }, { "epoch": 0.17, "learning_rate": 1.8986984725148138e-05, "loss": 1.034, "step": 5114 }, { "epoch": 0.17, "learning_rate": 1.8986511949750273e-05, "loss": 0.9897, "step": 5115 }, { "epoch": 0.17, "learning_rate": 1.89860390699446e-05, "loss": 0.9577, "step": 5116 }, { "epoch": 0.17, "learning_rate": 1.8985566085736614e-05, "loss": 0.9719, "step": 5117 }, { "epoch": 0.17, "learning_rate": 1.8985092997131814e-05, "loss": 0.9421, "step": 5118 }, { "epoch": 0.17, "learning_rate": 1.8984619804135685e-05, "loss": 1.0177, "step": 5119 }, { "epoch": 0.17, "learning_rate": 1.898414650675373e-05, "loss": 1.0244, "step": 5120 }, { "epoch": 0.17, "learning_rate": 1.8983673104991457e-05, "loss": 0.979, "step": 5121 }, { "epoch": 0.17, "learning_rate": 1.898319959885435e-05, "loss": 0.9131, "step": 5122 }, { "epoch": 0.17, "learning_rate": 1.898272598834792e-05, "loss": 0.9813, "step": 5123 }, { "epoch": 0.17, "learning_rate": 1.898225227347767e-05, "loss": 0.9656, "step": 5124 }, { "epoch": 0.17, "learning_rate": 1.89817784542491e-05, "loss": 1.0398, "step": 5125 }, { "epoch": 0.17, "learning_rate": 1.8981304530667715e-05, "loss": 1.0253, "step": 5126 }, { "epoch": 0.17, "learning_rate": 1.8980830502739026e-05, "loss": 0.9875, "step": 5127 }, { "epoch": 0.17, "learning_rate": 1.8980356370468537e-05, "loss": 0.931, "step": 5128 }, { "epoch": 0.17, "learning_rate": 1.8979882133861755e-05, "loss": 0.9708, "step": 5129 }, { "epoch": 0.17, "learning_rate": 1.8979407792924194e-05, "loss": 1.0022, "step": 5130 }, { "epoch": 0.17, "learning_rate": 1.8978933347661362e-05, "loss": 0.975, "step": 5131 }, { "epoch": 0.17, "learning_rate": 1.897845879807877e-05, "loss": 0.9534, "step": 5132 }, { "epoch": 0.17, "learning_rate": 1.8977984144181935e-05, "loss": 0.9574, "step": 5133 }, { "epoch": 0.17, "learning_rate": 1.8977509385976367e-05, "loss": 0.9986, "step": 5134 }, { "epoch": 0.17, "learning_rate": 1.897703452346759e-05, "loss": 0.9611, "step": 5135 }, { "epoch": 0.17, "learning_rate": 1.897655955666111e-05, "loss": 0.973, "step": 5136 }, { "epoch": 0.17, "learning_rate": 1.8976084485562453e-05, "loss": 0.9653, "step": 5137 }, { "epoch": 0.17, "learning_rate": 1.897560931017714e-05, "loss": 0.9555, "step": 5138 }, { "epoch": 0.17, "learning_rate": 1.8975134030510686e-05, "loss": 0.9806, "step": 5139 }, { "epoch": 0.17, "learning_rate": 1.8974658646568612e-05, "loss": 0.9988, "step": 5140 }, { "epoch": 0.17, "learning_rate": 1.897418315835645e-05, "loss": 0.9964, "step": 5141 }, { "epoch": 0.17, "learning_rate": 1.8973707565879717e-05, "loss": 0.9986, "step": 5142 }, { "epoch": 0.17, "learning_rate": 1.897323186914394e-05, "loss": 0.9801, "step": 5143 }, { "epoch": 0.17, "learning_rate": 1.8972756068154648e-05, "loss": 0.9665, "step": 5144 }, { "epoch": 0.17, "learning_rate": 1.8972280162917368e-05, "loss": 0.9489, "step": 5145 }, { "epoch": 0.17, "learning_rate": 1.8971804153437624e-05, "loss": 0.9577, "step": 5146 }, { "epoch": 0.17, "learning_rate": 1.8971328039720952e-05, "loss": 1.0101, "step": 5147 }, { "epoch": 0.17, "learning_rate": 1.8970851821772884e-05, "loss": 1.0026, "step": 5148 }, { "epoch": 0.17, "learning_rate": 1.897037549959895e-05, "loss": 1.0107, "step": 5149 }, { "epoch": 0.17, "learning_rate": 1.8969899073204687e-05, "loss": 1.0077, "step": 5150 }, { "epoch": 0.17, "learning_rate": 1.8969422542595627e-05, "loss": 1.0078, "step": 5151 }, { "epoch": 0.17, "learning_rate": 1.896894590777731e-05, "loss": 0.9555, "step": 5152 }, { "epoch": 0.17, "learning_rate": 1.896846916875527e-05, "loss": 0.9977, "step": 5153 }, { "epoch": 0.17, "learning_rate": 1.8967992325535046e-05, "loss": 0.9373, "step": 5154 }, { "epoch": 0.17, "learning_rate": 1.8967515378122184e-05, "loss": 0.9809, "step": 5155 }, { "epoch": 0.17, "learning_rate": 1.896703832652222e-05, "loss": 1.0074, "step": 5156 }, { "epoch": 0.17, "learning_rate": 1.8966561170740692e-05, "loss": 1.028, "step": 5157 }, { "epoch": 0.17, "learning_rate": 1.896608391078316e-05, "loss": 1.0163, "step": 5158 }, { "epoch": 0.17, "learning_rate": 1.896560654665515e-05, "loss": 0.9982, "step": 5159 }, { "epoch": 0.17, "learning_rate": 1.8965129078362218e-05, "loss": 0.942, "step": 5160 }, { "epoch": 0.17, "learning_rate": 1.8964651505909907e-05, "loss": 0.9973, "step": 5161 }, { "epoch": 0.17, "learning_rate": 1.8964173829303775e-05, "loss": 1.0004, "step": 5162 }, { "epoch": 0.17, "learning_rate": 1.896369604854936e-05, "loss": 0.9879, "step": 5163 }, { "epoch": 0.17, "learning_rate": 1.8963218163652218e-05, "loss": 0.9656, "step": 5164 }, { "epoch": 0.17, "learning_rate": 1.8962740174617902e-05, "loss": 0.9828, "step": 5165 }, { "epoch": 0.17, "learning_rate": 1.8962262081451965e-05, "loss": 0.9592, "step": 5166 }, { "epoch": 0.17, "learning_rate": 1.8961783884159962e-05, "loss": 0.943, "step": 5167 }, { "epoch": 0.17, "learning_rate": 1.8961305582747447e-05, "loss": 0.9411, "step": 5168 }, { "epoch": 0.17, "learning_rate": 1.8960827177219977e-05, "loss": 0.9693, "step": 5169 }, { "epoch": 0.17, "learning_rate": 1.8960348667583112e-05, "loss": 0.9639, "step": 5170 }, { "epoch": 0.17, "learning_rate": 1.8959870053842414e-05, "loss": 1.0062, "step": 5171 }, { "epoch": 0.17, "learning_rate": 1.8959391336003437e-05, "loss": 0.9369, "step": 5172 }, { "epoch": 0.17, "learning_rate": 1.895891251407175e-05, "loss": 0.9814, "step": 5173 }, { "epoch": 0.17, "learning_rate": 1.895843358805291e-05, "loss": 0.9922, "step": 5174 }, { "epoch": 0.17, "learning_rate": 1.8957954557952486e-05, "loss": 1.0128, "step": 5175 }, { "epoch": 0.17, "learning_rate": 1.8957475423776036e-05, "loss": 0.9607, "step": 5176 }, { "epoch": 0.17, "learning_rate": 1.8956996185529134e-05, "loss": 0.9719, "step": 5177 }, { "epoch": 0.17, "learning_rate": 1.895651684321735e-05, "loss": 0.9685, "step": 5178 }, { "epoch": 0.17, "learning_rate": 1.8956037396846246e-05, "loss": 1.0027, "step": 5179 }, { "epoch": 0.17, "learning_rate": 1.8955557846421394e-05, "loss": 0.9584, "step": 5180 }, { "epoch": 0.17, "learning_rate": 1.895507819194837e-05, "loss": 0.965, "step": 5181 }, { "epoch": 0.17, "learning_rate": 1.8954598433432744e-05, "loss": 0.9384, "step": 5182 }, { "epoch": 0.17, "learning_rate": 1.8954118570880086e-05, "loss": 0.9684, "step": 5183 }, { "epoch": 0.17, "learning_rate": 1.8953638604295976e-05, "loss": 0.9651, "step": 5184 }, { "epoch": 0.17, "learning_rate": 1.8953158533685992e-05, "loss": 0.9795, "step": 5185 }, { "epoch": 0.17, "learning_rate": 1.8952678359055707e-05, "loss": 0.9781, "step": 5186 }, { "epoch": 0.17, "learning_rate": 1.8952198080410704e-05, "loss": 0.9758, "step": 5187 }, { "epoch": 0.17, "learning_rate": 1.8951717697756557e-05, "loss": 0.9706, "step": 5188 }, { "epoch": 0.17, "learning_rate": 1.8951237211098852e-05, "loss": 0.9612, "step": 5189 }, { "epoch": 0.17, "learning_rate": 1.895075662044317e-05, "loss": 1.0273, "step": 5190 }, { "epoch": 0.17, "learning_rate": 1.8950275925795094e-05, "loss": 1.0061, "step": 5191 }, { "epoch": 0.17, "learning_rate": 1.894979512716021e-05, "loss": 0.9699, "step": 5192 }, { "epoch": 0.17, "learning_rate": 1.8949314224544108e-05, "loss": 0.9579, "step": 5193 }, { "epoch": 0.17, "learning_rate": 1.894883321795237e-05, "loss": 1.007, "step": 5194 }, { "epoch": 0.17, "learning_rate": 1.8948352107390583e-05, "loss": 1.0028, "step": 5195 }, { "epoch": 0.17, "learning_rate": 1.8947870892864338e-05, "loss": 0.9535, "step": 5196 }, { "epoch": 0.17, "learning_rate": 1.8947389574379232e-05, "loss": 0.9701, "step": 5197 }, { "epoch": 0.17, "learning_rate": 1.8946908151940846e-05, "loss": 0.9419, "step": 5198 }, { "epoch": 0.17, "learning_rate": 1.8946426625554784e-05, "loss": 1.0178, "step": 5199 }, { "epoch": 0.17, "learning_rate": 1.8945944995226635e-05, "loss": 0.9755, "step": 5200 }, { "epoch": 0.17, "learning_rate": 1.8945463260961997e-05, "loss": 1.0079, "step": 5201 }, { "epoch": 0.17, "learning_rate": 1.894498142276646e-05, "loss": 1.0039, "step": 5202 }, { "epoch": 0.17, "learning_rate": 1.8944499480645636e-05, "loss": 0.9729, "step": 5203 }, { "epoch": 0.17, "learning_rate": 1.8944017434605114e-05, "loss": 0.9447, "step": 5204 }, { "epoch": 0.17, "learning_rate": 1.8943535284650496e-05, "loss": 1.0242, "step": 5205 }, { "epoch": 0.17, "learning_rate": 1.894305303078738e-05, "loss": 0.9511, "step": 5206 }, { "epoch": 0.17, "learning_rate": 1.8942570673021377e-05, "loss": 1.003, "step": 5207 }, { "epoch": 0.17, "learning_rate": 1.8942088211358088e-05, "loss": 0.9392, "step": 5208 }, { "epoch": 0.17, "learning_rate": 1.8941605645803115e-05, "loss": 1.0141, "step": 5209 }, { "epoch": 0.17, "learning_rate": 1.8941122976362072e-05, "loss": 0.9163, "step": 5210 }, { "epoch": 0.17, "learning_rate": 1.8940640203040563e-05, "loss": 0.9649, "step": 5211 }, { "epoch": 0.17, "learning_rate": 1.894015732584419e-05, "loss": 0.9496, "step": 5212 }, { "epoch": 0.17, "learning_rate": 1.8939674344778574e-05, "loss": 0.9226, "step": 5213 }, { "epoch": 0.17, "learning_rate": 1.893919125984932e-05, "loss": 0.9803, "step": 5214 }, { "epoch": 0.17, "learning_rate": 1.8938708071062043e-05, "loss": 1.0112, "step": 5215 }, { "epoch": 0.17, "learning_rate": 1.8938224778422357e-05, "loss": 0.9702, "step": 5216 }, { "epoch": 0.17, "learning_rate": 1.8937741381935876e-05, "loss": 0.96, "step": 5217 }, { "epoch": 0.17, "learning_rate": 1.8937257881608217e-05, "loss": 0.946, "step": 5218 }, { "epoch": 0.17, "learning_rate": 1.8936774277445e-05, "loss": 0.9966, "step": 5219 }, { "epoch": 0.17, "learning_rate": 1.8936290569451837e-05, "loss": 0.966, "step": 5220 }, { "epoch": 0.17, "learning_rate": 1.8935806757634348e-05, "loss": 0.998, "step": 5221 }, { "epoch": 0.17, "learning_rate": 1.893532284199816e-05, "loss": 0.9556, "step": 5222 }, { "epoch": 0.17, "learning_rate": 1.89348388225489e-05, "loss": 0.9876, "step": 5223 }, { "epoch": 0.17, "learning_rate": 1.8934354699292177e-05, "loss": 0.9685, "step": 5224 }, { "epoch": 0.17, "learning_rate": 1.8933870472233624e-05, "loss": 0.9515, "step": 5225 }, { "epoch": 0.17, "learning_rate": 1.8933386141378868e-05, "loss": 0.9527, "step": 5226 }, { "epoch": 0.17, "learning_rate": 1.8932901706733533e-05, "loss": 0.9604, "step": 5227 }, { "epoch": 0.17, "learning_rate": 1.8932417168303246e-05, "loss": 0.9963, "step": 5228 }, { "epoch": 0.17, "learning_rate": 1.8931932526093642e-05, "loss": 1.004, "step": 5229 }, { "epoch": 0.17, "learning_rate": 1.893144778011035e-05, "loss": 0.944, "step": 5230 }, { "epoch": 0.17, "learning_rate": 1.8930962930358995e-05, "loss": 0.9468, "step": 5231 }, { "epoch": 0.17, "learning_rate": 1.893047797684522e-05, "loss": 0.9677, "step": 5232 }, { "epoch": 0.17, "learning_rate": 1.8929992919574656e-05, "loss": 0.9767, "step": 5233 }, { "epoch": 0.17, "learning_rate": 1.8929507758552934e-05, "loss": 0.9489, "step": 5234 }, { "epoch": 0.17, "learning_rate": 1.8929022493785695e-05, "loss": 0.936, "step": 5235 }, { "epoch": 0.17, "learning_rate": 1.8928537125278576e-05, "loss": 0.9386, "step": 5236 }, { "epoch": 0.17, "learning_rate": 1.8928051653037217e-05, "loss": 0.9887, "step": 5237 }, { "epoch": 0.17, "learning_rate": 1.8927566077067257e-05, "loss": 0.9847, "step": 5238 }, { "epoch": 0.17, "learning_rate": 1.892708039737434e-05, "loss": 1.0226, "step": 5239 }, { "epoch": 0.17, "learning_rate": 1.8926594613964103e-05, "loss": 0.941, "step": 5240 }, { "epoch": 0.17, "learning_rate": 1.8926108726842198e-05, "loss": 0.9975, "step": 5241 }, { "epoch": 0.17, "learning_rate": 1.8925622736014264e-05, "loss": 0.9509, "step": 5242 }, { "epoch": 0.17, "learning_rate": 1.8925136641485947e-05, "loss": 0.9724, "step": 5243 }, { "epoch": 0.17, "learning_rate": 1.89246504432629e-05, "loss": 0.9695, "step": 5244 }, { "epoch": 0.17, "learning_rate": 1.892416414135077e-05, "loss": 1.0293, "step": 5245 }, { "epoch": 0.17, "learning_rate": 1.8923677735755204e-05, "loss": 1.0138, "step": 5246 }, { "epoch": 0.17, "learning_rate": 1.8923191226481856e-05, "loss": 0.9918, "step": 5247 }, { "epoch": 0.17, "learning_rate": 1.8922704613536374e-05, "loss": 0.9711, "step": 5248 }, { "epoch": 0.17, "learning_rate": 1.892221789692442e-05, "loss": 0.9793, "step": 5249 }, { "epoch": 0.17, "learning_rate": 1.892173107665164e-05, "loss": 0.9592, "step": 5250 }, { "epoch": 0.17, "learning_rate": 1.8921244152723694e-05, "loss": 0.9958, "step": 5251 }, { "epoch": 0.17, "learning_rate": 1.8920757125146244e-05, "loss": 0.9889, "step": 5252 }, { "epoch": 0.17, "learning_rate": 1.8920269993924938e-05, "loss": 0.9547, "step": 5253 }, { "epoch": 0.17, "learning_rate": 1.8919782759065444e-05, "loss": 1.0018, "step": 5254 }, { "epoch": 0.17, "learning_rate": 1.8919295420573417e-05, "loss": 0.988, "step": 5255 }, { "epoch": 0.17, "learning_rate": 1.8918807978454525e-05, "loss": 0.9581, "step": 5256 }, { "epoch": 0.17, "learning_rate": 1.8918320432714424e-05, "loss": 1.0036, "step": 5257 }, { "epoch": 0.17, "learning_rate": 1.8917832783358788e-05, "loss": 0.9629, "step": 5258 }, { "epoch": 0.18, "learning_rate": 1.8917345030393276e-05, "loss": 0.9908, "step": 5259 }, { "epoch": 0.18, "learning_rate": 1.8916857173823552e-05, "loss": 0.9668, "step": 5260 }, { "epoch": 0.18, "learning_rate": 1.8916369213655295e-05, "loss": 0.9945, "step": 5261 }, { "epoch": 0.18, "learning_rate": 1.8915881149894163e-05, "loss": 1.0325, "step": 5262 }, { "epoch": 0.18, "learning_rate": 1.8915392982545832e-05, "loss": 1.0076, "step": 5263 }, { "epoch": 0.18, "learning_rate": 1.8914904711615972e-05, "loss": 0.9611, "step": 5264 }, { "epoch": 0.18, "learning_rate": 1.8914416337110257e-05, "loss": 0.9831, "step": 5265 }, { "epoch": 0.18, "learning_rate": 1.8913927859034356e-05, "loss": 0.9779, "step": 5266 }, { "epoch": 0.18, "learning_rate": 1.8913439277393953e-05, "loss": 0.988, "step": 5267 }, { "epoch": 0.18, "learning_rate": 1.891295059219472e-05, "loss": 0.9905, "step": 5268 }, { "epoch": 0.18, "learning_rate": 1.8912461803442337e-05, "loss": 0.9955, "step": 5269 }, { "epoch": 0.18, "learning_rate": 1.891197291114248e-05, "loss": 0.9976, "step": 5270 }, { "epoch": 0.18, "learning_rate": 1.8911483915300828e-05, "loss": 0.9739, "step": 5271 }, { "epoch": 0.18, "learning_rate": 1.8910994815923066e-05, "loss": 0.9568, "step": 5272 }, { "epoch": 0.18, "learning_rate": 1.8910505613014873e-05, "loss": 0.9811, "step": 5273 }, { "epoch": 0.18, "learning_rate": 1.8910016306581936e-05, "loss": 0.9633, "step": 5274 }, { "epoch": 0.18, "learning_rate": 1.8909526896629935e-05, "loss": 1.0176, "step": 5275 }, { "epoch": 0.18, "learning_rate": 1.8909037383164562e-05, "loss": 0.9695, "step": 5276 }, { "epoch": 0.18, "learning_rate": 1.89085477661915e-05, "loss": 0.967, "step": 5277 }, { "epoch": 0.18, "learning_rate": 1.8908058045716443e-05, "loss": 0.934, "step": 5278 }, { "epoch": 0.18, "learning_rate": 1.8907568221745075e-05, "loss": 0.9716, "step": 5279 }, { "epoch": 0.18, "learning_rate": 1.890707829428309e-05, "loss": 0.9514, "step": 5280 }, { "epoch": 0.18, "learning_rate": 1.8906588263336177e-05, "loss": 0.9745, "step": 5281 }, { "epoch": 0.18, "learning_rate": 1.8906098128910032e-05, "loss": 0.955, "step": 5282 }, { "epoch": 0.18, "learning_rate": 1.8905607891010345e-05, "loss": 0.9754, "step": 5283 }, { "epoch": 0.18, "learning_rate": 1.8905117549642823e-05, "loss": 0.9936, "step": 5284 }, { "epoch": 0.18, "learning_rate": 1.890462710481315e-05, "loss": 0.9704, "step": 5285 }, { "epoch": 0.18, "learning_rate": 1.8904136556527032e-05, "loss": 0.9697, "step": 5286 }, { "epoch": 0.18, "learning_rate": 1.8903645904790165e-05, "loss": 0.9796, "step": 5287 }, { "epoch": 0.18, "learning_rate": 1.890315514960825e-05, "loss": 0.9957, "step": 5288 }, { "epoch": 0.18, "learning_rate": 1.890266429098699e-05, "loss": 0.9918, "step": 5289 }, { "epoch": 0.18, "learning_rate": 1.8902173328932085e-05, "loss": 0.9501, "step": 5290 }, { "epoch": 0.18, "learning_rate": 1.8901682263449242e-05, "loss": 0.9961, "step": 5291 }, { "epoch": 0.18, "learning_rate": 1.8901191094544167e-05, "loss": 0.953, "step": 5292 }, { "epoch": 0.18, "learning_rate": 1.8900699822222567e-05, "loss": 1.0043, "step": 5293 }, { "epoch": 0.18, "learning_rate": 1.8900208446490143e-05, "loss": 0.9659, "step": 5294 }, { "epoch": 0.18, "learning_rate": 1.889971696735261e-05, "loss": 1.0004, "step": 5295 }, { "epoch": 0.18, "learning_rate": 1.889922538481568e-05, "loss": 0.9568, "step": 5296 }, { "epoch": 0.18, "learning_rate": 1.8898733698885058e-05, "loss": 0.9391, "step": 5297 }, { "epoch": 0.18, "learning_rate": 1.8898241909566463e-05, "loss": 1.0137, "step": 5298 }, { "epoch": 0.18, "learning_rate": 1.8897750016865602e-05, "loss": 0.9948, "step": 5299 }, { "epoch": 0.18, "learning_rate": 1.8897258020788196e-05, "loss": 0.9075, "step": 5300 }, { "epoch": 0.18, "learning_rate": 1.8896765921339955e-05, "loss": 0.9794, "step": 5301 }, { "epoch": 0.18, "learning_rate": 1.8896273718526604e-05, "loss": 0.9826, "step": 5302 }, { "epoch": 0.18, "learning_rate": 1.8895781412353854e-05, "loss": 0.9675, "step": 5303 }, { "epoch": 0.18, "learning_rate": 1.889528900282743e-05, "loss": 0.9911, "step": 5304 }, { "epoch": 0.18, "learning_rate": 1.8894796489953052e-05, "loss": 1.004, "step": 5305 }, { "epoch": 0.18, "learning_rate": 1.8894303873736438e-05, "loss": 0.983, "step": 5306 }, { "epoch": 0.18, "learning_rate": 1.8893811154183322e-05, "loss": 0.9826, "step": 5307 }, { "epoch": 0.18, "learning_rate": 1.8893318331299414e-05, "loss": 0.9463, "step": 5308 }, { "epoch": 0.18, "learning_rate": 1.8892825405090452e-05, "loss": 1.0192, "step": 5309 }, { "epoch": 0.18, "learning_rate": 1.8892332375562153e-05, "loss": 0.9407, "step": 5310 }, { "epoch": 0.18, "learning_rate": 1.8891839242720257e-05, "loss": 0.9624, "step": 5311 }, { "epoch": 0.18, "learning_rate": 1.889134600657048e-05, "loss": 0.974, "step": 5312 }, { "epoch": 0.18, "learning_rate": 1.8890852667118563e-05, "loss": 0.9769, "step": 5313 }, { "epoch": 0.18, "learning_rate": 1.889035922437023e-05, "loss": 0.947, "step": 5314 }, { "epoch": 0.18, "learning_rate": 1.8889865678331222e-05, "loss": 0.9968, "step": 5315 }, { "epoch": 0.18, "learning_rate": 1.888937202900727e-05, "loss": 0.9968, "step": 5316 }, { "epoch": 0.18, "learning_rate": 1.8888878276404104e-05, "loss": 0.9392, "step": 5317 }, { "epoch": 0.18, "learning_rate": 1.8888384420527467e-05, "loss": 0.9818, "step": 5318 }, { "epoch": 0.18, "learning_rate": 1.8887890461383094e-05, "loss": 0.9763, "step": 5319 }, { "epoch": 0.18, "learning_rate": 1.8887396398976723e-05, "loss": 0.9711, "step": 5320 }, { "epoch": 0.18, "learning_rate": 1.8886902233314096e-05, "loss": 0.9359, "step": 5321 }, { "epoch": 0.18, "learning_rate": 1.8886407964400958e-05, "loss": 0.9393, "step": 5322 }, { "epoch": 0.18, "learning_rate": 1.8885913592243045e-05, "loss": 0.9642, "step": 5323 }, { "epoch": 0.18, "learning_rate": 1.8885419116846104e-05, "loss": 0.9546, "step": 5324 }, { "epoch": 0.18, "learning_rate": 1.8884924538215877e-05, "loss": 1.0292, "step": 5325 }, { "epoch": 0.18, "learning_rate": 1.8884429856358114e-05, "loss": 0.9779, "step": 5326 }, { "epoch": 0.18, "learning_rate": 1.888393507127856e-05, "loss": 0.9603, "step": 5327 }, { "epoch": 0.18, "learning_rate": 1.8883440182982965e-05, "loss": 0.9513, "step": 5328 }, { "epoch": 0.18, "learning_rate": 1.8882945191477078e-05, "loss": 0.9638, "step": 5329 }, { "epoch": 0.18, "learning_rate": 1.888245009676665e-05, "loss": 0.9471, "step": 5330 }, { "epoch": 0.18, "learning_rate": 1.8881954898857436e-05, "loss": 0.9989, "step": 5331 }, { "epoch": 0.18, "learning_rate": 1.8881459597755186e-05, "loss": 0.9488, "step": 5332 }, { "epoch": 0.18, "learning_rate": 1.888096419346565e-05, "loss": 0.991, "step": 5333 }, { "epoch": 0.18, "learning_rate": 1.8880468685994592e-05, "loss": 0.9753, "step": 5334 }, { "epoch": 0.18, "learning_rate": 1.8879973075347767e-05, "loss": 0.9873, "step": 5335 }, { "epoch": 0.18, "learning_rate": 1.8879477361530933e-05, "loss": 0.9777, "step": 5336 }, { "epoch": 0.18, "learning_rate": 1.8878981544549843e-05, "loss": 0.9678, "step": 5337 }, { "epoch": 0.18, "learning_rate": 1.8878485624410267e-05, "loss": 0.9735, "step": 5338 }, { "epoch": 0.18, "learning_rate": 1.8877989601117962e-05, "loss": 0.9699, "step": 5339 }, { "epoch": 0.18, "learning_rate": 1.887749347467869e-05, "loss": 0.9946, "step": 5340 }, { "epoch": 0.18, "learning_rate": 1.887699724509822e-05, "loss": 1.0082, "step": 5341 }, { "epoch": 0.18, "learning_rate": 1.8876500912382313e-05, "loss": 0.9663, "step": 5342 }, { "epoch": 0.18, "learning_rate": 1.8876004476536733e-05, "loss": 0.9726, "step": 5343 }, { "epoch": 0.18, "learning_rate": 1.8875507937567256e-05, "loss": 0.9739, "step": 5344 }, { "epoch": 0.18, "learning_rate": 1.8875011295479642e-05, "loss": 1.0048, "step": 5345 }, { "epoch": 0.18, "learning_rate": 1.887451455027967e-05, "loss": 0.9512, "step": 5346 }, { "epoch": 0.18, "learning_rate": 1.8874017701973103e-05, "loss": 0.9982, "step": 5347 }, { "epoch": 0.18, "learning_rate": 1.8873520750565716e-05, "loss": 0.9806, "step": 5348 }, { "epoch": 0.18, "learning_rate": 1.8873023696063287e-05, "loss": 1.0194, "step": 5349 }, { "epoch": 0.18, "learning_rate": 1.8872526538471588e-05, "loss": 0.9836, "step": 5350 }, { "epoch": 0.18, "learning_rate": 1.8872029277796392e-05, "loss": 0.9627, "step": 5351 }, { "epoch": 0.18, "learning_rate": 1.887153191404348e-05, "loss": 0.9501, "step": 5352 }, { "epoch": 0.18, "learning_rate": 1.8871034447218633e-05, "loss": 1.0169, "step": 5353 }, { "epoch": 0.18, "learning_rate": 1.8870536877327623e-05, "loss": 0.9354, "step": 5354 }, { "epoch": 0.18, "learning_rate": 1.8870039204376237e-05, "loss": 0.9766, "step": 5355 }, { "epoch": 0.18, "learning_rate": 1.886954142837026e-05, "loss": 0.9716, "step": 5356 }, { "epoch": 0.18, "learning_rate": 1.8869043549315465e-05, "loss": 0.9997, "step": 5357 }, { "epoch": 0.18, "learning_rate": 1.8868545567217643e-05, "loss": 0.9538, "step": 5358 }, { "epoch": 0.18, "learning_rate": 1.886804748208258e-05, "loss": 0.9786, "step": 5359 }, { "epoch": 0.18, "learning_rate": 1.886754929391606e-05, "loss": 0.9844, "step": 5360 }, { "epoch": 0.18, "learning_rate": 1.8867051002723875e-05, "loss": 1.0081, "step": 5361 }, { "epoch": 0.18, "learning_rate": 1.886655260851181e-05, "loss": 0.9666, "step": 5362 }, { "epoch": 0.18, "learning_rate": 1.886605411128566e-05, "loss": 1.0146, "step": 5363 }, { "epoch": 0.18, "learning_rate": 1.8865555511051215e-05, "loss": 0.9541, "step": 5364 }, { "epoch": 0.18, "learning_rate": 1.8865056807814264e-05, "loss": 1.0071, "step": 5365 }, { "epoch": 0.18, "learning_rate": 1.8864558001580608e-05, "loss": 0.9483, "step": 5366 }, { "epoch": 0.18, "learning_rate": 1.8864059092356037e-05, "loss": 1.0017, "step": 5367 }, { "epoch": 0.18, "learning_rate": 1.886356008014635e-05, "loss": 0.9999, "step": 5368 }, { "epoch": 0.18, "learning_rate": 1.886306096495734e-05, "loss": 0.9771, "step": 5369 }, { "epoch": 0.18, "learning_rate": 1.8862561746794815e-05, "loss": 0.9604, "step": 5370 }, { "epoch": 0.18, "learning_rate": 1.8862062425664566e-05, "loss": 0.9652, "step": 5371 }, { "epoch": 0.18, "learning_rate": 1.88615630015724e-05, "loss": 0.9643, "step": 5372 }, { "epoch": 0.18, "learning_rate": 1.8861063474524113e-05, "loss": 0.9671, "step": 5373 }, { "epoch": 0.18, "learning_rate": 1.8860563844525515e-05, "loss": 0.9565, "step": 5374 }, { "epoch": 0.18, "learning_rate": 1.8860064111582416e-05, "loss": 0.9495, "step": 5375 }, { "epoch": 0.18, "learning_rate": 1.8859564275700605e-05, "loss": 0.9099, "step": 5376 }, { "epoch": 0.18, "learning_rate": 1.8859064336885903e-05, "loss": 0.9768, "step": 5377 }, { "epoch": 0.18, "learning_rate": 1.8858564295144115e-05, "loss": 0.9596, "step": 5378 }, { "epoch": 0.18, "learning_rate": 1.885806415048105e-05, "loss": 0.9927, "step": 5379 }, { "epoch": 0.18, "learning_rate": 1.885756390290252e-05, "loss": 0.9949, "step": 5380 }, { "epoch": 0.18, "learning_rate": 1.885706355241433e-05, "loss": 0.976, "step": 5381 }, { "epoch": 0.18, "learning_rate": 1.8856563099022306e-05, "loss": 0.9483, "step": 5382 }, { "epoch": 0.18, "learning_rate": 1.8856062542732255e-05, "loss": 1.0187, "step": 5383 }, { "epoch": 0.18, "learning_rate": 1.885556188354999e-05, "loss": 0.9381, "step": 5384 }, { "epoch": 0.18, "learning_rate": 1.8855061121481335e-05, "loss": 0.9538, "step": 5385 }, { "epoch": 0.18, "learning_rate": 1.8854560256532098e-05, "loss": 0.9802, "step": 5386 }, { "epoch": 0.18, "learning_rate": 1.885405928870811e-05, "loss": 0.9811, "step": 5387 }, { "epoch": 0.18, "learning_rate": 1.8853558218015185e-05, "loss": 0.9265, "step": 5388 }, { "epoch": 0.18, "learning_rate": 1.8853057044459143e-05, "loss": 0.9749, "step": 5389 }, { "epoch": 0.18, "learning_rate": 1.885255576804581e-05, "loss": 0.9832, "step": 5390 }, { "epoch": 0.18, "learning_rate": 1.8852054388781007e-05, "loss": 0.931, "step": 5391 }, { "epoch": 0.18, "learning_rate": 1.885155290667056e-05, "loss": 1.0143, "step": 5392 }, { "epoch": 0.18, "learning_rate": 1.88510513217203e-05, "loss": 1.006, "step": 5393 }, { "epoch": 0.18, "learning_rate": 1.8850549633936052e-05, "loss": 0.9724, "step": 5394 }, { "epoch": 0.18, "learning_rate": 1.885004784332364e-05, "loss": 0.9714, "step": 5395 }, { "epoch": 0.18, "learning_rate": 1.88495459498889e-05, "loss": 0.9645, "step": 5396 }, { "epoch": 0.18, "learning_rate": 1.884904395363766e-05, "loss": 0.9829, "step": 5397 }, { "epoch": 0.18, "learning_rate": 1.8848541854575755e-05, "loss": 0.9491, "step": 5398 }, { "epoch": 0.18, "learning_rate": 1.8848039652709013e-05, "loss": 1.0018, "step": 5399 }, { "epoch": 0.18, "learning_rate": 1.8847537348043277e-05, "loss": 0.9565, "step": 5400 }, { "epoch": 0.18, "learning_rate": 1.8847034940584376e-05, "loss": 0.9457, "step": 5401 }, { "epoch": 0.18, "learning_rate": 1.8846532430338146e-05, "loss": 0.9778, "step": 5402 }, { "epoch": 0.18, "learning_rate": 1.8846029817310434e-05, "loss": 0.982, "step": 5403 }, { "epoch": 0.18, "learning_rate": 1.884552710150707e-05, "loss": 0.9476, "step": 5404 }, { "epoch": 0.18, "learning_rate": 1.8845024282933903e-05, "loss": 0.9548, "step": 5405 }, { "epoch": 0.18, "learning_rate": 1.8844521361596766e-05, "loss": 0.9965, "step": 5406 }, { "epoch": 0.18, "learning_rate": 1.884401833750151e-05, "loss": 0.9684, "step": 5407 }, { "epoch": 0.18, "learning_rate": 1.8843515210653978e-05, "loss": 0.9566, "step": 5408 }, { "epoch": 0.18, "learning_rate": 1.884301198106001e-05, "loss": 0.9611, "step": 5409 }, { "epoch": 0.18, "learning_rate": 1.8842508648725458e-05, "loss": 0.9885, "step": 5410 }, { "epoch": 0.18, "learning_rate": 1.8842005213656167e-05, "loss": 1.0081, "step": 5411 }, { "epoch": 0.18, "learning_rate": 1.884150167585799e-05, "loss": 0.9609, "step": 5412 }, { "epoch": 0.18, "learning_rate": 1.884099803533677e-05, "loss": 0.9693, "step": 5413 }, { "epoch": 0.18, "learning_rate": 1.8840494292098367e-05, "loss": 0.9967, "step": 5414 }, { "epoch": 0.18, "learning_rate": 1.8839990446148627e-05, "loss": 0.9769, "step": 5415 }, { "epoch": 0.18, "learning_rate": 1.8839486497493407e-05, "loss": 0.9558, "step": 5416 }, { "epoch": 0.18, "learning_rate": 1.8838982446138563e-05, "loss": 0.9714, "step": 5417 }, { "epoch": 0.18, "learning_rate": 1.8838478292089946e-05, "loss": 0.9859, "step": 5418 }, { "epoch": 0.18, "learning_rate": 1.8837974035353418e-05, "loss": 1.0016, "step": 5419 }, { "epoch": 0.18, "learning_rate": 1.883746967593484e-05, "loss": 0.9429, "step": 5420 }, { "epoch": 0.18, "learning_rate": 1.8836965213840065e-05, "loss": 0.9818, "step": 5421 }, { "epoch": 0.18, "learning_rate": 1.8836460649074956e-05, "loss": 0.9254, "step": 5422 }, { "epoch": 0.18, "learning_rate": 1.8835955981645382e-05, "loss": 0.9682, "step": 5423 }, { "epoch": 0.18, "learning_rate": 1.8835451211557197e-05, "loss": 0.9737, "step": 5424 }, { "epoch": 0.18, "learning_rate": 1.883494633881627e-05, "loss": 1.0219, "step": 5425 }, { "epoch": 0.18, "learning_rate": 1.8834441363428464e-05, "loss": 0.9911, "step": 5426 }, { "epoch": 0.18, "learning_rate": 1.883393628539965e-05, "loss": 0.9528, "step": 5427 }, { "epoch": 0.18, "learning_rate": 1.8833431104735695e-05, "loss": 0.9853, "step": 5428 }, { "epoch": 0.18, "learning_rate": 1.8832925821442468e-05, "loss": 0.9529, "step": 5429 }, { "epoch": 0.18, "learning_rate": 1.8832420435525836e-05, "loss": 0.9625, "step": 5430 }, { "epoch": 0.18, "learning_rate": 1.8831914946991678e-05, "loss": 0.9815, "step": 5431 }, { "epoch": 0.18, "learning_rate": 1.883140935584586e-05, "loss": 0.9642, "step": 5432 }, { "epoch": 0.18, "learning_rate": 1.8830903662094258e-05, "loss": 0.9589, "step": 5433 }, { "epoch": 0.18, "learning_rate": 1.8830397865742753e-05, "loss": 0.9348, "step": 5434 }, { "epoch": 0.18, "learning_rate": 1.8829891966797212e-05, "loss": 0.981, "step": 5435 }, { "epoch": 0.18, "learning_rate": 1.882938596526352e-05, "loss": 0.9748, "step": 5436 }, { "epoch": 0.18, "learning_rate": 1.882887986114755e-05, "loss": 0.9629, "step": 5437 }, { "epoch": 0.18, "learning_rate": 1.882837365445519e-05, "loss": 0.9786, "step": 5438 }, { "epoch": 0.18, "learning_rate": 1.8827867345192312e-05, "loss": 0.9607, "step": 5439 }, { "epoch": 0.18, "learning_rate": 1.8827360933364804e-05, "loss": 0.9178, "step": 5440 }, { "epoch": 0.18, "learning_rate": 1.8826854418978547e-05, "loss": 0.9708, "step": 5441 }, { "epoch": 0.18, "learning_rate": 1.882634780203943e-05, "loss": 0.9879, "step": 5442 }, { "epoch": 0.18, "learning_rate": 1.8825841082553336e-05, "loss": 0.9395, "step": 5443 }, { "epoch": 0.18, "learning_rate": 1.8825334260526152e-05, "loss": 0.9656, "step": 5444 }, { "epoch": 0.18, "learning_rate": 1.8824827335963767e-05, "loss": 0.9953, "step": 5445 }, { "epoch": 0.18, "learning_rate": 1.8824320308872072e-05, "loss": 0.929, "step": 5446 }, { "epoch": 0.18, "learning_rate": 1.8823813179256952e-05, "loss": 0.963, "step": 5447 }, { "epoch": 0.18, "learning_rate": 1.882330594712431e-05, "loss": 0.9727, "step": 5448 }, { "epoch": 0.18, "learning_rate": 1.8822798612480027e-05, "loss": 1.0261, "step": 5449 }, { "epoch": 0.18, "learning_rate": 1.8822291175330006e-05, "loss": 1.009, "step": 5450 }, { "epoch": 0.18, "learning_rate": 1.8821783635680138e-05, "loss": 0.9671, "step": 5451 }, { "epoch": 0.18, "learning_rate": 1.8821275993536322e-05, "loss": 0.9784, "step": 5452 }, { "epoch": 0.18, "learning_rate": 1.8820768248904454e-05, "loss": 1.0056, "step": 5453 }, { "epoch": 0.18, "learning_rate": 1.8820260401790435e-05, "loss": 0.9981, "step": 5454 }, { "epoch": 0.18, "learning_rate": 1.881975245220016e-05, "loss": 0.9659, "step": 5455 }, { "epoch": 0.18, "learning_rate": 1.881924440013954e-05, "loss": 0.9705, "step": 5456 }, { "epoch": 0.18, "learning_rate": 1.8818736245614474e-05, "loss": 0.9728, "step": 5457 }, { "epoch": 0.18, "learning_rate": 1.881822798863086e-05, "loss": 0.9479, "step": 5458 }, { "epoch": 0.18, "learning_rate": 1.8817719629194613e-05, "loss": 1.0144, "step": 5459 }, { "epoch": 0.18, "learning_rate": 1.881721116731163e-05, "loss": 0.9679, "step": 5460 }, { "epoch": 0.18, "learning_rate": 1.8816702602987825e-05, "loss": 1.0043, "step": 5461 }, { "epoch": 0.18, "learning_rate": 1.88161939362291e-05, "loss": 0.953, "step": 5462 }, { "epoch": 0.18, "learning_rate": 1.881568516704137e-05, "loss": 0.9634, "step": 5463 }, { "epoch": 0.18, "learning_rate": 1.8815176295430547e-05, "loss": 0.9749, "step": 5464 }, { "epoch": 0.18, "learning_rate": 1.881466732140254e-05, "loss": 0.9669, "step": 5465 }, { "epoch": 0.18, "learning_rate": 1.8814158244963262e-05, "loss": 0.9874, "step": 5466 }, { "epoch": 0.18, "learning_rate": 1.8813649066118633e-05, "loss": 0.9476, "step": 5467 }, { "epoch": 0.18, "learning_rate": 1.881313978487456e-05, "loss": 1.0061, "step": 5468 }, { "epoch": 0.18, "learning_rate": 1.8812630401236968e-05, "loss": 0.9501, "step": 5469 }, { "epoch": 0.18, "learning_rate": 1.881212091521177e-05, "loss": 0.9555, "step": 5470 }, { "epoch": 0.18, "learning_rate": 1.881161132680489e-05, "loss": 0.9904, "step": 5471 }, { "epoch": 0.18, "learning_rate": 1.8811101636022244e-05, "loss": 0.9665, "step": 5472 }, { "epoch": 0.18, "learning_rate": 1.8810591842869758e-05, "loss": 0.9566, "step": 5473 }, { "epoch": 0.18, "learning_rate": 1.881008194735335e-05, "loss": 0.9439, "step": 5474 }, { "epoch": 0.18, "learning_rate": 1.8809571949478947e-05, "loss": 0.9683, "step": 5475 }, { "epoch": 0.18, "learning_rate": 1.880906184925247e-05, "loss": 0.9437, "step": 5476 }, { "epoch": 0.18, "learning_rate": 1.8808551646679854e-05, "loss": 0.9999, "step": 5477 }, { "epoch": 0.18, "learning_rate": 1.8808041341767022e-05, "loss": 0.8992, "step": 5478 }, { "epoch": 0.18, "learning_rate": 1.8807530934519905e-05, "loss": 1.0314, "step": 5479 }, { "epoch": 0.18, "learning_rate": 1.8807020424944428e-05, "loss": 0.9689, "step": 5480 }, { "epoch": 0.18, "learning_rate": 1.8806509813046524e-05, "loss": 1.002, "step": 5481 }, { "epoch": 0.18, "learning_rate": 1.880599909883213e-05, "loss": 0.9592, "step": 5482 }, { "epoch": 0.18, "learning_rate": 1.8805488282307176e-05, "loss": 0.9973, "step": 5483 }, { "epoch": 0.18, "learning_rate": 1.8804977363477593e-05, "loss": 0.9509, "step": 5484 }, { "epoch": 0.18, "learning_rate": 1.8804466342349324e-05, "loss": 0.9506, "step": 5485 }, { "epoch": 0.18, "learning_rate": 1.8803955218928305e-05, "loss": 0.96, "step": 5486 }, { "epoch": 0.18, "learning_rate": 1.880344399322047e-05, "loss": 0.9894, "step": 5487 }, { "epoch": 0.18, "learning_rate": 1.8802932665231762e-05, "loss": 0.9671, "step": 5488 }, { "epoch": 0.18, "learning_rate": 1.880242123496812e-05, "loss": 1.0182, "step": 5489 }, { "epoch": 0.18, "learning_rate": 1.880190970243549e-05, "loss": 0.9555, "step": 5490 }, { "epoch": 0.18, "learning_rate": 1.880139806763981e-05, "loss": 0.978, "step": 5491 }, { "epoch": 0.18, "learning_rate": 1.880088633058703e-05, "loss": 0.9479, "step": 5492 }, { "epoch": 0.18, "learning_rate": 1.8800374491283083e-05, "loss": 0.9687, "step": 5493 }, { "epoch": 0.18, "learning_rate": 1.879986254973393e-05, "loss": 0.9544, "step": 5494 }, { "epoch": 0.18, "learning_rate": 1.8799350505945514e-05, "loss": 0.9814, "step": 5495 }, { "epoch": 0.18, "learning_rate": 1.8798838359923782e-05, "loss": 0.9752, "step": 5496 }, { "epoch": 0.18, "learning_rate": 1.8798326111674688e-05, "loss": 0.9996, "step": 5497 }, { "epoch": 0.18, "learning_rate": 1.879781376120418e-05, "loss": 0.9879, "step": 5498 }, { "epoch": 0.18, "learning_rate": 1.879730130851821e-05, "loss": 0.9984, "step": 5499 }, { "epoch": 0.18, "learning_rate": 1.8796788753622736e-05, "loss": 0.9555, "step": 5500 }, { "epoch": 0.18, "learning_rate": 1.879627609652371e-05, "loss": 0.9449, "step": 5501 }, { "epoch": 0.18, "learning_rate": 1.879576333722709e-05, "loss": 0.9999, "step": 5502 }, { "epoch": 0.18, "learning_rate": 1.8795250475738832e-05, "loss": 0.945, "step": 5503 }, { "epoch": 0.18, "learning_rate": 1.879473751206489e-05, "loss": 0.9414, "step": 5504 }, { "epoch": 0.18, "learning_rate": 1.8794224446211237e-05, "loss": 0.954, "step": 5505 }, { "epoch": 0.18, "learning_rate": 1.8793711278183817e-05, "loss": 0.9721, "step": 5506 }, { "epoch": 0.18, "learning_rate": 1.8793198007988608e-05, "loss": 0.9742, "step": 5507 }, { "epoch": 0.18, "learning_rate": 1.8792684635631562e-05, "loss": 0.9531, "step": 5508 }, { "epoch": 0.18, "learning_rate": 1.879217116111865e-05, "loss": 0.9791, "step": 5509 }, { "epoch": 0.18, "learning_rate": 1.8791657584455828e-05, "loss": 0.9556, "step": 5510 }, { "epoch": 0.18, "learning_rate": 1.8791143905649075e-05, "loss": 0.9872, "step": 5511 }, { "epoch": 0.18, "learning_rate": 1.8790630124704356e-05, "loss": 0.947, "step": 5512 }, { "epoch": 0.18, "learning_rate": 1.8790116241627636e-05, "loss": 0.9822, "step": 5513 }, { "epoch": 0.18, "learning_rate": 1.878960225642489e-05, "loss": 0.9651, "step": 5514 }, { "epoch": 0.18, "learning_rate": 1.8789088169102084e-05, "loss": 0.938, "step": 5515 }, { "epoch": 0.18, "learning_rate": 1.8788573979665196e-05, "loss": 0.9066, "step": 5516 }, { "epoch": 0.18, "learning_rate": 1.87880596881202e-05, "loss": 0.9756, "step": 5517 }, { "epoch": 0.18, "learning_rate": 1.878754529447307e-05, "loss": 0.9823, "step": 5518 }, { "epoch": 0.18, "learning_rate": 1.8787030798729776e-05, "loss": 0.9695, "step": 5519 }, { "epoch": 0.18, "learning_rate": 1.8786516200896304e-05, "loss": 0.9892, "step": 5520 }, { "epoch": 0.18, "learning_rate": 1.878600150097863e-05, "loss": 1.0163, "step": 5521 }, { "epoch": 0.18, "learning_rate": 1.878548669898274e-05, "loss": 0.9535, "step": 5522 }, { "epoch": 0.18, "learning_rate": 1.8784971794914598e-05, "loss": 0.9618, "step": 5523 }, { "epoch": 0.18, "learning_rate": 1.8784456788780207e-05, "loss": 0.9824, "step": 5524 }, { "epoch": 0.18, "learning_rate": 1.8783941680585536e-05, "loss": 0.9957, "step": 5525 }, { "epoch": 0.18, "learning_rate": 1.8783426470336574e-05, "loss": 0.9949, "step": 5526 }, { "epoch": 0.18, "learning_rate": 1.878291115803931e-05, "loss": 0.9922, "step": 5527 }, { "epoch": 0.18, "learning_rate": 1.878239574369973e-05, "loss": 0.9798, "step": 5528 }, { "epoch": 0.18, "learning_rate": 1.8781880227323818e-05, "loss": 1.0074, "step": 5529 }, { "epoch": 0.18, "learning_rate": 1.8781364608917563e-05, "loss": 0.9818, "step": 5530 }, { "epoch": 0.18, "learning_rate": 1.8780848888486963e-05, "loss": 0.9962, "step": 5531 }, { "epoch": 0.18, "learning_rate": 1.8780333066038006e-05, "loss": 0.9765, "step": 5532 }, { "epoch": 0.18, "learning_rate": 1.877981714157668e-05, "loss": 1.0136, "step": 5533 }, { "epoch": 0.18, "learning_rate": 1.877930111510899e-05, "loss": 0.9687, "step": 5534 }, { "epoch": 0.18, "learning_rate": 1.8778784986640918e-05, "loss": 0.953, "step": 5535 }, { "epoch": 0.18, "learning_rate": 1.8778268756178474e-05, "loss": 0.9093, "step": 5536 }, { "epoch": 0.18, "learning_rate": 1.8777752423727643e-05, "loss": 1.0219, "step": 5537 }, { "epoch": 0.18, "learning_rate": 1.8777235989294434e-05, "loss": 0.9728, "step": 5538 }, { "epoch": 0.18, "learning_rate": 1.8776719452884845e-05, "loss": 0.9742, "step": 5539 }, { "epoch": 0.18, "learning_rate": 1.877620281450487e-05, "loss": 0.974, "step": 5540 }, { "epoch": 0.18, "learning_rate": 1.8775686074160522e-05, "loss": 0.9958, "step": 5541 }, { "epoch": 0.18, "learning_rate": 1.8775169231857795e-05, "loss": 0.9653, "step": 5542 }, { "epoch": 0.18, "learning_rate": 1.8774652287602702e-05, "loss": 0.939, "step": 5543 }, { "epoch": 0.18, "learning_rate": 1.8774135241401242e-05, "loss": 0.9578, "step": 5544 }, { "epoch": 0.18, "learning_rate": 1.877361809325943e-05, "loss": 0.9511, "step": 5545 }, { "epoch": 0.18, "learning_rate": 1.877310084318327e-05, "loss": 0.9643, "step": 5546 }, { "epoch": 0.18, "learning_rate": 1.8772583491178764e-05, "loss": 0.9266, "step": 5547 }, { "epoch": 0.18, "learning_rate": 1.8772066037251936e-05, "loss": 0.9497, "step": 5548 }, { "epoch": 0.18, "learning_rate": 1.877154848140879e-05, "loss": 0.9601, "step": 5549 }, { "epoch": 0.18, "learning_rate": 1.8771030823655342e-05, "loss": 0.9669, "step": 5550 }, { "epoch": 0.18, "learning_rate": 1.8770513063997606e-05, "loss": 0.9758, "step": 5551 }, { "epoch": 0.18, "learning_rate": 1.8769995202441594e-05, "loss": 0.9436, "step": 5552 }, { "epoch": 0.18, "learning_rate": 1.876947723899333e-05, "loss": 0.9591, "step": 5553 }, { "epoch": 0.18, "learning_rate": 1.8768959173658824e-05, "loss": 0.8964, "step": 5554 }, { "epoch": 0.18, "learning_rate": 1.87684410064441e-05, "loss": 0.961, "step": 5555 }, { "epoch": 0.18, "learning_rate": 1.876792273735518e-05, "loss": 0.9852, "step": 5556 }, { "epoch": 0.18, "learning_rate": 1.8767404366398078e-05, "loss": 0.9824, "step": 5557 }, { "epoch": 0.18, "learning_rate": 1.876688589357882e-05, "loss": 0.9976, "step": 5558 }, { "epoch": 0.19, "learning_rate": 1.8766367318903432e-05, "loss": 0.9081, "step": 5559 }, { "epoch": 0.19, "learning_rate": 1.8765848642377936e-05, "loss": 0.9543, "step": 5560 }, { "epoch": 0.19, "learning_rate": 1.8765329864008362e-05, "loss": 0.9792, "step": 5561 }, { "epoch": 0.19, "learning_rate": 1.8764810983800736e-05, "loss": 0.9694, "step": 5562 }, { "epoch": 0.19, "learning_rate": 1.876429200176108e-05, "loss": 0.9543, "step": 5563 }, { "epoch": 0.19, "learning_rate": 1.8763772917895434e-05, "loss": 0.9675, "step": 5564 }, { "epoch": 0.19, "learning_rate": 1.8763253732209823e-05, "loss": 0.9807, "step": 5565 }, { "epoch": 0.19, "learning_rate": 1.876273444471028e-05, "loss": 0.9672, "step": 5566 }, { "epoch": 0.19, "learning_rate": 1.876221505540284e-05, "loss": 0.9218, "step": 5567 }, { "epoch": 0.19, "learning_rate": 1.876169556429353e-05, "loss": 0.9504, "step": 5568 }, { "epoch": 0.19, "learning_rate": 1.8761175971388393e-05, "loss": 0.9766, "step": 5569 }, { "epoch": 0.19, "learning_rate": 1.8760656276693464e-05, "loss": 0.9469, "step": 5570 }, { "epoch": 0.19, "learning_rate": 1.8760136480214784e-05, "loss": 0.974, "step": 5571 }, { "epoch": 0.19, "learning_rate": 1.8759616581958387e-05, "loss": 0.9434, "step": 5572 }, { "epoch": 0.19, "learning_rate": 1.8759096581930314e-05, "loss": 0.9718, "step": 5573 }, { "epoch": 0.19, "learning_rate": 1.875857648013661e-05, "loss": 1.0011, "step": 5574 }, { "epoch": 0.19, "learning_rate": 1.8758056276583314e-05, "loss": 0.9602, "step": 5575 }, { "epoch": 0.19, "learning_rate": 1.8757535971276472e-05, "loss": 0.9119, "step": 5576 }, { "epoch": 0.19, "learning_rate": 1.875701556422213e-05, "loss": 0.9851, "step": 5577 }, { "epoch": 0.19, "learning_rate": 1.8756495055426328e-05, "loss": 0.961, "step": 5578 }, { "epoch": 0.19, "learning_rate": 1.875597444489512e-05, "loss": 0.993, "step": 5579 }, { "epoch": 0.19, "learning_rate": 1.8755453732634556e-05, "loss": 0.9645, "step": 5580 }, { "epoch": 0.19, "learning_rate": 1.875493291865068e-05, "loss": 1.0215, "step": 5581 }, { "epoch": 0.19, "learning_rate": 1.8754412002949546e-05, "loss": 0.9718, "step": 5582 }, { "epoch": 0.19, "learning_rate": 1.8753890985537206e-05, "loss": 0.9956, "step": 5583 }, { "epoch": 0.19, "learning_rate": 1.8753369866419713e-05, "loss": 0.9736, "step": 5584 }, { "epoch": 0.19, "learning_rate": 1.875284864560312e-05, "loss": 0.9785, "step": 5585 }, { "epoch": 0.19, "learning_rate": 1.8752327323093485e-05, "loss": 0.9631, "step": 5586 }, { "epoch": 0.19, "learning_rate": 1.8751805898896865e-05, "loss": 0.9866, "step": 5587 }, { "epoch": 0.19, "learning_rate": 1.8751284373019314e-05, "loss": 0.9748, "step": 5588 }, { "epoch": 0.19, "learning_rate": 1.8750762745466896e-05, "loss": 0.9554, "step": 5589 }, { "epoch": 0.19, "learning_rate": 1.8750241016245673e-05, "loss": 0.9091, "step": 5590 }, { "epoch": 0.19, "learning_rate": 1.8749719185361698e-05, "loss": 0.9821, "step": 5591 }, { "epoch": 0.19, "learning_rate": 1.8749197252821044e-05, "loss": 0.9732, "step": 5592 }, { "epoch": 0.19, "learning_rate": 1.8748675218629765e-05, "loss": 0.9807, "step": 5593 }, { "epoch": 0.19, "learning_rate": 1.8748153082793933e-05, "loss": 0.9478, "step": 5594 }, { "epoch": 0.19, "learning_rate": 1.874763084531961e-05, "loss": 0.9505, "step": 5595 }, { "epoch": 0.19, "learning_rate": 1.8747108506212874e-05, "loss": 0.984, "step": 5596 }, { "epoch": 0.19, "learning_rate": 1.874658606547978e-05, "loss": 1.0078, "step": 5597 }, { "epoch": 0.19, "learning_rate": 1.8746063523126405e-05, "loss": 0.9536, "step": 5598 }, { "epoch": 0.19, "learning_rate": 1.8745540879158818e-05, "loss": 1.0207, "step": 5599 }, { "epoch": 0.19, "learning_rate": 1.874501813358309e-05, "loss": 0.8933, "step": 5600 }, { "epoch": 0.19, "learning_rate": 1.87444952864053e-05, "loss": 1.0313, "step": 5601 }, { "epoch": 0.19, "learning_rate": 1.8743972337631516e-05, "loss": 0.9645, "step": 5602 }, { "epoch": 0.19, "learning_rate": 1.8743449287267817e-05, "loss": 0.9803, "step": 5603 }, { "epoch": 0.19, "learning_rate": 1.874292613532028e-05, "loss": 0.9233, "step": 5604 }, { "epoch": 0.19, "learning_rate": 1.874240288179498e-05, "loss": 0.9779, "step": 5605 }, { "epoch": 0.19, "learning_rate": 1.8741879526698002e-05, "loss": 0.9342, "step": 5606 }, { "epoch": 0.19, "learning_rate": 1.8741356070035425e-05, "loss": 0.958, "step": 5607 }, { "epoch": 0.19, "learning_rate": 1.8740832511813325e-05, "loss": 0.9172, "step": 5608 }, { "epoch": 0.19, "learning_rate": 1.8740308852037793e-05, "loss": 0.9799, "step": 5609 }, { "epoch": 0.19, "learning_rate": 1.8739785090714906e-05, "loss": 0.9477, "step": 5610 }, { "epoch": 0.19, "learning_rate": 1.873926122785075e-05, "loss": 0.9957, "step": 5611 }, { "epoch": 0.19, "learning_rate": 1.873873726345142e-05, "loss": 0.9708, "step": 5612 }, { "epoch": 0.19, "learning_rate": 1.8738213197522993e-05, "loss": 0.9639, "step": 5613 }, { "epoch": 0.19, "learning_rate": 1.873768903007156e-05, "loss": 0.9447, "step": 5614 }, { "epoch": 0.19, "learning_rate": 1.8737164761103216e-05, "loss": 0.9443, "step": 5615 }, { "epoch": 0.19, "learning_rate": 1.873664039062405e-05, "loss": 0.9726, "step": 5616 }, { "epoch": 0.19, "learning_rate": 1.873611591864015e-05, "loss": 0.9397, "step": 5617 }, { "epoch": 0.19, "learning_rate": 1.8735591345157614e-05, "loss": 0.9679, "step": 5618 }, { "epoch": 0.19, "learning_rate": 1.8735066670182537e-05, "loss": 0.9595, "step": 5619 }, { "epoch": 0.19, "learning_rate": 1.873454189372101e-05, "loss": 0.9623, "step": 5620 }, { "epoch": 0.19, "learning_rate": 1.8734017015779133e-05, "loss": 0.996, "step": 5621 }, { "epoch": 0.19, "learning_rate": 1.8733492036363007e-05, "loss": 0.9692, "step": 5622 }, { "epoch": 0.19, "learning_rate": 1.8732966955478727e-05, "loss": 1.0135, "step": 5623 }, { "epoch": 0.19, "learning_rate": 1.8732441773132395e-05, "loss": 0.952, "step": 5624 }, { "epoch": 0.19, "learning_rate": 1.8731916489330113e-05, "loss": 0.9895, "step": 5625 }, { "epoch": 0.19, "learning_rate": 1.8731391104077984e-05, "loss": 0.9788, "step": 5626 }, { "epoch": 0.19, "learning_rate": 1.873086561738211e-05, "loss": 0.9653, "step": 5627 }, { "epoch": 0.19, "learning_rate": 1.8730340029248602e-05, "loss": 0.9429, "step": 5628 }, { "epoch": 0.19, "learning_rate": 1.872981433968356e-05, "loss": 0.9989, "step": 5629 }, { "epoch": 0.19, "learning_rate": 1.8729288548693094e-05, "loss": 0.9608, "step": 5630 }, { "epoch": 0.19, "learning_rate": 1.872876265628331e-05, "loss": 1.0191, "step": 5631 }, { "epoch": 0.19, "learning_rate": 1.8728236662460328e-05, "loss": 0.9905, "step": 5632 }, { "epoch": 0.19, "learning_rate": 1.8727710567230247e-05, "loss": 0.9687, "step": 5633 }, { "epoch": 0.19, "learning_rate": 1.8727184370599187e-05, "loss": 0.9496, "step": 5634 }, { "epoch": 0.19, "learning_rate": 1.8726658072573258e-05, "loss": 0.9804, "step": 5635 }, { "epoch": 0.19, "learning_rate": 1.8726131673158576e-05, "loss": 0.9769, "step": 5636 }, { "epoch": 0.19, "learning_rate": 1.8725605172361253e-05, "loss": 0.9813, "step": 5637 }, { "epoch": 0.19, "learning_rate": 1.8725078570187412e-05, "loss": 0.9692, "step": 5638 }, { "epoch": 0.19, "learning_rate": 1.872455186664317e-05, "loss": 0.9568, "step": 5639 }, { "epoch": 0.19, "learning_rate": 1.8724025061734642e-05, "loss": 0.9705, "step": 5640 }, { "epoch": 0.19, "learning_rate": 1.8723498155467953e-05, "loss": 0.9889, "step": 5641 }, { "epoch": 0.19, "learning_rate": 1.8722971147849226e-05, "loss": 0.957, "step": 5642 }, { "epoch": 0.19, "learning_rate": 1.8722444038884578e-05, "loss": 0.9552, "step": 5643 }, { "epoch": 0.19, "learning_rate": 1.8721916828580136e-05, "loss": 0.9325, "step": 5644 }, { "epoch": 0.19, "learning_rate": 1.872138951694203e-05, "loss": 0.9859, "step": 5645 }, { "epoch": 0.19, "learning_rate": 1.8720862103976376e-05, "loss": 0.9858, "step": 5646 }, { "epoch": 0.19, "learning_rate": 1.8720334589689312e-05, "loss": 0.9573, "step": 5647 }, { "epoch": 0.19, "learning_rate": 1.871980697408696e-05, "loss": 0.9684, "step": 5648 }, { "epoch": 0.19, "learning_rate": 1.8719279257175454e-05, "loss": 0.9737, "step": 5649 }, { "epoch": 0.19, "learning_rate": 1.8718751438960924e-05, "loss": 0.9526, "step": 5650 }, { "epoch": 0.19, "learning_rate": 1.87182235194495e-05, "loss": 0.9481, "step": 5651 }, { "epoch": 0.19, "learning_rate": 1.871769549864732e-05, "loss": 0.9654, "step": 5652 }, { "epoch": 0.19, "learning_rate": 1.8717167376560515e-05, "loss": 0.9659, "step": 5653 }, { "epoch": 0.19, "learning_rate": 1.871663915319522e-05, "loss": 0.9698, "step": 5654 }, { "epoch": 0.19, "learning_rate": 1.8716110828557576e-05, "loss": 0.9677, "step": 5655 }, { "epoch": 0.19, "learning_rate": 1.8715582402653722e-05, "loss": 0.9598, "step": 5656 }, { "epoch": 0.19, "learning_rate": 1.8715053875489793e-05, "loss": 0.9838, "step": 5657 }, { "epoch": 0.19, "learning_rate": 1.8714525247071928e-05, "loss": 0.9431, "step": 5658 }, { "epoch": 0.19, "learning_rate": 1.8713996517406278e-05, "loss": 0.9993, "step": 5659 }, { "epoch": 0.19, "learning_rate": 1.8713467686498973e-05, "loss": 0.9317, "step": 5660 }, { "epoch": 0.19, "learning_rate": 1.8712938754356168e-05, "loss": 1.0124, "step": 5661 }, { "epoch": 0.19, "learning_rate": 1.8712409720984008e-05, "loss": 0.9736, "step": 5662 }, { "epoch": 0.19, "learning_rate": 1.871188058638863e-05, "loss": 1.0039, "step": 5663 }, { "epoch": 0.19, "learning_rate": 1.8711351350576187e-05, "loss": 0.9468, "step": 5664 }, { "epoch": 0.19, "learning_rate": 1.8710822013552833e-05, "loss": 0.9205, "step": 5665 }, { "epoch": 0.19, "learning_rate": 1.871029257532471e-05, "loss": 0.9616, "step": 5666 }, { "epoch": 0.19, "learning_rate": 1.8709763035897975e-05, "loss": 0.9376, "step": 5667 }, { "epoch": 0.19, "learning_rate": 1.8709233395278778e-05, "loss": 0.9296, "step": 5668 }, { "epoch": 0.19, "learning_rate": 1.870870365347327e-05, "loss": 0.9985, "step": 5669 }, { "epoch": 0.19, "learning_rate": 1.870817381048761e-05, "loss": 0.9688, "step": 5670 }, { "epoch": 0.19, "learning_rate": 1.870764386632795e-05, "loss": 0.9589, "step": 5671 }, { "epoch": 0.19, "learning_rate": 1.8707113821000448e-05, "loss": 0.9757, "step": 5672 }, { "epoch": 0.19, "learning_rate": 1.8706583674511264e-05, "loss": 0.9451, "step": 5673 }, { "epoch": 0.19, "learning_rate": 1.870605342686656e-05, "loss": 0.9778, "step": 5674 }, { "epoch": 0.19, "learning_rate": 1.8705523078072486e-05, "loss": 0.9877, "step": 5675 }, { "epoch": 0.19, "learning_rate": 1.8704992628135218e-05, "loss": 0.9463, "step": 5676 }, { "epoch": 0.19, "learning_rate": 1.870446207706091e-05, "loss": 0.9476, "step": 5677 }, { "epoch": 0.19, "learning_rate": 1.8703931424855725e-05, "loss": 0.9383, "step": 5678 }, { "epoch": 0.19, "learning_rate": 1.8703400671525835e-05, "loss": 1.0088, "step": 5679 }, { "epoch": 0.19, "learning_rate": 1.87028698170774e-05, "loss": 0.9674, "step": 5680 }, { "epoch": 0.19, "learning_rate": 1.8702338861516587e-05, "loss": 0.9589, "step": 5681 }, { "epoch": 0.19, "learning_rate": 1.870180780484957e-05, "loss": 0.9448, "step": 5682 }, { "epoch": 0.19, "learning_rate": 1.870127664708252e-05, "loss": 0.9562, "step": 5683 }, { "epoch": 0.19, "learning_rate": 1.8700745388221602e-05, "loss": 0.9522, "step": 5684 }, { "epoch": 0.19, "learning_rate": 1.8700214028272993e-05, "loss": 0.9938, "step": 5685 }, { "epoch": 0.19, "learning_rate": 1.8699682567242865e-05, "loss": 0.943, "step": 5686 }, { "epoch": 0.19, "learning_rate": 1.8699151005137392e-05, "loss": 1.0272, "step": 5687 }, { "epoch": 0.19, "learning_rate": 1.869861934196275e-05, "loss": 0.9915, "step": 5688 }, { "epoch": 0.19, "learning_rate": 1.8698087577725114e-05, "loss": 0.9948, "step": 5689 }, { "epoch": 0.19, "learning_rate": 1.869755571243067e-05, "loss": 0.9463, "step": 5690 }, { "epoch": 0.19, "learning_rate": 1.8697023746085587e-05, "loss": 1.0203, "step": 5691 }, { "epoch": 0.19, "learning_rate": 1.8696491678696054e-05, "loss": 0.9568, "step": 5692 }, { "epoch": 0.19, "learning_rate": 1.8695959510268246e-05, "loss": 0.9825, "step": 5693 }, { "epoch": 0.19, "learning_rate": 1.869542724080835e-05, "loss": 1.0107, "step": 5694 }, { "epoch": 0.19, "learning_rate": 1.8694894870322548e-05, "loss": 0.9902, "step": 5695 }, { "epoch": 0.19, "learning_rate": 1.869436239881703e-05, "loss": 0.9505, "step": 5696 }, { "epoch": 0.19, "learning_rate": 1.8693829826297977e-05, "loss": 0.9605, "step": 5697 }, { "epoch": 0.19, "learning_rate": 1.8693297152771575e-05, "loss": 0.965, "step": 5698 }, { "epoch": 0.19, "learning_rate": 1.869276437824402e-05, "loss": 1.0219, "step": 5699 }, { "epoch": 0.19, "learning_rate": 1.86922315027215e-05, "loss": 0.9315, "step": 5700 }, { "epoch": 0.19, "learning_rate": 1.8691698526210196e-05, "loss": 0.9663, "step": 5701 }, { "epoch": 0.19, "learning_rate": 1.8691165448716316e-05, "loss": 1.0077, "step": 5702 }, { "epoch": 0.19, "learning_rate": 1.8690632270246043e-05, "loss": 0.9523, "step": 5703 }, { "epoch": 0.19, "learning_rate": 1.8690098990805576e-05, "loss": 0.9379, "step": 5704 }, { "epoch": 0.19, "learning_rate": 1.8689565610401102e-05, "loss": 0.9526, "step": 5705 }, { "epoch": 0.19, "learning_rate": 1.8689032129038833e-05, "loss": 1.001, "step": 5706 }, { "epoch": 0.19, "learning_rate": 1.8688498546724955e-05, "loss": 0.9748, "step": 5707 }, { "epoch": 0.19, "learning_rate": 1.8687964863465674e-05, "loss": 0.9699, "step": 5708 }, { "epoch": 0.19, "learning_rate": 1.8687431079267188e-05, "loss": 1.0022, "step": 5709 }, { "epoch": 0.19, "learning_rate": 1.8686897194135697e-05, "loss": 0.9863, "step": 5710 }, { "epoch": 0.19, "learning_rate": 1.8686363208077408e-05, "loss": 0.9794, "step": 5711 }, { "epoch": 0.19, "learning_rate": 1.868582912109852e-05, "loss": 0.9313, "step": 5712 }, { "epoch": 0.19, "learning_rate": 1.8685294933205243e-05, "loss": 0.9449, "step": 5713 }, { "epoch": 0.19, "learning_rate": 1.868476064440378e-05, "loss": 0.9826, "step": 5714 }, { "epoch": 0.19, "learning_rate": 1.8684226254700338e-05, "loss": 0.9148, "step": 5715 }, { "epoch": 0.19, "learning_rate": 1.8683691764101128e-05, "loss": 0.9842, "step": 5716 }, { "epoch": 0.19, "learning_rate": 1.8683157172612362e-05, "loss": 0.9922, "step": 5717 }, { "epoch": 0.19, "learning_rate": 1.8682622480240247e-05, "loss": 0.9343, "step": 5718 }, { "epoch": 0.19, "learning_rate": 1.8682087686990994e-05, "loss": 0.972, "step": 5719 }, { "epoch": 0.19, "learning_rate": 1.8681552792870817e-05, "loss": 0.9885, "step": 5720 }, { "epoch": 0.19, "learning_rate": 1.8681017797885937e-05, "loss": 0.9711, "step": 5721 }, { "epoch": 0.19, "learning_rate": 1.8680482702042563e-05, "loss": 0.9562, "step": 5722 }, { "epoch": 0.19, "learning_rate": 1.8679947505346913e-05, "loss": 1.0208, "step": 5723 }, { "epoch": 0.19, "learning_rate": 1.8679412207805207e-05, "loss": 0.917, "step": 5724 }, { "epoch": 0.19, "learning_rate": 1.8678876809423667e-05, "loss": 0.9681, "step": 5725 }, { "epoch": 0.19, "learning_rate": 1.8678341310208506e-05, "loss": 0.9122, "step": 5726 }, { "epoch": 0.19, "learning_rate": 1.8677805710165948e-05, "loss": 1.0044, "step": 5727 }, { "epoch": 0.19, "learning_rate": 1.8677270009302217e-05, "loss": 0.9288, "step": 5728 }, { "epoch": 0.19, "learning_rate": 1.867673420762354e-05, "loss": 0.9933, "step": 5729 }, { "epoch": 0.19, "learning_rate": 1.8676198305136135e-05, "loss": 0.9605, "step": 5730 }, { "epoch": 0.19, "learning_rate": 1.8675662301846236e-05, "loss": 0.947, "step": 5731 }, { "epoch": 0.19, "learning_rate": 1.8675126197760063e-05, "loss": 0.9362, "step": 5732 }, { "epoch": 0.19, "learning_rate": 1.867458999288385e-05, "loss": 0.9729, "step": 5733 }, { "epoch": 0.19, "learning_rate": 1.8674053687223827e-05, "loss": 0.9635, "step": 5734 }, { "epoch": 0.19, "learning_rate": 1.867351728078622e-05, "loss": 0.9522, "step": 5735 }, { "epoch": 0.19, "learning_rate": 1.8672980773577265e-05, "loss": 0.954, "step": 5736 }, { "epoch": 0.19, "learning_rate": 1.8672444165603193e-05, "loss": 0.9681, "step": 5737 }, { "epoch": 0.19, "learning_rate": 1.867190745687024e-05, "loss": 0.9774, "step": 5738 }, { "epoch": 0.19, "learning_rate": 1.8671370647384646e-05, "loss": 0.9635, "step": 5739 }, { "epoch": 0.19, "learning_rate": 1.867083373715264e-05, "loss": 0.9516, "step": 5740 }, { "epoch": 0.19, "learning_rate": 1.8670296726180464e-05, "loss": 0.9481, "step": 5741 }, { "epoch": 0.19, "learning_rate": 1.8669759614474356e-05, "loss": 0.9573, "step": 5742 }, { "epoch": 0.19, "learning_rate": 1.8669222402040557e-05, "loss": 0.9973, "step": 5743 }, { "epoch": 0.19, "learning_rate": 1.866868508888531e-05, "loss": 0.9645, "step": 5744 }, { "epoch": 0.19, "learning_rate": 1.866814767501485e-05, "loss": 0.9726, "step": 5745 }, { "epoch": 0.19, "learning_rate": 1.866761016043543e-05, "loss": 0.9715, "step": 5746 }, { "epoch": 0.19, "learning_rate": 1.8667072545153295e-05, "loss": 1.0132, "step": 5747 }, { "epoch": 0.19, "learning_rate": 1.8666534829174682e-05, "loss": 0.9777, "step": 5748 }, { "epoch": 0.19, "learning_rate": 1.8665997012505847e-05, "loss": 1.0107, "step": 5749 }, { "epoch": 0.19, "learning_rate": 1.8665459095153036e-05, "loss": 0.9592, "step": 5750 }, { "epoch": 0.19, "learning_rate": 1.86649210771225e-05, "loss": 0.9587, "step": 5751 }, { "epoch": 0.19, "learning_rate": 1.8664382958420485e-05, "loss": 0.9996, "step": 5752 }, { "epoch": 0.19, "learning_rate": 1.8663844739053244e-05, "loss": 0.998, "step": 5753 }, { "epoch": 0.19, "learning_rate": 1.8663306419027037e-05, "loss": 0.9203, "step": 5754 }, { "epoch": 0.19, "learning_rate": 1.8662767998348116e-05, "loss": 1.0419, "step": 5755 }, { "epoch": 0.19, "learning_rate": 1.866222947702273e-05, "loss": 0.9333, "step": 5756 }, { "epoch": 0.19, "learning_rate": 1.8661690855057138e-05, "loss": 0.9535, "step": 5757 }, { "epoch": 0.19, "learning_rate": 1.8661152132457605e-05, "loss": 0.9414, "step": 5758 }, { "epoch": 0.19, "learning_rate": 1.866061330923038e-05, "loss": 0.9721, "step": 5759 }, { "epoch": 0.19, "learning_rate": 1.8660074385381732e-05, "loss": 0.9992, "step": 5760 }, { "epoch": 0.19, "learning_rate": 1.865953536091792e-05, "loss": 0.9789, "step": 5761 }, { "epoch": 0.19, "learning_rate": 1.86589962358452e-05, "loss": 0.9763, "step": 5762 }, { "epoch": 0.19, "learning_rate": 1.8658457010169844e-05, "loss": 0.9938, "step": 5763 }, { "epoch": 0.19, "learning_rate": 1.865791768389811e-05, "loss": 0.9711, "step": 5764 }, { "epoch": 0.19, "learning_rate": 1.8657378257036273e-05, "loss": 1.0328, "step": 5765 }, { "epoch": 0.19, "learning_rate": 1.865683872959059e-05, "loss": 0.9608, "step": 5766 }, { "epoch": 0.19, "learning_rate": 1.8656299101567335e-05, "loss": 0.9522, "step": 5767 }, { "epoch": 0.19, "learning_rate": 1.865575937297278e-05, "loss": 0.9554, "step": 5768 }, { "epoch": 0.19, "learning_rate": 1.8655219543813186e-05, "loss": 0.9698, "step": 5769 }, { "epoch": 0.19, "learning_rate": 1.8654679614094837e-05, "loss": 0.9186, "step": 5770 }, { "epoch": 0.19, "learning_rate": 1.8654139583824e-05, "loss": 0.9818, "step": 5771 }, { "epoch": 0.19, "learning_rate": 1.8653599453006947e-05, "loss": 0.9479, "step": 5772 }, { "epoch": 0.19, "learning_rate": 1.865305922164996e-05, "loss": 0.9679, "step": 5773 }, { "epoch": 0.19, "learning_rate": 1.8652518889759306e-05, "loss": 0.959, "step": 5774 }, { "epoch": 0.19, "learning_rate": 1.865197845734127e-05, "loss": 0.9836, "step": 5775 }, { "epoch": 0.19, "learning_rate": 1.865143792440213e-05, "loss": 0.9867, "step": 5776 }, { "epoch": 0.19, "learning_rate": 1.8650897290948167e-05, "loss": 1.0145, "step": 5777 }, { "epoch": 0.19, "learning_rate": 1.8650356556985657e-05, "loss": 0.9626, "step": 5778 }, { "epoch": 0.19, "learning_rate": 1.864981572252089e-05, "loss": 1.0063, "step": 5779 }, { "epoch": 0.19, "learning_rate": 1.864927478756014e-05, "loss": 0.9827, "step": 5780 }, { "epoch": 0.19, "learning_rate": 1.8648733752109702e-05, "loss": 0.9239, "step": 5781 }, { "epoch": 0.19, "learning_rate": 1.864819261617586e-05, "loss": 0.9636, "step": 5782 }, { "epoch": 0.19, "learning_rate": 1.864765137976489e-05, "loss": 0.9611, "step": 5783 }, { "epoch": 0.19, "learning_rate": 1.8647110042883093e-05, "loss": 0.9167, "step": 5784 }, { "epoch": 0.19, "learning_rate": 1.8646568605536754e-05, "loss": 1.0189, "step": 5785 }, { "epoch": 0.19, "learning_rate": 1.864602706773216e-05, "loss": 0.9422, "step": 5786 }, { "epoch": 0.19, "learning_rate": 1.864548542947561e-05, "loss": 1.0089, "step": 5787 }, { "epoch": 0.19, "learning_rate": 1.8644943690773393e-05, "loss": 0.9552, "step": 5788 }, { "epoch": 0.19, "learning_rate": 1.8644401851631797e-05, "loss": 0.9454, "step": 5789 }, { "epoch": 0.19, "learning_rate": 1.864385991205713e-05, "loss": 0.9494, "step": 5790 }, { "epoch": 0.19, "learning_rate": 1.864331787205568e-05, "loss": 0.9498, "step": 5791 }, { "epoch": 0.19, "learning_rate": 1.8642775731633743e-05, "loss": 0.9534, "step": 5792 }, { "epoch": 0.19, "learning_rate": 1.8642233490797625e-05, "loss": 0.9433, "step": 5793 }, { "epoch": 0.19, "learning_rate": 1.864169114955362e-05, "loss": 0.9595, "step": 5794 }, { "epoch": 0.19, "learning_rate": 1.864114870790803e-05, "loss": 1.011, "step": 5795 }, { "epoch": 0.19, "learning_rate": 1.864060616586716e-05, "loss": 0.9224, "step": 5796 }, { "epoch": 0.19, "learning_rate": 1.864006352343731e-05, "loss": 0.9441, "step": 5797 }, { "epoch": 0.19, "learning_rate": 1.863952078062479e-05, "loss": 0.9368, "step": 5798 }, { "epoch": 0.19, "learning_rate": 1.86389779374359e-05, "loss": 0.9799, "step": 5799 }, { "epoch": 0.19, "learning_rate": 1.863843499387695e-05, "loss": 0.9621, "step": 5800 }, { "epoch": 0.19, "learning_rate": 1.8637891949954246e-05, "loss": 1.0015, "step": 5801 }, { "epoch": 0.19, "learning_rate": 1.8637348805674097e-05, "loss": 0.9823, "step": 5802 }, { "epoch": 0.19, "learning_rate": 1.8636805561042814e-05, "loss": 0.9553, "step": 5803 }, { "epoch": 0.19, "learning_rate": 1.863626221606671e-05, "loss": 0.9443, "step": 5804 }, { "epoch": 0.19, "learning_rate": 1.86357187707521e-05, "loss": 0.9382, "step": 5805 }, { "epoch": 0.19, "learning_rate": 1.863517522510529e-05, "loss": 0.9728, "step": 5806 }, { "epoch": 0.19, "learning_rate": 1.8634631579132602e-05, "loss": 0.9715, "step": 5807 }, { "epoch": 0.19, "learning_rate": 1.8634087832840347e-05, "loss": 0.9711, "step": 5808 }, { "epoch": 0.19, "learning_rate": 1.863354398623485e-05, "loss": 1.0136, "step": 5809 }, { "epoch": 0.19, "learning_rate": 1.8633000039322423e-05, "loss": 0.9269, "step": 5810 }, { "epoch": 0.19, "learning_rate": 1.8632455992109388e-05, "loss": 0.9781, "step": 5811 }, { "epoch": 0.19, "learning_rate": 1.8631911844602065e-05, "loss": 0.9664, "step": 5812 }, { "epoch": 0.19, "learning_rate": 1.863136759680678e-05, "loss": 0.9478, "step": 5813 }, { "epoch": 0.19, "learning_rate": 1.8630823248729847e-05, "loss": 0.9892, "step": 5814 }, { "epoch": 0.19, "learning_rate": 1.86302788003776e-05, "loss": 0.9692, "step": 5815 }, { "epoch": 0.19, "learning_rate": 1.862973425175636e-05, "loss": 0.9599, "step": 5816 }, { "epoch": 0.19, "learning_rate": 1.8629189602872455e-05, "loss": 0.9627, "step": 5817 }, { "epoch": 0.19, "learning_rate": 1.8628644853732212e-05, "loss": 0.9375, "step": 5818 }, { "epoch": 0.19, "learning_rate": 1.862810000434196e-05, "loss": 0.9562, "step": 5819 }, { "epoch": 0.19, "learning_rate": 1.8627555054708033e-05, "loss": 0.9727, "step": 5820 }, { "epoch": 0.19, "learning_rate": 1.8627010004836755e-05, "loss": 0.951, "step": 5821 }, { "epoch": 0.19, "learning_rate": 1.8626464854734466e-05, "loss": 0.9444, "step": 5822 }, { "epoch": 0.19, "learning_rate": 1.8625919604407494e-05, "loss": 0.9658, "step": 5823 }, { "epoch": 0.19, "learning_rate": 1.8625374253862176e-05, "loss": 1.0031, "step": 5824 }, { "epoch": 0.19, "learning_rate": 1.862482880310485e-05, "loss": 0.981, "step": 5825 }, { "epoch": 0.19, "learning_rate": 1.862428325214185e-05, "loss": 0.9838, "step": 5826 }, { "epoch": 0.19, "learning_rate": 1.8623737600979516e-05, "loss": 0.9624, "step": 5827 }, { "epoch": 0.19, "learning_rate": 1.862319184962419e-05, "loss": 0.9389, "step": 5828 }, { "epoch": 0.19, "learning_rate": 1.862264599808221e-05, "loss": 0.994, "step": 5829 }, { "epoch": 0.19, "learning_rate": 1.8622100046359915e-05, "loss": 0.9732, "step": 5830 }, { "epoch": 0.19, "learning_rate": 1.862155399446365e-05, "loss": 0.9929, "step": 5831 }, { "epoch": 0.19, "learning_rate": 1.862100784239976e-05, "loss": 0.9619, "step": 5832 }, { "epoch": 0.19, "learning_rate": 1.8620461590174596e-05, "loss": 0.9408, "step": 5833 }, { "epoch": 0.19, "learning_rate": 1.8619915237794494e-05, "loss": 0.945, "step": 5834 }, { "epoch": 0.19, "learning_rate": 1.8619368785265806e-05, "loss": 0.96, "step": 5835 }, { "epoch": 0.19, "learning_rate": 1.8618822232594884e-05, "loss": 0.9202, "step": 5836 }, { "epoch": 0.19, "learning_rate": 1.8618275579788075e-05, "loss": 0.9489, "step": 5837 }, { "epoch": 0.19, "learning_rate": 1.861772882685173e-05, "loss": 0.9692, "step": 5838 }, { "epoch": 0.19, "learning_rate": 1.8617181973792204e-05, "loss": 0.9675, "step": 5839 }, { "epoch": 0.19, "learning_rate": 1.8616635020615846e-05, "loss": 0.971, "step": 5840 }, { "epoch": 0.19, "learning_rate": 1.8616087967329014e-05, "loss": 0.9866, "step": 5841 }, { "epoch": 0.19, "learning_rate": 1.8615540813938063e-05, "loss": 0.9696, "step": 5842 }, { "epoch": 0.19, "learning_rate": 1.861499356044935e-05, "loss": 0.9929, "step": 5843 }, { "epoch": 0.19, "learning_rate": 1.8614446206869232e-05, "loss": 0.9165, "step": 5844 }, { "epoch": 0.19, "learning_rate": 1.8613898753204068e-05, "loss": 0.9458, "step": 5845 }, { "epoch": 0.19, "learning_rate": 1.8613351199460226e-05, "loss": 0.9677, "step": 5846 }, { "epoch": 0.19, "learning_rate": 1.8612803545644056e-05, "loss": 1.0132, "step": 5847 }, { "epoch": 0.19, "learning_rate": 1.8612255791761925e-05, "loss": 0.9313, "step": 5848 }, { "epoch": 0.19, "learning_rate": 1.86117079378202e-05, "loss": 0.9856, "step": 5849 }, { "epoch": 0.19, "learning_rate": 1.8611159983825248e-05, "loss": 0.9336, "step": 5850 }, { "epoch": 0.19, "learning_rate": 1.8610611929783432e-05, "loss": 0.9687, "step": 5851 }, { "epoch": 0.19, "learning_rate": 1.8610063775701113e-05, "loss": 0.9478, "step": 5852 }, { "epoch": 0.19, "learning_rate": 1.860951552158467e-05, "loss": 0.9382, "step": 5853 }, { "epoch": 0.19, "learning_rate": 1.8608967167440467e-05, "loss": 0.9655, "step": 5854 }, { "epoch": 0.19, "learning_rate": 1.8608418713274873e-05, "loss": 0.9763, "step": 5855 }, { "epoch": 0.19, "learning_rate": 1.860787015909427e-05, "loss": 0.959, "step": 5856 }, { "epoch": 0.19, "learning_rate": 1.860732150490502e-05, "loss": 0.9944, "step": 5857 }, { "epoch": 0.19, "learning_rate": 1.8606772750713503e-05, "loss": 0.9899, "step": 5858 }, { "epoch": 0.19, "learning_rate": 1.8606223896526097e-05, "loss": 1.0061, "step": 5859 }, { "epoch": 0.2, "learning_rate": 1.860567494234917e-05, "loss": 0.9936, "step": 5860 }, { "epoch": 0.2, "learning_rate": 1.860512588818911e-05, "loss": 0.9684, "step": 5861 }, { "epoch": 0.2, "learning_rate": 1.860457673405229e-05, "loss": 0.9779, "step": 5862 }, { "epoch": 0.2, "learning_rate": 1.8604027479945094e-05, "loss": 0.9372, "step": 5863 }, { "epoch": 0.2, "learning_rate": 1.8603478125873896e-05, "loss": 0.9737, "step": 5864 }, { "epoch": 0.2, "learning_rate": 1.8602928671845088e-05, "loss": 0.9631, "step": 5865 }, { "epoch": 0.2, "learning_rate": 1.8602379117865044e-05, "loss": 1.0053, "step": 5866 }, { "epoch": 0.2, "learning_rate": 1.8601829463940157e-05, "loss": 0.9705, "step": 5867 }, { "epoch": 0.2, "learning_rate": 1.860127971007681e-05, "loss": 0.9252, "step": 5868 }, { "epoch": 0.2, "learning_rate": 1.8600729856281394e-05, "loss": 0.9876, "step": 5869 }, { "epoch": 0.2, "learning_rate": 1.8600179902560288e-05, "loss": 1.0039, "step": 5870 }, { "epoch": 0.2, "learning_rate": 1.8599629848919893e-05, "loss": 1.0105, "step": 5871 }, { "epoch": 0.2, "learning_rate": 1.859907969536659e-05, "loss": 0.9326, "step": 5872 }, { "epoch": 0.2, "learning_rate": 1.8598529441906776e-05, "loss": 0.9784, "step": 5873 }, { "epoch": 0.2, "learning_rate": 1.8597979088546842e-05, "loss": 1.0026, "step": 5874 }, { "epoch": 0.2, "learning_rate": 1.8597428635293185e-05, "loss": 0.9129, "step": 5875 }, { "epoch": 0.2, "learning_rate": 1.8596878082152195e-05, "loss": 0.9645, "step": 5876 }, { "epoch": 0.2, "learning_rate": 1.8596327429130273e-05, "loss": 0.9544, "step": 5877 }, { "epoch": 0.2, "learning_rate": 1.859577667623382e-05, "loss": 0.9704, "step": 5878 }, { "epoch": 0.2, "learning_rate": 1.8595225823469225e-05, "loss": 1.0134, "step": 5879 }, { "epoch": 0.2, "learning_rate": 1.8594674870842893e-05, "loss": 0.9348, "step": 5880 }, { "epoch": 0.2, "learning_rate": 1.8594123818361226e-05, "loss": 0.9652, "step": 5881 }, { "epoch": 0.2, "learning_rate": 1.8593572666030627e-05, "loss": 1.0283, "step": 5882 }, { "epoch": 0.2, "learning_rate": 1.8593021413857497e-05, "loss": 0.9752, "step": 5883 }, { "epoch": 0.2, "learning_rate": 1.859247006184824e-05, "loss": 1.0195, "step": 5884 }, { "epoch": 0.2, "learning_rate": 1.8591918610009266e-05, "loss": 0.9992, "step": 5885 }, { "epoch": 0.2, "learning_rate": 1.859136705834698e-05, "loss": 0.9526, "step": 5886 }, { "epoch": 0.2, "learning_rate": 1.8590815406867786e-05, "loss": 0.9565, "step": 5887 }, { "epoch": 0.2, "learning_rate": 1.85902636555781e-05, "loss": 0.9384, "step": 5888 }, { "epoch": 0.2, "learning_rate": 1.8589711804484327e-05, "loss": 1.01, "step": 5889 }, { "epoch": 0.2, "learning_rate": 1.858915985359288e-05, "loss": 0.9804, "step": 5890 }, { "epoch": 0.2, "learning_rate": 1.8588607802910174e-05, "loss": 0.9678, "step": 5891 }, { "epoch": 0.2, "learning_rate": 1.858805565244262e-05, "loss": 0.9628, "step": 5892 }, { "epoch": 0.2, "learning_rate": 1.8587503402196633e-05, "loss": 0.9656, "step": 5893 }, { "epoch": 0.2, "learning_rate": 1.858695105217863e-05, "loss": 0.989, "step": 5894 }, { "epoch": 0.2, "learning_rate": 1.8586398602395033e-05, "loss": 0.9946, "step": 5895 }, { "epoch": 0.2, "learning_rate": 1.8585846052852254e-05, "loss": 0.9666, "step": 5896 }, { "epoch": 0.2, "learning_rate": 1.8585293403556715e-05, "loss": 0.9588, "step": 5897 }, { "epoch": 0.2, "learning_rate": 1.8584740654514838e-05, "loss": 0.9975, "step": 5898 }, { "epoch": 0.2, "learning_rate": 1.858418780573304e-05, "loss": 1.0261, "step": 5899 }, { "epoch": 0.2, "learning_rate": 1.8583634857217752e-05, "loss": 0.9444, "step": 5900 }, { "epoch": 0.2, "learning_rate": 1.858308180897539e-05, "loss": 0.9816, "step": 5901 }, { "epoch": 0.2, "learning_rate": 1.8582528661012386e-05, "loss": 0.9835, "step": 5902 }, { "epoch": 0.2, "learning_rate": 1.8581975413335166e-05, "loss": 0.9717, "step": 5903 }, { "epoch": 0.2, "learning_rate": 1.8581422065950153e-05, "loss": 0.9613, "step": 5904 }, { "epoch": 0.2, "learning_rate": 1.8580868618863782e-05, "loss": 1.0022, "step": 5905 }, { "epoch": 0.2, "learning_rate": 1.858031507208248e-05, "loss": 0.9337, "step": 5906 }, { "epoch": 0.2, "learning_rate": 1.8579761425612678e-05, "loss": 0.9935, "step": 5907 }, { "epoch": 0.2, "learning_rate": 1.8579207679460807e-05, "loss": 0.9492, "step": 5908 }, { "epoch": 0.2, "learning_rate": 1.8578653833633303e-05, "loss": 0.9732, "step": 5909 }, { "epoch": 0.2, "learning_rate": 1.85780998881366e-05, "loss": 0.9526, "step": 5910 }, { "epoch": 0.2, "learning_rate": 1.8577545842977136e-05, "loss": 0.985, "step": 5911 }, { "epoch": 0.2, "learning_rate": 1.8576991698161344e-05, "loss": 0.945, "step": 5912 }, { "epoch": 0.2, "learning_rate": 1.8576437453695666e-05, "loss": 0.9742, "step": 5913 }, { "epoch": 0.2, "learning_rate": 1.8575883109586537e-05, "loss": 0.9254, "step": 5914 }, { "epoch": 0.2, "learning_rate": 1.8575328665840407e-05, "loss": 0.952, "step": 5915 }, { "epoch": 0.2, "learning_rate": 1.8574774122463704e-05, "loss": 0.9723, "step": 5916 }, { "epoch": 0.2, "learning_rate": 1.857421947946288e-05, "loss": 1.0256, "step": 5917 }, { "epoch": 0.2, "learning_rate": 1.8573664736844376e-05, "loss": 0.9911, "step": 5918 }, { "epoch": 0.2, "learning_rate": 1.8573109894614638e-05, "loss": 0.985, "step": 5919 }, { "epoch": 0.2, "learning_rate": 1.857255495278011e-05, "loss": 0.9755, "step": 5920 }, { "epoch": 0.2, "learning_rate": 1.8571999911347246e-05, "loss": 0.986, "step": 5921 }, { "epoch": 0.2, "learning_rate": 1.857144477032249e-05, "loss": 0.9493, "step": 5922 }, { "epoch": 0.2, "learning_rate": 1.857088952971229e-05, "loss": 0.9649, "step": 5923 }, { "epoch": 0.2, "learning_rate": 1.8570334189523097e-05, "loss": 0.9655, "step": 5924 }, { "epoch": 0.2, "learning_rate": 1.856977874976137e-05, "loss": 0.9527, "step": 5925 }, { "epoch": 0.2, "learning_rate": 1.8569223210433557e-05, "loss": 0.9596, "step": 5926 }, { "epoch": 0.2, "learning_rate": 1.856866757154611e-05, "loss": 1.0001, "step": 5927 }, { "epoch": 0.2, "learning_rate": 1.8568111833105485e-05, "loss": 0.9267, "step": 5928 }, { "epoch": 0.2, "learning_rate": 1.8567555995118147e-05, "loss": 0.9452, "step": 5929 }, { "epoch": 0.2, "learning_rate": 1.8567000057590546e-05, "loss": 0.9893, "step": 5930 }, { "epoch": 0.2, "learning_rate": 1.856644402052914e-05, "loss": 0.9859, "step": 5931 }, { "epoch": 0.2, "learning_rate": 1.8565887883940394e-05, "loss": 0.9993, "step": 5932 }, { "epoch": 0.2, "learning_rate": 1.856533164783077e-05, "loss": 0.9624, "step": 5933 }, { "epoch": 0.2, "learning_rate": 1.8564775312206726e-05, "loss": 0.9321, "step": 5934 }, { "epoch": 0.2, "learning_rate": 1.8564218877074726e-05, "loss": 0.9922, "step": 5935 }, { "epoch": 0.2, "learning_rate": 1.8563662342441237e-05, "loss": 0.9222, "step": 5936 }, { "epoch": 0.2, "learning_rate": 1.8563105708312727e-05, "loss": 0.9515, "step": 5937 }, { "epoch": 0.2, "learning_rate": 1.856254897469566e-05, "loss": 0.9912, "step": 5938 }, { "epoch": 0.2, "learning_rate": 1.8561992141596503e-05, "loss": 0.9802, "step": 5939 }, { "epoch": 0.2, "learning_rate": 1.856143520902173e-05, "loss": 0.9459, "step": 5940 }, { "epoch": 0.2, "learning_rate": 1.8560878176977807e-05, "loss": 0.9455, "step": 5941 }, { "epoch": 0.2, "learning_rate": 1.856032104547121e-05, "loss": 0.9872, "step": 5942 }, { "epoch": 0.2, "learning_rate": 1.8559763814508404e-05, "loss": 0.9785, "step": 5943 }, { "epoch": 0.2, "learning_rate": 1.8559206484095877e-05, "loss": 0.9378, "step": 5944 }, { "epoch": 0.2, "learning_rate": 1.8558649054240086e-05, "loss": 0.9587, "step": 5945 }, { "epoch": 0.2, "learning_rate": 1.8558091524947527e-05, "loss": 0.958, "step": 5946 }, { "epoch": 0.2, "learning_rate": 1.8557533896224664e-05, "loss": 0.9759, "step": 5947 }, { "epoch": 0.2, "learning_rate": 1.8556976168077977e-05, "loss": 0.9878, "step": 5948 }, { "epoch": 0.2, "learning_rate": 1.855641834051395e-05, "loss": 1.0097, "step": 5949 }, { "epoch": 0.2, "learning_rate": 1.8555860413539063e-05, "loss": 0.9264, "step": 5950 }, { "epoch": 0.2, "learning_rate": 1.85553023871598e-05, "loss": 0.9894, "step": 5951 }, { "epoch": 0.2, "learning_rate": 1.8554744261382635e-05, "loss": 0.9349, "step": 5952 }, { "epoch": 0.2, "learning_rate": 1.8554186036214065e-05, "loss": 0.9703, "step": 5953 }, { "epoch": 0.2, "learning_rate": 1.855362771166057e-05, "loss": 0.9424, "step": 5954 }, { "epoch": 0.2, "learning_rate": 1.8553069287728634e-05, "loss": 0.9462, "step": 5955 }, { "epoch": 0.2, "learning_rate": 1.8552510764424748e-05, "loss": 0.9289, "step": 5956 }, { "epoch": 0.2, "learning_rate": 1.85519521417554e-05, "loss": 0.969, "step": 5957 }, { "epoch": 0.2, "learning_rate": 1.8551393419727083e-05, "loss": 0.9508, "step": 5958 }, { "epoch": 0.2, "learning_rate": 1.8550834598346285e-05, "loss": 0.9685, "step": 5959 }, { "epoch": 0.2, "learning_rate": 1.85502756776195e-05, "loss": 0.8893, "step": 5960 }, { "epoch": 0.2, "learning_rate": 1.8549716657553222e-05, "loss": 0.9451, "step": 5961 }, { "epoch": 0.2, "learning_rate": 1.8549157538153944e-05, "loss": 0.9684, "step": 5962 }, { "epoch": 0.2, "learning_rate": 1.8548598319428164e-05, "loss": 0.9717, "step": 5963 }, { "epoch": 0.2, "learning_rate": 1.8548039001382377e-05, "loss": 0.9574, "step": 5964 }, { "epoch": 0.2, "learning_rate": 1.8547479584023085e-05, "loss": 0.9503, "step": 5965 }, { "epoch": 0.2, "learning_rate": 1.8546920067356783e-05, "loss": 0.9757, "step": 5966 }, { "epoch": 0.2, "learning_rate": 1.8546360451389976e-05, "loss": 0.983, "step": 5967 }, { "epoch": 0.2, "learning_rate": 1.8545800736129166e-05, "loss": 0.9302, "step": 5968 }, { "epoch": 0.2, "learning_rate": 1.8545240921580848e-05, "loss": 0.9553, "step": 5969 }, { "epoch": 0.2, "learning_rate": 1.8544681007751538e-05, "loss": 0.9577, "step": 5970 }, { "epoch": 0.2, "learning_rate": 1.854412099464773e-05, "loss": 0.9223, "step": 5971 }, { "epoch": 0.2, "learning_rate": 1.854356088227594e-05, "loss": 0.9229, "step": 5972 }, { "epoch": 0.2, "learning_rate": 1.8543000670642664e-05, "loss": 0.9476, "step": 5973 }, { "epoch": 0.2, "learning_rate": 1.8542440359754422e-05, "loss": 0.9499, "step": 5974 }, { "epoch": 0.2, "learning_rate": 1.854187994961772e-05, "loss": 0.9735, "step": 5975 }, { "epoch": 0.2, "learning_rate": 1.8541319440239066e-05, "loss": 0.9284, "step": 5976 }, { "epoch": 0.2, "learning_rate": 1.854075883162498e-05, "loss": 0.9633, "step": 5977 }, { "epoch": 0.2, "learning_rate": 1.8540198123781966e-05, "loss": 0.955, "step": 5978 }, { "epoch": 0.2, "learning_rate": 1.8539637316716538e-05, "loss": 0.983, "step": 5979 }, { "epoch": 0.2, "learning_rate": 1.8539076410435223e-05, "loss": 0.978, "step": 5980 }, { "epoch": 0.2, "learning_rate": 1.8538515404944527e-05, "loss": 0.9762, "step": 5981 }, { "epoch": 0.2, "learning_rate": 1.8537954300250973e-05, "loss": 0.9202, "step": 5982 }, { "epoch": 0.2, "learning_rate": 1.8537393096361078e-05, "loss": 0.9654, "step": 5983 }, { "epoch": 0.2, "learning_rate": 1.853683179328136e-05, "loss": 0.9708, "step": 5984 }, { "epoch": 0.2, "learning_rate": 1.8536270391018346e-05, "loss": 0.9473, "step": 5985 }, { "epoch": 0.2, "learning_rate": 1.853570888957856e-05, "loss": 0.9342, "step": 5986 }, { "epoch": 0.2, "learning_rate": 1.8535147288968515e-05, "loss": 0.9779, "step": 5987 }, { "epoch": 0.2, "learning_rate": 1.8534585589194745e-05, "loss": 0.9605, "step": 5988 }, { "epoch": 0.2, "learning_rate": 1.8534023790263775e-05, "loss": 0.9668, "step": 5989 }, { "epoch": 0.2, "learning_rate": 1.853346189218213e-05, "loss": 0.9477, "step": 5990 }, { "epoch": 0.2, "learning_rate": 1.8532899894956337e-05, "loss": 0.9859, "step": 5991 }, { "epoch": 0.2, "learning_rate": 1.853233779859293e-05, "loss": 0.9402, "step": 5992 }, { "epoch": 0.2, "learning_rate": 1.8531775603098435e-05, "loss": 0.9993, "step": 5993 }, { "epoch": 0.2, "learning_rate": 1.8531213308479383e-05, "loss": 0.9911, "step": 5994 }, { "epoch": 0.2, "learning_rate": 1.8530650914742313e-05, "loss": 0.9971, "step": 5995 }, { "epoch": 0.2, "learning_rate": 1.8530088421893756e-05, "loss": 0.942, "step": 5996 }, { "epoch": 0.2, "learning_rate": 1.8529525829940244e-05, "loss": 0.961, "step": 5997 }, { "epoch": 0.2, "learning_rate": 1.8528963138888315e-05, "loss": 0.9791, "step": 5998 }, { "epoch": 0.2, "learning_rate": 1.8528400348744512e-05, "loss": 0.9748, "step": 5999 }, { "epoch": 0.2, "learning_rate": 1.8527837459515364e-05, "loss": 0.9382, "step": 6000 }, { "epoch": 0.2, "learning_rate": 1.852727447120742e-05, "loss": 0.963, "step": 6001 }, { "epoch": 0.2, "learning_rate": 1.8526711383827217e-05, "loss": 0.9474, "step": 6002 }, { "epoch": 0.2, "learning_rate": 1.8526148197381296e-05, "loss": 0.9868, "step": 6003 }, { "epoch": 0.2, "learning_rate": 1.85255849118762e-05, "loss": 0.9413, "step": 6004 }, { "epoch": 0.2, "learning_rate": 1.8525021527318478e-05, "loss": 0.9745, "step": 6005 }, { "epoch": 0.2, "learning_rate": 1.852445804371467e-05, "loss": 0.9617, "step": 6006 }, { "epoch": 0.2, "learning_rate": 1.8523894461071323e-05, "loss": 0.9685, "step": 6007 }, { "epoch": 0.2, "learning_rate": 1.852333077939499e-05, "loss": 0.953, "step": 6008 }, { "epoch": 0.2, "learning_rate": 1.8522766998692216e-05, "loss": 1.01, "step": 6009 }, { "epoch": 0.2, "learning_rate": 1.852220311896955e-05, "loss": 0.9737, "step": 6010 }, { "epoch": 0.2, "learning_rate": 1.852163914023355e-05, "loss": 0.9613, "step": 6011 }, { "epoch": 0.2, "learning_rate": 1.852107506249076e-05, "loss": 0.9609, "step": 6012 }, { "epoch": 0.2, "learning_rate": 1.8520510885747737e-05, "loss": 0.9424, "step": 6013 }, { "epoch": 0.2, "learning_rate": 1.851994661001104e-05, "loss": 0.9783, "step": 6014 }, { "epoch": 0.2, "learning_rate": 1.8519382235287218e-05, "loss": 0.9885, "step": 6015 }, { "epoch": 0.2, "learning_rate": 1.8518817761582833e-05, "loss": 0.937, "step": 6016 }, { "epoch": 0.2, "learning_rate": 1.8518253188904437e-05, "loss": 1.0065, "step": 6017 }, { "epoch": 0.2, "learning_rate": 1.8517688517258597e-05, "loss": 0.9753, "step": 6018 }, { "epoch": 0.2, "learning_rate": 1.851712374665187e-05, "loss": 0.9831, "step": 6019 }, { "epoch": 0.2, "learning_rate": 1.851655887709082e-05, "loss": 0.9299, "step": 6020 }, { "epoch": 0.2, "learning_rate": 1.8515993908582005e-05, "loss": 0.9998, "step": 6021 }, { "epoch": 0.2, "learning_rate": 1.8515428841131996e-05, "loss": 0.9631, "step": 6022 }, { "epoch": 0.2, "learning_rate": 1.8514863674747347e-05, "loss": 0.9511, "step": 6023 }, { "epoch": 0.2, "learning_rate": 1.8514298409434633e-05, "loss": 0.9663, "step": 6024 }, { "epoch": 0.2, "learning_rate": 1.8513733045200425e-05, "loss": 0.9705, "step": 6025 }, { "epoch": 0.2, "learning_rate": 1.851316758205128e-05, "loss": 0.9581, "step": 6026 }, { "epoch": 0.2, "learning_rate": 1.8512602019993777e-05, "loss": 0.9342, "step": 6027 }, { "epoch": 0.2, "learning_rate": 1.851203635903448e-05, "loss": 0.944, "step": 6028 }, { "epoch": 0.2, "learning_rate": 1.8511470599179972e-05, "loss": 0.9773, "step": 6029 }, { "epoch": 0.2, "learning_rate": 1.851090474043681e-05, "loss": 0.9474, "step": 6030 }, { "epoch": 0.2, "learning_rate": 1.8510338782811582e-05, "loss": 0.9837, "step": 6031 }, { "epoch": 0.2, "learning_rate": 1.8509772726310857e-05, "loss": 0.91, "step": 6032 }, { "epoch": 0.2, "learning_rate": 1.8509206570941215e-05, "loss": 0.9473, "step": 6033 }, { "epoch": 0.2, "learning_rate": 1.850864031670923e-05, "loss": 0.9579, "step": 6034 }, { "epoch": 0.2, "learning_rate": 1.8508073963621482e-05, "loss": 0.9455, "step": 6035 }, { "epoch": 0.2, "learning_rate": 1.850750751168455e-05, "loss": 0.9358, "step": 6036 }, { "epoch": 0.2, "learning_rate": 1.850694096090502e-05, "loss": 0.9862, "step": 6037 }, { "epoch": 0.2, "learning_rate": 1.8506374311289472e-05, "loss": 0.913, "step": 6038 }, { "epoch": 0.2, "learning_rate": 1.850580756284449e-05, "loss": 0.9927, "step": 6039 }, { "epoch": 0.2, "learning_rate": 1.8505240715576656e-05, "loss": 0.9755, "step": 6040 }, { "epoch": 0.2, "learning_rate": 1.8504673769492555e-05, "loss": 0.9607, "step": 6041 }, { "epoch": 0.2, "learning_rate": 1.8504106724598778e-05, "loss": 0.9712, "step": 6042 }, { "epoch": 0.2, "learning_rate": 1.8503539580901914e-05, "loss": 0.9739, "step": 6043 }, { "epoch": 0.2, "learning_rate": 1.8502972338408547e-05, "loss": 0.9575, "step": 6044 }, { "epoch": 0.2, "learning_rate": 1.850240499712527e-05, "loss": 0.9599, "step": 6045 }, { "epoch": 0.2, "learning_rate": 1.8501837557058673e-05, "loss": 0.9403, "step": 6046 }, { "epoch": 0.2, "learning_rate": 1.8501270018215352e-05, "loss": 0.9748, "step": 6047 }, { "epoch": 0.2, "learning_rate": 1.8500702380601898e-05, "loss": 0.9497, "step": 6048 }, { "epoch": 0.2, "learning_rate": 1.8500134644224907e-05, "loss": 0.9576, "step": 6049 }, { "epoch": 0.2, "learning_rate": 1.8499566809090976e-05, "loss": 0.9684, "step": 6050 }, { "epoch": 0.2, "learning_rate": 1.84989988752067e-05, "loss": 0.9469, "step": 6051 }, { "epoch": 0.2, "learning_rate": 1.849843084257868e-05, "loss": 0.9621, "step": 6052 }, { "epoch": 0.2, "learning_rate": 1.8497862711213515e-05, "loss": 0.9882, "step": 6053 }, { "epoch": 0.2, "learning_rate": 1.84972944811178e-05, "loss": 0.9504, "step": 6054 }, { "epoch": 0.2, "learning_rate": 1.8496726152298147e-05, "loss": 0.9706, "step": 6055 }, { "epoch": 0.2, "learning_rate": 1.8496157724761156e-05, "loss": 0.9807, "step": 6056 }, { "epoch": 0.2, "learning_rate": 1.8495589198513423e-05, "loss": 0.9921, "step": 6057 }, { "epoch": 0.2, "learning_rate": 1.8495020573561563e-05, "loss": 0.9799, "step": 6058 }, { "epoch": 0.2, "learning_rate": 1.8494451849912178e-05, "loss": 0.9548, "step": 6059 }, { "epoch": 0.2, "learning_rate": 1.8493883027571877e-05, "loss": 0.9531, "step": 6060 }, { "epoch": 0.2, "learning_rate": 1.8493314106547268e-05, "loss": 0.974, "step": 6061 }, { "epoch": 0.2, "learning_rate": 1.849274508684496e-05, "loss": 0.9716, "step": 6062 }, { "epoch": 0.2, "learning_rate": 1.8492175968471563e-05, "loss": 0.9694, "step": 6063 }, { "epoch": 0.2, "learning_rate": 1.8491606751433692e-05, "loss": 0.9518, "step": 6064 }, { "epoch": 0.2, "learning_rate": 1.8491037435737964e-05, "loss": 0.9729, "step": 6065 }, { "epoch": 0.2, "learning_rate": 1.8490468021390985e-05, "loss": 0.9649, "step": 6066 }, { "epoch": 0.2, "learning_rate": 1.848989850839937e-05, "loss": 0.94, "step": 6067 }, { "epoch": 0.2, "learning_rate": 1.848932889676975e-05, "loss": 0.996, "step": 6068 }, { "epoch": 0.2, "learning_rate": 1.848875918650873e-05, "loss": 0.9428, "step": 6069 }, { "epoch": 0.2, "learning_rate": 1.8488189377622928e-05, "loss": 0.9394, "step": 6070 }, { "epoch": 0.2, "learning_rate": 1.8487619470118974e-05, "loss": 0.9865, "step": 6071 }, { "epoch": 0.2, "learning_rate": 1.848704946400348e-05, "loss": 0.9811, "step": 6072 }, { "epoch": 0.2, "learning_rate": 1.8486479359283078e-05, "loss": 0.9555, "step": 6073 }, { "epoch": 0.2, "learning_rate": 1.848590915596438e-05, "loss": 0.9594, "step": 6074 }, { "epoch": 0.2, "learning_rate": 1.8485338854054018e-05, "loss": 0.9562, "step": 6075 }, { "epoch": 0.2, "learning_rate": 1.8484768453558617e-05, "loss": 0.9559, "step": 6076 }, { "epoch": 0.2, "learning_rate": 1.8484197954484803e-05, "loss": 0.952, "step": 6077 }, { "epoch": 0.2, "learning_rate": 1.8483627356839206e-05, "loss": 0.9426, "step": 6078 }, { "epoch": 0.2, "learning_rate": 1.8483056660628455e-05, "loss": 0.9645, "step": 6079 }, { "epoch": 0.2, "learning_rate": 1.848248586585918e-05, "loss": 0.9539, "step": 6080 }, { "epoch": 0.2, "learning_rate": 1.848191497253801e-05, "loss": 0.9813, "step": 6081 }, { "epoch": 0.2, "learning_rate": 1.8481343980671585e-05, "loss": 0.9649, "step": 6082 }, { "epoch": 0.2, "learning_rate": 1.848077289026653e-05, "loss": 0.9721, "step": 6083 }, { "epoch": 0.2, "learning_rate": 1.8480201701329483e-05, "loss": 0.9573, "step": 6084 }, { "epoch": 0.2, "learning_rate": 1.8479630413867085e-05, "loss": 1.0008, "step": 6085 }, { "epoch": 0.2, "learning_rate": 1.847905902788597e-05, "loss": 0.9333, "step": 6086 }, { "epoch": 0.2, "learning_rate": 1.8478487543392775e-05, "loss": 0.9684, "step": 6087 }, { "epoch": 0.2, "learning_rate": 1.8477915960394144e-05, "loss": 0.9353, "step": 6088 }, { "epoch": 0.2, "learning_rate": 1.8477344278896708e-05, "loss": 0.9718, "step": 6089 }, { "epoch": 0.2, "learning_rate": 1.847677249890712e-05, "loss": 0.981, "step": 6090 }, { "epoch": 0.2, "learning_rate": 1.8476200620432024e-05, "loss": 0.9894, "step": 6091 }, { "epoch": 0.2, "learning_rate": 1.8475628643478055e-05, "loss": 0.9335, "step": 6092 }, { "epoch": 0.2, "learning_rate": 1.847505656805186e-05, "loss": 0.9306, "step": 6093 }, { "epoch": 0.2, "learning_rate": 1.847448439416009e-05, "loss": 0.9276, "step": 6094 }, { "epoch": 0.2, "learning_rate": 1.847391212180939e-05, "loss": 0.9477, "step": 6095 }, { "epoch": 0.2, "learning_rate": 1.8473339751006413e-05, "loss": 0.9455, "step": 6096 }, { "epoch": 0.2, "learning_rate": 1.8472767281757805e-05, "loss": 0.97, "step": 6097 }, { "epoch": 0.2, "learning_rate": 1.8472194714070214e-05, "loss": 0.9438, "step": 6098 }, { "epoch": 0.2, "learning_rate": 1.84716220479503e-05, "loss": 0.9633, "step": 6099 }, { "epoch": 0.2, "learning_rate": 1.8471049283404712e-05, "loss": 0.9445, "step": 6100 }, { "epoch": 0.2, "learning_rate": 1.84704764204401e-05, "loss": 0.9912, "step": 6101 }, { "epoch": 0.2, "learning_rate": 1.846990345906313e-05, "loss": 0.9581, "step": 6102 }, { "epoch": 0.2, "learning_rate": 1.846933039928045e-05, "loss": 0.9525, "step": 6103 }, { "epoch": 0.2, "learning_rate": 1.8468757241098724e-05, "loss": 0.9622, "step": 6104 }, { "epoch": 0.2, "learning_rate": 1.8468183984524606e-05, "loss": 0.9938, "step": 6105 }, { "epoch": 0.2, "learning_rate": 1.8467610629564763e-05, "loss": 0.9766, "step": 6106 }, { "epoch": 0.2, "learning_rate": 1.8467037176225848e-05, "loss": 0.9631, "step": 6107 }, { "epoch": 0.2, "learning_rate": 1.846646362451453e-05, "loss": 0.9286, "step": 6108 }, { "epoch": 0.2, "learning_rate": 1.846588997443747e-05, "loss": 0.9979, "step": 6109 }, { "epoch": 0.2, "learning_rate": 1.846531622600133e-05, "loss": 0.9339, "step": 6110 }, { "epoch": 0.2, "learning_rate": 1.8464742379212786e-05, "loss": 0.9615, "step": 6111 }, { "epoch": 0.2, "learning_rate": 1.846416843407849e-05, "loss": 0.9474, "step": 6112 }, { "epoch": 0.2, "learning_rate": 1.8463594390605124e-05, "loss": 0.9481, "step": 6113 }, { "epoch": 0.2, "learning_rate": 1.8463020248799354e-05, "loss": 0.9932, "step": 6114 }, { "epoch": 0.2, "learning_rate": 1.846244600866784e-05, "loss": 0.9811, "step": 6115 }, { "epoch": 0.2, "learning_rate": 1.846187167021727e-05, "loss": 0.9614, "step": 6116 }, { "epoch": 0.2, "learning_rate": 1.846129723345431e-05, "loss": 0.9694, "step": 6117 }, { "epoch": 0.2, "learning_rate": 1.8460722698385625e-05, "loss": 0.9334, "step": 6118 }, { "epoch": 0.2, "learning_rate": 1.84601480650179e-05, "loss": 0.9599, "step": 6119 }, { "epoch": 0.2, "learning_rate": 1.8459573333357814e-05, "loss": 0.9892, "step": 6120 }, { "epoch": 0.2, "learning_rate": 1.8458998503412035e-05, "loss": 0.9565, "step": 6121 }, { "epoch": 0.2, "learning_rate": 1.845842357518725e-05, "loss": 0.9421, "step": 6122 }, { "epoch": 0.2, "learning_rate": 1.8457848548690135e-05, "loss": 0.9934, "step": 6123 }, { "epoch": 0.2, "learning_rate": 1.8457273423927366e-05, "loss": 0.9251, "step": 6124 }, { "epoch": 0.2, "learning_rate": 1.8456698200905638e-05, "loss": 0.9955, "step": 6125 }, { "epoch": 0.2, "learning_rate": 1.8456122879631618e-05, "loss": 0.9435, "step": 6126 }, { "epoch": 0.2, "learning_rate": 1.8455547460112e-05, "loss": 0.976, "step": 6127 }, { "epoch": 0.2, "learning_rate": 1.8454971942353467e-05, "loss": 0.9329, "step": 6128 }, { "epoch": 0.2, "learning_rate": 1.845439632636271e-05, "loss": 0.9739, "step": 6129 }, { "epoch": 0.2, "learning_rate": 1.845382061214641e-05, "loss": 0.9636, "step": 6130 }, { "epoch": 0.2, "learning_rate": 1.8453244799711255e-05, "loss": 0.961, "step": 6131 }, { "epoch": 0.2, "learning_rate": 1.845266888906394e-05, "loss": 0.9401, "step": 6132 }, { "epoch": 0.2, "learning_rate": 1.8452092880211157e-05, "loss": 0.9078, "step": 6133 }, { "epoch": 0.2, "learning_rate": 1.8451516773159597e-05, "loss": 0.9104, "step": 6134 }, { "epoch": 0.2, "learning_rate": 1.845094056791595e-05, "loss": 0.9831, "step": 6135 }, { "epoch": 0.2, "learning_rate": 1.8450364264486912e-05, "loss": 0.9254, "step": 6136 }, { "epoch": 0.2, "learning_rate": 1.8449787862879177e-05, "loss": 0.9703, "step": 6137 }, { "epoch": 0.2, "learning_rate": 1.844921136309945e-05, "loss": 0.9432, "step": 6138 }, { "epoch": 0.2, "learning_rate": 1.8448634765154417e-05, "loss": 0.9728, "step": 6139 }, { "epoch": 0.2, "learning_rate": 1.844805806905079e-05, "loss": 0.9253, "step": 6140 }, { "epoch": 0.2, "learning_rate": 1.8447481274795253e-05, "loss": 0.9608, "step": 6141 }, { "epoch": 0.2, "learning_rate": 1.8446904382394524e-05, "loss": 0.9368, "step": 6142 }, { "epoch": 0.2, "learning_rate": 1.8446327391855294e-05, "loss": 0.9745, "step": 6143 }, { "epoch": 0.2, "learning_rate": 1.8445750303184275e-05, "loss": 0.9908, "step": 6144 }, { "epoch": 0.2, "learning_rate": 1.844517311638816e-05, "loss": 0.9675, "step": 6145 }, { "epoch": 0.2, "learning_rate": 1.844459583147367e-05, "loss": 0.9322, "step": 6146 }, { "epoch": 0.2, "learning_rate": 1.8444018448447504e-05, "loss": 0.995, "step": 6147 }, { "epoch": 0.2, "learning_rate": 1.8443440967316365e-05, "loss": 0.9329, "step": 6148 }, { "epoch": 0.2, "learning_rate": 1.8442863388086974e-05, "loss": 0.9472, "step": 6149 }, { "epoch": 0.2, "learning_rate": 1.844228571076603e-05, "loss": 0.9548, "step": 6150 }, { "epoch": 0.2, "learning_rate": 1.8441707935360255e-05, "loss": 0.9528, "step": 6151 }, { "epoch": 0.2, "learning_rate": 1.8441130061876357e-05, "loss": 0.9336, "step": 6152 }, { "epoch": 0.2, "learning_rate": 1.8440552090321047e-05, "loss": 0.9679, "step": 6153 }, { "epoch": 0.2, "learning_rate": 1.8439974020701043e-05, "loss": 0.9085, "step": 6154 }, { "epoch": 0.2, "learning_rate": 1.8439395853023064e-05, "loss": 0.9616, "step": 6155 }, { "epoch": 0.2, "learning_rate": 1.843881758729382e-05, "loss": 0.9459, "step": 6156 }, { "epoch": 0.2, "learning_rate": 1.8438239223520037e-05, "loss": 0.9812, "step": 6157 }, { "epoch": 0.2, "learning_rate": 1.843766076170843e-05, "loss": 1.0032, "step": 6158 }, { "epoch": 0.2, "learning_rate": 1.8437082201865723e-05, "loss": 0.9924, "step": 6159 }, { "epoch": 0.21, "learning_rate": 1.843650354399863e-05, "loss": 0.9365, "step": 6160 }, { "epoch": 0.21, "learning_rate": 1.8435924788113887e-05, "loss": 0.9843, "step": 6161 }, { "epoch": 0.21, "learning_rate": 1.8435345934218207e-05, "loss": 0.9281, "step": 6162 }, { "epoch": 0.21, "learning_rate": 1.8434766982318324e-05, "loss": 0.9585, "step": 6163 }, { "epoch": 0.21, "learning_rate": 1.8434187932420955e-05, "loss": 0.9425, "step": 6164 }, { "epoch": 0.21, "learning_rate": 1.8433608784532833e-05, "loss": 1.0197, "step": 6165 }, { "epoch": 0.21, "learning_rate": 1.843302953866069e-05, "loss": 0.9814, "step": 6166 }, { "epoch": 0.21, "learning_rate": 1.8432450194811245e-05, "loss": 0.9782, "step": 6167 }, { "epoch": 0.21, "learning_rate": 1.843187075299124e-05, "loss": 0.9788, "step": 6168 }, { "epoch": 0.21, "learning_rate": 1.8431291213207405e-05, "loss": 0.9367, "step": 6169 }, { "epoch": 0.21, "learning_rate": 1.843071157546647e-05, "loss": 0.9749, "step": 6170 }, { "epoch": 0.21, "learning_rate": 1.843013183977517e-05, "loss": 0.9286, "step": 6171 }, { "epoch": 0.21, "learning_rate": 1.8429552006140243e-05, "loss": 0.9648, "step": 6172 }, { "epoch": 0.21, "learning_rate": 1.842897207456842e-05, "loss": 0.9591, "step": 6173 }, { "epoch": 0.21, "learning_rate": 1.8428392045066445e-05, "loss": 0.9391, "step": 6174 }, { "epoch": 0.21, "learning_rate": 1.842781191764105e-05, "loss": 1.0009, "step": 6175 }, { "epoch": 0.21, "learning_rate": 1.8427231692298985e-05, "loss": 0.9477, "step": 6176 }, { "epoch": 0.21, "learning_rate": 1.8426651369046987e-05, "loss": 0.9375, "step": 6177 }, { "epoch": 0.21, "learning_rate": 1.8426070947891797e-05, "loss": 0.9557, "step": 6178 }, { "epoch": 0.21, "learning_rate": 1.8425490428840155e-05, "loss": 0.9845, "step": 6179 }, { "epoch": 0.21, "learning_rate": 1.8424909811898812e-05, "loss": 0.9283, "step": 6180 }, { "epoch": 0.21, "learning_rate": 1.842432909707451e-05, "loss": 1.0166, "step": 6181 }, { "epoch": 0.21, "learning_rate": 1.8423748284373998e-05, "loss": 0.9269, "step": 6182 }, { "epoch": 0.21, "learning_rate": 1.8423167373804025e-05, "loss": 0.9663, "step": 6183 }, { "epoch": 0.21, "learning_rate": 1.8422586365371335e-05, "loss": 0.9454, "step": 6184 }, { "epoch": 0.21, "learning_rate": 1.8422005259082684e-05, "loss": 0.9738, "step": 6185 }, { "epoch": 0.21, "learning_rate": 1.842142405494482e-05, "loss": 0.9937, "step": 6186 }, { "epoch": 0.21, "learning_rate": 1.8420842752964497e-05, "loss": 1.0063, "step": 6187 }, { "epoch": 0.21, "learning_rate": 1.8420261353148467e-05, "loss": 0.937, "step": 6188 }, { "epoch": 0.21, "learning_rate": 1.841967985550349e-05, "loss": 0.9562, "step": 6189 }, { "epoch": 0.21, "learning_rate": 1.8419098260036312e-05, "loss": 0.9847, "step": 6190 }, { "epoch": 0.21, "learning_rate": 1.84185165667537e-05, "loss": 0.962, "step": 6191 }, { "epoch": 0.21, "learning_rate": 1.841793477566241e-05, "loss": 0.954, "step": 6192 }, { "epoch": 0.21, "learning_rate": 1.8417352886769197e-05, "loss": 0.9636, "step": 6193 }, { "epoch": 0.21, "learning_rate": 1.8416770900080826e-05, "loss": 0.904, "step": 6194 }, { "epoch": 0.21, "learning_rate": 1.8416188815604063e-05, "loss": 0.9391, "step": 6195 }, { "epoch": 0.21, "learning_rate": 1.8415606633345657e-05, "loss": 0.9305, "step": 6196 }, { "epoch": 0.21, "learning_rate": 1.8415024353312382e-05, "loss": 0.9719, "step": 6197 }, { "epoch": 0.21, "learning_rate": 1.8414441975511003e-05, "loss": 0.9582, "step": 6198 }, { "epoch": 0.21, "learning_rate": 1.8413859499948286e-05, "loss": 1.0248, "step": 6199 }, { "epoch": 0.21, "learning_rate": 1.8413276926630995e-05, "loss": 0.9751, "step": 6200 }, { "epoch": 0.21, "learning_rate": 1.84126942555659e-05, "loss": 0.9762, "step": 6201 }, { "epoch": 0.21, "learning_rate": 1.8412111486759768e-05, "loss": 0.9719, "step": 6202 }, { "epoch": 0.21, "learning_rate": 1.8411528620219375e-05, "loss": 0.9801, "step": 6203 }, { "epoch": 0.21, "learning_rate": 1.8410945655951493e-05, "loss": 0.9504, "step": 6204 }, { "epoch": 0.21, "learning_rate": 1.8410362593962893e-05, "loss": 0.9424, "step": 6205 }, { "epoch": 0.21, "learning_rate": 1.8409779434260345e-05, "loss": 0.9452, "step": 6206 }, { "epoch": 0.21, "learning_rate": 1.840919617685063e-05, "loss": 0.9591, "step": 6207 }, { "epoch": 0.21, "learning_rate": 1.840861282174052e-05, "loss": 0.9688, "step": 6208 }, { "epoch": 0.21, "learning_rate": 1.84080293689368e-05, "loss": 1.0259, "step": 6209 }, { "epoch": 0.21, "learning_rate": 1.840744581844624e-05, "loss": 0.9409, "step": 6210 }, { "epoch": 0.21, "learning_rate": 1.8406862170275626e-05, "loss": 0.9778, "step": 6211 }, { "epoch": 0.21, "learning_rate": 1.8406278424431737e-05, "loss": 0.9514, "step": 6212 }, { "epoch": 0.21, "learning_rate": 1.8405694580921354e-05, "loss": 0.9751, "step": 6213 }, { "epoch": 0.21, "learning_rate": 1.840511063975126e-05, "loss": 0.9022, "step": 6214 }, { "epoch": 0.21, "learning_rate": 1.840452660092824e-05, "loss": 0.9636, "step": 6215 }, { "epoch": 0.21, "learning_rate": 1.8403942464459085e-05, "loss": 0.9456, "step": 6216 }, { "epoch": 0.21, "learning_rate": 1.8403358230350572e-05, "loss": 0.9938, "step": 6217 }, { "epoch": 0.21, "learning_rate": 1.8402773898609494e-05, "loss": 0.9684, "step": 6218 }, { "epoch": 0.21, "learning_rate": 1.840218946924264e-05, "loss": 0.961, "step": 6219 }, { "epoch": 0.21, "learning_rate": 1.8401604942256802e-05, "loss": 0.9404, "step": 6220 }, { "epoch": 0.21, "learning_rate": 1.8401020317658765e-05, "loss": 0.9513, "step": 6221 }, { "epoch": 0.21, "learning_rate": 1.8400435595455326e-05, "loss": 0.9321, "step": 6222 }, { "epoch": 0.21, "learning_rate": 1.8399850775653277e-05, "loss": 1.0051, "step": 6223 }, { "epoch": 0.21, "learning_rate": 1.8399265858259413e-05, "loss": 0.928, "step": 6224 }, { "epoch": 0.21, "learning_rate": 1.8398680843280533e-05, "loss": 0.9662, "step": 6225 }, { "epoch": 0.21, "learning_rate": 1.839809573072343e-05, "loss": 0.9408, "step": 6226 }, { "epoch": 0.21, "learning_rate": 1.8397510520594898e-05, "loss": 0.9716, "step": 6227 }, { "epoch": 0.21, "learning_rate": 1.8396925212901747e-05, "loss": 0.9268, "step": 6228 }, { "epoch": 0.21, "learning_rate": 1.8396339807650766e-05, "loss": 0.9812, "step": 6229 }, { "epoch": 0.21, "learning_rate": 1.8395754304848762e-05, "loss": 0.9774, "step": 6230 }, { "epoch": 0.21, "learning_rate": 1.839516870450254e-05, "loss": 0.9904, "step": 6231 }, { "epoch": 0.21, "learning_rate": 1.8394583006618897e-05, "loss": 0.942, "step": 6232 }, { "epoch": 0.21, "learning_rate": 1.8393997211204645e-05, "loss": 0.9639, "step": 6233 }, { "epoch": 0.21, "learning_rate": 1.8393411318266588e-05, "loss": 0.9384, "step": 6234 }, { "epoch": 0.21, "learning_rate": 1.8392825327811525e-05, "loss": 0.9901, "step": 6235 }, { "epoch": 0.21, "learning_rate": 1.8392239239846276e-05, "loss": 0.9552, "step": 6236 }, { "epoch": 0.21, "learning_rate": 1.8391653054377642e-05, "loss": 0.9773, "step": 6237 }, { "epoch": 0.21, "learning_rate": 1.839106677141244e-05, "loss": 0.9886, "step": 6238 }, { "epoch": 0.21, "learning_rate": 1.8390480390957474e-05, "loss": 0.9487, "step": 6239 }, { "epoch": 0.21, "learning_rate": 1.8389893913019564e-05, "loss": 0.9017, "step": 6240 }, { "epoch": 0.21, "learning_rate": 1.838930733760552e-05, "loss": 0.9347, "step": 6241 }, { "epoch": 0.21, "learning_rate": 1.8388720664722157e-05, "loss": 0.9616, "step": 6242 }, { "epoch": 0.21, "learning_rate": 1.838813389437629e-05, "loss": 0.963, "step": 6243 }, { "epoch": 0.21, "learning_rate": 1.8387547026574742e-05, "loss": 0.9227, "step": 6244 }, { "epoch": 0.21, "learning_rate": 1.8386960061324327e-05, "loss": 0.9611, "step": 6245 }, { "epoch": 0.21, "learning_rate": 1.8386372998631862e-05, "loss": 0.9366, "step": 6246 }, { "epoch": 0.21, "learning_rate": 1.838578583850417e-05, "loss": 1.0045, "step": 6247 }, { "epoch": 0.21, "learning_rate": 1.8385198580948076e-05, "loss": 0.9067, "step": 6248 }, { "epoch": 0.21, "learning_rate": 1.8384611225970403e-05, "loss": 1.0052, "step": 6249 }, { "epoch": 0.21, "learning_rate": 1.8384023773577967e-05, "loss": 0.946, "step": 6250 }, { "epoch": 0.21, "learning_rate": 1.8383436223777603e-05, "loss": 1.0105, "step": 6251 }, { "epoch": 0.21, "learning_rate": 1.8382848576576128e-05, "loss": 0.9891, "step": 6252 }, { "epoch": 0.21, "learning_rate": 1.8382260831980378e-05, "loss": 0.9413, "step": 6253 }, { "epoch": 0.21, "learning_rate": 1.8381672989997178e-05, "loss": 0.9555, "step": 6254 }, { "epoch": 0.21, "learning_rate": 1.8381085050633358e-05, "loss": 0.9726, "step": 6255 }, { "epoch": 0.21, "learning_rate": 1.8380497013895748e-05, "loss": 0.933, "step": 6256 }, { "epoch": 0.21, "learning_rate": 1.837990887979118e-05, "loss": 1.0082, "step": 6257 }, { "epoch": 0.21, "learning_rate": 1.8379320648326485e-05, "loss": 0.9676, "step": 6258 }, { "epoch": 0.21, "learning_rate": 1.8378732319508503e-05, "loss": 0.9664, "step": 6259 }, { "epoch": 0.21, "learning_rate": 1.8378143893344067e-05, "loss": 0.98, "step": 6260 }, { "epoch": 0.21, "learning_rate": 1.837755536984001e-05, "loss": 0.9658, "step": 6261 }, { "epoch": 0.21, "learning_rate": 1.8376966749003172e-05, "loss": 0.9589, "step": 6262 }, { "epoch": 0.21, "learning_rate": 1.8376378030840397e-05, "loss": 0.9458, "step": 6263 }, { "epoch": 0.21, "learning_rate": 1.8375789215358518e-05, "loss": 0.8938, "step": 6264 }, { "epoch": 0.21, "learning_rate": 1.8375200302564377e-05, "loss": 0.961, "step": 6265 }, { "epoch": 0.21, "learning_rate": 1.8374611292464815e-05, "loss": 0.9646, "step": 6266 }, { "epoch": 0.21, "learning_rate": 1.837402218506668e-05, "loss": 0.9512, "step": 6267 }, { "epoch": 0.21, "learning_rate": 1.8373432980376814e-05, "loss": 0.9415, "step": 6268 }, { "epoch": 0.21, "learning_rate": 1.837284367840206e-05, "loss": 0.9424, "step": 6269 }, { "epoch": 0.21, "learning_rate": 1.837225427914927e-05, "loss": 0.9722, "step": 6270 }, { "epoch": 0.21, "learning_rate": 1.8371664782625287e-05, "loss": 0.9725, "step": 6271 }, { "epoch": 0.21, "learning_rate": 1.8371075188836962e-05, "loss": 0.9703, "step": 6272 }, { "epoch": 0.21, "learning_rate": 1.8370485497791145e-05, "loss": 0.9898, "step": 6273 }, { "epoch": 0.21, "learning_rate": 1.8369895709494686e-05, "loss": 0.9597, "step": 6274 }, { "epoch": 0.21, "learning_rate": 1.836930582395444e-05, "loss": 0.9667, "step": 6275 }, { "epoch": 0.21, "learning_rate": 1.8368715841177263e-05, "loss": 0.9299, "step": 6276 }, { "epoch": 0.21, "learning_rate": 1.836812576117e-05, "loss": 0.9721, "step": 6277 }, { "epoch": 0.21, "learning_rate": 1.8367535583939512e-05, "loss": 0.963, "step": 6278 }, { "epoch": 0.21, "learning_rate": 1.8366945309492657e-05, "loss": 0.9726, "step": 6279 }, { "epoch": 0.21, "learning_rate": 1.8366354937836294e-05, "loss": 0.9212, "step": 6280 }, { "epoch": 0.21, "learning_rate": 1.836576446897728e-05, "loss": 0.9914, "step": 6281 }, { "epoch": 0.21, "learning_rate": 1.8365173902922473e-05, "loss": 0.9803, "step": 6282 }, { "epoch": 0.21, "learning_rate": 1.8364583239678737e-05, "loss": 0.974, "step": 6283 }, { "epoch": 0.21, "learning_rate": 1.8363992479252936e-05, "loss": 0.9513, "step": 6284 }, { "epoch": 0.21, "learning_rate": 1.8363401621651927e-05, "loss": 0.9408, "step": 6285 }, { "epoch": 0.21, "learning_rate": 1.8362810666882586e-05, "loss": 0.9471, "step": 6286 }, { "epoch": 0.21, "learning_rate": 1.836221961495177e-05, "loss": 0.9951, "step": 6287 }, { "epoch": 0.21, "learning_rate": 1.8361628465866346e-05, "loss": 0.9712, "step": 6288 }, { "epoch": 0.21, "learning_rate": 1.8361037219633187e-05, "loss": 0.9546, "step": 6289 }, { "epoch": 0.21, "learning_rate": 1.836044587625916e-05, "loss": 0.9568, "step": 6290 }, { "epoch": 0.21, "learning_rate": 1.8359854435751134e-05, "loss": 0.9385, "step": 6291 }, { "epoch": 0.21, "learning_rate": 1.835926289811598e-05, "loss": 0.9635, "step": 6292 }, { "epoch": 0.21, "learning_rate": 1.8358671263360577e-05, "loss": 0.9298, "step": 6293 }, { "epoch": 0.21, "learning_rate": 1.8358079531491794e-05, "loss": 0.9383, "step": 6294 }, { "epoch": 0.21, "learning_rate": 1.8357487702516502e-05, "loss": 0.9613, "step": 6295 }, { "epoch": 0.21, "learning_rate": 1.8356895776441584e-05, "loss": 0.9653, "step": 6296 }, { "epoch": 0.21, "learning_rate": 1.8356303753273912e-05, "loss": 0.9132, "step": 6297 }, { "epoch": 0.21, "learning_rate": 1.8355711633020367e-05, "loss": 0.9363, "step": 6298 }, { "epoch": 0.21, "learning_rate": 1.835511941568783e-05, "loss": 0.9742, "step": 6299 }, { "epoch": 0.21, "learning_rate": 1.8354527101283178e-05, "loss": 0.9137, "step": 6300 }, { "epoch": 0.21, "learning_rate": 1.8353934689813292e-05, "loss": 0.9601, "step": 6301 }, { "epoch": 0.21, "learning_rate": 1.8353342181285062e-05, "loss": 0.9469, "step": 6302 }, { "epoch": 0.21, "learning_rate": 1.8352749575705365e-05, "loss": 0.9565, "step": 6303 }, { "epoch": 0.21, "learning_rate": 1.8352156873081087e-05, "loss": 0.9581, "step": 6304 }, { "epoch": 0.21, "learning_rate": 1.8351564073419115e-05, "loss": 1.0168, "step": 6305 }, { "epoch": 0.21, "learning_rate": 1.8350971176726338e-05, "loss": 0.9479, "step": 6306 }, { "epoch": 0.21, "learning_rate": 1.835037818300964e-05, "loss": 0.9708, "step": 6307 }, { "epoch": 0.21, "learning_rate": 1.8349785092275916e-05, "loss": 0.9336, "step": 6308 }, { "epoch": 0.21, "learning_rate": 1.8349191904532057e-05, "loss": 0.9639, "step": 6309 }, { "epoch": 0.21, "learning_rate": 1.8348598619784948e-05, "loss": 0.9322, "step": 6310 }, { "epoch": 0.21, "learning_rate": 1.8348005238041484e-05, "loss": 0.9306, "step": 6311 }, { "epoch": 0.21, "learning_rate": 1.8347411759308564e-05, "loss": 0.911, "step": 6312 }, { "epoch": 0.21, "learning_rate": 1.834681818359308e-05, "loss": 0.9678, "step": 6313 }, { "epoch": 0.21, "learning_rate": 1.8346224510901928e-05, "loss": 0.9427, "step": 6314 }, { "epoch": 0.21, "learning_rate": 1.8345630741242006e-05, "loss": 0.9325, "step": 6315 }, { "epoch": 0.21, "learning_rate": 1.8345036874620214e-05, "loss": 0.938, "step": 6316 }, { "epoch": 0.21, "learning_rate": 1.834444291104345e-05, "loss": 0.9556, "step": 6317 }, { "epoch": 0.21, "learning_rate": 1.8343848850518613e-05, "loss": 0.9009, "step": 6318 }, { "epoch": 0.21, "learning_rate": 1.834325469305261e-05, "loss": 0.9993, "step": 6319 }, { "epoch": 0.21, "learning_rate": 1.834266043865234e-05, "loss": 0.9063, "step": 6320 }, { "epoch": 0.21, "learning_rate": 1.8342066087324708e-05, "loss": 0.9922, "step": 6321 }, { "epoch": 0.21, "learning_rate": 1.834147163907662e-05, "loss": 0.9317, "step": 6322 }, { "epoch": 0.21, "learning_rate": 1.834087709391498e-05, "loss": 0.9739, "step": 6323 }, { "epoch": 0.21, "learning_rate": 1.8340282451846702e-05, "loss": 0.9328, "step": 6324 }, { "epoch": 0.21, "learning_rate": 1.8339687712878688e-05, "loss": 0.9587, "step": 6325 }, { "epoch": 0.21, "learning_rate": 1.8339092877017853e-05, "loss": 0.9869, "step": 6326 }, { "epoch": 0.21, "learning_rate": 1.8338497944271102e-05, "loss": 0.9814, "step": 6327 }, { "epoch": 0.21, "learning_rate": 1.833790291464535e-05, "loss": 0.9835, "step": 6328 }, { "epoch": 0.21, "learning_rate": 1.8337307788147512e-05, "loss": 0.9522, "step": 6329 }, { "epoch": 0.21, "learning_rate": 1.8336712564784506e-05, "loss": 0.9751, "step": 6330 }, { "epoch": 0.21, "learning_rate": 1.8336117244563235e-05, "loss": 0.9651, "step": 6331 }, { "epoch": 0.21, "learning_rate": 1.8335521827490628e-05, "loss": 0.9403, "step": 6332 }, { "epoch": 0.21, "learning_rate": 1.8334926313573597e-05, "loss": 0.9504, "step": 6333 }, { "epoch": 0.21, "learning_rate": 1.8334330702819056e-05, "loss": 0.9965, "step": 6334 }, { "epoch": 0.21, "learning_rate": 1.8333734995233935e-05, "loss": 0.9572, "step": 6335 }, { "epoch": 0.21, "learning_rate": 1.833313919082515e-05, "loss": 0.9403, "step": 6336 }, { "epoch": 0.21, "learning_rate": 1.8332543289599628e-05, "loss": 0.9755, "step": 6337 }, { "epoch": 0.21, "learning_rate": 1.8331947291564284e-05, "loss": 0.9884, "step": 6338 }, { "epoch": 0.21, "learning_rate": 1.8331351196726046e-05, "loss": 1.0097, "step": 6339 }, { "epoch": 0.21, "learning_rate": 1.833075500509184e-05, "loss": 0.9435, "step": 6340 }, { "epoch": 0.21, "learning_rate": 1.8330158716668598e-05, "loss": 0.9721, "step": 6341 }, { "epoch": 0.21, "learning_rate": 1.832956233146324e-05, "loss": 0.9612, "step": 6342 }, { "epoch": 0.21, "learning_rate": 1.8328965849482696e-05, "loss": 0.9526, "step": 6343 }, { "epoch": 0.21, "learning_rate": 1.8328369270733903e-05, "loss": 0.9433, "step": 6344 }, { "epoch": 0.21, "learning_rate": 1.8327772595223782e-05, "loss": 0.9575, "step": 6345 }, { "epoch": 0.21, "learning_rate": 1.8327175822959273e-05, "loss": 0.9437, "step": 6346 }, { "epoch": 0.21, "learning_rate": 1.8326578953947304e-05, "loss": 0.9862, "step": 6347 }, { "epoch": 0.21, "learning_rate": 1.8325981988194815e-05, "loss": 0.935, "step": 6348 }, { "epoch": 0.21, "learning_rate": 1.832538492570874e-05, "loss": 0.9872, "step": 6349 }, { "epoch": 0.21, "learning_rate": 1.8324787766496013e-05, "loss": 0.9522, "step": 6350 }, { "epoch": 0.21, "learning_rate": 1.832419051056358e-05, "loss": 1.0004, "step": 6351 }, { "epoch": 0.21, "learning_rate": 1.8323593157918366e-05, "loss": 0.9381, "step": 6352 }, { "epoch": 0.21, "learning_rate": 1.8322995708567323e-05, "loss": 0.9473, "step": 6353 }, { "epoch": 0.21, "learning_rate": 1.832239816251739e-05, "loss": 0.9268, "step": 6354 }, { "epoch": 0.21, "learning_rate": 1.8321800519775508e-05, "loss": 0.9442, "step": 6355 }, { "epoch": 0.21, "learning_rate": 1.832120278034862e-05, "loss": 0.9751, "step": 6356 }, { "epoch": 0.21, "learning_rate": 1.8320604944243672e-05, "loss": 0.9839, "step": 6357 }, { "epoch": 0.21, "learning_rate": 1.8320007011467606e-05, "loss": 0.9739, "step": 6358 }, { "epoch": 0.21, "learning_rate": 1.8319408982027374e-05, "loss": 0.9945, "step": 6359 }, { "epoch": 0.21, "learning_rate": 1.8318810855929926e-05, "loss": 0.9146, "step": 6360 }, { "epoch": 0.21, "learning_rate": 1.8318212633182203e-05, "loss": 0.943, "step": 6361 }, { "epoch": 0.21, "learning_rate": 1.8317614313791162e-05, "loss": 0.9416, "step": 6362 }, { "epoch": 0.21, "learning_rate": 1.831701589776375e-05, "loss": 0.9406, "step": 6363 }, { "epoch": 0.21, "learning_rate": 1.8316417385106927e-05, "loss": 0.9495, "step": 6364 }, { "epoch": 0.21, "learning_rate": 1.831581877582764e-05, "loss": 0.9806, "step": 6365 }, { "epoch": 0.21, "learning_rate": 1.831522006993284e-05, "loss": 0.956, "step": 6366 }, { "epoch": 0.21, "learning_rate": 1.8314621267429494e-05, "loss": 0.9611, "step": 6367 }, { "epoch": 0.21, "learning_rate": 1.831402236832455e-05, "loss": 0.9394, "step": 6368 }, { "epoch": 0.21, "learning_rate": 1.831342337262497e-05, "loss": 0.9718, "step": 6369 }, { "epoch": 0.21, "learning_rate": 1.8312824280337715e-05, "loss": 0.9843, "step": 6370 }, { "epoch": 0.21, "learning_rate": 1.8312225091469742e-05, "loss": 0.9966, "step": 6371 }, { "epoch": 0.21, "learning_rate": 1.8311625806028012e-05, "loss": 0.9345, "step": 6372 }, { "epoch": 0.21, "learning_rate": 1.8311026424019492e-05, "loss": 0.9744, "step": 6373 }, { "epoch": 0.21, "learning_rate": 1.8310426945451143e-05, "loss": 0.9395, "step": 6374 }, { "epoch": 0.21, "learning_rate": 1.830982737032993e-05, "loss": 0.9575, "step": 6375 }, { "epoch": 0.21, "learning_rate": 1.8309227698662815e-05, "loss": 0.9235, "step": 6376 }, { "epoch": 0.21, "learning_rate": 1.8308627930456775e-05, "loss": 0.941, "step": 6377 }, { "epoch": 0.21, "learning_rate": 1.8308028065718774e-05, "loss": 1.01, "step": 6378 }, { "epoch": 0.21, "learning_rate": 1.8307428104455775e-05, "loss": 0.9707, "step": 6379 }, { "epoch": 0.21, "learning_rate": 1.8306828046674753e-05, "loss": 0.9436, "step": 6380 }, { "epoch": 0.21, "learning_rate": 1.8306227892382683e-05, "loss": 0.9773, "step": 6381 }, { "epoch": 0.21, "learning_rate": 1.8305627641586533e-05, "loss": 0.9297, "step": 6382 }, { "epoch": 0.21, "learning_rate": 1.8305027294293282e-05, "loss": 1.0436, "step": 6383 }, { "epoch": 0.21, "learning_rate": 1.8304426850509897e-05, "loss": 1.0011, "step": 6384 }, { "epoch": 0.21, "learning_rate": 1.830382631024336e-05, "loss": 0.9689, "step": 6385 }, { "epoch": 0.21, "learning_rate": 1.830322567350065e-05, "loss": 0.9716, "step": 6386 }, { "epoch": 0.21, "learning_rate": 1.8302624940288743e-05, "loss": 0.9509, "step": 6387 }, { "epoch": 0.21, "learning_rate": 1.8302024110614616e-05, "loss": 0.9584, "step": 6388 }, { "epoch": 0.21, "learning_rate": 1.8301423184485253e-05, "loss": 0.9607, "step": 6389 }, { "epoch": 0.21, "learning_rate": 1.8300822161907633e-05, "loss": 0.9515, "step": 6390 }, { "epoch": 0.21, "learning_rate": 1.8300221042888743e-05, "loss": 0.9636, "step": 6391 }, { "epoch": 0.21, "learning_rate": 1.8299619827435558e-05, "loss": 0.9452, "step": 6392 }, { "epoch": 0.21, "learning_rate": 1.8299018515555075e-05, "loss": 0.9418, "step": 6393 }, { "epoch": 0.21, "learning_rate": 1.829841710725427e-05, "loss": 0.9318, "step": 6394 }, { "epoch": 0.21, "learning_rate": 1.829781560254014e-05, "loss": 0.9865, "step": 6395 }, { "epoch": 0.21, "learning_rate": 1.8297214001419664e-05, "loss": 0.9393, "step": 6396 }, { "epoch": 0.21, "learning_rate": 1.829661230389984e-05, "loss": 0.9537, "step": 6397 }, { "epoch": 0.21, "learning_rate": 1.829601050998765e-05, "loss": 0.9736, "step": 6398 }, { "epoch": 0.21, "learning_rate": 1.829540861969009e-05, "loss": 1.021, "step": 6399 }, { "epoch": 0.21, "learning_rate": 1.8294806633014155e-05, "loss": 0.9366, "step": 6400 }, { "epoch": 0.21, "learning_rate": 1.829420454996684e-05, "loss": 0.9354, "step": 6401 }, { "epoch": 0.21, "learning_rate": 1.8293602370555135e-05, "loss": 0.9989, "step": 6402 }, { "epoch": 0.21, "learning_rate": 1.829300009478604e-05, "loss": 0.9588, "step": 6403 }, { "epoch": 0.21, "learning_rate": 1.829239772266655e-05, "loss": 0.9404, "step": 6404 }, { "epoch": 0.21, "learning_rate": 1.8291795254203664e-05, "loss": 0.9745, "step": 6405 }, { "epoch": 0.21, "learning_rate": 1.829119268940438e-05, "loss": 0.9644, "step": 6406 }, { "epoch": 0.21, "learning_rate": 1.8290590028275708e-05, "loss": 0.9948, "step": 6407 }, { "epoch": 0.21, "learning_rate": 1.8289987270824637e-05, "loss": 0.9344, "step": 6408 }, { "epoch": 0.21, "learning_rate": 1.8289384417058177e-05, "loss": 0.9893, "step": 6409 }, { "epoch": 0.21, "learning_rate": 1.8288781466983333e-05, "loss": 0.9546, "step": 6410 }, { "epoch": 0.21, "learning_rate": 1.828817842060711e-05, "loss": 0.9639, "step": 6411 }, { "epoch": 0.21, "learning_rate": 1.828757527793651e-05, "loss": 0.9404, "step": 6412 }, { "epoch": 0.21, "learning_rate": 1.828697203897854e-05, "loss": 0.9467, "step": 6413 }, { "epoch": 0.21, "learning_rate": 1.828636870374022e-05, "loss": 0.9351, "step": 6414 }, { "epoch": 0.21, "learning_rate": 1.8285765272228544e-05, "loss": 0.9513, "step": 6415 }, { "epoch": 0.21, "learning_rate": 1.8285161744450534e-05, "loss": 0.9422, "step": 6416 }, { "epoch": 0.21, "learning_rate": 1.8284558120413197e-05, "loss": 0.9612, "step": 6417 }, { "epoch": 0.21, "learning_rate": 1.828395440012355e-05, "loss": 0.9765, "step": 6418 }, { "epoch": 0.21, "learning_rate": 1.8283350583588602e-05, "loss": 0.9588, "step": 6419 }, { "epoch": 0.21, "learning_rate": 1.8282746670815372e-05, "loss": 0.8999, "step": 6420 }, { "epoch": 0.21, "learning_rate": 1.8282142661810874e-05, "loss": 0.9659, "step": 6421 }, { "epoch": 0.21, "learning_rate": 1.828153855658213e-05, "loss": 0.9287, "step": 6422 }, { "epoch": 0.21, "learning_rate": 1.828093435513615e-05, "loss": 0.9901, "step": 6423 }, { "epoch": 0.21, "learning_rate": 1.8280330057479965e-05, "loss": 0.9236, "step": 6424 }, { "epoch": 0.21, "learning_rate": 1.8279725663620586e-05, "loss": 0.9961, "step": 6425 }, { "epoch": 0.21, "learning_rate": 1.8279121173565043e-05, "loss": 0.9834, "step": 6426 }, { "epoch": 0.21, "learning_rate": 1.8278516587320352e-05, "loss": 0.9833, "step": 6427 }, { "epoch": 0.21, "learning_rate": 1.827791190489354e-05, "loss": 0.9478, "step": 6428 }, { "epoch": 0.21, "learning_rate": 1.8277307126291638e-05, "loss": 0.9688, "step": 6429 }, { "epoch": 0.21, "learning_rate": 1.8276702251521663e-05, "loss": 0.9651, "step": 6430 }, { "epoch": 0.21, "learning_rate": 1.827609728059065e-05, "loss": 0.9564, "step": 6431 }, { "epoch": 0.21, "learning_rate": 1.827549221350562e-05, "loss": 0.9216, "step": 6432 }, { "epoch": 0.21, "learning_rate": 1.8274887050273612e-05, "loss": 0.9885, "step": 6433 }, { "epoch": 0.21, "learning_rate": 1.827428179090165e-05, "loss": 0.9464, "step": 6434 }, { "epoch": 0.21, "learning_rate": 1.827367643539677e-05, "loss": 0.9558, "step": 6435 }, { "epoch": 0.21, "learning_rate": 1.8273070983766004e-05, "loss": 0.9354, "step": 6436 }, { "epoch": 0.21, "learning_rate": 1.8272465436016385e-05, "loss": 0.9938, "step": 6437 }, { "epoch": 0.21, "learning_rate": 1.827185979215495e-05, "loss": 0.9593, "step": 6438 }, { "epoch": 0.21, "learning_rate": 1.827125405218874e-05, "loss": 0.9679, "step": 6439 }, { "epoch": 0.21, "learning_rate": 1.8270648216124782e-05, "loss": 0.9783, "step": 6440 }, { "epoch": 0.21, "learning_rate": 1.827004228397012e-05, "loss": 0.9629, "step": 6441 }, { "epoch": 0.21, "learning_rate": 1.8269436255731797e-05, "loss": 0.9431, "step": 6442 }, { "epoch": 0.21, "learning_rate": 1.8268830131416852e-05, "loss": 0.9442, "step": 6443 }, { "epoch": 0.21, "learning_rate": 1.8268223911032326e-05, "loss": 0.9552, "step": 6444 }, { "epoch": 0.21, "learning_rate": 1.826761759458526e-05, "loss": 0.9351, "step": 6445 }, { "epoch": 0.21, "learning_rate": 1.8267011182082705e-05, "loss": 0.9667, "step": 6446 }, { "epoch": 0.21, "learning_rate": 1.8266404673531698e-05, "loss": 0.96, "step": 6447 }, { "epoch": 0.21, "learning_rate": 1.8265798068939295e-05, "loss": 0.9663, "step": 6448 }, { "epoch": 0.21, "learning_rate": 1.826519136831254e-05, "loss": 0.991, "step": 6449 }, { "epoch": 0.21, "learning_rate": 1.8264584571658475e-05, "loss": 0.9889, "step": 6450 }, { "epoch": 0.21, "learning_rate": 1.826397767898416e-05, "loss": 0.9595, "step": 6451 }, { "epoch": 0.21, "learning_rate": 1.826337069029664e-05, "loss": 0.919, "step": 6452 }, { "epoch": 0.21, "learning_rate": 1.826276360560297e-05, "loss": 0.9646, "step": 6453 }, { "epoch": 0.21, "learning_rate": 1.82621564249102e-05, "loss": 0.9254, "step": 6454 }, { "epoch": 0.21, "learning_rate": 1.8261549148225387e-05, "loss": 0.9433, "step": 6455 }, { "epoch": 0.21, "learning_rate": 1.8260941775555586e-05, "loss": 0.9245, "step": 6456 }, { "epoch": 0.21, "learning_rate": 1.8260334306907855e-05, "loss": 0.9959, "step": 6457 }, { "epoch": 0.21, "learning_rate": 1.825972674228925e-05, "loss": 0.9238, "step": 6458 }, { "epoch": 0.21, "learning_rate": 1.825911908170683e-05, "loss": 1.0011, "step": 6459 }, { "epoch": 0.21, "learning_rate": 1.8258511325167656e-05, "loss": 0.9375, "step": 6460 }, { "epoch": 0.22, "learning_rate": 1.8257903472678786e-05, "loss": 0.9496, "step": 6461 }, { "epoch": 0.22, "learning_rate": 1.8257295524247287e-05, "loss": 0.9541, "step": 6462 }, { "epoch": 0.22, "learning_rate": 1.8256687479880223e-05, "loss": 0.9728, "step": 6463 }, { "epoch": 0.22, "learning_rate": 1.825607933958465e-05, "loss": 0.9574, "step": 6464 }, { "epoch": 0.22, "learning_rate": 1.8255471103367638e-05, "loss": 0.9568, "step": 6465 }, { "epoch": 0.22, "learning_rate": 1.8254862771236258e-05, "loss": 0.9115, "step": 6466 }, { "epoch": 0.22, "learning_rate": 1.8254254343197574e-05, "loss": 0.9556, "step": 6467 }, { "epoch": 0.22, "learning_rate": 1.8253645819258653e-05, "loss": 0.9261, "step": 6468 }, { "epoch": 0.22, "learning_rate": 1.8253037199426566e-05, "loss": 0.9603, "step": 6469 }, { "epoch": 0.22, "learning_rate": 1.8252428483708388e-05, "loss": 0.9219, "step": 6470 }, { "epoch": 0.22, "learning_rate": 1.825181967211119e-05, "loss": 0.9673, "step": 6471 }, { "epoch": 0.22, "learning_rate": 1.8251210764642042e-05, "loss": 0.9041, "step": 6472 }, { "epoch": 0.22, "learning_rate": 1.8250601761308016e-05, "loss": 1.0178, "step": 6473 }, { "epoch": 0.22, "learning_rate": 1.82499926621162e-05, "loss": 0.9521, "step": 6474 }, { "epoch": 0.22, "learning_rate": 1.8249383467073658e-05, "loss": 0.9515, "step": 6475 }, { "epoch": 0.22, "learning_rate": 1.824877417618747e-05, "loss": 0.9461, "step": 6476 }, { "epoch": 0.22, "learning_rate": 1.8248164789464723e-05, "loss": 0.9915, "step": 6477 }, { "epoch": 0.22, "learning_rate": 1.824755530691249e-05, "loss": 0.9448, "step": 6478 }, { "epoch": 0.22, "learning_rate": 1.8246945728537852e-05, "loss": 0.9804, "step": 6479 }, { "epoch": 0.22, "learning_rate": 1.824633605434789e-05, "loss": 0.9632, "step": 6480 }, { "epoch": 0.22, "learning_rate": 1.8245726284349694e-05, "loss": 0.9523, "step": 6481 }, { "epoch": 0.22, "learning_rate": 1.8245116418550343e-05, "loss": 0.9539, "step": 6482 }, { "epoch": 0.22, "learning_rate": 1.8244506456956922e-05, "loss": 0.9723, "step": 6483 }, { "epoch": 0.22, "learning_rate": 1.8243896399576523e-05, "loss": 0.9099, "step": 6484 }, { "epoch": 0.22, "learning_rate": 1.824328624641623e-05, "loss": 0.953, "step": 6485 }, { "epoch": 0.22, "learning_rate": 1.824267599748313e-05, "loss": 0.9738, "step": 6486 }, { "epoch": 0.22, "learning_rate": 1.824206565278432e-05, "loss": 0.9504, "step": 6487 }, { "epoch": 0.22, "learning_rate": 1.8241455212326886e-05, "loss": 0.9735, "step": 6488 }, { "epoch": 0.22, "learning_rate": 1.8240844676117918e-05, "loss": 0.9876, "step": 6489 }, { "epoch": 0.22, "learning_rate": 1.8240234044164513e-05, "loss": 0.9899, "step": 6490 }, { "epoch": 0.22, "learning_rate": 1.8239623316473766e-05, "loss": 0.9609, "step": 6491 }, { "epoch": 0.22, "learning_rate": 1.823901249305277e-05, "loss": 0.9573, "step": 6492 }, { "epoch": 0.22, "learning_rate": 1.8238401573908628e-05, "loss": 0.9389, "step": 6493 }, { "epoch": 0.22, "learning_rate": 1.8237790559048426e-05, "loss": 0.9432, "step": 6494 }, { "epoch": 0.22, "learning_rate": 1.8237179448479275e-05, "loss": 0.9299, "step": 6495 }, { "epoch": 0.22, "learning_rate": 1.823656824220827e-05, "loss": 0.9379, "step": 6496 }, { "epoch": 0.22, "learning_rate": 1.823595694024251e-05, "loss": 0.9628, "step": 6497 }, { "epoch": 0.22, "learning_rate": 1.82353455425891e-05, "loss": 0.9599, "step": 6498 }, { "epoch": 0.22, "learning_rate": 1.8234734049255142e-05, "loss": 0.9714, "step": 6499 }, { "epoch": 0.22, "learning_rate": 1.823412246024774e-05, "loss": 0.9616, "step": 6500 }, { "epoch": 0.22, "learning_rate": 1.8233510775574007e-05, "loss": 0.9898, "step": 6501 }, { "epoch": 0.22, "learning_rate": 1.823289899524104e-05, "loss": 0.9358, "step": 6502 }, { "epoch": 0.22, "learning_rate": 1.823228711925595e-05, "loss": 0.9723, "step": 6503 }, { "epoch": 0.22, "learning_rate": 1.823167514762585e-05, "loss": 0.9587, "step": 6504 }, { "epoch": 0.22, "learning_rate": 1.8231063080357844e-05, "loss": 0.9756, "step": 6505 }, { "epoch": 0.22, "learning_rate": 1.8230450917459045e-05, "loss": 0.967, "step": 6506 }, { "epoch": 0.22, "learning_rate": 1.8229838658936566e-05, "loss": 0.9696, "step": 6507 }, { "epoch": 0.22, "learning_rate": 1.822922630479752e-05, "loss": 0.946, "step": 6508 }, { "epoch": 0.22, "learning_rate": 1.8228613855049026e-05, "loss": 0.9471, "step": 6509 }, { "epoch": 0.22, "learning_rate": 1.822800130969819e-05, "loss": 0.9376, "step": 6510 }, { "epoch": 0.22, "learning_rate": 1.8227388668752137e-05, "loss": 1.0111, "step": 6511 }, { "epoch": 0.22, "learning_rate": 1.822677593221798e-05, "loss": 0.9152, "step": 6512 }, { "epoch": 0.22, "learning_rate": 1.8226163100102842e-05, "loss": 0.9537, "step": 6513 }, { "epoch": 0.22, "learning_rate": 1.8225550172413838e-05, "loss": 0.95, "step": 6514 }, { "epoch": 0.22, "learning_rate": 1.8224937149158098e-05, "loss": 0.9519, "step": 6515 }, { "epoch": 0.22, "learning_rate": 1.8224324030342736e-05, "loss": 0.96, "step": 6516 }, { "epoch": 0.22, "learning_rate": 1.8223710815974875e-05, "loss": 0.937, "step": 6517 }, { "epoch": 0.22, "learning_rate": 1.8223097506061647e-05, "loss": 0.949, "step": 6518 }, { "epoch": 0.22, "learning_rate": 1.822248410061017e-05, "loss": 0.9965, "step": 6519 }, { "epoch": 0.22, "learning_rate": 1.8221870599627577e-05, "loss": 0.9115, "step": 6520 }, { "epoch": 0.22, "learning_rate": 1.8221257003120988e-05, "loss": 0.9549, "step": 6521 }, { "epoch": 0.22, "learning_rate": 1.8220643311097536e-05, "loss": 0.9709, "step": 6522 }, { "epoch": 0.22, "learning_rate": 1.8220029523564354e-05, "loss": 0.9514, "step": 6523 }, { "epoch": 0.22, "learning_rate": 1.8219415640528573e-05, "loss": 0.995, "step": 6524 }, { "epoch": 0.22, "learning_rate": 1.821880166199732e-05, "loss": 0.9621, "step": 6525 }, { "epoch": 0.22, "learning_rate": 1.821818758797773e-05, "loss": 0.9319, "step": 6526 }, { "epoch": 0.22, "learning_rate": 1.8217573418476942e-05, "loss": 0.9845, "step": 6527 }, { "epoch": 0.22, "learning_rate": 1.821695915350209e-05, "loss": 0.9261, "step": 6528 }, { "epoch": 0.22, "learning_rate": 1.82163447930603e-05, "loss": 0.9753, "step": 6529 }, { "epoch": 0.22, "learning_rate": 1.8215730337158728e-05, "loss": 0.9214, "step": 6530 }, { "epoch": 0.22, "learning_rate": 1.82151157858045e-05, "loss": 0.9839, "step": 6531 }, { "epoch": 0.22, "learning_rate": 1.821450113900476e-05, "loss": 0.948, "step": 6532 }, { "epoch": 0.22, "learning_rate": 1.8213886396766648e-05, "loss": 0.9781, "step": 6533 }, { "epoch": 0.22, "learning_rate": 1.8213271559097312e-05, "loss": 0.9895, "step": 6534 }, { "epoch": 0.22, "learning_rate": 1.8212656626003886e-05, "loss": 0.9712, "step": 6535 }, { "epoch": 0.22, "learning_rate": 1.821204159749352e-05, "loss": 0.9547, "step": 6536 }, { "epoch": 0.22, "learning_rate": 1.821142647357336e-05, "loss": 0.9558, "step": 6537 }, { "epoch": 0.22, "learning_rate": 1.8210811254250548e-05, "loss": 0.9536, "step": 6538 }, { "epoch": 0.22, "learning_rate": 1.821019593953224e-05, "loss": 0.9531, "step": 6539 }, { "epoch": 0.22, "learning_rate": 1.8209580529425578e-05, "loss": 0.9282, "step": 6540 }, { "epoch": 0.22, "learning_rate": 1.8208965023937712e-05, "loss": 0.9518, "step": 6541 }, { "epoch": 0.22, "learning_rate": 1.8208349423075797e-05, "loss": 0.9449, "step": 6542 }, { "epoch": 0.22, "learning_rate": 1.8207733726846984e-05, "loss": 0.9581, "step": 6543 }, { "epoch": 0.22, "learning_rate": 1.8207117935258422e-05, "loss": 0.9354, "step": 6544 }, { "epoch": 0.22, "learning_rate": 1.8206502048317275e-05, "loss": 0.9822, "step": 6545 }, { "epoch": 0.22, "learning_rate": 1.8205886066030687e-05, "loss": 0.9087, "step": 6546 }, { "epoch": 0.22, "learning_rate": 1.820526998840582e-05, "loss": 0.9682, "step": 6547 }, { "epoch": 0.22, "learning_rate": 1.8204653815449837e-05, "loss": 0.9233, "step": 6548 }, { "epoch": 0.22, "learning_rate": 1.820403754716989e-05, "loss": 0.9485, "step": 6549 }, { "epoch": 0.22, "learning_rate": 1.820342118357314e-05, "loss": 0.9451, "step": 6550 }, { "epoch": 0.22, "learning_rate": 1.8202804724666748e-05, "loss": 0.9863, "step": 6551 }, { "epoch": 0.22, "learning_rate": 1.8202188170457883e-05, "loss": 0.9712, "step": 6552 }, { "epoch": 0.22, "learning_rate": 1.8201571520953695e-05, "loss": 0.9266, "step": 6553 }, { "epoch": 0.22, "learning_rate": 1.820095477616136e-05, "loss": 0.9793, "step": 6554 }, { "epoch": 0.22, "learning_rate": 1.820033793608804e-05, "loss": 0.9725, "step": 6555 }, { "epoch": 0.22, "learning_rate": 1.81997210007409e-05, "loss": 0.9875, "step": 6556 }, { "epoch": 0.22, "learning_rate": 1.8199103970127104e-05, "loss": 0.9592, "step": 6557 }, { "epoch": 0.22, "learning_rate": 1.8198486844253833e-05, "loss": 0.9652, "step": 6558 }, { "epoch": 0.22, "learning_rate": 1.8197869623128247e-05, "loss": 0.9871, "step": 6559 }, { "epoch": 0.22, "learning_rate": 1.8197252306757517e-05, "loss": 0.9422, "step": 6560 }, { "epoch": 0.22, "learning_rate": 1.819663489514882e-05, "loss": 0.957, "step": 6561 }, { "epoch": 0.22, "learning_rate": 1.819601738830933e-05, "loss": 0.9062, "step": 6562 }, { "epoch": 0.22, "learning_rate": 1.8195399786246213e-05, "loss": 0.957, "step": 6563 }, { "epoch": 0.22, "learning_rate": 1.819478208896665e-05, "loss": 0.936, "step": 6564 }, { "epoch": 0.22, "learning_rate": 1.819416429647782e-05, "loss": 0.949, "step": 6565 }, { "epoch": 0.22, "learning_rate": 1.81935464087869e-05, "loss": 0.9121, "step": 6566 }, { "epoch": 0.22, "learning_rate": 1.819292842590106e-05, "loss": 0.9141, "step": 6567 }, { "epoch": 0.22, "learning_rate": 1.8192310347827496e-05, "loss": 0.9495, "step": 6568 }, { "epoch": 0.22, "learning_rate": 1.819169217457337e-05, "loss": 0.966, "step": 6569 }, { "epoch": 0.22, "learning_rate": 1.819107390614588e-05, "loss": 0.9514, "step": 6570 }, { "epoch": 0.22, "learning_rate": 1.8190455542552204e-05, "loss": 0.9308, "step": 6571 }, { "epoch": 0.22, "learning_rate": 1.8189837083799525e-05, "loss": 0.9289, "step": 6572 }, { "epoch": 0.22, "learning_rate": 1.8189218529895028e-05, "loss": 0.9105, "step": 6573 }, { "epoch": 0.22, "learning_rate": 1.8188599880845897e-05, "loss": 0.9737, "step": 6574 }, { "epoch": 0.22, "learning_rate": 1.8187981136659327e-05, "loss": 0.9467, "step": 6575 }, { "epoch": 0.22, "learning_rate": 1.81873622973425e-05, "loss": 0.9815, "step": 6576 }, { "epoch": 0.22, "learning_rate": 1.8186743362902614e-05, "loss": 0.9329, "step": 6577 }, { "epoch": 0.22, "learning_rate": 1.818612433334685e-05, "loss": 0.9904, "step": 6578 }, { "epoch": 0.22, "learning_rate": 1.8185505208682404e-05, "loss": 0.9578, "step": 6579 }, { "epoch": 0.22, "learning_rate": 1.818488598891647e-05, "loss": 0.9351, "step": 6580 }, { "epoch": 0.22, "learning_rate": 1.8184266674056244e-05, "loss": 0.9633, "step": 6581 }, { "epoch": 0.22, "learning_rate": 1.818364726410892e-05, "loss": 0.9325, "step": 6582 }, { "epoch": 0.22, "learning_rate": 1.818302775908169e-05, "loss": 0.9809, "step": 6583 }, { "epoch": 0.22, "learning_rate": 1.818240815898176e-05, "loss": 0.9462, "step": 6584 }, { "epoch": 0.22, "learning_rate": 1.8181788463816322e-05, "loss": 0.9622, "step": 6585 }, { "epoch": 0.22, "learning_rate": 1.818116867359258e-05, "loss": 0.9532, "step": 6586 }, { "epoch": 0.22, "learning_rate": 1.8180548788317732e-05, "loss": 0.9598, "step": 6587 }, { "epoch": 0.22, "learning_rate": 1.8179928807998977e-05, "loss": 0.9412, "step": 6588 }, { "epoch": 0.22, "learning_rate": 1.8179308732643525e-05, "loss": 0.9364, "step": 6589 }, { "epoch": 0.22, "learning_rate": 1.817868856225858e-05, "loss": 0.9323, "step": 6590 }, { "epoch": 0.22, "learning_rate": 1.817806829685134e-05, "loss": 0.9781, "step": 6591 }, { "epoch": 0.22, "learning_rate": 1.817744793642902e-05, "loss": 0.9582, "step": 6592 }, { "epoch": 0.22, "learning_rate": 1.8176827480998822e-05, "loss": 0.9444, "step": 6593 }, { "epoch": 0.22, "learning_rate": 1.8176206930567953e-05, "loss": 0.9695, "step": 6594 }, { "epoch": 0.22, "learning_rate": 1.817558628514363e-05, "loss": 0.961, "step": 6595 }, { "epoch": 0.22, "learning_rate": 1.8174965544733057e-05, "loss": 0.9528, "step": 6596 }, { "epoch": 0.22, "learning_rate": 1.817434470934345e-05, "loss": 0.9516, "step": 6597 }, { "epoch": 0.22, "learning_rate": 1.817372377898202e-05, "loss": 0.9944, "step": 6598 }, { "epoch": 0.22, "learning_rate": 1.817310275365598e-05, "loss": 0.9703, "step": 6599 }, { "epoch": 0.22, "learning_rate": 1.817248163337255e-05, "loss": 0.9451, "step": 6600 }, { "epoch": 0.22, "learning_rate": 1.817186041813894e-05, "loss": 0.9689, "step": 6601 }, { "epoch": 0.22, "learning_rate": 1.8171239107962372e-05, "loss": 0.9273, "step": 6602 }, { "epoch": 0.22, "learning_rate": 1.8170617702850066e-05, "loss": 0.9436, "step": 6603 }, { "epoch": 0.22, "learning_rate": 1.816999620280924e-05, "loss": 0.9445, "step": 6604 }, { "epoch": 0.22, "learning_rate": 1.8169374607847107e-05, "loss": 0.9592, "step": 6605 }, { "epoch": 0.22, "learning_rate": 1.8168752917970903e-05, "loss": 0.9915, "step": 6606 }, { "epoch": 0.22, "learning_rate": 1.8168131133187838e-05, "loss": 0.9708, "step": 6607 }, { "epoch": 0.22, "learning_rate": 1.8167509253505143e-05, "loss": 0.925, "step": 6608 }, { "epoch": 0.22, "learning_rate": 1.816688727893004e-05, "loss": 0.9548, "step": 6609 }, { "epoch": 0.22, "learning_rate": 1.8166265209469764e-05, "loss": 0.9296, "step": 6610 }, { "epoch": 0.22, "learning_rate": 1.816564304513153e-05, "loss": 0.9752, "step": 6611 }, { "epoch": 0.22, "learning_rate": 1.8165020785922577e-05, "loss": 0.9817, "step": 6612 }, { "epoch": 0.22, "learning_rate": 1.8164398431850125e-05, "loss": 0.9358, "step": 6613 }, { "epoch": 0.22, "learning_rate": 1.816377598292141e-05, "loss": 0.9835, "step": 6614 }, { "epoch": 0.22, "learning_rate": 1.8163153439143663e-05, "loss": 0.9753, "step": 6615 }, { "epoch": 0.22, "learning_rate": 1.8162530800524116e-05, "loss": 0.9692, "step": 6616 }, { "epoch": 0.22, "learning_rate": 1.8161908067070005e-05, "loss": 0.9978, "step": 6617 }, { "epoch": 0.22, "learning_rate": 1.8161285238788563e-05, "loss": 0.9392, "step": 6618 }, { "epoch": 0.22, "learning_rate": 1.816066231568703e-05, "loss": 0.9538, "step": 6619 }, { "epoch": 0.22, "learning_rate": 1.816003929777264e-05, "loss": 0.933, "step": 6620 }, { "epoch": 0.22, "learning_rate": 1.815941618505263e-05, "loss": 0.9818, "step": 6621 }, { "epoch": 0.22, "learning_rate": 1.815879297753424e-05, "loss": 0.9413, "step": 6622 }, { "epoch": 0.22, "learning_rate": 1.8158169675224715e-05, "loss": 1.0004, "step": 6623 }, { "epoch": 0.22, "learning_rate": 1.8157546278131294e-05, "loss": 0.928, "step": 6624 }, { "epoch": 0.22, "learning_rate": 1.815692278626122e-05, "loss": 1.0212, "step": 6625 }, { "epoch": 0.22, "learning_rate": 1.8156299199621733e-05, "loss": 0.9696, "step": 6626 }, { "epoch": 0.22, "learning_rate": 1.815567551822008e-05, "loss": 0.9657, "step": 6627 }, { "epoch": 0.22, "learning_rate": 1.8155051742063513e-05, "loss": 0.9562, "step": 6628 }, { "epoch": 0.22, "learning_rate": 1.8154427871159272e-05, "loss": 0.9779, "step": 6629 }, { "epoch": 0.22, "learning_rate": 1.815380390551461e-05, "loss": 0.9667, "step": 6630 }, { "epoch": 0.22, "learning_rate": 1.8153179845136773e-05, "loss": 0.9703, "step": 6631 }, { "epoch": 0.22, "learning_rate": 1.8152555690033012e-05, "loss": 0.9629, "step": 6632 }, { "epoch": 0.22, "learning_rate": 1.815193144021058e-05, "loss": 0.9403, "step": 6633 }, { "epoch": 0.22, "learning_rate": 1.815130709567673e-05, "loss": 0.9726, "step": 6634 }, { "epoch": 0.22, "learning_rate": 1.8150682656438716e-05, "loss": 1.0028, "step": 6635 }, { "epoch": 0.22, "learning_rate": 1.8150058122503793e-05, "loss": 0.9358, "step": 6636 }, { "epoch": 0.22, "learning_rate": 1.8149433493879212e-05, "loss": 0.9942, "step": 6637 }, { "epoch": 0.22, "learning_rate": 1.814880877057224e-05, "loss": 0.961, "step": 6638 }, { "epoch": 0.22, "learning_rate": 1.8148183952590122e-05, "loss": 0.9415, "step": 6639 }, { "epoch": 0.22, "learning_rate": 1.814755903994013e-05, "loss": 0.9553, "step": 6640 }, { "epoch": 0.22, "learning_rate": 1.814693403262952e-05, "loss": 0.9674, "step": 6641 }, { "epoch": 0.22, "learning_rate": 1.814630893066555e-05, "loss": 0.9612, "step": 6642 }, { "epoch": 0.22, "learning_rate": 1.814568373405548e-05, "loss": 0.9574, "step": 6643 }, { "epoch": 0.22, "learning_rate": 1.8145058442806584e-05, "loss": 0.9524, "step": 6644 }, { "epoch": 0.22, "learning_rate": 1.8144433056926122e-05, "loss": 0.933, "step": 6645 }, { "epoch": 0.22, "learning_rate": 1.8143807576421357e-05, "loss": 0.9841, "step": 6646 }, { "epoch": 0.22, "learning_rate": 1.8143182001299566e-05, "loss": 0.9654, "step": 6647 }, { "epoch": 0.22, "learning_rate": 1.8142556331568002e-05, "loss": 0.9468, "step": 6648 }, { "epoch": 0.22, "learning_rate": 1.814193056723394e-05, "loss": 0.955, "step": 6649 }, { "epoch": 0.22, "learning_rate": 1.8141304708304657e-05, "loss": 0.9439, "step": 6650 }, { "epoch": 0.22, "learning_rate": 1.814067875478742e-05, "loss": 0.941, "step": 6651 }, { "epoch": 0.22, "learning_rate": 1.81400527066895e-05, "loss": 0.9698, "step": 6652 }, { "epoch": 0.22, "learning_rate": 1.813942656401817e-05, "loss": 0.9728, "step": 6653 }, { "epoch": 0.22, "learning_rate": 1.8138800326780706e-05, "loss": 0.9448, "step": 6654 }, { "epoch": 0.22, "learning_rate": 1.813817399498439e-05, "loss": 0.9591, "step": 6655 }, { "epoch": 0.22, "learning_rate": 1.8137547568636486e-05, "loss": 0.9645, "step": 6656 }, { "epoch": 0.22, "learning_rate": 1.8136921047744286e-05, "loss": 0.9529, "step": 6657 }, { "epoch": 0.22, "learning_rate": 1.8136294432315057e-05, "loss": 0.948, "step": 6658 }, { "epoch": 0.22, "learning_rate": 1.8135667722356084e-05, "loss": 0.9865, "step": 6659 }, { "epoch": 0.22, "learning_rate": 1.813504091787465e-05, "loss": 0.9459, "step": 6660 }, { "epoch": 0.22, "learning_rate": 1.8134414018878035e-05, "loss": 0.9711, "step": 6661 }, { "epoch": 0.22, "learning_rate": 1.813378702537353e-05, "loss": 0.9552, "step": 6662 }, { "epoch": 0.22, "learning_rate": 1.8133159937368407e-05, "loss": 0.9827, "step": 6663 }, { "epoch": 0.22, "learning_rate": 1.8132532754869956e-05, "loss": 0.9498, "step": 6664 }, { "epoch": 0.22, "learning_rate": 1.813190547788547e-05, "loss": 0.9745, "step": 6665 }, { "epoch": 0.22, "learning_rate": 1.813127810642223e-05, "loss": 0.9541, "step": 6666 }, { "epoch": 0.22, "learning_rate": 1.813065064048753e-05, "loss": 0.9911, "step": 6667 }, { "epoch": 0.22, "learning_rate": 1.8130023080088653e-05, "loss": 0.9498, "step": 6668 }, { "epoch": 0.22, "learning_rate": 1.81293954252329e-05, "loss": 0.9655, "step": 6669 }, { "epoch": 0.22, "learning_rate": 1.8128767675927553e-05, "loss": 0.9782, "step": 6670 }, { "epoch": 0.22, "learning_rate": 1.8128139832179913e-05, "loss": 0.9585, "step": 6671 }, { "epoch": 0.22, "learning_rate": 1.8127511893997274e-05, "loss": 0.9346, "step": 6672 }, { "epoch": 0.22, "learning_rate": 1.8126883861386928e-05, "loss": 0.958, "step": 6673 }, { "epoch": 0.22, "learning_rate": 1.812625573435617e-05, "loss": 0.9288, "step": 6674 }, { "epoch": 0.22, "learning_rate": 1.8125627512912302e-05, "loss": 0.9626, "step": 6675 }, { "epoch": 0.22, "learning_rate": 1.8124999197062622e-05, "loss": 0.9341, "step": 6676 }, { "epoch": 0.22, "learning_rate": 1.8124370786814434e-05, "loss": 0.9637, "step": 6677 }, { "epoch": 0.22, "learning_rate": 1.812374228217503e-05, "loss": 0.9289, "step": 6678 }, { "epoch": 0.22, "learning_rate": 1.8123113683151718e-05, "loss": 0.9916, "step": 6679 }, { "epoch": 0.22, "learning_rate": 1.81224849897518e-05, "loss": 0.9565, "step": 6680 }, { "epoch": 0.22, "learning_rate": 1.812185620198258e-05, "loss": 0.9507, "step": 6681 }, { "epoch": 0.22, "learning_rate": 1.812122731985136e-05, "loss": 0.9165, "step": 6682 }, { "epoch": 0.22, "learning_rate": 1.8120598343365454e-05, "loss": 0.9341, "step": 6683 }, { "epoch": 0.22, "learning_rate": 1.8119969272532164e-05, "loss": 0.9279, "step": 6684 }, { "epoch": 0.22, "learning_rate": 1.8119340107358804e-05, "loss": 1.0118, "step": 6685 }, { "epoch": 0.22, "learning_rate": 1.811871084785268e-05, "loss": 0.9461, "step": 6686 }, { "epoch": 0.22, "learning_rate": 1.81180814940211e-05, "loss": 0.9304, "step": 6687 }, { "epoch": 0.22, "learning_rate": 1.8117452045871384e-05, "loss": 0.9851, "step": 6688 }, { "epoch": 0.22, "learning_rate": 1.8116822503410836e-05, "loss": 0.979, "step": 6689 }, { "epoch": 0.22, "learning_rate": 1.8116192866646777e-05, "loss": 0.9649, "step": 6690 }, { "epoch": 0.22, "learning_rate": 1.811556313558652e-05, "loss": 0.9886, "step": 6691 }, { "epoch": 0.22, "learning_rate": 1.811493331023738e-05, "loss": 0.9176, "step": 6692 }, { "epoch": 0.22, "learning_rate": 1.8114303390606677e-05, "loss": 0.9212, "step": 6693 }, { "epoch": 0.22, "learning_rate": 1.8113673376701727e-05, "loss": 0.9378, "step": 6694 }, { "epoch": 0.22, "learning_rate": 1.8113043268529856e-05, "loss": 0.9928, "step": 6695 }, { "epoch": 0.22, "learning_rate": 1.8112413066098377e-05, "loss": 0.9105, "step": 6696 }, { "epoch": 0.22, "learning_rate": 1.811178276941461e-05, "loss": 0.9505, "step": 6697 }, { "epoch": 0.22, "learning_rate": 1.8111152378485887e-05, "loss": 0.9658, "step": 6698 }, { "epoch": 0.22, "learning_rate": 1.811052189331953e-05, "loss": 0.9518, "step": 6699 }, { "epoch": 0.22, "learning_rate": 1.8109891313922858e-05, "loss": 0.9396, "step": 6700 }, { "epoch": 0.22, "learning_rate": 1.81092606403032e-05, "loss": 0.9473, "step": 6701 }, { "epoch": 0.22, "learning_rate": 1.810862987246789e-05, "loss": 0.9783, "step": 6702 }, { "epoch": 0.22, "learning_rate": 1.8107999010424245e-05, "loss": 0.9773, "step": 6703 }, { "epoch": 0.22, "learning_rate": 1.8107368054179607e-05, "loss": 0.9876, "step": 6704 }, { "epoch": 0.22, "learning_rate": 1.8106737003741296e-05, "loss": 0.9449, "step": 6705 }, { "epoch": 0.22, "learning_rate": 1.8106105859116648e-05, "loss": 0.9858, "step": 6706 }, { "epoch": 0.22, "learning_rate": 1.8105474620313e-05, "loss": 0.9623, "step": 6707 }, { "epoch": 0.22, "learning_rate": 1.8104843287337676e-05, "loss": 0.9601, "step": 6708 }, { "epoch": 0.22, "learning_rate": 1.8104211860198016e-05, "loss": 0.9299, "step": 6709 }, { "epoch": 0.22, "learning_rate": 1.810358033890136e-05, "loss": 0.9436, "step": 6710 }, { "epoch": 0.22, "learning_rate": 1.8102948723455042e-05, "loss": 0.9947, "step": 6711 }, { "epoch": 0.22, "learning_rate": 1.8102317013866398e-05, "loss": 0.9642, "step": 6712 }, { "epoch": 0.22, "learning_rate": 1.810168521014277e-05, "loss": 0.9633, "step": 6713 }, { "epoch": 0.22, "learning_rate": 1.81010533122915e-05, "loss": 0.9739, "step": 6714 }, { "epoch": 0.22, "learning_rate": 1.8100421320319923e-05, "loss": 0.9778, "step": 6715 }, { "epoch": 0.22, "learning_rate": 1.8099789234235393e-05, "loss": 0.9561, "step": 6716 }, { "epoch": 0.22, "learning_rate": 1.8099157054045242e-05, "loss": 0.9934, "step": 6717 }, { "epoch": 0.22, "learning_rate": 1.8098524779756822e-05, "loss": 0.9769, "step": 6718 }, { "epoch": 0.22, "learning_rate": 1.8097892411377475e-05, "loss": 0.9653, "step": 6719 }, { "epoch": 0.22, "learning_rate": 1.809725994891455e-05, "loss": 0.9521, "step": 6720 }, { "epoch": 0.22, "learning_rate": 1.8096627392375398e-05, "loss": 0.9936, "step": 6721 }, { "epoch": 0.22, "learning_rate": 1.809599474176736e-05, "loss": 0.9678, "step": 6722 }, { "epoch": 0.22, "learning_rate": 1.8095361997097796e-05, "loss": 0.9261, "step": 6723 }, { "epoch": 0.22, "learning_rate": 1.809472915837405e-05, "loss": 0.9424, "step": 6724 }, { "epoch": 0.22, "learning_rate": 1.809409622560348e-05, "loss": 0.9667, "step": 6725 }, { "epoch": 0.22, "learning_rate": 1.8093463198793433e-05, "loss": 0.9494, "step": 6726 }, { "epoch": 0.22, "learning_rate": 1.809283007795127e-05, "loss": 0.9449, "step": 6727 }, { "epoch": 0.22, "learning_rate": 1.8092196863084342e-05, "loss": 0.9368, "step": 6728 }, { "epoch": 0.22, "learning_rate": 1.809156355420001e-05, "loss": 0.9805, "step": 6729 }, { "epoch": 0.22, "learning_rate": 1.8090930151305633e-05, "loss": 0.9552, "step": 6730 }, { "epoch": 0.22, "learning_rate": 1.8090296654408566e-05, "loss": 0.9652, "step": 6731 }, { "epoch": 0.22, "learning_rate": 1.8089663063516168e-05, "loss": 0.9393, "step": 6732 }, { "epoch": 0.22, "learning_rate": 1.8089029378635805e-05, "loss": 0.9513, "step": 6733 }, { "epoch": 0.22, "learning_rate": 1.8088395599774835e-05, "loss": 0.9418, "step": 6734 }, { "epoch": 0.22, "learning_rate": 1.8087761726940623e-05, "loss": 0.9309, "step": 6735 }, { "epoch": 0.22, "learning_rate": 1.8087127760140535e-05, "loss": 0.9419, "step": 6736 }, { "epoch": 0.22, "learning_rate": 1.8086493699381936e-05, "loss": 0.988, "step": 6737 }, { "epoch": 0.22, "learning_rate": 1.8085859544672194e-05, "loss": 0.9759, "step": 6738 }, { "epoch": 0.22, "learning_rate": 1.8085225296018675e-05, "loss": 0.9527, "step": 6739 }, { "epoch": 0.22, "learning_rate": 1.808459095342874e-05, "loss": 0.9251, "step": 6740 }, { "epoch": 0.22, "learning_rate": 1.8083956516909775e-05, "loss": 0.9458, "step": 6741 }, { "epoch": 0.22, "learning_rate": 1.8083321986469144e-05, "loss": 0.9569, "step": 6742 }, { "epoch": 0.22, "learning_rate": 1.808268736211421e-05, "loss": 0.9573, "step": 6743 }, { "epoch": 0.22, "learning_rate": 1.8082052643852362e-05, "loss": 0.9604, "step": 6744 }, { "epoch": 0.22, "learning_rate": 1.8081417831690964e-05, "loss": 0.9702, "step": 6745 }, { "epoch": 0.22, "learning_rate": 1.8080782925637394e-05, "loss": 0.9547, "step": 6746 }, { "epoch": 0.22, "learning_rate": 1.8080147925699027e-05, "loss": 0.9638, "step": 6747 }, { "epoch": 0.22, "learning_rate": 1.8079512831883243e-05, "loss": 0.9638, "step": 6748 }, { "epoch": 0.22, "learning_rate": 1.8078877644197418e-05, "loss": 0.9382, "step": 6749 }, { "epoch": 0.22, "learning_rate": 1.8078242362648937e-05, "loss": 0.9883, "step": 6750 }, { "epoch": 0.22, "learning_rate": 1.807760698724518e-05, "loss": 0.9643, "step": 6751 }, { "epoch": 0.22, "learning_rate": 1.807697151799352e-05, "loss": 0.9402, "step": 6752 }, { "epoch": 0.22, "learning_rate": 1.807633595490135e-05, "loss": 0.9255, "step": 6753 }, { "epoch": 0.22, "learning_rate": 1.807570029797605e-05, "loss": 0.9459, "step": 6754 }, { "epoch": 0.22, "learning_rate": 1.8075064547225006e-05, "loss": 0.9738, "step": 6755 }, { "epoch": 0.22, "learning_rate": 1.8074428702655606e-05, "loss": 0.955, "step": 6756 }, { "epoch": 0.22, "learning_rate": 1.8073792764275234e-05, "loss": 0.9209, "step": 6757 }, { "epoch": 0.22, "learning_rate": 1.807315673209128e-05, "loss": 0.9203, "step": 6758 }, { "epoch": 0.22, "learning_rate": 1.8072520606111134e-05, "loss": 0.9586, "step": 6759 }, { "epoch": 0.22, "learning_rate": 1.8071884386342188e-05, "loss": 0.9862, "step": 6760 }, { "epoch": 0.23, "learning_rate": 1.8071248072791828e-05, "loss": 0.9641, "step": 6761 }, { "epoch": 0.23, "learning_rate": 1.8070611665467453e-05, "loss": 0.9341, "step": 6762 }, { "epoch": 0.23, "learning_rate": 1.806997516437646e-05, "loss": 0.9783, "step": 6763 }, { "epoch": 0.23, "learning_rate": 1.8069338569526235e-05, "loss": 0.9441, "step": 6764 }, { "epoch": 0.23, "learning_rate": 1.8068701880924177e-05, "loss": 0.9872, "step": 6765 }, { "epoch": 0.23, "learning_rate": 1.806806509857769e-05, "loss": 0.9532, "step": 6766 }, { "epoch": 0.23, "learning_rate": 1.806742822249416e-05, "loss": 0.9447, "step": 6767 }, { "epoch": 0.23, "learning_rate": 1.8066791252681e-05, "loss": 0.9368, "step": 6768 }, { "epoch": 0.23, "learning_rate": 1.80661541891456e-05, "loss": 0.9581, "step": 6769 }, { "epoch": 0.23, "learning_rate": 1.8065517031895363e-05, "loss": 0.9243, "step": 6770 }, { "epoch": 0.23, "learning_rate": 1.80648797809377e-05, "loss": 0.9923, "step": 6771 }, { "epoch": 0.23, "learning_rate": 1.8064242436280004e-05, "loss": 0.9446, "step": 6772 }, { "epoch": 0.23, "learning_rate": 1.8063604997929683e-05, "loss": 0.9636, "step": 6773 }, { "epoch": 0.23, "learning_rate": 1.806296746589415e-05, "loss": 0.9906, "step": 6774 }, { "epoch": 0.23, "learning_rate": 1.80623298401808e-05, "loss": 0.9648, "step": 6775 }, { "epoch": 0.23, "learning_rate": 1.8061692120797054e-05, "loss": 0.9439, "step": 6776 }, { "epoch": 0.23, "learning_rate": 1.8061054307750315e-05, "loss": 0.9473, "step": 6777 }, { "epoch": 0.23, "learning_rate": 1.8060416401047987e-05, "loss": 0.9541, "step": 6778 }, { "epoch": 0.23, "learning_rate": 1.8059778400697493e-05, "loss": 0.9374, "step": 6779 }, { "epoch": 0.23, "learning_rate": 1.805914030670624e-05, "loss": 0.9656, "step": 6780 }, { "epoch": 0.23, "learning_rate": 1.8058502119081637e-05, "loss": 0.9962, "step": 6781 }, { "epoch": 0.23, "learning_rate": 1.8057863837831107e-05, "loss": 0.9304, "step": 6782 }, { "epoch": 0.23, "learning_rate": 1.805722546296206e-05, "loss": 0.9607, "step": 6783 }, { "epoch": 0.23, "learning_rate": 1.8056586994481916e-05, "loss": 0.9426, "step": 6784 }, { "epoch": 0.23, "learning_rate": 1.8055948432398094e-05, "loss": 0.9917, "step": 6785 }, { "epoch": 0.23, "learning_rate": 1.8055309776718006e-05, "loss": 0.9867, "step": 6786 }, { "epoch": 0.23, "learning_rate": 1.805467102744908e-05, "loss": 0.9426, "step": 6787 }, { "epoch": 0.23, "learning_rate": 1.8054032184598733e-05, "loss": 0.9376, "step": 6788 }, { "epoch": 0.23, "learning_rate": 1.8053393248174385e-05, "loss": 1.0047, "step": 6789 }, { "epoch": 0.23, "learning_rate": 1.8052754218183465e-05, "loss": 1.0002, "step": 6790 }, { "epoch": 0.23, "learning_rate": 1.8052115094633398e-05, "loss": 0.9384, "step": 6791 }, { "epoch": 0.23, "learning_rate": 1.8051475877531605e-05, "loss": 0.9042, "step": 6792 }, { "epoch": 0.23, "learning_rate": 1.8050836566885513e-05, "loss": 0.9847, "step": 6793 }, { "epoch": 0.23, "learning_rate": 1.805019716270255e-05, "loss": 0.9897, "step": 6794 }, { "epoch": 0.23, "learning_rate": 1.8049557664990147e-05, "loss": 0.9801, "step": 6795 }, { "epoch": 0.23, "learning_rate": 1.804891807375573e-05, "loss": 0.9472, "step": 6796 }, { "epoch": 0.23, "learning_rate": 1.8048278389006736e-05, "loss": 0.9457, "step": 6797 }, { "epoch": 0.23, "learning_rate": 1.804763861075059e-05, "loss": 0.9468, "step": 6798 }, { "epoch": 0.23, "learning_rate": 1.8046998738994728e-05, "loss": 0.9714, "step": 6799 }, { "epoch": 0.23, "learning_rate": 1.8046358773746585e-05, "loss": 0.9506, "step": 6800 }, { "epoch": 0.23, "learning_rate": 1.8045718715013598e-05, "loss": 0.9703, "step": 6801 }, { "epoch": 0.23, "learning_rate": 1.8045078562803203e-05, "loss": 0.9342, "step": 6802 }, { "epoch": 0.23, "learning_rate": 1.8044438317122834e-05, "loss": 1.0181, "step": 6803 }, { "epoch": 0.23, "learning_rate": 1.804379797797993e-05, "loss": 0.9475, "step": 6804 }, { "epoch": 0.23, "learning_rate": 1.8043157545381932e-05, "loss": 1.0141, "step": 6805 }, { "epoch": 0.23, "learning_rate": 1.8042517019336284e-05, "loss": 0.9169, "step": 6806 }, { "epoch": 0.23, "learning_rate": 1.8041876399850422e-05, "loss": 0.9656, "step": 6807 }, { "epoch": 0.23, "learning_rate": 1.804123568693179e-05, "loss": 0.939, "step": 6808 }, { "epoch": 0.23, "learning_rate": 1.8040594880587837e-05, "loss": 0.9929, "step": 6809 }, { "epoch": 0.23, "learning_rate": 1.8039953980826e-05, "loss": 0.9501, "step": 6810 }, { "epoch": 0.23, "learning_rate": 1.8039312987653733e-05, "loss": 0.9503, "step": 6811 }, { "epoch": 0.23, "learning_rate": 1.8038671901078482e-05, "loss": 0.9513, "step": 6812 }, { "epoch": 0.23, "learning_rate": 1.8038030721107687e-05, "loss": 0.962, "step": 6813 }, { "epoch": 0.23, "learning_rate": 1.803738944774881e-05, "loss": 0.9654, "step": 6814 }, { "epoch": 0.23, "learning_rate": 1.8036748081009294e-05, "loss": 0.9934, "step": 6815 }, { "epoch": 0.23, "learning_rate": 1.803610662089659e-05, "loss": 0.9458, "step": 6816 }, { "epoch": 0.23, "learning_rate": 1.8035465067418152e-05, "loss": 0.9213, "step": 6817 }, { "epoch": 0.23, "learning_rate": 1.8034823420581432e-05, "loss": 0.9306, "step": 6818 }, { "epoch": 0.23, "learning_rate": 1.803418168039389e-05, "loss": 0.9146, "step": 6819 }, { "epoch": 0.23, "learning_rate": 1.803353984686298e-05, "loss": 0.9294, "step": 6820 }, { "epoch": 0.23, "learning_rate": 1.8032897919996155e-05, "loss": 0.9554, "step": 6821 }, { "epoch": 0.23, "learning_rate": 1.8032255899800878e-05, "loss": 0.9305, "step": 6822 }, { "epoch": 0.23, "learning_rate": 1.8031613786284606e-05, "loss": 1.0135, "step": 6823 }, { "epoch": 0.23, "learning_rate": 1.8030971579454797e-05, "loss": 0.9774, "step": 6824 }, { "epoch": 0.23, "learning_rate": 1.8030329279318918e-05, "loss": 0.9451, "step": 6825 }, { "epoch": 0.23, "learning_rate": 1.8029686885884427e-05, "loss": 0.9701, "step": 6826 }, { "epoch": 0.23, "learning_rate": 1.8029044399158788e-05, "loss": 0.9482, "step": 6827 }, { "epoch": 0.23, "learning_rate": 1.8028401819149467e-05, "loss": 0.9261, "step": 6828 }, { "epoch": 0.23, "learning_rate": 1.8027759145863927e-05, "loss": 0.9635, "step": 6829 }, { "epoch": 0.23, "learning_rate": 1.8027116379309637e-05, "loss": 0.939, "step": 6830 }, { "epoch": 0.23, "learning_rate": 1.8026473519494065e-05, "loss": 0.9516, "step": 6831 }, { "epoch": 0.23, "learning_rate": 1.8025830566424682e-05, "loss": 0.9603, "step": 6832 }, { "epoch": 0.23, "learning_rate": 1.8025187520108952e-05, "loss": 0.9492, "step": 6833 }, { "epoch": 0.23, "learning_rate": 1.8024544380554353e-05, "loss": 0.9185, "step": 6834 }, { "epoch": 0.23, "learning_rate": 1.802390114776835e-05, "loss": 0.9467, "step": 6835 }, { "epoch": 0.23, "learning_rate": 1.8023257821758422e-05, "loss": 0.9952, "step": 6836 }, { "epoch": 0.23, "learning_rate": 1.8022614402532043e-05, "loss": 0.9255, "step": 6837 }, { "epoch": 0.23, "learning_rate": 1.8021970890096684e-05, "loss": 0.9313, "step": 6838 }, { "epoch": 0.23, "learning_rate": 1.8021327284459827e-05, "loss": 0.9884, "step": 6839 }, { "epoch": 0.23, "learning_rate": 1.8020683585628942e-05, "loss": 0.9455, "step": 6840 }, { "epoch": 0.23, "learning_rate": 1.8020039793611515e-05, "loss": 0.957, "step": 6841 }, { "epoch": 0.23, "learning_rate": 1.8019395908415025e-05, "loss": 0.9578, "step": 6842 }, { "epoch": 0.23, "learning_rate": 1.801875193004695e-05, "loss": 0.9099, "step": 6843 }, { "epoch": 0.23, "learning_rate": 1.8018107858514772e-05, "loss": 0.9553, "step": 6844 }, { "epoch": 0.23, "learning_rate": 1.8017463693825975e-05, "loss": 0.9597, "step": 6845 }, { "epoch": 0.23, "learning_rate": 1.8016819435988043e-05, "loss": 0.9398, "step": 6846 }, { "epoch": 0.23, "learning_rate": 1.8016175085008464e-05, "loss": 0.9697, "step": 6847 }, { "epoch": 0.23, "learning_rate": 1.801553064089472e-05, "loss": 0.961, "step": 6848 }, { "epoch": 0.23, "learning_rate": 1.80148861036543e-05, "loss": 0.9436, "step": 6849 }, { "epoch": 0.23, "learning_rate": 1.8014241473294692e-05, "loss": 0.9582, "step": 6850 }, { "epoch": 0.23, "learning_rate": 1.8013596749823383e-05, "loss": 0.9268, "step": 6851 }, { "epoch": 0.23, "learning_rate": 1.801295193324787e-05, "loss": 0.9573, "step": 6852 }, { "epoch": 0.23, "learning_rate": 1.801230702357564e-05, "loss": 0.9583, "step": 6853 }, { "epoch": 0.23, "learning_rate": 1.8011662020814186e-05, "loss": 0.9405, "step": 6854 }, { "epoch": 0.23, "learning_rate": 1.8011016924971005e-05, "loss": 0.9583, "step": 6855 }, { "epoch": 0.23, "learning_rate": 1.8010371736053586e-05, "loss": 0.9506, "step": 6856 }, { "epoch": 0.23, "learning_rate": 1.8009726454069432e-05, "loss": 0.9771, "step": 6857 }, { "epoch": 0.23, "learning_rate": 1.8009081079026034e-05, "loss": 0.9184, "step": 6858 }, { "epoch": 0.23, "learning_rate": 1.8008435610930893e-05, "loss": 0.9537, "step": 6859 }, { "epoch": 0.23, "learning_rate": 1.800779004979151e-05, "loss": 0.9127, "step": 6860 }, { "epoch": 0.23, "learning_rate": 1.800714439561538e-05, "loss": 0.9675, "step": 6861 }, { "epoch": 0.23, "learning_rate": 1.8006498648410012e-05, "loss": 0.907, "step": 6862 }, { "epoch": 0.23, "learning_rate": 1.80058528081829e-05, "loss": 0.9902, "step": 6863 }, { "epoch": 0.23, "learning_rate": 1.8005206874941553e-05, "loss": 0.906, "step": 6864 }, { "epoch": 0.23, "learning_rate": 1.800456084869347e-05, "loss": 0.9951, "step": 6865 }, { "epoch": 0.23, "learning_rate": 1.8003914729446167e-05, "loss": 0.9283, "step": 6866 }, { "epoch": 0.23, "learning_rate": 1.8003268517207144e-05, "loss": 1.0094, "step": 6867 }, { "epoch": 0.23, "learning_rate": 1.800262221198391e-05, "loss": 0.9362, "step": 6868 }, { "epoch": 0.23, "learning_rate": 1.800197581378397e-05, "loss": 0.9677, "step": 6869 }, { "epoch": 0.23, "learning_rate": 1.8001329322614842e-05, "loss": 0.9343, "step": 6870 }, { "epoch": 0.23, "learning_rate": 1.8000682738484028e-05, "loss": 0.9285, "step": 6871 }, { "epoch": 0.23, "learning_rate": 1.800003606139905e-05, "loss": 0.9801, "step": 6872 }, { "epoch": 0.23, "learning_rate": 1.799938929136741e-05, "loss": 0.9556, "step": 6873 }, { "epoch": 0.23, "learning_rate": 1.7998742428396633e-05, "loss": 1.0012, "step": 6874 }, { "epoch": 0.23, "learning_rate": 1.799809547249423e-05, "loss": 0.9569, "step": 6875 }, { "epoch": 0.23, "learning_rate": 1.7997448423667716e-05, "loss": 0.9404, "step": 6876 }, { "epoch": 0.23, "learning_rate": 1.7996801281924613e-05, "loss": 0.9548, "step": 6877 }, { "epoch": 0.23, "learning_rate": 1.7996154047272434e-05, "loss": 0.9568, "step": 6878 }, { "epoch": 0.23, "learning_rate": 1.7995506719718702e-05, "loss": 0.8977, "step": 6879 }, { "epoch": 0.23, "learning_rate": 1.7994859299270937e-05, "loss": 0.9234, "step": 6880 }, { "epoch": 0.23, "learning_rate": 1.7994211785936665e-05, "loss": 0.9561, "step": 6881 }, { "epoch": 0.23, "learning_rate": 1.7993564179723398e-05, "loss": 0.9755, "step": 6882 }, { "epoch": 0.23, "learning_rate": 1.7992916480638674e-05, "loss": 0.9097, "step": 6883 }, { "epoch": 0.23, "learning_rate": 1.799226868869001e-05, "loss": 0.9339, "step": 6884 }, { "epoch": 0.23, "learning_rate": 1.799162080388493e-05, "loss": 0.9807, "step": 6885 }, { "epoch": 0.23, "learning_rate": 1.7990972826230968e-05, "loss": 0.9556, "step": 6886 }, { "epoch": 0.23, "learning_rate": 1.7990324755735652e-05, "loss": 0.979, "step": 6887 }, { "epoch": 0.23, "learning_rate": 1.7989676592406504e-05, "loss": 0.9422, "step": 6888 }, { "epoch": 0.23, "learning_rate": 1.798902833625106e-05, "loss": 0.968, "step": 6889 }, { "epoch": 0.23, "learning_rate": 1.7988379987276854e-05, "loss": 0.9087, "step": 6890 }, { "epoch": 0.23, "learning_rate": 1.7987731545491412e-05, "loss": 0.9626, "step": 6891 }, { "epoch": 0.23, "learning_rate": 1.7987083010902276e-05, "loss": 0.9396, "step": 6892 }, { "epoch": 0.23, "learning_rate": 1.7986434383516977e-05, "loss": 0.9388, "step": 6893 }, { "epoch": 0.23, "learning_rate": 1.7985785663343047e-05, "loss": 0.9202, "step": 6894 }, { "epoch": 0.23, "learning_rate": 1.7985136850388025e-05, "loss": 0.9758, "step": 6895 }, { "epoch": 0.23, "learning_rate": 1.7984487944659452e-05, "loss": 0.9655, "step": 6896 }, { "epoch": 0.23, "learning_rate": 1.7983838946164865e-05, "loss": 0.9609, "step": 6897 }, { "epoch": 0.23, "learning_rate": 1.7983189854911808e-05, "loss": 0.9697, "step": 6898 }, { "epoch": 0.23, "learning_rate": 1.7982540670907817e-05, "loss": 0.9302, "step": 6899 }, { "epoch": 0.23, "learning_rate": 1.7981891394160436e-05, "loss": 0.942, "step": 6900 }, { "epoch": 0.23, "learning_rate": 1.798124202467721e-05, "loss": 0.95, "step": 6901 }, { "epoch": 0.23, "learning_rate": 1.798059256246568e-05, "loss": 0.9486, "step": 6902 }, { "epoch": 0.23, "learning_rate": 1.7979943007533397e-05, "loss": 0.9001, "step": 6903 }, { "epoch": 0.23, "learning_rate": 1.7979293359887904e-05, "loss": 0.9568, "step": 6904 }, { "epoch": 0.23, "learning_rate": 1.797864361953675e-05, "loss": 0.9691, "step": 6905 }, { "epoch": 0.23, "learning_rate": 1.7977993786487483e-05, "loss": 0.919, "step": 6906 }, { "epoch": 0.23, "learning_rate": 1.7977343860747653e-05, "loss": 1.0135, "step": 6907 }, { "epoch": 0.23, "learning_rate": 1.7976693842324814e-05, "loss": 0.9508, "step": 6908 }, { "epoch": 0.23, "learning_rate": 1.7976043731226514e-05, "loss": 0.9475, "step": 6909 }, { "epoch": 0.23, "learning_rate": 1.7975393527460306e-05, "loss": 0.9567, "step": 6910 }, { "epoch": 0.23, "learning_rate": 1.7974743231033752e-05, "loss": 1.0002, "step": 6911 }, { "epoch": 0.23, "learning_rate": 1.7974092841954395e-05, "loss": 0.9543, "step": 6912 }, { "epoch": 0.23, "learning_rate": 1.7973442360229802e-05, "loss": 0.9606, "step": 6913 }, { "epoch": 0.23, "learning_rate": 1.7972791785867523e-05, "loss": 0.9601, "step": 6914 }, { "epoch": 0.23, "learning_rate": 1.797214111887512e-05, "loss": 0.9555, "step": 6915 }, { "epoch": 0.23, "learning_rate": 1.7971490359260156e-05, "loss": 0.927, "step": 6916 }, { "epoch": 0.23, "learning_rate": 1.7970839507030186e-05, "loss": 0.9288, "step": 6917 }, { "epoch": 0.23, "learning_rate": 1.7970188562192776e-05, "loss": 0.9563, "step": 6918 }, { "epoch": 0.23, "learning_rate": 1.796953752475548e-05, "loss": 0.9496, "step": 6919 }, { "epoch": 0.23, "learning_rate": 1.7968886394725876e-05, "loss": 0.9205, "step": 6920 }, { "epoch": 0.23, "learning_rate": 1.796823517211152e-05, "loss": 0.984, "step": 6921 }, { "epoch": 0.23, "learning_rate": 1.7967583856919977e-05, "loss": 0.941, "step": 6922 }, { "epoch": 0.23, "learning_rate": 1.796693244915882e-05, "loss": 0.9431, "step": 6923 }, { "epoch": 0.23, "learning_rate": 1.7966280948835614e-05, "loss": 1.0095, "step": 6924 }, { "epoch": 0.23, "learning_rate": 1.7965629355957928e-05, "loss": 0.9187, "step": 6925 }, { "epoch": 0.23, "learning_rate": 1.7964977670533333e-05, "loss": 0.9442, "step": 6926 }, { "epoch": 0.23, "learning_rate": 1.79643258925694e-05, "loss": 0.9502, "step": 6927 }, { "epoch": 0.23, "learning_rate": 1.7963674022073703e-05, "loss": 0.9607, "step": 6928 }, { "epoch": 0.23, "learning_rate": 1.7963022059053813e-05, "loss": 0.9545, "step": 6929 }, { "epoch": 0.23, "learning_rate": 1.796237000351731e-05, "loss": 0.9408, "step": 6930 }, { "epoch": 0.23, "learning_rate": 1.796171785547176e-05, "loss": 0.9427, "step": 6931 }, { "epoch": 0.23, "learning_rate": 1.7961065614924748e-05, "loss": 0.9441, "step": 6932 }, { "epoch": 0.23, "learning_rate": 1.796041328188385e-05, "loss": 0.9517, "step": 6933 }, { "epoch": 0.23, "learning_rate": 1.7959760856356646e-05, "loss": 0.9233, "step": 6934 }, { "epoch": 0.23, "learning_rate": 1.7959108338350713e-05, "loss": 0.9455, "step": 6935 }, { "epoch": 0.23, "learning_rate": 1.7958455727873638e-05, "loss": 0.9266, "step": 6936 }, { "epoch": 0.23, "learning_rate": 1.7957803024932995e-05, "loss": 1.0018, "step": 6937 }, { "epoch": 0.23, "learning_rate": 1.7957150229536372e-05, "loss": 0.9175, "step": 6938 }, { "epoch": 0.23, "learning_rate": 1.7956497341691355e-05, "loss": 0.9482, "step": 6939 }, { "epoch": 0.23, "learning_rate": 1.7955844361405525e-05, "loss": 0.9912, "step": 6940 }, { "epoch": 0.23, "learning_rate": 1.795519128868647e-05, "loss": 0.9747, "step": 6941 }, { "epoch": 0.23, "learning_rate": 1.795453812354178e-05, "loss": 0.968, "step": 6942 }, { "epoch": 0.23, "learning_rate": 1.7953884865979042e-05, "loss": 0.9443, "step": 6943 }, { "epoch": 0.23, "learning_rate": 1.795323151600585e-05, "loss": 0.9443, "step": 6944 }, { "epoch": 0.23, "learning_rate": 1.7952578073629784e-05, "loss": 0.9647, "step": 6945 }, { "epoch": 0.23, "learning_rate": 1.7951924538858443e-05, "loss": 0.9407, "step": 6946 }, { "epoch": 0.23, "learning_rate": 1.7951270911699423e-05, "loss": 0.9591, "step": 6947 }, { "epoch": 0.23, "learning_rate": 1.795061719216031e-05, "loss": 0.9435, "step": 6948 }, { "epoch": 0.23, "learning_rate": 1.7949963380248707e-05, "loss": 0.9247, "step": 6949 }, { "epoch": 0.23, "learning_rate": 1.7949309475972207e-05, "loss": 0.9778, "step": 6950 }, { "epoch": 0.23, "learning_rate": 1.7948655479338405e-05, "loss": 0.9324, "step": 6951 }, { "epoch": 0.23, "learning_rate": 1.7948001390354898e-05, "loss": 0.9388, "step": 6952 }, { "epoch": 0.23, "learning_rate": 1.7947347209029295e-05, "loss": 0.9484, "step": 6953 }, { "epoch": 0.23, "learning_rate": 1.7946692935369187e-05, "loss": 0.9491, "step": 6954 }, { "epoch": 0.23, "learning_rate": 1.794603856938218e-05, "loss": 0.9532, "step": 6955 }, { "epoch": 0.23, "learning_rate": 1.794538411107587e-05, "loss": 0.9204, "step": 6956 }, { "epoch": 0.23, "learning_rate": 1.794472956045787e-05, "loss": 0.9975, "step": 6957 }, { "epoch": 0.23, "learning_rate": 1.794407491753578e-05, "loss": 0.9642, "step": 6958 }, { "epoch": 0.23, "learning_rate": 1.7943420182317208e-05, "loss": 0.9715, "step": 6959 }, { "epoch": 0.23, "learning_rate": 1.7942765354809757e-05, "loss": 0.9511, "step": 6960 }, { "epoch": 0.23, "learning_rate": 1.7942110435021037e-05, "loss": 0.9669, "step": 6961 }, { "epoch": 0.23, "learning_rate": 1.794145542295866e-05, "loss": 0.9187, "step": 6962 }, { "epoch": 0.23, "learning_rate": 1.7940800318630233e-05, "loss": 0.9581, "step": 6963 }, { "epoch": 0.23, "learning_rate": 1.7940145122043368e-05, "loss": 0.9384, "step": 6964 }, { "epoch": 0.23, "learning_rate": 1.7939489833205673e-05, "loss": 0.9659, "step": 6965 }, { "epoch": 0.23, "learning_rate": 1.7938834452124766e-05, "loss": 0.9861, "step": 6966 }, { "epoch": 0.23, "learning_rate": 1.7938178978808265e-05, "loss": 0.9557, "step": 6967 }, { "epoch": 0.23, "learning_rate": 1.793752341326378e-05, "loss": 0.9419, "step": 6968 }, { "epoch": 0.23, "learning_rate": 1.7936867755498925e-05, "loss": 0.9658, "step": 6969 }, { "epoch": 0.23, "learning_rate": 1.7936212005521322e-05, "loss": 0.9573, "step": 6970 }, { "epoch": 0.23, "learning_rate": 1.7935556163338592e-05, "loss": 0.9559, "step": 6971 }, { "epoch": 0.23, "learning_rate": 1.793490022895835e-05, "loss": 0.9327, "step": 6972 }, { "epoch": 0.23, "learning_rate": 1.7934244202388218e-05, "loss": 0.9715, "step": 6973 }, { "epoch": 0.23, "learning_rate": 1.793358808363582e-05, "loss": 0.9887, "step": 6974 }, { "epoch": 0.23, "learning_rate": 1.7932931872708773e-05, "loss": 1.0107, "step": 6975 }, { "epoch": 0.23, "learning_rate": 1.793227556961471e-05, "loss": 0.9541, "step": 6976 }, { "epoch": 0.23, "learning_rate": 1.7931619174361253e-05, "loss": 0.9711, "step": 6977 }, { "epoch": 0.23, "learning_rate": 1.793096268695602e-05, "loss": 0.9855, "step": 6978 }, { "epoch": 0.23, "learning_rate": 1.793030610740665e-05, "loss": 0.9856, "step": 6979 }, { "epoch": 0.23, "learning_rate": 1.7929649435720768e-05, "loss": 0.9408, "step": 6980 }, { "epoch": 0.23, "learning_rate": 1.7928992671905997e-05, "loss": 0.9389, "step": 6981 }, { "epoch": 0.23, "learning_rate": 1.7928335815969977e-05, "loss": 0.941, "step": 6982 }, { "epoch": 0.23, "learning_rate": 1.7927678867920333e-05, "loss": 0.942, "step": 6983 }, { "epoch": 0.23, "learning_rate": 1.7927021827764696e-05, "loss": 0.9528, "step": 6984 }, { "epoch": 0.23, "learning_rate": 1.7926364695510706e-05, "loss": 0.9768, "step": 6985 }, { "epoch": 0.23, "learning_rate": 1.7925707471165993e-05, "loss": 0.931, "step": 6986 }, { "epoch": 0.23, "learning_rate": 1.79250501547382e-05, "loss": 0.9881, "step": 6987 }, { "epoch": 0.23, "learning_rate": 1.792439274623495e-05, "loss": 0.9766, "step": 6988 }, { "epoch": 0.23, "learning_rate": 1.7923735245663897e-05, "loss": 0.9864, "step": 6989 }, { "epoch": 0.23, "learning_rate": 1.7923077653032666e-05, "loss": 0.9543, "step": 6990 }, { "epoch": 0.23, "learning_rate": 1.7922419968348907e-05, "loss": 0.9334, "step": 6991 }, { "epoch": 0.23, "learning_rate": 1.792176219162026e-05, "loss": 0.9669, "step": 6992 }, { "epoch": 0.23, "learning_rate": 1.792110432285436e-05, "loss": 0.9551, "step": 6993 }, { "epoch": 0.23, "learning_rate": 1.7920446362058858e-05, "loss": 0.962, "step": 6994 }, { "epoch": 0.23, "learning_rate": 1.7919788309241397e-05, "loss": 0.9525, "step": 6995 }, { "epoch": 0.23, "learning_rate": 1.7919130164409618e-05, "loss": 0.9371, "step": 6996 }, { "epoch": 0.23, "learning_rate": 1.791847192757117e-05, "loss": 1.0002, "step": 6997 }, { "epoch": 0.23, "learning_rate": 1.79178135987337e-05, "loss": 0.9175, "step": 6998 }, { "epoch": 0.23, "learning_rate": 1.7917155177904863e-05, "loss": 0.9504, "step": 6999 }, { "epoch": 0.23, "learning_rate": 1.7916496665092302e-05, "loss": 0.9739, "step": 7000 }, { "epoch": 0.0, "learning_rate": 1.7915838060303666e-05, "loss": 0.9546, "step": 7001 }, { "epoch": 0.0, "learning_rate": 1.7915179363546612e-05, "loss": 0.9574, "step": 7002 }, { "epoch": 0.0, "learning_rate": 1.7914520574828797e-05, "loss": 0.9645, "step": 7003 }, { "epoch": 0.0, "learning_rate": 1.791386169415786e-05, "loss": 0.9419, "step": 7004 }, { "epoch": 0.0, "learning_rate": 1.791320272154147e-05, "loss": 0.9585, "step": 7005 }, { "epoch": 0.0, "learning_rate": 1.7912543656987274e-05, "loss": 0.9252, "step": 7006 }, { "epoch": 0.0, "learning_rate": 1.791188450050294e-05, "loss": 1.0023, "step": 7007 }, { "epoch": 0.0, "learning_rate": 1.7911225252096115e-05, "loss": 0.9314, "step": 7008 }, { "epoch": 0.0, "learning_rate": 1.7910565911774466e-05, "loss": 1.0044, "step": 7009 }, { "epoch": 0.0, "learning_rate": 1.7909906479545646e-05, "loss": 0.9386, "step": 7010 }, { "epoch": 0.0, "learning_rate": 1.7909246955417322e-05, "loss": 0.9501, "step": 7011 }, { "epoch": 0.0, "learning_rate": 1.790858733939716e-05, "loss": 0.9283, "step": 7012 }, { "epoch": 0.0, "learning_rate": 1.7907927631492816e-05, "loss": 0.96, "step": 7013 }, { "epoch": 0.0, "learning_rate": 1.7907267831711954e-05, "loss": 0.9672, "step": 7014 }, { "epoch": 0.0, "learning_rate": 1.790660794006225e-05, "loss": 0.9579, "step": 7015 }, { "epoch": 0.0, "learning_rate": 1.7905947956551362e-05, "loss": 0.9479, "step": 7016 }, { "epoch": 0.0, "learning_rate": 1.790528788118696e-05, "loss": 0.9694, "step": 7017 }, { "epoch": 0.0, "learning_rate": 1.790462771397671e-05, "loss": 0.9661, "step": 7018 }, { "epoch": 0.0, "learning_rate": 1.7903967454928287e-05, "loss": 0.9755, "step": 7019 }, { "epoch": 0.0, "learning_rate": 1.7903307104049362e-05, "loss": 0.9127, "step": 7020 }, { "epoch": 0.0, "learning_rate": 1.7902646661347604e-05, "loss": 0.9619, "step": 7021 }, { "epoch": 0.0, "learning_rate": 1.7901986126830686e-05, "loss": 0.9304, "step": 7022 }, { "epoch": 0.0, "learning_rate": 1.7901325500506286e-05, "loss": 0.9711, "step": 7023 }, { "epoch": 0.0, "learning_rate": 1.7900664782382074e-05, "loss": 0.9557, "step": 7024 }, { "epoch": 0.0, "learning_rate": 1.7900003972465736e-05, "loss": 0.991, "step": 7025 }, { "epoch": 0.0, "learning_rate": 1.7899343070764938e-05, "loss": 0.922, "step": 7026 }, { "epoch": 0.0, "learning_rate": 1.7898682077287365e-05, "loss": 0.9658, "step": 7027 }, { "epoch": 0.0, "learning_rate": 1.7898020992040693e-05, "loss": 0.9471, "step": 7028 }, { "epoch": 0.0, "learning_rate": 1.7897359815032605e-05, "loss": 0.9577, "step": 7029 }, { "epoch": 0.0, "learning_rate": 1.7896698546270784e-05, "loss": 0.9155, "step": 7030 }, { "epoch": 0.0, "learning_rate": 1.7896037185762915e-05, "loss": 0.9609, "step": 7031 }, { "epoch": 0.0, "learning_rate": 1.7895375733516675e-05, "loss": 0.8877, "step": 7032 }, { "epoch": 0.0, "learning_rate": 1.789471418953975e-05, "loss": 0.9569, "step": 7033 }, { "epoch": 0.0, "learning_rate": 1.789405255383983e-05, "loss": 0.9616, "step": 7034 }, { "epoch": 0.0, "learning_rate": 1.7893390826424598e-05, "loss": 0.9833, "step": 7035 }, { "epoch": 0.0, "learning_rate": 1.7892729007301748e-05, "loss": 0.9494, "step": 7036 }, { "epoch": 0.0, "learning_rate": 1.789206709647896e-05, "loss": 0.9612, "step": 7037 }, { "epoch": 0.0, "learning_rate": 1.789140509396394e-05, "loss": 0.9677, "step": 7038 }, { "epoch": 0.0, "learning_rate": 1.789074299976436e-05, "loss": 1.006, "step": 7039 }, { "epoch": 0.0, "learning_rate": 1.7890080813887927e-05, "loss": 0.9434, "step": 7040 }, { "epoch": 0.0, "learning_rate": 1.7889418536342324e-05, "loss": 0.962, "step": 7041 }, { "epoch": 0.0, "learning_rate": 1.7888756167135254e-05, "loss": 0.9442, "step": 7042 }, { "epoch": 0.0, "learning_rate": 1.788809370627441e-05, "loss": 0.9597, "step": 7043 }, { "epoch": 0.0, "learning_rate": 1.7887431153767484e-05, "loss": 0.9275, "step": 7044 }, { "epoch": 0.0, "learning_rate": 1.7886768509622183e-05, "loss": 1.0039, "step": 7045 }, { "epoch": 0.0, "learning_rate": 1.7886105773846195e-05, "loss": 0.9967, "step": 7046 }, { "epoch": 0.0, "learning_rate": 1.7885442946447227e-05, "loss": 0.9714, "step": 7047 }, { "epoch": 0.0, "learning_rate": 1.788478002743298e-05, "loss": 0.9773, "step": 7048 }, { "epoch": 0.0, "learning_rate": 1.788411701681115e-05, "loss": 1.0003, "step": 7049 }, { "epoch": 0.0, "learning_rate": 1.7883453914589446e-05, "loss": 0.9128, "step": 7050 }, { "epoch": 0.0, "learning_rate": 1.788279072077557e-05, "loss": 0.9562, "step": 7051 }, { "epoch": 0.0, "learning_rate": 1.788212743537723e-05, "loss": 0.9806, "step": 7052 }, { "epoch": 0.0, "learning_rate": 1.788146405840213e-05, "loss": 0.9878, "step": 7053 }, { "epoch": 0.0, "learning_rate": 1.7880800589857973e-05, "loss": 0.9884, "step": 7054 }, { "epoch": 0.0, "learning_rate": 1.7880137029752472e-05, "loss": 1.0008, "step": 7055 }, { "epoch": 0.0, "learning_rate": 1.7879473378093336e-05, "loss": 0.9473, "step": 7056 }, { "epoch": 0.0, "learning_rate": 1.7878809634888274e-05, "loss": 0.9687, "step": 7057 }, { "epoch": 0.0, "learning_rate": 1.7878145800145002e-05, "loss": 0.9217, "step": 7058 }, { "epoch": 0.0, "learning_rate": 1.7877481873871228e-05, "loss": 0.9494, "step": 7059 }, { "epoch": 0.0, "learning_rate": 1.7876817856074663e-05, "loss": 0.9395, "step": 7060 }, { "epoch": 0.0, "learning_rate": 1.787615374676303e-05, "loss": 0.9737, "step": 7061 }, { "epoch": 0.0, "learning_rate": 1.7875489545944043e-05, "loss": 0.9827, "step": 7062 }, { "epoch": 0.0, "learning_rate": 1.7874825253625414e-05, "loss": 0.8962, "step": 7063 }, { "epoch": 0.0, "learning_rate": 1.787416086981486e-05, "loss": 0.9192, "step": 7064 }, { "epoch": 0.0, "learning_rate": 1.787349639452011e-05, "loss": 0.9421, "step": 7065 }, { "epoch": 0.0, "learning_rate": 1.7872831827748874e-05, "loss": 0.9815, "step": 7066 }, { "epoch": 0.0, "learning_rate": 1.7872167169508875e-05, "loss": 0.9737, "step": 7067 }, { "epoch": 0.0, "learning_rate": 1.7871502419807837e-05, "loss": 0.9293, "step": 7068 }, { "epoch": 0.0, "learning_rate": 1.7870837578653485e-05, "loss": 0.9617, "step": 7069 }, { "epoch": 0.0, "learning_rate": 1.787017264605354e-05, "loss": 0.9157, "step": 7070 }, { "epoch": 0.0, "learning_rate": 1.786950762201573e-05, "loss": 0.9636, "step": 7071 }, { "epoch": 0.0, "learning_rate": 1.786884250654778e-05, "loss": 0.9366, "step": 7072 }, { "epoch": 0.0, "learning_rate": 1.7868177299657414e-05, "loss": 0.9991, "step": 7073 }, { "epoch": 0.0, "learning_rate": 1.7867512001352367e-05, "loss": 0.9757, "step": 7074 }, { "epoch": 0.0, "learning_rate": 1.7866846611640366e-05, "loss": 1.0334, "step": 7075 }, { "epoch": 0.0, "learning_rate": 1.7866181130529142e-05, "loss": 0.918, "step": 7076 }, { "epoch": 0.0, "learning_rate": 1.786551555802643e-05, "loss": 0.9743, "step": 7077 }, { "epoch": 0.0, "learning_rate": 1.7864849894139952e-05, "loss": 0.9464, "step": 7078 }, { "epoch": 0.0, "learning_rate": 1.786418413887745e-05, "loss": 0.9842, "step": 7079 }, { "epoch": 0.0, "learning_rate": 1.786351829224666e-05, "loss": 0.935, "step": 7080 }, { "epoch": 0.0, "learning_rate": 1.7862852354255315e-05, "loss": 0.949, "step": 7081 }, { "epoch": 0.0, "learning_rate": 1.7862186324911154e-05, "loss": 0.9701, "step": 7082 }, { "epoch": 0.0, "learning_rate": 1.786152020422191e-05, "loss": 0.9308, "step": 7083 }, { "epoch": 0.0, "learning_rate": 1.786085399219533e-05, "loss": 0.9372, "step": 7084 }, { "epoch": 0.0, "learning_rate": 1.7860187688839153e-05, "loss": 0.9817, "step": 7085 }, { "epoch": 0.0, "learning_rate": 1.7859521294161114e-05, "loss": 0.9327, "step": 7086 }, { "epoch": 0.0, "learning_rate": 1.7858854808168958e-05, "loss": 0.9862, "step": 7087 }, { "epoch": 0.0, "learning_rate": 1.785818823087043e-05, "loss": 0.9504, "step": 7088 }, { "epoch": 0.0, "learning_rate": 1.7857521562273278e-05, "loss": 0.968, "step": 7089 }, { "epoch": 0.0, "learning_rate": 1.7856854802385237e-05, "loss": 0.954, "step": 7090 }, { "epoch": 0.0, "learning_rate": 1.7856187951214064e-05, "loss": 0.9892, "step": 7091 }, { "epoch": 0.0, "learning_rate": 1.7855521008767502e-05, "loss": 0.9564, "step": 7092 }, { "epoch": 0.0, "learning_rate": 1.7854853975053302e-05, "loss": 0.9123, "step": 7093 }, { "epoch": 0.0, "learning_rate": 1.785418685007921e-05, "loss": 0.9656, "step": 7094 }, { "epoch": 0.0, "learning_rate": 1.7853519633852982e-05, "loss": 0.9596, "step": 7095 }, { "epoch": 0.0, "learning_rate": 1.7852852326382362e-05, "loss": 0.9504, "step": 7096 }, { "epoch": 0.0, "learning_rate": 1.7852184927675113e-05, "loss": 0.9869, "step": 7097 }, { "epoch": 0.0, "learning_rate": 1.7851517437738983e-05, "loss": 0.97, "step": 7098 }, { "epoch": 0.0, "learning_rate": 1.785084985658173e-05, "loss": 0.9604, "step": 7099 }, { "epoch": 0.0, "learning_rate": 1.7850182184211104e-05, "loss": 0.9285, "step": 7100 }, { "epoch": 0.0, "learning_rate": 1.7849514420634868e-05, "loss": 0.9173, "step": 7101 }, { "epoch": 0.0, "learning_rate": 1.784884656586078e-05, "loss": 0.9741, "step": 7102 }, { "epoch": 0.0, "learning_rate": 1.7848178619896596e-05, "loss": 0.9556, "step": 7103 }, { "epoch": 0.0, "learning_rate": 1.784751058275008e-05, "loss": 0.9287, "step": 7104 }, { "epoch": 0.0, "learning_rate": 1.784684245442899e-05, "loss": 0.9828, "step": 7105 }, { "epoch": 0.0, "learning_rate": 1.7846174234941092e-05, "loss": 0.952, "step": 7106 }, { "epoch": 0.0, "learning_rate": 1.7845505924294147e-05, "loss": 1.0023, "step": 7107 }, { "epoch": 0.0, "learning_rate": 1.7844837522495923e-05, "loss": 0.9437, "step": 7108 }, { "epoch": 0.0, "learning_rate": 1.784416902955418e-05, "loss": 0.976, "step": 7109 }, { "epoch": 0.0, "learning_rate": 1.784350044547669e-05, "loss": 0.9881, "step": 7110 }, { "epoch": 0.0, "learning_rate": 1.784283177027122e-05, "loss": 0.9513, "step": 7111 }, { "epoch": 0.0, "learning_rate": 1.7842163003945532e-05, "loss": 0.9701, "step": 7112 }, { "epoch": 0.0, "learning_rate": 1.7841494146507408e-05, "loss": 0.9808, "step": 7113 }, { "epoch": 0.0, "learning_rate": 1.784082519796461e-05, "loss": 0.957, "step": 7114 }, { "epoch": 0.0, "learning_rate": 1.784015615832491e-05, "loss": 0.952, "step": 7115 }, { "epoch": 0.0, "learning_rate": 1.7839487027596088e-05, "loss": 0.9717, "step": 7116 }, { "epoch": 0.0, "learning_rate": 1.783881780578591e-05, "loss": 0.9888, "step": 7117 }, { "epoch": 0.0, "learning_rate": 1.7838148492902156e-05, "loss": 0.914, "step": 7118 }, { "epoch": 0.0, "learning_rate": 1.7837479088952604e-05, "loss": 1.0293, "step": 7119 }, { "epoch": 0.0, "learning_rate": 1.7836809593945025e-05, "loss": 0.9329, "step": 7120 }, { "epoch": 0.0, "learning_rate": 1.7836140007887205e-05, "loss": 0.9921, "step": 7121 }, { "epoch": 0.0, "learning_rate": 1.7835470330786915e-05, "loss": 0.9369, "step": 7122 }, { "epoch": 0.0, "learning_rate": 1.7834800562651943e-05, "loss": 0.9859, "step": 7123 }, { "epoch": 0.0, "learning_rate": 1.7834130703490066e-05, "loss": 0.9371, "step": 7124 }, { "epoch": 0.0, "learning_rate": 1.783346075330907e-05, "loss": 0.9782, "step": 7125 }, { "epoch": 0.0, "learning_rate": 1.7832790712116735e-05, "loss": 0.9171, "step": 7126 }, { "epoch": 0.0, "learning_rate": 1.7832120579920843e-05, "loss": 0.9515, "step": 7127 }, { "epoch": 0.0, "learning_rate": 1.7831450356729193e-05, "loss": 0.9427, "step": 7128 }, { "epoch": 0.0, "learning_rate": 1.783078004254956e-05, "loss": 0.9112, "step": 7129 }, { "epoch": 0.0, "learning_rate": 1.7830109637389733e-05, "loss": 0.9316, "step": 7130 }, { "epoch": 0.0, "learning_rate": 1.7829439141257502e-05, "loss": 0.9462, "step": 7131 }, { "epoch": 0.0, "learning_rate": 1.782876855416066e-05, "loss": 0.9284, "step": 7132 }, { "epoch": 0.0, "learning_rate": 1.7828097876106992e-05, "loss": 0.9578, "step": 7133 }, { "epoch": 0.0, "learning_rate": 1.78274271071043e-05, "loss": 0.9637, "step": 7134 }, { "epoch": 0.0, "learning_rate": 1.782675624716037e-05, "loss": 0.9642, "step": 7135 }, { "epoch": 0.0, "learning_rate": 1.7826085296282998e-05, "loss": 0.9645, "step": 7136 }, { "epoch": 0.0, "learning_rate": 1.7825414254479975e-05, "loss": 0.9464, "step": 7137 }, { "epoch": 0.0, "learning_rate": 1.7824743121759106e-05, "loss": 0.9393, "step": 7138 }, { "epoch": 0.0, "learning_rate": 1.7824071898128182e-05, "loss": 0.9529, "step": 7139 }, { "epoch": 0.0, "learning_rate": 1.7823400583595e-05, "loss": 0.9343, "step": 7140 }, { "epoch": 0.0, "learning_rate": 1.7822729178167364e-05, "loss": 0.9339, "step": 7141 }, { "epoch": 0.0, "learning_rate": 1.7822057681853075e-05, "loss": 0.9581, "step": 7142 }, { "epoch": 0.0, "learning_rate": 1.782138609465993e-05, "loss": 0.9551, "step": 7143 }, { "epoch": 0.0, "learning_rate": 1.782071441659574e-05, "loss": 0.9229, "step": 7144 }, { "epoch": 0.0, "learning_rate": 1.7820042647668297e-05, "loss": 0.9555, "step": 7145 }, { "epoch": 0.0, "learning_rate": 1.7819370787885418e-05, "loss": 0.9454, "step": 7146 }, { "epoch": 0.0, "learning_rate": 1.7818698837254898e-05, "loss": 0.9745, "step": 7147 }, { "epoch": 0.0, "learning_rate": 1.781802679578455e-05, "loss": 0.9716, "step": 7148 }, { "epoch": 0.0, "learning_rate": 1.7817354663482182e-05, "loss": 0.9573, "step": 7149 }, { "epoch": 0.0, "learning_rate": 1.78166824403556e-05, "loss": 0.9739, "step": 7150 }, { "epoch": 0.01, "learning_rate": 1.7816010126412618e-05, "loss": 0.9796, "step": 7151 }, { "epoch": 0.01, "learning_rate": 1.7815337721661043e-05, "loss": 0.9252, "step": 7152 }, { "epoch": 0.01, "learning_rate": 1.781466522610869e-05, "loss": 0.9485, "step": 7153 }, { "epoch": 0.01, "learning_rate": 1.781399263976337e-05, "loss": 0.9362, "step": 7154 }, { "epoch": 0.01, "learning_rate": 1.78133199626329e-05, "loss": 0.9397, "step": 7155 }, { "epoch": 0.01, "learning_rate": 1.7812647194725093e-05, "loss": 0.9688, "step": 7156 }, { "epoch": 0.01, "learning_rate": 1.781197433604777e-05, "loss": 0.9512, "step": 7157 }, { "epoch": 0.01, "learning_rate": 1.781130138660874e-05, "loss": 0.9507, "step": 7158 }, { "epoch": 0.01, "learning_rate": 1.781062834641583e-05, "loss": 0.953, "step": 7159 }, { "epoch": 0.01, "learning_rate": 1.7809955215476855e-05, "loss": 0.9578, "step": 7160 }, { "epoch": 0.01, "learning_rate": 1.7809281993799637e-05, "loss": 0.9393, "step": 7161 }, { "epoch": 0.01, "learning_rate": 1.7808608681391998e-05, "loss": 0.943, "step": 7162 }, { "epoch": 0.01, "learning_rate": 1.780793527826176e-05, "loss": 0.9694, "step": 7163 }, { "epoch": 0.01, "learning_rate": 1.7807261784416743e-05, "loss": 0.9494, "step": 7164 }, { "epoch": 0.01, "learning_rate": 1.780658819986478e-05, "loss": 0.9774, "step": 7165 }, { "epoch": 0.01, "learning_rate": 1.7805914524613688e-05, "loss": 0.9633, "step": 7166 }, { "epoch": 0.01, "learning_rate": 1.7805240758671303e-05, "loss": 0.9366, "step": 7167 }, { "epoch": 0.01, "learning_rate": 1.780456690204545e-05, "loss": 0.979, "step": 7168 }, { "epoch": 0.01, "learning_rate": 1.780389295474395e-05, "loss": 0.9467, "step": 7169 }, { "epoch": 0.01, "learning_rate": 1.7803218916774643e-05, "loss": 0.9519, "step": 7170 }, { "epoch": 0.01, "learning_rate": 1.7802544788145358e-05, "loss": 0.9184, "step": 7171 }, { "epoch": 0.01, "learning_rate": 1.7801870568863923e-05, "loss": 0.962, "step": 7172 }, { "epoch": 0.01, "learning_rate": 1.780119625893818e-05, "loss": 0.9724, "step": 7173 }, { "epoch": 0.01, "learning_rate": 1.7800521858375953e-05, "loss": 0.9897, "step": 7174 }, { "epoch": 0.01, "learning_rate": 1.779984736718508e-05, "loss": 0.9583, "step": 7175 }, { "epoch": 0.01, "learning_rate": 1.7799172785373405e-05, "loss": 0.9078, "step": 7176 }, { "epoch": 0.01, "learning_rate": 1.7798498112948757e-05, "loss": 0.9576, "step": 7177 }, { "epoch": 0.01, "learning_rate": 1.7797823349918977e-05, "loss": 0.9355, "step": 7178 }, { "epoch": 0.01, "learning_rate": 1.7797148496291905e-05, "loss": 0.944, "step": 7179 }, { "epoch": 0.01, "learning_rate": 1.7796473552075385e-05, "loss": 0.954, "step": 7180 }, { "epoch": 0.01, "learning_rate": 1.7795798517277252e-05, "loss": 0.9498, "step": 7181 }, { "epoch": 0.01, "learning_rate": 1.7795123391905354e-05, "loss": 0.9567, "step": 7182 }, { "epoch": 0.01, "learning_rate": 1.7794448175967532e-05, "loss": 0.9455, "step": 7183 }, { "epoch": 0.01, "learning_rate": 1.7793772869471632e-05, "loss": 0.942, "step": 7184 }, { "epoch": 0.01, "learning_rate": 1.7793097472425497e-05, "loss": 0.9598, "step": 7185 }, { "epoch": 0.01, "learning_rate": 1.779242198483698e-05, "loss": 0.9165, "step": 7186 }, { "epoch": 0.01, "learning_rate": 1.7791746406713922e-05, "loss": 0.9684, "step": 7187 }, { "epoch": 0.01, "learning_rate": 1.7791070738064182e-05, "loss": 0.9252, "step": 7188 }, { "epoch": 0.01, "learning_rate": 1.77903949788956e-05, "loss": 1.0044, "step": 7189 }, { "epoch": 0.01, "learning_rate": 1.778971912921603e-05, "loss": 0.941, "step": 7190 }, { "epoch": 0.01, "learning_rate": 1.7789043189033325e-05, "loss": 0.9568, "step": 7191 }, { "epoch": 0.01, "learning_rate": 1.778836715835534e-05, "loss": 0.9217, "step": 7192 }, { "epoch": 0.01, "learning_rate": 1.7787691037189928e-05, "loss": 0.9227, "step": 7193 }, { "epoch": 0.01, "learning_rate": 1.778701482554494e-05, "loss": 0.923, "step": 7194 }, { "epoch": 0.01, "learning_rate": 1.7786338523428245e-05, "loss": 0.9537, "step": 7195 }, { "epoch": 0.01, "learning_rate": 1.7785662130847685e-05, "loss": 0.9646, "step": 7196 }, { "epoch": 0.01, "learning_rate": 1.7784985647811126e-05, "loss": 0.9105, "step": 7197 }, { "epoch": 0.01, "learning_rate": 1.778430907432643e-05, "loss": 0.9627, "step": 7198 }, { "epoch": 0.01, "learning_rate": 1.7783632410401454e-05, "loss": 0.918, "step": 7199 }, { "epoch": 0.01, "learning_rate": 1.7782955656044056e-05, "loss": 0.9989, "step": 7200 }, { "epoch": 0.01, "learning_rate": 1.778227881126211e-05, "loss": 0.938, "step": 7201 }, { "epoch": 0.01, "learning_rate": 1.778160187606347e-05, "loss": 0.9841, "step": 7202 }, { "epoch": 0.01, "learning_rate": 1.7780924850456e-05, "loss": 0.9189, "step": 7203 }, { "epoch": 0.01, "learning_rate": 1.7780247734447572e-05, "loss": 0.922, "step": 7204 }, { "epoch": 0.01, "learning_rate": 1.777957052804605e-05, "loss": 0.9436, "step": 7205 }, { "epoch": 0.01, "learning_rate": 1.7778893231259305e-05, "loss": 0.9654, "step": 7206 }, { "epoch": 0.01, "learning_rate": 1.7778215844095203e-05, "loss": 0.9363, "step": 7207 }, { "epoch": 0.01, "learning_rate": 1.777753836656161e-05, "loss": 0.9304, "step": 7208 }, { "epoch": 0.01, "learning_rate": 1.7776860798666407e-05, "loss": 0.9555, "step": 7209 }, { "epoch": 0.01, "learning_rate": 1.777618314041746e-05, "loss": 0.9577, "step": 7210 }, { "epoch": 0.01, "learning_rate": 1.7775505391822642e-05, "loss": 0.9247, "step": 7211 }, { "epoch": 0.01, "learning_rate": 1.7774827552889825e-05, "loss": 0.9102, "step": 7212 }, { "epoch": 0.01, "learning_rate": 1.777414962362689e-05, "loss": 0.9106, "step": 7213 }, { "epoch": 0.01, "learning_rate": 1.7773471604041715e-05, "loss": 0.9497, "step": 7214 }, { "epoch": 0.01, "learning_rate": 1.777279349414217e-05, "loss": 0.9595, "step": 7215 }, { "epoch": 0.01, "learning_rate": 1.7772115293936137e-05, "loss": 0.9067, "step": 7216 }, { "epoch": 0.01, "learning_rate": 1.7771437003431495e-05, "loss": 0.9617, "step": 7217 }, { "epoch": 0.01, "learning_rate": 1.7770758622636126e-05, "loss": 0.9108, "step": 7218 }, { "epoch": 0.01, "learning_rate": 1.7770080151557908e-05, "loss": 0.9441, "step": 7219 }, { "epoch": 0.01, "learning_rate": 1.776940159020473e-05, "loss": 0.9035, "step": 7220 }, { "epoch": 0.01, "learning_rate": 1.776872293858447e-05, "loss": 0.9332, "step": 7221 }, { "epoch": 0.01, "learning_rate": 1.7768044196705015e-05, "loss": 0.9573, "step": 7222 }, { "epoch": 0.01, "learning_rate": 1.776736536457425e-05, "loss": 0.9506, "step": 7223 }, { "epoch": 0.01, "learning_rate": 1.7766686442200064e-05, "loss": 0.8981, "step": 7224 }, { "epoch": 0.01, "learning_rate": 1.7766007429590342e-05, "loss": 0.9206, "step": 7225 }, { "epoch": 0.01, "learning_rate": 1.7765328326752975e-05, "loss": 0.8936, "step": 7226 }, { "epoch": 0.01, "learning_rate": 1.7764649133695853e-05, "loss": 0.9451, "step": 7227 }, { "epoch": 0.01, "learning_rate": 1.7763969850426862e-05, "loss": 0.9561, "step": 7228 }, { "epoch": 0.01, "learning_rate": 1.7763290476953904e-05, "loss": 0.9681, "step": 7229 }, { "epoch": 0.01, "learning_rate": 1.7762611013284867e-05, "loss": 0.9355, "step": 7230 }, { "epoch": 0.01, "learning_rate": 1.776193145942764e-05, "loss": 0.9699, "step": 7231 }, { "epoch": 0.01, "learning_rate": 1.7761251815390125e-05, "loss": 0.9072, "step": 7232 }, { "epoch": 0.01, "learning_rate": 1.7760572081180218e-05, "loss": 0.9645, "step": 7233 }, { "epoch": 0.01, "learning_rate": 1.7759892256805814e-05, "loss": 0.8983, "step": 7234 }, { "epoch": 0.01, "learning_rate": 1.775921234227481e-05, "loss": 0.9347, "step": 7235 }, { "epoch": 0.01, "learning_rate": 1.775853233759511e-05, "loss": 0.9079, "step": 7236 }, { "epoch": 0.01, "learning_rate": 1.775785224277461e-05, "loss": 0.9531, "step": 7237 }, { "epoch": 0.01, "learning_rate": 1.7757172057821213e-05, "loss": 0.9713, "step": 7238 }, { "epoch": 0.01, "learning_rate": 1.7756491782742825e-05, "loss": 0.9465, "step": 7239 }, { "epoch": 0.01, "learning_rate": 1.7755811417547345e-05, "loss": 0.9554, "step": 7240 }, { "epoch": 0.01, "learning_rate": 1.775513096224268e-05, "loss": 0.9199, "step": 7241 }, { "epoch": 0.01, "learning_rate": 1.7754450416836735e-05, "loss": 0.9482, "step": 7242 }, { "epoch": 0.01, "learning_rate": 1.775376978133742e-05, "loss": 0.9282, "step": 7243 }, { "epoch": 0.01, "learning_rate": 1.7753089055752635e-05, "loss": 0.917, "step": 7244 }, { "epoch": 0.01, "learning_rate": 1.7752408240090294e-05, "loss": 0.94, "step": 7245 }, { "epoch": 0.01, "learning_rate": 1.775172733435831e-05, "loss": 0.9122, "step": 7246 }, { "epoch": 0.01, "learning_rate": 1.7751046338564587e-05, "loss": 0.9781, "step": 7247 }, { "epoch": 0.01, "learning_rate": 1.775036525271704e-05, "loss": 0.9385, "step": 7248 }, { "epoch": 0.01, "learning_rate": 1.7749684076823584e-05, "loss": 0.9754, "step": 7249 }, { "epoch": 0.01, "learning_rate": 1.7749002810892133e-05, "loss": 0.9594, "step": 7250 }, { "epoch": 0.01, "learning_rate": 1.77483214549306e-05, "loss": 0.9325, "step": 7251 }, { "epoch": 0.01, "learning_rate": 1.77476400089469e-05, "loss": 0.9557, "step": 7252 }, { "epoch": 0.01, "learning_rate": 1.7746958472948954e-05, "loss": 0.9311, "step": 7253 }, { "epoch": 0.01, "learning_rate": 1.7746276846944677e-05, "loss": 0.9144, "step": 7254 }, { "epoch": 0.01, "learning_rate": 1.774559513094199e-05, "loss": 0.9531, "step": 7255 }, { "epoch": 0.01, "learning_rate": 1.7744913324948814e-05, "loss": 0.9458, "step": 7256 }, { "epoch": 0.01, "learning_rate": 1.774423142897307e-05, "loss": 0.9535, "step": 7257 }, { "epoch": 0.01, "learning_rate": 1.7743549443022678e-05, "loss": 0.9545, "step": 7258 }, { "epoch": 0.01, "learning_rate": 1.7742867367105566e-05, "loss": 0.9396, "step": 7259 }, { "epoch": 0.01, "learning_rate": 1.7742185201229654e-05, "loss": 0.9439, "step": 7260 }, { "epoch": 0.01, "learning_rate": 1.774150294540287e-05, "loss": 0.958, "step": 7261 }, { "epoch": 0.01, "learning_rate": 1.774082059963314e-05, "loss": 0.9423, "step": 7262 }, { "epoch": 0.01, "learning_rate": 1.7740138163928394e-05, "loss": 0.938, "step": 7263 }, { "epoch": 0.01, "learning_rate": 1.7739455638296553e-05, "loss": 0.9822, "step": 7264 }, { "epoch": 0.01, "learning_rate": 1.773877302274556e-05, "loss": 0.9862, "step": 7265 }, { "epoch": 0.01, "learning_rate": 1.7738090317283334e-05, "loss": 0.9477, "step": 7266 }, { "epoch": 0.01, "learning_rate": 1.7737407521917812e-05, "loss": 0.9481, "step": 7267 }, { "epoch": 0.01, "learning_rate": 1.7736724636656927e-05, "loss": 0.9624, "step": 7268 }, { "epoch": 0.01, "learning_rate": 1.7736041661508608e-05, "loss": 0.9198, "step": 7269 }, { "epoch": 0.01, "learning_rate": 1.77353585964808e-05, "loss": 0.9631, "step": 7270 }, { "epoch": 0.01, "learning_rate": 1.773467544158143e-05, "loss": 0.9797, "step": 7271 }, { "epoch": 0.01, "learning_rate": 1.7733992196818434e-05, "loss": 0.9208, "step": 7272 }, { "epoch": 0.01, "learning_rate": 1.7733308862199757e-05, "loss": 0.9777, "step": 7273 }, { "epoch": 0.01, "learning_rate": 1.7732625437733338e-05, "loss": 0.9598, "step": 7274 }, { "epoch": 0.01, "learning_rate": 1.7731941923427114e-05, "loss": 0.961, "step": 7275 }, { "epoch": 0.01, "learning_rate": 1.7731258319289028e-05, "loss": 0.9672, "step": 7276 }, { "epoch": 0.01, "learning_rate": 1.7730574625327018e-05, "loss": 0.9347, "step": 7277 }, { "epoch": 0.01, "learning_rate": 1.7729890841549033e-05, "loss": 0.9336, "step": 7278 }, { "epoch": 0.01, "learning_rate": 1.772920696796301e-05, "loss": 0.9321, "step": 7279 }, { "epoch": 0.01, "learning_rate": 1.7728523004576903e-05, "loss": 0.9461, "step": 7280 }, { "epoch": 0.01, "learning_rate": 1.7727838951398655e-05, "loss": 0.9457, "step": 7281 }, { "epoch": 0.01, "learning_rate": 1.7727154808436213e-05, "loss": 0.937, "step": 7282 }, { "epoch": 0.01, "learning_rate": 1.772647057569753e-05, "loss": 0.9514, "step": 7283 }, { "epoch": 0.01, "learning_rate": 1.7725786253190546e-05, "loss": 0.9666, "step": 7284 }, { "epoch": 0.01, "learning_rate": 1.7725101840923214e-05, "loss": 0.9623, "step": 7285 }, { "epoch": 0.01, "learning_rate": 1.7724417338903495e-05, "loss": 0.9256, "step": 7286 }, { "epoch": 0.01, "learning_rate": 1.7723732747139335e-05, "loss": 0.9429, "step": 7287 }, { "epoch": 0.01, "learning_rate": 1.7723048065638688e-05, "loss": 0.91, "step": 7288 }, { "epoch": 0.01, "learning_rate": 1.7722363294409507e-05, "loss": 0.9754, "step": 7289 }, { "epoch": 0.01, "learning_rate": 1.7721678433459753e-05, "loss": 0.9309, "step": 7290 }, { "epoch": 0.01, "learning_rate": 1.7720993482797378e-05, "loss": 0.9459, "step": 7291 }, { "epoch": 0.01, "learning_rate": 1.7720308442430342e-05, "loss": 0.956, "step": 7292 }, { "epoch": 0.01, "learning_rate": 1.7719623312366606e-05, "loss": 0.9584, "step": 7293 }, { "epoch": 0.01, "learning_rate": 1.7718938092614126e-05, "loss": 0.9057, "step": 7294 }, { "epoch": 0.01, "learning_rate": 1.7718252783180868e-05, "loss": 0.9708, "step": 7295 }, { "epoch": 0.01, "learning_rate": 1.771756738407479e-05, "loss": 0.8876, "step": 7296 }, { "epoch": 0.01, "learning_rate": 1.7716881895303852e-05, "loss": 0.9244, "step": 7297 }, { "epoch": 0.01, "learning_rate": 1.7716196316876026e-05, "loss": 0.9313, "step": 7298 }, { "epoch": 0.01, "learning_rate": 1.7715510648799273e-05, "loss": 1.0229, "step": 7299 }, { "epoch": 0.01, "learning_rate": 1.7714824891081558e-05, "loss": 0.9292, "step": 7300 }, { "epoch": 0.01, "learning_rate": 1.7714139043730855e-05, "loss": 0.9566, "step": 7301 }, { "epoch": 0.01, "learning_rate": 1.7713453106755124e-05, "loss": 0.9045, "step": 7302 }, { "epoch": 0.01, "learning_rate": 1.771276708016234e-05, "loss": 0.9528, "step": 7303 }, { "epoch": 0.01, "learning_rate": 1.771208096396047e-05, "loss": 0.9454, "step": 7304 }, { "epoch": 0.01, "learning_rate": 1.7711394758157488e-05, "loss": 0.9406, "step": 7305 }, { "epoch": 0.01, "learning_rate": 1.7710708462761365e-05, "loss": 0.9644, "step": 7306 }, { "epoch": 0.01, "learning_rate": 1.7710022077780076e-05, "loss": 0.9032, "step": 7307 }, { "epoch": 0.01, "learning_rate": 1.7709335603221597e-05, "loss": 0.9419, "step": 7308 }, { "epoch": 0.01, "learning_rate": 1.7708649039093897e-05, "loss": 0.9771, "step": 7309 }, { "epoch": 0.01, "learning_rate": 1.7707962385404958e-05, "loss": 0.8961, "step": 7310 }, { "epoch": 0.01, "learning_rate": 1.770727564216276e-05, "loss": 1.0151, "step": 7311 }, { "epoch": 0.01, "learning_rate": 1.7706588809375277e-05, "loss": 0.9494, "step": 7312 }, { "epoch": 0.01, "learning_rate": 1.770590188705049e-05, "loss": 0.9606, "step": 7313 }, { "epoch": 0.01, "learning_rate": 1.770521487519638e-05, "loss": 0.957, "step": 7314 }, { "epoch": 0.01, "learning_rate": 1.770452777382093e-05, "loss": 0.9284, "step": 7315 }, { "epoch": 0.01, "learning_rate": 1.770384058293212e-05, "loss": 0.9645, "step": 7316 }, { "epoch": 0.01, "learning_rate": 1.7703153302537937e-05, "loss": 0.9354, "step": 7317 }, { "epoch": 0.01, "learning_rate": 1.7702465932646365e-05, "loss": 0.9693, "step": 7318 }, { "epoch": 0.01, "learning_rate": 1.7701778473265392e-05, "loss": 0.9367, "step": 7319 }, { "epoch": 0.01, "learning_rate": 1.7701090924403e-05, "loss": 0.8892, "step": 7320 }, { "epoch": 0.01, "learning_rate": 1.7700403286067186e-05, "loss": 0.9358, "step": 7321 }, { "epoch": 0.01, "learning_rate": 1.7699715558265928e-05, "loss": 0.9286, "step": 7322 }, { "epoch": 0.01, "learning_rate": 1.7699027741007223e-05, "loss": 0.9309, "step": 7323 }, { "epoch": 0.01, "learning_rate": 1.7698339834299064e-05, "loss": 0.9358, "step": 7324 }, { "epoch": 0.01, "learning_rate": 1.769765183814944e-05, "loss": 0.9737, "step": 7325 }, { "epoch": 0.01, "learning_rate": 1.769696375256634e-05, "loss": 0.9413, "step": 7326 }, { "epoch": 0.01, "learning_rate": 1.7696275577557767e-05, "loss": 0.927, "step": 7327 }, { "epoch": 0.01, "learning_rate": 1.769558731313171e-05, "loss": 0.939, "step": 7328 }, { "epoch": 0.01, "learning_rate": 1.769489895929617e-05, "loss": 0.9311, "step": 7329 }, { "epoch": 0.01, "learning_rate": 1.769421051605914e-05, "loss": 0.904, "step": 7330 }, { "epoch": 0.01, "learning_rate": 1.7693521983428627e-05, "loss": 0.9517, "step": 7331 }, { "epoch": 0.01, "learning_rate": 1.769283336141262e-05, "loss": 0.899, "step": 7332 }, { "epoch": 0.01, "learning_rate": 1.7692144650019125e-05, "loss": 0.9573, "step": 7333 }, { "epoch": 0.01, "learning_rate": 1.769145584925614e-05, "loss": 0.9686, "step": 7334 }, { "epoch": 0.01, "learning_rate": 1.7690766959131673e-05, "loss": 0.9623, "step": 7335 }, { "epoch": 0.01, "learning_rate": 1.7690077979653725e-05, "loss": 0.912, "step": 7336 }, { "epoch": 0.01, "learning_rate": 1.76893889108303e-05, "loss": 0.9373, "step": 7337 }, { "epoch": 0.01, "learning_rate": 1.7688699752669407e-05, "loss": 0.8891, "step": 7338 }, { "epoch": 0.01, "learning_rate": 1.7688010505179047e-05, "loss": 0.9197, "step": 7339 }, { "epoch": 0.01, "learning_rate": 1.7687321168367233e-05, "loss": 0.9164, "step": 7340 }, { "epoch": 0.01, "learning_rate": 1.768663174224197e-05, "loss": 0.9504, "step": 7341 }, { "epoch": 0.01, "learning_rate": 1.7685942226811274e-05, "loss": 0.9199, "step": 7342 }, { "epoch": 0.01, "learning_rate": 1.7685252622083152e-05, "loss": 0.9426, "step": 7343 }, { "epoch": 0.01, "learning_rate": 1.7684562928065614e-05, "loss": 0.9136, "step": 7344 }, { "epoch": 0.01, "learning_rate": 1.7683873144766677e-05, "loss": 0.9329, "step": 7345 }, { "epoch": 0.01, "learning_rate": 1.768318327219435e-05, "loss": 0.9403, "step": 7346 }, { "epoch": 0.01, "learning_rate": 1.7682493310356656e-05, "loss": 0.9476, "step": 7347 }, { "epoch": 0.01, "learning_rate": 1.7681803259261605e-05, "loss": 0.9194, "step": 7348 }, { "epoch": 0.01, "learning_rate": 1.7681113118917217e-05, "loss": 0.9875, "step": 7349 }, { "epoch": 0.01, "learning_rate": 1.7680422889331507e-05, "loss": 0.9579, "step": 7350 }, { "epoch": 0.01, "learning_rate": 1.76797325705125e-05, "loss": 0.9323, "step": 7351 }, { "epoch": 0.01, "learning_rate": 1.7679042162468207e-05, "loss": 0.9113, "step": 7352 }, { "epoch": 0.01, "learning_rate": 1.7678351665206655e-05, "loss": 0.9036, "step": 7353 }, { "epoch": 0.01, "learning_rate": 1.767766107873587e-05, "loss": 0.9146, "step": 7354 }, { "epoch": 0.01, "learning_rate": 1.7676970403063872e-05, "loss": 0.9335, "step": 7355 }, { "epoch": 0.01, "learning_rate": 1.7676279638198687e-05, "loss": 0.9457, "step": 7356 }, { "epoch": 0.01, "learning_rate": 1.7675588784148338e-05, "loss": 0.9426, "step": 7357 }, { "epoch": 0.01, "learning_rate": 1.767489784092085e-05, "loss": 0.9606, "step": 7358 }, { "epoch": 0.01, "learning_rate": 1.7674206808524252e-05, "loss": 1.0047, "step": 7359 }, { "epoch": 0.01, "learning_rate": 1.7673515686966577e-05, "loss": 0.9303, "step": 7360 }, { "epoch": 0.01, "learning_rate": 1.767282447625585e-05, "loss": 0.9054, "step": 7361 }, { "epoch": 0.01, "learning_rate": 1.7672133176400103e-05, "loss": 0.937, "step": 7362 }, { "epoch": 0.01, "learning_rate": 1.7671441787407368e-05, "loss": 0.9324, "step": 7363 }, { "epoch": 0.01, "learning_rate": 1.7670750309285675e-05, "loss": 0.9195, "step": 7364 }, { "epoch": 0.01, "learning_rate": 1.7670058742043065e-05, "loss": 0.9498, "step": 7365 }, { "epoch": 0.01, "learning_rate": 1.7669367085687564e-05, "loss": 0.9769, "step": 7366 }, { "epoch": 0.01, "learning_rate": 1.7668675340227213e-05, "loss": 0.9409, "step": 7367 }, { "epoch": 0.01, "learning_rate": 1.7667983505670044e-05, "loss": 0.9444, "step": 7368 }, { "epoch": 0.01, "learning_rate": 1.7667291582024105e-05, "loss": 0.9298, "step": 7369 }, { "epoch": 0.01, "learning_rate": 1.7666599569297425e-05, "loss": 0.9286, "step": 7370 }, { "epoch": 0.01, "learning_rate": 1.766590746749805e-05, "loss": 0.9747, "step": 7371 }, { "epoch": 0.01, "learning_rate": 1.7665215276634014e-05, "loss": 0.9461, "step": 7372 }, { "epoch": 0.01, "learning_rate": 1.766452299671337e-05, "loss": 0.9405, "step": 7373 }, { "epoch": 0.01, "learning_rate": 1.766383062774415e-05, "loss": 0.9277, "step": 7374 }, { "epoch": 0.01, "learning_rate": 1.7663138169734402e-05, "loss": 0.9131, "step": 7375 }, { "epoch": 0.01, "learning_rate": 1.7662445622692177e-05, "loss": 0.9481, "step": 7376 }, { "epoch": 0.01, "learning_rate": 1.7661752986625515e-05, "loss": 0.975, "step": 7377 }, { "epoch": 0.01, "learning_rate": 1.7661060261542463e-05, "loss": 0.9415, "step": 7378 }, { "epoch": 0.01, "learning_rate": 1.766036744745107e-05, "loss": 0.9328, "step": 7379 }, { "epoch": 0.01, "learning_rate": 1.7659674544359386e-05, "loss": 0.8919, "step": 7380 }, { "epoch": 0.01, "learning_rate": 1.7658981552275466e-05, "loss": 0.9672, "step": 7381 }, { "epoch": 0.01, "learning_rate": 1.7658288471207353e-05, "loss": 0.9153, "step": 7382 }, { "epoch": 0.01, "learning_rate": 1.7657595301163105e-05, "loss": 0.8978, "step": 7383 }, { "epoch": 0.01, "learning_rate": 1.765690204215077e-05, "loss": 0.9129, "step": 7384 }, { "epoch": 0.01, "learning_rate": 1.7656208694178412e-05, "loss": 0.9475, "step": 7385 }, { "epoch": 0.01, "learning_rate": 1.7655515257254076e-05, "loss": 0.9605, "step": 7386 }, { "epoch": 0.01, "learning_rate": 1.7654821731385824e-05, "loss": 0.9406, "step": 7387 }, { "epoch": 0.01, "learning_rate": 1.7654128116581716e-05, "loss": 0.9879, "step": 7388 }, { "epoch": 0.01, "learning_rate": 1.7653434412849804e-05, "loss": 0.9366, "step": 7389 }, { "epoch": 0.01, "learning_rate": 1.7652740620198155e-05, "loss": 0.9431, "step": 7390 }, { "epoch": 0.01, "learning_rate": 1.7652046738634823e-05, "loss": 0.9588, "step": 7391 }, { "epoch": 0.01, "learning_rate": 1.765135276816787e-05, "loss": 0.9485, "step": 7392 }, { "epoch": 0.01, "learning_rate": 1.7650658708805365e-05, "loss": 0.9463, "step": 7393 }, { "epoch": 0.01, "learning_rate": 1.7649964560555366e-05, "loss": 0.953, "step": 7394 }, { "epoch": 0.01, "learning_rate": 1.764927032342594e-05, "loss": 0.9014, "step": 7395 }, { "epoch": 0.01, "learning_rate": 1.7648575997425155e-05, "loss": 0.8451, "step": 7396 }, { "epoch": 0.01, "learning_rate": 1.7647881582561074e-05, "loss": 0.9407, "step": 7397 }, { "epoch": 0.01, "learning_rate": 1.7647187078841766e-05, "loss": 0.9576, "step": 7398 }, { "epoch": 0.01, "learning_rate": 1.76464924862753e-05, "loss": 0.9371, "step": 7399 }, { "epoch": 0.01, "learning_rate": 1.7645797804869746e-05, "loss": 0.9334, "step": 7400 }, { "epoch": 0.01, "learning_rate": 1.7645103034633175e-05, "loss": 0.9538, "step": 7401 }, { "epoch": 0.01, "learning_rate": 1.7644408175573662e-05, "loss": 0.9144, "step": 7402 }, { "epoch": 0.01, "learning_rate": 1.7643713227699276e-05, "loss": 0.9344, "step": 7403 }, { "epoch": 0.01, "learning_rate": 1.7643018191018093e-05, "loss": 0.9455, "step": 7404 }, { "epoch": 0.01, "learning_rate": 1.7642323065538187e-05, "loss": 0.9608, "step": 7405 }, { "epoch": 0.01, "learning_rate": 1.7641627851267635e-05, "loss": 0.9835, "step": 7406 }, { "epoch": 0.01, "learning_rate": 1.7640932548214513e-05, "loss": 0.9487, "step": 7407 }, { "epoch": 0.01, "learning_rate": 1.7640237156386902e-05, "loss": 0.9115, "step": 7408 }, { "epoch": 0.01, "learning_rate": 1.763954167579288e-05, "loss": 0.9777, "step": 7409 }, { "epoch": 0.01, "learning_rate": 1.7638846106440522e-05, "loss": 0.8872, "step": 7410 }, { "epoch": 0.01, "learning_rate": 1.7638150448337917e-05, "loss": 0.9519, "step": 7411 }, { "epoch": 0.01, "learning_rate": 1.7637454701493147e-05, "loss": 0.911, "step": 7412 }, { "epoch": 0.01, "learning_rate": 1.763675886591429e-05, "loss": 0.9235, "step": 7413 }, { "epoch": 0.01, "learning_rate": 1.7636062941609436e-05, "loss": 0.9451, "step": 7414 }, { "epoch": 0.01, "learning_rate": 1.7635366928586663e-05, "loss": 0.9741, "step": 7415 }, { "epoch": 0.01, "learning_rate": 1.7634670826854065e-05, "loss": 0.9425, "step": 7416 }, { "epoch": 0.01, "learning_rate": 1.763397463641973e-05, "loss": 0.9091, "step": 7417 }, { "epoch": 0.01, "learning_rate": 1.7633278357291737e-05, "loss": 0.938, "step": 7418 }, { "epoch": 0.01, "learning_rate": 1.7632581989478186e-05, "loss": 0.9217, "step": 7419 }, { "epoch": 0.01, "learning_rate": 1.763188553298716e-05, "loss": 0.9005, "step": 7420 }, { "epoch": 0.01, "learning_rate": 1.7631188987826756e-05, "loss": 0.9624, "step": 7421 }, { "epoch": 0.01, "learning_rate": 1.7630492354005065e-05, "loss": 0.9159, "step": 7422 }, { "epoch": 0.01, "learning_rate": 1.7629795631530183e-05, "loss": 0.9458, "step": 7423 }, { "epoch": 0.01, "learning_rate": 1.7629098820410195e-05, "loss": 0.9055, "step": 7424 }, { "epoch": 0.01, "learning_rate": 1.7628401920653207e-05, "loss": 0.9872, "step": 7425 }, { "epoch": 0.01, "learning_rate": 1.7627704932267316e-05, "loss": 0.9213, "step": 7426 }, { "epoch": 0.01, "learning_rate": 1.7627007855260614e-05, "loss": 0.9465, "step": 7427 }, { "epoch": 0.01, "learning_rate": 1.7626310689641203e-05, "loss": 0.9408, "step": 7428 }, { "epoch": 0.01, "learning_rate": 1.762561343541718e-05, "loss": 0.9614, "step": 7429 }, { "epoch": 0.01, "learning_rate": 1.7624916092596647e-05, "loss": 0.966, "step": 7430 }, { "epoch": 0.01, "learning_rate": 1.762421866118771e-05, "loss": 0.9613, "step": 7431 }, { "epoch": 0.01, "learning_rate": 1.762352114119847e-05, "loss": 0.9672, "step": 7432 }, { "epoch": 0.01, "learning_rate": 1.7622823532637024e-05, "loss": 0.939, "step": 7433 }, { "epoch": 0.01, "learning_rate": 1.7622125835511487e-05, "loss": 0.9762, "step": 7434 }, { "epoch": 0.01, "learning_rate": 1.7621428049829963e-05, "loss": 0.9849, "step": 7435 }, { "epoch": 0.01, "learning_rate": 1.7620730175600553e-05, "loss": 0.9324, "step": 7436 }, { "epoch": 0.01, "learning_rate": 1.762003221283137e-05, "loss": 0.9941, "step": 7437 }, { "epoch": 0.01, "learning_rate": 1.7619334161530524e-05, "loss": 0.9192, "step": 7438 }, { "epoch": 0.01, "learning_rate": 1.761863602170612e-05, "loss": 0.924, "step": 7439 }, { "epoch": 0.01, "learning_rate": 1.7617937793366275e-05, "loss": 0.925, "step": 7440 }, { "epoch": 0.01, "learning_rate": 1.7617239476519097e-05, "loss": 0.9067, "step": 7441 }, { "epoch": 0.01, "learning_rate": 1.7616541071172704e-05, "loss": 0.8885, "step": 7442 }, { "epoch": 0.01, "learning_rate": 1.7615842577335204e-05, "loss": 0.9168, "step": 7443 }, { "epoch": 0.01, "learning_rate": 1.7615143995014717e-05, "loss": 0.9375, "step": 7444 }, { "epoch": 0.01, "learning_rate": 1.761444532421936e-05, "loss": 0.9384, "step": 7445 }, { "epoch": 0.01, "learning_rate": 1.7613746564957247e-05, "loss": 0.9257, "step": 7446 }, { "epoch": 0.01, "learning_rate": 1.7613047717236496e-05, "loss": 0.9868, "step": 7447 }, { "epoch": 0.01, "learning_rate": 1.7612348781065227e-05, "loss": 0.9152, "step": 7448 }, { "epoch": 0.01, "learning_rate": 1.7611649756451562e-05, "loss": 0.9106, "step": 7449 }, { "epoch": 0.01, "learning_rate": 1.7610950643403622e-05, "loss": 0.8854, "step": 7450 }, { "epoch": 0.02, "learning_rate": 1.7610251441929532e-05, "loss": 0.9932, "step": 7451 }, { "epoch": 0.02, "learning_rate": 1.7609552152037413e-05, "loss": 0.9309, "step": 7452 }, { "epoch": 0.02, "learning_rate": 1.7608852773735387e-05, "loss": 0.9441, "step": 7453 }, { "epoch": 0.02, "learning_rate": 1.7608153307031585e-05, "loss": 0.9468, "step": 7454 }, { "epoch": 0.02, "learning_rate": 1.7607453751934124e-05, "loss": 0.9492, "step": 7455 }, { "epoch": 0.02, "learning_rate": 1.7606754108451144e-05, "loss": 0.9453, "step": 7456 }, { "epoch": 0.02, "learning_rate": 1.7606054376590765e-05, "loss": 0.9434, "step": 7457 }, { "epoch": 0.02, "learning_rate": 1.760535455636112e-05, "loss": 0.9463, "step": 7458 }, { "epoch": 0.02, "learning_rate": 1.7604654647770342e-05, "loss": 0.9586, "step": 7459 }, { "epoch": 0.02, "learning_rate": 1.760395465082656e-05, "loss": 0.9247, "step": 7460 }, { "epoch": 0.02, "learning_rate": 1.7603254565537904e-05, "loss": 0.9852, "step": 7461 }, { "epoch": 0.02, "learning_rate": 1.7602554391912513e-05, "loss": 0.9888, "step": 7462 }, { "epoch": 0.02, "learning_rate": 1.760185412995852e-05, "loss": 0.9571, "step": 7463 }, { "epoch": 0.02, "learning_rate": 1.7601153779684055e-05, "loss": 0.9459, "step": 7464 }, { "epoch": 0.02, "learning_rate": 1.7600453341097264e-05, "loss": 0.9614, "step": 7465 }, { "epoch": 0.02, "learning_rate": 1.7599752814206283e-05, "loss": 0.9534, "step": 7466 }, { "epoch": 0.02, "learning_rate": 1.7599052199019245e-05, "loss": 0.9119, "step": 7467 }, { "epoch": 0.02, "learning_rate": 1.7598351495544293e-05, "loss": 0.9184, "step": 7468 }, { "epoch": 0.02, "learning_rate": 1.759765070378957e-05, "loss": 0.9452, "step": 7469 }, { "epoch": 0.02, "learning_rate": 1.7596949823763222e-05, "loss": 0.937, "step": 7470 }, { "epoch": 0.02, "learning_rate": 1.759624885547338e-05, "loss": 0.9548, "step": 7471 }, { "epoch": 0.02, "learning_rate": 1.7595547798928198e-05, "loss": 0.9116, "step": 7472 }, { "epoch": 0.02, "learning_rate": 1.759484665413582e-05, "loss": 0.9353, "step": 7473 }, { "epoch": 0.02, "learning_rate": 1.7594145421104386e-05, "loss": 0.9393, "step": 7474 }, { "epoch": 0.02, "learning_rate": 1.759344409984205e-05, "loss": 0.9518, "step": 7475 }, { "epoch": 0.02, "learning_rate": 1.7592742690356955e-05, "loss": 0.9786, "step": 7476 }, { "epoch": 0.02, "learning_rate": 1.7592041192657253e-05, "loss": 0.9553, "step": 7477 }, { "epoch": 0.02, "learning_rate": 1.7591339606751095e-05, "loss": 0.9717, "step": 7478 }, { "epoch": 0.02, "learning_rate": 1.7590637932646628e-05, "loss": 0.9555, "step": 7479 }, { "epoch": 0.02, "learning_rate": 1.758993617035201e-05, "loss": 0.9379, "step": 7480 }, { "epoch": 0.02, "learning_rate": 1.758923431987539e-05, "loss": 0.9172, "step": 7481 }, { "epoch": 0.02, "learning_rate": 1.7588532381224925e-05, "loss": 0.9397, "step": 7482 }, { "epoch": 0.02, "learning_rate": 1.758783035440877e-05, "loss": 0.9924, "step": 7483 }, { "epoch": 0.02, "learning_rate": 1.758712823943508e-05, "loss": 0.9009, "step": 7484 }, { "epoch": 0.02, "learning_rate": 1.758642603631201e-05, "loss": 0.9448, "step": 7485 }, { "epoch": 0.02, "learning_rate": 1.7585723745047724e-05, "loss": 0.94, "step": 7486 }, { "epoch": 0.02, "learning_rate": 1.7585021365650377e-05, "loss": 0.9404, "step": 7487 }, { "epoch": 0.02, "learning_rate": 1.758431889812813e-05, "loss": 0.9318, "step": 7488 }, { "epoch": 0.02, "learning_rate": 1.758361634248915e-05, "loss": 0.946, "step": 7489 }, { "epoch": 0.02, "learning_rate": 1.7582913698741588e-05, "loss": 0.8883, "step": 7490 }, { "epoch": 0.02, "learning_rate": 1.7582210966893618e-05, "loss": 0.9521, "step": 7491 }, { "epoch": 0.02, "learning_rate": 1.7581508146953402e-05, "loss": 0.9237, "step": 7492 }, { "epoch": 0.02, "learning_rate": 1.7580805238929102e-05, "loss": 0.9344, "step": 7493 }, { "epoch": 0.02, "learning_rate": 1.758010224282889e-05, "loss": 0.9417, "step": 7494 }, { "epoch": 0.02, "learning_rate": 1.7579399158660927e-05, "loss": 0.9738, "step": 7495 }, { "epoch": 0.02, "learning_rate": 1.757869598643339e-05, "loss": 0.9196, "step": 7496 }, { "epoch": 0.02, "learning_rate": 1.757799272615444e-05, "loss": 0.9417, "step": 7497 }, { "epoch": 0.02, "learning_rate": 1.7577289377832255e-05, "loss": 0.9334, "step": 7498 }, { "epoch": 0.02, "learning_rate": 1.7576585941475003e-05, "loss": 0.9623, "step": 7499 }, { "epoch": 0.02, "learning_rate": 1.7575882417090854e-05, "loss": 0.947, "step": 7500 }, { "epoch": 0.02, "learning_rate": 1.757517880468799e-05, "loss": 0.9368, "step": 7501 }, { "epoch": 0.02, "learning_rate": 1.7574475104274575e-05, "loss": 0.9228, "step": 7502 }, { "epoch": 0.02, "learning_rate": 1.7573771315858794e-05, "loss": 0.9807, "step": 7503 }, { "epoch": 0.02, "learning_rate": 1.757306743944882e-05, "loss": 0.8705, "step": 7504 }, { "epoch": 0.02, "learning_rate": 1.7572363475052828e-05, "loss": 0.9598, "step": 7505 }, { "epoch": 0.02, "learning_rate": 1.7571659422679003e-05, "loss": 0.9258, "step": 7506 }, { "epoch": 0.02, "learning_rate": 1.757095528233552e-05, "loss": 0.9267, "step": 7507 }, { "epoch": 0.02, "learning_rate": 1.7570251054030564e-05, "loss": 0.9318, "step": 7508 }, { "epoch": 0.02, "learning_rate": 1.7569546737772315e-05, "loss": 0.9448, "step": 7509 }, { "epoch": 0.02, "learning_rate": 1.7568842333568952e-05, "loss": 0.943, "step": 7510 }, { "epoch": 0.02, "learning_rate": 1.7568137841428662e-05, "loss": 0.9559, "step": 7511 }, { "epoch": 0.02, "learning_rate": 1.7567433261359633e-05, "loss": 0.9468, "step": 7512 }, { "epoch": 0.02, "learning_rate": 1.7566728593370048e-05, "loss": 0.9444, "step": 7513 }, { "epoch": 0.02, "learning_rate": 1.7566023837468096e-05, "loss": 0.919, "step": 7514 }, { "epoch": 0.02, "learning_rate": 1.7565318993661966e-05, "loss": 0.9115, "step": 7515 }, { "epoch": 0.02, "learning_rate": 1.756461406195984e-05, "loss": 0.9148, "step": 7516 }, { "epoch": 0.02, "learning_rate": 1.756390904236991e-05, "loss": 0.9275, "step": 7517 }, { "epoch": 0.02, "learning_rate": 1.7563203934900377e-05, "loss": 0.898, "step": 7518 }, { "epoch": 0.02, "learning_rate": 1.756249873955942e-05, "loss": 0.9475, "step": 7519 }, { "epoch": 0.02, "learning_rate": 1.7561793456355242e-05, "loss": 0.904, "step": 7520 }, { "epoch": 0.02, "learning_rate": 1.7561088085296032e-05, "loss": 0.9451, "step": 7521 }, { "epoch": 0.02, "learning_rate": 1.7560382626389988e-05, "loss": 0.9226, "step": 7522 }, { "epoch": 0.02, "learning_rate": 1.7559677079645303e-05, "loss": 0.8928, "step": 7523 }, { "epoch": 0.02, "learning_rate": 1.7558971445070175e-05, "loss": 0.9091, "step": 7524 }, { "epoch": 0.02, "learning_rate": 1.7558265722672805e-05, "loss": 0.9416, "step": 7525 }, { "epoch": 0.02, "learning_rate": 1.755755991246139e-05, "loss": 0.9161, "step": 7526 }, { "epoch": 0.02, "learning_rate": 1.755685401444413e-05, "loss": 0.9768, "step": 7527 }, { "epoch": 0.02, "learning_rate": 1.755614802862923e-05, "loss": 0.919, "step": 7528 }, { "epoch": 0.02, "learning_rate": 1.7555441955024888e-05, "loss": 0.8766, "step": 7529 }, { "epoch": 0.02, "learning_rate": 1.755473579363931e-05, "loss": 0.9148, "step": 7530 }, { "epoch": 0.02, "learning_rate": 1.7554029544480698e-05, "loss": 0.9453, "step": 7531 }, { "epoch": 0.02, "learning_rate": 1.7553323207557258e-05, "loss": 0.901, "step": 7532 }, { "epoch": 0.02, "learning_rate": 1.7552616782877194e-05, "loss": 1.0203, "step": 7533 }, { "epoch": 0.02, "learning_rate": 1.755191027044872e-05, "loss": 0.9492, "step": 7534 }, { "epoch": 0.02, "learning_rate": 1.7551203670280045e-05, "loss": 0.9457, "step": 7535 }, { "epoch": 0.02, "learning_rate": 1.755049698237937e-05, "loss": 0.9521, "step": 7536 }, { "epoch": 0.02, "learning_rate": 1.754979020675491e-05, "loss": 0.9529, "step": 7537 }, { "epoch": 0.02, "learning_rate": 1.7549083343414873e-05, "loss": 0.8806, "step": 7538 }, { "epoch": 0.02, "learning_rate": 1.754837639236748e-05, "loss": 0.9457, "step": 7539 }, { "epoch": 0.02, "learning_rate": 1.7547669353620938e-05, "loss": 0.9552, "step": 7540 }, { "epoch": 0.02, "learning_rate": 1.7546962227183465e-05, "loss": 0.912, "step": 7541 }, { "epoch": 0.02, "learning_rate": 1.7546255013063273e-05, "loss": 0.9299, "step": 7542 }, { "epoch": 0.02, "learning_rate": 1.754554771126858e-05, "loss": 0.9471, "step": 7543 }, { "epoch": 0.02, "learning_rate": 1.7544840321807605e-05, "loss": 0.9582, "step": 7544 }, { "epoch": 0.02, "learning_rate": 1.7544132844688565e-05, "loss": 0.9355, "step": 7545 }, { "epoch": 0.02, "learning_rate": 1.754342527991968e-05, "loss": 0.9082, "step": 7546 }, { "epoch": 0.02, "learning_rate": 1.754271762750917e-05, "loss": 0.9125, "step": 7547 }, { "epoch": 0.02, "learning_rate": 1.754200988746526e-05, "loss": 0.9249, "step": 7548 }, { "epoch": 0.02, "learning_rate": 1.754130205979617e-05, "loss": 0.9295, "step": 7549 }, { "epoch": 0.02, "learning_rate": 1.7540594144510122e-05, "loss": 0.8901, "step": 7550 }, { "epoch": 0.02, "learning_rate": 1.7539886141615343e-05, "loss": 0.9206, "step": 7551 }, { "epoch": 0.02, "learning_rate": 1.753917805112006e-05, "loss": 0.9185, "step": 7552 }, { "epoch": 0.02, "learning_rate": 1.7538469873032498e-05, "loss": 0.9431, "step": 7553 }, { "epoch": 0.02, "learning_rate": 1.753776160736089e-05, "loss": 0.9448, "step": 7554 }, { "epoch": 0.02, "learning_rate": 1.7537053254113452e-05, "loss": 0.9748, "step": 7555 }, { "epoch": 0.02, "learning_rate": 1.7536344813298427e-05, "loss": 0.9432, "step": 7556 }, { "epoch": 0.02, "learning_rate": 1.753563628492404e-05, "loss": 0.9672, "step": 7557 }, { "epoch": 0.02, "learning_rate": 1.7534927668998522e-05, "loss": 0.9009, "step": 7558 }, { "epoch": 0.02, "learning_rate": 1.7534218965530107e-05, "loss": 0.9773, "step": 7559 }, { "epoch": 0.02, "learning_rate": 1.753351017452703e-05, "loss": 0.9432, "step": 7560 }, { "epoch": 0.02, "learning_rate": 1.753280129599753e-05, "loss": 0.9227, "step": 7561 }, { "epoch": 0.02, "learning_rate": 1.7532092329949832e-05, "loss": 0.8963, "step": 7562 }, { "epoch": 0.02, "learning_rate": 1.7531383276392183e-05, "loss": 0.9551, "step": 7563 }, { "epoch": 0.02, "learning_rate": 1.7530674135332816e-05, "loss": 0.9094, "step": 7564 }, { "epoch": 0.02, "learning_rate": 1.752996490677997e-05, "loss": 0.9554, "step": 7565 }, { "epoch": 0.02, "learning_rate": 1.752925559074189e-05, "loss": 0.9437, "step": 7566 }, { "epoch": 0.02, "learning_rate": 1.752854618722681e-05, "loss": 0.9592, "step": 7567 }, { "epoch": 0.02, "learning_rate": 1.752783669624298e-05, "loss": 0.9825, "step": 7568 }, { "epoch": 0.02, "learning_rate": 1.7527127117798635e-05, "loss": 0.9336, "step": 7569 }, { "epoch": 0.02, "learning_rate": 1.7526417451902024e-05, "loss": 0.9248, "step": 7570 }, { "epoch": 0.02, "learning_rate": 1.7525707698561383e-05, "loss": 0.93, "step": 7571 }, { "epoch": 0.02, "learning_rate": 1.7524997857784978e-05, "loss": 0.8836, "step": 7572 }, { "epoch": 0.02, "learning_rate": 1.7524287929581036e-05, "loss": 0.9686, "step": 7573 }, { "epoch": 0.02, "learning_rate": 1.7523577913957813e-05, "loss": 0.9388, "step": 7574 }, { "epoch": 0.02, "learning_rate": 1.7522867810923564e-05, "loss": 0.9114, "step": 7575 }, { "epoch": 0.02, "learning_rate": 1.7522157620486527e-05, "loss": 0.9261, "step": 7576 }, { "epoch": 0.02, "learning_rate": 1.7521447342654964e-05, "loss": 0.9325, "step": 7577 }, { "epoch": 0.02, "learning_rate": 1.752073697743712e-05, "loss": 0.9583, "step": 7578 }, { "epoch": 0.02, "learning_rate": 1.752002652484125e-05, "loss": 0.9438, "step": 7579 }, { "epoch": 0.02, "learning_rate": 1.7519315984875614e-05, "loss": 0.9451, "step": 7580 }, { "epoch": 0.02, "learning_rate": 1.751860535754846e-05, "loss": 0.9066, "step": 7581 }, { "epoch": 0.02, "learning_rate": 1.7517894642868046e-05, "loss": 0.9129, "step": 7582 }, { "epoch": 0.02, "learning_rate": 1.751718384084263e-05, "loss": 0.9482, "step": 7583 }, { "epoch": 0.02, "learning_rate": 1.7516472951480472e-05, "loss": 0.8877, "step": 7584 }, { "epoch": 0.02, "learning_rate": 1.7515761974789828e-05, "loss": 0.9357, "step": 7585 }, { "epoch": 0.02, "learning_rate": 1.751505091077896e-05, "loss": 0.9337, "step": 7586 }, { "epoch": 0.02, "learning_rate": 1.751433975945613e-05, "loss": 0.9202, "step": 7587 }, { "epoch": 0.02, "learning_rate": 1.75136285208296e-05, "loss": 0.9361, "step": 7588 }, { "epoch": 0.02, "learning_rate": 1.7512917194907634e-05, "loss": 0.9428, "step": 7589 }, { "epoch": 0.02, "learning_rate": 1.7512205781698493e-05, "loss": 0.9307, "step": 7590 }, { "epoch": 0.02, "learning_rate": 1.7511494281210443e-05, "loss": 0.9391, "step": 7591 }, { "epoch": 0.02, "learning_rate": 1.7510782693451758e-05, "loss": 0.9021, "step": 7592 }, { "epoch": 0.02, "learning_rate": 1.7510071018430696e-05, "loss": 0.9493, "step": 7593 }, { "epoch": 0.02, "learning_rate": 1.750935925615553e-05, "loss": 0.9274, "step": 7594 }, { "epoch": 0.02, "learning_rate": 1.7508647406634527e-05, "loss": 0.9657, "step": 7595 }, { "epoch": 0.02, "learning_rate": 1.750793546987596e-05, "loss": 0.9721, "step": 7596 }, { "epoch": 0.02, "learning_rate": 1.7507223445888098e-05, "loss": 0.9618, "step": 7597 }, { "epoch": 0.02, "learning_rate": 1.7506511334679217e-05, "loss": 0.9216, "step": 7598 }, { "epoch": 0.02, "learning_rate": 1.7505799136257587e-05, "loss": 0.9207, "step": 7599 }, { "epoch": 0.02, "learning_rate": 1.7505086850631487e-05, "loss": 0.9301, "step": 7600 }, { "epoch": 0.02, "learning_rate": 1.750437447780919e-05, "loss": 0.9449, "step": 7601 }, { "epoch": 0.02, "learning_rate": 1.7503662017798967e-05, "loss": 0.9206, "step": 7602 }, { "epoch": 0.02, "learning_rate": 1.7502949470609102e-05, "loss": 0.9393, "step": 7603 }, { "epoch": 0.02, "learning_rate": 1.750223683624788e-05, "loss": 0.9559, "step": 7604 }, { "epoch": 0.02, "learning_rate": 1.7501524114723563e-05, "loss": 0.9089, "step": 7605 }, { "epoch": 0.02, "learning_rate": 1.7500811306044442e-05, "loss": 0.9013, "step": 7606 }, { "epoch": 0.02, "learning_rate": 1.75000984102188e-05, "loss": 0.9273, "step": 7607 }, { "epoch": 0.02, "learning_rate": 1.749938542725492e-05, "loss": 0.9287, "step": 7608 }, { "epoch": 0.02, "learning_rate": 1.749867235716108e-05, "loss": 0.9471, "step": 7609 }, { "epoch": 0.02, "learning_rate": 1.7497959199945575e-05, "loss": 0.934, "step": 7610 }, { "epoch": 0.02, "learning_rate": 1.7497245955616675e-05, "loss": 0.9179, "step": 7611 }, { "epoch": 0.02, "learning_rate": 1.7496532624182677e-05, "loss": 0.9521, "step": 7612 }, { "epoch": 0.02, "learning_rate": 1.749581920565187e-05, "loss": 0.9146, "step": 7613 }, { "epoch": 0.02, "learning_rate": 1.749510570003254e-05, "loss": 0.8917, "step": 7614 }, { "epoch": 0.02, "learning_rate": 1.749439210733297e-05, "loss": 0.9155, "step": 7615 }, { "epoch": 0.02, "learning_rate": 1.7493678427561464e-05, "loss": 0.9468, "step": 7616 }, { "epoch": 0.02, "learning_rate": 1.7492964660726304e-05, "loss": 0.9567, "step": 7617 }, { "epoch": 0.02, "learning_rate": 1.7492250806835788e-05, "loss": 0.9355, "step": 7618 }, { "epoch": 0.02, "learning_rate": 1.7491536865898203e-05, "loss": 0.9783, "step": 7619 }, { "epoch": 0.02, "learning_rate": 1.749082283792185e-05, "loss": 0.9807, "step": 7620 }, { "epoch": 0.02, "learning_rate": 1.7490108722915023e-05, "loss": 0.8983, "step": 7621 }, { "epoch": 0.02, "learning_rate": 1.7489394520886017e-05, "loss": 0.9815, "step": 7622 }, { "epoch": 0.02, "learning_rate": 1.7488680231843132e-05, "loss": 0.9397, "step": 7623 }, { "epoch": 0.02, "learning_rate": 1.748796585579467e-05, "loss": 0.8981, "step": 7624 }, { "epoch": 0.02, "learning_rate": 1.7487251392748924e-05, "loss": 0.9269, "step": 7625 }, { "epoch": 0.02, "learning_rate": 1.74865368427142e-05, "loss": 0.9403, "step": 7626 }, { "epoch": 0.02, "learning_rate": 1.748582220569879e-05, "loss": 0.9201, "step": 7627 }, { "epoch": 0.02, "learning_rate": 1.7485107481711014e-05, "loss": 0.9017, "step": 7628 }, { "epoch": 0.02, "learning_rate": 1.748439267075916e-05, "loss": 0.9287, "step": 7629 }, { "epoch": 0.02, "learning_rate": 1.7483677772851544e-05, "loss": 0.9546, "step": 7630 }, { "epoch": 0.02, "learning_rate": 1.7482962787996463e-05, "loss": 0.9615, "step": 7631 }, { "epoch": 0.02, "learning_rate": 1.7482247716202232e-05, "loss": 0.9604, "step": 7632 }, { "epoch": 0.02, "learning_rate": 1.7481532557477153e-05, "loss": 0.9259, "step": 7633 }, { "epoch": 0.02, "learning_rate": 1.748081731182954e-05, "loss": 0.9122, "step": 7634 }, { "epoch": 0.02, "learning_rate": 1.7480101979267694e-05, "loss": 0.9319, "step": 7635 }, { "epoch": 0.02, "learning_rate": 1.7479386559799937e-05, "loss": 0.9279, "step": 7636 }, { "epoch": 0.02, "learning_rate": 1.7478671053434572e-05, "loss": 0.9154, "step": 7637 }, { "epoch": 0.02, "learning_rate": 1.7477955460179917e-05, "loss": 0.9621, "step": 7638 }, { "epoch": 0.02, "learning_rate": 1.7477239780044285e-05, "loss": 0.9688, "step": 7639 }, { "epoch": 0.02, "learning_rate": 1.747652401303599e-05, "loss": 0.8878, "step": 7640 }, { "epoch": 0.02, "learning_rate": 1.7475808159163347e-05, "loss": 0.974, "step": 7641 }, { "epoch": 0.02, "learning_rate": 1.747509221843468e-05, "loss": 0.9591, "step": 7642 }, { "epoch": 0.02, "learning_rate": 1.7474376190858296e-05, "loss": 0.9536, "step": 7643 }, { "epoch": 0.02, "learning_rate": 1.7473660076442524e-05, "loss": 0.9575, "step": 7644 }, { "epoch": 0.02, "learning_rate": 1.7472943875195676e-05, "loss": 0.9537, "step": 7645 }, { "epoch": 0.02, "learning_rate": 1.7472227587126077e-05, "loss": 0.8859, "step": 7646 }, { "epoch": 0.02, "learning_rate": 1.747151121224205e-05, "loss": 0.9496, "step": 7647 }, { "epoch": 0.02, "learning_rate": 1.7470794750551916e-05, "loss": 0.9232, "step": 7648 }, { "epoch": 0.02, "learning_rate": 1.7470078202064e-05, "loss": 0.9218, "step": 7649 }, { "epoch": 0.02, "learning_rate": 1.7469361566786624e-05, "loss": 0.8959, "step": 7650 }, { "epoch": 0.02, "learning_rate": 1.746864484472812e-05, "loss": 0.962, "step": 7651 }, { "epoch": 0.02, "learning_rate": 1.746792803589681e-05, "loss": 0.9334, "step": 7652 }, { "epoch": 0.02, "learning_rate": 1.7467211140301027e-05, "loss": 0.9479, "step": 7653 }, { "epoch": 0.02, "learning_rate": 1.74664941579491e-05, "loss": 0.9225, "step": 7654 }, { "epoch": 0.02, "learning_rate": 1.746577708884935e-05, "loss": 0.9558, "step": 7655 }, { "epoch": 0.02, "learning_rate": 1.7465059933010116e-05, "loss": 0.9363, "step": 7656 }, { "epoch": 0.02, "learning_rate": 1.746434269043973e-05, "loss": 0.8989, "step": 7657 }, { "epoch": 0.02, "learning_rate": 1.746362536114652e-05, "loss": 0.9558, "step": 7658 }, { "epoch": 0.02, "learning_rate": 1.746290794513883e-05, "loss": 1.0171, "step": 7659 }, { "epoch": 0.02, "learning_rate": 1.7462190442424983e-05, "loss": 0.9609, "step": 7660 }, { "epoch": 0.02, "learning_rate": 1.746147285301332e-05, "loss": 0.9412, "step": 7661 }, { "epoch": 0.02, "learning_rate": 1.7460755176912183e-05, "loss": 0.9488, "step": 7662 }, { "epoch": 0.02, "learning_rate": 1.7460037414129905e-05, "loss": 0.9304, "step": 7663 }, { "epoch": 0.02, "learning_rate": 1.745931956467483e-05, "loss": 0.9495, "step": 7664 }, { "epoch": 0.02, "learning_rate": 1.7458601628555288e-05, "loss": 0.9297, "step": 7665 }, { "epoch": 0.02, "learning_rate": 1.7457883605779632e-05, "loss": 0.8936, "step": 7666 }, { "epoch": 0.02, "learning_rate": 1.7457165496356195e-05, "loss": 0.976, "step": 7667 }, { "epoch": 0.02, "learning_rate": 1.745644730029333e-05, "loss": 0.924, "step": 7668 }, { "epoch": 0.02, "learning_rate": 1.745572901759937e-05, "loss": 0.9602, "step": 7669 }, { "epoch": 0.02, "learning_rate": 1.745501064828267e-05, "loss": 0.9221, "step": 7670 }, { "epoch": 0.02, "learning_rate": 1.7454292192351568e-05, "loss": 0.9269, "step": 7671 }, { "epoch": 0.02, "learning_rate": 1.7453573649814415e-05, "loss": 0.9404, "step": 7672 }, { "epoch": 0.02, "learning_rate": 1.745285502067956e-05, "loss": 0.914, "step": 7673 }, { "epoch": 0.02, "learning_rate": 1.7452136304955355e-05, "loss": 0.9576, "step": 7674 }, { "epoch": 0.02, "learning_rate": 1.7451417502650144e-05, "loss": 0.9443, "step": 7675 }, { "epoch": 0.02, "learning_rate": 1.7450698613772284e-05, "loss": 0.9342, "step": 7676 }, { "epoch": 0.02, "learning_rate": 1.7449979638330117e-05, "loss": 0.9154, "step": 7677 }, { "epoch": 0.02, "learning_rate": 1.744926057633201e-05, "loss": 0.9054, "step": 7678 }, { "epoch": 0.02, "learning_rate": 1.744854142778631e-05, "loss": 0.9821, "step": 7679 }, { "epoch": 0.02, "learning_rate": 1.744782219270137e-05, "loss": 0.9413, "step": 7680 }, { "epoch": 0.02, "learning_rate": 1.7447102871085552e-05, "loss": 0.9607, "step": 7681 }, { "epoch": 0.02, "learning_rate": 1.7446383462947212e-05, "loss": 0.911, "step": 7682 }, { "epoch": 0.02, "learning_rate": 1.7445663968294703e-05, "loss": 1.0076, "step": 7683 }, { "epoch": 0.02, "learning_rate": 1.7444944387136388e-05, "loss": 0.9273, "step": 7684 }, { "epoch": 0.02, "learning_rate": 1.744422471948063e-05, "loss": 0.9476, "step": 7685 }, { "epoch": 0.02, "learning_rate": 1.744350496533579e-05, "loss": 0.9186, "step": 7686 }, { "epoch": 0.02, "learning_rate": 1.7442785124710227e-05, "loss": 0.9224, "step": 7687 }, { "epoch": 0.02, "learning_rate": 1.74420651976123e-05, "loss": 0.954, "step": 7688 }, { "epoch": 0.02, "learning_rate": 1.7441345184050385e-05, "loss": 0.9731, "step": 7689 }, { "epoch": 0.02, "learning_rate": 1.7440625084032836e-05, "loss": 0.9096, "step": 7690 }, { "epoch": 0.02, "learning_rate": 1.743990489756803e-05, "loss": 0.9627, "step": 7691 }, { "epoch": 0.02, "learning_rate": 1.7439184624664325e-05, "loss": 0.9145, "step": 7692 }, { "epoch": 0.02, "learning_rate": 1.7438464265330097e-05, "loss": 0.9756, "step": 7693 }, { "epoch": 0.02, "learning_rate": 1.7437743819573706e-05, "loss": 0.93, "step": 7694 }, { "epoch": 0.02, "learning_rate": 1.7437023287403526e-05, "loss": 0.9647, "step": 7695 }, { "epoch": 0.02, "learning_rate": 1.7436302668827935e-05, "loss": 0.9641, "step": 7696 }, { "epoch": 0.02, "learning_rate": 1.74355819638553e-05, "loss": 0.9535, "step": 7697 }, { "epoch": 0.02, "learning_rate": 1.7434861172493995e-05, "loss": 0.9142, "step": 7698 }, { "epoch": 0.02, "learning_rate": 1.7434140294752394e-05, "loss": 0.9714, "step": 7699 }, { "epoch": 0.02, "learning_rate": 1.743341933063887e-05, "loss": 0.9579, "step": 7700 }, { "epoch": 0.02, "learning_rate": 1.7432698280161803e-05, "loss": 0.9409, "step": 7701 }, { "epoch": 0.02, "learning_rate": 1.7431977143329568e-05, "loss": 0.9265, "step": 7702 }, { "epoch": 0.02, "learning_rate": 1.7431255920150546e-05, "loss": 0.9238, "step": 7703 }, { "epoch": 0.02, "learning_rate": 1.7430534610633118e-05, "loss": 0.9645, "step": 7704 }, { "epoch": 0.02, "learning_rate": 1.7429813214785657e-05, "loss": 0.9117, "step": 7705 }, { "epoch": 0.02, "learning_rate": 1.742909173261655e-05, "loss": 0.9297, "step": 7706 }, { "epoch": 0.02, "learning_rate": 1.7428370164134177e-05, "loss": 0.9612, "step": 7707 }, { "epoch": 0.02, "learning_rate": 1.7427648509346926e-05, "loss": 0.916, "step": 7708 }, { "epoch": 0.02, "learning_rate": 1.742692676826317e-05, "loss": 0.9577, "step": 7709 }, { "epoch": 0.02, "learning_rate": 1.7426204940891313e-05, "loss": 0.9145, "step": 7710 }, { "epoch": 0.02, "learning_rate": 1.7425483027239724e-05, "loss": 0.9218, "step": 7711 }, { "epoch": 0.02, "learning_rate": 1.7424761027316795e-05, "loss": 0.9742, "step": 7712 }, { "epoch": 0.02, "learning_rate": 1.7424038941130923e-05, "loss": 0.9533, "step": 7713 }, { "epoch": 0.02, "learning_rate": 1.7423316768690487e-05, "loss": 0.952, "step": 7714 }, { "epoch": 0.02, "learning_rate": 1.742259451000388e-05, "loss": 0.9602, "step": 7715 }, { "epoch": 0.02, "learning_rate": 1.7421872165079498e-05, "loss": 0.9098, "step": 7716 }, { "epoch": 0.02, "learning_rate": 1.742114973392573e-05, "loss": 0.9338, "step": 7717 }, { "epoch": 0.02, "learning_rate": 1.7420427216550967e-05, "loss": 0.9218, "step": 7718 }, { "epoch": 0.02, "learning_rate": 1.7419704612963605e-05, "loss": 0.9602, "step": 7719 }, { "epoch": 0.02, "learning_rate": 1.7418981923172043e-05, "loss": 0.9142, "step": 7720 }, { "epoch": 0.02, "learning_rate": 1.7418259147184673e-05, "loss": 0.9638, "step": 7721 }, { "epoch": 0.02, "learning_rate": 1.7417536285009896e-05, "loss": 0.9501, "step": 7722 }, { "epoch": 0.02, "learning_rate": 1.7416813336656106e-05, "loss": 0.9289, "step": 7723 }, { "epoch": 0.02, "learning_rate": 1.7416090302131706e-05, "loss": 0.9335, "step": 7724 }, { "epoch": 0.02, "learning_rate": 1.7415367181445095e-05, "loss": 0.9301, "step": 7725 }, { "epoch": 0.02, "learning_rate": 1.7414643974604676e-05, "loss": 0.9292, "step": 7726 }, { "epoch": 0.02, "learning_rate": 1.741392068161885e-05, "loss": 0.9404, "step": 7727 }, { "epoch": 0.02, "learning_rate": 1.7413197302496018e-05, "loss": 0.9767, "step": 7728 }, { "epoch": 0.02, "learning_rate": 1.741247383724459e-05, "loss": 0.9638, "step": 7729 }, { "epoch": 0.02, "learning_rate": 1.7411750285872967e-05, "loss": 0.929, "step": 7730 }, { "epoch": 0.02, "learning_rate": 1.7411026648389556e-05, "loss": 0.9597, "step": 7731 }, { "epoch": 0.02, "learning_rate": 1.7410302924802766e-05, "loss": 0.9098, "step": 7732 }, { "epoch": 0.02, "learning_rate": 1.7409579115121006e-05, "loss": 0.9094, "step": 7733 }, { "epoch": 0.02, "learning_rate": 1.7408855219352682e-05, "loss": 0.9341, "step": 7734 }, { "epoch": 0.02, "learning_rate": 1.7408131237506207e-05, "loss": 1.011, "step": 7735 }, { "epoch": 0.02, "learning_rate": 1.7407407169589994e-05, "loss": 0.908, "step": 7736 }, { "epoch": 0.02, "learning_rate": 1.7406683015612454e-05, "loss": 0.9365, "step": 7737 }, { "epoch": 0.02, "learning_rate": 1.7405958775581995e-05, "loss": 0.9238, "step": 7738 }, { "epoch": 0.02, "learning_rate": 1.740523444950704e-05, "loss": 0.9264, "step": 7739 }, { "epoch": 0.02, "learning_rate": 1.7404510037396e-05, "loss": 0.9069, "step": 7740 }, { "epoch": 0.02, "learning_rate": 1.740378553925729e-05, "loss": 0.9515, "step": 7741 }, { "epoch": 0.02, "learning_rate": 1.7403060955099332e-05, "loss": 0.9316, "step": 7742 }, { "epoch": 0.02, "learning_rate": 1.740233628493054e-05, "loss": 0.9607, "step": 7743 }, { "epoch": 0.02, "learning_rate": 1.7401611528759338e-05, "loss": 0.9189, "step": 7744 }, { "epoch": 0.02, "learning_rate": 1.7400886686594143e-05, "loss": 0.9622, "step": 7745 }, { "epoch": 0.02, "learning_rate": 1.7400161758443377e-05, "loss": 0.9305, "step": 7746 }, { "epoch": 0.02, "learning_rate": 1.739943674431546e-05, "loss": 0.9345, "step": 7747 }, { "epoch": 0.02, "learning_rate": 1.739871164421882e-05, "loss": 0.9205, "step": 7748 }, { "epoch": 0.02, "learning_rate": 1.7397986458161883e-05, "loss": 0.9449, "step": 7749 }, { "epoch": 0.02, "learning_rate": 1.739726118615306e-05, "loss": 0.9605, "step": 7750 }, { "epoch": 0.02, "learning_rate": 1.73965358282008e-05, "loss": 0.9456, "step": 7751 }, { "epoch": 0.03, "learning_rate": 1.7395810384313513e-05, "loss": 0.9468, "step": 7752 }, { "epoch": 0.03, "learning_rate": 1.7395084854499636e-05, "loss": 0.9573, "step": 7753 }, { "epoch": 0.03, "learning_rate": 1.7394359238767595e-05, "loss": 0.9087, "step": 7754 }, { "epoch": 0.03, "learning_rate": 1.7393633537125817e-05, "loss": 0.9339, "step": 7755 }, { "epoch": 0.03, "learning_rate": 1.7392907749582742e-05, "loss": 0.9434, "step": 7756 }, { "epoch": 0.03, "learning_rate": 1.7392181876146795e-05, "loss": 0.9554, "step": 7757 }, { "epoch": 0.03, "learning_rate": 1.7391455916826414e-05, "loss": 0.9371, "step": 7758 }, { "epoch": 0.03, "learning_rate": 1.7390729871630027e-05, "loss": 0.9565, "step": 7759 }, { "epoch": 0.03, "learning_rate": 1.739000374056608e-05, "loss": 0.9143, "step": 7760 }, { "epoch": 0.03, "learning_rate": 1.7389277523643e-05, "loss": 0.9399, "step": 7761 }, { "epoch": 0.03, "learning_rate": 1.738855122086923e-05, "loss": 0.9698, "step": 7762 }, { "epoch": 0.03, "learning_rate": 1.73878248322532e-05, "loss": 0.9474, "step": 7763 }, { "epoch": 0.03, "learning_rate": 1.7387098357803362e-05, "loss": 0.9053, "step": 7764 }, { "epoch": 0.03, "learning_rate": 1.7386371797528148e-05, "loss": 0.9303, "step": 7765 }, { "epoch": 0.03, "learning_rate": 1.7385645151436e-05, "loss": 0.9242, "step": 7766 }, { "epoch": 0.03, "learning_rate": 1.738491841953536e-05, "loss": 0.9473, "step": 7767 }, { "epoch": 0.03, "learning_rate": 1.738419160183468e-05, "loss": 0.9331, "step": 7768 }, { "epoch": 0.03, "learning_rate": 1.738346469834239e-05, "loss": 0.9448, "step": 7769 }, { "epoch": 0.03, "learning_rate": 1.7382737709066947e-05, "loss": 0.9218, "step": 7770 }, { "epoch": 0.03, "learning_rate": 1.738201063401679e-05, "loss": 1.0258, "step": 7771 }, { "epoch": 0.03, "learning_rate": 1.738128347320037e-05, "loss": 0.8939, "step": 7772 }, { "epoch": 0.03, "learning_rate": 1.738055622662614e-05, "loss": 0.9482, "step": 7773 }, { "epoch": 0.03, "learning_rate": 1.7379828894302538e-05, "loss": 0.93, "step": 7774 }, { "epoch": 0.03, "learning_rate": 1.737910147623802e-05, "loss": 0.9725, "step": 7775 }, { "epoch": 0.03, "learning_rate": 1.737837397244104e-05, "loss": 0.8757, "step": 7776 }, { "epoch": 0.03, "learning_rate": 1.737764638292005e-05, "loss": 0.9449, "step": 7777 }, { "epoch": 0.03, "learning_rate": 1.73769187076835e-05, "loss": 0.9602, "step": 7778 }, { "epoch": 0.03, "learning_rate": 1.7376190946739842e-05, "loss": 0.9381, "step": 7779 }, { "epoch": 0.03, "learning_rate": 1.737546310009754e-05, "loss": 0.914, "step": 7780 }, { "epoch": 0.03, "learning_rate": 1.7374735167765044e-05, "loss": 0.9244, "step": 7781 }, { "epoch": 0.03, "learning_rate": 1.7374007149750816e-05, "loss": 0.9205, "step": 7782 }, { "epoch": 0.03, "learning_rate": 1.737327904606331e-05, "loss": 0.924, "step": 7783 }, { "epoch": 0.03, "learning_rate": 1.737255085671098e-05, "loss": 0.9289, "step": 7784 }, { "epoch": 0.03, "learning_rate": 1.73718225817023e-05, "loss": 0.9387, "step": 7785 }, { "epoch": 0.03, "learning_rate": 1.737109422104572e-05, "loss": 0.9494, "step": 7786 }, { "epoch": 0.03, "learning_rate": 1.737036577474971e-05, "loss": 0.9027, "step": 7787 }, { "epoch": 0.03, "learning_rate": 1.7369637242822728e-05, "loss": 0.9375, "step": 7788 }, { "epoch": 0.03, "learning_rate": 1.736890862527324e-05, "loss": 0.9522, "step": 7789 }, { "epoch": 0.03, "learning_rate": 1.7368179922109713e-05, "loss": 0.9836, "step": 7790 }, { "epoch": 0.03, "learning_rate": 1.7367451133340606e-05, "loss": 1.0005, "step": 7791 }, { "epoch": 0.03, "learning_rate": 1.7366722258974396e-05, "loss": 0.9363, "step": 7792 }, { "epoch": 0.03, "learning_rate": 1.7365993299019548e-05, "loss": 0.9602, "step": 7793 }, { "epoch": 0.03, "learning_rate": 1.7365264253484526e-05, "loss": 0.914, "step": 7794 }, { "epoch": 0.03, "learning_rate": 1.7364535122377808e-05, "loss": 0.9474, "step": 7795 }, { "epoch": 0.03, "learning_rate": 1.7363805905707863e-05, "loss": 0.9164, "step": 7796 }, { "epoch": 0.03, "learning_rate": 1.736307660348316e-05, "loss": 0.98, "step": 7797 }, { "epoch": 0.03, "learning_rate": 1.7362347215712173e-05, "loss": 0.9212, "step": 7798 }, { "epoch": 0.03, "learning_rate": 1.736161774240338e-05, "loss": 0.955, "step": 7799 }, { "epoch": 0.03, "learning_rate": 1.736088818356525e-05, "loss": 0.9507, "step": 7800 }, { "epoch": 0.03, "learning_rate": 1.7360158539206265e-05, "loss": 0.9351, "step": 7801 }, { "epoch": 0.03, "learning_rate": 1.73594288093349e-05, "loss": 0.9419, "step": 7802 }, { "epoch": 0.03, "learning_rate": 1.7358698993959636e-05, "loss": 1.0117, "step": 7803 }, { "epoch": 0.03, "learning_rate": 1.7357969093088945e-05, "loss": 0.9402, "step": 7804 }, { "epoch": 0.03, "learning_rate": 1.735723910673132e-05, "loss": 0.9446, "step": 7805 }, { "epoch": 0.03, "learning_rate": 1.7356509034895227e-05, "loss": 0.9224, "step": 7806 }, { "epoch": 0.03, "learning_rate": 1.7355778877589157e-05, "loss": 0.9345, "step": 7807 }, { "epoch": 0.03, "learning_rate": 1.735504863482159e-05, "loss": 0.9532, "step": 7808 }, { "epoch": 0.03, "learning_rate": 1.735431830660101e-05, "loss": 0.9146, "step": 7809 }, { "epoch": 0.03, "learning_rate": 1.735358789293591e-05, "loss": 0.9523, "step": 7810 }, { "epoch": 0.03, "learning_rate": 1.7352857393834766e-05, "loss": 0.983, "step": 7811 }, { "epoch": 0.03, "learning_rate": 1.7352126809306066e-05, "loss": 0.9331, "step": 7812 }, { "epoch": 0.03, "learning_rate": 1.7351396139358306e-05, "loss": 0.9408, "step": 7813 }, { "epoch": 0.03, "learning_rate": 1.7350665383999968e-05, "loss": 0.9088, "step": 7814 }, { "epoch": 0.03, "learning_rate": 1.7349934543239544e-05, "loss": 0.9752, "step": 7815 }, { "epoch": 0.03, "learning_rate": 1.7349203617085523e-05, "loss": 0.9034, "step": 7816 }, { "epoch": 0.03, "learning_rate": 1.7348472605546404e-05, "loss": 0.9344, "step": 7817 }, { "epoch": 0.03, "learning_rate": 1.7347741508630673e-05, "loss": 0.958, "step": 7818 }, { "epoch": 0.03, "learning_rate": 1.7347010326346828e-05, "loss": 0.9663, "step": 7819 }, { "epoch": 0.03, "learning_rate": 1.7346279058703364e-05, "loss": 0.9235, "step": 7820 }, { "epoch": 0.03, "learning_rate": 1.7345547705708773e-05, "loss": 0.9266, "step": 7821 }, { "epoch": 0.03, "learning_rate": 1.7344816267371558e-05, "loss": 0.9698, "step": 7822 }, { "epoch": 0.03, "learning_rate": 1.734408474370021e-05, "loss": 0.8951, "step": 7823 }, { "epoch": 0.03, "learning_rate": 1.7343353134703235e-05, "loss": 0.9284, "step": 7824 }, { "epoch": 0.03, "learning_rate": 1.734262144038913e-05, "loss": 0.9501, "step": 7825 }, { "epoch": 0.03, "learning_rate": 1.7341889660766395e-05, "loss": 0.9137, "step": 7826 }, { "epoch": 0.03, "learning_rate": 1.7341157795843533e-05, "loss": 0.9621, "step": 7827 }, { "epoch": 0.03, "learning_rate": 1.734042584562905e-05, "loss": 0.9251, "step": 7828 }, { "epoch": 0.03, "learning_rate": 1.7339693810131446e-05, "loss": 0.9489, "step": 7829 }, { "epoch": 0.03, "learning_rate": 1.7338961689359222e-05, "loss": 0.9027, "step": 7830 }, { "epoch": 0.03, "learning_rate": 1.7338229483320895e-05, "loss": 0.9242, "step": 7831 }, { "epoch": 0.03, "learning_rate": 1.7337497192024964e-05, "loss": 0.8984, "step": 7832 }, { "epoch": 0.03, "learning_rate": 1.7336764815479938e-05, "loss": 0.9196, "step": 7833 }, { "epoch": 0.03, "learning_rate": 1.733603235369433e-05, "loss": 0.9905, "step": 7834 }, { "epoch": 0.03, "learning_rate": 1.7335299806676646e-05, "loss": 1.001, "step": 7835 }, { "epoch": 0.03, "learning_rate": 1.7334567174435395e-05, "loss": 0.9875, "step": 7836 }, { "epoch": 0.03, "learning_rate": 1.7333834456979093e-05, "loss": 0.9565, "step": 7837 }, { "epoch": 0.03, "learning_rate": 1.7333101654316253e-05, "loss": 0.9309, "step": 7838 }, { "epoch": 0.03, "learning_rate": 1.7332368766455387e-05, "loss": 0.931, "step": 7839 }, { "epoch": 0.03, "learning_rate": 1.7331635793405012e-05, "loss": 0.9302, "step": 7840 }, { "epoch": 0.03, "learning_rate": 1.733090273517364e-05, "loss": 0.9476, "step": 7841 }, { "epoch": 0.03, "learning_rate": 1.733016959176979e-05, "loss": 0.934, "step": 7842 }, { "epoch": 0.03, "learning_rate": 1.7329436363201984e-05, "loss": 0.9293, "step": 7843 }, { "epoch": 0.03, "learning_rate": 1.7328703049478736e-05, "loss": 0.9266, "step": 7844 }, { "epoch": 0.03, "learning_rate": 1.7327969650608564e-05, "loss": 0.9049, "step": 7845 }, { "epoch": 0.03, "learning_rate": 1.732723616659999e-05, "loss": 0.9215, "step": 7846 }, { "epoch": 0.03, "learning_rate": 1.7326502597461547e-05, "loss": 0.9569, "step": 7847 }, { "epoch": 0.03, "learning_rate": 1.732576894320174e-05, "loss": 0.9425, "step": 7848 }, { "epoch": 0.03, "learning_rate": 1.73250352038291e-05, "loss": 0.9803, "step": 7849 }, { "epoch": 0.03, "learning_rate": 1.7324301379352158e-05, "loss": 0.9555, "step": 7850 }, { "epoch": 0.03, "learning_rate": 1.7323567469779436e-05, "loss": 0.9526, "step": 7851 }, { "epoch": 0.03, "learning_rate": 1.7322833475119455e-05, "loss": 0.9615, "step": 7852 }, { "epoch": 0.03, "learning_rate": 1.7322099395380748e-05, "loss": 0.9513, "step": 7853 }, { "epoch": 0.03, "learning_rate": 1.732136523057184e-05, "loss": 0.9492, "step": 7854 }, { "epoch": 0.03, "learning_rate": 1.732063098070127e-05, "loss": 0.9328, "step": 7855 }, { "epoch": 0.03, "learning_rate": 1.7319896645777555e-05, "loss": 0.9437, "step": 7856 }, { "epoch": 0.03, "learning_rate": 1.731916222580924e-05, "loss": 0.9234, "step": 7857 }, { "epoch": 0.03, "learning_rate": 1.7318427720804852e-05, "loss": 0.9222, "step": 7858 }, { "epoch": 0.03, "learning_rate": 1.7317693130772925e-05, "loss": 0.9424, "step": 7859 }, { "epoch": 0.03, "learning_rate": 1.731695845572199e-05, "loss": 0.9567, "step": 7860 }, { "epoch": 0.03, "learning_rate": 1.7316223695660585e-05, "loss": 0.9484, "step": 7861 }, { "epoch": 0.03, "learning_rate": 1.731548885059725e-05, "loss": 0.8992, "step": 7862 }, { "epoch": 0.03, "learning_rate": 1.731475392054052e-05, "loss": 0.9295, "step": 7863 }, { "epoch": 0.03, "learning_rate": 1.7314018905498932e-05, "loss": 0.9206, "step": 7864 }, { "epoch": 0.03, "learning_rate": 1.731328380548103e-05, "loss": 0.9177, "step": 7865 }, { "epoch": 0.03, "learning_rate": 1.731254862049535e-05, "loss": 0.8811, "step": 7866 }, { "epoch": 0.03, "learning_rate": 1.7311813350550438e-05, "loss": 0.896, "step": 7867 }, { "epoch": 0.03, "learning_rate": 1.731107799565483e-05, "loss": 0.9063, "step": 7868 }, { "epoch": 0.03, "learning_rate": 1.7310342555817078e-05, "loss": 0.9665, "step": 7869 }, { "epoch": 0.03, "learning_rate": 1.730960703104572e-05, "loss": 0.9155, "step": 7870 }, { "epoch": 0.03, "learning_rate": 1.7308871421349303e-05, "loss": 0.959, "step": 7871 }, { "epoch": 0.03, "learning_rate": 1.7308135726736376e-05, "loss": 0.9211, "step": 7872 }, { "epoch": 0.03, "learning_rate": 1.730739994721548e-05, "loss": 0.9551, "step": 7873 }, { "epoch": 0.03, "learning_rate": 1.7306664082795173e-05, "loss": 0.9502, "step": 7874 }, { "epoch": 0.03, "learning_rate": 1.7305928133484e-05, "loss": 0.97, "step": 7875 }, { "epoch": 0.03, "learning_rate": 1.730519209929051e-05, "loss": 0.9188, "step": 7876 }, { "epoch": 0.03, "learning_rate": 1.7304455980223254e-05, "loss": 0.9475, "step": 7877 }, { "epoch": 0.03, "learning_rate": 1.7303719776290792e-05, "loss": 0.9255, "step": 7878 }, { "epoch": 0.03, "learning_rate": 1.7302983487501665e-05, "loss": 0.9076, "step": 7879 }, { "epoch": 0.03, "learning_rate": 1.7302247113864434e-05, "loss": 0.9482, "step": 7880 }, { "epoch": 0.03, "learning_rate": 1.7301510655387658e-05, "loss": 0.9354, "step": 7881 }, { "epoch": 0.03, "learning_rate": 1.730077411207989e-05, "loss": 0.8495, "step": 7882 }, { "epoch": 0.03, "learning_rate": 1.7300037483949684e-05, "loss": 0.9495, "step": 7883 }, { "epoch": 0.03, "learning_rate": 1.7299300771005603e-05, "loss": 0.9498, "step": 7884 }, { "epoch": 0.03, "learning_rate": 1.7298563973256204e-05, "loss": 0.9465, "step": 7885 }, { "epoch": 0.03, "learning_rate": 1.729782709071005e-05, "loss": 0.9329, "step": 7886 }, { "epoch": 0.03, "learning_rate": 1.7297090123375695e-05, "loss": 0.9322, "step": 7887 }, { "epoch": 0.03, "learning_rate": 1.7296353071261713e-05, "loss": 0.9059, "step": 7888 }, { "epoch": 0.03, "learning_rate": 1.7295615934376656e-05, "loss": 0.9322, "step": 7889 }, { "epoch": 0.03, "learning_rate": 1.7294878712729096e-05, "loss": 0.9365, "step": 7890 }, { "epoch": 0.03, "learning_rate": 1.7294141406327597e-05, "loss": 0.959, "step": 7891 }, { "epoch": 0.03, "learning_rate": 1.729340401518072e-05, "loss": 0.9625, "step": 7892 }, { "epoch": 0.03, "learning_rate": 1.729266653929704e-05, "loss": 0.9499, "step": 7893 }, { "epoch": 0.03, "learning_rate": 1.7291928978685116e-05, "loss": 0.9573, "step": 7894 }, { "epoch": 0.03, "learning_rate": 1.7291191333353527e-05, "loss": 0.9495, "step": 7895 }, { "epoch": 0.03, "learning_rate": 1.7290453603310836e-05, "loss": 0.9813, "step": 7896 }, { "epoch": 0.03, "learning_rate": 1.7289715788565616e-05, "loss": 0.9764, "step": 7897 }, { "epoch": 0.03, "learning_rate": 1.7288977889126443e-05, "loss": 0.9242, "step": 7898 }, { "epoch": 0.03, "learning_rate": 1.7288239905001885e-05, "loss": 0.894, "step": 7899 }, { "epoch": 0.03, "learning_rate": 1.7287501836200517e-05, "loss": 0.9738, "step": 7900 }, { "epoch": 0.03, "learning_rate": 1.7286763682730914e-05, "loss": 0.9396, "step": 7901 }, { "epoch": 0.03, "learning_rate": 1.7286025444601657e-05, "loss": 0.9429, "step": 7902 }, { "epoch": 0.03, "learning_rate": 1.7285287121821315e-05, "loss": 0.9626, "step": 7903 }, { "epoch": 0.03, "learning_rate": 1.7284548714398472e-05, "loss": 0.9277, "step": 7904 }, { "epoch": 0.03, "learning_rate": 1.728381022234171e-05, "loss": 0.9093, "step": 7905 }, { "epoch": 0.03, "learning_rate": 1.7283071645659597e-05, "loss": 0.9419, "step": 7906 }, { "epoch": 0.03, "learning_rate": 1.7282332984360725e-05, "loss": 0.8948, "step": 7907 }, { "epoch": 0.03, "learning_rate": 1.728159423845367e-05, "loss": 0.9603, "step": 7908 }, { "epoch": 0.03, "learning_rate": 1.728085540794702e-05, "loss": 0.9474, "step": 7909 }, { "epoch": 0.03, "learning_rate": 1.7280116492849354e-05, "loss": 0.9363, "step": 7910 }, { "epoch": 0.03, "learning_rate": 1.727937749316926e-05, "loss": 0.9377, "step": 7911 }, { "epoch": 0.03, "learning_rate": 1.727863840891532e-05, "loss": 0.8801, "step": 7912 }, { "epoch": 0.03, "learning_rate": 1.727789924009613e-05, "loss": 0.9109, "step": 7913 }, { "epoch": 0.03, "learning_rate": 1.727715998672027e-05, "loss": 0.9201, "step": 7914 }, { "epoch": 0.03, "learning_rate": 1.727642064879633e-05, "loss": 0.9044, "step": 7915 }, { "epoch": 0.03, "learning_rate": 1.7275681226332898e-05, "loss": 0.9277, "step": 7916 }, { "epoch": 0.03, "learning_rate": 1.7274941719338568e-05, "loss": 0.9352, "step": 7917 }, { "epoch": 0.03, "learning_rate": 1.7274202127821934e-05, "loss": 0.8981, "step": 7918 }, { "epoch": 0.03, "learning_rate": 1.727346245179158e-05, "loss": 0.9647, "step": 7919 }, { "epoch": 0.03, "learning_rate": 1.727272269125611e-05, "loss": 0.9241, "step": 7920 }, { "epoch": 0.03, "learning_rate": 1.7271982846224113e-05, "loss": 0.9604, "step": 7921 }, { "epoch": 0.03, "learning_rate": 1.727124291670419e-05, "loss": 0.9222, "step": 7922 }, { "epoch": 0.03, "learning_rate": 1.7270502902704925e-05, "loss": 0.9279, "step": 7923 }, { "epoch": 0.03, "learning_rate": 1.7269762804234932e-05, "loss": 0.9196, "step": 7924 }, { "epoch": 0.03, "learning_rate": 1.7269022621302797e-05, "loss": 0.9066, "step": 7925 }, { "epoch": 0.03, "learning_rate": 1.7268282353917128e-05, "loss": 0.8905, "step": 7926 }, { "epoch": 0.03, "learning_rate": 1.7267542002086525e-05, "loss": 0.9564, "step": 7927 }, { "epoch": 0.03, "learning_rate": 1.7266801565819582e-05, "loss": 0.8975, "step": 7928 }, { "epoch": 0.03, "learning_rate": 1.726606104512491e-05, "loss": 0.9514, "step": 7929 }, { "epoch": 0.03, "learning_rate": 1.7265320440011106e-05, "loss": 0.9261, "step": 7930 }, { "epoch": 0.03, "learning_rate": 1.726457975048678e-05, "loss": 0.9263, "step": 7931 }, { "epoch": 0.03, "learning_rate": 1.7263838976560534e-05, "loss": 0.9552, "step": 7932 }, { "epoch": 0.03, "learning_rate": 1.7263098118240977e-05, "loss": 0.9456, "step": 7933 }, { "epoch": 0.03, "learning_rate": 1.726235717553671e-05, "loss": 0.9261, "step": 7934 }, { "epoch": 0.03, "learning_rate": 1.7261616148456356e-05, "loss": 0.9351, "step": 7935 }, { "epoch": 0.03, "learning_rate": 1.726087503700851e-05, "loss": 0.9269, "step": 7936 }, { "epoch": 0.03, "learning_rate": 1.726013384120179e-05, "loss": 0.9353, "step": 7937 }, { "epoch": 0.03, "learning_rate": 1.7259392561044803e-05, "loss": 0.9274, "step": 7938 }, { "epoch": 0.03, "learning_rate": 1.7258651196546163e-05, "loss": 0.9513, "step": 7939 }, { "epoch": 0.03, "learning_rate": 1.7257909747714486e-05, "loss": 0.9908, "step": 7940 }, { "epoch": 0.03, "learning_rate": 1.7257168214558383e-05, "loss": 0.9351, "step": 7941 }, { "epoch": 0.03, "learning_rate": 1.7256426597086468e-05, "loss": 0.9362, "step": 7942 }, { "epoch": 0.03, "learning_rate": 1.7255684895307365e-05, "loss": 0.9288, "step": 7943 }, { "epoch": 0.03, "learning_rate": 1.7254943109229683e-05, "loss": 0.928, "step": 7944 }, { "epoch": 0.03, "learning_rate": 1.7254201238862042e-05, "loss": 0.9604, "step": 7945 }, { "epoch": 0.03, "learning_rate": 1.7253459284213067e-05, "loss": 0.924, "step": 7946 }, { "epoch": 0.03, "learning_rate": 1.725271724529137e-05, "loss": 0.9367, "step": 7947 }, { "epoch": 0.03, "learning_rate": 1.7251975122105578e-05, "loss": 0.9057, "step": 7948 }, { "epoch": 0.03, "learning_rate": 1.7251232914664313e-05, "loss": 0.953, "step": 7949 }, { "epoch": 0.03, "learning_rate": 1.7250490622976195e-05, "loss": 0.9126, "step": 7950 }, { "epoch": 0.03, "learning_rate": 1.7249748247049847e-05, "loss": 0.9435, "step": 7951 }, { "epoch": 0.03, "learning_rate": 1.7249005786893903e-05, "loss": 0.9386, "step": 7952 }, { "epoch": 0.03, "learning_rate": 1.724826324251698e-05, "loss": 0.9355, "step": 7953 }, { "epoch": 0.03, "learning_rate": 1.7247520613927707e-05, "loss": 0.9229, "step": 7954 }, { "epoch": 0.03, "learning_rate": 1.724677790113471e-05, "loss": 0.9208, "step": 7955 }, { "epoch": 0.03, "learning_rate": 1.7246035104146627e-05, "loss": 0.9002, "step": 7956 }, { "epoch": 0.03, "learning_rate": 1.724529222297208e-05, "loss": 0.9465, "step": 7957 }, { "epoch": 0.03, "learning_rate": 1.7244549257619702e-05, "loss": 0.9802, "step": 7958 }, { "epoch": 0.03, "learning_rate": 1.7243806208098125e-05, "loss": 0.9867, "step": 7959 }, { "epoch": 0.03, "learning_rate": 1.724306307441598e-05, "loss": 0.8892, "step": 7960 }, { "epoch": 0.03, "learning_rate": 1.7242319856581907e-05, "loss": 0.9412, "step": 7961 }, { "epoch": 0.03, "learning_rate": 1.7241576554604535e-05, "loss": 0.9298, "step": 7962 }, { "epoch": 0.03, "learning_rate": 1.72408331684925e-05, "loss": 0.9681, "step": 7963 }, { "epoch": 0.03, "learning_rate": 1.7240089698254444e-05, "loss": 0.8933, "step": 7964 }, { "epoch": 0.03, "learning_rate": 1.7239346143899e-05, "loss": 0.9206, "step": 7965 }, { "epoch": 0.03, "learning_rate": 1.723860250543481e-05, "loss": 0.9488, "step": 7966 }, { "epoch": 0.03, "learning_rate": 1.723785878287051e-05, "loss": 0.9459, "step": 7967 }, { "epoch": 0.03, "learning_rate": 1.7237114976214745e-05, "loss": 0.9115, "step": 7968 }, { "epoch": 0.03, "learning_rate": 1.7236371085476152e-05, "loss": 0.9364, "step": 7969 }, { "epoch": 0.03, "learning_rate": 1.723562711066338e-05, "loss": 0.9158, "step": 7970 }, { "epoch": 0.03, "learning_rate": 1.7234883051785064e-05, "loss": 0.9528, "step": 7971 }, { "epoch": 0.03, "learning_rate": 1.7234138908849858e-05, "loss": 0.9057, "step": 7972 }, { "epoch": 0.03, "learning_rate": 1.7233394681866402e-05, "loss": 0.9405, "step": 7973 }, { "epoch": 0.03, "learning_rate": 1.7232650370843345e-05, "loss": 0.8993, "step": 7974 }, { "epoch": 0.03, "learning_rate": 1.7231905975789335e-05, "loss": 0.9463, "step": 7975 }, { "epoch": 0.03, "learning_rate": 1.7231161496713017e-05, "loss": 0.9516, "step": 7976 }, { "epoch": 0.03, "learning_rate": 1.7230416933623042e-05, "loss": 0.9812, "step": 7977 }, { "epoch": 0.03, "learning_rate": 1.7229672286528063e-05, "loss": 0.9358, "step": 7978 }, { "epoch": 0.03, "learning_rate": 1.722892755543673e-05, "loss": 0.9555, "step": 7979 }, { "epoch": 0.03, "learning_rate": 1.7228182740357695e-05, "loss": 0.9565, "step": 7980 }, { "epoch": 0.03, "learning_rate": 1.7227437841299613e-05, "loss": 0.9336, "step": 7981 }, { "epoch": 0.03, "learning_rate": 1.7226692858271133e-05, "loss": 0.8545, "step": 7982 }, { "epoch": 0.03, "learning_rate": 1.722594779128092e-05, "loss": 0.9265, "step": 7983 }, { "epoch": 0.03, "learning_rate": 1.7225202640337624e-05, "loss": 0.9067, "step": 7984 }, { "epoch": 0.03, "learning_rate": 1.7224457405449904e-05, "loss": 0.9634, "step": 7985 }, { "epoch": 0.03, "learning_rate": 1.7223712086626417e-05, "loss": 0.9352, "step": 7986 }, { "epoch": 0.03, "learning_rate": 1.7222966683875822e-05, "loss": 0.9507, "step": 7987 }, { "epoch": 0.03, "learning_rate": 1.722222119720678e-05, "loss": 0.9114, "step": 7988 }, { "epoch": 0.03, "learning_rate": 1.7221475626627957e-05, "loss": 0.9836, "step": 7989 }, { "epoch": 0.03, "learning_rate": 1.722072997214801e-05, "loss": 0.933, "step": 7990 }, { "epoch": 0.03, "learning_rate": 1.72199842337756e-05, "loss": 0.9187, "step": 7991 }, { "epoch": 0.03, "learning_rate": 1.7219238411519397e-05, "loss": 0.9296, "step": 7992 }, { "epoch": 0.03, "learning_rate": 1.721849250538806e-05, "loss": 0.9496, "step": 7993 }, { "epoch": 0.03, "learning_rate": 1.7217746515390263e-05, "loss": 0.9616, "step": 7994 }, { "epoch": 0.03, "learning_rate": 1.7217000441534665e-05, "loss": 0.9565, "step": 7995 }, { "epoch": 0.03, "learning_rate": 1.7216254283829943e-05, "loss": 0.942, "step": 7996 }, { "epoch": 0.03, "learning_rate": 1.721550804228476e-05, "loss": 0.9337, "step": 7997 }, { "epoch": 0.03, "learning_rate": 1.7214761716907785e-05, "loss": 0.96, "step": 7998 }, { "epoch": 0.03, "learning_rate": 1.721401530770769e-05, "loss": 0.9532, "step": 7999 }, { "epoch": 0.03, "learning_rate": 1.721326881469315e-05, "loss": 0.9545, "step": 8000 }, { "epoch": 0.03, "learning_rate": 1.7212522237872837e-05, "loss": 0.9671, "step": 8001 }, { "epoch": 0.03, "learning_rate": 1.7211775577255422e-05, "loss": 0.9667, "step": 8002 }, { "epoch": 0.03, "learning_rate": 1.7211028832849582e-05, "loss": 0.952, "step": 8003 }, { "epoch": 0.03, "learning_rate": 1.7210282004663995e-05, "loss": 0.9387, "step": 8004 }, { "epoch": 0.03, "learning_rate": 1.7209535092707333e-05, "loss": 0.9298, "step": 8005 }, { "epoch": 0.03, "learning_rate": 1.7208788096988282e-05, "loss": 0.9658, "step": 8006 }, { "epoch": 0.03, "learning_rate": 1.7208041017515507e-05, "loss": 0.9255, "step": 8007 }, { "epoch": 0.03, "learning_rate": 1.72072938542977e-05, "loss": 0.9306, "step": 8008 }, { "epoch": 0.03, "learning_rate": 1.7206546607343536e-05, "loss": 0.9459, "step": 8009 }, { "epoch": 0.03, "learning_rate": 1.7205799276661704e-05, "loss": 0.9404, "step": 8010 }, { "epoch": 0.03, "learning_rate": 1.7205051862260877e-05, "loss": 0.9414, "step": 8011 }, { "epoch": 0.03, "learning_rate": 1.7204304364149745e-05, "loss": 0.893, "step": 8012 }, { "epoch": 0.03, "learning_rate": 1.7203556782336986e-05, "loss": 0.9596, "step": 8013 }, { "epoch": 0.03, "learning_rate": 1.7202809116831297e-05, "loss": 0.9218, "step": 8014 }, { "epoch": 0.03, "learning_rate": 1.7202061367641355e-05, "loss": 0.9393, "step": 8015 }, { "epoch": 0.03, "learning_rate": 1.7201313534775847e-05, "loss": 0.9125, "step": 8016 }, { "epoch": 0.03, "learning_rate": 1.7200565618243467e-05, "loss": 0.948, "step": 8017 }, { "epoch": 0.03, "learning_rate": 1.71998176180529e-05, "loss": 0.935, "step": 8018 }, { "epoch": 0.03, "learning_rate": 1.7199069534212846e-05, "loss": 0.9481, "step": 8019 }, { "epoch": 0.03, "learning_rate": 1.7198321366731986e-05, "loss": 0.9223, "step": 8020 }, { "epoch": 0.03, "learning_rate": 1.7197573115619014e-05, "loss": 0.9233, "step": 8021 }, { "epoch": 0.03, "learning_rate": 1.7196824780882625e-05, "loss": 0.9152, "step": 8022 }, { "epoch": 0.03, "learning_rate": 1.7196076362531515e-05, "loss": 0.951, "step": 8023 }, { "epoch": 0.03, "learning_rate": 1.7195327860574377e-05, "loss": 0.9116, "step": 8024 }, { "epoch": 0.03, "learning_rate": 1.7194579275019908e-05, "loss": 0.9198, "step": 8025 }, { "epoch": 0.03, "learning_rate": 1.7193830605876806e-05, "loss": 0.94, "step": 8026 }, { "epoch": 0.03, "learning_rate": 1.7193081853153768e-05, "loss": 0.9513, "step": 8027 }, { "epoch": 0.03, "learning_rate": 1.719233301685949e-05, "loss": 0.95, "step": 8028 }, { "epoch": 0.03, "learning_rate": 1.7191584097002683e-05, "loss": 0.9724, "step": 8029 }, { "epoch": 0.03, "learning_rate": 1.719083509359204e-05, "loss": 0.9195, "step": 8030 }, { "epoch": 0.03, "learning_rate": 1.719008600663626e-05, "loss": 0.9803, "step": 8031 }, { "epoch": 0.03, "learning_rate": 1.7189336836144054e-05, "loss": 0.9427, "step": 8032 }, { "epoch": 0.03, "learning_rate": 1.7188587582124123e-05, "loss": 0.9269, "step": 8033 }, { "epoch": 0.03, "learning_rate": 1.7187838244585167e-05, "loss": 0.9233, "step": 8034 }, { "epoch": 0.03, "learning_rate": 1.7187088823535898e-05, "loss": 0.9545, "step": 8035 }, { "epoch": 0.03, "learning_rate": 1.7186339318985027e-05, "loss": 0.9706, "step": 8036 }, { "epoch": 0.03, "learning_rate": 1.718558973094125e-05, "loss": 0.9424, "step": 8037 }, { "epoch": 0.03, "learning_rate": 1.7184840059413284e-05, "loss": 0.9576, "step": 8038 }, { "epoch": 0.03, "learning_rate": 1.718409030440984e-05, "loss": 0.9418, "step": 8039 }, { "epoch": 0.03, "learning_rate": 1.7183340465939622e-05, "loss": 0.9502, "step": 8040 }, { "epoch": 0.03, "learning_rate": 1.718259054401135e-05, "loss": 0.9785, "step": 8041 }, { "epoch": 0.03, "learning_rate": 1.7181840538633728e-05, "loss": 0.9348, "step": 8042 }, { "epoch": 0.03, "learning_rate": 1.718109044981548e-05, "loss": 0.9402, "step": 8043 }, { "epoch": 0.03, "learning_rate": 1.7180340277565312e-05, "loss": 0.9472, "step": 8044 }, { "epoch": 0.03, "learning_rate": 1.7179590021891946e-05, "loss": 0.9632, "step": 8045 }, { "epoch": 0.03, "learning_rate": 1.7178839682804093e-05, "loss": 0.9623, "step": 8046 }, { "epoch": 0.03, "learning_rate": 1.7178089260310474e-05, "loss": 0.9394, "step": 8047 }, { "epoch": 0.03, "learning_rate": 1.7177338754419808e-05, "loss": 0.9158, "step": 8048 }, { "epoch": 0.03, "learning_rate": 1.717658816514081e-05, "loss": 0.9521, "step": 8049 }, { "epoch": 0.03, "learning_rate": 1.7175837492482208e-05, "loss": 0.9301, "step": 8050 }, { "epoch": 0.03, "learning_rate": 1.7175086736452717e-05, "loss": 0.9896, "step": 8051 }, { "epoch": 0.04, "learning_rate": 1.7174335897061066e-05, "loss": 0.9326, "step": 8052 }, { "epoch": 0.04, "learning_rate": 1.717358497431597e-05, "loss": 0.9196, "step": 8053 }, { "epoch": 0.04, "learning_rate": 1.717283396822616e-05, "loss": 0.9618, "step": 8054 }, { "epoch": 0.04, "learning_rate": 1.717208287880036e-05, "loss": 0.9517, "step": 8055 }, { "epoch": 0.04, "learning_rate": 1.7171331706047296e-05, "loss": 0.9653, "step": 8056 }, { "epoch": 0.04, "learning_rate": 1.7170580449975692e-05, "loss": 0.9729, "step": 8057 }, { "epoch": 0.04, "learning_rate": 1.7169829110594282e-05, "loss": 0.9278, "step": 8058 }, { "epoch": 0.04, "learning_rate": 1.7169077687911793e-05, "loss": 0.9336, "step": 8059 }, { "epoch": 0.04, "learning_rate": 1.7168326181936955e-05, "loss": 0.928, "step": 8060 }, { "epoch": 0.04, "learning_rate": 1.71675745926785e-05, "loss": 0.9632, "step": 8061 }, { "epoch": 0.04, "learning_rate": 1.716682292014516e-05, "loss": 0.9585, "step": 8062 }, { "epoch": 0.04, "learning_rate": 1.7166071164345665e-05, "loss": 0.9767, "step": 8063 }, { "epoch": 0.04, "learning_rate": 1.7165319325288748e-05, "loss": 0.9313, "step": 8064 }, { "epoch": 0.04, "learning_rate": 1.7164567402983153e-05, "loss": 0.9468, "step": 8065 }, { "epoch": 0.04, "learning_rate": 1.716381539743761e-05, "loss": 0.9116, "step": 8066 }, { "epoch": 0.04, "learning_rate": 1.7163063308660856e-05, "loss": 0.906, "step": 8067 }, { "epoch": 0.04, "learning_rate": 1.716231113666163e-05, "loss": 0.9126, "step": 8068 }, { "epoch": 0.04, "learning_rate": 1.716155888144867e-05, "loss": 0.9221, "step": 8069 }, { "epoch": 0.04, "learning_rate": 1.7160806543030718e-05, "loss": 0.9367, "step": 8070 }, { "epoch": 0.04, "learning_rate": 1.716005412141651e-05, "loss": 0.9649, "step": 8071 }, { "epoch": 0.04, "learning_rate": 1.7159301616614794e-05, "loss": 0.8979, "step": 8072 }, { "epoch": 0.04, "learning_rate": 1.715854902863431e-05, "loss": 0.9538, "step": 8073 }, { "epoch": 0.04, "learning_rate": 1.7157796357483804e-05, "loss": 0.9204, "step": 8074 }, { "epoch": 0.04, "learning_rate": 1.7157043603172012e-05, "loss": 0.9188, "step": 8075 }, { "epoch": 0.04, "learning_rate": 1.7156290765707694e-05, "loss": 0.9275, "step": 8076 }, { "epoch": 0.04, "learning_rate": 1.7155537845099584e-05, "loss": 0.9133, "step": 8077 }, { "epoch": 0.04, "learning_rate": 1.7154784841356434e-05, "loss": 0.9374, "step": 8078 }, { "epoch": 0.04, "learning_rate": 1.7154031754486995e-05, "loss": 0.9206, "step": 8079 }, { "epoch": 0.04, "learning_rate": 1.7153278584500015e-05, "loss": 0.9309, "step": 8080 }, { "epoch": 0.04, "learning_rate": 1.7152525331404243e-05, "loss": 0.9466, "step": 8081 }, { "epoch": 0.04, "learning_rate": 1.7151771995208434e-05, "loss": 0.9556, "step": 8082 }, { "epoch": 0.04, "learning_rate": 1.7151018575921338e-05, "loss": 0.9386, "step": 8083 }, { "epoch": 0.04, "learning_rate": 1.7150265073551708e-05, "loss": 0.9325, "step": 8084 }, { "epoch": 0.04, "learning_rate": 1.71495114881083e-05, "loss": 0.9247, "step": 8085 }, { "epoch": 0.04, "learning_rate": 1.7148757819599867e-05, "loss": 0.9195, "step": 8086 }, { "epoch": 0.04, "learning_rate": 1.7148004068035167e-05, "loss": 0.9566, "step": 8087 }, { "epoch": 0.04, "learning_rate": 1.714725023342296e-05, "loss": 0.924, "step": 8088 }, { "epoch": 0.04, "learning_rate": 1.7146496315771997e-05, "loss": 0.9609, "step": 8089 }, { "epoch": 0.04, "learning_rate": 1.7145742315091044e-05, "loss": 0.9276, "step": 8090 }, { "epoch": 0.04, "learning_rate": 1.714498823138886e-05, "loss": 0.921, "step": 8091 }, { "epoch": 0.04, "learning_rate": 1.7144234064674202e-05, "loss": 0.9317, "step": 8092 }, { "epoch": 0.04, "learning_rate": 1.714347981495584e-05, "loss": 0.9427, "step": 8093 }, { "epoch": 0.04, "learning_rate": 1.7142725482242528e-05, "loss": 0.9657, "step": 8094 }, { "epoch": 0.04, "learning_rate": 1.7141971066543034e-05, "loss": 0.9599, "step": 8095 }, { "epoch": 0.04, "learning_rate": 1.7141216567866122e-05, "loss": 0.9412, "step": 8096 }, { "epoch": 0.04, "learning_rate": 1.7140461986220565e-05, "loss": 0.9738, "step": 8097 }, { "epoch": 0.04, "learning_rate": 1.713970732161512e-05, "loss": 0.9724, "step": 8098 }, { "epoch": 0.04, "learning_rate": 1.713895257405856e-05, "loss": 0.9181, "step": 8099 }, { "epoch": 0.04, "learning_rate": 1.7138197743559656e-05, "loss": 0.9092, "step": 8100 }, { "epoch": 0.04, "learning_rate": 1.713744283012717e-05, "loss": 0.9353, "step": 8101 }, { "epoch": 0.04, "learning_rate": 1.713668783376988e-05, "loss": 0.9388, "step": 8102 }, { "epoch": 0.04, "learning_rate": 1.7135932754496553e-05, "loss": 0.9665, "step": 8103 }, { "epoch": 0.04, "learning_rate": 1.7135177592315966e-05, "loss": 0.9344, "step": 8104 }, { "epoch": 0.04, "learning_rate": 1.713442234723689e-05, "loss": 0.9386, "step": 8105 }, { "epoch": 0.04, "learning_rate": 1.71336670192681e-05, "loss": 0.9386, "step": 8106 }, { "epoch": 0.04, "learning_rate": 1.7132911608418372e-05, "loss": 0.9345, "step": 8107 }, { "epoch": 0.04, "learning_rate": 1.7132156114696485e-05, "loss": 0.9616, "step": 8108 }, { "epoch": 0.04, "learning_rate": 1.7131400538111212e-05, "loss": 0.946, "step": 8109 }, { "epoch": 0.04, "learning_rate": 1.713064487867133e-05, "loss": 0.9234, "step": 8110 }, { "epoch": 0.04, "learning_rate": 1.712988913638563e-05, "loss": 0.9406, "step": 8111 }, { "epoch": 0.04, "learning_rate": 1.7129133311262882e-05, "loss": 0.9399, "step": 8112 }, { "epoch": 0.04, "learning_rate": 1.7128377403311865e-05, "loss": 0.974, "step": 8113 }, { "epoch": 0.04, "learning_rate": 1.7127621412541372e-05, "loss": 0.9119, "step": 8114 }, { "epoch": 0.04, "learning_rate": 1.7126865338960178e-05, "loss": 0.9133, "step": 8115 }, { "epoch": 0.04, "learning_rate": 1.7126109182577067e-05, "loss": 0.8869, "step": 8116 }, { "epoch": 0.04, "learning_rate": 1.7125352943400833e-05, "loss": 0.9866, "step": 8117 }, { "epoch": 0.04, "learning_rate": 1.712459662144025e-05, "loss": 0.9459, "step": 8118 }, { "epoch": 0.04, "learning_rate": 1.7123840216704116e-05, "loss": 0.9131, "step": 8119 }, { "epoch": 0.04, "learning_rate": 1.7123083729201214e-05, "loss": 0.927, "step": 8120 }, { "epoch": 0.04, "learning_rate": 1.7122327158940333e-05, "loss": 0.9212, "step": 8121 }, { "epoch": 0.04, "learning_rate": 1.7121570505930262e-05, "loss": 0.9348, "step": 8122 }, { "epoch": 0.04, "learning_rate": 1.7120813770179797e-05, "loss": 0.9176, "step": 8123 }, { "epoch": 0.04, "learning_rate": 1.7120056951697724e-05, "loss": 0.9464, "step": 8124 }, { "epoch": 0.04, "learning_rate": 1.711930005049284e-05, "loss": 0.9218, "step": 8125 }, { "epoch": 0.04, "learning_rate": 1.7118543066573937e-05, "loss": 0.8727, "step": 8126 }, { "epoch": 0.04, "learning_rate": 1.7117785999949807e-05, "loss": 0.956, "step": 8127 }, { "epoch": 0.04, "learning_rate": 1.7117028850629255e-05, "loss": 0.9445, "step": 8128 }, { "epoch": 0.04, "learning_rate": 1.711627161862107e-05, "loss": 0.9882, "step": 8129 }, { "epoch": 0.04, "learning_rate": 1.711551430393405e-05, "loss": 0.97, "step": 8130 }, { "epoch": 0.04, "learning_rate": 1.7114756906576996e-05, "loss": 0.9692, "step": 8131 }, { "epoch": 0.04, "learning_rate": 1.7113999426558704e-05, "loss": 0.9616, "step": 8132 }, { "epoch": 0.04, "learning_rate": 1.711324186388798e-05, "loss": 0.9504, "step": 8133 }, { "epoch": 0.04, "learning_rate": 1.7112484218573625e-05, "loss": 0.9412, "step": 8134 }, { "epoch": 0.04, "learning_rate": 1.711172649062444e-05, "loss": 0.9307, "step": 8135 }, { "epoch": 0.04, "learning_rate": 1.7110968680049226e-05, "loss": 0.9177, "step": 8136 }, { "epoch": 0.04, "learning_rate": 1.711021078685679e-05, "loss": 0.9538, "step": 8137 }, { "epoch": 0.04, "learning_rate": 1.710945281105594e-05, "loss": 0.9422, "step": 8138 }, { "epoch": 0.04, "learning_rate": 1.7108694752655474e-05, "loss": 0.9547, "step": 8139 }, { "epoch": 0.04, "learning_rate": 1.7107936611664207e-05, "loss": 0.9605, "step": 8140 }, { "epoch": 0.04, "learning_rate": 1.7107178388090944e-05, "loss": 0.9578, "step": 8141 }, { "epoch": 0.04, "learning_rate": 1.7106420081944496e-05, "loss": 0.9218, "step": 8142 }, { "epoch": 0.04, "learning_rate": 1.7105661693233676e-05, "loss": 0.9345, "step": 8143 }, { "epoch": 0.04, "learning_rate": 1.710490322196729e-05, "loss": 0.935, "step": 8144 }, { "epoch": 0.04, "learning_rate": 1.710414466815415e-05, "loss": 0.9689, "step": 8145 }, { "epoch": 0.04, "learning_rate": 1.710338603180307e-05, "loss": 0.9231, "step": 8146 }, { "epoch": 0.04, "learning_rate": 1.7102627312922866e-05, "loss": 0.9593, "step": 8147 }, { "epoch": 0.04, "learning_rate": 1.7101868511522352e-05, "loss": 0.9759, "step": 8148 }, { "epoch": 0.04, "learning_rate": 1.7101109627610346e-05, "loss": 0.9467, "step": 8149 }, { "epoch": 0.04, "learning_rate": 1.7100350661195663e-05, "loss": 0.9172, "step": 8150 }, { "epoch": 0.04, "learning_rate": 1.7099591612287117e-05, "loss": 0.9098, "step": 8151 }, { "epoch": 0.04, "learning_rate": 1.7098832480893534e-05, "loss": 0.9415, "step": 8152 }, { "epoch": 0.04, "learning_rate": 1.709807326702373e-05, "loss": 0.9437, "step": 8153 }, { "epoch": 0.04, "learning_rate": 1.7097313970686526e-05, "loss": 0.9462, "step": 8154 }, { "epoch": 0.04, "learning_rate": 1.7096554591890742e-05, "loss": 0.9348, "step": 8155 }, { "epoch": 0.04, "learning_rate": 1.7095795130645203e-05, "loss": 0.9431, "step": 8156 }, { "epoch": 0.04, "learning_rate": 1.7095035586958735e-05, "loss": 0.9335, "step": 8157 }, { "epoch": 0.04, "learning_rate": 1.7094275960840157e-05, "loss": 0.9206, "step": 8158 }, { "epoch": 0.04, "learning_rate": 1.7093516252298296e-05, "loss": 0.9754, "step": 8159 }, { "epoch": 0.04, "learning_rate": 1.7092756461341983e-05, "loss": 0.9186, "step": 8160 }, { "epoch": 0.04, "learning_rate": 1.709199658798004e-05, "loss": 0.9433, "step": 8161 }, { "epoch": 0.04, "learning_rate": 1.70912366322213e-05, "loss": 0.9195, "step": 8162 }, { "epoch": 0.04, "learning_rate": 1.709047659407459e-05, "loss": 0.9764, "step": 8163 }, { "epoch": 0.04, "learning_rate": 1.708971647354874e-05, "loss": 0.9012, "step": 8164 }, { "epoch": 0.04, "learning_rate": 1.7088956270652578e-05, "loss": 0.9243, "step": 8165 }, { "epoch": 0.04, "learning_rate": 1.7088195985394945e-05, "loss": 0.9513, "step": 8166 }, { "epoch": 0.04, "learning_rate": 1.7087435617784667e-05, "loss": 0.9359, "step": 8167 }, { "epoch": 0.04, "learning_rate": 1.7086675167830578e-05, "loss": 0.9629, "step": 8168 }, { "epoch": 0.04, "learning_rate": 1.708591463554152e-05, "loss": 0.9725, "step": 8169 }, { "epoch": 0.04, "learning_rate": 1.7085154020926323e-05, "loss": 0.9493, "step": 8170 }, { "epoch": 0.04, "learning_rate": 1.7084393323993825e-05, "loss": 0.9481, "step": 8171 }, { "epoch": 0.04, "learning_rate": 1.7083632544752866e-05, "loss": 0.9444, "step": 8172 }, { "epoch": 0.04, "learning_rate": 1.7082871683212282e-05, "loss": 0.9362, "step": 8173 }, { "epoch": 0.04, "learning_rate": 1.7082110739380912e-05, "loss": 0.9144, "step": 8174 }, { "epoch": 0.04, "learning_rate": 1.7081349713267603e-05, "loss": 0.9737, "step": 8175 }, { "epoch": 0.04, "learning_rate": 1.7080588604881194e-05, "loss": 0.9519, "step": 8176 }, { "epoch": 0.04, "learning_rate": 1.7079827414230522e-05, "loss": 0.9782, "step": 8177 }, { "epoch": 0.04, "learning_rate": 1.707906614132444e-05, "loss": 0.9095, "step": 8178 }, { "epoch": 0.04, "learning_rate": 1.7078304786171787e-05, "loss": 0.9095, "step": 8179 }, { "epoch": 0.04, "learning_rate": 1.707754334878141e-05, "loss": 0.9172, "step": 8180 }, { "epoch": 0.04, "learning_rate": 1.7076781829162155e-05, "loss": 0.9267, "step": 8181 }, { "epoch": 0.04, "learning_rate": 1.707602022732287e-05, "loss": 0.9462, "step": 8182 }, { "epoch": 0.04, "learning_rate": 1.70752585432724e-05, "loss": 0.9407, "step": 8183 }, { "epoch": 0.04, "learning_rate": 1.7074496777019606e-05, "loss": 0.9404, "step": 8184 }, { "epoch": 0.04, "learning_rate": 1.7073734928573325e-05, "loss": 0.9585, "step": 8185 }, { "epoch": 0.04, "learning_rate": 1.7072972997942415e-05, "loss": 0.9392, "step": 8186 }, { "epoch": 0.04, "learning_rate": 1.7072210985135726e-05, "loss": 0.9345, "step": 8187 }, { "epoch": 0.04, "learning_rate": 1.7071448890162112e-05, "loss": 0.8955, "step": 8188 }, { "epoch": 0.04, "learning_rate": 1.707068671303043e-05, "loss": 0.9995, "step": 8189 }, { "epoch": 0.04, "learning_rate": 1.7069924453749532e-05, "loss": 0.9516, "step": 8190 }, { "epoch": 0.04, "learning_rate": 1.7069162112328273e-05, "loss": 0.9714, "step": 8191 }, { "epoch": 0.04, "learning_rate": 1.7068399688775515e-05, "loss": 0.9589, "step": 8192 }, { "epoch": 0.04, "learning_rate": 1.706763718310011e-05, "loss": 0.921, "step": 8193 }, { "epoch": 0.04, "learning_rate": 1.7066874595310923e-05, "loss": 0.9031, "step": 8194 }, { "epoch": 0.04, "learning_rate": 1.706611192541681e-05, "loss": 0.9475, "step": 8195 }, { "epoch": 0.04, "learning_rate": 1.7065349173426633e-05, "loss": 0.9563, "step": 8196 }, { "epoch": 0.04, "learning_rate": 1.706458633934925e-05, "loss": 0.9331, "step": 8197 }, { "epoch": 0.04, "learning_rate": 1.7063823423193534e-05, "loss": 0.9145, "step": 8198 }, { "epoch": 0.04, "learning_rate": 1.706306042496834e-05, "loss": 0.97, "step": 8199 }, { "epoch": 0.04, "learning_rate": 1.7062297344682534e-05, "loss": 0.977, "step": 8200 }, { "epoch": 0.04, "learning_rate": 1.7061534182344983e-05, "loss": 0.9345, "step": 8201 }, { "epoch": 0.04, "learning_rate": 1.7060770937964556e-05, "loss": 0.9395, "step": 8202 }, { "epoch": 0.04, "learning_rate": 1.7060007611550116e-05, "loss": 0.914, "step": 8203 }, { "epoch": 0.04, "learning_rate": 1.7059244203110533e-05, "loss": 0.9535, "step": 8204 }, { "epoch": 0.04, "learning_rate": 1.7058480712654678e-05, "loss": 0.9455, "step": 8205 }, { "epoch": 0.04, "learning_rate": 1.705771714019142e-05, "loss": 0.9551, "step": 8206 }, { "epoch": 0.04, "learning_rate": 1.7056953485729632e-05, "loss": 0.9784, "step": 8207 }, { "epoch": 0.04, "learning_rate": 1.7056189749278184e-05, "loss": 0.8767, "step": 8208 }, { "epoch": 0.04, "learning_rate": 1.7055425930845954e-05, "loss": 0.9613, "step": 8209 }, { "epoch": 0.04, "learning_rate": 1.7054662030441806e-05, "loss": 0.942, "step": 8210 }, { "epoch": 0.04, "learning_rate": 1.705389804807463e-05, "loss": 0.9579, "step": 8211 }, { "epoch": 0.04, "learning_rate": 1.705313398375329e-05, "loss": 0.9532, "step": 8212 }, { "epoch": 0.04, "learning_rate": 1.7052369837486664e-05, "loss": 0.9594, "step": 8213 }, { "epoch": 0.04, "learning_rate": 1.7051605609283638e-05, "loss": 0.9043, "step": 8214 }, { "epoch": 0.04, "learning_rate": 1.7050841299153085e-05, "loss": 0.9783, "step": 8215 }, { "epoch": 0.04, "learning_rate": 1.7050076907103886e-05, "loss": 0.8908, "step": 8216 }, { "epoch": 0.04, "learning_rate": 1.7049312433144923e-05, "loss": 0.9468, "step": 8217 }, { "epoch": 0.04, "learning_rate": 1.7048547877285078e-05, "loss": 0.943, "step": 8218 }, { "epoch": 0.04, "learning_rate": 1.704778323953323e-05, "loss": 0.9042, "step": 8219 }, { "epoch": 0.04, "learning_rate": 1.704701851989827e-05, "loss": 0.8937, "step": 8220 }, { "epoch": 0.04, "learning_rate": 1.7046253718389074e-05, "loss": 0.9293, "step": 8221 }, { "epoch": 0.04, "learning_rate": 1.7045488835014534e-05, "loss": 0.9531, "step": 8222 }, { "epoch": 0.04, "learning_rate": 1.7044723869783537e-05, "loss": 0.9337, "step": 8223 }, { "epoch": 0.04, "learning_rate": 1.7043958822704968e-05, "loss": 0.9379, "step": 8224 }, { "epoch": 0.04, "learning_rate": 1.7043193693787713e-05, "loss": 0.9833, "step": 8225 }, { "epoch": 0.04, "learning_rate": 1.7042428483040665e-05, "loss": 0.918, "step": 8226 }, { "epoch": 0.04, "learning_rate": 1.7041663190472714e-05, "loss": 0.9913, "step": 8227 }, { "epoch": 0.04, "learning_rate": 1.704089781609275e-05, "loss": 0.9497, "step": 8228 }, { "epoch": 0.04, "learning_rate": 1.704013235990967e-05, "loss": 0.9111, "step": 8229 }, { "epoch": 0.04, "learning_rate": 1.7039366821932364e-05, "loss": 0.929, "step": 8230 }, { "epoch": 0.04, "learning_rate": 1.703860120216972e-05, "loss": 0.9006, "step": 8231 }, { "epoch": 0.04, "learning_rate": 1.7037835500630645e-05, "loss": 0.9371, "step": 8232 }, { "epoch": 0.04, "learning_rate": 1.703706971732403e-05, "loss": 0.9443, "step": 8233 }, { "epoch": 0.04, "learning_rate": 1.7036303852258768e-05, "loss": 0.9875, "step": 8234 }, { "epoch": 0.04, "learning_rate": 1.703553790544376e-05, "loss": 0.939, "step": 8235 }, { "epoch": 0.04, "learning_rate": 1.703477187688791e-05, "loss": 0.9147, "step": 8236 }, { "epoch": 0.04, "learning_rate": 1.703400576660011e-05, "loss": 0.9611, "step": 8237 }, { "epoch": 0.04, "learning_rate": 1.7033239574589267e-05, "loss": 0.948, "step": 8238 }, { "epoch": 0.04, "learning_rate": 1.7032473300864276e-05, "loss": 0.9677, "step": 8239 }, { "epoch": 0.04, "learning_rate": 1.7031706945434047e-05, "loss": 0.9512, "step": 8240 }, { "epoch": 0.04, "learning_rate": 1.703094050830748e-05, "loss": 0.9584, "step": 8241 }, { "epoch": 0.04, "learning_rate": 1.703017398949348e-05, "loss": 0.9415, "step": 8242 }, { "epoch": 0.04, "learning_rate": 1.702940738900095e-05, "loss": 0.9584, "step": 8243 }, { "epoch": 0.04, "learning_rate": 1.70286407068388e-05, "loss": 0.9627, "step": 8244 }, { "epoch": 0.04, "learning_rate": 1.7027873943015942e-05, "loss": 0.9687, "step": 8245 }, { "epoch": 0.04, "learning_rate": 1.702710709754128e-05, "loss": 0.921, "step": 8246 }, { "epoch": 0.04, "learning_rate": 1.702634017042372e-05, "loss": 0.9601, "step": 8247 }, { "epoch": 0.04, "learning_rate": 1.7025573161672174e-05, "loss": 0.9441, "step": 8248 }, { "epoch": 0.04, "learning_rate": 1.7024806071295556e-05, "loss": 0.929, "step": 8249 }, { "epoch": 0.04, "learning_rate": 1.7024038899302773e-05, "loss": 0.9146, "step": 8250 }, { "epoch": 0.04, "learning_rate": 1.7023271645702747e-05, "loss": 0.9541, "step": 8251 }, { "epoch": 0.04, "learning_rate": 1.702250431050439e-05, "loss": 0.9647, "step": 8252 }, { "epoch": 0.04, "learning_rate": 1.7021736893716608e-05, "loss": 0.986, "step": 8253 }, { "epoch": 0.04, "learning_rate": 1.7020969395348326e-05, "loss": 0.9279, "step": 8254 }, { "epoch": 0.04, "learning_rate": 1.702020181540846e-05, "loss": 0.9155, "step": 8255 }, { "epoch": 0.04, "learning_rate": 1.7019434153905922e-05, "loss": 0.9168, "step": 8256 }, { "epoch": 0.04, "learning_rate": 1.701866641084964e-05, "loss": 0.9869, "step": 8257 }, { "epoch": 0.04, "learning_rate": 1.7017898586248528e-05, "loss": 0.9508, "step": 8258 }, { "epoch": 0.04, "learning_rate": 1.7017130680111507e-05, "loss": 0.969, "step": 8259 }, { "epoch": 0.04, "learning_rate": 1.7016362692447498e-05, "loss": 0.955, "step": 8260 }, { "epoch": 0.04, "learning_rate": 1.701559462326543e-05, "loss": 0.9258, "step": 8261 }, { "epoch": 0.04, "learning_rate": 1.7014826472574223e-05, "loss": 0.93, "step": 8262 }, { "epoch": 0.04, "learning_rate": 1.7014058240382792e-05, "loss": 0.947, "step": 8263 }, { "epoch": 0.04, "learning_rate": 1.701328992670008e-05, "loss": 0.9251, "step": 8264 }, { "epoch": 0.04, "learning_rate": 1.7012521531535002e-05, "loss": 0.9664, "step": 8265 }, { "epoch": 0.04, "learning_rate": 1.7011753054896487e-05, "loss": 0.9311, "step": 8266 }, { "epoch": 0.04, "learning_rate": 1.7010984496793467e-05, "loss": 0.9334, "step": 8267 }, { "epoch": 0.04, "learning_rate": 1.7010215857234866e-05, "loss": 0.9501, "step": 8268 }, { "epoch": 0.04, "learning_rate": 1.700944713622962e-05, "loss": 0.9524, "step": 8269 }, { "epoch": 0.04, "learning_rate": 1.7008678333786653e-05, "loss": 0.9235, "step": 8270 }, { "epoch": 0.04, "learning_rate": 1.7007909449914904e-05, "loss": 0.9337, "step": 8271 }, { "epoch": 0.04, "learning_rate": 1.7007140484623302e-05, "loss": 0.9094, "step": 8272 }, { "epoch": 0.04, "learning_rate": 1.700637143792079e-05, "loss": 0.9339, "step": 8273 }, { "epoch": 0.04, "learning_rate": 1.7005602309816284e-05, "loss": 0.9423, "step": 8274 }, { "epoch": 0.04, "learning_rate": 1.7004833100318737e-05, "loss": 0.9534, "step": 8275 }, { "epoch": 0.04, "learning_rate": 1.7004063809437082e-05, "loss": 0.9567, "step": 8276 }, { "epoch": 0.04, "learning_rate": 1.7003294437180254e-05, "loss": 0.9628, "step": 8277 }, { "epoch": 0.04, "learning_rate": 1.7002524983557193e-05, "loss": 0.96, "step": 8278 }, { "epoch": 0.04, "learning_rate": 1.7001755448576842e-05, "loss": 0.9234, "step": 8279 }, { "epoch": 0.04, "learning_rate": 1.7000985832248137e-05, "loss": 0.9594, "step": 8280 }, { "epoch": 0.04, "learning_rate": 1.7000216134580017e-05, "loss": 0.9885, "step": 8281 }, { "epoch": 0.04, "learning_rate": 1.699944635558143e-05, "loss": 0.9308, "step": 8282 }, { "epoch": 0.04, "learning_rate": 1.6998676495261318e-05, "loss": 0.9321, "step": 8283 }, { "epoch": 0.04, "learning_rate": 1.699790655362863e-05, "loss": 0.9391, "step": 8284 }, { "epoch": 0.04, "learning_rate": 1.6997136530692303e-05, "loss": 0.9677, "step": 8285 }, { "epoch": 0.04, "learning_rate": 1.699636642646129e-05, "loss": 0.9638, "step": 8286 }, { "epoch": 0.04, "learning_rate": 1.6995596240944535e-05, "loss": 0.9666, "step": 8287 }, { "epoch": 0.04, "learning_rate": 1.6994825974150987e-05, "loss": 0.9588, "step": 8288 }, { "epoch": 0.04, "learning_rate": 1.6994055626089594e-05, "loss": 0.9535, "step": 8289 }, { "epoch": 0.04, "learning_rate": 1.6993285196769305e-05, "loss": 0.9516, "step": 8290 }, { "epoch": 0.04, "learning_rate": 1.699251468619908e-05, "loss": 0.9677, "step": 8291 }, { "epoch": 0.04, "learning_rate": 1.699174409438786e-05, "loss": 0.9479, "step": 8292 }, { "epoch": 0.04, "learning_rate": 1.69909734213446e-05, "loss": 0.9361, "step": 8293 }, { "epoch": 0.04, "learning_rate": 1.699020266707826e-05, "loss": 0.9355, "step": 8294 }, { "epoch": 0.04, "learning_rate": 1.698943183159779e-05, "loss": 0.9149, "step": 8295 }, { "epoch": 0.04, "learning_rate": 1.6988660914912144e-05, "loss": 0.9173, "step": 8296 }, { "epoch": 0.04, "learning_rate": 1.6987889917030283e-05, "loss": 1.0045, "step": 8297 }, { "epoch": 0.04, "learning_rate": 1.698711883796116e-05, "loss": 0.9211, "step": 8298 }, { "epoch": 0.04, "learning_rate": 1.6986347677713743e-05, "loss": 0.8997, "step": 8299 }, { "epoch": 0.04, "learning_rate": 1.698557643629698e-05, "loss": 0.9455, "step": 8300 }, { "epoch": 0.04, "learning_rate": 1.6984805113719846e-05, "loss": 0.9271, "step": 8301 }, { "epoch": 0.04, "learning_rate": 1.6984033709991282e-05, "loss": 0.9257, "step": 8302 }, { "epoch": 0.04, "learning_rate": 1.698326222512027e-05, "loss": 0.9408, "step": 8303 }, { "epoch": 0.04, "learning_rate": 1.698249065911576e-05, "loss": 0.9243, "step": 8304 }, { "epoch": 0.04, "learning_rate": 1.698171901198672e-05, "loss": 0.9791, "step": 8305 }, { "epoch": 0.04, "learning_rate": 1.698094728374212e-05, "loss": 0.9429, "step": 8306 }, { "epoch": 0.04, "learning_rate": 1.698017547439092e-05, "loss": 0.9783, "step": 8307 }, { "epoch": 0.04, "learning_rate": 1.6979403583942095e-05, "loss": 0.929, "step": 8308 }, { "epoch": 0.04, "learning_rate": 1.6978631612404603e-05, "loss": 0.9663, "step": 8309 }, { "epoch": 0.04, "learning_rate": 1.697785955978742e-05, "loss": 0.9673, "step": 8310 }, { "epoch": 0.04, "learning_rate": 1.697708742609951e-05, "loss": 0.9345, "step": 8311 }, { "epoch": 0.04, "learning_rate": 1.6976315211349848e-05, "loss": 0.9699, "step": 8312 }, { "epoch": 0.04, "learning_rate": 1.6975542915547408e-05, "loss": 0.931, "step": 8313 }, { "epoch": 0.04, "learning_rate": 1.6974770538701157e-05, "loss": 0.9127, "step": 8314 }, { "epoch": 0.04, "learning_rate": 1.6973998080820074e-05, "loss": 0.9808, "step": 8315 }, { "epoch": 0.04, "learning_rate": 1.6973225541913133e-05, "loss": 0.9365, "step": 8316 }, { "epoch": 0.04, "learning_rate": 1.6972452921989308e-05, "loss": 0.9419, "step": 8317 }, { "epoch": 0.04, "learning_rate": 1.697168022105757e-05, "loss": 0.9157, "step": 8318 }, { "epoch": 0.04, "learning_rate": 1.6970907439126904e-05, "loss": 0.9408, "step": 8319 }, { "epoch": 0.04, "learning_rate": 1.6970134576206287e-05, "loss": 0.8882, "step": 8320 }, { "epoch": 0.04, "learning_rate": 1.6969361632304702e-05, "loss": 0.9868, "step": 8321 }, { "epoch": 0.04, "learning_rate": 1.6968588607431117e-05, "loss": 0.9606, "step": 8322 }, { "epoch": 0.04, "learning_rate": 1.696781550159453e-05, "loss": 0.9763, "step": 8323 }, { "epoch": 0.04, "learning_rate": 1.6967042314803908e-05, "loss": 0.966, "step": 8324 }, { "epoch": 0.04, "learning_rate": 1.696626904706824e-05, "loss": 0.9823, "step": 8325 }, { "epoch": 0.04, "learning_rate": 1.6965495698396512e-05, "loss": 0.9192, "step": 8326 }, { "epoch": 0.04, "learning_rate": 1.6964722268797707e-05, "loss": 0.9599, "step": 8327 }, { "epoch": 0.04, "learning_rate": 1.6963948758280815e-05, "loss": 0.9288, "step": 8328 }, { "epoch": 0.04, "learning_rate": 1.6963175166854813e-05, "loss": 1.0074, "step": 8329 }, { "epoch": 0.04, "learning_rate": 1.6962401494528702e-05, "loss": 0.9128, "step": 8330 }, { "epoch": 0.04, "learning_rate": 1.696162774131146e-05, "loss": 0.9664, "step": 8331 }, { "epoch": 0.04, "learning_rate": 1.696085390721208e-05, "loss": 0.9505, "step": 8332 }, { "epoch": 0.04, "learning_rate": 1.6960079992239554e-05, "loss": 0.9911, "step": 8333 }, { "epoch": 0.04, "learning_rate": 1.695930599640287e-05, "loss": 0.9379, "step": 8334 }, { "epoch": 0.04, "learning_rate": 1.695853191971103e-05, "loss": 0.9657, "step": 8335 }, { "epoch": 0.04, "learning_rate": 1.695775776217301e-05, "loss": 0.8894, "step": 8336 }, { "epoch": 0.04, "learning_rate": 1.6956983523797823e-05, "loss": 0.9641, "step": 8337 }, { "epoch": 0.04, "learning_rate": 1.6956209204594452e-05, "loss": 0.9053, "step": 8338 }, { "epoch": 0.04, "learning_rate": 1.69554348045719e-05, "loss": 0.9477, "step": 8339 }, { "epoch": 0.04, "learning_rate": 1.6954660323739163e-05, "loss": 0.9374, "step": 8340 }, { "epoch": 0.04, "learning_rate": 1.6953885762105232e-05, "loss": 0.9643, "step": 8341 }, { "epoch": 0.04, "learning_rate": 1.695311111967912e-05, "loss": 0.8999, "step": 8342 }, { "epoch": 0.04, "learning_rate": 1.6952336396469813e-05, "loss": 0.9028, "step": 8343 }, { "epoch": 0.04, "learning_rate": 1.695156159248632e-05, "loss": 0.9107, "step": 8344 }, { "epoch": 0.04, "learning_rate": 1.695078670773764e-05, "loss": 0.9118, "step": 8345 }, { "epoch": 0.04, "learning_rate": 1.6950011742232778e-05, "loss": 0.9751, "step": 8346 }, { "epoch": 0.04, "learning_rate": 1.6949236695980734e-05, "loss": 0.9187, "step": 8347 }, { "epoch": 0.04, "learning_rate": 1.694846156899052e-05, "loss": 0.9367, "step": 8348 }, { "epoch": 0.04, "learning_rate": 1.694768636127113e-05, "loss": 0.9377, "step": 8349 }, { "epoch": 0.04, "learning_rate": 1.694691107283158e-05, "loss": 0.9467, "step": 8350 }, { "epoch": 0.04, "learning_rate": 1.6946135703680874e-05, "loss": 0.9394, "step": 8351 }, { "epoch": 0.04, "learning_rate": 1.694536025382802e-05, "loss": 0.9237, "step": 8352 }, { "epoch": 0.05, "learning_rate": 1.6944584723282033e-05, "loss": 0.9345, "step": 8353 }, { "epoch": 0.05, "learning_rate": 1.6943809112051922e-05, "loss": 0.9506, "step": 8354 }, { "epoch": 0.05, "learning_rate": 1.6943033420146686e-05, "loss": 0.9539, "step": 8355 }, { "epoch": 0.05, "learning_rate": 1.694225764757535e-05, "loss": 0.9398, "step": 8356 }, { "epoch": 0.05, "learning_rate": 1.6941481794346927e-05, "loss": 0.9302, "step": 8357 }, { "epoch": 0.05, "learning_rate": 1.6940705860470426e-05, "loss": 0.9151, "step": 8358 }, { "epoch": 0.05, "learning_rate": 1.693992984595486e-05, "loss": 0.9057, "step": 8359 }, { "epoch": 0.05, "learning_rate": 1.6939153750809252e-05, "loss": 0.9279, "step": 8360 }, { "epoch": 0.05, "learning_rate": 1.6938377575042615e-05, "loss": 0.9514, "step": 8361 }, { "epoch": 0.05, "learning_rate": 1.6937601318663967e-05, "loss": 0.9159, "step": 8362 }, { "epoch": 0.05, "learning_rate": 1.693682498168233e-05, "loss": 0.9508, "step": 8363 }, { "epoch": 0.05, "learning_rate": 1.6936048564106714e-05, "loss": 0.9125, "step": 8364 }, { "epoch": 0.05, "learning_rate": 1.6935272065946154e-05, "loss": 0.9777, "step": 8365 }, { "epoch": 0.05, "learning_rate": 1.6934495487209656e-05, "loss": 0.9409, "step": 8366 }, { "epoch": 0.05, "learning_rate": 1.6933718827906255e-05, "loss": 0.9764, "step": 8367 }, { "epoch": 0.05, "learning_rate": 1.6932942088044972e-05, "loss": 0.9698, "step": 8368 }, { "epoch": 0.05, "learning_rate": 1.6932165267634826e-05, "loss": 0.9543, "step": 8369 }, { "epoch": 0.05, "learning_rate": 1.6931388366684848e-05, "loss": 0.9273, "step": 8370 }, { "epoch": 0.05, "learning_rate": 1.6930611385204062e-05, "loss": 0.9426, "step": 8371 }, { "epoch": 0.05, "learning_rate": 1.6929834323201496e-05, "loss": 0.907, "step": 8372 }, { "epoch": 0.05, "learning_rate": 1.6929057180686174e-05, "loss": 0.9247, "step": 8373 }, { "epoch": 0.05, "learning_rate": 1.692827995766713e-05, "loss": 0.9498, "step": 8374 }, { "epoch": 0.05, "learning_rate": 1.692750265415339e-05, "loss": 0.9297, "step": 8375 }, { "epoch": 0.05, "learning_rate": 1.692672527015399e-05, "loss": 0.9452, "step": 8376 }, { "epoch": 0.05, "learning_rate": 1.692594780567796e-05, "loss": 0.9497, "step": 8377 }, { "epoch": 0.05, "learning_rate": 1.692517026073433e-05, "loss": 0.9027, "step": 8378 }, { "epoch": 0.05, "learning_rate": 1.6924392635332134e-05, "loss": 0.9734, "step": 8379 }, { "epoch": 0.05, "learning_rate": 1.692361492948041e-05, "loss": 0.9381, "step": 8380 }, { "epoch": 0.05, "learning_rate": 1.692283714318819e-05, "loss": 0.9381, "step": 8381 }, { "epoch": 0.05, "learning_rate": 1.6922059276464516e-05, "loss": 0.899, "step": 8382 }, { "epoch": 0.05, "learning_rate": 1.692128132931842e-05, "loss": 0.9265, "step": 8383 }, { "epoch": 0.05, "learning_rate": 1.6920503301758943e-05, "loss": 0.906, "step": 8384 }, { "epoch": 0.05, "learning_rate": 1.6919725193795123e-05, "loss": 0.9572, "step": 8385 }, { "epoch": 0.05, "learning_rate": 1.6918947005436e-05, "loss": 0.9133, "step": 8386 }, { "epoch": 0.05, "learning_rate": 1.691816873669062e-05, "loss": 0.9478, "step": 8387 }, { "epoch": 0.05, "learning_rate": 1.691739038756802e-05, "loss": 0.9184, "step": 8388 }, { "epoch": 0.05, "learning_rate": 1.691661195807724e-05, "loss": 0.9294, "step": 8389 }, { "epoch": 0.05, "learning_rate": 1.6915833448227334e-05, "loss": 0.9758, "step": 8390 }, { "epoch": 0.05, "learning_rate": 1.6915054858027334e-05, "loss": 0.9644, "step": 8391 }, { "epoch": 0.05, "learning_rate": 1.6914276187486304e-05, "loss": 0.9676, "step": 8392 }, { "epoch": 0.05, "learning_rate": 1.6913497436613272e-05, "loss": 0.9687, "step": 8393 }, { "epoch": 0.05, "learning_rate": 1.6912718605417295e-05, "loss": 0.9476, "step": 8394 }, { "epoch": 0.05, "learning_rate": 1.6911939693907422e-05, "loss": 0.9846, "step": 8395 }, { "epoch": 0.05, "learning_rate": 1.69111607020927e-05, "loss": 0.9376, "step": 8396 }, { "epoch": 0.05, "learning_rate": 1.691038162998218e-05, "loss": 0.9471, "step": 8397 }, { "epoch": 0.05, "learning_rate": 1.6909602477584915e-05, "loss": 0.9132, "step": 8398 }, { "epoch": 0.05, "learning_rate": 1.6908823244909957e-05, "loss": 0.9737, "step": 8399 }, { "epoch": 0.05, "learning_rate": 1.690804393196636e-05, "loss": 0.8891, "step": 8400 }, { "epoch": 0.05, "learning_rate": 1.690726453876317e-05, "loss": 0.9206, "step": 8401 }, { "epoch": 0.05, "learning_rate": 1.6906485065309456e-05, "loss": 0.9732, "step": 8402 }, { "epoch": 0.05, "learning_rate": 1.6905705511614267e-05, "loss": 0.9653, "step": 8403 }, { "epoch": 0.05, "learning_rate": 1.690492587768666e-05, "loss": 0.9163, "step": 8404 }, { "epoch": 0.05, "learning_rate": 1.6904146163535694e-05, "loss": 0.9351, "step": 8405 }, { "epoch": 0.05, "learning_rate": 1.6903366369170425e-05, "loss": 0.9286, "step": 8406 }, { "epoch": 0.05, "learning_rate": 1.6902586494599917e-05, "loss": 0.9127, "step": 8407 }, { "epoch": 0.05, "learning_rate": 1.690180653983323e-05, "loss": 0.8857, "step": 8408 }, { "epoch": 0.05, "learning_rate": 1.6901026504879424e-05, "loss": 0.9807, "step": 8409 }, { "epoch": 0.05, "learning_rate": 1.690024638974756e-05, "loss": 0.9634, "step": 8410 }, { "epoch": 0.05, "learning_rate": 1.6899466194446706e-05, "loss": 0.9557, "step": 8411 }, { "epoch": 0.05, "learning_rate": 1.6898685918985926e-05, "loss": 0.9342, "step": 8412 }, { "epoch": 0.05, "learning_rate": 1.6897905563374285e-05, "loss": 0.9636, "step": 8413 }, { "epoch": 0.05, "learning_rate": 1.6897125127620844e-05, "loss": 0.9234, "step": 8414 }, { "epoch": 0.05, "learning_rate": 1.689634461173468e-05, "loss": 0.9327, "step": 8415 }, { "epoch": 0.05, "learning_rate": 1.6895564015724854e-05, "loss": 0.9112, "step": 8416 }, { "epoch": 0.05, "learning_rate": 1.6894783339600435e-05, "loss": 0.9577, "step": 8417 }, { "epoch": 0.05, "learning_rate": 1.68940025833705e-05, "loss": 0.9362, "step": 8418 }, { "epoch": 0.05, "learning_rate": 1.6893221747044113e-05, "loss": 0.9366, "step": 8419 }, { "epoch": 0.05, "learning_rate": 1.6892440830630348e-05, "loss": 0.9214, "step": 8420 }, { "epoch": 0.05, "learning_rate": 1.6891659834138277e-05, "loss": 0.9238, "step": 8421 }, { "epoch": 0.05, "learning_rate": 1.689087875757698e-05, "loss": 1.0083, "step": 8422 }, { "epoch": 0.05, "learning_rate": 1.6890097600955522e-05, "loss": 0.9384, "step": 8423 }, { "epoch": 0.05, "learning_rate": 1.6889316364282983e-05, "loss": 0.9248, "step": 8424 }, { "epoch": 0.05, "learning_rate": 1.6888535047568444e-05, "loss": 0.9602, "step": 8425 }, { "epoch": 0.05, "learning_rate": 1.6887753650820976e-05, "loss": 0.8895, "step": 8426 }, { "epoch": 0.05, "learning_rate": 1.688697217404966e-05, "loss": 0.916, "step": 8427 }, { "epoch": 0.05, "learning_rate": 1.688619061726358e-05, "loss": 0.9627, "step": 8428 }, { "epoch": 0.05, "learning_rate": 1.688540898047181e-05, "loss": 0.9587, "step": 8429 }, { "epoch": 0.05, "learning_rate": 1.6884627263683428e-05, "loss": 0.9285, "step": 8430 }, { "epoch": 0.05, "learning_rate": 1.6883845466907528e-05, "loss": 0.9536, "step": 8431 }, { "epoch": 0.05, "learning_rate": 1.6883063590153184e-05, "loss": 0.8827, "step": 8432 }, { "epoch": 0.05, "learning_rate": 1.6882281633429484e-05, "loss": 0.8787, "step": 8433 }, { "epoch": 0.05, "learning_rate": 1.688149959674551e-05, "loss": 0.9685, "step": 8434 }, { "epoch": 0.05, "learning_rate": 1.688071748011035e-05, "loss": 0.9498, "step": 8435 }, { "epoch": 0.05, "learning_rate": 1.687993528353309e-05, "loss": 0.9275, "step": 8436 }, { "epoch": 0.05, "learning_rate": 1.687915300702282e-05, "loss": 0.9467, "step": 8437 }, { "epoch": 0.05, "learning_rate": 1.6878370650588624e-05, "loss": 0.9596, "step": 8438 }, { "epoch": 0.05, "learning_rate": 1.68775882142396e-05, "loss": 0.9488, "step": 8439 }, { "epoch": 0.05, "learning_rate": 1.687680569798483e-05, "loss": 0.9208, "step": 8440 }, { "epoch": 0.05, "learning_rate": 1.6876023101833406e-05, "loss": 0.9316, "step": 8441 }, { "epoch": 0.05, "learning_rate": 1.6875240425794427e-05, "loss": 0.9485, "step": 8442 }, { "epoch": 0.05, "learning_rate": 1.6874457669876984e-05, "loss": 0.9812, "step": 8443 }, { "epoch": 0.05, "learning_rate": 1.687367483409016e-05, "loss": 0.9426, "step": 8444 }, { "epoch": 0.05, "learning_rate": 1.687289191844307e-05, "loss": 0.9371, "step": 8445 }, { "epoch": 0.05, "learning_rate": 1.6872108922944798e-05, "loss": 0.9696, "step": 8446 }, { "epoch": 0.05, "learning_rate": 1.6871325847604438e-05, "loss": 0.9669, "step": 8447 }, { "epoch": 0.05, "learning_rate": 1.68705426924311e-05, "loss": 0.9249, "step": 8448 }, { "epoch": 0.05, "learning_rate": 1.6869759457433872e-05, "loss": 0.9401, "step": 8449 }, { "epoch": 0.05, "learning_rate": 1.6868976142621858e-05, "loss": 0.8876, "step": 8450 }, { "epoch": 0.05, "learning_rate": 1.6868192748004157e-05, "loss": 0.9171, "step": 8451 }, { "epoch": 0.05, "learning_rate": 1.6867409273589875e-05, "loss": 0.9585, "step": 8452 }, { "epoch": 0.05, "learning_rate": 1.686662571938811e-05, "loss": 0.9469, "step": 8453 }, { "epoch": 0.05, "learning_rate": 1.686584208540797e-05, "loss": 0.9081, "step": 8454 }, { "epoch": 0.05, "learning_rate": 1.686505837165856e-05, "loss": 0.9716, "step": 8455 }, { "epoch": 0.05, "learning_rate": 1.6864274578148975e-05, "loss": 0.9243, "step": 8456 }, { "epoch": 0.05, "learning_rate": 1.686349070488833e-05, "loss": 0.9614, "step": 8457 }, { "epoch": 0.05, "learning_rate": 1.6862706751885733e-05, "loss": 0.9558, "step": 8458 }, { "epoch": 0.05, "learning_rate": 1.686192271915029e-05, "loss": 1.0038, "step": 8459 }, { "epoch": 0.05, "learning_rate": 1.686113860669111e-05, "loss": 0.9429, "step": 8460 }, { "epoch": 0.05, "learning_rate": 1.6860354414517304e-05, "loss": 0.936, "step": 8461 }, { "epoch": 0.05, "learning_rate": 1.6859570142637985e-05, "loss": 0.9078, "step": 8462 }, { "epoch": 0.05, "learning_rate": 1.6858785791062254e-05, "loss": 0.9307, "step": 8463 }, { "epoch": 0.05, "learning_rate": 1.685800135979924e-05, "loss": 0.9769, "step": 8464 }, { "epoch": 0.05, "learning_rate": 1.6857216848858045e-05, "loss": 0.9426, "step": 8465 }, { "epoch": 0.05, "learning_rate": 1.6856432258247787e-05, "loss": 0.9119, "step": 8466 }, { "epoch": 0.05, "learning_rate": 1.6855647587977584e-05, "loss": 0.9289, "step": 8467 }, { "epoch": 0.05, "learning_rate": 1.6854862838056546e-05, "loss": 0.9227, "step": 8468 }, { "epoch": 0.05, "learning_rate": 1.68540780084938e-05, "loss": 0.9502, "step": 8469 }, { "epoch": 0.05, "learning_rate": 1.685329309929846e-05, "loss": 0.9169, "step": 8470 }, { "epoch": 0.05, "learning_rate": 1.685250811047964e-05, "loss": 0.9529, "step": 8471 }, { "epoch": 0.05, "learning_rate": 1.6851723042046468e-05, "loss": 0.9284, "step": 8472 }, { "epoch": 0.05, "learning_rate": 1.685093789400806e-05, "loss": 0.9413, "step": 8473 }, { "epoch": 0.05, "learning_rate": 1.685015266637354e-05, "loss": 0.9124, "step": 8474 }, { "epoch": 0.05, "learning_rate": 1.684936735915203e-05, "loss": 0.9534, "step": 8475 }, { "epoch": 0.05, "learning_rate": 1.684858197235266e-05, "loss": 0.9402, "step": 8476 }, { "epoch": 0.05, "learning_rate": 1.6847796505984547e-05, "loss": 0.906, "step": 8477 }, { "epoch": 0.05, "learning_rate": 1.6847010960056816e-05, "loss": 0.8918, "step": 8478 }, { "epoch": 0.05, "learning_rate": 1.68462253345786e-05, "loss": 0.9892, "step": 8479 }, { "epoch": 0.05, "learning_rate": 1.6845439629559027e-05, "loss": 0.9542, "step": 8480 }, { "epoch": 0.05, "learning_rate": 1.6844653845007217e-05, "loss": 0.9207, "step": 8481 }, { "epoch": 0.05, "learning_rate": 1.6843867980932308e-05, "loss": 0.9028, "step": 8482 }, { "epoch": 0.05, "learning_rate": 1.6843082037343425e-05, "loss": 0.9408, "step": 8483 }, { "epoch": 0.05, "learning_rate": 1.6842296014249705e-05, "loss": 0.9208, "step": 8484 }, { "epoch": 0.05, "learning_rate": 1.6841509911660273e-05, "loss": 0.9758, "step": 8485 }, { "epoch": 0.05, "learning_rate": 1.684072372958427e-05, "loss": 0.9191, "step": 8486 }, { "epoch": 0.05, "learning_rate": 1.6839937468030826e-05, "loss": 0.9122, "step": 8487 }, { "epoch": 0.05, "learning_rate": 1.6839151127009072e-05, "loss": 0.9622, "step": 8488 }, { "epoch": 0.05, "learning_rate": 1.6838364706528148e-05, "loss": 0.9606, "step": 8489 }, { "epoch": 0.05, "learning_rate": 1.6837578206597193e-05, "loss": 0.9222, "step": 8490 }, { "epoch": 0.05, "learning_rate": 1.6836791627225343e-05, "loss": 0.9332, "step": 8491 }, { "epoch": 0.05, "learning_rate": 1.6836004968421737e-05, "loss": 0.9112, "step": 8492 }, { "epoch": 0.05, "learning_rate": 1.6835218230195512e-05, "loss": 0.9343, "step": 8493 }, { "epoch": 0.05, "learning_rate": 1.683443141255581e-05, "loss": 0.9554, "step": 8494 }, { "epoch": 0.05, "learning_rate": 1.6833644515511775e-05, "loss": 0.9472, "step": 8495 }, { "epoch": 0.05, "learning_rate": 1.6832857539072545e-05, "loss": 0.9338, "step": 8496 }, { "epoch": 0.05, "learning_rate": 1.683207048324727e-05, "loss": 0.9144, "step": 8497 }, { "epoch": 0.05, "learning_rate": 1.6831283348045085e-05, "loss": 0.9213, "step": 8498 }, { "epoch": 0.05, "learning_rate": 1.6830496133475143e-05, "loss": 0.9488, "step": 8499 }, { "epoch": 0.05, "learning_rate": 1.6829708839546587e-05, "loss": 0.9495, "step": 8500 }, { "epoch": 0.05, "learning_rate": 1.6828921466268562e-05, "loss": 0.9534, "step": 8501 }, { "epoch": 0.05, "learning_rate": 1.6828134013650216e-05, "loss": 0.9123, "step": 8502 }, { "epoch": 0.05, "learning_rate": 1.6827346481700705e-05, "loss": 0.9392, "step": 8503 }, { "epoch": 0.05, "learning_rate": 1.682655887042917e-05, "loss": 0.9148, "step": 8504 }, { "epoch": 0.05, "learning_rate": 1.682577117984477e-05, "loss": 0.9688, "step": 8505 }, { "epoch": 0.05, "learning_rate": 1.6824983409956648e-05, "loss": 0.9614, "step": 8506 }, { "epoch": 0.05, "learning_rate": 1.6824195560773962e-05, "loss": 0.9469, "step": 8507 }, { "epoch": 0.05, "learning_rate": 1.6823407632305865e-05, "loss": 0.9544, "step": 8508 }, { "epoch": 0.05, "learning_rate": 1.6822619624561508e-05, "loss": 0.9366, "step": 8509 }, { "epoch": 0.05, "learning_rate": 1.682183153755005e-05, "loss": 0.9362, "step": 8510 }, { "epoch": 0.05, "learning_rate": 1.6821043371280644e-05, "loss": 0.9528, "step": 8511 }, { "epoch": 0.05, "learning_rate": 1.682025512576245e-05, "loss": 0.9451, "step": 8512 }, { "epoch": 0.05, "learning_rate": 1.6819466801004622e-05, "loss": 0.9611, "step": 8513 }, { "epoch": 0.05, "learning_rate": 1.6818678397016325e-05, "loss": 0.886, "step": 8514 }, { "epoch": 0.05, "learning_rate": 1.681788991380672e-05, "loss": 0.9605, "step": 8515 }, { "epoch": 0.05, "learning_rate": 1.681710135138496e-05, "loss": 0.9173, "step": 8516 }, { "epoch": 0.05, "learning_rate": 1.6816312709760206e-05, "loss": 0.9623, "step": 8517 }, { "epoch": 0.05, "learning_rate": 1.681552398894163e-05, "loss": 0.9184, "step": 8518 }, { "epoch": 0.05, "learning_rate": 1.681473518893839e-05, "loss": 0.9618, "step": 8519 }, { "epoch": 0.05, "learning_rate": 1.681394630975965e-05, "loss": 0.9473, "step": 8520 }, { "epoch": 0.05, "learning_rate": 1.681315735141458e-05, "loss": 0.9551, "step": 8521 }, { "epoch": 0.05, "learning_rate": 1.6812368313912336e-05, "loss": 0.9499, "step": 8522 }, { "epoch": 0.05, "learning_rate": 1.6811579197262097e-05, "loss": 0.9198, "step": 8523 }, { "epoch": 0.05, "learning_rate": 1.681079000147303e-05, "loss": 0.911, "step": 8524 }, { "epoch": 0.05, "learning_rate": 1.6810000726554294e-05, "loss": 0.9929, "step": 8525 }, { "epoch": 0.05, "learning_rate": 1.680921137251507e-05, "loss": 0.9415, "step": 8526 }, { "epoch": 0.05, "learning_rate": 1.680842193936452e-05, "loss": 0.9277, "step": 8527 }, { "epoch": 0.05, "learning_rate": 1.6807632427111823e-05, "loss": 0.9336, "step": 8528 }, { "epoch": 0.05, "learning_rate": 1.6806842835766147e-05, "loss": 0.9529, "step": 8529 }, { "epoch": 0.05, "learning_rate": 1.680605316533667e-05, "loss": 0.9071, "step": 8530 }, { "epoch": 0.05, "learning_rate": 1.6805263415832565e-05, "loss": 0.9299, "step": 8531 }, { "epoch": 0.05, "learning_rate": 1.6804473587263004e-05, "loss": 0.9345, "step": 8532 }, { "epoch": 0.05, "learning_rate": 1.680368367963717e-05, "loss": 0.9417, "step": 8533 }, { "epoch": 0.05, "learning_rate": 1.680289369296423e-05, "loss": 0.9561, "step": 8534 }, { "epoch": 0.05, "learning_rate": 1.6802103627253377e-05, "loss": 0.9414, "step": 8535 }, { "epoch": 0.05, "learning_rate": 1.6801313482513775e-05, "loss": 0.9252, "step": 8536 }, { "epoch": 0.05, "learning_rate": 1.680052325875462e-05, "loss": 0.9318, "step": 8537 }, { "epoch": 0.05, "learning_rate": 1.6799732955985076e-05, "loss": 0.8999, "step": 8538 }, { "epoch": 0.05, "learning_rate": 1.6798942574214336e-05, "loss": 0.9347, "step": 8539 }, { "epoch": 0.05, "learning_rate": 1.679815211345158e-05, "loss": 0.9128, "step": 8540 }, { "epoch": 0.05, "learning_rate": 1.6797361573705996e-05, "loss": 0.9648, "step": 8541 }, { "epoch": 0.05, "learning_rate": 1.679657095498676e-05, "loss": 0.9084, "step": 8542 }, { "epoch": 0.05, "learning_rate": 1.6795780257303064e-05, "loss": 0.8894, "step": 8543 }, { "epoch": 0.05, "learning_rate": 1.6794989480664093e-05, "loss": 0.9459, "step": 8544 }, { "epoch": 0.05, "learning_rate": 1.6794198625079033e-05, "loss": 0.9104, "step": 8545 }, { "epoch": 0.05, "learning_rate": 1.6793407690557075e-05, "loss": 0.9365, "step": 8546 }, { "epoch": 0.05, "learning_rate": 1.6792616677107407e-05, "loss": 0.9339, "step": 8547 }, { "epoch": 0.05, "learning_rate": 1.679182558473922e-05, "loss": 0.9172, "step": 8548 }, { "epoch": 0.05, "learning_rate": 1.67910344134617e-05, "loss": 0.9529, "step": 8549 }, { "epoch": 0.05, "learning_rate": 1.6790243163284047e-05, "loss": 0.9336, "step": 8550 }, { "epoch": 0.05, "learning_rate": 1.678945183421545e-05, "loss": 0.9647, "step": 8551 }, { "epoch": 0.05, "learning_rate": 1.6788660426265104e-05, "loss": 0.9149, "step": 8552 }, { "epoch": 0.05, "learning_rate": 1.6787868939442204e-05, "loss": 0.9231, "step": 8553 }, { "epoch": 0.05, "learning_rate": 1.6787077373755942e-05, "loss": 0.9281, "step": 8554 }, { "epoch": 0.05, "learning_rate": 1.678628572921552e-05, "loss": 0.9609, "step": 8555 }, { "epoch": 0.05, "learning_rate": 1.678549400583013e-05, "loss": 0.9399, "step": 8556 }, { "epoch": 0.05, "learning_rate": 1.6784702203608974e-05, "loss": 0.952, "step": 8557 }, { "epoch": 0.05, "learning_rate": 1.6783910322561254e-05, "loss": 0.9547, "step": 8558 }, { "epoch": 0.05, "learning_rate": 1.6783118362696162e-05, "loss": 0.9816, "step": 8559 }, { "epoch": 0.05, "learning_rate": 1.6782326324022908e-05, "loss": 0.9296, "step": 8560 }, { "epoch": 0.05, "learning_rate": 1.678153420655069e-05, "loss": 0.9541, "step": 8561 }, { "epoch": 0.05, "learning_rate": 1.678074201028871e-05, "loss": 0.8935, "step": 8562 }, { "epoch": 0.05, "learning_rate": 1.6779949735246176e-05, "loss": 0.8986, "step": 8563 }, { "epoch": 0.05, "learning_rate": 1.6779157381432285e-05, "loss": 0.9341, "step": 8564 }, { "epoch": 0.05, "learning_rate": 1.6778364948856255e-05, "loss": 0.9144, "step": 8565 }, { "epoch": 0.05, "learning_rate": 1.677757243752728e-05, "loss": 0.9059, "step": 8566 }, { "epoch": 0.05, "learning_rate": 1.6776779847454577e-05, "loss": 0.9678, "step": 8567 }, { "epoch": 0.05, "learning_rate": 1.6775987178647347e-05, "loss": 0.9239, "step": 8568 }, { "epoch": 0.05, "learning_rate": 1.6775194431114808e-05, "loss": 0.9483, "step": 8569 }, { "epoch": 0.05, "learning_rate": 1.6774401604866163e-05, "loss": 0.9334, "step": 8570 }, { "epoch": 0.05, "learning_rate": 1.6773608699910626e-05, "loss": 0.9436, "step": 8571 }, { "epoch": 0.05, "learning_rate": 1.6772815716257414e-05, "loss": 0.9121, "step": 8572 }, { "epoch": 0.05, "learning_rate": 1.677202265391573e-05, "loss": 0.9236, "step": 8573 }, { "epoch": 0.05, "learning_rate": 1.6771229512894796e-05, "loss": 0.9451, "step": 8574 }, { "epoch": 0.05, "learning_rate": 1.677043629320382e-05, "loss": 0.9188, "step": 8575 }, { "epoch": 0.05, "learning_rate": 1.6769642994852026e-05, "loss": 0.9274, "step": 8576 }, { "epoch": 0.05, "learning_rate": 1.676884961784863e-05, "loss": 0.9259, "step": 8577 }, { "epoch": 0.05, "learning_rate": 1.6768056162202838e-05, "loss": 0.9229, "step": 8578 }, { "epoch": 0.05, "learning_rate": 1.6767262627923885e-05, "loss": 0.9454, "step": 8579 }, { "epoch": 0.05, "learning_rate": 1.676646901502098e-05, "loss": 0.9455, "step": 8580 }, { "epoch": 0.05, "learning_rate": 1.6765675323503346e-05, "loss": 0.9164, "step": 8581 }, { "epoch": 0.05, "learning_rate": 1.67648815533802e-05, "loss": 0.9263, "step": 8582 }, { "epoch": 0.05, "learning_rate": 1.6764087704660775e-05, "loss": 0.9111, "step": 8583 }, { "epoch": 0.05, "learning_rate": 1.6763293777354286e-05, "loss": 0.9141, "step": 8584 }, { "epoch": 0.05, "learning_rate": 1.676249977146996e-05, "loss": 0.9312, "step": 8585 }, { "epoch": 0.05, "learning_rate": 1.6761705687017015e-05, "loss": 0.9172, "step": 8586 }, { "epoch": 0.05, "learning_rate": 1.676091152400469e-05, "loss": 0.9024, "step": 8587 }, { "epoch": 0.05, "learning_rate": 1.67601172824422e-05, "loss": 0.9376, "step": 8588 }, { "epoch": 0.05, "learning_rate": 1.6759322962338777e-05, "loss": 0.9791, "step": 8589 }, { "epoch": 0.05, "learning_rate": 1.6758528563703654e-05, "loss": 0.9473, "step": 8590 }, { "epoch": 0.05, "learning_rate": 1.6757734086546052e-05, "loss": 0.9694, "step": 8591 }, { "epoch": 0.05, "learning_rate": 1.6756939530875208e-05, "loss": 0.9101, "step": 8592 }, { "epoch": 0.05, "learning_rate": 1.675614489670035e-05, "loss": 0.9483, "step": 8593 }, { "epoch": 0.05, "learning_rate": 1.6755350184030714e-05, "loss": 0.9084, "step": 8594 }, { "epoch": 0.05, "learning_rate": 1.675455539287553e-05, "loss": 0.9372, "step": 8595 }, { "epoch": 0.05, "learning_rate": 1.675376052324403e-05, "loss": 0.9395, "step": 8596 }, { "epoch": 0.05, "learning_rate": 1.6752965575145456e-05, "loss": 0.9019, "step": 8597 }, { "epoch": 0.05, "learning_rate": 1.6752170548589037e-05, "loss": 0.9354, "step": 8598 }, { "epoch": 0.05, "learning_rate": 1.6751375443584014e-05, "loss": 0.9288, "step": 8599 }, { "epoch": 0.05, "learning_rate": 1.6750580260139628e-05, "loss": 0.9333, "step": 8600 }, { "epoch": 0.05, "learning_rate": 1.6749784998265103e-05, "loss": 0.9578, "step": 8601 }, { "epoch": 0.05, "learning_rate": 1.6748989657969696e-05, "loss": 0.914, "step": 8602 }, { "epoch": 0.05, "learning_rate": 1.6748194239262637e-05, "loss": 0.9348, "step": 8603 }, { "epoch": 0.05, "learning_rate": 1.6747398742153175e-05, "loss": 0.9537, "step": 8604 }, { "epoch": 0.05, "learning_rate": 1.6746603166650542e-05, "loss": 0.9803, "step": 8605 }, { "epoch": 0.05, "learning_rate": 1.674580751276399e-05, "loss": 0.9401, "step": 8606 }, { "epoch": 0.05, "learning_rate": 1.6745011780502758e-05, "loss": 0.9289, "step": 8607 }, { "epoch": 0.05, "learning_rate": 1.6744215969876095e-05, "loss": 0.9198, "step": 8608 }, { "epoch": 0.05, "learning_rate": 1.6743420080893247e-05, "loss": 0.9314, "step": 8609 }, { "epoch": 0.05, "learning_rate": 1.6742624113563456e-05, "loss": 0.932, "step": 8610 }, { "epoch": 0.05, "learning_rate": 1.6741828067895972e-05, "loss": 0.918, "step": 8611 }, { "epoch": 0.05, "learning_rate": 1.6741031943900048e-05, "loss": 0.9171, "step": 8612 }, { "epoch": 0.05, "learning_rate": 1.674023574158493e-05, "loss": 0.9735, "step": 8613 }, { "epoch": 0.05, "learning_rate": 1.6739439460959864e-05, "loss": 0.9543, "step": 8614 }, { "epoch": 0.05, "learning_rate": 1.673864310203411e-05, "loss": 0.9415, "step": 8615 }, { "epoch": 0.05, "learning_rate": 1.6737846664816916e-05, "loss": 0.9214, "step": 8616 }, { "epoch": 0.05, "learning_rate": 1.6737050149317532e-05, "loss": 0.9633, "step": 8617 }, { "epoch": 0.05, "learning_rate": 1.673625355554522e-05, "loss": 0.9243, "step": 8618 }, { "epoch": 0.05, "learning_rate": 1.6735456883509227e-05, "loss": 0.9349, "step": 8619 }, { "epoch": 0.05, "learning_rate": 1.673466013321882e-05, "loss": 0.9609, "step": 8620 }, { "epoch": 0.05, "learning_rate": 1.673386330468324e-05, "loss": 0.9359, "step": 8621 }, { "epoch": 0.05, "learning_rate": 1.6733066397911756e-05, "loss": 0.9376, "step": 8622 }, { "epoch": 0.05, "learning_rate": 1.6732269412913628e-05, "loss": 0.9354, "step": 8623 }, { "epoch": 0.05, "learning_rate": 1.6731472349698105e-05, "loss": 0.9458, "step": 8624 }, { "epoch": 0.05, "learning_rate": 1.673067520827446e-05, "loss": 0.9394, "step": 8625 }, { "epoch": 0.05, "learning_rate": 1.6729877988651944e-05, "loss": 0.8834, "step": 8626 }, { "epoch": 0.05, "learning_rate": 1.672908069083983e-05, "loss": 0.9883, "step": 8627 }, { "epoch": 0.05, "learning_rate": 1.6728283314847368e-05, "loss": 0.9544, "step": 8628 }, { "epoch": 0.05, "learning_rate": 1.6727485860683833e-05, "loss": 0.9491, "step": 8629 }, { "epoch": 0.05, "learning_rate": 1.6726688328358486e-05, "loss": 0.945, "step": 8630 }, { "epoch": 0.05, "learning_rate": 1.672589071788059e-05, "loss": 0.9123, "step": 8631 }, { "epoch": 0.05, "learning_rate": 1.672509302925942e-05, "loss": 0.9085, "step": 8632 }, { "epoch": 0.05, "learning_rate": 1.6724295262504236e-05, "loss": 0.9169, "step": 8633 }, { "epoch": 0.05, "learning_rate": 1.6723497417624308e-05, "loss": 0.9414, "step": 8634 }, { "epoch": 0.05, "learning_rate": 1.672269949462891e-05, "loss": 0.8983, "step": 8635 }, { "epoch": 0.05, "learning_rate": 1.6721901493527307e-05, "loss": 0.9564, "step": 8636 }, { "epoch": 0.05, "learning_rate": 1.6721103414328778e-05, "loss": 0.9381, "step": 8637 }, { "epoch": 0.05, "learning_rate": 1.6720305257042583e-05, "loss": 0.9376, "step": 8638 }, { "epoch": 0.05, "learning_rate": 1.671950702167801e-05, "loss": 0.9492, "step": 8639 }, { "epoch": 0.05, "learning_rate": 1.671870870824432e-05, "loss": 0.939, "step": 8640 }, { "epoch": 0.05, "learning_rate": 1.67179103167508e-05, "loss": 0.9453, "step": 8641 }, { "epoch": 0.05, "learning_rate": 1.6717111847206714e-05, "loss": 0.9231, "step": 8642 }, { "epoch": 0.05, "learning_rate": 1.6716313299621344e-05, "loss": 0.967, "step": 8643 }, { "epoch": 0.05, "learning_rate": 1.6715514674003972e-05, "loss": 0.9239, "step": 8644 }, { "epoch": 0.05, "learning_rate": 1.671471597036387e-05, "loss": 0.9105, "step": 8645 }, { "epoch": 0.05, "learning_rate": 1.6713917188710324e-05, "loss": 0.9103, "step": 8646 }, { "epoch": 0.05, "learning_rate": 1.6713118329052605e-05, "loss": 0.9604, "step": 8647 }, { "epoch": 0.05, "learning_rate": 1.6712319391400004e-05, "loss": 0.9194, "step": 8648 }, { "epoch": 0.05, "learning_rate": 1.67115203757618e-05, "loss": 0.9334, "step": 8649 }, { "epoch": 0.05, "learning_rate": 1.6710721282147274e-05, "loss": 0.9266, "step": 8650 }, { "epoch": 0.05, "learning_rate": 1.6709922110565712e-05, "loss": 0.9404, "step": 8651 }, { "epoch": 0.05, "learning_rate": 1.6709122861026398e-05, "loss": 0.9235, "step": 8652 }, { "epoch": 0.06, "learning_rate": 1.6708323533538623e-05, "loss": 0.9008, "step": 8653 }, { "epoch": 0.06, "learning_rate": 1.6707524128111664e-05, "loss": 0.878, "step": 8654 }, { "epoch": 0.06, "learning_rate": 1.6706724644754818e-05, "loss": 0.927, "step": 8655 }, { "epoch": 0.06, "learning_rate": 1.670592508347737e-05, "loss": 0.9174, "step": 8656 }, { "epoch": 0.06, "learning_rate": 1.6705125444288607e-05, "loss": 0.928, "step": 8657 }, { "epoch": 0.06, "learning_rate": 1.6704325727197823e-05, "loss": 0.9236, "step": 8658 }, { "epoch": 0.06, "learning_rate": 1.670352593221431e-05, "loss": 0.8797, "step": 8659 }, { "epoch": 0.06, "learning_rate": 1.6702726059347356e-05, "loss": 0.9263, "step": 8660 }, { "epoch": 0.06, "learning_rate": 1.6701926108606257e-05, "loss": 0.9019, "step": 8661 }, { "epoch": 0.06, "learning_rate": 1.6701126080000306e-05, "loss": 0.9229, "step": 8662 }, { "epoch": 0.06, "learning_rate": 1.6700325973538797e-05, "loss": 0.908, "step": 8663 }, { "epoch": 0.06, "learning_rate": 1.6699525789231033e-05, "loss": 0.8879, "step": 8664 }, { "epoch": 0.06, "learning_rate": 1.66987255270863e-05, "loss": 0.9513, "step": 8665 }, { "epoch": 0.06, "learning_rate": 1.6697925187113905e-05, "loss": 0.9257, "step": 8666 }, { "epoch": 0.06, "learning_rate": 1.6697124769323138e-05, "loss": 0.9206, "step": 8667 }, { "epoch": 0.06, "learning_rate": 1.6696324273723305e-05, "loss": 0.93, "step": 8668 }, { "epoch": 0.06, "learning_rate": 1.6695523700323706e-05, "loss": 0.9137, "step": 8669 }, { "epoch": 0.06, "learning_rate": 1.6694723049133638e-05, "loss": 0.9132, "step": 8670 }, { "epoch": 0.06, "learning_rate": 1.6693922320162408e-05, "loss": 0.9288, "step": 8671 }, { "epoch": 0.06, "learning_rate": 1.6693121513419316e-05, "loss": 0.9172, "step": 8672 }, { "epoch": 0.06, "learning_rate": 1.6692320628913667e-05, "loss": 0.8985, "step": 8673 }, { "epoch": 0.06, "learning_rate": 1.6691519666654763e-05, "loss": 0.9316, "step": 8674 }, { "epoch": 0.06, "learning_rate": 1.6690718626651917e-05, "loss": 1.0028, "step": 8675 }, { "epoch": 0.06, "learning_rate": 1.6689917508914428e-05, "loss": 0.9203, "step": 8676 }, { "epoch": 0.06, "learning_rate": 1.6689116313451606e-05, "loss": 0.9684, "step": 8677 }, { "epoch": 0.06, "learning_rate": 1.6688315040272766e-05, "loss": 0.9223, "step": 8678 }, { "epoch": 0.06, "learning_rate": 1.6687513689387205e-05, "loss": 0.9648, "step": 8679 }, { "epoch": 0.06, "learning_rate": 1.6686712260804247e-05, "loss": 0.9154, "step": 8680 }, { "epoch": 0.06, "learning_rate": 1.668591075453319e-05, "loss": 0.9407, "step": 8681 }, { "epoch": 0.06, "learning_rate": 1.6685109170583357e-05, "loss": 0.9335, "step": 8682 }, { "epoch": 0.06, "learning_rate": 1.668430750896405e-05, "loss": 0.9593, "step": 8683 }, { "epoch": 0.06, "learning_rate": 1.6683505769684596e-05, "loss": 0.9282, "step": 8684 }, { "epoch": 0.06, "learning_rate": 1.66827039527543e-05, "loss": 0.9776, "step": 8685 }, { "epoch": 0.06, "learning_rate": 1.6681902058182486e-05, "loss": 0.9316, "step": 8686 }, { "epoch": 0.06, "learning_rate": 1.6681100085978457e-05, "loss": 0.9719, "step": 8687 }, { "epoch": 0.06, "learning_rate": 1.6680298036151544e-05, "loss": 0.9199, "step": 8688 }, { "epoch": 0.06, "learning_rate": 1.6679495908711062e-05, "loss": 0.9244, "step": 8689 }, { "epoch": 0.06, "learning_rate": 1.6678693703666327e-05, "loss": 0.9523, "step": 8690 }, { "epoch": 0.06, "learning_rate": 1.667789142102666e-05, "loss": 0.902, "step": 8691 }, { "epoch": 0.06, "learning_rate": 1.6677089060801387e-05, "loss": 0.9574, "step": 8692 }, { "epoch": 0.06, "learning_rate": 1.6676286622999824e-05, "loss": 0.9043, "step": 8693 }, { "epoch": 0.06, "learning_rate": 1.66754841076313e-05, "loss": 0.9383, "step": 8694 }, { "epoch": 0.06, "learning_rate": 1.667468151470513e-05, "loss": 0.9453, "step": 8695 }, { "epoch": 0.06, "learning_rate": 1.6673878844230643e-05, "loss": 0.9284, "step": 8696 }, { "epoch": 0.06, "learning_rate": 1.667307609621717e-05, "loss": 0.9492, "step": 8697 }, { "epoch": 0.06, "learning_rate": 1.667227327067403e-05, "loss": 0.9081, "step": 8698 }, { "epoch": 0.06, "learning_rate": 1.6671470367610557e-05, "loss": 0.9313, "step": 8699 }, { "epoch": 0.06, "learning_rate": 1.6670667387036075e-05, "loss": 0.9515, "step": 8700 }, { "epoch": 0.06, "learning_rate": 1.666986432895991e-05, "loss": 0.9638, "step": 8701 }, { "epoch": 0.06, "learning_rate": 1.66690611933914e-05, "loss": 0.9114, "step": 8702 }, { "epoch": 0.06, "learning_rate": 1.6668257980339873e-05, "loss": 0.9184, "step": 8703 }, { "epoch": 0.06, "learning_rate": 1.666745468981466e-05, "loss": 0.9311, "step": 8704 }, { "epoch": 0.06, "learning_rate": 1.6666651321825093e-05, "loss": 0.9306, "step": 8705 }, { "epoch": 0.06, "learning_rate": 1.6665847876380505e-05, "loss": 0.9409, "step": 8706 }, { "epoch": 0.06, "learning_rate": 1.6665044353490235e-05, "loss": 0.9637, "step": 8707 }, { "epoch": 0.06, "learning_rate": 1.666424075316362e-05, "loss": 0.9302, "step": 8708 }, { "epoch": 0.06, "learning_rate": 1.6663437075409986e-05, "loss": 0.8982, "step": 8709 }, { "epoch": 0.06, "learning_rate": 1.666263332023868e-05, "loss": 0.8918, "step": 8710 }, { "epoch": 0.06, "learning_rate": 1.6661829487659033e-05, "loss": 0.9312, "step": 8711 }, { "epoch": 0.06, "learning_rate": 1.6661025577680394e-05, "loss": 0.9404, "step": 8712 }, { "epoch": 0.06, "learning_rate": 1.6660221590312095e-05, "loss": 0.9227, "step": 8713 }, { "epoch": 0.06, "learning_rate": 1.665941752556348e-05, "loss": 0.9574, "step": 8714 }, { "epoch": 0.06, "learning_rate": 1.6658613383443893e-05, "loss": 1.0024, "step": 8715 }, { "epoch": 0.06, "learning_rate": 1.6657809163962668e-05, "loss": 0.9274, "step": 8716 }, { "epoch": 0.06, "learning_rate": 1.6657004867129157e-05, "loss": 0.9665, "step": 8717 }, { "epoch": 0.06, "learning_rate": 1.6656200492952702e-05, "loss": 0.9024, "step": 8718 }, { "epoch": 0.06, "learning_rate": 1.6655396041442655e-05, "loss": 0.9478, "step": 8719 }, { "epoch": 0.06, "learning_rate": 1.6654591512608344e-05, "loss": 0.9246, "step": 8720 }, { "epoch": 0.06, "learning_rate": 1.6653786906459137e-05, "loss": 0.9139, "step": 8721 }, { "epoch": 0.06, "learning_rate": 1.6652982223004373e-05, "loss": 0.9639, "step": 8722 }, { "epoch": 0.06, "learning_rate": 1.6652177462253398e-05, "loss": 0.9412, "step": 8723 }, { "epoch": 0.06, "learning_rate": 1.665137262421557e-05, "loss": 0.9184, "step": 8724 }, { "epoch": 0.06, "learning_rate": 1.6650567708900226e-05, "loss": 0.9278, "step": 8725 }, { "epoch": 0.06, "learning_rate": 1.6649762716316733e-05, "loss": 0.8867, "step": 8726 }, { "epoch": 0.06, "learning_rate": 1.6648957646474436e-05, "loss": 0.9379, "step": 8727 }, { "epoch": 0.06, "learning_rate": 1.6648152499382694e-05, "loss": 0.9441, "step": 8728 }, { "epoch": 0.06, "learning_rate": 1.664734727505085e-05, "loss": 0.8977, "step": 8729 }, { "epoch": 0.06, "learning_rate": 1.6646541973488274e-05, "loss": 0.9167, "step": 8730 }, { "epoch": 0.06, "learning_rate": 1.6645736594704313e-05, "loss": 0.9279, "step": 8731 }, { "epoch": 0.06, "learning_rate": 1.6644931138708322e-05, "loss": 0.9401, "step": 8732 }, { "epoch": 0.06, "learning_rate": 1.6644125605509668e-05, "loss": 0.9329, "step": 8733 }, { "epoch": 0.06, "learning_rate": 1.66433199951177e-05, "loss": 0.9472, "step": 8734 }, { "epoch": 0.06, "learning_rate": 1.6642514307541787e-05, "loss": 0.9374, "step": 8735 }, { "epoch": 0.06, "learning_rate": 1.6641708542791285e-05, "loss": 0.9752, "step": 8736 }, { "epoch": 0.06, "learning_rate": 1.6640902700875556e-05, "loss": 0.9163, "step": 8737 }, { "epoch": 0.06, "learning_rate": 1.6640096781803966e-05, "loss": 0.8979, "step": 8738 }, { "epoch": 0.06, "learning_rate": 1.6639290785585868e-05, "loss": 0.9345, "step": 8739 }, { "epoch": 0.06, "learning_rate": 1.6638484712230635e-05, "loss": 0.9211, "step": 8740 }, { "epoch": 0.06, "learning_rate": 1.663767856174763e-05, "loss": 0.9465, "step": 8741 }, { "epoch": 0.06, "learning_rate": 1.6636872334146227e-05, "loss": 0.9268, "step": 8742 }, { "epoch": 0.06, "learning_rate": 1.663606602943578e-05, "loss": 0.9369, "step": 8743 }, { "epoch": 0.06, "learning_rate": 1.663525964762566e-05, "loss": 0.8734, "step": 8744 }, { "epoch": 0.06, "learning_rate": 1.6634453188725242e-05, "loss": 0.9494, "step": 8745 }, { "epoch": 0.06, "learning_rate": 1.663364665274389e-05, "loss": 0.9257, "step": 8746 }, { "epoch": 0.06, "learning_rate": 1.663284003969098e-05, "loss": 0.9557, "step": 8747 }, { "epoch": 0.06, "learning_rate": 1.6632033349575874e-05, "loss": 0.9471, "step": 8748 }, { "epoch": 0.06, "learning_rate": 1.6631226582407954e-05, "loss": 0.9377, "step": 8749 }, { "epoch": 0.06, "learning_rate": 1.6630419738196587e-05, "loss": 0.8894, "step": 8750 }, { "epoch": 0.06, "learning_rate": 1.6629612816951152e-05, "loss": 0.9245, "step": 8751 }, { "epoch": 0.06, "learning_rate": 1.6628805818681023e-05, "loss": 0.9672, "step": 8752 }, { "epoch": 0.06, "learning_rate": 1.662799874339557e-05, "loss": 0.948, "step": 8753 }, { "epoch": 0.06, "learning_rate": 1.662719159110418e-05, "loss": 0.919, "step": 8754 }, { "epoch": 0.06, "learning_rate": 1.662638436181622e-05, "loss": 0.9202, "step": 8755 }, { "epoch": 0.06, "learning_rate": 1.6625577055541074e-05, "loss": 0.9188, "step": 8756 }, { "epoch": 0.06, "learning_rate": 1.6624769672288125e-05, "loss": 0.9498, "step": 8757 }, { "epoch": 0.06, "learning_rate": 1.6623962212066748e-05, "loss": 0.9093, "step": 8758 }, { "epoch": 0.06, "learning_rate": 1.6623154674886323e-05, "loss": 0.9023, "step": 8759 }, { "epoch": 0.06, "learning_rate": 1.6622347060756237e-05, "loss": 0.9172, "step": 8760 }, { "epoch": 0.06, "learning_rate": 1.6621539369685873e-05, "loss": 0.939, "step": 8761 }, { "epoch": 0.06, "learning_rate": 1.6620731601684612e-05, "loss": 0.9261, "step": 8762 }, { "epoch": 0.06, "learning_rate": 1.6619923756761844e-05, "loss": 0.9148, "step": 8763 }, { "epoch": 0.06, "learning_rate": 1.661911583492694e-05, "loss": 0.9147, "step": 8764 }, { "epoch": 0.06, "learning_rate": 1.6618307836189308e-05, "loss": 0.9014, "step": 8765 }, { "epoch": 0.06, "learning_rate": 1.6617499760558322e-05, "loss": 0.9089, "step": 8766 }, { "epoch": 0.06, "learning_rate": 1.6616691608043375e-05, "loss": 0.943, "step": 8767 }, { "epoch": 0.06, "learning_rate": 1.6615883378653853e-05, "loss": 0.9577, "step": 8768 }, { "epoch": 0.06, "learning_rate": 1.6615075072399146e-05, "loss": 0.9723, "step": 8769 }, { "epoch": 0.06, "learning_rate": 1.661426668928865e-05, "loss": 0.9115, "step": 8770 }, { "epoch": 0.06, "learning_rate": 1.6613458229331753e-05, "loss": 0.9293, "step": 8771 }, { "epoch": 0.06, "learning_rate": 1.661264969253785e-05, "loss": 0.9626, "step": 8772 }, { "epoch": 0.06, "learning_rate": 1.6611841078916333e-05, "loss": 0.911, "step": 8773 }, { "epoch": 0.06, "learning_rate": 1.66110323884766e-05, "loss": 0.909, "step": 8774 }, { "epoch": 0.06, "learning_rate": 1.6610223621228043e-05, "loss": 0.9248, "step": 8775 }, { "epoch": 0.06, "learning_rate": 1.660941477718006e-05, "loss": 0.9285, "step": 8776 }, { "epoch": 0.06, "learning_rate": 1.660860585634205e-05, "loss": 0.9412, "step": 8777 }, { "epoch": 0.06, "learning_rate": 1.66077968587234e-05, "loss": 0.9414, "step": 8778 }, { "epoch": 0.06, "learning_rate": 1.660698778433353e-05, "loss": 0.9314, "step": 8779 }, { "epoch": 0.06, "learning_rate": 1.660617863318182e-05, "loss": 0.9571, "step": 8780 }, { "epoch": 0.06, "learning_rate": 1.6605369405277686e-05, "loss": 0.922, "step": 8781 }, { "epoch": 0.06, "learning_rate": 1.6604560100630524e-05, "loss": 0.9038, "step": 8782 }, { "epoch": 0.06, "learning_rate": 1.6603750719249728e-05, "loss": 0.9626, "step": 8783 }, { "epoch": 0.06, "learning_rate": 1.660294126114472e-05, "loss": 0.9285, "step": 8784 }, { "epoch": 0.06, "learning_rate": 1.6602131726324887e-05, "loss": 0.9069, "step": 8785 }, { "epoch": 0.06, "learning_rate": 1.6601322114799642e-05, "loss": 0.9323, "step": 8786 }, { "epoch": 0.06, "learning_rate": 1.6600512426578394e-05, "loss": 0.9682, "step": 8787 }, { "epoch": 0.06, "learning_rate": 1.6599702661670546e-05, "loss": 0.9116, "step": 8788 }, { "epoch": 0.06, "learning_rate": 1.6598892820085505e-05, "loss": 0.9234, "step": 8789 }, { "epoch": 0.06, "learning_rate": 1.6598082901832686e-05, "loss": 0.9324, "step": 8790 }, { "epoch": 0.06, "learning_rate": 1.6597272906921495e-05, "loss": 0.9555, "step": 8791 }, { "epoch": 0.06, "learning_rate": 1.6596462835361337e-05, "loss": 0.9021, "step": 8792 }, { "epoch": 0.06, "learning_rate": 1.6595652687161634e-05, "loss": 0.9144, "step": 8793 }, { "epoch": 0.06, "learning_rate": 1.6594842462331793e-05, "loss": 0.8864, "step": 8794 }, { "epoch": 0.06, "learning_rate": 1.659403216088123e-05, "loss": 0.9783, "step": 8795 }, { "epoch": 0.06, "learning_rate": 1.6593221782819356e-05, "loss": 0.921, "step": 8796 }, { "epoch": 0.06, "learning_rate": 1.6592411328155587e-05, "loss": 0.9001, "step": 8797 }, { "epoch": 0.06, "learning_rate": 1.659160079689934e-05, "loss": 0.8813, "step": 8798 }, { "epoch": 0.06, "learning_rate": 1.6590790189060034e-05, "loss": 0.9451, "step": 8799 }, { "epoch": 0.06, "learning_rate": 1.6589979504647086e-05, "loss": 0.9467, "step": 8800 }, { "epoch": 0.06, "learning_rate": 1.658916874366991e-05, "loss": 0.9226, "step": 8801 }, { "epoch": 0.06, "learning_rate": 1.658835790613793e-05, "loss": 0.9106, "step": 8802 }, { "epoch": 0.06, "learning_rate": 1.6587546992060566e-05, "loss": 0.9081, "step": 8803 }, { "epoch": 0.06, "learning_rate": 1.658673600144724e-05, "loss": 0.9264, "step": 8804 }, { "epoch": 0.06, "learning_rate": 1.6585924934307372e-05, "loss": 0.9344, "step": 8805 }, { "epoch": 0.06, "learning_rate": 1.658511379065039e-05, "loss": 0.9299, "step": 8806 }, { "epoch": 0.06, "learning_rate": 1.658430257048571e-05, "loss": 0.9355, "step": 8807 }, { "epoch": 0.06, "learning_rate": 1.6583491273822763e-05, "loss": 0.9421, "step": 8808 }, { "epoch": 0.06, "learning_rate": 1.6582679900670975e-05, "loss": 0.9272, "step": 8809 }, { "epoch": 0.06, "learning_rate": 1.6581868451039773e-05, "loss": 0.9305, "step": 8810 }, { "epoch": 0.06, "learning_rate": 1.6581056924938578e-05, "loss": 0.907, "step": 8811 }, { "epoch": 0.06, "learning_rate": 1.6580245322376827e-05, "loss": 0.9007, "step": 8812 }, { "epoch": 0.06, "learning_rate": 1.657943364336395e-05, "loss": 0.9494, "step": 8813 }, { "epoch": 0.06, "learning_rate": 1.6578621887909366e-05, "loss": 0.8863, "step": 8814 }, { "epoch": 0.06, "learning_rate": 1.6577810056022515e-05, "loss": 0.9438, "step": 8815 }, { "epoch": 0.06, "learning_rate": 1.657699814771283e-05, "loss": 0.9303, "step": 8816 }, { "epoch": 0.06, "learning_rate": 1.657618616298974e-05, "loss": 0.9657, "step": 8817 }, { "epoch": 0.06, "learning_rate": 1.657537410186268e-05, "loss": 0.9363, "step": 8818 }, { "epoch": 0.06, "learning_rate": 1.657456196434109e-05, "loss": 0.9086, "step": 8819 }, { "epoch": 0.06, "learning_rate": 1.6573749750434397e-05, "loss": 0.8888, "step": 8820 }, { "epoch": 0.06, "learning_rate": 1.657293746015204e-05, "loss": 0.9511, "step": 8821 }, { "epoch": 0.06, "learning_rate": 1.657212509350346e-05, "loss": 0.9451, "step": 8822 }, { "epoch": 0.06, "learning_rate": 1.6571312650498096e-05, "loss": 0.9404, "step": 8823 }, { "epoch": 0.06, "learning_rate": 1.6570500131145376e-05, "loss": 0.9444, "step": 8824 }, { "epoch": 0.06, "learning_rate": 1.6569687535454753e-05, "loss": 0.9213, "step": 8825 }, { "epoch": 0.06, "learning_rate": 1.6568874863435664e-05, "loss": 0.9267, "step": 8826 }, { "epoch": 0.06, "learning_rate": 1.656806211509755e-05, "loss": 0.979, "step": 8827 }, { "epoch": 0.06, "learning_rate": 1.6567249290449854e-05, "loss": 0.9066, "step": 8828 }, { "epoch": 0.06, "learning_rate": 1.6566436389502022e-05, "loss": 0.9277, "step": 8829 }, { "epoch": 0.06, "learning_rate": 1.656562341226349e-05, "loss": 0.9135, "step": 8830 }, { "epoch": 0.06, "learning_rate": 1.6564810358743713e-05, "loss": 0.938, "step": 8831 }, { "epoch": 0.06, "learning_rate": 1.6563997228952133e-05, "loss": 0.8923, "step": 8832 }, { "epoch": 0.06, "learning_rate": 1.6563184022898198e-05, "loss": 0.939, "step": 8833 }, { "epoch": 0.06, "learning_rate": 1.6562370740591355e-05, "loss": 0.9174, "step": 8834 }, { "epoch": 0.06, "learning_rate": 1.6561557382041055e-05, "loss": 0.9382, "step": 8835 }, { "epoch": 0.06, "learning_rate": 1.656074394725675e-05, "loss": 0.9046, "step": 8836 }, { "epoch": 0.06, "learning_rate": 1.6559930436247883e-05, "loss": 0.8664, "step": 8837 }, { "epoch": 0.06, "learning_rate": 1.655911684902391e-05, "loss": 0.9334, "step": 8838 }, { "epoch": 0.06, "learning_rate": 1.6558303185594285e-05, "loss": 0.9369, "step": 8839 }, { "epoch": 0.06, "learning_rate": 1.6557489445968456e-05, "loss": 0.9375, "step": 8840 }, { "epoch": 0.06, "learning_rate": 1.655667563015589e-05, "loss": 0.9498, "step": 8841 }, { "epoch": 0.06, "learning_rate": 1.6555861738166027e-05, "loss": 0.8608, "step": 8842 }, { "epoch": 0.06, "learning_rate": 1.6555047770008327e-05, "loss": 0.9221, "step": 8843 }, { "epoch": 0.06, "learning_rate": 1.6554233725692254e-05, "loss": 0.9166, "step": 8844 }, { "epoch": 0.06, "learning_rate": 1.655341960522726e-05, "loss": 0.9393, "step": 8845 }, { "epoch": 0.06, "learning_rate": 1.6552605408622804e-05, "loss": 0.9167, "step": 8846 }, { "epoch": 0.06, "learning_rate": 1.6551791135888345e-05, "loss": 0.9575, "step": 8847 }, { "epoch": 0.06, "learning_rate": 1.6550976787033347e-05, "loss": 0.9404, "step": 8848 }, { "epoch": 0.06, "learning_rate": 1.6550162362067268e-05, "loss": 0.9642, "step": 8849 }, { "epoch": 0.06, "learning_rate": 1.654934786099957e-05, "loss": 0.9083, "step": 8850 }, { "epoch": 0.06, "learning_rate": 1.6548533283839717e-05, "loss": 0.9408, "step": 8851 }, { "epoch": 0.06, "learning_rate": 1.6547718630597178e-05, "loss": 0.963, "step": 8852 }, { "epoch": 0.06, "learning_rate": 1.654690390128141e-05, "loss": 0.9126, "step": 8853 }, { "epoch": 0.06, "learning_rate": 1.6546089095901884e-05, "loss": 0.9179, "step": 8854 }, { "epoch": 0.06, "learning_rate": 1.6545274214468062e-05, "loss": 0.9049, "step": 8855 }, { "epoch": 0.06, "learning_rate": 1.6544459256989415e-05, "loss": 0.907, "step": 8856 }, { "epoch": 0.06, "learning_rate": 1.654364422347541e-05, "loss": 0.9067, "step": 8857 }, { "epoch": 0.06, "learning_rate": 1.6542829113935516e-05, "loss": 0.9545, "step": 8858 }, { "epoch": 0.06, "learning_rate": 1.6542013928379207e-05, "loss": 0.9271, "step": 8859 }, { "epoch": 0.06, "learning_rate": 1.654119866681595e-05, "loss": 0.924, "step": 8860 }, { "epoch": 0.06, "learning_rate": 1.6540383329255214e-05, "loss": 0.9137, "step": 8861 }, { "epoch": 0.06, "learning_rate": 1.6539567915706482e-05, "loss": 0.8955, "step": 8862 }, { "epoch": 0.06, "learning_rate": 1.6538752426179217e-05, "loss": 0.9357, "step": 8863 }, { "epoch": 0.06, "learning_rate": 1.6537936860682898e-05, "loss": 0.9076, "step": 8864 }, { "epoch": 0.06, "learning_rate": 1.6537121219227003e-05, "loss": 0.9252, "step": 8865 }, { "epoch": 0.06, "learning_rate": 1.6536305501821007e-05, "loss": 0.8906, "step": 8866 }, { "epoch": 0.06, "learning_rate": 1.653548970847438e-05, "loss": 0.9266, "step": 8867 }, { "epoch": 0.06, "learning_rate": 1.653467383919661e-05, "loss": 0.9087, "step": 8868 }, { "epoch": 0.06, "learning_rate": 1.6533857893997174e-05, "loss": 0.9269, "step": 8869 }, { "epoch": 0.06, "learning_rate": 1.653304187288555e-05, "loss": 0.9345, "step": 8870 }, { "epoch": 0.06, "learning_rate": 1.6532225775871215e-05, "loss": 0.9376, "step": 8871 }, { "epoch": 0.06, "learning_rate": 1.6531409602963656e-05, "loss": 0.9057, "step": 8872 }, { "epoch": 0.06, "learning_rate": 1.6530593354172358e-05, "loss": 0.9175, "step": 8873 }, { "epoch": 0.06, "learning_rate": 1.6529777029506795e-05, "loss": 0.9619, "step": 8874 }, { "epoch": 0.06, "learning_rate": 1.652896062897646e-05, "loss": 0.9792, "step": 8875 }, { "epoch": 0.06, "learning_rate": 1.6528144152590836e-05, "loss": 0.9202, "step": 8876 }, { "epoch": 0.06, "learning_rate": 1.6527327600359403e-05, "loss": 0.9133, "step": 8877 }, { "epoch": 0.06, "learning_rate": 1.6526510972291657e-05, "loss": 0.9274, "step": 8878 }, { "epoch": 0.06, "learning_rate": 1.652569426839708e-05, "loss": 0.8911, "step": 8879 }, { "epoch": 0.06, "learning_rate": 1.6524877488685165e-05, "loss": 0.9285, "step": 8880 }, { "epoch": 0.06, "learning_rate": 1.65240606331654e-05, "loss": 0.943, "step": 8881 }, { "epoch": 0.06, "learning_rate": 1.6523243701847267e-05, "loss": 0.8939, "step": 8882 }, { "epoch": 0.06, "learning_rate": 1.652242669474027e-05, "loss": 0.9809, "step": 8883 }, { "epoch": 0.06, "learning_rate": 1.6521609611853896e-05, "loss": 0.9031, "step": 8884 }, { "epoch": 0.06, "learning_rate": 1.6520792453197637e-05, "loss": 0.9353, "step": 8885 }, { "epoch": 0.06, "learning_rate": 1.651997521878099e-05, "loss": 0.9292, "step": 8886 }, { "epoch": 0.06, "learning_rate": 1.6519157908613444e-05, "loss": 0.9631, "step": 8887 }, { "epoch": 0.06, "learning_rate": 1.65183405227045e-05, "loss": 0.9636, "step": 8888 }, { "epoch": 0.06, "learning_rate": 1.6517523061063657e-05, "loss": 0.9285, "step": 8889 }, { "epoch": 0.06, "learning_rate": 1.65167055237004e-05, "loss": 0.9501, "step": 8890 }, { "epoch": 0.06, "learning_rate": 1.6515887910624247e-05, "loss": 0.939, "step": 8891 }, { "epoch": 0.06, "learning_rate": 1.6515070221844683e-05, "loss": 0.9176, "step": 8892 }, { "epoch": 0.06, "learning_rate": 1.6514252457371206e-05, "loss": 0.9708, "step": 8893 }, { "epoch": 0.06, "learning_rate": 1.651343461721333e-05, "loss": 0.9223, "step": 8894 }, { "epoch": 0.06, "learning_rate": 1.6512616701380545e-05, "loss": 0.9174, "step": 8895 }, { "epoch": 0.06, "learning_rate": 1.651179870988236e-05, "loss": 0.9331, "step": 8896 }, { "epoch": 0.06, "learning_rate": 1.6510980642728275e-05, "loss": 0.9277, "step": 8897 }, { "epoch": 0.06, "learning_rate": 1.6510162499927796e-05, "loss": 0.9358, "step": 8898 }, { "epoch": 0.06, "learning_rate": 1.6509344281490433e-05, "loss": 0.9237, "step": 8899 }, { "epoch": 0.06, "learning_rate": 1.6508525987425685e-05, "loss": 0.9342, "step": 8900 }, { "epoch": 0.06, "learning_rate": 1.6507707617743063e-05, "loss": 0.9458, "step": 8901 }, { "epoch": 0.06, "learning_rate": 1.6506889172452072e-05, "loss": 0.9331, "step": 8902 }, { "epoch": 0.06, "learning_rate": 1.6506070651562228e-05, "loss": 0.924, "step": 8903 }, { "epoch": 0.06, "learning_rate": 1.650525205508303e-05, "loss": 0.9401, "step": 8904 }, { "epoch": 0.06, "learning_rate": 1.6504433383024e-05, "loss": 0.9063, "step": 8905 }, { "epoch": 0.06, "learning_rate": 1.650361463539464e-05, "loss": 0.9271, "step": 8906 }, { "epoch": 0.06, "learning_rate": 1.6502795812204474e-05, "loss": 0.907, "step": 8907 }, { "epoch": 0.06, "learning_rate": 1.6501976913463004e-05, "loss": 0.906, "step": 8908 }, { "epoch": 0.06, "learning_rate": 1.6501157939179747e-05, "loss": 0.9092, "step": 8909 }, { "epoch": 0.06, "learning_rate": 1.6500338889364218e-05, "loss": 0.931, "step": 8910 }, { "epoch": 0.06, "learning_rate": 1.6499519764025936e-05, "loss": 0.931, "step": 8911 }, { "epoch": 0.06, "learning_rate": 1.6498700563174417e-05, "loss": 0.9162, "step": 8912 }, { "epoch": 0.06, "learning_rate": 1.6497881286819178e-05, "loss": 0.9656, "step": 8913 }, { "epoch": 0.06, "learning_rate": 1.6497061934969734e-05, "loss": 0.9661, "step": 8914 }, { "epoch": 0.06, "learning_rate": 1.649624250763561e-05, "loss": 0.9703, "step": 8915 }, { "epoch": 0.06, "learning_rate": 1.6495423004826327e-05, "loss": 0.9163, "step": 8916 }, { "epoch": 0.06, "learning_rate": 1.6494603426551397e-05, "loss": 0.9139, "step": 8917 }, { "epoch": 0.06, "learning_rate": 1.6493783772820357e-05, "loss": 0.8948, "step": 8918 }, { "epoch": 0.06, "learning_rate": 1.6492964043642716e-05, "loss": 0.9673, "step": 8919 }, { "epoch": 0.06, "learning_rate": 1.6492144239028003e-05, "loss": 0.8951, "step": 8920 }, { "epoch": 0.06, "learning_rate": 1.6491324358985744e-05, "loss": 0.9452, "step": 8921 }, { "epoch": 0.06, "learning_rate": 1.6490504403525464e-05, "loss": 0.9441, "step": 8922 }, { "epoch": 0.06, "learning_rate": 1.648968437265669e-05, "loss": 0.9654, "step": 8923 }, { "epoch": 0.06, "learning_rate": 1.6488864266388947e-05, "loss": 0.8756, "step": 8924 }, { "epoch": 0.06, "learning_rate": 1.6488044084731764e-05, "loss": 0.8958, "step": 8925 }, { "epoch": 0.06, "learning_rate": 1.6487223827694673e-05, "loss": 0.9044, "step": 8926 }, { "epoch": 0.06, "learning_rate": 1.6486403495287198e-05, "loss": 0.9598, "step": 8927 }, { "epoch": 0.06, "learning_rate": 1.6485583087518877e-05, "loss": 0.9071, "step": 8928 }, { "epoch": 0.06, "learning_rate": 1.648476260439924e-05, "loss": 0.9246, "step": 8929 }, { "epoch": 0.06, "learning_rate": 1.6483942045937813e-05, "loss": 0.9315, "step": 8930 }, { "epoch": 0.06, "learning_rate": 1.6483121412144137e-05, "loss": 0.9314, "step": 8931 }, { "epoch": 0.06, "learning_rate": 1.6482300703027744e-05, "loss": 0.9295, "step": 8932 }, { "epoch": 0.06, "learning_rate": 1.648147991859817e-05, "loss": 0.957, "step": 8933 }, { "epoch": 0.06, "learning_rate": 1.648065905886495e-05, "loss": 0.903, "step": 8934 }, { "epoch": 0.06, "learning_rate": 1.6479838123837618e-05, "loss": 0.9505, "step": 8935 }, { "epoch": 0.06, "learning_rate": 1.6479017113525714e-05, "loss": 0.9278, "step": 8936 }, { "epoch": 0.06, "learning_rate": 1.6478196027938782e-05, "loss": 0.972, "step": 8937 }, { "epoch": 0.06, "learning_rate": 1.6477374867086355e-05, "loss": 0.9671, "step": 8938 }, { "epoch": 0.06, "learning_rate": 1.6476553630977975e-05, "loss": 0.9884, "step": 8939 }, { "epoch": 0.06, "learning_rate": 1.6475732319623186e-05, "loss": 0.9224, "step": 8940 }, { "epoch": 0.06, "learning_rate": 1.6474910933031526e-05, "loss": 0.9392, "step": 8941 }, { "epoch": 0.06, "learning_rate": 1.6474089471212545e-05, "loss": 0.922, "step": 8942 }, { "epoch": 0.06, "learning_rate": 1.6473267934175775e-05, "loss": 0.9743, "step": 8943 }, { "epoch": 0.06, "learning_rate": 1.6472446321930776e-05, "loss": 0.9439, "step": 8944 }, { "epoch": 0.06, "learning_rate": 1.6471624634487084e-05, "loss": 0.9423, "step": 8945 }, { "epoch": 0.06, "learning_rate": 1.6470802871854242e-05, "loss": 0.9361, "step": 8946 }, { "epoch": 0.06, "learning_rate": 1.6469981034041808e-05, "loss": 0.9228, "step": 8947 }, { "epoch": 0.06, "learning_rate": 1.6469159121059324e-05, "loss": 0.9275, "step": 8948 }, { "epoch": 0.06, "learning_rate": 1.646833713291634e-05, "loss": 0.93, "step": 8949 }, { "epoch": 0.06, "learning_rate": 1.6467515069622407e-05, "loss": 0.9411, "step": 8950 }, { "epoch": 0.06, "learning_rate": 1.6466692931187078e-05, "loss": 0.9575, "step": 8951 }, { "epoch": 0.06, "learning_rate": 1.6465870717619898e-05, "loss": 0.9153, "step": 8952 }, { "epoch": 0.06, "learning_rate": 1.6465048428930427e-05, "loss": 0.9246, "step": 8953 }, { "epoch": 0.07, "learning_rate": 1.6464226065128216e-05, "loss": 0.9097, "step": 8954 }, { "epoch": 0.07, "learning_rate": 1.646340362622282e-05, "loss": 0.8944, "step": 8955 }, { "epoch": 0.07, "learning_rate": 1.646258111222379e-05, "loss": 0.9107, "step": 8956 }, { "epoch": 0.07, "learning_rate": 1.6461758523140688e-05, "loss": 0.9282, "step": 8957 }, { "epoch": 0.07, "learning_rate": 1.6460935858983067e-05, "loss": 0.8974, "step": 8958 }, { "epoch": 0.07, "learning_rate": 1.6460113119760488e-05, "loss": 0.9349, "step": 8959 }, { "epoch": 0.07, "learning_rate": 1.645929030548251e-05, "loss": 0.9598, "step": 8960 }, { "epoch": 0.07, "learning_rate": 1.645846741615869e-05, "loss": 0.9137, "step": 8961 }, { "epoch": 0.07, "learning_rate": 1.645764445179859e-05, "loss": 0.8967, "step": 8962 }, { "epoch": 0.07, "learning_rate": 1.645682141241177e-05, "loss": 0.9319, "step": 8963 }, { "epoch": 0.07, "learning_rate": 1.6455998298007794e-05, "loss": 0.9143, "step": 8964 }, { "epoch": 0.07, "learning_rate": 1.6455175108596223e-05, "loss": 0.9442, "step": 8965 }, { "epoch": 0.07, "learning_rate": 1.6454351844186624e-05, "loss": 0.9239, "step": 8966 }, { "epoch": 0.07, "learning_rate": 1.6453528504788564e-05, "loss": 0.908, "step": 8967 }, { "epoch": 0.07, "learning_rate": 1.6452705090411602e-05, "loss": 0.9185, "step": 8968 }, { "epoch": 0.07, "learning_rate": 1.645188160106531e-05, "loss": 0.9272, "step": 8969 }, { "epoch": 0.07, "learning_rate": 1.6451058036759252e-05, "loss": 0.8911, "step": 8970 }, { "epoch": 0.07, "learning_rate": 1.6450234397502994e-05, "loss": 0.9259, "step": 8971 }, { "epoch": 0.07, "learning_rate": 1.6449410683306116e-05, "loss": 0.9291, "step": 8972 }, { "epoch": 0.07, "learning_rate": 1.6448586894178176e-05, "loss": 0.9303, "step": 8973 }, { "epoch": 0.07, "learning_rate": 1.6447763030128756e-05, "loss": 0.9246, "step": 8974 }, { "epoch": 0.07, "learning_rate": 1.6446939091167423e-05, "loss": 0.9394, "step": 8975 }, { "epoch": 0.07, "learning_rate": 1.6446115077303743e-05, "loss": 0.9224, "step": 8976 }, { "epoch": 0.07, "learning_rate": 1.64452909885473e-05, "loss": 0.9353, "step": 8977 }, { "epoch": 0.07, "learning_rate": 1.6444466824907662e-05, "loss": 0.9276, "step": 8978 }, { "epoch": 0.07, "learning_rate": 1.644364258639441e-05, "loss": 0.9142, "step": 8979 }, { "epoch": 0.07, "learning_rate": 1.644281827301711e-05, "loss": 0.9081, "step": 8980 }, { "epoch": 0.07, "learning_rate": 1.6441993884785352e-05, "loss": 0.9272, "step": 8981 }, { "epoch": 0.07, "learning_rate": 1.6441169421708705e-05, "loss": 0.922, "step": 8982 }, { "epoch": 0.07, "learning_rate": 1.644034488379675e-05, "loss": 0.9399, "step": 8983 }, { "epoch": 0.07, "learning_rate": 1.643952027105907e-05, "loss": 0.8893, "step": 8984 }, { "epoch": 0.07, "learning_rate": 1.643869558350524e-05, "loss": 0.9438, "step": 8985 }, { "epoch": 0.07, "learning_rate": 1.6437870821144847e-05, "loss": 0.9281, "step": 8986 }, { "epoch": 0.07, "learning_rate": 1.6437045983987474e-05, "loss": 0.905, "step": 8987 }, { "epoch": 0.07, "learning_rate": 1.6436221072042694e-05, "loss": 0.9434, "step": 8988 }, { "epoch": 0.07, "learning_rate": 1.64353960853201e-05, "loss": 0.9367, "step": 8989 }, { "epoch": 0.07, "learning_rate": 1.6434571023829277e-05, "loss": 0.8844, "step": 8990 }, { "epoch": 0.07, "learning_rate": 1.643374588757981e-05, "loss": 0.9255, "step": 8991 }, { "epoch": 0.07, "learning_rate": 1.643292067658128e-05, "loss": 0.9061, "step": 8992 }, { "epoch": 0.07, "learning_rate": 1.643209539084328e-05, "loss": 0.9301, "step": 8993 }, { "epoch": 0.07, "learning_rate": 1.64312700303754e-05, "loss": 0.8853, "step": 8994 }, { "epoch": 0.07, "learning_rate": 1.6430444595187224e-05, "loss": 0.9364, "step": 8995 }, { "epoch": 0.07, "learning_rate": 1.6429619085288345e-05, "loss": 0.89, "step": 8996 }, { "epoch": 0.07, "learning_rate": 1.6428793500688356e-05, "loss": 0.9444, "step": 8997 }, { "epoch": 0.07, "learning_rate": 1.6427967841396847e-05, "loss": 0.8848, "step": 8998 }, { "epoch": 0.07, "learning_rate": 1.6427142107423407e-05, "loss": 0.9412, "step": 8999 }, { "epoch": 0.07, "learning_rate": 1.6426316298777634e-05, "loss": 0.9066, "step": 9000 }, { "epoch": 0.07, "learning_rate": 1.6425490415469122e-05, "loss": 0.9245, "step": 9001 }, { "epoch": 0.07, "learning_rate": 1.6424664457507466e-05, "loss": 0.9169, "step": 9002 }, { "epoch": 0.07, "learning_rate": 1.6423838424902258e-05, "loss": 0.9461, "step": 9003 }, { "epoch": 0.07, "learning_rate": 1.6423012317663106e-05, "loss": 0.9362, "step": 9004 }, { "epoch": 0.07, "learning_rate": 1.6422186135799597e-05, "loss": 0.977, "step": 9005 }, { "epoch": 0.07, "learning_rate": 1.642135987932133e-05, "loss": 0.8621, "step": 9006 }, { "epoch": 0.07, "learning_rate": 1.6420533548237914e-05, "loss": 0.8971, "step": 9007 }, { "epoch": 0.07, "learning_rate": 1.6419707142558946e-05, "loss": 0.9104, "step": 9008 }, { "epoch": 0.07, "learning_rate": 1.6418880662294023e-05, "loss": 0.9271, "step": 9009 }, { "epoch": 0.07, "learning_rate": 1.641805410745275e-05, "loss": 0.9321, "step": 9010 }, { "epoch": 0.07, "learning_rate": 1.6417227478044728e-05, "loss": 0.9266, "step": 9011 }, { "epoch": 0.07, "learning_rate": 1.6416400774079562e-05, "loss": 0.8989, "step": 9012 }, { "epoch": 0.07, "learning_rate": 1.641557399556686e-05, "loss": 0.9397, "step": 9013 }, { "epoch": 0.07, "learning_rate": 1.6414747142516227e-05, "loss": 0.9185, "step": 9014 }, { "epoch": 0.07, "learning_rate": 1.6413920214937267e-05, "loss": 0.9186, "step": 9015 }, { "epoch": 0.07, "learning_rate": 1.6413093212839587e-05, "loss": 0.9581, "step": 9016 }, { "epoch": 0.07, "learning_rate": 1.6412266136232797e-05, "loss": 0.9646, "step": 9017 }, { "epoch": 0.07, "learning_rate": 1.6411438985126508e-05, "loss": 0.9374, "step": 9018 }, { "epoch": 0.07, "learning_rate": 1.641061175953033e-05, "loss": 0.9405, "step": 9019 }, { "epoch": 0.07, "learning_rate": 1.6409784459453865e-05, "loss": 0.9385, "step": 9020 }, { "epoch": 0.07, "learning_rate": 1.640895708490674e-05, "loss": 0.9535, "step": 9021 }, { "epoch": 0.07, "learning_rate": 1.6408129635898556e-05, "loss": 0.8862, "step": 9022 }, { "epoch": 0.07, "learning_rate": 1.640730211243893e-05, "loss": 0.965, "step": 9023 }, { "epoch": 0.07, "learning_rate": 1.6406474514537477e-05, "loss": 0.8913, "step": 9024 }, { "epoch": 0.07, "learning_rate": 1.640564684220381e-05, "loss": 0.9031, "step": 9025 }, { "epoch": 0.07, "learning_rate": 1.640481909544755e-05, "loss": 0.9174, "step": 9026 }, { "epoch": 0.07, "learning_rate": 1.6403991274278307e-05, "loss": 0.9557, "step": 9027 }, { "epoch": 0.07, "learning_rate": 1.6403163378705708e-05, "loss": 0.8976, "step": 9028 }, { "epoch": 0.07, "learning_rate": 1.6402335408739363e-05, "loss": 0.923, "step": 9029 }, { "epoch": 0.07, "learning_rate": 1.6401507364388896e-05, "loss": 0.9482, "step": 9030 }, { "epoch": 0.07, "learning_rate": 1.6400679245663925e-05, "loss": 0.9803, "step": 9031 }, { "epoch": 0.07, "learning_rate": 1.6399851052574076e-05, "loss": 0.9267, "step": 9032 }, { "epoch": 0.07, "learning_rate": 1.6399022785128968e-05, "loss": 0.9235, "step": 9033 }, { "epoch": 0.07, "learning_rate": 1.639819444333822e-05, "loss": 0.9241, "step": 9034 }, { "epoch": 0.07, "learning_rate": 1.6397366027211462e-05, "loss": 0.9224, "step": 9035 }, { "epoch": 0.07, "learning_rate": 1.639653753675832e-05, "loss": 0.9551, "step": 9036 }, { "epoch": 0.07, "learning_rate": 1.6395708971988415e-05, "loss": 0.9165, "step": 9037 }, { "epoch": 0.07, "learning_rate": 1.639488033291137e-05, "loss": 0.9438, "step": 9038 }, { "epoch": 0.07, "learning_rate": 1.639405161953682e-05, "loss": 0.9118, "step": 9039 }, { "epoch": 0.07, "learning_rate": 1.6393222831874392e-05, "loss": 0.9089, "step": 9040 }, { "epoch": 0.07, "learning_rate": 1.6392393969933714e-05, "loss": 0.9559, "step": 9041 }, { "epoch": 0.07, "learning_rate": 1.6391565033724415e-05, "loss": 0.9202, "step": 9042 }, { "epoch": 0.07, "learning_rate": 1.6390736023256127e-05, "loss": 0.9289, "step": 9043 }, { "epoch": 0.07, "learning_rate": 1.638990693853848e-05, "loss": 0.8605, "step": 9044 }, { "epoch": 0.07, "learning_rate": 1.6389077779581107e-05, "loss": 0.9252, "step": 9045 }, { "epoch": 0.07, "learning_rate": 1.6388248546393644e-05, "loss": 1.0048, "step": 9046 }, { "epoch": 0.07, "learning_rate": 1.638741923898572e-05, "loss": 0.9576, "step": 9047 }, { "epoch": 0.07, "learning_rate": 1.6386589857366976e-05, "loss": 0.9011, "step": 9048 }, { "epoch": 0.07, "learning_rate": 1.6385760401547047e-05, "loss": 0.9233, "step": 9049 }, { "epoch": 0.07, "learning_rate": 1.6384930871535566e-05, "loss": 0.9036, "step": 9050 }, { "epoch": 0.07, "learning_rate": 1.638410126734217e-05, "loss": 0.9491, "step": 9051 }, { "epoch": 0.07, "learning_rate": 1.6383271588976503e-05, "loss": 0.9073, "step": 9052 }, { "epoch": 0.07, "learning_rate": 1.6382441836448203e-05, "loss": 0.9056, "step": 9053 }, { "epoch": 0.07, "learning_rate": 1.638161200976691e-05, "loss": 0.9384, "step": 9054 }, { "epoch": 0.07, "learning_rate": 1.638078210894226e-05, "loss": 0.9317, "step": 9055 }, { "epoch": 0.07, "learning_rate": 1.6379952133983903e-05, "loss": 0.912, "step": 9056 }, { "epoch": 0.07, "learning_rate": 1.6379122084901474e-05, "loss": 0.9584, "step": 9057 }, { "epoch": 0.07, "learning_rate": 1.6378291961704628e-05, "loss": 0.9349, "step": 9058 }, { "epoch": 0.07, "learning_rate": 1.6377461764402996e-05, "loss": 0.9407, "step": 9059 }, { "epoch": 0.07, "learning_rate": 1.6376631493006233e-05, "loss": 0.9216, "step": 9060 }, { "epoch": 0.07, "learning_rate": 1.637580114752398e-05, "loss": 0.9032, "step": 9061 }, { "epoch": 0.07, "learning_rate": 1.637497072796589e-05, "loss": 0.907, "step": 9062 }, { "epoch": 0.07, "learning_rate": 1.6374140234341608e-05, "loss": 0.947, "step": 9063 }, { "epoch": 0.07, "learning_rate": 1.637330966666078e-05, "loss": 0.9172, "step": 9064 }, { "epoch": 0.07, "learning_rate": 1.637247902493306e-05, "loss": 0.965, "step": 9065 }, { "epoch": 0.07, "learning_rate": 1.63716483091681e-05, "loss": 0.8793, "step": 9066 }, { "epoch": 0.07, "learning_rate": 1.6370817519375544e-05, "loss": 0.914, "step": 9067 }, { "epoch": 0.07, "learning_rate": 1.636998665556505e-05, "loss": 0.911, "step": 9068 }, { "epoch": 0.07, "learning_rate": 1.636915571774627e-05, "loss": 0.9412, "step": 9069 }, { "epoch": 0.07, "learning_rate": 1.6368324705928852e-05, "loss": 0.9013, "step": 9070 }, { "epoch": 0.07, "learning_rate": 1.6367493620122466e-05, "loss": 0.9313, "step": 9071 }, { "epoch": 0.07, "learning_rate": 1.636666246033675e-05, "loss": 0.9269, "step": 9072 }, { "epoch": 0.07, "learning_rate": 1.6365831226581378e-05, "loss": 0.9596, "step": 9073 }, { "epoch": 0.07, "learning_rate": 1.6364999918865997e-05, "loss": 0.918, "step": 9074 }, { "epoch": 0.07, "learning_rate": 1.6364168537200266e-05, "loss": 0.9748, "step": 9075 }, { "epoch": 0.07, "learning_rate": 1.6363337081593848e-05, "loss": 0.9219, "step": 9076 }, { "epoch": 0.07, "learning_rate": 1.6362505552056395e-05, "loss": 0.9306, "step": 9077 }, { "epoch": 0.07, "learning_rate": 1.6361673948597577e-05, "loss": 0.9331, "step": 9078 }, { "epoch": 0.07, "learning_rate": 1.636084227122705e-05, "loss": 0.9161, "step": 9079 }, { "epoch": 0.07, "learning_rate": 1.6360010519954485e-05, "loss": 0.9486, "step": 9080 }, { "epoch": 0.07, "learning_rate": 1.635917869478953e-05, "loss": 0.9161, "step": 9081 }, { "epoch": 0.07, "learning_rate": 1.6358346795741868e-05, "loss": 0.908, "step": 9082 }, { "epoch": 0.07, "learning_rate": 1.6357514822821153e-05, "loss": 0.9637, "step": 9083 }, { "epoch": 0.07, "learning_rate": 1.635668277603705e-05, "loss": 0.8818, "step": 9084 }, { "epoch": 0.07, "learning_rate": 1.6355850655399228e-05, "loss": 0.9251, "step": 9085 }, { "epoch": 0.07, "learning_rate": 1.635501846091736e-05, "loss": 0.9439, "step": 9086 }, { "epoch": 0.07, "learning_rate": 1.6354186192601108e-05, "loss": 0.9324, "step": 9087 }, { "epoch": 0.07, "learning_rate": 1.6353353850460147e-05, "loss": 0.9261, "step": 9088 }, { "epoch": 0.07, "learning_rate": 1.635252143450414e-05, "loss": 0.9156, "step": 9089 }, { "epoch": 0.07, "learning_rate": 1.6351688944742767e-05, "loss": 0.9218, "step": 9090 }, { "epoch": 0.07, "learning_rate": 1.6350856381185696e-05, "loss": 0.9649, "step": 9091 }, { "epoch": 0.07, "learning_rate": 1.63500237438426e-05, "loss": 0.9242, "step": 9092 }, { "epoch": 0.07, "learning_rate": 1.6349191032723147e-05, "loss": 0.9354, "step": 9093 }, { "epoch": 0.07, "learning_rate": 1.6348358247837025e-05, "loss": 0.9184, "step": 9094 }, { "epoch": 0.07, "learning_rate": 1.63475253891939e-05, "loss": 0.9678, "step": 9095 }, { "epoch": 0.07, "learning_rate": 1.6346692456803447e-05, "loss": 0.8881, "step": 9096 }, { "epoch": 0.07, "learning_rate": 1.634585945067535e-05, "loss": 0.9141, "step": 9097 }, { "epoch": 0.07, "learning_rate": 1.6345026370819282e-05, "loss": 0.903, "step": 9098 }, { "epoch": 0.07, "learning_rate": 1.6344193217244923e-05, "loss": 0.9383, "step": 9099 }, { "epoch": 0.07, "learning_rate": 1.6343359989961952e-05, "loss": 0.9104, "step": 9100 }, { "epoch": 0.07, "learning_rate": 1.6342526688980054e-05, "loss": 0.9825, "step": 9101 }, { "epoch": 0.07, "learning_rate": 1.6341693314308906e-05, "loss": 0.8978, "step": 9102 }, { "epoch": 0.07, "learning_rate": 1.6340859865958193e-05, "loss": 0.9053, "step": 9103 }, { "epoch": 0.07, "learning_rate": 1.6340026343937597e-05, "loss": 0.9167, "step": 9104 }, { "epoch": 0.07, "learning_rate": 1.6339192748256802e-05, "loss": 0.9604, "step": 9105 }, { "epoch": 0.07, "learning_rate": 1.633835907892549e-05, "loss": 0.879, "step": 9106 }, { "epoch": 0.07, "learning_rate": 1.6337525335953354e-05, "loss": 0.9456, "step": 9107 }, { "epoch": 0.07, "learning_rate": 1.6336691519350077e-05, "loss": 0.8973, "step": 9108 }, { "epoch": 0.07, "learning_rate": 1.6335857629125342e-05, "loss": 0.9209, "step": 9109 }, { "epoch": 0.07, "learning_rate": 1.6335023665288847e-05, "loss": 0.9374, "step": 9110 }, { "epoch": 0.07, "learning_rate": 1.6334189627850274e-05, "loss": 0.8982, "step": 9111 }, { "epoch": 0.07, "learning_rate": 1.633335551681931e-05, "loss": 0.9109, "step": 9112 }, { "epoch": 0.07, "learning_rate": 1.6332521332205655e-05, "loss": 0.9176, "step": 9113 }, { "epoch": 0.07, "learning_rate": 1.6331687074018997e-05, "loss": 0.9037, "step": 9114 }, { "epoch": 0.07, "learning_rate": 1.633085274226903e-05, "loss": 0.8893, "step": 9115 }, { "epoch": 0.07, "learning_rate": 1.633001833696544e-05, "loss": 0.903, "step": 9116 }, { "epoch": 0.07, "learning_rate": 1.6329183858117935e-05, "loss": 0.9341, "step": 9117 }, { "epoch": 0.07, "learning_rate": 1.6328349305736195e-05, "loss": 0.8948, "step": 9118 }, { "epoch": 0.07, "learning_rate": 1.6327514679829925e-05, "loss": 0.9526, "step": 9119 }, { "epoch": 0.07, "learning_rate": 1.6326679980408822e-05, "loss": 0.8988, "step": 9120 }, { "epoch": 0.07, "learning_rate": 1.632584520748258e-05, "loss": 0.9596, "step": 9121 }, { "epoch": 0.07, "learning_rate": 1.6325010361060904e-05, "loss": 0.9262, "step": 9122 }, { "epoch": 0.07, "learning_rate": 1.6324175441153485e-05, "loss": 0.9662, "step": 9123 }, { "epoch": 0.07, "learning_rate": 1.6323340447770028e-05, "loss": 0.8688, "step": 9124 }, { "epoch": 0.07, "learning_rate": 1.6322505380920236e-05, "loss": 0.9366, "step": 9125 }, { "epoch": 0.07, "learning_rate": 1.6321670240613805e-05, "loss": 0.9075, "step": 9126 }, { "epoch": 0.07, "learning_rate": 1.6320835026860442e-05, "loss": 0.9148, "step": 9127 }, { "epoch": 0.07, "learning_rate": 1.6319999739669854e-05, "loss": 0.8907, "step": 9128 }, { "epoch": 0.07, "learning_rate": 1.631916437905174e-05, "loss": 0.9305, "step": 9129 }, { "epoch": 0.07, "learning_rate": 1.6318328945015807e-05, "loss": 0.9405, "step": 9130 }, { "epoch": 0.07, "learning_rate": 1.6317493437571764e-05, "loss": 0.9204, "step": 9131 }, { "epoch": 0.07, "learning_rate": 1.631665785672931e-05, "loss": 0.9263, "step": 9132 }, { "epoch": 0.07, "learning_rate": 1.6315822202498165e-05, "loss": 0.9245, "step": 9133 }, { "epoch": 0.07, "learning_rate": 1.631498647488803e-05, "loss": 0.9199, "step": 9134 }, { "epoch": 0.07, "learning_rate": 1.6314150673908615e-05, "loss": 0.9613, "step": 9135 }, { "epoch": 0.07, "learning_rate": 1.6313314799569632e-05, "loss": 0.9101, "step": 9136 }, { "epoch": 0.07, "learning_rate": 1.6312478851880795e-05, "loss": 0.9186, "step": 9137 }, { "epoch": 0.07, "learning_rate": 1.6311642830851812e-05, "loss": 0.9483, "step": 9138 }, { "epoch": 0.07, "learning_rate": 1.6310806736492396e-05, "loss": 0.9293, "step": 9139 }, { "epoch": 0.07, "learning_rate": 1.6309970568812266e-05, "loss": 0.9328, "step": 9140 }, { "epoch": 0.07, "learning_rate": 1.6309134327821133e-05, "loss": 0.9801, "step": 9141 }, { "epoch": 0.07, "learning_rate": 1.630829801352871e-05, "loss": 0.8698, "step": 9142 }, { "epoch": 0.07, "learning_rate": 1.630746162594472e-05, "loss": 0.9456, "step": 9143 }, { "epoch": 0.07, "learning_rate": 1.6306625165078877e-05, "loss": 0.9348, "step": 9144 }, { "epoch": 0.07, "learning_rate": 1.63057886309409e-05, "loss": 0.8844, "step": 9145 }, { "epoch": 0.07, "learning_rate": 1.630495202354051e-05, "loss": 0.8969, "step": 9146 }, { "epoch": 0.07, "learning_rate": 1.6304115342887417e-05, "loss": 0.9253, "step": 9147 }, { "epoch": 0.07, "learning_rate": 1.6303278588991356e-05, "loss": 0.864, "step": 9148 }, { "epoch": 0.07, "learning_rate": 1.630244176186204e-05, "loss": 0.9124, "step": 9149 }, { "epoch": 0.07, "learning_rate": 1.6301604861509192e-05, "loss": 0.937, "step": 9150 }, { "epoch": 0.07, "learning_rate": 1.630076788794254e-05, "loss": 0.939, "step": 9151 }, { "epoch": 0.07, "learning_rate": 1.6299930841171803e-05, "loss": 0.9018, "step": 9152 }, { "epoch": 0.07, "learning_rate": 1.6299093721206706e-05, "loss": 0.9597, "step": 9153 }, { "epoch": 0.07, "learning_rate": 1.629825652805698e-05, "loss": 0.9182, "step": 9154 }, { "epoch": 0.07, "learning_rate": 1.6297419261732347e-05, "loss": 0.9376, "step": 9155 }, { "epoch": 0.07, "learning_rate": 1.6296581922242535e-05, "loss": 0.9284, "step": 9156 }, { "epoch": 0.07, "learning_rate": 1.629574450959728e-05, "loss": 0.9357, "step": 9157 }, { "epoch": 0.07, "learning_rate": 1.62949070238063e-05, "loss": 0.8921, "step": 9158 }, { "epoch": 0.07, "learning_rate": 1.6294069464879328e-05, "loss": 0.9533, "step": 9159 }, { "epoch": 0.07, "learning_rate": 1.62932318328261e-05, "loss": 0.8924, "step": 9160 }, { "epoch": 0.07, "learning_rate": 1.6292394127656346e-05, "loss": 0.9415, "step": 9161 }, { "epoch": 0.07, "learning_rate": 1.6291556349379794e-05, "loss": 0.874, "step": 9162 }, { "epoch": 0.07, "learning_rate": 1.6290718498006186e-05, "loss": 0.8952, "step": 9163 }, { "epoch": 0.07, "learning_rate": 1.6289880573545246e-05, "loss": 0.9202, "step": 9164 }, { "epoch": 0.07, "learning_rate": 1.628904257600672e-05, "loss": 0.921, "step": 9165 }, { "epoch": 0.07, "learning_rate": 1.6288204505400336e-05, "loss": 0.9459, "step": 9166 }, { "epoch": 0.07, "learning_rate": 1.6287366361735838e-05, "loss": 0.9732, "step": 9167 }, { "epoch": 0.07, "learning_rate": 1.6286528145022954e-05, "loss": 0.9026, "step": 9168 }, { "epoch": 0.07, "learning_rate": 1.6285689855271432e-05, "loss": 0.9289, "step": 9169 }, { "epoch": 0.07, "learning_rate": 1.6284851492491007e-05, "loss": 0.9388, "step": 9170 }, { "epoch": 0.07, "learning_rate": 1.6284013056691418e-05, "loss": 0.9015, "step": 9171 }, { "epoch": 0.07, "learning_rate": 1.628317454788241e-05, "loss": 0.8879, "step": 9172 }, { "epoch": 0.07, "learning_rate": 1.6282335966073725e-05, "loss": 0.9368, "step": 9173 }, { "epoch": 0.07, "learning_rate": 1.62814973112751e-05, "loss": 0.9173, "step": 9174 }, { "epoch": 0.07, "learning_rate": 1.6280658583496288e-05, "loss": 0.9386, "step": 9175 }, { "epoch": 0.07, "learning_rate": 1.6279819782747024e-05, "loss": 0.9135, "step": 9176 }, { "epoch": 0.07, "learning_rate": 1.627898090903706e-05, "loss": 0.937, "step": 9177 }, { "epoch": 0.07, "learning_rate": 1.627814196237614e-05, "loss": 0.9206, "step": 9178 }, { "epoch": 0.07, "learning_rate": 1.627730294277401e-05, "loss": 0.9215, "step": 9179 }, { "epoch": 0.07, "learning_rate": 1.6276463850240422e-05, "loss": 0.9226, "step": 9180 }, { "epoch": 0.07, "learning_rate": 1.6275624684785117e-05, "loss": 0.897, "step": 9181 }, { "epoch": 0.07, "learning_rate": 1.6274785446417857e-05, "loss": 0.9391, "step": 9182 }, { "epoch": 0.07, "learning_rate": 1.627394613514838e-05, "loss": 0.9255, "step": 9183 }, { "epoch": 0.07, "learning_rate": 1.6273106750986446e-05, "loss": 0.8977, "step": 9184 }, { "epoch": 0.07, "learning_rate": 1.62722672939418e-05, "loss": 0.8705, "step": 9185 }, { "epoch": 0.07, "learning_rate": 1.6271427764024205e-05, "loss": 0.9442, "step": 9186 }, { "epoch": 0.07, "learning_rate": 1.62705881612434e-05, "loss": 0.9197, "step": 9187 }, { "epoch": 0.07, "learning_rate": 1.6269748485609155e-05, "loss": 0.9568, "step": 9188 }, { "epoch": 0.07, "learning_rate": 1.6268908737131216e-05, "loss": 0.894, "step": 9189 }, { "epoch": 0.07, "learning_rate": 1.6268068915819344e-05, "loss": 0.9051, "step": 9190 }, { "epoch": 0.07, "learning_rate": 1.6267229021683296e-05, "loss": 0.9226, "step": 9191 }, { "epoch": 0.07, "learning_rate": 1.6266389054732827e-05, "loss": 0.8791, "step": 9192 }, { "epoch": 0.07, "learning_rate": 1.6265549014977697e-05, "loss": 0.9103, "step": 9193 }, { "epoch": 0.07, "learning_rate": 1.6264708902427668e-05, "loss": 0.9293, "step": 9194 }, { "epoch": 0.07, "learning_rate": 1.62638687170925e-05, "loss": 0.9305, "step": 9195 }, { "epoch": 0.07, "learning_rate": 1.626302845898195e-05, "loss": 0.8943, "step": 9196 }, { "epoch": 0.07, "learning_rate": 1.626218812810579e-05, "loss": 0.9111, "step": 9197 }, { "epoch": 0.07, "learning_rate": 1.6261347724473776e-05, "loss": 0.9356, "step": 9198 }, { "epoch": 0.07, "learning_rate": 1.626050724809567e-05, "loss": 0.9043, "step": 9199 }, { "epoch": 0.07, "learning_rate": 1.625966669898124e-05, "loss": 0.9672, "step": 9200 }, { "epoch": 0.07, "learning_rate": 1.6258826077140258e-05, "loss": 0.9758, "step": 9201 }, { "epoch": 0.07, "learning_rate": 1.625798538258248e-05, "loss": 0.8921, "step": 9202 }, { "epoch": 0.07, "learning_rate": 1.625714461531768e-05, "loss": 0.9506, "step": 9203 }, { "epoch": 0.07, "learning_rate": 1.625630377535562e-05, "loss": 0.8906, "step": 9204 }, { "epoch": 0.07, "learning_rate": 1.625546286270608e-05, "loss": 0.9159, "step": 9205 }, { "epoch": 0.07, "learning_rate": 1.625462187737882e-05, "loss": 0.8973, "step": 9206 }, { "epoch": 0.07, "learning_rate": 1.6253780819383615e-05, "loss": 0.9103, "step": 9207 }, { "epoch": 0.07, "learning_rate": 1.6252939688730236e-05, "loss": 0.8973, "step": 9208 }, { "epoch": 0.07, "learning_rate": 1.6252098485428457e-05, "loss": 0.9221, "step": 9209 }, { "epoch": 0.07, "learning_rate": 1.6251257209488047e-05, "loss": 0.9013, "step": 9210 }, { "epoch": 0.07, "learning_rate": 1.6250415860918782e-05, "loss": 0.9061, "step": 9211 }, { "epoch": 0.07, "learning_rate": 1.624957443973044e-05, "loss": 0.9205, "step": 9212 }, { "epoch": 0.07, "learning_rate": 1.6248732945932795e-05, "loss": 0.8999, "step": 9213 }, { "epoch": 0.07, "learning_rate": 1.6247891379535622e-05, "loss": 0.9543, "step": 9214 }, { "epoch": 0.07, "learning_rate": 1.62470497405487e-05, "loss": 0.9554, "step": 9215 }, { "epoch": 0.07, "learning_rate": 1.624620802898181e-05, "loss": 0.9169, "step": 9216 }, { "epoch": 0.07, "learning_rate": 1.624536624484473e-05, "loss": 0.8937, "step": 9217 }, { "epoch": 0.07, "learning_rate": 1.6244524388147233e-05, "loss": 0.8749, "step": 9218 }, { "epoch": 0.07, "learning_rate": 1.6243682458899106e-05, "loss": 0.9437, "step": 9219 }, { "epoch": 0.07, "learning_rate": 1.6242840457110135e-05, "loss": 0.9005, "step": 9220 }, { "epoch": 0.07, "learning_rate": 1.6241998382790095e-05, "loss": 0.9003, "step": 9221 }, { "epoch": 0.07, "learning_rate": 1.6241156235948775e-05, "loss": 0.9121, "step": 9222 }, { "epoch": 0.07, "learning_rate": 1.6240314016595956e-05, "loss": 0.9278, "step": 9223 }, { "epoch": 0.07, "learning_rate": 1.623947172474142e-05, "loss": 0.8708, "step": 9224 }, { "epoch": 0.07, "learning_rate": 1.6238629360394962e-05, "loss": 0.9229, "step": 9225 }, { "epoch": 0.07, "learning_rate": 1.623778692356636e-05, "loss": 0.9086, "step": 9226 }, { "epoch": 0.07, "learning_rate": 1.6236944414265407e-05, "loss": 0.9672, "step": 9227 }, { "epoch": 0.07, "learning_rate": 1.6236101832501892e-05, "loss": 0.8895, "step": 9228 }, { "epoch": 0.07, "learning_rate": 1.6235259178285597e-05, "loss": 0.9064, "step": 9229 }, { "epoch": 0.07, "learning_rate": 1.6234416451626325e-05, "loss": 0.9006, "step": 9230 }, { "epoch": 0.07, "learning_rate": 1.6233573652533853e-05, "loss": 0.9258, "step": 9231 }, { "epoch": 0.07, "learning_rate": 1.623273078101798e-05, "loss": 0.9348, "step": 9232 }, { "epoch": 0.07, "learning_rate": 1.6231887837088503e-05, "loss": 0.918, "step": 9233 }, { "epoch": 0.07, "learning_rate": 1.6231044820755205e-05, "loss": 0.9408, "step": 9234 }, { "epoch": 0.07, "learning_rate": 1.623020173202789e-05, "loss": 0.8952, "step": 9235 }, { "epoch": 0.07, "learning_rate": 1.622935857091635e-05, "loss": 0.9245, "step": 9236 }, { "epoch": 0.07, "learning_rate": 1.6228515337430377e-05, "loss": 0.9037, "step": 9237 }, { "epoch": 0.07, "learning_rate": 1.6227672031579774e-05, "loss": 0.919, "step": 9238 }, { "epoch": 0.07, "learning_rate": 1.6226828653374336e-05, "loss": 0.9242, "step": 9239 }, { "epoch": 0.07, "learning_rate": 1.6225985202823858e-05, "loss": 0.9096, "step": 9240 }, { "epoch": 0.07, "learning_rate": 1.6225141679938148e-05, "loss": 0.9075, "step": 9241 }, { "epoch": 0.07, "learning_rate": 1.6224298084726996e-05, "loss": 0.9023, "step": 9242 }, { "epoch": 0.07, "learning_rate": 1.6223454417200212e-05, "loss": 0.8865, "step": 9243 }, { "epoch": 0.07, "learning_rate": 1.6222610677367596e-05, "loss": 0.8798, "step": 9244 }, { "epoch": 0.07, "learning_rate": 1.6221766865238946e-05, "loss": 0.9105, "step": 9245 }, { "epoch": 0.07, "learning_rate": 1.622092298082407e-05, "loss": 0.8735, "step": 9246 }, { "epoch": 0.07, "learning_rate": 1.622007902413277e-05, "loss": 0.9194, "step": 9247 }, { "epoch": 0.07, "learning_rate": 1.6219234995174858e-05, "loss": 0.9236, "step": 9248 }, { "epoch": 0.07, "learning_rate": 1.621839089396013e-05, "loss": 0.9318, "step": 9249 }, { "epoch": 0.07, "learning_rate": 1.62175467204984e-05, "loss": 0.9354, "step": 9250 }, { "epoch": 0.07, "learning_rate": 1.6216702474799472e-05, "loss": 0.9179, "step": 9251 }, { "epoch": 0.07, "learning_rate": 1.6215858156873155e-05, "loss": 0.8736, "step": 9252 }, { "epoch": 0.07, "learning_rate": 1.6215013766729266e-05, "loss": 0.9267, "step": 9253 }, { "epoch": 0.08, "learning_rate": 1.6214169304377607e-05, "loss": 0.8823, "step": 9254 }, { "epoch": 0.08, "learning_rate": 1.621332476982799e-05, "loss": 0.9674, "step": 9255 }, { "epoch": 0.08, "learning_rate": 1.6212480163090227e-05, "loss": 0.9186, "step": 9256 }, { "epoch": 0.08, "learning_rate": 1.6211635484174134e-05, "loss": 0.9003, "step": 9257 }, { "epoch": 0.08, "learning_rate": 1.6210790733089523e-05, "loss": 0.9135, "step": 9258 }, { "epoch": 0.08, "learning_rate": 1.620994590984621e-05, "loss": 0.9131, "step": 9259 }, { "epoch": 0.08, "learning_rate": 1.620910101445401e-05, "loss": 0.8991, "step": 9260 }, { "epoch": 0.08, "learning_rate": 1.6208256046922737e-05, "loss": 0.9433, "step": 9261 }, { "epoch": 0.08, "learning_rate": 1.6207411007262207e-05, "loss": 0.9227, "step": 9262 }, { "epoch": 0.08, "learning_rate": 1.6206565895482242e-05, "loss": 0.945, "step": 9263 }, { "epoch": 0.08, "learning_rate": 1.620572071159266e-05, "loss": 0.8936, "step": 9264 }, { "epoch": 0.08, "learning_rate": 1.620487545560328e-05, "loss": 0.929, "step": 9265 }, { "epoch": 0.08, "learning_rate": 1.6204030127523922e-05, "loss": 0.9236, "step": 9266 }, { "epoch": 0.08, "learning_rate": 1.6203184727364407e-05, "loss": 0.9558, "step": 9267 }, { "epoch": 0.08, "learning_rate": 1.6202339255134556e-05, "loss": 0.9414, "step": 9268 }, { "epoch": 0.08, "learning_rate": 1.6201493710844193e-05, "loss": 0.9388, "step": 9269 }, { "epoch": 0.08, "learning_rate": 1.6200648094503144e-05, "loss": 0.8995, "step": 9270 }, { "epoch": 0.08, "learning_rate": 1.619980240612123e-05, "loss": 0.9282, "step": 9271 }, { "epoch": 0.08, "learning_rate": 1.6198956645708278e-05, "loss": 0.8928, "step": 9272 }, { "epoch": 0.08, "learning_rate": 1.6198110813274118e-05, "loss": 0.9147, "step": 9273 }, { "epoch": 0.08, "learning_rate": 1.619726490882857e-05, "loss": 0.9293, "step": 9274 }, { "epoch": 0.08, "learning_rate": 1.6196418932381466e-05, "loss": 0.9008, "step": 9275 }, { "epoch": 0.08, "learning_rate": 1.619557288394264e-05, "loss": 0.9614, "step": 9276 }, { "epoch": 0.08, "learning_rate": 1.619472676352191e-05, "loss": 0.8945, "step": 9277 }, { "epoch": 0.08, "learning_rate": 1.619388057112911e-05, "loss": 0.8986, "step": 9278 }, { "epoch": 0.08, "learning_rate": 1.6193034306774082e-05, "loss": 0.9341, "step": 9279 }, { "epoch": 0.08, "learning_rate": 1.6192187970466646e-05, "loss": 0.861, "step": 9280 }, { "epoch": 0.08, "learning_rate": 1.6191341562216636e-05, "loss": 0.9407, "step": 9281 }, { "epoch": 0.08, "learning_rate": 1.619049508203389e-05, "loss": 0.8393, "step": 9282 }, { "epoch": 0.08, "learning_rate": 1.618964852992824e-05, "loss": 0.9305, "step": 9283 }, { "epoch": 0.08, "learning_rate": 1.6188801905909526e-05, "loss": 0.9314, "step": 9284 }, { "epoch": 0.08, "learning_rate": 1.618795520998758e-05, "loss": 0.9491, "step": 9285 }, { "epoch": 0.08, "learning_rate": 1.618710844217224e-05, "loss": 0.9107, "step": 9286 }, { "epoch": 0.08, "learning_rate": 1.6186261602473344e-05, "loss": 0.9208, "step": 9287 }, { "epoch": 0.08, "learning_rate": 1.618541469090073e-05, "loss": 0.9101, "step": 9288 }, { "epoch": 0.08, "learning_rate": 1.618456770746424e-05, "loss": 0.9443, "step": 9289 }, { "epoch": 0.08, "learning_rate": 1.6183720652173716e-05, "loss": 0.9122, "step": 9290 }, { "epoch": 0.08, "learning_rate": 1.618287352503899e-05, "loss": 0.9422, "step": 9291 }, { "epoch": 0.08, "learning_rate": 1.6182026326069913e-05, "loss": 0.9213, "step": 9292 }, { "epoch": 0.08, "learning_rate": 1.618117905527633e-05, "loss": 0.8941, "step": 9293 }, { "epoch": 0.08, "learning_rate": 1.6180331712668077e-05, "loss": 0.8974, "step": 9294 }, { "epoch": 0.08, "learning_rate": 1.6179484298254998e-05, "loss": 0.9506, "step": 9295 }, { "epoch": 0.08, "learning_rate": 1.617863681204695e-05, "loss": 0.9015, "step": 9296 }, { "epoch": 0.08, "learning_rate": 1.617778925405377e-05, "loss": 0.9196, "step": 9297 }, { "epoch": 0.08, "learning_rate": 1.6176941624285305e-05, "loss": 0.8764, "step": 9298 }, { "epoch": 0.08, "learning_rate": 1.6176093922751408e-05, "loss": 0.9066, "step": 9299 }, { "epoch": 0.08, "learning_rate": 1.617524614946192e-05, "loss": 0.9095, "step": 9300 }, { "epoch": 0.08, "learning_rate": 1.61743983044267e-05, "loss": 0.9194, "step": 9301 }, { "epoch": 0.08, "learning_rate": 1.6173550387655593e-05, "loss": 0.8701, "step": 9302 }, { "epoch": 0.08, "learning_rate": 1.617270239915845e-05, "loss": 0.937, "step": 9303 }, { "epoch": 0.08, "learning_rate": 1.6171854338945124e-05, "loss": 0.9053, "step": 9304 }, { "epoch": 0.08, "learning_rate": 1.6171006207025475e-05, "loss": 0.8903, "step": 9305 }, { "epoch": 0.08, "learning_rate": 1.6170158003409343e-05, "loss": 0.8804, "step": 9306 }, { "epoch": 0.08, "learning_rate": 1.6169309728106593e-05, "loss": 0.8495, "step": 9307 }, { "epoch": 0.08, "learning_rate": 1.616846138112708e-05, "loss": 0.9366, "step": 9308 }, { "epoch": 0.08, "learning_rate": 1.6167612962480655e-05, "loss": 0.9536, "step": 9309 }, { "epoch": 0.08, "learning_rate": 1.616676447217718e-05, "loss": 0.8655, "step": 9310 }, { "epoch": 0.08, "learning_rate": 1.616591591022651e-05, "loss": 0.9467, "step": 9311 }, { "epoch": 0.08, "learning_rate": 1.6165067276638503e-05, "loss": 0.9265, "step": 9312 }, { "epoch": 0.08, "learning_rate": 1.6164218571423025e-05, "loss": 0.9426, "step": 9313 }, { "epoch": 0.08, "learning_rate": 1.6163369794589926e-05, "loss": 0.9144, "step": 9314 }, { "epoch": 0.08, "learning_rate": 1.6162520946149078e-05, "loss": 0.9064, "step": 9315 }, { "epoch": 0.08, "learning_rate": 1.6161672026110338e-05, "loss": 0.9127, "step": 9316 }, { "epoch": 0.08, "learning_rate": 1.616082303448357e-05, "loss": 0.9563, "step": 9317 }, { "epoch": 0.08, "learning_rate": 1.6159973971278637e-05, "loss": 0.8921, "step": 9318 }, { "epoch": 0.08, "learning_rate": 1.6159124836505403e-05, "loss": 0.8833, "step": 9319 }, { "epoch": 0.08, "learning_rate": 1.6158275630173735e-05, "loss": 0.9458, "step": 9320 }, { "epoch": 0.08, "learning_rate": 1.61574263522935e-05, "loss": 0.9061, "step": 9321 }, { "epoch": 0.08, "learning_rate": 1.6156577002874566e-05, "loss": 0.9461, "step": 9322 }, { "epoch": 0.08, "learning_rate": 1.6155727581926792e-05, "loss": 0.9195, "step": 9323 }, { "epoch": 0.08, "learning_rate": 1.615487808946006e-05, "loss": 0.9317, "step": 9324 }, { "epoch": 0.08, "learning_rate": 1.6154028525484233e-05, "loss": 0.9263, "step": 9325 }, { "epoch": 0.08, "learning_rate": 1.615317889000918e-05, "loss": 0.921, "step": 9326 }, { "epoch": 0.08, "learning_rate": 1.6152329183044772e-05, "loss": 0.91, "step": 9327 }, { "epoch": 0.08, "learning_rate": 1.6151479404600888e-05, "loss": 0.9153, "step": 9328 }, { "epoch": 0.08, "learning_rate": 1.6150629554687393e-05, "loss": 0.9303, "step": 9329 }, { "epoch": 0.08, "learning_rate": 1.6149779633314166e-05, "loss": 0.8925, "step": 9330 }, { "epoch": 0.08, "learning_rate": 1.614892964049108e-05, "loss": 0.9569, "step": 9331 }, { "epoch": 0.08, "learning_rate": 1.6148079576228008e-05, "loss": 0.9309, "step": 9332 }, { "epoch": 0.08, "learning_rate": 1.6147229440534832e-05, "loss": 0.9029, "step": 9333 }, { "epoch": 0.08, "learning_rate": 1.614637923342142e-05, "loss": 0.944, "step": 9334 }, { "epoch": 0.08, "learning_rate": 1.614552895489766e-05, "loss": 0.9015, "step": 9335 }, { "epoch": 0.08, "learning_rate": 1.6144678604973426e-05, "loss": 0.9126, "step": 9336 }, { "epoch": 0.08, "learning_rate": 1.6143828183658596e-05, "loss": 0.9109, "step": 9337 }, { "epoch": 0.08, "learning_rate": 1.6142977690963056e-05, "loss": 0.9132, "step": 9338 }, { "epoch": 0.08, "learning_rate": 1.6142127126896682e-05, "loss": 0.9088, "step": 9339 }, { "epoch": 0.08, "learning_rate": 1.6141276491469355e-05, "loss": 0.9191, "step": 9340 }, { "epoch": 0.08, "learning_rate": 1.6140425784690964e-05, "loss": 0.946, "step": 9341 }, { "epoch": 0.08, "learning_rate": 1.6139575006571385e-05, "loss": 0.897, "step": 9342 }, { "epoch": 0.08, "learning_rate": 1.613872415712051e-05, "loss": 0.8916, "step": 9343 }, { "epoch": 0.08, "learning_rate": 1.613787323634822e-05, "loss": 0.9173, "step": 9344 }, { "epoch": 0.08, "learning_rate": 1.6137022244264405e-05, "loss": 0.9309, "step": 9345 }, { "epoch": 0.08, "learning_rate": 1.6136171180878946e-05, "loss": 0.9028, "step": 9346 }, { "epoch": 0.08, "learning_rate": 1.613532004620174e-05, "loss": 0.9217, "step": 9347 }, { "epoch": 0.08, "learning_rate": 1.6134468840242666e-05, "loss": 0.9221, "step": 9348 }, { "epoch": 0.08, "learning_rate": 1.613361756301162e-05, "loss": 0.9355, "step": 9349 }, { "epoch": 0.08, "learning_rate": 1.6132766214518486e-05, "loss": 0.8904, "step": 9350 }, { "epoch": 0.08, "learning_rate": 1.6131914794773165e-05, "loss": 0.9321, "step": 9351 }, { "epoch": 0.08, "learning_rate": 1.6131063303785542e-05, "loss": 0.9006, "step": 9352 }, { "epoch": 0.08, "learning_rate": 1.6130211741565507e-05, "loss": 0.9108, "step": 9353 }, { "epoch": 0.08, "learning_rate": 1.6129360108122965e-05, "loss": 0.9141, "step": 9354 }, { "epoch": 0.08, "learning_rate": 1.61285084034678e-05, "loss": 0.9641, "step": 9355 }, { "epoch": 0.08, "learning_rate": 1.612765662760991e-05, "loss": 0.9083, "step": 9356 }, { "epoch": 0.08, "learning_rate": 1.612680478055919e-05, "loss": 0.9247, "step": 9357 }, { "epoch": 0.08, "learning_rate": 1.612595286232554e-05, "loss": 0.8939, "step": 9358 }, { "epoch": 0.08, "learning_rate": 1.6125100872918863e-05, "loss": 0.9366, "step": 9359 }, { "epoch": 0.08, "learning_rate": 1.6124248812349044e-05, "loss": 0.92, "step": 9360 }, { "epoch": 0.08, "learning_rate": 1.612339668062599e-05, "loss": 0.9095, "step": 9361 }, { "epoch": 0.08, "learning_rate": 1.6122544477759604e-05, "loss": 0.9329, "step": 9362 }, { "epoch": 0.08, "learning_rate": 1.6121692203759786e-05, "loss": 0.9439, "step": 9363 }, { "epoch": 0.08, "learning_rate": 1.6120839858636435e-05, "loss": 0.9199, "step": 9364 }, { "epoch": 0.08, "learning_rate": 1.6119987442399456e-05, "loss": 0.883, "step": 9365 }, { "epoch": 0.08, "learning_rate": 1.6119134955058748e-05, "loss": 0.8638, "step": 9366 }, { "epoch": 0.08, "learning_rate": 1.6118282396624223e-05, "loss": 0.9345, "step": 9367 }, { "epoch": 0.08, "learning_rate": 1.6117429767105782e-05, "loss": 0.9201, "step": 9368 }, { "epoch": 0.08, "learning_rate": 1.611657706651333e-05, "loss": 0.9398, "step": 9369 }, { "epoch": 0.08, "learning_rate": 1.6115724294856775e-05, "loss": 0.8931, "step": 9370 }, { "epoch": 0.08, "learning_rate": 1.6114871452146026e-05, "loss": 0.8976, "step": 9371 }, { "epoch": 0.08, "learning_rate": 1.6114018538390995e-05, "loss": 0.9102, "step": 9372 }, { "epoch": 0.08, "learning_rate": 1.611316555360158e-05, "loss": 0.9176, "step": 9373 }, { "epoch": 0.08, "learning_rate": 1.6112312497787705e-05, "loss": 0.9375, "step": 9374 }, { "epoch": 0.08, "learning_rate": 1.611145937095927e-05, "loss": 0.9502, "step": 9375 }, { "epoch": 0.08, "learning_rate": 1.611060617312619e-05, "loss": 0.9164, "step": 9376 }, { "epoch": 0.08, "learning_rate": 1.6109752904298383e-05, "loss": 0.8915, "step": 9377 }, { "epoch": 0.08, "learning_rate": 1.6108899564485753e-05, "loss": 0.8966, "step": 9378 }, { "epoch": 0.08, "learning_rate": 1.6108046153698226e-05, "loss": 0.9684, "step": 9379 }, { "epoch": 0.08, "learning_rate": 1.6107192671945706e-05, "loss": 0.9095, "step": 9380 }, { "epoch": 0.08, "learning_rate": 1.6106339119238117e-05, "loss": 0.8776, "step": 9381 }, { "epoch": 0.08, "learning_rate": 1.6105485495585373e-05, "loss": 0.9001, "step": 9382 }, { "epoch": 0.08, "learning_rate": 1.610463180099739e-05, "loss": 0.9239, "step": 9383 }, { "epoch": 0.08, "learning_rate": 1.610377803548409e-05, "loss": 0.8872, "step": 9384 }, { "epoch": 0.08, "learning_rate": 1.610292419905539e-05, "loss": 0.9127, "step": 9385 }, { "epoch": 0.08, "learning_rate": 1.6102070291721207e-05, "loss": 0.9651, "step": 9386 }, { "epoch": 0.08, "learning_rate": 1.610121631349147e-05, "loss": 0.9336, "step": 9387 }, { "epoch": 0.08, "learning_rate": 1.610036226437609e-05, "loss": 0.9039, "step": 9388 }, { "epoch": 0.08, "learning_rate": 1.6099508144384998e-05, "loss": 0.9112, "step": 9389 }, { "epoch": 0.08, "learning_rate": 1.6098653953528116e-05, "loss": 0.9295, "step": 9390 }, { "epoch": 0.08, "learning_rate": 1.6097799691815367e-05, "loss": 0.9542, "step": 9391 }, { "epoch": 0.08, "learning_rate": 1.6096945359256676e-05, "loss": 0.888, "step": 9392 }, { "epoch": 0.08, "learning_rate": 1.6096090955861966e-05, "loss": 0.8813, "step": 9393 }, { "epoch": 0.08, "learning_rate": 1.609523648164117e-05, "loss": 0.9234, "step": 9394 }, { "epoch": 0.08, "learning_rate": 1.6094381936604216e-05, "loss": 0.8967, "step": 9395 }, { "epoch": 0.08, "learning_rate": 1.609352732076102e-05, "loss": 0.8766, "step": 9396 }, { "epoch": 0.08, "learning_rate": 1.6092672634121525e-05, "loss": 0.9225, "step": 9397 }, { "epoch": 0.08, "learning_rate": 1.6091817876695655e-05, "loss": 0.9328, "step": 9398 }, { "epoch": 0.08, "learning_rate": 1.6090963048493345e-05, "loss": 0.9441, "step": 9399 }, { "epoch": 0.08, "learning_rate": 1.609010814952452e-05, "loss": 0.9382, "step": 9400 }, { "epoch": 0.08, "learning_rate": 1.6089253179799115e-05, "loss": 0.921, "step": 9401 }, { "epoch": 0.08, "learning_rate": 1.608839813932707e-05, "loss": 0.8714, "step": 9402 }, { "epoch": 0.08, "learning_rate": 1.60875430281183e-05, "loss": 0.9045, "step": 9403 }, { "epoch": 0.08, "learning_rate": 1.6086687846182764e-05, "loss": 0.918, "step": 9404 }, { "epoch": 0.08, "learning_rate": 1.6085832593530388e-05, "loss": 0.9254, "step": 9405 }, { "epoch": 0.08, "learning_rate": 1.60849772701711e-05, "loss": 0.9192, "step": 9406 }, { "epoch": 0.08, "learning_rate": 1.608412187611485e-05, "loss": 0.9184, "step": 9407 }, { "epoch": 0.08, "learning_rate": 1.608326641137157e-05, "loss": 0.8864, "step": 9408 }, { "epoch": 0.08, "learning_rate": 1.60824108759512e-05, "loss": 0.9248, "step": 9409 }, { "epoch": 0.08, "learning_rate": 1.6081555269863676e-05, "loss": 0.8982, "step": 9410 }, { "epoch": 0.08, "learning_rate": 1.6080699593118945e-05, "loss": 0.9293, "step": 9411 }, { "epoch": 0.08, "learning_rate": 1.6079843845726943e-05, "loss": 0.8756, "step": 9412 }, { "epoch": 0.08, "learning_rate": 1.607898802769762e-05, "loss": 0.9335, "step": 9413 }, { "epoch": 0.08, "learning_rate": 1.607813213904091e-05, "loss": 0.8893, "step": 9414 }, { "epoch": 0.08, "learning_rate": 1.607727617976677e-05, "loss": 0.9039, "step": 9415 }, { "epoch": 0.08, "learning_rate": 1.6076420149885125e-05, "loss": 0.911, "step": 9416 }, { "epoch": 0.08, "learning_rate": 1.6075564049405938e-05, "loss": 0.9459, "step": 9417 }, { "epoch": 0.08, "learning_rate": 1.607470787833915e-05, "loss": 0.9359, "step": 9418 }, { "epoch": 0.08, "learning_rate": 1.6073851636694705e-05, "loss": 0.9475, "step": 9419 }, { "epoch": 0.08, "learning_rate": 1.6072995324482553e-05, "loss": 0.8636, "step": 9420 }, { "epoch": 0.08, "learning_rate": 1.6072138941712646e-05, "loss": 0.9252, "step": 9421 }, { "epoch": 0.08, "learning_rate": 1.6071282488394925e-05, "loss": 0.933, "step": 9422 }, { "epoch": 0.08, "learning_rate": 1.6070425964539354e-05, "loss": 0.9412, "step": 9423 }, { "epoch": 0.08, "learning_rate": 1.606956937015587e-05, "loss": 0.9004, "step": 9424 }, { "epoch": 0.08, "learning_rate": 1.606871270525444e-05, "loss": 0.9209, "step": 9425 }, { "epoch": 0.08, "learning_rate": 1.6067855969845002e-05, "loss": 0.9114, "step": 9426 }, { "epoch": 0.08, "learning_rate": 1.606699916393752e-05, "loss": 0.95, "step": 9427 }, { "epoch": 0.08, "learning_rate": 1.6066142287541943e-05, "loss": 0.9224, "step": 9428 }, { "epoch": 0.08, "learning_rate": 1.606528534066823e-05, "loss": 0.9007, "step": 9429 }, { "epoch": 0.08, "learning_rate": 1.606442832332634e-05, "loss": 0.9128, "step": 9430 }, { "epoch": 0.08, "learning_rate": 1.6063571235526222e-05, "loss": 0.9362, "step": 9431 }, { "epoch": 0.08, "learning_rate": 1.6062714077277837e-05, "loss": 0.903, "step": 9432 }, { "epoch": 0.08, "learning_rate": 1.6061856848591145e-05, "loss": 0.9217, "step": 9433 }, { "epoch": 0.08, "learning_rate": 1.606099954947611e-05, "loss": 0.8937, "step": 9434 }, { "epoch": 0.08, "learning_rate": 1.606014217994268e-05, "loss": 0.9243, "step": 9435 }, { "epoch": 0.08, "learning_rate": 1.6059284740000827e-05, "loss": 0.8945, "step": 9436 }, { "epoch": 0.08, "learning_rate": 1.6058427229660514e-05, "loss": 0.8945, "step": 9437 }, { "epoch": 0.08, "learning_rate": 1.6057569648931696e-05, "loss": 0.9203, "step": 9438 }, { "epoch": 0.08, "learning_rate": 1.605671199782434e-05, "loss": 0.9556, "step": 9439 }, { "epoch": 0.08, "learning_rate": 1.605585427634841e-05, "loss": 0.9064, "step": 9440 }, { "epoch": 0.08, "learning_rate": 1.605499648451387e-05, "loss": 0.8702, "step": 9441 }, { "epoch": 0.08, "learning_rate": 1.6054138622330692e-05, "loss": 0.9137, "step": 9442 }, { "epoch": 0.08, "learning_rate": 1.6053280689808836e-05, "loss": 0.8888, "step": 9443 }, { "epoch": 0.08, "learning_rate": 1.6052422686958275e-05, "loss": 0.9112, "step": 9444 }, { "epoch": 0.08, "learning_rate": 1.6051564613788973e-05, "loss": 0.9189, "step": 9445 }, { "epoch": 0.08, "learning_rate": 1.6050706470310902e-05, "loss": 0.9114, "step": 9446 }, { "epoch": 0.08, "learning_rate": 1.604984825653403e-05, "loss": 0.9287, "step": 9447 }, { "epoch": 0.08, "learning_rate": 1.6048989972468333e-05, "loss": 0.9015, "step": 9448 }, { "epoch": 0.08, "learning_rate": 1.6048131618123775e-05, "loss": 0.9443, "step": 9449 }, { "epoch": 0.08, "learning_rate": 1.6047273193510332e-05, "loss": 0.9199, "step": 9450 }, { "epoch": 0.08, "learning_rate": 1.6046414698637983e-05, "loss": 0.9252, "step": 9451 }, { "epoch": 0.08, "learning_rate": 1.6045556133516696e-05, "loss": 0.8943, "step": 9452 }, { "epoch": 0.08, "learning_rate": 1.604469749815645e-05, "loss": 0.9364, "step": 9453 }, { "epoch": 0.08, "learning_rate": 1.6043838792567212e-05, "loss": 0.9176, "step": 9454 }, { "epoch": 0.08, "learning_rate": 1.6042980016758968e-05, "loss": 0.8926, "step": 9455 }, { "epoch": 0.08, "learning_rate": 1.604212117074169e-05, "loss": 0.9208, "step": 9456 }, { "epoch": 0.08, "learning_rate": 1.6041262254525362e-05, "loss": 0.9555, "step": 9457 }, { "epoch": 0.08, "learning_rate": 1.6040403268119963e-05, "loss": 0.8731, "step": 9458 }, { "epoch": 0.08, "learning_rate": 1.6039544211535462e-05, "loss": 0.9225, "step": 9459 }, { "epoch": 0.08, "learning_rate": 1.6038685084781854e-05, "loss": 0.9534, "step": 9460 }, { "epoch": 0.08, "learning_rate": 1.6037825887869113e-05, "loss": 0.8774, "step": 9461 }, { "epoch": 0.08, "learning_rate": 1.603696662080722e-05, "loss": 0.8741, "step": 9462 }, { "epoch": 0.08, "learning_rate": 1.6036107283606166e-05, "loss": 0.9179, "step": 9463 }, { "epoch": 0.08, "learning_rate": 1.6035247876275927e-05, "loss": 0.9161, "step": 9464 }, { "epoch": 0.08, "learning_rate": 1.603438839882649e-05, "loss": 0.9175, "step": 9465 }, { "epoch": 0.08, "learning_rate": 1.6033528851267842e-05, "loss": 0.894, "step": 9466 }, { "epoch": 0.08, "learning_rate": 1.6032669233609967e-05, "loss": 0.9206, "step": 9467 }, { "epoch": 0.08, "learning_rate": 1.6031809545862857e-05, "loss": 0.9085, "step": 9468 }, { "epoch": 0.08, "learning_rate": 1.6030949788036496e-05, "loss": 0.8897, "step": 9469 }, { "epoch": 0.08, "learning_rate": 1.6030089960140873e-05, "loss": 0.9259, "step": 9470 }, { "epoch": 0.08, "learning_rate": 1.6029230062185982e-05, "loss": 0.9657, "step": 9471 }, { "epoch": 0.08, "learning_rate": 1.6028370094181805e-05, "loss": 0.8713, "step": 9472 }, { "epoch": 0.08, "learning_rate": 1.6027510056138345e-05, "loss": 0.8965, "step": 9473 }, { "epoch": 0.08, "learning_rate": 1.6026649948065584e-05, "loss": 0.923, "step": 9474 }, { "epoch": 0.08, "learning_rate": 1.602578976997352e-05, "loss": 0.9205, "step": 9475 }, { "epoch": 0.08, "learning_rate": 1.6024929521872143e-05, "loss": 0.9306, "step": 9476 }, { "epoch": 0.08, "learning_rate": 1.6024069203771455e-05, "loss": 0.9323, "step": 9477 }, { "epoch": 0.08, "learning_rate": 1.6023208815681444e-05, "loss": 0.9085, "step": 9478 }, { "epoch": 0.08, "learning_rate": 1.6022348357612108e-05, "loss": 0.9339, "step": 9479 }, { "epoch": 0.08, "learning_rate": 1.6021487829573447e-05, "loss": 0.8969, "step": 9480 }, { "epoch": 0.08, "learning_rate": 1.6020627231575454e-05, "loss": 0.8899, "step": 9481 }, { "epoch": 0.08, "learning_rate": 1.6019766563628133e-05, "loss": 0.8687, "step": 9482 }, { "epoch": 0.08, "learning_rate": 1.6018905825741484e-05, "loss": 0.8925, "step": 9483 }, { "epoch": 0.08, "learning_rate": 1.6018045017925495e-05, "loss": 0.9215, "step": 9484 }, { "epoch": 0.08, "learning_rate": 1.6017184140190183e-05, "loss": 0.9292, "step": 9485 }, { "epoch": 0.08, "learning_rate": 1.601632319254554e-05, "loss": 0.8761, "step": 9486 }, { "epoch": 0.08, "learning_rate": 1.6015462175001576e-05, "loss": 0.9356, "step": 9487 }, { "epoch": 0.08, "learning_rate": 1.601460108756829e-05, "loss": 0.9311, "step": 9488 }, { "epoch": 0.08, "learning_rate": 1.6013739930255685e-05, "loss": 0.88, "step": 9489 }, { "epoch": 0.08, "learning_rate": 1.601287870307377e-05, "loss": 0.9484, "step": 9490 }, { "epoch": 0.08, "learning_rate": 1.6012017406032546e-05, "loss": 0.899, "step": 9491 }, { "epoch": 0.08, "learning_rate": 1.6011156039142022e-05, "loss": 0.8687, "step": 9492 }, { "epoch": 0.08, "learning_rate": 1.601029460241221e-05, "loss": 0.9455, "step": 9493 }, { "epoch": 0.08, "learning_rate": 1.6009433095853115e-05, "loss": 0.9106, "step": 9494 }, { "epoch": 0.08, "learning_rate": 1.6008571519474744e-05, "loss": 0.8989, "step": 9495 }, { "epoch": 0.08, "learning_rate": 1.6007709873287107e-05, "loss": 0.8842, "step": 9496 }, { "epoch": 0.08, "learning_rate": 1.600684815730022e-05, "loss": 0.9394, "step": 9497 }, { "epoch": 0.08, "learning_rate": 1.600598637152409e-05, "loss": 0.8826, "step": 9498 }, { "epoch": 0.08, "learning_rate": 1.6005124515968737e-05, "loss": 0.9199, "step": 9499 }, { "epoch": 0.08, "learning_rate": 1.600426259064416e-05, "loss": 0.9029, "step": 9500 }, { "epoch": 0.08, "learning_rate": 1.6003400595560385e-05, "loss": 0.8953, "step": 9501 }, { "epoch": 0.08, "learning_rate": 1.600253853072742e-05, "loss": 0.9037, "step": 9502 }, { "epoch": 0.08, "learning_rate": 1.600167639615529e-05, "loss": 0.9033, "step": 9503 }, { "epoch": 0.08, "learning_rate": 1.6000814191854002e-05, "loss": 0.9267, "step": 9504 }, { "epoch": 0.08, "learning_rate": 1.599995191783358e-05, "loss": 0.8892, "step": 9505 }, { "epoch": 0.08, "learning_rate": 1.5999089574104038e-05, "loss": 0.905, "step": 9506 }, { "epoch": 0.08, "learning_rate": 1.5998227160675393e-05, "loss": 0.9375, "step": 9507 }, { "epoch": 0.08, "learning_rate": 1.599736467755767e-05, "loss": 0.919, "step": 9508 }, { "epoch": 0.08, "learning_rate": 1.5996502124760883e-05, "loss": 0.9074, "step": 9509 }, { "epoch": 0.08, "learning_rate": 1.5995639502295067e-05, "loss": 0.8934, "step": 9510 }, { "epoch": 0.08, "learning_rate": 1.599477681017023e-05, "loss": 0.9434, "step": 9511 }, { "epoch": 0.08, "learning_rate": 1.5993914048396394e-05, "loss": 0.8834, "step": 9512 }, { "epoch": 0.08, "learning_rate": 1.5993051216983597e-05, "loss": 0.9161, "step": 9513 }, { "epoch": 0.08, "learning_rate": 1.5992188315941852e-05, "loss": 0.9384, "step": 9514 }, { "epoch": 0.08, "learning_rate": 1.599132534528119e-05, "loss": 0.9405, "step": 9515 }, { "epoch": 0.08, "learning_rate": 1.599046230501163e-05, "loss": 0.9101, "step": 9516 }, { "epoch": 0.08, "learning_rate": 1.5989599195143205e-05, "loss": 0.9085, "step": 9517 }, { "epoch": 0.08, "learning_rate": 1.5988736015685942e-05, "loss": 0.9281, "step": 9518 }, { "epoch": 0.08, "learning_rate": 1.5987872766649873e-05, "loss": 0.9736, "step": 9519 }, { "epoch": 0.08, "learning_rate": 1.5987009448045022e-05, "loss": 0.9145, "step": 9520 }, { "epoch": 0.08, "learning_rate": 1.5986146059881423e-05, "loss": 0.9215, "step": 9521 }, { "epoch": 0.08, "learning_rate": 1.5985282602169104e-05, "loss": 0.8902, "step": 9522 }, { "epoch": 0.08, "learning_rate": 1.5984419074918097e-05, "loss": 0.9562, "step": 9523 }, { "epoch": 0.08, "learning_rate": 1.598355547813844e-05, "loss": 0.8935, "step": 9524 }, { "epoch": 0.08, "learning_rate": 1.5982691811840155e-05, "loss": 0.9167, "step": 9525 }, { "epoch": 0.08, "learning_rate": 1.598182807603329e-05, "loss": 0.9336, "step": 9526 }, { "epoch": 0.08, "learning_rate": 1.598096427072787e-05, "loss": 0.9414, "step": 9527 }, { "epoch": 0.08, "learning_rate": 1.598010039593394e-05, "loss": 0.8927, "step": 9528 }, { "epoch": 0.08, "learning_rate": 1.5979236451661528e-05, "loss": 0.9201, "step": 9529 }, { "epoch": 0.08, "learning_rate": 1.597837243792067e-05, "loss": 0.9126, "step": 9530 }, { "epoch": 0.08, "learning_rate": 1.597750835472142e-05, "loss": 0.9239, "step": 9531 }, { "epoch": 0.08, "learning_rate": 1.59766442020738e-05, "loss": 0.9491, "step": 9532 }, { "epoch": 0.08, "learning_rate": 1.597577997998786e-05, "loss": 0.9277, "step": 9533 }, { "epoch": 0.08, "learning_rate": 1.5974915688473634e-05, "loss": 0.9307, "step": 9534 }, { "epoch": 0.08, "learning_rate": 1.597405132754117e-05, "loss": 0.9312, "step": 9535 }, { "epoch": 0.08, "learning_rate": 1.5973186897200506e-05, "loss": 0.9058, "step": 9536 }, { "epoch": 0.08, "learning_rate": 1.5972322397461682e-05, "loss": 0.8844, "step": 9537 }, { "epoch": 0.08, "learning_rate": 1.5971457828334756e-05, "loss": 0.8932, "step": 9538 }, { "epoch": 0.08, "learning_rate": 1.5970593189829757e-05, "loss": 0.9471, "step": 9539 }, { "epoch": 0.08, "learning_rate": 1.596972848195674e-05, "loss": 0.8957, "step": 9540 }, { "epoch": 0.08, "learning_rate": 1.5968863704725745e-05, "loss": 0.8978, "step": 9541 }, { "epoch": 0.08, "learning_rate": 1.5967998858146824e-05, "loss": 0.8645, "step": 9542 }, { "epoch": 0.08, "learning_rate": 1.5967133942230023e-05, "loss": 0.9371, "step": 9543 }, { "epoch": 0.08, "learning_rate": 1.5966268956985392e-05, "loss": 0.9012, "step": 9544 }, { "epoch": 0.08, "learning_rate": 1.596540390242298e-05, "loss": 0.9469, "step": 9545 }, { "epoch": 0.08, "learning_rate": 1.5964538778552837e-05, "loss": 0.8997, "step": 9546 }, { "epoch": 0.08, "learning_rate": 1.5963673585385016e-05, "loss": 0.92, "step": 9547 }, { "epoch": 0.08, "learning_rate": 1.5962808322929567e-05, "loss": 0.8802, "step": 9548 }, { "epoch": 0.08, "learning_rate": 1.5961942991196544e-05, "loss": 0.9232, "step": 9549 }, { "epoch": 0.08, "learning_rate": 1.5961077590196e-05, "loss": 0.89, "step": 9550 }, { "epoch": 0.08, "learning_rate": 1.596021211993799e-05, "loss": 0.9273, "step": 9551 }, { "epoch": 0.08, "learning_rate": 1.595934658043257e-05, "loss": 0.8848, "step": 9552 }, { "epoch": 0.08, "learning_rate": 1.5958480971689795e-05, "loss": 0.9252, "step": 9553 }, { "epoch": 0.08, "learning_rate": 1.5957615293719722e-05, "loss": 0.924, "step": 9554 }, { "epoch": 0.09, "learning_rate": 1.5956749546532407e-05, "loss": 0.9124, "step": 9555 }, { "epoch": 0.09, "learning_rate": 1.5955883730137912e-05, "loss": 0.8779, "step": 9556 }, { "epoch": 0.09, "learning_rate": 1.5955017844546294e-05, "loss": 0.9048, "step": 9557 }, { "epoch": 0.09, "learning_rate": 1.5954151889767616e-05, "loss": 0.9131, "step": 9558 }, { "epoch": 0.09, "learning_rate": 1.5953285865811933e-05, "loss": 0.9386, "step": 9559 }, { "epoch": 0.09, "learning_rate": 1.5952419772689308e-05, "loss": 0.8967, "step": 9560 }, { "epoch": 0.09, "learning_rate": 1.5951553610409814e-05, "loss": 0.9058, "step": 9561 }, { "epoch": 0.09, "learning_rate": 1.5950687378983497e-05, "loss": 0.9029, "step": 9562 }, { "epoch": 0.09, "learning_rate": 1.5949821078420436e-05, "loss": 0.9424, "step": 9563 }, { "epoch": 0.09, "learning_rate": 1.594895470873069e-05, "loss": 0.8851, "step": 9564 }, { "epoch": 0.09, "learning_rate": 1.5948088269924326e-05, "loss": 0.9136, "step": 9565 }, { "epoch": 0.09, "learning_rate": 1.5947221762011405e-05, "loss": 0.8933, "step": 9566 }, { "epoch": 0.09, "learning_rate": 1.5946355185002e-05, "loss": 0.8922, "step": 9567 }, { "epoch": 0.09, "learning_rate": 1.594548853890618e-05, "loss": 0.9506, "step": 9568 }, { "epoch": 0.09, "learning_rate": 1.594462182373401e-05, "loss": 0.9481, "step": 9569 }, { "epoch": 0.09, "learning_rate": 1.5943755039495562e-05, "loss": 0.9198, "step": 9570 }, { "epoch": 0.09, "learning_rate": 1.5942888186200905e-05, "loss": 0.9455, "step": 9571 }, { "epoch": 0.09, "learning_rate": 1.5942021263860113e-05, "loss": 0.8759, "step": 9572 }, { "epoch": 0.09, "learning_rate": 1.5941154272483255e-05, "loss": 0.9089, "step": 9573 }, { "epoch": 0.09, "learning_rate": 1.594028721208041e-05, "loss": 0.9215, "step": 9574 }, { "epoch": 0.09, "learning_rate": 1.593942008266164e-05, "loss": 0.9023, "step": 9575 }, { "epoch": 0.09, "learning_rate": 1.5938552884237032e-05, "loss": 0.9289, "step": 9576 }, { "epoch": 0.09, "learning_rate": 1.5937685616816653e-05, "loss": 0.9052, "step": 9577 }, { "epoch": 0.09, "learning_rate": 1.5936818280410585e-05, "loss": 0.931, "step": 9578 }, { "epoch": 0.09, "learning_rate": 1.5935950875028902e-05, "loss": 0.8909, "step": 9579 }, { "epoch": 0.09, "learning_rate": 1.5935083400681678e-05, "loss": 0.8679, "step": 9580 }, { "epoch": 0.09, "learning_rate": 1.5934215857379002e-05, "loss": 0.9051, "step": 9581 }, { "epoch": 0.09, "learning_rate": 1.5933348245130945e-05, "loss": 0.8952, "step": 9582 }, { "epoch": 0.09, "learning_rate": 1.5932480563947586e-05, "loss": 0.9184, "step": 9583 }, { "epoch": 0.09, "learning_rate": 1.593161281383901e-05, "loss": 0.9281, "step": 9584 }, { "epoch": 0.09, "learning_rate": 1.59307449948153e-05, "loss": 0.9486, "step": 9585 }, { "epoch": 0.09, "learning_rate": 1.5929877106886534e-05, "loss": 0.8583, "step": 9586 }, { "epoch": 0.09, "learning_rate": 1.59290091500628e-05, "loss": 0.951, "step": 9587 }, { "epoch": 0.09, "learning_rate": 1.592814112435418e-05, "loss": 0.9109, "step": 9588 }, { "epoch": 0.09, "learning_rate": 1.5927273029770757e-05, "loss": 0.9255, "step": 9589 }, { "epoch": 0.09, "learning_rate": 1.592640486632262e-05, "loss": 0.9138, "step": 9590 }, { "epoch": 0.09, "learning_rate": 1.5925536634019855e-05, "loss": 0.9035, "step": 9591 }, { "epoch": 0.09, "learning_rate": 1.5924668332872547e-05, "loss": 0.9082, "step": 9592 }, { "epoch": 0.09, "learning_rate": 1.5923799962890787e-05, "loss": 0.9336, "step": 9593 }, { "epoch": 0.09, "learning_rate": 1.592293152408466e-05, "loss": 0.9185, "step": 9594 }, { "epoch": 0.09, "learning_rate": 1.592206301646426e-05, "loss": 0.9646, "step": 9595 }, { "epoch": 0.09, "learning_rate": 1.5921194440039677e-05, "loss": 0.9133, "step": 9596 }, { "epoch": 0.09, "learning_rate": 1.5920325794821002e-05, "loss": 0.9316, "step": 9597 }, { "epoch": 0.09, "learning_rate": 1.5919457080818322e-05, "loss": 0.8981, "step": 9598 }, { "epoch": 0.09, "learning_rate": 1.591858829804174e-05, "loss": 0.9182, "step": 9599 }, { "epoch": 0.09, "learning_rate": 1.5917719446501343e-05, "loss": 0.8781, "step": 9600 }, { "epoch": 0.09, "learning_rate": 1.5916850526207224e-05, "loss": 0.8939, "step": 9601 }, { "epoch": 0.09, "learning_rate": 1.5915981537169484e-05, "loss": 0.9019, "step": 9602 }, { "epoch": 0.09, "learning_rate": 1.5915112479398218e-05, "loss": 0.9408, "step": 9603 }, { "epoch": 0.09, "learning_rate": 1.591424335290352e-05, "loss": 0.895, "step": 9604 }, { "epoch": 0.09, "learning_rate": 1.591337415769549e-05, "loss": 0.8979, "step": 9605 }, { "epoch": 0.09, "learning_rate": 1.591250489378422e-05, "loss": 0.8934, "step": 9606 }, { "epoch": 0.09, "learning_rate": 1.591163556117982e-05, "loss": 0.9423, "step": 9607 }, { "epoch": 0.09, "learning_rate": 1.5910766159892384e-05, "loss": 0.949, "step": 9608 }, { "epoch": 0.09, "learning_rate": 1.5909896689932012e-05, "loss": 0.8845, "step": 9609 }, { "epoch": 0.09, "learning_rate": 1.5909027151308813e-05, "loss": 0.9152, "step": 9610 }, { "epoch": 0.09, "learning_rate": 1.590815754403288e-05, "loss": 0.9037, "step": 9611 }, { "epoch": 0.09, "learning_rate": 1.5907287868114324e-05, "loss": 0.9429, "step": 9612 }, { "epoch": 0.09, "learning_rate": 1.5906418123563242e-05, "loss": 0.9291, "step": 9613 }, { "epoch": 0.09, "learning_rate": 1.590554831038975e-05, "loss": 0.882, "step": 9614 }, { "epoch": 0.09, "learning_rate": 1.5904678428603937e-05, "loss": 0.9089, "step": 9615 }, { "epoch": 0.09, "learning_rate": 1.5903808478215924e-05, "loss": 0.9197, "step": 9616 }, { "epoch": 0.09, "learning_rate": 1.590293845923581e-05, "loss": 0.8725, "step": 9617 }, { "epoch": 0.09, "learning_rate": 1.5902068371673714e-05, "loss": 0.908, "step": 9618 }, { "epoch": 0.09, "learning_rate": 1.590119821553973e-05, "loss": 0.9388, "step": 9619 }, { "epoch": 0.09, "learning_rate": 1.5900327990843978e-05, "loss": 0.8884, "step": 9620 }, { "epoch": 0.09, "learning_rate": 1.5899457697596564e-05, "loss": 0.9067, "step": 9621 }, { "epoch": 0.09, "learning_rate": 1.5898587335807603e-05, "loss": 0.9148, "step": 9622 }, { "epoch": 0.09, "learning_rate": 1.5897716905487205e-05, "loss": 0.9015, "step": 9623 }, { "epoch": 0.09, "learning_rate": 1.5896846406645484e-05, "loss": 0.8978, "step": 9624 }, { "epoch": 0.09, "learning_rate": 1.589597583929255e-05, "loss": 0.9317, "step": 9625 }, { "epoch": 0.09, "learning_rate": 1.589510520343852e-05, "loss": 0.905, "step": 9626 }, { "epoch": 0.09, "learning_rate": 1.5894234499093514e-05, "loss": 0.899, "step": 9627 }, { "epoch": 0.09, "learning_rate": 1.589336372626764e-05, "loss": 0.9171, "step": 9628 }, { "epoch": 0.09, "learning_rate": 1.5892492884971016e-05, "loss": 0.8856, "step": 9629 }, { "epoch": 0.09, "learning_rate": 1.589162197521377e-05, "loss": 0.8411, "step": 9630 }, { "epoch": 0.09, "learning_rate": 1.5890750997006002e-05, "loss": 0.9816, "step": 9631 }, { "epoch": 0.09, "learning_rate": 1.588987995035785e-05, "loss": 0.9226, "step": 9632 }, { "epoch": 0.09, "learning_rate": 1.588900883527942e-05, "loss": 0.9197, "step": 9633 }, { "epoch": 0.09, "learning_rate": 1.5888137651780847e-05, "loss": 0.8954, "step": 9634 }, { "epoch": 0.09, "learning_rate": 1.5887266399872237e-05, "loss": 0.914, "step": 9635 }, { "epoch": 0.09, "learning_rate": 1.5886395079563723e-05, "loss": 0.8911, "step": 9636 }, { "epoch": 0.09, "learning_rate": 1.5885523690865424e-05, "loss": 0.9079, "step": 9637 }, { "epoch": 0.09, "learning_rate": 1.5884652233787466e-05, "loss": 0.8725, "step": 9638 }, { "epoch": 0.09, "learning_rate": 1.5883780708339976e-05, "loss": 0.94, "step": 9639 }, { "epoch": 0.09, "learning_rate": 1.588290911453307e-05, "loss": 0.9066, "step": 9640 }, { "epoch": 0.09, "learning_rate": 1.5882037452376888e-05, "loss": 0.8753, "step": 9641 }, { "epoch": 0.09, "learning_rate": 1.5881165721881543e-05, "loss": 0.8873, "step": 9642 }, { "epoch": 0.09, "learning_rate": 1.5880293923057175e-05, "loss": 0.9592, "step": 9643 }, { "epoch": 0.09, "learning_rate": 1.5879422055913908e-05, "loss": 0.9064, "step": 9644 }, { "epoch": 0.09, "learning_rate": 1.587855012046187e-05, "loss": 0.9095, "step": 9645 }, { "epoch": 0.09, "learning_rate": 1.587767811671119e-05, "loss": 0.9011, "step": 9646 }, { "epoch": 0.09, "learning_rate": 1.587680604467201e-05, "loss": 0.8756, "step": 9647 }, { "epoch": 0.09, "learning_rate": 1.587593390435445e-05, "loss": 0.9038, "step": 9648 }, { "epoch": 0.09, "learning_rate": 1.5875061695768645e-05, "loss": 0.8922, "step": 9649 }, { "epoch": 0.09, "learning_rate": 1.5874189418924736e-05, "loss": 0.9418, "step": 9650 }, { "epoch": 0.09, "learning_rate": 1.587331707383285e-05, "loss": 0.9004, "step": 9651 }, { "epoch": 0.09, "learning_rate": 1.587244466050312e-05, "loss": 0.9224, "step": 9652 }, { "epoch": 0.09, "learning_rate": 1.5871572178945687e-05, "loss": 0.9217, "step": 9653 }, { "epoch": 0.09, "learning_rate": 1.587069962917069e-05, "loss": 0.8807, "step": 9654 }, { "epoch": 0.09, "learning_rate": 1.586982701118826e-05, "loss": 0.9565, "step": 9655 }, { "epoch": 0.09, "learning_rate": 1.5868954325008542e-05, "loss": 0.9062, "step": 9656 }, { "epoch": 0.09, "learning_rate": 1.5868081570641668e-05, "loss": 0.9397, "step": 9657 }, { "epoch": 0.09, "learning_rate": 1.5867208748097783e-05, "loss": 0.8878, "step": 9658 }, { "epoch": 0.09, "learning_rate": 1.5866335857387027e-05, "loss": 0.9076, "step": 9659 }, { "epoch": 0.09, "learning_rate": 1.5865462898519542e-05, "loss": 0.9119, "step": 9660 }, { "epoch": 0.09, "learning_rate": 1.5864589871505468e-05, "loss": 0.9504, "step": 9661 }, { "epoch": 0.09, "learning_rate": 1.586371677635495e-05, "loss": 0.9601, "step": 9662 }, { "epoch": 0.09, "learning_rate": 1.586284361307813e-05, "loss": 0.9233, "step": 9663 }, { "epoch": 0.09, "learning_rate": 1.586197038168515e-05, "loss": 0.895, "step": 9664 }, { "epoch": 0.09, "learning_rate": 1.586109708218616e-05, "loss": 0.9188, "step": 9665 }, { "epoch": 0.09, "learning_rate": 1.586022371459131e-05, "loss": 0.9044, "step": 9666 }, { "epoch": 0.09, "learning_rate": 1.585935027891074e-05, "loss": 0.921, "step": 9667 }, { "epoch": 0.09, "learning_rate": 1.58584767751546e-05, "loss": 0.9131, "step": 9668 }, { "epoch": 0.09, "learning_rate": 1.585760320333304e-05, "loss": 0.9326, "step": 9669 }, { "epoch": 0.09, "learning_rate": 1.5856729563456206e-05, "loss": 0.9306, "step": 9670 }, { "epoch": 0.09, "learning_rate": 1.5855855855534253e-05, "loss": 0.9161, "step": 9671 }, { "epoch": 0.09, "learning_rate": 1.5854982079577326e-05, "loss": 0.8899, "step": 9672 }, { "epoch": 0.09, "learning_rate": 1.5854108235595583e-05, "loss": 0.8888, "step": 9673 }, { "epoch": 0.09, "learning_rate": 1.5853234323599175e-05, "loss": 0.8635, "step": 9674 }, { "epoch": 0.09, "learning_rate": 1.585236034359825e-05, "loss": 0.9248, "step": 9675 }, { "epoch": 0.09, "learning_rate": 1.5851486295602966e-05, "loss": 0.9136, "step": 9676 }, { "epoch": 0.09, "learning_rate": 1.585061217962348e-05, "loss": 0.912, "step": 9677 }, { "epoch": 0.09, "learning_rate": 1.5849737995669946e-05, "loss": 0.9046, "step": 9678 }, { "epoch": 0.09, "learning_rate": 1.584886374375252e-05, "loss": 0.9342, "step": 9679 }, { "epoch": 0.09, "learning_rate": 1.5847989423881363e-05, "loss": 0.9158, "step": 9680 }, { "epoch": 0.09, "learning_rate": 1.5847115036066626e-05, "loss": 0.9492, "step": 9681 }, { "epoch": 0.09, "learning_rate": 1.584624058031847e-05, "loss": 0.921, "step": 9682 }, { "epoch": 0.09, "learning_rate": 1.5845366056647062e-05, "loss": 0.9458, "step": 9683 }, { "epoch": 0.09, "learning_rate": 1.5844491465062555e-05, "loss": 0.9195, "step": 9684 }, { "epoch": 0.09, "learning_rate": 1.584361680557511e-05, "loss": 0.9049, "step": 9685 }, { "epoch": 0.09, "learning_rate": 1.584274207819489e-05, "loss": 0.8876, "step": 9686 }, { "epoch": 0.09, "learning_rate": 1.5841867282932062e-05, "loss": 0.9555, "step": 9687 }, { "epoch": 0.09, "learning_rate": 1.5840992419796784e-05, "loss": 0.928, "step": 9688 }, { "epoch": 0.09, "learning_rate": 1.5840117488799225e-05, "loss": 0.9187, "step": 9689 }, { "epoch": 0.09, "learning_rate": 1.5839242489949546e-05, "loss": 0.8978, "step": 9690 }, { "epoch": 0.09, "learning_rate": 1.5838367423257916e-05, "loss": 0.8787, "step": 9691 }, { "epoch": 0.09, "learning_rate": 1.58374922887345e-05, "loss": 0.9375, "step": 9692 }, { "epoch": 0.09, "learning_rate": 1.583661708638947e-05, "loss": 0.8994, "step": 9693 }, { "epoch": 0.09, "learning_rate": 1.5835741816232986e-05, "loss": 0.9253, "step": 9694 }, { "epoch": 0.09, "learning_rate": 1.5834866478275226e-05, "loss": 0.9404, "step": 9695 }, { "epoch": 0.09, "learning_rate": 1.5833991072526352e-05, "loss": 0.903, "step": 9696 }, { "epoch": 0.09, "learning_rate": 1.5833115598996544e-05, "loss": 0.9236, "step": 9697 }, { "epoch": 0.09, "learning_rate": 1.5832240057695964e-05, "loss": 0.9365, "step": 9698 }, { "epoch": 0.09, "learning_rate": 1.583136444863479e-05, "loss": 0.9425, "step": 9699 }, { "epoch": 0.09, "learning_rate": 1.5830488771823192e-05, "loss": 0.8886, "step": 9700 }, { "epoch": 0.09, "learning_rate": 1.5829613027271344e-05, "loss": 0.939, "step": 9701 }, { "epoch": 0.09, "learning_rate": 1.582873721498942e-05, "loss": 0.9016, "step": 9702 }, { "epoch": 0.09, "learning_rate": 1.5827861334987603e-05, "loss": 0.9192, "step": 9703 }, { "epoch": 0.09, "learning_rate": 1.5826985387276062e-05, "loss": 0.9118, "step": 9704 }, { "epoch": 0.09, "learning_rate": 1.5826109371864976e-05, "loss": 0.8842, "step": 9705 }, { "epoch": 0.09, "learning_rate": 1.5825233288764522e-05, "loss": 0.9185, "step": 9706 }, { "epoch": 0.09, "learning_rate": 1.5824357137984875e-05, "loss": 0.9144, "step": 9707 }, { "epoch": 0.09, "learning_rate": 1.5823480919536224e-05, "loss": 0.9409, "step": 9708 }, { "epoch": 0.09, "learning_rate": 1.582260463342874e-05, "loss": 0.9209, "step": 9709 }, { "epoch": 0.09, "learning_rate": 1.5821728279672605e-05, "loss": 0.9041, "step": 9710 }, { "epoch": 0.09, "learning_rate": 1.5820851858278006e-05, "loss": 0.9541, "step": 9711 }, { "epoch": 0.09, "learning_rate": 1.5819975369255126e-05, "loss": 0.9407, "step": 9712 }, { "epoch": 0.09, "learning_rate": 1.5819098812614143e-05, "loss": 0.9033, "step": 9713 }, { "epoch": 0.09, "learning_rate": 1.581822218836524e-05, "loss": 0.9101, "step": 9714 }, { "epoch": 0.09, "learning_rate": 1.5817345496518606e-05, "loss": 0.8608, "step": 9715 }, { "epoch": 0.09, "learning_rate": 1.5816468737084428e-05, "loss": 0.908, "step": 9716 }, { "epoch": 0.09, "learning_rate": 1.5815591910072887e-05, "loss": 0.9159, "step": 9717 }, { "epoch": 0.09, "learning_rate": 1.5814715015494175e-05, "loss": 0.9207, "step": 9718 }, { "epoch": 0.09, "learning_rate": 1.5813838053358476e-05, "loss": 0.942, "step": 9719 }, { "epoch": 0.09, "learning_rate": 1.581296102367598e-05, "loss": 0.9039, "step": 9720 }, { "epoch": 0.09, "learning_rate": 1.581208392645688e-05, "loss": 0.8534, "step": 9721 }, { "epoch": 0.09, "learning_rate": 1.5811206761711364e-05, "loss": 0.8938, "step": 9722 }, { "epoch": 0.09, "learning_rate": 1.5810329529449622e-05, "loss": 0.9222, "step": 9723 }, { "epoch": 0.09, "learning_rate": 1.5809452229681847e-05, "loss": 0.9097, "step": 9724 }, { "epoch": 0.09, "learning_rate": 1.5808574862418232e-05, "loss": 0.9206, "step": 9725 }, { "epoch": 0.09, "learning_rate": 1.580769742766897e-05, "loss": 0.9071, "step": 9726 }, { "epoch": 0.09, "learning_rate": 1.5806819925444255e-05, "loss": 0.9094, "step": 9727 }, { "epoch": 0.09, "learning_rate": 1.5805942355754287e-05, "loss": 0.8903, "step": 9728 }, { "epoch": 0.09, "learning_rate": 1.5805064718609255e-05, "loss": 0.9276, "step": 9729 }, { "epoch": 0.09, "learning_rate": 1.5804187014019352e-05, "loss": 0.941, "step": 9730 }, { "epoch": 0.09, "learning_rate": 1.580330924199479e-05, "loss": 0.9086, "step": 9731 }, { "epoch": 0.09, "learning_rate": 1.580243140254575e-05, "loss": 0.9134, "step": 9732 }, { "epoch": 0.09, "learning_rate": 1.5801553495682447e-05, "loss": 0.9038, "step": 9733 }, { "epoch": 0.09, "learning_rate": 1.580067552141507e-05, "loss": 0.8877, "step": 9734 }, { "epoch": 0.09, "learning_rate": 1.5799797479753826e-05, "loss": 0.9586, "step": 9735 }, { "epoch": 0.09, "learning_rate": 1.5798919370708907e-05, "loss": 0.8936, "step": 9736 }, { "epoch": 0.09, "learning_rate": 1.579804119429053e-05, "loss": 0.9009, "step": 9737 }, { "epoch": 0.09, "learning_rate": 1.5797162950508882e-05, "loss": 0.9368, "step": 9738 }, { "epoch": 0.09, "learning_rate": 1.5796284639374177e-05, "loss": 0.9123, "step": 9739 }, { "epoch": 0.09, "learning_rate": 1.579540626089661e-05, "loss": 0.8974, "step": 9740 }, { "epoch": 0.09, "learning_rate": 1.57945278150864e-05, "loss": 0.883, "step": 9741 }, { "epoch": 0.09, "learning_rate": 1.5793649301953745e-05, "loss": 0.9066, "step": 9742 }, { "epoch": 0.09, "learning_rate": 1.579277072150885e-05, "loss": 0.8997, "step": 9743 }, { "epoch": 0.09, "learning_rate": 1.5791892073761928e-05, "loss": 0.9295, "step": 9744 }, { "epoch": 0.09, "learning_rate": 1.5791013358723187e-05, "loss": 0.9173, "step": 9745 }, { "epoch": 0.09, "learning_rate": 1.5790134576402827e-05, "loss": 0.8705, "step": 9746 }, { "epoch": 0.09, "learning_rate": 1.578925572681107e-05, "loss": 0.9141, "step": 9747 }, { "epoch": 0.09, "learning_rate": 1.5788376809958116e-05, "loss": 0.9208, "step": 9748 }, { "epoch": 0.09, "learning_rate": 1.5787497825854185e-05, "loss": 0.9204, "step": 9749 }, { "epoch": 0.09, "learning_rate": 1.5786618774509483e-05, "loss": 0.9329, "step": 9750 }, { "epoch": 0.09, "learning_rate": 1.578573965593423e-05, "loss": 0.9139, "step": 9751 }, { "epoch": 0.09, "learning_rate": 1.5784860470138633e-05, "loss": 0.9011, "step": 9752 }, { "epoch": 0.09, "learning_rate": 1.5783981217132913e-05, "loss": 0.9205, "step": 9753 }, { "epoch": 0.09, "learning_rate": 1.5783101896927278e-05, "loss": 0.8971, "step": 9754 }, { "epoch": 0.09, "learning_rate": 1.5782222509531954e-05, "loss": 0.8907, "step": 9755 }, { "epoch": 0.09, "learning_rate": 1.578134305495715e-05, "loss": 0.9095, "step": 9756 }, { "epoch": 0.09, "learning_rate": 1.5780463533213083e-05, "loss": 0.9146, "step": 9757 }, { "epoch": 0.09, "learning_rate": 1.5779583944309976e-05, "loss": 0.8702, "step": 9758 }, { "epoch": 0.09, "learning_rate": 1.5778704288258047e-05, "loss": 0.9221, "step": 9759 }, { "epoch": 0.09, "learning_rate": 1.5777824565067513e-05, "loss": 0.9215, "step": 9760 }, { "epoch": 0.09, "learning_rate": 1.57769447747486e-05, "loss": 0.9078, "step": 9761 }, { "epoch": 0.09, "learning_rate": 1.577606491731153e-05, "loss": 0.9081, "step": 9762 }, { "epoch": 0.09, "learning_rate": 1.577518499276652e-05, "loss": 0.957, "step": 9763 }, { "epoch": 0.09, "learning_rate": 1.5774305001123795e-05, "loss": 0.9008, "step": 9764 }, { "epoch": 0.09, "learning_rate": 1.5773424942393583e-05, "loss": 0.9237, "step": 9765 }, { "epoch": 0.09, "learning_rate": 1.5772544816586103e-05, "loss": 0.8914, "step": 9766 }, { "epoch": 0.09, "learning_rate": 1.5771664623711584e-05, "loss": 0.9172, "step": 9767 }, { "epoch": 0.09, "learning_rate": 1.5770784363780252e-05, "loss": 0.9049, "step": 9768 }, { "epoch": 0.09, "learning_rate": 1.5769904036802335e-05, "loss": 0.9212, "step": 9769 }, { "epoch": 0.09, "learning_rate": 1.576902364278806e-05, "loss": 0.8814, "step": 9770 }, { "epoch": 0.09, "learning_rate": 1.576814318174765e-05, "loss": 0.9293, "step": 9771 }, { "epoch": 0.09, "learning_rate": 1.5767262653691348e-05, "loss": 0.9032, "step": 9772 }, { "epoch": 0.09, "learning_rate": 1.576638205862937e-05, "loss": 0.914, "step": 9773 }, { "epoch": 0.09, "learning_rate": 1.5765501396571955e-05, "loss": 0.8952, "step": 9774 }, { "epoch": 0.09, "learning_rate": 1.576462066752933e-05, "loss": 0.9151, "step": 9775 }, { "epoch": 0.09, "learning_rate": 1.576373987151173e-05, "loss": 0.8773, "step": 9776 }, { "epoch": 0.09, "learning_rate": 1.5762859008529394e-05, "loss": 0.8876, "step": 9777 }, { "epoch": 0.09, "learning_rate": 1.5761978078592547e-05, "loss": 0.9173, "step": 9778 }, { "epoch": 0.09, "learning_rate": 1.576109708171143e-05, "loss": 0.9098, "step": 9779 }, { "epoch": 0.09, "learning_rate": 1.5760216017896274e-05, "loss": 0.9207, "step": 9780 }, { "epoch": 0.09, "learning_rate": 1.5759334887157318e-05, "loss": 0.9277, "step": 9781 }, { "epoch": 0.09, "learning_rate": 1.57584536895048e-05, "loss": 0.8937, "step": 9782 }, { "epoch": 0.09, "learning_rate": 1.575757242494896e-05, "loss": 0.8934, "step": 9783 }, { "epoch": 0.09, "learning_rate": 1.5756691093500026e-05, "loss": 0.8957, "step": 9784 }, { "epoch": 0.09, "learning_rate": 1.575580969516825e-05, "loss": 0.9229, "step": 9785 }, { "epoch": 0.09, "learning_rate": 1.575492822996387e-05, "loss": 0.894, "step": 9786 }, { "epoch": 0.09, "learning_rate": 1.5754046697897118e-05, "loss": 0.909, "step": 9787 }, { "epoch": 0.09, "learning_rate": 1.575316509897825e-05, "loss": 0.8712, "step": 9788 }, { "epoch": 0.09, "learning_rate": 1.5752283433217493e-05, "loss": 0.9245, "step": 9789 }, { "epoch": 0.09, "learning_rate": 1.5751401700625107e-05, "loss": 0.8886, "step": 9790 }, { "epoch": 0.09, "learning_rate": 1.5750519901211322e-05, "loss": 0.9503, "step": 9791 }, { "epoch": 0.09, "learning_rate": 1.574963803498639e-05, "loss": 0.8956, "step": 9792 }, { "epoch": 0.09, "learning_rate": 1.574875610196056e-05, "loss": 0.9177, "step": 9793 }, { "epoch": 0.09, "learning_rate": 1.5747874102144073e-05, "loss": 0.9162, "step": 9794 }, { "epoch": 0.09, "learning_rate": 1.5746992035547176e-05, "loss": 0.9239, "step": 9795 }, { "epoch": 0.09, "learning_rate": 1.5746109902180114e-05, "loss": 0.8967, "step": 9796 }, { "epoch": 0.09, "learning_rate": 1.5745227702053148e-05, "loss": 0.8576, "step": 9797 }, { "epoch": 0.09, "learning_rate": 1.5744345435176517e-05, "loss": 0.8983, "step": 9798 }, { "epoch": 0.09, "learning_rate": 1.5743463101560475e-05, "loss": 0.9325, "step": 9799 }, { "epoch": 0.09, "learning_rate": 1.5742580701215273e-05, "loss": 0.9307, "step": 9800 }, { "epoch": 0.09, "learning_rate": 1.5741698234151162e-05, "loss": 0.8958, "step": 9801 }, { "epoch": 0.09, "learning_rate": 1.5740815700378392e-05, "loss": 0.9157, "step": 9802 }, { "epoch": 0.09, "learning_rate": 1.5739933099907226e-05, "loss": 0.9441, "step": 9803 }, { "epoch": 0.09, "learning_rate": 1.573905043274791e-05, "loss": 0.8972, "step": 9804 }, { "epoch": 0.09, "learning_rate": 1.5738167698910704e-05, "loss": 0.9257, "step": 9805 }, { "epoch": 0.09, "learning_rate": 1.5737284898405857e-05, "loss": 0.9208, "step": 9806 }, { "epoch": 0.09, "learning_rate": 1.5736402031243633e-05, "loss": 0.957, "step": 9807 }, { "epoch": 0.09, "learning_rate": 1.5735519097434285e-05, "loss": 0.8785, "step": 9808 }, { "epoch": 0.09, "learning_rate": 1.5734636096988073e-05, "loss": 0.9421, "step": 9809 }, { "epoch": 0.09, "learning_rate": 1.5733753029915252e-05, "loss": 0.8804, "step": 9810 }, { "epoch": 0.09, "learning_rate": 1.573286989622609e-05, "loss": 0.864, "step": 9811 }, { "epoch": 0.09, "learning_rate": 1.5731986695930844e-05, "loss": 0.9162, "step": 9812 }, { "epoch": 0.09, "learning_rate": 1.573110342903977e-05, "loss": 0.926, "step": 9813 }, { "epoch": 0.09, "learning_rate": 1.5730220095563137e-05, "loss": 0.8923, "step": 9814 }, { "epoch": 0.09, "learning_rate": 1.5729336695511198e-05, "loss": 0.9219, "step": 9815 }, { "epoch": 0.09, "learning_rate": 1.572845322889423e-05, "loss": 0.9157, "step": 9816 }, { "epoch": 0.09, "learning_rate": 1.572756969572249e-05, "loss": 0.8957, "step": 9817 }, { "epoch": 0.09, "learning_rate": 1.572668609600624e-05, "loss": 0.9006, "step": 9818 }, { "epoch": 0.09, "learning_rate": 1.5725802429755752e-05, "loss": 0.9363, "step": 9819 }, { "epoch": 0.09, "learning_rate": 1.572491869698129e-05, "loss": 0.9109, "step": 9820 }, { "epoch": 0.09, "learning_rate": 1.5724034897693124e-05, "loss": 0.8983, "step": 9821 }, { "epoch": 0.09, "learning_rate": 1.572315103190152e-05, "loss": 0.9308, "step": 9822 }, { "epoch": 0.09, "learning_rate": 1.5722267099616746e-05, "loss": 0.9217, "step": 9823 }, { "epoch": 0.09, "learning_rate": 1.572138310084907e-05, "loss": 0.8688, "step": 9824 }, { "epoch": 0.09, "learning_rate": 1.5720499035608767e-05, "loss": 0.8958, "step": 9825 }, { "epoch": 0.09, "learning_rate": 1.571961490390611e-05, "loss": 0.8781, "step": 9826 }, { "epoch": 0.09, "learning_rate": 1.5718730705751363e-05, "loss": 0.9383, "step": 9827 }, { "epoch": 0.09, "learning_rate": 1.5717846441154806e-05, "loss": 0.9176, "step": 9828 }, { "epoch": 0.09, "learning_rate": 1.5716962110126712e-05, "loss": 0.8908, "step": 9829 }, { "epoch": 0.09, "learning_rate": 1.5716077712677354e-05, "loss": 0.902, "step": 9830 }, { "epoch": 0.09, "learning_rate": 1.5715193248817e-05, "loss": 0.9472, "step": 9831 }, { "epoch": 0.09, "learning_rate": 1.5714308718555943e-05, "loss": 0.8578, "step": 9832 }, { "epoch": 0.09, "learning_rate": 1.5713424121904443e-05, "loss": 0.9225, "step": 9833 }, { "epoch": 0.09, "learning_rate": 1.5712539458872783e-05, "loss": 0.8799, "step": 9834 }, { "epoch": 0.09, "learning_rate": 1.5711654729471247e-05, "loss": 0.9259, "step": 9835 }, { "epoch": 0.09, "learning_rate": 1.5710769933710107e-05, "loss": 0.9198, "step": 9836 }, { "epoch": 0.09, "learning_rate": 1.5709885071599647e-05, "loss": 0.8796, "step": 9837 }, { "epoch": 0.09, "learning_rate": 1.5709000143150145e-05, "loss": 0.8804, "step": 9838 }, { "epoch": 0.09, "learning_rate": 1.5708115148371882e-05, "loss": 0.9301, "step": 9839 }, { "epoch": 0.09, "learning_rate": 1.570723008727514e-05, "loss": 0.8847, "step": 9840 }, { "epoch": 0.09, "learning_rate": 1.5706344959870207e-05, "loss": 0.9275, "step": 9841 }, { "epoch": 0.09, "learning_rate": 1.570545976616736e-05, "loss": 0.8951, "step": 9842 }, { "epoch": 0.09, "learning_rate": 1.570457450617689e-05, "loss": 0.922, "step": 9843 }, { "epoch": 0.09, "learning_rate": 1.5703689179909073e-05, "loss": 0.9006, "step": 9844 }, { "epoch": 0.09, "learning_rate": 1.57028037873742e-05, "loss": 0.8863, "step": 9845 }, { "epoch": 0.09, "learning_rate": 1.570191832858256e-05, "loss": 0.8864, "step": 9846 }, { "epoch": 0.09, "learning_rate": 1.570103280354444e-05, "loss": 0.9336, "step": 9847 }, { "epoch": 0.09, "learning_rate": 1.5700147212270127e-05, "loss": 0.9027, "step": 9848 }, { "epoch": 0.09, "learning_rate": 1.5699261554769908e-05, "loss": 0.9205, "step": 9849 }, { "epoch": 0.09, "learning_rate": 1.5698375831054078e-05, "loss": 0.8496, "step": 9850 }, { "epoch": 0.09, "learning_rate": 1.5697490041132922e-05, "loss": 0.8906, "step": 9851 }, { "epoch": 0.09, "learning_rate": 1.569660418501673e-05, "loss": 0.8849, "step": 9852 }, { "epoch": 0.09, "learning_rate": 1.5695718262715805e-05, "loss": 0.8748, "step": 9853 }, { "epoch": 0.09, "learning_rate": 1.5694832274240427e-05, "loss": 0.8545, "step": 9854 }, { "epoch": 0.1, "learning_rate": 1.5693946219600896e-05, "loss": 0.9115, "step": 9855 }, { "epoch": 0.1, "learning_rate": 1.569306009880751e-05, "loss": 0.9244, "step": 9856 }, { "epoch": 0.1, "learning_rate": 1.5692173911870556e-05, "loss": 0.9335, "step": 9857 }, { "epoch": 0.1, "learning_rate": 1.5691287658800337e-05, "loss": 0.8889, "step": 9858 }, { "epoch": 0.1, "learning_rate": 1.5690401339607144e-05, "loss": 0.8964, "step": 9859 }, { "epoch": 0.1, "learning_rate": 1.568951495430128e-05, "loss": 0.8963, "step": 9860 }, { "epoch": 0.1, "learning_rate": 1.5688628502893036e-05, "loss": 0.9116, "step": 9861 }, { "epoch": 0.1, "learning_rate": 1.568774198539272e-05, "loss": 0.9008, "step": 9862 }, { "epoch": 0.1, "learning_rate": 1.5686855401810625e-05, "loss": 0.9029, "step": 9863 }, { "epoch": 0.1, "learning_rate": 1.5685968752157052e-05, "loss": 0.8996, "step": 9864 }, { "epoch": 0.1, "learning_rate": 1.5685082036442306e-05, "loss": 0.8723, "step": 9865 }, { "epoch": 0.1, "learning_rate": 1.568419525467669e-05, "loss": 0.8914, "step": 9866 }, { "epoch": 0.1, "learning_rate": 1.5683308406870502e-05, "loss": 0.9279, "step": 9867 }, { "epoch": 0.1, "learning_rate": 1.5682421493034045e-05, "loss": 0.9008, "step": 9868 }, { "epoch": 0.1, "learning_rate": 1.5681534513177634e-05, "loss": 0.9213, "step": 9869 }, { "epoch": 0.1, "learning_rate": 1.568064746731156e-05, "loss": 0.8921, "step": 9870 }, { "epoch": 0.1, "learning_rate": 1.5679760355446137e-05, "loss": 0.9294, "step": 9871 }, { "epoch": 0.1, "learning_rate": 1.5678873177591668e-05, "loss": 0.9011, "step": 9872 }, { "epoch": 0.1, "learning_rate": 1.5677985933758466e-05, "loss": 0.9434, "step": 9873 }, { "epoch": 0.1, "learning_rate": 1.5677098623956836e-05, "loss": 0.8693, "step": 9874 }, { "epoch": 0.1, "learning_rate": 1.5676211248197083e-05, "loss": 0.8965, "step": 9875 }, { "epoch": 0.1, "learning_rate": 1.5675323806489524e-05, "loss": 0.8802, "step": 9876 }, { "epoch": 0.1, "learning_rate": 1.5674436298844468e-05, "loss": 0.917, "step": 9877 }, { "epoch": 0.1, "learning_rate": 1.567354872527222e-05, "loss": 0.9136, "step": 9878 }, { "epoch": 0.1, "learning_rate": 1.56726610857831e-05, "loss": 0.9094, "step": 9879 }, { "epoch": 0.1, "learning_rate": 1.5671773380387413e-05, "loss": 0.9269, "step": 9880 }, { "epoch": 0.1, "learning_rate": 1.5670885609095482e-05, "loss": 0.9397, "step": 9881 }, { "epoch": 0.1, "learning_rate": 1.5669997771917614e-05, "loss": 0.8975, "step": 9882 }, { "epoch": 0.1, "learning_rate": 1.5669109868864127e-05, "loss": 0.9295, "step": 9883 }, { "epoch": 0.1, "learning_rate": 1.5668221899945338e-05, "loss": 0.9074, "step": 9884 }, { "epoch": 0.1, "learning_rate": 1.566733386517156e-05, "loss": 0.919, "step": 9885 }, { "epoch": 0.1, "learning_rate": 1.566644576455311e-05, "loss": 0.876, "step": 9886 }, { "epoch": 0.1, "learning_rate": 1.5665557598100315e-05, "loss": 0.9411, "step": 9887 }, { "epoch": 0.1, "learning_rate": 1.5664669365823482e-05, "loss": 0.9357, "step": 9888 }, { "epoch": 0.1, "learning_rate": 1.566378106773294e-05, "loss": 0.885, "step": 9889 }, { "epoch": 0.1, "learning_rate": 1.5662892703839006e-05, "loss": 0.8997, "step": 9890 }, { "epoch": 0.1, "learning_rate": 1.5662004274152e-05, "loss": 0.9558, "step": 9891 }, { "epoch": 0.1, "learning_rate": 1.5661115778682246e-05, "loss": 0.906, "step": 9892 }, { "epoch": 0.1, "learning_rate": 1.5660227217440063e-05, "loss": 0.8793, "step": 9893 }, { "epoch": 0.1, "learning_rate": 1.565933859043578e-05, "loss": 0.8727, "step": 9894 }, { "epoch": 0.1, "learning_rate": 1.565844989767972e-05, "loss": 0.9068, "step": 9895 }, { "epoch": 0.1, "learning_rate": 1.5657561139182205e-05, "loss": 0.91, "step": 9896 }, { "epoch": 0.1, "learning_rate": 1.5656672314953564e-05, "loss": 0.8865, "step": 9897 }, { "epoch": 0.1, "learning_rate": 1.565578342500412e-05, "loss": 0.8667, "step": 9898 }, { "epoch": 0.1, "learning_rate": 1.5654894469344207e-05, "loss": 0.8672, "step": 9899 }, { "epoch": 0.1, "learning_rate": 1.5654005447984145e-05, "loss": 0.8893, "step": 9900 }, { "epoch": 0.1, "learning_rate": 1.565311636093427e-05, "loss": 0.8908, "step": 9901 }, { "epoch": 0.1, "learning_rate": 1.5652227208204906e-05, "loss": 0.8963, "step": 9902 }, { "epoch": 0.1, "learning_rate": 1.565133798980639e-05, "loss": 0.9301, "step": 9903 }, { "epoch": 0.1, "learning_rate": 1.5650448705749047e-05, "loss": 0.8817, "step": 9904 }, { "epoch": 0.1, "learning_rate": 1.5649559356043206e-05, "loss": 0.9106, "step": 9905 }, { "epoch": 0.1, "learning_rate": 1.564866994069921e-05, "loss": 0.9291, "step": 9906 }, { "epoch": 0.1, "learning_rate": 1.5647780459727388e-05, "loss": 0.9085, "step": 9907 }, { "epoch": 0.1, "learning_rate": 1.5646890913138072e-05, "loss": 0.9284, "step": 9908 }, { "epoch": 0.1, "learning_rate": 1.56460013009416e-05, "loss": 0.8906, "step": 9909 }, { "epoch": 0.1, "learning_rate": 1.5645111623148303e-05, "loss": 0.9016, "step": 9910 }, { "epoch": 0.1, "learning_rate": 1.564422187976852e-05, "loss": 0.948, "step": 9911 }, { "epoch": 0.1, "learning_rate": 1.5643332070812592e-05, "loss": 0.9162, "step": 9912 }, { "epoch": 0.1, "learning_rate": 1.5642442196290856e-05, "loss": 0.9137, "step": 9913 }, { "epoch": 0.1, "learning_rate": 1.5641552256213644e-05, "loss": 0.8755, "step": 9914 }, { "epoch": 0.1, "learning_rate": 1.5640662250591306e-05, "loss": 0.9051, "step": 9915 }, { "epoch": 0.1, "learning_rate": 1.563977217943417e-05, "loss": 0.924, "step": 9916 }, { "epoch": 0.1, "learning_rate": 1.563888204275259e-05, "loss": 0.9152, "step": 9917 }, { "epoch": 0.1, "learning_rate": 1.56379918405569e-05, "loss": 0.8879, "step": 9918 }, { "epoch": 0.1, "learning_rate": 1.563710157285744e-05, "loss": 0.938, "step": 9919 }, { "epoch": 0.1, "learning_rate": 1.5636211239664565e-05, "loss": 0.9399, "step": 9920 }, { "epoch": 0.1, "learning_rate": 1.563532084098861e-05, "loss": 0.9012, "step": 9921 }, { "epoch": 0.1, "learning_rate": 1.563443037683992e-05, "loss": 0.9058, "step": 9922 }, { "epoch": 0.1, "learning_rate": 1.563353984722884e-05, "loss": 0.9333, "step": 9923 }, { "epoch": 0.1, "learning_rate": 1.5632649252165727e-05, "loss": 0.9155, "step": 9924 }, { "epoch": 0.1, "learning_rate": 1.563175859166091e-05, "loss": 0.9315, "step": 9925 }, { "epoch": 0.1, "learning_rate": 1.5630867865724755e-05, "loss": 0.9011, "step": 9926 }, { "epoch": 0.1, "learning_rate": 1.56299770743676e-05, "loss": 0.8873, "step": 9927 }, { "epoch": 0.1, "learning_rate": 1.5629086217599794e-05, "loss": 0.8522, "step": 9928 }, { "epoch": 0.1, "learning_rate": 1.5628195295431696e-05, "loss": 0.8982, "step": 9929 }, { "epoch": 0.1, "learning_rate": 1.5627304307873647e-05, "loss": 0.8969, "step": 9930 }, { "epoch": 0.1, "learning_rate": 1.5626413254936004e-05, "loss": 0.9225, "step": 9931 }, { "epoch": 0.1, "learning_rate": 1.562552213662912e-05, "loss": 0.8913, "step": 9932 }, { "epoch": 0.1, "learning_rate": 1.5624630952963347e-05, "loss": 0.9089, "step": 9933 }, { "epoch": 0.1, "learning_rate": 1.5623739703949036e-05, "loss": 0.9088, "step": 9934 }, { "epoch": 0.1, "learning_rate": 1.562284838959655e-05, "loss": 0.9295, "step": 9935 }, { "epoch": 0.1, "learning_rate": 1.5621957009916238e-05, "loss": 0.9013, "step": 9936 }, { "epoch": 0.1, "learning_rate": 1.5621065564918452e-05, "loss": 0.9273, "step": 9937 }, { "epoch": 0.1, "learning_rate": 1.5620174054613563e-05, "loss": 0.9102, "step": 9938 }, { "epoch": 0.1, "learning_rate": 1.5619282479011913e-05, "loss": 0.9478, "step": 9939 }, { "epoch": 0.1, "learning_rate": 1.561839083812387e-05, "loss": 0.902, "step": 9940 }, { "epoch": 0.1, "learning_rate": 1.5617499131959793e-05, "loss": 0.8856, "step": 9941 }, { "epoch": 0.1, "learning_rate": 1.561660736053004e-05, "loss": 0.894, "step": 9942 }, { "epoch": 0.1, "learning_rate": 1.5615715523844973e-05, "loss": 0.8968, "step": 9943 }, { "epoch": 0.1, "learning_rate": 1.5614823621914952e-05, "loss": 0.8927, "step": 9944 }, { "epoch": 0.1, "learning_rate": 1.561393165475034e-05, "loss": 0.9208, "step": 9945 }, { "epoch": 0.1, "learning_rate": 1.56130396223615e-05, "loss": 0.9085, "step": 9946 }, { "epoch": 0.1, "learning_rate": 1.5612147524758796e-05, "loss": 0.911, "step": 9947 }, { "epoch": 0.1, "learning_rate": 1.5611255361952595e-05, "loss": 0.9358, "step": 9948 }, { "epoch": 0.1, "learning_rate": 1.561036313395326e-05, "loss": 0.8811, "step": 9949 }, { "epoch": 0.1, "learning_rate": 1.5609470840771156e-05, "loss": 0.9172, "step": 9950 }, { "epoch": 0.1, "learning_rate": 1.560857848241665e-05, "loss": 0.8978, "step": 9951 }, { "epoch": 0.1, "learning_rate": 1.5607686058900115e-05, "loss": 0.9121, "step": 9952 }, { "epoch": 0.1, "learning_rate": 1.560679357023191e-05, "loss": 0.9044, "step": 9953 }, { "epoch": 0.1, "learning_rate": 1.5605901016422416e-05, "loss": 0.9001, "step": 9954 }, { "epoch": 0.1, "learning_rate": 1.5605008397481993e-05, "loss": 0.9413, "step": 9955 }, { "epoch": 0.1, "learning_rate": 1.5604115713421014e-05, "loss": 0.9051, "step": 9956 }, { "epoch": 0.1, "learning_rate": 1.5603222964249853e-05, "loss": 0.9412, "step": 9957 }, { "epoch": 0.1, "learning_rate": 1.5602330149978883e-05, "loss": 0.9626, "step": 9958 }, { "epoch": 0.1, "learning_rate": 1.5601437270618472e-05, "loss": 0.9067, "step": 9959 }, { "epoch": 0.1, "learning_rate": 1.5600544326179e-05, "loss": 0.8993, "step": 9960 }, { "epoch": 0.1, "learning_rate": 1.5599651316670835e-05, "loss": 0.9001, "step": 9961 }, { "epoch": 0.1, "learning_rate": 1.5598758242104353e-05, "loss": 0.8981, "step": 9962 }, { "epoch": 0.1, "learning_rate": 1.559786510248994e-05, "loss": 0.9055, "step": 9963 }, { "epoch": 0.1, "learning_rate": 1.5596971897837957e-05, "loss": 0.9309, "step": 9964 }, { "epoch": 0.1, "learning_rate": 1.5596078628158795e-05, "loss": 0.8554, "step": 9965 }, { "epoch": 0.1, "learning_rate": 1.5595185293462827e-05, "loss": 0.9211, "step": 9966 }, { "epoch": 0.1, "learning_rate": 1.559429189376043e-05, "loss": 0.9096, "step": 9967 }, { "epoch": 0.1, "learning_rate": 1.5593398429061985e-05, "loss": 0.9004, "step": 9968 }, { "epoch": 0.1, "learning_rate": 1.5592504899377874e-05, "loss": 0.8851, "step": 9969 }, { "epoch": 0.1, "learning_rate": 1.5591611304718478e-05, "loss": 0.9229, "step": 9970 }, { "epoch": 0.1, "learning_rate": 1.5590717645094177e-05, "loss": 0.9598, "step": 9971 }, { "epoch": 0.1, "learning_rate": 1.5589823920515353e-05, "loss": 0.8874, "step": 9972 }, { "epoch": 0.1, "learning_rate": 1.55889301309924e-05, "loss": 0.9133, "step": 9973 }, { "epoch": 0.1, "learning_rate": 1.5588036276535684e-05, "loss": 0.9308, "step": 9974 }, { "epoch": 0.1, "learning_rate": 1.5587142357155606e-05, "loss": 0.9125, "step": 9975 }, { "epoch": 0.1, "learning_rate": 1.5586248372862543e-05, "loss": 0.9042, "step": 9976 }, { "epoch": 0.1, "learning_rate": 1.5585354323666887e-05, "loss": 0.8948, "step": 9977 }, { "epoch": 0.1, "learning_rate": 1.558446020957902e-05, "loss": 0.9337, "step": 9978 }, { "epoch": 0.1, "learning_rate": 1.5583566030609336e-05, "loss": 0.9204, "step": 9979 }, { "epoch": 0.1, "learning_rate": 1.558267178676822e-05, "loss": 0.9394, "step": 9980 }, { "epoch": 0.1, "learning_rate": 1.5581777478066064e-05, "loss": 0.9374, "step": 9981 }, { "epoch": 0.1, "learning_rate": 1.5580883104513253e-05, "loss": 0.8971, "step": 9982 }, { "epoch": 0.1, "learning_rate": 1.557998866612018e-05, "loss": 0.8841, "step": 9983 }, { "epoch": 0.1, "learning_rate": 1.5579094162897244e-05, "loss": 0.9158, "step": 9984 }, { "epoch": 0.1, "learning_rate": 1.557819959485483e-05, "loss": 0.8709, "step": 9985 }, { "epoch": 0.1, "learning_rate": 1.5577304962003333e-05, "loss": 0.9216, "step": 9986 }, { "epoch": 0.1, "learning_rate": 1.5576410264353148e-05, "loss": 0.9109, "step": 9987 }, { "epoch": 0.1, "learning_rate": 1.557551550191467e-05, "loss": 0.9004, "step": 9988 }, { "epoch": 0.1, "learning_rate": 1.557462067469829e-05, "loss": 0.9374, "step": 9989 }, { "epoch": 0.1, "learning_rate": 1.5573725782714413e-05, "loss": 0.8551, "step": 9990 }, { "epoch": 0.1, "learning_rate": 1.5572830825973428e-05, "loss": 0.9285, "step": 9991 }, { "epoch": 0.1, "learning_rate": 1.557193580448574e-05, "loss": 0.927, "step": 9992 }, { "epoch": 0.1, "learning_rate": 1.5571040718261744e-05, "loss": 0.9222, "step": 9993 }, { "epoch": 0.1, "learning_rate": 1.557014556731183e-05, "loss": 0.8812, "step": 9994 }, { "epoch": 0.1, "learning_rate": 1.556925035164642e-05, "loss": 0.9137, "step": 9995 }, { "epoch": 0.1, "learning_rate": 1.5568355071275896e-05, "loss": 0.9181, "step": 9996 }, { "epoch": 0.1, "learning_rate": 1.5567459726210668e-05, "loss": 0.9047, "step": 9997 }, { "epoch": 0.1, "learning_rate": 1.5566564316461134e-05, "loss": 0.8537, "step": 9998 }, { "epoch": 0.1, "learning_rate": 1.55656688420377e-05, "loss": 0.9322, "step": 9999 }, { "epoch": 0.1, "learning_rate": 1.556477330295077e-05, "loss": 0.8987, "step": 10000 }, { "epoch": 0.1, "learning_rate": 1.5563877699210745e-05, "loss": 0.9173, "step": 10001 }, { "epoch": 0.1, "learning_rate": 1.5562982030828036e-05, "loss": 0.9192, "step": 10002 }, { "epoch": 0.1, "learning_rate": 1.5562086297813043e-05, "loss": 0.9067, "step": 10003 }, { "epoch": 0.1, "learning_rate": 1.5561190500176183e-05, "loss": 0.9233, "step": 10004 }, { "epoch": 0.1, "learning_rate": 1.5560294637927852e-05, "loss": 0.8997, "step": 10005 }, { "epoch": 0.1, "learning_rate": 1.5559398711078464e-05, "loss": 0.8728, "step": 10006 }, { "epoch": 0.1, "learning_rate": 1.5558502719638425e-05, "loss": 0.9134, "step": 10007 }, { "epoch": 0.1, "learning_rate": 1.5557606663618152e-05, "loss": 0.9141, "step": 10008 }, { "epoch": 0.1, "learning_rate": 1.555671054302805e-05, "loss": 0.9002, "step": 10009 }, { "epoch": 0.1, "learning_rate": 1.5555814357878528e-05, "loss": 0.9161, "step": 10010 }, { "epoch": 0.1, "learning_rate": 1.5554918108180004e-05, "loss": 0.8996, "step": 10011 }, { "epoch": 0.1, "learning_rate": 1.555402179394289e-05, "loss": 0.9206, "step": 10012 }, { "epoch": 0.1, "learning_rate": 1.5553125415177594e-05, "loss": 0.973, "step": 10013 }, { "epoch": 0.1, "learning_rate": 1.5552228971894535e-05, "loss": 0.8851, "step": 10014 }, { "epoch": 0.1, "learning_rate": 1.5551332464104126e-05, "loss": 0.962, "step": 10015 }, { "epoch": 0.1, "learning_rate": 1.555043589181679e-05, "loss": 0.8923, "step": 10016 }, { "epoch": 0.1, "learning_rate": 1.5549539255042932e-05, "loss": 0.9382, "step": 10017 }, { "epoch": 0.1, "learning_rate": 1.5548642553792977e-05, "loss": 0.8605, "step": 10018 }, { "epoch": 0.1, "learning_rate": 1.5547745788077343e-05, "loss": 0.9034, "step": 10019 }, { "epoch": 0.1, "learning_rate": 1.5546848957906446e-05, "loss": 0.9392, "step": 10020 }, { "epoch": 0.1, "learning_rate": 1.5545952063290707e-05, "loss": 0.8923, "step": 10021 }, { "epoch": 0.1, "learning_rate": 1.5545055104240548e-05, "loss": 0.8827, "step": 10022 }, { "epoch": 0.1, "learning_rate": 1.554415808076639e-05, "loss": 0.9167, "step": 10023 }, { "epoch": 0.1, "learning_rate": 1.5543260992878646e-05, "loss": 0.9219, "step": 10024 }, { "epoch": 0.1, "learning_rate": 1.5542363840587753e-05, "loss": 0.9113, "step": 10025 }, { "epoch": 0.1, "learning_rate": 1.554146662390412e-05, "loss": 0.9373, "step": 10026 }, { "epoch": 0.1, "learning_rate": 1.5540569342838185e-05, "loss": 0.9286, "step": 10027 }, { "epoch": 0.1, "learning_rate": 1.5539671997400364e-05, "loss": 0.8835, "step": 10028 }, { "epoch": 0.1, "learning_rate": 1.553877458760108e-05, "loss": 0.8979, "step": 10029 }, { "epoch": 0.1, "learning_rate": 1.553787711345077e-05, "loss": 0.8888, "step": 10030 }, { "epoch": 0.1, "learning_rate": 1.5536979574959855e-05, "loss": 0.9799, "step": 10031 }, { "epoch": 0.1, "learning_rate": 1.553608197213876e-05, "loss": 0.8882, "step": 10032 }, { "epoch": 0.1, "learning_rate": 1.5535184304997922e-05, "loss": 0.8944, "step": 10033 }, { "epoch": 0.1, "learning_rate": 1.5534286573547763e-05, "loss": 0.9013, "step": 10034 }, { "epoch": 0.1, "learning_rate": 1.5533388777798714e-05, "loss": 0.9467, "step": 10035 }, { "epoch": 0.1, "learning_rate": 1.5532490917761208e-05, "loss": 0.9313, "step": 10036 }, { "epoch": 0.1, "learning_rate": 1.5531592993445676e-05, "loss": 0.9542, "step": 10037 }, { "epoch": 0.1, "learning_rate": 1.5530695004862545e-05, "loss": 0.8998, "step": 10038 }, { "epoch": 0.1, "learning_rate": 1.5529796952022258e-05, "loss": 0.9257, "step": 10039 }, { "epoch": 0.1, "learning_rate": 1.552889883493524e-05, "loss": 0.9309, "step": 10040 }, { "epoch": 0.1, "learning_rate": 1.5528000653611935e-05, "loss": 0.8983, "step": 10041 }, { "epoch": 0.1, "learning_rate": 1.5527102408062772e-05, "loss": 0.9184, "step": 10042 }, { "epoch": 0.1, "learning_rate": 1.5526204098298182e-05, "loss": 0.9427, "step": 10043 }, { "epoch": 0.1, "learning_rate": 1.5525305724328613e-05, "loss": 0.9549, "step": 10044 }, { "epoch": 0.1, "learning_rate": 1.5524407286164497e-05, "loss": 0.8914, "step": 10045 }, { "epoch": 0.1, "learning_rate": 1.5523508783816273e-05, "loss": 0.8912, "step": 10046 }, { "epoch": 0.1, "learning_rate": 1.5522610217294377e-05, "loss": 0.9095, "step": 10047 }, { "epoch": 0.1, "learning_rate": 1.552171158660925e-05, "loss": 0.9079, "step": 10048 }, { "epoch": 0.1, "learning_rate": 1.552081289177134e-05, "loss": 0.9107, "step": 10049 }, { "epoch": 0.1, "learning_rate": 1.5519914132791075e-05, "loss": 0.8954, "step": 10050 }, { "epoch": 0.1, "learning_rate": 1.5519015309678907e-05, "loss": 0.928, "step": 10051 }, { "epoch": 0.1, "learning_rate": 1.551811642244528e-05, "loss": 0.8765, "step": 10052 }, { "epoch": 0.1, "learning_rate": 1.5517217471100627e-05, "loss": 0.9361, "step": 10053 }, { "epoch": 0.1, "learning_rate": 1.5516318455655402e-05, "loss": 0.8868, "step": 10054 }, { "epoch": 0.1, "learning_rate": 1.5515419376120046e-05, "loss": 0.915, "step": 10055 }, { "epoch": 0.1, "learning_rate": 1.5514520232505004e-05, "loss": 0.9008, "step": 10056 }, { "epoch": 0.1, "learning_rate": 1.551362102482073e-05, "loss": 0.9135, "step": 10057 }, { "epoch": 0.1, "learning_rate": 1.5512721753077663e-05, "loss": 0.9071, "step": 10058 }, { "epoch": 0.1, "learning_rate": 1.551182241728625e-05, "loss": 0.9025, "step": 10059 }, { "epoch": 0.1, "learning_rate": 1.5510923017456946e-05, "loss": 0.9048, "step": 10060 }, { "epoch": 0.1, "learning_rate": 1.5510023553600194e-05, "loss": 0.9044, "step": 10061 }, { "epoch": 0.1, "learning_rate": 1.5509124025726453e-05, "loss": 0.9141, "step": 10062 }, { "epoch": 0.1, "learning_rate": 1.5508224433846163e-05, "loss": 0.9043, "step": 10063 }, { "epoch": 0.1, "learning_rate": 1.5507324777969786e-05, "loss": 0.8956, "step": 10064 }, { "epoch": 0.1, "learning_rate": 1.5506425058107768e-05, "loss": 0.8706, "step": 10065 }, { "epoch": 0.1, "learning_rate": 1.5505525274270567e-05, "loss": 0.9348, "step": 10066 }, { "epoch": 0.1, "learning_rate": 1.550462542646863e-05, "loss": 0.9142, "step": 10067 }, { "epoch": 0.1, "learning_rate": 1.5503725514712415e-05, "loss": 0.8906, "step": 10068 }, { "epoch": 0.1, "learning_rate": 1.5502825539012382e-05, "loss": 0.8918, "step": 10069 }, { "epoch": 0.1, "learning_rate": 1.550192549937898e-05, "loss": 0.868, "step": 10070 }, { "epoch": 0.1, "learning_rate": 1.550102539582267e-05, "loss": 0.916, "step": 10071 }, { "epoch": 0.1, "learning_rate": 1.5500125228353912e-05, "loss": 0.9371, "step": 10072 }, { "epoch": 0.1, "learning_rate": 1.5499224996983157e-05, "loss": 0.9207, "step": 10073 }, { "epoch": 0.1, "learning_rate": 1.5498324701720873e-05, "loss": 0.8326, "step": 10074 }, { "epoch": 0.1, "learning_rate": 1.549742434257751e-05, "loss": 0.8762, "step": 10075 }, { "epoch": 0.1, "learning_rate": 1.5496523919563538e-05, "loss": 0.9028, "step": 10076 }, { "epoch": 0.1, "learning_rate": 1.5495623432689414e-05, "loss": 0.8992, "step": 10077 }, { "epoch": 0.1, "learning_rate": 1.54947228819656e-05, "loss": 0.8903, "step": 10078 }, { "epoch": 0.1, "learning_rate": 1.5493822267402556e-05, "loss": 0.8983, "step": 10079 }, { "epoch": 0.1, "learning_rate": 1.549292158901075e-05, "loss": 0.9026, "step": 10080 }, { "epoch": 0.1, "learning_rate": 1.5492020846800645e-05, "loss": 0.9254, "step": 10081 }, { "epoch": 0.1, "learning_rate": 1.5491120040782708e-05, "loss": 0.9238, "step": 10082 }, { "epoch": 0.1, "learning_rate": 1.5490219170967405e-05, "loss": 0.9083, "step": 10083 }, { "epoch": 0.1, "learning_rate": 1.5489318237365195e-05, "loss": 0.8801, "step": 10084 }, { "epoch": 0.1, "learning_rate": 1.5488417239986557e-05, "loss": 0.8938, "step": 10085 }, { "epoch": 0.1, "learning_rate": 1.5487516178841948e-05, "loss": 0.9043, "step": 10086 }, { "epoch": 0.1, "learning_rate": 1.5486615053941846e-05, "loss": 0.9344, "step": 10087 }, { "epoch": 0.1, "learning_rate": 1.548571386529671e-05, "loss": 0.9114, "step": 10088 }, { "epoch": 0.1, "learning_rate": 1.5484812612917022e-05, "loss": 0.9204, "step": 10089 }, { "epoch": 0.1, "learning_rate": 1.5483911296813246e-05, "loss": 0.9433, "step": 10090 }, { "epoch": 0.1, "learning_rate": 1.5483009916995855e-05, "loss": 0.913, "step": 10091 }, { "epoch": 0.1, "learning_rate": 1.5482108473475324e-05, "loss": 0.8994, "step": 10092 }, { "epoch": 0.1, "learning_rate": 1.5481206966262118e-05, "loss": 0.9665, "step": 10093 }, { "epoch": 0.1, "learning_rate": 1.548030539536672e-05, "loss": 0.886, "step": 10094 }, { "epoch": 0.1, "learning_rate": 1.5479403760799604e-05, "loss": 0.947, "step": 10095 }, { "epoch": 0.1, "learning_rate": 1.5478502062571242e-05, "loss": 0.9111, "step": 10096 }, { "epoch": 0.1, "learning_rate": 1.547760030069211e-05, "loss": 0.9009, "step": 10097 }, { "epoch": 0.1, "learning_rate": 1.5476698475172688e-05, "loss": 0.8762, "step": 10098 }, { "epoch": 0.1, "learning_rate": 1.5475796586023452e-05, "loss": 0.9493, "step": 10099 }, { "epoch": 0.1, "learning_rate": 1.5474894633254876e-05, "loss": 0.8669, "step": 10100 }, { "epoch": 0.1, "learning_rate": 1.547399261687745e-05, "loss": 0.8791, "step": 10101 }, { "epoch": 0.1, "learning_rate": 1.547309053690164e-05, "loss": 0.8792, "step": 10102 }, { "epoch": 0.1, "learning_rate": 1.5472188393337938e-05, "loss": 0.9335, "step": 10103 }, { "epoch": 0.1, "learning_rate": 1.5471286186196823e-05, "loss": 0.9346, "step": 10104 }, { "epoch": 0.1, "learning_rate": 1.547038391548877e-05, "loss": 0.8698, "step": 10105 }, { "epoch": 0.1, "learning_rate": 1.5469481581224274e-05, "loss": 0.9234, "step": 10106 }, { "epoch": 0.1, "learning_rate": 1.5468579183413803e-05, "loss": 0.9297, "step": 10107 }, { "epoch": 0.1, "learning_rate": 1.5467676722067858e-05, "loss": 0.8931, "step": 10108 }, { "epoch": 0.1, "learning_rate": 1.5466774197196908e-05, "loss": 0.9415, "step": 10109 }, { "epoch": 0.1, "learning_rate": 1.5465871608811454e-05, "loss": 0.9084, "step": 10110 }, { "epoch": 0.1, "learning_rate": 1.546496895692197e-05, "loss": 0.8959, "step": 10111 }, { "epoch": 0.1, "learning_rate": 1.5464066241538957e-05, "loss": 0.9184, "step": 10112 }, { "epoch": 0.1, "learning_rate": 1.546316346267289e-05, "loss": 0.8796, "step": 10113 }, { "epoch": 0.1, "learning_rate": 1.546226062033426e-05, "loss": 0.9033, "step": 10114 }, { "epoch": 0.1, "learning_rate": 1.5461357714533562e-05, "loss": 0.9026, "step": 10115 }, { "epoch": 0.1, "learning_rate": 1.5460454745281278e-05, "loss": 0.8691, "step": 10116 }, { "epoch": 0.1, "learning_rate": 1.545955171258791e-05, "loss": 0.9191, "step": 10117 }, { "epoch": 0.1, "learning_rate": 1.5458648616463938e-05, "loss": 0.9252, "step": 10118 }, { "epoch": 0.1, "learning_rate": 1.5457745456919868e-05, "loss": 0.9521, "step": 10119 }, { "epoch": 0.1, "learning_rate": 1.545684223396618e-05, "loss": 0.919, "step": 10120 }, { "epoch": 0.1, "learning_rate": 1.5455938947613373e-05, "loss": 0.926, "step": 10121 }, { "epoch": 0.1, "learning_rate": 1.545503559787194e-05, "loss": 0.843, "step": 10122 }, { "epoch": 0.1, "learning_rate": 1.5454132184752383e-05, "loss": 0.8826, "step": 10123 }, { "epoch": 0.1, "learning_rate": 1.5453228708265193e-05, "loss": 0.9198, "step": 10124 }, { "epoch": 0.1, "learning_rate": 1.5452325168420863e-05, "loss": 0.9219, "step": 10125 }, { "epoch": 0.1, "learning_rate": 1.5451421565229897e-05, "loss": 0.9021, "step": 10126 }, { "epoch": 0.1, "learning_rate": 1.5450517898702794e-05, "loss": 0.9561, "step": 10127 }, { "epoch": 0.1, "learning_rate": 1.544961416885005e-05, "loss": 0.8969, "step": 10128 }, { "epoch": 0.1, "learning_rate": 1.544871037568216e-05, "loss": 0.8814, "step": 10129 }, { "epoch": 0.1, "learning_rate": 1.544780651920963e-05, "loss": 0.8761, "step": 10130 }, { "epoch": 0.1, "learning_rate": 1.5446902599442962e-05, "loss": 0.9367, "step": 10131 }, { "epoch": 0.1, "learning_rate": 1.5445998616392657e-05, "loss": 0.8873, "step": 10132 }, { "epoch": 0.1, "learning_rate": 1.5445094570069215e-05, "loss": 0.8873, "step": 10133 }, { "epoch": 0.1, "learning_rate": 1.5444190460483146e-05, "loss": 0.9182, "step": 10134 }, { "epoch": 0.1, "learning_rate": 1.544328628764495e-05, "loss": 0.927, "step": 10135 }, { "epoch": 0.1, "learning_rate": 1.5442382051565127e-05, "loss": 0.8906, "step": 10136 }, { "epoch": 0.1, "learning_rate": 1.544147775225419e-05, "loss": 0.8875, "step": 10137 }, { "epoch": 0.1, "learning_rate": 1.5440573389722646e-05, "loss": 0.8995, "step": 10138 }, { "epoch": 0.1, "learning_rate": 1.5439668963980994e-05, "loss": 0.8816, "step": 10139 }, { "epoch": 0.1, "learning_rate": 1.543876447503975e-05, "loss": 0.8518, "step": 10140 }, { "epoch": 0.1, "learning_rate": 1.5437859922909418e-05, "loss": 0.8766, "step": 10141 }, { "epoch": 0.1, "learning_rate": 1.543695530760051e-05, "loss": 0.8808, "step": 10142 }, { "epoch": 0.1, "learning_rate": 1.543605062912354e-05, "loss": 0.8837, "step": 10143 }, { "epoch": 0.1, "learning_rate": 1.5435145887489006e-05, "loss": 0.8456, "step": 10144 }, { "epoch": 0.1, "learning_rate": 1.5434241082707432e-05, "loss": 0.9455, "step": 10145 }, { "epoch": 0.1, "learning_rate": 1.5433336214789326e-05, "loss": 0.9143, "step": 10146 }, { "epoch": 0.1, "learning_rate": 1.54324312837452e-05, "loss": 0.9375, "step": 10147 }, { "epoch": 0.1, "learning_rate": 1.5431526289585566e-05, "loss": 0.9238, "step": 10148 }, { "epoch": 0.1, "learning_rate": 1.5430621232320943e-05, "loss": 0.9088, "step": 10149 }, { "epoch": 0.1, "learning_rate": 1.5429716111961844e-05, "loss": 0.8946, "step": 10150 }, { "epoch": 0.1, "learning_rate": 1.542881092851878e-05, "loss": 0.905, "step": 10151 }, { "epoch": 0.1, "learning_rate": 1.542790568200228e-05, "loss": 0.9414, "step": 10152 }, { "epoch": 0.1, "learning_rate": 1.542700037242285e-05, "loss": 0.9048, "step": 10153 }, { "epoch": 0.1, "learning_rate": 1.5426094999791014e-05, "loss": 0.9443, "step": 10154 }, { "epoch": 0.1, "learning_rate": 1.5425189564117288e-05, "loss": 0.9232, "step": 10155 }, { "epoch": 0.11, "learning_rate": 1.5424284065412198e-05, "loss": 0.9231, "step": 10156 }, { "epoch": 0.11, "learning_rate": 1.5423378503686253e-05, "loss": 0.8777, "step": 10157 }, { "epoch": 0.11, "learning_rate": 1.5422472878949982e-05, "loss": 0.885, "step": 10158 }, { "epoch": 0.11, "learning_rate": 1.5421567191213905e-05, "loss": 0.9523, "step": 10159 }, { "epoch": 0.11, "learning_rate": 1.5420661440488544e-05, "loss": 0.8681, "step": 10160 }, { "epoch": 0.11, "learning_rate": 1.5419755626784426e-05, "loss": 0.8483, "step": 10161 }, { "epoch": 0.11, "learning_rate": 1.5418849750112066e-05, "loss": 0.9458, "step": 10162 }, { "epoch": 0.11, "learning_rate": 1.5417943810482e-05, "loss": 0.9424, "step": 10163 }, { "epoch": 0.11, "learning_rate": 1.541703780790474e-05, "loss": 0.8616, "step": 10164 }, { "epoch": 0.11, "learning_rate": 1.5416131742390827e-05, "loss": 0.9123, "step": 10165 }, { "epoch": 0.11, "learning_rate": 1.541522561395078e-05, "loss": 0.9335, "step": 10166 }, { "epoch": 0.11, "learning_rate": 1.541431942259513e-05, "loss": 0.9032, "step": 10167 }, { "epoch": 0.11, "learning_rate": 1.54134131683344e-05, "loss": 0.8931, "step": 10168 }, { "epoch": 0.11, "learning_rate": 1.541250685117912e-05, "loss": 0.953, "step": 10169 }, { "epoch": 0.11, "learning_rate": 1.5411600471139827e-05, "loss": 0.8815, "step": 10170 }, { "epoch": 0.11, "learning_rate": 1.5410694028227046e-05, "loss": 0.9658, "step": 10171 }, { "epoch": 0.11, "learning_rate": 1.5409787522451308e-05, "loss": 0.9164, "step": 10172 }, { "epoch": 0.11, "learning_rate": 1.5408880953823144e-05, "loss": 0.9475, "step": 10173 }, { "epoch": 0.11, "learning_rate": 1.540797432235309e-05, "loss": 0.9028, "step": 10174 }, { "epoch": 0.11, "learning_rate": 1.5407067628051678e-05, "loss": 0.9069, "step": 10175 }, { "epoch": 0.11, "learning_rate": 1.540616087092944e-05, "loss": 0.921, "step": 10176 }, { "epoch": 0.11, "learning_rate": 1.5405254050996917e-05, "loss": 0.9279, "step": 10177 }, { "epoch": 0.11, "learning_rate": 1.540434716826464e-05, "loss": 0.8845, "step": 10178 }, { "epoch": 0.11, "learning_rate": 1.5403440222743148e-05, "loss": 0.9197, "step": 10179 }, { "epoch": 0.11, "learning_rate": 1.5402533214442976e-05, "loss": 0.9122, "step": 10180 }, { "epoch": 0.11, "learning_rate": 1.540162614337466e-05, "loss": 0.9048, "step": 10181 }, { "epoch": 0.11, "learning_rate": 1.5400719009548747e-05, "loss": 0.9101, "step": 10182 }, { "epoch": 0.11, "learning_rate": 1.5399811812975763e-05, "loss": 0.961, "step": 10183 }, { "epoch": 0.11, "learning_rate": 1.539890455366626e-05, "loss": 0.9179, "step": 10184 }, { "epoch": 0.11, "learning_rate": 1.5397997231630774e-05, "loss": 0.9086, "step": 10185 }, { "epoch": 0.11, "learning_rate": 1.5397089846879846e-05, "loss": 0.8938, "step": 10186 }, { "epoch": 0.11, "learning_rate": 1.539618239942402e-05, "loss": 0.929, "step": 10187 }, { "epoch": 0.11, "learning_rate": 1.5395274889273837e-05, "loss": 0.9105, "step": 10188 }, { "epoch": 0.11, "learning_rate": 1.5394367316439843e-05, "loss": 0.9122, "step": 10189 }, { "epoch": 0.11, "learning_rate": 1.539345968093258e-05, "loss": 0.9104, "step": 10190 }, { "epoch": 0.11, "learning_rate": 1.5392551982762596e-05, "loss": 0.9155, "step": 10191 }, { "epoch": 0.11, "learning_rate": 1.5391644221940434e-05, "loss": 0.884, "step": 10192 }, { "epoch": 0.11, "learning_rate": 1.5390736398476645e-05, "loss": 0.9143, "step": 10193 }, { "epoch": 0.11, "learning_rate": 1.538982851238177e-05, "loss": 0.864, "step": 10194 }, { "epoch": 0.11, "learning_rate": 1.538892056366636e-05, "loss": 0.9274, "step": 10195 }, { "epoch": 0.11, "learning_rate": 1.5388012552340964e-05, "loss": 0.9156, "step": 10196 }, { "epoch": 0.11, "learning_rate": 1.5387104478416133e-05, "loss": 0.9441, "step": 10197 }, { "epoch": 0.11, "learning_rate": 1.5386196341902417e-05, "loss": 0.8669, "step": 10198 }, { "epoch": 0.11, "learning_rate": 1.5385288142810365e-05, "loss": 0.8716, "step": 10199 }, { "epoch": 0.11, "learning_rate": 1.5384379881150527e-05, "loss": 0.8893, "step": 10200 }, { "epoch": 0.11, "learning_rate": 1.5383471556933463e-05, "loss": 0.934, "step": 10201 }, { "epoch": 0.11, "learning_rate": 1.5382563170169718e-05, "loss": 0.9408, "step": 10202 }, { "epoch": 0.11, "learning_rate": 1.538165472086985e-05, "loss": 0.974, "step": 10203 }, { "epoch": 0.11, "learning_rate": 1.5380746209044414e-05, "loss": 0.8704, "step": 10204 }, { "epoch": 0.11, "learning_rate": 1.537983763470396e-05, "loss": 0.9105, "step": 10205 }, { "epoch": 0.11, "learning_rate": 1.537892899785905e-05, "loss": 0.887, "step": 10206 }, { "epoch": 0.11, "learning_rate": 1.537802029852024e-05, "loss": 0.9273, "step": 10207 }, { "epoch": 0.11, "learning_rate": 1.5377111536698086e-05, "loss": 0.9113, "step": 10208 }, { "epoch": 0.11, "learning_rate": 1.5376202712403148e-05, "loss": 0.8891, "step": 10209 }, { "epoch": 0.11, "learning_rate": 1.537529382564598e-05, "loss": 0.8871, "step": 10210 }, { "epoch": 0.11, "learning_rate": 1.537438487643715e-05, "loss": 0.9108, "step": 10211 }, { "epoch": 0.11, "learning_rate": 1.537347586478721e-05, "loss": 0.8863, "step": 10212 }, { "epoch": 0.11, "learning_rate": 1.5372566790706726e-05, "loss": 0.8582, "step": 10213 }, { "epoch": 0.11, "learning_rate": 1.5371657654206262e-05, "loss": 0.8822, "step": 10214 }, { "epoch": 0.11, "learning_rate": 1.537074845529637e-05, "loss": 0.9085, "step": 10215 }, { "epoch": 0.11, "learning_rate": 1.536983919398763e-05, "loss": 0.9433, "step": 10216 }, { "epoch": 0.11, "learning_rate": 1.5368929870290592e-05, "loss": 0.8877, "step": 10217 }, { "epoch": 0.11, "learning_rate": 1.5368020484215828e-05, "loss": 0.8839, "step": 10218 }, { "epoch": 0.11, "learning_rate": 1.53671110357739e-05, "loss": 0.9406, "step": 10219 }, { "epoch": 0.11, "learning_rate": 1.5366201524975372e-05, "loss": 0.9028, "step": 10220 }, { "epoch": 0.11, "learning_rate": 1.5365291951830817e-05, "loss": 0.9275, "step": 10221 }, { "epoch": 0.11, "learning_rate": 1.5364382316350803e-05, "loss": 0.8935, "step": 10222 }, { "epoch": 0.11, "learning_rate": 1.536347261854589e-05, "loss": 0.8994, "step": 10223 }, { "epoch": 0.11, "learning_rate": 1.5362562858426655e-05, "loss": 0.8802, "step": 10224 }, { "epoch": 0.11, "learning_rate": 1.5361653036003664e-05, "loss": 0.925, "step": 10225 }, { "epoch": 0.11, "learning_rate": 1.5360743151287492e-05, "loss": 0.9054, "step": 10226 }, { "epoch": 0.11, "learning_rate": 1.5359833204288702e-05, "loss": 0.9136, "step": 10227 }, { "epoch": 0.11, "learning_rate": 1.5358923195017877e-05, "loss": 0.8776, "step": 10228 }, { "epoch": 0.11, "learning_rate": 1.535801312348558e-05, "loss": 0.8921, "step": 10229 }, { "epoch": 0.11, "learning_rate": 1.5357102989702387e-05, "loss": 0.8755, "step": 10230 }, { "epoch": 0.11, "learning_rate": 1.5356192793678876e-05, "loss": 0.8981, "step": 10231 }, { "epoch": 0.11, "learning_rate": 1.5355282535425626e-05, "loss": 0.952, "step": 10232 }, { "epoch": 0.11, "learning_rate": 1.5354372214953198e-05, "loss": 0.8982, "step": 10233 }, { "epoch": 0.11, "learning_rate": 1.5353461832272176e-05, "loss": 0.8945, "step": 10234 }, { "epoch": 0.11, "learning_rate": 1.5352551387393143e-05, "loss": 0.9198, "step": 10235 }, { "epoch": 0.11, "learning_rate": 1.5351640880326668e-05, "loss": 0.8835, "step": 10236 }, { "epoch": 0.11, "learning_rate": 1.5350730311083333e-05, "loss": 0.9469, "step": 10237 }, { "epoch": 0.11, "learning_rate": 1.534981967967372e-05, "loss": 0.9291, "step": 10238 }, { "epoch": 0.11, "learning_rate": 1.5348908986108398e-05, "loss": 0.9304, "step": 10239 }, { "epoch": 0.11, "learning_rate": 1.5347998230397963e-05, "loss": 0.8886, "step": 10240 }, { "epoch": 0.11, "learning_rate": 1.534708741255299e-05, "loss": 0.8944, "step": 10241 }, { "epoch": 0.11, "learning_rate": 1.5346176532584058e-05, "loss": 0.9126, "step": 10242 }, { "epoch": 0.11, "learning_rate": 1.5345265590501753e-05, "loss": 0.9377, "step": 10243 }, { "epoch": 0.11, "learning_rate": 1.534435458631666e-05, "loss": 0.9618, "step": 10244 }, { "epoch": 0.11, "learning_rate": 1.5343443520039355e-05, "loss": 0.8975, "step": 10245 }, { "epoch": 0.11, "learning_rate": 1.5342532391680434e-05, "loss": 0.9027, "step": 10246 }, { "epoch": 0.11, "learning_rate": 1.5341621201250478e-05, "loss": 0.9175, "step": 10247 }, { "epoch": 0.11, "learning_rate": 1.534070994876007e-05, "loss": 0.9404, "step": 10248 }, { "epoch": 0.11, "learning_rate": 1.53397986342198e-05, "loss": 0.892, "step": 10249 }, { "epoch": 0.11, "learning_rate": 1.533888725764026e-05, "loss": 0.9211, "step": 10250 }, { "epoch": 0.11, "learning_rate": 1.533797581903204e-05, "loss": 0.9418, "step": 10251 }, { "epoch": 0.11, "learning_rate": 1.5337064318405713e-05, "loss": 0.9227, "step": 10252 }, { "epoch": 0.11, "learning_rate": 1.5336152755771886e-05, "loss": 0.925, "step": 10253 }, { "epoch": 0.11, "learning_rate": 1.5335241131141145e-05, "loss": 0.9073, "step": 10254 }, { "epoch": 0.11, "learning_rate": 1.533432944452408e-05, "loss": 0.9316, "step": 10255 }, { "epoch": 0.11, "learning_rate": 1.5333417695931283e-05, "loss": 0.9102, "step": 10256 }, { "epoch": 0.11, "learning_rate": 1.533250588537335e-05, "loss": 0.9228, "step": 10257 }, { "epoch": 0.11, "learning_rate": 1.533159401286087e-05, "loss": 0.906, "step": 10258 }, { "epoch": 0.11, "learning_rate": 1.5330682078404442e-05, "loss": 0.9211, "step": 10259 }, { "epoch": 0.11, "learning_rate": 1.532977008201466e-05, "loss": 0.8934, "step": 10260 }, { "epoch": 0.11, "learning_rate": 1.5328858023702116e-05, "loss": 0.9037, "step": 10261 }, { "epoch": 0.11, "learning_rate": 1.5327945903477416e-05, "loss": 0.9096, "step": 10262 }, { "epoch": 0.11, "learning_rate": 1.5327033721351145e-05, "loss": 0.9114, "step": 10263 }, { "epoch": 0.11, "learning_rate": 1.5326121477333908e-05, "loss": 0.8646, "step": 10264 }, { "epoch": 0.11, "learning_rate": 1.5325209171436302e-05, "loss": 0.9574, "step": 10265 }, { "epoch": 0.11, "learning_rate": 1.5324296803668927e-05, "loss": 0.8861, "step": 10266 }, { "epoch": 0.11, "learning_rate": 1.5323384374042383e-05, "loss": 0.9118, "step": 10267 }, { "epoch": 0.11, "learning_rate": 1.5322471882567272e-05, "loss": 0.9527, "step": 10268 }, { "epoch": 0.11, "learning_rate": 1.5321559329254192e-05, "loss": 0.9088, "step": 10269 }, { "epoch": 0.11, "learning_rate": 1.5320646714113748e-05, "loss": 0.8913, "step": 10270 }, { "epoch": 0.11, "learning_rate": 1.5319734037156545e-05, "loss": 0.9158, "step": 10271 }, { "epoch": 0.11, "learning_rate": 1.5318821298393184e-05, "loss": 0.883, "step": 10272 }, { "epoch": 0.11, "learning_rate": 1.531790849783427e-05, "loss": 0.8763, "step": 10273 }, { "epoch": 0.11, "learning_rate": 1.531699563549041e-05, "loss": 0.8862, "step": 10274 }, { "epoch": 0.11, "learning_rate": 1.5316082711372206e-05, "loss": 0.9248, "step": 10275 }, { "epoch": 0.11, "learning_rate": 1.5315169725490266e-05, "loss": 0.95, "step": 10276 }, { "epoch": 0.11, "learning_rate": 1.53142566778552e-05, "loss": 0.9025, "step": 10277 }, { "epoch": 0.11, "learning_rate": 1.5313343568477613e-05, "loss": 0.9186, "step": 10278 }, { "epoch": 0.11, "learning_rate": 1.5312430397368114e-05, "loss": 0.9025, "step": 10279 }, { "epoch": 0.11, "learning_rate": 1.5311517164537316e-05, "loss": 0.9418, "step": 10280 }, { "epoch": 0.11, "learning_rate": 1.5310603869995824e-05, "loss": 0.9116, "step": 10281 }, { "epoch": 0.11, "learning_rate": 1.530969051375425e-05, "loss": 0.887, "step": 10282 }, { "epoch": 0.11, "learning_rate": 1.530877709582321e-05, "loss": 0.9406, "step": 10283 }, { "epoch": 0.11, "learning_rate": 1.5307863616213313e-05, "loss": 0.884, "step": 10284 }, { "epoch": 0.11, "learning_rate": 1.530695007493517e-05, "loss": 0.8983, "step": 10285 }, { "epoch": 0.11, "learning_rate": 1.5306036471999397e-05, "loss": 0.9164, "step": 10286 }, { "epoch": 0.11, "learning_rate": 1.5305122807416614e-05, "loss": 0.9169, "step": 10287 }, { "epoch": 0.11, "learning_rate": 1.5304209081197425e-05, "loss": 0.8527, "step": 10288 }, { "epoch": 0.11, "learning_rate": 1.5303295293352456e-05, "loss": 0.9442, "step": 10289 }, { "epoch": 0.11, "learning_rate": 1.5302381443892318e-05, "loss": 0.9167, "step": 10290 }, { "epoch": 0.11, "learning_rate": 1.530146753282763e-05, "loss": 0.8787, "step": 10291 }, { "epoch": 0.11, "learning_rate": 1.530055356016901e-05, "loss": 0.8692, "step": 10292 }, { "epoch": 0.11, "learning_rate": 1.5299639525927075e-05, "loss": 0.9336, "step": 10293 }, { "epoch": 0.11, "learning_rate": 1.529872543011245e-05, "loss": 0.9368, "step": 10294 }, { "epoch": 0.11, "learning_rate": 1.5297811272735747e-05, "loss": 0.9326, "step": 10295 }, { "epoch": 0.11, "learning_rate": 1.5296897053807593e-05, "loss": 0.8614, "step": 10296 }, { "epoch": 0.11, "learning_rate": 1.529598277333861e-05, "loss": 0.8903, "step": 10297 }, { "epoch": 0.11, "learning_rate": 1.5295068431339414e-05, "loss": 0.9143, "step": 10298 }, { "epoch": 0.11, "learning_rate": 1.5294154027820635e-05, "loss": 0.9051, "step": 10299 }, { "epoch": 0.11, "learning_rate": 1.5293239562792893e-05, "loss": 0.9291, "step": 10300 }, { "epoch": 0.11, "learning_rate": 1.5292325036266818e-05, "loss": 0.9092, "step": 10301 }, { "epoch": 0.11, "learning_rate": 1.5291410448253025e-05, "loss": 0.8681, "step": 10302 }, { "epoch": 0.11, "learning_rate": 1.5290495798762154e-05, "loss": 0.8913, "step": 10303 }, { "epoch": 0.11, "learning_rate": 1.5289581087804816e-05, "loss": 0.9158, "step": 10304 }, { "epoch": 0.11, "learning_rate": 1.5288666315391646e-05, "loss": 0.9302, "step": 10305 }, { "epoch": 0.11, "learning_rate": 1.5287751481533276e-05, "loss": 0.9056, "step": 10306 }, { "epoch": 0.11, "learning_rate": 1.5286836586240328e-05, "loss": 0.9194, "step": 10307 }, { "epoch": 0.11, "learning_rate": 1.5285921629523438e-05, "loss": 0.9451, "step": 10308 }, { "epoch": 0.11, "learning_rate": 1.528500661139323e-05, "loss": 0.9257, "step": 10309 }, { "epoch": 0.11, "learning_rate": 1.5284091531860336e-05, "loss": 0.9143, "step": 10310 }, { "epoch": 0.11, "learning_rate": 1.5283176390935396e-05, "loss": 0.9093, "step": 10311 }, { "epoch": 0.11, "learning_rate": 1.5282261188629027e-05, "loss": 0.9079, "step": 10312 }, { "epoch": 0.11, "learning_rate": 1.5281345924951873e-05, "loss": 0.8787, "step": 10313 }, { "epoch": 0.11, "learning_rate": 1.5280430599914567e-05, "loss": 0.9045, "step": 10314 }, { "epoch": 0.11, "learning_rate": 1.5279515213527742e-05, "loss": 0.912, "step": 10315 }, { "epoch": 0.11, "learning_rate": 1.527859976580203e-05, "loss": 0.9404, "step": 10316 }, { "epoch": 0.11, "learning_rate": 1.5277684256748074e-05, "loss": 0.9163, "step": 10317 }, { "epoch": 0.11, "learning_rate": 1.5276768686376505e-05, "loss": 0.8779, "step": 10318 }, { "epoch": 0.11, "learning_rate": 1.527585305469796e-05, "loss": 0.9498, "step": 10319 }, { "epoch": 0.11, "learning_rate": 1.5274937361723082e-05, "loss": 0.8965, "step": 10320 }, { "epoch": 0.11, "learning_rate": 1.5274021607462506e-05, "loss": 0.8955, "step": 10321 }, { "epoch": 0.11, "learning_rate": 1.5273105791926872e-05, "loss": 0.8688, "step": 10322 }, { "epoch": 0.11, "learning_rate": 1.5272189915126823e-05, "loss": 0.9313, "step": 10323 }, { "epoch": 0.11, "learning_rate": 1.5271273977072993e-05, "loss": 0.9197, "step": 10324 }, { "epoch": 0.11, "learning_rate": 1.5270357977776035e-05, "loss": 0.9261, "step": 10325 }, { "epoch": 0.11, "learning_rate": 1.5269441917246573e-05, "loss": 0.8734, "step": 10326 }, { "epoch": 0.11, "learning_rate": 1.526852579549527e-05, "loss": 0.9208, "step": 10327 }, { "epoch": 0.11, "learning_rate": 1.526760961253276e-05, "loss": 0.8949, "step": 10328 }, { "epoch": 0.11, "learning_rate": 1.526669336836969e-05, "loss": 0.9322, "step": 10329 }, { "epoch": 0.11, "learning_rate": 1.5265777063016703e-05, "loss": 0.9014, "step": 10330 }, { "epoch": 0.11, "learning_rate": 1.5264860696484443e-05, "loss": 0.9218, "step": 10331 }, { "epoch": 0.11, "learning_rate": 1.5263944268783563e-05, "loss": 0.8951, "step": 10332 }, { "epoch": 0.11, "learning_rate": 1.5263027779924705e-05, "loss": 0.9183, "step": 10333 }, { "epoch": 0.11, "learning_rate": 1.5262111229918516e-05, "loss": 0.8872, "step": 10334 }, { "epoch": 0.11, "learning_rate": 1.526119461877565e-05, "loss": 0.9576, "step": 10335 }, { "epoch": 0.11, "learning_rate": 1.5260277946506756e-05, "loss": 0.8973, "step": 10336 }, { "epoch": 0.11, "learning_rate": 1.525936121312248e-05, "loss": 0.9303, "step": 10337 }, { "epoch": 0.11, "learning_rate": 1.5258444418633474e-05, "loss": 0.9316, "step": 10338 }, { "epoch": 0.11, "learning_rate": 1.525752756305039e-05, "loss": 0.9223, "step": 10339 }, { "epoch": 0.11, "learning_rate": 1.5256610646383884e-05, "loss": 0.8876, "step": 10340 }, { "epoch": 0.11, "learning_rate": 1.5255693668644603e-05, "loss": 0.9068, "step": 10341 }, { "epoch": 0.11, "learning_rate": 1.5254776629843204e-05, "loss": 0.9299, "step": 10342 }, { "epoch": 0.11, "learning_rate": 1.525385952999034e-05, "loss": 0.9055, "step": 10343 }, { "epoch": 0.11, "learning_rate": 1.5252942369096668e-05, "loss": 0.9045, "step": 10344 }, { "epoch": 0.11, "learning_rate": 1.5252025147172844e-05, "loss": 0.9354, "step": 10345 }, { "epoch": 0.11, "learning_rate": 1.5251107864229521e-05, "loss": 0.9102, "step": 10346 }, { "epoch": 0.11, "learning_rate": 1.525019052027736e-05, "loss": 0.9199, "step": 10347 }, { "epoch": 0.11, "learning_rate": 1.524927311532702e-05, "loss": 0.9126, "step": 10348 }, { "epoch": 0.11, "learning_rate": 1.5248355649389152e-05, "loss": 0.9241, "step": 10349 }, { "epoch": 0.11, "learning_rate": 1.5247438122474425e-05, "loss": 0.8991, "step": 10350 }, { "epoch": 0.11, "learning_rate": 1.5246520534593491e-05, "loss": 0.9345, "step": 10351 }, { "epoch": 0.11, "learning_rate": 1.5245602885757017e-05, "loss": 0.9255, "step": 10352 }, { "epoch": 0.11, "learning_rate": 1.5244685175975663e-05, "loss": 0.9101, "step": 10353 }, { "epoch": 0.11, "learning_rate": 1.5243767405260089e-05, "loss": 0.878, "step": 10354 }, { "epoch": 0.11, "learning_rate": 1.5242849573620957e-05, "loss": 0.8936, "step": 10355 }, { "epoch": 0.11, "learning_rate": 1.5241931681068936e-05, "loss": 0.8926, "step": 10356 }, { "epoch": 0.11, "learning_rate": 1.5241013727614686e-05, "loss": 0.8693, "step": 10357 }, { "epoch": 0.11, "learning_rate": 1.5240095713268872e-05, "loss": 0.8827, "step": 10358 }, { "epoch": 0.11, "learning_rate": 1.5239177638042164e-05, "loss": 0.9121, "step": 10359 }, { "epoch": 0.11, "learning_rate": 1.523825950194522e-05, "loss": 0.9138, "step": 10360 }, { "epoch": 0.11, "learning_rate": 1.5237341304988718e-05, "loss": 0.8866, "step": 10361 }, { "epoch": 0.11, "learning_rate": 1.5236423047183315e-05, "loss": 0.8863, "step": 10362 }, { "epoch": 0.11, "learning_rate": 1.523550472853969e-05, "loss": 0.9282, "step": 10363 }, { "epoch": 0.11, "learning_rate": 1.5234586349068505e-05, "loss": 0.9208, "step": 10364 }, { "epoch": 0.11, "learning_rate": 1.523366790878043e-05, "loss": 0.9087, "step": 10365 }, { "epoch": 0.11, "learning_rate": 1.5232749407686142e-05, "loss": 0.8867, "step": 10366 }, { "epoch": 0.11, "learning_rate": 1.5231830845796307e-05, "loss": 0.8887, "step": 10367 }, { "epoch": 0.11, "learning_rate": 1.5230912223121597e-05, "loss": 0.9385, "step": 10368 }, { "epoch": 0.11, "learning_rate": 1.5229993539672684e-05, "loss": 0.8896, "step": 10369 }, { "epoch": 0.11, "learning_rate": 1.522907479546025e-05, "loss": 0.8829, "step": 10370 }, { "epoch": 0.11, "learning_rate": 1.522815599049496e-05, "loss": 0.9, "step": 10371 }, { "epoch": 0.11, "learning_rate": 1.522723712478749e-05, "loss": 0.9501, "step": 10372 }, { "epoch": 0.11, "learning_rate": 1.5226318198348522e-05, "loss": 0.9231, "step": 10373 }, { "epoch": 0.11, "learning_rate": 1.5225399211188723e-05, "loss": 0.886, "step": 10374 }, { "epoch": 0.11, "learning_rate": 1.5224480163318777e-05, "loss": 0.949, "step": 10375 }, { "epoch": 0.11, "learning_rate": 1.5223561054749361e-05, "loss": 0.9117, "step": 10376 }, { "epoch": 0.11, "learning_rate": 1.522264188549115e-05, "loss": 0.9405, "step": 10377 }, { "epoch": 0.11, "learning_rate": 1.5221722655554824e-05, "loss": 0.8736, "step": 10378 }, { "epoch": 0.11, "learning_rate": 1.5220803364951069e-05, "loss": 0.9058, "step": 10379 }, { "epoch": 0.11, "learning_rate": 1.5219884013690557e-05, "loss": 0.9165, "step": 10380 }, { "epoch": 0.11, "learning_rate": 1.5218964601783974e-05, "loss": 0.9378, "step": 10381 }, { "epoch": 0.11, "learning_rate": 1.5218045129242004e-05, "loss": 0.9213, "step": 10382 }, { "epoch": 0.11, "learning_rate": 1.5217125596075324e-05, "loss": 0.9201, "step": 10383 }, { "epoch": 0.11, "learning_rate": 1.5216206002294617e-05, "loss": 0.8928, "step": 10384 }, { "epoch": 0.11, "learning_rate": 1.5215286347910574e-05, "loss": 0.9583, "step": 10385 }, { "epoch": 0.11, "learning_rate": 1.5214366632933877e-05, "loss": 0.9224, "step": 10386 }, { "epoch": 0.11, "learning_rate": 1.5213446857375208e-05, "loss": 0.9114, "step": 10387 }, { "epoch": 0.11, "learning_rate": 1.5212527021245256e-05, "loss": 0.9149, "step": 10388 }, { "epoch": 0.11, "learning_rate": 1.5211607124554709e-05, "loss": 0.9299, "step": 10389 }, { "epoch": 0.11, "learning_rate": 1.5210687167314251e-05, "loss": 0.9314, "step": 10390 }, { "epoch": 0.11, "learning_rate": 1.5209767149534576e-05, "loss": 0.9136, "step": 10391 }, { "epoch": 0.11, "learning_rate": 1.5208847071226366e-05, "loss": 0.9029, "step": 10392 }, { "epoch": 0.11, "learning_rate": 1.5207926932400314e-05, "loss": 0.8577, "step": 10393 }, { "epoch": 0.11, "learning_rate": 1.5207006733067114e-05, "loss": 0.917, "step": 10394 }, { "epoch": 0.11, "learning_rate": 1.5206086473237453e-05, "loss": 0.9203, "step": 10395 }, { "epoch": 0.11, "learning_rate": 1.5205166152922024e-05, "loss": 0.8975, "step": 10396 }, { "epoch": 0.11, "learning_rate": 1.5204245772131518e-05, "loss": 0.9185, "step": 10397 }, { "epoch": 0.11, "learning_rate": 1.520332533087663e-05, "loss": 0.9139, "step": 10398 }, { "epoch": 0.11, "learning_rate": 1.5202404829168051e-05, "loss": 0.9277, "step": 10399 }, { "epoch": 0.11, "learning_rate": 1.5201484267016482e-05, "loss": 0.9005, "step": 10400 }, { "epoch": 0.11, "learning_rate": 1.5200563644432614e-05, "loss": 0.9409, "step": 10401 }, { "epoch": 0.11, "learning_rate": 1.5199642961427141e-05, "loss": 0.8816, "step": 10402 }, { "epoch": 0.11, "learning_rate": 1.5198722218010766e-05, "loss": 0.911, "step": 10403 }, { "epoch": 0.11, "learning_rate": 1.519780141419418e-05, "loss": 0.8996, "step": 10404 }, { "epoch": 0.11, "learning_rate": 1.5196880549988081e-05, "loss": 0.8954, "step": 10405 }, { "epoch": 0.11, "learning_rate": 1.5195959625403176e-05, "loss": 0.9111, "step": 10406 }, { "epoch": 0.11, "learning_rate": 1.519503864045016e-05, "loss": 0.8958, "step": 10407 }, { "epoch": 0.11, "learning_rate": 1.5194117595139725e-05, "loss": 0.9458, "step": 10408 }, { "epoch": 0.11, "learning_rate": 1.5193196489482586e-05, "loss": 0.9104, "step": 10409 }, { "epoch": 0.11, "learning_rate": 1.5192275323489437e-05, "loss": 0.9061, "step": 10410 }, { "epoch": 0.11, "learning_rate": 1.519135409717098e-05, "loss": 0.871, "step": 10411 }, { "epoch": 0.11, "learning_rate": 1.5190432810537923e-05, "loss": 0.8933, "step": 10412 }, { "epoch": 0.11, "learning_rate": 1.5189511463600964e-05, "loss": 0.9251, "step": 10413 }, { "epoch": 0.11, "learning_rate": 1.5188590056370808e-05, "loss": 0.8889, "step": 10414 }, { "epoch": 0.11, "learning_rate": 1.5187668588858165e-05, "loss": 0.8861, "step": 10415 }, { "epoch": 0.11, "learning_rate": 1.5186747061073737e-05, "loss": 0.8853, "step": 10416 }, { "epoch": 0.11, "learning_rate": 1.518582547302823e-05, "loss": 0.8829, "step": 10417 }, { "epoch": 0.11, "learning_rate": 1.5184903824732353e-05, "loss": 0.8579, "step": 10418 }, { "epoch": 0.11, "learning_rate": 1.5183982116196818e-05, "loss": 0.9271, "step": 10419 }, { "epoch": 0.11, "learning_rate": 1.5183060347432323e-05, "loss": 0.9322, "step": 10420 }, { "epoch": 0.11, "learning_rate": 1.5182138518449587e-05, "loss": 0.931, "step": 10421 }, { "epoch": 0.11, "learning_rate": 1.5181216629259315e-05, "loss": 0.8988, "step": 10422 }, { "epoch": 0.11, "learning_rate": 1.5180294679872221e-05, "loss": 0.9576, "step": 10423 }, { "epoch": 0.11, "learning_rate": 1.5179372670299014e-05, "loss": 0.9002, "step": 10424 }, { "epoch": 0.11, "learning_rate": 1.5178450600550409e-05, "loss": 0.8749, "step": 10425 }, { "epoch": 0.11, "learning_rate": 1.5177528470637115e-05, "loss": 0.8969, "step": 10426 }, { "epoch": 0.11, "learning_rate": 1.517660628056985e-05, "loss": 0.8905, "step": 10427 }, { "epoch": 0.11, "learning_rate": 1.5175684030359325e-05, "loss": 0.9244, "step": 10428 }, { "epoch": 0.11, "learning_rate": 1.5174761720016253e-05, "loss": 0.9185, "step": 10429 }, { "epoch": 0.11, "learning_rate": 1.5173839349551356e-05, "loss": 0.883, "step": 10430 }, { "epoch": 0.11, "learning_rate": 1.5172916918975344e-05, "loss": 0.9174, "step": 10431 }, { "epoch": 0.11, "learning_rate": 1.517199442829894e-05, "loss": 0.908, "step": 10432 }, { "epoch": 0.11, "learning_rate": 1.5171071877532854e-05, "loss": 0.9244, "step": 10433 }, { "epoch": 0.11, "learning_rate": 1.5170149266687815e-05, "loss": 0.925, "step": 10434 }, { "epoch": 0.11, "learning_rate": 1.5169226595774533e-05, "loss": 0.9508, "step": 10435 }, { "epoch": 0.11, "learning_rate": 1.5168303864803729e-05, "loss": 0.914, "step": 10436 }, { "epoch": 0.11, "learning_rate": 1.5167381073786129e-05, "loss": 0.8852, "step": 10437 }, { "epoch": 0.11, "learning_rate": 1.5166458222732451e-05, "loss": 0.921, "step": 10438 }, { "epoch": 0.11, "learning_rate": 1.5165535311653416e-05, "loss": 0.9348, "step": 10439 }, { "epoch": 0.11, "learning_rate": 1.5164612340559746e-05, "loss": 0.911, "step": 10440 }, { "epoch": 0.11, "learning_rate": 1.5163689309462168e-05, "loss": 0.9676, "step": 10441 }, { "epoch": 0.11, "learning_rate": 1.5162766218371401e-05, "loss": 0.8891, "step": 10442 }, { "epoch": 0.11, "learning_rate": 1.5161843067298176e-05, "loss": 0.9017, "step": 10443 }, { "epoch": 0.11, "learning_rate": 1.5160919856253214e-05, "loss": 0.871, "step": 10444 }, { "epoch": 0.11, "learning_rate": 1.5159996585247238e-05, "loss": 0.9122, "step": 10445 }, { "epoch": 0.11, "learning_rate": 1.5159073254290983e-05, "loss": 0.916, "step": 10446 }, { "epoch": 0.11, "learning_rate": 1.5158149863395173e-05, "loss": 0.9175, "step": 10447 }, { "epoch": 0.11, "learning_rate": 1.5157226412570533e-05, "loss": 0.9396, "step": 10448 }, { "epoch": 0.11, "learning_rate": 1.51563029018278e-05, "loss": 0.9218, "step": 10449 }, { "epoch": 0.11, "learning_rate": 1.5155379331177692e-05, "loss": 0.9353, "step": 10450 }, { "epoch": 0.11, "learning_rate": 1.5154455700630948e-05, "loss": 0.8985, "step": 10451 }, { "epoch": 0.11, "learning_rate": 1.5153532010198296e-05, "loss": 0.8684, "step": 10452 }, { "epoch": 0.11, "learning_rate": 1.5152608259890469e-05, "loss": 0.8799, "step": 10453 }, { "epoch": 0.11, "learning_rate": 1.5151684449718197e-05, "loss": 0.8985, "step": 10454 }, { "epoch": 0.11, "learning_rate": 1.5150760579692217e-05, "loss": 0.9411, "step": 10455 }, { "epoch": 0.12, "learning_rate": 1.5149836649823257e-05, "loss": 0.9406, "step": 10456 }, { "epoch": 0.12, "learning_rate": 1.5148912660122057e-05, "loss": 0.9499, "step": 10457 }, { "epoch": 0.12, "learning_rate": 1.5147988610599352e-05, "loss": 0.8774, "step": 10458 }, { "epoch": 0.12, "learning_rate": 1.5147064501265871e-05, "loss": 0.9174, "step": 10459 }, { "epoch": 0.12, "learning_rate": 1.5146140332132359e-05, "loss": 0.9243, "step": 10460 }, { "epoch": 0.12, "learning_rate": 1.514521610320955e-05, "loss": 0.9145, "step": 10461 }, { "epoch": 0.12, "learning_rate": 1.5144291814508184e-05, "loss": 0.88, "step": 10462 }, { "epoch": 0.12, "learning_rate": 1.5143367466038989e-05, "loss": 0.9024, "step": 10463 }, { "epoch": 0.12, "learning_rate": 1.514244305781272e-05, "loss": 0.9334, "step": 10464 }, { "epoch": 0.12, "learning_rate": 1.5141518589840107e-05, "loss": 0.9007, "step": 10465 }, { "epoch": 0.12, "learning_rate": 1.5140594062131895e-05, "loss": 0.8747, "step": 10466 }, { "epoch": 0.12, "learning_rate": 1.5139669474698821e-05, "loss": 0.882, "step": 10467 }, { "epoch": 0.12, "learning_rate": 1.5138744827551635e-05, "loss": 0.9201, "step": 10468 }, { "epoch": 0.12, "learning_rate": 1.5137820120701067e-05, "loss": 0.8999, "step": 10469 }, { "epoch": 0.12, "learning_rate": 1.5136895354157874e-05, "loss": 0.9002, "step": 10470 }, { "epoch": 0.12, "learning_rate": 1.5135970527932795e-05, "loss": 0.9308, "step": 10471 }, { "epoch": 0.12, "learning_rate": 1.5135045642036574e-05, "loss": 0.911, "step": 10472 }, { "epoch": 0.12, "learning_rate": 1.5134120696479955e-05, "loss": 0.9066, "step": 10473 }, { "epoch": 0.12, "learning_rate": 1.5133195691273687e-05, "loss": 0.8728, "step": 10474 }, { "epoch": 0.12, "learning_rate": 1.5132270626428516e-05, "loss": 0.8914, "step": 10475 }, { "epoch": 0.12, "learning_rate": 1.513134550195519e-05, "loss": 0.9004, "step": 10476 }, { "epoch": 0.12, "learning_rate": 1.5130420317864458e-05, "loss": 0.9073, "step": 10477 }, { "epoch": 0.12, "learning_rate": 1.5129495074167067e-05, "loss": 0.9228, "step": 10478 }, { "epoch": 0.12, "learning_rate": 1.5128569770873767e-05, "loss": 0.9334, "step": 10479 }, { "epoch": 0.12, "learning_rate": 1.5127644407995314e-05, "loss": 0.8641, "step": 10480 }, { "epoch": 0.12, "learning_rate": 1.512671898554245e-05, "loss": 0.908, "step": 10481 }, { "epoch": 0.12, "learning_rate": 1.5125793503525935e-05, "loss": 0.8884, "step": 10482 }, { "epoch": 0.12, "learning_rate": 1.5124867961956517e-05, "loss": 0.9299, "step": 10483 }, { "epoch": 0.12, "learning_rate": 1.5123942360844947e-05, "loss": 0.938, "step": 10484 }, { "epoch": 0.12, "learning_rate": 1.5123016700201984e-05, "loss": 0.9492, "step": 10485 }, { "epoch": 0.12, "learning_rate": 1.512209098003838e-05, "loss": 0.9423, "step": 10486 }, { "epoch": 0.12, "learning_rate": 1.5121165200364893e-05, "loss": 0.9069, "step": 10487 }, { "epoch": 0.12, "learning_rate": 1.5120239361192275e-05, "loss": 0.9535, "step": 10488 }, { "epoch": 0.12, "learning_rate": 1.5119313462531284e-05, "loss": 0.9327, "step": 10489 }, { "epoch": 0.12, "learning_rate": 1.5118387504392682e-05, "loss": 0.8958, "step": 10490 }, { "epoch": 0.12, "learning_rate": 1.5117461486787215e-05, "loss": 0.908, "step": 10491 }, { "epoch": 0.12, "learning_rate": 1.5116535409725657e-05, "loss": 0.8861, "step": 10492 }, { "epoch": 0.12, "learning_rate": 1.5115609273218757e-05, "loss": 0.9427, "step": 10493 }, { "epoch": 0.12, "learning_rate": 1.5114683077277277e-05, "loss": 0.909, "step": 10494 }, { "epoch": 0.12, "learning_rate": 1.511375682191198e-05, "loss": 0.9239, "step": 10495 }, { "epoch": 0.12, "learning_rate": 1.5112830507133627e-05, "loss": 0.8702, "step": 10496 }, { "epoch": 0.12, "learning_rate": 1.511190413295298e-05, "loss": 0.9509, "step": 10497 }, { "epoch": 0.12, "learning_rate": 1.51109776993808e-05, "loss": 0.9008, "step": 10498 }, { "epoch": 0.12, "learning_rate": 1.5110051206427854e-05, "loss": 0.9342, "step": 10499 }, { "epoch": 0.12, "learning_rate": 1.5109124654104901e-05, "loss": 0.917, "step": 10500 }, { "epoch": 0.12, "learning_rate": 1.5108198042422713e-05, "loss": 0.9059, "step": 10501 }, { "epoch": 0.12, "learning_rate": 1.5107271371392049e-05, "loss": 0.8481, "step": 10502 }, { "epoch": 0.12, "learning_rate": 1.5106344641023677e-05, "loss": 0.9146, "step": 10503 }, { "epoch": 0.12, "learning_rate": 1.5105417851328371e-05, "loss": 0.9661, "step": 10504 }, { "epoch": 0.12, "learning_rate": 1.5104491002316887e-05, "loss": 0.9349, "step": 10505 }, { "epoch": 0.12, "learning_rate": 1.5103564094000005e-05, "loss": 0.8756, "step": 10506 }, { "epoch": 0.12, "learning_rate": 1.5102637126388484e-05, "loss": 0.924, "step": 10507 }, { "epoch": 0.12, "learning_rate": 1.51017100994931e-05, "loss": 0.9037, "step": 10508 }, { "epoch": 0.12, "learning_rate": 1.5100783013324617e-05, "loss": 0.9051, "step": 10509 }, { "epoch": 0.12, "learning_rate": 1.509985586789382e-05, "loss": 0.9054, "step": 10510 }, { "epoch": 0.12, "learning_rate": 1.5098928663211467e-05, "loss": 0.9226, "step": 10511 }, { "epoch": 0.12, "learning_rate": 1.509800139928833e-05, "loss": 0.9253, "step": 10512 }, { "epoch": 0.12, "learning_rate": 1.5097074076135196e-05, "loss": 0.9347, "step": 10513 }, { "epoch": 0.12, "learning_rate": 1.5096146693762825e-05, "loss": 0.8863, "step": 10514 }, { "epoch": 0.12, "learning_rate": 1.5095219252182e-05, "loss": 0.9255, "step": 10515 }, { "epoch": 0.12, "learning_rate": 1.5094291751403487e-05, "loss": 0.8809, "step": 10516 }, { "epoch": 0.12, "learning_rate": 1.5093364191438074e-05, "loss": 0.9376, "step": 10517 }, { "epoch": 0.12, "learning_rate": 1.509243657229653e-05, "loss": 0.9174, "step": 10518 }, { "epoch": 0.12, "learning_rate": 1.5091508893989633e-05, "loss": 0.9383, "step": 10519 }, { "epoch": 0.12, "learning_rate": 1.5090581156528164e-05, "loss": 0.9064, "step": 10520 }, { "epoch": 0.12, "learning_rate": 1.50896533599229e-05, "loss": 0.8764, "step": 10521 }, { "epoch": 0.12, "learning_rate": 1.5088725504184619e-05, "loss": 0.9009, "step": 10522 }, { "epoch": 0.12, "learning_rate": 1.50877975893241e-05, "loss": 0.9289, "step": 10523 }, { "epoch": 0.12, "learning_rate": 1.508686961535213e-05, "loss": 0.8527, "step": 10524 }, { "epoch": 0.12, "learning_rate": 1.5085941582279482e-05, "loss": 0.9177, "step": 10525 }, { "epoch": 0.12, "learning_rate": 1.5085013490116947e-05, "loss": 0.9323, "step": 10526 }, { "epoch": 0.12, "learning_rate": 1.5084085338875301e-05, "loss": 0.9121, "step": 10527 }, { "epoch": 0.12, "learning_rate": 1.5083157128565331e-05, "loss": 0.914, "step": 10528 }, { "epoch": 0.12, "learning_rate": 1.508222885919782e-05, "loss": 0.9245, "step": 10529 }, { "epoch": 0.12, "learning_rate": 1.5081300530783553e-05, "loss": 0.9409, "step": 10530 }, { "epoch": 0.12, "learning_rate": 1.5080372143333315e-05, "loss": 0.9251, "step": 10531 }, { "epoch": 0.12, "learning_rate": 1.5079443696857895e-05, "loss": 0.9003, "step": 10532 }, { "epoch": 0.12, "learning_rate": 1.507851519136808e-05, "loss": 0.9264, "step": 10533 }, { "epoch": 0.12, "learning_rate": 1.507758662687465e-05, "loss": 0.9088, "step": 10534 }, { "epoch": 0.12, "learning_rate": 1.50766580033884e-05, "loss": 0.9349, "step": 10535 }, { "epoch": 0.12, "learning_rate": 1.5075729320920117e-05, "loss": 0.8951, "step": 10536 }, { "epoch": 0.12, "learning_rate": 1.5074800579480594e-05, "loss": 0.8902, "step": 10537 }, { "epoch": 0.12, "learning_rate": 1.5073871779080622e-05, "loss": 0.9379, "step": 10538 }, { "epoch": 0.12, "learning_rate": 1.5072942919730986e-05, "loss": 0.9277, "step": 10539 }, { "epoch": 0.12, "learning_rate": 1.5072014001442479e-05, "loss": 0.9379, "step": 10540 }, { "epoch": 0.12, "learning_rate": 1.50710850242259e-05, "loss": 0.8974, "step": 10541 }, { "epoch": 0.12, "learning_rate": 1.5070155988092033e-05, "loss": 0.8869, "step": 10542 }, { "epoch": 0.12, "learning_rate": 1.5069226893051675e-05, "loss": 0.9122, "step": 10543 }, { "epoch": 0.12, "learning_rate": 1.5068297739115623e-05, "loss": 0.9164, "step": 10544 }, { "epoch": 0.12, "learning_rate": 1.5067368526294672e-05, "loss": 0.9378, "step": 10545 }, { "epoch": 0.12, "learning_rate": 1.5066439254599617e-05, "loss": 0.924, "step": 10546 }, { "epoch": 0.12, "learning_rate": 1.5065509924041252e-05, "loss": 0.9463, "step": 10547 }, { "epoch": 0.12, "learning_rate": 1.5064580534630378e-05, "loss": 0.8768, "step": 10548 }, { "epoch": 0.12, "learning_rate": 1.5063651086377791e-05, "loss": 0.9225, "step": 10549 }, { "epoch": 0.12, "learning_rate": 1.506272157929429e-05, "loss": 0.9031, "step": 10550 }, { "epoch": 0.12, "learning_rate": 1.5061792013390673e-05, "loss": 0.927, "step": 10551 }, { "epoch": 0.12, "learning_rate": 1.5060862388677744e-05, "loss": 0.9225, "step": 10552 }, { "epoch": 0.12, "learning_rate": 1.5059932705166298e-05, "loss": 0.9158, "step": 10553 }, { "epoch": 0.12, "learning_rate": 1.505900296286714e-05, "loss": 0.9229, "step": 10554 }, { "epoch": 0.12, "learning_rate": 1.5058073161791069e-05, "loss": 0.9572, "step": 10555 }, { "epoch": 0.12, "learning_rate": 1.5057143301948892e-05, "loss": 0.9423, "step": 10556 }, { "epoch": 0.12, "learning_rate": 1.505621338335141e-05, "loss": 0.9529, "step": 10557 }, { "epoch": 0.12, "learning_rate": 1.5055283406009427e-05, "loss": 0.8991, "step": 10558 }, { "epoch": 0.12, "learning_rate": 1.505435336993375e-05, "loss": 0.9543, "step": 10559 }, { "epoch": 0.12, "learning_rate": 1.505342327513518e-05, "loss": 0.9187, "step": 10560 }, { "epoch": 0.12, "learning_rate": 1.5052493121624524e-05, "loss": 0.9479, "step": 10561 }, { "epoch": 0.12, "learning_rate": 1.5051562909412591e-05, "loss": 0.9372, "step": 10562 }, { "epoch": 0.12, "learning_rate": 1.505063263851019e-05, "loss": 0.9021, "step": 10563 }, { "epoch": 0.12, "learning_rate": 1.5049702308928125e-05, "loss": 0.9096, "step": 10564 }, { "epoch": 0.12, "learning_rate": 1.5048771920677205e-05, "loss": 0.8899, "step": 10565 }, { "epoch": 0.12, "learning_rate": 1.5047841473768245e-05, "loss": 0.8725, "step": 10566 }, { "epoch": 0.12, "learning_rate": 1.5046910968212048e-05, "loss": 0.9351, "step": 10567 }, { "epoch": 0.12, "learning_rate": 1.5045980404019432e-05, "loss": 0.9118, "step": 10568 }, { "epoch": 0.12, "learning_rate": 1.5045049781201203e-05, "loss": 0.9133, "step": 10569 }, { "epoch": 0.12, "learning_rate": 1.5044119099768173e-05, "loss": 0.905, "step": 10570 }, { "epoch": 0.12, "learning_rate": 1.5043188359731158e-05, "loss": 0.9279, "step": 10571 }, { "epoch": 0.12, "learning_rate": 1.5042257561100973e-05, "loss": 0.9388, "step": 10572 }, { "epoch": 0.12, "learning_rate": 1.5041326703888429e-05, "loss": 0.9073, "step": 10573 }, { "epoch": 0.12, "learning_rate": 1.504039578810434e-05, "loss": 0.8824, "step": 10574 }, { "epoch": 0.12, "learning_rate": 1.5039464813759525e-05, "loss": 0.9006, "step": 10575 }, { "epoch": 0.12, "learning_rate": 1.5038533780864799e-05, "loss": 0.9279, "step": 10576 }, { "epoch": 0.12, "learning_rate": 1.5037602689430975e-05, "loss": 0.8965, "step": 10577 }, { "epoch": 0.12, "learning_rate": 1.5036671539468879e-05, "loss": 0.9159, "step": 10578 }, { "epoch": 0.12, "learning_rate": 1.5035740330989324e-05, "loss": 0.946, "step": 10579 }, { "epoch": 0.12, "learning_rate": 1.503480906400313e-05, "loss": 0.8847, "step": 10580 }, { "epoch": 0.12, "learning_rate": 1.5033877738521114e-05, "loss": 0.9182, "step": 10581 }, { "epoch": 0.12, "learning_rate": 1.50329463545541e-05, "loss": 0.8607, "step": 10582 }, { "epoch": 0.12, "learning_rate": 1.5032014912112907e-05, "loss": 0.9156, "step": 10583 }, { "epoch": 0.12, "learning_rate": 1.503108341120836e-05, "loss": 0.9173, "step": 10584 }, { "epoch": 0.12, "learning_rate": 1.5030151851851279e-05, "loss": 0.9021, "step": 10585 }, { "epoch": 0.12, "learning_rate": 1.5029220234052486e-05, "loss": 0.9177, "step": 10586 }, { "epoch": 0.12, "learning_rate": 1.5028288557822807e-05, "loss": 0.9144, "step": 10587 }, { "epoch": 0.12, "learning_rate": 1.5027356823173067e-05, "loss": 0.905, "step": 10588 }, { "epoch": 0.12, "learning_rate": 1.5026425030114087e-05, "loss": 0.9108, "step": 10589 }, { "epoch": 0.12, "learning_rate": 1.5025493178656695e-05, "loss": 0.9243, "step": 10590 }, { "epoch": 0.12, "learning_rate": 1.5024561268811723e-05, "loss": 0.9827, "step": 10591 }, { "epoch": 0.12, "learning_rate": 1.5023629300589988e-05, "loss": 0.8895, "step": 10592 }, { "epoch": 0.12, "learning_rate": 1.5022697274002327e-05, "loss": 0.9073, "step": 10593 }, { "epoch": 0.12, "learning_rate": 1.5021765189059563e-05, "loss": 0.9327, "step": 10594 }, { "epoch": 0.12, "learning_rate": 1.5020833045772528e-05, "loss": 0.9158, "step": 10595 }, { "epoch": 0.12, "learning_rate": 1.501990084415205e-05, "loss": 0.9174, "step": 10596 }, { "epoch": 0.12, "learning_rate": 1.5018968584208961e-05, "loss": 0.9243, "step": 10597 }, { "epoch": 0.12, "learning_rate": 1.5018036265954091e-05, "loss": 0.9197, "step": 10598 }, { "epoch": 0.12, "learning_rate": 1.5017103889398273e-05, "loss": 0.949, "step": 10599 }, { "epoch": 0.12, "learning_rate": 1.5016171454552338e-05, "loss": 0.8664, "step": 10600 }, { "epoch": 0.12, "learning_rate": 1.5015238961427119e-05, "loss": 0.9396, "step": 10601 }, { "epoch": 0.12, "learning_rate": 1.5014306410033457e-05, "loss": 0.9262, "step": 10602 }, { "epoch": 0.12, "learning_rate": 1.5013373800382178e-05, "loss": 0.9068, "step": 10603 }, { "epoch": 0.12, "learning_rate": 1.501244113248412e-05, "loss": 0.9278, "step": 10604 }, { "epoch": 0.12, "learning_rate": 1.501150840635012e-05, "loss": 0.9183, "step": 10605 }, { "epoch": 0.12, "learning_rate": 1.5010575621991014e-05, "loss": 0.8789, "step": 10606 }, { "epoch": 0.12, "learning_rate": 1.5009642779417639e-05, "loss": 0.9103, "step": 10607 }, { "epoch": 0.12, "learning_rate": 1.5008709878640834e-05, "loss": 0.8668, "step": 10608 }, { "epoch": 0.12, "learning_rate": 1.500777691967144e-05, "loss": 0.9199, "step": 10609 }, { "epoch": 0.12, "learning_rate": 1.5006843902520288e-05, "loss": 0.9036, "step": 10610 }, { "epoch": 0.12, "learning_rate": 1.5005910827198226e-05, "loss": 0.9555, "step": 10611 }, { "epoch": 0.12, "learning_rate": 1.5004977693716094e-05, "loss": 0.9094, "step": 10612 }, { "epoch": 0.12, "learning_rate": 1.5004044502084728e-05, "loss": 0.9338, "step": 10613 }, { "epoch": 0.12, "learning_rate": 1.500311125231498e-05, "loss": 0.8805, "step": 10614 }, { "epoch": 0.12, "learning_rate": 1.5002177944417681e-05, "loss": 0.8995, "step": 10615 }, { "epoch": 0.12, "learning_rate": 1.5001244578403682e-05, "loss": 0.9435, "step": 10616 }, { "epoch": 0.12, "learning_rate": 1.5000311154283824e-05, "loss": 0.9244, "step": 10617 }, { "epoch": 0.12, "learning_rate": 1.4999377672068955e-05, "loss": 0.8832, "step": 10618 }, { "epoch": 0.12, "learning_rate": 1.4998444131769918e-05, "loss": 0.9137, "step": 10619 }, { "epoch": 0.12, "learning_rate": 1.4997510533397559e-05, "loss": 0.9051, "step": 10620 }, { "epoch": 0.12, "learning_rate": 1.4996576876962725e-05, "loss": 0.9132, "step": 10621 }, { "epoch": 0.12, "learning_rate": 1.4995643162476266e-05, "loss": 0.8739, "step": 10622 }, { "epoch": 0.12, "learning_rate": 1.4994709389949024e-05, "loss": 0.8804, "step": 10623 }, { "epoch": 0.12, "learning_rate": 1.4993775559391855e-05, "loss": 0.9315, "step": 10624 }, { "epoch": 0.12, "learning_rate": 1.4992841670815602e-05, "loss": 0.9171, "step": 10625 }, { "epoch": 0.12, "learning_rate": 1.4991907724231122e-05, "loss": 0.9153, "step": 10626 }, { "epoch": 0.12, "learning_rate": 1.499097371964926e-05, "loss": 0.9507, "step": 10627 }, { "epoch": 0.12, "learning_rate": 1.4990039657080871e-05, "loss": 0.874, "step": 10628 }, { "epoch": 0.12, "learning_rate": 1.4989105536536805e-05, "loss": 0.9075, "step": 10629 }, { "epoch": 0.12, "learning_rate": 1.4988171358027919e-05, "loss": 0.8727, "step": 10630 }, { "epoch": 0.12, "learning_rate": 1.498723712156506e-05, "loss": 0.9114, "step": 10631 }, { "epoch": 0.12, "learning_rate": 1.4986302827159086e-05, "loss": 0.892, "step": 10632 }, { "epoch": 0.12, "learning_rate": 1.4985368474820854e-05, "loss": 0.9338, "step": 10633 }, { "epoch": 0.12, "learning_rate": 1.4984434064561215e-05, "loss": 0.87, "step": 10634 }, { "epoch": 0.12, "learning_rate": 1.4983499596391026e-05, "loss": 0.9131, "step": 10635 }, { "epoch": 0.12, "learning_rate": 1.4982565070321147e-05, "loss": 0.9011, "step": 10636 }, { "epoch": 0.12, "learning_rate": 1.4981630486362435e-05, "loss": 0.8784, "step": 10637 }, { "epoch": 0.12, "learning_rate": 1.4980695844525745e-05, "loss": 0.9753, "step": 10638 }, { "epoch": 0.12, "learning_rate": 1.4979761144821941e-05, "loss": 0.9806, "step": 10639 }, { "epoch": 0.12, "learning_rate": 1.4978826387261877e-05, "loss": 0.8809, "step": 10640 }, { "epoch": 0.12, "learning_rate": 1.4977891571856415e-05, "loss": 0.91, "step": 10641 }, { "epoch": 0.12, "learning_rate": 1.4976956698616422e-05, "loss": 0.8823, "step": 10642 }, { "epoch": 0.12, "learning_rate": 1.497602176755275e-05, "loss": 0.8885, "step": 10643 }, { "epoch": 0.12, "learning_rate": 1.4975086778676269e-05, "loss": 0.8891, "step": 10644 }, { "epoch": 0.12, "learning_rate": 1.4974151731997835e-05, "loss": 0.9477, "step": 10645 }, { "epoch": 0.12, "learning_rate": 1.4973216627528318e-05, "loss": 0.918, "step": 10646 }, { "epoch": 0.12, "learning_rate": 1.4972281465278578e-05, "loss": 0.9216, "step": 10647 }, { "epoch": 0.12, "learning_rate": 1.4971346245259484e-05, "loss": 0.9379, "step": 10648 }, { "epoch": 0.12, "learning_rate": 1.4970410967481899e-05, "loss": 0.8963, "step": 10649 }, { "epoch": 0.12, "learning_rate": 1.4969475631956685e-05, "loss": 0.9033, "step": 10650 }, { "epoch": 0.12, "learning_rate": 1.496854023869472e-05, "loss": 0.9278, "step": 10651 }, { "epoch": 0.12, "learning_rate": 1.4967604787706863e-05, "loss": 0.8979, "step": 10652 }, { "epoch": 0.12, "learning_rate": 1.4966669279003983e-05, "loss": 0.9051, "step": 10653 }, { "epoch": 0.12, "learning_rate": 1.496573371259695e-05, "loss": 0.901, "step": 10654 }, { "epoch": 0.12, "learning_rate": 1.4964798088496638e-05, "loss": 0.9317, "step": 10655 }, { "epoch": 0.12, "learning_rate": 1.4963862406713912e-05, "loss": 0.9184, "step": 10656 }, { "epoch": 0.12, "learning_rate": 1.4962926667259643e-05, "loss": 0.9345, "step": 10657 }, { "epoch": 0.12, "learning_rate": 1.4961990870144706e-05, "loss": 0.8865, "step": 10658 }, { "epoch": 0.12, "learning_rate": 1.496105501537997e-05, "loss": 0.9076, "step": 10659 }, { "epoch": 0.12, "learning_rate": 1.4960119102976314e-05, "loss": 0.9075, "step": 10660 }, { "epoch": 0.12, "learning_rate": 1.4959183132944604e-05, "loss": 0.9193, "step": 10661 }, { "epoch": 0.12, "learning_rate": 1.4958247105295718e-05, "loss": 0.9483, "step": 10662 }, { "epoch": 0.12, "learning_rate": 1.495731102004053e-05, "loss": 0.9365, "step": 10663 }, { "epoch": 0.12, "learning_rate": 1.4956374877189919e-05, "loss": 0.9064, "step": 10664 }, { "epoch": 0.12, "learning_rate": 1.4955438676754757e-05, "loss": 0.9422, "step": 10665 }, { "epoch": 0.12, "learning_rate": 1.4954502418745924e-05, "loss": 0.8747, "step": 10666 }, { "epoch": 0.12, "learning_rate": 1.4953566103174297e-05, "loss": 0.98, "step": 10667 }, { "epoch": 0.12, "learning_rate": 1.4952629730050752e-05, "loss": 0.8841, "step": 10668 }, { "epoch": 0.12, "learning_rate": 1.4951693299386172e-05, "loss": 0.9445, "step": 10669 }, { "epoch": 0.12, "learning_rate": 1.4950756811191431e-05, "loss": 0.8771, "step": 10670 }, { "epoch": 0.12, "learning_rate": 1.4949820265477417e-05, "loss": 0.9157, "step": 10671 }, { "epoch": 0.12, "learning_rate": 1.4948883662255007e-05, "loss": 0.9536, "step": 10672 }, { "epoch": 0.12, "learning_rate": 1.4947947001535081e-05, "loss": 0.9477, "step": 10673 }, { "epoch": 0.12, "learning_rate": 1.4947010283328525e-05, "loss": 0.8819, "step": 10674 }, { "epoch": 0.12, "learning_rate": 1.4946073507646219e-05, "loss": 0.9088, "step": 10675 }, { "epoch": 0.12, "learning_rate": 1.494513667449905e-05, "loss": 0.9108, "step": 10676 }, { "epoch": 0.12, "learning_rate": 1.4944199783897899e-05, "loss": 0.8824, "step": 10677 }, { "epoch": 0.12, "learning_rate": 1.4943262835853651e-05, "loss": 0.9044, "step": 10678 }, { "epoch": 0.12, "learning_rate": 1.4942325830377197e-05, "loss": 0.8982, "step": 10679 }, { "epoch": 0.12, "learning_rate": 1.4941388767479417e-05, "loss": 0.8987, "step": 10680 }, { "epoch": 0.12, "learning_rate": 1.4940451647171201e-05, "loss": 0.9331, "step": 10681 }, { "epoch": 0.12, "learning_rate": 1.4939514469463437e-05, "loss": 0.8768, "step": 10682 }, { "epoch": 0.12, "learning_rate": 1.4938577234367013e-05, "loss": 0.8923, "step": 10683 }, { "epoch": 0.12, "learning_rate": 1.4937639941892815e-05, "loss": 0.8811, "step": 10684 }, { "epoch": 0.12, "learning_rate": 1.493670259205174e-05, "loss": 0.9773, "step": 10685 }, { "epoch": 0.12, "learning_rate": 1.4935765184854669e-05, "loss": 0.9296, "step": 10686 }, { "epoch": 0.12, "learning_rate": 1.4934827720312502e-05, "loss": 0.9632, "step": 10687 }, { "epoch": 0.12, "learning_rate": 1.4933890198436127e-05, "loss": 0.929, "step": 10688 }, { "epoch": 0.12, "learning_rate": 1.4932952619236434e-05, "loss": 0.8776, "step": 10689 }, { "epoch": 0.12, "learning_rate": 1.4932014982724319e-05, "loss": 0.8917, "step": 10690 }, { "epoch": 0.12, "learning_rate": 1.4931077288910674e-05, "loss": 0.9522, "step": 10691 }, { "epoch": 0.12, "learning_rate": 1.4930139537806397e-05, "loss": 0.8829, "step": 10692 }, { "epoch": 0.12, "learning_rate": 1.4929201729422375e-05, "loss": 0.9472, "step": 10693 }, { "epoch": 0.12, "learning_rate": 1.4928263863769515e-05, "loss": 0.9153, "step": 10694 }, { "epoch": 0.12, "learning_rate": 1.4927325940858706e-05, "loss": 0.8974, "step": 10695 }, { "epoch": 0.12, "learning_rate": 1.4926387960700843e-05, "loss": 0.9334, "step": 10696 }, { "epoch": 0.12, "learning_rate": 1.4925449923306831e-05, "loss": 0.934, "step": 10697 }, { "epoch": 0.12, "learning_rate": 1.4924511828687565e-05, "loss": 0.9148, "step": 10698 }, { "epoch": 0.12, "learning_rate": 1.492357367685394e-05, "loss": 0.9244, "step": 10699 }, { "epoch": 0.12, "learning_rate": 1.4922635467816862e-05, "loss": 0.9074, "step": 10700 }, { "epoch": 0.12, "learning_rate": 1.4921697201587231e-05, "loss": 0.9305, "step": 10701 }, { "epoch": 0.12, "learning_rate": 1.4920758878175941e-05, "loss": 0.909, "step": 10702 }, { "epoch": 0.12, "learning_rate": 1.4919820497593901e-05, "loss": 0.8862, "step": 10703 }, { "epoch": 0.12, "learning_rate": 1.4918882059852012e-05, "loss": 0.9194, "step": 10704 }, { "epoch": 0.12, "learning_rate": 1.4917943564961173e-05, "loss": 0.9073, "step": 10705 }, { "epoch": 0.12, "learning_rate": 1.4917005012932292e-05, "loss": 0.9123, "step": 10706 }, { "epoch": 0.12, "learning_rate": 1.4916066403776272e-05, "loss": 0.9287, "step": 10707 }, { "epoch": 0.12, "learning_rate": 1.4915127737504017e-05, "loss": 0.8763, "step": 10708 }, { "epoch": 0.12, "learning_rate": 1.4914189014126435e-05, "loss": 0.9057, "step": 10709 }, { "epoch": 0.12, "learning_rate": 1.4913250233654433e-05, "loss": 0.895, "step": 10710 }, { "epoch": 0.12, "learning_rate": 1.4912311396098913e-05, "loss": 0.9122, "step": 10711 }, { "epoch": 0.12, "learning_rate": 1.4911372501470784e-05, "loss": 0.8939, "step": 10712 }, { "epoch": 0.12, "learning_rate": 1.4910433549780962e-05, "loss": 0.9196, "step": 10713 }, { "epoch": 0.12, "learning_rate": 1.4909494541040346e-05, "loss": 0.9053, "step": 10714 }, { "epoch": 0.12, "learning_rate": 1.490855547525985e-05, "loss": 0.8992, "step": 10715 }, { "epoch": 0.12, "learning_rate": 1.4907616352450382e-05, "loss": 0.8986, "step": 10716 }, { "epoch": 0.12, "learning_rate": 1.490667717262286e-05, "loss": 0.9146, "step": 10717 }, { "epoch": 0.12, "learning_rate": 1.4905737935788187e-05, "loss": 0.9146, "step": 10718 }, { "epoch": 0.12, "learning_rate": 1.4904798641957282e-05, "loss": 0.922, "step": 10719 }, { "epoch": 0.12, "learning_rate": 1.4903859291141053e-05, "loss": 0.9253, "step": 10720 }, { "epoch": 0.12, "learning_rate": 1.4902919883350413e-05, "loss": 0.9049, "step": 10721 }, { "epoch": 0.12, "learning_rate": 1.4901980418596282e-05, "loss": 0.8829, "step": 10722 }, { "epoch": 0.12, "learning_rate": 1.4901040896889574e-05, "loss": 0.9206, "step": 10723 }, { "epoch": 0.12, "learning_rate": 1.4900101318241197e-05, "loss": 0.884, "step": 10724 }, { "epoch": 0.12, "learning_rate": 1.4899161682662078e-05, "loss": 0.9138, "step": 10725 }, { "epoch": 0.12, "learning_rate": 1.4898221990163126e-05, "loss": 0.9325, "step": 10726 }, { "epoch": 0.12, "learning_rate": 1.4897282240755262e-05, "loss": 0.9328, "step": 10727 }, { "epoch": 0.12, "learning_rate": 1.4896342434449402e-05, "loss": 0.9175, "step": 10728 }, { "epoch": 0.12, "learning_rate": 1.489540257125647e-05, "loss": 0.924, "step": 10729 }, { "epoch": 0.12, "learning_rate": 1.4894462651187379e-05, "loss": 0.8836, "step": 10730 }, { "epoch": 0.12, "learning_rate": 1.4893522674253053e-05, "loss": 0.9526, "step": 10731 }, { "epoch": 0.12, "learning_rate": 1.4892582640464415e-05, "loss": 0.9122, "step": 10732 }, { "epoch": 0.12, "learning_rate": 1.4891642549832381e-05, "loss": 0.9382, "step": 10733 }, { "epoch": 0.12, "learning_rate": 1.4890702402367879e-05, "loss": 0.9288, "step": 10734 }, { "epoch": 0.12, "learning_rate": 1.4889762198081828e-05, "loss": 0.9229, "step": 10735 }, { "epoch": 0.12, "learning_rate": 1.4888821936985149e-05, "loss": 0.9024, "step": 10736 }, { "epoch": 0.12, "learning_rate": 1.4887881619088773e-05, "loss": 0.9337, "step": 10737 }, { "epoch": 0.12, "learning_rate": 1.4886941244403624e-05, "loss": 0.8979, "step": 10738 }, { "epoch": 0.12, "learning_rate": 1.488600081294062e-05, "loss": 0.962, "step": 10739 }, { "epoch": 0.12, "learning_rate": 1.4885060324710697e-05, "loss": 0.9078, "step": 10740 }, { "epoch": 0.12, "learning_rate": 1.4884119779724774e-05, "loss": 0.9184, "step": 10741 }, { "epoch": 0.12, "learning_rate": 1.4883179177993783e-05, "loss": 0.9167, "step": 10742 }, { "epoch": 0.12, "learning_rate": 1.4882238519528651e-05, "loss": 0.9325, "step": 10743 }, { "epoch": 0.12, "learning_rate": 1.4881297804340306e-05, "loss": 0.909, "step": 10744 }, { "epoch": 0.12, "learning_rate": 1.4880357032439681e-05, "loss": 0.8816, "step": 10745 }, { "epoch": 0.12, "learning_rate": 1.48794162038377e-05, "loss": 0.9053, "step": 10746 }, { "epoch": 0.12, "learning_rate": 1.48784753185453e-05, "loss": 0.9177, "step": 10747 }, { "epoch": 0.12, "learning_rate": 1.4877534376573409e-05, "loss": 0.9427, "step": 10748 }, { "epoch": 0.12, "learning_rate": 1.4876593377932957e-05, "loss": 0.9125, "step": 10749 }, { "epoch": 0.12, "learning_rate": 1.4875652322634884e-05, "loss": 0.8765, "step": 10750 }, { "epoch": 0.12, "learning_rate": 1.4874711210690117e-05, "loss": 0.9021, "step": 10751 }, { "epoch": 0.12, "learning_rate": 1.4873770042109592e-05, "loss": 0.9598, "step": 10752 }, { "epoch": 0.12, "learning_rate": 1.4872828816904244e-05, "loss": 0.8942, "step": 10753 }, { "epoch": 0.12, "learning_rate": 1.4871887535085007e-05, "loss": 0.9344, "step": 10754 }, { "epoch": 0.12, "learning_rate": 1.4870946196662822e-05, "loss": 0.9039, "step": 10755 }, { "epoch": 0.12, "learning_rate": 1.4870004801648618e-05, "loss": 0.8817, "step": 10756 }, { "epoch": 0.13, "learning_rate": 1.4869063350053342e-05, "loss": 0.8967, "step": 10757 }, { "epoch": 0.13, "learning_rate": 1.4868121841887922e-05, "loss": 0.9422, "step": 10758 }, { "epoch": 0.13, "learning_rate": 1.4867180277163305e-05, "loss": 0.9356, "step": 10759 }, { "epoch": 0.13, "learning_rate": 1.486623865589042e-05, "loss": 0.8911, "step": 10760 }, { "epoch": 0.13, "learning_rate": 1.4865296978080219e-05, "loss": 0.913, "step": 10761 }, { "epoch": 0.13, "learning_rate": 1.4864355243743637e-05, "loss": 0.9122, "step": 10762 }, { "epoch": 0.13, "learning_rate": 1.4863413452891613e-05, "loss": 0.9461, "step": 10763 }, { "epoch": 0.13, "learning_rate": 1.4862471605535097e-05, "loss": 0.9139, "step": 10764 }, { "epoch": 0.13, "learning_rate": 1.4861529701685022e-05, "loss": 0.9426, "step": 10765 }, { "epoch": 0.13, "learning_rate": 1.4860587741352337e-05, "loss": 0.9379, "step": 10766 }, { "epoch": 0.13, "learning_rate": 1.4859645724547984e-05, "loss": 0.8924, "step": 10767 }, { "epoch": 0.13, "learning_rate": 1.4858703651282909e-05, "loss": 0.9073, "step": 10768 }, { "epoch": 0.13, "learning_rate": 1.4857761521568054e-05, "loss": 0.9096, "step": 10769 }, { "epoch": 0.13, "learning_rate": 1.485681933541437e-05, "loss": 0.8804, "step": 10770 }, { "epoch": 0.13, "learning_rate": 1.4855877092832802e-05, "loss": 0.9, "step": 10771 }, { "epoch": 0.13, "learning_rate": 1.4854934793834295e-05, "loss": 0.9309, "step": 10772 }, { "epoch": 0.13, "learning_rate": 1.4853992438429797e-05, "loss": 0.9086, "step": 10773 }, { "epoch": 0.13, "learning_rate": 1.4853050026630258e-05, "loss": 0.922, "step": 10774 }, { "epoch": 0.13, "learning_rate": 1.4852107558446627e-05, "loss": 0.9173, "step": 10775 }, { "epoch": 0.13, "learning_rate": 1.4851165033889855e-05, "loss": 0.9795, "step": 10776 }, { "epoch": 0.13, "learning_rate": 1.485022245297089e-05, "loss": 0.931, "step": 10777 }, { "epoch": 0.13, "learning_rate": 1.4849279815700683e-05, "loss": 0.8955, "step": 10778 }, { "epoch": 0.13, "learning_rate": 1.484833712209019e-05, "loss": 0.9294, "step": 10779 }, { "epoch": 0.13, "learning_rate": 1.4847394372150361e-05, "loss": 0.8901, "step": 10780 }, { "epoch": 0.13, "learning_rate": 1.4846451565892146e-05, "loss": 0.8984, "step": 10781 }, { "epoch": 0.13, "learning_rate": 1.4845508703326504e-05, "loss": 0.9193, "step": 10782 }, { "epoch": 0.13, "learning_rate": 1.4844565784464385e-05, "loss": 0.976, "step": 10783 }, { "epoch": 0.13, "learning_rate": 1.4843622809316749e-05, "loss": 0.9133, "step": 10784 }, { "epoch": 0.13, "learning_rate": 1.4842679777894544e-05, "loss": 0.9255, "step": 10785 }, { "epoch": 0.13, "learning_rate": 1.4841736690208736e-05, "loss": 0.8879, "step": 10786 }, { "epoch": 0.13, "learning_rate": 1.4840793546270278e-05, "loss": 0.9313, "step": 10787 }, { "epoch": 0.13, "learning_rate": 1.4839850346090124e-05, "loss": 0.9136, "step": 10788 }, { "epoch": 0.13, "learning_rate": 1.4838907089679238e-05, "loss": 0.8876, "step": 10789 }, { "epoch": 0.13, "learning_rate": 1.4837963777048574e-05, "loss": 0.9053, "step": 10790 }, { "epoch": 0.13, "learning_rate": 1.4837020408209096e-05, "loss": 0.8912, "step": 10791 }, { "epoch": 0.13, "learning_rate": 1.4836076983171758e-05, "loss": 0.9089, "step": 10792 }, { "epoch": 0.13, "learning_rate": 1.4835133501947532e-05, "loss": 0.9703, "step": 10793 }, { "epoch": 0.13, "learning_rate": 1.483418996454737e-05, "loss": 0.8862, "step": 10794 }, { "epoch": 0.13, "learning_rate": 1.4833246370982238e-05, "loss": 0.9331, "step": 10795 }, { "epoch": 0.13, "learning_rate": 1.4832302721263097e-05, "loss": 0.9064, "step": 10796 }, { "epoch": 0.13, "learning_rate": 1.483135901540091e-05, "loss": 0.9532, "step": 10797 }, { "epoch": 0.13, "learning_rate": 1.4830415253406647e-05, "loss": 0.8846, "step": 10798 }, { "epoch": 0.13, "learning_rate": 1.4829471435291267e-05, "loss": 0.9103, "step": 10799 }, { "epoch": 0.13, "learning_rate": 1.4828527561065738e-05, "loss": 0.8764, "step": 10800 }, { "epoch": 0.13, "learning_rate": 1.4827583630741024e-05, "loss": 0.9307, "step": 10801 }, { "epoch": 0.13, "learning_rate": 1.4826639644328098e-05, "loss": 0.8779, "step": 10802 }, { "epoch": 0.13, "learning_rate": 1.4825695601837921e-05, "loss": 0.9169, "step": 10803 }, { "epoch": 0.13, "learning_rate": 1.4824751503281463e-05, "loss": 0.9112, "step": 10804 }, { "epoch": 0.13, "learning_rate": 1.4823807348669694e-05, "loss": 0.9424, "step": 10805 }, { "epoch": 0.13, "learning_rate": 1.4822863138013579e-05, "loss": 0.9211, "step": 10806 }, { "epoch": 0.13, "learning_rate": 1.4821918871324094e-05, "loss": 0.9283, "step": 10807 }, { "epoch": 0.13, "learning_rate": 1.4820974548612208e-05, "loss": 0.8837, "step": 10808 }, { "epoch": 0.13, "learning_rate": 1.482003016988889e-05, "loss": 0.9429, "step": 10809 }, { "epoch": 0.13, "learning_rate": 1.4819085735165117e-05, "loss": 0.911, "step": 10810 }, { "epoch": 0.13, "learning_rate": 1.4818141244451855e-05, "loss": 0.9002, "step": 10811 }, { "epoch": 0.13, "learning_rate": 1.4817196697760082e-05, "loss": 0.9284, "step": 10812 }, { "epoch": 0.13, "learning_rate": 1.4816252095100771e-05, "loss": 0.8943, "step": 10813 }, { "epoch": 0.13, "learning_rate": 1.4815307436484898e-05, "loss": 0.9092, "step": 10814 }, { "epoch": 0.13, "learning_rate": 1.4814362721923434e-05, "loss": 0.8983, "step": 10815 }, { "epoch": 0.13, "learning_rate": 1.481341795142736e-05, "loss": 0.8894, "step": 10816 }, { "epoch": 0.13, "learning_rate": 1.4812473125007652e-05, "loss": 0.9174, "step": 10817 }, { "epoch": 0.13, "learning_rate": 1.4811528242675279e-05, "loss": 0.8865, "step": 10818 }, { "epoch": 0.13, "learning_rate": 1.4810583304441233e-05, "loss": 0.9002, "step": 10819 }, { "epoch": 0.13, "learning_rate": 1.480963831031648e-05, "loss": 0.9411, "step": 10820 }, { "epoch": 0.13, "learning_rate": 1.4808693260312005e-05, "loss": 0.8729, "step": 10821 }, { "epoch": 0.13, "learning_rate": 1.480774815443879e-05, "loss": 0.9133, "step": 10822 }, { "epoch": 0.13, "learning_rate": 1.4806802992707812e-05, "loss": 0.9116, "step": 10823 }, { "epoch": 0.13, "learning_rate": 1.4805857775130052e-05, "loss": 0.9343, "step": 10824 }, { "epoch": 0.13, "learning_rate": 1.480491250171649e-05, "loss": 0.9441, "step": 10825 }, { "epoch": 0.13, "learning_rate": 1.4803967172478118e-05, "loss": 0.9198, "step": 10826 }, { "epoch": 0.13, "learning_rate": 1.4803021787425907e-05, "loss": 0.9142, "step": 10827 }, { "epoch": 0.13, "learning_rate": 1.4802076346570847e-05, "loss": 0.8983, "step": 10828 }, { "epoch": 0.13, "learning_rate": 1.480113084992392e-05, "loss": 0.8855, "step": 10829 }, { "epoch": 0.13, "learning_rate": 1.4800185297496115e-05, "loss": 0.9392, "step": 10830 }, { "epoch": 0.13, "learning_rate": 1.479923968929841e-05, "loss": 0.928, "step": 10831 }, { "epoch": 0.13, "learning_rate": 1.4798294025341802e-05, "loss": 0.9333, "step": 10832 }, { "epoch": 0.13, "learning_rate": 1.479734830563727e-05, "loss": 0.8786, "step": 10833 }, { "epoch": 0.13, "learning_rate": 1.4796402530195804e-05, "loss": 0.9136, "step": 10834 }, { "epoch": 0.13, "learning_rate": 1.4795456699028394e-05, "loss": 0.891, "step": 10835 }, { "epoch": 0.13, "learning_rate": 1.4794510812146028e-05, "loss": 0.8964, "step": 10836 }, { "epoch": 0.13, "learning_rate": 1.4793564869559692e-05, "loss": 0.8961, "step": 10837 }, { "epoch": 0.13, "learning_rate": 1.479261887128038e-05, "loss": 0.8948, "step": 10838 }, { "epoch": 0.13, "learning_rate": 1.4791672817319083e-05, "loss": 0.9568, "step": 10839 }, { "epoch": 0.13, "learning_rate": 1.479072670768679e-05, "loss": 0.9294, "step": 10840 }, { "epoch": 0.13, "learning_rate": 1.4789780542394494e-05, "loss": 0.9077, "step": 10841 }, { "epoch": 0.13, "learning_rate": 1.4788834321453189e-05, "loss": 0.8703, "step": 10842 }, { "epoch": 0.13, "learning_rate": 1.4787888044873868e-05, "loss": 0.9492, "step": 10843 }, { "epoch": 0.13, "learning_rate": 1.4786941712667526e-05, "loss": 0.9155, "step": 10844 }, { "epoch": 0.13, "learning_rate": 1.4785995324845157e-05, "loss": 0.9475, "step": 10845 }, { "epoch": 0.13, "learning_rate": 1.4785048881417755e-05, "loss": 0.9128, "step": 10846 }, { "epoch": 0.13, "learning_rate": 1.4784102382396314e-05, "loss": 0.89, "step": 10847 }, { "epoch": 0.13, "learning_rate": 1.478315582779184e-05, "loss": 0.8686, "step": 10848 }, { "epoch": 0.13, "learning_rate": 1.4782209217615322e-05, "loss": 0.9217, "step": 10849 }, { "epoch": 0.13, "learning_rate": 1.478126255187776e-05, "loss": 0.883, "step": 10850 }, { "epoch": 0.13, "learning_rate": 1.4780315830590153e-05, "loss": 0.919, "step": 10851 }, { "epoch": 0.13, "learning_rate": 1.4779369053763497e-05, "loss": 0.8812, "step": 10852 }, { "epoch": 0.13, "learning_rate": 1.4778422221408799e-05, "loss": 0.9203, "step": 10853 }, { "epoch": 0.13, "learning_rate": 1.4777475333537052e-05, "loss": 0.932, "step": 10854 }, { "epoch": 0.13, "learning_rate": 1.4776528390159264e-05, "loss": 0.9113, "step": 10855 }, { "epoch": 0.13, "learning_rate": 1.4775581391286436e-05, "loss": 0.9209, "step": 10856 }, { "epoch": 0.13, "learning_rate": 1.4774634336929563e-05, "loss": 0.9161, "step": 10857 }, { "epoch": 0.13, "learning_rate": 1.4773687227099655e-05, "loss": 0.895, "step": 10858 }, { "epoch": 0.13, "learning_rate": 1.4772740061807714e-05, "loss": 0.9388, "step": 10859 }, { "epoch": 0.13, "learning_rate": 1.4771792841064744e-05, "loss": 0.8873, "step": 10860 }, { "epoch": 0.13, "learning_rate": 1.4770845564881753e-05, "loss": 0.9474, "step": 10861 }, { "epoch": 0.13, "learning_rate": 1.4769898233269742e-05, "loss": 0.9148, "step": 10862 }, { "epoch": 0.13, "learning_rate": 1.476895084623972e-05, "loss": 0.9059, "step": 10863 }, { "epoch": 0.13, "learning_rate": 1.4768003403802693e-05, "loss": 0.8974, "step": 10864 }, { "epoch": 0.13, "learning_rate": 1.4767055905969669e-05, "loss": 0.9053, "step": 10865 }, { "epoch": 0.13, "learning_rate": 1.4766108352751661e-05, "loss": 0.9505, "step": 10866 }, { "epoch": 0.13, "learning_rate": 1.4765160744159667e-05, "loss": 0.8995, "step": 10867 }, { "epoch": 0.13, "learning_rate": 1.4764213080204708e-05, "loss": 0.8832, "step": 10868 }, { "epoch": 0.13, "learning_rate": 1.4763265360897787e-05, "loss": 0.9305, "step": 10869 }, { "epoch": 0.13, "learning_rate": 1.476231758624992e-05, "loss": 0.9478, "step": 10870 }, { "epoch": 0.13, "learning_rate": 1.476136975627211e-05, "loss": 0.9484, "step": 10871 }, { "epoch": 0.13, "learning_rate": 1.4760421870975381e-05, "loss": 0.8897, "step": 10872 }, { "epoch": 0.13, "learning_rate": 1.4759473930370738e-05, "loss": 0.9161, "step": 10873 }, { "epoch": 0.13, "learning_rate": 1.4758525934469195e-05, "loss": 0.9186, "step": 10874 }, { "epoch": 0.13, "learning_rate": 1.4757577883281766e-05, "loss": 0.8887, "step": 10875 }, { "epoch": 0.13, "learning_rate": 1.475662977681947e-05, "loss": 0.9147, "step": 10876 }, { "epoch": 0.13, "learning_rate": 1.4755681615093315e-05, "loss": 0.9211, "step": 10877 }, { "epoch": 0.13, "learning_rate": 1.4754733398114326e-05, "loss": 0.8966, "step": 10878 }, { "epoch": 0.13, "learning_rate": 1.4753785125893514e-05, "loss": 0.923, "step": 10879 }, { "epoch": 0.13, "learning_rate": 1.4752836798441897e-05, "loss": 0.8952, "step": 10880 }, { "epoch": 0.13, "learning_rate": 1.4751888415770492e-05, "loss": 0.9269, "step": 10881 }, { "epoch": 0.13, "learning_rate": 1.475093997789032e-05, "loss": 0.9019, "step": 10882 }, { "epoch": 0.13, "learning_rate": 1.4749991484812399e-05, "loss": 0.925, "step": 10883 }, { "epoch": 0.13, "learning_rate": 1.4749042936547745e-05, "loss": 0.8934, "step": 10884 }, { "epoch": 0.13, "learning_rate": 1.4748094333107388e-05, "loss": 0.8926, "step": 10885 }, { "epoch": 0.13, "learning_rate": 1.474714567450234e-05, "loss": 0.9186, "step": 10886 }, { "epoch": 0.13, "learning_rate": 1.4746196960743626e-05, "loss": 0.9475, "step": 10887 }, { "epoch": 0.13, "learning_rate": 1.4745248191842272e-05, "loss": 0.8523, "step": 10888 }, { "epoch": 0.13, "learning_rate": 1.4744299367809291e-05, "loss": 0.9344, "step": 10889 }, { "epoch": 0.13, "learning_rate": 1.4743350488655721e-05, "loss": 0.9611, "step": 10890 }, { "epoch": 0.13, "learning_rate": 1.4742401554392575e-05, "loss": 0.8983, "step": 10891 }, { "epoch": 0.13, "learning_rate": 1.4741452565030883e-05, "loss": 0.8975, "step": 10892 }, { "epoch": 0.13, "learning_rate": 1.4740503520581667e-05, "loss": 0.9195, "step": 10893 }, { "epoch": 0.13, "learning_rate": 1.473955442105596e-05, "loss": 0.8792, "step": 10894 }, { "epoch": 0.13, "learning_rate": 1.4738605266464783e-05, "loss": 0.9258, "step": 10895 }, { "epoch": 0.13, "learning_rate": 1.4737656056819162e-05, "loss": 0.9527, "step": 10896 }, { "epoch": 0.13, "learning_rate": 1.4736706792130134e-05, "loss": 0.9467, "step": 10897 }, { "epoch": 0.13, "learning_rate": 1.4735757472408714e-05, "loss": 0.8978, "step": 10898 }, { "epoch": 0.13, "learning_rate": 1.4734808097665947e-05, "loss": 0.9343, "step": 10899 }, { "epoch": 0.13, "learning_rate": 1.473385866791285e-05, "loss": 0.9057, "step": 10900 }, { "epoch": 0.13, "learning_rate": 1.4732909183160462e-05, "loss": 0.8835, "step": 10901 }, { "epoch": 0.13, "learning_rate": 1.4731959643419815e-05, "loss": 0.9179, "step": 10902 }, { "epoch": 0.13, "learning_rate": 1.4731010048701935e-05, "loss": 0.8935, "step": 10903 }, { "epoch": 0.13, "learning_rate": 1.4730060399017855e-05, "loss": 0.8745, "step": 10904 }, { "epoch": 0.13, "learning_rate": 1.4729110694378613e-05, "loss": 0.9288, "step": 10905 }, { "epoch": 0.13, "learning_rate": 1.4728160934795242e-05, "loss": 0.8742, "step": 10906 }, { "epoch": 0.13, "learning_rate": 1.4727211120278772e-05, "loss": 0.9365, "step": 10907 }, { "epoch": 0.13, "learning_rate": 1.4726261250840244e-05, "loss": 0.9084, "step": 10908 }, { "epoch": 0.13, "learning_rate": 1.4725311326490692e-05, "loss": 0.8801, "step": 10909 }, { "epoch": 0.13, "learning_rate": 1.4724361347241147e-05, "loss": 0.881, "step": 10910 }, { "epoch": 0.13, "learning_rate": 1.4723411313102658e-05, "loss": 0.9428, "step": 10911 }, { "epoch": 0.13, "learning_rate": 1.472246122408625e-05, "loss": 0.9339, "step": 10912 }, { "epoch": 0.13, "learning_rate": 1.4721511080202971e-05, "loss": 0.85, "step": 10913 }, { "epoch": 0.13, "learning_rate": 1.4720560881463852e-05, "loss": 0.8787, "step": 10914 }, { "epoch": 0.13, "learning_rate": 1.4719610627879941e-05, "loss": 0.8955, "step": 10915 }, { "epoch": 0.13, "learning_rate": 1.4718660319462272e-05, "loss": 0.8949, "step": 10916 }, { "epoch": 0.13, "learning_rate": 1.4717709956221889e-05, "loss": 0.9557, "step": 10917 }, { "epoch": 0.13, "learning_rate": 1.471675953816983e-05, "loss": 0.9238, "step": 10918 }, { "epoch": 0.13, "learning_rate": 1.4715809065317143e-05, "loss": 0.9319, "step": 10919 }, { "epoch": 0.13, "learning_rate": 1.4714858537674867e-05, "loss": 0.9021, "step": 10920 }, { "epoch": 0.13, "learning_rate": 1.4713907955254045e-05, "loss": 0.8985, "step": 10921 }, { "epoch": 0.13, "learning_rate": 1.4712957318065724e-05, "loss": 0.9133, "step": 10922 }, { "epoch": 0.13, "learning_rate": 1.4712006626120944e-05, "loss": 0.9154, "step": 10923 }, { "epoch": 0.13, "learning_rate": 1.4711055879430757e-05, "loss": 0.9161, "step": 10924 }, { "epoch": 0.13, "learning_rate": 1.4710105078006205e-05, "loss": 0.9705, "step": 10925 }, { "epoch": 0.13, "learning_rate": 1.4709154221858334e-05, "loss": 0.8559, "step": 10926 }, { "epoch": 0.13, "learning_rate": 1.4708203310998195e-05, "loss": 0.8992, "step": 10927 }, { "epoch": 0.13, "learning_rate": 1.470725234543683e-05, "loss": 0.8893, "step": 10928 }, { "epoch": 0.13, "learning_rate": 1.4706301325185294e-05, "loss": 0.9064, "step": 10929 }, { "epoch": 0.13, "learning_rate": 1.470535025025463e-05, "loss": 0.8795, "step": 10930 }, { "epoch": 0.13, "learning_rate": 1.4704399120655896e-05, "loss": 0.9557, "step": 10931 }, { "epoch": 0.13, "learning_rate": 1.4703447936400135e-05, "loss": 0.9077, "step": 10932 }, { "epoch": 0.13, "learning_rate": 1.47024966974984e-05, "loss": 0.9253, "step": 10933 }, { "epoch": 0.13, "learning_rate": 1.4701545403961745e-05, "loss": 0.9215, "step": 10934 }, { "epoch": 0.13, "learning_rate": 1.4700594055801219e-05, "loss": 0.905, "step": 10935 }, { "epoch": 0.13, "learning_rate": 1.469964265302788e-05, "loss": 0.9317, "step": 10936 }, { "epoch": 0.13, "learning_rate": 1.4698691195652775e-05, "loss": 0.9124, "step": 10937 }, { "epoch": 0.13, "learning_rate": 1.4697739683686965e-05, "loss": 0.9363, "step": 10938 }, { "epoch": 0.13, "learning_rate": 1.4696788117141499e-05, "loss": 0.932, "step": 10939 }, { "epoch": 0.13, "learning_rate": 1.4695836496027438e-05, "loss": 0.967, "step": 10940 }, { "epoch": 0.13, "learning_rate": 1.4694884820355834e-05, "loss": 0.8878, "step": 10941 }, { "epoch": 0.13, "learning_rate": 1.4693933090137747e-05, "loss": 0.9091, "step": 10942 }, { "epoch": 0.13, "learning_rate": 1.4692981305384234e-05, "loss": 0.9219, "step": 10943 }, { "epoch": 0.13, "learning_rate": 1.4692029466106348e-05, "loss": 0.8709, "step": 10944 }, { "epoch": 0.13, "learning_rate": 1.4691077572315156e-05, "loss": 0.9388, "step": 10945 }, { "epoch": 0.13, "learning_rate": 1.4690125624021709e-05, "loss": 0.9471, "step": 10946 }, { "epoch": 0.13, "learning_rate": 1.4689173621237073e-05, "loss": 0.9283, "step": 10947 }, { "epoch": 0.13, "learning_rate": 1.4688221563972307e-05, "loss": 0.9142, "step": 10948 }, { "epoch": 0.13, "learning_rate": 1.4687269452238473e-05, "loss": 0.9098, "step": 10949 }, { "epoch": 0.13, "learning_rate": 1.468631728604663e-05, "loss": 0.9253, "step": 10950 }, { "epoch": 0.13, "learning_rate": 1.4685365065407845e-05, "loss": 0.9122, "step": 10951 }, { "epoch": 0.13, "learning_rate": 1.4684412790333177e-05, "loss": 0.9007, "step": 10952 }, { "epoch": 0.13, "learning_rate": 1.4683460460833687e-05, "loss": 0.9483, "step": 10953 }, { "epoch": 0.13, "learning_rate": 1.4682508076920452e-05, "loss": 0.8884, "step": 10954 }, { "epoch": 0.13, "learning_rate": 1.4681555638604526e-05, "loss": 0.9569, "step": 10955 }, { "epoch": 0.13, "learning_rate": 1.4680603145896974e-05, "loss": 0.9073, "step": 10956 }, { "epoch": 0.13, "learning_rate": 1.467965059880887e-05, "loss": 0.9198, "step": 10957 }, { "epoch": 0.13, "learning_rate": 1.4678697997351277e-05, "loss": 0.9133, "step": 10958 }, { "epoch": 0.13, "learning_rate": 1.4677745341535258e-05, "loss": 0.8994, "step": 10959 }, { "epoch": 0.13, "learning_rate": 1.467679263137189e-05, "loss": 0.911, "step": 10960 }, { "epoch": 0.13, "learning_rate": 1.4675839866872239e-05, "loss": 0.9269, "step": 10961 }, { "epoch": 0.13, "learning_rate": 1.4674887048047368e-05, "loss": 0.9445, "step": 10962 }, { "epoch": 0.13, "learning_rate": 1.4673934174908355e-05, "loss": 0.9705, "step": 10963 }, { "epoch": 0.13, "learning_rate": 1.4672981247466269e-05, "loss": 0.9291, "step": 10964 }, { "epoch": 0.13, "learning_rate": 1.4672028265732178e-05, "loss": 0.9318, "step": 10965 }, { "epoch": 0.13, "learning_rate": 1.4671075229717157e-05, "loss": 0.882, "step": 10966 }, { "epoch": 0.13, "learning_rate": 1.467012213943228e-05, "loss": 0.9027, "step": 10967 }, { "epoch": 0.13, "learning_rate": 1.4669168994888614e-05, "loss": 0.8798, "step": 10968 }, { "epoch": 0.13, "learning_rate": 1.466821579609724e-05, "loss": 0.8943, "step": 10969 }, { "epoch": 0.13, "learning_rate": 1.4667262543069232e-05, "loss": 0.8991, "step": 10970 }, { "epoch": 0.13, "learning_rate": 1.4666309235815658e-05, "loss": 0.9439, "step": 10971 }, { "epoch": 0.13, "learning_rate": 1.46653558743476e-05, "loss": 0.9076, "step": 10972 }, { "epoch": 0.13, "learning_rate": 1.4664402458676134e-05, "loss": 0.892, "step": 10973 }, { "epoch": 0.13, "learning_rate": 1.4663448988812336e-05, "loss": 0.9066, "step": 10974 }, { "epoch": 0.13, "learning_rate": 1.4662495464767283e-05, "loss": 0.9752, "step": 10975 }, { "epoch": 0.13, "learning_rate": 1.4661541886552053e-05, "loss": 0.8614, "step": 10976 }, { "epoch": 0.13, "learning_rate": 1.466058825417773e-05, "loss": 0.8703, "step": 10977 }, { "epoch": 0.13, "learning_rate": 1.4659634567655385e-05, "loss": 0.9389, "step": 10978 }, { "epoch": 0.13, "learning_rate": 1.4658680826996106e-05, "loss": 0.9353, "step": 10979 }, { "epoch": 0.13, "learning_rate": 1.4657727032210971e-05, "loss": 0.905, "step": 10980 }, { "epoch": 0.13, "learning_rate": 1.4656773183311057e-05, "loss": 0.9666, "step": 10981 }, { "epoch": 0.13, "learning_rate": 1.4655819280307453e-05, "loss": 0.9266, "step": 10982 }, { "epoch": 0.13, "learning_rate": 1.4654865323211239e-05, "loss": 0.9666, "step": 10983 }, { "epoch": 0.13, "learning_rate": 1.4653911312033496e-05, "loss": 0.8924, "step": 10984 }, { "epoch": 0.13, "learning_rate": 1.4652957246785312e-05, "loss": 0.9184, "step": 10985 }, { "epoch": 0.13, "learning_rate": 1.465200312747777e-05, "loss": 0.9003, "step": 10986 }, { "epoch": 0.13, "learning_rate": 1.4651048954121954e-05, "loss": 0.9255, "step": 10987 }, { "epoch": 0.13, "learning_rate": 1.465009472672895e-05, "loss": 0.9311, "step": 10988 }, { "epoch": 0.13, "learning_rate": 1.4649140445309848e-05, "loss": 0.915, "step": 10989 }, { "epoch": 0.13, "learning_rate": 1.4648186109875728e-05, "loss": 0.9025, "step": 10990 }, { "epoch": 0.13, "learning_rate": 1.4647231720437687e-05, "loss": 0.9135, "step": 10991 }, { "epoch": 0.13, "learning_rate": 1.4646277277006805e-05, "loss": 0.9148, "step": 10992 }, { "epoch": 0.13, "learning_rate": 1.4645322779594175e-05, "loss": 0.9169, "step": 10993 }, { "epoch": 0.13, "learning_rate": 1.464436822821089e-05, "loss": 0.9018, "step": 10994 }, { "epoch": 0.13, "learning_rate": 1.4643413622868033e-05, "loss": 0.9239, "step": 10995 }, { "epoch": 0.13, "learning_rate": 1.4642458963576698e-05, "loss": 0.9105, "step": 10996 }, { "epoch": 0.13, "learning_rate": 1.4641504250347978e-05, "loss": 0.9208, "step": 10997 }, { "epoch": 0.13, "learning_rate": 1.4640549483192966e-05, "loss": 0.9443, "step": 10998 }, { "epoch": 0.13, "learning_rate": 1.4639594662122744e-05, "loss": 0.9442, "step": 10999 }, { "epoch": 0.13, "learning_rate": 1.4638639787148423e-05, "loss": 0.8775, "step": 11000 }, { "epoch": 0.0, "learning_rate": 1.4637684858281084e-05, "loss": 0.8894, "step": 11001 }, { "epoch": 0.0, "learning_rate": 1.4636729875531826e-05, "loss": 0.891, "step": 11002 }, { "epoch": 0.0, "learning_rate": 1.4635774838911745e-05, "loss": 0.9228, "step": 11003 }, { "epoch": 0.0, "learning_rate": 1.4634819748431934e-05, "loss": 0.919, "step": 11004 }, { "epoch": 0.0, "learning_rate": 1.4633864604103494e-05, "loss": 0.9022, "step": 11005 }, { "epoch": 0.0, "learning_rate": 1.4632909405937516e-05, "loss": 0.8954, "step": 11006 }, { "epoch": 0.0, "learning_rate": 1.4631954153945105e-05, "loss": 0.9558, "step": 11007 }, { "epoch": 0.0, "learning_rate": 1.4630998848137352e-05, "loss": 0.9318, "step": 11008 }, { "epoch": 0.0, "learning_rate": 1.4630043488525361e-05, "loss": 0.9174, "step": 11009 }, { "epoch": 0.0, "learning_rate": 1.462908807512023e-05, "loss": 0.9792, "step": 11010 }, { "epoch": 0.0, "learning_rate": 1.4628132607933058e-05, "loss": 0.9918, "step": 11011 }, { "epoch": 0.0, "learning_rate": 1.462717708697495e-05, "loss": 0.952, "step": 11012 }, { "epoch": 0.0, "learning_rate": 1.4626221512257003e-05, "loss": 0.8739, "step": 11013 }, { "epoch": 0.0, "learning_rate": 1.4625265883790323e-05, "loss": 0.9699, "step": 11014 }, { "epoch": 0.0, "learning_rate": 1.4624310201586007e-05, "loss": 0.9046, "step": 11015 }, { "epoch": 0.0, "learning_rate": 1.4623354465655166e-05, "loss": 0.9559, "step": 11016 }, { "epoch": 0.0, "learning_rate": 1.4622398676008898e-05, "loss": 0.9654, "step": 11017 }, { "epoch": 0.0, "learning_rate": 1.462144283265831e-05, "loss": 0.9721, "step": 11018 }, { "epoch": 0.0, "learning_rate": 1.4620486935614508e-05, "loss": 0.9455, "step": 11019 }, { "epoch": 0.0, "learning_rate": 1.4619530984888596e-05, "loss": 0.9225, "step": 11020 }, { "epoch": 0.0, "learning_rate": 1.4618574980491682e-05, "loss": 0.9395, "step": 11021 }, { "epoch": 0.0, "learning_rate": 1.4617618922434873e-05, "loss": 0.9465, "step": 11022 }, { "epoch": 0.0, "learning_rate": 1.4616662810729276e-05, "loss": 0.9083, "step": 11023 }, { "epoch": 0.0, "learning_rate": 1.4615706645385996e-05, "loss": 0.9319, "step": 11024 }, { "epoch": 0.0, "learning_rate": 1.4614750426416151e-05, "loss": 0.9628, "step": 11025 }, { "epoch": 0.0, "learning_rate": 1.4613794153830842e-05, "loss": 0.9496, "step": 11026 }, { "epoch": 0.0, "learning_rate": 1.4612837827641185e-05, "loss": 0.9663, "step": 11027 }, { "epoch": 0.0, "learning_rate": 1.4611881447858287e-05, "loss": 0.9707, "step": 11028 }, { "epoch": 0.0, "learning_rate": 1.461092501449326e-05, "loss": 0.9833, "step": 11029 }, { "epoch": 0.0, "learning_rate": 1.4609968527557217e-05, "loss": 0.9518, "step": 11030 }, { "epoch": 0.0, "learning_rate": 1.460901198706127e-05, "loss": 0.9924, "step": 11031 }, { "epoch": 0.0, "learning_rate": 1.4608055393016539e-05, "loss": 0.93, "step": 11032 }, { "epoch": 0.0, "learning_rate": 1.4607098745434126e-05, "loss": 0.9841, "step": 11033 }, { "epoch": 0.0, "learning_rate": 1.4606142044325152e-05, "loss": 0.9262, "step": 11034 }, { "epoch": 0.0, "learning_rate": 1.4605185289700734e-05, "loss": 0.9652, "step": 11035 }, { "epoch": 0.0, "learning_rate": 1.4604228481571983e-05, "loss": 1.006, "step": 11036 }, { "epoch": 0.0, "learning_rate": 1.4603271619950021e-05, "loss": 0.9555, "step": 11037 }, { "epoch": 0.0, "learning_rate": 1.460231470484596e-05, "loss": 0.9635, "step": 11038 }, { "epoch": 0.0, "learning_rate": 1.460135773627092e-05, "loss": 0.9439, "step": 11039 }, { "epoch": 0.0, "learning_rate": 1.4600400714236023e-05, "loss": 0.9433, "step": 11040 }, { "epoch": 0.0, "learning_rate": 1.459944363875238e-05, "loss": 0.9294, "step": 11041 }, { "epoch": 0.0, "learning_rate": 1.4598486509831116e-05, "loss": 0.9306, "step": 11042 }, { "epoch": 0.0, "learning_rate": 1.459752932748335e-05, "loss": 0.9176, "step": 11043 }, { "epoch": 0.0, "learning_rate": 1.4596572091720204e-05, "loss": 0.9402, "step": 11044 }, { "epoch": 0.0, "learning_rate": 1.4595614802552795e-05, "loss": 0.9566, "step": 11045 }, { "epoch": 0.0, "learning_rate": 1.4594657459992252e-05, "loss": 0.967, "step": 11046 }, { "epoch": 0.0, "learning_rate": 1.4593700064049691e-05, "loss": 0.9906, "step": 11047 }, { "epoch": 0.0, "learning_rate": 1.459274261473624e-05, "loss": 0.9468, "step": 11048 }, { "epoch": 0.0, "learning_rate": 1.4591785112063023e-05, "loss": 0.9518, "step": 11049 }, { "epoch": 0.0, "learning_rate": 1.4590827556041158e-05, "loss": 0.9631, "step": 11050 }, { "epoch": 0.0, "learning_rate": 1.4589869946681778e-05, "loss": 0.9691, "step": 11051 }, { "epoch": 0.0, "learning_rate": 1.4588912283996007e-05, "loss": 0.9306, "step": 11052 }, { "epoch": 0.0, "learning_rate": 1.4587954567994969e-05, "loss": 0.947, "step": 11053 }, { "epoch": 0.0, "learning_rate": 1.458699679868979e-05, "loss": 0.9376, "step": 11054 }, { "epoch": 0.0, "learning_rate": 1.4586038976091601e-05, "loss": 1.0074, "step": 11055 }, { "epoch": 0.0, "learning_rate": 1.458508110021153e-05, "loss": 0.9714, "step": 11056 }, { "epoch": 0.0, "learning_rate": 1.45841231710607e-05, "loss": 0.9465, "step": 11057 }, { "epoch": 0.0, "learning_rate": 1.458316518865025e-05, "loss": 0.923, "step": 11058 }, { "epoch": 0.0, "learning_rate": 1.4582207152991303e-05, "loss": 0.9397, "step": 11059 }, { "epoch": 0.0, "learning_rate": 1.4581249064094993e-05, "loss": 0.9549, "step": 11060 }, { "epoch": 0.0, "learning_rate": 1.458029092197245e-05, "loss": 0.9059, "step": 11061 }, { "epoch": 0.0, "learning_rate": 1.4579332726634809e-05, "loss": 0.9415, "step": 11062 }, { "epoch": 0.0, "learning_rate": 1.4578374478093198e-05, "loss": 0.9823, "step": 11063 }, { "epoch": 0.0, "learning_rate": 1.4577416176358752e-05, "loss": 0.9514, "step": 11064 }, { "epoch": 0.0, "learning_rate": 1.4576457821442607e-05, "loss": 0.9684, "step": 11065 }, { "epoch": 0.0, "learning_rate": 1.4575499413355894e-05, "loss": 0.9639, "step": 11066 }, { "epoch": 0.0, "learning_rate": 1.4574540952109748e-05, "loss": 0.9772, "step": 11067 }, { "epoch": 0.0, "learning_rate": 1.4573582437715309e-05, "loss": 0.9457, "step": 11068 }, { "epoch": 0.0, "learning_rate": 1.457262387018371e-05, "loss": 0.913, "step": 11069 }, { "epoch": 0.0, "learning_rate": 1.4571665249526083e-05, "loss": 0.9478, "step": 11070 }, { "epoch": 0.0, "learning_rate": 1.4570706575753577e-05, "loss": 0.9674, "step": 11071 }, { "epoch": 0.0, "learning_rate": 1.456974784887732e-05, "loss": 0.9463, "step": 11072 }, { "epoch": 0.0, "learning_rate": 1.4568789068908458e-05, "loss": 0.9442, "step": 11073 }, { "epoch": 0.0, "learning_rate": 1.4567830235858127e-05, "loss": 0.9905, "step": 11074 }, { "epoch": 0.0, "learning_rate": 1.4566871349737466e-05, "loss": 0.9284, "step": 11075 }, { "epoch": 0.0, "learning_rate": 1.4565912410557617e-05, "loss": 0.9899, "step": 11076 }, { "epoch": 0.0, "learning_rate": 1.4564953418329722e-05, "loss": 0.9371, "step": 11077 }, { "epoch": 0.0, "learning_rate": 1.456399437306492e-05, "loss": 0.9618, "step": 11078 }, { "epoch": 0.0, "learning_rate": 1.4563035274774358e-05, "loss": 0.9865, "step": 11079 }, { "epoch": 0.0, "learning_rate": 1.4562076123469174e-05, "loss": 0.9507, "step": 11080 }, { "epoch": 0.0, "learning_rate": 1.4561116919160514e-05, "loss": 0.9846, "step": 11081 }, { "epoch": 0.0, "learning_rate": 1.456015766185952e-05, "loss": 0.9765, "step": 11082 }, { "epoch": 0.0, "learning_rate": 1.4559198351577344e-05, "loss": 0.968, "step": 11083 }, { "epoch": 0.0, "learning_rate": 1.4558238988325124e-05, "loss": 0.9391, "step": 11084 }, { "epoch": 0.0, "learning_rate": 1.455727957211401e-05, "loss": 0.9854, "step": 11085 }, { "epoch": 0.0, "learning_rate": 1.4556320102955148e-05, "loss": 0.9631, "step": 11086 }, { "epoch": 0.0, "learning_rate": 1.4555360580859686e-05, "loss": 0.9415, "step": 11087 }, { "epoch": 0.0, "learning_rate": 1.4554401005838768e-05, "loss": 0.9358, "step": 11088 }, { "epoch": 0.0, "learning_rate": 1.4553441377903547e-05, "loss": 0.9569, "step": 11089 }, { "epoch": 0.0, "learning_rate": 1.4552481697065172e-05, "loss": 0.9374, "step": 11090 }, { "epoch": 0.0, "learning_rate": 1.455152196333479e-05, "loss": 0.983, "step": 11091 }, { "epoch": 0.0, "learning_rate": 1.4550562176723555e-05, "loss": 0.945, "step": 11092 }, { "epoch": 0.0, "learning_rate": 1.4549602337242616e-05, "loss": 0.9845, "step": 11093 }, { "epoch": 0.0, "learning_rate": 1.4548642444903122e-05, "loss": 0.8924, "step": 11094 }, { "epoch": 0.0, "learning_rate": 1.4547682499716234e-05, "loss": 0.9653, "step": 11095 }, { "epoch": 0.0, "learning_rate": 1.4546722501693096e-05, "loss": 0.9191, "step": 11096 }, { "epoch": 0.0, "learning_rate": 1.4545762450844865e-05, "loss": 0.9473, "step": 11097 }, { "epoch": 0.0, "learning_rate": 1.4544802347182692e-05, "loss": 0.9136, "step": 11098 }, { "epoch": 0.0, "learning_rate": 1.4543842190717738e-05, "loss": 0.9822, "step": 11099 }, { "epoch": 0.0, "learning_rate": 1.4542881981461154e-05, "loss": 0.9584, "step": 11100 }, { "epoch": 0.0, "learning_rate": 1.4541921719424098e-05, "loss": 0.9829, "step": 11101 }, { "epoch": 0.0, "learning_rate": 1.4540961404617725e-05, "loss": 0.9626, "step": 11102 }, { "epoch": 0.0, "learning_rate": 1.4540001037053191e-05, "loss": 0.9545, "step": 11103 }, { "epoch": 0.0, "learning_rate": 1.4539040616741657e-05, "loss": 0.9296, "step": 11104 }, { "epoch": 0.0, "learning_rate": 1.453808014369428e-05, "loss": 0.9564, "step": 11105 }, { "epoch": 0.0, "learning_rate": 1.453711961792222e-05, "loss": 0.9136, "step": 11106 }, { "epoch": 0.0, "learning_rate": 1.4536159039436632e-05, "loss": 1.0012, "step": 11107 }, { "epoch": 0.0, "learning_rate": 1.4535198408248685e-05, "loss": 0.9712, "step": 11108 }, { "epoch": 0.0, "learning_rate": 1.4534237724369534e-05, "loss": 0.957, "step": 11109 }, { "epoch": 0.0, "learning_rate": 1.453327698781034e-05, "loss": 0.9134, "step": 11110 }, { "epoch": 0.0, "learning_rate": 1.4532316198582265e-05, "loss": 0.9225, "step": 11111 }, { "epoch": 0.0, "learning_rate": 1.4531355356696475e-05, "loss": 0.9753, "step": 11112 }, { "epoch": 0.0, "learning_rate": 1.453039446216413e-05, "loss": 0.9342, "step": 11113 }, { "epoch": 0.0, "learning_rate": 1.4529433514996397e-05, "loss": 0.965, "step": 11114 }, { "epoch": 0.0, "learning_rate": 1.4528472515204438e-05, "loss": 0.9584, "step": 11115 }, { "epoch": 0.0, "learning_rate": 1.4527511462799417e-05, "loss": 0.93, "step": 11116 }, { "epoch": 0.0, "learning_rate": 1.4526550357792506e-05, "loss": 0.9535, "step": 11117 }, { "epoch": 0.0, "learning_rate": 1.4525589200194865e-05, "loss": 0.9119, "step": 11118 }, { "epoch": 0.0, "learning_rate": 1.4524627990017664e-05, "loss": 0.9796, "step": 11119 }, { "epoch": 0.0, "learning_rate": 1.4523666727272072e-05, "loss": 0.9019, "step": 11120 }, { "epoch": 0.0, "learning_rate": 1.452270541196925e-05, "loss": 0.9513, "step": 11121 }, { "epoch": 0.0, "learning_rate": 1.4521744044120376e-05, "loss": 0.9406, "step": 11122 }, { "epoch": 0.0, "learning_rate": 1.4520782623736613e-05, "loss": 0.9553, "step": 11123 }, { "epoch": 0.0, "learning_rate": 1.4519821150829138e-05, "loss": 0.9453, "step": 11124 }, { "epoch": 0.0, "learning_rate": 1.4518859625409115e-05, "loss": 0.935, "step": 11125 }, { "epoch": 0.0, "learning_rate": 1.4517898047487715e-05, "loss": 0.9416, "step": 11126 }, { "epoch": 0.0, "learning_rate": 1.4516936417076115e-05, "loss": 0.9396, "step": 11127 }, { "epoch": 0.0, "learning_rate": 1.451597473418548e-05, "loss": 0.9239, "step": 11128 }, { "epoch": 0.0, "learning_rate": 1.4515012998826995e-05, "loss": 0.9304, "step": 11129 }, { "epoch": 0.0, "learning_rate": 1.4514051211011821e-05, "loss": 0.9399, "step": 11130 }, { "epoch": 0.0, "learning_rate": 1.4513089370751139e-05, "loss": 0.9676, "step": 11131 }, { "epoch": 0.0, "learning_rate": 1.4512127478056126e-05, "loss": 0.911, "step": 11132 }, { "epoch": 0.0, "learning_rate": 1.4511165532937952e-05, "loss": 0.9897, "step": 11133 }, { "epoch": 0.0, "learning_rate": 1.4510203535407795e-05, "loss": 0.9529, "step": 11134 }, { "epoch": 0.0, "learning_rate": 1.4509241485476833e-05, "loss": 0.9627, "step": 11135 }, { "epoch": 0.0, "learning_rate": 1.4508279383156244e-05, "loss": 0.8998, "step": 11136 }, { "epoch": 0.0, "learning_rate": 1.4507317228457199e-05, "loss": 0.958, "step": 11137 }, { "epoch": 0.0, "learning_rate": 1.4506355021390888e-05, "loss": 0.9426, "step": 11138 }, { "epoch": 0.0, "learning_rate": 1.4505392761968483e-05, "loss": 0.9254, "step": 11139 }, { "epoch": 0.0, "learning_rate": 1.4504430450201164e-05, "loss": 0.956, "step": 11140 }, { "epoch": 0.0, "learning_rate": 1.4503468086100115e-05, "loss": 0.9438, "step": 11141 }, { "epoch": 0.0, "learning_rate": 1.4502505669676512e-05, "loss": 0.917, "step": 11142 }, { "epoch": 0.0, "learning_rate": 1.4501543200941539e-05, "loss": 0.9421, "step": 11143 }, { "epoch": 0.0, "learning_rate": 1.4500580679906379e-05, "loss": 0.9183, "step": 11144 }, { "epoch": 0.0, "learning_rate": 1.4499618106582217e-05, "loss": 0.9391, "step": 11145 }, { "epoch": 0.0, "learning_rate": 1.4498655480980228e-05, "loss": 0.9469, "step": 11146 }, { "epoch": 0.0, "learning_rate": 1.4497692803111609e-05, "loss": 0.9821, "step": 11147 }, { "epoch": 0.0, "learning_rate": 1.4496730072987535e-05, "loss": 0.9331, "step": 11148 }, { "epoch": 0.0, "learning_rate": 1.4495767290619189e-05, "loss": 0.9625, "step": 11149 }, { "epoch": 0.0, "learning_rate": 1.4494804456017768e-05, "loss": 0.9063, "step": 11150 }, { "epoch": 0.01, "learning_rate": 1.4493841569194449e-05, "loss": 0.9461, "step": 11151 }, { "epoch": 0.01, "learning_rate": 1.449287863016042e-05, "loss": 0.9459, "step": 11152 }, { "epoch": 0.01, "learning_rate": 1.4491915638926875e-05, "loss": 0.9383, "step": 11153 }, { "epoch": 0.01, "learning_rate": 1.4490952595504997e-05, "loss": 0.9022, "step": 11154 }, { "epoch": 0.01, "learning_rate": 1.4489989499905974e-05, "loss": 0.901, "step": 11155 }, { "epoch": 0.01, "learning_rate": 1.4489026352141001e-05, "loss": 0.9191, "step": 11156 }, { "epoch": 0.01, "learning_rate": 1.4488063152221263e-05, "loss": 0.9144, "step": 11157 }, { "epoch": 0.01, "learning_rate": 1.4487099900157952e-05, "loss": 0.9457, "step": 11158 }, { "epoch": 0.01, "learning_rate": 1.448613659596226e-05, "loss": 0.9489, "step": 11159 }, { "epoch": 0.01, "learning_rate": 1.4485173239645378e-05, "loss": 0.9396, "step": 11160 }, { "epoch": 0.01, "learning_rate": 1.4484209831218499e-05, "loss": 0.9397, "step": 11161 }, { "epoch": 0.01, "learning_rate": 1.4483246370692819e-05, "loss": 0.9719, "step": 11162 }, { "epoch": 0.01, "learning_rate": 1.4482282858079528e-05, "loss": 0.9155, "step": 11163 }, { "epoch": 0.01, "learning_rate": 1.4481319293389819e-05, "loss": 0.9345, "step": 11164 }, { "epoch": 0.01, "learning_rate": 1.4480355676634892e-05, "loss": 0.9267, "step": 11165 }, { "epoch": 0.01, "learning_rate": 1.4479392007825941e-05, "loss": 0.9501, "step": 11166 }, { "epoch": 0.01, "learning_rate": 1.4478428286974158e-05, "loss": 0.9539, "step": 11167 }, { "epoch": 0.01, "learning_rate": 1.4477464514090745e-05, "loss": 0.9485, "step": 11168 }, { "epoch": 0.01, "learning_rate": 1.4476500689186897e-05, "loss": 0.9645, "step": 11169 }, { "epoch": 0.01, "learning_rate": 1.4475536812273814e-05, "loss": 0.9602, "step": 11170 }, { "epoch": 0.01, "learning_rate": 1.4474572883362692e-05, "loss": 0.972, "step": 11171 }, { "epoch": 0.01, "learning_rate": 1.4473608902464728e-05, "loss": 0.9021, "step": 11172 }, { "epoch": 0.01, "learning_rate": 1.447264486959113e-05, "loss": 0.9412, "step": 11173 }, { "epoch": 0.01, "learning_rate": 1.4471680784753089e-05, "loss": 0.9611, "step": 11174 }, { "epoch": 0.01, "learning_rate": 1.4470716647961813e-05, "loss": 0.9493, "step": 11175 }, { "epoch": 0.01, "learning_rate": 1.44697524592285e-05, "loss": 0.9905, "step": 11176 }, { "epoch": 0.01, "learning_rate": 1.4468788218564353e-05, "loss": 0.9518, "step": 11177 }, { "epoch": 0.01, "learning_rate": 1.4467823925980575e-05, "loss": 0.8996, "step": 11178 }, { "epoch": 0.01, "learning_rate": 1.446685958148837e-05, "loss": 0.9918, "step": 11179 }, { "epoch": 0.01, "learning_rate": 1.4465895185098942e-05, "loss": 0.9116, "step": 11180 }, { "epoch": 0.01, "learning_rate": 1.4464930736823493e-05, "loss": 0.9604, "step": 11181 }, { "epoch": 0.01, "learning_rate": 1.4463966236673233e-05, "loss": 0.925, "step": 11182 }, { "epoch": 0.01, "learning_rate": 1.4463001684659362e-05, "loss": 0.94, "step": 11183 }, { "epoch": 0.01, "learning_rate": 1.4462037080793092e-05, "loss": 0.9653, "step": 11184 }, { "epoch": 0.01, "learning_rate": 1.4461072425085627e-05, "loss": 0.9338, "step": 11185 }, { "epoch": 0.01, "learning_rate": 1.4460107717548172e-05, "loss": 0.9541, "step": 11186 }, { "epoch": 0.01, "learning_rate": 1.4459142958191946e-05, "loss": 0.8944, "step": 11187 }, { "epoch": 0.01, "learning_rate": 1.4458178147028146e-05, "loss": 0.9243, "step": 11188 }, { "epoch": 0.01, "learning_rate": 1.4457213284067984e-05, "loss": 0.9773, "step": 11189 }, { "epoch": 0.01, "learning_rate": 1.4456248369322672e-05, "loss": 0.9008, "step": 11190 }, { "epoch": 0.01, "learning_rate": 1.4455283402803424e-05, "loss": 0.944, "step": 11191 }, { "epoch": 0.01, "learning_rate": 1.4454318384521444e-05, "loss": 0.9282, "step": 11192 }, { "epoch": 0.01, "learning_rate": 1.445335331448795e-05, "loss": 0.9561, "step": 11193 }, { "epoch": 0.01, "learning_rate": 1.4452388192714152e-05, "loss": 0.9219, "step": 11194 }, { "epoch": 0.01, "learning_rate": 1.4451423019211261e-05, "loss": 0.9915, "step": 11195 }, { "epoch": 0.01, "learning_rate": 1.4450457793990496e-05, "loss": 0.9107, "step": 11196 }, { "epoch": 0.01, "learning_rate": 1.4449492517063066e-05, "loss": 0.9503, "step": 11197 }, { "epoch": 0.01, "learning_rate": 1.4448527188440187e-05, "loss": 0.9274, "step": 11198 }, { "epoch": 0.01, "learning_rate": 1.4447561808133075e-05, "loss": 0.9225, "step": 11199 }, { "epoch": 0.01, "learning_rate": 1.444659637615295e-05, "loss": 0.919, "step": 11200 }, { "epoch": 0.01, "learning_rate": 1.4445630892511022e-05, "loss": 0.9599, "step": 11201 }, { "epoch": 0.01, "learning_rate": 1.4444665357218512e-05, "loss": 0.9224, "step": 11202 }, { "epoch": 0.01, "learning_rate": 1.4443699770286639e-05, "loss": 0.9383, "step": 11203 }, { "epoch": 0.01, "learning_rate": 1.4442734131726616e-05, "loss": 0.9253, "step": 11204 }, { "epoch": 0.01, "learning_rate": 1.4441768441549668e-05, "loss": 0.9338, "step": 11205 }, { "epoch": 0.01, "learning_rate": 1.444080269976701e-05, "loss": 0.877, "step": 11206 }, { "epoch": 0.01, "learning_rate": 1.4439836906389867e-05, "loss": 0.9778, "step": 11207 }, { "epoch": 0.01, "learning_rate": 1.4438871061429456e-05, "loss": 0.9351, "step": 11208 }, { "epoch": 0.01, "learning_rate": 1.4437905164897004e-05, "loss": 0.933, "step": 11209 }, { "epoch": 0.01, "learning_rate": 1.4436939216803723e-05, "loss": 0.9181, "step": 11210 }, { "epoch": 0.01, "learning_rate": 1.4435973217160843e-05, "loss": 0.9657, "step": 11211 }, { "epoch": 0.01, "learning_rate": 1.4435007165979591e-05, "loss": 0.959, "step": 11212 }, { "epoch": 0.01, "learning_rate": 1.443404106327118e-05, "loss": 0.9221, "step": 11213 }, { "epoch": 0.01, "learning_rate": 1.443307490904684e-05, "loss": 0.9126, "step": 11214 }, { "epoch": 0.01, "learning_rate": 1.44321087033178e-05, "loss": 0.9288, "step": 11215 }, { "epoch": 0.01, "learning_rate": 1.4431142446095282e-05, "loss": 0.9502, "step": 11216 }, { "epoch": 0.01, "learning_rate": 1.4430176137390508e-05, "loss": 0.9396, "step": 11217 }, { "epoch": 0.01, "learning_rate": 1.442920977721471e-05, "loss": 0.9641, "step": 11218 }, { "epoch": 0.01, "learning_rate": 1.4428243365579115e-05, "loss": 0.9131, "step": 11219 }, { "epoch": 0.01, "learning_rate": 1.4427276902494951e-05, "loss": 0.937, "step": 11220 }, { "epoch": 0.01, "learning_rate": 1.4426310387973448e-05, "loss": 0.9595, "step": 11221 }, { "epoch": 0.01, "learning_rate": 1.442534382202583e-05, "loss": 0.9267, "step": 11222 }, { "epoch": 0.01, "learning_rate": 1.4424377204663332e-05, "loss": 0.9156, "step": 11223 }, { "epoch": 0.01, "learning_rate": 1.4423410535897184e-05, "loss": 0.9245, "step": 11224 }, { "epoch": 0.01, "learning_rate": 1.4422443815738612e-05, "loss": 0.9279, "step": 11225 }, { "epoch": 0.01, "learning_rate": 1.4421477044198853e-05, "loss": 0.9175, "step": 11226 }, { "epoch": 0.01, "learning_rate": 1.4420510221289137e-05, "loss": 0.9569, "step": 11227 }, { "epoch": 0.01, "learning_rate": 1.4419543347020698e-05, "loss": 0.9217, "step": 11228 }, { "epoch": 0.01, "learning_rate": 1.4418576421404767e-05, "loss": 0.8974, "step": 11229 }, { "epoch": 0.01, "learning_rate": 1.441760944445258e-05, "loss": 0.9539, "step": 11230 }, { "epoch": 0.01, "learning_rate": 1.4416642416175375e-05, "loss": 0.9463, "step": 11231 }, { "epoch": 0.01, "learning_rate": 1.4415675336584378e-05, "loss": 0.9545, "step": 11232 }, { "epoch": 0.01, "learning_rate": 1.4414708205690834e-05, "loss": 0.948, "step": 11233 }, { "epoch": 0.01, "learning_rate": 1.4413741023505977e-05, "loss": 0.9585, "step": 11234 }, { "epoch": 0.01, "learning_rate": 1.441277379004104e-05, "loss": 0.959, "step": 11235 }, { "epoch": 0.01, "learning_rate": 1.4411806505307262e-05, "loss": 0.9288, "step": 11236 }, { "epoch": 0.01, "learning_rate": 1.4410839169315885e-05, "loss": 0.9423, "step": 11237 }, { "epoch": 0.01, "learning_rate": 1.4409871782078142e-05, "loss": 0.8921, "step": 11238 }, { "epoch": 0.01, "learning_rate": 1.440890434360528e-05, "loss": 0.9835, "step": 11239 }, { "epoch": 0.01, "learning_rate": 1.4407936853908535e-05, "loss": 0.8949, "step": 11240 }, { "epoch": 0.01, "learning_rate": 1.4406969312999141e-05, "loss": 0.9326, "step": 11241 }, { "epoch": 0.01, "learning_rate": 1.4406001720888352e-05, "loss": 0.9484, "step": 11242 }, { "epoch": 0.01, "learning_rate": 1.4405034077587398e-05, "loss": 0.9385, "step": 11243 }, { "epoch": 0.01, "learning_rate": 1.440406638310753e-05, "loss": 0.9686, "step": 11244 }, { "epoch": 0.01, "learning_rate": 1.4403098637459985e-05, "loss": 0.9724, "step": 11245 }, { "epoch": 0.01, "learning_rate": 1.4402130840656012e-05, "loss": 0.9282, "step": 11246 }, { "epoch": 0.01, "learning_rate": 1.4401162992706849e-05, "loss": 0.9633, "step": 11247 }, { "epoch": 0.01, "learning_rate": 1.4400195093623745e-05, "loss": 0.9326, "step": 11248 }, { "epoch": 0.01, "learning_rate": 1.4399227143417946e-05, "loss": 0.9163, "step": 11249 }, { "epoch": 0.01, "learning_rate": 1.439825914210069e-05, "loss": 0.9437, "step": 11250 }, { "epoch": 0.01, "learning_rate": 1.4397291089683236e-05, "loss": 0.9582, "step": 11251 }, { "epoch": 0.01, "learning_rate": 1.4396322986176823e-05, "loss": 0.9415, "step": 11252 }, { "epoch": 0.01, "learning_rate": 1.4395354831592699e-05, "loss": 0.9268, "step": 11253 }, { "epoch": 0.01, "learning_rate": 1.4394386625942112e-05, "loss": 0.9511, "step": 11254 }, { "epoch": 0.01, "learning_rate": 1.4393418369236316e-05, "loss": 0.8992, "step": 11255 }, { "epoch": 0.01, "learning_rate": 1.4392450061486555e-05, "loss": 0.9376, "step": 11256 }, { "epoch": 0.01, "learning_rate": 1.4391481702704082e-05, "loss": 0.9102, "step": 11257 }, { "epoch": 0.01, "learning_rate": 1.4390513292900147e-05, "loss": 0.954, "step": 11258 }, { "epoch": 0.01, "learning_rate": 1.4389544832086e-05, "loss": 0.9476, "step": 11259 }, { "epoch": 0.01, "learning_rate": 1.4388576320272894e-05, "loss": 0.9301, "step": 11260 }, { "epoch": 0.01, "learning_rate": 1.4387607757472081e-05, "loss": 0.9653, "step": 11261 }, { "epoch": 0.01, "learning_rate": 1.4386639143694817e-05, "loss": 0.9306, "step": 11262 }, { "epoch": 0.01, "learning_rate": 1.4385670478952349e-05, "loss": 0.9656, "step": 11263 }, { "epoch": 0.01, "learning_rate": 1.4384701763255936e-05, "loss": 0.9143, "step": 11264 }, { "epoch": 0.01, "learning_rate": 1.4383732996616833e-05, "loss": 0.9301, "step": 11265 }, { "epoch": 0.01, "learning_rate": 1.4382764179046294e-05, "loss": 0.9013, "step": 11266 }, { "epoch": 0.01, "learning_rate": 1.4381795310555577e-05, "loss": 0.9477, "step": 11267 }, { "epoch": 0.01, "learning_rate": 1.4380826391155935e-05, "loss": 0.9696, "step": 11268 }, { "epoch": 0.01, "learning_rate": 1.4379857420858625e-05, "loss": 0.9492, "step": 11269 }, { "epoch": 0.01, "learning_rate": 1.4378888399674913e-05, "loss": 0.9182, "step": 11270 }, { "epoch": 0.01, "learning_rate": 1.4377919327616046e-05, "loss": 0.9459, "step": 11271 }, { "epoch": 0.01, "learning_rate": 1.4376950204693289e-05, "loss": 0.9278, "step": 11272 }, { "epoch": 0.01, "learning_rate": 1.4375981030917903e-05, "loss": 0.9357, "step": 11273 }, { "epoch": 0.01, "learning_rate": 1.4375011806301144e-05, "loss": 0.9209, "step": 11274 }, { "epoch": 0.01, "learning_rate": 1.4374042530854272e-05, "loss": 0.9865, "step": 11275 }, { "epoch": 0.01, "learning_rate": 1.4373073204588556e-05, "loss": 0.9884, "step": 11276 }, { "epoch": 0.01, "learning_rate": 1.4372103827515249e-05, "loss": 0.9707, "step": 11277 }, { "epoch": 0.01, "learning_rate": 1.4371134399645618e-05, "loss": 0.904, "step": 11278 }, { "epoch": 0.01, "learning_rate": 1.437016492099093e-05, "loss": 0.9211, "step": 11279 }, { "epoch": 0.01, "learning_rate": 1.4369195391562438e-05, "loss": 0.939, "step": 11280 }, { "epoch": 0.01, "learning_rate": 1.4368225811371414e-05, "loss": 0.9333, "step": 11281 }, { "epoch": 0.01, "learning_rate": 1.436725618042912e-05, "loss": 0.9171, "step": 11282 }, { "epoch": 0.01, "learning_rate": 1.4366286498746828e-05, "loss": 0.9564, "step": 11283 }, { "epoch": 0.01, "learning_rate": 1.4365316766335793e-05, "loss": 0.9474, "step": 11284 }, { "epoch": 0.01, "learning_rate": 1.4364346983207292e-05, "loss": 0.9595, "step": 11285 }, { "epoch": 0.01, "learning_rate": 1.4363377149372584e-05, "loss": 0.8801, "step": 11286 }, { "epoch": 0.01, "learning_rate": 1.436240726484294e-05, "loss": 0.9619, "step": 11287 }, { "epoch": 0.01, "learning_rate": 1.4361437329629628e-05, "loss": 0.9491, "step": 11288 }, { "epoch": 0.01, "learning_rate": 1.4360467343743921e-05, "loss": 0.9446, "step": 11289 }, { "epoch": 0.01, "learning_rate": 1.4359497307197083e-05, "loss": 0.8968, "step": 11290 }, { "epoch": 0.01, "learning_rate": 1.4358527220000386e-05, "loss": 0.9207, "step": 11291 }, { "epoch": 0.01, "learning_rate": 1.4357557082165103e-05, "loss": 0.9205, "step": 11292 }, { "epoch": 0.01, "learning_rate": 1.4356586893702502e-05, "loss": 0.9924, "step": 11293 }, { "epoch": 0.01, "learning_rate": 1.4355616654623856e-05, "loss": 0.9145, "step": 11294 }, { "epoch": 0.01, "learning_rate": 1.4354646364940441e-05, "loss": 0.9371, "step": 11295 }, { "epoch": 0.01, "learning_rate": 1.4353676024663521e-05, "loss": 0.9542, "step": 11296 }, { "epoch": 0.01, "learning_rate": 1.435270563380438e-05, "loss": 0.9306, "step": 11297 }, { "epoch": 0.01, "learning_rate": 1.4351735192374287e-05, "loss": 0.951, "step": 11298 }, { "epoch": 0.01, "learning_rate": 1.4350764700384518e-05, "loss": 0.9589, "step": 11299 }, { "epoch": 0.01, "learning_rate": 1.4349794157846348e-05, "loss": 0.9465, "step": 11300 }, { "epoch": 0.01, "learning_rate": 1.4348823564771054e-05, "loss": 0.9378, "step": 11301 }, { "epoch": 0.01, "learning_rate": 1.4347852921169912e-05, "loss": 0.9475, "step": 11302 }, { "epoch": 0.01, "learning_rate": 1.4346882227054197e-05, "loss": 0.9194, "step": 11303 }, { "epoch": 0.01, "learning_rate": 1.4345911482435194e-05, "loss": 0.9255, "step": 11304 }, { "epoch": 0.01, "learning_rate": 1.4344940687324172e-05, "loss": 0.9676, "step": 11305 }, { "epoch": 0.01, "learning_rate": 1.4343969841732412e-05, "loss": 0.9341, "step": 11306 }, { "epoch": 0.01, "learning_rate": 1.4342998945671199e-05, "loss": 0.9348, "step": 11307 }, { "epoch": 0.01, "learning_rate": 1.4342027999151811e-05, "loss": 0.9278, "step": 11308 }, { "epoch": 0.01, "learning_rate": 1.4341057002185526e-05, "loss": 0.9482, "step": 11309 }, { "epoch": 0.01, "learning_rate": 1.4340085954783626e-05, "loss": 0.9066, "step": 11310 }, { "epoch": 0.01, "learning_rate": 1.4339114856957395e-05, "loss": 0.959, "step": 11311 }, { "epoch": 0.01, "learning_rate": 1.4338143708718115e-05, "loss": 0.9373, "step": 11312 }, { "epoch": 0.01, "learning_rate": 1.4337172510077068e-05, "loss": 0.9466, "step": 11313 }, { "epoch": 0.01, "learning_rate": 1.4336201261045538e-05, "loss": 0.8991, "step": 11314 }, { "epoch": 0.01, "learning_rate": 1.4335229961634808e-05, "loss": 0.9596, "step": 11315 }, { "epoch": 0.01, "learning_rate": 1.4334258611856168e-05, "loss": 0.9133, "step": 11316 }, { "epoch": 0.01, "learning_rate": 1.4333287211720896e-05, "loss": 0.9187, "step": 11317 }, { "epoch": 0.01, "learning_rate": 1.4332315761240283e-05, "loss": 0.8946, "step": 11318 }, { "epoch": 0.01, "learning_rate": 1.4331344260425612e-05, "loss": 0.918, "step": 11319 }, { "epoch": 0.01, "learning_rate": 1.4330372709288178e-05, "loss": 0.9083, "step": 11320 }, { "epoch": 0.01, "learning_rate": 1.4329401107839256e-05, "loss": 0.952, "step": 11321 }, { "epoch": 0.01, "learning_rate": 1.4328429456090146e-05, "loss": 0.9423, "step": 11322 }, { "epoch": 0.01, "learning_rate": 1.4327457754052132e-05, "loss": 0.9191, "step": 11323 }, { "epoch": 0.01, "learning_rate": 1.4326486001736501e-05, "loss": 0.9209, "step": 11324 }, { "epoch": 0.01, "learning_rate": 1.4325514199154551e-05, "loss": 0.9761, "step": 11325 }, { "epoch": 0.01, "learning_rate": 1.4324542346317562e-05, "loss": 0.9433, "step": 11326 }, { "epoch": 0.01, "learning_rate": 1.4323570443236835e-05, "loss": 0.9828, "step": 11327 }, { "epoch": 0.01, "learning_rate": 1.4322598489923655e-05, "loss": 0.9653, "step": 11328 }, { "epoch": 0.01, "learning_rate": 1.432162648638932e-05, "loss": 0.9019, "step": 11329 }, { "epoch": 0.01, "learning_rate": 1.4320654432645116e-05, "loss": 0.9612, "step": 11330 }, { "epoch": 0.01, "learning_rate": 1.4319682328702342e-05, "loss": 0.9451, "step": 11331 }, { "epoch": 0.01, "learning_rate": 1.4318710174572294e-05, "loss": 0.92, "step": 11332 }, { "epoch": 0.01, "learning_rate": 1.4317737970266261e-05, "loss": 0.9471, "step": 11333 }, { "epoch": 0.01, "learning_rate": 1.4316765715795542e-05, "loss": 0.9223, "step": 11334 }, { "epoch": 0.01, "learning_rate": 1.4315793411171433e-05, "loss": 0.8919, "step": 11335 }, { "epoch": 0.01, "learning_rate": 1.4314821056405226e-05, "loss": 0.9372, "step": 11336 }, { "epoch": 0.01, "learning_rate": 1.4313848651508224e-05, "loss": 0.9234, "step": 11337 }, { "epoch": 0.01, "learning_rate": 1.4312876196491722e-05, "loss": 0.9038, "step": 11338 }, { "epoch": 0.01, "learning_rate": 1.4311903691367017e-05, "loss": 0.8983, "step": 11339 }, { "epoch": 0.01, "learning_rate": 1.4310931136145412e-05, "loss": 0.8779, "step": 11340 }, { "epoch": 0.01, "learning_rate": 1.4309958530838205e-05, "loss": 0.9162, "step": 11341 }, { "epoch": 0.01, "learning_rate": 1.4308985875456687e-05, "loss": 0.9228, "step": 11342 }, { "epoch": 0.01, "learning_rate": 1.4308013170012174e-05, "loss": 0.9341, "step": 11343 }, { "epoch": 0.01, "learning_rate": 1.4307040414515956e-05, "loss": 0.9237, "step": 11344 }, { "epoch": 0.01, "learning_rate": 1.430606760897934e-05, "loss": 0.975, "step": 11345 }, { "epoch": 0.01, "learning_rate": 1.4305094753413624e-05, "loss": 0.9607, "step": 11346 }, { "epoch": 0.01, "learning_rate": 1.4304121847830117e-05, "loss": 0.9507, "step": 11347 }, { "epoch": 0.01, "learning_rate": 1.4303148892240115e-05, "loss": 0.9445, "step": 11348 }, { "epoch": 0.01, "learning_rate": 1.4302175886654928e-05, "loss": 0.9074, "step": 11349 }, { "epoch": 0.01, "learning_rate": 1.4301202831085861e-05, "loss": 0.9647, "step": 11350 }, { "epoch": 0.01, "learning_rate": 1.4300229725544213e-05, "loss": 0.9193, "step": 11351 }, { "epoch": 0.01, "learning_rate": 1.4299256570041295e-05, "loss": 0.9635, "step": 11352 }, { "epoch": 0.01, "learning_rate": 1.4298283364588412e-05, "loss": 0.9902, "step": 11353 }, { "epoch": 0.01, "learning_rate": 1.4297310109196872e-05, "loss": 0.933, "step": 11354 }, { "epoch": 0.01, "learning_rate": 1.4296336803877981e-05, "loss": 0.9107, "step": 11355 }, { "epoch": 0.01, "learning_rate": 1.4295363448643048e-05, "loss": 0.9455, "step": 11356 }, { "epoch": 0.01, "learning_rate": 1.429439004350338e-05, "loss": 0.933, "step": 11357 }, { "epoch": 0.01, "learning_rate": 1.4293416588470289e-05, "loss": 0.9474, "step": 11358 }, { "epoch": 0.01, "learning_rate": 1.4292443083555086e-05, "loss": 0.9343, "step": 11359 }, { "epoch": 0.01, "learning_rate": 1.4291469528769074e-05, "loss": 0.9151, "step": 11360 }, { "epoch": 0.01, "learning_rate": 1.429049592412357e-05, "loss": 0.9199, "step": 11361 }, { "epoch": 0.01, "learning_rate": 1.428952226962989e-05, "loss": 0.9354, "step": 11362 }, { "epoch": 0.01, "learning_rate": 1.4288548565299337e-05, "loss": 0.933, "step": 11363 }, { "epoch": 0.01, "learning_rate": 1.4287574811143227e-05, "loss": 0.9365, "step": 11364 }, { "epoch": 0.01, "learning_rate": 1.4286601007172876e-05, "loss": 0.9286, "step": 11365 }, { "epoch": 0.01, "learning_rate": 1.4285627153399594e-05, "loss": 0.9266, "step": 11366 }, { "epoch": 0.01, "learning_rate": 1.4284653249834698e-05, "loss": 0.9447, "step": 11367 }, { "epoch": 0.01, "learning_rate": 1.4283679296489505e-05, "loss": 0.9233, "step": 11368 }, { "epoch": 0.01, "learning_rate": 1.4282705293375325e-05, "loss": 0.9423, "step": 11369 }, { "epoch": 0.01, "learning_rate": 1.4281731240503478e-05, "loss": 0.9376, "step": 11370 }, { "epoch": 0.01, "learning_rate": 1.4280757137885281e-05, "loss": 0.9669, "step": 11371 }, { "epoch": 0.01, "learning_rate": 1.427978298553205e-05, "loss": 0.9144, "step": 11372 }, { "epoch": 0.01, "learning_rate": 1.4278808783455106e-05, "loss": 0.9482, "step": 11373 }, { "epoch": 0.01, "learning_rate": 1.4277834531665761e-05, "loss": 0.9099, "step": 11374 }, { "epoch": 0.01, "learning_rate": 1.4276860230175345e-05, "loss": 0.9685, "step": 11375 }, { "epoch": 0.01, "learning_rate": 1.4275885878995164e-05, "loss": 0.8894, "step": 11376 }, { "epoch": 0.01, "learning_rate": 1.4274911478136551e-05, "loss": 0.909, "step": 11377 }, { "epoch": 0.01, "learning_rate": 1.4273937027610818e-05, "loss": 0.9228, "step": 11378 }, { "epoch": 0.01, "learning_rate": 1.4272962527429288e-05, "loss": 0.9348, "step": 11379 }, { "epoch": 0.01, "learning_rate": 1.4271987977603287e-05, "loss": 0.9402, "step": 11380 }, { "epoch": 0.01, "learning_rate": 1.4271013378144134e-05, "loss": 0.9016, "step": 11381 }, { "epoch": 0.01, "learning_rate": 1.4270038729063153e-05, "loss": 0.9461, "step": 11382 }, { "epoch": 0.01, "learning_rate": 1.4269064030371671e-05, "loss": 0.9575, "step": 11383 }, { "epoch": 0.01, "learning_rate": 1.426808928208101e-05, "loss": 0.9748, "step": 11384 }, { "epoch": 0.01, "learning_rate": 1.426711448420249e-05, "loss": 0.9245, "step": 11385 }, { "epoch": 0.01, "learning_rate": 1.4266139636747442e-05, "loss": 0.9287, "step": 11386 }, { "epoch": 0.01, "learning_rate": 1.4265164739727195e-05, "loss": 0.9286, "step": 11387 }, { "epoch": 0.01, "learning_rate": 1.4264189793153068e-05, "loss": 0.9339, "step": 11388 }, { "epoch": 0.01, "learning_rate": 1.4263214797036391e-05, "loss": 0.9465, "step": 11389 }, { "epoch": 0.01, "learning_rate": 1.4262239751388493e-05, "loss": 0.9633, "step": 11390 }, { "epoch": 0.01, "learning_rate": 1.4261264656220704e-05, "loss": 0.9211, "step": 11391 }, { "epoch": 0.01, "learning_rate": 1.4260289511544347e-05, "loss": 0.9618, "step": 11392 }, { "epoch": 0.01, "learning_rate": 1.4259314317370757e-05, "loss": 0.9457, "step": 11393 }, { "epoch": 0.01, "learning_rate": 1.4258339073711266e-05, "loss": 0.912, "step": 11394 }, { "epoch": 0.01, "learning_rate": 1.4257363780577196e-05, "loss": 0.9518, "step": 11395 }, { "epoch": 0.01, "learning_rate": 1.4256388437979887e-05, "loss": 0.881, "step": 11396 }, { "epoch": 0.01, "learning_rate": 1.4255413045930664e-05, "loss": 0.897, "step": 11397 }, { "epoch": 0.01, "learning_rate": 1.4254437604440863e-05, "loss": 0.9103, "step": 11398 }, { "epoch": 0.01, "learning_rate": 1.4253462113521816e-05, "loss": 0.938, "step": 11399 }, { "epoch": 0.01, "learning_rate": 1.4252486573184859e-05, "loss": 0.9147, "step": 11400 }, { "epoch": 0.01, "learning_rate": 1.4251510983441323e-05, "loss": 0.9701, "step": 11401 }, { "epoch": 0.01, "learning_rate": 1.4250535344302543e-05, "loss": 0.8995, "step": 11402 }, { "epoch": 0.01, "learning_rate": 1.4249559655779855e-05, "loss": 0.9311, "step": 11403 }, { "epoch": 0.01, "learning_rate": 1.4248583917884595e-05, "loss": 0.9135, "step": 11404 }, { "epoch": 0.01, "learning_rate": 1.42476081306281e-05, "loss": 0.95, "step": 11405 }, { "epoch": 0.01, "learning_rate": 1.4246632294021705e-05, "loss": 0.9261, "step": 11406 }, { "epoch": 0.01, "learning_rate": 1.424565640807675e-05, "loss": 0.9453, "step": 11407 }, { "epoch": 0.01, "learning_rate": 1.424468047280457e-05, "loss": 0.9496, "step": 11408 }, { "epoch": 0.01, "learning_rate": 1.4243704488216507e-05, "loss": 0.9237, "step": 11409 }, { "epoch": 0.01, "learning_rate": 1.4242728454323897e-05, "loss": 0.8895, "step": 11410 }, { "epoch": 0.01, "learning_rate": 1.4241752371138082e-05, "loss": 0.9432, "step": 11411 }, { "epoch": 0.01, "learning_rate": 1.4240776238670402e-05, "loss": 0.9265, "step": 11412 }, { "epoch": 0.01, "learning_rate": 1.4239800056932199e-05, "loss": 0.9511, "step": 11413 }, { "epoch": 0.01, "learning_rate": 1.4238823825934813e-05, "loss": 0.9019, "step": 11414 }, { "epoch": 0.01, "learning_rate": 1.4237847545689585e-05, "loss": 0.9233, "step": 11415 }, { "epoch": 0.01, "learning_rate": 1.4236871216207862e-05, "loss": 0.9302, "step": 11416 }, { "epoch": 0.01, "learning_rate": 1.4235894837500985e-05, "loss": 0.8988, "step": 11417 }, { "epoch": 0.01, "learning_rate": 1.4234918409580294e-05, "loss": 0.9155, "step": 11418 }, { "epoch": 0.01, "learning_rate": 1.4233941932457138e-05, "loss": 0.9517, "step": 11419 }, { "epoch": 0.01, "learning_rate": 1.423296540614286e-05, "loss": 0.9284, "step": 11420 }, { "epoch": 0.01, "learning_rate": 1.423198883064881e-05, "loss": 0.9234, "step": 11421 }, { "epoch": 0.01, "learning_rate": 1.4231012205986328e-05, "loss": 0.9412, "step": 11422 }, { "epoch": 0.01, "learning_rate": 1.4230035532166764e-05, "loss": 0.9388, "step": 11423 }, { "epoch": 0.01, "learning_rate": 1.4229058809201463e-05, "loss": 0.9423, "step": 11424 }, { "epoch": 0.01, "learning_rate": 1.4228082037101774e-05, "loss": 0.9353, "step": 11425 }, { "epoch": 0.01, "learning_rate": 1.4227105215879049e-05, "loss": 0.9279, "step": 11426 }, { "epoch": 0.01, "learning_rate": 1.422612834554463e-05, "loss": 0.9021, "step": 11427 }, { "epoch": 0.01, "learning_rate": 1.4225151426109873e-05, "loss": 0.9169, "step": 11428 }, { "epoch": 0.01, "learning_rate": 1.4224174457586121e-05, "loss": 0.9817, "step": 11429 }, { "epoch": 0.01, "learning_rate": 1.4223197439984735e-05, "loss": 0.9429, "step": 11430 }, { "epoch": 0.01, "learning_rate": 1.4222220373317056e-05, "loss": 0.9681, "step": 11431 }, { "epoch": 0.01, "learning_rate": 1.4221243257594443e-05, "loss": 0.9015, "step": 11432 }, { "epoch": 0.01, "learning_rate": 1.4220266092828245e-05, "loss": 0.9079, "step": 11433 }, { "epoch": 0.01, "learning_rate": 1.4219288879029813e-05, "loss": 0.9256, "step": 11434 }, { "epoch": 0.01, "learning_rate": 1.4218311616210507e-05, "loss": 0.9528, "step": 11435 }, { "epoch": 0.01, "learning_rate": 1.4217334304381676e-05, "loss": 0.9384, "step": 11436 }, { "epoch": 0.01, "learning_rate": 1.4216356943554672e-05, "loss": 0.9304, "step": 11437 }, { "epoch": 0.01, "learning_rate": 1.4215379533740858e-05, "loss": 0.9112, "step": 11438 }, { "epoch": 0.01, "learning_rate": 1.4214402074951585e-05, "loss": 0.9223, "step": 11439 }, { "epoch": 0.01, "learning_rate": 1.4213424567198212e-05, "loss": 0.9024, "step": 11440 }, { "epoch": 0.01, "learning_rate": 1.4212447010492092e-05, "loss": 0.944, "step": 11441 }, { "epoch": 0.01, "learning_rate": 1.4211469404844587e-05, "loss": 0.9337, "step": 11442 }, { "epoch": 0.01, "learning_rate": 1.4210491750267049e-05, "loss": 0.9177, "step": 11443 }, { "epoch": 0.01, "learning_rate": 1.4209514046770845e-05, "loss": 0.926, "step": 11444 }, { "epoch": 0.01, "learning_rate": 1.4208536294367327e-05, "loss": 0.9225, "step": 11445 }, { "epoch": 0.01, "learning_rate": 1.420755849306786e-05, "loss": 0.9264, "step": 11446 }, { "epoch": 0.01, "learning_rate": 1.42065806428838e-05, "loss": 0.9237, "step": 11447 }, { "epoch": 0.01, "learning_rate": 1.420560274382651e-05, "loss": 0.9265, "step": 11448 }, { "epoch": 0.01, "learning_rate": 1.4204624795907352e-05, "loss": 0.9234, "step": 11449 }, { "epoch": 0.01, "learning_rate": 1.4203646799137688e-05, "loss": 0.9407, "step": 11450 }, { "epoch": 0.02, "learning_rate": 1.4202668753528878e-05, "loss": 0.9361, "step": 11451 }, { "epoch": 0.02, "learning_rate": 1.420169065909229e-05, "loss": 0.9639, "step": 11452 }, { "epoch": 0.02, "learning_rate": 1.4200712515839283e-05, "loss": 0.9235, "step": 11453 }, { "epoch": 0.02, "learning_rate": 1.4199734323781228e-05, "loss": 0.9438, "step": 11454 }, { "epoch": 0.02, "learning_rate": 1.4198756082929479e-05, "loss": 0.9372, "step": 11455 }, { "epoch": 0.02, "learning_rate": 1.419777779329541e-05, "loss": 0.92, "step": 11456 }, { "epoch": 0.02, "learning_rate": 1.4196799454890386e-05, "loss": 0.9058, "step": 11457 }, { "epoch": 0.02, "learning_rate": 1.4195821067725773e-05, "loss": 0.9099, "step": 11458 }, { "epoch": 0.02, "learning_rate": 1.4194842631812936e-05, "loss": 0.9207, "step": 11459 }, { "epoch": 0.02, "learning_rate": 1.4193864147163248e-05, "loss": 0.9521, "step": 11460 }, { "epoch": 0.02, "learning_rate": 1.419288561378807e-05, "loss": 0.9453, "step": 11461 }, { "epoch": 0.02, "learning_rate": 1.4191907031698774e-05, "loss": 0.9036, "step": 11462 }, { "epoch": 0.02, "learning_rate": 1.4190928400906731e-05, "loss": 0.9679, "step": 11463 }, { "epoch": 0.02, "learning_rate": 1.4189949721423313e-05, "loss": 0.9492, "step": 11464 }, { "epoch": 0.02, "learning_rate": 1.4188970993259883e-05, "loss": 0.9358, "step": 11465 }, { "epoch": 0.02, "learning_rate": 1.418799221642782e-05, "loss": 0.9126, "step": 11466 }, { "epoch": 0.02, "learning_rate": 1.4187013390938493e-05, "loss": 0.9248, "step": 11467 }, { "epoch": 0.02, "learning_rate": 1.4186034516803269e-05, "loss": 0.8972, "step": 11468 }, { "epoch": 0.02, "learning_rate": 1.418505559403353e-05, "loss": 0.9368, "step": 11469 }, { "epoch": 0.02, "learning_rate": 1.4184076622640646e-05, "loss": 0.9289, "step": 11470 }, { "epoch": 0.02, "learning_rate": 1.4183097602635985e-05, "loss": 0.9186, "step": 11471 }, { "epoch": 0.02, "learning_rate": 1.4182118534030931e-05, "loss": 0.8917, "step": 11472 }, { "epoch": 0.02, "learning_rate": 1.418113941683685e-05, "loss": 0.9199, "step": 11473 }, { "epoch": 0.02, "learning_rate": 1.4180160251065127e-05, "loss": 0.9276, "step": 11474 }, { "epoch": 0.02, "learning_rate": 1.417918103672713e-05, "loss": 0.9108, "step": 11475 }, { "epoch": 0.02, "learning_rate": 1.417820177383424e-05, "loss": 0.9399, "step": 11476 }, { "epoch": 0.02, "learning_rate": 1.4177222462397834e-05, "loss": 0.9404, "step": 11477 }, { "epoch": 0.02, "learning_rate": 1.417624310242929e-05, "loss": 0.9385, "step": 11478 }, { "epoch": 0.02, "learning_rate": 1.4175263693939987e-05, "loss": 0.9599, "step": 11479 }, { "epoch": 0.02, "learning_rate": 1.4174284236941301e-05, "loss": 0.9139, "step": 11480 }, { "epoch": 0.02, "learning_rate": 1.4173304731444615e-05, "loss": 0.9308, "step": 11481 }, { "epoch": 0.02, "learning_rate": 1.4172325177461305e-05, "loss": 0.9455, "step": 11482 }, { "epoch": 0.02, "learning_rate": 1.4171345575002757e-05, "loss": 0.9318, "step": 11483 }, { "epoch": 0.02, "learning_rate": 1.4170365924080347e-05, "loss": 0.8726, "step": 11484 }, { "epoch": 0.02, "learning_rate": 1.4169386224705464e-05, "loss": 0.9531, "step": 11485 }, { "epoch": 0.02, "learning_rate": 1.4168406476889484e-05, "loss": 0.9467, "step": 11486 }, { "epoch": 0.02, "learning_rate": 1.4167426680643793e-05, "loss": 0.9454, "step": 11487 }, { "epoch": 0.02, "learning_rate": 1.4166446835979772e-05, "loss": 0.9164, "step": 11488 }, { "epoch": 0.02, "learning_rate": 1.4165466942908805e-05, "loss": 0.9375, "step": 11489 }, { "epoch": 0.02, "learning_rate": 1.4164487001442283e-05, "loss": 0.9718, "step": 11490 }, { "epoch": 0.02, "learning_rate": 1.4163507011591583e-05, "loss": 0.9262, "step": 11491 }, { "epoch": 0.02, "learning_rate": 1.4162526973368096e-05, "loss": 0.9431, "step": 11492 }, { "epoch": 0.02, "learning_rate": 1.4161546886783208e-05, "loss": 0.9729, "step": 11493 }, { "epoch": 0.02, "learning_rate": 1.41605667518483e-05, "loss": 0.9032, "step": 11494 }, { "epoch": 0.02, "learning_rate": 1.4159586568574768e-05, "loss": 0.9248, "step": 11495 }, { "epoch": 0.02, "learning_rate": 1.4158606336973996e-05, "loss": 0.9306, "step": 11496 }, { "epoch": 0.02, "learning_rate": 1.4157626057057374e-05, "loss": 0.9198, "step": 11497 }, { "epoch": 0.02, "learning_rate": 1.4156645728836287e-05, "loss": 0.9258, "step": 11498 }, { "epoch": 0.02, "learning_rate": 1.415566535232213e-05, "loss": 0.9509, "step": 11499 }, { "epoch": 0.02, "learning_rate": 1.4154684927526291e-05, "loss": 0.9426, "step": 11500 }, { "epoch": 0.02, "learning_rate": 1.415370445446016e-05, "loss": 0.9047, "step": 11501 }, { "epoch": 0.02, "learning_rate": 1.415272393313513e-05, "loss": 0.9142, "step": 11502 }, { "epoch": 0.02, "learning_rate": 1.4151743363562589e-05, "loss": 0.9212, "step": 11503 }, { "epoch": 0.02, "learning_rate": 1.4150762745753937e-05, "loss": 0.9036, "step": 11504 }, { "epoch": 0.02, "learning_rate": 1.4149782079720558e-05, "loss": 0.9533, "step": 11505 }, { "epoch": 0.02, "learning_rate": 1.4148801365473856e-05, "loss": 0.9098, "step": 11506 }, { "epoch": 0.02, "learning_rate": 1.4147820603025217e-05, "loss": 0.911, "step": 11507 }, { "epoch": 0.02, "learning_rate": 1.4146839792386037e-05, "loss": 0.889, "step": 11508 }, { "epoch": 0.02, "learning_rate": 1.4145858933567718e-05, "loss": 0.9355, "step": 11509 }, { "epoch": 0.02, "learning_rate": 1.4144878026581648e-05, "loss": 0.9062, "step": 11510 }, { "epoch": 0.02, "learning_rate": 1.4143897071439223e-05, "loss": 0.9565, "step": 11511 }, { "epoch": 0.02, "learning_rate": 1.4142916068151846e-05, "loss": 0.8894, "step": 11512 }, { "epoch": 0.02, "learning_rate": 1.4141935016730912e-05, "loss": 0.9603, "step": 11513 }, { "epoch": 0.02, "learning_rate": 1.4140953917187816e-05, "loss": 0.9243, "step": 11514 }, { "epoch": 0.02, "learning_rate": 1.4139972769533964e-05, "loss": 0.9717, "step": 11515 }, { "epoch": 0.02, "learning_rate": 1.4138991573780749e-05, "loss": 0.8876, "step": 11516 }, { "epoch": 0.02, "learning_rate": 1.4138010329939572e-05, "loss": 0.9567, "step": 11517 }, { "epoch": 0.02, "learning_rate": 1.4137029038021835e-05, "loss": 0.9653, "step": 11518 }, { "epoch": 0.02, "learning_rate": 1.4136047698038937e-05, "loss": 0.9758, "step": 11519 }, { "epoch": 0.02, "learning_rate": 1.4135066310002281e-05, "loss": 0.9217, "step": 11520 }, { "epoch": 0.02, "learning_rate": 1.4134084873923265e-05, "loss": 0.92, "step": 11521 }, { "epoch": 0.02, "learning_rate": 1.4133103389813302e-05, "loss": 0.9156, "step": 11522 }, { "epoch": 0.02, "learning_rate": 1.4132121857683782e-05, "loss": 0.9505, "step": 11523 }, { "epoch": 0.02, "learning_rate": 1.4131140277546119e-05, "loss": 0.8991, "step": 11524 }, { "epoch": 0.02, "learning_rate": 1.4130158649411713e-05, "loss": 0.9189, "step": 11525 }, { "epoch": 0.02, "learning_rate": 1.4129176973291964e-05, "loss": 0.9051, "step": 11526 }, { "epoch": 0.02, "learning_rate": 1.4128195249198288e-05, "loss": 1.0151, "step": 11527 }, { "epoch": 0.02, "learning_rate": 1.4127213477142084e-05, "loss": 0.944, "step": 11528 }, { "epoch": 0.02, "learning_rate": 1.4126231657134757e-05, "loss": 0.9287, "step": 11529 }, { "epoch": 0.02, "learning_rate": 1.4125249789187718e-05, "loss": 0.9418, "step": 11530 }, { "epoch": 0.02, "learning_rate": 1.4124267873312376e-05, "loss": 0.9576, "step": 11531 }, { "epoch": 0.02, "learning_rate": 1.4123285909520136e-05, "loss": 0.8994, "step": 11532 }, { "epoch": 0.02, "learning_rate": 1.4122303897822405e-05, "loss": 0.9716, "step": 11533 }, { "epoch": 0.02, "learning_rate": 1.41213218382306e-05, "loss": 0.9154, "step": 11534 }, { "epoch": 0.02, "learning_rate": 1.412033973075612e-05, "loss": 0.951, "step": 11535 }, { "epoch": 0.02, "learning_rate": 1.4119357575410382e-05, "loss": 0.9186, "step": 11536 }, { "epoch": 0.02, "learning_rate": 1.4118375372204798e-05, "loss": 0.895, "step": 11537 }, { "epoch": 0.02, "learning_rate": 1.4117393121150776e-05, "loss": 0.9167, "step": 11538 }, { "epoch": 0.02, "learning_rate": 1.4116410822259729e-05, "loss": 0.9376, "step": 11539 }, { "epoch": 0.02, "learning_rate": 1.4115428475543072e-05, "loss": 0.9586, "step": 11540 }, { "epoch": 0.02, "learning_rate": 1.4114446081012214e-05, "loss": 0.9501, "step": 11541 }, { "epoch": 0.02, "learning_rate": 1.4113463638678573e-05, "loss": 0.9304, "step": 11542 }, { "epoch": 0.02, "learning_rate": 1.411248114855356e-05, "loss": 0.9513, "step": 11543 }, { "epoch": 0.02, "learning_rate": 1.4111498610648591e-05, "loss": 0.9418, "step": 11544 }, { "epoch": 0.02, "learning_rate": 1.4110516024975084e-05, "loss": 0.9163, "step": 11545 }, { "epoch": 0.02, "learning_rate": 1.4109533391544452e-05, "loss": 0.9081, "step": 11546 }, { "epoch": 0.02, "learning_rate": 1.4108550710368107e-05, "loss": 0.9539, "step": 11547 }, { "epoch": 0.02, "learning_rate": 1.4107567981457477e-05, "loss": 0.9317, "step": 11548 }, { "epoch": 0.02, "learning_rate": 1.410658520482397e-05, "loss": 0.8971, "step": 11549 }, { "epoch": 0.02, "learning_rate": 1.4105602380479009e-05, "loss": 0.9536, "step": 11550 }, { "epoch": 0.02, "learning_rate": 1.4104619508434013e-05, "loss": 0.8987, "step": 11551 }, { "epoch": 0.02, "learning_rate": 1.41036365887004e-05, "loss": 0.9156, "step": 11552 }, { "epoch": 0.02, "learning_rate": 1.4102653621289588e-05, "loss": 0.8764, "step": 11553 }, { "epoch": 0.02, "learning_rate": 1.4101670606213e-05, "loss": 0.9042, "step": 11554 }, { "epoch": 0.02, "learning_rate": 1.4100687543482056e-05, "loss": 0.9573, "step": 11555 }, { "epoch": 0.02, "learning_rate": 1.4099704433108177e-05, "loss": 0.8964, "step": 11556 }, { "epoch": 0.02, "learning_rate": 1.4098721275102784e-05, "loss": 0.8761, "step": 11557 }, { "epoch": 0.02, "learning_rate": 1.4097738069477304e-05, "loss": 0.9466, "step": 11558 }, { "epoch": 0.02, "learning_rate": 1.4096754816243158e-05, "loss": 0.942, "step": 11559 }, { "epoch": 0.02, "learning_rate": 1.4095771515411766e-05, "loss": 0.8914, "step": 11560 }, { "epoch": 0.02, "learning_rate": 1.4094788166994558e-05, "loss": 0.9508, "step": 11561 }, { "epoch": 0.02, "learning_rate": 1.4093804771002955e-05, "loss": 0.9498, "step": 11562 }, { "epoch": 0.02, "learning_rate": 1.4092821327448384e-05, "loss": 0.9335, "step": 11563 }, { "epoch": 0.02, "learning_rate": 1.409183783634227e-05, "loss": 0.8832, "step": 11564 }, { "epoch": 0.02, "learning_rate": 1.4090854297696041e-05, "loss": 0.9384, "step": 11565 }, { "epoch": 0.02, "learning_rate": 1.4089870711521121e-05, "loss": 0.9141, "step": 11566 }, { "epoch": 0.02, "learning_rate": 1.408888707782894e-05, "loss": 0.9503, "step": 11567 }, { "epoch": 0.02, "learning_rate": 1.4087903396630929e-05, "loss": 0.9174, "step": 11568 }, { "epoch": 0.02, "learning_rate": 1.4086919667938511e-05, "loss": 0.9298, "step": 11569 }, { "epoch": 0.02, "learning_rate": 1.4085935891763117e-05, "loss": 0.9428, "step": 11570 }, { "epoch": 0.02, "learning_rate": 1.4084952068116177e-05, "loss": 0.9502, "step": 11571 }, { "epoch": 0.02, "learning_rate": 1.4083968197009122e-05, "loss": 0.9304, "step": 11572 }, { "epoch": 0.02, "learning_rate": 1.4082984278453382e-05, "loss": 0.9025, "step": 11573 }, { "epoch": 0.02, "learning_rate": 1.4082000312460392e-05, "loss": 0.9493, "step": 11574 }, { "epoch": 0.02, "learning_rate": 1.4081016299041577e-05, "loss": 0.9553, "step": 11575 }, { "epoch": 0.02, "learning_rate": 1.4080032238208374e-05, "loss": 0.9106, "step": 11576 }, { "epoch": 0.02, "learning_rate": 1.407904812997222e-05, "loss": 0.9508, "step": 11577 }, { "epoch": 0.02, "learning_rate": 1.4078063974344542e-05, "loss": 0.9512, "step": 11578 }, { "epoch": 0.02, "learning_rate": 1.4077079771336774e-05, "loss": 0.9863, "step": 11579 }, { "epoch": 0.02, "learning_rate": 1.4076095520960357e-05, "loss": 0.9147, "step": 11580 }, { "epoch": 0.02, "learning_rate": 1.4075111223226721e-05, "loss": 0.937, "step": 11581 }, { "epoch": 0.02, "learning_rate": 1.4074126878147302e-05, "loss": 0.9172, "step": 11582 }, { "epoch": 0.02, "learning_rate": 1.407314248573354e-05, "loss": 0.9493, "step": 11583 }, { "epoch": 0.02, "learning_rate": 1.407215804599687e-05, "loss": 0.8641, "step": 11584 }, { "epoch": 0.02, "learning_rate": 1.4071173558948727e-05, "loss": 0.958, "step": 11585 }, { "epoch": 0.02, "learning_rate": 1.4070189024600554e-05, "loss": 0.901, "step": 11586 }, { "epoch": 0.02, "learning_rate": 1.4069204442963784e-05, "loss": 0.9533, "step": 11587 }, { "epoch": 0.02, "learning_rate": 1.406821981404986e-05, "loss": 0.88, "step": 11588 }, { "epoch": 0.02, "learning_rate": 1.4067235137870223e-05, "loss": 0.9357, "step": 11589 }, { "epoch": 0.02, "learning_rate": 1.4066250414436309e-05, "loss": 0.9289, "step": 11590 }, { "epoch": 0.02, "learning_rate": 1.406526564375956e-05, "loss": 0.9287, "step": 11591 }, { "epoch": 0.02, "learning_rate": 1.406428082585142e-05, "loss": 0.9067, "step": 11592 }, { "epoch": 0.02, "learning_rate": 1.4063295960723325e-05, "loss": 0.91, "step": 11593 }, { "epoch": 0.02, "learning_rate": 1.4062311048386726e-05, "loss": 0.84, "step": 11594 }, { "epoch": 0.02, "learning_rate": 1.406132608885306e-05, "loss": 0.9591, "step": 11595 }, { "epoch": 0.02, "learning_rate": 1.406034108213377e-05, "loss": 0.9105, "step": 11596 }, { "epoch": 0.02, "learning_rate": 1.4059356028240303e-05, "loss": 0.9441, "step": 11597 }, { "epoch": 0.02, "learning_rate": 1.4058370927184106e-05, "loss": 0.8941, "step": 11598 }, { "epoch": 0.02, "learning_rate": 1.4057385778976617e-05, "loss": 0.9261, "step": 11599 }, { "epoch": 0.02, "learning_rate": 1.4056400583629285e-05, "loss": 0.9594, "step": 11600 }, { "epoch": 0.02, "learning_rate": 1.4055415341153562e-05, "loss": 0.9293, "step": 11601 }, { "epoch": 0.02, "learning_rate": 1.4054430051560885e-05, "loss": 0.9256, "step": 11602 }, { "epoch": 0.02, "learning_rate": 1.4053444714862707e-05, "loss": 0.9386, "step": 11603 }, { "epoch": 0.02, "learning_rate": 1.4052459331070475e-05, "loss": 0.9113, "step": 11604 }, { "epoch": 0.02, "learning_rate": 1.405147390019564e-05, "loss": 0.9558, "step": 11605 }, { "epoch": 0.02, "learning_rate": 1.4050488422249644e-05, "loss": 0.9332, "step": 11606 }, { "epoch": 0.02, "learning_rate": 1.4049502897243943e-05, "loss": 0.9342, "step": 11607 }, { "epoch": 0.02, "learning_rate": 1.4048517325189985e-05, "loss": 0.9255, "step": 11608 }, { "epoch": 0.02, "learning_rate": 1.404753170609922e-05, "loss": 0.9448, "step": 11609 }, { "epoch": 0.02, "learning_rate": 1.4046546039983104e-05, "loss": 0.9294, "step": 11610 }, { "epoch": 0.02, "learning_rate": 1.4045560326853081e-05, "loss": 0.954, "step": 11611 }, { "epoch": 0.02, "learning_rate": 1.4044574566720605e-05, "loss": 0.9548, "step": 11612 }, { "epoch": 0.02, "learning_rate": 1.4043588759597135e-05, "loss": 0.9509, "step": 11613 }, { "epoch": 0.02, "learning_rate": 1.404260290549412e-05, "loss": 0.8938, "step": 11614 }, { "epoch": 0.02, "learning_rate": 1.4041617004423012e-05, "loss": 0.965, "step": 11615 }, { "epoch": 0.02, "learning_rate": 1.4040631056395268e-05, "loss": 0.8787, "step": 11616 }, { "epoch": 0.02, "learning_rate": 1.4039645061422345e-05, "loss": 0.9609, "step": 11617 }, { "epoch": 0.02, "learning_rate": 1.403865901951569e-05, "loss": 0.8726, "step": 11618 }, { "epoch": 0.02, "learning_rate": 1.4037672930686773e-05, "loss": 0.9121, "step": 11619 }, { "epoch": 0.02, "learning_rate": 1.4036686794947039e-05, "loss": 0.9266, "step": 11620 }, { "epoch": 0.02, "learning_rate": 1.4035700612307948e-05, "loss": 0.9711, "step": 11621 }, { "epoch": 0.02, "learning_rate": 1.4034714382780958e-05, "loss": 0.9477, "step": 11622 }, { "epoch": 0.02, "learning_rate": 1.4033728106377534e-05, "loss": 0.8974, "step": 11623 }, { "epoch": 0.02, "learning_rate": 1.4032741783109125e-05, "loss": 0.9307, "step": 11624 }, { "epoch": 0.02, "learning_rate": 1.4031755412987194e-05, "loss": 0.9582, "step": 11625 }, { "epoch": 0.02, "learning_rate": 1.4030768996023204e-05, "loss": 0.8702, "step": 11626 }, { "epoch": 0.02, "learning_rate": 1.4029782532228609e-05, "loss": 0.89, "step": 11627 }, { "epoch": 0.02, "learning_rate": 1.4028796021614877e-05, "loss": 0.9095, "step": 11628 }, { "epoch": 0.02, "learning_rate": 1.4027809464193465e-05, "loss": 0.9148, "step": 11629 }, { "epoch": 0.02, "learning_rate": 1.4026822859975837e-05, "loss": 0.9497, "step": 11630 }, { "epoch": 0.02, "learning_rate": 1.4025836208973456e-05, "loss": 0.9485, "step": 11631 }, { "epoch": 0.02, "learning_rate": 1.4024849511197783e-05, "loss": 0.9048, "step": 11632 }, { "epoch": 0.02, "learning_rate": 1.4023862766660284e-05, "loss": 0.92, "step": 11633 }, { "epoch": 0.02, "learning_rate": 1.4022875975372419e-05, "loss": 0.9608, "step": 11634 }, { "epoch": 0.02, "learning_rate": 1.402188913734566e-05, "loss": 0.955, "step": 11635 }, { "epoch": 0.02, "learning_rate": 1.4020902252591464e-05, "loss": 0.932, "step": 11636 }, { "epoch": 0.02, "learning_rate": 1.4019915321121307e-05, "loss": 0.9524, "step": 11637 }, { "epoch": 0.02, "learning_rate": 1.401892834294665e-05, "loss": 0.949, "step": 11638 }, { "epoch": 0.02, "learning_rate": 1.4017941318078953e-05, "loss": 0.9137, "step": 11639 }, { "epoch": 0.02, "learning_rate": 1.4016954246529697e-05, "loss": 0.9413, "step": 11640 }, { "epoch": 0.02, "learning_rate": 1.4015967128310339e-05, "loss": 0.9542, "step": 11641 }, { "epoch": 0.02, "learning_rate": 1.4014979963432355e-05, "loss": 0.9564, "step": 11642 }, { "epoch": 0.02, "learning_rate": 1.401399275190721e-05, "loss": 0.96, "step": 11643 }, { "epoch": 0.02, "learning_rate": 1.4013005493746374e-05, "loss": 0.9016, "step": 11644 }, { "epoch": 0.02, "learning_rate": 1.401201818896132e-05, "loss": 0.9165, "step": 11645 }, { "epoch": 0.02, "learning_rate": 1.4011030837563518e-05, "loss": 0.9429, "step": 11646 }, { "epoch": 0.02, "learning_rate": 1.4010043439564435e-05, "loss": 0.9163, "step": 11647 }, { "epoch": 0.02, "learning_rate": 1.4009055994975547e-05, "loss": 0.9382, "step": 11648 }, { "epoch": 0.02, "learning_rate": 1.4008068503808327e-05, "loss": 0.9529, "step": 11649 }, { "epoch": 0.02, "learning_rate": 1.4007080966074245e-05, "loss": 0.9052, "step": 11650 }, { "epoch": 0.02, "learning_rate": 1.4006093381784775e-05, "loss": 0.9476, "step": 11651 }, { "epoch": 0.02, "learning_rate": 1.4005105750951394e-05, "loss": 0.9129, "step": 11652 }, { "epoch": 0.02, "learning_rate": 1.4004118073585575e-05, "loss": 0.8989, "step": 11653 }, { "epoch": 0.02, "learning_rate": 1.4003130349698791e-05, "loss": 0.9519, "step": 11654 }, { "epoch": 0.02, "learning_rate": 1.4002142579302518e-05, "loss": 0.915, "step": 11655 }, { "epoch": 0.02, "learning_rate": 1.4001154762408238e-05, "loss": 0.949, "step": 11656 }, { "epoch": 0.02, "learning_rate": 1.4000166899027422e-05, "loss": 0.9298, "step": 11657 }, { "epoch": 0.02, "learning_rate": 1.3999178989171545e-05, "loss": 0.9266, "step": 11658 }, { "epoch": 0.02, "learning_rate": 1.399819103285209e-05, "loss": 0.9346, "step": 11659 }, { "epoch": 0.02, "learning_rate": 1.3997203030080536e-05, "loss": 0.9005, "step": 11660 }, { "epoch": 0.02, "learning_rate": 1.3996214980868354e-05, "loss": 0.9465, "step": 11661 }, { "epoch": 0.02, "learning_rate": 1.3995226885227036e-05, "loss": 0.9242, "step": 11662 }, { "epoch": 0.02, "learning_rate": 1.399423874316805e-05, "loss": 0.9534, "step": 11663 }, { "epoch": 0.02, "learning_rate": 1.3993250554702882e-05, "loss": 0.9425, "step": 11664 }, { "epoch": 0.02, "learning_rate": 1.3992262319843015e-05, "loss": 0.9603, "step": 11665 }, { "epoch": 0.02, "learning_rate": 1.3991274038599927e-05, "loss": 0.9127, "step": 11666 }, { "epoch": 0.02, "learning_rate": 1.3990285710985102e-05, "loss": 0.9104, "step": 11667 }, { "epoch": 0.02, "learning_rate": 1.3989297337010019e-05, "loss": 0.9156, "step": 11668 }, { "epoch": 0.02, "learning_rate": 1.3988308916686169e-05, "loss": 0.9251, "step": 11669 }, { "epoch": 0.02, "learning_rate": 1.3987320450025026e-05, "loss": 0.9514, "step": 11670 }, { "epoch": 0.02, "learning_rate": 1.398633193703808e-05, "loss": 0.9608, "step": 11671 }, { "epoch": 0.02, "learning_rate": 1.3985343377736817e-05, "loss": 0.9427, "step": 11672 }, { "epoch": 0.02, "learning_rate": 1.3984354772132718e-05, "loss": 0.9321, "step": 11673 }, { "epoch": 0.02, "learning_rate": 1.3983366120237273e-05, "loss": 0.9175, "step": 11674 }, { "epoch": 0.02, "learning_rate": 1.3982377422061968e-05, "loss": 0.899, "step": 11675 }, { "epoch": 0.02, "learning_rate": 1.3981388677618286e-05, "loss": 0.9134, "step": 11676 }, { "epoch": 0.02, "learning_rate": 1.398039988691772e-05, "loss": 0.9288, "step": 11677 }, { "epoch": 0.02, "learning_rate": 1.3979411049971754e-05, "loss": 0.8865, "step": 11678 }, { "epoch": 0.02, "learning_rate": 1.3978422166791877e-05, "loss": 0.9324, "step": 11679 }, { "epoch": 0.02, "learning_rate": 1.3977433237389577e-05, "loss": 0.9149, "step": 11680 }, { "epoch": 0.02, "learning_rate": 1.397644426177635e-05, "loss": 0.925, "step": 11681 }, { "epoch": 0.02, "learning_rate": 1.3975455239963676e-05, "loss": 0.9359, "step": 11682 }, { "epoch": 0.02, "learning_rate": 1.3974466171963057e-05, "loss": 0.9411, "step": 11683 }, { "epoch": 0.02, "learning_rate": 1.3973477057785977e-05, "loss": 0.913, "step": 11684 }, { "epoch": 0.02, "learning_rate": 1.3972487897443927e-05, "loss": 0.899, "step": 11685 }, { "epoch": 0.02, "learning_rate": 1.3971498690948403e-05, "loss": 0.9439, "step": 11686 }, { "epoch": 0.02, "learning_rate": 1.39705094383109e-05, "loss": 0.9407, "step": 11687 }, { "epoch": 0.02, "learning_rate": 1.3969520139542905e-05, "loss": 0.8869, "step": 11688 }, { "epoch": 0.02, "learning_rate": 1.3968530794655917e-05, "loss": 0.9214, "step": 11689 }, { "epoch": 0.02, "learning_rate": 1.3967541403661429e-05, "loss": 0.884, "step": 11690 }, { "epoch": 0.02, "learning_rate": 1.3966551966570933e-05, "loss": 0.9214, "step": 11691 }, { "epoch": 0.02, "learning_rate": 1.3965562483395926e-05, "loss": 0.8993, "step": 11692 }, { "epoch": 0.02, "learning_rate": 1.396457295414791e-05, "loss": 0.9074, "step": 11693 }, { "epoch": 0.02, "learning_rate": 1.3963583378838374e-05, "loss": 0.8866, "step": 11694 }, { "epoch": 0.02, "learning_rate": 1.396259375747882e-05, "loss": 0.927, "step": 11695 }, { "epoch": 0.02, "learning_rate": 1.3961604090080742e-05, "loss": 0.9355, "step": 11696 }, { "epoch": 0.02, "learning_rate": 1.396061437665564e-05, "loss": 0.9495, "step": 11697 }, { "epoch": 0.02, "learning_rate": 1.3959624617215013e-05, "loss": 0.8948, "step": 11698 }, { "epoch": 0.02, "learning_rate": 1.3958634811770361e-05, "loss": 0.9082, "step": 11699 }, { "epoch": 0.02, "learning_rate": 1.3957644960333183e-05, "loss": 0.9089, "step": 11700 }, { "epoch": 0.02, "learning_rate": 1.3956655062914978e-05, "loss": 0.8961, "step": 11701 }, { "epoch": 0.02, "learning_rate": 1.395566511952725e-05, "loss": 0.9226, "step": 11702 }, { "epoch": 0.02, "learning_rate": 1.3954675130181499e-05, "loss": 0.9561, "step": 11703 }, { "epoch": 0.02, "learning_rate": 1.3953685094889222e-05, "loss": 0.9267, "step": 11704 }, { "epoch": 0.02, "learning_rate": 1.395269501366193e-05, "loss": 0.9384, "step": 11705 }, { "epoch": 0.02, "learning_rate": 1.3951704886511123e-05, "loss": 0.9038, "step": 11706 }, { "epoch": 0.02, "learning_rate": 1.39507147134483e-05, "loss": 0.8833, "step": 11707 }, { "epoch": 0.02, "learning_rate": 1.3949724494484974e-05, "loss": 0.9191, "step": 11708 }, { "epoch": 0.02, "learning_rate": 1.3948734229632638e-05, "loss": 0.9514, "step": 11709 }, { "epoch": 0.02, "learning_rate": 1.3947743918902808e-05, "loss": 0.9251, "step": 11710 }, { "epoch": 0.02, "learning_rate": 1.3946753562306984e-05, "loss": 0.9447, "step": 11711 }, { "epoch": 0.02, "learning_rate": 1.3945763159856674e-05, "loss": 0.9397, "step": 11712 }, { "epoch": 0.02, "learning_rate": 1.3944772711563386e-05, "loss": 0.9685, "step": 11713 }, { "epoch": 0.02, "learning_rate": 1.3943782217438622e-05, "loss": 0.89, "step": 11714 }, { "epoch": 0.02, "learning_rate": 1.3942791677493897e-05, "loss": 0.9168, "step": 11715 }, { "epoch": 0.02, "learning_rate": 1.3941801091740714e-05, "loss": 0.957, "step": 11716 }, { "epoch": 0.02, "learning_rate": 1.3940810460190584e-05, "loss": 0.9319, "step": 11717 }, { "epoch": 0.02, "learning_rate": 1.3939819782855018e-05, "loss": 0.9465, "step": 11718 }, { "epoch": 0.02, "learning_rate": 1.393882905974552e-05, "loss": 0.9449, "step": 11719 }, { "epoch": 0.02, "learning_rate": 1.3937838290873609e-05, "loss": 0.9245, "step": 11720 }, { "epoch": 0.02, "learning_rate": 1.393684747625079e-05, "loss": 0.931, "step": 11721 }, { "epoch": 0.02, "learning_rate": 1.3935856615888574e-05, "loss": 0.9292, "step": 11722 }, { "epoch": 0.02, "learning_rate": 1.3934865709798482e-05, "loss": 0.9753, "step": 11723 }, { "epoch": 0.02, "learning_rate": 1.3933874757992015e-05, "loss": 0.9353, "step": 11724 }, { "epoch": 0.02, "learning_rate": 1.3932883760480688e-05, "loss": 0.892, "step": 11725 }, { "epoch": 0.02, "learning_rate": 1.3931892717276024e-05, "loss": 0.9247, "step": 11726 }, { "epoch": 0.02, "learning_rate": 1.393090162838953e-05, "loss": 0.899, "step": 11727 }, { "epoch": 0.02, "learning_rate": 1.3929910493832716e-05, "loss": 0.9107, "step": 11728 }, { "epoch": 0.02, "learning_rate": 1.3928919313617112e-05, "loss": 0.8995, "step": 11729 }, { "epoch": 0.02, "learning_rate": 1.3927928087754218e-05, "loss": 0.923, "step": 11730 }, { "epoch": 0.02, "learning_rate": 1.392693681625556e-05, "loss": 0.927, "step": 11731 }, { "epoch": 0.02, "learning_rate": 1.3925945499132653e-05, "loss": 0.914, "step": 11732 }, { "epoch": 0.02, "learning_rate": 1.3924954136397011e-05, "loss": 0.9337, "step": 11733 }, { "epoch": 0.02, "learning_rate": 1.3923962728060157e-05, "loss": 0.9079, "step": 11734 }, { "epoch": 0.02, "learning_rate": 1.3922971274133603e-05, "loss": 0.9082, "step": 11735 }, { "epoch": 0.02, "learning_rate": 1.3921979774628878e-05, "loss": 0.9178, "step": 11736 }, { "epoch": 0.02, "learning_rate": 1.392098822955749e-05, "loss": 0.9254, "step": 11737 }, { "epoch": 0.02, "learning_rate": 1.3919996638930967e-05, "loss": 0.9237, "step": 11738 }, { "epoch": 0.02, "learning_rate": 1.3919005002760826e-05, "loss": 0.9182, "step": 11739 }, { "epoch": 0.02, "learning_rate": 1.3918013321058587e-05, "loss": 0.9085, "step": 11740 }, { "epoch": 0.02, "learning_rate": 1.3917021593835778e-05, "loss": 0.9032, "step": 11741 }, { "epoch": 0.02, "learning_rate": 1.3916029821103916e-05, "loss": 0.9333, "step": 11742 }, { "epoch": 0.02, "learning_rate": 1.3915038002874522e-05, "loss": 0.9661, "step": 11743 }, { "epoch": 0.02, "learning_rate": 1.3914046139159123e-05, "loss": 0.9688, "step": 11744 }, { "epoch": 0.02, "learning_rate": 1.3913054229969243e-05, "loss": 0.91, "step": 11745 }, { "epoch": 0.02, "learning_rate": 1.3912062275316406e-05, "loss": 0.9642, "step": 11746 }, { "epoch": 0.02, "learning_rate": 1.391107027521213e-05, "loss": 0.9399, "step": 11747 }, { "epoch": 0.02, "learning_rate": 1.3910078229667953e-05, "loss": 0.9503, "step": 11748 }, { "epoch": 0.02, "learning_rate": 1.390908613869539e-05, "loss": 0.9359, "step": 11749 }, { "epoch": 0.02, "learning_rate": 1.3908094002305971e-05, "loss": 0.8924, "step": 11750 }, { "epoch": 0.02, "learning_rate": 1.3907101820511221e-05, "loss": 0.926, "step": 11751 }, { "epoch": 0.03, "learning_rate": 1.3906109593322676e-05, "loss": 0.9338, "step": 11752 }, { "epoch": 0.03, "learning_rate": 1.3905117320751851e-05, "loss": 0.9438, "step": 11753 }, { "epoch": 0.03, "learning_rate": 1.3904125002810287e-05, "loss": 0.898, "step": 11754 }, { "epoch": 0.03, "learning_rate": 1.3903132639509504e-05, "loss": 0.9416, "step": 11755 }, { "epoch": 0.03, "learning_rate": 1.3902140230861033e-05, "loss": 0.9548, "step": 11756 }, { "epoch": 0.03, "learning_rate": 1.3901147776876412e-05, "loss": 0.96, "step": 11757 }, { "epoch": 0.03, "learning_rate": 1.3900155277567157e-05, "loss": 0.9182, "step": 11758 }, { "epoch": 0.03, "learning_rate": 1.3899162732944812e-05, "loss": 0.9365, "step": 11759 }, { "epoch": 0.03, "learning_rate": 1.3898170143020903e-05, "loss": 0.8874, "step": 11760 }, { "epoch": 0.03, "learning_rate": 1.3897177507806967e-05, "loss": 0.9314, "step": 11761 }, { "epoch": 0.03, "learning_rate": 1.3896184827314529e-05, "loss": 0.9251, "step": 11762 }, { "epoch": 0.03, "learning_rate": 1.3895192101555129e-05, "loss": 0.9609, "step": 11763 }, { "epoch": 0.03, "learning_rate": 1.3894199330540297e-05, "loss": 0.9111, "step": 11764 }, { "epoch": 0.03, "learning_rate": 1.3893206514281566e-05, "loss": 0.9367, "step": 11765 }, { "epoch": 0.03, "learning_rate": 1.3892213652790475e-05, "loss": 0.8956, "step": 11766 }, { "epoch": 0.03, "learning_rate": 1.3891220746078558e-05, "loss": 0.9148, "step": 11767 }, { "epoch": 0.03, "learning_rate": 1.389022779415735e-05, "loss": 0.8667, "step": 11768 }, { "epoch": 0.03, "learning_rate": 1.3889234797038388e-05, "loss": 0.9181, "step": 11769 }, { "epoch": 0.03, "learning_rate": 1.388824175473321e-05, "loss": 0.8871, "step": 11770 }, { "epoch": 0.03, "learning_rate": 1.3887248667253348e-05, "loss": 0.9077, "step": 11771 }, { "epoch": 0.03, "learning_rate": 1.3886255534610347e-05, "loss": 0.8707, "step": 11772 }, { "epoch": 0.03, "learning_rate": 1.3885262356815743e-05, "loss": 0.9211, "step": 11773 }, { "epoch": 0.03, "learning_rate": 1.3884269133881073e-05, "loss": 0.9026, "step": 11774 }, { "epoch": 0.03, "learning_rate": 1.388327586581788e-05, "loss": 0.933, "step": 11775 }, { "epoch": 0.03, "learning_rate": 1.38822825526377e-05, "loss": 0.9257, "step": 11776 }, { "epoch": 0.03, "learning_rate": 1.3881289194352076e-05, "loss": 0.9477, "step": 11777 }, { "epoch": 0.03, "learning_rate": 1.388029579097255e-05, "loss": 0.9418, "step": 11778 }, { "epoch": 0.03, "learning_rate": 1.3879302342510666e-05, "loss": 0.9145, "step": 11779 }, { "epoch": 0.03, "learning_rate": 1.3878308848977958e-05, "loss": 0.9302, "step": 11780 }, { "epoch": 0.03, "learning_rate": 1.3877315310385973e-05, "loss": 0.9768, "step": 11781 }, { "epoch": 0.03, "learning_rate": 1.3876321726746258e-05, "loss": 0.9374, "step": 11782 }, { "epoch": 0.03, "learning_rate": 1.3875328098070351e-05, "loss": 0.9031, "step": 11783 }, { "epoch": 0.03, "learning_rate": 1.38743344243698e-05, "loss": 0.9072, "step": 11784 }, { "epoch": 0.03, "learning_rate": 1.387334070565615e-05, "loss": 0.9907, "step": 11785 }, { "epoch": 0.03, "learning_rate": 1.3872346941940938e-05, "loss": 0.8821, "step": 11786 }, { "epoch": 0.03, "learning_rate": 1.3871353133235723e-05, "loss": 0.9447, "step": 11787 }, { "epoch": 0.03, "learning_rate": 1.3870359279552041e-05, "loss": 0.9147, "step": 11788 }, { "epoch": 0.03, "learning_rate": 1.3869365380901445e-05, "loss": 0.9424, "step": 11789 }, { "epoch": 0.03, "learning_rate": 1.3868371437295479e-05, "loss": 0.9076, "step": 11790 }, { "epoch": 0.03, "learning_rate": 1.3867377448745693e-05, "loss": 0.956, "step": 11791 }, { "epoch": 0.03, "learning_rate": 1.3866383415263635e-05, "loss": 0.9389, "step": 11792 }, { "epoch": 0.03, "learning_rate": 1.3865389336860852e-05, "loss": 0.9417, "step": 11793 }, { "epoch": 0.03, "learning_rate": 1.3864395213548899e-05, "loss": 0.9342, "step": 11794 }, { "epoch": 0.03, "learning_rate": 1.3863401045339315e-05, "loss": 0.9188, "step": 11795 }, { "epoch": 0.03, "learning_rate": 1.3862406832243662e-05, "loss": 0.9164, "step": 11796 }, { "epoch": 0.03, "learning_rate": 1.3861412574273485e-05, "loss": 0.9285, "step": 11797 }, { "epoch": 0.03, "learning_rate": 1.3860418271440339e-05, "loss": 0.8965, "step": 11798 }, { "epoch": 0.03, "learning_rate": 1.385942392375577e-05, "loss": 0.9403, "step": 11799 }, { "epoch": 0.03, "learning_rate": 1.3858429531231338e-05, "loss": 0.8925, "step": 11800 }, { "epoch": 0.03, "learning_rate": 1.3857435093878592e-05, "loss": 0.974, "step": 11801 }, { "epoch": 0.03, "learning_rate": 1.3856440611709086e-05, "loss": 0.9441, "step": 11802 }, { "epoch": 0.03, "learning_rate": 1.3855446084734376e-05, "loss": 0.9115, "step": 11803 }, { "epoch": 0.03, "learning_rate": 1.3854451512966016e-05, "loss": 0.8939, "step": 11804 }, { "epoch": 0.03, "learning_rate": 1.3853456896415558e-05, "loss": 0.9076, "step": 11805 }, { "epoch": 0.03, "learning_rate": 1.3852462235094561e-05, "loss": 0.8826, "step": 11806 }, { "epoch": 0.03, "learning_rate": 1.3851467529014584e-05, "loss": 0.9472, "step": 11807 }, { "epoch": 0.03, "learning_rate": 1.3850472778187176e-05, "loss": 0.931, "step": 11808 }, { "epoch": 0.03, "learning_rate": 1.38494779826239e-05, "loss": 0.9142, "step": 11809 }, { "epoch": 0.03, "learning_rate": 1.3848483142336316e-05, "loss": 0.9176, "step": 11810 }, { "epoch": 0.03, "learning_rate": 1.3847488257335973e-05, "loss": 0.9285, "step": 11811 }, { "epoch": 0.03, "learning_rate": 1.384649332763444e-05, "loss": 0.8968, "step": 11812 }, { "epoch": 0.03, "learning_rate": 1.3845498353243269e-05, "loss": 0.9335, "step": 11813 }, { "epoch": 0.03, "learning_rate": 1.3844503334174024e-05, "loss": 0.9308, "step": 11814 }, { "epoch": 0.03, "learning_rate": 1.3843508270438266e-05, "loss": 0.9333, "step": 11815 }, { "epoch": 0.03, "learning_rate": 1.3842513162047553e-05, "loss": 0.9112, "step": 11816 }, { "epoch": 0.03, "learning_rate": 1.3841518009013446e-05, "loss": 0.9188, "step": 11817 }, { "epoch": 0.03, "learning_rate": 1.384052281134751e-05, "loss": 0.9297, "step": 11818 }, { "epoch": 0.03, "learning_rate": 1.3839527569061309e-05, "loss": 0.9194, "step": 11819 }, { "epoch": 0.03, "learning_rate": 1.3838532282166398e-05, "loss": 0.9389, "step": 11820 }, { "epoch": 0.03, "learning_rate": 1.3837536950674348e-05, "loss": 0.9241, "step": 11821 }, { "epoch": 0.03, "learning_rate": 1.3836541574596722e-05, "loss": 0.9106, "step": 11822 }, { "epoch": 0.03, "learning_rate": 1.3835546153945081e-05, "loss": 0.9235, "step": 11823 }, { "epoch": 0.03, "learning_rate": 1.3834550688730997e-05, "loss": 0.9236, "step": 11824 }, { "epoch": 0.03, "learning_rate": 1.3833555178966026e-05, "loss": 0.9343, "step": 11825 }, { "epoch": 0.03, "learning_rate": 1.3832559624661742e-05, "loss": 0.9313, "step": 11826 }, { "epoch": 0.03, "learning_rate": 1.3831564025829707e-05, "loss": 0.9596, "step": 11827 }, { "epoch": 0.03, "learning_rate": 1.383056838248149e-05, "loss": 0.9019, "step": 11828 }, { "epoch": 0.03, "learning_rate": 1.3829572694628658e-05, "loss": 0.9441, "step": 11829 }, { "epoch": 0.03, "learning_rate": 1.382857696228278e-05, "loss": 0.933, "step": 11830 }, { "epoch": 0.03, "learning_rate": 1.3827581185455427e-05, "loss": 0.9338, "step": 11831 }, { "epoch": 0.03, "learning_rate": 1.3826585364158161e-05, "loss": 0.945, "step": 11832 }, { "epoch": 0.03, "learning_rate": 1.382558949840256e-05, "loss": 0.9445, "step": 11833 }, { "epoch": 0.03, "learning_rate": 1.3824593588200191e-05, "loss": 0.9213, "step": 11834 }, { "epoch": 0.03, "learning_rate": 1.3823597633562619e-05, "loss": 0.8975, "step": 11835 }, { "epoch": 0.03, "learning_rate": 1.3822601634501425e-05, "loss": 0.9519, "step": 11836 }, { "epoch": 0.03, "learning_rate": 1.3821605591028178e-05, "loss": 0.9597, "step": 11837 }, { "epoch": 0.03, "learning_rate": 1.3820609503154446e-05, "loss": 0.9312, "step": 11838 }, { "epoch": 0.03, "learning_rate": 1.3819613370891802e-05, "loss": 0.9383, "step": 11839 }, { "epoch": 0.03, "learning_rate": 1.381861719425183e-05, "loss": 0.9497, "step": 11840 }, { "epoch": 0.03, "learning_rate": 1.3817620973246089e-05, "loss": 0.948, "step": 11841 }, { "epoch": 0.03, "learning_rate": 1.3816624707886159e-05, "loss": 0.9217, "step": 11842 }, { "epoch": 0.03, "learning_rate": 1.3815628398183617e-05, "loss": 0.9251, "step": 11843 }, { "epoch": 0.03, "learning_rate": 1.381463204415004e-05, "loss": 0.913, "step": 11844 }, { "epoch": 0.03, "learning_rate": 1.3813635645797001e-05, "loss": 0.9225, "step": 11845 }, { "epoch": 0.03, "learning_rate": 1.3812639203136077e-05, "loss": 0.932, "step": 11846 }, { "epoch": 0.03, "learning_rate": 1.3811642716178844e-05, "loss": 0.9406, "step": 11847 }, { "epoch": 0.03, "learning_rate": 1.381064618493688e-05, "loss": 0.8838, "step": 11848 }, { "epoch": 0.03, "learning_rate": 1.3809649609421765e-05, "loss": 0.9113, "step": 11849 }, { "epoch": 0.03, "learning_rate": 1.3808652989645073e-05, "loss": 0.9092, "step": 11850 }, { "epoch": 0.03, "learning_rate": 1.3807656325618386e-05, "loss": 0.9252, "step": 11851 }, { "epoch": 0.03, "learning_rate": 1.3806659617353285e-05, "loss": 0.9771, "step": 11852 }, { "epoch": 0.03, "learning_rate": 1.3805662864861346e-05, "loss": 0.956, "step": 11853 }, { "epoch": 0.03, "learning_rate": 1.3804666068154154e-05, "loss": 0.9381, "step": 11854 }, { "epoch": 0.03, "learning_rate": 1.3803669227243286e-05, "loss": 0.9643, "step": 11855 }, { "epoch": 0.03, "learning_rate": 1.3802672342140327e-05, "loss": 0.927, "step": 11856 }, { "epoch": 0.03, "learning_rate": 1.3801675412856854e-05, "loss": 0.9375, "step": 11857 }, { "epoch": 0.03, "learning_rate": 1.3800678439404455e-05, "loss": 0.9306, "step": 11858 }, { "epoch": 0.03, "learning_rate": 1.3799681421794712e-05, "loss": 0.9226, "step": 11859 }, { "epoch": 0.03, "learning_rate": 1.3798684360039208e-05, "loss": 0.9158, "step": 11860 }, { "epoch": 0.03, "learning_rate": 1.3797687254149528e-05, "loss": 0.9425, "step": 11861 }, { "epoch": 0.03, "learning_rate": 1.3796690104137253e-05, "loss": 0.9136, "step": 11862 }, { "epoch": 0.03, "learning_rate": 1.3795692910013974e-05, "loss": 0.9133, "step": 11863 }, { "epoch": 0.03, "learning_rate": 1.379469567179127e-05, "loss": 0.9241, "step": 11864 }, { "epoch": 0.03, "learning_rate": 1.3793698389480734e-05, "loss": 0.9087, "step": 11865 }, { "epoch": 0.03, "learning_rate": 1.3792701063093945e-05, "loss": 0.9018, "step": 11866 }, { "epoch": 0.03, "learning_rate": 1.3791703692642497e-05, "loss": 0.9286, "step": 11867 }, { "epoch": 0.03, "learning_rate": 1.3790706278137977e-05, "loss": 0.9252, "step": 11868 }, { "epoch": 0.03, "learning_rate": 1.378970881959197e-05, "loss": 0.9469, "step": 11869 }, { "epoch": 0.03, "learning_rate": 1.3788711317016067e-05, "loss": 0.894, "step": 11870 }, { "epoch": 0.03, "learning_rate": 1.3787713770421855e-05, "loss": 0.9231, "step": 11871 }, { "epoch": 0.03, "learning_rate": 1.3786716179820927e-05, "loss": 0.9024, "step": 11872 }, { "epoch": 0.03, "learning_rate": 1.3785718545224871e-05, "loss": 0.9145, "step": 11873 }, { "epoch": 0.03, "learning_rate": 1.378472086664528e-05, "loss": 0.9289, "step": 11874 }, { "epoch": 0.03, "learning_rate": 1.378372314409374e-05, "loss": 0.9323, "step": 11875 }, { "epoch": 0.03, "learning_rate": 1.378272537758185e-05, "loss": 0.9118, "step": 11876 }, { "epoch": 0.03, "learning_rate": 1.3781727567121199e-05, "loss": 0.8832, "step": 11877 }, { "epoch": 0.03, "learning_rate": 1.3780729712723377e-05, "loss": 0.9119, "step": 11878 }, { "epoch": 0.03, "learning_rate": 1.377973181439998e-05, "loss": 0.9441, "step": 11879 }, { "epoch": 0.03, "learning_rate": 1.3778733872162607e-05, "loss": 0.8983, "step": 11880 }, { "epoch": 0.03, "learning_rate": 1.3777735886022842e-05, "loss": 0.9226, "step": 11881 }, { "epoch": 0.03, "learning_rate": 1.3776737855992287e-05, "loss": 0.9304, "step": 11882 }, { "epoch": 0.03, "learning_rate": 1.3775739782082538e-05, "loss": 0.9651, "step": 11883 }, { "epoch": 0.03, "learning_rate": 1.3774741664305185e-05, "loss": 0.9412, "step": 11884 }, { "epoch": 0.03, "learning_rate": 1.3773743502671829e-05, "loss": 0.8985, "step": 11885 }, { "epoch": 0.03, "learning_rate": 1.3772745297194069e-05, "loss": 0.9098, "step": 11886 }, { "epoch": 0.03, "learning_rate": 1.3771747047883495e-05, "loss": 0.9631, "step": 11887 }, { "epoch": 0.03, "learning_rate": 1.3770748754751708e-05, "loss": 0.932, "step": 11888 }, { "epoch": 0.03, "learning_rate": 1.376975041781031e-05, "loss": 0.9414, "step": 11889 }, { "epoch": 0.03, "learning_rate": 1.3768752037070898e-05, "loss": 0.9324, "step": 11890 }, { "epoch": 0.03, "learning_rate": 1.3767753612545071e-05, "loss": 0.9342, "step": 11891 }, { "epoch": 0.03, "learning_rate": 1.376675514424443e-05, "loss": 0.9166, "step": 11892 }, { "epoch": 0.03, "learning_rate": 1.3765756632180573e-05, "loss": 0.9673, "step": 11893 }, { "epoch": 0.03, "learning_rate": 1.3764758076365103e-05, "loss": 0.927, "step": 11894 }, { "epoch": 0.03, "learning_rate": 1.3763759476809621e-05, "loss": 0.9366, "step": 11895 }, { "epoch": 0.03, "learning_rate": 1.376276083352573e-05, "loss": 0.919, "step": 11896 }, { "epoch": 0.03, "learning_rate": 1.376176214652503e-05, "loss": 0.8837, "step": 11897 }, { "epoch": 0.03, "learning_rate": 1.3760763415819125e-05, "loss": 0.9236, "step": 11898 }, { "epoch": 0.03, "learning_rate": 1.3759764641419624e-05, "loss": 0.9284, "step": 11899 }, { "epoch": 0.03, "learning_rate": 1.375876582333812e-05, "loss": 0.9467, "step": 11900 }, { "epoch": 0.03, "learning_rate": 1.3757766961586224e-05, "loss": 0.9472, "step": 11901 }, { "epoch": 0.03, "learning_rate": 1.3756768056175545e-05, "loss": 0.901, "step": 11902 }, { "epoch": 0.03, "learning_rate": 1.3755769107117682e-05, "loss": 0.9184, "step": 11903 }, { "epoch": 0.03, "learning_rate": 1.3754770114424245e-05, "loss": 0.9206, "step": 11904 }, { "epoch": 0.03, "learning_rate": 1.3753771078106837e-05, "loss": 0.9713, "step": 11905 }, { "epoch": 0.03, "learning_rate": 1.3752771998177068e-05, "loss": 0.9288, "step": 11906 }, { "epoch": 0.03, "learning_rate": 1.3751772874646544e-05, "loss": 0.9109, "step": 11907 }, { "epoch": 0.03, "learning_rate": 1.3750773707526875e-05, "loss": 0.9059, "step": 11908 }, { "epoch": 0.03, "learning_rate": 1.374977449682967e-05, "loss": 0.9233, "step": 11909 }, { "epoch": 0.03, "learning_rate": 1.374877524256653e-05, "loss": 0.9269, "step": 11910 }, { "epoch": 0.03, "learning_rate": 1.3747775944749078e-05, "loss": 0.9242, "step": 11911 }, { "epoch": 0.03, "learning_rate": 1.3746776603388912e-05, "loss": 0.912, "step": 11912 }, { "epoch": 0.03, "learning_rate": 1.3745777218497651e-05, "loss": 0.938, "step": 11913 }, { "epoch": 0.03, "learning_rate": 1.3744777790086903e-05, "loss": 0.9171, "step": 11914 }, { "epoch": 0.03, "learning_rate": 1.3743778318168277e-05, "loss": 0.9277, "step": 11915 }, { "epoch": 0.03, "learning_rate": 1.3742778802753392e-05, "loss": 0.8894, "step": 11916 }, { "epoch": 0.03, "learning_rate": 1.3741779243853852e-05, "loss": 0.9589, "step": 11917 }, { "epoch": 0.03, "learning_rate": 1.3740779641481275e-05, "loss": 0.9576, "step": 11918 }, { "epoch": 0.03, "learning_rate": 1.3739779995647276e-05, "loss": 0.8916, "step": 11919 }, { "epoch": 0.03, "learning_rate": 1.3738780306363469e-05, "loss": 0.9018, "step": 11920 }, { "epoch": 0.03, "learning_rate": 1.3737780573641462e-05, "loss": 0.9079, "step": 11921 }, { "epoch": 0.03, "learning_rate": 1.373678079749288e-05, "loss": 0.9296, "step": 11922 }, { "epoch": 0.03, "learning_rate": 1.3735780977929331e-05, "loss": 0.9686, "step": 11923 }, { "epoch": 0.03, "learning_rate": 1.3734781114962434e-05, "loss": 0.935, "step": 11924 }, { "epoch": 0.03, "learning_rate": 1.3733781208603807e-05, "loss": 0.948, "step": 11925 }, { "epoch": 0.03, "learning_rate": 1.3732781258865066e-05, "loss": 0.9074, "step": 11926 }, { "epoch": 0.03, "learning_rate": 1.3731781265757828e-05, "loss": 0.9101, "step": 11927 }, { "epoch": 0.03, "learning_rate": 1.3730781229293713e-05, "loss": 0.9215, "step": 11928 }, { "epoch": 0.03, "learning_rate": 1.3729781149484339e-05, "loss": 0.9095, "step": 11929 }, { "epoch": 0.03, "learning_rate": 1.3728781026341323e-05, "loss": 0.9231, "step": 11930 }, { "epoch": 0.03, "learning_rate": 1.3727780859876285e-05, "loss": 0.9536, "step": 11931 }, { "epoch": 0.03, "learning_rate": 1.372678065010085e-05, "loss": 0.9453, "step": 11932 }, { "epoch": 0.03, "learning_rate": 1.3725780397026633e-05, "loss": 0.9218, "step": 11933 }, { "epoch": 0.03, "learning_rate": 1.3724780100665261e-05, "loss": 0.9151, "step": 11934 }, { "epoch": 0.03, "learning_rate": 1.3723779761028349e-05, "loss": 0.956, "step": 11935 }, { "epoch": 0.03, "learning_rate": 1.3722779378127521e-05, "loss": 0.8907, "step": 11936 }, { "epoch": 0.03, "learning_rate": 1.3721778951974405e-05, "loss": 0.919, "step": 11937 }, { "epoch": 0.03, "learning_rate": 1.372077848258062e-05, "loss": 0.9191, "step": 11938 }, { "epoch": 0.03, "learning_rate": 1.371977796995779e-05, "loss": 0.941, "step": 11939 }, { "epoch": 0.03, "learning_rate": 1.3718777414117539e-05, "loss": 0.903, "step": 11940 }, { "epoch": 0.03, "learning_rate": 1.3717776815071493e-05, "loss": 0.9445, "step": 11941 }, { "epoch": 0.03, "learning_rate": 1.3716776172831274e-05, "loss": 0.9395, "step": 11942 }, { "epoch": 0.03, "learning_rate": 1.3715775487408514e-05, "loss": 0.9372, "step": 11943 }, { "epoch": 0.03, "learning_rate": 1.3714774758814834e-05, "loss": 0.9565, "step": 11944 }, { "epoch": 0.03, "learning_rate": 1.3713773987061861e-05, "loss": 0.943, "step": 11945 }, { "epoch": 0.03, "learning_rate": 1.3712773172161225e-05, "loss": 0.8945, "step": 11946 }, { "epoch": 0.03, "learning_rate": 1.371177231412455e-05, "loss": 0.9143, "step": 11947 }, { "epoch": 0.03, "learning_rate": 1.3710771412963466e-05, "loss": 0.893, "step": 11948 }, { "epoch": 0.03, "learning_rate": 1.3709770468689603e-05, "loss": 0.9133, "step": 11949 }, { "epoch": 0.03, "learning_rate": 1.3708769481314592e-05, "loss": 0.9204, "step": 11950 }, { "epoch": 0.03, "learning_rate": 1.370776845085006e-05, "loss": 0.935, "step": 11951 }, { "epoch": 0.03, "learning_rate": 1.3706767377307635e-05, "loss": 0.9311, "step": 11952 }, { "epoch": 0.03, "learning_rate": 1.3705766260698952e-05, "loss": 0.9375, "step": 11953 }, { "epoch": 0.03, "learning_rate": 1.3704765101035639e-05, "loss": 0.9032, "step": 11954 }, { "epoch": 0.03, "learning_rate": 1.370376389832933e-05, "loss": 0.9172, "step": 11955 }, { "epoch": 0.03, "learning_rate": 1.3702762652591655e-05, "loss": 0.9319, "step": 11956 }, { "epoch": 0.03, "learning_rate": 1.3701761363834252e-05, "loss": 0.8981, "step": 11957 }, { "epoch": 0.03, "learning_rate": 1.3700760032068745e-05, "loss": 0.9162, "step": 11958 }, { "epoch": 0.03, "learning_rate": 1.3699758657306777e-05, "loss": 0.9071, "step": 11959 }, { "epoch": 0.03, "learning_rate": 1.3698757239559978e-05, "loss": 0.9384, "step": 11960 }, { "epoch": 0.03, "learning_rate": 1.3697755778839983e-05, "loss": 0.9368, "step": 11961 }, { "epoch": 0.03, "learning_rate": 1.3696754275158427e-05, "loss": 0.9335, "step": 11962 }, { "epoch": 0.03, "learning_rate": 1.3695752728526948e-05, "loss": 0.968, "step": 11963 }, { "epoch": 0.03, "learning_rate": 1.3694751138957179e-05, "loss": 0.8878, "step": 11964 }, { "epoch": 0.03, "learning_rate": 1.3693749506460757e-05, "loss": 0.936, "step": 11965 }, { "epoch": 0.03, "learning_rate": 1.3692747831049322e-05, "loss": 0.9354, "step": 11966 }, { "epoch": 0.03, "learning_rate": 1.3691746112734507e-05, "loss": 0.9399, "step": 11967 }, { "epoch": 0.03, "learning_rate": 1.3690744351527958e-05, "loss": 0.9231, "step": 11968 }, { "epoch": 0.03, "learning_rate": 1.3689742547441309e-05, "loss": 0.9241, "step": 11969 }, { "epoch": 0.03, "learning_rate": 1.3688740700486196e-05, "loss": 0.9449, "step": 11970 }, { "epoch": 0.03, "learning_rate": 1.3687738810674266e-05, "loss": 0.9274, "step": 11971 }, { "epoch": 0.03, "learning_rate": 1.3686736878017152e-05, "loss": 0.8942, "step": 11972 }, { "epoch": 0.03, "learning_rate": 1.36857349025265e-05, "loss": 0.8826, "step": 11973 }, { "epoch": 0.03, "learning_rate": 1.3684732884213947e-05, "loss": 0.9111, "step": 11974 }, { "epoch": 0.03, "learning_rate": 1.3683730823091142e-05, "loss": 0.9427, "step": 11975 }, { "epoch": 0.03, "learning_rate": 1.3682728719169717e-05, "loss": 0.931, "step": 11976 }, { "epoch": 0.03, "learning_rate": 1.3681726572461325e-05, "loss": 0.9221, "step": 11977 }, { "epoch": 0.03, "learning_rate": 1.3680724382977602e-05, "loss": 0.9503, "step": 11978 }, { "epoch": 0.03, "learning_rate": 1.367972215073019e-05, "loss": 0.9423, "step": 11979 }, { "epoch": 0.03, "learning_rate": 1.3678719875730744e-05, "loss": 0.9121, "step": 11980 }, { "epoch": 0.03, "learning_rate": 1.3677717557990897e-05, "loss": 0.9405, "step": 11981 }, { "epoch": 0.03, "learning_rate": 1.3676715197522302e-05, "loss": 0.9174, "step": 11982 }, { "epoch": 0.03, "learning_rate": 1.36757127943366e-05, "loss": 0.9503, "step": 11983 }, { "epoch": 0.03, "learning_rate": 1.3674710348445441e-05, "loss": 0.9046, "step": 11984 }, { "epoch": 0.03, "learning_rate": 1.3673707859860468e-05, "loss": 0.9265, "step": 11985 }, { "epoch": 0.03, "learning_rate": 1.3672705328593329e-05, "loss": 0.8935, "step": 11986 }, { "epoch": 0.03, "learning_rate": 1.3671702754655675e-05, "loss": 0.9451, "step": 11987 }, { "epoch": 0.03, "learning_rate": 1.3670700138059151e-05, "loss": 0.9275, "step": 11988 }, { "epoch": 0.03, "learning_rate": 1.3669697478815407e-05, "loss": 0.9288, "step": 11989 }, { "epoch": 0.03, "learning_rate": 1.366869477693609e-05, "loss": 0.9162, "step": 11990 }, { "epoch": 0.03, "learning_rate": 1.3667692032432854e-05, "loss": 0.9116, "step": 11991 }, { "epoch": 0.03, "learning_rate": 1.3666689245317344e-05, "loss": 0.954, "step": 11992 }, { "epoch": 0.03, "learning_rate": 1.3665686415601214e-05, "loss": 0.9657, "step": 11993 }, { "epoch": 0.03, "learning_rate": 1.3664683543296114e-05, "loss": 0.9391, "step": 11994 }, { "epoch": 0.03, "learning_rate": 1.3663680628413694e-05, "loss": 0.9128, "step": 11995 }, { "epoch": 0.03, "learning_rate": 1.3662677670965612e-05, "loss": 0.8776, "step": 11996 }, { "epoch": 0.03, "learning_rate": 1.3661674670963514e-05, "loss": 0.8806, "step": 11997 }, { "epoch": 0.03, "learning_rate": 1.3660671628419055e-05, "loss": 0.9465, "step": 11998 }, { "epoch": 0.03, "learning_rate": 1.3659668543343893e-05, "loss": 0.9281, "step": 11999 }, { "epoch": 0.03, "learning_rate": 1.3658665415749677e-05, "loss": 0.9192, "step": 12000 }, { "epoch": 0.03, "learning_rate": 1.3657662245648063e-05, "loss": 0.9213, "step": 12001 }, { "epoch": 0.03, "learning_rate": 1.3656659033050706e-05, "loss": 0.9368, "step": 12002 }, { "epoch": 0.03, "learning_rate": 1.3655655777969263e-05, "loss": 0.9325, "step": 12003 }, { "epoch": 0.03, "learning_rate": 1.3654652480415387e-05, "loss": 0.8775, "step": 12004 }, { "epoch": 0.03, "learning_rate": 1.365364914040074e-05, "loss": 0.9315, "step": 12005 }, { "epoch": 0.03, "learning_rate": 1.3652645757936974e-05, "loss": 0.9692, "step": 12006 }, { "epoch": 0.03, "learning_rate": 1.3651642333035747e-05, "loss": 0.9731, "step": 12007 }, { "epoch": 0.03, "learning_rate": 1.3650638865708721e-05, "loss": 0.9201, "step": 12008 }, { "epoch": 0.03, "learning_rate": 1.364963535596755e-05, "loss": 0.9702, "step": 12009 }, { "epoch": 0.03, "learning_rate": 1.3648631803823892e-05, "loss": 0.9187, "step": 12010 }, { "epoch": 0.03, "learning_rate": 1.3647628209289415e-05, "loss": 0.9338, "step": 12011 }, { "epoch": 0.03, "learning_rate": 1.3646624572375773e-05, "loss": 0.9155, "step": 12012 }, { "epoch": 0.03, "learning_rate": 1.3645620893094623e-05, "loss": 0.9848, "step": 12013 }, { "epoch": 0.03, "learning_rate": 1.3644617171457632e-05, "loss": 0.9169, "step": 12014 }, { "epoch": 0.03, "learning_rate": 1.3643613407476463e-05, "loss": 0.915, "step": 12015 }, { "epoch": 0.03, "learning_rate": 1.3642609601162771e-05, "loss": 0.884, "step": 12016 }, { "epoch": 0.03, "learning_rate": 1.3641605752528225e-05, "loss": 0.965, "step": 12017 }, { "epoch": 0.03, "learning_rate": 1.3640601861584481e-05, "loss": 0.9134, "step": 12018 }, { "epoch": 0.03, "learning_rate": 1.3639597928343208e-05, "loss": 0.9363, "step": 12019 }, { "epoch": 0.03, "learning_rate": 1.363859395281607e-05, "loss": 0.8849, "step": 12020 }, { "epoch": 0.03, "learning_rate": 1.3637589935014732e-05, "loss": 0.9219, "step": 12021 }, { "epoch": 0.03, "learning_rate": 1.3636585874950853e-05, "loss": 0.9387, "step": 12022 }, { "epoch": 0.03, "learning_rate": 1.3635581772636104e-05, "loss": 0.9346, "step": 12023 }, { "epoch": 0.03, "learning_rate": 1.3634577628082152e-05, "loss": 0.9146, "step": 12024 }, { "epoch": 0.03, "learning_rate": 1.3633573441300656e-05, "loss": 0.9458, "step": 12025 }, { "epoch": 0.03, "learning_rate": 1.3632569212303293e-05, "loss": 0.9115, "step": 12026 }, { "epoch": 0.03, "learning_rate": 1.3631564941101723e-05, "loss": 0.9502, "step": 12027 }, { "epoch": 0.03, "learning_rate": 1.3630560627707615e-05, "loss": 0.9341, "step": 12028 }, { "epoch": 0.03, "learning_rate": 1.3629556272132638e-05, "loss": 0.8919, "step": 12029 }, { "epoch": 0.03, "learning_rate": 1.3628551874388465e-05, "loss": 0.9272, "step": 12030 }, { "epoch": 0.03, "learning_rate": 1.3627547434486758e-05, "loss": 0.9202, "step": 12031 }, { "epoch": 0.03, "learning_rate": 1.3626542952439192e-05, "loss": 0.8886, "step": 12032 }, { "epoch": 0.03, "learning_rate": 1.3625538428257438e-05, "loss": 0.9276, "step": 12033 }, { "epoch": 0.03, "learning_rate": 1.3624533861953162e-05, "loss": 0.9388, "step": 12034 }, { "epoch": 0.03, "learning_rate": 1.3623529253538039e-05, "loss": 0.9021, "step": 12035 }, { "epoch": 0.03, "learning_rate": 1.3622524603023739e-05, "loss": 0.9312, "step": 12036 }, { "epoch": 0.03, "learning_rate": 1.3621519910421937e-05, "loss": 0.9196, "step": 12037 }, { "epoch": 0.03, "learning_rate": 1.3620515175744307e-05, "loss": 0.9542, "step": 12038 }, { "epoch": 0.03, "learning_rate": 1.3619510399002513e-05, "loss": 0.9299, "step": 12039 }, { "epoch": 0.03, "learning_rate": 1.3618505580208238e-05, "loss": 0.8837, "step": 12040 }, { "epoch": 0.03, "learning_rate": 1.3617500719373154e-05, "loss": 0.941, "step": 12041 }, { "epoch": 0.03, "learning_rate": 1.3616495816508936e-05, "loss": 0.9402, "step": 12042 }, { "epoch": 0.03, "learning_rate": 1.3615490871627257e-05, "loss": 0.9119, "step": 12043 }, { "epoch": 0.03, "learning_rate": 1.3614485884739795e-05, "loss": 0.9143, "step": 12044 }, { "epoch": 0.03, "learning_rate": 1.3613480855858226e-05, "loss": 0.8633, "step": 12045 }, { "epoch": 0.03, "learning_rate": 1.3612475784994223e-05, "loss": 0.9001, "step": 12046 }, { "epoch": 0.03, "learning_rate": 1.3611470672159468e-05, "loss": 0.9443, "step": 12047 }, { "epoch": 0.03, "learning_rate": 1.3610465517365636e-05, "loss": 0.9525, "step": 12048 }, { "epoch": 0.03, "learning_rate": 1.360946032062441e-05, "loss": 0.9206, "step": 12049 }, { "epoch": 0.03, "learning_rate": 1.3608455081947458e-05, "loss": 0.9169, "step": 12050 }, { "epoch": 0.03, "learning_rate": 1.3607449801346472e-05, "loss": 0.9198, "step": 12051 }, { "epoch": 0.04, "learning_rate": 1.3606444478833123e-05, "loss": 0.9139, "step": 12052 }, { "epoch": 0.04, "learning_rate": 1.3605439114419095e-05, "loss": 0.9836, "step": 12053 }, { "epoch": 0.04, "learning_rate": 1.3604433708116068e-05, "loss": 0.9291, "step": 12054 }, { "epoch": 0.04, "learning_rate": 1.3603428259935718e-05, "loss": 0.8965, "step": 12055 }, { "epoch": 0.04, "learning_rate": 1.3602422769889734e-05, "loss": 0.9408, "step": 12056 }, { "epoch": 0.04, "learning_rate": 1.3601417237989793e-05, "loss": 0.9325, "step": 12057 }, { "epoch": 0.04, "learning_rate": 1.3600411664247584e-05, "loss": 0.8902, "step": 12058 }, { "epoch": 0.04, "learning_rate": 1.359940604867478e-05, "loss": 0.9038, "step": 12059 }, { "epoch": 0.04, "learning_rate": 1.3598400391283075e-05, "loss": 0.936, "step": 12060 }, { "epoch": 0.04, "learning_rate": 1.3597394692084148e-05, "loss": 0.9218, "step": 12061 }, { "epoch": 0.04, "learning_rate": 1.3596388951089679e-05, "loss": 0.8965, "step": 12062 }, { "epoch": 0.04, "learning_rate": 1.3595383168311363e-05, "loss": 0.9444, "step": 12063 }, { "epoch": 0.04, "learning_rate": 1.3594377343760876e-05, "loss": 0.9171, "step": 12064 }, { "epoch": 0.04, "learning_rate": 1.359337147744991e-05, "loss": 0.9271, "step": 12065 }, { "epoch": 0.04, "learning_rate": 1.3592365569390146e-05, "loss": 0.8895, "step": 12066 }, { "epoch": 0.04, "learning_rate": 1.359135961959328e-05, "loss": 0.9359, "step": 12067 }, { "epoch": 0.04, "learning_rate": 1.3590353628070991e-05, "loss": 0.9306, "step": 12068 }, { "epoch": 0.04, "learning_rate": 1.358934759483497e-05, "loss": 0.9163, "step": 12069 }, { "epoch": 0.04, "learning_rate": 1.3588341519896906e-05, "loss": 0.9228, "step": 12070 }, { "epoch": 0.04, "learning_rate": 1.3587335403268482e-05, "loss": 0.896, "step": 12071 }, { "epoch": 0.04, "learning_rate": 1.3586329244961398e-05, "loss": 0.9317, "step": 12072 }, { "epoch": 0.04, "learning_rate": 1.3585323044987338e-05, "loss": 0.9275, "step": 12073 }, { "epoch": 0.04, "learning_rate": 1.3584316803357989e-05, "loss": 0.9181, "step": 12074 }, { "epoch": 0.04, "learning_rate": 1.3583310520085046e-05, "loss": 0.9394, "step": 12075 }, { "epoch": 0.04, "learning_rate": 1.3582304195180202e-05, "loss": 0.908, "step": 12076 }, { "epoch": 0.04, "learning_rate": 1.3581297828655144e-05, "loss": 0.9453, "step": 12077 }, { "epoch": 0.04, "learning_rate": 1.3580291420521565e-05, "loss": 0.9147, "step": 12078 }, { "epoch": 0.04, "learning_rate": 1.3579284970791164e-05, "loss": 0.9109, "step": 12079 }, { "epoch": 0.04, "learning_rate": 1.3578278479475626e-05, "loss": 0.9306, "step": 12080 }, { "epoch": 0.04, "learning_rate": 1.357727194658665e-05, "loss": 0.9462, "step": 12081 }, { "epoch": 0.04, "learning_rate": 1.3576265372135926e-05, "loss": 0.9102, "step": 12082 }, { "epoch": 0.04, "learning_rate": 1.357525875613515e-05, "loss": 0.9388, "step": 12083 }, { "epoch": 0.04, "learning_rate": 1.3574252098596024e-05, "loss": 0.9295, "step": 12084 }, { "epoch": 0.04, "learning_rate": 1.3573245399530232e-05, "loss": 0.9065, "step": 12085 }, { "epoch": 0.04, "learning_rate": 1.3572238658949477e-05, "loss": 0.8884, "step": 12086 }, { "epoch": 0.04, "learning_rate": 1.3571231876865456e-05, "loss": 0.9517, "step": 12087 }, { "epoch": 0.04, "learning_rate": 1.3570225053289864e-05, "loss": 0.8838, "step": 12088 }, { "epoch": 0.04, "learning_rate": 1.3569218188234399e-05, "loss": 0.9052, "step": 12089 }, { "epoch": 0.04, "learning_rate": 1.356821128171076e-05, "loss": 0.9484, "step": 12090 }, { "epoch": 0.04, "learning_rate": 1.3567204333730643e-05, "loss": 0.9101, "step": 12091 }, { "epoch": 0.04, "learning_rate": 1.356619734430575e-05, "loss": 0.922, "step": 12092 }, { "epoch": 0.04, "learning_rate": 1.3565190313447777e-05, "loss": 0.9179, "step": 12093 }, { "epoch": 0.04, "learning_rate": 1.3564183241168427e-05, "loss": 0.8875, "step": 12094 }, { "epoch": 0.04, "learning_rate": 1.3563176127479402e-05, "loss": 0.9361, "step": 12095 }, { "epoch": 0.04, "learning_rate": 1.3562168972392398e-05, "loss": 0.9109, "step": 12096 }, { "epoch": 0.04, "learning_rate": 1.356116177591912e-05, "loss": 0.9066, "step": 12097 }, { "epoch": 0.04, "learning_rate": 1.3560154538071268e-05, "loss": 0.9313, "step": 12098 }, { "epoch": 0.04, "learning_rate": 1.3559147258860544e-05, "loss": 0.9769, "step": 12099 }, { "epoch": 0.04, "learning_rate": 1.3558139938298655e-05, "loss": 0.9093, "step": 12100 }, { "epoch": 0.04, "learning_rate": 1.35571325763973e-05, "loss": 0.9579, "step": 12101 }, { "epoch": 0.04, "learning_rate": 1.3556125173168182e-05, "loss": 0.9306, "step": 12102 }, { "epoch": 0.04, "learning_rate": 1.3555117728623011e-05, "loss": 0.9152, "step": 12103 }, { "epoch": 0.04, "learning_rate": 1.3554110242773489e-05, "loss": 0.9059, "step": 12104 }, { "epoch": 0.04, "learning_rate": 1.3553102715631315e-05, "loss": 0.924, "step": 12105 }, { "epoch": 0.04, "learning_rate": 1.3552095147208204e-05, "loss": 0.934, "step": 12106 }, { "epoch": 0.04, "learning_rate": 1.3551087537515859e-05, "loss": 0.9395, "step": 12107 }, { "epoch": 0.04, "learning_rate": 1.355007988656598e-05, "loss": 0.8906, "step": 12108 }, { "epoch": 0.04, "learning_rate": 1.3549072194370289e-05, "loss": 0.9264, "step": 12109 }, { "epoch": 0.04, "learning_rate": 1.354806446094048e-05, "loss": 0.9, "step": 12110 }, { "epoch": 0.04, "learning_rate": 1.3547056686288267e-05, "loss": 0.9384, "step": 12111 }, { "epoch": 0.04, "learning_rate": 1.3546048870425356e-05, "loss": 0.8978, "step": 12112 }, { "epoch": 0.04, "learning_rate": 1.3545041013363463e-05, "loss": 0.9018, "step": 12113 }, { "epoch": 0.04, "learning_rate": 1.3544033115114288e-05, "loss": 0.9311, "step": 12114 }, { "epoch": 0.04, "learning_rate": 1.3543025175689545e-05, "loss": 0.9399, "step": 12115 }, { "epoch": 0.04, "learning_rate": 1.3542017195100948e-05, "loss": 0.9334, "step": 12116 }, { "epoch": 0.04, "learning_rate": 1.35410091733602e-05, "loss": 0.9303, "step": 12117 }, { "epoch": 0.04, "learning_rate": 1.3540001110479023e-05, "loss": 0.9118, "step": 12118 }, { "epoch": 0.04, "learning_rate": 1.3538993006469119e-05, "loss": 0.9276, "step": 12119 }, { "epoch": 0.04, "learning_rate": 1.3537984861342205e-05, "loss": 0.9656, "step": 12120 }, { "epoch": 0.04, "learning_rate": 1.3536976675109994e-05, "loss": 0.887, "step": 12121 }, { "epoch": 0.04, "learning_rate": 1.3535968447784203e-05, "loss": 0.9034, "step": 12122 }, { "epoch": 0.04, "learning_rate": 1.3534960179376537e-05, "loss": 0.9526, "step": 12123 }, { "epoch": 0.04, "learning_rate": 1.3533951869898716e-05, "loss": 0.9418, "step": 12124 }, { "epoch": 0.04, "learning_rate": 1.3532943519362458e-05, "loss": 0.9475, "step": 12125 }, { "epoch": 0.04, "learning_rate": 1.353193512777947e-05, "loss": 0.8762, "step": 12126 }, { "epoch": 0.04, "learning_rate": 1.3530926695161475e-05, "loss": 0.9041, "step": 12127 }, { "epoch": 0.04, "learning_rate": 1.3529918221520185e-05, "loss": 0.8722, "step": 12128 }, { "epoch": 0.04, "learning_rate": 1.3528909706867317e-05, "loss": 0.9205, "step": 12129 }, { "epoch": 0.04, "learning_rate": 1.3527901151214593e-05, "loss": 0.9499, "step": 12130 }, { "epoch": 0.04, "learning_rate": 1.3526892554573723e-05, "loss": 0.9799, "step": 12131 }, { "epoch": 0.04, "learning_rate": 1.352588391695643e-05, "loss": 0.9298, "step": 12132 }, { "epoch": 0.04, "learning_rate": 1.3524875238374433e-05, "loss": 0.9326, "step": 12133 }, { "epoch": 0.04, "learning_rate": 1.352386651883945e-05, "loss": 0.9205, "step": 12134 }, { "epoch": 0.04, "learning_rate": 1.3522857758363198e-05, "loss": 0.9523, "step": 12135 }, { "epoch": 0.04, "learning_rate": 1.3521848956957401e-05, "loss": 0.9065, "step": 12136 }, { "epoch": 0.04, "learning_rate": 1.3520840114633782e-05, "loss": 0.9222, "step": 12137 }, { "epoch": 0.04, "learning_rate": 1.3519831231404055e-05, "loss": 0.9224, "step": 12138 }, { "epoch": 0.04, "learning_rate": 1.3518822307279943e-05, "loss": 0.9378, "step": 12139 }, { "epoch": 0.04, "learning_rate": 1.3517813342273171e-05, "loss": 0.8904, "step": 12140 }, { "epoch": 0.04, "learning_rate": 1.3516804336395459e-05, "loss": 0.8978, "step": 12141 }, { "epoch": 0.04, "learning_rate": 1.351579528965853e-05, "loss": 0.9138, "step": 12142 }, { "epoch": 0.04, "learning_rate": 1.3514786202074116e-05, "loss": 0.9297, "step": 12143 }, { "epoch": 0.04, "learning_rate": 1.3513777073653924e-05, "loss": 0.8893, "step": 12144 }, { "epoch": 0.04, "learning_rate": 1.3512767904409692e-05, "loss": 0.9466, "step": 12145 }, { "epoch": 0.04, "learning_rate": 1.3511758694353141e-05, "loss": 0.9261, "step": 12146 }, { "epoch": 0.04, "learning_rate": 1.3510749443495992e-05, "loss": 0.9149, "step": 12147 }, { "epoch": 0.04, "learning_rate": 1.3509740151849976e-05, "loss": 0.9388, "step": 12148 }, { "epoch": 0.04, "learning_rate": 1.3508730819426816e-05, "loss": 0.9306, "step": 12149 }, { "epoch": 0.04, "learning_rate": 1.3507721446238244e-05, "loss": 0.8977, "step": 12150 }, { "epoch": 0.04, "learning_rate": 1.3506712032295979e-05, "loss": 0.9071, "step": 12151 }, { "epoch": 0.04, "learning_rate": 1.3505702577611759e-05, "loss": 0.9422, "step": 12152 }, { "epoch": 0.04, "learning_rate": 1.3504693082197303e-05, "loss": 0.9228, "step": 12153 }, { "epoch": 0.04, "learning_rate": 1.3503683546064341e-05, "loss": 0.9335, "step": 12154 }, { "epoch": 0.04, "learning_rate": 1.3502673969224607e-05, "loss": 0.9059, "step": 12155 }, { "epoch": 0.04, "learning_rate": 1.3501664351689824e-05, "loss": 0.8875, "step": 12156 }, { "epoch": 0.04, "learning_rate": 1.3500654693471729e-05, "loss": 0.9729, "step": 12157 }, { "epoch": 0.04, "learning_rate": 1.3499644994582047e-05, "loss": 0.9713, "step": 12158 }, { "epoch": 0.04, "learning_rate": 1.3498635255032512e-05, "loss": 0.9251, "step": 12159 }, { "epoch": 0.04, "learning_rate": 1.3497625474834855e-05, "loss": 0.9259, "step": 12160 }, { "epoch": 0.04, "learning_rate": 1.3496615654000805e-05, "loss": 0.931, "step": 12161 }, { "epoch": 0.04, "learning_rate": 1.3495605792542098e-05, "loss": 0.928, "step": 12162 }, { "epoch": 0.04, "learning_rate": 1.3494595890470463e-05, "loss": 0.9191, "step": 12163 }, { "epoch": 0.04, "learning_rate": 1.349358594779764e-05, "loss": 0.9155, "step": 12164 }, { "epoch": 0.04, "learning_rate": 1.3492575964535357e-05, "loss": 0.9161, "step": 12165 }, { "epoch": 0.04, "learning_rate": 1.349156594069535e-05, "loss": 0.9281, "step": 12166 }, { "epoch": 0.04, "learning_rate": 1.3490555876289353e-05, "loss": 0.9047, "step": 12167 }, { "epoch": 0.04, "learning_rate": 1.3489545771329105e-05, "loss": 0.9155, "step": 12168 }, { "epoch": 0.04, "learning_rate": 1.3488535625826336e-05, "loss": 0.895, "step": 12169 }, { "epoch": 0.04, "learning_rate": 1.3487525439792784e-05, "loss": 0.9164, "step": 12170 }, { "epoch": 0.04, "learning_rate": 1.3486515213240188e-05, "loss": 0.89, "step": 12171 }, { "epoch": 0.04, "learning_rate": 1.3485504946180279e-05, "loss": 0.9286, "step": 12172 }, { "epoch": 0.04, "learning_rate": 1.3484494638624803e-05, "loss": 0.9433, "step": 12173 }, { "epoch": 0.04, "learning_rate": 1.3483484290585493e-05, "loss": 0.9355, "step": 12174 }, { "epoch": 0.04, "learning_rate": 1.348247390207409e-05, "loss": 0.9419, "step": 12175 }, { "epoch": 0.04, "learning_rate": 1.3481463473102331e-05, "loss": 0.9209, "step": 12176 }, { "epoch": 0.04, "learning_rate": 1.3480453003681955e-05, "loss": 0.9081, "step": 12177 }, { "epoch": 0.04, "learning_rate": 1.3479442493824704e-05, "loss": 0.929, "step": 12178 }, { "epoch": 0.04, "learning_rate": 1.3478431943542315e-05, "loss": 0.9088, "step": 12179 }, { "epoch": 0.04, "learning_rate": 1.3477421352846534e-05, "loss": 0.9177, "step": 12180 }, { "epoch": 0.04, "learning_rate": 1.3476410721749098e-05, "loss": 0.9642, "step": 12181 }, { "epoch": 0.04, "learning_rate": 1.3475400050261749e-05, "loss": 0.938, "step": 12182 }, { "epoch": 0.04, "learning_rate": 1.3474389338396235e-05, "loss": 0.9009, "step": 12183 }, { "epoch": 0.04, "learning_rate": 1.347337858616429e-05, "loss": 0.9374, "step": 12184 }, { "epoch": 0.04, "learning_rate": 1.3472367793577662e-05, "loss": 0.9605, "step": 12185 }, { "epoch": 0.04, "learning_rate": 1.3471356960648093e-05, "loss": 0.8997, "step": 12186 }, { "epoch": 0.04, "learning_rate": 1.347034608738733e-05, "loss": 0.9151, "step": 12187 }, { "epoch": 0.04, "learning_rate": 1.3469335173807114e-05, "loss": 0.9414, "step": 12188 }, { "epoch": 0.04, "learning_rate": 1.3468324219919196e-05, "loss": 0.9229, "step": 12189 }, { "epoch": 0.04, "learning_rate": 1.3467313225735312e-05, "loss": 0.9876, "step": 12190 }, { "epoch": 0.04, "learning_rate": 1.3466302191267218e-05, "loss": 0.9337, "step": 12191 }, { "epoch": 0.04, "learning_rate": 1.3465291116526655e-05, "loss": 0.935, "step": 12192 }, { "epoch": 0.04, "learning_rate": 1.346428000152537e-05, "loss": 0.9011, "step": 12193 }, { "epoch": 0.04, "learning_rate": 1.3463268846275113e-05, "loss": 0.8786, "step": 12194 }, { "epoch": 0.04, "learning_rate": 1.3462257650787626e-05, "loss": 0.9481, "step": 12195 }, { "epoch": 0.04, "learning_rate": 1.3461246415074667e-05, "loss": 0.8861, "step": 12196 }, { "epoch": 0.04, "learning_rate": 1.3460235139147973e-05, "loss": 0.9389, "step": 12197 }, { "epoch": 0.04, "learning_rate": 1.3459223823019305e-05, "loss": 0.9116, "step": 12198 }, { "epoch": 0.04, "learning_rate": 1.3458212466700406e-05, "loss": 0.8879, "step": 12199 }, { "epoch": 0.04, "learning_rate": 1.3457201070203027e-05, "loss": 0.9059, "step": 12200 }, { "epoch": 0.04, "learning_rate": 1.3456189633538921e-05, "loss": 0.9355, "step": 12201 }, { "epoch": 0.04, "learning_rate": 1.3455178156719838e-05, "loss": 0.9147, "step": 12202 }, { "epoch": 0.04, "learning_rate": 1.3454166639757525e-05, "loss": 0.917, "step": 12203 }, { "epoch": 0.04, "learning_rate": 1.345315508266374e-05, "loss": 0.9019, "step": 12204 }, { "epoch": 0.04, "learning_rate": 1.3452143485450236e-05, "loss": 0.8695, "step": 12205 }, { "epoch": 0.04, "learning_rate": 1.3451131848128763e-05, "loss": 0.9056, "step": 12206 }, { "epoch": 0.04, "learning_rate": 1.3450120170711074e-05, "loss": 0.92, "step": 12207 }, { "epoch": 0.04, "learning_rate": 1.3449108453208926e-05, "loss": 0.926, "step": 12208 }, { "epoch": 0.04, "learning_rate": 1.3448096695634068e-05, "loss": 0.8994, "step": 12209 }, { "epoch": 0.04, "learning_rate": 1.3447084897998259e-05, "loss": 0.8522, "step": 12210 }, { "epoch": 0.04, "learning_rate": 1.3446073060313257e-05, "loss": 0.8655, "step": 12211 }, { "epoch": 0.04, "learning_rate": 1.3445061182590813e-05, "loss": 0.8835, "step": 12212 }, { "epoch": 0.04, "learning_rate": 1.3444049264842684e-05, "loss": 0.9464, "step": 12213 }, { "epoch": 0.04, "learning_rate": 1.344303730708063e-05, "loss": 0.9388, "step": 12214 }, { "epoch": 0.04, "learning_rate": 1.3442025309316405e-05, "loss": 0.9365, "step": 12215 }, { "epoch": 0.04, "learning_rate": 1.3441013271561768e-05, "loss": 0.9031, "step": 12216 }, { "epoch": 0.04, "learning_rate": 1.3440001193828477e-05, "loss": 0.8962, "step": 12217 }, { "epoch": 0.04, "learning_rate": 1.3438989076128287e-05, "loss": 0.8744, "step": 12218 }, { "epoch": 0.04, "learning_rate": 1.3437976918472963e-05, "loss": 0.9429, "step": 12219 }, { "epoch": 0.04, "learning_rate": 1.3436964720874261e-05, "loss": 0.8969, "step": 12220 }, { "epoch": 0.04, "learning_rate": 1.3435952483343942e-05, "loss": 0.9278, "step": 12221 }, { "epoch": 0.04, "learning_rate": 1.3434940205893767e-05, "loss": 0.8549, "step": 12222 }, { "epoch": 0.04, "learning_rate": 1.3433927888535496e-05, "loss": 0.9534, "step": 12223 }, { "epoch": 0.04, "learning_rate": 1.343291553128089e-05, "loss": 0.8944, "step": 12224 }, { "epoch": 0.04, "learning_rate": 1.3431903134141714e-05, "loss": 0.9091, "step": 12225 }, { "epoch": 0.04, "learning_rate": 1.3430890697129727e-05, "loss": 0.9267, "step": 12226 }, { "epoch": 0.04, "learning_rate": 1.342987822025669e-05, "loss": 0.9443, "step": 12227 }, { "epoch": 0.04, "learning_rate": 1.342886570353437e-05, "loss": 0.9189, "step": 12228 }, { "epoch": 0.04, "learning_rate": 1.3427853146974532e-05, "loss": 0.9094, "step": 12229 }, { "epoch": 0.04, "learning_rate": 1.3426840550588933e-05, "loss": 0.9327, "step": 12230 }, { "epoch": 0.04, "learning_rate": 1.3425827914389346e-05, "loss": 0.9029, "step": 12231 }, { "epoch": 0.04, "learning_rate": 1.3424815238387531e-05, "loss": 0.922, "step": 12232 }, { "epoch": 0.04, "learning_rate": 1.3423802522595255e-05, "loss": 0.9221, "step": 12233 }, { "epoch": 0.04, "learning_rate": 1.3422789767024285e-05, "loss": 0.9235, "step": 12234 }, { "epoch": 0.04, "learning_rate": 1.3421776971686387e-05, "loss": 0.9513, "step": 12235 }, { "epoch": 0.04, "learning_rate": 1.3420764136593324e-05, "loss": 0.9026, "step": 12236 }, { "epoch": 0.04, "learning_rate": 1.341975126175687e-05, "loss": 0.9093, "step": 12237 }, { "epoch": 0.04, "learning_rate": 1.3418738347188791e-05, "loss": 0.9254, "step": 12238 }, { "epoch": 0.04, "learning_rate": 1.3417725392900851e-05, "loss": 0.9355, "step": 12239 }, { "epoch": 0.04, "learning_rate": 1.3416712398904821e-05, "loss": 0.9338, "step": 12240 }, { "epoch": 0.04, "learning_rate": 1.3415699365212473e-05, "loss": 0.9241, "step": 12241 }, { "epoch": 0.04, "learning_rate": 1.3414686291835577e-05, "loss": 0.9207, "step": 12242 }, { "epoch": 0.04, "learning_rate": 1.3413673178785893e-05, "loss": 0.9621, "step": 12243 }, { "epoch": 0.04, "learning_rate": 1.3412660026075208e-05, "loss": 0.8962, "step": 12244 }, { "epoch": 0.04, "learning_rate": 1.341164683371528e-05, "loss": 0.9071, "step": 12245 }, { "epoch": 0.04, "learning_rate": 1.3410633601717886e-05, "loss": 0.9407, "step": 12246 }, { "epoch": 0.04, "learning_rate": 1.3409620330094798e-05, "loss": 0.937, "step": 12247 }, { "epoch": 0.04, "learning_rate": 1.3408607018857789e-05, "loss": 0.9235, "step": 12248 }, { "epoch": 0.04, "learning_rate": 1.3407593668018628e-05, "loss": 0.8977, "step": 12249 }, { "epoch": 0.04, "learning_rate": 1.3406580277589092e-05, "loss": 0.904, "step": 12250 }, { "epoch": 0.04, "learning_rate": 1.3405566847580955e-05, "loss": 0.9256, "step": 12251 }, { "epoch": 0.04, "learning_rate": 1.3404553378005988e-05, "loss": 0.8985, "step": 12252 }, { "epoch": 0.04, "learning_rate": 1.340353986887597e-05, "loss": 0.9299, "step": 12253 }, { "epoch": 0.04, "learning_rate": 1.3402526320202676e-05, "loss": 0.9394, "step": 12254 }, { "epoch": 0.04, "learning_rate": 1.3401512731997875e-05, "loss": 0.9337, "step": 12255 }, { "epoch": 0.04, "learning_rate": 1.340049910427335e-05, "loss": 0.913, "step": 12256 }, { "epoch": 0.04, "learning_rate": 1.3399485437040876e-05, "loss": 0.9456, "step": 12257 }, { "epoch": 0.04, "learning_rate": 1.3398471730312228e-05, "loss": 0.8721, "step": 12258 }, { "epoch": 0.04, "learning_rate": 1.3397457984099188e-05, "loss": 0.8793, "step": 12259 }, { "epoch": 0.04, "learning_rate": 1.3396444198413529e-05, "loss": 0.9263, "step": 12260 }, { "epoch": 0.04, "learning_rate": 1.3395430373267033e-05, "loss": 0.9458, "step": 12261 }, { "epoch": 0.04, "learning_rate": 1.3394416508671478e-05, "loss": 0.9113, "step": 12262 }, { "epoch": 0.04, "learning_rate": 1.3393402604638642e-05, "loss": 0.9556, "step": 12263 }, { "epoch": 0.04, "learning_rate": 1.3392388661180303e-05, "loss": 0.9301, "step": 12264 }, { "epoch": 0.04, "learning_rate": 1.3391374678308249e-05, "loss": 0.9466, "step": 12265 }, { "epoch": 0.04, "learning_rate": 1.3390360656034252e-05, "loss": 0.9049, "step": 12266 }, { "epoch": 0.04, "learning_rate": 1.3389346594370098e-05, "loss": 0.9046, "step": 12267 }, { "epoch": 0.04, "learning_rate": 1.338833249332757e-05, "loss": 0.8813, "step": 12268 }, { "epoch": 0.04, "learning_rate": 1.3387318352918447e-05, "loss": 0.9408, "step": 12269 }, { "epoch": 0.04, "learning_rate": 1.3386304173154509e-05, "loss": 0.8896, "step": 12270 }, { "epoch": 0.04, "learning_rate": 1.3385289954047544e-05, "loss": 0.9213, "step": 12271 }, { "epoch": 0.04, "learning_rate": 1.3384275695609335e-05, "loss": 0.8829, "step": 12272 }, { "epoch": 0.04, "learning_rate": 1.3383261397851663e-05, "loss": 0.889, "step": 12273 }, { "epoch": 0.04, "learning_rate": 1.3382247060786316e-05, "loss": 0.9242, "step": 12274 }, { "epoch": 0.04, "learning_rate": 1.3381232684425077e-05, "loss": 0.8722, "step": 12275 }, { "epoch": 0.04, "learning_rate": 1.3380218268779728e-05, "loss": 0.8658, "step": 12276 }, { "epoch": 0.04, "learning_rate": 1.337920381386206e-05, "loss": 0.9367, "step": 12277 }, { "epoch": 0.04, "learning_rate": 1.3378189319683858e-05, "loss": 0.8876, "step": 12278 }, { "epoch": 0.04, "learning_rate": 1.3377174786256909e-05, "loss": 0.9127, "step": 12279 }, { "epoch": 0.04, "learning_rate": 1.3376160213592995e-05, "loss": 0.915, "step": 12280 }, { "epoch": 0.04, "learning_rate": 1.3375145601703911e-05, "loss": 0.9126, "step": 12281 }, { "epoch": 0.04, "learning_rate": 1.3374130950601441e-05, "loss": 0.8989, "step": 12282 }, { "epoch": 0.04, "learning_rate": 1.3373116260297375e-05, "loss": 0.8823, "step": 12283 }, { "epoch": 0.04, "learning_rate": 1.3372101530803501e-05, "loss": 0.9128, "step": 12284 }, { "epoch": 0.04, "learning_rate": 1.3371086762131606e-05, "loss": 0.9413, "step": 12285 }, { "epoch": 0.04, "learning_rate": 1.3370071954293485e-05, "loss": 0.9358, "step": 12286 }, { "epoch": 0.04, "learning_rate": 1.3369057107300925e-05, "loss": 0.9686, "step": 12287 }, { "epoch": 0.04, "learning_rate": 1.3368042221165718e-05, "loss": 0.9166, "step": 12288 }, { "epoch": 0.04, "learning_rate": 1.3367027295899652e-05, "loss": 0.9231, "step": 12289 }, { "epoch": 0.04, "learning_rate": 1.3366012331514525e-05, "loss": 0.9235, "step": 12290 }, { "epoch": 0.04, "learning_rate": 1.3364997328022122e-05, "loss": 0.9638, "step": 12291 }, { "epoch": 0.04, "learning_rate": 1.3363982285434242e-05, "loss": 0.9225, "step": 12292 }, { "epoch": 0.04, "learning_rate": 1.3362967203762676e-05, "loss": 0.8594, "step": 12293 }, { "epoch": 0.04, "learning_rate": 1.3361952083019216e-05, "loss": 0.9114, "step": 12294 }, { "epoch": 0.04, "learning_rate": 1.3360936923215654e-05, "loss": 0.939, "step": 12295 }, { "epoch": 0.04, "learning_rate": 1.335992172436379e-05, "loss": 0.8831, "step": 12296 }, { "epoch": 0.04, "learning_rate": 1.3358906486475418e-05, "loss": 0.8792, "step": 12297 }, { "epoch": 0.04, "learning_rate": 1.3357891209562328e-05, "loss": 0.8812, "step": 12298 }, { "epoch": 0.04, "learning_rate": 1.3356875893636318e-05, "loss": 0.882, "step": 12299 }, { "epoch": 0.04, "learning_rate": 1.335586053870919e-05, "loss": 0.921, "step": 12300 }, { "epoch": 0.04, "learning_rate": 1.335484514479273e-05, "loss": 0.8946, "step": 12301 }, { "epoch": 0.04, "learning_rate": 1.3353829711898744e-05, "loss": 0.895, "step": 12302 }, { "epoch": 0.04, "learning_rate": 1.3352814240039027e-05, "loss": 0.9135, "step": 12303 }, { "epoch": 0.04, "learning_rate": 1.3351798729225377e-05, "loss": 0.9327, "step": 12304 }, { "epoch": 0.04, "learning_rate": 1.335078317946959e-05, "loss": 0.9652, "step": 12305 }, { "epoch": 0.04, "learning_rate": 1.3349767590783472e-05, "loss": 0.899, "step": 12306 }, { "epoch": 0.04, "learning_rate": 1.3348751963178812e-05, "loss": 0.9259, "step": 12307 }, { "epoch": 0.04, "learning_rate": 1.3347736296667418e-05, "loss": 0.8628, "step": 12308 }, { "epoch": 0.04, "learning_rate": 1.3346720591261088e-05, "loss": 0.8972, "step": 12309 }, { "epoch": 0.04, "learning_rate": 1.334570484697162e-05, "loss": 0.8955, "step": 12310 }, { "epoch": 0.04, "learning_rate": 1.334468906381082e-05, "loss": 0.9143, "step": 12311 }, { "epoch": 0.04, "learning_rate": 1.3343673241790485e-05, "loss": 0.907, "step": 12312 }, { "epoch": 0.04, "learning_rate": 1.3342657380922422e-05, "loss": 0.9427, "step": 12313 }, { "epoch": 0.04, "learning_rate": 1.334164148121843e-05, "loss": 0.961, "step": 12314 }, { "epoch": 0.04, "learning_rate": 1.3340625542690314e-05, "loss": 0.9309, "step": 12315 }, { "epoch": 0.04, "learning_rate": 1.3339609565349874e-05, "loss": 0.8669, "step": 12316 }, { "epoch": 0.04, "learning_rate": 1.3338593549208916e-05, "loss": 0.923, "step": 12317 }, { "epoch": 0.04, "learning_rate": 1.3337577494279247e-05, "loss": 0.8817, "step": 12318 }, { "epoch": 0.04, "learning_rate": 1.3336561400572667e-05, "loss": 0.9696, "step": 12319 }, { "epoch": 0.04, "learning_rate": 1.3335545268100987e-05, "loss": 0.8965, "step": 12320 }, { "epoch": 0.04, "learning_rate": 1.333452909687601e-05, "loss": 0.9514, "step": 12321 }, { "epoch": 0.04, "learning_rate": 1.3333512886909537e-05, "loss": 0.9191, "step": 12322 }, { "epoch": 0.04, "learning_rate": 1.3332496638213383e-05, "loss": 0.9348, "step": 12323 }, { "epoch": 0.04, "learning_rate": 1.3331480350799349e-05, "loss": 0.9278, "step": 12324 }, { "epoch": 0.04, "learning_rate": 1.3330464024679246e-05, "loss": 0.9794, "step": 12325 }, { "epoch": 0.04, "learning_rate": 1.332944765986488e-05, "loss": 0.9302, "step": 12326 }, { "epoch": 0.04, "learning_rate": 1.3328431256368061e-05, "loss": 0.9521, "step": 12327 }, { "epoch": 0.04, "learning_rate": 1.3327414814200597e-05, "loss": 0.9594, "step": 12328 }, { "epoch": 0.04, "learning_rate": 1.3326398333374297e-05, "loss": 0.9037, "step": 12329 }, { "epoch": 0.04, "learning_rate": 1.3325381813900973e-05, "loss": 0.9185, "step": 12330 }, { "epoch": 0.04, "learning_rate": 1.332436525579243e-05, "loss": 0.9059, "step": 12331 }, { "epoch": 0.04, "learning_rate": 1.3323348659060483e-05, "loss": 0.9087, "step": 12332 }, { "epoch": 0.04, "learning_rate": 1.3322332023716942e-05, "loss": 0.9275, "step": 12333 }, { "epoch": 0.04, "learning_rate": 1.3321315349773621e-05, "loss": 0.9403, "step": 12334 }, { "epoch": 0.04, "learning_rate": 1.3320298637242325e-05, "loss": 0.9256, "step": 12335 }, { "epoch": 0.04, "learning_rate": 1.3319281886134876e-05, "loss": 0.9182, "step": 12336 }, { "epoch": 0.04, "learning_rate": 1.3318265096463078e-05, "loss": 0.8879, "step": 12337 }, { "epoch": 0.04, "learning_rate": 1.3317248268238751e-05, "loss": 0.9369, "step": 12338 }, { "epoch": 0.04, "learning_rate": 1.3316231401473708e-05, "loss": 0.9457, "step": 12339 }, { "epoch": 0.04, "learning_rate": 1.3315214496179758e-05, "loss": 0.9375, "step": 12340 }, { "epoch": 0.04, "learning_rate": 1.3314197552368717e-05, "loss": 0.9345, "step": 12341 }, { "epoch": 0.04, "learning_rate": 1.3313180570052406e-05, "loss": 0.8782, "step": 12342 }, { "epoch": 0.04, "learning_rate": 1.3312163549242636e-05, "loss": 0.9424, "step": 12343 }, { "epoch": 0.04, "learning_rate": 1.3311146489951224e-05, "loss": 0.8853, "step": 12344 }, { "epoch": 0.04, "learning_rate": 1.3310129392189986e-05, "loss": 0.9391, "step": 12345 }, { "epoch": 0.04, "learning_rate": 1.330911225597074e-05, "loss": 0.9136, "step": 12346 }, { "epoch": 0.04, "learning_rate": 1.3308095081305297e-05, "loss": 0.9354, "step": 12347 }, { "epoch": 0.04, "learning_rate": 1.3307077868205487e-05, "loss": 0.9234, "step": 12348 }, { "epoch": 0.04, "learning_rate": 1.3306060616683118e-05, "loss": 0.9224, "step": 12349 }, { "epoch": 0.04, "learning_rate": 1.3305043326750013e-05, "loss": 0.8947, "step": 12350 }, { "epoch": 0.04, "learning_rate": 1.3304025998417993e-05, "loss": 0.9172, "step": 12351 }, { "epoch": 0.04, "learning_rate": 1.3303008631698874e-05, "loss": 0.9363, "step": 12352 }, { "epoch": 0.05, "learning_rate": 1.3301991226604476e-05, "loss": 0.9294, "step": 12353 }, { "epoch": 0.05, "learning_rate": 1.3300973783146622e-05, "loss": 0.9067, "step": 12354 }, { "epoch": 0.05, "learning_rate": 1.3299956301337132e-05, "loss": 0.9233, "step": 12355 }, { "epoch": 0.05, "learning_rate": 1.3298938781187824e-05, "loss": 0.9235, "step": 12356 }, { "epoch": 0.05, "learning_rate": 1.3297921222710528e-05, "loss": 0.9144, "step": 12357 }, { "epoch": 0.05, "learning_rate": 1.3296903625917058e-05, "loss": 0.8746, "step": 12358 }, { "epoch": 0.05, "learning_rate": 1.3295885990819238e-05, "loss": 0.9367, "step": 12359 }, { "epoch": 0.05, "learning_rate": 1.3294868317428895e-05, "loss": 0.9337, "step": 12360 }, { "epoch": 0.05, "learning_rate": 1.3293850605757852e-05, "loss": 0.9522, "step": 12361 }, { "epoch": 0.05, "learning_rate": 1.329283285581793e-05, "loss": 0.9302, "step": 12362 }, { "epoch": 0.05, "learning_rate": 1.3291815067620954e-05, "loss": 0.9116, "step": 12363 }, { "epoch": 0.05, "learning_rate": 1.3290797241178752e-05, "loss": 0.9262, "step": 12364 }, { "epoch": 0.05, "learning_rate": 1.3289779376503147e-05, "loss": 0.9195, "step": 12365 }, { "epoch": 0.05, "learning_rate": 1.3288761473605967e-05, "loss": 0.8909, "step": 12366 }, { "epoch": 0.05, "learning_rate": 1.3287743532499033e-05, "loss": 0.9441, "step": 12367 }, { "epoch": 0.05, "learning_rate": 1.3286725553194178e-05, "loss": 0.8942, "step": 12368 }, { "epoch": 0.05, "learning_rate": 1.3285707535703228e-05, "loss": 0.9019, "step": 12369 }, { "epoch": 0.05, "learning_rate": 1.3284689480038006e-05, "loss": 0.8874, "step": 12370 }, { "epoch": 0.05, "learning_rate": 1.3283671386210344e-05, "loss": 0.8886, "step": 12371 }, { "epoch": 0.05, "learning_rate": 1.3282653254232073e-05, "loss": 0.8887, "step": 12372 }, { "epoch": 0.05, "learning_rate": 1.3281635084115014e-05, "loss": 0.8947, "step": 12373 }, { "epoch": 0.05, "learning_rate": 1.3280616875871004e-05, "loss": 0.933, "step": 12374 }, { "epoch": 0.05, "learning_rate": 1.3279598629511868e-05, "loss": 0.9172, "step": 12375 }, { "epoch": 0.05, "learning_rate": 1.3278580345049441e-05, "loss": 0.9158, "step": 12376 }, { "epoch": 0.05, "learning_rate": 1.3277562022495548e-05, "loss": 0.9317, "step": 12377 }, { "epoch": 0.05, "learning_rate": 1.3276543661862023e-05, "loss": 0.9022, "step": 12378 }, { "epoch": 0.05, "learning_rate": 1.3275525263160697e-05, "loss": 0.9101, "step": 12379 }, { "epoch": 0.05, "learning_rate": 1.3274506826403407e-05, "loss": 0.8947, "step": 12380 }, { "epoch": 0.05, "learning_rate": 1.3273488351601975e-05, "loss": 0.9422, "step": 12381 }, { "epoch": 0.05, "learning_rate": 1.3272469838768244e-05, "loss": 0.9601, "step": 12382 }, { "epoch": 0.05, "learning_rate": 1.3271451287914043e-05, "loss": 0.9205, "step": 12383 }, { "epoch": 0.05, "learning_rate": 1.3270432699051206e-05, "loss": 0.9285, "step": 12384 }, { "epoch": 0.05, "learning_rate": 1.3269414072191568e-05, "loss": 0.9266, "step": 12385 }, { "epoch": 0.05, "learning_rate": 1.3268395407346963e-05, "loss": 0.9471, "step": 12386 }, { "epoch": 0.05, "learning_rate": 1.3267376704529227e-05, "loss": 0.9229, "step": 12387 }, { "epoch": 0.05, "learning_rate": 1.3266357963750192e-05, "loss": 0.8885, "step": 12388 }, { "epoch": 0.05, "learning_rate": 1.32653391850217e-05, "loss": 0.9005, "step": 12389 }, { "epoch": 0.05, "learning_rate": 1.3264320368355581e-05, "loss": 0.9124, "step": 12390 }, { "epoch": 0.05, "learning_rate": 1.326330151376368e-05, "loss": 0.9103, "step": 12391 }, { "epoch": 0.05, "learning_rate": 1.3262282621257827e-05, "loss": 0.9157, "step": 12392 }, { "epoch": 0.05, "learning_rate": 1.326126369084986e-05, "loss": 0.928, "step": 12393 }, { "epoch": 0.05, "learning_rate": 1.3260244722551623e-05, "loss": 0.9247, "step": 12394 }, { "epoch": 0.05, "learning_rate": 1.325922571637495e-05, "loss": 0.9302, "step": 12395 }, { "epoch": 0.05, "learning_rate": 1.3258206672331681e-05, "loss": 0.9211, "step": 12396 }, { "epoch": 0.05, "learning_rate": 1.3257187590433655e-05, "loss": 0.919, "step": 12397 }, { "epoch": 0.05, "learning_rate": 1.3256168470692717e-05, "loss": 0.9443, "step": 12398 }, { "epoch": 0.05, "learning_rate": 1.3255149313120698e-05, "loss": 0.912, "step": 12399 }, { "epoch": 0.05, "learning_rate": 1.3254130117729448e-05, "loss": 0.9033, "step": 12400 }, { "epoch": 0.05, "learning_rate": 1.3253110884530804e-05, "loss": 0.8921, "step": 12401 }, { "epoch": 0.05, "learning_rate": 1.3252091613536604e-05, "loss": 0.8928, "step": 12402 }, { "epoch": 0.05, "learning_rate": 1.3251072304758701e-05, "loss": 0.9303, "step": 12403 }, { "epoch": 0.05, "learning_rate": 1.3250052958208926e-05, "loss": 0.9343, "step": 12404 }, { "epoch": 0.05, "learning_rate": 1.3249033573899128e-05, "loss": 0.9316, "step": 12405 }, { "epoch": 0.05, "learning_rate": 1.3248014151841151e-05, "loss": 0.9361, "step": 12406 }, { "epoch": 0.05, "learning_rate": 1.3246994692046837e-05, "loss": 0.9474, "step": 12407 }, { "epoch": 0.05, "learning_rate": 1.324597519452803e-05, "loss": 0.8993, "step": 12408 }, { "epoch": 0.05, "learning_rate": 1.3244955659296575e-05, "loss": 0.9164, "step": 12409 }, { "epoch": 0.05, "learning_rate": 1.324393608636432e-05, "loss": 0.881, "step": 12410 }, { "epoch": 0.05, "learning_rate": 1.3242916475743104e-05, "loss": 0.962, "step": 12411 }, { "epoch": 0.05, "learning_rate": 1.3241896827444785e-05, "loss": 0.8709, "step": 12412 }, { "epoch": 0.05, "learning_rate": 1.3240877141481197e-05, "loss": 0.9079, "step": 12413 }, { "epoch": 0.05, "learning_rate": 1.3239857417864193e-05, "loss": 0.9191, "step": 12414 }, { "epoch": 0.05, "learning_rate": 1.3238837656605621e-05, "loss": 0.9299, "step": 12415 }, { "epoch": 0.05, "learning_rate": 1.3237817857717326e-05, "loss": 0.9155, "step": 12416 }, { "epoch": 0.05, "learning_rate": 1.3236798021211156e-05, "loss": 0.9045, "step": 12417 }, { "epoch": 0.05, "learning_rate": 1.3235778147098965e-05, "loss": 0.9252, "step": 12418 }, { "epoch": 0.05, "learning_rate": 1.32347582353926e-05, "loss": 0.9619, "step": 12419 }, { "epoch": 0.05, "learning_rate": 1.3233738286103905e-05, "loss": 0.9149, "step": 12420 }, { "epoch": 0.05, "learning_rate": 1.3232718299244738e-05, "loss": 0.8989, "step": 12421 }, { "epoch": 0.05, "learning_rate": 1.3231698274826945e-05, "loss": 0.9097, "step": 12422 }, { "epoch": 0.05, "learning_rate": 1.3230678212862375e-05, "loss": 0.8981, "step": 12423 }, { "epoch": 0.05, "learning_rate": 1.3229658113362886e-05, "loss": 0.9246, "step": 12424 }, { "epoch": 0.05, "learning_rate": 1.3228637976340325e-05, "loss": 0.94, "step": 12425 }, { "epoch": 0.05, "learning_rate": 1.3227617801806543e-05, "loss": 0.8835, "step": 12426 }, { "epoch": 0.05, "learning_rate": 1.3226597589773396e-05, "loss": 0.911, "step": 12427 }, { "epoch": 0.05, "learning_rate": 1.322557734025274e-05, "loss": 0.9308, "step": 12428 }, { "epoch": 0.05, "learning_rate": 1.3224557053256422e-05, "loss": 0.9014, "step": 12429 }, { "epoch": 0.05, "learning_rate": 1.3223536728796296e-05, "loss": 0.9583, "step": 12430 }, { "epoch": 0.05, "learning_rate": 1.3222516366884221e-05, "loss": 0.9211, "step": 12431 }, { "epoch": 0.05, "learning_rate": 1.322149596753205e-05, "loss": 0.8973, "step": 12432 }, { "epoch": 0.05, "learning_rate": 1.3220475530751637e-05, "loss": 0.9081, "step": 12433 }, { "epoch": 0.05, "learning_rate": 1.321945505655484e-05, "loss": 0.8675, "step": 12434 }, { "epoch": 0.05, "learning_rate": 1.3218434544953515e-05, "loss": 0.9354, "step": 12435 }, { "epoch": 0.05, "learning_rate": 1.3217413995959514e-05, "loss": 0.9349, "step": 12436 }, { "epoch": 0.05, "learning_rate": 1.3216393409584703e-05, "loss": 0.9184, "step": 12437 }, { "epoch": 0.05, "learning_rate": 1.321537278584093e-05, "loss": 0.8889, "step": 12438 }, { "epoch": 0.05, "learning_rate": 1.3214352124740057e-05, "loss": 0.9179, "step": 12439 }, { "epoch": 0.05, "learning_rate": 1.3213331426293945e-05, "loss": 0.9165, "step": 12440 }, { "epoch": 0.05, "learning_rate": 1.3212310690514446e-05, "loss": 0.9324, "step": 12441 }, { "epoch": 0.05, "learning_rate": 1.3211289917413426e-05, "loss": 0.9293, "step": 12442 }, { "epoch": 0.05, "learning_rate": 1.321026910700274e-05, "loss": 0.9044, "step": 12443 }, { "epoch": 0.05, "learning_rate": 1.3209248259294255e-05, "loss": 0.8986, "step": 12444 }, { "epoch": 0.05, "learning_rate": 1.320822737429982e-05, "loss": 0.9587, "step": 12445 }, { "epoch": 0.05, "learning_rate": 1.3207206452031305e-05, "loss": 0.917, "step": 12446 }, { "epoch": 0.05, "learning_rate": 1.3206185492500571e-05, "loss": 0.8979, "step": 12447 }, { "epoch": 0.05, "learning_rate": 1.3205164495719471e-05, "loss": 0.9273, "step": 12448 }, { "epoch": 0.05, "learning_rate": 1.320414346169988e-05, "loss": 0.8898, "step": 12449 }, { "epoch": 0.05, "learning_rate": 1.3203122390453653e-05, "loss": 0.9001, "step": 12450 }, { "epoch": 0.05, "learning_rate": 1.3202101281992652e-05, "loss": 0.9352, "step": 12451 }, { "epoch": 0.05, "learning_rate": 1.3201080136328748e-05, "loss": 0.9132, "step": 12452 }, { "epoch": 0.05, "learning_rate": 1.3200058953473795e-05, "loss": 0.8966, "step": 12453 }, { "epoch": 0.05, "learning_rate": 1.3199037733439664e-05, "loss": 0.9034, "step": 12454 }, { "epoch": 0.05, "learning_rate": 1.3198016476238218e-05, "loss": 0.922, "step": 12455 }, { "epoch": 0.05, "learning_rate": 1.3196995181881325e-05, "loss": 0.9635, "step": 12456 }, { "epoch": 0.05, "learning_rate": 1.3195973850380841e-05, "loss": 0.8911, "step": 12457 }, { "epoch": 0.05, "learning_rate": 1.3194952481748646e-05, "loss": 0.9225, "step": 12458 }, { "epoch": 0.05, "learning_rate": 1.3193931075996598e-05, "loss": 0.9217, "step": 12459 }, { "epoch": 0.05, "learning_rate": 1.3192909633136564e-05, "loss": 0.9091, "step": 12460 }, { "epoch": 0.05, "learning_rate": 1.3191888153180417e-05, "loss": 0.9028, "step": 12461 }, { "epoch": 0.05, "learning_rate": 1.3190866636140017e-05, "loss": 0.9453, "step": 12462 }, { "epoch": 0.05, "learning_rate": 1.3189845082027237e-05, "loss": 0.9374, "step": 12463 }, { "epoch": 0.05, "learning_rate": 1.3188823490853946e-05, "loss": 0.8951, "step": 12464 }, { "epoch": 0.05, "learning_rate": 1.3187801862632013e-05, "loss": 0.9184, "step": 12465 }, { "epoch": 0.05, "learning_rate": 1.3186780197373306e-05, "loss": 0.8956, "step": 12466 }, { "epoch": 0.05, "learning_rate": 1.3185758495089694e-05, "loss": 0.8891, "step": 12467 }, { "epoch": 0.05, "learning_rate": 1.3184736755793051e-05, "loss": 0.9085, "step": 12468 }, { "epoch": 0.05, "learning_rate": 1.3183714979495243e-05, "loss": 0.8978, "step": 12469 }, { "epoch": 0.05, "learning_rate": 1.3182693166208146e-05, "loss": 0.9008, "step": 12470 }, { "epoch": 0.05, "learning_rate": 1.3181671315943632e-05, "loss": 0.9092, "step": 12471 }, { "epoch": 0.05, "learning_rate": 1.3180649428713567e-05, "loss": 0.8823, "step": 12472 }, { "epoch": 0.05, "learning_rate": 1.317962750452983e-05, "loss": 0.9133, "step": 12473 }, { "epoch": 0.05, "learning_rate": 1.3178605543404292e-05, "loss": 0.9201, "step": 12474 }, { "epoch": 0.05, "learning_rate": 1.3177583545348827e-05, "loss": 0.9768, "step": 12475 }, { "epoch": 0.05, "learning_rate": 1.3176561510375305e-05, "loss": 0.9761, "step": 12476 }, { "epoch": 0.05, "learning_rate": 1.3175539438495605e-05, "loss": 0.9121, "step": 12477 }, { "epoch": 0.05, "learning_rate": 1.31745173297216e-05, "loss": 0.9005, "step": 12478 }, { "epoch": 0.05, "learning_rate": 1.3173495184065165e-05, "loss": 0.9136, "step": 12479 }, { "epoch": 0.05, "learning_rate": 1.3172473001538175e-05, "loss": 0.9376, "step": 12480 }, { "epoch": 0.05, "learning_rate": 1.3171450782152508e-05, "loss": 0.8827, "step": 12481 }, { "epoch": 0.05, "learning_rate": 1.3170428525920036e-05, "loss": 0.9154, "step": 12482 }, { "epoch": 0.05, "learning_rate": 1.3169406232852641e-05, "loss": 0.8884, "step": 12483 }, { "epoch": 0.05, "learning_rate": 1.3168383902962198e-05, "loss": 0.9433, "step": 12484 }, { "epoch": 0.05, "learning_rate": 1.3167361536260586e-05, "loss": 0.9277, "step": 12485 }, { "epoch": 0.05, "learning_rate": 1.3166339132759682e-05, "loss": 0.8849, "step": 12486 }, { "epoch": 0.05, "learning_rate": 1.3165316692471364e-05, "loss": 0.9315, "step": 12487 }, { "epoch": 0.05, "learning_rate": 1.3164294215407511e-05, "loss": 0.8964, "step": 12488 }, { "epoch": 0.05, "learning_rate": 1.3163271701580003e-05, "loss": 0.904, "step": 12489 }, { "epoch": 0.05, "learning_rate": 1.3162249151000721e-05, "loss": 0.9164, "step": 12490 }, { "epoch": 0.05, "learning_rate": 1.3161226563681544e-05, "loss": 0.9311, "step": 12491 }, { "epoch": 0.05, "learning_rate": 1.3160203939634351e-05, "loss": 0.904, "step": 12492 }, { "epoch": 0.05, "learning_rate": 1.3159181278871028e-05, "loss": 0.9174, "step": 12493 }, { "epoch": 0.05, "learning_rate": 1.3158158581403448e-05, "loss": 0.8878, "step": 12494 }, { "epoch": 0.05, "learning_rate": 1.3157135847243505e-05, "loss": 0.8722, "step": 12495 }, { "epoch": 0.05, "learning_rate": 1.3156113076403071e-05, "loss": 0.9262, "step": 12496 }, { "epoch": 0.05, "learning_rate": 1.3155090268894038e-05, "loss": 0.9663, "step": 12497 }, { "epoch": 0.05, "learning_rate": 1.3154067424728281e-05, "loss": 0.8913, "step": 12498 }, { "epoch": 0.05, "learning_rate": 1.3153044543917685e-05, "loss": 0.8942, "step": 12499 }, { "epoch": 0.05, "learning_rate": 1.3152021626474136e-05, "loss": 0.9406, "step": 12500 }, { "epoch": 0.05, "learning_rate": 1.315099867240952e-05, "loss": 0.9479, "step": 12501 }, { "epoch": 0.05, "learning_rate": 1.3149975681735722e-05, "loss": 0.9406, "step": 12502 }, { "epoch": 0.05, "learning_rate": 1.3148952654464622e-05, "loss": 0.922, "step": 12503 }, { "epoch": 0.05, "learning_rate": 1.3147929590608113e-05, "loss": 0.925, "step": 12504 }, { "epoch": 0.05, "learning_rate": 1.3146906490178076e-05, "loss": 0.9196, "step": 12505 }, { "epoch": 0.05, "learning_rate": 1.3145883353186399e-05, "loss": 0.9076, "step": 12506 }, { "epoch": 0.05, "learning_rate": 1.3144860179644972e-05, "loss": 0.8903, "step": 12507 }, { "epoch": 0.05, "learning_rate": 1.3143836969565678e-05, "loss": 0.9083, "step": 12508 }, { "epoch": 0.05, "learning_rate": 1.3142813722960406e-05, "loss": 0.9635, "step": 12509 }, { "epoch": 0.05, "learning_rate": 1.3141790439841046e-05, "loss": 0.9134, "step": 12510 }, { "epoch": 0.05, "learning_rate": 1.314076712021949e-05, "loss": 0.9205, "step": 12511 }, { "epoch": 0.05, "learning_rate": 1.3139743764107617e-05, "loss": 0.9172, "step": 12512 }, { "epoch": 0.05, "learning_rate": 1.3138720371517331e-05, "loss": 0.9406, "step": 12513 }, { "epoch": 0.05, "learning_rate": 1.3137696942460511e-05, "loss": 0.8766, "step": 12514 }, { "epoch": 0.05, "learning_rate": 1.3136673476949047e-05, "loss": 0.9495, "step": 12515 }, { "epoch": 0.05, "learning_rate": 1.3135649974994837e-05, "loss": 0.9458, "step": 12516 }, { "epoch": 0.05, "learning_rate": 1.3134626436609767e-05, "loss": 0.9493, "step": 12517 }, { "epoch": 0.05, "learning_rate": 1.3133602861805732e-05, "loss": 0.9283, "step": 12518 }, { "epoch": 0.05, "learning_rate": 1.313257925059462e-05, "loss": 0.8936, "step": 12519 }, { "epoch": 0.05, "learning_rate": 1.3131555602988334e-05, "loss": 0.9073, "step": 12520 }, { "epoch": 0.05, "learning_rate": 1.3130531918998752e-05, "loss": 0.9254, "step": 12521 }, { "epoch": 0.05, "learning_rate": 1.3129508198637777e-05, "loss": 0.9231, "step": 12522 }, { "epoch": 0.05, "learning_rate": 1.3128484441917301e-05, "loss": 0.9483, "step": 12523 }, { "epoch": 0.05, "learning_rate": 1.3127460648849218e-05, "loss": 0.892, "step": 12524 }, { "epoch": 0.05, "learning_rate": 1.3126436819445423e-05, "loss": 0.9563, "step": 12525 }, { "epoch": 0.05, "learning_rate": 1.3125412953717808e-05, "loss": 0.8911, "step": 12526 }, { "epoch": 0.05, "learning_rate": 1.3124389051678274e-05, "loss": 0.905, "step": 12527 }, { "epoch": 0.05, "learning_rate": 1.3123365113338715e-05, "loss": 0.9255, "step": 12528 }, { "epoch": 0.05, "learning_rate": 1.3122341138711025e-05, "loss": 0.8758, "step": 12529 }, { "epoch": 0.05, "learning_rate": 1.3121317127807105e-05, "loss": 0.8838, "step": 12530 }, { "epoch": 0.05, "learning_rate": 1.3120293080638849e-05, "loss": 0.9061, "step": 12531 }, { "epoch": 0.05, "learning_rate": 1.3119268997218156e-05, "loss": 0.9066, "step": 12532 }, { "epoch": 0.05, "learning_rate": 1.3118244877556923e-05, "loss": 0.8995, "step": 12533 }, { "epoch": 0.05, "learning_rate": 1.3117220721667048e-05, "loss": 0.9459, "step": 12534 }, { "epoch": 0.05, "learning_rate": 1.3116196529560432e-05, "loss": 0.931, "step": 12535 }, { "epoch": 0.05, "learning_rate": 1.3115172301248975e-05, "loss": 0.9092, "step": 12536 }, { "epoch": 0.05, "learning_rate": 1.3114148036744574e-05, "loss": 0.8878, "step": 12537 }, { "epoch": 0.05, "learning_rate": 1.3113123736059128e-05, "loss": 0.8769, "step": 12538 }, { "epoch": 0.05, "learning_rate": 1.3112099399204543e-05, "loss": 0.91, "step": 12539 }, { "epoch": 0.05, "learning_rate": 1.3111075026192714e-05, "loss": 0.8903, "step": 12540 }, { "epoch": 0.05, "learning_rate": 1.311005061703555e-05, "loss": 0.902, "step": 12541 }, { "epoch": 0.05, "learning_rate": 1.3109026171744945e-05, "loss": 0.9216, "step": 12542 }, { "epoch": 0.05, "learning_rate": 1.3108001690332806e-05, "loss": 0.9462, "step": 12543 }, { "epoch": 0.05, "learning_rate": 1.3106977172811036e-05, "loss": 0.9049, "step": 12544 }, { "epoch": 0.05, "learning_rate": 1.3105952619191533e-05, "loss": 0.9299, "step": 12545 }, { "epoch": 0.05, "learning_rate": 1.3104928029486203e-05, "loss": 0.9252, "step": 12546 }, { "epoch": 0.05, "learning_rate": 1.3103903403706955e-05, "loss": 0.9378, "step": 12547 }, { "epoch": 0.05, "learning_rate": 1.3102878741865688e-05, "loss": 0.9692, "step": 12548 }, { "epoch": 0.05, "learning_rate": 1.3101854043974307e-05, "loss": 0.9561, "step": 12549 }, { "epoch": 0.05, "learning_rate": 1.310082931004472e-05, "loss": 0.8775, "step": 12550 }, { "epoch": 0.05, "learning_rate": 1.309980454008883e-05, "loss": 0.9059, "step": 12551 }, { "epoch": 0.05, "learning_rate": 1.3098779734118546e-05, "loss": 0.9165, "step": 12552 }, { "epoch": 0.05, "learning_rate": 1.309775489214577e-05, "loss": 0.9356, "step": 12553 }, { "epoch": 0.05, "learning_rate": 1.3096730014182415e-05, "loss": 0.9393, "step": 12554 }, { "epoch": 0.05, "learning_rate": 1.309570510024038e-05, "loss": 0.914, "step": 12555 }, { "epoch": 0.05, "learning_rate": 1.3094680150331582e-05, "loss": 0.9625, "step": 12556 }, { "epoch": 0.05, "learning_rate": 1.3093655164467925e-05, "loss": 0.9528, "step": 12557 }, { "epoch": 0.05, "learning_rate": 1.3092630142661312e-05, "loss": 0.9072, "step": 12558 }, { "epoch": 0.05, "learning_rate": 1.3091605084923663e-05, "loss": 0.9067, "step": 12559 }, { "epoch": 0.05, "learning_rate": 1.309057999126688e-05, "loss": 0.9333, "step": 12560 }, { "epoch": 0.05, "learning_rate": 1.3089554861702872e-05, "loss": 0.8903, "step": 12561 }, { "epoch": 0.05, "learning_rate": 1.3088529696243555e-05, "loss": 0.8648, "step": 12562 }, { "epoch": 0.05, "learning_rate": 1.3087504494900833e-05, "loss": 0.8725, "step": 12563 }, { "epoch": 0.05, "learning_rate": 1.3086479257686623e-05, "loss": 0.9309, "step": 12564 }, { "epoch": 0.05, "learning_rate": 1.3085453984612833e-05, "loss": 0.9348, "step": 12565 }, { "epoch": 0.05, "learning_rate": 1.3084428675691378e-05, "loss": 0.9305, "step": 12566 }, { "epoch": 0.05, "learning_rate": 1.3083403330934167e-05, "loss": 0.9341, "step": 12567 }, { "epoch": 0.05, "learning_rate": 1.3082377950353114e-05, "loss": 0.893, "step": 12568 }, { "epoch": 0.05, "learning_rate": 1.3081352533960132e-05, "loss": 0.9398, "step": 12569 }, { "epoch": 0.05, "learning_rate": 1.3080327081767134e-05, "loss": 0.9297, "step": 12570 }, { "epoch": 0.05, "learning_rate": 1.3079301593786035e-05, "loss": 0.8973, "step": 12571 }, { "epoch": 0.05, "learning_rate": 1.3078276070028748e-05, "loss": 0.9597, "step": 12572 }, { "epoch": 0.05, "learning_rate": 1.3077250510507192e-05, "loss": 0.9403, "step": 12573 }, { "epoch": 0.05, "learning_rate": 1.3076224915233275e-05, "loss": 0.9007, "step": 12574 }, { "epoch": 0.05, "learning_rate": 1.3075199284218923e-05, "loss": 0.896, "step": 12575 }, { "epoch": 0.05, "learning_rate": 1.307417361747604e-05, "loss": 0.9058, "step": 12576 }, { "epoch": 0.05, "learning_rate": 1.307314791501655e-05, "loss": 0.9551, "step": 12577 }, { "epoch": 0.05, "learning_rate": 1.3072122176852372e-05, "loss": 0.929, "step": 12578 }, { "epoch": 0.05, "learning_rate": 1.3071096402995415e-05, "loss": 0.9407, "step": 12579 }, { "epoch": 0.05, "learning_rate": 1.3070070593457602e-05, "loss": 0.9126, "step": 12580 }, { "epoch": 0.05, "learning_rate": 1.3069044748250852e-05, "loss": 0.8939, "step": 12581 }, { "epoch": 0.05, "learning_rate": 1.3068018867387083e-05, "loss": 0.9026, "step": 12582 }, { "epoch": 0.05, "learning_rate": 1.306699295087821e-05, "loss": 0.9506, "step": 12583 }, { "epoch": 0.05, "learning_rate": 1.3065966998736155e-05, "loss": 0.9104, "step": 12584 }, { "epoch": 0.05, "learning_rate": 1.3064941010972842e-05, "loss": 0.9051, "step": 12585 }, { "epoch": 0.05, "learning_rate": 1.3063914987600183e-05, "loss": 0.9224, "step": 12586 }, { "epoch": 0.05, "learning_rate": 1.3062888928630107e-05, "loss": 0.9257, "step": 12587 }, { "epoch": 0.05, "learning_rate": 1.3061862834074528e-05, "loss": 0.885, "step": 12588 }, { "epoch": 0.05, "learning_rate": 1.306083670394537e-05, "loss": 0.9431, "step": 12589 }, { "epoch": 0.05, "learning_rate": 1.3059810538254558e-05, "loss": 0.8822, "step": 12590 }, { "epoch": 0.05, "learning_rate": 1.3058784337014012e-05, "loss": 0.908, "step": 12591 }, { "epoch": 0.05, "learning_rate": 1.3057758100235652e-05, "loss": 0.9029, "step": 12592 }, { "epoch": 0.05, "learning_rate": 1.3056731827931404e-05, "loss": 0.9281, "step": 12593 }, { "epoch": 0.05, "learning_rate": 1.3055705520113193e-05, "loss": 0.9146, "step": 12594 }, { "epoch": 0.05, "learning_rate": 1.3054679176792937e-05, "loss": 0.9537, "step": 12595 }, { "epoch": 0.05, "learning_rate": 1.3053652797982567e-05, "loss": 0.9546, "step": 12596 }, { "epoch": 0.05, "learning_rate": 1.3052626383694004e-05, "loss": 0.9294, "step": 12597 }, { "epoch": 0.05, "learning_rate": 1.3051599933939174e-05, "loss": 0.8893, "step": 12598 }, { "epoch": 0.05, "learning_rate": 1.3050573448730006e-05, "loss": 0.8798, "step": 12599 }, { "epoch": 0.05, "learning_rate": 1.3049546928078418e-05, "loss": 0.9435, "step": 12600 }, { "epoch": 0.05, "learning_rate": 1.3048520371996342e-05, "loss": 0.9434, "step": 12601 }, { "epoch": 0.05, "learning_rate": 1.3047493780495706e-05, "loss": 0.8998, "step": 12602 }, { "epoch": 0.05, "learning_rate": 1.3046467153588436e-05, "loss": 0.9577, "step": 12603 }, { "epoch": 0.05, "learning_rate": 1.3045440491286455e-05, "loss": 0.9459, "step": 12604 }, { "epoch": 0.05, "learning_rate": 1.3044413793601698e-05, "loss": 0.8885, "step": 12605 }, { "epoch": 0.05, "learning_rate": 1.3043387060546092e-05, "loss": 0.918, "step": 12606 }, { "epoch": 0.05, "learning_rate": 1.3042360292131558e-05, "loss": 0.9092, "step": 12607 }, { "epoch": 0.05, "learning_rate": 1.3041333488370036e-05, "loss": 0.8978, "step": 12608 }, { "epoch": 0.05, "learning_rate": 1.304030664927345e-05, "loss": 0.922, "step": 12609 }, { "epoch": 0.05, "learning_rate": 1.303927977485373e-05, "loss": 0.9195, "step": 12610 }, { "epoch": 0.05, "learning_rate": 1.3038252865122809e-05, "loss": 0.894, "step": 12611 }, { "epoch": 0.05, "learning_rate": 1.303722592009262e-05, "loss": 0.9277, "step": 12612 }, { "epoch": 0.05, "learning_rate": 1.3036198939775086e-05, "loss": 0.9075, "step": 12613 }, { "epoch": 0.05, "learning_rate": 1.3035171924182145e-05, "loss": 0.8901, "step": 12614 }, { "epoch": 0.05, "learning_rate": 1.303414487332573e-05, "loss": 0.8988, "step": 12615 }, { "epoch": 0.05, "learning_rate": 1.3033117787217768e-05, "loss": 0.9105, "step": 12616 }, { "epoch": 0.05, "learning_rate": 1.30320906658702e-05, "loss": 0.9262, "step": 12617 }, { "epoch": 0.05, "learning_rate": 1.3031063509294951e-05, "loss": 0.8855, "step": 12618 }, { "epoch": 0.05, "learning_rate": 1.303003631750396e-05, "loss": 0.9247, "step": 12619 }, { "epoch": 0.05, "learning_rate": 1.3029009090509162e-05, "loss": 0.897, "step": 12620 }, { "epoch": 0.05, "learning_rate": 1.3027981828322487e-05, "loss": 0.9068, "step": 12621 }, { "epoch": 0.05, "learning_rate": 1.3026954530955872e-05, "loss": 0.9038, "step": 12622 }, { "epoch": 0.05, "learning_rate": 1.3025927198421255e-05, "loss": 0.9239, "step": 12623 }, { "epoch": 0.05, "learning_rate": 1.302489983073057e-05, "loss": 0.9102, "step": 12624 }, { "epoch": 0.05, "learning_rate": 1.3023872427895752e-05, "loss": 0.9294, "step": 12625 }, { "epoch": 0.05, "learning_rate": 1.3022844989928737e-05, "loss": 0.9622, "step": 12626 }, { "epoch": 0.05, "learning_rate": 1.3021817516841465e-05, "loss": 0.9131, "step": 12627 }, { "epoch": 0.05, "learning_rate": 1.3020790008645874e-05, "loss": 0.9018, "step": 12628 }, { "epoch": 0.05, "learning_rate": 1.3019762465353897e-05, "loss": 0.9331, "step": 12629 }, { "epoch": 0.05, "learning_rate": 1.3018734886977477e-05, "loss": 0.9104, "step": 12630 }, { "epoch": 0.05, "learning_rate": 1.301770727352855e-05, "loss": 0.9543, "step": 12631 }, { "epoch": 0.05, "learning_rate": 1.3016679625019057e-05, "loss": 0.9006, "step": 12632 }, { "epoch": 0.05, "learning_rate": 1.3015651941460938e-05, "loss": 0.9112, "step": 12633 }, { "epoch": 0.05, "learning_rate": 1.301462422286613e-05, "loss": 0.9004, "step": 12634 }, { "epoch": 0.05, "learning_rate": 1.3013596469246574e-05, "loss": 0.9007, "step": 12635 }, { "epoch": 0.05, "learning_rate": 1.3012568680614215e-05, "loss": 0.9468, "step": 12636 }, { "epoch": 0.05, "learning_rate": 1.3011540856980987e-05, "loss": 0.9327, "step": 12637 }, { "epoch": 0.05, "learning_rate": 1.3010512998358835e-05, "loss": 0.9207, "step": 12638 }, { "epoch": 0.05, "learning_rate": 1.3009485104759705e-05, "loss": 0.9383, "step": 12639 }, { "epoch": 0.05, "learning_rate": 1.3008457176195535e-05, "loss": 0.9134, "step": 12640 }, { "epoch": 0.05, "learning_rate": 1.3007429212678261e-05, "loss": 0.8873, "step": 12641 }, { "epoch": 0.05, "learning_rate": 1.3006401214219844e-05, "loss": 0.9085, "step": 12642 }, { "epoch": 0.05, "learning_rate": 1.300537318083221e-05, "loss": 0.9076, "step": 12643 }, { "epoch": 0.05, "learning_rate": 1.3004345112527312e-05, "loss": 0.8767, "step": 12644 }, { "epoch": 0.05, "learning_rate": 1.3003317009317096e-05, "loss": 0.9046, "step": 12645 }, { "epoch": 0.05, "learning_rate": 1.30022888712135e-05, "loss": 0.9091, "step": 12646 }, { "epoch": 0.05, "learning_rate": 1.3001260698228473e-05, "loss": 0.8943, "step": 12647 }, { "epoch": 0.05, "learning_rate": 1.3000232490373959e-05, "loss": 0.8783, "step": 12648 }, { "epoch": 0.05, "learning_rate": 1.2999204247661905e-05, "loss": 0.9099, "step": 12649 }, { "epoch": 0.05, "learning_rate": 1.2998175970104256e-05, "loss": 0.9093, "step": 12650 }, { "epoch": 0.05, "learning_rate": 1.2997147657712965e-05, "loss": 0.9485, "step": 12651 }, { "epoch": 0.05, "learning_rate": 1.2996119310499972e-05, "loss": 0.8956, "step": 12652 }, { "epoch": 0.06, "learning_rate": 1.2995090928477225e-05, "loss": 0.8974, "step": 12653 }, { "epoch": 0.06, "learning_rate": 1.299406251165668e-05, "loss": 0.9051, "step": 12654 }, { "epoch": 0.06, "learning_rate": 1.2993034060050276e-05, "loss": 0.9303, "step": 12655 }, { "epoch": 0.06, "learning_rate": 1.2992005573669964e-05, "loss": 0.9409, "step": 12656 }, { "epoch": 0.06, "learning_rate": 1.2990977052527696e-05, "loss": 0.9313, "step": 12657 }, { "epoch": 0.06, "learning_rate": 1.2989948496635423e-05, "loss": 0.8839, "step": 12658 }, { "epoch": 0.06, "learning_rate": 1.298891990600509e-05, "loss": 0.9546, "step": 12659 }, { "epoch": 0.06, "learning_rate": 1.2987891280648648e-05, "loss": 0.9461, "step": 12660 }, { "epoch": 0.06, "learning_rate": 1.298686262057805e-05, "loss": 0.9269, "step": 12661 }, { "epoch": 0.06, "learning_rate": 1.2985833925805247e-05, "loss": 0.9287, "step": 12662 }, { "epoch": 0.06, "learning_rate": 1.2984805196342193e-05, "loss": 0.9165, "step": 12663 }, { "epoch": 0.06, "learning_rate": 1.2983776432200837e-05, "loss": 0.9012, "step": 12664 }, { "epoch": 0.06, "learning_rate": 1.298274763339313e-05, "loss": 0.9402, "step": 12665 }, { "epoch": 0.06, "learning_rate": 1.298171879993103e-05, "loss": 0.9118, "step": 12666 }, { "epoch": 0.06, "learning_rate": 1.2980689931826487e-05, "loss": 0.928, "step": 12667 }, { "epoch": 0.06, "learning_rate": 1.2979661029091452e-05, "loss": 0.8871, "step": 12668 }, { "epoch": 0.06, "learning_rate": 1.2978632091737885e-05, "loss": 0.9025, "step": 12669 }, { "epoch": 0.06, "learning_rate": 1.2977603119777737e-05, "loss": 0.8865, "step": 12670 }, { "epoch": 0.06, "learning_rate": 1.2976574113222964e-05, "loss": 0.9301, "step": 12671 }, { "epoch": 0.06, "learning_rate": 1.2975545072085519e-05, "loss": 0.8797, "step": 12672 }, { "epoch": 0.06, "learning_rate": 1.297451599637736e-05, "loss": 0.9249, "step": 12673 }, { "epoch": 0.06, "learning_rate": 1.2973486886110442e-05, "loss": 0.9692, "step": 12674 }, { "epoch": 0.06, "learning_rate": 1.2972457741296722e-05, "loss": 0.9093, "step": 12675 }, { "epoch": 0.06, "learning_rate": 1.2971428561948158e-05, "loss": 0.943, "step": 12676 }, { "epoch": 0.06, "learning_rate": 1.2970399348076705e-05, "loss": 0.9129, "step": 12677 }, { "epoch": 0.06, "learning_rate": 1.2969370099694322e-05, "loss": 0.901, "step": 12678 }, { "epoch": 0.06, "learning_rate": 1.296834081681297e-05, "loss": 0.9533, "step": 12679 }, { "epoch": 0.06, "learning_rate": 1.2967311499444599e-05, "loss": 0.9023, "step": 12680 }, { "epoch": 0.06, "learning_rate": 1.2966282147601177e-05, "loss": 0.9099, "step": 12681 }, { "epoch": 0.06, "learning_rate": 1.296525276129466e-05, "loss": 0.8933, "step": 12682 }, { "epoch": 0.06, "learning_rate": 1.2964223340537006e-05, "loss": 0.9063, "step": 12683 }, { "epoch": 0.06, "learning_rate": 1.2963193885340176e-05, "loss": 0.9354, "step": 12684 }, { "epoch": 0.06, "learning_rate": 1.296216439571613e-05, "loss": 0.8839, "step": 12685 }, { "epoch": 0.06, "learning_rate": 1.2961134871676832e-05, "loss": 0.8891, "step": 12686 }, { "epoch": 0.06, "learning_rate": 1.296010531323424e-05, "loss": 0.9016, "step": 12687 }, { "epoch": 0.06, "learning_rate": 1.2959075720400316e-05, "loss": 0.9086, "step": 12688 }, { "epoch": 0.06, "learning_rate": 1.2958046093187024e-05, "loss": 0.9041, "step": 12689 }, { "epoch": 0.06, "learning_rate": 1.2957016431606326e-05, "loss": 0.9102, "step": 12690 }, { "epoch": 0.06, "learning_rate": 1.2955986735670186e-05, "loss": 0.9069, "step": 12691 }, { "epoch": 0.06, "learning_rate": 1.2954957005390563e-05, "loss": 0.9496, "step": 12692 }, { "epoch": 0.06, "learning_rate": 1.2953927240779422e-05, "loss": 0.9193, "step": 12693 }, { "epoch": 0.06, "learning_rate": 1.295289744184873e-05, "loss": 0.8803, "step": 12694 }, { "epoch": 0.06, "learning_rate": 1.295186760861045e-05, "loss": 0.9188, "step": 12695 }, { "epoch": 0.06, "learning_rate": 1.2950837741076545e-05, "loss": 0.9051, "step": 12696 }, { "epoch": 0.06, "learning_rate": 1.2949807839258986e-05, "loss": 0.9083, "step": 12697 }, { "epoch": 0.06, "learning_rate": 1.2948777903169733e-05, "loss": 0.9217, "step": 12698 }, { "epoch": 0.06, "learning_rate": 1.294774793282075e-05, "loss": 0.9689, "step": 12699 }, { "epoch": 0.06, "learning_rate": 1.2946717928224012e-05, "loss": 0.8825, "step": 12700 }, { "epoch": 0.06, "learning_rate": 1.2945687889391483e-05, "loss": 0.9148, "step": 12701 }, { "epoch": 0.06, "learning_rate": 1.2944657816335124e-05, "loss": 0.9094, "step": 12702 }, { "epoch": 0.06, "learning_rate": 1.2943627709066908e-05, "loss": 0.9071, "step": 12703 }, { "epoch": 0.06, "learning_rate": 1.2942597567598805e-05, "loss": 0.8977, "step": 12704 }, { "epoch": 0.06, "learning_rate": 1.2941567391942778e-05, "loss": 0.9062, "step": 12705 }, { "epoch": 0.06, "learning_rate": 1.2940537182110797e-05, "loss": 0.9089, "step": 12706 }, { "epoch": 0.06, "learning_rate": 1.2939506938114836e-05, "loss": 0.9493, "step": 12707 }, { "epoch": 0.06, "learning_rate": 1.2938476659966858e-05, "loss": 0.8729, "step": 12708 }, { "epoch": 0.06, "learning_rate": 1.293744634767884e-05, "loss": 0.8853, "step": 12709 }, { "epoch": 0.06, "learning_rate": 1.2936416001262749e-05, "loss": 0.8718, "step": 12710 }, { "epoch": 0.06, "learning_rate": 1.2935385620730554e-05, "loss": 0.8861, "step": 12711 }, { "epoch": 0.06, "learning_rate": 1.2934355206094229e-05, "loss": 0.9121, "step": 12712 }, { "epoch": 0.06, "learning_rate": 1.2933324757365746e-05, "loss": 0.9309, "step": 12713 }, { "epoch": 0.06, "learning_rate": 1.2932294274557075e-05, "loss": 0.9036, "step": 12714 }, { "epoch": 0.06, "learning_rate": 1.2931263757680189e-05, "loss": 0.8815, "step": 12715 }, { "epoch": 0.06, "learning_rate": 1.293023320674706e-05, "loss": 0.9172, "step": 12716 }, { "epoch": 0.06, "learning_rate": 1.2929202621769663e-05, "loss": 0.91, "step": 12717 }, { "epoch": 0.06, "learning_rate": 1.2928172002759972e-05, "loss": 0.9403, "step": 12718 }, { "epoch": 0.06, "learning_rate": 1.2927141349729957e-05, "loss": 0.8873, "step": 12719 }, { "epoch": 0.06, "learning_rate": 1.2926110662691597e-05, "loss": 0.8737, "step": 12720 }, { "epoch": 0.06, "learning_rate": 1.2925079941656868e-05, "loss": 0.8898, "step": 12721 }, { "epoch": 0.06, "learning_rate": 1.292404918663774e-05, "loss": 0.8908, "step": 12722 }, { "epoch": 0.06, "learning_rate": 1.2923018397646189e-05, "loss": 0.9348, "step": 12723 }, { "epoch": 0.06, "learning_rate": 1.2921987574694196e-05, "loss": 0.8956, "step": 12724 }, { "epoch": 0.06, "learning_rate": 1.2920956717793736e-05, "loss": 0.9622, "step": 12725 }, { "epoch": 0.06, "learning_rate": 1.291992582695678e-05, "loss": 0.9047, "step": 12726 }, { "epoch": 0.06, "learning_rate": 1.2918894902195312e-05, "loss": 0.9009, "step": 12727 }, { "epoch": 0.06, "learning_rate": 1.2917863943521305e-05, "loss": 0.9154, "step": 12728 }, { "epoch": 0.06, "learning_rate": 1.291683295094674e-05, "loss": 0.9113, "step": 12729 }, { "epoch": 0.06, "learning_rate": 1.2915801924483594e-05, "loss": 0.8861, "step": 12730 }, { "epoch": 0.06, "learning_rate": 1.2914770864143845e-05, "loss": 0.9546, "step": 12731 }, { "epoch": 0.06, "learning_rate": 1.2913739769939476e-05, "loss": 0.9085, "step": 12732 }, { "epoch": 0.06, "learning_rate": 1.2912708641882458e-05, "loss": 0.9203, "step": 12733 }, { "epoch": 0.06, "learning_rate": 1.2911677479984781e-05, "loss": 0.8606, "step": 12734 }, { "epoch": 0.06, "learning_rate": 1.2910646284258422e-05, "loss": 0.9124, "step": 12735 }, { "epoch": 0.06, "learning_rate": 1.2909615054715357e-05, "loss": 0.8674, "step": 12736 }, { "epoch": 0.06, "learning_rate": 1.2908583791367572e-05, "loss": 0.9097, "step": 12737 }, { "epoch": 0.06, "learning_rate": 1.2907552494227049e-05, "loss": 0.8953, "step": 12738 }, { "epoch": 0.06, "learning_rate": 1.2906521163305765e-05, "loss": 0.9368, "step": 12739 }, { "epoch": 0.06, "learning_rate": 1.2905489798615709e-05, "loss": 0.8747, "step": 12740 }, { "epoch": 0.06, "learning_rate": 1.2904458400168859e-05, "loss": 0.9377, "step": 12741 }, { "epoch": 0.06, "learning_rate": 1.2903426967977196e-05, "loss": 0.9072, "step": 12742 }, { "epoch": 0.06, "learning_rate": 1.2902395502052711e-05, "loss": 0.9543, "step": 12743 }, { "epoch": 0.06, "learning_rate": 1.2901364002407384e-05, "loss": 0.8978, "step": 12744 }, { "epoch": 0.06, "learning_rate": 1.2900332469053194e-05, "loss": 0.9297, "step": 12745 }, { "epoch": 0.06, "learning_rate": 1.2899300902002137e-05, "loss": 0.8838, "step": 12746 }, { "epoch": 0.06, "learning_rate": 1.2898269301266187e-05, "loss": 0.9222, "step": 12747 }, { "epoch": 0.06, "learning_rate": 1.2897237666857332e-05, "loss": 0.9056, "step": 12748 }, { "epoch": 0.06, "learning_rate": 1.2896205998787564e-05, "loss": 0.9051, "step": 12749 }, { "epoch": 0.06, "learning_rate": 1.2895174297068861e-05, "loss": 0.9246, "step": 12750 }, { "epoch": 0.06, "learning_rate": 1.2894142561713218e-05, "loss": 0.914, "step": 12751 }, { "epoch": 0.06, "learning_rate": 1.2893110792732615e-05, "loss": 0.8811, "step": 12752 }, { "epoch": 0.06, "learning_rate": 1.2892078990139043e-05, "loss": 0.924, "step": 12753 }, { "epoch": 0.06, "learning_rate": 1.2891047153944485e-05, "loss": 0.8621, "step": 12754 }, { "epoch": 0.06, "learning_rate": 1.2890015284160937e-05, "loss": 0.9241, "step": 12755 }, { "epoch": 0.06, "learning_rate": 1.2888983380800382e-05, "loss": 0.9165, "step": 12756 }, { "epoch": 0.06, "learning_rate": 1.288795144387481e-05, "loss": 0.9287, "step": 12757 }, { "epoch": 0.06, "learning_rate": 1.2886919473396212e-05, "loss": 0.9585, "step": 12758 }, { "epoch": 0.06, "learning_rate": 1.2885887469376575e-05, "loss": 0.8941, "step": 12759 }, { "epoch": 0.06, "learning_rate": 1.2884855431827892e-05, "loss": 0.9318, "step": 12760 }, { "epoch": 0.06, "learning_rate": 1.2883823360762149e-05, "loss": 0.9099, "step": 12761 }, { "epoch": 0.06, "learning_rate": 1.2882791256191344e-05, "loss": 0.9047, "step": 12762 }, { "epoch": 0.06, "learning_rate": 1.2881759118127461e-05, "loss": 0.9032, "step": 12763 }, { "epoch": 0.06, "learning_rate": 1.2880726946582496e-05, "loss": 0.9058, "step": 12764 }, { "epoch": 0.06, "learning_rate": 1.2879694741568438e-05, "loss": 0.8898, "step": 12765 }, { "epoch": 0.06, "learning_rate": 1.2878662503097283e-05, "loss": 0.9146, "step": 12766 }, { "epoch": 0.06, "learning_rate": 1.2877630231181021e-05, "loss": 0.9602, "step": 12767 }, { "epoch": 0.06, "learning_rate": 1.2876597925831648e-05, "loss": 0.9124, "step": 12768 }, { "epoch": 0.06, "learning_rate": 1.2875565587061156e-05, "loss": 0.9084, "step": 12769 }, { "epoch": 0.06, "learning_rate": 1.2874533214881537e-05, "loss": 0.9215, "step": 12770 }, { "epoch": 0.06, "learning_rate": 1.2873500809304792e-05, "loss": 0.8823, "step": 12771 }, { "epoch": 0.06, "learning_rate": 1.2872468370342908e-05, "loss": 0.9146, "step": 12772 }, { "epoch": 0.06, "learning_rate": 1.2871435898007882e-05, "loss": 0.917, "step": 12773 }, { "epoch": 0.06, "learning_rate": 1.2870403392311715e-05, "loss": 0.9193, "step": 12774 }, { "epoch": 0.06, "learning_rate": 1.2869370853266394e-05, "loss": 0.8845, "step": 12775 }, { "epoch": 0.06, "learning_rate": 1.2868338280883924e-05, "loss": 0.9005, "step": 12776 }, { "epoch": 0.06, "learning_rate": 1.2867305675176293e-05, "loss": 0.8846, "step": 12777 }, { "epoch": 0.06, "learning_rate": 1.2866273036155508e-05, "loss": 0.9392, "step": 12778 }, { "epoch": 0.06, "learning_rate": 1.2865240363833556e-05, "loss": 0.9574, "step": 12779 }, { "epoch": 0.06, "learning_rate": 1.2864207658222444e-05, "loss": 0.9525, "step": 12780 }, { "epoch": 0.06, "learning_rate": 1.2863174919334166e-05, "loss": 0.9064, "step": 12781 }, { "epoch": 0.06, "learning_rate": 1.2862142147180723e-05, "loss": 0.9149, "step": 12782 }, { "epoch": 0.06, "learning_rate": 1.2861109341774114e-05, "loss": 0.9763, "step": 12783 }, { "epoch": 0.06, "learning_rate": 1.2860076503126332e-05, "loss": 0.8948, "step": 12784 }, { "epoch": 0.06, "learning_rate": 1.2859043631249382e-05, "loss": 0.9087, "step": 12785 }, { "epoch": 0.06, "learning_rate": 1.2858010726155263e-05, "loss": 0.9026, "step": 12786 }, { "epoch": 0.06, "learning_rate": 1.2856977787855983e-05, "loss": 0.9473, "step": 12787 }, { "epoch": 0.06, "learning_rate": 1.2855944816363528e-05, "loss": 0.9039, "step": 12788 }, { "epoch": 0.06, "learning_rate": 1.2854911811689913e-05, "loss": 0.9523, "step": 12789 }, { "epoch": 0.06, "learning_rate": 1.2853878773847133e-05, "loss": 0.8933, "step": 12790 }, { "epoch": 0.06, "learning_rate": 1.2852845702847187e-05, "loss": 0.9293, "step": 12791 }, { "epoch": 0.06, "learning_rate": 1.2851812598702087e-05, "loss": 0.9123, "step": 12792 }, { "epoch": 0.06, "learning_rate": 1.285077946142383e-05, "loss": 0.9485, "step": 12793 }, { "epoch": 0.06, "learning_rate": 1.284974629102442e-05, "loss": 0.8931, "step": 12794 }, { "epoch": 0.06, "learning_rate": 1.2848713087515858e-05, "loss": 0.9314, "step": 12795 }, { "epoch": 0.06, "learning_rate": 1.2847679850910156e-05, "loss": 0.8933, "step": 12796 }, { "epoch": 0.06, "learning_rate": 1.2846646581219311e-05, "loss": 0.9447, "step": 12797 }, { "epoch": 0.06, "learning_rate": 1.2845613278455329e-05, "loss": 0.8979, "step": 12798 }, { "epoch": 0.06, "learning_rate": 1.2844579942630217e-05, "loss": 0.9167, "step": 12799 }, { "epoch": 0.06, "learning_rate": 1.2843546573755979e-05, "loss": 0.8689, "step": 12800 }, { "epoch": 0.06, "learning_rate": 1.2842513171844623e-05, "loss": 0.8789, "step": 12801 }, { "epoch": 0.06, "learning_rate": 1.2841479736908154e-05, "loss": 0.9263, "step": 12802 }, { "epoch": 0.06, "learning_rate": 1.2840446268958577e-05, "loss": 0.8992, "step": 12803 }, { "epoch": 0.06, "learning_rate": 1.2839412768007905e-05, "loss": 0.9166, "step": 12804 }, { "epoch": 0.06, "learning_rate": 1.283837923406814e-05, "loss": 0.8923, "step": 12805 }, { "epoch": 0.06, "learning_rate": 1.283734566715129e-05, "loss": 0.8986, "step": 12806 }, { "epoch": 0.06, "learning_rate": 1.2836312067269365e-05, "loss": 0.9256, "step": 12807 }, { "epoch": 0.06, "learning_rate": 1.2835278434434375e-05, "loss": 0.893, "step": 12808 }, { "epoch": 0.06, "learning_rate": 1.2834244768658324e-05, "loss": 0.8883, "step": 12809 }, { "epoch": 0.06, "learning_rate": 1.2833211069953228e-05, "loss": 0.9061, "step": 12810 }, { "epoch": 0.06, "learning_rate": 1.2832177338331093e-05, "loss": 0.9425, "step": 12811 }, { "epoch": 0.06, "learning_rate": 1.2831143573803931e-05, "loss": 0.9169, "step": 12812 }, { "epoch": 0.06, "learning_rate": 1.2830109776383751e-05, "loss": 0.9468, "step": 12813 }, { "epoch": 0.06, "learning_rate": 1.2829075946082563e-05, "loss": 0.8842, "step": 12814 }, { "epoch": 0.06, "learning_rate": 1.2828042082912377e-05, "loss": 0.9249, "step": 12815 }, { "epoch": 0.06, "learning_rate": 1.282700818688521e-05, "loss": 0.8972, "step": 12816 }, { "epoch": 0.06, "learning_rate": 1.2825974258013074e-05, "loss": 0.899, "step": 12817 }, { "epoch": 0.06, "learning_rate": 1.2824940296307977e-05, "loss": 0.9161, "step": 12818 }, { "epoch": 0.06, "learning_rate": 1.2823906301781933e-05, "loss": 0.9227, "step": 12819 }, { "epoch": 0.06, "learning_rate": 1.2822872274446958e-05, "loss": 0.9271, "step": 12820 }, { "epoch": 0.06, "learning_rate": 1.282183821431506e-05, "loss": 0.9057, "step": 12821 }, { "epoch": 0.06, "learning_rate": 1.282080412139826e-05, "loss": 0.9091, "step": 12822 }, { "epoch": 0.06, "learning_rate": 1.2819769995708567e-05, "loss": 0.9132, "step": 12823 }, { "epoch": 0.06, "learning_rate": 1.2818735837257998e-05, "loss": 0.9426, "step": 12824 }, { "epoch": 0.06, "learning_rate": 1.2817701646058568e-05, "loss": 0.911, "step": 12825 }, { "epoch": 0.06, "learning_rate": 1.2816667422122295e-05, "loss": 0.9276, "step": 12826 }, { "epoch": 0.06, "learning_rate": 1.281563316546119e-05, "loss": 0.8808, "step": 12827 }, { "epoch": 0.06, "learning_rate": 1.281459887608727e-05, "loss": 0.9029, "step": 12828 }, { "epoch": 0.06, "learning_rate": 1.2813564554012557e-05, "loss": 0.9351, "step": 12829 }, { "epoch": 0.06, "learning_rate": 1.2812530199249063e-05, "loss": 0.9101, "step": 12830 }, { "epoch": 0.06, "learning_rate": 1.2811495811808805e-05, "loss": 0.9494, "step": 12831 }, { "epoch": 0.06, "learning_rate": 1.2810461391703804e-05, "loss": 0.9048, "step": 12832 }, { "epoch": 0.06, "learning_rate": 1.2809426938946077e-05, "loss": 0.9174, "step": 12833 }, { "epoch": 0.06, "learning_rate": 1.280839245354764e-05, "loss": 0.8903, "step": 12834 }, { "epoch": 0.06, "learning_rate": 1.2807357935520517e-05, "loss": 0.9358, "step": 12835 }, { "epoch": 0.06, "learning_rate": 1.2806323384876727e-05, "loss": 0.9403, "step": 12836 }, { "epoch": 0.06, "learning_rate": 1.280528880162828e-05, "loss": 0.9497, "step": 12837 }, { "epoch": 0.06, "learning_rate": 1.2804254185787208e-05, "loss": 0.9107, "step": 12838 }, { "epoch": 0.06, "learning_rate": 1.2803219537365528e-05, "loss": 0.9176, "step": 12839 }, { "epoch": 0.06, "learning_rate": 1.2802184856375255e-05, "loss": 0.9361, "step": 12840 }, { "epoch": 0.06, "learning_rate": 1.2801150142828419e-05, "loss": 0.8905, "step": 12841 }, { "epoch": 0.06, "learning_rate": 1.2800115396737036e-05, "loss": 0.9145, "step": 12842 }, { "epoch": 0.06, "learning_rate": 1.279908061811313e-05, "loss": 0.9213, "step": 12843 }, { "epoch": 0.06, "learning_rate": 1.2798045806968722e-05, "loss": 0.935, "step": 12844 }, { "epoch": 0.06, "learning_rate": 1.2797010963315836e-05, "loss": 0.9086, "step": 12845 }, { "epoch": 0.06, "learning_rate": 1.279597608716649e-05, "loss": 0.9214, "step": 12846 }, { "epoch": 0.06, "learning_rate": 1.2794941178532721e-05, "loss": 0.9178, "step": 12847 }, { "epoch": 0.06, "learning_rate": 1.279390623742654e-05, "loss": 0.8847, "step": 12848 }, { "epoch": 0.06, "learning_rate": 1.2792871263859972e-05, "loss": 0.9686, "step": 12849 }, { "epoch": 0.06, "learning_rate": 1.2791836257845046e-05, "loss": 0.8969, "step": 12850 }, { "epoch": 0.06, "learning_rate": 1.2790801219393789e-05, "loss": 0.9571, "step": 12851 }, { "epoch": 0.06, "learning_rate": 1.2789766148518221e-05, "loss": 0.8989, "step": 12852 }, { "epoch": 0.06, "learning_rate": 1.278873104523037e-05, "loss": 0.896, "step": 12853 }, { "epoch": 0.06, "learning_rate": 1.2787695909542262e-05, "loss": 0.9097, "step": 12854 }, { "epoch": 0.06, "learning_rate": 1.278666074146592e-05, "loss": 0.9298, "step": 12855 }, { "epoch": 0.06, "learning_rate": 1.278562554101338e-05, "loss": 0.8947, "step": 12856 }, { "epoch": 0.06, "learning_rate": 1.278459030819666e-05, "loss": 0.9244, "step": 12857 }, { "epoch": 0.06, "learning_rate": 1.2783555043027794e-05, "loss": 0.9335, "step": 12858 }, { "epoch": 0.06, "learning_rate": 1.2782519745518807e-05, "loss": 0.9444, "step": 12859 }, { "epoch": 0.06, "learning_rate": 1.2781484415681725e-05, "loss": 0.9241, "step": 12860 }, { "epoch": 0.06, "learning_rate": 1.278044905352858e-05, "loss": 0.8963, "step": 12861 }, { "epoch": 0.06, "learning_rate": 1.27794136590714e-05, "loss": 0.8946, "step": 12862 }, { "epoch": 0.06, "learning_rate": 1.2778378232322219e-05, "loss": 0.9178, "step": 12863 }, { "epoch": 0.06, "learning_rate": 1.277734277329306e-05, "loss": 0.9293, "step": 12864 }, { "epoch": 0.06, "learning_rate": 1.2776307281995954e-05, "loss": 0.9187, "step": 12865 }, { "epoch": 0.06, "learning_rate": 1.2775271758442937e-05, "loss": 0.9712, "step": 12866 }, { "epoch": 0.06, "learning_rate": 1.2774236202646035e-05, "loss": 0.8992, "step": 12867 }, { "epoch": 0.06, "learning_rate": 1.2773200614617283e-05, "loss": 0.9309, "step": 12868 }, { "epoch": 0.06, "learning_rate": 1.2772164994368707e-05, "loss": 0.932, "step": 12869 }, { "epoch": 0.06, "learning_rate": 1.2771129341912345e-05, "loss": 0.8579, "step": 12870 }, { "epoch": 0.06, "learning_rate": 1.2770093657260229e-05, "loss": 0.9116, "step": 12871 }, { "epoch": 0.06, "learning_rate": 1.2769057940424391e-05, "loss": 0.9116, "step": 12872 }, { "epoch": 0.06, "learning_rate": 1.2768022191416861e-05, "loss": 0.936, "step": 12873 }, { "epoch": 0.06, "learning_rate": 1.2766986410249676e-05, "loss": 0.9198, "step": 12874 }, { "epoch": 0.06, "learning_rate": 1.2765950596934871e-05, "loss": 0.8896, "step": 12875 }, { "epoch": 0.06, "learning_rate": 1.2764914751484477e-05, "loss": 0.9062, "step": 12876 }, { "epoch": 0.06, "learning_rate": 1.2763878873910529e-05, "loss": 0.8705, "step": 12877 }, { "epoch": 0.06, "learning_rate": 1.2762842964225066e-05, "loss": 0.8978, "step": 12878 }, { "epoch": 0.06, "learning_rate": 1.276180702244012e-05, "loss": 0.9046, "step": 12879 }, { "epoch": 0.06, "learning_rate": 1.2760771048567726e-05, "loss": 0.9167, "step": 12880 }, { "epoch": 0.06, "learning_rate": 1.2759735042619925e-05, "loss": 0.9046, "step": 12881 }, { "epoch": 0.06, "learning_rate": 1.275869900460875e-05, "loss": 0.8794, "step": 12882 }, { "epoch": 0.06, "learning_rate": 1.2757662934546237e-05, "loss": 0.9126, "step": 12883 }, { "epoch": 0.06, "learning_rate": 1.2756626832444428e-05, "loss": 0.9047, "step": 12884 }, { "epoch": 0.06, "learning_rate": 1.2755590698315354e-05, "loss": 0.9209, "step": 12885 }, { "epoch": 0.06, "learning_rate": 1.2754554532171057e-05, "loss": 0.9347, "step": 12886 }, { "epoch": 0.06, "learning_rate": 1.2753518334023579e-05, "loss": 0.938, "step": 12887 }, { "epoch": 0.06, "learning_rate": 1.2752482103884953e-05, "loss": 0.8596, "step": 12888 }, { "epoch": 0.06, "learning_rate": 1.275144584176722e-05, "loss": 0.9198, "step": 12889 }, { "epoch": 0.06, "learning_rate": 1.2750409547682419e-05, "loss": 0.8722, "step": 12890 }, { "epoch": 0.06, "learning_rate": 1.2749373221642594e-05, "loss": 0.9209, "step": 12891 }, { "epoch": 0.06, "learning_rate": 1.2748336863659775e-05, "loss": 0.8942, "step": 12892 }, { "epoch": 0.06, "learning_rate": 1.2747300473746018e-05, "loss": 0.9024, "step": 12893 }, { "epoch": 0.06, "learning_rate": 1.2746264051913353e-05, "loss": 0.9085, "step": 12894 }, { "epoch": 0.06, "learning_rate": 1.2745227598173824e-05, "loss": 0.9106, "step": 12895 }, { "epoch": 0.06, "learning_rate": 1.2744191112539475e-05, "loss": 0.9034, "step": 12896 }, { "epoch": 0.06, "learning_rate": 1.2743154595022344e-05, "loss": 0.9224, "step": 12897 }, { "epoch": 0.06, "learning_rate": 1.2742118045634476e-05, "loss": 0.8665, "step": 12898 }, { "epoch": 0.06, "learning_rate": 1.2741081464387914e-05, "loss": 0.9191, "step": 12899 }, { "epoch": 0.06, "learning_rate": 1.2740044851294704e-05, "loss": 0.9264, "step": 12900 }, { "epoch": 0.06, "learning_rate": 1.2739008206366884e-05, "loss": 0.9153, "step": 12901 }, { "epoch": 0.06, "learning_rate": 1.2737971529616501e-05, "loss": 0.8888, "step": 12902 }, { "epoch": 0.06, "learning_rate": 1.27369348210556e-05, "loss": 0.9485, "step": 12903 }, { "epoch": 0.06, "learning_rate": 1.2735898080696223e-05, "loss": 0.9423, "step": 12904 }, { "epoch": 0.06, "learning_rate": 1.2734861308550422e-05, "loss": 0.9221, "step": 12905 }, { "epoch": 0.06, "learning_rate": 1.2733824504630233e-05, "loss": 0.9209, "step": 12906 }, { "epoch": 0.06, "learning_rate": 1.2732787668947708e-05, "loss": 0.8983, "step": 12907 }, { "epoch": 0.06, "learning_rate": 1.273175080151489e-05, "loss": 0.8857, "step": 12908 }, { "epoch": 0.06, "learning_rate": 1.2730713902343832e-05, "loss": 0.9254, "step": 12909 }, { "epoch": 0.06, "learning_rate": 1.2729676971446572e-05, "loss": 0.9267, "step": 12910 }, { "epoch": 0.06, "learning_rate": 1.2728640008835163e-05, "loss": 0.9289, "step": 12911 }, { "epoch": 0.06, "learning_rate": 1.2727603014521655e-05, "loss": 0.8952, "step": 12912 }, { "epoch": 0.06, "learning_rate": 1.2726565988518085e-05, "loss": 0.9254, "step": 12913 }, { "epoch": 0.06, "learning_rate": 1.2725528930836515e-05, "loss": 0.881, "step": 12914 }, { "epoch": 0.06, "learning_rate": 1.2724491841488986e-05, "loss": 0.9293, "step": 12915 }, { "epoch": 0.06, "learning_rate": 1.272345472048755e-05, "loss": 0.896, "step": 12916 }, { "epoch": 0.06, "learning_rate": 1.2722417567844252e-05, "loss": 0.9066, "step": 12917 }, { "epoch": 0.06, "learning_rate": 1.2721380383571149e-05, "loss": 0.9046, "step": 12918 }, { "epoch": 0.06, "learning_rate": 1.2720343167680289e-05, "loss": 0.925, "step": 12919 }, { "epoch": 0.06, "learning_rate": 1.2719305920183718e-05, "loss": 0.8889, "step": 12920 }, { "epoch": 0.06, "learning_rate": 1.2718268641093494e-05, "loss": 0.921, "step": 12921 }, { "epoch": 0.06, "learning_rate": 1.2717231330421661e-05, "loss": 0.9015, "step": 12922 }, { "epoch": 0.06, "learning_rate": 1.2716193988180276e-05, "loss": 0.9379, "step": 12923 }, { "epoch": 0.06, "learning_rate": 1.271515661438139e-05, "loss": 0.8892, "step": 12924 }, { "epoch": 0.06, "learning_rate": 1.2714119209037056e-05, "loss": 0.9192, "step": 12925 }, { "epoch": 0.06, "learning_rate": 1.2713081772159324e-05, "loss": 0.906, "step": 12926 }, { "epoch": 0.06, "learning_rate": 1.2712044303760253e-05, "loss": 0.9109, "step": 12927 }, { "epoch": 0.06, "learning_rate": 1.271100680385189e-05, "loss": 0.9201, "step": 12928 }, { "epoch": 0.06, "learning_rate": 1.2709969272446295e-05, "loss": 0.9394, "step": 12929 }, { "epoch": 0.06, "learning_rate": 1.2708931709555517e-05, "loss": 0.9194, "step": 12930 }, { "epoch": 0.06, "learning_rate": 1.2707894115191613e-05, "loss": 0.9142, "step": 12931 }, { "epoch": 0.06, "learning_rate": 1.2706856489366638e-05, "loss": 0.8575, "step": 12932 }, { "epoch": 0.06, "learning_rate": 1.2705818832092647e-05, "loss": 0.9225, "step": 12933 }, { "epoch": 0.06, "learning_rate": 1.2704781143381699e-05, "loss": 0.96, "step": 12934 }, { "epoch": 0.06, "learning_rate": 1.2703743423245844e-05, "loss": 0.9493, "step": 12935 }, { "epoch": 0.06, "learning_rate": 1.2702705671697144e-05, "loss": 0.9247, "step": 12936 }, { "epoch": 0.06, "learning_rate": 1.2701667888747653e-05, "loss": 0.9057, "step": 12937 }, { "epoch": 0.06, "learning_rate": 1.2700630074409427e-05, "loss": 0.9328, "step": 12938 }, { "epoch": 0.06, "learning_rate": 1.2699592228694531e-05, "loss": 0.9221, "step": 12939 }, { "epoch": 0.06, "learning_rate": 1.2698554351615013e-05, "loss": 0.885, "step": 12940 }, { "epoch": 0.06, "learning_rate": 1.2697516443182938e-05, "loss": 0.8917, "step": 12941 }, { "epoch": 0.06, "learning_rate": 1.2696478503410364e-05, "loss": 0.9447, "step": 12942 }, { "epoch": 0.06, "learning_rate": 1.2695440532309346e-05, "loss": 0.9165, "step": 12943 }, { "epoch": 0.06, "learning_rate": 1.2694402529891947e-05, "loss": 0.9121, "step": 12944 }, { "epoch": 0.06, "learning_rate": 1.2693364496170226e-05, "loss": 0.925, "step": 12945 }, { "epoch": 0.06, "learning_rate": 1.2692326431156245e-05, "loss": 0.9424, "step": 12946 }, { "epoch": 0.06, "learning_rate": 1.269128833486206e-05, "loss": 0.8994, "step": 12947 }, { "epoch": 0.06, "learning_rate": 1.269025020729974e-05, "loss": 0.8763, "step": 12948 }, { "epoch": 0.06, "learning_rate": 1.2689212048481333e-05, "loss": 0.91, "step": 12949 }, { "epoch": 0.06, "learning_rate": 1.2688173858418913e-05, "loss": 0.863, "step": 12950 }, { "epoch": 0.06, "learning_rate": 1.2687135637124539e-05, "loss": 0.9033, "step": 12951 }, { "epoch": 0.06, "learning_rate": 1.268609738461027e-05, "loss": 0.8755, "step": 12952 }, { "epoch": 0.06, "learning_rate": 1.2685059100888171e-05, "loss": 0.9237, "step": 12953 }, { "epoch": 0.07, "learning_rate": 1.2684020785970302e-05, "loss": 0.8766, "step": 12954 }, { "epoch": 0.07, "learning_rate": 1.2682982439868732e-05, "loss": 0.9062, "step": 12955 }, { "epoch": 0.07, "learning_rate": 1.2681944062595521e-05, "loss": 0.9531, "step": 12956 }, { "epoch": 0.07, "learning_rate": 1.2680905654162733e-05, "loss": 0.9288, "step": 12957 }, { "epoch": 0.07, "learning_rate": 1.2679867214582438e-05, "loss": 0.903, "step": 12958 }, { "epoch": 0.07, "learning_rate": 1.2678828743866689e-05, "loss": 0.9214, "step": 12959 }, { "epoch": 0.07, "learning_rate": 1.2677790242027564e-05, "loss": 0.9236, "step": 12960 }, { "epoch": 0.07, "learning_rate": 1.2676751709077123e-05, "loss": 0.919, "step": 12961 }, { "epoch": 0.07, "learning_rate": 1.2675713145027428e-05, "loss": 0.8804, "step": 12962 }, { "epoch": 0.07, "learning_rate": 1.2674674549890553e-05, "loss": 0.9099, "step": 12963 }, { "epoch": 0.07, "learning_rate": 1.2673635923678562e-05, "loss": 0.8764, "step": 12964 }, { "epoch": 0.07, "learning_rate": 1.2672597266403522e-05, "loss": 0.8955, "step": 12965 }, { "epoch": 0.07, "learning_rate": 1.2671558578077497e-05, "loss": 0.9157, "step": 12966 }, { "epoch": 0.07, "learning_rate": 1.2670519858712558e-05, "loss": 0.917, "step": 12967 }, { "epoch": 0.07, "learning_rate": 1.2669481108320773e-05, "loss": 0.9222, "step": 12968 }, { "epoch": 0.07, "learning_rate": 1.266844232691421e-05, "loss": 0.9174, "step": 12969 }, { "epoch": 0.07, "learning_rate": 1.2667403514504937e-05, "loss": 0.8862, "step": 12970 }, { "epoch": 0.07, "learning_rate": 1.2666364671105027e-05, "loss": 0.9322, "step": 12971 }, { "epoch": 0.07, "learning_rate": 1.2665325796726541e-05, "loss": 0.9083, "step": 12972 }, { "epoch": 0.07, "learning_rate": 1.2664286891381561e-05, "loss": 0.9107, "step": 12973 }, { "epoch": 0.07, "learning_rate": 1.2663247955082147e-05, "loss": 0.9008, "step": 12974 }, { "epoch": 0.07, "learning_rate": 1.2662208987840375e-05, "loss": 0.9171, "step": 12975 }, { "epoch": 0.07, "learning_rate": 1.2661169989668315e-05, "loss": 0.8888, "step": 12976 }, { "epoch": 0.07, "learning_rate": 1.2660130960578039e-05, "loss": 0.9214, "step": 12977 }, { "epoch": 0.07, "learning_rate": 1.2659091900581615e-05, "loss": 0.8968, "step": 12978 }, { "epoch": 0.07, "learning_rate": 1.2658052809691119e-05, "loss": 0.9155, "step": 12979 }, { "epoch": 0.07, "learning_rate": 1.2657013687918624e-05, "loss": 0.891, "step": 12980 }, { "epoch": 0.07, "learning_rate": 1.2655974535276199e-05, "loss": 0.9298, "step": 12981 }, { "epoch": 0.07, "learning_rate": 1.265493535177592e-05, "loss": 0.8938, "step": 12982 }, { "epoch": 0.07, "learning_rate": 1.2653896137429859e-05, "loss": 0.9276, "step": 12983 }, { "epoch": 0.07, "learning_rate": 1.265285689225009e-05, "loss": 0.8896, "step": 12984 }, { "epoch": 0.07, "learning_rate": 1.2651817616248692e-05, "loss": 0.9204, "step": 12985 }, { "epoch": 0.07, "learning_rate": 1.265077830943773e-05, "loss": 0.9192, "step": 12986 }, { "epoch": 0.07, "learning_rate": 1.2649738971829289e-05, "loss": 0.9055, "step": 12987 }, { "epoch": 0.07, "learning_rate": 1.2648699603435435e-05, "loss": 0.9059, "step": 12988 }, { "epoch": 0.07, "learning_rate": 1.2647660204268253e-05, "loss": 0.9355, "step": 12989 }, { "epoch": 0.07, "learning_rate": 1.2646620774339812e-05, "loss": 0.9155, "step": 12990 }, { "epoch": 0.07, "learning_rate": 1.2645581313662192e-05, "loss": 0.9321, "step": 12991 }, { "epoch": 0.07, "learning_rate": 1.2644541822247468e-05, "loss": 0.9159, "step": 12992 }, { "epoch": 0.07, "learning_rate": 1.2643502300107718e-05, "loss": 0.9032, "step": 12993 }, { "epoch": 0.07, "learning_rate": 1.2642462747255018e-05, "loss": 0.8733, "step": 12994 }, { "epoch": 0.07, "learning_rate": 1.2641423163701447e-05, "loss": 0.9313, "step": 12995 }, { "epoch": 0.07, "learning_rate": 1.2640383549459086e-05, "loss": 0.9004, "step": 12996 }, { "epoch": 0.07, "learning_rate": 1.2639343904540008e-05, "loss": 0.9489, "step": 12997 }, { "epoch": 0.07, "learning_rate": 1.2638304228956295e-05, "loss": 0.8922, "step": 12998 }, { "epoch": 0.07, "learning_rate": 1.2637264522720025e-05, "loss": 0.9519, "step": 12999 }, { "epoch": 0.07, "learning_rate": 1.263622478584328e-05, "loss": 0.901, "step": 13000 }, { "epoch": 0.07, "learning_rate": 1.263518501833814e-05, "loss": 0.921, "step": 13001 }, { "epoch": 0.07, "learning_rate": 1.263414522021668e-05, "loss": 0.9042, "step": 13002 }, { "epoch": 0.07, "learning_rate": 1.2633105391490984e-05, "loss": 0.9228, "step": 13003 }, { "epoch": 0.07, "learning_rate": 1.2632065532173137e-05, "loss": 0.8915, "step": 13004 }, { "epoch": 0.07, "learning_rate": 1.2631025642275212e-05, "loss": 0.8798, "step": 13005 }, { "epoch": 0.07, "learning_rate": 1.26299857218093e-05, "loss": 0.9236, "step": 13006 }, { "epoch": 0.07, "learning_rate": 1.2628945770787478e-05, "loss": 0.9211, "step": 13007 }, { "epoch": 0.07, "learning_rate": 1.2627905789221826e-05, "loss": 0.8801, "step": 13008 }, { "epoch": 0.07, "learning_rate": 1.262686577712443e-05, "loss": 0.9397, "step": 13009 }, { "epoch": 0.07, "learning_rate": 1.2625825734507377e-05, "loss": 0.9209, "step": 13010 }, { "epoch": 0.07, "learning_rate": 1.2624785661382743e-05, "loss": 0.9166, "step": 13011 }, { "epoch": 0.07, "learning_rate": 1.2623745557762616e-05, "loss": 0.9139, "step": 13012 }, { "epoch": 0.07, "learning_rate": 1.2622705423659082e-05, "loss": 0.9303, "step": 13013 }, { "epoch": 0.07, "learning_rate": 1.262166525908422e-05, "loss": 0.9159, "step": 13014 }, { "epoch": 0.07, "learning_rate": 1.2620625064050118e-05, "loss": 0.9113, "step": 13015 }, { "epoch": 0.07, "learning_rate": 1.2619584838568862e-05, "loss": 0.9086, "step": 13016 }, { "epoch": 0.07, "learning_rate": 1.2618544582652536e-05, "loss": 0.8565, "step": 13017 }, { "epoch": 0.07, "learning_rate": 1.2617504296313228e-05, "loss": 0.9134, "step": 13018 }, { "epoch": 0.07, "learning_rate": 1.2616463979563025e-05, "loss": 0.8964, "step": 13019 }, { "epoch": 0.07, "learning_rate": 1.261542363241401e-05, "loss": 0.8966, "step": 13020 }, { "epoch": 0.07, "learning_rate": 1.2614383254878271e-05, "loss": 0.9377, "step": 13021 }, { "epoch": 0.07, "learning_rate": 1.2613342846967898e-05, "loss": 0.9465, "step": 13022 }, { "epoch": 0.07, "learning_rate": 1.2612302408694976e-05, "loss": 0.9108, "step": 13023 }, { "epoch": 0.07, "learning_rate": 1.2611261940071592e-05, "loss": 0.8453, "step": 13024 }, { "epoch": 0.07, "learning_rate": 1.261022144110984e-05, "loss": 0.897, "step": 13025 }, { "epoch": 0.07, "learning_rate": 1.2609180911821808e-05, "loss": 0.9381, "step": 13026 }, { "epoch": 0.07, "learning_rate": 1.2608140352219577e-05, "loss": 0.9485, "step": 13027 }, { "epoch": 0.07, "learning_rate": 1.2607099762315245e-05, "loss": 0.9093, "step": 13028 }, { "epoch": 0.07, "learning_rate": 1.2606059142120901e-05, "loss": 0.9299, "step": 13029 }, { "epoch": 0.07, "learning_rate": 1.2605018491648626e-05, "loss": 0.9286, "step": 13030 }, { "epoch": 0.07, "learning_rate": 1.2603977810910526e-05, "loss": 0.9143, "step": 13031 }, { "epoch": 0.07, "learning_rate": 1.260293709991868e-05, "loss": 0.8857, "step": 13032 }, { "epoch": 0.07, "learning_rate": 1.2601896358685181e-05, "loss": 0.9478, "step": 13033 }, { "epoch": 0.07, "learning_rate": 1.2600855587222125e-05, "loss": 0.8991, "step": 13034 }, { "epoch": 0.07, "learning_rate": 1.2599814785541604e-05, "loss": 0.9183, "step": 13035 }, { "epoch": 0.07, "learning_rate": 1.2598773953655705e-05, "loss": 0.9109, "step": 13036 }, { "epoch": 0.07, "learning_rate": 1.2597733091576523e-05, "loss": 0.9481, "step": 13037 }, { "epoch": 0.07, "learning_rate": 1.2596692199316154e-05, "loss": 0.8738, "step": 13038 }, { "epoch": 0.07, "learning_rate": 1.2595651276886684e-05, "loss": 0.9251, "step": 13039 }, { "epoch": 0.07, "learning_rate": 1.2594610324300218e-05, "loss": 0.8787, "step": 13040 }, { "epoch": 0.07, "learning_rate": 1.259356934156884e-05, "loss": 0.8722, "step": 13041 }, { "epoch": 0.07, "learning_rate": 1.2592528328704649e-05, "loss": 0.9046, "step": 13042 }, { "epoch": 0.07, "learning_rate": 1.2591487285719742e-05, "loss": 0.9025, "step": 13043 }, { "epoch": 0.07, "learning_rate": 1.2590446212626208e-05, "loss": 0.8992, "step": 13044 }, { "epoch": 0.07, "learning_rate": 1.2589405109436146e-05, "loss": 0.9013, "step": 13045 }, { "epoch": 0.07, "learning_rate": 1.258836397616165e-05, "loss": 0.916, "step": 13046 }, { "epoch": 0.07, "learning_rate": 1.2587322812814821e-05, "loss": 0.8727, "step": 13047 }, { "epoch": 0.07, "learning_rate": 1.2586281619407747e-05, "loss": 0.8961, "step": 13048 }, { "epoch": 0.07, "learning_rate": 1.2585240395952536e-05, "loss": 0.9182, "step": 13049 }, { "epoch": 0.07, "learning_rate": 1.2584199142461275e-05, "loss": 0.9276, "step": 13050 }, { "epoch": 0.07, "learning_rate": 1.2583157858946067e-05, "loss": 0.9532, "step": 13051 }, { "epoch": 0.07, "learning_rate": 1.2582116545419012e-05, "loss": 0.8859, "step": 13052 }, { "epoch": 0.07, "learning_rate": 1.25810752018922e-05, "loss": 0.9053, "step": 13053 }, { "epoch": 0.07, "learning_rate": 1.2580033828377736e-05, "loss": 0.9007, "step": 13054 }, { "epoch": 0.07, "learning_rate": 1.2578992424887719e-05, "loss": 0.8811, "step": 13055 }, { "epoch": 0.07, "learning_rate": 1.2577950991434249e-05, "loss": 0.9158, "step": 13056 }, { "epoch": 0.07, "learning_rate": 1.2576909528029419e-05, "loss": 0.9427, "step": 13057 }, { "epoch": 0.07, "learning_rate": 1.2575868034685335e-05, "loss": 0.9011, "step": 13058 }, { "epoch": 0.07, "learning_rate": 1.2574826511414097e-05, "loss": 0.9017, "step": 13059 }, { "epoch": 0.07, "learning_rate": 1.2573784958227803e-05, "loss": 0.9536, "step": 13060 }, { "epoch": 0.07, "learning_rate": 1.2572743375138558e-05, "loss": 0.954, "step": 13061 }, { "epoch": 0.07, "learning_rate": 1.2571701762158458e-05, "loss": 0.8826, "step": 13062 }, { "epoch": 0.07, "learning_rate": 1.2570660119299612e-05, "loss": 0.9355, "step": 13063 }, { "epoch": 0.07, "learning_rate": 1.2569618446574115e-05, "loss": 0.9388, "step": 13064 }, { "epoch": 0.07, "learning_rate": 1.2568576743994074e-05, "loss": 0.9279, "step": 13065 }, { "epoch": 0.07, "learning_rate": 1.2567535011571588e-05, "loss": 0.8824, "step": 13066 }, { "epoch": 0.07, "learning_rate": 1.2566493249318762e-05, "loss": 0.9056, "step": 13067 }, { "epoch": 0.07, "learning_rate": 1.2565451457247705e-05, "loss": 0.8752, "step": 13068 }, { "epoch": 0.07, "learning_rate": 1.256440963537051e-05, "loss": 0.9272, "step": 13069 }, { "epoch": 0.07, "learning_rate": 1.256336778369929e-05, "loss": 0.935, "step": 13070 }, { "epoch": 0.07, "learning_rate": 1.2562325902246145e-05, "loss": 0.8986, "step": 13071 }, { "epoch": 0.07, "learning_rate": 1.2561283991023183e-05, "loss": 0.8662, "step": 13072 }, { "epoch": 0.07, "learning_rate": 1.2560242050042504e-05, "loss": 0.9111, "step": 13073 }, { "epoch": 0.07, "learning_rate": 1.2559200079316221e-05, "loss": 0.8864, "step": 13074 }, { "epoch": 0.07, "learning_rate": 1.2558158078856434e-05, "loss": 0.9003, "step": 13075 }, { "epoch": 0.07, "learning_rate": 1.255711604867525e-05, "loss": 0.8588, "step": 13076 }, { "epoch": 0.07, "learning_rate": 1.2556073988784781e-05, "loss": 0.8707, "step": 13077 }, { "epoch": 0.07, "learning_rate": 1.2555031899197125e-05, "loss": 0.8766, "step": 13078 }, { "epoch": 0.07, "learning_rate": 1.2553989779924396e-05, "loss": 0.9325, "step": 13079 }, { "epoch": 0.07, "learning_rate": 1.2552947630978698e-05, "loss": 0.9385, "step": 13080 }, { "epoch": 0.07, "learning_rate": 1.2551905452372145e-05, "loss": 0.9157, "step": 13081 }, { "epoch": 0.07, "learning_rate": 1.2550863244116836e-05, "loss": 0.9385, "step": 13082 }, { "epoch": 0.07, "learning_rate": 1.2549821006224886e-05, "loss": 0.9146, "step": 13083 }, { "epoch": 0.07, "learning_rate": 1.2548778738708405e-05, "loss": 0.9087, "step": 13084 }, { "epoch": 0.07, "learning_rate": 1.2547736441579494e-05, "loss": 0.9328, "step": 13085 }, { "epoch": 0.07, "learning_rate": 1.2546694114850275e-05, "loss": 0.9112, "step": 13086 }, { "epoch": 0.07, "learning_rate": 1.2545651758532851e-05, "loss": 0.9212, "step": 13087 }, { "epoch": 0.07, "learning_rate": 1.2544609372639331e-05, "loss": 0.9497, "step": 13088 }, { "epoch": 0.07, "learning_rate": 1.254356695718183e-05, "loss": 0.9223, "step": 13089 }, { "epoch": 0.07, "learning_rate": 1.2542524512172454e-05, "loss": 0.8489, "step": 13090 }, { "epoch": 0.07, "learning_rate": 1.2541482037623319e-05, "loss": 0.9392, "step": 13091 }, { "epoch": 0.07, "learning_rate": 1.2540439533546534e-05, "loss": 0.9149, "step": 13092 }, { "epoch": 0.07, "learning_rate": 1.2539396999954214e-05, "loss": 0.9111, "step": 13093 }, { "epoch": 0.07, "learning_rate": 1.2538354436858467e-05, "loss": 0.9153, "step": 13094 }, { "epoch": 0.07, "learning_rate": 1.2537311844271411e-05, "loss": 0.9118, "step": 13095 }, { "epoch": 0.07, "learning_rate": 1.2536269222205156e-05, "loss": 0.8974, "step": 13096 }, { "epoch": 0.07, "learning_rate": 1.2535226570671815e-05, "loss": 0.9333, "step": 13097 }, { "epoch": 0.07, "learning_rate": 1.2534183889683504e-05, "loss": 0.8748, "step": 13098 }, { "epoch": 0.07, "learning_rate": 1.2533141179252335e-05, "loss": 0.9251, "step": 13099 }, { "epoch": 0.07, "learning_rate": 1.2532098439390423e-05, "loss": 0.9096, "step": 13100 }, { "epoch": 0.07, "learning_rate": 1.2531055670109884e-05, "loss": 0.9429, "step": 13101 }, { "epoch": 0.07, "learning_rate": 1.2530012871422834e-05, "loss": 0.9063, "step": 13102 }, { "epoch": 0.07, "learning_rate": 1.2528970043341386e-05, "loss": 0.9284, "step": 13103 }, { "epoch": 0.07, "learning_rate": 1.2527927185877658e-05, "loss": 0.8615, "step": 13104 }, { "epoch": 0.07, "learning_rate": 1.2526884299043764e-05, "loss": 0.9343, "step": 13105 }, { "epoch": 0.07, "learning_rate": 1.252584138285182e-05, "loss": 0.9168, "step": 13106 }, { "epoch": 0.07, "learning_rate": 1.2524798437313948e-05, "loss": 0.9337, "step": 13107 }, { "epoch": 0.07, "learning_rate": 1.252375546244226e-05, "loss": 0.9187, "step": 13108 }, { "epoch": 0.07, "learning_rate": 1.2522712458248874e-05, "loss": 0.9236, "step": 13109 }, { "epoch": 0.07, "learning_rate": 1.252166942474591e-05, "loss": 0.9385, "step": 13110 }, { "epoch": 0.07, "learning_rate": 1.2520626361945487e-05, "loss": 0.9156, "step": 13111 }, { "epoch": 0.07, "learning_rate": 1.2519583269859723e-05, "loss": 0.8719, "step": 13112 }, { "epoch": 0.07, "learning_rate": 1.2518540148500734e-05, "loss": 0.9276, "step": 13113 }, { "epoch": 0.07, "learning_rate": 1.2517496997880642e-05, "loss": 0.9024, "step": 13114 }, { "epoch": 0.07, "learning_rate": 1.2516453818011567e-05, "loss": 0.9406, "step": 13115 }, { "epoch": 0.07, "learning_rate": 1.2515410608905626e-05, "loss": 0.9016, "step": 13116 }, { "epoch": 0.07, "learning_rate": 1.2514367370574942e-05, "loss": 0.9364, "step": 13117 }, { "epoch": 0.07, "learning_rate": 1.2513324103031636e-05, "loss": 0.9553, "step": 13118 }, { "epoch": 0.07, "learning_rate": 1.2512280806287824e-05, "loss": 0.9217, "step": 13119 }, { "epoch": 0.07, "learning_rate": 1.2511237480355636e-05, "loss": 0.8655, "step": 13120 }, { "epoch": 0.07, "learning_rate": 1.2510194125247185e-05, "loss": 0.9193, "step": 13121 }, { "epoch": 0.07, "learning_rate": 1.2509150740974599e-05, "loss": 0.8832, "step": 13122 }, { "epoch": 0.07, "learning_rate": 1.250810732755e-05, "loss": 0.9277, "step": 13123 }, { "epoch": 0.07, "learning_rate": 1.2507063884985507e-05, "loss": 0.89, "step": 13124 }, { "epoch": 0.07, "learning_rate": 1.2506020413293243e-05, "loss": 0.9002, "step": 13125 }, { "epoch": 0.07, "learning_rate": 1.2504976912485334e-05, "loss": 0.9238, "step": 13126 }, { "epoch": 0.07, "learning_rate": 1.2503933382573905e-05, "loss": 0.8947, "step": 13127 }, { "epoch": 0.07, "learning_rate": 1.2502889823571077e-05, "loss": 0.894, "step": 13128 }, { "epoch": 0.07, "learning_rate": 1.2501846235488973e-05, "loss": 0.9457, "step": 13129 }, { "epoch": 0.07, "learning_rate": 1.2500802618339723e-05, "loss": 0.9455, "step": 13130 }, { "epoch": 0.07, "learning_rate": 1.2499758972135445e-05, "loss": 0.8754, "step": 13131 }, { "epoch": 0.07, "learning_rate": 1.249871529688827e-05, "loss": 0.8826, "step": 13132 }, { "epoch": 0.07, "learning_rate": 1.2497671592610323e-05, "loss": 0.9454, "step": 13133 }, { "epoch": 0.07, "learning_rate": 1.2496627859313728e-05, "loss": 0.899, "step": 13134 }, { "epoch": 0.07, "learning_rate": 1.2495584097010616e-05, "loss": 0.8949, "step": 13135 }, { "epoch": 0.07, "learning_rate": 1.2494540305713105e-05, "loss": 0.9169, "step": 13136 }, { "epoch": 0.07, "learning_rate": 1.2493496485433329e-05, "loss": 0.9191, "step": 13137 }, { "epoch": 0.07, "learning_rate": 1.2492452636183416e-05, "loss": 0.8858, "step": 13138 }, { "epoch": 0.07, "learning_rate": 1.2491408757975489e-05, "loss": 0.9083, "step": 13139 }, { "epoch": 0.07, "learning_rate": 1.249036485082168e-05, "loss": 0.8923, "step": 13140 }, { "epoch": 0.07, "learning_rate": 1.2489320914734116e-05, "loss": 0.9144, "step": 13141 }, { "epoch": 0.07, "learning_rate": 1.2488276949724923e-05, "loss": 0.9176, "step": 13142 }, { "epoch": 0.07, "learning_rate": 1.2487232955806234e-05, "loss": 0.9153, "step": 13143 }, { "epoch": 0.07, "learning_rate": 1.2486188932990179e-05, "loss": 0.8879, "step": 13144 }, { "epoch": 0.07, "learning_rate": 1.2485144881288881e-05, "loss": 0.9385, "step": 13145 }, { "epoch": 0.07, "learning_rate": 1.248410080071448e-05, "loss": 0.8959, "step": 13146 }, { "epoch": 0.07, "learning_rate": 1.2483056691279098e-05, "loss": 0.9079, "step": 13147 }, { "epoch": 0.07, "learning_rate": 1.2482012552994873e-05, "loss": 0.8798, "step": 13148 }, { "epoch": 0.07, "learning_rate": 1.2480968385873928e-05, "loss": 0.9316, "step": 13149 }, { "epoch": 0.07, "learning_rate": 1.24799241899284e-05, "loss": 0.882, "step": 13150 }, { "epoch": 0.07, "learning_rate": 1.2478879965170422e-05, "loss": 0.9214, "step": 13151 }, { "epoch": 0.07, "learning_rate": 1.2477835711612118e-05, "loss": 0.9173, "step": 13152 }, { "epoch": 0.07, "learning_rate": 1.2476791429265631e-05, "loss": 0.8992, "step": 13153 }, { "epoch": 0.07, "learning_rate": 1.2475747118143087e-05, "loss": 0.8944, "step": 13154 }, { "epoch": 0.07, "learning_rate": 1.247470277825662e-05, "loss": 0.8883, "step": 13155 }, { "epoch": 0.07, "learning_rate": 1.2473658409618363e-05, "loss": 0.8879, "step": 13156 }, { "epoch": 0.07, "learning_rate": 1.2472614012240456e-05, "loss": 0.8945, "step": 13157 }, { "epoch": 0.07, "learning_rate": 1.2471569586135024e-05, "loss": 0.8867, "step": 13158 }, { "epoch": 0.07, "learning_rate": 1.2470525131314207e-05, "loss": 0.8906, "step": 13159 }, { "epoch": 0.07, "learning_rate": 1.2469480647790137e-05, "loss": 0.8754, "step": 13160 }, { "epoch": 0.07, "learning_rate": 1.2468436135574951e-05, "loss": 0.9386, "step": 13161 }, { "epoch": 0.07, "learning_rate": 1.2467391594680782e-05, "loss": 0.8911, "step": 13162 }, { "epoch": 0.07, "learning_rate": 1.246634702511977e-05, "loss": 0.9284, "step": 13163 }, { "epoch": 0.07, "learning_rate": 1.246530242690405e-05, "loss": 0.9302, "step": 13164 }, { "epoch": 0.07, "learning_rate": 1.2464257800045752e-05, "loss": 0.9237, "step": 13165 }, { "epoch": 0.07, "learning_rate": 1.246321314455702e-05, "loss": 0.8964, "step": 13166 }, { "epoch": 0.07, "learning_rate": 1.246216846044999e-05, "loss": 0.9254, "step": 13167 }, { "epoch": 0.07, "learning_rate": 1.2461123747736799e-05, "loss": 0.899, "step": 13168 }, { "epoch": 0.07, "learning_rate": 1.2460079006429584e-05, "loss": 0.9238, "step": 13169 }, { "epoch": 0.07, "learning_rate": 1.2459034236540481e-05, "loss": 0.9232, "step": 13170 }, { "epoch": 0.07, "learning_rate": 1.2457989438081632e-05, "loss": 0.9351, "step": 13171 }, { "epoch": 0.07, "learning_rate": 1.2456944611065175e-05, "loss": 0.8977, "step": 13172 }, { "epoch": 0.07, "learning_rate": 1.245589975550325e-05, "loss": 0.9002, "step": 13173 }, { "epoch": 0.07, "learning_rate": 1.2454854871407993e-05, "loss": 0.892, "step": 13174 }, { "epoch": 0.07, "learning_rate": 1.2453809958791546e-05, "loss": 0.9281, "step": 13175 }, { "epoch": 0.07, "learning_rate": 1.2452765017666051e-05, "loss": 0.88, "step": 13176 }, { "epoch": 0.07, "learning_rate": 1.245172004804364e-05, "loss": 0.9066, "step": 13177 }, { "epoch": 0.07, "learning_rate": 1.2450675049936467e-05, "loss": 0.8845, "step": 13178 }, { "epoch": 0.07, "learning_rate": 1.2449630023356663e-05, "loss": 0.9022, "step": 13179 }, { "epoch": 0.07, "learning_rate": 1.244858496831637e-05, "loss": 0.9322, "step": 13180 }, { "epoch": 0.07, "learning_rate": 1.2447539884827738e-05, "loss": 0.9296, "step": 13181 }, { "epoch": 0.07, "learning_rate": 1.24464947729029e-05, "loss": 0.9219, "step": 13182 }, { "epoch": 0.07, "learning_rate": 1.2445449632554e-05, "loss": 0.9623, "step": 13183 }, { "epoch": 0.07, "learning_rate": 1.2444404463793185e-05, "loss": 0.8737, "step": 13184 }, { "epoch": 0.07, "learning_rate": 1.2443359266632595e-05, "loss": 0.9009, "step": 13185 }, { "epoch": 0.07, "learning_rate": 1.2442314041084372e-05, "loss": 0.9071, "step": 13186 }, { "epoch": 0.07, "learning_rate": 1.2441268787160664e-05, "loss": 0.8849, "step": 13187 }, { "epoch": 0.07, "learning_rate": 1.2440223504873612e-05, "loss": 0.918, "step": 13188 }, { "epoch": 0.07, "learning_rate": 1.2439178194235359e-05, "loss": 0.8974, "step": 13189 }, { "epoch": 0.07, "learning_rate": 1.2438132855258055e-05, "loss": 0.8848, "step": 13190 }, { "epoch": 0.07, "learning_rate": 1.243708748795384e-05, "loss": 0.9112, "step": 13191 }, { "epoch": 0.07, "learning_rate": 1.243604209233486e-05, "loss": 0.8848, "step": 13192 }, { "epoch": 0.07, "learning_rate": 1.2434996668413265e-05, "loss": 0.8912, "step": 13193 }, { "epoch": 0.07, "learning_rate": 1.2433951216201197e-05, "loss": 0.8976, "step": 13194 }, { "epoch": 0.07, "learning_rate": 1.2432905735710801e-05, "loss": 0.8901, "step": 13195 }, { "epoch": 0.07, "learning_rate": 1.2431860226954227e-05, "loss": 0.903, "step": 13196 }, { "epoch": 0.07, "learning_rate": 1.2430814689943623e-05, "loss": 0.9583, "step": 13197 }, { "epoch": 0.07, "learning_rate": 1.2429769124691128e-05, "loss": 0.9011, "step": 13198 }, { "epoch": 0.07, "learning_rate": 1.2428723531208905e-05, "loss": 0.8975, "step": 13199 }, { "epoch": 0.07, "learning_rate": 1.2427677909509086e-05, "loss": 0.9041, "step": 13200 }, { "epoch": 0.07, "learning_rate": 1.242663225960383e-05, "loss": 0.8914, "step": 13201 }, { "epoch": 0.07, "learning_rate": 1.242558658150528e-05, "loss": 0.8852, "step": 13202 }, { "epoch": 0.07, "learning_rate": 1.2424540875225587e-05, "loss": 0.899, "step": 13203 }, { "epoch": 0.07, "learning_rate": 1.2423495140776901e-05, "loss": 0.9089, "step": 13204 }, { "epoch": 0.07, "learning_rate": 1.242244937817137e-05, "loss": 0.9276, "step": 13205 }, { "epoch": 0.07, "learning_rate": 1.2421403587421148e-05, "loss": 0.9017, "step": 13206 }, { "epoch": 0.07, "learning_rate": 1.2420357768538378e-05, "loss": 0.8774, "step": 13207 }, { "epoch": 0.07, "learning_rate": 1.2419311921535217e-05, "loss": 0.8907, "step": 13208 }, { "epoch": 0.07, "learning_rate": 1.2418266046423812e-05, "loss": 0.9203, "step": 13209 }, { "epoch": 0.07, "learning_rate": 1.2417220143216315e-05, "loss": 0.8984, "step": 13210 }, { "epoch": 0.07, "learning_rate": 1.241617421192488e-05, "loss": 0.8915, "step": 13211 }, { "epoch": 0.07, "learning_rate": 1.241512825256166e-05, "loss": 0.8971, "step": 13212 }, { "epoch": 0.07, "learning_rate": 1.2414082265138803e-05, "loss": 0.8966, "step": 13213 }, { "epoch": 0.07, "learning_rate": 1.2413036249668462e-05, "loss": 0.9237, "step": 13214 }, { "epoch": 0.07, "learning_rate": 1.2411990206162791e-05, "loss": 0.9218, "step": 13215 }, { "epoch": 0.07, "learning_rate": 1.2410944134633944e-05, "loss": 0.9231, "step": 13216 }, { "epoch": 0.07, "learning_rate": 1.2409898035094075e-05, "loss": 0.9012, "step": 13217 }, { "epoch": 0.07, "learning_rate": 1.2408851907555335e-05, "loss": 0.9037, "step": 13218 }, { "epoch": 0.07, "learning_rate": 1.2407805752029883e-05, "loss": 0.896, "step": 13219 }, { "epoch": 0.07, "learning_rate": 1.2406759568529867e-05, "loss": 0.9221, "step": 13220 }, { "epoch": 0.07, "learning_rate": 1.2405713357067448e-05, "loss": 0.943, "step": 13221 }, { "epoch": 0.07, "learning_rate": 1.2404667117654779e-05, "loss": 0.907, "step": 13222 }, { "epoch": 0.07, "learning_rate": 1.240362085030401e-05, "loss": 0.907, "step": 13223 }, { "epoch": 0.07, "learning_rate": 1.2402574555027306e-05, "loss": 0.9132, "step": 13224 }, { "epoch": 0.07, "learning_rate": 1.2401528231836819e-05, "loss": 0.9092, "step": 13225 }, { "epoch": 0.07, "learning_rate": 1.2400481880744705e-05, "loss": 0.8699, "step": 13226 }, { "epoch": 0.07, "learning_rate": 1.2399435501763123e-05, "loss": 0.9079, "step": 13227 }, { "epoch": 0.07, "learning_rate": 1.2398389094904226e-05, "loss": 0.8793, "step": 13228 }, { "epoch": 0.07, "learning_rate": 1.2397342660180173e-05, "loss": 0.9239, "step": 13229 }, { "epoch": 0.07, "learning_rate": 1.2396296197603126e-05, "loss": 0.9373, "step": 13230 }, { "epoch": 0.07, "learning_rate": 1.239524970718524e-05, "loss": 0.8872, "step": 13231 }, { "epoch": 0.07, "learning_rate": 1.239420318893867e-05, "loss": 0.928, "step": 13232 }, { "epoch": 0.07, "learning_rate": 1.2393156642875579e-05, "loss": 0.9327, "step": 13233 }, { "epoch": 0.07, "learning_rate": 1.2392110069008125e-05, "loss": 0.8646, "step": 13234 }, { "epoch": 0.07, "learning_rate": 1.2391063467348468e-05, "loss": 0.8708, "step": 13235 }, { "epoch": 0.07, "learning_rate": 1.2390016837908766e-05, "loss": 0.8779, "step": 13236 }, { "epoch": 0.07, "learning_rate": 1.2388970180701183e-05, "loss": 0.8886, "step": 13237 }, { "epoch": 0.07, "learning_rate": 1.2387923495737872e-05, "loss": 0.9278, "step": 13238 }, { "epoch": 0.07, "learning_rate": 1.2386876783031002e-05, "loss": 0.8993, "step": 13239 }, { "epoch": 0.07, "learning_rate": 1.238583004259273e-05, "loss": 0.912, "step": 13240 }, { "epoch": 0.07, "learning_rate": 1.2384783274435213e-05, "loss": 0.9306, "step": 13241 }, { "epoch": 0.07, "learning_rate": 1.2383736478570623e-05, "loss": 0.9328, "step": 13242 }, { "epoch": 0.07, "learning_rate": 1.2382689655011113e-05, "loss": 0.9425, "step": 13243 }, { "epoch": 0.07, "learning_rate": 1.2381642803768847e-05, "loss": 0.9236, "step": 13244 }, { "epoch": 0.07, "learning_rate": 1.2380595924855994e-05, "loss": 0.9172, "step": 13245 }, { "epoch": 0.07, "learning_rate": 1.2379549018284707e-05, "loss": 0.9025, "step": 13246 }, { "epoch": 0.07, "learning_rate": 1.2378502084067155e-05, "loss": 0.9001, "step": 13247 }, { "epoch": 0.07, "learning_rate": 1.23774551222155e-05, "loss": 0.9406, "step": 13248 }, { "epoch": 0.07, "learning_rate": 1.237640813274191e-05, "loss": 0.9272, "step": 13249 }, { "epoch": 0.07, "learning_rate": 1.2375361115658544e-05, "loss": 0.9221, "step": 13250 }, { "epoch": 0.07, "learning_rate": 1.2374314070977569e-05, "loss": 0.8948, "step": 13251 }, { "epoch": 0.07, "learning_rate": 1.2373266998711152e-05, "loss": 0.8995, "step": 13252 }, { "epoch": 0.07, "learning_rate": 1.2372219898871452e-05, "loss": 0.9229, "step": 13253 }, { "epoch": 0.08, "learning_rate": 1.2371172771470637e-05, "loss": 0.8924, "step": 13254 }, { "epoch": 0.08, "learning_rate": 1.2370125616520875e-05, "loss": 0.9131, "step": 13255 }, { "epoch": 0.08, "learning_rate": 1.2369078434034329e-05, "loss": 0.9083, "step": 13256 }, { "epoch": 0.08, "learning_rate": 1.236803122402317e-05, "loss": 0.9113, "step": 13257 }, { "epoch": 0.08, "learning_rate": 1.2366983986499562e-05, "loss": 0.9046, "step": 13258 }, { "epoch": 0.08, "learning_rate": 1.2365936721475671e-05, "loss": 0.9566, "step": 13259 }, { "epoch": 0.08, "learning_rate": 1.2364889428963667e-05, "loss": 0.9222, "step": 13260 }, { "epoch": 0.08, "learning_rate": 1.2363842108975717e-05, "loss": 0.8932, "step": 13261 }, { "epoch": 0.08, "learning_rate": 1.2362794761523987e-05, "loss": 0.8954, "step": 13262 }, { "epoch": 0.08, "learning_rate": 1.2361747386620646e-05, "loss": 0.9255, "step": 13263 }, { "epoch": 0.08, "learning_rate": 1.2360699984277864e-05, "loss": 0.8793, "step": 13264 }, { "epoch": 0.08, "learning_rate": 1.235965255450781e-05, "loss": 0.9057, "step": 13265 }, { "epoch": 0.08, "learning_rate": 1.2358605097322653e-05, "loss": 0.9128, "step": 13266 }, { "epoch": 0.08, "learning_rate": 1.2357557612734561e-05, "loss": 0.9318, "step": 13267 }, { "epoch": 0.08, "learning_rate": 1.235651010075571e-05, "loss": 0.8992, "step": 13268 }, { "epoch": 0.08, "learning_rate": 1.2355462561398259e-05, "loss": 0.9005, "step": 13269 }, { "epoch": 0.08, "learning_rate": 1.2354414994674391e-05, "loss": 0.9024, "step": 13270 }, { "epoch": 0.08, "learning_rate": 1.2353367400596268e-05, "loss": 0.9268, "step": 13271 }, { "epoch": 0.08, "learning_rate": 1.2352319779176064e-05, "loss": 0.9236, "step": 13272 }, { "epoch": 0.08, "learning_rate": 1.2351272130425955e-05, "loss": 0.878, "step": 13273 }, { "epoch": 0.08, "learning_rate": 1.2350224454358109e-05, "loss": 0.9227, "step": 13274 }, { "epoch": 0.08, "learning_rate": 1.2349176750984694e-05, "loss": 0.8965, "step": 13275 }, { "epoch": 0.08, "learning_rate": 1.234812902031789e-05, "loss": 0.8731, "step": 13276 }, { "epoch": 0.08, "learning_rate": 1.2347081262369867e-05, "loss": 0.9119, "step": 13277 }, { "epoch": 0.08, "learning_rate": 1.2346033477152792e-05, "loss": 0.9183, "step": 13278 }, { "epoch": 0.08, "learning_rate": 1.234498566467885e-05, "loss": 0.9396, "step": 13279 }, { "epoch": 0.08, "learning_rate": 1.2343937824960206e-05, "loss": 0.8891, "step": 13280 }, { "epoch": 0.08, "learning_rate": 1.2342889958009037e-05, "loss": 0.9364, "step": 13281 }, { "epoch": 0.08, "learning_rate": 1.234184206383752e-05, "loss": 0.8701, "step": 13282 }, { "epoch": 0.08, "learning_rate": 1.2340794142457824e-05, "loss": 0.9554, "step": 13283 }, { "epoch": 0.08, "learning_rate": 1.2339746193882127e-05, "loss": 0.8665, "step": 13284 }, { "epoch": 0.08, "learning_rate": 1.2338698218122607e-05, "loss": 0.9181, "step": 13285 }, { "epoch": 0.08, "learning_rate": 1.233765021519144e-05, "loss": 0.9293, "step": 13286 }, { "epoch": 0.08, "learning_rate": 1.233660218510079e-05, "loss": 0.9323, "step": 13287 }, { "epoch": 0.08, "learning_rate": 1.233555412786285e-05, "loss": 0.8937, "step": 13288 }, { "epoch": 0.08, "learning_rate": 1.2334506043489789e-05, "loss": 0.9106, "step": 13289 }, { "epoch": 0.08, "learning_rate": 1.233345793199378e-05, "loss": 0.875, "step": 13290 }, { "epoch": 0.08, "learning_rate": 1.2332409793387009e-05, "loss": 0.894, "step": 13291 }, { "epoch": 0.08, "learning_rate": 1.2331361627681645e-05, "loss": 0.9247, "step": 13292 }, { "epoch": 0.08, "learning_rate": 1.233031343488987e-05, "loss": 0.9087, "step": 13293 }, { "epoch": 0.08, "learning_rate": 1.2329265215023862e-05, "loss": 0.9148, "step": 13294 }, { "epoch": 0.08, "learning_rate": 1.2328216968095803e-05, "loss": 0.8888, "step": 13295 }, { "epoch": 0.08, "learning_rate": 1.2327168694117864e-05, "loss": 0.9021, "step": 13296 }, { "epoch": 0.08, "learning_rate": 1.232612039310223e-05, "loss": 0.9036, "step": 13297 }, { "epoch": 0.08, "learning_rate": 1.2325072065061082e-05, "loss": 0.9185, "step": 13298 }, { "epoch": 0.08, "learning_rate": 1.232402371000659e-05, "loss": 0.9167, "step": 13299 }, { "epoch": 0.08, "learning_rate": 1.2322975327950946e-05, "loss": 0.8858, "step": 13300 }, { "epoch": 0.08, "learning_rate": 1.2321926918906325e-05, "loss": 0.9302, "step": 13301 }, { "epoch": 0.08, "learning_rate": 1.2320878482884904e-05, "loss": 0.8761, "step": 13302 }, { "epoch": 0.08, "learning_rate": 1.2319830019898871e-05, "loss": 0.9121, "step": 13303 }, { "epoch": 0.08, "learning_rate": 1.2318781529960405e-05, "loss": 0.8853, "step": 13304 }, { "epoch": 0.08, "learning_rate": 1.2317733013081683e-05, "loss": 0.9069, "step": 13305 }, { "epoch": 0.08, "learning_rate": 1.2316684469274894e-05, "loss": 0.8922, "step": 13306 }, { "epoch": 0.08, "learning_rate": 1.2315635898552215e-05, "loss": 0.9424, "step": 13307 }, { "epoch": 0.08, "learning_rate": 1.2314587300925834e-05, "loss": 0.9121, "step": 13308 }, { "epoch": 0.08, "learning_rate": 1.2313538676407929e-05, "loss": 0.9253, "step": 13309 }, { "epoch": 0.08, "learning_rate": 1.2312490025010681e-05, "loss": 0.9171, "step": 13310 }, { "epoch": 0.08, "learning_rate": 1.2311441346746284e-05, "loss": 0.939, "step": 13311 }, { "epoch": 0.08, "learning_rate": 1.2310392641626912e-05, "loss": 0.9105, "step": 13312 }, { "epoch": 0.08, "learning_rate": 1.230934390966475e-05, "loss": 0.9317, "step": 13313 }, { "epoch": 0.08, "learning_rate": 1.2308295150871987e-05, "loss": 0.919, "step": 13314 }, { "epoch": 0.08, "learning_rate": 1.2307246365260804e-05, "loss": 0.8917, "step": 13315 }, { "epoch": 0.08, "learning_rate": 1.2306197552843391e-05, "loss": 0.9138, "step": 13316 }, { "epoch": 0.08, "learning_rate": 1.2305148713631927e-05, "loss": 0.9172, "step": 13317 }, { "epoch": 0.08, "learning_rate": 1.2304099847638601e-05, "loss": 0.9325, "step": 13318 }, { "epoch": 0.08, "learning_rate": 1.2303050954875602e-05, "loss": 0.9079, "step": 13319 }, { "epoch": 0.08, "learning_rate": 1.230200203535511e-05, "loss": 0.8482, "step": 13320 }, { "epoch": 0.08, "learning_rate": 1.2300953089089314e-05, "loss": 0.935, "step": 13321 }, { "epoch": 0.08, "learning_rate": 1.2299904116090402e-05, "loss": 0.8966, "step": 13322 }, { "epoch": 0.08, "learning_rate": 1.2298855116370565e-05, "loss": 0.9035, "step": 13323 }, { "epoch": 0.08, "learning_rate": 1.229780608994198e-05, "loss": 0.9058, "step": 13324 }, { "epoch": 0.08, "learning_rate": 1.2296757036816846e-05, "loss": 0.9009, "step": 13325 }, { "epoch": 0.08, "learning_rate": 1.2295707957007345e-05, "loss": 0.917, "step": 13326 }, { "epoch": 0.08, "learning_rate": 1.2294658850525667e-05, "loss": 0.9314, "step": 13327 }, { "epoch": 0.08, "learning_rate": 1.2293609717384002e-05, "loss": 0.9114, "step": 13328 }, { "epoch": 0.08, "learning_rate": 1.2292560557594537e-05, "loss": 0.935, "step": 13329 }, { "epoch": 0.08, "learning_rate": 1.2291511371169462e-05, "loss": 0.923, "step": 13330 }, { "epoch": 0.08, "learning_rate": 1.2290462158120968e-05, "loss": 0.9253, "step": 13331 }, { "epoch": 0.08, "learning_rate": 1.2289412918461246e-05, "loss": 0.888, "step": 13332 }, { "epoch": 0.08, "learning_rate": 1.2288363652202479e-05, "loss": 0.9262, "step": 13333 }, { "epoch": 0.08, "learning_rate": 1.228731435935687e-05, "loss": 0.902, "step": 13334 }, { "epoch": 0.08, "learning_rate": 1.22862650399366e-05, "loss": 0.9308, "step": 13335 }, { "epoch": 0.08, "learning_rate": 1.2285215693953863e-05, "loss": 0.8869, "step": 13336 }, { "epoch": 0.08, "learning_rate": 1.2284166321420852e-05, "loss": 0.9168, "step": 13337 }, { "epoch": 0.08, "learning_rate": 1.2283116922349758e-05, "loss": 0.926, "step": 13338 }, { "epoch": 0.08, "learning_rate": 1.2282067496752772e-05, "loss": 0.9458, "step": 13339 }, { "epoch": 0.08, "learning_rate": 1.2281018044642088e-05, "loss": 0.8895, "step": 13340 }, { "epoch": 0.08, "learning_rate": 1.2279968566029901e-05, "loss": 0.8809, "step": 13341 }, { "epoch": 0.08, "learning_rate": 1.22789190609284e-05, "loss": 0.9475, "step": 13342 }, { "epoch": 0.08, "learning_rate": 1.227786952934978e-05, "loss": 0.9009, "step": 13343 }, { "epoch": 0.08, "learning_rate": 1.2276819971306238e-05, "loss": 0.9484, "step": 13344 }, { "epoch": 0.08, "learning_rate": 1.2275770386809957e-05, "loss": 0.892, "step": 13345 }, { "epoch": 0.08, "learning_rate": 1.2274720775873148e-05, "loss": 0.9365, "step": 13346 }, { "epoch": 0.08, "learning_rate": 1.2273671138507993e-05, "loss": 0.9349, "step": 13347 }, { "epoch": 0.08, "learning_rate": 1.2272621474726692e-05, "loss": 0.9039, "step": 13348 }, { "epoch": 0.08, "learning_rate": 1.2271571784541437e-05, "loss": 0.9422, "step": 13349 }, { "epoch": 0.08, "learning_rate": 1.227052206796443e-05, "loss": 0.9116, "step": 13350 }, { "epoch": 0.08, "learning_rate": 1.2269472325007858e-05, "loss": 0.9216, "step": 13351 }, { "epoch": 0.08, "learning_rate": 1.2268422555683925e-05, "loss": 0.9329, "step": 13352 }, { "epoch": 0.08, "learning_rate": 1.2267372760004825e-05, "loss": 0.915, "step": 13353 }, { "epoch": 0.08, "learning_rate": 1.2266322937982752e-05, "loss": 0.9159, "step": 13354 }, { "epoch": 0.08, "learning_rate": 1.2265273089629908e-05, "loss": 0.8959, "step": 13355 }, { "epoch": 0.08, "learning_rate": 1.2264223214958486e-05, "loss": 0.9106, "step": 13356 }, { "epoch": 0.08, "learning_rate": 1.2263173313980688e-05, "loss": 0.9259, "step": 13357 }, { "epoch": 0.08, "learning_rate": 1.2262123386708706e-05, "loss": 0.919, "step": 13358 }, { "epoch": 0.08, "learning_rate": 1.2261073433154744e-05, "loss": 0.9747, "step": 13359 }, { "epoch": 0.08, "learning_rate": 1.2260023453330997e-05, "loss": 0.8932, "step": 13360 }, { "epoch": 0.08, "learning_rate": 1.2258973447249666e-05, "loss": 0.8946, "step": 13361 }, { "epoch": 0.08, "learning_rate": 1.2257923414922955e-05, "loss": 0.9225, "step": 13362 }, { "epoch": 0.08, "learning_rate": 1.2256873356363053e-05, "loss": 0.913, "step": 13363 }, { "epoch": 0.08, "learning_rate": 1.2255823271582166e-05, "loss": 0.9434, "step": 13364 }, { "epoch": 0.08, "learning_rate": 1.2254773160592494e-05, "loss": 0.92, "step": 13365 }, { "epoch": 0.08, "learning_rate": 1.2253723023406238e-05, "loss": 0.9358, "step": 13366 }, { "epoch": 0.08, "learning_rate": 1.2252672860035596e-05, "loss": 0.8854, "step": 13367 }, { "epoch": 0.08, "learning_rate": 1.2251622670492772e-05, "loss": 0.886, "step": 13368 }, { "epoch": 0.08, "learning_rate": 1.2250572454789968e-05, "loss": 0.9245, "step": 13369 }, { "epoch": 0.08, "learning_rate": 1.224952221293938e-05, "loss": 0.9223, "step": 13370 }, { "epoch": 0.08, "learning_rate": 1.2248471944953217e-05, "loss": 0.9337, "step": 13371 }, { "epoch": 0.08, "learning_rate": 1.2247421650843676e-05, "loss": 0.8794, "step": 13372 }, { "epoch": 0.08, "learning_rate": 1.2246371330622964e-05, "loss": 0.9141, "step": 13373 }, { "epoch": 0.08, "learning_rate": 1.2245320984303282e-05, "loss": 0.8781, "step": 13374 }, { "epoch": 0.08, "learning_rate": 1.2244270611896832e-05, "loss": 0.8916, "step": 13375 }, { "epoch": 0.08, "learning_rate": 1.2243220213415816e-05, "loss": 0.8909, "step": 13376 }, { "epoch": 0.08, "learning_rate": 1.2242169788872444e-05, "loss": 0.8887, "step": 13377 }, { "epoch": 0.08, "learning_rate": 1.2241119338278917e-05, "loss": 0.9078, "step": 13378 }, { "epoch": 0.08, "learning_rate": 1.2240068861647434e-05, "loss": 0.8904, "step": 13379 }, { "epoch": 0.08, "learning_rate": 1.2239018358990209e-05, "loss": 0.9146, "step": 13380 }, { "epoch": 0.08, "learning_rate": 1.2237967830319444e-05, "loss": 0.9063, "step": 13381 }, { "epoch": 0.08, "learning_rate": 1.2236917275647338e-05, "loss": 0.8909, "step": 13382 }, { "epoch": 0.08, "learning_rate": 1.2235866694986104e-05, "loss": 0.9324, "step": 13383 }, { "epoch": 0.08, "learning_rate": 1.2234816088347948e-05, "loss": 0.9288, "step": 13384 }, { "epoch": 0.08, "learning_rate": 1.223376545574507e-05, "loss": 0.9322, "step": 13385 }, { "epoch": 0.08, "learning_rate": 1.2232714797189682e-05, "loss": 0.9101, "step": 13386 }, { "epoch": 0.08, "learning_rate": 1.223166411269399e-05, "loss": 0.9731, "step": 13387 }, { "epoch": 0.08, "learning_rate": 1.2230613402270199e-05, "loss": 0.9259, "step": 13388 }, { "epoch": 0.08, "learning_rate": 1.2229562665930517e-05, "loss": 0.8701, "step": 13389 }, { "epoch": 0.08, "learning_rate": 1.2228511903687158e-05, "loss": 0.8845, "step": 13390 }, { "epoch": 0.08, "learning_rate": 1.2227461115552317e-05, "loss": 0.9459, "step": 13391 }, { "epoch": 0.08, "learning_rate": 1.2226410301538214e-05, "loss": 0.8459, "step": 13392 }, { "epoch": 0.08, "learning_rate": 1.2225359461657054e-05, "loss": 0.9229, "step": 13393 }, { "epoch": 0.08, "learning_rate": 1.2224308595921044e-05, "loss": 0.9064, "step": 13394 }, { "epoch": 0.08, "learning_rate": 1.2223257704342395e-05, "loss": 0.9233, "step": 13395 }, { "epoch": 0.08, "learning_rate": 1.2222206786933318e-05, "loss": 0.8824, "step": 13396 }, { "epoch": 0.08, "learning_rate": 1.2221155843706018e-05, "loss": 0.9092, "step": 13397 }, { "epoch": 0.08, "learning_rate": 1.222010487467271e-05, "loss": 0.8906, "step": 13398 }, { "epoch": 0.08, "learning_rate": 1.2219053879845605e-05, "loss": 0.9104, "step": 13399 }, { "epoch": 0.08, "learning_rate": 1.2218002859236911e-05, "loss": 0.8977, "step": 13400 }, { "epoch": 0.08, "learning_rate": 1.2216951812858837e-05, "loss": 0.9385, "step": 13401 }, { "epoch": 0.08, "learning_rate": 1.2215900740723598e-05, "loss": 0.9133, "step": 13402 }, { "epoch": 0.08, "learning_rate": 1.2214849642843405e-05, "loss": 0.9211, "step": 13403 }, { "epoch": 0.08, "learning_rate": 1.2213798519230471e-05, "loss": 0.8976, "step": 13404 }, { "epoch": 0.08, "learning_rate": 1.2212747369897004e-05, "loss": 0.907, "step": 13405 }, { "epoch": 0.08, "learning_rate": 1.2211696194855221e-05, "loss": 0.8872, "step": 13406 }, { "epoch": 0.08, "learning_rate": 1.2210644994117334e-05, "loss": 0.9227, "step": 13407 }, { "epoch": 0.08, "learning_rate": 1.2209593767695554e-05, "loss": 0.9175, "step": 13408 }, { "epoch": 0.08, "learning_rate": 1.2208542515602096e-05, "loss": 0.9356, "step": 13409 }, { "epoch": 0.08, "learning_rate": 1.2207491237849174e-05, "loss": 0.9009, "step": 13410 }, { "epoch": 0.08, "learning_rate": 1.2206439934449001e-05, "loss": 0.8825, "step": 13411 }, { "epoch": 0.08, "learning_rate": 1.2205388605413793e-05, "loss": 0.9098, "step": 13412 }, { "epoch": 0.08, "learning_rate": 1.2204337250755763e-05, "loss": 0.9117, "step": 13413 }, { "epoch": 0.08, "learning_rate": 1.2203285870487125e-05, "loss": 0.9004, "step": 13414 }, { "epoch": 0.08, "learning_rate": 1.22022344646201e-05, "loss": 0.8985, "step": 13415 }, { "epoch": 0.08, "learning_rate": 1.2201183033166893e-05, "loss": 0.8667, "step": 13416 }, { "epoch": 0.08, "learning_rate": 1.220013157613973e-05, "loss": 0.8682, "step": 13417 }, { "epoch": 0.08, "learning_rate": 1.2199080093550822e-05, "loss": 0.9385, "step": 13418 }, { "epoch": 0.08, "learning_rate": 1.2198028585412387e-05, "loss": 0.9168, "step": 13419 }, { "epoch": 0.08, "learning_rate": 1.2196977051736643e-05, "loss": 0.9485, "step": 13420 }, { "epoch": 0.08, "learning_rate": 1.2195925492535803e-05, "loss": 0.9113, "step": 13421 }, { "epoch": 0.08, "learning_rate": 1.2194873907822086e-05, "loss": 0.896, "step": 13422 }, { "epoch": 0.08, "learning_rate": 1.2193822297607711e-05, "loss": 0.9014, "step": 13423 }, { "epoch": 0.08, "learning_rate": 1.2192770661904899e-05, "loss": 0.9357, "step": 13424 }, { "epoch": 0.08, "learning_rate": 1.2191719000725859e-05, "loss": 0.9326, "step": 13425 }, { "epoch": 0.08, "learning_rate": 1.2190667314082818e-05, "loss": 0.9249, "step": 13426 }, { "epoch": 0.08, "learning_rate": 1.2189615601987991e-05, "loss": 0.9395, "step": 13427 }, { "epoch": 0.08, "learning_rate": 1.2188563864453595e-05, "loss": 0.9007, "step": 13428 }, { "epoch": 0.08, "learning_rate": 1.2187512101491856e-05, "loss": 0.907, "step": 13429 }, { "epoch": 0.08, "learning_rate": 1.218646031311499e-05, "loss": 0.8903, "step": 13430 }, { "epoch": 0.08, "learning_rate": 1.2185408499335213e-05, "loss": 0.9325, "step": 13431 }, { "epoch": 0.08, "learning_rate": 1.2184356660164751e-05, "loss": 0.8921, "step": 13432 }, { "epoch": 0.08, "learning_rate": 1.2183304795615825e-05, "loss": 0.9236, "step": 13433 }, { "epoch": 0.08, "learning_rate": 1.2182252905700649e-05, "loss": 0.8726, "step": 13434 }, { "epoch": 0.08, "learning_rate": 1.218120099043145e-05, "loss": 0.9106, "step": 13435 }, { "epoch": 0.08, "learning_rate": 1.2180149049820451e-05, "loss": 0.9005, "step": 13436 }, { "epoch": 0.08, "learning_rate": 1.2179097083879865e-05, "loss": 0.8995, "step": 13437 }, { "epoch": 0.08, "learning_rate": 1.2178045092621924e-05, "loss": 0.8899, "step": 13438 }, { "epoch": 0.08, "learning_rate": 1.2176993076058845e-05, "loss": 0.9025, "step": 13439 }, { "epoch": 0.08, "learning_rate": 1.2175941034202852e-05, "loss": 0.9318, "step": 13440 }, { "epoch": 0.08, "learning_rate": 1.2174888967066167e-05, "loss": 0.9554, "step": 13441 }, { "epoch": 0.08, "learning_rate": 1.2173836874661016e-05, "loss": 0.9083, "step": 13442 }, { "epoch": 0.08, "learning_rate": 1.2172784756999619e-05, "loss": 0.8745, "step": 13443 }, { "epoch": 0.08, "learning_rate": 1.2171732614094201e-05, "loss": 0.9002, "step": 13444 }, { "epoch": 0.08, "learning_rate": 1.2170680445956987e-05, "loss": 0.9524, "step": 13445 }, { "epoch": 0.08, "learning_rate": 1.2169628252600202e-05, "loss": 0.8842, "step": 13446 }, { "epoch": 0.08, "learning_rate": 1.2168576034036064e-05, "loss": 0.9136, "step": 13447 }, { "epoch": 0.08, "learning_rate": 1.2167523790276808e-05, "loss": 0.8976, "step": 13448 }, { "epoch": 0.08, "learning_rate": 1.216647152133465e-05, "loss": 0.9151, "step": 13449 }, { "epoch": 0.08, "learning_rate": 1.2165419227221827e-05, "loss": 0.9448, "step": 13450 }, { "epoch": 0.08, "learning_rate": 1.2164366907950553e-05, "loss": 0.9142, "step": 13451 }, { "epoch": 0.08, "learning_rate": 1.216331456353306e-05, "loss": 0.9186, "step": 13452 }, { "epoch": 0.08, "learning_rate": 1.2162262193981573e-05, "loss": 0.9261, "step": 13453 }, { "epoch": 0.08, "learning_rate": 1.2161209799308322e-05, "loss": 0.9399, "step": 13454 }, { "epoch": 0.08, "learning_rate": 1.216015737952553e-05, "loss": 0.9184, "step": 13455 }, { "epoch": 0.08, "learning_rate": 1.2159104934645425e-05, "loss": 0.9192, "step": 13456 }, { "epoch": 0.08, "learning_rate": 1.2158052464680237e-05, "loss": 0.9073, "step": 13457 }, { "epoch": 0.08, "learning_rate": 1.2156999969642189e-05, "loss": 0.8899, "step": 13458 }, { "epoch": 0.08, "learning_rate": 1.2155947449543513e-05, "loss": 0.9025, "step": 13459 }, { "epoch": 0.08, "learning_rate": 1.2154894904396437e-05, "loss": 0.887, "step": 13460 }, { "epoch": 0.08, "learning_rate": 1.215384233421319e-05, "loss": 0.905, "step": 13461 }, { "epoch": 0.08, "learning_rate": 1.2152789739005999e-05, "loss": 0.8998, "step": 13462 }, { "epoch": 0.08, "learning_rate": 1.21517371187871e-05, "loss": 0.905, "step": 13463 }, { "epoch": 0.08, "learning_rate": 1.2150684473568713e-05, "loss": 0.9115, "step": 13464 }, { "epoch": 0.08, "learning_rate": 1.214963180336307e-05, "loss": 0.901, "step": 13465 }, { "epoch": 0.08, "learning_rate": 1.214857910818241e-05, "loss": 0.8883, "step": 13466 }, { "epoch": 0.08, "learning_rate": 1.2147526388038952e-05, "loss": 0.8988, "step": 13467 }, { "epoch": 0.08, "learning_rate": 1.2146473642944933e-05, "loss": 0.9389, "step": 13468 }, { "epoch": 0.08, "learning_rate": 1.2145420872912586e-05, "loss": 0.9451, "step": 13469 }, { "epoch": 0.08, "learning_rate": 1.2144368077954139e-05, "loss": 0.8891, "step": 13470 }, { "epoch": 0.08, "learning_rate": 1.214331525808182e-05, "loss": 0.9165, "step": 13471 }, { "epoch": 0.08, "learning_rate": 1.2142262413307868e-05, "loss": 0.9157, "step": 13472 }, { "epoch": 0.08, "learning_rate": 1.2141209543644514e-05, "loss": 0.9264, "step": 13473 }, { "epoch": 0.08, "learning_rate": 1.2140156649103984e-05, "loss": 0.9143, "step": 13474 }, { "epoch": 0.08, "learning_rate": 1.213910372969852e-05, "loss": 0.9182, "step": 13475 }, { "epoch": 0.08, "learning_rate": 1.213805078544035e-05, "loss": 0.887, "step": 13476 }, { "epoch": 0.08, "learning_rate": 1.2136997816341707e-05, "loss": 0.8866, "step": 13477 }, { "epoch": 0.08, "learning_rate": 1.2135944822414825e-05, "loss": 0.9306, "step": 13478 }, { "epoch": 0.08, "learning_rate": 1.2134891803671942e-05, "loss": 0.9685, "step": 13479 }, { "epoch": 0.08, "learning_rate": 1.2133838760125288e-05, "loss": 0.8643, "step": 13480 }, { "epoch": 0.08, "learning_rate": 1.2132785691787096e-05, "loss": 0.9228, "step": 13481 }, { "epoch": 0.08, "learning_rate": 1.2131732598669608e-05, "loss": 0.9305, "step": 13482 }, { "epoch": 0.08, "learning_rate": 1.2130679480785049e-05, "loss": 0.924, "step": 13483 }, { "epoch": 0.08, "learning_rate": 1.2129626338145664e-05, "loss": 0.882, "step": 13484 }, { "epoch": 0.08, "learning_rate": 1.2128573170763684e-05, "loss": 0.9172, "step": 13485 }, { "epoch": 0.08, "learning_rate": 1.2127519978651346e-05, "loss": 0.9175, "step": 13486 }, { "epoch": 0.08, "learning_rate": 1.2126466761820882e-05, "loss": 0.9407, "step": 13487 }, { "epoch": 0.08, "learning_rate": 1.212541352028454e-05, "loss": 0.864, "step": 13488 }, { "epoch": 0.08, "learning_rate": 1.2124360254054545e-05, "loss": 0.9334, "step": 13489 }, { "epoch": 0.08, "learning_rate": 1.2123306963143137e-05, "loss": 0.9104, "step": 13490 }, { "epoch": 0.08, "learning_rate": 1.212225364756256e-05, "loss": 0.9078, "step": 13491 }, { "epoch": 0.08, "learning_rate": 1.2121200307325044e-05, "loss": 0.9191, "step": 13492 }, { "epoch": 0.08, "learning_rate": 1.212014694244283e-05, "loss": 0.8811, "step": 13493 }, { "epoch": 0.08, "learning_rate": 1.2119093552928154e-05, "loss": 0.9162, "step": 13494 }, { "epoch": 0.08, "learning_rate": 1.211804013879326e-05, "loss": 0.9285, "step": 13495 }, { "epoch": 0.08, "learning_rate": 1.2116986700050382e-05, "loss": 0.8484, "step": 13496 }, { "epoch": 0.08, "learning_rate": 1.211593323671176e-05, "loss": 0.9223, "step": 13497 }, { "epoch": 0.08, "learning_rate": 1.2114879748789633e-05, "loss": 0.9079, "step": 13498 }, { "epoch": 0.08, "learning_rate": 1.2113826236296245e-05, "loss": 0.9194, "step": 13499 }, { "epoch": 0.08, "learning_rate": 1.2112772699243832e-05, "loss": 0.888, "step": 13500 }, { "epoch": 0.08, "learning_rate": 1.2111719137644633e-05, "loss": 0.9093, "step": 13501 }, { "epoch": 0.08, "learning_rate": 1.2110665551510892e-05, "loss": 0.9003, "step": 13502 }, { "epoch": 0.08, "learning_rate": 1.2109611940854848e-05, "loss": 0.9085, "step": 13503 }, { "epoch": 0.08, "learning_rate": 1.2108558305688742e-05, "loss": 0.8995, "step": 13504 }, { "epoch": 0.08, "learning_rate": 1.2107504646024817e-05, "loss": 0.8706, "step": 13505 }, { "epoch": 0.08, "learning_rate": 1.2106450961875314e-05, "loss": 0.8588, "step": 13506 }, { "epoch": 0.08, "learning_rate": 1.2105397253252473e-05, "loss": 0.8989, "step": 13507 }, { "epoch": 0.08, "learning_rate": 1.210434352016854e-05, "loss": 0.9087, "step": 13508 }, { "epoch": 0.08, "learning_rate": 1.2103289762635756e-05, "loss": 0.8765, "step": 13509 }, { "epoch": 0.08, "learning_rate": 1.2102235980666362e-05, "loss": 0.926, "step": 13510 }, { "epoch": 0.08, "learning_rate": 1.2101182174272603e-05, "loss": 0.8862, "step": 13511 }, { "epoch": 0.08, "learning_rate": 1.2100128343466723e-05, "loss": 0.89, "step": 13512 }, { "epoch": 0.08, "learning_rate": 1.2099074488260965e-05, "loss": 0.9751, "step": 13513 }, { "epoch": 0.08, "learning_rate": 1.2098020608667569e-05, "loss": 0.8654, "step": 13514 }, { "epoch": 0.08, "learning_rate": 1.2096966704698784e-05, "loss": 0.9003, "step": 13515 }, { "epoch": 0.08, "learning_rate": 1.2095912776366859e-05, "loss": 0.8939, "step": 13516 }, { "epoch": 0.08, "learning_rate": 1.2094858823684025e-05, "loss": 0.9203, "step": 13517 }, { "epoch": 0.08, "learning_rate": 1.2093804846662542e-05, "loss": 0.8596, "step": 13518 }, { "epoch": 0.08, "learning_rate": 1.209275084531465e-05, "loss": 0.9001, "step": 13519 }, { "epoch": 0.08, "learning_rate": 1.2091696819652586e-05, "loss": 0.9128, "step": 13520 }, { "epoch": 0.08, "learning_rate": 1.209064276968861e-05, "loss": 0.9225, "step": 13521 }, { "epoch": 0.08, "learning_rate": 1.2089588695434957e-05, "loss": 0.8928, "step": 13522 }, { "epoch": 0.08, "learning_rate": 1.208853459690388e-05, "loss": 0.9246, "step": 13523 }, { "epoch": 0.08, "learning_rate": 1.2087480474107625e-05, "loss": 0.8933, "step": 13524 }, { "epoch": 0.08, "learning_rate": 1.208642632705844e-05, "loss": 0.9443, "step": 13525 }, { "epoch": 0.08, "learning_rate": 1.208537215576857e-05, "loss": 0.8929, "step": 13526 }, { "epoch": 0.08, "learning_rate": 1.208431796025026e-05, "loss": 0.8753, "step": 13527 }, { "epoch": 0.08, "learning_rate": 1.2083263740515764e-05, "loss": 0.8922, "step": 13528 }, { "epoch": 0.08, "learning_rate": 1.2082209496577325e-05, "loss": 0.9002, "step": 13529 }, { "epoch": 0.08, "learning_rate": 1.2081155228447194e-05, "loss": 0.8713, "step": 13530 }, { "epoch": 0.08, "learning_rate": 1.2080100936137621e-05, "loss": 0.886, "step": 13531 }, { "epoch": 0.08, "learning_rate": 1.2079046619660854e-05, "loss": 0.8949, "step": 13532 }, { "epoch": 0.08, "learning_rate": 1.2077992279029141e-05, "loss": 0.9247, "step": 13533 }, { "epoch": 0.08, "learning_rate": 1.2076937914254736e-05, "loss": 0.9191, "step": 13534 }, { "epoch": 0.08, "learning_rate": 1.207588352534988e-05, "loss": 0.9688, "step": 13535 }, { "epoch": 0.08, "learning_rate": 1.2074829112326834e-05, "loss": 0.9211, "step": 13536 }, { "epoch": 0.08, "learning_rate": 1.207377467519784e-05, "loss": 0.9166, "step": 13537 }, { "epoch": 0.08, "learning_rate": 1.2072720213975151e-05, "loss": 0.9095, "step": 13538 }, { "epoch": 0.08, "learning_rate": 1.2071665728671022e-05, "loss": 0.8799, "step": 13539 }, { "epoch": 0.08, "learning_rate": 1.20706112192977e-05, "loss": 0.9228, "step": 13540 }, { "epoch": 0.08, "learning_rate": 1.2069556685867438e-05, "loss": 0.9172, "step": 13541 }, { "epoch": 0.08, "learning_rate": 1.206850212839249e-05, "loss": 0.9024, "step": 13542 }, { "epoch": 0.08, "learning_rate": 1.2067447546885102e-05, "loss": 0.9536, "step": 13543 }, { "epoch": 0.08, "learning_rate": 1.2066392941357532e-05, "loss": 0.9044, "step": 13544 }, { "epoch": 0.08, "learning_rate": 1.2065338311822032e-05, "loss": 0.9607, "step": 13545 }, { "epoch": 0.08, "learning_rate": 1.2064283658290855e-05, "loss": 0.9442, "step": 13546 }, { "epoch": 0.08, "learning_rate": 1.2063228980776251e-05, "loss": 0.9048, "step": 13547 }, { "epoch": 0.08, "learning_rate": 1.2062174279290475e-05, "loss": 0.9144, "step": 13548 }, { "epoch": 0.08, "learning_rate": 1.2061119553845787e-05, "loss": 0.9066, "step": 13549 }, { "epoch": 0.08, "learning_rate": 1.206006480445443e-05, "loss": 0.8853, "step": 13550 }, { "epoch": 0.08, "learning_rate": 1.2059010031128664e-05, "loss": 0.9334, "step": 13551 }, { "epoch": 0.08, "learning_rate": 1.2057955233880748e-05, "loss": 0.9055, "step": 13552 }, { "epoch": 0.08, "learning_rate": 1.2056900412722927e-05, "loss": 0.9275, "step": 13553 }, { "epoch": 0.08, "learning_rate": 1.2055845567667465e-05, "loss": 0.8932, "step": 13554 }, { "epoch": 0.09, "learning_rate": 1.2054790698726617e-05, "loss": 0.8603, "step": 13555 }, { "epoch": 0.09, "learning_rate": 1.2053735805912634e-05, "loss": 0.9319, "step": 13556 }, { "epoch": 0.09, "learning_rate": 1.2052680889237773e-05, "loss": 0.8936, "step": 13557 }, { "epoch": 0.09, "learning_rate": 1.2051625948714294e-05, "loss": 0.9442, "step": 13558 }, { "epoch": 0.09, "learning_rate": 1.2050570984354448e-05, "loss": 0.9111, "step": 13559 }, { "epoch": 0.09, "learning_rate": 1.2049515996170497e-05, "loss": 0.903, "step": 13560 }, { "epoch": 0.09, "learning_rate": 1.2048460984174695e-05, "loss": 0.8588, "step": 13561 }, { "epoch": 0.09, "learning_rate": 1.2047405948379302e-05, "loss": 0.887, "step": 13562 }, { "epoch": 0.09, "learning_rate": 1.2046350888796571e-05, "loss": 0.9079, "step": 13563 }, { "epoch": 0.09, "learning_rate": 1.2045295805438769e-05, "loss": 0.9283, "step": 13564 }, { "epoch": 0.09, "learning_rate": 1.2044240698318146e-05, "loss": 0.9275, "step": 13565 }, { "epoch": 0.09, "learning_rate": 1.2043185567446957e-05, "loss": 0.9041, "step": 13566 }, { "epoch": 0.09, "learning_rate": 1.2042130412837473e-05, "loss": 0.9438, "step": 13567 }, { "epoch": 0.09, "learning_rate": 1.2041075234501946e-05, "loss": 0.8772, "step": 13568 }, { "epoch": 0.09, "learning_rate": 1.2040020032452637e-05, "loss": 0.9179, "step": 13569 }, { "epoch": 0.09, "learning_rate": 1.2038964806701803e-05, "loss": 0.9353, "step": 13570 }, { "epoch": 0.09, "learning_rate": 1.2037909557261709e-05, "loss": 0.9263, "step": 13571 }, { "epoch": 0.09, "learning_rate": 1.2036854284144608e-05, "loss": 0.9093, "step": 13572 }, { "epoch": 0.09, "learning_rate": 1.2035798987362766e-05, "loss": 0.9445, "step": 13573 }, { "epoch": 0.09, "learning_rate": 1.2034743666928443e-05, "loss": 0.9262, "step": 13574 }, { "epoch": 0.09, "learning_rate": 1.2033688322853896e-05, "loss": 0.8946, "step": 13575 }, { "epoch": 0.09, "learning_rate": 1.2032632955151394e-05, "loss": 0.8657, "step": 13576 }, { "epoch": 0.09, "learning_rate": 1.203157756383319e-05, "loss": 0.931, "step": 13577 }, { "epoch": 0.09, "learning_rate": 1.2030522148911553e-05, "loss": 0.8944, "step": 13578 }, { "epoch": 0.09, "learning_rate": 1.2029466710398738e-05, "loss": 0.9203, "step": 13579 }, { "epoch": 0.09, "learning_rate": 1.2028411248307016e-05, "loss": 0.8951, "step": 13580 }, { "epoch": 0.09, "learning_rate": 1.2027355762648643e-05, "loss": 0.8893, "step": 13581 }, { "epoch": 0.09, "learning_rate": 1.2026300253435883e-05, "loss": 0.8957, "step": 13582 }, { "epoch": 0.09, "learning_rate": 1.2025244720681002e-05, "loss": 0.925, "step": 13583 }, { "epoch": 0.09, "learning_rate": 1.2024189164396258e-05, "loss": 0.9033, "step": 13584 }, { "epoch": 0.09, "learning_rate": 1.2023133584593923e-05, "loss": 0.9006, "step": 13585 }, { "epoch": 0.09, "learning_rate": 1.2022077981286252e-05, "loss": 0.8792, "step": 13586 }, { "epoch": 0.09, "learning_rate": 1.2021022354485514e-05, "loss": 0.9443, "step": 13587 }, { "epoch": 0.09, "learning_rate": 1.2019966704203979e-05, "loss": 0.9062, "step": 13588 }, { "epoch": 0.09, "learning_rate": 1.2018911030453899e-05, "loss": 0.9085, "step": 13589 }, { "epoch": 0.09, "learning_rate": 1.201785533324755e-05, "loss": 0.89, "step": 13590 }, { "epoch": 0.09, "learning_rate": 1.2016799612597193e-05, "loss": 0.9282, "step": 13591 }, { "epoch": 0.09, "learning_rate": 1.2015743868515095e-05, "loss": 0.9083, "step": 13592 }, { "epoch": 0.09, "learning_rate": 1.2014688101013518e-05, "loss": 0.919, "step": 13593 }, { "epoch": 0.09, "learning_rate": 1.2013632310104732e-05, "loss": 0.9141, "step": 13594 }, { "epoch": 0.09, "learning_rate": 1.2012576495801007e-05, "loss": 0.9057, "step": 13595 }, { "epoch": 0.09, "learning_rate": 1.20115206581146e-05, "loss": 0.9291, "step": 13596 }, { "epoch": 0.09, "learning_rate": 1.2010464797057788e-05, "loss": 0.909, "step": 13597 }, { "epoch": 0.09, "learning_rate": 1.2009408912642831e-05, "loss": 0.8728, "step": 13598 }, { "epoch": 0.09, "learning_rate": 1.2008353004882001e-05, "loss": 0.918, "step": 13599 }, { "epoch": 0.09, "learning_rate": 1.2007297073787562e-05, "loss": 0.9406, "step": 13600 }, { "epoch": 0.09, "learning_rate": 1.2006241119371788e-05, "loss": 0.8797, "step": 13601 }, { "epoch": 0.09, "learning_rate": 1.2005185141646942e-05, "loss": 0.9028, "step": 13602 }, { "epoch": 0.09, "learning_rate": 1.2004129140625294e-05, "loss": 0.8881, "step": 13603 }, { "epoch": 0.09, "learning_rate": 1.2003073116319114e-05, "loss": 0.9002, "step": 13604 }, { "epoch": 0.09, "learning_rate": 1.200201706874067e-05, "loss": 0.9058, "step": 13605 }, { "epoch": 0.09, "learning_rate": 1.200096099790223e-05, "loss": 0.9133, "step": 13606 }, { "epoch": 0.09, "learning_rate": 1.1999904903816069e-05, "loss": 0.8581, "step": 13607 }, { "epoch": 0.09, "learning_rate": 1.1998848786494453e-05, "loss": 0.9328, "step": 13608 }, { "epoch": 0.09, "learning_rate": 1.1997792645949648e-05, "loss": 0.9486, "step": 13609 }, { "epoch": 0.09, "learning_rate": 1.1996736482193936e-05, "loss": 0.8636, "step": 13610 }, { "epoch": 0.09, "learning_rate": 1.1995680295239578e-05, "loss": 0.9351, "step": 13611 }, { "epoch": 0.09, "learning_rate": 1.199462408509885e-05, "loss": 0.9249, "step": 13612 }, { "epoch": 0.09, "learning_rate": 1.1993567851784022e-05, "loss": 0.9091, "step": 13613 }, { "epoch": 0.09, "learning_rate": 1.1992511595307363e-05, "loss": 0.9461, "step": 13614 }, { "epoch": 0.09, "learning_rate": 1.1991455315681148e-05, "loss": 0.8898, "step": 13615 }, { "epoch": 0.09, "learning_rate": 1.1990399012917649e-05, "loss": 0.8491, "step": 13616 }, { "epoch": 0.09, "learning_rate": 1.1989342687029137e-05, "loss": 0.9402, "step": 13617 }, { "epoch": 0.09, "learning_rate": 1.1988286338027888e-05, "loss": 0.9037, "step": 13618 }, { "epoch": 0.09, "learning_rate": 1.1987229965926168e-05, "loss": 0.9042, "step": 13619 }, { "epoch": 0.09, "learning_rate": 1.198617357073626e-05, "loss": 0.9551, "step": 13620 }, { "epoch": 0.09, "learning_rate": 1.1985117152470425e-05, "loss": 0.9244, "step": 13621 }, { "epoch": 0.09, "learning_rate": 1.1984060711140951e-05, "loss": 0.9434, "step": 13622 }, { "epoch": 0.09, "learning_rate": 1.19830042467601e-05, "loss": 0.9191, "step": 13623 }, { "epoch": 0.09, "learning_rate": 1.1981947759340153e-05, "loss": 0.9272, "step": 13624 }, { "epoch": 0.09, "learning_rate": 1.1980891248893382e-05, "loss": 0.8957, "step": 13625 }, { "epoch": 0.09, "learning_rate": 1.1979834715432064e-05, "loss": 0.9058, "step": 13626 }, { "epoch": 0.09, "learning_rate": 1.1978778158968471e-05, "loss": 0.9144, "step": 13627 }, { "epoch": 0.09, "learning_rate": 1.1977721579514884e-05, "loss": 0.8877, "step": 13628 }, { "epoch": 0.09, "learning_rate": 1.1976664977083572e-05, "loss": 0.943, "step": 13629 }, { "epoch": 0.09, "learning_rate": 1.1975608351686809e-05, "loss": 0.9127, "step": 13630 }, { "epoch": 0.09, "learning_rate": 1.1974551703336882e-05, "loss": 0.9374, "step": 13631 }, { "epoch": 0.09, "learning_rate": 1.1973495032046061e-05, "loss": 0.9029, "step": 13632 }, { "epoch": 0.09, "learning_rate": 1.1972438337826619e-05, "loss": 0.8806, "step": 13633 }, { "epoch": 0.09, "learning_rate": 1.1971381620690843e-05, "loss": 0.8899, "step": 13634 }, { "epoch": 0.09, "learning_rate": 1.1970324880650999e-05, "loss": 0.8309, "step": 13635 }, { "epoch": 0.09, "learning_rate": 1.196926811771937e-05, "loss": 0.91, "step": 13636 }, { "epoch": 0.09, "learning_rate": 1.1968211331908234e-05, "loss": 0.9306, "step": 13637 }, { "epoch": 0.09, "learning_rate": 1.1967154523229872e-05, "loss": 0.8916, "step": 13638 }, { "epoch": 0.09, "learning_rate": 1.1966097691696555e-05, "loss": 0.8916, "step": 13639 }, { "epoch": 0.09, "learning_rate": 1.1965040837320565e-05, "loss": 0.8765, "step": 13640 }, { "epoch": 0.09, "learning_rate": 1.1963983960114183e-05, "loss": 0.8714, "step": 13641 }, { "epoch": 0.09, "learning_rate": 1.1962927060089682e-05, "loss": 0.9187, "step": 13642 }, { "epoch": 0.09, "learning_rate": 1.196187013725935e-05, "loss": 0.9446, "step": 13643 }, { "epoch": 0.09, "learning_rate": 1.1960813191635461e-05, "loss": 0.9086, "step": 13644 }, { "epoch": 0.09, "learning_rate": 1.1959756223230297e-05, "loss": 0.9025, "step": 13645 }, { "epoch": 0.09, "learning_rate": 1.1958699232056135e-05, "loss": 0.8894, "step": 13646 }, { "epoch": 0.09, "learning_rate": 1.195764221812526e-05, "loss": 0.8938, "step": 13647 }, { "epoch": 0.09, "learning_rate": 1.1956585181449948e-05, "loss": 0.93, "step": 13648 }, { "epoch": 0.09, "learning_rate": 1.1955528122042485e-05, "loss": 0.8989, "step": 13649 }, { "epoch": 0.09, "learning_rate": 1.195447103991515e-05, "loss": 0.9084, "step": 13650 }, { "epoch": 0.09, "learning_rate": 1.1953413935080223e-05, "loss": 0.9065, "step": 13651 }, { "epoch": 0.09, "learning_rate": 1.1952356807549986e-05, "loss": 0.8942, "step": 13652 }, { "epoch": 0.09, "learning_rate": 1.195129965733672e-05, "loss": 0.9159, "step": 13653 }, { "epoch": 0.09, "learning_rate": 1.1950242484452713e-05, "loss": 0.906, "step": 13654 }, { "epoch": 0.09, "learning_rate": 1.1949185288910236e-05, "loss": 0.883, "step": 13655 }, { "epoch": 0.09, "learning_rate": 1.1948128070721588e-05, "loss": 0.9132, "step": 13656 }, { "epoch": 0.09, "learning_rate": 1.1947070829899039e-05, "loss": 0.9245, "step": 13657 }, { "epoch": 0.09, "learning_rate": 1.1946013566454877e-05, "loss": 0.8606, "step": 13658 }, { "epoch": 0.09, "learning_rate": 1.1944956280401387e-05, "loss": 0.8881, "step": 13659 }, { "epoch": 0.09, "learning_rate": 1.194389897175085e-05, "loss": 0.8663, "step": 13660 }, { "epoch": 0.09, "learning_rate": 1.194284164051555e-05, "loss": 0.8794, "step": 13661 }, { "epoch": 0.09, "learning_rate": 1.1941784286707774e-05, "loss": 0.9259, "step": 13662 }, { "epoch": 0.09, "learning_rate": 1.1940726910339805e-05, "loss": 0.9354, "step": 13663 }, { "epoch": 0.09, "learning_rate": 1.1939669511423928e-05, "loss": 0.8929, "step": 13664 }, { "epoch": 0.09, "learning_rate": 1.1938612089972428e-05, "loss": 0.9204, "step": 13665 }, { "epoch": 0.09, "learning_rate": 1.1937554645997593e-05, "loss": 0.8901, "step": 13666 }, { "epoch": 0.09, "learning_rate": 1.1936497179511701e-05, "loss": 0.9035, "step": 13667 }, { "epoch": 0.09, "learning_rate": 1.1935439690527048e-05, "loss": 0.9102, "step": 13668 }, { "epoch": 0.09, "learning_rate": 1.1934382179055914e-05, "loss": 0.8855, "step": 13669 }, { "epoch": 0.09, "learning_rate": 1.193332464511059e-05, "loss": 0.891, "step": 13670 }, { "epoch": 0.09, "learning_rate": 1.1932267088703356e-05, "loss": 0.9349, "step": 13671 }, { "epoch": 0.09, "learning_rate": 1.1931209509846503e-05, "loss": 0.9124, "step": 13672 }, { "epoch": 0.09, "learning_rate": 1.1930151908552319e-05, "loss": 0.8952, "step": 13673 }, { "epoch": 0.09, "learning_rate": 1.192909428483309e-05, "loss": 0.8975, "step": 13674 }, { "epoch": 0.09, "learning_rate": 1.1928036638701106e-05, "loss": 0.9077, "step": 13675 }, { "epoch": 0.09, "learning_rate": 1.192697897016865e-05, "loss": 0.9004, "step": 13676 }, { "epoch": 0.09, "learning_rate": 1.1925921279248016e-05, "loss": 0.9289, "step": 13677 }, { "epoch": 0.09, "learning_rate": 1.1924863565951488e-05, "loss": 0.8984, "step": 13678 }, { "epoch": 0.09, "learning_rate": 1.1923805830291358e-05, "loss": 0.9678, "step": 13679 }, { "epoch": 0.09, "learning_rate": 1.1922748072279917e-05, "loss": 0.8805, "step": 13680 }, { "epoch": 0.09, "learning_rate": 1.1921690291929447e-05, "loss": 0.9188, "step": 13681 }, { "epoch": 0.09, "learning_rate": 1.1920632489252242e-05, "loss": 0.9304, "step": 13682 }, { "epoch": 0.09, "learning_rate": 1.1919574664260594e-05, "loss": 0.9132, "step": 13683 }, { "epoch": 0.09, "learning_rate": 1.191851681696679e-05, "loss": 0.8635, "step": 13684 }, { "epoch": 0.09, "learning_rate": 1.191745894738312e-05, "loss": 0.9084, "step": 13685 }, { "epoch": 0.09, "learning_rate": 1.1916401055521877e-05, "loss": 0.8973, "step": 13686 }, { "epoch": 0.09, "learning_rate": 1.1915343141395352e-05, "loss": 0.9027, "step": 13687 }, { "epoch": 0.09, "learning_rate": 1.1914285205015832e-05, "loss": 0.8725, "step": 13688 }, { "epoch": 0.09, "learning_rate": 1.1913227246395613e-05, "loss": 0.9085, "step": 13689 }, { "epoch": 0.09, "learning_rate": 1.1912169265546985e-05, "loss": 0.8729, "step": 13690 }, { "epoch": 0.09, "learning_rate": 1.191111126248224e-05, "loss": 0.8943, "step": 13691 }, { "epoch": 0.09, "learning_rate": 1.1910053237213669e-05, "loss": 0.9159, "step": 13692 }, { "epoch": 0.09, "learning_rate": 1.1908995189753568e-05, "loss": 0.8863, "step": 13693 }, { "epoch": 0.09, "learning_rate": 1.1907937120114223e-05, "loss": 0.8681, "step": 13694 }, { "epoch": 0.09, "learning_rate": 1.1906879028307935e-05, "loss": 0.9619, "step": 13695 }, { "epoch": 0.09, "learning_rate": 1.1905820914346991e-05, "loss": 0.8633, "step": 13696 }, { "epoch": 0.09, "learning_rate": 1.1904762778243687e-05, "loss": 0.9242, "step": 13697 }, { "epoch": 0.09, "learning_rate": 1.1903704620010315e-05, "loss": 0.8942, "step": 13698 }, { "epoch": 0.09, "learning_rate": 1.1902646439659171e-05, "loss": 0.8781, "step": 13699 }, { "epoch": 0.09, "learning_rate": 1.1901588237202551e-05, "loss": 0.8805, "step": 13700 }, { "epoch": 0.09, "learning_rate": 1.1900530012652744e-05, "loss": 0.8836, "step": 13701 }, { "epoch": 0.09, "learning_rate": 1.1899471766022053e-05, "loss": 0.9061, "step": 13702 }, { "epoch": 0.09, "learning_rate": 1.1898413497322762e-05, "loss": 0.8958, "step": 13703 }, { "epoch": 0.09, "learning_rate": 1.1897355206567176e-05, "loss": 0.8997, "step": 13704 }, { "epoch": 0.09, "learning_rate": 1.1896296893767588e-05, "loss": 0.8921, "step": 13705 }, { "epoch": 0.09, "learning_rate": 1.1895238558936288e-05, "loss": 0.9126, "step": 13706 }, { "epoch": 0.09, "learning_rate": 1.1894180202085579e-05, "loss": 0.9525, "step": 13707 }, { "epoch": 0.09, "learning_rate": 1.1893121823227754e-05, "loss": 0.8894, "step": 13708 }, { "epoch": 0.09, "learning_rate": 1.1892063422375113e-05, "loss": 0.8892, "step": 13709 }, { "epoch": 0.09, "learning_rate": 1.1891004999539946e-05, "loss": 0.9226, "step": 13710 }, { "epoch": 0.09, "learning_rate": 1.1889946554734554e-05, "loss": 0.888, "step": 13711 }, { "epoch": 0.09, "learning_rate": 1.188888808797124e-05, "loss": 0.8873, "step": 13712 }, { "epoch": 0.09, "learning_rate": 1.188782959926229e-05, "loss": 0.911, "step": 13713 }, { "epoch": 0.09, "learning_rate": 1.1886771088620012e-05, "loss": 0.9218, "step": 13714 }, { "epoch": 0.09, "learning_rate": 1.18857125560567e-05, "loss": 0.9274, "step": 13715 }, { "epoch": 0.09, "learning_rate": 1.1884654001584649e-05, "loss": 0.8793, "step": 13716 }, { "epoch": 0.09, "learning_rate": 1.1883595425216161e-05, "loss": 0.8538, "step": 13717 }, { "epoch": 0.09, "learning_rate": 1.1882536826963537e-05, "loss": 0.9081, "step": 13718 }, { "epoch": 0.09, "learning_rate": 1.1881478206839074e-05, "loss": 0.9326, "step": 13719 }, { "epoch": 0.09, "learning_rate": 1.188041956485507e-05, "loss": 0.9058, "step": 13720 }, { "epoch": 0.09, "learning_rate": 1.1879360901023826e-05, "loss": 0.9277, "step": 13721 }, { "epoch": 0.09, "learning_rate": 1.187830221535764e-05, "loss": 0.9647, "step": 13722 }, { "epoch": 0.09, "learning_rate": 1.1877243507868817e-05, "loss": 0.8911, "step": 13723 }, { "epoch": 0.09, "learning_rate": 1.1876184778569652e-05, "loss": 0.9257, "step": 13724 }, { "epoch": 0.09, "learning_rate": 1.1875126027472448e-05, "loss": 0.8847, "step": 13725 }, { "epoch": 0.09, "learning_rate": 1.1874067254589505e-05, "loss": 0.8952, "step": 13726 }, { "epoch": 0.09, "learning_rate": 1.1873008459933127e-05, "loss": 0.9054, "step": 13727 }, { "epoch": 0.09, "learning_rate": 1.187194964351561e-05, "loss": 0.9239, "step": 13728 }, { "epoch": 0.09, "learning_rate": 1.1870890805349258e-05, "loss": 0.9114, "step": 13729 }, { "epoch": 0.09, "learning_rate": 1.1869831945446377e-05, "loss": 0.92, "step": 13730 }, { "epoch": 0.09, "learning_rate": 1.186877306381926e-05, "loss": 0.9338, "step": 13731 }, { "epoch": 0.09, "learning_rate": 1.186771416048022e-05, "loss": 0.9277, "step": 13732 }, { "epoch": 0.09, "learning_rate": 1.1866655235441557e-05, "loss": 0.913, "step": 13733 }, { "epoch": 0.09, "learning_rate": 1.1865596288715565e-05, "loss": 0.91, "step": 13734 }, { "epoch": 0.09, "learning_rate": 1.186453732031456e-05, "loss": 0.9125, "step": 13735 }, { "epoch": 0.09, "learning_rate": 1.1863478330250835e-05, "loss": 0.9164, "step": 13736 }, { "epoch": 0.09, "learning_rate": 1.1862419318536698e-05, "loss": 0.9264, "step": 13737 }, { "epoch": 0.09, "learning_rate": 1.1861360285184454e-05, "loss": 0.9108, "step": 13738 }, { "epoch": 0.09, "learning_rate": 1.1860301230206406e-05, "loss": 0.9344, "step": 13739 }, { "epoch": 0.09, "learning_rate": 1.1859242153614856e-05, "loss": 0.868, "step": 13740 }, { "epoch": 0.09, "learning_rate": 1.1858183055422112e-05, "loss": 0.927, "step": 13741 }, { "epoch": 0.09, "learning_rate": 1.185712393564048e-05, "loss": 0.9027, "step": 13742 }, { "epoch": 0.09, "learning_rate": 1.1856064794282261e-05, "loss": 0.922, "step": 13743 }, { "epoch": 0.09, "learning_rate": 1.185500563135976e-05, "loss": 0.8778, "step": 13744 }, { "epoch": 0.09, "learning_rate": 1.1853946446885285e-05, "loss": 0.933, "step": 13745 }, { "epoch": 0.09, "learning_rate": 1.1852887240871145e-05, "loss": 0.8759, "step": 13746 }, { "epoch": 0.09, "learning_rate": 1.1851828013329641e-05, "loss": 0.9271, "step": 13747 }, { "epoch": 0.09, "learning_rate": 1.1850768764273082e-05, "loss": 0.8969, "step": 13748 }, { "epoch": 0.09, "learning_rate": 1.1849709493713774e-05, "loss": 0.8566, "step": 13749 }, { "epoch": 0.09, "learning_rate": 1.1848650201664022e-05, "loss": 0.9135, "step": 13750 }, { "epoch": 0.09, "learning_rate": 1.184759088813614e-05, "loss": 0.9112, "step": 13751 }, { "epoch": 0.09, "learning_rate": 1.1846531553142426e-05, "loss": 0.9164, "step": 13752 }, { "epoch": 0.09, "learning_rate": 1.1845472196695193e-05, "loss": 0.9132, "step": 13753 }, { "epoch": 0.09, "learning_rate": 1.1844412818806749e-05, "loss": 0.8975, "step": 13754 }, { "epoch": 0.09, "learning_rate": 1.1843353419489403e-05, "loss": 0.8756, "step": 13755 }, { "epoch": 0.09, "learning_rate": 1.1842293998755457e-05, "loss": 0.8981, "step": 13756 }, { "epoch": 0.09, "learning_rate": 1.1841234556617228e-05, "loss": 0.9161, "step": 13757 }, { "epoch": 0.09, "learning_rate": 1.184017509308702e-05, "loss": 0.9171, "step": 13758 }, { "epoch": 0.09, "learning_rate": 1.1839115608177141e-05, "loss": 0.9437, "step": 13759 }, { "epoch": 0.09, "learning_rate": 1.1838056101899906e-05, "loss": 0.8809, "step": 13760 }, { "epoch": 0.09, "learning_rate": 1.183699657426762e-05, "loss": 0.9171, "step": 13761 }, { "epoch": 0.09, "learning_rate": 1.1835937025292593e-05, "loss": 0.8747, "step": 13762 }, { "epoch": 0.09, "learning_rate": 1.1834877454987137e-05, "loss": 0.8675, "step": 13763 }, { "epoch": 0.09, "learning_rate": 1.1833817863363563e-05, "loss": 0.8794, "step": 13764 }, { "epoch": 0.09, "learning_rate": 1.183275825043418e-05, "loss": 0.911, "step": 13765 }, { "epoch": 0.09, "learning_rate": 1.1831698616211296e-05, "loss": 0.9262, "step": 13766 }, { "epoch": 0.09, "learning_rate": 1.183063896070723e-05, "loss": 0.8844, "step": 13767 }, { "epoch": 0.09, "learning_rate": 1.1829579283934286e-05, "loss": 0.9095, "step": 13768 }, { "epoch": 0.09, "learning_rate": 1.1828519585904779e-05, "loss": 0.8826, "step": 13769 }, { "epoch": 0.09, "learning_rate": 1.1827459866631019e-05, "loss": 0.925, "step": 13770 }, { "epoch": 0.09, "learning_rate": 1.1826400126125321e-05, "loss": 0.8813, "step": 13771 }, { "epoch": 0.09, "learning_rate": 1.1825340364399996e-05, "loss": 0.8988, "step": 13772 }, { "epoch": 0.09, "learning_rate": 1.1824280581467356e-05, "loss": 0.9045, "step": 13773 }, { "epoch": 0.09, "learning_rate": 1.1823220777339712e-05, "loss": 0.9166, "step": 13774 }, { "epoch": 0.09, "learning_rate": 1.182216095202938e-05, "loss": 0.9026, "step": 13775 }, { "epoch": 0.09, "learning_rate": 1.1821101105548676e-05, "loss": 0.9099, "step": 13776 }, { "epoch": 0.09, "learning_rate": 1.1820041237909902e-05, "loss": 0.8838, "step": 13777 }, { "epoch": 0.09, "learning_rate": 1.181898134912539e-05, "loss": 0.8757, "step": 13778 }, { "epoch": 0.09, "learning_rate": 1.1817921439207439e-05, "loss": 0.9229, "step": 13779 }, { "epoch": 0.09, "learning_rate": 1.1816861508168367e-05, "loss": 0.8792, "step": 13780 }, { "epoch": 0.09, "learning_rate": 1.1815801556020495e-05, "loss": 0.925, "step": 13781 }, { "epoch": 0.09, "learning_rate": 1.1814741582776128e-05, "loss": 0.9012, "step": 13782 }, { "epoch": 0.09, "learning_rate": 1.1813681588447588e-05, "loss": 0.9164, "step": 13783 }, { "epoch": 0.09, "learning_rate": 1.1812621573047188e-05, "loss": 0.9104, "step": 13784 }, { "epoch": 0.09, "learning_rate": 1.1811561536587244e-05, "loss": 0.9168, "step": 13785 }, { "epoch": 0.09, "learning_rate": 1.1810501479080072e-05, "loss": 0.8863, "step": 13786 }, { "epoch": 0.09, "learning_rate": 1.1809441400537984e-05, "loss": 0.921, "step": 13787 }, { "epoch": 0.09, "learning_rate": 1.1808381300973307e-05, "loss": 0.9242, "step": 13788 }, { "epoch": 0.09, "learning_rate": 1.1807321180398343e-05, "loss": 0.9331, "step": 13789 }, { "epoch": 0.09, "learning_rate": 1.180626103882542e-05, "loss": 0.8621, "step": 13790 }, { "epoch": 0.09, "learning_rate": 1.180520087626685e-05, "loss": 0.9171, "step": 13791 }, { "epoch": 0.09, "learning_rate": 1.1804140692734951e-05, "loss": 0.9098, "step": 13792 }, { "epoch": 0.09, "learning_rate": 1.1803080488242043e-05, "loss": 0.9036, "step": 13793 }, { "epoch": 0.09, "learning_rate": 1.1802020262800442e-05, "loss": 0.8767, "step": 13794 }, { "epoch": 0.09, "learning_rate": 1.1800960016422463e-05, "loss": 0.9196, "step": 13795 }, { "epoch": 0.09, "learning_rate": 1.1799899749120429e-05, "loss": 0.9363, "step": 13796 }, { "epoch": 0.09, "learning_rate": 1.1798839460906657e-05, "loss": 0.8782, "step": 13797 }, { "epoch": 0.09, "learning_rate": 1.1797779151793463e-05, "loss": 0.8856, "step": 13798 }, { "epoch": 0.09, "learning_rate": 1.1796718821793169e-05, "loss": 0.9178, "step": 13799 }, { "epoch": 0.09, "learning_rate": 1.1795658470918094e-05, "loss": 0.8684, "step": 13800 }, { "epoch": 0.09, "learning_rate": 1.1794598099180558e-05, "loss": 0.9102, "step": 13801 }, { "epoch": 0.09, "learning_rate": 1.1793537706592875e-05, "loss": 0.954, "step": 13802 }, { "epoch": 0.09, "learning_rate": 1.1792477293167374e-05, "loss": 0.9127, "step": 13803 }, { "epoch": 0.09, "learning_rate": 1.179141685891637e-05, "loss": 0.9131, "step": 13804 }, { "epoch": 0.09, "learning_rate": 1.1790356403852181e-05, "loss": 0.9276, "step": 13805 }, { "epoch": 0.09, "learning_rate": 1.1789295927987134e-05, "loss": 0.9051, "step": 13806 }, { "epoch": 0.09, "learning_rate": 1.1788235431333545e-05, "loss": 0.9172, "step": 13807 }, { "epoch": 0.09, "learning_rate": 1.1787174913903735e-05, "loss": 0.9005, "step": 13808 }, { "epoch": 0.09, "learning_rate": 1.1786114375710031e-05, "loss": 0.8418, "step": 13809 }, { "epoch": 0.09, "learning_rate": 1.1785053816764752e-05, "loss": 0.8811, "step": 13810 }, { "epoch": 0.09, "learning_rate": 1.1783993237080214e-05, "loss": 0.9129, "step": 13811 }, { "epoch": 0.09, "learning_rate": 1.1782932636668745e-05, "loss": 0.9297, "step": 13812 }, { "epoch": 0.09, "learning_rate": 1.1781872015542668e-05, "loss": 0.9107, "step": 13813 }, { "epoch": 0.09, "learning_rate": 1.17808113737143e-05, "loss": 0.9192, "step": 13814 }, { "epoch": 0.09, "learning_rate": 1.1779750711195971e-05, "loss": 0.9103, "step": 13815 }, { "epoch": 0.09, "learning_rate": 1.1778690028e-05, "loss": 0.8914, "step": 13816 }, { "epoch": 0.09, "learning_rate": 1.1777629324138709e-05, "loss": 0.9039, "step": 13817 }, { "epoch": 0.09, "learning_rate": 1.1776568599624428e-05, "loss": 0.9389, "step": 13818 }, { "epoch": 0.09, "learning_rate": 1.1775507854469471e-05, "loss": 0.9067, "step": 13819 }, { "epoch": 0.09, "learning_rate": 1.177444708868617e-05, "loss": 0.8645, "step": 13820 }, { "epoch": 0.09, "learning_rate": 1.1773386302286845e-05, "loss": 0.8858, "step": 13821 }, { "epoch": 0.09, "learning_rate": 1.1772325495283824e-05, "loss": 0.8784, "step": 13822 }, { "epoch": 0.09, "learning_rate": 1.1771264667689428e-05, "loss": 0.887, "step": 13823 }, { "epoch": 0.09, "learning_rate": 1.1770203819515987e-05, "loss": 0.9216, "step": 13824 }, { "epoch": 0.09, "learning_rate": 1.176914295077582e-05, "loss": 0.9133, "step": 13825 }, { "epoch": 0.09, "learning_rate": 1.1768082061481257e-05, "loss": 0.9048, "step": 13826 }, { "epoch": 0.09, "learning_rate": 1.1767021151644622e-05, "loss": 0.894, "step": 13827 }, { "epoch": 0.09, "learning_rate": 1.1765960221278241e-05, "loss": 0.8852, "step": 13828 }, { "epoch": 0.09, "learning_rate": 1.1764899270394439e-05, "loss": 0.8739, "step": 13829 }, { "epoch": 0.09, "learning_rate": 1.1763838299005546e-05, "loss": 0.9186, "step": 13830 }, { "epoch": 0.09, "learning_rate": 1.1762777307123886e-05, "loss": 0.9258, "step": 13831 }, { "epoch": 0.09, "learning_rate": 1.1761716294761787e-05, "loss": 0.9371, "step": 13832 }, { "epoch": 0.09, "learning_rate": 1.1760655261931572e-05, "loss": 0.9196, "step": 13833 }, { "epoch": 0.09, "learning_rate": 1.1759594208645577e-05, "loss": 0.885, "step": 13834 }, { "epoch": 0.09, "learning_rate": 1.1758533134916119e-05, "loss": 0.9266, "step": 13835 }, { "epoch": 0.09, "learning_rate": 1.1757472040755535e-05, "loss": 0.9231, "step": 13836 }, { "epoch": 0.09, "learning_rate": 1.1756410926176149e-05, "loss": 0.9222, "step": 13837 }, { "epoch": 0.09, "learning_rate": 1.1755349791190289e-05, "loss": 0.9276, "step": 13838 }, { "epoch": 0.09, "learning_rate": 1.1754288635810285e-05, "loss": 0.8992, "step": 13839 }, { "epoch": 0.09, "learning_rate": 1.1753227460048466e-05, "loss": 0.91, "step": 13840 }, { "epoch": 0.09, "learning_rate": 1.175216626391716e-05, "loss": 0.9498, "step": 13841 }, { "epoch": 0.09, "learning_rate": 1.1751105047428695e-05, "loss": 0.9078, "step": 13842 }, { "epoch": 0.09, "learning_rate": 1.1750043810595404e-05, "loss": 0.92, "step": 13843 }, { "epoch": 0.09, "learning_rate": 1.1748982553429611e-05, "loss": 0.9295, "step": 13844 }, { "epoch": 0.09, "learning_rate": 1.1747921275943651e-05, "loss": 0.9347, "step": 13845 }, { "epoch": 0.09, "learning_rate": 1.1746859978149853e-05, "loss": 0.879, "step": 13846 }, { "epoch": 0.09, "learning_rate": 1.174579866006055e-05, "loss": 0.8994, "step": 13847 }, { "epoch": 0.09, "learning_rate": 1.1744737321688067e-05, "loss": 0.8741, "step": 13848 }, { "epoch": 0.09, "learning_rate": 1.174367596304474e-05, "loss": 0.9226, "step": 13849 }, { "epoch": 0.09, "learning_rate": 1.1742614584142897e-05, "loss": 0.9194, "step": 13850 }, { "epoch": 0.09, "learning_rate": 1.174155318499487e-05, "loss": 0.9304, "step": 13851 }, { "epoch": 0.09, "learning_rate": 1.1740491765612994e-05, "loss": 0.8953, "step": 13852 }, { "epoch": 0.09, "learning_rate": 1.1739430326009595e-05, "loss": 0.9287, "step": 13853 }, { "epoch": 0.09, "learning_rate": 1.173836886619701e-05, "loss": 0.8557, "step": 13854 }, { "epoch": 0.1, "learning_rate": 1.1737307386187567e-05, "loss": 0.9148, "step": 13855 }, { "epoch": 0.1, "learning_rate": 1.1736245885993604e-05, "loss": 0.9307, "step": 13856 }, { "epoch": 0.1, "learning_rate": 1.1735184365627447e-05, "loss": 0.9125, "step": 13857 }, { "epoch": 0.1, "learning_rate": 1.1734122825101437e-05, "loss": 0.9342, "step": 13858 }, { "epoch": 0.1, "learning_rate": 1.1733061264427904e-05, "loss": 0.8754, "step": 13859 }, { "epoch": 0.1, "learning_rate": 1.1731999683619173e-05, "loss": 0.9097, "step": 13860 }, { "epoch": 0.1, "learning_rate": 1.1730938082687592e-05, "loss": 0.9054, "step": 13861 }, { "epoch": 0.1, "learning_rate": 1.1729876461645486e-05, "loss": 0.9306, "step": 13862 }, { "epoch": 0.1, "learning_rate": 1.1728814820505194e-05, "loss": 0.8973, "step": 13863 }, { "epoch": 0.1, "learning_rate": 1.1727753159279048e-05, "loss": 0.895, "step": 13864 }, { "epoch": 0.1, "learning_rate": 1.1726691477979379e-05, "loss": 0.9039, "step": 13865 }, { "epoch": 0.1, "learning_rate": 1.1725629776618528e-05, "loss": 0.934, "step": 13866 }, { "epoch": 0.1, "learning_rate": 1.1724568055208827e-05, "loss": 0.9466, "step": 13867 }, { "epoch": 0.1, "learning_rate": 1.1723506313762615e-05, "loss": 0.8857, "step": 13868 }, { "epoch": 0.1, "learning_rate": 1.1722444552292221e-05, "loss": 0.9019, "step": 13869 }, { "epoch": 0.1, "learning_rate": 1.1721382770809986e-05, "loss": 0.8824, "step": 13870 }, { "epoch": 0.1, "learning_rate": 1.1720320969328245e-05, "loss": 0.8798, "step": 13871 }, { "epoch": 0.1, "learning_rate": 1.1719259147859333e-05, "loss": 0.8522, "step": 13872 }, { "epoch": 0.1, "learning_rate": 1.1718197306415591e-05, "loss": 0.9144, "step": 13873 }, { "epoch": 0.1, "learning_rate": 1.171713544500935e-05, "loss": 0.9006, "step": 13874 }, { "epoch": 0.1, "learning_rate": 1.1716073563652946e-05, "loss": 0.9199, "step": 13875 }, { "epoch": 0.1, "learning_rate": 1.1715011662358722e-05, "loss": 0.9109, "step": 13876 }, { "epoch": 0.1, "learning_rate": 1.1713949741139015e-05, "loss": 0.8905, "step": 13877 }, { "epoch": 0.1, "learning_rate": 1.1712887800006158e-05, "loss": 0.9304, "step": 13878 }, { "epoch": 0.1, "learning_rate": 1.1711825838972492e-05, "loss": 0.8981, "step": 13879 }, { "epoch": 0.1, "learning_rate": 1.1710763858050358e-05, "loss": 0.9067, "step": 13880 }, { "epoch": 0.1, "learning_rate": 1.1709701857252086e-05, "loss": 0.9372, "step": 13881 }, { "epoch": 0.1, "learning_rate": 1.1708639836590024e-05, "loss": 0.8989, "step": 13882 }, { "epoch": 0.1, "learning_rate": 1.1707577796076504e-05, "loss": 0.8815, "step": 13883 }, { "epoch": 0.1, "learning_rate": 1.170651573572387e-05, "loss": 0.881, "step": 13884 }, { "epoch": 0.1, "learning_rate": 1.1705453655544456e-05, "loss": 0.9231, "step": 13885 }, { "epoch": 0.1, "learning_rate": 1.1704391555550607e-05, "loss": 0.9062, "step": 13886 }, { "epoch": 0.1, "learning_rate": 1.170332943575466e-05, "loss": 0.9081, "step": 13887 }, { "epoch": 0.1, "learning_rate": 1.1702267296168954e-05, "loss": 0.9206, "step": 13888 }, { "epoch": 0.1, "learning_rate": 1.1701205136805834e-05, "loss": 0.8989, "step": 13889 }, { "epoch": 0.1, "learning_rate": 1.1700142957677634e-05, "loss": 0.9194, "step": 13890 }, { "epoch": 0.1, "learning_rate": 1.1699080758796697e-05, "loss": 0.9459, "step": 13891 }, { "epoch": 0.1, "learning_rate": 1.1698018540175367e-05, "loss": 0.8877, "step": 13892 }, { "epoch": 0.1, "learning_rate": 1.169695630182598e-05, "loss": 0.8761, "step": 13893 }, { "epoch": 0.1, "learning_rate": 1.1695894043760883e-05, "loss": 0.8769, "step": 13894 }, { "epoch": 0.1, "learning_rate": 1.1694831765992416e-05, "loss": 0.8847, "step": 13895 }, { "epoch": 0.1, "learning_rate": 1.1693769468532915e-05, "loss": 0.9168, "step": 13896 }, { "epoch": 0.1, "learning_rate": 1.1692707151394732e-05, "loss": 0.8988, "step": 13897 }, { "epoch": 0.1, "learning_rate": 1.1691644814590202e-05, "loss": 0.855, "step": 13898 }, { "epoch": 0.1, "learning_rate": 1.169058245813167e-05, "loss": 0.9167, "step": 13899 }, { "epoch": 0.1, "learning_rate": 1.1689520082031475e-05, "loss": 0.8743, "step": 13900 }, { "epoch": 0.1, "learning_rate": 1.1688457686301966e-05, "loss": 0.924, "step": 13901 }, { "epoch": 0.1, "learning_rate": 1.1687395270955485e-05, "loss": 0.9056, "step": 13902 }, { "epoch": 0.1, "learning_rate": 1.1686332836004371e-05, "loss": 0.9431, "step": 13903 }, { "epoch": 0.1, "learning_rate": 1.1685270381460972e-05, "loss": 0.8803, "step": 13904 }, { "epoch": 0.1, "learning_rate": 1.168420790733763e-05, "loss": 0.9363, "step": 13905 }, { "epoch": 0.1, "learning_rate": 1.1683145413646688e-05, "loss": 0.8734, "step": 13906 }, { "epoch": 0.1, "learning_rate": 1.1682082900400496e-05, "loss": 0.8706, "step": 13907 }, { "epoch": 0.1, "learning_rate": 1.1681020367611391e-05, "loss": 0.885, "step": 13908 }, { "epoch": 0.1, "learning_rate": 1.1679957815291723e-05, "loss": 0.9695, "step": 13909 }, { "epoch": 0.1, "learning_rate": 1.1678895243453836e-05, "loss": 0.8953, "step": 13910 }, { "epoch": 0.1, "learning_rate": 1.1677832652110072e-05, "loss": 0.9096, "step": 13911 }, { "epoch": 0.1, "learning_rate": 1.1676770041272783e-05, "loss": 0.9042, "step": 13912 }, { "epoch": 0.1, "learning_rate": 1.1675707410954307e-05, "loss": 0.9327, "step": 13913 }, { "epoch": 0.1, "learning_rate": 1.1674644761166999e-05, "loss": 0.8979, "step": 13914 }, { "epoch": 0.1, "learning_rate": 1.1673582091923192e-05, "loss": 0.9266, "step": 13915 }, { "epoch": 0.1, "learning_rate": 1.1672519403235248e-05, "loss": 0.8923, "step": 13916 }, { "epoch": 0.1, "learning_rate": 1.1671456695115499e-05, "loss": 0.8977, "step": 13917 }, { "epoch": 0.1, "learning_rate": 1.1670393967576303e-05, "loss": 0.841, "step": 13918 }, { "epoch": 0.1, "learning_rate": 1.1669331220630004e-05, "loss": 0.9078, "step": 13919 }, { "epoch": 0.1, "learning_rate": 1.1668268454288947e-05, "loss": 0.8654, "step": 13920 }, { "epoch": 0.1, "learning_rate": 1.1667205668565478e-05, "loss": 0.91, "step": 13921 }, { "epoch": 0.1, "learning_rate": 1.1666142863471949e-05, "loss": 0.8993, "step": 13922 }, { "epoch": 0.1, "learning_rate": 1.1665080039020709e-05, "loss": 0.8945, "step": 13923 }, { "epoch": 0.1, "learning_rate": 1.1664017195224098e-05, "loss": 0.8779, "step": 13924 }, { "epoch": 0.1, "learning_rate": 1.1662954332094475e-05, "loss": 0.8992, "step": 13925 }, { "epoch": 0.1, "learning_rate": 1.1661891449644185e-05, "loss": 0.9161, "step": 13926 }, { "epoch": 0.1, "learning_rate": 1.1660828547885568e-05, "loss": 0.8989, "step": 13927 }, { "epoch": 0.1, "learning_rate": 1.1659765626830987e-05, "loss": 0.9271, "step": 13928 }, { "epoch": 0.1, "learning_rate": 1.1658702686492781e-05, "loss": 0.8764, "step": 13929 }, { "epoch": 0.1, "learning_rate": 1.1657639726883306e-05, "loss": 0.9391, "step": 13930 }, { "epoch": 0.1, "learning_rate": 1.1656576748014908e-05, "loss": 0.8876, "step": 13931 }, { "epoch": 0.1, "learning_rate": 1.165551374989994e-05, "loss": 0.9011, "step": 13932 }, { "epoch": 0.1, "learning_rate": 1.1654450732550749e-05, "loss": 0.8539, "step": 13933 }, { "epoch": 0.1, "learning_rate": 1.1653387695979685e-05, "loss": 0.8776, "step": 13934 }, { "epoch": 0.1, "learning_rate": 1.1652324640199103e-05, "loss": 0.9485, "step": 13935 }, { "epoch": 0.1, "learning_rate": 1.1651261565221352e-05, "loss": 0.9068, "step": 13936 }, { "epoch": 0.1, "learning_rate": 1.165019847105878e-05, "loss": 0.9098, "step": 13937 }, { "epoch": 0.1, "learning_rate": 1.1649135357723742e-05, "loss": 0.8896, "step": 13938 }, { "epoch": 0.1, "learning_rate": 1.1648072225228586e-05, "loss": 0.9005, "step": 13939 }, { "epoch": 0.1, "learning_rate": 1.1647009073585669e-05, "loss": 0.8864, "step": 13940 }, { "epoch": 0.1, "learning_rate": 1.164594590280734e-05, "loss": 0.9083, "step": 13941 }, { "epoch": 0.1, "learning_rate": 1.164488271290595e-05, "loss": 0.9024, "step": 13942 }, { "epoch": 0.1, "learning_rate": 1.1643819503893855e-05, "loss": 0.8837, "step": 13943 }, { "epoch": 0.1, "learning_rate": 1.1642756275783405e-05, "loss": 0.9402, "step": 13944 }, { "epoch": 0.1, "learning_rate": 1.1641693028586948e-05, "loss": 0.9226, "step": 13945 }, { "epoch": 0.1, "learning_rate": 1.1640629762316848e-05, "loss": 0.8932, "step": 13946 }, { "epoch": 0.1, "learning_rate": 1.163956647698545e-05, "loss": 0.9151, "step": 13947 }, { "epoch": 0.1, "learning_rate": 1.1638503172605113e-05, "loss": 0.9267, "step": 13948 }, { "epoch": 0.1, "learning_rate": 1.1637439849188185e-05, "loss": 0.9296, "step": 13949 }, { "epoch": 0.1, "learning_rate": 1.1636376506747024e-05, "loss": 0.8866, "step": 13950 }, { "epoch": 0.1, "learning_rate": 1.1635313145293985e-05, "loss": 0.9063, "step": 13951 }, { "epoch": 0.1, "learning_rate": 1.1634249764841417e-05, "loss": 0.8717, "step": 13952 }, { "epoch": 0.1, "learning_rate": 1.1633186365401683e-05, "loss": 0.8939, "step": 13953 }, { "epoch": 0.1, "learning_rate": 1.163212294698713e-05, "loss": 0.878, "step": 13954 }, { "epoch": 0.1, "learning_rate": 1.1631059509610116e-05, "loss": 0.9111, "step": 13955 }, { "epoch": 0.1, "learning_rate": 1.1629996053282998e-05, "loss": 0.9267, "step": 13956 }, { "epoch": 0.1, "learning_rate": 1.1628932578018128e-05, "loss": 0.892, "step": 13957 }, { "epoch": 0.1, "learning_rate": 1.1627869083827865e-05, "loss": 0.8874, "step": 13958 }, { "epoch": 0.1, "learning_rate": 1.1626805570724565e-05, "loss": 0.9097, "step": 13959 }, { "epoch": 0.1, "learning_rate": 1.1625742038720584e-05, "loss": 0.8723, "step": 13960 }, { "epoch": 0.1, "learning_rate": 1.1624678487828272e-05, "loss": 0.9221, "step": 13961 }, { "epoch": 0.1, "learning_rate": 1.1623614918059997e-05, "loss": 0.9009, "step": 13962 }, { "epoch": 0.1, "learning_rate": 1.1622551329428107e-05, "loss": 0.89, "step": 13963 }, { "epoch": 0.1, "learning_rate": 1.1621487721944961e-05, "loss": 0.8499, "step": 13964 }, { "epoch": 0.1, "learning_rate": 1.1620424095622922e-05, "loss": 0.9233, "step": 13965 }, { "epoch": 0.1, "learning_rate": 1.1619360450474339e-05, "loss": 0.8939, "step": 13966 }, { "epoch": 0.1, "learning_rate": 1.1618296786511573e-05, "loss": 0.9191, "step": 13967 }, { "epoch": 0.1, "learning_rate": 1.1617233103746984e-05, "loss": 0.8845, "step": 13968 }, { "epoch": 0.1, "learning_rate": 1.161616940219293e-05, "loss": 0.8609, "step": 13969 }, { "epoch": 0.1, "learning_rate": 1.1615105681861762e-05, "loss": 0.8907, "step": 13970 }, { "epoch": 0.1, "learning_rate": 1.1614041942765852e-05, "loss": 0.9232, "step": 13971 }, { "epoch": 0.1, "learning_rate": 1.1612978184917549e-05, "loss": 0.8585, "step": 13972 }, { "epoch": 0.1, "learning_rate": 1.1611914408329212e-05, "loss": 0.8949, "step": 13973 }, { "epoch": 0.1, "learning_rate": 1.1610850613013205e-05, "loss": 0.9111, "step": 13974 }, { "epoch": 0.1, "learning_rate": 1.1609786798981886e-05, "loss": 0.877, "step": 13975 }, { "epoch": 0.1, "learning_rate": 1.1608722966247611e-05, "loss": 0.8746, "step": 13976 }, { "epoch": 0.1, "learning_rate": 1.1607659114822743e-05, "loss": 0.872, "step": 13977 }, { "epoch": 0.1, "learning_rate": 1.1606595244719644e-05, "loss": 0.9327, "step": 13978 }, { "epoch": 0.1, "learning_rate": 1.1605531355950672e-05, "loss": 0.9768, "step": 13979 }, { "epoch": 0.1, "learning_rate": 1.1604467448528188e-05, "loss": 0.8837, "step": 13980 }, { "epoch": 0.1, "learning_rate": 1.1603403522464553e-05, "loss": 0.8955, "step": 13981 }, { "epoch": 0.1, "learning_rate": 1.1602339577772126e-05, "loss": 0.8968, "step": 13982 }, { "epoch": 0.1, "learning_rate": 1.160127561446327e-05, "loss": 0.9047, "step": 13983 }, { "epoch": 0.1, "learning_rate": 1.1600211632550344e-05, "loss": 0.9262, "step": 13984 }, { "epoch": 0.1, "learning_rate": 1.1599147632045713e-05, "loss": 0.8922, "step": 13985 }, { "epoch": 0.1, "learning_rate": 1.1598083612961738e-05, "loss": 0.9215, "step": 13986 }, { "epoch": 0.1, "learning_rate": 1.1597019575310784e-05, "loss": 0.9037, "step": 13987 }, { "epoch": 0.1, "learning_rate": 1.1595955519105208e-05, "loss": 0.9178, "step": 13988 }, { "epoch": 0.1, "learning_rate": 1.1594891444357374e-05, "loss": 0.9115, "step": 13989 }, { "epoch": 0.1, "learning_rate": 1.1593827351079647e-05, "loss": 0.9042, "step": 13990 }, { "epoch": 0.1, "learning_rate": 1.1592763239284386e-05, "loss": 0.9135, "step": 13991 }, { "epoch": 0.1, "learning_rate": 1.1591699108983957e-05, "loss": 0.8948, "step": 13992 }, { "epoch": 0.1, "learning_rate": 1.1590634960190722e-05, "loss": 0.9232, "step": 13993 }, { "epoch": 0.1, "learning_rate": 1.1589570792917047e-05, "loss": 0.8857, "step": 13994 }, { "epoch": 0.1, "learning_rate": 1.1588506607175291e-05, "loss": 0.9079, "step": 13995 }, { "epoch": 0.1, "learning_rate": 1.1587442402977824e-05, "loss": 0.9028, "step": 13996 }, { "epoch": 0.1, "learning_rate": 1.1586378180337005e-05, "loss": 0.9191, "step": 13997 }, { "epoch": 0.1, "learning_rate": 1.1585313939265204e-05, "loss": 0.8929, "step": 13998 }, { "epoch": 0.1, "learning_rate": 1.1584249679774781e-05, "loss": 0.8872, "step": 13999 }, { "epoch": 0.1, "learning_rate": 1.15831854018781e-05, "loss": 0.9305, "step": 14000 }, { "epoch": 0.1, "learning_rate": 1.1582121105587532e-05, "loss": 0.8844, "step": 14001 }, { "epoch": 0.1, "learning_rate": 1.1581056790915437e-05, "loss": 0.8981, "step": 14002 }, { "epoch": 0.1, "learning_rate": 1.1579992457874182e-05, "loss": 0.9261, "step": 14003 }, { "epoch": 0.1, "learning_rate": 1.1578928106476133e-05, "loss": 0.907, "step": 14004 }, { "epoch": 0.1, "learning_rate": 1.1577863736733654e-05, "loss": 0.9013, "step": 14005 }, { "epoch": 0.1, "learning_rate": 1.1576799348659118e-05, "loss": 0.9128, "step": 14006 }, { "epoch": 0.1, "learning_rate": 1.157573494226488e-05, "loss": 0.9116, "step": 14007 }, { "epoch": 0.1, "learning_rate": 1.1574670517563315e-05, "loss": 0.879, "step": 14008 }, { "epoch": 0.1, "learning_rate": 1.157360607456679e-05, "loss": 0.8636, "step": 14009 }, { "epoch": 0.1, "learning_rate": 1.1572541613287667e-05, "loss": 0.912, "step": 14010 }, { "epoch": 0.1, "learning_rate": 1.1571477133738318e-05, "loss": 0.8871, "step": 14011 }, { "epoch": 0.1, "learning_rate": 1.1570412635931108e-05, "loss": 0.8612, "step": 14012 }, { "epoch": 0.1, "learning_rate": 1.1569348119878404e-05, "loss": 0.9135, "step": 14013 }, { "epoch": 0.1, "learning_rate": 1.1568283585592573e-05, "loss": 0.8895, "step": 14014 }, { "epoch": 0.1, "learning_rate": 1.1567219033085988e-05, "loss": 0.9123, "step": 14015 }, { "epoch": 0.1, "learning_rate": 1.156615446237101e-05, "loss": 0.9012, "step": 14016 }, { "epoch": 0.1, "learning_rate": 1.1565089873460016e-05, "loss": 0.8861, "step": 14017 }, { "epoch": 0.1, "learning_rate": 1.156402526636537e-05, "loss": 0.8963, "step": 14018 }, { "epoch": 0.1, "learning_rate": 1.1562960641099437e-05, "loss": 0.867, "step": 14019 }, { "epoch": 0.1, "learning_rate": 1.1561895997674594e-05, "loss": 0.8643, "step": 14020 }, { "epoch": 0.1, "learning_rate": 1.1560831336103205e-05, "loss": 0.8954, "step": 14021 }, { "epoch": 0.1, "learning_rate": 1.1559766656397641e-05, "loss": 0.9045, "step": 14022 }, { "epoch": 0.1, "learning_rate": 1.1558701958570274e-05, "loss": 0.9512, "step": 14023 }, { "epoch": 0.1, "learning_rate": 1.155763724263347e-05, "loss": 0.9269, "step": 14024 }, { "epoch": 0.1, "learning_rate": 1.1556572508599602e-05, "loss": 0.9084, "step": 14025 }, { "epoch": 0.1, "learning_rate": 1.1555507756481036e-05, "loss": 0.8933, "step": 14026 }, { "epoch": 0.1, "learning_rate": 1.155444298629015e-05, "loss": 0.8845, "step": 14027 }, { "epoch": 0.1, "learning_rate": 1.1553378198039308e-05, "loss": 0.8793, "step": 14028 }, { "epoch": 0.1, "learning_rate": 1.1552313391740888e-05, "loss": 0.8903, "step": 14029 }, { "epoch": 0.1, "learning_rate": 1.1551248567407254e-05, "loss": 0.9054, "step": 14030 }, { "epoch": 0.1, "learning_rate": 1.1550183725050779e-05, "loss": 0.9375, "step": 14031 }, { "epoch": 0.1, "learning_rate": 1.1549118864683837e-05, "loss": 0.9151, "step": 14032 }, { "epoch": 0.1, "learning_rate": 1.1548053986318801e-05, "loss": 0.9342, "step": 14033 }, { "epoch": 0.1, "learning_rate": 1.154698908996804e-05, "loss": 0.9393, "step": 14034 }, { "epoch": 0.1, "learning_rate": 1.1545924175643926e-05, "loss": 0.9331, "step": 14035 }, { "epoch": 0.1, "learning_rate": 1.1544859243358835e-05, "loss": 0.9169, "step": 14036 }, { "epoch": 0.1, "learning_rate": 1.1543794293125135e-05, "loss": 0.9003, "step": 14037 }, { "epoch": 0.1, "learning_rate": 1.1542729324955201e-05, "loss": 0.8854, "step": 14038 }, { "epoch": 0.1, "learning_rate": 1.1541664338861407e-05, "loss": 0.8821, "step": 14039 }, { "epoch": 0.1, "learning_rate": 1.1540599334856129e-05, "loss": 0.8681, "step": 14040 }, { "epoch": 0.1, "learning_rate": 1.1539534312951731e-05, "loss": 0.9069, "step": 14041 }, { "epoch": 0.1, "learning_rate": 1.1538469273160597e-05, "loss": 0.8996, "step": 14042 }, { "epoch": 0.1, "learning_rate": 1.1537404215495095e-05, "loss": 0.964, "step": 14043 }, { "epoch": 0.1, "learning_rate": 1.1536339139967604e-05, "loss": 0.9121, "step": 14044 }, { "epoch": 0.1, "learning_rate": 1.1535274046590492e-05, "loss": 0.9125, "step": 14045 }, { "epoch": 0.1, "learning_rate": 1.1534208935376138e-05, "loss": 0.8787, "step": 14046 }, { "epoch": 0.1, "learning_rate": 1.1533143806336915e-05, "loss": 0.8674, "step": 14047 }, { "epoch": 0.1, "learning_rate": 1.1532078659485202e-05, "loss": 0.9088, "step": 14048 }, { "epoch": 0.1, "learning_rate": 1.1531013494833367e-05, "loss": 0.8756, "step": 14049 }, { "epoch": 0.1, "learning_rate": 1.152994831239379e-05, "loss": 0.8936, "step": 14050 }, { "epoch": 0.1, "learning_rate": 1.1528883112178846e-05, "loss": 0.9127, "step": 14051 }, { "epoch": 0.1, "learning_rate": 1.1527817894200912e-05, "loss": 0.87, "step": 14052 }, { "epoch": 0.1, "learning_rate": 1.1526752658472358e-05, "loss": 0.8527, "step": 14053 }, { "epoch": 0.1, "learning_rate": 1.152568740500557e-05, "loss": 0.8701, "step": 14054 }, { "epoch": 0.1, "learning_rate": 1.1524622133812915e-05, "loss": 0.9056, "step": 14055 }, { "epoch": 0.1, "learning_rate": 1.1523556844906777e-05, "loss": 0.8957, "step": 14056 }, { "epoch": 0.1, "learning_rate": 1.1522491538299527e-05, "loss": 0.8973, "step": 14057 }, { "epoch": 0.1, "learning_rate": 1.1521426214003546e-05, "loss": 0.9087, "step": 14058 }, { "epoch": 0.1, "learning_rate": 1.1520360872031208e-05, "loss": 0.9, "step": 14059 }, { "epoch": 0.1, "learning_rate": 1.1519295512394895e-05, "loss": 0.9424, "step": 14060 }, { "epoch": 0.1, "learning_rate": 1.1518230135106983e-05, "loss": 0.8589, "step": 14061 }, { "epoch": 0.1, "learning_rate": 1.1517164740179843e-05, "loss": 0.9174, "step": 14062 }, { "epoch": 0.1, "learning_rate": 1.1516099327625865e-05, "loss": 0.9181, "step": 14063 }, { "epoch": 0.1, "learning_rate": 1.151503389745742e-05, "loss": 0.9081, "step": 14064 }, { "epoch": 0.1, "learning_rate": 1.1513968449686882e-05, "loss": 0.8845, "step": 14065 }, { "epoch": 0.1, "learning_rate": 1.1512902984326644e-05, "loss": 0.9207, "step": 14066 }, { "epoch": 0.1, "learning_rate": 1.151183750138907e-05, "loss": 0.9025, "step": 14067 }, { "epoch": 0.1, "learning_rate": 1.1510772000886545e-05, "loss": 0.9035, "step": 14068 }, { "epoch": 0.1, "learning_rate": 1.150970648283145e-05, "loss": 0.8895, "step": 14069 }, { "epoch": 0.1, "learning_rate": 1.1508640947236165e-05, "loss": 0.9179, "step": 14070 }, { "epoch": 0.1, "learning_rate": 1.1507575394113066e-05, "loss": 0.9376, "step": 14071 }, { "epoch": 0.1, "learning_rate": 1.1506509823474534e-05, "loss": 0.8965, "step": 14072 }, { "epoch": 0.1, "learning_rate": 1.1505444235332951e-05, "loss": 0.8962, "step": 14073 }, { "epoch": 0.1, "learning_rate": 1.1504378629700693e-05, "loss": 0.9025, "step": 14074 }, { "epoch": 0.1, "learning_rate": 1.1503313006590146e-05, "loss": 0.8888, "step": 14075 }, { "epoch": 0.1, "learning_rate": 1.1502247366013689e-05, "loss": 0.9232, "step": 14076 }, { "epoch": 0.1, "learning_rate": 1.1501181707983699e-05, "loss": 0.8943, "step": 14077 }, { "epoch": 0.1, "learning_rate": 1.1500116032512562e-05, "loss": 0.8925, "step": 14078 }, { "epoch": 0.1, "learning_rate": 1.1499050339612659e-05, "loss": 0.8821, "step": 14079 }, { "epoch": 0.1, "learning_rate": 1.1497984629296367e-05, "loss": 0.8994, "step": 14080 }, { "epoch": 0.1, "learning_rate": 1.1496918901576073e-05, "loss": 0.8866, "step": 14081 }, { "epoch": 0.1, "learning_rate": 1.1495853156464156e-05, "loss": 0.9043, "step": 14082 }, { "epoch": 0.1, "learning_rate": 1.1494787393973e-05, "loss": 0.8942, "step": 14083 }, { "epoch": 0.1, "learning_rate": 1.1493721614114983e-05, "loss": 0.9062, "step": 14084 }, { "epoch": 0.1, "learning_rate": 1.1492655816902492e-05, "loss": 0.9501, "step": 14085 }, { "epoch": 0.1, "learning_rate": 1.149159000234791e-05, "loss": 0.927, "step": 14086 }, { "epoch": 0.1, "learning_rate": 1.1490524170463617e-05, "loss": 0.9128, "step": 14087 }, { "epoch": 0.1, "learning_rate": 1.1489458321261996e-05, "loss": 0.8999, "step": 14088 }, { "epoch": 0.1, "learning_rate": 1.1488392454755433e-05, "loss": 0.8944, "step": 14089 }, { "epoch": 0.1, "learning_rate": 1.148732657095631e-05, "loss": 0.9222, "step": 14090 }, { "epoch": 0.1, "learning_rate": 1.1486260669877014e-05, "loss": 0.9234, "step": 14091 }, { "epoch": 0.1, "learning_rate": 1.1485194751529921e-05, "loss": 0.8986, "step": 14092 }, { "epoch": 0.1, "learning_rate": 1.1484128815927425e-05, "loss": 0.9631, "step": 14093 }, { "epoch": 0.1, "learning_rate": 1.1483062863081904e-05, "loss": 0.8954, "step": 14094 }, { "epoch": 0.1, "learning_rate": 1.1481996893005742e-05, "loss": 0.8786, "step": 14095 }, { "epoch": 0.1, "learning_rate": 1.1480930905711327e-05, "loss": 0.9352, "step": 14096 }, { "epoch": 0.1, "learning_rate": 1.1479864901211041e-05, "loss": 0.8983, "step": 14097 }, { "epoch": 0.1, "learning_rate": 1.1478798879517275e-05, "loss": 0.8974, "step": 14098 }, { "epoch": 0.1, "learning_rate": 1.1477732840642403e-05, "loss": 0.8926, "step": 14099 }, { "epoch": 0.1, "learning_rate": 1.1476666784598825e-05, "loss": 0.9218, "step": 14100 }, { "epoch": 0.1, "learning_rate": 1.1475600711398916e-05, "loss": 0.882, "step": 14101 }, { "epoch": 0.1, "learning_rate": 1.1474534621055065e-05, "loss": 0.9166, "step": 14102 }, { "epoch": 0.1, "learning_rate": 1.1473468513579663e-05, "loss": 0.8974, "step": 14103 }, { "epoch": 0.1, "learning_rate": 1.1472402388985086e-05, "loss": 0.9088, "step": 14104 }, { "epoch": 0.1, "learning_rate": 1.1471336247283729e-05, "loss": 0.8966, "step": 14105 }, { "epoch": 0.1, "learning_rate": 1.1470270088487974e-05, "loss": 0.9188, "step": 14106 }, { "epoch": 0.1, "learning_rate": 1.1469203912610214e-05, "loss": 0.8756, "step": 14107 }, { "epoch": 0.1, "learning_rate": 1.1468137719662827e-05, "loss": 0.8704, "step": 14108 }, { "epoch": 0.1, "learning_rate": 1.146707150965821e-05, "loss": 0.878, "step": 14109 }, { "epoch": 0.1, "learning_rate": 1.1466005282608745e-05, "loss": 0.8963, "step": 14110 }, { "epoch": 0.1, "learning_rate": 1.1464939038526819e-05, "loss": 0.8983, "step": 14111 }, { "epoch": 0.1, "learning_rate": 1.1463872777424822e-05, "loss": 0.9194, "step": 14112 }, { "epoch": 0.1, "learning_rate": 1.1462806499315144e-05, "loss": 0.9104, "step": 14113 }, { "epoch": 0.1, "learning_rate": 1.1461740204210167e-05, "loss": 0.917, "step": 14114 }, { "epoch": 0.1, "learning_rate": 1.1460673892122288e-05, "loss": 0.9126, "step": 14115 }, { "epoch": 0.1, "learning_rate": 1.145960756306389e-05, "loss": 0.8979, "step": 14116 }, { "epoch": 0.1, "learning_rate": 1.1458541217047364e-05, "loss": 0.894, "step": 14117 }, { "epoch": 0.1, "learning_rate": 1.1457474854085095e-05, "loss": 0.8571, "step": 14118 }, { "epoch": 0.1, "learning_rate": 1.145640847418948e-05, "loss": 0.9069, "step": 14119 }, { "epoch": 0.1, "learning_rate": 1.1455342077372897e-05, "loss": 0.908, "step": 14120 }, { "epoch": 0.1, "learning_rate": 1.145427566364775e-05, "loss": 0.9002, "step": 14121 }, { "epoch": 0.1, "learning_rate": 1.145320923302642e-05, "loss": 0.8909, "step": 14122 }, { "epoch": 0.1, "learning_rate": 1.1452142785521296e-05, "loss": 0.86, "step": 14123 }, { "epoch": 0.1, "learning_rate": 1.1451076321144774e-05, "loss": 0.9113, "step": 14124 }, { "epoch": 0.1, "learning_rate": 1.1450009839909241e-05, "loss": 0.8795, "step": 14125 }, { "epoch": 0.1, "learning_rate": 1.144894334182709e-05, "loss": 0.8684, "step": 14126 }, { "epoch": 0.1, "learning_rate": 1.1447876826910706e-05, "loss": 0.9224, "step": 14127 }, { "epoch": 0.1, "learning_rate": 1.144681029517249e-05, "loss": 0.8797, "step": 14128 }, { "epoch": 0.1, "learning_rate": 1.1445743746624822e-05, "loss": 0.9306, "step": 14129 }, { "epoch": 0.1, "learning_rate": 1.1444677181280102e-05, "loss": 0.8838, "step": 14130 }, { "epoch": 0.1, "learning_rate": 1.1443610599150716e-05, "loss": 0.9145, "step": 14131 }, { "epoch": 0.1, "learning_rate": 1.1442544000249058e-05, "loss": 0.9002, "step": 14132 }, { "epoch": 0.1, "learning_rate": 1.1441477384587525e-05, "loss": 0.9005, "step": 14133 }, { "epoch": 0.1, "learning_rate": 1.1440410752178502e-05, "loss": 0.8693, "step": 14134 }, { "epoch": 0.1, "learning_rate": 1.1439344103034381e-05, "loss": 0.9227, "step": 14135 }, { "epoch": 0.1, "learning_rate": 1.1438277437167562e-05, "loss": 0.9016, "step": 14136 }, { "epoch": 0.1, "learning_rate": 1.1437210754590435e-05, "loss": 0.9088, "step": 14137 }, { "epoch": 0.1, "learning_rate": 1.1436144055315388e-05, "loss": 0.8758, "step": 14138 }, { "epoch": 0.1, "learning_rate": 1.1435077339354818e-05, "loss": 0.8911, "step": 14139 }, { "epoch": 0.1, "learning_rate": 1.1434010606721121e-05, "loss": 0.8972, "step": 14140 }, { "epoch": 0.1, "learning_rate": 1.1432943857426686e-05, "loss": 0.9241, "step": 14141 }, { "epoch": 0.1, "learning_rate": 1.1431877091483907e-05, "loss": 0.9138, "step": 14142 }, { "epoch": 0.1, "learning_rate": 1.1430810308905182e-05, "loss": 0.9368, "step": 14143 }, { "epoch": 0.1, "learning_rate": 1.1429743509702904e-05, "loss": 0.8818, "step": 14144 }, { "epoch": 0.1, "learning_rate": 1.142867669388946e-05, "loss": 0.8873, "step": 14145 }, { "epoch": 0.1, "learning_rate": 1.1427609861477258e-05, "loss": 0.8722, "step": 14146 }, { "epoch": 0.1, "learning_rate": 1.1426543012478682e-05, "loss": 0.9346, "step": 14147 }, { "epoch": 0.1, "learning_rate": 1.1425476146906129e-05, "loss": 0.9113, "step": 14148 }, { "epoch": 0.1, "learning_rate": 1.1424409264771998e-05, "loss": 0.9246, "step": 14149 }, { "epoch": 0.1, "learning_rate": 1.1423342366088681e-05, "loss": 0.9063, "step": 14150 }, { "epoch": 0.1, "learning_rate": 1.1422275450868574e-05, "loss": 0.9167, "step": 14151 }, { "epoch": 0.1, "learning_rate": 1.1421208519124074e-05, "loss": 0.9193, "step": 14152 }, { "epoch": 0.1, "learning_rate": 1.1420141570867579e-05, "loss": 0.9369, "step": 14153 }, { "epoch": 0.1, "learning_rate": 1.1419074606111474e-05, "loss": 0.911, "step": 14154 }, { "epoch": 0.1, "learning_rate": 1.1418007624868171e-05, "loss": 0.8719, "step": 14155 }, { "epoch": 0.11, "learning_rate": 1.1416940627150057e-05, "loss": 0.9116, "step": 14156 }, { "epoch": 0.11, "learning_rate": 1.1415873612969527e-05, "loss": 0.8787, "step": 14157 }, { "epoch": 0.11, "learning_rate": 1.1414806582338986e-05, "loss": 0.9308, "step": 14158 }, { "epoch": 0.11, "learning_rate": 1.1413739535270825e-05, "loss": 0.943, "step": 14159 }, { "epoch": 0.11, "learning_rate": 1.1412672471777441e-05, "loss": 0.9328, "step": 14160 }, { "epoch": 0.11, "learning_rate": 1.1411605391871236e-05, "loss": 0.9143, "step": 14161 }, { "epoch": 0.11, "learning_rate": 1.1410538295564606e-05, "loss": 0.9141, "step": 14162 }, { "epoch": 0.11, "learning_rate": 1.1409471182869945e-05, "loss": 0.9291, "step": 14163 }, { "epoch": 0.11, "learning_rate": 1.1408404053799653e-05, "loss": 0.8836, "step": 14164 }, { "epoch": 0.11, "learning_rate": 1.140733690836613e-05, "loss": 0.9317, "step": 14165 }, { "epoch": 0.11, "learning_rate": 1.1406269746581773e-05, "loss": 0.8901, "step": 14166 }, { "epoch": 0.11, "learning_rate": 1.1405202568458982e-05, "loss": 0.9102, "step": 14167 }, { "epoch": 0.11, "learning_rate": 1.1404135374010155e-05, "loss": 0.8987, "step": 14168 }, { "epoch": 0.11, "learning_rate": 1.1403068163247688e-05, "loss": 0.9538, "step": 14169 }, { "epoch": 0.11, "learning_rate": 1.1402000936183983e-05, "loss": 0.8825, "step": 14170 }, { "epoch": 0.11, "learning_rate": 1.1400933692831444e-05, "loss": 0.8805, "step": 14171 }, { "epoch": 0.11, "learning_rate": 1.139986643320246e-05, "loss": 0.878, "step": 14172 }, { "epoch": 0.11, "learning_rate": 1.139879915730944e-05, "loss": 0.8664, "step": 14173 }, { "epoch": 0.11, "learning_rate": 1.1397731865164778e-05, "loss": 0.8729, "step": 14174 }, { "epoch": 0.11, "learning_rate": 1.1396664556780879e-05, "loss": 0.8865, "step": 14175 }, { "epoch": 0.11, "learning_rate": 1.1395597232170137e-05, "loss": 0.8528, "step": 14176 }, { "epoch": 0.11, "learning_rate": 1.139452989134496e-05, "loss": 0.9532, "step": 14177 }, { "epoch": 0.11, "learning_rate": 1.139346253431774e-05, "loss": 0.8904, "step": 14178 }, { "epoch": 0.11, "learning_rate": 1.1392395161100889e-05, "loss": 0.9029, "step": 14179 }, { "epoch": 0.11, "learning_rate": 1.1391327771706798e-05, "loss": 0.9008, "step": 14180 }, { "epoch": 0.11, "learning_rate": 1.139026036614787e-05, "loss": 0.9023, "step": 14181 }, { "epoch": 0.11, "learning_rate": 1.1389192944436509e-05, "loss": 0.9262, "step": 14182 }, { "epoch": 0.11, "learning_rate": 1.1388125506585119e-05, "loss": 0.8593, "step": 14183 }, { "epoch": 0.11, "learning_rate": 1.1387058052606097e-05, "loss": 0.9194, "step": 14184 }, { "epoch": 0.11, "learning_rate": 1.1385990582511847e-05, "loss": 0.9107, "step": 14185 }, { "epoch": 0.11, "learning_rate": 1.1384923096314773e-05, "loss": 0.8891, "step": 14186 }, { "epoch": 0.11, "learning_rate": 1.1383855594027274e-05, "loss": 0.8746, "step": 14187 }, { "epoch": 0.11, "learning_rate": 1.1382788075661751e-05, "loss": 0.9197, "step": 14188 }, { "epoch": 0.11, "learning_rate": 1.138172054123061e-05, "loss": 0.9308, "step": 14189 }, { "epoch": 0.11, "learning_rate": 1.1380652990746256e-05, "loss": 0.9327, "step": 14190 }, { "epoch": 0.11, "learning_rate": 1.1379585424221087e-05, "loss": 0.8972, "step": 14191 }, { "epoch": 0.11, "learning_rate": 1.1378517841667511e-05, "loss": 0.9026, "step": 14192 }, { "epoch": 0.11, "learning_rate": 1.1377450243097928e-05, "loss": 0.9289, "step": 14193 }, { "epoch": 0.11, "learning_rate": 1.1376382628524743e-05, "loss": 0.8693, "step": 14194 }, { "epoch": 0.11, "learning_rate": 1.1375314997960362e-05, "loss": 0.8736, "step": 14195 }, { "epoch": 0.11, "learning_rate": 1.1374247351417186e-05, "loss": 0.8864, "step": 14196 }, { "epoch": 0.11, "learning_rate": 1.1373179688907616e-05, "loss": 0.863, "step": 14197 }, { "epoch": 0.11, "learning_rate": 1.1372112010444062e-05, "loss": 0.8897, "step": 14198 }, { "epoch": 0.11, "learning_rate": 1.137104431603893e-05, "loss": 0.882, "step": 14199 }, { "epoch": 0.11, "learning_rate": 1.136997660570462e-05, "loss": 0.9466, "step": 14200 }, { "epoch": 0.11, "learning_rate": 1.1368908879453539e-05, "loss": 0.8856, "step": 14201 }, { "epoch": 0.11, "learning_rate": 1.1367841137298095e-05, "loss": 0.8846, "step": 14202 }, { "epoch": 0.11, "learning_rate": 1.1366773379250685e-05, "loss": 0.8945, "step": 14203 }, { "epoch": 0.11, "learning_rate": 1.1365705605323723e-05, "loss": 0.8865, "step": 14204 }, { "epoch": 0.11, "learning_rate": 1.1364637815529609e-05, "loss": 0.8645, "step": 14205 }, { "epoch": 0.11, "learning_rate": 1.1363570009880752e-05, "loss": 0.8696, "step": 14206 }, { "epoch": 0.11, "learning_rate": 1.1362502188389557e-05, "loss": 0.8807, "step": 14207 }, { "epoch": 0.11, "learning_rate": 1.1361434351068432e-05, "loss": 0.8787, "step": 14208 }, { "epoch": 0.11, "learning_rate": 1.1360366497929783e-05, "loss": 0.8572, "step": 14209 }, { "epoch": 0.11, "learning_rate": 1.1359298628986014e-05, "loss": 0.8732, "step": 14210 }, { "epoch": 0.11, "learning_rate": 1.1358230744249534e-05, "loss": 0.8849, "step": 14211 }, { "epoch": 0.11, "learning_rate": 1.1357162843732747e-05, "loss": 0.8775, "step": 14212 }, { "epoch": 0.11, "learning_rate": 1.1356094927448065e-05, "loss": 0.894, "step": 14213 }, { "epoch": 0.11, "learning_rate": 1.1355026995407893e-05, "loss": 0.8957, "step": 14214 }, { "epoch": 0.11, "learning_rate": 1.1353959047624639e-05, "loss": 0.8598, "step": 14215 }, { "epoch": 0.11, "learning_rate": 1.1352891084110707e-05, "loss": 0.8886, "step": 14216 }, { "epoch": 0.11, "learning_rate": 1.1351823104878511e-05, "loss": 0.887, "step": 14217 }, { "epoch": 0.11, "learning_rate": 1.1350755109940457e-05, "loss": 0.9059, "step": 14218 }, { "epoch": 0.11, "learning_rate": 1.134968709930895e-05, "loss": 0.9291, "step": 14219 }, { "epoch": 0.11, "learning_rate": 1.1348619072996405e-05, "loss": 0.873, "step": 14220 }, { "epoch": 0.11, "learning_rate": 1.134755103101522e-05, "loss": 0.9101, "step": 14221 }, { "epoch": 0.11, "learning_rate": 1.1346482973377816e-05, "loss": 0.9326, "step": 14222 }, { "epoch": 0.11, "learning_rate": 1.1345414900096595e-05, "loss": 0.8953, "step": 14223 }, { "epoch": 0.11, "learning_rate": 1.1344346811183967e-05, "loss": 0.9092, "step": 14224 }, { "epoch": 0.11, "learning_rate": 1.134327870665234e-05, "loss": 0.8948, "step": 14225 }, { "epoch": 0.11, "learning_rate": 1.134221058651413e-05, "loss": 0.9039, "step": 14226 }, { "epoch": 0.11, "learning_rate": 1.1341142450781736e-05, "loss": 0.9426, "step": 14227 }, { "epoch": 0.11, "learning_rate": 1.1340074299467578e-05, "loss": 0.919, "step": 14228 }, { "epoch": 0.11, "learning_rate": 1.1339006132584062e-05, "loss": 0.8429, "step": 14229 }, { "epoch": 0.11, "learning_rate": 1.1337937950143598e-05, "loss": 0.891, "step": 14230 }, { "epoch": 0.11, "learning_rate": 1.1336869752158596e-05, "loss": 0.9219, "step": 14231 }, { "epoch": 0.11, "learning_rate": 1.1335801538641468e-05, "loss": 0.8779, "step": 14232 }, { "epoch": 0.11, "learning_rate": 1.1334733309604624e-05, "loss": 0.916, "step": 14233 }, { "epoch": 0.11, "learning_rate": 1.1333665065060476e-05, "loss": 0.9036, "step": 14234 }, { "epoch": 0.11, "learning_rate": 1.1332596805021434e-05, "loss": 0.8705, "step": 14235 }, { "epoch": 0.11, "learning_rate": 1.1331528529499909e-05, "loss": 0.9087, "step": 14236 }, { "epoch": 0.11, "learning_rate": 1.1330460238508311e-05, "loss": 0.9354, "step": 14237 }, { "epoch": 0.11, "learning_rate": 1.1329391932059059e-05, "loss": 0.8902, "step": 14238 }, { "epoch": 0.11, "learning_rate": 1.1328323610164559e-05, "loss": 0.9028, "step": 14239 }, { "epoch": 0.11, "learning_rate": 1.1327255272837221e-05, "loss": 0.8707, "step": 14240 }, { "epoch": 0.11, "learning_rate": 1.1326186920089463e-05, "loss": 0.9491, "step": 14241 }, { "epoch": 0.11, "learning_rate": 1.1325118551933692e-05, "loss": 0.9125, "step": 14242 }, { "epoch": 0.11, "learning_rate": 1.1324050168382324e-05, "loss": 0.8947, "step": 14243 }, { "epoch": 0.11, "learning_rate": 1.1322981769447771e-05, "loss": 0.9418, "step": 14244 }, { "epoch": 0.11, "learning_rate": 1.1321913355142446e-05, "loss": 0.8788, "step": 14245 }, { "epoch": 0.11, "learning_rate": 1.1320844925478763e-05, "loss": 0.9158, "step": 14246 }, { "epoch": 0.11, "learning_rate": 1.1319776480469133e-05, "loss": 0.9304, "step": 14247 }, { "epoch": 0.11, "learning_rate": 1.1318708020125973e-05, "loss": 0.9169, "step": 14248 }, { "epoch": 0.11, "learning_rate": 1.131763954446169e-05, "loss": 0.8948, "step": 14249 }, { "epoch": 0.11, "learning_rate": 1.1316571053488708e-05, "loss": 0.8725, "step": 14250 }, { "epoch": 0.11, "learning_rate": 1.1315502547219431e-05, "loss": 0.8623, "step": 14251 }, { "epoch": 0.11, "learning_rate": 1.1314434025666278e-05, "loss": 0.8975, "step": 14252 }, { "epoch": 0.11, "learning_rate": 1.1313365488841665e-05, "loss": 0.8623, "step": 14253 }, { "epoch": 0.11, "learning_rate": 1.1312296936758005e-05, "loss": 0.9234, "step": 14254 }, { "epoch": 0.11, "learning_rate": 1.1311228369427712e-05, "loss": 0.9164, "step": 14255 }, { "epoch": 0.11, "learning_rate": 1.1310159786863198e-05, "loss": 0.897, "step": 14256 }, { "epoch": 0.11, "learning_rate": 1.1309091189076887e-05, "loss": 0.8864, "step": 14257 }, { "epoch": 0.11, "learning_rate": 1.1308022576081181e-05, "loss": 0.8961, "step": 14258 }, { "epoch": 0.11, "learning_rate": 1.1306953947888508e-05, "loss": 0.8909, "step": 14259 }, { "epoch": 0.11, "learning_rate": 1.1305885304511277e-05, "loss": 0.8789, "step": 14260 }, { "epoch": 0.11, "learning_rate": 1.1304816645961904e-05, "loss": 0.868, "step": 14261 }, { "epoch": 0.11, "learning_rate": 1.130374797225281e-05, "loss": 0.9292, "step": 14262 }, { "epoch": 0.11, "learning_rate": 1.1302679283396408e-05, "loss": 0.8886, "step": 14263 }, { "epoch": 0.11, "learning_rate": 1.130161057940511e-05, "loss": 0.8862, "step": 14264 }, { "epoch": 0.11, "learning_rate": 1.1300541860291339e-05, "loss": 0.8931, "step": 14265 }, { "epoch": 0.11, "learning_rate": 1.1299473126067509e-05, "loss": 0.9082, "step": 14266 }, { "epoch": 0.11, "learning_rate": 1.1298404376746033e-05, "loss": 0.9033, "step": 14267 }, { "epoch": 0.11, "learning_rate": 1.1297335612339335e-05, "loss": 0.8801, "step": 14268 }, { "epoch": 0.11, "learning_rate": 1.129626683285983e-05, "loss": 0.9306, "step": 14269 }, { "epoch": 0.11, "learning_rate": 1.1295198038319933e-05, "loss": 0.8694, "step": 14270 }, { "epoch": 0.11, "learning_rate": 1.1294129228732064e-05, "loss": 0.9053, "step": 14271 }, { "epoch": 0.11, "learning_rate": 1.129306040410864e-05, "loss": 0.865, "step": 14272 }, { "epoch": 0.11, "learning_rate": 1.1291991564462077e-05, "loss": 0.9238, "step": 14273 }, { "epoch": 0.11, "learning_rate": 1.1290922709804795e-05, "loss": 0.9052, "step": 14274 }, { "epoch": 0.11, "learning_rate": 1.1289853840149214e-05, "loss": 0.9169, "step": 14275 }, { "epoch": 0.11, "learning_rate": 1.128878495550775e-05, "loss": 0.9061, "step": 14276 }, { "epoch": 0.11, "learning_rate": 1.1287716055892822e-05, "loss": 0.885, "step": 14277 }, { "epoch": 0.11, "learning_rate": 1.1286647141316849e-05, "loss": 0.8905, "step": 14278 }, { "epoch": 0.11, "learning_rate": 1.1285578211792248e-05, "loss": 0.8665, "step": 14279 }, { "epoch": 0.11, "learning_rate": 1.128450926733144e-05, "loss": 0.8932, "step": 14280 }, { "epoch": 0.11, "learning_rate": 1.1283440307946845e-05, "loss": 0.923, "step": 14281 }, { "epoch": 0.11, "learning_rate": 1.128237133365088e-05, "loss": 0.8722, "step": 14282 }, { "epoch": 0.11, "learning_rate": 1.1281302344455967e-05, "loss": 0.8912, "step": 14283 }, { "epoch": 0.11, "learning_rate": 1.128023334037453e-05, "loss": 0.8836, "step": 14284 }, { "epoch": 0.11, "learning_rate": 1.1279164321418978e-05, "loss": 0.8669, "step": 14285 }, { "epoch": 0.11, "learning_rate": 1.1278095287601737e-05, "loss": 0.9172, "step": 14286 }, { "epoch": 0.11, "learning_rate": 1.1277026238935233e-05, "loss": 0.9081, "step": 14287 }, { "epoch": 0.11, "learning_rate": 1.1275957175431878e-05, "loss": 0.8995, "step": 14288 }, { "epoch": 0.11, "learning_rate": 1.1274888097104096e-05, "loss": 0.9329, "step": 14289 }, { "epoch": 0.11, "learning_rate": 1.1273819003964306e-05, "loss": 0.9114, "step": 14290 }, { "epoch": 0.11, "learning_rate": 1.1272749896024932e-05, "loss": 0.8663, "step": 14291 }, { "epoch": 0.11, "learning_rate": 1.1271680773298393e-05, "loss": 0.8898, "step": 14292 }, { "epoch": 0.11, "learning_rate": 1.1270611635797113e-05, "loss": 0.9272, "step": 14293 }, { "epoch": 0.11, "learning_rate": 1.126954248353351e-05, "loss": 0.8586, "step": 14294 }, { "epoch": 0.11, "learning_rate": 1.1268473316520007e-05, "loss": 0.8865, "step": 14295 }, { "epoch": 0.11, "learning_rate": 1.1267404134769031e-05, "loss": 0.9105, "step": 14296 }, { "epoch": 0.11, "learning_rate": 1.1266334938292996e-05, "loss": 0.8805, "step": 14297 }, { "epoch": 0.11, "learning_rate": 1.1265265727104327e-05, "loss": 0.8714, "step": 14298 }, { "epoch": 0.11, "learning_rate": 1.1264196501215448e-05, "loss": 0.9185, "step": 14299 }, { "epoch": 0.11, "learning_rate": 1.1263127260638781e-05, "loss": 0.9333, "step": 14300 }, { "epoch": 0.11, "learning_rate": 1.1262058005386747e-05, "loss": 0.8884, "step": 14301 }, { "epoch": 0.11, "learning_rate": 1.1260988735471773e-05, "loss": 0.9016, "step": 14302 }, { "epoch": 0.11, "learning_rate": 1.1259919450906276e-05, "loss": 0.9286, "step": 14303 }, { "epoch": 0.11, "learning_rate": 1.125885015170268e-05, "loss": 0.8903, "step": 14304 }, { "epoch": 0.11, "learning_rate": 1.1257780837873417e-05, "loss": 0.8698, "step": 14305 }, { "epoch": 0.11, "learning_rate": 1.1256711509430902e-05, "loss": 0.8693, "step": 14306 }, { "epoch": 0.11, "learning_rate": 1.1255642166387562e-05, "loss": 0.9024, "step": 14307 }, { "epoch": 0.11, "learning_rate": 1.1254572808755819e-05, "loss": 0.9312, "step": 14308 }, { "epoch": 0.11, "learning_rate": 1.12535034365481e-05, "loss": 0.9109, "step": 14309 }, { "epoch": 0.11, "learning_rate": 1.1252434049776825e-05, "loss": 0.8367, "step": 14310 }, { "epoch": 0.11, "learning_rate": 1.1251364648454423e-05, "loss": 0.917, "step": 14311 }, { "epoch": 0.11, "learning_rate": 1.1250295232593317e-05, "loss": 0.9038, "step": 14312 }, { "epoch": 0.11, "learning_rate": 1.124922580220593e-05, "loss": 0.9341, "step": 14313 }, { "epoch": 0.11, "learning_rate": 1.1248156357304692e-05, "loss": 0.9159, "step": 14314 }, { "epoch": 0.11, "learning_rate": 1.124708689790202e-05, "loss": 0.8713, "step": 14315 }, { "epoch": 0.11, "learning_rate": 1.1246017424010345e-05, "loss": 0.9208, "step": 14316 }, { "epoch": 0.11, "learning_rate": 1.1244947935642092e-05, "loss": 0.9143, "step": 14317 }, { "epoch": 0.11, "learning_rate": 1.1243878432809684e-05, "loss": 0.8892, "step": 14318 }, { "epoch": 0.11, "learning_rate": 1.124280891552555e-05, "loss": 0.9064, "step": 14319 }, { "epoch": 0.11, "learning_rate": 1.1241739383802112e-05, "loss": 0.8974, "step": 14320 }, { "epoch": 0.11, "learning_rate": 1.1240669837651806e-05, "loss": 0.8939, "step": 14321 }, { "epoch": 0.11, "learning_rate": 1.1239600277087044e-05, "loss": 0.8796, "step": 14322 }, { "epoch": 0.11, "learning_rate": 1.123853070212026e-05, "loss": 0.8715, "step": 14323 }, { "epoch": 0.11, "learning_rate": 1.1237461112763884e-05, "loss": 0.8855, "step": 14324 }, { "epoch": 0.11, "learning_rate": 1.1236391509030333e-05, "loss": 0.868, "step": 14325 }, { "epoch": 0.11, "learning_rate": 1.1235321890932046e-05, "loss": 0.8848, "step": 14326 }, { "epoch": 0.11, "learning_rate": 1.1234252258481439e-05, "loss": 0.8657, "step": 14327 }, { "epoch": 0.11, "learning_rate": 1.1233182611690945e-05, "loss": 0.9135, "step": 14328 }, { "epoch": 0.11, "learning_rate": 1.1232112950572991e-05, "loss": 0.907, "step": 14329 }, { "epoch": 0.11, "learning_rate": 1.1231043275140005e-05, "loss": 0.9051, "step": 14330 }, { "epoch": 0.11, "learning_rate": 1.1229973585404414e-05, "loss": 0.9041, "step": 14331 }, { "epoch": 0.11, "learning_rate": 1.1228903881378646e-05, "loss": 0.937, "step": 14332 }, { "epoch": 0.11, "learning_rate": 1.122783416307513e-05, "loss": 0.8959, "step": 14333 }, { "epoch": 0.11, "learning_rate": 1.1226764430506292e-05, "loss": 0.9065, "step": 14334 }, { "epoch": 0.11, "learning_rate": 1.1225694683684563e-05, "loss": 0.9309, "step": 14335 }, { "epoch": 0.11, "learning_rate": 1.1224624922622367e-05, "loss": 0.8833, "step": 14336 }, { "epoch": 0.11, "learning_rate": 1.122355514733214e-05, "loss": 0.8717, "step": 14337 }, { "epoch": 0.11, "learning_rate": 1.1222485357826304e-05, "loss": 0.913, "step": 14338 }, { "epoch": 0.11, "learning_rate": 1.1221415554117297e-05, "loss": 0.8935, "step": 14339 }, { "epoch": 0.11, "learning_rate": 1.1220345736217537e-05, "loss": 0.8825, "step": 14340 }, { "epoch": 0.11, "learning_rate": 1.1219275904139459e-05, "loss": 0.8702, "step": 14341 }, { "epoch": 0.11, "learning_rate": 1.1218206057895498e-05, "loss": 0.9004, "step": 14342 }, { "epoch": 0.11, "learning_rate": 1.1217136197498073e-05, "loss": 0.9006, "step": 14343 }, { "epoch": 0.11, "learning_rate": 1.1216066322959622e-05, "loss": 0.8924, "step": 14344 }, { "epoch": 0.11, "learning_rate": 1.1214996434292572e-05, "loss": 0.9279, "step": 14345 }, { "epoch": 0.11, "learning_rate": 1.1213926531509353e-05, "loss": 0.9144, "step": 14346 }, { "epoch": 0.11, "learning_rate": 1.1212856614622396e-05, "loss": 0.8824, "step": 14347 }, { "epoch": 0.11, "learning_rate": 1.1211786683644132e-05, "loss": 0.8867, "step": 14348 }, { "epoch": 0.11, "learning_rate": 1.1210716738586995e-05, "loss": 0.9305, "step": 14349 }, { "epoch": 0.11, "learning_rate": 1.1209646779463406e-05, "loss": 0.8845, "step": 14350 }, { "epoch": 0.11, "learning_rate": 1.1208576806285806e-05, "loss": 0.9365, "step": 14351 }, { "epoch": 0.11, "learning_rate": 1.1207506819066622e-05, "loss": 0.9141, "step": 14352 }, { "epoch": 0.11, "learning_rate": 1.1206436817818286e-05, "loss": 0.9191, "step": 14353 }, { "epoch": 0.11, "learning_rate": 1.1205366802553231e-05, "loss": 0.8892, "step": 14354 }, { "epoch": 0.11, "learning_rate": 1.1204296773283889e-05, "loss": 0.9134, "step": 14355 }, { "epoch": 0.11, "learning_rate": 1.1203226730022687e-05, "loss": 0.8927, "step": 14356 }, { "epoch": 0.11, "learning_rate": 1.120215667278206e-05, "loss": 0.9322, "step": 14357 }, { "epoch": 0.11, "learning_rate": 1.1201086601574444e-05, "loss": 0.8868, "step": 14358 }, { "epoch": 0.11, "learning_rate": 1.1200016516412264e-05, "loss": 0.9136, "step": 14359 }, { "epoch": 0.11, "learning_rate": 1.119894641730796e-05, "loss": 0.8974, "step": 14360 }, { "epoch": 0.11, "learning_rate": 1.119787630427396e-05, "loss": 0.8884, "step": 14361 }, { "epoch": 0.11, "learning_rate": 1.11968061773227e-05, "loss": 0.9075, "step": 14362 }, { "epoch": 0.11, "learning_rate": 1.1195736036466609e-05, "loss": 0.8616, "step": 14363 }, { "epoch": 0.11, "learning_rate": 1.1194665881718124e-05, "loss": 0.8639, "step": 14364 }, { "epoch": 0.11, "learning_rate": 1.1193595713089676e-05, "loss": 0.8901, "step": 14365 }, { "epoch": 0.11, "learning_rate": 1.1192525530593696e-05, "loss": 0.9103, "step": 14366 }, { "epoch": 0.11, "learning_rate": 1.1191455334242626e-05, "loss": 0.9314, "step": 14367 }, { "epoch": 0.11, "learning_rate": 1.1190385124048893e-05, "loss": 0.9349, "step": 14368 }, { "epoch": 0.11, "learning_rate": 1.1189314900024932e-05, "loss": 0.9106, "step": 14369 }, { "epoch": 0.11, "learning_rate": 1.1188244662183181e-05, "loss": 0.9428, "step": 14370 }, { "epoch": 0.11, "learning_rate": 1.1187174410536065e-05, "loss": 0.8865, "step": 14371 }, { "epoch": 0.11, "learning_rate": 1.1186104145096031e-05, "loss": 0.9246, "step": 14372 }, { "epoch": 0.11, "learning_rate": 1.1185033865875503e-05, "loss": 0.908, "step": 14373 }, { "epoch": 0.11, "learning_rate": 1.1183963572886923e-05, "loss": 0.8548, "step": 14374 }, { "epoch": 0.11, "learning_rate": 1.1182893266142721e-05, "loss": 0.8581, "step": 14375 }, { "epoch": 0.11, "learning_rate": 1.1181822945655338e-05, "loss": 0.8888, "step": 14376 }, { "epoch": 0.11, "learning_rate": 1.1180752611437203e-05, "loss": 0.9106, "step": 14377 }, { "epoch": 0.11, "learning_rate": 1.1179682263500754e-05, "loss": 0.9007, "step": 14378 }, { "epoch": 0.11, "learning_rate": 1.1178611901858427e-05, "loss": 0.9233, "step": 14379 }, { "epoch": 0.11, "learning_rate": 1.1177541526522658e-05, "loss": 0.8755, "step": 14380 }, { "epoch": 0.11, "learning_rate": 1.117647113750588e-05, "loss": 0.8978, "step": 14381 }, { "epoch": 0.11, "learning_rate": 1.1175400734820531e-05, "loss": 0.8978, "step": 14382 }, { "epoch": 0.11, "learning_rate": 1.117433031847905e-05, "loss": 0.8883, "step": 14383 }, { "epoch": 0.11, "learning_rate": 1.117325988849387e-05, "loss": 0.893, "step": 14384 }, { "epoch": 0.11, "learning_rate": 1.1172189444877431e-05, "loss": 0.9213, "step": 14385 }, { "epoch": 0.11, "learning_rate": 1.1171118987642164e-05, "loss": 0.9114, "step": 14386 }, { "epoch": 0.11, "learning_rate": 1.1170048516800511e-05, "loss": 0.9025, "step": 14387 }, { "epoch": 0.11, "learning_rate": 1.1168978032364908e-05, "loss": 0.8731, "step": 14388 }, { "epoch": 0.11, "learning_rate": 1.116790753434779e-05, "loss": 0.8931, "step": 14389 }, { "epoch": 0.11, "learning_rate": 1.1166837022761595e-05, "loss": 0.8964, "step": 14390 }, { "epoch": 0.11, "learning_rate": 1.1165766497618762e-05, "loss": 0.9086, "step": 14391 }, { "epoch": 0.11, "learning_rate": 1.1164695958931729e-05, "loss": 0.8947, "step": 14392 }, { "epoch": 0.11, "learning_rate": 1.1163625406712932e-05, "loss": 0.8645, "step": 14393 }, { "epoch": 0.11, "learning_rate": 1.1162554840974808e-05, "loss": 0.9084, "step": 14394 }, { "epoch": 0.11, "learning_rate": 1.11614842617298e-05, "loss": 0.8642, "step": 14395 }, { "epoch": 0.11, "learning_rate": 1.1160413668990338e-05, "loss": 0.8627, "step": 14396 }, { "epoch": 0.11, "learning_rate": 1.1159343062768871e-05, "loss": 0.8812, "step": 14397 }, { "epoch": 0.11, "learning_rate": 1.1158272443077828e-05, "loss": 0.881, "step": 14398 }, { "epoch": 0.11, "learning_rate": 1.1157201809929655e-05, "loss": 0.8951, "step": 14399 }, { "epoch": 0.11, "learning_rate": 1.1156131163336785e-05, "loss": 0.9301, "step": 14400 }, { "epoch": 0.11, "learning_rate": 1.1155060503311663e-05, "loss": 0.9196, "step": 14401 }, { "epoch": 0.11, "learning_rate": 1.1153989829866725e-05, "loss": 0.8869, "step": 14402 }, { "epoch": 0.11, "learning_rate": 1.1152919143014406e-05, "loss": 0.8991, "step": 14403 }, { "epoch": 0.11, "learning_rate": 1.1151848442767158e-05, "loss": 0.8758, "step": 14404 }, { "epoch": 0.11, "learning_rate": 1.1150777729137403e-05, "loss": 0.9132, "step": 14405 }, { "epoch": 0.11, "learning_rate": 1.1149707002137598e-05, "loss": 0.9023, "step": 14406 }, { "epoch": 0.11, "learning_rate": 1.1148636261780173e-05, "loss": 0.9154, "step": 14407 }, { "epoch": 0.11, "learning_rate": 1.114756550807757e-05, "loss": 0.8595, "step": 14408 }, { "epoch": 0.11, "learning_rate": 1.1146494741042232e-05, "loss": 0.8827, "step": 14409 }, { "epoch": 0.11, "learning_rate": 1.1145423960686596e-05, "loss": 0.884, "step": 14410 }, { "epoch": 0.11, "learning_rate": 1.1144353167023102e-05, "loss": 0.9064, "step": 14411 }, { "epoch": 0.11, "learning_rate": 1.1143282360064197e-05, "loss": 0.8906, "step": 14412 }, { "epoch": 0.11, "learning_rate": 1.1142211539822318e-05, "loss": 0.8876, "step": 14413 }, { "epoch": 0.11, "learning_rate": 1.11411407063099e-05, "loss": 0.9033, "step": 14414 }, { "epoch": 0.11, "learning_rate": 1.1140069859539398e-05, "loss": 0.8818, "step": 14415 }, { "epoch": 0.11, "learning_rate": 1.1138998999523243e-05, "loss": 0.8741, "step": 14416 }, { "epoch": 0.11, "learning_rate": 1.1137928126273877e-05, "loss": 0.9065, "step": 14417 }, { "epoch": 0.11, "learning_rate": 1.1136857239803747e-05, "loss": 0.9316, "step": 14418 }, { "epoch": 0.11, "learning_rate": 1.1135786340125288e-05, "loss": 0.9254, "step": 14419 }, { "epoch": 0.11, "learning_rate": 1.1134715427250948e-05, "loss": 0.8966, "step": 14420 }, { "epoch": 0.11, "learning_rate": 1.1133644501193166e-05, "loss": 0.8962, "step": 14421 }, { "epoch": 0.11, "learning_rate": 1.1132573561964385e-05, "loss": 0.929, "step": 14422 }, { "epoch": 0.11, "learning_rate": 1.113150260957705e-05, "loss": 0.9164, "step": 14423 }, { "epoch": 0.11, "learning_rate": 1.11304316440436e-05, "loss": 0.8682, "step": 14424 }, { "epoch": 0.11, "learning_rate": 1.1129360665376479e-05, "loss": 0.9357, "step": 14425 }, { "epoch": 0.11, "learning_rate": 1.1128289673588131e-05, "loss": 0.8829, "step": 14426 }, { "epoch": 0.11, "learning_rate": 1.1127218668690995e-05, "loss": 0.9097, "step": 14427 }, { "epoch": 0.11, "learning_rate": 1.1126147650697519e-05, "loss": 0.9007, "step": 14428 }, { "epoch": 0.11, "learning_rate": 1.1125076619620145e-05, "loss": 0.8518, "step": 14429 }, { "epoch": 0.11, "learning_rate": 1.1124005575471316e-05, "loss": 0.8998, "step": 14430 }, { "epoch": 0.11, "learning_rate": 1.1122934518263478e-05, "loss": 0.8885, "step": 14431 }, { "epoch": 0.11, "learning_rate": 1.112186344800907e-05, "loss": 0.893, "step": 14432 }, { "epoch": 0.11, "learning_rate": 1.1120792364720539e-05, "loss": 0.8792, "step": 14433 }, { "epoch": 0.11, "learning_rate": 1.111972126841033e-05, "loss": 0.881, "step": 14434 }, { "epoch": 0.11, "learning_rate": 1.1118650159090888e-05, "loss": 0.8763, "step": 14435 }, { "epoch": 0.11, "learning_rate": 1.1117579036774653e-05, "loss": 0.8907, "step": 14436 }, { "epoch": 0.11, "learning_rate": 1.1116507901474072e-05, "loss": 0.9113, "step": 14437 }, { "epoch": 0.11, "learning_rate": 1.1115436753201592e-05, "loss": 0.9065, "step": 14438 }, { "epoch": 0.11, "learning_rate": 1.1114365591969652e-05, "loss": 0.9219, "step": 14439 }, { "epoch": 0.11, "learning_rate": 1.1113294417790705e-05, "loss": 0.9117, "step": 14440 }, { "epoch": 0.11, "learning_rate": 1.1112223230677193e-05, "loss": 0.9305, "step": 14441 }, { "epoch": 0.11, "learning_rate": 1.1111152030641556e-05, "loss": 0.8799, "step": 14442 }, { "epoch": 0.11, "learning_rate": 1.111008081769625e-05, "loss": 0.9084, "step": 14443 }, { "epoch": 0.11, "learning_rate": 1.110900959185371e-05, "loss": 0.8657, "step": 14444 }, { "epoch": 0.11, "learning_rate": 1.1107938353126385e-05, "loss": 0.863, "step": 14445 }, { "epoch": 0.11, "learning_rate": 1.1106867101526726e-05, "loss": 0.8895, "step": 14446 }, { "epoch": 0.11, "learning_rate": 1.1105795837067177e-05, "loss": 0.8914, "step": 14447 }, { "epoch": 0.11, "learning_rate": 1.1104724559760181e-05, "loss": 0.9122, "step": 14448 }, { "epoch": 0.11, "learning_rate": 1.1103653269618185e-05, "loss": 0.9329, "step": 14449 }, { "epoch": 0.11, "learning_rate": 1.1102581966653638e-05, "loss": 0.9032, "step": 14450 }, { "epoch": 0.11, "learning_rate": 1.1101510650878981e-05, "loss": 0.9133, "step": 14451 }, { "epoch": 0.11, "learning_rate": 1.1100439322306671e-05, "loss": 0.8619, "step": 14452 }, { "epoch": 0.11, "learning_rate": 1.1099367980949146e-05, "loss": 0.9008, "step": 14453 }, { "epoch": 0.11, "learning_rate": 1.109829662681886e-05, "loss": 0.8936, "step": 14454 }, { "epoch": 0.11, "learning_rate": 1.1097225259928255e-05, "loss": 0.8561, "step": 14455 }, { "epoch": 0.12, "learning_rate": 1.109615388028978e-05, "loss": 0.8873, "step": 14456 }, { "epoch": 0.12, "learning_rate": 1.1095082487915881e-05, "loss": 0.8976, "step": 14457 }, { "epoch": 0.12, "learning_rate": 1.1094011082819006e-05, "loss": 0.9493, "step": 14458 }, { "epoch": 0.12, "learning_rate": 1.109293966501161e-05, "loss": 0.8864, "step": 14459 }, { "epoch": 0.12, "learning_rate": 1.1091868234506132e-05, "loss": 0.8561, "step": 14460 }, { "epoch": 0.12, "learning_rate": 1.1090796791315023e-05, "loss": 0.8956, "step": 14461 }, { "epoch": 0.12, "learning_rate": 1.1089725335450734e-05, "loss": 0.8954, "step": 14462 }, { "epoch": 0.12, "learning_rate": 1.108865386692571e-05, "loss": 0.9116, "step": 14463 }, { "epoch": 0.12, "learning_rate": 1.1087582385752402e-05, "loss": 0.9125, "step": 14464 }, { "epoch": 0.12, "learning_rate": 1.1086510891943258e-05, "loss": 0.915, "step": 14465 }, { "epoch": 0.12, "learning_rate": 1.1085439385510726e-05, "loss": 0.8816, "step": 14466 }, { "epoch": 0.12, "learning_rate": 1.1084367866467254e-05, "loss": 0.8896, "step": 14467 }, { "epoch": 0.12, "learning_rate": 1.1083296334825295e-05, "loss": 0.888, "step": 14468 }, { "epoch": 0.12, "learning_rate": 1.1082224790597296e-05, "loss": 0.9067, "step": 14469 }, { "epoch": 0.12, "learning_rate": 1.1081153233795708e-05, "loss": 0.9064, "step": 14470 }, { "epoch": 0.12, "learning_rate": 1.1080081664432978e-05, "loss": 0.8638, "step": 14471 }, { "epoch": 0.12, "learning_rate": 1.1079010082521557e-05, "loss": 0.8906, "step": 14472 }, { "epoch": 0.12, "learning_rate": 1.1077938488073894e-05, "loss": 0.9204, "step": 14473 }, { "epoch": 0.12, "learning_rate": 1.1076866881102441e-05, "loss": 0.8918, "step": 14474 }, { "epoch": 0.12, "learning_rate": 1.1075795261619649e-05, "loss": 0.9081, "step": 14475 }, { "epoch": 0.12, "learning_rate": 1.1074723629637964e-05, "loss": 0.8627, "step": 14476 }, { "epoch": 0.12, "learning_rate": 1.1073651985169845e-05, "loss": 0.901, "step": 14477 }, { "epoch": 0.12, "learning_rate": 1.1072580328227732e-05, "loss": 0.9071, "step": 14478 }, { "epoch": 0.12, "learning_rate": 1.1071508658824082e-05, "loss": 0.924, "step": 14479 }, { "epoch": 0.12, "learning_rate": 1.1070436976971345e-05, "loss": 0.9189, "step": 14480 }, { "epoch": 0.12, "learning_rate": 1.1069365282681973e-05, "loss": 0.9033, "step": 14481 }, { "epoch": 0.12, "learning_rate": 1.1068293575968414e-05, "loss": 0.8944, "step": 14482 }, { "epoch": 0.12, "learning_rate": 1.1067221856843122e-05, "loss": 0.9198, "step": 14483 }, { "epoch": 0.12, "learning_rate": 1.106615012531855e-05, "loss": 0.9177, "step": 14484 }, { "epoch": 0.12, "learning_rate": 1.1065078381407144e-05, "loss": 0.8709, "step": 14485 }, { "epoch": 0.12, "learning_rate": 1.1064006625121365e-05, "loss": 0.9204, "step": 14486 }, { "epoch": 0.12, "learning_rate": 1.1062934856473655e-05, "loss": 0.8817, "step": 14487 }, { "epoch": 0.12, "learning_rate": 1.106186307547647e-05, "loss": 0.8518, "step": 14488 }, { "epoch": 0.12, "learning_rate": 1.106079128214227e-05, "loss": 0.91, "step": 14489 }, { "epoch": 0.12, "learning_rate": 1.1059719476483492e-05, "loss": 0.9167, "step": 14490 }, { "epoch": 0.12, "learning_rate": 1.10586476585126e-05, "loss": 0.8885, "step": 14491 }, { "epoch": 0.12, "learning_rate": 1.1057575828242044e-05, "loss": 0.8873, "step": 14492 }, { "epoch": 0.12, "learning_rate": 1.1056503985684277e-05, "loss": 0.898, "step": 14493 }, { "epoch": 0.12, "learning_rate": 1.105543213085175e-05, "loss": 0.9324, "step": 14494 }, { "epoch": 0.12, "learning_rate": 1.1054360263756918e-05, "loss": 0.8992, "step": 14495 }, { "epoch": 0.12, "learning_rate": 1.1053288384412236e-05, "loss": 0.9042, "step": 14496 }, { "epoch": 0.12, "learning_rate": 1.1052216492830148e-05, "loss": 0.8962, "step": 14497 }, { "epoch": 0.12, "learning_rate": 1.1051144589023122e-05, "loss": 0.9141, "step": 14498 }, { "epoch": 0.12, "learning_rate": 1.10500726730036e-05, "loss": 0.8662, "step": 14499 }, { "epoch": 0.12, "learning_rate": 1.1049000744784041e-05, "loss": 0.8976, "step": 14500 }, { "epoch": 0.12, "learning_rate": 1.10479288043769e-05, "loss": 0.9164, "step": 14501 }, { "epoch": 0.12, "learning_rate": 1.1046856851794627e-05, "loss": 0.9099, "step": 14502 }, { "epoch": 0.12, "learning_rate": 1.1045784887049677e-05, "loss": 0.8766, "step": 14503 }, { "epoch": 0.12, "learning_rate": 1.1044712910154509e-05, "loss": 0.8626, "step": 14504 }, { "epoch": 0.12, "learning_rate": 1.1043640921121574e-05, "loss": 0.8957, "step": 14505 }, { "epoch": 0.12, "learning_rate": 1.1042568919963323e-05, "loss": 0.9132, "step": 14506 }, { "epoch": 0.12, "learning_rate": 1.1041496906692219e-05, "loss": 0.9444, "step": 14507 }, { "epoch": 0.12, "learning_rate": 1.1040424881320713e-05, "loss": 0.9085, "step": 14508 }, { "epoch": 0.12, "learning_rate": 1.1039352843861258e-05, "loss": 0.9095, "step": 14509 }, { "epoch": 0.12, "learning_rate": 1.1038280794326313e-05, "loss": 0.8888, "step": 14510 }, { "epoch": 0.12, "learning_rate": 1.1037208732728328e-05, "loss": 0.9215, "step": 14511 }, { "epoch": 0.12, "learning_rate": 1.1036136659079764e-05, "loss": 0.9037, "step": 14512 }, { "epoch": 0.12, "learning_rate": 1.1035064573393072e-05, "loss": 0.8979, "step": 14513 }, { "epoch": 0.12, "learning_rate": 1.1033992475680718e-05, "loss": 0.8709, "step": 14514 }, { "epoch": 0.12, "learning_rate": 1.1032920365955142e-05, "loss": 0.91, "step": 14515 }, { "epoch": 0.12, "learning_rate": 1.103184824422881e-05, "loss": 0.8701, "step": 14516 }, { "epoch": 0.12, "learning_rate": 1.1030776110514181e-05, "loss": 0.9101, "step": 14517 }, { "epoch": 0.12, "learning_rate": 1.10297039648237e-05, "loss": 0.9229, "step": 14518 }, { "epoch": 0.12, "learning_rate": 1.1028631807169838e-05, "loss": 0.8883, "step": 14519 }, { "epoch": 0.12, "learning_rate": 1.1027559637565039e-05, "loss": 0.8753, "step": 14520 }, { "epoch": 0.12, "learning_rate": 1.1026487456021766e-05, "loss": 0.8859, "step": 14521 }, { "epoch": 0.12, "learning_rate": 1.1025415262552474e-05, "loss": 0.884, "step": 14522 }, { "epoch": 0.12, "learning_rate": 1.1024343057169625e-05, "loss": 0.9167, "step": 14523 }, { "epoch": 0.12, "learning_rate": 1.1023270839885668e-05, "loss": 0.8787, "step": 14524 }, { "epoch": 0.12, "learning_rate": 1.1022198610713064e-05, "loss": 0.9279, "step": 14525 }, { "epoch": 0.12, "learning_rate": 1.1021126369664275e-05, "loss": 0.936, "step": 14526 }, { "epoch": 0.12, "learning_rate": 1.1020054116751749e-05, "loss": 0.9094, "step": 14527 }, { "epoch": 0.12, "learning_rate": 1.1018981851987951e-05, "loss": 0.9012, "step": 14528 }, { "epoch": 0.12, "learning_rate": 1.1017909575385337e-05, "loss": 0.9198, "step": 14529 }, { "epoch": 0.12, "learning_rate": 1.1016837286956365e-05, "loss": 0.8774, "step": 14530 }, { "epoch": 0.12, "learning_rate": 1.101576498671349e-05, "loss": 0.8702, "step": 14531 }, { "epoch": 0.12, "learning_rate": 1.1014692674669179e-05, "loss": 0.8907, "step": 14532 }, { "epoch": 0.12, "learning_rate": 1.1013620350835884e-05, "loss": 0.9166, "step": 14533 }, { "epoch": 0.12, "learning_rate": 1.101254801522606e-05, "loss": 0.8749, "step": 14534 }, { "epoch": 0.12, "learning_rate": 1.1011475667852175e-05, "loss": 0.8989, "step": 14535 }, { "epoch": 0.12, "learning_rate": 1.101040330872668e-05, "loss": 0.8868, "step": 14536 }, { "epoch": 0.12, "learning_rate": 1.1009330937862037e-05, "loss": 0.9208, "step": 14537 }, { "epoch": 0.12, "learning_rate": 1.1008258555270705e-05, "loss": 0.8919, "step": 14538 }, { "epoch": 0.12, "learning_rate": 1.1007186160965144e-05, "loss": 0.8866, "step": 14539 }, { "epoch": 0.12, "learning_rate": 1.1006113754957813e-05, "loss": 0.8902, "step": 14540 }, { "epoch": 0.12, "learning_rate": 1.1005041337261168e-05, "loss": 0.9073, "step": 14541 }, { "epoch": 0.12, "learning_rate": 1.1003968907887676e-05, "loss": 0.8937, "step": 14542 }, { "epoch": 0.12, "learning_rate": 1.1002896466849788e-05, "loss": 0.9051, "step": 14543 }, { "epoch": 0.12, "learning_rate": 1.1001824014159974e-05, "loss": 0.8778, "step": 14544 }, { "epoch": 0.12, "learning_rate": 1.1000751549830684e-05, "loss": 0.9061, "step": 14545 }, { "epoch": 0.12, "learning_rate": 1.0999679073874383e-05, "loss": 0.8707, "step": 14546 }, { "epoch": 0.12, "learning_rate": 1.0998606586303536e-05, "loss": 0.883, "step": 14547 }, { "epoch": 0.12, "learning_rate": 1.0997534087130592e-05, "loss": 0.8881, "step": 14548 }, { "epoch": 0.12, "learning_rate": 1.099646157636802e-05, "loss": 0.8666, "step": 14549 }, { "epoch": 0.12, "learning_rate": 1.099538905402828e-05, "loss": 0.901, "step": 14550 }, { "epoch": 0.12, "learning_rate": 1.0994316520123833e-05, "loss": 0.9434, "step": 14551 }, { "epoch": 0.12, "learning_rate": 1.0993243974667134e-05, "loss": 0.8899, "step": 14552 }, { "epoch": 0.12, "learning_rate": 1.0992171417670652e-05, "loss": 0.9085, "step": 14553 }, { "epoch": 0.12, "learning_rate": 1.0991098849146843e-05, "loss": 0.8746, "step": 14554 }, { "epoch": 0.12, "learning_rate": 1.0990026269108174e-05, "loss": 0.8626, "step": 14555 }, { "epoch": 0.12, "learning_rate": 1.0988953677567102e-05, "loss": 0.9029, "step": 14556 }, { "epoch": 0.12, "learning_rate": 1.0987881074536089e-05, "loss": 0.9056, "step": 14557 }, { "epoch": 0.12, "learning_rate": 1.0986808460027597e-05, "loss": 0.8663, "step": 14558 }, { "epoch": 0.12, "learning_rate": 1.0985735834054088e-05, "loss": 0.8882, "step": 14559 }, { "epoch": 0.12, "learning_rate": 1.0984663196628025e-05, "loss": 0.9561, "step": 14560 }, { "epoch": 0.12, "learning_rate": 1.0983590547761869e-05, "loss": 0.8966, "step": 14561 }, { "epoch": 0.12, "learning_rate": 1.0982517887468083e-05, "loss": 0.8972, "step": 14562 }, { "epoch": 0.12, "learning_rate": 1.0981445215759132e-05, "loss": 0.8776, "step": 14563 }, { "epoch": 0.12, "learning_rate": 1.0980372532647473e-05, "loss": 0.8983, "step": 14564 }, { "epoch": 0.12, "learning_rate": 1.0979299838145573e-05, "loss": 0.9071, "step": 14565 }, { "epoch": 0.12, "learning_rate": 1.0978227132265892e-05, "loss": 0.8795, "step": 14566 }, { "epoch": 0.12, "learning_rate": 1.0977154415020899e-05, "loss": 0.9051, "step": 14567 }, { "epoch": 0.12, "learning_rate": 1.0976081686423049e-05, "loss": 0.8958, "step": 14568 }, { "epoch": 0.12, "learning_rate": 1.0975008946484811e-05, "loss": 0.8874, "step": 14569 }, { "epoch": 0.12, "learning_rate": 1.0973936195218647e-05, "loss": 0.9051, "step": 14570 }, { "epoch": 0.12, "learning_rate": 1.0972863432637018e-05, "loss": 0.862, "step": 14571 }, { "epoch": 0.12, "learning_rate": 1.097179065875239e-05, "loss": 0.8977, "step": 14572 }, { "epoch": 0.12, "learning_rate": 1.0970717873577228e-05, "loss": 0.8799, "step": 14573 }, { "epoch": 0.12, "learning_rate": 1.0969645077123992e-05, "loss": 0.9142, "step": 14574 }, { "epoch": 0.12, "learning_rate": 1.0968572269405148e-05, "loss": 0.9018, "step": 14575 }, { "epoch": 0.12, "learning_rate": 1.0967499450433163e-05, "loss": 0.8512, "step": 14576 }, { "epoch": 0.12, "learning_rate": 1.0966426620220496e-05, "loss": 0.9297, "step": 14577 }, { "epoch": 0.12, "learning_rate": 1.0965353778779617e-05, "loss": 0.8942, "step": 14578 }, { "epoch": 0.12, "learning_rate": 1.0964280926122988e-05, "loss": 0.906, "step": 14579 }, { "epoch": 0.12, "learning_rate": 1.0963208062263072e-05, "loss": 0.915, "step": 14580 }, { "epoch": 0.12, "learning_rate": 1.0962135187212337e-05, "loss": 0.8835, "step": 14581 }, { "epoch": 0.12, "learning_rate": 1.0961062300983244e-05, "loss": 0.8632, "step": 14582 }, { "epoch": 0.12, "learning_rate": 1.0959989403588263e-05, "loss": 0.9207, "step": 14583 }, { "epoch": 0.12, "learning_rate": 1.0958916495039856e-05, "loss": 0.9239, "step": 14584 }, { "epoch": 0.12, "learning_rate": 1.0957843575350488e-05, "loss": 0.9054, "step": 14585 }, { "epoch": 0.12, "learning_rate": 1.0956770644532626e-05, "loss": 0.9387, "step": 14586 }, { "epoch": 0.12, "learning_rate": 1.0955697702598735e-05, "loss": 0.9333, "step": 14587 }, { "epoch": 0.12, "learning_rate": 1.0954624749561282e-05, "loss": 0.8576, "step": 14588 }, { "epoch": 0.12, "learning_rate": 1.0953551785432727e-05, "loss": 0.9265, "step": 14589 }, { "epoch": 0.12, "learning_rate": 1.095247881022555e-05, "loss": 0.9161, "step": 14590 }, { "epoch": 0.12, "learning_rate": 1.09514058239522e-05, "loss": 0.9622, "step": 14591 }, { "epoch": 0.12, "learning_rate": 1.0950332826625154e-05, "loss": 0.8868, "step": 14592 }, { "epoch": 0.12, "learning_rate": 1.0949259818256876e-05, "loss": 0.8694, "step": 14593 }, { "epoch": 0.12, "learning_rate": 1.094818679885983e-05, "loss": 0.8699, "step": 14594 }, { "epoch": 0.12, "learning_rate": 1.0947113768446487e-05, "loss": 0.8952, "step": 14595 }, { "epoch": 0.12, "learning_rate": 1.0946040727029308e-05, "loss": 0.8868, "step": 14596 }, { "epoch": 0.12, "learning_rate": 1.0944967674620767e-05, "loss": 0.9036, "step": 14597 }, { "epoch": 0.12, "learning_rate": 1.0943894611233324e-05, "loss": 0.9048, "step": 14598 }, { "epoch": 0.12, "learning_rate": 1.0942821536879454e-05, "loss": 0.9375, "step": 14599 }, { "epoch": 0.12, "learning_rate": 1.0941748451571616e-05, "loss": 0.8956, "step": 14600 }, { "epoch": 0.12, "learning_rate": 1.0940675355322281e-05, "loss": 0.8958, "step": 14601 }, { "epoch": 0.12, "learning_rate": 1.0939602248143921e-05, "loss": 0.905, "step": 14602 }, { "epoch": 0.12, "learning_rate": 1.0938529130048997e-05, "loss": 0.8906, "step": 14603 }, { "epoch": 0.12, "learning_rate": 1.0937456001049975e-05, "loss": 0.8838, "step": 14604 }, { "epoch": 0.12, "learning_rate": 1.093638286115933e-05, "loss": 0.9095, "step": 14605 }, { "epoch": 0.12, "learning_rate": 1.093530971038953e-05, "loss": 0.8869, "step": 14606 }, { "epoch": 0.12, "learning_rate": 1.0934236548753037e-05, "loss": 0.9301, "step": 14607 }, { "epoch": 0.12, "learning_rate": 1.0933163376262324e-05, "loss": 0.8884, "step": 14608 }, { "epoch": 0.12, "learning_rate": 1.093209019292986e-05, "loss": 0.8613, "step": 14609 }, { "epoch": 0.12, "learning_rate": 1.0931016998768106e-05, "loss": 0.8919, "step": 14610 }, { "epoch": 0.12, "learning_rate": 1.0929943793789542e-05, "loss": 0.8994, "step": 14611 }, { "epoch": 0.12, "learning_rate": 1.0928870578006627e-05, "loss": 0.896, "step": 14612 }, { "epoch": 0.12, "learning_rate": 1.0927797351431834e-05, "loss": 0.8902, "step": 14613 }, { "epoch": 0.12, "learning_rate": 1.0926724114077634e-05, "loss": 0.8842, "step": 14614 }, { "epoch": 0.12, "learning_rate": 1.0925650865956494e-05, "loss": 0.9022, "step": 14615 }, { "epoch": 0.12, "learning_rate": 1.0924577607080882e-05, "loss": 0.9194, "step": 14616 }, { "epoch": 0.12, "learning_rate": 1.0923504337463268e-05, "loss": 0.8741, "step": 14617 }, { "epoch": 0.12, "learning_rate": 1.0922431057116127e-05, "loss": 0.821, "step": 14618 }, { "epoch": 0.12, "learning_rate": 1.0921357766051921e-05, "loss": 0.917, "step": 14619 }, { "epoch": 0.12, "learning_rate": 1.0920284464283121e-05, "loss": 0.9218, "step": 14620 }, { "epoch": 0.12, "learning_rate": 1.09192111518222e-05, "loss": 0.9004, "step": 14621 }, { "epoch": 0.12, "learning_rate": 1.0918137828681627e-05, "loss": 0.8967, "step": 14622 }, { "epoch": 0.12, "learning_rate": 1.0917064494873872e-05, "loss": 0.8946, "step": 14623 }, { "epoch": 0.12, "learning_rate": 1.0915991150411407e-05, "loss": 0.8819, "step": 14624 }, { "epoch": 0.12, "learning_rate": 1.09149177953067e-05, "loss": 0.8551, "step": 14625 }, { "epoch": 0.12, "learning_rate": 1.091384442957222e-05, "loss": 0.9158, "step": 14626 }, { "epoch": 0.12, "learning_rate": 1.0912771053220443e-05, "loss": 0.8923, "step": 14627 }, { "epoch": 0.12, "learning_rate": 1.0911697666263837e-05, "loss": 0.9323, "step": 14628 }, { "epoch": 0.12, "learning_rate": 1.091062426871487e-05, "loss": 0.9114, "step": 14629 }, { "epoch": 0.12, "learning_rate": 1.0909550860586016e-05, "loss": 0.8871, "step": 14630 }, { "epoch": 0.12, "learning_rate": 1.0908477441889746e-05, "loss": 0.8869, "step": 14631 }, { "epoch": 0.12, "learning_rate": 1.0907404012638532e-05, "loss": 0.8686, "step": 14632 }, { "epoch": 0.12, "learning_rate": 1.0906330572844843e-05, "loss": 0.9266, "step": 14633 }, { "epoch": 0.12, "learning_rate": 1.0905257122521155e-05, "loss": 0.931, "step": 14634 }, { "epoch": 0.12, "learning_rate": 1.0904183661679932e-05, "loss": 0.8759, "step": 14635 }, { "epoch": 0.12, "learning_rate": 1.0903110190333655e-05, "loss": 0.8962, "step": 14636 }, { "epoch": 0.12, "learning_rate": 1.0902036708494788e-05, "loss": 0.9018, "step": 14637 }, { "epoch": 0.12, "learning_rate": 1.0900963216175807e-05, "loss": 0.8492, "step": 14638 }, { "epoch": 0.12, "learning_rate": 1.0899889713389187e-05, "loss": 0.8961, "step": 14639 }, { "epoch": 0.12, "learning_rate": 1.0898816200147392e-05, "loss": 0.8633, "step": 14640 }, { "epoch": 0.12, "learning_rate": 1.08977426764629e-05, "loss": 0.9352, "step": 14641 }, { "epoch": 0.12, "learning_rate": 1.0896669142348182e-05, "loss": 0.9246, "step": 14642 }, { "epoch": 0.12, "learning_rate": 1.0895595597815716e-05, "loss": 0.909, "step": 14643 }, { "epoch": 0.12, "learning_rate": 1.0894522042877962e-05, "loss": 0.8735, "step": 14644 }, { "epoch": 0.12, "learning_rate": 1.0893448477547409e-05, "loss": 0.8638, "step": 14645 }, { "epoch": 0.12, "learning_rate": 1.0892374901836516e-05, "loss": 0.8594, "step": 14646 }, { "epoch": 0.12, "learning_rate": 1.0891301315757763e-05, "loss": 0.9375, "step": 14647 }, { "epoch": 0.12, "learning_rate": 1.0890227719323625e-05, "loss": 0.9114, "step": 14648 }, { "epoch": 0.12, "learning_rate": 1.0889154112546569e-05, "loss": 0.9291, "step": 14649 }, { "epoch": 0.12, "learning_rate": 1.0888080495439073e-05, "loss": 0.8818, "step": 14650 }, { "epoch": 0.12, "learning_rate": 1.088700686801361e-05, "loss": 0.9204, "step": 14651 }, { "epoch": 0.12, "learning_rate": 1.0885933230282655e-05, "loss": 0.8803, "step": 14652 }, { "epoch": 0.12, "learning_rate": 1.0884859582258675e-05, "loss": 0.8691, "step": 14653 }, { "epoch": 0.12, "learning_rate": 1.0883785923954154e-05, "loss": 0.8281, "step": 14654 }, { "epoch": 0.12, "learning_rate": 1.088271225538156e-05, "loss": 0.8839, "step": 14655 }, { "epoch": 0.12, "learning_rate": 1.0881638576553365e-05, "loss": 0.9029, "step": 14656 }, { "epoch": 0.12, "learning_rate": 1.088056488748205e-05, "loss": 0.9037, "step": 14657 }, { "epoch": 0.12, "learning_rate": 1.0879491188180084e-05, "loss": 0.926, "step": 14658 }, { "epoch": 0.12, "learning_rate": 1.0878417478659944e-05, "loss": 0.8598, "step": 14659 }, { "epoch": 0.12, "learning_rate": 1.0877343758934105e-05, "loss": 0.9334, "step": 14660 }, { "epoch": 0.12, "learning_rate": 1.0876270029015042e-05, "loss": 0.8694, "step": 14661 }, { "epoch": 0.12, "learning_rate": 1.0875196288915226e-05, "loss": 0.8882, "step": 14662 }, { "epoch": 0.12, "learning_rate": 1.0874122538647133e-05, "loss": 0.9047, "step": 14663 }, { "epoch": 0.12, "learning_rate": 1.0873048778223247e-05, "loss": 0.9144, "step": 14664 }, { "epoch": 0.12, "learning_rate": 1.0871975007656031e-05, "loss": 0.8955, "step": 14665 }, { "epoch": 0.12, "learning_rate": 1.0870901226957966e-05, "loss": 0.8492, "step": 14666 }, { "epoch": 0.12, "learning_rate": 1.0869827436141526e-05, "loss": 0.9188, "step": 14667 }, { "epoch": 0.12, "learning_rate": 1.0868753635219189e-05, "loss": 0.8824, "step": 14668 }, { "epoch": 0.12, "learning_rate": 1.086767982420343e-05, "loss": 0.8819, "step": 14669 }, { "epoch": 0.12, "learning_rate": 1.0866606003106725e-05, "loss": 0.8645, "step": 14670 }, { "epoch": 0.12, "learning_rate": 1.0865532171941546e-05, "loss": 0.9073, "step": 14671 }, { "epoch": 0.12, "learning_rate": 1.0864458330720376e-05, "loss": 0.8874, "step": 14672 }, { "epoch": 0.12, "learning_rate": 1.0863384479455686e-05, "loss": 0.9026, "step": 14673 }, { "epoch": 0.12, "learning_rate": 1.0862310618159952e-05, "loss": 0.9065, "step": 14674 }, { "epoch": 0.12, "learning_rate": 1.0861236746845654e-05, "loss": 0.8502, "step": 14675 }, { "epoch": 0.12, "learning_rate": 1.0860162865525266e-05, "loss": 0.876, "step": 14676 }, { "epoch": 0.12, "learning_rate": 1.0859088974211268e-05, "loss": 0.8737, "step": 14677 }, { "epoch": 0.12, "learning_rate": 1.0858015072916128e-05, "loss": 0.9237, "step": 14678 }, { "epoch": 0.12, "learning_rate": 1.0856941161652333e-05, "loss": 0.8796, "step": 14679 }, { "epoch": 0.12, "learning_rate": 1.0855867240432354e-05, "loss": 0.8968, "step": 14680 }, { "epoch": 0.12, "learning_rate": 1.0854793309268672e-05, "loss": 0.9149, "step": 14681 }, { "epoch": 0.12, "learning_rate": 1.0853719368173763e-05, "loss": 0.9239, "step": 14682 }, { "epoch": 0.12, "learning_rate": 1.0852645417160101e-05, "loss": 0.8871, "step": 14683 }, { "epoch": 0.12, "learning_rate": 1.0851571456240169e-05, "loss": 0.9311, "step": 14684 }, { "epoch": 0.12, "learning_rate": 1.0850497485426441e-05, "loss": 0.911, "step": 14685 }, { "epoch": 0.12, "learning_rate": 1.0849423504731396e-05, "loss": 0.8948, "step": 14686 }, { "epoch": 0.12, "learning_rate": 1.0848349514167508e-05, "loss": 0.9032, "step": 14687 }, { "epoch": 0.12, "learning_rate": 1.084727551374726e-05, "loss": 0.8925, "step": 14688 }, { "epoch": 0.12, "learning_rate": 1.0846201503483129e-05, "loss": 0.8608, "step": 14689 }, { "epoch": 0.12, "learning_rate": 1.0845127483387589e-05, "loss": 0.8965, "step": 14690 }, { "epoch": 0.12, "learning_rate": 1.0844053453473127e-05, "loss": 0.8974, "step": 14691 }, { "epoch": 0.12, "learning_rate": 1.0842979413752212e-05, "loss": 0.888, "step": 14692 }, { "epoch": 0.12, "learning_rate": 1.084190536423733e-05, "loss": 0.9319, "step": 14693 }, { "epoch": 0.12, "learning_rate": 1.0840831304940952e-05, "loss": 0.9074, "step": 14694 }, { "epoch": 0.12, "learning_rate": 1.0839757235875564e-05, "loss": 0.9121, "step": 14695 }, { "epoch": 0.12, "learning_rate": 1.083868315705364e-05, "loss": 0.8917, "step": 14696 }, { "epoch": 0.12, "learning_rate": 1.083760906848766e-05, "loss": 0.8876, "step": 14697 }, { "epoch": 0.12, "learning_rate": 1.0836534970190106e-05, "loss": 0.8872, "step": 14698 }, { "epoch": 0.12, "learning_rate": 1.083546086217345e-05, "loss": 0.8843, "step": 14699 }, { "epoch": 0.12, "learning_rate": 1.0834386744450182e-05, "loss": 0.9166, "step": 14700 }, { "epoch": 0.12, "learning_rate": 1.0833312617032775e-05, "loss": 0.9093, "step": 14701 }, { "epoch": 0.12, "learning_rate": 1.0832238479933703e-05, "loss": 0.858, "step": 14702 }, { "epoch": 0.12, "learning_rate": 1.0831164333165458e-05, "loss": 0.8763, "step": 14703 }, { "epoch": 0.12, "learning_rate": 1.0830090176740512e-05, "loss": 0.8643, "step": 14704 }, { "epoch": 0.12, "learning_rate": 1.0829016010671342e-05, "loss": 0.9017, "step": 14705 }, { "epoch": 0.12, "learning_rate": 1.0827941834970436e-05, "loss": 0.9261, "step": 14706 }, { "epoch": 0.12, "learning_rate": 1.0826867649650272e-05, "loss": 0.9029, "step": 14707 }, { "epoch": 0.12, "learning_rate": 1.0825793454723325e-05, "loss": 0.8725, "step": 14708 }, { "epoch": 0.12, "learning_rate": 1.0824719250202078e-05, "loss": 0.9051, "step": 14709 }, { "epoch": 0.12, "learning_rate": 1.0823645036099017e-05, "loss": 0.9158, "step": 14710 }, { "epoch": 0.12, "learning_rate": 1.0822570812426611e-05, "loss": 0.9085, "step": 14711 }, { "epoch": 0.12, "learning_rate": 1.0821496579197355e-05, "loss": 0.8932, "step": 14712 }, { "epoch": 0.12, "learning_rate": 1.0820422336423717e-05, "loss": 0.9217, "step": 14713 }, { "epoch": 0.12, "learning_rate": 1.0819348084118183e-05, "loss": 0.8892, "step": 14714 }, { "epoch": 0.12, "learning_rate": 1.0818273822293235e-05, "loss": 0.9278, "step": 14715 }, { "epoch": 0.12, "learning_rate": 1.0817199550961352e-05, "loss": 0.9089, "step": 14716 }, { "epoch": 0.12, "learning_rate": 1.0816125270135019e-05, "loss": 0.925, "step": 14717 }, { "epoch": 0.12, "learning_rate": 1.0815050979826711e-05, "loss": 0.898, "step": 14718 }, { "epoch": 0.12, "learning_rate": 1.0813976680048914e-05, "loss": 0.8732, "step": 14719 }, { "epoch": 0.12, "learning_rate": 1.0812902370814109e-05, "loss": 0.8896, "step": 14720 }, { "epoch": 0.12, "learning_rate": 1.0811828052134777e-05, "loss": 0.9127, "step": 14721 }, { "epoch": 0.12, "learning_rate": 1.0810753724023397e-05, "loss": 0.9418, "step": 14722 }, { "epoch": 0.12, "learning_rate": 1.0809679386492457e-05, "loss": 0.9, "step": 14723 }, { "epoch": 0.12, "learning_rate": 1.0808605039554437e-05, "loss": 0.8798, "step": 14724 }, { "epoch": 0.12, "learning_rate": 1.0807530683221812e-05, "loss": 0.8976, "step": 14725 }, { "epoch": 0.12, "learning_rate": 1.0806456317507072e-05, "loss": 0.8846, "step": 14726 }, { "epoch": 0.12, "learning_rate": 1.0805381942422697e-05, "loss": 0.883, "step": 14727 }, { "epoch": 0.12, "learning_rate": 1.0804307557981174e-05, "loss": 0.874, "step": 14728 }, { "epoch": 0.12, "learning_rate": 1.0803233164194976e-05, "loss": 0.9093, "step": 14729 }, { "epoch": 0.12, "learning_rate": 1.0802158761076589e-05, "loss": 0.9261, "step": 14730 }, { "epoch": 0.12, "learning_rate": 1.0801084348638502e-05, "loss": 0.8843, "step": 14731 }, { "epoch": 0.12, "learning_rate": 1.0800009926893188e-05, "loss": 0.9332, "step": 14732 }, { "epoch": 0.12, "learning_rate": 1.0798935495853138e-05, "loss": 0.8861, "step": 14733 }, { "epoch": 0.12, "learning_rate": 1.0797861055530832e-05, "loss": 0.8841, "step": 14734 }, { "epoch": 0.12, "learning_rate": 1.0796786605938754e-05, "loss": 0.8644, "step": 14735 }, { "epoch": 0.12, "learning_rate": 1.0795712147089384e-05, "loss": 0.9595, "step": 14736 }, { "epoch": 0.12, "learning_rate": 1.079463767899521e-05, "loss": 0.8898, "step": 14737 }, { "epoch": 0.12, "learning_rate": 1.079356320166871e-05, "loss": 0.8796, "step": 14738 }, { "epoch": 0.12, "learning_rate": 1.0792488715122373e-05, "loss": 0.8865, "step": 14739 }, { "epoch": 0.12, "learning_rate": 1.0791414219368684e-05, "loss": 0.8941, "step": 14740 }, { "epoch": 0.12, "learning_rate": 1.079033971442012e-05, "loss": 0.942, "step": 14741 }, { "epoch": 0.12, "learning_rate": 1.0789265200289168e-05, "loss": 0.9107, "step": 14742 }, { "epoch": 0.12, "learning_rate": 1.078819067698831e-05, "loss": 0.8691, "step": 14743 }, { "epoch": 0.12, "learning_rate": 1.0787116144530037e-05, "loss": 0.9, "step": 14744 }, { "epoch": 0.12, "learning_rate": 1.0786041602926824e-05, "loss": 0.8903, "step": 14745 }, { "epoch": 0.12, "learning_rate": 1.0784967052191164e-05, "loss": 0.8917, "step": 14746 }, { "epoch": 0.12, "learning_rate": 1.0783892492335537e-05, "loss": 0.8837, "step": 14747 }, { "epoch": 0.12, "learning_rate": 1.0782817923372425e-05, "loss": 0.8988, "step": 14748 }, { "epoch": 0.12, "learning_rate": 1.0781743345314317e-05, "loss": 0.9135, "step": 14749 }, { "epoch": 0.12, "learning_rate": 1.0780668758173696e-05, "loss": 0.8864, "step": 14750 }, { "epoch": 0.12, "learning_rate": 1.0779594161963048e-05, "loss": 0.8918, "step": 14751 }, { "epoch": 0.12, "learning_rate": 1.0778519556694853e-05, "loss": 0.8676, "step": 14752 }, { "epoch": 0.12, "learning_rate": 1.0777444942381605e-05, "loss": 0.9096, "step": 14753 }, { "epoch": 0.12, "learning_rate": 1.0776370319035783e-05, "loss": 0.9001, "step": 14754 }, { "epoch": 0.12, "learning_rate": 1.0775295686669872e-05, "loss": 0.867, "step": 14755 }, { "epoch": 0.12, "learning_rate": 1.0774221045296363e-05, "loss": 0.8711, "step": 14756 }, { "epoch": 0.13, "learning_rate": 1.0773146394927732e-05, "loss": 0.9495, "step": 14757 }, { "epoch": 0.13, "learning_rate": 1.0772071735576475e-05, "loss": 0.8928, "step": 14758 }, { "epoch": 0.13, "learning_rate": 1.0770997067255069e-05, "loss": 0.908, "step": 14759 }, { "epoch": 0.13, "learning_rate": 1.0769922389976004e-05, "loss": 0.9013, "step": 14760 }, { "epoch": 0.13, "learning_rate": 1.0768847703751766e-05, "loss": 0.9003, "step": 14761 }, { "epoch": 0.13, "learning_rate": 1.0767773008594842e-05, "loss": 0.8384, "step": 14762 }, { "epoch": 0.13, "learning_rate": 1.0766698304517715e-05, "loss": 0.8963, "step": 14763 }, { "epoch": 0.13, "learning_rate": 1.0765623591532872e-05, "loss": 0.8593, "step": 14764 }, { "epoch": 0.13, "learning_rate": 1.0764548869652803e-05, "loss": 0.8756, "step": 14765 }, { "epoch": 0.13, "learning_rate": 1.0763474138889988e-05, "loss": 0.8805, "step": 14766 }, { "epoch": 0.13, "learning_rate": 1.0762399399256917e-05, "loss": 0.9236, "step": 14767 }, { "epoch": 0.13, "learning_rate": 1.0761324650766078e-05, "loss": 0.8925, "step": 14768 }, { "epoch": 0.13, "learning_rate": 1.076024989342996e-05, "loss": 0.9191, "step": 14769 }, { "epoch": 0.13, "learning_rate": 1.075917512726104e-05, "loss": 0.8991, "step": 14770 }, { "epoch": 0.13, "learning_rate": 1.0758100352271814e-05, "loss": 0.8933, "step": 14771 }, { "epoch": 0.13, "learning_rate": 1.0757025568474764e-05, "loss": 0.8939, "step": 14772 }, { "epoch": 0.13, "learning_rate": 1.075595077588238e-05, "loss": 0.9147, "step": 14773 }, { "epoch": 0.13, "learning_rate": 1.0754875974507153e-05, "loss": 0.9029, "step": 14774 }, { "epoch": 0.13, "learning_rate": 1.0753801164361561e-05, "loss": 0.8905, "step": 14775 }, { "epoch": 0.13, "learning_rate": 1.0752726345458097e-05, "loss": 0.8593, "step": 14776 }, { "epoch": 0.13, "learning_rate": 1.075165151780925e-05, "loss": 0.9064, "step": 14777 }, { "epoch": 0.13, "learning_rate": 1.0750576681427504e-05, "loss": 0.9263, "step": 14778 }, { "epoch": 0.13, "learning_rate": 1.0749501836325347e-05, "loss": 0.8667, "step": 14779 }, { "epoch": 0.13, "learning_rate": 1.074842698251527e-05, "loss": 0.8946, "step": 14780 }, { "epoch": 0.13, "learning_rate": 1.074735212000976e-05, "loss": 0.8768, "step": 14781 }, { "epoch": 0.13, "learning_rate": 1.07462772488213e-05, "loss": 0.9006, "step": 14782 }, { "epoch": 0.13, "learning_rate": 1.0745202368962387e-05, "loss": 0.8886, "step": 14783 }, { "epoch": 0.13, "learning_rate": 1.07441274804455e-05, "loss": 0.9228, "step": 14784 }, { "epoch": 0.13, "learning_rate": 1.0743052583283133e-05, "loss": 0.9302, "step": 14785 }, { "epoch": 0.13, "learning_rate": 1.0741977677487778e-05, "loss": 0.8871, "step": 14786 }, { "epoch": 0.13, "learning_rate": 1.0740902763071914e-05, "loss": 0.9161, "step": 14787 }, { "epoch": 0.13, "learning_rate": 1.0739827840048036e-05, "loss": 0.8848, "step": 14788 }, { "epoch": 0.13, "learning_rate": 1.0738752908428631e-05, "loss": 0.8969, "step": 14789 }, { "epoch": 0.13, "learning_rate": 1.073767796822619e-05, "loss": 0.8957, "step": 14790 }, { "epoch": 0.13, "learning_rate": 1.0736603019453196e-05, "loss": 0.9079, "step": 14791 }, { "epoch": 0.13, "learning_rate": 1.0735528062122148e-05, "loss": 0.8965, "step": 14792 }, { "epoch": 0.13, "learning_rate": 1.0734453096245525e-05, "loss": 0.9251, "step": 14793 }, { "epoch": 0.13, "learning_rate": 1.0733378121835821e-05, "loss": 0.9001, "step": 14794 }, { "epoch": 0.13, "learning_rate": 1.0732303138905526e-05, "loss": 0.8732, "step": 14795 }, { "epoch": 0.13, "learning_rate": 1.0731228147467128e-05, "loss": 0.9156, "step": 14796 }, { "epoch": 0.13, "learning_rate": 1.0730153147533116e-05, "loss": 0.8874, "step": 14797 }, { "epoch": 0.13, "learning_rate": 1.072907813911598e-05, "loss": 0.8941, "step": 14798 }, { "epoch": 0.13, "learning_rate": 1.0728003122228214e-05, "loss": 0.8699, "step": 14799 }, { "epoch": 0.13, "learning_rate": 1.0726928096882301e-05, "loss": 0.8985, "step": 14800 }, { "epoch": 0.13, "learning_rate": 1.0725853063090734e-05, "loss": 0.911, "step": 14801 }, { "epoch": 0.13, "learning_rate": 1.0724778020866007e-05, "loss": 0.8791, "step": 14802 }, { "epoch": 0.13, "learning_rate": 1.07237029702206e-05, "loss": 0.837, "step": 14803 }, { "epoch": 0.13, "learning_rate": 1.0722627911167014e-05, "loss": 0.8813, "step": 14804 }, { "epoch": 0.13, "learning_rate": 1.0721552843717736e-05, "loss": 0.9046, "step": 14805 }, { "epoch": 0.13, "learning_rate": 1.072047776788525e-05, "loss": 0.9023, "step": 14806 }, { "epoch": 0.13, "learning_rate": 1.0719402683682057e-05, "loss": 0.9064, "step": 14807 }, { "epoch": 0.13, "learning_rate": 1.071832759112064e-05, "loss": 0.9079, "step": 14808 }, { "epoch": 0.13, "learning_rate": 1.0717252490213493e-05, "loss": 0.9052, "step": 14809 }, { "epoch": 0.13, "learning_rate": 1.0716177380973105e-05, "loss": 0.9094, "step": 14810 }, { "epoch": 0.13, "learning_rate": 1.0715102263411972e-05, "loss": 0.9187, "step": 14811 }, { "epoch": 0.13, "learning_rate": 1.0714027137542575e-05, "loss": 0.8842, "step": 14812 }, { "epoch": 0.13, "learning_rate": 1.0712952003377414e-05, "loss": 0.9084, "step": 14813 }, { "epoch": 0.13, "learning_rate": 1.0711876860928976e-05, "loss": 0.8604, "step": 14814 }, { "epoch": 0.13, "learning_rate": 1.0710801710209755e-05, "loss": 0.9083, "step": 14815 }, { "epoch": 0.13, "learning_rate": 1.0709726551232242e-05, "loss": 0.8345, "step": 14816 }, { "epoch": 0.13, "learning_rate": 1.0708651384008926e-05, "loss": 0.8814, "step": 14817 }, { "epoch": 0.13, "learning_rate": 1.07075762085523e-05, "loss": 0.9292, "step": 14818 }, { "epoch": 0.13, "learning_rate": 1.0706501024874857e-05, "loss": 0.8863, "step": 14819 }, { "epoch": 0.13, "learning_rate": 1.0705425832989084e-05, "loss": 0.8588, "step": 14820 }, { "epoch": 0.13, "learning_rate": 1.070435063290748e-05, "loss": 0.9144, "step": 14821 }, { "epoch": 0.13, "learning_rate": 1.0703275424642532e-05, "loss": 0.8723, "step": 14822 }, { "epoch": 0.13, "learning_rate": 1.0702200208206737e-05, "loss": 0.9164, "step": 14823 }, { "epoch": 0.13, "learning_rate": 1.0701124983612579e-05, "loss": 0.8849, "step": 14824 }, { "epoch": 0.13, "learning_rate": 1.0700049750872556e-05, "loss": 0.9478, "step": 14825 }, { "epoch": 0.13, "learning_rate": 1.0698974509999159e-05, "loss": 0.8935, "step": 14826 }, { "epoch": 0.13, "learning_rate": 1.0697899261004883e-05, "loss": 0.9343, "step": 14827 }, { "epoch": 0.13, "learning_rate": 1.0696824003902213e-05, "loss": 0.8811, "step": 14828 }, { "epoch": 0.13, "learning_rate": 1.0695748738703653e-05, "loss": 0.9369, "step": 14829 }, { "epoch": 0.13, "learning_rate": 1.0694673465421686e-05, "loss": 0.9047, "step": 14830 }, { "epoch": 0.13, "learning_rate": 1.069359818406881e-05, "loss": 0.9102, "step": 14831 }, { "epoch": 0.13, "learning_rate": 1.0692522894657517e-05, "loss": 0.8797, "step": 14832 }, { "epoch": 0.13, "learning_rate": 1.0691447597200297e-05, "loss": 0.8783, "step": 14833 }, { "epoch": 0.13, "learning_rate": 1.0690372291709648e-05, "loss": 0.8953, "step": 14834 }, { "epoch": 0.13, "learning_rate": 1.0689296978198058e-05, "loss": 0.908, "step": 14835 }, { "epoch": 0.13, "learning_rate": 1.0688221656678026e-05, "loss": 0.9141, "step": 14836 }, { "epoch": 0.13, "learning_rate": 1.068714632716204e-05, "loss": 0.8939, "step": 14837 }, { "epoch": 0.13, "learning_rate": 1.0686070989662596e-05, "loss": 0.8949, "step": 14838 }, { "epoch": 0.13, "learning_rate": 1.068499564419219e-05, "loss": 0.8824, "step": 14839 }, { "epoch": 0.13, "learning_rate": 1.068392029076331e-05, "loss": 0.9273, "step": 14840 }, { "epoch": 0.13, "learning_rate": 1.0682844929388454e-05, "loss": 0.8905, "step": 14841 }, { "epoch": 0.13, "learning_rate": 1.0681769560080115e-05, "loss": 0.9218, "step": 14842 }, { "epoch": 0.13, "learning_rate": 1.0680694182850786e-05, "loss": 0.8647, "step": 14843 }, { "epoch": 0.13, "learning_rate": 1.0679618797712961e-05, "loss": 0.8632, "step": 14844 }, { "epoch": 0.13, "learning_rate": 1.0678543404679137e-05, "loss": 0.8837, "step": 14845 }, { "epoch": 0.13, "learning_rate": 1.0677468003761804e-05, "loss": 0.8509, "step": 14846 }, { "epoch": 0.13, "learning_rate": 1.0676392594973459e-05, "loss": 0.9012, "step": 14847 }, { "epoch": 0.13, "learning_rate": 1.06753171783266e-05, "loss": 0.8596, "step": 14848 }, { "epoch": 0.13, "learning_rate": 1.0674241753833708e-05, "loss": 0.8688, "step": 14849 }, { "epoch": 0.13, "learning_rate": 1.0673166321507293e-05, "loss": 0.9576, "step": 14850 }, { "epoch": 0.13, "learning_rate": 1.0672090881359842e-05, "loss": 0.9083, "step": 14851 }, { "epoch": 0.13, "learning_rate": 1.0671015433403851e-05, "loss": 0.9219, "step": 14852 }, { "epoch": 0.13, "learning_rate": 1.0669939977651815e-05, "loss": 0.9566, "step": 14853 }, { "epoch": 0.13, "learning_rate": 1.0668864514116231e-05, "loss": 0.9063, "step": 14854 }, { "epoch": 0.13, "learning_rate": 1.066778904280959e-05, "loss": 0.9409, "step": 14855 }, { "epoch": 0.13, "learning_rate": 1.0666713563744389e-05, "loss": 0.8906, "step": 14856 }, { "epoch": 0.13, "learning_rate": 1.0665638076933127e-05, "loss": 0.9285, "step": 14857 }, { "epoch": 0.13, "learning_rate": 1.0664562582388291e-05, "loss": 0.9397, "step": 14858 }, { "epoch": 0.13, "learning_rate": 1.0663487080122383e-05, "loss": 0.8784, "step": 14859 }, { "epoch": 0.13, "learning_rate": 1.0662411570147897e-05, "loss": 0.8653, "step": 14860 }, { "epoch": 0.13, "learning_rate": 1.0661336052477327e-05, "loss": 0.8774, "step": 14861 }, { "epoch": 0.13, "learning_rate": 1.0660260527123171e-05, "loss": 0.8806, "step": 14862 }, { "epoch": 0.13, "learning_rate": 1.0659184994097924e-05, "loss": 0.8549, "step": 14863 }, { "epoch": 0.13, "learning_rate": 1.065810945341408e-05, "loss": 0.9103, "step": 14864 }, { "epoch": 0.13, "learning_rate": 1.0657033905084135e-05, "loss": 0.8851, "step": 14865 }, { "epoch": 0.13, "learning_rate": 1.0655958349120592e-05, "loss": 0.8965, "step": 14866 }, { "epoch": 0.13, "learning_rate": 1.0654882785535935e-05, "loss": 0.8844, "step": 14867 }, { "epoch": 0.13, "learning_rate": 1.065380721434267e-05, "loss": 0.8713, "step": 14868 }, { "epoch": 0.13, "learning_rate": 1.0652731635553293e-05, "loss": 0.9302, "step": 14869 }, { "epoch": 0.13, "learning_rate": 1.0651656049180293e-05, "loss": 0.9104, "step": 14870 }, { "epoch": 0.13, "learning_rate": 1.065058045523617e-05, "loss": 0.8925, "step": 14871 }, { "epoch": 0.13, "learning_rate": 1.0649504853733423e-05, "loss": 0.8908, "step": 14872 }, { "epoch": 0.13, "learning_rate": 1.0648429244684548e-05, "loss": 0.8959, "step": 14873 }, { "epoch": 0.13, "learning_rate": 1.0647353628102038e-05, "loss": 0.9035, "step": 14874 }, { "epoch": 0.13, "learning_rate": 1.0646278003998395e-05, "loss": 0.9122, "step": 14875 }, { "epoch": 0.13, "learning_rate": 1.0645202372386114e-05, "loss": 0.8899, "step": 14876 }, { "epoch": 0.13, "learning_rate": 1.064412673327769e-05, "loss": 0.9434, "step": 14877 }, { "epoch": 0.13, "learning_rate": 1.0643051086685624e-05, "loss": 0.8878, "step": 14878 }, { "epoch": 0.13, "learning_rate": 1.0641975432622407e-05, "loss": 0.9123, "step": 14879 }, { "epoch": 0.13, "learning_rate": 1.0640899771100539e-05, "loss": 0.8837, "step": 14880 }, { "epoch": 0.13, "learning_rate": 1.063982410213252e-05, "loss": 0.9372, "step": 14881 }, { "epoch": 0.13, "learning_rate": 1.063874842573085e-05, "loss": 0.9074, "step": 14882 }, { "epoch": 0.13, "learning_rate": 1.0637672741908017e-05, "loss": 0.8627, "step": 14883 }, { "epoch": 0.13, "learning_rate": 1.0636597050676525e-05, "loss": 0.8878, "step": 14884 }, { "epoch": 0.13, "learning_rate": 1.0635521352048873e-05, "loss": 0.8916, "step": 14885 }, { "epoch": 0.13, "learning_rate": 1.063444564603755e-05, "loss": 0.8969, "step": 14886 }, { "epoch": 0.13, "learning_rate": 1.0633369932655067e-05, "loss": 0.895, "step": 14887 }, { "epoch": 0.13, "learning_rate": 1.0632294211913911e-05, "loss": 0.9137, "step": 14888 }, { "epoch": 0.13, "learning_rate": 1.0631218483826585e-05, "loss": 0.8774, "step": 14889 }, { "epoch": 0.13, "learning_rate": 1.0630142748405585e-05, "loss": 0.9175, "step": 14890 }, { "epoch": 0.13, "learning_rate": 1.0629067005663414e-05, "loss": 0.896, "step": 14891 }, { "epoch": 0.13, "learning_rate": 1.0627991255612564e-05, "loss": 0.8767, "step": 14892 }, { "epoch": 0.13, "learning_rate": 1.0626915498265535e-05, "loss": 0.9019, "step": 14893 }, { "epoch": 0.13, "learning_rate": 1.062583973363483e-05, "loss": 0.8677, "step": 14894 }, { "epoch": 0.13, "learning_rate": 1.062476396173294e-05, "loss": 0.8759, "step": 14895 }, { "epoch": 0.13, "learning_rate": 1.0623688182572372e-05, "loss": 0.8416, "step": 14896 }, { "epoch": 0.13, "learning_rate": 1.0622612396165616e-05, "loss": 0.9166, "step": 14897 }, { "epoch": 0.13, "learning_rate": 1.0621536602525176e-05, "loss": 0.9048, "step": 14898 }, { "epoch": 0.13, "learning_rate": 1.062046080166355e-05, "loss": 0.9372, "step": 14899 }, { "epoch": 0.13, "learning_rate": 1.061938499359324e-05, "loss": 0.8919, "step": 14900 }, { "epoch": 0.13, "learning_rate": 1.0618309178326738e-05, "loss": 0.9149, "step": 14901 }, { "epoch": 0.13, "learning_rate": 1.0617233355876548e-05, "loss": 0.9012, "step": 14902 }, { "epoch": 0.13, "learning_rate": 1.061615752625517e-05, "loss": 0.9243, "step": 14903 }, { "epoch": 0.13, "learning_rate": 1.06150816894751e-05, "loss": 0.8528, "step": 14904 }, { "epoch": 0.13, "learning_rate": 1.0614005845548841e-05, "loss": 0.9196, "step": 14905 }, { "epoch": 0.13, "learning_rate": 1.0612929994488887e-05, "loss": 0.88, "step": 14906 }, { "epoch": 0.13, "learning_rate": 1.0611854136307744e-05, "loss": 0.8907, "step": 14907 }, { "epoch": 0.13, "learning_rate": 1.061077827101791e-05, "loss": 0.9016, "step": 14908 }, { "epoch": 0.13, "learning_rate": 1.060970239863188e-05, "loss": 0.9052, "step": 14909 }, { "epoch": 0.13, "learning_rate": 1.0608626519162157e-05, "loss": 0.8584, "step": 14910 }, { "epoch": 0.13, "learning_rate": 1.0607550632621242e-05, "loss": 0.8687, "step": 14911 }, { "epoch": 0.13, "learning_rate": 1.0606474739021635e-05, "loss": 0.8829, "step": 14912 }, { "epoch": 0.13, "learning_rate": 1.0605398838375833e-05, "loss": 0.8812, "step": 14913 }, { "epoch": 0.13, "learning_rate": 1.0604322930696338e-05, "loss": 0.8992, "step": 14914 }, { "epoch": 0.13, "learning_rate": 1.0603247015995652e-05, "loss": 0.9019, "step": 14915 }, { "epoch": 0.13, "learning_rate": 1.060217109428627e-05, "loss": 0.8714, "step": 14916 }, { "epoch": 0.13, "learning_rate": 1.06010951655807e-05, "loss": 0.9229, "step": 14917 }, { "epoch": 0.13, "learning_rate": 1.0600019229891436e-05, "loss": 0.886, "step": 14918 }, { "epoch": 0.13, "learning_rate": 1.059894328723098e-05, "loss": 0.8555, "step": 14919 }, { "epoch": 0.13, "learning_rate": 1.0597867337611834e-05, "loss": 0.8871, "step": 14920 }, { "epoch": 0.13, "learning_rate": 1.0596791381046501e-05, "loss": 0.9026, "step": 14921 }, { "epoch": 0.13, "learning_rate": 1.0595715417547477e-05, "loss": 0.9167, "step": 14922 }, { "epoch": 0.13, "learning_rate": 1.0594639447127265e-05, "loss": 0.9151, "step": 14923 }, { "epoch": 0.13, "learning_rate": 1.0593563469798364e-05, "loss": 0.873, "step": 14924 }, { "epoch": 0.13, "learning_rate": 1.0592487485573277e-05, "loss": 0.9415, "step": 14925 }, { "epoch": 0.13, "learning_rate": 1.0591411494464506e-05, "loss": 0.8934, "step": 14926 }, { "epoch": 0.13, "learning_rate": 1.0590335496484549e-05, "loss": 0.882, "step": 14927 }, { "epoch": 0.13, "learning_rate": 1.0589259491645914e-05, "loss": 0.91, "step": 14928 }, { "epoch": 0.13, "learning_rate": 1.0588183479961089e-05, "loss": 0.9137, "step": 14929 }, { "epoch": 0.13, "learning_rate": 1.0587107461442592e-05, "loss": 0.8925, "step": 14930 }, { "epoch": 0.13, "learning_rate": 1.0586031436102915e-05, "loss": 0.886, "step": 14931 }, { "epoch": 0.13, "learning_rate": 1.0584955403954556e-05, "loss": 0.8859, "step": 14932 }, { "epoch": 0.13, "learning_rate": 1.0583879365010025e-05, "loss": 0.8854, "step": 14933 }, { "epoch": 0.13, "learning_rate": 1.058280331928182e-05, "loss": 0.8587, "step": 14934 }, { "epoch": 0.13, "learning_rate": 1.0581727266782442e-05, "loss": 0.8996, "step": 14935 }, { "epoch": 0.13, "learning_rate": 1.0580651207524395e-05, "loss": 0.9047, "step": 14936 }, { "epoch": 0.13, "learning_rate": 1.0579575141520182e-05, "loss": 0.935, "step": 14937 }, { "epoch": 0.13, "learning_rate": 1.05784990687823e-05, "loss": 0.9343, "step": 14938 }, { "epoch": 0.13, "learning_rate": 1.0577422989323254e-05, "loss": 0.8725, "step": 14939 }, { "epoch": 0.13, "learning_rate": 1.057634690315555e-05, "loss": 0.8621, "step": 14940 }, { "epoch": 0.13, "learning_rate": 1.057527081029168e-05, "loss": 0.8768, "step": 14941 }, { "epoch": 0.13, "learning_rate": 1.0574194710744158e-05, "loss": 0.887, "step": 14942 }, { "epoch": 0.13, "learning_rate": 1.0573118604525481e-05, "loss": 0.9182, "step": 14943 }, { "epoch": 0.13, "learning_rate": 1.057204249164815e-05, "loss": 0.88, "step": 14944 }, { "epoch": 0.13, "learning_rate": 1.0570966372124669e-05, "loss": 0.8845, "step": 14945 }, { "epoch": 0.13, "learning_rate": 1.0569890245967544e-05, "loss": 0.9195, "step": 14946 }, { "epoch": 0.13, "learning_rate": 1.0568814113189271e-05, "loss": 0.928, "step": 14947 }, { "epoch": 0.13, "learning_rate": 1.056773797380236e-05, "loss": 0.8731, "step": 14948 }, { "epoch": 0.13, "learning_rate": 1.0566661827819311e-05, "loss": 0.8797, "step": 14949 }, { "epoch": 0.13, "learning_rate": 1.056558567525262e-05, "loss": 0.8642, "step": 14950 }, { "epoch": 0.13, "learning_rate": 1.0564509516114803e-05, "loss": 0.8885, "step": 14951 }, { "epoch": 0.13, "learning_rate": 1.0563433350418355e-05, "loss": 0.8931, "step": 14952 }, { "epoch": 0.13, "learning_rate": 1.0562357178175779e-05, "loss": 0.9056, "step": 14953 }, { "epoch": 0.13, "learning_rate": 1.0561280999399584e-05, "loss": 0.8857, "step": 14954 }, { "epoch": 0.13, "learning_rate": 1.0560204814102267e-05, "loss": 0.8616, "step": 14955 }, { "epoch": 0.13, "learning_rate": 1.0559128622296333e-05, "loss": 0.8976, "step": 14956 }, { "epoch": 0.13, "learning_rate": 1.0558052423994285e-05, "loss": 0.929, "step": 14957 }, { "epoch": 0.13, "learning_rate": 1.0556976219208632e-05, "loss": 0.9636, "step": 14958 }, { "epoch": 0.13, "learning_rate": 1.055590000795187e-05, "loss": 0.9003, "step": 14959 }, { "epoch": 0.13, "learning_rate": 1.0554823790236509e-05, "loss": 0.8366, "step": 14960 }, { "epoch": 0.13, "learning_rate": 1.055374756607505e-05, "loss": 0.912, "step": 14961 }, { "epoch": 0.13, "learning_rate": 1.0552671335479996e-05, "loss": 0.8947, "step": 14962 }, { "epoch": 0.13, "learning_rate": 1.0551595098463848e-05, "loss": 0.8671, "step": 14963 }, { "epoch": 0.13, "learning_rate": 1.0550518855039117e-05, "loss": 0.9145, "step": 14964 }, { "epoch": 0.13, "learning_rate": 1.0549442605218305e-05, "loss": 0.8901, "step": 14965 }, { "epoch": 0.13, "learning_rate": 1.0548366349013915e-05, "loss": 0.9108, "step": 14966 }, { "epoch": 0.13, "learning_rate": 1.0547290086438453e-05, "loss": 0.9384, "step": 14967 }, { "epoch": 0.13, "learning_rate": 1.0546213817504421e-05, "loss": 0.8924, "step": 14968 }, { "epoch": 0.13, "learning_rate": 1.0545137542224322e-05, "loss": 0.8871, "step": 14969 }, { "epoch": 0.13, "learning_rate": 1.0544061260610665e-05, "loss": 0.8604, "step": 14970 }, { "epoch": 0.13, "learning_rate": 1.054298497267595e-05, "loss": 0.9447, "step": 14971 }, { "epoch": 0.13, "learning_rate": 1.0541908678432684e-05, "loss": 0.8956, "step": 14972 }, { "epoch": 0.13, "learning_rate": 1.0540832377893372e-05, "loss": 0.9064, "step": 14973 }, { "epoch": 0.13, "learning_rate": 1.053975607107052e-05, "loss": 0.8926, "step": 14974 }, { "epoch": 0.13, "learning_rate": 1.0538679757976627e-05, "loss": 0.8981, "step": 14975 }, { "epoch": 0.13, "learning_rate": 1.0537603438624205e-05, "loss": 0.8782, "step": 14976 }, { "epoch": 0.13, "learning_rate": 1.0536527113025756e-05, "loss": 0.9181, "step": 14977 }, { "epoch": 0.13, "learning_rate": 1.0535450781193784e-05, "loss": 0.8854, "step": 14978 }, { "epoch": 0.13, "learning_rate": 1.0534374443140796e-05, "loss": 0.8773, "step": 14979 }, { "epoch": 0.13, "learning_rate": 1.0533298098879295e-05, "loss": 0.9081, "step": 14980 }, { "epoch": 0.13, "learning_rate": 1.0532221748421786e-05, "loss": 0.8901, "step": 14981 }, { "epoch": 0.13, "learning_rate": 1.053114539178078e-05, "loss": 0.8707, "step": 14982 }, { "epoch": 0.13, "learning_rate": 1.0530069028968777e-05, "loss": 0.8912, "step": 14983 }, { "epoch": 0.13, "learning_rate": 1.0528992659998283e-05, "loss": 0.9133, "step": 14984 }, { "epoch": 0.13, "learning_rate": 1.0527916284881805e-05, "loss": 0.9071, "step": 14985 }, { "epoch": 0.13, "learning_rate": 1.0526839903631849e-05, "loss": 0.8933, "step": 14986 }, { "epoch": 0.13, "learning_rate": 1.0525763516260917e-05, "loss": 0.9013, "step": 14987 }, { "epoch": 0.13, "learning_rate": 1.0524687122781519e-05, "loss": 0.8784, "step": 14988 }, { "epoch": 0.13, "learning_rate": 1.0523610723206159e-05, "loss": 0.8673, "step": 14989 }, { "epoch": 0.13, "learning_rate": 1.0522534317547342e-05, "loss": 0.8726, "step": 14990 }, { "epoch": 0.13, "learning_rate": 1.0521457905817576e-05, "loss": 0.9018, "step": 14991 }, { "epoch": 0.13, "learning_rate": 1.0520381488029369e-05, "loss": 0.928, "step": 14992 }, { "epoch": 0.13, "learning_rate": 1.0519305064195222e-05, "loss": 0.9103, "step": 14993 }, { "epoch": 0.13, "learning_rate": 1.0518228634327643e-05, "loss": 0.8916, "step": 14994 }, { "epoch": 0.13, "learning_rate": 1.051715219843914e-05, "loss": 0.8913, "step": 14995 }, { "epoch": 0.13, "learning_rate": 1.0516075756542216e-05, "loss": 0.8973, "step": 14996 }, { "epoch": 0.13, "learning_rate": 1.0514999308649383e-05, "loss": 0.9162, "step": 14997 }, { "epoch": 0.13, "learning_rate": 1.0513922854773142e-05, "loss": 0.912, "step": 14998 }, { "epoch": 0.13, "learning_rate": 1.0512846394926003e-05, "loss": 0.8748, "step": 14999 }, { "epoch": 0.13, "learning_rate": 1.051176992912047e-05, "loss": 0.8853, "step": 15000 }, { "epoch": 0.13, "learning_rate": 1.051069345736905e-05, "loss": 0.9004, "step": 15001 }, { "epoch": 0.13, "learning_rate": 1.0509616979684252e-05, "loss": 0.8959, "step": 15002 }, { "epoch": 0.13, "learning_rate": 1.0508540496078582e-05, "loss": 0.8556, "step": 15003 }, { "epoch": 0.13, "learning_rate": 1.0507464006564545e-05, "loss": 0.8563, "step": 15004 }, { "epoch": 0.13, "learning_rate": 1.050638751115465e-05, "loss": 0.9, "step": 15005 }, { "epoch": 0.13, "learning_rate": 1.05053110098614e-05, "loss": 0.8821, "step": 15006 }, { "epoch": 0.13, "learning_rate": 1.0504234502697313e-05, "loss": 0.8811, "step": 15007 }, { "epoch": 0.13, "learning_rate": 1.050315798967488e-05, "loss": 0.8714, "step": 15008 }, { "epoch": 0.13, "learning_rate": 1.0502081470806622e-05, "loss": 0.9208, "step": 15009 }, { "epoch": 0.13, "learning_rate": 1.050100494610504e-05, "loss": 0.8718, "step": 15010 }, { "epoch": 0.13, "learning_rate": 1.049992841558264e-05, "loss": 0.8986, "step": 15011 }, { "epoch": 0.13, "learning_rate": 1.0498851879251934e-05, "loss": 0.8691, "step": 15012 }, { "epoch": 0.13, "learning_rate": 1.0497775337125431e-05, "loss": 0.8889, "step": 15013 }, { "epoch": 0.13, "learning_rate": 1.0496698789215629e-05, "loss": 0.8874, "step": 15014 }, { "epoch": 0.13, "learning_rate": 1.0495622235535046e-05, "loss": 0.8948, "step": 15015 }, { "epoch": 0.13, "learning_rate": 1.0494545676096183e-05, "loss": 0.898, "step": 15016 }, { "epoch": 0.13, "learning_rate": 1.049346911091155e-05, "loss": 0.9272, "step": 15017 }, { "epoch": 0.13, "learning_rate": 1.0492392539993654e-05, "loss": 0.9, "step": 15018 }, { "epoch": 0.13, "learning_rate": 1.0491315963355006e-05, "loss": 0.8694, "step": 15019 }, { "epoch": 0.13, "learning_rate": 1.0490239381008112e-05, "loss": 0.9293, "step": 15020 }, { "epoch": 0.13, "learning_rate": 1.0489162792965478e-05, "loss": 0.9038, "step": 15021 }, { "epoch": 0.13, "learning_rate": 1.0488086199239615e-05, "loss": 0.9425, "step": 15022 }, { "epoch": 0.13, "learning_rate": 1.048700959984303e-05, "loss": 0.8711, "step": 15023 }, { "epoch": 0.13, "learning_rate": 1.0485932994788228e-05, "loss": 0.8924, "step": 15024 }, { "epoch": 0.13, "learning_rate": 1.0484856384087727e-05, "loss": 0.9111, "step": 15025 }, { "epoch": 0.13, "learning_rate": 1.0483779767754025e-05, "loss": 0.8851, "step": 15026 }, { "epoch": 0.13, "learning_rate": 1.0482703145799632e-05, "loss": 0.9093, "step": 15027 }, { "epoch": 0.13, "learning_rate": 1.0481626518237062e-05, "loss": 0.8768, "step": 15028 }, { "epoch": 0.13, "learning_rate": 1.0480549885078823e-05, "loss": 0.897, "step": 15029 }, { "epoch": 0.13, "learning_rate": 1.0479473246337416e-05, "loss": 0.9165, "step": 15030 }, { "epoch": 0.13, "learning_rate": 1.0478396602025357e-05, "loss": 0.9078, "step": 15031 }, { "epoch": 0.13, "learning_rate": 1.0477319952155156e-05, "loss": 0.9032, "step": 15032 }, { "epoch": 0.13, "learning_rate": 1.0476243296739312e-05, "loss": 0.8819, "step": 15033 }, { "epoch": 0.13, "learning_rate": 1.0475166635790343e-05, "loss": 0.88, "step": 15034 }, { "epoch": 0.13, "learning_rate": 1.0474089969320756e-05, "loss": 0.8754, "step": 15035 }, { "epoch": 0.13, "learning_rate": 1.0473013297343057e-05, "loss": 0.8889, "step": 15036 }, { "epoch": 0.13, "learning_rate": 1.047193661986976e-05, "loss": 0.9114, "step": 15037 }, { "epoch": 0.13, "learning_rate": 1.0470859936913371e-05, "loss": 0.8927, "step": 15038 }, { "epoch": 0.13, "learning_rate": 1.0469783248486398e-05, "loss": 0.9269, "step": 15039 }, { "epoch": 0.13, "learning_rate": 1.0468706554601353e-05, "loss": 0.8557, "step": 15040 }, { "epoch": 0.13, "learning_rate": 1.0467629855270745e-05, "loss": 0.9009, "step": 15041 }, { "epoch": 0.13, "learning_rate": 1.046655315050708e-05, "loss": 0.908, "step": 15042 }, { "epoch": 0.13, "learning_rate": 1.0465476440322872e-05, "loss": 0.8868, "step": 15043 }, { "epoch": 0.13, "learning_rate": 1.0464399724730627e-05, "loss": 0.8772, "step": 15044 }, { "epoch": 0.13, "learning_rate": 1.0463323003742857e-05, "loss": 0.8871, "step": 15045 }, { "epoch": 0.13, "learning_rate": 1.0462246277372072e-05, "loss": 0.8752, "step": 15046 }, { "epoch": 0.13, "learning_rate": 1.046116954563078e-05, "loss": 0.8569, "step": 15047 }, { "epoch": 0.13, "learning_rate": 1.046009280853149e-05, "loss": 0.9187, "step": 15048 }, { "epoch": 0.13, "learning_rate": 1.0459016066086713e-05, "loss": 0.8871, "step": 15049 }, { "epoch": 0.13, "learning_rate": 1.0457939318308963e-05, "loss": 0.8829, "step": 15050 }, { "epoch": 0.13, "learning_rate": 1.0456862565210741e-05, "loss": 0.8811, "step": 15051 }, { "epoch": 0.13, "learning_rate": 1.0455785806804564e-05, "loss": 0.8711, "step": 15052 }, { "epoch": 0.13, "learning_rate": 1.0454709043102942e-05, "loss": 0.8602, "step": 15053 }, { "epoch": 0.13, "learning_rate": 1.0453632274118376e-05, "loss": 0.8793, "step": 15054 }, { "epoch": 0.13, "learning_rate": 1.0452555499863393e-05, "loss": 0.9112, "step": 15055 }, { "epoch": 0.13, "learning_rate": 1.0451478720350488e-05, "loss": 0.8405, "step": 15056 }, { "epoch": 0.14, "learning_rate": 1.0450401935592177e-05, "loss": 0.8966, "step": 15057 }, { "epoch": 0.14, "learning_rate": 1.044932514560097e-05, "loss": 0.8568, "step": 15058 }, { "epoch": 0.14, "learning_rate": 1.0448248350389382e-05, "loss": 0.9352, "step": 15059 }, { "epoch": 0.14, "learning_rate": 1.0447171549969917e-05, "loss": 0.9192, "step": 15060 }, { "epoch": 0.14, "learning_rate": 1.0446094744355087e-05, "loss": 0.885, "step": 15061 }, { "epoch": 0.14, "learning_rate": 1.0445017933557404e-05, "loss": 0.8513, "step": 15062 }, { "epoch": 0.14, "learning_rate": 1.0443941117589378e-05, "loss": 0.8889, "step": 15063 }, { "epoch": 0.14, "learning_rate": 1.044286429646352e-05, "loss": 0.897, "step": 15064 }, { "epoch": 0.14, "learning_rate": 1.0441787470192341e-05, "loss": 0.8709, "step": 15065 }, { "epoch": 0.14, "learning_rate": 1.0440710638788354e-05, "loss": 0.882, "step": 15066 }, { "epoch": 0.14, "learning_rate": 1.0439633802264062e-05, "loss": 0.8743, "step": 15067 }, { "epoch": 0.14, "learning_rate": 1.0438556960631987e-05, "loss": 0.8486, "step": 15068 }, { "epoch": 0.14, "learning_rate": 1.0437480113904632e-05, "loss": 0.8766, "step": 15069 }, { "epoch": 0.14, "learning_rate": 1.043640326209451e-05, "loss": 0.9083, "step": 15070 }, { "epoch": 0.14, "learning_rate": 1.0435326405214137e-05, "loss": 0.9361, "step": 15071 }, { "epoch": 0.14, "learning_rate": 1.0434249543276016e-05, "loss": 0.8721, "step": 15072 }, { "epoch": 0.14, "learning_rate": 1.0433172676292662e-05, "loss": 0.8931, "step": 15073 }, { "epoch": 0.14, "learning_rate": 1.0432095804276587e-05, "loss": 0.9131, "step": 15074 }, { "epoch": 0.14, "learning_rate": 1.0431018927240306e-05, "loss": 0.8762, "step": 15075 }, { "epoch": 0.14, "learning_rate": 1.0429942045196324e-05, "loss": 0.8874, "step": 15076 }, { "epoch": 0.14, "learning_rate": 1.0428865158157155e-05, "loss": 0.9552, "step": 15077 }, { "epoch": 0.14, "learning_rate": 1.0427788266135312e-05, "loss": 0.9354, "step": 15078 }, { "epoch": 0.14, "learning_rate": 1.04267113691433e-05, "loss": 0.8593, "step": 15079 }, { "epoch": 0.14, "learning_rate": 1.0425634467193643e-05, "loss": 0.874, "step": 15080 }, { "epoch": 0.14, "learning_rate": 1.0424557560298843e-05, "loss": 0.8851, "step": 15081 }, { "epoch": 0.14, "learning_rate": 1.0423480648471414e-05, "loss": 0.8848, "step": 15082 }, { "epoch": 0.14, "learning_rate": 1.0422403731723867e-05, "loss": 0.8753, "step": 15083 }, { "epoch": 0.14, "learning_rate": 1.042132681006872e-05, "loss": 0.8966, "step": 15084 }, { "epoch": 0.14, "learning_rate": 1.0420249883518476e-05, "loss": 0.9295, "step": 15085 }, { "epoch": 0.14, "learning_rate": 1.0419172952085651e-05, "loss": 0.9092, "step": 15086 }, { "epoch": 0.14, "learning_rate": 1.0418096015782763e-05, "loss": 0.9397, "step": 15087 }, { "epoch": 0.14, "learning_rate": 1.0417019074622312e-05, "loss": 0.9169, "step": 15088 }, { "epoch": 0.14, "learning_rate": 1.0415942128616821e-05, "loss": 0.9049, "step": 15089 }, { "epoch": 0.14, "learning_rate": 1.0414865177778797e-05, "loss": 0.8689, "step": 15090 }, { "epoch": 0.14, "learning_rate": 1.041378822212075e-05, "loss": 0.8831, "step": 15091 }, { "epoch": 0.14, "learning_rate": 1.0412711261655202e-05, "loss": 0.9209, "step": 15092 }, { "epoch": 0.14, "learning_rate": 1.0411634296394655e-05, "loss": 0.9081, "step": 15093 }, { "epoch": 0.14, "learning_rate": 1.0410557326351627e-05, "loss": 0.9026, "step": 15094 }, { "epoch": 0.14, "learning_rate": 1.0409480351538627e-05, "loss": 0.8796, "step": 15095 }, { "epoch": 0.14, "learning_rate": 1.0408403371968175e-05, "loss": 0.8774, "step": 15096 }, { "epoch": 0.14, "learning_rate": 1.0407326387652772e-05, "loss": 0.9164, "step": 15097 }, { "epoch": 0.14, "learning_rate": 1.0406249398604941e-05, "loss": 0.8871, "step": 15098 }, { "epoch": 0.14, "learning_rate": 1.0405172404837193e-05, "loss": 0.9056, "step": 15099 }, { "epoch": 0.14, "learning_rate": 1.0404095406362034e-05, "loss": 0.8771, "step": 15100 }, { "epoch": 0.14, "learning_rate": 1.0403018403191985e-05, "loss": 0.8804, "step": 15101 }, { "epoch": 0.14, "learning_rate": 1.0401941395339552e-05, "loss": 0.8761, "step": 15102 }, { "epoch": 0.14, "learning_rate": 1.0400864382817253e-05, "loss": 0.9069, "step": 15103 }, { "epoch": 0.14, "learning_rate": 1.0399787365637599e-05, "loss": 0.8539, "step": 15104 }, { "epoch": 0.14, "learning_rate": 1.0398710343813108e-05, "loss": 0.8757, "step": 15105 }, { "epoch": 0.14, "learning_rate": 1.0397633317356283e-05, "loss": 0.9106, "step": 15106 }, { "epoch": 0.14, "learning_rate": 1.0396556286279644e-05, "loss": 0.8979, "step": 15107 }, { "epoch": 0.14, "learning_rate": 1.0395479250595709e-05, "loss": 0.8905, "step": 15108 }, { "epoch": 0.14, "learning_rate": 1.0394402210316979e-05, "loss": 0.8897, "step": 15109 }, { "epoch": 0.14, "learning_rate": 1.0393325165455975e-05, "loss": 0.9202, "step": 15110 }, { "epoch": 0.14, "learning_rate": 1.039224811602521e-05, "loss": 0.8849, "step": 15111 }, { "epoch": 0.14, "learning_rate": 1.0391171062037195e-05, "loss": 0.9001, "step": 15112 }, { "epoch": 0.14, "learning_rate": 1.0390094003504447e-05, "loss": 0.916, "step": 15113 }, { "epoch": 0.14, "learning_rate": 1.0389016940439479e-05, "loss": 0.9014, "step": 15114 }, { "epoch": 0.14, "learning_rate": 1.0387939872854802e-05, "loss": 0.9058, "step": 15115 }, { "epoch": 0.14, "learning_rate": 1.0386862800762932e-05, "loss": 0.9281, "step": 15116 }, { "epoch": 0.14, "learning_rate": 1.038578572417638e-05, "loss": 0.8589, "step": 15117 }, { "epoch": 0.14, "learning_rate": 1.0384708643107664e-05, "loss": 0.8497, "step": 15118 }, { "epoch": 0.14, "learning_rate": 1.0383631557569293e-05, "loss": 0.8582, "step": 15119 }, { "epoch": 0.14, "learning_rate": 1.0382554467573785e-05, "loss": 0.876, "step": 15120 }, { "epoch": 0.14, "learning_rate": 1.0381477373133652e-05, "loss": 0.8501, "step": 15121 }, { "epoch": 0.14, "learning_rate": 1.0380400274261406e-05, "loss": 0.9123, "step": 15122 }, { "epoch": 0.14, "learning_rate": 1.0379323170969567e-05, "loss": 0.8632, "step": 15123 }, { "epoch": 0.14, "learning_rate": 1.0378246063270643e-05, "loss": 0.9049, "step": 15124 }, { "epoch": 0.14, "learning_rate": 1.0377168951177148e-05, "loss": 0.8725, "step": 15125 }, { "epoch": 0.14, "learning_rate": 1.0376091834701605e-05, "loss": 0.8838, "step": 15126 }, { "epoch": 0.14, "learning_rate": 1.0375014713856516e-05, "loss": 0.9351, "step": 15127 }, { "epoch": 0.14, "learning_rate": 1.0373937588654404e-05, "loss": 0.8983, "step": 15128 }, { "epoch": 0.14, "learning_rate": 1.037286045910778e-05, "loss": 0.9254, "step": 15129 }, { "epoch": 0.14, "learning_rate": 1.037178332522916e-05, "loss": 0.8848, "step": 15130 }, { "epoch": 0.14, "learning_rate": 1.0370706187031057e-05, "loss": 0.9252, "step": 15131 }, { "epoch": 0.14, "learning_rate": 1.0369629044525983e-05, "loss": 0.8765, "step": 15132 }, { "epoch": 0.14, "learning_rate": 1.0368551897726462e-05, "loss": 0.8892, "step": 15133 }, { "epoch": 0.14, "learning_rate": 1.0367474746644993e-05, "loss": 0.897, "step": 15134 }, { "epoch": 0.14, "learning_rate": 1.0366397591294107e-05, "loss": 0.8968, "step": 15135 }, { "epoch": 0.14, "learning_rate": 1.0365320431686308e-05, "loss": 0.8593, "step": 15136 }, { "epoch": 0.14, "learning_rate": 1.0364243267834116e-05, "loss": 0.9114, "step": 15137 }, { "epoch": 0.14, "learning_rate": 1.0363166099750044e-05, "loss": 0.8722, "step": 15138 }, { "epoch": 0.14, "learning_rate": 1.0362088927446604e-05, "loss": 0.9377, "step": 15139 }, { "epoch": 0.14, "learning_rate": 1.0361011750936314e-05, "loss": 0.8568, "step": 15140 }, { "epoch": 0.14, "learning_rate": 1.035993457023169e-05, "loss": 0.886, "step": 15141 }, { "epoch": 0.14, "learning_rate": 1.0358857385345246e-05, "loss": 0.8934, "step": 15142 }, { "epoch": 0.14, "learning_rate": 1.0357780196289494e-05, "loss": 0.9226, "step": 15143 }, { "epoch": 0.14, "learning_rate": 1.0356703003076954e-05, "loss": 0.8905, "step": 15144 }, { "epoch": 0.14, "learning_rate": 1.0355625805720138e-05, "loss": 0.8705, "step": 15145 }, { "epoch": 0.14, "learning_rate": 1.0354548604231563e-05, "loss": 0.8885, "step": 15146 }, { "epoch": 0.14, "learning_rate": 1.0353471398623743e-05, "loss": 0.9168, "step": 15147 }, { "epoch": 0.14, "learning_rate": 1.0352394188909192e-05, "loss": 0.8798, "step": 15148 }, { "epoch": 0.14, "learning_rate": 1.0351316975100427e-05, "loss": 0.9124, "step": 15149 }, { "epoch": 0.14, "learning_rate": 1.0350239757209963e-05, "loss": 0.8838, "step": 15150 }, { "epoch": 0.14, "learning_rate": 1.0349162535250316e-05, "loss": 0.8452, "step": 15151 }, { "epoch": 0.14, "learning_rate": 1.0348085309234001e-05, "loss": 0.898, "step": 15152 }, { "epoch": 0.14, "learning_rate": 1.0347008079173535e-05, "loss": 0.9124, "step": 15153 }, { "epoch": 0.14, "learning_rate": 1.0345930845081432e-05, "loss": 0.8877, "step": 15154 }, { "epoch": 0.14, "learning_rate": 1.0344853606970206e-05, "loss": 0.899, "step": 15155 }, { "epoch": 0.14, "learning_rate": 1.0343776364852373e-05, "loss": 0.8776, "step": 15156 }, { "epoch": 0.14, "learning_rate": 1.0342699118740454e-05, "loss": 0.9225, "step": 15157 }, { "epoch": 0.14, "learning_rate": 1.0341621868646958e-05, "loss": 0.8391, "step": 15158 }, { "epoch": 0.14, "learning_rate": 1.0340544614584402e-05, "loss": 0.8904, "step": 15159 }, { "epoch": 0.14, "learning_rate": 1.0339467356565309e-05, "loss": 0.8659, "step": 15160 }, { "epoch": 0.14, "learning_rate": 1.0338390094602186e-05, "loss": 0.9061, "step": 15161 }, { "epoch": 0.14, "learning_rate": 1.033731282870755e-05, "loss": 0.879, "step": 15162 }, { "epoch": 0.14, "learning_rate": 1.0336235558893926e-05, "loss": 0.9052, "step": 15163 }, { "epoch": 0.14, "learning_rate": 1.0335158285173819e-05, "loss": 0.8711, "step": 15164 }, { "epoch": 0.14, "learning_rate": 1.0334081007559747e-05, "loss": 0.8995, "step": 15165 }, { "epoch": 0.14, "learning_rate": 1.033300372606423e-05, "loss": 0.885, "step": 15166 }, { "epoch": 0.14, "learning_rate": 1.0331926440699787e-05, "loss": 0.8676, "step": 15167 }, { "epoch": 0.14, "learning_rate": 1.0330849151478924e-05, "loss": 0.8271, "step": 15168 }, { "epoch": 0.14, "learning_rate": 1.0329771858414166e-05, "loss": 0.8513, "step": 15169 }, { "epoch": 0.14, "learning_rate": 1.0328694561518028e-05, "loss": 0.9071, "step": 15170 }, { "epoch": 0.14, "learning_rate": 1.0327617260803017e-05, "loss": 0.9011, "step": 15171 }, { "epoch": 0.14, "learning_rate": 1.0326539956281665e-05, "loss": 0.8821, "step": 15172 }, { "epoch": 0.14, "learning_rate": 1.0325462647966478e-05, "loss": 0.8795, "step": 15173 }, { "epoch": 0.14, "learning_rate": 1.0324385335869973e-05, "loss": 0.9254, "step": 15174 }, { "epoch": 0.14, "learning_rate": 1.032330802000467e-05, "loss": 0.8948, "step": 15175 }, { "epoch": 0.14, "learning_rate": 1.0322230700383087e-05, "loss": 0.9047, "step": 15176 }, { "epoch": 0.14, "learning_rate": 1.0321153377017734e-05, "loss": 0.9147, "step": 15177 }, { "epoch": 0.14, "learning_rate": 1.0320076049921133e-05, "loss": 0.8623, "step": 15178 }, { "epoch": 0.14, "learning_rate": 1.03189987191058e-05, "loss": 0.89, "step": 15179 }, { "epoch": 0.14, "learning_rate": 1.0317921384584245e-05, "loss": 0.885, "step": 15180 }, { "epoch": 0.14, "learning_rate": 1.0316844046368997e-05, "loss": 0.9202, "step": 15181 }, { "epoch": 0.14, "learning_rate": 1.0315766704472563e-05, "loss": 0.8904, "step": 15182 }, { "epoch": 0.14, "learning_rate": 1.0314689358907461e-05, "loss": 0.8642, "step": 15183 }, { "epoch": 0.14, "learning_rate": 1.0313612009686217e-05, "loss": 0.9044, "step": 15184 }, { "epoch": 0.14, "learning_rate": 1.0312534656821336e-05, "loss": 0.8812, "step": 15185 }, { "epoch": 0.14, "learning_rate": 1.0311457300325339e-05, "loss": 0.8549, "step": 15186 }, { "epoch": 0.14, "learning_rate": 1.0310379940210744e-05, "loss": 0.8933, "step": 15187 }, { "epoch": 0.14, "learning_rate": 1.0309302576490072e-05, "loss": 0.8799, "step": 15188 }, { "epoch": 0.14, "learning_rate": 1.0308225209175832e-05, "loss": 0.8734, "step": 15189 }, { "epoch": 0.14, "learning_rate": 1.030714783828055e-05, "loss": 0.8906, "step": 15190 }, { "epoch": 0.14, "learning_rate": 1.0306070463816735e-05, "loss": 0.924, "step": 15191 }, { "epoch": 0.14, "learning_rate": 1.0304993085796909e-05, "loss": 0.902, "step": 15192 }, { "epoch": 0.14, "learning_rate": 1.0303915704233586e-05, "loss": 0.8684, "step": 15193 }, { "epoch": 0.14, "learning_rate": 1.0302838319139288e-05, "loss": 0.9292, "step": 15194 }, { "epoch": 0.14, "learning_rate": 1.0301760930526527e-05, "loss": 0.9114, "step": 15195 }, { "epoch": 0.14, "learning_rate": 1.0300683538407826e-05, "loss": 0.8685, "step": 15196 }, { "epoch": 0.14, "learning_rate": 1.02996061427957e-05, "loss": 0.9007, "step": 15197 }, { "epoch": 0.14, "learning_rate": 1.0298528743702664e-05, "loss": 0.8658, "step": 15198 }, { "epoch": 0.14, "learning_rate": 1.0297451341141237e-05, "loss": 0.8719, "step": 15199 }, { "epoch": 0.14, "learning_rate": 1.029637393512394e-05, "loss": 0.8861, "step": 15200 }, { "epoch": 0.14, "learning_rate": 1.0295296525663286e-05, "loss": 0.8915, "step": 15201 }, { "epoch": 0.14, "learning_rate": 1.0294219112771796e-05, "loss": 0.8888, "step": 15202 }, { "epoch": 0.14, "learning_rate": 1.0293141696461984e-05, "loss": 0.9245, "step": 15203 }, { "epoch": 0.14, "learning_rate": 1.0292064276746368e-05, "loss": 0.843, "step": 15204 }, { "epoch": 0.14, "learning_rate": 1.029098685363747e-05, "loss": 0.9119, "step": 15205 }, { "epoch": 0.14, "learning_rate": 1.0289909427147806e-05, "loss": 0.9132, "step": 15206 }, { "epoch": 0.14, "learning_rate": 1.0288831997289892e-05, "loss": 0.8871, "step": 15207 }, { "epoch": 0.14, "learning_rate": 1.0287754564076249e-05, "loss": 0.8908, "step": 15208 }, { "epoch": 0.14, "learning_rate": 1.028667712751939e-05, "loss": 0.8911, "step": 15209 }, { "epoch": 0.14, "learning_rate": 1.028559968763184e-05, "loss": 0.8917, "step": 15210 }, { "epoch": 0.14, "learning_rate": 1.028452224442611e-05, "loss": 0.9305, "step": 15211 }, { "epoch": 0.14, "learning_rate": 1.028344479791472e-05, "loss": 0.8658, "step": 15212 }, { "epoch": 0.14, "learning_rate": 1.0282367348110194e-05, "loss": 0.9238, "step": 15213 }, { "epoch": 0.14, "learning_rate": 1.0281289895025037e-05, "loss": 0.8342, "step": 15214 }, { "epoch": 0.14, "learning_rate": 1.0280212438671783e-05, "loss": 0.9183, "step": 15215 }, { "epoch": 0.14, "learning_rate": 1.0279134979062938e-05, "loss": 0.9167, "step": 15216 }, { "epoch": 0.14, "learning_rate": 1.0278057516211026e-05, "loss": 0.8977, "step": 15217 }, { "epoch": 0.14, "learning_rate": 1.0276980050128567e-05, "loss": 0.906, "step": 15218 }, { "epoch": 0.14, "learning_rate": 1.0275902580828074e-05, "loss": 0.9099, "step": 15219 }, { "epoch": 0.14, "learning_rate": 1.0274825108322067e-05, "loss": 0.9076, "step": 15220 }, { "epoch": 0.14, "learning_rate": 1.0273747632623065e-05, "loss": 0.9297, "step": 15221 }, { "epoch": 0.14, "learning_rate": 1.0272670153743589e-05, "loss": 0.8827, "step": 15222 }, { "epoch": 0.14, "learning_rate": 1.0271592671696151e-05, "loss": 0.8597, "step": 15223 }, { "epoch": 0.14, "learning_rate": 1.0270515186493276e-05, "loss": 0.8958, "step": 15224 }, { "epoch": 0.14, "learning_rate": 1.0269437698147482e-05, "loss": 0.898, "step": 15225 }, { "epoch": 0.14, "learning_rate": 1.026836020667128e-05, "loss": 0.9049, "step": 15226 }, { "epoch": 0.14, "learning_rate": 1.0267282712077198e-05, "loss": 0.92, "step": 15227 }, { "epoch": 0.14, "learning_rate": 1.026620521437775e-05, "loss": 0.9235, "step": 15228 }, { "epoch": 0.14, "learning_rate": 1.0265127713585454e-05, "loss": 0.8726, "step": 15229 }, { "epoch": 0.14, "learning_rate": 1.0264050209712832e-05, "loss": 0.8698, "step": 15230 }, { "epoch": 0.14, "learning_rate": 1.0262972702772401e-05, "loss": 0.8875, "step": 15231 }, { "epoch": 0.14, "learning_rate": 1.0261895192776676e-05, "loss": 0.8748, "step": 15232 }, { "epoch": 0.14, "learning_rate": 1.0260817679738182e-05, "loss": 0.8701, "step": 15233 }, { "epoch": 0.14, "learning_rate": 1.0259740163669438e-05, "loss": 0.8997, "step": 15234 }, { "epoch": 0.14, "learning_rate": 1.0258662644582954e-05, "loss": 0.8808, "step": 15235 }, { "epoch": 0.14, "learning_rate": 1.0257585122491261e-05, "loss": 0.9305, "step": 15236 }, { "epoch": 0.14, "learning_rate": 1.025650759740687e-05, "loss": 0.9373, "step": 15237 }, { "epoch": 0.14, "learning_rate": 1.0255430069342299e-05, "loss": 0.8935, "step": 15238 }, { "epoch": 0.14, "learning_rate": 1.0254352538310075e-05, "loss": 0.8975, "step": 15239 }, { "epoch": 0.14, "learning_rate": 1.025327500432271e-05, "loss": 0.881, "step": 15240 }, { "epoch": 0.14, "learning_rate": 1.0252197467392724e-05, "loss": 0.9059, "step": 15241 }, { "epoch": 0.14, "learning_rate": 1.0251119927532635e-05, "loss": 0.9235, "step": 15242 }, { "epoch": 0.14, "learning_rate": 1.025004238475497e-05, "loss": 0.912, "step": 15243 }, { "epoch": 0.14, "learning_rate": 1.024896483907224e-05, "loss": 0.8492, "step": 15244 }, { "epoch": 0.14, "learning_rate": 1.0247887290496968e-05, "loss": 0.8559, "step": 15245 }, { "epoch": 0.14, "learning_rate": 1.0246809739041672e-05, "loss": 0.9108, "step": 15246 }, { "epoch": 0.14, "learning_rate": 1.0245732184718868e-05, "loss": 0.8915, "step": 15247 }, { "epoch": 0.14, "learning_rate": 1.0244654627541084e-05, "loss": 0.8289, "step": 15248 }, { "epoch": 0.14, "learning_rate": 1.0243577067520832e-05, "loss": 0.8913, "step": 15249 }, { "epoch": 0.14, "learning_rate": 1.024249950467063e-05, "loss": 0.8752, "step": 15250 }, { "epoch": 0.14, "learning_rate": 1.0241421939003004e-05, "loss": 0.8925, "step": 15251 }, { "epoch": 0.14, "learning_rate": 1.0240344370530473e-05, "loss": 0.8932, "step": 15252 }, { "epoch": 0.14, "learning_rate": 1.023926679926555e-05, "loss": 0.8983, "step": 15253 }, { "epoch": 0.14, "learning_rate": 1.0238189225220761e-05, "loss": 0.8729, "step": 15254 }, { "epoch": 0.14, "learning_rate": 1.0237111648408621e-05, "loss": 0.8621, "step": 15255 }, { "epoch": 0.14, "learning_rate": 1.0236034068841654e-05, "loss": 0.8907, "step": 15256 }, { "epoch": 0.14, "learning_rate": 1.0234956486532375e-05, "loss": 0.8708, "step": 15257 }, { "epoch": 0.14, "learning_rate": 1.0233878901493305e-05, "loss": 0.8678, "step": 15258 }, { "epoch": 0.14, "learning_rate": 1.023280131373697e-05, "loss": 0.9142, "step": 15259 }, { "epoch": 0.14, "learning_rate": 1.0231723723275877e-05, "loss": 0.8893, "step": 15260 }, { "epoch": 0.14, "learning_rate": 1.0230646130122557e-05, "loss": 0.8808, "step": 15261 }, { "epoch": 0.14, "learning_rate": 1.0229568534289525e-05, "loss": 0.865, "step": 15262 }, { "epoch": 0.14, "learning_rate": 1.0228490935789301e-05, "loss": 0.8506, "step": 15263 }, { "epoch": 0.14, "learning_rate": 1.0227413334634406e-05, "loss": 0.921, "step": 15264 }, { "epoch": 0.14, "learning_rate": 1.022633573083736e-05, "loss": 0.9023, "step": 15265 }, { "epoch": 0.14, "learning_rate": 1.0225258124410682e-05, "loss": 0.8743, "step": 15266 }, { "epoch": 0.14, "learning_rate": 1.022418051536689e-05, "loss": 0.8946, "step": 15267 }, { "epoch": 0.14, "learning_rate": 1.0223102903718509e-05, "loss": 0.8903, "step": 15268 }, { "epoch": 0.14, "learning_rate": 1.0222025289478055e-05, "loss": 0.8886, "step": 15269 }, { "epoch": 0.14, "learning_rate": 1.0220947672658048e-05, "loss": 0.8684, "step": 15270 }, { "epoch": 0.14, "learning_rate": 1.0219870053271013e-05, "loss": 0.8708, "step": 15271 }, { "epoch": 0.14, "learning_rate": 1.0218792431329459e-05, "loss": 0.8594, "step": 15272 }, { "epoch": 0.14, "learning_rate": 1.021771480684592e-05, "loss": 0.9041, "step": 15273 }, { "epoch": 0.14, "learning_rate": 1.0216637179832907e-05, "loss": 0.8637, "step": 15274 }, { "epoch": 0.14, "learning_rate": 1.0215559550302942e-05, "loss": 0.892, "step": 15275 }, { "epoch": 0.14, "learning_rate": 1.0214481918268548e-05, "loss": 0.9149, "step": 15276 }, { "epoch": 0.14, "learning_rate": 1.0213404283742241e-05, "loss": 0.9004, "step": 15277 }, { "epoch": 0.14, "learning_rate": 1.0212326646736543e-05, "loss": 0.8772, "step": 15278 }, { "epoch": 0.14, "learning_rate": 1.0211249007263977e-05, "loss": 0.9075, "step": 15279 }, { "epoch": 0.14, "learning_rate": 1.021017136533706e-05, "loss": 0.8919, "step": 15280 }, { "epoch": 0.14, "learning_rate": 1.020909372096831e-05, "loss": 0.8809, "step": 15281 }, { "epoch": 0.14, "learning_rate": 1.0208016074170257e-05, "loss": 0.8872, "step": 15282 }, { "epoch": 0.14, "learning_rate": 1.0206938424955411e-05, "loss": 0.8762, "step": 15283 }, { "epoch": 0.14, "learning_rate": 1.0205860773336297e-05, "loss": 0.8774, "step": 15284 }, { "epoch": 0.14, "learning_rate": 1.0204783119325437e-05, "loss": 0.8647, "step": 15285 }, { "epoch": 0.14, "learning_rate": 1.0203705462935348e-05, "loss": 0.8737, "step": 15286 }, { "epoch": 0.14, "learning_rate": 1.020262780417855e-05, "loss": 0.8668, "step": 15287 }, { "epoch": 0.14, "learning_rate": 1.0201550143067567e-05, "loss": 0.9173, "step": 15288 }, { "epoch": 0.14, "learning_rate": 1.020047247961492e-05, "loss": 0.8987, "step": 15289 }, { "epoch": 0.14, "learning_rate": 1.0199394813833123e-05, "loss": 0.889, "step": 15290 }, { "epoch": 0.14, "learning_rate": 1.0198317145734707e-05, "loss": 0.8929, "step": 15291 }, { "epoch": 0.14, "learning_rate": 1.0197239475332186e-05, "loss": 0.9164, "step": 15292 }, { "epoch": 0.14, "learning_rate": 1.0196161802638078e-05, "loss": 0.8814, "step": 15293 }, { "epoch": 0.14, "learning_rate": 1.0195084127664912e-05, "loss": 0.9479, "step": 15294 }, { "epoch": 0.14, "learning_rate": 1.01940064504252e-05, "loss": 0.8989, "step": 15295 }, { "epoch": 0.14, "learning_rate": 1.0192928770931467e-05, "loss": 0.8835, "step": 15296 }, { "epoch": 0.14, "learning_rate": 1.0191851089196234e-05, "loss": 0.8824, "step": 15297 }, { "epoch": 0.14, "learning_rate": 1.0190773405232024e-05, "loss": 0.8759, "step": 15298 }, { "epoch": 0.14, "learning_rate": 1.0189695719051354e-05, "loss": 0.9018, "step": 15299 }, { "epoch": 0.14, "learning_rate": 1.0188618030666743e-05, "loss": 0.8862, "step": 15300 }, { "epoch": 0.14, "learning_rate": 1.018754034009072e-05, "loss": 0.8986, "step": 15301 }, { "epoch": 0.14, "learning_rate": 1.0186462647335798e-05, "loss": 0.8671, "step": 15302 }, { "epoch": 0.14, "learning_rate": 1.01853849524145e-05, "loss": 0.8854, "step": 15303 }, { "epoch": 0.14, "learning_rate": 1.018430725533935e-05, "loss": 0.8797, "step": 15304 }, { "epoch": 0.14, "learning_rate": 1.0183229556122865e-05, "loss": 0.8883, "step": 15305 }, { "epoch": 0.14, "learning_rate": 1.018215185477757e-05, "loss": 0.9156, "step": 15306 }, { "epoch": 0.14, "learning_rate": 1.0181074151315982e-05, "loss": 0.8714, "step": 15307 }, { "epoch": 0.14, "learning_rate": 1.0179996445750624e-05, "loss": 0.8671, "step": 15308 }, { "epoch": 0.14, "learning_rate": 1.0178918738094016e-05, "loss": 0.9148, "step": 15309 }, { "epoch": 0.14, "learning_rate": 1.0177841028358683e-05, "loss": 0.9013, "step": 15310 }, { "epoch": 0.14, "learning_rate": 1.0176763316557142e-05, "loss": 0.8876, "step": 15311 }, { "epoch": 0.14, "learning_rate": 1.0175685602701915e-05, "loss": 0.8775, "step": 15312 }, { "epoch": 0.14, "learning_rate": 1.0174607886805522e-05, "loss": 0.8894, "step": 15313 }, { "epoch": 0.14, "learning_rate": 1.017353016888049e-05, "loss": 0.8672, "step": 15314 }, { "epoch": 0.14, "learning_rate": 1.0172452448939332e-05, "loss": 0.8662, "step": 15315 }, { "epoch": 0.14, "learning_rate": 1.0171374726994573e-05, "loss": 0.8859, "step": 15316 }, { "epoch": 0.14, "learning_rate": 1.0170297003058738e-05, "loss": 0.8974, "step": 15317 }, { "epoch": 0.14, "learning_rate": 1.016921927714434e-05, "loss": 0.8829, "step": 15318 }, { "epoch": 0.14, "learning_rate": 1.0168141549263911e-05, "loss": 0.8946, "step": 15319 }, { "epoch": 0.14, "learning_rate": 1.0167063819429961e-05, "loss": 0.8991, "step": 15320 }, { "epoch": 0.14, "learning_rate": 1.016598608765502e-05, "loss": 0.8887, "step": 15321 }, { "epoch": 0.14, "learning_rate": 1.0164908353951606e-05, "loss": 0.871, "step": 15322 }, { "epoch": 0.14, "learning_rate": 1.016383061833224e-05, "loss": 0.8915, "step": 15323 }, { "epoch": 0.14, "learning_rate": 1.0162752880809444e-05, "loss": 0.86, "step": 15324 }, { "epoch": 0.14, "learning_rate": 1.0161675141395736e-05, "loss": 0.9214, "step": 15325 }, { "epoch": 0.14, "learning_rate": 1.0160597400103649e-05, "loss": 0.9255, "step": 15326 }, { "epoch": 0.14, "learning_rate": 1.015951965694569e-05, "loss": 0.9178, "step": 15327 }, { "epoch": 0.14, "learning_rate": 1.0158441911934388e-05, "loss": 0.8825, "step": 15328 }, { "epoch": 0.14, "learning_rate": 1.0157364165082266e-05, "loss": 0.9031, "step": 15329 }, { "epoch": 0.14, "learning_rate": 1.0156286416401842e-05, "loss": 0.8427, "step": 15330 }, { "epoch": 0.14, "learning_rate": 1.0155208665905638e-05, "loss": 0.9127, "step": 15331 }, { "epoch": 0.14, "learning_rate": 1.0154130913606178e-05, "loss": 0.8966, "step": 15332 }, { "epoch": 0.14, "learning_rate": 1.015305315951598e-05, "loss": 0.8642, "step": 15333 }, { "epoch": 0.14, "learning_rate": 1.0151975403647568e-05, "loss": 0.905, "step": 15334 }, { "epoch": 0.14, "learning_rate": 1.0150897646013465e-05, "loss": 0.8693, "step": 15335 }, { "epoch": 0.14, "learning_rate": 1.0149819886626187e-05, "loss": 0.8658, "step": 15336 }, { "epoch": 0.14, "learning_rate": 1.0148742125498263e-05, "loss": 0.8951, "step": 15337 }, { "epoch": 0.14, "learning_rate": 1.0147664362642214e-05, "loss": 0.8556, "step": 15338 }, { "epoch": 0.14, "learning_rate": 1.0146586598070554e-05, "loss": 0.8814, "step": 15339 }, { "epoch": 0.14, "learning_rate": 1.0145508831795812e-05, "loss": 0.8737, "step": 15340 }, { "epoch": 0.14, "learning_rate": 1.0144431063830506e-05, "loss": 0.8766, "step": 15341 }, { "epoch": 0.14, "learning_rate": 1.0143353294187162e-05, "loss": 0.8915, "step": 15342 }, { "epoch": 0.14, "learning_rate": 1.0142275522878298e-05, "loss": 0.93, "step": 15343 }, { "epoch": 0.14, "learning_rate": 1.0141197749916438e-05, "loss": 0.8597, "step": 15344 }, { "epoch": 0.14, "learning_rate": 1.0140119975314102e-05, "loss": 0.9033, "step": 15345 }, { "epoch": 0.14, "learning_rate": 1.0139042199083812e-05, "loss": 0.8602, "step": 15346 }, { "epoch": 0.14, "learning_rate": 1.0137964421238096e-05, "loss": 0.8691, "step": 15347 }, { "epoch": 0.14, "learning_rate": 1.0136886641789467e-05, "loss": 0.9182, "step": 15348 }, { "epoch": 0.14, "learning_rate": 1.0135808860750448e-05, "loss": 0.8996, "step": 15349 }, { "epoch": 0.14, "learning_rate": 1.0134731078133568e-05, "loss": 0.9093, "step": 15350 }, { "epoch": 0.14, "learning_rate": 1.0133653293951342e-05, "loss": 0.9136, "step": 15351 }, { "epoch": 0.14, "learning_rate": 1.0132575508216296e-05, "loss": 0.8701, "step": 15352 }, { "epoch": 0.14, "learning_rate": 1.0131497720940952e-05, "loss": 0.8802, "step": 15353 }, { "epoch": 0.14, "learning_rate": 1.013041993213783e-05, "loss": 0.8794, "step": 15354 }, { "epoch": 0.14, "learning_rate": 1.0129342141819451e-05, "loss": 0.8513, "step": 15355 }, { "epoch": 0.14, "learning_rate": 1.012826434999834e-05, "loss": 0.8817, "step": 15356 }, { "epoch": 0.15, "learning_rate": 1.012718655668702e-05, "loss": 0.9068, "step": 15357 }, { "epoch": 0.15, "learning_rate": 1.0126108761898006e-05, "loss": 0.8863, "step": 15358 }, { "epoch": 0.15, "learning_rate": 1.0125030965643827e-05, "loss": 0.86, "step": 15359 }, { "epoch": 0.15, "learning_rate": 1.0123953167937006e-05, "loss": 0.9287, "step": 15360 }, { "epoch": 0.15, "learning_rate": 1.0122875368790062e-05, "loss": 0.8904, "step": 15361 }, { "epoch": 0.15, "learning_rate": 1.0121797568215514e-05, "loss": 0.8738, "step": 15362 }, { "epoch": 0.15, "learning_rate": 1.0120719766225892e-05, "loss": 0.8742, "step": 15363 }, { "epoch": 0.15, "learning_rate": 1.011964196283371e-05, "loss": 0.8654, "step": 15364 }, { "epoch": 0.15, "learning_rate": 1.0118564158051498e-05, "loss": 0.9043, "step": 15365 }, { "epoch": 0.15, "learning_rate": 1.0117486351891772e-05, "loss": 0.8877, "step": 15366 }, { "epoch": 0.15, "learning_rate": 1.0116408544367057e-05, "loss": 0.8777, "step": 15367 }, { "epoch": 0.15, "learning_rate": 1.0115330735489877e-05, "loss": 0.8638, "step": 15368 }, { "epoch": 0.15, "learning_rate": 1.0114252925272751e-05, "loss": 0.8819, "step": 15369 }, { "epoch": 0.15, "learning_rate": 1.0113175113728202e-05, "loss": 0.8397, "step": 15370 }, { "epoch": 0.15, "learning_rate": 1.0112097300868753e-05, "loss": 0.832, "step": 15371 }, { "epoch": 0.15, "learning_rate": 1.0111019486706928e-05, "loss": 0.8662, "step": 15372 }, { "epoch": 0.15, "learning_rate": 1.0109941671255244e-05, "loss": 0.894, "step": 15373 }, { "epoch": 0.15, "learning_rate": 1.0108863854526229e-05, "loss": 0.9052, "step": 15374 }, { "epoch": 0.15, "learning_rate": 1.0107786036532404e-05, "loss": 0.9282, "step": 15375 }, { "epoch": 0.15, "learning_rate": 1.0106708217286287e-05, "loss": 0.8567, "step": 15376 }, { "epoch": 0.15, "learning_rate": 1.010563039680041e-05, "loss": 0.9583, "step": 15377 }, { "epoch": 0.15, "learning_rate": 1.0104552575087285e-05, "loss": 0.9297, "step": 15378 }, { "epoch": 0.15, "learning_rate": 1.0103474752159441e-05, "loss": 0.896, "step": 15379 }, { "epoch": 0.15, "learning_rate": 1.0102396928029396e-05, "loss": 0.8736, "step": 15380 }, { "epoch": 0.15, "learning_rate": 1.0101319102709678e-05, "loss": 0.8845, "step": 15381 }, { "epoch": 0.15, "learning_rate": 1.0100241276212803e-05, "loss": 0.8804, "step": 15382 }, { "epoch": 0.15, "learning_rate": 1.0099163448551302e-05, "loss": 0.9139, "step": 15383 }, { "epoch": 0.15, "learning_rate": 1.009808561973769e-05, "loss": 0.8844, "step": 15384 }, { "epoch": 0.15, "learning_rate": 1.009700778978449e-05, "loss": 0.8991, "step": 15385 }, { "epoch": 0.15, "learning_rate": 1.0095929958704226e-05, "loss": 0.8652, "step": 15386 }, { "epoch": 0.15, "learning_rate": 1.0094852126509425e-05, "loss": 0.8788, "step": 15387 }, { "epoch": 0.15, "learning_rate": 1.0093774293212602e-05, "loss": 0.871, "step": 15388 }, { "epoch": 0.15, "learning_rate": 1.0092696458826284e-05, "loss": 0.8729, "step": 15389 }, { "epoch": 0.15, "learning_rate": 1.0091618623362994e-05, "loss": 0.8716, "step": 15390 }, { "epoch": 0.15, "learning_rate": 1.009054078683525e-05, "loss": 0.9151, "step": 15391 }, { "epoch": 0.15, "learning_rate": 1.008946294925558e-05, "loss": 0.89, "step": 15392 }, { "epoch": 0.15, "learning_rate": 1.0088385110636505e-05, "loss": 0.9285, "step": 15393 }, { "epoch": 0.15, "learning_rate": 1.0087307270990543e-05, "loss": 0.8541, "step": 15394 }, { "epoch": 0.15, "learning_rate": 1.0086229430330226e-05, "loss": 0.8982, "step": 15395 }, { "epoch": 0.15, "learning_rate": 1.0085151588668066e-05, "loss": 0.8915, "step": 15396 }, { "epoch": 0.15, "learning_rate": 1.0084073746016596e-05, "loss": 0.9163, "step": 15397 }, { "epoch": 0.15, "learning_rate": 1.008299590238833e-05, "loss": 0.8866, "step": 15398 }, { "epoch": 0.15, "learning_rate": 1.0081918057795799e-05, "loss": 0.8785, "step": 15399 }, { "epoch": 0.15, "learning_rate": 1.0080840212251518e-05, "loss": 0.846, "step": 15400 }, { "epoch": 0.15, "learning_rate": 1.0079762365768013e-05, "loss": 0.9139, "step": 15401 }, { "epoch": 0.15, "learning_rate": 1.007868451835781e-05, "loss": 0.8365, "step": 15402 }, { "epoch": 0.15, "learning_rate": 1.0077606670033422e-05, "loss": 0.8864, "step": 15403 }, { "epoch": 0.15, "learning_rate": 1.0076528820807382e-05, "loss": 0.8836, "step": 15404 }, { "epoch": 0.15, "learning_rate": 1.0075450970692207e-05, "loss": 0.8803, "step": 15405 }, { "epoch": 0.15, "learning_rate": 1.0074373119700422e-05, "loss": 0.8818, "step": 15406 }, { "epoch": 0.15, "learning_rate": 1.0073295267844549e-05, "loss": 0.9293, "step": 15407 }, { "epoch": 0.15, "learning_rate": 1.0072217415137112e-05, "loss": 0.8758, "step": 15408 }, { "epoch": 0.15, "learning_rate": 1.0071139561590633e-05, "loss": 0.8938, "step": 15409 }, { "epoch": 0.15, "learning_rate": 1.0070061707217634e-05, "loss": 0.8884, "step": 15410 }, { "epoch": 0.15, "learning_rate": 1.0068983852030638e-05, "loss": 0.9104, "step": 15411 }, { "epoch": 0.15, "learning_rate": 1.006790599604217e-05, "loss": 0.8914, "step": 15412 }, { "epoch": 0.15, "learning_rate": 1.006682813926475e-05, "loss": 0.8781, "step": 15413 }, { "epoch": 0.15, "learning_rate": 1.0065750281710904e-05, "loss": 0.8645, "step": 15414 }, { "epoch": 0.15, "learning_rate": 1.006467242339315e-05, "loss": 0.9127, "step": 15415 }, { "epoch": 0.15, "learning_rate": 1.0063594564324014e-05, "loss": 0.9048, "step": 15416 }, { "epoch": 0.15, "learning_rate": 1.006251670451602e-05, "loss": 0.8911, "step": 15417 }, { "epoch": 0.15, "learning_rate": 1.006143884398169e-05, "loss": 0.8813, "step": 15418 }, { "epoch": 0.15, "learning_rate": 1.0060360982733541e-05, "loss": 0.9164, "step": 15419 }, { "epoch": 0.15, "learning_rate": 1.0059283120784107e-05, "loss": 0.8811, "step": 15420 }, { "epoch": 0.15, "learning_rate": 1.0058205258145903e-05, "loss": 0.9025, "step": 15421 }, { "epoch": 0.15, "learning_rate": 1.0057127394831453e-05, "loss": 0.9006, "step": 15422 }, { "epoch": 0.15, "learning_rate": 1.0056049530853284e-05, "loss": 0.8837, "step": 15423 }, { "epoch": 0.15, "learning_rate": 1.0054971666223915e-05, "loss": 0.8742, "step": 15424 }, { "epoch": 0.15, "learning_rate": 1.0053893800955869e-05, "loss": 0.9121, "step": 15425 }, { "epoch": 0.15, "learning_rate": 1.0052815935061669e-05, "loss": 0.896, "step": 15426 }, { "epoch": 0.15, "learning_rate": 1.0051738068553842e-05, "loss": 0.8679, "step": 15427 }, { "epoch": 0.15, "learning_rate": 1.0050660201444902e-05, "loss": 0.8765, "step": 15428 }, { "epoch": 0.15, "learning_rate": 1.0049582333747383e-05, "loss": 0.8835, "step": 15429 }, { "epoch": 0.15, "learning_rate": 1.0048504465473801e-05, "loss": 0.87, "step": 15430 }, { "epoch": 0.15, "learning_rate": 1.0047426596636678e-05, "loss": 0.925, "step": 15431 }, { "epoch": 0.15, "learning_rate": 1.0046348727248543e-05, "loss": 0.8922, "step": 15432 }, { "epoch": 0.15, "learning_rate": 1.0045270857321914e-05, "loss": 0.8507, "step": 15433 }, { "epoch": 0.15, "learning_rate": 1.0044192986869314e-05, "loss": 0.8744, "step": 15434 }, { "epoch": 0.15, "learning_rate": 1.0043115115903269e-05, "loss": 0.9042, "step": 15435 }, { "epoch": 0.15, "learning_rate": 1.0042037244436303e-05, "loss": 0.8825, "step": 15436 }, { "epoch": 0.15, "learning_rate": 1.0040959372480935e-05, "loss": 0.8559, "step": 15437 }, { "epoch": 0.15, "learning_rate": 1.0039881500049687e-05, "loss": 0.8902, "step": 15438 }, { "epoch": 0.15, "learning_rate": 1.0038803627155088e-05, "loss": 0.9074, "step": 15439 }, { "epoch": 0.15, "learning_rate": 1.0037725753809654e-05, "loss": 0.8818, "step": 15440 }, { "epoch": 0.15, "learning_rate": 1.0036647880025914e-05, "loss": 0.887, "step": 15441 }, { "epoch": 0.15, "learning_rate": 1.0035570005816386e-05, "loss": 0.8817, "step": 15442 }, { "epoch": 0.15, "learning_rate": 1.0034492131193598e-05, "loss": 0.8608, "step": 15443 }, { "epoch": 0.15, "learning_rate": 1.0033414256170071e-05, "loss": 0.8708, "step": 15444 }, { "epoch": 0.15, "learning_rate": 1.0032336380758329e-05, "loss": 0.9012, "step": 15445 }, { "epoch": 0.15, "learning_rate": 1.0031258504970894e-05, "loss": 0.8728, "step": 15446 }, { "epoch": 0.15, "learning_rate": 1.0030180628820285e-05, "loss": 0.8631, "step": 15447 }, { "epoch": 0.15, "learning_rate": 1.0029102752319033e-05, "loss": 0.8911, "step": 15448 }, { "epoch": 0.15, "learning_rate": 1.0028024875479656e-05, "loss": 0.9236, "step": 15449 }, { "epoch": 0.15, "learning_rate": 1.0026946998314678e-05, "loss": 0.8618, "step": 15450 }, { "epoch": 0.15, "learning_rate": 1.002586912083662e-05, "loss": 0.9055, "step": 15451 }, { "epoch": 0.15, "learning_rate": 1.002479124305801e-05, "loss": 0.8799, "step": 15452 }, { "epoch": 0.15, "learning_rate": 1.0023713364991368e-05, "loss": 0.8816, "step": 15453 }, { "epoch": 0.15, "learning_rate": 1.0022635486649217e-05, "loss": 0.8701, "step": 15454 }, { "epoch": 0.15, "learning_rate": 1.002155760804408e-05, "loss": 0.8988, "step": 15455 }, { "epoch": 0.15, "learning_rate": 1.0020479729188481e-05, "loss": 0.8891, "step": 15456 }, { "epoch": 0.15, "learning_rate": 1.0019401850094945e-05, "loss": 0.9209, "step": 15457 }, { "epoch": 0.15, "learning_rate": 1.0018323970775989e-05, "loss": 0.9136, "step": 15458 }, { "epoch": 0.15, "learning_rate": 1.0017246091244141e-05, "loss": 0.871, "step": 15459 }, { "epoch": 0.15, "learning_rate": 1.0016168211511928e-05, "loss": 0.8528, "step": 15460 }, { "epoch": 0.15, "learning_rate": 1.0015090331591863e-05, "loss": 0.8941, "step": 15461 }, { "epoch": 0.15, "learning_rate": 1.0014012451496476e-05, "loss": 0.8832, "step": 15462 }, { "epoch": 0.15, "learning_rate": 1.0012934571238286e-05, "loss": 0.8793, "step": 15463 }, { "epoch": 0.15, "learning_rate": 1.0011856690829822e-05, "loss": 0.8269, "step": 15464 }, { "epoch": 0.15, "learning_rate": 1.0010778810283602e-05, "loss": 0.8963, "step": 15465 }, { "epoch": 0.15, "learning_rate": 1.0009700929612151e-05, "loss": 0.8677, "step": 15466 }, { "epoch": 0.15, "learning_rate": 1.0008623048827992e-05, "loss": 0.8936, "step": 15467 }, { "epoch": 0.15, "learning_rate": 1.0007545167943647e-05, "loss": 0.8701, "step": 15468 }, { "epoch": 0.15, "learning_rate": 1.0006467286971642e-05, "loss": 0.9178, "step": 15469 }, { "epoch": 0.15, "learning_rate": 1.0005389405924498e-05, "loss": 0.8854, "step": 15470 }, { "epoch": 0.15, "learning_rate": 1.0004311524814736e-05, "loss": 0.8771, "step": 15471 }, { "epoch": 0.15, "learning_rate": 1.0003233643654883e-05, "loss": 0.87, "step": 15472 }, { "epoch": 0.15, "learning_rate": 1.0002155762457462e-05, "loss": 0.8702, "step": 15473 }, { "epoch": 0.15, "learning_rate": 1.0001077881234991e-05, "loss": 0.8431, "step": 15474 }, { "epoch": 0.15, "learning_rate": 1e-05, "loss": 0.8611, "step": 15475 }, { "epoch": 0.15, "learning_rate": 9.99892211876501e-06, "loss": 0.863, "step": 15476 }, { "epoch": 0.15, "learning_rate": 9.997844237542543e-06, "loss": 0.9109, "step": 15477 }, { "epoch": 0.15, "learning_rate": 9.99676635634512e-06, "loss": 0.8551, "step": 15478 }, { "epoch": 0.15, "learning_rate": 9.995688475185267e-06, "loss": 0.8876, "step": 15479 }, { "epoch": 0.15, "learning_rate": 9.994610594075505e-06, "loss": 0.8966, "step": 15480 }, { "epoch": 0.15, "learning_rate": 9.993532713028362e-06, "loss": 0.9292, "step": 15481 }, { "epoch": 0.15, "learning_rate": 9.992454832056357e-06, "loss": 0.8736, "step": 15482 }, { "epoch": 0.15, "learning_rate": 9.99137695117201e-06, "loss": 0.9117, "step": 15483 }, { "epoch": 0.15, "learning_rate": 9.990299070387854e-06, "loss": 0.9054, "step": 15484 }, { "epoch": 0.15, "learning_rate": 9.989221189716403e-06, "loss": 0.939, "step": 15485 }, { "epoch": 0.15, "learning_rate": 9.98814330917018e-06, "loss": 0.8898, "step": 15486 }, { "epoch": 0.15, "learning_rate": 9.987065428761714e-06, "loss": 0.8514, "step": 15487 }, { "epoch": 0.15, "learning_rate": 9.985987548503529e-06, "loss": 0.9076, "step": 15488 }, { "epoch": 0.15, "learning_rate": 9.98490966840814e-06, "loss": 0.8527, "step": 15489 }, { "epoch": 0.15, "learning_rate": 9.983831788488075e-06, "loss": 0.8767, "step": 15490 }, { "epoch": 0.15, "learning_rate": 9.982753908755864e-06, "loss": 0.8869, "step": 15491 }, { "epoch": 0.15, "learning_rate": 9.981676029224013e-06, "loss": 0.8719, "step": 15492 }, { "epoch": 0.15, "learning_rate": 9.980598149905059e-06, "loss": 0.9203, "step": 15493 }, { "epoch": 0.15, "learning_rate": 9.97952027081152e-06, "loss": 0.907, "step": 15494 }, { "epoch": 0.15, "learning_rate": 9.978442391955923e-06, "loss": 0.8962, "step": 15495 }, { "epoch": 0.15, "learning_rate": 9.977364513350788e-06, "loss": 0.8597, "step": 15496 }, { "epoch": 0.15, "learning_rate": 9.976286635008635e-06, "loss": 0.9095, "step": 15497 }, { "epoch": 0.15, "learning_rate": 9.975208756941992e-06, "loss": 0.9188, "step": 15498 }, { "epoch": 0.15, "learning_rate": 9.974130879163382e-06, "loss": 0.8911, "step": 15499 }, { "epoch": 0.15, "learning_rate": 9.973053001685327e-06, "loss": 0.8864, "step": 15500 }, { "epoch": 0.15, "learning_rate": 9.971975124520347e-06, "loss": 0.8758, "step": 15501 }, { "epoch": 0.15, "learning_rate": 9.970897247680972e-06, "loss": 0.8829, "step": 15502 }, { "epoch": 0.15, "learning_rate": 9.969819371179718e-06, "loss": 0.9037, "step": 15503 }, { "epoch": 0.15, "learning_rate": 9.968741495029111e-06, "loss": 0.9312, "step": 15504 }, { "epoch": 0.15, "learning_rate": 9.967663619241671e-06, "loss": 0.934, "step": 15505 }, { "epoch": 0.15, "learning_rate": 9.966585743829932e-06, "loss": 0.9004, "step": 15506 }, { "epoch": 0.15, "learning_rate": 9.965507868806404e-06, "loss": 0.8982, "step": 15507 }, { "epoch": 0.15, "learning_rate": 9.964429994183614e-06, "loss": 0.8874, "step": 15508 }, { "epoch": 0.15, "learning_rate": 9.963352119974091e-06, "loss": 0.9199, "step": 15509 }, { "epoch": 0.15, "learning_rate": 9.962274246190351e-06, "loss": 0.8852, "step": 15510 }, { "epoch": 0.15, "learning_rate": 9.961196372844916e-06, "loss": 0.9228, "step": 15511 }, { "epoch": 0.15, "learning_rate": 9.960118499950313e-06, "loss": 0.9258, "step": 15512 }, { "epoch": 0.15, "learning_rate": 9.95904062751907e-06, "loss": 0.925, "step": 15513 }, { "epoch": 0.15, "learning_rate": 9.9579627555637e-06, "loss": 0.8729, "step": 15514 }, { "epoch": 0.15, "learning_rate": 9.956884884096731e-06, "loss": 0.9039, "step": 15515 }, { "epoch": 0.15, "learning_rate": 9.955807013130689e-06, "loss": 0.8931, "step": 15516 }, { "epoch": 0.15, "learning_rate": 9.95472914267809e-06, "loss": 0.9116, "step": 15517 }, { "epoch": 0.15, "learning_rate": 9.95365127275146e-06, "loss": 0.9258, "step": 15518 }, { "epoch": 0.15, "learning_rate": 9.952573403363323e-06, "loss": 0.9113, "step": 15519 }, { "epoch": 0.15, "learning_rate": 9.951495534526204e-06, "loss": 0.9022, "step": 15520 }, { "epoch": 0.15, "learning_rate": 9.950417666252622e-06, "loss": 0.9285, "step": 15521 }, { "epoch": 0.15, "learning_rate": 9.949339798555098e-06, "loss": 0.8602, "step": 15522 }, { "epoch": 0.15, "learning_rate": 9.948261931446163e-06, "loss": 0.8799, "step": 15523 }, { "epoch": 0.15, "learning_rate": 9.947184064938333e-06, "loss": 0.9021, "step": 15524 }, { "epoch": 0.15, "learning_rate": 9.946106199044134e-06, "loss": 0.847, "step": 15525 }, { "epoch": 0.15, "learning_rate": 9.945028333776088e-06, "loss": 0.8842, "step": 15526 }, { "epoch": 0.15, "learning_rate": 9.94395046914672e-06, "loss": 0.8959, "step": 15527 }, { "epoch": 0.15, "learning_rate": 9.942872605168548e-06, "loss": 0.8999, "step": 15528 }, { "epoch": 0.15, "learning_rate": 9.941794741854097e-06, "loss": 0.9026, "step": 15529 }, { "epoch": 0.15, "learning_rate": 9.940716879215898e-06, "loss": 0.9, "step": 15530 }, { "epoch": 0.15, "learning_rate": 9.939639017266462e-06, "loss": 0.9243, "step": 15531 }, { "epoch": 0.15, "learning_rate": 9.938561156018314e-06, "loss": 0.8941, "step": 15532 }, { "epoch": 0.15, "learning_rate": 9.937483295483982e-06, "loss": 0.9021, "step": 15533 }, { "epoch": 0.15, "learning_rate": 9.936405435675991e-06, "loss": 0.8744, "step": 15534 }, { "epoch": 0.15, "learning_rate": 9.935327576606854e-06, "loss": 0.888, "step": 15535 }, { "epoch": 0.15, "learning_rate": 9.934249718289099e-06, "loss": 0.896, "step": 15536 }, { "epoch": 0.15, "learning_rate": 9.933171860735256e-06, "loss": 0.8552, "step": 15537 }, { "epoch": 0.15, "learning_rate": 9.932094003957832e-06, "loss": 0.9091, "step": 15538 }, { "epoch": 0.15, "learning_rate": 9.931016147969364e-06, "loss": 0.8627, "step": 15539 }, { "epoch": 0.15, "learning_rate": 9.929938292782368e-06, "loss": 0.8739, "step": 15540 }, { "epoch": 0.15, "learning_rate": 9.92886043840937e-06, "loss": 0.874, "step": 15541 }, { "epoch": 0.15, "learning_rate": 9.92778258486289e-06, "loss": 0.8596, "step": 15542 }, { "epoch": 0.15, "learning_rate": 9.926704732155453e-06, "loss": 0.903, "step": 15543 }, { "epoch": 0.15, "learning_rate": 9.92562688029958e-06, "loss": 0.9235, "step": 15544 }, { "epoch": 0.15, "learning_rate": 9.924549029307798e-06, "loss": 0.9377, "step": 15545 }, { "epoch": 0.15, "learning_rate": 9.923471179192622e-06, "loss": 0.8971, "step": 15546 }, { "epoch": 0.15, "learning_rate": 9.922393329966581e-06, "loss": 0.9054, "step": 15547 }, { "epoch": 0.15, "learning_rate": 9.921315481642197e-06, "loss": 0.8787, "step": 15548 }, { "epoch": 0.15, "learning_rate": 9.92023763423199e-06, "loss": 0.8727, "step": 15549 }, { "epoch": 0.15, "learning_rate": 9.919159787748485e-06, "loss": 0.8644, "step": 15550 }, { "epoch": 0.15, "learning_rate": 9.918081942204203e-06, "loss": 0.8933, "step": 15551 }, { "epoch": 0.15, "learning_rate": 9.917004097611671e-06, "loss": 0.8608, "step": 15552 }, { "epoch": 0.15, "learning_rate": 9.915926253983407e-06, "loss": 0.8726, "step": 15553 }, { "epoch": 0.15, "learning_rate": 9.914848411331933e-06, "loss": 0.8952, "step": 15554 }, { "epoch": 0.15, "learning_rate": 9.913770569669779e-06, "loss": 0.8884, "step": 15555 }, { "epoch": 0.15, "learning_rate": 9.91269272900946e-06, "loss": 0.8988, "step": 15556 }, { "epoch": 0.15, "learning_rate": 9.911614889363498e-06, "loss": 0.8797, "step": 15557 }, { "epoch": 0.15, "learning_rate": 9.910537050744421e-06, "loss": 0.8612, "step": 15558 }, { "epoch": 0.15, "learning_rate": 9.909459213164752e-06, "loss": 0.8932, "step": 15559 }, { "epoch": 0.15, "learning_rate": 9.908381376637011e-06, "loss": 0.8777, "step": 15560 }, { "epoch": 0.15, "learning_rate": 9.907303541173718e-06, "loss": 0.8959, "step": 15561 }, { "epoch": 0.15, "learning_rate": 9.906225706787402e-06, "loss": 0.8624, "step": 15562 }, { "epoch": 0.15, "learning_rate": 9.90514787349058e-06, "loss": 0.8719, "step": 15563 }, { "epoch": 0.15, "learning_rate": 9.904070041295775e-06, "loss": 0.8791, "step": 15564 }, { "epoch": 0.15, "learning_rate": 9.902992210215512e-06, "loss": 0.8656, "step": 15565 }, { "epoch": 0.15, "learning_rate": 9.901914380262313e-06, "loss": 0.8932, "step": 15566 }, { "epoch": 0.15, "learning_rate": 9.900836551448702e-06, "loss": 0.866, "step": 15567 }, { "epoch": 0.15, "learning_rate": 9.899758723787197e-06, "loss": 0.8178, "step": 15568 }, { "epoch": 0.15, "learning_rate": 9.898680897290325e-06, "loss": 0.899, "step": 15569 }, { "epoch": 0.15, "learning_rate": 9.897603071970605e-06, "loss": 0.8333, "step": 15570 }, { "epoch": 0.15, "learning_rate": 9.896525247840562e-06, "loss": 0.8832, "step": 15571 }, { "epoch": 0.15, "learning_rate": 9.895447424912716e-06, "loss": 0.885, "step": 15572 }, { "epoch": 0.15, "learning_rate": 9.894369603199595e-06, "loss": 0.9272, "step": 15573 }, { "epoch": 0.15, "learning_rate": 9.893291782713715e-06, "loss": 0.8803, "step": 15574 }, { "epoch": 0.15, "learning_rate": 9.892213963467598e-06, "loss": 0.8571, "step": 15575 }, { "epoch": 0.15, "learning_rate": 9.891136145473775e-06, "loss": 0.8785, "step": 15576 }, { "epoch": 0.15, "learning_rate": 9.89005832874476e-06, "loss": 0.9247, "step": 15577 }, { "epoch": 0.15, "learning_rate": 9.888980513293076e-06, "loss": 0.8837, "step": 15578 }, { "epoch": 0.15, "learning_rate": 9.887902699131248e-06, "loss": 0.886, "step": 15579 }, { "epoch": 0.15, "learning_rate": 9.886824886271803e-06, "loss": 0.8713, "step": 15580 }, { "epoch": 0.15, "learning_rate": 9.885747074727252e-06, "loss": 0.9341, "step": 15581 }, { "epoch": 0.15, "learning_rate": 9.884669264510125e-06, "loss": 0.8656, "step": 15582 }, { "epoch": 0.15, "learning_rate": 9.883591455632948e-06, "loss": 0.8828, "step": 15583 }, { "epoch": 0.15, "learning_rate": 9.88251364810823e-06, "loss": 0.9049, "step": 15584 }, { "epoch": 0.15, "learning_rate": 9.881435841948504e-06, "loss": 0.9006, "step": 15585 }, { "epoch": 0.15, "learning_rate": 9.88035803716629e-06, "loss": 0.8674, "step": 15586 }, { "epoch": 0.15, "learning_rate": 9.879280233774113e-06, "loss": 0.8774, "step": 15587 }, { "epoch": 0.15, "learning_rate": 9.878202431784488e-06, "loss": 0.8637, "step": 15588 }, { "epoch": 0.15, "learning_rate": 9.877124631209941e-06, "loss": 0.9128, "step": 15589 }, { "epoch": 0.15, "learning_rate": 9.876046832062996e-06, "loss": 0.8794, "step": 15590 }, { "epoch": 0.15, "learning_rate": 9.874969034356174e-06, "loss": 0.9144, "step": 15591 }, { "epoch": 0.15, "learning_rate": 9.873891238101996e-06, "loss": 0.8701, "step": 15592 }, { "epoch": 0.15, "learning_rate": 9.872813443312984e-06, "loss": 0.9094, "step": 15593 }, { "epoch": 0.15, "learning_rate": 9.871735650001664e-06, "loss": 0.8697, "step": 15594 }, { "epoch": 0.15, "learning_rate": 9.870657858180552e-06, "loss": 0.9219, "step": 15595 }, { "epoch": 0.15, "learning_rate": 9.869580067862174e-06, "loss": 0.8668, "step": 15596 }, { "epoch": 0.15, "learning_rate": 9.86850227905905e-06, "loss": 0.9228, "step": 15597 }, { "epoch": 0.15, "learning_rate": 9.867424491783708e-06, "loss": 0.8568, "step": 15598 }, { "epoch": 0.15, "learning_rate": 9.86634670604866e-06, "loss": 0.8649, "step": 15599 }, { "epoch": 0.15, "learning_rate": 9.865268921866434e-06, "loss": 0.8765, "step": 15600 }, { "epoch": 0.15, "learning_rate": 9.864191139249557e-06, "loss": 0.9527, "step": 15601 }, { "epoch": 0.15, "learning_rate": 9.863113358210536e-06, "loss": 0.8814, "step": 15602 }, { "epoch": 0.15, "learning_rate": 9.862035578761908e-06, "loss": 0.8749, "step": 15603 }, { "epoch": 0.15, "learning_rate": 9.860957800916186e-06, "loss": 0.8462, "step": 15604 }, { "epoch": 0.15, "learning_rate": 9.8598800246859e-06, "loss": 0.8966, "step": 15605 }, { "epoch": 0.15, "learning_rate": 9.858802250083564e-06, "loss": 0.8527, "step": 15606 }, { "epoch": 0.15, "learning_rate": 9.857724477121704e-06, "loss": 0.9011, "step": 15607 }, { "epoch": 0.15, "learning_rate": 9.856646705812842e-06, "loss": 0.9277, "step": 15608 }, { "epoch": 0.15, "learning_rate": 9.855568936169496e-06, "loss": 0.8798, "step": 15609 }, { "epoch": 0.15, "learning_rate": 9.854491168204191e-06, "loss": 0.8971, "step": 15610 }, { "epoch": 0.15, "learning_rate": 9.853413401929448e-06, "loss": 0.9509, "step": 15611 }, { "epoch": 0.15, "learning_rate": 9.852335637357791e-06, "loss": 0.8918, "step": 15612 }, { "epoch": 0.15, "learning_rate": 9.851257874501739e-06, "loss": 0.8936, "step": 15613 }, { "epoch": 0.15, "learning_rate": 9.850180113373815e-06, "loss": 0.8809, "step": 15614 }, { "epoch": 0.15, "learning_rate": 9.84910235398654e-06, "loss": 0.914, "step": 15615 }, { "epoch": 0.15, "learning_rate": 9.848024596352434e-06, "loss": 0.8783, "step": 15616 }, { "epoch": 0.15, "learning_rate": 9.846946840484024e-06, "loss": 0.9312, "step": 15617 }, { "epoch": 0.15, "learning_rate": 9.845869086393824e-06, "loss": 0.8794, "step": 15618 }, { "epoch": 0.15, "learning_rate": 9.844791334094367e-06, "loss": 0.8635, "step": 15619 }, { "epoch": 0.15, "learning_rate": 9.843713583598163e-06, "loss": 0.8745, "step": 15620 }, { "epoch": 0.15, "learning_rate": 9.842635834917735e-06, "loss": 0.8919, "step": 15621 }, { "epoch": 0.15, "learning_rate": 9.841558088065617e-06, "loss": 0.8462, "step": 15622 }, { "epoch": 0.15, "learning_rate": 9.840480343054314e-06, "loss": 0.8902, "step": 15623 }, { "epoch": 0.15, "learning_rate": 9.839402599896355e-06, "loss": 0.8666, "step": 15624 }, { "epoch": 0.15, "learning_rate": 9.838324858604262e-06, "loss": 0.9142, "step": 15625 }, { "epoch": 0.15, "learning_rate": 9.837247119190561e-06, "loss": 0.8454, "step": 15626 }, { "epoch": 0.15, "learning_rate": 9.836169381667763e-06, "loss": 0.9113, "step": 15627 }, { "epoch": 0.15, "learning_rate": 9.835091646048397e-06, "loss": 0.905, "step": 15628 }, { "epoch": 0.15, "learning_rate": 9.834013912344986e-06, "loss": 0.897, "step": 15629 }, { "epoch": 0.15, "learning_rate": 9.832936180570042e-06, "loss": 0.874, "step": 15630 }, { "epoch": 0.15, "learning_rate": 9.831858450736094e-06, "loss": 0.9165, "step": 15631 }, { "epoch": 0.15, "learning_rate": 9.83078072285566e-06, "loss": 0.8628, "step": 15632 }, { "epoch": 0.15, "learning_rate": 9.829702996941267e-06, "loss": 0.8953, "step": 15633 }, { "epoch": 0.15, "learning_rate": 9.82862527300543e-06, "loss": 0.8597, "step": 15634 }, { "epoch": 0.15, "learning_rate": 9.827547551060671e-06, "loss": 0.9012, "step": 15635 }, { "epoch": 0.15, "learning_rate": 9.826469831119514e-06, "loss": 0.8875, "step": 15636 }, { "epoch": 0.15, "learning_rate": 9.82539211319448e-06, "loss": 0.8899, "step": 15637 }, { "epoch": 0.15, "learning_rate": 9.824314397298088e-06, "loss": 0.8752, "step": 15638 }, { "epoch": 0.15, "learning_rate": 9.823236683442861e-06, "loss": 0.8758, "step": 15639 }, { "epoch": 0.15, "learning_rate": 9.822158971641322e-06, "loss": 0.892, "step": 15640 }, { "epoch": 0.15, "learning_rate": 9.821081261905985e-06, "loss": 0.921, "step": 15641 }, { "epoch": 0.15, "learning_rate": 9.82000355424938e-06, "loss": 0.8713, "step": 15642 }, { "epoch": 0.15, "learning_rate": 9.81892584868402e-06, "loss": 0.8921, "step": 15643 }, { "epoch": 0.15, "learning_rate": 9.817848145222436e-06, "loss": 0.9038, "step": 15644 }, { "epoch": 0.15, "learning_rate": 9.816770443877138e-06, "loss": 0.8825, "step": 15645 }, { "epoch": 0.15, "learning_rate": 9.815692744660652e-06, "loss": 0.8886, "step": 15646 }, { "epoch": 0.15, "learning_rate": 9.814615047585505e-06, "loss": 0.9127, "step": 15647 }, { "epoch": 0.15, "learning_rate": 9.813537352664204e-06, "loss": 0.8828, "step": 15648 }, { "epoch": 0.15, "learning_rate": 9.812459659909282e-06, "loss": 0.8749, "step": 15649 }, { "epoch": 0.15, "learning_rate": 9.811381969333257e-06, "loss": 0.8476, "step": 15650 }, { "epoch": 0.15, "learning_rate": 9.810304280948649e-06, "loss": 0.8751, "step": 15651 }, { "epoch": 0.15, "learning_rate": 9.809226594767979e-06, "loss": 0.9045, "step": 15652 }, { "epoch": 0.15, "learning_rate": 9.808148910803766e-06, "loss": 0.9257, "step": 15653 }, { "epoch": 0.15, "learning_rate": 9.807071229068536e-06, "loss": 0.8589, "step": 15654 }, { "epoch": 0.15, "learning_rate": 9.805993549574803e-06, "loss": 0.8868, "step": 15655 }, { "epoch": 0.15, "learning_rate": 9.804915872335093e-06, "loss": 0.8819, "step": 15656 }, { "epoch": 0.15, "learning_rate": 9.803838197361924e-06, "loss": 0.8703, "step": 15657 }, { "epoch": 0.16, "learning_rate": 9.802760524667818e-06, "loss": 0.8906, "step": 15658 }, { "epoch": 0.16, "learning_rate": 9.801682854265294e-06, "loss": 0.8626, "step": 15659 }, { "epoch": 0.16, "learning_rate": 9.800605186166875e-06, "loss": 0.9214, "step": 15660 }, { "epoch": 0.16, "learning_rate": 9.799527520385083e-06, "loss": 0.883, "step": 15661 }, { "epoch": 0.16, "learning_rate": 9.798449856932435e-06, "loss": 0.8694, "step": 15662 }, { "epoch": 0.16, "learning_rate": 9.797372195821451e-06, "loss": 0.8869, "step": 15663 }, { "epoch": 0.16, "learning_rate": 9.796294537064654e-06, "loss": 0.8989, "step": 15664 }, { "epoch": 0.16, "learning_rate": 9.795216880674568e-06, "loss": 0.9116, "step": 15665 }, { "epoch": 0.16, "learning_rate": 9.794139226663706e-06, "loss": 0.8943, "step": 15666 }, { "epoch": 0.16, "learning_rate": 9.793061575044589e-06, "loss": 0.89, "step": 15667 }, { "epoch": 0.16, "learning_rate": 9.791983925829748e-06, "loss": 0.8701, "step": 15668 }, { "epoch": 0.16, "learning_rate": 9.790906279031692e-06, "loss": 0.8983, "step": 15669 }, { "epoch": 0.16, "learning_rate": 9.789828634662942e-06, "loss": 0.8414, "step": 15670 }, { "epoch": 0.16, "learning_rate": 9.788750992736023e-06, "loss": 0.8743, "step": 15671 }, { "epoch": 0.16, "learning_rate": 9.78767335326346e-06, "loss": 0.8959, "step": 15672 }, { "epoch": 0.16, "learning_rate": 9.78659571625776e-06, "loss": 0.8894, "step": 15673 }, { "epoch": 0.16, "learning_rate": 9.785518081731455e-06, "loss": 0.9193, "step": 15674 }, { "epoch": 0.16, "learning_rate": 9.784440449697061e-06, "loss": 0.9057, "step": 15675 }, { "epoch": 0.16, "learning_rate": 9.783362820167096e-06, "loss": 0.8878, "step": 15676 }, { "epoch": 0.16, "learning_rate": 9.782285193154083e-06, "loss": 0.9007, "step": 15677 }, { "epoch": 0.16, "learning_rate": 9.781207568670541e-06, "loss": 0.9424, "step": 15678 }, { "epoch": 0.16, "learning_rate": 9.780129946728994e-06, "loss": 0.898, "step": 15679 }, { "epoch": 0.16, "learning_rate": 9.779052327341955e-06, "loss": 0.8912, "step": 15680 }, { "epoch": 0.16, "learning_rate": 9.777974710521949e-06, "loss": 0.8902, "step": 15681 }, { "epoch": 0.16, "learning_rate": 9.776897096281493e-06, "loss": 0.8268, "step": 15682 }, { "epoch": 0.16, "learning_rate": 9.775819484633114e-06, "loss": 0.8839, "step": 15683 }, { "epoch": 0.16, "learning_rate": 9.774741875589322e-06, "loss": 0.8966, "step": 15684 }, { "epoch": 0.16, "learning_rate": 9.773664269162643e-06, "loss": 0.8937, "step": 15685 }, { "epoch": 0.16, "learning_rate": 9.772586665365597e-06, "loss": 0.8726, "step": 15686 }, { "epoch": 0.16, "learning_rate": 9.771509064210702e-06, "loss": 0.8729, "step": 15687 }, { "epoch": 0.16, "learning_rate": 9.770431465710478e-06, "loss": 0.8988, "step": 15688 }, { "epoch": 0.16, "learning_rate": 9.769353869877444e-06, "loss": 0.9209, "step": 15689 }, { "epoch": 0.16, "learning_rate": 9.768276276724126e-06, "loss": 0.9012, "step": 15690 }, { "epoch": 0.16, "learning_rate": 9.767198686263034e-06, "loss": 0.8857, "step": 15691 }, { "epoch": 0.16, "learning_rate": 9.766121098506695e-06, "loss": 0.8789, "step": 15692 }, { "epoch": 0.16, "learning_rate": 9.76504351346763e-06, "loss": 0.9476, "step": 15693 }, { "epoch": 0.16, "learning_rate": 9.76396593115835e-06, "loss": 0.8666, "step": 15694 }, { "epoch": 0.16, "learning_rate": 9.76288835159138e-06, "loss": 0.9088, "step": 15695 }, { "epoch": 0.16, "learning_rate": 9.76181077477924e-06, "loss": 0.9028, "step": 15696 }, { "epoch": 0.16, "learning_rate": 9.760733200734452e-06, "loss": 0.9174, "step": 15697 }, { "epoch": 0.16, "learning_rate": 9.75965562946953e-06, "loss": 0.8432, "step": 15698 }, { "epoch": 0.16, "learning_rate": 9.758578060996996e-06, "loss": 0.9015, "step": 15699 }, { "epoch": 0.16, "learning_rate": 9.757500495329371e-06, "loss": 0.8478, "step": 15700 }, { "epoch": 0.16, "learning_rate": 9.756422932479173e-06, "loss": 0.8913, "step": 15701 }, { "epoch": 0.16, "learning_rate": 9.75534537245892e-06, "loss": 0.8804, "step": 15702 }, { "epoch": 0.16, "learning_rate": 9.754267815281132e-06, "loss": 0.8538, "step": 15703 }, { "epoch": 0.16, "learning_rate": 9.753190260958333e-06, "loss": 0.8884, "step": 15704 }, { "epoch": 0.16, "learning_rate": 9.752112709503036e-06, "loss": 0.9201, "step": 15705 }, { "epoch": 0.16, "learning_rate": 9.751035160927762e-06, "loss": 0.917, "step": 15706 }, { "epoch": 0.16, "learning_rate": 9.749957615245035e-06, "loss": 0.9092, "step": 15707 }, { "epoch": 0.16, "learning_rate": 9.748880072467366e-06, "loss": 0.9179, "step": 15708 }, { "epoch": 0.16, "learning_rate": 9.74780253260728e-06, "loss": 0.9142, "step": 15709 }, { "epoch": 0.16, "learning_rate": 9.746724995677292e-06, "loss": 0.8947, "step": 15710 }, { "epoch": 0.16, "learning_rate": 9.745647461689932e-06, "loss": 0.9505, "step": 15711 }, { "epoch": 0.16, "learning_rate": 9.744569930657703e-06, "loss": 0.8595, "step": 15712 }, { "epoch": 0.16, "learning_rate": 9.743492402593132e-06, "loss": 0.8822, "step": 15713 }, { "epoch": 0.16, "learning_rate": 9.742414877508744e-06, "loss": 0.9086, "step": 15714 }, { "epoch": 0.16, "learning_rate": 9.741337355417048e-06, "loss": 0.9228, "step": 15715 }, { "epoch": 0.16, "learning_rate": 9.740259836330566e-06, "loss": 0.9162, "step": 15716 }, { "epoch": 0.16, "learning_rate": 9.739182320261818e-06, "loss": 0.8951, "step": 15717 }, { "epoch": 0.16, "learning_rate": 9.738104807223329e-06, "loss": 0.8908, "step": 15718 }, { "epoch": 0.16, "learning_rate": 9.737027297227604e-06, "loss": 0.9004, "step": 15719 }, { "epoch": 0.16, "learning_rate": 9.735949790287171e-06, "loss": 0.8514, "step": 15720 }, { "epoch": 0.16, "learning_rate": 9.73487228641455e-06, "loss": 0.9251, "step": 15721 }, { "epoch": 0.16, "learning_rate": 9.733794785622254e-06, "loss": 0.8612, "step": 15722 }, { "epoch": 0.16, "learning_rate": 9.732717287922805e-06, "loss": 0.903, "step": 15723 }, { "epoch": 0.16, "learning_rate": 9.73163979332872e-06, "loss": 0.8761, "step": 15724 }, { "epoch": 0.16, "learning_rate": 9.730562301852524e-06, "loss": 0.8939, "step": 15725 }, { "epoch": 0.16, "learning_rate": 9.729484813506726e-06, "loss": 0.9001, "step": 15726 }, { "epoch": 0.16, "learning_rate": 9.72840732830385e-06, "loss": 0.9035, "step": 15727 }, { "epoch": 0.16, "learning_rate": 9.727329846256413e-06, "loss": 0.8493, "step": 15728 }, { "epoch": 0.16, "learning_rate": 9.726252367376938e-06, "loss": 0.9445, "step": 15729 }, { "epoch": 0.16, "learning_rate": 9.725174891677937e-06, "loss": 0.8777, "step": 15730 }, { "epoch": 0.16, "learning_rate": 9.724097419171929e-06, "loss": 0.8765, "step": 15731 }, { "epoch": 0.16, "learning_rate": 9.723019949871438e-06, "loss": 0.8976, "step": 15732 }, { "epoch": 0.16, "learning_rate": 9.721942483788976e-06, "loss": 0.8916, "step": 15733 }, { "epoch": 0.16, "learning_rate": 9.720865020937063e-06, "loss": 0.8769, "step": 15734 }, { "epoch": 0.16, "learning_rate": 9.719787561328218e-06, "loss": 0.8684, "step": 15735 }, { "epoch": 0.16, "learning_rate": 9.718710104974965e-06, "loss": 0.854, "step": 15736 }, { "epoch": 0.16, "learning_rate": 9.71763265188981e-06, "loss": 0.8952, "step": 15737 }, { "epoch": 0.16, "learning_rate": 9.716555202085281e-06, "loss": 0.9058, "step": 15738 }, { "epoch": 0.16, "learning_rate": 9.715477755573895e-06, "loss": 0.9062, "step": 15739 }, { "epoch": 0.16, "learning_rate": 9.714400312368165e-06, "loss": 0.901, "step": 15740 }, { "epoch": 0.16, "learning_rate": 9.713322872480611e-06, "loss": 0.8907, "step": 15741 }, { "epoch": 0.16, "learning_rate": 9.712245435923753e-06, "loss": 0.8823, "step": 15742 }, { "epoch": 0.16, "learning_rate": 9.71116800271011e-06, "loss": 0.8951, "step": 15743 }, { "epoch": 0.16, "learning_rate": 9.710090572852197e-06, "loss": 0.866, "step": 15744 }, { "epoch": 0.16, "learning_rate": 9.709013146362532e-06, "loss": 0.8578, "step": 15745 }, { "epoch": 0.16, "learning_rate": 9.707935723253634e-06, "loss": 0.8755, "step": 15746 }, { "epoch": 0.16, "learning_rate": 9.706858303538021e-06, "loss": 0.8868, "step": 15747 }, { "epoch": 0.16, "learning_rate": 9.70578088722821e-06, "loss": 0.9014, "step": 15748 }, { "epoch": 0.16, "learning_rate": 9.704703474336716e-06, "loss": 0.8789, "step": 15749 }, { "epoch": 0.16, "learning_rate": 9.703626064876064e-06, "loss": 0.8573, "step": 15750 }, { "epoch": 0.16, "learning_rate": 9.702548658858764e-06, "loss": 0.8892, "step": 15751 }, { "epoch": 0.16, "learning_rate": 9.70147125629734e-06, "loss": 0.9018, "step": 15752 }, { "epoch": 0.16, "learning_rate": 9.700393857204304e-06, "loss": 0.9045, "step": 15753 }, { "epoch": 0.16, "learning_rate": 9.699316461592176e-06, "loss": 0.8861, "step": 15754 }, { "epoch": 0.16, "learning_rate": 9.698239069473474e-06, "loss": 0.8943, "step": 15755 }, { "epoch": 0.16, "learning_rate": 9.697161680860713e-06, "loss": 0.9184, "step": 15756 }, { "epoch": 0.16, "learning_rate": 9.696084295766417e-06, "loss": 0.9136, "step": 15757 }, { "epoch": 0.16, "learning_rate": 9.695006914203096e-06, "loss": 0.8776, "step": 15758 }, { "epoch": 0.16, "learning_rate": 9.693929536183267e-06, "loss": 0.899, "step": 15759 }, { "epoch": 0.16, "learning_rate": 9.692852161719456e-06, "loss": 0.8994, "step": 15760 }, { "epoch": 0.16, "learning_rate": 9.691774790824171e-06, "loss": 0.9231, "step": 15761 }, { "epoch": 0.16, "learning_rate": 9.69069742350993e-06, "loss": 0.8616, "step": 15762 }, { "epoch": 0.16, "learning_rate": 9.689620059789255e-06, "loss": 0.8869, "step": 15763 }, { "epoch": 0.16, "learning_rate": 9.688542699674663e-06, "loss": 0.8377, "step": 15764 }, { "epoch": 0.16, "learning_rate": 9.687465343178667e-06, "loss": 0.9212, "step": 15765 }, { "epoch": 0.16, "learning_rate": 9.686387990313787e-06, "loss": 0.8863, "step": 15766 }, { "epoch": 0.16, "learning_rate": 9.685310641092542e-06, "loss": 0.9245, "step": 15767 }, { "epoch": 0.16, "learning_rate": 9.68423329552744e-06, "loss": 0.8815, "step": 15768 }, { "epoch": 0.16, "learning_rate": 9.683155953631008e-06, "loss": 0.8703, "step": 15769 }, { "epoch": 0.16, "learning_rate": 9.682078615415755e-06, "loss": 0.8996, "step": 15770 }, { "epoch": 0.16, "learning_rate": 9.681001280894206e-06, "loss": 0.8989, "step": 15771 }, { "epoch": 0.16, "learning_rate": 9.67992395007887e-06, "loss": 0.8712, "step": 15772 }, { "epoch": 0.16, "learning_rate": 9.678846622982269e-06, "loss": 0.8653, "step": 15773 }, { "epoch": 0.16, "learning_rate": 9.677769299616917e-06, "loss": 0.8681, "step": 15774 }, { "epoch": 0.16, "learning_rate": 9.676691979995331e-06, "loss": 0.888, "step": 15775 }, { "epoch": 0.16, "learning_rate": 9.675614664130029e-06, "loss": 0.8763, "step": 15776 }, { "epoch": 0.16, "learning_rate": 9.674537352033525e-06, "loss": 0.8933, "step": 15777 }, { "epoch": 0.16, "learning_rate": 9.67346004371834e-06, "loss": 0.8977, "step": 15778 }, { "epoch": 0.16, "learning_rate": 9.672382739196984e-06, "loss": 0.893, "step": 15779 }, { "epoch": 0.16, "learning_rate": 9.671305438481976e-06, "loss": 0.8859, "step": 15780 }, { "epoch": 0.16, "learning_rate": 9.670228141585836e-06, "loss": 0.9103, "step": 15781 }, { "epoch": 0.16, "learning_rate": 9.669150848521081e-06, "loss": 0.8615, "step": 15782 }, { "epoch": 0.16, "learning_rate": 9.668073559300217e-06, "loss": 0.9484, "step": 15783 }, { "epoch": 0.16, "learning_rate": 9.666996273935771e-06, "loss": 0.8778, "step": 15784 }, { "epoch": 0.16, "learning_rate": 9.665918992440257e-06, "loss": 0.9158, "step": 15785 }, { "epoch": 0.16, "learning_rate": 9.664841714826185e-06, "loss": 0.8334, "step": 15786 }, { "epoch": 0.16, "learning_rate": 9.663764441106079e-06, "loss": 0.8691, "step": 15787 }, { "epoch": 0.16, "learning_rate": 9.662687171292448e-06, "loss": 0.872, "step": 15788 }, { "epoch": 0.16, "learning_rate": 9.661609905397817e-06, "loss": 0.8875, "step": 15789 }, { "epoch": 0.16, "learning_rate": 9.660532643434693e-06, "loss": 0.8266, "step": 15790 }, { "epoch": 0.16, "learning_rate": 9.659455385415597e-06, "loss": 0.8729, "step": 15791 }, { "epoch": 0.16, "learning_rate": 9.658378131353045e-06, "loss": 0.8864, "step": 15792 }, { "epoch": 0.16, "learning_rate": 9.65730088125955e-06, "loss": 0.9068, "step": 15793 }, { "epoch": 0.16, "learning_rate": 9.656223635147629e-06, "loss": 0.9045, "step": 15794 }, { "epoch": 0.16, "learning_rate": 9.655146393029797e-06, "loss": 0.8561, "step": 15795 }, { "epoch": 0.16, "learning_rate": 9.654069154918574e-06, "loss": 0.8992, "step": 15796 }, { "epoch": 0.16, "learning_rate": 9.652991920826469e-06, "loss": 0.8943, "step": 15797 }, { "epoch": 0.16, "learning_rate": 9.651914690766e-06, "loss": 0.8771, "step": 15798 }, { "epoch": 0.16, "learning_rate": 9.650837464749687e-06, "loss": 0.9016, "step": 15799 }, { "epoch": 0.16, "learning_rate": 9.64976024279004e-06, "loss": 0.8558, "step": 15800 }, { "epoch": 0.16, "learning_rate": 9.648683024899577e-06, "loss": 0.893, "step": 15801 }, { "epoch": 0.16, "learning_rate": 9.64760581109081e-06, "loss": 0.8918, "step": 15802 }, { "epoch": 0.16, "learning_rate": 9.646528601376262e-06, "loss": 0.8864, "step": 15803 }, { "epoch": 0.16, "learning_rate": 9.645451395768442e-06, "loss": 0.8838, "step": 15804 }, { "epoch": 0.16, "learning_rate": 9.644374194279864e-06, "loss": 0.8924, "step": 15805 }, { "epoch": 0.16, "learning_rate": 9.643296996923051e-06, "loss": 0.9072, "step": 15806 }, { "epoch": 0.16, "learning_rate": 9.64221980371051e-06, "loss": 0.8571, "step": 15807 }, { "epoch": 0.16, "learning_rate": 9.641142614654755e-06, "loss": 0.8994, "step": 15808 }, { "epoch": 0.16, "learning_rate": 9.64006542976831e-06, "loss": 0.9153, "step": 15809 }, { "epoch": 0.16, "learning_rate": 9.638988249063688e-06, "loss": 0.8785, "step": 15810 }, { "epoch": 0.16, "learning_rate": 9.637911072553398e-06, "loss": 0.9145, "step": 15811 }, { "epoch": 0.16, "learning_rate": 9.63683390024996e-06, "loss": 0.8545, "step": 15812 }, { "epoch": 0.16, "learning_rate": 9.63575673216589e-06, "loss": 0.9112, "step": 15813 }, { "epoch": 0.16, "learning_rate": 9.634679568313693e-06, "loss": 0.9044, "step": 15814 }, { "epoch": 0.16, "learning_rate": 9.633602408705895e-06, "loss": 0.8948, "step": 15815 }, { "epoch": 0.16, "learning_rate": 9.632525253355006e-06, "loss": 0.8845, "step": 15816 }, { "epoch": 0.16, "learning_rate": 9.631448102273543e-06, "loss": 0.8899, "step": 15817 }, { "epoch": 0.16, "learning_rate": 9.630370955474018e-06, "loss": 0.8889, "step": 15818 }, { "epoch": 0.16, "learning_rate": 9.629293812968947e-06, "loss": 0.8938, "step": 15819 }, { "epoch": 0.16, "learning_rate": 9.628216674770841e-06, "loss": 0.8736, "step": 15820 }, { "epoch": 0.16, "learning_rate": 9.627139540892223e-06, "loss": 0.9216, "step": 15821 }, { "epoch": 0.16, "learning_rate": 9.6260624113456e-06, "loss": 0.8788, "step": 15822 }, { "epoch": 0.16, "learning_rate": 9.624985286143485e-06, "loss": 0.8681, "step": 15823 }, { "epoch": 0.16, "learning_rate": 9.623908165298401e-06, "loss": 0.8709, "step": 15824 }, { "epoch": 0.16, "learning_rate": 9.622831048822854e-06, "loss": 0.9122, "step": 15825 }, { "epoch": 0.16, "learning_rate": 9.621753936729359e-06, "loss": 0.8526, "step": 15826 }, { "epoch": 0.16, "learning_rate": 9.620676829030435e-06, "loss": 0.8793, "step": 15827 }, { "epoch": 0.16, "learning_rate": 9.619599725738597e-06, "loss": 0.848, "step": 15828 }, { "epoch": 0.16, "learning_rate": 9.618522626866351e-06, "loss": 0.899, "step": 15829 }, { "epoch": 0.16, "learning_rate": 9.617445532426217e-06, "loss": 0.8851, "step": 15830 }, { "epoch": 0.16, "learning_rate": 9.616368442430712e-06, "loss": 0.8758, "step": 15831 }, { "epoch": 0.16, "learning_rate": 9.61529135689234e-06, "loss": 0.9219, "step": 15832 }, { "epoch": 0.16, "learning_rate": 9.614214275823621e-06, "loss": 0.9004, "step": 15833 }, { "epoch": 0.16, "learning_rate": 9.61313719923707e-06, "loss": 0.8927, "step": 15834 }, { "epoch": 0.16, "learning_rate": 9.612060127145202e-06, "loss": 0.9226, "step": 15835 }, { "epoch": 0.16, "learning_rate": 9.610983059560523e-06, "loss": 0.8888, "step": 15836 }, { "epoch": 0.16, "learning_rate": 9.609905996495554e-06, "loss": 0.8379, "step": 15837 }, { "epoch": 0.16, "learning_rate": 9.608828937962807e-06, "loss": 0.9041, "step": 15838 }, { "epoch": 0.16, "learning_rate": 9.607751883974794e-06, "loss": 0.8846, "step": 15839 }, { "epoch": 0.16, "learning_rate": 9.606674834544027e-06, "loss": 0.9039, "step": 15840 }, { "epoch": 0.16, "learning_rate": 9.605597789683025e-06, "loss": 0.9142, "step": 15841 }, { "epoch": 0.16, "learning_rate": 9.604520749404298e-06, "loss": 0.8998, "step": 15842 }, { "epoch": 0.16, "learning_rate": 9.603443713720357e-06, "loss": 0.891, "step": 15843 }, { "epoch": 0.16, "learning_rate": 9.602366682643719e-06, "loss": 0.9189, "step": 15844 }, { "epoch": 0.16, "learning_rate": 9.601289656186899e-06, "loss": 0.8847, "step": 15845 }, { "epoch": 0.16, "learning_rate": 9.600212634362403e-06, "loss": 0.8742, "step": 15846 }, { "epoch": 0.16, "learning_rate": 9.599135617182749e-06, "loss": 0.9148, "step": 15847 }, { "epoch": 0.16, "learning_rate": 9.598058604660448e-06, "loss": 0.8976, "step": 15848 }, { "epoch": 0.16, "learning_rate": 9.596981596808021e-06, "loss": 0.9128, "step": 15849 }, { "epoch": 0.16, "learning_rate": 9.59590459363797e-06, "loss": 0.9017, "step": 15850 }, { "epoch": 0.16, "learning_rate": 9.59482759516281e-06, "loss": 0.8798, "step": 15851 }, { "epoch": 0.16, "learning_rate": 9.593750601395064e-06, "loss": 0.8821, "step": 15852 }, { "epoch": 0.16, "learning_rate": 9.592673612347231e-06, "loss": 0.8582, "step": 15853 }, { "epoch": 0.16, "learning_rate": 9.591596628031829e-06, "loss": 0.8446, "step": 15854 }, { "epoch": 0.16, "learning_rate": 9.590519648461373e-06, "loss": 0.9062, "step": 15855 }, { "epoch": 0.16, "learning_rate": 9.589442673648376e-06, "loss": 0.8989, "step": 15856 }, { "epoch": 0.16, "learning_rate": 9.588365703605347e-06, "loss": 0.8742, "step": 15857 }, { "epoch": 0.16, "learning_rate": 9.5872887383448e-06, "loss": 0.9002, "step": 15858 }, { "epoch": 0.16, "learning_rate": 9.586211777879251e-06, "loss": 0.8763, "step": 15859 }, { "epoch": 0.16, "learning_rate": 9.585134822221206e-06, "loss": 0.8829, "step": 15860 }, { "epoch": 0.16, "learning_rate": 9.58405787138318e-06, "loss": 0.9114, "step": 15861 }, { "epoch": 0.16, "learning_rate": 9.58298092537769e-06, "loss": 0.9139, "step": 15862 }, { "epoch": 0.16, "learning_rate": 9.581903984217242e-06, "loss": 0.8858, "step": 15863 }, { "epoch": 0.16, "learning_rate": 9.58082704791435e-06, "loss": 0.8662, "step": 15864 }, { "epoch": 0.16, "learning_rate": 9.579750116481527e-06, "loss": 0.8889, "step": 15865 }, { "epoch": 0.16, "learning_rate": 9.578673189931283e-06, "loss": 0.8827, "step": 15866 }, { "epoch": 0.16, "learning_rate": 9.577596268276134e-06, "loss": 0.8852, "step": 15867 }, { "epoch": 0.16, "learning_rate": 9.57651935152859e-06, "loss": 0.867, "step": 15868 }, { "epoch": 0.16, "learning_rate": 9.57544243970116e-06, "loss": 0.8892, "step": 15869 }, { "epoch": 0.16, "learning_rate": 9.574365532806362e-06, "loss": 0.9038, "step": 15870 }, { "epoch": 0.16, "learning_rate": 9.573288630856703e-06, "loss": 0.9322, "step": 15871 }, { "epoch": 0.16, "learning_rate": 9.572211733864692e-06, "loss": 0.8989, "step": 15872 }, { "epoch": 0.16, "learning_rate": 9.571134841842846e-06, "loss": 0.8803, "step": 15873 }, { "epoch": 0.16, "learning_rate": 9.57005795480368e-06, "loss": 0.8704, "step": 15874 }, { "epoch": 0.16, "learning_rate": 9.568981072759697e-06, "loss": 0.868, "step": 15875 }, { "epoch": 0.16, "learning_rate": 9.567904195723411e-06, "loss": 0.8897, "step": 15876 }, { "epoch": 0.16, "learning_rate": 9.566827323707343e-06, "loss": 0.885, "step": 15877 }, { "epoch": 0.16, "learning_rate": 9.565750456723988e-06, "loss": 0.8478, "step": 15878 }, { "epoch": 0.16, "learning_rate": 9.564673594785868e-06, "loss": 0.9086, "step": 15879 }, { "epoch": 0.16, "learning_rate": 9.56359673790549e-06, "loss": 0.8902, "step": 15880 }, { "epoch": 0.16, "learning_rate": 9.562519886095372e-06, "loss": 0.894, "step": 15881 }, { "epoch": 0.16, "learning_rate": 9.561443039368017e-06, "loss": 0.8888, "step": 15882 }, { "epoch": 0.16, "learning_rate": 9.560366197735938e-06, "loss": 0.897, "step": 15883 }, { "epoch": 0.16, "learning_rate": 9.559289361211652e-06, "loss": 0.9026, "step": 15884 }, { "epoch": 0.16, "learning_rate": 9.558212529807662e-06, "loss": 0.9194, "step": 15885 }, { "epoch": 0.16, "learning_rate": 9.557135703536482e-06, "loss": 0.8687, "step": 15886 }, { "epoch": 0.16, "learning_rate": 9.556058882410623e-06, "loss": 0.8822, "step": 15887 }, { "epoch": 0.16, "learning_rate": 9.554982066442601e-06, "loss": 0.8836, "step": 15888 }, { "epoch": 0.16, "learning_rate": 9.553905255644917e-06, "loss": 0.8783, "step": 15889 }, { "epoch": 0.16, "learning_rate": 9.552828450030087e-06, "loss": 0.8825, "step": 15890 }, { "epoch": 0.16, "learning_rate": 9.551751649610623e-06, "loss": 0.8821, "step": 15891 }, { "epoch": 0.16, "learning_rate": 9.550674854399032e-06, "loss": 0.9172, "step": 15892 }, { "epoch": 0.16, "learning_rate": 9.549598064407825e-06, "loss": 0.896, "step": 15893 }, { "epoch": 0.16, "learning_rate": 9.548521279649512e-06, "loss": 0.8904, "step": 15894 }, { "epoch": 0.16, "learning_rate": 9.547444500136614e-06, "loss": 0.9114, "step": 15895 }, { "epoch": 0.16, "learning_rate": 9.546367725881625e-06, "loss": 0.9298, "step": 15896 }, { "epoch": 0.16, "learning_rate": 9.545290956897062e-06, "loss": 0.8926, "step": 15897 }, { "epoch": 0.16, "learning_rate": 9.544214193195441e-06, "loss": 0.8924, "step": 15898 }, { "epoch": 0.16, "learning_rate": 9.543137434789262e-06, "loss": 0.8692, "step": 15899 }, { "epoch": 0.16, "learning_rate": 9.54206068169104e-06, "loss": 0.8922, "step": 15900 }, { "epoch": 0.16, "learning_rate": 9.540983933913287e-06, "loss": 0.9046, "step": 15901 }, { "epoch": 0.16, "learning_rate": 9.539907191468512e-06, "loss": 0.9037, "step": 15902 }, { "epoch": 0.16, "learning_rate": 9.538830454369223e-06, "loss": 0.8817, "step": 15903 }, { "epoch": 0.16, "learning_rate": 9.537753722627931e-06, "loss": 0.9221, "step": 15904 }, { "epoch": 0.16, "learning_rate": 9.536676996257146e-06, "loss": 0.8438, "step": 15905 }, { "epoch": 0.16, "learning_rate": 9.535600275269376e-06, "loss": 0.9024, "step": 15906 }, { "epoch": 0.16, "learning_rate": 9.534523559677131e-06, "loss": 0.9273, "step": 15907 }, { "epoch": 0.16, "learning_rate": 9.533446849492922e-06, "loss": 0.8527, "step": 15908 }, { "epoch": 0.16, "learning_rate": 9.532370144729259e-06, "loss": 0.8647, "step": 15909 }, { "epoch": 0.16, "learning_rate": 9.53129344539865e-06, "loss": 0.8843, "step": 15910 }, { "epoch": 0.16, "learning_rate": 9.530216751513605e-06, "loss": 0.8495, "step": 15911 }, { "epoch": 0.16, "learning_rate": 9.529140063086632e-06, "loss": 0.9486, "step": 15912 }, { "epoch": 0.16, "learning_rate": 9.528063380130243e-06, "loss": 0.9357, "step": 15913 }, { "epoch": 0.16, "learning_rate": 9.526986702656944e-06, "loss": 0.8667, "step": 15914 }, { "epoch": 0.16, "learning_rate": 9.525910030679244e-06, "loss": 0.885, "step": 15915 }, { "epoch": 0.16, "learning_rate": 9.52483336420966e-06, "loss": 0.8955, "step": 15916 }, { "epoch": 0.16, "learning_rate": 9.523756703260692e-06, "loss": 0.9117, "step": 15917 }, { "epoch": 0.16, "learning_rate": 9.522680047844847e-06, "loss": 0.8898, "step": 15918 }, { "epoch": 0.16, "learning_rate": 9.521603397974643e-06, "loss": 0.9055, "step": 15919 }, { "epoch": 0.16, "learning_rate": 9.520526753662587e-06, "loss": 0.9164, "step": 15920 }, { "epoch": 0.16, "learning_rate": 9.51945011492118e-06, "loss": 0.8965, "step": 15921 }, { "epoch": 0.16, "learning_rate": 9.518373481762938e-06, "loss": 0.8471, "step": 15922 }, { "epoch": 0.16, "learning_rate": 9.517296854200371e-06, "loss": 0.9326, "step": 15923 }, { "epoch": 0.16, "learning_rate": 9.516220232245979e-06, "loss": 0.866, "step": 15924 }, { "epoch": 0.16, "learning_rate": 9.515143615912277e-06, "loss": 0.863, "step": 15925 }, { "epoch": 0.16, "learning_rate": 9.51406700521177e-06, "loss": 0.8763, "step": 15926 }, { "epoch": 0.16, "learning_rate": 9.512990400156973e-06, "loss": 0.8469, "step": 15927 }, { "epoch": 0.16, "learning_rate": 9.511913800760388e-06, "loss": 0.8752, "step": 15928 }, { "epoch": 0.16, "learning_rate": 9.510837207034524e-06, "loss": 0.8587, "step": 15929 }, { "epoch": 0.16, "learning_rate": 9.509760618991892e-06, "loss": 0.907, "step": 15930 }, { "epoch": 0.16, "learning_rate": 9.508684036644997e-06, "loss": 0.8949, "step": 15931 }, { "epoch": 0.16, "learning_rate": 9.507607460006347e-06, "loss": 0.9082, "step": 15932 }, { "epoch": 0.16, "learning_rate": 9.506530889088452e-06, "loss": 0.8425, "step": 15933 }, { "epoch": 0.16, "learning_rate": 9.505454323903822e-06, "loss": 0.8752, "step": 15934 }, { "epoch": 0.16, "learning_rate": 9.504377764464959e-06, "loss": 0.8467, "step": 15935 }, { "epoch": 0.16, "learning_rate": 9.503301210784373e-06, "loss": 0.9012, "step": 15936 }, { "epoch": 0.16, "learning_rate": 9.502224662874575e-06, "loss": 0.8572, "step": 15937 }, { "epoch": 0.16, "learning_rate": 9.501148120748068e-06, "loss": 0.8107, "step": 15938 }, { "epoch": 0.16, "learning_rate": 9.500071584417361e-06, "loss": 0.904, "step": 15939 }, { "epoch": 0.16, "learning_rate": 9.498995053894961e-06, "loss": 0.8411, "step": 15940 }, { "epoch": 0.16, "learning_rate": 9.497918529193383e-06, "loss": 0.8764, "step": 15941 }, { "epoch": 0.16, "learning_rate": 9.496842010325123e-06, "loss": 0.8925, "step": 15942 }, { "epoch": 0.16, "learning_rate": 9.495765497302692e-06, "loss": 0.8876, "step": 15943 }, { "epoch": 0.16, "learning_rate": 9.494688990138603e-06, "loss": 0.8877, "step": 15944 }, { "epoch": 0.16, "learning_rate": 9.493612488845355e-06, "loss": 0.8794, "step": 15945 }, { "epoch": 0.16, "learning_rate": 9.492535993435459e-06, "loss": 0.9123, "step": 15946 }, { "epoch": 0.16, "learning_rate": 9.49145950392142e-06, "loss": 0.9015, "step": 15947 }, { "epoch": 0.16, "learning_rate": 9.490383020315751e-06, "loss": 0.8768, "step": 15948 }, { "epoch": 0.16, "learning_rate": 9.489306542630953e-06, "loss": 0.8973, "step": 15949 }, { "epoch": 0.16, "learning_rate": 9.488230070879533e-06, "loss": 0.9223, "step": 15950 }, { "epoch": 0.16, "learning_rate": 9.487153605074002e-06, "loss": 0.9031, "step": 15951 }, { "epoch": 0.16, "learning_rate": 9.486077145226861e-06, "loss": 0.9055, "step": 15952 }, { "epoch": 0.16, "learning_rate": 9.485000691350619e-06, "loss": 0.9535, "step": 15953 }, { "epoch": 0.16, "learning_rate": 9.483924243457786e-06, "loss": 0.8551, "step": 15954 }, { "epoch": 0.16, "learning_rate": 9.482847801560863e-06, "loss": 0.9105, "step": 15955 }, { "epoch": 0.16, "learning_rate": 9.48177136567236e-06, "loss": 0.8864, "step": 15956 }, { "epoch": 0.16, "learning_rate": 9.480694935804781e-06, "loss": 0.9165, "step": 15957 }, { "epoch": 0.17, "learning_rate": 9.479618511970634e-06, "loss": 0.8708, "step": 15958 }, { "epoch": 0.17, "learning_rate": 9.478542094182428e-06, "loss": 0.8603, "step": 15959 }, { "epoch": 0.17, "learning_rate": 9.47746568245266e-06, "loss": 0.8819, "step": 15960 }, { "epoch": 0.17, "learning_rate": 9.476389276793843e-06, "loss": 0.8881, "step": 15961 }, { "epoch": 0.17, "learning_rate": 9.475312877218486e-06, "loss": 0.8512, "step": 15962 }, { "epoch": 0.17, "learning_rate": 9.474236483739088e-06, "loss": 0.8849, "step": 15963 }, { "epoch": 0.17, "learning_rate": 9.473160096368153e-06, "loss": 0.9144, "step": 15964 }, { "epoch": 0.17, "learning_rate": 9.472083715118196e-06, "loss": 0.9114, "step": 15965 }, { "epoch": 0.17, "learning_rate": 9.471007340001722e-06, "loss": 0.8801, "step": 15966 }, { "epoch": 0.17, "learning_rate": 9.469930971031225e-06, "loss": 0.908, "step": 15967 }, { "epoch": 0.17, "learning_rate": 9.46885460821922e-06, "loss": 0.8759, "step": 15968 }, { "epoch": 0.17, "learning_rate": 9.467778251578217e-06, "loss": 0.9062, "step": 15969 }, { "epoch": 0.17, "learning_rate": 9.466701901120707e-06, "loss": 0.871, "step": 15970 }, { "epoch": 0.17, "learning_rate": 9.465625556859207e-06, "loss": 0.9141, "step": 15971 }, { "epoch": 0.17, "learning_rate": 9.464549218806218e-06, "loss": 0.8697, "step": 15972 }, { "epoch": 0.17, "learning_rate": 9.463472886974247e-06, "loss": 0.9053, "step": 15973 }, { "epoch": 0.17, "learning_rate": 9.462396561375797e-06, "loss": 0.8599, "step": 15974 }, { "epoch": 0.17, "learning_rate": 9.461320242023375e-06, "loss": 0.899, "step": 15975 }, { "epoch": 0.17, "learning_rate": 9.460243928929485e-06, "loss": 0.8597, "step": 15976 }, { "epoch": 0.17, "learning_rate": 9.45916762210663e-06, "loss": 0.9188, "step": 15977 }, { "epoch": 0.17, "learning_rate": 9.45809132156732e-06, "loss": 0.8604, "step": 15978 }, { "epoch": 0.17, "learning_rate": 9.457015027324052e-06, "loss": 0.898, "step": 15979 }, { "epoch": 0.17, "learning_rate": 9.455938739389342e-06, "loss": 0.885, "step": 15980 }, { "epoch": 0.17, "learning_rate": 9.454862457775683e-06, "loss": 0.8791, "step": 15981 }, { "epoch": 0.17, "learning_rate": 9.453786182495584e-06, "loss": 0.8732, "step": 15982 }, { "epoch": 0.17, "learning_rate": 9.452709913561552e-06, "loss": 0.8822, "step": 15983 }, { "epoch": 0.17, "learning_rate": 9.451633650986087e-06, "loss": 0.8664, "step": 15984 }, { "epoch": 0.17, "learning_rate": 9.450557394781697e-06, "loss": 0.8476, "step": 15985 }, { "epoch": 0.17, "learning_rate": 9.449481144960881e-06, "loss": 0.8911, "step": 15986 }, { "epoch": 0.17, "learning_rate": 9.448404901536155e-06, "loss": 0.8954, "step": 15987 }, { "epoch": 0.17, "learning_rate": 9.447328664520009e-06, "loss": 0.9004, "step": 15988 }, { "epoch": 0.17, "learning_rate": 9.446252433924953e-06, "loss": 0.8511, "step": 15989 }, { "epoch": 0.17, "learning_rate": 9.445176209763496e-06, "loss": 0.9148, "step": 15990 }, { "epoch": 0.17, "learning_rate": 9.444099992048133e-06, "loss": 0.8533, "step": 15991 }, { "epoch": 0.17, "learning_rate": 9.44302378079137e-06, "loss": 0.8715, "step": 15992 }, { "epoch": 0.17, "learning_rate": 9.441947576005715e-06, "loss": 0.8637, "step": 15993 }, { "epoch": 0.17, "learning_rate": 9.44087137770367e-06, "loss": 0.8476, "step": 15994 }, { "epoch": 0.17, "learning_rate": 9.439795185897737e-06, "loss": 0.8822, "step": 15995 }, { "epoch": 0.17, "learning_rate": 9.43871900060042e-06, "loss": 0.8607, "step": 15996 }, { "epoch": 0.17, "learning_rate": 9.437642821824223e-06, "loss": 0.8729, "step": 15997 }, { "epoch": 0.17, "learning_rate": 9.436566649581648e-06, "loss": 0.8976, "step": 15998 }, { "epoch": 0.17, "learning_rate": 9.435490483885199e-06, "loss": 0.8821, "step": 15999 }, { "epoch": 0.17, "learning_rate": 9.43441432474738e-06, "loss": 0.875, "step": 16000 }, { "epoch": 0.17, "learning_rate": 9.433338172180694e-06, "loss": 0.8961, "step": 16001 }, { "epoch": 0.17, "learning_rate": 9.432262026197644e-06, "loss": 0.8753, "step": 16002 }, { "epoch": 0.17, "learning_rate": 9.43118588681073e-06, "loss": 0.8713, "step": 16003 }, { "epoch": 0.17, "learning_rate": 9.430109754032458e-06, "loss": 0.8738, "step": 16004 }, { "epoch": 0.17, "learning_rate": 9.429033627875333e-06, "loss": 0.8881, "step": 16005 }, { "epoch": 0.17, "learning_rate": 9.427957508351852e-06, "loss": 0.9137, "step": 16006 }, { "epoch": 0.17, "learning_rate": 9.42688139547452e-06, "loss": 0.9032, "step": 16007 }, { "epoch": 0.17, "learning_rate": 9.425805289255847e-06, "loss": 0.8968, "step": 16008 }, { "epoch": 0.17, "learning_rate": 9.424729189708323e-06, "loss": 0.8984, "step": 16009 }, { "epoch": 0.17, "learning_rate": 9.423653096844452e-06, "loss": 0.8775, "step": 16010 }, { "epoch": 0.17, "learning_rate": 9.422577010676746e-06, "loss": 0.8808, "step": 16011 }, { "epoch": 0.17, "learning_rate": 9.421500931217705e-06, "loss": 0.8724, "step": 16012 }, { "epoch": 0.17, "learning_rate": 9.420424858479821e-06, "loss": 0.8764, "step": 16013 }, { "epoch": 0.17, "learning_rate": 9.419348792475605e-06, "loss": 0.9008, "step": 16014 }, { "epoch": 0.17, "learning_rate": 9.418272733217561e-06, "loss": 0.8726, "step": 16015 }, { "epoch": 0.17, "learning_rate": 9.417196680718183e-06, "loss": 0.8678, "step": 16016 }, { "epoch": 0.17, "learning_rate": 9.416120634989976e-06, "loss": 0.8926, "step": 16017 }, { "epoch": 0.17, "learning_rate": 9.415044596045446e-06, "loss": 0.8554, "step": 16018 }, { "epoch": 0.17, "learning_rate": 9.413968563897091e-06, "loss": 0.8616, "step": 16019 }, { "epoch": 0.17, "learning_rate": 9.412892538557411e-06, "loss": 0.9314, "step": 16020 }, { "epoch": 0.17, "learning_rate": 9.411816520038911e-06, "loss": 0.911, "step": 16021 }, { "epoch": 0.17, "learning_rate": 9.410740508354093e-06, "loss": 0.8908, "step": 16022 }, { "epoch": 0.17, "learning_rate": 9.409664503515453e-06, "loss": 0.8405, "step": 16023 }, { "epoch": 0.17, "learning_rate": 9.408588505535498e-06, "loss": 0.9387, "step": 16024 }, { "epoch": 0.17, "learning_rate": 9.407512514426725e-06, "loss": 0.8568, "step": 16025 }, { "epoch": 0.17, "learning_rate": 9.406436530201641e-06, "loss": 0.8714, "step": 16026 }, { "epoch": 0.17, "learning_rate": 9.40536055287274e-06, "loss": 0.9027, "step": 16027 }, { "epoch": 0.17, "learning_rate": 9.404284582452527e-06, "loss": 0.8618, "step": 16028 }, { "epoch": 0.17, "learning_rate": 9.403208618953504e-06, "loss": 0.8715, "step": 16029 }, { "epoch": 0.17, "learning_rate": 9.402132662388168e-06, "loss": 0.8927, "step": 16030 }, { "epoch": 0.17, "learning_rate": 9.401056712769021e-06, "loss": 0.8686, "step": 16031 }, { "epoch": 0.17, "learning_rate": 9.399980770108566e-06, "loss": 0.8707, "step": 16032 }, { "epoch": 0.17, "learning_rate": 9.398904834419305e-06, "loss": 0.918, "step": 16033 }, { "epoch": 0.17, "learning_rate": 9.397828905713731e-06, "loss": 0.866, "step": 16034 }, { "epoch": 0.17, "learning_rate": 9.396752984004351e-06, "loss": 0.8578, "step": 16035 }, { "epoch": 0.17, "learning_rate": 9.395677069303667e-06, "loss": 0.8441, "step": 16036 }, { "epoch": 0.17, "learning_rate": 9.39460116162417e-06, "loss": 0.9, "step": 16037 }, { "epoch": 0.17, "learning_rate": 9.393525260978368e-06, "loss": 0.8637, "step": 16038 }, { "epoch": 0.17, "learning_rate": 9.39244936737876e-06, "loss": 0.9133, "step": 16039 }, { "epoch": 0.17, "learning_rate": 9.391373480837847e-06, "loss": 0.8497, "step": 16040 }, { "epoch": 0.17, "learning_rate": 9.390297601368123e-06, "loss": 0.8797, "step": 16041 }, { "epoch": 0.17, "learning_rate": 9.389221728982095e-06, "loss": 0.8424, "step": 16042 }, { "epoch": 0.17, "learning_rate": 9.388145863692258e-06, "loss": 0.8847, "step": 16043 }, { "epoch": 0.17, "learning_rate": 9.387070005511115e-06, "loss": 0.867, "step": 16044 }, { "epoch": 0.17, "learning_rate": 9.38599415445116e-06, "loss": 0.9209, "step": 16045 }, { "epoch": 0.17, "learning_rate": 9.3849183105249e-06, "loss": 0.9176, "step": 16046 }, { "epoch": 0.17, "learning_rate": 9.383842473744833e-06, "loss": 0.9176, "step": 16047 }, { "epoch": 0.17, "learning_rate": 9.382766644123453e-06, "loss": 0.8705, "step": 16048 }, { "epoch": 0.17, "learning_rate": 9.381690821673263e-06, "loss": 0.854, "step": 16049 }, { "epoch": 0.17, "learning_rate": 9.380615006406761e-06, "loss": 0.8842, "step": 16050 }, { "epoch": 0.17, "learning_rate": 9.379539198336453e-06, "loss": 0.9162, "step": 16051 }, { "epoch": 0.17, "learning_rate": 9.378463397474827e-06, "loss": 0.8581, "step": 16052 }, { "epoch": 0.17, "learning_rate": 9.377387603834384e-06, "loss": 0.899, "step": 16053 }, { "epoch": 0.17, "learning_rate": 9.376311817427635e-06, "loss": 0.8834, "step": 16054 }, { "epoch": 0.17, "learning_rate": 9.375236038267065e-06, "loss": 0.8767, "step": 16055 }, { "epoch": 0.17, "learning_rate": 9.374160266365172e-06, "loss": 0.8725, "step": 16056 }, { "epoch": 0.17, "learning_rate": 9.373084501734465e-06, "loss": 0.8889, "step": 16057 }, { "epoch": 0.17, "learning_rate": 9.372008744387441e-06, "loss": 0.8724, "step": 16058 }, { "epoch": 0.17, "learning_rate": 9.37093299433659e-06, "loss": 0.904, "step": 16059 }, { "epoch": 0.17, "learning_rate": 9.369857251594415e-06, "loss": 0.9182, "step": 16060 }, { "epoch": 0.17, "learning_rate": 9.368781516173417e-06, "loss": 0.905, "step": 16061 }, { "epoch": 0.17, "learning_rate": 9.367705788086092e-06, "loss": 0.8939, "step": 16062 }, { "epoch": 0.17, "learning_rate": 9.366630067344937e-06, "loss": 0.8999, "step": 16063 }, { "epoch": 0.17, "learning_rate": 9.36555435396245e-06, "loss": 0.897, "step": 16064 }, { "epoch": 0.17, "learning_rate": 9.364478647951132e-06, "loss": 0.9184, "step": 16065 }, { "epoch": 0.17, "learning_rate": 9.363402949323478e-06, "loss": 0.8735, "step": 16066 }, { "epoch": 0.17, "learning_rate": 9.362327258091985e-06, "loss": 0.9022, "step": 16067 }, { "epoch": 0.17, "learning_rate": 9.361251574269155e-06, "loss": 0.8723, "step": 16068 }, { "epoch": 0.17, "learning_rate": 9.360175897867481e-06, "loss": 0.9078, "step": 16069 }, { "epoch": 0.17, "learning_rate": 9.359100228899463e-06, "loss": 0.8428, "step": 16070 }, { "epoch": 0.17, "learning_rate": 9.358024567377598e-06, "loss": 0.891, "step": 16071 }, { "epoch": 0.17, "learning_rate": 9.356948913314383e-06, "loss": 0.898, "step": 16072 }, { "epoch": 0.17, "learning_rate": 9.355873266722314e-06, "loss": 0.8903, "step": 16073 }, { "epoch": 0.17, "learning_rate": 9.35479762761389e-06, "loss": 0.8492, "step": 16074 }, { "epoch": 0.17, "learning_rate": 9.35372199600161e-06, "loss": 0.915, "step": 16075 }, { "epoch": 0.17, "learning_rate": 9.352646371897966e-06, "loss": 0.8785, "step": 16076 }, { "epoch": 0.17, "learning_rate": 9.351570755315454e-06, "loss": 0.8918, "step": 16077 }, { "epoch": 0.17, "learning_rate": 9.350495146266577e-06, "loss": 0.8617, "step": 16078 }, { "epoch": 0.17, "learning_rate": 9.349419544763835e-06, "loss": 0.9205, "step": 16079 }, { "epoch": 0.17, "learning_rate": 9.34834395081971e-06, "loss": 0.885, "step": 16080 }, { "epoch": 0.17, "learning_rate": 9.34726836444671e-06, "loss": 0.9151, "step": 16081 }, { "epoch": 0.17, "learning_rate": 9.346192785657335e-06, "loss": 0.8717, "step": 16082 }, { "epoch": 0.17, "learning_rate": 9.345117214464066e-06, "loss": 0.8966, "step": 16083 }, { "epoch": 0.17, "learning_rate": 9.344041650879412e-06, "loss": 0.9171, "step": 16084 }, { "epoch": 0.17, "learning_rate": 9.342966094915865e-06, "loss": 0.8712, "step": 16085 }, { "epoch": 0.17, "learning_rate": 9.341890546585923e-06, "loss": 0.8808, "step": 16086 }, { "epoch": 0.17, "learning_rate": 9.34081500590208e-06, "loss": 0.9033, "step": 16087 }, { "epoch": 0.17, "learning_rate": 9.33973947287683e-06, "loss": 0.919, "step": 16088 }, { "epoch": 0.17, "learning_rate": 9.338663947522676e-06, "loss": 0.878, "step": 16089 }, { "epoch": 0.17, "learning_rate": 9.337588429852106e-06, "loss": 0.9256, "step": 16090 }, { "epoch": 0.17, "learning_rate": 9.336512919877619e-06, "loss": 0.8853, "step": 16091 }, { "epoch": 0.17, "learning_rate": 9.335437417611712e-06, "loss": 0.9014, "step": 16092 }, { "epoch": 0.17, "learning_rate": 9.334361923066878e-06, "loss": 0.8673, "step": 16093 }, { "epoch": 0.17, "learning_rate": 9.333286436255613e-06, "loss": 0.8773, "step": 16094 }, { "epoch": 0.17, "learning_rate": 9.332210957190413e-06, "loss": 0.8431, "step": 16095 }, { "epoch": 0.17, "learning_rate": 9.331135485883769e-06, "loss": 0.9155, "step": 16096 }, { "epoch": 0.17, "learning_rate": 9.330060022348188e-06, "loss": 0.8983, "step": 16097 }, { "epoch": 0.17, "learning_rate": 9.328984566596152e-06, "loss": 0.9124, "step": 16098 }, { "epoch": 0.17, "learning_rate": 9.327909118640158e-06, "loss": 0.8666, "step": 16099 }, { "epoch": 0.17, "learning_rate": 9.326833678492712e-06, "loss": 0.8911, "step": 16100 }, { "epoch": 0.17, "learning_rate": 9.325758246166296e-06, "loss": 0.9026, "step": 16101 }, { "epoch": 0.17, "learning_rate": 9.324682821673406e-06, "loss": 0.8608, "step": 16102 }, { "epoch": 0.17, "learning_rate": 9.323607405026541e-06, "loss": 0.895, "step": 16103 }, { "epoch": 0.17, "learning_rate": 9.3225319962382e-06, "loss": 0.8927, "step": 16104 }, { "epoch": 0.17, "learning_rate": 9.321456595320866e-06, "loss": 0.9332, "step": 16105 }, { "epoch": 0.17, "learning_rate": 9.320381202287039e-06, "loss": 0.8555, "step": 16106 }, { "epoch": 0.17, "learning_rate": 9.319305817149218e-06, "loss": 0.877, "step": 16107 }, { "epoch": 0.17, "learning_rate": 9.318230439919889e-06, "loss": 0.8663, "step": 16108 }, { "epoch": 0.17, "learning_rate": 9.31715507061155e-06, "loss": 0.878, "step": 16109 }, { "epoch": 0.17, "learning_rate": 9.316079709236692e-06, "loss": 0.883, "step": 16110 }, { "epoch": 0.17, "learning_rate": 9.315004355807815e-06, "loss": 0.8845, "step": 16111 }, { "epoch": 0.17, "learning_rate": 9.313929010337405e-06, "loss": 0.8898, "step": 16112 }, { "epoch": 0.17, "learning_rate": 9.312853672837962e-06, "loss": 0.8932, "step": 16113 }, { "epoch": 0.17, "learning_rate": 9.311778343321979e-06, "loss": 0.8589, "step": 16114 }, { "epoch": 0.17, "learning_rate": 9.310703021801944e-06, "loss": 0.8772, "step": 16115 }, { "epoch": 0.17, "learning_rate": 9.309627708290357e-06, "loss": 0.8637, "step": 16116 }, { "epoch": 0.17, "learning_rate": 9.308552402799706e-06, "loss": 0.9139, "step": 16117 }, { "epoch": 0.17, "learning_rate": 9.307477105342488e-06, "loss": 0.8472, "step": 16118 }, { "epoch": 0.17, "learning_rate": 9.306401815931192e-06, "loss": 0.8955, "step": 16119 }, { "epoch": 0.17, "learning_rate": 9.305326534578316e-06, "loss": 0.8965, "step": 16120 }, { "epoch": 0.17, "learning_rate": 9.304251261296352e-06, "loss": 0.8962, "step": 16121 }, { "epoch": 0.17, "learning_rate": 9.303175996097789e-06, "loss": 0.8359, "step": 16122 }, { "epoch": 0.17, "learning_rate": 9.30210073899512e-06, "loss": 0.8474, "step": 16123 }, { "epoch": 0.17, "learning_rate": 9.301025490000843e-06, "loss": 0.8529, "step": 16124 }, { "epoch": 0.17, "learning_rate": 9.299950249127449e-06, "loss": 0.8245, "step": 16125 }, { "epoch": 0.17, "learning_rate": 9.298875016387424e-06, "loss": 0.8913, "step": 16126 }, { "epoch": 0.17, "learning_rate": 9.297799791793267e-06, "loss": 0.9042, "step": 16127 }, { "epoch": 0.17, "learning_rate": 9.296724575357473e-06, "loss": 0.8565, "step": 16128 }, { "epoch": 0.17, "learning_rate": 9.295649367092523e-06, "loss": 0.8908, "step": 16129 }, { "epoch": 0.17, "learning_rate": 9.294574167010917e-06, "loss": 0.8629, "step": 16130 }, { "epoch": 0.17, "learning_rate": 9.293498975125145e-06, "loss": 0.9, "step": 16131 }, { "epoch": 0.17, "learning_rate": 9.292423791447703e-06, "loss": 0.8395, "step": 16132 }, { "epoch": 0.17, "learning_rate": 9.291348615991077e-06, "loss": 0.9027, "step": 16133 }, { "epoch": 0.17, "learning_rate": 9.29027344876776e-06, "loss": 0.8855, "step": 16134 }, { "epoch": 0.17, "learning_rate": 9.289198289790248e-06, "loss": 0.9015, "step": 16135 }, { "epoch": 0.17, "learning_rate": 9.288123139071026e-06, "loss": 0.8736, "step": 16136 }, { "epoch": 0.17, "learning_rate": 9.287047996622589e-06, "loss": 0.8957, "step": 16137 }, { "epoch": 0.17, "learning_rate": 9.285972862457426e-06, "loss": 0.8859, "step": 16138 }, { "epoch": 0.17, "learning_rate": 9.284897736588035e-06, "loss": 0.8932, "step": 16139 }, { "epoch": 0.17, "learning_rate": 9.283822619026898e-06, "loss": 0.8422, "step": 16140 }, { "epoch": 0.17, "learning_rate": 9.282747509786509e-06, "loss": 0.902, "step": 16141 }, { "epoch": 0.17, "learning_rate": 9.28167240887936e-06, "loss": 0.9093, "step": 16142 }, { "epoch": 0.17, "learning_rate": 9.280597316317948e-06, "loss": 0.8975, "step": 16143 }, { "epoch": 0.17, "learning_rate": 9.279522232114751e-06, "loss": 0.8609, "step": 16144 }, { "epoch": 0.17, "learning_rate": 9.278447156282266e-06, "loss": 0.8781, "step": 16145 }, { "epoch": 0.17, "learning_rate": 9.27737208883299e-06, "loss": 0.8631, "step": 16146 }, { "epoch": 0.17, "learning_rate": 9.276297029779403e-06, "loss": 0.8632, "step": 16147 }, { "epoch": 0.17, "learning_rate": 9.275221979133996e-06, "loss": 0.8745, "step": 16148 }, { "epoch": 0.17, "learning_rate": 9.274146936909266e-06, "loss": 0.8799, "step": 16149 }, { "epoch": 0.17, "learning_rate": 9.273071903117704e-06, "loss": 0.867, "step": 16150 }, { "epoch": 0.17, "learning_rate": 9.27199687777179e-06, "loss": 0.8814, "step": 16151 }, { "epoch": 0.17, "learning_rate": 9.27092186088402e-06, "loss": 0.8958, "step": 16152 }, { "epoch": 0.17, "learning_rate": 9.269846852466887e-06, "loss": 0.9352, "step": 16153 }, { "epoch": 0.17, "learning_rate": 9.268771852532877e-06, "loss": 0.881, "step": 16154 }, { "epoch": 0.17, "learning_rate": 9.267696861094477e-06, "loss": 0.9083, "step": 16155 }, { "epoch": 0.17, "learning_rate": 9.26662187816418e-06, "loss": 0.8837, "step": 16156 }, { "epoch": 0.17, "learning_rate": 9.26554690375448e-06, "loss": 0.8511, "step": 16157 }, { "epoch": 0.17, "learning_rate": 9.264471937877857e-06, "loss": 0.8856, "step": 16158 }, { "epoch": 0.17, "learning_rate": 9.263396980546804e-06, "loss": 0.8673, "step": 16159 }, { "epoch": 0.17, "learning_rate": 9.262322031773815e-06, "loss": 0.8625, "step": 16160 }, { "epoch": 0.17, "learning_rate": 9.261247091571372e-06, "loss": 0.8948, "step": 16161 }, { "epoch": 0.17, "learning_rate": 9.260172159951967e-06, "loss": 0.8964, "step": 16162 }, { "epoch": 0.17, "learning_rate": 9.259097236928088e-06, "loss": 0.8427, "step": 16163 }, { "epoch": 0.17, "learning_rate": 9.258022322512228e-06, "loss": 0.8623, "step": 16164 }, { "epoch": 0.17, "learning_rate": 9.256947416716869e-06, "loss": 0.86, "step": 16165 }, { "epoch": 0.17, "learning_rate": 9.255872519554501e-06, "loss": 0.8418, "step": 16166 }, { "epoch": 0.17, "learning_rate": 9.25479763103762e-06, "loss": 0.8587, "step": 16167 }, { "epoch": 0.17, "learning_rate": 9.253722751178704e-06, "loss": 0.8784, "step": 16168 }, { "epoch": 0.17, "learning_rate": 9.252647879990242e-06, "loss": 0.92, "step": 16169 }, { "epoch": 0.17, "learning_rate": 9.251573017484731e-06, "loss": 0.8975, "step": 16170 }, { "epoch": 0.17, "learning_rate": 9.250498163674657e-06, "loss": 0.8973, "step": 16171 }, { "epoch": 0.17, "learning_rate": 9.2494233185725e-06, "loss": 0.8823, "step": 16172 }, { "epoch": 0.17, "learning_rate": 9.248348482190751e-06, "loss": 0.8932, "step": 16173 }, { "epoch": 0.17, "learning_rate": 9.247273654541908e-06, "loss": 0.8978, "step": 16174 }, { "epoch": 0.17, "learning_rate": 9.246198835638442e-06, "loss": 0.8882, "step": 16175 }, { "epoch": 0.17, "learning_rate": 9.24512402549285e-06, "loss": 0.8526, "step": 16176 }, { "epoch": 0.17, "learning_rate": 9.244049224117619e-06, "loss": 0.8604, "step": 16177 }, { "epoch": 0.17, "learning_rate": 9.242974431525238e-06, "loss": 0.8878, "step": 16178 }, { "epoch": 0.17, "learning_rate": 9.24189964772819e-06, "loss": 0.8635, "step": 16179 }, { "epoch": 0.17, "learning_rate": 9.240824872738963e-06, "loss": 0.8679, "step": 16180 }, { "epoch": 0.17, "learning_rate": 9.239750106570044e-06, "loss": 0.9001, "step": 16181 }, { "epoch": 0.17, "learning_rate": 9.238675349233924e-06, "loss": 0.8723, "step": 16182 }, { "epoch": 0.17, "learning_rate": 9.237600600743086e-06, "loss": 0.9094, "step": 16183 }, { "epoch": 0.17, "learning_rate": 9.236525861110015e-06, "loss": 0.8913, "step": 16184 }, { "epoch": 0.17, "learning_rate": 9.235451130347203e-06, "loss": 0.9074, "step": 16185 }, { "epoch": 0.17, "learning_rate": 9.234376408467131e-06, "loss": 0.8826, "step": 16186 }, { "epoch": 0.17, "learning_rate": 9.233301695482288e-06, "loss": 0.871, "step": 16187 }, { "epoch": 0.17, "learning_rate": 9.232226991405158e-06, "loss": 0.9135, "step": 16188 }, { "epoch": 0.17, "learning_rate": 9.231152296248236e-06, "loss": 0.9162, "step": 16189 }, { "epoch": 0.17, "learning_rate": 9.230077610023998e-06, "loss": 0.8764, "step": 16190 }, { "epoch": 0.17, "learning_rate": 9.229002932744933e-06, "loss": 0.9068, "step": 16191 }, { "epoch": 0.17, "learning_rate": 9.227928264423532e-06, "loss": 0.8465, "step": 16192 }, { "epoch": 0.17, "learning_rate": 9.226853605072271e-06, "loss": 0.8463, "step": 16193 }, { "epoch": 0.17, "learning_rate": 9.22577895470364e-06, "loss": 0.8786, "step": 16194 }, { "epoch": 0.17, "learning_rate": 9.224704313330128e-06, "loss": 0.8292, "step": 16195 }, { "epoch": 0.17, "learning_rate": 9.22362968096422e-06, "loss": 0.8605, "step": 16196 }, { "epoch": 0.17, "learning_rate": 9.222555057618397e-06, "loss": 0.9155, "step": 16197 }, { "epoch": 0.17, "learning_rate": 9.221480443305147e-06, "loss": 0.8503, "step": 16198 }, { "epoch": 0.17, "learning_rate": 9.220405838036957e-06, "loss": 0.903, "step": 16199 }, { "epoch": 0.17, "learning_rate": 9.219331241826307e-06, "loss": 0.9319, "step": 16200 }, { "epoch": 0.17, "learning_rate": 9.218256654685686e-06, "loss": 0.8536, "step": 16201 }, { "epoch": 0.17, "learning_rate": 9.217182076627577e-06, "loss": 0.8638, "step": 16202 }, { "epoch": 0.17, "learning_rate": 9.216107507664468e-06, "loss": 0.8815, "step": 16203 }, { "epoch": 0.17, "learning_rate": 9.21503294780884e-06, "loss": 0.8809, "step": 16204 }, { "epoch": 0.17, "learning_rate": 9.213958397073178e-06, "loss": 0.8808, "step": 16205 }, { "epoch": 0.17, "learning_rate": 9.212883855469968e-06, "loss": 0.8588, "step": 16206 }, { "epoch": 0.17, "learning_rate": 9.211809323011693e-06, "loss": 0.87, "step": 16207 }, { "epoch": 0.17, "learning_rate": 9.210734799710837e-06, "loss": 0.862, "step": 16208 }, { "epoch": 0.17, "learning_rate": 9.209660285579884e-06, "loss": 0.9251, "step": 16209 }, { "epoch": 0.17, "learning_rate": 9.208585780631321e-06, "loss": 0.886, "step": 16210 }, { "epoch": 0.17, "learning_rate": 9.207511284877629e-06, "loss": 0.9043, "step": 16211 }, { "epoch": 0.17, "learning_rate": 9.20643679833129e-06, "loss": 0.8825, "step": 16212 }, { "epoch": 0.17, "learning_rate": 9.205362321004795e-06, "loss": 0.8928, "step": 16213 }, { "epoch": 0.17, "learning_rate": 9.20428785291062e-06, "loss": 0.8947, "step": 16214 }, { "epoch": 0.17, "learning_rate": 9.203213394061248e-06, "loss": 0.8954, "step": 16215 }, { "epoch": 0.17, "learning_rate": 9.202138944469168e-06, "loss": 0.8944, "step": 16216 }, { "epoch": 0.17, "learning_rate": 9.201064504146865e-06, "loss": 0.8939, "step": 16217 }, { "epoch": 0.17, "learning_rate": 9.199990073106813e-06, "loss": 0.8671, "step": 16218 }, { "epoch": 0.17, "learning_rate": 9.198915651361501e-06, "loss": 0.9044, "step": 16219 }, { "epoch": 0.17, "learning_rate": 9.197841238923416e-06, "loss": 0.8423, "step": 16220 }, { "epoch": 0.17, "learning_rate": 9.19676683580503e-06, "loss": 0.926, "step": 16221 }, { "epoch": 0.17, "learning_rate": 9.195692442018831e-06, "loss": 0.8719, "step": 16222 }, { "epoch": 0.17, "learning_rate": 9.194618057577304e-06, "loss": 0.88, "step": 16223 }, { "epoch": 0.17, "learning_rate": 9.19354368249293e-06, "loss": 0.8587, "step": 16224 }, { "epoch": 0.17, "learning_rate": 9.19246931677819e-06, "loss": 0.9323, "step": 16225 }, { "epoch": 0.17, "learning_rate": 9.191394960445568e-06, "loss": 0.9008, "step": 16226 }, { "epoch": 0.17, "learning_rate": 9.190320613507545e-06, "loss": 0.8777, "step": 16227 }, { "epoch": 0.17, "learning_rate": 9.189246275976605e-06, "loss": 0.8804, "step": 16228 }, { "epoch": 0.17, "learning_rate": 9.188171947865227e-06, "loss": 0.9076, "step": 16229 }, { "epoch": 0.17, "learning_rate": 9.187097629185894e-06, "loss": 0.923, "step": 16230 }, { "epoch": 0.17, "learning_rate": 9.18602331995109e-06, "loss": 0.8514, "step": 16231 }, { "epoch": 0.17, "learning_rate": 9.184949020173292e-06, "loss": 0.9121, "step": 16232 }, { "epoch": 0.17, "learning_rate": 9.183874729864986e-06, "loss": 0.89, "step": 16233 }, { "epoch": 0.17, "learning_rate": 9.182800449038647e-06, "loss": 0.8825, "step": 16234 }, { "epoch": 0.17, "learning_rate": 9.181726177706769e-06, "loss": 0.8682, "step": 16235 }, { "epoch": 0.17, "learning_rate": 9.180651915881819e-06, "loss": 0.8695, "step": 16236 }, { "epoch": 0.17, "learning_rate": 9.179577663576285e-06, "loss": 0.9233, "step": 16237 }, { "epoch": 0.17, "learning_rate": 9.178503420802651e-06, "loss": 0.9303, "step": 16238 }, { "epoch": 0.17, "learning_rate": 9.177429187573392e-06, "loss": 0.9223, "step": 16239 }, { "epoch": 0.17, "learning_rate": 9.176354963900987e-06, "loss": 0.8802, "step": 16240 }, { "epoch": 0.17, "learning_rate": 9.17528074979792e-06, "loss": 0.9144, "step": 16241 }, { "epoch": 0.17, "learning_rate": 9.174206545276678e-06, "loss": 0.8794, "step": 16242 }, { "epoch": 0.17, "learning_rate": 9.173132350349732e-06, "loss": 0.8303, "step": 16243 }, { "epoch": 0.17, "learning_rate": 9.172058165029565e-06, "loss": 0.8879, "step": 16244 }, { "epoch": 0.17, "learning_rate": 9.170983989328661e-06, "loss": 0.9005, "step": 16245 }, { "epoch": 0.17, "learning_rate": 9.169909823259493e-06, "loss": 0.8754, "step": 16246 }, { "epoch": 0.17, "learning_rate": 9.168835666834545e-06, "loss": 0.8711, "step": 16247 }, { "epoch": 0.17, "learning_rate": 9.167761520066297e-06, "loss": 0.8959, "step": 16248 }, { "epoch": 0.17, "learning_rate": 9.16668738296723e-06, "loss": 0.8986, "step": 16249 }, { "epoch": 0.17, "learning_rate": 9.165613255549821e-06, "loss": 0.9115, "step": 16250 }, { "epoch": 0.17, "learning_rate": 9.16453913782655e-06, "loss": 0.919, "step": 16251 }, { "epoch": 0.17, "learning_rate": 9.163465029809899e-06, "loss": 0.8574, "step": 16252 }, { "epoch": 0.17, "learning_rate": 9.162390931512342e-06, "loss": 0.8732, "step": 16253 }, { "epoch": 0.17, "learning_rate": 9.161316842946364e-06, "loss": 0.8971, "step": 16254 }, { "epoch": 0.17, "learning_rate": 9.16024276412444e-06, "loss": 0.8676, "step": 16255 }, { "epoch": 0.17, "learning_rate": 9.159168695059051e-06, "loss": 0.8628, "step": 16256 }, { "epoch": 0.17, "learning_rate": 9.158094635762674e-06, "loss": 0.8716, "step": 16257 }, { "epoch": 0.17, "learning_rate": 9.157020586247788e-06, "loss": 0.8952, "step": 16258 }, { "epoch": 0.18, "learning_rate": 9.155946546526878e-06, "loss": 0.8569, "step": 16259 }, { "epoch": 0.18, "learning_rate": 9.154872516612412e-06, "loss": 0.8869, "step": 16260 }, { "epoch": 0.18, "learning_rate": 9.153798496516873e-06, "loss": 0.9004, "step": 16261 }, { "epoch": 0.18, "learning_rate": 9.152724486252742e-06, "loss": 0.9228, "step": 16262 }, { "epoch": 0.18, "learning_rate": 9.151650485832497e-06, "loss": 0.9144, "step": 16263 }, { "epoch": 0.18, "learning_rate": 9.150576495268609e-06, "loss": 0.8352, "step": 16264 }, { "epoch": 0.18, "learning_rate": 9.149502514573562e-06, "loss": 0.8715, "step": 16265 }, { "epoch": 0.18, "learning_rate": 9.148428543759836e-06, "loss": 0.8899, "step": 16266 }, { "epoch": 0.18, "learning_rate": 9.1473545828399e-06, "loss": 0.9236, "step": 16267 }, { "epoch": 0.18, "learning_rate": 9.14628063182624e-06, "loss": 0.8834, "step": 16268 }, { "epoch": 0.18, "learning_rate": 9.145206690731328e-06, "loss": 0.9081, "step": 16269 }, { "epoch": 0.18, "learning_rate": 9.144132759567648e-06, "loss": 0.8984, "step": 16270 }, { "epoch": 0.18, "learning_rate": 9.143058838347669e-06, "loss": 0.8746, "step": 16271 }, { "epoch": 0.18, "learning_rate": 9.141984927083874e-06, "loss": 0.8833, "step": 16272 }, { "epoch": 0.18, "learning_rate": 9.140911025788737e-06, "loss": 0.9263, "step": 16273 }, { "epoch": 0.18, "learning_rate": 9.139837134474739e-06, "loss": 0.8753, "step": 16274 }, { "epoch": 0.18, "learning_rate": 9.13876325315435e-06, "loss": 0.8926, "step": 16275 }, { "epoch": 0.18, "learning_rate": 9.137689381840051e-06, "loss": 0.8745, "step": 16276 }, { "epoch": 0.18, "learning_rate": 9.136615520544319e-06, "loss": 0.8498, "step": 16277 }, { "epoch": 0.18, "learning_rate": 9.135541669279629e-06, "loss": 0.9175, "step": 16278 }, { "epoch": 0.18, "learning_rate": 9.134467828058456e-06, "loss": 0.8764, "step": 16279 }, { "epoch": 0.18, "learning_rate": 9.133393996893276e-06, "loss": 0.8732, "step": 16280 }, { "epoch": 0.18, "learning_rate": 9.132320175796574e-06, "loss": 0.9176, "step": 16281 }, { "epoch": 0.18, "learning_rate": 9.131246364780814e-06, "loss": 0.8795, "step": 16282 }, { "epoch": 0.18, "learning_rate": 9.130172563858474e-06, "loss": 0.8671, "step": 16283 }, { "epoch": 0.18, "learning_rate": 9.12909877304204e-06, "loss": 0.8665, "step": 16284 }, { "epoch": 0.18, "learning_rate": 9.128024992343972e-06, "loss": 0.8692, "step": 16285 }, { "epoch": 0.18, "learning_rate": 9.126951221776756e-06, "loss": 0.8963, "step": 16286 }, { "epoch": 0.18, "learning_rate": 9.125877461352865e-06, "loss": 0.8686, "step": 16287 }, { "epoch": 0.18, "learning_rate": 9.124803711084777e-06, "loss": 0.8925, "step": 16288 }, { "epoch": 0.18, "learning_rate": 9.123729970984962e-06, "loss": 0.8916, "step": 16289 }, { "epoch": 0.18, "learning_rate": 9.122656241065897e-06, "loss": 0.87, "step": 16290 }, { "epoch": 0.18, "learning_rate": 9.12158252134006e-06, "loss": 0.9171, "step": 16291 }, { "epoch": 0.18, "learning_rate": 9.120508811819918e-06, "loss": 0.8997, "step": 16292 }, { "epoch": 0.18, "learning_rate": 9.119435112517953e-06, "loss": 0.9084, "step": 16293 }, { "epoch": 0.18, "learning_rate": 9.118361423446636e-06, "loss": 0.8682, "step": 16294 }, { "epoch": 0.18, "learning_rate": 9.117287744618444e-06, "loss": 0.9102, "step": 16295 }, { "epoch": 0.18, "learning_rate": 9.116214076045848e-06, "loss": 0.8594, "step": 16296 }, { "epoch": 0.18, "learning_rate": 9.115140417741325e-06, "loss": 0.8834, "step": 16297 }, { "epoch": 0.18, "learning_rate": 9.11406676971735e-06, "loss": 0.8943, "step": 16298 }, { "epoch": 0.18, "learning_rate": 9.112993131986393e-06, "loss": 0.8848, "step": 16299 }, { "epoch": 0.18, "learning_rate": 9.111919504560928e-06, "loss": 0.9059, "step": 16300 }, { "epoch": 0.18, "learning_rate": 9.11084588745343e-06, "loss": 0.8874, "step": 16301 }, { "epoch": 0.18, "learning_rate": 9.10977228067638e-06, "loss": 0.8741, "step": 16302 }, { "epoch": 0.18, "learning_rate": 9.10869868424224e-06, "loss": 0.8975, "step": 16303 }, { "epoch": 0.18, "learning_rate": 9.107625098163484e-06, "loss": 0.8633, "step": 16304 }, { "epoch": 0.18, "learning_rate": 9.106551522452598e-06, "loss": 0.8885, "step": 16305 }, { "epoch": 0.18, "learning_rate": 9.10547795712204e-06, "loss": 0.8862, "step": 16306 }, { "epoch": 0.18, "learning_rate": 9.104404402184287e-06, "loss": 0.8771, "step": 16307 }, { "epoch": 0.18, "learning_rate": 9.103330857651818e-06, "loss": 0.859, "step": 16308 }, { "epoch": 0.18, "learning_rate": 9.102257323537104e-06, "loss": 0.9013, "step": 16309 }, { "epoch": 0.18, "learning_rate": 9.101183799852611e-06, "loss": 0.8691, "step": 16310 }, { "epoch": 0.18, "learning_rate": 9.100110286610816e-06, "loss": 0.8875, "step": 16311 }, { "epoch": 0.18, "learning_rate": 9.099036783824198e-06, "loss": 0.8511, "step": 16312 }, { "epoch": 0.18, "learning_rate": 9.097963291505215e-06, "loss": 0.8642, "step": 16313 }, { "epoch": 0.18, "learning_rate": 9.096889809666349e-06, "loss": 0.8534, "step": 16314 }, { "epoch": 0.18, "learning_rate": 9.09581633832007e-06, "loss": 0.9148, "step": 16315 }, { "epoch": 0.18, "learning_rate": 9.09474287747885e-06, "loss": 0.8516, "step": 16316 }, { "epoch": 0.18, "learning_rate": 9.09366942715516e-06, "loss": 0.9308, "step": 16317 }, { "epoch": 0.18, "learning_rate": 9.09259598736147e-06, "loss": 0.8773, "step": 16318 }, { "epoch": 0.18, "learning_rate": 9.091522558110256e-06, "loss": 0.8449, "step": 16319 }, { "epoch": 0.18, "learning_rate": 9.090449139413987e-06, "loss": 0.901, "step": 16320 }, { "epoch": 0.18, "learning_rate": 9.089375731285134e-06, "loss": 0.8667, "step": 16321 }, { "epoch": 0.18, "learning_rate": 9.088302333736166e-06, "loss": 0.8687, "step": 16322 }, { "epoch": 0.18, "learning_rate": 9.08722894677956e-06, "loss": 0.8795, "step": 16323 }, { "epoch": 0.18, "learning_rate": 9.086155570427782e-06, "loss": 0.9, "step": 16324 }, { "epoch": 0.18, "learning_rate": 9.085082204693302e-06, "loss": 0.8755, "step": 16325 }, { "epoch": 0.18, "learning_rate": 9.084008849588593e-06, "loss": 0.8779, "step": 16326 }, { "epoch": 0.18, "learning_rate": 9.082935505126131e-06, "loss": 0.8824, "step": 16327 }, { "epoch": 0.18, "learning_rate": 9.081862171318375e-06, "loss": 0.8687, "step": 16328 }, { "epoch": 0.18, "learning_rate": 9.080788848177799e-06, "loss": 0.8597, "step": 16329 }, { "epoch": 0.18, "learning_rate": 9.079715535716884e-06, "loss": 0.8664, "step": 16330 }, { "epoch": 0.18, "learning_rate": 9.078642233948082e-06, "loss": 0.8989, "step": 16331 }, { "epoch": 0.18, "learning_rate": 9.077568942883877e-06, "loss": 0.8757, "step": 16332 }, { "epoch": 0.18, "learning_rate": 9.076495662536732e-06, "loss": 0.896, "step": 16333 }, { "epoch": 0.18, "learning_rate": 9.07542239291912e-06, "loss": 0.8953, "step": 16334 }, { "epoch": 0.18, "learning_rate": 9.074349134043509e-06, "loss": 0.8552, "step": 16335 }, { "epoch": 0.18, "learning_rate": 9.073275885922368e-06, "loss": 0.8918, "step": 16336 }, { "epoch": 0.18, "learning_rate": 9.072202648568167e-06, "loss": 0.8547, "step": 16337 }, { "epoch": 0.18, "learning_rate": 9.071129421993376e-06, "loss": 0.8501, "step": 16338 }, { "epoch": 0.18, "learning_rate": 9.070056206210462e-06, "loss": 0.8956, "step": 16339 }, { "epoch": 0.18, "learning_rate": 9.068983001231895e-06, "loss": 0.8902, "step": 16340 }, { "epoch": 0.18, "learning_rate": 9.067909807070146e-06, "loss": 0.8698, "step": 16341 }, { "epoch": 0.18, "learning_rate": 9.066836623737677e-06, "loss": 0.9223, "step": 16342 }, { "epoch": 0.18, "learning_rate": 9.065763451246964e-06, "loss": 0.901, "step": 16343 }, { "epoch": 0.18, "learning_rate": 9.064690289610474e-06, "loss": 0.9127, "step": 16344 }, { "epoch": 0.18, "learning_rate": 9.063617138840671e-06, "loss": 0.8696, "step": 16345 }, { "epoch": 0.18, "learning_rate": 9.062543998950027e-06, "loss": 0.8638, "step": 16346 }, { "epoch": 0.18, "learning_rate": 9.061470869951005e-06, "loss": 0.9085, "step": 16347 }, { "epoch": 0.18, "learning_rate": 9.060397751856084e-06, "loss": 0.8659, "step": 16348 }, { "epoch": 0.18, "learning_rate": 9.05932464467772e-06, "loss": 0.941, "step": 16349 }, { "epoch": 0.18, "learning_rate": 9.058251548428385e-06, "loss": 0.899, "step": 16350 }, { "epoch": 0.18, "learning_rate": 9.057178463120551e-06, "loss": 0.8559, "step": 16351 }, { "epoch": 0.18, "learning_rate": 9.056105388766678e-06, "loss": 0.8496, "step": 16352 }, { "epoch": 0.18, "learning_rate": 9.055032325379234e-06, "loss": 0.8675, "step": 16353 }, { "epoch": 0.18, "learning_rate": 9.053959272970692e-06, "loss": 0.8694, "step": 16354 }, { "epoch": 0.18, "learning_rate": 9.05288623155352e-06, "loss": 0.8875, "step": 16355 }, { "epoch": 0.18, "learning_rate": 9.051813201140171e-06, "loss": 0.8695, "step": 16356 }, { "epoch": 0.18, "learning_rate": 9.050740181743127e-06, "loss": 0.9091, "step": 16357 }, { "epoch": 0.18, "learning_rate": 9.04966717337485e-06, "loss": 0.8527, "step": 16358 }, { "epoch": 0.18, "learning_rate": 9.048594176047802e-06, "loss": 0.8441, "step": 16359 }, { "epoch": 0.18, "learning_rate": 9.047521189774456e-06, "loss": 0.8652, "step": 16360 }, { "epoch": 0.18, "learning_rate": 9.046448214567272e-06, "loss": 0.8786, "step": 16361 }, { "epoch": 0.18, "learning_rate": 9.045375250438723e-06, "loss": 0.9016, "step": 16362 }, { "epoch": 0.18, "learning_rate": 9.044302297401268e-06, "loss": 0.8703, "step": 16363 }, { "epoch": 0.18, "learning_rate": 9.043229355467377e-06, "loss": 0.9223, "step": 16364 }, { "epoch": 0.18, "learning_rate": 9.042156424649515e-06, "loss": 0.8431, "step": 16365 }, { "epoch": 0.18, "learning_rate": 9.041083504960149e-06, "loss": 0.8712, "step": 16366 }, { "epoch": 0.18, "learning_rate": 9.040010596411742e-06, "loss": 0.8811, "step": 16367 }, { "epoch": 0.18, "learning_rate": 9.038937699016757e-06, "loss": 0.8903, "step": 16368 }, { "epoch": 0.18, "learning_rate": 9.037864812787668e-06, "loss": 0.9153, "step": 16369 }, { "epoch": 0.18, "learning_rate": 9.036791937736932e-06, "loss": 0.8564, "step": 16370 }, { "epoch": 0.18, "learning_rate": 9.035719073877017e-06, "loss": 0.8406, "step": 16371 }, { "epoch": 0.18, "learning_rate": 9.034646221220383e-06, "loss": 0.8425, "step": 16372 }, { "epoch": 0.18, "learning_rate": 9.033573379779506e-06, "loss": 0.8879, "step": 16373 }, { "epoch": 0.18, "learning_rate": 9.03250054956684e-06, "loss": 0.8839, "step": 16374 }, { "epoch": 0.18, "learning_rate": 9.031427730594854e-06, "loss": 0.9233, "step": 16375 }, { "epoch": 0.18, "learning_rate": 9.030354922876013e-06, "loss": 0.8719, "step": 16376 }, { "epoch": 0.18, "learning_rate": 9.029282126422775e-06, "loss": 0.8668, "step": 16377 }, { "epoch": 0.18, "learning_rate": 9.028209341247612e-06, "loss": 0.8689, "step": 16378 }, { "epoch": 0.18, "learning_rate": 9.027136567362984e-06, "loss": 0.9027, "step": 16379 }, { "epoch": 0.18, "learning_rate": 9.026063804781357e-06, "loss": 0.9225, "step": 16380 }, { "epoch": 0.18, "learning_rate": 9.024991053515192e-06, "loss": 0.8866, "step": 16381 }, { "epoch": 0.18, "learning_rate": 9.023918313576953e-06, "loss": 0.9023, "step": 16382 }, { "epoch": 0.18, "learning_rate": 9.022845584979104e-06, "loss": 0.8975, "step": 16383 }, { "epoch": 0.18, "learning_rate": 9.02177286773411e-06, "loss": 0.8987, "step": 16384 }, { "epoch": 0.18, "learning_rate": 9.020700161854428e-06, "loss": 0.8929, "step": 16385 }, { "epoch": 0.18, "learning_rate": 9.019627467352529e-06, "loss": 0.8541, "step": 16386 }, { "epoch": 0.18, "learning_rate": 9.018554784240873e-06, "loss": 0.9454, "step": 16387 }, { "epoch": 0.18, "learning_rate": 9.017482112531919e-06, "loss": 0.8524, "step": 16388 }, { "epoch": 0.18, "learning_rate": 9.016409452238133e-06, "loss": 0.8664, "step": 16389 }, { "epoch": 0.18, "learning_rate": 9.01533680337198e-06, "loss": 0.8653, "step": 16390 }, { "epoch": 0.18, "learning_rate": 9.014264165945917e-06, "loss": 0.8841, "step": 16391 }, { "epoch": 0.18, "learning_rate": 9.013191539972407e-06, "loss": 0.8722, "step": 16392 }, { "epoch": 0.18, "learning_rate": 9.012118925463913e-06, "loss": 0.8547, "step": 16393 }, { "epoch": 0.18, "learning_rate": 9.011046322432903e-06, "loss": 0.8791, "step": 16394 }, { "epoch": 0.18, "learning_rate": 9.00997373089183e-06, "loss": 0.9302, "step": 16395 }, { "epoch": 0.18, "learning_rate": 9.008901150853157e-06, "loss": 0.8732, "step": 16396 }, { "epoch": 0.18, "learning_rate": 9.007828582329353e-06, "loss": 0.8534, "step": 16397 }, { "epoch": 0.18, "learning_rate": 9.00675602533287e-06, "loss": 0.8857, "step": 16398 }, { "epoch": 0.18, "learning_rate": 9.00568347987617e-06, "loss": 0.9027, "step": 16399 }, { "epoch": 0.18, "learning_rate": 9.00461094597172e-06, "loss": 0.9076, "step": 16400 }, { "epoch": 0.18, "learning_rate": 9.003538423631985e-06, "loss": 0.9004, "step": 16401 }, { "epoch": 0.18, "learning_rate": 9.00246591286941e-06, "loss": 0.9081, "step": 16402 }, { "epoch": 0.18, "learning_rate": 9.001393413696469e-06, "loss": 0.9046, "step": 16403 }, { "epoch": 0.18, "learning_rate": 9.000320926125618e-06, "loss": 0.8749, "step": 16404 }, { "epoch": 0.18, "learning_rate": 8.999248450169319e-06, "loss": 0.9003, "step": 16405 }, { "epoch": 0.18, "learning_rate": 8.998175985840029e-06, "loss": 0.878, "step": 16406 }, { "epoch": 0.18, "learning_rate": 8.997103533150212e-06, "loss": 0.8889, "step": 16407 }, { "epoch": 0.18, "learning_rate": 8.996031092112329e-06, "loss": 0.9019, "step": 16408 }, { "epoch": 0.18, "learning_rate": 8.994958662738833e-06, "loss": 0.8874, "step": 16409 }, { "epoch": 0.18, "learning_rate": 8.99388624504219e-06, "loss": 0.8905, "step": 16410 }, { "epoch": 0.18, "learning_rate": 8.992813839034858e-06, "loss": 0.9228, "step": 16411 }, { "epoch": 0.18, "learning_rate": 8.991741444729299e-06, "loss": 0.8973, "step": 16412 }, { "epoch": 0.18, "learning_rate": 8.990669062137966e-06, "loss": 0.9092, "step": 16413 }, { "epoch": 0.18, "learning_rate": 8.989596691273324e-06, "loss": 0.8679, "step": 16414 }, { "epoch": 0.18, "learning_rate": 8.98852433214783e-06, "loss": 0.9012, "step": 16415 }, { "epoch": 0.18, "learning_rate": 8.987451984773943e-06, "loss": 0.9394, "step": 16416 }, { "epoch": 0.18, "learning_rate": 8.986379649164121e-06, "loss": 0.8697, "step": 16417 }, { "epoch": 0.18, "learning_rate": 8.985307325330821e-06, "loss": 0.866, "step": 16418 }, { "epoch": 0.18, "learning_rate": 8.984235013286512e-06, "loss": 0.9344, "step": 16419 }, { "epoch": 0.18, "learning_rate": 8.983162713043638e-06, "loss": 0.8588, "step": 16420 }, { "epoch": 0.18, "learning_rate": 8.982090424614665e-06, "loss": 0.8639, "step": 16421 }, { "epoch": 0.18, "learning_rate": 8.981018148012054e-06, "loss": 0.9159, "step": 16422 }, { "epoch": 0.18, "learning_rate": 8.979945883248253e-06, "loss": 0.8834, "step": 16423 }, { "epoch": 0.18, "learning_rate": 8.97887363033573e-06, "loss": 0.8422, "step": 16424 }, { "epoch": 0.18, "learning_rate": 8.977801389286937e-06, "loss": 0.9002, "step": 16425 }, { "epoch": 0.18, "learning_rate": 8.976729160114334e-06, "loss": 0.8767, "step": 16426 }, { "epoch": 0.18, "learning_rate": 8.975656942830379e-06, "loss": 0.8981, "step": 16427 }, { "epoch": 0.18, "learning_rate": 8.974584737447526e-06, "loss": 0.9069, "step": 16428 }, { "epoch": 0.18, "learning_rate": 8.973512543978238e-06, "loss": 0.8238, "step": 16429 }, { "epoch": 0.18, "learning_rate": 8.972440362434963e-06, "loss": 0.8822, "step": 16430 }, { "epoch": 0.18, "learning_rate": 8.971368192830165e-06, "loss": 0.8763, "step": 16431 }, { "epoch": 0.18, "learning_rate": 8.9702960351763e-06, "loss": 0.9214, "step": 16432 }, { "epoch": 0.18, "learning_rate": 8.969223889485824e-06, "loss": 0.9013, "step": 16433 }, { "epoch": 0.18, "learning_rate": 8.968151755771191e-06, "loss": 0.867, "step": 16434 }, { "epoch": 0.18, "learning_rate": 8.96707963404486e-06, "loss": 0.8339, "step": 16435 }, { "epoch": 0.18, "learning_rate": 8.966007524319289e-06, "loss": 0.8715, "step": 16436 }, { "epoch": 0.18, "learning_rate": 8.96493542660693e-06, "loss": 0.8787, "step": 16437 }, { "epoch": 0.18, "learning_rate": 8.963863340920239e-06, "loss": 0.8474, "step": 16438 }, { "epoch": 0.18, "learning_rate": 8.962791267271672e-06, "loss": 0.8848, "step": 16439 }, { "epoch": 0.18, "learning_rate": 8.961719205673692e-06, "loss": 0.9074, "step": 16440 }, { "epoch": 0.18, "learning_rate": 8.960647156138745e-06, "loss": 0.8953, "step": 16441 }, { "epoch": 0.18, "learning_rate": 8.959575118679289e-06, "loss": 0.8684, "step": 16442 }, { "epoch": 0.18, "learning_rate": 8.958503093307785e-06, "loss": 0.9298, "step": 16443 }, { "epoch": 0.18, "learning_rate": 8.95743108003668e-06, "loss": 0.8586, "step": 16444 }, { "epoch": 0.18, "learning_rate": 8.956359078878428e-06, "loss": 0.8726, "step": 16445 }, { "epoch": 0.18, "learning_rate": 8.955287089845491e-06, "loss": 0.8869, "step": 16446 }, { "epoch": 0.18, "learning_rate": 8.954215112950326e-06, "loss": 0.8901, "step": 16447 }, { "epoch": 0.18, "learning_rate": 8.953143148205375e-06, "loss": 0.8713, "step": 16448 }, { "epoch": 0.18, "learning_rate": 8.952071195623102e-06, "loss": 0.8861, "step": 16449 }, { "epoch": 0.18, "learning_rate": 8.950999255215962e-06, "loss": 0.9079, "step": 16450 }, { "epoch": 0.18, "learning_rate": 8.949927326996403e-06, "loss": 0.8798, "step": 16451 }, { "epoch": 0.18, "learning_rate": 8.948855410976881e-06, "loss": 0.8992, "step": 16452 }, { "epoch": 0.18, "learning_rate": 8.947783507169852e-06, "loss": 0.8729, "step": 16453 }, { "epoch": 0.18, "learning_rate": 8.94671161558777e-06, "loss": 0.8543, "step": 16454 }, { "epoch": 0.18, "learning_rate": 8.945639736243085e-06, "loss": 0.8812, "step": 16455 }, { "epoch": 0.18, "learning_rate": 8.944567869148251e-06, "loss": 0.8726, "step": 16456 }, { "epoch": 0.18, "learning_rate": 8.943496014315725e-06, "loss": 0.8822, "step": 16457 }, { "epoch": 0.18, "learning_rate": 8.942424171757958e-06, "loss": 0.8948, "step": 16458 }, { "epoch": 0.18, "learning_rate": 8.941352341487402e-06, "loss": 0.9039, "step": 16459 }, { "epoch": 0.18, "learning_rate": 8.94028052351651e-06, "loss": 0.8753, "step": 16460 }, { "epoch": 0.18, "learning_rate": 8.939208717857737e-06, "loss": 0.8792, "step": 16461 }, { "epoch": 0.18, "learning_rate": 8.938136924523531e-06, "loss": 0.8673, "step": 16462 }, { "epoch": 0.18, "learning_rate": 8.937065143526349e-06, "loss": 0.8904, "step": 16463 }, { "epoch": 0.18, "learning_rate": 8.935993374878637e-06, "loss": 0.8766, "step": 16464 }, { "epoch": 0.18, "learning_rate": 8.93492161859286e-06, "loss": 0.858, "step": 16465 }, { "epoch": 0.18, "learning_rate": 8.933849874681453e-06, "loss": 0.9143, "step": 16466 }, { "epoch": 0.18, "learning_rate": 8.932778143156878e-06, "loss": 0.8652, "step": 16467 }, { "epoch": 0.18, "learning_rate": 8.931706424031591e-06, "loss": 0.924, "step": 16468 }, { "epoch": 0.18, "learning_rate": 8.93063471731803e-06, "loss": 0.8583, "step": 16469 }, { "epoch": 0.18, "learning_rate": 8.929563023028656e-06, "loss": 0.8948, "step": 16470 }, { "epoch": 0.18, "learning_rate": 8.92849134117592e-06, "loss": 0.8605, "step": 16471 }, { "epoch": 0.18, "learning_rate": 8.927419671772271e-06, "loss": 0.8767, "step": 16472 }, { "epoch": 0.18, "learning_rate": 8.92634801483016e-06, "loss": 0.8908, "step": 16473 }, { "epoch": 0.18, "learning_rate": 8.925276370362036e-06, "loss": 0.8179, "step": 16474 }, { "epoch": 0.18, "learning_rate": 8.924204738380354e-06, "loss": 0.8909, "step": 16475 }, { "epoch": 0.18, "learning_rate": 8.92313311889756e-06, "loss": 0.8768, "step": 16476 }, { "epoch": 0.18, "learning_rate": 8.922061511926107e-06, "loss": 0.901, "step": 16477 }, { "epoch": 0.18, "learning_rate": 8.920989917478446e-06, "loss": 0.8565, "step": 16478 }, { "epoch": 0.18, "learning_rate": 8.919918335567027e-06, "loss": 0.8722, "step": 16479 }, { "epoch": 0.18, "learning_rate": 8.918846766204297e-06, "loss": 0.8817, "step": 16480 }, { "epoch": 0.18, "learning_rate": 8.917775209402705e-06, "loss": 0.8396, "step": 16481 }, { "epoch": 0.18, "learning_rate": 8.916703665174708e-06, "loss": 0.8709, "step": 16482 }, { "epoch": 0.18, "learning_rate": 8.91563213353275e-06, "loss": 0.8946, "step": 16483 }, { "epoch": 0.18, "learning_rate": 8.914560614489279e-06, "loss": 0.8766, "step": 16484 }, { "epoch": 0.18, "learning_rate": 8.913489108056744e-06, "loss": 0.8881, "step": 16485 }, { "epoch": 0.18, "learning_rate": 8.912417614247601e-06, "loss": 0.8632, "step": 16486 }, { "epoch": 0.18, "learning_rate": 8.911346133074292e-06, "loss": 0.8834, "step": 16487 }, { "epoch": 0.18, "learning_rate": 8.910274664549266e-06, "loss": 0.9174, "step": 16488 }, { "epoch": 0.18, "learning_rate": 8.90920320868498e-06, "loss": 0.8744, "step": 16489 }, { "epoch": 0.18, "learning_rate": 8.908131765493873e-06, "loss": 0.863, "step": 16490 }, { "epoch": 0.18, "learning_rate": 8.907060334988392e-06, "loss": 0.8974, "step": 16491 }, { "epoch": 0.18, "learning_rate": 8.905988917180992e-06, "loss": 0.8763, "step": 16492 }, { "epoch": 0.18, "learning_rate": 8.904917512084122e-06, "loss": 0.88, "step": 16493 }, { "epoch": 0.18, "learning_rate": 8.903846119710225e-06, "loss": 0.9019, "step": 16494 }, { "epoch": 0.18, "learning_rate": 8.902774740071749e-06, "loss": 0.8898, "step": 16495 }, { "epoch": 0.18, "learning_rate": 8.901703373181144e-06, "loss": 0.902, "step": 16496 }, { "epoch": 0.18, "learning_rate": 8.900632019050855e-06, "loss": 0.894, "step": 16497 }, { "epoch": 0.18, "learning_rate": 8.89956067769333e-06, "loss": 0.8736, "step": 16498 }, { "epoch": 0.18, "learning_rate": 8.898489349121019e-06, "loss": 0.8601, "step": 16499 }, { "epoch": 0.18, "learning_rate": 8.897418033346367e-06, "loss": 0.8663, "step": 16500 }, { "epoch": 0.18, "learning_rate": 8.89634673038182e-06, "loss": 0.9205, "step": 16501 }, { "epoch": 0.18, "learning_rate": 8.895275440239824e-06, "loss": 0.8715, "step": 16502 }, { "epoch": 0.18, "learning_rate": 8.894204162932826e-06, "loss": 0.8792, "step": 16503 }, { "epoch": 0.18, "learning_rate": 8.893132898473277e-06, "loss": 0.8901, "step": 16504 }, { "epoch": 0.18, "learning_rate": 8.892061646873616e-06, "loss": 0.8485, "step": 16505 }, { "epoch": 0.18, "learning_rate": 8.890990408146294e-06, "loss": 0.8936, "step": 16506 }, { "epoch": 0.18, "learning_rate": 8.889919182303757e-06, "loss": 0.8979, "step": 16507 }, { "epoch": 0.18, "learning_rate": 8.888847969358446e-06, "loss": 0.907, "step": 16508 }, { "epoch": 0.18, "learning_rate": 8.887776769322808e-06, "loss": 0.9146, "step": 16509 }, { "epoch": 0.18, "learning_rate": 8.886705582209295e-06, "loss": 0.8757, "step": 16510 }, { "epoch": 0.18, "learning_rate": 8.88563440803035e-06, "loss": 0.9136, "step": 16511 }, { "epoch": 0.18, "learning_rate": 8.884563246798411e-06, "loss": 0.8978, "step": 16512 }, { "epoch": 0.18, "learning_rate": 8.88349209852593e-06, "loss": 0.89, "step": 16513 }, { "epoch": 0.18, "learning_rate": 8.882420963225354e-06, "loss": 0.8937, "step": 16514 }, { "epoch": 0.18, "learning_rate": 8.881349840909116e-06, "loss": 0.8936, "step": 16515 }, { "epoch": 0.18, "learning_rate": 8.880278731589672e-06, "loss": 0.8785, "step": 16516 }, { "epoch": 0.18, "learning_rate": 8.879207635279463e-06, "loss": 0.8856, "step": 16517 }, { "epoch": 0.18, "learning_rate": 8.878136551990934e-06, "loss": 0.8992, "step": 16518 }, { "epoch": 0.18, "learning_rate": 8.877065481736525e-06, "loss": 0.9083, "step": 16519 }, { "epoch": 0.18, "learning_rate": 8.875994424528686e-06, "loss": 0.8603, "step": 16520 }, { "epoch": 0.18, "learning_rate": 8.874923380379859e-06, "loss": 0.8934, "step": 16521 }, { "epoch": 0.18, "learning_rate": 8.873852349302484e-06, "loss": 0.8549, "step": 16522 }, { "epoch": 0.18, "learning_rate": 8.872781331309008e-06, "loss": 0.9234, "step": 16523 }, { "epoch": 0.18, "learning_rate": 8.871710326411874e-06, "loss": 0.8881, "step": 16524 }, { "epoch": 0.18, "learning_rate": 8.870639334623525e-06, "loss": 0.8781, "step": 16525 }, { "epoch": 0.18, "learning_rate": 8.869568355956404e-06, "loss": 0.8686, "step": 16526 }, { "epoch": 0.18, "learning_rate": 8.868497390422954e-06, "loss": 0.8757, "step": 16527 }, { "epoch": 0.18, "learning_rate": 8.867426438035618e-06, "loss": 0.8302, "step": 16528 }, { "epoch": 0.18, "learning_rate": 8.866355498806837e-06, "loss": 0.8991, "step": 16529 }, { "epoch": 0.18, "learning_rate": 8.865284572749055e-06, "loss": 0.8725, "step": 16530 }, { "epoch": 0.18, "learning_rate": 8.864213659874712e-06, "loss": 0.8774, "step": 16531 }, { "epoch": 0.18, "learning_rate": 8.86314276019626e-06, "loss": 0.8526, "step": 16532 }, { "epoch": 0.18, "learning_rate": 8.862071873726127e-06, "loss": 0.8835, "step": 16533 }, { "epoch": 0.18, "learning_rate": 8.86100100047676e-06, "loss": 0.8846, "step": 16534 }, { "epoch": 0.18, "learning_rate": 8.859930140460607e-06, "loss": 0.8709, "step": 16535 }, { "epoch": 0.18, "learning_rate": 8.858859293690101e-06, "loss": 0.8815, "step": 16536 }, { "epoch": 0.18, "learning_rate": 8.857788460177685e-06, "loss": 0.8841, "step": 16537 }, { "epoch": 0.18, "learning_rate": 8.856717639935803e-06, "loss": 0.8879, "step": 16538 }, { "epoch": 0.18, "learning_rate": 8.8556468329769e-06, "loss": 0.8407, "step": 16539 }, { "epoch": 0.18, "learning_rate": 8.854576039313407e-06, "loss": 0.8656, "step": 16540 }, { "epoch": 0.18, "learning_rate": 8.85350525895777e-06, "loss": 0.8828, "step": 16541 }, { "epoch": 0.18, "learning_rate": 8.852434491922434e-06, "loss": 0.8802, "step": 16542 }, { "epoch": 0.18, "learning_rate": 8.851363738219832e-06, "loss": 0.8328, "step": 16543 }, { "epoch": 0.18, "learning_rate": 8.850292997862405e-06, "loss": 0.8589, "step": 16544 }, { "epoch": 0.18, "learning_rate": 8.849222270862597e-06, "loss": 0.8424, "step": 16545 }, { "epoch": 0.18, "learning_rate": 8.848151557232849e-06, "loss": 0.8837, "step": 16546 }, { "epoch": 0.18, "learning_rate": 8.847080856985595e-06, "loss": 0.881, "step": 16547 }, { "epoch": 0.18, "learning_rate": 8.84601017013328e-06, "loss": 0.8736, "step": 16548 }, { "epoch": 0.18, "learning_rate": 8.844939496688339e-06, "loss": 0.8936, "step": 16549 }, { "epoch": 0.18, "learning_rate": 8.843868836663217e-06, "loss": 0.9269, "step": 16550 }, { "epoch": 0.18, "learning_rate": 8.842798190070348e-06, "loss": 0.8638, "step": 16551 }, { "epoch": 0.18, "learning_rate": 8.841727556922173e-06, "loss": 0.864, "step": 16552 }, { "epoch": 0.18, "learning_rate": 8.840656937231134e-06, "loss": 0.9033, "step": 16553 }, { "epoch": 0.18, "learning_rate": 8.839586331009663e-06, "loss": 0.8743, "step": 16554 }, { "epoch": 0.18, "learning_rate": 8.838515738270203e-06, "loss": 0.8765, "step": 16555 }, { "epoch": 0.18, "learning_rate": 8.837445159025194e-06, "loss": 0.9, "step": 16556 }, { "epoch": 0.18, "learning_rate": 8.836374593287073e-06, "loss": 0.8939, "step": 16557 }, { "epoch": 0.18, "learning_rate": 8.835304041068273e-06, "loss": 0.9005, "step": 16558 }, { "epoch": 0.19, "learning_rate": 8.83423350238124e-06, "loss": 0.871, "step": 16559 }, { "epoch": 0.19, "learning_rate": 8.83316297723841e-06, "loss": 0.8574, "step": 16560 }, { "epoch": 0.19, "learning_rate": 8.832092465652213e-06, "loss": 0.8621, "step": 16561 }, { "epoch": 0.19, "learning_rate": 8.831021967635094e-06, "loss": 0.878, "step": 16562 }, { "epoch": 0.19, "learning_rate": 8.829951483199489e-06, "loss": 0.8623, "step": 16563 }, { "epoch": 0.19, "learning_rate": 8.828881012357839e-06, "loss": 0.875, "step": 16564 }, { "epoch": 0.19, "learning_rate": 8.82781055512257e-06, "loss": 0.8944, "step": 16565 }, { "epoch": 0.19, "learning_rate": 8.82674011150613e-06, "loss": 0.856, "step": 16566 }, { "epoch": 0.19, "learning_rate": 8.825669681520954e-06, "loss": 0.8687, "step": 16567 }, { "epoch": 0.19, "learning_rate": 8.82459926517947e-06, "loss": 0.8796, "step": 16568 }, { "epoch": 0.19, "learning_rate": 8.823528862494124e-06, "loss": 0.8914, "step": 16569 }, { "epoch": 0.19, "learning_rate": 8.822458473477346e-06, "loss": 0.8846, "step": 16570 }, { "epoch": 0.19, "learning_rate": 8.821388098141578e-06, "loss": 0.8582, "step": 16571 }, { "epoch": 0.19, "learning_rate": 8.82031773649925e-06, "loss": 0.8887, "step": 16572 }, { "epoch": 0.19, "learning_rate": 8.8192473885628e-06, "loss": 0.8696, "step": 16573 }, { "epoch": 0.19, "learning_rate": 8.818177054344668e-06, "loss": 0.8827, "step": 16574 }, { "epoch": 0.19, "learning_rate": 8.81710673385728e-06, "loss": 0.8794, "step": 16575 }, { "epoch": 0.19, "learning_rate": 8.81603642711308e-06, "loss": 0.8449, "step": 16576 }, { "epoch": 0.19, "learning_rate": 8.814966134124497e-06, "loss": 0.9112, "step": 16577 }, { "epoch": 0.19, "learning_rate": 8.813895854903974e-06, "loss": 0.894, "step": 16578 }, { "epoch": 0.19, "learning_rate": 8.812825589463938e-06, "loss": 0.8918, "step": 16579 }, { "epoch": 0.19, "learning_rate": 8.811755337816824e-06, "loss": 0.9194, "step": 16580 }, { "epoch": 0.19, "learning_rate": 8.810685099975074e-06, "loss": 0.9164, "step": 16581 }, { "epoch": 0.19, "learning_rate": 8.80961487595111e-06, "loss": 0.8586, "step": 16582 }, { "epoch": 0.19, "learning_rate": 8.808544665757376e-06, "loss": 0.8816, "step": 16583 }, { "epoch": 0.19, "learning_rate": 8.807474469406303e-06, "loss": 0.8615, "step": 16584 }, { "epoch": 0.19, "learning_rate": 8.806404286910329e-06, "loss": 0.9133, "step": 16585 }, { "epoch": 0.19, "learning_rate": 8.80533411828188e-06, "loss": 0.8898, "step": 16586 }, { "epoch": 0.19, "learning_rate": 8.804263963533393e-06, "loss": 0.8745, "step": 16587 }, { "epoch": 0.19, "learning_rate": 8.803193822677304e-06, "loss": 0.8726, "step": 16588 }, { "epoch": 0.19, "learning_rate": 8.802123695726042e-06, "loss": 0.8877, "step": 16589 }, { "epoch": 0.19, "learning_rate": 8.801053582692041e-06, "loss": 0.8585, "step": 16590 }, { "epoch": 0.19, "learning_rate": 8.799983483587736e-06, "loss": 0.8882, "step": 16591 }, { "epoch": 0.19, "learning_rate": 8.79891339842556e-06, "loss": 0.8621, "step": 16592 }, { "epoch": 0.19, "learning_rate": 8.797843327217942e-06, "loss": 0.8674, "step": 16593 }, { "epoch": 0.19, "learning_rate": 8.796773269977315e-06, "loss": 0.8684, "step": 16594 }, { "epoch": 0.19, "learning_rate": 8.795703226716115e-06, "loss": 0.8703, "step": 16595 }, { "epoch": 0.19, "learning_rate": 8.79463319744677e-06, "loss": 0.8948, "step": 16596 }, { "epoch": 0.19, "learning_rate": 8.793563182181715e-06, "loss": 0.863, "step": 16597 }, { "epoch": 0.19, "learning_rate": 8.792493180933378e-06, "loss": 0.9452, "step": 16598 }, { "epoch": 0.19, "learning_rate": 8.791423193714199e-06, "loss": 0.8427, "step": 16599 }, { "epoch": 0.19, "learning_rate": 8.790353220536598e-06, "loss": 0.8677, "step": 16600 }, { "epoch": 0.19, "learning_rate": 8.789283261413008e-06, "loss": 0.8515, "step": 16601 }, { "epoch": 0.19, "learning_rate": 8.788213316355868e-06, "loss": 0.8903, "step": 16602 }, { "epoch": 0.19, "learning_rate": 8.787143385377608e-06, "loss": 0.8813, "step": 16603 }, { "epoch": 0.19, "learning_rate": 8.786073468490648e-06, "loss": 0.8612, "step": 16604 }, { "epoch": 0.19, "learning_rate": 8.78500356570743e-06, "loss": 0.8946, "step": 16605 }, { "epoch": 0.19, "learning_rate": 8.783933677040383e-06, "loss": 0.8459, "step": 16606 }, { "epoch": 0.19, "learning_rate": 8.782863802501929e-06, "loss": 0.8865, "step": 16607 }, { "epoch": 0.19, "learning_rate": 8.781793942104505e-06, "loss": 0.9103, "step": 16608 }, { "epoch": 0.19, "learning_rate": 8.78072409586054e-06, "loss": 0.8752, "step": 16609 }, { "epoch": 0.19, "learning_rate": 8.779654263782466e-06, "loss": 0.867, "step": 16610 }, { "epoch": 0.19, "learning_rate": 8.778584445882708e-06, "loss": 0.8757, "step": 16611 }, { "epoch": 0.19, "learning_rate": 8.777514642173696e-06, "loss": 0.8753, "step": 16612 }, { "epoch": 0.19, "learning_rate": 8.776444852667864e-06, "loss": 0.9028, "step": 16613 }, { "epoch": 0.19, "learning_rate": 8.775375077377634e-06, "loss": 0.8931, "step": 16614 }, { "epoch": 0.19, "learning_rate": 8.774305316315442e-06, "loss": 0.8993, "step": 16615 }, { "epoch": 0.19, "learning_rate": 8.773235569493712e-06, "loss": 0.8753, "step": 16616 }, { "epoch": 0.19, "learning_rate": 8.772165836924874e-06, "loss": 0.8515, "step": 16617 }, { "epoch": 0.19, "learning_rate": 8.771096118621357e-06, "loss": 0.9224, "step": 16618 }, { "epoch": 0.19, "learning_rate": 8.770026414595589e-06, "loss": 0.9015, "step": 16619 }, { "epoch": 0.19, "learning_rate": 8.768956724859998e-06, "loss": 0.8633, "step": 16620 }, { "epoch": 0.19, "learning_rate": 8.767887049427012e-06, "loss": 0.8691, "step": 16621 }, { "epoch": 0.19, "learning_rate": 8.766817388309058e-06, "loss": 0.8823, "step": 16622 }, { "epoch": 0.19, "learning_rate": 8.765747741518563e-06, "loss": 0.8722, "step": 16623 }, { "epoch": 0.19, "learning_rate": 8.76467810906796e-06, "loss": 0.9082, "step": 16624 }, { "epoch": 0.19, "learning_rate": 8.76360849096967e-06, "loss": 0.849, "step": 16625 }, { "epoch": 0.19, "learning_rate": 8.76253888723612e-06, "loss": 0.8635, "step": 16626 }, { "epoch": 0.19, "learning_rate": 8.761469297879745e-06, "loss": 0.888, "step": 16627 }, { "epoch": 0.19, "learning_rate": 8.76039972291296e-06, "loss": 0.859, "step": 16628 }, { "epoch": 0.19, "learning_rate": 8.7593301623482e-06, "loss": 0.9046, "step": 16629 }, { "epoch": 0.19, "learning_rate": 8.758260616197886e-06, "loss": 0.9143, "step": 16630 }, { "epoch": 0.19, "learning_rate": 8.757191084474452e-06, "loss": 0.8696, "step": 16631 }, { "epoch": 0.19, "learning_rate": 8.756121567190317e-06, "loss": 0.9069, "step": 16632 }, { "epoch": 0.19, "learning_rate": 8.755052064357911e-06, "loss": 0.9167, "step": 16633 }, { "epoch": 0.19, "learning_rate": 8.753982575989658e-06, "loss": 0.9097, "step": 16634 }, { "epoch": 0.19, "learning_rate": 8.752913102097985e-06, "loss": 0.9067, "step": 16635 }, { "epoch": 0.19, "learning_rate": 8.751843642695313e-06, "loss": 0.8715, "step": 16636 }, { "epoch": 0.19, "learning_rate": 8.75077419779407e-06, "loss": 0.9073, "step": 16637 }, { "epoch": 0.19, "learning_rate": 8.749704767406686e-06, "loss": 0.9099, "step": 16638 }, { "epoch": 0.19, "learning_rate": 8.748635351545579e-06, "loss": 0.8847, "step": 16639 }, { "epoch": 0.19, "learning_rate": 8.747565950223177e-06, "loss": 0.8623, "step": 16640 }, { "epoch": 0.19, "learning_rate": 8.746496563451903e-06, "loss": 0.8759, "step": 16641 }, { "epoch": 0.19, "learning_rate": 8.745427191244185e-06, "loss": 0.817, "step": 16642 }, { "epoch": 0.19, "learning_rate": 8.744357833612442e-06, "loss": 0.9461, "step": 16643 }, { "epoch": 0.19, "learning_rate": 8.743288490569098e-06, "loss": 0.832, "step": 16644 }, { "epoch": 0.19, "learning_rate": 8.742219162126588e-06, "loss": 0.9263, "step": 16645 }, { "epoch": 0.19, "learning_rate": 8.741149848297321e-06, "loss": 0.9146, "step": 16646 }, { "epoch": 0.19, "learning_rate": 8.740080549093726e-06, "loss": 0.9197, "step": 16647 }, { "epoch": 0.19, "learning_rate": 8.739011264528229e-06, "loss": 0.8936, "step": 16648 }, { "epoch": 0.19, "learning_rate": 8.737941994613256e-06, "loss": 0.9036, "step": 16649 }, { "epoch": 0.19, "learning_rate": 8.736872739361222e-06, "loss": 0.9316, "step": 16650 }, { "epoch": 0.19, "learning_rate": 8.735803498784553e-06, "loss": 0.8781, "step": 16651 }, { "epoch": 0.19, "learning_rate": 8.734734272895678e-06, "loss": 0.9361, "step": 16652 }, { "epoch": 0.19, "learning_rate": 8.733665061707008e-06, "loss": 0.8946, "step": 16653 }, { "epoch": 0.19, "learning_rate": 8.732595865230972e-06, "loss": 0.8824, "step": 16654 }, { "epoch": 0.19, "learning_rate": 8.731526683479991e-06, "loss": 0.8753, "step": 16655 }, { "epoch": 0.19, "learning_rate": 8.730457516466493e-06, "loss": 0.8641, "step": 16656 }, { "epoch": 0.19, "learning_rate": 8.729388364202888e-06, "loss": 0.8596, "step": 16657 }, { "epoch": 0.19, "learning_rate": 8.728319226701608e-06, "loss": 0.8477, "step": 16658 }, { "epoch": 0.19, "learning_rate": 8.727250103975072e-06, "loss": 0.8968, "step": 16659 }, { "epoch": 0.19, "learning_rate": 8.726180996035697e-06, "loss": 0.9206, "step": 16660 }, { "epoch": 0.19, "learning_rate": 8.725111902895907e-06, "loss": 0.8584, "step": 16661 }, { "epoch": 0.19, "learning_rate": 8.724042824568125e-06, "loss": 0.8509, "step": 16662 }, { "epoch": 0.19, "learning_rate": 8.722973761064772e-06, "loss": 0.8675, "step": 16663 }, { "epoch": 0.19, "learning_rate": 8.721904712398264e-06, "loss": 0.8937, "step": 16664 }, { "epoch": 0.19, "learning_rate": 8.720835678581025e-06, "loss": 0.8544, "step": 16665 }, { "epoch": 0.19, "learning_rate": 8.719766659625477e-06, "loss": 0.8788, "step": 16666 }, { "epoch": 0.19, "learning_rate": 8.718697655544035e-06, "loss": 0.8781, "step": 16667 }, { "epoch": 0.19, "learning_rate": 8.717628666349123e-06, "loss": 0.8358, "step": 16668 }, { "epoch": 0.19, "learning_rate": 8.716559692053157e-06, "loss": 0.9179, "step": 16669 }, { "epoch": 0.19, "learning_rate": 8.715490732668565e-06, "loss": 0.917, "step": 16670 }, { "epoch": 0.19, "learning_rate": 8.714421788207756e-06, "loss": 0.8742, "step": 16671 }, { "epoch": 0.19, "learning_rate": 8.713352858683154e-06, "loss": 0.8727, "step": 16672 }, { "epoch": 0.19, "learning_rate": 8.712283944107185e-06, "loss": 0.8475, "step": 16673 }, { "epoch": 0.19, "learning_rate": 8.711215044492254e-06, "loss": 0.8796, "step": 16674 }, { "epoch": 0.19, "learning_rate": 8.710146159850788e-06, "loss": 0.8546, "step": 16675 }, { "epoch": 0.19, "learning_rate": 8.709077290195205e-06, "loss": 0.8979, "step": 16676 }, { "epoch": 0.19, "learning_rate": 8.708008435537927e-06, "loss": 0.9093, "step": 16677 }, { "epoch": 0.19, "learning_rate": 8.706939595891364e-06, "loss": 0.8584, "step": 16678 }, { "epoch": 0.19, "learning_rate": 8.705870771267938e-06, "loss": 0.9224, "step": 16679 }, { "epoch": 0.19, "learning_rate": 8.70480196168007e-06, "loss": 0.8969, "step": 16680 }, { "epoch": 0.19, "learning_rate": 8.703733167140173e-06, "loss": 0.9104, "step": 16681 }, { "epoch": 0.19, "learning_rate": 8.702664387660666e-06, "loss": 0.8605, "step": 16682 }, { "epoch": 0.19, "learning_rate": 8.701595623253969e-06, "loss": 0.8663, "step": 16683 }, { "epoch": 0.19, "learning_rate": 8.700526873932496e-06, "loss": 0.9369, "step": 16684 }, { "epoch": 0.19, "learning_rate": 8.699458139708666e-06, "loss": 0.8793, "step": 16685 }, { "epoch": 0.19, "learning_rate": 8.698389420594892e-06, "loss": 0.8844, "step": 16686 }, { "epoch": 0.19, "learning_rate": 8.697320716603594e-06, "loss": 0.8942, "step": 16687 }, { "epoch": 0.19, "learning_rate": 8.696252027747194e-06, "loss": 0.8668, "step": 16688 }, { "epoch": 0.19, "learning_rate": 8.695183354038098e-06, "loss": 0.8785, "step": 16689 }, { "epoch": 0.19, "learning_rate": 8.694114695488723e-06, "loss": 0.8526, "step": 16690 }, { "epoch": 0.19, "learning_rate": 8.693046052111497e-06, "loss": 0.8947, "step": 16691 }, { "epoch": 0.19, "learning_rate": 8.691977423918822e-06, "loss": 0.8783, "step": 16692 }, { "epoch": 0.19, "learning_rate": 8.690908810923118e-06, "loss": 0.8796, "step": 16693 }, { "epoch": 0.19, "learning_rate": 8.689840213136802e-06, "loss": 0.8677, "step": 16694 }, { "epoch": 0.19, "learning_rate": 8.688771630572295e-06, "loss": 0.8808, "step": 16695 }, { "epoch": 0.19, "learning_rate": 8.687703063241999e-06, "loss": 0.8501, "step": 16696 }, { "epoch": 0.19, "learning_rate": 8.686634511158336e-06, "loss": 0.8636, "step": 16697 }, { "epoch": 0.19, "learning_rate": 8.685565974333727e-06, "loss": 0.8729, "step": 16698 }, { "epoch": 0.19, "learning_rate": 8.68449745278057e-06, "loss": 0.8563, "step": 16699 }, { "epoch": 0.19, "learning_rate": 8.683428946511296e-06, "loss": 0.8801, "step": 16700 }, { "epoch": 0.19, "learning_rate": 8.682360455538311e-06, "loss": 0.8743, "step": 16701 }, { "epoch": 0.19, "learning_rate": 8.681291979874032e-06, "loss": 0.885, "step": 16702 }, { "epoch": 0.19, "learning_rate": 8.68022351953087e-06, "loss": 0.9108, "step": 16703 }, { "epoch": 0.19, "learning_rate": 8.679155074521239e-06, "loss": 0.8708, "step": 16704 }, { "epoch": 0.19, "learning_rate": 8.678086644857558e-06, "loss": 0.9115, "step": 16705 }, { "epoch": 0.19, "learning_rate": 8.677018230552232e-06, "loss": 0.8764, "step": 16706 }, { "epoch": 0.19, "learning_rate": 8.675949831617678e-06, "loss": 0.8536, "step": 16707 }, { "epoch": 0.19, "learning_rate": 8.67488144806631e-06, "loss": 0.8925, "step": 16708 }, { "epoch": 0.19, "learning_rate": 8.673813079910542e-06, "loss": 0.8305, "step": 16709 }, { "epoch": 0.19, "learning_rate": 8.672744727162782e-06, "loss": 0.8669, "step": 16710 }, { "epoch": 0.19, "learning_rate": 8.671676389835445e-06, "loss": 0.8555, "step": 16711 }, { "epoch": 0.19, "learning_rate": 8.670608067940946e-06, "loss": 0.828, "step": 16712 }, { "epoch": 0.19, "learning_rate": 8.66953976149169e-06, "loss": 0.87, "step": 16713 }, { "epoch": 0.19, "learning_rate": 8.668471470500094e-06, "loss": 0.8593, "step": 16714 }, { "epoch": 0.19, "learning_rate": 8.667403194978568e-06, "loss": 0.9027, "step": 16715 }, { "epoch": 0.19, "learning_rate": 8.66633493493953e-06, "loss": 0.8591, "step": 16716 }, { "epoch": 0.19, "learning_rate": 8.665266690395378e-06, "loss": 0.8571, "step": 16717 }, { "epoch": 0.19, "learning_rate": 8.664198461358533e-06, "loss": 0.9169, "step": 16718 }, { "epoch": 0.19, "learning_rate": 8.663130247841409e-06, "loss": 0.896, "step": 16719 }, { "epoch": 0.19, "learning_rate": 8.662062049856406e-06, "loss": 0.8817, "step": 16720 }, { "epoch": 0.19, "learning_rate": 8.66099386741594e-06, "loss": 0.86, "step": 16721 }, { "epoch": 0.19, "learning_rate": 8.659925700532424e-06, "loss": 0.8857, "step": 16722 }, { "epoch": 0.19, "learning_rate": 8.658857549218265e-06, "loss": 0.8917, "step": 16723 }, { "epoch": 0.19, "learning_rate": 8.657789413485876e-06, "loss": 0.9044, "step": 16724 }, { "epoch": 0.19, "learning_rate": 8.656721293347661e-06, "loss": 0.8551, "step": 16725 }, { "epoch": 0.19, "learning_rate": 8.655653188816038e-06, "loss": 0.8634, "step": 16726 }, { "epoch": 0.19, "learning_rate": 8.654585099903408e-06, "loss": 0.8671, "step": 16727 }, { "epoch": 0.19, "learning_rate": 8.653517026622187e-06, "loss": 0.9289, "step": 16728 }, { "epoch": 0.19, "learning_rate": 8.65244896898478e-06, "loss": 0.8989, "step": 16729 }, { "epoch": 0.19, "learning_rate": 8.651380927003601e-06, "loss": 0.9005, "step": 16730 }, { "epoch": 0.19, "learning_rate": 8.650312900691052e-06, "loss": 0.8952, "step": 16731 }, { "epoch": 0.19, "learning_rate": 8.649244890059547e-06, "loss": 0.8776, "step": 16732 }, { "epoch": 0.19, "learning_rate": 8.648176895121487e-06, "loss": 0.8741, "step": 16733 }, { "epoch": 0.19, "learning_rate": 8.647108915889295e-06, "loss": 0.8494, "step": 16734 }, { "epoch": 0.19, "learning_rate": 8.646040952375365e-06, "loss": 0.8879, "step": 16735 }, { "epoch": 0.19, "learning_rate": 8.644973004592108e-06, "loss": 0.8848, "step": 16736 }, { "epoch": 0.19, "learning_rate": 8.643905072551939e-06, "loss": 0.9053, "step": 16737 }, { "epoch": 0.19, "learning_rate": 8.642837156267256e-06, "loss": 0.854, "step": 16738 }, { "epoch": 0.19, "learning_rate": 8.641769255750468e-06, "loss": 0.879, "step": 16739 }, { "epoch": 0.19, "learning_rate": 8.640701371013988e-06, "loss": 0.8828, "step": 16740 }, { "epoch": 0.19, "learning_rate": 8.639633502070222e-06, "loss": 0.8614, "step": 16741 }, { "epoch": 0.19, "learning_rate": 8.63856564893157e-06, "loss": 0.9264, "step": 16742 }, { "epoch": 0.19, "learning_rate": 8.637497811610443e-06, "loss": 0.8985, "step": 16743 }, { "epoch": 0.19, "learning_rate": 8.636429990119254e-06, "loss": 0.8956, "step": 16744 }, { "epoch": 0.19, "learning_rate": 8.635362184470394e-06, "loss": 0.8889, "step": 16745 }, { "epoch": 0.19, "learning_rate": 8.63429439467628e-06, "loss": 0.8855, "step": 16746 }, { "epoch": 0.19, "learning_rate": 8.633226620749317e-06, "loss": 0.8704, "step": 16747 }, { "epoch": 0.19, "learning_rate": 8.632158862701912e-06, "loss": 0.9133, "step": 16748 }, { "epoch": 0.19, "learning_rate": 8.631091120546463e-06, "loss": 0.869, "step": 16749 }, { "epoch": 0.19, "learning_rate": 8.630023394295382e-06, "loss": 0.8802, "step": 16750 }, { "epoch": 0.19, "learning_rate": 8.628955683961073e-06, "loss": 0.8767, "step": 16751 }, { "epoch": 0.19, "learning_rate": 8.62788798955594e-06, "loss": 0.8589, "step": 16752 }, { "epoch": 0.19, "learning_rate": 8.626820311092386e-06, "loss": 0.8829, "step": 16753 }, { "epoch": 0.19, "learning_rate": 8.62575264858282e-06, "loss": 0.8872, "step": 16754 }, { "epoch": 0.19, "learning_rate": 8.624685002039643e-06, "loss": 0.8645, "step": 16755 }, { "epoch": 0.19, "learning_rate": 8.62361737147526e-06, "loss": 0.8833, "step": 16756 }, { "epoch": 0.19, "learning_rate": 8.622549756902074e-06, "loss": 0.8882, "step": 16757 }, { "epoch": 0.19, "learning_rate": 8.621482158332494e-06, "loss": 0.8836, "step": 16758 }, { "epoch": 0.19, "learning_rate": 8.620414575778916e-06, "loss": 0.8533, "step": 16759 }, { "epoch": 0.19, "learning_rate": 8.619347009253748e-06, "loss": 0.883, "step": 16760 }, { "epoch": 0.19, "learning_rate": 8.61827945876939e-06, "loss": 0.8604, "step": 16761 }, { "epoch": 0.19, "learning_rate": 8.617211924338254e-06, "loss": 0.8936, "step": 16762 }, { "epoch": 0.19, "learning_rate": 8.616144405972731e-06, "loss": 0.8828, "step": 16763 }, { "epoch": 0.19, "learning_rate": 8.61507690368523e-06, "loss": 0.8633, "step": 16764 }, { "epoch": 0.19, "learning_rate": 8.614009417488158e-06, "loss": 0.8678, "step": 16765 }, { "epoch": 0.19, "learning_rate": 8.612941947393904e-06, "loss": 0.8459, "step": 16766 }, { "epoch": 0.19, "learning_rate": 8.611874493414883e-06, "loss": 0.9268, "step": 16767 }, { "epoch": 0.19, "learning_rate": 8.610807055563491e-06, "loss": 0.9093, "step": 16768 }, { "epoch": 0.19, "learning_rate": 8.609739633852132e-06, "loss": 0.8419, "step": 16769 }, { "epoch": 0.19, "learning_rate": 8.608672228293207e-06, "loss": 0.8599, "step": 16770 }, { "epoch": 0.19, "learning_rate": 8.607604838899115e-06, "loss": 0.8826, "step": 16771 }, { "epoch": 0.19, "learning_rate": 8.606537465682263e-06, "loss": 0.8953, "step": 16772 }, { "epoch": 0.19, "learning_rate": 8.605470108655046e-06, "loss": 0.8735, "step": 16773 }, { "epoch": 0.19, "learning_rate": 8.604402767829865e-06, "loss": 0.8916, "step": 16774 }, { "epoch": 0.19, "learning_rate": 8.603335443219126e-06, "loss": 0.8871, "step": 16775 }, { "epoch": 0.19, "learning_rate": 8.602268134835226e-06, "loss": 0.8516, "step": 16776 }, { "epoch": 0.19, "learning_rate": 8.601200842690563e-06, "loss": 0.8996, "step": 16777 }, { "epoch": 0.19, "learning_rate": 8.600133566797541e-06, "loss": 0.9094, "step": 16778 }, { "epoch": 0.19, "learning_rate": 8.599066307168557e-06, "loss": 0.8465, "step": 16779 }, { "epoch": 0.19, "learning_rate": 8.597999063816018e-06, "loss": 0.8454, "step": 16780 }, { "epoch": 0.19, "learning_rate": 8.596931836752314e-06, "loss": 0.8606, "step": 16781 }, { "epoch": 0.19, "learning_rate": 8.595864625989847e-06, "loss": 0.8906, "step": 16782 }, { "epoch": 0.19, "learning_rate": 8.594797431541023e-06, "loss": 0.8759, "step": 16783 }, { "epoch": 0.19, "learning_rate": 8.593730253418232e-06, "loss": 0.8388, "step": 16784 }, { "epoch": 0.19, "learning_rate": 8.592663091633871e-06, "loss": 0.8968, "step": 16785 }, { "epoch": 0.19, "learning_rate": 8.591595946200348e-06, "loss": 0.866, "step": 16786 }, { "epoch": 0.19, "learning_rate": 8.590528817130062e-06, "loss": 0.8935, "step": 16787 }, { "epoch": 0.19, "learning_rate": 8.589461704435397e-06, "loss": 0.8497, "step": 16788 }, { "epoch": 0.19, "learning_rate": 8.588394608128766e-06, "loss": 0.8537, "step": 16789 }, { "epoch": 0.19, "learning_rate": 8.58732752822256e-06, "loss": 0.915, "step": 16790 }, { "epoch": 0.19, "learning_rate": 8.586260464729178e-06, "loss": 0.92, "step": 16791 }, { "epoch": 0.19, "learning_rate": 8.585193417661016e-06, "loss": 0.8444, "step": 16792 }, { "epoch": 0.19, "learning_rate": 8.584126387030474e-06, "loss": 0.9065, "step": 16793 }, { "epoch": 0.19, "learning_rate": 8.583059372849948e-06, "loss": 0.9001, "step": 16794 }, { "epoch": 0.19, "learning_rate": 8.581992375131832e-06, "loss": 0.8734, "step": 16795 }, { "epoch": 0.19, "learning_rate": 8.580925393888526e-06, "loss": 0.8908, "step": 16796 }, { "epoch": 0.19, "learning_rate": 8.579858429132428e-06, "loss": 0.8546, "step": 16797 }, { "epoch": 0.19, "learning_rate": 8.578791480875929e-06, "loss": 0.8637, "step": 16798 }, { "epoch": 0.19, "learning_rate": 8.577724549131429e-06, "loss": 0.8813, "step": 16799 }, { "epoch": 0.19, "learning_rate": 8.576657633911322e-06, "loss": 0.9049, "step": 16800 }, { "epoch": 0.19, "learning_rate": 8.575590735228007e-06, "loss": 0.8853, "step": 16801 }, { "epoch": 0.19, "learning_rate": 8.574523853093875e-06, "loss": 0.857, "step": 16802 }, { "epoch": 0.19, "learning_rate": 8.573456987521322e-06, "loss": 0.8957, "step": 16803 }, { "epoch": 0.19, "learning_rate": 8.572390138522749e-06, "loss": 0.8703, "step": 16804 }, { "epoch": 0.19, "learning_rate": 8.571323306110543e-06, "loss": 0.9339, "step": 16805 }, { "epoch": 0.19, "learning_rate": 8.5702564902971e-06, "loss": 0.8795, "step": 16806 }, { "epoch": 0.19, "learning_rate": 8.56918969109482e-06, "loss": 0.8705, "step": 16807 }, { "epoch": 0.19, "learning_rate": 8.568122908516096e-06, "loss": 0.8978, "step": 16808 }, { "epoch": 0.19, "learning_rate": 8.567056142573317e-06, "loss": 0.8969, "step": 16809 }, { "epoch": 0.19, "learning_rate": 8.565989393278882e-06, "loss": 0.8438, "step": 16810 }, { "epoch": 0.19, "learning_rate": 8.564922660645187e-06, "loss": 0.9028, "step": 16811 }, { "epoch": 0.19, "learning_rate": 8.563855944684614e-06, "loss": 0.8312, "step": 16812 }, { "epoch": 0.19, "learning_rate": 8.562789245409569e-06, "loss": 0.8887, "step": 16813 }, { "epoch": 0.19, "learning_rate": 8.561722562832438e-06, "loss": 0.9013, "step": 16814 }, { "epoch": 0.19, "learning_rate": 8.56065589696562e-06, "loss": 0.8053, "step": 16815 }, { "epoch": 0.19, "learning_rate": 8.559589247821501e-06, "loss": 0.8879, "step": 16816 }, { "epoch": 0.19, "learning_rate": 8.558522615412479e-06, "loss": 0.8543, "step": 16817 }, { "epoch": 0.19, "learning_rate": 8.557455999750944e-06, "loss": 0.9023, "step": 16818 }, { "epoch": 0.19, "learning_rate": 8.556389400849287e-06, "loss": 0.8986, "step": 16819 }, { "epoch": 0.19, "learning_rate": 8.555322818719901e-06, "loss": 0.9085, "step": 16820 }, { "epoch": 0.19, "learning_rate": 8.55425625337518e-06, "loss": 0.8659, "step": 16821 }, { "epoch": 0.19, "learning_rate": 8.553189704827516e-06, "loss": 0.8695, "step": 16822 }, { "epoch": 0.19, "learning_rate": 8.552123173089296e-06, "loss": 0.822, "step": 16823 }, { "epoch": 0.19, "learning_rate": 8.551056658172914e-06, "loss": 0.8887, "step": 16824 }, { "epoch": 0.19, "learning_rate": 8.549990160090757e-06, "loss": 0.8712, "step": 16825 }, { "epoch": 0.19, "learning_rate": 8.548923678855228e-06, "loss": 0.884, "step": 16826 }, { "epoch": 0.19, "learning_rate": 8.547857214478706e-06, "loss": 0.8863, "step": 16827 }, { "epoch": 0.19, "learning_rate": 8.54679076697358e-06, "loss": 0.8869, "step": 16828 }, { "epoch": 0.19, "learning_rate": 8.545724336352254e-06, "loss": 0.87, "step": 16829 }, { "epoch": 0.19, "learning_rate": 8.544657922627104e-06, "loss": 0.8479, "step": 16830 }, { "epoch": 0.19, "learning_rate": 8.543591525810524e-06, "loss": 0.8646, "step": 16831 }, { "epoch": 0.19, "learning_rate": 8.542525145914907e-06, "loss": 0.8496, "step": 16832 }, { "epoch": 0.19, "learning_rate": 8.541458782952643e-06, "loss": 0.9231, "step": 16833 }, { "epoch": 0.19, "learning_rate": 8.540392436936112e-06, "loss": 0.8389, "step": 16834 }, { "epoch": 0.19, "learning_rate": 8.539326107877714e-06, "loss": 0.9145, "step": 16835 }, { "epoch": 0.19, "learning_rate": 8.538259795789835e-06, "loss": 0.8696, "step": 16836 }, { "epoch": 0.19, "learning_rate": 8.537193500684859e-06, "loss": 0.8692, "step": 16837 }, { "epoch": 0.19, "learning_rate": 8.53612722257518e-06, "loss": 0.8628, "step": 16838 }, { "epoch": 0.19, "learning_rate": 8.535060961473183e-06, "loss": 0.8728, "step": 16839 }, { "epoch": 0.19, "learning_rate": 8.53399471739126e-06, "loss": 0.8567, "step": 16840 }, { "epoch": 0.19, "learning_rate": 8.532928490341794e-06, "loss": 0.8658, "step": 16841 }, { "epoch": 0.19, "learning_rate": 8.531862280337175e-06, "loss": 0.9073, "step": 16842 }, { "epoch": 0.19, "learning_rate": 8.53079608738979e-06, "loss": 0.8951, "step": 16843 }, { "epoch": 0.19, "learning_rate": 8.529729911512028e-06, "loss": 0.9023, "step": 16844 }, { "epoch": 0.19, "learning_rate": 8.528663752716275e-06, "loss": 0.9037, "step": 16845 }, { "epoch": 0.19, "learning_rate": 8.527597611014917e-06, "loss": 0.8599, "step": 16846 }, { "epoch": 0.19, "learning_rate": 8.526531486420344e-06, "loss": 0.8788, "step": 16847 }, { "epoch": 0.19, "learning_rate": 8.525465378944937e-06, "loss": 0.8743, "step": 16848 }, { "epoch": 0.19, "learning_rate": 8.524399288601086e-06, "loss": 0.8753, "step": 16849 }, { "epoch": 0.19, "learning_rate": 8.52333321540118e-06, "loss": 0.8518, "step": 16850 }, { "epoch": 0.19, "learning_rate": 8.522267159357598e-06, "loss": 0.8772, "step": 16851 }, { "epoch": 0.19, "learning_rate": 8.521201120482729e-06, "loss": 0.9165, "step": 16852 }, { "epoch": 0.19, "learning_rate": 8.52013509878896e-06, "loss": 0.9039, "step": 16853 }, { "epoch": 0.19, "learning_rate": 8.519069094288678e-06, "loss": 0.8945, "step": 16854 }, { "epoch": 0.19, "learning_rate": 8.51800310699426e-06, "loss": 0.8752, "step": 16855 }, { "epoch": 0.19, "learning_rate": 8.5169371369181e-06, "loss": 0.9215, "step": 16856 }, { "epoch": 0.19, "learning_rate": 8.515871184072582e-06, "loss": 0.8976, "step": 16857 }, { "epoch": 0.19, "learning_rate": 8.51480524847008e-06, "loss": 0.8561, "step": 16858 }, { "epoch": 0.19, "learning_rate": 8.51373933012299e-06, "loss": 0.8881, "step": 16859 }, { "epoch": 0.2, "learning_rate": 8.51267342904369e-06, "loss": 0.9092, "step": 16860 }, { "epoch": 0.2, "learning_rate": 8.511607545244569e-06, "loss": 0.903, "step": 16861 }, { "epoch": 0.2, "learning_rate": 8.510541678738007e-06, "loss": 0.8433, "step": 16862 }, { "epoch": 0.2, "learning_rate": 8.509475829536386e-06, "loss": 0.873, "step": 16863 }, { "epoch": 0.2, "learning_rate": 8.508409997652093e-06, "loss": 0.9005, "step": 16864 }, { "epoch": 0.2, "learning_rate": 8.507344183097511e-06, "loss": 0.8855, "step": 16865 }, { "epoch": 0.2, "learning_rate": 8.50627838588502e-06, "loss": 0.9074, "step": 16866 }, { "epoch": 0.2, "learning_rate": 8.505212606027004e-06, "loss": 0.8365, "step": 16867 }, { "epoch": 0.2, "learning_rate": 8.50414684353585e-06, "loss": 0.8685, "step": 16868 }, { "epoch": 0.2, "learning_rate": 8.50308109842393e-06, "loss": 0.8847, "step": 16869 }, { "epoch": 0.2, "learning_rate": 8.502015370703636e-06, "loss": 0.8916, "step": 16870 }, { "epoch": 0.2, "learning_rate": 8.500949660387343e-06, "loss": 0.8566, "step": 16871 }, { "epoch": 0.2, "learning_rate": 8.499883967487441e-06, "loss": 0.8757, "step": 16872 }, { "epoch": 0.2, "learning_rate": 8.498818292016304e-06, "loss": 0.8735, "step": 16873 }, { "epoch": 0.2, "learning_rate": 8.497752633986313e-06, "loss": 0.9207, "step": 16874 }, { "epoch": 0.2, "learning_rate": 8.496686993409859e-06, "loss": 0.9042, "step": 16875 }, { "epoch": 0.2, "learning_rate": 8.495621370299309e-06, "loss": 0.8719, "step": 16876 }, { "epoch": 0.2, "learning_rate": 8.494555764667052e-06, "loss": 0.8768, "step": 16877 }, { "epoch": 0.2, "learning_rate": 8.493490176525468e-06, "loss": 0.8769, "step": 16878 }, { "epoch": 0.2, "learning_rate": 8.492424605886937e-06, "loss": 0.9085, "step": 16879 }, { "epoch": 0.2, "learning_rate": 8.491359052763836e-06, "loss": 0.8595, "step": 16880 }, { "epoch": 0.2, "learning_rate": 8.490293517168551e-06, "loss": 0.8515, "step": 16881 }, { "epoch": 0.2, "learning_rate": 8.489227999113458e-06, "loss": 0.8842, "step": 16882 }, { "epoch": 0.2, "learning_rate": 8.488162498610934e-06, "loss": 0.8645, "step": 16883 }, { "epoch": 0.2, "learning_rate": 8.48709701567336e-06, "loss": 0.8603, "step": 16884 }, { "epoch": 0.2, "learning_rate": 8.486031550313118e-06, "loss": 0.8995, "step": 16885 }, { "epoch": 0.2, "learning_rate": 8.484966102542584e-06, "loss": 0.8463, "step": 16886 }, { "epoch": 0.2, "learning_rate": 8.483900672374139e-06, "loss": 0.9064, "step": 16887 }, { "epoch": 0.2, "learning_rate": 8.482835259820157e-06, "loss": 0.8913, "step": 16888 }, { "epoch": 0.2, "learning_rate": 8.481769864893022e-06, "loss": 0.9247, "step": 16889 }, { "epoch": 0.2, "learning_rate": 8.480704487605108e-06, "loss": 0.8965, "step": 16890 }, { "epoch": 0.2, "learning_rate": 8.479639127968793e-06, "loss": 0.8676, "step": 16891 }, { "epoch": 0.2, "learning_rate": 8.478573785996456e-06, "loss": 0.8777, "step": 16892 }, { "epoch": 0.2, "learning_rate": 8.477508461700477e-06, "loss": 0.8727, "step": 16893 }, { "epoch": 0.2, "learning_rate": 8.476443155093228e-06, "loss": 0.8525, "step": 16894 }, { "epoch": 0.2, "learning_rate": 8.475377866187085e-06, "loss": 0.9353, "step": 16895 }, { "epoch": 0.2, "learning_rate": 8.474312594994435e-06, "loss": 0.8547, "step": 16896 }, { "epoch": 0.2, "learning_rate": 8.473247341527646e-06, "loss": 0.8773, "step": 16897 }, { "epoch": 0.2, "learning_rate": 8.472182105799091e-06, "loss": 0.874, "step": 16898 }, { "epoch": 0.2, "learning_rate": 8.471116887821156e-06, "loss": 0.8945, "step": 16899 }, { "epoch": 0.2, "learning_rate": 8.470051687606215e-06, "loss": 0.8842, "step": 16900 }, { "epoch": 0.2, "learning_rate": 8.468986505166636e-06, "loss": 0.8728, "step": 16901 }, { "epoch": 0.2, "learning_rate": 8.467921340514802e-06, "loss": 0.8911, "step": 16902 }, { "epoch": 0.2, "learning_rate": 8.466856193663089e-06, "loss": 0.8723, "step": 16903 }, { "epoch": 0.2, "learning_rate": 8.465791064623865e-06, "loss": 0.884, "step": 16904 }, { "epoch": 0.2, "learning_rate": 8.46472595340951e-06, "loss": 0.9171, "step": 16905 }, { "epoch": 0.2, "learning_rate": 8.463660860032398e-06, "loss": 0.8455, "step": 16906 }, { "epoch": 0.2, "learning_rate": 8.462595784504908e-06, "loss": 0.9314, "step": 16907 }, { "epoch": 0.2, "learning_rate": 8.461530726839405e-06, "loss": 0.8764, "step": 16908 }, { "epoch": 0.2, "learning_rate": 8.46046568704827e-06, "loss": 0.9004, "step": 16909 }, { "epoch": 0.2, "learning_rate": 8.459400665143876e-06, "loss": 0.8643, "step": 16910 }, { "epoch": 0.2, "learning_rate": 8.458335661138596e-06, "loss": 0.8843, "step": 16911 }, { "epoch": 0.2, "learning_rate": 8.457270675044802e-06, "loss": 0.8704, "step": 16912 }, { "epoch": 0.2, "learning_rate": 8.456205706874869e-06, "loss": 0.872, "step": 16913 }, { "epoch": 0.2, "learning_rate": 8.45514075664117e-06, "loss": 0.8706, "step": 16914 }, { "epoch": 0.2, "learning_rate": 8.454075824356077e-06, "loss": 0.8892, "step": 16915 }, { "epoch": 0.2, "learning_rate": 8.453010910031964e-06, "loss": 0.8795, "step": 16916 }, { "epoch": 0.2, "learning_rate": 8.451946013681199e-06, "loss": 0.8962, "step": 16917 }, { "epoch": 0.2, "learning_rate": 8.450881135316165e-06, "loss": 0.9068, "step": 16918 }, { "epoch": 0.2, "learning_rate": 8.449816274949223e-06, "loss": 0.9289, "step": 16919 }, { "epoch": 0.2, "learning_rate": 8.448751432592747e-06, "loss": 0.8516, "step": 16920 }, { "epoch": 0.2, "learning_rate": 8.447686608259119e-06, "loss": 0.8878, "step": 16921 }, { "epoch": 0.2, "learning_rate": 8.446621801960693e-06, "loss": 0.8529, "step": 16922 }, { "epoch": 0.2, "learning_rate": 8.445557013709852e-06, "loss": 0.882, "step": 16923 }, { "epoch": 0.2, "learning_rate": 8.444492243518962e-06, "loss": 0.8415, "step": 16924 }, { "epoch": 0.2, "learning_rate": 8.443427491400401e-06, "loss": 0.9021, "step": 16925 }, { "epoch": 0.2, "learning_rate": 8.442362757366532e-06, "loss": 0.8636, "step": 16926 }, { "epoch": 0.2, "learning_rate": 8.441298041429728e-06, "loss": 0.879, "step": 16927 }, { "epoch": 0.2, "learning_rate": 8.44023334360236e-06, "loss": 0.8664, "step": 16928 }, { "epoch": 0.2, "learning_rate": 8.439168663896797e-06, "loss": 0.8487, "step": 16929 }, { "epoch": 0.2, "learning_rate": 8.438104002325408e-06, "loss": 0.886, "step": 16930 }, { "epoch": 0.2, "learning_rate": 8.437039358900564e-06, "loss": 0.9231, "step": 16931 }, { "epoch": 0.2, "learning_rate": 8.435974733634635e-06, "loss": 0.8497, "step": 16932 }, { "epoch": 0.2, "learning_rate": 8.434910126539987e-06, "loss": 0.8957, "step": 16933 }, { "epoch": 0.2, "learning_rate": 8.433845537628991e-06, "loss": 0.9006, "step": 16934 }, { "epoch": 0.2, "learning_rate": 8.432780966914017e-06, "loss": 0.927, "step": 16935 }, { "epoch": 0.2, "learning_rate": 8.43171641440743e-06, "loss": 0.8552, "step": 16936 }, { "epoch": 0.2, "learning_rate": 8.430651880121601e-06, "loss": 0.8876, "step": 16937 }, { "epoch": 0.2, "learning_rate": 8.429587364068897e-06, "loss": 0.8884, "step": 16938 }, { "epoch": 0.2, "learning_rate": 8.428522866261687e-06, "loss": 0.9017, "step": 16939 }, { "epoch": 0.2, "learning_rate": 8.427458386712337e-06, "loss": 0.8946, "step": 16940 }, { "epoch": 0.2, "learning_rate": 8.426393925433211e-06, "loss": 0.8883, "step": 16941 }, { "epoch": 0.2, "learning_rate": 8.425329482436688e-06, "loss": 0.8676, "step": 16942 }, { "epoch": 0.2, "learning_rate": 8.424265057735124e-06, "loss": 0.8707, "step": 16943 }, { "epoch": 0.2, "learning_rate": 8.423200651340887e-06, "loss": 0.8609, "step": 16944 }, { "epoch": 0.2, "learning_rate": 8.422136263266346e-06, "loss": 0.8491, "step": 16945 }, { "epoch": 0.2, "learning_rate": 8.421071893523872e-06, "loss": 0.866, "step": 16946 }, { "epoch": 0.2, "learning_rate": 8.420007542125821e-06, "loss": 0.9001, "step": 16947 }, { "epoch": 0.2, "learning_rate": 8.418943209084565e-06, "loss": 0.8867, "step": 16948 }, { "epoch": 0.2, "learning_rate": 8.417878894412475e-06, "loss": 0.8866, "step": 16949 }, { "epoch": 0.2, "learning_rate": 8.416814598121901e-06, "loss": 0.8878, "step": 16950 }, { "epoch": 0.2, "learning_rate": 8.415750320225222e-06, "loss": 0.8447, "step": 16951 }, { "epoch": 0.2, "learning_rate": 8.414686060734798e-06, "loss": 0.8775, "step": 16952 }, { "epoch": 0.2, "learning_rate": 8.413621819662996e-06, "loss": 0.9008, "step": 16953 }, { "epoch": 0.2, "learning_rate": 8.412557597022178e-06, "loss": 0.8669, "step": 16954 }, { "epoch": 0.2, "learning_rate": 8.41149339282471e-06, "loss": 0.866, "step": 16955 }, { "epoch": 0.2, "learning_rate": 8.410429207082955e-06, "loss": 0.8742, "step": 16956 }, { "epoch": 0.2, "learning_rate": 8.409365039809282e-06, "loss": 0.8644, "step": 16957 }, { "epoch": 0.2, "learning_rate": 8.408300891016047e-06, "loss": 0.8973, "step": 16958 }, { "epoch": 0.2, "learning_rate": 8.407236760715618e-06, "loss": 0.8762, "step": 16959 }, { "epoch": 0.2, "learning_rate": 8.406172648920358e-06, "loss": 0.8543, "step": 16960 }, { "epoch": 0.2, "learning_rate": 8.40510855564263e-06, "loss": 0.9204, "step": 16961 }, { "epoch": 0.2, "learning_rate": 8.404044480894795e-06, "loss": 0.8399, "step": 16962 }, { "epoch": 0.2, "learning_rate": 8.402980424689215e-06, "loss": 0.8824, "step": 16963 }, { "epoch": 0.2, "learning_rate": 8.401916387038263e-06, "loss": 0.8757, "step": 16964 }, { "epoch": 0.2, "learning_rate": 8.400852367954289e-06, "loss": 0.8673, "step": 16965 }, { "epoch": 0.2, "learning_rate": 8.399788367449656e-06, "loss": 0.8776, "step": 16966 }, { "epoch": 0.2, "learning_rate": 8.398724385536736e-06, "loss": 0.8685, "step": 16967 }, { "epoch": 0.2, "learning_rate": 8.397660422227877e-06, "loss": 0.8378, "step": 16968 }, { "epoch": 0.2, "learning_rate": 8.39659647753545e-06, "loss": 0.884, "step": 16969 }, { "epoch": 0.2, "learning_rate": 8.395532551471814e-06, "loss": 0.8921, "step": 16970 }, { "epoch": 0.2, "learning_rate": 8.394468644049331e-06, "loss": 0.8693, "step": 16971 }, { "epoch": 0.2, "learning_rate": 8.393404755280357e-06, "loss": 0.849, "step": 16972 }, { "epoch": 0.2, "learning_rate": 8.392340885177257e-06, "loss": 0.8988, "step": 16973 }, { "epoch": 0.2, "learning_rate": 8.391277033752392e-06, "loss": 0.8624, "step": 16974 }, { "epoch": 0.2, "learning_rate": 8.390213201018119e-06, "loss": 0.9142, "step": 16975 }, { "epoch": 0.2, "learning_rate": 8.389149386986797e-06, "loss": 0.8336, "step": 16976 }, { "epoch": 0.2, "learning_rate": 8.38808559167079e-06, "loss": 0.8762, "step": 16977 }, { "epoch": 0.2, "learning_rate": 8.387021815082455e-06, "loss": 0.9245, "step": 16978 }, { "epoch": 0.2, "learning_rate": 8.385958057234151e-06, "loss": 0.899, "step": 16979 }, { "epoch": 0.2, "learning_rate": 8.384894318138237e-06, "loss": 0.8353, "step": 16980 }, { "epoch": 0.2, "learning_rate": 8.383830597807075e-06, "loss": 0.8459, "step": 16981 }, { "epoch": 0.2, "learning_rate": 8.38276689625302e-06, "loss": 0.8782, "step": 16982 }, { "epoch": 0.2, "learning_rate": 8.381703213488428e-06, "loss": 0.9195, "step": 16983 }, { "epoch": 0.2, "learning_rate": 8.380639549525663e-06, "loss": 0.896, "step": 16984 }, { "epoch": 0.2, "learning_rate": 8.379575904377083e-06, "loss": 0.8834, "step": 16985 }, { "epoch": 0.2, "learning_rate": 8.37851227805504e-06, "loss": 0.9087, "step": 16986 }, { "epoch": 0.2, "learning_rate": 8.377448670571893e-06, "loss": 0.829, "step": 16987 }, { "epoch": 0.2, "learning_rate": 8.376385081940008e-06, "loss": 0.9304, "step": 16988 }, { "epoch": 0.2, "learning_rate": 8.37532151217173e-06, "loss": 0.8351, "step": 16989 }, { "epoch": 0.2, "learning_rate": 8.374257961279419e-06, "loss": 0.8628, "step": 16990 }, { "epoch": 0.2, "learning_rate": 8.373194429275435e-06, "loss": 0.8966, "step": 16991 }, { "epoch": 0.2, "learning_rate": 8.372130916172138e-06, "loss": 0.8961, "step": 16992 }, { "epoch": 0.2, "learning_rate": 8.371067421981873e-06, "loss": 0.9065, "step": 16993 }, { "epoch": 0.2, "learning_rate": 8.370003946717005e-06, "loss": 0.8247, "step": 16994 }, { "epoch": 0.2, "learning_rate": 8.368940490389889e-06, "loss": 0.8864, "step": 16995 }, { "epoch": 0.2, "learning_rate": 8.367877053012874e-06, "loss": 0.8596, "step": 16996 }, { "epoch": 0.2, "learning_rate": 8.36681363459832e-06, "loss": 0.9311, "step": 16997 }, { "epoch": 0.2, "learning_rate": 8.365750235158584e-06, "loss": 0.8774, "step": 16998 }, { "epoch": 0.2, "learning_rate": 8.364686854706019e-06, "loss": 0.8866, "step": 16999 }, { "epoch": 0.2, "learning_rate": 8.363623493252978e-06, "loss": 0.8734, "step": 17000 }, { "epoch": 0.2, "learning_rate": 8.362560150811817e-06, "loss": 0.8618, "step": 17001 }, { "epoch": 0.2, "learning_rate": 8.36149682739489e-06, "loss": 0.86, "step": 17002 }, { "epoch": 0.2, "learning_rate": 8.360433523014553e-06, "loss": 0.8926, "step": 17003 }, { "epoch": 0.2, "learning_rate": 8.359370237683155e-06, "loss": 0.8965, "step": 17004 }, { "epoch": 0.2, "learning_rate": 8.358306971413053e-06, "loss": 0.8714, "step": 17005 }, { "epoch": 0.2, "learning_rate": 8.357243724216602e-06, "loss": 0.8892, "step": 17006 }, { "epoch": 0.2, "learning_rate": 8.35618049610615e-06, "loss": 0.911, "step": 17007 }, { "epoch": 0.2, "learning_rate": 8.355117287094052e-06, "loss": 0.8675, "step": 17008 }, { "epoch": 0.2, "learning_rate": 8.35405409719266e-06, "loss": 0.8443, "step": 17009 }, { "epoch": 0.2, "learning_rate": 8.352990926414334e-06, "loss": 0.8564, "step": 17010 }, { "epoch": 0.2, "learning_rate": 8.351927774771416e-06, "loss": 0.9048, "step": 17011 }, { "epoch": 0.2, "learning_rate": 8.35086464227626e-06, "loss": 0.8911, "step": 17012 }, { "epoch": 0.2, "learning_rate": 8.349801528941225e-06, "loss": 0.8581, "step": 17013 }, { "epoch": 0.2, "learning_rate": 8.348738434778652e-06, "loss": 0.8404, "step": 17014 }, { "epoch": 0.2, "learning_rate": 8.347675359800898e-06, "loss": 0.8793, "step": 17015 }, { "epoch": 0.2, "learning_rate": 8.346612304020313e-06, "loss": 0.8559, "step": 17016 }, { "epoch": 0.2, "learning_rate": 8.345549267449253e-06, "loss": 0.9432, "step": 17017 }, { "epoch": 0.2, "learning_rate": 8.344486250100062e-06, "loss": 0.8928, "step": 17018 }, { "epoch": 0.2, "learning_rate": 8.343423251985092e-06, "loss": 0.8954, "step": 17019 }, { "epoch": 0.2, "learning_rate": 8.342360273116696e-06, "loss": 0.8646, "step": 17020 }, { "epoch": 0.2, "learning_rate": 8.34129731350722e-06, "loss": 0.8935, "step": 17021 }, { "epoch": 0.2, "learning_rate": 8.340234373169016e-06, "loss": 0.833, "step": 17022 }, { "epoch": 0.2, "learning_rate": 8.339171452114432e-06, "loss": 0.8996, "step": 17023 }, { "epoch": 0.2, "learning_rate": 8.338108550355821e-06, "loss": 0.874, "step": 17024 }, { "epoch": 0.2, "learning_rate": 8.337045667905528e-06, "loss": 0.877, "step": 17025 }, { "epoch": 0.2, "learning_rate": 8.335982804775902e-06, "loss": 0.8345, "step": 17026 }, { "epoch": 0.2, "learning_rate": 8.334919960979296e-06, "loss": 0.8993, "step": 17027 }, { "epoch": 0.2, "learning_rate": 8.333857136528054e-06, "loss": 0.8938, "step": 17028 }, { "epoch": 0.2, "learning_rate": 8.332794331434524e-06, "loss": 0.8668, "step": 17029 }, { "epoch": 0.2, "learning_rate": 8.331731545711054e-06, "loss": 0.8995, "step": 17030 }, { "epoch": 0.2, "learning_rate": 8.330668779370001e-06, "loss": 0.8671, "step": 17031 }, { "epoch": 0.2, "learning_rate": 8.3296060324237e-06, "loss": 0.866, "step": 17032 }, { "epoch": 0.2, "learning_rate": 8.3285433048845e-06, "loss": 0.8784, "step": 17033 }, { "epoch": 0.2, "learning_rate": 8.327480596764759e-06, "loss": 0.8396, "step": 17034 }, { "epoch": 0.2, "learning_rate": 8.326417908076811e-06, "loss": 0.8976, "step": 17035 }, { "epoch": 0.2, "learning_rate": 8.325355238833006e-06, "loss": 0.8688, "step": 17036 }, { "epoch": 0.2, "learning_rate": 8.324292589045694e-06, "loss": 0.857, "step": 17037 }, { "epoch": 0.2, "learning_rate": 8.323229958727224e-06, "loss": 0.8799, "step": 17038 }, { "epoch": 0.2, "learning_rate": 8.32216734788993e-06, "loss": 0.877, "step": 17039 }, { "epoch": 0.2, "learning_rate": 8.321104756546167e-06, "loss": 0.8673, "step": 17040 }, { "epoch": 0.2, "learning_rate": 8.320042184708282e-06, "loss": 0.8838, "step": 17041 }, { "epoch": 0.2, "learning_rate": 8.318979632388612e-06, "loss": 0.8725, "step": 17042 }, { "epoch": 0.2, "learning_rate": 8.317917099599508e-06, "loss": 0.8486, "step": 17043 }, { "epoch": 0.2, "learning_rate": 8.316854586353312e-06, "loss": 0.8837, "step": 17044 }, { "epoch": 0.2, "learning_rate": 8.315792092662374e-06, "loss": 0.9104, "step": 17045 }, { "epoch": 0.2, "learning_rate": 8.314729618539033e-06, "loss": 0.8743, "step": 17046 }, { "epoch": 0.2, "learning_rate": 8.313667163995632e-06, "loss": 0.8551, "step": 17047 }, { "epoch": 0.2, "learning_rate": 8.312604729044519e-06, "loss": 0.9024, "step": 17048 }, { "epoch": 0.2, "learning_rate": 8.311542313698039e-06, "loss": 0.8991, "step": 17049 }, { "epoch": 0.2, "learning_rate": 8.310479917968528e-06, "loss": 0.8499, "step": 17050 }, { "epoch": 0.2, "learning_rate": 8.309417541868334e-06, "loss": 0.8615, "step": 17051 }, { "epoch": 0.2, "learning_rate": 8.308355185409803e-06, "loss": 0.8657, "step": 17052 }, { "epoch": 0.2, "learning_rate": 8.307292848605271e-06, "loss": 0.8714, "step": 17053 }, { "epoch": 0.2, "learning_rate": 8.306230531467086e-06, "loss": 0.8819, "step": 17054 }, { "epoch": 0.2, "learning_rate": 8.305168234007584e-06, "loss": 0.8697, "step": 17055 }, { "epoch": 0.2, "learning_rate": 8.30410595623912e-06, "loss": 0.894, "step": 17056 }, { "epoch": 0.2, "learning_rate": 8.303043698174021e-06, "loss": 0.9044, "step": 17057 }, { "epoch": 0.2, "learning_rate": 8.301981459824633e-06, "loss": 0.8327, "step": 17058 }, { "epoch": 0.2, "learning_rate": 8.300919241203307e-06, "loss": 0.8915, "step": 17059 }, { "epoch": 0.2, "learning_rate": 8.29985704232237e-06, "loss": 0.9091, "step": 17060 }, { "epoch": 0.2, "learning_rate": 8.298794863194168e-06, "loss": 0.8458, "step": 17061 }, { "epoch": 0.2, "learning_rate": 8.297732703831046e-06, "loss": 0.9007, "step": 17062 }, { "epoch": 0.2, "learning_rate": 8.296670564245342e-06, "loss": 0.859, "step": 17063 }, { "epoch": 0.2, "learning_rate": 8.295608444449395e-06, "loss": 0.8549, "step": 17064 }, { "epoch": 0.2, "learning_rate": 8.294546344455546e-06, "loss": 0.8775, "step": 17065 }, { "epoch": 0.2, "learning_rate": 8.293484264276135e-06, "loss": 0.8883, "step": 17066 }, { "epoch": 0.2, "learning_rate": 8.292422203923498e-06, "loss": 0.8829, "step": 17067 }, { "epoch": 0.2, "learning_rate": 8.291360163409978e-06, "loss": 0.8753, "step": 17068 }, { "epoch": 0.2, "learning_rate": 8.290298142747915e-06, "loss": 0.8883, "step": 17069 }, { "epoch": 0.2, "learning_rate": 8.289236141949647e-06, "loss": 0.8871, "step": 17070 }, { "epoch": 0.2, "learning_rate": 8.28817416102751e-06, "loss": 0.8621, "step": 17071 }, { "epoch": 0.2, "learning_rate": 8.287112199993844e-06, "loss": 0.8528, "step": 17072 }, { "epoch": 0.2, "learning_rate": 8.286050258860989e-06, "loss": 0.8711, "step": 17073 }, { "epoch": 0.2, "learning_rate": 8.28498833764128e-06, "loss": 0.846, "step": 17074 }, { "epoch": 0.2, "learning_rate": 8.283926436347056e-06, "loss": 0.8566, "step": 17075 }, { "epoch": 0.2, "learning_rate": 8.282864554990652e-06, "loss": 0.8715, "step": 17076 }, { "epoch": 0.2, "learning_rate": 8.281802693584416e-06, "loss": 0.877, "step": 17077 }, { "epoch": 0.2, "learning_rate": 8.280740852140669e-06, "loss": 0.8906, "step": 17078 }, { "epoch": 0.2, "learning_rate": 8.279679030671755e-06, "loss": 0.8946, "step": 17079 }, { "epoch": 0.2, "learning_rate": 8.278617229190018e-06, "loss": 0.8664, "step": 17080 }, { "epoch": 0.2, "learning_rate": 8.277555447707784e-06, "loss": 0.8895, "step": 17081 }, { "epoch": 0.2, "learning_rate": 8.276493686237388e-06, "loss": 0.8673, "step": 17082 }, { "epoch": 0.2, "learning_rate": 8.275431944791173e-06, "loss": 0.8556, "step": 17083 }, { "epoch": 0.2, "learning_rate": 8.274370223381477e-06, "loss": 0.8767, "step": 17084 }, { "epoch": 0.2, "learning_rate": 8.273308522020623e-06, "loss": 0.8533, "step": 17085 }, { "epoch": 0.2, "learning_rate": 8.272246840720955e-06, "loss": 0.8543, "step": 17086 }, { "epoch": 0.2, "learning_rate": 8.27118517949481e-06, "loss": 0.8521, "step": 17087 }, { "epoch": 0.2, "learning_rate": 8.270123538354516e-06, "loss": 0.833, "step": 17088 }, { "epoch": 0.2, "learning_rate": 8.26906191731241e-06, "loss": 0.8524, "step": 17089 }, { "epoch": 0.2, "learning_rate": 8.268000316380827e-06, "loss": 0.9086, "step": 17090 }, { "epoch": 0.2, "learning_rate": 8.266938735572103e-06, "loss": 0.8896, "step": 17091 }, { "epoch": 0.2, "learning_rate": 8.265877174898567e-06, "loss": 0.8676, "step": 17092 }, { "epoch": 0.2, "learning_rate": 8.264815634372555e-06, "loss": 0.8727, "step": 17093 }, { "epoch": 0.2, "learning_rate": 8.263754114006399e-06, "loss": 0.8572, "step": 17094 }, { "epoch": 0.2, "learning_rate": 8.262692613812437e-06, "loss": 0.9168, "step": 17095 }, { "epoch": 0.2, "learning_rate": 8.261631133802995e-06, "loss": 0.8969, "step": 17096 }, { "epoch": 0.2, "learning_rate": 8.260569673990408e-06, "loss": 0.908, "step": 17097 }, { "epoch": 0.2, "learning_rate": 8.259508234387011e-06, "loss": 0.8877, "step": 17098 }, { "epoch": 0.2, "learning_rate": 8.258446815005132e-06, "loss": 0.8889, "step": 17099 }, { "epoch": 0.2, "learning_rate": 8.257385415857106e-06, "loss": 0.9245, "step": 17100 }, { "epoch": 0.2, "learning_rate": 8.256324036955261e-06, "loss": 0.8703, "step": 17101 }, { "epoch": 0.2, "learning_rate": 8.255262678311936e-06, "loss": 0.9234, "step": 17102 }, { "epoch": 0.2, "learning_rate": 8.254201339939452e-06, "loss": 0.8874, "step": 17103 }, { "epoch": 0.2, "learning_rate": 8.253140021850147e-06, "loss": 0.8655, "step": 17104 }, { "epoch": 0.2, "learning_rate": 8.252078724056352e-06, "loss": 0.8597, "step": 17105 }, { "epoch": 0.2, "learning_rate": 8.25101744657039e-06, "loss": 0.8373, "step": 17106 }, { "epoch": 0.2, "learning_rate": 8.2499561894046e-06, "loss": 0.8941, "step": 17107 }, { "epoch": 0.2, "learning_rate": 8.248894952571306e-06, "loss": 0.8916, "step": 17108 }, { "epoch": 0.2, "learning_rate": 8.247833736082844e-06, "loss": 0.8882, "step": 17109 }, { "epoch": 0.2, "learning_rate": 8.246772539951536e-06, "loss": 0.8923, "step": 17110 }, { "epoch": 0.2, "learning_rate": 8.245711364189717e-06, "loss": 0.8935, "step": 17111 }, { "epoch": 0.2, "learning_rate": 8.244650208809714e-06, "loss": 0.8494, "step": 17112 }, { "epoch": 0.2, "learning_rate": 8.243589073823853e-06, "loss": 0.9045, "step": 17113 }, { "epoch": 0.2, "learning_rate": 8.242527959244467e-06, "loss": 0.8978, "step": 17114 }, { "epoch": 0.2, "learning_rate": 8.241466865083881e-06, "loss": 0.9209, "step": 17115 }, { "epoch": 0.2, "learning_rate": 8.240405791354428e-06, "loss": 0.8871, "step": 17116 }, { "epoch": 0.2, "learning_rate": 8.23934473806843e-06, "loss": 0.8654, "step": 17117 }, { "epoch": 0.2, "learning_rate": 8.238283705238216e-06, "loss": 0.8809, "step": 17118 }, { "epoch": 0.2, "learning_rate": 8.237222692876119e-06, "loss": 0.8928, "step": 17119 }, { "epoch": 0.2, "learning_rate": 8.236161700994458e-06, "loss": 0.8902, "step": 17120 }, { "epoch": 0.2, "learning_rate": 8.235100729605563e-06, "loss": 0.8757, "step": 17121 }, { "epoch": 0.2, "learning_rate": 8.234039778721759e-06, "loss": 0.8923, "step": 17122 }, { "epoch": 0.2, "learning_rate": 8.232978848355383e-06, "loss": 0.8468, "step": 17123 }, { "epoch": 0.2, "learning_rate": 8.231917938518748e-06, "loss": 0.882, "step": 17124 }, { "epoch": 0.2, "learning_rate": 8.230857049224181e-06, "loss": 0.8821, "step": 17125 }, { "epoch": 0.2, "learning_rate": 8.229796180484018e-06, "loss": 0.8669, "step": 17126 }, { "epoch": 0.2, "learning_rate": 8.228735332310575e-06, "loss": 0.8682, "step": 17127 }, { "epoch": 0.2, "learning_rate": 8.227674504716177e-06, "loss": 0.8712, "step": 17128 }, { "epoch": 0.2, "learning_rate": 8.226613697713155e-06, "loss": 0.8603, "step": 17129 }, { "epoch": 0.2, "learning_rate": 8.225552911313835e-06, "loss": 0.8816, "step": 17130 }, { "epoch": 0.2, "learning_rate": 8.224492145530532e-06, "loss": 0.9118, "step": 17131 }, { "epoch": 0.2, "learning_rate": 8.223431400375576e-06, "loss": 0.8407, "step": 17132 }, { "epoch": 0.2, "learning_rate": 8.222370675861293e-06, "loss": 0.8975, "step": 17133 }, { "epoch": 0.2, "learning_rate": 8.221309972000004e-06, "loss": 0.8947, "step": 17134 }, { "epoch": 0.2, "learning_rate": 8.220249288804032e-06, "loss": 0.8728, "step": 17135 }, { "epoch": 0.2, "learning_rate": 8.219188626285702e-06, "loss": 0.9001, "step": 17136 }, { "epoch": 0.2, "learning_rate": 8.218127984457337e-06, "loss": 0.8617, "step": 17137 }, { "epoch": 0.2, "learning_rate": 8.217067363331258e-06, "loss": 0.8829, "step": 17138 }, { "epoch": 0.2, "learning_rate": 8.21600676291979e-06, "loss": 0.878, "step": 17139 }, { "epoch": 0.2, "learning_rate": 8.214946183235253e-06, "loss": 0.916, "step": 17140 }, { "epoch": 0.2, "learning_rate": 8.213885624289972e-06, "loss": 0.9059, "step": 17141 }, { "epoch": 0.2, "learning_rate": 8.212825086096266e-06, "loss": 0.8656, "step": 17142 }, { "epoch": 0.2, "learning_rate": 8.211764568666458e-06, "loss": 0.8739, "step": 17143 }, { "epoch": 0.2, "learning_rate": 8.210704072012871e-06, "loss": 0.8735, "step": 17144 }, { "epoch": 0.2, "learning_rate": 8.209643596147822e-06, "loss": 0.8879, "step": 17145 }, { "epoch": 0.2, "learning_rate": 8.208583141083635e-06, "loss": 0.9039, "step": 17146 }, { "epoch": 0.2, "learning_rate": 8.207522706832627e-06, "loss": 0.8602, "step": 17147 }, { "epoch": 0.2, "learning_rate": 8.206462293407129e-06, "loss": 0.8805, "step": 17148 }, { "epoch": 0.2, "learning_rate": 8.205401900819445e-06, "loss": 0.8647, "step": 17149 }, { "epoch": 0.2, "learning_rate": 8.204341529081908e-06, "loss": 0.8682, "step": 17150 }, { "epoch": 0.2, "learning_rate": 8.203281178206836e-06, "loss": 0.9318, "step": 17151 }, { "epoch": 0.2, "learning_rate": 8.202220848206539e-06, "loss": 0.9021, "step": 17152 }, { "epoch": 0.2, "learning_rate": 8.201160539093346e-06, "loss": 0.9094, "step": 17153 }, { "epoch": 0.2, "learning_rate": 8.200100250879573e-06, "loss": 0.8996, "step": 17154 }, { "epoch": 0.2, "learning_rate": 8.199039983577539e-06, "loss": 0.8966, "step": 17155 }, { "epoch": 0.2, "learning_rate": 8.197979737199562e-06, "loss": 0.8752, "step": 17156 }, { "epoch": 0.2, "learning_rate": 8.196919511757959e-06, "loss": 0.8826, "step": 17157 }, { "epoch": 0.2, "learning_rate": 8.195859307265052e-06, "loss": 0.8789, "step": 17158 }, { "epoch": 0.2, "learning_rate": 8.194799123733152e-06, "loss": 0.894, "step": 17159 }, { "epoch": 0.21, "learning_rate": 8.193738961174584e-06, "loss": 0.8919, "step": 17160 }, { "epoch": 0.21, "learning_rate": 8.192678819601658e-06, "loss": 0.874, "step": 17161 }, { "epoch": 0.21, "learning_rate": 8.1916186990267e-06, "loss": 0.8623, "step": 17162 }, { "epoch": 0.21, "learning_rate": 8.190558599462018e-06, "loss": 0.8942, "step": 17163 }, { "epoch": 0.21, "learning_rate": 8.189498520919931e-06, "loss": 0.8323, "step": 17164 }, { "epoch": 0.21, "learning_rate": 8.18843846341276e-06, "loss": 0.8852, "step": 17165 }, { "epoch": 0.21, "learning_rate": 8.187378426952816e-06, "loss": 0.8859, "step": 17166 }, { "epoch": 0.21, "learning_rate": 8.186318411552415e-06, "loss": 0.881, "step": 17167 }, { "epoch": 0.21, "learning_rate": 8.185258417223872e-06, "loss": 0.9058, "step": 17168 }, { "epoch": 0.21, "learning_rate": 8.18419844397951e-06, "loss": 0.8685, "step": 17169 }, { "epoch": 0.21, "learning_rate": 8.183138491831636e-06, "loss": 0.8485, "step": 17170 }, { "epoch": 0.21, "learning_rate": 8.182078560792563e-06, "loss": 0.8889, "step": 17171 }, { "epoch": 0.21, "learning_rate": 8.181018650874616e-06, "loss": 0.8329, "step": 17172 }, { "epoch": 0.21, "learning_rate": 8.1799587620901e-06, "loss": 0.8548, "step": 17173 }, { "epoch": 0.21, "learning_rate": 8.178898894451327e-06, "loss": 0.877, "step": 17174 }, { "epoch": 0.21, "learning_rate": 8.17783904797062e-06, "loss": 0.8693, "step": 17175 }, { "epoch": 0.21, "learning_rate": 8.176779222660291e-06, "loss": 0.8946, "step": 17176 }, { "epoch": 0.21, "learning_rate": 8.175719418532648e-06, "loss": 0.9028, "step": 17177 }, { "epoch": 0.21, "learning_rate": 8.174659635600006e-06, "loss": 0.9203, "step": 17178 }, { "epoch": 0.21, "learning_rate": 8.173599873874682e-06, "loss": 0.8792, "step": 17179 }, { "epoch": 0.21, "learning_rate": 8.172540133368984e-06, "loss": 0.8995, "step": 17180 }, { "epoch": 0.21, "learning_rate": 8.171480414095224e-06, "loss": 0.863, "step": 17181 }, { "epoch": 0.21, "learning_rate": 8.170420716065716e-06, "loss": 0.858, "step": 17182 }, { "epoch": 0.21, "learning_rate": 8.169361039292774e-06, "loss": 0.8531, "step": 17183 }, { "epoch": 0.21, "learning_rate": 8.168301383788706e-06, "loss": 0.8587, "step": 17184 }, { "epoch": 0.21, "learning_rate": 8.167241749565824e-06, "loss": 0.8931, "step": 17185 }, { "epoch": 0.21, "learning_rate": 8.16618213663644e-06, "loss": 0.9214, "step": 17186 }, { "epoch": 0.21, "learning_rate": 8.165122545012866e-06, "loss": 0.8725, "step": 17187 }, { "epoch": 0.21, "learning_rate": 8.164062974707409e-06, "loss": 0.8466, "step": 17188 }, { "epoch": 0.21, "learning_rate": 8.163003425732382e-06, "loss": 0.8765, "step": 17189 }, { "epoch": 0.21, "learning_rate": 8.1619438981001e-06, "loss": 0.8655, "step": 17190 }, { "epoch": 0.21, "learning_rate": 8.160884391822862e-06, "loss": 0.8674, "step": 17191 }, { "epoch": 0.21, "learning_rate": 8.159824906912983e-06, "loss": 0.8288, "step": 17192 }, { "epoch": 0.21, "learning_rate": 8.158765443382774e-06, "loss": 0.8793, "step": 17193 }, { "epoch": 0.21, "learning_rate": 8.157706001244547e-06, "loss": 0.8936, "step": 17194 }, { "epoch": 0.21, "learning_rate": 8.156646580510602e-06, "loss": 0.8884, "step": 17195 }, { "epoch": 0.21, "learning_rate": 8.155587181193253e-06, "loss": 0.9114, "step": 17196 }, { "epoch": 0.21, "learning_rate": 8.154527803304812e-06, "loss": 0.8492, "step": 17197 }, { "epoch": 0.21, "learning_rate": 8.153468446857576e-06, "loss": 0.9008, "step": 17198 }, { "epoch": 0.21, "learning_rate": 8.152409111863863e-06, "loss": 0.8523, "step": 17199 }, { "epoch": 0.21, "learning_rate": 8.151349798335976e-06, "loss": 0.8774, "step": 17200 }, { "epoch": 0.21, "learning_rate": 8.15029050628623e-06, "loss": 0.9432, "step": 17201 }, { "epoch": 0.21, "learning_rate": 8.14923123572692e-06, "loss": 0.8786, "step": 17202 }, { "epoch": 0.21, "learning_rate": 8.14817198667036e-06, "loss": 0.8595, "step": 17203 }, { "epoch": 0.21, "learning_rate": 8.147112759128859e-06, "loss": 0.8735, "step": 17204 }, { "epoch": 0.21, "learning_rate": 8.146053553114716e-06, "loss": 0.856, "step": 17205 }, { "epoch": 0.21, "learning_rate": 8.144994368640244e-06, "loss": 0.8463, "step": 17206 }, { "epoch": 0.21, "learning_rate": 8.143935205717744e-06, "loss": 0.9266, "step": 17207 }, { "epoch": 0.21, "learning_rate": 8.142876064359526e-06, "loss": 0.8913, "step": 17208 }, { "epoch": 0.21, "learning_rate": 8.141816944577891e-06, "loss": 0.8875, "step": 17209 }, { "epoch": 0.21, "learning_rate": 8.140757846385147e-06, "loss": 0.8759, "step": 17210 }, { "epoch": 0.21, "learning_rate": 8.139698769793599e-06, "loss": 0.8947, "step": 17211 }, { "epoch": 0.21, "learning_rate": 8.13863971481555e-06, "loss": 0.9044, "step": 17212 }, { "epoch": 0.21, "learning_rate": 8.137580681463305e-06, "loss": 0.8743, "step": 17213 }, { "epoch": 0.21, "learning_rate": 8.136521669749165e-06, "loss": 0.896, "step": 17214 }, { "epoch": 0.21, "learning_rate": 8.135462679685445e-06, "loss": 0.939, "step": 17215 }, { "epoch": 0.21, "learning_rate": 8.134403711284436e-06, "loss": 0.9001, "step": 17216 }, { "epoch": 0.21, "learning_rate": 8.133344764558444e-06, "loss": 0.8602, "step": 17217 }, { "epoch": 0.21, "learning_rate": 8.132285839519783e-06, "loss": 0.8301, "step": 17218 }, { "epoch": 0.21, "learning_rate": 8.131226936180741e-06, "loss": 0.8945, "step": 17219 }, { "epoch": 0.21, "learning_rate": 8.130168054553626e-06, "loss": 0.8646, "step": 17220 }, { "epoch": 0.21, "learning_rate": 8.129109194650742e-06, "loss": 0.886, "step": 17221 }, { "epoch": 0.21, "learning_rate": 8.128050356484394e-06, "loss": 0.9065, "step": 17222 }, { "epoch": 0.21, "learning_rate": 8.126991540066878e-06, "loss": 0.9118, "step": 17223 }, { "epoch": 0.21, "learning_rate": 8.125932745410497e-06, "loss": 0.8353, "step": 17224 }, { "epoch": 0.21, "learning_rate": 8.124873972527557e-06, "loss": 0.8648, "step": 17225 }, { "epoch": 0.21, "learning_rate": 8.123815221430351e-06, "loss": 0.8792, "step": 17226 }, { "epoch": 0.21, "learning_rate": 8.122756492131186e-06, "loss": 0.8285, "step": 17227 }, { "epoch": 0.21, "learning_rate": 8.121697784642361e-06, "loss": 0.8429, "step": 17228 }, { "epoch": 0.21, "learning_rate": 8.120639098976177e-06, "loss": 0.8706, "step": 17229 }, { "epoch": 0.21, "learning_rate": 8.119580435144933e-06, "loss": 0.8883, "step": 17230 }, { "epoch": 0.21, "learning_rate": 8.11852179316093e-06, "loss": 0.9182, "step": 17231 }, { "epoch": 0.21, "learning_rate": 8.117463173036465e-06, "loss": 0.8777, "step": 17232 }, { "epoch": 0.21, "learning_rate": 8.116404574783842e-06, "loss": 0.8405, "step": 17233 }, { "epoch": 0.21, "learning_rate": 8.115345998415355e-06, "loss": 0.9069, "step": 17234 }, { "epoch": 0.21, "learning_rate": 8.114287443943306e-06, "loss": 0.8795, "step": 17235 }, { "epoch": 0.21, "learning_rate": 8.113228911379993e-06, "loss": 0.9024, "step": 17236 }, { "epoch": 0.21, "learning_rate": 8.112170400737713e-06, "loss": 0.8797, "step": 17237 }, { "epoch": 0.21, "learning_rate": 8.111111912028763e-06, "loss": 0.8778, "step": 17238 }, { "epoch": 0.21, "learning_rate": 8.110053445265444e-06, "loss": 0.8784, "step": 17239 }, { "epoch": 0.21, "learning_rate": 8.108995000460059e-06, "loss": 0.8713, "step": 17240 }, { "epoch": 0.21, "learning_rate": 8.10793657762489e-06, "loss": 0.8758, "step": 17241 }, { "epoch": 0.21, "learning_rate": 8.106878176772247e-06, "loss": 0.8993, "step": 17242 }, { "epoch": 0.21, "learning_rate": 8.105819797914426e-06, "loss": 0.8243, "step": 17243 }, { "epoch": 0.21, "learning_rate": 8.104761441063714e-06, "loss": 0.861, "step": 17244 }, { "epoch": 0.21, "learning_rate": 8.103703106232416e-06, "loss": 0.9015, "step": 17245 }, { "epoch": 0.21, "learning_rate": 8.102644793432826e-06, "loss": 0.8858, "step": 17246 }, { "epoch": 0.21, "learning_rate": 8.10158650267724e-06, "loss": 0.8496, "step": 17247 }, { "epoch": 0.21, "learning_rate": 8.10052823397795e-06, "loss": 0.8991, "step": 17248 }, { "epoch": 0.21, "learning_rate": 8.099469987347256e-06, "loss": 0.8587, "step": 17249 }, { "epoch": 0.21, "learning_rate": 8.098411762797452e-06, "loss": 0.8977, "step": 17250 }, { "epoch": 0.21, "learning_rate": 8.09735356034083e-06, "loss": 0.9145, "step": 17251 }, { "epoch": 0.21, "learning_rate": 8.096295379989687e-06, "loss": 0.8884, "step": 17252 }, { "epoch": 0.21, "learning_rate": 8.095237221756317e-06, "loss": 0.8897, "step": 17253 }, { "epoch": 0.21, "learning_rate": 8.094179085653014e-06, "loss": 0.8424, "step": 17254 }, { "epoch": 0.21, "learning_rate": 8.09312097169207e-06, "loss": 0.8886, "step": 17255 }, { "epoch": 0.21, "learning_rate": 8.09206287988578e-06, "loss": 0.9039, "step": 17256 }, { "epoch": 0.21, "learning_rate": 8.091004810246439e-06, "loss": 0.8322, "step": 17257 }, { "epoch": 0.21, "learning_rate": 8.089946762786335e-06, "loss": 0.8916, "step": 17258 }, { "epoch": 0.21, "learning_rate": 8.088888737517764e-06, "loss": 0.8765, "step": 17259 }, { "epoch": 0.21, "learning_rate": 8.087830734453015e-06, "loss": 0.8754, "step": 17260 }, { "epoch": 0.21, "learning_rate": 8.08677275360439e-06, "loss": 0.8677, "step": 17261 }, { "epoch": 0.21, "learning_rate": 8.085714794984171e-06, "loss": 0.8859, "step": 17262 }, { "epoch": 0.21, "learning_rate": 8.08465685860465e-06, "loss": 0.9039, "step": 17263 }, { "epoch": 0.21, "learning_rate": 8.083598944478128e-06, "loss": 0.9126, "step": 17264 }, { "epoch": 0.21, "learning_rate": 8.082541052616882e-06, "loss": 0.8901, "step": 17265 }, { "epoch": 0.21, "learning_rate": 8.081483183033212e-06, "loss": 0.9123, "step": 17266 }, { "epoch": 0.21, "learning_rate": 8.080425335739406e-06, "loss": 0.893, "step": 17267 }, { "epoch": 0.21, "learning_rate": 8.07936751074776e-06, "loss": 0.8826, "step": 17268 }, { "epoch": 0.21, "learning_rate": 8.078309708070555e-06, "loss": 0.9123, "step": 17269 }, { "epoch": 0.21, "learning_rate": 8.077251927720087e-06, "loss": 0.8571, "step": 17270 }, { "epoch": 0.21, "learning_rate": 8.076194169708644e-06, "loss": 0.9119, "step": 17271 }, { "epoch": 0.21, "learning_rate": 8.075136434048514e-06, "loss": 0.8174, "step": 17272 }, { "epoch": 0.21, "learning_rate": 8.074078720751986e-06, "loss": 0.8942, "step": 17273 }, { "epoch": 0.21, "learning_rate": 8.073021029831351e-06, "loss": 0.8895, "step": 17274 }, { "epoch": 0.21, "learning_rate": 8.071963361298899e-06, "loss": 0.8762, "step": 17275 }, { "epoch": 0.21, "learning_rate": 8.070905715166912e-06, "loss": 0.8629, "step": 17276 }, { "epoch": 0.21, "learning_rate": 8.069848091447683e-06, "loss": 0.812, "step": 17277 }, { "epoch": 0.21, "learning_rate": 8.068790490153498e-06, "loss": 0.8767, "step": 17278 }, { "epoch": 0.21, "learning_rate": 8.067732911296649e-06, "loss": 0.8776, "step": 17279 }, { "epoch": 0.21, "learning_rate": 8.066675354889415e-06, "loss": 0.864, "step": 17280 }, { "epoch": 0.21, "learning_rate": 8.065617820944084e-06, "loss": 0.9043, "step": 17281 }, { "epoch": 0.21, "learning_rate": 8.064560309472955e-06, "loss": 0.8514, "step": 17282 }, { "epoch": 0.21, "learning_rate": 8.0635028204883e-06, "loss": 0.8745, "step": 17283 }, { "epoch": 0.21, "learning_rate": 8.06244535400241e-06, "loss": 0.8767, "step": 17284 }, { "epoch": 0.21, "learning_rate": 8.061387910027572e-06, "loss": 0.8891, "step": 17285 }, { "epoch": 0.21, "learning_rate": 8.060330488576077e-06, "loss": 0.8892, "step": 17286 }, { "epoch": 0.21, "learning_rate": 8.059273089660196e-06, "loss": 0.8458, "step": 17287 }, { "epoch": 0.21, "learning_rate": 8.058215713292228e-06, "loss": 0.8687, "step": 17288 }, { "epoch": 0.21, "learning_rate": 8.057158359484455e-06, "loss": 0.8932, "step": 17289 }, { "epoch": 0.21, "learning_rate": 8.056101028249154e-06, "loss": 0.8695, "step": 17290 }, { "epoch": 0.21, "learning_rate": 8.055043719598617e-06, "loss": 0.882, "step": 17291 }, { "epoch": 0.21, "learning_rate": 8.053986433545125e-06, "loss": 0.8353, "step": 17292 }, { "epoch": 0.21, "learning_rate": 8.052929170100965e-06, "loss": 0.8627, "step": 17293 }, { "epoch": 0.21, "learning_rate": 8.051871929278417e-06, "loss": 0.8905, "step": 17294 }, { "epoch": 0.21, "learning_rate": 8.050814711089764e-06, "loss": 0.9197, "step": 17295 }, { "epoch": 0.21, "learning_rate": 8.049757515547294e-06, "loss": 0.9528, "step": 17296 }, { "epoch": 0.21, "learning_rate": 8.048700342663284e-06, "loss": 0.9165, "step": 17297 }, { "epoch": 0.21, "learning_rate": 8.047643192450019e-06, "loss": 0.8682, "step": 17298 }, { "epoch": 0.21, "learning_rate": 8.046586064919782e-06, "loss": 0.8587, "step": 17299 }, { "epoch": 0.21, "learning_rate": 8.045528960084855e-06, "loss": 0.9075, "step": 17300 }, { "epoch": 0.21, "learning_rate": 8.044471877957518e-06, "loss": 0.903, "step": 17301 }, { "epoch": 0.21, "learning_rate": 8.043414818550053e-06, "loss": 0.8638, "step": 17302 }, { "epoch": 0.21, "learning_rate": 8.042357781874745e-06, "loss": 0.8958, "step": 17303 }, { "epoch": 0.21, "learning_rate": 8.041300767943867e-06, "loss": 0.8818, "step": 17304 }, { "epoch": 0.21, "learning_rate": 8.040243776769707e-06, "loss": 0.8514, "step": 17305 }, { "epoch": 0.21, "learning_rate": 8.039186808364539e-06, "loss": 0.8524, "step": 17306 }, { "epoch": 0.21, "learning_rate": 8.038129862740653e-06, "loss": 0.9026, "step": 17307 }, { "epoch": 0.21, "learning_rate": 8.03707293991032e-06, "loss": 0.8779, "step": 17308 }, { "epoch": 0.21, "learning_rate": 8.03601603988582e-06, "loss": 0.863, "step": 17309 }, { "epoch": 0.21, "learning_rate": 8.034959162679442e-06, "loss": 0.8698, "step": 17310 }, { "epoch": 0.21, "learning_rate": 8.03390230830345e-06, "loss": 0.8799, "step": 17311 }, { "epoch": 0.21, "learning_rate": 8.032845476770133e-06, "loss": 0.869, "step": 17312 }, { "epoch": 0.21, "learning_rate": 8.031788668091767e-06, "loss": 0.8753, "step": 17313 }, { "epoch": 0.21, "learning_rate": 8.030731882280632e-06, "loss": 0.8529, "step": 17314 }, { "epoch": 0.21, "learning_rate": 8.029675119349004e-06, "loss": 0.8708, "step": 17315 }, { "epoch": 0.21, "learning_rate": 8.02861837930916e-06, "loss": 0.8725, "step": 17316 }, { "epoch": 0.21, "learning_rate": 8.027561662173383e-06, "loss": 0.8733, "step": 17317 }, { "epoch": 0.21, "learning_rate": 8.026504967953944e-06, "loss": 0.8805, "step": 17318 }, { "epoch": 0.21, "learning_rate": 8.02544829666312e-06, "loss": 0.9068, "step": 17319 }, { "epoch": 0.21, "learning_rate": 8.02439164831319e-06, "loss": 0.8653, "step": 17320 }, { "epoch": 0.21, "learning_rate": 8.023335022916435e-06, "loss": 0.916, "step": 17321 }, { "epoch": 0.21, "learning_rate": 8.022278420485122e-06, "loss": 0.85, "step": 17322 }, { "epoch": 0.21, "learning_rate": 8.02122184103153e-06, "loss": 0.876, "step": 17323 }, { "epoch": 0.21, "learning_rate": 8.020165284567938e-06, "loss": 0.8683, "step": 17324 }, { "epoch": 0.21, "learning_rate": 8.019108751106621e-06, "loss": 0.8734, "step": 17325 }, { "epoch": 0.21, "learning_rate": 8.01805224065985e-06, "loss": 0.9032, "step": 17326 }, { "epoch": 0.21, "learning_rate": 8.0169957532399e-06, "loss": 0.8937, "step": 17327 }, { "epoch": 0.21, "learning_rate": 8.015939288859056e-06, "loss": 0.8716, "step": 17328 }, { "epoch": 0.21, "learning_rate": 8.014882847529577e-06, "loss": 0.8857, "step": 17329 }, { "epoch": 0.21, "learning_rate": 8.013826429263743e-06, "loss": 0.8582, "step": 17330 }, { "epoch": 0.21, "learning_rate": 8.012770034073832e-06, "loss": 0.8434, "step": 17331 }, { "epoch": 0.21, "learning_rate": 8.011713661972119e-06, "loss": 0.9172, "step": 17332 }, { "epoch": 0.21, "learning_rate": 8.010657312970864e-06, "loss": 0.8676, "step": 17333 }, { "epoch": 0.21, "learning_rate": 8.009600987082353e-06, "loss": 0.887, "step": 17334 }, { "epoch": 0.21, "learning_rate": 8.008544684318857e-06, "loss": 0.9256, "step": 17335 }, { "epoch": 0.21, "learning_rate": 8.00748840469264e-06, "loss": 0.8447, "step": 17336 }, { "epoch": 0.21, "learning_rate": 8.006432148215981e-06, "loss": 0.9324, "step": 17337 }, { "epoch": 0.21, "learning_rate": 8.005375914901152e-06, "loss": 0.8639, "step": 17338 }, { "epoch": 0.21, "learning_rate": 8.004319704760424e-06, "loss": 0.9091, "step": 17339 }, { "epoch": 0.21, "learning_rate": 8.003263517806066e-06, "loss": 0.8754, "step": 17340 }, { "epoch": 0.21, "learning_rate": 8.002207354050351e-06, "loss": 0.8711, "step": 17341 }, { "epoch": 0.21, "learning_rate": 8.001151213505552e-06, "loss": 0.9103, "step": 17342 }, { "epoch": 0.21, "learning_rate": 8.000095096183935e-06, "loss": 0.8696, "step": 17343 }, { "epoch": 0.21, "learning_rate": 7.999039002097771e-06, "loss": 0.8843, "step": 17344 }, { "epoch": 0.21, "learning_rate": 7.997982931259333e-06, "loss": 0.8699, "step": 17345 }, { "epoch": 0.21, "learning_rate": 7.99692688368089e-06, "loss": 0.8707, "step": 17346 }, { "epoch": 0.21, "learning_rate": 7.995870859374709e-06, "loss": 0.8713, "step": 17347 }, { "epoch": 0.21, "learning_rate": 7.994814858353062e-06, "loss": 0.8521, "step": 17348 }, { "epoch": 0.21, "learning_rate": 7.993758880628217e-06, "loss": 0.8421, "step": 17349 }, { "epoch": 0.21, "learning_rate": 7.99270292621244e-06, "loss": 0.8867, "step": 17350 }, { "epoch": 0.21, "learning_rate": 7.991646995118004e-06, "loss": 0.8577, "step": 17351 }, { "epoch": 0.21, "learning_rate": 7.990591087357169e-06, "loss": 0.8963, "step": 17352 }, { "epoch": 0.21, "learning_rate": 7.989535202942217e-06, "loss": 0.8516, "step": 17353 }, { "epoch": 0.21, "learning_rate": 7.988479341885402e-06, "loss": 0.8883, "step": 17354 }, { "epoch": 0.21, "learning_rate": 7.987423504198995e-06, "loss": 0.8641, "step": 17355 }, { "epoch": 0.21, "learning_rate": 7.986367689895271e-06, "loss": 0.8614, "step": 17356 }, { "epoch": 0.21, "learning_rate": 7.985311898986484e-06, "loss": 0.8598, "step": 17357 }, { "epoch": 0.21, "learning_rate": 7.98425613148491e-06, "loss": 0.8456, "step": 17358 }, { "epoch": 0.21, "learning_rate": 7.983200387402809e-06, "loss": 0.8888, "step": 17359 }, { "epoch": 0.21, "learning_rate": 7.982144666752453e-06, "loss": 0.842, "step": 17360 }, { "epoch": 0.21, "learning_rate": 7.981088969546103e-06, "loss": 0.8691, "step": 17361 }, { "epoch": 0.21, "learning_rate": 7.980033295796026e-06, "loss": 0.882, "step": 17362 }, { "epoch": 0.21, "learning_rate": 7.978977645514488e-06, "loss": 0.9082, "step": 17363 }, { "epoch": 0.21, "learning_rate": 7.97792201871375e-06, "loss": 0.8902, "step": 17364 }, { "epoch": 0.21, "learning_rate": 7.97686641540608e-06, "loss": 0.9026, "step": 17365 }, { "epoch": 0.21, "learning_rate": 7.975810835603743e-06, "loss": 0.8454, "step": 17366 }, { "epoch": 0.21, "learning_rate": 7.974755279319005e-06, "loss": 0.8661, "step": 17367 }, { "epoch": 0.21, "learning_rate": 7.973699746564122e-06, "loss": 0.8843, "step": 17368 }, { "epoch": 0.21, "learning_rate": 7.972644237351362e-06, "loss": 0.8753, "step": 17369 }, { "epoch": 0.21, "learning_rate": 7.971588751692986e-06, "loss": 0.9119, "step": 17370 }, { "epoch": 0.21, "learning_rate": 7.970533289601264e-06, "loss": 0.8677, "step": 17371 }, { "epoch": 0.21, "learning_rate": 7.969477851088452e-06, "loss": 0.8245, "step": 17372 }, { "epoch": 0.21, "learning_rate": 7.96842243616681e-06, "loss": 0.8872, "step": 17373 }, { "epoch": 0.21, "learning_rate": 7.967367044848611e-06, "loss": 0.8945, "step": 17374 }, { "epoch": 0.21, "learning_rate": 7.966311677146107e-06, "loss": 0.8989, "step": 17375 }, { "epoch": 0.21, "learning_rate": 7.96525633307156e-06, "loss": 0.8857, "step": 17376 }, { "epoch": 0.21, "learning_rate": 7.964201012637235e-06, "loss": 0.8325, "step": 17377 }, { "epoch": 0.21, "learning_rate": 7.963145715855397e-06, "loss": 0.8836, "step": 17378 }, { "epoch": 0.21, "learning_rate": 7.962090442738294e-06, "loss": 0.8776, "step": 17379 }, { "epoch": 0.21, "learning_rate": 7.961035193298199e-06, "loss": 0.8351, "step": 17380 }, { "epoch": 0.21, "learning_rate": 7.95997996754737e-06, "loss": 0.9033, "step": 17381 }, { "epoch": 0.21, "learning_rate": 7.958924765498055e-06, "loss": 0.9035, "step": 17382 }, { "epoch": 0.21, "learning_rate": 7.957869587162528e-06, "loss": 0.8632, "step": 17383 }, { "epoch": 0.21, "learning_rate": 7.956814432553043e-06, "loss": 0.8858, "step": 17384 }, { "epoch": 0.21, "learning_rate": 7.95575930168186e-06, "loss": 0.8793, "step": 17385 }, { "epoch": 0.21, "learning_rate": 7.954704194561236e-06, "loss": 0.8627, "step": 17386 }, { "epoch": 0.21, "learning_rate": 7.95364911120343e-06, "loss": 0.8633, "step": 17387 }, { "epoch": 0.21, "learning_rate": 7.952594051620703e-06, "loss": 0.8973, "step": 17388 }, { "epoch": 0.21, "learning_rate": 7.951539015825308e-06, "loss": 0.8916, "step": 17389 }, { "epoch": 0.21, "learning_rate": 7.950484003829506e-06, "loss": 0.8226, "step": 17390 }, { "epoch": 0.21, "learning_rate": 7.949429015645554e-06, "loss": 0.9032, "step": 17391 }, { "epoch": 0.21, "learning_rate": 7.948374051285711e-06, "loss": 0.9179, "step": 17392 }, { "epoch": 0.21, "learning_rate": 7.94731911076223e-06, "loss": 0.8899, "step": 17393 }, { "epoch": 0.21, "learning_rate": 7.94626419408737e-06, "loss": 0.8892, "step": 17394 }, { "epoch": 0.21, "learning_rate": 7.945209301273388e-06, "loss": 0.9297, "step": 17395 }, { "epoch": 0.21, "learning_rate": 7.944154432332537e-06, "loss": 0.8509, "step": 17396 }, { "epoch": 0.21, "learning_rate": 7.943099587277074e-06, "loss": 0.8882, "step": 17397 }, { "epoch": 0.21, "learning_rate": 7.942044766119254e-06, "loss": 0.869, "step": 17398 }, { "epoch": 0.21, "learning_rate": 7.94098996887134e-06, "loss": 0.8967, "step": 17399 }, { "epoch": 0.21, "learning_rate": 7.939935195545573e-06, "loss": 0.9182, "step": 17400 }, { "epoch": 0.21, "learning_rate": 7.938880446154218e-06, "loss": 0.8537, "step": 17401 }, { "epoch": 0.21, "learning_rate": 7.937825720709529e-06, "loss": 0.9038, "step": 17402 }, { "epoch": 0.21, "learning_rate": 7.936771019223752e-06, "loss": 0.8426, "step": 17403 }, { "epoch": 0.21, "learning_rate": 7.935716341709147e-06, "loss": 0.8719, "step": 17404 }, { "epoch": 0.21, "learning_rate": 7.93466168817797e-06, "loss": 0.8608, "step": 17405 }, { "epoch": 0.21, "learning_rate": 7.93360705864247e-06, "loss": 0.8686, "step": 17406 }, { "epoch": 0.21, "learning_rate": 7.9325524531149e-06, "loss": 0.8421, "step": 17407 }, { "epoch": 0.21, "learning_rate": 7.931497871607513e-06, "loss": 0.9181, "step": 17408 }, { "epoch": 0.21, "learning_rate": 7.930443314132564e-06, "loss": 0.8528, "step": 17409 }, { "epoch": 0.21, "learning_rate": 7.929388780702301e-06, "loss": 0.894, "step": 17410 }, { "epoch": 0.21, "learning_rate": 7.92833427132898e-06, "loss": 0.8678, "step": 17411 }, { "epoch": 0.21, "learning_rate": 7.927279786024849e-06, "loss": 0.8634, "step": 17412 }, { "epoch": 0.21, "learning_rate": 7.926225324802163e-06, "loss": 0.8914, "step": 17413 }, { "epoch": 0.21, "learning_rate": 7.925170887673171e-06, "loss": 0.8668, "step": 17414 }, { "epoch": 0.21, "learning_rate": 7.924116474650121e-06, "loss": 0.8996, "step": 17415 }, { "epoch": 0.21, "learning_rate": 7.923062085745265e-06, "loss": 0.8415, "step": 17416 }, { "epoch": 0.21, "learning_rate": 7.922007720970862e-06, "loss": 0.8826, "step": 17417 }, { "epoch": 0.21, "learning_rate": 7.92095338033915e-06, "loss": 0.8302, "step": 17418 }, { "epoch": 0.21, "learning_rate": 7.91989906386238e-06, "loss": 0.918, "step": 17419 }, { "epoch": 0.21, "learning_rate": 7.91884477155281e-06, "loss": 0.857, "step": 17420 }, { "epoch": 0.21, "learning_rate": 7.91779050342268e-06, "loss": 0.9079, "step": 17421 }, { "epoch": 0.21, "learning_rate": 7.916736259484239e-06, "loss": 0.8516, "step": 17422 }, { "epoch": 0.21, "learning_rate": 7.915682039749742e-06, "loss": 0.8935, "step": 17423 }, { "epoch": 0.21, "learning_rate": 7.914627844231438e-06, "loss": 0.8742, "step": 17424 }, { "epoch": 0.21, "learning_rate": 7.913573672941562e-06, "loss": 0.8795, "step": 17425 }, { "epoch": 0.21, "learning_rate": 7.912519525892375e-06, "loss": 0.8935, "step": 17426 }, { "epoch": 0.21, "learning_rate": 7.911465403096123e-06, "loss": 0.8998, "step": 17427 }, { "epoch": 0.21, "learning_rate": 7.910411304565045e-06, "loss": 0.8913, "step": 17428 }, { "epoch": 0.21, "learning_rate": 7.909357230311394e-06, "loss": 0.9001, "step": 17429 }, { "epoch": 0.21, "learning_rate": 7.908303180347414e-06, "loss": 0.9166, "step": 17430 }, { "epoch": 0.21, "learning_rate": 7.907249154685357e-06, "loss": 0.8899, "step": 17431 }, { "epoch": 0.21, "learning_rate": 7.90619515333746e-06, "loss": 0.8456, "step": 17432 }, { "epoch": 0.21, "learning_rate": 7.905141176315975e-06, "loss": 0.8573, "step": 17433 }, { "epoch": 0.21, "learning_rate": 7.904087223633146e-06, "loss": 0.8848, "step": 17434 }, { "epoch": 0.21, "learning_rate": 7.903033295301217e-06, "loss": 0.9087, "step": 17435 }, { "epoch": 0.21, "learning_rate": 7.901979391332433e-06, "loss": 0.8911, "step": 17436 }, { "epoch": 0.21, "learning_rate": 7.900925511739039e-06, "loss": 0.8935, "step": 17437 }, { "epoch": 0.21, "learning_rate": 7.899871656533282e-06, "loss": 0.8854, "step": 17438 }, { "epoch": 0.21, "learning_rate": 7.8988178257274e-06, "loss": 0.8861, "step": 17439 }, { "epoch": 0.21, "learning_rate": 7.89776401933364e-06, "loss": 0.8654, "step": 17440 }, { "epoch": 0.21, "learning_rate": 7.896710237364249e-06, "loss": 0.9031, "step": 17441 }, { "epoch": 0.21, "learning_rate": 7.895656479831463e-06, "loss": 0.86, "step": 17442 }, { "epoch": 0.21, "learning_rate": 7.894602746747529e-06, "loss": 0.8912, "step": 17443 }, { "epoch": 0.21, "learning_rate": 7.893549038124688e-06, "loss": 0.8627, "step": 17444 }, { "epoch": 0.21, "learning_rate": 7.892495353975186e-06, "loss": 0.8918, "step": 17445 }, { "epoch": 0.21, "learning_rate": 7.89144169431126e-06, "loss": 0.8585, "step": 17446 }, { "epoch": 0.21, "learning_rate": 7.890388059145153e-06, "loss": 0.8889, "step": 17447 }, { "epoch": 0.21, "learning_rate": 7.889334448489115e-06, "loss": 0.8862, "step": 17448 }, { "epoch": 0.21, "learning_rate": 7.888280862355369e-06, "loss": 0.8812, "step": 17449 }, { "epoch": 0.21, "learning_rate": 7.887227300756172e-06, "loss": 0.8628, "step": 17450 }, { "epoch": 0.21, "learning_rate": 7.886173763703757e-06, "loss": 0.8556, "step": 17451 }, { "epoch": 0.21, "learning_rate": 7.885120251210368e-06, "loss": 0.8632, "step": 17452 }, { "epoch": 0.21, "learning_rate": 7.884066763288243e-06, "loss": 0.8296, "step": 17453 }, { "epoch": 0.21, "learning_rate": 7.88301329994962e-06, "loss": 0.8824, "step": 17454 }, { "epoch": 0.21, "learning_rate": 7.881959861206746e-06, "loss": 0.8718, "step": 17455 }, { "epoch": 0.21, "learning_rate": 7.880906447071847e-06, "loss": 0.8765, "step": 17456 }, { "epoch": 0.21, "learning_rate": 7.879853057557175e-06, "loss": 0.8893, "step": 17457 }, { "epoch": 0.21, "learning_rate": 7.878799692674959e-06, "loss": 0.841, "step": 17458 }, { "epoch": 0.21, "learning_rate": 7.877746352437445e-06, "loss": 0.843, "step": 17459 }, { "epoch": 0.21, "learning_rate": 7.876693036856865e-06, "loss": 0.8699, "step": 17460 }, { "epoch": 0.22, "learning_rate": 7.875639745945458e-06, "loss": 0.8722, "step": 17461 }, { "epoch": 0.22, "learning_rate": 7.87458647971546e-06, "loss": 0.8379, "step": 17462 }, { "epoch": 0.22, "learning_rate": 7.87353323817912e-06, "loss": 0.8675, "step": 17463 }, { "epoch": 0.22, "learning_rate": 7.872480021348659e-06, "loss": 0.8456, "step": 17464 }, { "epoch": 0.22, "learning_rate": 7.871426829236317e-06, "loss": 0.9053, "step": 17465 }, { "epoch": 0.22, "learning_rate": 7.87037366185434e-06, "loss": 0.8776, "step": 17466 }, { "epoch": 0.22, "learning_rate": 7.869320519214955e-06, "loss": 0.8677, "step": 17467 }, { "epoch": 0.22, "learning_rate": 7.868267401330396e-06, "loss": 0.9007, "step": 17468 }, { "epoch": 0.22, "learning_rate": 7.867214308212906e-06, "loss": 0.8749, "step": 17469 }, { "epoch": 0.22, "learning_rate": 7.866161239874719e-06, "loss": 0.9024, "step": 17470 }, { "epoch": 0.22, "learning_rate": 7.86510819632806e-06, "loss": 0.8504, "step": 17471 }, { "epoch": 0.22, "learning_rate": 7.864055177585175e-06, "loss": 0.8537, "step": 17472 }, { "epoch": 0.22, "learning_rate": 7.863002183658298e-06, "loss": 0.8575, "step": 17473 }, { "epoch": 0.22, "learning_rate": 7.861949214559653e-06, "loss": 0.8548, "step": 17474 }, { "epoch": 0.22, "learning_rate": 7.860896270301482e-06, "loss": 0.9034, "step": 17475 }, { "epoch": 0.22, "learning_rate": 7.859843350896016e-06, "loss": 0.8628, "step": 17476 }, { "epoch": 0.22, "learning_rate": 7.858790456355491e-06, "loss": 0.867, "step": 17477 }, { "epoch": 0.22, "learning_rate": 7.857737586692133e-06, "loss": 0.9161, "step": 17478 }, { "epoch": 0.22, "learning_rate": 7.856684741918181e-06, "loss": 0.8841, "step": 17479 }, { "epoch": 0.22, "learning_rate": 7.855631922045866e-06, "loss": 0.8773, "step": 17480 }, { "epoch": 0.22, "learning_rate": 7.854579127087418e-06, "loss": 0.8766, "step": 17481 }, { "epoch": 0.22, "learning_rate": 7.853526357055068e-06, "loss": 0.8633, "step": 17482 }, { "epoch": 0.22, "learning_rate": 7.85247361196105e-06, "loss": 0.8595, "step": 17483 }, { "epoch": 0.22, "learning_rate": 7.851420891817596e-06, "loss": 0.9037, "step": 17484 }, { "epoch": 0.22, "learning_rate": 7.850368196636931e-06, "loss": 0.8787, "step": 17485 }, { "epoch": 0.22, "learning_rate": 7.84931552643129e-06, "loss": 0.8833, "step": 17486 }, { "epoch": 0.22, "learning_rate": 7.848262881212907e-06, "loss": 0.9128, "step": 17487 }, { "epoch": 0.22, "learning_rate": 7.847210260994005e-06, "loss": 0.8832, "step": 17488 }, { "epoch": 0.22, "learning_rate": 7.846157665786811e-06, "loss": 0.888, "step": 17489 }, { "epoch": 0.22, "learning_rate": 7.845105095603563e-06, "loss": 0.8757, "step": 17490 }, { "epoch": 0.22, "learning_rate": 7.84405255045649e-06, "loss": 0.8778, "step": 17491 }, { "epoch": 0.22, "learning_rate": 7.843000030357814e-06, "loss": 0.8276, "step": 17492 }, { "epoch": 0.22, "learning_rate": 7.841947535319766e-06, "loss": 0.8945, "step": 17493 }, { "epoch": 0.22, "learning_rate": 7.840895065354582e-06, "loss": 0.8922, "step": 17494 }, { "epoch": 0.22, "learning_rate": 7.839842620474472e-06, "loss": 0.8579, "step": 17495 }, { "epoch": 0.22, "learning_rate": 7.83879020069168e-06, "loss": 0.8698, "step": 17496 }, { "epoch": 0.22, "learning_rate": 7.837737806018427e-06, "loss": 0.8947, "step": 17497 }, { "epoch": 0.22, "learning_rate": 7.836685436466941e-06, "loss": 0.8442, "step": 17498 }, { "epoch": 0.22, "learning_rate": 7.835633092049449e-06, "loss": 0.8897, "step": 17499 }, { "epoch": 0.22, "learning_rate": 7.834580772778176e-06, "loss": 0.8707, "step": 17500 }, { "epoch": 0.22, "learning_rate": 7.833528478665351e-06, "loss": 0.8415, "step": 17501 }, { "epoch": 0.22, "learning_rate": 7.832476209723195e-06, "loss": 0.894, "step": 17502 }, { "epoch": 0.22, "learning_rate": 7.831423965963938e-06, "loss": 0.8457, "step": 17503 }, { "epoch": 0.22, "learning_rate": 7.830371747399803e-06, "loss": 0.8824, "step": 17504 }, { "epoch": 0.22, "learning_rate": 7.829319554043018e-06, "loss": 0.8287, "step": 17505 }, { "epoch": 0.22, "learning_rate": 7.828267385905802e-06, "loss": 0.8464, "step": 17506 }, { "epoch": 0.22, "learning_rate": 7.827215243000384e-06, "loss": 0.8451, "step": 17507 }, { "epoch": 0.22, "learning_rate": 7.826163125338984e-06, "loss": 0.8423, "step": 17508 }, { "epoch": 0.22, "learning_rate": 7.825111032933836e-06, "loss": 0.9331, "step": 17509 }, { "epoch": 0.22, "learning_rate": 7.824058965797151e-06, "loss": 0.8691, "step": 17510 }, { "epoch": 0.22, "learning_rate": 7.823006923941156e-06, "loss": 0.873, "step": 17511 }, { "epoch": 0.22, "learning_rate": 7.82195490737808e-06, "loss": 0.8583, "step": 17512 }, { "epoch": 0.22, "learning_rate": 7.820902916120138e-06, "loss": 0.8636, "step": 17513 }, { "epoch": 0.22, "learning_rate": 7.819850950179552e-06, "loss": 0.8288, "step": 17514 }, { "epoch": 0.22, "learning_rate": 7.818799009568551e-06, "loss": 0.8723, "step": 17515 }, { "epoch": 0.22, "learning_rate": 7.817747094299356e-06, "loss": 0.881, "step": 17516 }, { "epoch": 0.22, "learning_rate": 7.816695204384179e-06, "loss": 0.8923, "step": 17517 }, { "epoch": 0.22, "learning_rate": 7.81564333983525e-06, "loss": 0.8583, "step": 17518 }, { "epoch": 0.22, "learning_rate": 7.81459150066479e-06, "loss": 0.8553, "step": 17519 }, { "epoch": 0.22, "learning_rate": 7.813539686885015e-06, "loss": 0.8408, "step": 17520 }, { "epoch": 0.22, "learning_rate": 7.812487898508147e-06, "loss": 0.9165, "step": 17521 }, { "epoch": 0.22, "learning_rate": 7.811436135546405e-06, "loss": 0.8962, "step": 17522 }, { "epoch": 0.22, "learning_rate": 7.810384398012014e-06, "loss": 0.8729, "step": 17523 }, { "epoch": 0.22, "learning_rate": 7.809332685917185e-06, "loss": 0.8651, "step": 17524 }, { "epoch": 0.22, "learning_rate": 7.808280999274143e-06, "loss": 0.9145, "step": 17525 }, { "epoch": 0.22, "learning_rate": 7.807229338095106e-06, "loss": 0.8855, "step": 17526 }, { "epoch": 0.22, "learning_rate": 7.80617770239229e-06, "loss": 0.874, "step": 17527 }, { "epoch": 0.22, "learning_rate": 7.805126092177915e-06, "loss": 0.8423, "step": 17528 }, { "epoch": 0.22, "learning_rate": 7.8040745074642e-06, "loss": 0.8165, "step": 17529 }, { "epoch": 0.22, "learning_rate": 7.803022948263362e-06, "loss": 0.8875, "step": 17530 }, { "epoch": 0.22, "learning_rate": 7.801971414587616e-06, "loss": 0.8713, "step": 17531 }, { "epoch": 0.22, "learning_rate": 7.80091990644918e-06, "loss": 0.8671, "step": 17532 }, { "epoch": 0.22, "learning_rate": 7.799868423860275e-06, "loss": 0.9123, "step": 17533 }, { "epoch": 0.22, "learning_rate": 7.79881696683311e-06, "loss": 0.8662, "step": 17534 }, { "epoch": 0.22, "learning_rate": 7.797765535379905e-06, "loss": 0.8545, "step": 17535 }, { "epoch": 0.22, "learning_rate": 7.796714129512875e-06, "loss": 0.9276, "step": 17536 }, { "epoch": 0.22, "learning_rate": 7.795662749244242e-06, "loss": 0.8475, "step": 17537 }, { "epoch": 0.22, "learning_rate": 7.79461139458621e-06, "loss": 0.8689, "step": 17538 }, { "epoch": 0.22, "learning_rate": 7.793560065551e-06, "loss": 0.8521, "step": 17539 }, { "epoch": 0.22, "learning_rate": 7.792508762150833e-06, "loss": 0.8788, "step": 17540 }, { "epoch": 0.22, "learning_rate": 7.791457484397906e-06, "loss": 0.8513, "step": 17541 }, { "epoch": 0.22, "learning_rate": 7.790406232304448e-06, "loss": 0.8896, "step": 17542 }, { "epoch": 0.22, "learning_rate": 7.789355005882668e-06, "loss": 0.894, "step": 17543 }, { "epoch": 0.22, "learning_rate": 7.78830380514478e-06, "loss": 0.8514, "step": 17544 }, { "epoch": 0.22, "learning_rate": 7.787252630102998e-06, "loss": 0.8969, "step": 17545 }, { "epoch": 0.22, "learning_rate": 7.786201480769532e-06, "loss": 0.8647, "step": 17546 }, { "epoch": 0.22, "learning_rate": 7.785150357156598e-06, "loss": 0.8386, "step": 17547 }, { "epoch": 0.22, "learning_rate": 7.784099259276405e-06, "loss": 0.8673, "step": 17548 }, { "epoch": 0.22, "learning_rate": 7.783048187141165e-06, "loss": 0.8799, "step": 17549 }, { "epoch": 0.22, "learning_rate": 7.781997140763092e-06, "loss": 0.8554, "step": 17550 }, { "epoch": 0.22, "learning_rate": 7.7809461201544e-06, "loss": 0.8561, "step": 17551 }, { "epoch": 0.22, "learning_rate": 7.779895125327293e-06, "loss": 0.8943, "step": 17552 }, { "epoch": 0.22, "learning_rate": 7.778844156293984e-06, "loss": 0.8778, "step": 17553 }, { "epoch": 0.22, "learning_rate": 7.777793213066684e-06, "loss": 0.854, "step": 17554 }, { "epoch": 0.22, "learning_rate": 7.776742295657609e-06, "loss": 0.8846, "step": 17555 }, { "epoch": 0.22, "learning_rate": 7.77569140407896e-06, "loss": 0.8631, "step": 17556 }, { "epoch": 0.22, "learning_rate": 7.774640538342948e-06, "loss": 0.9055, "step": 17557 }, { "epoch": 0.22, "learning_rate": 7.77358969846179e-06, "loss": 0.92, "step": 17558 }, { "epoch": 0.22, "learning_rate": 7.772538884447686e-06, "loss": 0.9058, "step": 17559 }, { "epoch": 0.22, "learning_rate": 7.771488096312847e-06, "loss": 0.8605, "step": 17560 }, { "epoch": 0.22, "learning_rate": 7.770437334069483e-06, "loss": 0.8946, "step": 17561 }, { "epoch": 0.22, "learning_rate": 7.769386597729805e-06, "loss": 0.8814, "step": 17562 }, { "epoch": 0.22, "learning_rate": 7.768335887306014e-06, "loss": 0.8724, "step": 17563 }, { "epoch": 0.22, "learning_rate": 7.76728520281032e-06, "loss": 0.8666, "step": 17564 }, { "epoch": 0.22, "learning_rate": 7.766234544254933e-06, "loss": 0.866, "step": 17565 }, { "epoch": 0.22, "learning_rate": 7.765183911652057e-06, "loss": 0.8769, "step": 17566 }, { "epoch": 0.22, "learning_rate": 7.764133305013897e-06, "loss": 0.8541, "step": 17567 }, { "epoch": 0.22, "learning_rate": 7.763082724352664e-06, "loss": 0.8819, "step": 17568 }, { "epoch": 0.22, "learning_rate": 7.762032169680561e-06, "loss": 0.833, "step": 17569 }, { "epoch": 0.22, "learning_rate": 7.760981641009793e-06, "loss": 0.8631, "step": 17570 }, { "epoch": 0.22, "learning_rate": 7.759931138352567e-06, "loss": 0.875, "step": 17571 }, { "epoch": 0.22, "learning_rate": 7.758880661721088e-06, "loss": 0.8622, "step": 17572 }, { "epoch": 0.22, "learning_rate": 7.757830211127559e-06, "loss": 0.8401, "step": 17573 }, { "epoch": 0.22, "learning_rate": 7.756779786584186e-06, "loss": 0.8856, "step": 17574 }, { "epoch": 0.22, "learning_rate": 7.755729388103172e-06, "loss": 0.8956, "step": 17575 }, { "epoch": 0.22, "learning_rate": 7.754679015696723e-06, "loss": 0.8572, "step": 17576 }, { "epoch": 0.22, "learning_rate": 7.753628669377041e-06, "loss": 0.8831, "step": 17577 }, { "epoch": 0.22, "learning_rate": 7.752578349156324e-06, "loss": 0.8662, "step": 17578 }, { "epoch": 0.22, "learning_rate": 7.75152805504679e-06, "loss": 0.9054, "step": 17579 }, { "epoch": 0.22, "learning_rate": 7.750477787060624e-06, "loss": 0.8577, "step": 17580 }, { "epoch": 0.22, "learning_rate": 7.749427545210036e-06, "loss": 0.8989, "step": 17581 }, { "epoch": 0.22, "learning_rate": 7.748377329507228e-06, "loss": 0.8766, "step": 17582 }, { "epoch": 0.22, "learning_rate": 7.747327139964409e-06, "loss": 0.8887, "step": 17583 }, { "epoch": 0.22, "learning_rate": 7.746276976593765e-06, "loss": 0.8312, "step": 17584 }, { "epoch": 0.22, "learning_rate": 7.745226839407507e-06, "loss": 0.8739, "step": 17585 }, { "epoch": 0.22, "learning_rate": 7.744176728417839e-06, "loss": 0.8584, "step": 17586 }, { "epoch": 0.22, "learning_rate": 7.74312664363695e-06, "loss": 0.8564, "step": 17587 }, { "epoch": 0.22, "learning_rate": 7.74207658507705e-06, "loss": 0.8537, "step": 17588 }, { "epoch": 0.22, "learning_rate": 7.741026552750334e-06, "loss": 0.8833, "step": 17589 }, { "epoch": 0.22, "learning_rate": 7.739976546669006e-06, "loss": 0.8971, "step": 17590 }, { "epoch": 0.22, "learning_rate": 7.738926566845259e-06, "loss": 0.8767, "step": 17591 }, { "epoch": 0.22, "learning_rate": 7.737876613291297e-06, "loss": 0.8698, "step": 17592 }, { "epoch": 0.22, "learning_rate": 7.736826686019317e-06, "loss": 0.9022, "step": 17593 }, { "epoch": 0.22, "learning_rate": 7.735776785041517e-06, "loss": 0.8847, "step": 17594 }, { "epoch": 0.22, "learning_rate": 7.734726910370097e-06, "loss": 0.8377, "step": 17595 }, { "epoch": 0.22, "learning_rate": 7.73367706201725e-06, "loss": 0.8578, "step": 17596 }, { "epoch": 0.22, "learning_rate": 7.73262723999518e-06, "loss": 0.8561, "step": 17597 }, { "epoch": 0.22, "learning_rate": 7.731577444316079e-06, "loss": 0.8851, "step": 17598 }, { "epoch": 0.22, "learning_rate": 7.730527674992143e-06, "loss": 0.8257, "step": 17599 }, { "epoch": 0.22, "learning_rate": 7.729477932035571e-06, "loss": 0.8386, "step": 17600 }, { "epoch": 0.22, "learning_rate": 7.728428215458565e-06, "loss": 0.8826, "step": 17601 }, { "epoch": 0.22, "learning_rate": 7.727378525273313e-06, "loss": 0.8914, "step": 17602 }, { "epoch": 0.22, "learning_rate": 7.726328861492009e-06, "loss": 0.8756, "step": 17603 }, { "epoch": 0.22, "learning_rate": 7.725279224126857e-06, "loss": 0.8489, "step": 17604 }, { "epoch": 0.22, "learning_rate": 7.724229613190045e-06, "loss": 0.8875, "step": 17605 }, { "epoch": 0.22, "learning_rate": 7.723180028693766e-06, "loss": 0.8838, "step": 17606 }, { "epoch": 0.22, "learning_rate": 7.72213047065022e-06, "loss": 0.8809, "step": 17607 }, { "epoch": 0.22, "learning_rate": 7.721080939071603e-06, "loss": 0.8733, "step": 17608 }, { "epoch": 0.22, "learning_rate": 7.7200314339701e-06, "loss": 0.8866, "step": 17609 }, { "epoch": 0.22, "learning_rate": 7.718981955357912e-06, "loss": 0.8465, "step": 17610 }, { "epoch": 0.22, "learning_rate": 7.717932503247231e-06, "loss": 0.9099, "step": 17611 }, { "epoch": 0.22, "learning_rate": 7.716883077650245e-06, "loss": 0.8572, "step": 17612 }, { "epoch": 0.22, "learning_rate": 7.71583367857915e-06, "loss": 0.8848, "step": 17613 }, { "epoch": 0.22, "learning_rate": 7.714784306046139e-06, "loss": 0.8788, "step": 17614 }, { "epoch": 0.22, "learning_rate": 7.713734960063404e-06, "loss": 0.9059, "step": 17615 }, { "epoch": 0.22, "learning_rate": 7.712685640643135e-06, "loss": 0.834, "step": 17616 }, { "epoch": 0.22, "learning_rate": 7.711636347797521e-06, "loss": 0.8641, "step": 17617 }, { "epoch": 0.22, "learning_rate": 7.710587081538759e-06, "loss": 0.8847, "step": 17618 }, { "epoch": 0.22, "learning_rate": 7.709537841879034e-06, "loss": 0.8802, "step": 17619 }, { "epoch": 0.22, "learning_rate": 7.708488628830541e-06, "loss": 0.867, "step": 17620 }, { "epoch": 0.22, "learning_rate": 7.707439442405466e-06, "loss": 0.8707, "step": 17621 }, { "epoch": 0.22, "learning_rate": 7.706390282616003e-06, "loss": 0.8581, "step": 17622 }, { "epoch": 0.22, "learning_rate": 7.705341149474337e-06, "loss": 0.8757, "step": 17623 }, { "epoch": 0.22, "learning_rate": 7.704292042992656e-06, "loss": 0.8415, "step": 17624 }, { "epoch": 0.22, "learning_rate": 7.703242963183158e-06, "loss": 0.853, "step": 17625 }, { "epoch": 0.22, "learning_rate": 7.702193910058022e-06, "loss": 0.8589, "step": 17626 }, { "epoch": 0.22, "learning_rate": 7.70114488362944e-06, "loss": 0.9052, "step": 17627 }, { "epoch": 0.22, "learning_rate": 7.700095883909598e-06, "loss": 0.8603, "step": 17628 }, { "epoch": 0.22, "learning_rate": 7.69904691091069e-06, "loss": 0.8419, "step": 17629 }, { "epoch": 0.22, "learning_rate": 7.697997964644893e-06, "loss": 0.8719, "step": 17630 }, { "epoch": 0.22, "learning_rate": 7.696949045124401e-06, "loss": 0.9247, "step": 17631 }, { "epoch": 0.22, "learning_rate": 7.695900152361404e-06, "loss": 0.8628, "step": 17632 }, { "epoch": 0.22, "learning_rate": 7.694851286368074e-06, "loss": 0.8753, "step": 17633 }, { "epoch": 0.22, "learning_rate": 7.693802447156612e-06, "loss": 0.8765, "step": 17634 }, { "epoch": 0.22, "learning_rate": 7.692753634739196e-06, "loss": 0.8857, "step": 17635 }, { "epoch": 0.22, "learning_rate": 7.691704849128016e-06, "loss": 0.9201, "step": 17636 }, { "epoch": 0.22, "learning_rate": 7.690656090335253e-06, "loss": 0.8476, "step": 17637 }, { "epoch": 0.22, "learning_rate": 7.689607358373092e-06, "loss": 0.862, "step": 17638 }, { "epoch": 0.22, "learning_rate": 7.68855865325372e-06, "loss": 0.8637, "step": 17639 }, { "epoch": 0.22, "learning_rate": 7.68750997498932e-06, "loss": 0.8664, "step": 17640 }, { "epoch": 0.22, "learning_rate": 7.686461323592076e-06, "loss": 0.8416, "step": 17641 }, { "epoch": 0.22, "learning_rate": 7.68541269907417e-06, "loss": 0.8706, "step": 17642 }, { "epoch": 0.22, "learning_rate": 7.684364101447788e-06, "loss": 0.9039, "step": 17643 }, { "epoch": 0.22, "learning_rate": 7.68331553072511e-06, "loss": 0.9088, "step": 17644 }, { "epoch": 0.22, "learning_rate": 7.682266986918319e-06, "loss": 0.8564, "step": 17645 }, { "epoch": 0.22, "learning_rate": 7.681218470039597e-06, "loss": 0.8769, "step": 17646 }, { "epoch": 0.22, "learning_rate": 7.680169980101134e-06, "loss": 0.8833, "step": 17647 }, { "epoch": 0.22, "learning_rate": 7.6791215171151e-06, "loss": 0.8759, "step": 17648 }, { "epoch": 0.22, "learning_rate": 7.678073081093677e-06, "loss": 0.9036, "step": 17649 }, { "epoch": 0.22, "learning_rate": 7.67702467204906e-06, "loss": 0.865, "step": 17650 }, { "epoch": 0.22, "learning_rate": 7.675976289993414e-06, "loss": 0.8861, "step": 17651 }, { "epoch": 0.22, "learning_rate": 7.674927934938923e-06, "loss": 0.9019, "step": 17652 }, { "epoch": 0.22, "learning_rate": 7.67387960689777e-06, "loss": 0.8671, "step": 17653 }, { "epoch": 0.22, "learning_rate": 7.672831305882137e-06, "loss": 0.8794, "step": 17654 }, { "epoch": 0.22, "learning_rate": 7.6717830319042e-06, "loss": 0.8834, "step": 17655 }, { "epoch": 0.22, "learning_rate": 7.67073478497614e-06, "loss": 0.8704, "step": 17656 }, { "epoch": 0.22, "learning_rate": 7.669686565110132e-06, "loss": 0.8637, "step": 17657 }, { "epoch": 0.22, "learning_rate": 7.668638372318359e-06, "loss": 0.8942, "step": 17658 }, { "epoch": 0.22, "learning_rate": 7.667590206612996e-06, "loss": 0.8537, "step": 17659 }, { "epoch": 0.22, "learning_rate": 7.666542068006222e-06, "loss": 0.8667, "step": 17660 }, { "epoch": 0.22, "learning_rate": 7.665493956510216e-06, "loss": 0.8687, "step": 17661 }, { "epoch": 0.22, "learning_rate": 7.664445872137153e-06, "loss": 0.8869, "step": 17662 }, { "epoch": 0.22, "learning_rate": 7.66339781489921e-06, "loss": 0.8356, "step": 17663 }, { "epoch": 0.22, "learning_rate": 7.662349784808567e-06, "loss": 0.8174, "step": 17664 }, { "epoch": 0.22, "learning_rate": 7.661301781877396e-06, "loss": 0.9117, "step": 17665 }, { "epoch": 0.22, "learning_rate": 7.660253806117874e-06, "loss": 0.9115, "step": 17666 }, { "epoch": 0.22, "learning_rate": 7.659205857542178e-06, "loss": 0.8499, "step": 17667 }, { "epoch": 0.22, "learning_rate": 7.658157936162485e-06, "loss": 0.8527, "step": 17668 }, { "epoch": 0.22, "learning_rate": 7.657110041990966e-06, "loss": 0.8794, "step": 17669 }, { "epoch": 0.22, "learning_rate": 7.656062175039796e-06, "loss": 0.8646, "step": 17670 }, { "epoch": 0.22, "learning_rate": 7.655014335321155e-06, "loss": 0.9212, "step": 17671 }, { "epoch": 0.22, "learning_rate": 7.653966522847211e-06, "loss": 0.8789, "step": 17672 }, { "epoch": 0.22, "learning_rate": 7.652918737630138e-06, "loss": 0.886, "step": 17673 }, { "epoch": 0.22, "learning_rate": 7.651870979682112e-06, "loss": 0.861, "step": 17674 }, { "epoch": 0.22, "learning_rate": 7.65082324901531e-06, "loss": 0.8719, "step": 17675 }, { "epoch": 0.22, "learning_rate": 7.649775545641895e-06, "loss": 0.8605, "step": 17676 }, { "epoch": 0.22, "learning_rate": 7.648727869574046e-06, "loss": 0.9122, "step": 17677 }, { "epoch": 0.22, "learning_rate": 7.647680220823939e-06, "loss": 0.8861, "step": 17678 }, { "epoch": 0.22, "learning_rate": 7.646632599403734e-06, "loss": 0.88, "step": 17679 }, { "epoch": 0.22, "learning_rate": 7.645585005325612e-06, "loss": 0.8661, "step": 17680 }, { "epoch": 0.22, "learning_rate": 7.644537438601741e-06, "loss": 0.8281, "step": 17681 }, { "epoch": 0.22, "learning_rate": 7.643489899244296e-06, "loss": 0.8735, "step": 17682 }, { "epoch": 0.22, "learning_rate": 7.64244238726544e-06, "loss": 0.8487, "step": 17683 }, { "epoch": 0.22, "learning_rate": 7.64139490267735e-06, "loss": 0.8924, "step": 17684 }, { "epoch": 0.22, "learning_rate": 7.640347445492192e-06, "loss": 0.9165, "step": 17685 }, { "epoch": 0.22, "learning_rate": 7.63930001572214e-06, "loss": 0.8684, "step": 17686 }, { "epoch": 0.22, "learning_rate": 7.638252613379357e-06, "loss": 0.8717, "step": 17687 }, { "epoch": 0.22, "learning_rate": 7.637205238476018e-06, "loss": 0.8638, "step": 17688 }, { "epoch": 0.22, "learning_rate": 7.636157891024288e-06, "loss": 0.8608, "step": 17689 }, { "epoch": 0.22, "learning_rate": 7.635110571036336e-06, "loss": 0.9051, "step": 17690 }, { "epoch": 0.22, "learning_rate": 7.634063278524332e-06, "loss": 0.8196, "step": 17691 }, { "epoch": 0.22, "learning_rate": 7.633016013500438e-06, "loss": 0.8694, "step": 17692 }, { "epoch": 0.22, "learning_rate": 7.631968775976834e-06, "loss": 0.8877, "step": 17693 }, { "epoch": 0.22, "learning_rate": 7.630921565965673e-06, "loss": 0.8827, "step": 17694 }, { "epoch": 0.22, "learning_rate": 7.629874383479127e-06, "loss": 0.9087, "step": 17695 }, { "epoch": 0.22, "learning_rate": 7.628827228529368e-06, "loss": 0.9078, "step": 17696 }, { "epoch": 0.22, "learning_rate": 7.627780101128552e-06, "loss": 0.8914, "step": 17697 }, { "epoch": 0.22, "learning_rate": 7.626733001288852e-06, "loss": 0.8741, "step": 17698 }, { "epoch": 0.22, "learning_rate": 7.625685929022432e-06, "loss": 0.8963, "step": 17699 }, { "epoch": 0.22, "learning_rate": 7.6246388843414574e-06, "loss": 0.8665, "step": 17700 }, { "epoch": 0.22, "learning_rate": 7.623591867258091e-06, "loss": 0.8613, "step": 17701 }, { "epoch": 0.22, "learning_rate": 7.622544877784499e-06, "loss": 0.8213, "step": 17702 }, { "epoch": 0.22, "learning_rate": 7.6214979159328475e-06, "loss": 0.8665, "step": 17703 }, { "epoch": 0.22, "learning_rate": 7.620450981715296e-06, "loss": 0.84, "step": 17704 }, { "epoch": 0.22, "learning_rate": 7.61940407514401e-06, "loss": 0.8452, "step": 17705 }, { "epoch": 0.22, "learning_rate": 7.6183571962311535e-06, "loss": 0.8671, "step": 17706 }, { "epoch": 0.22, "learning_rate": 7.617310344988891e-06, "loss": 0.9064, "step": 17707 }, { "epoch": 0.22, "learning_rate": 7.616263521429381e-06, "loss": 0.8612, "step": 17708 }, { "epoch": 0.22, "learning_rate": 7.615216725564788e-06, "loss": 0.8854, "step": 17709 }, { "epoch": 0.22, "learning_rate": 7.6141699574072755e-06, "loss": 0.87, "step": 17710 }, { "epoch": 0.22, "learning_rate": 7.613123216969002e-06, "loss": 0.8427, "step": 17711 }, { "epoch": 0.22, "learning_rate": 7.612076504262129e-06, "loss": 0.896, "step": 17712 }, { "epoch": 0.22, "learning_rate": 7.611029819298818e-06, "loss": 0.8766, "step": 17713 }, { "epoch": 0.22, "learning_rate": 7.609983162091237e-06, "loss": 0.8644, "step": 17714 }, { "epoch": 0.22, "learning_rate": 7.608936532651535e-06, "loss": 0.83, "step": 17715 }, { "epoch": 0.22, "learning_rate": 7.607889930991875e-06, "loss": 0.8853, "step": 17716 }, { "epoch": 0.22, "learning_rate": 7.606843357124426e-06, "loss": 0.8816, "step": 17717 }, { "epoch": 0.22, "learning_rate": 7.605796811061334e-06, "loss": 0.8551, "step": 17718 }, { "epoch": 0.22, "learning_rate": 7.604750292814763e-06, "loss": 0.8575, "step": 17719 }, { "epoch": 0.22, "learning_rate": 7.6037038023968745e-06, "loss": 0.8611, "step": 17720 }, { "epoch": 0.22, "learning_rate": 7.60265733981983e-06, "loss": 0.9261, "step": 17721 }, { "epoch": 0.22, "learning_rate": 7.601610905095777e-06, "loss": 0.8872, "step": 17722 }, { "epoch": 0.22, "learning_rate": 7.60056449823688e-06, "loss": 0.8567, "step": 17723 }, { "epoch": 0.22, "learning_rate": 7.5995181192553e-06, "loss": 0.8624, "step": 17724 }, { "epoch": 0.22, "learning_rate": 7.598471768163183e-06, "loss": 0.8947, "step": 17725 }, { "epoch": 0.22, "learning_rate": 7.597425444972694e-06, "loss": 0.8898, "step": 17726 }, { "epoch": 0.22, "learning_rate": 7.59637914969599e-06, "loss": 0.8403, "step": 17727 }, { "epoch": 0.22, "learning_rate": 7.5953328823452265e-06, "loss": 0.8856, "step": 17728 }, { "epoch": 0.22, "learning_rate": 7.594286642932556e-06, "loss": 0.8622, "step": 17729 }, { "epoch": 0.22, "learning_rate": 7.593240431470135e-06, "loss": 0.8891, "step": 17730 }, { "epoch": 0.22, "learning_rate": 7.5921942479701194e-06, "loss": 0.9051, "step": 17731 }, { "epoch": 0.22, "learning_rate": 7.591148092444668e-06, "loss": 0.8941, "step": 17732 }, { "epoch": 0.22, "learning_rate": 7.590101964905928e-06, "loss": 0.873, "step": 17733 }, { "epoch": 0.22, "learning_rate": 7.589055865366057e-06, "loss": 0.8819, "step": 17734 }, { "epoch": 0.22, "learning_rate": 7.588009793837212e-06, "loss": 0.9006, "step": 17735 }, { "epoch": 0.22, "learning_rate": 7.586963750331542e-06, "loss": 0.9044, "step": 17736 }, { "epoch": 0.22, "learning_rate": 7.585917734861201e-06, "loss": 0.8834, "step": 17737 }, { "epoch": 0.22, "learning_rate": 7.5848717474383405e-06, "loss": 0.9011, "step": 17738 }, { "epoch": 0.22, "learning_rate": 7.583825788075122e-06, "loss": 0.8853, "step": 17739 }, { "epoch": 0.22, "learning_rate": 7.582779856783686e-06, "loss": 0.9114, "step": 17740 }, { "epoch": 0.22, "learning_rate": 7.5817339535761895e-06, "loss": 0.8501, "step": 17741 }, { "epoch": 0.22, "learning_rate": 7.5806880784647884e-06, "loss": 0.8554, "step": 17742 }, { "epoch": 0.22, "learning_rate": 7.5796422314616234e-06, "loss": 0.8722, "step": 17743 }, { "epoch": 0.22, "learning_rate": 7.5785964125788545e-06, "loss": 0.8678, "step": 17744 }, { "epoch": 0.22, "learning_rate": 7.57755062182863e-06, "loss": 0.8507, "step": 17745 }, { "epoch": 0.22, "learning_rate": 7.576504859223102e-06, "loss": 0.8839, "step": 17746 }, { "epoch": 0.22, "learning_rate": 7.575459124774414e-06, "loss": 0.8793, "step": 17747 }, { "epoch": 0.22, "learning_rate": 7.5744134184947215e-06, "loss": 0.8551, "step": 17748 }, { "epoch": 0.22, "learning_rate": 7.573367740396174e-06, "loss": 0.8749, "step": 17749 }, { "epoch": 0.22, "learning_rate": 7.5723220904909155e-06, "loss": 0.869, "step": 17750 }, { "epoch": 0.0, "learning_rate": 7.571276468791099e-06, "loss": 0.8921, "step": 17751 }, { "epoch": 0.0, "learning_rate": 7.570230875308871e-06, "loss": 0.9095, "step": 17752 }, { "epoch": 0.0, "learning_rate": 7.569185310056382e-06, "loss": 0.8641, "step": 17753 }, { "epoch": 0.0, "learning_rate": 7.568139773045775e-06, "loss": 0.9384, "step": 17754 }, { "epoch": 0.0, "learning_rate": 7.567094264289201e-06, "loss": 0.8981, "step": 17755 }, { "epoch": 0.0, "learning_rate": 7.5660487837988085e-06, "loss": 0.8981, "step": 17756 }, { "epoch": 0.0, "learning_rate": 7.565003331586739e-06, "loss": 0.8989, "step": 17757 }, { "epoch": 0.0, "learning_rate": 7.563957907665142e-06, "loss": 0.9693, "step": 17758 }, { "epoch": 0.0, "learning_rate": 7.562912512046161e-06, "loss": 0.8869, "step": 17759 }, { "epoch": 0.0, "learning_rate": 7.56186714474195e-06, "loss": 0.9224, "step": 17760 }, { "epoch": 0.0, "learning_rate": 7.5608218057646445e-06, "loss": 0.889, "step": 17761 }, { "epoch": 0.0, "learning_rate": 7.55977649512639e-06, "loss": 0.8934, "step": 17762 }, { "epoch": 0.0, "learning_rate": 7.5587312128393405e-06, "loss": 0.8944, "step": 17763 }, { "epoch": 0.0, "learning_rate": 7.557685958915632e-06, "loss": 0.9323, "step": 17764 }, { "epoch": 0.0, "learning_rate": 7.556640733367407e-06, "loss": 0.9064, "step": 17765 }, { "epoch": 0.0, "learning_rate": 7.555595536206817e-06, "loss": 0.8817, "step": 17766 }, { "epoch": 0.0, "learning_rate": 7.554550367446005e-06, "loss": 0.8883, "step": 17767 }, { "epoch": 0.0, "learning_rate": 7.553505227097104e-06, "loss": 0.9118, "step": 17768 }, { "epoch": 0.0, "learning_rate": 7.552460115172266e-06, "loss": 0.8953, "step": 17769 }, { "epoch": 0.0, "learning_rate": 7.551415031683632e-06, "loss": 0.893, "step": 17770 }, { "epoch": 0.0, "learning_rate": 7.550369976643341e-06, "loss": 0.9247, "step": 17771 }, { "epoch": 0.0, "learning_rate": 7.549324950063537e-06, "loss": 0.9247, "step": 17772 }, { "epoch": 0.0, "learning_rate": 7.5482799519563595e-06, "loss": 0.8798, "step": 17773 }, { "epoch": 0.0, "learning_rate": 7.547234982333955e-06, "loss": 0.9218, "step": 17774 }, { "epoch": 0.0, "learning_rate": 7.546190041208457e-06, "loss": 0.8848, "step": 17775 }, { "epoch": 0.0, "learning_rate": 7.545145128592009e-06, "loss": 0.925, "step": 17776 }, { "epoch": 0.0, "learning_rate": 7.544100244496753e-06, "loss": 0.9136, "step": 17777 }, { "epoch": 0.0, "learning_rate": 7.543055388934826e-06, "loss": 0.9266, "step": 17778 }, { "epoch": 0.0, "learning_rate": 7.542010561918369e-06, "loss": 0.9105, "step": 17779 }, { "epoch": 0.0, "learning_rate": 7.54096576345952e-06, "loss": 0.9263, "step": 17780 }, { "epoch": 0.0, "learning_rate": 7.53992099357042e-06, "loss": 0.8879, "step": 17781 }, { "epoch": 0.0, "learning_rate": 7.538876252263204e-06, "loss": 0.9406, "step": 17782 }, { "epoch": 0.0, "learning_rate": 7.537831539550013e-06, "loss": 0.9262, "step": 17783 }, { "epoch": 0.0, "learning_rate": 7.5367868554429794e-06, "loss": 0.8849, "step": 17784 }, { "epoch": 0.0, "learning_rate": 7.535742199954252e-06, "loss": 0.8734, "step": 17785 }, { "epoch": 0.0, "learning_rate": 7.534697573095954e-06, "loss": 0.8705, "step": 17786 }, { "epoch": 0.0, "learning_rate": 7.533652974880231e-06, "loss": 0.9276, "step": 17787 }, { "epoch": 0.0, "learning_rate": 7.5326084053192226e-06, "loss": 0.8581, "step": 17788 }, { "epoch": 0.0, "learning_rate": 7.531563864425052e-06, "loss": 0.8727, "step": 17789 }, { "epoch": 0.0, "learning_rate": 7.530519352209866e-06, "loss": 0.8398, "step": 17790 }, { "epoch": 0.0, "learning_rate": 7.5294748686857955e-06, "loss": 0.8971, "step": 17791 }, { "epoch": 0.0, "learning_rate": 7.52843041386498e-06, "loss": 0.8995, "step": 17792 }, { "epoch": 0.0, "learning_rate": 7.527385987759548e-06, "loss": 0.9139, "step": 17793 }, { "epoch": 0.0, "learning_rate": 7.526341590381636e-06, "loss": 0.9499, "step": 17794 }, { "epoch": 0.0, "learning_rate": 7.525297221743383e-06, "loss": 0.8929, "step": 17795 }, { "epoch": 0.0, "learning_rate": 7.524252881856916e-06, "loss": 0.8549, "step": 17796 }, { "epoch": 0.0, "learning_rate": 7.523208570734371e-06, "loss": 0.9356, "step": 17797 }, { "epoch": 0.0, "learning_rate": 7.5221642883878824e-06, "loss": 0.8931, "step": 17798 }, { "epoch": 0.0, "learning_rate": 7.5211200348295836e-06, "loss": 0.8872, "step": 17799 }, { "epoch": 0.0, "learning_rate": 7.520075810071602e-06, "loss": 0.8683, "step": 17800 }, { "epoch": 0.0, "learning_rate": 7.519031614126073e-06, "loss": 0.8485, "step": 17801 }, { "epoch": 0.0, "learning_rate": 7.517987447005132e-06, "loss": 0.9114, "step": 17802 }, { "epoch": 0.0, "learning_rate": 7.516943308720905e-06, "loss": 0.9286, "step": 17803 }, { "epoch": 0.0, "learning_rate": 7.5158991992855234e-06, "loss": 0.9337, "step": 17804 }, { "epoch": 0.0, "learning_rate": 7.514855118711117e-06, "loss": 0.8786, "step": 17805 }, { "epoch": 0.0, "learning_rate": 7.513811067009826e-06, "loss": 0.8991, "step": 17806 }, { "epoch": 0.0, "learning_rate": 7.5127670441937696e-06, "loss": 0.9159, "step": 17807 }, { "epoch": 0.0, "learning_rate": 7.511723050275078e-06, "loss": 0.9219, "step": 17808 }, { "epoch": 0.0, "learning_rate": 7.510679085265891e-06, "loss": 0.8177, "step": 17809 }, { "epoch": 0.0, "learning_rate": 7.5096351491783246e-06, "loss": 0.8895, "step": 17810 }, { "epoch": 0.0, "learning_rate": 7.508591242024512e-06, "loss": 0.8973, "step": 17811 }, { "epoch": 0.0, "learning_rate": 7.507547363816585e-06, "loss": 0.8969, "step": 17812 }, { "epoch": 0.0, "learning_rate": 7.506503514566674e-06, "loss": 0.8761, "step": 17813 }, { "epoch": 0.0, "learning_rate": 7.505459694286896e-06, "loss": 0.9503, "step": 17814 }, { "epoch": 0.0, "learning_rate": 7.504415902989387e-06, "loss": 0.9146, "step": 17815 }, { "epoch": 0.0, "learning_rate": 7.503372140686274e-06, "loss": 0.8665, "step": 17816 }, { "epoch": 0.0, "learning_rate": 7.50232840738968e-06, "loss": 0.9043, "step": 17817 }, { "epoch": 0.0, "learning_rate": 7.501284703111731e-06, "loss": 0.9077, "step": 17818 }, { "epoch": 0.0, "learning_rate": 7.500241027864557e-06, "loss": 0.8827, "step": 17819 }, { "epoch": 0.0, "learning_rate": 7.499197381660282e-06, "loss": 0.9389, "step": 17820 }, { "epoch": 0.0, "learning_rate": 7.49815376451103e-06, "loss": 0.8657, "step": 17821 }, { "epoch": 0.0, "learning_rate": 7.4971101764289276e-06, "loss": 0.8848, "step": 17822 }, { "epoch": 0.0, "learning_rate": 7.496066617426099e-06, "loss": 0.8817, "step": 17823 }, { "epoch": 0.0, "learning_rate": 7.495023087514669e-06, "loss": 0.8728, "step": 17824 }, { "epoch": 0.0, "learning_rate": 7.49397958670676e-06, "loss": 0.8668, "step": 17825 }, { "epoch": 0.0, "learning_rate": 7.4929361150144975e-06, "loss": 0.9018, "step": 17826 }, { "epoch": 0.0, "learning_rate": 7.491892672450005e-06, "loss": 0.917, "step": 17827 }, { "epoch": 0.0, "learning_rate": 7.490849259025404e-06, "loss": 0.9002, "step": 17828 }, { "epoch": 0.0, "learning_rate": 7.4898058747528155e-06, "loss": 0.9269, "step": 17829 }, { "epoch": 0.0, "learning_rate": 7.4887625196443645e-06, "loss": 0.9152, "step": 17830 }, { "epoch": 0.0, "learning_rate": 7.4877191937121785e-06, "loss": 0.9084, "step": 17831 }, { "epoch": 0.0, "learning_rate": 7.4866758969683675e-06, "loss": 0.9201, "step": 17832 }, { "epoch": 0.0, "learning_rate": 7.485632629425059e-06, "loss": 0.8524, "step": 17833 }, { "epoch": 0.0, "learning_rate": 7.4845893910943786e-06, "loss": 0.8951, "step": 17834 }, { "epoch": 0.0, "learning_rate": 7.483546181988437e-06, "loss": 0.8988, "step": 17835 }, { "epoch": 0.0, "learning_rate": 7.482503002119359e-06, "loss": 0.8847, "step": 17836 }, { "epoch": 0.0, "learning_rate": 7.481459851499266e-06, "loss": 0.8739, "step": 17837 }, { "epoch": 0.0, "learning_rate": 7.48041673014028e-06, "loss": 0.9131, "step": 17838 }, { "epoch": 0.0, "learning_rate": 7.479373638054514e-06, "loss": 0.8928, "step": 17839 }, { "epoch": 0.0, "learning_rate": 7.47833057525409e-06, "loss": 0.9159, "step": 17840 }, { "epoch": 0.0, "learning_rate": 7.4772875417511285e-06, "loss": 0.8812, "step": 17841 }, { "epoch": 0.0, "learning_rate": 7.476244537557744e-06, "loss": 0.8837, "step": 17842 }, { "epoch": 0.0, "learning_rate": 7.475201562686056e-06, "loss": 0.8988, "step": 17843 }, { "epoch": 0.0, "learning_rate": 7.474158617148181e-06, "loss": 0.9013, "step": 17844 }, { "epoch": 0.0, "learning_rate": 7.47311570095624e-06, "loss": 0.8426, "step": 17845 }, { "epoch": 0.0, "learning_rate": 7.472072814122346e-06, "loss": 0.8822, "step": 17846 }, { "epoch": 0.0, "learning_rate": 7.471029956658616e-06, "loss": 0.9117, "step": 17847 }, { "epoch": 0.0, "learning_rate": 7.4699871285771695e-06, "loss": 0.8782, "step": 17848 }, { "epoch": 0.0, "learning_rate": 7.468944329890119e-06, "loss": 0.8767, "step": 17849 }, { "epoch": 0.0, "learning_rate": 7.46790156060958e-06, "loss": 0.9169, "step": 17850 }, { "epoch": 0.0, "learning_rate": 7.4668588207476655e-06, "loss": 0.8426, "step": 17851 }, { "epoch": 0.0, "learning_rate": 7.4658161103165014e-06, "loss": 0.8612, "step": 17852 }, { "epoch": 0.0, "learning_rate": 7.464773429328189e-06, "loss": 0.8585, "step": 17853 }, { "epoch": 0.0, "learning_rate": 7.463730777794847e-06, "loss": 0.8667, "step": 17854 }, { "epoch": 0.0, "learning_rate": 7.462688155728595e-06, "loss": 0.8537, "step": 17855 }, { "epoch": 0.0, "learning_rate": 7.4616455631415376e-06, "loss": 0.8489, "step": 17856 }, { "epoch": 0.0, "learning_rate": 7.460603000045788e-06, "loss": 0.8396, "step": 17857 }, { "epoch": 0.0, "learning_rate": 7.459560466453467e-06, "loss": 0.8786, "step": 17858 }, { "epoch": 0.0, "learning_rate": 7.458517962376684e-06, "loss": 0.8611, "step": 17859 }, { "epoch": 0.0, "learning_rate": 7.457475487827548e-06, "loss": 0.9034, "step": 17860 }, { "epoch": 0.0, "learning_rate": 7.456433042818174e-06, "loss": 0.891, "step": 17861 }, { "epoch": 0.0, "learning_rate": 7.455390627360672e-06, "loss": 0.9031, "step": 17862 }, { "epoch": 0.0, "learning_rate": 7.454348241467153e-06, "loss": 0.8961, "step": 17863 }, { "epoch": 0.0, "learning_rate": 7.453305885149726e-06, "loss": 0.8891, "step": 17864 }, { "epoch": 0.0, "learning_rate": 7.452263558420505e-06, "loss": 0.9007, "step": 17865 }, { "epoch": 0.0, "learning_rate": 7.4512212612916e-06, "loss": 0.9231, "step": 17866 }, { "epoch": 0.0, "learning_rate": 7.450178993775117e-06, "loss": 0.9318, "step": 17867 }, { "epoch": 0.0, "learning_rate": 7.4491367558831665e-06, "loss": 0.8815, "step": 17868 }, { "epoch": 0.0, "learning_rate": 7.4480945476278595e-06, "loss": 0.8547, "step": 17869 }, { "epoch": 0.0, "learning_rate": 7.447052369021305e-06, "loss": 0.9282, "step": 17870 }, { "epoch": 0.0, "learning_rate": 7.446010220075608e-06, "loss": 0.9058, "step": 17871 }, { "epoch": 0.0, "learning_rate": 7.444968100802879e-06, "loss": 0.8798, "step": 17872 }, { "epoch": 0.0, "learning_rate": 7.443926011215226e-06, "loss": 0.852, "step": 17873 }, { "epoch": 0.0, "learning_rate": 7.442883951324753e-06, "loss": 0.8985, "step": 17874 }, { "epoch": 0.0, "learning_rate": 7.4418419211435666e-06, "loss": 0.9033, "step": 17875 }, { "epoch": 0.0, "learning_rate": 7.44079992068378e-06, "loss": 0.9085, "step": 17876 }, { "epoch": 0.0, "learning_rate": 7.439757949957499e-06, "loss": 0.8709, "step": 17877 }, { "epoch": 0.0, "learning_rate": 7.43871600897682e-06, "loss": 0.8886, "step": 17878 }, { "epoch": 0.0, "learning_rate": 7.437674097753855e-06, "loss": 0.8833, "step": 17879 }, { "epoch": 0.0, "learning_rate": 7.4366322163007145e-06, "loss": 0.8757, "step": 17880 }, { "epoch": 0.0, "learning_rate": 7.435590364629491e-06, "loss": 0.9268, "step": 17881 }, { "epoch": 0.0, "learning_rate": 7.434548542752298e-06, "loss": 0.8919, "step": 17882 }, { "epoch": 0.0, "learning_rate": 7.433506750681237e-06, "loss": 0.887, "step": 17883 }, { "epoch": 0.0, "learning_rate": 7.432464988428414e-06, "loss": 0.8509, "step": 17884 }, { "epoch": 0.0, "learning_rate": 7.4314232560059294e-06, "loss": 0.8334, "step": 17885 }, { "epoch": 0.0, "learning_rate": 7.430381553425887e-06, "loss": 0.8873, "step": 17886 }, { "epoch": 0.0, "learning_rate": 7.429339880700392e-06, "loss": 0.8976, "step": 17887 }, { "epoch": 0.0, "learning_rate": 7.428298237841543e-06, "loss": 0.8929, "step": 17888 }, { "epoch": 0.0, "learning_rate": 7.4272566248614454e-06, "loss": 0.8858, "step": 17889 }, { "epoch": 0.0, "learning_rate": 7.4262150417721976e-06, "loss": 0.9108, "step": 17890 }, { "epoch": 0.0, "learning_rate": 7.425173488585906e-06, "loss": 0.879, "step": 17891 }, { "epoch": 0.0, "learning_rate": 7.424131965314667e-06, "loss": 0.8863, "step": 17892 }, { "epoch": 0.0, "learning_rate": 7.4230904719705845e-06, "loss": 0.8735, "step": 17893 }, { "epoch": 0.0, "learning_rate": 7.422049008565757e-06, "loss": 0.9076, "step": 17894 }, { "epoch": 0.0, "learning_rate": 7.421007575112284e-06, "loss": 0.8843, "step": 17895 }, { "epoch": 0.0, "learning_rate": 7.419966171622266e-06, "loss": 0.8616, "step": 17896 }, { "epoch": 0.0, "learning_rate": 7.418924798107801e-06, "loss": 0.8878, "step": 17897 }, { "epoch": 0.0, "learning_rate": 7.417883454580994e-06, "loss": 0.915, "step": 17898 }, { "epoch": 0.0, "learning_rate": 7.416842141053935e-06, "loss": 0.8884, "step": 17899 }, { "epoch": 0.0, "learning_rate": 7.415800857538725e-06, "loss": 0.8812, "step": 17900 }, { "epoch": 0.01, "learning_rate": 7.41475960404747e-06, "loss": 0.8487, "step": 17901 }, { "epoch": 0.01, "learning_rate": 7.413718380592256e-06, "loss": 0.857, "step": 17902 }, { "epoch": 0.01, "learning_rate": 7.412677187185182e-06, "loss": 0.8878, "step": 17903 }, { "epoch": 0.01, "learning_rate": 7.411636023838351e-06, "loss": 0.9068, "step": 17904 }, { "epoch": 0.01, "learning_rate": 7.410594890563858e-06, "loss": 0.9002, "step": 17905 }, { "epoch": 0.01, "learning_rate": 7.409553787373795e-06, "loss": 0.8995, "step": 17906 }, { "epoch": 0.01, "learning_rate": 7.408512714280261e-06, "loss": 0.8929, "step": 17907 }, { "epoch": 0.01, "learning_rate": 7.407471671295353e-06, "loss": 0.8477, "step": 17908 }, { "epoch": 0.01, "learning_rate": 7.406430658431162e-06, "loss": 0.9302, "step": 17909 }, { "epoch": 0.01, "learning_rate": 7.405389675699784e-06, "loss": 0.8534, "step": 17910 }, { "epoch": 0.01, "learning_rate": 7.404348723113315e-06, "loss": 0.8731, "step": 17911 }, { "epoch": 0.01, "learning_rate": 7.40330780068385e-06, "loss": 0.8802, "step": 17912 }, { "epoch": 0.01, "learning_rate": 7.4022669084234784e-06, "loss": 0.8367, "step": 17913 }, { "epoch": 0.01, "learning_rate": 7.401226046344299e-06, "loss": 0.8898, "step": 17914 }, { "epoch": 0.01, "learning_rate": 7.4001852144584e-06, "loss": 0.9072, "step": 17915 }, { "epoch": 0.01, "learning_rate": 7.399144412777877e-06, "loss": 0.8523, "step": 17916 }, { "epoch": 0.01, "learning_rate": 7.39810364131482e-06, "loss": 0.9047, "step": 17917 }, { "epoch": 0.01, "learning_rate": 7.397062900081323e-06, "loss": 0.8722, "step": 17918 }, { "epoch": 0.01, "learning_rate": 7.396022189089479e-06, "loss": 0.9047, "step": 17919 }, { "epoch": 0.01, "learning_rate": 7.394981508351375e-06, "loss": 0.8708, "step": 17920 }, { "epoch": 0.01, "learning_rate": 7.393940857879102e-06, "loss": 0.8684, "step": 17921 }, { "epoch": 0.01, "learning_rate": 7.392900237684755e-06, "loss": 0.8386, "step": 17922 }, { "epoch": 0.01, "learning_rate": 7.391859647780426e-06, "loss": 0.9014, "step": 17923 }, { "epoch": 0.01, "learning_rate": 7.390819088178196e-06, "loss": 0.89, "step": 17924 }, { "epoch": 0.01, "learning_rate": 7.389778558890159e-06, "loss": 0.8485, "step": 17925 }, { "epoch": 0.01, "learning_rate": 7.388738059928411e-06, "loss": 0.8809, "step": 17926 }, { "epoch": 0.01, "learning_rate": 7.3876975913050274e-06, "loss": 0.8883, "step": 17927 }, { "epoch": 0.01, "learning_rate": 7.386657153032104e-06, "loss": 0.8828, "step": 17928 }, { "epoch": 0.01, "learning_rate": 7.385616745121731e-06, "loss": 0.8361, "step": 17929 }, { "epoch": 0.01, "learning_rate": 7.384576367585994e-06, "loss": 0.8705, "step": 17930 }, { "epoch": 0.01, "learning_rate": 7.383536020436979e-06, "loss": 0.8732, "step": 17931 }, { "epoch": 0.01, "learning_rate": 7.382495703686774e-06, "loss": 0.9039, "step": 17932 }, { "epoch": 0.01, "learning_rate": 7.381455417347467e-06, "loss": 0.8684, "step": 17933 }, { "epoch": 0.01, "learning_rate": 7.3804151614311405e-06, "loss": 0.8628, "step": 17934 }, { "epoch": 0.01, "learning_rate": 7.379374935949884e-06, "loss": 0.888, "step": 17935 }, { "epoch": 0.01, "learning_rate": 7.378334740915783e-06, "loss": 0.8671, "step": 17936 }, { "epoch": 0.01, "learning_rate": 7.3772945763409234e-06, "loss": 0.8678, "step": 17937 }, { "epoch": 0.01, "learning_rate": 7.3762544422373875e-06, "loss": 0.8971, "step": 17938 }, { "epoch": 0.01, "learning_rate": 7.3752143386172594e-06, "loss": 0.8929, "step": 17939 }, { "epoch": 0.01, "learning_rate": 7.374174265492627e-06, "loss": 0.8819, "step": 17940 }, { "epoch": 0.01, "learning_rate": 7.373134222875571e-06, "loss": 0.8781, "step": 17941 }, { "epoch": 0.01, "learning_rate": 7.372094210778177e-06, "loss": 0.8905, "step": 17942 }, { "epoch": 0.01, "learning_rate": 7.371054229212524e-06, "loss": 0.8837, "step": 17943 }, { "epoch": 0.01, "learning_rate": 7.3700142781907045e-06, "loss": 0.896, "step": 17944 }, { "epoch": 0.01, "learning_rate": 7.368974357724789e-06, "loss": 0.8512, "step": 17945 }, { "epoch": 0.01, "learning_rate": 7.367934467826865e-06, "loss": 0.8611, "step": 17946 }, { "epoch": 0.01, "learning_rate": 7.36689460850902e-06, "loss": 0.8694, "step": 17947 }, { "epoch": 0.01, "learning_rate": 7.365854779783324e-06, "loss": 0.8721, "step": 17948 }, { "epoch": 0.01, "learning_rate": 7.364814981661863e-06, "loss": 0.8503, "step": 17949 }, { "epoch": 0.01, "learning_rate": 7.363775214156721e-06, "loss": 0.9129, "step": 17950 }, { "epoch": 0.01, "learning_rate": 7.362735477279976e-06, "loss": 0.8632, "step": 17951 }, { "epoch": 0.01, "learning_rate": 7.361695771043708e-06, "loss": 0.9166, "step": 17952 }, { "epoch": 0.01, "learning_rate": 7.360656095459995e-06, "loss": 0.8692, "step": 17953 }, { "epoch": 0.01, "learning_rate": 7.359616450540919e-06, "loss": 0.8864, "step": 17954 }, { "epoch": 0.01, "learning_rate": 7.358576836298554e-06, "loss": 0.8521, "step": 17955 }, { "epoch": 0.01, "learning_rate": 7.357537252744984e-06, "loss": 0.854, "step": 17956 }, { "epoch": 0.01, "learning_rate": 7.356497699892284e-06, "loss": 0.868, "step": 17957 }, { "epoch": 0.01, "learning_rate": 7.355458177752537e-06, "loss": 0.8611, "step": 17958 }, { "epoch": 0.01, "learning_rate": 7.354418686337812e-06, "loss": 0.8701, "step": 17959 }, { "epoch": 0.01, "learning_rate": 7.353379225660191e-06, "loss": 0.8595, "step": 17960 }, { "epoch": 0.01, "learning_rate": 7.35233979573175e-06, "loss": 0.8859, "step": 17961 }, { "epoch": 0.01, "learning_rate": 7.3513003965645665e-06, "loss": 0.889, "step": 17962 }, { "epoch": 0.01, "learning_rate": 7.3502610281707155e-06, "loss": 0.8773, "step": 17963 }, { "epoch": 0.01, "learning_rate": 7.349221690562269e-06, "loss": 0.8825, "step": 17964 }, { "epoch": 0.01, "learning_rate": 7.348182383751314e-06, "loss": 0.8639, "step": 17965 }, { "epoch": 0.01, "learning_rate": 7.347143107749912e-06, "loss": 0.9013, "step": 17966 }, { "epoch": 0.01, "learning_rate": 7.346103862570142e-06, "loss": 0.8793, "step": 17967 }, { "epoch": 0.01, "learning_rate": 7.345064648224081e-06, "loss": 0.8258, "step": 17968 }, { "epoch": 0.01, "learning_rate": 7.344025464723805e-06, "loss": 0.8528, "step": 17969 }, { "epoch": 0.01, "learning_rate": 7.34298631208138e-06, "loss": 0.93, "step": 17970 }, { "epoch": 0.01, "learning_rate": 7.341947190308883e-06, "loss": 0.8647, "step": 17971 }, { "epoch": 0.01, "learning_rate": 7.340908099418389e-06, "loss": 0.8825, "step": 17972 }, { "epoch": 0.01, "learning_rate": 7.339869039421965e-06, "loss": 0.8501, "step": 17973 }, { "epoch": 0.01, "learning_rate": 7.338830010331686e-06, "loss": 0.8566, "step": 17974 }, { "epoch": 0.01, "learning_rate": 7.337791012159625e-06, "loss": 0.872, "step": 17975 }, { "epoch": 0.01, "learning_rate": 7.336752044917855e-06, "loss": 0.886, "step": 17976 }, { "epoch": 0.01, "learning_rate": 7.335713108618443e-06, "loss": 0.9287, "step": 17977 }, { "epoch": 0.01, "learning_rate": 7.3346742032734595e-06, "loss": 0.8399, "step": 17978 }, { "epoch": 0.01, "learning_rate": 7.333635328894979e-06, "loss": 0.8964, "step": 17979 }, { "epoch": 0.01, "learning_rate": 7.332596485495066e-06, "loss": 0.8765, "step": 17980 }, { "epoch": 0.01, "learning_rate": 7.331557673085794e-06, "loss": 0.9043, "step": 17981 }, { "epoch": 0.01, "learning_rate": 7.33051889167923e-06, "loss": 0.9353, "step": 17982 }, { "epoch": 0.01, "learning_rate": 7.329480141287446e-06, "loss": 0.8804, "step": 17983 }, { "epoch": 0.01, "learning_rate": 7.328441421922507e-06, "loss": 0.8721, "step": 17984 }, { "epoch": 0.01, "learning_rate": 7.327402733596482e-06, "loss": 0.8435, "step": 17985 }, { "epoch": 0.01, "learning_rate": 7.326364076321442e-06, "loss": 0.8847, "step": 17986 }, { "epoch": 0.01, "learning_rate": 7.325325450109449e-06, "loss": 0.8562, "step": 17987 }, { "epoch": 0.01, "learning_rate": 7.324286854972573e-06, "loss": 0.8653, "step": 17988 }, { "epoch": 0.01, "learning_rate": 7.32324829092288e-06, "loss": 0.8796, "step": 17989 }, { "epoch": 0.01, "learning_rate": 7.32220975797244e-06, "loss": 0.8739, "step": 17990 }, { "epoch": 0.01, "learning_rate": 7.321171256133314e-06, "loss": 0.8167, "step": 17991 }, { "epoch": 0.01, "learning_rate": 7.320132785417566e-06, "loss": 0.8308, "step": 17992 }, { "epoch": 0.01, "learning_rate": 7.319094345837271e-06, "loss": 0.8631, "step": 17993 }, { "epoch": 0.01, "learning_rate": 7.318055937404481e-06, "loss": 0.9116, "step": 17994 }, { "epoch": 0.01, "learning_rate": 7.3170175601312695e-06, "loss": 0.8363, "step": 17995 }, { "epoch": 0.01, "learning_rate": 7.315979214029698e-06, "loss": 0.8692, "step": 17996 }, { "epoch": 0.01, "learning_rate": 7.314940899111833e-06, "loss": 0.8591, "step": 17997 }, { "epoch": 0.01, "learning_rate": 7.313902615389733e-06, "loss": 0.9067, "step": 17998 }, { "epoch": 0.01, "learning_rate": 7.312864362875464e-06, "loss": 0.92, "step": 17999 }, { "epoch": 0.01, "learning_rate": 7.311826141581089e-06, "loss": 0.883, "step": 18000 }, { "epoch": 0.01, "learning_rate": 7.310787951518668e-06, "loss": 0.8869, "step": 18001 }, { "epoch": 0.01, "learning_rate": 7.309749792700265e-06, "loss": 0.8774, "step": 18002 }, { "epoch": 0.01, "learning_rate": 7.30871166513794e-06, "loss": 0.8864, "step": 18003 }, { "epoch": 0.01, "learning_rate": 7.3076735688437585e-06, "loss": 0.8876, "step": 18004 }, { "epoch": 0.01, "learning_rate": 7.306635503829776e-06, "loss": 0.8601, "step": 18005 }, { "epoch": 0.01, "learning_rate": 7.305597470108055e-06, "loss": 0.8799, "step": 18006 }, { "epoch": 0.01, "learning_rate": 7.304559467690656e-06, "loss": 0.8613, "step": 18007 }, { "epoch": 0.01, "learning_rate": 7.3035214965896405e-06, "loss": 0.8931, "step": 18008 }, { "epoch": 0.01, "learning_rate": 7.302483556817065e-06, "loss": 0.8451, "step": 18009 }, { "epoch": 0.01, "learning_rate": 7.301445648384986e-06, "loss": 0.878, "step": 18010 }, { "epoch": 0.01, "learning_rate": 7.300407771305475e-06, "loss": 0.88, "step": 18011 }, { "epoch": 0.01, "learning_rate": 7.299369925590575e-06, "loss": 0.8741, "step": 18012 }, { "epoch": 0.01, "learning_rate": 7.298332111252348e-06, "loss": 0.8988, "step": 18013 }, { "epoch": 0.01, "learning_rate": 7.297294328302858e-06, "loss": 0.9149, "step": 18014 }, { "epoch": 0.01, "learning_rate": 7.296256576754161e-06, "loss": 0.892, "step": 18015 }, { "epoch": 0.01, "learning_rate": 7.295218856618304e-06, "loss": 0.8622, "step": 18016 }, { "epoch": 0.01, "learning_rate": 7.294181167907354e-06, "loss": 0.849, "step": 18017 }, { "epoch": 0.01, "learning_rate": 7.293143510633367e-06, "loss": 0.8928, "step": 18018 }, { "epoch": 0.01, "learning_rate": 7.292105884808389e-06, "loss": 0.8412, "step": 18019 }, { "epoch": 0.01, "learning_rate": 7.291068290444485e-06, "loss": 0.859, "step": 18020 }, { "epoch": 0.01, "learning_rate": 7.290030727553708e-06, "loss": 0.8598, "step": 18021 }, { "epoch": 0.01, "learning_rate": 7.288993196148112e-06, "loss": 0.8882, "step": 18022 }, { "epoch": 0.01, "learning_rate": 7.28795569623975e-06, "loss": 0.8384, "step": 18023 }, { "epoch": 0.01, "learning_rate": 7.286918227840676e-06, "loss": 0.8853, "step": 18024 }, { "epoch": 0.01, "learning_rate": 7.285880790962947e-06, "loss": 0.8574, "step": 18025 }, { "epoch": 0.01, "learning_rate": 7.284843385618613e-06, "loss": 0.8712, "step": 18026 }, { "epoch": 0.01, "learning_rate": 7.283806011819727e-06, "loss": 0.8932, "step": 18027 }, { "epoch": 0.01, "learning_rate": 7.282768669578342e-06, "loss": 0.8756, "step": 18028 }, { "epoch": 0.01, "learning_rate": 7.281731358906512e-06, "loss": 0.8714, "step": 18029 }, { "epoch": 0.01, "learning_rate": 7.2806940798162855e-06, "loss": 0.8525, "step": 18030 }, { "epoch": 0.01, "learning_rate": 7.279656832319715e-06, "loss": 0.8569, "step": 18031 }, { "epoch": 0.01, "learning_rate": 7.2786196164288546e-06, "loss": 0.9023, "step": 18032 }, { "epoch": 0.01, "learning_rate": 7.27758243215575e-06, "loss": 0.8557, "step": 18033 }, { "epoch": 0.01, "learning_rate": 7.276545279512455e-06, "loss": 0.8694, "step": 18034 }, { "epoch": 0.01, "learning_rate": 7.275508158511015e-06, "loss": 0.8728, "step": 18035 }, { "epoch": 0.01, "learning_rate": 7.274471069163489e-06, "loss": 0.8652, "step": 18036 }, { "epoch": 0.01, "learning_rate": 7.273434011481917e-06, "loss": 0.8634, "step": 18037 }, { "epoch": 0.01, "learning_rate": 7.2723969854783496e-06, "loss": 0.8612, "step": 18038 }, { "epoch": 0.01, "learning_rate": 7.271359991164842e-06, "loss": 0.8861, "step": 18039 }, { "epoch": 0.01, "learning_rate": 7.27032302855343e-06, "loss": 0.8639, "step": 18040 }, { "epoch": 0.01, "learning_rate": 7.269286097656171e-06, "loss": 0.8767, "step": 18041 }, { "epoch": 0.01, "learning_rate": 7.2682491984851086e-06, "loss": 0.8543, "step": 18042 }, { "epoch": 0.01, "learning_rate": 7.2672123310522946e-06, "loss": 0.8645, "step": 18043 }, { "epoch": 0.01, "learning_rate": 7.266175495369769e-06, "loss": 0.824, "step": 18044 }, { "epoch": 0.01, "learning_rate": 7.265138691449581e-06, "loss": 0.893, "step": 18045 }, { "epoch": 0.01, "learning_rate": 7.2641019193037785e-06, "loss": 0.8815, "step": 18046 }, { "epoch": 0.01, "learning_rate": 7.263065178944403e-06, "loss": 0.8569, "step": 18047 }, { "epoch": 0.01, "learning_rate": 7.262028470383501e-06, "loss": 0.8632, "step": 18048 }, { "epoch": 0.01, "learning_rate": 7.260991793633119e-06, "loss": 0.8422, "step": 18049 }, { "epoch": 0.01, "learning_rate": 7.2599551487053e-06, "loss": 0.8859, "step": 18050 }, { "epoch": 0.01, "learning_rate": 7.258918535612087e-06, "loss": 0.8481, "step": 18051 }, { "epoch": 0.01, "learning_rate": 7.257881954365525e-06, "loss": 0.8747, "step": 18052 }, { "epoch": 0.01, "learning_rate": 7.256845404977658e-06, "loss": 0.8191, "step": 18053 }, { "epoch": 0.01, "learning_rate": 7.255808887460529e-06, "loss": 0.8658, "step": 18054 }, { "epoch": 0.01, "learning_rate": 7.254772401826178e-06, "loss": 0.8729, "step": 18055 }, { "epoch": 0.01, "learning_rate": 7.2537359480866465e-06, "loss": 0.8855, "step": 18056 }, { "epoch": 0.01, "learning_rate": 7.252699526253986e-06, "loss": 0.9058, "step": 18057 }, { "epoch": 0.01, "learning_rate": 7.251663136340226e-06, "loss": 0.8717, "step": 18058 }, { "epoch": 0.01, "learning_rate": 7.250626778357409e-06, "loss": 0.8555, "step": 18059 }, { "epoch": 0.01, "learning_rate": 7.249590452317581e-06, "loss": 0.8863, "step": 18060 }, { "epoch": 0.01, "learning_rate": 7.248554158232785e-06, "loss": 0.8563, "step": 18061 }, { "epoch": 0.01, "learning_rate": 7.24751789611505e-06, "loss": 0.8586, "step": 18062 }, { "epoch": 0.01, "learning_rate": 7.246481665976424e-06, "loss": 0.8767, "step": 18063 }, { "epoch": 0.01, "learning_rate": 7.245445467828946e-06, "loss": 0.8755, "step": 18064 }, { "epoch": 0.01, "learning_rate": 7.2444093016846485e-06, "loss": 0.8483, "step": 18065 }, { "epoch": 0.01, "learning_rate": 7.243373167555576e-06, "loss": 0.8949, "step": 18066 }, { "epoch": 0.01, "learning_rate": 7.242337065453764e-06, "loss": 0.8617, "step": 18067 }, { "epoch": 0.01, "learning_rate": 7.241300995391254e-06, "loss": 0.9066, "step": 18068 }, { "epoch": 0.01, "learning_rate": 7.240264957380078e-06, "loss": 0.8411, "step": 18069 }, { "epoch": 0.01, "learning_rate": 7.239228951432275e-06, "loss": 0.9014, "step": 18070 }, { "epoch": 0.01, "learning_rate": 7.2381929775598835e-06, "loss": 0.8677, "step": 18071 }, { "epoch": 0.01, "learning_rate": 7.237157035774938e-06, "loss": 0.8773, "step": 18072 }, { "epoch": 0.01, "learning_rate": 7.236121126089473e-06, "loss": 0.8759, "step": 18073 }, { "epoch": 0.01, "learning_rate": 7.235085248515526e-06, "loss": 0.8894, "step": 18074 }, { "epoch": 0.01, "learning_rate": 7.234049403065133e-06, "loss": 0.8535, "step": 18075 }, { "epoch": 0.01, "learning_rate": 7.2330135897503264e-06, "loss": 0.8851, "step": 18076 }, { "epoch": 0.01, "learning_rate": 7.231977808583142e-06, "loss": 0.8865, "step": 18077 }, { "epoch": 0.01, "learning_rate": 7.230942059575614e-06, "loss": 0.8855, "step": 18078 }, { "epoch": 0.01, "learning_rate": 7.229906342739774e-06, "loss": 0.8615, "step": 18079 }, { "epoch": 0.01, "learning_rate": 7.2288706580876565e-06, "loss": 0.8525, "step": 18080 }, { "epoch": 0.01, "learning_rate": 7.2278350056312914e-06, "loss": 0.8868, "step": 18081 }, { "epoch": 0.01, "learning_rate": 7.226799385382723e-06, "loss": 0.8882, "step": 18082 }, { "epoch": 0.01, "learning_rate": 7.225763797353966e-06, "loss": 0.8849, "step": 18083 }, { "epoch": 0.01, "learning_rate": 7.2247282415570655e-06, "loss": 0.8882, "step": 18084 }, { "epoch": 0.01, "learning_rate": 7.2236927180040505e-06, "loss": 0.8576, "step": 18085 }, { "epoch": 0.01, "learning_rate": 7.222657226706944e-06, "loss": 0.8682, "step": 18086 }, { "epoch": 0.01, "learning_rate": 7.221621767677784e-06, "loss": 0.8934, "step": 18087 }, { "epoch": 0.01, "learning_rate": 7.2205863409286005e-06, "loss": 0.8653, "step": 18088 }, { "epoch": 0.01, "learning_rate": 7.219550946471423e-06, "loss": 0.8596, "step": 18089 }, { "epoch": 0.01, "learning_rate": 7.218515584318277e-06, "loss": 0.8958, "step": 18090 }, { "epoch": 0.01, "learning_rate": 7.217480254481197e-06, "loss": 0.8725, "step": 18091 }, { "epoch": 0.01, "learning_rate": 7.216444956972211e-06, "loss": 0.8743, "step": 18092 }, { "epoch": 0.01, "learning_rate": 7.215409691803342e-06, "loss": 0.932, "step": 18093 }, { "epoch": 0.01, "learning_rate": 7.214374458986624e-06, "loss": 0.857, "step": 18094 }, { "epoch": 0.01, "learning_rate": 7.21333925853408e-06, "loss": 0.8678, "step": 18095 }, { "epoch": 0.01, "learning_rate": 7.212304090457742e-06, "loss": 0.8294, "step": 18096 }, { "epoch": 0.01, "learning_rate": 7.211268954769635e-06, "loss": 0.8776, "step": 18097 }, { "epoch": 0.01, "learning_rate": 7.210233851481782e-06, "loss": 0.8568, "step": 18098 }, { "epoch": 0.01, "learning_rate": 7.209198780606213e-06, "loss": 0.8815, "step": 18099 }, { "epoch": 0.01, "learning_rate": 7.208163742154957e-06, "loss": 0.8392, "step": 18100 }, { "epoch": 0.01, "learning_rate": 7.207128736140032e-06, "loss": 0.8381, "step": 18101 }, { "epoch": 0.01, "learning_rate": 7.206093762573463e-06, "loss": 0.8407, "step": 18102 }, { "epoch": 0.01, "learning_rate": 7.2050588214672855e-06, "loss": 0.8992, "step": 18103 }, { "epoch": 0.01, "learning_rate": 7.204023912833511e-06, "loss": 0.8882, "step": 18104 }, { "epoch": 0.01, "learning_rate": 7.202989036684167e-06, "loss": 0.8811, "step": 18105 }, { "epoch": 0.01, "learning_rate": 7.20195419303128e-06, "loss": 0.898, "step": 18106 }, { "epoch": 0.01, "learning_rate": 7.200919381886875e-06, "loss": 0.8389, "step": 18107 }, { "epoch": 0.01, "learning_rate": 7.199884603262965e-06, "loss": 0.9053, "step": 18108 }, { "epoch": 0.01, "learning_rate": 7.198849857171583e-06, "loss": 0.8765, "step": 18109 }, { "epoch": 0.01, "learning_rate": 7.197815143624749e-06, "loss": 0.8647, "step": 18110 }, { "epoch": 0.01, "learning_rate": 7.196780462634476e-06, "loss": 0.8598, "step": 18111 }, { "epoch": 0.01, "learning_rate": 7.195745814212793e-06, "loss": 0.8662, "step": 18112 }, { "epoch": 0.01, "learning_rate": 7.19471119837172e-06, "loss": 0.8686, "step": 18113 }, { "epoch": 0.01, "learning_rate": 7.1936766151232795e-06, "loss": 0.9108, "step": 18114 }, { "epoch": 0.01, "learning_rate": 7.192642064479485e-06, "loss": 0.8671, "step": 18115 }, { "epoch": 0.01, "learning_rate": 7.191607546452361e-06, "loss": 0.9186, "step": 18116 }, { "epoch": 0.01, "learning_rate": 7.190573061053927e-06, "loss": 0.8591, "step": 18117 }, { "epoch": 0.01, "learning_rate": 7.1895386082962e-06, "loss": 0.8788, "step": 18118 }, { "epoch": 0.01, "learning_rate": 7.188504188191198e-06, "loss": 0.8413, "step": 18119 }, { "epoch": 0.01, "learning_rate": 7.18746980075094e-06, "loss": 0.8338, "step": 18120 }, { "epoch": 0.01, "learning_rate": 7.186435445987447e-06, "loss": 0.8508, "step": 18121 }, { "epoch": 0.01, "learning_rate": 7.185401123912732e-06, "loss": 0.8812, "step": 18122 }, { "epoch": 0.01, "learning_rate": 7.184366834538814e-06, "loss": 0.8359, "step": 18123 }, { "epoch": 0.01, "learning_rate": 7.183332577877711e-06, "loss": 0.8517, "step": 18124 }, { "epoch": 0.01, "learning_rate": 7.182298353941435e-06, "loss": 0.8679, "step": 18125 }, { "epoch": 0.01, "learning_rate": 7.1812641627420055e-06, "loss": 0.8679, "step": 18126 }, { "epoch": 0.01, "learning_rate": 7.180230004291433e-06, "loss": 0.8862, "step": 18127 }, { "epoch": 0.01, "learning_rate": 7.179195878601746e-06, "loss": 0.847, "step": 18128 }, { "epoch": 0.01, "learning_rate": 7.178161785684942e-06, "loss": 0.8265, "step": 18129 }, { "epoch": 0.01, "learning_rate": 7.1771277255530456e-06, "loss": 0.8986, "step": 18130 }, { "epoch": 0.01, "learning_rate": 7.176093698218072e-06, "loss": 0.8471, "step": 18131 }, { "epoch": 0.01, "learning_rate": 7.175059703692027e-06, "loss": 0.8435, "step": 18132 }, { "epoch": 0.01, "learning_rate": 7.174025741986928e-06, "loss": 0.8506, "step": 18133 }, { "epoch": 0.01, "learning_rate": 7.1729918131147895e-06, "loss": 0.86, "step": 18134 }, { "epoch": 0.01, "learning_rate": 7.1719579170876245e-06, "loss": 0.8555, "step": 18135 }, { "epoch": 0.01, "learning_rate": 7.170924053917441e-06, "loss": 0.8359, "step": 18136 }, { "epoch": 0.01, "learning_rate": 7.1698902236162524e-06, "loss": 0.847, "step": 18137 }, { "epoch": 0.01, "learning_rate": 7.168856426196074e-06, "loss": 0.8584, "step": 18138 }, { "epoch": 0.01, "learning_rate": 7.16782266166891e-06, "loss": 0.8661, "step": 18139 }, { "epoch": 0.01, "learning_rate": 7.166788930046773e-06, "loss": 0.8835, "step": 18140 }, { "epoch": 0.01, "learning_rate": 7.165755231341675e-06, "loss": 0.8772, "step": 18141 }, { "epoch": 0.01, "learning_rate": 7.164721565565629e-06, "loss": 0.8898, "step": 18142 }, { "epoch": 0.01, "learning_rate": 7.163687932730638e-06, "loss": 0.8412, "step": 18143 }, { "epoch": 0.01, "learning_rate": 7.1626543328487134e-06, "loss": 0.8698, "step": 18144 }, { "epoch": 0.01, "learning_rate": 7.161620765931862e-06, "loss": 0.8589, "step": 18145 }, { "epoch": 0.01, "learning_rate": 7.1605872319921e-06, "loss": 0.8796, "step": 18146 }, { "epoch": 0.01, "learning_rate": 7.159553731041424e-06, "loss": 0.8782, "step": 18147 }, { "epoch": 0.01, "learning_rate": 7.158520263091847e-06, "loss": 0.866, "step": 18148 }, { "epoch": 0.01, "learning_rate": 7.157486828155382e-06, "loss": 0.8336, "step": 18149 }, { "epoch": 0.01, "learning_rate": 7.156453426244025e-06, "loss": 0.9034, "step": 18150 }, { "epoch": 0.01, "learning_rate": 7.155420057369785e-06, "loss": 0.9031, "step": 18151 }, { "epoch": 0.01, "learning_rate": 7.154386721544673e-06, "loss": 0.865, "step": 18152 }, { "epoch": 0.01, "learning_rate": 7.153353418780695e-06, "loss": 0.8742, "step": 18153 }, { "epoch": 0.01, "learning_rate": 7.152320149089847e-06, "loss": 0.8915, "step": 18154 }, { "epoch": 0.01, "learning_rate": 7.151286912484141e-06, "loss": 0.8583, "step": 18155 }, { "epoch": 0.01, "learning_rate": 7.150253708975583e-06, "loss": 0.9203, "step": 18156 }, { "epoch": 0.01, "learning_rate": 7.149220538576174e-06, "loss": 0.8752, "step": 18157 }, { "epoch": 0.01, "learning_rate": 7.148187401297915e-06, "loss": 0.8631, "step": 18158 }, { "epoch": 0.01, "learning_rate": 7.147154297152813e-06, "loss": 0.8928, "step": 18159 }, { "epoch": 0.01, "learning_rate": 7.146121226152872e-06, "loss": 0.9066, "step": 18160 }, { "epoch": 0.01, "learning_rate": 7.145088188310092e-06, "loss": 0.8817, "step": 18161 }, { "epoch": 0.01, "learning_rate": 7.144055183636474e-06, "loss": 0.8702, "step": 18162 }, { "epoch": 0.01, "learning_rate": 7.143022212144024e-06, "loss": 0.8428, "step": 18163 }, { "epoch": 0.01, "learning_rate": 7.141989273844738e-06, "loss": 0.8572, "step": 18164 }, { "epoch": 0.01, "learning_rate": 7.14095636875062e-06, "loss": 0.9121, "step": 18165 }, { "epoch": 0.01, "learning_rate": 7.139923496873671e-06, "loss": 0.9093, "step": 18166 }, { "epoch": 0.01, "learning_rate": 7.138890658225892e-06, "loss": 0.9106, "step": 18167 }, { "epoch": 0.01, "learning_rate": 7.137857852819279e-06, "loss": 0.8522, "step": 18168 }, { "epoch": 0.01, "learning_rate": 7.136825080665834e-06, "loss": 0.8822, "step": 18169 }, { "epoch": 0.01, "learning_rate": 7.135792341777559e-06, "loss": 0.9088, "step": 18170 }, { "epoch": 0.01, "learning_rate": 7.134759636166446e-06, "loss": 0.8857, "step": 18171 }, { "epoch": 0.01, "learning_rate": 7.133726963844495e-06, "loss": 0.8611, "step": 18172 }, { "epoch": 0.01, "learning_rate": 7.132694324823708e-06, "loss": 0.8224, "step": 18173 }, { "epoch": 0.01, "learning_rate": 7.131661719116083e-06, "loss": 0.8674, "step": 18174 }, { "epoch": 0.01, "learning_rate": 7.130629146733609e-06, "loss": 0.8536, "step": 18175 }, { "epoch": 0.01, "learning_rate": 7.129596607688289e-06, "loss": 0.8256, "step": 18176 }, { "epoch": 0.01, "learning_rate": 7.128564101992124e-06, "loss": 0.836, "step": 18177 }, { "epoch": 0.01, "learning_rate": 7.127531629657096e-06, "loss": 0.8316, "step": 18178 }, { "epoch": 0.01, "learning_rate": 7.126499190695211e-06, "loss": 0.8666, "step": 18179 }, { "epoch": 0.01, "learning_rate": 7.1254667851184625e-06, "loss": 0.8982, "step": 18180 }, { "epoch": 0.01, "learning_rate": 7.124434412938847e-06, "loss": 0.8193, "step": 18181 }, { "epoch": 0.01, "learning_rate": 7.123402074168353e-06, "loss": 0.8866, "step": 18182 }, { "epoch": 0.01, "learning_rate": 7.1223697688189794e-06, "loss": 0.8975, "step": 18183 }, { "epoch": 0.01, "learning_rate": 7.121337496902721e-06, "loss": 0.8734, "step": 18184 }, { "epoch": 0.01, "learning_rate": 7.120305258431565e-06, "loss": 0.8437, "step": 18185 }, { "epoch": 0.01, "learning_rate": 7.119273053417508e-06, "loss": 0.8424, "step": 18186 }, { "epoch": 0.01, "learning_rate": 7.1182408818725425e-06, "loss": 0.8527, "step": 18187 }, { "epoch": 0.01, "learning_rate": 7.117208743808661e-06, "loss": 0.8616, "step": 18188 }, { "epoch": 0.01, "learning_rate": 7.116176639237853e-06, "loss": 0.8578, "step": 18189 }, { "epoch": 0.01, "learning_rate": 7.115144568172112e-06, "loss": 0.8903, "step": 18190 }, { "epoch": 0.01, "learning_rate": 7.114112530623425e-06, "loss": 0.8947, "step": 18191 }, { "epoch": 0.01, "learning_rate": 7.113080526603793e-06, "loss": 0.8104, "step": 18192 }, { "epoch": 0.01, "learning_rate": 7.1120485561251925e-06, "loss": 0.9144, "step": 18193 }, { "epoch": 0.01, "learning_rate": 7.111016619199619e-06, "loss": 0.8937, "step": 18194 }, { "epoch": 0.01, "learning_rate": 7.109984715839067e-06, "loss": 0.8747, "step": 18195 }, { "epoch": 0.01, "learning_rate": 7.1089528460555165e-06, "loss": 0.863, "step": 18196 }, { "epoch": 0.01, "learning_rate": 7.1079210098609595e-06, "loss": 0.8554, "step": 18197 }, { "epoch": 0.01, "learning_rate": 7.106889207267387e-06, "loss": 0.8722, "step": 18198 }, { "epoch": 0.01, "learning_rate": 7.1058574382867876e-06, "loss": 0.8752, "step": 18199 }, { "epoch": 0.01, "learning_rate": 7.1048257029311395e-06, "loss": 0.8662, "step": 18200 }, { "epoch": 0.02, "learning_rate": 7.103794001212439e-06, "loss": 0.8789, "step": 18201 }, { "epoch": 0.02, "learning_rate": 7.10276233314267e-06, "loss": 0.8792, "step": 18202 }, { "epoch": 0.02, "learning_rate": 7.101730698733817e-06, "loss": 0.8554, "step": 18203 }, { "epoch": 0.02, "learning_rate": 7.1006990979978675e-06, "loss": 0.8819, "step": 18204 }, { "epoch": 0.02, "learning_rate": 7.099667530946806e-06, "loss": 0.8688, "step": 18205 }, { "epoch": 0.02, "learning_rate": 7.09863599759262e-06, "loss": 0.8762, "step": 18206 }, { "epoch": 0.02, "learning_rate": 7.097604497947291e-06, "loss": 0.835, "step": 18207 }, { "epoch": 0.02, "learning_rate": 7.096573032022804e-06, "loss": 0.8823, "step": 18208 }, { "epoch": 0.02, "learning_rate": 7.095541599831146e-06, "loss": 0.8419, "step": 18209 }, { "epoch": 0.02, "learning_rate": 7.094510201384295e-06, "loss": 0.8799, "step": 18210 }, { "epoch": 0.02, "learning_rate": 7.0934788366942365e-06, "loss": 0.8802, "step": 18211 }, { "epoch": 0.02, "learning_rate": 7.092447505772955e-06, "loss": 0.8415, "step": 18212 }, { "epoch": 0.02, "learning_rate": 7.091416208632431e-06, "loss": 0.865, "step": 18213 }, { "epoch": 0.02, "learning_rate": 7.090384945284646e-06, "loss": 0.8782, "step": 18214 }, { "epoch": 0.02, "learning_rate": 7.089353715741583e-06, "loss": 0.8336, "step": 18215 }, { "epoch": 0.02, "learning_rate": 7.088322520015223e-06, "loss": 0.8439, "step": 18216 }, { "epoch": 0.02, "learning_rate": 7.087291358117545e-06, "loss": 0.8394, "step": 18217 }, { "epoch": 0.02, "learning_rate": 7.086260230060529e-06, "loss": 0.905, "step": 18218 }, { "epoch": 0.02, "learning_rate": 7.085229135856156e-06, "loss": 0.8803, "step": 18219 }, { "epoch": 0.02, "learning_rate": 7.084198075516412e-06, "loss": 0.8458, "step": 18220 }, { "epoch": 0.02, "learning_rate": 7.0831670490532636e-06, "loss": 0.8301, "step": 18221 }, { "epoch": 0.02, "learning_rate": 7.082136056478697e-06, "loss": 0.8678, "step": 18222 }, { "epoch": 0.02, "learning_rate": 7.081105097804695e-06, "loss": 0.8562, "step": 18223 }, { "epoch": 0.02, "learning_rate": 7.080074173043223e-06, "loss": 0.873, "step": 18224 }, { "epoch": 0.02, "learning_rate": 7.079043282206268e-06, "loss": 0.8487, "step": 18225 }, { "epoch": 0.02, "learning_rate": 7.078012425305805e-06, "loss": 0.8563, "step": 18226 }, { "epoch": 0.02, "learning_rate": 7.076981602353812e-06, "loss": 0.8826, "step": 18227 }, { "epoch": 0.02, "learning_rate": 7.075950813362262e-06, "loss": 0.8352, "step": 18228 }, { "epoch": 0.02, "learning_rate": 7.074920058343134e-06, "loss": 0.8782, "step": 18229 }, { "epoch": 0.02, "learning_rate": 7.073889337308404e-06, "loss": 0.8614, "step": 18230 }, { "epoch": 0.02, "learning_rate": 7.0728586502700445e-06, "loss": 0.878, "step": 18231 }, { "epoch": 0.02, "learning_rate": 7.071827997240031e-06, "loss": 0.8717, "step": 18232 }, { "epoch": 0.02, "learning_rate": 7.0707973782303385e-06, "loss": 0.8848, "step": 18233 }, { "epoch": 0.02, "learning_rate": 7.069766793252943e-06, "loss": 0.8604, "step": 18234 }, { "epoch": 0.02, "learning_rate": 7.068736242319816e-06, "loss": 0.8284, "step": 18235 }, { "epoch": 0.02, "learning_rate": 7.067705725442928e-06, "loss": 0.8498, "step": 18236 }, { "epoch": 0.02, "learning_rate": 7.066675242634254e-06, "loss": 0.8053, "step": 18237 }, { "epoch": 0.02, "learning_rate": 7.065644793905774e-06, "loss": 0.8285, "step": 18238 }, { "epoch": 0.02, "learning_rate": 7.064614379269448e-06, "loss": 0.8746, "step": 18239 }, { "epoch": 0.02, "learning_rate": 7.063583998737251e-06, "loss": 0.8675, "step": 18240 }, { "epoch": 0.02, "learning_rate": 7.0625536523211635e-06, "loss": 0.8736, "step": 18241 }, { "epoch": 0.02, "learning_rate": 7.061523340033144e-06, "loss": 0.8789, "step": 18242 }, { "epoch": 0.02, "learning_rate": 7.060493061885165e-06, "loss": 0.8282, "step": 18243 }, { "epoch": 0.02, "learning_rate": 7.059462817889203e-06, "loss": 0.8565, "step": 18244 }, { "epoch": 0.02, "learning_rate": 7.058432608057226e-06, "loss": 0.8069, "step": 18245 }, { "epoch": 0.02, "learning_rate": 7.057402432401199e-06, "loss": 0.8538, "step": 18246 }, { "epoch": 0.02, "learning_rate": 7.056372290933093e-06, "loss": 0.8514, "step": 18247 }, { "epoch": 0.02, "learning_rate": 7.05534218366488e-06, "loss": 0.8718, "step": 18248 }, { "epoch": 0.02, "learning_rate": 7.054312110608521e-06, "loss": 0.8799, "step": 18249 }, { "epoch": 0.02, "learning_rate": 7.053282071775989e-06, "loss": 0.925, "step": 18250 }, { "epoch": 0.02, "learning_rate": 7.052252067179248e-06, "loss": 0.8813, "step": 18251 }, { "epoch": 0.02, "learning_rate": 7.05122209683027e-06, "loss": 0.8956, "step": 18252 }, { "epoch": 0.02, "learning_rate": 7.050192160741016e-06, "loss": 0.8694, "step": 18253 }, { "epoch": 0.02, "learning_rate": 7.049162258923455e-06, "loss": 0.8339, "step": 18254 }, { "epoch": 0.02, "learning_rate": 7.048132391389554e-06, "loss": 0.9007, "step": 18255 }, { "epoch": 0.02, "learning_rate": 7.047102558151273e-06, "loss": 0.8887, "step": 18256 }, { "epoch": 0.02, "learning_rate": 7.046072759220581e-06, "loss": 0.9083, "step": 18257 }, { "epoch": 0.02, "learning_rate": 7.045042994609441e-06, "loss": 0.9115, "step": 18258 }, { "epoch": 0.02, "learning_rate": 7.04401326432982e-06, "loss": 0.876, "step": 18259 }, { "epoch": 0.02, "learning_rate": 7.042983568393678e-06, "loss": 0.8431, "step": 18260 }, { "epoch": 0.02, "learning_rate": 7.041953906812976e-06, "loss": 0.8518, "step": 18261 }, { "epoch": 0.02, "learning_rate": 7.040924279599688e-06, "loss": 0.8407, "step": 18262 }, { "epoch": 0.02, "learning_rate": 7.039894686765765e-06, "loss": 0.8725, "step": 18263 }, { "epoch": 0.02, "learning_rate": 7.03886512832317e-06, "loss": 0.8809, "step": 18264 }, { "epoch": 0.02, "learning_rate": 7.037835604283871e-06, "loss": 0.8643, "step": 18265 }, { "epoch": 0.02, "learning_rate": 7.03680611465983e-06, "loss": 0.8567, "step": 18266 }, { "epoch": 0.02, "learning_rate": 7.035776659462998e-06, "loss": 0.8698, "step": 18267 }, { "epoch": 0.02, "learning_rate": 7.034747238705344e-06, "loss": 0.8632, "step": 18268 }, { "epoch": 0.02, "learning_rate": 7.033717852398829e-06, "loss": 0.8739, "step": 18269 }, { "epoch": 0.02, "learning_rate": 7.032688500555403e-06, "loss": 0.8714, "step": 18270 }, { "epoch": 0.02, "learning_rate": 7.031659183187034e-06, "loss": 0.868, "step": 18271 }, { "epoch": 0.02, "learning_rate": 7.030629900305679e-06, "loss": 0.8894, "step": 18272 }, { "epoch": 0.02, "learning_rate": 7.0296006519232985e-06, "loss": 0.8621, "step": 18273 }, { "epoch": 0.02, "learning_rate": 7.028571438051846e-06, "loss": 0.9268, "step": 18274 }, { "epoch": 0.02, "learning_rate": 7.027542258703279e-06, "loss": 0.8764, "step": 18275 }, { "epoch": 0.02, "learning_rate": 7.026513113889559e-06, "loss": 0.8576, "step": 18276 }, { "epoch": 0.02, "learning_rate": 7.025484003622643e-06, "loss": 0.8734, "step": 18277 }, { "epoch": 0.02, "learning_rate": 7.024454927914484e-06, "loss": 0.8678, "step": 18278 }, { "epoch": 0.02, "learning_rate": 7.02342588677704e-06, "loss": 0.8762, "step": 18279 }, { "epoch": 0.02, "learning_rate": 7.022396880222266e-06, "loss": 0.8608, "step": 18280 }, { "epoch": 0.02, "learning_rate": 7.021367908262118e-06, "loss": 0.8618, "step": 18281 }, { "epoch": 0.02, "learning_rate": 7.020338970908549e-06, "loss": 0.8778, "step": 18282 }, { "epoch": 0.02, "learning_rate": 7.019310068173513e-06, "loss": 0.8552, "step": 18283 }, { "epoch": 0.02, "learning_rate": 7.018281200068973e-06, "loss": 0.874, "step": 18284 }, { "epoch": 0.02, "learning_rate": 7.017252366606871e-06, "loss": 0.823, "step": 18285 }, { "epoch": 0.02, "learning_rate": 7.016223567799163e-06, "loss": 0.8879, "step": 18286 }, { "epoch": 0.02, "learning_rate": 7.015194803657811e-06, "loss": 0.8844, "step": 18287 }, { "epoch": 0.02, "learning_rate": 7.014166074194755e-06, "loss": 0.8924, "step": 18288 }, { "epoch": 0.02, "learning_rate": 7.01313737942195e-06, "loss": 0.8554, "step": 18289 }, { "epoch": 0.02, "learning_rate": 7.012108719351354e-06, "loss": 0.8959, "step": 18290 }, { "epoch": 0.02, "learning_rate": 7.011080093994915e-06, "loss": 0.8422, "step": 18291 }, { "epoch": 0.02, "learning_rate": 7.010051503364581e-06, "loss": 0.8731, "step": 18292 }, { "epoch": 0.02, "learning_rate": 7.009022947472305e-06, "loss": 0.8337, "step": 18293 }, { "epoch": 0.02, "learning_rate": 7.007994426330038e-06, "loss": 0.8798, "step": 18294 }, { "epoch": 0.02, "learning_rate": 7.006965939949728e-06, "loss": 0.8707, "step": 18295 }, { "epoch": 0.02, "learning_rate": 7.005937488343324e-06, "loss": 0.8916, "step": 18296 }, { "epoch": 0.02, "learning_rate": 7.004909071522774e-06, "loss": 0.8242, "step": 18297 }, { "epoch": 0.02, "learning_rate": 7.003880689500031e-06, "loss": 0.8921, "step": 18298 }, { "epoch": 0.02, "learning_rate": 7.002852342287038e-06, "loss": 0.8474, "step": 18299 }, { "epoch": 0.02, "learning_rate": 7.001824029895743e-06, "loss": 0.8859, "step": 18300 }, { "epoch": 0.02, "learning_rate": 7.000795752338099e-06, "loss": 0.8826, "step": 18301 }, { "epoch": 0.02, "learning_rate": 6.9997675096260455e-06, "loss": 0.8409, "step": 18302 }, { "epoch": 0.02, "learning_rate": 6.998739301771531e-06, "loss": 0.8791, "step": 18303 }, { "epoch": 0.02, "learning_rate": 6.9977111287865045e-06, "loss": 0.8622, "step": 18304 }, { "epoch": 0.02, "learning_rate": 6.996682990682909e-06, "loss": 0.8501, "step": 18305 }, { "epoch": 0.02, "learning_rate": 6.995654887472689e-06, "loss": 0.852, "step": 18306 }, { "epoch": 0.02, "learning_rate": 6.994626819167789e-06, "loss": 0.889, "step": 18307 }, { "epoch": 0.02, "learning_rate": 6.993598785780163e-06, "loss": 0.8715, "step": 18308 }, { "epoch": 0.02, "learning_rate": 6.992570787321739e-06, "loss": 0.8575, "step": 18309 }, { "epoch": 0.02, "learning_rate": 6.9915428238044694e-06, "loss": 0.8698, "step": 18310 }, { "epoch": 0.02, "learning_rate": 6.990514895240297e-06, "loss": 0.8656, "step": 18311 }, { "epoch": 0.02, "learning_rate": 6.989487001641169e-06, "loss": 0.846, "step": 18312 }, { "epoch": 0.02, "learning_rate": 6.988459143019015e-06, "loss": 0.8552, "step": 18313 }, { "epoch": 0.02, "learning_rate": 6.987431319385789e-06, "loss": 0.8961, "step": 18314 }, { "epoch": 0.02, "learning_rate": 6.98640353075343e-06, "loss": 0.8809, "step": 18315 }, { "epoch": 0.02, "learning_rate": 6.985375777133873e-06, "loss": 0.8409, "step": 18316 }, { "epoch": 0.02, "learning_rate": 6.984348058539065e-06, "loss": 0.8631, "step": 18317 }, { "epoch": 0.02, "learning_rate": 6.983320374980944e-06, "loss": 0.8994, "step": 18318 }, { "epoch": 0.02, "learning_rate": 6.982292726471453e-06, "loss": 0.8919, "step": 18319 }, { "epoch": 0.02, "learning_rate": 6.981265113022526e-06, "loss": 0.8655, "step": 18320 }, { "epoch": 0.02, "learning_rate": 6.9802375346461045e-06, "loss": 0.8316, "step": 18321 }, { "epoch": 0.02, "learning_rate": 6.979209991354129e-06, "loss": 0.8927, "step": 18322 }, { "epoch": 0.02, "learning_rate": 6.978182483158538e-06, "loss": 0.8647, "step": 18323 }, { "epoch": 0.02, "learning_rate": 6.977155010071266e-06, "loss": 0.8547, "step": 18324 }, { "epoch": 0.02, "learning_rate": 6.976127572104251e-06, "loss": 0.8463, "step": 18325 }, { "epoch": 0.02, "learning_rate": 6.975100169269434e-06, "loss": 0.8722, "step": 18326 }, { "epoch": 0.02, "learning_rate": 6.974072801578749e-06, "loss": 0.8728, "step": 18327 }, { "epoch": 0.02, "learning_rate": 6.973045469044128e-06, "loss": 0.8432, "step": 18328 }, { "epoch": 0.02, "learning_rate": 6.972018171677512e-06, "loss": 0.8613, "step": 18329 }, { "epoch": 0.02, "learning_rate": 6.970990909490843e-06, "loss": 0.8917, "step": 18330 }, { "epoch": 0.02, "learning_rate": 6.9699636824960415e-06, "loss": 0.8652, "step": 18331 }, { "epoch": 0.02, "learning_rate": 6.9689364907050485e-06, "loss": 0.8487, "step": 18332 }, { "epoch": 0.02, "learning_rate": 6.967909334129804e-06, "loss": 0.871, "step": 18333 }, { "epoch": 0.02, "learning_rate": 6.966882212782234e-06, "loss": 0.877, "step": 18334 }, { "epoch": 0.02, "learning_rate": 6.965855126674271e-06, "loss": 0.8476, "step": 18335 }, { "epoch": 0.02, "learning_rate": 6.9648280758178545e-06, "loss": 0.8799, "step": 18336 }, { "epoch": 0.02, "learning_rate": 6.963801060224916e-06, "loss": 0.8395, "step": 18337 }, { "epoch": 0.02, "learning_rate": 6.962774079907384e-06, "loss": 0.8443, "step": 18338 }, { "epoch": 0.02, "learning_rate": 6.9617471348771906e-06, "loss": 0.8877, "step": 18339 }, { "epoch": 0.02, "learning_rate": 6.960720225146272e-06, "loss": 0.8744, "step": 18340 }, { "epoch": 0.02, "learning_rate": 6.959693350726553e-06, "loss": 0.8465, "step": 18341 }, { "epoch": 0.02, "learning_rate": 6.958666511629966e-06, "loss": 0.9087, "step": 18342 }, { "epoch": 0.02, "learning_rate": 6.957639707868442e-06, "loss": 0.8704, "step": 18343 }, { "epoch": 0.02, "learning_rate": 6.956612939453914e-06, "loss": 0.8566, "step": 18344 }, { "epoch": 0.02, "learning_rate": 6.955586206398305e-06, "loss": 0.8455, "step": 18345 }, { "epoch": 0.02, "learning_rate": 6.954559508713545e-06, "loss": 0.8531, "step": 18346 }, { "epoch": 0.02, "learning_rate": 6.953532846411568e-06, "loss": 0.8661, "step": 18347 }, { "epoch": 0.02, "learning_rate": 6.952506219504297e-06, "loss": 0.864, "step": 18348 }, { "epoch": 0.02, "learning_rate": 6.951479628003659e-06, "loss": 0.8477, "step": 18349 }, { "epoch": 0.02, "learning_rate": 6.950453071921583e-06, "loss": 0.8878, "step": 18350 }, { "epoch": 0.02, "learning_rate": 6.9494265512699996e-06, "loss": 0.8637, "step": 18351 }, { "epoch": 0.02, "learning_rate": 6.948400066060828e-06, "loss": 0.8438, "step": 18352 }, { "epoch": 0.02, "learning_rate": 6.947373616305997e-06, "loss": 0.852, "step": 18353 }, { "epoch": 0.02, "learning_rate": 6.9463472020174385e-06, "loss": 0.8532, "step": 18354 }, { "epoch": 0.02, "learning_rate": 6.945320823207067e-06, "loss": 0.8847, "step": 18355 }, { "epoch": 0.02, "learning_rate": 6.94429447988681e-06, "loss": 0.89, "step": 18356 }, { "epoch": 0.02, "learning_rate": 6.943268172068597e-06, "loss": 0.8625, "step": 18357 }, { "epoch": 0.02, "learning_rate": 6.942241899764352e-06, "loss": 0.8691, "step": 18358 }, { "epoch": 0.02, "learning_rate": 6.941215662985992e-06, "loss": 0.8779, "step": 18359 }, { "epoch": 0.02, "learning_rate": 6.940189461745444e-06, "loss": 0.8767, "step": 18360 }, { "epoch": 0.02, "learning_rate": 6.939163296054634e-06, "loss": 0.8386, "step": 18361 }, { "epoch": 0.02, "learning_rate": 6.938137165925475e-06, "loss": 0.8843, "step": 18362 }, { "epoch": 0.02, "learning_rate": 6.937111071369898e-06, "loss": 0.8767, "step": 18363 }, { "epoch": 0.02, "learning_rate": 6.936085012399818e-06, "loss": 0.8581, "step": 18364 }, { "epoch": 0.02, "learning_rate": 6.935058989027164e-06, "loss": 0.878, "step": 18365 }, { "epoch": 0.02, "learning_rate": 6.934033001263847e-06, "loss": 0.8448, "step": 18366 }, { "epoch": 0.02, "learning_rate": 6.933007049121794e-06, "loss": 0.8336, "step": 18367 }, { "epoch": 0.02, "learning_rate": 6.931981132612922e-06, "loss": 0.8734, "step": 18368 }, { "epoch": 0.02, "learning_rate": 6.930955251749152e-06, "loss": 0.8524, "step": 18369 }, { "epoch": 0.02, "learning_rate": 6.9299294065424e-06, "loss": 0.8898, "step": 18370 }, { "epoch": 0.02, "learning_rate": 6.9289035970045885e-06, "loss": 0.8551, "step": 18371 }, { "epoch": 0.02, "learning_rate": 6.927877823147633e-06, "loss": 0.8093, "step": 18372 }, { "epoch": 0.02, "learning_rate": 6.926852084983452e-06, "loss": 0.8335, "step": 18373 }, { "epoch": 0.02, "learning_rate": 6.925826382523961e-06, "loss": 0.8498, "step": 18374 }, { "epoch": 0.02, "learning_rate": 6.924800715781079e-06, "loss": 0.832, "step": 18375 }, { "epoch": 0.02, "learning_rate": 6.923775084766726e-06, "loss": 0.902, "step": 18376 }, { "epoch": 0.02, "learning_rate": 6.9227494894928115e-06, "loss": 0.8293, "step": 18377 }, { "epoch": 0.02, "learning_rate": 6.921723929971251e-06, "loss": 0.8755, "step": 18378 }, { "epoch": 0.02, "learning_rate": 6.920698406213969e-06, "loss": 0.8752, "step": 18379 }, { "epoch": 0.02, "learning_rate": 6.919672918232868e-06, "loss": 0.8893, "step": 18380 }, { "epoch": 0.02, "learning_rate": 6.918647466039869e-06, "loss": 0.8361, "step": 18381 }, { "epoch": 0.02, "learning_rate": 6.917622049646888e-06, "loss": 0.8539, "step": 18382 }, { "epoch": 0.02, "learning_rate": 6.916596669065836e-06, "loss": 0.8868, "step": 18383 }, { "epoch": 0.02, "learning_rate": 6.915571324308624e-06, "loss": 0.8576, "step": 18384 }, { "epoch": 0.02, "learning_rate": 6.914546015387167e-06, "loss": 0.8733, "step": 18385 }, { "epoch": 0.02, "learning_rate": 6.91352074231338e-06, "loss": 0.8354, "step": 18386 }, { "epoch": 0.02, "learning_rate": 6.9124955050991685e-06, "loss": 0.8818, "step": 18387 }, { "epoch": 0.02, "learning_rate": 6.911470303756449e-06, "loss": 0.8493, "step": 18388 }, { "epoch": 0.02, "learning_rate": 6.910445138297129e-06, "loss": 0.8538, "step": 18389 }, { "epoch": 0.02, "learning_rate": 6.909420008733126e-06, "loss": 0.8818, "step": 18390 }, { "epoch": 0.02, "learning_rate": 6.908394915076341e-06, "loss": 0.8843, "step": 18391 }, { "epoch": 0.02, "learning_rate": 6.907369857338688e-06, "loss": 0.8741, "step": 18392 }, { "epoch": 0.02, "learning_rate": 6.906344835532082e-06, "loss": 0.856, "step": 18393 }, { "epoch": 0.02, "learning_rate": 6.905319849668422e-06, "loss": 0.8579, "step": 18394 }, { "epoch": 0.02, "learning_rate": 6.904294899759621e-06, "loss": 0.8604, "step": 18395 }, { "epoch": 0.02, "learning_rate": 6.903269985817587e-06, "loss": 0.8575, "step": 18396 }, { "epoch": 0.02, "learning_rate": 6.902245107854233e-06, "loss": 0.9177, "step": 18397 }, { "epoch": 0.02, "learning_rate": 6.901220265881458e-06, "loss": 0.8916, "step": 18398 }, { "epoch": 0.02, "learning_rate": 6.90019545991117e-06, "loss": 0.8644, "step": 18399 }, { "epoch": 0.02, "learning_rate": 6.899170689955284e-06, "loss": 0.8221, "step": 18400 }, { "epoch": 0.02, "learning_rate": 6.898145956025697e-06, "loss": 0.8848, "step": 18401 }, { "epoch": 0.02, "learning_rate": 6.897121258134314e-06, "loss": 0.8929, "step": 18402 }, { "epoch": 0.02, "learning_rate": 6.896096596293047e-06, "loss": 0.8629, "step": 18403 }, { "epoch": 0.02, "learning_rate": 6.8950719705138e-06, "loss": 0.8626, "step": 18404 }, { "epoch": 0.02, "learning_rate": 6.894047380808471e-06, "loss": 0.8939, "step": 18405 }, { "epoch": 0.02, "learning_rate": 6.893022827188969e-06, "loss": 0.9126, "step": 18406 }, { "epoch": 0.02, "learning_rate": 6.8919983096672e-06, "loss": 0.9128, "step": 18407 }, { "epoch": 0.02, "learning_rate": 6.890973828255059e-06, "loss": 0.8781, "step": 18408 }, { "epoch": 0.02, "learning_rate": 6.889949382964453e-06, "loss": 0.8478, "step": 18409 }, { "epoch": 0.02, "learning_rate": 6.888924973807287e-06, "loss": 0.8712, "step": 18410 }, { "epoch": 0.02, "learning_rate": 6.887900600795461e-06, "loss": 0.8593, "step": 18411 }, { "epoch": 0.02, "learning_rate": 6.886876263940875e-06, "loss": 0.8928, "step": 18412 }, { "epoch": 0.02, "learning_rate": 6.88585196325543e-06, "loss": 0.8632, "step": 18413 }, { "epoch": 0.02, "learning_rate": 6.8848276987510286e-06, "loss": 0.8725, "step": 18414 }, { "epoch": 0.02, "learning_rate": 6.883803470439572e-06, "loss": 0.8635, "step": 18415 }, { "epoch": 0.02, "learning_rate": 6.882779278332955e-06, "loss": 0.8887, "step": 18416 }, { "epoch": 0.02, "learning_rate": 6.881755122443081e-06, "loss": 0.874, "step": 18417 }, { "epoch": 0.02, "learning_rate": 6.880731002781848e-06, "loss": 0.8809, "step": 18418 }, { "epoch": 0.02, "learning_rate": 6.879706919361154e-06, "loss": 0.8353, "step": 18419 }, { "epoch": 0.02, "learning_rate": 6.878682872192898e-06, "loss": 0.8873, "step": 18420 }, { "epoch": 0.02, "learning_rate": 6.877658861288974e-06, "loss": 0.8721, "step": 18421 }, { "epoch": 0.02, "learning_rate": 6.876634886661289e-06, "loss": 0.9083, "step": 18422 }, { "epoch": 0.02, "learning_rate": 6.875610948321728e-06, "loss": 0.885, "step": 18423 }, { "epoch": 0.02, "learning_rate": 6.874587046282192e-06, "loss": 0.895, "step": 18424 }, { "epoch": 0.02, "learning_rate": 6.873563180554583e-06, "loss": 0.8823, "step": 18425 }, { "epoch": 0.02, "learning_rate": 6.8725393511507845e-06, "loss": 0.8712, "step": 18426 }, { "epoch": 0.02, "learning_rate": 6.871515558082701e-06, "loss": 0.8689, "step": 18427 }, { "epoch": 0.02, "learning_rate": 6.870491801362224e-06, "loss": 0.8603, "step": 18428 }, { "epoch": 0.02, "learning_rate": 6.869468081001252e-06, "loss": 0.8354, "step": 18429 }, { "epoch": 0.02, "learning_rate": 6.868444397011671e-06, "loss": 0.8786, "step": 18430 }, { "epoch": 0.02, "learning_rate": 6.867420749405379e-06, "loss": 0.8991, "step": 18431 }, { "epoch": 0.02, "learning_rate": 6.866397138194271e-06, "loss": 0.8863, "step": 18432 }, { "epoch": 0.02, "learning_rate": 6.865373563390235e-06, "loss": 0.8483, "step": 18433 }, { "epoch": 0.02, "learning_rate": 6.864350025005165e-06, "loss": 0.8542, "step": 18434 }, { "epoch": 0.02, "learning_rate": 6.863326523050954e-06, "loss": 0.862, "step": 18435 }, { "epoch": 0.02, "learning_rate": 6.862303057539495e-06, "loss": 0.873, "step": 18436 }, { "epoch": 0.02, "learning_rate": 6.861279628482673e-06, "loss": 0.8529, "step": 18437 }, { "epoch": 0.02, "learning_rate": 6.860256235892382e-06, "loss": 0.8578, "step": 18438 }, { "epoch": 0.02, "learning_rate": 6.859232879780515e-06, "loss": 0.8563, "step": 18439 }, { "epoch": 0.02, "learning_rate": 6.858209560158955e-06, "loss": 0.8942, "step": 18440 }, { "epoch": 0.02, "learning_rate": 6.857186277039596e-06, "loss": 0.8488, "step": 18441 }, { "epoch": 0.02, "learning_rate": 6.856163030434323e-06, "loss": 0.9002, "step": 18442 }, { "epoch": 0.02, "learning_rate": 6.855139820355034e-06, "loss": 0.8535, "step": 18443 }, { "epoch": 0.02, "learning_rate": 6.8541166468136046e-06, "loss": 0.8776, "step": 18444 }, { "epoch": 0.02, "learning_rate": 6.853093509821925e-06, "loss": 0.9263, "step": 18445 }, { "epoch": 0.02, "learning_rate": 6.852070409391893e-06, "loss": 0.8706, "step": 18446 }, { "epoch": 0.02, "learning_rate": 6.851047345535382e-06, "loss": 0.8916, "step": 18447 }, { "epoch": 0.02, "learning_rate": 6.850024318264281e-06, "loss": 0.8984, "step": 18448 }, { "epoch": 0.02, "learning_rate": 6.849001327590481e-06, "loss": 0.8865, "step": 18449 }, { "epoch": 0.02, "learning_rate": 6.847978373525867e-06, "loss": 0.8702, "step": 18450 }, { "epoch": 0.02, "learning_rate": 6.846955456082318e-06, "loss": 0.8421, "step": 18451 }, { "epoch": 0.02, "learning_rate": 6.845932575271722e-06, "loss": 0.8585, "step": 18452 }, { "epoch": 0.02, "learning_rate": 6.8449097311059665e-06, "loss": 0.8543, "step": 18453 }, { "epoch": 0.02, "learning_rate": 6.8438869235969294e-06, "loss": 0.8423, "step": 18454 }, { "epoch": 0.02, "learning_rate": 6.842864152756497e-06, "loss": 0.9172, "step": 18455 }, { "epoch": 0.02, "learning_rate": 6.841841418596551e-06, "loss": 0.8678, "step": 18456 }, { "epoch": 0.02, "learning_rate": 6.840818721128977e-06, "loss": 0.9008, "step": 18457 }, { "epoch": 0.02, "learning_rate": 6.839796060365651e-06, "loss": 0.8932, "step": 18458 }, { "epoch": 0.02, "learning_rate": 6.83877343631846e-06, "loss": 0.8714, "step": 18459 }, { "epoch": 0.02, "learning_rate": 6.837750848999282e-06, "loss": 0.9271, "step": 18460 }, { "epoch": 0.02, "learning_rate": 6.836728298420001e-06, "loss": 0.8415, "step": 18461 }, { "epoch": 0.02, "learning_rate": 6.835705784592492e-06, "loss": 0.8716, "step": 18462 }, { "epoch": 0.02, "learning_rate": 6.83468330752864e-06, "loss": 0.8292, "step": 18463 }, { "epoch": 0.02, "learning_rate": 6.833660867240323e-06, "loss": 0.8565, "step": 18464 }, { "epoch": 0.02, "learning_rate": 6.832638463739418e-06, "loss": 0.8246, "step": 18465 }, { "epoch": 0.02, "learning_rate": 6.831616097037804e-06, "loss": 0.882, "step": 18466 }, { "epoch": 0.02, "learning_rate": 6.8305937671473586e-06, "loss": 0.8766, "step": 18467 }, { "epoch": 0.02, "learning_rate": 6.829571474079968e-06, "loss": 0.8587, "step": 18468 }, { "epoch": 0.02, "learning_rate": 6.828549217847495e-06, "loss": 0.8241, "step": 18469 }, { "epoch": 0.02, "learning_rate": 6.827526998461826e-06, "loss": 0.9023, "step": 18470 }, { "epoch": 0.02, "learning_rate": 6.82650481593484e-06, "loss": 0.8811, "step": 18471 }, { "epoch": 0.02, "learning_rate": 6.825482670278403e-06, "loss": 0.8665, "step": 18472 }, { "epoch": 0.02, "learning_rate": 6.8244605615043954e-06, "loss": 0.8547, "step": 18473 }, { "epoch": 0.02, "learning_rate": 6.823438489624695e-06, "loss": 0.8905, "step": 18474 }, { "epoch": 0.02, "learning_rate": 6.822416454651176e-06, "loss": 0.8605, "step": 18475 }, { "epoch": 0.02, "learning_rate": 6.821394456595709e-06, "loss": 0.8767, "step": 18476 }, { "epoch": 0.02, "learning_rate": 6.82037249547017e-06, "loss": 0.8982, "step": 18477 }, { "epoch": 0.02, "learning_rate": 6.819350571286434e-06, "loss": 0.8407, "step": 18478 }, { "epoch": 0.02, "learning_rate": 6.818328684056371e-06, "loss": 0.8657, "step": 18479 }, { "epoch": 0.02, "learning_rate": 6.817306833791854e-06, "loss": 0.8359, "step": 18480 }, { "epoch": 0.02, "learning_rate": 6.8162850205047585e-06, "loss": 0.885, "step": 18481 }, { "epoch": 0.02, "learning_rate": 6.815263244206953e-06, "loss": 0.9187, "step": 18482 }, { "epoch": 0.02, "learning_rate": 6.8142415049103085e-06, "loss": 0.8742, "step": 18483 }, { "epoch": 0.02, "learning_rate": 6.813219802626698e-06, "loss": 0.8484, "step": 18484 }, { "epoch": 0.02, "learning_rate": 6.812198137367991e-06, "loss": 0.8888, "step": 18485 }, { "epoch": 0.02, "learning_rate": 6.811176509146057e-06, "loss": 0.8842, "step": 18486 }, { "epoch": 0.02, "learning_rate": 6.810154917972764e-06, "loss": 0.8814, "step": 18487 }, { "epoch": 0.02, "learning_rate": 6.809133363859983e-06, "loss": 0.8382, "step": 18488 }, { "epoch": 0.02, "learning_rate": 6.808111846819588e-06, "loss": 0.8504, "step": 18489 }, { "epoch": 0.02, "learning_rate": 6.807090366863438e-06, "loss": 0.834, "step": 18490 }, { "epoch": 0.02, "learning_rate": 6.806068924003404e-06, "loss": 0.866, "step": 18491 }, { "epoch": 0.02, "learning_rate": 6.805047518251359e-06, "loss": 0.8687, "step": 18492 }, { "epoch": 0.02, "learning_rate": 6.804026149619161e-06, "loss": 0.8681, "step": 18493 }, { "epoch": 0.02, "learning_rate": 6.803004818118679e-06, "loss": 0.8321, "step": 18494 }, { "epoch": 0.02, "learning_rate": 6.801983523761784e-06, "loss": 0.8644, "step": 18495 }, { "epoch": 0.02, "learning_rate": 6.800962266560342e-06, "loss": 0.9043, "step": 18496 }, { "epoch": 0.02, "learning_rate": 6.799941046526208e-06, "loss": 0.868, "step": 18497 }, { "epoch": 0.02, "learning_rate": 6.798919863671256e-06, "loss": 0.8682, "step": 18498 }, { "epoch": 0.02, "learning_rate": 6.79789871800735e-06, "loss": 0.8928, "step": 18499 }, { "epoch": 0.02, "learning_rate": 6.79687760954635e-06, "loss": 0.8703, "step": 18500 }, { "epoch": 0.02, "learning_rate": 6.7958565383001225e-06, "loss": 0.8695, "step": 18501 }, { "epoch": 0.03, "learning_rate": 6.794835504280528e-06, "loss": 0.847, "step": 18502 }, { "epoch": 0.03, "learning_rate": 6.793814507499435e-06, "loss": 0.8159, "step": 18503 }, { "epoch": 0.03, "learning_rate": 6.7927935479686965e-06, "loss": 0.8964, "step": 18504 }, { "epoch": 0.03, "learning_rate": 6.791772625700181e-06, "loss": 0.8876, "step": 18505 }, { "epoch": 0.03, "learning_rate": 6.79075174070575e-06, "loss": 0.8859, "step": 18506 }, { "epoch": 0.03, "learning_rate": 6.789730892997261e-06, "loss": 0.8471, "step": 18507 }, { "epoch": 0.03, "learning_rate": 6.788710082586577e-06, "loss": 0.8607, "step": 18508 }, { "epoch": 0.03, "learning_rate": 6.787689309485555e-06, "loss": 0.8711, "step": 18509 }, { "epoch": 0.03, "learning_rate": 6.786668573706061e-06, "loss": 0.8818, "step": 18510 }, { "epoch": 0.03, "learning_rate": 6.785647875259947e-06, "loss": 0.827, "step": 18511 }, { "epoch": 0.03, "learning_rate": 6.784627214159074e-06, "loss": 0.9016, "step": 18512 }, { "epoch": 0.03, "learning_rate": 6.7836065904152995e-06, "loss": 0.8735, "step": 18513 }, { "epoch": 0.03, "learning_rate": 6.782586004040489e-06, "loss": 0.8941, "step": 18514 }, { "epoch": 0.03, "learning_rate": 6.781565455046487e-06, "loss": 0.9015, "step": 18515 }, { "epoch": 0.03, "learning_rate": 6.78054494344516e-06, "loss": 0.8763, "step": 18516 }, { "epoch": 0.03, "learning_rate": 6.779524469248367e-06, "loss": 0.8405, "step": 18517 }, { "epoch": 0.03, "learning_rate": 6.778504032467952e-06, "loss": 0.8826, "step": 18518 }, { "epoch": 0.03, "learning_rate": 6.77748363311578e-06, "loss": 0.8376, "step": 18519 }, { "epoch": 0.03, "learning_rate": 6.776463271203705e-06, "loss": 0.8829, "step": 18520 }, { "epoch": 0.03, "learning_rate": 6.775442946743582e-06, "loss": 0.8626, "step": 18521 }, { "epoch": 0.03, "learning_rate": 6.774422659747264e-06, "loss": 0.8725, "step": 18522 }, { "epoch": 0.03, "learning_rate": 6.773402410226604e-06, "loss": 0.856, "step": 18523 }, { "epoch": 0.03, "learning_rate": 6.772382198193459e-06, "loss": 0.831, "step": 18524 }, { "epoch": 0.03, "learning_rate": 6.771362023659678e-06, "loss": 0.8718, "step": 18525 }, { "epoch": 0.03, "learning_rate": 6.770341886637118e-06, "loss": 0.8659, "step": 18526 }, { "epoch": 0.03, "learning_rate": 6.769321787137626e-06, "loss": 0.8534, "step": 18527 }, { "epoch": 0.03, "learning_rate": 6.7683017251730584e-06, "loss": 0.8853, "step": 18528 }, { "epoch": 0.03, "learning_rate": 6.767281700755266e-06, "loss": 0.901, "step": 18529 }, { "epoch": 0.03, "learning_rate": 6.766261713896096e-06, "loss": 0.8457, "step": 18530 }, { "epoch": 0.03, "learning_rate": 6.765241764607405e-06, "loss": 0.8351, "step": 18531 }, { "epoch": 0.03, "learning_rate": 6.764221852901038e-06, "loss": 0.8415, "step": 18532 }, { "epoch": 0.03, "learning_rate": 6.763201978788844e-06, "loss": 0.8677, "step": 18533 }, { "epoch": 0.03, "learning_rate": 6.762182142282675e-06, "loss": 0.8511, "step": 18534 }, { "epoch": 0.03, "learning_rate": 6.761162343394384e-06, "loss": 0.8587, "step": 18535 }, { "epoch": 0.03, "learning_rate": 6.760142582135811e-06, "loss": 0.8623, "step": 18536 }, { "epoch": 0.03, "learning_rate": 6.759122858518805e-06, "loss": 0.8864, "step": 18537 }, { "epoch": 0.03, "learning_rate": 6.758103172555221e-06, "loss": 0.8557, "step": 18538 }, { "epoch": 0.03, "learning_rate": 6.7570835242568975e-06, "loss": 0.852, "step": 18539 }, { "epoch": 0.03, "learning_rate": 6.756063913635682e-06, "loss": 0.8764, "step": 18540 }, { "epoch": 0.03, "learning_rate": 6.755044340703425e-06, "loss": 0.917, "step": 18541 }, { "epoch": 0.03, "learning_rate": 6.754024805471973e-06, "loss": 0.8569, "step": 18542 }, { "epoch": 0.03, "learning_rate": 6.7530053079531664e-06, "loss": 0.8638, "step": 18543 }, { "epoch": 0.03, "learning_rate": 6.751985848158851e-06, "loss": 0.846, "step": 18544 }, { "epoch": 0.03, "learning_rate": 6.750966426100874e-06, "loss": 0.8782, "step": 18545 }, { "epoch": 0.03, "learning_rate": 6.7499470417910765e-06, "loss": 0.8771, "step": 18546 }, { "epoch": 0.03, "learning_rate": 6.748927695241303e-06, "loss": 0.8607, "step": 18547 }, { "epoch": 0.03, "learning_rate": 6.7479083864633956e-06, "loss": 0.867, "step": 18548 }, { "epoch": 0.03, "learning_rate": 6.7468891154692005e-06, "loss": 0.8762, "step": 18549 }, { "epoch": 0.03, "learning_rate": 6.745869882270555e-06, "loss": 0.8627, "step": 18550 }, { "epoch": 0.03, "learning_rate": 6.744850686879303e-06, "loss": 0.8602, "step": 18551 }, { "epoch": 0.03, "learning_rate": 6.743831529307287e-06, "loss": 0.8395, "step": 18552 }, { "epoch": 0.03, "learning_rate": 6.742812409566347e-06, "loss": 0.8194, "step": 18553 }, { "epoch": 0.03, "learning_rate": 6.741793327668322e-06, "loss": 0.8994, "step": 18554 }, { "epoch": 0.03, "learning_rate": 6.7407742836250535e-06, "loss": 0.8607, "step": 18555 }, { "epoch": 0.03, "learning_rate": 6.739755277448381e-06, "loss": 0.8558, "step": 18556 }, { "epoch": 0.03, "learning_rate": 6.738736309150142e-06, "loss": 0.8485, "step": 18557 }, { "epoch": 0.03, "learning_rate": 6.737717378742175e-06, "loss": 0.849, "step": 18558 }, { "epoch": 0.03, "learning_rate": 6.736698486236321e-06, "loss": 0.847, "step": 18559 }, { "epoch": 0.03, "learning_rate": 6.735679631644423e-06, "loss": 0.8772, "step": 18560 }, { "epoch": 0.03, "learning_rate": 6.734660814978302e-06, "loss": 0.8169, "step": 18561 }, { "epoch": 0.03, "learning_rate": 6.733642036249809e-06, "loss": 0.8324, "step": 18562 }, { "epoch": 0.03, "learning_rate": 6.732623295470779e-06, "loss": 0.8504, "step": 18563 }, { "epoch": 0.03, "learning_rate": 6.731604592653039e-06, "loss": 0.8373, "step": 18564 }, { "epoch": 0.03, "learning_rate": 6.730585927808433e-06, "loss": 0.8761, "step": 18565 }, { "epoch": 0.03, "learning_rate": 6.729567300948794e-06, "loss": 0.861, "step": 18566 }, { "epoch": 0.03, "learning_rate": 6.728548712085959e-06, "loss": 0.8748, "step": 18567 }, { "epoch": 0.03, "learning_rate": 6.727530161231757e-06, "loss": 0.8612, "step": 18568 }, { "epoch": 0.03, "learning_rate": 6.726511648398026e-06, "loss": 0.8718, "step": 18569 }, { "epoch": 0.03, "learning_rate": 6.725493173596597e-06, "loss": 0.876, "step": 18570 }, { "epoch": 0.03, "learning_rate": 6.724474736839305e-06, "loss": 0.8803, "step": 18571 }, { "epoch": 0.03, "learning_rate": 6.723456338137979e-06, "loss": 0.8563, "step": 18572 }, { "epoch": 0.03, "learning_rate": 6.722437977504455e-06, "loss": 0.8698, "step": 18573 }, { "epoch": 0.03, "learning_rate": 6.721419654950565e-06, "loss": 0.9088, "step": 18574 }, { "epoch": 0.03, "learning_rate": 6.720401370488135e-06, "loss": 0.8859, "step": 18575 }, { "epoch": 0.03, "learning_rate": 6.7193831241289995e-06, "loss": 0.8664, "step": 18576 }, { "epoch": 0.03, "learning_rate": 6.718364915884989e-06, "loss": 0.8318, "step": 18577 }, { "epoch": 0.03, "learning_rate": 6.717346745767932e-06, "loss": 0.8947, "step": 18578 }, { "epoch": 0.03, "learning_rate": 6.716328613789657e-06, "loss": 0.8697, "step": 18579 }, { "epoch": 0.03, "learning_rate": 6.7153105199619936e-06, "loss": 0.8758, "step": 18580 }, { "epoch": 0.03, "learning_rate": 6.714292464296777e-06, "loss": 0.8944, "step": 18581 }, { "epoch": 0.03, "learning_rate": 6.713274446805824e-06, "loss": 0.8721, "step": 18582 }, { "epoch": 0.03, "learning_rate": 6.7122564675009665e-06, "loss": 0.8685, "step": 18583 }, { "epoch": 0.03, "learning_rate": 6.711238526394039e-06, "loss": 0.871, "step": 18584 }, { "epoch": 0.03, "learning_rate": 6.710220623496857e-06, "loss": 0.8255, "step": 18585 }, { "epoch": 0.03, "learning_rate": 6.709202758821249e-06, "loss": 0.8877, "step": 18586 }, { "epoch": 0.03, "learning_rate": 6.708184932379047e-06, "loss": 0.8906, "step": 18587 }, { "epoch": 0.03, "learning_rate": 6.707167144182074e-06, "loss": 0.8561, "step": 18588 }, { "epoch": 0.03, "learning_rate": 6.706149394242152e-06, "loss": 0.9031, "step": 18589 }, { "epoch": 0.03, "learning_rate": 6.705131682571107e-06, "loss": 0.8803, "step": 18590 }, { "epoch": 0.03, "learning_rate": 6.704114009180765e-06, "loss": 0.8764, "step": 18591 }, { "epoch": 0.03, "learning_rate": 6.7030963740829466e-06, "loss": 0.8211, "step": 18592 }, { "epoch": 0.03, "learning_rate": 6.702078777289475e-06, "loss": 0.8807, "step": 18593 }, { "epoch": 0.03, "learning_rate": 6.701061218812177e-06, "loss": 0.8692, "step": 18594 }, { "epoch": 0.03, "learning_rate": 6.700043698662873e-06, "loss": 0.8625, "step": 18595 }, { "epoch": 0.03, "learning_rate": 6.699026216853381e-06, "loss": 0.859, "step": 18596 }, { "epoch": 0.03, "learning_rate": 6.698008773395526e-06, "loss": 0.8643, "step": 18597 }, { "epoch": 0.03, "learning_rate": 6.6969913683011274e-06, "loss": 0.8403, "step": 18598 }, { "epoch": 0.03, "learning_rate": 6.695974001582011e-06, "loss": 0.8845, "step": 18599 }, { "epoch": 0.03, "learning_rate": 6.694956673249988e-06, "loss": 0.8829, "step": 18600 }, { "epoch": 0.03, "learning_rate": 6.693939383316883e-06, "loss": 0.8762, "step": 18601 }, { "epoch": 0.03, "learning_rate": 6.692922131794517e-06, "loss": 0.8994, "step": 18602 }, { "epoch": 0.03, "learning_rate": 6.691904918694705e-06, "loss": 0.8557, "step": 18603 }, { "epoch": 0.03, "learning_rate": 6.690887744029264e-06, "loss": 0.8946, "step": 18604 }, { "epoch": 0.03, "learning_rate": 6.689870607810016e-06, "loss": 0.8602, "step": 18605 }, { "epoch": 0.03, "learning_rate": 6.688853510048782e-06, "loss": 0.8759, "step": 18606 }, { "epoch": 0.03, "learning_rate": 6.687836450757366e-06, "loss": 0.905, "step": 18607 }, { "epoch": 0.03, "learning_rate": 6.686819429947595e-06, "loss": 0.8852, "step": 18608 }, { "epoch": 0.03, "learning_rate": 6.6858024476312866e-06, "loss": 0.8334, "step": 18609 }, { "epoch": 0.03, "learning_rate": 6.684785503820246e-06, "loss": 0.8871, "step": 18610 }, { "epoch": 0.03, "learning_rate": 6.683768598526296e-06, "loss": 0.8914, "step": 18611 }, { "epoch": 0.03, "learning_rate": 6.68275173176125e-06, "loss": 0.8365, "step": 18612 }, { "epoch": 0.03, "learning_rate": 6.681734903536923e-06, "loss": 0.8553, "step": 18613 }, { "epoch": 0.03, "learning_rate": 6.680718113865127e-06, "loss": 0.874, "step": 18614 }, { "epoch": 0.03, "learning_rate": 6.6797013627576755e-06, "loss": 0.8423, "step": 18615 }, { "epoch": 0.03, "learning_rate": 6.678684650226383e-06, "loss": 0.8657, "step": 18616 }, { "epoch": 0.03, "learning_rate": 6.67766797628306e-06, "loss": 0.8678, "step": 18617 }, { "epoch": 0.03, "learning_rate": 6.67665134093952e-06, "loss": 0.8896, "step": 18618 }, { "epoch": 0.03, "learning_rate": 6.675634744207572e-06, "loss": 0.8744, "step": 18619 }, { "epoch": 0.03, "learning_rate": 6.674618186099032e-06, "loss": 0.8545, "step": 18620 }, { "epoch": 0.03, "learning_rate": 6.6736016666257065e-06, "loss": 0.8964, "step": 18621 }, { "epoch": 0.03, "learning_rate": 6.672585185799405e-06, "loss": 0.8609, "step": 18622 }, { "epoch": 0.03, "learning_rate": 6.671568743631943e-06, "loss": 0.8915, "step": 18623 }, { "epoch": 0.03, "learning_rate": 6.670552340135123e-06, "loss": 0.8645, "step": 18624 }, { "epoch": 0.03, "learning_rate": 6.669535975320758e-06, "loss": 0.8762, "step": 18625 }, { "epoch": 0.03, "learning_rate": 6.668519649200651e-06, "loss": 0.8892, "step": 18626 }, { "epoch": 0.03, "learning_rate": 6.6675033617866225e-06, "loss": 0.8532, "step": 18627 }, { "epoch": 0.03, "learning_rate": 6.666487113090466e-06, "loss": 0.8419, "step": 18628 }, { "epoch": 0.03, "learning_rate": 6.665470903123994e-06, "loss": 0.8592, "step": 18629 }, { "epoch": 0.03, "learning_rate": 6.664454731899017e-06, "loss": 0.9093, "step": 18630 }, { "epoch": 0.03, "learning_rate": 6.663438599427336e-06, "loss": 0.8747, "step": 18631 }, { "epoch": 0.03, "learning_rate": 6.6624225057207545e-06, "loss": 0.8688, "step": 18632 }, { "epoch": 0.03, "learning_rate": 6.6614064507910845e-06, "loss": 0.861, "step": 18633 }, { "epoch": 0.03, "learning_rate": 6.6603904346501304e-06, "loss": 0.9168, "step": 18634 }, { "epoch": 0.03, "learning_rate": 6.659374457309691e-06, "loss": 0.8941, "step": 18635 }, { "epoch": 0.03, "learning_rate": 6.658358518781573e-06, "loss": 0.8674, "step": 18636 }, { "epoch": 0.03, "learning_rate": 6.657342619077583e-06, "loss": 0.8822, "step": 18637 }, { "epoch": 0.03, "learning_rate": 6.656326758209517e-06, "loss": 0.881, "step": 18638 }, { "epoch": 0.03, "learning_rate": 6.6553109361891834e-06, "loss": 0.8885, "step": 18639 }, { "epoch": 0.03, "learning_rate": 6.654295153028383e-06, "loss": 0.8727, "step": 18640 }, { "epoch": 0.03, "learning_rate": 6.653279408738918e-06, "loss": 0.8673, "step": 18641 }, { "epoch": 0.03, "learning_rate": 6.652263703332587e-06, "loss": 0.8637, "step": 18642 }, { "epoch": 0.03, "learning_rate": 6.651248036821192e-06, "loss": 0.8789, "step": 18643 }, { "epoch": 0.03, "learning_rate": 6.6502324092165326e-06, "loss": 0.8892, "step": 18644 }, { "epoch": 0.03, "learning_rate": 6.649216820530413e-06, "loss": 0.8724, "step": 18645 }, { "epoch": 0.03, "learning_rate": 6.648201270774628e-06, "loss": 0.8631, "step": 18646 }, { "epoch": 0.03, "learning_rate": 6.647185759960977e-06, "loss": 0.8493, "step": 18647 }, { "epoch": 0.03, "learning_rate": 6.64617028810126e-06, "loss": 0.8774, "step": 18648 }, { "epoch": 0.03, "learning_rate": 6.645154855207275e-06, "loss": 0.8808, "step": 18649 }, { "epoch": 0.03, "learning_rate": 6.644139461290814e-06, "loss": 0.8721, "step": 18650 }, { "epoch": 0.03, "learning_rate": 6.643124106363683e-06, "loss": 0.8397, "step": 18651 }, { "epoch": 0.03, "learning_rate": 6.642108790437678e-06, "loss": 0.8308, "step": 18652 }, { "epoch": 0.03, "learning_rate": 6.641093513524586e-06, "loss": 0.867, "step": 18653 }, { "epoch": 0.03, "learning_rate": 6.64007827563621e-06, "loss": 0.892, "step": 18654 }, { "epoch": 0.03, "learning_rate": 6.639063076784349e-06, "loss": 0.8592, "step": 18655 }, { "epoch": 0.03, "learning_rate": 6.638047916980787e-06, "loss": 0.8792, "step": 18656 }, { "epoch": 0.03, "learning_rate": 6.637032796237326e-06, "loss": 0.87, "step": 18657 }, { "epoch": 0.03, "learning_rate": 6.636017714565758e-06, "loss": 0.9239, "step": 18658 }, { "epoch": 0.03, "learning_rate": 6.635002671977879e-06, "loss": 0.8733, "step": 18659 }, { "epoch": 0.03, "learning_rate": 6.633987668485478e-06, "loss": 0.9203, "step": 18660 }, { "epoch": 0.03, "learning_rate": 6.632972704100349e-06, "loss": 0.8755, "step": 18661 }, { "epoch": 0.03, "learning_rate": 6.631957778834287e-06, "loss": 0.8977, "step": 18662 }, { "epoch": 0.03, "learning_rate": 6.63094289269908e-06, "loss": 0.889, "step": 18663 }, { "epoch": 0.03, "learning_rate": 6.6299280457065185e-06, "loss": 0.8868, "step": 18664 }, { "epoch": 0.03, "learning_rate": 6.628913237868397e-06, "loss": 0.8231, "step": 18665 }, { "epoch": 0.03, "learning_rate": 6.627898469196506e-06, "loss": 0.8476, "step": 18666 }, { "epoch": 0.03, "learning_rate": 6.6268837397026295e-06, "loss": 0.8828, "step": 18667 }, { "epoch": 0.03, "learning_rate": 6.625869049398562e-06, "loss": 0.8654, "step": 18668 }, { "epoch": 0.03, "learning_rate": 6.624854398296094e-06, "loss": 0.8278, "step": 18669 }, { "epoch": 0.03, "learning_rate": 6.623839786407008e-06, "loss": 0.8132, "step": 18670 }, { "epoch": 0.03, "learning_rate": 6.622825213743096e-06, "loss": 0.8912, "step": 18671 }, { "epoch": 0.03, "learning_rate": 6.621810680316143e-06, "loss": 0.8554, "step": 18672 }, { "epoch": 0.03, "learning_rate": 6.620796186137943e-06, "loss": 0.8727, "step": 18673 }, { "epoch": 0.03, "learning_rate": 6.6197817312202765e-06, "loss": 0.8611, "step": 18674 }, { "epoch": 0.03, "learning_rate": 6.618767315574926e-06, "loss": 0.8825, "step": 18675 }, { "epoch": 0.03, "learning_rate": 6.61775293921369e-06, "loss": 0.8827, "step": 18676 }, { "epoch": 0.03, "learning_rate": 6.616738602148339e-06, "loss": 0.8503, "step": 18677 }, { "epoch": 0.03, "learning_rate": 6.615724304390667e-06, "loss": 0.8765, "step": 18678 }, { "epoch": 0.03, "learning_rate": 6.6147100459524575e-06, "loss": 0.8917, "step": 18679 }, { "epoch": 0.03, "learning_rate": 6.6136958268454945e-06, "loss": 0.8748, "step": 18680 }, { "epoch": 0.03, "learning_rate": 6.612681647081558e-06, "loss": 0.886, "step": 18681 }, { "epoch": 0.03, "learning_rate": 6.611667506672433e-06, "loss": 0.8636, "step": 18682 }, { "epoch": 0.03, "learning_rate": 6.610653405629904e-06, "loss": 0.8339, "step": 18683 }, { "epoch": 0.03, "learning_rate": 6.60963934396575e-06, "loss": 0.8446, "step": 18684 }, { "epoch": 0.03, "learning_rate": 6.608625321691755e-06, "loss": 0.8312, "step": 18685 }, { "epoch": 0.03, "learning_rate": 6.607611338819697e-06, "loss": 0.857, "step": 18686 }, { "epoch": 0.03, "learning_rate": 6.6065973953613624e-06, "loss": 0.8399, "step": 18687 }, { "epoch": 0.03, "learning_rate": 6.605583491328526e-06, "loss": 0.8568, "step": 18688 }, { "epoch": 0.03, "learning_rate": 6.604569626732971e-06, "loss": 0.8564, "step": 18689 }, { "epoch": 0.03, "learning_rate": 6.603555801586473e-06, "loss": 0.8573, "step": 18690 }, { "epoch": 0.03, "learning_rate": 6.6025420159008165e-06, "loss": 0.7985, "step": 18691 }, { "epoch": 0.03, "learning_rate": 6.601528269687775e-06, "loss": 0.8274, "step": 18692 }, { "epoch": 0.03, "learning_rate": 6.600514562959126e-06, "loss": 0.8671, "step": 18693 }, { "epoch": 0.03, "learning_rate": 6.599500895726655e-06, "loss": 0.8397, "step": 18694 }, { "epoch": 0.03, "learning_rate": 6.59848726800213e-06, "loss": 0.8643, "step": 18695 }, { "epoch": 0.03, "learning_rate": 6.597473679797328e-06, "loss": 0.8562, "step": 18696 }, { "epoch": 0.03, "learning_rate": 6.596460131124031e-06, "loss": 0.8685, "step": 18697 }, { "epoch": 0.03, "learning_rate": 6.595446621994015e-06, "loss": 0.8791, "step": 18698 }, { "epoch": 0.03, "learning_rate": 6.594433152419047e-06, "loss": 0.8711, "step": 18699 }, { "epoch": 0.03, "learning_rate": 6.593419722410907e-06, "loss": 0.8812, "step": 18700 }, { "epoch": 0.03, "learning_rate": 6.592406331981376e-06, "loss": 0.8766, "step": 18701 }, { "epoch": 0.03, "learning_rate": 6.5913929811422126e-06, "loss": 0.8899, "step": 18702 }, { "epoch": 0.03, "learning_rate": 6.590379669905202e-06, "loss": 0.8569, "step": 18703 }, { "epoch": 0.03, "learning_rate": 6.589366398282113e-06, "loss": 0.8806, "step": 18704 }, { "epoch": 0.03, "learning_rate": 6.5883531662847225e-06, "loss": 0.8386, "step": 18705 }, { "epoch": 0.03, "learning_rate": 6.587339973924795e-06, "loss": 0.8833, "step": 18706 }, { "epoch": 0.03, "learning_rate": 6.586326821214106e-06, "loss": 0.8504, "step": 18707 }, { "epoch": 0.03, "learning_rate": 6.585313708164429e-06, "loss": 0.845, "step": 18708 }, { "epoch": 0.03, "learning_rate": 6.584300634787531e-06, "loss": 0.8996, "step": 18709 }, { "epoch": 0.03, "learning_rate": 6.583287601095181e-06, "loss": 0.8563, "step": 18710 }, { "epoch": 0.03, "learning_rate": 6.582274607099152e-06, "loss": 0.9121, "step": 18711 }, { "epoch": 0.03, "learning_rate": 6.581261652811215e-06, "loss": 0.9125, "step": 18712 }, { "epoch": 0.03, "learning_rate": 6.580248738243134e-06, "loss": 0.8817, "step": 18713 }, { "epoch": 0.03, "learning_rate": 6.579235863406678e-06, "loss": 0.8585, "step": 18714 }, { "epoch": 0.03, "learning_rate": 6.578223028313618e-06, "loss": 0.8437, "step": 18715 }, { "epoch": 0.03, "learning_rate": 6.5772102329757194e-06, "loss": 0.8529, "step": 18716 }, { "epoch": 0.03, "learning_rate": 6.576197477404748e-06, "loss": 0.8685, "step": 18717 }, { "epoch": 0.03, "learning_rate": 6.5751847616124695e-06, "loss": 0.8926, "step": 18718 }, { "epoch": 0.03, "learning_rate": 6.574172085610659e-06, "loss": 0.8409, "step": 18719 }, { "epoch": 0.03, "learning_rate": 6.573159449411071e-06, "loss": 0.846, "step": 18720 }, { "epoch": 0.03, "learning_rate": 6.572146853025472e-06, "loss": 0.9162, "step": 18721 }, { "epoch": 0.03, "learning_rate": 6.5711342964656356e-06, "loss": 0.8852, "step": 18722 }, { "epoch": 0.03, "learning_rate": 6.570121779743312e-06, "loss": 0.8376, "step": 18723 }, { "epoch": 0.03, "learning_rate": 6.569109302870277e-06, "loss": 0.9168, "step": 18724 }, { "epoch": 0.03, "learning_rate": 6.568096865858289e-06, "loss": 0.8533, "step": 18725 }, { "epoch": 0.03, "learning_rate": 6.567084468719112e-06, "loss": 0.8739, "step": 18726 }, { "epoch": 0.03, "learning_rate": 6.566072111464507e-06, "loss": 0.8919, "step": 18727 }, { "epoch": 0.03, "learning_rate": 6.5650597941062345e-06, "loss": 0.878, "step": 18728 }, { "epoch": 0.03, "learning_rate": 6.564047516656061e-06, "loss": 0.8657, "step": 18729 }, { "epoch": 0.03, "learning_rate": 6.563035279125742e-06, "loss": 0.8563, "step": 18730 }, { "epoch": 0.03, "learning_rate": 6.56202308152704e-06, "loss": 0.8946, "step": 18731 }, { "epoch": 0.03, "learning_rate": 6.561010923871716e-06, "loss": 0.8545, "step": 18732 }, { "epoch": 0.03, "learning_rate": 6.559998806171529e-06, "loss": 0.8343, "step": 18733 }, { "epoch": 0.03, "learning_rate": 6.5589867284382366e-06, "loss": 0.8826, "step": 18734 }, { "epoch": 0.03, "learning_rate": 6.557974690683598e-06, "loss": 0.8699, "step": 18735 }, { "epoch": 0.03, "learning_rate": 6.556962692919373e-06, "loss": 0.8247, "step": 18736 }, { "epoch": 0.03, "learning_rate": 6.5559507351573195e-06, "loss": 0.8457, "step": 18737 }, { "epoch": 0.03, "learning_rate": 6.554938817409191e-06, "loss": 0.9049, "step": 18738 }, { "epoch": 0.03, "learning_rate": 6.5539269396867434e-06, "loss": 0.8838, "step": 18739 }, { "epoch": 0.03, "learning_rate": 6.552915102001744e-06, "loss": 0.8845, "step": 18740 }, { "epoch": 0.03, "learning_rate": 6.551903304365936e-06, "loss": 0.8598, "step": 18741 }, { "epoch": 0.03, "learning_rate": 6.550891546791079e-06, "loss": 0.8471, "step": 18742 }, { "epoch": 0.03, "learning_rate": 6.549879829288927e-06, "loss": 0.8502, "step": 18743 }, { "epoch": 0.03, "learning_rate": 6.548868151871244e-06, "loss": 0.8458, "step": 18744 }, { "epoch": 0.03, "learning_rate": 6.547856514549766e-06, "loss": 0.8405, "step": 18745 }, { "epoch": 0.03, "learning_rate": 6.54684491733626e-06, "loss": 0.8942, "step": 18746 }, { "epoch": 0.03, "learning_rate": 6.545833360242479e-06, "loss": 0.8591, "step": 18747 }, { "epoch": 0.03, "learning_rate": 6.544821843280167e-06, "loss": 0.8533, "step": 18748 }, { "epoch": 0.03, "learning_rate": 6.543810366461081e-06, "loss": 0.8472, "step": 18749 }, { "epoch": 0.03, "learning_rate": 6.542798929796974e-06, "loss": 0.8771, "step": 18750 }, { "epoch": 0.03, "learning_rate": 6.5417875332995974e-06, "loss": 0.875, "step": 18751 }, { "epoch": 0.03, "learning_rate": 6.5407761769806965e-06, "loss": 0.845, "step": 18752 }, { "epoch": 0.03, "learning_rate": 6.539764860852027e-06, "loss": 0.8561, "step": 18753 }, { "epoch": 0.03, "learning_rate": 6.538753584925338e-06, "loss": 0.8567, "step": 18754 }, { "epoch": 0.03, "learning_rate": 6.537742349212376e-06, "loss": 0.8898, "step": 18755 }, { "epoch": 0.03, "learning_rate": 6.536731153724891e-06, "loss": 0.8781, "step": 18756 }, { "epoch": 0.03, "learning_rate": 6.535719998474632e-06, "loss": 0.8897, "step": 18757 }, { "epoch": 0.03, "learning_rate": 6.5347088834733496e-06, "loss": 0.8815, "step": 18758 }, { "epoch": 0.03, "learning_rate": 6.5336978087327855e-06, "loss": 0.8599, "step": 18759 }, { "epoch": 0.03, "learning_rate": 6.532686774264689e-06, "loss": 0.84, "step": 18760 }, { "epoch": 0.03, "learning_rate": 6.53167578008081e-06, "loss": 0.8851, "step": 18761 }, { "epoch": 0.03, "learning_rate": 6.530664826192888e-06, "loss": 0.8843, "step": 18762 }, { "epoch": 0.03, "learning_rate": 6.529653912612674e-06, "loss": 0.8412, "step": 18763 }, { "epoch": 0.03, "learning_rate": 6.528643039351907e-06, "loss": 0.8665, "step": 18764 }, { "epoch": 0.03, "learning_rate": 6.527632206422344e-06, "loss": 0.841, "step": 18765 }, { "epoch": 0.03, "learning_rate": 6.526621413835714e-06, "loss": 0.8573, "step": 18766 }, { "epoch": 0.03, "learning_rate": 6.52561066160377e-06, "loss": 0.8932, "step": 18767 }, { "epoch": 0.03, "learning_rate": 6.524599949738255e-06, "loss": 0.8855, "step": 18768 }, { "epoch": 0.03, "learning_rate": 6.5235892782509055e-06, "loss": 0.8511, "step": 18769 }, { "epoch": 0.03, "learning_rate": 6.522578647153468e-06, "loss": 0.8934, "step": 18770 }, { "epoch": 0.03, "learning_rate": 6.5215680564576855e-06, "loss": 0.8516, "step": 18771 }, { "epoch": 0.03, "learning_rate": 6.5205575061753e-06, "loss": 0.8833, "step": 18772 }, { "epoch": 0.03, "learning_rate": 6.519546996318047e-06, "loss": 0.8483, "step": 18773 }, { "epoch": 0.03, "learning_rate": 6.518536526897672e-06, "loss": 0.908, "step": 18774 }, { "epoch": 0.03, "learning_rate": 6.517526097925914e-06, "loss": 0.8803, "step": 18775 }, { "epoch": 0.03, "learning_rate": 6.516515709414508e-06, "loss": 0.8893, "step": 18776 }, { "epoch": 0.03, "learning_rate": 6.515505361375198e-06, "loss": 0.8714, "step": 18777 }, { "epoch": 0.03, "learning_rate": 6.514495053819722e-06, "loss": 0.8855, "step": 18778 }, { "epoch": 0.03, "learning_rate": 6.513484786759818e-06, "loss": 0.917, "step": 18779 }, { "epoch": 0.03, "learning_rate": 6.512474560207221e-06, "loss": 0.8613, "step": 18780 }, { "epoch": 0.03, "learning_rate": 6.511464374173668e-06, "loss": 0.8494, "step": 18781 }, { "epoch": 0.03, "learning_rate": 6.510454228670897e-06, "loss": 0.8611, "step": 18782 }, { "epoch": 0.03, "learning_rate": 6.50944412371065e-06, "loss": 0.8519, "step": 18783 }, { "epoch": 0.03, "learning_rate": 6.5084340593046526e-06, "loss": 0.8958, "step": 18784 }, { "epoch": 0.03, "learning_rate": 6.507424035464642e-06, "loss": 0.8704, "step": 18785 }, { "epoch": 0.03, "learning_rate": 6.506414052202364e-06, "loss": 0.9196, "step": 18786 }, { "epoch": 0.03, "learning_rate": 6.505404109529539e-06, "loss": 0.862, "step": 18787 }, { "epoch": 0.03, "learning_rate": 6.504394207457902e-06, "loss": 0.8375, "step": 18788 }, { "epoch": 0.03, "learning_rate": 6.503384345999195e-06, "loss": 0.8749, "step": 18789 }, { "epoch": 0.03, "learning_rate": 6.502374525165151e-06, "loss": 0.8978, "step": 18790 }, { "epoch": 0.03, "learning_rate": 6.50136474496749e-06, "loss": 0.8457, "step": 18791 }, { "epoch": 0.03, "learning_rate": 6.500355005417955e-06, "loss": 0.8571, "step": 18792 }, { "epoch": 0.03, "learning_rate": 6.499345306528276e-06, "loss": 0.855, "step": 18793 }, { "epoch": 0.03, "learning_rate": 6.4983356483101765e-06, "loss": 0.8635, "step": 18794 }, { "epoch": 0.03, "learning_rate": 6.497326030775396e-06, "loss": 0.8903, "step": 18795 }, { "epoch": 0.03, "learning_rate": 6.496316453935661e-06, "loss": 0.8525, "step": 18796 }, { "epoch": 0.03, "learning_rate": 6.495306917802702e-06, "loss": 0.8534, "step": 18797 }, { "epoch": 0.03, "learning_rate": 6.494297422388245e-06, "loss": 0.887, "step": 18798 }, { "epoch": 0.03, "learning_rate": 6.493287967704021e-06, "loss": 0.8497, "step": 18799 }, { "epoch": 0.03, "learning_rate": 6.49227855376176e-06, "loss": 0.8939, "step": 18800 }, { "epoch": 0.03, "learning_rate": 6.491269180573185e-06, "loss": 0.8365, "step": 18801 }, { "epoch": 0.04, "learning_rate": 6.490259848150027e-06, "loss": 0.8503, "step": 18802 }, { "epoch": 0.04, "learning_rate": 6.48925055650401e-06, "loss": 0.8851, "step": 18803 }, { "epoch": 0.04, "learning_rate": 6.488241305646865e-06, "loss": 0.9104, "step": 18804 }, { "epoch": 0.04, "learning_rate": 6.487232095590312e-06, "loss": 0.8659, "step": 18805 }, { "epoch": 0.04, "learning_rate": 6.4862229263460775e-06, "loss": 0.9022, "step": 18806 }, { "epoch": 0.04, "learning_rate": 6.485213797925891e-06, "loss": 0.9038, "step": 18807 }, { "epoch": 0.04, "learning_rate": 6.48420471034147e-06, "loss": 0.8134, "step": 18808 }, { "epoch": 0.04, "learning_rate": 6.483195663604544e-06, "loss": 0.8489, "step": 18809 }, { "epoch": 0.04, "learning_rate": 6.48218665772683e-06, "loss": 0.8997, "step": 18810 }, { "epoch": 0.04, "learning_rate": 6.481177692720062e-06, "loss": 0.8465, "step": 18811 }, { "epoch": 0.04, "learning_rate": 6.480168768595948e-06, "loss": 0.8519, "step": 18812 }, { "epoch": 0.04, "learning_rate": 6.47915988536622e-06, "loss": 0.863, "step": 18813 }, { "epoch": 0.04, "learning_rate": 6.478151043042602e-06, "loss": 0.8727, "step": 18814 }, { "epoch": 0.04, "learning_rate": 6.477142241636802e-06, "loss": 0.8549, "step": 18815 }, { "epoch": 0.04, "learning_rate": 6.476133481160553e-06, "loss": 0.8635, "step": 18816 }, { "epoch": 0.04, "learning_rate": 6.475124761625569e-06, "loss": 0.8082, "step": 18817 }, { "epoch": 0.04, "learning_rate": 6.4741160830435735e-06, "loss": 0.8794, "step": 18818 }, { "epoch": 0.04, "learning_rate": 6.47310744542628e-06, "loss": 0.9069, "step": 18819 }, { "epoch": 0.04, "learning_rate": 6.472098848785411e-06, "loss": 0.8296, "step": 18820 }, { "epoch": 0.04, "learning_rate": 6.471090293132687e-06, "loss": 0.9161, "step": 18821 }, { "epoch": 0.04, "learning_rate": 6.4700817784798195e-06, "loss": 0.8589, "step": 18822 }, { "epoch": 0.04, "learning_rate": 6.469073304838529e-06, "loss": 0.8549, "step": 18823 }, { "epoch": 0.04, "learning_rate": 6.468064872220532e-06, "loss": 0.8583, "step": 18824 }, { "epoch": 0.04, "learning_rate": 6.467056480637547e-06, "loss": 0.8492, "step": 18825 }, { "epoch": 0.04, "learning_rate": 6.466048130101286e-06, "loss": 0.8643, "step": 18826 }, { "epoch": 0.04, "learning_rate": 6.465039820623465e-06, "loss": 0.8469, "step": 18827 }, { "epoch": 0.04, "learning_rate": 6.464031552215799e-06, "loss": 0.8619, "step": 18828 }, { "epoch": 0.04, "learning_rate": 6.463023324890007e-06, "loss": 0.8418, "step": 18829 }, { "epoch": 0.04, "learning_rate": 6.462015138657798e-06, "loss": 0.8832, "step": 18830 }, { "epoch": 0.04, "learning_rate": 6.461006993530883e-06, "loss": 0.9003, "step": 18831 }, { "epoch": 0.04, "learning_rate": 6.459998889520984e-06, "loss": 0.8517, "step": 18832 }, { "epoch": 0.04, "learning_rate": 6.458990826639804e-06, "loss": 0.9, "step": 18833 }, { "epoch": 0.04, "learning_rate": 6.457982804899056e-06, "loss": 0.8414, "step": 18834 }, { "epoch": 0.04, "learning_rate": 6.456974824310456e-06, "loss": 0.8689, "step": 18835 }, { "epoch": 0.04, "learning_rate": 6.455966884885718e-06, "loss": 0.8799, "step": 18836 }, { "epoch": 0.04, "learning_rate": 6.454958986636541e-06, "loss": 0.9086, "step": 18837 }, { "epoch": 0.04, "learning_rate": 6.453951129574644e-06, "loss": 0.8775, "step": 18838 }, { "epoch": 0.04, "learning_rate": 6.452943313711736e-06, "loss": 0.8998, "step": 18839 }, { "epoch": 0.04, "learning_rate": 6.451935539059522e-06, "loss": 0.869, "step": 18840 }, { "epoch": 0.04, "learning_rate": 6.450927805629714e-06, "loss": 0.831, "step": 18841 }, { "epoch": 0.04, "learning_rate": 6.449920113434018e-06, "loss": 0.9004, "step": 18842 }, { "epoch": 0.04, "learning_rate": 6.448912462484146e-06, "loss": 0.887, "step": 18843 }, { "epoch": 0.04, "learning_rate": 6.4479048527917996e-06, "loss": 0.8564, "step": 18844 }, { "epoch": 0.04, "learning_rate": 6.446897284368687e-06, "loss": 0.882, "step": 18845 }, { "epoch": 0.04, "learning_rate": 6.445889757226517e-06, "loss": 0.8636, "step": 18846 }, { "epoch": 0.04, "learning_rate": 6.444882271376993e-06, "loss": 0.8487, "step": 18847 }, { "epoch": 0.04, "learning_rate": 6.44387482683182e-06, "loss": 0.8427, "step": 18848 }, { "epoch": 0.04, "learning_rate": 6.442867423602704e-06, "loss": 0.8527, "step": 18849 }, { "epoch": 0.04, "learning_rate": 6.441860061701349e-06, "loss": 0.8371, "step": 18850 }, { "epoch": 0.04, "learning_rate": 6.440852741139458e-06, "loss": 0.8836, "step": 18851 }, { "epoch": 0.04, "learning_rate": 6.439845461928735e-06, "loss": 0.8875, "step": 18852 }, { "epoch": 0.04, "learning_rate": 6.4388382240808855e-06, "loss": 0.8561, "step": 18853 }, { "epoch": 0.04, "learning_rate": 6.437831027607606e-06, "loss": 0.8909, "step": 18854 }, { "epoch": 0.04, "learning_rate": 6.4368238725206e-06, "loss": 0.8737, "step": 18855 }, { "epoch": 0.04, "learning_rate": 6.435816758831572e-06, "loss": 0.8497, "step": 18856 }, { "epoch": 0.04, "learning_rate": 6.434809686552227e-06, "loss": 0.8665, "step": 18857 }, { "epoch": 0.04, "learning_rate": 6.433802655694252e-06, "loss": 0.9141, "step": 18858 }, { "epoch": 0.04, "learning_rate": 6.432795666269359e-06, "loss": 0.8601, "step": 18859 }, { "epoch": 0.04, "learning_rate": 6.431788718289247e-06, "loss": 0.8958, "step": 18860 }, { "epoch": 0.04, "learning_rate": 6.430781811765604e-06, "loss": 0.9014, "step": 18861 }, { "epoch": 0.04, "learning_rate": 6.429774946710139e-06, "loss": 0.8524, "step": 18862 }, { "epoch": 0.04, "learning_rate": 6.428768123134546e-06, "loss": 0.8495, "step": 18863 }, { "epoch": 0.04, "learning_rate": 6.427761341050525e-06, "loss": 0.8384, "step": 18864 }, { "epoch": 0.04, "learning_rate": 6.426754600469771e-06, "loss": 0.8943, "step": 18865 }, { "epoch": 0.04, "learning_rate": 6.425747901403981e-06, "loss": 0.8403, "step": 18866 }, { "epoch": 0.04, "learning_rate": 6.4247412438648514e-06, "loss": 0.8707, "step": 18867 }, { "epoch": 0.04, "learning_rate": 6.423734627864078e-06, "loss": 0.8145, "step": 18868 }, { "epoch": 0.04, "learning_rate": 6.422728053413354e-06, "loss": 0.8605, "step": 18869 }, { "epoch": 0.04, "learning_rate": 6.421721520524377e-06, "loss": 0.9178, "step": 18870 }, { "epoch": 0.04, "learning_rate": 6.420715029208841e-06, "loss": 0.8466, "step": 18871 }, { "epoch": 0.04, "learning_rate": 6.419708579478436e-06, "loss": 0.8682, "step": 18872 }, { "epoch": 0.04, "learning_rate": 6.41870217134486e-06, "loss": 0.863, "step": 18873 }, { "epoch": 0.04, "learning_rate": 6.417695804819799e-06, "loss": 0.9179, "step": 18874 }, { "epoch": 0.04, "learning_rate": 6.416689479914957e-06, "loss": 0.8221, "step": 18875 }, { "epoch": 0.04, "learning_rate": 6.415683196642014e-06, "loss": 0.8432, "step": 18876 }, { "epoch": 0.04, "learning_rate": 6.414676955012664e-06, "loss": 0.8531, "step": 18877 }, { "epoch": 0.04, "learning_rate": 6.413670755038606e-06, "loss": 0.8612, "step": 18878 }, { "epoch": 0.04, "learning_rate": 6.412664596731519e-06, "loss": 0.8501, "step": 18879 }, { "epoch": 0.04, "learning_rate": 6.411658480103097e-06, "loss": 0.9421, "step": 18880 }, { "epoch": 0.04, "learning_rate": 6.410652405165031e-06, "loss": 0.8838, "step": 18881 }, { "epoch": 0.04, "learning_rate": 6.409646371929013e-06, "loss": 0.8877, "step": 18882 }, { "epoch": 0.04, "learning_rate": 6.408640380406722e-06, "loss": 0.8428, "step": 18883 }, { "epoch": 0.04, "learning_rate": 6.407634430609853e-06, "loss": 0.8758, "step": 18884 }, { "epoch": 0.04, "learning_rate": 6.406628522550093e-06, "loss": 0.8853, "step": 18885 }, { "epoch": 0.04, "learning_rate": 6.405622656239127e-06, "loss": 0.8696, "step": 18886 }, { "epoch": 0.04, "learning_rate": 6.404616831688641e-06, "loss": 0.8942, "step": 18887 }, { "epoch": 0.04, "learning_rate": 6.403611048910321e-06, "loss": 0.8165, "step": 18888 }, { "epoch": 0.04, "learning_rate": 6.402605307915857e-06, "loss": 0.8585, "step": 18889 }, { "epoch": 0.04, "learning_rate": 6.401599608716929e-06, "loss": 0.8564, "step": 18890 }, { "epoch": 0.04, "learning_rate": 6.40059395132522e-06, "loss": 0.8694, "step": 18891 }, { "epoch": 0.04, "learning_rate": 6.3995883357524205e-06, "loss": 0.8729, "step": 18892 }, { "epoch": 0.04, "learning_rate": 6.398582762010208e-06, "loss": 0.8473, "step": 18893 }, { "epoch": 0.04, "learning_rate": 6.397577230110269e-06, "loss": 0.8589, "step": 18894 }, { "epoch": 0.04, "learning_rate": 6.396571740064283e-06, "loss": 0.8921, "step": 18895 }, { "epoch": 0.04, "learning_rate": 6.395566291883939e-06, "loss": 0.9081, "step": 18896 }, { "epoch": 0.04, "learning_rate": 6.39456088558091e-06, "loss": 0.8471, "step": 18897 }, { "epoch": 0.04, "learning_rate": 6.39355552116688e-06, "loss": 0.8922, "step": 18898 }, { "epoch": 0.04, "learning_rate": 6.392550198653533e-06, "loss": 0.8424, "step": 18899 }, { "epoch": 0.04, "learning_rate": 6.391544918052544e-06, "loss": 0.872, "step": 18900 }, { "epoch": 0.04, "learning_rate": 6.3905396793755936e-06, "loss": 0.8848, "step": 18901 }, { "epoch": 0.04, "learning_rate": 6.389534482634365e-06, "loss": 0.9049, "step": 18902 }, { "epoch": 0.04, "learning_rate": 6.388529327840538e-06, "loss": 0.9055, "step": 18903 }, { "epoch": 0.04, "learning_rate": 6.387524215005781e-06, "loss": 0.9117, "step": 18904 }, { "epoch": 0.04, "learning_rate": 6.386519144141779e-06, "loss": 0.8959, "step": 18905 }, { "epoch": 0.04, "learning_rate": 6.385514115260212e-06, "loss": 0.8934, "step": 18906 }, { "epoch": 0.04, "learning_rate": 6.384509128372747e-06, "loss": 0.8604, "step": 18907 }, { "epoch": 0.04, "learning_rate": 6.383504183491067e-06, "loss": 0.8691, "step": 18908 }, { "epoch": 0.04, "learning_rate": 6.382499280626847e-06, "loss": 0.8297, "step": 18909 }, { "epoch": 0.04, "learning_rate": 6.381494419791765e-06, "loss": 0.8402, "step": 18910 }, { "epoch": 0.04, "learning_rate": 6.3804896009974886e-06, "loss": 0.8685, "step": 18911 }, { "epoch": 0.04, "learning_rate": 6.379484824255697e-06, "loss": 0.8882, "step": 18912 }, { "epoch": 0.04, "learning_rate": 6.378480089578065e-06, "loss": 0.8444, "step": 18913 }, { "epoch": 0.04, "learning_rate": 6.377475396976263e-06, "loss": 0.9031, "step": 18914 }, { "epoch": 0.04, "learning_rate": 6.376470746461963e-06, "loss": 0.8834, "step": 18915 }, { "epoch": 0.04, "learning_rate": 6.37546613804684e-06, "loss": 0.9179, "step": 18916 }, { "epoch": 0.04, "learning_rate": 6.374461571742568e-06, "loss": 0.87, "step": 18917 }, { "epoch": 0.04, "learning_rate": 6.37345704756081e-06, "loss": 0.847, "step": 18918 }, { "epoch": 0.04, "learning_rate": 6.372452565513244e-06, "loss": 0.8856, "step": 18919 }, { "epoch": 0.04, "learning_rate": 6.371448125611535e-06, "loss": 0.8968, "step": 18920 }, { "epoch": 0.04, "learning_rate": 6.370443727867363e-06, "loss": 0.8346, "step": 18921 }, { "epoch": 0.04, "learning_rate": 6.369439372292388e-06, "loss": 0.852, "step": 18922 }, { "epoch": 0.04, "learning_rate": 6.368435058898279e-06, "loss": 0.8724, "step": 18923 }, { "epoch": 0.04, "learning_rate": 6.367430787696711e-06, "loss": 0.8311, "step": 18924 }, { "epoch": 0.04, "learning_rate": 6.366426558699346e-06, "loss": 0.8403, "step": 18925 }, { "epoch": 0.04, "learning_rate": 6.36542237191785e-06, "loss": 0.8651, "step": 18926 }, { "epoch": 0.04, "learning_rate": 6.364418227363895e-06, "loss": 0.8526, "step": 18927 }, { "epoch": 0.04, "learning_rate": 6.36341412504915e-06, "loss": 0.8629, "step": 18928 }, { "epoch": 0.04, "learning_rate": 6.362410064985272e-06, "loss": 0.9066, "step": 18929 }, { "epoch": 0.04, "learning_rate": 6.36140604718393e-06, "loss": 0.8437, "step": 18930 }, { "epoch": 0.04, "learning_rate": 6.3604020716567924e-06, "loss": 0.8908, "step": 18931 }, { "epoch": 0.04, "learning_rate": 6.35939813841552e-06, "loss": 0.8984, "step": 18932 }, { "epoch": 0.04, "learning_rate": 6.358394247471779e-06, "loss": 0.861, "step": 18933 }, { "epoch": 0.04, "learning_rate": 6.357390398837231e-06, "loss": 0.8288, "step": 18934 }, { "epoch": 0.04, "learning_rate": 6.356386592523541e-06, "loss": 0.8679, "step": 18935 }, { "epoch": 0.04, "learning_rate": 6.355382828542369e-06, "loss": 0.8884, "step": 18936 }, { "epoch": 0.04, "learning_rate": 6.354379106905378e-06, "loss": 0.9218, "step": 18937 }, { "epoch": 0.04, "learning_rate": 6.3533754276242336e-06, "loss": 0.8253, "step": 18938 }, { "epoch": 0.04, "learning_rate": 6.3523717907105885e-06, "loss": 0.8555, "step": 18939 }, { "epoch": 0.04, "learning_rate": 6.3513681961761084e-06, "loss": 0.8553, "step": 18940 }, { "epoch": 0.04, "learning_rate": 6.350364644032454e-06, "loss": 0.8664, "step": 18941 }, { "epoch": 0.04, "learning_rate": 6.349361134291286e-06, "loss": 0.8706, "step": 18942 }, { "epoch": 0.04, "learning_rate": 6.348357666964257e-06, "loss": 0.8795, "step": 18943 }, { "epoch": 0.04, "learning_rate": 6.3473542420630305e-06, "loss": 0.8245, "step": 18944 }, { "epoch": 0.04, "learning_rate": 6.346350859599266e-06, "loss": 0.8701, "step": 18945 }, { "epoch": 0.04, "learning_rate": 6.3453475195846155e-06, "loss": 0.8749, "step": 18946 }, { "epoch": 0.04, "learning_rate": 6.344344222030738e-06, "loss": 0.8991, "step": 18947 }, { "epoch": 0.04, "learning_rate": 6.343340966949294e-06, "loss": 0.8761, "step": 18948 }, { "epoch": 0.04, "learning_rate": 6.342337754351942e-06, "loss": 0.8534, "step": 18949 }, { "epoch": 0.04, "learning_rate": 6.341334584250325e-06, "loss": 0.8574, "step": 18950 }, { "epoch": 0.04, "learning_rate": 6.340331456656108e-06, "loss": 0.8841, "step": 18951 }, { "epoch": 0.04, "learning_rate": 6.339328371580949e-06, "loss": 0.8396, "step": 18952 }, { "epoch": 0.04, "learning_rate": 6.338325329036489e-06, "loss": 0.871, "step": 18953 }, { "epoch": 0.04, "learning_rate": 6.337322329034391e-06, "loss": 0.8421, "step": 18954 }, { "epoch": 0.04, "learning_rate": 6.336319371586306e-06, "loss": 0.8448, "step": 18955 }, { "epoch": 0.04, "learning_rate": 6.335316456703891e-06, "loss": 0.8781, "step": 18956 }, { "epoch": 0.04, "learning_rate": 6.3343135843987905e-06, "loss": 0.8705, "step": 18957 }, { "epoch": 0.04, "learning_rate": 6.333310754682659e-06, "loss": 0.867, "step": 18958 }, { "epoch": 0.04, "learning_rate": 6.33230796756715e-06, "loss": 0.8859, "step": 18959 }, { "epoch": 0.04, "learning_rate": 6.331305223063913e-06, "loss": 0.8749, "step": 18960 }, { "epoch": 0.04, "learning_rate": 6.330302521184597e-06, "loss": 0.8702, "step": 18961 }, { "epoch": 0.04, "learning_rate": 6.329299861940852e-06, "loss": 0.9201, "step": 18962 }, { "epoch": 0.04, "learning_rate": 6.328297245344328e-06, "loss": 0.8475, "step": 18963 }, { "epoch": 0.04, "learning_rate": 6.327294671406673e-06, "loss": 0.8845, "step": 18964 }, { "epoch": 0.04, "learning_rate": 6.326292140139535e-06, "loss": 0.871, "step": 18965 }, { "epoch": 0.04, "learning_rate": 6.32528965155456e-06, "loss": 0.9033, "step": 18966 }, { "epoch": 0.04, "learning_rate": 6.324287205663403e-06, "loss": 0.895, "step": 18967 }, { "epoch": 0.04, "learning_rate": 6.323284802477702e-06, "loss": 0.8572, "step": 18968 }, { "epoch": 0.04, "learning_rate": 6.322282442009102e-06, "loss": 0.8542, "step": 18969 }, { "epoch": 0.04, "learning_rate": 6.321280124269261e-06, "loss": 0.8889, "step": 18970 }, { "epoch": 0.04, "learning_rate": 6.320277849269811e-06, "loss": 0.8661, "step": 18971 }, { "epoch": 0.04, "learning_rate": 6.3192756170224e-06, "loss": 0.8532, "step": 18972 }, { "epoch": 0.04, "learning_rate": 6.318273427538677e-06, "loss": 0.8843, "step": 18973 }, { "epoch": 0.04, "learning_rate": 6.317271280830284e-06, "loss": 0.8546, "step": 18974 }, { "epoch": 0.04, "learning_rate": 6.31626917690886e-06, "loss": 0.8619, "step": 18975 }, { "epoch": 0.04, "learning_rate": 6.315267115786052e-06, "loss": 0.8641, "step": 18976 }, { "epoch": 0.04, "learning_rate": 6.314265097473503e-06, "loss": 0.8835, "step": 18977 }, { "epoch": 0.04, "learning_rate": 6.31326312198285e-06, "loss": 0.8842, "step": 18978 }, { "epoch": 0.04, "learning_rate": 6.312261189325737e-06, "loss": 0.8412, "step": 18979 }, { "epoch": 0.04, "learning_rate": 6.311259299513804e-06, "loss": 0.8938, "step": 18980 }, { "epoch": 0.04, "learning_rate": 6.310257452558694e-06, "loss": 0.8796, "step": 18981 }, { "epoch": 0.04, "learning_rate": 6.309255648472044e-06, "loss": 0.9054, "step": 18982 }, { "epoch": 0.04, "learning_rate": 6.308253887265493e-06, "loss": 0.8616, "step": 18983 }, { "epoch": 0.04, "learning_rate": 6.307252168950682e-06, "loss": 0.8551, "step": 18984 }, { "epoch": 0.04, "learning_rate": 6.3062504935392465e-06, "loss": 0.8509, "step": 18985 }, { "epoch": 0.04, "learning_rate": 6.3052488610428255e-06, "loss": 0.885, "step": 18986 }, { "epoch": 0.04, "learning_rate": 6.304247271473057e-06, "loss": 0.9174, "step": 18987 }, { "epoch": 0.04, "learning_rate": 6.303245724841577e-06, "loss": 0.8812, "step": 18988 }, { "epoch": 0.04, "learning_rate": 6.302244221160021e-06, "loss": 0.8671, "step": 18989 }, { "epoch": 0.04, "learning_rate": 6.301242760440023e-06, "loss": 0.8416, "step": 18990 }, { "epoch": 0.04, "learning_rate": 6.3002413426932275e-06, "loss": 0.8703, "step": 18991 }, { "epoch": 0.04, "learning_rate": 6.299239967931258e-06, "loss": 0.9165, "step": 18992 }, { "epoch": 0.04, "learning_rate": 6.298238636165751e-06, "loss": 0.8832, "step": 18993 }, { "epoch": 0.04, "learning_rate": 6.297237347408345e-06, "loss": 0.8893, "step": 18994 }, { "epoch": 0.04, "learning_rate": 6.296236101670675e-06, "loss": 0.8567, "step": 18995 }, { "epoch": 0.04, "learning_rate": 6.295234898964363e-06, "loss": 0.8508, "step": 18996 }, { "epoch": 0.04, "learning_rate": 6.29423373930105e-06, "loss": 0.8968, "step": 18997 }, { "epoch": 0.04, "learning_rate": 6.29323262269237e-06, "loss": 0.8623, "step": 18998 }, { "epoch": 0.04, "learning_rate": 6.292231549149943e-06, "loss": 0.8644, "step": 18999 }, { "epoch": 0.04, "learning_rate": 6.29123051868541e-06, "loss": 0.8904, "step": 19000 }, { "epoch": 0.04, "learning_rate": 6.290229531310397e-06, "loss": 0.8744, "step": 19001 }, { "epoch": 0.04, "learning_rate": 6.289228587036535e-06, "loss": 0.8501, "step": 19002 }, { "epoch": 0.04, "learning_rate": 6.288227685875453e-06, "loss": 0.8485, "step": 19003 }, { "epoch": 0.04, "learning_rate": 6.287226827838779e-06, "loss": 0.9195, "step": 19004 }, { "epoch": 0.04, "learning_rate": 6.2862260129381415e-06, "loss": 0.8736, "step": 19005 }, { "epoch": 0.04, "learning_rate": 6.285225241185171e-06, "loss": 0.8559, "step": 19006 }, { "epoch": 0.04, "learning_rate": 6.284224512591491e-06, "loss": 0.8845, "step": 19007 }, { "epoch": 0.04, "learning_rate": 6.283223827168728e-06, "loss": 0.8742, "step": 19008 }, { "epoch": 0.04, "learning_rate": 6.2822231849285124e-06, "loss": 0.8743, "step": 19009 }, { "epoch": 0.04, "learning_rate": 6.2812225858824645e-06, "loss": 0.8313, "step": 19010 }, { "epoch": 0.04, "learning_rate": 6.280222030042214e-06, "loss": 0.8665, "step": 19011 }, { "epoch": 0.04, "learning_rate": 6.27922151741938e-06, "loss": 0.8181, "step": 19012 }, { "epoch": 0.04, "learning_rate": 6.278221048025599e-06, "loss": 0.886, "step": 19013 }, { "epoch": 0.04, "learning_rate": 6.27722062187248e-06, "loss": 0.8892, "step": 19014 }, { "epoch": 0.04, "learning_rate": 6.276220238971653e-06, "loss": 0.8501, "step": 19015 }, { "epoch": 0.04, "learning_rate": 6.275219899334746e-06, "loss": 0.8509, "step": 19016 }, { "epoch": 0.04, "learning_rate": 6.27421960297337e-06, "loss": 0.8513, "step": 19017 }, { "epoch": 0.04, "learning_rate": 6.273219349899152e-06, "loss": 0.882, "step": 19018 }, { "epoch": 0.04, "learning_rate": 6.272219140123715e-06, "loss": 0.8867, "step": 19019 }, { "epoch": 0.04, "learning_rate": 6.271218973658681e-06, "loss": 0.8475, "step": 19020 }, { "epoch": 0.04, "learning_rate": 6.2702188505156644e-06, "loss": 0.8551, "step": 19021 }, { "epoch": 0.04, "learning_rate": 6.269218770706289e-06, "loss": 0.8765, "step": 19022 }, { "epoch": 0.04, "learning_rate": 6.268218734242175e-06, "loss": 0.8512, "step": 19023 }, { "epoch": 0.04, "learning_rate": 6.2672187411349374e-06, "loss": 0.8314, "step": 19024 }, { "epoch": 0.04, "learning_rate": 6.266218791396195e-06, "loss": 0.8777, "step": 19025 }, { "epoch": 0.04, "learning_rate": 6.265218885037567e-06, "loss": 0.9198, "step": 19026 }, { "epoch": 0.04, "learning_rate": 6.264219022070672e-06, "loss": 0.9217, "step": 19027 }, { "epoch": 0.04, "learning_rate": 6.263219202507125e-06, "loss": 0.8632, "step": 19028 }, { "epoch": 0.04, "learning_rate": 6.2622194263585405e-06, "loss": 0.8782, "step": 19029 }, { "epoch": 0.04, "learning_rate": 6.261219693636538e-06, "loss": 0.8837, "step": 19030 }, { "epoch": 0.04, "learning_rate": 6.260220004352727e-06, "loss": 0.9229, "step": 19031 }, { "epoch": 0.04, "learning_rate": 6.259220358518727e-06, "loss": 0.8937, "step": 19032 }, { "epoch": 0.04, "learning_rate": 6.258220756146151e-06, "loss": 0.9042, "step": 19033 }, { "epoch": 0.04, "learning_rate": 6.257221197246614e-06, "loss": 0.8931, "step": 19034 }, { "epoch": 0.04, "learning_rate": 6.256221681831727e-06, "loss": 0.8527, "step": 19035 }, { "epoch": 0.04, "learning_rate": 6.255222209913099e-06, "loss": 0.8859, "step": 19036 }, { "epoch": 0.04, "learning_rate": 6.254222781502354e-06, "loss": 0.8839, "step": 19037 }, { "epoch": 0.04, "learning_rate": 6.253223396611092e-06, "loss": 0.9113, "step": 19038 }, { "epoch": 0.04, "learning_rate": 6.252224055250925e-06, "loss": 0.8623, "step": 19039 }, { "epoch": 0.04, "learning_rate": 6.251224757433469e-06, "loss": 0.8637, "step": 19040 }, { "epoch": 0.04, "learning_rate": 6.250225503170337e-06, "loss": 0.9105, "step": 19041 }, { "epoch": 0.04, "learning_rate": 6.2492262924731275e-06, "loss": 0.8717, "step": 19042 }, { "epoch": 0.04, "learning_rate": 6.2482271253534565e-06, "loss": 0.8634, "step": 19043 }, { "epoch": 0.04, "learning_rate": 6.247228001822937e-06, "loss": 0.8962, "step": 19044 }, { "epoch": 0.04, "learning_rate": 6.246228921893165e-06, "loss": 0.8838, "step": 19045 }, { "epoch": 0.04, "learning_rate": 6.245229885575758e-06, "loss": 0.8579, "step": 19046 }, { "epoch": 0.04, "learning_rate": 6.24423089288232e-06, "loss": 0.8048, "step": 19047 }, { "epoch": 0.04, "learning_rate": 6.243231943824457e-06, "loss": 0.8715, "step": 19048 }, { "epoch": 0.04, "learning_rate": 6.242233038413776e-06, "loss": 0.8745, "step": 19049 }, { "epoch": 0.04, "learning_rate": 6.241234176661881e-06, "loss": 0.875, "step": 19050 }, { "epoch": 0.04, "learning_rate": 6.24023535858038e-06, "loss": 0.9063, "step": 19051 }, { "epoch": 0.04, "learning_rate": 6.239236584180876e-06, "loss": 0.8462, "step": 19052 }, { "epoch": 0.04, "learning_rate": 6.2382378534749735e-06, "loss": 0.8958, "step": 19053 }, { "epoch": 0.04, "learning_rate": 6.237239166474273e-06, "loss": 0.8922, "step": 19054 }, { "epoch": 0.04, "learning_rate": 6.236240523190382e-06, "loss": 0.8566, "step": 19055 }, { "epoch": 0.04, "learning_rate": 6.2352419236349e-06, "loss": 0.8408, "step": 19056 }, { "epoch": 0.04, "learning_rate": 6.2342433678194305e-06, "loss": 0.8224, "step": 19057 }, { "epoch": 0.04, "learning_rate": 6.233244855755572e-06, "loss": 0.8846, "step": 19058 }, { "epoch": 0.04, "learning_rate": 6.232246387454932e-06, "loss": 0.8637, "step": 19059 }, { "epoch": 0.04, "learning_rate": 6.231247962929104e-06, "loss": 0.8573, "step": 19060 }, { "epoch": 0.04, "learning_rate": 6.23024958218969e-06, "loss": 0.8834, "step": 19061 }, { "epoch": 0.04, "learning_rate": 6.229251245248296e-06, "loss": 0.857, "step": 19062 }, { "epoch": 0.04, "learning_rate": 6.228252952116509e-06, "loss": 0.8575, "step": 19063 }, { "epoch": 0.04, "learning_rate": 6.227254702805936e-06, "loss": 0.8828, "step": 19064 }, { "epoch": 0.04, "learning_rate": 6.226256497328171e-06, "loss": 0.8483, "step": 19065 }, { "epoch": 0.04, "learning_rate": 6.225258335694818e-06, "loss": 0.8754, "step": 19066 }, { "epoch": 0.04, "learning_rate": 6.224260217917466e-06, "loss": 0.8684, "step": 19067 }, { "epoch": 0.04, "learning_rate": 6.223262144007713e-06, "loss": 0.8421, "step": 19068 }, { "epoch": 0.04, "learning_rate": 6.222264113977161e-06, "loss": 0.8933, "step": 19069 }, { "epoch": 0.04, "learning_rate": 6.221266127837398e-06, "loss": 0.8832, "step": 19070 }, { "epoch": 0.04, "learning_rate": 6.22026818560002e-06, "loss": 0.8432, "step": 19071 }, { "epoch": 0.04, "learning_rate": 6.219270287276624e-06, "loss": 0.8536, "step": 19072 }, { "epoch": 0.04, "learning_rate": 6.218272432878806e-06, "loss": 0.8333, "step": 19073 }, { "epoch": 0.04, "learning_rate": 6.2172746224181524e-06, "loss": 0.8286, "step": 19074 }, { "epoch": 0.04, "learning_rate": 6.216276855906261e-06, "loss": 0.8659, "step": 19075 }, { "epoch": 0.04, "learning_rate": 6.2152791333547256e-06, "loss": 0.8561, "step": 19076 }, { "epoch": 0.04, "learning_rate": 6.214281454775133e-06, "loss": 0.8493, "step": 19077 }, { "epoch": 0.04, "learning_rate": 6.213283820179076e-06, "loss": 0.9075, "step": 19078 }, { "epoch": 0.04, "learning_rate": 6.212286229578144e-06, "loss": 0.8647, "step": 19079 }, { "epoch": 0.04, "learning_rate": 6.211288682983938e-06, "loss": 0.879, "step": 19080 }, { "epoch": 0.04, "learning_rate": 6.210291180408034e-06, "loss": 0.857, "step": 19081 }, { "epoch": 0.04, "learning_rate": 6.209293721862024e-06, "loss": 0.8493, "step": 19082 }, { "epoch": 0.04, "learning_rate": 6.208296307357506e-06, "loss": 0.8657, "step": 19083 }, { "epoch": 0.04, "learning_rate": 6.207298936906059e-06, "loss": 0.904, "step": 19084 }, { "epoch": 0.04, "learning_rate": 6.20630161051927e-06, "loss": 0.8908, "step": 19085 }, { "epoch": 0.04, "learning_rate": 6.205304328208731e-06, "loss": 0.8862, "step": 19086 }, { "epoch": 0.04, "learning_rate": 6.2043070899860326e-06, "loss": 0.8799, "step": 19087 }, { "epoch": 0.04, "learning_rate": 6.203309895862748e-06, "loss": 0.8874, "step": 19088 }, { "epoch": 0.04, "learning_rate": 6.202312745850474e-06, "loss": 0.8423, "step": 19089 }, { "epoch": 0.04, "learning_rate": 6.201315639960796e-06, "loss": 0.876, "step": 19090 }, { "epoch": 0.04, "learning_rate": 6.20031857820529e-06, "loss": 0.8606, "step": 19091 }, { "epoch": 0.04, "learning_rate": 6.199321560595546e-06, "loss": 0.8968, "step": 19092 }, { "epoch": 0.04, "learning_rate": 6.198324587143148e-06, "loss": 0.8857, "step": 19093 }, { "epoch": 0.04, "learning_rate": 6.197327657859677e-06, "loss": 0.8691, "step": 19094 }, { "epoch": 0.04, "learning_rate": 6.196330772756718e-06, "loss": 0.8619, "step": 19095 }, { "epoch": 0.04, "learning_rate": 6.195333931845851e-06, "loss": 0.8525, "step": 19096 }, { "epoch": 0.04, "learning_rate": 6.194337135138656e-06, "loss": 0.8835, "step": 19097 }, { "epoch": 0.04, "learning_rate": 6.19334038264672e-06, "loss": 0.8706, "step": 19098 }, { "epoch": 0.04, "learning_rate": 6.192343674381617e-06, "loss": 0.8491, "step": 19099 }, { "epoch": 0.04, "learning_rate": 6.19134701035493e-06, "loss": 0.8769, "step": 19100 }, { "epoch": 0.04, "learning_rate": 6.1903503905782405e-06, "loss": 0.9058, "step": 19101 }, { "epoch": 0.04, "learning_rate": 6.189353815063124e-06, "loss": 0.8811, "step": 19102 }, { "epoch": 0.05, "learning_rate": 6.188357283821158e-06, "loss": 0.9096, "step": 19103 }, { "epoch": 0.05, "learning_rate": 6.1873607968639236e-06, "loss": 0.8807, "step": 19104 }, { "epoch": 0.05, "learning_rate": 6.186364354203003e-06, "loss": 0.8226, "step": 19105 }, { "epoch": 0.05, "learning_rate": 6.185367955849961e-06, "loss": 0.895, "step": 19106 }, { "epoch": 0.05, "learning_rate": 6.184371601816382e-06, "loss": 0.8641, "step": 19107 }, { "epoch": 0.05, "learning_rate": 6.183375292113844e-06, "loss": 0.8437, "step": 19108 }, { "epoch": 0.05, "learning_rate": 6.1823790267539155e-06, "loss": 0.8576, "step": 19109 }, { "epoch": 0.05, "learning_rate": 6.181382805748175e-06, "loss": 0.8609, "step": 19110 }, { "epoch": 0.05, "learning_rate": 6.180386629108197e-06, "loss": 0.8475, "step": 19111 }, { "epoch": 0.05, "learning_rate": 6.179390496845557e-06, "loss": 0.8684, "step": 19112 }, { "epoch": 0.05, "learning_rate": 6.178394408971826e-06, "loss": 0.8639, "step": 19113 }, { "epoch": 0.05, "learning_rate": 6.177398365498576e-06, "loss": 0.874, "step": 19114 }, { "epoch": 0.05, "learning_rate": 6.176402366437382e-06, "loss": 0.8683, "step": 19115 }, { "epoch": 0.05, "learning_rate": 6.175406411799813e-06, "loss": 0.8508, "step": 19116 }, { "epoch": 0.05, "learning_rate": 6.174410501597443e-06, "loss": 0.8582, "step": 19117 }, { "epoch": 0.05, "learning_rate": 6.173414635841839e-06, "loss": 0.9008, "step": 19118 }, { "epoch": 0.05, "learning_rate": 6.172418814544578e-06, "loss": 0.844, "step": 19119 }, { "epoch": 0.05, "learning_rate": 6.171423037717221e-06, "loss": 0.8556, "step": 19120 }, { "epoch": 0.05, "learning_rate": 6.170427305371344e-06, "loss": 0.7996, "step": 19121 }, { "epoch": 0.05, "learning_rate": 6.169431617518514e-06, "loss": 0.8785, "step": 19122 }, { "epoch": 0.05, "learning_rate": 6.168435974170297e-06, "loss": 0.8775, "step": 19123 }, { "epoch": 0.05, "learning_rate": 6.167440375338262e-06, "loss": 0.8599, "step": 19124 }, { "epoch": 0.05, "learning_rate": 6.166444821033974e-06, "loss": 0.8732, "step": 19125 }, { "epoch": 0.05, "learning_rate": 6.1654493112690085e-06, "loss": 0.8703, "step": 19126 }, { "epoch": 0.05, "learning_rate": 6.1644538460549204e-06, "loss": 0.8284, "step": 19127 }, { "epoch": 0.05, "learning_rate": 6.163458425403279e-06, "loss": 0.8727, "step": 19128 }, { "epoch": 0.05, "learning_rate": 6.162463049325655e-06, "loss": 0.8584, "step": 19129 }, { "epoch": 0.05, "learning_rate": 6.161467717833605e-06, "loss": 0.8959, "step": 19130 }, { "epoch": 0.05, "learning_rate": 6.1604724309386945e-06, "loss": 0.8325, "step": 19131 }, { "epoch": 0.05, "learning_rate": 6.15947718865249e-06, "loss": 0.8434, "step": 19132 }, { "epoch": 0.05, "learning_rate": 6.158481990986558e-06, "loss": 0.8633, "step": 19133 }, { "epoch": 0.05, "learning_rate": 6.15748683795245e-06, "loss": 0.8755, "step": 19134 }, { "epoch": 0.05, "learning_rate": 6.156491729561736e-06, "loss": 0.8814, "step": 19135 }, { "epoch": 0.05, "learning_rate": 6.155496665825978e-06, "loss": 0.9036, "step": 19136 }, { "epoch": 0.05, "learning_rate": 6.1545016467567345e-06, "loss": 0.8999, "step": 19137 }, { "epoch": 0.05, "learning_rate": 6.1535066723655634e-06, "loss": 0.8722, "step": 19138 }, { "epoch": 0.05, "learning_rate": 6.152511742664029e-06, "loss": 0.8558, "step": 19139 }, { "epoch": 0.05, "learning_rate": 6.15151685766369e-06, "loss": 0.8943, "step": 19140 }, { "epoch": 0.05, "learning_rate": 6.150522017376102e-06, "loss": 0.829, "step": 19141 }, { "epoch": 0.05, "learning_rate": 6.149527221812825e-06, "loss": 0.8252, "step": 19142 }, { "epoch": 0.05, "learning_rate": 6.14853247098542e-06, "loss": 0.8412, "step": 19143 }, { "epoch": 0.05, "learning_rate": 6.147537764905441e-06, "loss": 0.858, "step": 19144 }, { "epoch": 0.05, "learning_rate": 6.146543103584444e-06, "loss": 0.8526, "step": 19145 }, { "epoch": 0.05, "learning_rate": 6.145548487033987e-06, "loss": 0.8408, "step": 19146 }, { "epoch": 0.05, "learning_rate": 6.144553915265627e-06, "loss": 0.8479, "step": 19147 }, { "epoch": 0.05, "learning_rate": 6.143559388290916e-06, "loss": 0.8551, "step": 19148 }, { "epoch": 0.05, "learning_rate": 6.142564906121411e-06, "loss": 0.8523, "step": 19149 }, { "epoch": 0.05, "learning_rate": 6.141570468768663e-06, "loss": 0.8662, "step": 19150 }, { "epoch": 0.05, "learning_rate": 6.140576076244234e-06, "loss": 0.8682, "step": 19151 }, { "epoch": 0.05, "learning_rate": 6.139581728559665e-06, "loss": 0.8506, "step": 19152 }, { "epoch": 0.05, "learning_rate": 6.138587425726517e-06, "loss": 0.8955, "step": 19153 }, { "epoch": 0.05, "learning_rate": 6.1375931677563435e-06, "loss": 0.8354, "step": 19154 }, { "epoch": 0.05, "learning_rate": 6.136598954660687e-06, "loss": 0.8527, "step": 19155 }, { "epoch": 0.05, "learning_rate": 6.1356047864511056e-06, "loss": 0.8679, "step": 19156 }, { "epoch": 0.05, "learning_rate": 6.134610663139149e-06, "loss": 0.8598, "step": 19157 }, { "epoch": 0.05, "learning_rate": 6.133616584736368e-06, "loss": 0.9104, "step": 19158 }, { "epoch": 0.05, "learning_rate": 6.132622551254308e-06, "loss": 0.9122, "step": 19159 }, { "epoch": 0.05, "learning_rate": 6.131628562704521e-06, "loss": 0.8454, "step": 19160 }, { "epoch": 0.05, "learning_rate": 6.130634619098558e-06, "loss": 0.8777, "step": 19161 }, { "epoch": 0.05, "learning_rate": 6.12964072044796e-06, "loss": 0.8892, "step": 19162 }, { "epoch": 0.05, "learning_rate": 6.128646866764279e-06, "loss": 0.8981, "step": 19163 }, { "epoch": 0.05, "learning_rate": 6.127653058059062e-06, "loss": 0.8351, "step": 19164 }, { "epoch": 0.05, "learning_rate": 6.1266592943438565e-06, "loss": 0.8011, "step": 19165 }, { "epoch": 0.05, "learning_rate": 6.125665575630204e-06, "loss": 0.879, "step": 19166 }, { "epoch": 0.05, "learning_rate": 6.1246719019296504e-06, "loss": 0.8421, "step": 19167 }, { "epoch": 0.05, "learning_rate": 6.123678273253746e-06, "loss": 0.8712, "step": 19168 }, { "epoch": 0.05, "learning_rate": 6.122684689614029e-06, "loss": 0.9109, "step": 19169 }, { "epoch": 0.05, "learning_rate": 6.121691151022046e-06, "loss": 0.8764, "step": 19170 }, { "epoch": 0.05, "learning_rate": 6.120697657489336e-06, "loss": 0.8717, "step": 19171 }, { "epoch": 0.05, "learning_rate": 6.119704209027453e-06, "loss": 0.8395, "step": 19172 }, { "epoch": 0.05, "learning_rate": 6.118710805647926e-06, "loss": 0.8226, "step": 19173 }, { "epoch": 0.05, "learning_rate": 6.1177174473623e-06, "loss": 0.8971, "step": 19174 }, { "epoch": 0.05, "learning_rate": 6.1167241341821255e-06, "loss": 0.8383, "step": 19175 }, { "epoch": 0.05, "learning_rate": 6.115730866118931e-06, "loss": 0.8875, "step": 19176 }, { "epoch": 0.05, "learning_rate": 6.114737643184259e-06, "loss": 0.8578, "step": 19177 }, { "epoch": 0.05, "learning_rate": 6.113744465389654e-06, "loss": 0.8856, "step": 19178 }, { "epoch": 0.05, "learning_rate": 6.1127513327466555e-06, "loss": 0.857, "step": 19179 }, { "epoch": 0.05, "learning_rate": 6.111758245266795e-06, "loss": 0.8742, "step": 19180 }, { "epoch": 0.05, "learning_rate": 6.110765202961615e-06, "loss": 0.8875, "step": 19181 }, { "epoch": 0.05, "learning_rate": 6.109772205842653e-06, "loss": 0.8911, "step": 19182 }, { "epoch": 0.05, "learning_rate": 6.1087792539214455e-06, "loss": 0.8423, "step": 19183 }, { "epoch": 0.05, "learning_rate": 6.107786347209528e-06, "loss": 0.8778, "step": 19184 }, { "epoch": 0.05, "learning_rate": 6.106793485718436e-06, "loss": 0.857, "step": 19185 }, { "epoch": 0.05, "learning_rate": 6.105800669459708e-06, "loss": 0.862, "step": 19186 }, { "epoch": 0.05, "learning_rate": 6.104807898444875e-06, "loss": 0.8746, "step": 19187 }, { "epoch": 0.05, "learning_rate": 6.103815172685474e-06, "loss": 0.8184, "step": 19188 }, { "epoch": 0.05, "learning_rate": 6.1028224921930355e-06, "loss": 0.902, "step": 19189 }, { "epoch": 0.05, "learning_rate": 6.101829856979098e-06, "loss": 0.8707, "step": 19190 }, { "epoch": 0.05, "learning_rate": 6.10083726705519e-06, "loss": 0.8792, "step": 19191 }, { "epoch": 0.05, "learning_rate": 6.099844722432844e-06, "loss": 0.8748, "step": 19192 }, { "epoch": 0.05, "learning_rate": 6.098852223123595e-06, "loss": 0.8615, "step": 19193 }, { "epoch": 0.05, "learning_rate": 6.09785976913897e-06, "loss": 0.907, "step": 19194 }, { "epoch": 0.05, "learning_rate": 6.096867360490499e-06, "loss": 0.8402, "step": 19195 }, { "epoch": 0.05, "learning_rate": 6.095874997189714e-06, "loss": 0.882, "step": 19196 }, { "epoch": 0.05, "learning_rate": 6.09488267924815e-06, "loss": 0.8716, "step": 19197 }, { "epoch": 0.05, "learning_rate": 6.093890406677326e-06, "loss": 0.8892, "step": 19198 }, { "epoch": 0.05, "learning_rate": 6.092898179488778e-06, "loss": 0.8887, "step": 19199 }, { "epoch": 0.05, "learning_rate": 6.091905997694034e-06, "loss": 0.843, "step": 19200 }, { "epoch": 0.05, "learning_rate": 6.090913861304612e-06, "loss": 0.8862, "step": 19201 }, { "epoch": 0.05, "learning_rate": 6.08992177033205e-06, "loss": 0.872, "step": 19202 }, { "epoch": 0.05, "learning_rate": 6.088929724787869e-06, "loss": 0.8473, "step": 19203 }, { "epoch": 0.05, "learning_rate": 6.0879377246835994e-06, "loss": 0.8738, "step": 19204 }, { "epoch": 0.05, "learning_rate": 6.086945770030759e-06, "loss": 0.8618, "step": 19205 }, { "epoch": 0.05, "learning_rate": 6.085953860840877e-06, "loss": 0.9127, "step": 19206 }, { "epoch": 0.05, "learning_rate": 6.084961997125481e-06, "loss": 0.8521, "step": 19207 }, { "epoch": 0.05, "learning_rate": 6.083970178896088e-06, "loss": 0.8678, "step": 19208 }, { "epoch": 0.05, "learning_rate": 6.0829784061642234e-06, "loss": 0.8581, "step": 19209 }, { "epoch": 0.05, "learning_rate": 6.0819866789414124e-06, "loss": 0.8436, "step": 19210 }, { "epoch": 0.05, "learning_rate": 6.080994997239178e-06, "loss": 0.864, "step": 19211 }, { "epoch": 0.05, "learning_rate": 6.080003361069037e-06, "loss": 0.8389, "step": 19212 }, { "epoch": 0.05, "learning_rate": 6.079011770442512e-06, "loss": 0.8662, "step": 19213 }, { "epoch": 0.05, "learning_rate": 6.078020225371128e-06, "loss": 0.8764, "step": 19214 }, { "epoch": 0.05, "learning_rate": 6.077028725866398e-06, "loss": 0.8799, "step": 19215 }, { "epoch": 0.05, "learning_rate": 6.0760372719398476e-06, "loss": 0.8501, "step": 19216 }, { "epoch": 0.05, "learning_rate": 6.075045863602989e-06, "loss": 0.8552, "step": 19217 }, { "epoch": 0.05, "learning_rate": 6.074054500867352e-06, "loss": 0.8692, "step": 19218 }, { "epoch": 0.05, "learning_rate": 6.073063183744443e-06, "loss": 0.8298, "step": 19219 }, { "epoch": 0.05, "learning_rate": 6.072071912245783e-06, "loss": 0.8716, "step": 19220 }, { "epoch": 0.05, "learning_rate": 6.071080686382894e-06, "loss": 0.8954, "step": 19221 }, { "epoch": 0.05, "learning_rate": 6.070089506167285e-06, "loss": 0.9025, "step": 19222 }, { "epoch": 0.05, "learning_rate": 6.069098371610474e-06, "loss": 0.8118, "step": 19223 }, { "epoch": 0.05, "learning_rate": 6.068107282723977e-06, "loss": 0.8805, "step": 19224 }, { "epoch": 0.05, "learning_rate": 6.067116239519315e-06, "loss": 0.8857, "step": 19225 }, { "epoch": 0.05, "learning_rate": 6.066125242007989e-06, "loss": 0.8891, "step": 19226 }, { "epoch": 0.05, "learning_rate": 6.065134290201522e-06, "loss": 0.8474, "step": 19227 }, { "epoch": 0.05, "learning_rate": 6.064143384111427e-06, "loss": 0.9066, "step": 19228 }, { "epoch": 0.05, "learning_rate": 6.063152523749214e-06, "loss": 0.8574, "step": 19229 }, { "epoch": 0.05, "learning_rate": 6.062161709126394e-06, "loss": 0.878, "step": 19230 }, { "epoch": 0.05, "learning_rate": 6.061170940254481e-06, "loss": 0.8605, "step": 19231 }, { "epoch": 0.05, "learning_rate": 6.060180217144986e-06, "loss": 0.831, "step": 19232 }, { "epoch": 0.05, "learning_rate": 6.059189539809418e-06, "loss": 0.8891, "step": 19233 }, { "epoch": 0.05, "learning_rate": 6.058198908259288e-06, "loss": 0.8955, "step": 19234 }, { "epoch": 0.05, "learning_rate": 6.057208322506105e-06, "loss": 0.8584, "step": 19235 }, { "epoch": 0.05, "learning_rate": 6.05621778256138e-06, "loss": 0.8427, "step": 19236 }, { "epoch": 0.05, "learning_rate": 6.055227288436618e-06, "loss": 0.8462, "step": 19237 }, { "epoch": 0.05, "learning_rate": 6.054236840143328e-06, "loss": 0.8474, "step": 19238 }, { "epoch": 0.05, "learning_rate": 6.053246437693019e-06, "loss": 0.8841, "step": 19239 }, { "epoch": 0.05, "learning_rate": 6.052256081097195e-06, "loss": 0.8782, "step": 19240 }, { "epoch": 0.05, "learning_rate": 6.051265770367365e-06, "loss": 0.8275, "step": 19241 }, { "epoch": 0.05, "learning_rate": 6.050275505515029e-06, "loss": 0.8373, "step": 19242 }, { "epoch": 0.05, "learning_rate": 6.049285286551704e-06, "loss": 0.8339, "step": 19243 }, { "epoch": 0.05, "learning_rate": 6.04829511348888e-06, "loss": 0.9033, "step": 19244 }, { "epoch": 0.05, "learning_rate": 6.047304986338072e-06, "loss": 0.8534, "step": 19245 }, { "epoch": 0.05, "learning_rate": 6.046314905110782e-06, "loss": 0.874, "step": 19246 }, { "epoch": 0.05, "learning_rate": 6.045324869818506e-06, "loss": 0.8969, "step": 19247 }, { "epoch": 0.05, "learning_rate": 6.044334880472753e-06, "loss": 0.8393, "step": 19248 }, { "epoch": 0.05, "learning_rate": 6.043344937085023e-06, "loss": 0.8742, "step": 19249 }, { "epoch": 0.05, "learning_rate": 6.042355039666821e-06, "loss": 0.8697, "step": 19250 }, { "epoch": 0.05, "learning_rate": 6.041365188229641e-06, "loss": 0.8597, "step": 19251 }, { "epoch": 0.05, "learning_rate": 6.040375382784988e-06, "loss": 0.866, "step": 19252 }, { "epoch": 0.05, "learning_rate": 6.039385623344362e-06, "loss": 0.8645, "step": 19253 }, { "epoch": 0.05, "learning_rate": 6.038395909919261e-06, "loss": 0.8455, "step": 19254 }, { "epoch": 0.05, "learning_rate": 6.037406242521183e-06, "loss": 0.8684, "step": 19255 }, { "epoch": 0.05, "learning_rate": 6.036416621161627e-06, "loss": 0.8714, "step": 19256 }, { "epoch": 0.05, "learning_rate": 6.0354270458520936e-06, "loss": 0.8582, "step": 19257 }, { "epoch": 0.05, "learning_rate": 6.034437516604075e-06, "loss": 0.8863, "step": 19258 }, { "epoch": 0.05, "learning_rate": 6.03344803342907e-06, "loss": 0.8503, "step": 19259 }, { "epoch": 0.05, "learning_rate": 6.032458596338577e-06, "loss": 0.903, "step": 19260 }, { "epoch": 0.05, "learning_rate": 6.031469205344087e-06, "loss": 0.8666, "step": 19261 }, { "epoch": 0.05, "learning_rate": 6.030479860457098e-06, "loss": 0.8966, "step": 19262 }, { "epoch": 0.05, "learning_rate": 6.029490561689102e-06, "loss": 0.8687, "step": 19263 }, { "epoch": 0.05, "learning_rate": 6.0285013090516e-06, "loss": 0.8688, "step": 19264 }, { "epoch": 0.05, "learning_rate": 6.027512102556077e-06, "loss": 0.8703, "step": 19265 }, { "epoch": 0.05, "learning_rate": 6.026522942214026e-06, "loss": 0.8974, "step": 19266 }, { "epoch": 0.05, "learning_rate": 6.025533828036949e-06, "loss": 0.8891, "step": 19267 }, { "epoch": 0.05, "learning_rate": 6.024544760036327e-06, "loss": 0.8328, "step": 19268 }, { "epoch": 0.05, "learning_rate": 6.023555738223654e-06, "loss": 0.879, "step": 19269 }, { "epoch": 0.05, "learning_rate": 6.0225667626104245e-06, "loss": 0.8704, "step": 19270 }, { "epoch": 0.05, "learning_rate": 6.0215778332081275e-06, "loss": 0.8424, "step": 19271 }, { "epoch": 0.05, "learning_rate": 6.02058895002825e-06, "loss": 0.882, "step": 19272 }, { "epoch": 0.05, "learning_rate": 6.019600113082282e-06, "loss": 0.8487, "step": 19273 }, { "epoch": 0.05, "learning_rate": 6.018611322381717e-06, "loss": 0.834, "step": 19274 }, { "epoch": 0.05, "learning_rate": 6.017622577938035e-06, "loss": 0.8763, "step": 19275 }, { "epoch": 0.05, "learning_rate": 6.016633879762728e-06, "loss": 0.8848, "step": 19276 }, { "epoch": 0.05, "learning_rate": 6.015645227867282e-06, "loss": 0.8829, "step": 19277 }, { "epoch": 0.05, "learning_rate": 6.0146566222631864e-06, "loss": 0.8879, "step": 19278 }, { "epoch": 0.05, "learning_rate": 6.013668062961921e-06, "loss": 0.8708, "step": 19279 }, { "epoch": 0.05, "learning_rate": 6.012679549974976e-06, "loss": 0.8795, "step": 19280 }, { "epoch": 0.05, "learning_rate": 6.011691083313835e-06, "loss": 0.8557, "step": 19281 }, { "epoch": 0.05, "learning_rate": 6.010702662989984e-06, "loss": 0.8659, "step": 19282 }, { "epoch": 0.05, "learning_rate": 6.009714289014903e-06, "loss": 0.868, "step": 19283 }, { "epoch": 0.05, "learning_rate": 6.008725961400076e-06, "loss": 0.895, "step": 19284 }, { "epoch": 0.05, "learning_rate": 6.007737680156989e-06, "loss": 0.8621, "step": 19285 }, { "epoch": 0.05, "learning_rate": 6.00674944529712e-06, "loss": 0.8621, "step": 19286 }, { "epoch": 0.05, "learning_rate": 6.005761256831951e-06, "loss": 0.8997, "step": 19287 }, { "epoch": 0.05, "learning_rate": 6.004773114772966e-06, "loss": 0.8376, "step": 19288 }, { "epoch": 0.05, "learning_rate": 6.003785019131649e-06, "loss": 0.8817, "step": 19289 }, { "epoch": 0.05, "learning_rate": 6.002796969919468e-06, "loss": 0.8693, "step": 19290 }, { "epoch": 0.05, "learning_rate": 6.0018089671479105e-06, "loss": 0.8453, "step": 19291 }, { "epoch": 0.05, "learning_rate": 6.00082101082846e-06, "loss": 0.868, "step": 19292 }, { "epoch": 0.05, "learning_rate": 5.999833100972583e-06, "loss": 0.8619, "step": 19293 }, { "epoch": 0.05, "learning_rate": 5.998845237591765e-06, "loss": 0.8591, "step": 19294 }, { "epoch": 0.05, "learning_rate": 5.9978574206974816e-06, "loss": 0.857, "step": 19295 }, { "epoch": 0.05, "learning_rate": 5.9968696503012125e-06, "loss": 0.852, "step": 19296 }, { "epoch": 0.05, "learning_rate": 5.995881926414428e-06, "loss": 0.8442, "step": 19297 }, { "epoch": 0.05, "learning_rate": 5.9948942490486084e-06, "loss": 0.8546, "step": 19298 }, { "epoch": 0.05, "learning_rate": 5.993906618215228e-06, "loss": 0.8581, "step": 19299 }, { "epoch": 0.05, "learning_rate": 5.992919033925759e-06, "loss": 0.83, "step": 19300 }, { "epoch": 0.05, "learning_rate": 5.991931496191676e-06, "loss": 0.8342, "step": 19301 }, { "epoch": 0.05, "learning_rate": 5.990944005024455e-06, "loss": 0.8448, "step": 19302 }, { "epoch": 0.05, "learning_rate": 5.989956560435569e-06, "loss": 0.8993, "step": 19303 }, { "epoch": 0.05, "learning_rate": 5.9889691624364866e-06, "loss": 0.8677, "step": 19304 }, { "epoch": 0.05, "learning_rate": 5.987981811038683e-06, "loss": 0.8579, "step": 19305 }, { "epoch": 0.05, "learning_rate": 5.986994506253629e-06, "loss": 0.8622, "step": 19306 }, { "epoch": 0.05, "learning_rate": 5.986007248092793e-06, "loss": 0.8717, "step": 19307 }, { "epoch": 0.05, "learning_rate": 5.985020036567648e-06, "loss": 0.8858, "step": 19308 }, { "epoch": 0.05, "learning_rate": 5.98403287168966e-06, "loss": 0.845, "step": 19309 }, { "epoch": 0.05, "learning_rate": 5.983045753470308e-06, "loss": 0.8727, "step": 19310 }, { "epoch": 0.05, "learning_rate": 5.9820586819210495e-06, "loss": 0.8839, "step": 19311 }, { "epoch": 0.05, "learning_rate": 5.981071657053355e-06, "loss": 0.8348, "step": 19312 }, { "epoch": 0.05, "learning_rate": 5.980084678878697e-06, "loss": 0.8568, "step": 19313 }, { "epoch": 0.05, "learning_rate": 5.979097747408538e-06, "loss": 0.8826, "step": 19314 }, { "epoch": 0.05, "learning_rate": 5.9781108626543425e-06, "loss": 0.8516, "step": 19315 }, { "epoch": 0.05, "learning_rate": 5.977124024627581e-06, "loss": 0.8409, "step": 19316 }, { "epoch": 0.05, "learning_rate": 5.9761372333397205e-06, "loss": 0.8713, "step": 19317 }, { "epoch": 0.05, "learning_rate": 5.975150488802221e-06, "loss": 0.8959, "step": 19318 }, { "epoch": 0.05, "learning_rate": 5.974163791026547e-06, "loss": 0.8573, "step": 19319 }, { "epoch": 0.05, "learning_rate": 5.973177140024167e-06, "loss": 0.8366, "step": 19320 }, { "epoch": 0.05, "learning_rate": 5.972190535806538e-06, "loss": 0.8795, "step": 19321 }, { "epoch": 0.05, "learning_rate": 5.971203978385125e-06, "loss": 0.8635, "step": 19322 }, { "epoch": 0.05, "learning_rate": 5.970217467771392e-06, "loss": 0.8829, "step": 19323 }, { "epoch": 0.05, "learning_rate": 5.969231003976801e-06, "loss": 0.8323, "step": 19324 }, { "epoch": 0.05, "learning_rate": 5.968244587012808e-06, "loss": 0.8982, "step": 19325 }, { "epoch": 0.05, "learning_rate": 5.967258216890878e-06, "loss": 0.8834, "step": 19326 }, { "epoch": 0.05, "learning_rate": 5.96627189362247e-06, "loss": 0.8856, "step": 19327 }, { "epoch": 0.05, "learning_rate": 5.965285617219043e-06, "loss": 0.8733, "step": 19328 }, { "epoch": 0.05, "learning_rate": 5.964299387692055e-06, "loss": 0.8491, "step": 19329 }, { "epoch": 0.05, "learning_rate": 5.9633132050529655e-06, "loss": 0.8427, "step": 19330 }, { "epoch": 0.05, "learning_rate": 5.962327069313234e-06, "loss": 0.8585, "step": 19331 }, { "epoch": 0.05, "learning_rate": 5.961340980484311e-06, "loss": 0.8918, "step": 19332 }, { "epoch": 0.05, "learning_rate": 5.960354938577659e-06, "loss": 0.8548, "step": 19333 }, { "epoch": 0.05, "learning_rate": 5.959368943604733e-06, "loss": 0.8755, "step": 19334 }, { "epoch": 0.05, "learning_rate": 5.958382995576994e-06, "loss": 0.8675, "step": 19335 }, { "epoch": 0.05, "learning_rate": 5.957397094505883e-06, "loss": 0.8587, "step": 19336 }, { "epoch": 0.05, "learning_rate": 5.956411240402866e-06, "loss": 0.8472, "step": 19337 }, { "epoch": 0.05, "learning_rate": 5.955425433279398e-06, "loss": 0.8755, "step": 19338 }, { "epoch": 0.05, "learning_rate": 5.9544396731469224e-06, "loss": 0.8713, "step": 19339 }, { "epoch": 0.05, "learning_rate": 5.9534539600169e-06, "loss": 0.8412, "step": 19340 }, { "epoch": 0.05, "learning_rate": 5.952468293900779e-06, "loss": 0.8834, "step": 19341 }, { "epoch": 0.05, "learning_rate": 5.951482674810018e-06, "loss": 0.8895, "step": 19342 }, { "epoch": 0.05, "learning_rate": 5.950497102756059e-06, "loss": 0.8333, "step": 19343 }, { "epoch": 0.05, "learning_rate": 5.949511577750357e-06, "loss": 0.838, "step": 19344 }, { "epoch": 0.05, "learning_rate": 5.948526099804366e-06, "loss": 0.8571, "step": 19345 }, { "epoch": 0.05, "learning_rate": 5.947540668929528e-06, "loss": 0.8714, "step": 19346 }, { "epoch": 0.05, "learning_rate": 5.946555285137295e-06, "loss": 0.8694, "step": 19347 }, { "epoch": 0.05, "learning_rate": 5.945569948439117e-06, "loss": 0.899, "step": 19348 }, { "epoch": 0.05, "learning_rate": 5.944584658846443e-06, "loss": 0.8587, "step": 19349 }, { "epoch": 0.05, "learning_rate": 5.9435994163707156e-06, "loss": 0.8603, "step": 19350 }, { "epoch": 0.05, "learning_rate": 5.942614221023386e-06, "loss": 0.8349, "step": 19351 }, { "epoch": 0.05, "learning_rate": 5.9416290728159e-06, "loss": 0.8435, "step": 19352 }, { "epoch": 0.05, "learning_rate": 5.940643971759699e-06, "loss": 0.8835, "step": 19353 }, { "epoch": 0.05, "learning_rate": 5.939658917866233e-06, "loss": 0.8648, "step": 19354 }, { "epoch": 0.05, "learning_rate": 5.938673911146942e-06, "loss": 0.8789, "step": 19355 }, { "epoch": 0.05, "learning_rate": 5.937688951613278e-06, "loss": 0.8799, "step": 19356 }, { "epoch": 0.05, "learning_rate": 5.936704039276678e-06, "loss": 0.8822, "step": 19357 }, { "epoch": 0.05, "learning_rate": 5.935719174148583e-06, "loss": 0.8576, "step": 19358 }, { "epoch": 0.05, "learning_rate": 5.934734356240446e-06, "loss": 0.9011, "step": 19359 }, { "epoch": 0.05, "learning_rate": 5.933749585563695e-06, "loss": 0.8616, "step": 19360 }, { "epoch": 0.05, "learning_rate": 5.932764862129781e-06, "loss": 0.8806, "step": 19361 }, { "epoch": 0.05, "learning_rate": 5.93178018595014e-06, "loss": 0.8658, "step": 19362 }, { "epoch": 0.05, "learning_rate": 5.930795557036218e-06, "loss": 0.8647, "step": 19363 }, { "epoch": 0.05, "learning_rate": 5.929810975399449e-06, "loss": 0.8466, "step": 19364 }, { "epoch": 0.05, "learning_rate": 5.928826441051274e-06, "loss": 0.8584, "step": 19365 }, { "epoch": 0.05, "learning_rate": 5.927841954003134e-06, "loss": 0.8564, "step": 19366 }, { "epoch": 0.05, "learning_rate": 5.926857514266462e-06, "loss": 0.8864, "step": 19367 }, { "epoch": 0.05, "learning_rate": 5.9258731218526984e-06, "loss": 0.8495, "step": 19368 }, { "epoch": 0.05, "learning_rate": 5.924888776773281e-06, "loss": 0.8531, "step": 19369 }, { "epoch": 0.05, "learning_rate": 5.923904479039646e-06, "loss": 0.8477, "step": 19370 }, { "epoch": 0.05, "learning_rate": 5.922920228663227e-06, "loss": 0.8575, "step": 19371 }, { "epoch": 0.05, "learning_rate": 5.921936025655462e-06, "loss": 0.9163, "step": 19372 }, { "epoch": 0.05, "learning_rate": 5.9209518700277825e-06, "loss": 0.8866, "step": 19373 }, { "epoch": 0.05, "learning_rate": 5.919967761791627e-06, "loss": 0.8894, "step": 19374 }, { "epoch": 0.05, "learning_rate": 5.918983700958425e-06, "loss": 0.8445, "step": 19375 }, { "epoch": 0.05, "learning_rate": 5.917999687539611e-06, "loss": 0.841, "step": 19376 }, { "epoch": 0.05, "learning_rate": 5.917015721546622e-06, "loss": 0.8337, "step": 19377 }, { "epoch": 0.05, "learning_rate": 5.916031802990883e-06, "loss": 0.8437, "step": 19378 }, { "epoch": 0.05, "learning_rate": 5.915047931883825e-06, "loss": 0.8909, "step": 19379 }, { "epoch": 0.05, "learning_rate": 5.914064108236885e-06, "loss": 0.8789, "step": 19380 }, { "epoch": 0.05, "learning_rate": 5.913080332061495e-06, "loss": 0.8403, "step": 19381 }, { "epoch": 0.05, "learning_rate": 5.9120966033690745e-06, "loss": 0.8899, "step": 19382 }, { "epoch": 0.05, "learning_rate": 5.911112922171061e-06, "loss": 0.8934, "step": 19383 }, { "epoch": 0.05, "learning_rate": 5.910129288478883e-06, "loss": 0.8637, "step": 19384 }, { "epoch": 0.05, "learning_rate": 5.909145702303962e-06, "loss": 0.852, "step": 19385 }, { "epoch": 0.05, "learning_rate": 5.908162163657731e-06, "loss": 0.8781, "step": 19386 }, { "epoch": 0.05, "learning_rate": 5.907178672551616e-06, "loss": 0.8278, "step": 19387 }, { "epoch": 0.05, "learning_rate": 5.906195228997047e-06, "loss": 0.8756, "step": 19388 }, { "epoch": 0.05, "learning_rate": 5.905211833005445e-06, "loss": 0.8166, "step": 19389 }, { "epoch": 0.05, "learning_rate": 5.9042284845882345e-06, "loss": 0.9005, "step": 19390 }, { "epoch": 0.05, "learning_rate": 5.903245183756846e-06, "loss": 0.8522, "step": 19391 }, { "epoch": 0.05, "learning_rate": 5.902261930522698e-06, "loss": 0.8864, "step": 19392 }, { "epoch": 0.05, "learning_rate": 5.901278724897217e-06, "loss": 0.8528, "step": 19393 }, { "epoch": 0.05, "learning_rate": 5.900295566891825e-06, "loss": 0.8764, "step": 19394 }, { "epoch": 0.05, "learning_rate": 5.899312456517949e-06, "loss": 0.8683, "step": 19395 }, { "epoch": 0.05, "learning_rate": 5.898329393787003e-06, "loss": 0.8921, "step": 19396 }, { "epoch": 0.05, "learning_rate": 5.897346378710416e-06, "loss": 0.8132, "step": 19397 }, { "epoch": 0.05, "learning_rate": 5.896363411299606e-06, "loss": 0.8568, "step": 19398 }, { "epoch": 0.05, "learning_rate": 5.895380491565991e-06, "loss": 0.8716, "step": 19399 }, { "epoch": 0.05, "learning_rate": 5.894397619520993e-06, "loss": 0.8288, "step": 19400 }, { "epoch": 0.05, "learning_rate": 5.89341479517603e-06, "loss": 0.8486, "step": 19401 }, { "epoch": 0.05, "learning_rate": 5.892432018542528e-06, "loss": 0.89, "step": 19402 }, { "epoch": 0.06, "learning_rate": 5.8914492896318955e-06, "loss": 0.8293, "step": 19403 }, { "epoch": 0.06, "learning_rate": 5.890466608455552e-06, "loss": 0.8855, "step": 19404 }, { "epoch": 0.06, "learning_rate": 5.889483975024922e-06, "loss": 0.8954, "step": 19405 }, { "epoch": 0.06, "learning_rate": 5.8885013893514106e-06, "loss": 0.8974, "step": 19406 }, { "epoch": 0.06, "learning_rate": 5.887518851446442e-06, "loss": 0.8954, "step": 19407 }, { "epoch": 0.06, "learning_rate": 5.886536361321429e-06, "loss": 0.9199, "step": 19408 }, { "epoch": 0.06, "learning_rate": 5.885553918987789e-06, "loss": 0.8565, "step": 19409 }, { "epoch": 0.06, "learning_rate": 5.884571524456931e-06, "loss": 0.9216, "step": 19410 }, { "epoch": 0.06, "learning_rate": 5.883589177740272e-06, "loss": 0.8859, "step": 19411 }, { "epoch": 0.06, "learning_rate": 5.8826068788492284e-06, "loss": 0.8464, "step": 19412 }, { "epoch": 0.06, "learning_rate": 5.881624627795206e-06, "loss": 0.895, "step": 19413 }, { "epoch": 0.06, "learning_rate": 5.88064242458962e-06, "loss": 0.875, "step": 19414 }, { "epoch": 0.06, "learning_rate": 5.879660269243883e-06, "loss": 0.8486, "step": 19415 }, { "epoch": 0.06, "learning_rate": 5.878678161769406e-06, "loss": 0.8923, "step": 19416 }, { "epoch": 0.06, "learning_rate": 5.877696102177597e-06, "loss": 0.8572, "step": 19417 }, { "epoch": 0.06, "learning_rate": 5.876714090479867e-06, "loss": 0.8596, "step": 19418 }, { "epoch": 0.06, "learning_rate": 5.875732126687626e-06, "loss": 0.8723, "step": 19419 }, { "epoch": 0.06, "learning_rate": 5.874750210812283e-06, "loss": 0.8336, "step": 19420 }, { "epoch": 0.06, "learning_rate": 5.873768342865245e-06, "loss": 0.8878, "step": 19421 }, { "epoch": 0.06, "learning_rate": 5.872786522857918e-06, "loss": 0.8663, "step": 19422 }, { "epoch": 0.06, "learning_rate": 5.8718047508017174e-06, "loss": 0.8544, "step": 19423 }, { "epoch": 0.06, "learning_rate": 5.870823026708038e-06, "loss": 0.852, "step": 19424 }, { "epoch": 0.06, "learning_rate": 5.86984135058829e-06, "loss": 0.8732, "step": 19425 }, { "epoch": 0.06, "learning_rate": 5.868859722453883e-06, "loss": 0.8732, "step": 19426 }, { "epoch": 0.06, "learning_rate": 5.867878142316221e-06, "loss": 0.8449, "step": 19427 }, { "epoch": 0.06, "learning_rate": 5.866896610186701e-06, "loss": 0.8775, "step": 19428 }, { "epoch": 0.06, "learning_rate": 5.865915126076734e-06, "loss": 0.8916, "step": 19429 }, { "epoch": 0.06, "learning_rate": 5.864933689997725e-06, "loss": 0.8588, "step": 19430 }, { "epoch": 0.06, "learning_rate": 5.863952301961065e-06, "loss": 0.899, "step": 19431 }, { "epoch": 0.06, "learning_rate": 5.862970961978167e-06, "loss": 0.9116, "step": 19432 }, { "epoch": 0.06, "learning_rate": 5.86198967006043e-06, "loss": 0.903, "step": 19433 }, { "epoch": 0.06, "learning_rate": 5.861008426219255e-06, "loss": 0.884, "step": 19434 }, { "epoch": 0.06, "learning_rate": 5.860027230466039e-06, "loss": 0.8911, "step": 19435 }, { "epoch": 0.06, "learning_rate": 5.8590460828121845e-06, "loss": 0.8519, "step": 19436 }, { "epoch": 0.06, "learning_rate": 5.858064983269092e-06, "loss": 0.864, "step": 19437 }, { "epoch": 0.06, "learning_rate": 5.8570839318481574e-06, "loss": 0.8663, "step": 19438 }, { "epoch": 0.06, "learning_rate": 5.856102928560779e-06, "loss": 0.849, "step": 19439 }, { "epoch": 0.06, "learning_rate": 5.855121973418356e-06, "loss": 0.864, "step": 19440 }, { "epoch": 0.06, "learning_rate": 5.854141066432288e-06, "loss": 0.8477, "step": 19441 }, { "epoch": 0.06, "learning_rate": 5.8531602076139645e-06, "loss": 0.8622, "step": 19442 }, { "epoch": 0.06, "learning_rate": 5.8521793969747865e-06, "loss": 0.8628, "step": 19443 }, { "epoch": 0.06, "learning_rate": 5.85119863452615e-06, "loss": 0.8551, "step": 19444 }, { "epoch": 0.06, "learning_rate": 5.850217920279444e-06, "loss": 0.8513, "step": 19445 }, { "epoch": 0.06, "learning_rate": 5.849237254246067e-06, "loss": 0.8591, "step": 19446 }, { "epoch": 0.06, "learning_rate": 5.8482566364374114e-06, "loss": 0.9047, "step": 19447 }, { "epoch": 0.06, "learning_rate": 5.847276066864877e-06, "loss": 0.8755, "step": 19448 }, { "epoch": 0.06, "learning_rate": 5.8462955455398426e-06, "loss": 0.8386, "step": 19449 }, { "epoch": 0.06, "learning_rate": 5.845315072473712e-06, "loss": 0.877, "step": 19450 }, { "epoch": 0.06, "learning_rate": 5.844334647677875e-06, "loss": 0.8624, "step": 19451 }, { "epoch": 0.06, "learning_rate": 5.843354271163715e-06, "loss": 0.899, "step": 19452 }, { "epoch": 0.06, "learning_rate": 5.842373942942628e-06, "loss": 0.8585, "step": 19453 }, { "epoch": 0.06, "learning_rate": 5.841393663026005e-06, "loss": 0.8521, "step": 19454 }, { "epoch": 0.06, "learning_rate": 5.8404134314252336e-06, "loss": 0.853, "step": 19455 }, { "epoch": 0.06, "learning_rate": 5.8394332481517e-06, "loss": 0.8688, "step": 19456 }, { "epoch": 0.06, "learning_rate": 5.838453113216795e-06, "loss": 0.8726, "step": 19457 }, { "epoch": 0.06, "learning_rate": 5.837473026631909e-06, "loss": 0.8612, "step": 19458 }, { "epoch": 0.06, "learning_rate": 5.836492988408422e-06, "loss": 0.8819, "step": 19459 }, { "epoch": 0.06, "learning_rate": 5.835512998557717e-06, "loss": 0.8526, "step": 19460 }, { "epoch": 0.06, "learning_rate": 5.834533057091194e-06, "loss": 0.9357, "step": 19461 }, { "epoch": 0.06, "learning_rate": 5.8335531640202325e-06, "loss": 0.8861, "step": 19462 }, { "epoch": 0.06, "learning_rate": 5.832573319356211e-06, "loss": 0.8548, "step": 19463 }, { "epoch": 0.06, "learning_rate": 5.831593523110519e-06, "loss": 0.8589, "step": 19464 }, { "epoch": 0.06, "learning_rate": 5.830613775294538e-06, "loss": 0.8352, "step": 19465 }, { "epoch": 0.06, "learning_rate": 5.8296340759196526e-06, "loss": 0.9103, "step": 19466 }, { "epoch": 0.06, "learning_rate": 5.8286544249972445e-06, "loss": 0.8375, "step": 19467 }, { "epoch": 0.06, "learning_rate": 5.827674822538696e-06, "loss": 0.8421, "step": 19468 }, { "epoch": 0.06, "learning_rate": 5.826695268555391e-06, "loss": 0.8437, "step": 19469 }, { "epoch": 0.06, "learning_rate": 5.825715763058703e-06, "loss": 0.8442, "step": 19470 }, { "epoch": 0.06, "learning_rate": 5.824736306060017e-06, "loss": 0.8127, "step": 19471 }, { "epoch": 0.06, "learning_rate": 5.823756897570711e-06, "loss": 0.8357, "step": 19472 }, { "epoch": 0.06, "learning_rate": 5.8227775376021664e-06, "loss": 0.8525, "step": 19473 }, { "epoch": 0.06, "learning_rate": 5.82179822616576e-06, "loss": 0.8996, "step": 19474 }, { "epoch": 0.06, "learning_rate": 5.8208189632728705e-06, "loss": 0.8235, "step": 19475 }, { "epoch": 0.06, "learning_rate": 5.819839748934879e-06, "loss": 0.8782, "step": 19476 }, { "epoch": 0.06, "learning_rate": 5.818860583163153e-06, "loss": 0.8486, "step": 19477 }, { "epoch": 0.06, "learning_rate": 5.817881465969074e-06, "loss": 0.8824, "step": 19478 }, { "epoch": 0.06, "learning_rate": 5.8169023973640135e-06, "loss": 0.888, "step": 19479 }, { "epoch": 0.06, "learning_rate": 5.815923377359358e-06, "loss": 0.8458, "step": 19480 }, { "epoch": 0.06, "learning_rate": 5.814944405966471e-06, "loss": 0.8765, "step": 19481 }, { "epoch": 0.06, "learning_rate": 5.813965483196732e-06, "loss": 0.8969, "step": 19482 }, { "epoch": 0.06, "learning_rate": 5.812986609061515e-06, "loss": 0.8478, "step": 19483 }, { "epoch": 0.06, "learning_rate": 5.8120077835721855e-06, "loss": 0.8553, "step": 19484 }, { "epoch": 0.06, "learning_rate": 5.8110290067401165e-06, "loss": 0.8717, "step": 19485 }, { "epoch": 0.06, "learning_rate": 5.810050278576689e-06, "loss": 0.9059, "step": 19486 }, { "epoch": 0.06, "learning_rate": 5.809071599093272e-06, "loss": 0.8727, "step": 19487 }, { "epoch": 0.06, "learning_rate": 5.808092968301229e-06, "loss": 0.8697, "step": 19488 }, { "epoch": 0.06, "learning_rate": 5.807114386211934e-06, "loss": 0.8741, "step": 19489 }, { "epoch": 0.06, "learning_rate": 5.806135852836756e-06, "loss": 0.8499, "step": 19490 }, { "epoch": 0.06, "learning_rate": 5.805157368187066e-06, "loss": 0.8869, "step": 19491 }, { "epoch": 0.06, "learning_rate": 5.804178932274229e-06, "loss": 0.9051, "step": 19492 }, { "epoch": 0.06, "learning_rate": 5.803200545109614e-06, "loss": 0.8691, "step": 19493 }, { "epoch": 0.06, "learning_rate": 5.802222206704593e-06, "loss": 0.8715, "step": 19494 }, { "epoch": 0.06, "learning_rate": 5.801243917070525e-06, "loss": 0.8235, "step": 19495 }, { "epoch": 0.06, "learning_rate": 5.800265676218778e-06, "loss": 0.8458, "step": 19496 }, { "epoch": 0.06, "learning_rate": 5.799287484160718e-06, "loss": 0.8897, "step": 19497 }, { "epoch": 0.06, "learning_rate": 5.798309340907713e-06, "loss": 0.852, "step": 19498 }, { "epoch": 0.06, "learning_rate": 5.797331246471123e-06, "loss": 0.8381, "step": 19499 }, { "epoch": 0.06, "learning_rate": 5.796353200862315e-06, "loss": 0.8395, "step": 19500 }, { "epoch": 0.06, "learning_rate": 5.7953752040926526e-06, "loss": 0.8866, "step": 19501 }, { "epoch": 0.06, "learning_rate": 5.794397256173494e-06, "loss": 0.8549, "step": 19502 }, { "epoch": 0.06, "learning_rate": 5.7934193571162e-06, "loss": 0.868, "step": 19503 }, { "epoch": 0.06, "learning_rate": 5.792441506932145e-06, "loss": 0.8504, "step": 19504 }, { "epoch": 0.06, "learning_rate": 5.791463705632676e-06, "loss": 0.8545, "step": 19505 }, { "epoch": 0.06, "learning_rate": 5.790485953229158e-06, "loss": 0.8463, "step": 19506 }, { "epoch": 0.06, "learning_rate": 5.789508249732951e-06, "loss": 0.8767, "step": 19507 }, { "epoch": 0.06, "learning_rate": 5.788530595155419e-06, "loss": 0.8664, "step": 19508 }, { "epoch": 0.06, "learning_rate": 5.787552989507909e-06, "loss": 0.8491, "step": 19509 }, { "epoch": 0.06, "learning_rate": 5.786575432801788e-06, "loss": 0.8663, "step": 19510 }, { "epoch": 0.06, "learning_rate": 5.785597925048414e-06, "loss": 0.8696, "step": 19511 }, { "epoch": 0.06, "learning_rate": 5.7846204662591455e-06, "loss": 0.8959, "step": 19512 }, { "epoch": 0.06, "learning_rate": 5.78364305644533e-06, "loss": 0.8685, "step": 19513 }, { "epoch": 0.06, "learning_rate": 5.7826656956183284e-06, "loss": 0.859, "step": 19514 }, { "epoch": 0.06, "learning_rate": 5.781688383789496e-06, "loss": 0.8902, "step": 19515 }, { "epoch": 0.06, "learning_rate": 5.780711120970187e-06, "loss": 0.8673, "step": 19516 }, { "epoch": 0.06, "learning_rate": 5.7797339071717564e-06, "loss": 0.8893, "step": 19517 }, { "epoch": 0.06, "learning_rate": 5.778756742405558e-06, "loss": 0.8938, "step": 19518 }, { "epoch": 0.06, "learning_rate": 5.777779626682947e-06, "loss": 0.8473, "step": 19519 }, { "epoch": 0.06, "learning_rate": 5.776802560015269e-06, "loss": 0.8369, "step": 19520 }, { "epoch": 0.06, "learning_rate": 5.77582554241388e-06, "loss": 0.9092, "step": 19521 }, { "epoch": 0.06, "learning_rate": 5.774848573890131e-06, "loss": 0.8582, "step": 19522 }, { "epoch": 0.06, "learning_rate": 5.773871654455372e-06, "loss": 0.8257, "step": 19523 }, { "epoch": 0.06, "learning_rate": 5.7728947841209546e-06, "loss": 0.878, "step": 19524 }, { "epoch": 0.06, "learning_rate": 5.771917962898225e-06, "loss": 0.8846, "step": 19525 }, { "epoch": 0.06, "learning_rate": 5.770941190798543e-06, "loss": 0.8629, "step": 19526 }, { "epoch": 0.06, "learning_rate": 5.769964467833242e-06, "loss": 0.8665, "step": 19527 }, { "epoch": 0.06, "learning_rate": 5.768987794013672e-06, "loss": 0.8547, "step": 19528 }, { "epoch": 0.06, "learning_rate": 5.768011169351194e-06, "loss": 0.8499, "step": 19529 }, { "epoch": 0.06, "learning_rate": 5.767034593857142e-06, "loss": 0.9113, "step": 19530 }, { "epoch": 0.06, "learning_rate": 5.7660580675428645e-06, "loss": 0.8519, "step": 19531 }, { "epoch": 0.06, "learning_rate": 5.765081590419708e-06, "loss": 0.8607, "step": 19532 }, { "epoch": 0.06, "learning_rate": 5.764105162499019e-06, "loss": 0.8367, "step": 19533 }, { "epoch": 0.06, "learning_rate": 5.76312878379214e-06, "loss": 0.8874, "step": 19534 }, { "epoch": 0.06, "learning_rate": 5.762152454310415e-06, "loss": 0.8605, "step": 19535 }, { "epoch": 0.06, "learning_rate": 5.761176174065192e-06, "loss": 0.8723, "step": 19536 }, { "epoch": 0.06, "learning_rate": 5.760199943067806e-06, "loss": 0.8627, "step": 19537 }, { "epoch": 0.06, "learning_rate": 5.759223761329601e-06, "loss": 0.868, "step": 19538 }, { "epoch": 0.06, "learning_rate": 5.758247628861922e-06, "loss": 0.8841, "step": 19539 }, { "epoch": 0.06, "learning_rate": 5.757271545676106e-06, "loss": 0.8604, "step": 19540 }, { "epoch": 0.06, "learning_rate": 5.756295511783496e-06, "loss": 0.8558, "step": 19541 }, { "epoch": 0.06, "learning_rate": 5.755319527195431e-06, "loss": 0.8638, "step": 19542 }, { "epoch": 0.06, "learning_rate": 5.754343591923256e-06, "loss": 0.844, "step": 19543 }, { "epoch": 0.06, "learning_rate": 5.753367705978299e-06, "loss": 0.8695, "step": 19544 }, { "epoch": 0.06, "learning_rate": 5.752391869371905e-06, "loss": 0.8162, "step": 19545 }, { "epoch": 0.06, "learning_rate": 5.7514160821154085e-06, "loss": 0.904, "step": 19546 }, { "epoch": 0.06, "learning_rate": 5.750440344220147e-06, "loss": 0.8555, "step": 19547 }, { "epoch": 0.06, "learning_rate": 5.74946465569746e-06, "loss": 0.8606, "step": 19548 }, { "epoch": 0.06, "learning_rate": 5.748489016558679e-06, "loss": 0.8928, "step": 19549 }, { "epoch": 0.06, "learning_rate": 5.747513426815147e-06, "loss": 0.8563, "step": 19550 }, { "epoch": 0.06, "learning_rate": 5.746537886478189e-06, "loss": 0.8448, "step": 19551 }, { "epoch": 0.06, "learning_rate": 5.745562395559138e-06, "loss": 0.8746, "step": 19552 }, { "epoch": 0.06, "learning_rate": 5.744586954069336e-06, "loss": 0.8755, "step": 19553 }, { "epoch": 0.06, "learning_rate": 5.7436115620201185e-06, "loss": 0.8708, "step": 19554 }, { "epoch": 0.06, "learning_rate": 5.742636219422807e-06, "loss": 0.8663, "step": 19555 }, { "epoch": 0.06, "learning_rate": 5.741660926288738e-06, "loss": 0.8738, "step": 19556 }, { "epoch": 0.06, "learning_rate": 5.740685682629243e-06, "loss": 0.8547, "step": 19557 }, { "epoch": 0.06, "learning_rate": 5.739710488455654e-06, "loss": 0.8542, "step": 19558 }, { "epoch": 0.06, "learning_rate": 5.738735343779298e-06, "loss": 0.8574, "step": 19559 }, { "epoch": 0.06, "learning_rate": 5.737760248611507e-06, "loss": 0.8696, "step": 19560 }, { "epoch": 0.06, "learning_rate": 5.736785202963613e-06, "loss": 0.8378, "step": 19561 }, { "epoch": 0.06, "learning_rate": 5.7358102068469366e-06, "loss": 0.9082, "step": 19562 }, { "epoch": 0.06, "learning_rate": 5.73483526027281e-06, "loss": 0.8434, "step": 19563 }, { "epoch": 0.06, "learning_rate": 5.733860363252559e-06, "loss": 0.8506, "step": 19564 }, { "epoch": 0.06, "learning_rate": 5.732885515797512e-06, "loss": 0.8695, "step": 19565 }, { "epoch": 0.06, "learning_rate": 5.731910717918993e-06, "loss": 0.8666, "step": 19566 }, { "epoch": 0.06, "learning_rate": 5.730935969628331e-06, "loss": 0.8402, "step": 19567 }, { "epoch": 0.06, "learning_rate": 5.72996127093685e-06, "loss": 0.8765, "step": 19568 }, { "epoch": 0.06, "learning_rate": 5.72898662185587e-06, "loss": 0.856, "step": 19569 }, { "epoch": 0.06, "learning_rate": 5.728012022396716e-06, "loss": 0.9016, "step": 19570 }, { "epoch": 0.06, "learning_rate": 5.727037472570711e-06, "loss": 0.8553, "step": 19571 }, { "epoch": 0.06, "learning_rate": 5.7260629723891854e-06, "loss": 0.8412, "step": 19572 }, { "epoch": 0.06, "learning_rate": 5.725088521863453e-06, "loss": 0.8911, "step": 19573 }, { "epoch": 0.06, "learning_rate": 5.724114121004836e-06, "loss": 0.8651, "step": 19574 }, { "epoch": 0.06, "learning_rate": 5.723139769824663e-06, "loss": 0.8778, "step": 19575 }, { "epoch": 0.06, "learning_rate": 5.722165468334237e-06, "loss": 0.8916, "step": 19576 }, { "epoch": 0.06, "learning_rate": 5.7211912165448945e-06, "loss": 0.8753, "step": 19577 }, { "epoch": 0.06, "learning_rate": 5.720217014467948e-06, "loss": 0.8734, "step": 19578 }, { "epoch": 0.06, "learning_rate": 5.719242862114722e-06, "loss": 0.8626, "step": 19579 }, { "epoch": 0.06, "learning_rate": 5.718268759496524e-06, "loss": 0.8924, "step": 19580 }, { "epoch": 0.06, "learning_rate": 5.717294706624678e-06, "loss": 0.8431, "step": 19581 }, { "epoch": 0.06, "learning_rate": 5.716320703510498e-06, "loss": 0.9049, "step": 19582 }, { "epoch": 0.06, "learning_rate": 5.715346750165302e-06, "loss": 0.8409, "step": 19583 }, { "epoch": 0.06, "learning_rate": 5.7143728466004064e-06, "loss": 0.8818, "step": 19584 }, { "epoch": 0.06, "learning_rate": 5.713398992827126e-06, "loss": 0.8864, "step": 19585 }, { "epoch": 0.06, "learning_rate": 5.712425188856777e-06, "loss": 0.8551, "step": 19586 }, { "epoch": 0.06, "learning_rate": 5.7114514347006674e-06, "loss": 0.8311, "step": 19587 }, { "epoch": 0.06, "learning_rate": 5.7104777303701145e-06, "loss": 0.8654, "step": 19588 }, { "epoch": 0.06, "learning_rate": 5.709504075876431e-06, "loss": 0.8476, "step": 19589 }, { "epoch": 0.06, "learning_rate": 5.708530471230929e-06, "loss": 0.9017, "step": 19590 }, { "epoch": 0.06, "learning_rate": 5.707556916444919e-06, "loss": 0.8369, "step": 19591 }, { "epoch": 0.06, "learning_rate": 5.706583411529712e-06, "loss": 0.8695, "step": 19592 }, { "epoch": 0.06, "learning_rate": 5.705609956496625e-06, "loss": 0.8141, "step": 19593 }, { "epoch": 0.06, "learning_rate": 5.704636551356957e-06, "loss": 0.875, "step": 19594 }, { "epoch": 0.06, "learning_rate": 5.70366319612202e-06, "loss": 0.856, "step": 19595 }, { "epoch": 0.06, "learning_rate": 5.702689890803132e-06, "loss": 0.8479, "step": 19596 }, { "epoch": 0.06, "learning_rate": 5.701716635411591e-06, "loss": 0.8808, "step": 19597 }, { "epoch": 0.06, "learning_rate": 5.7007434299587085e-06, "loss": 0.8668, "step": 19598 }, { "epoch": 0.06, "learning_rate": 5.699770274455789e-06, "loss": 0.8303, "step": 19599 }, { "epoch": 0.06, "learning_rate": 5.698797168914146e-06, "loss": 0.8676, "step": 19600 }, { "epoch": 0.06, "learning_rate": 5.6978241133450715e-06, "loss": 0.8966, "step": 19601 }, { "epoch": 0.06, "learning_rate": 5.6968511077598845e-06, "loss": 0.8326, "step": 19602 }, { "epoch": 0.06, "learning_rate": 5.695878152169885e-06, "loss": 0.9034, "step": 19603 }, { "epoch": 0.06, "learning_rate": 5.69490524658638e-06, "loss": 0.8984, "step": 19604 }, { "epoch": 0.06, "learning_rate": 5.693932391020664e-06, "loss": 0.8802, "step": 19605 }, { "epoch": 0.06, "learning_rate": 5.692959585484046e-06, "loss": 0.8608, "step": 19606 }, { "epoch": 0.06, "learning_rate": 5.691986829987829e-06, "loss": 0.8365, "step": 19607 }, { "epoch": 0.06, "learning_rate": 5.6910141245433126e-06, "loss": 0.8616, "step": 19608 }, { "epoch": 0.06, "learning_rate": 5.6900414691618e-06, "loss": 0.8717, "step": 19609 }, { "epoch": 0.06, "learning_rate": 5.689068863854589e-06, "loss": 0.89, "step": 19610 }, { "epoch": 0.06, "learning_rate": 5.688096308632986e-06, "loss": 0.8681, "step": 19611 }, { "epoch": 0.06, "learning_rate": 5.687123803508282e-06, "loss": 0.8644, "step": 19612 }, { "epoch": 0.06, "learning_rate": 5.68615134849178e-06, "loss": 0.8495, "step": 19613 }, { "epoch": 0.06, "learning_rate": 5.685178943594776e-06, "loss": 0.8766, "step": 19614 }, { "epoch": 0.06, "learning_rate": 5.684206588828571e-06, "loss": 0.8765, "step": 19615 }, { "epoch": 0.06, "learning_rate": 5.68323428420446e-06, "loss": 0.8524, "step": 19616 }, { "epoch": 0.06, "learning_rate": 5.682262029733741e-06, "loss": 0.9085, "step": 19617 }, { "epoch": 0.06, "learning_rate": 5.681289825427712e-06, "loss": 0.8766, "step": 19618 }, { "epoch": 0.06, "learning_rate": 5.680317671297662e-06, "loss": 0.8651, "step": 19619 }, { "epoch": 0.06, "learning_rate": 5.679345567354884e-06, "loss": 0.8781, "step": 19620 }, { "epoch": 0.06, "learning_rate": 5.678373513610686e-06, "loss": 0.8545, "step": 19621 }, { "epoch": 0.06, "learning_rate": 5.677401510076348e-06, "loss": 0.8718, "step": 19622 }, { "epoch": 0.06, "learning_rate": 5.676429556763169e-06, "loss": 0.8733, "step": 19623 }, { "epoch": 0.06, "learning_rate": 5.675457653682439e-06, "loss": 0.8369, "step": 19624 }, { "epoch": 0.06, "learning_rate": 5.674485800845453e-06, "loss": 0.905, "step": 19625 }, { "epoch": 0.06, "learning_rate": 5.673513998263499e-06, "loss": 0.8547, "step": 19626 }, { "epoch": 0.06, "learning_rate": 5.67254224594787e-06, "loss": 0.8678, "step": 19627 }, { "epoch": 0.06, "learning_rate": 5.671570543909858e-06, "loss": 0.8346, "step": 19628 }, { "epoch": 0.06, "learning_rate": 5.670598892160746e-06, "loss": 0.8647, "step": 19629 }, { "epoch": 0.06, "learning_rate": 5.6696272907118274e-06, "loss": 0.8596, "step": 19630 }, { "epoch": 0.06, "learning_rate": 5.668655739574389e-06, "loss": 0.8621, "step": 19631 }, { "epoch": 0.06, "learning_rate": 5.667684238759719e-06, "loss": 0.8932, "step": 19632 }, { "epoch": 0.06, "learning_rate": 5.6667127882791055e-06, "loss": 0.9055, "step": 19633 }, { "epoch": 0.06, "learning_rate": 5.665741388143834e-06, "loss": 0.8973, "step": 19634 }, { "epoch": 0.06, "learning_rate": 5.664770038365195e-06, "loss": 0.875, "step": 19635 }, { "epoch": 0.06, "learning_rate": 5.663798738954466e-06, "loss": 0.8688, "step": 19636 }, { "epoch": 0.06, "learning_rate": 5.662827489922935e-06, "loss": 0.8863, "step": 19637 }, { "epoch": 0.06, "learning_rate": 5.661856291281884e-06, "loss": 0.8258, "step": 19638 }, { "epoch": 0.06, "learning_rate": 5.660885143042607e-06, "loss": 0.8816, "step": 19639 }, { "epoch": 0.06, "learning_rate": 5.659914045216376e-06, "loss": 0.8698, "step": 19640 }, { "epoch": 0.06, "learning_rate": 5.658942997814476e-06, "loss": 0.9005, "step": 19641 }, { "epoch": 0.06, "learning_rate": 5.657972000848191e-06, "loss": 0.8522, "step": 19642 }, { "epoch": 0.06, "learning_rate": 5.6570010543288055e-06, "loss": 0.8968, "step": 19643 }, { "epoch": 0.06, "learning_rate": 5.656030158267587e-06, "loss": 0.8985, "step": 19644 }, { "epoch": 0.06, "learning_rate": 5.65505931267583e-06, "loss": 0.8576, "step": 19645 }, { "epoch": 0.06, "learning_rate": 5.654088517564812e-06, "loss": 0.8462, "step": 19646 }, { "epoch": 0.06, "learning_rate": 5.653117772945804e-06, "loss": 0.8339, "step": 19647 }, { "epoch": 0.06, "learning_rate": 5.65214707883009e-06, "loss": 0.8782, "step": 19648 }, { "epoch": 0.06, "learning_rate": 5.6511764352289475e-06, "loss": 0.8701, "step": 19649 }, { "epoch": 0.06, "learning_rate": 5.650205842153653e-06, "loss": 0.8733, "step": 19650 }, { "epoch": 0.06, "learning_rate": 5.649235299615482e-06, "loss": 0.8366, "step": 19651 }, { "epoch": 0.06, "learning_rate": 5.648264807625712e-06, "loss": 0.8814, "step": 19652 }, { "epoch": 0.06, "learning_rate": 5.647294366195623e-06, "loss": 0.8669, "step": 19653 }, { "epoch": 0.06, "learning_rate": 5.64632397533648e-06, "loss": 0.836, "step": 19654 }, { "epoch": 0.06, "learning_rate": 5.645353635059564e-06, "loss": 0.8721, "step": 19655 }, { "epoch": 0.06, "learning_rate": 5.644383345376147e-06, "loss": 0.8635, "step": 19656 }, { "epoch": 0.06, "learning_rate": 5.643413106297501e-06, "loss": 0.879, "step": 19657 }, { "epoch": 0.06, "learning_rate": 5.642442917834899e-06, "loss": 0.8825, "step": 19658 }, { "epoch": 0.06, "learning_rate": 5.641472779999615e-06, "loss": 0.8159, "step": 19659 }, { "epoch": 0.06, "learning_rate": 5.6405026928029225e-06, "loss": 0.9047, "step": 19660 }, { "epoch": 0.06, "learning_rate": 5.639532656256083e-06, "loss": 0.8848, "step": 19661 }, { "epoch": 0.06, "learning_rate": 5.638562670370375e-06, "loss": 0.8764, "step": 19662 }, { "epoch": 0.06, "learning_rate": 5.637592735157061e-06, "loss": 0.8813, "step": 19663 }, { "epoch": 0.06, "learning_rate": 5.63662285062742e-06, "loss": 0.873, "step": 19664 }, { "epoch": 0.06, "learning_rate": 5.635653016792713e-06, "loss": 0.8572, "step": 19665 }, { "epoch": 0.06, "learning_rate": 5.634683233664209e-06, "loss": 0.8531, "step": 19666 }, { "epoch": 0.06, "learning_rate": 5.6337135012531795e-06, "loss": 0.8577, "step": 19667 }, { "epoch": 0.06, "learning_rate": 5.632743819570878e-06, "loss": 0.8485, "step": 19668 }, { "epoch": 0.06, "learning_rate": 5.631774188628586e-06, "loss": 0.8952, "step": 19669 }, { "epoch": 0.06, "learning_rate": 5.630804608437562e-06, "loss": 0.8673, "step": 19670 }, { "epoch": 0.06, "learning_rate": 5.629835079009076e-06, "loss": 0.8765, "step": 19671 }, { "epoch": 0.06, "learning_rate": 5.628865600354383e-06, "loss": 0.9024, "step": 19672 }, { "epoch": 0.06, "learning_rate": 5.627896172484753e-06, "loss": 0.875, "step": 19673 }, { "epoch": 0.06, "learning_rate": 5.626926795411447e-06, "loss": 0.8843, "step": 19674 }, { "epoch": 0.06, "learning_rate": 5.625957469145728e-06, "loss": 0.8601, "step": 19675 }, { "epoch": 0.06, "learning_rate": 5.624988193698859e-06, "loss": 0.8642, "step": 19676 }, { "epoch": 0.06, "learning_rate": 5.624018969082099e-06, "loss": 0.8895, "step": 19677 }, { "epoch": 0.06, "learning_rate": 5.623049795306715e-06, "loss": 0.8434, "step": 19678 }, { "epoch": 0.06, "learning_rate": 5.6220806723839585e-06, "loss": 0.8432, "step": 19679 }, { "epoch": 0.06, "learning_rate": 5.621111600325093e-06, "loss": 0.9003, "step": 19680 }, { "epoch": 0.06, "learning_rate": 5.620142579141377e-06, "loss": 0.8703, "step": 19681 }, { "epoch": 0.06, "learning_rate": 5.619173608844069e-06, "loss": 0.8313, "step": 19682 }, { "epoch": 0.06, "learning_rate": 5.6182046894444266e-06, "loss": 0.8787, "step": 19683 }, { "epoch": 0.06, "learning_rate": 5.617235820953707e-06, "loss": 0.8609, "step": 19684 }, { "epoch": 0.06, "learning_rate": 5.616267003383171e-06, "loss": 0.8491, "step": 19685 }, { "epoch": 0.06, "learning_rate": 5.615298236744068e-06, "loss": 0.8536, "step": 19686 }, { "epoch": 0.06, "learning_rate": 5.614329521047651e-06, "loss": 0.9098, "step": 19687 }, { "epoch": 0.06, "learning_rate": 5.613360856305184e-06, "loss": 0.8699, "step": 19688 }, { "epoch": 0.06, "learning_rate": 5.612392242527922e-06, "loss": 0.8347, "step": 19689 }, { "epoch": 0.06, "learning_rate": 5.6114236797271085e-06, "loss": 0.8643, "step": 19690 }, { "epoch": 0.06, "learning_rate": 5.6104551679140015e-06, "loss": 0.8723, "step": 19691 }, { "epoch": 0.06, "learning_rate": 5.609486707099858e-06, "loss": 0.8796, "step": 19692 }, { "epoch": 0.06, "learning_rate": 5.60851829729592e-06, "loss": 0.8811, "step": 19693 }, { "epoch": 0.06, "learning_rate": 5.607549938513446e-06, "loss": 0.8336, "step": 19694 }, { "epoch": 0.06, "learning_rate": 5.6065816307636835e-06, "loss": 0.8627, "step": 19695 }, { "epoch": 0.06, "learning_rate": 5.60561337405789e-06, "loss": 0.8866, "step": 19696 }, { "epoch": 0.06, "learning_rate": 5.604645168407305e-06, "loss": 0.836, "step": 19697 }, { "epoch": 0.06, "learning_rate": 5.603677013823181e-06, "loss": 0.8565, "step": 19698 }, { "epoch": 0.06, "learning_rate": 5.602708910316767e-06, "loss": 0.8668, "step": 19699 }, { "epoch": 0.06, "learning_rate": 5.60174085789931e-06, "loss": 0.8506, "step": 19700 }, { "epoch": 0.06, "learning_rate": 5.600772856582058e-06, "loss": 0.8628, "step": 19701 }, { "epoch": 0.06, "learning_rate": 5.599804906376256e-06, "loss": 0.8282, "step": 19702 }, { "epoch": 0.06, "learning_rate": 5.5988370072931545e-06, "loss": 0.8662, "step": 19703 }, { "epoch": 0.07, "learning_rate": 5.597869159343993e-06, "loss": 0.8363, "step": 19704 }, { "epoch": 0.07, "learning_rate": 5.596901362540017e-06, "loss": 0.8896, "step": 19705 }, { "epoch": 0.07, "learning_rate": 5.595933616892473e-06, "loss": 0.8598, "step": 19706 }, { "epoch": 0.07, "learning_rate": 5.594965922412603e-06, "loss": 0.8857, "step": 19707 }, { "epoch": 0.07, "learning_rate": 5.593998279111651e-06, "loss": 0.9028, "step": 19708 }, { "epoch": 0.07, "learning_rate": 5.5930306870008586e-06, "loss": 0.8687, "step": 19709 }, { "epoch": 0.07, "learning_rate": 5.592063146091472e-06, "loss": 0.8885, "step": 19710 }, { "epoch": 0.07, "learning_rate": 5.591095656394721e-06, "loss": 0.8386, "step": 19711 }, { "epoch": 0.07, "learning_rate": 5.590128217921855e-06, "loss": 0.9009, "step": 19712 }, { "epoch": 0.07, "learning_rate": 5.589160830684119e-06, "loss": 0.8212, "step": 19713 }, { "epoch": 0.07, "learning_rate": 5.5881934946927404e-06, "loss": 0.8715, "step": 19714 }, { "epoch": 0.07, "learning_rate": 5.587226209958965e-06, "loss": 0.8678, "step": 19715 }, { "epoch": 0.07, "learning_rate": 5.586258976494028e-06, "loss": 0.8669, "step": 19716 }, { "epoch": 0.07, "learning_rate": 5.585291794309166e-06, "loss": 0.8617, "step": 19717 }, { "epoch": 0.07, "learning_rate": 5.5843246634156215e-06, "loss": 0.8617, "step": 19718 }, { "epoch": 0.07, "learning_rate": 5.583357583824627e-06, "loss": 0.8914, "step": 19719 }, { "epoch": 0.07, "learning_rate": 5.582390555547424e-06, "loss": 0.9146, "step": 19720 }, { "epoch": 0.07, "learning_rate": 5.5814235785952365e-06, "loss": 0.8751, "step": 19721 }, { "epoch": 0.07, "learning_rate": 5.5804566529793065e-06, "loss": 0.8872, "step": 19722 }, { "epoch": 0.07, "learning_rate": 5.579489778710867e-06, "loss": 0.8556, "step": 19723 }, { "epoch": 0.07, "learning_rate": 5.5785229558011496e-06, "loss": 0.8696, "step": 19724 }, { "epoch": 0.07, "learning_rate": 5.57755618426139e-06, "loss": 0.8563, "step": 19725 }, { "epoch": 0.07, "learning_rate": 5.57658946410282e-06, "loss": 0.8877, "step": 19726 }, { "epoch": 0.07, "learning_rate": 5.575622795336673e-06, "loss": 0.8663, "step": 19727 }, { "epoch": 0.07, "learning_rate": 5.574656177974173e-06, "loss": 0.8331, "step": 19728 }, { "epoch": 0.07, "learning_rate": 5.573689612026557e-06, "loss": 0.9028, "step": 19729 }, { "epoch": 0.07, "learning_rate": 5.572723097505047e-06, "loss": 0.9174, "step": 19730 }, { "epoch": 0.07, "learning_rate": 5.571756634420886e-06, "loss": 0.8654, "step": 19731 }, { "epoch": 0.07, "learning_rate": 5.5707902227852914e-06, "loss": 0.8735, "step": 19732 }, { "epoch": 0.07, "learning_rate": 5.5698238626094935e-06, "loss": 0.8562, "step": 19733 }, { "epoch": 0.07, "learning_rate": 5.568857553904722e-06, "loss": 0.8682, "step": 19734 }, { "epoch": 0.07, "learning_rate": 5.5678912966822065e-06, "loss": 0.8626, "step": 19735 }, { "epoch": 0.07, "learning_rate": 5.566925090953159e-06, "loss": 0.8626, "step": 19736 }, { "epoch": 0.07, "learning_rate": 5.565958936728822e-06, "loss": 0.8524, "step": 19737 }, { "epoch": 0.07, "learning_rate": 5.564992834020415e-06, "loss": 0.8656, "step": 19738 }, { "epoch": 0.07, "learning_rate": 5.5640267828391585e-06, "loss": 0.8935, "step": 19739 }, { "epoch": 0.07, "learning_rate": 5.5630607831962794e-06, "loss": 0.8848, "step": 19740 }, { "epoch": 0.07, "learning_rate": 5.5620948351030006e-06, "loss": 0.8541, "step": 19741 }, { "epoch": 0.07, "learning_rate": 5.561128938570545e-06, "loss": 0.8849, "step": 19742 }, { "epoch": 0.07, "learning_rate": 5.5601630936101335e-06, "loss": 0.8883, "step": 19743 }, { "epoch": 0.07, "learning_rate": 5.55919730023299e-06, "loss": 0.8647, "step": 19744 }, { "epoch": 0.07, "learning_rate": 5.558231558450338e-06, "loss": 0.871, "step": 19745 }, { "epoch": 0.07, "learning_rate": 5.557265868273387e-06, "loss": 0.9088, "step": 19746 }, { "epoch": 0.07, "learning_rate": 5.5563002297133664e-06, "loss": 0.8755, "step": 19747 }, { "epoch": 0.07, "learning_rate": 5.555334642781491e-06, "loss": 0.8574, "step": 19748 }, { "epoch": 0.07, "learning_rate": 5.554369107488981e-06, "loss": 0.8783, "step": 19749 }, { "epoch": 0.07, "learning_rate": 5.553403623847053e-06, "loss": 0.8406, "step": 19750 }, { "epoch": 0.07, "learning_rate": 5.552438191866924e-06, "loss": 0.8393, "step": 19751 }, { "epoch": 0.07, "learning_rate": 5.5514728115598175e-06, "loss": 0.8487, "step": 19752 }, { "epoch": 0.07, "learning_rate": 5.55050748293694e-06, "loss": 0.8719, "step": 19753 }, { "epoch": 0.07, "learning_rate": 5.549542206009505e-06, "loss": 0.886, "step": 19754 }, { "epoch": 0.07, "learning_rate": 5.548576980788738e-06, "loss": 0.8617, "step": 19755 }, { "epoch": 0.07, "learning_rate": 5.547611807285851e-06, "loss": 0.8813, "step": 19756 }, { "epoch": 0.07, "learning_rate": 5.546646685512051e-06, "loss": 0.8604, "step": 19757 }, { "epoch": 0.07, "learning_rate": 5.5456816154785575e-06, "loss": 0.8607, "step": 19758 }, { "epoch": 0.07, "learning_rate": 5.544716597196582e-06, "loss": 0.8377, "step": 19759 }, { "epoch": 0.07, "learning_rate": 5.543751630677328e-06, "loss": 0.8918, "step": 19760 }, { "epoch": 0.07, "learning_rate": 5.542786715932017e-06, "loss": 0.836, "step": 19761 }, { "epoch": 0.07, "learning_rate": 5.541821852971856e-06, "loss": 0.8906, "step": 19762 }, { "epoch": 0.07, "learning_rate": 5.5408570418080585e-06, "loss": 0.8452, "step": 19763 }, { "epoch": 0.07, "learning_rate": 5.539892282451828e-06, "loss": 0.8746, "step": 19764 }, { "epoch": 0.07, "learning_rate": 5.538927574914377e-06, "loss": 0.8488, "step": 19765 }, { "epoch": 0.07, "learning_rate": 5.537962919206911e-06, "loss": 0.8856, "step": 19766 }, { "epoch": 0.07, "learning_rate": 5.536998315340639e-06, "loss": 0.8759, "step": 19767 }, { "epoch": 0.07, "learning_rate": 5.536033763326769e-06, "loss": 0.8774, "step": 19768 }, { "epoch": 0.07, "learning_rate": 5.535069263176508e-06, "loss": 0.8464, "step": 19769 }, { "epoch": 0.07, "learning_rate": 5.5341048149010635e-06, "loss": 0.8585, "step": 19770 }, { "epoch": 0.07, "learning_rate": 5.5331404185116345e-06, "loss": 0.8718, "step": 19771 }, { "epoch": 0.07, "learning_rate": 5.532176074019429e-06, "loss": 0.8638, "step": 19772 }, { "epoch": 0.07, "learning_rate": 5.53121178143565e-06, "loss": 0.9203, "step": 19773 }, { "epoch": 0.07, "learning_rate": 5.530247540771503e-06, "loss": 0.874, "step": 19774 }, { "epoch": 0.07, "learning_rate": 5.5292833520381905e-06, "loss": 0.8585, "step": 19775 }, { "epoch": 0.07, "learning_rate": 5.5283192152469135e-06, "loss": 0.8689, "step": 19776 }, { "epoch": 0.07, "learning_rate": 5.527355130408877e-06, "loss": 0.8997, "step": 19777 }, { "epoch": 0.07, "learning_rate": 5.526391097535275e-06, "loss": 0.884, "step": 19778 }, { "epoch": 0.07, "learning_rate": 5.5254271166373095e-06, "loss": 0.8703, "step": 19779 }, { "epoch": 0.07, "learning_rate": 5.5244631877261865e-06, "loss": 0.8931, "step": 19780 }, { "epoch": 0.07, "learning_rate": 5.523499310813105e-06, "loss": 0.92, "step": 19781 }, { "epoch": 0.07, "learning_rate": 5.522535485909258e-06, "loss": 0.8706, "step": 19782 }, { "epoch": 0.07, "learning_rate": 5.521571713025844e-06, "loss": 0.8381, "step": 19783 }, { "epoch": 0.07, "learning_rate": 5.520607992174062e-06, "loss": 0.8653, "step": 19784 }, { "epoch": 0.07, "learning_rate": 5.51964432336511e-06, "loss": 0.8486, "step": 19785 }, { "epoch": 0.07, "learning_rate": 5.518680706610181e-06, "loss": 0.8932, "step": 19786 }, { "epoch": 0.07, "learning_rate": 5.517717141920473e-06, "loss": 0.8881, "step": 19787 }, { "epoch": 0.07, "learning_rate": 5.5167536293071855e-06, "loss": 0.8951, "step": 19788 }, { "epoch": 0.07, "learning_rate": 5.515790168781503e-06, "loss": 0.8668, "step": 19789 }, { "epoch": 0.07, "learning_rate": 5.514826760354625e-06, "loss": 0.8766, "step": 19790 }, { "epoch": 0.07, "learning_rate": 5.513863404037743e-06, "loss": 0.8581, "step": 19791 }, { "epoch": 0.07, "learning_rate": 5.512900099842051e-06, "loss": 0.8356, "step": 19792 }, { "epoch": 0.07, "learning_rate": 5.511936847778739e-06, "loss": 0.8543, "step": 19793 }, { "epoch": 0.07, "learning_rate": 5.510973647859002e-06, "loss": 0.8638, "step": 19794 }, { "epoch": 0.07, "learning_rate": 5.510010500094029e-06, "loss": 0.8536, "step": 19795 }, { "epoch": 0.07, "learning_rate": 5.5090474044950074e-06, "loss": 0.8744, "step": 19796 }, { "epoch": 0.07, "learning_rate": 5.50808436107313e-06, "loss": 0.9501, "step": 19797 }, { "epoch": 0.07, "learning_rate": 5.507121369839581e-06, "loss": 0.812, "step": 19798 }, { "epoch": 0.07, "learning_rate": 5.506158430805555e-06, "loss": 0.849, "step": 19799 }, { "epoch": 0.07, "learning_rate": 5.505195543982235e-06, "loss": 0.8888, "step": 19800 }, { "epoch": 0.07, "learning_rate": 5.504232709380811e-06, "loss": 0.8702, "step": 19801 }, { "epoch": 0.07, "learning_rate": 5.503269927012472e-06, "loss": 0.8649, "step": 19802 }, { "epoch": 0.07, "learning_rate": 5.502307196888392e-06, "loss": 0.8692, "step": 19803 }, { "epoch": 0.07, "learning_rate": 5.501344519019769e-06, "loss": 0.9069, "step": 19804 }, { "epoch": 0.07, "learning_rate": 5.500381893417787e-06, "loss": 0.8839, "step": 19805 }, { "epoch": 0.07, "learning_rate": 5.499419320093624e-06, "loss": 0.8821, "step": 19806 }, { "epoch": 0.07, "learning_rate": 5.4984567990584635e-06, "loss": 0.8667, "step": 19807 }, { "epoch": 0.07, "learning_rate": 5.497494330323492e-06, "loss": 0.8872, "step": 19808 }, { "epoch": 0.07, "learning_rate": 5.4965319138998885e-06, "loss": 0.8648, "step": 19809 }, { "epoch": 0.07, "learning_rate": 5.495569549798838e-06, "loss": 0.8624, "step": 19810 }, { "epoch": 0.07, "learning_rate": 5.49460723803152e-06, "loss": 0.8637, "step": 19811 }, { "epoch": 0.07, "learning_rate": 5.493644978609116e-06, "loss": 0.8828, "step": 19812 }, { "epoch": 0.07, "learning_rate": 5.4926827715428035e-06, "loss": 0.8407, "step": 19813 }, { "epoch": 0.07, "learning_rate": 5.491720616843762e-06, "loss": 0.8558, "step": 19814 }, { "epoch": 0.07, "learning_rate": 5.490758514523171e-06, "loss": 0.8578, "step": 19815 }, { "epoch": 0.07, "learning_rate": 5.489796464592208e-06, "loss": 0.8877, "step": 19816 }, { "epoch": 0.07, "learning_rate": 5.4888344670620516e-06, "loss": 0.8497, "step": 19817 }, { "epoch": 0.07, "learning_rate": 5.487872521943878e-06, "loss": 0.8703, "step": 19818 }, { "epoch": 0.07, "learning_rate": 5.486910629248865e-06, "loss": 0.8473, "step": 19819 }, { "epoch": 0.07, "learning_rate": 5.485948788988183e-06, "loss": 0.8623, "step": 19820 }, { "epoch": 0.07, "learning_rate": 5.484987001173011e-06, "loss": 0.8619, "step": 19821 }, { "epoch": 0.07, "learning_rate": 5.484025265814518e-06, "loss": 0.8311, "step": 19822 }, { "epoch": 0.07, "learning_rate": 5.48306358292389e-06, "loss": 0.8532, "step": 19823 }, { "epoch": 0.07, "learning_rate": 5.482101952512289e-06, "loss": 0.8636, "step": 19824 }, { "epoch": 0.07, "learning_rate": 5.481140374590889e-06, "loss": 0.8843, "step": 19825 }, { "epoch": 0.07, "learning_rate": 5.4801788491708655e-06, "loss": 0.8661, "step": 19826 }, { "epoch": 0.07, "learning_rate": 5.47921737626339e-06, "loss": 0.8179, "step": 19827 }, { "epoch": 0.07, "learning_rate": 5.4782559558796235e-06, "loss": 0.845, "step": 19828 }, { "epoch": 0.07, "learning_rate": 5.4772945880307485e-06, "loss": 0.8962, "step": 19829 }, { "epoch": 0.07, "learning_rate": 5.476333272727934e-06, "loss": 0.8954, "step": 19830 }, { "epoch": 0.07, "learning_rate": 5.4753720099823385e-06, "loss": 0.8774, "step": 19831 }, { "epoch": 0.07, "learning_rate": 5.4744107998051365e-06, "loss": 0.8687, "step": 19832 }, { "epoch": 0.07, "learning_rate": 5.473449642207495e-06, "loss": 0.8431, "step": 19833 }, { "epoch": 0.07, "learning_rate": 5.472488537200583e-06, "loss": 0.8334, "step": 19834 }, { "epoch": 0.07, "learning_rate": 5.471527484795565e-06, "loss": 0.8643, "step": 19835 }, { "epoch": 0.07, "learning_rate": 5.470566485003606e-06, "loss": 0.8444, "step": 19836 }, { "epoch": 0.07, "learning_rate": 5.469605537835875e-06, "loss": 0.8785, "step": 19837 }, { "epoch": 0.07, "learning_rate": 5.468644643303532e-06, "loss": 0.876, "step": 19838 }, { "epoch": 0.07, "learning_rate": 5.467683801417739e-06, "loss": 0.8658, "step": 19839 }, { "epoch": 0.07, "learning_rate": 5.466723012189666e-06, "loss": 0.9309, "step": 19840 }, { "epoch": 0.07, "learning_rate": 5.465762275630471e-06, "loss": 0.8512, "step": 19841 }, { "epoch": 0.07, "learning_rate": 5.4648015917513175e-06, "loss": 0.8583, "step": 19842 }, { "epoch": 0.07, "learning_rate": 5.463840960563367e-06, "loss": 0.8411, "step": 19843 }, { "epoch": 0.07, "learning_rate": 5.462880382077786e-06, "loss": 0.8771, "step": 19844 }, { "epoch": 0.07, "learning_rate": 5.461919856305724e-06, "loss": 0.8591, "step": 19845 }, { "epoch": 0.07, "learning_rate": 5.460959383258342e-06, "loss": 0.8945, "step": 19846 }, { "epoch": 0.07, "learning_rate": 5.459998962946809e-06, "loss": 0.934, "step": 19847 }, { "epoch": 0.07, "learning_rate": 5.459038595382279e-06, "loss": 0.9165, "step": 19848 }, { "epoch": 0.07, "learning_rate": 5.458078280575905e-06, "loss": 0.8761, "step": 19849 }, { "epoch": 0.07, "learning_rate": 5.457118018538847e-06, "loss": 0.8959, "step": 19850 }, { "epoch": 0.07, "learning_rate": 5.456157809282268e-06, "loss": 0.8616, "step": 19851 }, { "epoch": 0.07, "learning_rate": 5.4551976528173074e-06, "loss": 0.8625, "step": 19852 }, { "epoch": 0.07, "learning_rate": 5.454237549155137e-06, "loss": 0.8687, "step": 19853 }, { "epoch": 0.07, "learning_rate": 5.453277498306906e-06, "loss": 0.8782, "step": 19854 }, { "epoch": 0.07, "learning_rate": 5.45231750028377e-06, "loss": 0.8106, "step": 19855 }, { "epoch": 0.07, "learning_rate": 5.451357555096879e-06, "loss": 0.8849, "step": 19856 }, { "epoch": 0.07, "learning_rate": 5.450397662757387e-06, "loss": 0.8787, "step": 19857 }, { "epoch": 0.07, "learning_rate": 5.4494378232764465e-06, "loss": 0.8878, "step": 19858 }, { "epoch": 0.07, "learning_rate": 5.448478036665211e-06, "loss": 0.8724, "step": 19859 }, { "epoch": 0.07, "learning_rate": 5.44751830293483e-06, "loss": 0.8335, "step": 19860 }, { "epoch": 0.07, "learning_rate": 5.446558622096454e-06, "loss": 0.8412, "step": 19861 }, { "epoch": 0.07, "learning_rate": 5.4455989941612365e-06, "loss": 0.8136, "step": 19862 }, { "epoch": 0.07, "learning_rate": 5.44463941914032e-06, "loss": 0.8406, "step": 19863 }, { "epoch": 0.07, "learning_rate": 5.443679897044856e-06, "loss": 0.8411, "step": 19864 }, { "epoch": 0.07, "learning_rate": 5.442720427885992e-06, "loss": 0.894, "step": 19865 }, { "epoch": 0.07, "learning_rate": 5.441761011674877e-06, "loss": 0.8791, "step": 19866 }, { "epoch": 0.07, "learning_rate": 5.440801648422658e-06, "loss": 0.8223, "step": 19867 }, { "epoch": 0.07, "learning_rate": 5.43984233814048e-06, "loss": 0.9015, "step": 19868 }, { "epoch": 0.07, "learning_rate": 5.438883080839492e-06, "loss": 0.8558, "step": 19869 }, { "epoch": 0.07, "learning_rate": 5.437923876530833e-06, "loss": 0.8855, "step": 19870 }, { "epoch": 0.07, "learning_rate": 5.436964725225644e-06, "loss": 0.8763, "step": 19871 }, { "epoch": 0.07, "learning_rate": 5.43600562693508e-06, "loss": 0.8673, "step": 19872 }, { "epoch": 0.07, "learning_rate": 5.4350465816702825e-06, "loss": 0.8741, "step": 19873 }, { "epoch": 0.07, "learning_rate": 5.434087589442386e-06, "loss": 0.8164, "step": 19874 }, { "epoch": 0.07, "learning_rate": 5.433128650262535e-06, "loss": 0.856, "step": 19875 }, { "epoch": 0.07, "learning_rate": 5.432169764141875e-06, "loss": 0.9128, "step": 19876 }, { "epoch": 0.07, "learning_rate": 5.431210931091543e-06, "loss": 0.8701, "step": 19877 }, { "epoch": 0.07, "learning_rate": 5.430252151122679e-06, "loss": 0.8288, "step": 19878 }, { "epoch": 0.07, "learning_rate": 5.4292934242464245e-06, "loss": 0.8616, "step": 19879 }, { "epoch": 0.07, "learning_rate": 5.42833475047392e-06, "loss": 0.909, "step": 19880 }, { "epoch": 0.07, "learning_rate": 5.427376129816297e-06, "loss": 0.8716, "step": 19881 }, { "epoch": 0.07, "learning_rate": 5.426417562284696e-06, "loss": 0.9001, "step": 19882 }, { "epoch": 0.07, "learning_rate": 5.425459047890255e-06, "loss": 0.8763, "step": 19883 }, { "epoch": 0.07, "learning_rate": 5.424500586644111e-06, "loss": 0.8292, "step": 19884 }, { "epoch": 0.07, "learning_rate": 5.423542178557396e-06, "loss": 0.863, "step": 19885 }, { "epoch": 0.07, "learning_rate": 5.422583823641249e-06, "loss": 0.8792, "step": 19886 }, { "epoch": 0.07, "learning_rate": 5.421625521906806e-06, "loss": 0.8513, "step": 19887 }, { "epoch": 0.07, "learning_rate": 5.420667273365195e-06, "loss": 0.8676, "step": 19888 }, { "epoch": 0.07, "learning_rate": 5.419709078027548e-06, "loss": 0.853, "step": 19889 }, { "epoch": 0.07, "learning_rate": 5.418750935905008e-06, "loss": 0.9046, "step": 19890 }, { "epoch": 0.07, "learning_rate": 5.417792847008698e-06, "loss": 0.8362, "step": 19891 }, { "epoch": 0.07, "learning_rate": 5.416834811349752e-06, "loss": 0.8466, "step": 19892 }, { "epoch": 0.07, "learning_rate": 5.4158768289393006e-06, "loss": 0.9003, "step": 19893 }, { "epoch": 0.07, "learning_rate": 5.414918899788477e-06, "loss": 0.8459, "step": 19894 }, { "epoch": 0.07, "learning_rate": 5.4139610239084e-06, "loss": 0.8863, "step": 19895 }, { "epoch": 0.07, "learning_rate": 5.413003201310212e-06, "loss": 0.8535, "step": 19896 }, { "epoch": 0.07, "learning_rate": 5.412045432005038e-06, "loss": 0.863, "step": 19897 }, { "epoch": 0.07, "learning_rate": 5.411087716003998e-06, "loss": 0.8859, "step": 19898 }, { "epoch": 0.07, "learning_rate": 5.410130053318223e-06, "loss": 0.8609, "step": 19899 }, { "epoch": 0.07, "learning_rate": 5.409172443958844e-06, "loss": 0.9009, "step": 19900 }, { "epoch": 0.07, "learning_rate": 5.40821488793698e-06, "loss": 0.8267, "step": 19901 }, { "epoch": 0.07, "learning_rate": 5.40725738526376e-06, "loss": 0.8753, "step": 19902 }, { "epoch": 0.07, "learning_rate": 5.406299935950309e-06, "loss": 0.8601, "step": 19903 }, { "epoch": 0.07, "learning_rate": 5.405342540007753e-06, "loss": 0.8606, "step": 19904 }, { "epoch": 0.07, "learning_rate": 5.404385197447209e-06, "loss": 0.8526, "step": 19905 }, { "epoch": 0.07, "learning_rate": 5.403427908279801e-06, "loss": 0.8823, "step": 19906 }, { "epoch": 0.07, "learning_rate": 5.4024706725166535e-06, "loss": 0.8919, "step": 19907 }, { "epoch": 0.07, "learning_rate": 5.401513490168887e-06, "loss": 0.8575, "step": 19908 }, { "epoch": 0.07, "learning_rate": 5.400556361247622e-06, "loss": 0.8143, "step": 19909 }, { "epoch": 0.07, "learning_rate": 5.399599285763981e-06, "loss": 0.9051, "step": 19910 }, { "epoch": 0.07, "learning_rate": 5.3986422637290836e-06, "loss": 0.8967, "step": 19911 }, { "epoch": 0.07, "learning_rate": 5.3976852951540445e-06, "loss": 0.8787, "step": 19912 }, { "epoch": 0.07, "learning_rate": 5.396728380049985e-06, "loss": 0.8693, "step": 19913 }, { "epoch": 0.07, "learning_rate": 5.395771518428017e-06, "loss": 0.8987, "step": 19914 }, { "epoch": 0.07, "learning_rate": 5.39481471029927e-06, "loss": 0.8623, "step": 19915 }, { "epoch": 0.07, "learning_rate": 5.393857955674851e-06, "loss": 0.896, "step": 19916 }, { "epoch": 0.07, "learning_rate": 5.3929012545658775e-06, "loss": 0.907, "step": 19917 }, { "epoch": 0.07, "learning_rate": 5.391944606983466e-06, "loss": 0.8936, "step": 19918 }, { "epoch": 0.07, "learning_rate": 5.390988012938729e-06, "loss": 0.8076, "step": 19919 }, { "epoch": 0.07, "learning_rate": 5.390031472442783e-06, "loss": 0.8912, "step": 19920 }, { "epoch": 0.07, "learning_rate": 5.38907498550674e-06, "loss": 0.8727, "step": 19921 }, { "epoch": 0.07, "learning_rate": 5.388118552141718e-06, "loss": 0.8803, "step": 19922 }, { "epoch": 0.07, "learning_rate": 5.3871621723588195e-06, "loss": 0.8744, "step": 19923 }, { "epoch": 0.07, "learning_rate": 5.3862058461691605e-06, "loss": 0.8355, "step": 19924 }, { "epoch": 0.07, "learning_rate": 5.385249573583852e-06, "loss": 0.8668, "step": 19925 }, { "epoch": 0.07, "learning_rate": 5.384293354614003e-06, "loss": 0.8625, "step": 19926 }, { "epoch": 0.07, "learning_rate": 5.383337189270727e-06, "loss": 0.8687, "step": 19927 }, { "epoch": 0.07, "learning_rate": 5.382381077565129e-06, "loss": 0.8407, "step": 19928 }, { "epoch": 0.07, "learning_rate": 5.3814250195083215e-06, "loss": 0.8758, "step": 19929 }, { "epoch": 0.07, "learning_rate": 5.380469015111407e-06, "loss": 0.8594, "step": 19930 }, { "epoch": 0.07, "learning_rate": 5.3795130643854956e-06, "loss": 0.8344, "step": 19931 }, { "epoch": 0.07, "learning_rate": 5.378557167341691e-06, "loss": 0.8918, "step": 19932 }, { "epoch": 0.07, "learning_rate": 5.377601323991104e-06, "loss": 0.8648, "step": 19933 }, { "epoch": 0.07, "learning_rate": 5.376645534344835e-06, "loss": 0.8262, "step": 19934 }, { "epoch": 0.07, "learning_rate": 5.375689798413993e-06, "loss": 0.8442, "step": 19935 }, { "epoch": 0.07, "learning_rate": 5.374734116209683e-06, "loss": 0.8645, "step": 19936 }, { "epoch": 0.07, "learning_rate": 5.373778487743001e-06, "loss": 0.8889, "step": 19937 }, { "epoch": 0.07, "learning_rate": 5.372822913025051e-06, "loss": 0.8658, "step": 19938 }, { "epoch": 0.07, "learning_rate": 5.371867392066941e-06, "loss": 0.8694, "step": 19939 }, { "epoch": 0.07, "learning_rate": 5.3709119248797735e-06, "loss": 0.8518, "step": 19940 }, { "epoch": 0.07, "learning_rate": 5.369956511474642e-06, "loss": 0.8727, "step": 19941 }, { "epoch": 0.07, "learning_rate": 5.369001151862651e-06, "loss": 0.878, "step": 19942 }, { "epoch": 0.07, "learning_rate": 5.368045846054902e-06, "loss": 0.8385, "step": 19943 }, { "epoch": 0.07, "learning_rate": 5.367090594062485e-06, "loss": 0.8867, "step": 19944 }, { "epoch": 0.07, "learning_rate": 5.366135395896508e-06, "loss": 0.8786, "step": 19945 }, { "epoch": 0.07, "learning_rate": 5.365180251568065e-06, "loss": 0.8765, "step": 19946 }, { "epoch": 0.07, "learning_rate": 5.364225161088259e-06, "loss": 0.8738, "step": 19947 }, { "epoch": 0.07, "learning_rate": 5.3632701244681776e-06, "loss": 0.8996, "step": 19948 }, { "epoch": 0.07, "learning_rate": 5.362315141718919e-06, "loss": 0.8609, "step": 19949 }, { "epoch": 0.07, "learning_rate": 5.361360212851581e-06, "loss": 0.8709, "step": 19950 }, { "epoch": 0.07, "learning_rate": 5.360405337877256e-06, "loss": 0.852, "step": 19951 }, { "epoch": 0.07, "learning_rate": 5.359450516807038e-06, "loss": 0.8622, "step": 19952 }, { "epoch": 0.07, "learning_rate": 5.358495749652024e-06, "loss": 0.847, "step": 19953 }, { "epoch": 0.07, "learning_rate": 5.357541036423306e-06, "loss": 0.8707, "step": 19954 }, { "epoch": 0.07, "learning_rate": 5.356586377131972e-06, "loss": 0.868, "step": 19955 }, { "epoch": 0.07, "learning_rate": 5.355631771789114e-06, "loss": 0.8921, "step": 19956 }, { "epoch": 0.07, "learning_rate": 5.354677220405826e-06, "loss": 0.8659, "step": 19957 }, { "epoch": 0.07, "learning_rate": 5.353722722993196e-06, "loss": 0.9093, "step": 19958 }, { "epoch": 0.07, "learning_rate": 5.352768279562315e-06, "loss": 0.8641, "step": 19959 }, { "epoch": 0.07, "learning_rate": 5.351813890124271e-06, "loss": 0.8408, "step": 19960 }, { "epoch": 0.07, "learning_rate": 5.350859554690158e-06, "loss": 0.8487, "step": 19961 }, { "epoch": 0.07, "learning_rate": 5.34990527327105e-06, "loss": 0.8862, "step": 19962 }, { "epoch": 0.07, "learning_rate": 5.348951045878047e-06, "loss": 0.8227, "step": 19963 }, { "epoch": 0.07, "learning_rate": 5.34799687252223e-06, "loss": 0.8379, "step": 19964 }, { "epoch": 0.07, "learning_rate": 5.347042753214691e-06, "loss": 0.8739, "step": 19965 }, { "epoch": 0.07, "learning_rate": 5.346088687966507e-06, "loss": 0.8312, "step": 19966 }, { "epoch": 0.07, "learning_rate": 5.3451346767887646e-06, "loss": 0.8544, "step": 19967 }, { "epoch": 0.07, "learning_rate": 5.344180719692549e-06, "loss": 0.8817, "step": 19968 }, { "epoch": 0.07, "learning_rate": 5.343226816688944e-06, "loss": 0.8587, "step": 19969 }, { "epoch": 0.07, "learning_rate": 5.342272967789033e-06, "loss": 0.8852, "step": 19970 }, { "epoch": 0.07, "learning_rate": 5.341319173003895e-06, "loss": 0.8533, "step": 19971 }, { "epoch": 0.07, "learning_rate": 5.3403654323446184e-06, "loss": 0.872, "step": 19972 }, { "epoch": 0.07, "learning_rate": 5.339411745822274e-06, "loss": 0.8395, "step": 19973 }, { "epoch": 0.07, "learning_rate": 5.338458113447949e-06, "loss": 0.8382, "step": 19974 }, { "epoch": 0.07, "learning_rate": 5.337504535232719e-06, "loss": 0.88, "step": 19975 }, { "epoch": 0.07, "learning_rate": 5.336551011187666e-06, "loss": 0.9072, "step": 19976 }, { "epoch": 0.07, "learning_rate": 5.335597541323867e-06, "loss": 0.8657, "step": 19977 }, { "epoch": 0.07, "learning_rate": 5.3346441256524005e-06, "loss": 0.8462, "step": 19978 }, { "epoch": 0.07, "learning_rate": 5.333690764184347e-06, "loss": 0.8321, "step": 19979 }, { "epoch": 0.07, "learning_rate": 5.332737456930774e-06, "loss": 0.8442, "step": 19980 }, { "epoch": 0.07, "learning_rate": 5.331784203902759e-06, "loss": 0.8798, "step": 19981 }, { "epoch": 0.07, "learning_rate": 5.330831005111388e-06, "loss": 0.8508, "step": 19982 }, { "epoch": 0.07, "learning_rate": 5.329877860567724e-06, "loss": 0.8492, "step": 19983 }, { "epoch": 0.07, "learning_rate": 5.3289247702828436e-06, "loss": 0.8978, "step": 19984 }, { "epoch": 0.07, "learning_rate": 5.327971734267823e-06, "loss": 0.8658, "step": 19985 }, { "epoch": 0.07, "learning_rate": 5.327018752533737e-06, "loss": 0.8414, "step": 19986 }, { "epoch": 0.07, "learning_rate": 5.326065825091645e-06, "loss": 0.857, "step": 19987 }, { "epoch": 0.07, "learning_rate": 5.325112951952631e-06, "loss": 0.8934, "step": 19988 }, { "epoch": 0.07, "learning_rate": 5.324160133127767e-06, "loss": 0.9113, "step": 19989 }, { "epoch": 0.07, "learning_rate": 5.323207368628112e-06, "loss": 0.819, "step": 19990 }, { "epoch": 0.07, "learning_rate": 5.322254658464743e-06, "loss": 0.8276, "step": 19991 }, { "epoch": 0.07, "learning_rate": 5.321302002648727e-06, "loss": 0.8346, "step": 19992 }, { "epoch": 0.07, "learning_rate": 5.320349401191132e-06, "loss": 0.8609, "step": 19993 }, { "epoch": 0.07, "learning_rate": 5.319396854103026e-06, "loss": 0.8846, "step": 19994 }, { "epoch": 0.07, "learning_rate": 5.318444361395476e-06, "loss": 0.8193, "step": 19995 }, { "epoch": 0.07, "learning_rate": 5.317491923079554e-06, "loss": 0.8554, "step": 19996 }, { "epoch": 0.07, "learning_rate": 5.316539539166314e-06, "loss": 0.8801, "step": 19997 }, { "epoch": 0.07, "learning_rate": 5.3155872096668284e-06, "loss": 0.8741, "step": 19998 }, { "epoch": 0.07, "learning_rate": 5.31463493459216e-06, "loss": 0.8352, "step": 19999 }, { "epoch": 0.07, "learning_rate": 5.313682713953372e-06, "loss": 0.8831, "step": 20000 }, { "epoch": 0.07, "learning_rate": 5.31273054776153e-06, "loss": 0.8536, "step": 20001 }, { "epoch": 0.07, "learning_rate": 5.311778436027694e-06, "loss": 0.8606, "step": 20002 }, { "epoch": 0.07, "learning_rate": 5.3108263787629325e-06, "loss": 0.8648, "step": 20003 }, { "epoch": 0.08, "learning_rate": 5.309874375978296e-06, "loss": 0.9, "step": 20004 }, { "epoch": 0.08, "learning_rate": 5.308922427684851e-06, "loss": 0.8452, "step": 20005 }, { "epoch": 0.08, "learning_rate": 5.3079705338936515e-06, "loss": 0.8577, "step": 20006 }, { "epoch": 0.08, "learning_rate": 5.3070186946157705e-06, "loss": 0.8496, "step": 20007 }, { "epoch": 0.08, "learning_rate": 5.306066909862255e-06, "loss": 0.8415, "step": 20008 }, { "epoch": 0.08, "learning_rate": 5.305115179644167e-06, "loss": 0.8621, "step": 20009 }, { "epoch": 0.08, "learning_rate": 5.3041635039725635e-06, "loss": 0.8651, "step": 20010 }, { "epoch": 0.08, "learning_rate": 5.303211882858501e-06, "loss": 0.8887, "step": 20011 }, { "epoch": 0.08, "learning_rate": 5.302260316313037e-06, "loss": 0.8613, "step": 20012 }, { "epoch": 0.08, "learning_rate": 5.301308804347225e-06, "loss": 0.8683, "step": 20013 }, { "epoch": 0.08, "learning_rate": 5.300357346972125e-06, "loss": 0.8689, "step": 20014 }, { "epoch": 0.08, "learning_rate": 5.299405944198783e-06, "loss": 0.9084, "step": 20015 }, { "epoch": 0.08, "learning_rate": 5.298454596038258e-06, "loss": 0.8748, "step": 20016 }, { "epoch": 0.08, "learning_rate": 5.297503302501603e-06, "loss": 0.8578, "step": 20017 }, { "epoch": 0.08, "learning_rate": 5.2965520635998676e-06, "loss": 0.8594, "step": 20018 }, { "epoch": 0.08, "learning_rate": 5.295600879344107e-06, "loss": 0.9094, "step": 20019 }, { "epoch": 0.08, "learning_rate": 5.294649749745369e-06, "loss": 0.8636, "step": 20020 }, { "epoch": 0.08, "learning_rate": 5.293698674814711e-06, "loss": 0.8389, "step": 20021 }, { "epoch": 0.08, "learning_rate": 5.292747654563173e-06, "loss": 0.8666, "step": 20022 }, { "epoch": 0.08, "learning_rate": 5.291796689001809e-06, "loss": 0.8364, "step": 20023 }, { "epoch": 0.08, "learning_rate": 5.290845778141665e-06, "loss": 0.8623, "step": 20024 }, { "epoch": 0.08, "learning_rate": 5.289894921993798e-06, "loss": 0.8646, "step": 20025 }, { "epoch": 0.08, "learning_rate": 5.288944120569244e-06, "loss": 0.8799, "step": 20026 }, { "epoch": 0.08, "learning_rate": 5.287993373879057e-06, "loss": 0.8595, "step": 20027 }, { "epoch": 0.08, "learning_rate": 5.287042681934282e-06, "loss": 0.8905, "step": 20028 }, { "epoch": 0.08, "learning_rate": 5.2860920447459595e-06, "loss": 0.8506, "step": 20029 }, { "epoch": 0.08, "learning_rate": 5.285141462325134e-06, "loss": 0.9126, "step": 20030 }, { "epoch": 0.08, "learning_rate": 5.2841909346828565e-06, "loss": 0.8702, "step": 20031 }, { "epoch": 0.08, "learning_rate": 5.2832404618301725e-06, "loss": 0.8347, "step": 20032 }, { "epoch": 0.08, "learning_rate": 5.282290043778116e-06, "loss": 0.8708, "step": 20033 }, { "epoch": 0.08, "learning_rate": 5.28133968053773e-06, "loss": 0.8823, "step": 20034 }, { "epoch": 0.08, "learning_rate": 5.280389372120061e-06, "loss": 0.8593, "step": 20035 }, { "epoch": 0.08, "learning_rate": 5.279439118536149e-06, "loss": 0.8756, "step": 20036 }, { "epoch": 0.08, "learning_rate": 5.278488919797031e-06, "loss": 0.9066, "step": 20037 }, { "epoch": 0.08, "learning_rate": 5.27753877591375e-06, "loss": 0.9093, "step": 20038 }, { "epoch": 0.08, "learning_rate": 5.276588686897347e-06, "loss": 0.8305, "step": 20039 }, { "epoch": 0.08, "learning_rate": 5.275638652758854e-06, "loss": 0.8618, "step": 20040 }, { "epoch": 0.08, "learning_rate": 5.274688673509312e-06, "loss": 0.9024, "step": 20041 }, { "epoch": 0.08, "learning_rate": 5.273738749159758e-06, "loss": 0.8916, "step": 20042 }, { "epoch": 0.08, "learning_rate": 5.272788879721229e-06, "loss": 0.8762, "step": 20043 }, { "epoch": 0.08, "learning_rate": 5.27183906520476e-06, "loss": 0.8877, "step": 20044 }, { "epoch": 0.08, "learning_rate": 5.270889305621387e-06, "loss": 0.8399, "step": 20045 }, { "epoch": 0.08, "learning_rate": 5.269939600982151e-06, "loss": 0.8807, "step": 20046 }, { "epoch": 0.08, "learning_rate": 5.268989951298071e-06, "loss": 0.8539, "step": 20047 }, { "epoch": 0.08, "learning_rate": 5.268040356580191e-06, "loss": 0.8272, "step": 20048 }, { "epoch": 0.08, "learning_rate": 5.26709081683954e-06, "loss": 0.8745, "step": 20049 }, { "epoch": 0.08, "learning_rate": 5.266141332087151e-06, "loss": 0.91, "step": 20050 }, { "epoch": 0.08, "learning_rate": 5.265191902334057e-06, "loss": 0.8459, "step": 20051 }, { "epoch": 0.08, "learning_rate": 5.264242527591287e-06, "loss": 0.891, "step": 20052 }, { "epoch": 0.08, "learning_rate": 5.263293207869875e-06, "loss": 0.8479, "step": 20053 }, { "epoch": 0.08, "learning_rate": 5.262343943180838e-06, "loss": 0.9117, "step": 20054 }, { "epoch": 0.08, "learning_rate": 5.2613947335352185e-06, "loss": 0.8278, "step": 20055 }, { "epoch": 0.08, "learning_rate": 5.26044557894404e-06, "loss": 0.8735, "step": 20056 }, { "epoch": 0.08, "learning_rate": 5.259496479418334e-06, "loss": 0.8606, "step": 20057 }, { "epoch": 0.08, "learning_rate": 5.2585474349691205e-06, "loss": 0.8939, "step": 20058 }, { "epoch": 0.08, "learning_rate": 5.257598445607427e-06, "loss": 0.8318, "step": 20059 }, { "epoch": 0.08, "learning_rate": 5.25664951134428e-06, "loss": 0.8775, "step": 20060 }, { "epoch": 0.08, "learning_rate": 5.255700632190708e-06, "loss": 0.8269, "step": 20061 }, { "epoch": 0.08, "learning_rate": 5.254751808157731e-06, "loss": 0.837, "step": 20062 }, { "epoch": 0.08, "learning_rate": 5.253803039256374e-06, "loss": 0.9156, "step": 20063 }, { "epoch": 0.08, "learning_rate": 5.252854325497665e-06, "loss": 0.8172, "step": 20064 }, { "epoch": 0.08, "learning_rate": 5.251905666892617e-06, "loss": 0.893, "step": 20065 }, { "epoch": 0.08, "learning_rate": 5.250957063452258e-06, "loss": 0.9052, "step": 20066 }, { "epoch": 0.08, "learning_rate": 5.250008515187605e-06, "loss": 0.8377, "step": 20067 }, { "epoch": 0.08, "learning_rate": 5.2490600221096835e-06, "loss": 0.8902, "step": 20068 }, { "epoch": 0.08, "learning_rate": 5.248111584229511e-06, "loss": 0.8522, "step": 20069 }, { "epoch": 0.08, "learning_rate": 5.247163201558105e-06, "loss": 0.822, "step": 20070 }, { "epoch": 0.08, "learning_rate": 5.246214874106491e-06, "loss": 0.8416, "step": 20071 }, { "epoch": 0.08, "learning_rate": 5.245266601885678e-06, "loss": 0.8461, "step": 20072 }, { "epoch": 0.08, "learning_rate": 5.244318384906682e-06, "loss": 0.8157, "step": 20073 }, { "epoch": 0.08, "learning_rate": 5.243370223180534e-06, "loss": 0.9228, "step": 20074 }, { "epoch": 0.08, "learning_rate": 5.242422116718236e-06, "loss": 0.8762, "step": 20075 }, { "epoch": 0.08, "learning_rate": 5.241474065530808e-06, "loss": 0.8748, "step": 20076 }, { "epoch": 0.08, "learning_rate": 5.240526069629265e-06, "loss": 0.8843, "step": 20077 }, { "epoch": 0.08, "learning_rate": 5.239578129024625e-06, "loss": 0.8405, "step": 20078 }, { "epoch": 0.08, "learning_rate": 5.238630243727889e-06, "loss": 0.8686, "step": 20079 }, { "epoch": 0.08, "learning_rate": 5.237682413750083e-06, "loss": 0.8647, "step": 20080 }, { "epoch": 0.08, "learning_rate": 5.236734639102217e-06, "loss": 0.8477, "step": 20081 }, { "epoch": 0.08, "learning_rate": 5.235786919795296e-06, "loss": 0.8566, "step": 20082 }, { "epoch": 0.08, "learning_rate": 5.234839255840335e-06, "loss": 0.835, "step": 20083 }, { "epoch": 0.08, "learning_rate": 5.233891647248344e-06, "loss": 0.8383, "step": 20084 }, { "epoch": 0.08, "learning_rate": 5.232944094030332e-06, "loss": 0.8839, "step": 20085 }, { "epoch": 0.08, "learning_rate": 5.231996596197309e-06, "loss": 0.8502, "step": 20086 }, { "epoch": 0.08, "learning_rate": 5.231049153760282e-06, "loss": 0.8685, "step": 20087 }, { "epoch": 0.08, "learning_rate": 5.230101766730264e-06, "loss": 0.868, "step": 20088 }, { "epoch": 0.08, "learning_rate": 5.229154435118253e-06, "loss": 0.8636, "step": 20089 }, { "epoch": 0.08, "learning_rate": 5.228207158935259e-06, "loss": 0.8788, "step": 20090 }, { "epoch": 0.08, "learning_rate": 5.227259938192289e-06, "loss": 0.8775, "step": 20091 }, { "epoch": 0.08, "learning_rate": 5.226312772900348e-06, "loss": 0.8767, "step": 20092 }, { "epoch": 0.08, "learning_rate": 5.2253656630704406e-06, "loss": 0.9117, "step": 20093 }, { "epoch": 0.08, "learning_rate": 5.2244186087135686e-06, "loss": 0.8842, "step": 20094 }, { "epoch": 0.08, "learning_rate": 5.22347160984074e-06, "loss": 0.8574, "step": 20095 }, { "epoch": 0.08, "learning_rate": 5.222524666462952e-06, "loss": 0.8769, "step": 20096 }, { "epoch": 0.08, "learning_rate": 5.2215777785912025e-06, "loss": 0.8785, "step": 20097 }, { "epoch": 0.08, "learning_rate": 5.2206309462365025e-06, "loss": 0.8781, "step": 20098 }, { "epoch": 0.08, "learning_rate": 5.219684169409852e-06, "loss": 0.8366, "step": 20099 }, { "epoch": 0.08, "learning_rate": 5.218737448122244e-06, "loss": 0.8472, "step": 20100 }, { "epoch": 0.08, "learning_rate": 5.217790782384682e-06, "loss": 0.8487, "step": 20101 }, { "epoch": 0.08, "learning_rate": 5.216844172208162e-06, "loss": 0.8813, "step": 20102 }, { "epoch": 0.08, "learning_rate": 5.215897617603684e-06, "loss": 0.8721, "step": 20103 }, { "epoch": 0.08, "learning_rate": 5.214951118582248e-06, "loss": 0.8908, "step": 20104 }, { "epoch": 0.08, "learning_rate": 5.214004675154845e-06, "loss": 0.8674, "step": 20105 }, { "epoch": 0.08, "learning_rate": 5.213058287332478e-06, "loss": 0.8897, "step": 20106 }, { "epoch": 0.08, "learning_rate": 5.212111955126136e-06, "loss": 0.9019, "step": 20107 }, { "epoch": 0.08, "learning_rate": 5.211165678546814e-06, "loss": 0.8471, "step": 20108 }, { "epoch": 0.08, "learning_rate": 5.21021945760551e-06, "loss": 0.8056, "step": 20109 }, { "epoch": 0.08, "learning_rate": 5.209273292313214e-06, "loss": 0.8707, "step": 20110 }, { "epoch": 0.08, "learning_rate": 5.208327182680921e-06, "loss": 0.8446, "step": 20111 }, { "epoch": 0.08, "learning_rate": 5.207381128719622e-06, "loss": 0.8673, "step": 20112 }, { "epoch": 0.08, "learning_rate": 5.206435130440314e-06, "loss": 0.8888, "step": 20113 }, { "epoch": 0.08, "learning_rate": 5.205489187853978e-06, "loss": 0.8632, "step": 20114 }, { "epoch": 0.08, "learning_rate": 5.204543300971609e-06, "loss": 0.8669, "step": 20115 }, { "epoch": 0.08, "learning_rate": 5.203597469804194e-06, "loss": 0.934, "step": 20116 }, { "epoch": 0.08, "learning_rate": 5.202651694362732e-06, "loss": 0.8868, "step": 20117 }, { "epoch": 0.08, "learning_rate": 5.2017059746582e-06, "loss": 0.8684, "step": 20118 }, { "epoch": 0.08, "learning_rate": 5.200760310701589e-06, "loss": 0.8888, "step": 20119 }, { "epoch": 0.08, "learning_rate": 5.199814702503891e-06, "loss": 0.8476, "step": 20120 }, { "epoch": 0.08, "learning_rate": 5.198869150076084e-06, "loss": 0.8541, "step": 20121 }, { "epoch": 0.08, "learning_rate": 5.197923653429155e-06, "loss": 0.8626, "step": 20122 }, { "epoch": 0.08, "learning_rate": 5.196978212574094e-06, "loss": 0.8452, "step": 20123 }, { "epoch": 0.08, "learning_rate": 5.196032827521889e-06, "loss": 0.8771, "step": 20124 }, { "epoch": 0.08, "learning_rate": 5.195087498283511e-06, "loss": 0.8575, "step": 20125 }, { "epoch": 0.08, "learning_rate": 5.194142224869951e-06, "loss": 0.8658, "step": 20126 }, { "epoch": 0.08, "learning_rate": 5.19319700729219e-06, "loss": 0.8668, "step": 20127 }, { "epoch": 0.08, "learning_rate": 5.192251845561212e-06, "loss": 0.8627, "step": 20128 }, { "epoch": 0.08, "learning_rate": 5.191306739687995e-06, "loss": 0.8364, "step": 20129 }, { "epoch": 0.08, "learning_rate": 5.190361689683521e-06, "loss": 0.8831, "step": 20130 }, { "epoch": 0.08, "learning_rate": 5.189416695558773e-06, "loss": 0.8636, "step": 20131 }, { "epoch": 0.08, "learning_rate": 5.188471757324722e-06, "loss": 0.8747, "step": 20132 }, { "epoch": 0.08, "learning_rate": 5.187526874992355e-06, "loss": 0.8855, "step": 20133 }, { "epoch": 0.08, "learning_rate": 5.186582048572644e-06, "loss": 0.8847, "step": 20134 }, { "epoch": 0.08, "learning_rate": 5.185637278076567e-06, "loss": 0.8205, "step": 20135 }, { "epoch": 0.08, "learning_rate": 5.184692563515104e-06, "loss": 0.8134, "step": 20136 }, { "epoch": 0.08, "learning_rate": 5.18374790489923e-06, "loss": 0.8321, "step": 20137 }, { "epoch": 0.08, "learning_rate": 5.182803302239923e-06, "loss": 0.8494, "step": 20138 }, { "epoch": 0.08, "learning_rate": 5.18185875554815e-06, "loss": 0.8475, "step": 20139 }, { "epoch": 0.08, "learning_rate": 5.180914264834887e-06, "loss": 0.8532, "step": 20140 }, { "epoch": 0.08, "learning_rate": 5.1799698301111125e-06, "loss": 0.865, "step": 20141 }, { "epoch": 0.08, "learning_rate": 5.1790254513877945e-06, "loss": 0.8819, "step": 20142 }, { "epoch": 0.08, "learning_rate": 5.178081128675907e-06, "loss": 0.8875, "step": 20143 }, { "epoch": 0.08, "learning_rate": 5.177136861986423e-06, "loss": 0.8669, "step": 20144 }, { "epoch": 0.08, "learning_rate": 5.176192651330313e-06, "loss": 0.8672, "step": 20145 }, { "epoch": 0.08, "learning_rate": 5.175248496718539e-06, "loss": 0.8485, "step": 20146 }, { "epoch": 0.08, "learning_rate": 5.1743043981620796e-06, "loss": 0.8651, "step": 20147 }, { "epoch": 0.08, "learning_rate": 5.173360355671901e-06, "loss": 0.8668, "step": 20148 }, { "epoch": 0.08, "learning_rate": 5.172416369258977e-06, "loss": 0.8649, "step": 20149 }, { "epoch": 0.08, "learning_rate": 5.171472438934264e-06, "loss": 0.8771, "step": 20150 }, { "epoch": 0.08, "learning_rate": 5.170528564708734e-06, "loss": 0.8235, "step": 20151 }, { "epoch": 0.08, "learning_rate": 5.169584746593355e-06, "loss": 0.8703, "step": 20152 }, { "epoch": 0.08, "learning_rate": 5.168640984599091e-06, "loss": 0.8641, "step": 20153 }, { "epoch": 0.08, "learning_rate": 5.167697278736905e-06, "loss": 0.8865, "step": 20154 }, { "epoch": 0.08, "learning_rate": 5.1667536290177645e-06, "loss": 0.8638, "step": 20155 }, { "epoch": 0.08, "learning_rate": 5.165810035452636e-06, "loss": 0.8347, "step": 20156 }, { "epoch": 0.08, "learning_rate": 5.164866498052473e-06, "loss": 0.8906, "step": 20157 }, { "epoch": 0.08, "learning_rate": 5.163923016828243e-06, "loss": 0.8869, "step": 20158 }, { "epoch": 0.08, "learning_rate": 5.162979591790909e-06, "loss": 0.8629, "step": 20159 }, { "epoch": 0.08, "learning_rate": 5.16203622295143e-06, "loss": 0.9047, "step": 20160 }, { "epoch": 0.08, "learning_rate": 5.161092910320767e-06, "loss": 0.865, "step": 20161 }, { "epoch": 0.08, "learning_rate": 5.160149653909878e-06, "loss": 0.9154, "step": 20162 }, { "epoch": 0.08, "learning_rate": 5.159206453729728e-06, "loss": 0.869, "step": 20163 }, { "epoch": 0.08, "learning_rate": 5.158263309791267e-06, "loss": 0.8964, "step": 20164 }, { "epoch": 0.08, "learning_rate": 5.157320222105454e-06, "loss": 0.8885, "step": 20165 }, { "epoch": 0.08, "learning_rate": 5.156377190683256e-06, "loss": 0.8719, "step": 20166 }, { "epoch": 0.08, "learning_rate": 5.155434215535617e-06, "loss": 0.8137, "step": 20167 }, { "epoch": 0.08, "learning_rate": 5.1544912966735e-06, "loss": 0.8977, "step": 20168 }, { "epoch": 0.08, "learning_rate": 5.153548434107855e-06, "loss": 0.8471, "step": 20169 }, { "epoch": 0.08, "learning_rate": 5.1526056278496424e-06, "loss": 0.8644, "step": 20170 }, { "epoch": 0.08, "learning_rate": 5.151662877909811e-06, "loss": 0.8757, "step": 20171 }, { "epoch": 0.08, "learning_rate": 5.150720184299316e-06, "loss": 0.8519, "step": 20172 }, { "epoch": 0.08, "learning_rate": 5.149777547029114e-06, "loss": 0.8884, "step": 20173 }, { "epoch": 0.08, "learning_rate": 5.14883496611015e-06, "loss": 0.844, "step": 20174 }, { "epoch": 0.08, "learning_rate": 5.147892441553376e-06, "loss": 0.87, "step": 20175 }, { "epoch": 0.08, "learning_rate": 5.146949973369745e-06, "loss": 0.8581, "step": 20176 }, { "epoch": 0.08, "learning_rate": 5.146007561570206e-06, "loss": 0.8942, "step": 20177 }, { "epoch": 0.08, "learning_rate": 5.145065206165708e-06, "loss": 0.8451, "step": 20178 }, { "epoch": 0.08, "learning_rate": 5.144122907167201e-06, "loss": 0.8573, "step": 20179 }, { "epoch": 0.08, "learning_rate": 5.143180664585634e-06, "loss": 0.8657, "step": 20180 }, { "epoch": 0.08, "learning_rate": 5.142238478431949e-06, "loss": 0.8665, "step": 20181 }, { "epoch": 0.08, "learning_rate": 5.141296348717096e-06, "loss": 0.8844, "step": 20182 }, { "epoch": 0.08, "learning_rate": 5.140354275452019e-06, "loss": 0.8082, "step": 20183 }, { "epoch": 0.08, "learning_rate": 5.139412258647667e-06, "loss": 0.8694, "step": 20184 }, { "epoch": 0.08, "learning_rate": 5.1384702983149815e-06, "loss": 0.8594, "step": 20185 }, { "epoch": 0.08, "learning_rate": 5.137528394464906e-06, "loss": 0.8714, "step": 20186 }, { "epoch": 0.08, "learning_rate": 5.136586547108391e-06, "loss": 0.8762, "step": 20187 }, { "epoch": 0.08, "learning_rate": 5.135644756256369e-06, "loss": 0.859, "step": 20188 }, { "epoch": 0.08, "learning_rate": 5.134703021919781e-06, "loss": 0.8946, "step": 20189 }, { "epoch": 0.08, "learning_rate": 5.1337613441095775e-06, "loss": 0.8817, "step": 20190 }, { "epoch": 0.08, "learning_rate": 5.132819722836701e-06, "loss": 0.8511, "step": 20191 }, { "epoch": 0.08, "learning_rate": 5.131878158112082e-06, "loss": 0.8355, "step": 20192 }, { "epoch": 0.08, "learning_rate": 5.130936649946663e-06, "loss": 0.8518, "step": 20193 }, { "epoch": 0.08, "learning_rate": 5.129995198351383e-06, "loss": 0.8879, "step": 20194 }, { "epoch": 0.08, "learning_rate": 5.129053803337181e-06, "loss": 0.8866, "step": 20195 }, { "epoch": 0.08, "learning_rate": 5.1281124649149925e-06, "loss": 0.8716, "step": 20196 }, { "epoch": 0.08, "learning_rate": 5.127171183095757e-06, "loss": 0.8579, "step": 20197 }, { "epoch": 0.08, "learning_rate": 5.126229957890413e-06, "loss": 0.8817, "step": 20198 }, { "epoch": 0.08, "learning_rate": 5.125288789309888e-06, "loss": 0.8549, "step": 20199 }, { "epoch": 0.08, "learning_rate": 5.124347677365121e-06, "loss": 0.8539, "step": 20200 }, { "epoch": 0.08, "learning_rate": 5.123406622067044e-06, "loss": 0.8338, "step": 20201 }, { "epoch": 0.08, "learning_rate": 5.122465623426595e-06, "loss": 0.8519, "step": 20202 }, { "epoch": 0.08, "learning_rate": 5.121524681454703e-06, "loss": 0.8866, "step": 20203 }, { "epoch": 0.08, "learning_rate": 5.1205837961623015e-06, "loss": 0.8989, "step": 20204 }, { "epoch": 0.08, "learning_rate": 5.119642967560326e-06, "loss": 0.8261, "step": 20205 }, { "epoch": 0.08, "learning_rate": 5.1187021956596975e-06, "loss": 0.869, "step": 20206 }, { "epoch": 0.08, "learning_rate": 5.117761480471353e-06, "loss": 0.8708, "step": 20207 }, { "epoch": 0.08, "learning_rate": 5.116820822006217e-06, "loss": 0.9134, "step": 20208 }, { "epoch": 0.08, "learning_rate": 5.115880220275229e-06, "loss": 0.8786, "step": 20209 }, { "epoch": 0.08, "learning_rate": 5.1149396752893075e-06, "loss": 0.8535, "step": 20210 }, { "epoch": 0.08, "learning_rate": 5.113999187059382e-06, "loss": 0.8542, "step": 20211 }, { "epoch": 0.08, "learning_rate": 5.113058755596383e-06, "loss": 0.8904, "step": 20212 }, { "epoch": 0.08, "learning_rate": 5.11211838091123e-06, "loss": 0.884, "step": 20213 }, { "epoch": 0.08, "learning_rate": 5.11117806301485e-06, "loss": 0.8912, "step": 20214 }, { "epoch": 0.08, "learning_rate": 5.110237801918174e-06, "loss": 0.8341, "step": 20215 }, { "epoch": 0.08, "learning_rate": 5.1092975976321255e-06, "loss": 0.8836, "step": 20216 }, { "epoch": 0.08, "learning_rate": 5.1083574501676216e-06, "loss": 0.8542, "step": 20217 }, { "epoch": 0.08, "learning_rate": 5.107417359535588e-06, "loss": 0.848, "step": 20218 }, { "epoch": 0.08, "learning_rate": 5.106477325746948e-06, "loss": 0.8213, "step": 20219 }, { "epoch": 0.08, "learning_rate": 5.105537348812621e-06, "loss": 0.8781, "step": 20220 }, { "epoch": 0.08, "learning_rate": 5.104597428743532e-06, "loss": 0.8526, "step": 20221 }, { "epoch": 0.08, "learning_rate": 5.103657565550597e-06, "loss": 0.8838, "step": 20222 }, { "epoch": 0.08, "learning_rate": 5.102717759244743e-06, "loss": 0.8845, "step": 20223 }, { "epoch": 0.08, "learning_rate": 5.101778009836877e-06, "loss": 0.8669, "step": 20224 }, { "epoch": 0.08, "learning_rate": 5.100838317337926e-06, "loss": 0.8786, "step": 20225 }, { "epoch": 0.08, "learning_rate": 5.099898681758805e-06, "loss": 0.8818, "step": 20226 }, { "epoch": 0.08, "learning_rate": 5.0989591031104304e-06, "loss": 0.9179, "step": 20227 }, { "epoch": 0.08, "learning_rate": 5.0980195814037195e-06, "loss": 0.8487, "step": 20228 }, { "epoch": 0.08, "learning_rate": 5.097080116649587e-06, "loss": 0.8891, "step": 20229 }, { "epoch": 0.08, "learning_rate": 5.096140708858954e-06, "loss": 0.8793, "step": 20230 }, { "epoch": 0.08, "learning_rate": 5.095201358042725e-06, "loss": 0.8921, "step": 20231 }, { "epoch": 0.08, "learning_rate": 5.0942620642118126e-06, "loss": 0.8641, "step": 20232 }, { "epoch": 0.08, "learning_rate": 5.093322827377145e-06, "loss": 0.8473, "step": 20233 }, { "epoch": 0.08, "learning_rate": 5.092383647549619e-06, "loss": 0.8723, "step": 20234 }, { "epoch": 0.08, "learning_rate": 5.091444524740154e-06, "loss": 0.8454, "step": 20235 }, { "epoch": 0.08, "learning_rate": 5.090505458959657e-06, "loss": 0.8549, "step": 20236 }, { "epoch": 0.08, "learning_rate": 5.089566450219046e-06, "loss": 0.8305, "step": 20237 }, { "epoch": 0.08, "learning_rate": 5.088627498529215e-06, "loss": 0.8421, "step": 20238 }, { "epoch": 0.08, "learning_rate": 5.087688603901089e-06, "loss": 0.8477, "step": 20239 }, { "epoch": 0.08, "learning_rate": 5.0867497663455685e-06, "loss": 0.8569, "step": 20240 }, { "epoch": 0.08, "learning_rate": 5.085810985873567e-06, "loss": 0.842, "step": 20241 }, { "epoch": 0.08, "learning_rate": 5.084872262495986e-06, "loss": 0.8915, "step": 20242 }, { "epoch": 0.08, "learning_rate": 5.083933596223731e-06, "loss": 0.8544, "step": 20243 }, { "epoch": 0.08, "learning_rate": 5.082994987067711e-06, "loss": 0.8774, "step": 20244 }, { "epoch": 0.08, "learning_rate": 5.082056435038829e-06, "loss": 0.8548, "step": 20245 }, { "epoch": 0.08, "learning_rate": 5.08111794014799e-06, "loss": 0.8542, "step": 20246 }, { "epoch": 0.08, "learning_rate": 5.080179502406099e-06, "loss": 0.8433, "step": 20247 }, { "epoch": 0.08, "learning_rate": 5.079241121824063e-06, "loss": 0.8703, "step": 20248 }, { "epoch": 0.08, "learning_rate": 5.078302798412776e-06, "loss": 0.8711, "step": 20249 }, { "epoch": 0.08, "learning_rate": 5.07736453218314e-06, "loss": 0.8909, "step": 20250 }, { "epoch": 0.08, "learning_rate": 5.076426323146062e-06, "loss": 0.8512, "step": 20251 }, { "epoch": 0.08, "learning_rate": 5.0754881713124394e-06, "loss": 0.876, "step": 20252 }, { "epoch": 0.08, "learning_rate": 5.07455007669317e-06, "loss": 0.8753, "step": 20253 }, { "epoch": 0.08, "learning_rate": 5.073612039299157e-06, "loss": 0.8721, "step": 20254 }, { "epoch": 0.08, "learning_rate": 5.0726740591413e-06, "loss": 0.8756, "step": 20255 }, { "epoch": 0.08, "learning_rate": 5.071736136230489e-06, "loss": 0.8544, "step": 20256 }, { "epoch": 0.08, "learning_rate": 5.070798270577624e-06, "loss": 0.8564, "step": 20257 }, { "epoch": 0.08, "learning_rate": 5.0698604621936075e-06, "loss": 0.8676, "step": 20258 }, { "epoch": 0.08, "learning_rate": 5.068922711089329e-06, "loss": 0.828, "step": 20259 }, { "epoch": 0.08, "learning_rate": 5.067985017275685e-06, "loss": 0.8766, "step": 20260 }, { "epoch": 0.08, "learning_rate": 5.067047380763569e-06, "loss": 0.9177, "step": 20261 }, { "epoch": 0.08, "learning_rate": 5.0661098015638766e-06, "loss": 0.9054, "step": 20262 }, { "epoch": 0.08, "learning_rate": 5.0651722796874995e-06, "loss": 0.8776, "step": 20263 }, { "epoch": 0.08, "learning_rate": 5.06423481514533e-06, "loss": 0.9032, "step": 20264 }, { "epoch": 0.08, "learning_rate": 5.063297407948267e-06, "loss": 0.8409, "step": 20265 }, { "epoch": 0.08, "learning_rate": 5.062360058107188e-06, "loss": 0.8936, "step": 20266 }, { "epoch": 0.08, "learning_rate": 5.061422765632993e-06, "loss": 0.8724, "step": 20267 }, { "epoch": 0.08, "learning_rate": 5.060485530536567e-06, "loss": 0.8455, "step": 20268 }, { "epoch": 0.08, "learning_rate": 5.059548352828803e-06, "loss": 0.8483, "step": 20269 }, { "epoch": 0.08, "learning_rate": 5.058611232520586e-06, "loss": 0.8452, "step": 20270 }, { "epoch": 0.08, "learning_rate": 5.057674169622806e-06, "loss": 0.8591, "step": 20271 }, { "epoch": 0.08, "learning_rate": 5.056737164146353e-06, "loss": 0.8603, "step": 20272 }, { "epoch": 0.08, "learning_rate": 5.055800216102107e-06, "loss": 0.8765, "step": 20273 }, { "epoch": 0.08, "learning_rate": 5.054863325500955e-06, "loss": 0.866, "step": 20274 }, { "epoch": 0.08, "learning_rate": 5.05392649235378e-06, "loss": 0.8814, "step": 20275 }, { "epoch": 0.08, "learning_rate": 5.052989716671478e-06, "loss": 0.8291, "step": 20276 }, { "epoch": 0.08, "learning_rate": 5.052052998464921e-06, "loss": 0.8649, "step": 20277 }, { "epoch": 0.08, "learning_rate": 5.051116337744996e-06, "loss": 0.8667, "step": 20278 }, { "epoch": 0.08, "learning_rate": 5.050179734522589e-06, "loss": 0.8787, "step": 20279 }, { "epoch": 0.08, "learning_rate": 5.049243188808573e-06, "loss": 0.8957, "step": 20280 }, { "epoch": 0.08, "learning_rate": 5.04830670061383e-06, "loss": 0.8595, "step": 20281 }, { "epoch": 0.08, "learning_rate": 5.0473702699492484e-06, "loss": 0.8476, "step": 20282 }, { "epoch": 0.08, "learning_rate": 5.046433896825709e-06, "loss": 0.8366, "step": 20283 }, { "epoch": 0.08, "learning_rate": 5.045497581254079e-06, "loss": 0.9143, "step": 20284 }, { "epoch": 0.08, "learning_rate": 5.044561323245245e-06, "loss": 0.8994, "step": 20285 }, { "epoch": 0.08, "learning_rate": 5.0436251228100826e-06, "loss": 0.8857, "step": 20286 }, { "epoch": 0.08, "learning_rate": 5.0426889799594714e-06, "loss": 0.8851, "step": 20287 }, { "epoch": 0.08, "learning_rate": 5.0417528947042835e-06, "loss": 0.8817, "step": 20288 }, { "epoch": 0.08, "learning_rate": 5.040816867055397e-06, "loss": 0.8396, "step": 20289 }, { "epoch": 0.08, "learning_rate": 5.039880897023691e-06, "loss": 0.8521, "step": 20290 }, { "epoch": 0.08, "learning_rate": 5.038944984620031e-06, "loss": 0.8653, "step": 20291 }, { "epoch": 0.08, "learning_rate": 5.038009129855297e-06, "loss": 0.8367, "step": 20292 }, { "epoch": 0.08, "learning_rate": 5.037073332740358e-06, "loss": 0.8411, "step": 20293 }, { "epoch": 0.08, "learning_rate": 5.0361375932860916e-06, "loss": 0.88, "step": 20294 }, { "epoch": 0.08, "learning_rate": 5.035201911503365e-06, "loss": 0.8877, "step": 20295 }, { "epoch": 0.08, "learning_rate": 5.03426628740305e-06, "loss": 0.8961, "step": 20296 }, { "epoch": 0.08, "learning_rate": 5.033330720996023e-06, "loss": 0.8266, "step": 20297 }, { "epoch": 0.08, "learning_rate": 5.032395212293143e-06, "loss": 0.8529, "step": 20298 }, { "epoch": 0.08, "learning_rate": 5.031459761305285e-06, "loss": 0.8642, "step": 20299 }, { "epoch": 0.08, "learning_rate": 5.030524368043313e-06, "loss": 0.8625, "step": 20300 }, { "epoch": 0.08, "learning_rate": 5.029589032518105e-06, "loss": 0.8849, "step": 20301 }, { "epoch": 0.08, "learning_rate": 5.02865375474052e-06, "loss": 0.8788, "step": 20302 }, { "epoch": 0.08, "learning_rate": 5.027718534721424e-06, "loss": 0.8414, "step": 20303 }, { "epoch": 0.08, "learning_rate": 5.026783372471688e-06, "loss": 0.8348, "step": 20304 }, { "epoch": 0.09, "learning_rate": 5.025848268002166e-06, "loss": 0.8589, "step": 20305 }, { "epoch": 0.09, "learning_rate": 5.024913221323733e-06, "loss": 0.8826, "step": 20306 }, { "epoch": 0.09, "learning_rate": 5.02397823244725e-06, "loss": 0.874, "step": 20307 }, { "epoch": 0.09, "learning_rate": 5.023043301383583e-06, "loss": 0.8829, "step": 20308 }, { "epoch": 0.09, "learning_rate": 5.022108428143586e-06, "loss": 0.8558, "step": 20309 }, { "epoch": 0.09, "learning_rate": 5.021173612738126e-06, "loss": 0.88, "step": 20310 }, { "epoch": 0.09, "learning_rate": 5.020238855178061e-06, "loss": 0.8306, "step": 20311 }, { "epoch": 0.09, "learning_rate": 5.019304155474256e-06, "loss": 0.8747, "step": 20312 }, { "epoch": 0.09, "learning_rate": 5.018369513637567e-06, "loss": 0.8518, "step": 20313 }, { "epoch": 0.09, "learning_rate": 5.0174349296788524e-06, "loss": 0.8753, "step": 20314 }, { "epoch": 0.09, "learning_rate": 5.016500403608977e-06, "loss": 0.8934, "step": 20315 }, { "epoch": 0.09, "learning_rate": 5.01556593543879e-06, "loss": 0.8903, "step": 20316 }, { "epoch": 0.09, "learning_rate": 5.014631525179151e-06, "loss": 0.8711, "step": 20317 }, { "epoch": 0.09, "learning_rate": 5.013697172840917e-06, "loss": 0.8397, "step": 20318 }, { "epoch": 0.09, "learning_rate": 5.012762878434943e-06, "loss": 0.8517, "step": 20319 }, { "epoch": 0.09, "learning_rate": 5.0118286419720856e-06, "loss": 0.8874, "step": 20320 }, { "epoch": 0.09, "learning_rate": 5.010894463463196e-06, "loss": 0.8431, "step": 20321 }, { "epoch": 0.09, "learning_rate": 5.009960342919133e-06, "loss": 0.8441, "step": 20322 }, { "epoch": 0.09, "learning_rate": 5.009026280350744e-06, "loss": 0.8813, "step": 20323 }, { "epoch": 0.09, "learning_rate": 5.0080922757688785e-06, "loss": 0.8839, "step": 20324 }, { "epoch": 0.09, "learning_rate": 5.007158329184401e-06, "loss": 0.8465, "step": 20325 }, { "epoch": 0.09, "learning_rate": 5.0062244406081495e-06, "loss": 0.8556, "step": 20326 }, { "epoch": 0.09, "learning_rate": 5.005290610050979e-06, "loss": 0.8793, "step": 20327 }, { "epoch": 0.09, "learning_rate": 5.004356837523738e-06, "loss": 0.9166, "step": 20328 }, { "epoch": 0.09, "learning_rate": 5.003423123037281e-06, "loss": 0.9011, "step": 20329 }, { "epoch": 0.09, "learning_rate": 5.0024894666024425e-06, "loss": 0.8715, "step": 20330 }, { "epoch": 0.09, "learning_rate": 5.001555868230084e-06, "loss": 0.8793, "step": 20331 }, { "epoch": 0.09, "learning_rate": 5.000622327931044e-06, "loss": 0.8655, "step": 20332 }, { "epoch": 0.09, "learning_rate": 4.999688845716179e-06, "loss": 0.8581, "step": 20333 }, { "epoch": 0.09, "learning_rate": 4.998755421596321e-06, "loss": 0.8303, "step": 20334 }, { "epoch": 0.09, "learning_rate": 4.997822055582321e-06, "loss": 0.846, "step": 20335 }, { "epoch": 0.09, "learning_rate": 4.996888747685024e-06, "loss": 0.8647, "step": 20336 }, { "epoch": 0.09, "learning_rate": 4.9959554979152715e-06, "loss": 0.8753, "step": 20337 }, { "epoch": 0.09, "learning_rate": 4.9950223062839075e-06, "loss": 0.8649, "step": 20338 }, { "epoch": 0.09, "learning_rate": 4.994089172801774e-06, "loss": 0.8673, "step": 20339 }, { "epoch": 0.09, "learning_rate": 4.993156097479716e-06, "loss": 0.8672, "step": 20340 }, { "epoch": 0.09, "learning_rate": 4.992223080328567e-06, "loss": 0.867, "step": 20341 }, { "epoch": 0.09, "learning_rate": 4.991290121359169e-06, "loss": 0.8462, "step": 20342 }, { "epoch": 0.09, "learning_rate": 4.990357220582364e-06, "loss": 0.8566, "step": 20343 }, { "epoch": 0.09, "learning_rate": 4.989424378008989e-06, "loss": 0.8661, "step": 20344 }, { "epoch": 0.09, "learning_rate": 4.988491593649883e-06, "loss": 0.8727, "step": 20345 }, { "epoch": 0.09, "learning_rate": 4.9875588675158815e-06, "loss": 0.9389, "step": 20346 }, { "epoch": 0.09, "learning_rate": 4.986626199617828e-06, "loss": 0.8758, "step": 20347 }, { "epoch": 0.09, "learning_rate": 4.985693589966545e-06, "loss": 0.8525, "step": 20348 }, { "epoch": 0.09, "learning_rate": 4.984761038572879e-06, "loss": 0.8279, "step": 20349 }, { "epoch": 0.09, "learning_rate": 4.983828545447666e-06, "loss": 0.8429, "step": 20350 }, { "epoch": 0.09, "learning_rate": 4.982896110601731e-06, "loss": 0.8461, "step": 20351 }, { "epoch": 0.09, "learning_rate": 4.9819637340459124e-06, "loss": 0.8389, "step": 20352 }, { "epoch": 0.09, "learning_rate": 4.981031415791042e-06, "loss": 0.8742, "step": 20353 }, { "epoch": 0.09, "learning_rate": 4.980099155847952e-06, "loss": 0.8844, "step": 20354 }, { "epoch": 0.09, "learning_rate": 4.979166954227475e-06, "loss": 0.8861, "step": 20355 }, { "epoch": 0.09, "learning_rate": 4.978234810940438e-06, "loss": 0.802, "step": 20356 }, { "epoch": 0.09, "learning_rate": 4.977302725997677e-06, "loss": 0.8499, "step": 20357 }, { "epoch": 0.09, "learning_rate": 4.976370699410015e-06, "loss": 0.8845, "step": 20358 }, { "epoch": 0.09, "learning_rate": 4.975438731188281e-06, "loss": 0.8739, "step": 20359 }, { "epoch": 0.09, "learning_rate": 4.9745068213433055e-06, "loss": 0.8466, "step": 20360 }, { "epoch": 0.09, "learning_rate": 4.973574969885916e-06, "loss": 0.8862, "step": 20361 }, { "epoch": 0.09, "learning_rate": 4.972643176826937e-06, "loss": 0.8811, "step": 20362 }, { "epoch": 0.09, "learning_rate": 4.971711442177195e-06, "loss": 0.881, "step": 20363 }, { "epoch": 0.09, "learning_rate": 4.970779765947519e-06, "loss": 0.8774, "step": 20364 }, { "epoch": 0.09, "learning_rate": 4.969848148148726e-06, "loss": 0.8742, "step": 20365 }, { "epoch": 0.09, "learning_rate": 4.968916588791644e-06, "loss": 0.8533, "step": 20366 }, { "epoch": 0.09, "learning_rate": 4.967985087887092e-06, "loss": 0.8754, "step": 20367 }, { "epoch": 0.09, "learning_rate": 4.967053645445904e-06, "loss": 0.8688, "step": 20368 }, { "epoch": 0.09, "learning_rate": 4.9661222614788895e-06, "loss": 0.8697, "step": 20369 }, { "epoch": 0.09, "learning_rate": 4.965190935996875e-06, "loss": 0.9122, "step": 20370 }, { "epoch": 0.09, "learning_rate": 4.96425966901068e-06, "loss": 0.8477, "step": 20371 }, { "epoch": 0.09, "learning_rate": 4.963328460531127e-06, "loss": 0.8664, "step": 20372 }, { "epoch": 0.09, "learning_rate": 4.962397310569025e-06, "loss": 0.8522, "step": 20373 }, { "epoch": 0.09, "learning_rate": 4.961466219135203e-06, "loss": 0.8732, "step": 20374 }, { "epoch": 0.09, "learning_rate": 4.960535186240479e-06, "loss": 0.8642, "step": 20375 }, { "epoch": 0.09, "learning_rate": 4.959604211895664e-06, "loss": 0.9023, "step": 20376 }, { "epoch": 0.09, "learning_rate": 4.958673296111575e-06, "loss": 0.8406, "step": 20377 }, { "epoch": 0.09, "learning_rate": 4.957742438899029e-06, "loss": 0.8679, "step": 20378 }, { "epoch": 0.09, "learning_rate": 4.956811640268843e-06, "loss": 0.8712, "step": 20379 }, { "epoch": 0.09, "learning_rate": 4.955880900231828e-06, "loss": 0.8922, "step": 20380 }, { "epoch": 0.09, "learning_rate": 4.954950218798798e-06, "loss": 0.8804, "step": 20381 }, { "epoch": 0.09, "learning_rate": 4.954019595980572e-06, "loss": 0.9192, "step": 20382 }, { "epoch": 0.09, "learning_rate": 4.953089031787955e-06, "loss": 0.8534, "step": 20383 }, { "epoch": 0.09, "learning_rate": 4.952158526231759e-06, "loss": 0.8259, "step": 20384 }, { "epoch": 0.09, "learning_rate": 4.951228079322797e-06, "loss": 0.8519, "step": 20385 }, { "epoch": 0.09, "learning_rate": 4.950297691071878e-06, "loss": 0.8861, "step": 20386 }, { "epoch": 0.09, "learning_rate": 4.949367361489813e-06, "loss": 0.8302, "step": 20387 }, { "epoch": 0.09, "learning_rate": 4.94843709058741e-06, "loss": 0.873, "step": 20388 }, { "epoch": 0.09, "learning_rate": 4.947506878375481e-06, "loss": 0.8065, "step": 20389 }, { "epoch": 0.09, "learning_rate": 4.946576724864827e-06, "loss": 0.8539, "step": 20390 }, { "epoch": 0.09, "learning_rate": 4.945646630066256e-06, "loss": 0.885, "step": 20391 }, { "epoch": 0.09, "learning_rate": 4.944716593990573e-06, "loss": 0.8394, "step": 20392 }, { "epoch": 0.09, "learning_rate": 4.943786616648592e-06, "loss": 0.8967, "step": 20393 }, { "epoch": 0.09, "learning_rate": 4.942856698051111e-06, "loss": 0.8534, "step": 20394 }, { "epoch": 0.09, "learning_rate": 4.941926838208933e-06, "loss": 0.8747, "step": 20395 }, { "epoch": 0.09, "learning_rate": 4.940997037132866e-06, "loss": 0.8471, "step": 20396 }, { "epoch": 0.09, "learning_rate": 4.9400672948337045e-06, "loss": 0.875, "step": 20397 }, { "epoch": 0.09, "learning_rate": 4.939137611322259e-06, "loss": 0.8559, "step": 20398 }, { "epoch": 0.09, "learning_rate": 4.938207986609327e-06, "loss": 0.8115, "step": 20399 }, { "epoch": 0.09, "learning_rate": 4.937278420705714e-06, "loss": 0.874, "step": 20400 }, { "epoch": 0.09, "learning_rate": 4.936348913622212e-06, "loss": 0.8909, "step": 20401 }, { "epoch": 0.09, "learning_rate": 4.935419465369625e-06, "loss": 0.8434, "step": 20402 }, { "epoch": 0.09, "learning_rate": 4.93449007595875e-06, "loss": 0.8527, "step": 20403 }, { "epoch": 0.09, "learning_rate": 4.9335607454003856e-06, "loss": 0.9142, "step": 20404 }, { "epoch": 0.09, "learning_rate": 4.9326314737053296e-06, "loss": 0.8643, "step": 20405 }, { "epoch": 0.09, "learning_rate": 4.931702260884377e-06, "loss": 0.8301, "step": 20406 }, { "epoch": 0.09, "learning_rate": 4.930773106948329e-06, "loss": 0.8598, "step": 20407 }, { "epoch": 0.09, "learning_rate": 4.929844011907973e-06, "loss": 0.838, "step": 20408 }, { "epoch": 0.09, "learning_rate": 4.928914975774106e-06, "loss": 0.8751, "step": 20409 }, { "epoch": 0.09, "learning_rate": 4.9279859985575225e-06, "loss": 0.8723, "step": 20410 }, { "epoch": 0.09, "learning_rate": 4.927057080269018e-06, "loss": 0.8518, "step": 20411 }, { "epoch": 0.09, "learning_rate": 4.92612822091938e-06, "loss": 0.8645, "step": 20412 }, { "epoch": 0.09, "learning_rate": 4.925199420519404e-06, "loss": 0.8548, "step": 20413 }, { "epoch": 0.09, "learning_rate": 4.924270679079884e-06, "loss": 0.9, "step": 20414 }, { "epoch": 0.09, "learning_rate": 4.923341996611604e-06, "loss": 0.8958, "step": 20415 }, { "epoch": 0.09, "learning_rate": 4.922413373125351e-06, "loss": 0.8833, "step": 20416 }, { "epoch": 0.09, "learning_rate": 4.921484808631922e-06, "loss": 0.887, "step": 20417 }, { "epoch": 0.09, "learning_rate": 4.920556303142108e-06, "loss": 0.8798, "step": 20418 }, { "epoch": 0.09, "learning_rate": 4.919627856666687e-06, "loss": 0.8412, "step": 20419 }, { "epoch": 0.09, "learning_rate": 4.918699469216449e-06, "loss": 0.8272, "step": 20420 }, { "epoch": 0.09, "learning_rate": 4.917771140802181e-06, "loss": 0.8422, "step": 20421 }, { "epoch": 0.09, "learning_rate": 4.9168428714346705e-06, "loss": 0.8671, "step": 20422 }, { "epoch": 0.09, "learning_rate": 4.9159146611247e-06, "loss": 0.9384, "step": 20423 }, { "epoch": 0.09, "learning_rate": 4.914986509883054e-06, "loss": 0.8794, "step": 20424 }, { "epoch": 0.09, "learning_rate": 4.9140584177205205e-06, "loss": 0.8674, "step": 20425 }, { "epoch": 0.09, "learning_rate": 4.913130384647875e-06, "loss": 0.8802, "step": 20426 }, { "epoch": 0.09, "learning_rate": 4.912202410675903e-06, "loss": 0.8623, "step": 20427 }, { "epoch": 0.09, "learning_rate": 4.9112744958153855e-06, "loss": 0.8816, "step": 20428 }, { "epoch": 0.09, "learning_rate": 4.910346640077104e-06, "loss": 0.8475, "step": 20429 }, { "epoch": 0.09, "learning_rate": 4.909418843471838e-06, "loss": 0.884, "step": 20430 }, { "epoch": 0.09, "learning_rate": 4.908491106010368e-06, "loss": 0.8698, "step": 20431 }, { "epoch": 0.09, "learning_rate": 4.907563427703476e-06, "loss": 0.8401, "step": 20432 }, { "epoch": 0.09, "learning_rate": 4.90663580856193e-06, "loss": 0.8443, "step": 20433 }, { "epoch": 0.09, "learning_rate": 4.905708248596514e-06, "loss": 0.8852, "step": 20434 }, { "epoch": 0.09, "learning_rate": 4.904780747818006e-06, "loss": 0.8484, "step": 20435 }, { "epoch": 0.09, "learning_rate": 4.903853306237178e-06, "loss": 0.8331, "step": 20436 }, { "epoch": 0.09, "learning_rate": 4.902925923864807e-06, "loss": 0.851, "step": 20437 }, { "epoch": 0.09, "learning_rate": 4.901998600711669e-06, "loss": 0.8937, "step": 20438 }, { "epoch": 0.09, "learning_rate": 4.90107133678854e-06, "loss": 0.8661, "step": 20439 }, { "epoch": 0.09, "learning_rate": 4.900144132106183e-06, "loss": 0.8483, "step": 20440 }, { "epoch": 0.09, "learning_rate": 4.89921698667538e-06, "loss": 0.8849, "step": 20441 }, { "epoch": 0.09, "learning_rate": 4.898289900506904e-06, "loss": 0.8527, "step": 20442 }, { "epoch": 0.09, "learning_rate": 4.897362873611519e-06, "loss": 0.8705, "step": 20443 }, { "epoch": 0.09, "learning_rate": 4.896435905999999e-06, "loss": 0.8584, "step": 20444 }, { "epoch": 0.09, "learning_rate": 4.895508997683114e-06, "loss": 0.8577, "step": 20445 }, { "epoch": 0.09, "learning_rate": 4.894582148671632e-06, "loss": 0.8537, "step": 20446 }, { "epoch": 0.09, "learning_rate": 4.893655358976321e-06, "loss": 0.8869, "step": 20447 }, { "epoch": 0.09, "learning_rate": 4.892728628607951e-06, "loss": 0.8476, "step": 20448 }, { "epoch": 0.09, "learning_rate": 4.8918019575772925e-06, "loss": 0.8628, "step": 20449 }, { "epoch": 0.09, "learning_rate": 4.890875345895102e-06, "loss": 0.871, "step": 20450 }, { "epoch": 0.09, "learning_rate": 4.88994879357215e-06, "loss": 0.8627, "step": 20451 }, { "epoch": 0.09, "learning_rate": 4.889022300619202e-06, "loss": 0.8601, "step": 20452 }, { "epoch": 0.09, "learning_rate": 4.888095867047023e-06, "loss": 0.8556, "step": 20453 }, { "epoch": 0.09, "learning_rate": 4.887169492866375e-06, "loss": 0.8884, "step": 20454 }, { "epoch": 0.09, "learning_rate": 4.886243178088022e-06, "loss": 0.9028, "step": 20455 }, { "epoch": 0.09, "learning_rate": 4.885316922722728e-06, "loss": 0.8601, "step": 20456 }, { "epoch": 0.09, "learning_rate": 4.884390726781249e-06, "loss": 0.8872, "step": 20457 }, { "epoch": 0.09, "learning_rate": 4.8834645902743485e-06, "loss": 0.8901, "step": 20458 }, { "epoch": 0.09, "learning_rate": 4.882538513212784e-06, "loss": 0.8666, "step": 20459 }, { "epoch": 0.09, "learning_rate": 4.881612495607324e-06, "loss": 0.8755, "step": 20460 }, { "epoch": 0.09, "learning_rate": 4.880686537468719e-06, "loss": 0.9002, "step": 20461 }, { "epoch": 0.09, "learning_rate": 4.879760638807728e-06, "loss": 0.8781, "step": 20462 }, { "epoch": 0.09, "learning_rate": 4.878834799635109e-06, "loss": 0.8848, "step": 20463 }, { "epoch": 0.09, "learning_rate": 4.877909019961624e-06, "loss": 0.8787, "step": 20464 }, { "epoch": 0.09, "learning_rate": 4.876983299798016e-06, "loss": 0.8649, "step": 20465 }, { "epoch": 0.09, "learning_rate": 4.876057639155053e-06, "loss": 0.884, "step": 20466 }, { "epoch": 0.09, "learning_rate": 4.875132038043489e-06, "loss": 0.8693, "step": 20467 }, { "epoch": 0.09, "learning_rate": 4.874206496474069e-06, "loss": 0.8551, "step": 20468 }, { "epoch": 0.09, "learning_rate": 4.8732810144575515e-06, "loss": 0.8804, "step": 20469 }, { "epoch": 0.09, "learning_rate": 4.872355592004689e-06, "loss": 0.8856, "step": 20470 }, { "epoch": 0.09, "learning_rate": 4.871430229126232e-06, "loss": 0.8506, "step": 20471 }, { "epoch": 0.09, "learning_rate": 4.870504925832934e-06, "loss": 0.8714, "step": 20472 }, { "epoch": 0.09, "learning_rate": 4.869579682135544e-06, "loss": 0.8461, "step": 20473 }, { "epoch": 0.09, "learning_rate": 4.8686544980448135e-06, "loss": 0.9224, "step": 20474 }, { "epoch": 0.09, "learning_rate": 4.867729373571487e-06, "loss": 0.841, "step": 20475 }, { "epoch": 0.09, "learning_rate": 4.866804308726316e-06, "loss": 0.8434, "step": 20476 }, { "epoch": 0.09, "learning_rate": 4.865879303520048e-06, "loss": 0.8759, "step": 20477 }, { "epoch": 0.09, "learning_rate": 4.864954357963429e-06, "loss": 0.8608, "step": 20478 }, { "epoch": 0.09, "learning_rate": 4.864029472067207e-06, "loss": 0.8413, "step": 20479 }, { "epoch": 0.09, "learning_rate": 4.863104645842126e-06, "loss": 0.8376, "step": 20480 }, { "epoch": 0.09, "learning_rate": 4.862179879298935e-06, "loss": 0.8571, "step": 20481 }, { "epoch": 0.09, "learning_rate": 4.861255172448372e-06, "loss": 0.8805, "step": 20482 }, { "epoch": 0.09, "learning_rate": 4.860330525301178e-06, "loss": 0.8822, "step": 20483 }, { "epoch": 0.09, "learning_rate": 4.859405937868106e-06, "loss": 0.8696, "step": 20484 }, { "epoch": 0.09, "learning_rate": 4.858481410159896e-06, "loss": 0.8684, "step": 20485 }, { "epoch": 0.09, "learning_rate": 4.857556942187283e-06, "loss": 0.8664, "step": 20486 }, { "epoch": 0.09, "learning_rate": 4.856632533961012e-06, "loss": 0.8634, "step": 20487 }, { "epoch": 0.09, "learning_rate": 4.855708185491825e-06, "loss": 0.8656, "step": 20488 }, { "epoch": 0.09, "learning_rate": 4.854783896790452e-06, "loss": 0.8629, "step": 20489 }, { "epoch": 0.09, "learning_rate": 4.853859667867641e-06, "loss": 0.9254, "step": 20490 }, { "epoch": 0.09, "learning_rate": 4.852935498734128e-06, "loss": 0.8481, "step": 20491 }, { "epoch": 0.09, "learning_rate": 4.852011389400653e-06, "loss": 0.8685, "step": 20492 }, { "epoch": 0.09, "learning_rate": 4.851087339877945e-06, "loss": 0.8589, "step": 20493 }, { "epoch": 0.09, "learning_rate": 4.850163350176745e-06, "loss": 0.8627, "step": 20494 }, { "epoch": 0.09, "learning_rate": 4.8492394203077876e-06, "loss": 0.8492, "step": 20495 }, { "epoch": 0.09, "learning_rate": 4.848315550281805e-06, "loss": 0.859, "step": 20496 }, { "epoch": 0.09, "learning_rate": 4.847391740109534e-06, "loss": 0.8845, "step": 20497 }, { "epoch": 0.09, "learning_rate": 4.846467989801706e-06, "loss": 0.86, "step": 20498 }, { "epoch": 0.09, "learning_rate": 4.845544299369057e-06, "loss": 0.877, "step": 20499 }, { "epoch": 0.09, "learning_rate": 4.844620668822312e-06, "loss": 0.8743, "step": 20500 }, { "epoch": 0.09, "learning_rate": 4.843697098172206e-06, "loss": 0.8677, "step": 20501 }, { "epoch": 0.09, "learning_rate": 4.842773587429469e-06, "loss": 0.8467, "step": 20502 }, { "epoch": 0.09, "learning_rate": 4.8418501366048295e-06, "loss": 0.8321, "step": 20503 }, { "epoch": 0.09, "learning_rate": 4.840926745709018e-06, "loss": 0.8463, "step": 20504 }, { "epoch": 0.09, "learning_rate": 4.840003414752762e-06, "loss": 0.881, "step": 20505 }, { "epoch": 0.09, "learning_rate": 4.839080143746793e-06, "loss": 0.896, "step": 20506 }, { "epoch": 0.09, "learning_rate": 4.838156932701829e-06, "loss": 0.8373, "step": 20507 }, { "epoch": 0.09, "learning_rate": 4.837233781628599e-06, "loss": 0.9063, "step": 20508 }, { "epoch": 0.09, "learning_rate": 4.836310690537833e-06, "loss": 0.8885, "step": 20509 }, { "epoch": 0.09, "learning_rate": 4.835387659440257e-06, "loss": 0.8614, "step": 20510 }, { "epoch": 0.09, "learning_rate": 4.834464688346587e-06, "loss": 0.8503, "step": 20511 }, { "epoch": 0.09, "learning_rate": 4.833541777267552e-06, "loss": 0.8652, "step": 20512 }, { "epoch": 0.09, "learning_rate": 4.832618926213872e-06, "loss": 0.86, "step": 20513 }, { "epoch": 0.09, "learning_rate": 4.831696135196271e-06, "loss": 0.8671, "step": 20514 }, { "epoch": 0.09, "learning_rate": 4.830773404225468e-06, "loss": 0.8989, "step": 20515 }, { "epoch": 0.09, "learning_rate": 4.829850733312187e-06, "loss": 0.8923, "step": 20516 }, { "epoch": 0.09, "learning_rate": 4.828928122467148e-06, "loss": 0.8652, "step": 20517 }, { "epoch": 0.09, "learning_rate": 4.828005571701064e-06, "loss": 0.8748, "step": 20518 }, { "epoch": 0.09, "learning_rate": 4.827083081024659e-06, "loss": 0.8189, "step": 20519 }, { "epoch": 0.09, "learning_rate": 4.826160650448647e-06, "loss": 0.8702, "step": 20520 }, { "epoch": 0.09, "learning_rate": 4.825238279983748e-06, "loss": 0.8933, "step": 20521 }, { "epoch": 0.09, "learning_rate": 4.824315969640678e-06, "loss": 0.9162, "step": 20522 }, { "epoch": 0.09, "learning_rate": 4.823393719430151e-06, "loss": 0.8696, "step": 20523 }, { "epoch": 0.09, "learning_rate": 4.822471529362889e-06, "loss": 0.8772, "step": 20524 }, { "epoch": 0.09, "learning_rate": 4.821549399449595e-06, "loss": 0.8513, "step": 20525 }, { "epoch": 0.09, "learning_rate": 4.820627329700989e-06, "loss": 0.8746, "step": 20526 }, { "epoch": 0.09, "learning_rate": 4.819705320127782e-06, "loss": 0.8635, "step": 20527 }, { "epoch": 0.09, "learning_rate": 4.818783370740687e-06, "loss": 0.8296, "step": 20528 }, { "epoch": 0.09, "learning_rate": 4.817861481550415e-06, "loss": 0.8683, "step": 20529 }, { "epoch": 0.09, "learning_rate": 4.81693965256768e-06, "loss": 0.9083, "step": 20530 }, { "epoch": 0.09, "learning_rate": 4.81601788380319e-06, "loss": 0.8417, "step": 20531 }, { "epoch": 0.09, "learning_rate": 4.815096175267647e-06, "loss": 0.864, "step": 20532 }, { "epoch": 0.09, "learning_rate": 4.81417452697177e-06, "loss": 0.843, "step": 20533 }, { "epoch": 0.09, "learning_rate": 4.813252938926267e-06, "loss": 0.8636, "step": 20534 }, { "epoch": 0.09, "learning_rate": 4.812331411141838e-06, "loss": 0.8881, "step": 20535 }, { "epoch": 0.09, "learning_rate": 4.811409943629194e-06, "loss": 0.9004, "step": 20536 }, { "epoch": 0.09, "learning_rate": 4.81048853639904e-06, "loss": 0.8639, "step": 20537 }, { "epoch": 0.09, "learning_rate": 4.80956718946208e-06, "loss": 0.8345, "step": 20538 }, { "epoch": 0.09, "learning_rate": 4.80864590282902e-06, "loss": 0.8518, "step": 20539 }, { "epoch": 0.09, "learning_rate": 4.807724676510563e-06, "loss": 0.8419, "step": 20540 }, { "epoch": 0.09, "learning_rate": 4.806803510517417e-06, "loss": 0.8805, "step": 20541 }, { "epoch": 0.09, "learning_rate": 4.805882404860276e-06, "loss": 0.8622, "step": 20542 }, { "epoch": 0.09, "learning_rate": 4.804961359549845e-06, "loss": 0.8556, "step": 20543 }, { "epoch": 0.09, "learning_rate": 4.804040374596826e-06, "loss": 0.8926, "step": 20544 }, { "epoch": 0.09, "learning_rate": 4.803119450011919e-06, "loss": 0.853, "step": 20545 }, { "epoch": 0.09, "learning_rate": 4.802198585805823e-06, "loss": 0.8511, "step": 20546 }, { "epoch": 0.09, "learning_rate": 4.801277781989238e-06, "loss": 0.8847, "step": 20547 }, { "epoch": 0.09, "learning_rate": 4.800357038572863e-06, "loss": 0.865, "step": 20548 }, { "epoch": 0.09, "learning_rate": 4.799436355567391e-06, "loss": 0.8855, "step": 20549 }, { "epoch": 0.09, "learning_rate": 4.798515732983522e-06, "loss": 0.8581, "step": 20550 }, { "epoch": 0.09, "learning_rate": 4.797595170831947e-06, "loss": 0.9174, "step": 20551 }, { "epoch": 0.09, "learning_rate": 4.796674669123373e-06, "loss": 0.8565, "step": 20552 }, { "epoch": 0.09, "learning_rate": 4.795754227868486e-06, "loss": 0.8511, "step": 20553 }, { "epoch": 0.09, "learning_rate": 4.7948338470779794e-06, "loss": 0.8715, "step": 20554 }, { "epoch": 0.09, "learning_rate": 4.793913526762549e-06, "loss": 0.8528, "step": 20555 }, { "epoch": 0.09, "learning_rate": 4.792993266932887e-06, "loss": 0.8796, "step": 20556 }, { "epoch": 0.09, "learning_rate": 4.792073067599685e-06, "loss": 0.8781, "step": 20557 }, { "epoch": 0.09, "learning_rate": 4.7911529287736345e-06, "loss": 0.8913, "step": 20558 }, { "epoch": 0.09, "learning_rate": 4.790232850465429e-06, "loss": 0.8643, "step": 20559 }, { "epoch": 0.09, "learning_rate": 4.789312832685752e-06, "loss": 0.853, "step": 20560 }, { "epoch": 0.09, "learning_rate": 4.788392875445295e-06, "loss": 0.8477, "step": 20561 }, { "epoch": 0.09, "learning_rate": 4.7874729787547465e-06, "loss": 0.9065, "step": 20562 }, { "epoch": 0.09, "learning_rate": 4.786553142624795e-06, "loss": 0.876, "step": 20563 }, { "epoch": 0.09, "learning_rate": 4.785633367066126e-06, "loss": 0.8921, "step": 20564 }, { "epoch": 0.09, "learning_rate": 4.7847136520894264e-06, "loss": 0.9021, "step": 20565 }, { "epoch": 0.09, "learning_rate": 4.783793997705386e-06, "loss": 0.8825, "step": 20566 }, { "epoch": 0.09, "learning_rate": 4.782874403924682e-06, "loss": 0.8857, "step": 20567 }, { "epoch": 0.09, "learning_rate": 4.781954870758001e-06, "loss": 0.9176, "step": 20568 }, { "epoch": 0.09, "learning_rate": 4.7810353982160285e-06, "loss": 0.844, "step": 20569 }, { "epoch": 0.09, "learning_rate": 4.780115986309445e-06, "loss": 0.9041, "step": 20570 }, { "epoch": 0.09, "learning_rate": 4.779196635048934e-06, "loss": 0.8643, "step": 20571 }, { "epoch": 0.09, "learning_rate": 4.7782773444451756e-06, "loss": 0.8675, "step": 20572 }, { "epoch": 0.09, "learning_rate": 4.7773581145088556e-06, "loss": 0.8804, "step": 20573 }, { "epoch": 0.09, "learning_rate": 4.7764389452506445e-06, "loss": 0.842, "step": 20574 }, { "epoch": 0.09, "learning_rate": 4.775519836681223e-06, "loss": 0.8497, "step": 20575 }, { "epoch": 0.09, "learning_rate": 4.774600788811277e-06, "loss": 0.8552, "step": 20576 }, { "epoch": 0.09, "learning_rate": 4.773681801651483e-06, "loss": 0.8713, "step": 20577 }, { "epoch": 0.09, "learning_rate": 4.772762875212511e-06, "loss": 0.8489, "step": 20578 }, { "epoch": 0.09, "learning_rate": 4.771844009505042e-06, "loss": 0.8783, "step": 20579 }, { "epoch": 0.09, "learning_rate": 4.7709252045397555e-06, "loss": 0.8578, "step": 20580 }, { "epoch": 0.09, "learning_rate": 4.770006460327314e-06, "loss": 0.8871, "step": 20581 }, { "epoch": 0.09, "learning_rate": 4.769087776878404e-06, "loss": 0.8494, "step": 20582 }, { "epoch": 0.09, "learning_rate": 4.7681691542036945e-06, "loss": 0.8585, "step": 20583 }, { "epoch": 0.09, "learning_rate": 4.7672505923138626e-06, "loss": 0.893, "step": 20584 }, { "epoch": 0.09, "learning_rate": 4.766332091219572e-06, "loss": 0.8689, "step": 20585 }, { "epoch": 0.09, "learning_rate": 4.7654136509314985e-06, "loss": 0.8453, "step": 20586 }, { "epoch": 0.09, "learning_rate": 4.7644952714603135e-06, "loss": 0.8334, "step": 20587 }, { "epoch": 0.09, "learning_rate": 4.763576952816685e-06, "loss": 0.8371, "step": 20588 }, { "epoch": 0.09, "learning_rate": 4.762658695011284e-06, "loss": 0.8704, "step": 20589 }, { "epoch": 0.09, "learning_rate": 4.76174049805478e-06, "loss": 0.9139, "step": 20590 }, { "epoch": 0.09, "learning_rate": 4.760822361957842e-06, "loss": 0.852, "step": 20591 }, { "epoch": 0.09, "learning_rate": 4.759904286731132e-06, "loss": 0.841, "step": 20592 }, { "epoch": 0.09, "learning_rate": 4.758986272385318e-06, "loss": 0.8881, "step": 20593 }, { "epoch": 0.09, "learning_rate": 4.758068318931067e-06, "loss": 0.861, "step": 20594 }, { "epoch": 0.09, "learning_rate": 4.757150426379045e-06, "loss": 0.8474, "step": 20595 }, { "epoch": 0.09, "learning_rate": 4.7562325947399135e-06, "loss": 0.8825, "step": 20596 }, { "epoch": 0.09, "learning_rate": 4.755314824024338e-06, "loss": 0.8641, "step": 20597 }, { "epoch": 0.09, "learning_rate": 4.754397114242987e-06, "loss": 0.8647, "step": 20598 }, { "epoch": 0.09, "learning_rate": 4.753479465406512e-06, "loss": 0.9118, "step": 20599 }, { "epoch": 0.09, "learning_rate": 4.752561877525575e-06, "loss": 0.8914, "step": 20600 }, { "epoch": 0.09, "learning_rate": 4.751644350610847e-06, "loss": 0.8471, "step": 20601 }, { "epoch": 0.09, "learning_rate": 4.750726884672984e-06, "loss": 0.8672, "step": 20602 }, { "epoch": 0.09, "learning_rate": 4.749809479722641e-06, "loss": 0.8613, "step": 20603 }, { "epoch": 0.09, "learning_rate": 4.74889213577048e-06, "loss": 0.8965, "step": 20604 }, { "epoch": 0.1, "learning_rate": 4.747974852827158e-06, "loss": 0.862, "step": 20605 }, { "epoch": 0.1, "learning_rate": 4.747057630903332e-06, "loss": 0.9084, "step": 20606 }, { "epoch": 0.1, "learning_rate": 4.74614047000966e-06, "loss": 0.8701, "step": 20607 }, { "epoch": 0.1, "learning_rate": 4.745223370156797e-06, "loss": 0.8834, "step": 20608 }, { "epoch": 0.1, "learning_rate": 4.744306331355401e-06, "loss": 0.887, "step": 20609 }, { "epoch": 0.1, "learning_rate": 4.743389353616122e-06, "loss": 0.8694, "step": 20610 }, { "epoch": 0.1, "learning_rate": 4.742472436949612e-06, "loss": 0.854, "step": 20611 }, { "epoch": 0.1, "learning_rate": 4.741555581366529e-06, "loss": 0.8684, "step": 20612 }, { "epoch": 0.1, "learning_rate": 4.740638786877524e-06, "loss": 0.9043, "step": 20613 }, { "epoch": 0.1, "learning_rate": 4.739722053493248e-06, "loss": 0.8401, "step": 20614 }, { "epoch": 0.1, "learning_rate": 4.7388053812243504e-06, "loss": 0.8571, "step": 20615 }, { "epoch": 0.1, "learning_rate": 4.737888770081488e-06, "loss": 0.8621, "step": 20616 }, { "epoch": 0.1, "learning_rate": 4.736972220075301e-06, "loss": 0.86, "step": 20617 }, { "epoch": 0.1, "learning_rate": 4.736055731216439e-06, "loss": 0.883, "step": 20618 }, { "epoch": 0.1, "learning_rate": 4.73513930351556e-06, "loss": 0.8343, "step": 20619 }, { "epoch": 0.1, "learning_rate": 4.734222936983302e-06, "loss": 0.8773, "step": 20620 }, { "epoch": 0.1, "learning_rate": 4.733306631630314e-06, "loss": 0.8887, "step": 20621 }, { "epoch": 0.1, "learning_rate": 4.732390387467241e-06, "loss": 0.9139, "step": 20622 }, { "epoch": 0.1, "learning_rate": 4.731474204504734e-06, "loss": 0.8296, "step": 20623 }, { "epoch": 0.1, "learning_rate": 4.730558082753425e-06, "loss": 0.8853, "step": 20624 }, { "epoch": 0.1, "learning_rate": 4.729642022223969e-06, "loss": 0.8636, "step": 20625 }, { "epoch": 0.1, "learning_rate": 4.72872602292701e-06, "loss": 0.8899, "step": 20626 }, { "epoch": 0.1, "learning_rate": 4.727810084873181e-06, "loss": 0.8781, "step": 20627 }, { "epoch": 0.1, "learning_rate": 4.726894208073129e-06, "loss": 0.8995, "step": 20628 }, { "epoch": 0.1, "learning_rate": 4.7259783925374955e-06, "loss": 0.855, "step": 20629 }, { "epoch": 0.1, "learning_rate": 4.72506263827692e-06, "loss": 0.8744, "step": 20630 }, { "epoch": 0.1, "learning_rate": 4.72414694530204e-06, "loss": 0.8862, "step": 20631 }, { "epoch": 0.1, "learning_rate": 4.723231313623496e-06, "loss": 0.8384, "step": 20632 }, { "epoch": 0.1, "learning_rate": 4.722315743251931e-06, "loss": 0.8421, "step": 20633 }, { "epoch": 0.1, "learning_rate": 4.721400234197973e-06, "loss": 0.8833, "step": 20634 }, { "epoch": 0.1, "learning_rate": 4.720484786472263e-06, "loss": 0.8808, "step": 20635 }, { "epoch": 0.1, "learning_rate": 4.7195694000854365e-06, "loss": 0.8602, "step": 20636 }, { "epoch": 0.1, "learning_rate": 4.718654075048129e-06, "loss": 0.9009, "step": 20637 }, { "epoch": 0.1, "learning_rate": 4.717738811370975e-06, "loss": 0.8663, "step": 20638 }, { "epoch": 0.1, "learning_rate": 4.716823609064609e-06, "loss": 0.8909, "step": 20639 }, { "epoch": 0.1, "learning_rate": 4.715908468139668e-06, "loss": 0.8616, "step": 20640 }, { "epoch": 0.1, "learning_rate": 4.714993388606775e-06, "loss": 0.8754, "step": 20641 }, { "epoch": 0.1, "learning_rate": 4.7140783704765666e-06, "loss": 0.8564, "step": 20642 }, { "epoch": 0.1, "learning_rate": 4.71316341375967e-06, "loss": 0.8491, "step": 20643 }, { "epoch": 0.1, "learning_rate": 4.712248518466727e-06, "loss": 0.8141, "step": 20644 }, { "epoch": 0.1, "learning_rate": 4.711333684608354e-06, "loss": 0.8741, "step": 20645 }, { "epoch": 0.1, "learning_rate": 4.710418912195187e-06, "loss": 0.8946, "step": 20646 }, { "epoch": 0.1, "learning_rate": 4.709504201237851e-06, "loss": 0.8234, "step": 20647 }, { "epoch": 0.1, "learning_rate": 4.708589551746975e-06, "loss": 0.8622, "step": 20648 }, { "epoch": 0.1, "learning_rate": 4.707674963733184e-06, "loss": 0.8411, "step": 20649 }, { "epoch": 0.1, "learning_rate": 4.706760437207106e-06, "loss": 0.8646, "step": 20650 }, { "epoch": 0.1, "learning_rate": 4.70584597217937e-06, "loss": 0.9179, "step": 20651 }, { "epoch": 0.1, "learning_rate": 4.704931568660589e-06, "loss": 0.8696, "step": 20652 }, { "epoch": 0.1, "learning_rate": 4.704017226661396e-06, "loss": 0.8869, "step": 20653 }, { "epoch": 0.1, "learning_rate": 4.7031029461924105e-06, "loss": 0.8275, "step": 20654 }, { "epoch": 0.1, "learning_rate": 4.702188727264256e-06, "loss": 0.8495, "step": 20655 }, { "epoch": 0.1, "learning_rate": 4.701274569887554e-06, "loss": 0.8483, "step": 20656 }, { "epoch": 0.1, "learning_rate": 4.700360474072926e-06, "loss": 0.896, "step": 20657 }, { "epoch": 0.1, "learning_rate": 4.699446439830996e-06, "loss": 0.8364, "step": 20658 }, { "epoch": 0.1, "learning_rate": 4.6985324671723756e-06, "loss": 0.8915, "step": 20659 }, { "epoch": 0.1, "learning_rate": 4.6976185561076855e-06, "loss": 0.8858, "step": 20660 }, { "epoch": 0.1, "learning_rate": 4.696704706647543e-06, "loss": 0.8554, "step": 20661 }, { "epoch": 0.1, "learning_rate": 4.695790918802577e-06, "loss": 0.8866, "step": 20662 }, { "epoch": 0.1, "learning_rate": 4.694877192583389e-06, "loss": 0.9013, "step": 20663 }, { "epoch": 0.1, "learning_rate": 4.693963528000602e-06, "loss": 0.9152, "step": 20664 }, { "epoch": 0.1, "learning_rate": 4.693049925064834e-06, "loss": 0.8662, "step": 20665 }, { "epoch": 0.1, "learning_rate": 4.692136383786693e-06, "loss": 0.8611, "step": 20666 }, { "epoch": 0.1, "learning_rate": 4.691222904176791e-06, "loss": 0.8544, "step": 20667 }, { "epoch": 0.1, "learning_rate": 4.69030948624575e-06, "loss": 0.9075, "step": 20668 }, { "epoch": 0.1, "learning_rate": 4.68939613000418e-06, "loss": 0.8792, "step": 20669 }, { "epoch": 0.1, "learning_rate": 4.688482835462689e-06, "loss": 0.8761, "step": 20670 }, { "epoch": 0.1, "learning_rate": 4.6875696026318886e-06, "loss": 0.8458, "step": 20671 }, { "epoch": 0.1, "learning_rate": 4.686656431522393e-06, "loss": 0.9108, "step": 20672 }, { "epoch": 0.1, "learning_rate": 4.685743322144802e-06, "loss": 0.8601, "step": 20673 }, { "epoch": 0.1, "learning_rate": 4.684830274509734e-06, "loss": 0.8732, "step": 20674 }, { "epoch": 0.1, "learning_rate": 4.683917288627795e-06, "loss": 0.8588, "step": 20675 }, { "epoch": 0.1, "learning_rate": 4.6830043645095945e-06, "loss": 0.8346, "step": 20676 }, { "epoch": 0.1, "learning_rate": 4.682091502165732e-06, "loss": 0.8402, "step": 20677 }, { "epoch": 0.1, "learning_rate": 4.681178701606818e-06, "loss": 0.87, "step": 20678 }, { "epoch": 0.1, "learning_rate": 4.680265962843457e-06, "loss": 0.8773, "step": 20679 }, { "epoch": 0.1, "learning_rate": 4.6793532858862526e-06, "loss": 0.8581, "step": 20680 }, { "epoch": 0.1, "learning_rate": 4.6784406707458094e-06, "loss": 0.8568, "step": 20681 }, { "epoch": 0.1, "learning_rate": 4.6775281174327304e-06, "loss": 0.888, "step": 20682 }, { "epoch": 0.1, "learning_rate": 4.676615625957622e-06, "loss": 0.8286, "step": 20683 }, { "epoch": 0.1, "learning_rate": 4.675703196331078e-06, "loss": 0.8532, "step": 20684 }, { "epoch": 0.1, "learning_rate": 4.674790828563704e-06, "loss": 0.8738, "step": 20685 }, { "epoch": 0.1, "learning_rate": 4.673878522666097e-06, "loss": 0.895, "step": 20686 }, { "epoch": 0.1, "learning_rate": 4.672966278648859e-06, "loss": 0.8864, "step": 20687 }, { "epoch": 0.1, "learning_rate": 4.672054096522589e-06, "loss": 0.8517, "step": 20688 }, { "epoch": 0.1, "learning_rate": 4.6711419762978836e-06, "loss": 0.8944, "step": 20689 }, { "epoch": 0.1, "learning_rate": 4.670229917985345e-06, "loss": 0.8644, "step": 20690 }, { "epoch": 0.1, "learning_rate": 4.669317921595558e-06, "loss": 0.8685, "step": 20691 }, { "epoch": 0.1, "learning_rate": 4.668405987139129e-06, "loss": 0.8596, "step": 20692 }, { "epoch": 0.1, "learning_rate": 4.6674941146266505e-06, "loss": 0.8622, "step": 20693 }, { "epoch": 0.1, "learning_rate": 4.6665823040687195e-06, "loss": 0.8626, "step": 20694 }, { "epoch": 0.1, "learning_rate": 4.665670555475923e-06, "loss": 0.8955, "step": 20695 }, { "epoch": 0.1, "learning_rate": 4.664758868858858e-06, "loss": 0.8836, "step": 20696 }, { "epoch": 0.1, "learning_rate": 4.663847244228116e-06, "loss": 0.8524, "step": 20697 }, { "epoch": 0.1, "learning_rate": 4.6629356815942885e-06, "loss": 0.893, "step": 20698 }, { "epoch": 0.1, "learning_rate": 4.662024180967966e-06, "loss": 0.8265, "step": 20699 }, { "epoch": 0.1, "learning_rate": 4.661112742359739e-06, "loss": 0.8488, "step": 20700 }, { "epoch": 0.1, "learning_rate": 4.660201365780202e-06, "loss": 0.8531, "step": 20701 }, { "epoch": 0.1, "learning_rate": 4.659290051239934e-06, "loss": 0.8341, "step": 20702 }, { "epoch": 0.1, "learning_rate": 4.658378798749527e-06, "loss": 0.8232, "step": 20703 }, { "epoch": 0.1, "learning_rate": 4.657467608319569e-06, "loss": 0.8805, "step": 20704 }, { "epoch": 0.1, "learning_rate": 4.656556479960647e-06, "loss": 0.9076, "step": 20705 }, { "epoch": 0.1, "learning_rate": 4.655645413683345e-06, "loss": 0.8549, "step": 20706 }, { "epoch": 0.1, "learning_rate": 4.654734409498248e-06, "loss": 0.8595, "step": 20707 }, { "epoch": 0.1, "learning_rate": 4.6538234674159465e-06, "loss": 0.8503, "step": 20708 }, { "epoch": 0.1, "learning_rate": 4.652912587447014e-06, "loss": 0.8601, "step": 20709 }, { "epoch": 0.1, "learning_rate": 4.652001769602035e-06, "loss": 0.883, "step": 20710 }, { "epoch": 0.1, "learning_rate": 4.651091013891602e-06, "loss": 0.825, "step": 20711 }, { "epoch": 0.1, "learning_rate": 4.650180320326285e-06, "loss": 0.8745, "step": 20712 }, { "epoch": 0.1, "learning_rate": 4.64926968891667e-06, "loss": 0.8293, "step": 20713 }, { "epoch": 0.1, "learning_rate": 4.648359119673334e-06, "loss": 0.8483, "step": 20714 }, { "epoch": 0.1, "learning_rate": 4.647448612606863e-06, "loss": 0.8474, "step": 20715 }, { "epoch": 0.1, "learning_rate": 4.646538167727823e-06, "loss": 0.8865, "step": 20716 }, { "epoch": 0.1, "learning_rate": 4.645627785046803e-06, "loss": 0.8636, "step": 20717 }, { "epoch": 0.1, "learning_rate": 4.644717464574381e-06, "loss": 0.8444, "step": 20718 }, { "epoch": 0.1, "learning_rate": 4.643807206321125e-06, "loss": 0.8414, "step": 20719 }, { "epoch": 0.1, "learning_rate": 4.6428970102976135e-06, "loss": 0.8838, "step": 20720 }, { "epoch": 0.1, "learning_rate": 4.641986876514423e-06, "loss": 0.8963, "step": 20721 }, { "epoch": 0.1, "learning_rate": 4.641076804982127e-06, "loss": 0.8649, "step": 20722 }, { "epoch": 0.1, "learning_rate": 4.6401667957112985e-06, "loss": 0.8248, "step": 20723 }, { "epoch": 0.1, "learning_rate": 4.639256848712511e-06, "loss": 0.8782, "step": 20724 }, { "epoch": 0.1, "learning_rate": 4.638346963996341e-06, "loss": 0.8646, "step": 20725 }, { "epoch": 0.1, "learning_rate": 4.63743714157335e-06, "loss": 0.8799, "step": 20726 }, { "epoch": 0.1, "learning_rate": 4.636527381454114e-06, "loss": 0.8513, "step": 20727 }, { "epoch": 0.1, "learning_rate": 4.6356176836492025e-06, "loss": 0.8537, "step": 20728 }, { "epoch": 0.1, "learning_rate": 4.634708048169185e-06, "loss": 0.8137, "step": 20729 }, { "epoch": 0.1, "learning_rate": 4.63379847502463e-06, "loss": 0.8479, "step": 20730 }, { "epoch": 0.1, "learning_rate": 4.632888964226104e-06, "loss": 0.8476, "step": 20731 }, { "epoch": 0.1, "learning_rate": 4.631979515784178e-06, "loss": 0.8366, "step": 20732 }, { "epoch": 0.1, "learning_rate": 4.631070129709413e-06, "loss": 0.8421, "step": 20733 }, { "epoch": 0.1, "learning_rate": 4.6301608060123706e-06, "loss": 0.8859, "step": 20734 }, { "epoch": 0.1, "learning_rate": 4.6292515447036255e-06, "loss": 0.8405, "step": 20735 }, { "epoch": 0.1, "learning_rate": 4.628342345793743e-06, "loss": 0.8684, "step": 20736 }, { "epoch": 0.1, "learning_rate": 4.627433209293276e-06, "loss": 0.861, "step": 20737 }, { "epoch": 0.1, "learning_rate": 4.626524135212792e-06, "loss": 0.8363, "step": 20738 }, { "epoch": 0.1, "learning_rate": 4.6256151235628535e-06, "loss": 0.8495, "step": 20739 }, { "epoch": 0.1, "learning_rate": 4.62470617435402e-06, "loss": 0.8898, "step": 20740 }, { "epoch": 0.1, "learning_rate": 4.623797287596854e-06, "loss": 0.8637, "step": 20741 }, { "epoch": 0.1, "learning_rate": 4.622888463301914e-06, "loss": 0.8499, "step": 20742 }, { "epoch": 0.1, "learning_rate": 4.6219797014797645e-06, "loss": 0.8986, "step": 20743 }, { "epoch": 0.1, "learning_rate": 4.621071002140953e-06, "loss": 0.9011, "step": 20744 }, { "epoch": 0.1, "learning_rate": 4.620162365296044e-06, "loss": 0.8687, "step": 20745 }, { "epoch": 0.1, "learning_rate": 4.6192537909555915e-06, "loss": 0.8578, "step": 20746 }, { "epoch": 0.1, "learning_rate": 4.618345279130153e-06, "loss": 0.8587, "step": 20747 }, { "epoch": 0.1, "learning_rate": 4.617436829830285e-06, "loss": 0.8415, "step": 20748 }, { "epoch": 0.1, "learning_rate": 4.61652844306654e-06, "loss": 0.8418, "step": 20749 }, { "epoch": 0.1, "learning_rate": 4.6156201188494765e-06, "loss": 0.902, "step": 20750 }, { "epoch": 0.1, "learning_rate": 4.61471185718964e-06, "loss": 0.8575, "step": 20751 }, { "epoch": 0.1, "learning_rate": 4.613803658097588e-06, "loss": 0.895, "step": 20752 }, { "epoch": 0.1, "learning_rate": 4.612895521583867e-06, "loss": 0.8651, "step": 20753 }, { "epoch": 0.1, "learning_rate": 4.611987447659039e-06, "loss": 0.8892, "step": 20754 }, { "epoch": 0.1, "learning_rate": 4.611079436333643e-06, "loss": 0.8398, "step": 20755 }, { "epoch": 0.1, "learning_rate": 4.610171487618233e-06, "loss": 0.85, "step": 20756 }, { "epoch": 0.1, "learning_rate": 4.609263601523363e-06, "loss": 0.8713, "step": 20757 }, { "epoch": 0.1, "learning_rate": 4.60835577805957e-06, "loss": 0.8637, "step": 20758 }, { "epoch": 0.1, "learning_rate": 4.607448017237405e-06, "loss": 0.8465, "step": 20759 }, { "epoch": 0.1, "learning_rate": 4.60654031906742e-06, "loss": 0.8486, "step": 20760 }, { "epoch": 0.1, "learning_rate": 4.60563268356016e-06, "loss": 0.8758, "step": 20761 }, { "epoch": 0.1, "learning_rate": 4.6047251107261655e-06, "loss": 0.8878, "step": 20762 }, { "epoch": 0.1, "learning_rate": 4.603817600575983e-06, "loss": 0.8511, "step": 20763 }, { "epoch": 0.1, "learning_rate": 4.602910153120156e-06, "loss": 0.8677, "step": 20764 }, { "epoch": 0.1, "learning_rate": 4.6020027683692285e-06, "loss": 0.8875, "step": 20765 }, { "epoch": 0.1, "learning_rate": 4.601095446333741e-06, "loss": 0.8417, "step": 20766 }, { "epoch": 0.1, "learning_rate": 4.600188187024237e-06, "loss": 0.8576, "step": 20767 }, { "epoch": 0.1, "learning_rate": 4.59928099045126e-06, "loss": 0.8574, "step": 20768 }, { "epoch": 0.1, "learning_rate": 4.598373856625342e-06, "loss": 0.8503, "step": 20769 }, { "epoch": 0.1, "learning_rate": 4.597466785557029e-06, "loss": 0.8839, "step": 20770 }, { "epoch": 0.1, "learning_rate": 4.5965597772568545e-06, "loss": 0.8762, "step": 20771 }, { "epoch": 0.1, "learning_rate": 4.595652831735362e-06, "loss": 0.9126, "step": 20772 }, { "epoch": 0.1, "learning_rate": 4.594745949003084e-06, "loss": 0.8721, "step": 20773 }, { "epoch": 0.1, "learning_rate": 4.5938391290705595e-06, "loss": 0.8687, "step": 20774 }, { "epoch": 0.1, "learning_rate": 4.592932371948327e-06, "loss": 0.8741, "step": 20775 }, { "epoch": 0.1, "learning_rate": 4.5920256776469164e-06, "loss": 0.8769, "step": 20776 }, { "epoch": 0.1, "learning_rate": 4.591119046176862e-06, "loss": 0.8607, "step": 20777 }, { "epoch": 0.1, "learning_rate": 4.590212477548698e-06, "loss": 0.8895, "step": 20778 }, { "epoch": 0.1, "learning_rate": 4.589305971772959e-06, "loss": 0.8563, "step": 20779 }, { "epoch": 0.1, "learning_rate": 4.5883995288601755e-06, "loss": 0.8406, "step": 20780 }, { "epoch": 0.1, "learning_rate": 4.58749314882088e-06, "loss": 0.8713, "step": 20781 }, { "epoch": 0.1, "learning_rate": 4.586586831665606e-06, "loss": 0.8619, "step": 20782 }, { "epoch": 0.1, "learning_rate": 4.585680577404872e-06, "loss": 0.8539, "step": 20783 }, { "epoch": 0.1, "learning_rate": 4.584774386049219e-06, "loss": 0.8645, "step": 20784 }, { "epoch": 0.1, "learning_rate": 4.583868257609171e-06, "loss": 0.8502, "step": 20785 }, { "epoch": 0.1, "learning_rate": 4.582962192095259e-06, "loss": 0.8387, "step": 20786 }, { "epoch": 0.1, "learning_rate": 4.582056189518005e-06, "loss": 0.8699, "step": 20787 }, { "epoch": 0.1, "learning_rate": 4.581150249887936e-06, "loss": 0.8588, "step": 20788 }, { "epoch": 0.1, "learning_rate": 4.580244373215578e-06, "loss": 0.8131, "step": 20789 }, { "epoch": 0.1, "learning_rate": 4.579338559511457e-06, "loss": 0.87, "step": 20790 }, { "epoch": 0.1, "learning_rate": 4.578432808786096e-06, "loss": 0.8668, "step": 20791 }, { "epoch": 0.1, "learning_rate": 4.577527121050018e-06, "loss": 0.8879, "step": 20792 }, { "epoch": 0.1, "learning_rate": 4.57662149631375e-06, "loss": 0.8848, "step": 20793 }, { "epoch": 0.1, "learning_rate": 4.575715934587807e-06, "loss": 0.8825, "step": 20794 }, { "epoch": 0.1, "learning_rate": 4.574810435882713e-06, "loss": 0.8755, "step": 20795 }, { "epoch": 0.1, "learning_rate": 4.573905000208987e-06, "loss": 0.821, "step": 20796 }, { "epoch": 0.1, "learning_rate": 4.57299962757715e-06, "loss": 0.8594, "step": 20797 }, { "epoch": 0.1, "learning_rate": 4.572094317997722e-06, "loss": 0.8744, "step": 20798 }, { "epoch": 0.1, "learning_rate": 4.571189071481218e-06, "loss": 0.8549, "step": 20799 }, { "epoch": 0.1, "learning_rate": 4.5702838880381615e-06, "loss": 0.8718, "step": 20800 }, { "epoch": 0.1, "learning_rate": 4.5693787676790625e-06, "loss": 0.8403, "step": 20801 }, { "epoch": 0.1, "learning_rate": 4.568473710414435e-06, "loss": 0.8782, "step": 20802 }, { "epoch": 0.1, "learning_rate": 4.567568716254806e-06, "loss": 0.8761, "step": 20803 }, { "epoch": 0.1, "learning_rate": 4.5666637852106775e-06, "loss": 0.8252, "step": 20804 }, { "epoch": 0.1, "learning_rate": 4.56575891729257e-06, "loss": 0.9045, "step": 20805 }, { "epoch": 0.1, "learning_rate": 4.564854112510994e-06, "loss": 0.8914, "step": 20806 }, { "epoch": 0.1, "learning_rate": 4.563949370876467e-06, "loss": 0.8202, "step": 20807 }, { "epoch": 0.1, "learning_rate": 4.563044692399489e-06, "loss": 0.8416, "step": 20808 }, { "epoch": 0.1, "learning_rate": 4.562140077090581e-06, "loss": 0.8673, "step": 20809 }, { "epoch": 0.1, "learning_rate": 4.5612355249602535e-06, "loss": 0.8834, "step": 20810 }, { "epoch": 0.1, "learning_rate": 4.560331036019009e-06, "loss": 0.8481, "step": 20811 }, { "epoch": 0.1, "learning_rate": 4.559426610277359e-06, "loss": 0.8874, "step": 20812 }, { "epoch": 0.1, "learning_rate": 4.558522247745812e-06, "loss": 0.8909, "step": 20813 }, { "epoch": 0.1, "learning_rate": 4.557617948434875e-06, "loss": 0.9011, "step": 20814 }, { "epoch": 0.1, "learning_rate": 4.5567137123550556e-06, "loss": 0.8789, "step": 20815 }, { "epoch": 0.1, "learning_rate": 4.555809539516857e-06, "loss": 0.8806, "step": 20816 }, { "epoch": 0.1, "learning_rate": 4.554905429930788e-06, "loss": 0.8847, "step": 20817 }, { "epoch": 0.1, "learning_rate": 4.5540013836073484e-06, "loss": 0.869, "step": 20818 }, { "epoch": 0.1, "learning_rate": 4.553097400557043e-06, "loss": 0.8634, "step": 20819 }, { "epoch": 0.1, "learning_rate": 4.552193480790374e-06, "loss": 0.9013, "step": 20820 }, { "epoch": 0.1, "learning_rate": 4.551289624317845e-06, "loss": 0.9056, "step": 20821 }, { "epoch": 0.1, "learning_rate": 4.5503858311499575e-06, "loss": 0.8677, "step": 20822 }, { "epoch": 0.1, "learning_rate": 4.54948210129721e-06, "loss": 0.8907, "step": 20823 }, { "epoch": 0.1, "learning_rate": 4.548578434770108e-06, "loss": 0.9059, "step": 20824 }, { "epoch": 0.1, "learning_rate": 4.5476748315791405e-06, "loss": 0.8647, "step": 20825 }, { "epoch": 0.1, "learning_rate": 4.546771291734809e-06, "loss": 0.8401, "step": 20826 }, { "epoch": 0.1, "learning_rate": 4.545867815247616e-06, "loss": 0.8609, "step": 20827 }, { "epoch": 0.1, "learning_rate": 4.544964402128061e-06, "loss": 0.867, "step": 20828 }, { "epoch": 0.1, "learning_rate": 4.54406105238663e-06, "loss": 0.8845, "step": 20829 }, { "epoch": 0.1, "learning_rate": 4.543157766033824e-06, "loss": 0.857, "step": 20830 }, { "epoch": 0.1, "learning_rate": 4.542254543080137e-06, "loss": 0.843, "step": 20831 }, { "epoch": 0.1, "learning_rate": 4.5413513835360615e-06, "loss": 0.8562, "step": 20832 }, { "epoch": 0.1, "learning_rate": 4.540448287412093e-06, "loss": 0.8525, "step": 20833 }, { "epoch": 0.1, "learning_rate": 4.539545254718721e-06, "loss": 0.8885, "step": 20834 }, { "epoch": 0.1, "learning_rate": 4.538642285466445e-06, "loss": 0.8852, "step": 20835 }, { "epoch": 0.1, "learning_rate": 4.537739379665745e-06, "loss": 0.853, "step": 20836 }, { "epoch": 0.1, "learning_rate": 4.536836537327116e-06, "loss": 0.8845, "step": 20837 }, { "epoch": 0.1, "learning_rate": 4.5359337584610476e-06, "loss": 0.8831, "step": 20838 }, { "epoch": 0.1, "learning_rate": 4.5350310430780285e-06, "loss": 0.8603, "step": 20839 }, { "epoch": 0.1, "learning_rate": 4.534128391188548e-06, "loss": 0.8852, "step": 20840 }, { "epoch": 0.1, "learning_rate": 4.533225802803091e-06, "loss": 0.8435, "step": 20841 }, { "epoch": 0.1, "learning_rate": 4.532323277932149e-06, "loss": 0.8716, "step": 20842 }, { "epoch": 0.1, "learning_rate": 4.531420816586201e-06, "loss": 0.8434, "step": 20843 }, { "epoch": 0.1, "learning_rate": 4.530518418775734e-06, "loss": 0.8616, "step": 20844 }, { "epoch": 0.1, "learning_rate": 4.529616084511229e-06, "loss": 0.8723, "step": 20845 }, { "epoch": 0.1, "learning_rate": 4.5287138138031824e-06, "loss": 0.8755, "step": 20846 }, { "epoch": 0.1, "learning_rate": 4.527811606662065e-06, "loss": 0.8556, "step": 20847 }, { "epoch": 0.1, "learning_rate": 4.526909463098362e-06, "loss": 0.888, "step": 20848 }, { "epoch": 0.1, "learning_rate": 4.526007383122558e-06, "loss": 0.889, "step": 20849 }, { "epoch": 0.1, "learning_rate": 4.525105366745128e-06, "loss": 0.8875, "step": 20850 }, { "epoch": 0.1, "learning_rate": 4.5242034139765504e-06, "loss": 0.8451, "step": 20851 }, { "epoch": 0.1, "learning_rate": 4.523301524827313e-06, "loss": 0.8401, "step": 20852 }, { "epoch": 0.1, "learning_rate": 4.522399699307892e-06, "loss": 0.8706, "step": 20853 }, { "epoch": 0.1, "learning_rate": 4.521497937428761e-06, "loss": 0.8809, "step": 20854 }, { "epoch": 0.1, "learning_rate": 4.520596239200398e-06, "loss": 0.8247, "step": 20855 }, { "epoch": 0.1, "learning_rate": 4.51969460463328e-06, "loss": 0.8733, "step": 20856 }, { "epoch": 0.1, "learning_rate": 4.518793033737881e-06, "loss": 0.8461, "step": 20857 }, { "epoch": 0.1, "learning_rate": 4.517891526524679e-06, "loss": 0.8653, "step": 20858 }, { "epoch": 0.1, "learning_rate": 4.516990083004146e-06, "loss": 0.8808, "step": 20859 }, { "epoch": 0.1, "learning_rate": 4.516088703186758e-06, "loss": 0.8592, "step": 20860 }, { "epoch": 0.1, "learning_rate": 4.515187387082982e-06, "loss": 0.8871, "step": 20861 }, { "epoch": 0.1, "learning_rate": 4.5142861347032915e-06, "loss": 0.8625, "step": 20862 }, { "epoch": 0.1, "learning_rate": 4.513384946058159e-06, "loss": 0.8527, "step": 20863 }, { "epoch": 0.1, "learning_rate": 4.512483821158055e-06, "loss": 0.8643, "step": 20864 }, { "epoch": 0.1, "learning_rate": 4.511582760013447e-06, "loss": 0.8704, "step": 20865 }, { "epoch": 0.1, "learning_rate": 4.510681762634805e-06, "loss": 0.8429, "step": 20866 }, { "epoch": 0.1, "learning_rate": 4.509780829032602e-06, "loss": 0.8825, "step": 20867 }, { "epoch": 0.1, "learning_rate": 4.508879959217295e-06, "loss": 0.8494, "step": 20868 }, { "epoch": 0.1, "learning_rate": 4.507979153199353e-06, "loss": 0.8589, "step": 20869 }, { "epoch": 0.1, "learning_rate": 4.507078410989253e-06, "loss": 0.9166, "step": 20870 }, { "epoch": 0.1, "learning_rate": 4.506177732597447e-06, "loss": 0.8449, "step": 20871 }, { "epoch": 0.1, "learning_rate": 4.505277118034406e-06, "loss": 0.8691, "step": 20872 }, { "epoch": 0.1, "learning_rate": 4.50437656731059e-06, "loss": 0.8551, "step": 20873 }, { "epoch": 0.1, "learning_rate": 4.503476080436468e-06, "loss": 0.9254, "step": 20874 }, { "epoch": 0.1, "learning_rate": 4.50257565742249e-06, "loss": 0.8224, "step": 20875 }, { "epoch": 0.1, "learning_rate": 4.501675298279129e-06, "loss": 0.904, "step": 20876 }, { "epoch": 0.1, "learning_rate": 4.500775003016842e-06, "loss": 0.9, "step": 20877 }, { "epoch": 0.1, "learning_rate": 4.499874771646092e-06, "loss": 0.8254, "step": 20878 }, { "epoch": 0.1, "learning_rate": 4.498974604177331e-06, "loss": 0.8869, "step": 20879 }, { "epoch": 0.1, "learning_rate": 4.498074500621021e-06, "loss": 0.8564, "step": 20880 }, { "epoch": 0.1, "learning_rate": 4.49717446098762e-06, "loss": 0.867, "step": 20881 }, { "epoch": 0.1, "learning_rate": 4.496274485287585e-06, "loss": 0.9022, "step": 20882 }, { "epoch": 0.1, "learning_rate": 4.495374573531371e-06, "loss": 0.8285, "step": 20883 }, { "epoch": 0.1, "learning_rate": 4.4944747257294356e-06, "loss": 0.8695, "step": 20884 }, { "epoch": 0.1, "learning_rate": 4.493574941892236e-06, "loss": 0.8783, "step": 20885 }, { "epoch": 0.1, "learning_rate": 4.492675222030217e-06, "loss": 0.8538, "step": 20886 }, { "epoch": 0.1, "learning_rate": 4.4917755661538386e-06, "loss": 0.8409, "step": 20887 }, { "epoch": 0.1, "learning_rate": 4.490875974273551e-06, "loss": 0.8519, "step": 20888 }, { "epoch": 0.1, "learning_rate": 4.489976446399806e-06, "loss": 0.8735, "step": 20889 }, { "epoch": 0.1, "learning_rate": 4.489076982543057e-06, "loss": 0.926, "step": 20890 }, { "epoch": 0.1, "learning_rate": 4.4881775827137505e-06, "loss": 0.8596, "step": 20891 }, { "epoch": 0.1, "learning_rate": 4.487278246922343e-06, "loss": 0.8688, "step": 20892 }, { "epoch": 0.1, "learning_rate": 4.486378975179274e-06, "loss": 0.8456, "step": 20893 }, { "epoch": 0.1, "learning_rate": 4.485479767494992e-06, "loss": 0.8572, "step": 20894 }, { "epoch": 0.1, "learning_rate": 4.484580623879956e-06, "loss": 0.8305, "step": 20895 }, { "epoch": 0.1, "learning_rate": 4.4836815443446e-06, "loss": 0.8705, "step": 20896 }, { "epoch": 0.1, "learning_rate": 4.482782528899375e-06, "loss": 0.8644, "step": 20897 }, { "epoch": 0.1, "learning_rate": 4.481883577554724e-06, "loss": 0.8303, "step": 20898 }, { "epoch": 0.1, "learning_rate": 4.480984690321094e-06, "loss": 0.8568, "step": 20899 }, { "epoch": 0.1, "learning_rate": 4.480085867208927e-06, "loss": 0.8349, "step": 20900 }, { "epoch": 0.1, "learning_rate": 4.479187108228664e-06, "loss": 0.8689, "step": 20901 }, { "epoch": 0.1, "learning_rate": 4.478288413390753e-06, "loss": 0.8578, "step": 20902 }, { "epoch": 0.1, "learning_rate": 4.477389782705628e-06, "loss": 0.901, "step": 20903 }, { "epoch": 0.1, "learning_rate": 4.476491216183733e-06, "loss": 0.8515, "step": 20904 }, { "epoch": 0.1, "learning_rate": 4.475592713835506e-06, "loss": 0.8867, "step": 20905 }, { "epoch": 0.11, "learning_rate": 4.474694275671388e-06, "loss": 0.9327, "step": 20906 }, { "epoch": 0.11, "learning_rate": 4.473795901701817e-06, "loss": 0.859, "step": 20907 }, { "epoch": 0.11, "learning_rate": 4.472897591937231e-06, "loss": 0.8583, "step": 20908 }, { "epoch": 0.11, "learning_rate": 4.4719993463880695e-06, "loss": 0.851, "step": 20909 }, { "epoch": 0.11, "learning_rate": 4.471101165064761e-06, "loss": 0.8868, "step": 20910 }, { "epoch": 0.11, "learning_rate": 4.470203047977746e-06, "loss": 0.8626, "step": 20911 }, { "epoch": 0.11, "learning_rate": 4.469304995137457e-06, "loss": 0.8625, "step": 20912 }, { "epoch": 0.11, "learning_rate": 4.468407006554329e-06, "loss": 0.8685, "step": 20913 }, { "epoch": 0.11, "learning_rate": 4.467509082238795e-06, "loss": 0.8702, "step": 20914 }, { "epoch": 0.11, "learning_rate": 4.466611222201289e-06, "loss": 0.8632, "step": 20915 }, { "epoch": 0.11, "learning_rate": 4.465713426452244e-06, "loss": 0.8482, "step": 20916 }, { "epoch": 0.11, "learning_rate": 4.464815695002083e-06, "loss": 0.8472, "step": 20917 }, { "epoch": 0.11, "learning_rate": 4.4639180278612385e-06, "loss": 0.8861, "step": 20918 }, { "epoch": 0.11, "learning_rate": 4.4630204250401444e-06, "loss": 0.8379, "step": 20919 }, { "epoch": 0.11, "learning_rate": 4.462122886549232e-06, "loss": 0.8843, "step": 20920 }, { "epoch": 0.11, "learning_rate": 4.46122541239892e-06, "loss": 0.8992, "step": 20921 }, { "epoch": 0.11, "learning_rate": 4.460328002599641e-06, "loss": 0.8649, "step": 20922 }, { "epoch": 0.11, "learning_rate": 4.459430657161819e-06, "loss": 0.8508, "step": 20923 }, { "epoch": 0.11, "learning_rate": 4.4585333760958804e-06, "loss": 0.8947, "step": 20924 }, { "epoch": 0.11, "learning_rate": 4.457636159412252e-06, "loss": 0.8799, "step": 20925 }, { "epoch": 0.11, "learning_rate": 4.456739007121356e-06, "loss": 0.8377, "step": 20926 }, { "epoch": 0.11, "learning_rate": 4.455841919233619e-06, "loss": 0.8531, "step": 20927 }, { "epoch": 0.11, "learning_rate": 4.454944895759457e-06, "loss": 0.8564, "step": 20928 }, { "epoch": 0.11, "learning_rate": 4.4540479367092955e-06, "loss": 0.8656, "step": 20929 }, { "epoch": 0.11, "learning_rate": 4.4531510420935565e-06, "loss": 0.8771, "step": 20930 }, { "epoch": 0.11, "learning_rate": 4.452254211922658e-06, "loss": 0.8403, "step": 20931 }, { "epoch": 0.11, "learning_rate": 4.4513574462070245e-06, "loss": 0.8376, "step": 20932 }, { "epoch": 0.11, "learning_rate": 4.450460744957069e-06, "loss": 0.8501, "step": 20933 }, { "epoch": 0.11, "learning_rate": 4.449564108183216e-06, "loss": 0.8614, "step": 20934 }, { "epoch": 0.11, "learning_rate": 4.448667535895876e-06, "loss": 0.8128, "step": 20935 }, { "epoch": 0.11, "learning_rate": 4.447771028105469e-06, "loss": 0.8893, "step": 20936 }, { "epoch": 0.11, "learning_rate": 4.446874584822406e-06, "loss": 0.8364, "step": 20937 }, { "epoch": 0.11, "learning_rate": 4.445978206057116e-06, "loss": 0.8827, "step": 20938 }, { "epoch": 0.11, "learning_rate": 4.445081891819998e-06, "loss": 0.8205, "step": 20939 }, { "epoch": 0.11, "learning_rate": 4.444185642121472e-06, "loss": 0.8384, "step": 20940 }, { "epoch": 0.11, "learning_rate": 4.443289456971956e-06, "loss": 0.8463, "step": 20941 }, { "epoch": 0.11, "learning_rate": 4.442393336381848e-06, "loss": 0.8688, "step": 20942 }, { "epoch": 0.11, "learning_rate": 4.441497280361573e-06, "loss": 0.8379, "step": 20943 }, { "epoch": 0.11, "learning_rate": 4.440601288921535e-06, "loss": 0.8379, "step": 20944 }, { "epoch": 0.11, "learning_rate": 4.439705362072151e-06, "loss": 0.871, "step": 20945 }, { "epoch": 0.11, "learning_rate": 4.4388094998238204e-06, "loss": 0.8506, "step": 20946 }, { "epoch": 0.11, "learning_rate": 4.4379137021869565e-06, "loss": 0.8681, "step": 20947 }, { "epoch": 0.11, "learning_rate": 4.437017969171966e-06, "loss": 0.8953, "step": 20948 }, { "epoch": 0.11, "learning_rate": 4.436122300789257e-06, "loss": 0.88, "step": 20949 }, { "epoch": 0.11, "learning_rate": 4.435226697049233e-06, "loss": 0.858, "step": 20950 }, { "epoch": 0.11, "learning_rate": 4.434331157962302e-06, "loss": 0.8672, "step": 20951 }, { "epoch": 0.11, "learning_rate": 4.433435683538871e-06, "loss": 0.8271, "step": 20952 }, { "epoch": 0.11, "learning_rate": 4.4325402737893375e-06, "loss": 0.8379, "step": 20953 }, { "epoch": 0.11, "learning_rate": 4.4316449287241085e-06, "loss": 0.8577, "step": 20954 }, { "epoch": 0.11, "learning_rate": 4.430749648353584e-06, "loss": 0.874, "step": 20955 }, { "epoch": 0.11, "learning_rate": 4.429854432688169e-06, "loss": 0.8348, "step": 20956 }, { "epoch": 0.11, "learning_rate": 4.428959281738261e-06, "loss": 0.892, "step": 20957 }, { "epoch": 0.11, "learning_rate": 4.428064195514262e-06, "loss": 0.8549, "step": 20958 }, { "epoch": 0.11, "learning_rate": 4.427169174026576e-06, "loss": 0.873, "step": 20959 }, { "epoch": 0.11, "learning_rate": 4.426274217285592e-06, "loss": 0.8787, "step": 20960 }, { "epoch": 0.11, "learning_rate": 4.425379325301708e-06, "loss": 0.8764, "step": 20961 }, { "epoch": 0.11, "learning_rate": 4.424484498085335e-06, "loss": 0.8913, "step": 20962 }, { "epoch": 0.11, "learning_rate": 4.423589735646856e-06, "loss": 0.9175, "step": 20963 }, { "epoch": 0.11, "learning_rate": 4.4226950379966695e-06, "loss": 0.88, "step": 20964 }, { "epoch": 0.11, "learning_rate": 4.421800405145172e-06, "loss": 0.8441, "step": 20965 }, { "epoch": 0.11, "learning_rate": 4.420905837102761e-06, "loss": 0.9031, "step": 20966 }, { "epoch": 0.11, "learning_rate": 4.420011333879819e-06, "loss": 0.9039, "step": 20967 }, { "epoch": 0.11, "learning_rate": 4.419116895486748e-06, "loss": 0.8705, "step": 20968 }, { "epoch": 0.11, "learning_rate": 4.418222521933938e-06, "loss": 0.8506, "step": 20969 }, { "epoch": 0.11, "learning_rate": 4.417328213231783e-06, "loss": 0.8381, "step": 20970 }, { "epoch": 0.11, "learning_rate": 4.4164339693906665e-06, "loss": 0.8543, "step": 20971 }, { "epoch": 0.11, "learning_rate": 4.4155397904209805e-06, "loss": 0.8283, "step": 20972 }, { "epoch": 0.11, "learning_rate": 4.4146456763331145e-06, "loss": 0.8459, "step": 20973 }, { "epoch": 0.11, "learning_rate": 4.413751627137458e-06, "loss": 0.9102, "step": 20974 }, { "epoch": 0.11, "learning_rate": 4.412857642844396e-06, "loss": 0.8478, "step": 20975 }, { "epoch": 0.11, "learning_rate": 4.4119637234643164e-06, "loss": 0.8957, "step": 20976 }, { "epoch": 0.11, "learning_rate": 4.411069869007608e-06, "loss": 0.8812, "step": 20977 }, { "epoch": 0.11, "learning_rate": 4.410176079484649e-06, "loss": 0.8386, "step": 20978 }, { "epoch": 0.11, "learning_rate": 4.409282354905828e-06, "loss": 0.8522, "step": 20979 }, { "epoch": 0.11, "learning_rate": 4.408388695281527e-06, "loss": 0.867, "step": 20980 }, { "epoch": 0.11, "learning_rate": 4.407495100622129e-06, "loss": 0.8531, "step": 20981 }, { "epoch": 0.11, "learning_rate": 4.406601570938017e-06, "loss": 0.8624, "step": 20982 }, { "epoch": 0.11, "learning_rate": 4.405708106239573e-06, "loss": 0.8278, "step": 20983 }, { "epoch": 0.11, "learning_rate": 4.4048147065371785e-06, "loss": 0.8986, "step": 20984 }, { "epoch": 0.11, "learning_rate": 4.403921371841209e-06, "loss": 0.872, "step": 20985 }, { "epoch": 0.11, "learning_rate": 4.4030281021620405e-06, "loss": 0.8569, "step": 20986 }, { "epoch": 0.11, "learning_rate": 4.402134897510064e-06, "loss": 0.8461, "step": 20987 }, { "epoch": 0.11, "learning_rate": 4.401241757895647e-06, "loss": 0.8971, "step": 20988 }, { "epoch": 0.11, "learning_rate": 4.400348683329168e-06, "loss": 0.8752, "step": 20989 }, { "epoch": 0.11, "learning_rate": 4.399455673821004e-06, "loss": 0.8573, "step": 20990 }, { "epoch": 0.11, "learning_rate": 4.398562729381529e-06, "loss": 0.8282, "step": 20991 }, { "epoch": 0.11, "learning_rate": 4.397669850021119e-06, "loss": 0.867, "step": 20992 }, { "epoch": 0.11, "learning_rate": 4.396777035750146e-06, "loss": 0.8669, "step": 20993 }, { "epoch": 0.11, "learning_rate": 4.39588428657899e-06, "loss": 0.8685, "step": 20994 }, { "epoch": 0.11, "learning_rate": 4.3949916025180116e-06, "loss": 0.8434, "step": 20995 }, { "epoch": 0.11, "learning_rate": 4.394098983577589e-06, "loss": 0.863, "step": 20996 }, { "epoch": 0.11, "learning_rate": 4.393206429768091e-06, "loss": 0.8508, "step": 20997 }, { "epoch": 0.11, "learning_rate": 4.392313941099888e-06, "loss": 0.8768, "step": 20998 }, { "epoch": 0.11, "learning_rate": 4.391421517583351e-06, "loss": 0.8787, "step": 20999 }, { "epoch": 0.11, "learning_rate": 4.3905291592288464e-06, "loss": 0.843, "step": 21000 }, { "epoch": 0.11, "learning_rate": 4.389636866046746e-06, "loss": 0.8858, "step": 21001 }, { "epoch": 0.11, "learning_rate": 4.38874463804741e-06, "loss": 0.9124, "step": 21002 }, { "epoch": 0.11, "learning_rate": 4.387852475241207e-06, "loss": 0.8907, "step": 21003 }, { "epoch": 0.11, "learning_rate": 4.3869603776385e-06, "loss": 0.893, "step": 21004 }, { "epoch": 0.11, "learning_rate": 4.386068345249663e-06, "loss": 0.8473, "step": 21005 }, { "epoch": 0.11, "learning_rate": 4.385176378085051e-06, "loss": 0.8617, "step": 21006 }, { "epoch": 0.11, "learning_rate": 4.38428447615503e-06, "loss": 0.9033, "step": 21007 }, { "epoch": 0.11, "learning_rate": 4.383392639469965e-06, "loss": 0.8716, "step": 21008 }, { "epoch": 0.11, "learning_rate": 4.3825008680402105e-06, "loss": 0.8636, "step": 21009 }, { "epoch": 0.11, "learning_rate": 4.38160916187613e-06, "loss": 0.8545, "step": 21010 }, { "epoch": 0.11, "learning_rate": 4.380717520988087e-06, "loss": 0.8788, "step": 21011 }, { "epoch": 0.11, "learning_rate": 4.379825945386443e-06, "loss": 0.8895, "step": 21012 }, { "epoch": 0.11, "learning_rate": 4.378934435081549e-06, "loss": 0.8757, "step": 21013 }, { "epoch": 0.11, "learning_rate": 4.378042990083766e-06, "loss": 0.8463, "step": 21014 }, { "epoch": 0.11, "learning_rate": 4.377151610403453e-06, "loss": 0.8596, "step": 21015 }, { "epoch": 0.11, "learning_rate": 4.3762602960509625e-06, "loss": 0.8154, "step": 21016 }, { "epoch": 0.11, "learning_rate": 4.375369047036654e-06, "loss": 0.7997, "step": 21017 }, { "epoch": 0.11, "learning_rate": 4.374477863370881e-06, "loss": 0.9018, "step": 21018 }, { "epoch": 0.11, "learning_rate": 4.373586745063999e-06, "loss": 0.8714, "step": 21019 }, { "epoch": 0.11, "learning_rate": 4.372695692126357e-06, "loss": 0.8725, "step": 21020 }, { "epoch": 0.11, "learning_rate": 4.371804704568309e-06, "loss": 0.8754, "step": 21021 }, { "epoch": 0.11, "learning_rate": 4.370913782400208e-06, "loss": 0.8627, "step": 21022 }, { "epoch": 0.11, "learning_rate": 4.370022925632405e-06, "loss": 0.8461, "step": 21023 }, { "epoch": 0.11, "learning_rate": 4.369132134275248e-06, "loss": 0.8402, "step": 21024 }, { "epoch": 0.11, "learning_rate": 4.36824140833909e-06, "loss": 0.8665, "step": 21025 }, { "epoch": 0.11, "learning_rate": 4.367350747834282e-06, "loss": 0.9088, "step": 21026 }, { "epoch": 0.11, "learning_rate": 4.366460152771162e-06, "loss": 0.8843, "step": 21027 }, { "epoch": 0.11, "learning_rate": 4.365569623160086e-06, "loss": 0.8817, "step": 21028 }, { "epoch": 0.11, "learning_rate": 4.364679159011391e-06, "loss": 0.8275, "step": 21029 }, { "epoch": 0.11, "learning_rate": 4.363788760335438e-06, "loss": 0.8156, "step": 21030 }, { "epoch": 0.11, "learning_rate": 4.36289842714256e-06, "loss": 0.8801, "step": 21031 }, { "epoch": 0.11, "learning_rate": 4.362008159443103e-06, "loss": 0.8678, "step": 21032 }, { "epoch": 0.11, "learning_rate": 4.361117957247416e-06, "loss": 0.8837, "step": 21033 }, { "epoch": 0.11, "learning_rate": 4.36022782056583e-06, "loss": 0.8713, "step": 21034 }, { "epoch": 0.11, "learning_rate": 4.359337749408696e-06, "loss": 0.8733, "step": 21035 }, { "epoch": 0.11, "learning_rate": 4.358447743786356e-06, "loss": 0.8565, "step": 21036 }, { "epoch": 0.11, "learning_rate": 4.3575578037091495e-06, "loss": 0.8287, "step": 21037 }, { "epoch": 0.11, "learning_rate": 4.3566679291874104e-06, "loss": 0.8641, "step": 21038 }, { "epoch": 0.11, "learning_rate": 4.355778120231481e-06, "loss": 0.8536, "step": 21039 }, { "epoch": 0.11, "learning_rate": 4.3548883768517e-06, "loss": 0.857, "step": 21040 }, { "epoch": 0.11, "learning_rate": 4.3539986990584035e-06, "loss": 0.9026, "step": 21041 }, { "epoch": 0.11, "learning_rate": 4.35310908686193e-06, "loss": 0.8879, "step": 21042 }, { "epoch": 0.11, "learning_rate": 4.352219540272613e-06, "loss": 0.8657, "step": 21043 }, { "epoch": 0.11, "learning_rate": 4.351330059300793e-06, "loss": 0.8785, "step": 21044 }, { "epoch": 0.11, "learning_rate": 4.350440643956796e-06, "loss": 0.8773, "step": 21045 }, { "epoch": 0.11, "learning_rate": 4.3495512942509585e-06, "loss": 0.8592, "step": 21046 }, { "epoch": 0.11, "learning_rate": 4.3486620101936135e-06, "loss": 0.8788, "step": 21047 }, { "epoch": 0.11, "learning_rate": 4.347772791795095e-06, "loss": 0.8848, "step": 21048 }, { "epoch": 0.11, "learning_rate": 4.346883639065732e-06, "loss": 0.8731, "step": 21049 }, { "epoch": 0.11, "learning_rate": 4.345994552015855e-06, "loss": 0.8692, "step": 21050 }, { "epoch": 0.11, "learning_rate": 4.345105530655798e-06, "loss": 0.8565, "step": 21051 }, { "epoch": 0.11, "learning_rate": 4.344216574995882e-06, "loss": 0.8873, "step": 21052 }, { "epoch": 0.11, "learning_rate": 4.343327685046437e-06, "loss": 0.8851, "step": 21053 }, { "epoch": 0.11, "learning_rate": 4.342438860817795e-06, "loss": 0.8765, "step": 21054 }, { "epoch": 0.11, "learning_rate": 4.341550102320283e-06, "loss": 0.8544, "step": 21055 }, { "epoch": 0.11, "learning_rate": 4.340661409564222e-06, "loss": 0.8907, "step": 21056 }, { "epoch": 0.11, "learning_rate": 4.339772782559939e-06, "loss": 0.8873, "step": 21057 }, { "epoch": 0.11, "learning_rate": 4.338884221317761e-06, "loss": 0.8808, "step": 21058 }, { "epoch": 0.11, "learning_rate": 4.337995725848002e-06, "loss": 0.89, "step": 21059 }, { "epoch": 0.11, "learning_rate": 4.337107296160996e-06, "loss": 0.8485, "step": 21060 }, { "epoch": 0.11, "learning_rate": 4.336218932267061e-06, "loss": 0.8427, "step": 21061 }, { "epoch": 0.11, "learning_rate": 4.3353306341765214e-06, "loss": 0.8784, "step": 21062 }, { "epoch": 0.11, "learning_rate": 4.33444240189969e-06, "loss": 0.8649, "step": 21063 }, { "epoch": 0.11, "learning_rate": 4.333554235446892e-06, "loss": 0.8701, "step": 21064 }, { "epoch": 0.11, "learning_rate": 4.332666134828444e-06, "loss": 0.9, "step": 21065 }, { "epoch": 0.11, "learning_rate": 4.331778100054665e-06, "loss": 0.8947, "step": 21066 }, { "epoch": 0.11, "learning_rate": 4.330890131135875e-06, "loss": 0.8359, "step": 21067 }, { "epoch": 0.11, "learning_rate": 4.330002228082388e-06, "loss": 0.8426, "step": 21068 }, { "epoch": 0.11, "learning_rate": 4.329114390904523e-06, "loss": 0.8445, "step": 21069 }, { "epoch": 0.11, "learning_rate": 4.328226619612589e-06, "loss": 0.8834, "step": 21070 }, { "epoch": 0.11, "learning_rate": 4.327338914216904e-06, "loss": 0.8351, "step": 21071 }, { "epoch": 0.11, "learning_rate": 4.3264512747277835e-06, "loss": 0.8414, "step": 21072 }, { "epoch": 0.11, "learning_rate": 4.3255637011555365e-06, "loss": 0.8559, "step": 21073 }, { "epoch": 0.11, "learning_rate": 4.324676193510478e-06, "loss": 0.883, "step": 21074 }, { "epoch": 0.11, "learning_rate": 4.3237887518029165e-06, "loss": 0.8526, "step": 21075 }, { "epoch": 0.11, "learning_rate": 4.3229013760431695e-06, "loss": 0.8527, "step": 21076 }, { "epoch": 0.11, "learning_rate": 4.322014066241534e-06, "loss": 0.879, "step": 21077 }, { "epoch": 0.11, "learning_rate": 4.32112682240833e-06, "loss": 0.8757, "step": 21078 }, { "epoch": 0.11, "learning_rate": 4.320239644553867e-06, "loss": 0.8403, "step": 21079 }, { "epoch": 0.11, "learning_rate": 4.319352532688444e-06, "loss": 0.8577, "step": 21080 }, { "epoch": 0.11, "learning_rate": 4.3184654868223705e-06, "loss": 0.8853, "step": 21081 }, { "epoch": 0.11, "learning_rate": 4.317578506965955e-06, "loss": 0.8583, "step": 21082 }, { "epoch": 0.11, "learning_rate": 4.3166915931295e-06, "loss": 0.8076, "step": 21083 }, { "epoch": 0.11, "learning_rate": 4.3158047453233125e-06, "loss": 0.8602, "step": 21084 }, { "epoch": 0.11, "learning_rate": 4.314917963557693e-06, "loss": 0.8685, "step": 21085 }, { "epoch": 0.11, "learning_rate": 4.314031247842951e-06, "loss": 0.8525, "step": 21086 }, { "epoch": 0.11, "learning_rate": 4.313144598189379e-06, "loss": 0.8494, "step": 21087 }, { "epoch": 0.11, "learning_rate": 4.312258014607284e-06, "loss": 0.8328, "step": 21088 }, { "epoch": 0.11, "learning_rate": 4.311371497106966e-06, "loss": 0.8812, "step": 21089 }, { "epoch": 0.11, "learning_rate": 4.310485045698725e-06, "loss": 0.919, "step": 21090 }, { "epoch": 0.11, "learning_rate": 4.309598660392859e-06, "loss": 0.8404, "step": 21091 }, { "epoch": 0.11, "learning_rate": 4.308712341199666e-06, "loss": 0.8614, "step": 21092 }, { "epoch": 0.11, "learning_rate": 4.307826088129447e-06, "loss": 0.8968, "step": 21093 }, { "epoch": 0.11, "learning_rate": 4.306939901192495e-06, "loss": 0.8767, "step": 21094 }, { "epoch": 0.11, "learning_rate": 4.306053780399106e-06, "loss": 0.8462, "step": 21095 }, { "epoch": 0.11, "learning_rate": 4.305167725759572e-06, "loss": 0.8647, "step": 21096 }, { "epoch": 0.11, "learning_rate": 4.3042817372842e-06, "loss": 0.883, "step": 21097 }, { "epoch": 0.11, "learning_rate": 4.30339581498327e-06, "loss": 0.9122, "step": 21098 }, { "epoch": 0.11, "learning_rate": 4.302509958867081e-06, "loss": 0.8659, "step": 21099 }, { "epoch": 0.11, "learning_rate": 4.301624168945925e-06, "loss": 0.8634, "step": 21100 }, { "epoch": 0.11, "learning_rate": 4.300738445230096e-06, "loss": 0.9047, "step": 21101 }, { "epoch": 0.11, "learning_rate": 4.299852787729873e-06, "loss": 0.8503, "step": 21102 }, { "epoch": 0.11, "learning_rate": 4.298967196455559e-06, "loss": 0.8792, "step": 21103 }, { "epoch": 0.11, "learning_rate": 4.298081671417442e-06, "loss": 0.8657, "step": 21104 }, { "epoch": 0.11, "learning_rate": 4.297196212625802e-06, "loss": 0.8519, "step": 21105 }, { "epoch": 0.11, "learning_rate": 4.296310820090931e-06, "loss": 0.8602, "step": 21106 }, { "epoch": 0.11, "learning_rate": 4.295425493823116e-06, "loss": 0.8687, "step": 21107 }, { "epoch": 0.11, "learning_rate": 4.294540233832642e-06, "loss": 0.8354, "step": 21108 }, { "epoch": 0.11, "learning_rate": 4.293655040129795e-06, "loss": 0.8564, "step": 21109 }, { "epoch": 0.11, "learning_rate": 4.292769912724859e-06, "loss": 0.9087, "step": 21110 }, { "epoch": 0.11, "learning_rate": 4.291884851628122e-06, "loss": 0.8549, "step": 21111 }, { "epoch": 0.11, "learning_rate": 4.290999856849859e-06, "loss": 0.824, "step": 21112 }, { "epoch": 0.11, "learning_rate": 4.290114928400357e-06, "loss": 0.9052, "step": 21113 }, { "epoch": 0.11, "learning_rate": 4.2892300662898944e-06, "loss": 0.8811, "step": 21114 }, { "epoch": 0.11, "learning_rate": 4.288345270528755e-06, "loss": 0.8819, "step": 21115 }, { "epoch": 0.11, "learning_rate": 4.287460541127217e-06, "loss": 0.8978, "step": 21116 }, { "epoch": 0.11, "learning_rate": 4.286575878095558e-06, "loss": 0.8072, "step": 21117 }, { "epoch": 0.11, "learning_rate": 4.285691281444063e-06, "loss": 0.8836, "step": 21118 }, { "epoch": 0.11, "learning_rate": 4.284806751183e-06, "loss": 0.8716, "step": 21119 }, { "epoch": 0.11, "learning_rate": 4.283922287322652e-06, "loss": 0.897, "step": 21120 }, { "epoch": 0.11, "learning_rate": 4.2830378898732884e-06, "loss": 0.8659, "step": 21121 }, { "epoch": 0.11, "learning_rate": 4.2821535588451955e-06, "loss": 0.9017, "step": 21122 }, { "epoch": 0.11, "learning_rate": 4.281269294248638e-06, "loss": 0.8655, "step": 21123 }, { "epoch": 0.11, "learning_rate": 4.2803850960938935e-06, "loss": 0.8662, "step": 21124 }, { "epoch": 0.11, "learning_rate": 4.279500964391238e-06, "loss": 0.8859, "step": 21125 }, { "epoch": 0.11, "learning_rate": 4.2786168991509315e-06, "loss": 0.8806, "step": 21126 }, { "epoch": 0.11, "learning_rate": 4.277732900383258e-06, "loss": 0.8226, "step": 21127 }, { "epoch": 0.11, "learning_rate": 4.276848968098483e-06, "loss": 0.8806, "step": 21128 }, { "epoch": 0.11, "learning_rate": 4.27596510230688e-06, "loss": 0.8318, "step": 21129 }, { "epoch": 0.11, "learning_rate": 4.275081303018712e-06, "loss": 0.8434, "step": 21130 }, { "epoch": 0.11, "learning_rate": 4.27419757024425e-06, "loss": 0.8664, "step": 21131 }, { "epoch": 0.11, "learning_rate": 4.273313903993762e-06, "loss": 0.8437, "step": 21132 }, { "epoch": 0.11, "learning_rate": 4.272430304277514e-06, "loss": 0.8842, "step": 21133 }, { "epoch": 0.11, "learning_rate": 4.2715467711057724e-06, "loss": 0.8748, "step": 21134 }, { "epoch": 0.11, "learning_rate": 4.270663304488801e-06, "loss": 0.88, "step": 21135 }, { "epoch": 0.11, "learning_rate": 4.26977990443687e-06, "loss": 0.8562, "step": 21136 }, { "epoch": 0.11, "learning_rate": 4.268896570960235e-06, "loss": 0.8735, "step": 21137 }, { "epoch": 0.11, "learning_rate": 4.268013304069161e-06, "loss": 0.8772, "step": 21138 }, { "epoch": 0.11, "learning_rate": 4.267130103773911e-06, "loss": 0.8478, "step": 21139 }, { "epoch": 0.11, "learning_rate": 4.266246970084747e-06, "loss": 0.8263, "step": 21140 }, { "epoch": 0.11, "learning_rate": 4.265363903011929e-06, "loss": 0.902, "step": 21141 }, { "epoch": 0.11, "learning_rate": 4.264480902565716e-06, "loss": 0.8256, "step": 21142 }, { "epoch": 0.11, "learning_rate": 4.263597968756372e-06, "loss": 0.8492, "step": 21143 }, { "epoch": 0.11, "learning_rate": 4.262715101594146e-06, "loss": 0.8397, "step": 21144 }, { "epoch": 0.11, "learning_rate": 4.261832301089297e-06, "loss": 0.8923, "step": 21145 }, { "epoch": 0.11, "learning_rate": 4.260949567252088e-06, "loss": 0.8967, "step": 21146 }, { "epoch": 0.11, "learning_rate": 4.260066900092776e-06, "loss": 0.8816, "step": 21147 }, { "epoch": 0.11, "learning_rate": 4.259184299621608e-06, "loss": 0.8653, "step": 21148 }, { "epoch": 0.11, "learning_rate": 4.258301765848841e-06, "loss": 0.8551, "step": 21149 }, { "epoch": 0.11, "learning_rate": 4.257419298784731e-06, "loss": 0.86, "step": 21150 }, { "epoch": 0.11, "learning_rate": 4.2565368984395276e-06, "loss": 0.8474, "step": 21151 }, { "epoch": 0.11, "learning_rate": 4.255654564823485e-06, "loss": 0.8519, "step": 21152 }, { "epoch": 0.11, "learning_rate": 4.254772297946854e-06, "loss": 0.912, "step": 21153 }, { "epoch": 0.11, "learning_rate": 4.253890097819888e-06, "loss": 0.8505, "step": 21154 }, { "epoch": 0.11, "learning_rate": 4.25300796445283e-06, "loss": 0.8088, "step": 21155 }, { "epoch": 0.11, "learning_rate": 4.2521258978559324e-06, "loss": 0.8777, "step": 21156 }, { "epoch": 0.11, "learning_rate": 4.251243898039444e-06, "loss": 0.8803, "step": 21157 }, { "epoch": 0.11, "learning_rate": 4.250361965013609e-06, "loss": 0.8651, "step": 21158 }, { "epoch": 0.11, "learning_rate": 4.2494800987886785e-06, "loss": 0.8532, "step": 21159 }, { "epoch": 0.11, "learning_rate": 4.248598299374895e-06, "loss": 0.8647, "step": 21160 }, { "epoch": 0.11, "learning_rate": 4.2477165667825095e-06, "loss": 0.8867, "step": 21161 }, { "epoch": 0.11, "learning_rate": 4.246834901021757e-06, "loss": 0.8631, "step": 21162 }, { "epoch": 0.11, "learning_rate": 4.245953302102884e-06, "loss": 0.8785, "step": 21163 }, { "epoch": 0.11, "learning_rate": 4.245071770036136e-06, "loss": 0.9077, "step": 21164 }, { "epoch": 0.11, "learning_rate": 4.244190304831752e-06, "loss": 0.8673, "step": 21165 }, { "epoch": 0.11, "learning_rate": 4.243308906499975e-06, "loss": 0.8772, "step": 21166 }, { "epoch": 0.11, "learning_rate": 4.242427575051046e-06, "loss": 0.8772, "step": 21167 }, { "epoch": 0.11, "learning_rate": 4.241546310495205e-06, "loss": 0.8822, "step": 21168 }, { "epoch": 0.11, "learning_rate": 4.240665112842682e-06, "loss": 0.9059, "step": 21169 }, { "epoch": 0.11, "learning_rate": 4.239783982103727e-06, "loss": 0.8608, "step": 21170 }, { "epoch": 0.11, "learning_rate": 4.238902918288575e-06, "loss": 0.8559, "step": 21171 }, { "epoch": 0.11, "learning_rate": 4.238021921407456e-06, "loss": 0.8975, "step": 21172 }, { "epoch": 0.11, "learning_rate": 4.237140991470609e-06, "loss": 0.8688, "step": 21173 }, { "epoch": 0.11, "learning_rate": 4.23626012848827e-06, "loss": 0.8793, "step": 21174 }, { "epoch": 0.11, "learning_rate": 4.235379332470672e-06, "loss": 0.8364, "step": 21175 }, { "epoch": 0.11, "learning_rate": 4.234498603428048e-06, "loss": 0.8448, "step": 21176 }, { "epoch": 0.11, "learning_rate": 4.233617941370632e-06, "loss": 0.8765, "step": 21177 }, { "epoch": 0.11, "learning_rate": 4.232737346308659e-06, "loss": 0.898, "step": 21178 }, { "epoch": 0.11, "learning_rate": 4.231856818252352e-06, "loss": 0.8615, "step": 21179 }, { "epoch": 0.11, "learning_rate": 4.230976357211946e-06, "loss": 0.8897, "step": 21180 }, { "epoch": 0.11, "learning_rate": 4.230095963197668e-06, "loss": 0.8918, "step": 21181 }, { "epoch": 0.11, "learning_rate": 4.2292156362197505e-06, "loss": 0.874, "step": 21182 }, { "epoch": 0.11, "learning_rate": 4.2283353762884175e-06, "loss": 0.8443, "step": 21183 }, { "epoch": 0.11, "learning_rate": 4.227455183413899e-06, "loss": 0.8712, "step": 21184 }, { "epoch": 0.11, "learning_rate": 4.226575057606422e-06, "loss": 0.9041, "step": 21185 }, { "epoch": 0.11, "learning_rate": 4.225694998876209e-06, "loss": 0.8683, "step": 21186 }, { "epoch": 0.11, "learning_rate": 4.224815007233485e-06, "loss": 0.8365, "step": 21187 }, { "epoch": 0.11, "learning_rate": 4.22393508268847e-06, "loss": 0.8324, "step": 21188 }, { "epoch": 0.11, "learning_rate": 4.223055225251401e-06, "loss": 0.869, "step": 21189 }, { "epoch": 0.11, "learning_rate": 4.222175434932488e-06, "loss": 0.8894, "step": 21190 }, { "epoch": 0.11, "learning_rate": 4.221295711741956e-06, "loss": 0.8491, "step": 21191 }, { "epoch": 0.11, "learning_rate": 4.220416055690027e-06, "loss": 0.8459, "step": 21192 }, { "epoch": 0.11, "learning_rate": 4.219536466786923e-06, "loss": 0.8979, "step": 21193 }, { "epoch": 0.11, "learning_rate": 4.218656945042854e-06, "loss": 0.8828, "step": 21194 }, { "epoch": 0.11, "learning_rate": 4.217777490468048e-06, "loss": 0.8586, "step": 21195 }, { "epoch": 0.11, "learning_rate": 4.216898103072723e-06, "loss": 0.8651, "step": 21196 }, { "epoch": 0.11, "learning_rate": 4.2160187828670895e-06, "loss": 0.8347, "step": 21197 }, { "epoch": 0.11, "learning_rate": 4.2151395298613675e-06, "loss": 0.8826, "step": 21198 }, { "epoch": 0.11, "learning_rate": 4.214260344065772e-06, "loss": 0.8922, "step": 21199 }, { "epoch": 0.11, "learning_rate": 4.213381225490517e-06, "loss": 0.8522, "step": 21200 }, { "epoch": 0.11, "learning_rate": 4.212502174145817e-06, "loss": 0.8611, "step": 21201 }, { "epoch": 0.11, "learning_rate": 4.211623190041885e-06, "loss": 0.8604, "step": 21202 }, { "epoch": 0.11, "learning_rate": 4.210744273188937e-06, "loss": 0.8582, "step": 21203 }, { "epoch": 0.11, "learning_rate": 4.209865423597177e-06, "loss": 0.8671, "step": 21204 }, { "epoch": 0.11, "learning_rate": 4.208986641276819e-06, "loss": 0.8528, "step": 21205 }, { "epoch": 0.12, "learning_rate": 4.2081079262380744e-06, "loss": 0.8389, "step": 21206 }, { "epoch": 0.12, "learning_rate": 4.207229278491149e-06, "loss": 0.851, "step": 21207 }, { "epoch": 0.12, "learning_rate": 4.206350698046256e-06, "loss": 0.8912, "step": 21208 }, { "epoch": 0.12, "learning_rate": 4.205472184913599e-06, "loss": 0.861, "step": 21209 }, { "epoch": 0.12, "learning_rate": 4.204593739103391e-06, "loss": 0.8498, "step": 21210 }, { "epoch": 0.12, "learning_rate": 4.203715360625829e-06, "loss": 0.884, "step": 21211 }, { "epoch": 0.12, "learning_rate": 4.20283704949112e-06, "loss": 0.8595, "step": 21212 }, { "epoch": 0.12, "learning_rate": 4.201958805709474e-06, "loss": 0.8876, "step": 21213 }, { "epoch": 0.12, "learning_rate": 4.201080629291096e-06, "loss": 0.8842, "step": 21214 }, { "epoch": 0.12, "learning_rate": 4.200202520246179e-06, "loss": 0.8431, "step": 21215 }, { "epoch": 0.12, "learning_rate": 4.199324478584933e-06, "loss": 0.857, "step": 21216 }, { "epoch": 0.12, "learning_rate": 4.198446504317559e-06, "loss": 0.8782, "step": 21217 }, { "epoch": 0.12, "learning_rate": 4.197568597454249e-06, "loss": 0.8263, "step": 21218 }, { "epoch": 0.12, "learning_rate": 4.196690758005214e-06, "loss": 0.8333, "step": 21219 }, { "epoch": 0.12, "learning_rate": 4.195812985980647e-06, "loss": 0.8421, "step": 21220 }, { "epoch": 0.12, "learning_rate": 4.194935281390751e-06, "loss": 0.8357, "step": 21221 }, { "epoch": 0.12, "learning_rate": 4.194057644245718e-06, "loss": 0.8936, "step": 21222 }, { "epoch": 0.12, "learning_rate": 4.193180074555746e-06, "loss": 0.8373, "step": 21223 }, { "epoch": 0.12, "learning_rate": 4.192302572331031e-06, "loss": 0.83, "step": 21224 }, { "epoch": 0.12, "learning_rate": 4.19142513758177e-06, "loss": 0.8639, "step": 21225 }, { "epoch": 0.12, "learning_rate": 4.190547770318154e-06, "loss": 0.8433, "step": 21226 }, { "epoch": 0.12, "learning_rate": 4.189670470550379e-06, "loss": 0.843, "step": 21227 }, { "epoch": 0.12, "learning_rate": 4.188793238288641e-06, "loss": 0.8468, "step": 21228 }, { "epoch": 0.12, "learning_rate": 4.187916073543124e-06, "loss": 0.863, "step": 21229 }, { "epoch": 0.12, "learning_rate": 4.187038976324023e-06, "loss": 0.8452, "step": 21230 }, { "epoch": 0.12, "learning_rate": 4.186161946641528e-06, "loss": 0.8678, "step": 21231 }, { "epoch": 0.12, "learning_rate": 4.18528498450583e-06, "loss": 0.8673, "step": 21232 }, { "epoch": 0.12, "learning_rate": 4.184408089927117e-06, "loss": 0.9046, "step": 21233 }, { "epoch": 0.12, "learning_rate": 4.183531262915576e-06, "loss": 0.8419, "step": 21234 }, { "epoch": 0.12, "learning_rate": 4.182654503481399e-06, "loss": 0.8456, "step": 21235 }, { "epoch": 0.12, "learning_rate": 4.181777811634765e-06, "loss": 0.7956, "step": 21236 }, { "epoch": 0.12, "learning_rate": 4.180901187385859e-06, "loss": 0.8584, "step": 21237 }, { "epoch": 0.12, "learning_rate": 4.180024630744874e-06, "loss": 0.8452, "step": 21238 }, { "epoch": 0.12, "learning_rate": 4.179148141721995e-06, "loss": 0.8462, "step": 21239 }, { "epoch": 0.12, "learning_rate": 4.178271720327396e-06, "loss": 0.8699, "step": 21240 }, { "epoch": 0.12, "learning_rate": 4.177395366571263e-06, "loss": 0.9036, "step": 21241 }, { "epoch": 0.12, "learning_rate": 4.176519080463779e-06, "loss": 0.8869, "step": 21242 }, { "epoch": 0.12, "learning_rate": 4.175642862015126e-06, "loss": 0.878, "step": 21243 }, { "epoch": 0.12, "learning_rate": 4.1747667112354805e-06, "loss": 0.8449, "step": 21244 }, { "epoch": 0.12, "learning_rate": 4.173890628135026e-06, "loss": 0.8354, "step": 21245 }, { "epoch": 0.12, "learning_rate": 4.173014612723941e-06, "loss": 0.8894, "step": 21246 }, { "epoch": 0.12, "learning_rate": 4.1721386650124e-06, "loss": 0.8679, "step": 21247 }, { "epoch": 0.12, "learning_rate": 4.171262785010579e-06, "loss": 0.8845, "step": 21248 }, { "epoch": 0.12, "learning_rate": 4.170386972728658e-06, "loss": 0.8724, "step": 21249 }, { "epoch": 0.12, "learning_rate": 4.1695112281768115e-06, "loss": 0.8712, "step": 21250 }, { "epoch": 0.12, "learning_rate": 4.168635551365214e-06, "loss": 0.8971, "step": 21251 }, { "epoch": 0.12, "learning_rate": 4.167759942304038e-06, "loss": 0.8641, "step": 21252 }, { "epoch": 0.12, "learning_rate": 4.166884401003461e-06, "loss": 0.8661, "step": 21253 }, { "epoch": 0.12, "learning_rate": 4.166008927473649e-06, "loss": 0.8547, "step": 21254 }, { "epoch": 0.12, "learning_rate": 4.165133521724773e-06, "loss": 0.884, "step": 21255 }, { "epoch": 0.12, "learning_rate": 4.164258183767015e-06, "loss": 0.8937, "step": 21256 }, { "epoch": 0.12, "learning_rate": 4.163382913610533e-06, "loss": 0.8771, "step": 21257 }, { "epoch": 0.12, "learning_rate": 4.1625077112655e-06, "loss": 0.8719, "step": 21258 }, { "epoch": 0.12, "learning_rate": 4.161632576742085e-06, "loss": 0.8638, "step": 21259 }, { "epoch": 0.12, "learning_rate": 4.160757510050458e-06, "loss": 0.869, "step": 21260 }, { "epoch": 0.12, "learning_rate": 4.159882511200777e-06, "loss": 0.8467, "step": 21261 }, { "epoch": 0.12, "learning_rate": 4.159007580203216e-06, "loss": 0.9032, "step": 21262 }, { "epoch": 0.12, "learning_rate": 4.158132717067942e-06, "loss": 0.8235, "step": 21263 }, { "epoch": 0.12, "learning_rate": 4.1572579218051126e-06, "loss": 0.8607, "step": 21264 }, { "epoch": 0.12, "learning_rate": 4.156383194424894e-06, "loss": 0.8506, "step": 21265 }, { "epoch": 0.12, "learning_rate": 4.155508534937449e-06, "loss": 0.9091, "step": 21266 }, { "epoch": 0.12, "learning_rate": 4.15463394335294e-06, "loss": 0.8942, "step": 21267 }, { "epoch": 0.12, "learning_rate": 4.153759419681529e-06, "loss": 0.8483, "step": 21268 }, { "epoch": 0.12, "learning_rate": 4.152884963933375e-06, "loss": 0.8619, "step": 21269 }, { "epoch": 0.12, "learning_rate": 4.152010576118643e-06, "loss": 0.8864, "step": 21270 }, { "epoch": 0.12, "learning_rate": 4.151136256247481e-06, "loss": 0.8458, "step": 21271 }, { "epoch": 0.12, "learning_rate": 4.150262004330055e-06, "loss": 0.8683, "step": 21272 }, { "epoch": 0.12, "learning_rate": 4.149387820376521e-06, "loss": 0.8687, "step": 21273 }, { "epoch": 0.12, "learning_rate": 4.148513704397035e-06, "loss": 0.8593, "step": 21274 }, { "epoch": 0.12, "learning_rate": 4.1476396564017526e-06, "loss": 0.8596, "step": 21275 }, { "epoch": 0.12, "learning_rate": 4.146765676400829e-06, "loss": 0.898, "step": 21276 }, { "epoch": 0.12, "learning_rate": 4.1458917644044216e-06, "loss": 0.878, "step": 21277 }, { "epoch": 0.12, "learning_rate": 4.145017920422678e-06, "loss": 0.8537, "step": 21278 }, { "epoch": 0.12, "learning_rate": 4.144144144465753e-06, "loss": 0.919, "step": 21279 }, { "epoch": 0.12, "learning_rate": 4.143270436543794e-06, "loss": 0.9016, "step": 21280 }, { "epoch": 0.12, "learning_rate": 4.142396796666964e-06, "loss": 0.9069, "step": 21281 }, { "epoch": 0.12, "learning_rate": 4.141523224845403e-06, "loss": 0.8743, "step": 21282 }, { "epoch": 0.12, "learning_rate": 4.140649721089263e-06, "loss": 0.8627, "step": 21283 }, { "epoch": 0.12, "learning_rate": 4.139776285408692e-06, "loss": 0.8707, "step": 21284 }, { "epoch": 0.12, "learning_rate": 4.138902917813839e-06, "loss": 0.8414, "step": 21285 }, { "epoch": 0.12, "learning_rate": 4.138029618314852e-06, "loss": 0.8395, "step": 21286 }, { "epoch": 0.12, "learning_rate": 4.137156386921874e-06, "loss": 0.8135, "step": 21287 }, { "epoch": 0.12, "learning_rate": 4.136283223645057e-06, "loss": 0.8625, "step": 21288 }, { "epoch": 0.12, "learning_rate": 4.1354101284945365e-06, "loss": 0.8939, "step": 21289 }, { "epoch": 0.12, "learning_rate": 4.134537101480462e-06, "loss": 0.8541, "step": 21290 }, { "epoch": 0.12, "learning_rate": 4.133664142612975e-06, "loss": 0.8697, "step": 21291 }, { "epoch": 0.12, "learning_rate": 4.132791251902218e-06, "loss": 0.8285, "step": 21292 }, { "epoch": 0.12, "learning_rate": 4.131918429358334e-06, "loss": 0.8491, "step": 21293 }, { "epoch": 0.12, "learning_rate": 4.131045674991461e-06, "loss": 0.828, "step": 21294 }, { "epoch": 0.12, "learning_rate": 4.130172988811744e-06, "loss": 0.8613, "step": 21295 }, { "epoch": 0.12, "learning_rate": 4.129300370829315e-06, "loss": 0.8285, "step": 21296 }, { "epoch": 0.12, "learning_rate": 4.128427821054316e-06, "loss": 0.8628, "step": 21297 }, { "epoch": 0.12, "learning_rate": 4.127555339496885e-06, "loss": 0.8853, "step": 21298 }, { "epoch": 0.12, "learning_rate": 4.126682926167157e-06, "loss": 0.8331, "step": 21299 }, { "epoch": 0.12, "learning_rate": 4.125810581075269e-06, "loss": 0.8517, "step": 21300 }, { "epoch": 0.12, "learning_rate": 4.124938304231356e-06, "loss": 0.8782, "step": 21301 }, { "epoch": 0.12, "learning_rate": 4.1240660956455555e-06, "loss": 0.8515, "step": 21302 }, { "epoch": 0.12, "learning_rate": 4.123193955327996e-06, "loss": 0.8879, "step": 21303 }, { "epoch": 0.12, "learning_rate": 4.122321883288808e-06, "loss": 0.855, "step": 21304 }, { "epoch": 0.12, "learning_rate": 4.1214498795381304e-06, "loss": 0.8354, "step": 21305 }, { "epoch": 0.12, "learning_rate": 4.120577944086097e-06, "loss": 0.8329, "step": 21306 }, { "epoch": 0.12, "learning_rate": 4.119706076942829e-06, "loss": 0.8814, "step": 21307 }, { "epoch": 0.12, "learning_rate": 4.118834278118459e-06, "loss": 0.9245, "step": 21308 }, { "epoch": 0.12, "learning_rate": 4.1179625476231215e-06, "loss": 0.8577, "step": 21309 }, { "epoch": 0.12, "learning_rate": 4.117090885466931e-06, "loss": 0.8386, "step": 21310 }, { "epoch": 0.12, "learning_rate": 4.116219291660028e-06, "loss": 0.9216, "step": 21311 }, { "epoch": 0.12, "learning_rate": 4.115347766212534e-06, "loss": 0.8327, "step": 21312 }, { "epoch": 0.12, "learning_rate": 4.114476309134579e-06, "loss": 0.8457, "step": 21313 }, { "epoch": 0.12, "learning_rate": 4.11360492043628e-06, "loss": 0.8338, "step": 21314 }, { "epoch": 0.12, "learning_rate": 4.112733600127765e-06, "loss": 0.8529, "step": 21315 }, { "epoch": 0.12, "learning_rate": 4.111862348219158e-06, "loss": 0.9021, "step": 21316 }, { "epoch": 0.12, "learning_rate": 4.110991164720579e-06, "loss": 0.8618, "step": 21317 }, { "epoch": 0.12, "learning_rate": 4.110120049642153e-06, "loss": 0.8411, "step": 21318 }, { "epoch": 0.12, "learning_rate": 4.109249002993997e-06, "loss": 0.8734, "step": 21319 }, { "epoch": 0.12, "learning_rate": 4.108378024786238e-06, "loss": 0.8632, "step": 21320 }, { "epoch": 0.12, "learning_rate": 4.107507115028986e-06, "loss": 0.897, "step": 21321 }, { "epoch": 0.12, "learning_rate": 4.106636273732365e-06, "loss": 0.825, "step": 21322 }, { "epoch": 0.12, "learning_rate": 4.105765500906491e-06, "loss": 0.8623, "step": 21323 }, { "epoch": 0.12, "learning_rate": 4.104894796561482e-06, "loss": 0.8412, "step": 21324 }, { "epoch": 0.12, "learning_rate": 4.104024160707452e-06, "loss": 0.9165, "step": 21325 }, { "epoch": 0.12, "learning_rate": 4.1031535933545185e-06, "loss": 0.8592, "step": 21326 }, { "epoch": 0.12, "learning_rate": 4.1022830945128e-06, "loss": 0.8928, "step": 21327 }, { "epoch": 0.12, "learning_rate": 4.101412664192397e-06, "loss": 0.8608, "step": 21328 }, { "epoch": 0.12, "learning_rate": 4.100542302403435e-06, "loss": 0.8703, "step": 21329 }, { "epoch": 0.12, "learning_rate": 4.099672009156023e-06, "loss": 0.8701, "step": 21330 }, { "epoch": 0.12, "learning_rate": 4.098801784460274e-06, "loss": 0.843, "step": 21331 }, { "epoch": 0.12, "learning_rate": 4.097931628326292e-06, "loss": 0.8742, "step": 21332 }, { "epoch": 0.12, "learning_rate": 4.09706154076419e-06, "loss": 0.8738, "step": 21333 }, { "epoch": 0.12, "learning_rate": 4.096191521784079e-06, "loss": 0.8965, "step": 21334 }, { "epoch": 0.12, "learning_rate": 4.095321571396065e-06, "loss": 0.8665, "step": 21335 }, { "epoch": 0.12, "learning_rate": 4.094451689610256e-06, "loss": 0.8652, "step": 21336 }, { "epoch": 0.12, "learning_rate": 4.093581876436758e-06, "loss": 0.8756, "step": 21337 }, { "epoch": 0.12, "learning_rate": 4.092712131885681e-06, "loss": 0.8874, "step": 21338 }, { "epoch": 0.12, "learning_rate": 4.0918424559671235e-06, "loss": 0.8441, "step": 21339 }, { "epoch": 0.12, "learning_rate": 4.09097284869119e-06, "loss": 0.8993, "step": 21340 }, { "epoch": 0.12, "learning_rate": 4.090103310067989e-06, "loss": 0.8673, "step": 21341 }, { "epoch": 0.12, "learning_rate": 4.089233840107618e-06, "loss": 0.9073, "step": 21342 }, { "epoch": 0.12, "learning_rate": 4.088364438820182e-06, "loss": 0.8697, "step": 21343 }, { "epoch": 0.12, "learning_rate": 4.08749510621578e-06, "loss": 0.8801, "step": 21344 }, { "epoch": 0.12, "learning_rate": 4.086625842304517e-06, "loss": 0.8849, "step": 21345 }, { "epoch": 0.12, "learning_rate": 4.0857566470964846e-06, "loss": 0.8573, "step": 21346 }, { "epoch": 0.12, "learning_rate": 4.084887520601782e-06, "loss": 0.8411, "step": 21347 }, { "epoch": 0.12, "learning_rate": 4.084018462830517e-06, "loss": 0.8889, "step": 21348 }, { "epoch": 0.12, "learning_rate": 4.083149473792777e-06, "loss": 0.8791, "step": 21349 }, { "epoch": 0.12, "learning_rate": 4.082280553498658e-06, "loss": 0.8532, "step": 21350 }, { "epoch": 0.12, "learning_rate": 4.0814117019582615e-06, "loss": 0.8974, "step": 21351 }, { "epoch": 0.12, "learning_rate": 4.080542919181681e-06, "loss": 0.8555, "step": 21352 }, { "epoch": 0.12, "learning_rate": 4.079674205179e-06, "loss": 0.8739, "step": 21353 }, { "epoch": 0.12, "learning_rate": 4.078805559960323e-06, "loss": 0.9024, "step": 21354 }, { "epoch": 0.12, "learning_rate": 4.077936983535743e-06, "loss": 0.87, "step": 21355 }, { "epoch": 0.12, "learning_rate": 4.077068475915343e-06, "loss": 0.9233, "step": 21356 }, { "epoch": 0.12, "learning_rate": 4.076200037109217e-06, "loss": 0.8386, "step": 21357 }, { "epoch": 0.12, "learning_rate": 4.075331667127456e-06, "loss": 0.8863, "step": 21358 }, { "epoch": 0.12, "learning_rate": 4.074463365980148e-06, "loss": 0.8957, "step": 21359 }, { "epoch": 0.12, "learning_rate": 4.073595133677382e-06, "loss": 0.8745, "step": 21360 }, { "epoch": 0.12, "learning_rate": 4.072726970229244e-06, "loss": 0.9364, "step": 21361 }, { "epoch": 0.12, "learning_rate": 4.0718588756458244e-06, "loss": 0.8587, "step": 21362 }, { "epoch": 0.12, "learning_rate": 4.070990849937203e-06, "loss": 0.8659, "step": 21363 }, { "epoch": 0.12, "learning_rate": 4.070122893113469e-06, "loss": 0.8767, "step": 21364 }, { "epoch": 0.12, "learning_rate": 4.0692550051847035e-06, "loss": 0.8741, "step": 21365 }, { "epoch": 0.12, "learning_rate": 4.068387186160991e-06, "loss": 0.8674, "step": 21366 }, { "epoch": 0.12, "learning_rate": 4.067519436052416e-06, "loss": 0.8327, "step": 21367 }, { "epoch": 0.12, "learning_rate": 4.066651754869059e-06, "loss": 0.8661, "step": 21368 }, { "epoch": 0.12, "learning_rate": 4.065784142621004e-06, "loss": 0.8849, "step": 21369 }, { "epoch": 0.12, "learning_rate": 4.064916599318325e-06, "loss": 0.8961, "step": 21370 }, { "epoch": 0.12, "learning_rate": 4.064049124971103e-06, "loss": 0.8979, "step": 21371 }, { "epoch": 0.12, "learning_rate": 4.063181719589415e-06, "loss": 0.8252, "step": 21372 }, { "epoch": 0.12, "learning_rate": 4.06231438318335e-06, "loss": 0.8646, "step": 21373 }, { "epoch": 0.12, "learning_rate": 4.061447115762972e-06, "loss": 0.8605, "step": 21374 }, { "epoch": 0.12, "learning_rate": 4.060579917338362e-06, "loss": 0.8366, "step": 21375 }, { "epoch": 0.12, "learning_rate": 4.059712787919595e-06, "loss": 0.8493, "step": 21376 }, { "epoch": 0.12, "learning_rate": 4.0588457275167455e-06, "loss": 0.8853, "step": 21377 }, { "epoch": 0.12, "learning_rate": 4.057978736139889e-06, "loss": 0.868, "step": 21378 }, { "epoch": 0.12, "learning_rate": 4.057111813799096e-06, "loss": 0.9018, "step": 21379 }, { "epoch": 0.12, "learning_rate": 4.056244960504443e-06, "loss": 0.8476, "step": 21380 }, { "epoch": 0.12, "learning_rate": 4.055378176265995e-06, "loss": 0.8686, "step": 21381 }, { "epoch": 0.12, "learning_rate": 4.054511461093824e-06, "loss": 0.9221, "step": 21382 }, { "epoch": 0.12, "learning_rate": 4.053644814998002e-06, "loss": 0.8257, "step": 21383 }, { "epoch": 0.12, "learning_rate": 4.0527782379885985e-06, "loss": 0.866, "step": 21384 }, { "epoch": 0.12, "learning_rate": 4.051911730075678e-06, "loss": 0.8212, "step": 21385 }, { "epoch": 0.12, "learning_rate": 4.051045291269312e-06, "loss": 0.869, "step": 21386 }, { "epoch": 0.12, "learning_rate": 4.050178921579567e-06, "loss": 0.8341, "step": 21387 }, { "epoch": 0.12, "learning_rate": 4.049312621016505e-06, "loss": 0.8263, "step": 21388 }, { "epoch": 0.12, "learning_rate": 4.048446389590192e-06, "loss": 0.8371, "step": 21389 }, { "epoch": 0.12, "learning_rate": 4.047580227310689e-06, "loss": 0.8708, "step": 21390 }, { "epoch": 0.12, "learning_rate": 4.046714134188071e-06, "loss": 0.8602, "step": 21391 }, { "epoch": 0.12, "learning_rate": 4.045848110232389e-06, "loss": 0.8835, "step": 21392 }, { "epoch": 0.12, "learning_rate": 4.044982155453708e-06, "loss": 0.8451, "step": 21393 }, { "epoch": 0.12, "learning_rate": 4.044116269862094e-06, "loss": 0.8522, "step": 21394 }, { "epoch": 0.12, "learning_rate": 4.043250453467597e-06, "loss": 0.829, "step": 21395 }, { "epoch": 0.12, "learning_rate": 4.0423847062802805e-06, "loss": 0.8387, "step": 21396 }, { "epoch": 0.12, "learning_rate": 4.041519028310206e-06, "loss": 0.8706, "step": 21397 }, { "epoch": 0.12, "learning_rate": 4.040653419567434e-06, "loss": 0.8914, "step": 21398 }, { "epoch": 0.12, "learning_rate": 4.039787880062013e-06, "loss": 0.8612, "step": 21399 }, { "epoch": 0.12, "learning_rate": 4.038922409804003e-06, "loss": 0.836, "step": 21400 }, { "epoch": 0.12, "learning_rate": 4.038057008803462e-06, "loss": 0.8105, "step": 21401 }, { "epoch": 0.12, "learning_rate": 4.0371916770704344e-06, "loss": 0.8868, "step": 21402 }, { "epoch": 0.12, "learning_rate": 4.036326414614985e-06, "loss": 0.8647, "step": 21403 }, { "epoch": 0.12, "learning_rate": 4.0354612214471635e-06, "loss": 0.8542, "step": 21404 }, { "epoch": 0.12, "learning_rate": 4.034596097577023e-06, "loss": 0.8973, "step": 21405 }, { "epoch": 0.12, "learning_rate": 4.033731043014611e-06, "loss": 0.8971, "step": 21406 }, { "epoch": 0.12, "learning_rate": 4.032866057769981e-06, "loss": 0.8934, "step": 21407 }, { "epoch": 0.12, "learning_rate": 4.03200114185318e-06, "loss": 0.8486, "step": 21408 }, { "epoch": 0.12, "learning_rate": 4.0311362952742586e-06, "loss": 0.8639, "step": 21409 }, { "epoch": 0.12, "learning_rate": 4.0302715180432635e-06, "loss": 0.8603, "step": 21410 }, { "epoch": 0.12, "learning_rate": 4.029406810170246e-06, "loss": 0.8567, "step": 21411 }, { "epoch": 0.12, "learning_rate": 4.028542171665251e-06, "loss": 0.8953, "step": 21412 }, { "epoch": 0.12, "learning_rate": 4.027677602538319e-06, "loss": 0.8307, "step": 21413 }, { "epoch": 0.12, "learning_rate": 4.026813102799499e-06, "loss": 0.8872, "step": 21414 }, { "epoch": 0.12, "learning_rate": 4.025948672458834e-06, "loss": 0.8641, "step": 21415 }, { "epoch": 0.12, "learning_rate": 4.0250843115263685e-06, "loss": 0.8912, "step": 21416 }, { "epoch": 0.12, "learning_rate": 4.024220020012143e-06, "loss": 0.8801, "step": 21417 }, { "epoch": 0.12, "learning_rate": 4.023355797926201e-06, "loss": 0.8557, "step": 21418 }, { "epoch": 0.12, "learning_rate": 4.022491645278586e-06, "loss": 0.8539, "step": 21419 }, { "epoch": 0.12, "learning_rate": 4.021627562079327e-06, "loss": 0.8865, "step": 21420 }, { "epoch": 0.12, "learning_rate": 4.0207635483384735e-06, "loss": 0.8861, "step": 21421 }, { "epoch": 0.12, "learning_rate": 4.019899604066061e-06, "loss": 0.863, "step": 21422 }, { "epoch": 0.12, "learning_rate": 4.019035729272132e-06, "loss": 0.8447, "step": 21423 }, { "epoch": 0.12, "learning_rate": 4.018171923966713e-06, "loss": 0.8822, "step": 21424 }, { "epoch": 0.12, "learning_rate": 4.017308188159845e-06, "loss": 0.8792, "step": 21425 }, { "epoch": 0.12, "learning_rate": 4.016444521861564e-06, "loss": 0.887, "step": 21426 }, { "epoch": 0.12, "learning_rate": 4.015580925081904e-06, "loss": 0.8391, "step": 21427 }, { "epoch": 0.12, "learning_rate": 4.014717397830897e-06, "loss": 0.8535, "step": 21428 }, { "epoch": 0.12, "learning_rate": 4.013853940118579e-06, "loss": 0.8594, "step": 21429 }, { "epoch": 0.12, "learning_rate": 4.01299055195498e-06, "loss": 0.8625, "step": 21430 }, { "epoch": 0.12, "learning_rate": 4.01212723335013e-06, "loss": 0.8557, "step": 21431 }, { "epoch": 0.12, "learning_rate": 4.011263984314058e-06, "loss": 0.8777, "step": 21432 }, { "epoch": 0.12, "learning_rate": 4.010400804856796e-06, "loss": 0.8563, "step": 21433 }, { "epoch": 0.12, "learning_rate": 4.009537694988372e-06, "loss": 0.8714, "step": 21434 }, { "epoch": 0.12, "learning_rate": 4.008674654718815e-06, "loss": 0.8681, "step": 21435 }, { "epoch": 0.12, "learning_rate": 4.00781168405815e-06, "loss": 0.887, "step": 21436 }, { "epoch": 0.12, "learning_rate": 4.006948783016409e-06, "loss": 0.876, "step": 21437 }, { "epoch": 0.12, "learning_rate": 4.006085951603608e-06, "loss": 0.8922, "step": 21438 }, { "epoch": 0.12, "learning_rate": 4.005223189829773e-06, "loss": 0.8445, "step": 21439 }, { "epoch": 0.12, "learning_rate": 4.004360497704939e-06, "loss": 0.8943, "step": 21440 }, { "epoch": 0.12, "learning_rate": 4.003497875239117e-06, "loss": 0.8633, "step": 21441 }, { "epoch": 0.12, "learning_rate": 4.002635322442333e-06, "loss": 0.8807, "step": 21442 }, { "epoch": 0.12, "learning_rate": 4.001772839324609e-06, "loss": 0.8496, "step": 21443 }, { "epoch": 0.12, "learning_rate": 4.000910425895969e-06, "loss": 0.8291, "step": 21444 }, { "epoch": 0.12, "learning_rate": 4.000048082166422e-06, "loss": 0.8829, "step": 21445 }, { "epoch": 0.12, "learning_rate": 3.999185808145996e-06, "loss": 0.8763, "step": 21446 }, { "epoch": 0.12, "learning_rate": 3.998323603844713e-06, "loss": 0.8695, "step": 21447 }, { "epoch": 0.12, "learning_rate": 3.99746146927258e-06, "loss": 0.8809, "step": 21448 }, { "epoch": 0.12, "learning_rate": 3.996599404439618e-06, "loss": 0.8739, "step": 21449 }, { "epoch": 0.12, "learning_rate": 3.995737409355842e-06, "loss": 0.8665, "step": 21450 }, { "epoch": 0.12, "learning_rate": 3.9948754840312675e-06, "loss": 0.8673, "step": 21451 }, { "epoch": 0.12, "learning_rate": 3.99401362847591e-06, "loss": 0.839, "step": 21452 }, { "epoch": 0.12, "learning_rate": 3.99315184269978e-06, "loss": 0.8997, "step": 21453 }, { "epoch": 0.12, "learning_rate": 3.992290126712896e-06, "loss": 0.8589, "step": 21454 }, { "epoch": 0.12, "learning_rate": 3.991428480525262e-06, "loss": 0.8542, "step": 21455 }, { "epoch": 0.12, "learning_rate": 3.99056690414689e-06, "loss": 0.8762, "step": 21456 }, { "epoch": 0.12, "learning_rate": 3.989705397587793e-06, "loss": 0.8543, "step": 21457 }, { "epoch": 0.12, "learning_rate": 3.98884396085798e-06, "loss": 0.9092, "step": 21458 }, { "epoch": 0.12, "learning_rate": 3.987982593967458e-06, "loss": 0.8046, "step": 21459 }, { "epoch": 0.12, "learning_rate": 3.987121296926234e-06, "loss": 0.8906, "step": 21460 }, { "epoch": 0.12, "learning_rate": 3.986260069744321e-06, "loss": 0.8336, "step": 21461 }, { "epoch": 0.12, "learning_rate": 3.985398912431716e-06, "loss": 0.8489, "step": 21462 }, { "epoch": 0.12, "learning_rate": 3.9845378249984245e-06, "loss": 0.8619, "step": 21463 }, { "epoch": 0.12, "learning_rate": 3.983676807454458e-06, "loss": 0.8267, "step": 21464 }, { "epoch": 0.12, "learning_rate": 3.98281585980982e-06, "loss": 0.8646, "step": 21465 }, { "epoch": 0.12, "learning_rate": 3.981954982074506e-06, "loss": 0.8395, "step": 21466 }, { "epoch": 0.12, "learning_rate": 3.981094174258522e-06, "loss": 0.8477, "step": 21467 }, { "epoch": 0.12, "learning_rate": 3.980233436371869e-06, "loss": 0.9094, "step": 21468 }, { "epoch": 0.12, "learning_rate": 3.979372768424546e-06, "loss": 0.8828, "step": 21469 }, { "epoch": 0.12, "learning_rate": 3.978512170426555e-06, "loss": 0.8874, "step": 21470 }, { "epoch": 0.12, "learning_rate": 3.977651642387892e-06, "loss": 0.8724, "step": 21471 }, { "epoch": 0.12, "learning_rate": 3.976791184318561e-06, "loss": 0.8703, "step": 21472 }, { "epoch": 0.12, "learning_rate": 3.97593079622855e-06, "loss": 0.8321, "step": 21473 }, { "epoch": 0.12, "learning_rate": 3.975070478127858e-06, "loss": 0.8369, "step": 21474 }, { "epoch": 0.12, "learning_rate": 3.9742102300264825e-06, "loss": 0.8376, "step": 21475 }, { "epoch": 0.12, "learning_rate": 3.973350051934418e-06, "loss": 0.9131, "step": 21476 }, { "epoch": 0.12, "learning_rate": 3.972489943861658e-06, "loss": 0.8778, "step": 21477 }, { "epoch": 0.12, "learning_rate": 3.9716299058181955e-06, "loss": 0.8435, "step": 21478 }, { "epoch": 0.12, "learning_rate": 3.970769937814024e-06, "loss": 0.8779, "step": 21479 }, { "epoch": 0.12, "learning_rate": 3.969910039859131e-06, "loss": 0.8854, "step": 21480 }, { "epoch": 0.12, "learning_rate": 3.969050211963507e-06, "loss": 0.8569, "step": 21481 }, { "epoch": 0.12, "learning_rate": 3.968190454137143e-06, "loss": 0.8589, "step": 21482 }, { "epoch": 0.12, "learning_rate": 3.967330766390035e-06, "loss": 0.8753, "step": 21483 }, { "epoch": 0.12, "learning_rate": 3.966471148732162e-06, "loss": 0.8959, "step": 21484 }, { "epoch": 0.12, "learning_rate": 3.965611601173512e-06, "loss": 0.8797, "step": 21485 }, { "epoch": 0.12, "learning_rate": 3.964752123724079e-06, "loss": 0.8833, "step": 21486 }, { "epoch": 0.12, "learning_rate": 3.963892716393839e-06, "loss": 0.8537, "step": 21487 }, { "epoch": 0.12, "learning_rate": 3.963033379192779e-06, "loss": 0.9009, "step": 21488 }, { "epoch": 0.12, "learning_rate": 3.962174112130886e-06, "loss": 0.8686, "step": 21489 }, { "epoch": 0.12, "learning_rate": 3.961314915218149e-06, "loss": 0.886, "step": 21490 }, { "epoch": 0.12, "learning_rate": 3.960455788464538e-06, "loss": 0.85, "step": 21491 }, { "epoch": 0.12, "learning_rate": 3.959596731880041e-06, "loss": 0.8619, "step": 21492 }, { "epoch": 0.12, "learning_rate": 3.958737745474638e-06, "loss": 0.8753, "step": 21493 }, { "epoch": 0.12, "learning_rate": 3.9578788292583095e-06, "loss": 0.8553, "step": 21494 }, { "epoch": 0.12, "learning_rate": 3.957019983241034e-06, "loss": 0.8783, "step": 21495 }, { "epoch": 0.12, "learning_rate": 3.95616120743279e-06, "loss": 0.8434, "step": 21496 }, { "epoch": 0.12, "learning_rate": 3.955302501843557e-06, "loss": 0.8377, "step": 21497 }, { "epoch": 0.12, "learning_rate": 3.954443866483309e-06, "loss": 0.8574, "step": 21498 }, { "epoch": 0.12, "learning_rate": 3.95358530136202e-06, "loss": 0.85, "step": 21499 }, { "epoch": 0.12, "learning_rate": 3.952726806489668e-06, "loss": 0.9028, "step": 21500 }, { "epoch": 0.12, "learning_rate": 3.951868381876227e-06, "loss": 0.8692, "step": 21501 }, { "epoch": 0.12, "learning_rate": 3.951010027531671e-06, "loss": 0.887, "step": 21502 }, { "epoch": 0.12, "learning_rate": 3.9501517434659705e-06, "loss": 0.8365, "step": 21503 }, { "epoch": 0.12, "learning_rate": 3.949293529689103e-06, "loss": 0.8432, "step": 21504 }, { "epoch": 0.12, "learning_rate": 3.948435386211031e-06, "loss": 0.8599, "step": 21505 }, { "epoch": 0.12, "learning_rate": 3.947577313041729e-06, "loss": 0.8386, "step": 21506 }, { "epoch": 0.13, "learning_rate": 3.946719310191166e-06, "loss": 0.8687, "step": 21507 }, { "epoch": 0.13, "learning_rate": 3.945861377669311e-06, "loss": 0.8932, "step": 21508 }, { "epoch": 0.13, "learning_rate": 3.945003515486131e-06, "loss": 0.9067, "step": 21509 }, { "epoch": 0.13, "learning_rate": 3.944145723651594e-06, "loss": 0.8846, "step": 21510 }, { "epoch": 0.13, "learning_rate": 3.943288002175667e-06, "loss": 0.8863, "step": 21511 }, { "epoch": 0.13, "learning_rate": 3.942430351068307e-06, "loss": 0.8845, "step": 21512 }, { "epoch": 0.13, "learning_rate": 3.941572770339489e-06, "loss": 0.8692, "step": 21513 }, { "epoch": 0.13, "learning_rate": 3.940715259999172e-06, "loss": 0.8617, "step": 21514 }, { "epoch": 0.13, "learning_rate": 3.939857820057322e-06, "loss": 0.8717, "step": 21515 }, { "epoch": 0.13, "learning_rate": 3.939000450523896e-06, "loss": 0.8997, "step": 21516 }, { "epoch": 0.13, "learning_rate": 3.938143151408858e-06, "loss": 0.8682, "step": 21517 }, { "epoch": 0.13, "learning_rate": 3.937285922722166e-06, "loss": 0.8893, "step": 21518 }, { "epoch": 0.13, "learning_rate": 3.936428764473782e-06, "loss": 0.799, "step": 21519 }, { "epoch": 0.13, "learning_rate": 3.935571676673664e-06, "loss": 0.8889, "step": 21520 }, { "epoch": 0.13, "learning_rate": 3.93471465933177e-06, "loss": 0.8385, "step": 21521 }, { "epoch": 0.13, "learning_rate": 3.93385771245806e-06, "loss": 0.8601, "step": 21522 }, { "epoch": 0.13, "learning_rate": 3.933000836062485e-06, "loss": 0.8471, "step": 21523 }, { "epoch": 0.13, "learning_rate": 3.932144030155002e-06, "loss": 0.8505, "step": 21524 }, { "epoch": 0.13, "learning_rate": 3.931287294745566e-06, "loss": 0.8883, "step": 21525 }, { "epoch": 0.13, "learning_rate": 3.930430629844131e-06, "loss": 0.8306, "step": 21526 }, { "epoch": 0.13, "learning_rate": 3.929574035460649e-06, "loss": 0.9004, "step": 21527 }, { "epoch": 0.13, "learning_rate": 3.9287175116050745e-06, "loss": 0.8342, "step": 21528 }, { "epoch": 0.13, "learning_rate": 3.927861058287361e-06, "loss": 0.8708, "step": 21529 }, { "epoch": 0.13, "learning_rate": 3.9270046755174515e-06, "loss": 0.86, "step": 21530 }, { "epoch": 0.13, "learning_rate": 3.926148363305297e-06, "loss": 0.8707, "step": 21531 }, { "epoch": 0.13, "learning_rate": 3.925292121660855e-06, "loss": 0.8287, "step": 21532 }, { "epoch": 0.13, "learning_rate": 3.924435950594064e-06, "loss": 0.8659, "step": 21533 }, { "epoch": 0.13, "learning_rate": 3.923579850114876e-06, "loss": 0.8943, "step": 21534 }, { "epoch": 0.13, "learning_rate": 3.922723820233237e-06, "loss": 0.8507, "step": 21535 }, { "epoch": 0.13, "learning_rate": 3.92186786095909e-06, "loss": 0.8769, "step": 21536 }, { "epoch": 0.13, "learning_rate": 3.921011972302382e-06, "loss": 0.8219, "step": 21537 }, { "epoch": 0.13, "learning_rate": 3.920156154273056e-06, "loss": 0.8509, "step": 21538 }, { "epoch": 0.13, "learning_rate": 3.919300406881061e-06, "loss": 0.8634, "step": 21539 }, { "epoch": 0.13, "learning_rate": 3.918444730136329e-06, "loss": 0.8823, "step": 21540 }, { "epoch": 0.13, "learning_rate": 3.917589124048807e-06, "loss": 0.866, "step": 21541 }, { "epoch": 0.13, "learning_rate": 3.916733588628435e-06, "loss": 0.8641, "step": 21542 }, { "epoch": 0.13, "learning_rate": 3.915878123885154e-06, "loss": 0.8516, "step": 21543 }, { "epoch": 0.13, "learning_rate": 3.915022729828901e-06, "loss": 0.8547, "step": 21544 }, { "epoch": 0.13, "learning_rate": 3.9141674064696165e-06, "loss": 0.8465, "step": 21545 }, { "epoch": 0.13, "learning_rate": 3.913312153817239e-06, "loss": 0.8687, "step": 21546 }, { "epoch": 0.13, "learning_rate": 3.9124569718816996e-06, "loss": 0.8208, "step": 21547 }, { "epoch": 0.13, "learning_rate": 3.9116018606729375e-06, "loss": 0.8662, "step": 21548 }, { "epoch": 0.13, "learning_rate": 3.910746820200887e-06, "loss": 0.8954, "step": 21549 }, { "epoch": 0.13, "learning_rate": 3.909891850475482e-06, "loss": 0.8669, "step": 21550 }, { "epoch": 0.13, "learning_rate": 3.909036951506657e-06, "loss": 0.8293, "step": 21551 }, { "epoch": 0.13, "learning_rate": 3.908182123304344e-06, "loss": 0.8844, "step": 21552 }, { "epoch": 0.13, "learning_rate": 3.907327365878478e-06, "loss": 0.8408, "step": 21553 }, { "epoch": 0.13, "learning_rate": 3.906472679238982e-06, "loss": 0.8722, "step": 21554 }, { "epoch": 0.13, "learning_rate": 3.905618063395787e-06, "loss": 0.8425, "step": 21555 }, { "epoch": 0.13, "learning_rate": 3.904763518358828e-06, "loss": 0.9006, "step": 21556 }, { "epoch": 0.13, "learning_rate": 3.903909044138035e-06, "loss": 0.8377, "step": 21557 }, { "epoch": 0.13, "learning_rate": 3.903054640743328e-06, "loss": 0.867, "step": 21558 }, { "epoch": 0.13, "learning_rate": 3.902200308184636e-06, "loss": 0.8674, "step": 21559 }, { "epoch": 0.13, "learning_rate": 3.901346046471887e-06, "loss": 0.8716, "step": 21560 }, { "epoch": 0.13, "learning_rate": 3.900491855615004e-06, "loss": 0.8532, "step": 21561 }, { "epoch": 0.13, "learning_rate": 3.899637735623911e-06, "loss": 0.8445, "step": 21562 }, { "epoch": 0.13, "learning_rate": 3.898783686508534e-06, "loss": 0.8613, "step": 21563 }, { "epoch": 0.13, "learning_rate": 3.897929708278797e-06, "loss": 0.8983, "step": 21564 }, { "epoch": 0.13, "learning_rate": 3.897075800944615e-06, "loss": 0.8239, "step": 21565 }, { "epoch": 0.13, "learning_rate": 3.896221964515914e-06, "loss": 0.8591, "step": 21566 }, { "epoch": 0.13, "learning_rate": 3.895368199002612e-06, "loss": 0.8746, "step": 21567 }, { "epoch": 0.13, "learning_rate": 3.894514504414629e-06, "loss": 0.8977, "step": 21568 }, { "epoch": 0.13, "learning_rate": 3.893660880761883e-06, "loss": 0.8841, "step": 21569 }, { "epoch": 0.13, "learning_rate": 3.892807328054294e-06, "loss": 0.9036, "step": 21570 }, { "epoch": 0.13, "learning_rate": 3.891953846301779e-06, "loss": 0.826, "step": 21571 }, { "epoch": 0.13, "learning_rate": 3.891100435514249e-06, "loss": 0.8353, "step": 21572 }, { "epoch": 0.13, "learning_rate": 3.890247095701623e-06, "loss": 0.8556, "step": 21573 }, { "epoch": 0.13, "learning_rate": 3.88939382687381e-06, "loss": 0.8893, "step": 21574 }, { "epoch": 0.13, "learning_rate": 3.888540629040735e-06, "loss": 0.8554, "step": 21575 }, { "epoch": 0.13, "learning_rate": 3.887687502212299e-06, "loss": 0.8514, "step": 21576 }, { "epoch": 0.13, "learning_rate": 3.886834446398421e-06, "loss": 0.8834, "step": 21577 }, { "epoch": 0.13, "learning_rate": 3.885981461609012e-06, "loss": 0.8416, "step": 21578 }, { "epoch": 0.13, "learning_rate": 3.885128547853977e-06, "loss": 0.8086, "step": 21579 }, { "epoch": 0.13, "learning_rate": 3.8842757051432245e-06, "loss": 0.8549, "step": 21580 }, { "epoch": 0.13, "learning_rate": 3.8834229334866704e-06, "loss": 0.8987, "step": 21581 }, { "epoch": 0.13, "learning_rate": 3.8825702328942215e-06, "loss": 0.8639, "step": 21582 }, { "epoch": 0.13, "learning_rate": 3.8817176033757794e-06, "loss": 0.8608, "step": 21583 }, { "epoch": 0.13, "learning_rate": 3.880865044941253e-06, "loss": 0.8704, "step": 21584 }, { "epoch": 0.13, "learning_rate": 3.880012557600546e-06, "loss": 0.8499, "step": 21585 }, { "epoch": 0.13, "learning_rate": 3.879160141363566e-06, "loss": 0.8792, "step": 21586 }, { "epoch": 0.13, "learning_rate": 3.878307796240215e-06, "loss": 0.8654, "step": 21587 }, { "epoch": 0.13, "learning_rate": 3.877455522240394e-06, "loss": 0.8572, "step": 21588 }, { "epoch": 0.13, "learning_rate": 3.876603319374012e-06, "loss": 0.8305, "step": 21589 }, { "epoch": 0.13, "learning_rate": 3.87575118765096e-06, "loss": 0.9173, "step": 21590 }, { "epoch": 0.13, "learning_rate": 3.874899127081143e-06, "loss": 0.8926, "step": 21591 }, { "epoch": 0.13, "learning_rate": 3.874047137674462e-06, "loss": 0.8782, "step": 21592 }, { "epoch": 0.13, "learning_rate": 3.873195219440812e-06, "loss": 0.9074, "step": 21593 }, { "epoch": 0.13, "learning_rate": 3.872343372390094e-06, "loss": 0.8369, "step": 21594 }, { "epoch": 0.13, "learning_rate": 3.871491596532204e-06, "loss": 0.8552, "step": 21595 }, { "epoch": 0.13, "learning_rate": 3.8706398918770414e-06, "loss": 0.8592, "step": 21596 }, { "epoch": 0.13, "learning_rate": 3.869788258434496e-06, "loss": 0.8416, "step": 21597 }, { "epoch": 0.13, "learning_rate": 3.868936696214461e-06, "loss": 0.9229, "step": 21598 }, { "epoch": 0.13, "learning_rate": 3.8680852052268395e-06, "loss": 0.8466, "step": 21599 }, { "epoch": 0.13, "learning_rate": 3.8672337854815146e-06, "loss": 0.8475, "step": 21600 }, { "epoch": 0.13, "learning_rate": 3.866382436988384e-06, "loss": 0.9193, "step": 21601 }, { "epoch": 0.13, "learning_rate": 3.865531159757336e-06, "loss": 0.8542, "step": 21602 }, { "epoch": 0.13, "learning_rate": 3.864679953798266e-06, "loss": 0.8099, "step": 21603 }, { "epoch": 0.13, "learning_rate": 3.863828819121053e-06, "loss": 0.8773, "step": 21604 }, { "epoch": 0.13, "learning_rate": 3.8629777557355964e-06, "loss": 0.8641, "step": 21605 }, { "epoch": 0.13, "learning_rate": 3.862126763651779e-06, "loss": 0.8557, "step": 21606 }, { "epoch": 0.13, "learning_rate": 3.861275842879493e-06, "loss": 0.874, "step": 21607 }, { "epoch": 0.13, "learning_rate": 3.860424993428617e-06, "loss": 0.8834, "step": 21608 }, { "epoch": 0.13, "learning_rate": 3.859574215309041e-06, "loss": 0.8274, "step": 21609 }, { "epoch": 0.13, "learning_rate": 3.858723508530648e-06, "loss": 0.9165, "step": 21610 }, { "epoch": 0.13, "learning_rate": 3.857872873103322e-06, "loss": 0.873, "step": 21611 }, { "epoch": 0.13, "learning_rate": 3.857022309036946e-06, "loss": 0.8749, "step": 21612 }, { "epoch": 0.13, "learning_rate": 3.856171816341404e-06, "loss": 0.8563, "step": 21613 }, { "epoch": 0.13, "learning_rate": 3.855321395026578e-06, "loss": 0.851, "step": 21614 }, { "epoch": 0.13, "learning_rate": 3.854471045102343e-06, "loss": 0.8633, "step": 21615 }, { "epoch": 0.13, "learning_rate": 3.853620766578582e-06, "loss": 0.8624, "step": 21616 }, { "epoch": 0.13, "learning_rate": 3.852770559465172e-06, "loss": 0.8638, "step": 21617 }, { "epoch": 0.13, "learning_rate": 3.851920423771993e-06, "loss": 0.853, "step": 21618 }, { "epoch": 0.13, "learning_rate": 3.851070359508923e-06, "loss": 0.8661, "step": 21619 }, { "epoch": 0.13, "learning_rate": 3.850220366685835e-06, "loss": 0.8655, "step": 21620 }, { "epoch": 0.13, "learning_rate": 3.849370445312611e-06, "loss": 0.9065, "step": 21621 }, { "epoch": 0.13, "learning_rate": 3.848520595399117e-06, "loss": 0.8522, "step": 21622 }, { "epoch": 0.13, "learning_rate": 3.847670816955227e-06, "loss": 0.8573, "step": 21623 }, { "epoch": 0.13, "learning_rate": 3.846821109990825e-06, "loss": 0.8814, "step": 21624 }, { "epoch": 0.13, "learning_rate": 3.845971474515772e-06, "loss": 0.8724, "step": 21625 }, { "epoch": 0.13, "learning_rate": 3.845121910539943e-06, "loss": 0.845, "step": 21626 }, { "epoch": 0.13, "learning_rate": 3.8442724180732085e-06, "loss": 0.8434, "step": 21627 }, { "epoch": 0.13, "learning_rate": 3.843422997125438e-06, "loss": 0.86, "step": 21628 }, { "epoch": 0.13, "learning_rate": 3.842573647706501e-06, "loss": 0.8458, "step": 21629 }, { "epoch": 0.13, "learning_rate": 3.841724369826265e-06, "loss": 0.8483, "step": 21630 }, { "epoch": 0.13, "learning_rate": 3.840875163494601e-06, "loss": 0.8604, "step": 21631 }, { "epoch": 0.13, "learning_rate": 3.840026028721367e-06, "loss": 0.8415, "step": 21632 }, { "epoch": 0.13, "learning_rate": 3.839176965516432e-06, "loss": 0.8965, "step": 21633 }, { "epoch": 0.13, "learning_rate": 3.838327973889664e-06, "loss": 0.9044, "step": 21634 }, { "epoch": 0.13, "learning_rate": 3.837479053850924e-06, "loss": 0.8639, "step": 21635 }, { "epoch": 0.13, "learning_rate": 3.836630205410073e-06, "loss": 0.8646, "step": 21636 }, { "epoch": 0.13, "learning_rate": 3.835781428576978e-06, "loss": 0.8423, "step": 21637 }, { "epoch": 0.13, "learning_rate": 3.834932723361501e-06, "loss": 0.8998, "step": 21638 }, { "epoch": 0.13, "learning_rate": 3.834084089773496e-06, "loss": 0.8318, "step": 21639 }, { "epoch": 0.13, "learning_rate": 3.833235527822825e-06, "loss": 0.857, "step": 21640 }, { "epoch": 0.13, "learning_rate": 3.832387037519345e-06, "loss": 0.9165, "step": 21641 }, { "epoch": 0.13, "learning_rate": 3.831538618872924e-06, "loss": 0.8971, "step": 21642 }, { "epoch": 0.13, "learning_rate": 3.830690271893408e-06, "loss": 0.8787, "step": 21643 }, { "epoch": 0.13, "learning_rate": 3.829841996590658e-06, "loss": 0.863, "step": 21644 }, { "epoch": 0.13, "learning_rate": 3.828993792974532e-06, "loss": 0.8412, "step": 21645 }, { "epoch": 0.13, "learning_rate": 3.828145661054879e-06, "loss": 0.8598, "step": 21646 }, { "epoch": 0.13, "learning_rate": 3.8272976008415505e-06, "loss": 0.8329, "step": 21647 }, { "epoch": 0.13, "learning_rate": 3.826449612344408e-06, "loss": 0.9166, "step": 21648 }, { "epoch": 0.13, "learning_rate": 3.825601695573304e-06, "loss": 0.8602, "step": 21649 }, { "epoch": 0.13, "learning_rate": 3.824753850538082e-06, "loss": 0.864, "step": 21650 }, { "epoch": 0.13, "learning_rate": 3.823906077248597e-06, "loss": 0.8206, "step": 21651 }, { "epoch": 0.13, "learning_rate": 3.823058375714698e-06, "loss": 0.8294, "step": 21652 }, { "epoch": 0.13, "learning_rate": 3.822210745946233e-06, "loss": 0.8338, "step": 21653 }, { "epoch": 0.13, "learning_rate": 3.821363187953052e-06, "loss": 0.8684, "step": 21654 }, { "epoch": 0.13, "learning_rate": 3.820515701745e-06, "loss": 0.8639, "step": 21655 }, { "epoch": 0.13, "learning_rate": 3.819668287331929e-06, "loss": 0.8559, "step": 21656 }, { "epoch": 0.13, "learning_rate": 3.818820944723676e-06, "loss": 0.8168, "step": 21657 }, { "epoch": 0.13, "learning_rate": 3.817973673930089e-06, "loss": 0.8308, "step": 21658 }, { "epoch": 0.13, "learning_rate": 3.817126474961013e-06, "loss": 0.8725, "step": 21659 }, { "epoch": 0.13, "learning_rate": 3.816279347826289e-06, "loss": 0.8845, "step": 21660 }, { "epoch": 0.13, "learning_rate": 3.815432292535761e-06, "loss": 0.8791, "step": 21661 }, { "epoch": 0.13, "learning_rate": 3.8145853090992712e-06, "loss": 0.884, "step": 21662 }, { "epoch": 0.13, "learning_rate": 3.813738397526661e-06, "loss": 0.8514, "step": 21663 }, { "epoch": 0.13, "learning_rate": 3.8128915578277648e-06, "loss": 0.8663, "step": 21664 }, { "epoch": 0.13, "learning_rate": 3.8120447900124236e-06, "loss": 0.8433, "step": 21665 }, { "epoch": 0.13, "learning_rate": 3.8111980940904734e-06, "loss": 0.8118, "step": 21666 }, { "epoch": 0.13, "learning_rate": 3.8103514700717614e-06, "loss": 0.8608, "step": 21667 }, { "epoch": 0.13, "learning_rate": 3.8095049179661127e-06, "loss": 0.8294, "step": 21668 }, { "epoch": 0.13, "learning_rate": 3.808658437783367e-06, "loss": 0.8812, "step": 21669 }, { "epoch": 0.13, "learning_rate": 3.807812029533362e-06, "loss": 0.8725, "step": 21670 }, { "epoch": 0.13, "learning_rate": 3.806965693225921e-06, "loss": 0.8879, "step": 21671 }, { "epoch": 0.13, "learning_rate": 3.806119428870888e-06, "loss": 0.8484, "step": 21672 }, { "epoch": 0.13, "learning_rate": 3.805273236478091e-06, "loss": 0.9042, "step": 21673 }, { "epoch": 0.13, "learning_rate": 3.8044271160573654e-06, "loss": 0.8768, "step": 21674 }, { "epoch": 0.13, "learning_rate": 3.8035810676185335e-06, "loss": 0.8389, "step": 21675 }, { "epoch": 0.13, "learning_rate": 3.802735091171431e-06, "loss": 0.8639, "step": 21676 }, { "epoch": 0.13, "learning_rate": 3.8018891867258834e-06, "loss": 0.8604, "step": 21677 }, { "epoch": 0.13, "learning_rate": 3.801043354291721e-06, "loss": 0.8887, "step": 21678 }, { "epoch": 0.13, "learning_rate": 3.8001975938787693e-06, "loss": 0.8373, "step": 21679 }, { "epoch": 0.13, "learning_rate": 3.7993519054968563e-06, "loss": 0.8703, "step": 21680 }, { "epoch": 0.13, "learning_rate": 3.79850628915581e-06, "loss": 0.8619, "step": 21681 }, { "epoch": 0.13, "learning_rate": 3.797660744865448e-06, "loss": 0.9068, "step": 21682 }, { "epoch": 0.13, "learning_rate": 3.796815272635598e-06, "loss": 0.872, "step": 21683 }, { "epoch": 0.13, "learning_rate": 3.7959698724760818e-06, "loss": 0.8505, "step": 21684 }, { "epoch": 0.13, "learning_rate": 3.7951245443967223e-06, "loss": 0.8283, "step": 21685 }, { "epoch": 0.13, "learning_rate": 3.7942792884073408e-06, "loss": 0.8709, "step": 21686 }, { "epoch": 0.13, "learning_rate": 3.7934341045177582e-06, "loss": 0.8705, "step": 21687 }, { "epoch": 0.13, "learning_rate": 3.792588992737797e-06, "loss": 0.8956, "step": 21688 }, { "epoch": 0.13, "learning_rate": 3.791743953077268e-06, "loss": 0.8577, "step": 21689 }, { "epoch": 0.13, "learning_rate": 3.790898985545991e-06, "loss": 0.8661, "step": 21690 }, { "epoch": 0.13, "learning_rate": 3.790054090153793e-06, "loss": 0.872, "step": 21691 }, { "epoch": 0.13, "learning_rate": 3.7892092669104787e-06, "loss": 0.8362, "step": 21692 }, { "epoch": 0.13, "learning_rate": 3.7883645158258674e-06, "loss": 0.843, "step": 21693 }, { "epoch": 0.13, "learning_rate": 3.7875198369097745e-06, "loss": 0.8417, "step": 21694 }, { "epoch": 0.13, "learning_rate": 3.7866752301720165e-06, "loss": 0.8806, "step": 21695 }, { "epoch": 0.13, "learning_rate": 3.7858306956223967e-06, "loss": 0.8527, "step": 21696 }, { "epoch": 0.13, "learning_rate": 3.7849862332707356e-06, "loss": 0.866, "step": 21697 }, { "epoch": 0.13, "learning_rate": 3.784141843126843e-06, "loss": 0.8694, "step": 21698 }, { "epoch": 0.13, "learning_rate": 3.783297525200531e-06, "loss": 0.8478, "step": 21699 }, { "epoch": 0.13, "learning_rate": 3.7824532795016043e-06, "loss": 0.8678, "step": 21700 }, { "epoch": 0.13, "learning_rate": 3.781609106039873e-06, "loss": 0.8882, "step": 21701 }, { "epoch": 0.13, "learning_rate": 3.7807650048251456e-06, "loss": 0.8973, "step": 21702 }, { "epoch": 0.13, "learning_rate": 3.7799209758672296e-06, "loss": 0.8814, "step": 21703 }, { "epoch": 0.13, "learning_rate": 3.7790770191759306e-06, "loss": 0.8292, "step": 21704 }, { "epoch": 0.13, "learning_rate": 3.7782331347610547e-06, "loss": 0.8508, "step": 21705 }, { "epoch": 0.13, "learning_rate": 3.7773893226324097e-06, "loss": 0.8792, "step": 21706 }, { "epoch": 0.13, "learning_rate": 3.776545582799791e-06, "loss": 0.823, "step": 21707 }, { "epoch": 0.13, "learning_rate": 3.7757019152730055e-06, "loss": 0.8769, "step": 21708 }, { "epoch": 0.13, "learning_rate": 3.774858320061857e-06, "loss": 0.8372, "step": 21709 }, { "epoch": 0.13, "learning_rate": 3.7740147971761443e-06, "loss": 0.8483, "step": 21710 }, { "epoch": 0.13, "learning_rate": 3.7731713466256683e-06, "loss": 0.8476, "step": 21711 }, { "epoch": 0.13, "learning_rate": 3.7723279684202284e-06, "loss": 0.8781, "step": 21712 }, { "epoch": 0.13, "learning_rate": 3.7714846625696266e-06, "loss": 0.8681, "step": 21713 }, { "epoch": 0.13, "learning_rate": 3.7706414290836544e-06, "loss": 0.9083, "step": 21714 }, { "epoch": 0.13, "learning_rate": 3.769798267972109e-06, "loss": 0.8599, "step": 21715 }, { "epoch": 0.13, "learning_rate": 3.7689551792447963e-06, "loss": 0.8799, "step": 21716 }, { "epoch": 0.13, "learning_rate": 3.7681121629115005e-06, "loss": 0.8587, "step": 21717 }, { "epoch": 0.13, "learning_rate": 3.76726921898202e-06, "loss": 0.8205, "step": 21718 }, { "epoch": 0.13, "learning_rate": 3.766426347466149e-06, "loss": 0.8339, "step": 21719 }, { "epoch": 0.13, "learning_rate": 3.765583548373678e-06, "loss": 0.859, "step": 21720 }, { "epoch": 0.13, "learning_rate": 3.7647408217144023e-06, "loss": 0.8322, "step": 21721 }, { "epoch": 0.13, "learning_rate": 3.76389816749811e-06, "loss": 0.88, "step": 21722 }, { "epoch": 0.13, "learning_rate": 3.7630555857345963e-06, "loss": 0.8132, "step": 21723 }, { "epoch": 0.13, "learning_rate": 3.762213076433643e-06, "loss": 0.8517, "step": 21724 }, { "epoch": 0.13, "learning_rate": 3.7613706396050433e-06, "loss": 0.8842, "step": 21725 }, { "epoch": 0.13, "learning_rate": 3.7605282752585826e-06, "loss": 0.8429, "step": 21726 }, { "epoch": 0.13, "learning_rate": 3.759685983404049e-06, "loss": 0.8501, "step": 21727 }, { "epoch": 0.13, "learning_rate": 3.758843764051229e-06, "loss": 0.9024, "step": 21728 }, { "epoch": 0.13, "learning_rate": 3.7580016172099067e-06, "loss": 0.884, "step": 21729 }, { "epoch": 0.13, "learning_rate": 3.757159542889871e-06, "loss": 0.875, "step": 21730 }, { "epoch": 0.13, "learning_rate": 3.756317541100897e-06, "loss": 0.8216, "step": 21731 }, { "epoch": 0.13, "learning_rate": 3.7554756118527725e-06, "loss": 0.8387, "step": 21732 }, { "epoch": 0.13, "learning_rate": 3.7546337551552736e-06, "loss": 0.8608, "step": 21733 }, { "epoch": 0.13, "learning_rate": 3.7537919710181936e-06, "loss": 0.8658, "step": 21734 }, { "epoch": 0.13, "learning_rate": 3.7529502594513013e-06, "loss": 0.8169, "step": 21735 }, { "epoch": 0.13, "learning_rate": 3.75210862046438e-06, "loss": 0.8655, "step": 21736 }, { "epoch": 0.13, "learning_rate": 3.751267054067207e-06, "loss": 0.8755, "step": 21737 }, { "epoch": 0.13, "learning_rate": 3.750425560269565e-06, "loss": 0.8685, "step": 21738 }, { "epoch": 0.13, "learning_rate": 3.7495841390812184e-06, "loss": 0.8772, "step": 21739 }, { "epoch": 0.13, "learning_rate": 3.7487427905119545e-06, "loss": 0.8487, "step": 21740 }, { "epoch": 0.13, "learning_rate": 3.747901514571548e-06, "loss": 0.8191, "step": 21741 }, { "epoch": 0.13, "learning_rate": 3.7470603112697667e-06, "loss": 0.9149, "step": 21742 }, { "epoch": 0.13, "learning_rate": 3.746219180616387e-06, "loss": 0.8319, "step": 21743 }, { "epoch": 0.13, "learning_rate": 3.7453781226211816e-06, "loss": 0.8766, "step": 21744 }, { "epoch": 0.13, "learning_rate": 3.744537137293922e-06, "loss": 0.8733, "step": 21745 }, { "epoch": 0.13, "learning_rate": 3.7436962246443787e-06, "loss": 0.8579, "step": 21746 }, { "epoch": 0.13, "learning_rate": 3.742855384682321e-06, "loss": 0.8397, "step": 21747 }, { "epoch": 0.13, "learning_rate": 3.7420146174175245e-06, "loss": 0.8807, "step": 21748 }, { "epoch": 0.13, "learning_rate": 3.7411739228597464e-06, "loss": 0.8695, "step": 21749 }, { "epoch": 0.13, "learning_rate": 3.7403333010187613e-06, "loss": 0.8739, "step": 21750 }, { "epoch": 0.13, "learning_rate": 3.7394927519043335e-06, "loss": 0.8196, "step": 21751 }, { "epoch": 0.13, "learning_rate": 3.738652275526229e-06, "loss": 0.8215, "step": 21752 }, { "epoch": 0.13, "learning_rate": 3.7378118718942135e-06, "loss": 0.8809, "step": 21753 }, { "epoch": 0.13, "learning_rate": 3.7369715410180494e-06, "loss": 0.8694, "step": 21754 }, { "epoch": 0.13, "learning_rate": 3.736131282907506e-06, "loss": 0.8603, "step": 21755 }, { "epoch": 0.13, "learning_rate": 3.7352910975723366e-06, "loss": 0.8632, "step": 21756 }, { "epoch": 0.13, "learning_rate": 3.7344509850223074e-06, "loss": 0.8742, "step": 21757 }, { "epoch": 0.13, "learning_rate": 3.7336109452671744e-06, "loss": 0.8582, "step": 21758 }, { "epoch": 0.13, "learning_rate": 3.7327709783167086e-06, "loss": 0.8771, "step": 21759 }, { "epoch": 0.13, "learning_rate": 3.731931084180659e-06, "loss": 0.8817, "step": 21760 }, { "epoch": 0.13, "learning_rate": 3.7310912628687856e-06, "loss": 0.8775, "step": 21761 }, { "epoch": 0.13, "learning_rate": 3.7302515143908513e-06, "loss": 0.8786, "step": 21762 }, { "epoch": 0.13, "learning_rate": 3.7294118387566e-06, "loss": 0.8453, "step": 21763 }, { "epoch": 0.13, "learning_rate": 3.7285722359757993e-06, "loss": 0.8463, "step": 21764 }, { "epoch": 0.13, "learning_rate": 3.7277327060582002e-06, "loss": 0.9223, "step": 21765 }, { "epoch": 0.13, "learning_rate": 3.726893249013559e-06, "loss": 0.8715, "step": 21766 }, { "epoch": 0.13, "learning_rate": 3.726053864851623e-06, "loss": 0.845, "step": 21767 }, { "epoch": 0.13, "learning_rate": 3.725214553582147e-06, "loss": 0.8922, "step": 21768 }, { "epoch": 0.13, "learning_rate": 3.724375315214883e-06, "loss": 0.8618, "step": 21769 }, { "epoch": 0.13, "learning_rate": 3.723536149759581e-06, "loss": 0.8496, "step": 21770 }, { "epoch": 0.13, "learning_rate": 3.722697057225991e-06, "loss": 0.8766, "step": 21771 }, { "epoch": 0.13, "learning_rate": 3.7218580376238622e-06, "loss": 0.8761, "step": 21772 }, { "epoch": 0.13, "learning_rate": 3.721019090962945e-06, "loss": 0.8816, "step": 21773 }, { "epoch": 0.13, "learning_rate": 3.72018021725298e-06, "loss": 0.882, "step": 21774 }, { "epoch": 0.13, "learning_rate": 3.719341416503718e-06, "loss": 0.8776, "step": 21775 }, { "epoch": 0.13, "learning_rate": 3.718502688724902e-06, "loss": 0.8857, "step": 21776 }, { "epoch": 0.13, "learning_rate": 3.717664033926279e-06, "loss": 0.8543, "step": 21777 }, { "epoch": 0.13, "learning_rate": 3.716825452117593e-06, "loss": 0.8813, "step": 21778 }, { "epoch": 0.13, "learning_rate": 3.7159869433085837e-06, "loss": 0.8633, "step": 21779 }, { "epoch": 0.13, "learning_rate": 3.715148507509e-06, "loss": 0.8715, "step": 21780 }, { "epoch": 0.13, "learning_rate": 3.7143101447285734e-06, "loss": 0.8706, "step": 21781 }, { "epoch": 0.13, "learning_rate": 3.7134718549770464e-06, "loss": 0.8786, "step": 21782 }, { "epoch": 0.13, "learning_rate": 3.7126336382641647e-06, "loss": 0.8324, "step": 21783 }, { "epoch": 0.13, "learning_rate": 3.711795494599666e-06, "loss": 0.8316, "step": 21784 }, { "epoch": 0.13, "learning_rate": 3.7109574239932833e-06, "loss": 0.8454, "step": 21785 }, { "epoch": 0.13, "learning_rate": 3.710119426454755e-06, "loss": 0.8494, "step": 21786 }, { "epoch": 0.13, "learning_rate": 3.709281501993821e-06, "loss": 0.8724, "step": 21787 }, { "epoch": 0.13, "learning_rate": 3.708443650620206e-06, "loss": 0.86, "step": 21788 }, { "epoch": 0.13, "learning_rate": 3.7076058723436557e-06, "loss": 0.8901, "step": 21789 }, { "epoch": 0.13, "learning_rate": 3.706768167173901e-06, "loss": 0.8731, "step": 21790 }, { "epoch": 0.13, "learning_rate": 3.7059305351206754e-06, "loss": 0.8718, "step": 21791 }, { "epoch": 0.13, "learning_rate": 3.705092976193705e-06, "loss": 0.863, "step": 21792 }, { "epoch": 0.13, "learning_rate": 3.704255490402725e-06, "loss": 0.8426, "step": 21793 }, { "epoch": 0.13, "learning_rate": 3.703418077757466e-06, "loss": 0.8582, "step": 21794 }, { "epoch": 0.13, "learning_rate": 3.702580738267655e-06, "loss": 0.8527, "step": 21795 }, { "epoch": 0.13, "learning_rate": 3.7017434719430224e-06, "loss": 0.8379, "step": 21796 }, { "epoch": 0.13, "learning_rate": 3.7009062787932948e-06, "loss": 0.8558, "step": 21797 }, { "epoch": 0.13, "learning_rate": 3.7000691588282022e-06, "loss": 0.8705, "step": 21798 }, { "epoch": 0.13, "learning_rate": 3.699232112057465e-06, "loss": 0.8523, "step": 21799 }, { "epoch": 0.13, "learning_rate": 3.6983951384908102e-06, "loss": 0.8768, "step": 21800 }, { "epoch": 0.13, "learning_rate": 3.6975582381379625e-06, "loss": 0.8865, "step": 21801 }, { "epoch": 0.13, "learning_rate": 3.696721411008647e-06, "loss": 0.8865, "step": 21802 }, { "epoch": 0.13, "learning_rate": 3.6958846571125827e-06, "loss": 0.8809, "step": 21803 }, { "epoch": 0.13, "learning_rate": 3.695047976459494e-06, "loss": 0.8377, "step": 21804 }, { "epoch": 0.13, "learning_rate": 3.694211369059104e-06, "loss": 0.8583, "step": 21805 }, { "epoch": 0.13, "learning_rate": 3.693374834921123e-06, "loss": 0.8438, "step": 21806 }, { "epoch": 0.14, "learning_rate": 3.692538374055279e-06, "loss": 0.8252, "step": 21807 }, { "epoch": 0.14, "learning_rate": 3.691701986471291e-06, "loss": 0.8627, "step": 21808 }, { "epoch": 0.14, "learning_rate": 3.6908656721788706e-06, "loss": 0.86, "step": 21809 }, { "epoch": 0.14, "learning_rate": 3.690029431187737e-06, "loss": 0.8333, "step": 21810 }, { "epoch": 0.14, "learning_rate": 3.6891932635076057e-06, "loss": 0.848, "step": 21811 }, { "epoch": 0.14, "learning_rate": 3.688357169148191e-06, "loss": 0.887, "step": 21812 }, { "epoch": 0.14, "learning_rate": 3.6875211481192075e-06, "loss": 0.857, "step": 21813 }, { "epoch": 0.14, "learning_rate": 3.686685200430369e-06, "loss": 0.8601, "step": 21814 }, { "epoch": 0.14, "learning_rate": 3.6858493260913897e-06, "loss": 0.847, "step": 21815 }, { "epoch": 0.14, "learning_rate": 3.6850135251119745e-06, "loss": 0.9034, "step": 21816 }, { "epoch": 0.14, "learning_rate": 3.6841777975018387e-06, "loss": 0.8672, "step": 21817 }, { "epoch": 0.14, "learning_rate": 3.6833421432706906e-06, "loss": 0.8842, "step": 21818 }, { "epoch": 0.14, "learning_rate": 3.682506562428241e-06, "loss": 0.8248, "step": 21819 }, { "epoch": 0.14, "learning_rate": 3.6816710549841948e-06, "loss": 0.8652, "step": 21820 }, { "epoch": 0.14, "learning_rate": 3.680835620948262e-06, "loss": 0.8358, "step": 21821 }, { "epoch": 0.14, "learning_rate": 3.680000260330151e-06, "loss": 0.8901, "step": 21822 }, { "epoch": 0.14, "learning_rate": 3.6791649731395607e-06, "loss": 0.8802, "step": 21823 }, { "epoch": 0.14, "learning_rate": 3.678329759386199e-06, "loss": 0.8405, "step": 21824 }, { "epoch": 0.14, "learning_rate": 3.6774946190797655e-06, "loss": 0.8759, "step": 21825 }, { "epoch": 0.14, "learning_rate": 3.6766595522299755e-06, "loss": 0.8538, "step": 21826 }, { "epoch": 0.14, "learning_rate": 3.675824558846518e-06, "loss": 0.8675, "step": 21827 }, { "epoch": 0.14, "learning_rate": 3.6749896389390994e-06, "loss": 0.8538, "step": 21828 }, { "epoch": 0.14, "learning_rate": 3.6741547925174205e-06, "loss": 0.8606, "step": 21829 }, { "epoch": 0.14, "learning_rate": 3.6733200195911823e-06, "loss": 0.8703, "step": 21830 }, { "epoch": 0.14, "learning_rate": 3.672485320170075e-06, "loss": 0.8425, "step": 21831 }, { "epoch": 0.14, "learning_rate": 3.6716506942638055e-06, "loss": 0.8974, "step": 21832 }, { "epoch": 0.14, "learning_rate": 3.670816141882071e-06, "loss": 0.8468, "step": 21833 }, { "epoch": 0.14, "learning_rate": 3.6699816630345607e-06, "loss": 0.8569, "step": 21834 }, { "epoch": 0.14, "learning_rate": 3.6691472577309737e-06, "loss": 0.8523, "step": 21835 }, { "epoch": 0.14, "learning_rate": 3.6683129259810046e-06, "loss": 0.858, "step": 21836 }, { "epoch": 0.14, "learning_rate": 3.6674786677943452e-06, "loss": 0.8562, "step": 21837 }, { "epoch": 0.14, "learning_rate": 3.666644483180689e-06, "loss": 0.8685, "step": 21838 }, { "epoch": 0.14, "learning_rate": 3.6658103721497284e-06, "loss": 0.8488, "step": 21839 }, { "epoch": 0.14, "learning_rate": 3.664976334711158e-06, "loss": 0.8685, "step": 21840 }, { "epoch": 0.14, "learning_rate": 3.6641423708746606e-06, "loss": 0.8388, "step": 21841 }, { "epoch": 0.14, "learning_rate": 3.6633084806499275e-06, "loss": 0.8604, "step": 21842 }, { "epoch": 0.14, "learning_rate": 3.662474664046648e-06, "loss": 0.866, "step": 21843 }, { "epoch": 0.14, "learning_rate": 3.661640921074511e-06, "loss": 0.873, "step": 21844 }, { "epoch": 0.14, "learning_rate": 3.6608072517432014e-06, "loss": 0.8471, "step": 21845 }, { "epoch": 0.14, "learning_rate": 3.6599736560624054e-06, "loss": 0.9061, "step": 21846 }, { "epoch": 0.14, "learning_rate": 3.659140134041812e-06, "loss": 0.8454, "step": 21847 }, { "epoch": 0.14, "learning_rate": 3.6583066856910978e-06, "loss": 0.8651, "step": 21848 }, { "epoch": 0.14, "learning_rate": 3.6574733110199468e-06, "loss": 0.8668, "step": 21849 }, { "epoch": 0.14, "learning_rate": 3.6566400100380474e-06, "loss": 0.8897, "step": 21850 }, { "epoch": 0.14, "learning_rate": 3.655806782755081e-06, "loss": 0.8313, "step": 21851 }, { "epoch": 0.14, "learning_rate": 3.6549736291807214e-06, "loss": 0.8805, "step": 21852 }, { "epoch": 0.14, "learning_rate": 3.654140549324653e-06, "loss": 0.9043, "step": 21853 }, { "epoch": 0.14, "learning_rate": 3.653307543196558e-06, "loss": 0.8649, "step": 21854 }, { "epoch": 0.14, "learning_rate": 3.652474610806103e-06, "loss": 0.8216, "step": 21855 }, { "epoch": 0.14, "learning_rate": 3.6516417521629757e-06, "loss": 0.8943, "step": 21856 }, { "epoch": 0.14, "learning_rate": 3.65080896727685e-06, "loss": 0.89, "step": 21857 }, { "epoch": 0.14, "learning_rate": 3.6499762561574046e-06, "loss": 0.8679, "step": 21858 }, { "epoch": 0.14, "learning_rate": 3.649143618814307e-06, "loss": 0.8908, "step": 21859 }, { "epoch": 0.14, "learning_rate": 3.6483110552572344e-06, "loss": 0.8642, "step": 21860 }, { "epoch": 0.14, "learning_rate": 3.647478565495859e-06, "loss": 0.861, "step": 21861 }, { "epoch": 0.14, "learning_rate": 3.646646149539855e-06, "loss": 0.8575, "step": 21862 }, { "epoch": 0.14, "learning_rate": 3.6458138073988925e-06, "loss": 0.8351, "step": 21863 }, { "epoch": 0.14, "learning_rate": 3.6449815390826415e-06, "loss": 0.8785, "step": 21864 }, { "epoch": 0.14, "learning_rate": 3.644149344600775e-06, "loss": 0.8672, "step": 21865 }, { "epoch": 0.14, "learning_rate": 3.6433172239629556e-06, "loss": 0.8702, "step": 21866 }, { "epoch": 0.14, "learning_rate": 3.6424851771788537e-06, "loss": 0.8453, "step": 21867 }, { "epoch": 0.14, "learning_rate": 3.6416532042581366e-06, "loss": 0.8393, "step": 21868 }, { "epoch": 0.14, "learning_rate": 3.6408213052104704e-06, "loss": 0.8781, "step": 21869 }, { "epoch": 0.14, "learning_rate": 3.6399894800455202e-06, "loss": 0.8569, "step": 21870 }, { "epoch": 0.14, "learning_rate": 3.639157728772951e-06, "loss": 0.8568, "step": 21871 }, { "epoch": 0.14, "learning_rate": 3.638326051402429e-06, "loss": 0.8283, "step": 21872 }, { "epoch": 0.14, "learning_rate": 3.6374944479436093e-06, "loss": 0.8557, "step": 21873 }, { "epoch": 0.14, "learning_rate": 3.636662918406155e-06, "loss": 0.8666, "step": 21874 }, { "epoch": 0.14, "learning_rate": 3.635831462799735e-06, "loss": 0.8249, "step": 21875 }, { "epoch": 0.14, "learning_rate": 3.635000081134006e-06, "loss": 0.899, "step": 21876 }, { "epoch": 0.14, "learning_rate": 3.6341687734186236e-06, "loss": 0.873, "step": 21877 }, { "epoch": 0.14, "learning_rate": 3.6333375396632475e-06, "loss": 0.8442, "step": 21878 }, { "epoch": 0.14, "learning_rate": 3.632506379877536e-06, "loss": 0.8381, "step": 21879 }, { "epoch": 0.14, "learning_rate": 3.6316752940711464e-06, "loss": 0.841, "step": 21880 }, { "epoch": 0.14, "learning_rate": 3.6308442822537337e-06, "loss": 0.8896, "step": 21881 }, { "epoch": 0.14, "learning_rate": 3.6300133444349528e-06, "loss": 0.8711, "step": 21882 }, { "epoch": 0.14, "learning_rate": 3.6291824806244614e-06, "loss": 0.8259, "step": 21883 }, { "epoch": 0.14, "learning_rate": 3.628351690831906e-06, "loss": 0.9045, "step": 21884 }, { "epoch": 0.14, "learning_rate": 3.6275209750669414e-06, "loss": 0.8589, "step": 21885 }, { "epoch": 0.14, "learning_rate": 3.62669033333922e-06, "loss": 0.8905, "step": 21886 }, { "epoch": 0.14, "learning_rate": 3.6258597656583927e-06, "loss": 0.8587, "step": 21887 }, { "epoch": 0.14, "learning_rate": 3.6250292720341095e-06, "loss": 0.8593, "step": 21888 }, { "epoch": 0.14, "learning_rate": 3.6241988524760173e-06, "loss": 0.8315, "step": 21889 }, { "epoch": 0.14, "learning_rate": 3.6233685069937698e-06, "loss": 0.8977, "step": 21890 }, { "epoch": 0.14, "learning_rate": 3.622538235597006e-06, "loss": 0.8784, "step": 21891 }, { "epoch": 0.14, "learning_rate": 3.6217080382953775e-06, "loss": 0.9148, "step": 21892 }, { "epoch": 0.14, "learning_rate": 3.6208779150985264e-06, "loss": 0.8989, "step": 21893 }, { "epoch": 0.14, "learning_rate": 3.620047866016101e-06, "loss": 0.8459, "step": 21894 }, { "epoch": 0.14, "learning_rate": 3.6192178910577425e-06, "loss": 0.902, "step": 21895 }, { "epoch": 0.14, "learning_rate": 3.6183879902330944e-06, "loss": 0.885, "step": 21896 }, { "epoch": 0.14, "learning_rate": 3.617558163551802e-06, "loss": 0.8428, "step": 21897 }, { "epoch": 0.14, "learning_rate": 3.616728411023498e-06, "loss": 0.915, "step": 21898 }, { "epoch": 0.14, "learning_rate": 3.61589873265783e-06, "loss": 0.85, "step": 21899 }, { "epoch": 0.14, "learning_rate": 3.61506912846444e-06, "loss": 0.8873, "step": 21900 }, { "epoch": 0.14, "learning_rate": 3.6142395984529576e-06, "loss": 0.8417, "step": 21901 }, { "epoch": 0.14, "learning_rate": 3.613410142633026e-06, "loss": 0.8775, "step": 21902 }, { "epoch": 0.14, "learning_rate": 3.6125807610142805e-06, "loss": 0.8367, "step": 21903 }, { "epoch": 0.14, "learning_rate": 3.611751453606358e-06, "loss": 0.8715, "step": 21904 }, { "epoch": 0.14, "learning_rate": 3.610922220418893e-06, "loss": 0.8148, "step": 21905 }, { "epoch": 0.14, "learning_rate": 3.6100930614615204e-06, "loss": 0.9464, "step": 21906 }, { "epoch": 0.14, "learning_rate": 3.6092639767438763e-06, "loss": 0.8833, "step": 21907 }, { "epoch": 0.14, "learning_rate": 3.608434966275587e-06, "loss": 0.8817, "step": 21908 }, { "epoch": 0.14, "learning_rate": 3.6076060300662876e-06, "loss": 0.8484, "step": 21909 }, { "epoch": 0.14, "learning_rate": 3.606777168125608e-06, "loss": 0.8474, "step": 21910 }, { "epoch": 0.14, "learning_rate": 3.60594838046318e-06, "loss": 0.8894, "step": 21911 }, { "epoch": 0.14, "learning_rate": 3.60511966708863e-06, "loss": 0.8375, "step": 21912 }, { "epoch": 0.14, "learning_rate": 3.6042910280115883e-06, "loss": 0.853, "step": 21913 }, { "epoch": 0.14, "learning_rate": 3.603462463241686e-06, "loss": 0.8597, "step": 21914 }, { "epoch": 0.14, "learning_rate": 3.6026339727885407e-06, "loss": 0.8215, "step": 21915 }, { "epoch": 0.14, "learning_rate": 3.601805556661783e-06, "loss": 0.8637, "step": 21916 }, { "epoch": 0.14, "learning_rate": 3.6009772148710333e-06, "loss": 0.848, "step": 21917 }, { "epoch": 0.14, "learning_rate": 3.600148947425928e-06, "loss": 0.8591, "step": 21918 }, { "epoch": 0.14, "learning_rate": 3.5993207543360765e-06, "loss": 0.8891, "step": 21919 }, { "epoch": 0.14, "learning_rate": 3.598492635611106e-06, "loss": 0.8847, "step": 21920 }, { "epoch": 0.14, "learning_rate": 3.597664591260639e-06, "loss": 0.8514, "step": 21921 }, { "epoch": 0.14, "learning_rate": 3.596836621294295e-06, "loss": 0.9207, "step": 21922 }, { "epoch": 0.14, "learning_rate": 3.5960087257216926e-06, "loss": 0.8462, "step": 21923 }, { "epoch": 0.14, "learning_rate": 3.5951809045524523e-06, "loss": 0.8865, "step": 21924 }, { "epoch": 0.14, "learning_rate": 3.5943531577961932e-06, "loss": 0.8769, "step": 21925 }, { "epoch": 0.14, "learning_rate": 3.593525485462528e-06, "loss": 0.8445, "step": 21926 }, { "epoch": 0.14, "learning_rate": 3.592697887561074e-06, "loss": 0.8386, "step": 21927 }, { "epoch": 0.14, "learning_rate": 3.5918703641014474e-06, "loss": 0.8621, "step": 21928 }, { "epoch": 0.14, "learning_rate": 3.5910429150932633e-06, "loss": 0.8683, "step": 21929 }, { "epoch": 0.14, "learning_rate": 3.5902155405461346e-06, "loss": 0.8491, "step": 21930 }, { "epoch": 0.14, "learning_rate": 3.5893882404696734e-06, "loss": 0.871, "step": 21931 }, { "epoch": 0.14, "learning_rate": 3.588561014873496e-06, "loss": 0.8762, "step": 21932 }, { "epoch": 0.14, "learning_rate": 3.5877338637672053e-06, "loss": 0.8489, "step": 21933 }, { "epoch": 0.14, "learning_rate": 3.586906787160416e-06, "loss": 0.894, "step": 21934 }, { "epoch": 0.14, "learning_rate": 3.5860797850627373e-06, "loss": 0.8593, "step": 21935 }, { "epoch": 0.14, "learning_rate": 3.5852528574837754e-06, "loss": 0.8982, "step": 21936 }, { "epoch": 0.14, "learning_rate": 3.5844260044331403e-06, "loss": 0.8832, "step": 21937 }, { "epoch": 0.14, "learning_rate": 3.5835992259204376e-06, "loss": 0.8939, "step": 21938 }, { "epoch": 0.14, "learning_rate": 3.5827725219552765e-06, "loss": 0.8405, "step": 21939 }, { "epoch": 0.14, "learning_rate": 3.581945892547256e-06, "loss": 0.8759, "step": 21940 }, { "epoch": 0.14, "learning_rate": 3.5811193377059783e-06, "loss": 0.885, "step": 21941 }, { "epoch": 0.14, "learning_rate": 3.5802928574410544e-06, "loss": 0.9192, "step": 21942 }, { "epoch": 0.14, "learning_rate": 3.579466451762086e-06, "loss": 0.8697, "step": 21943 }, { "epoch": 0.14, "learning_rate": 3.578640120678668e-06, "loss": 0.8647, "step": 21944 }, { "epoch": 0.14, "learning_rate": 3.5778138642004056e-06, "loss": 0.8435, "step": 21945 }, { "epoch": 0.14, "learning_rate": 3.5769876823369e-06, "loss": 0.8792, "step": 21946 }, { "epoch": 0.14, "learning_rate": 3.5761615750977406e-06, "loss": 0.8636, "step": 21947 }, { "epoch": 0.14, "learning_rate": 3.5753355424925364e-06, "loss": 0.8761, "step": 21948 }, { "epoch": 0.14, "learning_rate": 3.57450958453088e-06, "loss": 0.8755, "step": 21949 }, { "epoch": 0.14, "learning_rate": 3.57368370122237e-06, "loss": 0.8841, "step": 21950 }, { "epoch": 0.14, "learning_rate": 3.5728578925765968e-06, "loss": 0.8269, "step": 21951 }, { "epoch": 0.14, "learning_rate": 3.572032158603158e-06, "loss": 0.8825, "step": 21952 }, { "epoch": 0.14, "learning_rate": 3.571206499311647e-06, "loss": 0.8209, "step": 21953 }, { "epoch": 0.14, "learning_rate": 3.5703809147116552e-06, "loss": 0.8683, "step": 21954 }, { "epoch": 0.14, "learning_rate": 3.569555404812777e-06, "loss": 0.8576, "step": 21955 }, { "epoch": 0.14, "learning_rate": 3.5687299696246012e-06, "loss": 0.8775, "step": 21956 }, { "epoch": 0.14, "learning_rate": 3.5679046091567225e-06, "loss": 0.8841, "step": 21957 }, { "epoch": 0.14, "learning_rate": 3.567079323418724e-06, "loss": 0.8347, "step": 21958 }, { "epoch": 0.14, "learning_rate": 3.5662541124201953e-06, "loss": 0.841, "step": 21959 }, { "epoch": 0.14, "learning_rate": 3.565428976170725e-06, "loss": 0.8892, "step": 21960 }, { "epoch": 0.14, "learning_rate": 3.5646039146799004e-06, "loss": 0.8594, "step": 21961 }, { "epoch": 0.14, "learning_rate": 3.5637789279573077e-06, "loss": 0.8375, "step": 21962 }, { "epoch": 0.14, "learning_rate": 3.56295401601253e-06, "loss": 0.8488, "step": 21963 }, { "epoch": 0.14, "learning_rate": 3.562129178855156e-06, "loss": 0.8792, "step": 21964 }, { "epoch": 0.14, "learning_rate": 3.5613044164947617e-06, "loss": 0.8784, "step": 21965 }, { "epoch": 0.14, "learning_rate": 3.5604797289409297e-06, "loss": 0.82, "step": 21966 }, { "epoch": 0.14, "learning_rate": 3.5596551162032487e-06, "loss": 0.8038, "step": 21967 }, { "epoch": 0.14, "learning_rate": 3.558830578291298e-06, "loss": 0.9121, "step": 21968 }, { "epoch": 0.14, "learning_rate": 3.5580061152146516e-06, "loss": 0.8519, "step": 21969 }, { "epoch": 0.14, "learning_rate": 3.557181726982891e-06, "loss": 0.8607, "step": 21970 }, { "epoch": 0.14, "learning_rate": 3.5563574136055946e-06, "loss": 0.8508, "step": 21971 }, { "epoch": 0.14, "learning_rate": 3.5555331750923396e-06, "loss": 0.8663, "step": 21972 }, { "epoch": 0.14, "learning_rate": 3.5547090114527016e-06, "loss": 0.8369, "step": 21973 }, { "epoch": 0.14, "learning_rate": 3.553884922696257e-06, "loss": 0.8244, "step": 21974 }, { "epoch": 0.14, "learning_rate": 3.553060908832583e-06, "loss": 0.8594, "step": 21975 }, { "epoch": 0.14, "learning_rate": 3.5522369698712465e-06, "loss": 0.8266, "step": 21976 }, { "epoch": 0.14, "learning_rate": 3.5514131058218235e-06, "loss": 0.8815, "step": 21977 }, { "epoch": 0.14, "learning_rate": 3.5505893166938853e-06, "loss": 0.8674, "step": 21978 }, { "epoch": 0.14, "learning_rate": 3.549765602497005e-06, "loss": 0.8563, "step": 21979 }, { "epoch": 0.14, "learning_rate": 3.5489419632407508e-06, "loss": 0.8339, "step": 21980 }, { "epoch": 0.14, "learning_rate": 3.5481183989346924e-06, "loss": 0.8578, "step": 21981 }, { "epoch": 0.14, "learning_rate": 3.5472949095884024e-06, "loss": 0.8468, "step": 21982 }, { "epoch": 0.14, "learning_rate": 3.546471495211441e-06, "loss": 0.8272, "step": 21983 }, { "epoch": 0.14, "learning_rate": 3.545648155813374e-06, "loss": 0.8618, "step": 21984 }, { "epoch": 0.14, "learning_rate": 3.5448248914037786e-06, "loss": 0.8822, "step": 21985 }, { "epoch": 0.14, "learning_rate": 3.544001701992209e-06, "loss": 0.9111, "step": 21986 }, { "epoch": 0.14, "learning_rate": 3.5431785875882328e-06, "loss": 0.8582, "step": 21987 }, { "epoch": 0.14, "learning_rate": 3.5423555482014137e-06, "loss": 0.8885, "step": 21988 }, { "epoch": 0.14, "learning_rate": 3.5415325838413164e-06, "loss": 0.8746, "step": 21989 }, { "epoch": 0.14, "learning_rate": 3.5407096945174914e-06, "loss": 0.8683, "step": 21990 }, { "epoch": 0.14, "learning_rate": 3.5398868802395113e-06, "loss": 0.8038, "step": 21991 }, { "epoch": 0.14, "learning_rate": 3.5390641410169355e-06, "loss": 0.8429, "step": 21992 }, { "epoch": 0.14, "learning_rate": 3.5382414768593153e-06, "loss": 0.8441, "step": 21993 }, { "epoch": 0.14, "learning_rate": 3.5374188877762127e-06, "loss": 0.8913, "step": 21994 }, { "epoch": 0.14, "learning_rate": 3.5365963737771845e-06, "loss": 0.8379, "step": 21995 }, { "epoch": 0.14, "learning_rate": 3.535773934871787e-06, "loss": 0.881, "step": 21996 }, { "epoch": 0.14, "learning_rate": 3.534951571069574e-06, "loss": 0.8528, "step": 21997 }, { "epoch": 0.14, "learning_rate": 3.534129282380102e-06, "loss": 0.8333, "step": 21998 }, { "epoch": 0.14, "learning_rate": 3.533307068812928e-06, "loss": 0.8566, "step": 21999 }, { "epoch": 0.14, "learning_rate": 3.5324849303775965e-06, "loss": 0.8774, "step": 22000 }, { "epoch": 0.14, "learning_rate": 3.531662867083663e-06, "loss": 0.8613, "step": 22001 }, { "epoch": 0.14, "learning_rate": 3.5308408789406788e-06, "loss": 0.8236, "step": 22002 }, { "epoch": 0.14, "learning_rate": 3.530018965958195e-06, "loss": 0.8222, "step": 22003 }, { "epoch": 0.14, "learning_rate": 3.5291971281457593e-06, "loss": 0.8339, "step": 22004 }, { "epoch": 0.14, "learning_rate": 3.5283753655129207e-06, "loss": 0.8717, "step": 22005 }, { "epoch": 0.14, "learning_rate": 3.52755367806923e-06, "loss": 0.8447, "step": 22006 }, { "epoch": 0.14, "learning_rate": 3.5267320658242264e-06, "loss": 0.8635, "step": 22007 }, { "epoch": 0.14, "learning_rate": 3.5259105287874608e-06, "loss": 0.8602, "step": 22008 }, { "epoch": 0.14, "learning_rate": 3.5250890669684723e-06, "loss": 0.8893, "step": 22009 }, { "epoch": 0.14, "learning_rate": 3.5242676803768174e-06, "loss": 0.8639, "step": 22010 }, { "epoch": 0.14, "learning_rate": 3.5234463690220266e-06, "loss": 0.8742, "step": 22011 }, { "epoch": 0.14, "learning_rate": 3.5226251329136475e-06, "loss": 0.8776, "step": 22012 }, { "epoch": 0.14, "learning_rate": 3.521803972061221e-06, "loss": 0.887, "step": 22013 }, { "epoch": 0.14, "learning_rate": 3.5209828864742866e-06, "loss": 0.8656, "step": 22014 }, { "epoch": 0.14, "learning_rate": 3.5201618761623847e-06, "loss": 0.8489, "step": 22015 }, { "epoch": 0.14, "learning_rate": 3.519340941135053e-06, "loss": 0.8549, "step": 22016 }, { "epoch": 0.14, "learning_rate": 3.518520081401835e-06, "loss": 0.9006, "step": 22017 }, { "epoch": 0.14, "learning_rate": 3.517699296972259e-06, "loss": 0.8342, "step": 22018 }, { "epoch": 0.14, "learning_rate": 3.5168785878558655e-06, "loss": 0.8429, "step": 22019 }, { "epoch": 0.14, "learning_rate": 3.5160579540621886e-06, "loss": 0.8463, "step": 22020 }, { "epoch": 0.14, "learning_rate": 3.5152373956007637e-06, "loss": 0.8673, "step": 22021 }, { "epoch": 0.14, "learning_rate": 3.5144169124811233e-06, "loss": 0.8663, "step": 22022 }, { "epoch": 0.14, "learning_rate": 3.513596504712801e-06, "loss": 0.8892, "step": 22023 }, { "epoch": 0.14, "learning_rate": 3.5127761723053313e-06, "loss": 0.873, "step": 22024 }, { "epoch": 0.14, "learning_rate": 3.511955915268239e-06, "loss": 0.8605, "step": 22025 }, { "epoch": 0.14, "learning_rate": 3.5111357336110564e-06, "loss": 0.8805, "step": 22026 }, { "epoch": 0.14, "learning_rate": 3.5103156273433136e-06, "loss": 0.8459, "step": 22027 }, { "epoch": 0.14, "learning_rate": 3.509495596474538e-06, "loss": 0.8584, "step": 22028 }, { "epoch": 0.14, "learning_rate": 3.5086756410142577e-06, "loss": 0.8729, "step": 22029 }, { "epoch": 0.14, "learning_rate": 3.5078557609719986e-06, "loss": 0.8622, "step": 22030 }, { "epoch": 0.14, "learning_rate": 3.50703595635729e-06, "loss": 0.8187, "step": 22031 }, { "epoch": 0.14, "learning_rate": 3.5062162271796486e-06, "loss": 0.9039, "step": 22032 }, { "epoch": 0.14, "learning_rate": 3.5053965734486006e-06, "loss": 0.9002, "step": 22033 }, { "epoch": 0.14, "learning_rate": 3.5045769951736743e-06, "loss": 0.8825, "step": 22034 }, { "epoch": 0.14, "learning_rate": 3.5037574923643912e-06, "loss": 0.8432, "step": 22035 }, { "epoch": 0.14, "learning_rate": 3.502938065030268e-06, "loss": 0.8741, "step": 22036 }, { "epoch": 0.14, "learning_rate": 3.5021187131808253e-06, "loss": 0.8812, "step": 22037 }, { "epoch": 0.14, "learning_rate": 3.5012994368255894e-06, "loss": 0.8637, "step": 22038 }, { "epoch": 0.14, "learning_rate": 3.500480235974065e-06, "loss": 0.8636, "step": 22039 }, { "epoch": 0.14, "learning_rate": 3.4996611106357837e-06, "loss": 0.8423, "step": 22040 }, { "epoch": 0.14, "learning_rate": 3.4988420608202557e-06, "loss": 0.8379, "step": 22041 }, { "epoch": 0.14, "learning_rate": 3.4980230865370023e-06, "loss": 0.875, "step": 22042 }, { "epoch": 0.14, "learning_rate": 3.497204187795531e-06, "loss": 0.8445, "step": 22043 }, { "epoch": 0.14, "learning_rate": 3.4963853646053604e-06, "loss": 0.8801, "step": 22044 }, { "epoch": 0.14, "learning_rate": 3.4955666169760027e-06, "loss": 0.8759, "step": 22045 }, { "epoch": 0.14, "learning_rate": 3.4947479449169695e-06, "loss": 0.8367, "step": 22046 }, { "epoch": 0.14, "learning_rate": 3.4939293484377757e-06, "loss": 0.8558, "step": 22047 }, { "epoch": 0.14, "learning_rate": 3.4931108275479287e-06, "loss": 0.8702, "step": 22048 }, { "epoch": 0.14, "learning_rate": 3.492292382256942e-06, "loss": 0.8626, "step": 22049 }, { "epoch": 0.14, "learning_rate": 3.49147401257432e-06, "loss": 0.8848, "step": 22050 }, { "epoch": 0.14, "learning_rate": 3.490655718509571e-06, "loss": 0.8183, "step": 22051 }, { "epoch": 0.14, "learning_rate": 3.4898375000722063e-06, "loss": 0.8612, "step": 22052 }, { "epoch": 0.14, "learning_rate": 3.489019357271728e-06, "loss": 0.872, "step": 22053 }, { "epoch": 0.14, "learning_rate": 3.4882012901176433e-06, "loss": 0.9057, "step": 22054 }, { "epoch": 0.14, "learning_rate": 3.4873832986194577e-06, "loss": 0.8291, "step": 22055 }, { "epoch": 0.14, "learning_rate": 3.486565382786676e-06, "loss": 0.874, "step": 22056 }, { "epoch": 0.14, "learning_rate": 3.4857475426287934e-06, "loss": 0.8936, "step": 22057 }, { "epoch": 0.14, "learning_rate": 3.4849297781553205e-06, "loss": 0.8576, "step": 22058 }, { "epoch": 0.14, "learning_rate": 3.484112089375754e-06, "loss": 0.8431, "step": 22059 }, { "epoch": 0.14, "learning_rate": 3.483294476299599e-06, "loss": 0.8654, "step": 22060 }, { "epoch": 0.14, "learning_rate": 3.4824769389363476e-06, "loss": 0.86, "step": 22061 }, { "epoch": 0.14, "learning_rate": 3.4816594772955013e-06, "loss": 0.8684, "step": 22062 }, { "epoch": 0.14, "learning_rate": 3.480842091386558e-06, "loss": 0.8467, "step": 22063 }, { "epoch": 0.14, "learning_rate": 3.4800247812190136e-06, "loss": 0.8603, "step": 22064 }, { "epoch": 0.14, "learning_rate": 3.4792075468023644e-06, "loss": 0.8478, "step": 22065 }, { "epoch": 0.14, "learning_rate": 3.4783903881461054e-06, "loss": 0.9064, "step": 22066 }, { "epoch": 0.14, "learning_rate": 3.4775733052597338e-06, "loss": 0.8284, "step": 22067 }, { "epoch": 0.14, "learning_rate": 3.4767562981527357e-06, "loss": 0.8498, "step": 22068 }, { "epoch": 0.14, "learning_rate": 3.4759393668346065e-06, "loss": 0.8216, "step": 22069 }, { "epoch": 0.14, "learning_rate": 3.475122511314838e-06, "loss": 0.8759, "step": 22070 }, { "epoch": 0.14, "learning_rate": 3.4743057316029207e-06, "loss": 0.8714, "step": 22071 }, { "epoch": 0.14, "learning_rate": 3.4734890277083444e-06, "loss": 0.9058, "step": 22072 }, { "epoch": 0.14, "learning_rate": 3.4726723996405964e-06, "loss": 0.866, "step": 22073 }, { "epoch": 0.14, "learning_rate": 3.47185584740917e-06, "loss": 0.9006, "step": 22074 }, { "epoch": 0.14, "learning_rate": 3.4710393710235434e-06, "loss": 0.8209, "step": 22075 }, { "epoch": 0.14, "learning_rate": 3.4702229704932043e-06, "loss": 0.8749, "step": 22076 }, { "epoch": 0.14, "learning_rate": 3.4694066458276465e-06, "loss": 0.8594, "step": 22077 }, { "epoch": 0.14, "learning_rate": 3.4685903970363457e-06, "loss": 0.8816, "step": 22078 }, { "epoch": 0.14, "learning_rate": 3.467774224128787e-06, "loss": 0.8251, "step": 22079 }, { "epoch": 0.14, "learning_rate": 3.4669581271144535e-06, "loss": 0.8461, "step": 22080 }, { "epoch": 0.14, "learning_rate": 3.466142106002831e-06, "loss": 0.8364, "step": 22081 }, { "epoch": 0.14, "learning_rate": 3.4653261608033893e-06, "loss": 0.839, "step": 22082 }, { "epoch": 0.14, "learning_rate": 3.46451029152562e-06, "loss": 0.8819, "step": 22083 }, { "epoch": 0.14, "learning_rate": 3.463694498179e-06, "loss": 0.8639, "step": 22084 }, { "epoch": 0.14, "learning_rate": 3.4628787807730014e-06, "loss": 0.8601, "step": 22085 }, { "epoch": 0.14, "learning_rate": 3.4620631393171046e-06, "loss": 0.9165, "step": 22086 }, { "epoch": 0.14, "learning_rate": 3.4612475738207865e-06, "loss": 0.8396, "step": 22087 }, { "epoch": 0.14, "learning_rate": 3.4604320842935226e-06, "loss": 0.8454, "step": 22088 }, { "epoch": 0.14, "learning_rate": 3.4596166707447876e-06, "loss": 0.8421, "step": 22089 }, { "epoch": 0.14, "learning_rate": 3.4588013331840542e-06, "loss": 0.8805, "step": 22090 }, { "epoch": 0.14, "learning_rate": 3.457986071620799e-06, "loss": 0.8676, "step": 22091 }, { "epoch": 0.14, "learning_rate": 3.457170886064488e-06, "loss": 0.8501, "step": 22092 }, { "epoch": 0.14, "learning_rate": 3.4563557765245948e-06, "loss": 0.8458, "step": 22093 }, { "epoch": 0.14, "learning_rate": 3.455540743010589e-06, "loss": 0.8627, "step": 22094 }, { "epoch": 0.14, "learning_rate": 3.4547257855319417e-06, "loss": 0.8087, "step": 22095 }, { "epoch": 0.14, "learning_rate": 3.45391090409812e-06, "loss": 0.8796, "step": 22096 }, { "epoch": 0.14, "learning_rate": 3.4530960987185923e-06, "loss": 0.8786, "step": 22097 }, { "epoch": 0.14, "learning_rate": 3.452281369402827e-06, "loss": 0.8205, "step": 22098 }, { "epoch": 0.14, "learning_rate": 3.4514667161602843e-06, "loss": 0.8572, "step": 22099 }, { "epoch": 0.14, "learning_rate": 3.4506521390004332e-06, "loss": 0.8615, "step": 22100 }, { "epoch": 0.14, "learning_rate": 3.4498376379327324e-06, "loss": 0.8752, "step": 22101 }, { "epoch": 0.14, "learning_rate": 3.4490232129666567e-06, "loss": 0.8691, "step": 22102 }, { "epoch": 0.14, "learning_rate": 3.4482088641116574e-06, "loss": 0.8431, "step": 22103 }, { "epoch": 0.14, "learning_rate": 3.447394591377199e-06, "loss": 0.8735, "step": 22104 }, { "epoch": 0.14, "learning_rate": 3.4465803947727426e-06, "loss": 0.8395, "step": 22105 }, { "epoch": 0.14, "learning_rate": 3.445766274307748e-06, "loss": 0.8257, "step": 22106 }, { "epoch": 0.15, "learning_rate": 3.4449522299916726e-06, "loss": 0.8408, "step": 22107 }, { "epoch": 0.15, "learning_rate": 3.444138261833976e-06, "loss": 0.8644, "step": 22108 }, { "epoch": 0.15, "learning_rate": 3.4433243698441165e-06, "loss": 0.859, "step": 22109 }, { "epoch": 0.15, "learning_rate": 3.4425105540315452e-06, "loss": 0.8901, "step": 22110 }, { "epoch": 0.15, "learning_rate": 3.4416968144057183e-06, "loss": 0.8624, "step": 22111 }, { "epoch": 0.15, "learning_rate": 3.4408831509760933e-06, "loss": 0.8584, "step": 22112 }, { "epoch": 0.15, "learning_rate": 3.440069563752121e-06, "loss": 0.9179, "step": 22113 }, { "epoch": 0.15, "learning_rate": 3.439256052743254e-06, "loss": 0.8345, "step": 22114 }, { "epoch": 0.15, "learning_rate": 3.4384426179589458e-06, "loss": 0.8782, "step": 22115 }, { "epoch": 0.15, "learning_rate": 3.437629259408648e-06, "loss": 0.8944, "step": 22116 }, { "epoch": 0.15, "learning_rate": 3.4368159771018063e-06, "loss": 0.9113, "step": 22117 }, { "epoch": 0.15, "learning_rate": 3.4360027710478707e-06, "loss": 0.882, "step": 22118 }, { "epoch": 0.15, "learning_rate": 3.435189641256287e-06, "loss": 0.8441, "step": 22119 }, { "epoch": 0.15, "learning_rate": 3.4343765877365133e-06, "loss": 0.8913, "step": 22120 }, { "epoch": 0.15, "learning_rate": 3.433563610497983e-06, "loss": 0.8909, "step": 22121 }, { "epoch": 0.15, "learning_rate": 3.4327507095501477e-06, "loss": 0.8957, "step": 22122 }, { "epoch": 0.15, "learning_rate": 3.4319378849024542e-06, "loss": 0.8831, "step": 22123 }, { "epoch": 0.15, "learning_rate": 3.4311251365643393e-06, "loss": 0.8621, "step": 22124 }, { "epoch": 0.15, "learning_rate": 3.4303124645452456e-06, "loss": 0.8486, "step": 22125 }, { "epoch": 0.15, "learning_rate": 3.429499868854623e-06, "loss": 0.9324, "step": 22126 }, { "epoch": 0.15, "learning_rate": 3.4286873495019103e-06, "loss": 0.8934, "step": 22127 }, { "epoch": 0.15, "learning_rate": 3.4278749064965422e-06, "loss": 0.8218, "step": 22128 }, { "epoch": 0.15, "learning_rate": 3.4270625398479616e-06, "loss": 0.8283, "step": 22129 }, { "epoch": 0.15, "learning_rate": 3.426250249565606e-06, "loss": 0.8524, "step": 22130 }, { "epoch": 0.15, "learning_rate": 3.4254380356589123e-06, "loss": 0.8687, "step": 22131 }, { "epoch": 0.15, "learning_rate": 3.4246258981373184e-06, "loss": 0.8474, "step": 22132 }, { "epoch": 0.15, "learning_rate": 3.423813837010259e-06, "loss": 0.8581, "step": 22133 }, { "epoch": 0.15, "learning_rate": 3.423001852287172e-06, "loss": 0.8456, "step": 22134 }, { "epoch": 0.15, "learning_rate": 3.4221899439774863e-06, "loss": 0.8442, "step": 22135 }, { "epoch": 0.15, "learning_rate": 3.4213781120906365e-06, "loss": 0.8856, "step": 22136 }, { "epoch": 0.15, "learning_rate": 3.4205663566360547e-06, "loss": 0.8697, "step": 22137 }, { "epoch": 0.15, "learning_rate": 3.4197546776231737e-06, "loss": 0.8584, "step": 22138 }, { "epoch": 0.15, "learning_rate": 3.418943075061422e-06, "loss": 0.8518, "step": 22139 }, { "epoch": 0.15, "learning_rate": 3.4181315489602296e-06, "loss": 0.8568, "step": 22140 }, { "epoch": 0.15, "learning_rate": 3.417320099329028e-06, "loss": 0.8535, "step": 22141 }, { "epoch": 0.15, "learning_rate": 3.41650872617724e-06, "loss": 0.8855, "step": 22142 }, { "epoch": 0.15, "learning_rate": 3.415697429514293e-06, "loss": 0.8774, "step": 22143 }, { "epoch": 0.15, "learning_rate": 3.414886209349615e-06, "loss": 0.8504, "step": 22144 }, { "epoch": 0.15, "learning_rate": 3.4140750656926304e-06, "loss": 0.8519, "step": 22145 }, { "epoch": 0.15, "learning_rate": 3.4132639985527627e-06, "loss": 0.8298, "step": 22146 }, { "epoch": 0.15, "learning_rate": 3.4124530079394356e-06, "loss": 0.8647, "step": 22147 }, { "epoch": 0.15, "learning_rate": 3.4116420938620743e-06, "loss": 0.8278, "step": 22148 }, { "epoch": 0.15, "learning_rate": 3.4108312563300916e-06, "loss": 0.8658, "step": 22149 }, { "epoch": 0.15, "learning_rate": 3.4100204953529157e-06, "loss": 0.8342, "step": 22150 }, { "epoch": 0.15, "learning_rate": 3.4092098109399653e-06, "loss": 0.8599, "step": 22151 }, { "epoch": 0.15, "learning_rate": 3.408399203100662e-06, "loss": 0.8875, "step": 22152 }, { "epoch": 0.15, "learning_rate": 3.407588671844415e-06, "loss": 0.8911, "step": 22153 }, { "epoch": 0.15, "learning_rate": 3.406778217180646e-06, "loss": 0.8703, "step": 22154 }, { "epoch": 0.15, "learning_rate": 3.4059678391187722e-06, "loss": 0.835, "step": 22155 }, { "epoch": 0.15, "learning_rate": 3.405157537668208e-06, "loss": 0.8626, "step": 22156 }, { "epoch": 0.15, "learning_rate": 3.4043473128383664e-06, "loss": 0.883, "step": 22157 }, { "epoch": 0.15, "learning_rate": 3.4035371646386627e-06, "loss": 0.8804, "step": 22158 }, { "epoch": 0.15, "learning_rate": 3.402727093078512e-06, "loss": 0.83, "step": 22159 }, { "epoch": 0.15, "learning_rate": 3.4019170981673178e-06, "loss": 0.8564, "step": 22160 }, { "epoch": 0.15, "learning_rate": 3.401107179914497e-06, "loss": 0.8978, "step": 22161 }, { "epoch": 0.15, "learning_rate": 3.400297338329457e-06, "loss": 0.8669, "step": 22162 }, { "epoch": 0.15, "learning_rate": 3.399487573421608e-06, "loss": 0.8544, "step": 22163 }, { "epoch": 0.15, "learning_rate": 3.3986778852003588e-06, "loss": 0.8484, "step": 22164 }, { "epoch": 0.15, "learning_rate": 3.3978682736751154e-06, "loss": 0.8286, "step": 22165 }, { "epoch": 0.15, "learning_rate": 3.3970587388552865e-06, "loss": 0.8812, "step": 22166 }, { "epoch": 0.15, "learning_rate": 3.3962492807502732e-06, "loss": 0.9021, "step": 22167 }, { "epoch": 0.15, "learning_rate": 3.3954398993694793e-06, "loss": 0.9027, "step": 22168 }, { "epoch": 0.15, "learning_rate": 3.394630594722317e-06, "loss": 0.8876, "step": 22169 }, { "epoch": 0.15, "learning_rate": 3.39382136681818e-06, "loss": 0.8227, "step": 22170 }, { "epoch": 0.15, "learning_rate": 3.3930122156664734e-06, "loss": 0.8322, "step": 22171 }, { "epoch": 0.15, "learning_rate": 3.3922031412765988e-06, "loss": 0.8597, "step": 22172 }, { "epoch": 0.15, "learning_rate": 3.391394143657959e-06, "loss": 0.8589, "step": 22173 }, { "epoch": 0.15, "learning_rate": 3.390585222819942e-06, "loss": 0.8656, "step": 22174 }, { "epoch": 0.15, "learning_rate": 3.3897763787719585e-06, "loss": 0.8594, "step": 22175 }, { "epoch": 0.15, "learning_rate": 3.3889676115234037e-06, "loss": 0.8512, "step": 22176 }, { "epoch": 0.15, "learning_rate": 3.3881589210836686e-06, "loss": 0.8479, "step": 22177 }, { "epoch": 0.15, "learning_rate": 3.387350307462153e-06, "loss": 0.8549, "step": 22178 }, { "epoch": 0.15, "learning_rate": 3.386541770668249e-06, "loss": 0.8616, "step": 22179 }, { "epoch": 0.15, "learning_rate": 3.385733310711352e-06, "loss": 0.8712, "step": 22180 }, { "epoch": 0.15, "learning_rate": 3.3849249276008546e-06, "loss": 0.8868, "step": 22181 }, { "epoch": 0.15, "learning_rate": 3.3841166213461497e-06, "loss": 0.8719, "step": 22182 }, { "epoch": 0.15, "learning_rate": 3.3833083919566313e-06, "loss": 0.8555, "step": 22183 }, { "epoch": 0.15, "learning_rate": 3.382500239441682e-06, "loss": 0.896, "step": 22184 }, { "epoch": 0.15, "learning_rate": 3.3816921638106946e-06, "loss": 0.8551, "step": 22185 }, { "epoch": 0.15, "learning_rate": 3.3808841650730584e-06, "loss": 0.8665, "step": 22186 }, { "epoch": 0.15, "learning_rate": 3.380076243238162e-06, "loss": 0.8834, "step": 22187 }, { "epoch": 0.15, "learning_rate": 3.3792683983153897e-06, "loss": 0.8676, "step": 22188 }, { "epoch": 0.15, "learning_rate": 3.3784606303141277e-06, "loss": 0.869, "step": 22189 }, { "epoch": 0.15, "learning_rate": 3.3776529392437663e-06, "loss": 0.8588, "step": 22190 }, { "epoch": 0.15, "learning_rate": 3.3768453251136803e-06, "loss": 0.8483, "step": 22191 }, { "epoch": 0.15, "learning_rate": 3.376037787933254e-06, "loss": 0.871, "step": 22192 }, { "epoch": 0.15, "learning_rate": 3.375230327711876e-06, "loss": 0.8692, "step": 22193 }, { "epoch": 0.15, "learning_rate": 3.3744229444589284e-06, "loss": 0.8367, "step": 22194 }, { "epoch": 0.15, "learning_rate": 3.3736156381837837e-06, "loss": 0.8384, "step": 22195 }, { "epoch": 0.15, "learning_rate": 3.3728084088958247e-06, "loss": 0.901, "step": 22196 }, { "epoch": 0.15, "learning_rate": 3.3720012566044312e-06, "loss": 0.8805, "step": 22197 }, { "epoch": 0.15, "learning_rate": 3.3711941813189807e-06, "loss": 0.8849, "step": 22198 }, { "epoch": 0.15, "learning_rate": 3.370387183048849e-06, "loss": 0.8335, "step": 22199 }, { "epoch": 0.15, "learning_rate": 3.369580261803412e-06, "loss": 0.8329, "step": 22200 }, { "epoch": 0.15, "learning_rate": 3.3687734175920505e-06, "loss": 0.8588, "step": 22201 }, { "epoch": 0.15, "learning_rate": 3.3679666504241294e-06, "loss": 0.8614, "step": 22202 }, { "epoch": 0.15, "learning_rate": 3.3671599603090253e-06, "loss": 0.849, "step": 22203 }, { "epoch": 0.15, "learning_rate": 3.3663533472561118e-06, "loss": 0.8783, "step": 22204 }, { "epoch": 0.15, "learning_rate": 3.3655468112747604e-06, "loss": 0.863, "step": 22205 }, { "epoch": 0.15, "learning_rate": 3.3647403523743403e-06, "loss": 0.8896, "step": 22206 }, { "epoch": 0.15, "learning_rate": 3.3639339705642225e-06, "loss": 0.844, "step": 22207 }, { "epoch": 0.15, "learning_rate": 3.363127665853779e-06, "loss": 0.8571, "step": 22208 }, { "epoch": 0.15, "learning_rate": 3.3623214382523705e-06, "loss": 0.8606, "step": 22209 }, { "epoch": 0.15, "learning_rate": 3.361515287769367e-06, "loss": 0.8143, "step": 22210 }, { "epoch": 0.15, "learning_rate": 3.360709214414132e-06, "loss": 0.8449, "step": 22211 }, { "epoch": 0.15, "learning_rate": 3.35990321819604e-06, "loss": 0.9153, "step": 22212 }, { "epoch": 0.15, "learning_rate": 3.3590972991244454e-06, "loss": 0.833, "step": 22213 }, { "epoch": 0.15, "learning_rate": 3.3582914572087156e-06, "loss": 0.8544, "step": 22214 }, { "epoch": 0.15, "learning_rate": 3.3574856924582167e-06, "loss": 0.8562, "step": 22215 }, { "epoch": 0.15, "learning_rate": 3.3566800048823024e-06, "loss": 0.8624, "step": 22216 }, { "epoch": 0.15, "learning_rate": 3.355874394490334e-06, "loss": 0.8635, "step": 22217 }, { "epoch": 0.15, "learning_rate": 3.355068861291677e-06, "loss": 0.8375, "step": 22218 }, { "epoch": 0.15, "learning_rate": 3.3542634052956924e-06, "loss": 0.8616, "step": 22219 }, { "epoch": 0.15, "learning_rate": 3.35345802651173e-06, "loss": 0.8681, "step": 22220 }, { "epoch": 0.15, "learning_rate": 3.35265272494915e-06, "loss": 0.8611, "step": 22221 }, { "epoch": 0.15, "learning_rate": 3.35184750061731e-06, "loss": 0.8346, "step": 22222 }, { "epoch": 0.15, "learning_rate": 3.3510423535255644e-06, "loss": 0.8592, "step": 22223 }, { "epoch": 0.15, "learning_rate": 3.350237283683269e-06, "loss": 0.903, "step": 22224 }, { "epoch": 0.15, "learning_rate": 3.3494322910997745e-06, "loss": 0.8821, "step": 22225 }, { "epoch": 0.15, "learning_rate": 3.348627375784439e-06, "loss": 0.8824, "step": 22226 }, { "epoch": 0.15, "learning_rate": 3.347822537746607e-06, "loss": 0.847, "step": 22227 }, { "epoch": 0.15, "learning_rate": 3.3470177769956326e-06, "loss": 0.8147, "step": 22228 }, { "epoch": 0.15, "learning_rate": 3.3462130935408655e-06, "loss": 0.9109, "step": 22229 }, { "epoch": 0.15, "learning_rate": 3.345408487391656e-06, "loss": 0.8662, "step": 22230 }, { "epoch": 0.15, "learning_rate": 3.344603958557351e-06, "loss": 0.869, "step": 22231 }, { "epoch": 0.15, "learning_rate": 3.343799507047297e-06, "loss": 0.8987, "step": 22232 }, { "epoch": 0.15, "learning_rate": 3.3429951328708466e-06, "loss": 0.8281, "step": 22233 }, { "epoch": 0.15, "learning_rate": 3.3421908360373357e-06, "loss": 0.8619, "step": 22234 }, { "epoch": 0.15, "learning_rate": 3.3413866165561094e-06, "loss": 0.842, "step": 22235 }, { "epoch": 0.15, "learning_rate": 3.3405824744365222e-06, "loss": 0.8576, "step": 22236 }, { "epoch": 0.15, "learning_rate": 3.339778409687907e-06, "loss": 0.8706, "step": 22237 }, { "epoch": 0.15, "learning_rate": 3.3389744223196087e-06, "loss": 0.881, "step": 22238 }, { "epoch": 0.15, "learning_rate": 3.338170512340967e-06, "loss": 0.8491, "step": 22239 }, { "epoch": 0.15, "learning_rate": 3.337366679761326e-06, "loss": 0.8929, "step": 22240 }, { "epoch": 0.15, "learning_rate": 3.3365629245900166e-06, "loss": 0.841, "step": 22241 }, { "epoch": 0.15, "learning_rate": 3.335759246836384e-06, "loss": 0.8404, "step": 22242 }, { "epoch": 0.15, "learning_rate": 3.3349556465097644e-06, "loss": 0.8135, "step": 22243 }, { "epoch": 0.15, "learning_rate": 3.3341521236194973e-06, "loss": 0.8855, "step": 22244 }, { "epoch": 0.15, "learning_rate": 3.3333486781749116e-06, "loss": 0.8198, "step": 22245 }, { "epoch": 0.15, "learning_rate": 3.3325453101853445e-06, "loss": 0.8706, "step": 22246 }, { "epoch": 0.15, "learning_rate": 3.3317420196601303e-06, "loss": 0.8524, "step": 22247 }, { "epoch": 0.15, "learning_rate": 3.3309388066086013e-06, "loss": 0.8628, "step": 22248 }, { "epoch": 0.15, "learning_rate": 3.33013567104009e-06, "loss": 0.8634, "step": 22249 }, { "epoch": 0.15, "learning_rate": 3.329332612963928e-06, "loss": 0.9095, "step": 22250 }, { "epoch": 0.15, "learning_rate": 3.328529632389448e-06, "loss": 0.8477, "step": 22251 }, { "epoch": 0.15, "learning_rate": 3.327726729325972e-06, "loss": 0.8976, "step": 22252 }, { "epoch": 0.15, "learning_rate": 3.3269239037828328e-06, "loss": 0.8416, "step": 22253 }, { "epoch": 0.15, "learning_rate": 3.3261211557693583e-06, "loss": 0.9084, "step": 22254 }, { "epoch": 0.15, "learning_rate": 3.3253184852948727e-06, "loss": 0.8451, "step": 22255 }, { "epoch": 0.15, "learning_rate": 3.3245158923687047e-06, "loss": 0.8161, "step": 22256 }, { "epoch": 0.15, "learning_rate": 3.323713377000177e-06, "loss": 0.8713, "step": 22257 }, { "epoch": 0.15, "learning_rate": 3.322910939198617e-06, "loss": 0.8999, "step": 22258 }, { "epoch": 0.15, "learning_rate": 3.322108578973342e-06, "loss": 0.8471, "step": 22259 }, { "epoch": 0.15, "learning_rate": 3.321306296333673e-06, "loss": 0.8245, "step": 22260 }, { "epoch": 0.15, "learning_rate": 3.3205040912889407e-06, "loss": 0.8316, "step": 22261 }, { "epoch": 0.15, "learning_rate": 3.3197019638484574e-06, "loss": 0.8852, "step": 22262 }, { "epoch": 0.15, "learning_rate": 3.318899914021544e-06, "loss": 0.8427, "step": 22263 }, { "epoch": 0.15, "learning_rate": 3.3180979418175186e-06, "loss": 0.8767, "step": 22264 }, { "epoch": 0.15, "learning_rate": 3.3172960472457007e-06, "loss": 0.8717, "step": 22265 }, { "epoch": 0.15, "learning_rate": 3.316494230315406e-06, "loss": 0.8688, "step": 22266 }, { "epoch": 0.15, "learning_rate": 3.3156924910359488e-06, "loss": 0.8597, "step": 22267 }, { "epoch": 0.15, "learning_rate": 3.3148908294166495e-06, "loss": 0.8813, "step": 22268 }, { "epoch": 0.15, "learning_rate": 3.314089245466814e-06, "loss": 0.8756, "step": 22269 }, { "epoch": 0.15, "learning_rate": 3.3132877391957597e-06, "loss": 0.8772, "step": 22270 }, { "epoch": 0.15, "learning_rate": 3.3124863106127967e-06, "loss": 0.86, "step": 22271 }, { "epoch": 0.15, "learning_rate": 3.311684959727238e-06, "loss": 0.8971, "step": 22272 }, { "epoch": 0.15, "learning_rate": 3.310883686548394e-06, "loss": 0.8242, "step": 22273 }, { "epoch": 0.15, "learning_rate": 3.3100824910855745e-06, "loss": 0.8578, "step": 22274 }, { "epoch": 0.15, "learning_rate": 3.309281373348089e-06, "loss": 0.8673, "step": 22275 }, { "epoch": 0.15, "learning_rate": 3.30848033334524e-06, "loss": 0.8602, "step": 22276 }, { "epoch": 0.15, "learning_rate": 3.3076793710863376e-06, "loss": 0.8695, "step": 22277 }, { "epoch": 0.15, "learning_rate": 3.306878486580688e-06, "loss": 0.8841, "step": 22278 }, { "epoch": 0.15, "learning_rate": 3.306077679837595e-06, "loss": 0.8618, "step": 22279 }, { "epoch": 0.15, "learning_rate": 3.305276950866364e-06, "loss": 0.8591, "step": 22280 }, { "epoch": 0.15, "learning_rate": 3.3044762996762956e-06, "loss": 0.8585, "step": 22281 }, { "epoch": 0.15, "learning_rate": 3.303675726276698e-06, "loss": 0.865, "step": 22282 }, { "epoch": 0.15, "learning_rate": 3.302875230676865e-06, "loss": 0.8212, "step": 22283 }, { "epoch": 0.15, "learning_rate": 3.302074812886097e-06, "loss": 0.8729, "step": 22284 }, { "epoch": 0.15, "learning_rate": 3.3012744729136993e-06, "loss": 0.8488, "step": 22285 }, { "epoch": 0.15, "learning_rate": 3.3004742107689702e-06, "loss": 0.8754, "step": 22286 }, { "epoch": 0.15, "learning_rate": 3.2996740264612026e-06, "loss": 0.8548, "step": 22287 }, { "epoch": 0.15, "learning_rate": 3.2988739199996956e-06, "loss": 0.863, "step": 22288 }, { "epoch": 0.15, "learning_rate": 3.298073891393745e-06, "loss": 0.8756, "step": 22289 }, { "epoch": 0.15, "learning_rate": 3.2972739406526466e-06, "loss": 0.8579, "step": 22290 }, { "epoch": 0.15, "learning_rate": 3.2964740677856923e-06, "loss": 0.8811, "step": 22291 }, { "epoch": 0.15, "learning_rate": 3.2956742728021774e-06, "loss": 0.8556, "step": 22292 }, { "epoch": 0.15, "learning_rate": 3.294874555711396e-06, "loss": 0.8635, "step": 22293 }, { "epoch": 0.15, "learning_rate": 3.2940749165226337e-06, "loss": 0.8804, "step": 22294 }, { "epoch": 0.15, "learning_rate": 3.2932753552451846e-06, "loss": 0.8062, "step": 22295 }, { "epoch": 0.15, "learning_rate": 3.2924758718883365e-06, "loss": 0.8723, "step": 22296 }, { "epoch": 0.15, "learning_rate": 3.29167646646138e-06, "loss": 0.8741, "step": 22297 }, { "epoch": 0.15, "learning_rate": 3.2908771389736017e-06, "loss": 0.8619, "step": 22298 }, { "epoch": 0.15, "learning_rate": 3.290077889434289e-06, "loss": 0.8735, "step": 22299 }, { "epoch": 0.15, "learning_rate": 3.2892787178527307e-06, "loss": 0.8792, "step": 22300 }, { "epoch": 0.15, "learning_rate": 3.2884796242382044e-06, "loss": 0.8651, "step": 22301 }, { "epoch": 0.15, "learning_rate": 3.287680608599999e-06, "loss": 0.8997, "step": 22302 }, { "epoch": 0.15, "learning_rate": 3.2868816709473937e-06, "loss": 0.8689, "step": 22303 }, { "epoch": 0.15, "learning_rate": 3.2860828112896812e-06, "loss": 0.8412, "step": 22304 }, { "epoch": 0.15, "learning_rate": 3.285284029636132e-06, "loss": 0.8665, "step": 22305 }, { "epoch": 0.15, "learning_rate": 3.2844853259960308e-06, "loss": 0.8642, "step": 22306 }, { "epoch": 0.15, "learning_rate": 3.2836867003786598e-06, "loss": 0.874, "step": 22307 }, { "epoch": 0.15, "learning_rate": 3.2828881527932923e-06, "loss": 0.8464, "step": 22308 }, { "epoch": 0.15, "learning_rate": 3.2820896832492033e-06, "loss": 0.8568, "step": 22309 }, { "epoch": 0.15, "learning_rate": 3.2812912917556784e-06, "loss": 0.8668, "step": 22310 }, { "epoch": 0.15, "learning_rate": 3.280492978321994e-06, "loss": 0.8541, "step": 22311 }, { "epoch": 0.15, "learning_rate": 3.2796947429574165e-06, "loss": 0.844, "step": 22312 }, { "epoch": 0.15, "learning_rate": 3.2788965856712265e-06, "loss": 0.9011, "step": 22313 }, { "epoch": 0.15, "learning_rate": 3.2780985064726934e-06, "loss": 0.8763, "step": 22314 }, { "epoch": 0.15, "learning_rate": 3.2773005053710915e-06, "loss": 0.8631, "step": 22315 }, { "epoch": 0.15, "learning_rate": 3.2765025823756923e-06, "loss": 0.873, "step": 22316 }, { "epoch": 0.15, "learning_rate": 3.2757047374957662e-06, "loss": 0.8651, "step": 22317 }, { "epoch": 0.15, "learning_rate": 3.2749069707405857e-06, "loss": 0.8692, "step": 22318 }, { "epoch": 0.15, "learning_rate": 3.274109282119413e-06, "loss": 0.8644, "step": 22319 }, { "epoch": 0.15, "learning_rate": 3.273311671641519e-06, "loss": 0.9014, "step": 22320 }, { "epoch": 0.15, "learning_rate": 3.272514139316171e-06, "loss": 0.8663, "step": 22321 }, { "epoch": 0.15, "learning_rate": 3.271716685152635e-06, "loss": 0.8912, "step": 22322 }, { "epoch": 0.15, "learning_rate": 3.2709193091601756e-06, "loss": 0.8199, "step": 22323 }, { "epoch": 0.15, "learning_rate": 3.270122011348057e-06, "loss": 0.8427, "step": 22324 }, { "epoch": 0.15, "learning_rate": 3.2693247917255456e-06, "loss": 0.8878, "step": 22325 }, { "epoch": 0.15, "learning_rate": 3.268527650301898e-06, "loss": 0.8833, "step": 22326 }, { "epoch": 0.15, "learning_rate": 3.2677305870863742e-06, "loss": 0.8747, "step": 22327 }, { "epoch": 0.15, "learning_rate": 3.2669336020882457e-06, "loss": 0.854, "step": 22328 }, { "epoch": 0.15, "learning_rate": 3.266136695316763e-06, "loss": 0.885, "step": 22329 }, { "epoch": 0.15, "learning_rate": 3.2653398667811853e-06, "loss": 0.871, "step": 22330 }, { "epoch": 0.15, "learning_rate": 3.2645431164907725e-06, "loss": 0.8792, "step": 22331 }, { "epoch": 0.15, "learning_rate": 3.2637464444547852e-06, "loss": 0.8628, "step": 22332 }, { "epoch": 0.15, "learning_rate": 3.2629498506824675e-06, "loss": 0.8941, "step": 22333 }, { "epoch": 0.15, "learning_rate": 3.262153335183086e-06, "loss": 0.8726, "step": 22334 }, { "epoch": 0.15, "learning_rate": 3.2613568979658904e-06, "loss": 0.84, "step": 22335 }, { "epoch": 0.15, "learning_rate": 3.2605605390401386e-06, "loss": 0.7927, "step": 22336 }, { "epoch": 0.15, "learning_rate": 3.2597642584150756e-06, "loss": 0.8567, "step": 22337 }, { "epoch": 0.15, "learning_rate": 3.258968056099956e-06, "loss": 0.866, "step": 22338 }, { "epoch": 0.15, "learning_rate": 3.2581719321040294e-06, "loss": 0.9004, "step": 22339 }, { "epoch": 0.15, "learning_rate": 3.2573758864365467e-06, "loss": 0.8645, "step": 22340 }, { "epoch": 0.15, "learning_rate": 3.2565799191067558e-06, "loss": 0.8632, "step": 22341 }, { "epoch": 0.15, "learning_rate": 3.2557840301239062e-06, "loss": 0.8467, "step": 22342 }, { "epoch": 0.15, "learning_rate": 3.254988219497246e-06, "loss": 0.8792, "step": 22343 }, { "epoch": 0.15, "learning_rate": 3.2541924872360152e-06, "loss": 0.8681, "step": 22344 }, { "epoch": 0.15, "learning_rate": 3.2533968333494615e-06, "loss": 0.8558, "step": 22345 }, { "epoch": 0.15, "learning_rate": 3.2526012578468303e-06, "loss": 0.816, "step": 22346 }, { "epoch": 0.15, "learning_rate": 3.2518057607373656e-06, "loss": 0.8373, "step": 22347 }, { "epoch": 0.15, "learning_rate": 3.2510103420303063e-06, "loss": 0.8693, "step": 22348 }, { "epoch": 0.15, "learning_rate": 3.2502150017348967e-06, "loss": 0.8814, "step": 22349 }, { "epoch": 0.15, "learning_rate": 3.24941973986038e-06, "loss": 0.8821, "step": 22350 }, { "epoch": 0.15, "learning_rate": 3.248624556415989e-06, "loss": 0.8638, "step": 22351 }, { "epoch": 0.15, "learning_rate": 3.247829451410962e-06, "loss": 0.8364, "step": 22352 }, { "epoch": 0.15, "learning_rate": 3.247034424854547e-06, "loss": 0.8458, "step": 22353 }, { "epoch": 0.15, "learning_rate": 3.2462394767559714e-06, "loss": 0.8629, "step": 22354 }, { "epoch": 0.15, "learning_rate": 3.2454446071244728e-06, "loss": 0.8503, "step": 22355 }, { "epoch": 0.15, "learning_rate": 3.2446498159692884e-06, "loss": 0.8593, "step": 22356 }, { "epoch": 0.15, "learning_rate": 3.24385510329965e-06, "loss": 0.8933, "step": 22357 }, { "epoch": 0.15, "learning_rate": 3.2430604691247933e-06, "loss": 0.8927, "step": 22358 }, { "epoch": 0.15, "learning_rate": 3.2422659134539493e-06, "loss": 0.8718, "step": 22359 }, { "epoch": 0.15, "learning_rate": 3.2414714362963518e-06, "loss": 0.87, "step": 22360 }, { "epoch": 0.15, "learning_rate": 3.2406770376612263e-06, "loss": 0.8525, "step": 22361 }, { "epoch": 0.15, "learning_rate": 3.239882717557804e-06, "loss": 0.8948, "step": 22362 }, { "epoch": 0.15, "learning_rate": 3.2390884759953156e-06, "loss": 0.8601, "step": 22363 }, { "epoch": 0.15, "learning_rate": 3.238294312982987e-06, "loss": 0.8385, "step": 22364 }, { "epoch": 0.15, "learning_rate": 3.237500228530045e-06, "loss": 0.8679, "step": 22365 }, { "epoch": 0.15, "learning_rate": 3.2367062226457167e-06, "loss": 0.8439, "step": 22366 }, { "epoch": 0.15, "learning_rate": 3.235912295339231e-06, "loss": 0.8342, "step": 22367 }, { "epoch": 0.15, "learning_rate": 3.2351184466198028e-06, "loss": 0.8254, "step": 22368 }, { "epoch": 0.15, "learning_rate": 3.2343246764966606e-06, "loss": 0.8281, "step": 22369 }, { "epoch": 0.15, "learning_rate": 3.233530984979022e-06, "loss": 0.8818, "step": 22370 }, { "epoch": 0.15, "learning_rate": 3.232737372076119e-06, "loss": 0.8375, "step": 22371 }, { "epoch": 0.15, "learning_rate": 3.231943837797162e-06, "loss": 0.889, "step": 22372 }, { "epoch": 0.15, "learning_rate": 3.2311503821513744e-06, "loss": 0.8577, "step": 22373 }, { "epoch": 0.15, "learning_rate": 3.230357005147978e-06, "loss": 0.8813, "step": 22374 }, { "epoch": 0.15, "learning_rate": 3.229563706796183e-06, "loss": 0.8272, "step": 22375 }, { "epoch": 0.15, "learning_rate": 3.2287704871052062e-06, "loss": 0.8719, "step": 22376 }, { "epoch": 0.15, "learning_rate": 3.2279773460842713e-06, "loss": 0.8667, "step": 22377 }, { "epoch": 0.15, "learning_rate": 3.2271842837425917e-06, "loss": 0.8858, "step": 22378 }, { "epoch": 0.15, "learning_rate": 3.226391300089374e-06, "loss": 0.8391, "step": 22379 }, { "epoch": 0.15, "learning_rate": 3.2255983951338387e-06, "loss": 0.8889, "step": 22380 }, { "epoch": 0.15, "learning_rate": 3.2248055688851942e-06, "loss": 0.8441, "step": 22381 }, { "epoch": 0.15, "learning_rate": 3.224012821352652e-06, "loss": 0.8744, "step": 22382 }, { "epoch": 0.15, "learning_rate": 3.2232201525454254e-06, "loss": 0.8847, "step": 22383 }, { "epoch": 0.15, "learning_rate": 3.2224275624727207e-06, "loss": 0.8758, "step": 22384 }, { "epoch": 0.15, "learning_rate": 3.2216350511437512e-06, "loss": 0.8685, "step": 22385 }, { "epoch": 0.15, "learning_rate": 3.2208426185677166e-06, "loss": 0.8633, "step": 22386 }, { "epoch": 0.15, "learning_rate": 3.220050264753829e-06, "loss": 0.868, "step": 22387 }, { "epoch": 0.15, "learning_rate": 3.219257989711293e-06, "loss": 0.815, "step": 22388 }, { "epoch": 0.15, "learning_rate": 3.2184657934493134e-06, "loss": 0.893, "step": 22389 }, { "epoch": 0.15, "learning_rate": 3.2176736759770945e-06, "loss": 0.8473, "step": 22390 }, { "epoch": 0.15, "learning_rate": 3.216881637303839e-06, "loss": 0.8648, "step": 22391 }, { "epoch": 0.15, "learning_rate": 3.216089677438753e-06, "loss": 0.9197, "step": 22392 }, { "epoch": 0.15, "learning_rate": 3.2152977963910302e-06, "loss": 0.8506, "step": 22393 }, { "epoch": 0.15, "learning_rate": 3.2145059941698744e-06, "loss": 0.8681, "step": 22394 }, { "epoch": 0.15, "learning_rate": 3.2137142707844814e-06, "loss": 0.8522, "step": 22395 }, { "epoch": 0.15, "learning_rate": 3.2129226262440614e-06, "loss": 0.8229, "step": 22396 }, { "epoch": 0.15, "learning_rate": 3.2121310605578006e-06, "loss": 0.8365, "step": 22397 }, { "epoch": 0.15, "learning_rate": 3.2113395737348984e-06, "loss": 0.8461, "step": 22398 }, { "epoch": 0.15, "learning_rate": 3.210548165784554e-06, "loss": 0.8913, "step": 22399 }, { "epoch": 0.15, "learning_rate": 3.2097568367159536e-06, "loss": 0.8502, "step": 22400 }, { "epoch": 0.15, "learning_rate": 3.2089655865382996e-06, "loss": 0.8563, "step": 22401 }, { "epoch": 0.15, "learning_rate": 3.208174415260783e-06, "loss": 0.8445, "step": 22402 }, { "epoch": 0.15, "learning_rate": 3.2073833228925977e-06, "loss": 0.8512, "step": 22403 }, { "epoch": 0.15, "learning_rate": 3.2065923094429286e-06, "loss": 0.8632, "step": 22404 }, { "epoch": 0.15, "learning_rate": 3.205801374920969e-06, "loss": 0.8771, "step": 22405 }, { "epoch": 0.15, "learning_rate": 3.2050105193359104e-06, "loss": 0.8697, "step": 22406 }, { "epoch": 0.15, "learning_rate": 3.204219742696938e-06, "loss": 0.8503, "step": 22407 }, { "epoch": 0.16, "learning_rate": 3.203429045013242e-06, "loss": 0.8854, "step": 22408 }, { "epoch": 0.16, "learning_rate": 3.2026384262940067e-06, "loss": 0.8699, "step": 22409 }, { "epoch": 0.16, "learning_rate": 3.2018478865484228e-06, "loss": 0.8534, "step": 22410 }, { "epoch": 0.16, "learning_rate": 3.2010574257856664e-06, "loss": 0.8804, "step": 22411 }, { "epoch": 0.16, "learning_rate": 3.2002670440149277e-06, "loss": 0.8565, "step": 22412 }, { "epoch": 0.16, "learning_rate": 3.1994767412453865e-06, "loss": 0.8502, "step": 22413 }, { "epoch": 0.16, "learning_rate": 3.198686517486226e-06, "loss": 0.8522, "step": 22414 }, { "epoch": 0.16, "learning_rate": 3.1978963727466272e-06, "loss": 0.8736, "step": 22415 }, { "epoch": 0.16, "learning_rate": 3.19710630703577e-06, "loss": 0.8106, "step": 22416 }, { "epoch": 0.16, "learning_rate": 3.196316320362837e-06, "loss": 0.9119, "step": 22417 }, { "epoch": 0.16, "learning_rate": 3.1955264127370002e-06, "loss": 0.8484, "step": 22418 }, { "epoch": 0.16, "learning_rate": 3.194736584167437e-06, "loss": 0.8488, "step": 22419 }, { "epoch": 0.16, "learning_rate": 3.193946834663331e-06, "loss": 0.8551, "step": 22420 }, { "epoch": 0.16, "learning_rate": 3.1931571642338554e-06, "loss": 0.825, "step": 22421 }, { "epoch": 0.16, "learning_rate": 3.1923675728881808e-06, "loss": 0.869, "step": 22422 }, { "epoch": 0.16, "learning_rate": 3.1915780606354828e-06, "loss": 0.869, "step": 22423 }, { "epoch": 0.16, "learning_rate": 3.1907886274849377e-06, "loss": 0.8283, "step": 22424 }, { "epoch": 0.16, "learning_rate": 3.1899992734457076e-06, "loss": 0.9016, "step": 22425 }, { "epoch": 0.16, "learning_rate": 3.1892099985269743e-06, "loss": 0.8358, "step": 22426 }, { "epoch": 0.16, "learning_rate": 3.1884208027379026e-06, "loss": 0.8611, "step": 22427 }, { "epoch": 0.16, "learning_rate": 3.1876316860876655e-06, "loss": 0.8818, "step": 22428 }, { "epoch": 0.16, "learning_rate": 3.1868426485854254e-06, "loss": 0.8554, "step": 22429 }, { "epoch": 0.16, "learning_rate": 3.1860536902403526e-06, "loss": 0.9048, "step": 22430 }, { "epoch": 0.16, "learning_rate": 3.1852648110616138e-06, "loss": 0.8352, "step": 22431 }, { "epoch": 0.16, "learning_rate": 3.184476011058373e-06, "loss": 0.8692, "step": 22432 }, { "epoch": 0.16, "learning_rate": 3.1836872902397954e-06, "loss": 0.8769, "step": 22433 }, { "epoch": 0.16, "learning_rate": 3.182898648615045e-06, "loss": 0.8492, "step": 22434 }, { "epoch": 0.16, "learning_rate": 3.1821100861932865e-06, "loss": 0.846, "step": 22435 }, { "epoch": 0.16, "learning_rate": 3.181321602983677e-06, "loss": 0.871, "step": 22436 }, { "epoch": 0.16, "learning_rate": 3.180533198995379e-06, "loss": 0.8798, "step": 22437 }, { "epoch": 0.16, "learning_rate": 3.1797448742375535e-06, "loss": 0.8525, "step": 22438 }, { "epoch": 0.16, "learning_rate": 3.1789566287193594e-06, "loss": 0.8514, "step": 22439 }, { "epoch": 0.16, "learning_rate": 3.178168462449954e-06, "loss": 0.8449, "step": 22440 }, { "epoch": 0.16, "learning_rate": 3.1773803754384945e-06, "loss": 0.8692, "step": 22441 }, { "epoch": 0.16, "learning_rate": 3.176592367694141e-06, "loss": 0.8478, "step": 22442 }, { "epoch": 0.16, "learning_rate": 3.175804439226039e-06, "loss": 0.8703, "step": 22443 }, { "epoch": 0.16, "learning_rate": 3.1750165900433515e-06, "loss": 0.8225, "step": 22444 }, { "epoch": 0.16, "learning_rate": 3.174228820155234e-06, "loss": 0.8715, "step": 22445 }, { "epoch": 0.16, "learning_rate": 3.17344112957083e-06, "loss": 0.8849, "step": 22446 }, { "epoch": 0.16, "learning_rate": 3.172653518299297e-06, "loss": 0.8554, "step": 22447 }, { "epoch": 0.16, "learning_rate": 3.1718659863497837e-06, "loss": 0.8595, "step": 22448 }, { "epoch": 0.16, "learning_rate": 3.17107853373144e-06, "loss": 0.8448, "step": 22449 }, { "epoch": 0.16, "learning_rate": 3.170291160453416e-06, "loss": 0.8969, "step": 22450 }, { "epoch": 0.16, "learning_rate": 3.1695038665248578e-06, "loss": 0.8317, "step": 22451 }, { "epoch": 0.16, "learning_rate": 3.1687166519549185e-06, "loss": 0.8711, "step": 22452 }, { "epoch": 0.16, "learning_rate": 3.1679295167527346e-06, "loss": 0.8374, "step": 22453 }, { "epoch": 0.16, "learning_rate": 3.1671424609274558e-06, "loss": 0.8456, "step": 22454 }, { "epoch": 0.16, "learning_rate": 3.166355484488227e-06, "loss": 0.8424, "step": 22455 }, { "epoch": 0.16, "learning_rate": 3.165568587444191e-06, "loss": 0.878, "step": 22456 }, { "epoch": 0.16, "learning_rate": 3.164781769804489e-06, "loss": 0.8782, "step": 22457 }, { "epoch": 0.16, "learning_rate": 3.1639950315782643e-06, "loss": 0.8337, "step": 22458 }, { "epoch": 0.16, "learning_rate": 3.16320837277466e-06, "loss": 0.83, "step": 22459 }, { "epoch": 0.16, "learning_rate": 3.162421793402809e-06, "loss": 0.8413, "step": 22460 }, { "epoch": 0.16, "learning_rate": 3.161635293471853e-06, "loss": 0.8783, "step": 22461 }, { "epoch": 0.16, "learning_rate": 3.1608488729909283e-06, "loss": 0.8668, "step": 22462 }, { "epoch": 0.16, "learning_rate": 3.160062531969179e-06, "loss": 0.8478, "step": 22463 }, { "epoch": 0.16, "learning_rate": 3.159276270415732e-06, "loss": 0.8475, "step": 22464 }, { "epoch": 0.16, "learning_rate": 3.1584900883397262e-06, "loss": 0.8622, "step": 22465 }, { "epoch": 0.16, "learning_rate": 3.1577039857502965e-06, "loss": 0.8706, "step": 22466 }, { "epoch": 0.16, "learning_rate": 3.156917962656577e-06, "loss": 0.8495, "step": 22467 }, { "epoch": 0.16, "learning_rate": 3.1561320190676927e-06, "loss": 0.8503, "step": 22468 }, { "epoch": 0.16, "learning_rate": 3.1553461549927823e-06, "loss": 0.8465, "step": 22469 }, { "epoch": 0.16, "learning_rate": 3.154560370440978e-06, "loss": 0.8621, "step": 22470 }, { "epoch": 0.16, "learning_rate": 3.153774665421402e-06, "loss": 0.825, "step": 22471 }, { "epoch": 0.16, "learning_rate": 3.152989039943185e-06, "loss": 0.8987, "step": 22472 }, { "epoch": 0.16, "learning_rate": 3.1522034940154565e-06, "loss": 0.8649, "step": 22473 }, { "epoch": 0.16, "learning_rate": 3.1514180276473427e-06, "loss": 0.8355, "step": 22474 }, { "epoch": 0.16, "learning_rate": 3.150632640847969e-06, "loss": 0.861, "step": 22475 }, { "epoch": 0.16, "learning_rate": 3.1498473336264613e-06, "loss": 0.8547, "step": 22476 }, { "epoch": 0.16, "learning_rate": 3.1490621059919447e-06, "loss": 0.8652, "step": 22477 }, { "epoch": 0.16, "learning_rate": 3.1482769579535365e-06, "loss": 0.8717, "step": 22478 }, { "epoch": 0.16, "learning_rate": 3.1474918895203633e-06, "loss": 0.8239, "step": 22479 }, { "epoch": 0.16, "learning_rate": 3.146706900701545e-06, "loss": 0.8677, "step": 22480 }, { "epoch": 0.16, "learning_rate": 3.145921991506202e-06, "loss": 0.8618, "step": 22481 }, { "epoch": 0.16, "learning_rate": 3.1451371619434534e-06, "loss": 0.8989, "step": 22482 }, { "epoch": 0.16, "learning_rate": 3.144352412022419e-06, "loss": 0.8591, "step": 22483 }, { "epoch": 0.16, "learning_rate": 3.143567741752217e-06, "loss": 0.8689, "step": 22484 }, { "epoch": 0.16, "learning_rate": 3.142783151141959e-06, "loss": 0.8907, "step": 22485 }, { "epoch": 0.16, "learning_rate": 3.1419986402007653e-06, "loss": 0.8469, "step": 22486 }, { "epoch": 0.16, "learning_rate": 3.141214208937744e-06, "loss": 0.8687, "step": 22487 }, { "epoch": 0.16, "learning_rate": 3.140429857362021e-06, "loss": 0.8323, "step": 22488 }, { "epoch": 0.16, "learning_rate": 3.139645585482698e-06, "loss": 0.8838, "step": 22489 }, { "epoch": 0.16, "learning_rate": 3.1388613933088906e-06, "loss": 0.8606, "step": 22490 }, { "epoch": 0.16, "learning_rate": 3.138077280849714e-06, "loss": 0.8947, "step": 22491 }, { "epoch": 0.16, "learning_rate": 3.137293248114267e-06, "loss": 0.8867, "step": 22492 }, { "epoch": 0.16, "learning_rate": 3.1365092951116693e-06, "loss": 0.8941, "step": 22493 }, { "epoch": 0.16, "learning_rate": 3.135725421851026e-06, "loss": 0.8893, "step": 22494 }, { "epoch": 0.16, "learning_rate": 3.134941628341447e-06, "loss": 0.8501, "step": 22495 }, { "epoch": 0.16, "learning_rate": 3.134157914592032e-06, "loss": 0.8124, "step": 22496 }, { "epoch": 0.16, "learning_rate": 3.1333742806118905e-06, "loss": 0.8666, "step": 22497 }, { "epoch": 0.16, "learning_rate": 3.1325907264101265e-06, "loss": 0.8763, "step": 22498 }, { "epoch": 0.16, "learning_rate": 3.131807251995843e-06, "loss": 0.841, "step": 22499 }, { "epoch": 0.16, "learning_rate": 3.1310238573781437e-06, "loss": 0.8686, "step": 22500 }, { "epoch": 0.16, "learning_rate": 3.1302405425661297e-06, "loss": 0.8602, "step": 22501 }, { "epoch": 0.16, "learning_rate": 3.129457307568905e-06, "loss": 0.9, "step": 22502 }, { "epoch": 0.16, "learning_rate": 3.128674152395563e-06, "loss": 0.8786, "step": 22503 }, { "epoch": 0.16, "learning_rate": 3.1278910770552074e-06, "loss": 0.8683, "step": 22504 }, { "epoch": 0.16, "learning_rate": 3.127108081556933e-06, "loss": 0.8818, "step": 22505 }, { "epoch": 0.16, "learning_rate": 3.1263251659098393e-06, "loss": 0.8719, "step": 22506 }, { "epoch": 0.16, "learning_rate": 3.1255423301230214e-06, "loss": 0.8647, "step": 22507 }, { "epoch": 0.16, "learning_rate": 3.1247595742055747e-06, "loss": 0.8278, "step": 22508 }, { "epoch": 0.16, "learning_rate": 3.1239768981665973e-06, "loss": 0.8691, "step": 22509 }, { "epoch": 0.16, "learning_rate": 3.1231943020151758e-06, "loss": 0.86, "step": 22510 }, { "epoch": 0.16, "learning_rate": 3.122411785760402e-06, "loss": 0.8817, "step": 22511 }, { "epoch": 0.16, "learning_rate": 3.1216293494113738e-06, "loss": 0.9068, "step": 22512 }, { "epoch": 0.16, "learning_rate": 3.120846992977182e-06, "loss": 0.828, "step": 22513 }, { "epoch": 0.16, "learning_rate": 3.120064716466912e-06, "loss": 0.8704, "step": 22514 }, { "epoch": 0.16, "learning_rate": 3.1192825198896516e-06, "loss": 0.879, "step": 22515 }, { "epoch": 0.16, "learning_rate": 3.118500403254492e-06, "loss": 0.8777, "step": 22516 }, { "epoch": 0.16, "learning_rate": 3.117718366570518e-06, "loss": 0.8996, "step": 22517 }, { "epoch": 0.16, "learning_rate": 3.116936409846817e-06, "loss": 0.864, "step": 22518 }, { "epoch": 0.16, "learning_rate": 3.1161545330924736e-06, "loss": 0.8631, "step": 22519 }, { "epoch": 0.16, "learning_rate": 3.1153727363165744e-06, "loss": 0.8587, "step": 22520 }, { "epoch": 0.16, "learning_rate": 3.1145910195281957e-06, "loss": 0.8816, "step": 22521 }, { "epoch": 0.16, "learning_rate": 3.1138093827364245e-06, "loss": 0.8663, "step": 22522 }, { "epoch": 0.16, "learning_rate": 3.113027825950341e-06, "loss": 0.8583, "step": 22523 }, { "epoch": 0.16, "learning_rate": 3.1122463491790254e-06, "loss": 0.8578, "step": 22524 }, { "epoch": 0.16, "learning_rate": 3.1114649524315587e-06, "loss": 0.8532, "step": 22525 }, { "epoch": 0.16, "learning_rate": 3.110683635717018e-06, "loss": 0.8507, "step": 22526 }, { "epoch": 0.16, "learning_rate": 3.109902399044483e-06, "loss": 0.8838, "step": 22527 }, { "epoch": 0.16, "learning_rate": 3.1091212424230267e-06, "loss": 0.853, "step": 22528 }, { "epoch": 0.16, "learning_rate": 3.1083401658617263e-06, "loss": 0.8854, "step": 22529 }, { "epoch": 0.16, "learning_rate": 3.107559169369656e-06, "loss": 0.8942, "step": 22530 }, { "epoch": 0.16, "learning_rate": 3.106778252955891e-06, "loss": 0.7975, "step": 22531 }, { "epoch": 0.16, "learning_rate": 3.1059974166295025e-06, "loss": 0.8953, "step": 22532 }, { "epoch": 0.16, "learning_rate": 3.1052166603995647e-06, "loss": 0.8756, "step": 22533 }, { "epoch": 0.16, "learning_rate": 3.1044359842751506e-06, "loss": 0.8294, "step": 22534 }, { "epoch": 0.16, "learning_rate": 3.1036553882653208e-06, "loss": 0.8511, "step": 22535 }, { "epoch": 0.16, "learning_rate": 3.1028748723791534e-06, "loss": 0.8315, "step": 22536 }, { "epoch": 0.16, "learning_rate": 3.102094436625719e-06, "loss": 0.8702, "step": 22537 }, { "epoch": 0.16, "learning_rate": 3.101314081014075e-06, "loss": 0.8632, "step": 22538 }, { "epoch": 0.16, "learning_rate": 3.100533805553294e-06, "loss": 0.8221, "step": 22539 }, { "epoch": 0.16, "learning_rate": 3.0997536102524404e-06, "loss": 0.8244, "step": 22540 }, { "epoch": 0.16, "learning_rate": 3.0989734951205785e-06, "loss": 0.8743, "step": 22541 }, { "epoch": 0.16, "learning_rate": 3.0981934601667717e-06, "loss": 0.8484, "step": 22542 }, { "epoch": 0.16, "learning_rate": 3.097413505400084e-06, "loss": 0.8236, "step": 22543 }, { "epoch": 0.16, "learning_rate": 3.0966336308295785e-06, "loss": 0.8745, "step": 22544 }, { "epoch": 0.16, "learning_rate": 3.095853836464311e-06, "loss": 0.843, "step": 22545 }, { "epoch": 0.16, "learning_rate": 3.095074122313343e-06, "loss": 0.8866, "step": 22546 }, { "epoch": 0.16, "learning_rate": 3.094294488385735e-06, "loss": 0.8469, "step": 22547 }, { "epoch": 0.16, "learning_rate": 3.0935149346905456e-06, "loss": 0.8205, "step": 22548 }, { "epoch": 0.16, "learning_rate": 3.09273546123683e-06, "loss": 0.8512, "step": 22549 }, { "epoch": 0.16, "learning_rate": 3.0919560680336447e-06, "loss": 0.8607, "step": 22550 }, { "epoch": 0.16, "learning_rate": 3.0911767550900483e-06, "loss": 0.8417, "step": 22551 }, { "epoch": 0.16, "learning_rate": 3.0903975224150896e-06, "loss": 0.8291, "step": 22552 }, { "epoch": 0.16, "learning_rate": 3.089618370017824e-06, "loss": 0.8331, "step": 22553 }, { "epoch": 0.16, "learning_rate": 3.088839297907301e-06, "loss": 0.867, "step": 22554 }, { "epoch": 0.16, "learning_rate": 3.088060306092582e-06, "loss": 0.8691, "step": 22555 }, { "epoch": 0.16, "learning_rate": 3.0872813945827076e-06, "loss": 0.8857, "step": 22556 }, { "epoch": 0.16, "learning_rate": 3.0865025633867308e-06, "loss": 0.8335, "step": 22557 }, { "epoch": 0.16, "learning_rate": 3.0857238125137013e-06, "loss": 0.9093, "step": 22558 }, { "epoch": 0.16, "learning_rate": 3.084945141972667e-06, "loss": 0.8584, "step": 22559 }, { "epoch": 0.16, "learning_rate": 3.084166551772668e-06, "loss": 0.9062, "step": 22560 }, { "epoch": 0.16, "learning_rate": 3.0833880419227604e-06, "loss": 0.8649, "step": 22561 }, { "epoch": 0.16, "learning_rate": 3.082609612431986e-06, "loss": 0.8509, "step": 22562 }, { "epoch": 0.16, "learning_rate": 3.0818312633093837e-06, "loss": 0.8376, "step": 22563 }, { "epoch": 0.16, "learning_rate": 3.081052994564001e-06, "loss": 0.8798, "step": 22564 }, { "epoch": 0.16, "learning_rate": 3.0802748062048793e-06, "loss": 0.8498, "step": 22565 }, { "epoch": 0.16, "learning_rate": 3.0794966982410592e-06, "loss": 0.876, "step": 22566 }, { "epoch": 0.16, "learning_rate": 3.0787186706815808e-06, "loss": 0.8211, "step": 22567 }, { "epoch": 0.16, "learning_rate": 3.077940723535485e-06, "loss": 0.8697, "step": 22568 }, { "epoch": 0.16, "learning_rate": 3.0771628568118115e-06, "loss": 0.852, "step": 22569 }, { "epoch": 0.16, "learning_rate": 3.0763850705195943e-06, "loss": 0.8506, "step": 22570 }, { "epoch": 0.16, "learning_rate": 3.0756073646678697e-06, "loss": 0.8764, "step": 22571 }, { "epoch": 0.16, "learning_rate": 3.0748297392656746e-06, "loss": 0.8729, "step": 22572 }, { "epoch": 0.16, "learning_rate": 3.074052194322045e-06, "loss": 0.8711, "step": 22573 }, { "epoch": 0.16, "learning_rate": 3.073274729846013e-06, "loss": 0.8333, "step": 22574 }, { "epoch": 0.16, "learning_rate": 3.0724973458466114e-06, "loss": 0.8519, "step": 22575 }, { "epoch": 0.16, "learning_rate": 3.071720042332876e-06, "loss": 0.8801, "step": 22576 }, { "epoch": 0.16, "learning_rate": 3.0709428193138313e-06, "loss": 0.8471, "step": 22577 }, { "epoch": 0.16, "learning_rate": 3.0701656767985075e-06, "loss": 0.8315, "step": 22578 }, { "epoch": 0.16, "learning_rate": 3.0693886147959384e-06, "loss": 0.8883, "step": 22579 }, { "epoch": 0.16, "learning_rate": 3.068611633315155e-06, "loss": 0.8591, "step": 22580 }, { "epoch": 0.16, "learning_rate": 3.0678347323651757e-06, "loss": 0.8413, "step": 22581 }, { "epoch": 0.16, "learning_rate": 3.0670579119550302e-06, "loss": 0.8718, "step": 22582 }, { "epoch": 0.16, "learning_rate": 3.066281172093748e-06, "loss": 0.8846, "step": 22583 }, { "epoch": 0.16, "learning_rate": 3.065504512790344e-06, "loss": 0.8366, "step": 22584 }, { "epoch": 0.16, "learning_rate": 3.0647279340538492e-06, "loss": 0.8476, "step": 22585 }, { "epoch": 0.16, "learning_rate": 3.0639514358932864e-06, "loss": 0.8839, "step": 22586 }, { "epoch": 0.16, "learning_rate": 3.0631750183176767e-06, "loss": 0.8577, "step": 22587 }, { "epoch": 0.16, "learning_rate": 3.062398681336036e-06, "loss": 0.8317, "step": 22588 }, { "epoch": 0.16, "learning_rate": 3.061622424957389e-06, "loss": 0.8498, "step": 22589 }, { "epoch": 0.16, "learning_rate": 3.060846249190751e-06, "loss": 0.8784, "step": 22590 }, { "epoch": 0.16, "learning_rate": 3.0600701540451414e-06, "loss": 0.8608, "step": 22591 }, { "epoch": 0.16, "learning_rate": 3.0592941395295782e-06, "loss": 0.8752, "step": 22592 }, { "epoch": 0.16, "learning_rate": 3.058518205653075e-06, "loss": 0.8441, "step": 22593 }, { "epoch": 0.16, "learning_rate": 3.0577423524246518e-06, "loss": 0.8766, "step": 22594 }, { "epoch": 0.16, "learning_rate": 3.0569665798533168e-06, "loss": 0.8739, "step": 22595 }, { "epoch": 0.16, "learning_rate": 3.056190887948084e-06, "loss": 0.8883, "step": 22596 }, { "epoch": 0.16, "learning_rate": 3.055415276717968e-06, "loss": 0.8805, "step": 22597 }, { "epoch": 0.16, "learning_rate": 3.054639746171979e-06, "loss": 0.8917, "step": 22598 }, { "epoch": 0.16, "learning_rate": 3.0538642963191266e-06, "loss": 0.861, "step": 22599 }, { "epoch": 0.16, "learning_rate": 3.0530889271684217e-06, "loss": 0.857, "step": 22600 }, { "epoch": 0.16, "learning_rate": 3.0523136387288745e-06, "loss": 0.866, "step": 22601 }, { "epoch": 0.16, "learning_rate": 3.0515384310094877e-06, "loss": 0.8313, "step": 22602 }, { "epoch": 0.16, "learning_rate": 3.050763304019266e-06, "loss": 0.8577, "step": 22603 }, { "epoch": 0.16, "learning_rate": 3.049988257767222e-06, "loss": 0.8834, "step": 22604 }, { "epoch": 0.16, "learning_rate": 3.049213292262363e-06, "loss": 0.8506, "step": 22605 }, { "epoch": 0.16, "learning_rate": 3.048438407513683e-06, "loss": 0.8655, "step": 22606 }, { "epoch": 0.16, "learning_rate": 3.0476636035301887e-06, "loss": 0.8751, "step": 22607 }, { "epoch": 0.16, "learning_rate": 3.046888880320883e-06, "loss": 0.8402, "step": 22608 }, { "epoch": 0.16, "learning_rate": 3.046114237894766e-06, "loss": 0.8277, "step": 22609 }, { "epoch": 0.16, "learning_rate": 3.045339676260839e-06, "loss": 0.8666, "step": 22610 }, { "epoch": 0.16, "learning_rate": 3.0445651954281e-06, "loss": 0.7911, "step": 22611 }, { "epoch": 0.16, "learning_rate": 3.04379079540555e-06, "loss": 0.895, "step": 22612 }, { "epoch": 0.16, "learning_rate": 3.043016476202181e-06, "loss": 0.8814, "step": 22613 }, { "epoch": 0.16, "learning_rate": 3.042242237826991e-06, "loss": 0.8697, "step": 22614 }, { "epoch": 0.16, "learning_rate": 3.041468080288976e-06, "loss": 0.8572, "step": 22615 }, { "epoch": 0.16, "learning_rate": 3.040694003597131e-06, "loss": 0.8275, "step": 22616 }, { "epoch": 0.16, "learning_rate": 3.0399200077604485e-06, "loss": 0.8827, "step": 22617 }, { "epoch": 0.16, "learning_rate": 3.0391460927879225e-06, "loss": 0.8832, "step": 22618 }, { "epoch": 0.16, "learning_rate": 3.0383722586885454e-06, "loss": 0.8558, "step": 22619 }, { "epoch": 0.16, "learning_rate": 3.037598505471303e-06, "loss": 0.864, "step": 22620 }, { "epoch": 0.16, "learning_rate": 3.036824833145188e-06, "loss": 0.8459, "step": 22621 }, { "epoch": 0.16, "learning_rate": 3.036051241719189e-06, "loss": 0.8579, "step": 22622 }, { "epoch": 0.16, "learning_rate": 3.035277731202293e-06, "loss": 0.8298, "step": 22623 }, { "epoch": 0.16, "learning_rate": 3.034504301603489e-06, "loss": 0.8757, "step": 22624 }, { "epoch": 0.16, "learning_rate": 3.0337309529317603e-06, "loss": 0.8783, "step": 22625 }, { "epoch": 0.16, "learning_rate": 3.032957685196097e-06, "loss": 0.8408, "step": 22626 }, { "epoch": 0.16, "learning_rate": 3.0321844984054737e-06, "loss": 0.8334, "step": 22627 }, { "epoch": 0.16, "learning_rate": 3.0314113925688814e-06, "loss": 0.867, "step": 22628 }, { "epoch": 0.16, "learning_rate": 3.0306383676953033e-06, "loss": 0.8843, "step": 22629 }, { "epoch": 0.16, "learning_rate": 3.0298654237937144e-06, "loss": 0.8872, "step": 22630 }, { "epoch": 0.16, "learning_rate": 3.029092560873097e-06, "loss": 0.8491, "step": 22631 }, { "epoch": 0.16, "learning_rate": 3.0283197789424324e-06, "loss": 0.8757, "step": 22632 }, { "epoch": 0.16, "learning_rate": 3.0275470780106973e-06, "loss": 0.8371, "step": 22633 }, { "epoch": 0.16, "learning_rate": 3.0267744580868694e-06, "loss": 0.8797, "step": 22634 }, { "epoch": 0.16, "learning_rate": 3.026001919179926e-06, "loss": 0.8595, "step": 22635 }, { "epoch": 0.16, "learning_rate": 3.0252294612988454e-06, "loss": 0.8825, "step": 22636 }, { "epoch": 0.16, "learning_rate": 3.0244570844525957e-06, "loss": 0.8035, "step": 22637 }, { "epoch": 0.16, "learning_rate": 3.023684788650154e-06, "loss": 0.8632, "step": 22638 }, { "epoch": 0.16, "learning_rate": 3.022912573900494e-06, "loss": 0.8838, "step": 22639 }, { "epoch": 0.16, "learning_rate": 3.0221404402125854e-06, "loss": 0.8602, "step": 22640 }, { "epoch": 0.16, "learning_rate": 3.0213683875954004e-06, "loss": 0.861, "step": 22641 }, { "epoch": 0.16, "learning_rate": 3.020596416057909e-06, "loss": 0.9069, "step": 22642 }, { "epoch": 0.16, "learning_rate": 3.019824525609083e-06, "loss": 0.8426, "step": 22643 }, { "epoch": 0.16, "learning_rate": 3.0190527162578843e-06, "loss": 0.8753, "step": 22644 }, { "epoch": 0.16, "learning_rate": 3.0182809880132824e-06, "loss": 0.85, "step": 22645 }, { "epoch": 0.16, "learning_rate": 3.0175093408842413e-06, "loss": 0.9007, "step": 22646 }, { "epoch": 0.16, "learning_rate": 3.0167377748797355e-06, "loss": 0.8543, "step": 22647 }, { "epoch": 0.16, "learning_rate": 3.0159662900087193e-06, "loss": 0.8585, "step": 22648 }, { "epoch": 0.16, "learning_rate": 3.0151948862801594e-06, "loss": 0.8698, "step": 22649 }, { "epoch": 0.16, "learning_rate": 3.014423563703018e-06, "loss": 0.9032, "step": 22650 }, { "epoch": 0.16, "learning_rate": 3.013652322286258e-06, "loss": 0.8696, "step": 22651 }, { "epoch": 0.16, "learning_rate": 3.0128811620388378e-06, "loss": 0.8516, "step": 22652 }, { "epoch": 0.16, "learning_rate": 3.012110082969718e-06, "loss": 0.857, "step": 22653 }, { "epoch": 0.16, "learning_rate": 3.01133908508786e-06, "loss": 0.894, "step": 22654 }, { "epoch": 0.16, "learning_rate": 3.010568168402216e-06, "loss": 0.9033, "step": 22655 }, { "epoch": 0.16, "learning_rate": 3.0097973329217447e-06, "loss": 0.882, "step": 22656 }, { "epoch": 0.16, "learning_rate": 3.009026578655402e-06, "loss": 0.9148, "step": 22657 }, { "epoch": 0.16, "learning_rate": 3.008255905612144e-06, "loss": 0.8628, "step": 22658 }, { "epoch": 0.16, "learning_rate": 3.0074853138009243e-06, "loss": 0.8506, "step": 22659 }, { "epoch": 0.16, "learning_rate": 3.0067148032306936e-06, "loss": 0.8694, "step": 22660 }, { "epoch": 0.16, "learning_rate": 3.00594437391041e-06, "loss": 0.8008, "step": 22661 }, { "epoch": 0.16, "learning_rate": 3.005174025849017e-06, "loss": 0.8535, "step": 22662 }, { "epoch": 0.16, "learning_rate": 3.0044037590554685e-06, "loss": 0.8172, "step": 22663 }, { "epoch": 0.16, "learning_rate": 3.003633573538712e-06, "loss": 0.8497, "step": 22664 }, { "epoch": 0.16, "learning_rate": 3.0028634693076974e-06, "loss": 0.8419, "step": 22665 }, { "epoch": 0.16, "learning_rate": 3.002093446371371e-06, "loss": 0.8426, "step": 22666 }, { "epoch": 0.16, "learning_rate": 3.00132350473868e-06, "loss": 0.8589, "step": 22667 }, { "epoch": 0.16, "learning_rate": 3.0005536444185724e-06, "loss": 0.8424, "step": 22668 }, { "epoch": 0.16, "learning_rate": 2.999783865419987e-06, "loss": 0.8571, "step": 22669 }, { "epoch": 0.16, "learning_rate": 2.9990141677518657e-06, "loss": 0.835, "step": 22670 }, { "epoch": 0.16, "learning_rate": 2.9982445514231593e-06, "loss": 0.896, "step": 22671 }, { "epoch": 0.16, "learning_rate": 2.9974750164428078e-06, "loss": 0.8745, "step": 22672 }, { "epoch": 0.16, "learning_rate": 2.996705562819747e-06, "loss": 0.8434, "step": 22673 }, { "epoch": 0.16, "learning_rate": 2.9959361905629192e-06, "loss": 0.8842, "step": 22674 }, { "epoch": 0.16, "learning_rate": 2.9951668996812657e-06, "loss": 0.8745, "step": 22675 }, { "epoch": 0.16, "learning_rate": 2.994397690183716e-06, "loss": 0.8852, "step": 22676 }, { "epoch": 0.16, "learning_rate": 2.993628562079215e-06, "loss": 0.8682, "step": 22677 }, { "epoch": 0.16, "learning_rate": 2.992859515376697e-06, "loss": 0.897, "step": 22678 }, { "epoch": 0.16, "learning_rate": 2.9920905500850992e-06, "loss": 0.8829, "step": 22679 }, { "epoch": 0.16, "learning_rate": 2.9913216662133503e-06, "loss": 0.8503, "step": 22680 }, { "epoch": 0.16, "learning_rate": 2.990552863770385e-06, "loss": 0.8531, "step": 22681 }, { "epoch": 0.16, "learning_rate": 2.9897841427651363e-06, "loss": 0.9014, "step": 22682 }, { "epoch": 0.16, "learning_rate": 2.989015503206536e-06, "loss": 0.877, "step": 22683 }, { "epoch": 0.16, "learning_rate": 2.9882469451035146e-06, "loss": 0.8479, "step": 22684 }, { "epoch": 0.16, "learning_rate": 2.9874784684649995e-06, "loss": 0.8841, "step": 22685 }, { "epoch": 0.16, "learning_rate": 2.9867100732999243e-06, "loss": 0.8572, "step": 22686 }, { "epoch": 0.16, "learning_rate": 2.985941759617208e-06, "loss": 0.8375, "step": 22687 }, { "epoch": 0.16, "learning_rate": 2.985173527425783e-06, "loss": 0.8936, "step": 22688 }, { "epoch": 0.16, "learning_rate": 2.9844053767345726e-06, "loss": 0.8759, "step": 22689 }, { "epoch": 0.16, "learning_rate": 2.9836373075525014e-06, "loss": 0.8805, "step": 22690 }, { "epoch": 0.16, "learning_rate": 2.982869319888495e-06, "loss": 0.85, "step": 22691 }, { "epoch": 0.16, "learning_rate": 2.982101413751474e-06, "loss": 0.8069, "step": 22692 }, { "epoch": 0.16, "learning_rate": 2.9813335891503648e-06, "loss": 0.8608, "step": 22693 }, { "epoch": 0.16, "learning_rate": 2.9805658460940802e-06, "loss": 0.8728, "step": 22694 }, { "epoch": 0.16, "learning_rate": 2.9797981845915424e-06, "loss": 0.844, "step": 22695 }, { "epoch": 0.16, "learning_rate": 2.9790306046516735e-06, "loss": 0.8411, "step": 22696 }, { "epoch": 0.16, "learning_rate": 2.9782631062833953e-06, "loss": 0.8543, "step": 22697 }, { "epoch": 0.16, "learning_rate": 2.9774956894956153e-06, "loss": 0.8396, "step": 22698 }, { "epoch": 0.16, "learning_rate": 2.9767283542972547e-06, "loss": 0.8688, "step": 22699 }, { "epoch": 0.16, "learning_rate": 2.9759611006972266e-06, "loss": 0.8412, "step": 22700 }, { "epoch": 0.16, "learning_rate": 2.975193928704447e-06, "loss": 0.8563, "step": 22701 }, { "epoch": 0.16, "learning_rate": 2.9744268383278284e-06, "loss": 0.8235, "step": 22702 }, { "epoch": 0.16, "learning_rate": 2.973659829576283e-06, "loss": 0.8522, "step": 22703 }, { "epoch": 0.16, "learning_rate": 2.972892902458726e-06, "loss": 0.8876, "step": 22704 }, { "epoch": 0.16, "learning_rate": 2.9721260569840604e-06, "loss": 0.8851, "step": 22705 }, { "epoch": 0.16, "learning_rate": 2.971359293161199e-06, "loss": 0.8511, "step": 22706 }, { "epoch": 0.16, "learning_rate": 2.9705926109990513e-06, "loss": 0.8781, "step": 22707 }, { "epoch": 0.17, "learning_rate": 2.9698260105065236e-06, "loss": 0.894, "step": 22708 }, { "epoch": 0.17, "learning_rate": 2.969059491692523e-06, "loss": 0.835, "step": 22709 }, { "epoch": 0.17, "learning_rate": 2.968293054565956e-06, "loss": 0.8404, "step": 22710 }, { "epoch": 0.17, "learning_rate": 2.967526699135729e-06, "loss": 0.8746, "step": 22711 }, { "epoch": 0.17, "learning_rate": 2.9667604254107386e-06, "loss": 0.8705, "step": 22712 }, { "epoch": 0.17, "learning_rate": 2.9659942333998904e-06, "loss": 0.8385, "step": 22713 }, { "epoch": 0.17, "learning_rate": 2.965228123112094e-06, "loss": 0.8325, "step": 22714 }, { "epoch": 0.17, "learning_rate": 2.9644620945562407e-06, "loss": 0.8511, "step": 22715 }, { "epoch": 0.17, "learning_rate": 2.963696147741234e-06, "loss": 0.8301, "step": 22716 }, { "epoch": 0.17, "learning_rate": 2.9629302826759734e-06, "loss": 0.85, "step": 22717 }, { "epoch": 0.17, "learning_rate": 2.9621644993693587e-06, "loss": 0.8495, "step": 22718 }, { "epoch": 0.17, "learning_rate": 2.961398797830278e-06, "loss": 0.8002, "step": 22719 }, { "epoch": 0.17, "learning_rate": 2.9606331780676376e-06, "loss": 0.8387, "step": 22720 }, { "epoch": 0.17, "learning_rate": 2.9598676400903324e-06, "loss": 0.8241, "step": 22721 }, { "epoch": 0.17, "learning_rate": 2.959102183907251e-06, "loss": 0.8248, "step": 22722 }, { "epoch": 0.17, "learning_rate": 2.958336809527288e-06, "loss": 0.8392, "step": 22723 }, { "epoch": 0.17, "learning_rate": 2.957571516959338e-06, "loss": 0.8963, "step": 22724 }, { "epoch": 0.17, "learning_rate": 2.9568063062122898e-06, "loss": 0.8502, "step": 22725 }, { "epoch": 0.17, "learning_rate": 2.956041177295036e-06, "loss": 0.8375, "step": 22726 }, { "epoch": 0.17, "learning_rate": 2.9552761302164655e-06, "loss": 0.8719, "step": 22727 }, { "epoch": 0.17, "learning_rate": 2.9545111649854687e-06, "loss": 0.8514, "step": 22728 }, { "epoch": 0.17, "learning_rate": 2.9537462816109287e-06, "loss": 0.8169, "step": 22729 }, { "epoch": 0.17, "learning_rate": 2.9529814801017355e-06, "loss": 0.8755, "step": 22730 }, { "epoch": 0.17, "learning_rate": 2.9522167604667716e-06, "loss": 0.8461, "step": 22731 }, { "epoch": 0.17, "learning_rate": 2.951452122714926e-06, "loss": 0.8731, "step": 22732 }, { "epoch": 0.17, "learning_rate": 2.9506875668550793e-06, "loss": 0.8774, "step": 22733 }, { "epoch": 0.17, "learning_rate": 2.9499230928961166e-06, "loss": 0.872, "step": 22734 }, { "epoch": 0.17, "learning_rate": 2.94915870084692e-06, "loss": 0.8247, "step": 22735 }, { "epoch": 0.17, "learning_rate": 2.948394390716367e-06, "loss": 0.8913, "step": 22736 }, { "epoch": 0.17, "learning_rate": 2.9476301625133395e-06, "loss": 0.8271, "step": 22737 }, { "epoch": 0.17, "learning_rate": 2.9468660162467132e-06, "loss": 0.8746, "step": 22738 }, { "epoch": 0.17, "learning_rate": 2.9461019519253763e-06, "loss": 0.8489, "step": 22739 }, { "epoch": 0.17, "learning_rate": 2.9453379695581964e-06, "loss": 0.848, "step": 22740 }, { "epoch": 0.17, "learning_rate": 2.9445740691540516e-06, "loss": 0.8535, "step": 22741 }, { "epoch": 0.17, "learning_rate": 2.943810250721818e-06, "loss": 0.8598, "step": 22742 }, { "epoch": 0.17, "learning_rate": 2.94304651427037e-06, "loss": 0.8101, "step": 22743 }, { "epoch": 0.17, "learning_rate": 2.9422828598085806e-06, "loss": 0.8601, "step": 22744 }, { "epoch": 0.17, "learning_rate": 2.9415192873453225e-06, "loss": 0.856, "step": 22745 }, { "epoch": 0.17, "learning_rate": 2.9407557968894706e-06, "loss": 0.8773, "step": 22746 }, { "epoch": 0.17, "learning_rate": 2.9399923884498873e-06, "loss": 0.8747, "step": 22747 }, { "epoch": 0.17, "learning_rate": 2.939229062035447e-06, "loss": 0.8631, "step": 22748 }, { "epoch": 0.17, "learning_rate": 2.9384658176550185e-06, "loss": 0.8751, "step": 22749 }, { "epoch": 0.17, "learning_rate": 2.937702655317467e-06, "loss": 0.8286, "step": 22750 }, { "epoch": 0.17, "learning_rate": 2.936939575031662e-06, "loss": 0.866, "step": 22751 }, { "epoch": 0.17, "learning_rate": 2.9361765768064665e-06, "loss": 0.883, "step": 22752 }, { "epoch": 0.17, "learning_rate": 2.935413660650751e-06, "loss": 0.8801, "step": 22753 }, { "epoch": 0.17, "learning_rate": 2.9346508265733708e-06, "loss": 0.8905, "step": 22754 }, { "epoch": 0.17, "learning_rate": 2.9338880745831933e-06, "loss": 0.8935, "step": 22755 }, { "epoch": 0.17, "learning_rate": 2.9331254046890767e-06, "loss": 0.8809, "step": 22756 }, { "epoch": 0.17, "learning_rate": 2.9323628168998906e-06, "loss": 0.8699, "step": 22757 }, { "epoch": 0.17, "learning_rate": 2.9316003112244874e-06, "loss": 0.8458, "step": 22758 }, { "epoch": 0.17, "learning_rate": 2.930837887671728e-06, "loss": 0.8749, "step": 22759 }, { "epoch": 0.17, "learning_rate": 2.9300755462504724e-06, "loss": 0.8614, "step": 22760 }, { "epoch": 0.17, "learning_rate": 2.929313286969574e-06, "loss": 0.8542, "step": 22761 }, { "epoch": 0.17, "learning_rate": 2.9285511098378873e-06, "loss": 0.8551, "step": 22762 }, { "epoch": 0.17, "learning_rate": 2.927789014864274e-06, "loss": 0.8353, "step": 22763 }, { "epoch": 0.17, "learning_rate": 2.9270270020575885e-06, "loss": 0.8644, "step": 22764 }, { "epoch": 0.17, "learning_rate": 2.9262650714266783e-06, "loss": 0.8385, "step": 22765 }, { "epoch": 0.17, "learning_rate": 2.925503222980398e-06, "loss": 0.8346, "step": 22766 }, { "epoch": 0.17, "learning_rate": 2.9247414567276023e-06, "loss": 0.8465, "step": 22767 }, { "epoch": 0.17, "learning_rate": 2.9239797726771323e-06, "loss": 0.8456, "step": 22768 }, { "epoch": 0.17, "learning_rate": 2.923218170837847e-06, "loss": 0.9226, "step": 22769 }, { "epoch": 0.17, "learning_rate": 2.9224566512185916e-06, "loss": 0.8544, "step": 22770 }, { "epoch": 0.17, "learning_rate": 2.921695213828217e-06, "loss": 0.8534, "step": 22771 }, { "epoch": 0.17, "learning_rate": 2.920933858675563e-06, "loss": 0.8609, "step": 22772 }, { "epoch": 0.17, "learning_rate": 2.9201725857694794e-06, "loss": 0.8136, "step": 22773 }, { "epoch": 0.17, "learning_rate": 2.91941139511881e-06, "loss": 0.8119, "step": 22774 }, { "epoch": 0.17, "learning_rate": 2.9186502867323985e-06, "loss": 0.8237, "step": 22775 }, { "epoch": 0.17, "learning_rate": 2.917889260619088e-06, "loss": 0.8676, "step": 22776 }, { "epoch": 0.17, "learning_rate": 2.9171283167877197e-06, "loss": 0.8657, "step": 22777 }, { "epoch": 0.17, "learning_rate": 2.9163674552471387e-06, "loss": 0.8563, "step": 22778 }, { "epoch": 0.17, "learning_rate": 2.9156066760061784e-06, "loss": 0.8521, "step": 22779 }, { "epoch": 0.17, "learning_rate": 2.9148459790736805e-06, "loss": 0.8488, "step": 22780 }, { "epoch": 0.17, "learning_rate": 2.914085364458482e-06, "loss": 0.9029, "step": 22781 }, { "epoch": 0.17, "learning_rate": 2.913324832169422e-06, "loss": 0.8576, "step": 22782 }, { "epoch": 0.17, "learning_rate": 2.912564382215335e-06, "loss": 0.8613, "step": 22783 }, { "epoch": 0.17, "learning_rate": 2.9118040146050574e-06, "loss": 0.8605, "step": 22784 }, { "epoch": 0.17, "learning_rate": 2.911043729347425e-06, "loss": 0.8579, "step": 22785 }, { "epoch": 0.17, "learning_rate": 2.910283526451262e-06, "loss": 0.8635, "step": 22786 }, { "epoch": 0.17, "learning_rate": 2.909523405925412e-06, "loss": 0.8122, "step": 22787 }, { "epoch": 0.17, "learning_rate": 2.9087633677786996e-06, "loss": 0.8292, "step": 22788 }, { "epoch": 0.17, "learning_rate": 2.9080034120199616e-06, "loss": 0.8319, "step": 22789 }, { "epoch": 0.17, "learning_rate": 2.907243538658019e-06, "loss": 0.8657, "step": 22790 }, { "epoch": 0.17, "learning_rate": 2.906483747701705e-06, "loss": 0.8872, "step": 22791 }, { "epoch": 0.17, "learning_rate": 2.9057240391598453e-06, "loss": 0.8596, "step": 22792 }, { "epoch": 0.17, "learning_rate": 2.904964413041268e-06, "loss": 0.8466, "step": 22793 }, { "epoch": 0.17, "learning_rate": 2.9042048693547974e-06, "loss": 0.8479, "step": 22794 }, { "epoch": 0.17, "learning_rate": 2.9034454081092588e-06, "loss": 0.8582, "step": 22795 }, { "epoch": 0.17, "learning_rate": 2.9026860293134786e-06, "loss": 0.856, "step": 22796 }, { "epoch": 0.17, "learning_rate": 2.9019267329762733e-06, "loss": 0.8577, "step": 22797 }, { "epoch": 0.17, "learning_rate": 2.9011675191064694e-06, "loss": 0.8499, "step": 22798 }, { "epoch": 0.17, "learning_rate": 2.900408387712884e-06, "loss": 0.8873, "step": 22799 }, { "epoch": 0.17, "learning_rate": 2.89964933880434e-06, "loss": 0.849, "step": 22800 }, { "epoch": 0.17, "learning_rate": 2.8988903723896556e-06, "loss": 0.9169, "step": 22801 }, { "epoch": 0.17, "learning_rate": 2.898131488477648e-06, "loss": 0.8652, "step": 22802 }, { "epoch": 0.17, "learning_rate": 2.897372687077137e-06, "loss": 0.8526, "step": 22803 }, { "epoch": 0.17, "learning_rate": 2.896613968196933e-06, "loss": 0.8532, "step": 22804 }, { "epoch": 0.17, "learning_rate": 2.8958553318458517e-06, "loss": 0.8656, "step": 22805 }, { "epoch": 0.17, "learning_rate": 2.8950967780327145e-06, "loss": 0.8261, "step": 22806 }, { "epoch": 0.17, "learning_rate": 2.894338306766328e-06, "loss": 0.8622, "step": 22807 }, { "epoch": 0.17, "learning_rate": 2.8935799180555047e-06, "loss": 0.8502, "step": 22808 }, { "epoch": 0.17, "learning_rate": 2.8928216119090567e-06, "loss": 0.8808, "step": 22809 }, { "epoch": 0.17, "learning_rate": 2.8920633883357986e-06, "loss": 0.8899, "step": 22810 }, { "epoch": 0.17, "learning_rate": 2.891305247344528e-06, "loss": 0.8453, "step": 22811 }, { "epoch": 0.17, "learning_rate": 2.890547188944064e-06, "loss": 0.8425, "step": 22812 }, { "epoch": 0.17, "learning_rate": 2.8897892131432138e-06, "loss": 0.8648, "step": 22813 }, { "epoch": 0.17, "learning_rate": 2.889031319950778e-06, "loss": 0.8093, "step": 22814 }, { "epoch": 0.17, "learning_rate": 2.888273509375564e-06, "loss": 0.8595, "step": 22815 }, { "epoch": 0.17, "learning_rate": 2.8875157814263768e-06, "loss": 0.8437, "step": 22816 }, { "epoch": 0.17, "learning_rate": 2.88675813611202e-06, "loss": 0.8374, "step": 22817 }, { "epoch": 0.17, "learning_rate": 2.8860005734412964e-06, "loss": 0.867, "step": 22818 }, { "epoch": 0.17, "learning_rate": 2.885243093423007e-06, "loss": 0.8542, "step": 22819 }, { "epoch": 0.17, "learning_rate": 2.884485696065955e-06, "loss": 0.8657, "step": 22820 }, { "epoch": 0.17, "learning_rate": 2.8837283813789354e-06, "loss": 0.8805, "step": 22821 }, { "epoch": 0.17, "learning_rate": 2.882971149370749e-06, "loss": 0.8877, "step": 22822 }, { "epoch": 0.17, "learning_rate": 2.882214000050194e-06, "loss": 0.8506, "step": 22823 }, { "epoch": 0.17, "learning_rate": 2.8814569334260667e-06, "loss": 0.8775, "step": 22824 }, { "epoch": 0.17, "learning_rate": 2.880699949507163e-06, "loss": 0.8523, "step": 22825 }, { "epoch": 0.17, "learning_rate": 2.879943048302278e-06, "loss": 0.8342, "step": 22826 }, { "epoch": 0.17, "learning_rate": 2.8791862298202076e-06, "loss": 0.8671, "step": 22827 }, { "epoch": 0.17, "learning_rate": 2.878429494069741e-06, "loss": 0.8799, "step": 22828 }, { "epoch": 0.17, "learning_rate": 2.8776728410596676e-06, "loss": 0.8836, "step": 22829 }, { "epoch": 0.17, "learning_rate": 2.8769162707987862e-06, "loss": 0.8453, "step": 22830 }, { "epoch": 0.17, "learning_rate": 2.8761597832958864e-06, "loss": 0.8271, "step": 22831 }, { "epoch": 0.17, "learning_rate": 2.87540337855975e-06, "loss": 0.8386, "step": 22832 }, { "epoch": 0.17, "learning_rate": 2.8746470565991703e-06, "loss": 0.8676, "step": 22833 }, { "epoch": 0.17, "learning_rate": 2.873890817422933e-06, "loss": 0.8861, "step": 22834 }, { "epoch": 0.17, "learning_rate": 2.873134661039825e-06, "loss": 0.8099, "step": 22835 }, { "epoch": 0.17, "learning_rate": 2.8723785874586295e-06, "loss": 0.8872, "step": 22836 }, { "epoch": 0.17, "learning_rate": 2.8716225966881338e-06, "loss": 0.8542, "step": 22837 }, { "epoch": 0.17, "learning_rate": 2.870866688737123e-06, "loss": 0.8224, "step": 22838 }, { "epoch": 0.17, "learning_rate": 2.8701108636143737e-06, "loss": 0.8541, "step": 22839 }, { "epoch": 0.17, "learning_rate": 2.8693551213286684e-06, "loss": 0.8285, "step": 22840 }, { "epoch": 0.17, "learning_rate": 2.8685994618887904e-06, "loss": 0.9063, "step": 22841 }, { "epoch": 0.17, "learning_rate": 2.8678438853035175e-06, "loss": 0.813, "step": 22842 }, { "epoch": 0.17, "learning_rate": 2.8670883915816283e-06, "loss": 0.8522, "step": 22843 }, { "epoch": 0.17, "learning_rate": 2.8663329807319006e-06, "loss": 0.878, "step": 22844 }, { "epoch": 0.17, "learning_rate": 2.865577652763114e-06, "loss": 0.8028, "step": 22845 }, { "epoch": 0.17, "learning_rate": 2.864822407684038e-06, "loss": 0.828, "step": 22846 }, { "epoch": 0.17, "learning_rate": 2.86406724550345e-06, "loss": 0.8677, "step": 22847 }, { "epoch": 0.17, "learning_rate": 2.863312166230121e-06, "loss": 0.8858, "step": 22848 }, { "epoch": 0.17, "learning_rate": 2.8625571698728326e-06, "loss": 0.8503, "step": 22849 }, { "epoch": 0.17, "learning_rate": 2.861802256440348e-06, "loss": 0.8507, "step": 22850 }, { "epoch": 0.17, "learning_rate": 2.8610474259414413e-06, "loss": 0.8543, "step": 22851 }, { "epoch": 0.17, "learning_rate": 2.860292678384884e-06, "loss": 0.8844, "step": 22852 }, { "epoch": 0.17, "learning_rate": 2.8595380137794393e-06, "loss": 0.8857, "step": 22853 }, { "epoch": 0.17, "learning_rate": 2.858783432133875e-06, "loss": 0.8825, "step": 22854 }, { "epoch": 0.17, "learning_rate": 2.8580289334569656e-06, "loss": 0.8673, "step": 22855 }, { "epoch": 0.17, "learning_rate": 2.8572745177574757e-06, "loss": 0.875, "step": 22856 }, { "epoch": 0.17, "learning_rate": 2.856520185044164e-06, "loss": 0.8514, "step": 22857 }, { "epoch": 0.17, "learning_rate": 2.855765935325798e-06, "loss": 0.841, "step": 22858 }, { "epoch": 0.17, "learning_rate": 2.855011768611141e-06, "loss": 0.8553, "step": 22859 }, { "epoch": 0.17, "learning_rate": 2.8542576849089563e-06, "loss": 0.8447, "step": 22860 }, { "epoch": 0.17, "learning_rate": 2.8535036842280027e-06, "loss": 0.864, "step": 22861 }, { "epoch": 0.17, "learning_rate": 2.852749766577041e-06, "loss": 0.849, "step": 22862 }, { "epoch": 0.17, "learning_rate": 2.851995931964835e-06, "loss": 0.8113, "step": 22863 }, { "epoch": 0.17, "learning_rate": 2.8512421804001354e-06, "loss": 0.8822, "step": 22864 }, { "epoch": 0.17, "learning_rate": 2.8504885118917025e-06, "loss": 0.8432, "step": 22865 }, { "epoch": 0.17, "learning_rate": 2.849734926448294e-06, "loss": 0.8589, "step": 22866 }, { "epoch": 0.17, "learning_rate": 2.848981424078664e-06, "loss": 0.8481, "step": 22867 }, { "epoch": 0.17, "learning_rate": 2.8482280047915667e-06, "loss": 0.8906, "step": 22868 }, { "epoch": 0.17, "learning_rate": 2.8474746685957567e-06, "loss": 0.8916, "step": 22869 }, { "epoch": 0.17, "learning_rate": 2.846721415499989e-06, "loss": 0.8763, "step": 22870 }, { "epoch": 0.17, "learning_rate": 2.845968245513008e-06, "loss": 0.8466, "step": 22871 }, { "epoch": 0.17, "learning_rate": 2.8452151586435693e-06, "loss": 0.8811, "step": 22872 }, { "epoch": 0.17, "learning_rate": 2.8444621549004203e-06, "loss": 0.8272, "step": 22873 }, { "epoch": 0.17, "learning_rate": 2.8437092342923113e-06, "loss": 0.893, "step": 22874 }, { "epoch": 0.17, "learning_rate": 2.8429563968279895e-06, "loss": 0.8137, "step": 22875 }, { "epoch": 0.17, "learning_rate": 2.8422036425162013e-06, "loss": 0.8598, "step": 22876 }, { "epoch": 0.17, "learning_rate": 2.841450971365696e-06, "loss": 0.8655, "step": 22877 }, { "epoch": 0.17, "learning_rate": 2.8406983833852076e-06, "loss": 0.8958, "step": 22878 }, { "epoch": 0.17, "learning_rate": 2.83994587858349e-06, "loss": 0.8592, "step": 22879 }, { "epoch": 0.17, "learning_rate": 2.8391934569692846e-06, "loss": 0.8726, "step": 22880 }, { "epoch": 0.17, "learning_rate": 2.838441118551334e-06, "loss": 0.8648, "step": 22881 }, { "epoch": 0.17, "learning_rate": 2.837688863338374e-06, "loss": 0.8414, "step": 22882 }, { "epoch": 0.17, "learning_rate": 2.836936691339147e-06, "loss": 0.8236, "step": 22883 }, { "epoch": 0.17, "learning_rate": 2.836184602562393e-06, "loss": 0.8716, "step": 22884 }, { "epoch": 0.17, "learning_rate": 2.8354325970168483e-06, "loss": 0.9114, "step": 22885 }, { "epoch": 0.17, "learning_rate": 2.834680674711252e-06, "loss": 0.83, "step": 22886 }, { "epoch": 0.17, "learning_rate": 2.8339288356543382e-06, "loss": 0.8203, "step": 22887 }, { "epoch": 0.17, "learning_rate": 2.833177079854846e-06, "loss": 0.9103, "step": 22888 }, { "epoch": 0.17, "learning_rate": 2.832425407321504e-06, "loss": 0.8329, "step": 22889 }, { "epoch": 0.17, "learning_rate": 2.8316738180630477e-06, "loss": 0.8667, "step": 22890 }, { "epoch": 0.17, "learning_rate": 2.830922312088209e-06, "loss": 0.8272, "step": 22891 }, { "epoch": 0.17, "learning_rate": 2.830170889405719e-06, "loss": 0.9057, "step": 22892 }, { "epoch": 0.17, "learning_rate": 2.829419550024308e-06, "loss": 0.8465, "step": 22893 }, { "epoch": 0.17, "learning_rate": 2.8286682939527064e-06, "loss": 0.8805, "step": 22894 }, { "epoch": 0.17, "learning_rate": 2.8279171211996446e-06, "loss": 0.8267, "step": 22895 }, { "epoch": 0.17, "learning_rate": 2.8271660317738437e-06, "loss": 0.8659, "step": 22896 }, { "epoch": 0.17, "learning_rate": 2.8264150256840307e-06, "loss": 0.8764, "step": 22897 }, { "epoch": 0.17, "learning_rate": 2.8256641029389385e-06, "loss": 0.8815, "step": 22898 }, { "epoch": 0.17, "learning_rate": 2.8249132635472855e-06, "loss": 0.836, "step": 22899 }, { "epoch": 0.17, "learning_rate": 2.8241625075177947e-06, "loss": 0.8455, "step": 22900 }, { "epoch": 0.17, "learning_rate": 2.823411834859191e-06, "loss": 0.8716, "step": 22901 }, { "epoch": 0.17, "learning_rate": 2.822661245580195e-06, "loss": 0.8633, "step": 22902 }, { "epoch": 0.17, "learning_rate": 2.8219107396895286e-06, "loss": 0.8125, "step": 22903 }, { "epoch": 0.17, "learning_rate": 2.821160317195909e-06, "loss": 0.8342, "step": 22904 }, { "epoch": 0.17, "learning_rate": 2.8204099781080587e-06, "loss": 0.8631, "step": 22905 }, { "epoch": 0.17, "learning_rate": 2.819659722434691e-06, "loss": 0.8305, "step": 22906 }, { "epoch": 0.17, "learning_rate": 2.818909550184523e-06, "loss": 0.8031, "step": 22907 }, { "epoch": 0.17, "learning_rate": 2.818159461366272e-06, "loss": 0.839, "step": 22908 }, { "epoch": 0.17, "learning_rate": 2.8174094559886535e-06, "loss": 0.8748, "step": 22909 }, { "epoch": 0.17, "learning_rate": 2.8166595340603787e-06, "loss": 0.8298, "step": 22910 }, { "epoch": 0.17, "learning_rate": 2.815909695590163e-06, "loss": 0.8371, "step": 22911 }, { "epoch": 0.17, "learning_rate": 2.8151599405867204e-06, "loss": 0.8289, "step": 22912 }, { "epoch": 0.17, "learning_rate": 2.814410269058754e-06, "loss": 0.8562, "step": 22913 }, { "epoch": 0.17, "learning_rate": 2.813660681014979e-06, "loss": 0.8782, "step": 22914 }, { "epoch": 0.17, "learning_rate": 2.812911176464104e-06, "loss": 0.8886, "step": 22915 }, { "epoch": 0.17, "learning_rate": 2.8121617554148363e-06, "loss": 0.8821, "step": 22916 }, { "epoch": 0.17, "learning_rate": 2.811412417875883e-06, "loss": 0.8671, "step": 22917 }, { "epoch": 0.17, "learning_rate": 2.8106631638559487e-06, "loss": 0.8603, "step": 22918 }, { "epoch": 0.17, "learning_rate": 2.809913993363744e-06, "loss": 0.8374, "step": 22919 }, { "epoch": 0.17, "learning_rate": 2.809164906407967e-06, "loss": 0.8502, "step": 22920 }, { "epoch": 0.17, "learning_rate": 2.8084159029973177e-06, "loss": 0.8691, "step": 22921 }, { "epoch": 0.17, "learning_rate": 2.807666983140508e-06, "loss": 0.8759, "step": 22922 }, { "epoch": 0.17, "learning_rate": 2.806918146846237e-06, "loss": 0.854, "step": 22923 }, { "epoch": 0.17, "learning_rate": 2.806169394123198e-06, "loss": 0.8697, "step": 22924 }, { "epoch": 0.17, "learning_rate": 2.805420724980096e-06, "loss": 0.8862, "step": 22925 }, { "epoch": 0.17, "learning_rate": 2.804672139425626e-06, "loss": 0.8825, "step": 22926 }, { "epoch": 0.17, "learning_rate": 2.8039236374684876e-06, "loss": 0.8355, "step": 22927 }, { "epoch": 0.17, "learning_rate": 2.803175219117377e-06, "loss": 0.8882, "step": 22928 }, { "epoch": 0.17, "learning_rate": 2.8024268843809875e-06, "loss": 0.8642, "step": 22929 }, { "epoch": 0.17, "learning_rate": 2.801678633268019e-06, "loss": 0.8535, "step": 22930 }, { "epoch": 0.17, "learning_rate": 2.800930465787157e-06, "loss": 0.7957, "step": 22931 }, { "epoch": 0.17, "learning_rate": 2.8001823819470987e-06, "loss": 0.8984, "step": 22932 }, { "epoch": 0.17, "learning_rate": 2.7994343817565338e-06, "loss": 0.8604, "step": 22933 }, { "epoch": 0.17, "learning_rate": 2.798686465224154e-06, "loss": 0.8706, "step": 22934 }, { "epoch": 0.17, "learning_rate": 2.797938632358649e-06, "loss": 0.813, "step": 22935 }, { "epoch": 0.17, "learning_rate": 2.7971908831687054e-06, "loss": 0.8657, "step": 22936 }, { "epoch": 0.17, "learning_rate": 2.796443217663015e-06, "loss": 0.8695, "step": 22937 }, { "epoch": 0.17, "learning_rate": 2.7956956358502594e-06, "loss": 0.8915, "step": 22938 }, { "epoch": 0.17, "learning_rate": 2.794948137739126e-06, "loss": 0.8485, "step": 22939 }, { "epoch": 0.17, "learning_rate": 2.794200723338296e-06, "loss": 0.9, "step": 22940 }, { "epoch": 0.17, "learning_rate": 2.793453392656463e-06, "loss": 0.8827, "step": 22941 }, { "epoch": 0.17, "learning_rate": 2.792706145702302e-06, "loss": 0.8652, "step": 22942 }, { "epoch": 0.17, "learning_rate": 2.7919589824844937e-06, "loss": 0.8416, "step": 22943 }, { "epoch": 0.17, "learning_rate": 2.7912119030117256e-06, "loss": 0.8847, "step": 22944 }, { "epoch": 0.17, "learning_rate": 2.7904649072926703e-06, "loss": 0.8478, "step": 22945 }, { "epoch": 0.17, "learning_rate": 2.7897179953360064e-06, "loss": 0.8113, "step": 22946 }, { "epoch": 0.17, "learning_rate": 2.788971167150417e-06, "loss": 0.8536, "step": 22947 }, { "epoch": 0.17, "learning_rate": 2.788224422744581e-06, "loss": 0.829, "step": 22948 }, { "epoch": 0.17, "learning_rate": 2.7874777621271663e-06, "loss": 0.8533, "step": 22949 }, { "epoch": 0.17, "learning_rate": 2.7867311853068524e-06, "loss": 0.8491, "step": 22950 }, { "epoch": 0.17, "learning_rate": 2.7859846922923115e-06, "loss": 0.8603, "step": 22951 }, { "epoch": 0.17, "learning_rate": 2.7852382830922174e-06, "loss": 0.8836, "step": 22952 }, { "epoch": 0.17, "learning_rate": 2.784491957715243e-06, "loss": 0.8409, "step": 22953 }, { "epoch": 0.17, "learning_rate": 2.7837457161700588e-06, "loss": 0.8433, "step": 22954 }, { "epoch": 0.17, "learning_rate": 2.7829995584653368e-06, "loss": 0.8507, "step": 22955 }, { "epoch": 0.17, "learning_rate": 2.7822534846097403e-06, "loss": 0.8741, "step": 22956 }, { "epoch": 0.17, "learning_rate": 2.7815074946119413e-06, "loss": 0.8837, "step": 22957 }, { "epoch": 0.17, "learning_rate": 2.780761588480607e-06, "loss": 0.8563, "step": 22958 }, { "epoch": 0.17, "learning_rate": 2.780015766224403e-06, "loss": 0.8544, "step": 22959 }, { "epoch": 0.17, "learning_rate": 2.7792700278519945e-06, "loss": 0.8546, "step": 22960 }, { "epoch": 0.17, "learning_rate": 2.7785243733720456e-06, "loss": 0.8491, "step": 22961 }, { "epoch": 0.17, "learning_rate": 2.777778802793223e-06, "loss": 0.8394, "step": 22962 }, { "epoch": 0.17, "learning_rate": 2.7770333161241814e-06, "loss": 0.8396, "step": 22963 }, { "epoch": 0.17, "learning_rate": 2.7762879133735844e-06, "loss": 0.843, "step": 22964 }, { "epoch": 0.17, "learning_rate": 2.7755425945501e-06, "loss": 0.8508, "step": 22965 }, { "epoch": 0.17, "learning_rate": 2.774797359662379e-06, "loss": 0.824, "step": 22966 }, { "epoch": 0.17, "learning_rate": 2.774052208719081e-06, "loss": 0.8783, "step": 22967 }, { "epoch": 0.17, "learning_rate": 2.773307141728867e-06, "loss": 0.8514, "step": 22968 }, { "epoch": 0.17, "learning_rate": 2.7725621587003926e-06, "loss": 0.8731, "step": 22969 }, { "epoch": 0.17, "learning_rate": 2.7718172596423067e-06, "loss": 0.8371, "step": 22970 }, { "epoch": 0.17, "learning_rate": 2.771072444563271e-06, "loss": 0.8919, "step": 22971 }, { "epoch": 0.17, "learning_rate": 2.7703277134719376e-06, "loss": 0.8729, "step": 22972 }, { "epoch": 0.17, "learning_rate": 2.769583066376961e-06, "loss": 0.8759, "step": 22973 }, { "epoch": 0.17, "learning_rate": 2.7688385032869867e-06, "loss": 0.8704, "step": 22974 }, { "epoch": 0.17, "learning_rate": 2.768094024210669e-06, "loss": 0.8644, "step": 22975 }, { "epoch": 0.17, "learning_rate": 2.767349629156657e-06, "loss": 0.8527, "step": 22976 }, { "epoch": 0.17, "learning_rate": 2.7666053181335995e-06, "loss": 0.857, "step": 22977 }, { "epoch": 0.17, "learning_rate": 2.7658610911501426e-06, "loss": 0.8551, "step": 22978 }, { "epoch": 0.17, "learning_rate": 2.7651169482149355e-06, "loss": 0.8373, "step": 22979 }, { "epoch": 0.17, "learning_rate": 2.7643728893366253e-06, "loss": 0.8411, "step": 22980 }, { "epoch": 0.17, "learning_rate": 2.7636289145238502e-06, "loss": 0.8563, "step": 22981 }, { "epoch": 0.17, "learning_rate": 2.762885023785259e-06, "loss": 0.8697, "step": 22982 }, { "epoch": 0.17, "learning_rate": 2.7621412171294924e-06, "loss": 0.8231, "step": 22983 }, { "epoch": 0.17, "learning_rate": 2.7613974945651924e-06, "loss": 0.8659, "step": 22984 }, { "epoch": 0.17, "learning_rate": 2.760653856101001e-06, "loss": 0.8779, "step": 22985 }, { "epoch": 0.17, "learning_rate": 2.7599103017455563e-06, "loss": 0.7769, "step": 22986 }, { "epoch": 0.17, "learning_rate": 2.7591668315075026e-06, "loss": 0.8123, "step": 22987 }, { "epoch": 0.17, "learning_rate": 2.758423445395468e-06, "loss": 0.8375, "step": 22988 }, { "epoch": 0.17, "learning_rate": 2.7576801434180933e-06, "loss": 0.8394, "step": 22989 }, { "epoch": 0.17, "learning_rate": 2.756936925584022e-06, "loss": 0.8533, "step": 22990 }, { "epoch": 0.17, "learning_rate": 2.7561937919018788e-06, "loss": 0.8862, "step": 22991 }, { "epoch": 0.17, "learning_rate": 2.7554507423803012e-06, "loss": 0.9169, "step": 22992 }, { "epoch": 0.17, "learning_rate": 2.754707777027923e-06, "loss": 0.89, "step": 22993 }, { "epoch": 0.17, "learning_rate": 2.7539648958533747e-06, "loss": 0.8703, "step": 22994 }, { "epoch": 0.17, "learning_rate": 2.753222098865289e-06, "loss": 0.8118, "step": 22995 }, { "epoch": 0.17, "learning_rate": 2.7524793860722954e-06, "loss": 0.8679, "step": 22996 }, { "epoch": 0.17, "learning_rate": 2.7517367574830257e-06, "loss": 0.9067, "step": 22997 }, { "epoch": 0.17, "learning_rate": 2.750994213106102e-06, "loss": 0.8648, "step": 22998 }, { "epoch": 0.17, "learning_rate": 2.7502517529501538e-06, "loss": 0.8213, "step": 22999 }, { "epoch": 0.17, "learning_rate": 2.749509377023808e-06, "loss": 0.8796, "step": 23000 }, { "epoch": 0.17, "learning_rate": 2.7487670853356895e-06, "loss": 0.89, "step": 23001 }, { "epoch": 0.17, "learning_rate": 2.748024877894423e-06, "loss": 0.8569, "step": 23002 }, { "epoch": 0.17, "learning_rate": 2.74728275470863e-06, "loss": 0.8233, "step": 23003 }, { "epoch": 0.17, "learning_rate": 2.746540715786937e-06, "loss": 0.8764, "step": 23004 }, { "epoch": 0.17, "learning_rate": 2.74579876113796e-06, "loss": 0.8627, "step": 23005 }, { "epoch": 0.17, "learning_rate": 2.7450568907703213e-06, "loss": 0.8743, "step": 23006 }, { "epoch": 0.17, "learning_rate": 2.74431510469264e-06, "loss": 0.8346, "step": 23007 }, { "epoch": 0.17, "learning_rate": 2.743573402913533e-06, "loss": 0.8566, "step": 23008 }, { "epoch": 0.18, "learning_rate": 2.7428317854416208e-06, "loss": 0.8493, "step": 23009 }, { "epoch": 0.18, "learning_rate": 2.7420902522855176e-06, "loss": 0.8611, "step": 23010 }, { "epoch": 0.18, "learning_rate": 2.7413488034538415e-06, "loss": 0.8724, "step": 23011 }, { "epoch": 0.18, "learning_rate": 2.740607438955203e-06, "loss": 0.8547, "step": 23012 }, { "epoch": 0.18, "learning_rate": 2.739866158798212e-06, "loss": 0.8956, "step": 23013 }, { "epoch": 0.18, "learning_rate": 2.7391249629914905e-06, "loss": 0.8798, "step": 23014 }, { "epoch": 0.18, "learning_rate": 2.7383838515436476e-06, "loss": 0.8128, "step": 23015 }, { "epoch": 0.18, "learning_rate": 2.737642824463289e-06, "loss": 0.8743, "step": 23016 }, { "epoch": 0.18, "learning_rate": 2.7369018817590266e-06, "loss": 0.8674, "step": 23017 }, { "epoch": 0.18, "learning_rate": 2.7361610234394686e-06, "loss": 0.8903, "step": 23018 }, { "epoch": 0.18, "learning_rate": 2.735420249513222e-06, "loss": 0.8389, "step": 23019 }, { "epoch": 0.18, "learning_rate": 2.734679559988895e-06, "loss": 0.8666, "step": 23020 }, { "epoch": 0.18, "learning_rate": 2.733938954875093e-06, "loss": 0.8966, "step": 23021 }, { "epoch": 0.18, "learning_rate": 2.733198434180422e-06, "loss": 0.8597, "step": 23022 }, { "epoch": 0.18, "learning_rate": 2.7324579979134802e-06, "loss": 0.8374, "step": 23023 }, { "epoch": 0.18, "learning_rate": 2.731717646082873e-06, "loss": 0.882, "step": 23024 }, { "epoch": 0.18, "learning_rate": 2.730977378697204e-06, "loss": 0.8669, "step": 23025 }, { "epoch": 0.18, "learning_rate": 2.7302371957650707e-06, "loss": 0.8174, "step": 23026 }, { "epoch": 0.18, "learning_rate": 2.729497097295075e-06, "loss": 0.8396, "step": 23027 }, { "epoch": 0.18, "learning_rate": 2.7287570832958143e-06, "loss": 0.8862, "step": 23028 }, { "epoch": 0.18, "learning_rate": 2.7280171537758905e-06, "loss": 0.8887, "step": 23029 }, { "epoch": 0.18, "learning_rate": 2.7272773087438944e-06, "loss": 0.8497, "step": 23030 }, { "epoch": 0.18, "learning_rate": 2.726537548208422e-06, "loss": 0.8347, "step": 23031 }, { "epoch": 0.18, "learning_rate": 2.7257978721780685e-06, "loss": 0.8874, "step": 23032 }, { "epoch": 0.18, "learning_rate": 2.7250582806614347e-06, "loss": 0.9246, "step": 23033 }, { "epoch": 0.18, "learning_rate": 2.724318773667105e-06, "loss": 0.878, "step": 23034 }, { "epoch": 0.18, "learning_rate": 2.7235793512036744e-06, "loss": 0.8663, "step": 23035 }, { "epoch": 0.18, "learning_rate": 2.7228400132797362e-06, "loss": 0.8397, "step": 23036 }, { "epoch": 0.18, "learning_rate": 2.7221007599038706e-06, "loss": 0.847, "step": 23037 }, { "epoch": 0.18, "learning_rate": 2.721361591084677e-06, "loss": 0.8798, "step": 23038 }, { "epoch": 0.18, "learning_rate": 2.7206225068307402e-06, "loss": 0.8507, "step": 23039 }, { "epoch": 0.18, "learning_rate": 2.7198835071506478e-06, "loss": 0.814, "step": 23040 }, { "epoch": 0.18, "learning_rate": 2.7191445920529822e-06, "loss": 0.8696, "step": 23041 }, { "epoch": 0.18, "learning_rate": 2.718405761546331e-06, "loss": 0.8462, "step": 23042 }, { "epoch": 0.18, "learning_rate": 2.717667015639277e-06, "loss": 0.85, "step": 23043 }, { "epoch": 0.18, "learning_rate": 2.7169283543404047e-06, "loss": 0.8509, "step": 23044 }, { "epoch": 0.18, "learning_rate": 2.7161897776582937e-06, "loss": 0.8068, "step": 23045 }, { "epoch": 0.18, "learning_rate": 2.715451285601527e-06, "loss": 0.8199, "step": 23046 }, { "epoch": 0.18, "learning_rate": 2.714712878178687e-06, "loss": 0.8216, "step": 23047 }, { "epoch": 0.18, "learning_rate": 2.7139745553983464e-06, "loss": 0.8735, "step": 23048 }, { "epoch": 0.18, "learning_rate": 2.713236317269088e-06, "loss": 0.8482, "step": 23049 }, { "epoch": 0.18, "learning_rate": 2.712498163799486e-06, "loss": 0.8562, "step": 23050 }, { "epoch": 0.18, "learning_rate": 2.711760094998118e-06, "loss": 0.8634, "step": 23051 }, { "epoch": 0.18, "learning_rate": 2.7110221108735602e-06, "loss": 0.8458, "step": 23052 }, { "epoch": 0.18, "learning_rate": 2.710284211434384e-06, "loss": 0.8724, "step": 23053 }, { "epoch": 0.18, "learning_rate": 2.7095463966891687e-06, "loss": 0.8429, "step": 23054 }, { "epoch": 0.18, "learning_rate": 2.7088086666464772e-06, "loss": 0.8659, "step": 23055 }, { "epoch": 0.18, "learning_rate": 2.708071021314883e-06, "loss": 0.8725, "step": 23056 }, { "epoch": 0.18, "learning_rate": 2.7073334607029642e-06, "loss": 0.8871, "step": 23057 }, { "epoch": 0.18, "learning_rate": 2.7065959848192823e-06, "loss": 0.8751, "step": 23058 }, { "epoch": 0.18, "learning_rate": 2.7058585936724068e-06, "loss": 0.8674, "step": 23059 }, { "epoch": 0.18, "learning_rate": 2.7051212872709053e-06, "loss": 0.8699, "step": 23060 }, { "epoch": 0.18, "learning_rate": 2.7043840656233476e-06, "loss": 0.853, "step": 23061 }, { "epoch": 0.18, "learning_rate": 2.7036469287382894e-06, "loss": 0.8724, "step": 23062 }, { "epoch": 0.18, "learning_rate": 2.7029098766243046e-06, "loss": 0.8399, "step": 23063 }, { "epoch": 0.18, "learning_rate": 2.702172909289953e-06, "loss": 0.8647, "step": 23064 }, { "epoch": 0.18, "learning_rate": 2.7014360267438e-06, "loss": 0.8706, "step": 23065 }, { "epoch": 0.18, "learning_rate": 2.700699228994401e-06, "loss": 0.874, "step": 23066 }, { "epoch": 0.18, "learning_rate": 2.6999625160503196e-06, "loss": 0.8243, "step": 23067 }, { "epoch": 0.18, "learning_rate": 2.699225887920115e-06, "loss": 0.8635, "step": 23068 }, { "epoch": 0.18, "learning_rate": 2.698489344612344e-06, "loss": 0.8528, "step": 23069 }, { "epoch": 0.18, "learning_rate": 2.697752886135566e-06, "loss": 0.896, "step": 23070 }, { "epoch": 0.18, "learning_rate": 2.6970165124983374e-06, "loss": 0.8245, "step": 23071 }, { "epoch": 0.18, "learning_rate": 2.6962802237092146e-06, "loss": 0.8452, "step": 23072 }, { "epoch": 0.18, "learning_rate": 2.695544019776748e-06, "loss": 0.8151, "step": 23073 }, { "epoch": 0.18, "learning_rate": 2.694807900709494e-06, "loss": 0.8489, "step": 23074 }, { "epoch": 0.18, "learning_rate": 2.6940718665160036e-06, "loss": 0.8287, "step": 23075 }, { "epoch": 0.18, "learning_rate": 2.693335917204829e-06, "loss": 0.7983, "step": 23076 }, { "epoch": 0.18, "learning_rate": 2.69260005278452e-06, "loss": 0.901, "step": 23077 }, { "epoch": 0.18, "learning_rate": 2.691864273263628e-06, "loss": 0.8013, "step": 23078 }, { "epoch": 0.18, "learning_rate": 2.6911285786507026e-06, "loss": 0.8459, "step": 23079 }, { "epoch": 0.18, "learning_rate": 2.6903929689542864e-06, "loss": 0.8593, "step": 23080 }, { "epoch": 0.18, "learning_rate": 2.689657444182925e-06, "loss": 0.8721, "step": 23081 }, { "epoch": 0.18, "learning_rate": 2.6889220043451727e-06, "loss": 0.9035, "step": 23082 }, { "epoch": 0.18, "learning_rate": 2.6881866494495667e-06, "loss": 0.859, "step": 23083 }, { "epoch": 0.18, "learning_rate": 2.687451379504652e-06, "loss": 0.8458, "step": 23084 }, { "epoch": 0.18, "learning_rate": 2.6867161945189726e-06, "loss": 0.812, "step": 23085 }, { "epoch": 0.18, "learning_rate": 2.6859810945010687e-06, "loss": 0.8503, "step": 23086 }, { "epoch": 0.18, "learning_rate": 2.6852460794594816e-06, "loss": 0.8588, "step": 23087 }, { "epoch": 0.18, "learning_rate": 2.684511149402751e-06, "loss": 0.8478, "step": 23088 }, { "epoch": 0.18, "learning_rate": 2.6837763043394184e-06, "loss": 0.8586, "step": 23089 }, { "epoch": 0.18, "learning_rate": 2.6830415442780143e-06, "loss": 0.8689, "step": 23090 }, { "epoch": 0.18, "learning_rate": 2.6823068692270806e-06, "loss": 0.8592, "step": 23091 }, { "epoch": 0.18, "learning_rate": 2.6815722791951503e-06, "loss": 0.8607, "step": 23092 }, { "epoch": 0.18, "learning_rate": 2.680837774190761e-06, "loss": 0.8665, "step": 23093 }, { "epoch": 0.18, "learning_rate": 2.680103354222444e-06, "loss": 0.8325, "step": 23094 }, { "epoch": 0.18, "learning_rate": 2.6793690192987334e-06, "loss": 0.8447, "step": 23095 }, { "epoch": 0.18, "learning_rate": 2.6786347694281623e-06, "loss": 0.8881, "step": 23096 }, { "epoch": 0.18, "learning_rate": 2.677900604619258e-06, "loss": 0.8277, "step": 23097 }, { "epoch": 0.18, "learning_rate": 2.677166524880551e-06, "loss": 0.8698, "step": 23098 }, { "epoch": 0.18, "learning_rate": 2.6764325302205673e-06, "loss": 0.8548, "step": 23099 }, { "epoch": 0.18, "learning_rate": 2.6756986206478452e-06, "loss": 0.845, "step": 23100 }, { "epoch": 0.18, "learning_rate": 2.6749647961709003e-06, "loss": 0.8564, "step": 23101 }, { "epoch": 0.18, "learning_rate": 2.674231056798263e-06, "loss": 0.8561, "step": 23102 }, { "epoch": 0.18, "learning_rate": 2.673497402538461e-06, "loss": 0.8482, "step": 23103 }, { "epoch": 0.18, "learning_rate": 2.6727638334000118e-06, "loss": 0.8384, "step": 23104 }, { "epoch": 0.18, "learning_rate": 2.6720303493914367e-06, "loss": 0.8638, "step": 23105 }, { "epoch": 0.18, "learning_rate": 2.671296950521266e-06, "loss": 0.9046, "step": 23106 }, { "epoch": 0.18, "learning_rate": 2.670563636798019e-06, "loss": 0.8447, "step": 23107 }, { "epoch": 0.18, "learning_rate": 2.669830408230211e-06, "loss": 0.9069, "step": 23108 }, { "epoch": 0.18, "learning_rate": 2.6690972648263626e-06, "loss": 0.9009, "step": 23109 }, { "epoch": 0.18, "learning_rate": 2.668364206594991e-06, "loss": 0.8814, "step": 23110 }, { "epoch": 0.18, "learning_rate": 2.667631233544614e-06, "loss": 0.8241, "step": 23111 }, { "epoch": 0.18, "learning_rate": 2.6668983456837484e-06, "loss": 0.8513, "step": 23112 }, { "epoch": 0.18, "learning_rate": 2.6661655430209078e-06, "loss": 0.8456, "step": 23113 }, { "epoch": 0.18, "learning_rate": 2.6654328255646088e-06, "loss": 0.8517, "step": 23114 }, { "epoch": 0.18, "learning_rate": 2.6647001933233597e-06, "loss": 0.853, "step": 23115 }, { "epoch": 0.18, "learning_rate": 2.6639676463056743e-06, "loss": 0.8576, "step": 23116 }, { "epoch": 0.18, "learning_rate": 2.663235184520064e-06, "loss": 0.8535, "step": 23117 }, { "epoch": 0.18, "learning_rate": 2.662502807975039e-06, "loss": 0.8857, "step": 23118 }, { "epoch": 0.18, "learning_rate": 2.6617705166791077e-06, "loss": 0.8728, "step": 23119 }, { "epoch": 0.18, "learning_rate": 2.661038310640779e-06, "loss": 0.8704, "step": 23120 }, { "epoch": 0.18, "learning_rate": 2.6603061898685612e-06, "loss": 0.8705, "step": 23121 }, { "epoch": 0.18, "learning_rate": 2.659574154370955e-06, "loss": 0.859, "step": 23122 }, { "epoch": 0.18, "learning_rate": 2.65884220415647e-06, "loss": 0.8314, "step": 23123 }, { "epoch": 0.18, "learning_rate": 2.658110339233605e-06, "loss": 0.8785, "step": 23124 }, { "epoch": 0.18, "learning_rate": 2.6573785596108725e-06, "loss": 0.8849, "step": 23125 }, { "epoch": 0.18, "learning_rate": 2.656646865296767e-06, "loss": 0.8647, "step": 23126 }, { "epoch": 0.18, "learning_rate": 2.6559152562997913e-06, "loss": 0.8185, "step": 23127 }, { "epoch": 0.18, "learning_rate": 2.655183732628448e-06, "loss": 0.8464, "step": 23128 }, { "epoch": 0.18, "learning_rate": 2.654452294291228e-06, "loss": 0.8419, "step": 23129 }, { "epoch": 0.18, "learning_rate": 2.653720941296638e-06, "loss": 0.8642, "step": 23130 }, { "epoch": 0.18, "learning_rate": 2.6529896736531712e-06, "loss": 0.8682, "step": 23131 }, { "epoch": 0.18, "learning_rate": 2.6522584913693295e-06, "loss": 0.8377, "step": 23132 }, { "epoch": 0.18, "learning_rate": 2.6515273944535993e-06, "loss": 0.8418, "step": 23133 }, { "epoch": 0.18, "learning_rate": 2.650796382914478e-06, "loss": 0.8669, "step": 23134 }, { "epoch": 0.18, "learning_rate": 2.650065456760459e-06, "loss": 0.8682, "step": 23135 }, { "epoch": 0.18, "learning_rate": 2.6493346160000356e-06, "loss": 0.8624, "step": 23136 }, { "epoch": 0.18, "learning_rate": 2.6486038606416965e-06, "loss": 0.8427, "step": 23137 }, { "epoch": 0.18, "learning_rate": 2.6478731906939335e-06, "loss": 0.8588, "step": 23138 }, { "epoch": 0.18, "learning_rate": 2.6471426061652394e-06, "loss": 0.8409, "step": 23139 }, { "epoch": 0.18, "learning_rate": 2.6464121070640947e-06, "loss": 0.8383, "step": 23140 }, { "epoch": 0.18, "learning_rate": 2.6456816933989905e-06, "loss": 0.853, "step": 23141 }, { "epoch": 0.18, "learning_rate": 2.6449513651784133e-06, "loss": 0.8624, "step": 23142 }, { "epoch": 0.18, "learning_rate": 2.644221122410846e-06, "loss": 0.8932, "step": 23143 }, { "epoch": 0.18, "learning_rate": 2.6434909651047758e-06, "loss": 0.9086, "step": 23144 }, { "epoch": 0.18, "learning_rate": 2.642760893268684e-06, "loss": 0.8628, "step": 23145 }, { "epoch": 0.18, "learning_rate": 2.6420309069110563e-06, "loss": 0.8319, "step": 23146 }, { "epoch": 0.18, "learning_rate": 2.6413010060403675e-06, "loss": 0.8829, "step": 23147 }, { "epoch": 0.18, "learning_rate": 2.6405711906650975e-06, "loss": 0.8255, "step": 23148 }, { "epoch": 0.18, "learning_rate": 2.639841460793734e-06, "loss": 0.8635, "step": 23149 }, { "epoch": 0.18, "learning_rate": 2.639111816434752e-06, "loss": 0.8922, "step": 23150 }, { "epoch": 0.18, "learning_rate": 2.6383822575966245e-06, "loss": 0.8581, "step": 23151 }, { "epoch": 0.18, "learning_rate": 2.63765278428783e-06, "loss": 0.8332, "step": 23152 }, { "epoch": 0.18, "learning_rate": 2.6369233965168462e-06, "loss": 0.8715, "step": 23153 }, { "epoch": 0.18, "learning_rate": 2.6361940942921406e-06, "loss": 0.884, "step": 23154 }, { "epoch": 0.18, "learning_rate": 2.6354648776221927e-06, "loss": 0.8476, "step": 23155 }, { "epoch": 0.18, "learning_rate": 2.634735746515473e-06, "loss": 0.9022, "step": 23156 }, { "epoch": 0.18, "learning_rate": 2.634006700980457e-06, "loss": 0.9017, "step": 23157 }, { "epoch": 0.18, "learning_rate": 2.6332777410256074e-06, "loss": 0.8456, "step": 23158 }, { "epoch": 0.18, "learning_rate": 2.632548866659396e-06, "loss": 0.8279, "step": 23159 }, { "epoch": 0.18, "learning_rate": 2.631820077890291e-06, "loss": 0.8839, "step": 23160 }, { "epoch": 0.18, "learning_rate": 2.631091374726762e-06, "loss": 0.8582, "step": 23161 }, { "epoch": 0.18, "learning_rate": 2.630362757177274e-06, "loss": 0.8447, "step": 23162 }, { "epoch": 0.18, "learning_rate": 2.6296342252502916e-06, "loss": 0.8416, "step": 23163 }, { "epoch": 0.18, "learning_rate": 2.6289057789542816e-06, "loss": 0.8548, "step": 23164 }, { "epoch": 0.18, "learning_rate": 2.6281774182977026e-06, "loss": 0.8835, "step": 23165 }, { "epoch": 0.18, "learning_rate": 2.6274491432890205e-06, "loss": 0.8758, "step": 23166 }, { "epoch": 0.18, "learning_rate": 2.6267209539366946e-06, "loss": 0.866, "step": 23167 }, { "epoch": 0.18, "learning_rate": 2.625992850249187e-06, "loss": 0.8893, "step": 23168 }, { "epoch": 0.18, "learning_rate": 2.625264832234956e-06, "loss": 0.8853, "step": 23169 }, { "epoch": 0.18, "learning_rate": 2.62453689990246e-06, "loss": 0.8694, "step": 23170 }, { "epoch": 0.18, "learning_rate": 2.623809053260159e-06, "loss": 0.8198, "step": 23171 }, { "epoch": 0.18, "learning_rate": 2.6230812923165017e-06, "loss": 0.8815, "step": 23172 }, { "epoch": 0.18, "learning_rate": 2.6223536170799513e-06, "loss": 0.8561, "step": 23173 }, { "epoch": 0.18, "learning_rate": 2.6216260275589622e-06, "loss": 0.8742, "step": 23174 }, { "epoch": 0.18, "learning_rate": 2.6208985237619824e-06, "loss": 0.836, "step": 23175 }, { "epoch": 0.18, "learning_rate": 2.6201711056974664e-06, "loss": 0.8549, "step": 23176 }, { "epoch": 0.18, "learning_rate": 2.6194437733738652e-06, "loss": 0.8266, "step": 23177 }, { "epoch": 0.18, "learning_rate": 2.6187165267996305e-06, "loss": 0.855, "step": 23178 }, { "epoch": 0.18, "learning_rate": 2.6179893659832113e-06, "loss": 0.824, "step": 23179 }, { "epoch": 0.18, "learning_rate": 2.617262290933056e-06, "loss": 0.9097, "step": 23180 }, { "epoch": 0.18, "learning_rate": 2.6165353016576134e-06, "loss": 0.8344, "step": 23181 }, { "epoch": 0.18, "learning_rate": 2.6158083981653258e-06, "loss": 0.8675, "step": 23182 }, { "epoch": 0.18, "learning_rate": 2.61508158046464e-06, "loss": 0.8736, "step": 23183 }, { "epoch": 0.18, "learning_rate": 2.6143548485640024e-06, "loss": 0.831, "step": 23184 }, { "epoch": 0.18, "learning_rate": 2.6136282024718552e-06, "loss": 0.8479, "step": 23185 }, { "epoch": 0.18, "learning_rate": 2.6129016421966402e-06, "loss": 0.8676, "step": 23186 }, { "epoch": 0.18, "learning_rate": 2.6121751677467997e-06, "loss": 0.7737, "step": 23187 }, { "epoch": 0.18, "learning_rate": 2.6114487791307765e-06, "loss": 0.8496, "step": 23188 }, { "epoch": 0.18, "learning_rate": 2.6107224763570037e-06, "loss": 0.8335, "step": 23189 }, { "epoch": 0.18, "learning_rate": 2.6099962594339245e-06, "loss": 0.8452, "step": 23190 }, { "epoch": 0.18, "learning_rate": 2.6092701283699707e-06, "loss": 0.8321, "step": 23191 }, { "epoch": 0.18, "learning_rate": 2.6085440831735897e-06, "loss": 0.8722, "step": 23192 }, { "epoch": 0.18, "learning_rate": 2.6078181238532074e-06, "loss": 0.8834, "step": 23193 }, { "epoch": 0.18, "learning_rate": 2.607092250417261e-06, "loss": 0.8588, "step": 23194 }, { "epoch": 0.18, "learning_rate": 2.6063664628741836e-06, "loss": 0.8668, "step": 23195 }, { "epoch": 0.18, "learning_rate": 2.6056407612324106e-06, "loss": 0.8972, "step": 23196 }, { "epoch": 0.18, "learning_rate": 2.6049151455003653e-06, "loss": 0.9021, "step": 23197 }, { "epoch": 0.18, "learning_rate": 2.6041896156864864e-06, "loss": 0.8527, "step": 23198 }, { "epoch": 0.18, "learning_rate": 2.6034641717992026e-06, "loss": 0.8466, "step": 23199 }, { "epoch": 0.18, "learning_rate": 2.602738813846938e-06, "loss": 0.8247, "step": 23200 }, { "epoch": 0.18, "learning_rate": 2.602013541838122e-06, "loss": 0.8746, "step": 23201 }, { "epoch": 0.18, "learning_rate": 2.601288355781182e-06, "loss": 0.8995, "step": 23202 }, { "epoch": 0.18, "learning_rate": 2.6005632556845406e-06, "loss": 0.8154, "step": 23203 }, { "epoch": 0.18, "learning_rate": 2.5998382415566258e-06, "loss": 0.852, "step": 23204 }, { "epoch": 0.18, "learning_rate": 2.599113313405859e-06, "loss": 0.876, "step": 23205 }, { "epoch": 0.18, "learning_rate": 2.598388471240666e-06, "loss": 0.8321, "step": 23206 }, { "epoch": 0.18, "learning_rate": 2.5976637150694617e-06, "loss": 0.8497, "step": 23207 }, { "epoch": 0.18, "learning_rate": 2.5969390449006703e-06, "loss": 0.8754, "step": 23208 }, { "epoch": 0.18, "learning_rate": 2.596214460742712e-06, "loss": 0.8865, "step": 23209 }, { "epoch": 0.18, "learning_rate": 2.5954899626040033e-06, "loss": 0.8903, "step": 23210 }, { "epoch": 0.18, "learning_rate": 2.5947655504929625e-06, "loss": 0.8765, "step": 23211 }, { "epoch": 0.18, "learning_rate": 2.5940412244180058e-06, "loss": 0.8646, "step": 23212 }, { "epoch": 0.18, "learning_rate": 2.5933169843875527e-06, "loss": 0.8357, "step": 23213 }, { "epoch": 0.18, "learning_rate": 2.59259283041001e-06, "loss": 0.8337, "step": 23214 }, { "epoch": 0.18, "learning_rate": 2.5918687624937924e-06, "loss": 0.8775, "step": 23215 }, { "epoch": 0.18, "learning_rate": 2.5911447806473177e-06, "loss": 0.873, "step": 23216 }, { "epoch": 0.18, "learning_rate": 2.5904208848789967e-06, "loss": 0.863, "step": 23217 }, { "epoch": 0.18, "learning_rate": 2.5896970751972363e-06, "loss": 0.8533, "step": 23218 }, { "epoch": 0.18, "learning_rate": 2.5889733516104453e-06, "loss": 0.8666, "step": 23219 }, { "epoch": 0.18, "learning_rate": 2.588249714127039e-06, "loss": 0.8642, "step": 23220 }, { "epoch": 0.18, "learning_rate": 2.587526162755413e-06, "loss": 0.8485, "step": 23221 }, { "epoch": 0.18, "learning_rate": 2.586802697503983e-06, "loss": 0.8396, "step": 23222 }, { "epoch": 0.18, "learning_rate": 2.586079318381153e-06, "loss": 0.8266, "step": 23223 }, { "epoch": 0.18, "learning_rate": 2.585356025395328e-06, "loss": 0.8566, "step": 23224 }, { "epoch": 0.18, "learning_rate": 2.5846328185549074e-06, "loss": 0.9153, "step": 23225 }, { "epoch": 0.18, "learning_rate": 2.5839096978682963e-06, "loss": 0.9181, "step": 23226 }, { "epoch": 0.18, "learning_rate": 2.5831866633438963e-06, "loss": 0.806, "step": 23227 }, { "epoch": 0.18, "learning_rate": 2.5824637149901067e-06, "loss": 0.8722, "step": 23228 }, { "epoch": 0.18, "learning_rate": 2.5817408528153287e-06, "loss": 0.8367, "step": 23229 }, { "epoch": 0.18, "learning_rate": 2.581018076827958e-06, "loss": 0.8249, "step": 23230 }, { "epoch": 0.18, "learning_rate": 2.580295387036398e-06, "loss": 0.8345, "step": 23231 }, { "epoch": 0.18, "learning_rate": 2.579572783449037e-06, "loss": 0.8578, "step": 23232 }, { "epoch": 0.18, "learning_rate": 2.5788502660742744e-06, "loss": 0.8545, "step": 23233 }, { "epoch": 0.18, "learning_rate": 2.578127834920504e-06, "loss": 0.9019, "step": 23234 }, { "epoch": 0.18, "learning_rate": 2.5774054899961196e-06, "loss": 0.8344, "step": 23235 }, { "epoch": 0.18, "learning_rate": 2.576683231309515e-06, "loss": 0.8524, "step": 23236 }, { "epoch": 0.18, "learning_rate": 2.5759610588690786e-06, "loss": 0.8565, "step": 23237 }, { "epoch": 0.18, "learning_rate": 2.575238972683206e-06, "loss": 0.8609, "step": 23238 }, { "epoch": 0.18, "learning_rate": 2.5745169727602805e-06, "loss": 0.8447, "step": 23239 }, { "epoch": 0.18, "learning_rate": 2.573795059108689e-06, "loss": 0.8948, "step": 23240 }, { "epoch": 0.18, "learning_rate": 2.5730732317368255e-06, "loss": 0.8088, "step": 23241 }, { "epoch": 0.18, "learning_rate": 2.572351490653078e-06, "loss": 0.8439, "step": 23242 }, { "epoch": 0.18, "learning_rate": 2.5716298358658244e-06, "loss": 0.8262, "step": 23243 }, { "epoch": 0.18, "learning_rate": 2.5709082673834517e-06, "loss": 0.8716, "step": 23244 }, { "epoch": 0.18, "learning_rate": 2.570186785214345e-06, "loss": 0.8431, "step": 23245 }, { "epoch": 0.18, "learning_rate": 2.5694653893668854e-06, "loss": 0.8712, "step": 23246 }, { "epoch": 0.18, "learning_rate": 2.5687440798494533e-06, "loss": 0.8859, "step": 23247 }, { "epoch": 0.18, "learning_rate": 2.5680228566704314e-06, "loss": 0.8937, "step": 23248 }, { "epoch": 0.18, "learning_rate": 2.5673017198382e-06, "loss": 0.8681, "step": 23249 }, { "epoch": 0.18, "learning_rate": 2.5665806693611327e-06, "loss": 0.8806, "step": 23250 }, { "epoch": 0.18, "learning_rate": 2.5658597052476107e-06, "loss": 0.8396, "step": 23251 }, { "epoch": 0.18, "learning_rate": 2.5651388275060083e-06, "loss": 0.8781, "step": 23252 }, { "epoch": 0.18, "learning_rate": 2.5644180361447013e-06, "loss": 0.8528, "step": 23253 }, { "epoch": 0.18, "learning_rate": 2.5636973311720657e-06, "loss": 0.8406, "step": 23254 }, { "epoch": 0.18, "learning_rate": 2.5629767125964735e-06, "loss": 0.7925, "step": 23255 }, { "epoch": 0.18, "learning_rate": 2.5622561804262993e-06, "loss": 0.8492, "step": 23256 }, { "epoch": 0.18, "learning_rate": 2.56153573466991e-06, "loss": 0.8358, "step": 23257 }, { "epoch": 0.18, "learning_rate": 2.560815375335679e-06, "loss": 0.8517, "step": 23258 }, { "epoch": 0.18, "learning_rate": 2.560095102431974e-06, "loss": 0.8229, "step": 23259 }, { "epoch": 0.18, "learning_rate": 2.5593749159671644e-06, "loss": 0.8746, "step": 23260 }, { "epoch": 0.18, "learning_rate": 2.5586548159496182e-06, "loss": 0.8735, "step": 23261 }, { "epoch": 0.18, "learning_rate": 2.5579348023877005e-06, "loss": 0.8604, "step": 23262 }, { "epoch": 0.18, "learning_rate": 2.55721487528978e-06, "loss": 0.8525, "step": 23263 }, { "epoch": 0.18, "learning_rate": 2.5564950346642115e-06, "loss": 0.8574, "step": 23264 }, { "epoch": 0.18, "learning_rate": 2.555775280519369e-06, "loss": 0.8727, "step": 23265 }, { "epoch": 0.18, "learning_rate": 2.555055612863613e-06, "loss": 0.8706, "step": 23266 }, { "epoch": 0.18, "learning_rate": 2.5543360317053e-06, "loss": 0.8288, "step": 23267 }, { "epoch": 0.18, "learning_rate": 2.553616537052792e-06, "loss": 0.8293, "step": 23268 }, { "epoch": 0.18, "learning_rate": 2.55289712891445e-06, "loss": 0.8739, "step": 23269 }, { "epoch": 0.18, "learning_rate": 2.5521778072986303e-06, "loss": 0.8251, "step": 23270 }, { "epoch": 0.18, "learning_rate": 2.551458572213692e-06, "loss": 0.8759, "step": 23271 }, { "epoch": 0.18, "learning_rate": 2.5507394236679905e-06, "loss": 0.8766, "step": 23272 }, { "epoch": 0.18, "learning_rate": 2.5500203616698847e-06, "loss": 0.8676, "step": 23273 }, { "epoch": 0.18, "learning_rate": 2.5493013862277215e-06, "loss": 0.8818, "step": 23274 }, { "epoch": 0.18, "learning_rate": 2.5485824973498585e-06, "loss": 0.8632, "step": 23275 }, { "epoch": 0.18, "learning_rate": 2.547863695044648e-06, "loss": 0.8864, "step": 23276 }, { "epoch": 0.18, "learning_rate": 2.5471449793204393e-06, "loss": 0.8404, "step": 23277 }, { "epoch": 0.18, "learning_rate": 2.5464263501855857e-06, "loss": 0.8416, "step": 23278 }, { "epoch": 0.18, "learning_rate": 2.5457078076484343e-06, "loss": 0.8003, "step": 23279 }, { "epoch": 0.18, "learning_rate": 2.5449893517173363e-06, "loss": 0.8647, "step": 23280 }, { "epoch": 0.18, "learning_rate": 2.544270982400633e-06, "loss": 0.8429, "step": 23281 }, { "epoch": 0.18, "learning_rate": 2.543552699706675e-06, "loss": 0.8908, "step": 23282 }, { "epoch": 0.18, "learning_rate": 2.542834503643803e-06, "loss": 0.8579, "step": 23283 }, { "epoch": 0.18, "learning_rate": 2.542116394220371e-06, "loss": 0.878, "step": 23284 }, { "epoch": 0.18, "learning_rate": 2.541398371444713e-06, "loss": 0.8357, "step": 23285 }, { "epoch": 0.18, "learning_rate": 2.5406804353251746e-06, "loss": 0.8371, "step": 23286 }, { "epoch": 0.18, "learning_rate": 2.539962585870096e-06, "loss": 0.8329, "step": 23287 }, { "epoch": 0.18, "learning_rate": 2.5392448230878206e-06, "loss": 0.8692, "step": 23288 }, { "epoch": 0.18, "learning_rate": 2.5385271469866802e-06, "loss": 0.8708, "step": 23289 }, { "epoch": 0.18, "learning_rate": 2.5378095575750197e-06, "loss": 0.8976, "step": 23290 }, { "epoch": 0.18, "learning_rate": 2.5370920548611767e-06, "loss": 0.901, "step": 23291 }, { "epoch": 0.18, "learning_rate": 2.5363746388534828e-06, "loss": 0.8616, "step": 23292 }, { "epoch": 0.18, "learning_rate": 2.5356573095602744e-06, "loss": 0.8568, "step": 23293 }, { "epoch": 0.18, "learning_rate": 2.534940066989887e-06, "loss": 0.8704, "step": 23294 }, { "epoch": 0.18, "learning_rate": 2.534222911150652e-06, "loss": 0.8264, "step": 23295 }, { "epoch": 0.18, "learning_rate": 2.533505842050904e-06, "loss": 0.8868, "step": 23296 }, { "epoch": 0.18, "learning_rate": 2.5327888596989725e-06, "loss": 0.8749, "step": 23297 }, { "epoch": 0.18, "learning_rate": 2.532071964103191e-06, "loss": 0.8595, "step": 23298 }, { "epoch": 0.18, "learning_rate": 2.531355155271882e-06, "loss": 0.8369, "step": 23299 }, { "epoch": 0.18, "learning_rate": 2.5306384332133773e-06, "loss": 0.872, "step": 23300 }, { "epoch": 0.18, "learning_rate": 2.5299217979360035e-06, "loss": 0.842, "step": 23301 }, { "epoch": 0.18, "learning_rate": 2.529205249448087e-06, "loss": 0.873, "step": 23302 }, { "epoch": 0.18, "learning_rate": 2.5284887877579523e-06, "loss": 0.837, "step": 23303 }, { "epoch": 0.18, "learning_rate": 2.5277724128739246e-06, "loss": 0.8813, "step": 23304 }, { "epoch": 0.18, "learning_rate": 2.527056124804329e-06, "loss": 0.8643, "step": 23305 }, { "epoch": 0.18, "learning_rate": 2.5263399235574827e-06, "loss": 0.8802, "step": 23306 }, { "epoch": 0.18, "learning_rate": 2.525623809141704e-06, "loss": 0.8574, "step": 23307 }, { "epoch": 0.18, "learning_rate": 2.5249077815653224e-06, "loss": 0.8352, "step": 23308 }, { "epoch": 0.19, "learning_rate": 2.5241918408366538e-06, "loss": 0.8855, "step": 23309 }, { "epoch": 0.19, "learning_rate": 2.5234759869640124e-06, "loss": 0.851, "step": 23310 }, { "epoch": 0.19, "learning_rate": 2.522760219955718e-06, "loss": 0.8418, "step": 23311 }, { "epoch": 0.19, "learning_rate": 2.5220445398200876e-06, "loss": 0.8374, "step": 23312 }, { "epoch": 0.19, "learning_rate": 2.52132894656543e-06, "loss": 0.8355, "step": 23313 }, { "epoch": 0.19, "learning_rate": 2.520613440200066e-06, "loss": 0.8984, "step": 23314 }, { "epoch": 0.19, "learning_rate": 2.5198980207323064e-06, "loss": 0.8538, "step": 23315 }, { "epoch": 0.19, "learning_rate": 2.519182688170466e-06, "loss": 0.8727, "step": 23316 }, { "epoch": 0.19, "learning_rate": 2.5184674425228497e-06, "loss": 0.8591, "step": 23317 }, { "epoch": 0.19, "learning_rate": 2.517752283797771e-06, "loss": 0.8664, "step": 23318 }, { "epoch": 0.19, "learning_rate": 2.517037212003538e-06, "loss": 0.8391, "step": 23319 }, { "epoch": 0.19, "learning_rate": 2.516322227148459e-06, "loss": 0.872, "step": 23320 }, { "epoch": 0.19, "learning_rate": 2.51560732924084e-06, "loss": 0.8648, "step": 23321 }, { "epoch": 0.19, "learning_rate": 2.514892518288988e-06, "loss": 0.8869, "step": 23322 }, { "epoch": 0.19, "learning_rate": 2.5141777943012114e-06, "loss": 0.8421, "step": 23323 }, { "epoch": 0.19, "learning_rate": 2.513463157285807e-06, "loss": 0.8554, "step": 23324 }, { "epoch": 0.19, "learning_rate": 2.512748607251081e-06, "loss": 0.8964, "step": 23325 }, { "epoch": 0.19, "learning_rate": 2.5120341442053344e-06, "loss": 0.8958, "step": 23326 }, { "epoch": 0.19, "learning_rate": 2.511319768156869e-06, "loss": 0.8239, "step": 23327 }, { "epoch": 0.19, "learning_rate": 2.5106054791139846e-06, "loss": 0.8576, "step": 23328 }, { "epoch": 0.19, "learning_rate": 2.5098912770849793e-06, "loss": 0.8601, "step": 23329 }, { "epoch": 0.19, "learning_rate": 2.5091771620781545e-06, "loss": 0.8653, "step": 23330 }, { "epoch": 0.19, "learning_rate": 2.5084631341018016e-06, "loss": 0.8289, "step": 23331 }, { "epoch": 0.19, "learning_rate": 2.507749193164215e-06, "loss": 0.8871, "step": 23332 }, { "epoch": 0.19, "learning_rate": 2.507035339273697e-06, "loss": 0.8821, "step": 23333 }, { "epoch": 0.19, "learning_rate": 2.5063215724385394e-06, "loss": 0.8299, "step": 23334 }, { "epoch": 0.19, "learning_rate": 2.50560789266703e-06, "loss": 0.8391, "step": 23335 }, { "epoch": 0.19, "learning_rate": 2.504894299967464e-06, "loss": 0.8244, "step": 23336 }, { "epoch": 0.19, "learning_rate": 2.5041807943481323e-06, "loss": 0.864, "step": 23337 }, { "epoch": 0.19, "learning_rate": 2.5034673758173234e-06, "loss": 0.8328, "step": 23338 }, { "epoch": 0.19, "learning_rate": 2.5027540443833266e-06, "loss": 0.8473, "step": 23339 }, { "epoch": 0.19, "learning_rate": 2.502040800054429e-06, "loss": 0.8585, "step": 23340 }, { "epoch": 0.19, "learning_rate": 2.501327642838921e-06, "loss": 0.8584, "step": 23341 }, { "epoch": 0.19, "learning_rate": 2.5006145727450825e-06, "loss": 0.8696, "step": 23342 }, { "epoch": 0.19, "learning_rate": 2.499901589781201e-06, "loss": 0.8413, "step": 23343 }, { "epoch": 0.19, "learning_rate": 2.4991886939555587e-06, "loss": 0.8771, "step": 23344 }, { "epoch": 0.19, "learning_rate": 2.49847588527644e-06, "loss": 0.8814, "step": 23345 }, { "epoch": 0.19, "learning_rate": 2.4977631637521262e-06, "loss": 0.8168, "step": 23346 }, { "epoch": 0.19, "learning_rate": 2.4970505293908974e-06, "loss": 0.8698, "step": 23347 }, { "epoch": 0.19, "learning_rate": 2.496337982201037e-06, "loss": 0.8667, "step": 23348 }, { "epoch": 0.19, "learning_rate": 2.4956255221908153e-06, "loss": 0.8722, "step": 23349 }, { "epoch": 0.19, "learning_rate": 2.494913149368513e-06, "loss": 0.8626, "step": 23350 }, { "epoch": 0.19, "learning_rate": 2.4942008637424143e-06, "loss": 0.8713, "step": 23351 }, { "epoch": 0.19, "learning_rate": 2.4934886653207847e-06, "loss": 0.8896, "step": 23352 }, { "epoch": 0.19, "learning_rate": 2.492776554111903e-06, "loss": 0.8626, "step": 23353 }, { "epoch": 0.19, "learning_rate": 2.492064530124042e-06, "loss": 0.873, "step": 23354 }, { "epoch": 0.19, "learning_rate": 2.491352593365478e-06, "loss": 0.846, "step": 23355 }, { "epoch": 0.19, "learning_rate": 2.4906407438444724e-06, "loss": 0.8633, "step": 23356 }, { "epoch": 0.19, "learning_rate": 2.489928981569306e-06, "loss": 0.8697, "step": 23357 }, { "epoch": 0.19, "learning_rate": 2.4892173065482473e-06, "loss": 0.8738, "step": 23358 }, { "epoch": 0.19, "learning_rate": 2.488505718789559e-06, "loss": 0.8929, "step": 23359 }, { "epoch": 0.19, "learning_rate": 2.4877942183015104e-06, "loss": 0.851, "step": 23360 }, { "epoch": 0.19, "learning_rate": 2.4870828050923703e-06, "loss": 0.8838, "step": 23361 }, { "epoch": 0.19, "learning_rate": 2.486371479170402e-06, "loss": 0.8561, "step": 23362 }, { "epoch": 0.19, "learning_rate": 2.4856602405438713e-06, "loss": 0.8748, "step": 23363 }, { "epoch": 0.19, "learning_rate": 2.4849490892210402e-06, "loss": 0.8433, "step": 23364 }, { "epoch": 0.19, "learning_rate": 2.484238025210176e-06, "loss": 0.8891, "step": 23365 }, { "epoch": 0.19, "learning_rate": 2.4835270485195318e-06, "loss": 0.8463, "step": 23366 }, { "epoch": 0.19, "learning_rate": 2.482816159157373e-06, "loss": 0.8221, "step": 23367 }, { "epoch": 0.19, "learning_rate": 2.4821053571319565e-06, "loss": 0.8717, "step": 23368 }, { "epoch": 0.19, "learning_rate": 2.481394642451542e-06, "loss": 0.8436, "step": 23369 }, { "epoch": 0.19, "learning_rate": 2.4806840151243884e-06, "loss": 0.8592, "step": 23370 }, { "epoch": 0.19, "learning_rate": 2.479973475158749e-06, "loss": 0.8545, "step": 23371 }, { "epoch": 0.19, "learning_rate": 2.4792630225628835e-06, "loss": 0.8845, "step": 23372 }, { "epoch": 0.19, "learning_rate": 2.47855265734504e-06, "loss": 0.8455, "step": 23373 }, { "epoch": 0.19, "learning_rate": 2.4778423795134755e-06, "loss": 0.8533, "step": 23374 }, { "epoch": 0.19, "learning_rate": 2.4771321890764377e-06, "loss": 0.8309, "step": 23375 }, { "epoch": 0.19, "learning_rate": 2.4764220860421873e-06, "loss": 0.8556, "step": 23376 }, { "epoch": 0.19, "learning_rate": 2.4757120704189664e-06, "loss": 0.8366, "step": 23377 }, { "epoch": 0.19, "learning_rate": 2.475002142215026e-06, "loss": 0.8445, "step": 23378 }, { "epoch": 0.19, "learning_rate": 2.4742923014386154e-06, "loss": 0.8453, "step": 23379 }, { "epoch": 0.19, "learning_rate": 2.4735825480979802e-06, "loss": 0.8595, "step": 23380 }, { "epoch": 0.19, "learning_rate": 2.4728728822013683e-06, "loss": 0.8402, "step": 23381 }, { "epoch": 0.19, "learning_rate": 2.4721633037570223e-06, "loss": 0.8271, "step": 23382 }, { "epoch": 0.19, "learning_rate": 2.4714538127731926e-06, "loss": 0.8428, "step": 23383 }, { "epoch": 0.19, "learning_rate": 2.4707444092581133e-06, "loss": 0.8293, "step": 23384 }, { "epoch": 0.19, "learning_rate": 2.4700350932200314e-06, "loss": 0.8926, "step": 23385 }, { "epoch": 0.19, "learning_rate": 2.469325864667186e-06, "loss": 0.8585, "step": 23386 }, { "epoch": 0.19, "learning_rate": 2.4686167236078196e-06, "loss": 0.8352, "step": 23387 }, { "epoch": 0.19, "learning_rate": 2.467907670050169e-06, "loss": 0.8713, "step": 23388 }, { "epoch": 0.19, "learning_rate": 2.4671987040024736e-06, "loss": 0.8192, "step": 23389 }, { "epoch": 0.19, "learning_rate": 2.466489825472973e-06, "loss": 0.8501, "step": 23390 }, { "epoch": 0.19, "learning_rate": 2.4657810344698964e-06, "loss": 0.849, "step": 23391 }, { "epoch": 0.19, "learning_rate": 2.4650723310014823e-06, "loss": 0.8496, "step": 23392 }, { "epoch": 0.19, "learning_rate": 2.4643637150759647e-06, "loss": 0.8917, "step": 23393 }, { "epoch": 0.19, "learning_rate": 2.4636551867015767e-06, "loss": 0.8331, "step": 23394 }, { "epoch": 0.19, "learning_rate": 2.46294674588655e-06, "loss": 0.835, "step": 23395 }, { "epoch": 0.19, "learning_rate": 2.462238392639116e-06, "loss": 0.8332, "step": 23396 }, { "epoch": 0.19, "learning_rate": 2.461530126967505e-06, "loss": 0.8653, "step": 23397 }, { "epoch": 0.19, "learning_rate": 2.4608219488799434e-06, "loss": 0.8631, "step": 23398 }, { "epoch": 0.19, "learning_rate": 2.4601138583846563e-06, "loss": 0.8529, "step": 23399 }, { "epoch": 0.19, "learning_rate": 2.459405855489878e-06, "loss": 0.8626, "step": 23400 }, { "epoch": 0.19, "learning_rate": 2.4586979402038347e-06, "loss": 0.8857, "step": 23401 }, { "epoch": 0.19, "learning_rate": 2.4579901125347426e-06, "loss": 0.8703, "step": 23402 }, { "epoch": 0.19, "learning_rate": 2.457282372490831e-06, "loss": 0.8711, "step": 23403 }, { "epoch": 0.19, "learning_rate": 2.456574720080326e-06, "loss": 0.8972, "step": 23404 }, { "epoch": 0.19, "learning_rate": 2.4558671553114375e-06, "loss": 0.8498, "step": 23405 }, { "epoch": 0.19, "learning_rate": 2.455159678192397e-06, "loss": 0.831, "step": 23406 }, { "epoch": 0.19, "learning_rate": 2.4544522887314203e-06, "loss": 0.8696, "step": 23407 }, { "epoch": 0.19, "learning_rate": 2.4537449869367314e-06, "loss": 0.8548, "step": 23408 }, { "epoch": 0.19, "learning_rate": 2.4530377728165388e-06, "loss": 0.8526, "step": 23409 }, { "epoch": 0.19, "learning_rate": 2.452330646379064e-06, "loss": 0.8368, "step": 23410 }, { "epoch": 0.19, "learning_rate": 2.4516236076325216e-06, "loss": 0.8655, "step": 23411 }, { "epoch": 0.19, "learning_rate": 2.4509166565851274e-06, "loss": 0.8507, "step": 23412 }, { "epoch": 0.19, "learning_rate": 2.450209793245093e-06, "loss": 0.8825, "step": 23413 }, { "epoch": 0.19, "learning_rate": 2.4495030176206336e-06, "loss": 0.8838, "step": 23414 }, { "epoch": 0.19, "learning_rate": 2.4487963297199614e-06, "loss": 0.8691, "step": 23415 }, { "epoch": 0.19, "learning_rate": 2.448089729551282e-06, "loss": 0.8398, "step": 23416 }, { "epoch": 0.19, "learning_rate": 2.4473832171228076e-06, "loss": 0.8645, "step": 23417 }, { "epoch": 0.19, "learning_rate": 2.446676792442747e-06, "loss": 0.9015, "step": 23418 }, { "epoch": 0.19, "learning_rate": 2.4459704555193066e-06, "loss": 0.8593, "step": 23419 }, { "epoch": 0.19, "learning_rate": 2.4452642063606947e-06, "loss": 0.8388, "step": 23420 }, { "epoch": 0.19, "learning_rate": 2.4445580449751138e-06, "loss": 0.8731, "step": 23421 }, { "epoch": 0.19, "learning_rate": 2.4438519713707742e-06, "loss": 0.8577, "step": 23422 }, { "epoch": 0.19, "learning_rate": 2.4431459855558694e-06, "loss": 0.8302, "step": 23423 }, { "epoch": 0.19, "learning_rate": 2.44244008753861e-06, "loss": 0.8594, "step": 23424 }, { "epoch": 0.19, "learning_rate": 2.4417342773271944e-06, "loss": 0.8791, "step": 23425 }, { "epoch": 0.19, "learning_rate": 2.4410285549298275e-06, "loss": 0.8308, "step": 23426 }, { "epoch": 0.19, "learning_rate": 2.440322920354701e-06, "loss": 0.882, "step": 23427 }, { "epoch": 0.19, "learning_rate": 2.439617373610016e-06, "loss": 0.8521, "step": 23428 }, { "epoch": 0.19, "learning_rate": 2.43891191470397e-06, "loss": 0.8605, "step": 23429 }, { "epoch": 0.19, "learning_rate": 2.43820654364476e-06, "loss": 0.8418, "step": 23430 }, { "epoch": 0.19, "learning_rate": 2.4375012604405802e-06, "loss": 0.8355, "step": 23431 }, { "epoch": 0.19, "learning_rate": 2.436796065099626e-06, "loss": 0.8569, "step": 23432 }, { "epoch": 0.19, "learning_rate": 2.4360909576300917e-06, "loss": 0.8504, "step": 23433 }, { "epoch": 0.19, "learning_rate": 2.4353859380401646e-06, "loss": 0.888, "step": 23434 }, { "epoch": 0.19, "learning_rate": 2.43468100633804e-06, "loss": 0.8592, "step": 23435 }, { "epoch": 0.19, "learning_rate": 2.433976162531906e-06, "loss": 0.839, "step": 23436 }, { "epoch": 0.19, "learning_rate": 2.4332714066299523e-06, "loss": 0.8468, "step": 23437 }, { "epoch": 0.19, "learning_rate": 2.432566738640367e-06, "loss": 0.8907, "step": 23438 }, { "epoch": 0.19, "learning_rate": 2.4318621585713374e-06, "loss": 0.8357, "step": 23439 }, { "epoch": 0.19, "learning_rate": 2.431157666431052e-06, "loss": 0.8601, "step": 23440 }, { "epoch": 0.19, "learning_rate": 2.430453262227691e-06, "loss": 0.86, "step": 23441 }, { "epoch": 0.19, "learning_rate": 2.4297489459694366e-06, "loss": 0.8694, "step": 23442 }, { "epoch": 0.19, "learning_rate": 2.4290447176644815e-06, "loss": 0.838, "step": 23443 }, { "epoch": 0.19, "learning_rate": 2.428340577320999e-06, "loss": 0.8598, "step": 23444 }, { "epoch": 0.19, "learning_rate": 2.427636524947173e-06, "loss": 0.8739, "step": 23445 }, { "epoch": 0.19, "learning_rate": 2.426932560551183e-06, "loss": 0.8497, "step": 23446 }, { "epoch": 0.19, "learning_rate": 2.426228684141211e-06, "loss": 0.8488, "step": 23447 }, { "epoch": 0.19, "learning_rate": 2.425524895725425e-06, "loss": 0.9002, "step": 23448 }, { "epoch": 0.19, "learning_rate": 2.4248211953120126e-06, "loss": 0.8674, "step": 23449 }, { "epoch": 0.19, "learning_rate": 2.424117582909148e-06, "loss": 0.8854, "step": 23450 }, { "epoch": 0.19, "learning_rate": 2.423414058525001e-06, "loss": 0.8389, "step": 23451 }, { "epoch": 0.19, "learning_rate": 2.4227106221677476e-06, "loss": 0.8853, "step": 23452 }, { "epoch": 0.19, "learning_rate": 2.4220072738455602e-06, "loss": 0.8612, "step": 23453 }, { "epoch": 0.19, "learning_rate": 2.421304013566612e-06, "loss": 0.8776, "step": 23454 }, { "epoch": 0.19, "learning_rate": 2.4206008413390725e-06, "loss": 0.8533, "step": 23455 }, { "epoch": 0.19, "learning_rate": 2.4198977571711114e-06, "loss": 0.8731, "step": 23456 }, { "epoch": 0.19, "learning_rate": 2.419194761070901e-06, "loss": 0.8977, "step": 23457 }, { "epoch": 0.19, "learning_rate": 2.4184918530466027e-06, "loss": 0.8841, "step": 23458 }, { "epoch": 0.19, "learning_rate": 2.4177890331063847e-06, "loss": 0.8536, "step": 23459 }, { "epoch": 0.19, "learning_rate": 2.4170863012584144e-06, "loss": 0.8377, "step": 23460 }, { "epoch": 0.19, "learning_rate": 2.4163836575108555e-06, "loss": 0.8742, "step": 23461 }, { "epoch": 0.19, "learning_rate": 2.415681101871873e-06, "loss": 0.839, "step": 23462 }, { "epoch": 0.19, "learning_rate": 2.4149786343496264e-06, "loss": 0.7967, "step": 23463 }, { "epoch": 0.19, "learning_rate": 2.4142762549522823e-06, "loss": 0.8722, "step": 23464 }, { "epoch": 0.19, "learning_rate": 2.413573963687994e-06, "loss": 0.833, "step": 23465 }, { "epoch": 0.19, "learning_rate": 2.4128717605649255e-06, "loss": 0.8621, "step": 23466 }, { "epoch": 0.19, "learning_rate": 2.4121696455912314e-06, "loss": 0.8155, "step": 23467 }, { "epoch": 0.19, "learning_rate": 2.4114676187750773e-06, "loss": 0.872, "step": 23468 }, { "epoch": 0.19, "learning_rate": 2.4107656801246115e-06, "loss": 0.885, "step": 23469 }, { "epoch": 0.19, "learning_rate": 2.4100638296479916e-06, "loss": 0.8569, "step": 23470 }, { "epoch": 0.19, "learning_rate": 2.4093620673533723e-06, "loss": 0.8435, "step": 23471 }, { "epoch": 0.19, "learning_rate": 2.408660393248907e-06, "loss": 0.9176, "step": 23472 }, { "epoch": 0.19, "learning_rate": 2.4079588073427483e-06, "loss": 0.8754, "step": 23473 }, { "epoch": 0.19, "learning_rate": 2.407257309643046e-06, "loss": 0.8596, "step": 23474 }, { "epoch": 0.19, "learning_rate": 2.406555900157955e-06, "loss": 0.8326, "step": 23475 }, { "epoch": 0.19, "learning_rate": 2.405854578895618e-06, "loss": 0.8708, "step": 23476 }, { "epoch": 0.19, "learning_rate": 2.405153345864184e-06, "loss": 0.8481, "step": 23477 }, { "epoch": 0.19, "learning_rate": 2.404452201071804e-06, "loss": 0.8426, "step": 23478 }, { "epoch": 0.19, "learning_rate": 2.4037511445266215e-06, "loss": 0.8478, "step": 23479 }, { "epoch": 0.19, "learning_rate": 2.403050176236781e-06, "loss": 0.8603, "step": 23480 }, { "epoch": 0.19, "learning_rate": 2.4023492962104288e-06, "loss": 0.8523, "step": 23481 }, { "epoch": 0.19, "learning_rate": 2.401648504455709e-06, "loss": 0.8615, "step": 23482 }, { "epoch": 0.19, "learning_rate": 2.4009478009807595e-06, "loss": 0.8368, "step": 23483 }, { "epoch": 0.19, "learning_rate": 2.4002471857937227e-06, "loss": 0.8516, "step": 23484 }, { "epoch": 0.19, "learning_rate": 2.399546658902736e-06, "loss": 0.8406, "step": 23485 }, { "epoch": 0.19, "learning_rate": 2.398846220315947e-06, "loss": 0.8112, "step": 23486 }, { "epoch": 0.19, "learning_rate": 2.398145870041485e-06, "loss": 0.8409, "step": 23487 }, { "epoch": 0.19, "learning_rate": 2.3974456080874896e-06, "loss": 0.897, "step": 23488 }, { "epoch": 0.19, "learning_rate": 2.3967454344621e-06, "loss": 0.8697, "step": 23489 }, { "epoch": 0.19, "learning_rate": 2.396045349173445e-06, "loss": 0.8294, "step": 23490 }, { "epoch": 0.19, "learning_rate": 2.3953453522296577e-06, "loss": 0.8367, "step": 23491 }, { "epoch": 0.19, "learning_rate": 2.3946454436388777e-06, "loss": 0.8974, "step": 23492 }, { "epoch": 0.19, "learning_rate": 2.3939456234092373e-06, "loss": 0.8209, "step": 23493 }, { "epoch": 0.19, "learning_rate": 2.3932458915488587e-06, "loss": 0.8909, "step": 23494 }, { "epoch": 0.19, "learning_rate": 2.3925462480658767e-06, "loss": 0.8832, "step": 23495 }, { "epoch": 0.19, "learning_rate": 2.3918466929684205e-06, "loss": 0.8488, "step": 23496 }, { "epoch": 0.19, "learning_rate": 2.391147226264615e-06, "loss": 0.8656, "step": 23497 }, { "epoch": 0.19, "learning_rate": 2.3904478479625903e-06, "loss": 0.8375, "step": 23498 }, { "epoch": 0.19, "learning_rate": 2.3897485580704684e-06, "loss": 0.8684, "step": 23499 }, { "epoch": 0.19, "learning_rate": 2.389049356596379e-06, "loss": 0.8892, "step": 23500 }, { "epoch": 0.19, "learning_rate": 2.3883502435484406e-06, "loss": 0.8788, "step": 23501 }, { "epoch": 0.19, "learning_rate": 2.3876512189347757e-06, "loss": 0.8694, "step": 23502 }, { "epoch": 0.19, "learning_rate": 2.3869522827635084e-06, "loss": 0.8591, "step": 23503 }, { "epoch": 0.19, "learning_rate": 2.386253435042757e-06, "loss": 0.8502, "step": 23504 }, { "epoch": 0.19, "learning_rate": 2.3855546757806426e-06, "loss": 0.8658, "step": 23505 }, { "epoch": 0.19, "learning_rate": 2.384856004985283e-06, "loss": 0.8597, "step": 23506 }, { "epoch": 0.19, "learning_rate": 2.3841574226647988e-06, "loss": 0.8581, "step": 23507 }, { "epoch": 0.19, "learning_rate": 2.3834589288273003e-06, "loss": 0.8545, "step": 23508 }, { "epoch": 0.19, "learning_rate": 2.382760523480905e-06, "loss": 0.8941, "step": 23509 }, { "epoch": 0.19, "learning_rate": 2.3820622066337275e-06, "loss": 0.8864, "step": 23510 }, { "epoch": 0.19, "learning_rate": 2.3813639782938812e-06, "loss": 0.8494, "step": 23511 }, { "epoch": 0.19, "learning_rate": 2.3806658384694793e-06, "loss": 0.8503, "step": 23512 }, { "epoch": 0.19, "learning_rate": 2.379967787168631e-06, "loss": 0.8941, "step": 23513 }, { "epoch": 0.19, "learning_rate": 2.3792698243994506e-06, "loss": 0.831, "step": 23514 }, { "epoch": 0.19, "learning_rate": 2.378571950170039e-06, "loss": 0.8499, "step": 23515 }, { "epoch": 0.19, "learning_rate": 2.377874164488512e-06, "loss": 0.8383, "step": 23516 }, { "epoch": 0.19, "learning_rate": 2.3771764673629737e-06, "loss": 0.8863, "step": 23517 }, { "epoch": 0.19, "learning_rate": 2.3764788588015344e-06, "loss": 0.8295, "step": 23518 }, { "epoch": 0.19, "learning_rate": 2.375781338812292e-06, "loss": 0.8195, "step": 23519 }, { "epoch": 0.19, "learning_rate": 2.375083907403354e-06, "loss": 0.8523, "step": 23520 }, { "epoch": 0.19, "learning_rate": 2.3743865645828222e-06, "loss": 0.8527, "step": 23521 }, { "epoch": 0.19, "learning_rate": 2.3736893103588e-06, "loss": 0.8851, "step": 23522 }, { "epoch": 0.19, "learning_rate": 2.3729921447393876e-06, "loss": 0.837, "step": 23523 }, { "epoch": 0.19, "learning_rate": 2.372295067732685e-06, "loss": 0.8311, "step": 23524 }, { "epoch": 0.19, "learning_rate": 2.371598079346794e-06, "loss": 0.8615, "step": 23525 }, { "epoch": 0.19, "learning_rate": 2.370901179589806e-06, "loss": 0.8475, "step": 23526 }, { "epoch": 0.19, "learning_rate": 2.3702043684698216e-06, "loss": 0.8501, "step": 23527 }, { "epoch": 0.19, "learning_rate": 2.3695076459949364e-06, "loss": 0.911, "step": 23528 }, { "epoch": 0.19, "learning_rate": 2.3688110121732443e-06, "loss": 0.8581, "step": 23529 }, { "epoch": 0.19, "learning_rate": 2.368114467012841e-06, "loss": 0.8387, "step": 23530 }, { "epoch": 0.19, "learning_rate": 2.367418010521816e-06, "loss": 0.868, "step": 23531 }, { "epoch": 0.19, "learning_rate": 2.366721642708266e-06, "loss": 0.8743, "step": 23532 }, { "epoch": 0.19, "learning_rate": 2.3660253635802753e-06, "loss": 0.8862, "step": 23533 }, { "epoch": 0.19, "learning_rate": 2.365329173145934e-06, "loss": 0.8693, "step": 23534 }, { "epoch": 0.19, "learning_rate": 2.3646330714133394e-06, "loss": 0.8543, "step": 23535 }, { "epoch": 0.19, "learning_rate": 2.3639370583905685e-06, "loss": 0.8195, "step": 23536 }, { "epoch": 0.19, "learning_rate": 2.363241134085712e-06, "loss": 0.8144, "step": 23537 }, { "epoch": 0.19, "learning_rate": 2.3625452985068552e-06, "loss": 0.8411, "step": 23538 }, { "epoch": 0.19, "learning_rate": 2.3618495516620853e-06, "loss": 0.8429, "step": 23539 }, { "epoch": 0.19, "learning_rate": 2.361153893559478e-06, "loss": 0.846, "step": 23540 }, { "epoch": 0.19, "learning_rate": 2.3604583242071234e-06, "loss": 0.8706, "step": 23541 }, { "epoch": 0.19, "learning_rate": 2.3597628436131026e-06, "loss": 0.903, "step": 23542 }, { "epoch": 0.19, "learning_rate": 2.3590674517854907e-06, "loss": 0.8371, "step": 23543 }, { "epoch": 0.19, "learning_rate": 2.3583721487323683e-06, "loss": 0.8399, "step": 23544 }, { "epoch": 0.19, "learning_rate": 2.3576769344618166e-06, "loss": 0.8518, "step": 23545 }, { "epoch": 0.19, "learning_rate": 2.356981808981911e-06, "loss": 0.8572, "step": 23546 }, { "epoch": 0.19, "learning_rate": 2.3562867723007264e-06, "loss": 0.8555, "step": 23547 }, { "epoch": 0.19, "learning_rate": 2.35559182442634e-06, "loss": 0.8099, "step": 23548 }, { "epoch": 0.19, "learning_rate": 2.3548969653668284e-06, "loss": 0.8366, "step": 23549 }, { "epoch": 0.19, "learning_rate": 2.3542021951302574e-06, "loss": 0.872, "step": 23550 }, { "epoch": 0.19, "learning_rate": 2.3535075137247045e-06, "loss": 0.9151, "step": 23551 }, { "epoch": 0.19, "learning_rate": 2.3528129211582384e-06, "loss": 0.8457, "step": 23552 }, { "epoch": 0.19, "learning_rate": 2.35211841743893e-06, "loss": 0.8452, "step": 23553 }, { "epoch": 0.19, "learning_rate": 2.351424002574848e-06, "loss": 0.8471, "step": 23554 }, { "epoch": 0.19, "learning_rate": 2.3507296765740616e-06, "loss": 0.8599, "step": 23555 }, { "epoch": 0.19, "learning_rate": 2.350035439444638e-06, "loss": 0.8892, "step": 23556 }, { "epoch": 0.19, "learning_rate": 2.34934129119464e-06, "loss": 0.9008, "step": 23557 }, { "epoch": 0.19, "learning_rate": 2.348647231832131e-06, "loss": 0.8426, "step": 23558 }, { "epoch": 0.19, "learning_rate": 2.347953261365179e-06, "loss": 0.8252, "step": 23559 }, { "epoch": 0.19, "learning_rate": 2.3472593798018506e-06, "loss": 0.8281, "step": 23560 }, { "epoch": 0.19, "learning_rate": 2.346565587150198e-06, "loss": 0.8843, "step": 23561 }, { "epoch": 0.19, "learning_rate": 2.345871883418287e-06, "loss": 0.8635, "step": 23562 }, { "epoch": 0.19, "learning_rate": 2.3451782686141768e-06, "loss": 0.8472, "step": 23563 }, { "epoch": 0.19, "learning_rate": 2.344484742745925e-06, "loss": 0.818, "step": 23564 }, { "epoch": 0.19, "learning_rate": 2.34379130582159e-06, "loss": 0.8183, "step": 23565 }, { "epoch": 0.19, "learning_rate": 2.3430979578492295e-06, "loss": 0.8384, "step": 23566 }, { "epoch": 0.19, "learning_rate": 2.3424046988369e-06, "loss": 0.8314, "step": 23567 }, { "epoch": 0.19, "learning_rate": 2.3417115287926507e-06, "loss": 0.814, "step": 23568 }, { "epoch": 0.19, "learning_rate": 2.341018447724538e-06, "loss": 0.8689, "step": 23569 }, { "epoch": 0.19, "learning_rate": 2.340325455640614e-06, "loss": 0.8345, "step": 23570 }, { "epoch": 0.19, "learning_rate": 2.3396325525489315e-06, "loss": 0.8401, "step": 23571 }, { "epoch": 0.19, "learning_rate": 2.3389397384575398e-06, "loss": 0.8772, "step": 23572 }, { "epoch": 0.19, "learning_rate": 2.3382470133744873e-06, "loss": 0.8742, "step": 23573 }, { "epoch": 0.19, "learning_rate": 2.337554377307827e-06, "loss": 0.8167, "step": 23574 }, { "epoch": 0.19, "learning_rate": 2.336861830265599e-06, "loss": 0.8594, "step": 23575 }, { "epoch": 0.19, "learning_rate": 2.336169372255853e-06, "loss": 0.8632, "step": 23576 }, { "epoch": 0.19, "learning_rate": 2.3354770032866313e-06, "loss": 0.851, "step": 23577 }, { "epoch": 0.19, "learning_rate": 2.334784723365987e-06, "loss": 0.8416, "step": 23578 }, { "epoch": 0.19, "learning_rate": 2.3340925325019537e-06, "loss": 0.8897, "step": 23579 }, { "epoch": 0.19, "learning_rate": 2.333400430702577e-06, "loss": 0.8718, "step": 23580 }, { "epoch": 0.19, "learning_rate": 2.3327084179758998e-06, "loss": 0.8811, "step": 23581 }, { "epoch": 0.19, "learning_rate": 2.3320164943299572e-06, "loss": 0.883, "step": 23582 }, { "epoch": 0.19, "learning_rate": 2.331324659772789e-06, "loss": 0.8569, "step": 23583 }, { "epoch": 0.19, "learning_rate": 2.3306329143124374e-06, "loss": 0.8932, "step": 23584 }, { "epoch": 0.19, "learning_rate": 2.32994125795694e-06, "loss": 0.8274, "step": 23585 }, { "epoch": 0.19, "learning_rate": 2.329249690714326e-06, "loss": 0.8663, "step": 23586 }, { "epoch": 0.19, "learning_rate": 2.3285582125926343e-06, "loss": 0.8606, "step": 23587 }, { "epoch": 0.19, "learning_rate": 2.327866823599899e-06, "loss": 0.8957, "step": 23588 }, { "epoch": 0.19, "learning_rate": 2.327175523744151e-06, "loss": 0.9137, "step": 23589 }, { "epoch": 0.19, "learning_rate": 2.3264843130334237e-06, "loss": 0.8617, "step": 23590 }, { "epoch": 0.19, "learning_rate": 2.325793191475747e-06, "loss": 0.8258, "step": 23591 }, { "epoch": 0.19, "learning_rate": 2.3251021590791534e-06, "loss": 0.8619, "step": 23592 }, { "epoch": 0.19, "learning_rate": 2.3244112158516665e-06, "loss": 0.8316, "step": 23593 }, { "epoch": 0.19, "learning_rate": 2.323720361801316e-06, "loss": 0.8529, "step": 23594 }, { "epoch": 0.19, "learning_rate": 2.3230295969361284e-06, "loss": 0.8288, "step": 23595 }, { "epoch": 0.19, "learning_rate": 2.322338921264129e-06, "loss": 0.8643, "step": 23596 }, { "epoch": 0.19, "learning_rate": 2.321648334793344e-06, "loss": 0.8631, "step": 23597 }, { "epoch": 0.19, "learning_rate": 2.320957837531794e-06, "loss": 0.8988, "step": 23598 }, { "epoch": 0.19, "learning_rate": 2.3202674294875072e-06, "loss": 0.8137, "step": 23599 }, { "epoch": 0.19, "learning_rate": 2.3195771106684974e-06, "loss": 0.8519, "step": 23600 }, { "epoch": 0.19, "learning_rate": 2.318886881082787e-06, "loss": 0.8825, "step": 23601 }, { "epoch": 0.19, "learning_rate": 2.3181967407383977e-06, "loss": 0.8312, "step": 23602 }, { "epoch": 0.19, "learning_rate": 2.3175066896433464e-06, "loss": 0.7975, "step": 23603 }, { "epoch": 0.19, "learning_rate": 2.3168167278056497e-06, "loss": 0.853, "step": 23604 }, { "epoch": 0.19, "learning_rate": 2.316126855233325e-06, "loss": 0.8413, "step": 23605 }, { "epoch": 0.19, "learning_rate": 2.3154370719343898e-06, "loss": 0.9007, "step": 23606 }, { "epoch": 0.19, "learning_rate": 2.3147473779168495e-06, "loss": 0.8638, "step": 23607 }, { "epoch": 0.19, "learning_rate": 2.3140577731887257e-06, "loss": 0.8844, "step": 23608 }, { "epoch": 0.19, "learning_rate": 2.3133682577580285e-06, "loss": 0.8551, "step": 23609 }, { "epoch": 0.2, "learning_rate": 2.31267883163277e-06, "loss": 0.8222, "step": 23610 }, { "epoch": 0.2, "learning_rate": 2.311989494820955e-06, "loss": 0.8365, "step": 23611 }, { "epoch": 0.2, "learning_rate": 2.311300247330597e-06, "loss": 0.8664, "step": 23612 }, { "epoch": 0.2, "learning_rate": 2.310611089169702e-06, "loss": 0.8724, "step": 23613 }, { "epoch": 0.2, "learning_rate": 2.3099220203462768e-06, "loss": 0.8809, "step": 23614 }, { "epoch": 0.2, "learning_rate": 2.3092330408683285e-06, "loss": 0.8469, "step": 23615 }, { "epoch": 0.2, "learning_rate": 2.30854415074386e-06, "loss": 0.8456, "step": 23616 }, { "epoch": 0.2, "learning_rate": 2.30785534998088e-06, "loss": 0.8644, "step": 23617 }, { "epoch": 0.2, "learning_rate": 2.3071666385873835e-06, "loss": 0.8073, "step": 23618 }, { "epoch": 0.2, "learning_rate": 2.3064780165713764e-06, "loss": 0.8534, "step": 23619 }, { "epoch": 0.2, "learning_rate": 2.305789483940859e-06, "loss": 0.8762, "step": 23620 }, { "epoch": 0.2, "learning_rate": 2.305101040703831e-06, "loss": 0.8602, "step": 23621 }, { "epoch": 0.2, "learning_rate": 2.3044126868682892e-06, "loss": 0.8465, "step": 23622 }, { "epoch": 0.2, "learning_rate": 2.303724422442234e-06, "loss": 0.8142, "step": 23623 }, { "epoch": 0.2, "learning_rate": 2.303036247433662e-06, "loss": 0.8387, "step": 23624 }, { "epoch": 0.2, "learning_rate": 2.302348161850565e-06, "loss": 0.8231, "step": 23625 }, { "epoch": 0.2, "learning_rate": 2.3016601657009364e-06, "loss": 0.8222, "step": 23626 }, { "epoch": 0.2, "learning_rate": 2.3009722589927785e-06, "loss": 0.8777, "step": 23627 }, { "epoch": 0.2, "learning_rate": 2.3002844417340743e-06, "loss": 0.8654, "step": 23628 }, { "epoch": 0.2, "learning_rate": 2.2995967139328177e-06, "loss": 0.8329, "step": 23629 }, { "epoch": 0.2, "learning_rate": 2.298909075597e-06, "loss": 0.8512, "step": 23630 }, { "epoch": 0.2, "learning_rate": 2.29822152673461e-06, "loss": 0.847, "step": 23631 }, { "epoch": 0.2, "learning_rate": 2.297534067353635e-06, "loss": 0.8474, "step": 23632 }, { "epoch": 0.2, "learning_rate": 2.296846697462064e-06, "loss": 0.885, "step": 23633 }, { "epoch": 0.2, "learning_rate": 2.2961594170678835e-06, "loss": 0.8669, "step": 23634 }, { "epoch": 0.2, "learning_rate": 2.295472226179075e-06, "loss": 0.8298, "step": 23635 }, { "epoch": 0.2, "learning_rate": 2.2947851248036245e-06, "loss": 0.8215, "step": 23636 }, { "epoch": 0.2, "learning_rate": 2.2940981129495143e-06, "loss": 0.8559, "step": 23637 }, { "epoch": 0.2, "learning_rate": 2.293411190624727e-06, "loss": 0.857, "step": 23638 }, { "epoch": 0.2, "learning_rate": 2.2927243578372428e-06, "loss": 0.8451, "step": 23639 }, { "epoch": 0.2, "learning_rate": 2.292037614595043e-06, "loss": 0.829, "step": 23640 }, { "epoch": 0.2, "learning_rate": 2.291350960906107e-06, "loss": 0.9023, "step": 23641 }, { "epoch": 0.2, "learning_rate": 2.290664396778408e-06, "loss": 0.8772, "step": 23642 }, { "epoch": 0.2, "learning_rate": 2.289977922219927e-06, "loss": 0.8869, "step": 23643 }, { "epoch": 0.2, "learning_rate": 2.289291537238637e-06, "loss": 0.8552, "step": 23644 }, { "epoch": 0.2, "learning_rate": 2.2886052418425143e-06, "loss": 0.8659, "step": 23645 }, { "epoch": 0.2, "learning_rate": 2.287919036039532e-06, "loss": 0.8593, "step": 23646 }, { "epoch": 0.2, "learning_rate": 2.287232919837663e-06, "loss": 0.8519, "step": 23647 }, { "epoch": 0.2, "learning_rate": 2.286546893244881e-06, "loss": 0.8751, "step": 23648 }, { "epoch": 0.2, "learning_rate": 2.2858609562691505e-06, "loss": 0.8698, "step": 23649 }, { "epoch": 0.2, "learning_rate": 2.2851751089184417e-06, "loss": 0.8408, "step": 23650 }, { "epoch": 0.2, "learning_rate": 2.284489351200728e-06, "loss": 0.8714, "step": 23651 }, { "epoch": 0.2, "learning_rate": 2.2838036831239785e-06, "loss": 0.8958, "step": 23652 }, { "epoch": 0.2, "learning_rate": 2.2831181046961516e-06, "loss": 0.8646, "step": 23653 }, { "epoch": 0.2, "learning_rate": 2.2824326159252164e-06, "loss": 0.8521, "step": 23654 }, { "epoch": 0.2, "learning_rate": 2.2817472168191358e-06, "loss": 0.8223, "step": 23655 }, { "epoch": 0.2, "learning_rate": 2.2810619073858752e-06, "loss": 0.8509, "step": 23656 }, { "epoch": 0.2, "learning_rate": 2.2803766876333956e-06, "loss": 0.889, "step": 23657 }, { "epoch": 0.2, "learning_rate": 2.2796915575696578e-06, "loss": 0.8609, "step": 23658 }, { "epoch": 0.2, "learning_rate": 2.2790065172026255e-06, "loss": 0.8283, "step": 23659 }, { "epoch": 0.2, "learning_rate": 2.2783215665402503e-06, "loss": 0.8653, "step": 23660 }, { "epoch": 0.2, "learning_rate": 2.2776367055904947e-06, "loss": 0.8153, "step": 23661 }, { "epoch": 0.2, "learning_rate": 2.2769519343613158e-06, "loss": 0.8982, "step": 23662 }, { "epoch": 0.2, "learning_rate": 2.2762672528606676e-06, "loss": 0.8441, "step": 23663 }, { "epoch": 0.2, "learning_rate": 2.275582661096506e-06, "loss": 0.8602, "step": 23664 }, { "epoch": 0.2, "learning_rate": 2.274898159076785e-06, "loss": 0.8535, "step": 23665 }, { "epoch": 0.2, "learning_rate": 2.2742137468094605e-06, "loss": 0.8797, "step": 23666 }, { "epoch": 0.2, "learning_rate": 2.2735294243024775e-06, "loss": 0.8676, "step": 23667 }, { "epoch": 0.2, "learning_rate": 2.2728451915637904e-06, "loss": 0.8736, "step": 23668 }, { "epoch": 0.2, "learning_rate": 2.2721610486013457e-06, "loss": 0.851, "step": 23669 }, { "epoch": 0.2, "learning_rate": 2.2714769954230996e-06, "loss": 0.8582, "step": 23670 }, { "epoch": 0.2, "learning_rate": 2.270793032036992e-06, "loss": 0.8346, "step": 23671 }, { "epoch": 0.2, "learning_rate": 2.2701091584509727e-06, "loss": 0.8461, "step": 23672 }, { "epoch": 0.2, "learning_rate": 2.269425374672989e-06, "loss": 0.8507, "step": 23673 }, { "epoch": 0.2, "learning_rate": 2.2687416807109787e-06, "loss": 0.8375, "step": 23674 }, { "epoch": 0.2, "learning_rate": 2.2680580765728877e-06, "loss": 0.8647, "step": 23675 }, { "epoch": 0.2, "learning_rate": 2.267374562266662e-06, "loss": 0.875, "step": 23676 }, { "epoch": 0.2, "learning_rate": 2.2666911378002443e-06, "loss": 0.8408, "step": 23677 }, { "epoch": 0.2, "learning_rate": 2.266007803181568e-06, "loss": 0.8383, "step": 23678 }, { "epoch": 0.2, "learning_rate": 2.265324558418576e-06, "loss": 0.8877, "step": 23679 }, { "epoch": 0.2, "learning_rate": 2.2646414035192056e-06, "loss": 0.8432, "step": 23680 }, { "epoch": 0.2, "learning_rate": 2.263958338491393e-06, "loss": 0.8578, "step": 23681 }, { "epoch": 0.2, "learning_rate": 2.2632753633430772e-06, "loss": 0.8593, "step": 23682 }, { "epoch": 0.2, "learning_rate": 2.2625924780821908e-06, "loss": 0.8495, "step": 23683 }, { "epoch": 0.2, "learning_rate": 2.261909682716671e-06, "loss": 0.8902, "step": 23684 }, { "epoch": 0.2, "learning_rate": 2.261226977254445e-06, "loss": 0.859, "step": 23685 }, { "epoch": 0.2, "learning_rate": 2.2605443617034474e-06, "loss": 0.8618, "step": 23686 }, { "epoch": 0.2, "learning_rate": 2.2598618360716105e-06, "loss": 0.8097, "step": 23687 }, { "epoch": 0.2, "learning_rate": 2.259179400366862e-06, "loss": 0.8581, "step": 23688 }, { "epoch": 0.2, "learning_rate": 2.2584970545971317e-06, "loss": 0.8162, "step": 23689 }, { "epoch": 0.2, "learning_rate": 2.257814798770348e-06, "loss": 0.8133, "step": 23690 }, { "epoch": 0.2, "learning_rate": 2.2571326328944386e-06, "loss": 0.8562, "step": 23691 }, { "epoch": 0.2, "learning_rate": 2.2564505569773254e-06, "loss": 0.8651, "step": 23692 }, { "epoch": 0.2, "learning_rate": 2.2557685710269307e-06, "loss": 0.8658, "step": 23693 }, { "epoch": 0.2, "learning_rate": 2.255086675051189e-06, "loss": 0.8645, "step": 23694 }, { "epoch": 0.2, "learning_rate": 2.254404869058012e-06, "loss": 0.8311, "step": 23695 }, { "epoch": 0.2, "learning_rate": 2.253723153055325e-06, "loss": 0.8575, "step": 23696 }, { "epoch": 0.2, "learning_rate": 2.253041527051049e-06, "loss": 0.8473, "step": 23697 }, { "epoch": 0.2, "learning_rate": 2.2523599910531048e-06, "loss": 0.8715, "step": 23698 }, { "epoch": 0.2, "learning_rate": 2.2516785450694025e-06, "loss": 0.8515, "step": 23699 }, { "epoch": 0.2, "learning_rate": 2.2509971891078684e-06, "loss": 0.8558, "step": 23700 }, { "epoch": 0.2, "learning_rate": 2.250315923176416e-06, "loss": 0.8616, "step": 23701 }, { "epoch": 0.2, "learning_rate": 2.2496347472829615e-06, "loss": 0.8741, "step": 23702 }, { "epoch": 0.2, "learning_rate": 2.2489536614354155e-06, "loss": 0.8185, "step": 23703 }, { "epoch": 0.2, "learning_rate": 2.2482726656416944e-06, "loss": 0.8719, "step": 23704 }, { "epoch": 0.2, "learning_rate": 2.2475917599097076e-06, "loss": 0.8803, "step": 23705 }, { "epoch": 0.2, "learning_rate": 2.246910944247367e-06, "loss": 0.8777, "step": 23706 }, { "epoch": 0.2, "learning_rate": 2.246230218662584e-06, "loss": 0.8619, "step": 23707 }, { "epoch": 0.2, "learning_rate": 2.2455495831632647e-06, "loss": 0.9052, "step": 23708 }, { "epoch": 0.2, "learning_rate": 2.2448690377573225e-06, "loss": 0.8472, "step": 23709 }, { "epoch": 0.2, "learning_rate": 2.2441885824526564e-06, "loss": 0.8415, "step": 23710 }, { "epoch": 0.2, "learning_rate": 2.2435082172571777e-06, "loss": 0.8069, "step": 23711 }, { "epoch": 0.2, "learning_rate": 2.2428279421787878e-06, "loss": 0.8787, "step": 23712 }, { "epoch": 0.2, "learning_rate": 2.2421477572253912e-06, "loss": 0.8468, "step": 23713 }, { "epoch": 0.2, "learning_rate": 2.2414676624048926e-06, "loss": 0.8745, "step": 23714 }, { "epoch": 0.2, "learning_rate": 2.240787657725191e-06, "loss": 0.884, "step": 23715 }, { "epoch": 0.2, "learning_rate": 2.240107743194191e-06, "loss": 0.8553, "step": 23716 }, { "epoch": 0.2, "learning_rate": 2.2394279188197864e-06, "loss": 0.8477, "step": 23717 }, { "epoch": 0.2, "learning_rate": 2.2387481846098745e-06, "loss": 0.8275, "step": 23718 }, { "epoch": 0.2, "learning_rate": 2.2380685405723622e-06, "loss": 0.8513, "step": 23719 }, { "epoch": 0.2, "learning_rate": 2.2373889867151377e-06, "loss": 0.8735, "step": 23720 }, { "epoch": 0.2, "learning_rate": 2.2367095230460977e-06, "loss": 0.8276, "step": 23721 }, { "epoch": 0.2, "learning_rate": 2.2360301495731375e-06, "loss": 0.8224, "step": 23722 }, { "epoch": 0.2, "learning_rate": 2.235350866304149e-06, "loss": 0.8506, "step": 23723 }, { "epoch": 0.2, "learning_rate": 2.2346716732470266e-06, "loss": 0.844, "step": 23724 }, { "epoch": 0.2, "learning_rate": 2.2339925704096587e-06, "loss": 0.8813, "step": 23725 }, { "epoch": 0.2, "learning_rate": 2.23331355779994e-06, "loss": 0.8934, "step": 23726 }, { "epoch": 0.2, "learning_rate": 2.232634635425752e-06, "loss": 0.8354, "step": 23727 }, { "epoch": 0.2, "learning_rate": 2.231955803294987e-06, "loss": 0.8592, "step": 23728 }, { "epoch": 0.2, "learning_rate": 2.231277061415532e-06, "loss": 0.8814, "step": 23729 }, { "epoch": 0.2, "learning_rate": 2.230598409795273e-06, "loss": 0.8906, "step": 23730 }, { "epoch": 0.2, "learning_rate": 2.2299198484420924e-06, "loss": 0.8368, "step": 23731 }, { "epoch": 0.2, "learning_rate": 2.2292413773638767e-06, "loss": 0.8445, "step": 23732 }, { "epoch": 0.2, "learning_rate": 2.2285629965685096e-06, "loss": 0.8197, "step": 23733 }, { "epoch": 0.2, "learning_rate": 2.2278847060638676e-06, "loss": 0.8482, "step": 23734 }, { "epoch": 0.2, "learning_rate": 2.227206505857834e-06, "loss": 0.8677, "step": 23735 }, { "epoch": 0.2, "learning_rate": 2.226528395958286e-06, "loss": 0.8511, "step": 23736 }, { "epoch": 0.2, "learning_rate": 2.225850376373111e-06, "loss": 0.8811, "step": 23737 }, { "epoch": 0.2, "learning_rate": 2.2251724471101753e-06, "loss": 0.8637, "step": 23738 }, { "epoch": 0.2, "learning_rate": 2.2244946081773623e-06, "loss": 0.8673, "step": 23739 }, { "epoch": 0.2, "learning_rate": 2.223816859582546e-06, "loss": 0.8033, "step": 23740 }, { "epoch": 0.2, "learning_rate": 2.2231392013335973e-06, "loss": 0.8507, "step": 23741 }, { "epoch": 0.2, "learning_rate": 2.2224616334383886e-06, "loss": 0.8213, "step": 23742 }, { "epoch": 0.2, "learning_rate": 2.221784155904799e-06, "loss": 0.8149, "step": 23743 }, { "epoch": 0.2, "learning_rate": 2.2211067687406983e-06, "loss": 0.8658, "step": 23744 }, { "epoch": 0.2, "learning_rate": 2.220429471953951e-06, "loss": 0.8187, "step": 23745 }, { "epoch": 0.2, "learning_rate": 2.2197522655524297e-06, "loss": 0.8419, "step": 23746 }, { "epoch": 0.2, "learning_rate": 2.2190751495440023e-06, "loss": 0.8762, "step": 23747 }, { "epoch": 0.2, "learning_rate": 2.218398123936535e-06, "loss": 0.8465, "step": 23748 }, { "epoch": 0.2, "learning_rate": 2.2177211887378933e-06, "loss": 0.8711, "step": 23749 }, { "epoch": 0.2, "learning_rate": 2.217044343955943e-06, "loss": 0.8508, "step": 23750 }, { "epoch": 0.2, "learning_rate": 2.2163675895985513e-06, "loss": 0.8421, "step": 23751 }, { "epoch": 0.2, "learning_rate": 2.2156909256735735e-06, "loss": 0.8643, "step": 23752 }, { "epoch": 0.2, "learning_rate": 2.2150143521888756e-06, "loss": 0.8825, "step": 23753 }, { "epoch": 0.2, "learning_rate": 2.2143378691523174e-06, "loss": 0.8118, "step": 23754 }, { "epoch": 0.2, "learning_rate": 2.213661476571759e-06, "loss": 0.8668, "step": 23755 }, { "epoch": 0.2, "learning_rate": 2.212985174455059e-06, "loss": 0.8417, "step": 23756 }, { "epoch": 0.2, "learning_rate": 2.2123089628100736e-06, "loss": 0.861, "step": 23757 }, { "epoch": 0.2, "learning_rate": 2.211632841644663e-06, "loss": 0.8444, "step": 23758 }, { "epoch": 0.2, "learning_rate": 2.210956810966678e-06, "loss": 0.8491, "step": 23759 }, { "epoch": 0.2, "learning_rate": 2.2102808707839727e-06, "loss": 0.8819, "step": 23760 }, { "epoch": 0.2, "learning_rate": 2.2096050211044005e-06, "loss": 0.85, "step": 23761 }, { "epoch": 0.2, "learning_rate": 2.2089292619358218e-06, "loss": 0.8855, "step": 23762 }, { "epoch": 0.2, "learning_rate": 2.208253593286077e-06, "loss": 0.8368, "step": 23763 }, { "epoch": 0.2, "learning_rate": 2.2075780151630223e-06, "loss": 0.8397, "step": 23764 }, { "epoch": 0.2, "learning_rate": 2.206902527574506e-06, "loss": 0.8589, "step": 23765 }, { "epoch": 0.2, "learning_rate": 2.20622713052837e-06, "loss": 0.8238, "step": 23766 }, { "epoch": 0.2, "learning_rate": 2.2055518240324704e-06, "loss": 0.8706, "step": 23767 }, { "epoch": 0.2, "learning_rate": 2.204876608094647e-06, "loss": 0.8233, "step": 23768 }, { "epoch": 0.2, "learning_rate": 2.204201482722751e-06, "loss": 0.8776, "step": 23769 }, { "epoch": 0.2, "learning_rate": 2.2035264479246187e-06, "loss": 0.8614, "step": 23770 }, { "epoch": 0.2, "learning_rate": 2.2028515037080955e-06, "loss": 0.8113, "step": 23771 }, { "epoch": 0.2, "learning_rate": 2.2021766500810247e-06, "loss": 0.8809, "step": 23772 }, { "epoch": 0.2, "learning_rate": 2.201501887051245e-06, "loss": 0.8412, "step": 23773 }, { "epoch": 0.2, "learning_rate": 2.2008272146265973e-06, "loss": 0.876, "step": 23774 }, { "epoch": 0.2, "learning_rate": 2.200152632814919e-06, "loss": 0.8219, "step": 23775 }, { "epoch": 0.2, "learning_rate": 2.199478141624052e-06, "loss": 0.8428, "step": 23776 }, { "epoch": 0.2, "learning_rate": 2.1988037410618257e-06, "loss": 0.8209, "step": 23777 }, { "epoch": 0.2, "learning_rate": 2.198129431136078e-06, "loss": 0.83, "step": 23778 }, { "epoch": 0.2, "learning_rate": 2.1974552118546456e-06, "loss": 0.8368, "step": 23779 }, { "epoch": 0.2, "learning_rate": 2.196781083225359e-06, "loss": 0.8542, "step": 23780 }, { "epoch": 0.2, "learning_rate": 2.1961070452560518e-06, "loss": 0.8612, "step": 23781 }, { "epoch": 0.2, "learning_rate": 2.1954330979545548e-06, "loss": 0.8647, "step": 23782 }, { "epoch": 0.2, "learning_rate": 2.1947592413287013e-06, "loss": 0.8605, "step": 23783 }, { "epoch": 0.2, "learning_rate": 2.1940854753863138e-06, "loss": 0.8501, "step": 23784 }, { "epoch": 0.2, "learning_rate": 2.193411800135222e-06, "loss": 0.8625, "step": 23785 }, { "epoch": 0.2, "learning_rate": 2.19273821558326e-06, "loss": 0.8343, "step": 23786 }, { "epoch": 0.2, "learning_rate": 2.192064721738245e-06, "loss": 0.8423, "step": 23787 }, { "epoch": 0.2, "learning_rate": 2.191391318608006e-06, "loss": 0.8362, "step": 23788 }, { "epoch": 0.2, "learning_rate": 2.190718006200365e-06, "loss": 0.8772, "step": 23789 }, { "epoch": 0.2, "learning_rate": 2.19004478452315e-06, "loss": 0.8429, "step": 23790 }, { "epoch": 0.2, "learning_rate": 2.1893716535841712e-06, "loss": 0.8103, "step": 23791 }, { "epoch": 0.2, "learning_rate": 2.18869861339126e-06, "loss": 0.8761, "step": 23792 }, { "epoch": 0.2, "learning_rate": 2.1880256639522324e-06, "loss": 0.8915, "step": 23793 }, { "epoch": 0.2, "learning_rate": 2.1873528052749094e-06, "loss": 0.8195, "step": 23794 }, { "epoch": 0.2, "learning_rate": 2.1866800373671025e-06, "loss": 0.8929, "step": 23795 }, { "epoch": 0.2, "learning_rate": 2.1860073602366317e-06, "loss": 0.8629, "step": 23796 }, { "epoch": 0.2, "learning_rate": 2.185334773891313e-06, "loss": 0.8708, "step": 23797 }, { "epoch": 0.2, "learning_rate": 2.1846622783389595e-06, "loss": 0.8528, "step": 23798 }, { "epoch": 0.2, "learning_rate": 2.1839898735873853e-06, "loss": 0.8602, "step": 23799 }, { "epoch": 0.2, "learning_rate": 2.1833175596444013e-06, "loss": 0.8624, "step": 23800 }, { "epoch": 0.2, "learning_rate": 2.1826453365178224e-06, "loss": 0.8571, "step": 23801 }, { "epoch": 0.2, "learning_rate": 2.181973204215453e-06, "loss": 0.8498, "step": 23802 }, { "epoch": 0.2, "learning_rate": 2.181301162745105e-06, "loss": 0.8309, "step": 23803 }, { "epoch": 0.2, "learning_rate": 2.1806292121145866e-06, "loss": 0.8938, "step": 23804 }, { "epoch": 0.2, "learning_rate": 2.1799573523317042e-06, "loss": 0.8428, "step": 23805 }, { "epoch": 0.2, "learning_rate": 2.1792855834042638e-06, "loss": 0.8609, "step": 23806 }, { "epoch": 0.2, "learning_rate": 2.178613905340069e-06, "loss": 0.8154, "step": 23807 }, { "epoch": 0.2, "learning_rate": 2.1779423181469284e-06, "loss": 0.8519, "step": 23808 }, { "epoch": 0.2, "learning_rate": 2.1772708218326353e-06, "loss": 0.8647, "step": 23809 }, { "epoch": 0.2, "learning_rate": 2.1765994164049995e-06, "loss": 0.8609, "step": 23810 }, { "epoch": 0.2, "learning_rate": 2.1759281018718226e-06, "loss": 0.8597, "step": 23811 }, { "epoch": 0.2, "learning_rate": 2.1752568782408977e-06, "loss": 0.8624, "step": 23812 }, { "epoch": 0.2, "learning_rate": 2.174585745520026e-06, "loss": 0.8588, "step": 23813 }, { "epoch": 0.2, "learning_rate": 2.173914703717005e-06, "loss": 0.857, "step": 23814 }, { "epoch": 0.2, "learning_rate": 2.173243752839632e-06, "loss": 0.8683, "step": 23815 }, { "epoch": 0.2, "learning_rate": 2.1725728928957e-06, "loss": 0.8029, "step": 23816 }, { "epoch": 0.2, "learning_rate": 2.171902123893006e-06, "loss": 0.8774, "step": 23817 }, { "epoch": 0.2, "learning_rate": 2.1712314458393437e-06, "loss": 0.8513, "step": 23818 }, { "epoch": 0.2, "learning_rate": 2.1705608587425007e-06, "loss": 0.8378, "step": 23819 }, { "epoch": 0.2, "learning_rate": 2.1698903626102717e-06, "loss": 0.8203, "step": 23820 }, { "epoch": 0.2, "learning_rate": 2.169219957450446e-06, "loss": 0.8198, "step": 23821 }, { "epoch": 0.2, "learning_rate": 2.1685496432708107e-06, "loss": 0.8271, "step": 23822 }, { "epoch": 0.2, "learning_rate": 2.1678794200791564e-06, "loss": 0.8157, "step": 23823 }, { "epoch": 0.2, "learning_rate": 2.1672092878832683e-06, "loss": 0.8798, "step": 23824 }, { "epoch": 0.2, "learning_rate": 2.1665392466909364e-06, "loss": 0.8516, "step": 23825 }, { "epoch": 0.2, "learning_rate": 2.1658692965099383e-06, "loss": 0.8785, "step": 23826 }, { "epoch": 0.2, "learning_rate": 2.165199437348061e-06, "loss": 0.8763, "step": 23827 }, { "epoch": 0.2, "learning_rate": 2.1645296692130844e-06, "loss": 0.8923, "step": 23828 }, { "epoch": 0.2, "learning_rate": 2.163859992112799e-06, "loss": 0.8435, "step": 23829 }, { "epoch": 0.2, "learning_rate": 2.163190406054976e-06, "loss": 0.8812, "step": 23830 }, { "epoch": 0.2, "learning_rate": 2.162520911047399e-06, "loss": 0.8811, "step": 23831 }, { "epoch": 0.2, "learning_rate": 2.161851507097844e-06, "loss": 0.8351, "step": 23832 }, { "epoch": 0.2, "learning_rate": 2.1611821942140932e-06, "loss": 0.8632, "step": 23833 }, { "epoch": 0.2, "learning_rate": 2.1605129724039144e-06, "loss": 0.8279, "step": 23834 }, { "epoch": 0.2, "learning_rate": 2.15984384167509e-06, "loss": 0.8329, "step": 23835 }, { "epoch": 0.2, "learning_rate": 2.159174802035394e-06, "loss": 0.8475, "step": 23836 }, { "epoch": 0.2, "learning_rate": 2.1585058534925964e-06, "loss": 0.8628, "step": 23837 }, { "epoch": 0.2, "learning_rate": 2.1578369960544687e-06, "loss": 0.8599, "step": 23838 }, { "epoch": 0.2, "learning_rate": 2.1571682297287843e-06, "loss": 0.8328, "step": 23839 }, { "epoch": 0.2, "learning_rate": 2.1564995545233125e-06, "loss": 0.8635, "step": 23840 }, { "epoch": 0.2, "learning_rate": 2.1558309704458215e-06, "loss": 0.8477, "step": 23841 }, { "epoch": 0.2, "learning_rate": 2.1551624775040793e-06, "loss": 0.8713, "step": 23842 }, { "epoch": 0.2, "learning_rate": 2.1544940757058564e-06, "loss": 0.8335, "step": 23843 }, { "epoch": 0.2, "learning_rate": 2.1538257650589114e-06, "loss": 0.8585, "step": 23844 }, { "epoch": 0.2, "learning_rate": 2.153157545571013e-06, "loss": 0.8981, "step": 23845 }, { "epoch": 0.2, "learning_rate": 2.152489417249923e-06, "loss": 0.8903, "step": 23846 }, { "epoch": 0.2, "learning_rate": 2.151821380103406e-06, "loss": 0.8649, "step": 23847 }, { "epoch": 0.2, "learning_rate": 2.1511534341392227e-06, "loss": 0.8895, "step": 23848 }, { "epoch": 0.2, "learning_rate": 2.1504855793651338e-06, "loss": 0.8466, "step": 23849 }, { "epoch": 0.2, "learning_rate": 2.1498178157889003e-06, "loss": 0.8108, "step": 23850 }, { "epoch": 0.2, "learning_rate": 2.149150143418276e-06, "loss": 0.8331, "step": 23851 }, { "epoch": 0.2, "learning_rate": 2.1484825622610195e-06, "loss": 0.8418, "step": 23852 }, { "epoch": 0.2, "learning_rate": 2.147815072324886e-06, "loss": 0.8805, "step": 23853 }, { "epoch": 0.2, "learning_rate": 2.1471476736176376e-06, "loss": 0.8987, "step": 23854 }, { "epoch": 0.2, "learning_rate": 2.1464803661470215e-06, "loss": 0.819, "step": 23855 }, { "epoch": 0.2, "learning_rate": 2.1458131499207913e-06, "loss": 0.8773, "step": 23856 }, { "epoch": 0.2, "learning_rate": 2.1451460249467026e-06, "loss": 0.8369, "step": 23857 }, { "epoch": 0.2, "learning_rate": 2.144478991232498e-06, "loss": 0.8644, "step": 23858 }, { "epoch": 0.2, "learning_rate": 2.143812048785936e-06, "loss": 0.8589, "step": 23859 }, { "epoch": 0.2, "learning_rate": 2.1431451976147623e-06, "loss": 0.8172, "step": 23860 }, { "epoch": 0.2, "learning_rate": 2.142478437726727e-06, "loss": 0.8453, "step": 23861 }, { "epoch": 0.2, "learning_rate": 2.1418117691295713e-06, "loss": 0.8417, "step": 23862 }, { "epoch": 0.2, "learning_rate": 2.1411451918310446e-06, "loss": 0.8657, "step": 23863 }, { "epoch": 0.2, "learning_rate": 2.14047870583889e-06, "loss": 0.8728, "step": 23864 }, { "epoch": 0.2, "learning_rate": 2.1398123111608504e-06, "loss": 0.8441, "step": 23865 }, { "epoch": 0.2, "learning_rate": 2.13914600780467e-06, "loss": 0.8747, "step": 23866 }, { "epoch": 0.2, "learning_rate": 2.138479795778089e-06, "loss": 0.881, "step": 23867 }, { "epoch": 0.2, "learning_rate": 2.13781367508885e-06, "loss": 0.8542, "step": 23868 }, { "epoch": 0.2, "learning_rate": 2.1371476457446883e-06, "loss": 0.8548, "step": 23869 }, { "epoch": 0.2, "learning_rate": 2.1364817077533427e-06, "loss": 0.8394, "step": 23870 }, { "epoch": 0.2, "learning_rate": 2.1358158611225523e-06, "loss": 0.8304, "step": 23871 }, { "epoch": 0.2, "learning_rate": 2.135150105860051e-06, "loss": 0.8329, "step": 23872 }, { "epoch": 0.2, "learning_rate": 2.1344844419735757e-06, "loss": 0.8529, "step": 23873 }, { "epoch": 0.2, "learning_rate": 2.1338188694708584e-06, "loss": 0.8612, "step": 23874 }, { "epoch": 0.2, "learning_rate": 2.1331533883596367e-06, "loss": 0.7943, "step": 23875 }, { "epoch": 0.2, "learning_rate": 2.1324879986476352e-06, "loss": 0.8636, "step": 23876 }, { "epoch": 0.2, "learning_rate": 2.1318227003425853e-06, "loss": 0.8625, "step": 23877 }, { "epoch": 0.2, "learning_rate": 2.1311574934522215e-06, "loss": 0.8567, "step": 23878 }, { "epoch": 0.2, "learning_rate": 2.1304923779842734e-06, "loss": 0.8209, "step": 23879 }, { "epoch": 0.2, "learning_rate": 2.129827353946462e-06, "loss": 0.8701, "step": 23880 }, { "epoch": 0.2, "learning_rate": 2.1291624213465168e-06, "loss": 0.8654, "step": 23881 }, { "epoch": 0.2, "learning_rate": 2.128497580192167e-06, "loss": 0.8664, "step": 23882 }, { "epoch": 0.2, "learning_rate": 2.1278328304911267e-06, "loss": 0.8237, "step": 23883 }, { "epoch": 0.2, "learning_rate": 2.1271681722511282e-06, "loss": 0.8943, "step": 23884 }, { "epoch": 0.2, "learning_rate": 2.1265036054798926e-06, "loss": 0.8704, "step": 23885 }, { "epoch": 0.2, "learning_rate": 2.125839130185141e-06, "loss": 0.8508, "step": 23886 }, { "epoch": 0.2, "learning_rate": 2.12517474637459e-06, "loss": 0.8701, "step": 23887 }, { "epoch": 0.2, "learning_rate": 2.1245104540559603e-06, "loss": 0.8426, "step": 23888 }, { "epoch": 0.2, "learning_rate": 2.1238462532369696e-06, "loss": 0.8661, "step": 23889 }, { "epoch": 0.2, "learning_rate": 2.1231821439253365e-06, "loss": 0.8389, "step": 23890 }, { "epoch": 0.2, "learning_rate": 2.1225181261287743e-06, "loss": 0.8097, "step": 23891 }, { "epoch": 0.2, "learning_rate": 2.121854199855e-06, "loss": 0.8639, "step": 23892 }, { "epoch": 0.2, "learning_rate": 2.121190365111728e-06, "loss": 0.8162, "step": 23893 }, { "epoch": 0.2, "learning_rate": 2.1205266219066676e-06, "loss": 0.841, "step": 23894 }, { "epoch": 0.2, "learning_rate": 2.1198629702475314e-06, "loss": 0.8567, "step": 23895 }, { "epoch": 0.2, "learning_rate": 2.119199410142031e-06, "loss": 0.8746, "step": 23896 }, { "epoch": 0.2, "learning_rate": 2.1185359415978756e-06, "loss": 0.8417, "step": 23897 }, { "epoch": 0.2, "learning_rate": 2.1178725646227726e-06, "loss": 0.8605, "step": 23898 }, { "epoch": 0.2, "learning_rate": 2.1172092792244304e-06, "loss": 0.8487, "step": 23899 }, { "epoch": 0.2, "learning_rate": 2.1165460854105566e-06, "loss": 0.8615, "step": 23900 }, { "epoch": 0.2, "learning_rate": 2.1158829831888504e-06, "loss": 0.852, "step": 23901 }, { "epoch": 0.2, "learning_rate": 2.115219972567023e-06, "loss": 0.844, "step": 23902 }, { "epoch": 0.2, "learning_rate": 2.114557053552776e-06, "loss": 0.8653, "step": 23903 }, { "epoch": 0.2, "learning_rate": 2.1138942261538074e-06, "loss": 0.8348, "step": 23904 }, { "epoch": 0.2, "learning_rate": 2.1132314903778216e-06, "loss": 0.8282, "step": 23905 }, { "epoch": 0.2, "learning_rate": 2.1125688462325167e-06, "loss": 0.8425, "step": 23906 }, { "epoch": 0.2, "learning_rate": 2.111906293725593e-06, "loss": 0.8434, "step": 23907 }, { "epoch": 0.2, "learning_rate": 2.1112438328647465e-06, "loss": 0.8545, "step": 23908 }, { "epoch": 0.2, "learning_rate": 2.110581463657676e-06, "loss": 0.8644, "step": 23909 }, { "epoch": 0.21, "learning_rate": 2.1099191861120784e-06, "loss": 0.8647, "step": 23910 }, { "epoch": 0.21, "learning_rate": 2.1092570002356426e-06, "loss": 0.8499, "step": 23911 }, { "epoch": 0.21, "learning_rate": 2.1085949060360654e-06, "loss": 0.8619, "step": 23912 }, { "epoch": 0.21, "learning_rate": 2.1079329035210394e-06, "loss": 0.8674, "step": 23913 }, { "epoch": 0.21, "learning_rate": 2.107270992698255e-06, "loss": 0.8789, "step": 23914 }, { "epoch": 0.21, "learning_rate": 2.106609173575402e-06, "loss": 0.8321, "step": 23915 }, { "epoch": 0.21, "learning_rate": 2.1059474461601726e-06, "loss": 0.8734, "step": 23916 }, { "epoch": 0.21, "learning_rate": 2.105285810460255e-06, "loss": 0.8532, "step": 23917 }, { "epoch": 0.21, "learning_rate": 2.1046242664833317e-06, "loss": 0.8768, "step": 23918 }, { "epoch": 0.21, "learning_rate": 2.1039628142370905e-06, "loss": 0.8058, "step": 23919 }, { "epoch": 0.21, "learning_rate": 2.1033014537292152e-06, "loss": 0.8253, "step": 23920 }, { "epoch": 0.21, "learning_rate": 2.102640184967396e-06, "loss": 0.8479, "step": 23921 }, { "epoch": 0.21, "learning_rate": 2.1019790079593094e-06, "loss": 0.8735, "step": 23922 }, { "epoch": 0.21, "learning_rate": 2.101317922712639e-06, "loss": 0.8387, "step": 23923 }, { "epoch": 0.21, "learning_rate": 2.1006569292350643e-06, "loss": 0.8392, "step": 23924 }, { "epoch": 0.21, "learning_rate": 2.09999602753427e-06, "loss": 0.9254, "step": 23925 }, { "epoch": 0.21, "learning_rate": 2.099335217617925e-06, "loss": 0.8752, "step": 23926 }, { "epoch": 0.21, "learning_rate": 2.0986744994937146e-06, "loss": 0.8373, "step": 23927 }, { "epoch": 0.21, "learning_rate": 2.0980138731693167e-06, "loss": 0.8302, "step": 23928 }, { "epoch": 0.21, "learning_rate": 2.0973533386524004e-06, "loss": 0.864, "step": 23929 }, { "epoch": 0.21, "learning_rate": 2.0966928959506415e-06, "loss": 0.8208, "step": 23930 }, { "epoch": 0.21, "learning_rate": 2.0960325450717146e-06, "loss": 0.8248, "step": 23931 }, { "epoch": 0.21, "learning_rate": 2.0953722860232916e-06, "loss": 0.8735, "step": 23932 }, { "epoch": 0.21, "learning_rate": 2.0947121188130446e-06, "loss": 0.8986, "step": 23933 }, { "epoch": 0.21, "learning_rate": 2.094052043448641e-06, "loss": 0.8659, "step": 23934 }, { "epoch": 0.21, "learning_rate": 2.0933920599377555e-06, "loss": 0.86, "step": 23935 }, { "epoch": 0.21, "learning_rate": 2.0927321682880476e-06, "loss": 0.8562, "step": 23936 }, { "epoch": 0.21, "learning_rate": 2.0920723685071885e-06, "loss": 0.8485, "step": 23937 }, { "epoch": 0.21, "learning_rate": 2.091412660602845e-06, "loss": 0.8624, "step": 23938 }, { "epoch": 0.21, "learning_rate": 2.0907530445826798e-06, "loss": 0.8035, "step": 23939 }, { "epoch": 0.21, "learning_rate": 2.090093520454356e-06, "loss": 0.8113, "step": 23940 }, { "epoch": 0.21, "learning_rate": 2.089434088225538e-06, "loss": 0.8361, "step": 23941 }, { "epoch": 0.21, "learning_rate": 2.08877474790389e-06, "loss": 0.8599, "step": 23942 }, { "epoch": 0.21, "learning_rate": 2.0881154994970656e-06, "loss": 0.9016, "step": 23943 }, { "epoch": 0.21, "learning_rate": 2.0874563430127246e-06, "loss": 0.874, "step": 23944 }, { "epoch": 0.21, "learning_rate": 2.0867972784585313e-06, "loss": 0.849, "step": 23945 }, { "epoch": 0.21, "learning_rate": 2.0861383058421423e-06, "loss": 0.8598, "step": 23946 }, { "epoch": 0.21, "learning_rate": 2.0854794251712086e-06, "loss": 0.8463, "step": 23947 }, { "epoch": 0.21, "learning_rate": 2.0848206364533873e-06, "loss": 0.8532, "step": 23948 }, { "epoch": 0.21, "learning_rate": 2.0841619396963365e-06, "loss": 0.8388, "step": 23949 }, { "epoch": 0.21, "learning_rate": 2.083503334907699e-06, "loss": 0.8405, "step": 23950 }, { "epoch": 0.21, "learning_rate": 2.082844822095137e-06, "loss": 0.8692, "step": 23951 }, { "epoch": 0.21, "learning_rate": 2.0821864012662973e-06, "loss": 0.8232, "step": 23952 }, { "epoch": 0.21, "learning_rate": 2.081528072428832e-06, "loss": 0.8285, "step": 23953 }, { "epoch": 0.21, "learning_rate": 2.0808698355903856e-06, "loss": 0.8813, "step": 23954 }, { "epoch": 0.21, "learning_rate": 2.0802116907586076e-06, "loss": 0.8669, "step": 23955 }, { "epoch": 0.21, "learning_rate": 2.079553637941144e-06, "loss": 0.8749, "step": 23956 }, { "epoch": 0.21, "learning_rate": 2.0788956771456414e-06, "loss": 0.8662, "step": 23957 }, { "epoch": 0.21, "learning_rate": 2.0782378083797437e-06, "loss": 0.8513, "step": 23958 }, { "epoch": 0.21, "learning_rate": 2.077580031651093e-06, "loss": 0.834, "step": 23959 }, { "epoch": 0.21, "learning_rate": 2.076922346967336e-06, "loss": 0.8405, "step": 23960 }, { "epoch": 0.21, "learning_rate": 2.076264754336108e-06, "loss": 0.8563, "step": 23961 }, { "epoch": 0.21, "learning_rate": 2.0756072537650508e-06, "loss": 0.8855, "step": 23962 }, { "epoch": 0.21, "learning_rate": 2.0749498452618057e-06, "loss": 0.8662, "step": 23963 }, { "epoch": 0.21, "learning_rate": 2.0742925288340075e-06, "loss": 0.859, "step": 23964 }, { "epoch": 0.21, "learning_rate": 2.0736353044892955e-06, "loss": 0.8469, "step": 23965 }, { "epoch": 0.21, "learning_rate": 2.072978172235305e-06, "loss": 0.8426, "step": 23966 }, { "epoch": 0.21, "learning_rate": 2.072321132079672e-06, "loss": 0.8679, "step": 23967 }, { "epoch": 0.21, "learning_rate": 2.0716641840300266e-06, "loss": 0.8679, "step": 23968 }, { "epoch": 0.21, "learning_rate": 2.0710073280940012e-06, "loss": 0.8642, "step": 23969 }, { "epoch": 0.21, "learning_rate": 2.070350564279232e-06, "loss": 0.932, "step": 23970 }, { "epoch": 0.21, "learning_rate": 2.0696938925933505e-06, "loss": 0.8557, "step": 23971 }, { "epoch": 0.21, "learning_rate": 2.069037313043979e-06, "loss": 0.8801, "step": 23972 }, { "epoch": 0.21, "learning_rate": 2.0683808256387507e-06, "loss": 0.9021, "step": 23973 }, { "epoch": 0.21, "learning_rate": 2.0677244303852907e-06, "loss": 0.9063, "step": 23974 }, { "epoch": 0.21, "learning_rate": 2.0670681272912254e-06, "loss": 0.8792, "step": 23975 }, { "epoch": 0.21, "learning_rate": 2.0664119163641817e-06, "loss": 0.8461, "step": 23976 }, { "epoch": 0.21, "learning_rate": 2.0657557976117825e-06, "loss": 0.8663, "step": 23977 }, { "epoch": 0.21, "learning_rate": 2.0650997710416533e-06, "loss": 0.8558, "step": 23978 }, { "epoch": 0.21, "learning_rate": 2.064443836661412e-06, "loss": 0.8546, "step": 23979 }, { "epoch": 0.21, "learning_rate": 2.06378799447868e-06, "loss": 0.8547, "step": 23980 }, { "epoch": 0.21, "learning_rate": 2.063132244501077e-06, "loss": 0.8852, "step": 23981 }, { "epoch": 0.21, "learning_rate": 2.062476586736224e-06, "loss": 0.8514, "step": 23982 }, { "epoch": 0.21, "learning_rate": 2.0618210211917376e-06, "loss": 0.8534, "step": 23983 }, { "epoch": 0.21, "learning_rate": 2.0611655478752336e-06, "loss": 0.8259, "step": 23984 }, { "epoch": 0.21, "learning_rate": 2.0605101667943304e-06, "loss": 0.8596, "step": 23985 }, { "epoch": 0.21, "learning_rate": 2.0598548779566387e-06, "loss": 0.8646, "step": 23986 }, { "epoch": 0.21, "learning_rate": 2.059199681369771e-06, "loss": 0.8563, "step": 23987 }, { "epoch": 0.21, "learning_rate": 2.0585445770413425e-06, "loss": 0.8595, "step": 23988 }, { "epoch": 0.21, "learning_rate": 2.0578895649789643e-06, "loss": 0.827, "step": 23989 }, { "epoch": 0.21, "learning_rate": 2.0572346451902446e-06, "loss": 0.8497, "step": 23990 }, { "epoch": 0.21, "learning_rate": 2.0565798176827944e-06, "loss": 0.8025, "step": 23991 }, { "epoch": 0.21, "learning_rate": 2.0559250824642243e-06, "loss": 0.8742, "step": 23992 }, { "epoch": 0.21, "learning_rate": 2.055270439542131e-06, "loss": 0.8487, "step": 23993 }, { "epoch": 0.21, "learning_rate": 2.054615888924131e-06, "loss": 0.8441, "step": 23994 }, { "epoch": 0.21, "learning_rate": 2.053961430617827e-06, "loss": 0.8403, "step": 23995 }, { "epoch": 0.21, "learning_rate": 2.0533070646308173e-06, "loss": 0.893, "step": 23996 }, { "epoch": 0.21, "learning_rate": 2.0526527909707094e-06, "loss": 0.8425, "step": 23997 }, { "epoch": 0.21, "learning_rate": 2.0519986096451028e-06, "loss": 0.8604, "step": 23998 }, { "epoch": 0.21, "learning_rate": 2.051344520661599e-06, "loss": 0.8194, "step": 23999 }, { "epoch": 0.21, "learning_rate": 2.0506905240277963e-06, "loss": 0.8561, "step": 24000 }, { "epoch": 0.21, "learning_rate": 2.0500366197512943e-06, "loss": 0.8396, "step": 24001 }, { "epoch": 0.21, "learning_rate": 2.0493828078396928e-06, "loss": 0.8587, "step": 24002 }, { "epoch": 0.21, "learning_rate": 2.0487290883005815e-06, "loss": 0.8252, "step": 24003 }, { "epoch": 0.21, "learning_rate": 2.048075461141559e-06, "loss": 0.8595, "step": 24004 }, { "epoch": 0.21, "learning_rate": 2.047421926370219e-06, "loss": 0.8471, "step": 24005 }, { "epoch": 0.21, "learning_rate": 2.0467684839941547e-06, "loss": 0.8608, "step": 24006 }, { "epoch": 0.21, "learning_rate": 2.0461151340209584e-06, "loss": 0.7918, "step": 24007 }, { "epoch": 0.21, "learning_rate": 2.04546187645822e-06, "loss": 0.8847, "step": 24008 }, { "epoch": 0.21, "learning_rate": 2.0448087113135316e-06, "loss": 0.882, "step": 24009 }, { "epoch": 0.21, "learning_rate": 2.0441556385944783e-06, "loss": 0.8858, "step": 24010 }, { "epoch": 0.21, "learning_rate": 2.0435026583086483e-06, "loss": 0.8592, "step": 24011 }, { "epoch": 0.21, "learning_rate": 2.0428497704636274e-06, "loss": 0.8504, "step": 24012 }, { "epoch": 0.21, "learning_rate": 2.0421969750670078e-06, "loss": 0.8547, "step": 24013 }, { "epoch": 0.21, "learning_rate": 2.041544272126366e-06, "loss": 0.8704, "step": 24014 }, { "epoch": 0.21, "learning_rate": 2.0408916616492867e-06, "loss": 0.8738, "step": 24015 }, { "epoch": 0.21, "learning_rate": 2.0402391436433556e-06, "loss": 0.8775, "step": 24016 }, { "epoch": 0.21, "learning_rate": 2.03958671811615e-06, "loss": 0.8501, "step": 24017 }, { "epoch": 0.21, "learning_rate": 2.038934385075253e-06, "loss": 0.8803, "step": 24018 }, { "epoch": 0.21, "learning_rate": 2.0382821445282407e-06, "loss": 0.8643, "step": 24019 }, { "epoch": 0.21, "learning_rate": 2.0376299964826963e-06, "loss": 0.8614, "step": 24020 }, { "epoch": 0.21, "learning_rate": 2.0369779409461898e-06, "loss": 0.8649, "step": 24021 }, { "epoch": 0.21, "learning_rate": 2.0363259779263012e-06, "loss": 0.8472, "step": 24022 }, { "epoch": 0.21, "learning_rate": 2.0356741074306032e-06, "loss": 0.8422, "step": 24023 }, { "epoch": 0.21, "learning_rate": 2.0350223294666694e-06, "loss": 0.8772, "step": 24024 }, { "epoch": 0.21, "learning_rate": 2.034370644042074e-06, "loss": 0.8573, "step": 24025 }, { "epoch": 0.21, "learning_rate": 2.0337190511643878e-06, "loss": 0.8331, "step": 24026 }, { "epoch": 0.21, "learning_rate": 2.0330675508411836e-06, "loss": 0.8389, "step": 24027 }, { "epoch": 0.21, "learning_rate": 2.0324161430800247e-06, "loss": 0.8687, "step": 24028 }, { "epoch": 0.21, "learning_rate": 2.031764827888484e-06, "loss": 0.8607, "step": 24029 }, { "epoch": 0.21, "learning_rate": 2.0311136052741274e-06, "loss": 0.8573, "step": 24030 }, { "epoch": 0.21, "learning_rate": 2.03046247524452e-06, "loss": 0.8766, "step": 24031 }, { "epoch": 0.21, "learning_rate": 2.0298114378072287e-06, "loss": 0.8632, "step": 24032 }, { "epoch": 0.21, "learning_rate": 2.029160492969816e-06, "loss": 0.8598, "step": 24033 }, { "epoch": 0.21, "learning_rate": 2.0285096407398485e-06, "loss": 0.8428, "step": 24034 }, { "epoch": 0.21, "learning_rate": 2.0278588811248822e-06, "loss": 0.858, "step": 24035 }, { "epoch": 0.21, "learning_rate": 2.027208214132478e-06, "loss": 0.8918, "step": 24036 }, { "epoch": 0.21, "learning_rate": 2.0265576397702003e-06, "loss": 0.8638, "step": 24037 }, { "epoch": 0.21, "learning_rate": 2.025907158045608e-06, "loss": 0.8668, "step": 24038 }, { "epoch": 0.21, "learning_rate": 2.0252567689662526e-06, "loss": 0.8433, "step": 24039 }, { "epoch": 0.21, "learning_rate": 2.0246064725396953e-06, "loss": 0.8143, "step": 24040 }, { "epoch": 0.21, "learning_rate": 2.0239562687734916e-06, "loss": 0.8309, "step": 24041 }, { "epoch": 0.21, "learning_rate": 2.0233061576751877e-06, "loss": 0.8534, "step": 24042 }, { "epoch": 0.21, "learning_rate": 2.022656139252348e-06, "loss": 0.8463, "step": 24043 }, { "epoch": 0.21, "learning_rate": 2.0220062135125183e-06, "loss": 0.8418, "step": 24044 }, { "epoch": 0.21, "learning_rate": 2.021356380463254e-06, "loss": 0.9105, "step": 24045 }, { "epoch": 0.21, "learning_rate": 2.0207066401121e-06, "loss": 0.8563, "step": 24046 }, { "epoch": 0.21, "learning_rate": 2.020056992466607e-06, "loss": 0.8592, "step": 24047 }, { "epoch": 0.21, "learning_rate": 2.019407437534322e-06, "loss": 0.8221, "step": 24048 }, { "epoch": 0.21, "learning_rate": 2.018757975322794e-06, "loss": 0.8915, "step": 24049 }, { "epoch": 0.21, "learning_rate": 2.018108605839567e-06, "loss": 0.8158, "step": 24050 }, { "epoch": 0.21, "learning_rate": 2.0174593290921863e-06, "loss": 0.8823, "step": 24051 }, { "epoch": 0.21, "learning_rate": 2.016810145088197e-06, "loss": 0.8634, "step": 24052 }, { "epoch": 0.21, "learning_rate": 2.0161610538351384e-06, "loss": 0.8257, "step": 24053 }, { "epoch": 0.21, "learning_rate": 2.0155120553405515e-06, "loss": 0.8809, "step": 24054 }, { "epoch": 0.21, "learning_rate": 2.0148631496119788e-06, "loss": 0.8584, "step": 24055 }, { "epoch": 0.21, "learning_rate": 2.014214336656958e-06, "loss": 0.8543, "step": 24056 }, { "epoch": 0.21, "learning_rate": 2.0135656164830276e-06, "loss": 0.8539, "step": 24057 }, { "epoch": 0.21, "learning_rate": 2.0129169890977264e-06, "loss": 0.8468, "step": 24058 }, { "epoch": 0.21, "learning_rate": 2.0122684545085902e-06, "loss": 0.8355, "step": 24059 }, { "epoch": 0.21, "learning_rate": 2.01162001272315e-06, "loss": 0.8467, "step": 24060 }, { "epoch": 0.21, "learning_rate": 2.0109716637489394e-06, "loss": 0.8309, "step": 24061 }, { "epoch": 0.21, "learning_rate": 2.0103234075934964e-06, "loss": 0.8479, "step": 24062 }, { "epoch": 0.21, "learning_rate": 2.009675244264353e-06, "loss": 0.8384, "step": 24063 }, { "epoch": 0.21, "learning_rate": 2.009027173769034e-06, "loss": 0.8205, "step": 24064 }, { "epoch": 0.21, "learning_rate": 2.008379196115071e-06, "loss": 0.7972, "step": 24065 }, { "epoch": 0.21, "learning_rate": 2.0077313113099938e-06, "loss": 0.8678, "step": 24066 }, { "epoch": 0.21, "learning_rate": 2.007083519361328e-06, "loss": 0.8371, "step": 24067 }, { "epoch": 0.21, "learning_rate": 2.0064358202766023e-06, "loss": 0.8704, "step": 24068 }, { "epoch": 0.21, "learning_rate": 2.0057882140633387e-06, "loss": 0.8459, "step": 24069 }, { "epoch": 0.21, "learning_rate": 2.0051407007290657e-06, "loss": 0.8203, "step": 24070 }, { "epoch": 0.21, "learning_rate": 2.004493280281301e-06, "loss": 0.8387, "step": 24071 }, { "epoch": 0.21, "learning_rate": 2.0038459527275687e-06, "loss": 0.845, "step": 24072 }, { "epoch": 0.21, "learning_rate": 2.0031987180753897e-06, "loss": 0.8661, "step": 24073 }, { "epoch": 0.21, "learning_rate": 2.0025515763322845e-06, "loss": 0.8525, "step": 24074 }, { "epoch": 0.21, "learning_rate": 2.0019045275057713e-06, "loss": 0.84, "step": 24075 }, { "epoch": 0.21, "learning_rate": 2.0012575716033676e-06, "loss": 0.8675, "step": 24076 }, { "epoch": 0.21, "learning_rate": 2.000610708632592e-06, "loss": 0.8485, "step": 24077 }, { "epoch": 0.21, "learning_rate": 1.9999639386009552e-06, "loss": 0.8564, "step": 24078 }, { "epoch": 0.21, "learning_rate": 1.9993172615159717e-06, "loss": 0.844, "step": 24079 }, { "epoch": 0.21, "learning_rate": 1.9986706773851627e-06, "loss": 0.8352, "step": 24080 }, { "epoch": 0.21, "learning_rate": 1.998024186216032e-06, "loss": 0.8654, "step": 24081 }, { "epoch": 0.21, "learning_rate": 1.9973777880160937e-06, "loss": 0.8142, "step": 24082 }, { "epoch": 0.21, "learning_rate": 1.9967314827928575e-06, "loss": 0.8698, "step": 24083 }, { "epoch": 0.21, "learning_rate": 1.996085270553836e-06, "loss": 0.8479, "step": 24084 }, { "epoch": 0.21, "learning_rate": 1.9954391513065288e-06, "loss": 0.8807, "step": 24085 }, { "epoch": 0.21, "learning_rate": 1.9947931250584485e-06, "loss": 0.8785, "step": 24086 }, { "epoch": 0.21, "learning_rate": 1.994147191817104e-06, "loss": 0.8772, "step": 24087 }, { "epoch": 0.21, "learning_rate": 1.9935013515899927e-06, "loss": 0.8626, "step": 24088 }, { "epoch": 0.21, "learning_rate": 1.9928556043846215e-06, "loss": 0.8747, "step": 24089 }, { "epoch": 0.21, "learning_rate": 1.9922099502084934e-06, "loss": 0.8455, "step": 24090 }, { "epoch": 0.21, "learning_rate": 1.991564389069108e-06, "loss": 0.8136, "step": 24091 }, { "epoch": 0.21, "learning_rate": 1.9909189209739675e-06, "loss": 0.8291, "step": 24092 }, { "epoch": 0.21, "learning_rate": 1.99027354593057e-06, "loss": 0.8722, "step": 24093 }, { "epoch": 0.21, "learning_rate": 1.989628263946416e-06, "loss": 0.8852, "step": 24094 }, { "epoch": 0.21, "learning_rate": 1.9889830750289986e-06, "loss": 0.8601, "step": 24095 }, { "epoch": 0.21, "learning_rate": 1.988337979185816e-06, "loss": 0.8837, "step": 24096 }, { "epoch": 0.21, "learning_rate": 1.9876929764243626e-06, "loss": 0.8481, "step": 24097 }, { "epoch": 0.21, "learning_rate": 1.987048066752132e-06, "loss": 0.8612, "step": 24098 }, { "epoch": 0.21, "learning_rate": 1.9864032501766173e-06, "loss": 0.8313, "step": 24099 }, { "epoch": 0.21, "learning_rate": 1.98575852670531e-06, "loss": 0.8645, "step": 24100 }, { "epoch": 0.21, "learning_rate": 1.985113896345705e-06, "loss": 0.8676, "step": 24101 }, { "epoch": 0.21, "learning_rate": 1.9844693591052842e-06, "loss": 0.8477, "step": 24102 }, { "epoch": 0.21, "learning_rate": 1.98382491499154e-06, "loss": 0.873, "step": 24103 }, { "epoch": 0.21, "learning_rate": 1.983180564011956e-06, "loss": 0.8747, "step": 24104 }, { "epoch": 0.21, "learning_rate": 1.9825363061740266e-06, "loss": 0.8704, "step": 24105 }, { "epoch": 0.21, "learning_rate": 1.98189214148523e-06, "loss": 0.8282, "step": 24106 }, { "epoch": 0.21, "learning_rate": 1.9812480699530535e-06, "loss": 0.8789, "step": 24107 }, { "epoch": 0.21, "learning_rate": 1.9806040915849777e-06, "loss": 0.8816, "step": 24108 }, { "epoch": 0.21, "learning_rate": 1.979960206388486e-06, "loss": 0.8733, "step": 24109 }, { "epoch": 0.21, "learning_rate": 1.9793164143710588e-06, "loss": 0.862, "step": 24110 }, { "epoch": 0.21, "learning_rate": 1.9786727155401762e-06, "loss": 0.8453, "step": 24111 }, { "epoch": 0.21, "learning_rate": 1.97802910990332e-06, "loss": 0.8531, "step": 24112 }, { "epoch": 0.21, "learning_rate": 1.977385597467961e-06, "loss": 0.8724, "step": 24113 }, { "epoch": 0.21, "learning_rate": 1.9767421782415797e-06, "loss": 0.8691, "step": 24114 }, { "epoch": 0.21, "learning_rate": 1.9760988522316517e-06, "loss": 0.8293, "step": 24115 }, { "epoch": 0.21, "learning_rate": 1.97545561944565e-06, "loss": 0.8916, "step": 24116 }, { "epoch": 0.21, "learning_rate": 1.9748124798910484e-06, "loss": 0.8286, "step": 24117 }, { "epoch": 0.21, "learning_rate": 1.9741694335753192e-06, "loss": 0.8225, "step": 24118 }, { "epoch": 0.21, "learning_rate": 1.973526480505937e-06, "loss": 0.8761, "step": 24119 }, { "epoch": 0.21, "learning_rate": 1.972883620690366e-06, "loss": 0.8658, "step": 24120 }, { "epoch": 0.21, "learning_rate": 1.972240854136076e-06, "loss": 0.8801, "step": 24121 }, { "epoch": 0.21, "learning_rate": 1.9715981808505346e-06, "loss": 0.8364, "step": 24122 }, { "epoch": 0.21, "learning_rate": 1.9709556008412157e-06, "loss": 0.8267, "step": 24123 }, { "epoch": 0.21, "learning_rate": 1.9703131141155763e-06, "loss": 0.87, "step": 24124 }, { "epoch": 0.21, "learning_rate": 1.969670720681085e-06, "loss": 0.8677, "step": 24125 }, { "epoch": 0.21, "learning_rate": 1.9690284205452058e-06, "loss": 0.8314, "step": 24126 }, { "epoch": 0.21, "learning_rate": 1.9683862137153985e-06, "loss": 0.8169, "step": 24127 }, { "epoch": 0.21, "learning_rate": 1.9677441001991227e-06, "loss": 0.8665, "step": 24128 }, { "epoch": 0.21, "learning_rate": 1.9671020800038452e-06, "loss": 0.8609, "step": 24129 }, { "epoch": 0.21, "learning_rate": 1.966460153137023e-06, "loss": 0.866, "step": 24130 }, { "epoch": 0.21, "learning_rate": 1.965818319606112e-06, "loss": 0.8834, "step": 24131 }, { "epoch": 0.21, "learning_rate": 1.9651765794185686e-06, "loss": 0.8625, "step": 24132 }, { "epoch": 0.21, "learning_rate": 1.964534932581854e-06, "loss": 0.8565, "step": 24133 }, { "epoch": 0.21, "learning_rate": 1.963893379103413e-06, "loss": 0.8594, "step": 24134 }, { "epoch": 0.21, "learning_rate": 1.9632519189907094e-06, "loss": 0.8451, "step": 24135 }, { "epoch": 0.21, "learning_rate": 1.9626105522511917e-06, "loss": 0.8607, "step": 24136 }, { "epoch": 0.21, "learning_rate": 1.961969278892314e-06, "loss": 0.8462, "step": 24137 }, { "epoch": 0.21, "learning_rate": 1.961328098921522e-06, "loss": 0.8364, "step": 24138 }, { "epoch": 0.21, "learning_rate": 1.9606870123462684e-06, "loss": 0.8629, "step": 24139 }, { "epoch": 0.21, "learning_rate": 1.960046019174e-06, "loss": 0.8535, "step": 24140 }, { "epoch": 0.21, "learning_rate": 1.9594051194121657e-06, "loss": 0.8585, "step": 24141 }, { "epoch": 0.21, "learning_rate": 1.958764313068211e-06, "loss": 0.9141, "step": 24142 }, { "epoch": 0.21, "learning_rate": 1.9581236001495805e-06, "loss": 0.8386, "step": 24143 }, { "epoch": 0.21, "learning_rate": 1.9574829806637206e-06, "loss": 0.8584, "step": 24144 }, { "epoch": 0.21, "learning_rate": 1.9568424546180707e-06, "loss": 0.8911, "step": 24145 }, { "epoch": 0.21, "learning_rate": 1.956202022020073e-06, "loss": 0.8323, "step": 24146 }, { "epoch": 0.21, "learning_rate": 1.9555616828771696e-06, "loss": 0.8471, "step": 24147 }, { "epoch": 0.21, "learning_rate": 1.9549214371968008e-06, "loss": 0.8929, "step": 24148 }, { "epoch": 0.21, "learning_rate": 1.9542812849864024e-06, "loss": 0.8504, "step": 24149 }, { "epoch": 0.21, "learning_rate": 1.9536412262534145e-06, "loss": 0.8814, "step": 24150 }, { "epoch": 0.21, "learning_rate": 1.953001261005275e-06, "loss": 0.8431, "step": 24151 }, { "epoch": 0.21, "learning_rate": 1.952361389249412e-06, "loss": 0.8785, "step": 24152 }, { "epoch": 0.21, "learning_rate": 1.951721610993267e-06, "loss": 0.8675, "step": 24153 }, { "epoch": 0.21, "learning_rate": 1.9510819262442705e-06, "loss": 0.8444, "step": 24154 }, { "epoch": 0.21, "learning_rate": 1.9504423350098577e-06, "loss": 0.828, "step": 24155 }, { "epoch": 0.21, "learning_rate": 1.949802837297453e-06, "loss": 0.8112, "step": 24156 }, { "epoch": 0.21, "learning_rate": 1.9491634331144906e-06, "loss": 0.895, "step": 24157 }, { "epoch": 0.21, "learning_rate": 1.9485241224683983e-06, "loss": 0.8649, "step": 24158 }, { "epoch": 0.21, "learning_rate": 1.9478849053666037e-06, "loss": 0.8297, "step": 24159 }, { "epoch": 0.21, "learning_rate": 1.9472457818165335e-06, "loss": 0.8268, "step": 24160 }, { "epoch": 0.21, "learning_rate": 1.9466067518256138e-06, "loss": 0.8379, "step": 24161 }, { "epoch": 0.21, "learning_rate": 1.9459678154012708e-06, "loss": 0.8761, "step": 24162 }, { "epoch": 0.21, "learning_rate": 1.945328972550924e-06, "loss": 0.8849, "step": 24163 }, { "epoch": 0.21, "learning_rate": 1.9446902232819963e-06, "loss": 0.8922, "step": 24164 }, { "epoch": 0.21, "learning_rate": 1.94405156760191e-06, "loss": 0.8533, "step": 24165 }, { "epoch": 0.21, "learning_rate": 1.9434130055180843e-06, "loss": 0.8627, "step": 24166 }, { "epoch": 0.21, "learning_rate": 1.9427745370379404e-06, "loss": 0.8487, "step": 24167 }, { "epoch": 0.21, "learning_rate": 1.942136162168894e-06, "loss": 0.8186, "step": 24168 }, { "epoch": 0.21, "learning_rate": 1.941497880918366e-06, "loss": 0.8715, "step": 24169 }, { "epoch": 0.21, "learning_rate": 1.940859693293765e-06, "loss": 0.8601, "step": 24170 }, { "epoch": 0.21, "learning_rate": 1.9402215993025074e-06, "loss": 0.8552, "step": 24171 }, { "epoch": 0.21, "learning_rate": 1.939583598952014e-06, "loss": 0.8598, "step": 24172 }, { "epoch": 0.21, "learning_rate": 1.9389456922496896e-06, "loss": 0.8747, "step": 24173 }, { "epoch": 0.21, "learning_rate": 1.9383078792029474e-06, "loss": 0.8264, "step": 24174 }, { "epoch": 0.21, "learning_rate": 1.9376701598191994e-06, "loss": 0.8452, "step": 24175 }, { "epoch": 0.21, "learning_rate": 1.937032534105855e-06, "loss": 0.8382, "step": 24176 }, { "epoch": 0.21, "learning_rate": 1.936395002070317e-06, "loss": 0.844, "step": 24177 }, { "epoch": 0.21, "learning_rate": 1.935757563719999e-06, "loss": 0.8226, "step": 24178 }, { "epoch": 0.21, "learning_rate": 1.9351202190623066e-06, "loss": 0.8451, "step": 24179 }, { "epoch": 0.21, "learning_rate": 1.9344829681046394e-06, "loss": 0.8478, "step": 24180 }, { "epoch": 0.21, "learning_rate": 1.9338458108544046e-06, "loss": 0.863, "step": 24181 }, { "epoch": 0.21, "learning_rate": 1.933208747319004e-06, "loss": 0.9084, "step": 24182 }, { "epoch": 0.21, "learning_rate": 1.9325717775058404e-06, "loss": 0.8124, "step": 24183 }, { "epoch": 0.21, "learning_rate": 1.9319349014223132e-06, "loss": 0.8357, "step": 24184 }, { "epoch": 0.21, "learning_rate": 1.9312981190758227e-06, "loss": 0.8626, "step": 24185 }, { "epoch": 0.21, "learning_rate": 1.930661430473768e-06, "loss": 0.9193, "step": 24186 }, { "epoch": 0.21, "learning_rate": 1.930024835623544e-06, "loss": 0.8436, "step": 24187 }, { "epoch": 0.21, "learning_rate": 1.929388334532547e-06, "loss": 0.8736, "step": 24188 }, { "epoch": 0.21, "learning_rate": 1.928751927208172e-06, "loss": 0.8583, "step": 24189 }, { "epoch": 0.21, "learning_rate": 1.928115613657816e-06, "loss": 0.8315, "step": 24190 }, { "epoch": 0.21, "learning_rate": 1.927479393888868e-06, "loss": 0.8439, "step": 24191 }, { "epoch": 0.21, "learning_rate": 1.926843267908722e-06, "loss": 0.8335, "step": 24192 }, { "epoch": 0.21, "learning_rate": 1.926207235724771e-06, "loss": 0.8562, "step": 24193 }, { "epoch": 0.21, "learning_rate": 1.9255712973443987e-06, "loss": 0.8521, "step": 24194 }, { "epoch": 0.21, "learning_rate": 1.9249354527749973e-06, "loss": 0.8602, "step": 24195 }, { "epoch": 0.21, "learning_rate": 1.9242997020239506e-06, "loss": 0.8521, "step": 24196 }, { "epoch": 0.21, "learning_rate": 1.9236640450986533e-06, "loss": 0.8353, "step": 24197 }, { "epoch": 0.21, "learning_rate": 1.9230284820064826e-06, "loss": 0.8558, "step": 24198 }, { "epoch": 0.21, "learning_rate": 1.922393012754825e-06, "loss": 0.8522, "step": 24199 }, { "epoch": 0.21, "learning_rate": 1.9217576373510637e-06, "loss": 0.8419, "step": 24200 }, { "epoch": 0.21, "learning_rate": 1.9211223558025815e-06, "loss": 0.909, "step": 24201 }, { "epoch": 0.21, "learning_rate": 1.9204871681167594e-06, "loss": 0.833, "step": 24202 }, { "epoch": 0.21, "learning_rate": 1.9198520743009754e-06, "loss": 0.8424, "step": 24203 }, { "epoch": 0.21, "learning_rate": 1.9192170743626115e-06, "loss": 0.8639, "step": 24204 }, { "epoch": 0.21, "learning_rate": 1.918582168309041e-06, "loss": 0.8655, "step": 24205 }, { "epoch": 0.21, "learning_rate": 1.9179473561476414e-06, "loss": 0.8734, "step": 24206 }, { "epoch": 0.21, "learning_rate": 1.917312637885791e-06, "loss": 0.8274, "step": 24207 }, { "epoch": 0.21, "learning_rate": 1.9166780135308605e-06, "loss": 0.866, "step": 24208 }, { "epoch": 0.21, "learning_rate": 1.916043483090226e-06, "loss": 0.8613, "step": 24209 }, { "epoch": 0.21, "learning_rate": 1.9154090465712584e-06, "loss": 0.8513, "step": 24210 }, { "epoch": 0.22, "learning_rate": 1.9147747039813314e-06, "loss": 0.8378, "step": 24211 }, { "epoch": 0.22, "learning_rate": 1.9141404553278097e-06, "loss": 0.8768, "step": 24212 }, { "epoch": 0.22, "learning_rate": 1.9135063006180655e-06, "loss": 0.838, "step": 24213 }, { "epoch": 0.22, "learning_rate": 1.9128722398594633e-06, "loss": 0.869, "step": 24214 }, { "epoch": 0.22, "learning_rate": 1.912238273059378e-06, "loss": 0.8085, "step": 24215 }, { "epoch": 0.22, "learning_rate": 1.9116044002251675e-06, "loss": 0.8915, "step": 24216 }, { "epoch": 0.22, "learning_rate": 1.9109706213641975e-06, "loss": 0.8645, "step": 24217 }, { "epoch": 0.22, "learning_rate": 1.9103369364838365e-06, "loss": 0.8396, "step": 24218 }, { "epoch": 0.22, "learning_rate": 1.9097033455914393e-06, "loss": 0.8648, "step": 24219 }, { "epoch": 0.22, "learning_rate": 1.9090698486943683e-06, "loss": 0.8217, "step": 24220 }, { "epoch": 0.22, "learning_rate": 1.908436445799988e-06, "loss": 0.8491, "step": 24221 }, { "epoch": 0.22, "learning_rate": 1.9078031369156593e-06, "loss": 0.9211, "step": 24222 }, { "epoch": 0.22, "learning_rate": 1.907169922048734e-06, "loss": 0.8322, "step": 24223 }, { "epoch": 0.22, "learning_rate": 1.90653680120657e-06, "loss": 0.8175, "step": 24224 }, { "epoch": 0.22, "learning_rate": 1.9059037743965247e-06, "loss": 0.8863, "step": 24225 }, { "epoch": 0.22, "learning_rate": 1.9052708416259525e-06, "loss": 0.846, "step": 24226 }, { "epoch": 0.22, "learning_rate": 1.9046380029022071e-06, "loss": 0.8577, "step": 24227 }, { "epoch": 0.22, "learning_rate": 1.9040052582326407e-06, "loss": 0.8448, "step": 24228 }, { "epoch": 0.22, "learning_rate": 1.9033726076246084e-06, "loss": 0.8416, "step": 24229 }, { "epoch": 0.22, "learning_rate": 1.9027400510854533e-06, "loss": 0.8584, "step": 24230 }, { "epoch": 0.22, "learning_rate": 1.902107588622528e-06, "loss": 0.8372, "step": 24231 }, { "epoch": 0.22, "learning_rate": 1.901475220243182e-06, "loss": 0.852, "step": 24232 }, { "epoch": 0.22, "learning_rate": 1.9008429459547606e-06, "loss": 0.8536, "step": 24233 }, { "epoch": 0.22, "learning_rate": 1.9002107657646107e-06, "loss": 0.854, "step": 24234 }, { "epoch": 0.22, "learning_rate": 1.8995786796800764e-06, "loss": 0.8182, "step": 24235 }, { "epoch": 0.22, "learning_rate": 1.898946687708505e-06, "loss": 0.8825, "step": 24236 }, { "epoch": 0.22, "learning_rate": 1.898314789857233e-06, "loss": 0.8983, "step": 24237 }, { "epoch": 0.22, "learning_rate": 1.8976829861336054e-06, "loss": 0.8378, "step": 24238 }, { "epoch": 0.22, "learning_rate": 1.8970512765449622e-06, "loss": 0.8455, "step": 24239 }, { "epoch": 0.22, "learning_rate": 1.896419661098643e-06, "loss": 0.8618, "step": 24240 }, { "epoch": 0.22, "learning_rate": 1.895788139801985e-06, "loss": 0.8475, "step": 24241 }, { "epoch": 0.22, "learning_rate": 1.8951567126623272e-06, "loss": 0.8463, "step": 24242 }, { "epoch": 0.22, "learning_rate": 1.8945253796870078e-06, "loss": 0.851, "step": 24243 }, { "epoch": 0.22, "learning_rate": 1.8938941408833521e-06, "loss": 0.8408, "step": 24244 }, { "epoch": 0.22, "learning_rate": 1.8932629962587046e-06, "loss": 0.8524, "step": 24245 }, { "epoch": 0.22, "learning_rate": 1.8926319458203946e-06, "loss": 0.8517, "step": 24246 }, { "epoch": 0.22, "learning_rate": 1.8920009895757551e-06, "loss": 0.8291, "step": 24247 }, { "epoch": 0.22, "learning_rate": 1.8913701275321128e-06, "loss": 0.8806, "step": 24248 }, { "epoch": 0.22, "learning_rate": 1.8907393596968005e-06, "loss": 0.8672, "step": 24249 }, { "epoch": 0.22, "learning_rate": 1.8901086860771445e-06, "loss": 0.8231, "step": 24250 }, { "epoch": 0.22, "learning_rate": 1.8894781066804735e-06, "loss": 0.802, "step": 24251 }, { "epoch": 0.22, "learning_rate": 1.888847621514114e-06, "loss": 0.8432, "step": 24252 }, { "epoch": 0.22, "learning_rate": 1.8882172305853896e-06, "loss": 0.8605, "step": 24253 }, { "epoch": 0.22, "learning_rate": 1.8875869339016296e-06, "loss": 0.8377, "step": 24254 }, { "epoch": 0.22, "learning_rate": 1.8869567314701498e-06, "loss": 0.8664, "step": 24255 }, { "epoch": 0.22, "learning_rate": 1.8863266232982746e-06, "loss": 0.8465, "step": 24256 }, { "epoch": 0.22, "learning_rate": 1.885696609393326e-06, "loss": 0.8657, "step": 24257 }, { "epoch": 0.22, "learning_rate": 1.8850666897626224e-06, "loss": 0.8718, "step": 24258 }, { "epoch": 0.22, "learning_rate": 1.8844368644134825e-06, "loss": 0.8848, "step": 24259 }, { "epoch": 0.22, "learning_rate": 1.8838071333532248e-06, "loss": 0.8678, "step": 24260 }, { "epoch": 0.22, "learning_rate": 1.883177496589168e-06, "loss": 0.8913, "step": 24261 }, { "epoch": 0.22, "learning_rate": 1.8825479541286218e-06, "loss": 0.8409, "step": 24262 }, { "epoch": 0.22, "learning_rate": 1.8819185059789003e-06, "loss": 0.8446, "step": 24263 }, { "epoch": 0.22, "learning_rate": 1.8812891521473242e-06, "loss": 0.8667, "step": 24264 }, { "epoch": 0.22, "learning_rate": 1.880659892641199e-06, "loss": 0.8683, "step": 24265 }, { "epoch": 0.22, "learning_rate": 1.8800307274678364e-06, "loss": 0.845, "step": 24266 }, { "epoch": 0.22, "learning_rate": 1.8794016566345486e-06, "loss": 0.8201, "step": 24267 }, { "epoch": 0.22, "learning_rate": 1.8787726801486438e-06, "loss": 0.8338, "step": 24268 }, { "epoch": 0.22, "learning_rate": 1.8781437980174244e-06, "loss": 0.8348, "step": 24269 }, { "epoch": 0.22, "learning_rate": 1.8775150102482042e-06, "loss": 0.8628, "step": 24270 }, { "epoch": 0.22, "learning_rate": 1.8768863168482875e-06, "loss": 0.8263, "step": 24271 }, { "epoch": 0.22, "learning_rate": 1.8762577178249753e-06, "loss": 0.8682, "step": 24272 }, { "epoch": 0.22, "learning_rate": 1.8756292131855714e-06, "loss": 0.8925, "step": 24273 }, { "epoch": 0.22, "learning_rate": 1.875000802937379e-06, "loss": 0.8534, "step": 24274 }, { "epoch": 0.22, "learning_rate": 1.8743724870876989e-06, "loss": 0.8504, "step": 24275 }, { "epoch": 0.22, "learning_rate": 1.873744265643831e-06, "loss": 0.8455, "step": 24276 }, { "epoch": 0.22, "learning_rate": 1.8731161386130748e-06, "loss": 0.8597, "step": 24277 }, { "epoch": 0.22, "learning_rate": 1.8724881060027301e-06, "loss": 0.8603, "step": 24278 }, { "epoch": 0.22, "learning_rate": 1.8718601678200887e-06, "loss": 0.8191, "step": 24279 }, { "epoch": 0.22, "learning_rate": 1.8712323240724483e-06, "loss": 0.869, "step": 24280 }, { "epoch": 0.22, "learning_rate": 1.8706045747671031e-06, "loss": 0.8577, "step": 24281 }, { "epoch": 0.22, "learning_rate": 1.869976919911347e-06, "loss": 0.8825, "step": 24282 }, { "epoch": 0.22, "learning_rate": 1.8693493595124734e-06, "loss": 0.827, "step": 24283 }, { "epoch": 0.22, "learning_rate": 1.8687218935777706e-06, "loss": 0.8706, "step": 24284 }, { "epoch": 0.22, "learning_rate": 1.868094522114534e-06, "loss": 0.8705, "step": 24285 }, { "epoch": 0.22, "learning_rate": 1.8674672451300469e-06, "loss": 0.8584, "step": 24286 }, { "epoch": 0.22, "learning_rate": 1.8668400626315953e-06, "loss": 0.8518, "step": 24287 }, { "epoch": 0.22, "learning_rate": 1.8662129746264735e-06, "loss": 0.888, "step": 24288 }, { "epoch": 0.22, "learning_rate": 1.8655859811219656e-06, "loss": 0.8492, "step": 24289 }, { "epoch": 0.22, "learning_rate": 1.8649590821253515e-06, "loss": 0.8684, "step": 24290 }, { "epoch": 0.22, "learning_rate": 1.8643322776439176e-06, "loss": 0.8597, "step": 24291 }, { "epoch": 0.22, "learning_rate": 1.8637055676849458e-06, "loss": 0.8516, "step": 24292 }, { "epoch": 0.22, "learning_rate": 1.863078952255718e-06, "loss": 0.8356, "step": 24293 }, { "epoch": 0.22, "learning_rate": 1.862452431363514e-06, "loss": 0.8878, "step": 24294 }, { "epoch": 0.22, "learning_rate": 1.8618260050156122e-06, "loss": 0.858, "step": 24295 }, { "epoch": 0.22, "learning_rate": 1.861199673219295e-06, "loss": 0.8791, "step": 24296 }, { "epoch": 0.22, "learning_rate": 1.8605734359818318e-06, "loss": 0.8578, "step": 24297 }, { "epoch": 0.22, "learning_rate": 1.8599472933105033e-06, "loss": 0.8624, "step": 24298 }, { "epoch": 0.22, "learning_rate": 1.859321245212582e-06, "loss": 0.8519, "step": 24299 }, { "epoch": 0.22, "learning_rate": 1.8586952916953438e-06, "loss": 0.8711, "step": 24300 }, { "epoch": 0.22, "learning_rate": 1.8580694327660597e-06, "loss": 0.8432, "step": 24301 }, { "epoch": 0.22, "learning_rate": 1.8574436684320008e-06, "loss": 0.8419, "step": 24302 }, { "epoch": 0.22, "learning_rate": 1.8568179987004408e-06, "loss": 0.8387, "step": 24303 }, { "epoch": 0.22, "learning_rate": 1.8561924235786445e-06, "loss": 0.8508, "step": 24304 }, { "epoch": 0.22, "learning_rate": 1.8555669430738809e-06, "loss": 0.859, "step": 24305 }, { "epoch": 0.22, "learning_rate": 1.8549415571934148e-06, "loss": 0.8616, "step": 24306 }, { "epoch": 0.22, "learning_rate": 1.8543162659445214e-06, "loss": 0.8298, "step": 24307 }, { "epoch": 0.22, "learning_rate": 1.8536910693344557e-06, "loss": 0.8579, "step": 24308 }, { "epoch": 0.22, "learning_rate": 1.8530659673704844e-06, "loss": 0.8736, "step": 24309 }, { "epoch": 0.22, "learning_rate": 1.8524409600598748e-06, "loss": 0.8503, "step": 24310 }, { "epoch": 0.22, "learning_rate": 1.8518160474098802e-06, "loss": 0.8816, "step": 24311 }, { "epoch": 0.22, "learning_rate": 1.8511912294277622e-06, "loss": 0.8134, "step": 24312 }, { "epoch": 0.22, "learning_rate": 1.8505665061207866e-06, "loss": 0.8488, "step": 24313 }, { "epoch": 0.22, "learning_rate": 1.8499418774962097e-06, "loss": 0.8677, "step": 24314 }, { "epoch": 0.22, "learning_rate": 1.8493173435612844e-06, "loss": 0.8706, "step": 24315 }, { "epoch": 0.22, "learning_rate": 1.8486929043232693e-06, "loss": 0.8638, "step": 24316 }, { "epoch": 0.22, "learning_rate": 1.84806855978942e-06, "loss": 0.8474, "step": 24317 }, { "epoch": 0.22, "learning_rate": 1.847444309966988e-06, "loss": 0.8826, "step": 24318 }, { "epoch": 0.22, "learning_rate": 1.846820154863228e-06, "loss": 0.8178, "step": 24319 }, { "epoch": 0.22, "learning_rate": 1.8461960944853918e-06, "loss": 0.8544, "step": 24320 }, { "epoch": 0.22, "learning_rate": 1.8455721288407313e-06, "loss": 0.8231, "step": 24321 }, { "epoch": 0.22, "learning_rate": 1.8449482579364908e-06, "loss": 0.8421, "step": 24322 }, { "epoch": 0.22, "learning_rate": 1.844324481779921e-06, "loss": 0.8279, "step": 24323 }, { "epoch": 0.22, "learning_rate": 1.843700800378271e-06, "loss": 0.8527, "step": 24324 }, { "epoch": 0.22, "learning_rate": 1.8430772137387853e-06, "loss": 0.7929, "step": 24325 }, { "epoch": 0.22, "learning_rate": 1.8424537218687088e-06, "loss": 0.834, "step": 24326 }, { "epoch": 0.22, "learning_rate": 1.8418303247752867e-06, "loss": 0.841, "step": 24327 }, { "epoch": 0.22, "learning_rate": 1.841207022465763e-06, "loss": 0.8453, "step": 24328 }, { "epoch": 0.22, "learning_rate": 1.8405838149473743e-06, "loss": 0.8921, "step": 24329 }, { "epoch": 0.22, "learning_rate": 1.8399607022273624e-06, "loss": 0.8707, "step": 24330 }, { "epoch": 0.22, "learning_rate": 1.839337684312973e-06, "loss": 0.821, "step": 24331 }, { "epoch": 0.22, "learning_rate": 1.8387147612114376e-06, "loss": 0.8087, "step": 24332 }, { "epoch": 0.22, "learning_rate": 1.8380919329299962e-06, "loss": 0.8588, "step": 24333 }, { "epoch": 0.22, "learning_rate": 1.8374691994758853e-06, "loss": 0.8717, "step": 24334 }, { "epoch": 0.22, "learning_rate": 1.8368465608563413e-06, "loss": 0.842, "step": 24335 }, { "epoch": 0.22, "learning_rate": 1.8362240170785916e-06, "loss": 0.882, "step": 24336 }, { "epoch": 0.22, "learning_rate": 1.8356015681498774e-06, "loss": 0.8769, "step": 24337 }, { "epoch": 0.22, "learning_rate": 1.8349792140774258e-06, "loss": 0.8652, "step": 24338 }, { "epoch": 0.22, "learning_rate": 1.8343569548684714e-06, "loss": 0.8943, "step": 24339 }, { "epoch": 0.22, "learning_rate": 1.8337347905302383e-06, "loss": 0.8561, "step": 24340 }, { "epoch": 0.22, "learning_rate": 1.8331127210699583e-06, "loss": 0.831, "step": 24341 }, { "epoch": 0.22, "learning_rate": 1.8324907464948571e-06, "loss": 0.8293, "step": 24342 }, { "epoch": 0.22, "learning_rate": 1.831868866812163e-06, "loss": 0.8649, "step": 24343 }, { "epoch": 0.22, "learning_rate": 1.8312470820290994e-06, "loss": 0.845, "step": 24344 }, { "epoch": 0.22, "learning_rate": 1.8306253921528927e-06, "loss": 0.8979, "step": 24345 }, { "epoch": 0.22, "learning_rate": 1.8300037971907658e-06, "loss": 0.8685, "step": 24346 }, { "epoch": 0.22, "learning_rate": 1.8293822971499363e-06, "loss": 0.8383, "step": 24347 }, { "epoch": 0.22, "learning_rate": 1.8287608920376288e-06, "loss": 0.8488, "step": 24348 }, { "epoch": 0.22, "learning_rate": 1.8281395818610615e-06, "loss": 0.8662, "step": 24349 }, { "epoch": 0.22, "learning_rate": 1.8275183666274532e-06, "loss": 0.8681, "step": 24350 }, { "epoch": 0.22, "learning_rate": 1.8268972463440216e-06, "loss": 0.8286, "step": 24351 }, { "epoch": 0.22, "learning_rate": 1.826276221017983e-06, "loss": 0.8617, "step": 24352 }, { "epoch": 0.22, "learning_rate": 1.825655290656555e-06, "loss": 0.853, "step": 24353 }, { "epoch": 0.22, "learning_rate": 1.8250344552669473e-06, "loss": 0.8626, "step": 24354 }, { "epoch": 0.22, "learning_rate": 1.8244137148563724e-06, "loss": 0.8383, "step": 24355 }, { "epoch": 0.22, "learning_rate": 1.8237930694320494e-06, "loss": 0.8822, "step": 24356 }, { "epoch": 0.22, "learning_rate": 1.8231725190011817e-06, "loss": 0.8689, "step": 24357 }, { "epoch": 0.22, "learning_rate": 1.8225520635709836e-06, "loss": 0.8565, "step": 24358 }, { "epoch": 0.22, "learning_rate": 1.8219317031486606e-06, "loss": 0.8729, "step": 24359 }, { "epoch": 0.22, "learning_rate": 1.821311437741422e-06, "loss": 0.8533, "step": 24360 }, { "epoch": 0.22, "learning_rate": 1.8206912673564746e-06, "loss": 0.8352, "step": 24361 }, { "epoch": 0.22, "learning_rate": 1.8200711920010228e-06, "loss": 0.8491, "step": 24362 }, { "epoch": 0.22, "learning_rate": 1.8194512116822726e-06, "loss": 0.8232, "step": 24363 }, { "epoch": 0.22, "learning_rate": 1.8188313264074243e-06, "loss": 0.8979, "step": 24364 }, { "epoch": 0.22, "learning_rate": 1.8182115361836795e-06, "loss": 0.8561, "step": 24365 }, { "epoch": 0.22, "learning_rate": 1.8175918410182425e-06, "loss": 0.7979, "step": 24366 }, { "epoch": 0.22, "learning_rate": 1.81697224091831e-06, "loss": 0.8577, "step": 24367 }, { "epoch": 0.22, "learning_rate": 1.8163527358910825e-06, "loss": 0.8198, "step": 24368 }, { "epoch": 0.22, "learning_rate": 1.815733325943757e-06, "loss": 0.8739, "step": 24369 }, { "epoch": 0.22, "learning_rate": 1.815114011083532e-06, "loss": 0.8622, "step": 24370 }, { "epoch": 0.22, "learning_rate": 1.8144947913175992e-06, "loss": 0.8699, "step": 24371 }, { "epoch": 0.22, "learning_rate": 1.8138756666531542e-06, "loss": 0.8609, "step": 24372 }, { "epoch": 0.22, "learning_rate": 1.8132566370973914e-06, "loss": 0.8949, "step": 24373 }, { "epoch": 0.22, "learning_rate": 1.8126377026575015e-06, "loss": 0.8253, "step": 24374 }, { "epoch": 0.22, "learning_rate": 1.8120188633406755e-06, "loss": 0.8206, "step": 24375 }, { "epoch": 0.22, "learning_rate": 1.8114001191541041e-06, "loss": 0.8533, "step": 24376 }, { "epoch": 0.22, "learning_rate": 1.8107814701049775e-06, "loss": 0.8435, "step": 24377 }, { "epoch": 0.22, "learning_rate": 1.8101629162004798e-06, "loss": 0.852, "step": 24378 }, { "epoch": 0.22, "learning_rate": 1.809544457447796e-06, "loss": 0.8142, "step": 24379 }, { "epoch": 0.22, "learning_rate": 1.8089260938541186e-06, "loss": 0.8748, "step": 24380 }, { "epoch": 0.22, "learning_rate": 1.8083078254266296e-06, "loss": 0.8879, "step": 24381 }, { "epoch": 0.22, "learning_rate": 1.8076896521725084e-06, "loss": 0.9106, "step": 24382 }, { "epoch": 0.22, "learning_rate": 1.8070715740989386e-06, "loss": 0.83, "step": 24383 }, { "epoch": 0.22, "learning_rate": 1.8064535912131032e-06, "loss": 0.8345, "step": 24384 }, { "epoch": 0.22, "learning_rate": 1.8058357035221807e-06, "loss": 0.8846, "step": 24385 }, { "epoch": 0.22, "learning_rate": 1.805217911033349e-06, "loss": 0.8975, "step": 24386 }, { "epoch": 0.22, "learning_rate": 1.8046002137537876e-06, "loss": 0.8286, "step": 24387 }, { "epoch": 0.22, "learning_rate": 1.8039826116906755e-06, "loss": 0.873, "step": 24388 }, { "epoch": 0.22, "learning_rate": 1.8033651048511824e-06, "loss": 0.8421, "step": 24389 }, { "epoch": 0.22, "learning_rate": 1.8027476932424848e-06, "loss": 0.852, "step": 24390 }, { "epoch": 0.22, "learning_rate": 1.802130376871757e-06, "loss": 0.8068, "step": 24391 }, { "epoch": 0.22, "learning_rate": 1.8015131557461695e-06, "loss": 0.812, "step": 24392 }, { "epoch": 0.22, "learning_rate": 1.800896029872895e-06, "loss": 0.85, "step": 24393 }, { "epoch": 0.22, "learning_rate": 1.800278999259103e-06, "loss": 0.889, "step": 24394 }, { "epoch": 0.22, "learning_rate": 1.7996620639119655e-06, "loss": 0.8409, "step": 24395 }, { "epoch": 0.22, "learning_rate": 1.7990452238386436e-06, "loss": 0.8721, "step": 24396 }, { "epoch": 0.22, "learning_rate": 1.798428479046307e-06, "loss": 0.8586, "step": 24397 }, { "epoch": 0.22, "learning_rate": 1.79781182954212e-06, "loss": 0.8853, "step": 24398 }, { "epoch": 0.22, "learning_rate": 1.7971952753332522e-06, "loss": 0.8321, "step": 24399 }, { "epoch": 0.22, "learning_rate": 1.7965788164268615e-06, "loss": 0.8691, "step": 24400 }, { "epoch": 0.22, "learning_rate": 1.7959624528301122e-06, "loss": 0.8355, "step": 24401 }, { "epoch": 0.22, "learning_rate": 1.7953461845501663e-06, "loss": 0.8179, "step": 24402 }, { "epoch": 0.22, "learning_rate": 1.794730011594179e-06, "loss": 0.8099, "step": 24403 }, { "epoch": 0.22, "learning_rate": 1.7941139339693137e-06, "loss": 0.8339, "step": 24404 }, { "epoch": 0.22, "learning_rate": 1.7934979516827278e-06, "loss": 0.8486, "step": 24405 }, { "epoch": 0.22, "learning_rate": 1.79288206474158e-06, "loss": 0.8905, "step": 24406 }, { "epoch": 0.22, "learning_rate": 1.7922662731530204e-06, "loss": 0.7867, "step": 24407 }, { "epoch": 0.22, "learning_rate": 1.7916505769242055e-06, "loss": 0.8712, "step": 24408 }, { "epoch": 0.22, "learning_rate": 1.7910349760622903e-06, "loss": 0.8626, "step": 24409 }, { "epoch": 0.22, "learning_rate": 1.7904194705744249e-06, "loss": 0.8868, "step": 24410 }, { "epoch": 0.22, "learning_rate": 1.7898040604677624e-06, "loss": 0.8685, "step": 24411 }, { "epoch": 0.22, "learning_rate": 1.7891887457494517e-06, "loss": 0.8353, "step": 24412 }, { "epoch": 0.22, "learning_rate": 1.7885735264266435e-06, "loss": 0.8548, "step": 24413 }, { "epoch": 0.22, "learning_rate": 1.7879584025064822e-06, "loss": 0.837, "step": 24414 }, { "epoch": 0.22, "learning_rate": 1.7873433739961166e-06, "loss": 0.8142, "step": 24415 }, { "epoch": 0.22, "learning_rate": 1.786728440902692e-06, "loss": 0.8471, "step": 24416 }, { "epoch": 0.22, "learning_rate": 1.7861136032333525e-06, "loss": 0.8402, "step": 24417 }, { "epoch": 0.22, "learning_rate": 1.7854988609952417e-06, "loss": 0.8511, "step": 24418 }, { "epoch": 0.22, "learning_rate": 1.7848842141955013e-06, "loss": 0.8415, "step": 24419 }, { "epoch": 0.22, "learning_rate": 1.784269662841277e-06, "loss": 0.8239, "step": 24420 }, { "epoch": 0.22, "learning_rate": 1.7836552069397006e-06, "loss": 0.8958, "step": 24421 }, { "epoch": 0.22, "learning_rate": 1.7830408464979144e-06, "loss": 0.8581, "step": 24422 }, { "epoch": 0.22, "learning_rate": 1.7824265815230613e-06, "loss": 0.8663, "step": 24423 }, { "epoch": 0.22, "learning_rate": 1.7818124120222712e-06, "loss": 0.8736, "step": 24424 }, { "epoch": 0.22, "learning_rate": 1.781198338002683e-06, "loss": 0.8583, "step": 24425 }, { "epoch": 0.22, "learning_rate": 1.78058435947143e-06, "loss": 0.8262, "step": 24426 }, { "epoch": 0.22, "learning_rate": 1.7799704764356485e-06, "loss": 0.8718, "step": 24427 }, { "epoch": 0.22, "learning_rate": 1.779356688902464e-06, "loss": 0.8758, "step": 24428 }, { "epoch": 0.22, "learning_rate": 1.7787429968790136e-06, "loss": 0.8521, "step": 24429 }, { "epoch": 0.22, "learning_rate": 1.778129400372427e-06, "loss": 0.8574, "step": 24430 }, { "epoch": 0.22, "learning_rate": 1.7775158993898335e-06, "loss": 0.8958, "step": 24431 }, { "epoch": 0.22, "learning_rate": 1.7769024939383561e-06, "loss": 0.8366, "step": 24432 }, { "epoch": 0.22, "learning_rate": 1.7762891840251262e-06, "loss": 0.8094, "step": 24433 }, { "epoch": 0.22, "learning_rate": 1.7756759696572668e-06, "loss": 0.8649, "step": 24434 }, { "epoch": 0.22, "learning_rate": 1.7750628508419042e-06, "loss": 0.832, "step": 24435 }, { "epoch": 0.22, "learning_rate": 1.7744498275861598e-06, "loss": 0.8618, "step": 24436 }, { "epoch": 0.22, "learning_rate": 1.7738368998971589e-06, "loss": 0.8367, "step": 24437 }, { "epoch": 0.22, "learning_rate": 1.7732240677820223e-06, "loss": 0.8368, "step": 24438 }, { "epoch": 0.22, "learning_rate": 1.7726113312478665e-06, "loss": 0.8681, "step": 24439 }, { "epoch": 0.22, "learning_rate": 1.7719986903018127e-06, "loss": 0.8639, "step": 24440 }, { "epoch": 0.22, "learning_rate": 1.7713861449509796e-06, "loss": 0.8669, "step": 24441 }, { "epoch": 0.22, "learning_rate": 1.7707736952024811e-06, "loss": 0.8453, "step": 24442 }, { "epoch": 0.22, "learning_rate": 1.7701613410634367e-06, "loss": 0.8348, "step": 24443 }, { "epoch": 0.22, "learning_rate": 1.7695490825409577e-06, "loss": 0.8529, "step": 24444 }, { "epoch": 0.22, "learning_rate": 1.7689369196421612e-06, "loss": 0.8509, "step": 24445 }, { "epoch": 0.22, "learning_rate": 1.7683248523741548e-06, "loss": 0.8261, "step": 24446 }, { "epoch": 0.22, "learning_rate": 1.7677128807440492e-06, "loss": 0.8554, "step": 24447 }, { "epoch": 0.22, "learning_rate": 1.767101004758962e-06, "loss": 0.8368, "step": 24448 }, { "epoch": 0.22, "learning_rate": 1.7664892244259958e-06, "loss": 0.856, "step": 24449 }, { "epoch": 0.22, "learning_rate": 1.7658775397522587e-06, "loss": 0.8408, "step": 24450 }, { "epoch": 0.22, "learning_rate": 1.7652659507448588e-06, "loss": 0.8263, "step": 24451 }, { "epoch": 0.22, "learning_rate": 1.7646544574109015e-06, "loss": 0.8599, "step": 24452 }, { "epoch": 0.22, "learning_rate": 1.764043059757492e-06, "loss": 0.8608, "step": 24453 }, { "epoch": 0.22, "learning_rate": 1.7634317577917325e-06, "loss": 0.8394, "step": 24454 }, { "epoch": 0.22, "learning_rate": 1.7628205515207275e-06, "loss": 0.8342, "step": 24455 }, { "epoch": 0.22, "learning_rate": 1.7622094409515754e-06, "loss": 0.878, "step": 24456 }, { "epoch": 0.22, "learning_rate": 1.7615984260913766e-06, "loss": 0.8587, "step": 24457 }, { "epoch": 0.22, "learning_rate": 1.7609875069472303e-06, "loss": 0.8345, "step": 24458 }, { "epoch": 0.22, "learning_rate": 1.7603766835262359e-06, "loss": 0.8819, "step": 24459 }, { "epoch": 0.22, "learning_rate": 1.7597659558354884e-06, "loss": 0.8768, "step": 24460 }, { "epoch": 0.22, "learning_rate": 1.7591553238820834e-06, "loss": 0.8672, "step": 24461 }, { "epoch": 0.22, "learning_rate": 1.7585447876731198e-06, "loss": 0.8519, "step": 24462 }, { "epoch": 0.22, "learning_rate": 1.757934347215684e-06, "loss": 0.84, "step": 24463 }, { "epoch": 0.22, "learning_rate": 1.7573240025168715e-06, "loss": 0.8497, "step": 24464 }, { "epoch": 0.22, "learning_rate": 1.7567137535837707e-06, "loss": 0.8664, "step": 24465 }, { "epoch": 0.22, "learning_rate": 1.75610360042348e-06, "loss": 0.8473, "step": 24466 }, { "epoch": 0.22, "learning_rate": 1.7554935430430786e-06, "loss": 0.8385, "step": 24467 }, { "epoch": 0.22, "learning_rate": 1.7548835814496601e-06, "loss": 0.8735, "step": 24468 }, { "epoch": 0.22, "learning_rate": 1.754273715650311e-06, "loss": 0.8396, "step": 24469 }, { "epoch": 0.22, "learning_rate": 1.753663945652113e-06, "loss": 0.8622, "step": 24470 }, { "epoch": 0.22, "learning_rate": 1.7530542714621512e-06, "loss": 0.8422, "step": 24471 }, { "epoch": 0.22, "learning_rate": 1.7524446930875127e-06, "loss": 0.863, "step": 24472 }, { "epoch": 0.22, "learning_rate": 1.7518352105352799e-06, "loss": 0.8821, "step": 24473 }, { "epoch": 0.22, "learning_rate": 1.7512258238125302e-06, "loss": 0.8296, "step": 24474 }, { "epoch": 0.22, "learning_rate": 1.7506165329263447e-06, "loss": 0.8506, "step": 24475 }, { "epoch": 0.22, "learning_rate": 1.7500073378838033e-06, "loss": 0.8419, "step": 24476 }, { "epoch": 0.22, "learning_rate": 1.7493982386919828e-06, "loss": 0.844, "step": 24477 }, { "epoch": 0.22, "learning_rate": 1.7487892353579605e-06, "loss": 0.8696, "step": 24478 }, { "epoch": 0.22, "learning_rate": 1.7481803278888122e-06, "loss": 0.871, "step": 24479 }, { "epoch": 0.22, "learning_rate": 1.747571516291614e-06, "loss": 0.8675, "step": 24480 }, { "epoch": 0.22, "learning_rate": 1.746962800573435e-06, "loss": 0.843, "step": 24481 }, { "epoch": 0.22, "learning_rate": 1.746354180741351e-06, "loss": 0.8442, "step": 24482 }, { "epoch": 0.22, "learning_rate": 1.74574565680243e-06, "loss": 0.8863, "step": 24483 }, { "epoch": 0.22, "learning_rate": 1.7451372287637446e-06, "loss": 0.8673, "step": 24484 }, { "epoch": 0.22, "learning_rate": 1.7445288966323637e-06, "loss": 0.8315, "step": 24485 }, { "epoch": 0.22, "learning_rate": 1.7439206604153535e-06, "loss": 0.8717, "step": 24486 }, { "epoch": 0.22, "learning_rate": 1.7433125201197842e-06, "loss": 0.8387, "step": 24487 }, { "epoch": 0.22, "learning_rate": 1.7427044757527157e-06, "loss": 0.8874, "step": 24488 }, { "epoch": 0.22, "learning_rate": 1.7420965273212155e-06, "loss": 0.8381, "step": 24489 }, { "epoch": 0.22, "learning_rate": 1.7414886748323445e-06, "loss": 0.855, "step": 24490 }, { "epoch": 0.22, "learning_rate": 1.7408809182931718e-06, "loss": 0.8642, "step": 24491 }, { "epoch": 0.22, "learning_rate": 1.7402732577107517e-06, "loss": 0.835, "step": 24492 }, { "epoch": 0.22, "learning_rate": 1.7396656930921475e-06, "loss": 0.8263, "step": 24493 }, { "epoch": 0.22, "learning_rate": 1.7390582244444177e-06, "loss": 0.8642, "step": 24494 }, { "epoch": 0.22, "learning_rate": 1.7384508517746147e-06, "loss": 0.7947, "step": 24495 }, { "epoch": 0.22, "learning_rate": 1.7378435750898015e-06, "loss": 0.9202, "step": 24496 }, { "epoch": 0.22, "learning_rate": 1.7372363943970327e-06, "loss": 0.8468, "step": 24497 }, { "epoch": 0.22, "learning_rate": 1.7366293097033638e-06, "loss": 0.8576, "step": 24498 }, { "epoch": 0.22, "learning_rate": 1.7360223210158446e-06, "loss": 0.8125, "step": 24499 }, { "epoch": 0.22, "learning_rate": 1.735415428341527e-06, "loss": 0.8648, "step": 24500 }, { "epoch": 0.0, "learning_rate": 1.7348086316874646e-06, "loss": 0.887, "step": 24501 }, { "epoch": 0.0, "learning_rate": 1.7342019310607062e-06, "loss": 0.8631, "step": 24502 }, { "epoch": 0.0, "learning_rate": 1.7335953264683014e-06, "loss": 0.8961, "step": 24503 }, { "epoch": 0.0, "learning_rate": 1.732988817917297e-06, "loss": 0.8537, "step": 24504 }, { "epoch": 0.0, "learning_rate": 1.732382405414742e-06, "loss": 0.8565, "step": 24505 }, { "epoch": 0.0, "learning_rate": 1.7317760889676781e-06, "loss": 0.8574, "step": 24506 }, { "epoch": 0.0, "learning_rate": 1.7311698685831513e-06, "loss": 0.8911, "step": 24507 }, { "epoch": 0.0, "learning_rate": 1.7305637442682056e-06, "loss": 0.9048, "step": 24508 }, { "epoch": 0.0, "learning_rate": 1.7299577160298809e-06, "loss": 0.8551, "step": 24509 }, { "epoch": 0.0, "learning_rate": 1.7293517838752205e-06, "loss": 0.8816, "step": 24510 }, { "epoch": 0.0, "learning_rate": 1.7287459478112644e-06, "loss": 0.8677, "step": 24511 }, { "epoch": 0.0, "learning_rate": 1.7281402078450516e-06, "loss": 0.8821, "step": 24512 }, { "epoch": 0.0, "learning_rate": 1.7275345639836171e-06, "loss": 0.8771, "step": 24513 }, { "epoch": 0.0, "learning_rate": 1.7269290162339958e-06, "loss": 0.8143, "step": 24514 }, { "epoch": 0.0, "learning_rate": 1.7263235646032296e-06, "loss": 0.9187, "step": 24515 }, { "epoch": 0.0, "learning_rate": 1.7257182090983515e-06, "loss": 0.8652, "step": 24516 }, { "epoch": 0.0, "learning_rate": 1.7251129497263908e-06, "loss": 0.8344, "step": 24517 }, { "epoch": 0.0, "learning_rate": 1.7245077864943816e-06, "loss": 0.8491, "step": 24518 }, { "epoch": 0.0, "learning_rate": 1.7239027194093572e-06, "loss": 0.8168, "step": 24519 }, { "epoch": 0.0, "learning_rate": 1.7232977484783398e-06, "loss": 0.8756, "step": 24520 }, { "epoch": 0.0, "learning_rate": 1.7226928737083649e-06, "loss": 0.8762, "step": 24521 }, { "epoch": 0.0, "learning_rate": 1.72208809510646e-06, "loss": 0.8945, "step": 24522 }, { "epoch": 0.0, "learning_rate": 1.721483412679652e-06, "loss": 0.891, "step": 24523 }, { "epoch": 0.0, "learning_rate": 1.7208788264349618e-06, "loss": 0.8856, "step": 24524 }, { "epoch": 0.0, "learning_rate": 1.720274336379416e-06, "loss": 0.8738, "step": 24525 }, { "epoch": 0.0, "learning_rate": 1.719669942520039e-06, "loss": 0.8591, "step": 24526 }, { "epoch": 0.0, "learning_rate": 1.7190656448638509e-06, "loss": 0.8595, "step": 24527 }, { "epoch": 0.0, "learning_rate": 1.7184614434178737e-06, "loss": 0.8839, "step": 24528 }, { "epoch": 0.0, "learning_rate": 1.7178573381891274e-06, "loss": 0.8613, "step": 24529 }, { "epoch": 0.0, "learning_rate": 1.7172533291846316e-06, "loss": 0.8633, "step": 24530 }, { "epoch": 0.0, "learning_rate": 1.716649416411401e-06, "loss": 0.8745, "step": 24531 }, { "epoch": 0.0, "learning_rate": 1.7160455998764537e-06, "loss": 0.8742, "step": 24532 }, { "epoch": 0.0, "learning_rate": 1.7154418795868044e-06, "loss": 0.8726, "step": 24533 }, { "epoch": 0.0, "learning_rate": 1.714838255549468e-06, "loss": 0.8908, "step": 24534 }, { "epoch": 0.0, "learning_rate": 1.7142347277714566e-06, "loss": 0.8774, "step": 24535 }, { "epoch": 0.0, "learning_rate": 1.7136312962597833e-06, "loss": 0.8474, "step": 24536 }, { "epoch": 0.0, "learning_rate": 1.7130279610214607e-06, "loss": 0.8253, "step": 24537 }, { "epoch": 0.0, "learning_rate": 1.7124247220634915e-06, "loss": 0.8702, "step": 24538 }, { "epoch": 0.0, "learning_rate": 1.7118215793928917e-06, "loss": 0.8643, "step": 24539 }, { "epoch": 0.0, "learning_rate": 1.7112185330166686e-06, "loss": 0.8874, "step": 24540 }, { "epoch": 0.0, "learning_rate": 1.7106155829418235e-06, "loss": 0.8558, "step": 24541 }, { "epoch": 0.0, "learning_rate": 1.7100127291753642e-06, "loss": 0.8531, "step": 24542 }, { "epoch": 0.0, "learning_rate": 1.709409971724295e-06, "loss": 0.8802, "step": 24543 }, { "epoch": 0.0, "learning_rate": 1.7088073105956193e-06, "loss": 0.859, "step": 24544 }, { "epoch": 0.0, "learning_rate": 1.708204745796338e-06, "loss": 0.8683, "step": 24545 }, { "epoch": 0.0, "learning_rate": 1.7076022773334522e-06, "loss": 0.8836, "step": 24546 }, { "epoch": 0.0, "learning_rate": 1.7069999052139653e-06, "loss": 0.8584, "step": 24547 }, { "epoch": 0.0, "learning_rate": 1.7063976294448681e-06, "loss": 0.8589, "step": 24548 }, { "epoch": 0.0, "learning_rate": 1.705795450033163e-06, "loss": 0.8247, "step": 24549 }, { "epoch": 0.0, "learning_rate": 1.7051933669858456e-06, "loss": 0.8484, "step": 24550 }, { "epoch": 0.0, "learning_rate": 1.7045913803099112e-06, "loss": 0.8971, "step": 24551 }, { "epoch": 0.0, "learning_rate": 1.7039894900123532e-06, "loss": 0.8676, "step": 24552 }, { "epoch": 0.0, "learning_rate": 1.7033876961001639e-06, "loss": 0.8936, "step": 24553 }, { "epoch": 0.0, "learning_rate": 1.7027859985803396e-06, "loss": 0.8546, "step": 24554 }, { "epoch": 0.0, "learning_rate": 1.702184397459865e-06, "loss": 0.8753, "step": 24555 }, { "epoch": 0.0, "learning_rate": 1.7015828927457324e-06, "loss": 0.8896, "step": 24556 }, { "epoch": 0.0, "learning_rate": 1.7009814844449268e-06, "loss": 0.8485, "step": 24557 }, { "epoch": 0.0, "learning_rate": 1.7003801725644432e-06, "loss": 0.8441, "step": 24558 }, { "epoch": 0.0, "learning_rate": 1.6997789571112622e-06, "loss": 0.8539, "step": 24559 }, { "epoch": 0.0, "learning_rate": 1.6991778380923696e-06, "loss": 0.876, "step": 24560 }, { "epoch": 0.0, "learning_rate": 1.6985768155147498e-06, "loss": 0.8468, "step": 24561 }, { "epoch": 0.0, "learning_rate": 1.697975889385387e-06, "loss": 0.8516, "step": 24562 }, { "epoch": 0.0, "learning_rate": 1.6973750597112581e-06, "loss": 0.8554, "step": 24563 }, { "epoch": 0.0, "learning_rate": 1.6967743264993486e-06, "loss": 0.8464, "step": 24564 }, { "epoch": 0.0, "learning_rate": 1.6961736897566405e-06, "loss": 0.8228, "step": 24565 }, { "epoch": 0.0, "learning_rate": 1.695573149490104e-06, "loss": 0.8223, "step": 24566 }, { "epoch": 0.0, "learning_rate": 1.6949727057067222e-06, "loss": 0.8532, "step": 24567 }, { "epoch": 0.0, "learning_rate": 1.6943723584134687e-06, "loss": 0.8526, "step": 24568 }, { "epoch": 0.0, "learning_rate": 1.6937721076173196e-06, "loss": 0.813, "step": 24569 }, { "epoch": 0.0, "learning_rate": 1.6931719533252477e-06, "loss": 0.8978, "step": 24570 }, { "epoch": 0.0, "learning_rate": 1.6925718955442283e-06, "loss": 0.87, "step": 24571 }, { "epoch": 0.0, "learning_rate": 1.6919719342812324e-06, "loss": 0.853, "step": 24572 }, { "epoch": 0.0, "learning_rate": 1.6913720695432279e-06, "loss": 0.822, "step": 24573 }, { "epoch": 0.0, "learning_rate": 1.6907723013371858e-06, "loss": 0.9059, "step": 24574 }, { "epoch": 0.0, "learning_rate": 1.6901726296700738e-06, "loss": 0.8837, "step": 24575 }, { "epoch": 0.0, "learning_rate": 1.6895730545488597e-06, "loss": 0.8543, "step": 24576 }, { "epoch": 0.0, "learning_rate": 1.6889735759805093e-06, "loss": 0.8209, "step": 24577 }, { "epoch": 0.0, "learning_rate": 1.6883741939719888e-06, "loss": 0.8933, "step": 24578 }, { "epoch": 0.0, "learning_rate": 1.6877749085302619e-06, "loss": 0.863, "step": 24579 }, { "epoch": 0.0, "learning_rate": 1.6871757196622885e-06, "loss": 0.8798, "step": 24580 }, { "epoch": 0.0, "learning_rate": 1.6865766273750318e-06, "loss": 0.8579, "step": 24581 }, { "epoch": 0.0, "learning_rate": 1.6859776316754506e-06, "loss": 0.8386, "step": 24582 }, { "epoch": 0.0, "learning_rate": 1.6853787325705096e-06, "loss": 0.9055, "step": 24583 }, { "epoch": 0.0, "learning_rate": 1.6847799300671608e-06, "loss": 0.8615, "step": 24584 }, { "epoch": 0.0, "learning_rate": 1.6841812241723642e-06, "loss": 0.834, "step": 24585 }, { "epoch": 0.0, "learning_rate": 1.6835826148930778e-06, "loss": 0.8977, "step": 24586 }, { "epoch": 0.0, "learning_rate": 1.6829841022362493e-06, "loss": 0.8841, "step": 24587 }, { "epoch": 0.0, "learning_rate": 1.6823856862088384e-06, "loss": 0.8946, "step": 24588 }, { "epoch": 0.0, "learning_rate": 1.6817873668177964e-06, "loss": 0.8638, "step": 24589 }, { "epoch": 0.0, "learning_rate": 1.681189144070078e-06, "loss": 0.8503, "step": 24590 }, { "epoch": 0.0, "learning_rate": 1.680591017972627e-06, "loss": 0.8652, "step": 24591 }, { "epoch": 0.0, "learning_rate": 1.6799929885323951e-06, "loss": 0.8724, "step": 24592 }, { "epoch": 0.0, "learning_rate": 1.6793950557563322e-06, "loss": 0.8387, "step": 24593 }, { "epoch": 0.0, "learning_rate": 1.6787972196513824e-06, "loss": 0.8876, "step": 24594 }, { "epoch": 0.0, "learning_rate": 1.678199480224494e-06, "loss": 0.8781, "step": 24595 }, { "epoch": 0.0, "learning_rate": 1.6776018374826108e-06, "loss": 0.8989, "step": 24596 }, { "epoch": 0.0, "learning_rate": 1.677004291432679e-06, "loss": 0.86, "step": 24597 }, { "epoch": 0.0, "learning_rate": 1.676406842081636e-06, "loss": 0.8834, "step": 24598 }, { "epoch": 0.0, "learning_rate": 1.6758094894364252e-06, "loss": 0.8553, "step": 24599 }, { "epoch": 0.0, "learning_rate": 1.6752122335039879e-06, "loss": 0.8722, "step": 24600 }, { "epoch": 0.0, "learning_rate": 1.6746150742912625e-06, "loss": 0.833, "step": 24601 }, { "epoch": 0.0, "learning_rate": 1.6740180118051863e-06, "loss": 0.8625, "step": 24602 }, { "epoch": 0.0, "learning_rate": 1.6734210460526967e-06, "loss": 0.8754, "step": 24603 }, { "epoch": 0.0, "learning_rate": 1.6728241770407328e-06, "loss": 0.8866, "step": 24604 }, { "epoch": 0.0, "learning_rate": 1.6722274047762222e-06, "loss": 0.8595, "step": 24605 }, { "epoch": 0.0, "learning_rate": 1.6716307292661005e-06, "loss": 0.8436, "step": 24606 }, { "epoch": 0.0, "learning_rate": 1.6710341505173044e-06, "loss": 0.8657, "step": 24607 }, { "epoch": 0.0, "learning_rate": 1.6704376685367641e-06, "loss": 0.8638, "step": 24608 }, { "epoch": 0.0, "learning_rate": 1.6698412833314048e-06, "loss": 0.8174, "step": 24609 }, { "epoch": 0.0, "learning_rate": 1.6692449949081601e-06, "loss": 0.87, "step": 24610 }, { "epoch": 0.0, "learning_rate": 1.6686488032739556e-06, "loss": 0.8979, "step": 24611 }, { "epoch": 0.0, "learning_rate": 1.668052708435719e-06, "loss": 0.894, "step": 24612 }, { "epoch": 0.0, "learning_rate": 1.6674567104003758e-06, "loss": 0.8685, "step": 24613 }, { "epoch": 0.0, "learning_rate": 1.6668608091748495e-06, "loss": 0.8839, "step": 24614 }, { "epoch": 0.0, "learning_rate": 1.6662650047660678e-06, "loss": 0.8795, "step": 24615 }, { "epoch": 0.0, "learning_rate": 1.6656692971809463e-06, "loss": 0.88, "step": 24616 }, { "epoch": 0.0, "learning_rate": 1.6650736864264082e-06, "loss": 0.83, "step": 24617 }, { "epoch": 0.0, "learning_rate": 1.664478172509376e-06, "loss": 0.8568, "step": 24618 }, { "epoch": 0.0, "learning_rate": 1.6638827554367665e-06, "loss": 0.8406, "step": 24619 }, { "epoch": 0.0, "learning_rate": 1.6632874352154982e-06, "loss": 0.8305, "step": 24620 }, { "epoch": 0.0, "learning_rate": 1.6626922118524868e-06, "loss": 0.8275, "step": 24621 }, { "epoch": 0.0, "learning_rate": 1.6620970853546515e-06, "loss": 0.8941, "step": 24622 }, { "epoch": 0.0, "learning_rate": 1.6615020557289007e-06, "loss": 0.8618, "step": 24623 }, { "epoch": 0.0, "learning_rate": 1.6609071229821517e-06, "loss": 0.867, "step": 24624 }, { "epoch": 0.0, "learning_rate": 1.660312287121314e-06, "loss": 0.8437, "step": 24625 }, { "epoch": 0.0, "learning_rate": 1.6597175481533001e-06, "loss": 0.8418, "step": 24626 }, { "epoch": 0.0, "learning_rate": 1.65912290608502e-06, "loss": 0.896, "step": 24627 }, { "epoch": 0.0, "learning_rate": 1.6585283609233816e-06, "loss": 0.858, "step": 24628 }, { "epoch": 0.0, "learning_rate": 1.6579339126752958e-06, "loss": 0.8611, "step": 24629 }, { "epoch": 0.0, "learning_rate": 1.6573395613476617e-06, "loss": 0.8428, "step": 24630 }, { "epoch": 0.0, "learning_rate": 1.6567453069473904e-06, "loss": 0.8717, "step": 24631 }, { "epoch": 0.0, "learning_rate": 1.6561511494813886e-06, "loss": 0.8689, "step": 24632 }, { "epoch": 0.0, "learning_rate": 1.655557088956553e-06, "loss": 0.87, "step": 24633 }, { "epoch": 0.0, "learning_rate": 1.6549631253797882e-06, "loss": 0.8857, "step": 24634 }, { "epoch": 0.0, "learning_rate": 1.6543692587579952e-06, "loss": 0.8576, "step": 24635 }, { "epoch": 0.0, "learning_rate": 1.653775489098073e-06, "loss": 0.8467, "step": 24636 }, { "epoch": 0.0, "learning_rate": 1.6531818164069214e-06, "loss": 0.8562, "step": 24637 }, { "epoch": 0.0, "learning_rate": 1.6525882406914372e-06, "loss": 0.8171, "step": 24638 }, { "epoch": 0.0, "learning_rate": 1.6519947619585185e-06, "loss": 0.8724, "step": 24639 }, { "epoch": 0.0, "learning_rate": 1.6514013802150575e-06, "loss": 0.9094, "step": 24640 }, { "epoch": 0.0, "learning_rate": 1.6508080954679484e-06, "loss": 0.8053, "step": 24641 }, { "epoch": 0.0, "learning_rate": 1.650214907724086e-06, "loss": 0.8451, "step": 24642 }, { "epoch": 0.0, "learning_rate": 1.649621816990361e-06, "loss": 0.8581, "step": 24643 }, { "epoch": 0.0, "learning_rate": 1.6490288232736652e-06, "loss": 0.8754, "step": 24644 }, { "epoch": 0.0, "learning_rate": 1.648435926580887e-06, "loss": 0.8096, "step": 24645 }, { "epoch": 0.0, "learning_rate": 1.6478431269189177e-06, "loss": 0.8467, "step": 24646 }, { "epoch": 0.0, "learning_rate": 1.6472504242946396e-06, "loss": 0.853, "step": 24647 }, { "epoch": 0.0, "learning_rate": 1.6466578187149419e-06, "loss": 0.879, "step": 24648 }, { "epoch": 0.0, "learning_rate": 1.6460653101867063e-06, "loss": 0.8544, "step": 24649 }, { "epoch": 0.0, "learning_rate": 1.6454728987168257e-06, "loss": 0.8656, "step": 24650 }, { "epoch": 0.01, "learning_rate": 1.644880584312173e-06, "loss": 0.8409, "step": 24651 }, { "epoch": 0.01, "learning_rate": 1.644288366979635e-06, "loss": 0.8649, "step": 24652 }, { "epoch": 0.01, "learning_rate": 1.6436962467260908e-06, "loss": 0.8634, "step": 24653 }, { "epoch": 0.01, "learning_rate": 1.6431042235584217e-06, "loss": 0.8519, "step": 24654 }, { "epoch": 0.01, "learning_rate": 1.6425122974834995e-06, "loss": 0.899, "step": 24655 }, { "epoch": 0.01, "learning_rate": 1.641920468508209e-06, "loss": 0.8678, "step": 24656 }, { "epoch": 0.01, "learning_rate": 1.641328736639426e-06, "loss": 0.8008, "step": 24657 }, { "epoch": 0.01, "learning_rate": 1.6407371018840212e-06, "loss": 0.8495, "step": 24658 }, { "epoch": 0.01, "learning_rate": 1.6401455642488683e-06, "loss": 0.8362, "step": 24659 }, { "epoch": 0.01, "learning_rate": 1.639554123740843e-06, "loss": 0.8728, "step": 24660 }, { "epoch": 0.01, "learning_rate": 1.638962780366815e-06, "loss": 0.8189, "step": 24661 }, { "epoch": 0.01, "learning_rate": 1.6383715341336547e-06, "loss": 0.855, "step": 24662 }, { "epoch": 0.01, "learning_rate": 1.637780385048232e-06, "loss": 0.8791, "step": 24663 }, { "epoch": 0.01, "learning_rate": 1.6371893331174182e-06, "loss": 0.8661, "step": 24664 }, { "epoch": 0.01, "learning_rate": 1.6365983783480744e-06, "loss": 0.8369, "step": 24665 }, { "epoch": 0.01, "learning_rate": 1.6360075207470683e-06, "loss": 0.8434, "step": 24666 }, { "epoch": 0.01, "learning_rate": 1.6354167603212657e-06, "loss": 0.8575, "step": 24667 }, { "epoch": 0.01, "learning_rate": 1.6348260970775299e-06, "loss": 0.8622, "step": 24668 }, { "epoch": 0.01, "learning_rate": 1.6342355310227231e-06, "loss": 0.8275, "step": 24669 }, { "epoch": 0.01, "learning_rate": 1.6336450621637079e-06, "loss": 0.8235, "step": 24670 }, { "epoch": 0.01, "learning_rate": 1.6330546905073453e-06, "loss": 0.8499, "step": 24671 }, { "epoch": 0.01, "learning_rate": 1.6324644160604908e-06, "loss": 0.8541, "step": 24672 }, { "epoch": 0.01, "learning_rate": 1.6318742388300014e-06, "loss": 0.8527, "step": 24673 }, { "epoch": 0.01, "learning_rate": 1.631284158822739e-06, "loss": 0.8903, "step": 24674 }, { "epoch": 0.01, "learning_rate": 1.6306941760455597e-06, "loss": 0.878, "step": 24675 }, { "epoch": 0.01, "learning_rate": 1.6301042905053133e-06, "loss": 0.8071, "step": 24676 }, { "epoch": 0.01, "learning_rate": 1.6295145022088554e-06, "loss": 0.8463, "step": 24677 }, { "epoch": 0.01, "learning_rate": 1.6289248111630419e-06, "loss": 0.9085, "step": 24678 }, { "epoch": 0.01, "learning_rate": 1.6283352173747148e-06, "loss": 0.8585, "step": 24679 }, { "epoch": 0.01, "learning_rate": 1.6277457208507319e-06, "loss": 0.8876, "step": 24680 }, { "epoch": 0.01, "learning_rate": 1.6271563215979402e-06, "loss": 0.8577, "step": 24681 }, { "epoch": 0.01, "learning_rate": 1.6265670196231898e-06, "loss": 0.8547, "step": 24682 }, { "epoch": 0.01, "learning_rate": 1.6259778149333227e-06, "loss": 0.8533, "step": 24683 }, { "epoch": 0.01, "learning_rate": 1.6253887075351871e-06, "loss": 0.8299, "step": 24684 }, { "epoch": 0.01, "learning_rate": 1.6247996974356273e-06, "loss": 0.8593, "step": 24685 }, { "epoch": 0.01, "learning_rate": 1.6242107846414856e-06, "loss": 0.8014, "step": 24686 }, { "epoch": 0.01, "learning_rate": 1.6236219691596045e-06, "loss": 0.8832, "step": 24687 }, { "epoch": 0.01, "learning_rate": 1.6230332509968262e-06, "loss": 0.8687, "step": 24688 }, { "epoch": 0.01, "learning_rate": 1.6224446301599927e-06, "loss": 0.8331, "step": 24689 }, { "epoch": 0.01, "learning_rate": 1.621856106655937e-06, "loss": 0.8531, "step": 24690 }, { "epoch": 0.01, "learning_rate": 1.6212676804914984e-06, "loss": 0.8713, "step": 24691 }, { "epoch": 0.01, "learning_rate": 1.6206793516735165e-06, "loss": 0.8815, "step": 24692 }, { "epoch": 0.01, "learning_rate": 1.6200911202088232e-06, "loss": 0.8384, "step": 24693 }, { "epoch": 0.01, "learning_rate": 1.6195029861042555e-06, "loss": 0.8673, "step": 24694 }, { "epoch": 0.01, "learning_rate": 1.6189149493666446e-06, "loss": 0.8662, "step": 24695 }, { "epoch": 0.01, "learning_rate": 1.6183270100028259e-06, "loss": 0.8707, "step": 24696 }, { "epoch": 0.01, "learning_rate": 1.6177391680196242e-06, "loss": 0.8305, "step": 24697 }, { "epoch": 0.01, "learning_rate": 1.6171514234238716e-06, "loss": 0.8312, "step": 24698 }, { "epoch": 0.01, "learning_rate": 1.6165637762223985e-06, "loss": 0.8611, "step": 24699 }, { "epoch": 0.01, "learning_rate": 1.6159762264220346e-06, "loss": 0.8736, "step": 24700 }, { "epoch": 0.01, "learning_rate": 1.6153887740296014e-06, "loss": 0.8321, "step": 24701 }, { "epoch": 0.01, "learning_rate": 1.6148014190519245e-06, "loss": 0.8818, "step": 24702 }, { "epoch": 0.01, "learning_rate": 1.6142141614958306e-06, "loss": 0.838, "step": 24703 }, { "epoch": 0.01, "learning_rate": 1.61362700136814e-06, "loss": 0.9074, "step": 24704 }, { "epoch": 0.01, "learning_rate": 1.6130399386756768e-06, "loss": 0.8255, "step": 24705 }, { "epoch": 0.01, "learning_rate": 1.6124529734252592e-06, "loss": 0.8738, "step": 24706 }, { "epoch": 0.01, "learning_rate": 1.6118661056237118e-06, "loss": 0.8535, "step": 24707 }, { "epoch": 0.01, "learning_rate": 1.6112793352778465e-06, "loss": 0.8682, "step": 24708 }, { "epoch": 0.01, "learning_rate": 1.6106926623944829e-06, "loss": 0.8344, "step": 24709 }, { "epoch": 0.01, "learning_rate": 1.6101060869804386e-06, "loss": 0.8228, "step": 24710 }, { "epoch": 0.01, "learning_rate": 1.6095196090425269e-06, "loss": 0.8617, "step": 24711 }, { "epoch": 0.01, "learning_rate": 1.6089332285875624e-06, "loss": 0.8651, "step": 24712 }, { "epoch": 0.01, "learning_rate": 1.6083469456223577e-06, "loss": 0.8673, "step": 24713 }, { "epoch": 0.01, "learning_rate": 1.6077607601537283e-06, "loss": 0.8352, "step": 24714 }, { "epoch": 0.01, "learning_rate": 1.6071746721884761e-06, "loss": 0.8583, "step": 24715 }, { "epoch": 0.01, "learning_rate": 1.6065886817334142e-06, "loss": 0.8793, "step": 24716 }, { "epoch": 0.01, "learning_rate": 1.6060027887953566e-06, "loss": 0.8456, "step": 24717 }, { "epoch": 0.01, "learning_rate": 1.6054169933811037e-06, "loss": 0.8437, "step": 24718 }, { "epoch": 0.01, "learning_rate": 1.604831295497462e-06, "loss": 0.8738, "step": 24719 }, { "epoch": 0.01, "learning_rate": 1.6042456951512387e-06, "loss": 0.8533, "step": 24720 }, { "epoch": 0.01, "learning_rate": 1.6036601923492378e-06, "loss": 0.8846, "step": 24721 }, { "epoch": 0.01, "learning_rate": 1.6030747870982565e-06, "loss": 0.8238, "step": 24722 }, { "epoch": 0.01, "learning_rate": 1.6024894794051015e-06, "loss": 0.8815, "step": 24723 }, { "epoch": 0.01, "learning_rate": 1.601904269276575e-06, "loss": 0.8703, "step": 24724 }, { "epoch": 0.01, "learning_rate": 1.6013191567194696e-06, "loss": 0.8553, "step": 24725 }, { "epoch": 0.01, "learning_rate": 1.6007341417405874e-06, "loss": 0.8754, "step": 24726 }, { "epoch": 0.01, "learning_rate": 1.6001492243467242e-06, "loss": 0.884, "step": 24727 }, { "epoch": 0.01, "learning_rate": 1.5995644045446756e-06, "loss": 0.8742, "step": 24728 }, { "epoch": 0.01, "learning_rate": 1.5989796823412362e-06, "loss": 0.8437, "step": 24729 }, { "epoch": 0.01, "learning_rate": 1.5983950577432007e-06, "loss": 0.828, "step": 24730 }, { "epoch": 0.01, "learning_rate": 1.597810530757362e-06, "loss": 0.8619, "step": 24731 }, { "epoch": 0.01, "learning_rate": 1.5972261013905087e-06, "loss": 0.8465, "step": 24732 }, { "epoch": 0.01, "learning_rate": 1.5966417696494306e-06, "loss": 0.8188, "step": 24733 }, { "epoch": 0.01, "learning_rate": 1.5960575355409191e-06, "loss": 0.8562, "step": 24734 }, { "epoch": 0.01, "learning_rate": 1.5954733990717598e-06, "loss": 0.8364, "step": 24735 }, { "epoch": 0.01, "learning_rate": 1.5948893602487415e-06, "loss": 0.8222, "step": 24736 }, { "epoch": 0.01, "learning_rate": 1.5943054190786478e-06, "loss": 0.8381, "step": 24737 }, { "epoch": 0.01, "learning_rate": 1.5937215755682667e-06, "loss": 0.8387, "step": 24738 }, { "epoch": 0.01, "learning_rate": 1.5931378297243772e-06, "loss": 0.863, "step": 24739 }, { "epoch": 0.01, "learning_rate": 1.5925541815537615e-06, "loss": 0.8618, "step": 24740 }, { "epoch": 0.01, "learning_rate": 1.591970631063201e-06, "loss": 0.8497, "step": 24741 }, { "epoch": 0.01, "learning_rate": 1.5913871782594803e-06, "loss": 0.8319, "step": 24742 }, { "epoch": 0.01, "learning_rate": 1.5908038231493729e-06, "loss": 0.839, "step": 24743 }, { "epoch": 0.01, "learning_rate": 1.5902205657396575e-06, "loss": 0.813, "step": 24744 }, { "epoch": 0.01, "learning_rate": 1.5896374060371112e-06, "loss": 0.8593, "step": 24745 }, { "epoch": 0.01, "learning_rate": 1.5890543440485085e-06, "loss": 0.8203, "step": 24746 }, { "epoch": 0.01, "learning_rate": 1.588471379780624e-06, "loss": 0.8432, "step": 24747 }, { "epoch": 0.01, "learning_rate": 1.587888513240231e-06, "loss": 0.851, "step": 24748 }, { "epoch": 0.01, "learning_rate": 1.5873057444341045e-06, "loss": 0.8127, "step": 24749 }, { "epoch": 0.01, "learning_rate": 1.5867230733690086e-06, "loss": 0.8262, "step": 24750 }, { "epoch": 0.01, "learning_rate": 1.5861405000517172e-06, "loss": 0.8808, "step": 24751 }, { "epoch": 0.01, "learning_rate": 1.585558024488998e-06, "loss": 0.8529, "step": 24752 }, { "epoch": 0.01, "learning_rate": 1.5849756466876186e-06, "loss": 0.8564, "step": 24753 }, { "epoch": 0.01, "learning_rate": 1.5843933666543443e-06, "loss": 0.8451, "step": 24754 }, { "epoch": 0.01, "learning_rate": 1.5838111843959414e-06, "loss": 0.8291, "step": 24755 }, { "epoch": 0.01, "learning_rate": 1.5832290999191747e-06, "loss": 0.7942, "step": 24756 }, { "epoch": 0.01, "learning_rate": 1.5826471132308042e-06, "loss": 0.8272, "step": 24757 }, { "epoch": 0.01, "learning_rate": 1.5820652243375934e-06, "loss": 0.8853, "step": 24758 }, { "epoch": 0.01, "learning_rate": 1.5814834332463015e-06, "loss": 0.8496, "step": 24759 }, { "epoch": 0.01, "learning_rate": 1.5809017399636895e-06, "loss": 0.8657, "step": 24760 }, { "epoch": 0.01, "learning_rate": 1.5803201444965144e-06, "loss": 0.8443, "step": 24761 }, { "epoch": 0.01, "learning_rate": 1.5797386468515352e-06, "loss": 0.8488, "step": 24762 }, { "epoch": 0.01, "learning_rate": 1.5791572470355077e-06, "loss": 0.8888, "step": 24763 }, { "epoch": 0.01, "learning_rate": 1.578575945055184e-06, "loss": 0.8857, "step": 24764 }, { "epoch": 0.01, "learning_rate": 1.577994740917318e-06, "loss": 0.8405, "step": 24765 }, { "epoch": 0.01, "learning_rate": 1.577413634628665e-06, "loss": 0.8432, "step": 24766 }, { "epoch": 0.01, "learning_rate": 1.5768326261959787e-06, "loss": 0.8574, "step": 24767 }, { "epoch": 0.01, "learning_rate": 1.5762517156260039e-06, "loss": 0.8468, "step": 24768 }, { "epoch": 0.01, "learning_rate": 1.5756709029254914e-06, "loss": 0.882, "step": 24769 }, { "epoch": 0.01, "learning_rate": 1.5750901881011916e-06, "loss": 0.8801, "step": 24770 }, { "epoch": 0.01, "learning_rate": 1.5745095711598457e-06, "loss": 0.8543, "step": 24771 }, { "epoch": 0.01, "learning_rate": 1.573929052108205e-06, "loss": 0.9249, "step": 24772 }, { "epoch": 0.01, "learning_rate": 1.5733486309530133e-06, "loss": 0.8414, "step": 24773 }, { "epoch": 0.01, "learning_rate": 1.5727683077010158e-06, "loss": 0.8491, "step": 24774 }, { "epoch": 0.01, "learning_rate": 1.5721880823589496e-06, "loss": 0.8595, "step": 24775 }, { "epoch": 0.01, "learning_rate": 1.5716079549335584e-06, "loss": 0.8843, "step": 24776 }, { "epoch": 0.01, "learning_rate": 1.571027925431583e-06, "loss": 0.8277, "step": 24777 }, { "epoch": 0.01, "learning_rate": 1.5704479938597617e-06, "loss": 0.8333, "step": 24778 }, { "epoch": 0.01, "learning_rate": 1.5698681602248323e-06, "loss": 0.8964, "step": 24779 }, { "epoch": 0.01, "learning_rate": 1.5692884245335326e-06, "loss": 0.8706, "step": 24780 }, { "epoch": 0.01, "learning_rate": 1.5687087867925988e-06, "loss": 0.8372, "step": 24781 }, { "epoch": 0.01, "learning_rate": 1.5681292470087617e-06, "loss": 0.8767, "step": 24782 }, { "epoch": 0.01, "learning_rate": 1.5675498051887562e-06, "loss": 0.9024, "step": 24783 }, { "epoch": 0.01, "learning_rate": 1.5669704613393145e-06, "loss": 0.863, "step": 24784 }, { "epoch": 0.01, "learning_rate": 1.566391215467169e-06, "loss": 0.8747, "step": 24785 }, { "epoch": 0.01, "learning_rate": 1.5658120675790478e-06, "loss": 0.8349, "step": 24786 }, { "epoch": 0.01, "learning_rate": 1.56523301768168e-06, "loss": 0.8331, "step": 24787 }, { "epoch": 0.01, "learning_rate": 1.5646540657817955e-06, "loss": 0.822, "step": 24788 }, { "epoch": 0.01, "learning_rate": 1.564075211886117e-06, "loss": 0.8467, "step": 24789 }, { "epoch": 0.01, "learning_rate": 1.5634964560013688e-06, "loss": 0.8832, "step": 24790 }, { "epoch": 0.01, "learning_rate": 1.5629177981342792e-06, "loss": 0.907, "step": 24791 }, { "epoch": 0.01, "learning_rate": 1.5623392382915726e-06, "loss": 0.8509, "step": 24792 }, { "epoch": 0.01, "learning_rate": 1.5617607764799658e-06, "loss": 0.8733, "step": 24793 }, { "epoch": 0.01, "learning_rate": 1.5611824127061814e-06, "loss": 0.844, "step": 24794 }, { "epoch": 0.01, "learning_rate": 1.5606041469769396e-06, "loss": 0.885, "step": 24795 }, { "epoch": 0.01, "learning_rate": 1.560025979298958e-06, "loss": 0.8322, "step": 24796 }, { "epoch": 0.01, "learning_rate": 1.559447909678954e-06, "loss": 0.8762, "step": 24797 }, { "epoch": 0.01, "learning_rate": 1.558869938123645e-06, "loss": 0.8549, "step": 24798 }, { "epoch": 0.01, "learning_rate": 1.558292064639747e-06, "loss": 0.8592, "step": 24799 }, { "epoch": 0.01, "learning_rate": 1.5577142892339702e-06, "loss": 0.837, "step": 24800 }, { "epoch": 0.01, "learning_rate": 1.5571366119130293e-06, "loss": 0.8324, "step": 24801 }, { "epoch": 0.01, "learning_rate": 1.5565590326836355e-06, "loss": 0.8852, "step": 24802 }, { "epoch": 0.01, "learning_rate": 1.5559815515525e-06, "loss": 0.8657, "step": 24803 }, { "epoch": 0.01, "learning_rate": 1.555404168526331e-06, "loss": 0.8895, "step": 24804 }, { "epoch": 0.01, "learning_rate": 1.5548268836118385e-06, "loss": 0.8293, "step": 24805 }, { "epoch": 0.01, "learning_rate": 1.5542496968157306e-06, "loss": 0.8626, "step": 24806 }, { "epoch": 0.01, "learning_rate": 1.5536726081447084e-06, "loss": 0.8654, "step": 24807 }, { "epoch": 0.01, "learning_rate": 1.5530956176054779e-06, "loss": 0.9013, "step": 24808 }, { "epoch": 0.01, "learning_rate": 1.552518725204748e-06, "loss": 0.8307, "step": 24809 }, { "epoch": 0.01, "learning_rate": 1.5519419309492156e-06, "loss": 0.8585, "step": 24810 }, { "epoch": 0.01, "learning_rate": 1.5513652348455843e-06, "loss": 0.8272, "step": 24811 }, { "epoch": 0.01, "learning_rate": 1.550788636900553e-06, "loss": 0.8495, "step": 24812 }, { "epoch": 0.01, "learning_rate": 1.5502121371208257e-06, "loss": 0.8434, "step": 24813 }, { "epoch": 0.01, "learning_rate": 1.5496357355130897e-06, "loss": 0.859, "step": 24814 }, { "epoch": 0.01, "learning_rate": 1.5490594320840525e-06, "loss": 0.8489, "step": 24815 }, { "epoch": 0.01, "learning_rate": 1.548483226840407e-06, "loss": 0.8369, "step": 24816 }, { "epoch": 0.01, "learning_rate": 1.547907119788844e-06, "loss": 0.8536, "step": 24817 }, { "epoch": 0.01, "learning_rate": 1.5473311109360601e-06, "loss": 0.8489, "step": 24818 }, { "epoch": 0.01, "learning_rate": 1.5467552002887464e-06, "loss": 0.8294, "step": 24819 }, { "epoch": 0.01, "learning_rate": 1.5461793878535936e-06, "loss": 0.8541, "step": 24820 }, { "epoch": 0.01, "learning_rate": 1.5456036736372938e-06, "loss": 0.8147, "step": 24821 }, { "epoch": 0.01, "learning_rate": 1.5450280576465337e-06, "loss": 0.8672, "step": 24822 }, { "epoch": 0.01, "learning_rate": 1.544452539888004e-06, "loss": 0.895, "step": 24823 }, { "epoch": 0.01, "learning_rate": 1.5438771203683856e-06, "loss": 0.8601, "step": 24824 }, { "epoch": 0.01, "learning_rate": 1.5433017990943677e-06, "loss": 0.8448, "step": 24825 }, { "epoch": 0.01, "learning_rate": 1.542726576072635e-06, "loss": 0.8773, "step": 24826 }, { "epoch": 0.01, "learning_rate": 1.542151451309869e-06, "loss": 0.8455, "step": 24827 }, { "epoch": 0.01, "learning_rate": 1.5415764248127518e-06, "loss": 0.8351, "step": 24828 }, { "epoch": 0.01, "learning_rate": 1.541001496587965e-06, "loss": 0.8272, "step": 24829 }, { "epoch": 0.01, "learning_rate": 1.5404266666421908e-06, "loss": 0.8543, "step": 24830 }, { "epoch": 0.01, "learning_rate": 1.5398519349821017e-06, "loss": 0.8459, "step": 24831 }, { "epoch": 0.01, "learning_rate": 1.5392773016143781e-06, "loss": 0.872, "step": 24832 }, { "epoch": 0.01, "learning_rate": 1.5387027665456945e-06, "loss": 0.8395, "step": 24833 }, { "epoch": 0.01, "learning_rate": 1.538128329782733e-06, "loss": 0.8758, "step": 24834 }, { "epoch": 0.01, "learning_rate": 1.5375539913321601e-06, "loss": 0.8664, "step": 24835 }, { "epoch": 0.01, "learning_rate": 1.5369797512006513e-06, "loss": 0.8432, "step": 24836 }, { "epoch": 0.01, "learning_rate": 1.5364056093948776e-06, "loss": 0.8149, "step": 24837 }, { "epoch": 0.01, "learning_rate": 1.5358315659215094e-06, "loss": 0.8531, "step": 24838 }, { "epoch": 0.01, "learning_rate": 1.5352576207872171e-06, "loss": 0.8501, "step": 24839 }, { "epoch": 0.01, "learning_rate": 1.5346837739986686e-06, "loss": 0.8406, "step": 24840 }, { "epoch": 0.01, "learning_rate": 1.5341100255625342e-06, "loss": 0.8405, "step": 24841 }, { "epoch": 0.01, "learning_rate": 1.5335363754854726e-06, "loss": 0.8658, "step": 24842 }, { "epoch": 0.01, "learning_rate": 1.5329628237741535e-06, "loss": 0.8662, "step": 24843 }, { "epoch": 0.01, "learning_rate": 1.53238937043524e-06, "loss": 0.8361, "step": 24844 }, { "epoch": 0.01, "learning_rate": 1.5318160154753937e-06, "loss": 0.8088, "step": 24845 }, { "epoch": 0.01, "learning_rate": 1.5312427589012769e-06, "loss": 0.8336, "step": 24846 }, { "epoch": 0.01, "learning_rate": 1.53066960071955e-06, "loss": 0.8683, "step": 24847 }, { "epoch": 0.01, "learning_rate": 1.5300965409368728e-06, "loss": 0.8443, "step": 24848 }, { "epoch": 0.01, "learning_rate": 1.5295235795599006e-06, "loss": 0.841, "step": 24849 }, { "epoch": 0.01, "learning_rate": 1.528950716595292e-06, "loss": 0.8501, "step": 24850 }, { "epoch": 0.01, "learning_rate": 1.5283779520497e-06, "loss": 0.8741, "step": 24851 }, { "epoch": 0.01, "learning_rate": 1.5278052859297866e-06, "loss": 0.8785, "step": 24852 }, { "epoch": 0.01, "learning_rate": 1.5272327182421975e-06, "loss": 0.8645, "step": 24853 }, { "epoch": 0.01, "learning_rate": 1.5266602489935878e-06, "loss": 0.8516, "step": 24854 }, { "epoch": 0.01, "learning_rate": 1.5260878781906107e-06, "loss": 0.8936, "step": 24855 }, { "epoch": 0.01, "learning_rate": 1.5255156058399124e-06, "loss": 0.885, "step": 24856 }, { "epoch": 0.01, "learning_rate": 1.5249434319481404e-06, "loss": 0.8973, "step": 24857 }, { "epoch": 0.01, "learning_rate": 1.5243713565219477e-06, "loss": 0.9063, "step": 24858 }, { "epoch": 0.01, "learning_rate": 1.5237993795679795e-06, "loss": 0.842, "step": 24859 }, { "epoch": 0.01, "learning_rate": 1.5232275010928789e-06, "loss": 0.8629, "step": 24860 }, { "epoch": 0.01, "learning_rate": 1.522655721103291e-06, "loss": 0.8818, "step": 24861 }, { "epoch": 0.01, "learning_rate": 1.5220840396058624e-06, "loss": 0.8012, "step": 24862 }, { "epoch": 0.01, "learning_rate": 1.5215124566072259e-06, "loss": 0.8399, "step": 24863 }, { "epoch": 0.01, "learning_rate": 1.520940972114031e-06, "loss": 0.9126, "step": 24864 }, { "epoch": 0.01, "learning_rate": 1.5203695861329148e-06, "loss": 0.8412, "step": 24865 }, { "epoch": 0.01, "learning_rate": 1.5197982986705185e-06, "loss": 0.8859, "step": 24866 }, { "epoch": 0.01, "learning_rate": 1.5192271097334731e-06, "loss": 0.8628, "step": 24867 }, { "epoch": 0.01, "learning_rate": 1.5186560193284194e-06, "loss": 0.8724, "step": 24868 }, { "epoch": 0.01, "learning_rate": 1.5180850274619907e-06, "loss": 0.882, "step": 24869 }, { "epoch": 0.01, "learning_rate": 1.5175141341408227e-06, "loss": 0.8745, "step": 24870 }, { "epoch": 0.01, "learning_rate": 1.516943339371546e-06, "loss": 0.8957, "step": 24871 }, { "epoch": 0.01, "learning_rate": 1.516372643160795e-06, "loss": 0.8577, "step": 24872 }, { "epoch": 0.01, "learning_rate": 1.5158020455152e-06, "loss": 0.8278, "step": 24873 }, { "epoch": 0.01, "learning_rate": 1.515231546441387e-06, "loss": 0.8571, "step": 24874 }, { "epoch": 0.01, "learning_rate": 1.514661145945986e-06, "loss": 0.8536, "step": 24875 }, { "epoch": 0.01, "learning_rate": 1.514090844035624e-06, "loss": 0.8621, "step": 24876 }, { "epoch": 0.01, "learning_rate": 1.5135206407169277e-06, "loss": 0.8767, "step": 24877 }, { "epoch": 0.01, "learning_rate": 1.5129505359965213e-06, "loss": 0.8645, "step": 24878 }, { "epoch": 0.01, "learning_rate": 1.5123805298810279e-06, "loss": 0.9088, "step": 24879 }, { "epoch": 0.01, "learning_rate": 1.5118106223770746e-06, "loss": 0.896, "step": 24880 }, { "epoch": 0.01, "learning_rate": 1.5112408134912725e-06, "loss": 0.8397, "step": 24881 }, { "epoch": 0.01, "learning_rate": 1.5106711032302512e-06, "loss": 0.8356, "step": 24882 }, { "epoch": 0.01, "learning_rate": 1.5101014916006262e-06, "loss": 0.8273, "step": 24883 }, { "epoch": 0.01, "learning_rate": 1.509531978609019e-06, "loss": 0.8543, "step": 24884 }, { "epoch": 0.01, "learning_rate": 1.5089625642620409e-06, "loss": 0.8527, "step": 24885 }, { "epoch": 0.01, "learning_rate": 1.508393248566309e-06, "loss": 0.8385, "step": 24886 }, { "epoch": 0.01, "learning_rate": 1.5078240315284387e-06, "loss": 0.8213, "step": 24887 }, { "epoch": 0.01, "learning_rate": 1.507254913155044e-06, "loss": 0.8947, "step": 24888 }, { "epoch": 0.01, "learning_rate": 1.5066858934527352e-06, "loss": 0.8666, "step": 24889 }, { "epoch": 0.01, "learning_rate": 1.5061169724281254e-06, "loss": 0.8622, "step": 24890 }, { "epoch": 0.01, "learning_rate": 1.5055481500878254e-06, "loss": 0.8583, "step": 24891 }, { "epoch": 0.01, "learning_rate": 1.5049794264384398e-06, "loss": 0.8822, "step": 24892 }, { "epoch": 0.01, "learning_rate": 1.5044108014865788e-06, "loss": 0.8322, "step": 24893 }, { "epoch": 0.01, "learning_rate": 1.503842275238848e-06, "loss": 0.8389, "step": 24894 }, { "epoch": 0.01, "learning_rate": 1.5032738477018527e-06, "loss": 0.8499, "step": 24895 }, { "epoch": 0.01, "learning_rate": 1.502705518882198e-06, "loss": 0.868, "step": 24896 }, { "epoch": 0.01, "learning_rate": 1.5021372887864872e-06, "loss": 0.8263, "step": 24897 }, { "epoch": 0.01, "learning_rate": 1.501569157421323e-06, "loss": 0.8229, "step": 24898 }, { "epoch": 0.01, "learning_rate": 1.501001124793302e-06, "loss": 0.8395, "step": 24899 }, { "epoch": 0.01, "learning_rate": 1.500433190909024e-06, "loss": 0.8636, "step": 24900 }, { "epoch": 0.01, "learning_rate": 1.4998653557750953e-06, "loss": 0.8939, "step": 24901 }, { "epoch": 0.01, "learning_rate": 1.4992976193981046e-06, "loss": 0.8511, "step": 24902 }, { "epoch": 0.01, "learning_rate": 1.4987299817846502e-06, "loss": 0.8889, "step": 24903 }, { "epoch": 0.01, "learning_rate": 1.498162442941329e-06, "loss": 0.8764, "step": 24904 }, { "epoch": 0.01, "learning_rate": 1.4975950028747355e-06, "loss": 0.8312, "step": 24905 }, { "epoch": 0.01, "learning_rate": 1.4970276615914559e-06, "loss": 0.8724, "step": 24906 }, { "epoch": 0.01, "learning_rate": 1.496460419098088e-06, "loss": 0.8317, "step": 24907 }, { "epoch": 0.01, "learning_rate": 1.4958932754012234e-06, "loss": 0.8662, "step": 24908 }, { "epoch": 0.01, "learning_rate": 1.4953262305074468e-06, "loss": 0.8253, "step": 24909 }, { "epoch": 0.01, "learning_rate": 1.4947592844233472e-06, "loss": 0.8392, "step": 24910 }, { "epoch": 0.01, "learning_rate": 1.4941924371555127e-06, "loss": 0.856, "step": 24911 }, { "epoch": 0.01, "learning_rate": 1.4936256887105283e-06, "loss": 0.848, "step": 24912 }, { "epoch": 0.01, "learning_rate": 1.4930590390949796e-06, "loss": 0.8534, "step": 24913 }, { "epoch": 0.01, "learning_rate": 1.492492488315449e-06, "loss": 0.8361, "step": 24914 }, { "epoch": 0.01, "learning_rate": 1.4919260363785215e-06, "loss": 0.8824, "step": 24915 }, { "epoch": 0.01, "learning_rate": 1.491359683290775e-06, "loss": 0.8834, "step": 24916 }, { "epoch": 0.01, "learning_rate": 1.49079342905879e-06, "loss": 0.8053, "step": 24917 }, { "epoch": 0.01, "learning_rate": 1.4902272736891455e-06, "loss": 0.8704, "step": 24918 }, { "epoch": 0.01, "learning_rate": 1.4896612171884205e-06, "loss": 0.8397, "step": 24919 }, { "epoch": 0.01, "learning_rate": 1.4890952595631913e-06, "loss": 0.8579, "step": 24920 }, { "epoch": 0.01, "learning_rate": 1.4885294008200324e-06, "loss": 0.8253, "step": 24921 }, { "epoch": 0.01, "learning_rate": 1.4879636409655219e-06, "loss": 0.8744, "step": 24922 }, { "epoch": 0.01, "learning_rate": 1.4873979800062266e-06, "loss": 0.831, "step": 24923 }, { "epoch": 0.01, "learning_rate": 1.4868324179487204e-06, "loss": 0.8242, "step": 24924 }, { "epoch": 0.01, "learning_rate": 1.4862669547995767e-06, "loss": 0.8518, "step": 24925 }, { "epoch": 0.01, "learning_rate": 1.485701590565367e-06, "loss": 0.8645, "step": 24926 }, { "epoch": 0.01, "learning_rate": 1.485136325252654e-06, "loss": 0.8408, "step": 24927 }, { "epoch": 0.01, "learning_rate": 1.484571158868008e-06, "loss": 0.8818, "step": 24928 }, { "epoch": 0.01, "learning_rate": 1.4840060914179954e-06, "loss": 0.8182, "step": 24929 }, { "epoch": 0.01, "learning_rate": 1.4834411229091816e-06, "loss": 0.889, "step": 24930 }, { "epoch": 0.01, "learning_rate": 1.4828762533481288e-06, "loss": 0.8597, "step": 24931 }, { "epoch": 0.01, "learning_rate": 1.482311482741402e-06, "loss": 0.8726, "step": 24932 }, { "epoch": 0.01, "learning_rate": 1.4817468110955646e-06, "loss": 0.8059, "step": 24933 }, { "epoch": 0.01, "learning_rate": 1.4811822384171714e-06, "loss": 0.8143, "step": 24934 }, { "epoch": 0.01, "learning_rate": 1.4806177647127851e-06, "loss": 0.8734, "step": 24935 }, { "epoch": 0.01, "learning_rate": 1.4800533899889635e-06, "loss": 0.8626, "step": 24936 }, { "epoch": 0.01, "learning_rate": 1.4794891142522639e-06, "loss": 0.8103, "step": 24937 }, { "epoch": 0.01, "learning_rate": 1.4789249375092418e-06, "loss": 0.8764, "step": 24938 }, { "epoch": 0.01, "learning_rate": 1.4783608597664533e-06, "loss": 0.8856, "step": 24939 }, { "epoch": 0.01, "learning_rate": 1.4777968810304522e-06, "loss": 0.8531, "step": 24940 }, { "epoch": 0.01, "learning_rate": 1.4772330013077874e-06, "loss": 0.8538, "step": 24941 }, { "epoch": 0.01, "learning_rate": 1.4766692206050137e-06, "loss": 0.8381, "step": 24942 }, { "epoch": 0.01, "learning_rate": 1.476105538928677e-06, "loss": 0.8717, "step": 24943 }, { "epoch": 0.01, "learning_rate": 1.4755419562853345e-06, "loss": 0.8508, "step": 24944 }, { "epoch": 0.01, "learning_rate": 1.4749784726815264e-06, "loss": 0.8471, "step": 24945 }, { "epoch": 0.01, "learning_rate": 1.4744150881238018e-06, "loss": 0.8513, "step": 24946 }, { "epoch": 0.01, "learning_rate": 1.4738518026187087e-06, "loss": 0.895, "step": 24947 }, { "epoch": 0.01, "learning_rate": 1.4732886161727877e-06, "loss": 0.8862, "step": 24948 }, { "epoch": 0.01, "learning_rate": 1.4727255287925801e-06, "loss": 0.8262, "step": 24949 }, { "epoch": 0.01, "learning_rate": 1.4721625404846351e-06, "loss": 0.8193, "step": 24950 }, { "epoch": 0.02, "learning_rate": 1.471599651255492e-06, "loss": 0.8957, "step": 24951 }, { "epoch": 0.02, "learning_rate": 1.4710368611116866e-06, "loss": 0.8794, "step": 24952 }, { "epoch": 0.02, "learning_rate": 1.470474170059759e-06, "loss": 0.8339, "step": 24953 }, { "epoch": 0.02, "learning_rate": 1.4699115781062478e-06, "loss": 0.8279, "step": 24954 }, { "epoch": 0.02, "learning_rate": 1.4693490852576885e-06, "loss": 0.859, "step": 24955 }, { "epoch": 0.02, "learning_rate": 1.4687866915206173e-06, "loss": 0.8598, "step": 24956 }, { "epoch": 0.02, "learning_rate": 1.4682243969015674e-06, "loss": 0.8373, "step": 24957 }, { "epoch": 0.02, "learning_rate": 1.467662201407074e-06, "loss": 0.8476, "step": 24958 }, { "epoch": 0.02, "learning_rate": 1.4671001050436651e-06, "loss": 0.8943, "step": 24959 }, { "epoch": 0.02, "learning_rate": 1.4665381078178731e-06, "loss": 0.8433, "step": 24960 }, { "epoch": 0.02, "learning_rate": 1.4659762097362262e-06, "loss": 0.8675, "step": 24961 }, { "epoch": 0.02, "learning_rate": 1.465414410805255e-06, "loss": 0.8872, "step": 24962 }, { "epoch": 0.02, "learning_rate": 1.4648527110314847e-06, "loss": 0.8165, "step": 24963 }, { "epoch": 0.02, "learning_rate": 1.464291110421442e-06, "loss": 0.8489, "step": 24964 }, { "epoch": 0.02, "learning_rate": 1.4637296089816543e-06, "loss": 0.832, "step": 24965 }, { "epoch": 0.02, "learning_rate": 1.4631682067186404e-06, "loss": 0.86, "step": 24966 }, { "epoch": 0.02, "learning_rate": 1.4626069036389256e-06, "loss": 0.8879, "step": 24967 }, { "epoch": 0.02, "learning_rate": 1.4620456997490306e-06, "loss": 0.8828, "step": 24968 }, { "epoch": 0.02, "learning_rate": 1.4614845950554756e-06, "loss": 0.8439, "step": 24969 }, { "epoch": 0.02, "learning_rate": 1.46092358956478e-06, "loss": 0.9202, "step": 24970 }, { "epoch": 0.02, "learning_rate": 1.4603626832834628e-06, "loss": 0.8655, "step": 24971 }, { "epoch": 0.02, "learning_rate": 1.4598018762180411e-06, "loss": 0.8322, "step": 24972 }, { "epoch": 0.02, "learning_rate": 1.459241168375024e-06, "loss": 0.8416, "step": 24973 }, { "epoch": 0.02, "learning_rate": 1.4586805597609333e-06, "loss": 0.903, "step": 24974 }, { "epoch": 0.02, "learning_rate": 1.4581200503822801e-06, "loss": 0.866, "step": 24975 }, { "epoch": 0.02, "learning_rate": 1.4575596402455794e-06, "loss": 0.8785, "step": 24976 }, { "epoch": 0.02, "learning_rate": 1.456999329357337e-06, "loss": 0.8787, "step": 24977 }, { "epoch": 0.02, "learning_rate": 1.4564391177240644e-06, "loss": 0.8398, "step": 24978 }, { "epoch": 0.02, "learning_rate": 1.455879005352271e-06, "loss": 0.8751, "step": 24979 }, { "epoch": 0.02, "learning_rate": 1.4553189922484645e-06, "loss": 0.8608, "step": 24980 }, { "epoch": 0.02, "learning_rate": 1.4547590784191511e-06, "loss": 0.8419, "step": 24981 }, { "epoch": 0.02, "learning_rate": 1.4541992638708358e-06, "loss": 0.8263, "step": 24982 }, { "epoch": 0.02, "learning_rate": 1.4536395486100251e-06, "loss": 0.8733, "step": 24983 }, { "epoch": 0.02, "learning_rate": 1.4530799326432177e-06, "loss": 0.8402, "step": 24984 }, { "epoch": 0.02, "learning_rate": 1.452520415976918e-06, "loss": 0.8435, "step": 24985 }, { "epoch": 0.02, "learning_rate": 1.4519609986176253e-06, "loss": 0.8553, "step": 24986 }, { "epoch": 0.02, "learning_rate": 1.4514016805718389e-06, "loss": 0.8914, "step": 24987 }, { "epoch": 0.02, "learning_rate": 1.4508424618460592e-06, "loss": 0.8816, "step": 24988 }, { "epoch": 0.02, "learning_rate": 1.4502833424467811e-06, "loss": 0.8623, "step": 24989 }, { "epoch": 0.02, "learning_rate": 1.4497243223805046e-06, "loss": 0.8965, "step": 24990 }, { "epoch": 0.02, "learning_rate": 1.4491654016537193e-06, "loss": 0.8566, "step": 24991 }, { "epoch": 0.02, "learning_rate": 1.4486065802729188e-06, "loss": 0.8214, "step": 24992 }, { "epoch": 0.02, "learning_rate": 1.4480478582446033e-06, "loss": 0.8391, "step": 24993 }, { "epoch": 0.02, "learning_rate": 1.4474892355752556e-06, "loss": 0.8362, "step": 24994 }, { "epoch": 0.02, "learning_rate": 1.4469307122713693e-06, "loss": 0.8915, "step": 24995 }, { "epoch": 0.02, "learning_rate": 1.4463722883394337e-06, "loss": 0.8505, "step": 24996 }, { "epoch": 0.02, "learning_rate": 1.445813963785937e-06, "loss": 0.8299, "step": 24997 }, { "epoch": 0.02, "learning_rate": 1.445255738617365e-06, "loss": 0.8219, "step": 24998 }, { "epoch": 0.02, "learning_rate": 1.4446976128402035e-06, "loss": 0.8534, "step": 24999 }, { "epoch": 0.02, "learning_rate": 1.4441395864609398e-06, "loss": 0.867, "step": 25000 }, { "epoch": 0.02, "learning_rate": 1.443581659486052e-06, "loss": 0.8133, "step": 25001 }, { "epoch": 0.02, "learning_rate": 1.4430238319220247e-06, "loss": 0.8276, "step": 25002 }, { "epoch": 0.02, "learning_rate": 1.4424661037753396e-06, "loss": 0.8632, "step": 25003 }, { "epoch": 0.02, "learning_rate": 1.4419084750524759e-06, "loss": 0.8455, "step": 25004 }, { "epoch": 0.02, "learning_rate": 1.4413509457599128e-06, "loss": 0.8592, "step": 25005 }, { "epoch": 0.02, "learning_rate": 1.4407935159041265e-06, "loss": 0.8977, "step": 25006 }, { "epoch": 0.02, "learning_rate": 1.440236185491597e-06, "loss": 0.8624, "step": 25007 }, { "epoch": 0.02, "learning_rate": 1.4396789545287949e-06, "loss": 0.8558, "step": 25008 }, { "epoch": 0.02, "learning_rate": 1.4391218230221958e-06, "loss": 0.8885, "step": 25009 }, { "epoch": 0.02, "learning_rate": 1.438564790978273e-06, "loss": 0.861, "step": 25010 }, { "epoch": 0.02, "learning_rate": 1.4380078584034985e-06, "loss": 0.8504, "step": 25011 }, { "epoch": 0.02, "learning_rate": 1.4374510253043416e-06, "loss": 0.8595, "step": 25012 }, { "epoch": 0.02, "learning_rate": 1.4368942916872742e-06, "loss": 0.811, "step": 25013 }, { "epoch": 0.02, "learning_rate": 1.436337657558765e-06, "loss": 0.8575, "step": 25014 }, { "epoch": 0.02, "learning_rate": 1.4357811229252772e-06, "loss": 0.8643, "step": 25015 }, { "epoch": 0.02, "learning_rate": 1.4352246877932753e-06, "loss": 0.8799, "step": 25016 }, { "epoch": 0.02, "learning_rate": 1.4346683521692316e-06, "loss": 0.8349, "step": 25017 }, { "epoch": 0.02, "learning_rate": 1.4341121160596073e-06, "loss": 0.8868, "step": 25018 }, { "epoch": 0.02, "learning_rate": 1.433555979470861e-06, "loss": 0.8751, "step": 25019 }, { "epoch": 0.02, "learning_rate": 1.4329999424094575e-06, "loss": 0.8288, "step": 25020 }, { "epoch": 0.02, "learning_rate": 1.4324440048818555e-06, "loss": 0.8554, "step": 25021 }, { "epoch": 0.02, "learning_rate": 1.4318881668945139e-06, "loss": 0.8499, "step": 25022 }, { "epoch": 0.02, "learning_rate": 1.431332428453892e-06, "loss": 0.8593, "step": 25023 }, { "epoch": 0.02, "learning_rate": 1.430776789566446e-06, "loss": 0.8479, "step": 25024 }, { "epoch": 0.02, "learning_rate": 1.4302212502386326e-06, "loss": 0.8715, "step": 25025 }, { "epoch": 0.02, "learning_rate": 1.4296658104769034e-06, "loss": 0.8392, "step": 25026 }, { "epoch": 0.02, "learning_rate": 1.4291104702877124e-06, "loss": 0.8506, "step": 25027 }, { "epoch": 0.02, "learning_rate": 1.428555229677513e-06, "loss": 0.8419, "step": 25028 }, { "epoch": 0.02, "learning_rate": 1.4280000886527545e-06, "loss": 0.8197, "step": 25029 }, { "epoch": 0.02, "learning_rate": 1.4274450472198886e-06, "loss": 0.8556, "step": 25030 }, { "epoch": 0.02, "learning_rate": 1.4268901053853634e-06, "loss": 0.8337, "step": 25031 }, { "epoch": 0.02, "learning_rate": 1.4263352631556272e-06, "loss": 0.8679, "step": 25032 }, { "epoch": 0.02, "learning_rate": 1.4257805205371233e-06, "loss": 0.8716, "step": 25033 }, { "epoch": 0.02, "learning_rate": 1.4252258775362992e-06, "loss": 0.8554, "step": 25034 }, { "epoch": 0.02, "learning_rate": 1.4246713341595953e-06, "loss": 0.8578, "step": 25035 }, { "epoch": 0.02, "learning_rate": 1.4241168904134627e-06, "loss": 0.8717, "step": 25036 }, { "epoch": 0.02, "learning_rate": 1.4235625463043357e-06, "loss": 0.8213, "step": 25037 }, { "epoch": 0.02, "learning_rate": 1.4230083018386565e-06, "loss": 0.8386, "step": 25038 }, { "epoch": 0.02, "learning_rate": 1.4224541570228677e-06, "loss": 0.8533, "step": 25039 }, { "epoch": 0.02, "learning_rate": 1.4219001118634025e-06, "loss": 0.8446, "step": 25040 }, { "epoch": 0.02, "learning_rate": 1.421346166366697e-06, "loss": 0.8283, "step": 25041 }, { "epoch": 0.02, "learning_rate": 1.4207923205391938e-06, "loss": 0.8797, "step": 25042 }, { "epoch": 0.02, "learning_rate": 1.4202385743873259e-06, "loss": 0.8816, "step": 25043 }, { "epoch": 0.02, "learning_rate": 1.4196849279175228e-06, "loss": 0.8659, "step": 25044 }, { "epoch": 0.02, "learning_rate": 1.4191313811362195e-06, "loss": 0.7898, "step": 25045 }, { "epoch": 0.02, "learning_rate": 1.4185779340498473e-06, "loss": 0.821, "step": 25046 }, { "epoch": 0.02, "learning_rate": 1.4180245866648356e-06, "loss": 0.8594, "step": 25047 }, { "epoch": 0.02, "learning_rate": 1.4174713389876137e-06, "loss": 0.8508, "step": 25048 }, { "epoch": 0.02, "learning_rate": 1.4169181910246099e-06, "loss": 0.8544, "step": 25049 }, { "epoch": 0.02, "learning_rate": 1.4163651427822522e-06, "loss": 0.8387, "step": 25050 }, { "epoch": 0.02, "learning_rate": 1.4158121942669622e-06, "loss": 0.8634, "step": 25051 }, { "epoch": 0.02, "learning_rate": 1.4152593454851672e-06, "loss": 0.8527, "step": 25052 }, { "epoch": 0.02, "learning_rate": 1.4147065964432883e-06, "loss": 0.7984, "step": 25053 }, { "epoch": 0.02, "learning_rate": 1.4141539471477482e-06, "loss": 0.825, "step": 25054 }, { "epoch": 0.02, "learning_rate": 1.413601397604969e-06, "loss": 0.8554, "step": 25055 }, { "epoch": 0.02, "learning_rate": 1.4130489478213694e-06, "loss": 0.8671, "step": 25056 }, { "epoch": 0.02, "learning_rate": 1.41249659780337e-06, "loss": 0.8481, "step": 25057 }, { "epoch": 0.02, "learning_rate": 1.4119443475573845e-06, "loss": 0.8504, "step": 25058 }, { "epoch": 0.02, "learning_rate": 1.4113921970898282e-06, "loss": 0.8815, "step": 25059 }, { "epoch": 0.02, "learning_rate": 1.410840146407123e-06, "loss": 0.8558, "step": 25060 }, { "epoch": 0.02, "learning_rate": 1.4102881955156765e-06, "loss": 0.8358, "step": 25061 }, { "epoch": 0.02, "learning_rate": 1.4097363444219037e-06, "loss": 0.843, "step": 25062 }, { "epoch": 0.02, "learning_rate": 1.4091845931322156e-06, "loss": 0.824, "step": 25063 }, { "epoch": 0.02, "learning_rate": 1.4086329416530252e-06, "loss": 0.8393, "step": 25064 }, { "epoch": 0.02, "learning_rate": 1.4080813899907353e-06, "loss": 0.8174, "step": 25065 }, { "epoch": 0.02, "learning_rate": 1.4075299381517604e-06, "loss": 0.8731, "step": 25066 }, { "epoch": 0.02, "learning_rate": 1.4069785861425045e-06, "loss": 0.9403, "step": 25067 }, { "epoch": 0.02, "learning_rate": 1.4064273339693768e-06, "loss": 0.8209, "step": 25068 }, { "epoch": 0.02, "learning_rate": 1.4058761816387766e-06, "loss": 0.8768, "step": 25069 }, { "epoch": 0.02, "learning_rate": 1.4053251291571101e-06, "loss": 0.855, "step": 25070 }, { "epoch": 0.02, "learning_rate": 1.4047741765307788e-06, "loss": 0.932, "step": 25071 }, { "epoch": 0.02, "learning_rate": 1.4042233237661852e-06, "loss": 0.8086, "step": 25072 }, { "epoch": 0.02, "learning_rate": 1.4036725708697273e-06, "loss": 0.8401, "step": 25073 }, { "epoch": 0.02, "learning_rate": 1.403121917847805e-06, "loss": 0.8847, "step": 25074 }, { "epoch": 0.02, "learning_rate": 1.4025713647068196e-06, "loss": 0.8551, "step": 25075 }, { "epoch": 0.02, "learning_rate": 1.4020209114531601e-06, "loss": 0.8304, "step": 25076 }, { "epoch": 0.02, "learning_rate": 1.4014705580932263e-06, "loss": 0.8443, "step": 25077 }, { "epoch": 0.02, "learning_rate": 1.400920304633413e-06, "loss": 0.8746, "step": 25078 }, { "epoch": 0.02, "learning_rate": 1.4003701510801105e-06, "loss": 0.8825, "step": 25079 }, { "epoch": 0.02, "learning_rate": 1.3998200974397125e-06, "loss": 0.8508, "step": 25080 }, { "epoch": 0.02, "learning_rate": 1.3992701437186085e-06, "loss": 0.8146, "step": 25081 }, { "epoch": 0.02, "learning_rate": 1.398720289923191e-06, "loss": 0.8721, "step": 25082 }, { "epoch": 0.02, "learning_rate": 1.3981705360598453e-06, "loss": 0.8699, "step": 25083 }, { "epoch": 0.02, "learning_rate": 1.3976208821349558e-06, "loss": 0.8817, "step": 25084 }, { "epoch": 0.02, "learning_rate": 1.3970713281549176e-06, "loss": 0.8558, "step": 25085 }, { "epoch": 0.02, "learning_rate": 1.3965218741261067e-06, "loss": 0.8347, "step": 25086 }, { "epoch": 0.02, "learning_rate": 1.3959725200549111e-06, "loss": 0.8496, "step": 25087 }, { "epoch": 0.02, "learning_rate": 1.3954232659477117e-06, "loss": 0.8475, "step": 25088 }, { "epoch": 0.02, "learning_rate": 1.394874111810891e-06, "loss": 0.8714, "step": 25089 }, { "epoch": 0.02, "learning_rate": 1.3943250576508293e-06, "loss": 0.8348, "step": 25090 }, { "epoch": 0.02, "learning_rate": 1.3937761034739051e-06, "loss": 0.8543, "step": 25091 }, { "epoch": 0.02, "learning_rate": 1.3932272492864984e-06, "loss": 0.8327, "step": 25092 }, { "epoch": 0.02, "learning_rate": 1.3926784950949823e-06, "loss": 0.8133, "step": 25093 }, { "epoch": 0.02, "learning_rate": 1.3921298409057338e-06, "loss": 0.8309, "step": 25094 }, { "epoch": 0.02, "learning_rate": 1.3915812867251265e-06, "loss": 0.8491, "step": 25095 }, { "epoch": 0.02, "learning_rate": 1.3910328325595358e-06, "loss": 0.8391, "step": 25096 }, { "epoch": 0.02, "learning_rate": 1.3904844784153325e-06, "loss": 0.8296, "step": 25097 }, { "epoch": 0.02, "learning_rate": 1.3899362242988878e-06, "loss": 0.8747, "step": 25098 }, { "epoch": 0.02, "learning_rate": 1.3893880702165742e-06, "loss": 0.8517, "step": 25099 }, { "epoch": 0.02, "learning_rate": 1.388840016174755e-06, "loss": 0.8762, "step": 25100 }, { "epoch": 0.02, "learning_rate": 1.388292062179799e-06, "loss": 0.7984, "step": 25101 }, { "epoch": 0.02, "learning_rate": 1.3877442082380755e-06, "loss": 0.8755, "step": 25102 }, { "epoch": 0.02, "learning_rate": 1.3871964543559468e-06, "loss": 0.8446, "step": 25103 }, { "epoch": 0.02, "learning_rate": 1.3866488005397783e-06, "loss": 0.8109, "step": 25104 }, { "epoch": 0.02, "learning_rate": 1.3861012467959313e-06, "loss": 0.871, "step": 25105 }, { "epoch": 0.02, "learning_rate": 1.3855537931307716e-06, "loss": 0.8259, "step": 25106 }, { "epoch": 0.02, "learning_rate": 1.3850064395506547e-06, "loss": 0.8479, "step": 25107 }, { "epoch": 0.02, "learning_rate": 1.3844591860619382e-06, "loss": 0.8229, "step": 25108 }, { "epoch": 0.02, "learning_rate": 1.3839120326709865e-06, "loss": 0.8593, "step": 25109 }, { "epoch": 0.02, "learning_rate": 1.3833649793841564e-06, "loss": 0.8584, "step": 25110 }, { "epoch": 0.02, "learning_rate": 1.3828180262077995e-06, "loss": 0.8874, "step": 25111 }, { "epoch": 0.02, "learning_rate": 1.3822711731482719e-06, "loss": 0.8486, "step": 25112 }, { "epoch": 0.02, "learning_rate": 1.3817244202119263e-06, "loss": 0.8301, "step": 25113 }, { "epoch": 0.02, "learning_rate": 1.3811777674051174e-06, "loss": 0.8274, "step": 25114 }, { "epoch": 0.02, "learning_rate": 1.3806312147341939e-06, "loss": 0.8599, "step": 25115 }, { "epoch": 0.02, "learning_rate": 1.3800847622055081e-06, "loss": 0.8513, "step": 25116 }, { "epoch": 0.02, "learning_rate": 1.3795384098254084e-06, "loss": 0.8698, "step": 25117 }, { "epoch": 0.02, "learning_rate": 1.3789921576002407e-06, "loss": 0.84, "step": 25118 }, { "epoch": 0.02, "learning_rate": 1.3784460055363524e-06, "loss": 0.8835, "step": 25119 }, { "epoch": 0.02, "learning_rate": 1.3778999536400883e-06, "loss": 0.8755, "step": 25120 }, { "epoch": 0.02, "learning_rate": 1.3773540019177945e-06, "loss": 0.861, "step": 25121 }, { "epoch": 0.02, "learning_rate": 1.3768081503758114e-06, "loss": 0.8844, "step": 25122 }, { "epoch": 0.02, "learning_rate": 1.3762623990204837e-06, "loss": 0.8433, "step": 25123 }, { "epoch": 0.02, "learning_rate": 1.3757167478581512e-06, "loss": 0.8752, "step": 25124 }, { "epoch": 0.02, "learning_rate": 1.375171196895152e-06, "loss": 0.8196, "step": 25125 }, { "epoch": 0.02, "learning_rate": 1.3746257461378253e-06, "loss": 0.8192, "step": 25126 }, { "epoch": 0.02, "learning_rate": 1.3740803955925053e-06, "loss": 0.8273, "step": 25127 }, { "epoch": 0.02, "learning_rate": 1.3735351452655365e-06, "loss": 0.8616, "step": 25128 }, { "epoch": 0.02, "learning_rate": 1.3729899951632453e-06, "loss": 0.8449, "step": 25129 }, { "epoch": 0.02, "learning_rate": 1.3724449452919685e-06, "loss": 0.8586, "step": 25130 }, { "epoch": 0.02, "learning_rate": 1.3718999956580415e-06, "loss": 0.8812, "step": 25131 }, { "epoch": 0.02, "learning_rate": 1.3713551462677876e-06, "loss": 0.8184, "step": 25132 }, { "epoch": 0.02, "learning_rate": 1.3708103971275456e-06, "loss": 0.8575, "step": 25133 }, { "epoch": 0.02, "learning_rate": 1.3702657482436398e-06, "loss": 0.8436, "step": 25134 }, { "epoch": 0.02, "learning_rate": 1.3697211996224025e-06, "loss": 0.8465, "step": 25135 }, { "epoch": 0.02, "learning_rate": 1.3691767512701548e-06, "loss": 0.8339, "step": 25136 }, { "epoch": 0.02, "learning_rate": 1.3686324031932241e-06, "loss": 0.8858, "step": 25137 }, { "epoch": 0.02, "learning_rate": 1.3680881553979363e-06, "loss": 0.8764, "step": 25138 }, { "epoch": 0.02, "learning_rate": 1.3675440078906144e-06, "loss": 0.834, "step": 25139 }, { "epoch": 0.02, "learning_rate": 1.3669999606775786e-06, "loss": 0.8651, "step": 25140 }, { "epoch": 0.02, "learning_rate": 1.3664560137651517e-06, "loss": 0.8447, "step": 25141 }, { "epoch": 0.02, "learning_rate": 1.3659121671596542e-06, "loss": 0.8135, "step": 25142 }, { "epoch": 0.02, "learning_rate": 1.3653684208674023e-06, "loss": 0.8568, "step": 25143 }, { "epoch": 0.02, "learning_rate": 1.364824774894714e-06, "loss": 0.8516, "step": 25144 }, { "epoch": 0.02, "learning_rate": 1.3642812292479058e-06, "loss": 0.8338, "step": 25145 }, { "epoch": 0.02, "learning_rate": 1.3637377839332921e-06, "loss": 0.8518, "step": 25146 }, { "epoch": 0.02, "learning_rate": 1.3631944389571882e-06, "loss": 0.8581, "step": 25147 }, { "epoch": 0.02, "learning_rate": 1.3626511943259058e-06, "loss": 0.8688, "step": 25148 }, { "epoch": 0.02, "learning_rate": 1.3621080500457595e-06, "loss": 0.8518, "step": 25149 }, { "epoch": 0.02, "learning_rate": 1.3615650061230545e-06, "loss": 0.8346, "step": 25150 }, { "epoch": 0.02, "learning_rate": 1.3610220625641002e-06, "loss": 0.8513, "step": 25151 }, { "epoch": 0.02, "learning_rate": 1.3604792193752126e-06, "loss": 0.8592, "step": 25152 }, { "epoch": 0.02, "learning_rate": 1.35993647656269e-06, "loss": 0.8859, "step": 25153 }, { "epoch": 0.02, "learning_rate": 1.3593938341328405e-06, "loss": 0.8645, "step": 25154 }, { "epoch": 0.02, "learning_rate": 1.3588512920919706e-06, "loss": 0.8929, "step": 25155 }, { "epoch": 0.02, "learning_rate": 1.3583088504463837e-06, "loss": 0.871, "step": 25156 }, { "epoch": 0.02, "learning_rate": 1.3577665092023772e-06, "loss": 0.8393, "step": 25157 }, { "epoch": 0.02, "learning_rate": 1.3572242683662573e-06, "loss": 0.8415, "step": 25158 }, { "epoch": 0.02, "learning_rate": 1.356682127944322e-06, "loss": 0.8739, "step": 25159 }, { "epoch": 0.02, "learning_rate": 1.3561400879428744e-06, "loss": 0.8539, "step": 25160 }, { "epoch": 0.02, "learning_rate": 1.3555981483682046e-06, "loss": 0.8543, "step": 25161 }, { "epoch": 0.02, "learning_rate": 1.3550563092266122e-06, "loss": 0.8652, "step": 25162 }, { "epoch": 0.02, "learning_rate": 1.3545145705243935e-06, "loss": 0.8315, "step": 25163 }, { "epoch": 0.02, "learning_rate": 1.3539729322678409e-06, "loss": 0.8425, "step": 25164 }, { "epoch": 0.02, "learning_rate": 1.3534313944632493e-06, "loss": 0.8361, "step": 25165 }, { "epoch": 0.02, "learning_rate": 1.3528899571169084e-06, "loss": 0.8618, "step": 25166 }, { "epoch": 0.02, "learning_rate": 1.3523486202351132e-06, "loss": 0.8704, "step": 25167 }, { "epoch": 0.02, "learning_rate": 1.3518073838241462e-06, "loss": 0.8856, "step": 25168 }, { "epoch": 0.02, "learning_rate": 1.3512662478902993e-06, "loss": 0.8532, "step": 25169 }, { "epoch": 0.02, "learning_rate": 1.3507252124398596e-06, "loss": 0.8447, "step": 25170 }, { "epoch": 0.02, "learning_rate": 1.350184277479113e-06, "loss": 0.8594, "step": 25171 }, { "epoch": 0.02, "learning_rate": 1.3496434430143436e-06, "loss": 0.8675, "step": 25172 }, { "epoch": 0.02, "learning_rate": 1.349102709051835e-06, "loss": 0.8119, "step": 25173 }, { "epoch": 0.02, "learning_rate": 1.3485620755978723e-06, "loss": 0.884, "step": 25174 }, { "epoch": 0.02, "learning_rate": 1.348021542658733e-06, "loss": 0.8625, "step": 25175 }, { "epoch": 0.02, "learning_rate": 1.3474811102406949e-06, "loss": 0.843, "step": 25176 }, { "epoch": 0.02, "learning_rate": 1.3469407783500453e-06, "loss": 0.8995, "step": 25177 }, { "epoch": 0.02, "learning_rate": 1.3464005469930564e-06, "loss": 0.8194, "step": 25178 }, { "epoch": 0.02, "learning_rate": 1.3458604161760036e-06, "loss": 0.8626, "step": 25179 }, { "epoch": 0.02, "learning_rate": 1.3453203859051657e-06, "loss": 0.8468, "step": 25180 }, { "epoch": 0.02, "learning_rate": 1.3447804561868138e-06, "loss": 0.8477, "step": 25181 }, { "epoch": 0.02, "learning_rate": 1.3442406270272235e-06, "loss": 0.8385, "step": 25182 }, { "epoch": 0.02, "learning_rate": 1.3437008984326661e-06, "loss": 0.8682, "step": 25183 }, { "epoch": 0.02, "learning_rate": 1.3431612704094145e-06, "loss": 0.8479, "step": 25184 }, { "epoch": 0.02, "learning_rate": 1.3426217429637322e-06, "loss": 0.8699, "step": 25185 }, { "epoch": 0.02, "learning_rate": 1.3420823161018913e-06, "loss": 0.8595, "step": 25186 }, { "epoch": 0.02, "learning_rate": 1.3415429898301602e-06, "loss": 0.8223, "step": 25187 }, { "epoch": 0.02, "learning_rate": 1.3410037641548023e-06, "loss": 0.8435, "step": 25188 }, { "epoch": 0.02, "learning_rate": 1.3404646390820841e-06, "loss": 0.842, "step": 25189 }, { "epoch": 0.02, "learning_rate": 1.3399256146182682e-06, "loss": 0.8777, "step": 25190 }, { "epoch": 0.02, "learning_rate": 1.3393866907696218e-06, "loss": 0.8567, "step": 25191 }, { "epoch": 0.02, "learning_rate": 1.338847867542399e-06, "loss": 0.8776, "step": 25192 }, { "epoch": 0.02, "learning_rate": 1.3383091449428632e-06, "loss": 0.8486, "step": 25193 }, { "epoch": 0.02, "learning_rate": 1.337770522977272e-06, "loss": 0.8665, "step": 25194 }, { "epoch": 0.02, "learning_rate": 1.337232001651888e-06, "loss": 0.8826, "step": 25195 }, { "epoch": 0.02, "learning_rate": 1.3366935809729641e-06, "loss": 0.8737, "step": 25196 }, { "epoch": 0.02, "learning_rate": 1.3361552609467553e-06, "loss": 0.8396, "step": 25197 }, { "epoch": 0.02, "learning_rate": 1.3356170415795178e-06, "loss": 0.8643, "step": 25198 }, { "epoch": 0.02, "learning_rate": 1.3350789228775064e-06, "loss": 0.865, "step": 25199 }, { "epoch": 0.02, "learning_rate": 1.3345409048469648e-06, "loss": 0.8527, "step": 25200 }, { "epoch": 0.02, "learning_rate": 1.334002987494154e-06, "loss": 0.7959, "step": 25201 }, { "epoch": 0.02, "learning_rate": 1.333465170825321e-06, "loss": 0.8453, "step": 25202 }, { "epoch": 0.02, "learning_rate": 1.332927454846711e-06, "loss": 0.8815, "step": 25203 }, { "epoch": 0.02, "learning_rate": 1.3323898395645718e-06, "loss": 0.8245, "step": 25204 }, { "epoch": 0.02, "learning_rate": 1.3318523249851522e-06, "loss": 0.8569, "step": 25205 }, { "epoch": 0.02, "learning_rate": 1.331314911114695e-06, "loss": 0.8511, "step": 25206 }, { "epoch": 0.02, "learning_rate": 1.3307775979594463e-06, "loss": 0.8898, "step": 25207 }, { "epoch": 0.02, "learning_rate": 1.3302403855256463e-06, "loss": 0.8598, "step": 25208 }, { "epoch": 0.02, "learning_rate": 1.3297032738195404e-06, "loss": 0.8191, "step": 25209 }, { "epoch": 0.02, "learning_rate": 1.3291662628473634e-06, "loss": 0.8254, "step": 25210 }, { "epoch": 0.02, "learning_rate": 1.328629352615357e-06, "loss": 0.8747, "step": 25211 }, { "epoch": 0.02, "learning_rate": 1.3280925431297597e-06, "loss": 0.8264, "step": 25212 }, { "epoch": 0.02, "learning_rate": 1.3275558343968075e-06, "loss": 0.8322, "step": 25213 }, { "epoch": 0.02, "learning_rate": 1.3270192264227366e-06, "loss": 0.8406, "step": 25214 }, { "epoch": 0.02, "learning_rate": 1.3264827192137818e-06, "loss": 0.8909, "step": 25215 }, { "epoch": 0.02, "learning_rate": 1.3259463127761784e-06, "loss": 0.8565, "step": 25216 }, { "epoch": 0.02, "learning_rate": 1.3254100071161524e-06, "loss": 0.8851, "step": 25217 }, { "epoch": 0.02, "learning_rate": 1.32487380223994e-06, "loss": 0.8762, "step": 25218 }, { "epoch": 0.02, "learning_rate": 1.3243376981537658e-06, "loss": 0.8749, "step": 25219 }, { "epoch": 0.02, "learning_rate": 1.3238016948638666e-06, "loss": 0.8502, "step": 25220 }, { "epoch": 0.02, "learning_rate": 1.3232657923764635e-06, "loss": 0.831, "step": 25221 }, { "epoch": 0.02, "learning_rate": 1.3227299906977843e-06, "loss": 0.8589, "step": 25222 }, { "epoch": 0.02, "learning_rate": 1.3221942898340557e-06, "loss": 0.8645, "step": 25223 }, { "epoch": 0.02, "learning_rate": 1.3216586897914963e-06, "loss": 0.8677, "step": 25224 }, { "epoch": 0.02, "learning_rate": 1.3211231905763356e-06, "loss": 0.848, "step": 25225 }, { "epoch": 0.02, "learning_rate": 1.3205877921947918e-06, "loss": 0.8704, "step": 25226 }, { "epoch": 0.02, "learning_rate": 1.320052494653088e-06, "loss": 0.8509, "step": 25227 }, { "epoch": 0.02, "learning_rate": 1.3195172979574388e-06, "loss": 0.8287, "step": 25228 }, { "epoch": 0.02, "learning_rate": 1.3189822021140642e-06, "loss": 0.8634, "step": 25229 }, { "epoch": 0.02, "learning_rate": 1.318447207129182e-06, "loss": 0.7948, "step": 25230 }, { "epoch": 0.02, "learning_rate": 1.3179123130090078e-06, "loss": 0.8751, "step": 25231 }, { "epoch": 0.02, "learning_rate": 1.3173775197597561e-06, "loss": 0.8605, "step": 25232 }, { "epoch": 0.02, "learning_rate": 1.3168428273876398e-06, "loss": 0.8584, "step": 25233 }, { "epoch": 0.02, "learning_rate": 1.3163082358988732e-06, "loss": 0.8551, "step": 25234 }, { "epoch": 0.02, "learning_rate": 1.3157737452996645e-06, "loss": 0.8793, "step": 25235 }, { "epoch": 0.02, "learning_rate": 1.3152393555962227e-06, "loss": 0.8561, "step": 25236 }, { "epoch": 0.02, "learning_rate": 1.3147050667947603e-06, "loss": 0.8193, "step": 25237 }, { "epoch": 0.02, "learning_rate": 1.3141708789014818e-06, "loss": 0.873, "step": 25238 }, { "epoch": 0.02, "learning_rate": 1.3136367919225946e-06, "loss": 0.808, "step": 25239 }, { "epoch": 0.02, "learning_rate": 1.3131028058643036e-06, "loss": 0.9068, "step": 25240 }, { "epoch": 0.02, "learning_rate": 1.312568920732815e-06, "loss": 0.8101, "step": 25241 }, { "epoch": 0.02, "learning_rate": 1.3120351365343286e-06, "loss": 0.8396, "step": 25242 }, { "epoch": 0.02, "learning_rate": 1.3115014532750447e-06, "loss": 0.8317, "step": 25243 }, { "epoch": 0.02, "learning_rate": 1.3109678709611672e-06, "loss": 0.8242, "step": 25244 }, { "epoch": 0.02, "learning_rate": 1.310434389598898e-06, "loss": 0.8515, "step": 25245 }, { "epoch": 0.02, "learning_rate": 1.3099010091944287e-06, "loss": 0.8615, "step": 25246 }, { "epoch": 0.02, "learning_rate": 1.3093677297539597e-06, "loss": 0.8733, "step": 25247 }, { "epoch": 0.02, "learning_rate": 1.3088345512836886e-06, "loss": 0.8662, "step": 25248 }, { "epoch": 0.02, "learning_rate": 1.3083014737898036e-06, "loss": 0.8587, "step": 25249 }, { "epoch": 0.02, "learning_rate": 1.3077684972785044e-06, "loss": 0.8485, "step": 25250 }, { "epoch": 0.02, "learning_rate": 1.30723562175598e-06, "loss": 0.8773, "step": 25251 }, { "epoch": 0.03, "learning_rate": 1.306702847228426e-06, "loss": 0.8438, "step": 25252 }, { "epoch": 0.03, "learning_rate": 1.306170173702026e-06, "loss": 0.8287, "step": 25253 }, { "epoch": 0.03, "learning_rate": 1.305637601182973e-06, "loss": 0.8603, "step": 25254 }, { "epoch": 0.03, "learning_rate": 1.305105129677452e-06, "loss": 0.827, "step": 25255 }, { "epoch": 0.03, "learning_rate": 1.3045727591916513e-06, "loss": 0.8498, "step": 25256 }, { "epoch": 0.03, "learning_rate": 1.3040404897317549e-06, "loss": 0.8591, "step": 25257 }, { "epoch": 0.03, "learning_rate": 1.303508321303949e-06, "loss": 0.8613, "step": 25258 }, { "epoch": 0.03, "learning_rate": 1.3029762539144153e-06, "loss": 0.898, "step": 25259 }, { "epoch": 0.03, "learning_rate": 1.3024442875693343e-06, "loss": 0.8452, "step": 25260 }, { "epoch": 0.03, "learning_rate": 1.301912422274887e-06, "loss": 0.8146, "step": 25261 }, { "epoch": 0.03, "learning_rate": 1.3013806580372534e-06, "loss": 0.8758, "step": 25262 }, { "epoch": 0.03, "learning_rate": 1.3008489948626113e-06, "loss": 0.8617, "step": 25263 }, { "epoch": 0.03, "learning_rate": 1.3003174327571378e-06, "loss": 0.834, "step": 25264 }, { "epoch": 0.03, "learning_rate": 1.2997859717270078e-06, "loss": 0.8362, "step": 25265 }, { "epoch": 0.03, "learning_rate": 1.299254611778401e-06, "loss": 0.8655, "step": 25266 }, { "epoch": 0.03, "learning_rate": 1.2987233529174815e-06, "loss": 0.8455, "step": 25267 }, { "epoch": 0.03, "learning_rate": 1.2981921951504295e-06, "loss": 0.8504, "step": 25268 }, { "epoch": 0.03, "learning_rate": 1.2976611384834148e-06, "loss": 0.8449, "step": 25269 }, { "epoch": 0.03, "learning_rate": 1.2971301829226057e-06, "loss": 0.8381, "step": 25270 }, { "epoch": 0.03, "learning_rate": 1.2965993284741696e-06, "loss": 0.907, "step": 25271 }, { "epoch": 0.03, "learning_rate": 1.296068575144277e-06, "loss": 0.8474, "step": 25272 }, { "epoch": 0.03, "learning_rate": 1.295537922939094e-06, "loss": 0.846, "step": 25273 }, { "epoch": 0.03, "learning_rate": 1.2950073718647838e-06, "loss": 0.8734, "step": 25274 }, { "epoch": 0.03, "learning_rate": 1.2944769219275132e-06, "loss": 0.8528, "step": 25275 }, { "epoch": 0.03, "learning_rate": 1.2939465731334444e-06, "loss": 0.8596, "step": 25276 }, { "epoch": 0.03, "learning_rate": 1.2934163254887378e-06, "loss": 0.8296, "step": 25277 }, { "epoch": 0.03, "learning_rate": 1.292886178999554e-06, "loss": 0.8235, "step": 25278 }, { "epoch": 0.03, "learning_rate": 1.2923561336720536e-06, "loss": 0.8519, "step": 25279 }, { "epoch": 0.03, "learning_rate": 1.291826189512394e-06, "loss": 0.8445, "step": 25280 }, { "epoch": 0.03, "learning_rate": 1.2912963465267324e-06, "loss": 0.8756, "step": 25281 }, { "epoch": 0.03, "learning_rate": 1.2907666047212252e-06, "loss": 0.8544, "step": 25282 }, { "epoch": 0.03, "learning_rate": 1.2902369641020285e-06, "loss": 0.858, "step": 25283 }, { "epoch": 0.03, "learning_rate": 1.2897074246752917e-06, "loss": 0.8364, "step": 25284 }, { "epoch": 0.03, "learning_rate": 1.28917798644717e-06, "loss": 0.8531, "step": 25285 }, { "epoch": 0.03, "learning_rate": 1.2886486494238115e-06, "loss": 0.8418, "step": 25286 }, { "epoch": 0.03, "learning_rate": 1.2881194136113727e-06, "loss": 0.836, "step": 25287 }, { "epoch": 0.03, "learning_rate": 1.2875902790159966e-06, "loss": 0.8378, "step": 25288 }, { "epoch": 0.03, "learning_rate": 1.2870612456438326e-06, "loss": 0.8067, "step": 25289 }, { "epoch": 0.03, "learning_rate": 1.2865323135010277e-06, "loss": 0.8319, "step": 25290 }, { "epoch": 0.03, "learning_rate": 1.2860034825937284e-06, "loss": 0.8839, "step": 25291 }, { "epoch": 0.03, "learning_rate": 1.285474752928072e-06, "loss": 0.8451, "step": 25292 }, { "epoch": 0.03, "learning_rate": 1.28494612451021e-06, "loss": 0.8599, "step": 25293 }, { "epoch": 0.03, "learning_rate": 1.2844175973462814e-06, "loss": 0.8445, "step": 25294 }, { "epoch": 0.03, "learning_rate": 1.2838891714424251e-06, "loss": 0.8178, "step": 25295 }, { "epoch": 0.03, "learning_rate": 1.283360846804781e-06, "loss": 0.8191, "step": 25296 }, { "epoch": 0.03, "learning_rate": 1.2828326234394884e-06, "loss": 0.8723, "step": 25297 }, { "epoch": 0.03, "learning_rate": 1.2823045013526824e-06, "loss": 0.8857, "step": 25298 }, { "epoch": 0.03, "learning_rate": 1.2817764805505017e-06, "loss": 0.8194, "step": 25299 }, { "epoch": 0.03, "learning_rate": 1.2812485610390778e-06, "loss": 0.8284, "step": 25300 }, { "epoch": 0.03, "learning_rate": 1.2807207428245495e-06, "loss": 0.8851, "step": 25301 }, { "epoch": 0.03, "learning_rate": 1.2801930259130425e-06, "loss": 0.8331, "step": 25302 }, { "epoch": 0.03, "learning_rate": 1.279665410310691e-06, "loss": 0.8408, "step": 25303 }, { "epoch": 0.03, "learning_rate": 1.2791378960236257e-06, "loss": 0.8467, "step": 25304 }, { "epoch": 0.03, "learning_rate": 1.2786104830579738e-06, "loss": 0.8174, "step": 25305 }, { "epoch": 0.03, "learning_rate": 1.2780831714198648e-06, "loss": 0.8395, "step": 25306 }, { "epoch": 0.03, "learning_rate": 1.2775559611154231e-06, "loss": 0.8246, "step": 25307 }, { "epoch": 0.03, "learning_rate": 1.2770288521507779e-06, "loss": 0.8559, "step": 25308 }, { "epoch": 0.03, "learning_rate": 1.2765018445320488e-06, "loss": 0.8407, "step": 25309 }, { "epoch": 0.03, "learning_rate": 1.2759749382653575e-06, "loss": 0.8543, "step": 25310 }, { "epoch": 0.03, "learning_rate": 1.2754481333568303e-06, "loss": 0.8605, "step": 25311 }, { "epoch": 0.03, "learning_rate": 1.274921429812589e-06, "loss": 0.8308, "step": 25312 }, { "epoch": 0.03, "learning_rate": 1.2743948276387476e-06, "loss": 0.8248, "step": 25313 }, { "epoch": 0.03, "learning_rate": 1.2738683268414276e-06, "loss": 0.8793, "step": 25314 }, { "epoch": 0.03, "learning_rate": 1.2733419274267467e-06, "loss": 0.8525, "step": 25315 }, { "epoch": 0.03, "learning_rate": 1.2728156294008143e-06, "loss": 0.903, "step": 25316 }, { "epoch": 0.03, "learning_rate": 1.2722894327697533e-06, "loss": 0.8214, "step": 25317 }, { "epoch": 0.03, "learning_rate": 1.2717633375396732e-06, "loss": 0.8766, "step": 25318 }, { "epoch": 0.03, "learning_rate": 1.2712373437166892e-06, "loss": 0.8973, "step": 25319 }, { "epoch": 0.03, "learning_rate": 1.2707114513069085e-06, "loss": 0.8526, "step": 25320 }, { "epoch": 0.03, "learning_rate": 1.270185660316443e-06, "loss": 0.8723, "step": 25321 }, { "epoch": 0.03, "learning_rate": 1.2696599707514012e-06, "loss": 0.8711, "step": 25322 }, { "epoch": 0.03, "learning_rate": 1.2691343826178903e-06, "loss": 0.8517, "step": 25323 }, { "epoch": 0.03, "learning_rate": 1.2686088959220189e-06, "loss": 0.8196, "step": 25324 }, { "epoch": 0.03, "learning_rate": 1.2680835106698886e-06, "loss": 0.8304, "step": 25325 }, { "epoch": 0.03, "learning_rate": 1.2675582268676089e-06, "loss": 0.8271, "step": 25326 }, { "epoch": 0.03, "learning_rate": 1.2670330445212775e-06, "loss": 0.8641, "step": 25327 }, { "epoch": 0.03, "learning_rate": 1.266507963636997e-06, "loss": 0.8618, "step": 25328 }, { "epoch": 0.03, "learning_rate": 1.2659829842208693e-06, "loss": 0.8581, "step": 25329 }, { "epoch": 0.03, "learning_rate": 1.2654581062789939e-06, "loss": 0.8469, "step": 25330 }, { "epoch": 0.03, "learning_rate": 1.264933329817467e-06, "loss": 0.8724, "step": 25331 }, { "epoch": 0.03, "learning_rate": 1.2644086548423884e-06, "loss": 0.8744, "step": 25332 }, { "epoch": 0.03, "learning_rate": 1.2638840813598542e-06, "loss": 0.8479, "step": 25333 }, { "epoch": 0.03, "learning_rate": 1.263359609375955e-06, "loss": 0.8598, "step": 25334 }, { "epoch": 0.03, "learning_rate": 1.2628352388967846e-06, "loss": 0.857, "step": 25335 }, { "epoch": 0.03, "learning_rate": 1.2623109699284396e-06, "loss": 0.868, "step": 25336 }, { "epoch": 0.03, "learning_rate": 1.2617868024770108e-06, "loss": 0.8414, "step": 25337 }, { "epoch": 0.03, "learning_rate": 1.261262736548584e-06, "loss": 0.836, "step": 25338 }, { "epoch": 0.03, "learning_rate": 1.2607387721492503e-06, "loss": 0.8583, "step": 25339 }, { "epoch": 0.03, "learning_rate": 1.2602149092850969e-06, "loss": 0.8443, "step": 25340 }, { "epoch": 0.03, "learning_rate": 1.25969114796221e-06, "loss": 0.8374, "step": 25341 }, { "epoch": 0.03, "learning_rate": 1.2591674881866756e-06, "loss": 0.8624, "step": 25342 }, { "epoch": 0.03, "learning_rate": 1.2586439299645769e-06, "loss": 0.8806, "step": 25343 }, { "epoch": 0.03, "learning_rate": 1.2581204733019991e-06, "loss": 0.841, "step": 25344 }, { "epoch": 0.03, "learning_rate": 1.2575971182050207e-06, "loss": 0.8632, "step": 25345 }, { "epoch": 0.03, "learning_rate": 1.2570738646797232e-06, "loss": 0.874, "step": 25346 }, { "epoch": 0.03, "learning_rate": 1.2565507127321851e-06, "loss": 0.8678, "step": 25347 }, { "epoch": 0.03, "learning_rate": 1.2560276623684863e-06, "loss": 0.8342, "step": 25348 }, { "epoch": 0.03, "learning_rate": 1.255504713594703e-06, "loss": 0.8487, "step": 25349 }, { "epoch": 0.03, "learning_rate": 1.25498186641691e-06, "loss": 0.8622, "step": 25350 }, { "epoch": 0.03, "learning_rate": 1.254459120841186e-06, "loss": 0.8123, "step": 25351 }, { "epoch": 0.03, "learning_rate": 1.2539364768735984e-06, "loss": 0.8282, "step": 25352 }, { "epoch": 0.03, "learning_rate": 1.2534139345202223e-06, "loss": 0.8603, "step": 25353 }, { "epoch": 0.03, "learning_rate": 1.2528914937871296e-06, "loss": 0.8324, "step": 25354 }, { "epoch": 0.03, "learning_rate": 1.2523691546803872e-06, "loss": 0.8552, "step": 25355 }, { "epoch": 0.03, "learning_rate": 1.2518469172060677e-06, "loss": 0.8716, "step": 25356 }, { "epoch": 0.03, "learning_rate": 1.2513247813702355e-06, "loss": 0.8422, "step": 25357 }, { "epoch": 0.03, "learning_rate": 1.2508027471789608e-06, "loss": 0.8629, "step": 25358 }, { "epoch": 0.03, "learning_rate": 1.2502808146383028e-06, "loss": 0.8432, "step": 25359 }, { "epoch": 0.03, "learning_rate": 1.2497589837543288e-06, "loss": 0.8568, "step": 25360 }, { "epoch": 0.03, "learning_rate": 1.2492372545331055e-06, "loss": 0.8716, "step": 25361 }, { "epoch": 0.03, "learning_rate": 1.2487156269806877e-06, "loss": 0.851, "step": 25362 }, { "epoch": 0.03, "learning_rate": 1.2481941011031384e-06, "loss": 0.8579, "step": 25363 }, { "epoch": 0.03, "learning_rate": 1.2476726769065172e-06, "loss": 0.8318, "step": 25364 }, { "epoch": 0.03, "learning_rate": 1.2471513543968816e-06, "loss": 0.8614, "step": 25365 }, { "epoch": 0.03, "learning_rate": 1.246630133580289e-06, "loss": 0.858, "step": 25366 }, { "epoch": 0.03, "learning_rate": 1.2461090144627952e-06, "loss": 0.8776, "step": 25367 }, { "epoch": 0.03, "learning_rate": 1.2455879970504569e-06, "loss": 0.8363, "step": 25368 }, { "epoch": 0.03, "learning_rate": 1.2450670813493226e-06, "loss": 0.8544, "step": 25369 }, { "epoch": 0.03, "learning_rate": 1.2445462673654464e-06, "loss": 0.8559, "step": 25370 }, { "epoch": 0.03, "learning_rate": 1.2440255551048808e-06, "loss": 0.891, "step": 25371 }, { "epoch": 0.03, "learning_rate": 1.2435049445736736e-06, "loss": 0.8255, "step": 25372 }, { "epoch": 0.03, "learning_rate": 1.2429844357778742e-06, "loss": 0.8494, "step": 25373 }, { "epoch": 0.03, "learning_rate": 1.2424640287235302e-06, "loss": 0.8457, "step": 25374 }, { "epoch": 0.03, "learning_rate": 1.2419437234166898e-06, "loss": 0.8403, "step": 25375 }, { "epoch": 0.03, "learning_rate": 1.241423519863395e-06, "loss": 0.8658, "step": 25376 }, { "epoch": 0.03, "learning_rate": 1.2409034180696887e-06, "loss": 0.8142, "step": 25377 }, { "epoch": 0.03, "learning_rate": 1.2403834180416152e-06, "loss": 0.8739, "step": 25378 }, { "epoch": 0.03, "learning_rate": 1.2398635197852193e-06, "loss": 0.8895, "step": 25379 }, { "epoch": 0.03, "learning_rate": 1.2393437233065374e-06, "loss": 0.8455, "step": 25380 }, { "epoch": 0.03, "learning_rate": 1.2388240286116082e-06, "loss": 0.8701, "step": 25381 }, { "epoch": 0.03, "learning_rate": 1.2383044357064722e-06, "loss": 0.8675, "step": 25382 }, { "epoch": 0.03, "learning_rate": 1.2377849445971646e-06, "loss": 0.8377, "step": 25383 }, { "epoch": 0.03, "learning_rate": 1.2372655552897218e-06, "loss": 0.8334, "step": 25384 }, { "epoch": 0.03, "learning_rate": 1.2367462677901775e-06, "loss": 0.8787, "step": 25385 }, { "epoch": 0.03, "learning_rate": 1.2362270821045674e-06, "loss": 0.9047, "step": 25386 }, { "epoch": 0.03, "learning_rate": 1.2357079982389197e-06, "loss": 0.8426, "step": 25387 }, { "epoch": 0.03, "learning_rate": 1.2351890161992664e-06, "loss": 0.8405, "step": 25388 }, { "epoch": 0.03, "learning_rate": 1.234670135991638e-06, "loss": 0.8417, "step": 25389 }, { "epoch": 0.03, "learning_rate": 1.2341513576220631e-06, "loss": 0.8618, "step": 25390 }, { "epoch": 0.03, "learning_rate": 1.2336326810965681e-06, "loss": 0.8677, "step": 25391 }, { "epoch": 0.03, "learning_rate": 1.2331141064211804e-06, "loss": 0.8309, "step": 25392 }, { "epoch": 0.03, "learning_rate": 1.2325956336019262e-06, "loss": 0.8604, "step": 25393 }, { "epoch": 0.03, "learning_rate": 1.232077262644824e-06, "loss": 0.8185, "step": 25394 }, { "epoch": 0.03, "learning_rate": 1.2315589935559014e-06, "loss": 0.8659, "step": 25395 }, { "epoch": 0.03, "learning_rate": 1.2310408263411766e-06, "loss": 0.846, "step": 25396 }, { "epoch": 0.03, "learning_rate": 1.2305227610066727e-06, "loss": 0.8583, "step": 25397 }, { "epoch": 0.03, "learning_rate": 1.2300047975584063e-06, "loss": 0.848, "step": 25398 }, { "epoch": 0.03, "learning_rate": 1.2294869360023963e-06, "loss": 0.9079, "step": 25399 }, { "epoch": 0.03, "learning_rate": 1.2289691763446622e-06, "loss": 0.8175, "step": 25400 }, { "epoch": 0.03, "learning_rate": 1.2284515185912138e-06, "loss": 0.83, "step": 25401 }, { "epoch": 0.03, "learning_rate": 1.2279339627480657e-06, "loss": 0.8579, "step": 25402 }, { "epoch": 0.03, "learning_rate": 1.2274165088212364e-06, "loss": 0.8573, "step": 25403 }, { "epoch": 0.03, "learning_rate": 1.2268991568167364e-06, "loss": 0.841, "step": 25404 }, { "epoch": 0.03, "learning_rate": 1.2263819067405736e-06, "loss": 0.8339, "step": 25405 }, { "epoch": 0.03, "learning_rate": 1.2258647585987594e-06, "loss": 0.8192, "step": 25406 }, { "epoch": 0.03, "learning_rate": 1.2253477123973023e-06, "loss": 0.8664, "step": 25407 }, { "epoch": 0.03, "learning_rate": 1.2248307681422055e-06, "loss": 0.8368, "step": 25408 }, { "epoch": 0.03, "learning_rate": 1.2243139258394809e-06, "loss": 0.8296, "step": 25409 }, { "epoch": 0.03, "learning_rate": 1.2237971854951302e-06, "loss": 0.875, "step": 25410 }, { "epoch": 0.03, "learning_rate": 1.22328054711516e-06, "loss": 0.8714, "step": 25411 }, { "epoch": 0.03, "learning_rate": 1.222764010705567e-06, "loss": 0.8393, "step": 25412 }, { "epoch": 0.03, "learning_rate": 1.2222475762723574e-06, "loss": 0.8687, "step": 25413 }, { "epoch": 0.03, "learning_rate": 1.221731243821529e-06, "loss": 0.853, "step": 25414 }, { "epoch": 0.03, "learning_rate": 1.2212150133590816e-06, "loss": 0.8772, "step": 25415 }, { "epoch": 0.03, "learning_rate": 1.2206988848910128e-06, "loss": 0.8343, "step": 25416 }, { "epoch": 0.03, "learning_rate": 1.2201828584233187e-06, "loss": 0.8646, "step": 25417 }, { "epoch": 0.03, "learning_rate": 1.2196669339619981e-06, "loss": 0.8174, "step": 25418 }, { "epoch": 0.03, "learning_rate": 1.2191511115130395e-06, "loss": 0.861, "step": 25419 }, { "epoch": 0.03, "learning_rate": 1.2186353910824377e-06, "loss": 0.8843, "step": 25420 }, { "epoch": 0.03, "learning_rate": 1.2181197726761862e-06, "loss": 0.819, "step": 25421 }, { "epoch": 0.03, "learning_rate": 1.2176042563002744e-06, "loss": 0.8273, "step": 25422 }, { "epoch": 0.03, "learning_rate": 1.2170888419606919e-06, "loss": 0.8956, "step": 25423 }, { "epoch": 0.03, "learning_rate": 1.2165735296634262e-06, "loss": 0.8513, "step": 25424 }, { "epoch": 0.03, "learning_rate": 1.2160583194144682e-06, "loss": 0.8053, "step": 25425 }, { "epoch": 0.03, "learning_rate": 1.2155432112197974e-06, "loss": 0.8257, "step": 25426 }, { "epoch": 0.03, "learning_rate": 1.2150282050854001e-06, "loss": 0.8887, "step": 25427 }, { "epoch": 0.03, "learning_rate": 1.2145133010172639e-06, "loss": 0.8325, "step": 25428 }, { "epoch": 0.03, "learning_rate": 1.2139984990213705e-06, "loss": 0.8202, "step": 25429 }, { "epoch": 0.03, "learning_rate": 1.2134837991036975e-06, "loss": 0.8461, "step": 25430 }, { "epoch": 0.03, "learning_rate": 1.2129692012702254e-06, "loss": 0.8171, "step": 25431 }, { "epoch": 0.03, "learning_rate": 1.2124547055269343e-06, "loss": 0.8117, "step": 25432 }, { "epoch": 0.03, "learning_rate": 1.2119403118798023e-06, "loss": 0.8847, "step": 25433 }, { "epoch": 0.03, "learning_rate": 1.2114260203348039e-06, "loss": 0.8887, "step": 25434 }, { "epoch": 0.03, "learning_rate": 1.2109118308979162e-06, "loss": 0.842, "step": 25435 }, { "epoch": 0.03, "learning_rate": 1.2103977435751136e-06, "loss": 0.8638, "step": 25436 }, { "epoch": 0.03, "learning_rate": 1.2098837583723654e-06, "loss": 0.8344, "step": 25437 }, { "epoch": 0.03, "learning_rate": 1.2093698752956462e-06, "loss": 0.8557, "step": 25438 }, { "epoch": 0.03, "learning_rate": 1.2088560943509253e-06, "loss": 0.8507, "step": 25439 }, { "epoch": 0.03, "learning_rate": 1.2083424155441714e-06, "loss": 0.8686, "step": 25440 }, { "epoch": 0.03, "learning_rate": 1.2078288388813553e-06, "loss": 0.8239, "step": 25441 }, { "epoch": 0.03, "learning_rate": 1.207315364368441e-06, "loss": 0.8191, "step": 25442 }, { "epoch": 0.03, "learning_rate": 1.206801992011397e-06, "loss": 0.8422, "step": 25443 }, { "epoch": 0.03, "learning_rate": 1.2062887218161845e-06, "loss": 0.8419, "step": 25444 }, { "epoch": 0.03, "learning_rate": 1.205775553788766e-06, "loss": 0.8336, "step": 25445 }, { "epoch": 0.03, "learning_rate": 1.2052624879351105e-06, "loss": 0.8806, "step": 25446 }, { "epoch": 0.03, "learning_rate": 1.2047495242611718e-06, "loss": 0.8873, "step": 25447 }, { "epoch": 0.03, "learning_rate": 1.204236662772913e-06, "loss": 0.8605, "step": 25448 }, { "epoch": 0.03, "learning_rate": 1.2037239034762915e-06, "loss": 0.815, "step": 25449 }, { "epoch": 0.03, "learning_rate": 1.203211246377267e-06, "loss": 0.8472, "step": 25450 }, { "epoch": 0.03, "learning_rate": 1.2026986914817895e-06, "loss": 0.8939, "step": 25451 }, { "epoch": 0.03, "learning_rate": 1.2021862387958217e-06, "loss": 0.8381, "step": 25452 }, { "epoch": 0.03, "learning_rate": 1.2016738883253153e-06, "loss": 0.7979, "step": 25453 }, { "epoch": 0.03, "learning_rate": 1.2011616400762194e-06, "loss": 0.8459, "step": 25454 }, { "epoch": 0.03, "learning_rate": 1.2006494940544877e-06, "loss": 0.8661, "step": 25455 }, { "epoch": 0.03, "learning_rate": 1.2001374502660711e-06, "loss": 0.8261, "step": 25456 }, { "epoch": 0.03, "learning_rate": 1.1996255087169174e-06, "loss": 0.8023, "step": 25457 }, { "epoch": 0.03, "learning_rate": 1.1991136694129756e-06, "loss": 0.869, "step": 25458 }, { "epoch": 0.03, "learning_rate": 1.1986019323601917e-06, "loss": 0.8108, "step": 25459 }, { "epoch": 0.03, "learning_rate": 1.1980902975645137e-06, "loss": 0.8462, "step": 25460 }, { "epoch": 0.03, "learning_rate": 1.1975787650318816e-06, "loss": 0.823, "step": 25461 }, { "epoch": 0.03, "learning_rate": 1.1970673347682405e-06, "loss": 0.8402, "step": 25462 }, { "epoch": 0.03, "learning_rate": 1.1965560067795322e-06, "loss": 0.7884, "step": 25463 }, { "epoch": 0.03, "learning_rate": 1.196044781071698e-06, "loss": 0.8634, "step": 25464 }, { "epoch": 0.03, "learning_rate": 1.195533657650677e-06, "loss": 0.8094, "step": 25465 }, { "epoch": 0.03, "learning_rate": 1.195022636522407e-06, "loss": 0.8549, "step": 25466 }, { "epoch": 0.03, "learning_rate": 1.194511717692829e-06, "loss": 0.8744, "step": 25467 }, { "epoch": 0.03, "learning_rate": 1.1940009011678734e-06, "loss": 0.8657, "step": 25468 }, { "epoch": 0.03, "learning_rate": 1.1934901869534775e-06, "loss": 0.8031, "step": 25469 }, { "epoch": 0.03, "learning_rate": 1.1929795750555727e-06, "loss": 0.8722, "step": 25470 }, { "epoch": 0.03, "learning_rate": 1.1924690654800986e-06, "loss": 0.8374, "step": 25471 }, { "epoch": 0.03, "learning_rate": 1.1919586582329789e-06, "loss": 0.8524, "step": 25472 }, { "epoch": 0.03, "learning_rate": 1.191448353320146e-06, "loss": 0.8501, "step": 25473 }, { "epoch": 0.03, "learning_rate": 1.1909381507475292e-06, "loss": 0.869, "step": 25474 }, { "epoch": 0.03, "learning_rate": 1.1904280505210552e-06, "loss": 0.8573, "step": 25475 }, { "epoch": 0.03, "learning_rate": 1.1899180526466526e-06, "loss": 0.8585, "step": 25476 }, { "epoch": 0.03, "learning_rate": 1.1894081571302451e-06, "loss": 0.8841, "step": 25477 }, { "epoch": 0.03, "learning_rate": 1.1888983639777586e-06, "loss": 0.8482, "step": 25478 }, { "epoch": 0.03, "learning_rate": 1.1883886731951122e-06, "loss": 0.8794, "step": 25479 }, { "epoch": 0.03, "learning_rate": 1.1878790847882305e-06, "loss": 0.8392, "step": 25480 }, { "epoch": 0.03, "learning_rate": 1.187369598763034e-06, "loss": 0.8335, "step": 25481 }, { "epoch": 0.03, "learning_rate": 1.1868602151254405e-06, "loss": 0.8275, "step": 25482 }, { "epoch": 0.03, "learning_rate": 1.1863509338813695e-06, "loss": 0.8724, "step": 25483 }, { "epoch": 0.03, "learning_rate": 1.1858417550367374e-06, "loss": 0.8688, "step": 25484 }, { "epoch": 0.03, "learning_rate": 1.1853326785974628e-06, "loss": 0.8649, "step": 25485 }, { "epoch": 0.03, "learning_rate": 1.1848237045694555e-06, "loss": 0.8569, "step": 25486 }, { "epoch": 0.03, "learning_rate": 1.1843148329586317e-06, "loss": 0.8627, "step": 25487 }, { "epoch": 0.03, "learning_rate": 1.1838060637709025e-06, "loss": 0.8217, "step": 25488 }, { "epoch": 0.03, "learning_rate": 1.1832973970121796e-06, "loss": 0.8406, "step": 25489 }, { "epoch": 0.03, "learning_rate": 1.1827888326883719e-06, "loss": 0.8503, "step": 25490 }, { "epoch": 0.03, "learning_rate": 1.1822803708053899e-06, "loss": 0.833, "step": 25491 }, { "epoch": 0.03, "learning_rate": 1.1817720113691421e-06, "loss": 0.8561, "step": 25492 }, { "epoch": 0.03, "learning_rate": 1.1812637543855298e-06, "loss": 0.8737, "step": 25493 }, { "epoch": 0.03, "learning_rate": 1.1807555998604592e-06, "loss": 0.8788, "step": 25494 }, { "epoch": 0.03, "learning_rate": 1.1802475477998377e-06, "loss": 0.8576, "step": 25495 }, { "epoch": 0.03, "learning_rate": 1.1797395982095695e-06, "loss": 0.8317, "step": 25496 }, { "epoch": 0.03, "learning_rate": 1.1792317510955497e-06, "loss": 0.8798, "step": 25497 }, { "epoch": 0.03, "learning_rate": 1.1787240064636818e-06, "loss": 0.8393, "step": 25498 }, { "epoch": 0.03, "learning_rate": 1.1782163643198664e-06, "loss": 0.8502, "step": 25499 }, { "epoch": 0.03, "learning_rate": 1.1777088246699963e-06, "loss": 0.8289, "step": 25500 }, { "epoch": 0.03, "learning_rate": 1.177201387519974e-06, "loss": 0.8319, "step": 25501 }, { "epoch": 0.03, "learning_rate": 1.176694052875692e-06, "loss": 0.8694, "step": 25502 }, { "epoch": 0.03, "learning_rate": 1.1761868207430494e-06, "loss": 0.8341, "step": 25503 }, { "epoch": 0.03, "learning_rate": 1.1756796911279322e-06, "loss": 0.834, "step": 25504 }, { "epoch": 0.03, "learning_rate": 1.175172664036235e-06, "loss": 0.8274, "step": 25505 }, { "epoch": 0.03, "learning_rate": 1.1746657394738504e-06, "loss": 0.8635, "step": 25506 }, { "epoch": 0.03, "learning_rate": 1.1741589174466662e-06, "loss": 0.8356, "step": 25507 }, { "epoch": 0.03, "learning_rate": 1.173652197960571e-06, "loss": 0.8271, "step": 25508 }, { "epoch": 0.03, "learning_rate": 1.1731455810214531e-06, "loss": 0.8843, "step": 25509 }, { "epoch": 0.03, "learning_rate": 1.1726390666351994e-06, "loss": 0.8524, "step": 25510 }, { "epoch": 0.03, "learning_rate": 1.1721326548076917e-06, "loss": 0.8923, "step": 25511 }, { "epoch": 0.03, "learning_rate": 1.171626345544815e-06, "loss": 0.902, "step": 25512 }, { "epoch": 0.03, "learning_rate": 1.1711201388524517e-06, "loss": 0.8509, "step": 25513 }, { "epoch": 0.03, "learning_rate": 1.1706140347364837e-06, "loss": 0.8653, "step": 25514 }, { "epoch": 0.03, "learning_rate": 1.1701080332027893e-06, "loss": 0.8185, "step": 25515 }, { "epoch": 0.03, "learning_rate": 1.1696021342572495e-06, "loss": 0.8359, "step": 25516 }, { "epoch": 0.03, "learning_rate": 1.169096337905743e-06, "loss": 0.82, "step": 25517 }, { "epoch": 0.03, "learning_rate": 1.1685906441541406e-06, "loss": 0.8186, "step": 25518 }, { "epoch": 0.03, "learning_rate": 1.1680850530083231e-06, "loss": 0.8876, "step": 25519 }, { "epoch": 0.03, "learning_rate": 1.1675795644741627e-06, "loss": 0.8731, "step": 25520 }, { "epoch": 0.03, "learning_rate": 1.1670741785575346e-06, "loss": 0.7945, "step": 25521 }, { "epoch": 0.03, "learning_rate": 1.1665688952643062e-06, "loss": 0.8294, "step": 25522 }, { "epoch": 0.03, "learning_rate": 1.1660637146003507e-06, "loss": 0.883, "step": 25523 }, { "epoch": 0.03, "learning_rate": 1.1655586365715365e-06, "loss": 0.8478, "step": 25524 }, { "epoch": 0.03, "learning_rate": 1.1650536611837327e-06, "loss": 0.8653, "step": 25525 }, { "epoch": 0.03, "learning_rate": 1.1645487884428053e-06, "loss": 0.8837, "step": 25526 }, { "epoch": 0.03, "learning_rate": 1.164044018354621e-06, "loss": 0.847, "step": 25527 }, { "epoch": 0.03, "learning_rate": 1.1635393509250449e-06, "loss": 0.8464, "step": 25528 }, { "epoch": 0.03, "learning_rate": 1.1630347861599377e-06, "loss": 0.8056, "step": 25529 }, { "epoch": 0.03, "learning_rate": 1.1625303240651641e-06, "loss": 0.8413, "step": 25530 }, { "epoch": 0.03, "learning_rate": 1.1620259646465826e-06, "loss": 0.9116, "step": 25531 }, { "epoch": 0.03, "learning_rate": 1.161521707910056e-06, "loss": 0.8577, "step": 25532 }, { "epoch": 0.03, "learning_rate": 1.16101755386144e-06, "loss": 0.8392, "step": 25533 }, { "epoch": 0.03, "learning_rate": 1.160513502506595e-06, "loss": 0.8722, "step": 25534 }, { "epoch": 0.03, "learning_rate": 1.1600095538513767e-06, "loss": 0.837, "step": 25535 }, { "epoch": 0.03, "learning_rate": 1.159505707901637e-06, "loss": 0.8342, "step": 25536 }, { "epoch": 0.03, "learning_rate": 1.15900196466323e-06, "loss": 0.854, "step": 25537 }, { "epoch": 0.03, "learning_rate": 1.1584983241420145e-06, "loss": 0.8683, "step": 25538 }, { "epoch": 0.03, "learning_rate": 1.1579947863438345e-06, "loss": 0.8765, "step": 25539 }, { "epoch": 0.03, "learning_rate": 1.1574913512745446e-06, "loss": 0.8328, "step": 25540 }, { "epoch": 0.03, "learning_rate": 1.156988018939992e-06, "loss": 0.839, "step": 25541 }, { "epoch": 0.03, "learning_rate": 1.1564847893460274e-06, "loss": 0.8599, "step": 25542 }, { "epoch": 0.03, "learning_rate": 1.155981662498491e-06, "loss": 0.8626, "step": 25543 }, { "epoch": 0.03, "learning_rate": 1.1554786384032334e-06, "loss": 0.8662, "step": 25544 }, { "epoch": 0.03, "learning_rate": 1.1549757170661012e-06, "loss": 0.8496, "step": 25545 }, { "epoch": 0.03, "learning_rate": 1.1544728984929322e-06, "loss": 0.8435, "step": 25546 }, { "epoch": 0.03, "learning_rate": 1.1539701826895699e-06, "loss": 0.8721, "step": 25547 }, { "epoch": 0.03, "learning_rate": 1.1534675696618557e-06, "loss": 0.845, "step": 25548 }, { "epoch": 0.03, "learning_rate": 1.1529650594156283e-06, "loss": 0.8376, "step": 25549 }, { "epoch": 0.03, "learning_rate": 1.1524626519567262e-06, "loss": 0.8318, "step": 25550 }, { "epoch": 0.03, "learning_rate": 1.151960347290988e-06, "loss": 0.8795, "step": 25551 }, { "epoch": 0.04, "learning_rate": 1.151458145424249e-06, "loss": 0.8414, "step": 25552 }, { "epoch": 0.04, "learning_rate": 1.1509560463623426e-06, "loss": 0.8149, "step": 25553 }, { "epoch": 0.04, "learning_rate": 1.1504540501111017e-06, "loss": 0.8191, "step": 25554 }, { "epoch": 0.04, "learning_rate": 1.1499521566763616e-06, "loss": 0.9008, "step": 25555 }, { "epoch": 0.04, "learning_rate": 1.1494503660639511e-06, "loss": 0.8233, "step": 25556 }, { "epoch": 0.04, "learning_rate": 1.1489486782797e-06, "loss": 0.8141, "step": 25557 }, { "epoch": 0.04, "learning_rate": 1.148447093329439e-06, "loss": 0.8832, "step": 25558 }, { "epoch": 0.04, "learning_rate": 1.1479456112189957e-06, "loss": 0.8664, "step": 25559 }, { "epoch": 0.04, "learning_rate": 1.1474442319541945e-06, "loss": 0.8399, "step": 25560 }, { "epoch": 0.04, "learning_rate": 1.1469429555408607e-06, "loss": 0.8288, "step": 25561 }, { "epoch": 0.04, "learning_rate": 1.1464417819848172e-06, "loss": 0.8588, "step": 25562 }, { "epoch": 0.04, "learning_rate": 1.1459407112918919e-06, "loss": 0.8844, "step": 25563 }, { "epoch": 0.04, "learning_rate": 1.1454397434679022e-06, "loss": 0.8504, "step": 25564 }, { "epoch": 0.04, "learning_rate": 1.1449388785186677e-06, "loss": 0.8287, "step": 25565 }, { "epoch": 0.04, "learning_rate": 1.1444381164500106e-06, "loss": 0.8384, "step": 25566 }, { "epoch": 0.04, "learning_rate": 1.1439374572677475e-06, "loss": 0.8806, "step": 25567 }, { "epoch": 0.04, "learning_rate": 1.1434369009776947e-06, "loss": 0.8429, "step": 25568 }, { "epoch": 0.04, "learning_rate": 1.1429364475856674e-06, "loss": 0.8231, "step": 25569 }, { "epoch": 0.04, "learning_rate": 1.1424360970974846e-06, "loss": 0.8268, "step": 25570 }, { "epoch": 0.04, "learning_rate": 1.1419358495189525e-06, "loss": 0.865, "step": 25571 }, { "epoch": 0.04, "learning_rate": 1.1414357048558867e-06, "loss": 0.8683, "step": 25572 }, { "epoch": 0.04, "learning_rate": 1.1409356631140977e-06, "loss": 0.8772, "step": 25573 }, { "epoch": 0.04, "learning_rate": 1.1404357242993958e-06, "loss": 0.8365, "step": 25574 }, { "epoch": 0.04, "learning_rate": 1.1399358884175881e-06, "loss": 0.8454, "step": 25575 }, { "epoch": 0.04, "learning_rate": 1.1394361554744837e-06, "loss": 0.84, "step": 25576 }, { "epoch": 0.04, "learning_rate": 1.1389365254758888e-06, "loss": 0.8734, "step": 25577 }, { "epoch": 0.04, "learning_rate": 1.1384369984276044e-06, "loss": 0.8513, "step": 25578 }, { "epoch": 0.04, "learning_rate": 1.137937574335437e-06, "loss": 0.8381, "step": 25579 }, { "epoch": 0.04, "learning_rate": 1.1374382532051864e-06, "loss": 0.7963, "step": 25580 }, { "epoch": 0.04, "learning_rate": 1.13693903504266e-06, "loss": 0.8379, "step": 25581 }, { "epoch": 0.04, "learning_rate": 1.1364399198536535e-06, "loss": 0.8543, "step": 25582 }, { "epoch": 0.04, "learning_rate": 1.1359409076439653e-06, "loss": 0.8437, "step": 25583 }, { "epoch": 0.04, "learning_rate": 1.1354419984193955e-06, "loss": 0.8636, "step": 25584 }, { "epoch": 0.04, "learning_rate": 1.134943192185738e-06, "loss": 0.8375, "step": 25585 }, { "epoch": 0.04, "learning_rate": 1.134444488948786e-06, "loss": 0.9018, "step": 25586 }, { "epoch": 0.04, "learning_rate": 1.1339458887143396e-06, "loss": 0.9331, "step": 25587 }, { "epoch": 0.04, "learning_rate": 1.1334473914881905e-06, "loss": 0.8077, "step": 25588 }, { "epoch": 0.04, "learning_rate": 1.1329489972761276e-06, "loss": 0.8759, "step": 25589 }, { "epoch": 0.04, "learning_rate": 1.1324507060839419e-06, "loss": 0.8141, "step": 25590 }, { "epoch": 0.04, "learning_rate": 1.1319525179174228e-06, "loss": 0.8037, "step": 25591 }, { "epoch": 0.04, "learning_rate": 1.1314544327823596e-06, "loss": 0.8797, "step": 25592 }, { "epoch": 0.04, "learning_rate": 1.1309564506845382e-06, "loss": 0.8686, "step": 25593 }, { "epoch": 0.04, "learning_rate": 1.1304585716297444e-06, "loss": 0.8149, "step": 25594 }, { "epoch": 0.04, "learning_rate": 1.1299607956237646e-06, "loss": 0.8916, "step": 25595 }, { "epoch": 0.04, "learning_rate": 1.1294631226723784e-06, "loss": 0.8373, "step": 25596 }, { "epoch": 0.04, "learning_rate": 1.1289655527813704e-06, "loss": 0.8488, "step": 25597 }, { "epoch": 0.04, "learning_rate": 1.128468085956521e-06, "loss": 0.8826, "step": 25598 }, { "epoch": 0.04, "learning_rate": 1.1279707222036095e-06, "loss": 0.8064, "step": 25599 }, { "epoch": 0.04, "learning_rate": 1.1274734615284145e-06, "loss": 0.873, "step": 25600 }, { "epoch": 0.04, "learning_rate": 1.1269763039367143e-06, "loss": 0.8156, "step": 25601 }, { "epoch": 0.04, "learning_rate": 1.1264792494342858e-06, "loss": 0.8698, "step": 25602 }, { "epoch": 0.04, "learning_rate": 1.1259822980269008e-06, "loss": 0.8843, "step": 25603 }, { "epoch": 0.04, "learning_rate": 1.1254854497203339e-06, "loss": 0.852, "step": 25604 }, { "epoch": 0.04, "learning_rate": 1.124988704520359e-06, "loss": 0.8479, "step": 25605 }, { "epoch": 0.04, "learning_rate": 1.124492062432747e-06, "loss": 0.8172, "step": 25606 }, { "epoch": 0.04, "learning_rate": 1.1239955234632672e-06, "loss": 0.8406, "step": 25607 }, { "epoch": 0.04, "learning_rate": 1.12349908761769e-06, "loss": 0.8481, "step": 25608 }, { "epoch": 0.04, "learning_rate": 1.123002754901783e-06, "loss": 0.8079, "step": 25609 }, { "epoch": 0.04, "learning_rate": 1.1225065253213096e-06, "loss": 0.8507, "step": 25610 }, { "epoch": 0.04, "learning_rate": 1.1220103988820385e-06, "loss": 0.8761, "step": 25611 }, { "epoch": 0.04, "learning_rate": 1.1215143755897328e-06, "loss": 0.839, "step": 25612 }, { "epoch": 0.04, "learning_rate": 1.121018455450158e-06, "loss": 0.8833, "step": 25613 }, { "epoch": 0.04, "learning_rate": 1.1205226384690704e-06, "loss": 0.871, "step": 25614 }, { "epoch": 0.04, "learning_rate": 1.1200269246522343e-06, "loss": 0.8691, "step": 25615 }, { "epoch": 0.04, "learning_rate": 1.1195313140054086e-06, "loss": 0.8754, "step": 25616 }, { "epoch": 0.04, "learning_rate": 1.119035806534351e-06, "loss": 0.8351, "step": 25617 }, { "epoch": 0.04, "learning_rate": 1.1185404022448177e-06, "loss": 0.8431, "step": 25618 }, { "epoch": 0.04, "learning_rate": 1.1180451011425664e-06, "loss": 0.8619, "step": 25619 }, { "epoch": 0.04, "learning_rate": 1.1175499032333514e-06, "loss": 0.8528, "step": 25620 }, { "epoch": 0.04, "learning_rate": 1.1170548085229239e-06, "loss": 0.8328, "step": 25621 }, { "epoch": 0.04, "learning_rate": 1.1165598170170367e-06, "loss": 0.8615, "step": 25622 }, { "epoch": 0.04, "learning_rate": 1.116064928721442e-06, "loss": 0.8639, "step": 25623 }, { "epoch": 0.04, "learning_rate": 1.1155701436418887e-06, "loss": 0.8122, "step": 25624 }, { "epoch": 0.04, "learning_rate": 1.1150754617841252e-06, "loss": 0.8427, "step": 25625 }, { "epoch": 0.04, "learning_rate": 1.1145808831538996e-06, "loss": 0.8705, "step": 25626 }, { "epoch": 0.04, "learning_rate": 1.1140864077569603e-06, "loss": 0.8339, "step": 25627 }, { "epoch": 0.04, "learning_rate": 1.1135920355990471e-06, "loss": 0.8424, "step": 25628 }, { "epoch": 0.04, "learning_rate": 1.1130977666859033e-06, "loss": 0.836, "step": 25629 }, { "epoch": 0.04, "learning_rate": 1.1126036010232798e-06, "loss": 0.8351, "step": 25630 }, { "epoch": 0.04, "learning_rate": 1.11210953861691e-06, "loss": 0.8242, "step": 25631 }, { "epoch": 0.04, "learning_rate": 1.111615579472537e-06, "loss": 0.8431, "step": 25632 }, { "epoch": 0.04, "learning_rate": 1.1111217235958982e-06, "loss": 0.8573, "step": 25633 }, { "epoch": 0.04, "learning_rate": 1.110627970992736e-06, "loss": 0.8164, "step": 25634 }, { "epoch": 0.04, "learning_rate": 1.1101343216687787e-06, "loss": 0.904, "step": 25635 }, { "epoch": 0.04, "learning_rate": 1.109640775629769e-06, "loss": 0.8224, "step": 25636 }, { "epoch": 0.04, "learning_rate": 1.1091473328814405e-06, "loss": 0.7939, "step": 25637 }, { "epoch": 0.04, "learning_rate": 1.1086539934295215e-06, "loss": 0.845, "step": 25638 }, { "epoch": 0.04, "learning_rate": 1.1081607572797471e-06, "loss": 0.8671, "step": 25639 }, { "epoch": 0.04, "learning_rate": 1.107667624437847e-06, "loss": 0.872, "step": 25640 }, { "epoch": 0.04, "learning_rate": 1.1071745949095502e-06, "loss": 0.8424, "step": 25641 }, { "epoch": 0.04, "learning_rate": 1.1066816687005866e-06, "loss": 0.8536, "step": 25642 }, { "epoch": 0.04, "learning_rate": 1.1061888458166814e-06, "loss": 0.8478, "step": 25643 }, { "epoch": 0.04, "learning_rate": 1.1056961262635623e-06, "loss": 0.8159, "step": 25644 }, { "epoch": 0.04, "learning_rate": 1.1052035100469515e-06, "loss": 0.865, "step": 25645 }, { "epoch": 0.04, "learning_rate": 1.1047109971725723e-06, "loss": 0.8567, "step": 25646 }, { "epoch": 0.04, "learning_rate": 1.1042185876461486e-06, "loss": 0.8224, "step": 25647 }, { "epoch": 0.04, "learning_rate": 1.1037262814733996e-06, "loss": 0.8142, "step": 25648 }, { "epoch": 0.04, "learning_rate": 1.103234078660047e-06, "loss": 0.8387, "step": 25649 }, { "epoch": 0.04, "learning_rate": 1.1027419792118088e-06, "loss": 0.8195, "step": 25650 }, { "epoch": 0.04, "learning_rate": 1.1022499831344035e-06, "loss": 0.8699, "step": 25651 }, { "epoch": 0.04, "learning_rate": 1.1017580904335434e-06, "loss": 0.8471, "step": 25652 }, { "epoch": 0.04, "learning_rate": 1.1012663011149438e-06, "loss": 0.806, "step": 25653 }, { "epoch": 0.04, "learning_rate": 1.1007746151843225e-06, "loss": 0.8459, "step": 25654 }, { "epoch": 0.04, "learning_rate": 1.1002830326473923e-06, "loss": 0.8519, "step": 25655 }, { "epoch": 0.04, "learning_rate": 1.0997915535098603e-06, "loss": 0.8537, "step": 25656 }, { "epoch": 0.04, "learning_rate": 1.099300177777438e-06, "loss": 0.8515, "step": 25657 }, { "epoch": 0.04, "learning_rate": 1.0988089054558348e-06, "loss": 0.8134, "step": 25658 }, { "epoch": 0.04, "learning_rate": 1.0983177365507591e-06, "loss": 0.8664, "step": 25659 }, { "epoch": 0.04, "learning_rate": 1.0978266710679163e-06, "loss": 0.8907, "step": 25660 }, { "epoch": 0.04, "learning_rate": 1.097335709013012e-06, "loss": 0.8818, "step": 25661 }, { "epoch": 0.04, "learning_rate": 1.0968448503917539e-06, "loss": 0.825, "step": 25662 }, { "epoch": 0.04, "learning_rate": 1.0963540952098385e-06, "loss": 0.8715, "step": 25663 }, { "epoch": 0.04, "learning_rate": 1.095863443472972e-06, "loss": 0.8778, "step": 25664 }, { "epoch": 0.04, "learning_rate": 1.0953728951868525e-06, "loss": 0.8706, "step": 25665 }, { "epoch": 0.04, "learning_rate": 1.0948824503571808e-06, "loss": 0.8652, "step": 25666 }, { "epoch": 0.04, "learning_rate": 1.0943921089896547e-06, "loss": 0.8889, "step": 25667 }, { "epoch": 0.04, "learning_rate": 1.0939018710899707e-06, "loss": 0.8343, "step": 25668 }, { "epoch": 0.04, "learning_rate": 1.0934117366638275e-06, "loss": 0.8314, "step": 25669 }, { "epoch": 0.04, "learning_rate": 1.0929217057169139e-06, "loss": 0.8971, "step": 25670 }, { "epoch": 0.04, "learning_rate": 1.0924317782549277e-06, "loss": 0.8595, "step": 25671 }, { "epoch": 0.04, "learning_rate": 1.0919419542835574e-06, "loss": 0.8446, "step": 25672 }, { "epoch": 0.04, "learning_rate": 1.0914522338084998e-06, "loss": 0.8242, "step": 25673 }, { "epoch": 0.04, "learning_rate": 1.0909626168354393e-06, "loss": 0.8327, "step": 25674 }, { "epoch": 0.04, "learning_rate": 1.0904731033700654e-06, "loss": 0.8592, "step": 25675 }, { "epoch": 0.04, "learning_rate": 1.0899836934180685e-06, "loss": 0.8446, "step": 25676 }, { "epoch": 0.04, "learning_rate": 1.0894943869851305e-06, "loss": 0.8461, "step": 25677 }, { "epoch": 0.04, "learning_rate": 1.0890051840769356e-06, "loss": 0.8377, "step": 25678 }, { "epoch": 0.04, "learning_rate": 1.0885160846991727e-06, "loss": 0.8697, "step": 25679 }, { "epoch": 0.04, "learning_rate": 1.0880270888575239e-06, "loss": 0.8538, "step": 25680 }, { "epoch": 0.04, "learning_rate": 1.0875381965576647e-06, "loss": 0.8728, "step": 25681 }, { "epoch": 0.04, "learning_rate": 1.0870494078052796e-06, "loss": 0.8635, "step": 25682 }, { "epoch": 0.04, "learning_rate": 1.0865607226060471e-06, "loss": 0.8147, "step": 25683 }, { "epoch": 0.04, "learning_rate": 1.0860721409656438e-06, "loss": 0.859, "step": 25684 }, { "epoch": 0.04, "learning_rate": 1.0855836628897465e-06, "loss": 0.8425, "step": 25685 }, { "epoch": 0.04, "learning_rate": 1.0850952883840304e-06, "loss": 0.8579, "step": 25686 }, { "epoch": 0.04, "learning_rate": 1.0846070174541724e-06, "loss": 0.8498, "step": 25687 }, { "epoch": 0.04, "learning_rate": 1.084118850105841e-06, "loss": 0.8465, "step": 25688 }, { "epoch": 0.04, "learning_rate": 1.0836307863447093e-06, "loss": 0.8551, "step": 25689 }, { "epoch": 0.04, "learning_rate": 1.0831428261764488e-06, "loss": 0.8679, "step": 25690 }, { "epoch": 0.04, "learning_rate": 1.0826549696067268e-06, "loss": 0.8585, "step": 25691 }, { "epoch": 0.04, "learning_rate": 1.0821672166412133e-06, "loss": 0.878, "step": 25692 }, { "epoch": 0.04, "learning_rate": 1.081679567285575e-06, "loss": 0.7961, "step": 25693 }, { "epoch": 0.04, "learning_rate": 1.0811920215454785e-06, "loss": 0.8644, "step": 25694 }, { "epoch": 0.04, "learning_rate": 1.0807045794265858e-06, "loss": 0.8304, "step": 25695 }, { "epoch": 0.04, "learning_rate": 1.0802172409345578e-06, "loss": 0.8257, "step": 25696 }, { "epoch": 0.04, "learning_rate": 1.079730006075065e-06, "loss": 0.8468, "step": 25697 }, { "epoch": 0.04, "learning_rate": 1.0792428748537597e-06, "loss": 0.8335, "step": 25698 }, { "epoch": 0.04, "learning_rate": 1.078755847276306e-06, "loss": 0.8641, "step": 25699 }, { "epoch": 0.04, "learning_rate": 1.0782689233483613e-06, "loss": 0.8294, "step": 25700 }, { "epoch": 0.04, "learning_rate": 1.0777821030755842e-06, "loss": 0.8336, "step": 25701 }, { "epoch": 0.04, "learning_rate": 1.077295386463626e-06, "loss": 0.8304, "step": 25702 }, { "epoch": 0.04, "learning_rate": 1.0768087735181465e-06, "loss": 0.8732, "step": 25703 }, { "epoch": 0.04, "learning_rate": 1.0763222642447968e-06, "loss": 0.808, "step": 25704 }, { "epoch": 0.04, "learning_rate": 1.0758358586492334e-06, "loss": 0.8466, "step": 25705 }, { "epoch": 0.04, "learning_rate": 1.0753495567371009e-06, "loss": 0.8473, "step": 25706 }, { "epoch": 0.04, "learning_rate": 1.0748633585140534e-06, "loss": 0.8832, "step": 25707 }, { "epoch": 0.04, "learning_rate": 1.0743772639857387e-06, "loss": 0.8353, "step": 25708 }, { "epoch": 0.04, "learning_rate": 1.0738912731578054e-06, "loss": 0.8282, "step": 25709 }, { "epoch": 0.04, "learning_rate": 1.0734053860358983e-06, "loss": 0.8283, "step": 25710 }, { "epoch": 0.04, "learning_rate": 1.0729196026256627e-06, "loss": 0.8155, "step": 25711 }, { "epoch": 0.04, "learning_rate": 1.0724339229327462e-06, "loss": 0.8826, "step": 25712 }, { "epoch": 0.04, "learning_rate": 1.0719483469627856e-06, "loss": 0.8263, "step": 25713 }, { "epoch": 0.04, "learning_rate": 1.0714628747214263e-06, "loss": 0.8589, "step": 25714 }, { "epoch": 0.04, "learning_rate": 1.0709775062143069e-06, "loss": 0.8605, "step": 25715 }, { "epoch": 0.04, "learning_rate": 1.0704922414470687e-06, "loss": 0.88, "step": 25716 }, { "epoch": 0.04, "learning_rate": 1.070007080425347e-06, "loss": 0.8301, "step": 25717 }, { "epoch": 0.04, "learning_rate": 1.0695220231547799e-06, "loss": 0.883, "step": 25718 }, { "epoch": 0.04, "learning_rate": 1.0690370696410057e-06, "loss": 0.8599, "step": 25719 }, { "epoch": 0.04, "learning_rate": 1.0685522198896536e-06, "loss": 0.8922, "step": 25720 }, { "epoch": 0.04, "learning_rate": 1.0680674739063569e-06, "loss": 0.804, "step": 25721 }, { "epoch": 0.04, "learning_rate": 1.067582831696754e-06, "loss": 0.8829, "step": 25722 }, { "epoch": 0.04, "learning_rate": 1.0670982932664698e-06, "loss": 0.8301, "step": 25723 }, { "epoch": 0.04, "learning_rate": 1.0666138586211339e-06, "loss": 0.8605, "step": 25724 }, { "epoch": 0.04, "learning_rate": 1.0661295277663763e-06, "loss": 0.8412, "step": 25725 }, { "epoch": 0.04, "learning_rate": 1.065645300707825e-06, "loss": 0.8582, "step": 25726 }, { "epoch": 0.04, "learning_rate": 1.065161177451104e-06, "loss": 0.8379, "step": 25727 }, { "epoch": 0.04, "learning_rate": 1.0646771580018377e-06, "loss": 0.8618, "step": 25728 }, { "epoch": 0.04, "learning_rate": 1.064193242365653e-06, "loss": 0.8299, "step": 25729 }, { "epoch": 0.04, "learning_rate": 1.0637094305481688e-06, "loss": 0.8505, "step": 25730 }, { "epoch": 0.04, "learning_rate": 1.0632257225550058e-06, "loss": 0.9128, "step": 25731 }, { "epoch": 0.04, "learning_rate": 1.0627421183917852e-06, "loss": 0.8363, "step": 25732 }, { "epoch": 0.04, "learning_rate": 1.062258618064126e-06, "loss": 0.8541, "step": 25733 }, { "epoch": 0.04, "learning_rate": 1.0617752215776444e-06, "loss": 0.8427, "step": 25734 }, { "epoch": 0.04, "learning_rate": 1.0612919289379586e-06, "loss": 0.8529, "step": 25735 }, { "epoch": 0.04, "learning_rate": 1.060808740150684e-06, "loss": 0.8282, "step": 25736 }, { "epoch": 0.04, "learning_rate": 1.0603256552214303e-06, "loss": 0.8761, "step": 25737 }, { "epoch": 0.04, "learning_rate": 1.0598426741558132e-06, "loss": 0.837, "step": 25738 }, { "epoch": 0.04, "learning_rate": 1.0593597969594427e-06, "loss": 0.8549, "step": 25739 }, { "epoch": 0.04, "learning_rate": 1.058877023637931e-06, "loss": 0.8903, "step": 25740 }, { "epoch": 0.04, "learning_rate": 1.0583943541968856e-06, "loss": 0.8556, "step": 25741 }, { "epoch": 0.04, "learning_rate": 1.0579117886419144e-06, "loss": 0.7829, "step": 25742 }, { "epoch": 0.04, "learning_rate": 1.0574293269786272e-06, "loss": 0.8297, "step": 25743 }, { "epoch": 0.04, "learning_rate": 1.0569469692126233e-06, "loss": 0.837, "step": 25744 }, { "epoch": 0.04, "learning_rate": 1.0564647153495089e-06, "loss": 0.833, "step": 25745 }, { "epoch": 0.04, "learning_rate": 1.0559825653948896e-06, "loss": 0.8312, "step": 25746 }, { "epoch": 0.04, "learning_rate": 1.0555005193543676e-06, "loss": 0.8461, "step": 25747 }, { "epoch": 0.04, "learning_rate": 1.0550185772335397e-06, "loss": 0.8335, "step": 25748 }, { "epoch": 0.04, "learning_rate": 1.0545367390380068e-06, "loss": 0.811, "step": 25749 }, { "epoch": 0.04, "learning_rate": 1.0540550047733667e-06, "loss": 0.8232, "step": 25750 }, { "epoch": 0.04, "learning_rate": 1.053573374445217e-06, "loss": 0.8791, "step": 25751 }, { "epoch": 0.04, "learning_rate": 1.0530918480591544e-06, "loss": 0.8393, "step": 25752 }, { "epoch": 0.04, "learning_rate": 1.0526104256207715e-06, "loss": 0.8844, "step": 25753 }, { "epoch": 0.04, "learning_rate": 1.0521291071356643e-06, "loss": 0.8557, "step": 25754 }, { "epoch": 0.04, "learning_rate": 1.0516478926094208e-06, "loss": 0.8226, "step": 25755 }, { "epoch": 0.04, "learning_rate": 1.0511667820476346e-06, "loss": 0.8724, "step": 25756 }, { "epoch": 0.04, "learning_rate": 1.0506857754558942e-06, "loss": 0.8396, "step": 25757 }, { "epoch": 0.04, "learning_rate": 1.0502048728397896e-06, "loss": 0.8589, "step": 25758 }, { "epoch": 0.04, "learning_rate": 1.0497240742049063e-06, "loss": 0.8797, "step": 25759 }, { "epoch": 0.04, "learning_rate": 1.0492433795568314e-06, "loss": 0.8423, "step": 25760 }, { "epoch": 0.04, "learning_rate": 1.048762788901152e-06, "loss": 0.8197, "step": 25761 }, { "epoch": 0.04, "learning_rate": 1.0482823022434463e-06, "loss": 0.8666, "step": 25762 }, { "epoch": 0.04, "learning_rate": 1.0478019195893008e-06, "loss": 0.8805, "step": 25763 }, { "epoch": 0.04, "learning_rate": 1.0473216409442931e-06, "loss": 0.8162, "step": 25764 }, { "epoch": 0.04, "learning_rate": 1.0468414663140103e-06, "loss": 0.8423, "step": 25765 }, { "epoch": 0.04, "learning_rate": 1.0463613957040242e-06, "loss": 0.8482, "step": 25766 }, { "epoch": 0.04, "learning_rate": 1.0458814291199148e-06, "loss": 0.871, "step": 25767 }, { "epoch": 0.04, "learning_rate": 1.0454015665672613e-06, "loss": 0.8288, "step": 25768 }, { "epoch": 0.04, "learning_rate": 1.0449218080516333e-06, "loss": 0.82, "step": 25769 }, { "epoch": 0.04, "learning_rate": 1.0444421535786053e-06, "loss": 0.8655, "step": 25770 }, { "epoch": 0.04, "learning_rate": 1.0439626031537554e-06, "loss": 0.8984, "step": 25771 }, { "epoch": 0.04, "learning_rate": 1.043483156782653e-06, "loss": 0.8106, "step": 25772 }, { "epoch": 0.04, "learning_rate": 1.0430038144708665e-06, "loss": 0.8018, "step": 25773 }, { "epoch": 0.04, "learning_rate": 1.0425245762239654e-06, "loss": 0.8713, "step": 25774 }, { "epoch": 0.04, "learning_rate": 1.042045442047518e-06, "loss": 0.817, "step": 25775 }, { "epoch": 0.04, "learning_rate": 1.0415664119470925e-06, "loss": 0.8463, "step": 25776 }, { "epoch": 0.04, "learning_rate": 1.0410874859282517e-06, "loss": 0.8292, "step": 25777 }, { "epoch": 0.04, "learning_rate": 1.0406086639965629e-06, "loss": 0.8871, "step": 25778 }, { "epoch": 0.04, "learning_rate": 1.0401299461575886e-06, "loss": 0.8623, "step": 25779 }, { "epoch": 0.04, "learning_rate": 1.039651332416889e-06, "loss": 0.8438, "step": 25780 }, { "epoch": 0.04, "learning_rate": 1.0391728227800246e-06, "loss": 0.869, "step": 25781 }, { "epoch": 0.04, "learning_rate": 1.0386944172525559e-06, "loss": 0.8797, "step": 25782 }, { "epoch": 0.04, "learning_rate": 1.0382161158400405e-06, "loss": 0.8448, "step": 25783 }, { "epoch": 0.04, "learning_rate": 1.0377379185480375e-06, "loss": 0.8505, "step": 25784 }, { "epoch": 0.04, "learning_rate": 1.0372598253820998e-06, "loss": 0.8102, "step": 25785 }, { "epoch": 0.04, "learning_rate": 1.0367818363477855e-06, "loss": 0.8407, "step": 25786 }, { "epoch": 0.04, "learning_rate": 1.0363039514506446e-06, "loss": 0.8511, "step": 25787 }, { "epoch": 0.04, "learning_rate": 1.035826170696229e-06, "loss": 0.8548, "step": 25788 }, { "epoch": 0.04, "learning_rate": 1.0353484940900949e-06, "loss": 0.8866, "step": 25789 }, { "epoch": 0.04, "learning_rate": 1.034870921637786e-06, "loss": 0.8669, "step": 25790 }, { "epoch": 0.04, "learning_rate": 1.034393453344854e-06, "loss": 0.8633, "step": 25791 }, { "epoch": 0.04, "learning_rate": 1.0339160892168453e-06, "loss": 0.8393, "step": 25792 }, { "epoch": 0.04, "learning_rate": 1.033438829259309e-06, "loss": 0.8283, "step": 25793 }, { "epoch": 0.04, "learning_rate": 1.0329616734777826e-06, "loss": 0.8719, "step": 25794 }, { "epoch": 0.04, "learning_rate": 1.0324846218778174e-06, "loss": 0.8556, "step": 25795 }, { "epoch": 0.04, "learning_rate": 1.0320076744649532e-06, "loss": 0.8843, "step": 25796 }, { "epoch": 0.04, "learning_rate": 1.0315308312447325e-06, "loss": 0.8645, "step": 25797 }, { "epoch": 0.04, "learning_rate": 1.031054092222693e-06, "loss": 0.8434, "step": 25798 }, { "epoch": 0.04, "learning_rate": 1.0305774574043737e-06, "loss": 0.8579, "step": 25799 }, { "epoch": 0.04, "learning_rate": 1.0301009267953145e-06, "loss": 0.8654, "step": 25800 }, { "epoch": 0.04, "learning_rate": 1.0296245004010508e-06, "loss": 0.8584, "step": 25801 }, { "epoch": 0.04, "learning_rate": 1.0291481782271174e-06, "loss": 0.8382, "step": 25802 }, { "epoch": 0.04, "learning_rate": 1.0286719602790486e-06, "loss": 0.8785, "step": 25803 }, { "epoch": 0.04, "learning_rate": 1.0281958465623788e-06, "loss": 0.8642, "step": 25804 }, { "epoch": 0.04, "learning_rate": 1.027719837082637e-06, "loss": 0.8376, "step": 25805 }, { "epoch": 0.04, "learning_rate": 1.0272439318453553e-06, "loss": 0.8014, "step": 25806 }, { "epoch": 0.04, "learning_rate": 1.0267681308560607e-06, "loss": 0.844, "step": 25807 }, { "epoch": 0.04, "learning_rate": 1.026292434120284e-06, "loss": 0.8575, "step": 25808 }, { "epoch": 0.04, "learning_rate": 1.025816841643551e-06, "loss": 0.8583, "step": 25809 }, { "epoch": 0.04, "learning_rate": 1.0253413534313872e-06, "loss": 0.8446, "step": 25810 }, { "epoch": 0.04, "learning_rate": 1.0248659694893183e-06, "loss": 0.8454, "step": 25811 }, { "epoch": 0.04, "learning_rate": 1.024390689822864e-06, "loss": 0.8331, "step": 25812 }, { "epoch": 0.04, "learning_rate": 1.0239155144375467e-06, "loss": 0.8349, "step": 25813 }, { "epoch": 0.04, "learning_rate": 1.0234404433388922e-06, "loss": 0.8608, "step": 25814 }, { "epoch": 0.04, "learning_rate": 1.0229654765324147e-06, "loss": 0.8302, "step": 25815 }, { "epoch": 0.04, "learning_rate": 1.0224906140236345e-06, "loss": 0.8628, "step": 25816 }, { "epoch": 0.04, "learning_rate": 1.0220158558180692e-06, "loss": 0.8354, "step": 25817 }, { "epoch": 0.04, "learning_rate": 1.0215412019212323e-06, "loss": 0.8205, "step": 25818 }, { "epoch": 0.04, "learning_rate": 1.0210666523386414e-06, "loss": 0.8656, "step": 25819 }, { "epoch": 0.04, "learning_rate": 1.0205922070758078e-06, "loss": 0.8667, "step": 25820 }, { "epoch": 0.04, "learning_rate": 1.0201178661382472e-06, "loss": 0.849, "step": 25821 }, { "epoch": 0.04, "learning_rate": 1.019643629531466e-06, "loss": 0.8734, "step": 25822 }, { "epoch": 0.04, "learning_rate": 1.0191694972609756e-06, "loss": 0.849, "step": 25823 }, { "epoch": 0.04, "learning_rate": 1.0186954693322847e-06, "loss": 0.8168, "step": 25824 }, { "epoch": 0.04, "learning_rate": 1.0182215457509014e-06, "loss": 0.8411, "step": 25825 }, { "epoch": 0.04, "learning_rate": 1.0177477265223323e-06, "loss": 0.8878, "step": 25826 }, { "epoch": 0.04, "learning_rate": 1.0172740116520807e-06, "loss": 0.8571, "step": 25827 }, { "epoch": 0.04, "learning_rate": 1.0168004011456533e-06, "loss": 0.8096, "step": 25828 }, { "epoch": 0.04, "learning_rate": 1.0163268950085493e-06, "loss": 0.8279, "step": 25829 }, { "epoch": 0.04, "learning_rate": 1.0158534932462704e-06, "loss": 0.8538, "step": 25830 }, { "epoch": 0.04, "learning_rate": 1.0153801958643161e-06, "loss": 0.8615, "step": 25831 }, { "epoch": 0.04, "learning_rate": 1.0149070028681906e-06, "loss": 0.8248, "step": 25832 }, { "epoch": 0.04, "learning_rate": 1.0144339142633853e-06, "loss": 0.855, "step": 25833 }, { "epoch": 0.04, "learning_rate": 1.0139609300553999e-06, "loss": 0.8865, "step": 25834 }, { "epoch": 0.04, "learning_rate": 1.0134880502497303e-06, "loss": 0.8688, "step": 25835 }, { "epoch": 0.04, "learning_rate": 1.0130152748518674e-06, "loss": 0.8996, "step": 25836 }, { "epoch": 0.04, "learning_rate": 1.0125426038673036e-06, "loss": 0.7918, "step": 25837 }, { "epoch": 0.04, "learning_rate": 1.0120700373015357e-06, "loss": 0.8239, "step": 25838 }, { "epoch": 0.04, "learning_rate": 1.0115975751600526e-06, "loss": 0.8522, "step": 25839 }, { "epoch": 0.04, "learning_rate": 1.0111252174483399e-06, "loss": 0.8268, "step": 25840 }, { "epoch": 0.04, "learning_rate": 1.0106529641718877e-06, "loss": 0.8244, "step": 25841 }, { "epoch": 0.04, "learning_rate": 1.0101808153361826e-06, "loss": 0.8643, "step": 25842 }, { "epoch": 0.04, "learning_rate": 1.0097087709467102e-06, "loss": 0.8663, "step": 25843 }, { "epoch": 0.04, "learning_rate": 1.0092368310089551e-06, "loss": 0.8685, "step": 25844 }, { "epoch": 0.04, "learning_rate": 1.0087649955284007e-06, "loss": 0.8085, "step": 25845 }, { "epoch": 0.04, "learning_rate": 1.0082932645105303e-06, "loss": 0.8065, "step": 25846 }, { "epoch": 0.04, "learning_rate": 1.0078216379608218e-06, "loss": 0.848, "step": 25847 }, { "epoch": 0.04, "learning_rate": 1.0073501158847544e-06, "loss": 0.8232, "step": 25848 }, { "epoch": 0.04, "learning_rate": 1.0068786982878087e-06, "loss": 0.8617, "step": 25849 }, { "epoch": 0.04, "learning_rate": 1.0064073851754597e-06, "loss": 0.845, "step": 25850 }, { "epoch": 0.04, "learning_rate": 1.005936176553185e-06, "loss": 0.851, "step": 25851 }, { "epoch": 0.04, "learning_rate": 1.0054650724264592e-06, "loss": 0.8709, "step": 25852 }, { "epoch": 0.05, "learning_rate": 1.0049940728007557e-06, "loss": 0.8822, "step": 25853 }, { "epoch": 0.05, "learning_rate": 1.0045231776815445e-06, "loss": 0.8403, "step": 25854 }, { "epoch": 0.05, "learning_rate": 1.004052387074299e-06, "loss": 0.8211, "step": 25855 }, { "epoch": 0.05, "learning_rate": 1.0035817009844851e-06, "loss": 0.8347, "step": 25856 }, { "epoch": 0.05, "learning_rate": 1.0031111194175792e-06, "loss": 0.7996, "step": 25857 }, { "epoch": 0.05, "learning_rate": 1.0026406423790413e-06, "loss": 0.8661, "step": 25858 }, { "epoch": 0.05, "learning_rate": 1.0021702698743408e-06, "loss": 0.8791, "step": 25859 }, { "epoch": 0.05, "learning_rate": 1.001700001908943e-06, "loss": 0.8923, "step": 25860 }, { "epoch": 0.05, "learning_rate": 1.0012298384883067e-06, "loss": 0.7997, "step": 25861 }, { "epoch": 0.05, "learning_rate": 1.0007597796179003e-06, "loss": 0.8568, "step": 25862 }, { "epoch": 0.05, "learning_rate": 1.0002898253031834e-06, "loss": 0.869, "step": 25863 }, { "epoch": 0.05, "learning_rate": 9.998199755496173e-07, "loss": 0.8427, "step": 25864 }, { "epoch": 0.05, "learning_rate": 9.993502303626567e-07, "loss": 0.8578, "step": 25865 }, { "epoch": 0.05, "learning_rate": 9.988805897477627e-07, "loss": 0.8392, "step": 25866 }, { "epoch": 0.05, "learning_rate": 9.984110537103896e-07, "loss": 0.825, "step": 25867 }, { "epoch": 0.05, "learning_rate": 9.979416222559945e-07, "loss": 0.8374, "step": 25868 }, { "epoch": 0.05, "learning_rate": 9.974722953900307e-07, "loss": 0.8235, "step": 25869 }, { "epoch": 0.05, "learning_rate": 9.970030731179514e-07, "loss": 0.8365, "step": 25870 }, { "epoch": 0.05, "learning_rate": 9.96533955445208e-07, "loss": 0.8943, "step": 25871 }, { "epoch": 0.05, "learning_rate": 9.960649423772494e-07, "loss": 0.8583, "step": 25872 }, { "epoch": 0.05, "learning_rate": 9.955960339195258e-07, "loss": 0.8405, "step": 25873 }, { "epoch": 0.05, "learning_rate": 9.95127230077486e-07, "loss": 0.8415, "step": 25874 }, { "epoch": 0.05, "learning_rate": 9.946585308565749e-07, "loss": 0.8576, "step": 25875 }, { "epoch": 0.05, "learning_rate": 9.941899362622387e-07, "loss": 0.8612, "step": 25876 }, { "epoch": 0.05, "learning_rate": 9.937214462999223e-07, "loss": 0.837, "step": 25877 }, { "epoch": 0.05, "learning_rate": 9.9325306097507e-07, "loss": 0.8698, "step": 25878 }, { "epoch": 0.05, "learning_rate": 9.9278478029312e-07, "loss": 0.8103, "step": 25879 }, { "epoch": 0.05, "learning_rate": 9.923166042595134e-07, "loss": 0.8497, "step": 25880 }, { "epoch": 0.05, "learning_rate": 9.918485328796946e-07, "loss": 0.8583, "step": 25881 }, { "epoch": 0.05, "learning_rate": 9.913805661590958e-07, "loss": 0.835, "step": 25882 }, { "epoch": 0.05, "learning_rate": 9.909127041031564e-07, "loss": 0.8626, "step": 25883 }, { "epoch": 0.05, "learning_rate": 9.904449467173127e-07, "loss": 0.8204, "step": 25884 }, { "epoch": 0.05, "learning_rate": 9.899772940070008e-07, "loss": 0.8819, "step": 25885 }, { "epoch": 0.05, "learning_rate": 9.89509745977648e-07, "loss": 0.8798, "step": 25886 }, { "epoch": 0.05, "learning_rate": 9.890423026346919e-07, "loss": 0.8884, "step": 25887 }, { "epoch": 0.05, "learning_rate": 9.885749639835618e-07, "loss": 0.8386, "step": 25888 }, { "epoch": 0.05, "learning_rate": 9.881077300296893e-07, "loss": 0.8749, "step": 25889 }, { "epoch": 0.05, "learning_rate": 9.876406007784989e-07, "loss": 0.8464, "step": 25890 }, { "epoch": 0.05, "learning_rate": 9.871735762354207e-07, "loss": 0.8705, "step": 25891 }, { "epoch": 0.05, "learning_rate": 9.867066564058791e-07, "loss": 0.8435, "step": 25892 }, { "epoch": 0.05, "learning_rate": 9.86239841295299e-07, "loss": 0.8209, "step": 25893 }, { "epoch": 0.05, "learning_rate": 9.857731309091057e-07, "loss": 0.8257, "step": 25894 }, { "epoch": 0.05, "learning_rate": 9.853065252527205e-07, "loss": 0.879, "step": 25895 }, { "epoch": 0.05, "learning_rate": 9.84840024331567e-07, "loss": 0.8553, "step": 25896 }, { "epoch": 0.05, "learning_rate": 9.843736281510595e-07, "loss": 0.8234, "step": 25897 }, { "epoch": 0.05, "learning_rate": 9.839073367166207e-07, "loss": 0.8385, "step": 25898 }, { "epoch": 0.05, "learning_rate": 9.834411500336672e-07, "loss": 0.9162, "step": 25899 }, { "epoch": 0.05, "learning_rate": 9.829750681076155e-07, "loss": 0.8567, "step": 25900 }, { "epoch": 0.05, "learning_rate": 9.825090909438817e-07, "loss": 0.8194, "step": 25901 }, { "epoch": 0.05, "learning_rate": 9.820432185478767e-07, "loss": 0.8791, "step": 25902 }, { "epoch": 0.05, "learning_rate": 9.815774509250186e-07, "loss": 0.8771, "step": 25903 }, { "epoch": 0.05, "learning_rate": 9.811117880807108e-07, "loss": 0.8286, "step": 25904 }, { "epoch": 0.05, "learning_rate": 9.80646230020369e-07, "loss": 0.8667, "step": 25905 }, { "epoch": 0.05, "learning_rate": 9.801807767494042e-07, "loss": 0.8283, "step": 25906 }, { "epoch": 0.05, "learning_rate": 9.797154282732191e-07, "loss": 0.8529, "step": 25907 }, { "epoch": 0.05, "learning_rate": 9.792501845972213e-07, "loss": 0.8634, "step": 25908 }, { "epoch": 0.05, "learning_rate": 9.787850457268177e-07, "loss": 0.8369, "step": 25909 }, { "epoch": 0.05, "learning_rate": 9.783200116674119e-07, "loss": 0.8264, "step": 25910 }, { "epoch": 0.05, "learning_rate": 9.77855082424407e-07, "loss": 0.888, "step": 25911 }, { "epoch": 0.05, "learning_rate": 9.773902580032035e-07, "loss": 0.8796, "step": 25912 }, { "epoch": 0.05, "learning_rate": 9.769255384092046e-07, "loss": 0.8449, "step": 25913 }, { "epoch": 0.05, "learning_rate": 9.764609236478063e-07, "loss": 0.8166, "step": 25914 }, { "epoch": 0.05, "learning_rate": 9.759964137244083e-07, "loss": 0.8835, "step": 25915 }, { "epoch": 0.05, "learning_rate": 9.755320086444064e-07, "loss": 0.8536, "step": 25916 }, { "epoch": 0.05, "learning_rate": 9.750677084131966e-07, "loss": 0.8858, "step": 25917 }, { "epoch": 0.05, "learning_rate": 9.746035130361741e-07, "loss": 0.8286, "step": 25918 }, { "epoch": 0.05, "learning_rate": 9.741394225187316e-07, "loss": 0.8778, "step": 25919 }, { "epoch": 0.05, "learning_rate": 9.736754368662626e-07, "loss": 0.8289, "step": 25920 }, { "epoch": 0.05, "learning_rate": 9.732115560841549e-07, "loss": 0.8454, "step": 25921 }, { "epoch": 0.05, "learning_rate": 9.727477801777985e-07, "loss": 0.8545, "step": 25922 }, { "epoch": 0.05, "learning_rate": 9.722841091525803e-07, "loss": 0.8697, "step": 25923 }, { "epoch": 0.05, "learning_rate": 9.71820543013894e-07, "loss": 0.8557, "step": 25924 }, { "epoch": 0.05, "learning_rate": 9.713570817671181e-07, "loss": 0.8284, "step": 25925 }, { "epoch": 0.05, "learning_rate": 9.70893725417641e-07, "loss": 0.8684, "step": 25926 }, { "epoch": 0.05, "learning_rate": 9.704304739708448e-07, "loss": 0.806, "step": 25927 }, { "epoch": 0.05, "learning_rate": 9.699673274321143e-07, "loss": 0.8593, "step": 25928 }, { "epoch": 0.05, "learning_rate": 9.695042858068238e-07, "loss": 0.8313, "step": 25929 }, { "epoch": 0.05, "learning_rate": 9.690413491003591e-07, "loss": 0.8274, "step": 25930 }, { "epoch": 0.05, "learning_rate": 9.685785173180994e-07, "loss": 0.878, "step": 25931 }, { "epoch": 0.05, "learning_rate": 9.681157904654181e-07, "loss": 0.8571, "step": 25932 }, { "epoch": 0.05, "learning_rate": 9.67653168547692e-07, "loss": 0.8606, "step": 25933 }, { "epoch": 0.05, "learning_rate": 9.671906515702979e-07, "loss": 0.8876, "step": 25934 }, { "epoch": 0.05, "learning_rate": 9.66728239538608e-07, "loss": 0.8766, "step": 25935 }, { "epoch": 0.05, "learning_rate": 9.662659324579948e-07, "loss": 0.8482, "step": 25936 }, { "epoch": 0.05, "learning_rate": 9.658037303338298e-07, "loss": 0.8217, "step": 25937 }, { "epoch": 0.05, "learning_rate": 9.65341633171486e-07, "loss": 0.8412, "step": 25938 }, { "epoch": 0.05, "learning_rate": 9.64879640976326e-07, "loss": 0.8432, "step": 25939 }, { "epoch": 0.05, "learning_rate": 9.644177537537214e-07, "loss": 0.8269, "step": 25940 }, { "epoch": 0.05, "learning_rate": 9.639559715090374e-07, "loss": 0.8452, "step": 25941 }, { "epoch": 0.05, "learning_rate": 9.634942942476388e-07, "loss": 0.8212, "step": 25942 }, { "epoch": 0.05, "learning_rate": 9.630327219748914e-07, "loss": 0.8238, "step": 25943 }, { "epoch": 0.05, "learning_rate": 9.625712546961552e-07, "loss": 0.8112, "step": 25944 }, { "epoch": 0.05, "learning_rate": 9.62109892416796e-07, "loss": 0.8407, "step": 25945 }, { "epoch": 0.05, "learning_rate": 9.616486351421683e-07, "loss": 0.8549, "step": 25946 }, { "epoch": 0.05, "learning_rate": 9.611874828776347e-07, "loss": 0.8228, "step": 25947 }, { "epoch": 0.05, "learning_rate": 9.607264356285495e-07, "loss": 0.8376, "step": 25948 }, { "epoch": 0.05, "learning_rate": 9.602654934002764e-07, "loss": 0.8673, "step": 25949 }, { "epoch": 0.05, "learning_rate": 9.598046561981645e-07, "loss": 0.8384, "step": 25950 }, { "epoch": 0.05, "learning_rate": 9.593439240275692e-07, "loss": 0.8689, "step": 25951 }, { "epoch": 0.05, "learning_rate": 9.588832968938466e-07, "loss": 0.8315, "step": 25952 }, { "epoch": 0.05, "learning_rate": 9.584227748023412e-07, "loss": 0.8404, "step": 25953 }, { "epoch": 0.05, "learning_rate": 9.579623577584108e-07, "loss": 0.8659, "step": 25954 }, { "epoch": 0.05, "learning_rate": 9.57502045767401e-07, "loss": 0.8722, "step": 25955 }, { "epoch": 0.05, "learning_rate": 9.570418388346625e-07, "loss": 0.8324, "step": 25956 }, { "epoch": 0.05, "learning_rate": 9.565817369655395e-07, "loss": 0.8327, "step": 25957 }, { "epoch": 0.05, "learning_rate": 9.561217401653778e-07, "loss": 0.8624, "step": 25958 }, { "epoch": 0.05, "learning_rate": 9.55661848439523e-07, "loss": 0.8804, "step": 25959 }, { "epoch": 0.05, "learning_rate": 9.552020617933166e-07, "loss": 0.8116, "step": 25960 }, { "epoch": 0.05, "learning_rate": 9.54742380232102e-07, "loss": 0.8032, "step": 25961 }, { "epoch": 0.05, "learning_rate": 9.542828037612195e-07, "loss": 0.8485, "step": 25962 }, { "epoch": 0.05, "learning_rate": 9.538233323860103e-07, "loss": 0.8694, "step": 25963 }, { "epoch": 0.05, "learning_rate": 9.533639661118099e-07, "loss": 0.8452, "step": 25964 }, { "epoch": 0.05, "learning_rate": 9.529047049439555e-07, "loss": 0.8787, "step": 25965 }, { "epoch": 0.05, "learning_rate": 9.524455488877837e-07, "loss": 0.8313, "step": 25966 }, { "epoch": 0.05, "learning_rate": 9.519864979486304e-07, "loss": 0.8555, "step": 25967 }, { "epoch": 0.05, "learning_rate": 9.515275521318268e-07, "loss": 0.8967, "step": 25968 }, { "epoch": 0.05, "learning_rate": 9.510687114427064e-07, "loss": 0.8541, "step": 25969 }, { "epoch": 0.05, "learning_rate": 9.506099758866017e-07, "loss": 0.8867, "step": 25970 }, { "epoch": 0.05, "learning_rate": 9.501513454688383e-07, "loss": 0.8444, "step": 25971 }, { "epoch": 0.05, "learning_rate": 9.496928201947453e-07, "loss": 0.827, "step": 25972 }, { "epoch": 0.05, "learning_rate": 9.492344000696529e-07, "loss": 0.8411, "step": 25973 }, { "epoch": 0.05, "learning_rate": 9.487760850988892e-07, "loss": 0.7988, "step": 25974 }, { "epoch": 0.05, "learning_rate": 9.48317875287772e-07, "loss": 0.8365, "step": 25975 }, { "epoch": 0.05, "learning_rate": 9.478597706416303e-07, "loss": 0.836, "step": 25976 }, { "epoch": 0.05, "learning_rate": 9.474017711657835e-07, "loss": 0.8377, "step": 25977 }, { "epoch": 0.05, "learning_rate": 9.469438768655548e-07, "loss": 0.8119, "step": 25978 }, { "epoch": 0.05, "learning_rate": 9.464860877462623e-07, "loss": 0.8462, "step": 25979 }, { "epoch": 0.05, "learning_rate": 9.460284038132273e-07, "loss": 0.8851, "step": 25980 }, { "epoch": 0.05, "learning_rate": 9.455708250717666e-07, "loss": 0.8647, "step": 25981 }, { "epoch": 0.05, "learning_rate": 9.451133515271937e-07, "loss": 0.8373, "step": 25982 }, { "epoch": 0.05, "learning_rate": 9.446559831848267e-07, "loss": 0.783, "step": 25983 }, { "epoch": 0.05, "learning_rate": 9.44198720049978e-07, "loss": 0.8328, "step": 25984 }, { "epoch": 0.05, "learning_rate": 9.4374156212796e-07, "loss": 0.8643, "step": 25985 }, { "epoch": 0.05, "learning_rate": 9.43284509424085e-07, "loss": 0.8206, "step": 25986 }, { "epoch": 0.05, "learning_rate": 9.428275619436633e-07, "loss": 0.9013, "step": 25987 }, { "epoch": 0.05, "learning_rate": 9.42370719692004e-07, "loss": 0.8223, "step": 25988 }, { "epoch": 0.05, "learning_rate": 9.419139826744139e-07, "loss": 0.8401, "step": 25989 }, { "epoch": 0.05, "learning_rate": 9.414573508961989e-07, "loss": 0.8247, "step": 25990 }, { "epoch": 0.05, "learning_rate": 9.410008243626645e-07, "loss": 0.8627, "step": 25991 }, { "epoch": 0.05, "learning_rate": 9.405444030791166e-07, "loss": 0.8645, "step": 25992 }, { "epoch": 0.05, "learning_rate": 9.400880870508566e-07, "loss": 0.8119, "step": 25993 }, { "epoch": 0.05, "learning_rate": 9.396318762831868e-07, "loss": 0.8789, "step": 25994 }, { "epoch": 0.05, "learning_rate": 9.391757707814075e-07, "loss": 0.8903, "step": 25995 }, { "epoch": 0.05, "learning_rate": 9.387197705508155e-07, "loss": 0.8607, "step": 25996 }, { "epoch": 0.05, "learning_rate": 9.382638755967122e-07, "loss": 0.8356, "step": 25997 }, { "epoch": 0.05, "learning_rate": 9.378080859243943e-07, "loss": 0.8773, "step": 25998 }, { "epoch": 0.05, "learning_rate": 9.373524015391533e-07, "loss": 0.8522, "step": 25999 }, { "epoch": 0.05, "learning_rate": 9.368968224462871e-07, "loss": 0.8632, "step": 26000 }, { "epoch": 0.05, "learning_rate": 9.36441348651087e-07, "loss": 0.8491, "step": 26001 }, { "epoch": 0.05, "learning_rate": 9.359859801588467e-07, "loss": 0.8474, "step": 26002 }, { "epoch": 0.05, "learning_rate": 9.355307169748539e-07, "loss": 0.8835, "step": 26003 }, { "epoch": 0.05, "learning_rate": 9.350755591044014e-07, "loss": 0.853, "step": 26004 }, { "epoch": 0.05, "learning_rate": 9.34620506552777e-07, "loss": 0.849, "step": 26005 }, { "epoch": 0.05, "learning_rate": 9.34165559325263e-07, "loss": 0.8707, "step": 26006 }, { "epoch": 0.05, "learning_rate": 9.337107174271498e-07, "loss": 0.8458, "step": 26007 }, { "epoch": 0.05, "learning_rate": 9.332559808637198e-07, "loss": 0.867, "step": 26008 }, { "epoch": 0.05, "learning_rate": 9.328013496402577e-07, "loss": 0.8443, "step": 26009 }, { "epoch": 0.05, "learning_rate": 9.323468237620436e-07, "loss": 0.8215, "step": 26010 }, { "epoch": 0.05, "learning_rate": 9.318924032343601e-07, "loss": 0.8961, "step": 26011 }, { "epoch": 0.05, "learning_rate": 9.314380880624874e-07, "loss": 0.8603, "step": 26012 }, { "epoch": 0.05, "learning_rate": 9.309838782517012e-07, "loss": 0.8407, "step": 26013 }, { "epoch": 0.05, "learning_rate": 9.305297738072805e-07, "loss": 0.8343, "step": 26014 }, { "epoch": 0.05, "learning_rate": 9.300757747344979e-07, "loss": 0.8319, "step": 26015 }, { "epoch": 0.05, "learning_rate": 9.296218810386348e-07, "loss": 0.8766, "step": 26016 }, { "epoch": 0.05, "learning_rate": 9.291680927249591e-07, "loss": 0.8601, "step": 26017 }, { "epoch": 0.05, "learning_rate": 9.287144097987455e-07, "loss": 0.8677, "step": 26018 }, { "epoch": 0.05, "learning_rate": 9.282608322652631e-07, "loss": 0.8574, "step": 26019 }, { "epoch": 0.05, "learning_rate": 9.278073601297854e-07, "loss": 0.8589, "step": 26020 }, { "epoch": 0.05, "learning_rate": 9.27353993397574e-07, "loss": 0.862, "step": 26021 }, { "epoch": 0.05, "learning_rate": 9.269007320739032e-07, "loss": 0.854, "step": 26022 }, { "epoch": 0.05, "learning_rate": 9.264475761640379e-07, "loss": 0.8373, "step": 26023 }, { "epoch": 0.05, "learning_rate": 9.259945256732394e-07, "loss": 0.8666, "step": 26024 }, { "epoch": 0.05, "learning_rate": 9.255415806067747e-07, "loss": 0.8575, "step": 26025 }, { "epoch": 0.05, "learning_rate": 9.250887409699038e-07, "loss": 0.8586, "step": 26026 }, { "epoch": 0.05, "learning_rate": 9.246360067678906e-07, "loss": 0.8741, "step": 26027 }, { "epoch": 0.05, "learning_rate": 9.241833780059928e-07, "loss": 0.7898, "step": 26028 }, { "epoch": 0.05, "learning_rate": 9.237308546894708e-07, "loss": 0.8475, "step": 26029 }, { "epoch": 0.05, "learning_rate": 9.232784368235826e-07, "loss": 0.8822, "step": 26030 }, { "epoch": 0.05, "learning_rate": 9.228261244135828e-07, "loss": 0.8821, "step": 26031 }, { "epoch": 0.05, "learning_rate": 9.223739174647262e-07, "loss": 0.8641, "step": 26032 }, { "epoch": 0.05, "learning_rate": 9.219218159822674e-07, "loss": 0.7972, "step": 26033 }, { "epoch": 0.05, "learning_rate": 9.214698199714589e-07, "loss": 0.8347, "step": 26034 }, { "epoch": 0.05, "learning_rate": 9.210179294375532e-07, "loss": 0.8526, "step": 26035 }, { "epoch": 0.05, "learning_rate": 9.205661443857994e-07, "loss": 0.8649, "step": 26036 }, { "epoch": 0.05, "learning_rate": 9.201144648214477e-07, "loss": 0.8369, "step": 26037 }, { "epoch": 0.05, "learning_rate": 9.19662890749744e-07, "loss": 0.875, "step": 26038 }, { "epoch": 0.05, "learning_rate": 9.192114221759341e-07, "loss": 0.8752, "step": 26039 }, { "epoch": 0.05, "learning_rate": 9.187600591052659e-07, "loss": 0.8733, "step": 26040 }, { "epoch": 0.05, "learning_rate": 9.183088015429853e-07, "loss": 0.8337, "step": 26041 }, { "epoch": 0.05, "learning_rate": 9.178576494943303e-07, "loss": 0.8252, "step": 26042 }, { "epoch": 0.05, "learning_rate": 9.174066029645434e-07, "loss": 0.8569, "step": 26043 }, { "epoch": 0.05, "learning_rate": 9.169556619588693e-07, "loss": 0.8098, "step": 26044 }, { "epoch": 0.05, "learning_rate": 9.165048264825394e-07, "loss": 0.8527, "step": 26045 }, { "epoch": 0.05, "learning_rate": 9.160540965407983e-07, "loss": 0.8632, "step": 26046 }, { "epoch": 0.05, "learning_rate": 9.156034721388806e-07, "loss": 0.8336, "step": 26047 }, { "epoch": 0.05, "learning_rate": 9.151529532820236e-07, "loss": 0.8474, "step": 26048 }, { "epoch": 0.05, "learning_rate": 9.147025399754583e-07, "loss": 0.8358, "step": 26049 }, { "epoch": 0.05, "learning_rate": 9.142522322244185e-07, "loss": 0.8508, "step": 26050 }, { "epoch": 0.05, "learning_rate": 9.138020300341366e-07, "loss": 0.8701, "step": 26051 }, { "epoch": 0.05, "learning_rate": 9.133519334098439e-07, "loss": 0.8391, "step": 26052 }, { "epoch": 0.05, "learning_rate": 9.129019423567686e-07, "loss": 0.8205, "step": 26053 }, { "epoch": 0.05, "learning_rate": 9.124520568801387e-07, "loss": 0.8668, "step": 26054 }, { "epoch": 0.05, "learning_rate": 9.120022769851843e-07, "loss": 0.8766, "step": 26055 }, { "epoch": 0.05, "learning_rate": 9.115526026771271e-07, "loss": 0.8714, "step": 26056 }, { "epoch": 0.05, "learning_rate": 9.111030339611915e-07, "loss": 0.8255, "step": 26057 }, { "epoch": 0.05, "learning_rate": 9.106535708426034e-07, "loss": 0.8479, "step": 26058 }, { "epoch": 0.05, "learning_rate": 9.102042133265831e-07, "loss": 0.846, "step": 26059 }, { "epoch": 0.05, "learning_rate": 9.097549614183521e-07, "loss": 0.8483, "step": 26060 }, { "epoch": 0.05, "learning_rate": 9.093058151231282e-07, "loss": 0.8119, "step": 26061 }, { "epoch": 0.05, "learning_rate": 9.08856774446134e-07, "loss": 0.8557, "step": 26062 }, { "epoch": 0.05, "learning_rate": 9.084078393925821e-07, "loss": 0.8555, "step": 26063 }, { "epoch": 0.05, "learning_rate": 9.079590099676882e-07, "loss": 0.8816, "step": 26064 }, { "epoch": 0.05, "learning_rate": 9.075102861766694e-07, "loss": 0.8167, "step": 26065 }, { "epoch": 0.05, "learning_rate": 9.070616680247412e-07, "loss": 0.8774, "step": 26066 }, { "epoch": 0.05, "learning_rate": 9.06613155517111e-07, "loss": 0.8655, "step": 26067 }, { "epoch": 0.05, "learning_rate": 9.06164748658992e-07, "loss": 0.823, "step": 26068 }, { "epoch": 0.05, "learning_rate": 9.057164474555924e-07, "loss": 0.8521, "step": 26069 }, { "epoch": 0.05, "learning_rate": 9.052682519121225e-07, "loss": 0.8269, "step": 26070 }, { "epoch": 0.05, "learning_rate": 9.048201620337893e-07, "loss": 0.8283, "step": 26071 }, { "epoch": 0.05, "learning_rate": 9.043721778257985e-07, "loss": 0.8492, "step": 26072 }, { "epoch": 0.05, "learning_rate": 9.039242992933561e-07, "loss": 0.8229, "step": 26073 }, { "epoch": 0.05, "learning_rate": 9.034765264416634e-07, "loss": 0.8698, "step": 26074 }, { "epoch": 0.05, "learning_rate": 9.030288592759228e-07, "loss": 0.842, "step": 26075 }, { "epoch": 0.05, "learning_rate": 9.025812978013371e-07, "loss": 0.8491, "step": 26076 }, { "epoch": 0.05, "learning_rate": 9.021338420231052e-07, "loss": 0.8404, "step": 26077 }, { "epoch": 0.05, "learning_rate": 9.016864919464263e-07, "loss": 0.8569, "step": 26078 }, { "epoch": 0.05, "learning_rate": 9.012392475764986e-07, "loss": 0.8623, "step": 26079 }, { "epoch": 0.05, "learning_rate": 9.007921089185178e-07, "loss": 0.8267, "step": 26080 }, { "epoch": 0.05, "learning_rate": 9.003450759776766e-07, "loss": 0.827, "step": 26081 }, { "epoch": 0.05, "learning_rate": 8.998981487591718e-07, "loss": 0.8986, "step": 26082 }, { "epoch": 0.05, "learning_rate": 8.994513272681937e-07, "loss": 0.8396, "step": 26083 }, { "epoch": 0.05, "learning_rate": 8.990046115099349e-07, "loss": 0.8016, "step": 26084 }, { "epoch": 0.05, "learning_rate": 8.985580014895856e-07, "loss": 0.8585, "step": 26085 }, { "epoch": 0.05, "learning_rate": 8.981114972123329e-07, "loss": 0.8268, "step": 26086 }, { "epoch": 0.05, "learning_rate": 8.97665098683369e-07, "loss": 0.8191, "step": 26087 }, { "epoch": 0.05, "learning_rate": 8.972188059078723e-07, "loss": 0.8463, "step": 26088 }, { "epoch": 0.05, "learning_rate": 8.967726188910353e-07, "loss": 0.8426, "step": 26089 }, { "epoch": 0.05, "learning_rate": 8.963265376380403e-07, "loss": 0.8562, "step": 26090 }, { "epoch": 0.05, "learning_rate": 8.958805621540679e-07, "loss": 0.8878, "step": 26091 }, { "epoch": 0.05, "learning_rate": 8.954346924443002e-07, "loss": 0.8445, "step": 26092 }, { "epoch": 0.05, "learning_rate": 8.949889285139179e-07, "loss": 0.8277, "step": 26093 }, { "epoch": 0.05, "learning_rate": 8.945432703681001e-07, "loss": 0.8867, "step": 26094 }, { "epoch": 0.05, "learning_rate": 8.940977180120247e-07, "loss": 0.8781, "step": 26095 }, { "epoch": 0.05, "learning_rate": 8.936522714508678e-07, "loss": 0.8661, "step": 26096 }, { "epoch": 0.05, "learning_rate": 8.932069306898072e-07, "loss": 0.8745, "step": 26097 }, { "epoch": 0.05, "learning_rate": 8.927616957340124e-07, "loss": 0.849, "step": 26098 }, { "epoch": 0.05, "learning_rate": 8.923165665886591e-07, "loss": 0.8511, "step": 26099 }, { "epoch": 0.05, "learning_rate": 8.918715432589175e-07, "loss": 0.8497, "step": 26100 }, { "epoch": 0.05, "learning_rate": 8.914266257499604e-07, "loss": 0.8217, "step": 26101 }, { "epoch": 0.05, "learning_rate": 8.909818140669546e-07, "loss": 0.8412, "step": 26102 }, { "epoch": 0.05, "learning_rate": 8.905371082150683e-07, "loss": 0.8416, "step": 26103 }, { "epoch": 0.05, "learning_rate": 8.900925081994716e-07, "loss": 0.8433, "step": 26104 }, { "epoch": 0.05, "learning_rate": 8.896480140253261e-07, "loss": 0.8404, "step": 26105 }, { "epoch": 0.05, "learning_rate": 8.892036256977965e-07, "loss": 0.8438, "step": 26106 }, { "epoch": 0.05, "learning_rate": 8.887593432220443e-07, "loss": 0.8466, "step": 26107 }, { "epoch": 0.05, "learning_rate": 8.883151666032375e-07, "loss": 0.8207, "step": 26108 }, { "epoch": 0.05, "learning_rate": 8.878710958465308e-07, "loss": 0.8408, "step": 26109 }, { "epoch": 0.05, "learning_rate": 8.874271309570858e-07, "loss": 0.851, "step": 26110 }, { "epoch": 0.05, "learning_rate": 8.869832719400606e-07, "loss": 0.8857, "step": 26111 }, { "epoch": 0.05, "learning_rate": 8.86539518800611e-07, "loss": 0.8473, "step": 26112 }, { "epoch": 0.05, "learning_rate": 8.860958715438939e-07, "loss": 0.8667, "step": 26113 }, { "epoch": 0.05, "learning_rate": 8.85652330175063e-07, "loss": 0.8527, "step": 26114 }, { "epoch": 0.05, "learning_rate": 8.852088946992743e-07, "loss": 0.8144, "step": 26115 }, { "epoch": 0.05, "learning_rate": 8.847655651216747e-07, "loss": 0.8314, "step": 26116 }, { "epoch": 0.05, "learning_rate": 8.843223414474167e-07, "loss": 0.8124, "step": 26117 }, { "epoch": 0.05, "learning_rate": 8.838792236816518e-07, "loss": 0.8748, "step": 26118 }, { "epoch": 0.05, "learning_rate": 8.834362118295259e-07, "loss": 0.8597, "step": 26119 }, { "epoch": 0.05, "learning_rate": 8.829933058961882e-07, "loss": 0.8472, "step": 26120 }, { "epoch": 0.05, "learning_rate": 8.825505058867822e-07, "loss": 0.8292, "step": 26121 }, { "epoch": 0.05, "learning_rate": 8.821078118064563e-07, "loss": 0.8508, "step": 26122 }, { "epoch": 0.05, "learning_rate": 8.816652236603485e-07, "loss": 0.8574, "step": 26123 }, { "epoch": 0.05, "learning_rate": 8.812227414536045e-07, "loss": 0.8483, "step": 26124 }, { "epoch": 0.05, "learning_rate": 8.807803651913638e-07, "loss": 0.859, "step": 26125 }, { "epoch": 0.05, "learning_rate": 8.803380948787676e-07, "loss": 0.8802, "step": 26126 }, { "epoch": 0.05, "learning_rate": 8.79895930520952e-07, "loss": 0.8431, "step": 26127 }, { "epoch": 0.05, "learning_rate": 8.794538721230561e-07, "loss": 0.833, "step": 26128 }, { "epoch": 0.05, "learning_rate": 8.79011919690217e-07, "loss": 0.8567, "step": 26129 }, { "epoch": 0.05, "learning_rate": 8.785700732275659e-07, "loss": 0.8657, "step": 26130 }, { "epoch": 0.05, "learning_rate": 8.781283327402357e-07, "loss": 0.8423, "step": 26131 }, { "epoch": 0.05, "learning_rate": 8.776866982333632e-07, "loss": 0.8286, "step": 26132 }, { "epoch": 0.05, "learning_rate": 8.772451697120788e-07, "loss": 0.8596, "step": 26133 }, { "epoch": 0.05, "learning_rate": 8.768037471815094e-07, "loss": 0.8835, "step": 26134 }, { "epoch": 0.05, "learning_rate": 8.763624306467844e-07, "loss": 0.8389, "step": 26135 }, { "epoch": 0.05, "learning_rate": 8.759212201130341e-07, "loss": 0.8246, "step": 26136 }, { "epoch": 0.05, "learning_rate": 8.754801155853787e-07, "loss": 0.8371, "step": 26137 }, { "epoch": 0.05, "learning_rate": 8.750391170689476e-07, "loss": 0.8188, "step": 26138 }, { "epoch": 0.05, "learning_rate": 8.745982245688634e-07, "loss": 0.8641, "step": 26139 }, { "epoch": 0.05, "learning_rate": 8.741574380902506e-07, "loss": 0.8608, "step": 26140 }, { "epoch": 0.05, "learning_rate": 8.737167576382266e-07, "loss": 0.8241, "step": 26141 }, { "epoch": 0.05, "learning_rate": 8.732761832179126e-07, "loss": 0.8577, "step": 26142 }, { "epoch": 0.05, "learning_rate": 8.728357148344291e-07, "loss": 0.8436, "step": 26143 }, { "epoch": 0.05, "learning_rate": 8.723953524928908e-07, "loss": 0.8223, "step": 26144 }, { "epoch": 0.05, "learning_rate": 8.719550961984158e-07, "loss": 0.8505, "step": 26145 }, { "epoch": 0.05, "learning_rate": 8.71514945956119e-07, "loss": 0.8402, "step": 26146 }, { "epoch": 0.05, "learning_rate": 8.710749017711151e-07, "loss": 0.828, "step": 26147 }, { "epoch": 0.05, "learning_rate": 8.706349636485134e-07, "loss": 0.8385, "step": 26148 }, { "epoch": 0.05, "learning_rate": 8.701951315934287e-07, "loss": 0.8426, "step": 26149 }, { "epoch": 0.05, "learning_rate": 8.697554056109681e-07, "loss": 0.8421, "step": 26150 }, { "epoch": 0.05, "learning_rate": 8.693157857062429e-07, "loss": 0.8594, "step": 26151 }, { "epoch": 0.05, "learning_rate": 8.688762718843591e-07, "loss": 0.8665, "step": 26152 }, { "epoch": 0.06, "learning_rate": 8.684368641504248e-07, "loss": 0.8453, "step": 26153 }, { "epoch": 0.06, "learning_rate": 8.67997562509546e-07, "loss": 0.8657, "step": 26154 }, { "epoch": 0.06, "learning_rate": 8.67558366966822e-07, "loss": 0.8716, "step": 26155 }, { "epoch": 0.06, "learning_rate": 8.671192775273574e-07, "loss": 0.863, "step": 26156 }, { "epoch": 0.06, "learning_rate": 8.66680294196256e-07, "loss": 0.8627, "step": 26157 }, { "epoch": 0.06, "learning_rate": 8.662414169786193e-07, "loss": 0.8601, "step": 26158 }, { "epoch": 0.06, "learning_rate": 8.65802645879541e-07, "loss": 0.8621, "step": 26159 }, { "epoch": 0.06, "learning_rate": 8.653639809041215e-07, "loss": 0.7987, "step": 26160 }, { "epoch": 0.06, "learning_rate": 8.649254220574577e-07, "loss": 0.8621, "step": 26161 }, { "epoch": 0.06, "learning_rate": 8.644869693446434e-07, "loss": 0.8708, "step": 26162 }, { "epoch": 0.06, "learning_rate": 8.640486227707745e-07, "loss": 0.855, "step": 26163 }, { "epoch": 0.06, "learning_rate": 8.636103823409436e-07, "loss": 0.8457, "step": 26164 }, { "epoch": 0.06, "learning_rate": 8.631722480602434e-07, "loss": 0.8447, "step": 26165 }, { "epoch": 0.06, "learning_rate": 8.627342199337607e-07, "loss": 0.8664, "step": 26166 }, { "epoch": 0.06, "learning_rate": 8.622962979665872e-07, "loss": 0.8856, "step": 26167 }, { "epoch": 0.06, "learning_rate": 8.618584821638098e-07, "loss": 0.8531, "step": 26168 }, { "epoch": 0.06, "learning_rate": 8.614207725305157e-07, "loss": 0.8117, "step": 26169 }, { "epoch": 0.06, "learning_rate": 8.609831690717896e-07, "loss": 0.853, "step": 26170 }, { "epoch": 0.06, "learning_rate": 8.605456717927164e-07, "loss": 0.8653, "step": 26171 }, { "epoch": 0.06, "learning_rate": 8.601082806983818e-07, "loss": 0.8438, "step": 26172 }, { "epoch": 0.06, "learning_rate": 8.596709957938621e-07, "loss": 0.8459, "step": 26173 }, { "epoch": 0.06, "learning_rate": 8.592338170842385e-07, "loss": 0.8293, "step": 26174 }, { "epoch": 0.06, "learning_rate": 8.587967445745959e-07, "loss": 0.8897, "step": 26175 }, { "epoch": 0.06, "learning_rate": 8.58359778270007e-07, "loss": 0.8212, "step": 26176 }, { "epoch": 0.06, "learning_rate": 8.579229181755511e-07, "loss": 0.839, "step": 26177 }, { "epoch": 0.06, "learning_rate": 8.574861642963017e-07, "loss": 0.8375, "step": 26178 }, { "epoch": 0.06, "learning_rate": 8.57049516637336e-07, "loss": 0.8339, "step": 26179 }, { "epoch": 0.06, "learning_rate": 8.566129752037223e-07, "loss": 0.8441, "step": 26180 }, { "epoch": 0.06, "learning_rate": 8.561765400005373e-07, "loss": 0.8519, "step": 26181 }, { "epoch": 0.06, "learning_rate": 8.557402110328516e-07, "loss": 0.8729, "step": 26182 }, { "epoch": 0.06, "learning_rate": 8.553039883057312e-07, "loss": 0.8627, "step": 26183 }, { "epoch": 0.06, "learning_rate": 8.548678718242453e-07, "loss": 0.8193, "step": 26184 }, { "epoch": 0.06, "learning_rate": 8.544318615934622e-07, "loss": 0.8327, "step": 26185 }, { "epoch": 0.06, "learning_rate": 8.539959576184464e-07, "loss": 0.8441, "step": 26186 }, { "epoch": 0.06, "learning_rate": 8.535601599042631e-07, "loss": 0.8445, "step": 26187 }, { "epoch": 0.06, "learning_rate": 8.531244684559759e-07, "loss": 0.8509, "step": 26188 }, { "epoch": 0.06, "learning_rate": 8.526888832786462e-07, "loss": 0.815, "step": 26189 }, { "epoch": 0.06, "learning_rate": 8.522534043773345e-07, "loss": 0.8504, "step": 26190 }, { "epoch": 0.06, "learning_rate": 8.51818031757099e-07, "loss": 0.8408, "step": 26191 }, { "epoch": 0.06, "learning_rate": 8.513827654230011e-07, "loss": 0.8808, "step": 26192 }, { "epoch": 0.06, "learning_rate": 8.509476053800947e-07, "loss": 0.8267, "step": 26193 }, { "epoch": 0.06, "learning_rate": 8.505125516334379e-07, "loss": 0.8365, "step": 26194 }, { "epoch": 0.06, "learning_rate": 8.500776041880843e-07, "loss": 0.8755, "step": 26195 }, { "epoch": 0.06, "learning_rate": 8.496427630490888e-07, "loss": 0.8138, "step": 26196 }, { "epoch": 0.06, "learning_rate": 8.49208028221501e-07, "loss": 0.8181, "step": 26197 }, { "epoch": 0.06, "learning_rate": 8.487733997103731e-07, "loss": 0.8233, "step": 26198 }, { "epoch": 0.06, "learning_rate": 8.483388775207523e-07, "loss": 0.8869, "step": 26199 }, { "epoch": 0.06, "learning_rate": 8.479044616576925e-07, "loss": 0.8336, "step": 26200 }, { "epoch": 0.06, "learning_rate": 8.474701521262363e-07, "loss": 0.839, "step": 26201 }, { "epoch": 0.06, "learning_rate": 8.470359489314306e-07, "loss": 0.8322, "step": 26202 }, { "epoch": 0.06, "learning_rate": 8.466018520783204e-07, "loss": 0.8387, "step": 26203 }, { "epoch": 0.06, "learning_rate": 8.461678615719493e-07, "loss": 0.8435, "step": 26204 }, { "epoch": 0.06, "learning_rate": 8.4573397741736e-07, "loss": 0.8688, "step": 26205 }, { "epoch": 0.06, "learning_rate": 8.45300199619592e-07, "loss": 0.8419, "step": 26206 }, { "epoch": 0.06, "learning_rate": 8.448665281836887e-07, "loss": 0.8819, "step": 26207 }, { "epoch": 0.06, "learning_rate": 8.44432963114683e-07, "loss": 0.8209, "step": 26208 }, { "epoch": 0.06, "learning_rate": 8.439995044176153e-07, "loss": 0.8479, "step": 26209 }, { "epoch": 0.06, "learning_rate": 8.435661520975213e-07, "loss": 0.8966, "step": 26210 }, { "epoch": 0.06, "learning_rate": 8.431329061594373e-07, "loss": 0.9043, "step": 26211 }, { "epoch": 0.06, "learning_rate": 8.426997666083936e-07, "loss": 0.8474, "step": 26212 }, { "epoch": 0.06, "learning_rate": 8.42266733449425e-07, "loss": 0.8402, "step": 26213 }, { "epoch": 0.06, "learning_rate": 8.418338066875642e-07, "loss": 0.8495, "step": 26214 }, { "epoch": 0.06, "learning_rate": 8.414009863278372e-07, "loss": 0.8383, "step": 26215 }, { "epoch": 0.06, "learning_rate": 8.409682723752743e-07, "loss": 0.8117, "step": 26216 }, { "epoch": 0.06, "learning_rate": 8.405356648349006e-07, "loss": 0.845, "step": 26217 }, { "epoch": 0.06, "learning_rate": 8.401031637117473e-07, "loss": 0.8491, "step": 26218 }, { "epoch": 0.06, "learning_rate": 8.396707690108363e-07, "loss": 0.8504, "step": 26219 }, { "epoch": 0.06, "learning_rate": 8.392384807371912e-07, "loss": 0.8309, "step": 26220 }, { "epoch": 0.06, "learning_rate": 8.388062988958357e-07, "loss": 0.8497, "step": 26221 }, { "epoch": 0.06, "learning_rate": 8.383742234917891e-07, "loss": 0.8141, "step": 26222 }, { "epoch": 0.06, "learning_rate": 8.379422545300708e-07, "loss": 0.8564, "step": 26223 }, { "epoch": 0.06, "learning_rate": 8.375103920157024e-07, "loss": 0.8568, "step": 26224 }, { "epoch": 0.06, "learning_rate": 8.37078635953702e-07, "loss": 0.8428, "step": 26225 }, { "epoch": 0.06, "learning_rate": 8.366469863490822e-07, "loss": 0.8462, "step": 26226 }, { "epoch": 0.06, "learning_rate": 8.362154432068603e-07, "loss": 0.8368, "step": 26227 }, { "epoch": 0.06, "learning_rate": 8.35784006532051e-07, "loss": 0.8389, "step": 26228 }, { "epoch": 0.06, "learning_rate": 8.353526763296616e-07, "loss": 0.82, "step": 26229 }, { "epoch": 0.06, "learning_rate": 8.349214526047089e-07, "loss": 0.864, "step": 26230 }, { "epoch": 0.06, "learning_rate": 8.344903353622013e-07, "loss": 0.8301, "step": 26231 }, { "epoch": 0.06, "learning_rate": 8.340593246071505e-07, "loss": 0.8449, "step": 26232 }, { "epoch": 0.06, "learning_rate": 8.336284203445588e-07, "loss": 0.8525, "step": 26233 }, { "epoch": 0.06, "learning_rate": 8.331976225794347e-07, "loss": 0.8564, "step": 26234 }, { "epoch": 0.06, "learning_rate": 8.327669313167841e-07, "loss": 0.8447, "step": 26235 }, { "epoch": 0.06, "learning_rate": 8.323363465616108e-07, "loss": 0.8322, "step": 26236 }, { "epoch": 0.06, "learning_rate": 8.319058683189163e-07, "loss": 0.8134, "step": 26237 }, { "epoch": 0.06, "learning_rate": 8.314754965937033e-07, "loss": 0.8949, "step": 26238 }, { "epoch": 0.06, "learning_rate": 8.310452313909734e-07, "loss": 0.909, "step": 26239 }, { "epoch": 0.06, "learning_rate": 8.306150727157214e-07, "loss": 0.87, "step": 26240 }, { "epoch": 0.06, "learning_rate": 8.301850205729478e-07, "loss": 0.8206, "step": 26241 }, { "epoch": 0.06, "learning_rate": 8.297550749676475e-07, "loss": 0.8433, "step": 26242 }, { "epoch": 0.06, "learning_rate": 8.293252359048176e-07, "loss": 0.829, "step": 26243 }, { "epoch": 0.06, "learning_rate": 8.288955033894519e-07, "loss": 0.8471, "step": 26244 }, { "epoch": 0.06, "learning_rate": 8.284658774265408e-07, "loss": 0.8623, "step": 26245 }, { "epoch": 0.06, "learning_rate": 8.280363580210804e-07, "loss": 0.8468, "step": 26246 }, { "epoch": 0.06, "learning_rate": 8.276069451780544e-07, "loss": 0.8262, "step": 26247 }, { "epoch": 0.06, "learning_rate": 8.271776389024577e-07, "loss": 0.8293, "step": 26248 }, { "epoch": 0.06, "learning_rate": 8.267484391992753e-07, "loss": 0.84, "step": 26249 }, { "epoch": 0.06, "learning_rate": 8.263193460734964e-07, "loss": 0.8791, "step": 26250 }, { "epoch": 0.06, "learning_rate": 8.258903595301026e-07, "loss": 0.8956, "step": 26251 }, { "epoch": 0.06, "learning_rate": 8.254614795740789e-07, "loss": 0.837, "step": 26252 }, { "epoch": 0.06, "learning_rate": 8.250327062104102e-07, "loss": 0.8606, "step": 26253 }, { "epoch": 0.06, "learning_rate": 8.246040394440757e-07, "loss": 0.8375, "step": 26254 }, { "epoch": 0.06, "learning_rate": 8.241754792800583e-07, "loss": 0.8695, "step": 26255 }, { "epoch": 0.06, "learning_rate": 8.237470257233349e-07, "loss": 0.8636, "step": 26256 }, { "epoch": 0.06, "learning_rate": 8.233186787788861e-07, "loss": 0.8774, "step": 26257 }, { "epoch": 0.06, "learning_rate": 8.228904384516845e-07, "loss": 0.8448, "step": 26258 }, { "epoch": 0.06, "learning_rate": 8.224623047467084e-07, "loss": 0.8613, "step": 26259 }, { "epoch": 0.06, "learning_rate": 8.220342776689305e-07, "loss": 0.8034, "step": 26260 }, { "epoch": 0.06, "learning_rate": 8.216063572233246e-07, "loss": 0.8451, "step": 26261 }, { "epoch": 0.06, "learning_rate": 8.211785434148634e-07, "loss": 0.8521, "step": 26262 }, { "epoch": 0.06, "learning_rate": 8.20750836248515e-07, "loss": 0.8411, "step": 26263 }, { "epoch": 0.06, "learning_rate": 8.203232357292523e-07, "loss": 0.8375, "step": 26264 }, { "epoch": 0.06, "learning_rate": 8.198957418620401e-07, "loss": 0.8236, "step": 26265 }, { "epoch": 0.06, "learning_rate": 8.194683546518433e-07, "loss": 0.8159, "step": 26266 }, { "epoch": 0.06, "learning_rate": 8.190410741036325e-07, "loss": 0.8694, "step": 26267 }, { "epoch": 0.06, "learning_rate": 8.186139002223692e-07, "loss": 0.856, "step": 26268 }, { "epoch": 0.06, "learning_rate": 8.181868330130161e-07, "loss": 0.8431, "step": 26269 }, { "epoch": 0.06, "learning_rate": 8.177598724805358e-07, "loss": 0.8195, "step": 26270 }, { "epoch": 0.06, "learning_rate": 8.173330186298911e-07, "loss": 0.8453, "step": 26271 }, { "epoch": 0.06, "learning_rate": 8.169062714660347e-07, "loss": 0.8467, "step": 26272 }, { "epoch": 0.06, "learning_rate": 8.164796309939304e-07, "loss": 0.8323, "step": 26273 }, { "epoch": 0.06, "learning_rate": 8.160530972185355e-07, "loss": 0.844, "step": 26274 }, { "epoch": 0.06, "learning_rate": 8.156266701448013e-07, "loss": 0.8399, "step": 26275 }, { "epoch": 0.06, "learning_rate": 8.15200349777685e-07, "loss": 0.812, "step": 26276 }, { "epoch": 0.06, "learning_rate": 8.147741361221384e-07, "loss": 0.8138, "step": 26277 }, { "epoch": 0.06, "learning_rate": 8.143480291831152e-07, "loss": 0.8232, "step": 26278 }, { "epoch": 0.06, "learning_rate": 8.139220289655647e-07, "loss": 0.8653, "step": 26279 }, { "epoch": 0.06, "learning_rate": 8.134961354744353e-07, "loss": 0.8116, "step": 26280 }, { "epoch": 0.06, "learning_rate": 8.130703487146796e-07, "loss": 0.8385, "step": 26281 }, { "epoch": 0.06, "learning_rate": 8.126446686912382e-07, "loss": 0.8042, "step": 26282 }, { "epoch": 0.06, "learning_rate": 8.122190954090614e-07, "loss": 0.8342, "step": 26283 }, { "epoch": 0.06, "learning_rate": 8.11793628873091e-07, "loss": 0.8192, "step": 26284 }, { "epoch": 0.06, "learning_rate": 8.113682690882707e-07, "loss": 0.83, "step": 26285 }, { "epoch": 0.06, "learning_rate": 8.109430160595444e-07, "loss": 0.7985, "step": 26286 }, { "epoch": 0.06, "learning_rate": 8.105178697918504e-07, "loss": 0.8962, "step": 26287 }, { "epoch": 0.06, "learning_rate": 8.100928302901311e-07, "loss": 0.8497, "step": 26288 }, { "epoch": 0.06, "learning_rate": 8.09667897559322e-07, "loss": 0.8576, "step": 26289 }, { "epoch": 0.06, "learning_rate": 8.092430716043576e-07, "loss": 0.8174, "step": 26290 }, { "epoch": 0.06, "learning_rate": 8.088183524301796e-07, "loss": 0.9045, "step": 26291 }, { "epoch": 0.06, "learning_rate": 8.083937400417219e-07, "loss": 0.8573, "step": 26292 }, { "epoch": 0.06, "learning_rate": 8.079692344439138e-07, "loss": 0.801, "step": 26293 }, { "epoch": 0.06, "learning_rate": 8.075448356416881e-07, "loss": 0.8504, "step": 26294 }, { "epoch": 0.06, "learning_rate": 8.071205436399776e-07, "loss": 0.8904, "step": 26295 }, { "epoch": 0.06, "learning_rate": 8.066963584437115e-07, "loss": 0.8245, "step": 26296 }, { "epoch": 0.06, "learning_rate": 8.062722800578171e-07, "loss": 0.8378, "step": 26297 }, { "epoch": 0.06, "learning_rate": 8.058483084872215e-07, "loss": 0.8489, "step": 26298 }, { "epoch": 0.06, "learning_rate": 8.05424443736853e-07, "loss": 0.8237, "step": 26299 }, { "epoch": 0.06, "learning_rate": 8.050006858116321e-07, "loss": 0.8492, "step": 26300 }, { "epoch": 0.06, "learning_rate": 8.04577034716485e-07, "loss": 0.8458, "step": 26301 }, { "epoch": 0.06, "learning_rate": 8.041534904563319e-07, "loss": 0.8587, "step": 26302 }, { "epoch": 0.06, "learning_rate": 8.037300530360959e-07, "loss": 0.8531, "step": 26303 }, { "epoch": 0.06, "learning_rate": 8.03306722460695e-07, "loss": 0.8656, "step": 26304 }, { "epoch": 0.06, "learning_rate": 8.028834987350476e-07, "loss": 0.819, "step": 26305 }, { "epoch": 0.06, "learning_rate": 8.02460381864073e-07, "loss": 0.8515, "step": 26306 }, { "epoch": 0.06, "learning_rate": 8.020373718526842e-07, "loss": 0.8648, "step": 26307 }, { "epoch": 0.06, "learning_rate": 8.016144687057959e-07, "loss": 0.8178, "step": 26308 }, { "epoch": 0.06, "learning_rate": 8.011916724283219e-07, "loss": 0.8193, "step": 26309 }, { "epoch": 0.06, "learning_rate": 8.007689830251775e-07, "loss": 0.8422, "step": 26310 }, { "epoch": 0.06, "learning_rate": 8.003464005012707e-07, "loss": 0.8548, "step": 26311 }, { "epoch": 0.06, "learning_rate": 7.99923924861511e-07, "loss": 0.8468, "step": 26312 }, { "epoch": 0.06, "learning_rate": 7.995015561108088e-07, "loss": 0.8097, "step": 26313 }, { "epoch": 0.06, "learning_rate": 7.990792942540693e-07, "loss": 0.8385, "step": 26314 }, { "epoch": 0.06, "learning_rate": 7.986571392961961e-07, "loss": 0.8721, "step": 26315 }, { "epoch": 0.06, "learning_rate": 7.982350912421e-07, "loss": 0.8527, "step": 26316 }, { "epoch": 0.06, "learning_rate": 7.978131500966823e-07, "loss": 0.8173, "step": 26317 }, { "epoch": 0.06, "learning_rate": 7.973913158648428e-07, "loss": 0.8413, "step": 26318 }, { "epoch": 0.06, "learning_rate": 7.969695885514838e-07, "loss": 0.8602, "step": 26319 }, { "epoch": 0.06, "learning_rate": 7.965479681615063e-07, "loss": 0.8314, "step": 26320 }, { "epoch": 0.06, "learning_rate": 7.961264546998071e-07, "loss": 0.8473, "step": 26321 }, { "epoch": 0.06, "learning_rate": 7.957050481712836e-07, "loss": 0.7971, "step": 26322 }, { "epoch": 0.06, "learning_rate": 7.952837485808329e-07, "loss": 0.8147, "step": 26323 }, { "epoch": 0.06, "learning_rate": 7.948625559333511e-07, "loss": 0.8266, "step": 26324 }, { "epoch": 0.06, "learning_rate": 7.944414702337288e-07, "loss": 0.8476, "step": 26325 }, { "epoch": 0.06, "learning_rate": 7.940204914868588e-07, "loss": 0.8685, "step": 26326 }, { "epoch": 0.06, "learning_rate": 7.935996196976325e-07, "loss": 0.817, "step": 26327 }, { "epoch": 0.06, "learning_rate": 7.931788548709396e-07, "loss": 0.8201, "step": 26328 }, { "epoch": 0.06, "learning_rate": 7.927581970116693e-07, "loss": 0.8074, "step": 26329 }, { "epoch": 0.06, "learning_rate": 7.92337646124709e-07, "loss": 0.8695, "step": 26330 }, { "epoch": 0.06, "learning_rate": 7.919172022149458e-07, "loss": 0.8859, "step": 26331 }, { "epoch": 0.06, "learning_rate": 7.914968652872624e-07, "loss": 0.809, "step": 26332 }, { "epoch": 0.06, "learning_rate": 7.910766353465416e-07, "loss": 0.8413, "step": 26333 }, { "epoch": 0.06, "learning_rate": 7.906565123976684e-07, "loss": 0.8534, "step": 26334 }, { "epoch": 0.06, "learning_rate": 7.902364964455223e-07, "loss": 0.8784, "step": 26335 }, { "epoch": 0.06, "learning_rate": 7.898165874949837e-07, "loss": 0.8454, "step": 26336 }, { "epoch": 0.06, "learning_rate": 7.89396785550931e-07, "loss": 0.8587, "step": 26337 }, { "epoch": 0.06, "learning_rate": 7.889770906182437e-07, "loss": 0.8107, "step": 26338 }, { "epoch": 0.06, "learning_rate": 7.885575027017922e-07, "loss": 0.8003, "step": 26339 }, { "epoch": 0.06, "learning_rate": 7.881380218064572e-07, "loss": 0.8602, "step": 26340 }, { "epoch": 0.06, "learning_rate": 7.877186479371102e-07, "loss": 0.8725, "step": 26341 }, { "epoch": 0.06, "learning_rate": 7.872993810986252e-07, "loss": 0.8664, "step": 26342 }, { "epoch": 0.06, "learning_rate": 7.868802212958704e-07, "loss": 0.8545, "step": 26343 }, { "epoch": 0.06, "learning_rate": 7.864611685337175e-07, "loss": 0.836, "step": 26344 }, { "epoch": 0.06, "learning_rate": 7.86042222817035e-07, "loss": 0.8871, "step": 26345 }, { "epoch": 0.06, "learning_rate": 7.8562338415069e-07, "loss": 0.8185, "step": 26346 }, { "epoch": 0.06, "learning_rate": 7.852046525395485e-07, "loss": 0.8597, "step": 26347 }, { "epoch": 0.06, "learning_rate": 7.847860279884756e-07, "loss": 0.8377, "step": 26348 }, { "epoch": 0.06, "learning_rate": 7.843675105023374e-07, "loss": 0.812, "step": 26349 }, { "epoch": 0.06, "learning_rate": 7.839491000859922e-07, "loss": 0.859, "step": 26350 }, { "epoch": 0.06, "learning_rate": 7.835307967443029e-07, "loss": 0.8798, "step": 26351 }, { "epoch": 0.06, "learning_rate": 7.83112600482131e-07, "loss": 0.8382, "step": 26352 }, { "epoch": 0.06, "learning_rate": 7.826945113043327e-07, "loss": 0.8524, "step": 26353 }, { "epoch": 0.06, "learning_rate": 7.822765292157675e-07, "loss": 0.8778, "step": 26354 }, { "epoch": 0.06, "learning_rate": 7.818586542212902e-07, "loss": 0.8416, "step": 26355 }, { "epoch": 0.06, "learning_rate": 7.814408863257583e-07, "loss": 0.8538, "step": 26356 }, { "epoch": 0.06, "learning_rate": 7.810232255340222e-07, "loss": 0.8432, "step": 26357 }, { "epoch": 0.06, "learning_rate": 7.806056718509347e-07, "loss": 0.8178, "step": 26358 }, { "epoch": 0.06, "learning_rate": 7.801882252813508e-07, "loss": 0.8784, "step": 26359 }, { "epoch": 0.06, "learning_rate": 7.797708858301167e-07, "loss": 0.8223, "step": 26360 }, { "epoch": 0.06, "learning_rate": 7.793536535020818e-07, "loss": 0.8405, "step": 26361 }, { "epoch": 0.06, "learning_rate": 7.789365283020956e-07, "loss": 0.8432, "step": 26362 }, { "epoch": 0.06, "learning_rate": 7.785195102350041e-07, "loss": 0.8174, "step": 26363 }, { "epoch": 0.06, "learning_rate": 7.781025993056468e-07, "loss": 0.8516, "step": 26364 }, { "epoch": 0.06, "learning_rate": 7.776857955188744e-07, "loss": 0.8528, "step": 26365 }, { "epoch": 0.06, "learning_rate": 7.772690988795295e-07, "loss": 0.8492, "step": 26366 }, { "epoch": 0.06, "learning_rate": 7.768525093924484e-07, "loss": 0.8651, "step": 26367 }, { "epoch": 0.06, "learning_rate": 7.764360270624738e-07, "loss": 0.8469, "step": 26368 }, { "epoch": 0.06, "learning_rate": 7.76019651894444e-07, "loss": 0.8394, "step": 26369 }, { "epoch": 0.06, "learning_rate": 7.756033838931975e-07, "loss": 0.8295, "step": 26370 }, { "epoch": 0.06, "learning_rate": 7.751872230635704e-07, "loss": 0.844, "step": 26371 }, { "epoch": 0.06, "learning_rate": 7.747711694103966e-07, "loss": 0.8193, "step": 26372 }, { "epoch": 0.06, "learning_rate": 7.74355222938512e-07, "loss": 0.8342, "step": 26373 }, { "epoch": 0.06, "learning_rate": 7.739393836527476e-07, "loss": 0.8388, "step": 26374 }, { "epoch": 0.06, "learning_rate": 7.735236515579336e-07, "loss": 0.8342, "step": 26375 }, { "epoch": 0.06, "learning_rate": 7.731080266589031e-07, "loss": 0.846, "step": 26376 }, { "epoch": 0.06, "learning_rate": 7.72692508960482e-07, "loss": 0.8224, "step": 26377 }, { "epoch": 0.06, "learning_rate": 7.722770984674999e-07, "loss": 0.8307, "step": 26378 }, { "epoch": 0.06, "learning_rate": 7.718617951847829e-07, "loss": 0.8921, "step": 26379 }, { "epoch": 0.06, "learning_rate": 7.714465991171571e-07, "loss": 0.8927, "step": 26380 }, { "epoch": 0.06, "learning_rate": 7.710315102694433e-07, "loss": 0.8763, "step": 26381 }, { "epoch": 0.06, "learning_rate": 7.706165286464639e-07, "loss": 0.8704, "step": 26382 }, { "epoch": 0.06, "learning_rate": 7.702016542530443e-07, "loss": 0.8424, "step": 26383 }, { "epoch": 0.06, "learning_rate": 7.697868870940039e-07, "loss": 0.8293, "step": 26384 }, { "epoch": 0.06, "learning_rate": 7.693722271741588e-07, "loss": 0.7858, "step": 26385 }, { "epoch": 0.06, "learning_rate": 7.689576744983285e-07, "loss": 0.8544, "step": 26386 }, { "epoch": 0.06, "learning_rate": 7.685432290713279e-07, "loss": 0.819, "step": 26387 }, { "epoch": 0.06, "learning_rate": 7.681288908979744e-07, "loss": 0.8452, "step": 26388 }, { "epoch": 0.06, "learning_rate": 7.677146599830798e-07, "loss": 0.8636, "step": 26389 }, { "epoch": 0.06, "learning_rate": 7.673005363314578e-07, "loss": 0.9095, "step": 26390 }, { "epoch": 0.06, "learning_rate": 7.668865199479214e-07, "loss": 0.866, "step": 26391 }, { "epoch": 0.06, "learning_rate": 7.664726108372777e-07, "loss": 0.8334, "step": 26392 }, { "epoch": 0.06, "learning_rate": 7.660588090043364e-07, "loss": 0.8244, "step": 26393 }, { "epoch": 0.06, "learning_rate": 7.656451144539057e-07, "loss": 0.8243, "step": 26394 }, { "epoch": 0.06, "learning_rate": 7.652315271907928e-07, "loss": 0.8412, "step": 26395 }, { "epoch": 0.06, "learning_rate": 7.648180472198008e-07, "loss": 0.8422, "step": 26396 }, { "epoch": 0.06, "learning_rate": 7.644046745457357e-07, "loss": 0.8773, "step": 26397 }, { "epoch": 0.06, "learning_rate": 7.639914091734013e-07, "loss": 0.8635, "step": 26398 }, { "epoch": 0.06, "learning_rate": 7.63578251107594e-07, "loss": 0.8494, "step": 26399 }, { "epoch": 0.06, "learning_rate": 7.631652003531187e-07, "loss": 0.8621, "step": 26400 }, { "epoch": 0.06, "learning_rate": 7.627522569147705e-07, "loss": 0.8314, "step": 26401 }, { "epoch": 0.06, "learning_rate": 7.623394207973522e-07, "loss": 0.8383, "step": 26402 }, { "epoch": 0.06, "learning_rate": 7.619266920056567e-07, "loss": 0.8985, "step": 26403 }, { "epoch": 0.06, "learning_rate": 7.61514070544479e-07, "loss": 0.8113, "step": 26404 }, { "epoch": 0.06, "learning_rate": 7.611015564186153e-07, "loss": 0.8345, "step": 26405 }, { "epoch": 0.06, "learning_rate": 7.606891496328561e-07, "loss": 0.8329, "step": 26406 }, { "epoch": 0.06, "learning_rate": 7.602768501919921e-07, "loss": 0.8721, "step": 26407 }, { "epoch": 0.06, "learning_rate": 7.598646581008162e-07, "loss": 0.8552, "step": 26408 }, { "epoch": 0.06, "learning_rate": 7.594525733641179e-07, "loss": 0.8248, "step": 26409 }, { "epoch": 0.06, "learning_rate": 7.59040595986682e-07, "loss": 0.8626, "step": 26410 }, { "epoch": 0.06, "learning_rate": 7.586287259732961e-07, "loss": 0.9014, "step": 26411 }, { "epoch": 0.06, "learning_rate": 7.58216963328745e-07, "loss": 0.8444, "step": 26412 }, { "epoch": 0.06, "learning_rate": 7.578053080578141e-07, "loss": 0.8663, "step": 26413 }, { "epoch": 0.06, "learning_rate": 7.573937601652848e-07, "loss": 0.8282, "step": 26414 }, { "epoch": 0.06, "learning_rate": 7.56982319655939e-07, "loss": 0.8262, "step": 26415 }, { "epoch": 0.06, "learning_rate": 7.565709865345594e-07, "loss": 0.87, "step": 26416 }, { "epoch": 0.06, "learning_rate": 7.561597608059201e-07, "loss": 0.8257, "step": 26417 }, { "epoch": 0.06, "learning_rate": 7.557486424748017e-07, "loss": 0.8671, "step": 26418 }, { "epoch": 0.06, "learning_rate": 7.553376315459792e-07, "loss": 0.8364, "step": 26419 }, { "epoch": 0.06, "learning_rate": 7.5492672802423e-07, "loss": 0.8368, "step": 26420 }, { "epoch": 0.06, "learning_rate": 7.545159319143269e-07, "loss": 0.8237, "step": 26421 }, { "epoch": 0.06, "learning_rate": 7.541052432210427e-07, "loss": 0.8411, "step": 26422 }, { "epoch": 0.06, "learning_rate": 7.536946619491503e-07, "loss": 0.8477, "step": 26423 }, { "epoch": 0.06, "learning_rate": 7.53284188103417e-07, "loss": 0.8369, "step": 26424 }, { "epoch": 0.06, "learning_rate": 7.528738216886122e-07, "loss": 0.8336, "step": 26425 }, { "epoch": 0.06, "learning_rate": 7.524635627095078e-07, "loss": 0.8403, "step": 26426 }, { "epoch": 0.06, "learning_rate": 7.520534111708656e-07, "loss": 0.852, "step": 26427 }, { "epoch": 0.06, "learning_rate": 7.516433670774526e-07, "loss": 0.8287, "step": 26428 }, { "epoch": 0.06, "learning_rate": 7.51233430434033e-07, "loss": 0.833, "step": 26429 }, { "epoch": 0.06, "learning_rate": 7.508236012453707e-07, "loss": 0.8366, "step": 26430 }, { "epoch": 0.06, "learning_rate": 7.504138795162231e-07, "loss": 0.8556, "step": 26431 }, { "epoch": 0.06, "learning_rate": 7.50004265251355e-07, "loss": 0.879, "step": 26432 }, { "epoch": 0.06, "learning_rate": 7.49594758455523e-07, "loss": 0.8093, "step": 26433 }, { "epoch": 0.06, "learning_rate": 7.491853591334874e-07, "loss": 0.853, "step": 26434 }, { "epoch": 0.06, "learning_rate": 7.487760672900013e-07, "loss": 0.8278, "step": 26435 }, { "epoch": 0.06, "learning_rate": 7.483668829298219e-07, "loss": 0.8618, "step": 26436 }, { "epoch": 0.06, "learning_rate": 7.479578060577019e-07, "loss": 0.8102, "step": 26437 }, { "epoch": 0.06, "learning_rate": 7.475488366783956e-07, "loss": 0.8328, "step": 26438 }, { "epoch": 0.06, "learning_rate": 7.471399747966546e-07, "loss": 0.8647, "step": 26439 }, { "epoch": 0.06, "learning_rate": 7.467312204172272e-07, "loss": 0.8452, "step": 26440 }, { "epoch": 0.06, "learning_rate": 7.463225735448654e-07, "loss": 0.8512, "step": 26441 }, { "epoch": 0.06, "learning_rate": 7.459140341843141e-07, "loss": 0.8319, "step": 26442 }, { "epoch": 0.06, "learning_rate": 7.455056023403217e-07, "loss": 0.8907, "step": 26443 }, { "epoch": 0.06, "learning_rate": 7.450972780176314e-07, "loss": 0.871, "step": 26444 }, { "epoch": 0.06, "learning_rate": 7.4468906122099e-07, "loss": 0.8174, "step": 26445 }, { "epoch": 0.06, "learning_rate": 7.442809519551375e-07, "loss": 0.828, "step": 26446 }, { "epoch": 0.06, "learning_rate": 7.438729502248177e-07, "loss": 0.871, "step": 26447 }, { "epoch": 0.06, "learning_rate": 7.434650560347712e-07, "loss": 0.8043, "step": 26448 }, { "epoch": 0.06, "learning_rate": 7.430572693897342e-07, "loss": 0.8256, "step": 26449 }, { "epoch": 0.06, "learning_rate": 7.426495902944453e-07, "loss": 0.8685, "step": 26450 }, { "epoch": 0.06, "learning_rate": 7.422420187536439e-07, "loss": 0.8683, "step": 26451 }, { "epoch": 0.06, "learning_rate": 7.418345547720618e-07, "loss": 0.8301, "step": 26452 }, { "epoch": 0.06, "learning_rate": 7.414271983544352e-07, "loss": 0.8523, "step": 26453 }, { "epoch": 0.07, "learning_rate": 7.41019949505496e-07, "loss": 0.8409, "step": 26454 }, { "epoch": 0.07, "learning_rate": 7.406128082299746e-07, "loss": 0.8684, "step": 26455 }, { "epoch": 0.07, "learning_rate": 7.402057745326042e-07, "loss": 0.869, "step": 26456 }, { "epoch": 0.07, "learning_rate": 7.397988484181107e-07, "loss": 0.8291, "step": 26457 }, { "epoch": 0.07, "learning_rate": 7.39392029891226e-07, "loss": 0.8596, "step": 26458 }, { "epoch": 0.07, "learning_rate": 7.38985318956672e-07, "loss": 0.8869, "step": 26459 }, { "epoch": 0.07, "learning_rate": 7.38578715619176e-07, "loss": 0.7849, "step": 26460 }, { "epoch": 0.07, "learning_rate": 7.381722198834618e-07, "loss": 0.8566, "step": 26461 }, { "epoch": 0.07, "learning_rate": 7.377658317542524e-07, "loss": 0.85, "step": 26462 }, { "epoch": 0.07, "learning_rate": 7.373595512362697e-07, "loss": 0.88, "step": 26463 }, { "epoch": 0.07, "learning_rate": 7.369533783342331e-07, "loss": 0.8324, "step": 26464 }, { "epoch": 0.07, "learning_rate": 7.365473130528644e-07, "loss": 0.8492, "step": 26465 }, { "epoch": 0.07, "learning_rate": 7.361413553968766e-07, "loss": 0.8615, "step": 26466 }, { "epoch": 0.07, "learning_rate": 7.357355053709892e-07, "loss": 0.8776, "step": 26467 }, { "epoch": 0.07, "learning_rate": 7.353297629799172e-07, "loss": 0.8577, "step": 26468 }, { "epoch": 0.07, "learning_rate": 7.349241282283736e-07, "loss": 0.8634, "step": 26469 }, { "epoch": 0.07, "learning_rate": 7.345186011210725e-07, "loss": 0.8789, "step": 26470 }, { "epoch": 0.07, "learning_rate": 7.341131816627256e-07, "loss": 0.8226, "step": 26471 }, { "epoch": 0.07, "learning_rate": 7.337078698580436e-07, "loss": 0.8679, "step": 26472 }, { "epoch": 0.07, "learning_rate": 7.333026657117326e-07, "loss": 0.8427, "step": 26473 }, { "epoch": 0.07, "learning_rate": 7.328975692285001e-07, "loss": 0.8567, "step": 26474 }, { "epoch": 0.07, "learning_rate": 7.324925804130567e-07, "loss": 0.8469, "step": 26475 }, { "epoch": 0.07, "learning_rate": 7.320876992701086e-07, "loss": 0.8241, "step": 26476 }, { "epoch": 0.07, "learning_rate": 7.316829258043534e-07, "loss": 0.8538, "step": 26477 }, { "epoch": 0.07, "learning_rate": 7.312782600204993e-07, "loss": 0.847, "step": 26478 }, { "epoch": 0.07, "learning_rate": 7.308737019232448e-07, "loss": 0.86, "step": 26479 }, { "epoch": 0.07, "learning_rate": 7.304692515172906e-07, "loss": 0.8779, "step": 26480 }, { "epoch": 0.07, "learning_rate": 7.300649088073375e-07, "loss": 0.8422, "step": 26481 }, { "epoch": 0.07, "learning_rate": 7.296606737980816e-07, "loss": 0.8377, "step": 26482 }, { "epoch": 0.07, "learning_rate": 7.292565464942225e-07, "loss": 0.8547, "step": 26483 }, { "epoch": 0.07, "learning_rate": 7.288525269004509e-07, "loss": 0.831, "step": 26484 }, { "epoch": 0.07, "learning_rate": 7.284486150214621e-07, "loss": 0.7886, "step": 26485 }, { "epoch": 0.07, "learning_rate": 7.280448108619509e-07, "loss": 0.8384, "step": 26486 }, { "epoch": 0.07, "learning_rate": 7.27641114426606e-07, "loss": 0.7959, "step": 26487 }, { "epoch": 0.07, "learning_rate": 7.272375257201203e-07, "loss": 0.8263, "step": 26488 }, { "epoch": 0.07, "learning_rate": 7.268340447471811e-07, "loss": 0.8164, "step": 26489 }, { "epoch": 0.07, "learning_rate": 7.264306715124791e-07, "loss": 0.8462, "step": 26490 }, { "epoch": 0.07, "learning_rate": 7.260274060206962e-07, "loss": 0.838, "step": 26491 }, { "epoch": 0.07, "learning_rate": 7.256242482765207e-07, "loss": 0.8433, "step": 26492 }, { "epoch": 0.07, "learning_rate": 7.252211982846335e-07, "loss": 0.8577, "step": 26493 }, { "epoch": 0.07, "learning_rate": 7.24818256049723e-07, "loss": 0.8451, "step": 26494 }, { "epoch": 0.07, "learning_rate": 7.244154215764665e-07, "loss": 0.857, "step": 26495 }, { "epoch": 0.07, "learning_rate": 7.240126948695437e-07, "loss": 0.8443, "step": 26496 }, { "epoch": 0.07, "learning_rate": 7.236100759336373e-07, "loss": 0.8686, "step": 26497 }, { "epoch": 0.07, "learning_rate": 7.232075647734194e-07, "loss": 0.8307, "step": 26498 }, { "epoch": 0.07, "learning_rate": 7.228051613935716e-07, "loss": 0.8245, "step": 26499 }, { "epoch": 0.07, "learning_rate": 7.22402865798768e-07, "loss": 0.8512, "step": 26500 }, { "epoch": 0.07, "learning_rate": 7.220006779936828e-07, "loss": 0.861, "step": 26501 }, { "epoch": 0.07, "learning_rate": 7.215985979829853e-07, "loss": 0.8454, "step": 26502 }, { "epoch": 0.07, "learning_rate": 7.21196625771351e-07, "loss": 0.8605, "step": 26503 }, { "epoch": 0.07, "learning_rate": 7.207947613634481e-07, "loss": 0.8249, "step": 26504 }, { "epoch": 0.07, "learning_rate": 7.203930047639463e-07, "loss": 0.8525, "step": 26505 }, { "epoch": 0.07, "learning_rate": 7.199913559775129e-07, "loss": 0.87, "step": 26506 }, { "epoch": 0.07, "learning_rate": 7.195898150088155e-07, "loss": 0.8906, "step": 26507 }, { "epoch": 0.07, "learning_rate": 7.19188381862519e-07, "loss": 0.8462, "step": 26508 }, { "epoch": 0.07, "learning_rate": 7.187870565432853e-07, "loss": 0.8488, "step": 26509 }, { "epoch": 0.07, "learning_rate": 7.183858390557796e-07, "loss": 0.886, "step": 26510 }, { "epoch": 0.07, "learning_rate": 7.179847294046615e-07, "loss": 0.8567, "step": 26511 }, { "epoch": 0.07, "learning_rate": 7.175837275945918e-07, "loss": 0.8523, "step": 26512 }, { "epoch": 0.07, "learning_rate": 7.1718283363023e-07, "loss": 0.8121, "step": 26513 }, { "epoch": 0.07, "learning_rate": 7.167820475162335e-07, "loss": 0.8491, "step": 26514 }, { "epoch": 0.07, "learning_rate": 7.163813692572607e-07, "loss": 0.8781, "step": 26515 }, { "epoch": 0.07, "learning_rate": 7.159807988579637e-07, "loss": 0.8647, "step": 26516 }, { "epoch": 0.07, "learning_rate": 7.155803363229952e-07, "loss": 0.8602, "step": 26517 }, { "epoch": 0.07, "learning_rate": 7.151799816570138e-07, "loss": 0.9096, "step": 26518 }, { "epoch": 0.07, "learning_rate": 7.147797348646657e-07, "loss": 0.8165, "step": 26519 }, { "epoch": 0.07, "learning_rate": 7.143795959506028e-07, "loss": 0.8468, "step": 26520 }, { "epoch": 0.07, "learning_rate": 7.139795649194736e-07, "loss": 0.8226, "step": 26521 }, { "epoch": 0.07, "learning_rate": 7.135796417759289e-07, "loss": 0.8676, "step": 26522 }, { "epoch": 0.07, "learning_rate": 7.131798265246082e-07, "loss": 0.8517, "step": 26523 }, { "epoch": 0.07, "learning_rate": 7.127801191701633e-07, "loss": 0.8252, "step": 26524 }, { "epoch": 0.07, "learning_rate": 7.123805197172351e-07, "loss": 0.8429, "step": 26525 }, { "epoch": 0.07, "learning_rate": 7.119810281704687e-07, "loss": 0.8925, "step": 26526 }, { "epoch": 0.07, "learning_rate": 7.115816445345014e-07, "loss": 0.8674, "step": 26527 }, { "epoch": 0.07, "learning_rate": 7.111823688139763e-07, "loss": 0.7822, "step": 26528 }, { "epoch": 0.07, "learning_rate": 7.107832010135319e-07, "loss": 0.8479, "step": 26529 }, { "epoch": 0.07, "learning_rate": 7.103841411378055e-07, "loss": 0.827, "step": 26530 }, { "epoch": 0.07, "learning_rate": 7.099851891914322e-07, "loss": 0.8771, "step": 26531 }, { "epoch": 0.07, "learning_rate": 7.095863451790497e-07, "loss": 0.7963, "step": 26532 }, { "epoch": 0.07, "learning_rate": 7.091876091052919e-07, "loss": 0.8482, "step": 26533 }, { "epoch": 0.07, "learning_rate": 7.087889809747883e-07, "loss": 0.8303, "step": 26534 }, { "epoch": 0.07, "learning_rate": 7.083904607921733e-07, "loss": 0.8657, "step": 26535 }, { "epoch": 0.07, "learning_rate": 7.079920485620751e-07, "loss": 0.8712, "step": 26536 }, { "epoch": 0.07, "learning_rate": 7.075937442891234e-07, "loss": 0.8144, "step": 26537 }, { "epoch": 0.07, "learning_rate": 7.071955479779457e-07, "loss": 0.8218, "step": 26538 }, { "epoch": 0.07, "learning_rate": 7.067974596331684e-07, "loss": 0.8273, "step": 26539 }, { "epoch": 0.07, "learning_rate": 7.063994792594175e-07, "loss": 0.8477, "step": 26540 }, { "epoch": 0.07, "learning_rate": 7.060016068613151e-07, "loss": 0.8061, "step": 26541 }, { "epoch": 0.07, "learning_rate": 7.05603842443483e-07, "loss": 0.8411, "step": 26542 }, { "epoch": 0.07, "learning_rate": 7.052061860105475e-07, "loss": 0.8329, "step": 26543 }, { "epoch": 0.07, "learning_rate": 7.048086375671237e-07, "loss": 0.8248, "step": 26544 }, { "epoch": 0.07, "learning_rate": 7.044111971178313e-07, "loss": 0.8131, "step": 26545 }, { "epoch": 0.07, "learning_rate": 7.040138646672901e-07, "loss": 0.8703, "step": 26546 }, { "epoch": 0.07, "learning_rate": 7.036166402201139e-07, "loss": 0.8185, "step": 26547 }, { "epoch": 0.07, "learning_rate": 7.032195237809192e-07, "loss": 0.8462, "step": 26548 }, { "epoch": 0.07, "learning_rate": 7.0282251535432e-07, "loss": 0.8432, "step": 26549 }, { "epoch": 0.07, "learning_rate": 7.024256149449293e-07, "loss": 0.8753, "step": 26550 }, { "epoch": 0.07, "learning_rate": 7.020288225573557e-07, "loss": 0.8875, "step": 26551 }, { "epoch": 0.07, "learning_rate": 7.01632138196211e-07, "loss": 0.8517, "step": 26552 }, { "epoch": 0.07, "learning_rate": 7.012355618661048e-07, "loss": 0.8462, "step": 26553 }, { "epoch": 0.07, "learning_rate": 7.008390935716425e-07, "loss": 0.8246, "step": 26554 }, { "epoch": 0.07, "learning_rate": 7.004427333174324e-07, "loss": 0.8821, "step": 26555 }, { "epoch": 0.07, "learning_rate": 7.000464811080787e-07, "loss": 0.832, "step": 26556 }, { "epoch": 0.07, "learning_rate": 6.996503369481867e-07, "loss": 0.8555, "step": 26557 }, { "epoch": 0.07, "learning_rate": 6.992543008423558e-07, "loss": 0.8195, "step": 26558 }, { "epoch": 0.07, "learning_rate": 6.988583727951892e-07, "loss": 0.8812, "step": 26559 }, { "epoch": 0.07, "learning_rate": 6.984625528112854e-07, "loss": 0.828, "step": 26560 }, { "epoch": 0.07, "learning_rate": 6.980668408952474e-07, "loss": 0.8648, "step": 26561 }, { "epoch": 0.07, "learning_rate": 6.97671237051668e-07, "loss": 0.8259, "step": 26562 }, { "epoch": 0.07, "learning_rate": 6.972757412851449e-07, "loss": 0.8746, "step": 26563 }, { "epoch": 0.07, "learning_rate": 6.968803536002744e-07, "loss": 0.8481, "step": 26564 }, { "epoch": 0.07, "learning_rate": 6.964850740016482e-07, "loss": 0.7913, "step": 26565 }, { "epoch": 0.07, "learning_rate": 6.960899024938583e-07, "loss": 0.8675, "step": 26566 }, { "epoch": 0.07, "learning_rate": 6.956948390814977e-07, "loss": 0.857, "step": 26567 }, { "epoch": 0.07, "learning_rate": 6.952998837691593e-07, "loss": 0.818, "step": 26568 }, { "epoch": 0.07, "learning_rate": 6.949050365614251e-07, "loss": 0.8346, "step": 26569 }, { "epoch": 0.07, "learning_rate": 6.94510297462887e-07, "loss": 0.8271, "step": 26570 }, { "epoch": 0.07, "learning_rate": 6.941156664781301e-07, "loss": 0.8447, "step": 26571 }, { "epoch": 0.07, "learning_rate": 6.937211436117386e-07, "loss": 0.8559, "step": 26572 }, { "epoch": 0.07, "learning_rate": 6.933267288682977e-07, "loss": 0.8198, "step": 26573 }, { "epoch": 0.07, "learning_rate": 6.929324222523892e-07, "loss": 0.8624, "step": 26574 }, { "epoch": 0.07, "learning_rate": 6.925382237685951e-07, "loss": 0.8564, "step": 26575 }, { "epoch": 0.07, "learning_rate": 6.921441334214929e-07, "loss": 0.8433, "step": 26576 }, { "epoch": 0.07, "learning_rate": 6.917501512156621e-07, "loss": 0.8306, "step": 26577 }, { "epoch": 0.07, "learning_rate": 6.913562771556814e-07, "loss": 0.8561, "step": 26578 }, { "epoch": 0.07, "learning_rate": 6.909625112461261e-07, "loss": 0.8994, "step": 26579 }, { "epoch": 0.07, "learning_rate": 6.9056885349157e-07, "loss": 0.8225, "step": 26580 }, { "epoch": 0.07, "learning_rate": 6.901753038965886e-07, "loss": 0.8243, "step": 26581 }, { "epoch": 0.07, "learning_rate": 6.89781862465756e-07, "loss": 0.8565, "step": 26582 }, { "epoch": 0.07, "learning_rate": 6.893885292036384e-07, "loss": 0.8396, "step": 26583 }, { "epoch": 0.07, "learning_rate": 6.889953041148079e-07, "loss": 0.8789, "step": 26584 }, { "epoch": 0.07, "learning_rate": 6.886021872038329e-07, "loss": 0.8367, "step": 26585 }, { "epoch": 0.07, "learning_rate": 6.882091784752831e-07, "loss": 0.8698, "step": 26586 }, { "epoch": 0.07, "learning_rate": 6.878162779337217e-07, "loss": 0.8791, "step": 26587 }, { "epoch": 0.07, "learning_rate": 6.874234855837148e-07, "loss": 0.8385, "step": 26588 }, { "epoch": 0.07, "learning_rate": 6.870308014298266e-07, "loss": 0.8558, "step": 26589 }, { "epoch": 0.07, "learning_rate": 6.866382254766158e-07, "loss": 0.832, "step": 26590 }, { "epoch": 0.07, "learning_rate": 6.862457577286474e-07, "loss": 0.8896, "step": 26591 }, { "epoch": 0.07, "learning_rate": 6.858533981904791e-07, "loss": 0.8468, "step": 26592 }, { "epoch": 0.07, "learning_rate": 6.854611468666727e-07, "loss": 0.834, "step": 26593 }, { "epoch": 0.07, "learning_rate": 6.850690037617824e-07, "loss": 0.8664, "step": 26594 }, { "epoch": 0.07, "learning_rate": 6.846769688803633e-07, "loss": 0.8977, "step": 26595 }, { "epoch": 0.07, "learning_rate": 6.842850422269731e-07, "loss": 0.8465, "step": 26596 }, { "epoch": 0.07, "learning_rate": 6.838932238061625e-07, "loss": 0.8608, "step": 26597 }, { "epoch": 0.07, "learning_rate": 6.835015136224865e-07, "loss": 0.8257, "step": 26598 }, { "epoch": 0.07, "learning_rate": 6.83109911680494e-07, "loss": 0.8559, "step": 26599 }, { "epoch": 0.07, "learning_rate": 6.827184179847379e-07, "loss": 0.8629, "step": 26600 }, { "epoch": 0.07, "learning_rate": 6.823270325397624e-07, "loss": 0.843, "step": 26601 }, { "epoch": 0.07, "learning_rate": 6.819357553501161e-07, "loss": 0.8445, "step": 26602 }, { "epoch": 0.07, "learning_rate": 6.815445864203463e-07, "loss": 0.8403, "step": 26603 }, { "epoch": 0.07, "learning_rate": 6.811535257549962e-07, "loss": 0.833, "step": 26604 }, { "epoch": 0.07, "learning_rate": 6.80762573358611e-07, "loss": 0.8042, "step": 26605 }, { "epoch": 0.07, "learning_rate": 6.803717292357304e-07, "loss": 0.8374, "step": 26606 }, { "epoch": 0.07, "learning_rate": 6.799809933908996e-07, "loss": 0.92, "step": 26607 }, { "epoch": 0.07, "learning_rate": 6.79590365828654e-07, "loss": 0.8693, "step": 26608 }, { "epoch": 0.07, "learning_rate": 6.79199846553531e-07, "loss": 0.831, "step": 26609 }, { "epoch": 0.07, "learning_rate": 6.788094355700725e-07, "loss": 0.8464, "step": 26610 }, { "epoch": 0.07, "learning_rate": 6.784191328828138e-07, "loss": 0.8222, "step": 26611 }, { "epoch": 0.07, "learning_rate": 6.780289384962868e-07, "loss": 0.8255, "step": 26612 }, { "epoch": 0.07, "learning_rate": 6.776388524150257e-07, "loss": 0.8062, "step": 26613 }, { "epoch": 0.07, "learning_rate": 6.772488746435646e-07, "loss": 0.845, "step": 26614 }, { "epoch": 0.07, "learning_rate": 6.768590051864288e-07, "loss": 0.8107, "step": 26615 }, { "epoch": 0.07, "learning_rate": 6.764692440481546e-07, "loss": 0.8454, "step": 26616 }, { "epoch": 0.07, "learning_rate": 6.760795912332663e-07, "loss": 0.8338, "step": 26617 }, { "epoch": 0.07, "learning_rate": 6.756900467462946e-07, "loss": 0.832, "step": 26618 }, { "epoch": 0.07, "learning_rate": 6.753006105917615e-07, "loss": 0.8596, "step": 26619 }, { "epoch": 0.07, "learning_rate": 6.749112827741932e-07, "loss": 0.8471, "step": 26620 }, { "epoch": 0.07, "learning_rate": 6.745220632981131e-07, "loss": 0.8532, "step": 26621 }, { "epoch": 0.07, "learning_rate": 6.741329521680417e-07, "loss": 0.8409, "step": 26622 }, { "epoch": 0.07, "learning_rate": 6.737439493885023e-07, "loss": 0.8439, "step": 26623 }, { "epoch": 0.07, "learning_rate": 6.733550549640133e-07, "loss": 0.8689, "step": 26624 }, { "epoch": 0.07, "learning_rate": 6.729662688990945e-07, "loss": 0.8725, "step": 26625 }, { "epoch": 0.07, "learning_rate": 6.725775911982602e-07, "loss": 0.8728, "step": 26626 }, { "epoch": 0.07, "learning_rate": 6.721890218660265e-07, "loss": 0.8318, "step": 26627 }, { "epoch": 0.07, "learning_rate": 6.7180056090691e-07, "loss": 0.8579, "step": 26628 }, { "epoch": 0.07, "learning_rate": 6.714122083254216e-07, "loss": 0.8576, "step": 26629 }, { "epoch": 0.07, "learning_rate": 6.710239641260752e-07, "loss": 0.8198, "step": 26630 }, { "epoch": 0.07, "learning_rate": 6.706358283133807e-07, "loss": 0.8219, "step": 26631 }, { "epoch": 0.07, "learning_rate": 6.70247800891849e-07, "loss": 0.8025, "step": 26632 }, { "epoch": 0.07, "learning_rate": 6.69859881865984e-07, "loss": 0.8269, "step": 26633 }, { "epoch": 0.07, "learning_rate": 6.694720712402969e-07, "loss": 0.825, "step": 26634 }, { "epoch": 0.07, "learning_rate": 6.690843690192938e-07, "loss": 0.8578, "step": 26635 }, { "epoch": 0.07, "learning_rate": 6.686967752074757e-07, "loss": 0.8522, "step": 26636 }, { "epoch": 0.07, "learning_rate": 6.68309289809348e-07, "loss": 0.857, "step": 26637 }, { "epoch": 0.07, "learning_rate": 6.679219128294123e-07, "loss": 0.8464, "step": 26638 }, { "epoch": 0.07, "learning_rate": 6.675346442721686e-07, "loss": 0.8596, "step": 26639 }, { "epoch": 0.07, "learning_rate": 6.671474841421166e-07, "loss": 0.8175, "step": 26640 }, { "epoch": 0.07, "learning_rate": 6.667604324437549e-07, "loss": 0.8256, "step": 26641 }, { "epoch": 0.07, "learning_rate": 6.66373489181581e-07, "loss": 0.8289, "step": 26642 }, { "epoch": 0.07, "learning_rate": 6.65986654360089e-07, "loss": 0.8621, "step": 26643 }, { "epoch": 0.07, "learning_rate": 6.655999279837733e-07, "loss": 0.8319, "step": 26644 }, { "epoch": 0.07, "learning_rate": 6.652133100571278e-07, "loss": 0.8139, "step": 26645 }, { "epoch": 0.07, "learning_rate": 6.648268005846436e-07, "loss": 0.8487, "step": 26646 }, { "epoch": 0.07, "learning_rate": 6.644403995708126e-07, "loss": 0.8483, "step": 26647 }, { "epoch": 0.07, "learning_rate": 6.640541070201223e-07, "loss": 0.8748, "step": 26648 }, { "epoch": 0.07, "learning_rate": 6.636679229370646e-07, "loss": 0.8544, "step": 26649 }, { "epoch": 0.07, "learning_rate": 6.632818473261205e-07, "loss": 0.8048, "step": 26650 }, { "epoch": 0.07, "learning_rate": 6.628958801917795e-07, "loss": 0.8812, "step": 26651 }, { "epoch": 0.07, "learning_rate": 6.625100215385227e-07, "loss": 0.8558, "step": 26652 }, { "epoch": 0.07, "learning_rate": 6.621242713708375e-07, "loss": 0.8338, "step": 26653 }, { "epoch": 0.07, "learning_rate": 6.617386296932038e-07, "loss": 0.8605, "step": 26654 }, { "epoch": 0.07, "learning_rate": 6.613530965101e-07, "loss": 0.8338, "step": 26655 }, { "epoch": 0.07, "learning_rate": 6.609676718260083e-07, "loss": 0.828, "step": 26656 }, { "epoch": 0.07, "learning_rate": 6.605823556454061e-07, "loss": 0.8498, "step": 26657 }, { "epoch": 0.07, "learning_rate": 6.601971479727665e-07, "loss": 0.8576, "step": 26658 }, { "epoch": 0.07, "learning_rate": 6.598120488125692e-07, "loss": 0.8502, "step": 26659 }, { "epoch": 0.07, "learning_rate": 6.594270581692885e-07, "loss": 0.821, "step": 26660 }, { "epoch": 0.07, "learning_rate": 6.590421760473953e-07, "loss": 0.8235, "step": 26661 }, { "epoch": 0.07, "learning_rate": 6.586574024513603e-07, "loss": 0.8742, "step": 26662 }, { "epoch": 0.07, "learning_rate": 6.582727373856568e-07, "loss": 0.8488, "step": 26663 }, { "epoch": 0.07, "learning_rate": 6.578881808547522e-07, "loss": 0.7964, "step": 26664 }, { "epoch": 0.07, "learning_rate": 6.575037328631151e-07, "loss": 0.8482, "step": 26665 }, { "epoch": 0.07, "learning_rate": 6.57119393415211e-07, "loss": 0.859, "step": 26666 }, { "epoch": 0.07, "learning_rate": 6.567351625155082e-07, "loss": 0.8455, "step": 26667 }, { "epoch": 0.07, "learning_rate": 6.56351040168467e-07, "loss": 0.8598, "step": 26668 }, { "epoch": 0.07, "learning_rate": 6.559670263785522e-07, "loss": 0.8443, "step": 26669 }, { "epoch": 0.07, "learning_rate": 6.555831211502239e-07, "loss": 0.8543, "step": 26670 }, { "epoch": 0.07, "learning_rate": 6.551993244879451e-07, "loss": 0.8586, "step": 26671 }, { "epoch": 0.07, "learning_rate": 6.548156363961722e-07, "loss": 0.8911, "step": 26672 }, { "epoch": 0.07, "learning_rate": 6.54432056879365e-07, "loss": 0.8185, "step": 26673 }, { "epoch": 0.07, "learning_rate": 6.54048585941981e-07, "loss": 0.8641, "step": 26674 }, { "epoch": 0.07, "learning_rate": 6.536652235884711e-07, "loss": 0.8577, "step": 26675 }, { "epoch": 0.07, "learning_rate": 6.532819698232918e-07, "loss": 0.8706, "step": 26676 }, { "epoch": 0.07, "learning_rate": 6.52898824650895e-07, "loss": 0.8242, "step": 26677 }, { "epoch": 0.07, "learning_rate": 6.525157880757361e-07, "loss": 0.8587, "step": 26678 }, { "epoch": 0.07, "learning_rate": 6.521328601022614e-07, "loss": 0.8943, "step": 26679 }, { "epoch": 0.07, "learning_rate": 6.517500407349198e-07, "loss": 0.8716, "step": 26680 }, { "epoch": 0.07, "learning_rate": 6.51367329978162e-07, "loss": 0.8196, "step": 26681 }, { "epoch": 0.07, "learning_rate": 6.50984727836429e-07, "loss": 0.8498, "step": 26682 }, { "epoch": 0.07, "learning_rate": 6.506022343141715e-07, "loss": 0.8267, "step": 26683 }, { "epoch": 0.07, "learning_rate": 6.502198494158318e-07, "loss": 0.8509, "step": 26684 }, { "epoch": 0.07, "learning_rate": 6.498375731458529e-07, "loss": 0.8172, "step": 26685 }, { "epoch": 0.07, "learning_rate": 6.494554055086732e-07, "loss": 0.8322, "step": 26686 }, { "epoch": 0.07, "learning_rate": 6.490733465087362e-07, "loss": 0.8688, "step": 26687 }, { "epoch": 0.07, "learning_rate": 6.486913961504793e-07, "loss": 0.869, "step": 26688 }, { "epoch": 0.07, "learning_rate": 6.4830955443834e-07, "loss": 0.8495, "step": 26689 }, { "epoch": 0.07, "learning_rate": 6.479278213767559e-07, "loss": 0.8176, "step": 26690 }, { "epoch": 0.07, "learning_rate": 6.475461969701602e-07, "loss": 0.8401, "step": 26691 }, { "epoch": 0.07, "learning_rate": 6.471646812229893e-07, "loss": 0.8784, "step": 26692 }, { "epoch": 0.07, "learning_rate": 6.46783274139674e-07, "loss": 0.8321, "step": 26693 }, { "epoch": 0.07, "learning_rate": 6.464019757246442e-07, "loss": 0.8494, "step": 26694 }, { "epoch": 0.07, "learning_rate": 6.46020785982332e-07, "loss": 0.83, "step": 26695 }, { "epoch": 0.07, "learning_rate": 6.456397049171658e-07, "loss": 0.8462, "step": 26696 }, { "epoch": 0.07, "learning_rate": 6.452587325335735e-07, "loss": 0.8415, "step": 26697 }, { "epoch": 0.07, "learning_rate": 6.448778688359803e-07, "loss": 0.8443, "step": 26698 }, { "epoch": 0.07, "learning_rate": 6.444971138288136e-07, "loss": 0.8551, "step": 26699 }, { "epoch": 0.07, "learning_rate": 6.441164675164935e-07, "loss": 0.8297, "step": 26700 }, { "epoch": 0.07, "learning_rate": 6.437359299034429e-07, "loss": 0.8482, "step": 26701 }, { "epoch": 0.07, "learning_rate": 6.433555009940851e-07, "loss": 0.8506, "step": 26702 }, { "epoch": 0.07, "learning_rate": 6.429751807928408e-07, "loss": 0.8492, "step": 26703 }, { "epoch": 0.07, "learning_rate": 6.425949693041267e-07, "loss": 0.8752, "step": 26704 }, { "epoch": 0.07, "learning_rate": 6.42214866532359e-07, "loss": 0.8468, "step": 26705 }, { "epoch": 0.07, "learning_rate": 6.418348724819568e-07, "loss": 0.8593, "step": 26706 }, { "epoch": 0.07, "learning_rate": 6.414549871573328e-07, "loss": 0.8887, "step": 26707 }, { "epoch": 0.07, "learning_rate": 6.410752105629026e-07, "loss": 0.812, "step": 26708 }, { "epoch": 0.07, "learning_rate": 6.406955427030759e-07, "loss": 0.844, "step": 26709 }, { "epoch": 0.07, "learning_rate": 6.40315983582267e-07, "loss": 0.8633, "step": 26710 }, { "epoch": 0.07, "learning_rate": 6.399365332048835e-07, "loss": 0.8466, "step": 26711 }, { "epoch": 0.07, "learning_rate": 6.39557191575334e-07, "loss": 0.8564, "step": 26712 }, { "epoch": 0.07, "learning_rate": 6.391779586980252e-07, "loss": 0.8281, "step": 26713 }, { "epoch": 0.07, "learning_rate": 6.387988345773654e-07, "loss": 0.8621, "step": 26714 }, { "epoch": 0.07, "learning_rate": 6.384198192177571e-07, "loss": 0.8592, "step": 26715 }, { "epoch": 0.07, "learning_rate": 6.380409126236064e-07, "loss": 0.8187, "step": 26716 }, { "epoch": 0.07, "learning_rate": 6.376621147993146e-07, "loss": 0.834, "step": 26717 }, { "epoch": 0.07, "learning_rate": 6.372834257492799e-07, "loss": 0.858, "step": 26718 }, { "epoch": 0.07, "learning_rate": 6.369048454779048e-07, "loss": 0.8818, "step": 26719 }, { "epoch": 0.07, "learning_rate": 6.365263739895878e-07, "loss": 0.8058, "step": 26720 }, { "epoch": 0.07, "learning_rate": 6.361480112887241e-07, "loss": 0.8511, "step": 26721 }, { "epoch": 0.07, "learning_rate": 6.357697573797117e-07, "loss": 0.8266, "step": 26722 }, { "epoch": 0.07, "learning_rate": 6.353916122669446e-07, "loss": 0.8501, "step": 26723 }, { "epoch": 0.07, "learning_rate": 6.35013575954817e-07, "loss": 0.8556, "step": 26724 }, { "epoch": 0.07, "learning_rate": 6.346356484477179e-07, "loss": 0.8317, "step": 26725 }, { "epoch": 0.07, "learning_rate": 6.342578297500412e-07, "loss": 0.8184, "step": 26726 }, { "epoch": 0.07, "learning_rate": 6.338801198661782e-07, "loss": 0.898, "step": 26727 }, { "epoch": 0.07, "learning_rate": 6.335025188005128e-07, "loss": 0.816, "step": 26728 }, { "epoch": 0.07, "learning_rate": 6.33125026557434e-07, "loss": 0.8202, "step": 26729 }, { "epoch": 0.07, "learning_rate": 6.32747643141327e-07, "loss": 0.8304, "step": 26730 }, { "epoch": 0.07, "learning_rate": 6.323703685565774e-07, "loss": 0.8594, "step": 26731 }, { "epoch": 0.07, "learning_rate": 6.319932028075682e-07, "loss": 0.8345, "step": 26732 }, { "epoch": 0.07, "learning_rate": 6.316161458986813e-07, "loss": 0.815, "step": 26733 }, { "epoch": 0.07, "learning_rate": 6.312391978342991e-07, "loss": 0.8553, "step": 26734 }, { "epoch": 0.07, "learning_rate": 6.308623586187978e-07, "loss": 0.8614, "step": 26735 }, { "epoch": 0.07, "learning_rate": 6.304856282565563e-07, "loss": 0.8664, "step": 26736 }, { "epoch": 0.07, "learning_rate": 6.301090067519533e-07, "loss": 0.8229, "step": 26737 }, { "epoch": 0.07, "learning_rate": 6.297324941093641e-07, "loss": 0.8456, "step": 26738 }, { "epoch": 0.07, "learning_rate": 6.293560903331619e-07, "loss": 0.8784, "step": 26739 }, { "epoch": 0.07, "learning_rate": 6.28979795427721e-07, "loss": 0.8601, "step": 26740 }, { "epoch": 0.07, "learning_rate": 6.286036093974146e-07, "loss": 0.8627, "step": 26741 }, { "epoch": 0.07, "learning_rate": 6.282275322466103e-07, "loss": 0.8324, "step": 26742 }, { "epoch": 0.07, "learning_rate": 6.27851563979679e-07, "loss": 0.8555, "step": 26743 }, { "epoch": 0.07, "learning_rate": 6.274757046009871e-07, "loss": 0.8504, "step": 26744 }, { "epoch": 0.07, "learning_rate": 6.270999541149059e-07, "loss": 0.8147, "step": 26745 }, { "epoch": 0.07, "learning_rate": 6.267243125257971e-07, "loss": 0.8553, "step": 26746 }, { "epoch": 0.07, "learning_rate": 6.263487798380252e-07, "loss": 0.8468, "step": 26747 }, { "epoch": 0.07, "learning_rate": 6.259733560559544e-07, "loss": 0.8183, "step": 26748 }, { "epoch": 0.07, "learning_rate": 6.25598041183948e-07, "loss": 0.8102, "step": 26749 }, { "epoch": 0.07, "learning_rate": 6.252228352263612e-07, "loss": 0.8575, "step": 26750 }, { "epoch": 0.07, "learning_rate": 6.248477381875573e-07, "loss": 0.8475, "step": 26751 }, { "epoch": 0.07, "learning_rate": 6.244727500718973e-07, "loss": 0.8685, "step": 26752 }, { "epoch": 0.07, "learning_rate": 6.240978708837308e-07, "loss": 0.86, "step": 26753 }, { "epoch": 0.08, "learning_rate": 6.237231006274181e-07, "loss": 0.8517, "step": 26754 }, { "epoch": 0.08, "learning_rate": 6.233484393073108e-07, "loss": 0.8342, "step": 26755 }, { "epoch": 0.08, "learning_rate": 6.229738869277647e-07, "loss": 0.8243, "step": 26756 }, { "epoch": 0.08, "learning_rate": 6.225994434931282e-07, "loss": 0.8671, "step": 26757 }, { "epoch": 0.08, "learning_rate": 6.222251090077547e-07, "loss": 0.7905, "step": 26758 }, { "epoch": 0.08, "learning_rate": 6.218508834759929e-07, "loss": 0.8369, "step": 26759 }, { "epoch": 0.08, "learning_rate": 6.214767669021871e-07, "loss": 0.8219, "step": 26760 }, { "epoch": 0.08, "learning_rate": 6.211027592906882e-07, "loss": 0.8189, "step": 26761 }, { "epoch": 0.08, "learning_rate": 6.207288606458384e-07, "loss": 0.8668, "step": 26762 }, { "epoch": 0.08, "learning_rate": 6.203550709719841e-07, "loss": 0.8528, "step": 26763 }, { "epoch": 0.08, "learning_rate": 6.199813902734664e-07, "loss": 0.8396, "step": 26764 }, { "epoch": 0.08, "learning_rate": 6.196078185546283e-07, "loss": 0.8513, "step": 26765 }, { "epoch": 0.08, "learning_rate": 6.1923435581981e-07, "loss": 0.8239, "step": 26766 }, { "epoch": 0.08, "learning_rate": 6.188610020733487e-07, "loss": 0.861, "step": 26767 }, { "epoch": 0.08, "learning_rate": 6.184877573195824e-07, "loss": 0.81, "step": 26768 }, { "epoch": 0.08, "learning_rate": 6.181146215628497e-07, "loss": 0.8092, "step": 26769 }, { "epoch": 0.08, "learning_rate": 6.17741594807485e-07, "loss": 0.8857, "step": 26770 }, { "epoch": 0.08, "learning_rate": 6.173686770578213e-07, "loss": 0.8757, "step": 26771 }, { "epoch": 0.08, "learning_rate": 6.169958683181909e-07, "loss": 0.8494, "step": 26772 }, { "epoch": 0.08, "learning_rate": 6.166231685929281e-07, "loss": 0.8228, "step": 26773 }, { "epoch": 0.08, "learning_rate": 6.162505778863582e-07, "loss": 0.835, "step": 26774 }, { "epoch": 0.08, "learning_rate": 6.158780962028132e-07, "loss": 0.8137, "step": 26775 }, { "epoch": 0.08, "learning_rate": 6.15505723546621e-07, "loss": 0.8526, "step": 26776 }, { "epoch": 0.08, "learning_rate": 6.151334599221093e-07, "loss": 0.7988, "step": 26777 }, { "epoch": 0.08, "learning_rate": 6.147613053335988e-07, "loss": 0.8237, "step": 26778 }, { "epoch": 0.08, "learning_rate": 6.143892597854151e-07, "loss": 0.8508, "step": 26779 }, { "epoch": 0.08, "learning_rate": 6.140173232818824e-07, "loss": 0.8437, "step": 26780 }, { "epoch": 0.08, "learning_rate": 6.136454958273196e-07, "loss": 0.8327, "step": 26781 }, { "epoch": 0.08, "learning_rate": 6.132737774260489e-07, "loss": 0.8189, "step": 26782 }, { "epoch": 0.08, "learning_rate": 6.129021680823876e-07, "loss": 0.8455, "step": 26783 }, { "epoch": 0.08, "learning_rate": 6.125306678006549e-07, "loss": 0.8738, "step": 26784 }, { "epoch": 0.08, "learning_rate": 6.121592765851647e-07, "loss": 0.7924, "step": 26785 }, { "epoch": 0.08, "learning_rate": 6.117879944402328e-07, "loss": 0.8785, "step": 26786 }, { "epoch": 0.08, "learning_rate": 6.114168213701722e-07, "loss": 0.8778, "step": 26787 }, { "epoch": 0.08, "learning_rate": 6.110457573792961e-07, "loss": 0.8634, "step": 26788 }, { "epoch": 0.08, "learning_rate": 6.106748024719156e-07, "loss": 0.8398, "step": 26789 }, { "epoch": 0.08, "learning_rate": 6.103039566523416e-07, "loss": 0.8267, "step": 26790 }, { "epoch": 0.08, "learning_rate": 6.099332199248819e-07, "loss": 0.8603, "step": 26791 }, { "epoch": 0.08, "learning_rate": 6.09562592293842e-07, "loss": 0.8165, "step": 26792 }, { "epoch": 0.08, "learning_rate": 6.091920737635293e-07, "loss": 0.8466, "step": 26793 }, { "epoch": 0.08, "learning_rate": 6.088216643382483e-07, "loss": 0.8229, "step": 26794 }, { "epoch": 0.08, "learning_rate": 6.084513640223067e-07, "loss": 0.8567, "step": 26795 }, { "epoch": 0.08, "learning_rate": 6.080811728199998e-07, "loss": 0.8711, "step": 26796 }, { "epoch": 0.08, "learning_rate": 6.07711090735632e-07, "loss": 0.84, "step": 26797 }, { "epoch": 0.08, "learning_rate": 6.073411177735034e-07, "loss": 0.8325, "step": 26798 }, { "epoch": 0.08, "learning_rate": 6.069712539379125e-07, "loss": 0.8473, "step": 26799 }, { "epoch": 0.08, "learning_rate": 6.066014992331548e-07, "loss": 0.8182, "step": 26800 }, { "epoch": 0.08, "learning_rate": 6.06231853663527e-07, "loss": 0.8234, "step": 26801 }, { "epoch": 0.08, "learning_rate": 6.058623172333266e-07, "loss": 0.8645, "step": 26802 }, { "epoch": 0.08, "learning_rate": 6.054928899468427e-07, "loss": 0.8628, "step": 26803 }, { "epoch": 0.08, "learning_rate": 6.051235718083681e-07, "loss": 0.8452, "step": 26804 }, { "epoch": 0.08, "learning_rate": 6.047543628221963e-07, "loss": 0.8392, "step": 26805 }, { "epoch": 0.08, "learning_rate": 6.043852629926139e-07, "loss": 0.8475, "step": 26806 }, { "epoch": 0.08, "learning_rate": 6.040162723239107e-07, "loss": 0.8538, "step": 26807 }, { "epoch": 0.08, "learning_rate": 6.036473908203744e-07, "loss": 0.8551, "step": 26808 }, { "epoch": 0.08, "learning_rate": 6.032786184862905e-07, "loss": 0.8398, "step": 26809 }, { "epoch": 0.08, "learning_rate": 6.029099553259421e-07, "loss": 0.8335, "step": 26810 }, { "epoch": 0.08, "learning_rate": 6.025414013436115e-07, "loss": 0.8338, "step": 26811 }, { "epoch": 0.08, "learning_rate": 6.021729565435852e-07, "loss": 0.8771, "step": 26812 }, { "epoch": 0.08, "learning_rate": 6.018046209301409e-07, "loss": 0.8572, "step": 26813 }, { "epoch": 0.08, "learning_rate": 6.014363945075563e-07, "loss": 0.877, "step": 26814 }, { "epoch": 0.08, "learning_rate": 6.010682772801135e-07, "loss": 0.8851, "step": 26815 }, { "epoch": 0.08, "learning_rate": 6.007002692520869e-07, "loss": 0.8258, "step": 26816 }, { "epoch": 0.08, "learning_rate": 6.003323704277508e-07, "loss": 0.8336, "step": 26817 }, { "epoch": 0.08, "learning_rate": 5.999645808113829e-07, "loss": 0.8499, "step": 26818 }, { "epoch": 0.08, "learning_rate": 5.995969004072566e-07, "loss": 0.8708, "step": 26819 }, { "epoch": 0.08, "learning_rate": 5.992293292196393e-07, "loss": 0.8624, "step": 26820 }, { "epoch": 0.08, "learning_rate": 5.988618672528046e-07, "loss": 0.8356, "step": 26821 }, { "epoch": 0.08, "learning_rate": 5.984945145110221e-07, "loss": 0.8213, "step": 26822 }, { "epoch": 0.08, "learning_rate": 5.981272709985575e-07, "loss": 0.8885, "step": 26823 }, { "epoch": 0.08, "learning_rate": 5.977601367196806e-07, "loss": 0.8506, "step": 26824 }, { "epoch": 0.08, "learning_rate": 5.973931116786546e-07, "loss": 0.8583, "step": 26825 }, { "epoch": 0.08, "learning_rate": 5.970261958797452e-07, "loss": 0.8454, "step": 26826 }, { "epoch": 0.08, "learning_rate": 5.966593893272143e-07, "loss": 0.8842, "step": 26827 }, { "epoch": 0.08, "learning_rate": 5.962926920253232e-07, "loss": 0.8379, "step": 26828 }, { "epoch": 0.08, "learning_rate": 5.959261039783326e-07, "loss": 0.8224, "step": 26829 }, { "epoch": 0.08, "learning_rate": 5.955596251905016e-07, "loss": 0.8534, "step": 26830 }, { "epoch": 0.08, "learning_rate": 5.951932556660888e-07, "loss": 0.8773, "step": 26831 }, { "epoch": 0.08, "learning_rate": 5.948269954093511e-07, "loss": 0.7908, "step": 26832 }, { "epoch": 0.08, "learning_rate": 5.944608444245436e-07, "loss": 0.8505, "step": 26833 }, { "epoch": 0.08, "learning_rate": 5.940948027159188e-07, "loss": 0.8315, "step": 26834 }, { "epoch": 0.08, "learning_rate": 5.937288702877309e-07, "loss": 0.8723, "step": 26835 }, { "epoch": 0.08, "learning_rate": 5.933630471442286e-07, "loss": 0.8321, "step": 26836 }, { "epoch": 0.08, "learning_rate": 5.929973332896677e-07, "loss": 0.8163, "step": 26837 }, { "epoch": 0.08, "learning_rate": 5.926317287282923e-07, "loss": 0.877, "step": 26838 }, { "epoch": 0.08, "learning_rate": 5.922662334643525e-07, "loss": 0.8195, "step": 26839 }, { "epoch": 0.08, "learning_rate": 5.919008475020937e-07, "loss": 0.7905, "step": 26840 }, { "epoch": 0.08, "learning_rate": 5.915355708457615e-07, "loss": 0.8688, "step": 26841 }, { "epoch": 0.08, "learning_rate": 5.91170403499599e-07, "loss": 0.8087, "step": 26842 }, { "epoch": 0.08, "learning_rate": 5.908053454678508e-07, "loss": 0.8393, "step": 26843 }, { "epoch": 0.08, "learning_rate": 5.904403967547567e-07, "loss": 0.8726, "step": 26844 }, { "epoch": 0.08, "learning_rate": 5.900755573645567e-07, "loss": 0.7742, "step": 26845 }, { "epoch": 0.08, "learning_rate": 5.897108273014896e-07, "loss": 0.817, "step": 26846 }, { "epoch": 0.08, "learning_rate": 5.89346206569793e-07, "loss": 0.8644, "step": 26847 }, { "epoch": 0.08, "learning_rate": 5.889816951737038e-07, "loss": 0.8831, "step": 26848 }, { "epoch": 0.08, "learning_rate": 5.886172931174572e-07, "loss": 0.8001, "step": 26849 }, { "epoch": 0.08, "learning_rate": 5.882530004052856e-07, "loss": 0.844, "step": 26850 }, { "epoch": 0.08, "learning_rate": 5.878888170414232e-07, "loss": 0.874, "step": 26851 }, { "epoch": 0.08, "learning_rate": 5.87524743030099e-07, "loss": 0.8528, "step": 26852 }, { "epoch": 0.08, "learning_rate": 5.87160778375544e-07, "loss": 0.8132, "step": 26853 }, { "epoch": 0.08, "learning_rate": 5.86796923081987e-07, "loss": 0.8598, "step": 26854 }, { "epoch": 0.08, "learning_rate": 5.864331771536558e-07, "loss": 0.8642, "step": 26855 }, { "epoch": 0.08, "learning_rate": 5.860695405947759e-07, "loss": 0.843, "step": 26856 }, { "epoch": 0.08, "learning_rate": 5.857060134095715e-07, "loss": 0.8267, "step": 26857 }, { "epoch": 0.08, "learning_rate": 5.853425956022684e-07, "loss": 0.816, "step": 26858 }, { "epoch": 0.08, "learning_rate": 5.849792871770865e-07, "loss": 0.8396, "step": 26859 }, { "epoch": 0.08, "learning_rate": 5.846160881382456e-07, "loss": 0.8652, "step": 26860 }, { "epoch": 0.08, "learning_rate": 5.842529984899692e-07, "loss": 0.8342, "step": 26861 }, { "epoch": 0.08, "learning_rate": 5.83890018236476e-07, "loss": 0.8144, "step": 26862 }, { "epoch": 0.08, "learning_rate": 5.835271473819793e-07, "loss": 0.8476, "step": 26863 }, { "epoch": 0.08, "learning_rate": 5.83164385930698e-07, "loss": 0.8325, "step": 26864 }, { "epoch": 0.08, "learning_rate": 5.828017338868463e-07, "loss": 0.872, "step": 26865 }, { "epoch": 0.08, "learning_rate": 5.824391912546345e-07, "loss": 0.8786, "step": 26866 }, { "epoch": 0.08, "learning_rate": 5.820767580382791e-07, "loss": 0.8495, "step": 26867 }, { "epoch": 0.08, "learning_rate": 5.817144342419889e-07, "loss": 0.7926, "step": 26868 }, { "epoch": 0.08, "learning_rate": 5.81352219869975e-07, "loss": 0.8397, "step": 26869 }, { "epoch": 0.08, "learning_rate": 5.80990114926444e-07, "loss": 0.8671, "step": 26870 }, { "epoch": 0.08, "learning_rate": 5.806281194156027e-07, "loss": 0.8394, "step": 26871 }, { "epoch": 0.08, "learning_rate": 5.802662333416575e-07, "loss": 0.8736, "step": 26872 }, { "epoch": 0.08, "learning_rate": 5.79904456708813e-07, "loss": 0.8475, "step": 26873 }, { "epoch": 0.08, "learning_rate": 5.795427895212725e-07, "loss": 0.8084, "step": 26874 }, { "epoch": 0.08, "learning_rate": 5.791812317832379e-07, "loss": 0.8197, "step": 26875 }, { "epoch": 0.08, "learning_rate": 5.788197834989106e-07, "loss": 0.8798, "step": 26876 }, { "epoch": 0.08, "learning_rate": 5.784584446724883e-07, "loss": 0.8305, "step": 26877 }, { "epoch": 0.08, "learning_rate": 5.780972153081698e-07, "loss": 0.8223, "step": 26878 }, { "epoch": 0.08, "learning_rate": 5.77736095410153e-07, "loss": 0.847, "step": 26879 }, { "epoch": 0.08, "learning_rate": 5.77375084982632e-07, "loss": 0.8456, "step": 26880 }, { "epoch": 0.08, "learning_rate": 5.770141840298027e-07, "loss": 0.8571, "step": 26881 }, { "epoch": 0.08, "learning_rate": 5.76653392555856e-07, "loss": 0.8308, "step": 26882 }, { "epoch": 0.08, "learning_rate": 5.762927105649874e-07, "loss": 0.8396, "step": 26883 }, { "epoch": 0.08, "learning_rate": 5.759321380613814e-07, "loss": 0.862, "step": 26884 }, { "epoch": 0.08, "learning_rate": 5.755716750492324e-07, "loss": 0.8376, "step": 26885 }, { "epoch": 0.08, "learning_rate": 5.752113215327271e-07, "loss": 0.8572, "step": 26886 }, { "epoch": 0.08, "learning_rate": 5.748510775160532e-07, "loss": 0.8409, "step": 26887 }, { "epoch": 0.08, "learning_rate": 5.74490943003394e-07, "loss": 0.8373, "step": 26888 }, { "epoch": 0.08, "learning_rate": 5.741309179989329e-07, "loss": 0.8486, "step": 26889 }, { "epoch": 0.08, "learning_rate": 5.737710025068555e-07, "loss": 0.8342, "step": 26890 }, { "epoch": 0.08, "learning_rate": 5.734111965313416e-07, "loss": 0.8934, "step": 26891 }, { "epoch": 0.08, "learning_rate": 5.730515000765724e-07, "loss": 0.8449, "step": 26892 }, { "epoch": 0.08, "learning_rate": 5.726919131467268e-07, "loss": 0.8788, "step": 26893 }, { "epoch": 0.08, "learning_rate": 5.723324357459836e-07, "loss": 0.8558, "step": 26894 }, { "epoch": 0.08, "learning_rate": 5.719730678785162e-07, "loss": 0.8649, "step": 26895 }, { "epoch": 0.08, "learning_rate": 5.716138095485024e-07, "loss": 0.8273, "step": 26896 }, { "epoch": 0.08, "learning_rate": 5.712546607601155e-07, "loss": 0.8727, "step": 26897 }, { "epoch": 0.08, "learning_rate": 5.708956215175276e-07, "loss": 0.8347, "step": 26898 }, { "epoch": 0.08, "learning_rate": 5.705366918249122e-07, "loss": 0.8294, "step": 26899 }, { "epoch": 0.08, "learning_rate": 5.70177871686437e-07, "loss": 0.8445, "step": 26900 }, { "epoch": 0.08, "learning_rate": 5.698191611062731e-07, "loss": 0.8088, "step": 26901 }, { "epoch": 0.08, "learning_rate": 5.694605600885861e-07, "loss": 0.8478, "step": 26902 }, { "epoch": 0.08, "learning_rate": 5.691020686375414e-07, "loss": 0.8749, "step": 26903 }, { "epoch": 0.08, "learning_rate": 5.687436867573081e-07, "loss": 0.8465, "step": 26904 }, { "epoch": 0.08, "learning_rate": 5.683854144520474e-07, "loss": 0.8442, "step": 26905 }, { "epoch": 0.08, "learning_rate": 5.680272517259211e-07, "loss": 0.8688, "step": 26906 }, { "epoch": 0.08, "learning_rate": 5.676691985830918e-07, "loss": 0.8806, "step": 26907 }, { "epoch": 0.08, "learning_rate": 5.673112550277194e-07, "loss": 0.8138, "step": 26908 }, { "epoch": 0.08, "learning_rate": 5.669534210639605e-07, "loss": 0.85, "step": 26909 }, { "epoch": 0.08, "learning_rate": 5.665956966959752e-07, "loss": 0.828, "step": 26910 }, { "epoch": 0.08, "learning_rate": 5.662380819279211e-07, "loss": 0.8713, "step": 26911 }, { "epoch": 0.08, "learning_rate": 5.658805767639486e-07, "loss": 0.8253, "step": 26912 }, { "epoch": 0.08, "learning_rate": 5.655231812082129e-07, "loss": 0.841, "step": 26913 }, { "epoch": 0.08, "learning_rate": 5.651658952648664e-07, "loss": 0.8012, "step": 26914 }, { "epoch": 0.08, "learning_rate": 5.648087189380613e-07, "loss": 0.8895, "step": 26915 }, { "epoch": 0.08, "learning_rate": 5.644516522319465e-07, "loss": 0.829, "step": 26916 }, { "epoch": 0.08, "learning_rate": 5.640946951506709e-07, "loss": 0.7747, "step": 26917 }, { "epoch": 0.08, "learning_rate": 5.637378476983823e-07, "loss": 0.8678, "step": 26918 }, { "epoch": 0.08, "learning_rate": 5.633811098792252e-07, "loss": 0.8412, "step": 26919 }, { "epoch": 0.08, "learning_rate": 5.63024481697344e-07, "loss": 0.8475, "step": 26920 }, { "epoch": 0.08, "learning_rate": 5.626679631568832e-07, "loss": 0.8394, "step": 26921 }, { "epoch": 0.08, "learning_rate": 5.62311554261985e-07, "loss": 0.8421, "step": 26922 }, { "epoch": 0.08, "learning_rate": 5.619552550167894e-07, "loss": 0.8306, "step": 26923 }, { "epoch": 0.08, "learning_rate": 5.615990654254366e-07, "loss": 0.8353, "step": 26924 }, { "epoch": 0.08, "learning_rate": 5.612429854920664e-07, "loss": 0.8315, "step": 26925 }, { "epoch": 0.08, "learning_rate": 5.608870152208135e-07, "loss": 0.8381, "step": 26926 }, { "epoch": 0.08, "learning_rate": 5.605311546158132e-07, "loss": 0.85, "step": 26927 }, { "epoch": 0.08, "learning_rate": 5.601754036812012e-07, "loss": 0.8446, "step": 26928 }, { "epoch": 0.08, "learning_rate": 5.598197624211121e-07, "loss": 0.8223, "step": 26929 }, { "epoch": 0.08, "learning_rate": 5.594642308396758e-07, "loss": 0.8462, "step": 26930 }, { "epoch": 0.08, "learning_rate": 5.591088089410246e-07, "loss": 0.8583, "step": 26931 }, { "epoch": 0.08, "learning_rate": 5.587534967292851e-07, "loss": 0.8571, "step": 26932 }, { "epoch": 0.08, "learning_rate": 5.583982942085886e-07, "loss": 0.8396, "step": 26933 }, { "epoch": 0.08, "learning_rate": 5.580432013830606e-07, "loss": 0.887, "step": 26934 }, { "epoch": 0.08, "learning_rate": 5.576882182568255e-07, "loss": 0.8537, "step": 26935 }, { "epoch": 0.08, "learning_rate": 5.573333448340113e-07, "loss": 0.8326, "step": 26936 }, { "epoch": 0.08, "learning_rate": 5.569785811187367e-07, "loss": 0.8311, "step": 26937 }, { "epoch": 0.08, "learning_rate": 5.566239271151253e-07, "loss": 0.844, "step": 26938 }, { "epoch": 0.08, "learning_rate": 5.562693828272969e-07, "loss": 0.8215, "step": 26939 }, { "epoch": 0.08, "learning_rate": 5.559149482593717e-07, "loss": 0.8655, "step": 26940 }, { "epoch": 0.08, "learning_rate": 5.555606234154676e-07, "loss": 0.8361, "step": 26941 }, { "epoch": 0.08, "learning_rate": 5.552064082997011e-07, "loss": 0.8919, "step": 26942 }, { "epoch": 0.08, "learning_rate": 5.548523029161878e-07, "loss": 0.8745, "step": 26943 }, { "epoch": 0.08, "learning_rate": 5.544983072690402e-07, "loss": 0.844, "step": 26944 }, { "epoch": 0.08, "learning_rate": 5.541444213623726e-07, "loss": 0.8894, "step": 26945 }, { "epoch": 0.08, "learning_rate": 5.537906452002951e-07, "loss": 0.8482, "step": 26946 }, { "epoch": 0.08, "learning_rate": 5.534369787869209e-07, "loss": 0.8559, "step": 26947 }, { "epoch": 0.08, "learning_rate": 5.530834221263559e-07, "loss": 0.8616, "step": 26948 }, { "epoch": 0.08, "learning_rate": 5.5272997522271e-07, "loss": 0.8591, "step": 26949 }, { "epoch": 0.08, "learning_rate": 5.523766380800888e-07, "loss": 0.8488, "step": 26950 }, { "epoch": 0.08, "learning_rate": 5.520234107025968e-07, "loss": 0.8291, "step": 26951 }, { "epoch": 0.08, "learning_rate": 5.516702930943362e-07, "loss": 0.824, "step": 26952 }, { "epoch": 0.08, "learning_rate": 5.513172852594151e-07, "loss": 0.8052, "step": 26953 }, { "epoch": 0.08, "learning_rate": 5.50964387201931e-07, "loss": 0.8602, "step": 26954 }, { "epoch": 0.08, "learning_rate": 5.506115989259842e-07, "loss": 0.8444, "step": 26955 }, { "epoch": 0.08, "learning_rate": 5.502589204356746e-07, "loss": 0.8303, "step": 26956 }, { "epoch": 0.08, "learning_rate": 5.499063517350989e-07, "loss": 0.8409, "step": 26957 }, { "epoch": 0.08, "learning_rate": 5.495538928283517e-07, "loss": 0.8902, "step": 26958 }, { "epoch": 0.08, "learning_rate": 5.492015437195308e-07, "loss": 0.8596, "step": 26959 }, { "epoch": 0.08, "learning_rate": 5.488493044127297e-07, "loss": 0.8393, "step": 26960 }, { "epoch": 0.08, "learning_rate": 5.484971749120404e-07, "loss": 0.8644, "step": 26961 }, { "epoch": 0.08, "learning_rate": 5.481451552215522e-07, "loss": 0.8722, "step": 26962 }, { "epoch": 0.08, "learning_rate": 5.47793245345356e-07, "loss": 0.8566, "step": 26963 }, { "epoch": 0.08, "learning_rate": 5.474414452875409e-07, "loss": 0.9023, "step": 26964 }, { "epoch": 0.08, "learning_rate": 5.470897550521947e-07, "loss": 0.8644, "step": 26965 }, { "epoch": 0.08, "learning_rate": 5.46738174643402e-07, "loss": 0.8314, "step": 26966 }, { "epoch": 0.08, "learning_rate": 5.463867040652493e-07, "loss": 0.8391, "step": 26967 }, { "epoch": 0.08, "learning_rate": 5.460353433218201e-07, "loss": 0.8445, "step": 26968 }, { "epoch": 0.08, "learning_rate": 5.456840924171936e-07, "loss": 0.7785, "step": 26969 }, { "epoch": 0.08, "learning_rate": 5.453329513554528e-07, "loss": 0.8244, "step": 26970 }, { "epoch": 0.08, "learning_rate": 5.449819201406781e-07, "loss": 0.8346, "step": 26971 }, { "epoch": 0.08, "learning_rate": 5.446309987769472e-07, "loss": 0.8343, "step": 26972 }, { "epoch": 0.08, "learning_rate": 5.442801872683356e-07, "loss": 0.8071, "step": 26973 }, { "epoch": 0.08, "learning_rate": 5.439294856189225e-07, "loss": 0.8368, "step": 26974 }, { "epoch": 0.08, "learning_rate": 5.435788938327802e-07, "loss": 0.8606, "step": 26975 }, { "epoch": 0.08, "learning_rate": 5.432284119139808e-07, "loss": 0.8503, "step": 26976 }, { "epoch": 0.08, "learning_rate": 5.42878039866599e-07, "loss": 0.8445, "step": 26977 }, { "epoch": 0.08, "learning_rate": 5.425277776947036e-07, "loss": 0.8425, "step": 26978 }, { "epoch": 0.08, "learning_rate": 5.421776254023669e-07, "loss": 0.8903, "step": 26979 }, { "epoch": 0.08, "learning_rate": 5.418275829936537e-07, "loss": 0.8029, "step": 26980 }, { "epoch": 0.08, "learning_rate": 5.414776504726315e-07, "loss": 0.7978, "step": 26981 }, { "epoch": 0.08, "learning_rate": 5.411278278433673e-07, "loss": 0.8556, "step": 26982 }, { "epoch": 0.08, "learning_rate": 5.407781151099245e-07, "loss": 0.8604, "step": 26983 }, { "epoch": 0.08, "learning_rate": 5.404285122763663e-07, "loss": 0.8459, "step": 26984 }, { "epoch": 0.08, "learning_rate": 5.400790193467543e-07, "loss": 0.8622, "step": 26985 }, { "epoch": 0.08, "learning_rate": 5.397296363251514e-07, "loss": 0.8424, "step": 26986 }, { "epoch": 0.08, "learning_rate": 5.393803632156124e-07, "loss": 0.8419, "step": 26987 }, { "epoch": 0.08, "learning_rate": 5.390312000221976e-07, "loss": 0.8284, "step": 26988 }, { "epoch": 0.08, "learning_rate": 5.386821467489645e-07, "loss": 0.8231, "step": 26989 }, { "epoch": 0.08, "learning_rate": 5.383332033999667e-07, "loss": 0.8172, "step": 26990 }, { "epoch": 0.08, "learning_rate": 5.379843699792597e-07, "loss": 0.849, "step": 26991 }, { "epoch": 0.08, "learning_rate": 5.37635646490896e-07, "loss": 0.82, "step": 26992 }, { "epoch": 0.08, "learning_rate": 5.37287032938928e-07, "loss": 0.8239, "step": 26993 }, { "epoch": 0.08, "learning_rate": 5.369385293274032e-07, "loss": 0.8521, "step": 26994 }, { "epoch": 0.08, "learning_rate": 5.36590135660372e-07, "loss": 0.8189, "step": 26995 }, { "epoch": 0.08, "learning_rate": 5.362418519418855e-07, "loss": 0.8328, "step": 26996 }, { "epoch": 0.08, "learning_rate": 5.358936781759849e-07, "loss": 0.8018, "step": 26997 }, { "epoch": 0.08, "learning_rate": 5.355456143667181e-07, "loss": 0.8284, "step": 26998 }, { "epoch": 0.08, "learning_rate": 5.351976605181286e-07, "loss": 0.8336, "step": 26999 }, { "epoch": 0.08, "learning_rate": 5.348498166342608e-07, "loss": 0.8447, "step": 27000 }, { "epoch": 0.08, "learning_rate": 5.345020827191516e-07, "loss": 0.8293, "step": 27001 }, { "epoch": 0.08, "learning_rate": 5.341544587768444e-07, "loss": 0.8415, "step": 27002 }, { "epoch": 0.08, "learning_rate": 5.338069448113792e-07, "loss": 0.8458, "step": 27003 }, { "epoch": 0.08, "learning_rate": 5.334595408267906e-07, "loss": 0.8866, "step": 27004 }, { "epoch": 0.08, "learning_rate": 5.331122468271155e-07, "loss": 0.8118, "step": 27005 }, { "epoch": 0.08, "learning_rate": 5.327650628163894e-07, "loss": 0.8001, "step": 27006 }, { "epoch": 0.08, "learning_rate": 5.324179887986458e-07, "loss": 0.8432, "step": 27007 }, { "epoch": 0.08, "learning_rate": 5.32071024777917e-07, "loss": 0.8769, "step": 27008 }, { "epoch": 0.08, "learning_rate": 5.317241707582355e-07, "loss": 0.8565, "step": 27009 }, { "epoch": 0.08, "learning_rate": 5.3137742674363e-07, "loss": 0.8426, "step": 27010 }, { "epoch": 0.08, "learning_rate": 5.310307927381276e-07, "loss": 0.8374, "step": 27011 }, { "epoch": 0.08, "learning_rate": 5.30684268745757e-07, "loss": 0.8607, "step": 27012 }, { "epoch": 0.08, "learning_rate": 5.303378547705452e-07, "loss": 0.8425, "step": 27013 }, { "epoch": 0.08, "learning_rate": 5.299915508165143e-07, "loss": 0.8253, "step": 27014 }, { "epoch": 0.08, "learning_rate": 5.296453568876903e-07, "loss": 0.8389, "step": 27015 }, { "epoch": 0.08, "learning_rate": 5.292992729880952e-07, "loss": 0.8639, "step": 27016 }, { "epoch": 0.08, "learning_rate": 5.289532991217494e-07, "loss": 0.8418, "step": 27017 }, { "epoch": 0.08, "learning_rate": 5.286074352926718e-07, "loss": 0.8288, "step": 27018 }, { "epoch": 0.08, "learning_rate": 5.282616815048791e-07, "loss": 0.8415, "step": 27019 }, { "epoch": 0.08, "learning_rate": 5.279160377623915e-07, "loss": 0.8632, "step": 27020 }, { "epoch": 0.08, "learning_rate": 5.275705040692259e-07, "loss": 0.8668, "step": 27021 }, { "epoch": 0.08, "learning_rate": 5.272250804293932e-07, "loss": 0.8615, "step": 27022 }, { "epoch": 0.08, "learning_rate": 5.268797668469083e-07, "loss": 0.8336, "step": 27023 }, { "epoch": 0.08, "learning_rate": 5.265345633257835e-07, "loss": 0.8407, "step": 27024 }, { "epoch": 0.08, "learning_rate": 5.261894698700276e-07, "loss": 0.8823, "step": 27025 }, { "epoch": 0.08, "learning_rate": 5.25844486483652e-07, "loss": 0.8198, "step": 27026 }, { "epoch": 0.08, "learning_rate": 5.254996131706635e-07, "loss": 0.8793, "step": 27027 }, { "epoch": 0.08, "learning_rate": 5.251548499350723e-07, "loss": 0.8398, "step": 27028 }, { "epoch": 0.08, "learning_rate": 5.248101967808794e-07, "loss": 0.8255, "step": 27029 }, { "epoch": 0.08, "learning_rate": 5.244656537120907e-07, "loss": 0.8328, "step": 27030 }, { "epoch": 0.08, "learning_rate": 5.241212207327085e-07, "loss": 0.8795, "step": 27031 }, { "epoch": 0.08, "learning_rate": 5.237768978467362e-07, "loss": 0.818, "step": 27032 }, { "epoch": 0.08, "learning_rate": 5.234326850581739e-07, "loss": 0.8615, "step": 27033 }, { "epoch": 0.08, "learning_rate": 5.230885823710197e-07, "loss": 0.834, "step": 27034 }, { "epoch": 0.08, "learning_rate": 5.227445897892735e-07, "loss": 0.9086, "step": 27035 }, { "epoch": 0.08, "learning_rate": 5.224007073169302e-07, "loss": 0.8582, "step": 27036 }, { "epoch": 0.08, "learning_rate": 5.22056934957984e-07, "loss": 0.8582, "step": 27037 }, { "epoch": 0.08, "learning_rate": 5.217132727164298e-07, "loss": 0.8205, "step": 27038 }, { "epoch": 0.08, "learning_rate": 5.213697205962631e-07, "loss": 0.8255, "step": 27039 }, { "epoch": 0.08, "learning_rate": 5.21026278601472e-07, "loss": 0.8817, "step": 27040 }, { "epoch": 0.08, "learning_rate": 5.206829467360486e-07, "loss": 0.8405, "step": 27041 }, { "epoch": 0.08, "learning_rate": 5.203397250039821e-07, "loss": 0.8219, "step": 27042 }, { "epoch": 0.08, "learning_rate": 5.199966134092571e-07, "loss": 0.8885, "step": 27043 }, { "epoch": 0.08, "learning_rate": 5.196536119558615e-07, "loss": 0.8082, "step": 27044 }, { "epoch": 0.08, "learning_rate": 5.193107206477821e-07, "loss": 0.851, "step": 27045 }, { "epoch": 0.08, "learning_rate": 5.189679394890034e-07, "loss": 0.8045, "step": 27046 }, { "epoch": 0.08, "learning_rate": 5.186252684835047e-07, "loss": 0.8596, "step": 27047 }, { "epoch": 0.08, "learning_rate": 5.182827076352681e-07, "loss": 0.8347, "step": 27048 }, { "epoch": 0.08, "learning_rate": 5.179402569482739e-07, "loss": 0.8473, "step": 27049 }, { "epoch": 0.08, "learning_rate": 5.175979164265021e-07, "loss": 0.8483, "step": 27050 }, { "epoch": 0.08, "learning_rate": 5.172556860739275e-07, "loss": 0.8662, "step": 27051 }, { "epoch": 0.08, "learning_rate": 5.16913565894529e-07, "loss": 0.8594, "step": 27052 }, { "epoch": 0.08, "learning_rate": 5.165715558922812e-07, "loss": 0.8429, "step": 27053 }, { "epoch": 0.08, "learning_rate": 5.162296560711544e-07, "loss": 0.8683, "step": 27054 }, { "epoch": 0.09, "learning_rate": 5.158878664351241e-07, "loss": 0.8416, "step": 27055 }, { "epoch": 0.09, "learning_rate": 5.155461869881606e-07, "loss": 0.8068, "step": 27056 }, { "epoch": 0.09, "learning_rate": 5.152046177342329e-07, "loss": 0.8232, "step": 27057 }, { "epoch": 0.09, "learning_rate": 5.148631586773089e-07, "loss": 0.8389, "step": 27058 }, { "epoch": 0.09, "learning_rate": 5.145218098213578e-07, "loss": 0.8342, "step": 27059 }, { "epoch": 0.09, "learning_rate": 5.141805711703463e-07, "loss": 0.8419, "step": 27060 }, { "epoch": 0.09, "learning_rate": 5.138394427282356e-07, "loss": 0.8626, "step": 27061 }, { "epoch": 0.09, "learning_rate": 5.134984244989904e-07, "loss": 0.8611, "step": 27062 }, { "epoch": 0.09, "learning_rate": 5.131575164865721e-07, "loss": 0.8632, "step": 27063 }, { "epoch": 0.09, "learning_rate": 5.128167186949428e-07, "loss": 0.8554, "step": 27064 }, { "epoch": 0.09, "learning_rate": 5.124760311280619e-07, "loss": 0.8571, "step": 27065 }, { "epoch": 0.09, "learning_rate": 5.12135453789887e-07, "loss": 0.8217, "step": 27066 }, { "epoch": 0.09, "learning_rate": 5.117949866843763e-07, "loss": 0.8259, "step": 27067 }, { "epoch": 0.09, "learning_rate": 5.114546298154821e-07, "loss": 0.8336, "step": 27068 }, { "epoch": 0.09, "learning_rate": 5.111143831871624e-07, "loss": 0.8127, "step": 27069 }, { "epoch": 0.09, "learning_rate": 5.107742468033683e-07, "loss": 0.8259, "step": 27070 }, { "epoch": 0.09, "learning_rate": 5.104342206680546e-07, "loss": 0.819, "step": 27071 }, { "epoch": 0.09, "learning_rate": 5.10094304785167e-07, "loss": 0.8108, "step": 27072 }, { "epoch": 0.09, "learning_rate": 5.097544991586578e-07, "loss": 0.8101, "step": 27073 }, { "epoch": 0.09, "learning_rate": 5.09414803792474e-07, "loss": 0.8438, "step": 27074 }, { "epoch": 0.09, "learning_rate": 5.090752186905623e-07, "loss": 0.7721, "step": 27075 }, { "epoch": 0.09, "learning_rate": 5.087357438568696e-07, "loss": 0.834, "step": 27076 }, { "epoch": 0.09, "learning_rate": 5.083963792953373e-07, "loss": 0.8402, "step": 27077 }, { "epoch": 0.09, "learning_rate": 5.08057125009912e-07, "loss": 0.848, "step": 27078 }, { "epoch": 0.09, "learning_rate": 5.077179810045319e-07, "loss": 0.8381, "step": 27079 }, { "epoch": 0.09, "learning_rate": 5.073789472831369e-07, "loss": 0.8398, "step": 27080 }, { "epoch": 0.09, "learning_rate": 5.070400238496687e-07, "loss": 0.8197, "step": 27081 }, { "epoch": 0.09, "learning_rate": 5.067012107080637e-07, "loss": 0.8439, "step": 27082 }, { "epoch": 0.09, "learning_rate": 5.06362507862258e-07, "loss": 0.8629, "step": 27083 }, { "epoch": 0.09, "learning_rate": 5.060239153161872e-07, "loss": 0.8291, "step": 27084 }, { "epoch": 0.09, "learning_rate": 5.05685433073786e-07, "loss": 0.8662, "step": 27085 }, { "epoch": 0.09, "learning_rate": 5.053470611389855e-07, "loss": 0.8695, "step": 27086 }, { "epoch": 0.09, "learning_rate": 5.05008799515716e-07, "loss": 0.8546, "step": 27087 }, { "epoch": 0.09, "learning_rate": 5.046706482079111e-07, "loss": 0.8101, "step": 27088 }, { "epoch": 0.09, "learning_rate": 5.043326072194965e-07, "loss": 0.8861, "step": 27089 }, { "epoch": 0.09, "learning_rate": 5.039946765544013e-07, "loss": 0.816, "step": 27090 }, { "epoch": 0.09, "learning_rate": 5.0365685621655e-07, "loss": 0.8595, "step": 27091 }, { "epoch": 0.09, "learning_rate": 5.033191462098697e-07, "loss": 0.8465, "step": 27092 }, { "epoch": 0.09, "learning_rate": 5.029815465382825e-07, "loss": 0.8408, "step": 27093 }, { "epoch": 0.09, "learning_rate": 5.026440572057101e-07, "loss": 0.8261, "step": 27094 }, { "epoch": 0.09, "learning_rate": 5.023066782160769e-07, "loss": 0.8682, "step": 27095 }, { "epoch": 0.09, "learning_rate": 5.019694095732985e-07, "loss": 0.8422, "step": 27096 }, { "epoch": 0.09, "learning_rate": 5.016322512812955e-07, "loss": 0.8502, "step": 27097 }, { "epoch": 0.09, "learning_rate": 5.012952033439844e-07, "loss": 0.8347, "step": 27098 }, { "epoch": 0.09, "learning_rate": 5.009582657652823e-07, "loss": 0.8402, "step": 27099 }, { "epoch": 0.09, "learning_rate": 5.006214385491015e-07, "loss": 0.8223, "step": 27100 }, { "epoch": 0.09, "learning_rate": 5.002847216993578e-07, "loss": 0.8723, "step": 27101 }, { "epoch": 0.09, "learning_rate": 4.999481152199637e-07, "loss": 0.8437, "step": 27102 }, { "epoch": 0.09, "learning_rate": 4.99611619114827e-07, "loss": 0.8517, "step": 27103 }, { "epoch": 0.09, "learning_rate": 4.992752333878592e-07, "loss": 0.8392, "step": 27104 }, { "epoch": 0.09, "learning_rate": 4.989389580429683e-07, "loss": 0.8136, "step": 27105 }, { "epoch": 0.09, "learning_rate": 4.986027930840609e-07, "loss": 0.8895, "step": 27106 }, { "epoch": 0.09, "learning_rate": 4.982667385150419e-07, "loss": 0.8329, "step": 27107 }, { "epoch": 0.09, "learning_rate": 4.979307943398182e-07, "loss": 0.8129, "step": 27108 }, { "epoch": 0.09, "learning_rate": 4.975949605622921e-07, "loss": 0.8618, "step": 27109 }, { "epoch": 0.09, "learning_rate": 4.972592371863627e-07, "loss": 0.8605, "step": 27110 }, { "epoch": 0.09, "learning_rate": 4.969236242159315e-07, "loss": 0.9054, "step": 27111 }, { "epoch": 0.09, "learning_rate": 4.965881216548996e-07, "loss": 0.8735, "step": 27112 }, { "epoch": 0.09, "learning_rate": 4.962527295071662e-07, "loss": 0.8232, "step": 27113 }, { "epoch": 0.09, "learning_rate": 4.959174477766237e-07, "loss": 0.8213, "step": 27114 }, { "epoch": 0.09, "learning_rate": 4.955822764671703e-07, "loss": 0.8458, "step": 27115 }, { "epoch": 0.09, "learning_rate": 4.952472155826993e-07, "loss": 0.8371, "step": 27116 }, { "epoch": 0.09, "learning_rate": 4.949122651271033e-07, "loss": 0.8265, "step": 27117 }, { "epoch": 0.09, "learning_rate": 4.945774251042746e-07, "loss": 0.8075, "step": 27118 }, { "epoch": 0.09, "learning_rate": 4.942426955181034e-07, "loss": 0.8742, "step": 27119 }, { "epoch": 0.09, "learning_rate": 4.93908076372479e-07, "loss": 0.8276, "step": 27120 }, { "epoch": 0.09, "learning_rate": 4.935735676712883e-07, "loss": 0.844, "step": 27121 }, { "epoch": 0.09, "learning_rate": 4.932391694184169e-07, "loss": 0.8355, "step": 27122 }, { "epoch": 0.09, "learning_rate": 4.929048816177517e-07, "loss": 0.8858, "step": 27123 }, { "epoch": 0.09, "learning_rate": 4.925707042731742e-07, "loss": 0.8601, "step": 27124 }, { "epoch": 0.09, "learning_rate": 4.922366373885701e-07, "loss": 0.8358, "step": 27125 }, { "epoch": 0.09, "learning_rate": 4.919026809678185e-07, "loss": 0.8593, "step": 27126 }, { "epoch": 0.09, "learning_rate": 4.915688350148018e-07, "loss": 0.8357, "step": 27127 }, { "epoch": 0.09, "learning_rate": 4.912350995333948e-07, "loss": 0.8491, "step": 27128 }, { "epoch": 0.09, "learning_rate": 4.909014745274776e-07, "loss": 0.8132, "step": 27129 }, { "epoch": 0.09, "learning_rate": 4.90567960000925e-07, "loss": 0.8405, "step": 27130 }, { "epoch": 0.09, "learning_rate": 4.90234555957615e-07, "loss": 0.8317, "step": 27131 }, { "epoch": 0.09, "learning_rate": 4.899012624014177e-07, "loss": 0.8481, "step": 27132 }, { "epoch": 0.09, "learning_rate": 4.895680793362057e-07, "loss": 0.8413, "step": 27133 }, { "epoch": 0.09, "learning_rate": 4.892350067658524e-07, "loss": 0.8743, "step": 27134 }, { "epoch": 0.09, "learning_rate": 4.88902044694225e-07, "loss": 0.8295, "step": 27135 }, { "epoch": 0.09, "learning_rate": 4.885691931251923e-07, "loss": 0.83, "step": 27136 }, { "epoch": 0.09, "learning_rate": 4.882364520626226e-07, "loss": 0.8526, "step": 27137 }, { "epoch": 0.09, "learning_rate": 4.879038215103826e-07, "loss": 0.8443, "step": 27138 }, { "epoch": 0.09, "learning_rate": 4.875713014723349e-07, "loss": 0.855, "step": 27139 }, { "epoch": 0.09, "learning_rate": 4.87238891952343e-07, "loss": 0.8317, "step": 27140 }, { "epoch": 0.09, "learning_rate": 4.869065929542694e-07, "loss": 0.8538, "step": 27141 }, { "epoch": 0.09, "learning_rate": 4.865744044819753e-07, "loss": 0.8228, "step": 27142 }, { "epoch": 0.09, "learning_rate": 4.86242326539319e-07, "loss": 0.8908, "step": 27143 }, { "epoch": 0.09, "learning_rate": 4.859103591301606e-07, "loss": 0.8497, "step": 27144 }, { "epoch": 0.09, "learning_rate": 4.855785022583559e-07, "loss": 0.8319, "step": 27145 }, { "epoch": 0.09, "learning_rate": 4.852467559277596e-07, "loss": 0.8438, "step": 27146 }, { "epoch": 0.09, "learning_rate": 4.849151201422264e-07, "loss": 0.8908, "step": 27147 }, { "epoch": 0.09, "learning_rate": 4.84583594905611e-07, "loss": 0.8228, "step": 27148 }, { "epoch": 0.09, "learning_rate": 4.842521802217626e-07, "loss": 0.84, "step": 27149 }, { "epoch": 0.09, "learning_rate": 4.839208760945335e-07, "loss": 0.8619, "step": 27150 }, { "epoch": 0.09, "learning_rate": 4.835896825277719e-07, "loss": 0.8734, "step": 27151 }, { "epoch": 0.09, "learning_rate": 4.832585995253281e-07, "loss": 0.8485, "step": 27152 }, { "epoch": 0.09, "learning_rate": 4.829276270910455e-07, "loss": 0.8518, "step": 27153 }, { "epoch": 0.09, "learning_rate": 4.825967652287688e-07, "loss": 0.846, "step": 27154 }, { "epoch": 0.09, "learning_rate": 4.822660139423474e-07, "loss": 0.8526, "step": 27155 }, { "epoch": 0.09, "learning_rate": 4.819353732356191e-07, "loss": 0.8633, "step": 27156 }, { "epoch": 0.09, "learning_rate": 4.816048431124265e-07, "loss": 0.8796, "step": 27157 }, { "epoch": 0.09, "learning_rate": 4.812744235766109e-07, "loss": 0.8128, "step": 27158 }, { "epoch": 0.09, "learning_rate": 4.809441146320104e-07, "loss": 0.853, "step": 27159 }, { "epoch": 0.09, "learning_rate": 4.806139162824619e-07, "loss": 0.8444, "step": 27160 }, { "epoch": 0.09, "learning_rate": 4.802838285318023e-07, "loss": 0.8632, "step": 27161 }, { "epoch": 0.09, "learning_rate": 4.799538513838675e-07, "loss": 0.836, "step": 27162 }, { "epoch": 0.09, "learning_rate": 4.796239848424922e-07, "loss": 0.8591, "step": 27163 }, { "epoch": 0.09, "learning_rate": 4.792942289115054e-07, "loss": 0.805, "step": 27164 }, { "epoch": 0.09, "learning_rate": 4.78964583594741e-07, "loss": 0.8355, "step": 27165 }, { "epoch": 0.09, "learning_rate": 4.786350488960278e-07, "loss": 0.8505, "step": 27166 }, { "epoch": 0.09, "learning_rate": 4.783056248191953e-07, "loss": 0.8484, "step": 27167 }, { "epoch": 0.09, "learning_rate": 4.779763113680702e-07, "loss": 0.8301, "step": 27168 }, { "epoch": 0.09, "learning_rate": 4.776471085464784e-07, "loss": 0.8587, "step": 27169 }, { "epoch": 0.09, "learning_rate": 4.773180163582458e-07, "loss": 0.8909, "step": 27170 }, { "epoch": 0.09, "learning_rate": 4.769890348071937e-07, "loss": 0.8239, "step": 27171 }, { "epoch": 0.09, "learning_rate": 4.7666016389714685e-07, "loss": 0.8623, "step": 27172 }, { "epoch": 0.09, "learning_rate": 4.763314036319244e-07, "loss": 0.8154, "step": 27173 }, { "epoch": 0.09, "learning_rate": 4.7600275401534555e-07, "loss": 0.8475, "step": 27174 }, { "epoch": 0.09, "learning_rate": 4.756742150512306e-07, "loss": 0.8193, "step": 27175 }, { "epoch": 0.09, "learning_rate": 4.753457867433953e-07, "loss": 0.8739, "step": 27176 }, { "epoch": 0.09, "learning_rate": 4.750174690956577e-07, "loss": 0.8349, "step": 27177 }, { "epoch": 0.09, "learning_rate": 4.746892621118282e-07, "loss": 0.8934, "step": 27178 }, { "epoch": 0.09, "learning_rate": 4.7436116579572145e-07, "loss": 0.847, "step": 27179 }, { "epoch": 0.09, "learning_rate": 4.740331801511522e-07, "loss": 0.8618, "step": 27180 }, { "epoch": 0.09, "learning_rate": 4.7370530518192736e-07, "loss": 0.8185, "step": 27181 }, { "epoch": 0.09, "learning_rate": 4.733775408918584e-07, "loss": 0.8307, "step": 27182 }, { "epoch": 0.09, "learning_rate": 4.7304988728475334e-07, "loss": 0.8718, "step": 27183 }, { "epoch": 0.09, "learning_rate": 4.7272234436441797e-07, "loss": 0.8739, "step": 27184 }, { "epoch": 0.09, "learning_rate": 4.7239491213465827e-07, "loss": 0.8417, "step": 27185 }, { "epoch": 0.09, "learning_rate": 4.720675905992778e-07, "loss": 0.8064, "step": 27186 }, { "epoch": 0.09, "learning_rate": 4.7174037976208233e-07, "loss": 0.8339, "step": 27187 }, { "epoch": 0.09, "learning_rate": 4.714132796268689e-07, "loss": 0.8462, "step": 27188 }, { "epoch": 0.09, "learning_rate": 4.710862901974411e-07, "loss": 0.8083, "step": 27189 }, { "epoch": 0.09, "learning_rate": 4.70759411477596e-07, "loss": 0.8517, "step": 27190 }, { "epoch": 0.09, "learning_rate": 4.704326434711337e-07, "loss": 0.8589, "step": 27191 }, { "epoch": 0.09, "learning_rate": 4.701059861818491e-07, "loss": 0.8188, "step": 27192 }, { "epoch": 0.09, "learning_rate": 4.6977943961353687e-07, "loss": 0.8119, "step": 27193 }, { "epoch": 0.09, "learning_rate": 4.6945300376999403e-07, "loss": 0.8315, "step": 27194 }, { "epoch": 0.09, "learning_rate": 4.691266786550097e-07, "loss": 0.8535, "step": 27195 }, { "epoch": 0.09, "learning_rate": 4.6880046427237536e-07, "loss": 0.8555, "step": 27196 }, { "epoch": 0.09, "learning_rate": 4.6847436062588127e-07, "loss": 0.8466, "step": 27197 }, { "epoch": 0.09, "learning_rate": 4.6814836771932105e-07, "loss": 0.813, "step": 27198 }, { "epoch": 0.09, "learning_rate": 4.67822485556475e-07, "loss": 0.8335, "step": 27199 }, { "epoch": 0.09, "learning_rate": 4.674967141411335e-07, "loss": 0.8559, "step": 27200 }, { "epoch": 0.09, "learning_rate": 4.6717105347708234e-07, "loss": 0.8122, "step": 27201 }, { "epoch": 0.09, "learning_rate": 4.668455035681008e-07, "loss": 0.8757, "step": 27202 }, { "epoch": 0.09, "learning_rate": 4.6652006441797237e-07, "loss": 0.8613, "step": 27203 }, { "epoch": 0.09, "learning_rate": 4.661947360304819e-07, "loss": 0.8841, "step": 27204 }, { "epoch": 0.09, "learning_rate": 4.658695184094075e-07, "loss": 0.8717, "step": 27205 }, { "epoch": 0.09, "learning_rate": 4.6554441155852504e-07, "loss": 0.8385, "step": 27206 }, { "epoch": 0.09, "learning_rate": 4.6521941548161364e-07, "loss": 0.8389, "step": 27207 }, { "epoch": 0.09, "learning_rate": 4.6489453018244925e-07, "loss": 0.8522, "step": 27208 }, { "epoch": 0.09, "learning_rate": 4.6456975566480545e-07, "loss": 0.8309, "step": 27209 }, { "epoch": 0.09, "learning_rate": 4.6424509193245704e-07, "loss": 0.8404, "step": 27210 }, { "epoch": 0.09, "learning_rate": 4.6392053898917543e-07, "loss": 0.8856, "step": 27211 }, { "epoch": 0.09, "learning_rate": 4.635960968387321e-07, "loss": 0.8132, "step": 27212 }, { "epoch": 0.09, "learning_rate": 4.6327176548489506e-07, "loss": 0.829, "step": 27213 }, { "epoch": 0.09, "learning_rate": 4.6294754493143246e-07, "loss": 0.8268, "step": 27214 }, { "epoch": 0.09, "learning_rate": 4.6262343518211236e-07, "loss": 0.849, "step": 27215 }, { "epoch": 0.09, "learning_rate": 4.6229943624069963e-07, "loss": 0.8351, "step": 27216 }, { "epoch": 0.09, "learning_rate": 4.619755481109589e-07, "loss": 0.7924, "step": 27217 }, { "epoch": 0.09, "learning_rate": 4.616517707966528e-07, "loss": 0.8716, "step": 27218 }, { "epoch": 0.09, "learning_rate": 4.6132810430154497e-07, "loss": 0.8575, "step": 27219 }, { "epoch": 0.09, "learning_rate": 4.610045486293924e-07, "loss": 0.8287, "step": 27220 }, { "epoch": 0.09, "learning_rate": 4.6068110378395647e-07, "loss": 0.8339, "step": 27221 }, { "epoch": 0.09, "learning_rate": 4.6035776976899425e-07, "loss": 0.8528, "step": 27222 }, { "epoch": 0.09, "learning_rate": 4.6003454658826385e-07, "loss": 0.8479, "step": 27223 }, { "epoch": 0.09, "learning_rate": 4.597114342455189e-07, "loss": 0.8312, "step": 27224 }, { "epoch": 0.09, "learning_rate": 4.593884327445142e-07, "loss": 0.8974, "step": 27225 }, { "epoch": 0.09, "learning_rate": 4.5906554208900333e-07, "loss": 0.8628, "step": 27226 }, { "epoch": 0.09, "learning_rate": 4.587427622827345e-07, "loss": 0.8626, "step": 27227 }, { "epoch": 0.09, "learning_rate": 4.5842009332946026e-07, "loss": 0.8364, "step": 27228 }, { "epoch": 0.09, "learning_rate": 4.580975352329298e-07, "loss": 0.8358, "step": 27229 }, { "epoch": 0.09, "learning_rate": 4.577750879968912e-07, "loss": 0.8229, "step": 27230 }, { "epoch": 0.09, "learning_rate": 4.574527516250893e-07, "loss": 0.8961, "step": 27231 }, { "epoch": 0.09, "learning_rate": 4.5713052612126883e-07, "loss": 0.8397, "step": 27232 }, { "epoch": 0.09, "learning_rate": 4.568084114891735e-07, "loss": 0.8679, "step": 27233 }, { "epoch": 0.09, "learning_rate": 4.5648640773254704e-07, "loss": 0.8416, "step": 27234 }, { "epoch": 0.09, "learning_rate": 4.561645148551297e-07, "loss": 0.8536, "step": 27235 }, { "epoch": 0.09, "learning_rate": 4.558427328606607e-07, "loss": 0.8416, "step": 27236 }, { "epoch": 0.09, "learning_rate": 4.5552106175288156e-07, "loss": 0.8284, "step": 27237 }, { "epoch": 0.09, "learning_rate": 4.5519950153552593e-07, "loss": 0.8368, "step": 27238 }, { "epoch": 0.09, "learning_rate": 4.548780522123319e-07, "loss": 0.8764, "step": 27239 }, { "epoch": 0.09, "learning_rate": 4.5455671378703215e-07, "loss": 0.8288, "step": 27240 }, { "epoch": 0.09, "learning_rate": 4.5423548626336246e-07, "loss": 0.8815, "step": 27241 }, { "epoch": 0.09, "learning_rate": 4.539143696450532e-07, "loss": 0.8344, "step": 27242 }, { "epoch": 0.09, "learning_rate": 4.53593363935837e-07, "loss": 0.863, "step": 27243 }, { "epoch": 0.09, "learning_rate": 4.532724691394419e-07, "loss": 0.832, "step": 27244 }, { "epoch": 0.09, "learning_rate": 4.529516852595972e-07, "loss": 0.8318, "step": 27245 }, { "epoch": 0.09, "learning_rate": 4.526310123000266e-07, "loss": 0.8274, "step": 27246 }, { "epoch": 0.09, "learning_rate": 4.523104502644604e-07, "loss": 0.84, "step": 27247 }, { "epoch": 0.09, "learning_rate": 4.519899991566212e-07, "loss": 0.8882, "step": 27248 }, { "epoch": 0.09, "learning_rate": 4.516696589802305e-07, "loss": 0.8442, "step": 27249 }, { "epoch": 0.09, "learning_rate": 4.5134942973901197e-07, "loss": 0.8424, "step": 27250 }, { "epoch": 0.09, "learning_rate": 4.5102931143668707e-07, "loss": 0.8249, "step": 27251 }, { "epoch": 0.09, "learning_rate": 4.507093040769706e-07, "loss": 0.8692, "step": 27252 }, { "epoch": 0.09, "learning_rate": 4.5038940766358506e-07, "loss": 0.8363, "step": 27253 }, { "epoch": 0.09, "learning_rate": 4.5006962220024543e-07, "loss": 0.8531, "step": 27254 }, { "epoch": 0.09, "learning_rate": 4.4974994769066747e-07, "loss": 0.8654, "step": 27255 }, { "epoch": 0.09, "learning_rate": 4.49430384138565e-07, "loss": 0.8288, "step": 27256 }, { "epoch": 0.09, "learning_rate": 4.491109315476494e-07, "loss": 0.8319, "step": 27257 }, { "epoch": 0.09, "learning_rate": 4.487915899216344e-07, "loss": 0.828, "step": 27258 }, { "epoch": 0.09, "learning_rate": 4.4847235926422925e-07, "loss": 0.8302, "step": 27259 }, { "epoch": 0.09, "learning_rate": 4.4815323957914215e-07, "loss": 0.8535, "step": 27260 }, { "epoch": 0.09, "learning_rate": 4.4783423087008225e-07, "loss": 0.8593, "step": 27261 }, { "epoch": 0.09, "learning_rate": 4.475153331407556e-07, "loss": 0.8325, "step": 27262 }, { "epoch": 0.09, "learning_rate": 4.4719654639486574e-07, "loss": 0.8475, "step": 27263 }, { "epoch": 0.09, "learning_rate": 4.4687787063611766e-07, "loss": 0.8346, "step": 27264 }, { "epoch": 0.09, "learning_rate": 4.465593058682127e-07, "loss": 0.8108, "step": 27265 }, { "epoch": 0.09, "learning_rate": 4.462408520948536e-07, "loss": 0.8599, "step": 27266 }, { "epoch": 0.09, "learning_rate": 4.459225093197395e-07, "loss": 0.9042, "step": 27267 }, { "epoch": 0.09, "learning_rate": 4.456042775465685e-07, "loss": 0.8232, "step": 27268 }, { "epoch": 0.09, "learning_rate": 4.452861567790401e-07, "loss": 0.8616, "step": 27269 }, { "epoch": 0.09, "learning_rate": 4.449681470208467e-07, "loss": 0.8257, "step": 27270 }, { "epoch": 0.09, "learning_rate": 4.4465024827568424e-07, "loss": 0.8867, "step": 27271 }, { "epoch": 0.09, "learning_rate": 4.4433246054724987e-07, "loss": 0.7878, "step": 27272 }, { "epoch": 0.09, "learning_rate": 4.4401478383923167e-07, "loss": 0.8441, "step": 27273 }, { "epoch": 0.09, "learning_rate": 4.4369721815532006e-07, "loss": 0.8484, "step": 27274 }, { "epoch": 0.09, "learning_rate": 4.4337976349920763e-07, "loss": 0.8523, "step": 27275 }, { "epoch": 0.09, "learning_rate": 4.430624198745803e-07, "loss": 0.8277, "step": 27276 }, { "epoch": 0.09, "learning_rate": 4.427451872851263e-07, "loss": 0.8236, "step": 27277 }, { "epoch": 0.09, "learning_rate": 4.4242806573453034e-07, "loss": 0.8441, "step": 27278 }, { "epoch": 0.09, "learning_rate": 4.4211105522647956e-07, "loss": 0.8394, "step": 27279 }, { "epoch": 0.09, "learning_rate": 4.4179415576465325e-07, "loss": 0.8506, "step": 27280 }, { "epoch": 0.09, "learning_rate": 4.4147736735273505e-07, "loss": 0.8637, "step": 27281 }, { "epoch": 0.09, "learning_rate": 4.4116068999440543e-07, "loss": 0.8465, "step": 27282 }, { "epoch": 0.09, "learning_rate": 4.4084412369334363e-07, "loss": 0.8624, "step": 27283 }, { "epoch": 0.09, "learning_rate": 4.405276684532267e-07, "loss": 0.8079, "step": 27284 }, { "epoch": 0.09, "learning_rate": 4.402113242777328e-07, "loss": 0.8568, "step": 27285 }, { "epoch": 0.09, "learning_rate": 4.3989509117053796e-07, "loss": 0.8612, "step": 27286 }, { "epoch": 0.09, "learning_rate": 4.395789691353136e-07, "loss": 0.8615, "step": 27287 }, { "epoch": 0.09, "learning_rate": 4.3926295817573463e-07, "loss": 0.8184, "step": 27288 }, { "epoch": 0.09, "learning_rate": 4.3894705829546914e-07, "loss": 0.8499, "step": 27289 }, { "epoch": 0.09, "learning_rate": 4.386312694981931e-07, "loss": 0.844, "step": 27290 }, { "epoch": 0.09, "learning_rate": 4.3831559178757146e-07, "loss": 0.8389, "step": 27291 }, { "epoch": 0.09, "learning_rate": 4.3800002516727224e-07, "loss": 0.8314, "step": 27292 }, { "epoch": 0.09, "learning_rate": 4.3768456964096263e-07, "loss": 0.8371, "step": 27293 }, { "epoch": 0.09, "learning_rate": 4.373692252123085e-07, "loss": 0.8281, "step": 27294 }, { "epoch": 0.09, "learning_rate": 4.370539918849703e-07, "loss": 0.8688, "step": 27295 }, { "epoch": 0.09, "learning_rate": 4.367388696626129e-07, "loss": 0.8583, "step": 27296 }, { "epoch": 0.09, "learning_rate": 4.364238585488989e-07, "loss": 0.8211, "step": 27297 }, { "epoch": 0.09, "learning_rate": 4.3610895854748645e-07, "loss": 0.83, "step": 27298 }, { "epoch": 0.09, "learning_rate": 4.357941696620327e-07, "loss": 0.8319, "step": 27299 }, { "epoch": 0.09, "learning_rate": 4.354794918961969e-07, "loss": 0.8288, "step": 27300 }, { "epoch": 0.09, "learning_rate": 4.35164925253635e-07, "loss": 0.8732, "step": 27301 }, { "epoch": 0.09, "learning_rate": 4.348504697380007e-07, "loss": 0.8501, "step": 27302 }, { "epoch": 0.09, "learning_rate": 4.345361253529479e-07, "loss": 0.8304, "step": 27303 }, { "epoch": 0.09, "learning_rate": 4.342218921021313e-07, "loss": 0.8595, "step": 27304 }, { "epoch": 0.09, "learning_rate": 4.3390776998919695e-07, "loss": 0.8411, "step": 27305 }, { "epoch": 0.09, "learning_rate": 4.335937590177974e-07, "loss": 0.8258, "step": 27306 }, { "epoch": 0.09, "learning_rate": 4.332798591915799e-07, "loss": 0.8504, "step": 27307 }, { "epoch": 0.09, "learning_rate": 4.329660705141925e-07, "loss": 0.8401, "step": 27308 }, { "epoch": 0.09, "learning_rate": 4.3265239298928006e-07, "loss": 0.8727, "step": 27309 }, { "epoch": 0.09, "learning_rate": 4.323388266204864e-07, "loss": 0.8821, "step": 27310 }, { "epoch": 0.09, "learning_rate": 4.320253714114575e-07, "loss": 0.8653, "step": 27311 }, { "epoch": 0.09, "learning_rate": 4.317120273658315e-07, "loss": 0.8461, "step": 27312 }, { "epoch": 0.09, "learning_rate": 4.313987944872511e-07, "loss": 0.8479, "step": 27313 }, { "epoch": 0.09, "learning_rate": 4.310856727793522e-07, "loss": 0.8293, "step": 27314 }, { "epoch": 0.09, "learning_rate": 4.3077266224577863e-07, "loss": 0.85, "step": 27315 }, { "epoch": 0.09, "learning_rate": 4.30459762890163e-07, "loss": 0.8528, "step": 27316 }, { "epoch": 0.09, "learning_rate": 4.301469747161413e-07, "loss": 0.7854, "step": 27317 }, { "epoch": 0.09, "learning_rate": 4.298342977273495e-07, "loss": 0.8327, "step": 27318 }, { "epoch": 0.09, "learning_rate": 4.295217319274159e-07, "loss": 0.842, "step": 27319 }, { "epoch": 0.09, "learning_rate": 4.292092773199752e-07, "loss": 0.8924, "step": 27320 }, { "epoch": 0.09, "learning_rate": 4.2889693390865796e-07, "loss": 0.8495, "step": 27321 }, { "epoch": 0.09, "learning_rate": 4.2858470169709345e-07, "loss": 0.8495, "step": 27322 }, { "epoch": 0.09, "learning_rate": 4.2827258068890545e-07, "loss": 0.8681, "step": 27323 }, { "epoch": 0.09, "learning_rate": 4.2796057088772434e-07, "loss": 0.8491, "step": 27324 }, { "epoch": 0.09, "learning_rate": 4.27648672297174e-07, "loss": 0.8808, "step": 27325 }, { "epoch": 0.09, "learning_rate": 4.27336884920877e-07, "loss": 0.8242, "step": 27326 }, { "epoch": 0.09, "learning_rate": 4.2702520876245603e-07, "loss": 0.8549, "step": 27327 }, { "epoch": 0.09, "learning_rate": 4.267136438255337e-07, "loss": 0.8313, "step": 27328 }, { "epoch": 0.09, "learning_rate": 4.2640219011373054e-07, "loss": 0.8056, "step": 27329 }, { "epoch": 0.09, "learning_rate": 4.260908476306613e-07, "loss": 0.8302, "step": 27330 }, { "epoch": 0.09, "learning_rate": 4.2577961637994544e-07, "loss": 0.8199, "step": 27331 }, { "epoch": 0.09, "learning_rate": 4.2546849636520006e-07, "loss": 0.8085, "step": 27332 }, { "epoch": 0.09, "learning_rate": 4.251574875900377e-07, "loss": 0.8244, "step": 27333 }, { "epoch": 0.09, "learning_rate": 4.248465900580734e-07, "loss": 0.8341, "step": 27334 }, { "epoch": 0.09, "learning_rate": 4.245358037729186e-07, "loss": 0.8225, "step": 27335 }, { "epoch": 0.09, "learning_rate": 4.2422512873818486e-07, "loss": 0.8298, "step": 27336 }, { "epoch": 0.09, "learning_rate": 4.239145649574805e-07, "loss": 0.8208, "step": 27337 }, { "epoch": 0.09, "learning_rate": 4.236041124344126e-07, "loss": 0.8211, "step": 27338 }, { "epoch": 0.09, "learning_rate": 4.232937711725904e-07, "loss": 0.8768, "step": 27339 }, { "epoch": 0.09, "learning_rate": 4.2298354117562114e-07, "loss": 0.8047, "step": 27340 }, { "epoch": 0.09, "learning_rate": 4.2267342244710516e-07, "loss": 0.8234, "step": 27341 }, { "epoch": 0.09, "learning_rate": 4.2236341499064635e-07, "loss": 0.8244, "step": 27342 }, { "epoch": 0.09, "learning_rate": 4.220535188098496e-07, "loss": 0.867, "step": 27343 }, { "epoch": 0.09, "learning_rate": 4.217437339083097e-07, "loss": 0.8216, "step": 27344 }, { "epoch": 0.09, "learning_rate": 4.2143406028963163e-07, "loss": 0.8616, "step": 27345 }, { "epoch": 0.09, "learning_rate": 4.211244979574103e-07, "loss": 0.8716, "step": 27346 }, { "epoch": 0.09, "learning_rate": 4.208150469152439e-07, "loss": 0.8231, "step": 27347 }, { "epoch": 0.09, "learning_rate": 4.2050570716672514e-07, "loss": 0.8727, "step": 27348 }, { "epoch": 0.09, "learning_rate": 4.2019647871545e-07, "loss": 0.8244, "step": 27349 }, { "epoch": 0.09, "learning_rate": 4.198873615650101e-07, "loss": 0.8775, "step": 27350 }, { "epoch": 0.09, "learning_rate": 4.1957835571899806e-07, "loss": 0.8815, "step": 27351 }, { "epoch": 0.09, "learning_rate": 4.192694611810033e-07, "loss": 0.8568, "step": 27352 }, { "epoch": 0.09, "learning_rate": 4.1896067795461403e-07, "loss": 0.8828, "step": 27353 }, { "epoch": 0.09, "learning_rate": 4.186520060434207e-07, "loss": 0.8381, "step": 27354 }, { "epoch": 0.1, "learning_rate": 4.183434454510049e-07, "loss": 0.8414, "step": 27355 }, { "epoch": 0.1, "learning_rate": 4.180349961809549e-07, "loss": 0.8427, "step": 27356 }, { "epoch": 0.1, "learning_rate": 4.177266582368533e-07, "loss": 0.8541, "step": 27357 }, { "epoch": 0.1, "learning_rate": 4.174184316222829e-07, "loss": 0.8646, "step": 27358 }, { "epoch": 0.1, "learning_rate": 4.1711031634082413e-07, "loss": 0.845, "step": 27359 }, { "epoch": 0.1, "learning_rate": 4.168023123960574e-07, "loss": 0.856, "step": 27360 }, { "epoch": 0.1, "learning_rate": 4.164944197915621e-07, "loss": 0.8895, "step": 27361 }, { "epoch": 0.1, "learning_rate": 4.16186638530911e-07, "loss": 0.8561, "step": 27362 }, { "epoch": 0.1, "learning_rate": 4.158789686176845e-07, "loss": 0.8366, "step": 27363 }, { "epoch": 0.1, "learning_rate": 4.155714100554575e-07, "loss": 0.8231, "step": 27364 }, { "epoch": 0.1, "learning_rate": 4.1526396284780057e-07, "loss": 0.8454, "step": 27365 }, { "epoch": 0.1, "learning_rate": 4.149566269982863e-07, "loss": 0.8578, "step": 27366 }, { "epoch": 0.1, "learning_rate": 4.1464940251048635e-07, "loss": 0.8713, "step": 27367 }, { "epoch": 0.1, "learning_rate": 4.1434228938797003e-07, "loss": 0.851, "step": 27368 }, { "epoch": 0.1, "learning_rate": 4.1403528763430456e-07, "loss": 0.8466, "step": 27369 }, { "epoch": 0.1, "learning_rate": 4.1372839725305813e-07, "loss": 0.8551, "step": 27370 }, { "epoch": 0.1, "learning_rate": 4.134216182477957e-07, "loss": 0.8669, "step": 27371 }, { "epoch": 0.1, "learning_rate": 4.1311495062208104e-07, "loss": 0.8553, "step": 27372 }, { "epoch": 0.1, "learning_rate": 4.1280839437947695e-07, "loss": 0.8365, "step": 27373 }, { "epoch": 0.1, "learning_rate": 4.125019495235449e-07, "loss": 0.8094, "step": 27374 }, { "epoch": 0.1, "learning_rate": 4.1219561605784666e-07, "loss": 0.8294, "step": 27375 }, { "epoch": 0.1, "learning_rate": 4.118893939859403e-07, "loss": 0.8682, "step": 27376 }, { "epoch": 0.1, "learning_rate": 4.115832833113842e-07, "loss": 0.8883, "step": 27377 }, { "epoch": 0.1, "learning_rate": 4.112772840377355e-07, "loss": 0.8491, "step": 27378 }, { "epoch": 0.1, "learning_rate": 4.1097139616854687e-07, "loss": 0.8828, "step": 27379 }, { "epoch": 0.1, "learning_rate": 4.106656197073733e-07, "loss": 0.8731, "step": 27380 }, { "epoch": 0.1, "learning_rate": 4.103599546577675e-07, "loss": 0.8315, "step": 27381 }, { "epoch": 0.1, "learning_rate": 4.100544010232821e-07, "loss": 0.8274, "step": 27382 }, { "epoch": 0.1, "learning_rate": 4.0974895880746656e-07, "loss": 0.8709, "step": 27383 }, { "epoch": 0.1, "learning_rate": 4.0944362801386807e-07, "loss": 0.8794, "step": 27384 }, { "epoch": 0.1, "learning_rate": 4.091384086460348e-07, "loss": 0.8108, "step": 27385 }, { "epoch": 0.1, "learning_rate": 4.08833300707514e-07, "loss": 0.8513, "step": 27386 }, { "epoch": 0.1, "learning_rate": 4.0852830420184844e-07, "loss": 0.8246, "step": 27387 }, { "epoch": 0.1, "learning_rate": 4.082234191325829e-07, "loss": 0.8637, "step": 27388 }, { "epoch": 0.1, "learning_rate": 4.079186455032613e-07, "loss": 0.8075, "step": 27389 }, { "epoch": 0.1, "learning_rate": 4.07613983317422e-07, "loss": 0.8645, "step": 27390 }, { "epoch": 0.1, "learning_rate": 4.073094325786053e-07, "loss": 0.8743, "step": 27391 }, { "epoch": 0.1, "learning_rate": 4.070049932903497e-07, "loss": 0.7855, "step": 27392 }, { "epoch": 0.1, "learning_rate": 4.0670066545619224e-07, "loss": 0.882, "step": 27393 }, { "epoch": 0.1, "learning_rate": 4.063964490796679e-07, "loss": 0.8281, "step": 27394 }, { "epoch": 0.1, "learning_rate": 4.060923441643128e-07, "loss": 0.8268, "step": 27395 }, { "epoch": 0.1, "learning_rate": 4.0578835071366065e-07, "loss": 0.8387, "step": 27396 }, { "epoch": 0.1, "learning_rate": 4.054844687312409e-07, "loss": 0.819, "step": 27397 }, { "epoch": 0.1, "learning_rate": 4.0518069822058413e-07, "loss": 0.8972, "step": 27398 }, { "epoch": 0.1, "learning_rate": 4.0487703918522193e-07, "loss": 0.8561, "step": 27399 }, { "epoch": 0.1, "learning_rate": 4.0457349162868033e-07, "loss": 0.8157, "step": 27400 }, { "epoch": 0.1, "learning_rate": 4.042700555544876e-07, "loss": 0.8021, "step": 27401 }, { "epoch": 0.1, "learning_rate": 4.039667309661677e-07, "loss": 0.7985, "step": 27402 }, { "epoch": 0.1, "learning_rate": 4.0366351786724654e-07, "loss": 0.8757, "step": 27403 }, { "epoch": 0.1, "learning_rate": 4.033604162612459e-07, "loss": 0.8271, "step": 27404 }, { "epoch": 0.1, "learning_rate": 4.0305742615168395e-07, "loss": 0.7815, "step": 27405 }, { "epoch": 0.1, "learning_rate": 4.0275454754208685e-07, "loss": 0.827, "step": 27406 }, { "epoch": 0.1, "learning_rate": 4.024517804359718e-07, "loss": 0.8754, "step": 27407 }, { "epoch": 0.1, "learning_rate": 4.0214912483685366e-07, "loss": 0.8455, "step": 27408 }, { "epoch": 0.1, "learning_rate": 4.018465807482508e-07, "loss": 0.8737, "step": 27409 }, { "epoch": 0.1, "learning_rate": 4.015441481736793e-07, "loss": 0.8518, "step": 27410 }, { "epoch": 0.1, "learning_rate": 4.012418271166485e-07, "loss": 0.8792, "step": 27411 }, { "epoch": 0.1, "learning_rate": 4.009396175806757e-07, "loss": 0.82, "step": 27412 }, { "epoch": 0.1, "learning_rate": 4.0063751956927133e-07, "loss": 0.871, "step": 27413 }, { "epoch": 0.1, "learning_rate": 4.0033553308594377e-07, "loss": 0.838, "step": 27414 }, { "epoch": 0.1, "learning_rate": 4.0003365813420236e-07, "loss": 0.8358, "step": 27415 }, { "epoch": 0.1, "learning_rate": 3.9973189471755325e-07, "loss": 0.834, "step": 27416 }, { "epoch": 0.1, "learning_rate": 3.994302428395036e-07, "loss": 0.836, "step": 27417 }, { "epoch": 0.1, "learning_rate": 3.991287025035573e-07, "loss": 0.8648, "step": 27418 }, { "epoch": 0.1, "learning_rate": 3.988272737132182e-07, "loss": 0.843, "step": 27419 }, { "epoch": 0.1, "learning_rate": 3.985259564719879e-07, "loss": 0.8282, "step": 27420 }, { "epoch": 0.1, "learning_rate": 3.982247507833703e-07, "loss": 0.8235, "step": 27421 }, { "epoch": 0.1, "learning_rate": 3.979236566508593e-07, "loss": 0.8467, "step": 27422 }, { "epoch": 0.1, "learning_rate": 3.976226740779576e-07, "loss": 0.8581, "step": 27423 }, { "epoch": 0.1, "learning_rate": 3.9732180306815914e-07, "loss": 0.8491, "step": 27424 }, { "epoch": 0.1, "learning_rate": 3.970210436249622e-07, "loss": 0.8312, "step": 27425 }, { "epoch": 0.1, "learning_rate": 3.967203957518584e-07, "loss": 0.8419, "step": 27426 }, { "epoch": 0.1, "learning_rate": 3.964198594523427e-07, "loss": 0.9188, "step": 27427 }, { "epoch": 0.1, "learning_rate": 3.9611943472990797e-07, "loss": 0.8539, "step": 27428 }, { "epoch": 0.1, "learning_rate": 3.9581912158804134e-07, "loss": 0.861, "step": 27429 }, { "epoch": 0.1, "learning_rate": 3.955189200302323e-07, "loss": 0.8962, "step": 27430 }, { "epoch": 0.1, "learning_rate": 3.952188300599702e-07, "loss": 0.8358, "step": 27431 }, { "epoch": 0.1, "learning_rate": 3.949188516807434e-07, "loss": 0.8878, "step": 27432 }, { "epoch": 0.1, "learning_rate": 3.946189848960336e-07, "loss": 0.8921, "step": 27433 }, { "epoch": 0.1, "learning_rate": 3.943192297093268e-07, "loss": 0.8178, "step": 27434 }, { "epoch": 0.1, "learning_rate": 3.940195861241036e-07, "loss": 0.8437, "step": 27435 }, { "epoch": 0.1, "learning_rate": 3.9372005414384794e-07, "loss": 0.8738, "step": 27436 }, { "epoch": 0.1, "learning_rate": 3.934206337720381e-07, "loss": 0.8457, "step": 27437 }, { "epoch": 0.1, "learning_rate": 3.931213250121535e-07, "loss": 0.8577, "step": 27438 }, { "epoch": 0.1, "learning_rate": 3.9282212786767247e-07, "loss": 0.8571, "step": 27439 }, { "epoch": 0.1, "learning_rate": 3.9252304234206896e-07, "loss": 0.8204, "step": 27440 }, { "epoch": 0.1, "learning_rate": 3.922240684388179e-07, "loss": 0.8191, "step": 27441 }, { "epoch": 0.1, "learning_rate": 3.919252061613954e-07, "loss": 0.8225, "step": 27442 }, { "epoch": 0.1, "learning_rate": 3.9162645551327203e-07, "loss": 0.8684, "step": 27443 }, { "epoch": 0.1, "learning_rate": 3.913278164979184e-07, "loss": 0.8103, "step": 27444 }, { "epoch": 0.1, "learning_rate": 3.910292891188061e-07, "loss": 0.8438, "step": 27445 }, { "epoch": 0.1, "learning_rate": 3.907308733794024e-07, "loss": 0.865, "step": 27446 }, { "epoch": 0.1, "learning_rate": 3.904325692831734e-07, "loss": 0.8662, "step": 27447 }, { "epoch": 0.1, "learning_rate": 3.9013437683358525e-07, "loss": 0.8287, "step": 27448 }, { "epoch": 0.1, "learning_rate": 3.898362960341029e-07, "loss": 0.845, "step": 27449 }, { "epoch": 0.1, "learning_rate": 3.8953832688818916e-07, "loss": 0.8196, "step": 27450 }, { "epoch": 0.1, "learning_rate": 3.892404693993068e-07, "loss": 0.8399, "step": 27451 }, { "epoch": 0.1, "learning_rate": 3.889427235709153e-07, "loss": 0.8432, "step": 27452 }, { "epoch": 0.1, "learning_rate": 3.886450894064764e-07, "loss": 0.8357, "step": 27453 }, { "epoch": 0.1, "learning_rate": 3.883475669094439e-07, "loss": 0.8451, "step": 27454 }, { "epoch": 0.1, "learning_rate": 3.8805015608327723e-07, "loss": 0.8675, "step": 27455 }, { "epoch": 0.1, "learning_rate": 3.8775285693143263e-07, "loss": 0.8187, "step": 27456 }, { "epoch": 0.1, "learning_rate": 3.874556694573628e-07, "loss": 0.8569, "step": 27457 }, { "epoch": 0.1, "learning_rate": 3.871585936645195e-07, "loss": 0.8397, "step": 27458 }, { "epoch": 0.1, "learning_rate": 3.868616295563554e-07, "loss": 0.8477, "step": 27459 }, { "epoch": 0.1, "learning_rate": 3.865647771363212e-07, "loss": 0.8175, "step": 27460 }, { "epoch": 0.1, "learning_rate": 3.8626803640786526e-07, "loss": 0.8609, "step": 27461 }, { "epoch": 0.1, "learning_rate": 3.859714073744358e-07, "loss": 0.8429, "step": 27462 }, { "epoch": 0.1, "learning_rate": 3.8567489003947914e-07, "loss": 0.844, "step": 27463 }, { "epoch": 0.1, "learning_rate": 3.85378484406439e-07, "loss": 0.849, "step": 27464 }, { "epoch": 0.1, "learning_rate": 3.850821904787594e-07, "loss": 0.8379, "step": 27465 }, { "epoch": 0.1, "learning_rate": 3.8478600825988423e-07, "loss": 0.8381, "step": 27466 }, { "epoch": 0.1, "learning_rate": 3.8448993775325185e-07, "loss": 0.844, "step": 27467 }, { "epoch": 0.1, "learning_rate": 3.8419397896230506e-07, "loss": 0.8462, "step": 27468 }, { "epoch": 0.1, "learning_rate": 3.8389813189048107e-07, "loss": 0.8098, "step": 27469 }, { "epoch": 0.1, "learning_rate": 3.8360239654121946e-07, "loss": 0.848, "step": 27470 }, { "epoch": 0.1, "learning_rate": 3.8330677291795184e-07, "loss": 0.858, "step": 27471 }, { "epoch": 0.1, "learning_rate": 3.8301126102411435e-07, "loss": 0.86, "step": 27472 }, { "epoch": 0.1, "learning_rate": 3.8271586086314097e-07, "loss": 0.8405, "step": 27473 }, { "epoch": 0.1, "learning_rate": 3.824205724384655e-07, "loss": 0.8372, "step": 27474 }, { "epoch": 0.1, "learning_rate": 3.821253957535154e-07, "loss": 0.8493, "step": 27475 }, { "epoch": 0.1, "learning_rate": 3.8183033081172216e-07, "loss": 0.8271, "step": 27476 }, { "epoch": 0.1, "learning_rate": 3.8153537761651317e-07, "loss": 0.8484, "step": 27477 }, { "epoch": 0.1, "learning_rate": 3.812405361713156e-07, "loss": 0.8667, "step": 27478 }, { "epoch": 0.1, "learning_rate": 3.809458064795546e-07, "loss": 0.8741, "step": 27479 }, { "epoch": 0.1, "learning_rate": 3.806511885446551e-07, "loss": 0.7904, "step": 27480 }, { "epoch": 0.1, "learning_rate": 3.8035668237004e-07, "loss": 0.7564, "step": 27481 }, { "epoch": 0.1, "learning_rate": 3.8006228795912983e-07, "loss": 0.8328, "step": 27482 }, { "epoch": 0.1, "learning_rate": 3.797680053153463e-07, "loss": 0.845, "step": 27483 }, { "epoch": 0.1, "learning_rate": 3.794738344421078e-07, "loss": 0.8356, "step": 27484 }, { "epoch": 0.1, "learning_rate": 3.791797753428317e-07, "loss": 0.8382, "step": 27485 }, { "epoch": 0.1, "learning_rate": 3.788858280209362e-07, "loss": 0.7995, "step": 27486 }, { "epoch": 0.1, "learning_rate": 3.7859199247983424e-07, "loss": 0.84, "step": 27487 }, { "epoch": 0.1, "learning_rate": 3.7829826872294194e-07, "loss": 0.8162, "step": 27488 }, { "epoch": 0.1, "learning_rate": 3.780046567536688e-07, "loss": 0.8744, "step": 27489 }, { "epoch": 0.1, "learning_rate": 3.777111565754288e-07, "loss": 0.8164, "step": 27490 }, { "epoch": 0.1, "learning_rate": 3.7741776819163025e-07, "loss": 0.8829, "step": 27491 }, { "epoch": 0.1, "learning_rate": 3.7712449160568374e-07, "loss": 0.8189, "step": 27492 }, { "epoch": 0.1, "learning_rate": 3.768313268209944e-07, "loss": 0.832, "step": 27493 }, { "epoch": 0.1, "learning_rate": 3.765382738409695e-07, "loss": 0.8918, "step": 27494 }, { "epoch": 0.1, "learning_rate": 3.762453326690152e-07, "loss": 0.9005, "step": 27495 }, { "epoch": 0.1, "learning_rate": 3.7595250330853207e-07, "loss": 0.8487, "step": 27496 }, { "epoch": 0.1, "learning_rate": 3.756597857629218e-07, "loss": 0.8457, "step": 27497 }, { "epoch": 0.1, "learning_rate": 3.7536718003558956e-07, "loss": 0.8305, "step": 27498 }, { "epoch": 0.1, "learning_rate": 3.7507468612993257e-07, "loss": 0.8662, "step": 27499 }, { "epoch": 0.1, "learning_rate": 3.747823040493481e-07, "loss": 0.8454, "step": 27500 }, { "epoch": 0.1, "learning_rate": 3.744900337972346e-07, "loss": 0.8509, "step": 27501 }, { "epoch": 0.1, "learning_rate": 3.7419787537698813e-07, "loss": 0.85, "step": 27502 }, { "epoch": 0.1, "learning_rate": 3.739058287919994e-07, "loss": 0.8565, "step": 27503 }, { "epoch": 0.1, "learning_rate": 3.7361389404566573e-07, "loss": 0.819, "step": 27504 }, { "epoch": 0.1, "learning_rate": 3.7332207114137764e-07, "loss": 0.8207, "step": 27505 }, { "epoch": 0.1, "learning_rate": 3.730303600825258e-07, "loss": 0.8295, "step": 27506 }, { "epoch": 0.1, "learning_rate": 3.727387608724986e-07, "loss": 0.8699, "step": 27507 }, { "epoch": 0.1, "learning_rate": 3.7244727351468447e-07, "loss": 0.8141, "step": 27508 }, { "epoch": 0.1, "learning_rate": 3.7215589801246845e-07, "loss": 0.8295, "step": 27509 }, { "epoch": 0.1, "learning_rate": 3.7186463436923783e-07, "loss": 0.8224, "step": 27510 }, { "epoch": 0.1, "learning_rate": 3.715734825883766e-07, "loss": 0.8378, "step": 27511 }, { "epoch": 0.1, "learning_rate": 3.712824426732653e-07, "loss": 0.8065, "step": 27512 }, { "epoch": 0.1, "learning_rate": 3.709915146272891e-07, "loss": 0.8391, "step": 27513 }, { "epoch": 0.1, "learning_rate": 3.7070069845382415e-07, "loss": 0.8469, "step": 27514 }, { "epoch": 0.1, "learning_rate": 3.7040999415625113e-07, "loss": 0.8885, "step": 27515 }, { "epoch": 0.1, "learning_rate": 3.7011940173794726e-07, "loss": 0.8621, "step": 27516 }, { "epoch": 0.1, "learning_rate": 3.6982892120228875e-07, "loss": 0.8213, "step": 27517 }, { "epoch": 0.1, "learning_rate": 3.695385525526507e-07, "loss": 0.8854, "step": 27518 }, { "epoch": 0.1, "learning_rate": 3.6924829579240595e-07, "loss": 0.8471, "step": 27519 }, { "epoch": 0.1, "learning_rate": 3.689581509249285e-07, "loss": 0.8516, "step": 27520 }, { "epoch": 0.1, "learning_rate": 3.686681179535867e-07, "loss": 0.8499, "step": 27521 }, { "epoch": 0.1, "learning_rate": 3.683781968817501e-07, "loss": 0.8181, "step": 27522 }, { "epoch": 0.1, "learning_rate": 3.6808838771279054e-07, "loss": 0.8393, "step": 27523 }, { "epoch": 0.1, "learning_rate": 3.677986904500741e-07, "loss": 0.8462, "step": 27524 }, { "epoch": 0.1, "learning_rate": 3.675091050969648e-07, "loss": 0.8106, "step": 27525 }, { "epoch": 0.1, "learning_rate": 3.6721963165682774e-07, "loss": 0.8122, "step": 27526 }, { "epoch": 0.1, "learning_rate": 3.6693027013302683e-07, "loss": 0.8315, "step": 27527 }, { "epoch": 0.1, "learning_rate": 3.666410205289228e-07, "loss": 0.8311, "step": 27528 }, { "epoch": 0.1, "learning_rate": 3.663518828478774e-07, "loss": 0.8524, "step": 27529 }, { "epoch": 0.1, "learning_rate": 3.66062857093249e-07, "loss": 0.8524, "step": 27530 }, { "epoch": 0.1, "learning_rate": 3.657739432683971e-07, "loss": 0.8657, "step": 27531 }, { "epoch": 0.1, "learning_rate": 3.654851413766758e-07, "loss": 0.824, "step": 27532 }, { "epoch": 0.1, "learning_rate": 3.651964514214423e-07, "loss": 0.8303, "step": 27533 }, { "epoch": 0.1, "learning_rate": 3.649078734060507e-07, "loss": 0.8859, "step": 27534 }, { "epoch": 0.1, "learning_rate": 3.646194073338527e-07, "loss": 0.8805, "step": 27535 }, { "epoch": 0.1, "learning_rate": 3.6433105320820116e-07, "loss": 0.8529, "step": 27536 }, { "epoch": 0.1, "learning_rate": 3.6404281103244565e-07, "loss": 0.8488, "step": 27537 }, { "epoch": 0.1, "learning_rate": 3.637546808099357e-07, "loss": 0.8723, "step": 27538 }, { "epoch": 0.1, "learning_rate": 3.634666625440175e-07, "loss": 0.84, "step": 27539 }, { "epoch": 0.1, "learning_rate": 3.631787562380362e-07, "loss": 0.8527, "step": 27540 }, { "epoch": 0.1, "learning_rate": 3.6289096189534133e-07, "loss": 0.8368, "step": 27541 }, { "epoch": 0.1, "learning_rate": 3.6260327951927245e-07, "loss": 0.8841, "step": 27542 }, { "epoch": 0.1, "learning_rate": 3.623157091131735e-07, "loss": 0.8653, "step": 27543 }, { "epoch": 0.1, "learning_rate": 3.6202825068038517e-07, "loss": 0.8263, "step": 27544 }, { "epoch": 0.1, "learning_rate": 3.617409042242481e-07, "loss": 0.8871, "step": 27545 }, { "epoch": 0.1, "learning_rate": 3.6145366974809856e-07, "loss": 0.8529, "step": 27546 }, { "epoch": 0.1, "learning_rate": 3.611665472552761e-07, "loss": 0.8362, "step": 27547 }, { "epoch": 0.1, "learning_rate": 3.608795367491169e-07, "loss": 0.8676, "step": 27548 }, { "epoch": 0.1, "learning_rate": 3.605926382329539e-07, "loss": 0.8354, "step": 27549 }, { "epoch": 0.1, "learning_rate": 3.603058517101199e-07, "loss": 0.7901, "step": 27550 }, { "epoch": 0.1, "learning_rate": 3.6001917718394786e-07, "loss": 0.8389, "step": 27551 }, { "epoch": 0.1, "learning_rate": 3.597326146577684e-07, "loss": 0.8766, "step": 27552 }, { "epoch": 0.1, "learning_rate": 3.594461641349112e-07, "loss": 0.8393, "step": 27553 }, { "epoch": 0.1, "learning_rate": 3.5915982561870346e-07, "loss": 0.8421, "step": 27554 }, { "epoch": 0.1, "learning_rate": 3.5887359911247366e-07, "loss": 0.8485, "step": 27555 }, { "epoch": 0.1, "learning_rate": 3.5858748461954474e-07, "loss": 0.8295, "step": 27556 }, { "epoch": 0.1, "learning_rate": 3.583014821432429e-07, "loss": 0.8269, "step": 27557 }, { "epoch": 0.1, "learning_rate": 3.5801559168688995e-07, "loss": 0.7884, "step": 27558 }, { "epoch": 0.1, "learning_rate": 3.577298132538076e-07, "loss": 0.8464, "step": 27559 }, { "epoch": 0.1, "learning_rate": 3.5744414684731666e-07, "loss": 0.7878, "step": 27560 }, { "epoch": 0.1, "learning_rate": 3.5715859247073547e-07, "loss": 0.8167, "step": 27561 }, { "epoch": 0.1, "learning_rate": 3.5687315012738257e-07, "loss": 0.8488, "step": 27562 }, { "epoch": 0.1, "learning_rate": 3.56587819820573e-07, "loss": 0.857, "step": 27563 }, { "epoch": 0.1, "learning_rate": 3.5630260155362306e-07, "loss": 0.8129, "step": 27564 }, { "epoch": 0.1, "learning_rate": 3.560174953298434e-07, "loss": 0.8438, "step": 27565 }, { "epoch": 0.1, "learning_rate": 3.5573250115255253e-07, "loss": 0.8552, "step": 27566 }, { "epoch": 0.1, "learning_rate": 3.554476190250555e-07, "loss": 0.8012, "step": 27567 }, { "epoch": 0.1, "learning_rate": 3.5516284895066533e-07, "loss": 0.8154, "step": 27568 }, { "epoch": 0.1, "learning_rate": 3.548781909326904e-07, "loss": 0.844, "step": 27569 }, { "epoch": 0.1, "learning_rate": 3.5459364497443696e-07, "loss": 0.825, "step": 27570 }, { "epoch": 0.1, "learning_rate": 3.5430921107921124e-07, "loss": 0.8786, "step": 27571 }, { "epoch": 0.1, "learning_rate": 3.540248892503184e-07, "loss": 0.8225, "step": 27572 }, { "epoch": 0.1, "learning_rate": 3.537406794910625e-07, "loss": 0.8195, "step": 27573 }, { "epoch": 0.1, "learning_rate": 3.5345658180474306e-07, "loss": 0.8399, "step": 27574 }, { "epoch": 0.1, "learning_rate": 3.53172596194663e-07, "loss": 0.8477, "step": 27575 }, { "epoch": 0.1, "learning_rate": 3.528887226641209e-07, "loss": 0.8225, "step": 27576 }, { "epoch": 0.1, "learning_rate": 3.5260496121641397e-07, "loss": 0.813, "step": 27577 }, { "epoch": 0.1, "learning_rate": 3.5232131185484075e-07, "loss": 0.8647, "step": 27578 }, { "epoch": 0.1, "learning_rate": 3.520377745826953e-07, "loss": 0.8425, "step": 27579 }, { "epoch": 0.1, "learning_rate": 3.517543494032749e-07, "loss": 0.8188, "step": 27580 }, { "epoch": 0.1, "learning_rate": 3.5147103631986813e-07, "loss": 0.8181, "step": 27581 }, { "epoch": 0.1, "learning_rate": 3.5118783533576893e-07, "loss": 0.8539, "step": 27582 }, { "epoch": 0.1, "learning_rate": 3.5090474645426584e-07, "loss": 0.8986, "step": 27583 }, { "epoch": 0.1, "learning_rate": 3.506217696786518e-07, "loss": 0.8508, "step": 27584 }, { "epoch": 0.1, "learning_rate": 3.5033890501221077e-07, "loss": 0.8639, "step": 27585 }, { "epoch": 0.1, "learning_rate": 3.5005615245823023e-07, "loss": 0.8468, "step": 27586 }, { "epoch": 0.1, "learning_rate": 3.497735120199974e-07, "loss": 0.8508, "step": 27587 }, { "epoch": 0.1, "learning_rate": 3.4949098370079203e-07, "loss": 0.8302, "step": 27588 }, { "epoch": 0.1, "learning_rate": 3.492085675038981e-07, "loss": 0.8518, "step": 27589 }, { "epoch": 0.1, "learning_rate": 3.489262634325985e-07, "loss": 0.8226, "step": 27590 }, { "epoch": 0.1, "learning_rate": 3.4864407149017286e-07, "loss": 0.8686, "step": 27591 }, { "epoch": 0.1, "learning_rate": 3.483619916798986e-07, "loss": 0.8172, "step": 27592 }, { "epoch": 0.1, "learning_rate": 3.4808002400505193e-07, "loss": 0.8346, "step": 27593 }, { "epoch": 0.1, "learning_rate": 3.4779816846891247e-07, "loss": 0.8114, "step": 27594 }, { "epoch": 0.1, "learning_rate": 3.475164250747498e-07, "loss": 0.7996, "step": 27595 }, { "epoch": 0.1, "learning_rate": 3.4723479382584134e-07, "loss": 0.8003, "step": 27596 }, { "epoch": 0.1, "learning_rate": 3.4695327472545893e-07, "loss": 0.815, "step": 27597 }, { "epoch": 0.1, "learning_rate": 3.466718677768721e-07, "loss": 0.8237, "step": 27598 }, { "epoch": 0.1, "learning_rate": 3.4639057298335054e-07, "loss": 0.8689, "step": 27599 }, { "epoch": 0.1, "learning_rate": 3.461093903481616e-07, "loss": 0.8129, "step": 27600 }, { "epoch": 0.1, "learning_rate": 3.4582831987457376e-07, "loss": 0.8565, "step": 27601 }, { "epoch": 0.1, "learning_rate": 3.455473615658511e-07, "loss": 0.84, "step": 27602 }, { "epoch": 0.1, "learning_rate": 3.452665154252599e-07, "loss": 0.841, "step": 27603 }, { "epoch": 0.1, "learning_rate": 3.4498578145606087e-07, "loss": 0.8675, "step": 27604 }, { "epoch": 0.1, "learning_rate": 3.4470515966151694e-07, "loss": 0.8296, "step": 27605 }, { "epoch": 0.1, "learning_rate": 3.444246500448878e-07, "loss": 0.8557, "step": 27606 }, { "epoch": 0.1, "learning_rate": 3.44144252609433e-07, "loss": 0.8486, "step": 27607 }, { "epoch": 0.1, "learning_rate": 3.4386396735840987e-07, "loss": 0.8439, "step": 27608 }, { "epoch": 0.1, "learning_rate": 3.4358379429507484e-07, "loss": 0.8351, "step": 27609 }, { "epoch": 0.1, "learning_rate": 3.433037334226841e-07, "loss": 0.821, "step": 27610 }, { "epoch": 0.1, "learning_rate": 3.4302378474448947e-07, "loss": 0.8915, "step": 27611 }, { "epoch": 0.1, "learning_rate": 3.4274394826374625e-07, "loss": 0.8431, "step": 27612 }, { "epoch": 0.1, "learning_rate": 3.424642239837017e-07, "loss": 0.8147, "step": 27613 }, { "epoch": 0.1, "learning_rate": 3.421846119076089e-07, "loss": 0.8301, "step": 27614 }, { "epoch": 0.1, "learning_rate": 3.4190511203871626e-07, "loss": 0.852, "step": 27615 }, { "epoch": 0.1, "learning_rate": 3.416257243802712e-07, "loss": 0.8908, "step": 27616 }, { "epoch": 0.1, "learning_rate": 3.413464489355178e-07, "loss": 0.8903, "step": 27617 }, { "epoch": 0.1, "learning_rate": 3.410672857077013e-07, "loss": 0.807, "step": 27618 }, { "epoch": 0.1, "learning_rate": 3.407882347000668e-07, "loss": 0.8917, "step": 27619 }, { "epoch": 0.1, "learning_rate": 3.405092959158551e-07, "loss": 0.8244, "step": 27620 }, { "epoch": 0.1, "learning_rate": 3.4023046935830693e-07, "loss": 0.8476, "step": 27621 }, { "epoch": 0.1, "learning_rate": 3.399517550306619e-07, "loss": 0.8167, "step": 27622 }, { "epoch": 0.1, "learning_rate": 3.396731529361597e-07, "loss": 0.8099, "step": 27623 }, { "epoch": 0.1, "learning_rate": 3.393946630780343e-07, "loss": 0.8615, "step": 27624 }, { "epoch": 0.1, "learning_rate": 3.391162854595231e-07, "loss": 0.8583, "step": 27625 }, { "epoch": 0.1, "learning_rate": 3.3883802008386034e-07, "loss": 0.8303, "step": 27626 }, { "epoch": 0.1, "learning_rate": 3.385598669542789e-07, "loss": 0.864, "step": 27627 }, { "epoch": 0.1, "learning_rate": 3.382818260740106e-07, "loss": 0.8163, "step": 27628 }, { "epoch": 0.1, "learning_rate": 3.38003897446284e-07, "loss": 0.859, "step": 27629 }, { "epoch": 0.1, "learning_rate": 3.3772608107433214e-07, "loss": 0.8869, "step": 27630 }, { "epoch": 0.1, "learning_rate": 3.374483769613779e-07, "loss": 0.8429, "step": 27631 }, { "epoch": 0.1, "learning_rate": 3.371707851106498e-07, "loss": 0.8407, "step": 27632 }, { "epoch": 0.1, "learning_rate": 3.368933055253754e-07, "loss": 0.8201, "step": 27633 }, { "epoch": 0.1, "learning_rate": 3.3661593820877524e-07, "loss": 0.8162, "step": 27634 }, { "epoch": 0.1, "learning_rate": 3.363386831640736e-07, "loss": 0.8292, "step": 27635 }, { "epoch": 0.1, "learning_rate": 3.3606154039449e-07, "loss": 0.8049, "step": 27636 }, { "epoch": 0.1, "learning_rate": 3.357845099032475e-07, "loss": 0.8639, "step": 27637 }, { "epoch": 0.1, "learning_rate": 3.3550759169356017e-07, "loss": 0.8434, "step": 27638 }, { "epoch": 0.1, "learning_rate": 3.3523078576864874e-07, "loss": 0.8759, "step": 27639 }, { "epoch": 0.1, "learning_rate": 3.349540921317296e-07, "loss": 0.8508, "step": 27640 }, { "epoch": 0.1, "learning_rate": 3.346775107860145e-07, "loss": 0.8669, "step": 27641 }, { "epoch": 0.1, "learning_rate": 3.3440104173471876e-07, "loss": 0.8409, "step": 27642 }, { "epoch": 0.1, "learning_rate": 3.3412468498105535e-07, "loss": 0.8168, "step": 27643 }, { "epoch": 0.1, "learning_rate": 3.338484405282327e-07, "loss": 0.8368, "step": 27644 }, { "epoch": 0.1, "learning_rate": 3.3357230837946176e-07, "loss": 0.8333, "step": 27645 }, { "epoch": 0.1, "learning_rate": 3.332962885379509e-07, "loss": 0.8702, "step": 27646 }, { "epoch": 0.1, "learning_rate": 3.330203810069077e-07, "loss": 0.8718, "step": 27647 }, { "epoch": 0.1, "learning_rate": 3.3274458578953504e-07, "loss": 0.8182, "step": 27648 }, { "epoch": 0.1, "learning_rate": 3.324689028890382e-07, "loss": 0.8251, "step": 27649 }, { "epoch": 0.1, "learning_rate": 3.3219333230862236e-07, "loss": 0.8229, "step": 27650 }, { "epoch": 0.1, "learning_rate": 3.3191787405148614e-07, "loss": 0.88, "step": 27651 }, { "epoch": 0.1, "learning_rate": 3.316425281208313e-07, "loss": 0.8592, "step": 27652 }, { "epoch": 0.1, "learning_rate": 3.313672945198576e-07, "loss": 0.8459, "step": 27653 }, { "epoch": 0.1, "learning_rate": 3.310921732517636e-07, "loss": 0.8413, "step": 27654 }, { "epoch": 0.1, "learning_rate": 3.308171643197422e-07, "loss": 0.8496, "step": 27655 }, { "epoch": 0.11, "learning_rate": 3.305422677269898e-07, "loss": 0.8159, "step": 27656 }, { "epoch": 0.11, "learning_rate": 3.302674834767006e-07, "loss": 0.8614, "step": 27657 }, { "epoch": 0.11, "learning_rate": 3.299928115720696e-07, "loss": 0.8342, "step": 27658 }, { "epoch": 0.11, "learning_rate": 3.297182520162856e-07, "loss": 0.8588, "step": 27659 }, { "epoch": 0.11, "learning_rate": 3.29443804812537e-07, "loss": 0.8111, "step": 27660 }, { "epoch": 0.11, "learning_rate": 3.291694699640158e-07, "loss": 0.8471, "step": 27661 }, { "epoch": 0.11, "learning_rate": 3.288952474739071e-07, "loss": 0.8401, "step": 27662 }, { "epoch": 0.11, "learning_rate": 3.2862113734539736e-07, "loss": 0.8511, "step": 27663 }, { "epoch": 0.11, "learning_rate": 3.2834713958167176e-07, "loss": 0.843, "step": 27664 }, { "epoch": 0.11, "learning_rate": 3.280732541859133e-07, "loss": 0.8244, "step": 27665 }, { "epoch": 0.11, "learning_rate": 3.277994811613039e-07, "loss": 0.8278, "step": 27666 }, { "epoch": 0.11, "learning_rate": 3.2752582051102435e-07, "loss": 0.8425, "step": 27667 }, { "epoch": 0.11, "learning_rate": 3.272522722382543e-07, "loss": 0.8085, "step": 27668 }, { "epoch": 0.11, "learning_rate": 3.2697883634617235e-07, "loss": 0.8466, "step": 27669 }, { "epoch": 0.11, "learning_rate": 3.267055128379537e-07, "loss": 0.8417, "step": 27670 }, { "epoch": 0.11, "learning_rate": 3.264323017167759e-07, "loss": 0.8618, "step": 27671 }, { "epoch": 0.11, "learning_rate": 3.2615920298581296e-07, "loss": 0.8204, "step": 27672 }, { "epoch": 0.11, "learning_rate": 3.2588621664823686e-07, "loss": 0.8018, "step": 27673 }, { "epoch": 0.11, "learning_rate": 3.256133427072183e-07, "loss": 0.8207, "step": 27674 }, { "epoch": 0.11, "learning_rate": 3.253405811659294e-07, "loss": 0.8568, "step": 27675 }, { "epoch": 0.11, "learning_rate": 3.2506793202753964e-07, "loss": 0.7897, "step": 27676 }, { "epoch": 0.11, "learning_rate": 3.247953952952154e-07, "loss": 0.8537, "step": 27677 }, { "epoch": 0.11, "learning_rate": 3.2452297097212317e-07, "loss": 0.8149, "step": 27678 }, { "epoch": 0.11, "learning_rate": 3.242506590614292e-07, "loss": 0.853, "step": 27679 }, { "epoch": 0.11, "learning_rate": 3.2397845956629535e-07, "loss": 0.8357, "step": 27680 }, { "epoch": 0.11, "learning_rate": 3.237063724898848e-07, "loss": 0.8196, "step": 27681 }, { "epoch": 0.11, "learning_rate": 3.2343439783535934e-07, "loss": 0.813, "step": 27682 }, { "epoch": 0.11, "learning_rate": 3.2316253560587984e-07, "loss": 0.8743, "step": 27683 }, { "epoch": 0.11, "learning_rate": 3.228907858046026e-07, "loss": 0.828, "step": 27684 }, { "epoch": 0.11, "learning_rate": 3.2261914843468634e-07, "loss": 0.8511, "step": 27685 }, { "epoch": 0.11, "learning_rate": 3.2234762349928614e-07, "loss": 0.8513, "step": 27686 }, { "epoch": 0.11, "learning_rate": 3.220762110015574e-07, "loss": 0.8584, "step": 27687 }, { "epoch": 0.11, "learning_rate": 3.2180491094465414e-07, "loss": 0.8344, "step": 27688 }, { "epoch": 0.11, "learning_rate": 3.2153372333172616e-07, "loss": 0.8076, "step": 27689 }, { "epoch": 0.11, "learning_rate": 3.212626481659264e-07, "loss": 0.8569, "step": 27690 }, { "epoch": 0.11, "learning_rate": 3.209916854504036e-07, "loss": 0.8528, "step": 27691 }, { "epoch": 0.11, "learning_rate": 3.207208351883051e-07, "loss": 0.7792, "step": 27692 }, { "epoch": 0.11, "learning_rate": 3.2045009738277733e-07, "loss": 0.8326, "step": 27693 }, { "epoch": 0.11, "learning_rate": 3.2017947203696774e-07, "loss": 0.8232, "step": 27694 }, { "epoch": 0.11, "learning_rate": 3.1990895915401943e-07, "loss": 0.8499, "step": 27695 }, { "epoch": 0.11, "learning_rate": 3.196385587370754e-07, "loss": 0.8288, "step": 27696 }, { "epoch": 0.11, "learning_rate": 3.193682707892776e-07, "loss": 0.8137, "step": 27697 }, { "epoch": 0.11, "learning_rate": 3.190980953137657e-07, "loss": 0.8766, "step": 27698 }, { "epoch": 0.11, "learning_rate": 3.1882803231367833e-07, "loss": 0.8501, "step": 27699 }, { "epoch": 0.11, "learning_rate": 3.185580817921541e-07, "loss": 0.8722, "step": 27700 }, { "epoch": 0.11, "learning_rate": 3.182882437523294e-07, "loss": 0.8364, "step": 27701 }, { "epoch": 0.11, "learning_rate": 3.180185181973383e-07, "loss": 0.8472, "step": 27702 }, { "epoch": 0.11, "learning_rate": 3.1774890513031617e-07, "loss": 0.8206, "step": 27703 }, { "epoch": 0.11, "learning_rate": 3.174794045543939e-07, "loss": 0.8492, "step": 27704 }, { "epoch": 0.11, "learning_rate": 3.172100164727021e-07, "loss": 0.8386, "step": 27705 }, { "epoch": 0.11, "learning_rate": 3.169407408883729e-07, "loss": 0.8189, "step": 27706 }, { "epoch": 0.11, "learning_rate": 3.1667157780453264e-07, "loss": 0.828, "step": 27707 }, { "epoch": 0.11, "learning_rate": 3.1640252722431096e-07, "loss": 0.8059, "step": 27708 }, { "epoch": 0.11, "learning_rate": 3.16133589150831e-07, "loss": 0.8609, "step": 27709 }, { "epoch": 0.11, "learning_rate": 3.158647635872192e-07, "loss": 0.8418, "step": 27710 }, { "epoch": 0.11, "learning_rate": 3.1559605053659734e-07, "loss": 0.8581, "step": 27711 }, { "epoch": 0.11, "learning_rate": 3.1532745000208976e-07, "loss": 0.817, "step": 27712 }, { "epoch": 0.11, "learning_rate": 3.150589619868138e-07, "loss": 0.8604, "step": 27713 }, { "epoch": 0.11, "learning_rate": 3.147905864938927e-07, "loss": 0.8235, "step": 27714 }, { "epoch": 0.11, "learning_rate": 3.145223235264427e-07, "loss": 0.8099, "step": 27715 }, { "epoch": 0.11, "learning_rate": 3.1425417308757923e-07, "loss": 0.8156, "step": 27716 }, { "epoch": 0.11, "learning_rate": 3.139861351804185e-07, "loss": 0.8503, "step": 27717 }, { "epoch": 0.11, "learning_rate": 3.13718209808076e-07, "loss": 0.8618, "step": 27718 }, { "epoch": 0.11, "learning_rate": 3.134503969736635e-07, "loss": 0.811, "step": 27719 }, { "epoch": 0.11, "learning_rate": 3.13182696680292e-07, "loss": 0.7913, "step": 27720 }, { "epoch": 0.11, "learning_rate": 3.1291510893107335e-07, "loss": 0.8258, "step": 27721 }, { "epoch": 0.11, "learning_rate": 3.126476337291162e-07, "loss": 0.8452, "step": 27722 }, { "epoch": 0.11, "learning_rate": 3.123802710775259e-07, "loss": 0.8304, "step": 27723 }, { "epoch": 0.11, "learning_rate": 3.1211302097940874e-07, "loss": 0.8721, "step": 27724 }, { "epoch": 0.11, "learning_rate": 3.118458834378746e-07, "loss": 0.8436, "step": 27725 }, { "epoch": 0.11, "learning_rate": 3.1157885845602197e-07, "loss": 0.813, "step": 27726 }, { "epoch": 0.11, "learning_rate": 3.113119460369562e-07, "loss": 0.806, "step": 27727 }, { "epoch": 0.11, "learning_rate": 3.11045146183776e-07, "loss": 0.8731, "step": 27728 }, { "epoch": 0.11, "learning_rate": 3.1077845889958324e-07, "loss": 0.8352, "step": 27729 }, { "epoch": 0.11, "learning_rate": 3.1051188418747433e-07, "loss": 0.89, "step": 27730 }, { "epoch": 0.11, "learning_rate": 3.10245422050548e-07, "loss": 0.8858, "step": 27731 }, { "epoch": 0.11, "learning_rate": 3.099790724919005e-07, "loss": 0.7987, "step": 27732 }, { "epoch": 0.11, "learning_rate": 3.0971283551462503e-07, "loss": 0.8571, "step": 27733 }, { "epoch": 0.11, "learning_rate": 3.0944671112181466e-07, "loss": 0.8243, "step": 27734 }, { "epoch": 0.11, "learning_rate": 3.091806993165625e-07, "loss": 0.8846, "step": 27735 }, { "epoch": 0.11, "learning_rate": 3.089148001019582e-07, "loss": 0.8311, "step": 27736 }, { "epoch": 0.11, "learning_rate": 3.0864901348109155e-07, "loss": 0.8478, "step": 27737 }, { "epoch": 0.11, "learning_rate": 3.0838333945705014e-07, "loss": 0.8448, "step": 27738 }, { "epoch": 0.11, "learning_rate": 3.0811777803292255e-07, "loss": 0.8614, "step": 27739 }, { "epoch": 0.11, "learning_rate": 3.0785232921179077e-07, "loss": 0.8603, "step": 27740 }, { "epoch": 0.11, "learning_rate": 3.075869929967401e-07, "loss": 0.8117, "step": 27741 }, { "epoch": 0.11, "learning_rate": 3.0732176939085476e-07, "loss": 0.847, "step": 27742 }, { "epoch": 0.11, "learning_rate": 3.0705665839721456e-07, "loss": 0.8871, "step": 27743 }, { "epoch": 0.11, "learning_rate": 3.0679166001890027e-07, "loss": 0.8345, "step": 27744 }, { "epoch": 0.11, "learning_rate": 3.065267742589917e-07, "loss": 0.8045, "step": 27745 }, { "epoch": 0.11, "learning_rate": 3.0626200112056524e-07, "loss": 0.8424, "step": 27746 }, { "epoch": 0.11, "learning_rate": 3.059973406066963e-07, "loss": 0.9032, "step": 27747 }, { "epoch": 0.11, "learning_rate": 3.0573279272046006e-07, "loss": 0.8457, "step": 27748 }, { "epoch": 0.11, "learning_rate": 3.0546835746493196e-07, "loss": 0.8236, "step": 27749 }, { "epoch": 0.11, "learning_rate": 3.0520403484318395e-07, "loss": 0.8116, "step": 27750 }, { "epoch": 0.11, "learning_rate": 3.0493982485828464e-07, "loss": 0.8191, "step": 27751 }, { "epoch": 0.11, "learning_rate": 3.046757275133061e-07, "loss": 0.8138, "step": 27752 }, { "epoch": 0.11, "learning_rate": 3.044117428113147e-07, "loss": 0.8397, "step": 27753 }, { "epoch": 0.11, "learning_rate": 3.041478707553791e-07, "loss": 0.8433, "step": 27754 }, { "epoch": 0.11, "learning_rate": 3.038841113485647e-07, "loss": 0.8701, "step": 27755 }, { "epoch": 0.11, "learning_rate": 3.036204645939356e-07, "loss": 0.8526, "step": 27756 }, { "epoch": 0.11, "learning_rate": 3.0335693049455603e-07, "loss": 0.8305, "step": 27757 }, { "epoch": 0.11, "learning_rate": 3.030935090534859e-07, "loss": 0.8409, "step": 27758 }, { "epoch": 0.11, "learning_rate": 3.0283020027378705e-07, "loss": 0.8483, "step": 27759 }, { "epoch": 0.11, "learning_rate": 3.0256700415851827e-07, "loss": 0.8246, "step": 27760 }, { "epoch": 0.11, "learning_rate": 3.02303920710737e-07, "loss": 0.8528, "step": 27761 }, { "epoch": 0.11, "learning_rate": 3.0204094993349974e-07, "loss": 0.8473, "step": 27762 }, { "epoch": 0.11, "learning_rate": 3.0177809182986296e-07, "loss": 0.8252, "step": 27763 }, { "epoch": 0.11, "learning_rate": 3.015153464028808e-07, "loss": 0.8397, "step": 27764 }, { "epoch": 0.11, "learning_rate": 3.0125271365560426e-07, "loss": 0.8499, "step": 27765 }, { "epoch": 0.11, "learning_rate": 3.009901935910853e-07, "loss": 0.8486, "step": 27766 }, { "epoch": 0.11, "learning_rate": 3.0072778621237253e-07, "loss": 0.8708, "step": 27767 }, { "epoch": 0.11, "learning_rate": 3.0046549152251916e-07, "loss": 0.8253, "step": 27768 }, { "epoch": 0.11, "learning_rate": 3.0020330952456823e-07, "loss": 0.8206, "step": 27769 }, { "epoch": 0.11, "learning_rate": 2.999412402215673e-07, "loss": 0.8684, "step": 27770 }, { "epoch": 0.11, "learning_rate": 2.9967928361656295e-07, "loss": 0.848, "step": 27771 }, { "epoch": 0.11, "learning_rate": 2.9941743971259483e-07, "loss": 0.8663, "step": 27772 }, { "epoch": 0.11, "learning_rate": 2.991557085127073e-07, "loss": 0.856, "step": 27773 }, { "epoch": 0.11, "learning_rate": 2.9889409001994107e-07, "loss": 0.8331, "step": 27774 }, { "epoch": 0.11, "learning_rate": 2.9863258423733723e-07, "loss": 0.8603, "step": 27775 }, { "epoch": 0.11, "learning_rate": 2.9837119116793214e-07, "loss": 0.8465, "step": 27776 }, { "epoch": 0.11, "learning_rate": 2.9810991081476227e-07, "loss": 0.8126, "step": 27777 }, { "epoch": 0.11, "learning_rate": 2.978487431808641e-07, "loss": 0.8395, "step": 27778 }, { "epoch": 0.11, "learning_rate": 2.9758768826927186e-07, "loss": 0.834, "step": 27779 }, { "epoch": 0.11, "learning_rate": 2.973267460830187e-07, "loss": 0.8173, "step": 27780 }, { "epoch": 0.11, "learning_rate": 2.9706591662513663e-07, "loss": 0.8583, "step": 27781 }, { "epoch": 0.11, "learning_rate": 2.968051998986576e-07, "loss": 0.8574, "step": 27782 }, { "epoch": 0.11, "learning_rate": 2.96544595906606e-07, "loss": 0.8588, "step": 27783 }, { "epoch": 0.11, "learning_rate": 2.9628410465201374e-07, "loss": 0.82, "step": 27784 }, { "epoch": 0.11, "learning_rate": 2.960237261379051e-07, "loss": 0.8633, "step": 27785 }, { "epoch": 0.11, "learning_rate": 2.957634603673065e-07, "loss": 0.8218, "step": 27786 }, { "epoch": 0.11, "learning_rate": 2.955033073432412e-07, "loss": 0.8649, "step": 27787 }, { "epoch": 0.11, "learning_rate": 2.952432670687322e-07, "loss": 0.8263, "step": 27788 }, { "epoch": 0.11, "learning_rate": 2.949833395468005e-07, "loss": 0.8472, "step": 27789 }, { "epoch": 0.11, "learning_rate": 2.9472352478046585e-07, "loss": 0.8554, "step": 27790 }, { "epoch": 0.11, "learning_rate": 2.9446382277274477e-07, "loss": 0.872, "step": 27791 }, { "epoch": 0.11, "learning_rate": 2.942042335266604e-07, "loss": 0.8342, "step": 27792 }, { "epoch": 0.11, "learning_rate": 2.939447570452225e-07, "loss": 0.8366, "step": 27793 }, { "epoch": 0.11, "learning_rate": 2.9368539333144875e-07, "loss": 0.8215, "step": 27794 }, { "epoch": 0.11, "learning_rate": 2.934261423883522e-07, "loss": 0.7962, "step": 27795 }, { "epoch": 0.11, "learning_rate": 2.9316700421894495e-07, "loss": 0.8286, "step": 27796 }, { "epoch": 0.11, "learning_rate": 2.9290797882623566e-07, "loss": 0.8045, "step": 27797 }, { "epoch": 0.11, "learning_rate": 2.9264906621323754e-07, "loss": 0.855, "step": 27798 }, { "epoch": 0.11, "learning_rate": 2.923902663829559e-07, "loss": 0.8655, "step": 27799 }, { "epoch": 0.11, "learning_rate": 2.921315793383994e-07, "loss": 0.8041, "step": 27800 }, { "epoch": 0.11, "learning_rate": 2.9187300508257245e-07, "loss": 0.8489, "step": 27801 }, { "epoch": 0.11, "learning_rate": 2.9161454361847806e-07, "loss": 0.8023, "step": 27802 }, { "epoch": 0.11, "learning_rate": 2.913561949491217e-07, "loss": 0.8326, "step": 27803 }, { "epoch": 0.11, "learning_rate": 2.9109795907750205e-07, "loss": 0.8115, "step": 27804 }, { "epoch": 0.11, "learning_rate": 2.908398360066222e-07, "loss": 0.8535, "step": 27805 }, { "epoch": 0.11, "learning_rate": 2.905818257394799e-07, "loss": 0.7982, "step": 27806 }, { "epoch": 0.11, "learning_rate": 2.903239282790726e-07, "loss": 0.8319, "step": 27807 }, { "epoch": 0.11, "learning_rate": 2.900661436283969e-07, "loss": 0.8061, "step": 27808 }, { "epoch": 0.11, "learning_rate": 2.89808471790447e-07, "loss": 0.8134, "step": 27809 }, { "epoch": 0.11, "learning_rate": 2.8955091276821724e-07, "loss": 0.8268, "step": 27810 }, { "epoch": 0.11, "learning_rate": 2.892934665647007e-07, "loss": 0.8965, "step": 27811 }, { "epoch": 0.11, "learning_rate": 2.890361331828873e-07, "loss": 0.834, "step": 27812 }, { "epoch": 0.11, "learning_rate": 2.8877891262576676e-07, "loss": 0.8664, "step": 27813 }, { "epoch": 0.11, "learning_rate": 2.885218048963301e-07, "loss": 0.8488, "step": 27814 }, { "epoch": 0.11, "learning_rate": 2.8826480999756043e-07, "loss": 0.8653, "step": 27815 }, { "epoch": 0.11, "learning_rate": 2.8800792793244547e-07, "loss": 0.8109, "step": 27816 }, { "epoch": 0.11, "learning_rate": 2.877511587039716e-07, "loss": 0.8716, "step": 27817 }, { "epoch": 0.11, "learning_rate": 2.8749450231511877e-07, "loss": 0.841, "step": 27818 }, { "epoch": 0.11, "learning_rate": 2.8723795876887116e-07, "loss": 0.8391, "step": 27819 }, { "epoch": 0.11, "learning_rate": 2.869815280682087e-07, "loss": 0.8291, "step": 27820 }, { "epoch": 0.11, "learning_rate": 2.8672521021611e-07, "loss": 0.8356, "step": 27821 }, { "epoch": 0.11, "learning_rate": 2.864690052155539e-07, "loss": 0.8179, "step": 27822 }, { "epoch": 0.11, "learning_rate": 2.8621291306951684e-07, "loss": 0.8459, "step": 27823 }, { "epoch": 0.11, "learning_rate": 2.8595693378097533e-07, "loss": 0.8693, "step": 27824 }, { "epoch": 0.11, "learning_rate": 2.857010673529015e-07, "loss": 0.8771, "step": 27825 }, { "epoch": 0.11, "learning_rate": 2.8544531378826733e-07, "loss": 0.8498, "step": 27826 }, { "epoch": 0.11, "learning_rate": 2.851896730900472e-07, "loss": 0.8487, "step": 27827 }, { "epoch": 0.11, "learning_rate": 2.8493414526120977e-07, "loss": 0.8258, "step": 27828 }, { "epoch": 0.11, "learning_rate": 2.8467873030472273e-07, "loss": 0.8518, "step": 27829 }, { "epoch": 0.11, "learning_rate": 2.8442342822355587e-07, "loss": 0.8649, "step": 27830 }, { "epoch": 0.11, "learning_rate": 2.841682390206735e-07, "loss": 0.8507, "step": 27831 }, { "epoch": 0.11, "learning_rate": 2.8391316269904103e-07, "loss": 0.8304, "step": 27832 }, { "epoch": 0.11, "learning_rate": 2.8365819926162277e-07, "loss": 0.8006, "step": 27833 }, { "epoch": 0.11, "learning_rate": 2.8340334871137965e-07, "loss": 0.8721, "step": 27834 }, { "epoch": 0.11, "learning_rate": 2.831486110512727e-07, "loss": 0.8346, "step": 27835 }, { "epoch": 0.11, "learning_rate": 2.828939862842628e-07, "loss": 0.8133, "step": 27836 }, { "epoch": 0.11, "learning_rate": 2.8263947441330654e-07, "loss": 0.8438, "step": 27837 }, { "epoch": 0.11, "learning_rate": 2.8238507544136373e-07, "loss": 0.8412, "step": 27838 }, { "epoch": 0.11, "learning_rate": 2.8213078937138647e-07, "loss": 0.8366, "step": 27839 }, { "epoch": 0.11, "learning_rate": 2.818766162063302e-07, "loss": 0.8672, "step": 27840 }, { "epoch": 0.11, "learning_rate": 2.8162255594914923e-07, "loss": 0.8328, "step": 27841 }, { "epoch": 0.11, "learning_rate": 2.813686086027967e-07, "loss": 0.8501, "step": 27842 }, { "epoch": 0.11, "learning_rate": 2.8111477417021917e-07, "loss": 0.8507, "step": 27843 }, { "epoch": 0.11, "learning_rate": 2.8086105265436757e-07, "loss": 0.7918, "step": 27844 }, { "epoch": 0.11, "learning_rate": 2.8060744405819076e-07, "loss": 0.8546, "step": 27845 }, { "epoch": 0.11, "learning_rate": 2.803539483846329e-07, "loss": 0.8592, "step": 27846 }, { "epoch": 0.11, "learning_rate": 2.8010056563664067e-07, "loss": 0.8496, "step": 27847 }, { "epoch": 0.11, "learning_rate": 2.798472958171583e-07, "loss": 0.843, "step": 27848 }, { "epoch": 0.11, "learning_rate": 2.7959413892912903e-07, "loss": 0.8573, "step": 27849 }, { "epoch": 0.11, "learning_rate": 2.7934109497549154e-07, "loss": 0.8293, "step": 27850 }, { "epoch": 0.11, "learning_rate": 2.7908816395918694e-07, "loss": 0.8506, "step": 27851 }, { "epoch": 0.11, "learning_rate": 2.7883534588315385e-07, "loss": 0.8271, "step": 27852 }, { "epoch": 0.11, "learning_rate": 2.7858264075033005e-07, "loss": 0.8443, "step": 27853 }, { "epoch": 0.11, "learning_rate": 2.78330048563652e-07, "loss": 0.8778, "step": 27854 }, { "epoch": 0.11, "learning_rate": 2.780775693260529e-07, "loss": 0.854, "step": 27855 }, { "epoch": 0.11, "learning_rate": 2.778252030404671e-07, "loss": 0.8093, "step": 27856 }, { "epoch": 0.11, "learning_rate": 2.7757294970982676e-07, "loss": 0.8557, "step": 27857 }, { "epoch": 0.11, "learning_rate": 2.7732080933706164e-07, "loss": 0.8235, "step": 27858 }, { "epoch": 0.11, "learning_rate": 2.7706878192510054e-07, "loss": 0.8411, "step": 27859 }, { "epoch": 0.11, "learning_rate": 2.7681686747687566e-07, "loss": 0.8313, "step": 27860 }, { "epoch": 0.11, "learning_rate": 2.7656506599530896e-07, "loss": 0.8197, "step": 27861 }, { "epoch": 0.11, "learning_rate": 2.763133774833293e-07, "loss": 0.839, "step": 27862 }, { "epoch": 0.11, "learning_rate": 2.760618019438599e-07, "loss": 0.8677, "step": 27863 }, { "epoch": 0.11, "learning_rate": 2.758103393798206e-07, "loss": 0.7828, "step": 27864 }, { "epoch": 0.11, "learning_rate": 2.7555898979413796e-07, "loss": 0.8529, "step": 27865 }, { "epoch": 0.11, "learning_rate": 2.753077531897297e-07, "loss": 0.8532, "step": 27866 }, { "epoch": 0.11, "learning_rate": 2.750566295695156e-07, "loss": 0.8067, "step": 27867 }, { "epoch": 0.11, "learning_rate": 2.748056189364112e-07, "loss": 0.8472, "step": 27868 }, { "epoch": 0.11, "learning_rate": 2.7455472129333524e-07, "loss": 0.8089, "step": 27869 }, { "epoch": 0.11, "learning_rate": 2.74303936643201e-07, "loss": 0.8095, "step": 27870 }, { "epoch": 0.11, "learning_rate": 2.740532649889238e-07, "loss": 0.8804, "step": 27871 }, { "epoch": 0.11, "learning_rate": 2.7380270633341477e-07, "loss": 0.8894, "step": 27872 }, { "epoch": 0.11, "learning_rate": 2.7355226067958486e-07, "loss": 0.8142, "step": 27873 }, { "epoch": 0.11, "learning_rate": 2.7330192803034617e-07, "loss": 0.8129, "step": 27874 }, { "epoch": 0.11, "learning_rate": 2.730517083886042e-07, "loss": 0.8269, "step": 27875 }, { "epoch": 0.11, "learning_rate": 2.7280160175726656e-07, "loss": 0.8232, "step": 27876 }, { "epoch": 0.11, "learning_rate": 2.725516081392399e-07, "loss": 0.819, "step": 27877 }, { "epoch": 0.11, "learning_rate": 2.7230172753742847e-07, "loss": 0.8781, "step": 27878 }, { "epoch": 0.11, "learning_rate": 2.7205195995473553e-07, "loss": 0.8551, "step": 27879 }, { "epoch": 0.11, "learning_rate": 2.7180230539406216e-07, "loss": 0.8047, "step": 27880 }, { "epoch": 0.11, "learning_rate": 2.7155276385831155e-07, "loss": 0.8471, "step": 27881 }, { "epoch": 0.11, "learning_rate": 2.7130333535038024e-07, "loss": 0.8429, "step": 27882 }, { "epoch": 0.11, "learning_rate": 2.710540198731648e-07, "loss": 0.882, "step": 27883 }, { "epoch": 0.11, "learning_rate": 2.708048174295663e-07, "loss": 0.8076, "step": 27884 }, { "epoch": 0.11, "learning_rate": 2.705557280224769e-07, "loss": 0.8553, "step": 27885 }, { "epoch": 0.11, "learning_rate": 2.7030675165479195e-07, "loss": 0.7851, "step": 27886 }, { "epoch": 0.11, "learning_rate": 2.700578883294036e-07, "loss": 0.8584, "step": 27887 }, { "epoch": 0.11, "learning_rate": 2.6980913804920293e-07, "loss": 0.8534, "step": 27888 }, { "epoch": 0.11, "learning_rate": 2.6956050081707985e-07, "loss": 0.8149, "step": 27889 }, { "epoch": 0.11, "learning_rate": 2.693119766359231e-07, "loss": 0.8424, "step": 27890 }, { "epoch": 0.11, "learning_rate": 2.6906356550862154e-07, "loss": 0.8766, "step": 27891 }, { "epoch": 0.11, "learning_rate": 2.6881526743806064e-07, "loss": 0.8297, "step": 27892 }, { "epoch": 0.11, "learning_rate": 2.685670824271247e-07, "loss": 0.8145, "step": 27893 }, { "epoch": 0.11, "learning_rate": 2.683190104786959e-07, "loss": 0.8347, "step": 27894 }, { "epoch": 0.11, "learning_rate": 2.680710515956586e-07, "loss": 0.8526, "step": 27895 }, { "epoch": 0.11, "learning_rate": 2.678232057808927e-07, "loss": 0.8339, "step": 27896 }, { "epoch": 0.11, "learning_rate": 2.6757547303727814e-07, "loss": 0.8508, "step": 27897 }, { "epoch": 0.11, "learning_rate": 2.673278533676937e-07, "loss": 0.8674, "step": 27898 }, { "epoch": 0.11, "learning_rate": 2.6708034677501494e-07, "loss": 0.8605, "step": 27899 }, { "epoch": 0.11, "learning_rate": 2.6683295326211835e-07, "loss": 0.8804, "step": 27900 }, { "epoch": 0.11, "learning_rate": 2.665856728318772e-07, "loss": 0.8208, "step": 27901 }, { "epoch": 0.11, "learning_rate": 2.663385054871659e-07, "loss": 0.8394, "step": 27902 }, { "epoch": 0.11, "learning_rate": 2.660914512308543e-07, "loss": 0.9113, "step": 27903 }, { "epoch": 0.11, "learning_rate": 2.658445100658147e-07, "loss": 0.8889, "step": 27904 }, { "epoch": 0.11, "learning_rate": 2.6559768199491463e-07, "loss": 0.8133, "step": 27905 }, { "epoch": 0.11, "learning_rate": 2.6535096702102414e-07, "loss": 0.8416, "step": 27906 }, { "epoch": 0.11, "learning_rate": 2.6510436514700645e-07, "loss": 0.8508, "step": 27907 }, { "epoch": 0.11, "learning_rate": 2.64857876375727e-07, "loss": 0.854, "step": 27908 }, { "epoch": 0.11, "learning_rate": 2.646115007100536e-07, "loss": 0.8351, "step": 27909 }, { "epoch": 0.11, "learning_rate": 2.643652381528439e-07, "loss": 0.8564, "step": 27910 }, { "epoch": 0.11, "learning_rate": 2.641190887069611e-07, "loss": 0.8474, "step": 27911 }, { "epoch": 0.11, "learning_rate": 2.638730523752653e-07, "loss": 0.8205, "step": 27912 }, { "epoch": 0.11, "learning_rate": 2.6362712916061407e-07, "loss": 0.8646, "step": 27913 }, { "epoch": 0.11, "learning_rate": 2.6338131906586516e-07, "loss": 0.8679, "step": 27914 }, { "epoch": 0.11, "learning_rate": 2.6313562209387525e-07, "loss": 0.8847, "step": 27915 }, { "epoch": 0.11, "learning_rate": 2.628900382474986e-07, "loss": 0.8327, "step": 27916 }, { "epoch": 0.11, "learning_rate": 2.6264456752958636e-07, "loss": 0.839, "step": 27917 }, { "epoch": 0.11, "learning_rate": 2.6239920994299284e-07, "loss": 0.8277, "step": 27918 }, { "epoch": 0.11, "learning_rate": 2.6215396549056806e-07, "loss": 0.8209, "step": 27919 }, { "epoch": 0.11, "learning_rate": 2.619088341751608e-07, "loss": 0.8429, "step": 27920 }, { "epoch": 0.11, "learning_rate": 2.616638159996199e-07, "loss": 0.8249, "step": 27921 }, { "epoch": 0.11, "learning_rate": 2.6141891096679196e-07, "loss": 0.8507, "step": 27922 }, { "epoch": 0.11, "learning_rate": 2.611741190795225e-07, "loss": 0.8629, "step": 27923 }, { "epoch": 0.11, "learning_rate": 2.609294403406537e-07, "loss": 0.8193, "step": 27924 }, { "epoch": 0.11, "learning_rate": 2.6068487475302995e-07, "loss": 0.8047, "step": 27925 }, { "epoch": 0.11, "learning_rate": 2.6044042231949116e-07, "loss": 0.8235, "step": 27926 }, { "epoch": 0.11, "learning_rate": 2.601960830428818e-07, "loss": 0.9054, "step": 27927 }, { "epoch": 0.11, "learning_rate": 2.599518569260362e-07, "loss": 0.7958, "step": 27928 }, { "epoch": 0.11, "learning_rate": 2.5970774397179323e-07, "loss": 0.8396, "step": 27929 }, { "epoch": 0.11, "learning_rate": 2.5946374418298945e-07, "loss": 0.8341, "step": 27930 }, { "epoch": 0.11, "learning_rate": 2.592198575624594e-07, "loss": 0.8327, "step": 27931 }, { "epoch": 0.11, "learning_rate": 2.58976084113034e-07, "loss": 0.8235, "step": 27932 }, { "epoch": 0.11, "learning_rate": 2.5873242383755104e-07, "loss": 0.8532, "step": 27933 }, { "epoch": 0.11, "learning_rate": 2.584888767388383e-07, "loss": 0.8474, "step": 27934 }, { "epoch": 0.11, "learning_rate": 2.582454428197245e-07, "loss": 0.8538, "step": 27935 }, { "epoch": 0.11, "learning_rate": 2.5800212208303864e-07, "loss": 0.8083, "step": 27936 }, { "epoch": 0.11, "learning_rate": 2.577589145316084e-07, "loss": 0.8095, "step": 27937 }, { "epoch": 0.11, "learning_rate": 2.5751582016825925e-07, "loss": 0.8225, "step": 27938 }, { "epoch": 0.11, "learning_rate": 2.5727283899581566e-07, "loss": 0.8406, "step": 27939 }, { "epoch": 0.11, "learning_rate": 2.570299710170998e-07, "loss": 0.8187, "step": 27940 }, { "epoch": 0.11, "learning_rate": 2.56787216234935e-07, "loss": 0.8159, "step": 27941 }, { "epoch": 0.11, "learning_rate": 2.565445746521389e-07, "loss": 0.8279, "step": 27942 }, { "epoch": 0.11, "learning_rate": 2.5630204627153377e-07, "loss": 0.8502, "step": 27943 }, { "epoch": 0.11, "learning_rate": 2.56059631095934e-07, "loss": 0.82, "step": 27944 }, { "epoch": 0.11, "learning_rate": 2.558173291281585e-07, "loss": 0.8674, "step": 27945 }, { "epoch": 0.11, "learning_rate": 2.5557514037102273e-07, "loss": 0.8373, "step": 27946 }, { "epoch": 0.11, "learning_rate": 2.553330648273389e-07, "loss": 0.911, "step": 27947 }, { "epoch": 0.11, "learning_rate": 2.550911024999203e-07, "loss": 0.8489, "step": 27948 }, { "epoch": 0.11, "learning_rate": 2.5484925339157806e-07, "loss": 0.8377, "step": 27949 }, { "epoch": 0.11, "learning_rate": 2.54607517505121e-07, "loss": 0.8466, "step": 27950 }, { "epoch": 0.11, "learning_rate": 2.5436589484335804e-07, "loss": 0.8775, "step": 27951 }, { "epoch": 0.11, "learning_rate": 2.541243854090991e-07, "loss": 0.8255, "step": 27952 }, { "epoch": 0.11, "learning_rate": 2.5388298920514755e-07, "loss": 0.84, "step": 27953 }, { "epoch": 0.11, "learning_rate": 2.536417062343077e-07, "loss": 0.8007, "step": 27954 }, { "epoch": 0.11, "learning_rate": 2.5340053649938523e-07, "loss": 0.8489, "step": 27955 }, { "epoch": 0.12, "learning_rate": 2.531594800031778e-07, "loss": 0.8314, "step": 27956 }, { "epoch": 0.12, "learning_rate": 2.529185367484899e-07, "loss": 0.8322, "step": 27957 }, { "epoch": 0.12, "learning_rate": 2.526777067381203e-07, "loss": 0.8537, "step": 27958 }, { "epoch": 0.12, "learning_rate": 2.524369899748669e-07, "loss": 0.8936, "step": 27959 }, { "epoch": 0.12, "learning_rate": 2.5219638646152513e-07, "loss": 0.7938, "step": 27960 }, { "epoch": 0.12, "learning_rate": 2.519558962008917e-07, "loss": 0.8524, "step": 27961 }, { "epoch": 0.12, "learning_rate": 2.517155191957599e-07, "loss": 0.8858, "step": 27962 }, { "epoch": 0.12, "learning_rate": 2.514752554489219e-07, "loss": 0.8417, "step": 27963 }, { "epoch": 0.12, "learning_rate": 2.5123510496317116e-07, "loss": 0.8105, "step": 27964 }, { "epoch": 0.12, "learning_rate": 2.5099506774129646e-07, "loss": 0.807, "step": 27965 }, { "epoch": 0.12, "learning_rate": 2.507551437860878e-07, "loss": 0.8366, "step": 27966 }, { "epoch": 0.12, "learning_rate": 2.505153331003307e-07, "loss": 0.8718, "step": 27967 }, { "epoch": 0.12, "learning_rate": 2.50275635686813e-07, "loss": 0.8027, "step": 27968 }, { "epoch": 0.12, "learning_rate": 2.50036051548318e-07, "loss": 0.8711, "step": 27969 }, { "epoch": 0.12, "learning_rate": 2.4979658068763125e-07, "loss": 0.8242, "step": 27970 }, { "epoch": 0.12, "learning_rate": 2.4955722310753385e-07, "loss": 0.8603, "step": 27971 }, { "epoch": 0.12, "learning_rate": 2.4931797881080687e-07, "loss": 0.806, "step": 27972 }, { "epoch": 0.12, "learning_rate": 2.490788478002304e-07, "loss": 0.8739, "step": 27973 }, { "epoch": 0.12, "learning_rate": 2.48839830078581e-07, "loss": 0.8628, "step": 27974 }, { "epoch": 0.12, "learning_rate": 2.4860092564863656e-07, "loss": 0.8024, "step": 27975 }, { "epoch": 0.12, "learning_rate": 2.483621345131748e-07, "loss": 0.8066, "step": 27976 }, { "epoch": 0.12, "learning_rate": 2.4812345667496796e-07, "loss": 0.8383, "step": 27977 }, { "epoch": 0.12, "learning_rate": 2.4788489213678934e-07, "loss": 0.8412, "step": 27978 }, { "epoch": 0.12, "learning_rate": 2.476464409014101e-07, "loss": 0.8532, "step": 27979 }, { "epoch": 0.12, "learning_rate": 2.4740810297160244e-07, "loss": 0.8477, "step": 27980 }, { "epoch": 0.12, "learning_rate": 2.4716987835013196e-07, "loss": 0.8224, "step": 27981 }, { "epoch": 0.12, "learning_rate": 2.4693176703977085e-07, "loss": 0.7998, "step": 27982 }, { "epoch": 0.12, "learning_rate": 2.4669376904328244e-07, "loss": 0.8699, "step": 27983 }, { "epoch": 0.12, "learning_rate": 2.4645588436343346e-07, "loss": 0.7754, "step": 27984 }, { "epoch": 0.12, "learning_rate": 2.4621811300298725e-07, "loss": 0.8393, "step": 27985 }, { "epoch": 0.12, "learning_rate": 2.4598045496470604e-07, "loss": 0.8021, "step": 27986 }, { "epoch": 0.12, "learning_rate": 2.457429102513509e-07, "loss": 0.8682, "step": 27987 }, { "epoch": 0.12, "learning_rate": 2.455054788656819e-07, "loss": 0.8342, "step": 27988 }, { "epoch": 0.12, "learning_rate": 2.452681608104568e-07, "loss": 0.8721, "step": 27989 }, { "epoch": 0.12, "learning_rate": 2.450309560884345e-07, "loss": 0.8515, "step": 27990 }, { "epoch": 0.12, "learning_rate": 2.447938647023718e-07, "loss": 0.8474, "step": 27991 }, { "epoch": 0.12, "learning_rate": 2.445568866550196e-07, "loss": 0.8313, "step": 27992 }, { "epoch": 0.12, "learning_rate": 2.4432002194913263e-07, "loss": 0.8528, "step": 27993 }, { "epoch": 0.12, "learning_rate": 2.4408327058746404e-07, "loss": 0.8552, "step": 27994 }, { "epoch": 0.12, "learning_rate": 2.4384663257276396e-07, "loss": 0.8662, "step": 27995 }, { "epoch": 0.12, "learning_rate": 2.4361010790778126e-07, "loss": 0.8208, "step": 27996 }, { "epoch": 0.12, "learning_rate": 2.433736965952638e-07, "loss": 0.8171, "step": 27997 }, { "epoch": 0.12, "learning_rate": 2.4313739863796036e-07, "loss": 0.8336, "step": 27998 }, { "epoch": 0.12, "learning_rate": 2.429012140386122e-07, "loss": 0.8681, "step": 27999 }, { "epoch": 0.12, "learning_rate": 2.4266514279996713e-07, "loss": 0.8237, "step": 28000 }, { "epoch": 0.12, "learning_rate": 2.424291849247673e-07, "loss": 0.8543, "step": 28001 }, { "epoch": 0.12, "learning_rate": 2.421933404157528e-07, "loss": 0.8319, "step": 28002 }, { "epoch": 0.12, "learning_rate": 2.419576092756648e-07, "loss": 0.8696, "step": 28003 }, { "epoch": 0.12, "learning_rate": 2.4172199150724107e-07, "loss": 0.8109, "step": 28004 }, { "epoch": 0.12, "learning_rate": 2.4148648711321944e-07, "loss": 0.8783, "step": 28005 }, { "epoch": 0.12, "learning_rate": 2.412510960963366e-07, "loss": 0.8273, "step": 28006 }, { "epoch": 0.12, "learning_rate": 2.4101581845932696e-07, "loss": 0.8722, "step": 28007 }, { "epoch": 0.12, "learning_rate": 2.407806542049251e-07, "loss": 0.8746, "step": 28008 }, { "epoch": 0.12, "learning_rate": 2.4054560333586106e-07, "loss": 0.8554, "step": 28009 }, { "epoch": 0.12, "learning_rate": 2.4031066585486704e-07, "loss": 0.8774, "step": 28010 }, { "epoch": 0.12, "learning_rate": 2.4007584176467315e-07, "loss": 0.8662, "step": 28011 }, { "epoch": 0.12, "learning_rate": 2.398411310680071e-07, "loss": 0.8631, "step": 28012 }, { "epoch": 0.12, "learning_rate": 2.396065337675946e-07, "loss": 0.8439, "step": 28013 }, { "epoch": 0.12, "learning_rate": 2.3937204986616337e-07, "loss": 0.8491, "step": 28014 }, { "epoch": 0.12, "learning_rate": 2.3913767936643795e-07, "loss": 0.8048, "step": 28015 }, { "epoch": 0.12, "learning_rate": 2.389034222711384e-07, "loss": 0.8939, "step": 28016 }, { "epoch": 0.12, "learning_rate": 2.3866927858298804e-07, "loss": 0.8554, "step": 28017 }, { "epoch": 0.12, "learning_rate": 2.38435248304707e-07, "loss": 0.8545, "step": 28018 }, { "epoch": 0.12, "learning_rate": 2.3820133143901526e-07, "loss": 0.8353, "step": 28019 }, { "epoch": 0.12, "learning_rate": 2.3796752798863066e-07, "loss": 0.7704, "step": 28020 }, { "epoch": 0.12, "learning_rate": 2.3773383795626769e-07, "loss": 0.862, "step": 28021 }, { "epoch": 0.12, "learning_rate": 2.375002613446431e-07, "loss": 0.8692, "step": 28022 }, { "epoch": 0.12, "learning_rate": 2.3726679815647024e-07, "loss": 0.8525, "step": 28023 }, { "epoch": 0.12, "learning_rate": 2.370334483944603e-07, "loss": 0.8564, "step": 28024 }, { "epoch": 0.12, "learning_rate": 2.3680021206132553e-07, "loss": 0.8474, "step": 28025 }, { "epoch": 0.12, "learning_rate": 2.365670891597771e-07, "loss": 0.7875, "step": 28026 }, { "epoch": 0.12, "learning_rate": 2.3633407969252175e-07, "loss": 0.85, "step": 28027 }, { "epoch": 0.12, "learning_rate": 2.361011836622662e-07, "loss": 0.8371, "step": 28028 }, { "epoch": 0.12, "learning_rate": 2.3586840107171715e-07, "loss": 0.8578, "step": 28029 }, { "epoch": 0.12, "learning_rate": 2.3563573192357914e-07, "loss": 0.8954, "step": 28030 }, { "epoch": 0.12, "learning_rate": 2.3540317622055553e-07, "loss": 0.8179, "step": 28031 }, { "epoch": 0.12, "learning_rate": 2.3517073396534862e-07, "loss": 0.8257, "step": 28032 }, { "epoch": 0.12, "learning_rate": 2.3493840516065846e-07, "loss": 0.8508, "step": 28033 }, { "epoch": 0.12, "learning_rate": 2.3470618980918403e-07, "loss": 0.8375, "step": 28034 }, { "epoch": 0.12, "learning_rate": 2.3447408791362203e-07, "loss": 0.8763, "step": 28035 }, { "epoch": 0.12, "learning_rate": 2.3424209947667253e-07, "loss": 0.8264, "step": 28036 }, { "epoch": 0.12, "learning_rate": 2.3401022450102672e-07, "loss": 0.8254, "step": 28037 }, { "epoch": 0.12, "learning_rate": 2.3377846298938244e-07, "loss": 0.857, "step": 28038 }, { "epoch": 0.12, "learning_rate": 2.3354681494442978e-07, "loss": 0.8483, "step": 28039 }, { "epoch": 0.12, "learning_rate": 2.3331528036886207e-07, "loss": 0.8209, "step": 28040 }, { "epoch": 0.12, "learning_rate": 2.3308385926536726e-07, "loss": 0.8268, "step": 28041 }, { "epoch": 0.12, "learning_rate": 2.3285255163663535e-07, "loss": 0.858, "step": 28042 }, { "epoch": 0.12, "learning_rate": 2.3262135748535197e-07, "loss": 0.8322, "step": 28043 }, { "epoch": 0.12, "learning_rate": 2.323902768142061e-07, "loss": 0.8229, "step": 28044 }, { "epoch": 0.12, "learning_rate": 2.3215930962588118e-07, "loss": 0.8559, "step": 28045 }, { "epoch": 0.12, "learning_rate": 2.3192845592305947e-07, "loss": 0.8727, "step": 28046 }, { "epoch": 0.12, "learning_rate": 2.3169771570842547e-07, "loss": 0.8427, "step": 28047 }, { "epoch": 0.12, "learning_rate": 2.3146708898465708e-07, "loss": 0.7926, "step": 28048 }, { "epoch": 0.12, "learning_rate": 2.312365757544366e-07, "loss": 0.8147, "step": 28049 }, { "epoch": 0.12, "learning_rate": 2.3100617602043962e-07, "loss": 0.8451, "step": 28050 }, { "epoch": 0.12, "learning_rate": 2.3077588978534626e-07, "loss": 0.846, "step": 28051 }, { "epoch": 0.12, "learning_rate": 2.3054571705182882e-07, "loss": 0.8626, "step": 28052 }, { "epoch": 0.12, "learning_rate": 2.3031565782256294e-07, "loss": 0.857, "step": 28053 }, { "epoch": 0.12, "learning_rate": 2.300857121002209e-07, "loss": 0.8309, "step": 28054 }, { "epoch": 0.12, "learning_rate": 2.2985587988747504e-07, "loss": 0.8183, "step": 28055 }, { "epoch": 0.12, "learning_rate": 2.2962616118699544e-07, "loss": 0.7988, "step": 28056 }, { "epoch": 0.12, "learning_rate": 2.2939655600145105e-07, "loss": 0.8362, "step": 28057 }, { "epoch": 0.12, "learning_rate": 2.2916706433351088e-07, "loss": 0.845, "step": 28058 }, { "epoch": 0.12, "learning_rate": 2.2893768618583724e-07, "loss": 0.8504, "step": 28059 }, { "epoch": 0.12, "learning_rate": 2.2870842156109906e-07, "loss": 0.8468, "step": 28060 }, { "epoch": 0.12, "learning_rate": 2.2847927046195763e-07, "loss": 0.8272, "step": 28061 }, { "epoch": 0.12, "learning_rate": 2.282502328910763e-07, "loss": 0.8708, "step": 28062 }, { "epoch": 0.12, "learning_rate": 2.2802130885111518e-07, "loss": 0.8795, "step": 28063 }, { "epoch": 0.12, "learning_rate": 2.277924983447355e-07, "loss": 0.8246, "step": 28064 }, { "epoch": 0.12, "learning_rate": 2.2756380137459622e-07, "loss": 0.8495, "step": 28065 }, { "epoch": 0.12, "learning_rate": 2.273352179433519e-07, "loss": 0.8552, "step": 28066 }, { "epoch": 0.12, "learning_rate": 2.2710674805365817e-07, "loss": 0.8259, "step": 28067 }, { "epoch": 0.12, "learning_rate": 2.2687839170817184e-07, "loss": 0.7942, "step": 28068 }, { "epoch": 0.12, "learning_rate": 2.266501489095463e-07, "loss": 0.8216, "step": 28069 }, { "epoch": 0.12, "learning_rate": 2.2642201966043053e-07, "loss": 0.8796, "step": 28070 }, { "epoch": 0.12, "learning_rate": 2.261940039634769e-07, "loss": 0.8749, "step": 28071 }, { "epoch": 0.12, "learning_rate": 2.2596610182133328e-07, "loss": 0.8299, "step": 28072 }, { "epoch": 0.12, "learning_rate": 2.2573831323664862e-07, "loss": 0.8346, "step": 28073 }, { "epoch": 0.12, "learning_rate": 2.2551063821206976e-07, "loss": 0.8268, "step": 28074 }, { "epoch": 0.12, "learning_rate": 2.252830767502412e-07, "loss": 0.8307, "step": 28075 }, { "epoch": 0.12, "learning_rate": 2.250556288538075e-07, "loss": 0.8183, "step": 28076 }, { "epoch": 0.12, "learning_rate": 2.248282945254099e-07, "loss": 0.8329, "step": 28077 }, { "epoch": 0.12, "learning_rate": 2.2460107376768959e-07, "loss": 0.8588, "step": 28078 }, { "epoch": 0.12, "learning_rate": 2.243739665832878e-07, "loss": 0.8415, "step": 28079 }, { "epoch": 0.12, "learning_rate": 2.2414697297484244e-07, "loss": 0.8132, "step": 28080 }, { "epoch": 0.12, "learning_rate": 2.239200929449903e-07, "loss": 0.8093, "step": 28081 }, { "epoch": 0.12, "learning_rate": 2.236933264963692e-07, "loss": 0.871, "step": 28082 }, { "epoch": 0.12, "learning_rate": 2.2346667363161267e-07, "loss": 0.8157, "step": 28083 }, { "epoch": 0.12, "learning_rate": 2.2324013435335302e-07, "loss": 0.8711, "step": 28084 }, { "epoch": 0.12, "learning_rate": 2.2301370866422255e-07, "loss": 0.8565, "step": 28085 }, { "epoch": 0.12, "learning_rate": 2.2278739656685256e-07, "loss": 0.8502, "step": 28086 }, { "epoch": 0.12, "learning_rate": 2.2256119806387312e-07, "loss": 0.9096, "step": 28087 }, { "epoch": 0.12, "learning_rate": 2.22335113157911e-07, "loss": 0.863, "step": 28088 }, { "epoch": 0.12, "learning_rate": 2.2210914185159417e-07, "loss": 0.8815, "step": 28089 }, { "epoch": 0.12, "learning_rate": 2.218832841475471e-07, "loss": 0.8623, "step": 28090 }, { "epoch": 0.12, "learning_rate": 2.2165754004839225e-07, "loss": 0.8259, "step": 28091 }, { "epoch": 0.12, "learning_rate": 2.214319095567552e-07, "loss": 0.8193, "step": 28092 }, { "epoch": 0.12, "learning_rate": 2.2120639267525723e-07, "loss": 0.8465, "step": 28093 }, { "epoch": 0.12, "learning_rate": 2.2098098940651734e-07, "loss": 0.8382, "step": 28094 }, { "epoch": 0.12, "learning_rate": 2.2075569975315347e-07, "loss": 0.8648, "step": 28095 }, { "epoch": 0.12, "learning_rate": 2.205305237177846e-07, "loss": 0.8092, "step": 28096 }, { "epoch": 0.12, "learning_rate": 2.2030546130302643e-07, "loss": 0.8343, "step": 28097 }, { "epoch": 0.12, "learning_rate": 2.2008051251149464e-07, "loss": 0.8459, "step": 28098 }, { "epoch": 0.12, "learning_rate": 2.1985567734580049e-07, "loss": 0.8409, "step": 28099 }, { "epoch": 0.12, "learning_rate": 2.1963095580855964e-07, "loss": 0.8558, "step": 28100 }, { "epoch": 0.12, "learning_rate": 2.1940634790238003e-07, "loss": 0.843, "step": 28101 }, { "epoch": 0.12, "learning_rate": 2.1918185362987176e-07, "loss": 0.8466, "step": 28102 }, { "epoch": 0.12, "learning_rate": 2.189574729936428e-07, "loss": 0.8189, "step": 28103 }, { "epoch": 0.12, "learning_rate": 2.18733205996301e-07, "loss": 0.866, "step": 28104 }, { "epoch": 0.12, "learning_rate": 2.1850905264045208e-07, "loss": 0.8552, "step": 28105 }, { "epoch": 0.12, "learning_rate": 2.1828501292870062e-07, "loss": 0.8384, "step": 28106 }, { "epoch": 0.12, "learning_rate": 2.18061086863649e-07, "loss": 0.8447, "step": 28107 }, { "epoch": 0.12, "learning_rate": 2.1783727444789737e-07, "loss": 0.8398, "step": 28108 }, { "epoch": 0.12, "learning_rate": 2.1761357568404806e-07, "loss": 0.8247, "step": 28109 }, { "epoch": 0.12, "learning_rate": 2.1738999057469789e-07, "loss": 0.8165, "step": 28110 }, { "epoch": 0.12, "learning_rate": 2.1716651912244812e-07, "loss": 0.8509, "step": 28111 }, { "epoch": 0.12, "learning_rate": 2.1694316132989224e-07, "loss": 0.8769, "step": 28112 }, { "epoch": 0.12, "learning_rate": 2.1671991719962593e-07, "loss": 0.8602, "step": 28113 }, { "epoch": 0.12, "learning_rate": 2.164967867342438e-07, "loss": 0.8084, "step": 28114 }, { "epoch": 0.12, "learning_rate": 2.162737699363371e-07, "loss": 0.8647, "step": 28115 }, { "epoch": 0.12, "learning_rate": 2.1605086680849596e-07, "loss": 0.842, "step": 28116 }, { "epoch": 0.12, "learning_rate": 2.158280773533128e-07, "loss": 0.8525, "step": 28117 }, { "epoch": 0.12, "learning_rate": 2.156054015733755e-07, "loss": 0.8281, "step": 28118 }, { "epoch": 0.12, "learning_rate": 2.153828394712687e-07, "loss": 0.8507, "step": 28119 }, { "epoch": 0.12, "learning_rate": 2.151603910495803e-07, "loss": 0.8344, "step": 28120 }, { "epoch": 0.12, "learning_rate": 2.1493805631089493e-07, "loss": 0.8064, "step": 28121 }, { "epoch": 0.12, "learning_rate": 2.1471583525779494e-07, "loss": 0.8327, "step": 28122 }, { "epoch": 0.12, "learning_rate": 2.1449372789286162e-07, "loss": 0.8597, "step": 28123 }, { "epoch": 0.12, "learning_rate": 2.1427173421867732e-07, "loss": 0.8546, "step": 28124 }, { "epoch": 0.12, "learning_rate": 2.1404985423782e-07, "loss": 0.8738, "step": 28125 }, { "epoch": 0.12, "learning_rate": 2.1382808795286646e-07, "loss": 0.829, "step": 28126 }, { "epoch": 0.12, "learning_rate": 2.136064353663958e-07, "loss": 0.8507, "step": 28127 }, { "epoch": 0.12, "learning_rate": 2.1338489648098037e-07, "loss": 0.8201, "step": 28128 }, { "epoch": 0.12, "learning_rate": 2.13163471299197e-07, "loss": 0.7726, "step": 28129 }, { "epoch": 0.12, "learning_rate": 2.1294215982361587e-07, "loss": 0.8623, "step": 28130 }, { "epoch": 0.12, "learning_rate": 2.1272096205680937e-07, "loss": 0.8197, "step": 28131 }, { "epoch": 0.12, "learning_rate": 2.1249987800134875e-07, "loss": 0.8402, "step": 28132 }, { "epoch": 0.12, "learning_rate": 2.1227890765979975e-07, "loss": 0.8484, "step": 28133 }, { "epoch": 0.12, "learning_rate": 2.1205805103473031e-07, "loss": 0.8555, "step": 28134 }, { "epoch": 0.12, "learning_rate": 2.118373081287073e-07, "loss": 0.8553, "step": 28135 }, { "epoch": 0.12, "learning_rate": 2.116166789442964e-07, "loss": 0.8033, "step": 28136 }, { "epoch": 0.12, "learning_rate": 2.1139616348406e-07, "loss": 0.851, "step": 28137 }, { "epoch": 0.12, "learning_rate": 2.1117576175055832e-07, "loss": 0.8515, "step": 28138 }, { "epoch": 0.12, "learning_rate": 2.1095547374635596e-07, "loss": 0.8506, "step": 28139 }, { "epoch": 0.12, "learning_rate": 2.107352994740075e-07, "loss": 0.8126, "step": 28140 }, { "epoch": 0.12, "learning_rate": 2.105152389360743e-07, "loss": 0.877, "step": 28141 }, { "epoch": 0.12, "learning_rate": 2.1029529213511202e-07, "loss": 0.8739, "step": 28142 }, { "epoch": 0.12, "learning_rate": 2.1007545907367755e-07, "loss": 0.8719, "step": 28143 }, { "epoch": 0.12, "learning_rate": 2.0985573975432215e-07, "loss": 0.8095, "step": 28144 }, { "epoch": 0.12, "learning_rate": 2.096361341796005e-07, "loss": 0.8397, "step": 28145 }, { "epoch": 0.12, "learning_rate": 2.0941664235206382e-07, "loss": 0.8318, "step": 28146 }, { "epoch": 0.12, "learning_rate": 2.0919726427426124e-07, "loss": 0.8421, "step": 28147 }, { "epoch": 0.12, "learning_rate": 2.089779999487429e-07, "loss": 0.8102, "step": 28148 }, { "epoch": 0.12, "learning_rate": 2.087588493780568e-07, "loss": 0.8514, "step": 28149 }, { "epoch": 0.12, "learning_rate": 2.0853981256474754e-07, "loss": 0.8203, "step": 28150 }, { "epoch": 0.12, "learning_rate": 2.0832088951135977e-07, "loss": 0.8223, "step": 28151 }, { "epoch": 0.12, "learning_rate": 2.081020802204381e-07, "loss": 0.7908, "step": 28152 }, { "epoch": 0.12, "learning_rate": 2.0788338469452384e-07, "loss": 0.8514, "step": 28153 }, { "epoch": 0.12, "learning_rate": 2.076648029361583e-07, "loss": 0.8268, "step": 28154 }, { "epoch": 0.12, "learning_rate": 2.0744633494788168e-07, "loss": 0.8522, "step": 28155 }, { "epoch": 0.12, "learning_rate": 2.0722798073223083e-07, "loss": 0.8742, "step": 28156 }, { "epoch": 0.12, "learning_rate": 2.070097402917448e-07, "loss": 0.8143, "step": 28157 }, { "epoch": 0.12, "learning_rate": 2.0679161362895605e-07, "loss": 0.8221, "step": 28158 }, { "epoch": 0.12, "learning_rate": 2.065736007464003e-07, "loss": 0.8488, "step": 28159 }, { "epoch": 0.12, "learning_rate": 2.063557016466111e-07, "loss": 0.8331, "step": 28160 }, { "epoch": 0.12, "learning_rate": 2.0613791633212088e-07, "loss": 0.8196, "step": 28161 }, { "epoch": 0.12, "learning_rate": 2.0592024480545868e-07, "loss": 0.8361, "step": 28162 }, { "epoch": 0.12, "learning_rate": 2.0570268706915254e-07, "loss": 0.8726, "step": 28163 }, { "epoch": 0.12, "learning_rate": 2.0548524312573148e-07, "loss": 0.8124, "step": 28164 }, { "epoch": 0.12, "learning_rate": 2.052679129777213e-07, "loss": 0.8416, "step": 28165 }, { "epoch": 0.12, "learning_rate": 2.0505069662764666e-07, "loss": 0.849, "step": 28166 }, { "epoch": 0.12, "learning_rate": 2.0483359407803217e-07, "loss": 0.8448, "step": 28167 }, { "epoch": 0.12, "learning_rate": 2.0461660533140027e-07, "loss": 0.7967, "step": 28168 }, { "epoch": 0.12, "learning_rate": 2.0439973039027118e-07, "loss": 0.8714, "step": 28169 }, { "epoch": 0.12, "learning_rate": 2.0418296925716396e-07, "loss": 0.8325, "step": 28170 }, { "epoch": 0.12, "learning_rate": 2.0396632193459886e-07, "loss": 0.8952, "step": 28171 }, { "epoch": 0.12, "learning_rate": 2.0374978842509164e-07, "loss": 0.8245, "step": 28172 }, { "epoch": 0.12, "learning_rate": 2.0353336873115803e-07, "loss": 0.8761, "step": 28173 }, { "epoch": 0.12, "learning_rate": 2.0331706285531384e-07, "loss": 0.8525, "step": 28174 }, { "epoch": 0.12, "learning_rate": 2.0310087080007145e-07, "loss": 0.8332, "step": 28175 }, { "epoch": 0.12, "learning_rate": 2.0288479256794114e-07, "loss": 0.8143, "step": 28176 }, { "epoch": 0.12, "learning_rate": 2.026688281614342e-07, "loss": 0.8634, "step": 28177 }, { "epoch": 0.12, "learning_rate": 2.0245297758306192e-07, "loss": 0.8332, "step": 28178 }, { "epoch": 0.12, "learning_rate": 2.0223724083532903e-07, "loss": 0.8611, "step": 28179 }, { "epoch": 0.12, "learning_rate": 2.020216179207435e-07, "loss": 0.8145, "step": 28180 }, { "epoch": 0.12, "learning_rate": 2.0180610884180995e-07, "loss": 0.8384, "step": 28181 }, { "epoch": 0.12, "learning_rate": 2.015907136010342e-07, "loss": 0.848, "step": 28182 }, { "epoch": 0.12, "learning_rate": 2.013754322009154e-07, "loss": 0.8652, "step": 28183 }, { "epoch": 0.12, "learning_rate": 2.01160264643957e-07, "loss": 0.8143, "step": 28184 }, { "epoch": 0.12, "learning_rate": 2.009452109326604e-07, "loss": 0.8359, "step": 28185 }, { "epoch": 0.12, "learning_rate": 2.0073027106952026e-07, "loss": 0.8386, "step": 28186 }, { "epoch": 0.12, "learning_rate": 2.005154450570357e-07, "loss": 0.869, "step": 28187 }, { "epoch": 0.12, "learning_rate": 2.0030073289770245e-07, "loss": 0.8365, "step": 28188 }, { "epoch": 0.12, "learning_rate": 2.000861345940164e-07, "loss": 0.8412, "step": 28189 }, { "epoch": 0.12, "learning_rate": 1.9987165014846876e-07, "loss": 0.8029, "step": 28190 }, { "epoch": 0.12, "learning_rate": 1.9965727956355318e-07, "loss": 0.8746, "step": 28191 }, { "epoch": 0.12, "learning_rate": 1.9944302284175987e-07, "loss": 0.8548, "step": 28192 }, { "epoch": 0.12, "learning_rate": 1.9922887998557683e-07, "loss": 0.8466, "step": 28193 }, { "epoch": 0.12, "learning_rate": 1.9901485099749207e-07, "loss": 0.8449, "step": 28194 }, { "epoch": 0.12, "learning_rate": 1.9880093587999472e-07, "loss": 0.8443, "step": 28195 }, { "epoch": 0.12, "learning_rate": 1.9858713463556721e-07, "loss": 0.8379, "step": 28196 }, { "epoch": 0.12, "learning_rate": 1.9837344726669648e-07, "loss": 0.8197, "step": 28197 }, { "epoch": 0.12, "learning_rate": 1.9815987377586276e-07, "loss": 0.8666, "step": 28198 }, { "epoch": 0.12, "learning_rate": 1.979464141655485e-07, "loss": 0.8201, "step": 28199 }, { "epoch": 0.12, "learning_rate": 1.9773306843823282e-07, "loss": 0.8229, "step": 28200 }, { "epoch": 0.12, "learning_rate": 1.9751983659639595e-07, "loss": 0.8406, "step": 28201 }, { "epoch": 0.12, "learning_rate": 1.973067186425126e-07, "loss": 0.8315, "step": 28202 }, { "epoch": 0.12, "learning_rate": 1.97093714579063e-07, "loss": 0.8551, "step": 28203 }, { "epoch": 0.12, "learning_rate": 1.9688082440851854e-07, "loss": 0.8664, "step": 28204 }, { "epoch": 0.12, "learning_rate": 1.9666804813335272e-07, "loss": 0.8299, "step": 28205 }, { "epoch": 0.12, "learning_rate": 1.9645538575603918e-07, "loss": 0.8786, "step": 28206 }, { "epoch": 0.12, "learning_rate": 1.9624283727904813e-07, "loss": 0.8577, "step": 28207 }, { "epoch": 0.12, "learning_rate": 1.9603040270484986e-07, "loss": 0.8889, "step": 28208 }, { "epoch": 0.12, "learning_rate": 1.9581808203591014e-07, "loss": 0.8467, "step": 28209 }, { "epoch": 0.12, "learning_rate": 1.956058752746992e-07, "loss": 0.8626, "step": 28210 }, { "epoch": 0.12, "learning_rate": 1.9539378242367845e-07, "loss": 0.861, "step": 28211 }, { "epoch": 0.12, "learning_rate": 1.951818034853148e-07, "loss": 0.8484, "step": 28212 }, { "epoch": 0.12, "learning_rate": 1.9496993846207068e-07, "loss": 0.8283, "step": 28213 }, { "epoch": 0.12, "learning_rate": 1.947581873564064e-07, "loss": 0.854, "step": 28214 }, { "epoch": 0.12, "learning_rate": 1.9454655017078438e-07, "loss": 0.8316, "step": 28215 }, { "epoch": 0.12, "learning_rate": 1.9433502690766048e-07, "loss": 0.8574, "step": 28216 }, { "epoch": 0.12, "learning_rate": 1.9412361756949605e-07, "loss": 0.8518, "step": 28217 }, { "epoch": 0.12, "learning_rate": 1.9391232215874356e-07, "loss": 0.8582, "step": 28218 }, { "epoch": 0.12, "learning_rate": 1.9370114067785995e-07, "loss": 0.8266, "step": 28219 }, { "epoch": 0.12, "learning_rate": 1.934900731292988e-07, "loss": 0.8169, "step": 28220 }, { "epoch": 0.12, "learning_rate": 1.9327911951551148e-07, "loss": 0.7908, "step": 28221 }, { "epoch": 0.12, "learning_rate": 1.930682798389494e-07, "loss": 0.8553, "step": 28222 }, { "epoch": 0.12, "learning_rate": 1.9285755410206164e-07, "loss": 0.8565, "step": 28223 }, { "epoch": 0.12, "learning_rate": 1.926469423072974e-07, "loss": 0.8271, "step": 28224 }, { "epoch": 0.12, "learning_rate": 1.9243644445710365e-07, "loss": 0.8329, "step": 28225 }, { "epoch": 0.12, "learning_rate": 1.922260605539239e-07, "loss": 0.8386, "step": 28226 }, { "epoch": 0.12, "learning_rate": 1.9201579060020402e-07, "loss": 0.857, "step": 28227 }, { "epoch": 0.12, "learning_rate": 1.9180563459838874e-07, "loss": 0.868, "step": 28228 }, { "epoch": 0.12, "learning_rate": 1.915955925509172e-07, "loss": 0.8265, "step": 28229 }, { "epoch": 0.12, "learning_rate": 1.9138566446023076e-07, "loss": 0.8748, "step": 28230 }, { "epoch": 0.12, "learning_rate": 1.9117585032876861e-07, "loss": 0.8245, "step": 28231 }, { "epoch": 0.12, "learning_rate": 1.9096615015896657e-07, "loss": 0.8304, "step": 28232 }, { "epoch": 0.12, "learning_rate": 1.907565639532627e-07, "loss": 0.8234, "step": 28233 }, { "epoch": 0.12, "learning_rate": 1.905470917140917e-07, "loss": 0.8247, "step": 28234 }, { "epoch": 0.12, "learning_rate": 1.903377334438894e-07, "loss": 0.8334, "step": 28235 }, { "epoch": 0.12, "learning_rate": 1.9012848914508387e-07, "loss": 0.8423, "step": 28236 }, { "epoch": 0.12, "learning_rate": 1.899193588201087e-07, "loss": 0.8419, "step": 28237 }, { "epoch": 0.12, "learning_rate": 1.8971034247139418e-07, "loss": 0.8233, "step": 28238 }, { "epoch": 0.12, "learning_rate": 1.895014401013673e-07, "loss": 0.8791, "step": 28239 }, { "epoch": 0.12, "learning_rate": 1.8929265171245604e-07, "loss": 0.844, "step": 28240 }, { "epoch": 0.12, "learning_rate": 1.8908397730708517e-07, "loss": 0.8263, "step": 28241 }, { "epoch": 0.12, "learning_rate": 1.8887541688768053e-07, "loss": 0.815, "step": 28242 }, { "epoch": 0.12, "learning_rate": 1.886669704566646e-07, "loss": 0.8642, "step": 28243 }, { "epoch": 0.12, "learning_rate": 1.884586380164577e-07, "loss": 0.8347, "step": 28244 }, { "epoch": 0.12, "learning_rate": 1.8825041956948343e-07, "loss": 0.8497, "step": 28245 }, { "epoch": 0.12, "learning_rate": 1.8804231511815762e-07, "loss": 0.8091, "step": 28246 }, { "epoch": 0.12, "learning_rate": 1.8783432466490058e-07, "loss": 0.8141, "step": 28247 }, { "epoch": 0.12, "learning_rate": 1.8762644821212707e-07, "loss": 0.8455, "step": 28248 }, { "epoch": 0.12, "learning_rate": 1.8741868576225508e-07, "loss": 0.9081, "step": 28249 }, { "epoch": 0.12, "learning_rate": 1.8721103731769498e-07, "loss": 0.8377, "step": 28250 }, { "epoch": 0.12, "learning_rate": 1.8700350288085923e-07, "loss": 0.8643, "step": 28251 }, { "epoch": 0.12, "learning_rate": 1.8679608245416147e-07, "loss": 0.8288, "step": 28252 }, { "epoch": 0.12, "learning_rate": 1.8658877604001202e-07, "loss": 0.8481, "step": 28253 }, { "epoch": 0.12, "learning_rate": 1.8638158364081672e-07, "loss": 0.8406, "step": 28254 }, { "epoch": 0.12, "learning_rate": 1.8617450525898473e-07, "loss": 0.8503, "step": 28255 }, { "epoch": 0.12, "learning_rate": 1.8596754089692082e-07, "loss": 0.7961, "step": 28256 }, { "epoch": 0.13, "learning_rate": 1.8576069055702973e-07, "loss": 0.8253, "step": 28257 }, { "epoch": 0.13, "learning_rate": 1.855539542417162e-07, "loss": 0.8195, "step": 28258 }, { "epoch": 0.13, "learning_rate": 1.853473319533794e-07, "loss": 0.8433, "step": 28259 }, { "epoch": 0.13, "learning_rate": 1.85140823694423e-07, "loss": 0.795, "step": 28260 }, { "epoch": 0.13, "learning_rate": 1.8493442946724394e-07, "loss": 0.8417, "step": 28261 }, { "epoch": 0.13, "learning_rate": 1.8472814927424142e-07, "loss": 0.8235, "step": 28262 }, { "epoch": 0.13, "learning_rate": 1.8452198311781133e-07, "loss": 0.8564, "step": 28263 }, { "epoch": 0.13, "learning_rate": 1.843159310003495e-07, "loss": 0.8767, "step": 28264 }, { "epoch": 0.13, "learning_rate": 1.8410999292424847e-07, "loss": 0.8652, "step": 28265 }, { "epoch": 0.13, "learning_rate": 1.8390416889190299e-07, "loss": 0.8186, "step": 28266 }, { "epoch": 0.13, "learning_rate": 1.8369845890570449e-07, "loss": 0.8492, "step": 28267 }, { "epoch": 0.13, "learning_rate": 1.8349286296804104e-07, "loss": 0.8275, "step": 28268 }, { "epoch": 0.13, "learning_rate": 1.8328738108130073e-07, "loss": 0.8269, "step": 28269 }, { "epoch": 0.13, "learning_rate": 1.83082013247875e-07, "loss": 0.8571, "step": 28270 }, { "epoch": 0.13, "learning_rate": 1.8287675947014638e-07, "loss": 0.8444, "step": 28271 }, { "epoch": 0.13, "learning_rate": 1.8267161975049962e-07, "loss": 0.8774, "step": 28272 }, { "epoch": 0.13, "learning_rate": 1.824665940913195e-07, "loss": 0.8786, "step": 28273 }, { "epoch": 0.13, "learning_rate": 1.8226168249498854e-07, "loss": 0.8361, "step": 28274 }, { "epoch": 0.13, "learning_rate": 1.8205688496388374e-07, "loss": 0.8905, "step": 28275 }, { "epoch": 0.13, "learning_rate": 1.8185220150038873e-07, "loss": 0.8449, "step": 28276 }, { "epoch": 0.13, "learning_rate": 1.8164763210688052e-07, "loss": 0.8461, "step": 28277 }, { "epoch": 0.13, "learning_rate": 1.8144317678573497e-07, "loss": 0.8083, "step": 28278 }, { "epoch": 0.13, "learning_rate": 1.8123883553932796e-07, "loss": 0.8336, "step": 28279 }, { "epoch": 0.13, "learning_rate": 1.8103460837003316e-07, "loss": 0.8563, "step": 28280 }, { "epoch": 0.13, "learning_rate": 1.8083049528022312e-07, "loss": 0.8294, "step": 28281 }, { "epoch": 0.13, "learning_rate": 1.8062649627227037e-07, "loss": 0.8158, "step": 28282 }, { "epoch": 0.13, "learning_rate": 1.804226113485441e-07, "loss": 0.8875, "step": 28283 }, { "epoch": 0.13, "learning_rate": 1.802188405114147e-07, "loss": 0.8254, "step": 28284 }, { "epoch": 0.13, "learning_rate": 1.80015183763248e-07, "loss": 0.8554, "step": 28285 }, { "epoch": 0.13, "learning_rate": 1.798116411064099e-07, "loss": 0.8161, "step": 28286 }, { "epoch": 0.13, "learning_rate": 1.7960821254326632e-07, "loss": 0.8374, "step": 28287 }, { "epoch": 0.13, "learning_rate": 1.7940489807618088e-07, "loss": 0.869, "step": 28288 }, { "epoch": 0.13, "learning_rate": 1.7920169770751395e-07, "loss": 0.8642, "step": 28289 }, { "epoch": 0.13, "learning_rate": 1.7899861143962804e-07, "loss": 0.8455, "step": 28290 }, { "epoch": 0.13, "learning_rate": 1.7879563927488352e-07, "loss": 0.8379, "step": 28291 }, { "epoch": 0.13, "learning_rate": 1.785927812156363e-07, "loss": 0.8278, "step": 28292 }, { "epoch": 0.13, "learning_rate": 1.783900372642433e-07, "loss": 0.8122, "step": 28293 }, { "epoch": 0.13, "learning_rate": 1.7818740742306162e-07, "loss": 0.873, "step": 28294 }, { "epoch": 0.13, "learning_rate": 1.77984891694446e-07, "loss": 0.8813, "step": 28295 }, { "epoch": 0.13, "learning_rate": 1.7778249008074678e-07, "loss": 0.825, "step": 28296 }, { "epoch": 0.13, "learning_rate": 1.7758020258431652e-07, "loss": 0.8171, "step": 28297 }, { "epoch": 0.13, "learning_rate": 1.7737802920750668e-07, "loss": 0.878, "step": 28298 }, { "epoch": 0.13, "learning_rate": 1.7717596995266539e-07, "loss": 0.8695, "step": 28299 }, { "epoch": 0.13, "learning_rate": 1.7697402482213964e-07, "loss": 0.82, "step": 28300 }, { "epoch": 0.13, "learning_rate": 1.7677219381827647e-07, "loss": 0.7886, "step": 28301 }, { "epoch": 0.13, "learning_rate": 1.7657047694342178e-07, "loss": 0.8398, "step": 28302 }, { "epoch": 0.13, "learning_rate": 1.7636887419991698e-07, "loss": 0.8829, "step": 28303 }, { "epoch": 0.13, "learning_rate": 1.761673855901047e-07, "loss": 0.8362, "step": 28304 }, { "epoch": 0.13, "learning_rate": 1.7596601111632639e-07, "loss": 0.8412, "step": 28305 }, { "epoch": 0.13, "learning_rate": 1.7576475078092238e-07, "loss": 0.8569, "step": 28306 }, { "epoch": 0.13, "learning_rate": 1.7556360458622967e-07, "loss": 0.8484, "step": 28307 }, { "epoch": 0.13, "learning_rate": 1.7536257253458644e-07, "loss": 0.8532, "step": 28308 }, { "epoch": 0.13, "learning_rate": 1.7516165462832857e-07, "loss": 0.8325, "step": 28309 }, { "epoch": 0.13, "learning_rate": 1.7496085086978974e-07, "loss": 0.8363, "step": 28310 }, { "epoch": 0.13, "learning_rate": 1.7476016126130147e-07, "loss": 0.8561, "step": 28311 }, { "epoch": 0.13, "learning_rate": 1.7455958580519627e-07, "loss": 0.8437, "step": 28312 }, { "epoch": 0.13, "learning_rate": 1.7435912450380676e-07, "loss": 0.8294, "step": 28313 }, { "epoch": 0.13, "learning_rate": 1.7415877735945995e-07, "loss": 0.811, "step": 28314 }, { "epoch": 0.13, "learning_rate": 1.739585443744829e-07, "loss": 0.8239, "step": 28315 }, { "epoch": 0.13, "learning_rate": 1.737584255512048e-07, "loss": 0.8087, "step": 28316 }, { "epoch": 0.13, "learning_rate": 1.735584208919472e-07, "loss": 0.844, "step": 28317 }, { "epoch": 0.13, "learning_rate": 1.7335853039903373e-07, "loss": 0.8368, "step": 28318 }, { "epoch": 0.13, "learning_rate": 1.731587540747903e-07, "loss": 0.8684, "step": 28319 }, { "epoch": 0.13, "learning_rate": 1.7295909192153626e-07, "loss": 0.8356, "step": 28320 }, { "epoch": 0.13, "learning_rate": 1.7275954394159079e-07, "loss": 0.8468, "step": 28321 }, { "epoch": 0.13, "learning_rate": 1.7256011013727313e-07, "loss": 0.836, "step": 28322 }, { "epoch": 0.13, "learning_rate": 1.7236079051089927e-07, "loss": 0.8492, "step": 28323 }, { "epoch": 0.13, "learning_rate": 1.7216158506478508e-07, "loss": 0.8727, "step": 28324 }, { "epoch": 0.13, "learning_rate": 1.7196249380124542e-07, "loss": 0.8597, "step": 28325 }, { "epoch": 0.13, "learning_rate": 1.7176351672259395e-07, "loss": 0.8386, "step": 28326 }, { "epoch": 0.13, "learning_rate": 1.715646538311433e-07, "loss": 0.868, "step": 28327 }, { "epoch": 0.13, "learning_rate": 1.713659051292016e-07, "loss": 0.8511, "step": 28328 }, { "epoch": 0.13, "learning_rate": 1.7116727061907813e-07, "loss": 0.8574, "step": 28329 }, { "epoch": 0.13, "learning_rate": 1.7096875030308212e-07, "loss": 0.8259, "step": 28330 }, { "epoch": 0.13, "learning_rate": 1.7077034418351847e-07, "loss": 0.8964, "step": 28331 }, { "epoch": 0.13, "learning_rate": 1.7057205226269414e-07, "loss": 0.8169, "step": 28332 }, { "epoch": 0.13, "learning_rate": 1.7037387454291176e-07, "loss": 0.873, "step": 28333 }, { "epoch": 0.13, "learning_rate": 1.7017581102647508e-07, "loss": 0.832, "step": 28334 }, { "epoch": 0.13, "learning_rate": 1.6997786171568332e-07, "loss": 0.8731, "step": 28335 }, { "epoch": 0.13, "learning_rate": 1.6978002661283688e-07, "loss": 0.8466, "step": 28336 }, { "epoch": 0.13, "learning_rate": 1.6958230572023504e-07, "loss": 0.8257, "step": 28337 }, { "epoch": 0.13, "learning_rate": 1.6938469904017374e-07, "loss": 0.905, "step": 28338 }, { "epoch": 0.13, "learning_rate": 1.6918720657495114e-07, "loss": 0.9068, "step": 28339 }, { "epoch": 0.13, "learning_rate": 1.6898982832685873e-07, "loss": 0.8278, "step": 28340 }, { "epoch": 0.13, "learning_rate": 1.6879256429819358e-07, "loss": 0.8106, "step": 28341 }, { "epoch": 0.13, "learning_rate": 1.6859541449124271e-07, "loss": 0.8839, "step": 28342 }, { "epoch": 0.13, "learning_rate": 1.6839837890829992e-07, "loss": 0.8909, "step": 28343 }, { "epoch": 0.13, "learning_rate": 1.682014575516544e-07, "loss": 0.8347, "step": 28344 }, { "epoch": 0.13, "learning_rate": 1.6800465042359326e-07, "loss": 0.8225, "step": 28345 }, { "epoch": 0.13, "learning_rate": 1.6780795752640243e-07, "loss": 0.8681, "step": 28346 }, { "epoch": 0.13, "learning_rate": 1.676113788623679e-07, "loss": 0.8678, "step": 28347 }, { "epoch": 0.13, "learning_rate": 1.6741491443377444e-07, "loss": 0.8144, "step": 28348 }, { "epoch": 0.13, "learning_rate": 1.6721856424290252e-07, "loss": 0.8348, "step": 28349 }, { "epoch": 0.13, "learning_rate": 1.670223282920358e-07, "loss": 0.8105, "step": 28350 }, { "epoch": 0.13, "learning_rate": 1.6682620658345138e-07, "loss": 0.8654, "step": 28351 }, { "epoch": 0.13, "learning_rate": 1.666301991194319e-07, "loss": 0.8383, "step": 28352 }, { "epoch": 0.13, "learning_rate": 1.6643430590224995e-07, "loss": 0.8101, "step": 28353 }, { "epoch": 0.13, "learning_rate": 1.6623852693418484e-07, "loss": 0.8423, "step": 28354 }, { "epoch": 0.13, "learning_rate": 1.660428622175092e-07, "loss": 0.8364, "step": 28355 }, { "epoch": 0.13, "learning_rate": 1.6584731175449786e-07, "loss": 0.8297, "step": 28356 }, { "epoch": 0.13, "learning_rate": 1.6565187554742122e-07, "loss": 0.8545, "step": 28357 }, { "epoch": 0.13, "learning_rate": 1.6545655359855196e-07, "loss": 0.8344, "step": 28358 }, { "epoch": 0.13, "learning_rate": 1.6526134591015818e-07, "loss": 0.8356, "step": 28359 }, { "epoch": 0.13, "learning_rate": 1.6506625248450702e-07, "loss": 0.7984, "step": 28360 }, { "epoch": 0.13, "learning_rate": 1.6487127332386554e-07, "loss": 0.833, "step": 28361 }, { "epoch": 0.13, "learning_rate": 1.646764084305019e-07, "loss": 0.8481, "step": 28362 }, { "epoch": 0.13, "learning_rate": 1.6448165780667546e-07, "loss": 0.8901, "step": 28363 }, { "epoch": 0.13, "learning_rate": 1.6428702145465215e-07, "loss": 0.8418, "step": 28364 }, { "epoch": 0.13, "learning_rate": 1.6409249937669235e-07, "loss": 0.8646, "step": 28365 }, { "epoch": 0.13, "learning_rate": 1.6389809157505653e-07, "loss": 0.8018, "step": 28366 }, { "epoch": 0.13, "learning_rate": 1.6370379805200065e-07, "loss": 0.8622, "step": 28367 }, { "epoch": 0.13, "learning_rate": 1.635096188097851e-07, "loss": 0.8207, "step": 28368 }, { "epoch": 0.13, "learning_rate": 1.63315553850667e-07, "loss": 0.8235, "step": 28369 }, { "epoch": 0.13, "learning_rate": 1.6312160317689784e-07, "loss": 0.8444, "step": 28370 }, { "epoch": 0.13, "learning_rate": 1.6292776679073142e-07, "loss": 0.8481, "step": 28371 }, { "epoch": 0.13, "learning_rate": 1.6273404469442144e-07, "loss": 0.8014, "step": 28372 }, { "epoch": 0.13, "learning_rate": 1.6254043689021836e-07, "loss": 0.854, "step": 28373 }, { "epoch": 0.13, "learning_rate": 1.6234694338037038e-07, "loss": 0.8626, "step": 28374 }, { "epoch": 0.13, "learning_rate": 1.621535641671268e-07, "loss": 0.8419, "step": 28375 }, { "epoch": 0.13, "learning_rate": 1.619602992527336e-07, "loss": 0.8227, "step": 28376 }, { "epoch": 0.13, "learning_rate": 1.6176714863943677e-07, "loss": 0.8609, "step": 28377 }, { "epoch": 0.13, "learning_rate": 1.6157411232947895e-07, "loss": 0.8546, "step": 28378 }, { "epoch": 0.13, "learning_rate": 1.613811903251039e-07, "loss": 0.8778, "step": 28379 }, { "epoch": 0.13, "learning_rate": 1.611883826285543e-07, "loss": 0.7994, "step": 28380 }, { "epoch": 0.13, "learning_rate": 1.609956892420683e-07, "loss": 0.8367, "step": 28381 }, { "epoch": 0.13, "learning_rate": 1.608031101678853e-07, "loss": 0.7929, "step": 28382 }, { "epoch": 0.13, "learning_rate": 1.6061064540824345e-07, "loss": 0.8169, "step": 28383 }, { "epoch": 0.13, "learning_rate": 1.6041829496537765e-07, "loss": 0.822, "step": 28384 }, { "epoch": 0.13, "learning_rate": 1.6022605884152164e-07, "loss": 0.8312, "step": 28385 }, { "epoch": 0.13, "learning_rate": 1.6003393703891256e-07, "loss": 0.8653, "step": 28386 }, { "epoch": 0.13, "learning_rate": 1.5984192955978085e-07, "loss": 0.8674, "step": 28387 }, { "epoch": 0.13, "learning_rate": 1.596500364063558e-07, "loss": 0.8129, "step": 28388 }, { "epoch": 0.13, "learning_rate": 1.5945825758086786e-07, "loss": 0.8536, "step": 28389 }, { "epoch": 0.13, "learning_rate": 1.5926659308554638e-07, "loss": 0.8388, "step": 28390 }, { "epoch": 0.13, "learning_rate": 1.5907504292261622e-07, "loss": 0.8967, "step": 28391 }, { "epoch": 0.13, "learning_rate": 1.588836070943034e-07, "loss": 0.8328, "step": 28392 }, { "epoch": 0.13, "learning_rate": 1.586922856028328e-07, "loss": 0.8599, "step": 28393 }, { "epoch": 0.13, "learning_rate": 1.5850107845042818e-07, "loss": 0.8746, "step": 28394 }, { "epoch": 0.13, "learning_rate": 1.5830998563930888e-07, "loss": 0.8721, "step": 28395 }, { "epoch": 0.13, "learning_rate": 1.5811900717169537e-07, "loss": 0.8277, "step": 28396 }, { "epoch": 0.13, "learning_rate": 1.5792814304980697e-07, "loss": 0.8691, "step": 28397 }, { "epoch": 0.13, "learning_rate": 1.577373932758619e-07, "loss": 0.862, "step": 28398 }, { "epoch": 0.13, "learning_rate": 1.5754675785207506e-07, "loss": 0.8521, "step": 28399 }, { "epoch": 0.13, "learning_rate": 1.5735623678066247e-07, "loss": 0.8737, "step": 28400 }, { "epoch": 0.13, "learning_rate": 1.571658300638379e-07, "loss": 0.8421, "step": 28401 }, { "epoch": 0.13, "learning_rate": 1.569755377038118e-07, "loss": 0.8376, "step": 28402 }, { "epoch": 0.13, "learning_rate": 1.5678535970279575e-07, "loss": 0.8098, "step": 28403 }, { "epoch": 0.13, "learning_rate": 1.5659529606299905e-07, "loss": 0.8359, "step": 28404 }, { "epoch": 0.13, "learning_rate": 1.564053467866322e-07, "loss": 0.8113, "step": 28405 }, { "epoch": 0.13, "learning_rate": 1.56215511875899e-07, "loss": 0.8204, "step": 28406 }, { "epoch": 0.13, "learning_rate": 1.5602579133300654e-07, "loss": 0.8716, "step": 28407 }, { "epoch": 0.13, "learning_rate": 1.5583618516015975e-07, "loss": 0.8551, "step": 28408 }, { "epoch": 0.13, "learning_rate": 1.5564669335956018e-07, "loss": 0.8119, "step": 28409 }, { "epoch": 0.13, "learning_rate": 1.554573159334083e-07, "loss": 0.8427, "step": 28410 }, { "epoch": 0.13, "learning_rate": 1.552680528839068e-07, "loss": 0.8409, "step": 28411 }, { "epoch": 0.13, "learning_rate": 1.55078904213255e-07, "loss": 0.8232, "step": 28412 }, { "epoch": 0.13, "learning_rate": 1.5488986992364785e-07, "loss": 0.8167, "step": 28413 }, { "epoch": 0.13, "learning_rate": 1.5470095001728358e-07, "loss": 0.839, "step": 28414 }, { "epoch": 0.13, "learning_rate": 1.5451214449635598e-07, "loss": 0.8644, "step": 28415 }, { "epoch": 0.13, "learning_rate": 1.5432345336305886e-07, "loss": 0.8328, "step": 28416 }, { "epoch": 0.13, "learning_rate": 1.5413487661958493e-07, "loss": 0.8287, "step": 28417 }, { "epoch": 0.13, "learning_rate": 1.5394641426812461e-07, "loss": 0.842, "step": 28418 }, { "epoch": 0.13, "learning_rate": 1.5375806631086842e-07, "loss": 0.8941, "step": 28419 }, { "epoch": 0.13, "learning_rate": 1.535698327500035e-07, "loss": 0.8474, "step": 28420 }, { "epoch": 0.13, "learning_rate": 1.5338171358771803e-07, "loss": 0.8169, "step": 28421 }, { "epoch": 0.13, "learning_rate": 1.531937088261959e-07, "loss": 0.8508, "step": 28422 }, { "epoch": 0.13, "learning_rate": 1.5300581846762308e-07, "loss": 0.8774, "step": 28423 }, { "epoch": 0.13, "learning_rate": 1.528180425141823e-07, "loss": 0.8584, "step": 28424 }, { "epoch": 0.13, "learning_rate": 1.5263038096805404e-07, "loss": 0.8392, "step": 28425 }, { "epoch": 0.13, "learning_rate": 1.524428338314199e-07, "loss": 0.8239, "step": 28426 }, { "epoch": 0.13, "learning_rate": 1.522554011064581e-07, "loss": 0.8394, "step": 28427 }, { "epoch": 0.13, "learning_rate": 1.520680827953469e-07, "loss": 0.8169, "step": 28428 }, { "epoch": 0.13, "learning_rate": 1.5188087890026125e-07, "loss": 0.8284, "step": 28429 }, { "epoch": 0.13, "learning_rate": 1.516937894233772e-07, "loss": 0.7878, "step": 28430 }, { "epoch": 0.13, "learning_rate": 1.5150681436686966e-07, "loss": 0.8547, "step": 28431 }, { "epoch": 0.13, "learning_rate": 1.5131995373290799e-07, "loss": 0.8299, "step": 28432 }, { "epoch": 0.13, "learning_rate": 1.5113320752366601e-07, "loss": 0.8228, "step": 28433 }, { "epoch": 0.13, "learning_rate": 1.5094657574131088e-07, "loss": 0.846, "step": 28434 }, { "epoch": 0.13, "learning_rate": 1.5076005838801312e-07, "loss": 0.8462, "step": 28435 }, { "epoch": 0.13, "learning_rate": 1.5057365546593872e-07, "loss": 0.8628, "step": 28436 }, { "epoch": 0.13, "learning_rate": 1.5038736697725377e-07, "loss": 0.8428, "step": 28437 }, { "epoch": 0.13, "learning_rate": 1.5020119292412204e-07, "loss": 0.8462, "step": 28438 }, { "epoch": 0.13, "learning_rate": 1.5001513330870742e-07, "loss": 0.8563, "step": 28439 }, { "epoch": 0.13, "learning_rate": 1.4982918813317037e-07, "loss": 0.8275, "step": 28440 }, { "epoch": 0.13, "learning_rate": 1.496433573996714e-07, "loss": 0.8226, "step": 28441 }, { "epoch": 0.13, "learning_rate": 1.4945764111037097e-07, "loss": 0.8544, "step": 28442 }, { "epoch": 0.13, "learning_rate": 1.4927203926742627e-07, "loss": 0.8428, "step": 28443 }, { "epoch": 0.13, "learning_rate": 1.4908655187299337e-07, "loss": 0.7958, "step": 28444 }, { "epoch": 0.13, "learning_rate": 1.4890117892922606e-07, "loss": 0.8789, "step": 28445 }, { "epoch": 0.13, "learning_rate": 1.4871592043827933e-07, "loss": 0.8557, "step": 28446 }, { "epoch": 0.13, "learning_rate": 1.4853077640230585e-07, "loss": 0.8715, "step": 28447 }, { "epoch": 0.13, "learning_rate": 1.4834574682345616e-07, "loss": 0.8412, "step": 28448 }, { "epoch": 0.13, "learning_rate": 1.4816083170388075e-07, "loss": 0.8609, "step": 28449 }, { "epoch": 0.13, "learning_rate": 1.4797603104572676e-07, "loss": 0.8675, "step": 28450 }, { "epoch": 0.13, "learning_rate": 1.4779134485114255e-07, "loss": 0.8439, "step": 28451 }, { "epoch": 0.13, "learning_rate": 1.47606773122273e-07, "loss": 0.8066, "step": 28452 }, { "epoch": 0.13, "learning_rate": 1.474223158612609e-07, "loss": 0.8806, "step": 28453 }, { "epoch": 0.13, "learning_rate": 1.4723797307025333e-07, "loss": 0.8522, "step": 28454 }, { "epoch": 0.13, "learning_rate": 1.4705374475138978e-07, "loss": 0.8409, "step": 28455 }, { "epoch": 0.13, "learning_rate": 1.4686963090680962e-07, "loss": 0.83, "step": 28456 }, { "epoch": 0.13, "learning_rate": 1.4668563153865335e-07, "loss": 0.8036, "step": 28457 }, { "epoch": 0.13, "learning_rate": 1.4650174664905814e-07, "loss": 0.8207, "step": 28458 }, { "epoch": 0.13, "learning_rate": 1.463179762401612e-07, "loss": 0.8788, "step": 28459 }, { "epoch": 0.13, "learning_rate": 1.4613432031409747e-07, "loss": 0.8219, "step": 28460 }, { "epoch": 0.13, "learning_rate": 1.459507788730008e-07, "loss": 0.8362, "step": 28461 }, { "epoch": 0.13, "learning_rate": 1.457673519190017e-07, "loss": 0.8208, "step": 28462 }, { "epoch": 0.13, "learning_rate": 1.4558403945423293e-07, "loss": 0.8282, "step": 28463 }, { "epoch": 0.13, "learning_rate": 1.45400841480825e-07, "loss": 0.8243, "step": 28464 }, { "epoch": 0.13, "learning_rate": 1.452177580009051e-07, "loss": 0.813, "step": 28465 }, { "epoch": 0.13, "learning_rate": 1.450347890166004e-07, "loss": 0.8549, "step": 28466 }, { "epoch": 0.13, "learning_rate": 1.4485193453003699e-07, "loss": 0.8664, "step": 28467 }, { "epoch": 0.13, "learning_rate": 1.4466919454333982e-07, "loss": 0.8175, "step": 28468 }, { "epoch": 0.13, "learning_rate": 1.4448656905863058e-07, "loss": 0.8127, "step": 28469 }, { "epoch": 0.13, "learning_rate": 1.4430405807803306e-07, "loss": 0.8241, "step": 28470 }, { "epoch": 0.13, "learning_rate": 1.441216616036656e-07, "loss": 0.8774, "step": 28471 }, { "epoch": 0.13, "learning_rate": 1.439393796376487e-07, "loss": 0.7873, "step": 28472 }, { "epoch": 0.13, "learning_rate": 1.437572121821007e-07, "loss": 0.87, "step": 28473 }, { "epoch": 0.13, "learning_rate": 1.4357515923913657e-07, "loss": 0.804, "step": 28474 }, { "epoch": 0.13, "learning_rate": 1.4339322081087237e-07, "loss": 0.8572, "step": 28475 }, { "epoch": 0.13, "learning_rate": 1.43211396899422e-07, "loss": 0.83, "step": 28476 }, { "epoch": 0.13, "learning_rate": 1.4302968750689594e-07, "loss": 0.8693, "step": 28477 }, { "epoch": 0.13, "learning_rate": 1.4284809263540812e-07, "loss": 0.7882, "step": 28478 }, { "epoch": 0.13, "learning_rate": 1.426666122870679e-07, "loss": 0.8332, "step": 28479 }, { "epoch": 0.13, "learning_rate": 1.4248524646398253e-07, "loss": 0.8107, "step": 28480 }, { "epoch": 0.13, "learning_rate": 1.4230399516825921e-07, "loss": 0.8479, "step": 28481 }, { "epoch": 0.13, "learning_rate": 1.4212285840200511e-07, "loss": 0.8591, "step": 28482 }, { "epoch": 0.13, "learning_rate": 1.4194183616732305e-07, "loss": 0.8848, "step": 28483 }, { "epoch": 0.13, "learning_rate": 1.4176092846631683e-07, "loss": 0.8256, "step": 28484 }, { "epoch": 0.13, "learning_rate": 1.4158013530108928e-07, "loss": 0.8766, "step": 28485 }, { "epoch": 0.13, "learning_rate": 1.4139945667374088e-07, "loss": 0.9166, "step": 28486 }, { "epoch": 0.13, "learning_rate": 1.412188925863689e-07, "loss": 0.8321, "step": 28487 }, { "epoch": 0.13, "learning_rate": 1.4103844304107272e-07, "loss": 0.8168, "step": 28488 }, { "epoch": 0.13, "learning_rate": 1.4085810803994848e-07, "loss": 0.8493, "step": 28489 }, { "epoch": 0.13, "learning_rate": 1.4067788758509116e-07, "loss": 0.8008, "step": 28490 }, { "epoch": 0.13, "learning_rate": 1.4049778167859573e-07, "loss": 0.8627, "step": 28491 }, { "epoch": 0.13, "learning_rate": 1.4031779032255277e-07, "loss": 0.8063, "step": 28492 }, { "epoch": 0.13, "learning_rate": 1.4013791351905502e-07, "loss": 0.8322, "step": 28493 }, { "epoch": 0.13, "learning_rate": 1.3995815127019198e-07, "loss": 0.8258, "step": 28494 }, { "epoch": 0.13, "learning_rate": 1.3977850357805189e-07, "loss": 0.8204, "step": 28495 }, { "epoch": 0.13, "learning_rate": 1.3959897044472094e-07, "loss": 0.8516, "step": 28496 }, { "epoch": 0.13, "learning_rate": 1.394195518722874e-07, "loss": 0.8469, "step": 28497 }, { "epoch": 0.13, "learning_rate": 1.3924024786283407e-07, "loss": 0.8458, "step": 28498 }, { "epoch": 0.13, "learning_rate": 1.3906105841844487e-07, "loss": 0.8665, "step": 28499 }, { "epoch": 0.13, "learning_rate": 1.388819835412014e-07, "loss": 0.8269, "step": 28500 }, { "epoch": 0.13, "learning_rate": 1.387030232331843e-07, "loss": 0.8532, "step": 28501 }, { "epoch": 0.13, "learning_rate": 1.3852417749647186e-07, "loss": 0.8362, "step": 28502 }, { "epoch": 0.13, "learning_rate": 1.383454463331435e-07, "loss": 0.7999, "step": 28503 }, { "epoch": 0.13, "learning_rate": 1.381668297452754e-07, "loss": 0.8515, "step": 28504 }, { "epoch": 0.13, "learning_rate": 1.3798832773494254e-07, "loss": 0.8284, "step": 28505 }, { "epoch": 0.13, "learning_rate": 1.378099403042177e-07, "loss": 0.8481, "step": 28506 }, { "epoch": 0.13, "learning_rate": 1.3763166745517476e-07, "loss": 0.8643, "step": 28507 }, { "epoch": 0.13, "learning_rate": 1.3745350918988544e-07, "loss": 0.8526, "step": 28508 }, { "epoch": 0.13, "learning_rate": 1.3727546551041804e-07, "loss": 0.8129, "step": 28509 }, { "epoch": 0.13, "learning_rate": 1.3709753641884205e-07, "loss": 0.853, "step": 28510 }, { "epoch": 0.13, "learning_rate": 1.369197219172258e-07, "loss": 0.833, "step": 28511 }, { "epoch": 0.13, "learning_rate": 1.3674202200763208e-07, "loss": 0.823, "step": 28512 }, { "epoch": 0.13, "learning_rate": 1.3656443669212814e-07, "loss": 0.8758, "step": 28513 }, { "epoch": 0.13, "learning_rate": 1.3638696597277678e-07, "loss": 0.8363, "step": 28514 }, { "epoch": 0.13, "learning_rate": 1.3620960985163857e-07, "loss": 0.8712, "step": 28515 }, { "epoch": 0.13, "learning_rate": 1.3603236833077515e-07, "loss": 0.8191, "step": 28516 }, { "epoch": 0.13, "learning_rate": 1.3585524141224603e-07, "loss": 0.8527, "step": 28517 }, { "epoch": 0.13, "learning_rate": 1.3567822909810957e-07, "loss": 0.835, "step": 28518 }, { "epoch": 0.13, "learning_rate": 1.3550133139042075e-07, "loss": 0.8864, "step": 28519 }, { "epoch": 0.13, "learning_rate": 1.3532454829123465e-07, "loss": 0.8692, "step": 28520 }, { "epoch": 0.13, "learning_rate": 1.3514787980260845e-07, "loss": 0.8244, "step": 28521 }, { "epoch": 0.13, "learning_rate": 1.3497132592659056e-07, "loss": 0.8201, "step": 28522 }, { "epoch": 0.13, "learning_rate": 1.3479488666523487e-07, "loss": 0.8323, "step": 28523 }, { "epoch": 0.13, "learning_rate": 1.3461856202058975e-07, "loss": 0.8581, "step": 28524 }, { "epoch": 0.13, "learning_rate": 1.344423519947058e-07, "loss": 0.8258, "step": 28525 }, { "epoch": 0.13, "learning_rate": 1.3426625658962688e-07, "loss": 0.7839, "step": 28526 }, { "epoch": 0.13, "learning_rate": 1.3409027580740252e-07, "loss": 0.8382, "step": 28527 }, { "epoch": 0.13, "learning_rate": 1.3391440965007552e-07, "loss": 0.7742, "step": 28528 }, { "epoch": 0.13, "learning_rate": 1.3373865811968979e-07, "loss": 0.8524, "step": 28529 }, { "epoch": 0.13, "learning_rate": 1.3356302121828702e-07, "loss": 0.8344, "step": 28530 }, { "epoch": 0.13, "learning_rate": 1.333874989479078e-07, "loss": 0.8568, "step": 28531 }, { "epoch": 0.13, "learning_rate": 1.3321209131059055e-07, "loss": 0.8137, "step": 28532 }, { "epoch": 0.13, "learning_rate": 1.330367983083747e-07, "loss": 0.8546, "step": 28533 }, { "epoch": 0.13, "learning_rate": 1.3286161994329637e-07, "loss": 0.8085, "step": 28534 }, { "epoch": 0.13, "learning_rate": 1.3268655621738958e-07, "loss": 0.8661, "step": 28535 }, { "epoch": 0.13, "learning_rate": 1.325116071326904e-07, "loss": 0.8233, "step": 28536 }, { "epoch": 0.13, "learning_rate": 1.3233677269122946e-07, "loss": 0.8208, "step": 28537 }, { "epoch": 0.13, "learning_rate": 1.321620528950396e-07, "loss": 0.8901, "step": 28538 }, { "epoch": 0.13, "learning_rate": 1.3198744774615025e-07, "loss": 0.844, "step": 28539 }, { "epoch": 0.13, "learning_rate": 1.318129572465887e-07, "loss": 0.823, "step": 28540 }, { "epoch": 0.13, "learning_rate": 1.3163858139838447e-07, "loss": 0.8113, "step": 28541 }, { "epoch": 0.13, "learning_rate": 1.314643202035615e-07, "loss": 0.8484, "step": 28542 }, { "epoch": 0.13, "learning_rate": 1.3129017366414588e-07, "loss": 0.8523, "step": 28543 }, { "epoch": 0.13, "learning_rate": 1.3111614178216047e-07, "loss": 0.8054, "step": 28544 }, { "epoch": 0.13, "learning_rate": 1.309422245596259e-07, "loss": 0.8122, "step": 28545 }, { "epoch": 0.13, "learning_rate": 1.307684219985661e-07, "loss": 0.7987, "step": 28546 }, { "epoch": 0.13, "learning_rate": 1.305947341009961e-07, "loss": 0.8549, "step": 28547 }, { "epoch": 0.13, "learning_rate": 1.3042116086893764e-07, "loss": 0.8962, "step": 28548 }, { "epoch": 0.13, "learning_rate": 1.3024770230440466e-07, "loss": 0.826, "step": 28549 }, { "epoch": 0.13, "learning_rate": 1.300743584094133e-07, "loss": 0.8525, "step": 28550 }, { "epoch": 0.13, "learning_rate": 1.2990112918597753e-07, "loss": 0.8408, "step": 28551 }, { "epoch": 0.13, "learning_rate": 1.297280146361113e-07, "loss": 0.8321, "step": 28552 }, { "epoch": 0.13, "learning_rate": 1.295550147618252e-07, "loss": 0.8213, "step": 28553 }, { "epoch": 0.13, "learning_rate": 1.2938212956512652e-07, "loss": 0.8527, "step": 28554 }, { "epoch": 0.13, "learning_rate": 1.2920935904802813e-07, "loss": 0.8474, "step": 28555 }, { "epoch": 0.13, "learning_rate": 1.290367032125339e-07, "loss": 0.8349, "step": 28556 }, { "epoch": 0.14, "learning_rate": 1.2886416206065123e-07, "loss": 0.8557, "step": 28557 }, { "epoch": 0.14, "learning_rate": 1.2869173559438509e-07, "loss": 0.8948, "step": 28558 }, { "epoch": 0.14, "learning_rate": 1.2851942381573835e-07, "loss": 0.8595, "step": 28559 }, { "epoch": 0.14, "learning_rate": 1.2834722672671385e-07, "loss": 0.8658, "step": 28560 }, { "epoch": 0.14, "learning_rate": 1.281751443293111e-07, "loss": 0.823, "step": 28561 }, { "epoch": 0.14, "learning_rate": 1.2800317662552853e-07, "loss": 0.8102, "step": 28562 }, { "epoch": 0.14, "learning_rate": 1.2783132361736672e-07, "loss": 0.8374, "step": 28563 }, { "epoch": 0.14, "learning_rate": 1.2765958530681966e-07, "loss": 0.8512, "step": 28564 }, { "epoch": 0.14, "learning_rate": 1.2748796169588463e-07, "loss": 0.8406, "step": 28565 }, { "epoch": 0.14, "learning_rate": 1.2731645278655448e-07, "loss": 0.8268, "step": 28566 }, { "epoch": 0.14, "learning_rate": 1.2714505858082316e-07, "loss": 0.8653, "step": 28567 }, { "epoch": 0.14, "learning_rate": 1.269737790806802e-07, "loss": 0.8439, "step": 28568 }, { "epoch": 0.14, "learning_rate": 1.2680261428811514e-07, "loss": 0.8187, "step": 28569 }, { "epoch": 0.14, "learning_rate": 1.2663156420511858e-07, "loss": 0.8365, "step": 28570 }, { "epoch": 0.14, "learning_rate": 1.2646062883367893e-07, "loss": 0.8497, "step": 28571 }, { "epoch": 0.14, "learning_rate": 1.2628980817577908e-07, "loss": 0.8396, "step": 28572 }, { "epoch": 0.14, "learning_rate": 1.2611910223340408e-07, "loss": 0.8502, "step": 28573 }, { "epoch": 0.14, "learning_rate": 1.25948511008539e-07, "loss": 0.8175, "step": 28574 }, { "epoch": 0.14, "learning_rate": 1.257780345031645e-07, "loss": 0.8563, "step": 28575 }, { "epoch": 0.14, "learning_rate": 1.2560767271926234e-07, "loss": 0.8262, "step": 28576 }, { "epoch": 0.14, "learning_rate": 1.254374256588109e-07, "loss": 0.86, "step": 28577 }, { "epoch": 0.14, "learning_rate": 1.2526729332378862e-07, "loss": 0.8289, "step": 28578 }, { "epoch": 0.14, "learning_rate": 1.2509727571617058e-07, "loss": 0.8451, "step": 28579 }, { "epoch": 0.14, "learning_rate": 1.2492737283793522e-07, "loss": 0.8258, "step": 28580 }, { "epoch": 0.14, "learning_rate": 1.2475758469105314e-07, "loss": 0.8344, "step": 28581 }, { "epoch": 0.14, "learning_rate": 1.2458791127749948e-07, "loss": 0.8388, "step": 28582 }, { "epoch": 0.14, "learning_rate": 1.2441835259924484e-07, "loss": 0.8083, "step": 28583 }, { "epoch": 0.14, "learning_rate": 1.2424890865825768e-07, "loss": 0.8597, "step": 28584 }, { "epoch": 0.14, "learning_rate": 1.2407957945650972e-07, "loss": 0.8428, "step": 28585 }, { "epoch": 0.14, "learning_rate": 1.239103649959661e-07, "loss": 0.8081, "step": 28586 }, { "epoch": 0.14, "learning_rate": 1.2374126527859298e-07, "loss": 0.8555, "step": 28587 }, { "epoch": 0.14, "learning_rate": 1.2357228030635437e-07, "loss": 0.8209, "step": 28588 }, { "epoch": 0.14, "learning_rate": 1.2340341008121536e-07, "loss": 0.8577, "step": 28589 }, { "epoch": 0.14, "learning_rate": 1.232346546051366e-07, "loss": 0.8722, "step": 28590 }, { "epoch": 0.14, "learning_rate": 1.2306601388007988e-07, "loss": 0.8397, "step": 28591 }, { "epoch": 0.14, "learning_rate": 1.228974879080036e-07, "loss": 0.8355, "step": 28592 }, { "epoch": 0.14, "learning_rate": 1.2272907669086508e-07, "loss": 0.806, "step": 28593 }, { "epoch": 0.14, "learning_rate": 1.2256078023062278e-07, "loss": 0.8586, "step": 28594 }, { "epoch": 0.14, "learning_rate": 1.2239259852923069e-07, "loss": 0.8621, "step": 28595 }, { "epoch": 0.14, "learning_rate": 1.222245315886439e-07, "loss": 0.7955, "step": 28596 }, { "epoch": 0.14, "learning_rate": 1.2205657941081307e-07, "loss": 0.8572, "step": 28597 }, { "epoch": 0.14, "learning_rate": 1.2188874199769218e-07, "loss": 0.8583, "step": 28598 }, { "epoch": 0.14, "learning_rate": 1.217210193512286e-07, "loss": 0.8645, "step": 28599 }, { "epoch": 0.14, "learning_rate": 1.2155341147337295e-07, "loss": 0.8407, "step": 28600 }, { "epoch": 0.14, "learning_rate": 1.2138591836607149e-07, "loss": 0.8355, "step": 28601 }, { "epoch": 0.14, "learning_rate": 1.2121854003127044e-07, "loss": 0.8329, "step": 28602 }, { "epoch": 0.14, "learning_rate": 1.2105127647091486e-07, "loss": 0.8749, "step": 28603 }, { "epoch": 0.14, "learning_rate": 1.208841276869477e-07, "loss": 0.812, "step": 28604 }, { "epoch": 0.14, "learning_rate": 1.207170936813107e-07, "loss": 0.7788, "step": 28605 }, { "epoch": 0.14, "learning_rate": 1.2055017445594452e-07, "loss": 0.8387, "step": 28606 }, { "epoch": 0.14, "learning_rate": 1.2038337001278876e-07, "loss": 0.868, "step": 28607 }, { "epoch": 0.14, "learning_rate": 1.2021668035378187e-07, "loss": 0.833, "step": 28608 }, { "epoch": 0.14, "learning_rate": 1.2005010548085894e-07, "loss": 0.8356, "step": 28609 }, { "epoch": 0.14, "learning_rate": 1.1988364539595841e-07, "loss": 0.8382, "step": 28610 }, { "epoch": 0.14, "learning_rate": 1.1971730010101102e-07, "loss": 0.8541, "step": 28611 }, { "epoch": 0.14, "learning_rate": 1.195510695979496e-07, "loss": 0.8243, "step": 28612 }, { "epoch": 0.14, "learning_rate": 1.1938495388870707e-07, "loss": 0.809, "step": 28613 }, { "epoch": 0.14, "learning_rate": 1.1921895297521303e-07, "loss": 0.8854, "step": 28614 }, { "epoch": 0.14, "learning_rate": 1.1905306685939588e-07, "loss": 0.8413, "step": 28615 }, { "epoch": 0.14, "learning_rate": 1.1888729554318301e-07, "loss": 0.8354, "step": 28616 }, { "epoch": 0.14, "learning_rate": 1.1872163902850064e-07, "loss": 0.8356, "step": 28617 }, { "epoch": 0.14, "learning_rate": 1.1855609731727169e-07, "loss": 0.8432, "step": 28618 }, { "epoch": 0.14, "learning_rate": 1.1839067041142238e-07, "loss": 0.8669, "step": 28619 }, { "epoch": 0.14, "learning_rate": 1.182253583128723e-07, "loss": 0.8612, "step": 28620 }, { "epoch": 0.14, "learning_rate": 1.1806016102354434e-07, "loss": 0.8295, "step": 28621 }, { "epoch": 0.14, "learning_rate": 1.1789507854535475e-07, "loss": 0.8, "step": 28622 }, { "epoch": 0.14, "learning_rate": 1.1773011088022424e-07, "loss": 0.8441, "step": 28623 }, { "epoch": 0.14, "learning_rate": 1.175652580300668e-07, "loss": 0.8467, "step": 28624 }, { "epoch": 0.14, "learning_rate": 1.1740051999680091e-07, "loss": 0.8269, "step": 28625 }, { "epoch": 0.14, "learning_rate": 1.1723589678233838e-07, "loss": 0.8122, "step": 28626 }, { "epoch": 0.14, "learning_rate": 1.1707138838859211e-07, "loss": 0.8604, "step": 28627 }, { "epoch": 0.14, "learning_rate": 1.1690699481747503e-07, "loss": 0.845, "step": 28628 }, { "epoch": 0.14, "learning_rate": 1.167427160708956e-07, "loss": 0.8494, "step": 28629 }, { "epoch": 0.14, "learning_rate": 1.1657855215076119e-07, "loss": 0.8609, "step": 28630 }, { "epoch": 0.14, "learning_rate": 1.1641450305898139e-07, "loss": 0.8596, "step": 28631 }, { "epoch": 0.14, "learning_rate": 1.1625056879746133e-07, "loss": 0.8344, "step": 28632 }, { "epoch": 0.14, "learning_rate": 1.1608674936810616e-07, "loss": 0.84, "step": 28633 }, { "epoch": 0.14, "learning_rate": 1.1592304477281768e-07, "loss": 0.8367, "step": 28634 }, { "epoch": 0.14, "learning_rate": 1.1575945501349994e-07, "loss": 0.8582, "step": 28635 }, { "epoch": 0.14, "learning_rate": 1.1559598009205253e-07, "loss": 0.8422, "step": 28636 }, { "epoch": 0.14, "learning_rate": 1.1543262001037282e-07, "loss": 0.8145, "step": 28637 }, { "epoch": 0.14, "learning_rate": 1.152693747703626e-07, "loss": 0.8293, "step": 28638 }, { "epoch": 0.14, "learning_rate": 1.1510624437391593e-07, "loss": 0.8612, "step": 28639 }, { "epoch": 0.14, "learning_rate": 1.1494322882292908e-07, "loss": 0.8085, "step": 28640 }, { "epoch": 0.14, "learning_rate": 1.1478032811929495e-07, "loss": 0.8473, "step": 28641 }, { "epoch": 0.14, "learning_rate": 1.1461754226490762e-07, "loss": 0.859, "step": 28642 }, { "epoch": 0.14, "learning_rate": 1.1445487126165778e-07, "loss": 0.8547, "step": 28643 }, { "epoch": 0.14, "learning_rate": 1.1429231511143501e-07, "loss": 0.8477, "step": 28644 }, { "epoch": 0.14, "learning_rate": 1.1412987381612783e-07, "loss": 0.8407, "step": 28645 }, { "epoch": 0.14, "learning_rate": 1.1396754737762472e-07, "loss": 0.8428, "step": 28646 }, { "epoch": 0.14, "learning_rate": 1.1380533579780972e-07, "loss": 0.8749, "step": 28647 }, { "epoch": 0.14, "learning_rate": 1.1364323907856911e-07, "loss": 0.8771, "step": 28648 }, { "epoch": 0.14, "learning_rate": 1.1348125722178471e-07, "loss": 0.8613, "step": 28649 }, { "epoch": 0.14, "learning_rate": 1.1331939022934057e-07, "loss": 0.7979, "step": 28650 }, { "epoch": 0.14, "learning_rate": 1.131576381031152e-07, "loss": 0.834, "step": 28651 }, { "epoch": 0.14, "learning_rate": 1.129960008449893e-07, "loss": 0.799, "step": 28652 }, { "epoch": 0.14, "learning_rate": 1.1283447845684026e-07, "loss": 0.8222, "step": 28653 }, { "epoch": 0.14, "learning_rate": 1.1267307094054436e-07, "loss": 0.8463, "step": 28654 }, { "epoch": 0.14, "learning_rate": 1.1251177829797676e-07, "loss": 0.8628, "step": 28655 }, { "epoch": 0.14, "learning_rate": 1.1235060053101376e-07, "loss": 0.8467, "step": 28656 }, { "epoch": 0.14, "learning_rate": 1.1218953764152496e-07, "loss": 0.8523, "step": 28657 }, { "epoch": 0.14, "learning_rate": 1.1202858963138219e-07, "loss": 0.8383, "step": 28658 }, { "epoch": 0.14, "learning_rate": 1.1186775650245729e-07, "loss": 0.8443, "step": 28659 }, { "epoch": 0.14, "learning_rate": 1.1170703825661655e-07, "loss": 0.8694, "step": 28660 }, { "epoch": 0.14, "learning_rate": 1.1154643489572736e-07, "loss": 0.8427, "step": 28661 }, { "epoch": 0.14, "learning_rate": 1.1138594642165712e-07, "loss": 0.8327, "step": 28662 }, { "epoch": 0.14, "learning_rate": 1.11225572836271e-07, "loss": 0.8959, "step": 28663 }, { "epoch": 0.14, "learning_rate": 1.1106531414142974e-07, "loss": 0.8501, "step": 28664 }, { "epoch": 0.14, "learning_rate": 1.1090517033899739e-07, "loss": 0.8585, "step": 28665 }, { "epoch": 0.14, "learning_rate": 1.1074514143083359e-07, "loss": 0.8412, "step": 28666 }, { "epoch": 0.14, "learning_rate": 1.1058522741879795e-07, "loss": 0.8597, "step": 28667 }, { "epoch": 0.14, "learning_rate": 1.104254283047479e-07, "loss": 0.8558, "step": 28668 }, { "epoch": 0.14, "learning_rate": 1.102657440905397e-07, "loss": 0.868, "step": 28669 }, { "epoch": 0.14, "learning_rate": 1.1010617477803076e-07, "loss": 0.8797, "step": 28670 }, { "epoch": 0.14, "learning_rate": 1.0994672036907294e-07, "loss": 0.8226, "step": 28671 }, { "epoch": 0.14, "learning_rate": 1.0978738086551921e-07, "loss": 0.8553, "step": 28672 }, { "epoch": 0.14, "learning_rate": 1.096281562692214e-07, "loss": 0.8397, "step": 28673 }, { "epoch": 0.14, "learning_rate": 1.0946904658202806e-07, "loss": 0.8254, "step": 28674 }, { "epoch": 0.14, "learning_rate": 1.0931005180578991e-07, "loss": 0.915, "step": 28675 }, { "epoch": 0.14, "learning_rate": 1.0915117194235214e-07, "loss": 0.8488, "step": 28676 }, { "epoch": 0.14, "learning_rate": 1.0899240699356217e-07, "loss": 0.78, "step": 28677 }, { "epoch": 0.14, "learning_rate": 1.0883375696126408e-07, "loss": 0.8266, "step": 28678 }, { "epoch": 0.14, "learning_rate": 1.0867522184729973e-07, "loss": 0.8588, "step": 28679 }, { "epoch": 0.14, "learning_rate": 1.085168016535132e-07, "loss": 0.8193, "step": 28680 }, { "epoch": 0.14, "learning_rate": 1.0835849638174411e-07, "loss": 0.8718, "step": 28681 }, { "epoch": 0.14, "learning_rate": 1.0820030603383214e-07, "loss": 0.8066, "step": 28682 }, { "epoch": 0.14, "learning_rate": 1.0804223061161357e-07, "loss": 0.8624, "step": 28683 }, { "epoch": 0.14, "learning_rate": 1.0788427011692803e-07, "loss": 0.8173, "step": 28684 }, { "epoch": 0.14, "learning_rate": 1.0772642455160632e-07, "loss": 0.8342, "step": 28685 }, { "epoch": 0.14, "learning_rate": 1.0756869391748692e-07, "loss": 0.8344, "step": 28686 }, { "epoch": 0.14, "learning_rate": 1.0741107821639951e-07, "loss": 0.8407, "step": 28687 }, { "epoch": 0.14, "learning_rate": 1.0725357745017706e-07, "loss": 0.776, "step": 28688 }, { "epoch": 0.14, "learning_rate": 1.070961916206481e-07, "loss": 0.8857, "step": 28689 }, { "epoch": 0.14, "learning_rate": 1.0693892072964119e-07, "loss": 0.8604, "step": 28690 }, { "epoch": 0.14, "learning_rate": 1.0678176477898372e-07, "loss": 0.8446, "step": 28691 }, { "epoch": 0.14, "learning_rate": 1.0662472377050314e-07, "loss": 0.7822, "step": 28692 }, { "epoch": 0.14, "learning_rate": 1.0646779770602134e-07, "loss": 0.8054, "step": 28693 }, { "epoch": 0.14, "learning_rate": 1.0631098658736349e-07, "loss": 0.8245, "step": 28694 }, { "epoch": 0.14, "learning_rate": 1.061542904163515e-07, "loss": 0.877, "step": 28695 }, { "epoch": 0.14, "learning_rate": 1.0599770919480502e-07, "loss": 0.8247, "step": 28696 }, { "epoch": 0.14, "learning_rate": 1.0584124292454368e-07, "loss": 0.8927, "step": 28697 }, { "epoch": 0.14, "learning_rate": 1.0568489160738493e-07, "loss": 0.8329, "step": 28698 }, { "epoch": 0.14, "learning_rate": 1.0552865524514511e-07, "loss": 0.8369, "step": 28699 }, { "epoch": 0.14, "learning_rate": 1.0537253383964052e-07, "loss": 0.8396, "step": 28700 }, { "epoch": 0.14, "learning_rate": 1.0521652739268417e-07, "loss": 0.8032, "step": 28701 }, { "epoch": 0.14, "learning_rate": 1.0506063590608905e-07, "loss": 0.8621, "step": 28702 }, { "epoch": 0.14, "learning_rate": 1.0490485938166594e-07, "loss": 0.8251, "step": 28703 }, { "epoch": 0.14, "learning_rate": 1.047491978212234e-07, "loss": 0.8524, "step": 28704 }, { "epoch": 0.14, "learning_rate": 1.0459365122657327e-07, "loss": 0.8252, "step": 28705 }, { "epoch": 0.14, "learning_rate": 1.0443821959952083e-07, "loss": 0.8369, "step": 28706 }, { "epoch": 0.14, "learning_rate": 1.0428290294187127e-07, "loss": 0.869, "step": 28707 }, { "epoch": 0.14, "learning_rate": 1.0412770125542982e-07, "loss": 0.7976, "step": 28708 }, { "epoch": 0.14, "learning_rate": 1.0397261454200058e-07, "loss": 0.8705, "step": 28709 }, { "epoch": 0.14, "learning_rate": 1.0381764280338213e-07, "loss": 0.8475, "step": 28710 }, { "epoch": 0.14, "learning_rate": 1.0366278604137858e-07, "loss": 0.862, "step": 28711 }, { "epoch": 0.14, "learning_rate": 1.0350804425778849e-07, "loss": 0.8155, "step": 28712 }, { "epoch": 0.14, "learning_rate": 1.033534174544093e-07, "loss": 0.8525, "step": 28713 }, { "epoch": 0.14, "learning_rate": 1.0319890563303625e-07, "loss": 0.8242, "step": 28714 }, { "epoch": 0.14, "learning_rate": 1.0304450879546568e-07, "loss": 0.8279, "step": 28715 }, { "epoch": 0.14, "learning_rate": 1.028902269434906e-07, "loss": 0.8948, "step": 28716 }, { "epoch": 0.14, "learning_rate": 1.0273606007890514e-07, "loss": 0.8688, "step": 28717 }, { "epoch": 0.14, "learning_rate": 1.0258200820349895e-07, "loss": 0.8034, "step": 28718 }, { "epoch": 0.14, "learning_rate": 1.0242807131906285e-07, "loss": 0.8512, "step": 28719 }, { "epoch": 0.14, "learning_rate": 1.0227424942738429e-07, "loss": 0.8444, "step": 28720 }, { "epoch": 0.14, "learning_rate": 1.0212054253025072e-07, "loss": 0.8491, "step": 28721 }, { "epoch": 0.14, "learning_rate": 1.019669506294485e-07, "loss": 0.7983, "step": 28722 }, { "epoch": 0.14, "learning_rate": 1.0181347372676175e-07, "loss": 0.8572, "step": 28723 }, { "epoch": 0.14, "learning_rate": 1.0166011182397351e-07, "loss": 0.86, "step": 28724 }, { "epoch": 0.14, "learning_rate": 1.0150686492286566e-07, "loss": 0.8022, "step": 28725 }, { "epoch": 0.14, "learning_rate": 1.0135373302521901e-07, "loss": 0.8212, "step": 28726 }, { "epoch": 0.14, "learning_rate": 1.0120071613281213e-07, "loss": 0.8173, "step": 28727 }, { "epoch": 0.14, "learning_rate": 1.010478142474225e-07, "loss": 0.8431, "step": 28728 }, { "epoch": 0.14, "learning_rate": 1.0089502737082757e-07, "loss": 0.8867, "step": 28729 }, { "epoch": 0.14, "learning_rate": 1.007423555048026e-07, "loss": 0.8593, "step": 28730 }, { "epoch": 0.14, "learning_rate": 1.005897986511195e-07, "loss": 0.8883, "step": 28731 }, { "epoch": 0.14, "learning_rate": 1.0043735681155243e-07, "loss": 0.8931, "step": 28732 }, { "epoch": 0.14, "learning_rate": 1.0028502998787326e-07, "loss": 0.8693, "step": 28733 }, { "epoch": 0.14, "learning_rate": 1.001328181818495e-07, "loss": 0.8324, "step": 28734 }, { "epoch": 0.14, "learning_rate": 9.998072139525083e-08, "loss": 0.8882, "step": 28735 }, { "epoch": 0.14, "learning_rate": 9.982873962984363e-08, "loss": 0.8559, "step": 28736 }, { "epoch": 0.14, "learning_rate": 9.967687288739536e-08, "loss": 0.8413, "step": 28737 }, { "epoch": 0.14, "learning_rate": 9.952512116966907e-08, "loss": 0.849, "step": 28738 }, { "epoch": 0.14, "learning_rate": 9.937348447842776e-08, "loss": 0.8856, "step": 28739 }, { "epoch": 0.14, "learning_rate": 9.92219628154334e-08, "loss": 0.8189, "step": 28740 }, { "epoch": 0.14, "learning_rate": 9.907055618244676e-08, "loss": 0.8182, "step": 28741 }, { "epoch": 0.14, "learning_rate": 9.891926458122647e-08, "loss": 0.8612, "step": 28742 }, { "epoch": 0.14, "learning_rate": 9.87680880135311e-08, "loss": 0.8831, "step": 28743 }, { "epoch": 0.14, "learning_rate": 9.861702648111594e-08, "loss": 0.834, "step": 28744 }, { "epoch": 0.14, "learning_rate": 9.846607998573732e-08, "loss": 0.8428, "step": 28745 }, { "epoch": 0.14, "learning_rate": 9.83152485291472e-08, "loss": 0.8306, "step": 28746 }, { "epoch": 0.14, "learning_rate": 9.816453211309973e-08, "loss": 0.8974, "step": 28747 }, { "epoch": 0.14, "learning_rate": 9.801393073934462e-08, "loss": 0.8162, "step": 28748 }, { "epoch": 0.14, "learning_rate": 9.78634444096338e-08, "loss": 0.8481, "step": 28749 }, { "epoch": 0.14, "learning_rate": 9.771307312571254e-08, "loss": 0.8672, "step": 28750 }, { "epoch": 0.14, "learning_rate": 9.756281688933056e-08, "loss": 0.847, "step": 28751 }, { "epoch": 0.14, "learning_rate": 9.7412675702232e-08, "loss": 0.8249, "step": 28752 }, { "epoch": 0.14, "learning_rate": 9.726264956616104e-08, "loss": 0.8438, "step": 28753 }, { "epoch": 0.14, "learning_rate": 9.711273848286185e-08, "loss": 0.8323, "step": 28754 }, { "epoch": 0.14, "learning_rate": 9.696294245407744e-08, "loss": 0.85, "step": 28755 }, { "epoch": 0.14, "learning_rate": 9.681326148154424e-08, "loss": 0.7933, "step": 28756 }, { "epoch": 0.14, "learning_rate": 9.666369556700416e-08, "loss": 0.8433, "step": 28757 }, { "epoch": 0.14, "learning_rate": 9.651424471219472e-08, "loss": 0.844, "step": 28758 }, { "epoch": 0.14, "learning_rate": 9.636490891885119e-08, "loss": 0.8656, "step": 28759 }, { "epoch": 0.14, "learning_rate": 9.621568818870997e-08, "loss": 0.8611, "step": 28760 }, { "epoch": 0.14, "learning_rate": 9.606658252350298e-08, "loss": 0.8537, "step": 28761 }, { "epoch": 0.14, "learning_rate": 9.591759192496552e-08, "loss": 0.8321, "step": 28762 }, { "epoch": 0.14, "learning_rate": 9.57687163948251e-08, "loss": 0.859, "step": 28763 }, { "epoch": 0.14, "learning_rate": 9.561995593481255e-08, "loss": 0.8031, "step": 28764 }, { "epoch": 0.14, "learning_rate": 9.547131054665759e-08, "loss": 0.8495, "step": 28765 }, { "epoch": 0.14, "learning_rate": 9.532278023208553e-08, "loss": 0.8181, "step": 28766 }, { "epoch": 0.14, "learning_rate": 9.517436499282385e-08, "loss": 0.8389, "step": 28767 }, { "epoch": 0.14, "learning_rate": 9.502606483059451e-08, "loss": 0.8176, "step": 28768 }, { "epoch": 0.14, "learning_rate": 9.487787974712281e-08, "loss": 0.8239, "step": 28769 }, { "epoch": 0.14, "learning_rate": 9.47298097441296e-08, "loss": 0.8757, "step": 28770 }, { "epoch": 0.14, "learning_rate": 9.458185482333349e-08, "loss": 0.8527, "step": 28771 }, { "epoch": 0.14, "learning_rate": 9.443401498645532e-08, "loss": 0.8414, "step": 28772 }, { "epoch": 0.14, "learning_rate": 9.428629023521374e-08, "loss": 0.7984, "step": 28773 }, { "epoch": 0.14, "learning_rate": 9.413868057132291e-08, "loss": 0.8607, "step": 28774 }, { "epoch": 0.14, "learning_rate": 9.399118599649815e-08, "loss": 0.8494, "step": 28775 }, { "epoch": 0.14, "learning_rate": 9.384380651245472e-08, "loss": 0.8668, "step": 28776 }, { "epoch": 0.14, "learning_rate": 9.36965421209024e-08, "loss": 0.894, "step": 28777 }, { "epoch": 0.14, "learning_rate": 9.354939282355313e-08, "loss": 0.8275, "step": 28778 }, { "epoch": 0.14, "learning_rate": 9.340235862211777e-08, "loss": 0.8917, "step": 28779 }, { "epoch": 0.14, "learning_rate": 9.325543951830495e-08, "loss": 0.8364, "step": 28780 }, { "epoch": 0.14, "learning_rate": 9.310863551381888e-08, "loss": 0.8432, "step": 28781 }, { "epoch": 0.14, "learning_rate": 9.296194661036594e-08, "loss": 0.8576, "step": 28782 }, { "epoch": 0.14, "learning_rate": 9.281537280965258e-08, "loss": 0.8292, "step": 28783 }, { "epoch": 0.14, "learning_rate": 9.266891411337963e-08, "loss": 0.7902, "step": 28784 }, { "epoch": 0.14, "learning_rate": 9.252257052324909e-08, "loss": 0.8182, "step": 28785 }, { "epoch": 0.14, "learning_rate": 9.23763420409618e-08, "loss": 0.8062, "step": 28786 }, { "epoch": 0.14, "learning_rate": 9.223022866821751e-08, "loss": 0.8605, "step": 28787 }, { "epoch": 0.14, "learning_rate": 9.208423040671045e-08, "loss": 0.8547, "step": 28788 }, { "epoch": 0.14, "learning_rate": 9.193834725814033e-08, "loss": 0.8262, "step": 28789 }, { "epoch": 0.14, "learning_rate": 9.179257922420027e-08, "loss": 0.8406, "step": 28790 }, { "epoch": 0.14, "learning_rate": 9.164692630658556e-08, "loss": 0.8554, "step": 28791 }, { "epoch": 0.14, "learning_rate": 9.150138850698598e-08, "loss": 0.8479, "step": 28792 }, { "epoch": 0.14, "learning_rate": 9.13559658270935e-08, "loss": 0.8589, "step": 28793 }, { "epoch": 0.14, "learning_rate": 9.121065826859899e-08, "loss": 0.8632, "step": 28794 }, { "epoch": 0.14, "learning_rate": 9.106546583318888e-08, "loss": 0.9041, "step": 28795 }, { "epoch": 0.14, "learning_rate": 9.092038852255069e-08, "loss": 0.8264, "step": 28796 }, { "epoch": 0.14, "learning_rate": 9.077542633836977e-08, "loss": 0.8199, "step": 28797 }, { "epoch": 0.14, "learning_rate": 9.063057928233143e-08, "loss": 0.8364, "step": 28798 }, { "epoch": 0.14, "learning_rate": 9.048584735611765e-08, "loss": 0.8402, "step": 28799 }, { "epoch": 0.14, "learning_rate": 9.034123056140931e-08, "loss": 0.8767, "step": 28800 }, { "epoch": 0.14, "learning_rate": 9.01967288998884e-08, "loss": 0.832, "step": 28801 }, { "epoch": 0.14, "learning_rate": 9.005234237323134e-08, "loss": 0.8564, "step": 28802 }, { "epoch": 0.14, "learning_rate": 8.990807098311794e-08, "loss": 0.8529, "step": 28803 }, { "epoch": 0.14, "learning_rate": 8.976391473122237e-08, "loss": 0.8173, "step": 28804 }, { "epoch": 0.14, "learning_rate": 8.96198736192222e-08, "loss": 0.8122, "step": 28805 }, { "epoch": 0.14, "learning_rate": 8.947594764878831e-08, "loss": 0.8353, "step": 28806 }, { "epoch": 0.14, "learning_rate": 8.933213682159381e-08, "loss": 0.8386, "step": 28807 }, { "epoch": 0.14, "learning_rate": 8.918844113930958e-08, "loss": 0.8563, "step": 28808 }, { "epoch": 0.14, "learning_rate": 8.90448606036054e-08, "loss": 0.8151, "step": 28809 }, { "epoch": 0.14, "learning_rate": 8.890139521614882e-08, "loss": 0.8182, "step": 28810 }, { "epoch": 0.14, "learning_rate": 8.87580449786074e-08, "loss": 0.8687, "step": 28811 }, { "epoch": 0.14, "learning_rate": 8.861480989264648e-08, "loss": 0.8256, "step": 28812 }, { "epoch": 0.14, "learning_rate": 8.847168995992916e-08, "loss": 0.8427, "step": 28813 }, { "epoch": 0.14, "learning_rate": 8.832868518211857e-08, "loss": 0.8362, "step": 28814 }, { "epoch": 0.14, "learning_rate": 8.818579556087781e-08, "loss": 0.8759, "step": 28815 }, { "epoch": 0.14, "learning_rate": 8.804302109786444e-08, "loss": 0.8519, "step": 28816 }, { "epoch": 0.14, "learning_rate": 8.790036179474048e-08, "loss": 0.824, "step": 28817 }, { "epoch": 0.14, "learning_rate": 8.775781765316016e-08, "loss": 0.8636, "step": 28818 }, { "epoch": 0.14, "learning_rate": 8.761538867478214e-08, "loss": 0.8542, "step": 28819 }, { "epoch": 0.14, "learning_rate": 8.747307486125845e-08, "loss": 0.8594, "step": 28820 }, { "epoch": 0.14, "learning_rate": 8.733087621424553e-08, "loss": 0.8466, "step": 28821 }, { "epoch": 0.14, "learning_rate": 8.718879273539316e-08, "loss": 0.8082, "step": 28822 }, { "epoch": 0.14, "learning_rate": 8.704682442635337e-08, "loss": 0.8625, "step": 28823 }, { "epoch": 0.14, "learning_rate": 8.690497128877595e-08, "loss": 0.8577, "step": 28824 }, { "epoch": 0.14, "learning_rate": 8.676323332430736e-08, "loss": 0.8543, "step": 28825 }, { "epoch": 0.14, "learning_rate": 8.662161053459516e-08, "loss": 0.8625, "step": 28826 }, { "epoch": 0.14, "learning_rate": 8.648010292128583e-08, "loss": 0.8367, "step": 28827 }, { "epoch": 0.14, "learning_rate": 8.633871048602139e-08, "loss": 0.8432, "step": 28828 }, { "epoch": 0.14, "learning_rate": 8.619743323044716e-08, "loss": 0.8534, "step": 28829 }, { "epoch": 0.14, "learning_rate": 8.605627115620185e-08, "loss": 0.869, "step": 28830 }, { "epoch": 0.14, "learning_rate": 8.591522426492749e-08, "loss": 0.8487, "step": 28831 }, { "epoch": 0.14, "learning_rate": 8.577429255826164e-08, "loss": 0.8569, "step": 28832 }, { "epoch": 0.14, "learning_rate": 8.563347603784189e-08, "loss": 0.8524, "step": 28833 }, { "epoch": 0.14, "learning_rate": 8.54927747053047e-08, "loss": 0.896, "step": 28834 }, { "epoch": 0.14, "learning_rate": 8.535218856228434e-08, "loss": 0.8713, "step": 28835 }, { "epoch": 0.14, "learning_rate": 8.521171761041502e-08, "loss": 0.8496, "step": 28836 }, { "epoch": 0.14, "learning_rate": 8.507136185132769e-08, "loss": 0.8657, "step": 28837 }, { "epoch": 0.14, "learning_rate": 8.493112128665326e-08, "loss": 0.8368, "step": 28838 }, { "epoch": 0.14, "learning_rate": 8.479099591802154e-08, "loss": 0.8777, "step": 28839 }, { "epoch": 0.14, "learning_rate": 8.46509857470601e-08, "loss": 0.8695, "step": 28840 }, { "epoch": 0.14, "learning_rate": 8.451109077539654e-08, "loss": 0.8853, "step": 28841 }, { "epoch": 0.14, "learning_rate": 8.437131100465512e-08, "loss": 0.8344, "step": 28842 }, { "epoch": 0.14, "learning_rate": 8.423164643646009e-08, "loss": 0.854, "step": 28843 }, { "epoch": 0.14, "learning_rate": 8.40920970724346e-08, "loss": 0.8262, "step": 28844 }, { "epoch": 0.14, "learning_rate": 8.395266291419845e-08, "loss": 0.8288, "step": 28845 }, { "epoch": 0.14, "learning_rate": 8.38133439633737e-08, "loss": 0.8655, "step": 28846 }, { "epoch": 0.14, "learning_rate": 8.367414022157904e-08, "loss": 0.8626, "step": 28847 }, { "epoch": 0.14, "learning_rate": 8.353505169042986e-08, "loss": 0.8386, "step": 28848 }, { "epoch": 0.14, "learning_rate": 8.339607837154373e-08, "loss": 0.8296, "step": 28849 }, { "epoch": 0.14, "learning_rate": 8.325722026653383e-08, "loss": 0.8751, "step": 28850 }, { "epoch": 0.14, "learning_rate": 8.311847737701439e-08, "loss": 0.8195, "step": 28851 }, { "epoch": 0.14, "learning_rate": 8.297984970459749e-08, "loss": 0.8775, "step": 28852 }, { "epoch": 0.14, "learning_rate": 8.284133725089294e-08, "loss": 0.827, "step": 28853 }, { "epoch": 0.14, "learning_rate": 8.270294001751167e-08, "loss": 0.8737, "step": 28854 }, { "epoch": 0.14, "learning_rate": 8.256465800606018e-08, "loss": 0.8559, "step": 28855 }, { "epoch": 0.14, "learning_rate": 8.242649121814494e-08, "loss": 0.8565, "step": 28856 }, { "epoch": 0.15, "learning_rate": 8.228843965537137e-08, "loss": 0.8338, "step": 28857 }, { "epoch": 0.15, "learning_rate": 8.215050331934482e-08, "loss": 0.8541, "step": 28858 }, { "epoch": 0.15, "learning_rate": 8.201268221166625e-08, "loss": 0.8203, "step": 28859 }, { "epoch": 0.15, "learning_rate": 8.18749763339377e-08, "loss": 0.8078, "step": 28860 }, { "epoch": 0.15, "learning_rate": 8.173738568775902e-08, "loss": 0.8519, "step": 28861 }, { "epoch": 0.15, "learning_rate": 8.15999102747278e-08, "loss": 0.8257, "step": 28862 }, { "epoch": 0.15, "learning_rate": 8.146255009644166e-08, "loss": 0.838, "step": 28863 }, { "epoch": 0.15, "learning_rate": 8.13253051544971e-08, "loss": 0.8533, "step": 28864 }, { "epoch": 0.15, "learning_rate": 8.118817545048951e-08, "loss": 0.8312, "step": 28865 }, { "epoch": 0.15, "learning_rate": 8.105116098600985e-08, "loss": 0.8161, "step": 28866 }, { "epoch": 0.15, "learning_rate": 8.091426176265127e-08, "loss": 0.8837, "step": 28867 }, { "epoch": 0.15, "learning_rate": 8.077747778200474e-08, "loss": 0.8192, "step": 28868 }, { "epoch": 0.15, "learning_rate": 8.064080904565896e-08, "loss": 0.8428, "step": 28869 }, { "epoch": 0.15, "learning_rate": 8.050425555520047e-08, "loss": 0.837, "step": 28870 }, { "epoch": 0.15, "learning_rate": 8.036781731221799e-08, "loss": 0.8697, "step": 28871 }, { "epoch": 0.15, "learning_rate": 8.023149431829691e-08, "loss": 0.8301, "step": 28872 }, { "epoch": 0.15, "learning_rate": 8.00952865750182e-08, "loss": 0.8592, "step": 28873 }, { "epoch": 0.15, "learning_rate": 7.995919408396614e-08, "loss": 0.8337, "step": 28874 }, { "epoch": 0.15, "learning_rate": 7.982321684672279e-08, "loss": 0.928, "step": 28875 }, { "epoch": 0.15, "learning_rate": 7.968735486486579e-08, "loss": 0.8225, "step": 28876 }, { "epoch": 0.15, "learning_rate": 7.955160813997609e-08, "loss": 0.8476, "step": 28877 }, { "epoch": 0.15, "learning_rate": 7.94159766736291e-08, "loss": 0.8149, "step": 28878 }, { "epoch": 0.15, "learning_rate": 7.928046046740134e-08, "loss": 0.8502, "step": 28879 }, { "epoch": 0.15, "learning_rate": 7.914505952286711e-08, "loss": 0.8502, "step": 28880 }, { "epoch": 0.15, "learning_rate": 7.900977384159846e-08, "loss": 0.819, "step": 28881 }, { "epoch": 0.15, "learning_rate": 7.887460342516861e-08, "loss": 0.8573, "step": 28882 }, { "epoch": 0.15, "learning_rate": 7.873954827514741e-08, "loss": 0.8742, "step": 28883 }, { "epoch": 0.15, "learning_rate": 7.860460839310468e-08, "loss": 0.8243, "step": 28884 }, { "epoch": 0.15, "learning_rate": 7.846978378060699e-08, "loss": 0.8628, "step": 28885 }, { "epoch": 0.15, "learning_rate": 7.833507443922306e-08, "loss": 0.8296, "step": 28886 }, { "epoch": 0.15, "learning_rate": 7.820048037051498e-08, "loss": 0.8773, "step": 28887 }, { "epoch": 0.15, "learning_rate": 7.806600157604704e-08, "loss": 0.8662, "step": 28888 }, { "epoch": 0.15, "learning_rate": 7.793163805738358e-08, "loss": 0.8431, "step": 28889 }, { "epoch": 0.15, "learning_rate": 7.779738981608554e-08, "loss": 0.8061, "step": 28890 }, { "epoch": 0.15, "learning_rate": 7.766325685371057e-08, "loss": 0.8871, "step": 28891 }, { "epoch": 0.15, "learning_rate": 7.752923917181743e-08, "loss": 0.8398, "step": 28892 }, { "epoch": 0.15, "learning_rate": 7.739533677196486e-08, "loss": 0.825, "step": 28893 }, { "epoch": 0.15, "learning_rate": 7.72615496557072e-08, "loss": 0.7959, "step": 28894 }, { "epoch": 0.15, "learning_rate": 7.712787782459985e-08, "loss": 0.8376, "step": 28895 }, { "epoch": 0.15, "learning_rate": 7.69943212801949e-08, "loss": 0.8422, "step": 28896 }, { "epoch": 0.15, "learning_rate": 7.686088002404446e-08, "loss": 0.8022, "step": 28897 }, { "epoch": 0.15, "learning_rate": 7.67275540576995e-08, "loss": 0.8717, "step": 28898 }, { "epoch": 0.15, "learning_rate": 7.659434338270766e-08, "loss": 0.8147, "step": 28899 }, { "epoch": 0.15, "learning_rate": 7.646124800061771e-08, "loss": 0.844, "step": 28900 }, { "epoch": 0.15, "learning_rate": 7.63282679129751e-08, "loss": 0.8151, "step": 28901 }, { "epoch": 0.15, "learning_rate": 7.619540312132524e-08, "loss": 0.8245, "step": 28902 }, { "epoch": 0.15, "learning_rate": 7.606265362721244e-08, "loss": 0.8244, "step": 28903 }, { "epoch": 0.15, "learning_rate": 7.593001943217881e-08, "loss": 0.8047, "step": 28904 }, { "epoch": 0.15, "learning_rate": 7.579750053776425e-08, "loss": 0.8467, "step": 28905 }, { "epoch": 0.15, "learning_rate": 7.566509694550861e-08, "loss": 0.8184, "step": 28906 }, { "epoch": 0.15, "learning_rate": 7.553280865695179e-08, "loss": 0.8635, "step": 28907 }, { "epoch": 0.15, "learning_rate": 7.540063567362921e-08, "loss": 0.8224, "step": 28908 }, { "epoch": 0.15, "learning_rate": 7.526857799707743e-08, "loss": 0.8511, "step": 28909 }, { "epoch": 0.15, "learning_rate": 7.513663562882967e-08, "loss": 0.7835, "step": 28910 }, { "epoch": 0.15, "learning_rate": 7.500480857042025e-08, "loss": 0.8381, "step": 28911 }, { "epoch": 0.15, "learning_rate": 7.487309682337906e-08, "loss": 0.8725, "step": 28912 }, { "epoch": 0.15, "learning_rate": 7.474150038923711e-08, "loss": 0.8412, "step": 28913 }, { "epoch": 0.15, "learning_rate": 7.461001926952537e-08, "loss": 0.8296, "step": 28914 }, { "epoch": 0.15, "learning_rate": 7.447865346576821e-08, "loss": 0.8192, "step": 28915 }, { "epoch": 0.15, "learning_rate": 7.434740297949328e-08, "loss": 0.82, "step": 28916 }, { "epoch": 0.15, "learning_rate": 7.421626781222602e-08, "loss": 0.8675, "step": 28917 }, { "epoch": 0.15, "learning_rate": 7.408524796548855e-08, "loss": 0.8232, "step": 28918 }, { "epoch": 0.15, "learning_rate": 7.39543434408041e-08, "loss": 0.8202, "step": 28919 }, { "epoch": 0.15, "learning_rate": 7.382355423969367e-08, "loss": 0.8527, "step": 28920 }, { "epoch": 0.15, "learning_rate": 7.369288036367717e-08, "loss": 0.8532, "step": 28921 }, { "epoch": 0.15, "learning_rate": 7.356232181427225e-08, "loss": 0.8375, "step": 28922 }, { "epoch": 0.15, "learning_rate": 7.343187859299439e-08, "loss": 0.8478, "step": 28923 }, { "epoch": 0.15, "learning_rate": 7.330155070136125e-08, "loss": 0.8711, "step": 28924 }, { "epoch": 0.15, "learning_rate": 7.317133814088606e-08, "loss": 0.8477, "step": 28925 }, { "epoch": 0.15, "learning_rate": 7.304124091308207e-08, "loss": 0.8533, "step": 28926 }, { "epoch": 0.15, "learning_rate": 7.291125901946027e-08, "loss": 0.9098, "step": 28927 }, { "epoch": 0.15, "learning_rate": 7.278139246153171e-08, "loss": 0.8264, "step": 28928 }, { "epoch": 0.15, "learning_rate": 7.265164124080404e-08, "loss": 0.8385, "step": 28929 }, { "epoch": 0.15, "learning_rate": 7.252200535878496e-08, "loss": 0.872, "step": 28930 }, { "epoch": 0.15, "learning_rate": 7.239248481698103e-08, "loss": 0.8574, "step": 28931 }, { "epoch": 0.15, "learning_rate": 7.226307961689771e-08, "loss": 0.8501, "step": 28932 }, { "epoch": 0.15, "learning_rate": 7.213378976003715e-08, "loss": 0.8559, "step": 28933 }, { "epoch": 0.15, "learning_rate": 7.200461524790258e-08, "loss": 0.8233, "step": 28934 }, { "epoch": 0.15, "learning_rate": 7.187555608199282e-08, "loss": 0.8425, "step": 28935 }, { "epoch": 0.15, "learning_rate": 7.174661226380997e-08, "loss": 0.8429, "step": 28936 }, { "epoch": 0.15, "learning_rate": 7.161778379485063e-08, "loss": 0.8405, "step": 28937 }, { "epoch": 0.15, "learning_rate": 7.148907067661249e-08, "loss": 0.8719, "step": 28938 }, { "epoch": 0.15, "learning_rate": 7.13604729105899e-08, "loss": 0.8204, "step": 28939 }, { "epoch": 0.15, "learning_rate": 7.123199049827723e-08, "loss": 0.8434, "step": 28940 }, { "epoch": 0.15, "learning_rate": 7.110362344116772e-08, "loss": 0.8553, "step": 28941 }, { "epoch": 0.15, "learning_rate": 7.097537174075242e-08, "loss": 0.8632, "step": 28942 }, { "epoch": 0.15, "learning_rate": 7.084723539852122e-08, "loss": 0.8582, "step": 28943 }, { "epoch": 0.15, "learning_rate": 7.071921441596297e-08, "loss": 0.8269, "step": 28944 }, { "epoch": 0.15, "learning_rate": 7.059130879456532e-08, "loss": 0.8115, "step": 28945 }, { "epoch": 0.15, "learning_rate": 7.04635185358149e-08, "loss": 0.8408, "step": 28946 }, { "epoch": 0.15, "learning_rate": 7.033584364119494e-08, "loss": 0.8509, "step": 28947 }, { "epoch": 0.15, "learning_rate": 7.020828411218872e-08, "loss": 0.8024, "step": 28948 }, { "epoch": 0.15, "learning_rate": 7.008083995027948e-08, "loss": 0.8185, "step": 28949 }, { "epoch": 0.15, "learning_rate": 6.995351115694826e-08, "loss": 0.8328, "step": 28950 }, { "epoch": 0.15, "learning_rate": 6.982629773367278e-08, "loss": 0.8718, "step": 28951 }, { "epoch": 0.15, "learning_rate": 6.969919968193184e-08, "loss": 0.8629, "step": 28952 }, { "epoch": 0.15, "learning_rate": 6.957221700320316e-08, "loss": 0.8853, "step": 28953 }, { "epoch": 0.15, "learning_rate": 6.944534969896e-08, "loss": 0.8544, "step": 28954 }, { "epoch": 0.15, "learning_rate": 6.931859777067673e-08, "loss": 0.866, "step": 28955 }, { "epoch": 0.15, "learning_rate": 6.919196121982774e-08, "loss": 0.8168, "step": 28956 }, { "epoch": 0.15, "learning_rate": 6.906544004788296e-08, "loss": 0.8594, "step": 28957 }, { "epoch": 0.15, "learning_rate": 6.893903425631233e-08, "loss": 0.8751, "step": 28958 }, { "epoch": 0.15, "learning_rate": 6.881274384658576e-08, "loss": 0.8443, "step": 28959 }, { "epoch": 0.15, "learning_rate": 6.868656882016877e-08, "loss": 0.8887, "step": 28960 }, { "epoch": 0.15, "learning_rate": 6.856050917852685e-08, "loss": 0.8641, "step": 28961 }, { "epoch": 0.15, "learning_rate": 6.843456492312772e-08, "loss": 0.8726, "step": 28962 }, { "epoch": 0.15, "learning_rate": 6.830873605543131e-08, "loss": 0.8473, "step": 28963 }, { "epoch": 0.15, "learning_rate": 6.818302257690201e-08, "loss": 0.8343, "step": 28964 }, { "epoch": 0.15, "learning_rate": 6.805742448899866e-08, "loss": 0.8169, "step": 28965 }, { "epoch": 0.15, "learning_rate": 6.793194179318119e-08, "loss": 0.8553, "step": 28966 }, { "epoch": 0.15, "learning_rate": 6.780657449090732e-08, "loss": 0.8566, "step": 28967 }, { "epoch": 0.15, "learning_rate": 6.768132258363369e-08, "loss": 0.8569, "step": 28968 }, { "epoch": 0.15, "learning_rate": 6.755618607281578e-08, "loss": 0.8845, "step": 28969 }, { "epoch": 0.15, "learning_rate": 6.743116495990798e-08, "loss": 0.8404, "step": 28970 }, { "epoch": 0.15, "learning_rate": 6.730625924636137e-08, "loss": 0.8845, "step": 28971 }, { "epoch": 0.15, "learning_rate": 6.718146893362809e-08, "loss": 0.854, "step": 28972 }, { "epoch": 0.15, "learning_rate": 6.705679402315702e-08, "loss": 0.8357, "step": 28973 }, { "epoch": 0.15, "learning_rate": 6.693223451639808e-08, "loss": 0.8389, "step": 28974 }, { "epoch": 0.15, "learning_rate": 6.680779041479679e-08, "loss": 0.8404, "step": 28975 }, { "epoch": 0.15, "learning_rate": 6.668346171980089e-08, "loss": 0.8318, "step": 28976 }, { "epoch": 0.15, "learning_rate": 6.655924843285367e-08, "loss": 0.8687, "step": 28977 }, { "epoch": 0.15, "learning_rate": 6.643515055539839e-08, "loss": 0.8144, "step": 28978 }, { "epoch": 0.15, "learning_rate": 6.631116808887617e-08, "loss": 0.8516, "step": 28979 }, { "epoch": 0.15, "learning_rate": 6.618730103472915e-08, "loss": 0.8864, "step": 28980 }, { "epoch": 0.15, "learning_rate": 6.606354939439508e-08, "loss": 0.8377, "step": 28981 }, { "epoch": 0.15, "learning_rate": 6.593991316931391e-08, "loss": 0.8509, "step": 28982 }, { "epoch": 0.15, "learning_rate": 6.581639236091897e-08, "loss": 0.8345, "step": 28983 }, { "epoch": 0.15, "learning_rate": 6.569298697064685e-08, "loss": 0.8363, "step": 28984 }, { "epoch": 0.15, "learning_rate": 6.556969699993087e-08, "loss": 0.858, "step": 28985 }, { "epoch": 0.15, "learning_rate": 6.544652245020433e-08, "loss": 0.844, "step": 28986 }, { "epoch": 0.15, "learning_rate": 6.53234633228983e-08, "loss": 0.8364, "step": 28987 }, { "epoch": 0.15, "learning_rate": 6.520051961944052e-08, "loss": 0.8784, "step": 28988 }, { "epoch": 0.15, "learning_rate": 6.507769134126318e-08, "loss": 0.8137, "step": 28989 }, { "epoch": 0.15, "learning_rate": 6.49549784897896e-08, "loss": 0.8493, "step": 28990 }, { "epoch": 0.15, "learning_rate": 6.48323810664464e-08, "loss": 0.8381, "step": 28991 }, { "epoch": 0.15, "learning_rate": 6.470989907265913e-08, "loss": 0.8321, "step": 28992 }, { "epoch": 0.15, "learning_rate": 6.458753250985105e-08, "loss": 0.8073, "step": 28993 }, { "epoch": 0.15, "learning_rate": 6.446528137944108e-08, "loss": 0.8483, "step": 28994 }, { "epoch": 0.15, "learning_rate": 6.434314568285249e-08, "loss": 0.8147, "step": 28995 }, { "epoch": 0.15, "learning_rate": 6.422112542150415e-08, "loss": 0.8378, "step": 28996 }, { "epoch": 0.15, "learning_rate": 6.409922059681162e-08, "loss": 0.841, "step": 28997 }, { "epoch": 0.15, "learning_rate": 6.397743121019262e-08, "loss": 0.8112, "step": 28998 }, { "epoch": 0.15, "learning_rate": 6.38557572630627e-08, "loss": 0.8859, "step": 28999 }, { "epoch": 0.15, "learning_rate": 6.373419875683406e-08, "loss": 0.8097, "step": 29000 }, { "epoch": 0.15, "learning_rate": 6.361275569291892e-08, "loss": 0.8432, "step": 29001 }, { "epoch": 0.15, "learning_rate": 6.349142807273056e-08, "loss": 0.84, "step": 29002 }, { "epoch": 0.15, "learning_rate": 6.337021589767677e-08, "loss": 0.877, "step": 29003 }, { "epoch": 0.15, "learning_rate": 6.324911916916531e-08, "loss": 0.8385, "step": 29004 }, { "epoch": 0.15, "learning_rate": 6.312813788860395e-08, "loss": 0.8414, "step": 29005 }, { "epoch": 0.15, "learning_rate": 6.300727205739932e-08, "loss": 0.8453, "step": 29006 }, { "epoch": 0.15, "learning_rate": 6.288652167695364e-08, "loss": 0.8652, "step": 29007 }, { "epoch": 0.15, "learning_rate": 6.276588674867023e-08, "loss": 0.8302, "step": 29008 }, { "epoch": 0.15, "learning_rate": 6.264536727395242e-08, "loss": 0.8988, "step": 29009 }, { "epoch": 0.15, "learning_rate": 6.25249632541991e-08, "loss": 0.8199, "step": 29010 }, { "epoch": 0.15, "learning_rate": 6.240467469080802e-08, "loss": 0.8231, "step": 29011 }, { "epoch": 0.15, "learning_rate": 6.228450158517918e-08, "loss": 0.8278, "step": 29012 }, { "epoch": 0.15, "learning_rate": 6.216444393870814e-08, "loss": 0.8456, "step": 29013 }, { "epoch": 0.15, "learning_rate": 6.20445017527882e-08, "loss": 0.8193, "step": 29014 }, { "epoch": 0.15, "learning_rate": 6.192467502881383e-08, "loss": 0.8685, "step": 29015 }, { "epoch": 0.15, "learning_rate": 6.180496376817835e-08, "loss": 0.8523, "step": 29016 }, { "epoch": 0.15, "learning_rate": 6.168536797227065e-08, "loss": 0.8415, "step": 29017 }, { "epoch": 0.15, "learning_rate": 6.156588764248072e-08, "loss": 0.8264, "step": 29018 }, { "epoch": 0.15, "learning_rate": 6.144652278019748e-08, "loss": 0.8512, "step": 29019 }, { "epoch": 0.15, "learning_rate": 6.132727338680866e-08, "loss": 0.8554, "step": 29020 }, { "epoch": 0.15, "learning_rate": 6.120813946369653e-08, "loss": 0.846, "step": 29021 }, { "epoch": 0.15, "learning_rate": 6.108912101224773e-08, "loss": 0.8526, "step": 29022 }, { "epoch": 0.15, "learning_rate": 6.09702180338445e-08, "loss": 0.9022, "step": 29023 }, { "epoch": 0.15, "learning_rate": 6.085143052986909e-08, "loss": 0.8047, "step": 29024 }, { "epoch": 0.15, "learning_rate": 6.073275850169924e-08, "loss": 0.8072, "step": 29025 }, { "epoch": 0.15, "learning_rate": 6.06142019507161e-08, "loss": 0.836, "step": 29026 }, { "epoch": 0.15, "learning_rate": 6.049576087829634e-08, "loss": 0.8502, "step": 29027 }, { "epoch": 0.15, "learning_rate": 6.037743528581552e-08, "loss": 0.8097, "step": 29028 }, { "epoch": 0.15, "learning_rate": 6.025922517464921e-08, "loss": 0.8938, "step": 29029 }, { "epoch": 0.15, "learning_rate": 6.014113054616965e-08, "loss": 0.8443, "step": 29030 }, { "epoch": 0.15, "learning_rate": 6.00231514017513e-08, "loss": 0.8742, "step": 29031 }, { "epoch": 0.15, "learning_rate": 5.990528774276194e-08, "loss": 0.8483, "step": 29032 }, { "epoch": 0.15, "learning_rate": 5.97875395705716e-08, "loss": 0.8454, "step": 29033 }, { "epoch": 0.15, "learning_rate": 5.966990688655027e-08, "loss": 0.8378, "step": 29034 }, { "epoch": 0.15, "learning_rate": 5.955238969206245e-08, "loss": 0.8428, "step": 29035 }, { "epoch": 0.15, "learning_rate": 5.94349879884748e-08, "loss": 0.8286, "step": 29036 }, { "epoch": 0.15, "learning_rate": 5.931770177715068e-08, "loss": 0.8277, "step": 29037 }, { "epoch": 0.15, "learning_rate": 5.920053105945345e-08, "loss": 0.8607, "step": 29038 }, { "epoch": 0.15, "learning_rate": 5.908347583674312e-08, "loss": 0.8801, "step": 29039 }, { "epoch": 0.15, "learning_rate": 5.896653611037972e-08, "loss": 0.8337, "step": 29040 }, { "epoch": 0.15, "learning_rate": 5.884971188172328e-08, "loss": 0.8775, "step": 29041 }, { "epoch": 0.15, "learning_rate": 5.8733003152131596e-08, "loss": 0.7965, "step": 29042 }, { "epoch": 0.15, "learning_rate": 5.861640992295803e-08, "loss": 0.8366, "step": 29043 }, { "epoch": 0.15, "learning_rate": 5.8499932195559274e-08, "loss": 0.8088, "step": 29044 }, { "epoch": 0.15, "learning_rate": 5.838356997128869e-08, "loss": 0.8609, "step": 29045 }, { "epoch": 0.15, "learning_rate": 5.826732325149631e-08, "loss": 0.7939, "step": 29046 }, { "epoch": 0.15, "learning_rate": 5.815119203753439e-08, "loss": 0.8226, "step": 29047 }, { "epoch": 0.15, "learning_rate": 5.803517633075073e-08, "loss": 0.8482, "step": 29048 }, { "epoch": 0.15, "learning_rate": 5.7919276132495374e-08, "loss": 0.8458, "step": 29049 }, { "epoch": 0.15, "learning_rate": 5.78034914441139e-08, "loss": 0.8661, "step": 29050 }, { "epoch": 0.15, "learning_rate": 5.76878222669508e-08, "loss": 0.8765, "step": 29051 }, { "epoch": 0.15, "learning_rate": 5.757226860235055e-08, "loss": 0.8617, "step": 29052 }, { "epoch": 0.15, "learning_rate": 5.74568304516554e-08, "loss": 0.8493, "step": 29053 }, { "epoch": 0.15, "learning_rate": 5.734150781620651e-08, "loss": 0.8538, "step": 29054 }, { "epoch": 0.15, "learning_rate": 5.722630069734392e-08, "loss": 0.8584, "step": 29055 }, { "epoch": 0.15, "learning_rate": 5.711120909640655e-08, "loss": 0.9139, "step": 29056 }, { "epoch": 0.15, "learning_rate": 5.6996233014731115e-08, "loss": 0.813, "step": 29057 }, { "epoch": 0.15, "learning_rate": 5.688137245365433e-08, "loss": 0.8411, "step": 29058 }, { "epoch": 0.15, "learning_rate": 5.676662741450845e-08, "loss": 0.8689, "step": 29059 }, { "epoch": 0.15, "learning_rate": 5.665199789862907e-08, "loss": 0.819, "step": 29060 }, { "epoch": 0.15, "learning_rate": 5.6537483907346256e-08, "loss": 0.8417, "step": 29061 }, { "epoch": 0.15, "learning_rate": 5.6423085441991155e-08, "loss": 0.8246, "step": 29062 }, { "epoch": 0.15, "learning_rate": 5.6308802503893814e-08, "loss": 0.8569, "step": 29063 }, { "epoch": 0.15, "learning_rate": 5.619463509437983e-08, "loss": 0.8358, "step": 29064 }, { "epoch": 0.15, "learning_rate": 5.608058321477816e-08, "loss": 0.8267, "step": 29065 }, { "epoch": 0.15, "learning_rate": 5.5966646866412175e-08, "loss": 0.8674, "step": 29066 }, { "epoch": 0.15, "learning_rate": 5.585282605060527e-08, "loss": 0.8756, "step": 29067 }, { "epoch": 0.15, "learning_rate": 5.5739120768680824e-08, "loss": 0.8521, "step": 29068 }, { "epoch": 0.15, "learning_rate": 5.562553102196e-08, "loss": 0.8491, "step": 29069 }, { "epoch": 0.15, "learning_rate": 5.551205681176286e-08, "loss": 0.8235, "step": 29070 }, { "epoch": 0.15, "learning_rate": 5.5398698139406127e-08, "loss": 0.8644, "step": 29071 }, { "epoch": 0.15, "learning_rate": 5.528545500620874e-08, "loss": 0.8471, "step": 29072 }, { "epoch": 0.15, "learning_rate": 5.5172327413484104e-08, "loss": 0.8447, "step": 29073 }, { "epoch": 0.15, "learning_rate": 5.505931536255005e-08, "loss": 0.828, "step": 29074 }, { "epoch": 0.15, "learning_rate": 5.4946418854715525e-08, "loss": 0.8274, "step": 29075 }, { "epoch": 0.15, "learning_rate": 5.483363789129503e-08, "loss": 0.8329, "step": 29076 }, { "epoch": 0.15, "learning_rate": 5.4720972473597537e-08, "loss": 0.8419, "step": 29077 }, { "epoch": 0.15, "learning_rate": 5.4608422602933084e-08, "loss": 0.8533, "step": 29078 }, { "epoch": 0.15, "learning_rate": 5.4495988280608426e-08, "loss": 0.8113, "step": 29079 }, { "epoch": 0.15, "learning_rate": 5.4383669507930284e-08, "loss": 0.8649, "step": 29080 }, { "epoch": 0.15, "learning_rate": 5.427146628620428e-08, "loss": 0.8506, "step": 29081 }, { "epoch": 0.15, "learning_rate": 5.4159378616731594e-08, "loss": 0.8873, "step": 29082 }, { "epoch": 0.15, "learning_rate": 5.4047406500816745e-08, "loss": 0.8647, "step": 29083 }, { "epoch": 0.15, "learning_rate": 5.393554993976091e-08, "loss": 0.8091, "step": 29084 }, { "epoch": 0.15, "learning_rate": 5.3823808934861944e-08, "loss": 0.827, "step": 29085 }, { "epoch": 0.15, "learning_rate": 5.3712183487418803e-08, "loss": 0.8641, "step": 29086 }, { "epoch": 0.15, "learning_rate": 5.3600673598728226e-08, "loss": 0.9074, "step": 29087 }, { "epoch": 0.15, "learning_rate": 5.348927927008696e-08, "loss": 0.8453, "step": 29088 }, { "epoch": 0.15, "learning_rate": 5.3378000502788404e-08, "loss": 0.8531, "step": 29089 }, { "epoch": 0.15, "learning_rate": 5.326683729812376e-08, "loss": 0.8715, "step": 29090 }, { "epoch": 0.15, "learning_rate": 5.315578965738755e-08, "loss": 0.8387, "step": 29091 }, { "epoch": 0.15, "learning_rate": 5.304485758186873e-08, "loss": 0.8651, "step": 29092 }, { "epoch": 0.15, "learning_rate": 5.2934041072855156e-08, "loss": 0.8583, "step": 29093 }, { "epoch": 0.15, "learning_rate": 5.28233401316347e-08, "loss": 0.8659, "step": 29094 }, { "epoch": 0.15, "learning_rate": 5.271275475949522e-08, "loss": 0.8541, "step": 29095 }, { "epoch": 0.15, "learning_rate": 5.260228495771902e-08, "loss": 0.8131, "step": 29096 }, { "epoch": 0.15, "learning_rate": 5.2491930727591735e-08, "loss": 0.8392, "step": 29097 }, { "epoch": 0.15, "learning_rate": 5.2381692070393454e-08, "loss": 0.8486, "step": 29098 }, { "epoch": 0.15, "learning_rate": 5.2271568987407596e-08, "loss": 0.8642, "step": 29099 }, { "epoch": 0.15, "learning_rate": 5.216156147991092e-08, "loss": 0.8602, "step": 29100 }, { "epoch": 0.15, "learning_rate": 5.205166954918239e-08, "loss": 0.8195, "step": 29101 }, { "epoch": 0.15, "learning_rate": 5.194189319649989e-08, "loss": 0.8445, "step": 29102 }, { "epoch": 0.15, "learning_rate": 5.183223242313684e-08, "loss": 0.8826, "step": 29103 }, { "epoch": 0.15, "learning_rate": 5.1722687230369995e-08, "loss": 0.8197, "step": 29104 }, { "epoch": 0.15, "learning_rate": 5.1613257619469446e-08, "loss": 0.8307, "step": 29105 }, { "epoch": 0.15, "learning_rate": 5.150394359170752e-08, "loss": 0.8243, "step": 29106 }, { "epoch": 0.15, "learning_rate": 5.1394745148355405e-08, "loss": 0.8413, "step": 29107 }, { "epoch": 0.15, "learning_rate": 5.128566229067988e-08, "loss": 0.8597, "step": 29108 }, { "epoch": 0.15, "learning_rate": 5.117669501994882e-08, "loss": 0.8571, "step": 29109 }, { "epoch": 0.15, "learning_rate": 5.106784333743009e-08, "loss": 0.8099, "step": 29110 }, { "epoch": 0.15, "learning_rate": 5.0959107244386016e-08, "loss": 0.8792, "step": 29111 }, { "epoch": 0.15, "learning_rate": 5.085048674208115e-08, "loss": 0.8333, "step": 29112 }, { "epoch": 0.15, "learning_rate": 5.07419818317767e-08, "loss": 0.8407, "step": 29113 }, { "epoch": 0.15, "learning_rate": 5.063359251473277e-08, "loss": 0.8496, "step": 29114 }, { "epoch": 0.15, "learning_rate": 5.0525318792210566e-08, "loss": 0.8342, "step": 29115 }, { "epoch": 0.15, "learning_rate": 5.041716066546798e-08, "loss": 0.8345, "step": 29116 }, { "epoch": 0.15, "learning_rate": 5.030911813575956e-08, "loss": 0.8087, "step": 29117 }, { "epoch": 0.15, "learning_rate": 5.020119120434208e-08, "loss": 0.8583, "step": 29118 }, { "epoch": 0.15, "learning_rate": 5.00933798724701e-08, "loss": 0.8578, "step": 29119 }, { "epoch": 0.15, "learning_rate": 4.998568414139371e-08, "loss": 0.8819, "step": 29120 }, { "epoch": 0.15, "learning_rate": 4.987810401236637e-08, "loss": 0.81, "step": 29121 }, { "epoch": 0.15, "learning_rate": 4.9770639486637075e-08, "loss": 0.8414, "step": 29122 }, { "epoch": 0.15, "learning_rate": 4.966329056545482e-08, "loss": 0.8937, "step": 29123 }, { "epoch": 0.15, "learning_rate": 4.9556057250066406e-08, "loss": 0.834, "step": 29124 }, { "epoch": 0.15, "learning_rate": 4.944893954171859e-08, "loss": 0.8508, "step": 29125 }, { "epoch": 0.15, "learning_rate": 4.934193744165483e-08, "loss": 0.8346, "step": 29126 }, { "epoch": 0.15, "learning_rate": 4.923505095111747e-08, "loss": 0.8156, "step": 29127 }, { "epoch": 0.15, "learning_rate": 4.912828007135106e-08, "loss": 0.8514, "step": 29128 }, { "epoch": 0.15, "learning_rate": 4.902162480359352e-08, "loss": 0.8308, "step": 29129 }, { "epoch": 0.15, "learning_rate": 4.891508514908605e-08, "loss": 0.8269, "step": 29130 }, { "epoch": 0.15, "learning_rate": 4.8808661109065456e-08, "loss": 0.8068, "step": 29131 }, { "epoch": 0.15, "learning_rate": 4.8702352684767416e-08, "loss": 0.8217, "step": 29132 }, { "epoch": 0.15, "learning_rate": 4.859615987742761e-08, "loss": 0.8421, "step": 29133 }, { "epoch": 0.15, "learning_rate": 4.8490082688280595e-08, "loss": 0.8943, "step": 29134 }, { "epoch": 0.15, "learning_rate": 4.8384121118557614e-08, "loss": 0.8325, "step": 29135 }, { "epoch": 0.15, "learning_rate": 4.827827516949102e-08, "loss": 0.8313, "step": 29136 }, { "epoch": 0.15, "learning_rate": 4.817254484231093e-08, "loss": 0.7935, "step": 29137 }, { "epoch": 0.15, "learning_rate": 4.806693013824304e-08, "loss": 0.8425, "step": 29138 }, { "epoch": 0.15, "learning_rate": 4.796143105851636e-08, "loss": 0.8543, "step": 29139 }, { "epoch": 0.15, "learning_rate": 4.785604760435658e-08, "loss": 0.8629, "step": 29140 }, { "epoch": 0.15, "learning_rate": 4.7750779776989385e-08, "loss": 0.8478, "step": 29141 }, { "epoch": 0.15, "learning_rate": 4.764562757763491e-08, "loss": 0.8505, "step": 29142 }, { "epoch": 0.15, "learning_rate": 4.754059100751551e-08, "loss": 0.8417, "step": 29143 }, { "epoch": 0.15, "learning_rate": 4.743567006785354e-08, "loss": 0.8258, "step": 29144 }, { "epoch": 0.15, "learning_rate": 4.733086475986581e-08, "loss": 0.8396, "step": 29145 }, { "epoch": 0.15, "learning_rate": 4.7226175084771344e-08, "loss": 0.8235, "step": 29146 }, { "epoch": 0.15, "learning_rate": 4.7121601043785825e-08, "loss": 0.8727, "step": 29147 }, { "epoch": 0.15, "learning_rate": 4.701714263812496e-08, "loss": 0.7974, "step": 29148 }, { "epoch": 0.15, "learning_rate": 4.69127998690011e-08, "loss": 0.8198, "step": 29149 }, { "epoch": 0.15, "learning_rate": 4.680857273762773e-08, "loss": 0.8822, "step": 29150 }, { "epoch": 0.15, "learning_rate": 4.6704461245214997e-08, "loss": 0.8437, "step": 29151 }, { "epoch": 0.15, "learning_rate": 4.6600465392971916e-08, "loss": 0.8198, "step": 29152 }, { "epoch": 0.15, "learning_rate": 4.649658518210864e-08, "loss": 0.8379, "step": 29153 }, { "epoch": 0.15, "learning_rate": 4.639282061383088e-08, "loss": 0.8888, "step": 29154 }, { "epoch": 0.15, "learning_rate": 4.628917168934543e-08, "loss": 0.8384, "step": 29155 }, { "epoch": 0.15, "learning_rate": 4.618563840985357e-08, "loss": 0.812, "step": 29156 }, { "epoch": 0.15, "learning_rate": 4.608222077656099e-08, "loss": 0.814, "step": 29157 }, { "epoch": 0.16, "learning_rate": 4.5978918790667846e-08, "loss": 0.8343, "step": 29158 }, { "epoch": 0.16, "learning_rate": 4.58757324533754e-08, "loss": 0.8368, "step": 29159 }, { "epoch": 0.16, "learning_rate": 4.5772661765881576e-08, "loss": 0.8256, "step": 29160 }, { "epoch": 0.16, "learning_rate": 4.566970672938431e-08, "loss": 0.8915, "step": 29161 }, { "epoch": 0.16, "learning_rate": 4.556686734507931e-08, "loss": 0.8499, "step": 29162 }, { "epoch": 0.16, "learning_rate": 4.5464143614162294e-08, "loss": 0.8688, "step": 29163 }, { "epoch": 0.16, "learning_rate": 4.536153553782563e-08, "loss": 0.8678, "step": 29164 }, { "epoch": 0.16, "learning_rate": 4.5259043117261705e-08, "loss": 0.8386, "step": 29165 }, { "epoch": 0.16, "learning_rate": 4.5156666353662895e-08, "loss": 0.8484, "step": 29166 }, { "epoch": 0.16, "learning_rate": 4.5054405248216025e-08, "loss": 0.8821, "step": 29167 }, { "epoch": 0.16, "learning_rate": 4.495225980211127e-08, "loss": 0.8257, "step": 29168 }, { "epoch": 0.16, "learning_rate": 4.485023001653432e-08, "loss": 0.8242, "step": 29169 }, { "epoch": 0.16, "learning_rate": 4.4748315892669814e-08, "loss": 0.8253, "step": 29170 }, { "epoch": 0.16, "learning_rate": 4.464651743170456e-08, "loss": 0.8356, "step": 29171 }, { "epoch": 0.16, "learning_rate": 4.4544834634818735e-08, "loss": 0.8544, "step": 29172 }, { "epoch": 0.16, "learning_rate": 4.444326750319472e-08, "loss": 0.8396, "step": 29173 }, { "epoch": 0.16, "learning_rate": 4.434181603801268e-08, "loss": 0.8528, "step": 29174 }, { "epoch": 0.16, "learning_rate": 4.4240480240450577e-08, "loss": 0.8205, "step": 29175 }, { "epoch": 0.16, "learning_rate": 4.413926011168634e-08, "loss": 0.8243, "step": 29176 }, { "epoch": 0.16, "learning_rate": 4.403815565289571e-08, "loss": 0.8414, "step": 29177 }, { "epoch": 0.16, "learning_rate": 4.39371668652544e-08, "loss": 0.8474, "step": 29178 }, { "epoch": 0.16, "learning_rate": 4.38362937499337e-08, "loss": 0.9042, "step": 29179 }, { "epoch": 0.16, "learning_rate": 4.3735536308107116e-08, "loss": 0.87, "step": 29180 }, { "epoch": 0.16, "learning_rate": 4.3634894540944826e-08, "loss": 0.8708, "step": 29181 }, { "epoch": 0.16, "learning_rate": 4.353436844961589e-08, "loss": 0.8085, "step": 29182 }, { "epoch": 0.16, "learning_rate": 4.343395803528938e-08, "loss": 0.8708, "step": 29183 }, { "epoch": 0.16, "learning_rate": 4.3333663299129915e-08, "loss": 0.8484, "step": 29184 }, { "epoch": 0.16, "learning_rate": 4.323348424230433e-08, "loss": 0.8496, "step": 29185 }, { "epoch": 0.16, "learning_rate": 4.313342086597616e-08, "loss": 0.8319, "step": 29186 }, { "epoch": 0.16, "learning_rate": 4.3033473171307795e-08, "loss": 0.8137, "step": 29187 }, { "epoch": 0.16, "learning_rate": 4.2933641159460524e-08, "loss": 0.8239, "step": 29188 }, { "epoch": 0.16, "learning_rate": 4.283392483159454e-08, "loss": 0.8997, "step": 29189 }, { "epoch": 0.16, "learning_rate": 4.27343241888678e-08, "loss": 0.819, "step": 29190 }, { "epoch": 0.16, "learning_rate": 4.2634839232437166e-08, "loss": 0.8666, "step": 29191 }, { "epoch": 0.16, "learning_rate": 4.2535469963459476e-08, "loss": 0.7958, "step": 29192 }, { "epoch": 0.16, "learning_rate": 4.243621638308937e-08, "loss": 0.8055, "step": 29193 }, { "epoch": 0.16, "learning_rate": 4.233707849247926e-08, "loss": 0.8437, "step": 29194 }, { "epoch": 0.16, "learning_rate": 4.223805629278044e-08, "loss": 0.8588, "step": 29195 }, { "epoch": 0.16, "learning_rate": 4.2139149785144216e-08, "loss": 0.8335, "step": 29196 }, { "epoch": 0.16, "learning_rate": 4.204035897072079e-08, "loss": 0.823, "step": 29197 }, { "epoch": 0.16, "learning_rate": 4.194168385065478e-08, "loss": 0.8643, "step": 29198 }, { "epoch": 0.16, "learning_rate": 4.1843124426096396e-08, "loss": 0.8398, "step": 29199 }, { "epoch": 0.16, "learning_rate": 4.174468069818805e-08, "loss": 0.8507, "step": 29200 }, { "epoch": 0.16, "learning_rate": 4.1646352668074375e-08, "loss": 0.8206, "step": 29201 }, { "epoch": 0.16, "learning_rate": 4.1548140336896695e-08, "loss": 0.841, "step": 29202 }, { "epoch": 0.16, "learning_rate": 4.145004370579853e-08, "loss": 0.9172, "step": 29203 }, { "epoch": 0.16, "learning_rate": 4.135206277591785e-08, "loss": 0.8172, "step": 29204 }, { "epoch": 0.16, "learning_rate": 4.1254197548393773e-08, "loss": 0.8841, "step": 29205 }, { "epoch": 0.16, "learning_rate": 4.115644802436203e-08, "loss": 0.8205, "step": 29206 }, { "epoch": 0.16, "learning_rate": 4.10588142049595e-08, "loss": 0.8354, "step": 29207 }, { "epoch": 0.16, "learning_rate": 4.0961296091320826e-08, "loss": 0.8413, "step": 29208 }, { "epoch": 0.16, "learning_rate": 4.0863893684577325e-08, "loss": 0.8276, "step": 29209 }, { "epoch": 0.16, "learning_rate": 4.0766606985862544e-08, "loss": 0.8343, "step": 29210 }, { "epoch": 0.16, "learning_rate": 4.066943599630557e-08, "loss": 0.8318, "step": 29211 }, { "epoch": 0.16, "learning_rate": 4.057238071703551e-08, "loss": 0.865, "step": 29212 }, { "epoch": 0.16, "learning_rate": 4.047544114918034e-08, "loss": 0.8049, "step": 29213 }, { "epoch": 0.16, "learning_rate": 4.037861729386694e-08, "loss": 0.8277, "step": 29214 }, { "epoch": 0.16, "learning_rate": 4.028190915221886e-08, "loss": 0.8552, "step": 29215 }, { "epoch": 0.16, "learning_rate": 4.018531672536075e-08, "loss": 0.8181, "step": 29216 }, { "epoch": 0.16, "learning_rate": 4.008884001441393e-08, "loss": 0.8488, "step": 29217 }, { "epoch": 0.16, "learning_rate": 3.999247902049974e-08, "loss": 0.8816, "step": 29218 }, { "epoch": 0.16, "learning_rate": 3.989623374473728e-08, "loss": 0.8528, "step": 29219 }, { "epoch": 0.16, "learning_rate": 3.980010418824565e-08, "loss": 0.8869, "step": 29220 }, { "epoch": 0.16, "learning_rate": 3.970409035214173e-08, "loss": 0.8266, "step": 29221 }, { "epoch": 0.16, "learning_rate": 3.96081922375402e-08, "loss": 0.8193, "step": 29222 }, { "epoch": 0.16, "learning_rate": 3.951240984555571e-08, "loss": 0.848, "step": 29223 }, { "epoch": 0.16, "learning_rate": 3.941674317730071e-08, "loss": 0.8569, "step": 29224 }, { "epoch": 0.16, "learning_rate": 3.932119223388764e-08, "loss": 0.8532, "step": 29225 }, { "epoch": 0.16, "learning_rate": 3.922575701642561e-08, "loss": 0.8336, "step": 29226 }, { "epoch": 0.16, "learning_rate": 3.913043752602374e-08, "loss": 0.8477, "step": 29227 }, { "epoch": 0.16, "learning_rate": 3.903523376379004e-08, "loss": 0.8659, "step": 29228 }, { "epoch": 0.16, "learning_rate": 3.894014573082916e-08, "loss": 0.845, "step": 29229 }, { "epoch": 0.16, "learning_rate": 3.8845173428248007e-08, "loss": 0.8247, "step": 29230 }, { "epoch": 0.16, "learning_rate": 3.8750316857146806e-08, "loss": 0.8356, "step": 29231 }, { "epoch": 0.16, "learning_rate": 3.8655576018631346e-08, "loss": 0.8143, "step": 29232 }, { "epoch": 0.16, "learning_rate": 3.856095091379963e-08, "loss": 0.8346, "step": 29233 }, { "epoch": 0.16, "learning_rate": 3.8466441543753006e-08, "loss": 0.8028, "step": 29234 }, { "epoch": 0.16, "learning_rate": 3.8372047909588375e-08, "loss": 0.8673, "step": 29235 }, { "epoch": 0.16, "learning_rate": 3.8277770012401515e-08, "loss": 0.7912, "step": 29236 }, { "epoch": 0.16, "learning_rate": 3.818360785328934e-08, "loss": 0.8436, "step": 29237 }, { "epoch": 0.16, "learning_rate": 3.808956143334541e-08, "loss": 0.8734, "step": 29238 }, { "epoch": 0.16, "learning_rate": 3.7995630753662195e-08, "loss": 0.8774, "step": 29239 }, { "epoch": 0.16, "learning_rate": 3.7901815815332145e-08, "loss": 0.8548, "step": 29240 }, { "epoch": 0.16, "learning_rate": 3.780811661944328e-08, "loss": 0.7942, "step": 29241 }, { "epoch": 0.16, "learning_rate": 3.771453316708584e-08, "loss": 0.8628, "step": 29242 }, { "epoch": 0.16, "learning_rate": 3.762106545934563e-08, "loss": 0.8662, "step": 29243 }, { "epoch": 0.16, "learning_rate": 3.752771349730955e-08, "loss": 0.8557, "step": 29244 }, { "epoch": 0.16, "learning_rate": 3.743447728206229e-08, "loss": 0.8336, "step": 29245 }, { "epoch": 0.16, "learning_rate": 3.734135681468631e-08, "loss": 0.8391, "step": 29246 }, { "epoch": 0.16, "learning_rate": 3.7248352096264094e-08, "loss": 0.8452, "step": 29247 }, { "epoch": 0.16, "learning_rate": 3.715546312787699e-08, "loss": 0.8227, "step": 29248 }, { "epoch": 0.16, "learning_rate": 3.706268991060191e-08, "loss": 0.8602, "step": 29249 }, { "epoch": 0.16, "learning_rate": 3.697003244551911e-08, "loss": 0.8327, "step": 29250 }, { "epoch": 0.16, "learning_rate": 3.687749073370439e-08, "loss": 0.8709, "step": 29251 }, { "epoch": 0.16, "learning_rate": 3.678506477623245e-08, "loss": 0.8436, "step": 29252 }, { "epoch": 0.16, "learning_rate": 3.669275457417687e-08, "loss": 0.8059, "step": 29253 }, { "epoch": 0.16, "learning_rate": 3.660056012861124e-08, "loss": 0.8096, "step": 29254 }, { "epoch": 0.16, "learning_rate": 3.6508481440604704e-08, "loss": 0.8508, "step": 29255 }, { "epoch": 0.16, "learning_rate": 3.641651851122974e-08, "loss": 0.8264, "step": 29256 }, { "epoch": 0.16, "learning_rate": 3.6324671341553265e-08, "loss": 0.8478, "step": 29257 }, { "epoch": 0.16, "learning_rate": 3.6232939932643316e-08, "loss": 0.8691, "step": 29258 }, { "epoch": 0.16, "learning_rate": 3.6141324285563496e-08, "loss": 0.8699, "step": 29259 }, { "epoch": 0.16, "learning_rate": 3.6049824401380715e-08, "loss": 0.8276, "step": 29260 }, { "epoch": 0.16, "learning_rate": 3.595844028115636e-08, "loss": 0.8398, "step": 29261 }, { "epoch": 0.16, "learning_rate": 3.5867171925954016e-08, "loss": 0.8319, "step": 29262 }, { "epoch": 0.16, "learning_rate": 3.577601933683172e-08, "loss": 0.889, "step": 29263 }, { "epoch": 0.16, "learning_rate": 3.568498251484975e-08, "loss": 0.8121, "step": 29264 }, { "epoch": 0.16, "learning_rate": 3.559406146106614e-08, "loss": 0.8205, "step": 29265 }, { "epoch": 0.16, "learning_rate": 3.5503256176536714e-08, "loss": 0.8317, "step": 29266 }, { "epoch": 0.16, "learning_rate": 3.5412566662316186e-08, "loss": 0.8645, "step": 29267 }, { "epoch": 0.16, "learning_rate": 3.532199291945815e-08, "loss": 0.8247, "step": 29268 }, { "epoch": 0.16, "learning_rate": 3.523153494901621e-08, "loss": 0.8476, "step": 29269 }, { "epoch": 0.16, "learning_rate": 3.514119275203953e-08, "loss": 0.8193, "step": 29270 }, { "epoch": 0.16, "learning_rate": 3.50509663295795e-08, "loss": 0.8537, "step": 29271 }, { "epoch": 0.16, "learning_rate": 3.4960855682684145e-08, "loss": 0.8387, "step": 29272 }, { "epoch": 0.16, "learning_rate": 3.487086081239821e-08, "loss": 0.8961, "step": 29273 }, { "epoch": 0.16, "learning_rate": 3.478098171976974e-08, "loss": 0.825, "step": 29274 }, { "epoch": 0.16, "learning_rate": 3.469121840584233e-08, "loss": 0.8322, "step": 29275 }, { "epoch": 0.16, "learning_rate": 3.4601570871658494e-08, "loss": 0.7945, "step": 29276 }, { "epoch": 0.16, "learning_rate": 3.4512039118259624e-08, "loss": 0.8456, "step": 29277 }, { "epoch": 0.16, "learning_rate": 3.442262314668599e-08, "loss": 0.8559, "step": 29278 }, { "epoch": 0.16, "learning_rate": 3.433332295797787e-08, "loss": 0.8286, "step": 29279 }, { "epoch": 0.16, "learning_rate": 3.424413855316999e-08, "loss": 0.8475, "step": 29280 }, { "epoch": 0.16, "learning_rate": 3.4155069933301535e-08, "loss": 0.875, "step": 29281 }, { "epoch": 0.16, "learning_rate": 3.40661170994061e-08, "loss": 0.8541, "step": 29282 }, { "epoch": 0.16, "learning_rate": 3.397728005251621e-08, "loss": 0.8504, "step": 29283 }, { "epoch": 0.16, "learning_rate": 3.388855879366437e-08, "loss": 0.8847, "step": 29284 }, { "epoch": 0.16, "learning_rate": 3.379995332388197e-08, "loss": 0.8614, "step": 29285 }, { "epoch": 0.16, "learning_rate": 3.37114636441993e-08, "loss": 0.8486, "step": 29286 }, { "epoch": 0.16, "learning_rate": 3.3623089755642216e-08, "loss": 0.8799, "step": 29287 }, { "epoch": 0.16, "learning_rate": 3.353483165923988e-08, "loss": 0.8493, "step": 29288 }, { "epoch": 0.16, "learning_rate": 3.3446689356015914e-08, "loss": 0.8212, "step": 29289 }, { "epoch": 0.16, "learning_rate": 3.335866284699507e-08, "loss": 0.8695, "step": 29290 }, { "epoch": 0.16, "learning_rate": 3.3270752133199856e-08, "loss": 0.8231, "step": 29291 }, { "epoch": 0.16, "learning_rate": 3.318295721565168e-08, "loss": 0.8254, "step": 29292 }, { "epoch": 0.16, "learning_rate": 3.309527809537083e-08, "loss": 0.8473, "step": 29293 }, { "epoch": 0.16, "learning_rate": 3.300771477337539e-08, "loss": 0.8068, "step": 29294 }, { "epoch": 0.16, "learning_rate": 3.2920267250683426e-08, "loss": 0.8393, "step": 29295 }, { "epoch": 0.16, "learning_rate": 3.28329355283108e-08, "loss": 0.8159, "step": 29296 }, { "epoch": 0.16, "learning_rate": 3.2745719607271133e-08, "loss": 0.793, "step": 29297 }, { "epoch": 0.16, "learning_rate": 3.265861948857918e-08, "loss": 0.8437, "step": 29298 }, { "epoch": 0.16, "learning_rate": 3.257163517324635e-08, "loss": 0.8939, "step": 29299 }, { "epoch": 0.16, "learning_rate": 3.248476666228295e-08, "loss": 0.8392, "step": 29300 }, { "epoch": 0.16, "learning_rate": 3.239801395669928e-08, "loss": 0.8093, "step": 29301 }, { "epoch": 0.16, "learning_rate": 3.2311377057501204e-08, "loss": 0.8222, "step": 29302 }, { "epoch": 0.16, "learning_rate": 3.22248559656968e-08, "loss": 0.859, "step": 29303 }, { "epoch": 0.16, "learning_rate": 3.213845068229082e-08, "loss": 0.8611, "step": 29304 }, { "epoch": 0.16, "learning_rate": 3.2052161208288025e-08, "loss": 0.8753, "step": 29305 }, { "epoch": 0.16, "learning_rate": 3.1965987544689826e-08, "loss": 0.8343, "step": 29306 }, { "epoch": 0.16, "learning_rate": 3.187992969249876e-08, "loss": 0.8325, "step": 29307 }, { "epoch": 0.16, "learning_rate": 3.179398765271291e-08, "loss": 0.8684, "step": 29308 }, { "epoch": 0.16, "learning_rate": 3.1708161426331485e-08, "loss": 0.8382, "step": 29309 }, { "epoch": 0.16, "learning_rate": 3.1622451014352575e-08, "loss": 0.8293, "step": 29310 }, { "epoch": 0.16, "learning_rate": 3.153685641777094e-08, "loss": 0.8565, "step": 29311 }, { "epoch": 0.16, "learning_rate": 3.1451377637580214e-08, "loss": 0.8599, "step": 29312 }, { "epoch": 0.16, "learning_rate": 3.136601467477629e-08, "loss": 0.8232, "step": 29313 }, { "epoch": 0.16, "learning_rate": 3.128076753034837e-08, "loss": 0.8728, "step": 29314 }, { "epoch": 0.16, "learning_rate": 3.119563620528787e-08, "loss": 0.8544, "step": 29315 }, { "epoch": 0.16, "learning_rate": 3.111062070058401e-08, "loss": 0.8624, "step": 29316 }, { "epoch": 0.16, "learning_rate": 3.102572101722268e-08, "loss": 0.8805, "step": 29317 }, { "epoch": 0.16, "learning_rate": 3.0940937156194175e-08, "loss": 0.8486, "step": 29318 }, { "epoch": 0.16, "learning_rate": 3.085626911847994e-08, "loss": 0.8274, "step": 29319 }, { "epoch": 0.16, "learning_rate": 3.077171690506475e-08, "loss": 0.8035, "step": 29320 }, { "epoch": 0.16, "learning_rate": 3.068728051693226e-08, "loss": 0.8318, "step": 29321 }, { "epoch": 0.16, "learning_rate": 3.0602959955061686e-08, "loss": 0.8227, "step": 29322 }, { "epoch": 0.16, "learning_rate": 3.051875522043335e-08, "loss": 0.8516, "step": 29323 }, { "epoch": 0.16, "learning_rate": 3.043466631402536e-08, "loss": 0.8252, "step": 29324 }, { "epoch": 0.16, "learning_rate": 3.035069323681583e-08, "loss": 0.8363, "step": 29325 }, { "epoch": 0.16, "learning_rate": 3.026683598977842e-08, "loss": 0.8482, "step": 29326 }, { "epoch": 0.16, "learning_rate": 3.018309457388902e-08, "loss": 0.8508, "step": 29327 }, { "epoch": 0.16, "learning_rate": 3.009946899012017e-08, "loss": 0.8649, "step": 29328 }, { "epoch": 0.16, "learning_rate": 3.0015959239442224e-08, "loss": 0.8421, "step": 29329 }, { "epoch": 0.16, "learning_rate": 2.993256532282773e-08, "loss": 0.875, "step": 29330 }, { "epoch": 0.16, "learning_rate": 2.98492872412437e-08, "loss": 0.837, "step": 29331 }, { "epoch": 0.16, "learning_rate": 2.976612499565934e-08, "loss": 0.8383, "step": 29332 }, { "epoch": 0.16, "learning_rate": 2.9683078587038338e-08, "loss": 0.8557, "step": 29333 }, { "epoch": 0.16, "learning_rate": 2.9600148016347697e-08, "loss": 0.8651, "step": 29334 }, { "epoch": 0.16, "learning_rate": 2.9517333284549978e-08, "loss": 0.8227, "step": 29335 }, { "epoch": 0.16, "learning_rate": 2.9434634392607743e-08, "loss": 0.8401, "step": 29336 }, { "epoch": 0.16, "learning_rate": 2.9352051341482445e-08, "loss": 0.8453, "step": 29337 }, { "epoch": 0.16, "learning_rate": 2.9269584132132213e-08, "loss": 0.845, "step": 29338 }, { "epoch": 0.16, "learning_rate": 2.918723276551627e-08, "loss": 0.8459, "step": 29339 }, { "epoch": 0.16, "learning_rate": 2.9104997242590528e-08, "loss": 0.8597, "step": 29340 }, { "epoch": 0.16, "learning_rate": 2.902287756430977e-08, "loss": 0.8476, "step": 29341 }, { "epoch": 0.16, "learning_rate": 2.8940873731631015e-08, "loss": 0.8132, "step": 29342 }, { "epoch": 0.16, "learning_rate": 2.8858985745503498e-08, "loss": 0.8855, "step": 29343 }, { "epoch": 0.16, "learning_rate": 2.8777213606880904e-08, "loss": 0.8045, "step": 29344 }, { "epoch": 0.16, "learning_rate": 2.8695557316712473e-08, "loss": 0.8797, "step": 29345 }, { "epoch": 0.16, "learning_rate": 2.861401687594745e-08, "loss": 0.8732, "step": 29346 }, { "epoch": 0.16, "learning_rate": 2.8532592285532845e-08, "loss": 0.8488, "step": 29347 }, { "epoch": 0.16, "learning_rate": 2.8451283546413465e-08, "loss": 0.8333, "step": 29348 }, { "epoch": 0.16, "learning_rate": 2.8370090659536332e-08, "loss": 0.8886, "step": 29349 }, { "epoch": 0.16, "learning_rate": 2.828901362584402e-08, "loss": 0.8406, "step": 29350 }, { "epoch": 0.16, "learning_rate": 2.820805244627689e-08, "loss": 0.8498, "step": 29351 }, { "epoch": 0.16, "learning_rate": 2.8127207121776412e-08, "loss": 0.8585, "step": 29352 }, { "epoch": 0.16, "learning_rate": 2.8046477653282944e-08, "loss": 0.816, "step": 29353 }, { "epoch": 0.16, "learning_rate": 2.7965864041733515e-08, "loss": 0.8817, "step": 29354 }, { "epoch": 0.16, "learning_rate": 2.788536628806515e-08, "loss": 0.841, "step": 29355 }, { "epoch": 0.16, "learning_rate": 2.7804984393212664e-08, "loss": 0.8225, "step": 29356 }, { "epoch": 0.16, "learning_rate": 2.7724718358110857e-08, "loss": 0.8261, "step": 29357 }, { "epoch": 0.16, "learning_rate": 2.76445681836901e-08, "loss": 0.8278, "step": 29358 }, { "epoch": 0.16, "learning_rate": 2.756453387088409e-08, "loss": 0.8447, "step": 29359 }, { "epoch": 0.16, "learning_rate": 2.7484615420620975e-08, "loss": 0.8123, "step": 29360 }, { "epoch": 0.16, "learning_rate": 2.740481283383001e-08, "loss": 0.7944, "step": 29361 }, { "epoch": 0.16, "learning_rate": 2.7325126111438227e-08, "loss": 0.827, "step": 29362 }, { "epoch": 0.16, "learning_rate": 2.724555525437156e-08, "loss": 0.8457, "step": 29363 }, { "epoch": 0.16, "learning_rate": 2.7166100263554816e-08, "loss": 0.8424, "step": 29364 }, { "epoch": 0.16, "learning_rate": 2.70867611399106e-08, "loss": 0.8482, "step": 29365 }, { "epoch": 0.16, "learning_rate": 2.700753788436039e-08, "loss": 0.8533, "step": 29366 }, { "epoch": 0.16, "learning_rate": 2.6928430497825674e-08, "loss": 0.8285, "step": 29367 }, { "epoch": 0.16, "learning_rate": 2.6849438981224607e-08, "loss": 0.8446, "step": 29368 }, { "epoch": 0.16, "learning_rate": 2.677056333547534e-08, "loss": 0.8946, "step": 29369 }, { "epoch": 0.16, "learning_rate": 2.669180356149381e-08, "loss": 0.8371, "step": 29370 }, { "epoch": 0.16, "learning_rate": 2.6613159660195954e-08, "loss": 0.8645, "step": 29371 }, { "epoch": 0.16, "learning_rate": 2.6534631632494368e-08, "loss": 0.7925, "step": 29372 }, { "epoch": 0.16, "learning_rate": 2.645621947930277e-08, "loss": 0.8501, "step": 29373 }, { "epoch": 0.16, "learning_rate": 2.637792320153154e-08, "loss": 0.8573, "step": 29374 }, { "epoch": 0.16, "learning_rate": 2.6299742800089956e-08, "loss": 0.8536, "step": 29375 }, { "epoch": 0.16, "learning_rate": 2.6221678275886176e-08, "loss": 0.7992, "step": 29376 }, { "epoch": 0.16, "learning_rate": 2.6143729629828363e-08, "loss": 0.83, "step": 29377 }, { "epoch": 0.16, "learning_rate": 2.606589686282024e-08, "loss": 0.8843, "step": 29378 }, { "epoch": 0.16, "learning_rate": 2.5988179975768858e-08, "loss": 0.8522, "step": 29379 }, { "epoch": 0.16, "learning_rate": 2.5910578969574607e-08, "loss": 0.8379, "step": 29380 }, { "epoch": 0.16, "learning_rate": 2.58330938451401e-08, "loss": 0.8632, "step": 29381 }, { "epoch": 0.16, "learning_rate": 2.575572460336573e-08, "loss": 0.8612, "step": 29382 }, { "epoch": 0.16, "learning_rate": 2.567847124514966e-08, "loss": 0.8391, "step": 29383 }, { "epoch": 0.16, "learning_rate": 2.560133377139007e-08, "loss": 0.8584, "step": 29384 }, { "epoch": 0.16, "learning_rate": 2.55243121829829e-08, "loss": 0.8346, "step": 29385 }, { "epoch": 0.16, "learning_rate": 2.5447406480823e-08, "loss": 0.8272, "step": 29386 }, { "epoch": 0.16, "learning_rate": 2.5370616665805202e-08, "loss": 0.8733, "step": 29387 }, { "epoch": 0.16, "learning_rate": 2.52939427388188e-08, "loss": 0.8115, "step": 29388 }, { "epoch": 0.16, "learning_rate": 2.5217384700757517e-08, "loss": 0.8546, "step": 29389 }, { "epoch": 0.16, "learning_rate": 2.514094255250954e-08, "loss": 0.829, "step": 29390 }, { "epoch": 0.16, "learning_rate": 2.506461629496304e-08, "loss": 0.8585, "step": 29391 }, { "epoch": 0.16, "learning_rate": 2.498840592900509e-08, "loss": 0.8579, "step": 29392 }, { "epoch": 0.16, "learning_rate": 2.4912311455520533e-08, "loss": 0.8351, "step": 29393 }, { "epoch": 0.16, "learning_rate": 2.483633287539311e-08, "loss": 0.8771, "step": 29394 }, { "epoch": 0.16, "learning_rate": 2.4760470189506558e-08, "loss": 0.8155, "step": 29395 }, { "epoch": 0.16, "learning_rate": 2.4684723398742394e-08, "loss": 0.8168, "step": 29396 }, { "epoch": 0.16, "learning_rate": 2.4609092503979914e-08, "loss": 0.8307, "step": 29397 }, { "epoch": 0.16, "learning_rate": 2.4533577506098415e-08, "loss": 0.8545, "step": 29398 }, { "epoch": 0.16, "learning_rate": 2.4458178405974974e-08, "loss": 0.8378, "step": 29399 }, { "epoch": 0.16, "learning_rate": 2.4382895204485556e-08, "loss": 0.8351, "step": 29400 }, { "epoch": 0.16, "learning_rate": 2.430772790250502e-08, "loss": 0.8733, "step": 29401 }, { "epoch": 0.16, "learning_rate": 2.4232676500905995e-08, "loss": 0.8803, "step": 29402 }, { "epoch": 0.16, "learning_rate": 2.415774100056112e-08, "loss": 0.8318, "step": 29403 }, { "epoch": 0.16, "learning_rate": 2.4082921402340808e-08, "loss": 0.8303, "step": 29404 }, { "epoch": 0.16, "learning_rate": 2.4008217707114366e-08, "loss": 0.9008, "step": 29405 }, { "epoch": 0.16, "learning_rate": 2.39336299157511e-08, "loss": 0.8415, "step": 29406 }, { "epoch": 0.16, "learning_rate": 2.385915802911476e-08, "loss": 0.8919, "step": 29407 }, { "epoch": 0.16, "learning_rate": 2.378480204807243e-08, "loss": 0.8312, "step": 29408 }, { "epoch": 0.16, "learning_rate": 2.3710561973486756e-08, "loss": 0.85, "step": 29409 }, { "epoch": 0.16, "learning_rate": 2.36364378062226e-08, "loss": 0.8428, "step": 29410 }, { "epoch": 0.16, "learning_rate": 2.3562429547139275e-08, "loss": 0.8195, "step": 29411 }, { "epoch": 0.16, "learning_rate": 2.3488537197096094e-08, "loss": 0.8087, "step": 29412 }, { "epoch": 0.16, "learning_rate": 2.341476075695348e-08, "loss": 0.7966, "step": 29413 }, { "epoch": 0.16, "learning_rate": 2.3341100227567414e-08, "loss": 0.8507, "step": 29414 }, { "epoch": 0.16, "learning_rate": 2.326755560979388e-08, "loss": 0.9038, "step": 29415 }, { "epoch": 0.16, "learning_rate": 2.3194126904487748e-08, "loss": 0.8664, "step": 29416 }, { "epoch": 0.16, "learning_rate": 2.3120814112501665e-08, "loss": 0.891, "step": 29417 }, { "epoch": 0.16, "learning_rate": 2.304761723468718e-08, "loss": 0.863, "step": 29418 }, { "epoch": 0.16, "learning_rate": 2.2974536271895832e-08, "loss": 0.8135, "step": 29419 }, { "epoch": 0.16, "learning_rate": 2.2901571224974718e-08, "loss": 0.8282, "step": 29420 }, { "epoch": 0.16, "learning_rate": 2.2828722094774268e-08, "loss": 0.8341, "step": 29421 }, { "epoch": 0.16, "learning_rate": 2.275598888213826e-08, "loss": 0.8416, "step": 29422 }, { "epoch": 0.16, "learning_rate": 2.2683371587912672e-08, "loss": 0.83, "step": 29423 }, { "epoch": 0.16, "learning_rate": 2.2610870212942394e-08, "loss": 0.7969, "step": 29424 }, { "epoch": 0.16, "learning_rate": 2.2538484758067858e-08, "loss": 0.8378, "step": 29425 }, { "epoch": 0.16, "learning_rate": 2.2466215224130615e-08, "loss": 0.8606, "step": 29426 }, { "epoch": 0.16, "learning_rate": 2.239406161197111e-08, "loss": 0.8506, "step": 29427 }, { "epoch": 0.16, "learning_rate": 2.232202392242755e-08, "loss": 0.859, "step": 29428 }, { "epoch": 0.16, "learning_rate": 2.225010215633594e-08, "loss": 0.8436, "step": 29429 }, { "epoch": 0.16, "learning_rate": 2.2178296314532278e-08, "loss": 0.8604, "step": 29430 }, { "epoch": 0.16, "learning_rate": 2.210660639785145e-08, "loss": 0.8865, "step": 29431 }, { "epoch": 0.16, "learning_rate": 2.2035032407126123e-08, "loss": 0.8216, "step": 29432 }, { "epoch": 0.16, "learning_rate": 2.1963574343186745e-08, "loss": 0.8388, "step": 29433 }, { "epoch": 0.16, "learning_rate": 2.189223220686487e-08, "loss": 0.843, "step": 29434 }, { "epoch": 0.16, "learning_rate": 2.182100599898873e-08, "loss": 0.8597, "step": 29435 }, { "epoch": 0.16, "learning_rate": 2.174989572038655e-08, "loss": 0.8349, "step": 29436 }, { "epoch": 0.16, "learning_rate": 2.167890137188322e-08, "loss": 0.83, "step": 29437 }, { "epoch": 0.16, "learning_rate": 2.1608022954305863e-08, "loss": 0.8506, "step": 29438 }, { "epoch": 0.16, "learning_rate": 2.153726046847493e-08, "loss": 0.8066, "step": 29439 }, { "epoch": 0.16, "learning_rate": 2.1466613915214208e-08, "loss": 0.8591, "step": 29440 }, { "epoch": 0.16, "learning_rate": 2.139608329534526e-08, "loss": 0.8904, "step": 29441 }, { "epoch": 0.16, "learning_rate": 2.1325668609686323e-08, "loss": 0.7997, "step": 29442 }, { "epoch": 0.16, "learning_rate": 2.125536985905563e-08, "loss": 0.8367, "step": 29443 }, { "epoch": 0.16, "learning_rate": 2.1185187044270307e-08, "loss": 0.7891, "step": 29444 }, { "epoch": 0.16, "learning_rate": 2.1115120166145254e-08, "loss": 0.8578, "step": 29445 }, { "epoch": 0.16, "learning_rate": 2.1045169225495376e-08, "loss": 0.8411, "step": 29446 }, { "epoch": 0.16, "learning_rate": 2.097533422313225e-08, "loss": 0.8214, "step": 29447 }, { "epoch": 0.16, "learning_rate": 2.0905615159868554e-08, "loss": 0.8333, "step": 29448 }, { "epoch": 0.16, "learning_rate": 2.0836012036512532e-08, "loss": 0.8432, "step": 29449 }, { "epoch": 0.16, "learning_rate": 2.0766524853874648e-08, "loss": 0.8701, "step": 29450 }, { "epoch": 0.16, "learning_rate": 2.0697153612762033e-08, "loss": 0.8667, "step": 29451 }, { "epoch": 0.16, "learning_rate": 2.062789831397849e-08, "loss": 0.8426, "step": 29452 }, { "epoch": 0.16, "learning_rate": 2.055875895833115e-08, "loss": 0.8548, "step": 29453 }, { "epoch": 0.16, "learning_rate": 2.0489735546622702e-08, "loss": 0.843, "step": 29454 }, { "epoch": 0.16, "learning_rate": 2.042082807965473e-08, "loss": 0.8677, "step": 29455 }, { "epoch": 0.16, "learning_rate": 2.0352036558227704e-08, "loss": 0.7909, "step": 29456 }, { "epoch": 0.16, "learning_rate": 2.0283360983140987e-08, "loss": 0.8238, "step": 29457 }, { "epoch": 0.17, "learning_rate": 2.0214801355192826e-08, "loss": 0.9052, "step": 29458 }, { "epoch": 0.17, "learning_rate": 2.0146357675179252e-08, "loss": 0.8402, "step": 29459 }, { "epoch": 0.17, "learning_rate": 2.0078029943895184e-08, "loss": 0.8302, "step": 29460 }, { "epoch": 0.17, "learning_rate": 2.0009818162135543e-08, "loss": 0.8247, "step": 29461 }, { "epoch": 0.17, "learning_rate": 1.9941722330691915e-08, "loss": 0.8459, "step": 29462 }, { "epoch": 0.17, "learning_rate": 1.9873742450355894e-08, "loss": 0.8503, "step": 29463 }, { "epoch": 0.17, "learning_rate": 1.9805878521917955e-08, "loss": 0.8387, "step": 29464 }, { "epoch": 0.17, "learning_rate": 1.9738130546165246e-08, "loss": 0.8709, "step": 29465 }, { "epoch": 0.17, "learning_rate": 1.9670498523884918e-08, "loss": 0.8305, "step": 29466 }, { "epoch": 0.17, "learning_rate": 1.9602982455864117e-08, "loss": 0.853, "step": 29467 }, { "epoch": 0.17, "learning_rate": 1.953558234288555e-08, "loss": 0.8144, "step": 29468 }, { "epoch": 0.17, "learning_rate": 1.9468298185734147e-08, "loss": 0.8633, "step": 29469 }, { "epoch": 0.17, "learning_rate": 1.9401129985189283e-08, "loss": 0.8635, "step": 29470 }, { "epoch": 0.17, "learning_rate": 1.9334077742033662e-08, "loss": 0.8693, "step": 29471 }, { "epoch": 0.17, "learning_rate": 1.9267141457045557e-08, "loss": 0.7831, "step": 29472 }, { "epoch": 0.17, "learning_rate": 1.920032113100212e-08, "loss": 0.839, "step": 29473 }, { "epoch": 0.17, "learning_rate": 1.9133616764679396e-08, "loss": 0.8, "step": 29474 }, { "epoch": 0.17, "learning_rate": 1.9067028358853433e-08, "loss": 0.8313, "step": 29475 }, { "epoch": 0.17, "learning_rate": 1.9000555914296948e-08, "loss": 0.8122, "step": 29476 }, { "epoch": 0.17, "learning_rate": 1.8934199431783763e-08, "loss": 0.8664, "step": 29477 }, { "epoch": 0.17, "learning_rate": 1.886795891208326e-08, "loss": 0.844, "step": 29478 }, { "epoch": 0.17, "learning_rate": 1.880183435596483e-08, "loss": 0.8466, "step": 29479 }, { "epoch": 0.17, "learning_rate": 1.8735825764197857e-08, "loss": 0.8243, "step": 29480 }, { "epoch": 0.17, "learning_rate": 1.86699331375495e-08, "loss": 0.8499, "step": 29481 }, { "epoch": 0.17, "learning_rate": 1.8604156476783595e-08, "loss": 0.8292, "step": 29482 }, { "epoch": 0.17, "learning_rate": 1.8538495782666198e-08, "loss": 0.9001, "step": 29483 }, { "epoch": 0.17, "learning_rate": 1.847295105595892e-08, "loss": 0.7874, "step": 29484 }, { "epoch": 0.17, "learning_rate": 1.8407522297424484e-08, "loss": 0.8186, "step": 29485 }, { "epoch": 0.17, "learning_rate": 1.8342209507822285e-08, "loss": 0.8676, "step": 29486 }, { "epoch": 0.17, "learning_rate": 1.82770126879106e-08, "loss": 0.8726, "step": 29487 }, { "epoch": 0.17, "learning_rate": 1.8211931838447715e-08, "loss": 0.8028, "step": 29488 }, { "epoch": 0.17, "learning_rate": 1.8146966960188584e-08, "loss": 0.8521, "step": 29489 }, { "epoch": 0.17, "learning_rate": 1.8082118053890374e-08, "loss": 0.8723, "step": 29490 }, { "epoch": 0.17, "learning_rate": 1.801738512030471e-08, "loss": 0.8731, "step": 29491 }, { "epoch": 0.17, "learning_rate": 1.795276816018321e-08, "loss": 0.8491, "step": 29492 }, { "epoch": 0.17, "learning_rate": 1.7888267174278607e-08, "loss": 0.8462, "step": 29493 }, { "epoch": 0.17, "learning_rate": 1.7823882163338078e-08, "loss": 0.835, "step": 29494 }, { "epoch": 0.17, "learning_rate": 1.775961312811103e-08, "loss": 0.8422, "step": 29495 }, { "epoch": 0.17, "learning_rate": 1.769546006934464e-08, "loss": 0.7854, "step": 29496 }, { "epoch": 0.17, "learning_rate": 1.7631422987782754e-08, "loss": 0.8863, "step": 29497 }, { "epoch": 0.17, "learning_rate": 1.7567501884170335e-08, "loss": 0.8234, "step": 29498 }, { "epoch": 0.17, "learning_rate": 1.7503696759249012e-08, "loss": 0.847, "step": 29499 }, { "epoch": 0.17, "learning_rate": 1.744000761376152e-08, "loss": 0.8364, "step": 29500 }, { "epoch": 0.17, "learning_rate": 1.7376434448447276e-08, "loss": 0.8768, "step": 29501 }, { "epoch": 0.17, "learning_rate": 1.7312977264044572e-08, "loss": 0.8697, "step": 29502 }, { "epoch": 0.17, "learning_rate": 1.7249636061290598e-08, "loss": 0.8455, "step": 29503 }, { "epoch": 0.17, "learning_rate": 1.718641084092143e-08, "loss": 0.8399, "step": 29504 }, { "epoch": 0.17, "learning_rate": 1.712330160367315e-08, "loss": 0.8467, "step": 29505 }, { "epoch": 0.17, "learning_rate": 1.706030835027628e-08, "loss": 0.8539, "step": 29506 }, { "epoch": 0.17, "learning_rate": 1.6997431081463567e-08, "loss": 0.8997, "step": 29507 }, { "epoch": 0.17, "learning_rate": 1.693466979796665e-08, "loss": 0.8092, "step": 29508 }, { "epoch": 0.17, "learning_rate": 1.6872024500513838e-08, "loss": 0.8621, "step": 29509 }, { "epoch": 0.17, "learning_rate": 1.6809495189833435e-08, "loss": 0.8237, "step": 29510 }, { "epoch": 0.17, "learning_rate": 1.6747081866651528e-08, "loss": 0.8692, "step": 29511 }, { "epoch": 0.17, "learning_rate": 1.6684784531693087e-08, "loss": 0.8726, "step": 29512 }, { "epoch": 0.17, "learning_rate": 1.6622603185683094e-08, "loss": 0.8322, "step": 29513 }, { "epoch": 0.17, "learning_rate": 1.6560537829342082e-08, "loss": 0.8478, "step": 29514 }, { "epoch": 0.17, "learning_rate": 1.6498588463392805e-08, "loss": 0.8396, "step": 29515 }, { "epoch": 0.17, "learning_rate": 1.643675508855358e-08, "loss": 0.7788, "step": 29516 }, { "epoch": 0.17, "learning_rate": 1.6375037705543827e-08, "loss": 0.8082, "step": 29517 }, { "epoch": 0.17, "learning_rate": 1.6313436315080754e-08, "loss": 0.8501, "step": 29518 }, { "epoch": 0.17, "learning_rate": 1.6251950917879345e-08, "loss": 0.8855, "step": 29519 }, { "epoch": 0.17, "learning_rate": 1.619058151465458e-08, "loss": 0.8001, "step": 29520 }, { "epoch": 0.17, "learning_rate": 1.6129328106118114e-08, "loss": 0.8464, "step": 29521 }, { "epoch": 0.17, "learning_rate": 1.606819069298271e-08, "loss": 0.8436, "step": 29522 }, { "epoch": 0.17, "learning_rate": 1.6007169275960022e-08, "loss": 0.8364, "step": 29523 }, { "epoch": 0.17, "learning_rate": 1.594626385575615e-08, "loss": 0.8236, "step": 29524 }, { "epoch": 0.17, "learning_rate": 1.5885474433080526e-08, "loss": 0.8216, "step": 29525 }, { "epoch": 0.17, "learning_rate": 1.5824801008639257e-08, "loss": 0.8173, "step": 29526 }, { "epoch": 0.17, "learning_rate": 1.5764243583137327e-08, "loss": 0.8635, "step": 29527 }, { "epoch": 0.17, "learning_rate": 1.5703802157277514e-08, "loss": 0.8827, "step": 29528 }, { "epoch": 0.17, "learning_rate": 1.5643476731762585e-08, "loss": 0.8524, "step": 29529 }, { "epoch": 0.17, "learning_rate": 1.558326730729309e-08, "loss": 0.8299, "step": 29530 }, { "epoch": 0.17, "learning_rate": 1.5523173884569588e-08, "loss": 0.8393, "step": 29531 }, { "epoch": 0.17, "learning_rate": 1.546319646428929e-08, "loss": 0.8384, "step": 29532 }, { "epoch": 0.17, "learning_rate": 1.5403335047148306e-08, "loss": 0.823, "step": 29533 }, { "epoch": 0.17, "learning_rate": 1.5343589633844968e-08, "loss": 0.8427, "step": 29534 }, { "epoch": 0.17, "learning_rate": 1.528396022506984e-08, "loss": 0.8636, "step": 29535 }, { "epoch": 0.17, "learning_rate": 1.5224446821517914e-08, "loss": 0.8357, "step": 29536 }, { "epoch": 0.17, "learning_rate": 1.5165049423880863e-08, "loss": 0.8319, "step": 29537 }, { "epoch": 0.17, "learning_rate": 1.5105768032847023e-08, "loss": 0.8175, "step": 29538 }, { "epoch": 0.17, "learning_rate": 1.5046602649106955e-08, "loss": 0.8353, "step": 29539 }, { "epoch": 0.17, "learning_rate": 1.4987553273346778e-08, "loss": 0.8266, "step": 29540 }, { "epoch": 0.17, "learning_rate": 1.4928619906252606e-08, "loss": 0.8535, "step": 29541 }, { "epoch": 0.17, "learning_rate": 1.486980254851056e-08, "loss": 0.8487, "step": 29542 }, { "epoch": 0.17, "learning_rate": 1.4811101200802314e-08, "loss": 0.8503, "step": 29543 }, { "epoch": 0.17, "learning_rate": 1.4752515863810657e-08, "loss": 0.829, "step": 29544 }, { "epoch": 0.17, "learning_rate": 1.4694046538216155e-08, "loss": 0.8281, "step": 29545 }, { "epoch": 0.17, "learning_rate": 1.4635693224698266e-08, "loss": 0.8746, "step": 29546 }, { "epoch": 0.17, "learning_rate": 1.4577455923935335e-08, "loss": 0.8527, "step": 29547 }, { "epoch": 0.17, "learning_rate": 1.4519334636602378e-08, "loss": 0.8271, "step": 29548 }, { "epoch": 0.17, "learning_rate": 1.446132936337663e-08, "loss": 0.8125, "step": 29549 }, { "epoch": 0.17, "learning_rate": 1.4403440104929778e-08, "loss": 0.8355, "step": 29550 }, { "epoch": 0.17, "learning_rate": 1.4345666861937945e-08, "loss": 0.8181, "step": 29551 }, { "epoch": 0.17, "learning_rate": 1.4288009635068378e-08, "loss": 0.8558, "step": 29552 }, { "epoch": 0.17, "learning_rate": 1.423046842499276e-08, "loss": 0.8527, "step": 29553 }, { "epoch": 0.17, "learning_rate": 1.4173043232380557e-08, "loss": 0.8331, "step": 29554 }, { "epoch": 0.17, "learning_rate": 1.4115734057897902e-08, "loss": 0.8429, "step": 29555 }, { "epoch": 0.17, "learning_rate": 1.4058540902209816e-08, "loss": 0.8426, "step": 29556 }, { "epoch": 0.17, "learning_rate": 1.4001463765982437e-08, "loss": 0.8482, "step": 29557 }, { "epoch": 0.17, "learning_rate": 1.3944502649878566e-08, "loss": 0.8484, "step": 29558 }, { "epoch": 0.17, "learning_rate": 1.3887657554558787e-08, "loss": 0.8701, "step": 29559 }, { "epoch": 0.17, "learning_rate": 1.383092848068368e-08, "loss": 0.8298, "step": 29560 }, { "epoch": 0.17, "learning_rate": 1.3774315428913831e-08, "loss": 0.8666, "step": 29561 }, { "epoch": 0.17, "learning_rate": 1.3717818399906491e-08, "loss": 0.8278, "step": 29562 }, { "epoch": 0.17, "learning_rate": 1.36614373943178e-08, "loss": 0.8657, "step": 29563 }, { "epoch": 0.17, "learning_rate": 1.3605172412801681e-08, "loss": 0.7964, "step": 29564 }, { "epoch": 0.17, "learning_rate": 1.3549023456013166e-08, "loss": 0.8355, "step": 29565 }, { "epoch": 0.17, "learning_rate": 1.3492990524605065e-08, "loss": 0.8396, "step": 29566 }, { "epoch": 0.17, "learning_rate": 1.3437073619226859e-08, "loss": 0.8313, "step": 29567 }, { "epoch": 0.17, "learning_rate": 1.3381272740529139e-08, "loss": 0.8186, "step": 29568 }, { "epoch": 0.17, "learning_rate": 1.3325587889160274e-08, "loss": 0.843, "step": 29569 }, { "epoch": 0.17, "learning_rate": 1.3270019065767526e-08, "loss": 0.8395, "step": 29570 }, { "epoch": 0.17, "learning_rate": 1.3214566270994821e-08, "loss": 0.8469, "step": 29571 }, { "epoch": 0.17, "learning_rate": 1.3159229505488313e-08, "loss": 0.8507, "step": 29572 }, { "epoch": 0.17, "learning_rate": 1.310400876989082e-08, "loss": 0.8867, "step": 29573 }, { "epoch": 0.17, "learning_rate": 1.3048904064841827e-08, "loss": 0.8262, "step": 29574 }, { "epoch": 0.17, "learning_rate": 1.2993915390984158e-08, "loss": 0.8719, "step": 29575 }, { "epoch": 0.17, "learning_rate": 1.2939042748955078e-08, "loss": 0.8855, "step": 29576 }, { "epoch": 0.17, "learning_rate": 1.2884286139392965e-08, "loss": 0.8453, "step": 29577 }, { "epoch": 0.17, "learning_rate": 1.2829645562933978e-08, "loss": 0.8066, "step": 29578 }, { "epoch": 0.17, "learning_rate": 1.2775121020212056e-08, "loss": 0.8055, "step": 29579 }, { "epoch": 0.17, "learning_rate": 1.2720712511861133e-08, "loss": 0.8476, "step": 29580 }, { "epoch": 0.17, "learning_rate": 1.2666420038512928e-08, "loss": 0.8731, "step": 29581 }, { "epoch": 0.17, "learning_rate": 1.2612243600799156e-08, "loss": 0.9014, "step": 29582 }, { "epoch": 0.17, "learning_rate": 1.2558183199349316e-08, "loss": 0.8648, "step": 29583 }, { "epoch": 0.17, "learning_rate": 1.2504238834790683e-08, "loss": 0.8387, "step": 29584 }, { "epoch": 0.17, "learning_rate": 1.2450410507750533e-08, "loss": 0.8378, "step": 29585 }, { "epoch": 0.17, "learning_rate": 1.239669821885281e-08, "loss": 0.8618, "step": 29586 }, { "epoch": 0.17, "learning_rate": 1.234310196872368e-08, "loss": 0.8591, "step": 29587 }, { "epoch": 0.17, "learning_rate": 1.2289621757985981e-08, "loss": 0.8192, "step": 29588 }, { "epoch": 0.17, "learning_rate": 1.2236257587258105e-08, "loss": 0.8807, "step": 29589 }, { "epoch": 0.17, "learning_rate": 1.2183009457163997e-08, "loss": 0.8432, "step": 29590 }, { "epoch": 0.17, "learning_rate": 1.212987736831872e-08, "loss": 0.8153, "step": 29591 }, { "epoch": 0.17, "learning_rate": 1.2076861321341783e-08, "loss": 0.8571, "step": 29592 }, { "epoch": 0.17, "learning_rate": 1.2023961316848243e-08, "loss": 0.845, "step": 29593 }, { "epoch": 0.17, "learning_rate": 1.1971177355453168e-08, "loss": 0.8431, "step": 29594 }, { "epoch": 0.17, "learning_rate": 1.1918509437769399e-08, "loss": 0.8306, "step": 29595 }, { "epoch": 0.17, "learning_rate": 1.186595756440978e-08, "loss": 0.8206, "step": 29596 }, { "epoch": 0.17, "learning_rate": 1.1813521735983825e-08, "loss": 0.8585, "step": 29597 }, { "epoch": 0.17, "learning_rate": 1.1761201953099931e-08, "loss": 0.7957, "step": 29598 }, { "epoch": 0.17, "learning_rate": 1.1708998216368728e-08, "loss": 0.8442, "step": 29599 }, { "epoch": 0.17, "learning_rate": 1.1656910526394171e-08, "loss": 0.8321, "step": 29600 }, { "epoch": 0.17, "learning_rate": 1.1604938883782447e-08, "loss": 0.8484, "step": 29601 }, { "epoch": 0.17, "learning_rate": 1.1553083289137512e-08, "loss": 0.8887, "step": 29602 }, { "epoch": 0.17, "learning_rate": 1.150134374306222e-08, "loss": 0.8617, "step": 29603 }, { "epoch": 0.17, "learning_rate": 1.1449720246156093e-08, "loss": 0.8246, "step": 29604 }, { "epoch": 0.17, "learning_rate": 1.1398212799020869e-08, "loss": 0.8717, "step": 29605 }, { "epoch": 0.17, "learning_rate": 1.1346821402252739e-08, "loss": 0.8436, "step": 29606 }, { "epoch": 0.17, "learning_rate": 1.1295546056451224e-08, "loss": 0.8686, "step": 29607 }, { "epoch": 0.17, "learning_rate": 1.124438676221029e-08, "loss": 0.8701, "step": 29608 }, { "epoch": 0.17, "learning_rate": 1.1193343520126132e-08, "loss": 0.8519, "step": 29609 }, { "epoch": 0.17, "learning_rate": 1.1142416330789385e-08, "loss": 0.8242, "step": 29610 }, { "epoch": 0.17, "learning_rate": 1.1091605194792909e-08, "loss": 0.85, "step": 29611 }, { "epoch": 0.17, "learning_rate": 1.1040910112727344e-08, "loss": 0.8543, "step": 29612 }, { "epoch": 0.17, "learning_rate": 1.0990331085181105e-08, "loss": 0.8063, "step": 29613 }, { "epoch": 0.17, "learning_rate": 1.0939868112742613e-08, "loss": 0.8083, "step": 29614 }, { "epoch": 0.17, "learning_rate": 1.0889521195998065e-08, "loss": 0.8609, "step": 29615 }, { "epoch": 0.17, "learning_rate": 1.0839290335531438e-08, "loss": 0.8489, "step": 29616 }, { "epoch": 0.17, "learning_rate": 1.078917553192671e-08, "loss": 0.8591, "step": 29617 }, { "epoch": 0.17, "learning_rate": 1.0739176785766747e-08, "loss": 0.8366, "step": 29618 }, { "epoch": 0.17, "learning_rate": 1.0689294097632197e-08, "loss": 0.7945, "step": 29619 }, { "epoch": 0.17, "learning_rate": 1.0639527468102595e-08, "loss": 0.8505, "step": 29620 }, { "epoch": 0.17, "learning_rate": 1.0589876897755258e-08, "loss": 0.8452, "step": 29621 }, { "epoch": 0.17, "learning_rate": 1.0540342387167501e-08, "loss": 0.8361, "step": 29622 }, { "epoch": 0.17, "learning_rate": 1.0490923936916642e-08, "loss": 0.869, "step": 29623 }, { "epoch": 0.17, "learning_rate": 1.0441621547574443e-08, "loss": 0.8346, "step": 29624 }, { "epoch": 0.17, "learning_rate": 1.0392435219713781e-08, "loss": 0.8151, "step": 29625 }, { "epoch": 0.17, "learning_rate": 1.0343364953908642e-08, "loss": 0.8392, "step": 29626 }, { "epoch": 0.17, "learning_rate": 1.0294410750726346e-08, "loss": 0.8497, "step": 29627 }, { "epoch": 0.17, "learning_rate": 1.0245572610736443e-08, "loss": 0.8406, "step": 29628 }, { "epoch": 0.17, "learning_rate": 1.0196850534507363e-08, "loss": 0.8663, "step": 29629 }, { "epoch": 0.17, "learning_rate": 1.0148244522603102e-08, "loss": 0.8124, "step": 29630 }, { "epoch": 0.17, "learning_rate": 1.0099754575590981e-08, "loss": 0.8835, "step": 29631 }, { "epoch": 0.17, "learning_rate": 1.0051380694032774e-08, "loss": 0.7868, "step": 29632 }, { "epoch": 0.17, "learning_rate": 1.0003122878491367e-08, "loss": 0.8684, "step": 29633 }, { "epoch": 0.17, "learning_rate": 9.954981129526309e-09, "loss": 0.8402, "step": 29634 }, { "epoch": 0.17, "learning_rate": 9.906955447697153e-09, "loss": 0.8031, "step": 29635 }, { "epoch": 0.17, "learning_rate": 9.859045833562341e-09, "loss": 0.8217, "step": 29636 }, { "epoch": 0.17, "learning_rate": 9.811252287679206e-09, "loss": 0.8351, "step": 29637 }, { "epoch": 0.17, "learning_rate": 9.76357481060175e-09, "loss": 0.8419, "step": 29638 }, { "epoch": 0.17, "learning_rate": 9.716013402883972e-09, "loss": 0.8444, "step": 29639 }, { "epoch": 0.17, "learning_rate": 9.668568065079876e-09, "loss": 0.8169, "step": 29640 }, { "epoch": 0.17, "learning_rate": 9.621238797737908e-09, "loss": 0.8332, "step": 29641 }, { "epoch": 0.17, "learning_rate": 9.574025601410963e-09, "loss": 0.8793, "step": 29642 }, { "epoch": 0.17, "learning_rate": 9.52692847664638e-09, "loss": 0.8443, "step": 29643 }, { "epoch": 0.17, "learning_rate": 9.479947423991498e-09, "loss": 0.8257, "step": 29644 }, { "epoch": 0.17, "learning_rate": 9.433082443991438e-09, "loss": 0.8636, "step": 29645 }, { "epoch": 0.17, "learning_rate": 9.386333537191317e-09, "loss": 0.8312, "step": 29646 }, { "epoch": 0.17, "learning_rate": 9.339700704134035e-09, "loss": 0.8821, "step": 29647 }, { "epoch": 0.17, "learning_rate": 9.293183945361384e-09, "loss": 0.8363, "step": 29648 }, { "epoch": 0.17, "learning_rate": 9.246783261414038e-09, "loss": 0.8391, "step": 29649 }, { "epoch": 0.17, "learning_rate": 9.200498652831569e-09, "loss": 0.8047, "step": 29650 }, { "epoch": 0.17, "learning_rate": 9.154330120150212e-09, "loss": 0.8529, "step": 29651 }, { "epoch": 0.17, "learning_rate": 9.108277663907316e-09, "loss": 0.8454, "step": 29652 }, { "epoch": 0.17, "learning_rate": 9.06234128463912e-09, "loss": 0.8234, "step": 29653 }, { "epoch": 0.17, "learning_rate": 9.01652098287742e-09, "loss": 0.8355, "step": 29654 }, { "epoch": 0.17, "learning_rate": 8.970816759155121e-09, "loss": 0.875, "step": 29655 }, { "epoch": 0.17, "learning_rate": 8.925228614002913e-09, "loss": 0.8402, "step": 29656 }, { "epoch": 0.17, "learning_rate": 8.879756547951479e-09, "loss": 0.8276, "step": 29657 }, { "epoch": 0.17, "learning_rate": 8.834400561529289e-09, "loss": 0.8392, "step": 29658 }, { "epoch": 0.17, "learning_rate": 8.789160655261475e-09, "loss": 0.901, "step": 29659 }, { "epoch": 0.17, "learning_rate": 8.744036829674284e-09, "loss": 0.8767, "step": 29660 }, { "epoch": 0.17, "learning_rate": 8.699029085293965e-09, "loss": 0.8098, "step": 29661 }, { "epoch": 0.17, "learning_rate": 8.654137422641207e-09, "loss": 0.8459, "step": 29662 }, { "epoch": 0.17, "learning_rate": 8.609361842237818e-09, "loss": 0.8438, "step": 29663 }, { "epoch": 0.17, "learning_rate": 8.564702344605602e-09, "loss": 0.7989, "step": 29664 }, { "epoch": 0.17, "learning_rate": 8.520158930260813e-09, "loss": 0.8259, "step": 29665 }, { "epoch": 0.17, "learning_rate": 8.475731599724147e-09, "loss": 0.871, "step": 29666 }, { "epoch": 0.17, "learning_rate": 8.431420353508523e-09, "loss": 0.8295, "step": 29667 }, { "epoch": 0.17, "learning_rate": 8.38722519213131e-09, "loss": 0.8676, "step": 29668 }, { "epoch": 0.17, "learning_rate": 8.343146116104317e-09, "loss": 0.819, "step": 29669 }, { "epoch": 0.17, "learning_rate": 8.29918312594047e-09, "loss": 0.8152, "step": 29670 }, { "epoch": 0.17, "learning_rate": 8.255336222150467e-09, "loss": 0.8798, "step": 29671 }, { "epoch": 0.17, "learning_rate": 8.211605405243906e-09, "loss": 0.814, "step": 29672 }, { "epoch": 0.17, "learning_rate": 8.167990675729265e-09, "loss": 0.8739, "step": 29673 }, { "epoch": 0.17, "learning_rate": 8.124492034111698e-09, "loss": 0.7856, "step": 29674 }, { "epoch": 0.17, "learning_rate": 8.081109480897465e-09, "loss": 0.8545, "step": 29675 }, { "epoch": 0.17, "learning_rate": 8.037843016590608e-09, "loss": 0.8265, "step": 29676 }, { "epoch": 0.17, "learning_rate": 7.994692641694058e-09, "loss": 0.8742, "step": 29677 }, { "epoch": 0.17, "learning_rate": 7.951658356709634e-09, "loss": 0.823, "step": 29678 }, { "epoch": 0.17, "learning_rate": 7.90874016213583e-09, "loss": 0.8288, "step": 29679 }, { "epoch": 0.17, "learning_rate": 7.865938058472245e-09, "loss": 0.8535, "step": 29680 }, { "epoch": 0.17, "learning_rate": 7.823252046216256e-09, "loss": 0.8466, "step": 29681 }, { "epoch": 0.17, "learning_rate": 7.780682125863026e-09, "loss": 0.8491, "step": 29682 }, { "epoch": 0.17, "learning_rate": 7.738228297907712e-09, "loss": 0.8602, "step": 29683 }, { "epoch": 0.17, "learning_rate": 7.695890562844365e-09, "loss": 0.8131, "step": 29684 }, { "epoch": 0.17, "learning_rate": 7.653668921162594e-09, "loss": 0.8449, "step": 29685 }, { "epoch": 0.17, "learning_rate": 7.611563373355335e-09, "loss": 0.8208, "step": 29686 }, { "epoch": 0.17, "learning_rate": 7.569573919911088e-09, "loss": 0.8824, "step": 29687 }, { "epoch": 0.17, "learning_rate": 7.527700561317241e-09, "loss": 0.8378, "step": 29688 }, { "epoch": 0.17, "learning_rate": 7.485943298060071e-09, "loss": 0.8504, "step": 29689 }, { "epoch": 0.17, "learning_rate": 7.444302130624747e-09, "loss": 0.8421, "step": 29690 }, { "epoch": 0.17, "learning_rate": 7.402777059496436e-09, "loss": 0.8507, "step": 29691 }, { "epoch": 0.17, "learning_rate": 7.361368085155862e-09, "loss": 0.8149, "step": 29692 }, { "epoch": 0.17, "learning_rate": 7.3200752080859746e-09, "loss": 0.8615, "step": 29693 }, { "epoch": 0.17, "learning_rate": 7.278898428764169e-09, "loss": 0.8509, "step": 29694 }, { "epoch": 0.17, "learning_rate": 7.237837747671173e-09, "loss": 0.849, "step": 29695 }, { "epoch": 0.17, "learning_rate": 7.196893165282159e-09, "loss": 0.8341, "step": 29696 }, { "epoch": 0.17, "learning_rate": 7.156064682073416e-09, "loss": 0.848, "step": 29697 }, { "epoch": 0.17, "learning_rate": 7.115352298520117e-09, "loss": 0.844, "step": 29698 }, { "epoch": 0.17, "learning_rate": 7.0747560150941086e-09, "loss": 0.8518, "step": 29699 }, { "epoch": 0.17, "learning_rate": 7.034275832268344e-09, "loss": 0.8665, "step": 29700 }, { "epoch": 0.17, "learning_rate": 6.99391175051245e-09, "loss": 0.8452, "step": 29701 }, { "epoch": 0.17, "learning_rate": 6.9536637702949385e-09, "loss": 0.8341, "step": 29702 }, { "epoch": 0.17, "learning_rate": 6.9135318920832136e-09, "loss": 0.8884, "step": 29703 }, { "epoch": 0.17, "learning_rate": 6.873516116345791e-09, "loss": 0.8176, "step": 29704 }, { "epoch": 0.17, "learning_rate": 6.8336164435445215e-09, "loss": 0.8516, "step": 29705 }, { "epoch": 0.17, "learning_rate": 6.793832874144591e-09, "loss": 0.8297, "step": 29706 }, { "epoch": 0.17, "learning_rate": 6.754165408608959e-09, "loss": 0.8801, "step": 29707 }, { "epoch": 0.17, "learning_rate": 6.714614047396151e-09, "loss": 0.8122, "step": 29708 }, { "epoch": 0.17, "learning_rate": 6.6751787909680175e-09, "loss": 0.8396, "step": 29709 }, { "epoch": 0.17, "learning_rate": 6.635859639781972e-09, "loss": 0.836, "step": 29710 }, { "epoch": 0.17, "learning_rate": 6.596656594294315e-09, "loss": 0.8586, "step": 29711 }, { "epoch": 0.17, "learning_rate": 6.557569654960238e-09, "loss": 0.7843, "step": 29712 }, { "epoch": 0.17, "learning_rate": 6.518598822234934e-09, "loss": 0.8384, "step": 29713 }, { "epoch": 0.17, "learning_rate": 6.479744096571372e-09, "loss": 0.8362, "step": 29714 }, { "epoch": 0.17, "learning_rate": 6.441005478420303e-09, "loss": 0.8665, "step": 29715 }, { "epoch": 0.17, "learning_rate": 6.402382968231369e-09, "loss": 0.808, "step": 29716 }, { "epoch": 0.17, "learning_rate": 6.3638765664530975e-09, "loss": 0.8715, "step": 29717 }, { "epoch": 0.17, "learning_rate": 6.3254862735351304e-09, "loss": 0.8295, "step": 29718 }, { "epoch": 0.17, "learning_rate": 6.287212089920447e-09, "loss": 0.8532, "step": 29719 }, { "epoch": 0.17, "learning_rate": 6.249054016056466e-09, "loss": 0.8304, "step": 29720 }, { "epoch": 0.17, "learning_rate": 6.211012052383947e-09, "loss": 0.7947, "step": 29721 }, { "epoch": 0.17, "learning_rate": 6.173086199346978e-09, "loss": 0.848, "step": 29722 }, { "epoch": 0.17, "learning_rate": 6.135276457385209e-09, "loss": 0.8666, "step": 29723 }, { "epoch": 0.17, "learning_rate": 6.097582826938286e-09, "loss": 0.8696, "step": 29724 }, { "epoch": 0.17, "learning_rate": 6.060005308443639e-09, "loss": 0.8446, "step": 29725 }, { "epoch": 0.17, "learning_rate": 6.022543902337586e-09, "loss": 0.8886, "step": 29726 }, { "epoch": 0.17, "learning_rate": 5.985198609057552e-09, "loss": 0.8612, "step": 29727 }, { "epoch": 0.17, "learning_rate": 5.947969429034306e-09, "loss": 0.8912, "step": 29728 }, { "epoch": 0.17, "learning_rate": 5.9108563627030545e-09, "loss": 0.8552, "step": 29729 }, { "epoch": 0.17, "learning_rate": 5.873859410492344e-09, "loss": 0.8351, "step": 29730 }, { "epoch": 0.17, "learning_rate": 5.8369785728351614e-09, "loss": 0.8658, "step": 29731 }, { "epoch": 0.17, "learning_rate": 5.800213850157832e-09, "loss": 0.8457, "step": 29732 }, { "epoch": 0.17, "learning_rate": 5.763565242887792e-09, "loss": 0.8597, "step": 29733 }, { "epoch": 0.17, "learning_rate": 5.727032751450256e-09, "loss": 0.8463, "step": 29734 }, { "epoch": 0.17, "learning_rate": 5.690616376271552e-09, "loss": 0.849, "step": 29735 }, { "epoch": 0.17, "learning_rate": 5.6543161177735615e-09, "loss": 0.822, "step": 29736 }, { "epoch": 0.17, "learning_rate": 5.618131976378172e-09, "loss": 0.8748, "step": 29737 }, { "epoch": 0.17, "learning_rate": 5.582063952505046e-09, "loss": 0.8554, "step": 29738 }, { "epoch": 0.17, "learning_rate": 5.546112046574958e-09, "loss": 0.8624, "step": 29739 }, { "epoch": 0.17, "learning_rate": 5.510276259004244e-09, "loss": 0.844, "step": 29740 }, { "epoch": 0.17, "learning_rate": 5.474556590209234e-09, "loss": 0.8752, "step": 29741 }, { "epoch": 0.17, "learning_rate": 5.438953040605155e-09, "loss": 0.8721, "step": 29742 }, { "epoch": 0.17, "learning_rate": 5.403465610606118e-09, "loss": 0.8559, "step": 29743 }, { "epoch": 0.17, "learning_rate": 5.368094300625126e-09, "loss": 0.83, "step": 29744 }, { "epoch": 0.17, "learning_rate": 5.332839111070743e-09, "loss": 0.854, "step": 29745 }, { "epoch": 0.17, "learning_rate": 5.2977000423548585e-09, "loss": 0.8669, "step": 29746 }, { "epoch": 0.17, "learning_rate": 5.262677094883817e-09, "loss": 0.8656, "step": 29747 }, { "epoch": 0.17, "learning_rate": 5.227770269066179e-09, "loss": 0.7977, "step": 29748 }, { "epoch": 0.17, "learning_rate": 5.1929795653071764e-09, "loss": 0.8553, "step": 29749 }, { "epoch": 0.17, "learning_rate": 5.158304984009821e-09, "loss": 0.8069, "step": 29750 }, { "epoch": 0.17, "learning_rate": 5.123746525578232e-09, "loss": 0.8206, "step": 29751 }, { "epoch": 0.17, "learning_rate": 5.089304190414313e-09, "loss": 0.8292, "step": 29752 }, { "epoch": 0.17, "learning_rate": 5.054977978916631e-09, "loss": 0.8018, "step": 29753 }, { "epoch": 0.17, "learning_rate": 5.0207678914848675e-09, "loss": 0.8776, "step": 29754 }, { "epoch": 0.17, "learning_rate": 4.986673928516483e-09, "loss": 0.8134, "step": 29755 }, { "epoch": 0.17, "learning_rate": 4.952696090407827e-09, "loss": 0.8435, "step": 29756 }, { "epoch": 0.17, "learning_rate": 4.918834377553028e-09, "loss": 0.86, "step": 29757 }, { "epoch": 0.17, "learning_rate": 4.8850887903462155e-09, "loss": 0.8586, "step": 29758 }, { "epoch": 0.18, "learning_rate": 4.851459329178187e-09, "loss": 0.8911, "step": 29759 }, { "epoch": 0.18, "learning_rate": 4.817945994441964e-09, "loss": 0.8582, "step": 29760 }, { "epoch": 0.18, "learning_rate": 4.7845487865239015e-09, "loss": 0.8546, "step": 29761 }, { "epoch": 0.18, "learning_rate": 4.751267705814799e-09, "loss": 0.7674, "step": 29762 }, { "epoch": 0.18, "learning_rate": 4.718102752699904e-09, "loss": 0.8797, "step": 29763 }, { "epoch": 0.18, "learning_rate": 4.685053927564465e-09, "loss": 0.8407, "step": 29764 }, { "epoch": 0.18, "learning_rate": 4.652121230793727e-09, "loss": 0.8269, "step": 29765 }, { "epoch": 0.18, "learning_rate": 4.619304662768498e-09, "loss": 0.8247, "step": 29766 }, { "epoch": 0.18, "learning_rate": 4.586604223870694e-09, "loss": 0.8455, "step": 29767 }, { "epoch": 0.18, "learning_rate": 4.554019914481123e-09, "loss": 0.8637, "step": 29768 }, { "epoch": 0.18, "learning_rate": 4.521551734977259e-09, "loss": 0.8583, "step": 29769 }, { "epoch": 0.18, "learning_rate": 4.489199685736579e-09, "loss": 0.8268, "step": 29770 }, { "epoch": 0.18, "learning_rate": 4.456963767135447e-09, "loss": 0.8958, "step": 29771 }, { "epoch": 0.18, "learning_rate": 4.4248439795480105e-09, "loss": 0.839, "step": 29772 }, { "epoch": 0.18, "learning_rate": 4.392840323347303e-09, "loss": 0.8086, "step": 29773 }, { "epoch": 0.18, "learning_rate": 4.36095279890636e-09, "loss": 0.8313, "step": 29774 }, { "epoch": 0.18, "learning_rate": 4.3291814065937745e-09, "loss": 0.803, "step": 29775 }, { "epoch": 0.18, "learning_rate": 4.297526146779252e-09, "loss": 0.8433, "step": 29776 }, { "epoch": 0.18, "learning_rate": 4.2659870198313855e-09, "loss": 0.8395, "step": 29777 }, { "epoch": 0.18, "learning_rate": 4.23456402611655e-09, "loss": 0.8279, "step": 29778 }, { "epoch": 0.18, "learning_rate": 4.203257165998898e-09, "loss": 0.8531, "step": 29779 }, { "epoch": 0.18, "learning_rate": 4.172066439842582e-09, "loss": 0.8162, "step": 29780 }, { "epoch": 0.18, "learning_rate": 4.140991848009535e-09, "loss": 0.8578, "step": 29781 }, { "epoch": 0.18, "learning_rate": 4.11003339086169e-09, "loss": 0.8421, "step": 29782 }, { "epoch": 0.18, "learning_rate": 4.079191068758759e-09, "loss": 0.8579, "step": 29783 }, { "epoch": 0.18, "learning_rate": 4.048464882058234e-09, "loss": 0.843, "step": 29784 }, { "epoch": 0.18, "learning_rate": 4.0178548311176066e-09, "loss": 0.8569, "step": 29785 }, { "epoch": 0.18, "learning_rate": 3.987360916293259e-09, "loss": 0.7879, "step": 29786 }, { "epoch": 0.18, "learning_rate": 3.9569831379371315e-09, "loss": 0.8447, "step": 29787 }, { "epoch": 0.18, "learning_rate": 3.926721496404496e-09, "loss": 0.8662, "step": 29788 }, { "epoch": 0.18, "learning_rate": 3.896575992046181e-09, "loss": 0.9003, "step": 29789 }, { "epoch": 0.18, "learning_rate": 3.866546625213019e-09, "loss": 0.7893, "step": 29790 }, { "epoch": 0.18, "learning_rate": 3.836633396252509e-09, "loss": 0.8847, "step": 29791 }, { "epoch": 0.18, "learning_rate": 3.806836305512152e-09, "loss": 0.847, "step": 29792 }, { "epoch": 0.18, "learning_rate": 3.777155353340556e-09, "loss": 0.8618, "step": 29793 }, { "epoch": 0.18, "learning_rate": 3.747590540079671e-09, "loss": 0.8144, "step": 29794 }, { "epoch": 0.18, "learning_rate": 3.7181418660747757e-09, "loss": 0.8635, "step": 29795 }, { "epoch": 0.18, "learning_rate": 3.6888093316667097e-09, "loss": 0.8496, "step": 29796 }, { "epoch": 0.18, "learning_rate": 3.6595929371974203e-09, "loss": 0.8608, "step": 29797 }, { "epoch": 0.18, "learning_rate": 3.630492683006637e-09, "loss": 0.8169, "step": 29798 }, { "epoch": 0.18, "learning_rate": 3.6015085694318662e-09, "loss": 0.8725, "step": 29799 }, { "epoch": 0.18, "learning_rate": 3.5726405968083968e-09, "loss": 0.8183, "step": 29800 }, { "epoch": 0.18, "learning_rate": 3.5438887654737355e-09, "loss": 0.8531, "step": 29801 }, { "epoch": 0.18, "learning_rate": 3.5152530757620596e-09, "loss": 0.8464, "step": 29802 }, { "epoch": 0.18, "learning_rate": 3.4867335280042157e-09, "loss": 0.8867, "step": 29803 }, { "epoch": 0.18, "learning_rate": 3.458330122533271e-09, "loss": 0.8328, "step": 29804 }, { "epoch": 0.18, "learning_rate": 3.4300428596778514e-09, "loss": 0.8544, "step": 29805 }, { "epoch": 0.18, "learning_rate": 3.4018717397665823e-09, "loss": 0.847, "step": 29806 }, { "epoch": 0.18, "learning_rate": 3.3738167631292006e-09, "loss": 0.8641, "step": 29807 }, { "epoch": 0.18, "learning_rate": 3.345877930088781e-09, "loss": 0.8322, "step": 29808 }, { "epoch": 0.18, "learning_rate": 3.318055240970619e-09, "loss": 0.8337, "step": 29809 }, { "epoch": 0.18, "learning_rate": 3.2903486960988995e-09, "loss": 0.8392, "step": 29810 }, { "epoch": 0.18, "learning_rate": 3.2627582957955872e-09, "loss": 0.8733, "step": 29811 }, { "epoch": 0.18, "learning_rate": 3.2352840403804264e-09, "loss": 0.8119, "step": 29812 }, { "epoch": 0.18, "learning_rate": 3.207925930172051e-09, "loss": 0.8305, "step": 29813 }, { "epoch": 0.18, "learning_rate": 3.1806839654890954e-09, "loss": 0.8263, "step": 29814 }, { "epoch": 0.18, "learning_rate": 3.153558146649083e-09, "loss": 0.7989, "step": 29815 }, { "epoch": 0.18, "learning_rate": 3.1265484739650966e-09, "loss": 0.8604, "step": 29816 }, { "epoch": 0.18, "learning_rate": 3.09965494775355e-09, "loss": 0.8113, "step": 29817 }, { "epoch": 0.18, "learning_rate": 3.0728775683241953e-09, "loss": 0.8246, "step": 29818 }, { "epoch": 0.18, "learning_rate": 3.046216335989005e-09, "loss": 0.8806, "step": 29819 }, { "epoch": 0.18, "learning_rate": 3.0196712510588422e-09, "loss": 0.7913, "step": 29820 }, { "epoch": 0.18, "learning_rate": 2.9932423138412382e-09, "loss": 0.8496, "step": 29821 }, { "epoch": 0.18, "learning_rate": 2.9669295246426146e-09, "loss": 0.8495, "step": 29822 }, { "epoch": 0.18, "learning_rate": 2.9407328837693926e-09, "loss": 0.8311, "step": 29823 }, { "epoch": 0.18, "learning_rate": 2.914652391526884e-09, "loss": 0.8686, "step": 29824 }, { "epoch": 0.18, "learning_rate": 2.8886880482159597e-09, "loss": 0.8336, "step": 29825 }, { "epoch": 0.18, "learning_rate": 2.8628398541408198e-09, "loss": 0.839, "step": 29826 }, { "epoch": 0.18, "learning_rate": 2.8371078095990047e-09, "loss": 0.8462, "step": 29827 }, { "epoch": 0.18, "learning_rate": 2.8114919148924948e-09, "loss": 0.8191, "step": 29828 }, { "epoch": 0.18, "learning_rate": 2.7859921703154993e-09, "loss": 0.8216, "step": 29829 }, { "epoch": 0.18, "learning_rate": 2.760608576167778e-09, "loss": 0.8168, "step": 29830 }, { "epoch": 0.18, "learning_rate": 2.7353411327413204e-09, "loss": 0.8509, "step": 29831 }, { "epoch": 0.18, "learning_rate": 2.7101898403314453e-09, "loss": 0.8409, "step": 29832 }, { "epoch": 0.18, "learning_rate": 2.6851546992301416e-09, "loss": 0.8411, "step": 29833 }, { "epoch": 0.18, "learning_rate": 2.660235709727177e-09, "loss": 0.8517, "step": 29834 }, { "epoch": 0.18, "learning_rate": 2.6354328721134304e-09, "loss": 0.8678, "step": 29835 }, { "epoch": 0.18, "learning_rate": 2.6107461866764496e-09, "loss": 0.8205, "step": 29836 }, { "epoch": 0.18, "learning_rate": 2.5861756537026715e-09, "loss": 0.8719, "step": 29837 }, { "epoch": 0.18, "learning_rate": 2.5617212734785346e-09, "loss": 0.8807, "step": 29838 }, { "epoch": 0.18, "learning_rate": 2.537383046287145e-09, "loss": 0.8585, "step": 29839 }, { "epoch": 0.18, "learning_rate": 2.5131609724116103e-09, "loss": 0.8317, "step": 29840 }, { "epoch": 0.18, "learning_rate": 2.4890550521339264e-09, "loss": 0.8737, "step": 29841 }, { "epoch": 0.18, "learning_rate": 2.4650652857338697e-09, "loss": 0.8538, "step": 29842 }, { "epoch": 0.18, "learning_rate": 2.4411916734901063e-09, "loss": 0.8374, "step": 29843 }, { "epoch": 0.18, "learning_rate": 2.417434215679082e-09, "loss": 0.8257, "step": 29844 }, { "epoch": 0.18, "learning_rate": 2.393792912578352e-09, "loss": 0.8297, "step": 29845 }, { "epoch": 0.18, "learning_rate": 2.3702677644610316e-09, "loss": 0.8505, "step": 29846 }, { "epoch": 0.18, "learning_rate": 2.346858771602456e-09, "loss": 0.8557, "step": 29847 }, { "epoch": 0.18, "learning_rate": 2.3235659342724093e-09, "loss": 0.8669, "step": 29848 }, { "epoch": 0.18, "learning_rate": 2.3003892527440066e-09, "loss": 0.8452, "step": 29849 }, { "epoch": 0.18, "learning_rate": 2.277328727283701e-09, "loss": 0.8642, "step": 29850 }, { "epoch": 0.18, "learning_rate": 2.2543843581612766e-09, "loss": 0.8191, "step": 29851 }, { "epoch": 0.18, "learning_rate": 2.2315561456420776e-09, "loss": 0.8323, "step": 29852 }, { "epoch": 0.18, "learning_rate": 2.208844089992557e-09, "loss": 0.8618, "step": 29853 }, { "epoch": 0.18, "learning_rate": 2.1862481914769475e-09, "loss": 0.8418, "step": 29854 }, { "epoch": 0.18, "learning_rate": 2.163768450356152e-09, "loss": 0.8534, "step": 29855 }, { "epoch": 0.18, "learning_rate": 2.1414048668910723e-09, "loss": 0.8264, "step": 29856 }, { "epoch": 0.18, "learning_rate": 2.119157441343722e-09, "loss": 0.8402, "step": 29857 }, { "epoch": 0.18, "learning_rate": 2.0970261739716724e-09, "loss": 0.8531, "step": 29858 }, { "epoch": 0.18, "learning_rate": 2.075011065031385e-09, "loss": 0.8733, "step": 29859 }, { "epoch": 0.18, "learning_rate": 2.053112114779321e-09, "loss": 0.8641, "step": 29860 }, { "epoch": 0.18, "learning_rate": 2.0313293234686116e-09, "loss": 0.8129, "step": 29861 }, { "epoch": 0.18, "learning_rate": 2.009662691354608e-09, "loss": 0.8235, "step": 29862 }, { "epoch": 0.18, "learning_rate": 1.9881122186871106e-09, "loss": 0.8564, "step": 29863 }, { "epoch": 0.18, "learning_rate": 1.9666779057170293e-09, "loss": 0.8143, "step": 29864 }, { "epoch": 0.18, "learning_rate": 1.945359752694165e-09, "loss": 0.8377, "step": 29865 }, { "epoch": 0.18, "learning_rate": 1.9241577598649864e-09, "loss": 0.8613, "step": 29866 }, { "epoch": 0.18, "learning_rate": 1.9030719274770737e-09, "loss": 0.887, "step": 29867 }, { "epoch": 0.18, "learning_rate": 1.882102255774676e-09, "loss": 0.8484, "step": 29868 }, { "epoch": 0.18, "learning_rate": 1.861248745000932e-09, "loss": 0.8527, "step": 29869 }, { "epoch": 0.18, "learning_rate": 1.8405113953989806e-09, "loss": 0.8624, "step": 29870 }, { "epoch": 0.18, "learning_rate": 1.8198902072097402e-09, "loss": 0.8613, "step": 29871 }, { "epoch": 0.18, "learning_rate": 1.7993851806719086e-09, "loss": 0.8452, "step": 29872 }, { "epoch": 0.18, "learning_rate": 1.778996316024184e-09, "loss": 0.8121, "step": 29873 }, { "epoch": 0.18, "learning_rate": 1.758723613504154e-09, "loss": 0.8619, "step": 29874 }, { "epoch": 0.18, "learning_rate": 1.738567073347186e-09, "loss": 0.8833, "step": 29875 }, { "epoch": 0.18, "learning_rate": 1.7185266957864266e-09, "loss": 0.863, "step": 29876 }, { "epoch": 0.18, "learning_rate": 1.698602481056133e-09, "loss": 0.87, "step": 29877 }, { "epoch": 0.18, "learning_rate": 1.6787944293861214e-09, "loss": 0.8554, "step": 29878 }, { "epoch": 0.18, "learning_rate": 1.6591025410073181e-09, "loss": 0.8949, "step": 29879 }, { "epoch": 0.18, "learning_rate": 1.6395268161495393e-09, "loss": 0.8002, "step": 29880 }, { "epoch": 0.18, "learning_rate": 1.6200672550381602e-09, "loss": 0.8328, "step": 29881 }, { "epoch": 0.18, "learning_rate": 1.6007238579007768e-09, "loss": 0.8246, "step": 29882 }, { "epoch": 0.18, "learning_rate": 1.581496624962764e-09, "loss": 0.914, "step": 29883 }, { "epoch": 0.18, "learning_rate": 1.5623855564450564e-09, "loss": 0.859, "step": 29884 }, { "epoch": 0.18, "learning_rate": 1.543390652571919e-09, "loss": 0.8456, "step": 29885 }, { "epoch": 0.18, "learning_rate": 1.5245119135631758e-09, "loss": 0.835, "step": 29886 }, { "epoch": 0.18, "learning_rate": 1.5057493396386514e-09, "loss": 0.8738, "step": 29887 }, { "epoch": 0.18, "learning_rate": 1.4871029310148387e-09, "loss": 0.8487, "step": 29888 }, { "epoch": 0.18, "learning_rate": 1.4685726879104522e-09, "loss": 0.8316, "step": 29889 }, { "epoch": 0.18, "learning_rate": 1.4501586105386545e-09, "loss": 0.8386, "step": 29890 }, { "epoch": 0.18, "learning_rate": 1.431860699115939e-09, "loss": 0.8941, "step": 29891 }, { "epoch": 0.18, "learning_rate": 1.4136789538521378e-09, "loss": 0.8408, "step": 29892 }, { "epoch": 0.18, "learning_rate": 1.395613374960414e-09, "loss": 0.8414, "step": 29893 }, { "epoch": 0.18, "learning_rate": 1.3776639626505994e-09, "loss": 0.8015, "step": 29894 }, { "epoch": 0.18, "learning_rate": 1.359830717130306e-09, "loss": 0.8696, "step": 29895 }, { "epoch": 0.18, "learning_rate": 1.3421136386071454e-09, "loss": 0.8423, "step": 29896 }, { "epoch": 0.18, "learning_rate": 1.3245127272865088e-09, "loss": 0.8297, "step": 29897 }, { "epoch": 0.18, "learning_rate": 1.3070279833737875e-09, "loss": 0.8439, "step": 29898 }, { "epoch": 0.18, "learning_rate": 1.2896594070721524e-09, "loss": 0.8563, "step": 29899 }, { "epoch": 0.18, "learning_rate": 1.2724069985825539e-09, "loss": 0.8366, "step": 29900 }, { "epoch": 0.18, "learning_rate": 1.2552707581048318e-09, "loss": 0.8331, "step": 29901 }, { "epoch": 0.18, "learning_rate": 1.2382506858399368e-09, "loss": 0.8244, "step": 29902 }, { "epoch": 0.18, "learning_rate": 1.2213467819843782e-09, "loss": 0.8476, "step": 29903 }, { "epoch": 0.18, "learning_rate": 1.2045590467357759e-09, "loss": 0.8088, "step": 29904 }, { "epoch": 0.18, "learning_rate": 1.1878874802873086e-09, "loss": 0.856, "step": 29905 }, { "epoch": 0.18, "learning_rate": 1.1713320828332653e-09, "loss": 0.8235, "step": 29906 }, { "epoch": 0.18, "learning_rate": 1.1548928545668247e-09, "loss": 0.8257, "step": 29907 }, { "epoch": 0.18, "learning_rate": 1.1385697956789454e-09, "loss": 0.8783, "step": 29908 }, { "epoch": 0.18, "learning_rate": 1.1223629063583652e-09, "loss": 0.8431, "step": 29909 }, { "epoch": 0.18, "learning_rate": 1.1062721867949322e-09, "loss": 0.8542, "step": 29910 }, { "epoch": 0.18, "learning_rate": 1.0902976371729435e-09, "loss": 0.8399, "step": 29911 }, { "epoch": 0.18, "learning_rate": 1.074439257681137e-09, "loss": 0.8456, "step": 29912 }, { "epoch": 0.18, "learning_rate": 1.058697048501589e-09, "loss": 0.802, "step": 29913 }, { "epoch": 0.18, "learning_rate": 1.0430710098174867e-09, "loss": 0.8354, "step": 29914 }, { "epoch": 0.18, "learning_rate": 1.0275611418109066e-09, "loss": 0.8684, "step": 29915 }, { "epoch": 0.18, "learning_rate": 1.0121674446617048e-09, "loss": 0.8171, "step": 29916 }, { "epoch": 0.18, "learning_rate": 9.96889918548627e-10, "loss": 0.8344, "step": 29917 }, { "epoch": 0.18, "learning_rate": 9.81728563649309e-10, "loss": 0.8935, "step": 29918 }, { "epoch": 0.18, "learning_rate": 9.666833801413867e-10, "loss": 0.873, "step": 29919 }, { "epoch": 0.18, "learning_rate": 9.517543681969443e-10, "loss": 0.8564, "step": 29920 }, { "epoch": 0.18, "learning_rate": 9.369415279913973e-10, "loss": 0.8398, "step": 29921 }, { "epoch": 0.18, "learning_rate": 9.222448596957201e-10, "loss": 0.8557, "step": 29922 }, { "epoch": 0.18, "learning_rate": 9.076643634819971e-10, "loss": 0.8694, "step": 29923 }, { "epoch": 0.18, "learning_rate": 8.932000395178719e-10, "loss": 0.8351, "step": 29924 }, { "epoch": 0.18, "learning_rate": 8.788518879720987e-10, "loss": 0.8478, "step": 29925 }, { "epoch": 0.18, "learning_rate": 8.646199090123209e-10, "loss": 0.8612, "step": 29926 }, { "epoch": 0.18, "learning_rate": 8.505041028028515e-10, "loss": 0.8841, "step": 29927 }, { "epoch": 0.18, "learning_rate": 8.365044695080037e-10, "loss": 0.8452, "step": 29928 }, { "epoch": 0.18, "learning_rate": 8.226210092909803e-10, "loss": 0.8803, "step": 29929 }, { "epoch": 0.18, "learning_rate": 8.088537223116533e-10, "loss": 0.8548, "step": 29930 }, { "epoch": 0.18, "learning_rate": 7.952026087321152e-10, "loss": 0.8617, "step": 29931 }, { "epoch": 0.18, "learning_rate": 7.816676687089076e-10, "loss": 0.8484, "step": 29932 }, { "epoch": 0.18, "learning_rate": 7.682489023996819e-10, "loss": 0.8494, "step": 29933 }, { "epoch": 0.18, "learning_rate": 7.5494630996098e-10, "loss": 0.8404, "step": 29934 }, { "epoch": 0.18, "learning_rate": 7.417598915471225e-10, "loss": 0.868, "step": 29935 }, { "epoch": 0.18, "learning_rate": 7.286896473113203e-10, "loss": 0.85, "step": 29936 }, { "epoch": 0.18, "learning_rate": 7.157355774056741e-10, "loss": 0.879, "step": 29937 }, { "epoch": 0.18, "learning_rate": 7.028976819800637e-10, "loss": 0.8111, "step": 29938 }, { "epoch": 0.18, "learning_rate": 6.901759611843695e-10, "loss": 0.8421, "step": 29939 }, { "epoch": 0.18, "learning_rate": 6.77570415166251e-10, "loss": 0.8432, "step": 29940 }, { "epoch": 0.18, "learning_rate": 6.650810440711475e-10, "loss": 0.8911, "step": 29941 }, { "epoch": 0.18, "learning_rate": 6.527078480456084e-10, "loss": 0.8444, "step": 29942 }, { "epoch": 0.18, "learning_rate": 6.404508272328524e-10, "loss": 0.8481, "step": 29943 }, { "epoch": 0.18, "learning_rate": 6.283099817749882e-10, "loss": 0.8445, "step": 29944 }, { "epoch": 0.18, "learning_rate": 6.162853118130141e-10, "loss": 0.8141, "step": 29945 }, { "epoch": 0.18, "learning_rate": 6.043768174868181e-10, "loss": 0.8405, "step": 29946 }, { "epoch": 0.18, "learning_rate": 5.925844989351781e-10, "loss": 0.8574, "step": 29947 }, { "epoch": 0.18, "learning_rate": 5.809083562946516e-10, "loss": 0.8176, "step": 29948 }, { "epoch": 0.18, "learning_rate": 5.693483897006857e-10, "loss": 0.8419, "step": 29949 }, { "epoch": 0.18, "learning_rate": 5.579045992876175e-10, "loss": 0.8137, "step": 29950 }, { "epoch": 0.18, "learning_rate": 5.465769851886738e-10, "loss": 0.8461, "step": 29951 }, { "epoch": 0.18, "learning_rate": 5.353655475359709e-10, "loss": 0.8379, "step": 29952 }, { "epoch": 0.18, "learning_rate": 5.242702864594051e-10, "loss": 0.8722, "step": 29953 }, { "epoch": 0.18, "learning_rate": 5.132912020877623e-10, "loss": 0.8706, "step": 29954 }, { "epoch": 0.18, "learning_rate": 5.024282945476077e-10, "loss": 0.8614, "step": 29955 }, { "epoch": 0.18, "learning_rate": 4.916815639677274e-10, "loss": 0.8416, "step": 29956 }, { "epoch": 0.18, "learning_rate": 4.810510104702459e-10, "loss": 0.8062, "step": 29957 }, { "epoch": 0.18, "learning_rate": 4.705366341795081e-10, "loss": 0.8685, "step": 29958 }, { "epoch": 0.18, "learning_rate": 4.6013843521874877e-10, "loss": 0.9118, "step": 29959 }, { "epoch": 0.18, "learning_rate": 4.49856413707872e-10, "loss": 0.8216, "step": 29960 }, { "epoch": 0.18, "learning_rate": 4.396905697656717e-10, "loss": 0.829, "step": 29961 }, { "epoch": 0.18, "learning_rate": 4.296409035120519e-10, "loss": 0.8182, "step": 29962 }, { "epoch": 0.18, "learning_rate": 4.197074150624758e-10, "loss": 0.8676, "step": 29963 }, { "epoch": 0.18, "learning_rate": 4.098901045324066e-10, "loss": 0.814, "step": 29964 }, { "epoch": 0.18, "learning_rate": 4.001889720361973e-10, "loss": 0.9031, "step": 29965 }, { "epoch": 0.18, "learning_rate": 3.9060401768598043e-10, "loss": 0.8829, "step": 29966 }, { "epoch": 0.18, "learning_rate": 3.8113524159388846e-10, "loss": 0.8321, "step": 29967 }, { "epoch": 0.18, "learning_rate": 3.717826438709438e-10, "loss": 0.8328, "step": 29968 }, { "epoch": 0.18, "learning_rate": 3.6254622462261745e-10, "loss": 0.835, "step": 29969 }, { "epoch": 0.18, "learning_rate": 3.534259839599319e-10, "loss": 0.8189, "step": 29970 }, { "epoch": 0.18, "learning_rate": 3.444219219861378e-10, "loss": 0.8289, "step": 29971 }, { "epoch": 0.18, "learning_rate": 3.3553403880670634e-10, "loss": 0.8454, "step": 29972 }, { "epoch": 0.18, "learning_rate": 3.2676233452488827e-10, "loss": 0.821, "step": 29973 }, { "epoch": 0.18, "learning_rate": 3.181068092428241e-10, "loss": 0.8514, "step": 29974 }, { "epoch": 0.18, "learning_rate": 3.0956746306043396e-10, "loss": 0.8712, "step": 29975 }, { "epoch": 0.18, "learning_rate": 3.0114429607763783e-10, "loss": 0.8102, "step": 29976 }, { "epoch": 0.18, "learning_rate": 2.9283730839324566e-10, "loss": 0.835, "step": 29977 }, { "epoch": 0.18, "learning_rate": 2.846465001016263e-10, "loss": 0.8851, "step": 29978 }, { "epoch": 0.18, "learning_rate": 2.76571871298259e-10, "loss": 0.8846, "step": 29979 }, { "epoch": 0.18, "learning_rate": 2.6861342207862294e-10, "loss": 0.8055, "step": 29980 }, { "epoch": 0.18, "learning_rate": 2.607711525337564e-10, "loss": 0.8361, "step": 29981 }, { "epoch": 0.18, "learning_rate": 2.530450627558079e-10, "loss": 0.7941, "step": 29982 }, { "epoch": 0.18, "learning_rate": 2.454351528335952e-10, "loss": 0.8499, "step": 29983 }, { "epoch": 0.18, "learning_rate": 2.3794142285482604e-10, "loss": 0.8559, "step": 29984 }, { "epoch": 0.18, "learning_rate": 2.3056387290942839e-10, "loss": 0.8114, "step": 29985 }, { "epoch": 0.18, "learning_rate": 2.233025030795588e-10, "loss": 0.8086, "step": 29986 }, { "epoch": 0.18, "learning_rate": 2.1615731345181468e-10, "loss": 0.8323, "step": 29987 }, { "epoch": 0.18, "learning_rate": 2.0912830410946273e-10, "loss": 0.8347, "step": 29988 }, { "epoch": 0.18, "learning_rate": 2.02215475132439e-10, "loss": 0.862, "step": 29989 }, { "epoch": 0.18, "learning_rate": 1.9541882660290001e-10, "loss": 0.8566, "step": 29990 }, { "epoch": 0.18, "learning_rate": 1.8873835859856137e-10, "loss": 0.9119, "step": 29991 }, { "epoch": 0.18, "learning_rate": 1.821740711971387e-10, "loss": 0.8513, "step": 29992 }, { "epoch": 0.18, "learning_rate": 1.757259644763476e-10, "loss": 0.8217, "step": 29993 }, { "epoch": 0.18, "learning_rate": 1.6939403850835257e-10, "loss": 0.8366, "step": 29994 }, { "epoch": 0.18, "learning_rate": 1.6317829336975898e-10, "loss": 0.856, "step": 29995 }, { "epoch": 0.18, "learning_rate": 1.5707872913051093e-10, "loss": 0.8307, "step": 29996 }, { "epoch": 0.18, "learning_rate": 1.5109534586166264e-10, "loss": 0.8518, "step": 29997 }, { "epoch": 0.18, "learning_rate": 1.452281436342684e-10, "loss": 0.8124, "step": 29998 }, { "epoch": 0.18, "learning_rate": 1.3947712251494162e-10, "loss": 0.8626, "step": 29999 }, { "epoch": 0.18, "learning_rate": 1.3384228257140586e-10, "loss": 0.869, "step": 30000 }, { "epoch": 0.18, "learning_rate": 1.283236238691643e-10, "loss": 0.8314, "step": 30001 }, { "epoch": 0.18, "learning_rate": 1.2292114647260988e-10, "loss": 0.8329, "step": 30002 }, { "epoch": 0.18, "learning_rate": 1.1763485044280486e-10, "loss": 0.8586, "step": 30003 }, { "epoch": 0.18, "learning_rate": 1.1246473584303197e-10, "loss": 0.8344, "step": 30004 }, { "epoch": 0.18, "learning_rate": 1.07410802732133e-10, "loss": 0.851, "step": 30005 }, { "epoch": 0.18, "learning_rate": 1.0247305116894979e-10, "loss": 0.8376, "step": 30006 }, { "epoch": 0.18, "learning_rate": 9.765148121232415e-11, "loss": 0.8142, "step": 30007 }, { "epoch": 0.18, "learning_rate": 9.294609291665702e-11, "loss": 0.8557, "step": 30008 }, { "epoch": 0.18, "learning_rate": 8.835688633634931e-11, "loss": 0.8357, "step": 30009 }, { "epoch": 0.18, "learning_rate": 8.388386152691219e-11, "loss": 0.8249, "step": 30010 }, { "epoch": 0.18, "learning_rate": 7.952701853830569e-11, "loss": 0.8305, "step": 30011 }, { "epoch": 0.18, "learning_rate": 7.528635742160007e-11, "loss": 0.8274, "step": 30012 }, { "epoch": 0.18, "learning_rate": 7.116187822564513e-11, "loss": 0.8208, "step": 30013 }, { "epoch": 0.18, "learning_rate": 6.715358100040093e-11, "loss": 0.854, "step": 30014 }, { "epoch": 0.18, "learning_rate": 6.326146578916614e-11, "loss": 0.867, "step": 30015 }, { "epoch": 0.18, "learning_rate": 5.94855326407906e-11, "loss": 0.8472, "step": 30016 }, { "epoch": 0.18, "learning_rate": 5.5825781596352544e-11, "loss": 0.8593, "step": 30017 }, { "epoch": 0.18, "learning_rate": 5.228221269915068e-11, "loss": 0.8657, "step": 30018 }, { "epoch": 0.18, "learning_rate": 4.8854825990263256e-11, "loss": 0.8631, "step": 30019 }, { "epoch": 0.18, "learning_rate": 4.554362151076852e-11, "loss": 0.8651, "step": 30020 }, { "epoch": 0.18, "learning_rate": 4.234859929730384e-11, "loss": 0.8696, "step": 30021 }, { "epoch": 0.18, "learning_rate": 3.9269759388727015e-11, "loss": 0.8783, "step": 30022 }, { "epoch": 0.18, "learning_rate": 3.6307101818344736e-11, "loss": 0.8995, "step": 30023 }, { "epoch": 0.18, "learning_rate": 3.346062662390459e-11, "loss": 0.8472, "step": 30024 }, { "epoch": 0.18, "learning_rate": 3.07303338353826e-11, "loss": 0.8398, "step": 30025 }, { "epoch": 0.18, "learning_rate": 2.811622348719567e-11, "loss": 0.8051, "step": 30026 }, { "epoch": 0.18, "learning_rate": 2.5618295607099387e-11, "loss": 0.879, "step": 30027 }, { "epoch": 0.18, "learning_rate": 2.323655022617999e-11, "loss": 0.8339, "step": 30028 }, { "epoch": 0.18, "learning_rate": 2.097098737108283e-11, "loss": 0.8849, "step": 30029 }, { "epoch": 0.18, "learning_rate": 1.8821607068453263e-11, "loss": 0.8159, "step": 30030 }, { "epoch": 0.18, "learning_rate": 1.6788409342716195e-11, "loss": 0.8219, "step": 30031 }, { "epoch": 0.18, "learning_rate": 1.487139421829653e-11, "loss": 0.8719, "step": 30032 }, { "epoch": 0.18, "learning_rate": 1.307056171628851e-11, "loss": 0.8485, "step": 30033 }, { "epoch": 0.18, "learning_rate": 1.1385911860006815e-11, "loss": 0.8639, "step": 30034 }, { "epoch": 0.18, "learning_rate": 9.817444666104791e-12, "loss": 0.8646, "step": 30035 }, { "epoch": 0.18, "learning_rate": 8.365160154566454e-12, "loss": 0.8185, "step": 30036 }, { "epoch": 0.18, "learning_rate": 7.029058342045148e-12, "loss": 0.8269, "step": 30037 }, { "epoch": 0.18, "learning_rate": 5.809139244083994e-12, "loss": 0.8305, "step": 30038 }, { "epoch": 0.18, "learning_rate": 4.705402874005671e-12, "loss": 0.8381, "step": 30039 }, { "epoch": 0.18, "learning_rate": 3.717849245132854e-12, "loss": 0.7974, "step": 30040 }, { "epoch": 0.18, "learning_rate": 2.8464783696779964e-12, "loss": 0.8916, "step": 30041 }, { "epoch": 0.18, "learning_rate": 2.091290257633105e-12, "loss": 0.8173, "step": 30042 }, { "epoch": 0.18, "learning_rate": 1.4522849167697417e-12, "loss": 0.8811, "step": 30043 }, { "epoch": 0.18, "learning_rate": 9.294623548594673e-13, "loss": 0.8222, "step": 30044 }, { "epoch": 0.18, "learning_rate": 5.228225774533969e-13, "loss": 0.8872, "step": 30045 }, { "epoch": 0.18, "learning_rate": 2.323655912128686e-13, "loss": 0.8544, "step": 30046 }, { "epoch": 0.18, "learning_rate": 5.809139835832866e-14, "loss": 0.8647, "step": 30047 }, { "epoch": 0.18, "learning_rate": 0.0, "loss": 0.8884, "step": 30048 }, { "epoch": 0.18, "step": 30048, "total_flos": 1.8108019980896843e+25, "train_loss": 0.1562760086954275, "train_runtime": 34539.9979, "train_samples_per_second": 890.827, "train_steps_per_second": 0.87 } ], "logging_steps": 1.0, "max_steps": 30048, "num_input_tokens_seen": 0, "num_train_epochs": 9223372036854775807, "save_steps": 250, "total_flos": 1.8108019980896843e+25, "train_batch_size": 1, "trial_name": null, "trial_params": null }