{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "global_step": 19684, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 3.3840947546531305e-08, "loss": 1.448, "step": 1 }, { "epoch": 0.0, "learning_rate": 6.768189509306261e-08, "loss": 1.5632, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.0152284263959391e-07, "loss": 1.5219, "step": 3 }, { "epoch": 0.0, "learning_rate": 1.3536379018612522e-07, "loss": 1.5868, "step": 4 }, { "epoch": 0.0, "learning_rate": 1.6920473773265654e-07, "loss": 1.6326, "step": 5 }, { "epoch": 0.0, "learning_rate": 2.0304568527918783e-07, "loss": 1.5404, "step": 6 }, { "epoch": 0.0, "learning_rate": 2.3688663282571915e-07, "loss": 1.634, "step": 7 }, { "epoch": 0.0, "learning_rate": 2.7072758037225044e-07, "loss": 1.5579, "step": 8 }, { "epoch": 0.0, "learning_rate": 3.0456852791878176e-07, "loss": 1.5142, "step": 9 }, { "epoch": 0.0, "learning_rate": 3.384094754653131e-07, "loss": 1.4871, "step": 10 }, { "epoch": 0.0, "learning_rate": 3.7225042301184434e-07, "loss": 1.623, "step": 11 }, { "epoch": 0.0, "learning_rate": 4.0609137055837566e-07, "loss": 1.4572, "step": 12 }, { "epoch": 0.0, "learning_rate": 4.39932318104907e-07, "loss": 1.6021, "step": 13 }, { "epoch": 0.0, "learning_rate": 4.737732656514383e-07, "loss": 1.5669, "step": 14 }, { "epoch": 0.0, "learning_rate": 5.076142131979696e-07, "loss": 1.519, "step": 15 }, { "epoch": 0.0, "learning_rate": 5.414551607445009e-07, "loss": 1.4891, "step": 16 }, { "epoch": 0.0, "learning_rate": 5.752961082910321e-07, "loss": 1.5603, "step": 17 }, { "epoch": 0.0, "learning_rate": 6.091370558375635e-07, "loss": 1.481, "step": 18 }, { "epoch": 0.0, "learning_rate": 6.429780033840948e-07, "loss": 1.4006, "step": 19 }, { "epoch": 0.0, "learning_rate": 6.768189509306262e-07, "loss": 1.383, "step": 20 }, { "epoch": 0.0, "learning_rate": 7.106598984771574e-07, "loss": 1.4547, "step": 21 }, { "epoch": 0.0, "learning_rate": 7.445008460236887e-07, "loss": 1.4856, "step": 22 }, { "epoch": 0.0, "learning_rate": 7.7834179357022e-07, "loss": 1.4553, "step": 23 }, { "epoch": 0.0, "learning_rate": 8.121827411167513e-07, "loss": 1.4917, "step": 24 }, { "epoch": 0.0, "learning_rate": 8.460236886632826e-07, "loss": 1.4277, "step": 25 }, { "epoch": 0.0, "learning_rate": 8.79864636209814e-07, "loss": 1.4171, "step": 26 }, { "epoch": 0.0, "learning_rate": 9.137055837563452e-07, "loss": 1.4017, "step": 27 }, { "epoch": 0.0, "learning_rate": 9.475465313028766e-07, "loss": 1.4778, "step": 28 }, { "epoch": 0.0, "learning_rate": 9.813874788494077e-07, "loss": 1.3636, "step": 29 }, { "epoch": 0.0, "learning_rate": 1.0152284263959392e-06, "loss": 1.4423, "step": 30 }, { "epoch": 0.0, "learning_rate": 1.0490693739424705e-06, "loss": 1.339, "step": 31 }, { "epoch": 0.0, "learning_rate": 1.0829103214890018e-06, "loss": 1.433, "step": 32 }, { "epoch": 0.0, "learning_rate": 1.116751269035533e-06, "loss": 1.324, "step": 33 }, { "epoch": 0.0, "learning_rate": 1.1505922165820643e-06, "loss": 1.3351, "step": 34 }, { "epoch": 0.0, "learning_rate": 1.1844331641285958e-06, "loss": 1.3518, "step": 35 }, { "epoch": 0.0, "learning_rate": 1.218274111675127e-06, "loss": 1.2889, "step": 36 }, { "epoch": 0.0, "learning_rate": 1.2521150592216583e-06, "loss": 1.2234, "step": 37 }, { "epoch": 0.0, "learning_rate": 1.2859560067681896e-06, "loss": 1.2856, "step": 38 }, { "epoch": 0.0, "learning_rate": 1.319796954314721e-06, "loss": 1.2034, "step": 39 }, { "epoch": 0.0, "learning_rate": 1.3536379018612523e-06, "loss": 1.2097, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.3874788494077834e-06, "loss": 1.3267, "step": 41 }, { "epoch": 0.0, "learning_rate": 1.4213197969543148e-06, "loss": 1.3889, "step": 42 }, { "epoch": 0.0, "learning_rate": 1.455160744500846e-06, "loss": 1.2998, "step": 43 }, { "epoch": 0.0, "learning_rate": 1.4890016920473774e-06, "loss": 1.3451, "step": 44 }, { "epoch": 0.0, "learning_rate": 1.5228426395939088e-06, "loss": 1.2987, "step": 45 }, { "epoch": 0.0, "learning_rate": 1.55668358714044e-06, "loss": 1.2329, "step": 46 }, { "epoch": 0.0, "learning_rate": 1.5905245346869714e-06, "loss": 1.2994, "step": 47 }, { "epoch": 0.0, "learning_rate": 1.6243654822335026e-06, "loss": 1.3464, "step": 48 }, { "epoch": 0.0, "learning_rate": 1.6582064297800341e-06, "loss": 1.2092, "step": 49 }, { "epoch": 0.0, "learning_rate": 1.6920473773265652e-06, "loss": 1.2929, "step": 50 }, { "epoch": 0.0, "learning_rate": 1.7258883248730964e-06, "loss": 1.155, "step": 51 }, { "epoch": 0.0, "learning_rate": 1.759729272419628e-06, "loss": 1.1709, "step": 52 }, { "epoch": 0.0, "learning_rate": 1.7935702199661592e-06, "loss": 1.2594, "step": 53 }, { "epoch": 0.0, "learning_rate": 1.8274111675126904e-06, "loss": 1.321, "step": 54 }, { "epoch": 0.0, "learning_rate": 1.861252115059222e-06, "loss": 1.2649, "step": 55 }, { "epoch": 0.0, "learning_rate": 1.8950930626057532e-06, "loss": 1.2646, "step": 56 }, { "epoch": 0.0, "learning_rate": 1.9289340101522844e-06, "loss": 1.3342, "step": 57 }, { "epoch": 0.0, "learning_rate": 1.9627749576988155e-06, "loss": 1.1941, "step": 58 }, { "epoch": 0.0, "learning_rate": 1.996615905245347e-06, "loss": 1.2374, "step": 59 }, { "epoch": 0.0, "learning_rate": 2.0304568527918785e-06, "loss": 1.2684, "step": 60 }, { "epoch": 0.0, "learning_rate": 2.0642978003384095e-06, "loss": 1.2772, "step": 61 }, { "epoch": 0.0, "learning_rate": 2.098138747884941e-06, "loss": 1.2433, "step": 62 }, { "epoch": 0.0, "learning_rate": 2.1319796954314725e-06, "loss": 1.2656, "step": 63 }, { "epoch": 0.0, "learning_rate": 2.1658206429780035e-06, "loss": 1.3159, "step": 64 }, { "epoch": 0.0, "learning_rate": 2.199661590524535e-06, "loss": 1.3613, "step": 65 }, { "epoch": 0.0, "learning_rate": 2.233502538071066e-06, "loss": 1.2695, "step": 66 }, { "epoch": 0.0, "learning_rate": 2.2673434856175975e-06, "loss": 1.196, "step": 67 }, { "epoch": 0.0, "learning_rate": 2.3011844331641286e-06, "loss": 1.3472, "step": 68 }, { "epoch": 0.0, "learning_rate": 2.33502538071066e-06, "loss": 1.1666, "step": 69 }, { "epoch": 0.0, "learning_rate": 2.3688663282571915e-06, "loss": 1.1459, "step": 70 }, { "epoch": 0.0, "learning_rate": 2.4027072758037226e-06, "loss": 1.2529, "step": 71 }, { "epoch": 0.0, "learning_rate": 2.436548223350254e-06, "loss": 1.1793, "step": 72 }, { "epoch": 0.0, "learning_rate": 2.470389170896785e-06, "loss": 1.2458, "step": 73 }, { "epoch": 0.0, "learning_rate": 2.5042301184433166e-06, "loss": 1.1085, "step": 74 }, { "epoch": 0.0, "learning_rate": 2.5380710659898476e-06, "loss": 1.2645, "step": 75 }, { "epoch": 0.0, "learning_rate": 2.571912013536379e-06, "loss": 1.1565, "step": 76 }, { "epoch": 0.0, "learning_rate": 2.6057529610829106e-06, "loss": 1.2052, "step": 77 }, { "epoch": 0.0, "learning_rate": 2.639593908629442e-06, "loss": 1.1194, "step": 78 }, { "epoch": 0.0, "learning_rate": 2.673434856175973e-06, "loss": 1.0579, "step": 79 }, { "epoch": 0.0, "learning_rate": 2.7072758037225046e-06, "loss": 1.2411, "step": 80 }, { "epoch": 0.0, "learning_rate": 2.7411167512690357e-06, "loss": 1.1326, "step": 81 }, { "epoch": 0.0, "learning_rate": 2.7749576988155667e-06, "loss": 1.3167, "step": 82 }, { "epoch": 0.0, "learning_rate": 2.808798646362098e-06, "loss": 1.2577, "step": 83 }, { "epoch": 0.0, "learning_rate": 2.8426395939086297e-06, "loss": 1.1776, "step": 84 }, { "epoch": 0.0, "learning_rate": 2.876480541455161e-06, "loss": 1.2847, "step": 85 }, { "epoch": 0.0, "learning_rate": 2.910321489001692e-06, "loss": 1.1949, "step": 86 }, { "epoch": 0.0, "learning_rate": 2.9441624365482237e-06, "loss": 1.1811, "step": 87 }, { "epoch": 0.0, "learning_rate": 2.9780033840947547e-06, "loss": 1.202, "step": 88 }, { "epoch": 0.0, "learning_rate": 3.0118443316412858e-06, "loss": 1.1462, "step": 89 }, { "epoch": 0.0, "learning_rate": 3.0456852791878177e-06, "loss": 1.2545, "step": 90 }, { "epoch": 0.0, "learning_rate": 3.079526226734349e-06, "loss": 1.2489, "step": 91 }, { "epoch": 0.0, "learning_rate": 3.11336717428088e-06, "loss": 1.1436, "step": 92 }, { "epoch": 0.0, "learning_rate": 3.1472081218274113e-06, "loss": 1.2322, "step": 93 }, { "epoch": 0.0, "learning_rate": 3.1810490693739427e-06, "loss": 1.3174, "step": 94 }, { "epoch": 0.0, "learning_rate": 3.214890016920474e-06, "loss": 1.162, "step": 95 }, { "epoch": 0.0, "learning_rate": 3.2487309644670053e-06, "loss": 1.1918, "step": 96 }, { "epoch": 0.0, "learning_rate": 3.2825719120135367e-06, "loss": 1.2322, "step": 97 }, { "epoch": 0.0, "learning_rate": 3.3164128595600682e-06, "loss": 1.1368, "step": 98 }, { "epoch": 0.01, "learning_rate": 3.3502538071065993e-06, "loss": 1.1288, "step": 99 }, { "epoch": 0.01, "learning_rate": 3.3840947546531303e-06, "loss": 1.1736, "step": 100 }, { "epoch": 0.01, "learning_rate": 3.417935702199662e-06, "loss": 1.2086, "step": 101 }, { "epoch": 0.01, "learning_rate": 3.451776649746193e-06, "loss": 1.1625, "step": 102 }, { "epoch": 0.01, "learning_rate": 3.4856175972927248e-06, "loss": 1.2712, "step": 103 }, { "epoch": 0.01, "learning_rate": 3.519458544839256e-06, "loss": 1.1124, "step": 104 }, { "epoch": 0.01, "learning_rate": 3.5532994923857873e-06, "loss": 1.0892, "step": 105 }, { "epoch": 0.01, "learning_rate": 3.5871404399323183e-06, "loss": 1.0984, "step": 106 }, { "epoch": 0.01, "learning_rate": 3.62098138747885e-06, "loss": 1.2337, "step": 107 }, { "epoch": 0.01, "learning_rate": 3.654822335025381e-06, "loss": 1.1016, "step": 108 }, { "epoch": 0.01, "learning_rate": 3.688663282571912e-06, "loss": 1.2626, "step": 109 }, { "epoch": 0.01, "learning_rate": 3.722504230118444e-06, "loss": 1.1873, "step": 110 }, { "epoch": 0.01, "learning_rate": 3.756345177664975e-06, "loss": 1.1531, "step": 111 }, { "epoch": 0.01, "learning_rate": 3.7901861252115064e-06, "loss": 1.1553, "step": 112 }, { "epoch": 0.01, "learning_rate": 3.824027072758038e-06, "loss": 1.1294, "step": 113 }, { "epoch": 0.01, "learning_rate": 3.857868020304569e-06, "loss": 1.1758, "step": 114 }, { "epoch": 0.01, "learning_rate": 3.8917089678511e-06, "loss": 1.1425, "step": 115 }, { "epoch": 0.01, "learning_rate": 3.925549915397631e-06, "loss": 1.1645, "step": 116 }, { "epoch": 0.01, "learning_rate": 3.959390862944163e-06, "loss": 1.1888, "step": 117 }, { "epoch": 0.01, "learning_rate": 3.993231810490694e-06, "loss": 1.0904, "step": 118 }, { "epoch": 0.01, "learning_rate": 4.027072758037225e-06, "loss": 1.0497, "step": 119 }, { "epoch": 0.01, "learning_rate": 4.060913705583757e-06, "loss": 1.1932, "step": 120 }, { "epoch": 0.01, "learning_rate": 4.094754653130288e-06, "loss": 1.0938, "step": 121 }, { "epoch": 0.01, "learning_rate": 4.128595600676819e-06, "loss": 1.1362, "step": 122 }, { "epoch": 0.01, "learning_rate": 4.162436548223351e-06, "loss": 1.1666, "step": 123 }, { "epoch": 0.01, "learning_rate": 4.196277495769882e-06, "loss": 1.1931, "step": 124 }, { "epoch": 0.01, "learning_rate": 4.230118443316413e-06, "loss": 1.1375, "step": 125 }, { "epoch": 0.01, "learning_rate": 4.263959390862945e-06, "loss": 1.1312, "step": 126 }, { "epoch": 0.01, "learning_rate": 4.297800338409476e-06, "loss": 1.1582, "step": 127 }, { "epoch": 0.01, "learning_rate": 4.331641285956007e-06, "loss": 1.0984, "step": 128 }, { "epoch": 0.01, "learning_rate": 4.365482233502538e-06, "loss": 1.1387, "step": 129 }, { "epoch": 0.01, "learning_rate": 4.39932318104907e-06, "loss": 1.1115, "step": 130 }, { "epoch": 0.01, "learning_rate": 4.433164128595601e-06, "loss": 1.1107, "step": 131 }, { "epoch": 0.01, "learning_rate": 4.467005076142132e-06, "loss": 1.1469, "step": 132 }, { "epoch": 0.01, "learning_rate": 4.500846023688664e-06, "loss": 1.1217, "step": 133 }, { "epoch": 0.01, "learning_rate": 4.534686971235195e-06, "loss": 1.2272, "step": 134 }, { "epoch": 0.01, "learning_rate": 4.568527918781726e-06, "loss": 1.1963, "step": 135 }, { "epoch": 0.01, "learning_rate": 4.602368866328257e-06, "loss": 1.1188, "step": 136 }, { "epoch": 0.01, "learning_rate": 4.636209813874789e-06, "loss": 1.0613, "step": 137 }, { "epoch": 0.01, "learning_rate": 4.67005076142132e-06, "loss": 1.1885, "step": 138 }, { "epoch": 0.01, "learning_rate": 4.703891708967851e-06, "loss": 1.1816, "step": 139 }, { "epoch": 0.01, "learning_rate": 4.737732656514383e-06, "loss": 1.0891, "step": 140 }, { "epoch": 0.01, "learning_rate": 4.771573604060914e-06, "loss": 1.1077, "step": 141 }, { "epoch": 0.01, "learning_rate": 4.805414551607445e-06, "loss": 1.1543, "step": 142 }, { "epoch": 0.01, "learning_rate": 4.839255499153977e-06, "loss": 1.1784, "step": 143 }, { "epoch": 0.01, "learning_rate": 4.873096446700508e-06, "loss": 1.2574, "step": 144 }, { "epoch": 0.01, "learning_rate": 4.906937394247039e-06, "loss": 1.1379, "step": 145 }, { "epoch": 0.01, "learning_rate": 4.94077834179357e-06, "loss": 1.0639, "step": 146 }, { "epoch": 0.01, "learning_rate": 4.974619289340102e-06, "loss": 1.1797, "step": 147 }, { "epoch": 0.01, "learning_rate": 5.008460236886633e-06, "loss": 1.1703, "step": 148 }, { "epoch": 0.01, "learning_rate": 5.042301184433165e-06, "loss": 1.0023, "step": 149 }, { "epoch": 0.01, "learning_rate": 5.076142131979695e-06, "loss": 1.1532, "step": 150 }, { "epoch": 0.01, "learning_rate": 5.109983079526227e-06, "loss": 1.1557, "step": 151 }, { "epoch": 0.01, "learning_rate": 5.143824027072758e-06, "loss": 1.1382, "step": 152 }, { "epoch": 0.01, "learning_rate": 5.17766497461929e-06, "loss": 1.1136, "step": 153 }, { "epoch": 0.01, "learning_rate": 5.211505922165821e-06, "loss": 1.1219, "step": 154 }, { "epoch": 0.01, "learning_rate": 5.245346869712352e-06, "loss": 1.1024, "step": 155 }, { "epoch": 0.01, "learning_rate": 5.279187817258884e-06, "loss": 1.0938, "step": 156 }, { "epoch": 0.01, "learning_rate": 5.313028764805414e-06, "loss": 1.1216, "step": 157 }, { "epoch": 0.01, "learning_rate": 5.346869712351946e-06, "loss": 1.2169, "step": 158 }, { "epoch": 0.01, "learning_rate": 5.380710659898477e-06, "loss": 1.1199, "step": 159 }, { "epoch": 0.01, "learning_rate": 5.414551607445009e-06, "loss": 1.1981, "step": 160 }, { "epoch": 0.01, "learning_rate": 5.44839255499154e-06, "loss": 1.0723, "step": 161 }, { "epoch": 0.01, "learning_rate": 5.482233502538071e-06, "loss": 1.1152, "step": 162 }, { "epoch": 0.01, "learning_rate": 5.516074450084603e-06, "loss": 1.1538, "step": 163 }, { "epoch": 0.01, "learning_rate": 5.549915397631133e-06, "loss": 1.0684, "step": 164 }, { "epoch": 0.01, "learning_rate": 5.583756345177665e-06, "loss": 1.1995, "step": 165 }, { "epoch": 0.01, "learning_rate": 5.617597292724196e-06, "loss": 1.1022, "step": 166 }, { "epoch": 0.01, "learning_rate": 5.651438240270728e-06, "loss": 1.1426, "step": 167 }, { "epoch": 0.01, "learning_rate": 5.685279187817259e-06, "loss": 1.1001, "step": 168 }, { "epoch": 0.01, "learning_rate": 5.71912013536379e-06, "loss": 1.1096, "step": 169 }, { "epoch": 0.01, "learning_rate": 5.752961082910322e-06, "loss": 1.1584, "step": 170 }, { "epoch": 0.01, "learning_rate": 5.7868020304568525e-06, "loss": 1.1859, "step": 171 }, { "epoch": 0.01, "learning_rate": 5.820642978003384e-06, "loss": 1.1006, "step": 172 }, { "epoch": 0.01, "learning_rate": 5.854483925549916e-06, "loss": 1.1191, "step": 173 }, { "epoch": 0.01, "learning_rate": 5.888324873096447e-06, "loss": 1.1677, "step": 174 }, { "epoch": 0.01, "learning_rate": 5.922165820642979e-06, "loss": 1.2213, "step": 175 }, { "epoch": 0.01, "learning_rate": 5.9560067681895094e-06, "loss": 1.1447, "step": 176 }, { "epoch": 0.01, "learning_rate": 5.989847715736041e-06, "loss": 1.1727, "step": 177 }, { "epoch": 0.01, "learning_rate": 6.0236886632825716e-06, "loss": 1.0198, "step": 178 }, { "epoch": 0.01, "learning_rate": 6.0575296108291035e-06, "loss": 1.1862, "step": 179 }, { "epoch": 0.01, "learning_rate": 6.091370558375635e-06, "loss": 1.309, "step": 180 }, { "epoch": 0.01, "learning_rate": 6.125211505922166e-06, "loss": 1.1835, "step": 181 }, { "epoch": 0.01, "learning_rate": 6.159052453468698e-06, "loss": 1.1657, "step": 182 }, { "epoch": 0.01, "learning_rate": 6.1928934010152285e-06, "loss": 1.1188, "step": 183 }, { "epoch": 0.01, "learning_rate": 6.22673434856176e-06, "loss": 1.166, "step": 184 }, { "epoch": 0.01, "learning_rate": 6.2605752961082915e-06, "loss": 1.2305, "step": 185 }, { "epoch": 0.01, "learning_rate": 6.2944162436548225e-06, "loss": 1.0416, "step": 186 }, { "epoch": 0.01, "learning_rate": 6.328257191201354e-06, "loss": 1.0795, "step": 187 }, { "epoch": 0.01, "learning_rate": 6.3620981387478855e-06, "loss": 1.1591, "step": 188 }, { "epoch": 0.01, "learning_rate": 6.395939086294417e-06, "loss": 1.093, "step": 189 }, { "epoch": 0.01, "learning_rate": 6.429780033840948e-06, "loss": 1.1073, "step": 190 }, { "epoch": 0.01, "learning_rate": 6.4636209813874795e-06, "loss": 1.1345, "step": 191 }, { "epoch": 0.01, "learning_rate": 6.4974619289340105e-06, "loss": 1.1649, "step": 192 }, { "epoch": 0.01, "learning_rate": 6.531302876480542e-06, "loss": 1.0215, "step": 193 }, { "epoch": 0.01, "learning_rate": 6.5651438240270735e-06, "loss": 1.0617, "step": 194 }, { "epoch": 0.01, "learning_rate": 6.5989847715736045e-06, "loss": 1.1704, "step": 195 }, { "epoch": 0.01, "learning_rate": 6.6328257191201364e-06, "loss": 1.1591, "step": 196 }, { "epoch": 0.01, "learning_rate": 6.666666666666667e-06, "loss": 1.1965, "step": 197 }, { "epoch": 0.01, "learning_rate": 6.7005076142131985e-06, "loss": 1.1277, "step": 198 }, { "epoch": 0.01, "learning_rate": 6.73434856175973e-06, "loss": 1.1141, "step": 199 }, { "epoch": 0.01, "learning_rate": 6.768189509306261e-06, "loss": 1.1211, "step": 200 }, { "epoch": 0.01, "learning_rate": 6.8020304568527926e-06, "loss": 1.0842, "step": 201 }, { "epoch": 0.01, "learning_rate": 6.835871404399324e-06, "loss": 1.015, "step": 202 }, { "epoch": 0.01, "learning_rate": 6.8697123519458555e-06, "loss": 1.1884, "step": 203 }, { "epoch": 0.01, "learning_rate": 6.903553299492386e-06, "loss": 1.1191, "step": 204 }, { "epoch": 0.01, "learning_rate": 6.937394247038918e-06, "loss": 1.0928, "step": 205 }, { "epoch": 0.01, "learning_rate": 6.9712351945854495e-06, "loss": 1.045, "step": 206 }, { "epoch": 0.01, "learning_rate": 7.0050761421319806e-06, "loss": 1.1631, "step": 207 }, { "epoch": 0.01, "learning_rate": 7.038917089678512e-06, "loss": 1.1674, "step": 208 }, { "epoch": 0.01, "learning_rate": 7.072758037225043e-06, "loss": 1.1972, "step": 209 }, { "epoch": 0.01, "learning_rate": 7.106598984771575e-06, "loss": 1.1184, "step": 210 }, { "epoch": 0.01, "learning_rate": 7.140439932318105e-06, "loss": 1.0305, "step": 211 }, { "epoch": 0.01, "learning_rate": 7.174280879864637e-06, "loss": 1.0558, "step": 212 }, { "epoch": 0.01, "learning_rate": 7.208121827411169e-06, "loss": 1.0589, "step": 213 }, { "epoch": 0.01, "learning_rate": 7.2419627749577e-06, "loss": 1.1506, "step": 214 }, { "epoch": 0.01, "learning_rate": 7.275803722504231e-06, "loss": 1.1431, "step": 215 }, { "epoch": 0.01, "learning_rate": 7.309644670050762e-06, "loss": 1.1212, "step": 216 }, { "epoch": 0.01, "learning_rate": 7.343485617597294e-06, "loss": 1.0443, "step": 217 }, { "epoch": 0.01, "learning_rate": 7.377326565143824e-06, "loss": 1.0178, "step": 218 }, { "epoch": 0.01, "learning_rate": 7.411167512690356e-06, "loss": 1.0731, "step": 219 }, { "epoch": 0.01, "learning_rate": 7.445008460236888e-06, "loss": 1.2195, "step": 220 }, { "epoch": 0.01, "learning_rate": 7.478849407783419e-06, "loss": 1.2437, "step": 221 }, { "epoch": 0.01, "learning_rate": 7.51269035532995e-06, "loss": 1.1066, "step": 222 }, { "epoch": 0.01, "learning_rate": 7.546531302876481e-06, "loss": 1.0491, "step": 223 }, { "epoch": 0.01, "learning_rate": 7.580372250423013e-06, "loss": 1.1185, "step": 224 }, { "epoch": 0.01, "learning_rate": 7.614213197969543e-06, "loss": 1.0459, "step": 225 }, { "epoch": 0.01, "learning_rate": 7.648054145516076e-06, "loss": 1.269, "step": 226 }, { "epoch": 0.01, "learning_rate": 7.681895093062607e-06, "loss": 1.1145, "step": 227 }, { "epoch": 0.01, "learning_rate": 7.715736040609138e-06, "loss": 1.1193, "step": 228 }, { "epoch": 0.01, "learning_rate": 7.749576988155669e-06, "loss": 1.0215, "step": 229 }, { "epoch": 0.01, "learning_rate": 7.7834179357022e-06, "loss": 1.1992, "step": 230 }, { "epoch": 0.01, "learning_rate": 7.817258883248731e-06, "loss": 1.0585, "step": 231 }, { "epoch": 0.01, "learning_rate": 7.851099830795262e-06, "loss": 1.1633, "step": 232 }, { "epoch": 0.01, "learning_rate": 7.884940778341795e-06, "loss": 1.1713, "step": 233 }, { "epoch": 0.01, "learning_rate": 7.918781725888326e-06, "loss": 1.0479, "step": 234 }, { "epoch": 0.01, "learning_rate": 7.952622673434857e-06, "loss": 1.0268, "step": 235 }, { "epoch": 0.01, "learning_rate": 7.986463620981388e-06, "loss": 1.1834, "step": 236 }, { "epoch": 0.01, "learning_rate": 8.020304568527919e-06, "loss": 1.1242, "step": 237 }, { "epoch": 0.01, "learning_rate": 8.05414551607445e-06, "loss": 1.1637, "step": 238 }, { "epoch": 0.01, "learning_rate": 8.087986463620981e-06, "loss": 1.1161, "step": 239 }, { "epoch": 0.01, "learning_rate": 8.121827411167514e-06, "loss": 1.053, "step": 240 }, { "epoch": 0.01, "learning_rate": 8.155668358714045e-06, "loss": 1.0939, "step": 241 }, { "epoch": 0.01, "learning_rate": 8.189509306260576e-06, "loss": 1.0377, "step": 242 }, { "epoch": 0.01, "learning_rate": 8.223350253807107e-06, "loss": 1.1816, "step": 243 }, { "epoch": 0.01, "learning_rate": 8.257191201353638e-06, "loss": 1.0858, "step": 244 }, { "epoch": 0.01, "learning_rate": 8.29103214890017e-06, "loss": 1.1753, "step": 245 }, { "epoch": 0.01, "learning_rate": 8.324873096446702e-06, "loss": 1.1442, "step": 246 }, { "epoch": 0.01, "learning_rate": 8.358714043993233e-06, "loss": 1.1555, "step": 247 }, { "epoch": 0.01, "learning_rate": 8.392554991539764e-06, "loss": 1.0542, "step": 248 }, { "epoch": 0.01, "learning_rate": 8.426395939086295e-06, "loss": 1.1046, "step": 249 }, { "epoch": 0.01, "learning_rate": 8.460236886632826e-06, "loss": 1.053, "step": 250 }, { "epoch": 0.01, "learning_rate": 8.494077834179357e-06, "loss": 1.069, "step": 251 }, { "epoch": 0.01, "learning_rate": 8.52791878172589e-06, "loss": 1.2141, "step": 252 }, { "epoch": 0.01, "learning_rate": 8.561759729272421e-06, "loss": 1.1837, "step": 253 }, { "epoch": 0.01, "learning_rate": 8.595600676818952e-06, "loss": 1.1487, "step": 254 }, { "epoch": 0.01, "learning_rate": 8.629441624365483e-06, "loss": 1.1827, "step": 255 }, { "epoch": 0.01, "learning_rate": 8.663282571912014e-06, "loss": 1.1032, "step": 256 }, { "epoch": 0.01, "learning_rate": 8.697123519458545e-06, "loss": 1.1272, "step": 257 }, { "epoch": 0.01, "learning_rate": 8.730964467005076e-06, "loss": 1.1545, "step": 258 }, { "epoch": 0.01, "learning_rate": 8.764805414551609e-06, "loss": 1.1901, "step": 259 }, { "epoch": 0.01, "learning_rate": 8.79864636209814e-06, "loss": 1.0879, "step": 260 }, { "epoch": 0.01, "learning_rate": 8.832487309644671e-06, "loss": 1.076, "step": 261 }, { "epoch": 0.01, "learning_rate": 8.866328257191202e-06, "loss": 1.1172, "step": 262 }, { "epoch": 0.01, "learning_rate": 8.900169204737733e-06, "loss": 1.1371, "step": 263 }, { "epoch": 0.01, "learning_rate": 8.934010152284264e-06, "loss": 1.185, "step": 264 }, { "epoch": 0.01, "learning_rate": 8.967851099830795e-06, "loss": 1.0508, "step": 265 }, { "epoch": 0.01, "learning_rate": 9.001692047377328e-06, "loss": 1.0719, "step": 266 }, { "epoch": 0.01, "learning_rate": 9.035532994923859e-06, "loss": 1.0963, "step": 267 }, { "epoch": 0.01, "learning_rate": 9.06937394247039e-06, "loss": 1.1142, "step": 268 }, { "epoch": 0.01, "learning_rate": 9.103214890016921e-06, "loss": 1.1553, "step": 269 }, { "epoch": 0.01, "learning_rate": 9.137055837563452e-06, "loss": 1.0908, "step": 270 }, { "epoch": 0.01, "learning_rate": 9.170896785109983e-06, "loss": 1.0871, "step": 271 }, { "epoch": 0.01, "learning_rate": 9.204737732656514e-06, "loss": 1.1193, "step": 272 }, { "epoch": 0.01, "learning_rate": 9.238578680203047e-06, "loss": 1.0833, "step": 273 }, { "epoch": 0.01, "learning_rate": 9.272419627749578e-06, "loss": 1.0786, "step": 274 }, { "epoch": 0.01, "learning_rate": 9.306260575296109e-06, "loss": 1.0884, "step": 275 }, { "epoch": 0.01, "learning_rate": 9.34010152284264e-06, "loss": 1.1246, "step": 276 }, { "epoch": 0.01, "learning_rate": 9.373942470389171e-06, "loss": 1.1676, "step": 277 }, { "epoch": 0.01, "learning_rate": 9.407783417935702e-06, "loss": 1.1963, "step": 278 }, { "epoch": 0.01, "learning_rate": 9.441624365482235e-06, "loss": 1.2129, "step": 279 }, { "epoch": 0.01, "learning_rate": 9.475465313028766e-06, "loss": 1.0446, "step": 280 }, { "epoch": 0.01, "learning_rate": 9.509306260575297e-06, "loss": 1.0901, "step": 281 }, { "epoch": 0.01, "learning_rate": 9.543147208121828e-06, "loss": 1.1193, "step": 282 }, { "epoch": 0.01, "learning_rate": 9.57698815566836e-06, "loss": 1.1492, "step": 283 }, { "epoch": 0.01, "learning_rate": 9.61082910321489e-06, "loss": 1.0817, "step": 284 }, { "epoch": 0.01, "learning_rate": 9.644670050761421e-06, "loss": 1.0798, "step": 285 }, { "epoch": 0.01, "learning_rate": 9.678510998307954e-06, "loss": 1.0994, "step": 286 }, { "epoch": 0.01, "learning_rate": 9.712351945854485e-06, "loss": 1.0548, "step": 287 }, { "epoch": 0.01, "learning_rate": 9.746192893401016e-06, "loss": 1.0789, "step": 288 }, { "epoch": 0.01, "learning_rate": 9.780033840947547e-06, "loss": 1.1345, "step": 289 }, { "epoch": 0.01, "learning_rate": 9.813874788494078e-06, "loss": 1.1024, "step": 290 }, { "epoch": 0.01, "learning_rate": 9.84771573604061e-06, "loss": 1.2678, "step": 291 }, { "epoch": 0.01, "learning_rate": 9.88155668358714e-06, "loss": 1.0565, "step": 292 }, { "epoch": 0.01, "learning_rate": 9.915397631133673e-06, "loss": 1.0598, "step": 293 }, { "epoch": 0.01, "learning_rate": 9.949238578680204e-06, "loss": 1.0153, "step": 294 }, { "epoch": 0.01, "learning_rate": 9.983079526226735e-06, "loss": 1.0908, "step": 295 }, { "epoch": 0.02, "learning_rate": 1.0016920473773266e-05, "loss": 1.124, "step": 296 }, { "epoch": 0.02, "learning_rate": 1.0050761421319797e-05, "loss": 1.1463, "step": 297 }, { "epoch": 0.02, "learning_rate": 1.008460236886633e-05, "loss": 1.1264, "step": 298 }, { "epoch": 0.02, "learning_rate": 1.0118443316412861e-05, "loss": 1.0638, "step": 299 }, { "epoch": 0.02, "learning_rate": 1.015228426395939e-05, "loss": 1.0941, "step": 300 }, { "epoch": 0.02, "learning_rate": 1.0186125211505923e-05, "loss": 1.1703, "step": 301 }, { "epoch": 0.02, "learning_rate": 1.0219966159052454e-05, "loss": 1.1026, "step": 302 }, { "epoch": 0.02, "learning_rate": 1.0253807106598985e-05, "loss": 1.0691, "step": 303 }, { "epoch": 0.02, "learning_rate": 1.0287648054145516e-05, "loss": 1.0837, "step": 304 }, { "epoch": 0.02, "learning_rate": 1.032148900169205e-05, "loss": 1.1953, "step": 305 }, { "epoch": 0.02, "learning_rate": 1.035532994923858e-05, "loss": 1.0894, "step": 306 }, { "epoch": 0.02, "learning_rate": 1.038917089678511e-05, "loss": 1.0795, "step": 307 }, { "epoch": 0.02, "learning_rate": 1.0423011844331642e-05, "loss": 1.0278, "step": 308 }, { "epoch": 0.02, "learning_rate": 1.0456852791878173e-05, "loss": 1.1061, "step": 309 }, { "epoch": 0.02, "learning_rate": 1.0490693739424704e-05, "loss": 1.1287, "step": 310 }, { "epoch": 0.02, "learning_rate": 1.0524534686971236e-05, "loss": 1.0847, "step": 311 }, { "epoch": 0.02, "learning_rate": 1.0558375634517768e-05, "loss": 1.0939, "step": 312 }, { "epoch": 0.02, "learning_rate": 1.05922165820643e-05, "loss": 1.16, "step": 313 }, { "epoch": 0.02, "learning_rate": 1.0626057529610829e-05, "loss": 1.1017, "step": 314 }, { "epoch": 0.02, "learning_rate": 1.0659898477157361e-05, "loss": 1.0303, "step": 315 }, { "epoch": 0.02, "learning_rate": 1.0693739424703892e-05, "loss": 1.1731, "step": 316 }, { "epoch": 0.02, "learning_rate": 1.0727580372250424e-05, "loss": 1.135, "step": 317 }, { "epoch": 0.02, "learning_rate": 1.0761421319796955e-05, "loss": 1.1246, "step": 318 }, { "epoch": 0.02, "learning_rate": 1.0795262267343487e-05, "loss": 1.1216, "step": 319 }, { "epoch": 0.02, "learning_rate": 1.0829103214890018e-05, "loss": 1.1133, "step": 320 }, { "epoch": 0.02, "learning_rate": 1.0862944162436548e-05, "loss": 1.0742, "step": 321 }, { "epoch": 0.02, "learning_rate": 1.089678510998308e-05, "loss": 1.1178, "step": 322 }, { "epoch": 0.02, "learning_rate": 1.0930626057529612e-05, "loss": 1.1025, "step": 323 }, { "epoch": 0.02, "learning_rate": 1.0964467005076143e-05, "loss": 1.1398, "step": 324 }, { "epoch": 0.02, "learning_rate": 1.0998307952622674e-05, "loss": 1.2119, "step": 325 }, { "epoch": 0.02, "learning_rate": 1.1032148900169206e-05, "loss": 1.0576, "step": 326 }, { "epoch": 0.02, "learning_rate": 1.1065989847715737e-05, "loss": 1.1766, "step": 327 }, { "epoch": 0.02, "learning_rate": 1.1099830795262267e-05, "loss": 1.0802, "step": 328 }, { "epoch": 0.02, "learning_rate": 1.11336717428088e-05, "loss": 1.0857, "step": 329 }, { "epoch": 0.02, "learning_rate": 1.116751269035533e-05, "loss": 1.0982, "step": 330 }, { "epoch": 0.02, "learning_rate": 1.1201353637901862e-05, "loss": 1.0403, "step": 331 }, { "epoch": 0.02, "learning_rate": 1.1235194585448393e-05, "loss": 1.1227, "step": 332 }, { "epoch": 0.02, "learning_rate": 1.1269035532994925e-05, "loss": 1.0579, "step": 333 }, { "epoch": 0.02, "learning_rate": 1.1302876480541457e-05, "loss": 1.1953, "step": 334 }, { "epoch": 0.02, "learning_rate": 1.1336717428087986e-05, "loss": 1.0146, "step": 335 }, { "epoch": 0.02, "learning_rate": 1.1370558375634519e-05, "loss": 1.0587, "step": 336 }, { "epoch": 0.02, "learning_rate": 1.140439932318105e-05, "loss": 1.1506, "step": 337 }, { "epoch": 0.02, "learning_rate": 1.143824027072758e-05, "loss": 1.1454, "step": 338 }, { "epoch": 0.02, "learning_rate": 1.1472081218274113e-05, "loss": 1.1225, "step": 339 }, { "epoch": 0.02, "learning_rate": 1.1505922165820645e-05, "loss": 1.0517, "step": 340 }, { "epoch": 0.02, "learning_rate": 1.1539763113367176e-05, "loss": 1.1058, "step": 341 }, { "epoch": 0.02, "learning_rate": 1.1573604060913705e-05, "loss": 1.1052, "step": 342 }, { "epoch": 0.02, "learning_rate": 1.160744500846024e-05, "loss": 1.0944, "step": 343 }, { "epoch": 0.02, "learning_rate": 1.1641285956006769e-05, "loss": 1.1598, "step": 344 }, { "epoch": 0.02, "learning_rate": 1.16751269035533e-05, "loss": 1.1602, "step": 345 }, { "epoch": 0.02, "learning_rate": 1.1708967851099833e-05, "loss": 1.0975, "step": 346 }, { "epoch": 0.02, "learning_rate": 1.1742808798646364e-05, "loss": 1.166, "step": 347 }, { "epoch": 0.02, "learning_rate": 1.1776649746192895e-05, "loss": 1.1298, "step": 348 }, { "epoch": 0.02, "learning_rate": 1.1810490693739424e-05, "loss": 1.061, "step": 349 }, { "epoch": 0.02, "learning_rate": 1.1844331641285958e-05, "loss": 1.0388, "step": 350 }, { "epoch": 0.02, "learning_rate": 1.1878172588832488e-05, "loss": 1.0829, "step": 351 }, { "epoch": 0.02, "learning_rate": 1.1912013536379019e-05, "loss": 1.113, "step": 352 }, { "epoch": 0.02, "learning_rate": 1.1945854483925552e-05, "loss": 1.1005, "step": 353 }, { "epoch": 0.02, "learning_rate": 1.1979695431472083e-05, "loss": 1.1343, "step": 354 }, { "epoch": 0.02, "learning_rate": 1.2013536379018614e-05, "loss": 1.048, "step": 355 }, { "epoch": 0.02, "learning_rate": 1.2047377326565143e-05, "loss": 1.1, "step": 356 }, { "epoch": 0.02, "learning_rate": 1.2081218274111678e-05, "loss": 1.1631, "step": 357 }, { "epoch": 0.02, "learning_rate": 1.2115059221658207e-05, "loss": 1.0826, "step": 358 }, { "epoch": 0.02, "learning_rate": 1.2148900169204738e-05, "loss": 1.0032, "step": 359 }, { "epoch": 0.02, "learning_rate": 1.218274111675127e-05, "loss": 1.1189, "step": 360 }, { "epoch": 0.02, "learning_rate": 1.2216582064297802e-05, "loss": 1.1063, "step": 361 }, { "epoch": 0.02, "learning_rate": 1.2250423011844333e-05, "loss": 1.0854, "step": 362 }, { "epoch": 0.02, "learning_rate": 1.2284263959390864e-05, "loss": 1.1046, "step": 363 }, { "epoch": 0.02, "learning_rate": 1.2318104906937397e-05, "loss": 1.0721, "step": 364 }, { "epoch": 0.02, "learning_rate": 1.2351945854483926e-05, "loss": 1.2584, "step": 365 }, { "epoch": 0.02, "learning_rate": 1.2385786802030457e-05, "loss": 1.1016, "step": 366 }, { "epoch": 0.02, "learning_rate": 1.241962774957699e-05, "loss": 1.0521, "step": 367 }, { "epoch": 0.02, "learning_rate": 1.245346869712352e-05, "loss": 1.1473, "step": 368 }, { "epoch": 0.02, "learning_rate": 1.2487309644670052e-05, "loss": 1.1489, "step": 369 }, { "epoch": 0.02, "learning_rate": 1.2521150592216583e-05, "loss": 1.0432, "step": 370 }, { "epoch": 0.02, "learning_rate": 1.2554991539763116e-05, "loss": 0.9692, "step": 371 }, { "epoch": 0.02, "learning_rate": 1.2588832487309645e-05, "loss": 0.9957, "step": 372 }, { "epoch": 0.02, "learning_rate": 1.2622673434856176e-05, "loss": 1.1006, "step": 373 }, { "epoch": 0.02, "learning_rate": 1.2656514382402709e-05, "loss": 1.0378, "step": 374 }, { "epoch": 0.02, "learning_rate": 1.269035532994924e-05, "loss": 1.0638, "step": 375 }, { "epoch": 0.02, "learning_rate": 1.2724196277495771e-05, "loss": 1.0375, "step": 376 }, { "epoch": 0.02, "learning_rate": 1.2758037225042302e-05, "loss": 1.0611, "step": 377 }, { "epoch": 0.02, "learning_rate": 1.2791878172588835e-05, "loss": 1.1016, "step": 378 }, { "epoch": 0.02, "learning_rate": 1.2825719120135364e-05, "loss": 1.2265, "step": 379 }, { "epoch": 0.02, "learning_rate": 1.2859560067681895e-05, "loss": 1.1303, "step": 380 }, { "epoch": 0.02, "learning_rate": 1.2893401015228428e-05, "loss": 1.182, "step": 381 }, { "epoch": 0.02, "learning_rate": 1.2927241962774959e-05, "loss": 1.0128, "step": 382 }, { "epoch": 0.02, "learning_rate": 1.296108291032149e-05, "loss": 1.0899, "step": 383 }, { "epoch": 0.02, "learning_rate": 1.2994923857868021e-05, "loss": 1.069, "step": 384 }, { "epoch": 0.02, "learning_rate": 1.3028764805414554e-05, "loss": 1.1205, "step": 385 }, { "epoch": 0.02, "learning_rate": 1.3062605752961083e-05, "loss": 1.0654, "step": 386 }, { "epoch": 0.02, "learning_rate": 1.3096446700507614e-05, "loss": 1.0814, "step": 387 }, { "epoch": 0.02, "learning_rate": 1.3130287648054147e-05, "loss": 1.0734, "step": 388 }, { "epoch": 0.02, "learning_rate": 1.3164128595600678e-05, "loss": 1.0605, "step": 389 }, { "epoch": 0.02, "learning_rate": 1.3197969543147209e-05, "loss": 1.0333, "step": 390 }, { "epoch": 0.02, "learning_rate": 1.323181049069374e-05, "loss": 1.0628, "step": 391 }, { "epoch": 0.02, "learning_rate": 1.3265651438240273e-05, "loss": 1.1465, "step": 392 }, { "epoch": 0.02, "learning_rate": 1.3299492385786802e-05, "loss": 1.0481, "step": 393 }, { "epoch": 0.02, "learning_rate": 1.3333333333333333e-05, "loss": 1.1539, "step": 394 }, { "epoch": 0.02, "learning_rate": 1.3367174280879866e-05, "loss": 1.0933, "step": 395 }, { "epoch": 0.02, "learning_rate": 1.3401015228426397e-05, "loss": 1.1057, "step": 396 }, { "epoch": 0.02, "learning_rate": 1.3434856175972928e-05, "loss": 1.1783, "step": 397 }, { "epoch": 0.02, "learning_rate": 1.346869712351946e-05, "loss": 1.0409, "step": 398 }, { "epoch": 0.02, "learning_rate": 1.3502538071065992e-05, "loss": 1.042, "step": 399 }, { "epoch": 0.02, "learning_rate": 1.3536379018612521e-05, "loss": 1.0578, "step": 400 }, { "epoch": 0.02, "learning_rate": 1.3570219966159052e-05, "loss": 1.0651, "step": 401 }, { "epoch": 0.02, "learning_rate": 1.3604060913705585e-05, "loss": 1.0302, "step": 402 }, { "epoch": 0.02, "learning_rate": 1.3637901861252116e-05, "loss": 1.1387, "step": 403 }, { "epoch": 0.02, "learning_rate": 1.3671742808798647e-05, "loss": 1.0977, "step": 404 }, { "epoch": 0.02, "learning_rate": 1.3705583756345178e-05, "loss": 1.1134, "step": 405 }, { "epoch": 0.02, "learning_rate": 1.3739424703891711e-05, "loss": 1.1311, "step": 406 }, { "epoch": 0.02, "learning_rate": 1.3773265651438242e-05, "loss": 1.076, "step": 407 }, { "epoch": 0.02, "learning_rate": 1.3807106598984771e-05, "loss": 1.0326, "step": 408 }, { "epoch": 0.02, "learning_rate": 1.3840947546531304e-05, "loss": 1.0938, "step": 409 }, { "epoch": 0.02, "learning_rate": 1.3874788494077835e-05, "loss": 1.0893, "step": 410 }, { "epoch": 0.02, "learning_rate": 1.3908629441624366e-05, "loss": 1.0333, "step": 411 }, { "epoch": 0.02, "learning_rate": 1.3942470389170899e-05, "loss": 1.1017, "step": 412 }, { "epoch": 0.02, "learning_rate": 1.397631133671743e-05, "loss": 1.0922, "step": 413 }, { "epoch": 0.02, "learning_rate": 1.4010152284263961e-05, "loss": 1.2079, "step": 414 }, { "epoch": 0.02, "learning_rate": 1.404399323181049e-05, "loss": 1.1675, "step": 415 }, { "epoch": 0.02, "learning_rate": 1.4077834179357023e-05, "loss": 1.0487, "step": 416 }, { "epoch": 0.02, "learning_rate": 1.4111675126903554e-05, "loss": 1.1093, "step": 417 }, { "epoch": 0.02, "learning_rate": 1.4145516074450085e-05, "loss": 1.1049, "step": 418 }, { "epoch": 0.02, "learning_rate": 1.4179357021996618e-05, "loss": 1.0616, "step": 419 }, { "epoch": 0.02, "learning_rate": 1.421319796954315e-05, "loss": 1.1114, "step": 420 }, { "epoch": 0.02, "learning_rate": 1.424703891708968e-05, "loss": 1.2056, "step": 421 }, { "epoch": 0.02, "learning_rate": 1.428087986463621e-05, "loss": 1.0641, "step": 422 }, { "epoch": 0.02, "learning_rate": 1.4314720812182742e-05, "loss": 1.1383, "step": 423 }, { "epoch": 0.02, "learning_rate": 1.4348561759729273e-05, "loss": 1.0797, "step": 424 }, { "epoch": 0.02, "learning_rate": 1.4382402707275804e-05, "loss": 1.1429, "step": 425 }, { "epoch": 0.02, "learning_rate": 1.4416243654822337e-05, "loss": 1.1678, "step": 426 }, { "epoch": 0.02, "learning_rate": 1.4450084602368868e-05, "loss": 1.1421, "step": 427 }, { "epoch": 0.02, "learning_rate": 1.44839255499154e-05, "loss": 1.0447, "step": 428 }, { "epoch": 0.02, "learning_rate": 1.4517766497461929e-05, "loss": 1.1721, "step": 429 }, { "epoch": 0.02, "learning_rate": 1.4551607445008461e-05, "loss": 1.1381, "step": 430 }, { "epoch": 0.02, "learning_rate": 1.4585448392554992e-05, "loss": 0.995, "step": 431 }, { "epoch": 0.02, "learning_rate": 1.4619289340101523e-05, "loss": 1.1015, "step": 432 }, { "epoch": 0.02, "learning_rate": 1.4653130287648056e-05, "loss": 1.0643, "step": 433 }, { "epoch": 0.02, "learning_rate": 1.4686971235194587e-05, "loss": 1.0829, "step": 434 }, { "epoch": 0.02, "learning_rate": 1.4720812182741118e-05, "loss": 1.1017, "step": 435 }, { "epoch": 0.02, "learning_rate": 1.4754653130287648e-05, "loss": 1.0871, "step": 436 }, { "epoch": 0.02, "learning_rate": 1.478849407783418e-05, "loss": 1.0673, "step": 437 }, { "epoch": 0.02, "learning_rate": 1.4822335025380712e-05, "loss": 1.1509, "step": 438 }, { "epoch": 0.02, "learning_rate": 1.4856175972927243e-05, "loss": 1.0891, "step": 439 }, { "epoch": 0.02, "learning_rate": 1.4890016920473775e-05, "loss": 1.1037, "step": 440 }, { "epoch": 0.02, "learning_rate": 1.4923857868020306e-05, "loss": 1.1002, "step": 441 }, { "epoch": 0.02, "learning_rate": 1.4957698815566837e-05, "loss": 1.0967, "step": 442 }, { "epoch": 0.02, "learning_rate": 1.4991539763113367e-05, "loss": 1.0668, "step": 443 }, { "epoch": 0.02, "learning_rate": 1.50253807106599e-05, "loss": 1.0313, "step": 444 }, { "epoch": 0.02, "learning_rate": 1.505922165820643e-05, "loss": 1.1226, "step": 445 }, { "epoch": 0.02, "learning_rate": 1.5093062605752962e-05, "loss": 1.0992, "step": 446 }, { "epoch": 0.02, "learning_rate": 1.5126903553299494e-05, "loss": 1.0233, "step": 447 }, { "epoch": 0.02, "learning_rate": 1.5160744500846025e-05, "loss": 1.1497, "step": 448 }, { "epoch": 0.02, "learning_rate": 1.5194585448392556e-05, "loss": 1.1161, "step": 449 }, { "epoch": 0.02, "learning_rate": 1.5228426395939086e-05, "loss": 1.1528, "step": 450 }, { "epoch": 0.02, "learning_rate": 1.526226734348562e-05, "loss": 1.1614, "step": 451 }, { "epoch": 0.02, "learning_rate": 1.529610829103215e-05, "loss": 1.1726, "step": 452 }, { "epoch": 0.02, "learning_rate": 1.5329949238578682e-05, "loss": 1.0222, "step": 453 }, { "epoch": 0.02, "learning_rate": 1.5363790186125213e-05, "loss": 1.115, "step": 454 }, { "epoch": 0.02, "learning_rate": 1.5397631133671744e-05, "loss": 1.1172, "step": 455 }, { "epoch": 0.02, "learning_rate": 1.5431472081218276e-05, "loss": 1.1425, "step": 456 }, { "epoch": 0.02, "learning_rate": 1.5465313028764807e-05, "loss": 1.12, "step": 457 }, { "epoch": 0.02, "learning_rate": 1.5499153976311338e-05, "loss": 1.1005, "step": 458 }, { "epoch": 0.02, "learning_rate": 1.553299492385787e-05, "loss": 1.0584, "step": 459 }, { "epoch": 0.02, "learning_rate": 1.55668358714044e-05, "loss": 1.05, "step": 460 }, { "epoch": 0.02, "learning_rate": 1.5600676818950934e-05, "loss": 1.0817, "step": 461 }, { "epoch": 0.02, "learning_rate": 1.5634517766497462e-05, "loss": 1.0432, "step": 462 }, { "epoch": 0.02, "learning_rate": 1.5668358714043993e-05, "loss": 0.9489, "step": 463 }, { "epoch": 0.02, "learning_rate": 1.5702199661590524e-05, "loss": 1.1783, "step": 464 }, { "epoch": 0.02, "learning_rate": 1.573604060913706e-05, "loss": 1.1507, "step": 465 }, { "epoch": 0.02, "learning_rate": 1.576988155668359e-05, "loss": 1.133, "step": 466 }, { "epoch": 0.02, "learning_rate": 1.580372250423012e-05, "loss": 1.0286, "step": 467 }, { "epoch": 0.02, "learning_rate": 1.583756345177665e-05, "loss": 1.0519, "step": 468 }, { "epoch": 0.02, "learning_rate": 1.5871404399323183e-05, "loss": 1.1332, "step": 469 }, { "epoch": 0.02, "learning_rate": 1.5905245346869714e-05, "loss": 1.186, "step": 470 }, { "epoch": 0.02, "learning_rate": 1.5939086294416245e-05, "loss": 1.1931, "step": 471 }, { "epoch": 0.02, "learning_rate": 1.5972927241962776e-05, "loss": 1.0645, "step": 472 }, { "epoch": 0.02, "learning_rate": 1.6006768189509307e-05, "loss": 1.1014, "step": 473 }, { "epoch": 0.02, "learning_rate": 1.6040609137055838e-05, "loss": 1.0775, "step": 474 }, { "epoch": 0.02, "learning_rate": 1.6074450084602372e-05, "loss": 1.1287, "step": 475 }, { "epoch": 0.02, "learning_rate": 1.61082910321489e-05, "loss": 1.1344, "step": 476 }, { "epoch": 0.02, "learning_rate": 1.614213197969543e-05, "loss": 1.0975, "step": 477 }, { "epoch": 0.02, "learning_rate": 1.6175972927241962e-05, "loss": 1.0605, "step": 478 }, { "epoch": 0.02, "learning_rate": 1.6209813874788497e-05, "loss": 1.0261, "step": 479 }, { "epoch": 0.02, "learning_rate": 1.6243654822335028e-05, "loss": 1.0137, "step": 480 }, { "epoch": 0.02, "learning_rate": 1.627749576988156e-05, "loss": 1.0762, "step": 481 }, { "epoch": 0.02, "learning_rate": 1.631133671742809e-05, "loss": 1.0467, "step": 482 }, { "epoch": 0.02, "learning_rate": 1.634517766497462e-05, "loss": 0.9897, "step": 483 }, { "epoch": 0.02, "learning_rate": 1.6379018612521152e-05, "loss": 1.083, "step": 484 }, { "epoch": 0.02, "learning_rate": 1.6412859560067683e-05, "loss": 1.1263, "step": 485 }, { "epoch": 0.02, "learning_rate": 1.6446700507614214e-05, "loss": 1.0652, "step": 486 }, { "epoch": 0.02, "learning_rate": 1.6480541455160745e-05, "loss": 1.099, "step": 487 }, { "epoch": 0.02, "learning_rate": 1.6514382402707276e-05, "loss": 1.1179, "step": 488 }, { "epoch": 0.02, "learning_rate": 1.654822335025381e-05, "loss": 1.095, "step": 489 }, { "epoch": 0.02, "learning_rate": 1.658206429780034e-05, "loss": 1.1686, "step": 490 }, { "epoch": 0.02, "learning_rate": 1.661590524534687e-05, "loss": 1.0413, "step": 491 }, { "epoch": 0.02, "learning_rate": 1.6649746192893404e-05, "loss": 1.0964, "step": 492 }, { "epoch": 0.03, "learning_rate": 1.6683587140439935e-05, "loss": 1.1198, "step": 493 }, { "epoch": 0.03, "learning_rate": 1.6717428087986466e-05, "loss": 1.1138, "step": 494 }, { "epoch": 0.03, "learning_rate": 1.6751269035532997e-05, "loss": 1.149, "step": 495 }, { "epoch": 0.03, "learning_rate": 1.6785109983079528e-05, "loss": 1.1252, "step": 496 }, { "epoch": 0.03, "learning_rate": 1.681895093062606e-05, "loss": 1.1082, "step": 497 }, { "epoch": 0.03, "learning_rate": 1.685279187817259e-05, "loss": 1.0084, "step": 498 }, { "epoch": 0.03, "learning_rate": 1.688663282571912e-05, "loss": 1.1481, "step": 499 }, { "epoch": 0.03, "learning_rate": 1.6920473773265652e-05, "loss": 1.0737, "step": 500 }, { "epoch": 0.03, "learning_rate": 1.6954314720812183e-05, "loss": 1.1058, "step": 501 }, { "epoch": 0.03, "learning_rate": 1.6988155668358714e-05, "loss": 1.0449, "step": 502 }, { "epoch": 0.03, "learning_rate": 1.702199661590525e-05, "loss": 1.0845, "step": 503 }, { "epoch": 0.03, "learning_rate": 1.705583756345178e-05, "loss": 1.1005, "step": 504 }, { "epoch": 0.03, "learning_rate": 1.7089678510998307e-05, "loss": 1.0213, "step": 505 }, { "epoch": 0.03, "learning_rate": 1.7123519458544842e-05, "loss": 1.1354, "step": 506 }, { "epoch": 0.03, "learning_rate": 1.7157360406091373e-05, "loss": 1.0913, "step": 507 }, { "epoch": 0.03, "learning_rate": 1.7191201353637904e-05, "loss": 1.1288, "step": 508 }, { "epoch": 0.03, "learning_rate": 1.7225042301184435e-05, "loss": 1.0518, "step": 509 }, { "epoch": 0.03, "learning_rate": 1.7258883248730966e-05, "loss": 1.1074, "step": 510 }, { "epoch": 0.03, "learning_rate": 1.7292724196277497e-05, "loss": 1.1058, "step": 511 }, { "epoch": 0.03, "learning_rate": 1.7326565143824028e-05, "loss": 1.1743, "step": 512 }, { "epoch": 0.03, "learning_rate": 1.736040609137056e-05, "loss": 1.0018, "step": 513 }, { "epoch": 0.03, "learning_rate": 1.739424703891709e-05, "loss": 1.1259, "step": 514 }, { "epoch": 0.03, "learning_rate": 1.742808798646362e-05, "loss": 1.1481, "step": 515 }, { "epoch": 0.03, "learning_rate": 1.7461928934010152e-05, "loss": 1.1777, "step": 516 }, { "epoch": 0.03, "learning_rate": 1.7495769881556687e-05, "loss": 1.0599, "step": 517 }, { "epoch": 0.03, "learning_rate": 1.7529610829103218e-05, "loss": 1.04, "step": 518 }, { "epoch": 0.03, "learning_rate": 1.7563451776649745e-05, "loss": 1.0997, "step": 519 }, { "epoch": 0.03, "learning_rate": 1.759729272419628e-05, "loss": 1.0933, "step": 520 }, { "epoch": 0.03, "learning_rate": 1.763113367174281e-05, "loss": 1.1401, "step": 521 }, { "epoch": 0.03, "learning_rate": 1.7664974619289342e-05, "loss": 1.0679, "step": 522 }, { "epoch": 0.03, "learning_rate": 1.7698815566835873e-05, "loss": 0.9808, "step": 523 }, { "epoch": 0.03, "learning_rate": 1.7732656514382404e-05, "loss": 1.0642, "step": 524 }, { "epoch": 0.03, "learning_rate": 1.7766497461928935e-05, "loss": 1.121, "step": 525 }, { "epoch": 0.03, "learning_rate": 1.7800338409475466e-05, "loss": 1.147, "step": 526 }, { "epoch": 0.03, "learning_rate": 1.7834179357022e-05, "loss": 1.1577, "step": 527 }, { "epoch": 0.03, "learning_rate": 1.786802030456853e-05, "loss": 1.0441, "step": 528 }, { "epoch": 0.03, "learning_rate": 1.790186125211506e-05, "loss": 1.0861, "step": 529 }, { "epoch": 0.03, "learning_rate": 1.793570219966159e-05, "loss": 1.1232, "step": 530 }, { "epoch": 0.03, "learning_rate": 1.7969543147208125e-05, "loss": 0.9844, "step": 531 }, { "epoch": 0.03, "learning_rate": 1.8003384094754656e-05, "loss": 1.1849, "step": 532 }, { "epoch": 0.03, "learning_rate": 1.8037225042301184e-05, "loss": 1.0762, "step": 533 }, { "epoch": 0.03, "learning_rate": 1.8071065989847718e-05, "loss": 1.077, "step": 534 }, { "epoch": 0.03, "learning_rate": 1.810490693739425e-05, "loss": 1.1552, "step": 535 }, { "epoch": 0.03, "learning_rate": 1.813874788494078e-05, "loss": 1.111, "step": 536 }, { "epoch": 0.03, "learning_rate": 1.817258883248731e-05, "loss": 1.0731, "step": 537 }, { "epoch": 0.03, "learning_rate": 1.8206429780033842e-05, "loss": 1.0916, "step": 538 }, { "epoch": 0.03, "learning_rate": 1.8240270727580373e-05, "loss": 1.1786, "step": 539 }, { "epoch": 0.03, "learning_rate": 1.8274111675126904e-05, "loss": 1.1787, "step": 540 }, { "epoch": 0.03, "learning_rate": 1.830795262267344e-05, "loss": 1.069, "step": 541 }, { "epoch": 0.03, "learning_rate": 1.8341793570219966e-05, "loss": 1.073, "step": 542 }, { "epoch": 0.03, "learning_rate": 1.8375634517766498e-05, "loss": 1.0657, "step": 543 }, { "epoch": 0.03, "learning_rate": 1.840947546531303e-05, "loss": 1.0064, "step": 544 }, { "epoch": 0.03, "learning_rate": 1.8443316412859563e-05, "loss": 0.9341, "step": 545 }, { "epoch": 0.03, "learning_rate": 1.8477157360406094e-05, "loss": 1.0438, "step": 546 }, { "epoch": 0.03, "learning_rate": 1.8510998307952622e-05, "loss": 1.1931, "step": 547 }, { "epoch": 0.03, "learning_rate": 1.8544839255499156e-05, "loss": 1.1111, "step": 548 }, { "epoch": 0.03, "learning_rate": 1.8578680203045687e-05, "loss": 1.2041, "step": 549 }, { "epoch": 0.03, "learning_rate": 1.8612521150592218e-05, "loss": 1.0181, "step": 550 }, { "epoch": 0.03, "learning_rate": 1.864636209813875e-05, "loss": 1.017, "step": 551 }, { "epoch": 0.03, "learning_rate": 1.868020304568528e-05, "loss": 1.1048, "step": 552 }, { "epoch": 0.03, "learning_rate": 1.871404399323181e-05, "loss": 1.0591, "step": 553 }, { "epoch": 0.03, "learning_rate": 1.8747884940778342e-05, "loss": 1.0289, "step": 554 }, { "epoch": 0.03, "learning_rate": 1.8781725888324877e-05, "loss": 1.009, "step": 555 }, { "epoch": 0.03, "learning_rate": 1.8815566835871405e-05, "loss": 1.0986, "step": 556 }, { "epoch": 0.03, "learning_rate": 1.8849407783417936e-05, "loss": 1.1133, "step": 557 }, { "epoch": 0.03, "learning_rate": 1.888324873096447e-05, "loss": 1.1318, "step": 558 }, { "epoch": 0.03, "learning_rate": 1.8917089678511e-05, "loss": 1.1292, "step": 559 }, { "epoch": 0.03, "learning_rate": 1.8950930626057532e-05, "loss": 1.0668, "step": 560 }, { "epoch": 0.03, "learning_rate": 1.8984771573604063e-05, "loss": 1.0292, "step": 561 }, { "epoch": 0.03, "learning_rate": 1.9018612521150594e-05, "loss": 1.1708, "step": 562 }, { "epoch": 0.03, "learning_rate": 1.9052453468697125e-05, "loss": 1.099, "step": 563 }, { "epoch": 0.03, "learning_rate": 1.9086294416243656e-05, "loss": 1.116, "step": 564 }, { "epoch": 0.03, "learning_rate": 1.9120135363790187e-05, "loss": 1.0877, "step": 565 }, { "epoch": 0.03, "learning_rate": 1.915397631133672e-05, "loss": 1.1372, "step": 566 }, { "epoch": 0.03, "learning_rate": 1.918781725888325e-05, "loss": 1.0779, "step": 567 }, { "epoch": 0.03, "learning_rate": 1.922165820642978e-05, "loss": 1.1566, "step": 568 }, { "epoch": 0.03, "learning_rate": 1.9255499153976315e-05, "loss": 1.0531, "step": 569 }, { "epoch": 0.03, "learning_rate": 1.9289340101522843e-05, "loss": 1.0644, "step": 570 }, { "epoch": 0.03, "learning_rate": 1.9323181049069374e-05, "loss": 1.0569, "step": 571 }, { "epoch": 0.03, "learning_rate": 1.9357021996615908e-05, "loss": 1.0647, "step": 572 }, { "epoch": 0.03, "learning_rate": 1.939086294416244e-05, "loss": 1.1623, "step": 573 }, { "epoch": 0.03, "learning_rate": 1.942470389170897e-05, "loss": 1.1082, "step": 574 }, { "epoch": 0.03, "learning_rate": 1.94585448392555e-05, "loss": 1.2148, "step": 575 }, { "epoch": 0.03, "learning_rate": 1.9492385786802032e-05, "loss": 0.9925, "step": 576 }, { "epoch": 0.03, "learning_rate": 1.9526226734348564e-05, "loss": 1.1084, "step": 577 }, { "epoch": 0.03, "learning_rate": 1.9560067681895095e-05, "loss": 1.1373, "step": 578 }, { "epoch": 0.03, "learning_rate": 1.9593908629441626e-05, "loss": 1.1691, "step": 579 }, { "epoch": 0.03, "learning_rate": 1.9627749576988157e-05, "loss": 1.062, "step": 580 }, { "epoch": 0.03, "learning_rate": 1.9661590524534688e-05, "loss": 1.0329, "step": 581 }, { "epoch": 0.03, "learning_rate": 1.969543147208122e-05, "loss": 1.0564, "step": 582 }, { "epoch": 0.03, "learning_rate": 1.9729272419627753e-05, "loss": 1.1058, "step": 583 }, { "epoch": 0.03, "learning_rate": 1.976311336717428e-05, "loss": 1.1299, "step": 584 }, { "epoch": 0.03, "learning_rate": 1.9796954314720812e-05, "loss": 1.0337, "step": 585 }, { "epoch": 0.03, "learning_rate": 1.9830795262267346e-05, "loss": 1.127, "step": 586 }, { "epoch": 0.03, "learning_rate": 1.9864636209813877e-05, "loss": 1.0555, "step": 587 }, { "epoch": 0.03, "learning_rate": 1.989847715736041e-05, "loss": 1.1456, "step": 588 }, { "epoch": 0.03, "learning_rate": 1.993231810490694e-05, "loss": 1.0802, "step": 589 }, { "epoch": 0.03, "learning_rate": 1.996615905245347e-05, "loss": 1.182, "step": 590 }, { "epoch": 0.03, "learning_rate": 2e-05, "loss": 1.1207, "step": 591 }, { "epoch": 0.03, "learning_rate": 1.9999999864630323e-05, "loss": 1.0754, "step": 592 }, { "epoch": 0.03, "learning_rate": 1.999999945852128e-05, "loss": 1.1532, "step": 593 }, { "epoch": 0.03, "learning_rate": 1.99999987816729e-05, "loss": 1.0477, "step": 594 }, { "epoch": 0.03, "learning_rate": 1.9999997834085188e-05, "loss": 1.137, "step": 595 }, { "epoch": 0.03, "learning_rate": 1.9999996615758174e-05, "loss": 1.122, "step": 596 }, { "epoch": 0.03, "learning_rate": 1.999999512669189e-05, "loss": 1.0753, "step": 597 }, { "epoch": 0.03, "learning_rate": 1.9999993366886377e-05, "loss": 1.0199, "step": 598 }, { "epoch": 0.03, "learning_rate": 1.9999991336341686e-05, "loss": 1.1921, "step": 599 }, { "epoch": 0.03, "learning_rate": 1.9999989035057865e-05, "loss": 1.0363, "step": 600 }, { "epoch": 0.03, "learning_rate": 1.9999986463034985e-05, "loss": 1.1208, "step": 601 }, { "epoch": 0.03, "learning_rate": 1.9999983620273108e-05, "loss": 1.0676, "step": 602 }, { "epoch": 0.03, "learning_rate": 1.999998050677231e-05, "loss": 1.0194, "step": 603 }, { "epoch": 0.03, "learning_rate": 1.9999977122532683e-05, "loss": 1.1245, "step": 604 }, { "epoch": 0.03, "learning_rate": 1.999997346755431e-05, "loss": 1.1207, "step": 605 }, { "epoch": 0.03, "learning_rate": 1.99999695418373e-05, "loss": 1.1417, "step": 606 }, { "epoch": 0.03, "learning_rate": 1.9999965345381753e-05, "loss": 1.0803, "step": 607 }, { "epoch": 0.03, "learning_rate": 1.999996087818778e-05, "loss": 1.1376, "step": 608 }, { "epoch": 0.03, "learning_rate": 1.9999956140255508e-05, "loss": 1.0836, "step": 609 }, { "epoch": 0.03, "learning_rate": 1.999995113158506e-05, "loss": 1.0614, "step": 610 }, { "epoch": 0.03, "learning_rate": 1.999994585217658e-05, "loss": 1.1736, "step": 611 }, { "epoch": 0.03, "learning_rate": 1.9999940302030198e-05, "loss": 1.1674, "step": 612 }, { "epoch": 0.03, "learning_rate": 1.9999934481146076e-05, "loss": 1.0868, "step": 613 }, { "epoch": 0.03, "learning_rate": 1.999992838952437e-05, "loss": 1.1446, "step": 614 }, { "epoch": 0.03, "learning_rate": 1.9999922027165236e-05, "loss": 1.123, "step": 615 }, { "epoch": 0.03, "learning_rate": 1.999991539406885e-05, "loss": 1.1146, "step": 616 }, { "epoch": 0.03, "learning_rate": 1.99999084902354e-05, "loss": 1.0461, "step": 617 }, { "epoch": 0.03, "learning_rate": 1.9999901315665064e-05, "loss": 1.0584, "step": 618 }, { "epoch": 0.03, "learning_rate": 1.9999893870358038e-05, "loss": 1.0854, "step": 619 }, { "epoch": 0.03, "learning_rate": 1.999988615431453e-05, "loss": 1.1158, "step": 620 }, { "epoch": 0.03, "learning_rate": 1.9999878167534736e-05, "loss": 1.1964, "step": 621 }, { "epoch": 0.03, "learning_rate": 1.9999869910018883e-05, "loss": 1.0526, "step": 622 }, { "epoch": 0.03, "learning_rate": 1.999986138176719e-05, "loss": 1.0735, "step": 623 }, { "epoch": 0.03, "learning_rate": 1.999985258277989e-05, "loss": 1.0861, "step": 624 }, { "epoch": 0.03, "learning_rate": 1.9999843513057223e-05, "loss": 1.1262, "step": 625 }, { "epoch": 0.03, "learning_rate": 1.9999834172599427e-05, "loss": 1.0642, "step": 626 }, { "epoch": 0.03, "learning_rate": 1.9999824561406763e-05, "loss": 1.1146, "step": 627 }, { "epoch": 0.03, "learning_rate": 1.999981467947949e-05, "loss": 1.1525, "step": 628 }, { "epoch": 0.03, "learning_rate": 1.9999804526817866e-05, "loss": 1.0577, "step": 629 }, { "epoch": 0.03, "learning_rate": 1.9999794103422182e-05, "loss": 1.0367, "step": 630 }, { "epoch": 0.03, "learning_rate": 1.9999783409292708e-05, "loss": 1.1918, "step": 631 }, { "epoch": 0.03, "learning_rate": 1.9999772444429735e-05, "loss": 1.1885, "step": 632 }, { "epoch": 0.03, "learning_rate": 1.9999761208833562e-05, "loss": 1.0505, "step": 633 }, { "epoch": 0.03, "learning_rate": 1.99997497025045e-05, "loss": 1.1175, "step": 634 }, { "epoch": 0.03, "learning_rate": 1.9999737925442847e-05, "loss": 1.0865, "step": 635 }, { "epoch": 0.03, "learning_rate": 1.999972587764893e-05, "loss": 1.082, "step": 636 }, { "epoch": 0.03, "learning_rate": 1.9999713559123074e-05, "loss": 1.0385, "step": 637 }, { "epoch": 0.03, "learning_rate": 1.999970096986561e-05, "loss": 1.0389, "step": 638 }, { "epoch": 0.03, "learning_rate": 1.9999688109876885e-05, "loss": 1.0507, "step": 639 }, { "epoch": 0.03, "learning_rate": 1.9999674979157242e-05, "loss": 1.0876, "step": 640 }, { "epoch": 0.03, "learning_rate": 1.9999661577707035e-05, "loss": 1.0881, "step": 641 }, { "epoch": 0.03, "learning_rate": 1.9999647905526633e-05, "loss": 1.0455, "step": 642 }, { "epoch": 0.03, "learning_rate": 1.99996339626164e-05, "loss": 1.1573, "step": 643 }, { "epoch": 0.03, "learning_rate": 1.999961974897672e-05, "loss": 1.1721, "step": 644 }, { "epoch": 0.03, "learning_rate": 1.999960526460797e-05, "loss": 1.0999, "step": 645 }, { "epoch": 0.03, "learning_rate": 1.999959050951055e-05, "loss": 1.0604, "step": 646 }, { "epoch": 0.03, "learning_rate": 1.999957548368485e-05, "loss": 1.0429, "step": 647 }, { "epoch": 0.03, "learning_rate": 1.999956018713129e-05, "loss": 1.0375, "step": 648 }, { "epoch": 0.03, "learning_rate": 1.9999544619850272e-05, "loss": 1.0473, "step": 649 }, { "epoch": 0.03, "learning_rate": 1.9999528781842222e-05, "loss": 1.0167, "step": 650 }, { "epoch": 0.03, "learning_rate": 1.999951267310757e-05, "loss": 1.0052, "step": 651 }, { "epoch": 0.03, "learning_rate": 1.9999496293646753e-05, "loss": 1.0732, "step": 652 }, { "epoch": 0.03, "learning_rate": 1.999947964346021e-05, "loss": 1.1296, "step": 653 }, { "epoch": 0.03, "learning_rate": 1.9999462722548394e-05, "loss": 1.0555, "step": 654 }, { "epoch": 0.03, "learning_rate": 1.9999445530911764e-05, "loss": 1.104, "step": 655 }, { "epoch": 0.03, "learning_rate": 1.9999428068550783e-05, "loss": 1.1855, "step": 656 }, { "epoch": 0.03, "learning_rate": 1.999941033546593e-05, "loss": 1.0343, "step": 657 }, { "epoch": 0.03, "learning_rate": 1.9999392331657677e-05, "loss": 1.1121, "step": 658 }, { "epoch": 0.03, "learning_rate": 1.9999374057126517e-05, "loss": 1.0448, "step": 659 }, { "epoch": 0.03, "learning_rate": 1.999935551187294e-05, "loss": 1.1361, "step": 660 }, { "epoch": 0.03, "learning_rate": 1.9999336695897457e-05, "loss": 1.1285, "step": 661 }, { "epoch": 0.03, "learning_rate": 1.9999317609200566e-05, "loss": 0.9792, "step": 662 }, { "epoch": 0.03, "learning_rate": 1.999929825178279e-05, "loss": 1.082, "step": 663 }, { "epoch": 0.03, "learning_rate": 1.9999278623644653e-05, "loss": 1.0235, "step": 664 }, { "epoch": 0.03, "learning_rate": 1.9999258724786688e-05, "loss": 1.0855, "step": 665 }, { "epoch": 0.03, "learning_rate": 1.999923855520943e-05, "loss": 1.0322, "step": 666 }, { "epoch": 0.03, "learning_rate": 1.9999218114913422e-05, "loss": 1.0231, "step": 667 }, { "epoch": 0.03, "learning_rate": 1.9999197403899226e-05, "loss": 1.0885, "step": 668 }, { "epoch": 0.03, "learning_rate": 1.99991764221674e-05, "loss": 1.0966, "step": 669 }, { "epoch": 0.03, "learning_rate": 1.9999155169718508e-05, "loss": 1.0101, "step": 670 }, { "epoch": 0.03, "learning_rate": 1.9999133646553127e-05, "loss": 1.0902, "step": 671 }, { "epoch": 0.03, "learning_rate": 1.9999111852671845e-05, "loss": 1.0496, "step": 672 }, { "epoch": 0.03, "learning_rate": 1.9999089788075242e-05, "loss": 1.1601, "step": 673 }, { "epoch": 0.03, "learning_rate": 1.9999067452763924e-05, "loss": 1.0241, "step": 674 }, { "epoch": 0.03, "learning_rate": 1.9999044846738495e-05, "loss": 1.1183, "step": 675 }, { "epoch": 0.03, "learning_rate": 1.9999021969999563e-05, "loss": 1.0813, "step": 676 }, { "epoch": 0.03, "learning_rate": 1.999899882254775e-05, "loss": 1.1419, "step": 677 }, { "epoch": 0.03, "learning_rate": 1.999897540438368e-05, "loss": 1.1176, "step": 678 }, { "epoch": 0.03, "learning_rate": 1.9998951715507994e-05, "loss": 1.0597, "step": 679 }, { "epoch": 0.03, "learning_rate": 1.9998927755921326e-05, "loss": 1.0388, "step": 680 }, { "epoch": 0.03, "learning_rate": 1.9998903525624324e-05, "loss": 1.0578, "step": 681 }, { "epoch": 0.03, "learning_rate": 1.999887902461765e-05, "loss": 1.0809, "step": 682 }, { "epoch": 0.03, "learning_rate": 1.999885425290196e-05, "loss": 1.1603, "step": 683 }, { "epoch": 0.03, "learning_rate": 1.999882921047793e-05, "loss": 1.0912, "step": 684 }, { "epoch": 0.03, "learning_rate": 1.9998803897346244e-05, "loss": 1.0501, "step": 685 }, { "epoch": 0.03, "learning_rate": 1.9998778313507574e-05, "loss": 1.1069, "step": 686 }, { "epoch": 0.03, "learning_rate": 1.999875245896262e-05, "loss": 1.0606, "step": 687 }, { "epoch": 0.03, "learning_rate": 1.9998726333712086e-05, "loss": 1.0348, "step": 688 }, { "epoch": 0.04, "learning_rate": 1.9998699937756667e-05, "loss": 1.0958, "step": 689 }, { "epoch": 0.04, "learning_rate": 1.9998673271097088e-05, "loss": 1.1158, "step": 690 }, { "epoch": 0.04, "learning_rate": 1.999864633373407e-05, "loss": 1.1144, "step": 691 }, { "epoch": 0.04, "learning_rate": 1.9998619125668336e-05, "loss": 1.1094, "step": 692 }, { "epoch": 0.04, "learning_rate": 1.9998591646900628e-05, "loss": 1.0053, "step": 693 }, { "epoch": 0.04, "learning_rate": 1.9998563897431693e-05, "loss": 0.9926, "step": 694 }, { "epoch": 0.04, "learning_rate": 1.9998535877262277e-05, "loss": 1.1002, "step": 695 }, { "epoch": 0.04, "learning_rate": 1.9998507586393138e-05, "loss": 1.046, "step": 696 }, { "epoch": 0.04, "learning_rate": 1.999847902482504e-05, "loss": 1.0941, "step": 697 }, { "epoch": 0.04, "learning_rate": 1.9998450192558766e-05, "loss": 1.1006, "step": 698 }, { "epoch": 0.04, "learning_rate": 1.9998421089595085e-05, "loss": 1.1482, "step": 699 }, { "epoch": 0.04, "learning_rate": 1.9998391715934795e-05, "loss": 1.1059, "step": 700 }, { "epoch": 0.04, "learning_rate": 1.9998362071578686e-05, "loss": 1.1066, "step": 701 }, { "epoch": 0.04, "learning_rate": 1.999833215652756e-05, "loss": 1.0872, "step": 702 }, { "epoch": 0.04, "learning_rate": 1.999830197078223e-05, "loss": 1.0973, "step": 703 }, { "epoch": 0.04, "learning_rate": 1.9998271514343507e-05, "loss": 1.1526, "step": 704 }, { "epoch": 0.04, "learning_rate": 1.9998240787212225e-05, "loss": 1.0615, "step": 705 }, { "epoch": 0.04, "learning_rate": 1.999820978938921e-05, "loss": 0.9477, "step": 706 }, { "epoch": 0.04, "learning_rate": 1.9998178520875302e-05, "loss": 1.0416, "step": 707 }, { "epoch": 0.04, "learning_rate": 1.9998146981671348e-05, "loss": 1.0635, "step": 708 }, { "epoch": 0.04, "learning_rate": 1.9998115171778202e-05, "loss": 1.0493, "step": 709 }, { "epoch": 0.04, "learning_rate": 1.999808309119672e-05, "loss": 1.1091, "step": 710 }, { "epoch": 0.04, "learning_rate": 1.999805073992778e-05, "loss": 1.1707, "step": 711 }, { "epoch": 0.04, "learning_rate": 1.9998018117972252e-05, "loss": 1.2068, "step": 712 }, { "epoch": 0.04, "learning_rate": 1.999798522533102e-05, "loss": 1.1017, "step": 713 }, { "epoch": 0.04, "learning_rate": 1.999795206200498e-05, "loss": 1.1187, "step": 714 }, { "epoch": 0.04, "learning_rate": 1.9997918627995018e-05, "loss": 1.1184, "step": 715 }, { "epoch": 0.04, "learning_rate": 1.999788492330205e-05, "loss": 1.1137, "step": 716 }, { "epoch": 0.04, "learning_rate": 1.9997850947926986e-05, "loss": 1.0391, "step": 717 }, { "epoch": 0.04, "learning_rate": 1.9997816701870742e-05, "loss": 1.0974, "step": 718 }, { "epoch": 0.04, "learning_rate": 1.9997782185134247e-05, "loss": 1.0336, "step": 719 }, { "epoch": 0.04, "learning_rate": 1.9997747397718437e-05, "loss": 1.1675, "step": 720 }, { "epoch": 0.04, "learning_rate": 1.999771233962425e-05, "loss": 1.0303, "step": 721 }, { "epoch": 0.04, "learning_rate": 1.9997677010852644e-05, "loss": 1.0168, "step": 722 }, { "epoch": 0.04, "learning_rate": 1.999764141140457e-05, "loss": 1.1395, "step": 723 }, { "epoch": 0.04, "learning_rate": 1.9997605541280986e-05, "loss": 1.1162, "step": 724 }, { "epoch": 0.04, "learning_rate": 1.999756940048287e-05, "loss": 1.0731, "step": 725 }, { "epoch": 0.04, "learning_rate": 1.9997532989011197e-05, "loss": 1.1816, "step": 726 }, { "epoch": 0.04, "learning_rate": 1.999749630686696e-05, "loss": 1.0609, "step": 727 }, { "epoch": 0.04, "learning_rate": 1.9997459354051143e-05, "loss": 0.998, "step": 728 }, { "epoch": 0.04, "learning_rate": 1.999742213056475e-05, "loss": 1.1089, "step": 729 }, { "epoch": 0.04, "learning_rate": 1.9997384636408795e-05, "loss": 1.1702, "step": 730 }, { "epoch": 0.04, "learning_rate": 1.9997346871584284e-05, "loss": 1.0665, "step": 731 }, { "epoch": 0.04, "learning_rate": 1.999730883609224e-05, "loss": 1.0908, "step": 732 }, { "epoch": 0.04, "learning_rate": 1.99972705299337e-05, "loss": 1.0026, "step": 733 }, { "epoch": 0.04, "learning_rate": 1.9997231953109696e-05, "loss": 1.0616, "step": 734 }, { "epoch": 0.04, "learning_rate": 1.999719310562127e-05, "loss": 1.1093, "step": 735 }, { "epoch": 0.04, "learning_rate": 1.999715398746948e-05, "loss": 1.0575, "step": 736 }, { "epoch": 0.04, "learning_rate": 1.999711459865538e-05, "loss": 1.103, "step": 737 }, { "epoch": 0.04, "learning_rate": 1.999707493918004e-05, "loss": 1.0872, "step": 738 }, { "epoch": 0.04, "learning_rate": 1.999703500904453e-05, "loss": 1.0217, "step": 739 }, { "epoch": 0.04, "learning_rate": 1.9996994808249933e-05, "loss": 1.1985, "step": 740 }, { "epoch": 0.04, "learning_rate": 1.999695433679734e-05, "loss": 1.0361, "step": 741 }, { "epoch": 0.04, "learning_rate": 1.999691359468784e-05, "loss": 1.0311, "step": 742 }, { "epoch": 0.04, "learning_rate": 1.9996872581922546e-05, "loss": 1.0684, "step": 743 }, { "epoch": 0.04, "learning_rate": 1.999683129850256e-05, "loss": 0.9824, "step": 744 }, { "epoch": 0.04, "learning_rate": 1.9996789744429e-05, "loss": 0.9875, "step": 745 }, { "epoch": 0.04, "learning_rate": 1.9996747919703e-05, "loss": 0.9984, "step": 746 }, { "epoch": 0.04, "learning_rate": 1.9996705824325676e-05, "loss": 1.0088, "step": 747 }, { "epoch": 0.04, "learning_rate": 1.9996663458298187e-05, "loss": 1.0873, "step": 748 }, { "epoch": 0.04, "learning_rate": 1.9996620821621663e-05, "loss": 1.1546, "step": 749 }, { "epoch": 0.04, "learning_rate": 1.999657791429727e-05, "loss": 1.0588, "step": 750 }, { "epoch": 0.04, "learning_rate": 1.999653473632617e-05, "loss": 1.1655, "step": 751 }, { "epoch": 0.04, "learning_rate": 1.999649128770952e-05, "loss": 1.1348, "step": 752 }, { "epoch": 0.04, "learning_rate": 1.9996447568448506e-05, "loss": 1.1124, "step": 753 }, { "epoch": 0.04, "learning_rate": 1.9996403578544308e-05, "loss": 1.0258, "step": 754 }, { "epoch": 0.04, "learning_rate": 1.999635931799812e-05, "loss": 1.0685, "step": 755 }, { "epoch": 0.04, "learning_rate": 1.999631478681114e-05, "loss": 1.0363, "step": 756 }, { "epoch": 0.04, "learning_rate": 1.999626998498457e-05, "loss": 1.0664, "step": 757 }, { "epoch": 0.04, "learning_rate": 1.9996224912519625e-05, "loss": 1.0998, "step": 758 }, { "epoch": 0.04, "learning_rate": 1.999617956941753e-05, "loss": 1.0458, "step": 759 }, { "epoch": 0.04, "learning_rate": 1.9996133955679502e-05, "loss": 1.1745, "step": 760 }, { "epoch": 0.04, "learning_rate": 1.9996088071306788e-05, "loss": 1.116, "step": 761 }, { "epoch": 0.04, "learning_rate": 1.999604191630062e-05, "loss": 0.9971, "step": 762 }, { "epoch": 0.04, "learning_rate": 1.9995995490662255e-05, "loss": 1.0526, "step": 763 }, { "epoch": 0.04, "learning_rate": 1.9995948794392944e-05, "loss": 1.0938, "step": 764 }, { "epoch": 0.04, "learning_rate": 1.9995901827493956e-05, "loss": 1.0703, "step": 765 }, { "epoch": 0.04, "learning_rate": 1.9995854589966563e-05, "loss": 1.0413, "step": 766 }, { "epoch": 0.04, "learning_rate": 1.999580708181204e-05, "loss": 1.066, "step": 767 }, { "epoch": 0.04, "learning_rate": 1.9995759303031675e-05, "loss": 1.1584, "step": 768 }, { "epoch": 0.04, "learning_rate": 1.9995711253626763e-05, "loss": 1.2044, "step": 769 }, { "epoch": 0.04, "learning_rate": 1.9995662933598598e-05, "loss": 1.0873, "step": 770 }, { "epoch": 0.04, "learning_rate": 1.99956143429485e-05, "loss": 1.0762, "step": 771 }, { "epoch": 0.04, "learning_rate": 1.9995565481677777e-05, "loss": 1.2427, "step": 772 }, { "epoch": 0.04, "learning_rate": 1.999551634978775e-05, "loss": 1.0964, "step": 773 }, { "epoch": 0.04, "learning_rate": 1.9995466947279753e-05, "loss": 1.0348, "step": 774 }, { "epoch": 0.04, "learning_rate": 1.9995417274155124e-05, "loss": 1.0527, "step": 775 }, { "epoch": 0.04, "learning_rate": 1.9995367330415204e-05, "loss": 1.0563, "step": 776 }, { "epoch": 0.04, "learning_rate": 1.9995317116061354e-05, "loss": 1.0819, "step": 777 }, { "epoch": 0.04, "learning_rate": 1.999526663109492e-05, "loss": 1.0904, "step": 778 }, { "epoch": 0.04, "learning_rate": 1.999521587551728e-05, "loss": 1.0733, "step": 779 }, { "epoch": 0.04, "learning_rate": 1.9995164849329805e-05, "loss": 1.1171, "step": 780 }, { "epoch": 0.04, "learning_rate": 1.9995113552533872e-05, "loss": 1.0212, "step": 781 }, { "epoch": 0.04, "learning_rate": 1.9995061985130876e-05, "loss": 1.1714, "step": 782 }, { "epoch": 0.04, "learning_rate": 1.999501014712221e-05, "loss": 1.1565, "step": 783 }, { "epoch": 0.04, "learning_rate": 1.999495803850928e-05, "loss": 1.0276, "step": 784 }, { "epoch": 0.04, "learning_rate": 1.9994905659293493e-05, "loss": 1.1122, "step": 785 }, { "epoch": 0.04, "learning_rate": 1.9994853009476268e-05, "loss": 1.0779, "step": 786 }, { "epoch": 0.04, "learning_rate": 1.9994800089059033e-05, "loss": 1.1192, "step": 787 }, { "epoch": 0.04, "learning_rate": 1.999474689804322e-05, "loss": 1.0533, "step": 788 }, { "epoch": 0.04, "learning_rate": 1.9994693436430268e-05, "loss": 1.0598, "step": 789 }, { "epoch": 0.04, "learning_rate": 1.9994639704221624e-05, "loss": 1.1196, "step": 790 }, { "epoch": 0.04, "learning_rate": 1.9994585701418746e-05, "loss": 1.0696, "step": 791 }, { "epoch": 0.04, "learning_rate": 1.9994531428023087e-05, "loss": 1.1039, "step": 792 }, { "epoch": 0.04, "learning_rate": 1.9994476884036128e-05, "loss": 1.147, "step": 793 }, { "epoch": 0.04, "learning_rate": 1.999442206945934e-05, "loss": 0.9722, "step": 794 }, { "epoch": 0.04, "learning_rate": 1.999436698429421e-05, "loss": 1.0412, "step": 795 }, { "epoch": 0.04, "learning_rate": 1.999431162854222e-05, "loss": 0.9798, "step": 796 }, { "epoch": 0.04, "learning_rate": 1.999425600220488e-05, "loss": 1.0438, "step": 797 }, { "epoch": 0.04, "learning_rate": 1.9994200105283694e-05, "loss": 1.0837, "step": 798 }, { "epoch": 0.04, "learning_rate": 1.9994143937780174e-05, "loss": 1.0518, "step": 799 }, { "epoch": 0.04, "learning_rate": 1.9994087499695832e-05, "loss": 1.1469, "step": 800 }, { "epoch": 0.04, "learning_rate": 1.9994030791032212e-05, "loss": 1.0536, "step": 801 }, { "epoch": 0.04, "learning_rate": 1.9993973811790835e-05, "loss": 1.0589, "step": 802 }, { "epoch": 0.04, "learning_rate": 1.9993916561973257e-05, "loss": 0.9896, "step": 803 }, { "epoch": 0.04, "learning_rate": 1.9993859041581016e-05, "loss": 1.0801, "step": 804 }, { "epoch": 0.04, "learning_rate": 1.9993801250615674e-05, "loss": 1.1248, "step": 805 }, { "epoch": 0.04, "learning_rate": 1.9993743189078796e-05, "loss": 1.0809, "step": 806 }, { "epoch": 0.04, "learning_rate": 1.9993684856971955e-05, "loss": 1.1277, "step": 807 }, { "epoch": 0.04, "learning_rate": 1.999362625429673e-05, "loss": 1.0825, "step": 808 }, { "epoch": 0.04, "learning_rate": 1.9993567381054703e-05, "loss": 1.1209, "step": 809 }, { "epoch": 0.04, "learning_rate": 1.999350823724747e-05, "loss": 1.0556, "step": 810 }, { "epoch": 0.04, "learning_rate": 1.9993448822876637e-05, "loss": 1.1036, "step": 811 }, { "epoch": 0.04, "learning_rate": 1.9993389137943812e-05, "loss": 1.05, "step": 812 }, { "epoch": 0.04, "learning_rate": 1.9993329182450602e-05, "loss": 0.9944, "step": 813 }, { "epoch": 0.04, "learning_rate": 1.9993268956398643e-05, "loss": 1.1144, "step": 814 }, { "epoch": 0.04, "learning_rate": 1.9993208459789554e-05, "loss": 1.0895, "step": 815 }, { "epoch": 0.04, "learning_rate": 1.999314769262498e-05, "loss": 1.0733, "step": 816 }, { "epoch": 0.04, "learning_rate": 1.999308665490656e-05, "loss": 1.1188, "step": 817 }, { "epoch": 0.04, "learning_rate": 1.9993025346635957e-05, "loss": 1.0663, "step": 818 }, { "epoch": 0.04, "learning_rate": 1.9992963767814823e-05, "loss": 1.0271, "step": 819 }, { "epoch": 0.04, "learning_rate": 1.9992901918444824e-05, "loss": 1.1093, "step": 820 }, { "epoch": 0.04, "learning_rate": 1.999283979852764e-05, "loss": 1.0186, "step": 821 }, { "epoch": 0.04, "learning_rate": 1.9992777408064946e-05, "loss": 1.0834, "step": 822 }, { "epoch": 0.04, "learning_rate": 1.999271474705844e-05, "loss": 1.1035, "step": 823 }, { "epoch": 0.04, "learning_rate": 1.9992651815509808e-05, "loss": 0.9177, "step": 824 }, { "epoch": 0.04, "learning_rate": 1.999258861342077e-05, "loss": 1.1384, "step": 825 }, { "epoch": 0.04, "learning_rate": 1.9992525140793014e-05, "loss": 1.0898, "step": 826 }, { "epoch": 0.04, "learning_rate": 1.9992461397628276e-05, "loss": 1.1021, "step": 827 }, { "epoch": 0.04, "learning_rate": 1.999239738392828e-05, "loss": 1.1097, "step": 828 }, { "epoch": 0.04, "learning_rate": 1.9992333099694753e-05, "loss": 0.9693, "step": 829 }, { "epoch": 0.04, "learning_rate": 1.999226854492944e-05, "loss": 1.1628, "step": 830 }, { "epoch": 0.04, "learning_rate": 1.9992203719634082e-05, "loss": 1.1608, "step": 831 }, { "epoch": 0.04, "learning_rate": 1.9992138623810443e-05, "loss": 1.0457, "step": 832 }, { "epoch": 0.04, "learning_rate": 1.9992073257460283e-05, "loss": 1.0998, "step": 833 }, { "epoch": 0.04, "learning_rate": 1.9992007620585367e-05, "loss": 0.9558, "step": 834 }, { "epoch": 0.04, "learning_rate": 1.9991941713187477e-05, "loss": 1.0634, "step": 835 }, { "epoch": 0.04, "learning_rate": 1.9991875535268396e-05, "loss": 1.0879, "step": 836 }, { "epoch": 0.04, "learning_rate": 1.9991809086829915e-05, "loss": 1.0361, "step": 837 }, { "epoch": 0.04, "learning_rate": 1.999174236787383e-05, "loss": 1.0549, "step": 838 }, { "epoch": 0.04, "learning_rate": 1.9991675378401956e-05, "loss": 1.1716, "step": 839 }, { "epoch": 0.04, "learning_rate": 1.99916081184161e-05, "loss": 1.0437, "step": 840 }, { "epoch": 0.04, "learning_rate": 1.9991540587918084e-05, "loss": 1.0698, "step": 841 }, { "epoch": 0.04, "learning_rate": 1.999147278690974e-05, "loss": 1.1247, "step": 842 }, { "epoch": 0.04, "learning_rate": 1.9991404715392895e-05, "loss": 1.1314, "step": 843 }, { "epoch": 0.04, "learning_rate": 1.99913363733694e-05, "loss": 1.0195, "step": 844 }, { "epoch": 0.04, "learning_rate": 1.99912677608411e-05, "loss": 1.0162, "step": 845 }, { "epoch": 0.04, "learning_rate": 1.999119887780986e-05, "loss": 1.0893, "step": 846 }, { "epoch": 0.04, "learning_rate": 1.9991129724277536e-05, "loss": 1.0427, "step": 847 }, { "epoch": 0.04, "learning_rate": 1.9991060300246007e-05, "loss": 1.1052, "step": 848 }, { "epoch": 0.04, "learning_rate": 1.9990990605717148e-05, "loss": 1.173, "step": 849 }, { "epoch": 0.04, "learning_rate": 1.9990920640692848e-05, "loss": 1.0124, "step": 850 }, { "epoch": 0.04, "learning_rate": 1.9990850405175e-05, "loss": 1.093, "step": 851 }, { "epoch": 0.04, "learning_rate": 1.9990779899165512e-05, "loss": 1.1827, "step": 852 }, { "epoch": 0.04, "learning_rate": 1.9990709122666284e-05, "loss": 0.9049, "step": 853 }, { "epoch": 0.04, "learning_rate": 1.9990638075679238e-05, "loss": 1.069, "step": 854 }, { "epoch": 0.04, "learning_rate": 1.9990566758206294e-05, "loss": 0.998, "step": 855 }, { "epoch": 0.04, "learning_rate": 1.999049517024938e-05, "loss": 1.0581, "step": 856 }, { "epoch": 0.04, "learning_rate": 1.9990423311810443e-05, "loss": 0.9783, "step": 857 }, { "epoch": 0.04, "learning_rate": 1.9990351182891422e-05, "loss": 1.0678, "step": 858 }, { "epoch": 0.04, "learning_rate": 1.9990278783494276e-05, "loss": 1.0731, "step": 859 }, { "epoch": 0.04, "learning_rate": 1.9990206113620956e-05, "loss": 1.1545, "step": 860 }, { "epoch": 0.04, "learning_rate": 1.9990133173273434e-05, "loss": 1.1487, "step": 861 }, { "epoch": 0.04, "learning_rate": 1.9990059962453687e-05, "loss": 1.0412, "step": 862 }, { "epoch": 0.04, "learning_rate": 1.9989986481163695e-05, "loss": 1.09, "step": 863 }, { "epoch": 0.04, "learning_rate": 1.9989912729405447e-05, "loss": 1.1138, "step": 864 }, { "epoch": 0.04, "learning_rate": 1.998983870718094e-05, "loss": 1.0358, "step": 865 }, { "epoch": 0.04, "learning_rate": 1.998976441449218e-05, "loss": 1.0919, "step": 866 }, { "epoch": 0.04, "learning_rate": 1.9989689851341173e-05, "loss": 1.0898, "step": 867 }, { "epoch": 0.04, "learning_rate": 1.9989615017729945e-05, "loss": 0.9461, "step": 868 }, { "epoch": 0.04, "learning_rate": 1.9989539913660517e-05, "loss": 1.0791, "step": 869 }, { "epoch": 0.04, "learning_rate": 1.9989464539134923e-05, "loss": 1.056, "step": 870 }, { "epoch": 0.04, "learning_rate": 1.998938889415521e-05, "loss": 1.1399, "step": 871 }, { "epoch": 0.04, "learning_rate": 1.9989312978723414e-05, "loss": 1.04, "step": 872 }, { "epoch": 0.04, "learning_rate": 1.99892367928416e-05, "loss": 1.0809, "step": 873 }, { "epoch": 0.04, "learning_rate": 1.9989160336511827e-05, "loss": 1.1028, "step": 874 }, { "epoch": 0.04, "learning_rate": 1.9989083609736164e-05, "loss": 1.0435, "step": 875 }, { "epoch": 0.04, "learning_rate": 1.998900661251669e-05, "loss": 1.0845, "step": 876 }, { "epoch": 0.04, "learning_rate": 1.9988929344855492e-05, "loss": 1.0586, "step": 877 }, { "epoch": 0.04, "learning_rate": 1.9988851806754657e-05, "loss": 1.0818, "step": 878 }, { "epoch": 0.04, "learning_rate": 1.9988773998216287e-05, "loss": 1.0147, "step": 879 }, { "epoch": 0.04, "learning_rate": 1.998869591924249e-05, "loss": 1.088, "step": 880 }, { "epoch": 0.04, "learning_rate": 1.9988617569835374e-05, "loss": 1.0806, "step": 881 }, { "epoch": 0.04, "learning_rate": 1.998853894999707e-05, "loss": 1.0923, "step": 882 }, { "epoch": 0.04, "learning_rate": 1.9988460059729698e-05, "loss": 1.0949, "step": 883 }, { "epoch": 0.04, "learning_rate": 1.9988380899035395e-05, "loss": 1.0878, "step": 884 }, { "epoch": 0.04, "learning_rate": 1.998830146791631e-05, "loss": 1.0624, "step": 885 }, { "epoch": 0.05, "learning_rate": 1.9988221766374587e-05, "loss": 1.1375, "step": 886 }, { "epoch": 0.05, "learning_rate": 1.998814179441239e-05, "loss": 1.04, "step": 887 }, { "epoch": 0.05, "learning_rate": 1.9988061552031876e-05, "loss": 1.0482, "step": 888 }, { "epoch": 0.05, "learning_rate": 1.9987981039235224e-05, "loss": 1.0831, "step": 889 }, { "epoch": 0.05, "learning_rate": 1.9987900256024608e-05, "loss": 1.0695, "step": 890 }, { "epoch": 0.05, "learning_rate": 1.998781920240222e-05, "loss": 1.0441, "step": 891 }, { "epoch": 0.05, "learning_rate": 1.9987737878370256e-05, "loss": 1.1785, "step": 892 }, { "epoch": 0.05, "learning_rate": 1.9987656283930914e-05, "loss": 1.0328, "step": 893 }, { "epoch": 0.05, "learning_rate": 1.9987574419086404e-05, "loss": 0.9986, "step": 894 }, { "epoch": 0.05, "learning_rate": 1.9987492283838944e-05, "loss": 1.062, "step": 895 }, { "epoch": 0.05, "learning_rate": 1.9987409878190752e-05, "loss": 1.1647, "step": 896 }, { "epoch": 0.05, "learning_rate": 1.998732720214407e-05, "loss": 1.0333, "step": 897 }, { "epoch": 0.05, "learning_rate": 1.9987244255701125e-05, "loss": 1.0323, "step": 898 }, { "epoch": 0.05, "learning_rate": 1.998716103886417e-05, "loss": 1.1173, "step": 899 }, { "epoch": 0.05, "learning_rate": 1.9987077551635453e-05, "loss": 1.0588, "step": 900 }, { "epoch": 0.05, "learning_rate": 1.9986993794017238e-05, "loss": 1.1291, "step": 901 }, { "epoch": 0.05, "learning_rate": 1.998690976601179e-05, "loss": 1.0656, "step": 902 }, { "epoch": 0.05, "learning_rate": 1.9986825467621386e-05, "loss": 1.0571, "step": 903 }, { "epoch": 0.05, "learning_rate": 1.9986740898848306e-05, "loss": 1.0539, "step": 904 }, { "epoch": 0.05, "learning_rate": 1.9986656059694846e-05, "loss": 1.0499, "step": 905 }, { "epoch": 0.05, "learning_rate": 1.9986570950163292e-05, "loss": 1.0912, "step": 906 }, { "epoch": 0.05, "learning_rate": 1.9986485570255958e-05, "loss": 1.0513, "step": 907 }, { "epoch": 0.05, "learning_rate": 1.9986399919975152e-05, "loss": 1.0336, "step": 908 }, { "epoch": 0.05, "learning_rate": 1.9986313999323194e-05, "loss": 1.1027, "step": 909 }, { "epoch": 0.05, "learning_rate": 1.9986227808302405e-05, "loss": 1.0446, "step": 910 }, { "epoch": 0.05, "learning_rate": 1.9986141346915127e-05, "loss": 1.0591, "step": 911 }, { "epoch": 0.05, "learning_rate": 1.998605461516369e-05, "loss": 1.0035, "step": 912 }, { "epoch": 0.05, "learning_rate": 1.9985967613050454e-05, "loss": 1.1802, "step": 913 }, { "epoch": 0.05, "learning_rate": 1.998588034057777e-05, "loss": 1.1063, "step": 914 }, { "epoch": 0.05, "learning_rate": 1.9985792797748e-05, "loss": 1.0193, "step": 915 }, { "epoch": 0.05, "learning_rate": 1.998570498456351e-05, "loss": 1.162, "step": 916 }, { "epoch": 0.05, "learning_rate": 1.9985616901026684e-05, "loss": 0.9886, "step": 917 }, { "epoch": 0.05, "learning_rate": 1.9985528547139906e-05, "loss": 1.0716, "step": 918 }, { "epoch": 0.05, "learning_rate": 1.9985439922905564e-05, "loss": 1.0498, "step": 919 }, { "epoch": 0.05, "learning_rate": 1.9985351028326055e-05, "loss": 1.0226, "step": 920 }, { "epoch": 0.05, "learning_rate": 1.9985261863403797e-05, "loss": 1.064, "step": 921 }, { "epoch": 0.05, "learning_rate": 1.99851724281412e-05, "loss": 1.1146, "step": 922 }, { "epoch": 0.05, "learning_rate": 1.9985082722540676e-05, "loss": 1.0739, "step": 923 }, { "epoch": 0.05, "learning_rate": 1.9984992746604663e-05, "loss": 1.1373, "step": 924 }, { "epoch": 0.05, "learning_rate": 1.9984902500335597e-05, "loss": 1.1937, "step": 925 }, { "epoch": 0.05, "learning_rate": 1.9984811983735915e-05, "loss": 1.1038, "step": 926 }, { "epoch": 0.05, "learning_rate": 1.9984721196808074e-05, "loss": 1.0383, "step": 927 }, { "epoch": 0.05, "learning_rate": 1.998463013955453e-05, "loss": 0.994, "step": 928 }, { "epoch": 0.05, "learning_rate": 1.9984538811977745e-05, "loss": 1.0839, "step": 929 }, { "epoch": 0.05, "learning_rate": 1.9984447214080198e-05, "loss": 1.1564, "step": 930 }, { "epoch": 0.05, "learning_rate": 1.9984355345864364e-05, "loss": 1.0356, "step": 931 }, { "epoch": 0.05, "learning_rate": 1.998426320733273e-05, "loss": 1.0726, "step": 932 }, { "epoch": 0.05, "learning_rate": 1.998417079848779e-05, "loss": 1.0859, "step": 933 }, { "epoch": 0.05, "learning_rate": 1.9984078119332055e-05, "loss": 1.0482, "step": 934 }, { "epoch": 0.05, "learning_rate": 1.9983985169868022e-05, "loss": 1.0632, "step": 935 }, { "epoch": 0.05, "learning_rate": 1.9983891950098214e-05, "loss": 1.082, "step": 936 }, { "epoch": 0.05, "learning_rate": 1.9983798460025153e-05, "loss": 1.0482, "step": 937 }, { "epoch": 0.05, "learning_rate": 1.9983704699651373e-05, "loss": 1.1338, "step": 938 }, { "epoch": 0.05, "learning_rate": 1.998361066897941e-05, "loss": 0.9652, "step": 939 }, { "epoch": 0.05, "learning_rate": 1.9983516368011804e-05, "loss": 0.9839, "step": 940 }, { "epoch": 0.05, "learning_rate": 1.998342179675112e-05, "loss": 1.1134, "step": 941 }, { "epoch": 0.05, "learning_rate": 1.9983326955199905e-05, "loss": 1.1529, "step": 942 }, { "epoch": 0.05, "learning_rate": 1.998323184336074e-05, "loss": 1.0514, "step": 943 }, { "epoch": 0.05, "learning_rate": 1.9983136461236196e-05, "loss": 1.192, "step": 944 }, { "epoch": 0.05, "learning_rate": 1.9983040808828852e-05, "loss": 1.0784, "step": 945 }, { "epoch": 0.05, "learning_rate": 1.9982944886141297e-05, "loss": 1.0179, "step": 946 }, { "epoch": 0.05, "learning_rate": 1.998284869317613e-05, "loss": 1.1181, "step": 947 }, { "epoch": 0.05, "learning_rate": 1.9982752229935957e-05, "loss": 0.9899, "step": 948 }, { "epoch": 0.05, "learning_rate": 1.998265549642339e-05, "loss": 1.0734, "step": 949 }, { "epoch": 0.05, "learning_rate": 1.9982558492641043e-05, "loss": 1.1393, "step": 950 }, { "epoch": 0.05, "learning_rate": 1.998246121859155e-05, "loss": 1.0129, "step": 951 }, { "epoch": 0.05, "learning_rate": 1.9982363674277538e-05, "loss": 1.0548, "step": 952 }, { "epoch": 0.05, "learning_rate": 1.9982265859701644e-05, "loss": 1.0804, "step": 953 }, { "epoch": 0.05, "learning_rate": 1.998216777486653e-05, "loss": 1.069, "step": 954 }, { "epoch": 0.05, "learning_rate": 1.998206941977484e-05, "loss": 1.1191, "step": 955 }, { "epoch": 0.05, "learning_rate": 1.998197079442924e-05, "loss": 1.0951, "step": 956 }, { "epoch": 0.05, "learning_rate": 1.99818718988324e-05, "loss": 1.1014, "step": 957 }, { "epoch": 0.05, "learning_rate": 1.9981772732987e-05, "loss": 1.0106, "step": 958 }, { "epoch": 0.05, "learning_rate": 1.9981673296895726e-05, "loss": 0.9753, "step": 959 }, { "epoch": 0.05, "learning_rate": 1.9981573590561262e-05, "loss": 1.0035, "step": 960 }, { "epoch": 0.05, "learning_rate": 1.9981473613986314e-05, "loss": 1.1234, "step": 961 }, { "epoch": 0.05, "learning_rate": 1.9981373367173588e-05, "loss": 0.9963, "step": 962 }, { "epoch": 0.05, "learning_rate": 1.9981272850125798e-05, "loss": 0.9964, "step": 963 }, { "epoch": 0.05, "learning_rate": 1.9981172062845665e-05, "loss": 0.9917, "step": 964 }, { "epoch": 0.05, "learning_rate": 1.998107100533592e-05, "loss": 1.1981, "step": 965 }, { "epoch": 0.05, "learning_rate": 1.9980969677599293e-05, "loss": 1.0596, "step": 966 }, { "epoch": 0.05, "learning_rate": 1.998086807963853e-05, "loss": 1.1217, "step": 967 }, { "epoch": 0.05, "learning_rate": 1.9980766211456385e-05, "loss": 1.141, "step": 968 }, { "epoch": 0.05, "learning_rate": 1.998066407305561e-05, "loss": 0.9832, "step": 969 }, { "epoch": 0.05, "learning_rate": 1.9980561664438976e-05, "loss": 1.1206, "step": 970 }, { "epoch": 0.05, "learning_rate": 1.9980458985609253e-05, "loss": 0.9806, "step": 971 }, { "epoch": 0.05, "learning_rate": 1.9980356036569223e-05, "loss": 1.0676, "step": 972 }, { "epoch": 0.05, "learning_rate": 1.9980252817321672e-05, "loss": 1.1203, "step": 973 }, { "epoch": 0.05, "learning_rate": 1.9980149327869394e-05, "loss": 1.0929, "step": 974 }, { "epoch": 0.05, "learning_rate": 1.9980045568215188e-05, "loss": 1.124, "step": 975 }, { "epoch": 0.05, "learning_rate": 1.997994153836187e-05, "loss": 1.0673, "step": 976 }, { "epoch": 0.05, "learning_rate": 1.9979837238312248e-05, "loss": 1.0654, "step": 977 }, { "epoch": 0.05, "learning_rate": 1.9979732668069155e-05, "loss": 1.0917, "step": 978 }, { "epoch": 0.05, "learning_rate": 1.9979627827635417e-05, "loss": 1.0363, "step": 979 }, { "epoch": 0.05, "learning_rate": 1.997952271701387e-05, "loss": 0.9998, "step": 980 }, { "epoch": 0.05, "learning_rate": 1.9979417336207364e-05, "loss": 1.102, "step": 981 }, { "epoch": 0.05, "learning_rate": 1.997931168521875e-05, "loss": 1.0447, "step": 982 }, { "epoch": 0.05, "learning_rate": 1.997920576405089e-05, "loss": 0.9732, "step": 983 }, { "epoch": 0.05, "learning_rate": 1.9979099572706654e-05, "loss": 1.1505, "step": 984 }, { "epoch": 0.05, "learning_rate": 1.997899311118891e-05, "loss": 1.0627, "step": 985 }, { "epoch": 0.05, "learning_rate": 1.9978886379500547e-05, "loss": 1.1428, "step": 986 }, { "epoch": 0.05, "learning_rate": 1.997877937764445e-05, "loss": 1.0733, "step": 987 }, { "epoch": 0.05, "learning_rate": 1.997867210562352e-05, "loss": 1.1234, "step": 988 }, { "epoch": 0.05, "learning_rate": 1.9978564563440655e-05, "loss": 1.126, "step": 989 }, { "epoch": 0.05, "learning_rate": 1.9978456751098776e-05, "loss": 1.0282, "step": 990 }, { "epoch": 0.05, "learning_rate": 1.9978348668600794e-05, "loss": 1.0585, "step": 991 }, { "epoch": 0.05, "learning_rate": 1.997824031594964e-05, "loss": 1.1229, "step": 992 }, { "epoch": 0.05, "learning_rate": 1.9978131693148242e-05, "loss": 1.1066, "step": 993 }, { "epoch": 0.05, "learning_rate": 1.9978022800199545e-05, "loss": 1.1466, "step": 994 }, { "epoch": 0.05, "learning_rate": 1.99779136371065e-05, "loss": 1.0903, "step": 995 }, { "epoch": 0.05, "learning_rate": 1.9977804203872056e-05, "loss": 1.0495, "step": 996 }, { "epoch": 0.05, "learning_rate": 1.9977694500499182e-05, "loss": 1.1046, "step": 997 }, { "epoch": 0.05, "learning_rate": 1.997758452699084e-05, "loss": 0.9282, "step": 998 }, { "epoch": 0.05, "learning_rate": 1.9977474283350012e-05, "loss": 1.1219, "step": 999 }, { "epoch": 0.05, "learning_rate": 1.9977363769579687e-05, "loss": 1.036, "step": 1000 }, { "epoch": 0.05, "learning_rate": 1.9977252985682854e-05, "loss": 1.0847, "step": 1001 }, { "epoch": 0.05, "learning_rate": 1.997714193166251e-05, "loss": 1.0497, "step": 1002 }, { "epoch": 0.05, "learning_rate": 1.997703060752166e-05, "loss": 0.961, "step": 1003 }, { "epoch": 0.05, "learning_rate": 1.9976919013263323e-05, "loss": 1.0059, "step": 1004 }, { "epoch": 0.05, "learning_rate": 1.9976807148890517e-05, "loss": 1.1633, "step": 1005 }, { "epoch": 0.05, "learning_rate": 1.9976695014406276e-05, "loss": 1.0706, "step": 1006 }, { "epoch": 0.05, "learning_rate": 1.9976582609813627e-05, "loss": 1.0645, "step": 1007 }, { "epoch": 0.05, "learning_rate": 1.997646993511562e-05, "loss": 0.9919, "step": 1008 }, { "epoch": 0.05, "learning_rate": 1.9976356990315303e-05, "loss": 1.1796, "step": 1009 }, { "epoch": 0.05, "learning_rate": 1.9976243775415737e-05, "loss": 1.1318, "step": 1010 }, { "epoch": 0.05, "learning_rate": 1.9976130290419983e-05, "loss": 1.043, "step": 1011 }, { "epoch": 0.05, "learning_rate": 1.9976016535331115e-05, "loss": 1.1196, "step": 1012 }, { "epoch": 0.05, "learning_rate": 1.997590251015221e-05, "loss": 1.0776, "step": 1013 }, { "epoch": 0.05, "learning_rate": 1.9975788214886363e-05, "loss": 1.0541, "step": 1014 }, { "epoch": 0.05, "learning_rate": 1.9975673649536664e-05, "loss": 1.2174, "step": 1015 }, { "epoch": 0.05, "learning_rate": 1.997555881410621e-05, "loss": 1.0857, "step": 1016 }, { "epoch": 0.05, "learning_rate": 1.9975443708598114e-05, "loss": 1.1106, "step": 1017 }, { "epoch": 0.05, "learning_rate": 1.9975328333015497e-05, "loss": 1.048, "step": 1018 }, { "epoch": 0.05, "learning_rate": 1.9975212687361474e-05, "loss": 1.0577, "step": 1019 }, { "epoch": 0.05, "learning_rate": 1.9975096771639185e-05, "loss": 0.9987, "step": 1020 }, { "epoch": 0.05, "learning_rate": 1.997498058585176e-05, "loss": 1.0855, "step": 1021 }, { "epoch": 0.05, "learning_rate": 1.9974864130002348e-05, "loss": 1.0559, "step": 1022 }, { "epoch": 0.05, "learning_rate": 1.9974747404094102e-05, "loss": 1.1, "step": 1023 }, { "epoch": 0.05, "learning_rate": 1.9974630408130187e-05, "loss": 1.022, "step": 1024 }, { "epoch": 0.05, "learning_rate": 1.997451314211376e-05, "loss": 1.0677, "step": 1025 }, { "epoch": 0.05, "learning_rate": 1.9974395606048004e-05, "loss": 1.0186, "step": 1026 }, { "epoch": 0.05, "learning_rate": 1.99742777999361e-05, "loss": 1.1408, "step": 1027 }, { "epoch": 0.05, "learning_rate": 1.9974159723781234e-05, "loss": 1.1599, "step": 1028 }, { "epoch": 0.05, "learning_rate": 1.9974041377586604e-05, "loss": 1.1699, "step": 1029 }, { "epoch": 0.05, "learning_rate": 1.997392276135542e-05, "loss": 1.0907, "step": 1030 }, { "epoch": 0.05, "learning_rate": 1.9973803875090886e-05, "loss": 1.0344, "step": 1031 }, { "epoch": 0.05, "learning_rate": 1.9973684718796226e-05, "loss": 1.1077, "step": 1032 }, { "epoch": 0.05, "learning_rate": 1.9973565292474658e-05, "loss": 1.0234, "step": 1033 }, { "epoch": 0.05, "learning_rate": 1.9973445596129426e-05, "loss": 1.1636, "step": 1034 }, { "epoch": 0.05, "learning_rate": 1.9973325629763764e-05, "loss": 1.0342, "step": 1035 }, { "epoch": 0.05, "learning_rate": 1.997320539338092e-05, "loss": 1.1443, "step": 1036 }, { "epoch": 0.05, "learning_rate": 1.997308488698415e-05, "loss": 1.198, "step": 1037 }, { "epoch": 0.05, "learning_rate": 1.997296411057672e-05, "loss": 1.0621, "step": 1038 }, { "epoch": 0.05, "learning_rate": 1.9972843064161896e-05, "loss": 1.0769, "step": 1039 }, { "epoch": 0.05, "learning_rate": 1.9972721747742953e-05, "loss": 1.1848, "step": 1040 }, { "epoch": 0.05, "learning_rate": 1.9972600161323184e-05, "loss": 1.1584, "step": 1041 }, { "epoch": 0.05, "learning_rate": 1.9972478304905873e-05, "loss": 1.0541, "step": 1042 }, { "epoch": 0.05, "learning_rate": 1.997235617849432e-05, "loss": 1.018, "step": 1043 }, { "epoch": 0.05, "learning_rate": 1.9972233782091838e-05, "loss": 0.9889, "step": 1044 }, { "epoch": 0.05, "learning_rate": 1.9972111115701734e-05, "loss": 1.0528, "step": 1045 }, { "epoch": 0.05, "learning_rate": 1.997198817932733e-05, "loss": 1.0794, "step": 1046 }, { "epoch": 0.05, "learning_rate": 1.9971864972971956e-05, "loss": 1.0971, "step": 1047 }, { "epoch": 0.05, "learning_rate": 1.9971741496638947e-05, "loss": 1.0524, "step": 1048 }, { "epoch": 0.05, "learning_rate": 1.9971617750331644e-05, "loss": 1.1306, "step": 1049 }, { "epoch": 0.05, "learning_rate": 1.9971493734053403e-05, "loss": 1.0099, "step": 1050 }, { "epoch": 0.05, "learning_rate": 1.9971369447807574e-05, "loss": 1.1146, "step": 1051 }, { "epoch": 0.05, "learning_rate": 1.997124489159753e-05, "loss": 1.1334, "step": 1052 }, { "epoch": 0.05, "learning_rate": 1.9971120065426638e-05, "loss": 1.0884, "step": 1053 }, { "epoch": 0.05, "learning_rate": 1.9970994969298276e-05, "loss": 1.029, "step": 1054 }, { "epoch": 0.05, "learning_rate": 1.9970869603215835e-05, "loss": 1.1647, "step": 1055 }, { "epoch": 0.05, "learning_rate": 1.997074396718271e-05, "loss": 1.0625, "step": 1056 }, { "epoch": 0.05, "learning_rate": 1.99706180612023e-05, "loss": 1.1623, "step": 1057 }, { "epoch": 0.05, "learning_rate": 1.997049188527801e-05, "loss": 1.0445, "step": 1058 }, { "epoch": 0.05, "learning_rate": 1.997036543941326e-05, "loss": 1.1296, "step": 1059 }, { "epoch": 0.05, "learning_rate": 1.9970238723611476e-05, "loss": 1.0276, "step": 1060 }, { "epoch": 0.05, "learning_rate": 1.9970111737876086e-05, "loss": 1.0132, "step": 1061 }, { "epoch": 0.05, "learning_rate": 1.996998448221053e-05, "loss": 1.1118, "step": 1062 }, { "epoch": 0.05, "learning_rate": 1.9969856956618248e-05, "loss": 1.1418, "step": 1063 }, { "epoch": 0.05, "learning_rate": 1.9969729161102694e-05, "loss": 1.1262, "step": 1064 }, { "epoch": 0.05, "learning_rate": 1.9969601095667335e-05, "loss": 1.1721, "step": 1065 }, { "epoch": 0.05, "learning_rate": 1.996947276031563e-05, "loss": 1.0797, "step": 1066 }, { "epoch": 0.05, "learning_rate": 1.996934415505106e-05, "loss": 1.0723, "step": 1067 }, { "epoch": 0.05, "learning_rate": 1.99692152798771e-05, "loss": 1.0676, "step": 1068 }, { "epoch": 0.05, "learning_rate": 1.9969086134797243e-05, "loss": 1.0807, "step": 1069 }, { "epoch": 0.05, "learning_rate": 1.996895671981499e-05, "loss": 1.0972, "step": 1070 }, { "epoch": 0.05, "learning_rate": 1.9968827034933836e-05, "loss": 1.046, "step": 1071 }, { "epoch": 0.05, "learning_rate": 1.9968697080157295e-05, "loss": 1.0616, "step": 1072 }, { "epoch": 0.05, "learning_rate": 1.9968566855488888e-05, "loss": 1.0195, "step": 1073 }, { "epoch": 0.05, "learning_rate": 1.996843636093214e-05, "loss": 0.9946, "step": 1074 }, { "epoch": 0.05, "learning_rate": 1.996830559649058e-05, "loss": 1.1397, "step": 1075 }, { "epoch": 0.05, "learning_rate": 1.9968174562167753e-05, "loss": 1.1355, "step": 1076 }, { "epoch": 0.05, "learning_rate": 1.9968043257967205e-05, "loss": 0.9948, "step": 1077 }, { "epoch": 0.05, "learning_rate": 1.996791168389249e-05, "loss": 1.0418, "step": 1078 }, { "epoch": 0.05, "learning_rate": 1.9967779839947172e-05, "loss": 1.0284, "step": 1079 }, { "epoch": 0.05, "learning_rate": 1.996764772613482e-05, "loss": 1.0479, "step": 1080 }, { "epoch": 0.05, "learning_rate": 1.996751534245901e-05, "loss": 1.0938, "step": 1081 }, { "epoch": 0.05, "learning_rate": 1.996738268892333e-05, "loss": 1.0529, "step": 1082 }, { "epoch": 0.06, "learning_rate": 1.9967249765531364e-05, "loss": 1.107, "step": 1083 }, { "epoch": 0.06, "learning_rate": 1.996711657228672e-05, "loss": 1.1292, "step": 1084 }, { "epoch": 0.06, "learning_rate": 1.9966983109192993e-05, "loss": 1.079, "step": 1085 }, { "epoch": 0.06, "learning_rate": 1.9966849376253804e-05, "loss": 1.1257, "step": 1086 }, { "epoch": 0.06, "learning_rate": 1.9966715373472775e-05, "loss": 1.0502, "step": 1087 }, { "epoch": 0.06, "learning_rate": 1.9966581100853527e-05, "loss": 1.1001, "step": 1088 }, { "epoch": 0.06, "learning_rate": 1.9966446558399703e-05, "loss": 1.0787, "step": 1089 }, { "epoch": 0.06, "learning_rate": 1.9966311746114937e-05, "loss": 1.0428, "step": 1090 }, { "epoch": 0.06, "learning_rate": 1.996617666400289e-05, "loss": 1.0246, "step": 1091 }, { "epoch": 0.06, "learning_rate": 1.996604131206721e-05, "loss": 1.0104, "step": 1092 }, { "epoch": 0.06, "learning_rate": 1.9965905690311562e-05, "loss": 1.0363, "step": 1093 }, { "epoch": 0.06, "learning_rate": 1.996576979873962e-05, "loss": 1.0704, "step": 1094 }, { "epoch": 0.06, "learning_rate": 1.996563363735507e-05, "loss": 1.0264, "step": 1095 }, { "epoch": 0.06, "learning_rate": 1.9965497206161588e-05, "loss": 1.0798, "step": 1096 }, { "epoch": 0.06, "learning_rate": 1.996536050516287e-05, "loss": 1.1345, "step": 1097 }, { "epoch": 0.06, "learning_rate": 1.996522353436262e-05, "loss": 1.1788, "step": 1098 }, { "epoch": 0.06, "learning_rate": 1.9965086293764546e-05, "loss": 1.0617, "step": 1099 }, { "epoch": 0.06, "learning_rate": 1.996494878337236e-05, "loss": 1.0288, "step": 1100 }, { "epoch": 0.06, "learning_rate": 1.9964811003189793e-05, "loss": 1.075, "step": 1101 }, { "epoch": 0.06, "learning_rate": 1.9964672953220564e-05, "loss": 0.9987, "step": 1102 }, { "epoch": 0.06, "learning_rate": 1.9964534633468417e-05, "loss": 1.0394, "step": 1103 }, { "epoch": 0.06, "learning_rate": 1.99643960439371e-05, "loss": 1.0238, "step": 1104 }, { "epoch": 0.06, "learning_rate": 1.9964257184630357e-05, "loss": 0.9629, "step": 1105 }, { "epoch": 0.06, "learning_rate": 1.9964118055551955e-05, "loss": 1.0516, "step": 1106 }, { "epoch": 0.06, "learning_rate": 1.9963978656705657e-05, "loss": 1.0699, "step": 1107 }, { "epoch": 0.06, "learning_rate": 1.9963838988095237e-05, "loss": 1.113, "step": 1108 }, { "epoch": 0.06, "learning_rate": 1.996369904972448e-05, "loss": 1.1166, "step": 1109 }, { "epoch": 0.06, "learning_rate": 1.9963558841597166e-05, "loss": 1.0597, "step": 1110 }, { "epoch": 0.06, "learning_rate": 1.9963418363717098e-05, "loss": 1.0218, "step": 1111 }, { "epoch": 0.06, "learning_rate": 1.9963277616088082e-05, "loss": 0.9646, "step": 1112 }, { "epoch": 0.06, "learning_rate": 1.9963136598713923e-05, "loss": 1.0923, "step": 1113 }, { "epoch": 0.06, "learning_rate": 1.9962995311598442e-05, "loss": 1.084, "step": 1114 }, { "epoch": 0.06, "learning_rate": 1.9962853754745458e-05, "loss": 1.0624, "step": 1115 }, { "epoch": 0.06, "learning_rate": 1.9962711928158813e-05, "loss": 1.0493, "step": 1116 }, { "epoch": 0.06, "learning_rate": 1.996256983184234e-05, "loss": 1.0609, "step": 1117 }, { "epoch": 0.06, "learning_rate": 1.996242746579989e-05, "loss": 1.0886, "step": 1118 }, { "epoch": 0.06, "learning_rate": 1.9962284830035313e-05, "loss": 1.0233, "step": 1119 }, { "epoch": 0.06, "learning_rate": 1.9962141924552473e-05, "loss": 1.1505, "step": 1120 }, { "epoch": 0.06, "learning_rate": 1.996199874935524e-05, "loss": 1.0687, "step": 1121 }, { "epoch": 0.06, "learning_rate": 1.996185530444749e-05, "loss": 1.0906, "step": 1122 }, { "epoch": 0.06, "learning_rate": 1.9961711589833106e-05, "loss": 1.0713, "step": 1123 }, { "epoch": 0.06, "learning_rate": 1.996156760551598e-05, "loss": 1.0805, "step": 1124 }, { "epoch": 0.06, "learning_rate": 1.996142335150001e-05, "loss": 1.1019, "step": 1125 }, { "epoch": 0.06, "learning_rate": 1.99612788277891e-05, "loss": 1.1892, "step": 1126 }, { "epoch": 0.06, "learning_rate": 1.9961134034387163e-05, "loss": 1.0647, "step": 1127 }, { "epoch": 0.06, "learning_rate": 1.996098897129812e-05, "loss": 1.0966, "step": 1128 }, { "epoch": 0.06, "learning_rate": 1.99608436385259e-05, "loss": 1.1299, "step": 1129 }, { "epoch": 0.06, "learning_rate": 1.9960698036074433e-05, "loss": 1.0879, "step": 1130 }, { "epoch": 0.06, "learning_rate": 1.9960552163947664e-05, "loss": 1.1039, "step": 1131 }, { "epoch": 0.06, "learning_rate": 1.9960406022149545e-05, "loss": 1.1222, "step": 1132 }, { "epoch": 0.06, "learning_rate": 1.9960259610684026e-05, "loss": 1.1716, "step": 1133 }, { "epoch": 0.06, "learning_rate": 1.9960112929555076e-05, "loss": 1.1078, "step": 1134 }, { "epoch": 0.06, "learning_rate": 1.995996597876667e-05, "loss": 0.9553, "step": 1135 }, { "epoch": 0.06, "learning_rate": 1.995981875832278e-05, "loss": 1.0577, "step": 1136 }, { "epoch": 0.06, "learning_rate": 1.995967126822739e-05, "loss": 1.092, "step": 1137 }, { "epoch": 0.06, "learning_rate": 1.99595235084845e-05, "loss": 1.1133, "step": 1138 }, { "epoch": 0.06, "learning_rate": 1.9959375479098107e-05, "loss": 1.2345, "step": 1139 }, { "epoch": 0.06, "learning_rate": 1.9959227180072216e-05, "loss": 1.0917, "step": 1140 }, { "epoch": 0.06, "learning_rate": 1.9959078611410848e-05, "loss": 1.0742, "step": 1141 }, { "epoch": 0.06, "learning_rate": 1.995892977311802e-05, "loss": 1.1108, "step": 1142 }, { "epoch": 0.06, "learning_rate": 1.9958780665197763e-05, "loss": 1.1181, "step": 1143 }, { "epoch": 0.06, "learning_rate": 1.9958631287654117e-05, "loss": 0.9818, "step": 1144 }, { "epoch": 0.06, "learning_rate": 1.9958481640491122e-05, "loss": 0.9664, "step": 1145 }, { "epoch": 0.06, "learning_rate": 1.9958331723712834e-05, "loss": 1.0541, "step": 1146 }, { "epoch": 0.06, "learning_rate": 1.9958181537323308e-05, "loss": 1.0013, "step": 1147 }, { "epoch": 0.06, "learning_rate": 1.995803108132661e-05, "loss": 1.0192, "step": 1148 }, { "epoch": 0.06, "learning_rate": 1.995788035572682e-05, "loss": 1.0526, "step": 1149 }, { "epoch": 0.06, "learning_rate": 1.995772936052801e-05, "loss": 1.0518, "step": 1150 }, { "epoch": 0.06, "learning_rate": 1.9957578095734273e-05, "loss": 1.0778, "step": 1151 }, { "epoch": 0.06, "learning_rate": 1.9957426561349704e-05, "loss": 1.0746, "step": 1152 }, { "epoch": 0.06, "learning_rate": 1.99572747573784e-05, "loss": 1.0684, "step": 1153 }, { "epoch": 0.06, "learning_rate": 1.995712268382448e-05, "loss": 1.0684, "step": 1154 }, { "epoch": 0.06, "learning_rate": 1.9956970340692057e-05, "loss": 1.1217, "step": 1155 }, { "epoch": 0.06, "learning_rate": 1.9956817727985253e-05, "loss": 1.0167, "step": 1156 }, { "epoch": 0.06, "learning_rate": 1.9956664845708204e-05, "loss": 1.012, "step": 1157 }, { "epoch": 0.06, "learning_rate": 1.9956511693865044e-05, "loss": 1.0292, "step": 1158 }, { "epoch": 0.06, "learning_rate": 1.9956358272459927e-05, "loss": 1.0772, "step": 1159 }, { "epoch": 0.06, "learning_rate": 1.9956204581497003e-05, "loss": 1.0019, "step": 1160 }, { "epoch": 0.06, "learning_rate": 1.9956050620980433e-05, "loss": 1.1351, "step": 1161 }, { "epoch": 0.06, "learning_rate": 1.995589639091438e-05, "loss": 1.062, "step": 1162 }, { "epoch": 0.06, "learning_rate": 1.9955741891303028e-05, "loss": 1.0402, "step": 1163 }, { "epoch": 0.06, "learning_rate": 1.9955587122150554e-05, "loss": 0.9939, "step": 1164 }, { "epoch": 0.06, "learning_rate": 1.995543208346115e-05, "loss": 1.0106, "step": 1165 }, { "epoch": 0.06, "learning_rate": 1.9955276775239017e-05, "loss": 1.0378, "step": 1166 }, { "epoch": 0.06, "learning_rate": 1.9955121197488356e-05, "loss": 1.048, "step": 1167 }, { "epoch": 0.06, "learning_rate": 1.995496535021338e-05, "loss": 1.108, "step": 1168 }, { "epoch": 0.06, "learning_rate": 1.9954809233418308e-05, "loss": 1.0759, "step": 1169 }, { "epoch": 0.06, "learning_rate": 1.9954652847107367e-05, "loss": 1.0784, "step": 1170 }, { "epoch": 0.06, "learning_rate": 1.9954496191284792e-05, "loss": 1.0669, "step": 1171 }, { "epoch": 0.06, "learning_rate": 1.9954339265954823e-05, "loss": 1.1693, "step": 1172 }, { "epoch": 0.06, "learning_rate": 1.9954182071121708e-05, "loss": 1.1003, "step": 1173 }, { "epoch": 0.06, "learning_rate": 1.9954024606789705e-05, "loss": 1.0477, "step": 1174 }, { "epoch": 0.06, "learning_rate": 1.9953866872963075e-05, "loss": 1.0994, "step": 1175 }, { "epoch": 0.06, "learning_rate": 1.995370886964609e-05, "loss": 1.1726, "step": 1176 }, { "epoch": 0.06, "learning_rate": 1.9953550596843028e-05, "loss": 0.971, "step": 1177 }, { "epoch": 0.06, "learning_rate": 1.9953392054558174e-05, "loss": 1.0438, "step": 1178 }, { "epoch": 0.06, "learning_rate": 1.995323324279582e-05, "loss": 1.0516, "step": 1179 }, { "epoch": 0.06, "learning_rate": 1.995307416156026e-05, "loss": 1.0466, "step": 1180 }, { "epoch": 0.06, "learning_rate": 1.9952914810855813e-05, "loss": 1.1179, "step": 1181 }, { "epoch": 0.06, "learning_rate": 1.995275519068678e-05, "loss": 1.0038, "step": 1182 }, { "epoch": 0.06, "learning_rate": 1.9952595301057497e-05, "loss": 1.143, "step": 1183 }, { "epoch": 0.06, "learning_rate": 1.9952435141972278e-05, "loss": 1.105, "step": 1184 }, { "epoch": 0.06, "learning_rate": 1.995227471343547e-05, "loss": 1.017, "step": 1185 }, { "epoch": 0.06, "learning_rate": 1.9952114015451412e-05, "loss": 0.9588, "step": 1186 }, { "epoch": 0.06, "learning_rate": 1.9951953048024454e-05, "loss": 1.0352, "step": 1187 }, { "epoch": 0.06, "learning_rate": 1.9951791811158957e-05, "loss": 1.0932, "step": 1188 }, { "epoch": 0.06, "learning_rate": 1.995163030485928e-05, "loss": 1.1176, "step": 1189 }, { "epoch": 0.06, "learning_rate": 1.9951468529129806e-05, "loss": 1.0626, "step": 1190 }, { "epoch": 0.06, "learning_rate": 1.9951306483974907e-05, "loss": 1.011, "step": 1191 }, { "epoch": 0.06, "learning_rate": 1.995114416939897e-05, "loss": 1.0674, "step": 1192 }, { "epoch": 0.06, "learning_rate": 1.9950981585406395e-05, "loss": 1.0283, "step": 1193 }, { "epoch": 0.06, "learning_rate": 1.9950818732001578e-05, "loss": 1.1021, "step": 1194 }, { "epoch": 0.06, "learning_rate": 1.995065560918893e-05, "loss": 1.0466, "step": 1195 }, { "epoch": 0.06, "learning_rate": 1.9950492216972873e-05, "loss": 0.9516, "step": 1196 }, { "epoch": 0.06, "learning_rate": 1.995032855535782e-05, "loss": 1.0654, "step": 1197 }, { "epoch": 0.06, "learning_rate": 1.9950164624348215e-05, "loss": 1.1244, "step": 1198 }, { "epoch": 0.06, "learning_rate": 1.995000042394848e-05, "loss": 1.0355, "step": 1199 }, { "epoch": 0.06, "learning_rate": 1.9949835954163075e-05, "loss": 1.0806, "step": 1200 }, { "epoch": 0.06, "learning_rate": 1.9949671214996448e-05, "loss": 0.9832, "step": 1201 }, { "epoch": 0.06, "learning_rate": 1.9949506206453055e-05, "loss": 1.0361, "step": 1202 }, { "epoch": 0.06, "learning_rate": 1.994934092853737e-05, "loss": 1.067, "step": 1203 }, { "epoch": 0.06, "learning_rate": 1.9949175381253863e-05, "loss": 1.0472, "step": 1204 }, { "epoch": 0.06, "learning_rate": 1.994900956460702e-05, "loss": 1.0741, "step": 1205 }, { "epoch": 0.06, "learning_rate": 1.9948843478601322e-05, "loss": 1.1432, "step": 1206 }, { "epoch": 0.06, "learning_rate": 1.9948677123241276e-05, "loss": 1.0517, "step": 1207 }, { "epoch": 0.06, "learning_rate": 1.9948510498531383e-05, "loss": 1.1292, "step": 1208 }, { "epoch": 0.06, "learning_rate": 1.994834360447615e-05, "loss": 1.2039, "step": 1209 }, { "epoch": 0.06, "learning_rate": 1.9948176441080096e-05, "loss": 1.0208, "step": 1210 }, { "epoch": 0.06, "learning_rate": 1.9948009008347754e-05, "loss": 1.0238, "step": 1211 }, { "epoch": 0.06, "learning_rate": 1.994784130628365e-05, "loss": 1.0046, "step": 1212 }, { "epoch": 0.06, "learning_rate": 1.9947673334892323e-05, "loss": 1.0363, "step": 1213 }, { "epoch": 0.06, "learning_rate": 1.994750509417833e-05, "loss": 1.0685, "step": 1214 }, { "epoch": 0.06, "learning_rate": 1.9947336584146212e-05, "loss": 1.1132, "step": 1215 }, { "epoch": 0.06, "learning_rate": 1.9947167804800546e-05, "loss": 1.0408, "step": 1216 }, { "epoch": 0.06, "learning_rate": 1.9946998756145894e-05, "loss": 1.1293, "step": 1217 }, { "epoch": 0.06, "learning_rate": 1.994682943818683e-05, "loss": 1.0869, "step": 1218 }, { "epoch": 0.06, "learning_rate": 1.994665985092794e-05, "loss": 1.0182, "step": 1219 }, { "epoch": 0.06, "learning_rate": 1.9946489994373823e-05, "loss": 0.9915, "step": 1220 }, { "epoch": 0.06, "learning_rate": 1.994631986852907e-05, "loss": 1.054, "step": 1221 }, { "epoch": 0.06, "learning_rate": 1.9946149473398287e-05, "loss": 1.0151, "step": 1222 }, { "epoch": 0.06, "learning_rate": 1.994597880898609e-05, "loss": 0.9518, "step": 1223 }, { "epoch": 0.06, "learning_rate": 1.9945807875297096e-05, "loss": 1.0652, "step": 1224 }, { "epoch": 0.06, "learning_rate": 1.994563667233594e-05, "loss": 1.0524, "step": 1225 }, { "epoch": 0.06, "learning_rate": 1.994546520010725e-05, "loss": 1.0703, "step": 1226 }, { "epoch": 0.06, "learning_rate": 1.9945293458615667e-05, "loss": 1.1505, "step": 1227 }, { "epoch": 0.06, "learning_rate": 1.994512144786585e-05, "loss": 1.0758, "step": 1228 }, { "epoch": 0.06, "learning_rate": 1.994494916786245e-05, "loss": 1.0319, "step": 1229 }, { "epoch": 0.06, "learning_rate": 1.9944776618610127e-05, "loss": 1.0282, "step": 1230 }, { "epoch": 0.06, "learning_rate": 1.9944603800113563e-05, "loss": 1.0412, "step": 1231 }, { "epoch": 0.06, "learning_rate": 1.994443071237743e-05, "loss": 1.0441, "step": 1232 }, { "epoch": 0.06, "learning_rate": 1.9944257355406413e-05, "loss": 1.0441, "step": 1233 }, { "epoch": 0.06, "learning_rate": 1.9944083729205207e-05, "loss": 1.1428, "step": 1234 }, { "epoch": 0.06, "learning_rate": 1.994390983377852e-05, "loss": 1.0013, "step": 1235 }, { "epoch": 0.06, "learning_rate": 1.994373566913105e-05, "loss": 1.0766, "step": 1236 }, { "epoch": 0.06, "learning_rate": 1.9943561235267514e-05, "loss": 1.0173, "step": 1237 }, { "epoch": 0.06, "learning_rate": 1.9943386532192642e-05, "loss": 1.0356, "step": 1238 }, { "epoch": 0.06, "learning_rate": 1.9943211559911153e-05, "loss": 1.1184, "step": 1239 }, { "epoch": 0.06, "learning_rate": 1.994303631842779e-05, "loss": 1.0822, "step": 1240 }, { "epoch": 0.06, "learning_rate": 1.99428608077473e-05, "loss": 0.9891, "step": 1241 }, { "epoch": 0.06, "learning_rate": 1.9942685027874435e-05, "loss": 1.005, "step": 1242 }, { "epoch": 0.06, "learning_rate": 1.9942508978813946e-05, "loss": 1.0307, "step": 1243 }, { "epoch": 0.06, "learning_rate": 1.9942332660570605e-05, "loss": 1.1491, "step": 1244 }, { "epoch": 0.06, "learning_rate": 1.9942156073149186e-05, "loss": 1.0892, "step": 1245 }, { "epoch": 0.06, "learning_rate": 1.994197921655447e-05, "loss": 1.0703, "step": 1246 }, { "epoch": 0.06, "learning_rate": 1.994180209079124e-05, "loss": 1.0209, "step": 1247 }, { "epoch": 0.06, "learning_rate": 1.99416246958643e-05, "loss": 1.0685, "step": 1248 }, { "epoch": 0.06, "learning_rate": 1.9941447031778447e-05, "loss": 1.0582, "step": 1249 }, { "epoch": 0.06, "learning_rate": 1.9941269098538494e-05, "loss": 1.0868, "step": 1250 }, { "epoch": 0.06, "learning_rate": 1.9941090896149254e-05, "loss": 0.9937, "step": 1251 }, { "epoch": 0.06, "learning_rate": 1.9940912424615553e-05, "loss": 1.0794, "step": 1252 }, { "epoch": 0.06, "learning_rate": 1.994073368394223e-05, "loss": 1.0811, "step": 1253 }, { "epoch": 0.06, "learning_rate": 1.9940554674134116e-05, "loss": 1.0682, "step": 1254 }, { "epoch": 0.06, "learning_rate": 1.994037539519606e-05, "loss": 1.1483, "step": 1255 }, { "epoch": 0.06, "learning_rate": 1.9940195847132916e-05, "loss": 1.1448, "step": 1256 }, { "epoch": 0.06, "learning_rate": 1.9940016029949546e-05, "loss": 0.9686, "step": 1257 }, { "epoch": 0.06, "learning_rate": 1.9939835943650815e-05, "loss": 1.051, "step": 1258 }, { "epoch": 0.06, "learning_rate": 1.9939655588241602e-05, "loss": 1.116, "step": 1259 }, { "epoch": 0.06, "learning_rate": 1.993947496372679e-05, "loss": 1.0713, "step": 1260 }, { "epoch": 0.06, "learning_rate": 1.9939294070111266e-05, "loss": 1.0694, "step": 1261 }, { "epoch": 0.06, "learning_rate": 1.993911290739993e-05, "loss": 1.1058, "step": 1262 }, { "epoch": 0.06, "learning_rate": 1.9938931475597688e-05, "loss": 0.9451, "step": 1263 }, { "epoch": 0.06, "learning_rate": 1.993874977470945e-05, "loss": 1.0532, "step": 1264 }, { "epoch": 0.06, "learning_rate": 1.9938567804740137e-05, "loss": 1.0875, "step": 1265 }, { "epoch": 0.06, "learning_rate": 1.993838556569467e-05, "loss": 1.0775, "step": 1266 }, { "epoch": 0.06, "learning_rate": 1.9938203057577993e-05, "loss": 1.1072, "step": 1267 }, { "epoch": 0.06, "learning_rate": 1.9938020280395037e-05, "loss": 1.109, "step": 1268 }, { "epoch": 0.06, "learning_rate": 1.9937837234150758e-05, "loss": 1.0414, "step": 1269 }, { "epoch": 0.06, "learning_rate": 1.9937653918850108e-05, "loss": 1.1854, "step": 1270 }, { "epoch": 0.06, "learning_rate": 1.9937470334498052e-05, "loss": 1.0607, "step": 1271 }, { "epoch": 0.06, "learning_rate": 1.9937286481099557e-05, "loss": 1.1491, "step": 1272 }, { "epoch": 0.06, "learning_rate": 1.9937102358659603e-05, "loss": 1.027, "step": 1273 }, { "epoch": 0.06, "learning_rate": 1.9936917967183177e-05, "loss": 1.0537, "step": 1274 }, { "epoch": 0.06, "learning_rate": 1.993673330667527e-05, "loss": 1.0742, "step": 1275 }, { "epoch": 0.06, "learning_rate": 1.9936548377140877e-05, "loss": 1.0132, "step": 1276 }, { "epoch": 0.06, "learning_rate": 1.993636317858501e-05, "loss": 1.0459, "step": 1277 }, { "epoch": 0.06, "learning_rate": 1.9936177711012684e-05, "loss": 1.0864, "step": 1278 }, { "epoch": 0.06, "learning_rate": 1.9935991974428917e-05, "loss": 1.0923, "step": 1279 }, { "epoch": 0.07, "learning_rate": 1.9935805968838735e-05, "loss": 1.0724, "step": 1280 }, { "epoch": 0.07, "learning_rate": 1.993561969424718e-05, "loss": 1.0069, "step": 1281 }, { "epoch": 0.07, "learning_rate": 1.993543315065929e-05, "loss": 1.0621, "step": 1282 }, { "epoch": 0.07, "learning_rate": 1.993524633808012e-05, "loss": 1.1085, "step": 1283 }, { "epoch": 0.07, "learning_rate": 1.9935059256514725e-05, "loss": 1.114, "step": 1284 }, { "epoch": 0.07, "learning_rate": 1.993487190596817e-05, "loss": 1.0458, "step": 1285 }, { "epoch": 0.07, "learning_rate": 1.993468428644553e-05, "loss": 1.0676, "step": 1286 }, { "epoch": 0.07, "learning_rate": 1.993449639795188e-05, "loss": 1.07, "step": 1287 }, { "epoch": 0.07, "learning_rate": 1.9934308240492313e-05, "loss": 1.1288, "step": 1288 }, { "epoch": 0.07, "learning_rate": 1.993411981407192e-05, "loss": 1.0466, "step": 1289 }, { "epoch": 0.07, "learning_rate": 1.9933931118695795e-05, "loss": 1.0812, "step": 1290 }, { "epoch": 0.07, "learning_rate": 1.993374215436906e-05, "loss": 1.0471, "step": 1291 }, { "epoch": 0.07, "learning_rate": 1.993355292109682e-05, "loss": 1.107, "step": 1292 }, { "epoch": 0.07, "learning_rate": 1.993336341888421e-05, "loss": 0.9891, "step": 1293 }, { "epoch": 0.07, "learning_rate": 1.9933173647736348e-05, "loss": 1.0419, "step": 1294 }, { "epoch": 0.07, "learning_rate": 1.993298360765838e-05, "loss": 0.9984, "step": 1295 }, { "epoch": 0.07, "learning_rate": 1.993279329865545e-05, "loss": 0.988, "step": 1296 }, { "epoch": 0.07, "learning_rate": 1.9932602720732706e-05, "loss": 1.04, "step": 1297 }, { "epoch": 0.07, "learning_rate": 1.9932411873895316e-05, "loss": 1.0848, "step": 1298 }, { "epoch": 0.07, "learning_rate": 1.9932220758148434e-05, "loss": 0.9808, "step": 1299 }, { "epoch": 0.07, "learning_rate": 1.9932029373497248e-05, "loss": 1.0299, "step": 1300 }, { "epoch": 0.07, "learning_rate": 1.9931837719946928e-05, "loss": 1.0851, "step": 1301 }, { "epoch": 0.07, "learning_rate": 1.9931645797502676e-05, "loss": 1.0971, "step": 1302 }, { "epoch": 0.07, "learning_rate": 1.9931453606169674e-05, "loss": 1.0769, "step": 1303 }, { "epoch": 0.07, "learning_rate": 1.9931261145953135e-05, "loss": 1.0382, "step": 1304 }, { "epoch": 0.07, "learning_rate": 1.9931068416858265e-05, "loss": 1.0302, "step": 1305 }, { "epoch": 0.07, "learning_rate": 1.9930875418890285e-05, "loss": 1.1189, "step": 1306 }, { "epoch": 0.07, "learning_rate": 1.9930682152054418e-05, "loss": 1.0314, "step": 1307 }, { "epoch": 0.07, "learning_rate": 1.9930488616355896e-05, "loss": 1.0164, "step": 1308 }, { "epoch": 0.07, "learning_rate": 1.9930294811799963e-05, "loss": 1.0384, "step": 1309 }, { "epoch": 0.07, "learning_rate": 1.993010073839186e-05, "loss": 1.1464, "step": 1310 }, { "epoch": 0.07, "learning_rate": 1.9929906396136845e-05, "loss": 1.0718, "step": 1311 }, { "epoch": 0.07, "learning_rate": 1.992971178504018e-05, "loss": 1.1027, "step": 1312 }, { "epoch": 0.07, "learning_rate": 1.9929516905107132e-05, "loss": 1.0468, "step": 1313 }, { "epoch": 0.07, "learning_rate": 1.992932175634298e-05, "loss": 1.0057, "step": 1314 }, { "epoch": 0.07, "learning_rate": 1.9929126338753002e-05, "loss": 1.0855, "step": 1315 }, { "epoch": 0.07, "learning_rate": 1.9928930652342492e-05, "loss": 0.9919, "step": 1316 }, { "epoch": 0.07, "learning_rate": 1.992873469711675e-05, "loss": 1.0474, "step": 1317 }, { "epoch": 0.07, "learning_rate": 1.992853847308108e-05, "loss": 1.0305, "step": 1318 }, { "epoch": 0.07, "learning_rate": 1.9928341980240793e-05, "loss": 1.0629, "step": 1319 }, { "epoch": 0.07, "learning_rate": 1.992814521860121e-05, "loss": 1.0147, "step": 1320 }, { "epoch": 0.07, "learning_rate": 1.9927948188167655e-05, "loss": 1.0303, "step": 1321 }, { "epoch": 0.07, "learning_rate": 1.9927750888945468e-05, "loss": 0.9474, "step": 1322 }, { "epoch": 0.07, "learning_rate": 1.992755332093999e-05, "loss": 1.1208, "step": 1323 }, { "epoch": 0.07, "learning_rate": 1.9927355484156567e-05, "loss": 1.0418, "step": 1324 }, { "epoch": 0.07, "learning_rate": 1.9927157378600553e-05, "loss": 1.0803, "step": 1325 }, { "epoch": 0.07, "learning_rate": 1.992695900427732e-05, "loss": 1.0945, "step": 1326 }, { "epoch": 0.07, "learning_rate": 1.9926760361192227e-05, "loss": 1.1232, "step": 1327 }, { "epoch": 0.07, "learning_rate": 1.9926561449350663e-05, "loss": 1.0497, "step": 1328 }, { "epoch": 0.07, "learning_rate": 1.9926362268758007e-05, "loss": 1.0254, "step": 1329 }, { "epoch": 0.07, "learning_rate": 1.9926162819419655e-05, "loss": 1.0459, "step": 1330 }, { "epoch": 0.07, "learning_rate": 1.9925963101341005e-05, "loss": 1.1171, "step": 1331 }, { "epoch": 0.07, "learning_rate": 1.992576311452746e-05, "loss": 1.0991, "step": 1332 }, { "epoch": 0.07, "learning_rate": 1.9925562858984444e-05, "loss": 1.0842, "step": 1333 }, { "epoch": 0.07, "learning_rate": 1.992536233471737e-05, "loss": 1.1017, "step": 1334 }, { "epoch": 0.07, "learning_rate": 1.9925161541731675e-05, "loss": 1.071, "step": 1335 }, { "epoch": 0.07, "learning_rate": 1.992496048003279e-05, "loss": 1.0225, "step": 1336 }, { "epoch": 0.07, "learning_rate": 1.9924759149626155e-05, "loss": 1.117, "step": 1337 }, { "epoch": 0.07, "learning_rate": 1.9924557550517227e-05, "loss": 1.0609, "step": 1338 }, { "epoch": 0.07, "learning_rate": 1.9924355682711465e-05, "loss": 1.038, "step": 1339 }, { "epoch": 0.07, "learning_rate": 1.992415354621433e-05, "loss": 1.0162, "step": 1340 }, { "epoch": 0.07, "learning_rate": 1.9923951141031294e-05, "loss": 1.0576, "step": 1341 }, { "epoch": 0.07, "learning_rate": 1.9923748467167844e-05, "loss": 1.0688, "step": 1342 }, { "epoch": 0.07, "learning_rate": 1.9923545524629456e-05, "loss": 1.0341, "step": 1343 }, { "epoch": 0.07, "learning_rate": 1.9923342313421633e-05, "loss": 1.0523, "step": 1344 }, { "epoch": 0.07, "learning_rate": 1.9923138833549876e-05, "loss": 1.071, "step": 1345 }, { "epoch": 0.07, "learning_rate": 1.9922935085019693e-05, "loss": 1.0282, "step": 1346 }, { "epoch": 0.07, "learning_rate": 1.99227310678366e-05, "loss": 1.0255, "step": 1347 }, { "epoch": 0.07, "learning_rate": 1.992252678200612e-05, "loss": 0.9415, "step": 1348 }, { "epoch": 0.07, "learning_rate": 1.9922322227533782e-05, "loss": 1.119, "step": 1349 }, { "epoch": 0.07, "learning_rate": 1.9922117404425127e-05, "loss": 1.1174, "step": 1350 }, { "epoch": 0.07, "learning_rate": 1.9921912312685698e-05, "loss": 1.1205, "step": 1351 }, { "epoch": 0.07, "learning_rate": 1.9921706952321053e-05, "loss": 1.146, "step": 1352 }, { "epoch": 0.07, "learning_rate": 1.9921501323336743e-05, "loss": 1.0568, "step": 1353 }, { "epoch": 0.07, "learning_rate": 1.9921295425738344e-05, "loss": 1.0899, "step": 1354 }, { "epoch": 0.07, "learning_rate": 1.992108925953143e-05, "loss": 1.0923, "step": 1355 }, { "epoch": 0.07, "learning_rate": 1.9920882824721573e-05, "loss": 1.0612, "step": 1356 }, { "epoch": 0.07, "learning_rate": 1.992067612131437e-05, "loss": 1.1335, "step": 1357 }, { "epoch": 0.07, "learning_rate": 1.9920469149315417e-05, "loss": 1.0041, "step": 1358 }, { "epoch": 0.07, "learning_rate": 1.992026190873032e-05, "loss": 1.1062, "step": 1359 }, { "epoch": 0.07, "learning_rate": 1.9920054399564683e-05, "loss": 0.9391, "step": 1360 }, { "epoch": 0.07, "learning_rate": 1.9919846621824123e-05, "loss": 1.1449, "step": 1361 }, { "epoch": 0.07, "learning_rate": 1.9919638575514275e-05, "loss": 1.0837, "step": 1362 }, { "epoch": 0.07, "learning_rate": 1.9919430260640765e-05, "loss": 1.0969, "step": 1363 }, { "epoch": 0.07, "learning_rate": 1.9919221677209233e-05, "loss": 1.052, "step": 1364 }, { "epoch": 0.07, "learning_rate": 1.9919012825225325e-05, "loss": 1.1101, "step": 1365 }, { "epoch": 0.07, "learning_rate": 1.9918803704694703e-05, "loss": 1.119, "step": 1366 }, { "epoch": 0.07, "learning_rate": 1.9918594315623022e-05, "loss": 1.0287, "step": 1367 }, { "epoch": 0.07, "learning_rate": 1.991838465801595e-05, "loss": 1.0806, "step": 1368 }, { "epoch": 0.07, "learning_rate": 1.9918174731879168e-05, "loss": 0.9515, "step": 1369 }, { "epoch": 0.07, "learning_rate": 1.9917964537218358e-05, "loss": 0.9533, "step": 1370 }, { "epoch": 0.07, "learning_rate": 1.9917754074039206e-05, "loss": 0.9998, "step": 1371 }, { "epoch": 0.07, "learning_rate": 1.991754334234742e-05, "loss": 1.0691, "step": 1372 }, { "epoch": 0.07, "learning_rate": 1.9917332342148694e-05, "loss": 1.0356, "step": 1373 }, { "epoch": 0.07, "learning_rate": 1.9917121073448748e-05, "loss": 1.0304, "step": 1374 }, { "epoch": 0.07, "learning_rate": 1.9916909536253303e-05, "loss": 1.0425, "step": 1375 }, { "epoch": 0.07, "learning_rate": 1.991669773056808e-05, "loss": 0.9816, "step": 1376 }, { "epoch": 0.07, "learning_rate": 1.9916485656398815e-05, "loss": 1.0533, "step": 1377 }, { "epoch": 0.07, "learning_rate": 1.991627331375125e-05, "loss": 1.0209, "step": 1378 }, { "epoch": 0.07, "learning_rate": 1.991606070263114e-05, "loss": 1.1205, "step": 1379 }, { "epoch": 0.07, "learning_rate": 1.9915847823044237e-05, "loss": 1.0068, "step": 1380 }, { "epoch": 0.07, "learning_rate": 1.99156346749963e-05, "loss": 1.0292, "step": 1381 }, { "epoch": 0.07, "learning_rate": 1.9915421258493103e-05, "loss": 0.9978, "step": 1382 }, { "epoch": 0.07, "learning_rate": 1.9915207573540426e-05, "loss": 1.1331, "step": 1383 }, { "epoch": 0.07, "learning_rate": 1.9914993620144055e-05, "loss": 1.1243, "step": 1384 }, { "epoch": 0.07, "learning_rate": 1.991477939830978e-05, "loss": 1.1185, "step": 1385 }, { "epoch": 0.07, "learning_rate": 1.9914564908043398e-05, "loss": 1.111, "step": 1386 }, { "epoch": 0.07, "learning_rate": 1.991435014935072e-05, "loss": 1.0302, "step": 1387 }, { "epoch": 0.07, "learning_rate": 1.9914135122237563e-05, "loss": 0.9488, "step": 1388 }, { "epoch": 0.07, "learning_rate": 1.991391982670974e-05, "loss": 1.0577, "step": 1389 }, { "epoch": 0.07, "learning_rate": 1.991370426277309e-05, "loss": 1.0908, "step": 1390 }, { "epoch": 0.07, "learning_rate": 1.9913488430433445e-05, "loss": 1.0417, "step": 1391 }, { "epoch": 0.07, "learning_rate": 1.9913272329696646e-05, "loss": 0.9745, "step": 1392 }, { "epoch": 0.07, "learning_rate": 1.9913055960568547e-05, "loss": 1.1107, "step": 1393 }, { "epoch": 0.07, "learning_rate": 1.9912839323055002e-05, "loss": 1.103, "step": 1394 }, { "epoch": 0.07, "learning_rate": 1.991262241716188e-05, "loss": 1.0616, "step": 1395 }, { "epoch": 0.07, "learning_rate": 1.9912405242895052e-05, "loss": 0.9946, "step": 1396 }, { "epoch": 0.07, "learning_rate": 1.9912187800260398e-05, "loss": 0.9727, "step": 1397 }, { "epoch": 0.07, "learning_rate": 1.9911970089263805e-05, "loss": 1.0863, "step": 1398 }, { "epoch": 0.07, "learning_rate": 1.9911752109911168e-05, "loss": 1.0073, "step": 1399 }, { "epoch": 0.07, "learning_rate": 1.9911533862208387e-05, "loss": 1.062, "step": 1400 }, { "epoch": 0.07, "learning_rate": 1.9911315346161374e-05, "loss": 1.104, "step": 1401 }, { "epoch": 0.07, "learning_rate": 1.991109656177604e-05, "loss": 1.0684, "step": 1402 }, { "epoch": 0.07, "learning_rate": 1.9910877509058314e-05, "loss": 1.0756, "step": 1403 }, { "epoch": 0.07, "learning_rate": 1.9910658188014122e-05, "loss": 1.0349, "step": 1404 }, { "epoch": 0.07, "learning_rate": 1.99104385986494e-05, "loss": 0.9493, "step": 1405 }, { "epoch": 0.07, "learning_rate": 1.99102187409701e-05, "loss": 1.1285, "step": 1406 }, { "epoch": 0.07, "learning_rate": 1.9909998614982175e-05, "loss": 1.0609, "step": 1407 }, { "epoch": 0.07, "learning_rate": 1.990977822069158e-05, "loss": 1.0032, "step": 1408 }, { "epoch": 0.07, "learning_rate": 1.990955755810428e-05, "loss": 1.037, "step": 1409 }, { "epoch": 0.07, "learning_rate": 1.990933662722625e-05, "loss": 0.9628, "step": 1410 }, { "epoch": 0.07, "learning_rate": 1.990911542806348e-05, "loss": 0.9125, "step": 1411 }, { "epoch": 0.07, "learning_rate": 1.9908893960621945e-05, "loss": 1.0149, "step": 1412 }, { "epoch": 0.07, "learning_rate": 1.990867222490765e-05, "loss": 1.159, "step": 1413 }, { "epoch": 0.07, "learning_rate": 1.9908450220926597e-05, "loss": 1.0415, "step": 1414 }, { "epoch": 0.07, "learning_rate": 1.9908227948684796e-05, "loss": 1.12, "step": 1415 }, { "epoch": 0.07, "learning_rate": 1.9908005408188267e-05, "loss": 0.9756, "step": 1416 }, { "epoch": 0.07, "learning_rate": 1.990778259944303e-05, "loss": 1.0844, "step": 1417 }, { "epoch": 0.07, "learning_rate": 1.9907559522455123e-05, "loss": 1.0153, "step": 1418 }, { "epoch": 0.07, "learning_rate": 1.9907336177230582e-05, "loss": 1.0831, "step": 1419 }, { "epoch": 0.07, "learning_rate": 1.9907112563775457e-05, "loss": 1.005, "step": 1420 }, { "epoch": 0.07, "learning_rate": 1.9906888682095796e-05, "loss": 0.8764, "step": 1421 }, { "epoch": 0.07, "learning_rate": 1.9906664532197664e-05, "loss": 0.9146, "step": 1422 }, { "epoch": 0.07, "learning_rate": 1.990644011408713e-05, "loss": 1.1236, "step": 1423 }, { "epoch": 0.07, "learning_rate": 1.990621542777027e-05, "loss": 0.9255, "step": 1424 }, { "epoch": 0.07, "learning_rate": 1.9905990473253166e-05, "loss": 1.055, "step": 1425 }, { "epoch": 0.07, "learning_rate": 1.9905765250541913e-05, "loss": 1.144, "step": 1426 }, { "epoch": 0.07, "learning_rate": 1.99055397596426e-05, "loss": 0.975, "step": 1427 }, { "epoch": 0.07, "learning_rate": 1.990531400056134e-05, "loss": 1.0132, "step": 1428 }, { "epoch": 0.07, "learning_rate": 1.990508797330424e-05, "loss": 1.09, "step": 1429 }, { "epoch": 0.07, "learning_rate": 1.990486167787742e-05, "loss": 1.0377, "step": 1430 }, { "epoch": 0.07, "learning_rate": 1.990463511428701e-05, "loss": 1.0164, "step": 1431 }, { "epoch": 0.07, "learning_rate": 1.990440828253914e-05, "loss": 1.1129, "step": 1432 }, { "epoch": 0.07, "learning_rate": 1.9904181182639958e-05, "loss": 1.0962, "step": 1433 }, { "epoch": 0.07, "learning_rate": 1.9903953814595603e-05, "loss": 1.0177, "step": 1434 }, { "epoch": 0.07, "learning_rate": 1.990372617841224e-05, "loss": 1.0772, "step": 1435 }, { "epoch": 0.07, "learning_rate": 1.990349827409602e-05, "loss": 1.0795, "step": 1436 }, { "epoch": 0.07, "learning_rate": 1.9903270101653126e-05, "loss": 1.0735, "step": 1437 }, { "epoch": 0.07, "learning_rate": 1.990304166108973e-05, "loss": 1.0825, "step": 1438 }, { "epoch": 0.07, "learning_rate": 1.9902812952412013e-05, "loss": 1.0768, "step": 1439 }, { "epoch": 0.07, "learning_rate": 1.9902583975626177e-05, "loss": 1.0428, "step": 1440 }, { "epoch": 0.07, "learning_rate": 1.9902354730738412e-05, "loss": 1.0604, "step": 1441 }, { "epoch": 0.07, "learning_rate": 1.9902125217754927e-05, "loss": 1.1062, "step": 1442 }, { "epoch": 0.07, "learning_rate": 1.990189543668194e-05, "loss": 1.0733, "step": 1443 }, { "epoch": 0.07, "learning_rate": 1.9901665387525668e-05, "loss": 1.1379, "step": 1444 }, { "epoch": 0.07, "learning_rate": 1.990143507029234e-05, "loss": 1.064, "step": 1445 }, { "epoch": 0.07, "learning_rate": 1.9901204484988192e-05, "loss": 1.0702, "step": 1446 }, { "epoch": 0.07, "learning_rate": 1.9900973631619466e-05, "loss": 1.069, "step": 1447 }, { "epoch": 0.07, "learning_rate": 1.990074251019241e-05, "loss": 1.0594, "step": 1448 }, { "epoch": 0.07, "learning_rate": 1.9900511120713288e-05, "loss": 1.1277, "step": 1449 }, { "epoch": 0.07, "learning_rate": 1.990027946318836e-05, "loss": 1.0991, "step": 1450 }, { "epoch": 0.07, "learning_rate": 1.9900047537623898e-05, "loss": 0.966, "step": 1451 }, { "epoch": 0.07, "learning_rate": 1.9899815344026182e-05, "loss": 0.9121, "step": 1452 }, { "epoch": 0.07, "learning_rate": 1.9899582882401498e-05, "loss": 1.1415, "step": 1453 }, { "epoch": 0.07, "learning_rate": 1.989935015275614e-05, "loss": 1.0716, "step": 1454 }, { "epoch": 0.07, "learning_rate": 1.989911715509641e-05, "loss": 1.0264, "step": 1455 }, { "epoch": 0.07, "learning_rate": 1.989888388942861e-05, "loss": 1.1538, "step": 1456 }, { "epoch": 0.07, "learning_rate": 1.9898650355759065e-05, "loss": 0.9383, "step": 1457 }, { "epoch": 0.07, "learning_rate": 1.989841655409409e-05, "loss": 1.1154, "step": 1458 }, { "epoch": 0.07, "learning_rate": 1.989818248444002e-05, "loss": 0.9104, "step": 1459 }, { "epoch": 0.07, "learning_rate": 1.989794814680319e-05, "loss": 1.0425, "step": 1460 }, { "epoch": 0.07, "learning_rate": 1.9897713541189943e-05, "loss": 1.0411, "step": 1461 }, { "epoch": 0.07, "learning_rate": 1.9897478667606633e-05, "loss": 0.9836, "step": 1462 }, { "epoch": 0.07, "learning_rate": 1.9897243526059618e-05, "loss": 1.0084, "step": 1463 }, { "epoch": 0.07, "learning_rate": 1.9897008116555265e-05, "loss": 0.9795, "step": 1464 }, { "epoch": 0.07, "learning_rate": 1.9896772439099944e-05, "loss": 1.0192, "step": 1465 }, { "epoch": 0.07, "learning_rate": 1.989653649370004e-05, "loss": 1.0682, "step": 1466 }, { "epoch": 0.07, "learning_rate": 1.9896300280361944e-05, "loss": 1.0395, "step": 1467 }, { "epoch": 0.07, "learning_rate": 1.9896063799092044e-05, "loss": 1.1432, "step": 1468 }, { "epoch": 0.07, "learning_rate": 1.9895827049896743e-05, "loss": 0.9347, "step": 1469 }, { "epoch": 0.07, "learning_rate": 1.9895590032782458e-05, "loss": 0.9807, "step": 1470 }, { "epoch": 0.07, "learning_rate": 1.9895352747755595e-05, "loss": 1.01, "step": 1471 }, { "epoch": 0.07, "learning_rate": 1.9895115194822585e-05, "loss": 1.0861, "step": 1472 }, { "epoch": 0.07, "learning_rate": 1.989487737398986e-05, "loss": 0.9834, "step": 1473 }, { "epoch": 0.07, "learning_rate": 1.989463928526386e-05, "loss": 1.0052, "step": 1474 }, { "epoch": 0.07, "learning_rate": 1.9894400928651027e-05, "loss": 1.1161, "step": 1475 }, { "epoch": 0.07, "learning_rate": 1.9894162304157815e-05, "loss": 1.024, "step": 1476 }, { "epoch": 0.08, "learning_rate": 1.9893923411790684e-05, "loss": 1.1368, "step": 1477 }, { "epoch": 0.08, "learning_rate": 1.9893684251556102e-05, "loss": 1.1113, "step": 1478 }, { "epoch": 0.08, "learning_rate": 1.9893444823460548e-05, "loss": 1.0486, "step": 1479 }, { "epoch": 0.08, "learning_rate": 1.98932051275105e-05, "loss": 1.0804, "step": 1480 }, { "epoch": 0.08, "learning_rate": 1.989296516371245e-05, "loss": 0.9521, "step": 1481 }, { "epoch": 0.08, "learning_rate": 1.9892724932072893e-05, "loss": 1.0304, "step": 1482 }, { "epoch": 0.08, "learning_rate": 1.989248443259833e-05, "loss": 1.0043, "step": 1483 }, { "epoch": 0.08, "learning_rate": 1.989224366529528e-05, "loss": 1.0645, "step": 1484 }, { "epoch": 0.08, "learning_rate": 1.9892002630170254e-05, "loss": 0.9926, "step": 1485 }, { "epoch": 0.08, "learning_rate": 1.9891761327229785e-05, "loss": 0.9899, "step": 1486 }, { "epoch": 0.08, "learning_rate": 1.9891519756480398e-05, "loss": 1.1496, "step": 1487 }, { "epoch": 0.08, "learning_rate": 1.989127791792864e-05, "loss": 1.1077, "step": 1488 }, { "epoch": 0.08, "learning_rate": 1.9891035811581053e-05, "loss": 1.0173, "step": 1489 }, { "epoch": 0.08, "learning_rate": 1.98907934374442e-05, "loss": 0.9975, "step": 1490 }, { "epoch": 0.08, "learning_rate": 1.9890550795524635e-05, "loss": 1.0761, "step": 1491 }, { "epoch": 0.08, "learning_rate": 1.9890307885828927e-05, "loss": 1.0614, "step": 1492 }, { "epoch": 0.08, "learning_rate": 1.9890064708363658e-05, "loss": 1.0249, "step": 1493 }, { "epoch": 0.08, "learning_rate": 1.9889821263135408e-05, "loss": 1.0416, "step": 1494 }, { "epoch": 0.08, "learning_rate": 1.9889577550150773e-05, "loss": 1.0993, "step": 1495 }, { "epoch": 0.08, "learning_rate": 1.9889333569416343e-05, "loss": 1.0206, "step": 1496 }, { "epoch": 0.08, "learning_rate": 1.988908932093873e-05, "loss": 1.0336, "step": 1497 }, { "epoch": 0.08, "learning_rate": 1.9888844804724545e-05, "loss": 1.0883, "step": 1498 }, { "epoch": 0.08, "learning_rate": 1.988860002078041e-05, "loss": 1.0143, "step": 1499 }, { "epoch": 0.08, "learning_rate": 1.9888354969112945e-05, "loss": 0.9705, "step": 1500 }, { "epoch": 0.08, "learning_rate": 1.988810964972879e-05, "loss": 1.0905, "step": 1501 }, { "epoch": 0.08, "learning_rate": 1.988786406263459e-05, "loss": 1.0503, "step": 1502 }, { "epoch": 0.08, "learning_rate": 1.9887618207836988e-05, "loss": 0.976, "step": 1503 }, { "epoch": 0.08, "learning_rate": 1.9887372085342644e-05, "loss": 1.0766, "step": 1504 }, { "epoch": 0.08, "learning_rate": 1.988712569515822e-05, "loss": 1.0528, "step": 1505 }, { "epoch": 0.08, "learning_rate": 1.9886879037290385e-05, "loss": 1.0795, "step": 1506 }, { "epoch": 0.08, "learning_rate": 1.9886632111745822e-05, "loss": 1.0917, "step": 1507 }, { "epoch": 0.08, "learning_rate": 1.988638491853121e-05, "loss": 1.0659, "step": 1508 }, { "epoch": 0.08, "learning_rate": 1.9886137457653247e-05, "loss": 1.0369, "step": 1509 }, { "epoch": 0.08, "learning_rate": 1.9885889729118628e-05, "loss": 1.0583, "step": 1510 }, { "epoch": 0.08, "learning_rate": 1.9885641732934066e-05, "loss": 0.9923, "step": 1511 }, { "epoch": 0.08, "learning_rate": 1.988539346910627e-05, "loss": 1.0728, "step": 1512 }, { "epoch": 0.08, "learning_rate": 1.988514493764196e-05, "loss": 1.1469, "step": 1513 }, { "epoch": 0.08, "learning_rate": 1.988489613854787e-05, "loss": 0.9911, "step": 1514 }, { "epoch": 0.08, "learning_rate": 1.9884647071830736e-05, "loss": 0.9809, "step": 1515 }, { "epoch": 0.08, "learning_rate": 1.9884397737497293e-05, "loss": 1.051, "step": 1516 }, { "epoch": 0.08, "learning_rate": 1.9884148135554304e-05, "loss": 0.9957, "step": 1517 }, { "epoch": 0.08, "learning_rate": 1.9883898266008518e-05, "loss": 1.0161, "step": 1518 }, { "epoch": 0.08, "learning_rate": 1.98836481288667e-05, "loss": 1.0781, "step": 1519 }, { "epoch": 0.08, "learning_rate": 1.9883397724135628e-05, "loss": 1.1032, "step": 1520 }, { "epoch": 0.08, "learning_rate": 1.9883147051822073e-05, "loss": 1.0318, "step": 1521 }, { "epoch": 0.08, "learning_rate": 1.988289611193283e-05, "loss": 1.0215, "step": 1522 }, { "epoch": 0.08, "learning_rate": 1.988264490447469e-05, "loss": 0.9861, "step": 1523 }, { "epoch": 0.08, "learning_rate": 1.988239342945445e-05, "loss": 1.0961, "step": 1524 }, { "epoch": 0.08, "learning_rate": 1.9882141686878925e-05, "loss": 1.0262, "step": 1525 }, { "epoch": 0.08, "learning_rate": 1.9881889676754924e-05, "loss": 1.1342, "step": 1526 }, { "epoch": 0.08, "learning_rate": 1.9881637399089277e-05, "loss": 1.0911, "step": 1527 }, { "epoch": 0.08, "learning_rate": 1.988138485388881e-05, "loss": 1.0782, "step": 1528 }, { "epoch": 0.08, "learning_rate": 1.988113204116036e-05, "loss": 0.9669, "step": 1529 }, { "epoch": 0.08, "learning_rate": 1.9880878960910772e-05, "loss": 1.0081, "step": 1530 }, { "epoch": 0.08, "learning_rate": 1.9880625613146902e-05, "loss": 1.0313, "step": 1531 }, { "epoch": 0.08, "learning_rate": 1.98803719978756e-05, "loss": 1.0602, "step": 1532 }, { "epoch": 0.08, "learning_rate": 1.9880118115103743e-05, "loss": 1.1197, "step": 1533 }, { "epoch": 0.08, "learning_rate": 1.98798639648382e-05, "loss": 1.0691, "step": 1534 }, { "epoch": 0.08, "learning_rate": 1.987960954708585e-05, "loss": 1.0297, "step": 1535 }, { "epoch": 0.08, "learning_rate": 1.987935486185358e-05, "loss": 1.0906, "step": 1536 }, { "epoch": 0.08, "learning_rate": 1.987909990914829e-05, "loss": 1.0443, "step": 1537 }, { "epoch": 0.08, "learning_rate": 1.987884468897688e-05, "loss": 1.0419, "step": 1538 }, { "epoch": 0.08, "learning_rate": 1.987858920134626e-05, "loss": 1.1149, "step": 1539 }, { "epoch": 0.08, "learning_rate": 1.9878333446263353e-05, "loss": 1.0552, "step": 1540 }, { "epoch": 0.08, "learning_rate": 1.987807742373507e-05, "loss": 1.1766, "step": 1541 }, { "epoch": 0.08, "learning_rate": 1.9877821133768355e-05, "loss": 0.9827, "step": 1542 }, { "epoch": 0.08, "learning_rate": 1.9877564576370138e-05, "loss": 1.0332, "step": 1543 }, { "epoch": 0.08, "learning_rate": 1.9877307751547373e-05, "loss": 1.0695, "step": 1544 }, { "epoch": 0.08, "learning_rate": 1.9877050659307007e-05, "loss": 1.0464, "step": 1545 }, { "epoch": 0.08, "learning_rate": 1.9876793299656005e-05, "loss": 1.0629, "step": 1546 }, { "epoch": 0.08, "learning_rate": 1.987653567260133e-05, "loss": 1.0466, "step": 1547 }, { "epoch": 0.08, "learning_rate": 1.987627777814996e-05, "loss": 1.0553, "step": 1548 }, { "epoch": 0.08, "learning_rate": 1.9876019616308876e-05, "loss": 0.9663, "step": 1549 }, { "epoch": 0.08, "learning_rate": 1.987576118708507e-05, "loss": 1.0319, "step": 1550 }, { "epoch": 0.08, "learning_rate": 1.9875502490485533e-05, "loss": 1.0628, "step": 1551 }, { "epoch": 0.08, "learning_rate": 1.9875243526517276e-05, "loss": 1.0019, "step": 1552 }, { "epoch": 0.08, "learning_rate": 1.9874984295187308e-05, "loss": 1.0094, "step": 1553 }, { "epoch": 0.08, "learning_rate": 1.9874724796502645e-05, "loss": 1.1035, "step": 1554 }, { "epoch": 0.08, "learning_rate": 1.9874465030470315e-05, "loss": 1.1677, "step": 1555 }, { "epoch": 0.08, "learning_rate": 1.9874204997097347e-05, "loss": 1.1106, "step": 1556 }, { "epoch": 0.08, "learning_rate": 1.9873944696390787e-05, "loss": 0.9952, "step": 1557 }, { "epoch": 0.08, "learning_rate": 1.9873684128357677e-05, "loss": 0.9113, "step": 1558 }, { "epoch": 0.08, "learning_rate": 1.9873423293005076e-05, "loss": 1.0377, "step": 1559 }, { "epoch": 0.08, "learning_rate": 1.9873162190340044e-05, "loss": 1.0315, "step": 1560 }, { "epoch": 0.08, "learning_rate": 1.987290082036965e-05, "loss": 1.0741, "step": 1561 }, { "epoch": 0.08, "learning_rate": 1.9872639183100964e-05, "loss": 0.9543, "step": 1562 }, { "epoch": 0.08, "learning_rate": 1.9872377278541083e-05, "loss": 1.0388, "step": 1563 }, { "epoch": 0.08, "learning_rate": 1.9872115106697088e-05, "loss": 1.0586, "step": 1564 }, { "epoch": 0.08, "learning_rate": 1.9871852667576082e-05, "loss": 1.0074, "step": 1565 }, { "epoch": 0.08, "learning_rate": 1.9871589961185166e-05, "loss": 0.9543, "step": 1566 }, { "epoch": 0.08, "learning_rate": 1.9871326987531453e-05, "loss": 1.05, "step": 1567 }, { "epoch": 0.08, "learning_rate": 1.9871063746622066e-05, "loss": 1.0461, "step": 1568 }, { "epoch": 0.08, "learning_rate": 1.987080023846413e-05, "loss": 0.9436, "step": 1569 }, { "epoch": 0.08, "learning_rate": 1.9870536463064778e-05, "loss": 1.007, "step": 1570 }, { "epoch": 0.08, "learning_rate": 1.9870272420431155e-05, "loss": 1.0423, "step": 1571 }, { "epoch": 0.08, "learning_rate": 1.9870008110570403e-05, "loss": 1.0388, "step": 1572 }, { "epoch": 0.08, "learning_rate": 1.9869743533489687e-05, "loss": 1.0383, "step": 1573 }, { "epoch": 0.08, "learning_rate": 1.9869478689196164e-05, "loss": 1.1031, "step": 1574 }, { "epoch": 0.08, "learning_rate": 1.9869213577697004e-05, "loss": 1.067, "step": 1575 }, { "epoch": 0.08, "learning_rate": 1.986894819899939e-05, "loss": 1.0145, "step": 1576 }, { "epoch": 0.08, "learning_rate": 1.98686825531105e-05, "loss": 1.1196, "step": 1577 }, { "epoch": 0.08, "learning_rate": 1.986841664003753e-05, "loss": 1.0693, "step": 1578 }, { "epoch": 0.08, "learning_rate": 1.986815045978768e-05, "loss": 1.0427, "step": 1579 }, { "epoch": 0.08, "learning_rate": 1.986788401236815e-05, "loss": 1.0171, "step": 1580 }, { "epoch": 0.08, "learning_rate": 1.9867617297786165e-05, "loss": 1.144, "step": 1581 }, { "epoch": 0.08, "learning_rate": 1.986735031604894e-05, "loss": 1.197, "step": 1582 }, { "epoch": 0.08, "learning_rate": 1.98670830671637e-05, "loss": 0.9821, "step": 1583 }, { "epoch": 0.08, "learning_rate": 1.9866815551137683e-05, "loss": 1.0661, "step": 1584 }, { "epoch": 0.08, "learning_rate": 1.9866547767978135e-05, "loss": 1.1329, "step": 1585 }, { "epoch": 0.08, "learning_rate": 1.9866279717692306e-05, "loss": 1.0083, "step": 1586 }, { "epoch": 0.08, "learning_rate": 1.9866011400287445e-05, "loss": 0.9608, "step": 1587 }, { "epoch": 0.08, "learning_rate": 1.9865742815770825e-05, "loss": 1.0396, "step": 1588 }, { "epoch": 0.08, "learning_rate": 1.9865473964149714e-05, "loss": 0.9839, "step": 1589 }, { "epoch": 0.08, "learning_rate": 1.9865204845431393e-05, "loss": 1.1854, "step": 1590 }, { "epoch": 0.08, "learning_rate": 1.9864935459623143e-05, "loss": 1.0293, "step": 1591 }, { "epoch": 0.08, "learning_rate": 1.9864665806732262e-05, "loss": 1.0636, "step": 1592 }, { "epoch": 0.08, "learning_rate": 1.986439588676605e-05, "loss": 1.0011, "step": 1593 }, { "epoch": 0.08, "learning_rate": 1.9864125699731814e-05, "loss": 1.0171, "step": 1594 }, { "epoch": 0.08, "learning_rate": 1.9863855245636873e-05, "loss": 1.0394, "step": 1595 }, { "epoch": 0.08, "learning_rate": 1.986358452448854e-05, "loss": 0.9771, "step": 1596 }, { "epoch": 0.08, "learning_rate": 1.9863313536294153e-05, "loss": 1.0949, "step": 1597 }, { "epoch": 0.08, "learning_rate": 1.9863042281061044e-05, "loss": 1.1072, "step": 1598 }, { "epoch": 0.08, "learning_rate": 1.986277075879656e-05, "loss": 0.9888, "step": 1599 }, { "epoch": 0.08, "learning_rate": 1.986249896950805e-05, "loss": 1.0126, "step": 1600 }, { "epoch": 0.08, "learning_rate": 1.9862226913202875e-05, "loss": 1.0461, "step": 1601 }, { "epoch": 0.08, "learning_rate": 1.9861954589888396e-05, "loss": 1.0473, "step": 1602 }, { "epoch": 0.08, "learning_rate": 1.986168199957199e-05, "loss": 1.0146, "step": 1603 }, { "epoch": 0.08, "learning_rate": 1.9861409142261037e-05, "loss": 0.9731, "step": 1604 }, { "epoch": 0.08, "learning_rate": 1.9861136017962923e-05, "loss": 1.0985, "step": 1605 }, { "epoch": 0.08, "learning_rate": 1.9860862626685042e-05, "loss": 1.0577, "step": 1606 }, { "epoch": 0.08, "learning_rate": 1.9860588968434796e-05, "loss": 1.0463, "step": 1607 }, { "epoch": 0.08, "learning_rate": 1.9860315043219597e-05, "loss": 1.0771, "step": 1608 }, { "epoch": 0.08, "learning_rate": 1.986004085104686e-05, "loss": 1.046, "step": 1609 }, { "epoch": 0.08, "learning_rate": 1.9859766391924e-05, "loss": 1.1452, "step": 1610 }, { "epoch": 0.08, "learning_rate": 1.9859491665858462e-05, "loss": 1.0754, "step": 1611 }, { "epoch": 0.08, "learning_rate": 1.9859216672857672e-05, "loss": 1.0524, "step": 1612 }, { "epoch": 0.08, "learning_rate": 1.9858941412929083e-05, "loss": 1.0984, "step": 1613 }, { "epoch": 0.08, "learning_rate": 1.985866588608014e-05, "loss": 1.0645, "step": 1614 }, { "epoch": 0.08, "learning_rate": 1.9858390092318313e-05, "loss": 1.0931, "step": 1615 }, { "epoch": 0.08, "learning_rate": 1.985811403165106e-05, "loss": 1.0726, "step": 1616 }, { "epoch": 0.08, "learning_rate": 1.9857837704085857e-05, "loss": 0.9615, "step": 1617 }, { "epoch": 0.08, "learning_rate": 1.9857561109630188e-05, "loss": 1.1266, "step": 1618 }, { "epoch": 0.08, "learning_rate": 1.9857284248291537e-05, "loss": 1.1344, "step": 1619 }, { "epoch": 0.08, "learning_rate": 1.9857007120077405e-05, "loss": 1.0604, "step": 1620 }, { "epoch": 0.08, "learning_rate": 1.985672972499529e-05, "loss": 1.0255, "step": 1621 }, { "epoch": 0.08, "learning_rate": 1.9856452063052706e-05, "loss": 1.0316, "step": 1622 }, { "epoch": 0.08, "learning_rate": 1.985617413425717e-05, "loss": 1.012, "step": 1623 }, { "epoch": 0.08, "learning_rate": 1.9855895938616203e-05, "loss": 1.0295, "step": 1624 }, { "epoch": 0.08, "learning_rate": 1.985561747613734e-05, "loss": 1.0714, "step": 1625 }, { "epoch": 0.08, "learning_rate": 1.985533874682812e-05, "loss": 1.0184, "step": 1626 }, { "epoch": 0.08, "learning_rate": 1.985505975069609e-05, "loss": 1.0123, "step": 1627 }, { "epoch": 0.08, "learning_rate": 1.98547804877488e-05, "loss": 1.1665, "step": 1628 }, { "epoch": 0.08, "learning_rate": 1.985450095799381e-05, "loss": 0.998, "step": 1629 }, { "epoch": 0.08, "learning_rate": 1.9854221161438697e-05, "loss": 0.985, "step": 1630 }, { "epoch": 0.08, "learning_rate": 1.985394109809103e-05, "loss": 1.0515, "step": 1631 }, { "epoch": 0.08, "learning_rate": 1.9853660767958386e-05, "loss": 1.0151, "step": 1632 }, { "epoch": 0.08, "learning_rate": 1.9853380171048364e-05, "loss": 1.0485, "step": 1633 }, { "epoch": 0.08, "learning_rate": 1.9853099307368553e-05, "loss": 1.0467, "step": 1634 }, { "epoch": 0.08, "learning_rate": 1.9852818176926565e-05, "loss": 1.0531, "step": 1635 }, { "epoch": 0.08, "learning_rate": 1.985253677973001e-05, "loss": 1.0459, "step": 1636 }, { "epoch": 0.08, "learning_rate": 1.9852255115786497e-05, "loss": 1.058, "step": 1637 }, { "epoch": 0.08, "learning_rate": 1.9851973185103664e-05, "loss": 1.0465, "step": 1638 }, { "epoch": 0.08, "learning_rate": 1.9851690987689135e-05, "loss": 1.0184, "step": 1639 }, { "epoch": 0.08, "learning_rate": 1.9851408523550556e-05, "loss": 1.1141, "step": 1640 }, { "epoch": 0.08, "learning_rate": 1.9851125792695572e-05, "loss": 1.04, "step": 1641 }, { "epoch": 0.08, "learning_rate": 1.9850842795131838e-05, "loss": 1.098, "step": 1642 }, { "epoch": 0.08, "learning_rate": 1.9850559530867018e-05, "loss": 1.0939, "step": 1643 }, { "epoch": 0.08, "learning_rate": 1.9850275999908778e-05, "loss": 1.0258, "step": 1644 }, { "epoch": 0.08, "learning_rate": 1.9849992202264793e-05, "loss": 1.0173, "step": 1645 }, { "epoch": 0.08, "learning_rate": 1.984970813794275e-05, "loss": 1.0405, "step": 1646 }, { "epoch": 0.08, "learning_rate": 1.9849423806950335e-05, "loss": 1.0376, "step": 1647 }, { "epoch": 0.08, "learning_rate": 1.9849139209295254e-05, "loss": 0.9503, "step": 1648 }, { "epoch": 0.08, "learning_rate": 1.9848854344985206e-05, "loss": 0.9719, "step": 1649 }, { "epoch": 0.08, "learning_rate": 1.9848569214027906e-05, "loss": 1.0734, "step": 1650 }, { "epoch": 0.08, "learning_rate": 1.9848283816431074e-05, "loss": 0.9858, "step": 1651 }, { "epoch": 0.08, "learning_rate": 1.984799815220243e-05, "loss": 1.0586, "step": 1652 }, { "epoch": 0.08, "learning_rate": 1.984771222134972e-05, "loss": 1.0709, "step": 1653 }, { "epoch": 0.08, "learning_rate": 1.9847426023880675e-05, "loss": 1.0063, "step": 1654 }, { "epoch": 0.08, "learning_rate": 1.9847139559803047e-05, "loss": 1.0524, "step": 1655 }, { "epoch": 0.08, "learning_rate": 1.984685282912459e-05, "loss": 1.0445, "step": 1656 }, { "epoch": 0.08, "learning_rate": 1.984656583185307e-05, "loss": 1.0765, "step": 1657 }, { "epoch": 0.08, "learning_rate": 1.984627856799626e-05, "loss": 1.0495, "step": 1658 }, { "epoch": 0.08, "learning_rate": 1.9845991037561928e-05, "loss": 1.0894, "step": 1659 }, { "epoch": 0.08, "learning_rate": 1.984570324055787e-05, "loss": 1.0522, "step": 1660 }, { "epoch": 0.08, "learning_rate": 1.9845415176991867e-05, "loss": 0.9446, "step": 1661 }, { "epoch": 0.08, "learning_rate": 1.9845126846871725e-05, "loss": 1.0602, "step": 1662 }, { "epoch": 0.08, "learning_rate": 1.9844838250205246e-05, "loss": 1.0661, "step": 1663 }, { "epoch": 0.08, "learning_rate": 1.9844549387000247e-05, "loss": 1.0102, "step": 1664 }, { "epoch": 0.08, "learning_rate": 1.9844260257264546e-05, "loss": 1.064, "step": 1665 }, { "epoch": 0.08, "learning_rate": 1.9843970861005975e-05, "loss": 1.0186, "step": 1666 }, { "epoch": 0.08, "learning_rate": 1.9843681198232363e-05, "loss": 0.9774, "step": 1667 }, { "epoch": 0.08, "learning_rate": 1.9843391268951557e-05, "loss": 1.0215, "step": 1668 }, { "epoch": 0.08, "learning_rate": 1.9843101073171405e-05, "loss": 1.0988, "step": 1669 }, { "epoch": 0.08, "learning_rate": 1.9842810610899762e-05, "loss": 1.0585, "step": 1670 }, { "epoch": 0.08, "learning_rate": 1.9842519882144496e-05, "loss": 1.0312, "step": 1671 }, { "epoch": 0.08, "learning_rate": 1.9842228886913472e-05, "loss": 0.9621, "step": 1672 }, { "epoch": 0.08, "learning_rate": 1.9841937625214576e-05, "loss": 1.1106, "step": 1673 }, { "epoch": 0.09, "learning_rate": 1.9841646097055687e-05, "loss": 1.0247, "step": 1674 }, { "epoch": 0.09, "learning_rate": 1.98413543024447e-05, "loss": 1.0309, "step": 1675 }, { "epoch": 0.09, "learning_rate": 1.984106224138952e-05, "loss": 1.0755, "step": 1676 }, { "epoch": 0.09, "learning_rate": 1.9840769913898047e-05, "loss": 1.063, "step": 1677 }, { "epoch": 0.09, "learning_rate": 1.98404773199782e-05, "loss": 0.9623, "step": 1678 }, { "epoch": 0.09, "learning_rate": 1.98401844596379e-05, "loss": 0.9933, "step": 1679 }, { "epoch": 0.09, "learning_rate": 1.983989133288507e-05, "loss": 1.013, "step": 1680 }, { "epoch": 0.09, "learning_rate": 1.9839597939727656e-05, "loss": 1.0693, "step": 1681 }, { "epoch": 0.09, "learning_rate": 1.9839304280173592e-05, "loss": 1.0254, "step": 1682 }, { "epoch": 0.09, "learning_rate": 1.9839010354230837e-05, "loss": 0.9701, "step": 1683 }, { "epoch": 0.09, "learning_rate": 1.983871616190734e-05, "loss": 1.0025, "step": 1684 }, { "epoch": 0.09, "learning_rate": 1.9838421703211073e-05, "loss": 1.0696, "step": 1685 }, { "epoch": 0.09, "learning_rate": 1.9838126978150005e-05, "loss": 1.0538, "step": 1686 }, { "epoch": 0.09, "learning_rate": 1.9837831986732116e-05, "loss": 1.0457, "step": 1687 }, { "epoch": 0.09, "learning_rate": 1.9837536728965394e-05, "loss": 1.0305, "step": 1688 }, { "epoch": 0.09, "learning_rate": 1.983724120485783e-05, "loss": 1.0616, "step": 1689 }, { "epoch": 0.09, "learning_rate": 1.983694541441743e-05, "loss": 1.1748, "step": 1690 }, { "epoch": 0.09, "learning_rate": 1.9836649357652194e-05, "loss": 1.1921, "step": 1691 }, { "epoch": 0.09, "learning_rate": 1.9836353034570142e-05, "loss": 0.932, "step": 1692 }, { "epoch": 0.09, "learning_rate": 1.9836056445179298e-05, "loss": 1.0797, "step": 1693 }, { "epoch": 0.09, "learning_rate": 1.9835759589487692e-05, "loss": 1.018, "step": 1694 }, { "epoch": 0.09, "learning_rate": 1.9835462467503358e-05, "loss": 1.0344, "step": 1695 }, { "epoch": 0.09, "learning_rate": 1.9835165079234342e-05, "loss": 1.018, "step": 1696 }, { "epoch": 0.09, "learning_rate": 1.9834867424688695e-05, "loss": 0.9636, "step": 1697 }, { "epoch": 0.09, "learning_rate": 1.9834569503874476e-05, "loss": 1.0645, "step": 1698 }, { "epoch": 0.09, "learning_rate": 1.9834271316799753e-05, "loss": 1.0832, "step": 1699 }, { "epoch": 0.09, "learning_rate": 1.9833972863472593e-05, "loss": 1.0688, "step": 1700 }, { "epoch": 0.09, "learning_rate": 1.9833674143901084e-05, "loss": 1.0195, "step": 1701 }, { "epoch": 0.09, "learning_rate": 1.9833375158093306e-05, "loss": 0.9785, "step": 1702 }, { "epoch": 0.09, "learning_rate": 1.983307590605736e-05, "loss": 1.0839, "step": 1703 }, { "epoch": 0.09, "learning_rate": 1.9832776387801345e-05, "loss": 1.0801, "step": 1704 }, { "epoch": 0.09, "learning_rate": 1.983247660333337e-05, "loss": 1.0378, "step": 1705 }, { "epoch": 0.09, "learning_rate": 1.983217655266155e-05, "loss": 1.0314, "step": 1706 }, { "epoch": 0.09, "learning_rate": 1.9831876235794016e-05, "loss": 1.1348, "step": 1707 }, { "epoch": 0.09, "learning_rate": 1.983157565273889e-05, "loss": 1.048, "step": 1708 }, { "epoch": 0.09, "learning_rate": 1.9831274803504314e-05, "loss": 0.9709, "step": 1709 }, { "epoch": 0.09, "learning_rate": 1.983097368809843e-05, "loss": 1.1, "step": 1710 }, { "epoch": 0.09, "learning_rate": 1.9830672306529392e-05, "loss": 1.0697, "step": 1711 }, { "epoch": 0.09, "learning_rate": 1.9830370658805364e-05, "loss": 0.9718, "step": 1712 }, { "epoch": 0.09, "learning_rate": 1.9830068744934507e-05, "loss": 1.127, "step": 1713 }, { "epoch": 0.09, "learning_rate": 1.9829766564925e-05, "loss": 1.0949, "step": 1714 }, { "epoch": 0.09, "learning_rate": 1.982946411878502e-05, "loss": 1.0564, "step": 1715 }, { "epoch": 0.09, "learning_rate": 1.9829161406522753e-05, "loss": 1.0758, "step": 1716 }, { "epoch": 0.09, "learning_rate": 1.98288584281464e-05, "loss": 1.0367, "step": 1717 }, { "epoch": 0.09, "learning_rate": 1.9828555183664165e-05, "loss": 1.1181, "step": 1718 }, { "epoch": 0.09, "learning_rate": 1.9828251673084253e-05, "loss": 0.9677, "step": 1719 }, { "epoch": 0.09, "learning_rate": 1.9827947896414886e-05, "loss": 1.0262, "step": 1720 }, { "epoch": 0.09, "learning_rate": 1.982764385366428e-05, "loss": 1.1259, "step": 1721 }, { "epoch": 0.09, "learning_rate": 1.9827339544840678e-05, "loss": 1.029, "step": 1722 }, { "epoch": 0.09, "learning_rate": 1.9827034969952313e-05, "loss": 1.0887, "step": 1723 }, { "epoch": 0.09, "learning_rate": 1.9826730129007427e-05, "loss": 1.0314, "step": 1724 }, { "epoch": 0.09, "learning_rate": 1.982642502201428e-05, "loss": 1.0226, "step": 1725 }, { "epoch": 0.09, "learning_rate": 1.9826119648981132e-05, "loss": 1.0087, "step": 1726 }, { "epoch": 0.09, "learning_rate": 1.9825814009916246e-05, "loss": 1.0162, "step": 1727 }, { "epoch": 0.09, "learning_rate": 1.9825508104827903e-05, "loss": 1.0906, "step": 1728 }, { "epoch": 0.09, "learning_rate": 1.9825201933724375e-05, "loss": 1.0637, "step": 1729 }, { "epoch": 0.09, "learning_rate": 1.9824895496613964e-05, "loss": 1.0485, "step": 1730 }, { "epoch": 0.09, "learning_rate": 1.9824588793504958e-05, "loss": 1.0305, "step": 1731 }, { "epoch": 0.09, "learning_rate": 1.9824281824405662e-05, "loss": 0.9864, "step": 1732 }, { "epoch": 0.09, "learning_rate": 1.982397458932439e-05, "loss": 0.9796, "step": 1733 }, { "epoch": 0.09, "learning_rate": 1.9823667088269457e-05, "loss": 1.0716, "step": 1734 }, { "epoch": 0.09, "learning_rate": 1.9823359321249187e-05, "loss": 1.0545, "step": 1735 }, { "epoch": 0.09, "learning_rate": 1.9823051288271918e-05, "loss": 1.0022, "step": 1736 }, { "epoch": 0.09, "learning_rate": 1.9822742989345986e-05, "loss": 1.0395, "step": 1737 }, { "epoch": 0.09, "learning_rate": 1.9822434424479735e-05, "loss": 1.1155, "step": 1738 }, { "epoch": 0.09, "learning_rate": 1.9822125593681526e-05, "loss": 1.0761, "step": 1739 }, { "epoch": 0.09, "learning_rate": 1.9821816496959713e-05, "loss": 1.0737, "step": 1740 }, { "epoch": 0.09, "learning_rate": 1.982150713432267e-05, "loss": 1.0524, "step": 1741 }, { "epoch": 0.09, "learning_rate": 1.982119750577877e-05, "loss": 0.9075, "step": 1742 }, { "epoch": 0.09, "learning_rate": 1.9820887611336397e-05, "loss": 1.1051, "step": 1743 }, { "epoch": 0.09, "learning_rate": 1.982057745100394e-05, "loss": 1.0215, "step": 1744 }, { "epoch": 0.09, "learning_rate": 1.9820267024789795e-05, "loss": 1.1129, "step": 1745 }, { "epoch": 0.09, "learning_rate": 1.981995633270237e-05, "loss": 0.9761, "step": 1746 }, { "epoch": 0.09, "learning_rate": 1.981964537475008e-05, "loss": 1.0906, "step": 1747 }, { "epoch": 0.09, "learning_rate": 1.981933415094133e-05, "loss": 0.9283, "step": 1748 }, { "epoch": 0.09, "learning_rate": 1.9819022661284563e-05, "loss": 1.0473, "step": 1749 }, { "epoch": 0.09, "learning_rate": 1.9818710905788195e-05, "loss": 1.0252, "step": 1750 }, { "epoch": 0.09, "learning_rate": 1.9818398884460685e-05, "loss": 1.0216, "step": 1751 }, { "epoch": 0.09, "learning_rate": 1.9818086597310464e-05, "loss": 0.9811, "step": 1752 }, { "epoch": 0.09, "learning_rate": 1.9817774044345994e-05, "loss": 1.0643, "step": 1753 }, { "epoch": 0.09, "learning_rate": 1.9817461225575744e-05, "loss": 0.9883, "step": 1754 }, { "epoch": 0.09, "learning_rate": 1.981714814100817e-05, "loss": 1.1211, "step": 1755 }, { "epoch": 0.09, "learning_rate": 1.9816834790651757e-05, "loss": 1.0262, "step": 1756 }, { "epoch": 0.09, "learning_rate": 1.9816521174514984e-05, "loss": 1.1196, "step": 1757 }, { "epoch": 0.09, "learning_rate": 1.9816207292606347e-05, "loss": 1.0921, "step": 1758 }, { "epoch": 0.09, "learning_rate": 1.981589314493434e-05, "loss": 0.9673, "step": 1759 }, { "epoch": 0.09, "learning_rate": 1.981557873150747e-05, "loss": 0.998, "step": 1760 }, { "epoch": 0.09, "learning_rate": 1.9815264052334247e-05, "loss": 1.0328, "step": 1761 }, { "epoch": 0.09, "learning_rate": 1.9814949107423195e-05, "loss": 1.0298, "step": 1762 }, { "epoch": 0.09, "learning_rate": 1.9814633896782836e-05, "loss": 1.0186, "step": 1763 }, { "epoch": 0.09, "learning_rate": 1.9814318420421707e-05, "loss": 1.025, "step": 1764 }, { "epoch": 0.09, "learning_rate": 1.9814002678348348e-05, "loss": 0.9973, "step": 1765 }, { "epoch": 0.09, "learning_rate": 1.9813686670571307e-05, "loss": 1.0393, "step": 1766 }, { "epoch": 0.09, "learning_rate": 1.9813370397099143e-05, "loss": 1.0194, "step": 1767 }, { "epoch": 0.09, "learning_rate": 1.9813053857940414e-05, "loss": 1.0978, "step": 1768 }, { "epoch": 0.09, "learning_rate": 1.9812737053103693e-05, "loss": 0.9918, "step": 1769 }, { "epoch": 0.09, "learning_rate": 1.9812419982597557e-05, "loss": 1.0739, "step": 1770 }, { "epoch": 0.09, "learning_rate": 1.9812102646430587e-05, "loss": 1.0898, "step": 1771 }, { "epoch": 0.09, "learning_rate": 1.9811785044611377e-05, "loss": 1.0569, "step": 1772 }, { "epoch": 0.09, "learning_rate": 1.981146717714853e-05, "loss": 1.084, "step": 1773 }, { "epoch": 0.09, "learning_rate": 1.9811149044050644e-05, "loss": 0.9487, "step": 1774 }, { "epoch": 0.09, "learning_rate": 1.9810830645326337e-05, "loss": 1.1286, "step": 1775 }, { "epoch": 0.09, "learning_rate": 1.981051198098423e-05, "loss": 0.9429, "step": 1776 }, { "epoch": 0.09, "learning_rate": 1.9810193051032945e-05, "loss": 0.991, "step": 1777 }, { "epoch": 0.09, "learning_rate": 1.9809873855481123e-05, "loss": 1.0519, "step": 1778 }, { "epoch": 0.09, "learning_rate": 1.9809554394337403e-05, "loss": 1.0302, "step": 1779 }, { "epoch": 0.09, "learning_rate": 1.9809234667610437e-05, "loss": 0.9532, "step": 1780 }, { "epoch": 0.09, "learning_rate": 1.9808914675308873e-05, "loss": 1.0453, "step": 1781 }, { "epoch": 0.09, "learning_rate": 1.9808594417441383e-05, "loss": 1.0417, "step": 1782 }, { "epoch": 0.09, "learning_rate": 1.9808273894016632e-05, "loss": 0.9173, "step": 1783 }, { "epoch": 0.09, "learning_rate": 1.9807953105043306e-05, "loss": 1.1086, "step": 1784 }, { "epoch": 0.09, "learning_rate": 1.980763205053008e-05, "loss": 0.9932, "step": 1785 }, { "epoch": 0.09, "learning_rate": 1.9807310730485652e-05, "loss": 1.0372, "step": 1786 }, { "epoch": 0.09, "learning_rate": 1.980698914491872e-05, "loss": 1.0852, "step": 1787 }, { "epoch": 0.09, "learning_rate": 1.980666729383799e-05, "loss": 1.0543, "step": 1788 }, { "epoch": 0.09, "learning_rate": 1.9806345177252178e-05, "loss": 1.0909, "step": 1789 }, { "epoch": 0.09, "learning_rate": 1.980602279517e-05, "loss": 1.0176, "step": 1790 }, { "epoch": 0.09, "learning_rate": 1.9805700147600193e-05, "loss": 0.9959, "step": 1791 }, { "epoch": 0.09, "learning_rate": 1.9805377234551482e-05, "loss": 1.1194, "step": 1792 }, { "epoch": 0.09, "learning_rate": 1.9805054056032617e-05, "loss": 1.0039, "step": 1793 }, { "epoch": 0.09, "learning_rate": 1.980473061205235e-05, "loss": 1.085, "step": 1794 }, { "epoch": 0.09, "learning_rate": 1.9804406902619423e-05, "loss": 1.0937, "step": 1795 }, { "epoch": 0.09, "learning_rate": 1.9804082927742617e-05, "loss": 0.9569, "step": 1796 }, { "epoch": 0.09, "learning_rate": 1.9803758687430693e-05, "loss": 1.0809, "step": 1797 }, { "epoch": 0.09, "learning_rate": 1.9803434181692434e-05, "loss": 1.0472, "step": 1798 }, { "epoch": 0.09, "learning_rate": 1.9803109410536624e-05, "loss": 0.9818, "step": 1799 }, { "epoch": 0.09, "learning_rate": 1.980278437397206e-05, "loss": 1.045, "step": 1800 }, { "epoch": 0.09, "learning_rate": 1.9802459072007535e-05, "loss": 0.9841, "step": 1801 }, { "epoch": 0.09, "learning_rate": 1.980213350465186e-05, "loss": 0.9514, "step": 1802 }, { "epoch": 0.09, "learning_rate": 1.9801807671913852e-05, "loss": 1.0331, "step": 1803 }, { "epoch": 0.09, "learning_rate": 1.9801481573802326e-05, "loss": 1.0599, "step": 1804 }, { "epoch": 0.09, "learning_rate": 1.9801155210326114e-05, "loss": 1.0089, "step": 1805 }, { "epoch": 0.09, "learning_rate": 1.9800828581494053e-05, "loss": 1.1105, "step": 1806 }, { "epoch": 0.09, "learning_rate": 1.980050168731499e-05, "loss": 1.076, "step": 1807 }, { "epoch": 0.09, "learning_rate": 1.9800174527797765e-05, "loss": 1.0484, "step": 1808 }, { "epoch": 0.09, "learning_rate": 1.9799847102951246e-05, "loss": 0.9605, "step": 1809 }, { "epoch": 0.09, "learning_rate": 1.979951941278429e-05, "loss": 1.0233, "step": 1810 }, { "epoch": 0.09, "learning_rate": 1.9799191457305767e-05, "loss": 1.0592, "step": 1811 }, { "epoch": 0.09, "learning_rate": 1.979886323652457e-05, "loss": 0.8958, "step": 1812 }, { "epoch": 0.09, "learning_rate": 1.9798534750449568e-05, "loss": 1.0848, "step": 1813 }, { "epoch": 0.09, "learning_rate": 1.979820599908967e-05, "loss": 1.1109, "step": 1814 }, { "epoch": 0.09, "learning_rate": 1.9797876982453763e-05, "loss": 1.0521, "step": 1815 }, { "epoch": 0.09, "learning_rate": 1.9797547700550762e-05, "loss": 1.0103, "step": 1816 }, { "epoch": 0.09, "learning_rate": 1.9797218153389583e-05, "loss": 1.1819, "step": 1817 }, { "epoch": 0.09, "learning_rate": 1.9796888340979144e-05, "loss": 1.0742, "step": 1818 }, { "epoch": 0.09, "learning_rate": 1.979655826332838e-05, "loss": 1.0688, "step": 1819 }, { "epoch": 0.09, "learning_rate": 1.9796227920446217e-05, "loss": 1.07, "step": 1820 }, { "epoch": 0.09, "learning_rate": 1.9795897312341614e-05, "loss": 0.9783, "step": 1821 }, { "epoch": 0.09, "learning_rate": 1.9795566439023503e-05, "loss": 1.0793, "step": 1822 }, { "epoch": 0.09, "learning_rate": 1.9795235300500858e-05, "loss": 1.0581, "step": 1823 }, { "epoch": 0.09, "learning_rate": 1.979490389678264e-05, "loss": 1.0413, "step": 1824 }, { "epoch": 0.09, "learning_rate": 1.9794572227877818e-05, "loss": 0.9957, "step": 1825 }, { "epoch": 0.09, "learning_rate": 1.9794240293795374e-05, "loss": 1.1102, "step": 1826 }, { "epoch": 0.09, "learning_rate": 1.9793908094544295e-05, "loss": 1.031, "step": 1827 }, { "epoch": 0.09, "learning_rate": 1.979357563013357e-05, "loss": 1.0435, "step": 1828 }, { "epoch": 0.09, "learning_rate": 1.979324290057221e-05, "loss": 1.063, "step": 1829 }, { "epoch": 0.09, "learning_rate": 1.9792909905869214e-05, "loss": 0.9657, "step": 1830 }, { "epoch": 0.09, "learning_rate": 1.9792576646033603e-05, "loss": 1.0156, "step": 1831 }, { "epoch": 0.09, "learning_rate": 1.9792243121074396e-05, "loss": 1.0503, "step": 1832 }, { "epoch": 0.09, "learning_rate": 1.979190933100063e-05, "loss": 1.1145, "step": 1833 }, { "epoch": 0.09, "learning_rate": 1.979157527582133e-05, "loss": 1.0328, "step": 1834 }, { "epoch": 0.09, "learning_rate": 1.9791240955545546e-05, "loss": 1.0803, "step": 1835 }, { "epoch": 0.09, "learning_rate": 1.9790906370182336e-05, "loss": 1.0691, "step": 1836 }, { "epoch": 0.09, "learning_rate": 1.9790571519740752e-05, "loss": 1.0875, "step": 1837 }, { "epoch": 0.09, "learning_rate": 1.9790236404229856e-05, "loss": 1.1047, "step": 1838 }, { "epoch": 0.09, "learning_rate": 1.978990102365873e-05, "loss": 1.0726, "step": 1839 }, { "epoch": 0.09, "learning_rate": 1.9789565378036448e-05, "loss": 1.1195, "step": 1840 }, { "epoch": 0.09, "learning_rate": 1.97892294673721e-05, "loss": 1.077, "step": 1841 }, { "epoch": 0.09, "learning_rate": 1.9788893291674776e-05, "loss": 0.8532, "step": 1842 }, { "epoch": 0.09, "learning_rate": 1.978855685095358e-05, "loss": 1.0346, "step": 1843 }, { "epoch": 0.09, "learning_rate": 1.978822014521762e-05, "loss": 1.0648, "step": 1844 }, { "epoch": 0.09, "learning_rate": 1.9787883174476018e-05, "loss": 1.0325, "step": 1845 }, { "epoch": 0.09, "learning_rate": 1.9787545938737894e-05, "loss": 0.9751, "step": 1846 }, { "epoch": 0.09, "learning_rate": 1.9787208438012376e-05, "loss": 1.0377, "step": 1847 }, { "epoch": 0.09, "learning_rate": 1.9786870672308597e-05, "loss": 1.0824, "step": 1848 }, { "epoch": 0.09, "learning_rate": 1.9786532641635713e-05, "loss": 1.1003, "step": 1849 }, { "epoch": 0.09, "learning_rate": 1.9786194346002866e-05, "loss": 1.0254, "step": 1850 }, { "epoch": 0.09, "learning_rate": 1.978585578541922e-05, "loss": 1.0253, "step": 1851 }, { "epoch": 0.09, "learning_rate": 1.978551695989394e-05, "loss": 1.0887, "step": 1852 }, { "epoch": 0.09, "learning_rate": 1.9785177869436198e-05, "loss": 1.0968, "step": 1853 }, { "epoch": 0.09, "learning_rate": 1.978483851405518e-05, "loss": 1.1157, "step": 1854 }, { "epoch": 0.09, "learning_rate": 1.9784498893760063e-05, "loss": 1.0715, "step": 1855 }, { "epoch": 0.09, "learning_rate": 1.9784159008560055e-05, "loss": 0.9908, "step": 1856 }, { "epoch": 0.09, "learning_rate": 1.978381885846435e-05, "loss": 1.0055, "step": 1857 }, { "epoch": 0.09, "learning_rate": 1.9783478443482153e-05, "loss": 1.187, "step": 1858 }, { "epoch": 0.09, "learning_rate": 1.9783137763622687e-05, "loss": 1.0385, "step": 1859 }, { "epoch": 0.09, "learning_rate": 1.978279681889518e-05, "loss": 1.004, "step": 1860 }, { "epoch": 0.09, "learning_rate": 1.9782455609308854e-05, "loss": 1.1626, "step": 1861 }, { "epoch": 0.09, "learning_rate": 1.978211413487295e-05, "loss": 1.0612, "step": 1862 }, { "epoch": 0.09, "learning_rate": 1.9781772395596714e-05, "loss": 1.064, "step": 1863 }, { "epoch": 0.09, "learning_rate": 1.9781430391489394e-05, "loss": 0.9982, "step": 1864 }, { "epoch": 0.09, "learning_rate": 1.978108812256026e-05, "loss": 1.0917, "step": 1865 }, { "epoch": 0.09, "learning_rate": 1.9780745588818567e-05, "loss": 1.093, "step": 1866 }, { "epoch": 0.09, "learning_rate": 1.978040279027359e-05, "loss": 1.0983, "step": 1867 }, { "epoch": 0.09, "learning_rate": 1.978005972693462e-05, "loss": 1.0023, "step": 1868 }, { "epoch": 0.09, "learning_rate": 1.977971639881093e-05, "loss": 0.9248, "step": 1869 }, { "epoch": 0.1, "learning_rate": 1.977937280591183e-05, "loss": 0.9965, "step": 1870 }, { "epoch": 0.1, "learning_rate": 1.9779028948246613e-05, "loss": 0.9631, "step": 1871 }, { "epoch": 0.1, "learning_rate": 1.977868482582459e-05, "loss": 0.9994, "step": 1872 }, { "epoch": 0.1, "learning_rate": 1.9778340438655083e-05, "loss": 1.0627, "step": 1873 }, { "epoch": 0.1, "learning_rate": 1.977799578674741e-05, "loss": 1.0168, "step": 1874 }, { "epoch": 0.1, "learning_rate": 1.9777650870110904e-05, "loss": 1.08, "step": 1875 }, { "epoch": 0.1, "learning_rate": 1.9777305688754902e-05, "loss": 1.1178, "step": 1876 }, { "epoch": 0.1, "learning_rate": 1.977696024268875e-05, "loss": 0.9547, "step": 1877 }, { "epoch": 0.1, "learning_rate": 1.97766145319218e-05, "loss": 1.0155, "step": 1878 }, { "epoch": 0.1, "learning_rate": 1.977626855646342e-05, "loss": 0.9836, "step": 1879 }, { "epoch": 0.1, "learning_rate": 1.9775922316322965e-05, "loss": 1.0237, "step": 1880 }, { "epoch": 0.1, "learning_rate": 1.9775575811509818e-05, "loss": 1.0419, "step": 1881 }, { "epoch": 0.1, "learning_rate": 1.9775229042033352e-05, "loss": 1.0506, "step": 1882 }, { "epoch": 0.1, "learning_rate": 1.977488200790296e-05, "loss": 1.009, "step": 1883 }, { "epoch": 0.1, "learning_rate": 1.9774534709128036e-05, "loss": 1.0519, "step": 1884 }, { "epoch": 0.1, "learning_rate": 1.977418714571799e-05, "loss": 1.0596, "step": 1885 }, { "epoch": 0.1, "learning_rate": 1.977383931768222e-05, "loss": 1.0248, "step": 1886 }, { "epoch": 0.1, "learning_rate": 1.9773491225030154e-05, "loss": 1.0711, "step": 1887 }, { "epoch": 0.1, "learning_rate": 1.9773142867771208e-05, "loss": 1.1119, "step": 1888 }, { "epoch": 0.1, "learning_rate": 1.9772794245914815e-05, "loss": 1.0791, "step": 1889 }, { "epoch": 0.1, "learning_rate": 1.977244535947042e-05, "loss": 0.9636, "step": 1890 }, { "epoch": 0.1, "learning_rate": 1.977209620844746e-05, "loss": 1.0015, "step": 1891 }, { "epoch": 0.1, "learning_rate": 1.9771746792855393e-05, "loss": 1.0871, "step": 1892 }, { "epoch": 0.1, "learning_rate": 1.977139711270368e-05, "loss": 1.1852, "step": 1893 }, { "epoch": 0.1, "learning_rate": 1.977104716800178e-05, "loss": 1.0324, "step": 1894 }, { "epoch": 0.1, "learning_rate": 1.977069695875918e-05, "loss": 1.0484, "step": 1895 }, { "epoch": 0.1, "learning_rate": 1.977034648498535e-05, "loss": 1.0347, "step": 1896 }, { "epoch": 0.1, "learning_rate": 1.976999574668979e-05, "loss": 1.0477, "step": 1897 }, { "epoch": 0.1, "learning_rate": 1.9769644743881985e-05, "loss": 1.1078, "step": 1898 }, { "epoch": 0.1, "learning_rate": 1.9769293476571444e-05, "loss": 1.0301, "step": 1899 }, { "epoch": 0.1, "learning_rate": 1.9768941944767673e-05, "loss": 1.0144, "step": 1900 }, { "epoch": 0.1, "learning_rate": 1.9768590148480195e-05, "loss": 1.1288, "step": 1901 }, { "epoch": 0.1, "learning_rate": 1.9768238087718534e-05, "loss": 1.0018, "step": 1902 }, { "epoch": 0.1, "learning_rate": 1.9767885762492215e-05, "loss": 1.0117, "step": 1903 }, { "epoch": 0.1, "learning_rate": 1.9767533172810785e-05, "loss": 1.042, "step": 1904 }, { "epoch": 0.1, "learning_rate": 1.9767180318683784e-05, "loss": 0.9973, "step": 1905 }, { "epoch": 0.1, "learning_rate": 1.9766827200120765e-05, "loss": 0.9963, "step": 1906 }, { "epoch": 0.1, "learning_rate": 1.9766473817131296e-05, "loss": 1.0436, "step": 1907 }, { "epoch": 0.1, "learning_rate": 1.9766120169724937e-05, "loss": 0.9847, "step": 1908 }, { "epoch": 0.1, "learning_rate": 1.9765766257911263e-05, "loss": 1.0809, "step": 1909 }, { "epoch": 0.1, "learning_rate": 1.976541208169986e-05, "loss": 1.0425, "step": 1910 }, { "epoch": 0.1, "learning_rate": 1.9765057641100312e-05, "loss": 1.0669, "step": 1911 }, { "epoch": 0.1, "learning_rate": 1.976470293612222e-05, "loss": 1.0179, "step": 1912 }, { "epoch": 0.1, "learning_rate": 1.976434796677518e-05, "loss": 1.072, "step": 1913 }, { "epoch": 0.1, "learning_rate": 1.9763992733068812e-05, "loss": 1.0181, "step": 1914 }, { "epoch": 0.1, "learning_rate": 1.976363723501273e-05, "loss": 1.0135, "step": 1915 }, { "epoch": 0.1, "learning_rate": 1.9763281472616553e-05, "loss": 1.0226, "step": 1916 }, { "epoch": 0.1, "learning_rate": 1.976292544588992e-05, "loss": 1.0243, "step": 1917 }, { "epoch": 0.1, "learning_rate": 1.9762569154842466e-05, "loss": 1.0324, "step": 1918 }, { "epoch": 0.1, "learning_rate": 1.976221259948384e-05, "loss": 1.0099, "step": 1919 }, { "epoch": 0.1, "learning_rate": 1.9761855779823693e-05, "loss": 1.0378, "step": 1920 }, { "epoch": 0.1, "learning_rate": 1.976149869587169e-05, "loss": 1.0386, "step": 1921 }, { "epoch": 0.1, "learning_rate": 1.9761141347637492e-05, "loss": 1.1158, "step": 1922 }, { "epoch": 0.1, "learning_rate": 1.9760783735130777e-05, "loss": 0.978, "step": 1923 }, { "epoch": 0.1, "learning_rate": 1.9760425858361227e-05, "loss": 1.072, "step": 1924 }, { "epoch": 0.1, "learning_rate": 1.9760067717338532e-05, "loss": 1.0438, "step": 1925 }, { "epoch": 0.1, "learning_rate": 1.9759709312072387e-05, "loss": 1.0487, "step": 1926 }, { "epoch": 0.1, "learning_rate": 1.9759350642572498e-05, "loss": 1.0116, "step": 1927 }, { "epoch": 0.1, "learning_rate": 1.975899170884857e-05, "loss": 1.0096, "step": 1928 }, { "epoch": 0.1, "learning_rate": 1.9758632510910326e-05, "loss": 0.9925, "step": 1929 }, { "epoch": 0.1, "learning_rate": 1.9758273048767488e-05, "loss": 1.1229, "step": 1930 }, { "epoch": 0.1, "learning_rate": 1.975791332242979e-05, "loss": 1.0646, "step": 1931 }, { "epoch": 0.1, "learning_rate": 1.975755333190697e-05, "loss": 1.0378, "step": 1932 }, { "epoch": 0.1, "learning_rate": 1.9757193077208776e-05, "loss": 1.0176, "step": 1933 }, { "epoch": 0.1, "learning_rate": 1.975683255834496e-05, "loss": 0.9973, "step": 1934 }, { "epoch": 0.1, "learning_rate": 1.9756471775325283e-05, "loss": 1.0211, "step": 1935 }, { "epoch": 0.1, "learning_rate": 1.975611072815951e-05, "loss": 1.0663, "step": 1936 }, { "epoch": 0.1, "learning_rate": 1.9755749416857424e-05, "loss": 1.0576, "step": 1937 }, { "epoch": 0.1, "learning_rate": 1.9755387841428796e-05, "loss": 1.0167, "step": 1938 }, { "epoch": 0.1, "learning_rate": 1.9755026001883423e-05, "loss": 1.1072, "step": 1939 }, { "epoch": 0.1, "learning_rate": 1.9754663898231103e-05, "loss": 1.0187, "step": 1940 }, { "epoch": 0.1, "learning_rate": 1.9754301530481635e-05, "loss": 1.1403, "step": 1941 }, { "epoch": 0.1, "learning_rate": 1.9753938898644828e-05, "loss": 1.0375, "step": 1942 }, { "epoch": 0.1, "learning_rate": 1.9753576002730503e-05, "loss": 1.0333, "step": 1943 }, { "epoch": 0.1, "learning_rate": 1.9753212842748487e-05, "loss": 1.0092, "step": 1944 }, { "epoch": 0.1, "learning_rate": 1.975284941870861e-05, "loss": 1.0673, "step": 1945 }, { "epoch": 0.1, "learning_rate": 1.975248573062071e-05, "loss": 1.0731, "step": 1946 }, { "epoch": 0.1, "learning_rate": 1.9752121778494634e-05, "loss": 1.0441, "step": 1947 }, { "epoch": 0.1, "learning_rate": 1.9751757562340236e-05, "loss": 1.1387, "step": 1948 }, { "epoch": 0.1, "learning_rate": 1.9751393082167376e-05, "loss": 1.0077, "step": 1949 }, { "epoch": 0.1, "learning_rate": 1.9751028337985927e-05, "loss": 1.1168, "step": 1950 }, { "epoch": 0.1, "learning_rate": 1.975066332980576e-05, "loss": 0.9646, "step": 1951 }, { "epoch": 0.1, "learning_rate": 1.975029805763675e-05, "loss": 1.0945, "step": 1952 }, { "epoch": 0.1, "learning_rate": 1.97499325214888e-05, "loss": 0.9898, "step": 1953 }, { "epoch": 0.1, "learning_rate": 1.97495667213718e-05, "loss": 1.0404, "step": 1954 }, { "epoch": 0.1, "learning_rate": 1.974920065729565e-05, "loss": 1.0415, "step": 1955 }, { "epoch": 0.1, "learning_rate": 1.974883432927027e-05, "loss": 1.0421, "step": 1956 }, { "epoch": 0.1, "learning_rate": 1.974846773730557e-05, "loss": 1.1164, "step": 1957 }, { "epoch": 0.1, "learning_rate": 1.9748100881411475e-05, "loss": 1.0023, "step": 1958 }, { "epoch": 0.1, "learning_rate": 1.9747733761597922e-05, "loss": 1.1531, "step": 1959 }, { "epoch": 0.1, "learning_rate": 1.974736637787485e-05, "loss": 1.0766, "step": 1960 }, { "epoch": 0.1, "learning_rate": 1.97469987302522e-05, "loss": 1.079, "step": 1961 }, { "epoch": 0.1, "learning_rate": 1.9746630818739933e-05, "loss": 1.1255, "step": 1962 }, { "epoch": 0.1, "learning_rate": 1.9746262643348002e-05, "loss": 1.1129, "step": 1963 }, { "epoch": 0.1, "learning_rate": 1.9745894204086382e-05, "loss": 1.0775, "step": 1964 }, { "epoch": 0.1, "learning_rate": 1.9745525500965047e-05, "loss": 1.0876, "step": 1965 }, { "epoch": 0.1, "learning_rate": 1.9745156533993973e-05, "loss": 1.075, "step": 1966 }, { "epoch": 0.1, "learning_rate": 1.9744787303183157e-05, "loss": 1.0287, "step": 1967 }, { "epoch": 0.1, "learning_rate": 1.974441780854259e-05, "loss": 1.0483, "step": 1968 }, { "epoch": 0.1, "learning_rate": 1.974404805008228e-05, "loss": 1.0281, "step": 1969 }, { "epoch": 0.1, "learning_rate": 1.9743678027812237e-05, "loss": 0.9768, "step": 1970 }, { "epoch": 0.1, "learning_rate": 1.9743307741742476e-05, "loss": 1.0293, "step": 1971 }, { "epoch": 0.1, "learning_rate": 1.9742937191883024e-05, "loss": 0.9738, "step": 1972 }, { "epoch": 0.1, "learning_rate": 1.974256637824391e-05, "loss": 1.0062, "step": 1973 }, { "epoch": 0.1, "learning_rate": 1.9742195300835185e-05, "loss": 1.0313, "step": 1974 }, { "epoch": 0.1, "learning_rate": 1.974182395966688e-05, "loss": 0.981, "step": 1975 }, { "epoch": 0.1, "learning_rate": 1.9741452354749056e-05, "loss": 1.0824, "step": 1976 }, { "epoch": 0.1, "learning_rate": 1.974108048609178e-05, "loss": 1.0848, "step": 1977 }, { "epoch": 0.1, "learning_rate": 1.9740708353705107e-05, "loss": 1.1029, "step": 1978 }, { "epoch": 0.1, "learning_rate": 1.9740335957599123e-05, "loss": 0.9731, "step": 1979 }, { "epoch": 0.1, "learning_rate": 1.9739963297783902e-05, "loss": 0.9326, "step": 1980 }, { "epoch": 0.1, "learning_rate": 1.973959037426954e-05, "loss": 0.9845, "step": 1981 }, { "epoch": 0.1, "learning_rate": 1.9739217187066127e-05, "loss": 1.1335, "step": 1982 }, { "epoch": 0.1, "learning_rate": 1.9738843736183774e-05, "loss": 1.0268, "step": 1983 }, { "epoch": 0.1, "learning_rate": 1.9738470021632587e-05, "loss": 1.0614, "step": 1984 }, { "epoch": 0.1, "learning_rate": 1.9738096043422686e-05, "loss": 1.0787, "step": 1985 }, { "epoch": 0.1, "learning_rate": 1.973772180156419e-05, "loss": 1.0862, "step": 1986 }, { "epoch": 0.1, "learning_rate": 1.9737347296067245e-05, "loss": 0.9778, "step": 1987 }, { "epoch": 0.1, "learning_rate": 1.9736972526941976e-05, "loss": 1.0338, "step": 1988 }, { "epoch": 0.1, "learning_rate": 1.9736597494198536e-05, "loss": 0.9945, "step": 1989 }, { "epoch": 0.1, "learning_rate": 1.9736222197847075e-05, "loss": 0.9753, "step": 1990 }, { "epoch": 0.1, "learning_rate": 1.9735846637897762e-05, "loss": 0.9735, "step": 1991 }, { "epoch": 0.1, "learning_rate": 1.9735470814360758e-05, "loss": 1.0171, "step": 1992 }, { "epoch": 0.1, "learning_rate": 1.973509472724624e-05, "loss": 1.0561, "step": 1993 }, { "epoch": 0.1, "learning_rate": 1.9734718376564386e-05, "loss": 1.0149, "step": 1994 }, { "epoch": 0.1, "learning_rate": 1.973434176232539e-05, "loss": 1.0545, "step": 1995 }, { "epoch": 0.1, "learning_rate": 1.973396488453945e-05, "loss": 1.0532, "step": 1996 }, { "epoch": 0.1, "learning_rate": 1.9733587743216765e-05, "loss": 1.0584, "step": 1997 }, { "epoch": 0.1, "learning_rate": 1.9733210338367548e-05, "loss": 1.0364, "step": 1998 }, { "epoch": 0.1, "learning_rate": 1.9732832670002017e-05, "loss": 1.1033, "step": 1999 }, { "epoch": 0.1, "learning_rate": 1.9732454738130394e-05, "loss": 1.1481, "step": 2000 }, { "epoch": 0.1, "learning_rate": 1.9732076542762915e-05, "loss": 1.1201, "step": 2001 }, { "epoch": 0.1, "learning_rate": 1.973169808390982e-05, "loss": 1.0161, "step": 2002 }, { "epoch": 0.1, "learning_rate": 1.9731319361581347e-05, "loss": 1.0846, "step": 2003 }, { "epoch": 0.1, "learning_rate": 1.973094037578776e-05, "loss": 0.9845, "step": 2004 }, { "epoch": 0.1, "learning_rate": 1.9730561126539316e-05, "loss": 1.0175, "step": 2005 }, { "epoch": 0.1, "learning_rate": 1.973018161384628e-05, "loss": 1.0012, "step": 2006 }, { "epoch": 0.1, "learning_rate": 1.972980183771893e-05, "loss": 1.0888, "step": 2007 }, { "epoch": 0.1, "learning_rate": 1.9729421798167547e-05, "loss": 0.9319, "step": 2008 }, { "epoch": 0.1, "learning_rate": 1.972904149520242e-05, "loss": 1.033, "step": 2009 }, { "epoch": 0.1, "learning_rate": 1.9728660928833843e-05, "loss": 0.9958, "step": 2010 }, { "epoch": 0.1, "learning_rate": 1.9728280099072122e-05, "loss": 1.0267, "step": 2011 }, { "epoch": 0.1, "learning_rate": 1.9727899005927572e-05, "loss": 1.0379, "step": 2012 }, { "epoch": 0.1, "learning_rate": 1.9727517649410505e-05, "loss": 1.0322, "step": 2013 }, { "epoch": 0.1, "learning_rate": 1.9727136029531243e-05, "loss": 1.065, "step": 2014 }, { "epoch": 0.1, "learning_rate": 1.9726754146300123e-05, "loss": 0.9319, "step": 2015 }, { "epoch": 0.1, "learning_rate": 1.9726371999727484e-05, "loss": 1.0028, "step": 2016 }, { "epoch": 0.1, "learning_rate": 1.9725989589823672e-05, "loss": 0.977, "step": 2017 }, { "epoch": 0.1, "learning_rate": 1.9725606916599038e-05, "loss": 1.0266, "step": 2018 }, { "epoch": 0.1, "learning_rate": 1.9725223980063946e-05, "loss": 0.9719, "step": 2019 }, { "epoch": 0.1, "learning_rate": 1.972484078022876e-05, "loss": 0.9911, "step": 2020 }, { "epoch": 0.1, "learning_rate": 1.9724457317103856e-05, "loss": 1.0073, "step": 2021 }, { "epoch": 0.1, "learning_rate": 1.9724073590699618e-05, "loss": 1.0067, "step": 2022 }, { "epoch": 0.1, "learning_rate": 1.9723689601026433e-05, "loss": 1.0321, "step": 2023 }, { "epoch": 0.1, "learning_rate": 1.9723305348094693e-05, "loss": 0.9436, "step": 2024 }, { "epoch": 0.1, "learning_rate": 1.972292083191481e-05, "loss": 1.0447, "step": 2025 }, { "epoch": 0.1, "learning_rate": 1.972253605249719e-05, "loss": 0.9905, "step": 2026 }, { "epoch": 0.1, "learning_rate": 1.9722151009852245e-05, "loss": 1.0956, "step": 2027 }, { "epoch": 0.1, "learning_rate": 1.972176570399041e-05, "loss": 1.0704, "step": 2028 }, { "epoch": 0.1, "learning_rate": 1.972138013492211e-05, "loss": 1.0037, "step": 2029 }, { "epoch": 0.1, "learning_rate": 1.9720994302657788e-05, "loss": 1.0427, "step": 2030 }, { "epoch": 0.1, "learning_rate": 1.9720608207207882e-05, "loss": 1.0426, "step": 2031 }, { "epoch": 0.1, "learning_rate": 1.9720221848582857e-05, "loss": 1.004, "step": 2032 }, { "epoch": 0.1, "learning_rate": 1.9719835226793163e-05, "loss": 0.9845, "step": 2033 }, { "epoch": 0.1, "learning_rate": 1.971944834184927e-05, "loss": 1.0873, "step": 2034 }, { "epoch": 0.1, "learning_rate": 1.971906119376166e-05, "loss": 1.0223, "step": 2035 }, { "epoch": 0.1, "learning_rate": 1.97186737825408e-05, "loss": 1.1003, "step": 2036 }, { "epoch": 0.1, "learning_rate": 1.9718286108197196e-05, "loss": 1.0557, "step": 2037 }, { "epoch": 0.1, "learning_rate": 1.971789817074133e-05, "loss": 0.998, "step": 2038 }, { "epoch": 0.1, "learning_rate": 1.971750997018371e-05, "loss": 1.0061, "step": 2039 }, { "epoch": 0.1, "learning_rate": 1.9717121506534853e-05, "loss": 1.0368, "step": 2040 }, { "epoch": 0.1, "learning_rate": 1.9716732779805258e-05, "loss": 1.0657, "step": 2041 }, { "epoch": 0.1, "learning_rate": 1.971634379000547e-05, "loss": 0.9645, "step": 2042 }, { "epoch": 0.1, "learning_rate": 1.9715954537146004e-05, "loss": 0.9874, "step": 2043 }, { "epoch": 0.1, "learning_rate": 1.9715565021237412e-05, "loss": 1.0083, "step": 2044 }, { "epoch": 0.1, "learning_rate": 1.971517524229023e-05, "loss": 0.9718, "step": 2045 }, { "epoch": 0.1, "learning_rate": 1.9714785200315017e-05, "loss": 1.0096, "step": 2046 }, { "epoch": 0.1, "learning_rate": 1.971439489532233e-05, "loss": 1.0569, "step": 2047 }, { "epoch": 0.1, "learning_rate": 1.9714004327322732e-05, "loss": 0.9932, "step": 2048 }, { "epoch": 0.1, "learning_rate": 1.9713613496326805e-05, "loss": 1.0945, "step": 2049 }, { "epoch": 0.1, "learning_rate": 1.9713222402345132e-05, "loss": 1.0406, "step": 2050 }, { "epoch": 0.1, "learning_rate": 1.971283104538829e-05, "loss": 1.0566, "step": 2051 }, { "epoch": 0.1, "learning_rate": 1.9712439425466883e-05, "loss": 1.0368, "step": 2052 }, { "epoch": 0.1, "learning_rate": 1.971204754259151e-05, "loss": 0.9731, "step": 2053 }, { "epoch": 0.1, "learning_rate": 1.9711655396772787e-05, "loss": 1.1065, "step": 2054 }, { "epoch": 0.1, "learning_rate": 1.9711262988021322e-05, "loss": 1.0823, "step": 2055 }, { "epoch": 0.1, "learning_rate": 1.9710870316347747e-05, "loss": 0.9621, "step": 2056 }, { "epoch": 0.1, "learning_rate": 1.9710477381762687e-05, "loss": 0.9, "step": 2057 }, { "epoch": 0.1, "learning_rate": 1.9710084184276784e-05, "loss": 1.0585, "step": 2058 }, { "epoch": 0.1, "learning_rate": 1.9709690723900684e-05, "loss": 0.9995, "step": 2059 }, { "epoch": 0.1, "learning_rate": 1.9709297000645032e-05, "loss": 1.0184, "step": 2060 }, { "epoch": 0.1, "learning_rate": 1.9708903014520498e-05, "loss": 0.9609, "step": 2061 }, { "epoch": 0.1, "learning_rate": 1.9708508765537747e-05, "loss": 1.0353, "step": 2062 }, { "epoch": 0.1, "learning_rate": 1.970811425370745e-05, "loss": 1.0361, "step": 2063 }, { "epoch": 0.1, "learning_rate": 1.9707719479040283e-05, "loss": 1.0481, "step": 2064 }, { "epoch": 0.1, "learning_rate": 1.9707324441546945e-05, "loss": 1.0632, "step": 2065 }, { "epoch": 0.1, "learning_rate": 1.970692914123812e-05, "loss": 1.0388, "step": 2066 }, { "epoch": 0.11, "learning_rate": 1.970653357812452e-05, "loss": 1.0172, "step": 2067 }, { "epoch": 0.11, "learning_rate": 1.970613775221685e-05, "loss": 0.989, "step": 2068 }, { "epoch": 0.11, "learning_rate": 1.9705741663525823e-05, "loss": 1.1538, "step": 2069 }, { "epoch": 0.11, "learning_rate": 1.9705345312062173e-05, "loss": 1.0948, "step": 2070 }, { "epoch": 0.11, "learning_rate": 1.9704948697836618e-05, "loss": 1.032, "step": 2071 }, { "epoch": 0.11, "learning_rate": 1.9704551820859904e-05, "loss": 0.9051, "step": 2072 }, { "epoch": 0.11, "learning_rate": 1.970415468114278e-05, "loss": 1.0177, "step": 2073 }, { "epoch": 0.11, "learning_rate": 1.9703757278695983e-05, "loss": 1.0603, "step": 2074 }, { "epoch": 0.11, "learning_rate": 1.970335961353029e-05, "loss": 1.0258, "step": 2075 }, { "epoch": 0.11, "learning_rate": 1.970296168565645e-05, "loss": 1.1506, "step": 2076 }, { "epoch": 0.11, "learning_rate": 1.9702563495085254e-05, "loss": 1.1154, "step": 2077 }, { "epoch": 0.11, "learning_rate": 1.970216504182747e-05, "loss": 1.0685, "step": 2078 }, { "epoch": 0.11, "learning_rate": 1.970176632589389e-05, "loss": 1.0893, "step": 2079 }, { "epoch": 0.11, "learning_rate": 1.970136734729531e-05, "loss": 0.9745, "step": 2080 }, { "epoch": 0.11, "learning_rate": 1.970096810604253e-05, "loss": 1.0464, "step": 2081 }, { "epoch": 0.11, "learning_rate": 1.970056860214636e-05, "loss": 0.9641, "step": 2082 }, { "epoch": 0.11, "learning_rate": 1.9700168835617614e-05, "loss": 1.0818, "step": 2083 }, { "epoch": 0.11, "learning_rate": 1.9699768806467116e-05, "loss": 1.0106, "step": 2084 }, { "epoch": 0.11, "learning_rate": 1.9699368514705702e-05, "loss": 1.0916, "step": 2085 }, { "epoch": 0.11, "learning_rate": 1.9698967960344202e-05, "loss": 1.0127, "step": 2086 }, { "epoch": 0.11, "learning_rate": 1.969856714339346e-05, "loss": 1.1547, "step": 2087 }, { "epoch": 0.11, "learning_rate": 1.9698166063864337e-05, "loss": 1.077, "step": 2088 }, { "epoch": 0.11, "learning_rate": 1.9697764721767685e-05, "loss": 1.0167, "step": 2089 }, { "epoch": 0.11, "learning_rate": 1.9697363117114368e-05, "loss": 1.0812, "step": 2090 }, { "epoch": 0.11, "learning_rate": 1.9696961249915263e-05, "loss": 0.9903, "step": 2091 }, { "epoch": 0.11, "learning_rate": 1.9696559120181248e-05, "loss": 1.042, "step": 2092 }, { "epoch": 0.11, "learning_rate": 1.9696156727923214e-05, "loss": 1.0786, "step": 2093 }, { "epoch": 0.11, "learning_rate": 1.969575407315205e-05, "loss": 0.991, "step": 2094 }, { "epoch": 0.11, "learning_rate": 1.969535115587866e-05, "loss": 0.9886, "step": 2095 }, { "epoch": 0.11, "learning_rate": 1.969494797611395e-05, "loss": 0.9495, "step": 2096 }, { "epoch": 0.11, "learning_rate": 1.9694544533868844e-05, "loss": 1.0717, "step": 2097 }, { "epoch": 0.11, "learning_rate": 1.9694140829154255e-05, "loss": 1.0049, "step": 2098 }, { "epoch": 0.11, "learning_rate": 1.969373686198112e-05, "loss": 1.0718, "step": 2099 }, { "epoch": 0.11, "learning_rate": 1.969333263236037e-05, "loss": 1.0099, "step": 2100 }, { "epoch": 0.11, "learning_rate": 1.9692928140302953e-05, "loss": 1.0806, "step": 2101 }, { "epoch": 0.11, "learning_rate": 1.9692523385819816e-05, "loss": 1.1443, "step": 2102 }, { "epoch": 0.11, "learning_rate": 1.969211836892192e-05, "loss": 1.1157, "step": 2103 }, { "epoch": 0.11, "learning_rate": 1.969171308962024e-05, "loss": 1.044, "step": 2104 }, { "epoch": 0.11, "learning_rate": 1.969130754792573e-05, "loss": 1.0569, "step": 2105 }, { "epoch": 0.11, "learning_rate": 1.9690901743849383e-05, "loss": 1.0591, "step": 2106 }, { "epoch": 0.11, "learning_rate": 1.9690495677402176e-05, "loss": 1.0244, "step": 2107 }, { "epoch": 0.11, "learning_rate": 1.9690089348595115e-05, "loss": 1.0236, "step": 2108 }, { "epoch": 0.11, "learning_rate": 1.9689682757439192e-05, "loss": 1.0966, "step": 2109 }, { "epoch": 0.11, "learning_rate": 1.9689275903945416e-05, "loss": 1.0091, "step": 2110 }, { "epoch": 0.11, "learning_rate": 1.9688868788124807e-05, "loss": 1.0276, "step": 2111 }, { "epoch": 0.11, "learning_rate": 1.968846140998838e-05, "loss": 1.0577, "step": 2112 }, { "epoch": 0.11, "learning_rate": 1.968805376954717e-05, "loss": 1.0397, "step": 2113 }, { "epoch": 0.11, "learning_rate": 1.9687645866812214e-05, "loss": 1.0142, "step": 2114 }, { "epoch": 0.11, "learning_rate": 1.9687237701794547e-05, "loss": 1.0974, "step": 2115 }, { "epoch": 0.11, "learning_rate": 1.968682927450523e-05, "loss": 0.9747, "step": 2116 }, { "epoch": 0.11, "learning_rate": 1.9686420584955315e-05, "loss": 1.0997, "step": 2117 }, { "epoch": 0.11, "learning_rate": 1.968601163315587e-05, "loss": 1.1115, "step": 2118 }, { "epoch": 0.11, "learning_rate": 1.9685602419117962e-05, "loss": 1.1005, "step": 2119 }, { "epoch": 0.11, "learning_rate": 1.9685192942852676e-05, "loss": 1.109, "step": 2120 }, { "epoch": 0.11, "learning_rate": 1.968478320437109e-05, "loss": 1.0953, "step": 2121 }, { "epoch": 0.11, "learning_rate": 1.968437320368431e-05, "loss": 1.0175, "step": 2122 }, { "epoch": 0.11, "learning_rate": 1.9683962940803424e-05, "loss": 1.0553, "step": 2123 }, { "epoch": 0.11, "learning_rate": 1.9683552415739547e-05, "loss": 1.0, "step": 2124 }, { "epoch": 0.11, "learning_rate": 1.968314162850379e-05, "loss": 1.1054, "step": 2125 }, { "epoch": 0.11, "learning_rate": 1.9682730579107274e-05, "loss": 1.0121, "step": 2126 }, { "epoch": 0.11, "learning_rate": 1.9682319267561128e-05, "loss": 1.0699, "step": 2127 }, { "epoch": 0.11, "learning_rate": 1.9681907693876493e-05, "loss": 1.076, "step": 2128 }, { "epoch": 0.11, "learning_rate": 1.9681495858064505e-05, "loss": 1.0427, "step": 2129 }, { "epoch": 0.11, "learning_rate": 1.968108376013632e-05, "loss": 1.0348, "step": 2130 }, { "epoch": 0.11, "learning_rate": 1.9680671400103087e-05, "loss": 1.0121, "step": 2131 }, { "epoch": 0.11, "learning_rate": 1.968025877797598e-05, "loss": 1.0167, "step": 2132 }, { "epoch": 0.11, "learning_rate": 1.9679845893766165e-05, "loss": 1.0636, "step": 2133 }, { "epoch": 0.11, "learning_rate": 1.9679432747484818e-05, "loss": 1.0698, "step": 2134 }, { "epoch": 0.11, "learning_rate": 1.967901933914313e-05, "loss": 0.9984, "step": 2135 }, { "epoch": 0.11, "learning_rate": 1.967860566875229e-05, "loss": 1.061, "step": 2136 }, { "epoch": 0.11, "learning_rate": 1.96781917363235e-05, "loss": 1.0218, "step": 2137 }, { "epoch": 0.11, "learning_rate": 1.9677777541867966e-05, "loss": 1.0071, "step": 2138 }, { "epoch": 0.11, "learning_rate": 1.9677363085396897e-05, "loss": 1.0341, "step": 2139 }, { "epoch": 0.11, "learning_rate": 1.9676948366921526e-05, "loss": 1.0221, "step": 2140 }, { "epoch": 0.11, "learning_rate": 1.9676533386453064e-05, "loss": 1.0212, "step": 2141 }, { "epoch": 0.11, "learning_rate": 1.9676118144002762e-05, "loss": 1.1111, "step": 2142 }, { "epoch": 0.11, "learning_rate": 1.9675702639581853e-05, "loss": 1.0072, "step": 2143 }, { "epoch": 0.11, "learning_rate": 1.9675286873201593e-05, "loss": 0.9249, "step": 2144 }, { "epoch": 0.11, "learning_rate": 1.967487084487323e-05, "loss": 1.0758, "step": 2145 }, { "epoch": 0.11, "learning_rate": 1.9674454554608036e-05, "loss": 0.9219, "step": 2146 }, { "epoch": 0.11, "learning_rate": 1.967403800241728e-05, "loss": 1.0657, "step": 2147 }, { "epoch": 0.11, "learning_rate": 1.9673621188312228e-05, "loss": 1.124, "step": 2148 }, { "epoch": 0.11, "learning_rate": 1.967320411230418e-05, "loss": 1.0996, "step": 2149 }, { "epoch": 0.11, "learning_rate": 1.9672786774404423e-05, "loss": 1.0988, "step": 2150 }, { "epoch": 0.11, "learning_rate": 1.967236917462425e-05, "loss": 0.9695, "step": 2151 }, { "epoch": 0.11, "learning_rate": 1.967195131297498e-05, "loss": 1.0436, "step": 2152 }, { "epoch": 0.11, "learning_rate": 1.9671533189467914e-05, "loss": 1.0589, "step": 2153 }, { "epoch": 0.11, "learning_rate": 1.9671114804114378e-05, "loss": 1.0198, "step": 2154 }, { "epoch": 0.11, "learning_rate": 1.9670696156925695e-05, "loss": 1.0437, "step": 2155 }, { "epoch": 0.11, "learning_rate": 1.9670277247913205e-05, "loss": 0.9982, "step": 2156 }, { "epoch": 0.11, "learning_rate": 1.9669858077088246e-05, "loss": 1.0468, "step": 2157 }, { "epoch": 0.11, "learning_rate": 1.9669438644462168e-05, "loss": 1.0441, "step": 2158 }, { "epoch": 0.11, "learning_rate": 1.9669018950046326e-05, "loss": 1.1055, "step": 2159 }, { "epoch": 0.11, "learning_rate": 1.9668598993852083e-05, "loss": 1.1068, "step": 2160 }, { "epoch": 0.11, "learning_rate": 1.9668178775890808e-05, "loss": 1.1112, "step": 2161 }, { "epoch": 0.11, "learning_rate": 1.966775829617388e-05, "loss": 0.9695, "step": 2162 }, { "epoch": 0.11, "learning_rate": 1.966733755471268e-05, "loss": 1.019, "step": 2163 }, { "epoch": 0.11, "learning_rate": 1.9666916551518605e-05, "loss": 1.1511, "step": 2164 }, { "epoch": 0.11, "learning_rate": 1.9666495286603046e-05, "loss": 1.1109, "step": 2165 }, { "epoch": 0.11, "learning_rate": 1.966607375997741e-05, "loss": 1.0818, "step": 2166 }, { "epoch": 0.11, "learning_rate": 1.9665651971653113e-05, "loss": 1.0651, "step": 2167 }, { "epoch": 0.11, "learning_rate": 1.966522992164157e-05, "loss": 1.0838, "step": 2168 }, { "epoch": 0.11, "learning_rate": 1.9664807609954213e-05, "loss": 1.1066, "step": 2169 }, { "epoch": 0.11, "learning_rate": 1.9664385036602472e-05, "loss": 1.0656, "step": 2170 }, { "epoch": 0.11, "learning_rate": 1.9663962201597785e-05, "loss": 1.158, "step": 2171 }, { "epoch": 0.11, "learning_rate": 1.9663539104951606e-05, "loss": 1.0912, "step": 2172 }, { "epoch": 0.11, "learning_rate": 1.966311574667539e-05, "loss": 1.0191, "step": 2173 }, { "epoch": 0.11, "learning_rate": 1.966269212678059e-05, "loss": 0.9847, "step": 2174 }, { "epoch": 0.11, "learning_rate": 1.9662268245278684e-05, "loss": 0.9652, "step": 2175 }, { "epoch": 0.11, "learning_rate": 1.9661844102181144e-05, "loss": 0.9993, "step": 2176 }, { "epoch": 0.11, "learning_rate": 1.9661419697499455e-05, "loss": 0.9966, "step": 2177 }, { "epoch": 0.11, "learning_rate": 1.9660995031245107e-05, "loss": 0.9346, "step": 2178 }, { "epoch": 0.11, "learning_rate": 1.9660570103429598e-05, "loss": 0.991, "step": 2179 }, { "epoch": 0.11, "learning_rate": 1.9660144914064426e-05, "loss": 0.9771, "step": 2180 }, { "epoch": 0.11, "learning_rate": 1.9659719463161115e-05, "loss": 0.9987, "step": 2181 }, { "epoch": 0.11, "learning_rate": 1.9659293750731174e-05, "loss": 0.9622, "step": 2182 }, { "epoch": 0.11, "learning_rate": 1.965886777678613e-05, "loss": 1.1303, "step": 2183 }, { "epoch": 0.11, "learning_rate": 1.965844154133752e-05, "loss": 1.0969, "step": 2184 }, { "epoch": 0.11, "learning_rate": 1.9658015044396878e-05, "loss": 0.9962, "step": 2185 }, { "epoch": 0.11, "learning_rate": 1.9657588285975755e-05, "loss": 1.0688, "step": 2186 }, { "epoch": 0.11, "learning_rate": 1.9657161266085706e-05, "loss": 1.0033, "step": 2187 }, { "epoch": 0.11, "learning_rate": 1.9656733984738286e-05, "loss": 0.9916, "step": 2188 }, { "epoch": 0.11, "learning_rate": 1.965630644194507e-05, "loss": 0.976, "step": 2189 }, { "epoch": 0.11, "learning_rate": 1.965587863771763e-05, "loss": 1.0747, "step": 2190 }, { "epoch": 0.11, "learning_rate": 1.9655450572067548e-05, "loss": 1.0491, "step": 2191 }, { "epoch": 0.11, "learning_rate": 1.9655022245006415e-05, "loss": 0.9955, "step": 2192 }, { "epoch": 0.11, "learning_rate": 1.965459365654583e-05, "loss": 1.0685, "step": 2193 }, { "epoch": 0.11, "learning_rate": 1.965416480669739e-05, "loss": 1.0081, "step": 2194 }, { "epoch": 0.11, "learning_rate": 1.965373569547271e-05, "loss": 0.93, "step": 2195 }, { "epoch": 0.11, "learning_rate": 1.9653306322883406e-05, "loss": 0.9272, "step": 2196 }, { "epoch": 0.11, "learning_rate": 1.9652876688941106e-05, "loss": 1.0991, "step": 2197 }, { "epoch": 0.11, "learning_rate": 1.9652446793657435e-05, "loss": 1.103, "step": 2198 }, { "epoch": 0.11, "learning_rate": 1.965201663704404e-05, "loss": 0.9875, "step": 2199 }, { "epoch": 0.11, "learning_rate": 1.9651586219112565e-05, "loss": 1.0479, "step": 2200 }, { "epoch": 0.11, "learning_rate": 1.965115553987466e-05, "loss": 0.99, "step": 2201 }, { "epoch": 0.11, "learning_rate": 1.9650724599341988e-05, "loss": 1.0, "step": 2202 }, { "epoch": 0.11, "learning_rate": 1.965029339752621e-05, "loss": 1.08, "step": 2203 }, { "epoch": 0.11, "learning_rate": 1.964986193443901e-05, "loss": 1.0171, "step": 2204 }, { "epoch": 0.11, "learning_rate": 1.9649430210092063e-05, "loss": 1.0057, "step": 2205 }, { "epoch": 0.11, "learning_rate": 1.964899822449706e-05, "loss": 1.0251, "step": 2206 }, { "epoch": 0.11, "learning_rate": 1.9648565977665693e-05, "loss": 1.0309, "step": 2207 }, { "epoch": 0.11, "learning_rate": 1.964813346960967e-05, "loss": 0.9948, "step": 2208 }, { "epoch": 0.11, "learning_rate": 1.96477007003407e-05, "loss": 0.9918, "step": 2209 }, { "epoch": 0.11, "learning_rate": 1.9647267669870494e-05, "loss": 0.9868, "step": 2210 }, { "epoch": 0.11, "learning_rate": 1.964683437821078e-05, "loss": 0.9952, "step": 2211 }, { "epoch": 0.11, "learning_rate": 1.9646400825373294e-05, "loss": 0.9777, "step": 2212 }, { "epoch": 0.11, "learning_rate": 1.9645967011369765e-05, "loss": 1.0166, "step": 2213 }, { "epoch": 0.11, "learning_rate": 1.9645532936211938e-05, "loss": 1.0135, "step": 2214 }, { "epoch": 0.11, "learning_rate": 1.9645098599911573e-05, "loss": 0.9822, "step": 2215 }, { "epoch": 0.11, "learning_rate": 1.9644664002480428e-05, "loss": 0.9597, "step": 2216 }, { "epoch": 0.11, "learning_rate": 1.964422914393026e-05, "loss": 1.002, "step": 2217 }, { "epoch": 0.11, "learning_rate": 1.964379402427285e-05, "loss": 1.0359, "step": 2218 }, { "epoch": 0.11, "learning_rate": 1.9643358643519977e-05, "loss": 1.0989, "step": 2219 }, { "epoch": 0.11, "learning_rate": 1.964292300168343e-05, "loss": 1.1055, "step": 2220 }, { "epoch": 0.11, "learning_rate": 1.9642487098775004e-05, "loss": 1.0895, "step": 2221 }, { "epoch": 0.11, "learning_rate": 1.9642050934806495e-05, "loss": 1.0334, "step": 2222 }, { "epoch": 0.11, "learning_rate": 1.9641614509789714e-05, "loss": 0.9433, "step": 2223 }, { "epoch": 0.11, "learning_rate": 1.9641177823736483e-05, "loss": 0.9563, "step": 2224 }, { "epoch": 0.11, "learning_rate": 1.964074087665862e-05, "loss": 1.0008, "step": 2225 }, { "epoch": 0.11, "learning_rate": 1.964030366856795e-05, "loss": 1.0921, "step": 2226 }, { "epoch": 0.11, "learning_rate": 1.9639866199476315e-05, "loss": 1.0103, "step": 2227 }, { "epoch": 0.11, "learning_rate": 1.963942846939556e-05, "loss": 1.0787, "step": 2228 }, { "epoch": 0.11, "learning_rate": 1.9638990478337536e-05, "loss": 1.0121, "step": 2229 }, { "epoch": 0.11, "learning_rate": 1.96385522263141e-05, "loss": 1.0389, "step": 2230 }, { "epoch": 0.11, "learning_rate": 1.9638113713337114e-05, "loss": 0.9941, "step": 2231 }, { "epoch": 0.11, "learning_rate": 1.9637674939418457e-05, "loss": 1.052, "step": 2232 }, { "epoch": 0.11, "learning_rate": 1.9637235904570003e-05, "loss": 1.0555, "step": 2233 }, { "epoch": 0.11, "learning_rate": 1.963679660880364e-05, "loss": 1.0312, "step": 2234 }, { "epoch": 0.11, "learning_rate": 1.9636357052131262e-05, "loss": 1.0075, "step": 2235 }, { "epoch": 0.11, "learning_rate": 1.9635917234564768e-05, "loss": 0.9149, "step": 2236 }, { "epoch": 0.11, "learning_rate": 1.963547715611607e-05, "loss": 1.0518, "step": 2237 }, { "epoch": 0.11, "learning_rate": 1.9635036816797072e-05, "loss": 1.0737, "step": 2238 }, { "epoch": 0.11, "learning_rate": 1.9634596216619707e-05, "loss": 1.0681, "step": 2239 }, { "epoch": 0.11, "learning_rate": 1.9634155355595903e-05, "loss": 1.0066, "step": 2240 }, { "epoch": 0.11, "learning_rate": 1.963371423373759e-05, "loss": 1.0204, "step": 2241 }, { "epoch": 0.11, "learning_rate": 1.9633272851056714e-05, "loss": 1.1426, "step": 2242 }, { "epoch": 0.11, "learning_rate": 1.9632831207565225e-05, "loss": 1.0381, "step": 2243 }, { "epoch": 0.11, "learning_rate": 1.963238930327508e-05, "loss": 1.0255, "step": 2244 }, { "epoch": 0.11, "learning_rate": 1.963194713819824e-05, "loss": 1.0499, "step": 2245 }, { "epoch": 0.11, "learning_rate": 1.963150471234668e-05, "loss": 1.1073, "step": 2246 }, { "epoch": 0.11, "learning_rate": 1.963106202573238e-05, "loss": 1.0544, "step": 2247 }, { "epoch": 0.11, "learning_rate": 1.963061907836732e-05, "loss": 0.9846, "step": 2248 }, { "epoch": 0.11, "learning_rate": 1.9630175870263492e-05, "loss": 1.0494, "step": 2249 }, { "epoch": 0.11, "learning_rate": 1.9629732401432902e-05, "loss": 1.1347, "step": 2250 }, { "epoch": 0.11, "learning_rate": 1.962928867188755e-05, "loss": 0.9826, "step": 2251 }, { "epoch": 0.11, "learning_rate": 1.9628844681639455e-05, "loss": 1.0381, "step": 2252 }, { "epoch": 0.11, "learning_rate": 1.962840043070063e-05, "loss": 1.0759, "step": 2253 }, { "epoch": 0.11, "learning_rate": 1.962795591908311e-05, "loss": 0.9539, "step": 2254 }, { "epoch": 0.11, "learning_rate": 1.962751114679893e-05, "loss": 0.9594, "step": 2255 }, { "epoch": 0.11, "learning_rate": 1.962706611386012e-05, "loss": 1.0485, "step": 2256 }, { "epoch": 0.11, "learning_rate": 1.9626620820278745e-05, "loss": 1.0085, "step": 2257 }, { "epoch": 0.11, "learning_rate": 1.962617526606685e-05, "loss": 1.1059, "step": 2258 }, { "epoch": 0.11, "learning_rate": 1.96257294512365e-05, "loss": 1.0655, "step": 2259 }, { "epoch": 0.11, "learning_rate": 1.9625283375799767e-05, "loss": 1.0092, "step": 2260 }, { "epoch": 0.11, "learning_rate": 1.9624837039768732e-05, "loss": 1.0012, "step": 2261 }, { "epoch": 0.11, "learning_rate": 1.962439044315547e-05, "loss": 1.1525, "step": 2262 }, { "epoch": 0.11, "learning_rate": 1.9623943585972075e-05, "loss": 1.0131, "step": 2263 }, { "epoch": 0.12, "learning_rate": 1.9623496468230647e-05, "loss": 0.9936, "step": 2264 }, { "epoch": 0.12, "learning_rate": 1.962304908994329e-05, "loss": 1.0776, "step": 2265 }, { "epoch": 0.12, "learning_rate": 1.9622601451122118e-05, "loss": 0.9745, "step": 2266 }, { "epoch": 0.12, "learning_rate": 1.9622153551779252e-05, "loss": 1.0546, "step": 2267 }, { "epoch": 0.12, "learning_rate": 1.9621705391926813e-05, "loss": 0.9545, "step": 2268 }, { "epoch": 0.12, "learning_rate": 1.962125697157694e-05, "loss": 1.0817, "step": 2269 }, { "epoch": 0.12, "learning_rate": 1.9620808290741768e-05, "loss": 1.0754, "step": 2270 }, { "epoch": 0.12, "learning_rate": 1.9620359349433448e-05, "loss": 1.0802, "step": 2271 }, { "epoch": 0.12, "learning_rate": 1.9619910147664134e-05, "loss": 1.0227, "step": 2272 }, { "epoch": 0.12, "learning_rate": 1.9619460685445984e-05, "loss": 1.0181, "step": 2273 }, { "epoch": 0.12, "learning_rate": 1.9619010962791174e-05, "loss": 0.947, "step": 2274 }, { "epoch": 0.12, "learning_rate": 1.961856097971188e-05, "loss": 1.0565, "step": 2275 }, { "epoch": 0.12, "learning_rate": 1.9618110736220274e-05, "loss": 1.0122, "step": 2276 }, { "epoch": 0.12, "learning_rate": 1.961766023232855e-05, "loss": 1.0217, "step": 2277 }, { "epoch": 0.12, "learning_rate": 1.9617209468048916e-05, "loss": 0.9559, "step": 2278 }, { "epoch": 0.12, "learning_rate": 1.9616758443393565e-05, "loss": 1.0322, "step": 2279 }, { "epoch": 0.12, "learning_rate": 1.9616307158374707e-05, "loss": 0.9819, "step": 2280 }, { "epoch": 0.12, "learning_rate": 1.9615855613004568e-05, "loss": 1.0818, "step": 2281 }, { "epoch": 0.12, "learning_rate": 1.961540380729537e-05, "loss": 1.0365, "step": 2282 }, { "epoch": 0.12, "learning_rate": 1.961495174125934e-05, "loss": 1.104, "step": 2283 }, { "epoch": 0.12, "learning_rate": 1.9614499414908722e-05, "loss": 1.0128, "step": 2284 }, { "epoch": 0.12, "learning_rate": 1.9614046828255763e-05, "loss": 1.0683, "step": 2285 }, { "epoch": 0.12, "learning_rate": 1.961359398131271e-05, "loss": 1.0483, "step": 2286 }, { "epoch": 0.12, "learning_rate": 1.961314087409183e-05, "loss": 0.9954, "step": 2287 }, { "epoch": 0.12, "learning_rate": 1.961268750660539e-05, "loss": 0.9697, "step": 2288 }, { "epoch": 0.12, "learning_rate": 1.9612233878865664e-05, "loss": 0.9615, "step": 2289 }, { "epoch": 0.12, "learning_rate": 1.961177999088493e-05, "loss": 1.0084, "step": 2290 }, { "epoch": 0.12, "learning_rate": 1.961132584267548e-05, "loss": 0.9388, "step": 2291 }, { "epoch": 0.12, "learning_rate": 1.9610871434249607e-05, "loss": 1.1702, "step": 2292 }, { "epoch": 0.12, "learning_rate": 1.9610416765619615e-05, "loss": 1.0428, "step": 2293 }, { "epoch": 0.12, "learning_rate": 1.9609961836797815e-05, "loss": 1.0318, "step": 2294 }, { "epoch": 0.12, "learning_rate": 1.9609506647796523e-05, "loss": 1.1179, "step": 2295 }, { "epoch": 0.12, "learning_rate": 1.9609051198628065e-05, "loss": 1.08, "step": 2296 }, { "epoch": 0.12, "learning_rate": 1.9608595489304764e-05, "loss": 0.9474, "step": 2297 }, { "epoch": 0.12, "learning_rate": 1.9608139519838963e-05, "loss": 1.1381, "step": 2298 }, { "epoch": 0.12, "learning_rate": 1.960768329024301e-05, "loss": 1.0026, "step": 2299 }, { "epoch": 0.12, "learning_rate": 1.960722680052925e-05, "loss": 1.0483, "step": 2300 }, { "epoch": 0.12, "learning_rate": 1.9606770050710048e-05, "loss": 0.9746, "step": 2301 }, { "epoch": 0.12, "learning_rate": 1.960631304079777e-05, "loss": 1.1232, "step": 2302 }, { "epoch": 0.12, "learning_rate": 1.960585577080478e-05, "loss": 1.0763, "step": 2303 }, { "epoch": 0.12, "learning_rate": 1.960539824074347e-05, "loss": 1.0118, "step": 2304 }, { "epoch": 0.12, "learning_rate": 1.960494045062622e-05, "loss": 1.0182, "step": 2305 }, { "epoch": 0.12, "learning_rate": 1.9604482400465427e-05, "loss": 1.0109, "step": 2306 }, { "epoch": 0.12, "learning_rate": 1.960402409027349e-05, "loss": 1.0256, "step": 2307 }, { "epoch": 0.12, "learning_rate": 1.960356552006282e-05, "loss": 1.0213, "step": 2308 }, { "epoch": 0.12, "learning_rate": 1.9603106689845833e-05, "loss": 0.98, "step": 2309 }, { "epoch": 0.12, "learning_rate": 1.960264759963495e-05, "loss": 1.0182, "step": 2310 }, { "epoch": 0.12, "learning_rate": 1.9602188249442594e-05, "loss": 0.9588, "step": 2311 }, { "epoch": 0.12, "learning_rate": 1.960172863928121e-05, "loss": 1.0303, "step": 2312 }, { "epoch": 0.12, "learning_rate": 1.9601268769163237e-05, "loss": 1.0743, "step": 2313 }, { "epoch": 0.12, "learning_rate": 1.960080863910113e-05, "loss": 1.0181, "step": 2314 }, { "epoch": 0.12, "learning_rate": 1.9600348249107342e-05, "loss": 0.9937, "step": 2315 }, { "epoch": 0.12, "learning_rate": 1.959988759919434e-05, "loss": 1.085, "step": 2316 }, { "epoch": 0.12, "learning_rate": 1.959942668937459e-05, "loss": 0.9206, "step": 2317 }, { "epoch": 0.12, "learning_rate": 1.959896551966058e-05, "loss": 0.8638, "step": 2318 }, { "epoch": 0.12, "learning_rate": 1.9598504090064793e-05, "loss": 1.1193, "step": 2319 }, { "epoch": 0.12, "learning_rate": 1.959804240059972e-05, "loss": 0.987, "step": 2320 }, { "epoch": 0.12, "learning_rate": 1.9597580451277852e-05, "loss": 1.1052, "step": 2321 }, { "epoch": 0.12, "learning_rate": 1.959711824211171e-05, "loss": 0.9811, "step": 2322 }, { "epoch": 0.12, "learning_rate": 1.9596655773113803e-05, "loss": 1.0554, "step": 2323 }, { "epoch": 0.12, "learning_rate": 1.959619304429665e-05, "loss": 1.0115, "step": 2324 }, { "epoch": 0.12, "learning_rate": 1.959573005567278e-05, "loss": 1.0118, "step": 2325 }, { "epoch": 0.12, "learning_rate": 1.9595266807254727e-05, "loss": 0.9609, "step": 2326 }, { "epoch": 0.12, "learning_rate": 1.9594803299055037e-05, "loss": 1.0193, "step": 2327 }, { "epoch": 0.12, "learning_rate": 1.959433953108625e-05, "loss": 1.0466, "step": 2328 }, { "epoch": 0.12, "learning_rate": 1.9593875503360934e-05, "loss": 1.0077, "step": 2329 }, { "epoch": 0.12, "learning_rate": 1.959341121589164e-05, "loss": 1.05, "step": 2330 }, { "epoch": 0.12, "learning_rate": 1.9592946668690945e-05, "loss": 1.1077, "step": 2331 }, { "epoch": 0.12, "learning_rate": 1.959248186177143e-05, "loss": 1.0836, "step": 2332 }, { "epoch": 0.12, "learning_rate": 1.959201679514567e-05, "loss": 1.0004, "step": 2333 }, { "epoch": 0.12, "learning_rate": 1.959155146882626e-05, "loss": 0.966, "step": 2334 }, { "epoch": 0.12, "learning_rate": 1.95910858828258e-05, "loss": 1.0944, "step": 2335 }, { "epoch": 0.12, "learning_rate": 1.9590620037156896e-05, "loss": 0.9666, "step": 2336 }, { "epoch": 0.12, "learning_rate": 1.9590153931832156e-05, "loss": 0.9592, "step": 2337 }, { "epoch": 0.12, "learning_rate": 1.95896875668642e-05, "loss": 0.9456, "step": 2338 }, { "epoch": 0.12, "learning_rate": 1.9589220942265658e-05, "loss": 1.0593, "step": 2339 }, { "epoch": 0.12, "learning_rate": 1.958875405804916e-05, "loss": 0.9533, "step": 2340 }, { "epoch": 0.12, "learning_rate": 1.9588286914227344e-05, "loss": 0.9974, "step": 2341 }, { "epoch": 0.12, "learning_rate": 1.9587819510812866e-05, "loss": 1.0015, "step": 2342 }, { "epoch": 0.12, "learning_rate": 1.9587351847818375e-05, "loss": 1.0765, "step": 2343 }, { "epoch": 0.12, "learning_rate": 1.958688392525653e-05, "loss": 1.0428, "step": 2344 }, { "epoch": 0.12, "learning_rate": 1.9586415743140006e-05, "loss": 1.0543, "step": 2345 }, { "epoch": 0.12, "learning_rate": 1.9585947301481472e-05, "loss": 1.0698, "step": 2346 }, { "epoch": 0.12, "learning_rate": 1.9585478600293616e-05, "loss": 0.9051, "step": 2347 }, { "epoch": 0.12, "learning_rate": 1.958500963958912e-05, "loss": 1.033, "step": 2348 }, { "epoch": 0.12, "learning_rate": 1.9584540419380688e-05, "loss": 1.0625, "step": 2349 }, { "epoch": 0.12, "learning_rate": 1.9584070939681027e-05, "loss": 0.994, "step": 2350 }, { "epoch": 0.12, "learning_rate": 1.9583601200502835e-05, "loss": 1.123, "step": 2351 }, { "epoch": 0.12, "learning_rate": 1.958313120185884e-05, "loss": 1.0175, "step": 2352 }, { "epoch": 0.12, "learning_rate": 1.958266094376176e-05, "loss": 1.0773, "step": 2353 }, { "epoch": 0.12, "learning_rate": 1.9582190426224336e-05, "loss": 1.0676, "step": 2354 }, { "epoch": 0.12, "learning_rate": 1.9581719649259297e-05, "loss": 0.9061, "step": 2355 }, { "epoch": 0.12, "learning_rate": 1.9581248612879393e-05, "loss": 1.0745, "step": 2356 }, { "epoch": 0.12, "learning_rate": 1.9580777317097376e-05, "loss": 0.997, "step": 2357 }, { "epoch": 0.12, "learning_rate": 1.9580305761926004e-05, "loss": 1.0183, "step": 2358 }, { "epoch": 0.12, "learning_rate": 1.957983394737805e-05, "loss": 1.0612, "step": 2359 }, { "epoch": 0.12, "learning_rate": 1.957936187346628e-05, "loss": 1.0397, "step": 2360 }, { "epoch": 0.12, "learning_rate": 1.9578889540203485e-05, "loss": 1.0814, "step": 2361 }, { "epoch": 0.12, "learning_rate": 1.9578416947602442e-05, "loss": 1.0172, "step": 2362 }, { "epoch": 0.12, "learning_rate": 1.9577944095675954e-05, "loss": 1.0157, "step": 2363 }, { "epoch": 0.12, "learning_rate": 1.9577470984436817e-05, "loss": 1.1025, "step": 2364 }, { "epoch": 0.12, "learning_rate": 1.9576997613897846e-05, "loss": 1.0253, "step": 2365 }, { "epoch": 0.12, "learning_rate": 1.957652398407185e-05, "loss": 1.0559, "step": 2366 }, { "epoch": 0.12, "learning_rate": 1.957605009497166e-05, "loss": 0.957, "step": 2367 }, { "epoch": 0.12, "learning_rate": 1.9575575946610098e-05, "loss": 0.9353, "step": 2368 }, { "epoch": 0.12, "learning_rate": 1.9575101539000005e-05, "loss": 1.0465, "step": 2369 }, { "epoch": 0.12, "learning_rate": 1.9574626872154225e-05, "loss": 1.1538, "step": 2370 }, { "epoch": 0.12, "learning_rate": 1.9574151946085613e-05, "loss": 1.1149, "step": 2371 }, { "epoch": 0.12, "learning_rate": 1.957367676080702e-05, "loss": 1.0499, "step": 2372 }, { "epoch": 0.12, "learning_rate": 1.9573201316331315e-05, "loss": 1.0409, "step": 2373 }, { "epoch": 0.12, "learning_rate": 1.9572725612671367e-05, "loss": 1.0851, "step": 2374 }, { "epoch": 0.12, "learning_rate": 1.9572249649840058e-05, "loss": 0.8864, "step": 2375 }, { "epoch": 0.12, "learning_rate": 1.9571773427850274e-05, "loss": 0.9655, "step": 2376 }, { "epoch": 0.12, "learning_rate": 1.9571296946714914e-05, "loss": 1.063, "step": 2377 }, { "epoch": 0.12, "learning_rate": 1.9570820206446863e-05, "loss": 0.9832, "step": 2378 }, { "epoch": 0.12, "learning_rate": 1.9570343207059043e-05, "loss": 0.9975, "step": 2379 }, { "epoch": 0.12, "learning_rate": 1.9569865948564366e-05, "loss": 1.0236, "step": 2380 }, { "epoch": 0.12, "learning_rate": 1.956938843097574e-05, "loss": 0.9977, "step": 2381 }, { "epoch": 0.12, "learning_rate": 1.956891065430611e-05, "loss": 1.0006, "step": 2382 }, { "epoch": 0.12, "learning_rate": 1.9568432618568403e-05, "loss": 1.0265, "step": 2383 }, { "epoch": 0.12, "learning_rate": 1.9567954323775566e-05, "loss": 1.0338, "step": 2384 }, { "epoch": 0.12, "learning_rate": 1.9567475769940542e-05, "loss": 1.0572, "step": 2385 }, { "epoch": 0.12, "learning_rate": 1.956699695707629e-05, "loss": 0.9932, "step": 2386 }, { "epoch": 0.12, "learning_rate": 1.9566517885195776e-05, "loss": 0.9216, "step": 2387 }, { "epoch": 0.12, "learning_rate": 1.9566038554311967e-05, "loss": 1.0219, "step": 2388 }, { "epoch": 0.12, "learning_rate": 1.956555896443784e-05, "loss": 1.0212, "step": 2389 }, { "epoch": 0.12, "learning_rate": 1.9565079115586383e-05, "loss": 1.0841, "step": 2390 }, { "epoch": 0.12, "learning_rate": 1.9564599007770585e-05, "loss": 1.0005, "step": 2391 }, { "epoch": 0.12, "learning_rate": 1.9564118641003445e-05, "loss": 1.0413, "step": 2392 }, { "epoch": 0.12, "learning_rate": 1.956363801529797e-05, "loss": 1.0145, "step": 2393 }, { "epoch": 0.12, "learning_rate": 1.956315713066717e-05, "loss": 1.0454, "step": 2394 }, { "epoch": 0.12, "learning_rate": 1.956267598712406e-05, "loss": 1.0702, "step": 2395 }, { "epoch": 0.12, "learning_rate": 1.9562194584681677e-05, "loss": 1.0237, "step": 2396 }, { "epoch": 0.12, "learning_rate": 1.9561712923353047e-05, "loss": 0.9858, "step": 2397 }, { "epoch": 0.12, "learning_rate": 1.9561231003151216e-05, "loss": 1.0519, "step": 2398 }, { "epoch": 0.12, "learning_rate": 1.9560748824089222e-05, "loss": 1.0683, "step": 2399 }, { "epoch": 0.12, "learning_rate": 1.956026638618013e-05, "loss": 0.9839, "step": 2400 }, { "epoch": 0.12, "learning_rate": 1.9559783689436994e-05, "loss": 1.033, "step": 2401 }, { "epoch": 0.12, "learning_rate": 1.9559300733872887e-05, "loss": 1.032, "step": 2402 }, { "epoch": 0.12, "learning_rate": 1.9558817519500885e-05, "loss": 1.0501, "step": 2403 }, { "epoch": 0.12, "learning_rate": 1.9558334046334063e-05, "loss": 0.955, "step": 2404 }, { "epoch": 0.12, "learning_rate": 1.955785031438552e-05, "loss": 1.0331, "step": 2405 }, { "epoch": 0.12, "learning_rate": 1.9557366323668345e-05, "loss": 1.0446, "step": 2406 }, { "epoch": 0.12, "learning_rate": 1.9556882074195648e-05, "loss": 1.0971, "step": 2407 }, { "epoch": 0.12, "learning_rate": 1.9556397565980537e-05, "loss": 0.9675, "step": 2408 }, { "epoch": 0.12, "learning_rate": 1.9555912799036127e-05, "loss": 1.0218, "step": 2409 }, { "epoch": 0.12, "learning_rate": 1.9555427773375545e-05, "loss": 0.9762, "step": 2410 }, { "epoch": 0.12, "learning_rate": 1.9554942489011927e-05, "loss": 0.9565, "step": 2411 }, { "epoch": 0.12, "learning_rate": 1.95544569459584e-05, "loss": 1.0876, "step": 2412 }, { "epoch": 0.12, "learning_rate": 1.955397114422812e-05, "loss": 1.0482, "step": 2413 }, { "epoch": 0.12, "learning_rate": 1.9553485083834233e-05, "loss": 1.0887, "step": 2414 }, { "epoch": 0.12, "learning_rate": 1.9552998764789906e-05, "loss": 1.0331, "step": 2415 }, { "epoch": 0.12, "learning_rate": 1.9552512187108296e-05, "loss": 0.9869, "step": 2416 }, { "epoch": 0.12, "learning_rate": 1.9552025350802585e-05, "loss": 0.92, "step": 2417 }, { "epoch": 0.12, "learning_rate": 1.9551538255885948e-05, "loss": 1.0651, "step": 2418 }, { "epoch": 0.12, "learning_rate": 1.9551050902371575e-05, "loss": 1.0735, "step": 2419 }, { "epoch": 0.12, "learning_rate": 1.9550563290272663e-05, "loss": 0.9648, "step": 2420 }, { "epoch": 0.12, "learning_rate": 1.955007541960241e-05, "loss": 1.1506, "step": 2421 }, { "epoch": 0.12, "learning_rate": 1.9549587290374025e-05, "loss": 1.0597, "step": 2422 }, { "epoch": 0.12, "learning_rate": 1.954909890260072e-05, "loss": 1.0245, "step": 2423 }, { "epoch": 0.12, "learning_rate": 1.954861025629573e-05, "loss": 0.9572, "step": 2424 }, { "epoch": 0.12, "learning_rate": 1.954812135147227e-05, "loss": 1.049, "step": 2425 }, { "epoch": 0.12, "learning_rate": 1.9547632188143585e-05, "loss": 1.0671, "step": 2426 }, { "epoch": 0.12, "learning_rate": 1.9547142766322914e-05, "loss": 1.1063, "step": 2427 }, { "epoch": 0.12, "learning_rate": 1.9546653086023513e-05, "loss": 1.0299, "step": 2428 }, { "epoch": 0.12, "learning_rate": 1.9546163147258638e-05, "loss": 1.0006, "step": 2429 }, { "epoch": 0.12, "learning_rate": 1.9545672950041547e-05, "loss": 1.1021, "step": 2430 }, { "epoch": 0.12, "learning_rate": 1.954518249438552e-05, "loss": 1.0468, "step": 2431 }, { "epoch": 0.12, "learning_rate": 1.9544691780303832e-05, "loss": 1.0202, "step": 2432 }, { "epoch": 0.12, "learning_rate": 1.9544200807809767e-05, "loss": 1.0068, "step": 2433 }, { "epoch": 0.12, "learning_rate": 1.9543709576916622e-05, "loss": 1.0153, "step": 2434 }, { "epoch": 0.12, "learning_rate": 1.9543218087637693e-05, "loss": 1.1455, "step": 2435 }, { "epoch": 0.12, "learning_rate": 1.9542726339986285e-05, "loss": 1.0714, "step": 2436 }, { "epoch": 0.12, "learning_rate": 1.9542234333975717e-05, "loss": 1.093, "step": 2437 }, { "epoch": 0.12, "learning_rate": 1.9541742069619306e-05, "loss": 1.0642, "step": 2438 }, { "epoch": 0.12, "learning_rate": 1.954124954693038e-05, "loss": 1.0333, "step": 2439 }, { "epoch": 0.12, "learning_rate": 1.9540756765922272e-05, "loss": 1.0226, "step": 2440 }, { "epoch": 0.12, "learning_rate": 1.9540263726608327e-05, "loss": 0.9754, "step": 2441 }, { "epoch": 0.12, "learning_rate": 1.953977042900189e-05, "loss": 1.0482, "step": 2442 }, { "epoch": 0.12, "learning_rate": 1.953927687311632e-05, "loss": 1.0559, "step": 2443 }, { "epoch": 0.12, "learning_rate": 1.9538783058964976e-05, "loss": 0.9726, "step": 2444 }, { "epoch": 0.12, "learning_rate": 1.953828898656123e-05, "loss": 1.0126, "step": 2445 }, { "epoch": 0.12, "learning_rate": 1.9537794655918457e-05, "loss": 0.9629, "step": 2446 }, { "epoch": 0.12, "learning_rate": 1.9537300067050044e-05, "loss": 1.0363, "step": 2447 }, { "epoch": 0.12, "learning_rate": 1.9536805219969374e-05, "loss": 1.0491, "step": 2448 }, { "epoch": 0.12, "learning_rate": 1.9536310114689852e-05, "loss": 0.978, "step": 2449 }, { "epoch": 0.12, "learning_rate": 1.953581475122488e-05, "loss": 0.9751, "step": 2450 }, { "epoch": 0.12, "learning_rate": 1.9535319129587866e-05, "loss": 0.9707, "step": 2451 }, { "epoch": 0.12, "learning_rate": 1.9534823249792233e-05, "loss": 1.0542, "step": 2452 }, { "epoch": 0.12, "learning_rate": 1.9534327111851406e-05, "loss": 1.048, "step": 2453 }, { "epoch": 0.12, "learning_rate": 1.9533830715778814e-05, "loss": 1.029, "step": 2454 }, { "epoch": 0.12, "learning_rate": 1.95333340615879e-05, "loss": 0.9936, "step": 2455 }, { "epoch": 0.12, "learning_rate": 1.9532837149292106e-05, "loss": 1.0468, "step": 2456 }, { "epoch": 0.12, "learning_rate": 1.953233997890489e-05, "loss": 1.024, "step": 2457 }, { "epoch": 0.12, "learning_rate": 1.9531842550439712e-05, "loss": 1.0824, "step": 2458 }, { "epoch": 0.12, "learning_rate": 1.953134486391004e-05, "loss": 0.9779, "step": 2459 }, { "epoch": 0.12, "learning_rate": 1.9530846919329342e-05, "loss": 1.1034, "step": 2460 }, { "epoch": 0.13, "learning_rate": 1.9530348716711107e-05, "loss": 1.0397, "step": 2461 }, { "epoch": 0.13, "learning_rate": 1.9529850256068818e-05, "loss": 1.0376, "step": 2462 }, { "epoch": 0.13, "learning_rate": 1.9529351537415973e-05, "loss": 1.0995, "step": 2463 }, { "epoch": 0.13, "learning_rate": 1.9528852560766075e-05, "loss": 1.0646, "step": 2464 }, { "epoch": 0.13, "learning_rate": 1.9528353326132627e-05, "loss": 0.9429, "step": 2465 }, { "epoch": 0.13, "learning_rate": 1.9527853833529153e-05, "loss": 1.0528, "step": 2466 }, { "epoch": 0.13, "learning_rate": 1.9527354082969174e-05, "loss": 1.04, "step": 2467 }, { "epoch": 0.13, "learning_rate": 1.9526854074466222e-05, "loss": 1.0168, "step": 2468 }, { "epoch": 0.13, "learning_rate": 1.9526353808033827e-05, "loss": 1.0327, "step": 2469 }, { "epoch": 0.13, "learning_rate": 1.9525853283685537e-05, "loss": 0.9568, "step": 2470 }, { "epoch": 0.13, "learning_rate": 1.9525352501434912e-05, "loss": 1.0885, "step": 2471 }, { "epoch": 0.13, "learning_rate": 1.9524851461295494e-05, "loss": 1.0021, "step": 2472 }, { "epoch": 0.13, "learning_rate": 1.952435016328086e-05, "loss": 0.9321, "step": 2473 }, { "epoch": 0.13, "learning_rate": 1.9523848607404578e-05, "loss": 0.995, "step": 2474 }, { "epoch": 0.13, "learning_rate": 1.9523346793680228e-05, "loss": 1.005, "step": 2475 }, { "epoch": 0.13, "learning_rate": 1.9522844722121394e-05, "loss": 1.0718, "step": 2476 }, { "epoch": 0.13, "learning_rate": 1.9522342392741672e-05, "loss": 1.0304, "step": 2477 }, { "epoch": 0.13, "learning_rate": 1.952183980555466e-05, "loss": 0.931, "step": 2478 }, { "epoch": 0.13, "learning_rate": 1.9521336960573964e-05, "loss": 1.0409, "step": 2479 }, { "epoch": 0.13, "learning_rate": 1.9520833857813205e-05, "loss": 1.0318, "step": 2480 }, { "epoch": 0.13, "learning_rate": 1.952033049728599e-05, "loss": 1.0053, "step": 2481 }, { "epoch": 0.13, "learning_rate": 1.9519826879005964e-05, "loss": 0.9926, "step": 2482 }, { "epoch": 0.13, "learning_rate": 1.951932300298675e-05, "loss": 1.0747, "step": 2483 }, { "epoch": 0.13, "learning_rate": 1.9518818869241995e-05, "loss": 1.0905, "step": 2484 }, { "epoch": 0.13, "learning_rate": 1.9518314477785343e-05, "loss": 1.0134, "step": 2485 }, { "epoch": 0.13, "learning_rate": 1.9517809828630458e-05, "loss": 1.0931, "step": 2486 }, { "epoch": 0.13, "learning_rate": 1.9517304921790992e-05, "loss": 1.0114, "step": 2487 }, { "epoch": 0.13, "learning_rate": 1.9516799757280624e-05, "loss": 0.9973, "step": 2488 }, { "epoch": 0.13, "learning_rate": 1.951629433511303e-05, "loss": 1.0046, "step": 2489 }, { "epoch": 0.13, "learning_rate": 1.951578865530189e-05, "loss": 0.9518, "step": 2490 }, { "epoch": 0.13, "learning_rate": 1.9515282717860894e-05, "loss": 1.0203, "step": 2491 }, { "epoch": 0.13, "learning_rate": 1.9514776522803744e-05, "loss": 1.0587, "step": 2492 }, { "epoch": 0.13, "learning_rate": 1.9514270070144142e-05, "loss": 0.9876, "step": 2493 }, { "epoch": 0.13, "learning_rate": 1.95137633598958e-05, "loss": 1.1228, "step": 2494 }, { "epoch": 0.13, "learning_rate": 1.9513256392072437e-05, "loss": 0.9784, "step": 2495 }, { "epoch": 0.13, "learning_rate": 1.9512749166687777e-05, "loss": 0.9819, "step": 2496 }, { "epoch": 0.13, "learning_rate": 1.9512241683755557e-05, "loss": 1.0588, "step": 2497 }, { "epoch": 0.13, "learning_rate": 1.9511733943289512e-05, "loss": 1.0147, "step": 2498 }, { "epoch": 0.13, "learning_rate": 1.951122594530339e-05, "loss": 1.0604, "step": 2499 }, { "epoch": 0.13, "learning_rate": 1.9510717689810946e-05, "loss": 1.0057, "step": 2500 }, { "epoch": 0.13, "learning_rate": 1.9510209176825937e-05, "loss": 0.9232, "step": 2501 }, { "epoch": 0.13, "learning_rate": 1.9509700406362136e-05, "loss": 1.0966, "step": 2502 }, { "epoch": 0.13, "learning_rate": 1.9509191378433313e-05, "loss": 1.0999, "step": 2503 }, { "epoch": 0.13, "learning_rate": 1.950868209305325e-05, "loss": 1.118, "step": 2504 }, { "epoch": 0.13, "learning_rate": 1.9508172550235734e-05, "loss": 0.9666, "step": 2505 }, { "epoch": 0.13, "learning_rate": 1.950766274999456e-05, "loss": 0.9759, "step": 2506 }, { "epoch": 0.13, "learning_rate": 1.950715269234354e-05, "loss": 1.0383, "step": 2507 }, { "epoch": 0.13, "learning_rate": 1.950664237729647e-05, "loss": 1.0632, "step": 2508 }, { "epoch": 0.13, "learning_rate": 1.9506131804867174e-05, "loss": 1.0635, "step": 2509 }, { "epoch": 0.13, "learning_rate": 1.950562097506947e-05, "loss": 0.9813, "step": 2510 }, { "epoch": 0.13, "learning_rate": 1.9505109887917197e-05, "loss": 1.1253, "step": 2511 }, { "epoch": 0.13, "learning_rate": 1.9504598543424183e-05, "loss": 1.0063, "step": 2512 }, { "epoch": 0.13, "learning_rate": 1.9504086941604274e-05, "loss": 1.053, "step": 2513 }, { "epoch": 0.13, "learning_rate": 1.9503575082471325e-05, "loss": 0.9447, "step": 2514 }, { "epoch": 0.13, "learning_rate": 1.9503062966039188e-05, "loss": 1.0838, "step": 2515 }, { "epoch": 0.13, "learning_rate": 1.9502550592321736e-05, "loss": 1.0359, "step": 2516 }, { "epoch": 0.13, "learning_rate": 1.9502037961332835e-05, "loss": 1.0052, "step": 2517 }, { "epoch": 0.13, "learning_rate": 1.9501525073086368e-05, "loss": 1.053, "step": 2518 }, { "epoch": 0.13, "learning_rate": 1.9501011927596215e-05, "loss": 1.0712, "step": 2519 }, { "epoch": 0.13, "learning_rate": 1.9500498524876273e-05, "loss": 0.9904, "step": 2520 }, { "epoch": 0.13, "learning_rate": 1.9499984864940443e-05, "loss": 1.0533, "step": 2521 }, { "epoch": 0.13, "learning_rate": 1.9499470947802627e-05, "loss": 0.9737, "step": 2522 }, { "epoch": 0.13, "learning_rate": 1.9498956773476742e-05, "loss": 1.0896, "step": 2523 }, { "epoch": 0.13, "learning_rate": 1.9498442341976713e-05, "loss": 1.0458, "step": 2524 }, { "epoch": 0.13, "learning_rate": 1.9497927653316457e-05, "loss": 1.0253, "step": 2525 }, { "epoch": 0.13, "learning_rate": 1.949741270750992e-05, "loss": 1.032, "step": 2526 }, { "epoch": 0.13, "learning_rate": 1.9496897504571036e-05, "loss": 1.0616, "step": 2527 }, { "epoch": 0.13, "learning_rate": 1.9496382044513756e-05, "loss": 0.9842, "step": 2528 }, { "epoch": 0.13, "learning_rate": 1.9495866327352035e-05, "loss": 1.0454, "step": 2529 }, { "epoch": 0.13, "learning_rate": 1.9495350353099836e-05, "loss": 1.1697, "step": 2530 }, { "epoch": 0.13, "learning_rate": 1.949483412177113e-05, "loss": 1.0516, "step": 2531 }, { "epoch": 0.13, "learning_rate": 1.949431763337989e-05, "loss": 1.1298, "step": 2532 }, { "epoch": 0.13, "learning_rate": 1.94938008879401e-05, "loss": 0.996, "step": 2533 }, { "epoch": 0.13, "learning_rate": 1.9493283885465755e-05, "loss": 0.9628, "step": 2534 }, { "epoch": 0.13, "learning_rate": 1.9492766625970847e-05, "loss": 1.0526, "step": 2535 }, { "epoch": 0.13, "learning_rate": 1.9492249109469386e-05, "loss": 1.0366, "step": 2536 }, { "epoch": 0.13, "learning_rate": 1.9491731335975374e-05, "loss": 1.0414, "step": 2537 }, { "epoch": 0.13, "learning_rate": 1.9491213305502835e-05, "loss": 0.9396, "step": 2538 }, { "epoch": 0.13, "learning_rate": 1.9490695018065796e-05, "loss": 0.9607, "step": 2539 }, { "epoch": 0.13, "learning_rate": 1.9490176473678286e-05, "loss": 0.986, "step": 2540 }, { "epoch": 0.13, "learning_rate": 1.9489657672354347e-05, "loss": 1.0342, "step": 2541 }, { "epoch": 0.13, "learning_rate": 1.9489138614108018e-05, "loss": 1.0471, "step": 2542 }, { "epoch": 0.13, "learning_rate": 1.948861929895336e-05, "loss": 1.0325, "step": 2543 }, { "epoch": 0.13, "learning_rate": 1.948809972690443e-05, "loss": 0.9741, "step": 2544 }, { "epoch": 0.13, "learning_rate": 1.948757989797529e-05, "loss": 1.0738, "step": 2545 }, { "epoch": 0.13, "learning_rate": 1.948705981218002e-05, "loss": 0.9464, "step": 2546 }, { "epoch": 0.13, "learning_rate": 1.9486539469532705e-05, "loss": 1.0225, "step": 2547 }, { "epoch": 0.13, "learning_rate": 1.9486018870047417e-05, "loss": 1.1007, "step": 2548 }, { "epoch": 0.13, "learning_rate": 1.9485498013738267e-05, "loss": 1.0017, "step": 2549 }, { "epoch": 0.13, "learning_rate": 1.948497690061935e-05, "loss": 0.9657, "step": 2550 }, { "epoch": 0.13, "learning_rate": 1.9484455530704773e-05, "loss": 1.101, "step": 2551 }, { "epoch": 0.13, "learning_rate": 1.948393390400865e-05, "loss": 0.9919, "step": 2552 }, { "epoch": 0.13, "learning_rate": 1.9483412020545113e-05, "loss": 1.0657, "step": 2553 }, { "epoch": 0.13, "learning_rate": 1.948288988032828e-05, "loss": 1.104, "step": 2554 }, { "epoch": 0.13, "learning_rate": 1.9482367483372295e-05, "loss": 1.0402, "step": 2555 }, { "epoch": 0.13, "learning_rate": 1.94818448296913e-05, "loss": 0.9504, "step": 2556 }, { "epoch": 0.13, "learning_rate": 1.948132191929944e-05, "loss": 0.959, "step": 2557 }, { "epoch": 0.13, "learning_rate": 1.948079875221088e-05, "loss": 1.0637, "step": 2558 }, { "epoch": 0.13, "learning_rate": 1.948027532843978e-05, "loss": 1.0111, "step": 2559 }, { "epoch": 0.13, "learning_rate": 1.9479751648000312e-05, "loss": 0.9634, "step": 2560 }, { "epoch": 0.13, "learning_rate": 1.947922771090665e-05, "loss": 1.0313, "step": 2561 }, { "epoch": 0.13, "learning_rate": 1.9478703517172986e-05, "loss": 0.9877, "step": 2562 }, { "epoch": 0.13, "learning_rate": 1.947817906681351e-05, "loss": 1.0961, "step": 2563 }, { "epoch": 0.13, "learning_rate": 1.9477654359842417e-05, "loss": 0.9694, "step": 2564 }, { "epoch": 0.13, "learning_rate": 1.947712939627392e-05, "loss": 1.0281, "step": 2565 }, { "epoch": 0.13, "learning_rate": 1.9476604176122223e-05, "loss": 1.0301, "step": 2566 }, { "epoch": 0.13, "learning_rate": 1.947607869940155e-05, "loss": 1.1006, "step": 2567 }, { "epoch": 0.13, "learning_rate": 1.947555296612613e-05, "loss": 1.0229, "step": 2568 }, { "epoch": 0.13, "learning_rate": 1.9475026976310196e-05, "loss": 1.0414, "step": 2569 }, { "epoch": 0.13, "learning_rate": 1.9474500729967984e-05, "loss": 1.0347, "step": 2570 }, { "epoch": 0.13, "learning_rate": 1.9473974227113744e-05, "loss": 1.0369, "step": 2571 }, { "epoch": 0.13, "learning_rate": 1.947344746776174e-05, "loss": 0.9747, "step": 2572 }, { "epoch": 0.13, "learning_rate": 1.9472920451926215e-05, "loss": 1.0021, "step": 2573 }, { "epoch": 0.13, "learning_rate": 1.947239317962145e-05, "loss": 0.9474, "step": 2574 }, { "epoch": 0.13, "learning_rate": 1.9471865650861716e-05, "loss": 1.0251, "step": 2575 }, { "epoch": 0.13, "learning_rate": 1.94713378656613e-05, "loss": 1.052, "step": 2576 }, { "epoch": 0.13, "learning_rate": 1.9470809824034486e-05, "loss": 1.1143, "step": 2577 }, { "epoch": 0.13, "learning_rate": 1.9470281525995575e-05, "loss": 0.9637, "step": 2578 }, { "epoch": 0.13, "learning_rate": 1.9469752971558866e-05, "loss": 0.9502, "step": 2579 }, { "epoch": 0.13, "learning_rate": 1.946922416073867e-05, "loss": 1.015, "step": 2580 }, { "epoch": 0.13, "learning_rate": 1.9468695093549302e-05, "loss": 0.9212, "step": 2581 }, { "epoch": 0.13, "learning_rate": 1.946816577000509e-05, "loss": 1.0161, "step": 2582 }, { "epoch": 0.13, "learning_rate": 1.946763619012036e-05, "loss": 1.0211, "step": 2583 }, { "epoch": 0.13, "learning_rate": 1.946710635390946e-05, "loss": 1.0184, "step": 2584 }, { "epoch": 0.13, "learning_rate": 1.9466576261386722e-05, "loss": 1.0162, "step": 2585 }, { "epoch": 0.13, "learning_rate": 1.9466045912566507e-05, "loss": 0.9628, "step": 2586 }, { "epoch": 0.13, "learning_rate": 1.9465515307463168e-05, "loss": 1.0435, "step": 2587 }, { "epoch": 0.13, "learning_rate": 1.946498444609107e-05, "loss": 1.0632, "step": 2588 }, { "epoch": 0.13, "learning_rate": 1.9464453328464592e-05, "loss": 1.0857, "step": 2589 }, { "epoch": 0.13, "learning_rate": 1.946392195459811e-05, "loss": 1.0288, "step": 2590 }, { "epoch": 0.13, "learning_rate": 1.9463390324506006e-05, "loss": 1.0324, "step": 2591 }, { "epoch": 0.13, "learning_rate": 1.946285843820268e-05, "loss": 1.0278, "step": 2592 }, { "epoch": 0.13, "learning_rate": 1.9462326295702532e-05, "loss": 1.0442, "step": 2593 }, { "epoch": 0.13, "learning_rate": 1.946179389701996e-05, "loss": 1.0337, "step": 2594 }, { "epoch": 0.13, "learning_rate": 1.9461261242169392e-05, "loss": 0.9743, "step": 2595 }, { "epoch": 0.13, "learning_rate": 1.9460728331165235e-05, "loss": 1.012, "step": 2596 }, { "epoch": 0.13, "learning_rate": 1.946019516402193e-05, "loss": 0.9401, "step": 2597 }, { "epoch": 0.13, "learning_rate": 1.94596617407539e-05, "loss": 0.9826, "step": 2598 }, { "epoch": 0.13, "learning_rate": 1.94591280613756e-05, "loss": 1.0459, "step": 2599 }, { "epoch": 0.13, "learning_rate": 1.9458594125901463e-05, "loss": 0.9854, "step": 2600 }, { "epoch": 0.13, "learning_rate": 1.9458059934345958e-05, "loss": 1.0625, "step": 2601 }, { "epoch": 0.13, "learning_rate": 1.9457525486723542e-05, "loss": 0.9125, "step": 2602 }, { "epoch": 0.13, "learning_rate": 1.9456990783048684e-05, "loss": 0.9185, "step": 2603 }, { "epoch": 0.13, "learning_rate": 1.945645582333587e-05, "loss": 0.968, "step": 2604 }, { "epoch": 0.13, "learning_rate": 1.9455920607599565e-05, "loss": 1.0562, "step": 2605 }, { "epoch": 0.13, "learning_rate": 1.9455385135854272e-05, "loss": 0.9589, "step": 2606 }, { "epoch": 0.13, "learning_rate": 1.945484940811449e-05, "loss": 1.0455, "step": 2607 }, { "epoch": 0.13, "learning_rate": 1.9454313424394718e-05, "loss": 0.9945, "step": 2608 }, { "epoch": 0.13, "learning_rate": 1.9453777184709467e-05, "loss": 1.0742, "step": 2609 }, { "epoch": 0.13, "learning_rate": 1.9453240689073257e-05, "loss": 1.0315, "step": 2610 }, { "epoch": 0.13, "learning_rate": 1.9452703937500615e-05, "loss": 1.0865, "step": 2611 }, { "epoch": 0.13, "learning_rate": 1.945216693000607e-05, "loss": 1.0466, "step": 2612 }, { "epoch": 0.13, "learning_rate": 1.945162966660416e-05, "loss": 0.9381, "step": 2613 }, { "epoch": 0.13, "learning_rate": 1.9451092147309433e-05, "loss": 1.1047, "step": 2614 }, { "epoch": 0.13, "learning_rate": 1.945055437213644e-05, "loss": 0.995, "step": 2615 }, { "epoch": 0.13, "learning_rate": 1.9450016341099747e-05, "loss": 1.0121, "step": 2616 }, { "epoch": 0.13, "learning_rate": 1.944947805421391e-05, "loss": 1.0514, "step": 2617 }, { "epoch": 0.13, "learning_rate": 1.9448939511493513e-05, "loss": 1.0021, "step": 2618 }, { "epoch": 0.13, "learning_rate": 1.944840071295313e-05, "loss": 1.0988, "step": 2619 }, { "epoch": 0.13, "learning_rate": 1.9447861658607348e-05, "loss": 1.1185, "step": 2620 }, { "epoch": 0.13, "learning_rate": 1.9447322348470763e-05, "loss": 1.0163, "step": 2621 }, { "epoch": 0.13, "learning_rate": 1.9446782782557977e-05, "loss": 0.9892, "step": 2622 }, { "epoch": 0.13, "learning_rate": 1.94462429608836e-05, "loss": 1.0499, "step": 2623 }, { "epoch": 0.13, "learning_rate": 1.9445702883462247e-05, "loss": 1.0159, "step": 2624 }, { "epoch": 0.13, "learning_rate": 1.944516255030853e-05, "loss": 1.1042, "step": 2625 }, { "epoch": 0.13, "learning_rate": 1.9444621961437094e-05, "loss": 1.0265, "step": 2626 }, { "epoch": 0.13, "learning_rate": 1.944408111686256e-05, "loss": 0.9884, "step": 2627 }, { "epoch": 0.13, "learning_rate": 1.9443540016599583e-05, "loss": 0.9225, "step": 2628 }, { "epoch": 0.13, "learning_rate": 1.9442998660662803e-05, "loss": 0.9558, "step": 2629 }, { "epoch": 0.13, "learning_rate": 1.9442457049066882e-05, "loss": 1.1388, "step": 2630 }, { "epoch": 0.13, "learning_rate": 1.9441915181826487e-05, "loss": 0.9638, "step": 2631 }, { "epoch": 0.13, "learning_rate": 1.9441373058956278e-05, "loss": 0.987, "step": 2632 }, { "epoch": 0.13, "learning_rate": 1.944083068047094e-05, "loss": 1.1131, "step": 2633 }, { "epoch": 0.13, "learning_rate": 1.9440288046385155e-05, "loss": 1.0434, "step": 2634 }, { "epoch": 0.13, "learning_rate": 1.9439745156713617e-05, "loss": 1.0328, "step": 2635 }, { "epoch": 0.13, "learning_rate": 1.943920201147102e-05, "loss": 0.8995, "step": 2636 }, { "epoch": 0.13, "learning_rate": 1.9438658610672072e-05, "loss": 1.1304, "step": 2637 }, { "epoch": 0.13, "learning_rate": 1.943811495433148e-05, "loss": 0.9987, "step": 2638 }, { "epoch": 0.13, "learning_rate": 1.9437571042463973e-05, "loss": 1.0003, "step": 2639 }, { "epoch": 0.13, "learning_rate": 1.9437026875084268e-05, "loss": 1.0419, "step": 2640 }, { "epoch": 0.13, "learning_rate": 1.9436482452207103e-05, "loss": 0.9104, "step": 2641 }, { "epoch": 0.13, "learning_rate": 1.943593777384721e-05, "loss": 1.1167, "step": 2642 }, { "epoch": 0.13, "learning_rate": 1.9435392840019343e-05, "loss": 1.0839, "step": 2643 }, { "epoch": 0.13, "learning_rate": 1.9434847650738257e-05, "loss": 1.0807, "step": 2644 }, { "epoch": 0.13, "learning_rate": 1.9434302206018704e-05, "loss": 0.9112, "step": 2645 }, { "epoch": 0.13, "learning_rate": 1.943375650587546e-05, "loss": 1.1041, "step": 2646 }, { "epoch": 0.13, "learning_rate": 1.9433210550323292e-05, "loss": 0.9075, "step": 2647 }, { "epoch": 0.13, "learning_rate": 1.9432664339376987e-05, "loss": 1.0281, "step": 2648 }, { "epoch": 0.13, "learning_rate": 1.943211787305133e-05, "loss": 1.0569, "step": 2649 }, { "epoch": 0.13, "learning_rate": 1.9431571151361118e-05, "loss": 1.0274, "step": 2650 }, { "epoch": 0.13, "learning_rate": 1.9431024174321146e-05, "loss": 1.0715, "step": 2651 }, { "epoch": 0.13, "learning_rate": 1.9430476941946232e-05, "loss": 1.0795, "step": 2652 }, { "epoch": 0.13, "learning_rate": 1.942992945425119e-05, "loss": 0.9985, "step": 2653 }, { "epoch": 0.13, "learning_rate": 1.9429381711250837e-05, "loss": 0.9894, "step": 2654 }, { "epoch": 0.13, "learning_rate": 1.9428833712960008e-05, "loss": 1.009, "step": 2655 }, { "epoch": 0.13, "learning_rate": 1.9428285459393537e-05, "loss": 0.9024, "step": 2656 }, { "epoch": 0.13, "learning_rate": 1.942773695056627e-05, "loss": 1.0131, "step": 2657 }, { "epoch": 0.14, "learning_rate": 1.9427188186493053e-05, "loss": 0.9841, "step": 2658 }, { "epoch": 0.14, "learning_rate": 1.9426639167188747e-05, "loss": 1.0015, "step": 2659 }, { "epoch": 0.14, "learning_rate": 1.9426089892668213e-05, "loss": 0.9008, "step": 2660 }, { "epoch": 0.14, "learning_rate": 1.942554036294633e-05, "loss": 1.0947, "step": 2661 }, { "epoch": 0.14, "learning_rate": 1.942499057803796e-05, "loss": 1.0615, "step": 2662 }, { "epoch": 0.14, "learning_rate": 1.9424440537958003e-05, "loss": 1.0618, "step": 2663 }, { "epoch": 0.14, "learning_rate": 1.9423890242721345e-05, "loss": 1.068, "step": 2664 }, { "epoch": 0.14, "learning_rate": 1.9423339692342885e-05, "loss": 1.0596, "step": 2665 }, { "epoch": 0.14, "learning_rate": 1.9422788886837526e-05, "loss": 1.0356, "step": 2666 }, { "epoch": 0.14, "learning_rate": 1.942223782622019e-05, "loss": 1.012, "step": 2667 }, { "epoch": 0.14, "learning_rate": 1.942168651050578e-05, "loss": 0.9052, "step": 2668 }, { "epoch": 0.14, "learning_rate": 1.9421134939709233e-05, "loss": 1.0086, "step": 2669 }, { "epoch": 0.14, "learning_rate": 1.9420583113845485e-05, "loss": 1.0844, "step": 2670 }, { "epoch": 0.14, "learning_rate": 1.942003103292947e-05, "loss": 1.051, "step": 2671 }, { "epoch": 0.14, "learning_rate": 1.941947869697614e-05, "loss": 1.0894, "step": 2672 }, { "epoch": 0.14, "learning_rate": 1.9418926106000443e-05, "loss": 0.9711, "step": 2673 }, { "epoch": 0.14, "learning_rate": 1.9418373260017342e-05, "loss": 1.0543, "step": 2674 }, { "epoch": 0.14, "learning_rate": 1.9417820159041807e-05, "loss": 1.0131, "step": 2675 }, { "epoch": 0.14, "learning_rate": 1.941726680308881e-05, "loss": 1.063, "step": 2676 }, { "epoch": 0.14, "learning_rate": 1.9416713192173333e-05, "loss": 0.9447, "step": 2677 }, { "epoch": 0.14, "learning_rate": 1.9416159326310366e-05, "loss": 1.0389, "step": 2678 }, { "epoch": 0.14, "learning_rate": 1.94156052055149e-05, "loss": 1.0392, "step": 2679 }, { "epoch": 0.14, "learning_rate": 1.9415050829801945e-05, "loss": 1.0123, "step": 2680 }, { "epoch": 0.14, "learning_rate": 1.9414496199186504e-05, "loss": 1.0043, "step": 2681 }, { "epoch": 0.14, "learning_rate": 1.9413941313683595e-05, "loss": 1.0195, "step": 2682 }, { "epoch": 0.14, "learning_rate": 1.941338617330824e-05, "loss": 1.0992, "step": 2683 }, { "epoch": 0.14, "learning_rate": 1.941283077807547e-05, "loss": 0.9886, "step": 2684 }, { "epoch": 0.14, "learning_rate": 1.941227512800032e-05, "loss": 0.9488, "step": 2685 }, { "epoch": 0.14, "learning_rate": 1.9411719223097837e-05, "loss": 1.0378, "step": 2686 }, { "epoch": 0.14, "learning_rate": 1.9411163063383066e-05, "loss": 1.0632, "step": 2687 }, { "epoch": 0.14, "learning_rate": 1.941060664887107e-05, "loss": 1.0387, "step": 2688 }, { "epoch": 0.14, "learning_rate": 1.941004997957691e-05, "loss": 0.9079, "step": 2689 }, { "epoch": 0.14, "learning_rate": 1.940949305551566e-05, "loss": 1.0272, "step": 2690 }, { "epoch": 0.14, "learning_rate": 1.9408935876702393e-05, "loss": 0.938, "step": 2691 }, { "epoch": 0.14, "learning_rate": 1.94083784431522e-05, "loss": 1.0826, "step": 2692 }, { "epoch": 0.14, "learning_rate": 1.940782075488017e-05, "loss": 1.0399, "step": 2693 }, { "epoch": 0.14, "learning_rate": 1.9407262811901404e-05, "loss": 1.0916, "step": 2694 }, { "epoch": 0.14, "learning_rate": 1.9406704614231003e-05, "loss": 0.9544, "step": 2695 }, { "epoch": 0.14, "learning_rate": 1.9406146161884086e-05, "loss": 0.9949, "step": 2696 }, { "epoch": 0.14, "learning_rate": 1.9405587454875765e-05, "loss": 1.1099, "step": 2697 }, { "epoch": 0.14, "learning_rate": 1.940502849322117e-05, "loss": 0.9388, "step": 2698 }, { "epoch": 0.14, "learning_rate": 1.940446927693544e-05, "loss": 1.0789, "step": 2699 }, { "epoch": 0.14, "learning_rate": 1.9403909806033703e-05, "loss": 1.1025, "step": 2700 }, { "epoch": 0.14, "learning_rate": 1.9403350080531117e-05, "loss": 1.0187, "step": 2701 }, { "epoch": 0.14, "learning_rate": 1.9402790100442833e-05, "loss": 0.9533, "step": 2702 }, { "epoch": 0.14, "learning_rate": 1.9402229865784012e-05, "loss": 0.9976, "step": 2703 }, { "epoch": 0.14, "learning_rate": 1.9401669376569818e-05, "loss": 1.0101, "step": 2704 }, { "epoch": 0.14, "learning_rate": 1.9401108632815426e-05, "loss": 1.0813, "step": 2705 }, { "epoch": 0.14, "learning_rate": 1.9400547634536023e-05, "loss": 0.9169, "step": 2706 }, { "epoch": 0.14, "learning_rate": 1.939998638174679e-05, "loss": 1.0627, "step": 2707 }, { "epoch": 0.14, "learning_rate": 1.9399424874462932e-05, "loss": 0.9254, "step": 2708 }, { "epoch": 0.14, "learning_rate": 1.9398863112699645e-05, "loss": 1.0757, "step": 2709 }, { "epoch": 0.14, "learning_rate": 1.9398301096472136e-05, "loss": 1.0586, "step": 2710 }, { "epoch": 0.14, "learning_rate": 1.9397738825795625e-05, "loss": 0.9902, "step": 2711 }, { "epoch": 0.14, "learning_rate": 1.9397176300685335e-05, "loss": 1.0413, "step": 2712 }, { "epoch": 0.14, "learning_rate": 1.9396613521156495e-05, "loss": 1.0186, "step": 2713 }, { "epoch": 0.14, "learning_rate": 1.9396050487224338e-05, "loss": 1.0612, "step": 2714 }, { "epoch": 0.14, "learning_rate": 1.939548719890411e-05, "loss": 1.0206, "step": 2715 }, { "epoch": 0.14, "learning_rate": 1.9394923656211064e-05, "loss": 1.0514, "step": 2716 }, { "epoch": 0.14, "learning_rate": 1.9394359859160456e-05, "loss": 0.9583, "step": 2717 }, { "epoch": 0.14, "learning_rate": 1.939379580776755e-05, "loss": 1.0383, "step": 2718 }, { "epoch": 0.14, "learning_rate": 1.9393231502047616e-05, "loss": 1.0659, "step": 2719 }, { "epoch": 0.14, "learning_rate": 1.939266694201593e-05, "loss": 0.9584, "step": 2720 }, { "epoch": 0.14, "learning_rate": 1.9392102127687783e-05, "loss": 0.9974, "step": 2721 }, { "epoch": 0.14, "learning_rate": 1.939153705907846e-05, "loss": 0.9498, "step": 2722 }, { "epoch": 0.14, "learning_rate": 1.939097173620327e-05, "loss": 1.0254, "step": 2723 }, { "epoch": 0.14, "learning_rate": 1.9390406159077507e-05, "loss": 1.0316, "step": 2724 }, { "epoch": 0.14, "learning_rate": 1.9389840327716486e-05, "loss": 1.026, "step": 2725 }, { "epoch": 0.14, "learning_rate": 1.9389274242135528e-05, "loss": 1.0938, "step": 2726 }, { "epoch": 0.14, "learning_rate": 1.9388707902349964e-05, "loss": 0.9837, "step": 2727 }, { "epoch": 0.14, "learning_rate": 1.9388141308375118e-05, "loss": 0.9405, "step": 2728 }, { "epoch": 0.14, "learning_rate": 1.938757446022634e-05, "loss": 0.9775, "step": 2729 }, { "epoch": 0.14, "learning_rate": 1.9387007357918964e-05, "loss": 0.9417, "step": 2730 }, { "epoch": 0.14, "learning_rate": 1.9386440001468353e-05, "loss": 1.0316, "step": 2731 }, { "epoch": 0.14, "learning_rate": 1.9385872390889866e-05, "loss": 1.0184, "step": 2732 }, { "epoch": 0.14, "learning_rate": 1.9385304526198873e-05, "loss": 1.0804, "step": 2733 }, { "epoch": 0.14, "learning_rate": 1.9384736407410738e-05, "loss": 1.0036, "step": 2734 }, { "epoch": 0.14, "learning_rate": 1.9384168034540854e-05, "loss": 0.9841, "step": 2735 }, { "epoch": 0.14, "learning_rate": 1.9383599407604606e-05, "loss": 1.0261, "step": 2736 }, { "epoch": 0.14, "learning_rate": 1.9383030526617384e-05, "loss": 0.9886, "step": 2737 }, { "epoch": 0.14, "learning_rate": 1.9382461391594593e-05, "loss": 0.9249, "step": 2738 }, { "epoch": 0.14, "learning_rate": 1.9381892002551646e-05, "loss": 1.0274, "step": 2739 }, { "epoch": 0.14, "learning_rate": 1.9381322359503948e-05, "loss": 0.9233, "step": 2740 }, { "epoch": 0.14, "learning_rate": 1.9380752462466933e-05, "loss": 0.9524, "step": 2741 }, { "epoch": 0.14, "learning_rate": 1.9380182311456022e-05, "loss": 1.0642, "step": 2742 }, { "epoch": 0.14, "learning_rate": 1.937961190648666e-05, "loss": 1.0671, "step": 2743 }, { "epoch": 0.14, "learning_rate": 1.937904124757428e-05, "loss": 1.0615, "step": 2744 }, { "epoch": 0.14, "learning_rate": 1.937847033473434e-05, "loss": 0.9678, "step": 2745 }, { "epoch": 0.14, "learning_rate": 1.9377899167982294e-05, "loss": 0.9456, "step": 2746 }, { "epoch": 0.14, "learning_rate": 1.9377327747333604e-05, "loss": 1.042, "step": 2747 }, { "epoch": 0.14, "learning_rate": 1.937675607280374e-05, "loss": 1.0145, "step": 2748 }, { "epoch": 0.14, "learning_rate": 1.9376184144408183e-05, "loss": 1.0176, "step": 2749 }, { "epoch": 0.14, "learning_rate": 1.9375611962162416e-05, "loss": 1.0692, "step": 2750 }, { "epoch": 0.14, "learning_rate": 1.9375039526081928e-05, "loss": 1.0422, "step": 2751 }, { "epoch": 0.14, "learning_rate": 1.937446683618222e-05, "loss": 1.0163, "step": 2752 }, { "epoch": 0.14, "learning_rate": 1.9373893892478798e-05, "loss": 1.0052, "step": 2753 }, { "epoch": 0.14, "learning_rate": 1.937332069498717e-05, "loss": 1.0137, "step": 2754 }, { "epoch": 0.14, "learning_rate": 1.937274724372286e-05, "loss": 0.9228, "step": 2755 }, { "epoch": 0.14, "learning_rate": 1.9372173538701384e-05, "loss": 1.0702, "step": 2756 }, { "epoch": 0.14, "learning_rate": 1.9371599579938285e-05, "loss": 0.9951, "step": 2757 }, { "epoch": 0.14, "learning_rate": 1.9371025367449096e-05, "loss": 1.0817, "step": 2758 }, { "epoch": 0.14, "learning_rate": 1.9370450901249368e-05, "loss": 0.9893, "step": 2759 }, { "epoch": 0.14, "learning_rate": 1.936987618135465e-05, "loss": 1.0106, "step": 2760 }, { "epoch": 0.14, "learning_rate": 1.9369301207780503e-05, "loss": 1.0425, "step": 2761 }, { "epoch": 0.14, "learning_rate": 1.9368725980542492e-05, "loss": 1.0775, "step": 2762 }, { "epoch": 0.14, "learning_rate": 1.9368150499656193e-05, "loss": 1.0408, "step": 2763 }, { "epoch": 0.14, "learning_rate": 1.936757476513719e-05, "loss": 1.0244, "step": 2764 }, { "epoch": 0.14, "learning_rate": 1.936699877700106e-05, "loss": 1.0305, "step": 2765 }, { "epoch": 0.14, "learning_rate": 1.936642253526341e-05, "loss": 0.9188, "step": 2766 }, { "epoch": 0.14, "learning_rate": 1.9365846039939832e-05, "loss": 0.8965, "step": 2767 }, { "epoch": 0.14, "learning_rate": 1.9365269291045936e-05, "loss": 1.0273, "step": 2768 }, { "epoch": 0.14, "learning_rate": 1.936469228859734e-05, "loss": 0.9186, "step": 2769 }, { "epoch": 0.14, "learning_rate": 1.936411503260966e-05, "loss": 1.0887, "step": 2770 }, { "epoch": 0.14, "learning_rate": 1.9363537523098535e-05, "loss": 0.9737, "step": 2771 }, { "epoch": 0.14, "learning_rate": 1.936295976007959e-05, "loss": 0.9867, "step": 2772 }, { "epoch": 0.14, "learning_rate": 1.9362381743568467e-05, "loss": 1.0632, "step": 2773 }, { "epoch": 0.14, "learning_rate": 1.9361803473580823e-05, "loss": 1.0785, "step": 2774 }, { "epoch": 0.14, "learning_rate": 1.9361224950132313e-05, "loss": 1.0648, "step": 2775 }, { "epoch": 0.14, "learning_rate": 1.936064617323859e-05, "loss": 1.0648, "step": 2776 }, { "epoch": 0.14, "learning_rate": 1.9360067142915337e-05, "loss": 1.0782, "step": 2777 }, { "epoch": 0.14, "learning_rate": 1.935948785917822e-05, "loss": 0.9337, "step": 2778 }, { "epoch": 0.14, "learning_rate": 1.935890832204293e-05, "loss": 0.9748, "step": 2779 }, { "epoch": 0.14, "learning_rate": 1.9358328531525155e-05, "loss": 1.0432, "step": 2780 }, { "epoch": 0.14, "learning_rate": 1.9357748487640585e-05, "loss": 1.0284, "step": 2781 }, { "epoch": 0.14, "learning_rate": 1.9357168190404937e-05, "loss": 1.0525, "step": 2782 }, { "epoch": 0.14, "learning_rate": 1.9356587639833914e-05, "loss": 1.0294, "step": 2783 }, { "epoch": 0.14, "learning_rate": 1.9356006835943235e-05, "loss": 1.0638, "step": 2784 }, { "epoch": 0.14, "learning_rate": 1.9355425778748627e-05, "loss": 1.1008, "step": 2785 }, { "epoch": 0.14, "learning_rate": 1.9354844468265814e-05, "loss": 1.1483, "step": 2786 }, { "epoch": 0.14, "learning_rate": 1.9354262904510544e-05, "loss": 1.0643, "step": 2787 }, { "epoch": 0.14, "learning_rate": 1.935368108749856e-05, "loss": 0.9733, "step": 2788 }, { "epoch": 0.14, "learning_rate": 1.9353099017245608e-05, "loss": 1.0195, "step": 2789 }, { "epoch": 0.14, "learning_rate": 1.935251669376745e-05, "loss": 1.0081, "step": 2790 }, { "epoch": 0.14, "learning_rate": 1.9351934117079855e-05, "loss": 1.0182, "step": 2791 }, { "epoch": 0.14, "learning_rate": 1.9351351287198598e-05, "loss": 1.1686, "step": 2792 }, { "epoch": 0.14, "learning_rate": 1.9350768204139447e-05, "loss": 1.1073, "step": 2793 }, { "epoch": 0.14, "learning_rate": 1.9350184867918197e-05, "loss": 1.0105, "step": 2794 }, { "epoch": 0.14, "learning_rate": 1.934960127855064e-05, "loss": 1.0421, "step": 2795 }, { "epoch": 0.14, "learning_rate": 1.9349017436052575e-05, "loss": 1.0297, "step": 2796 }, { "epoch": 0.14, "learning_rate": 1.934843334043981e-05, "loss": 0.9268, "step": 2797 }, { "epoch": 0.14, "learning_rate": 1.934784899172816e-05, "loss": 1.0079, "step": 2798 }, { "epoch": 0.14, "learning_rate": 1.9347264389933442e-05, "loss": 1.0004, "step": 2799 }, { "epoch": 0.14, "learning_rate": 1.9346679535071487e-05, "loss": 1.0011, "step": 2800 }, { "epoch": 0.14, "learning_rate": 1.9346094427158127e-05, "loss": 1.0153, "step": 2801 }, { "epoch": 0.14, "learning_rate": 1.9345509066209202e-05, "loss": 0.9487, "step": 2802 }, { "epoch": 0.14, "learning_rate": 1.934492345224056e-05, "loss": 1.0157, "step": 2803 }, { "epoch": 0.14, "learning_rate": 1.9344337585268065e-05, "loss": 1.0114, "step": 2804 }, { "epoch": 0.14, "learning_rate": 1.9343751465307567e-05, "loss": 0.8921, "step": 2805 }, { "epoch": 0.14, "learning_rate": 1.934316509237494e-05, "loss": 1.1151, "step": 2806 }, { "epoch": 0.14, "learning_rate": 1.934257846648606e-05, "loss": 0.9119, "step": 2807 }, { "epoch": 0.14, "learning_rate": 1.9341991587656803e-05, "loss": 1.0157, "step": 2808 }, { "epoch": 0.14, "learning_rate": 1.9341404455903065e-05, "loss": 0.9156, "step": 2809 }, { "epoch": 0.14, "learning_rate": 1.9340817071240742e-05, "loss": 0.8927, "step": 2810 }, { "epoch": 0.14, "learning_rate": 1.9340229433685732e-05, "loss": 1.0059, "step": 2811 }, { "epoch": 0.14, "learning_rate": 1.9339641543253948e-05, "loss": 0.9409, "step": 2812 }, { "epoch": 0.14, "learning_rate": 1.9339053399961308e-05, "loss": 0.9971, "step": 2813 }, { "epoch": 0.14, "learning_rate": 1.933846500382373e-05, "loss": 0.9906, "step": 2814 }, { "epoch": 0.14, "learning_rate": 1.933787635485715e-05, "loss": 1.0186, "step": 2815 }, { "epoch": 0.14, "learning_rate": 1.9337287453077496e-05, "loss": 0.994, "step": 2816 }, { "epoch": 0.14, "learning_rate": 1.9336698298500728e-05, "loss": 0.9576, "step": 2817 }, { "epoch": 0.14, "learning_rate": 1.9336108891142782e-05, "loss": 0.9951, "step": 2818 }, { "epoch": 0.14, "learning_rate": 1.933551923101962e-05, "loss": 1.0704, "step": 2819 }, { "epoch": 0.14, "learning_rate": 1.9334929318147208e-05, "loss": 1.0131, "step": 2820 }, { "epoch": 0.14, "learning_rate": 1.933433915254152e-05, "loss": 1.0758, "step": 2821 }, { "epoch": 0.14, "learning_rate": 1.9333748734218522e-05, "loss": 1.0027, "step": 2822 }, { "epoch": 0.14, "learning_rate": 1.9333158063194213e-05, "loss": 1.0573, "step": 2823 }, { "epoch": 0.14, "learning_rate": 1.933256713948458e-05, "loss": 0.9193, "step": 2824 }, { "epoch": 0.14, "learning_rate": 1.9331975963105618e-05, "loss": 1.0725, "step": 2825 }, { "epoch": 0.14, "learning_rate": 1.9331384534073337e-05, "loss": 1.0978, "step": 2826 }, { "epoch": 0.14, "learning_rate": 1.9330792852403745e-05, "loss": 1.0401, "step": 2827 }, { "epoch": 0.14, "learning_rate": 1.9330200918112866e-05, "loss": 0.915, "step": 2828 }, { "epoch": 0.14, "learning_rate": 1.9329608731216724e-05, "loss": 0.9493, "step": 2829 }, { "epoch": 0.14, "learning_rate": 1.932901629173135e-05, "loss": 1.0579, "step": 2830 }, { "epoch": 0.14, "learning_rate": 1.9328423599672786e-05, "loss": 1.0903, "step": 2831 }, { "epoch": 0.14, "learning_rate": 1.9327830655057077e-05, "loss": 0.9966, "step": 2832 }, { "epoch": 0.14, "learning_rate": 1.932723745790028e-05, "loss": 0.9666, "step": 2833 }, { "epoch": 0.14, "learning_rate": 1.9326644008218448e-05, "loss": 1.027, "step": 2834 }, { "epoch": 0.14, "learning_rate": 1.9326050306027655e-05, "loss": 1.0515, "step": 2835 }, { "epoch": 0.14, "learning_rate": 1.932545635134397e-05, "loss": 1.1839, "step": 2836 }, { "epoch": 0.14, "learning_rate": 1.9324862144183478e-05, "loss": 0.8802, "step": 2837 }, { "epoch": 0.14, "learning_rate": 1.9324267684562264e-05, "loss": 1.0845, "step": 2838 }, { "epoch": 0.14, "learning_rate": 1.9323672972496422e-05, "loss": 1.0037, "step": 2839 }, { "epoch": 0.14, "learning_rate": 1.9323078008002054e-05, "loss": 1.0536, "step": 2840 }, { "epoch": 0.14, "learning_rate": 1.932248279109527e-05, "loss": 1.1273, "step": 2841 }, { "epoch": 0.14, "learning_rate": 1.9321887321792177e-05, "loss": 1.0376, "step": 2842 }, { "epoch": 0.14, "learning_rate": 1.9321291600108907e-05, "loss": 1.0258, "step": 2843 }, { "epoch": 0.14, "learning_rate": 1.932069562606158e-05, "loss": 1.0699, "step": 2844 }, { "epoch": 0.14, "learning_rate": 1.932009939966634e-05, "loss": 1.0641, "step": 2845 }, { "epoch": 0.14, "learning_rate": 1.9319502920939322e-05, "loss": 0.9908, "step": 2846 }, { "epoch": 0.14, "learning_rate": 1.9318906189896678e-05, "loss": 1.0184, "step": 2847 }, { "epoch": 0.14, "learning_rate": 1.9318309206554567e-05, "loss": 1.0646, "step": 2848 }, { "epoch": 0.14, "learning_rate": 1.9317711970929144e-05, "loss": 0.957, "step": 2849 }, { "epoch": 0.14, "learning_rate": 1.9317114483036584e-05, "loss": 1.0016, "step": 2850 }, { "epoch": 0.14, "learning_rate": 1.931651674289306e-05, "loss": 1.047, "step": 2851 }, { "epoch": 0.14, "learning_rate": 1.931591875051476e-05, "loss": 1.0619, "step": 2852 }, { "epoch": 0.14, "learning_rate": 1.931532050591787e-05, "loss": 1.0785, "step": 2853 }, { "epoch": 0.14, "learning_rate": 1.9314722009118588e-05, "loss": 1.0354, "step": 2854 }, { "epoch": 0.15, "learning_rate": 1.931412326013312e-05, "loss": 1.0028, "step": 2855 }, { "epoch": 0.15, "learning_rate": 1.931352425897767e-05, "loss": 1.0593, "step": 2856 }, { "epoch": 0.15, "learning_rate": 1.931292500566846e-05, "loss": 0.9789, "step": 2857 }, { "epoch": 0.15, "learning_rate": 1.9312325500221717e-05, "loss": 1.0471, "step": 2858 }, { "epoch": 0.15, "learning_rate": 1.9311725742653665e-05, "loss": 1.1395, "step": 2859 }, { "epoch": 0.15, "learning_rate": 1.9311125732980548e-05, "loss": 1.0768, "step": 2860 }, { "epoch": 0.15, "learning_rate": 1.9310525471218607e-05, "loss": 1.1585, "step": 2861 }, { "epoch": 0.15, "learning_rate": 1.9309924957384097e-05, "loss": 1.081, "step": 2862 }, { "epoch": 0.15, "learning_rate": 1.930932419149327e-05, "loss": 1.0853, "step": 2863 }, { "epoch": 0.15, "learning_rate": 1.9308723173562397e-05, "loss": 0.9279, "step": 2864 }, { "epoch": 0.15, "learning_rate": 1.9308121903607748e-05, "loss": 1.072, "step": 2865 }, { "epoch": 0.15, "learning_rate": 1.9307520381645598e-05, "loss": 1.125, "step": 2866 }, { "epoch": 0.15, "learning_rate": 1.930691860769224e-05, "loss": 1.0567, "step": 2867 }, { "epoch": 0.15, "learning_rate": 1.9306316581763962e-05, "loss": 1.1276, "step": 2868 }, { "epoch": 0.15, "learning_rate": 1.930571430387706e-05, "loss": 0.9766, "step": 2869 }, { "epoch": 0.15, "learning_rate": 1.9305111774047848e-05, "loss": 0.9495, "step": 2870 }, { "epoch": 0.15, "learning_rate": 1.9304508992292633e-05, "loss": 1.0945, "step": 2871 }, { "epoch": 0.15, "learning_rate": 1.930390595862774e-05, "loss": 0.9844, "step": 2872 }, { "epoch": 0.15, "learning_rate": 1.9303302673069487e-05, "loss": 1.0553, "step": 2873 }, { "epoch": 0.15, "learning_rate": 1.930269913563421e-05, "loss": 1.0336, "step": 2874 }, { "epoch": 0.15, "learning_rate": 1.9302095346338256e-05, "loss": 1.0113, "step": 2875 }, { "epoch": 0.15, "learning_rate": 1.9301491305197965e-05, "loss": 1.0416, "step": 2876 }, { "epoch": 0.15, "learning_rate": 1.9300887012229693e-05, "loss": 0.9554, "step": 2877 }, { "epoch": 0.15, "learning_rate": 1.9300282467449802e-05, "loss": 0.9656, "step": 2878 }, { "epoch": 0.15, "learning_rate": 1.9299677670874656e-05, "loss": 1.0195, "step": 2879 }, { "epoch": 0.15, "learning_rate": 1.9299072622520633e-05, "loss": 1.0151, "step": 2880 }, { "epoch": 0.15, "learning_rate": 1.9298467322404112e-05, "loss": 1.0221, "step": 2881 }, { "epoch": 0.15, "learning_rate": 1.9297861770541478e-05, "loss": 0.9287, "step": 2882 }, { "epoch": 0.15, "learning_rate": 1.9297255966949132e-05, "loss": 1.0735, "step": 2883 }, { "epoch": 0.15, "learning_rate": 1.9296649911643473e-05, "loss": 1.0814, "step": 2884 }, { "epoch": 0.15, "learning_rate": 1.9296043604640907e-05, "loss": 1.0682, "step": 2885 }, { "epoch": 0.15, "learning_rate": 1.929543704595785e-05, "loss": 1.0885, "step": 2886 }, { "epoch": 0.15, "learning_rate": 1.9294830235610724e-05, "loss": 0.9973, "step": 2887 }, { "epoch": 0.15, "learning_rate": 1.929422317361596e-05, "loss": 1.0243, "step": 2888 }, { "epoch": 0.15, "learning_rate": 1.929361585998999e-05, "loss": 1.0176, "step": 2889 }, { "epoch": 0.15, "learning_rate": 1.929300829474926e-05, "loss": 1.1011, "step": 2890 }, { "epoch": 0.15, "learning_rate": 1.9292400477910217e-05, "loss": 1.0616, "step": 2891 }, { "epoch": 0.15, "learning_rate": 1.9291792409489315e-05, "loss": 0.9909, "step": 2892 }, { "epoch": 0.15, "learning_rate": 1.9291184089503023e-05, "loss": 1.2152, "step": 2893 }, { "epoch": 0.15, "learning_rate": 1.9290575517967806e-05, "loss": 1.0082, "step": 2894 }, { "epoch": 0.15, "learning_rate": 1.9289966694900138e-05, "loss": 1.1087, "step": 2895 }, { "epoch": 0.15, "learning_rate": 1.928935762031651e-05, "loss": 1.0415, "step": 2896 }, { "epoch": 0.15, "learning_rate": 1.9288748294233403e-05, "loss": 1.0728, "step": 2897 }, { "epoch": 0.15, "learning_rate": 1.9288138716667322e-05, "loss": 1.0652, "step": 2898 }, { "epoch": 0.15, "learning_rate": 1.9287528887634765e-05, "loss": 1.0005, "step": 2899 }, { "epoch": 0.15, "learning_rate": 1.9286918807152247e-05, "loss": 1.0346, "step": 2900 }, { "epoch": 0.15, "learning_rate": 1.928630847523628e-05, "loss": 1.0632, "step": 2901 }, { "epoch": 0.15, "learning_rate": 1.9285697891903392e-05, "loss": 1.0089, "step": 2902 }, { "epoch": 0.15, "learning_rate": 1.9285087057170113e-05, "loss": 1.0331, "step": 2903 }, { "epoch": 0.15, "learning_rate": 1.928447597105298e-05, "loss": 0.9492, "step": 2904 }, { "epoch": 0.15, "learning_rate": 1.9283864633568537e-05, "loss": 1.0176, "step": 2905 }, { "epoch": 0.15, "learning_rate": 1.9283253044733338e-05, "loss": 1.0303, "step": 2906 }, { "epoch": 0.15, "learning_rate": 1.9282641204563938e-05, "loss": 1.0862, "step": 2907 }, { "epoch": 0.15, "learning_rate": 1.9282029113076905e-05, "loss": 1.0229, "step": 2908 }, { "epoch": 0.15, "learning_rate": 1.9281416770288806e-05, "loss": 0.999, "step": 2909 }, { "epoch": 0.15, "learning_rate": 1.9280804176216227e-05, "loss": 0.9901, "step": 2910 }, { "epoch": 0.15, "learning_rate": 1.928019133087575e-05, "loss": 1.0881, "step": 2911 }, { "epoch": 0.15, "learning_rate": 1.927957823428396e-05, "loss": 1.0244, "step": 2912 }, { "epoch": 0.15, "learning_rate": 1.9278964886457467e-05, "loss": 0.9769, "step": 2913 }, { "epoch": 0.15, "learning_rate": 1.927835128741287e-05, "loss": 1.1292, "step": 2914 }, { "epoch": 0.15, "learning_rate": 1.927773743716678e-05, "loss": 0.9469, "step": 2915 }, { "epoch": 0.15, "learning_rate": 1.9277123335735822e-05, "loss": 1.0726, "step": 2916 }, { "epoch": 0.15, "learning_rate": 1.9276508983136623e-05, "loss": 1.0499, "step": 2917 }, { "epoch": 0.15, "learning_rate": 1.927589437938581e-05, "loss": 0.975, "step": 2918 }, { "epoch": 0.15, "learning_rate": 1.9275279524500027e-05, "loss": 0.9833, "step": 2919 }, { "epoch": 0.15, "learning_rate": 1.9274664418495918e-05, "loss": 1.1431, "step": 2920 }, { "epoch": 0.15, "learning_rate": 1.9274049061390135e-05, "loss": 1.0359, "step": 2921 }, { "epoch": 0.15, "learning_rate": 1.927343345319934e-05, "loss": 1.0251, "step": 2922 }, { "epoch": 0.15, "learning_rate": 1.9272817593940207e-05, "loss": 1.0374, "step": 2923 }, { "epoch": 0.15, "learning_rate": 1.9272201483629397e-05, "loss": 0.9897, "step": 2924 }, { "epoch": 0.15, "learning_rate": 1.9271585122283598e-05, "loss": 1.0551, "step": 2925 }, { "epoch": 0.15, "learning_rate": 1.9270968509919497e-05, "loss": 1.1021, "step": 2926 }, { "epoch": 0.15, "learning_rate": 1.9270351646553785e-05, "loss": 0.9787, "step": 2927 }, { "epoch": 0.15, "learning_rate": 1.9269734532203167e-05, "loss": 0.9567, "step": 2928 }, { "epoch": 0.15, "learning_rate": 1.926911716688435e-05, "loss": 1.0134, "step": 2929 }, { "epoch": 0.15, "learning_rate": 1.9268499550614046e-05, "loss": 1.1256, "step": 2930 }, { "epoch": 0.15, "learning_rate": 1.9267881683408974e-05, "loss": 0.9873, "step": 2931 }, { "epoch": 0.15, "learning_rate": 1.9267263565285872e-05, "loss": 1.157, "step": 2932 }, { "epoch": 0.15, "learning_rate": 1.9266645196261463e-05, "loss": 1.0144, "step": 2933 }, { "epoch": 0.15, "learning_rate": 1.92660265763525e-05, "loss": 1.0681, "step": 2934 }, { "epoch": 0.15, "learning_rate": 1.926540770557572e-05, "loss": 0.9729, "step": 2935 }, { "epoch": 0.15, "learning_rate": 1.9264788583947888e-05, "loss": 0.9545, "step": 2936 }, { "epoch": 0.15, "learning_rate": 1.926416921148576e-05, "loss": 0.9891, "step": 2937 }, { "epoch": 0.15, "learning_rate": 1.9263549588206106e-05, "loss": 1.1334, "step": 2938 }, { "epoch": 0.15, "learning_rate": 1.9262929714125706e-05, "loss": 0.8839, "step": 2939 }, { "epoch": 0.15, "learning_rate": 1.9262309589261338e-05, "loss": 0.9651, "step": 2940 }, { "epoch": 0.15, "learning_rate": 1.9261689213629792e-05, "loss": 0.999, "step": 2941 }, { "epoch": 0.15, "learning_rate": 1.926106858724786e-05, "loss": 0.9942, "step": 2942 }, { "epoch": 0.15, "learning_rate": 1.9260447710132353e-05, "loss": 0.9216, "step": 2943 }, { "epoch": 0.15, "learning_rate": 1.9259826582300077e-05, "loss": 1.0082, "step": 2944 }, { "epoch": 0.15, "learning_rate": 1.9259205203767847e-05, "loss": 0.9677, "step": 2945 }, { "epoch": 0.15, "learning_rate": 1.9258583574552486e-05, "loss": 0.9603, "step": 2946 }, { "epoch": 0.15, "learning_rate": 1.9257961694670827e-05, "loss": 1.0386, "step": 2947 }, { "epoch": 0.15, "learning_rate": 1.9257339564139704e-05, "loss": 1.0858, "step": 2948 }, { "epoch": 0.15, "learning_rate": 1.925671718297596e-05, "loss": 1.0966, "step": 2949 }, { "epoch": 0.15, "learning_rate": 1.925609455119645e-05, "loss": 0.8689, "step": 2950 }, { "epoch": 0.15, "learning_rate": 1.9255471668818024e-05, "loss": 0.899, "step": 2951 }, { "epoch": 0.15, "learning_rate": 1.925484853585755e-05, "loss": 1.0731, "step": 2952 }, { "epoch": 0.15, "learning_rate": 1.9254225152331904e-05, "loss": 1.1025, "step": 2953 }, { "epoch": 0.15, "learning_rate": 1.9253601518257953e-05, "loss": 1.0498, "step": 2954 }, { "epoch": 0.15, "learning_rate": 1.9252977633652582e-05, "loss": 1.0655, "step": 2955 }, { "epoch": 0.15, "learning_rate": 1.925235349853269e-05, "loss": 0.9952, "step": 2956 }, { "epoch": 0.15, "learning_rate": 1.9251729112915174e-05, "loss": 0.9828, "step": 2957 }, { "epoch": 0.15, "learning_rate": 1.925110447681693e-05, "loss": 1.004, "step": 2958 }, { "epoch": 0.15, "learning_rate": 1.925047959025488e-05, "loss": 1.1601, "step": 2959 }, { "epoch": 0.15, "learning_rate": 1.9249854453245932e-05, "loss": 1.0668, "step": 2960 }, { "epoch": 0.15, "learning_rate": 1.9249229065807018e-05, "loss": 0.9755, "step": 2961 }, { "epoch": 0.15, "learning_rate": 1.9248603427955067e-05, "loss": 1.1698, "step": 2962 }, { "epoch": 0.15, "learning_rate": 1.9247977539707015e-05, "loss": 1.1061, "step": 2963 }, { "epoch": 0.15, "learning_rate": 1.9247351401079815e-05, "loss": 0.9605, "step": 2964 }, { "epoch": 0.15, "learning_rate": 1.924672501209041e-05, "loss": 1.1241, "step": 2965 }, { "epoch": 0.15, "learning_rate": 1.9246098372755768e-05, "loss": 1.0376, "step": 2966 }, { "epoch": 0.15, "learning_rate": 1.9245471483092846e-05, "loss": 1.0792, "step": 2967 }, { "epoch": 0.15, "learning_rate": 1.9244844343118623e-05, "loss": 1.0018, "step": 2968 }, { "epoch": 0.15, "learning_rate": 1.924421695285007e-05, "loss": 1.0365, "step": 2969 }, { "epoch": 0.15, "learning_rate": 1.924358931230418e-05, "loss": 1.0449, "step": 2970 }, { "epoch": 0.15, "learning_rate": 1.924296142149795e-05, "loss": 1.0242, "step": 2971 }, { "epoch": 0.15, "learning_rate": 1.9242333280448368e-05, "loss": 0.9502, "step": 2972 }, { "epoch": 0.15, "learning_rate": 1.9241704889172443e-05, "loss": 1.0696, "step": 2973 }, { "epoch": 0.15, "learning_rate": 1.9241076247687195e-05, "loss": 1.0901, "step": 2974 }, { "epoch": 0.15, "learning_rate": 1.9240447356009637e-05, "loss": 0.9507, "step": 2975 }, { "epoch": 0.15, "learning_rate": 1.9239818214156798e-05, "loss": 1.0015, "step": 2976 }, { "epoch": 0.15, "learning_rate": 1.9239188822145714e-05, "loss": 0.9981, "step": 2977 }, { "epoch": 0.15, "learning_rate": 1.9238559179993417e-05, "loss": 1.0462, "step": 2978 }, { "epoch": 0.15, "learning_rate": 1.9237929287716964e-05, "loss": 1.039, "step": 2979 }, { "epoch": 0.15, "learning_rate": 1.92372991453334e-05, "loss": 1.0614, "step": 2980 }, { "epoch": 0.15, "learning_rate": 1.9236668752859794e-05, "loss": 1.0298, "step": 2981 }, { "epoch": 0.15, "learning_rate": 1.9236038110313205e-05, "loss": 0.952, "step": 2982 }, { "epoch": 0.15, "learning_rate": 1.923540721771071e-05, "loss": 1.0657, "step": 2983 }, { "epoch": 0.15, "learning_rate": 1.9234776075069395e-05, "loss": 1.015, "step": 2984 }, { "epoch": 0.15, "learning_rate": 1.923414468240634e-05, "loss": 1.0586, "step": 2985 }, { "epoch": 0.15, "learning_rate": 1.9233513039738643e-05, "loss": 0.9717, "step": 2986 }, { "epoch": 0.15, "learning_rate": 1.9232881147083404e-05, "loss": 1.0024, "step": 2987 }, { "epoch": 0.15, "learning_rate": 1.9232249004457727e-05, "loss": 1.0353, "step": 2988 }, { "epoch": 0.15, "learning_rate": 1.9231616611878733e-05, "loss": 1.0002, "step": 2989 }, { "epoch": 0.15, "learning_rate": 1.9230983969363542e-05, "loss": 0.9178, "step": 2990 }, { "epoch": 0.15, "learning_rate": 1.923035107692928e-05, "loss": 1.1345, "step": 2991 }, { "epoch": 0.15, "learning_rate": 1.9229717934593084e-05, "loss": 1.1476, "step": 2992 }, { "epoch": 0.15, "learning_rate": 1.9229084542372094e-05, "loss": 0.9598, "step": 2993 }, { "epoch": 0.15, "learning_rate": 1.922845090028346e-05, "loss": 1.0736, "step": 2994 }, { "epoch": 0.15, "learning_rate": 1.9227817008344334e-05, "loss": 0.8849, "step": 2995 }, { "epoch": 0.15, "learning_rate": 1.9227182866571882e-05, "loss": 1.0421, "step": 2996 }, { "epoch": 0.15, "learning_rate": 1.9226548474983267e-05, "loss": 1.0458, "step": 2997 }, { "epoch": 0.15, "learning_rate": 1.922591383359567e-05, "loss": 1.0792, "step": 2998 }, { "epoch": 0.15, "learning_rate": 1.922527894242628e-05, "loss": 0.9987, "step": 2999 }, { "epoch": 0.15, "learning_rate": 1.9224643801492268e-05, "loss": 0.9869, "step": 3000 }, { "epoch": 0.15, "learning_rate": 1.922400841081084e-05, "loss": 0.983, "step": 3001 }, { "epoch": 0.15, "learning_rate": 1.92233727703992e-05, "loss": 1.07, "step": 3002 }, { "epoch": 0.15, "learning_rate": 1.9222736880274555e-05, "loss": 0.9601, "step": 3003 }, { "epoch": 0.15, "learning_rate": 1.922210074045412e-05, "loss": 0.9576, "step": 3004 }, { "epoch": 0.15, "learning_rate": 1.922146435095512e-05, "loss": 1.044, "step": 3005 }, { "epoch": 0.15, "learning_rate": 1.9220827711794784e-05, "loss": 0.9608, "step": 3006 }, { "epoch": 0.15, "learning_rate": 1.922019082299035e-05, "loss": 1.0968, "step": 3007 }, { "epoch": 0.15, "learning_rate": 1.9219553684559054e-05, "loss": 1.0104, "step": 3008 }, { "epoch": 0.15, "learning_rate": 1.9218916296518154e-05, "loss": 1.0656, "step": 3009 }, { "epoch": 0.15, "learning_rate": 1.9218278658884902e-05, "loss": 0.9612, "step": 3010 }, { "epoch": 0.15, "learning_rate": 1.9217640771676566e-05, "loss": 0.9815, "step": 3011 }, { "epoch": 0.15, "learning_rate": 1.9217002634910406e-05, "loss": 1.0021, "step": 3012 }, { "epoch": 0.15, "learning_rate": 1.9216364248603712e-05, "loss": 0.9718, "step": 3013 }, { "epoch": 0.15, "learning_rate": 1.921572561277376e-05, "loss": 1.1105, "step": 3014 }, { "epoch": 0.15, "learning_rate": 1.921508672743784e-05, "loss": 1.0273, "step": 3015 }, { "epoch": 0.15, "learning_rate": 1.921444759261325e-05, "loss": 1.08, "step": 3016 }, { "epoch": 0.15, "learning_rate": 1.92138082083173e-05, "loss": 0.9097, "step": 3017 }, { "epoch": 0.15, "learning_rate": 1.9213168574567293e-05, "loss": 1.0308, "step": 3018 }, { "epoch": 0.15, "learning_rate": 1.9212528691380546e-05, "loss": 0.9902, "step": 3019 }, { "epoch": 0.15, "learning_rate": 1.9211888558774393e-05, "loss": 1.0248, "step": 3020 }, { "epoch": 0.15, "learning_rate": 1.9211248176766152e-05, "loss": 1.0474, "step": 3021 }, { "epoch": 0.15, "learning_rate": 1.9210607545373166e-05, "loss": 0.9886, "step": 3022 }, { "epoch": 0.15, "learning_rate": 1.9209966664612782e-05, "loss": 0.9888, "step": 3023 }, { "epoch": 0.15, "learning_rate": 1.920932553450235e-05, "loss": 0.9958, "step": 3024 }, { "epoch": 0.15, "learning_rate": 1.9208684155059228e-05, "loss": 0.9714, "step": 3025 }, { "epoch": 0.15, "learning_rate": 1.9208042526300776e-05, "loss": 1.0916, "step": 3026 }, { "epoch": 0.15, "learning_rate": 1.9207400648244372e-05, "loss": 0.9902, "step": 3027 }, { "epoch": 0.15, "learning_rate": 1.9206758520907392e-05, "loss": 0.9806, "step": 3028 }, { "epoch": 0.15, "learning_rate": 1.9206116144307217e-05, "loss": 1.11, "step": 3029 }, { "epoch": 0.15, "learning_rate": 1.9205473518461244e-05, "loss": 1.0349, "step": 3030 }, { "epoch": 0.15, "learning_rate": 1.920483064338687e-05, "loss": 0.9952, "step": 3031 }, { "epoch": 0.15, "learning_rate": 1.9204187519101498e-05, "loss": 0.9404, "step": 3032 }, { "epoch": 0.15, "learning_rate": 1.920354414562254e-05, "loss": 1.1005, "step": 3033 }, { "epoch": 0.15, "learning_rate": 1.920290052296742e-05, "loss": 0.9598, "step": 3034 }, { "epoch": 0.15, "learning_rate": 1.9202256651153556e-05, "loss": 1.0383, "step": 3035 }, { "epoch": 0.15, "learning_rate": 1.9201612530198388e-05, "loss": 1.0002, "step": 3036 }, { "epoch": 0.15, "learning_rate": 1.9200968160119348e-05, "loss": 0.9692, "step": 3037 }, { "epoch": 0.15, "learning_rate": 1.9200323540933884e-05, "loss": 0.9626, "step": 3038 }, { "epoch": 0.15, "learning_rate": 1.9199678672659446e-05, "loss": 0.9819, "step": 3039 }, { "epoch": 0.15, "learning_rate": 1.91990335553135e-05, "loss": 1.0684, "step": 3040 }, { "epoch": 0.15, "learning_rate": 1.9198388188913504e-05, "loss": 1.0875, "step": 3041 }, { "epoch": 0.15, "learning_rate": 1.9197742573476937e-05, "loss": 0.982, "step": 3042 }, { "epoch": 0.15, "learning_rate": 1.9197096709021273e-05, "loss": 0.9503, "step": 3043 }, { "epoch": 0.15, "learning_rate": 1.9196450595564e-05, "loss": 0.8961, "step": 3044 }, { "epoch": 0.15, "learning_rate": 1.9195804233122614e-05, "loss": 0.9086, "step": 3045 }, { "epoch": 0.15, "learning_rate": 1.9195157621714612e-05, "loss": 1.0072, "step": 3046 }, { "epoch": 0.15, "learning_rate": 1.9194510761357497e-05, "loss": 1.1077, "step": 3047 }, { "epoch": 0.15, "learning_rate": 1.919386365206879e-05, "loss": 1.0123, "step": 3048 }, { "epoch": 0.15, "learning_rate": 1.9193216293866004e-05, "loss": 0.9766, "step": 3049 }, { "epoch": 0.15, "learning_rate": 1.9192568686766666e-05, "loss": 1.011, "step": 3050 }, { "epoch": 0.15, "learning_rate": 1.9191920830788312e-05, "loss": 1.054, "step": 3051 }, { "epoch": 0.16, "learning_rate": 1.9191272725948483e-05, "loss": 0.9709, "step": 3052 }, { "epoch": 0.16, "learning_rate": 1.919062437226472e-05, "loss": 1.0517, "step": 3053 }, { "epoch": 0.16, "learning_rate": 1.9189975769754583e-05, "loss": 1.0031, "step": 3054 }, { "epoch": 0.16, "learning_rate": 1.918932691843563e-05, "loss": 1.0152, "step": 3055 }, { "epoch": 0.16, "learning_rate": 1.9188677818325424e-05, "loss": 0.9697, "step": 3056 }, { "epoch": 0.16, "learning_rate": 1.9188028469441543e-05, "loss": 0.9571, "step": 3057 }, { "epoch": 0.16, "learning_rate": 1.9187378871801568e-05, "loss": 1.0388, "step": 3058 }, { "epoch": 0.16, "learning_rate": 1.9186729025423084e-05, "loss": 1.0897, "step": 3059 }, { "epoch": 0.16, "learning_rate": 1.9186078930323685e-05, "loss": 1.0173, "step": 3060 }, { "epoch": 0.16, "learning_rate": 1.9185428586520972e-05, "loss": 0.9753, "step": 3061 }, { "epoch": 0.16, "learning_rate": 1.9184777994032554e-05, "loss": 0.9465, "step": 3062 }, { "epoch": 0.16, "learning_rate": 1.9184127152876046e-05, "loss": 1.0272, "step": 3063 }, { "epoch": 0.16, "learning_rate": 1.9183476063069058e-05, "loss": 1.1343, "step": 3064 }, { "epoch": 0.16, "learning_rate": 1.9182824724629234e-05, "loss": 0.9868, "step": 3065 }, { "epoch": 0.16, "learning_rate": 1.9182173137574197e-05, "loss": 1.0066, "step": 3066 }, { "epoch": 0.16, "learning_rate": 1.918152130192159e-05, "loss": 0.8587, "step": 3067 }, { "epoch": 0.16, "learning_rate": 1.9180869217689066e-05, "loss": 0.9761, "step": 3068 }, { "epoch": 0.16, "learning_rate": 1.9180216884894273e-05, "loss": 1.0244, "step": 3069 }, { "epoch": 0.16, "learning_rate": 1.9179564303554877e-05, "loss": 1.0439, "step": 3070 }, { "epoch": 0.16, "learning_rate": 1.9178911473688543e-05, "loss": 1.0342, "step": 3071 }, { "epoch": 0.16, "learning_rate": 1.9178258395312944e-05, "loss": 1.053, "step": 3072 }, { "epoch": 0.16, "learning_rate": 1.9177605068445766e-05, "loss": 0.9847, "step": 3073 }, { "epoch": 0.16, "learning_rate": 1.9176951493104695e-05, "loss": 1.0463, "step": 3074 }, { "epoch": 0.16, "learning_rate": 1.9176297669307423e-05, "loss": 0.9259, "step": 3075 }, { "epoch": 0.16, "learning_rate": 1.9175643597071662e-05, "loss": 1.0452, "step": 3076 }, { "epoch": 0.16, "learning_rate": 1.9174989276415106e-05, "loss": 0.9857, "step": 3077 }, { "epoch": 0.16, "learning_rate": 1.9174334707355476e-05, "loss": 1.0393, "step": 3078 }, { "epoch": 0.16, "learning_rate": 1.91736798899105e-05, "loss": 0.9876, "step": 3079 }, { "epoch": 0.16, "learning_rate": 1.9173024824097897e-05, "loss": 1.0536, "step": 3080 }, { "epoch": 0.16, "learning_rate": 1.9172369509935408e-05, "loss": 1.0163, "step": 3081 }, { "epoch": 0.16, "learning_rate": 1.9171713947440772e-05, "loss": 0.9052, "step": 3082 }, { "epoch": 0.16, "learning_rate": 1.9171058136631736e-05, "loss": 0.9701, "step": 3083 }, { "epoch": 0.16, "learning_rate": 1.9170402077526063e-05, "loss": 0.9276, "step": 3084 }, { "epoch": 0.16, "learning_rate": 1.916974577014151e-05, "loss": 1.0118, "step": 3085 }, { "epoch": 0.16, "learning_rate": 1.9169089214495843e-05, "loss": 1.0241, "step": 3086 }, { "epoch": 0.16, "learning_rate": 1.9168432410606843e-05, "loss": 0.8881, "step": 3087 }, { "epoch": 0.16, "learning_rate": 1.916777535849229e-05, "loss": 1.0092, "step": 3088 }, { "epoch": 0.16, "learning_rate": 1.916711805816997e-05, "loss": 1.0529, "step": 3089 }, { "epoch": 0.16, "learning_rate": 1.9166460509657686e-05, "loss": 0.9677, "step": 3090 }, { "epoch": 0.16, "learning_rate": 1.9165802712973233e-05, "loss": 0.9454, "step": 3091 }, { "epoch": 0.16, "learning_rate": 1.9165144668134426e-05, "loss": 0.9822, "step": 3092 }, { "epoch": 0.16, "learning_rate": 1.9164486375159076e-05, "loss": 1.0673, "step": 3093 }, { "epoch": 0.16, "learning_rate": 1.9163827834065007e-05, "loss": 1.0352, "step": 3094 }, { "epoch": 0.16, "learning_rate": 1.916316904487005e-05, "loss": 0.8568, "step": 3095 }, { "epoch": 0.16, "learning_rate": 1.916251000759204e-05, "loss": 1.0199, "step": 3096 }, { "epoch": 0.16, "learning_rate": 1.916185072224882e-05, "loss": 0.9852, "step": 3097 }, { "epoch": 0.16, "learning_rate": 1.9161191188858238e-05, "loss": 1.0619, "step": 3098 }, { "epoch": 0.16, "learning_rate": 1.916053140743815e-05, "loss": 1.0325, "step": 3099 }, { "epoch": 0.16, "learning_rate": 1.915987137800642e-05, "loss": 1.0019, "step": 3100 }, { "epoch": 0.16, "learning_rate": 1.9159211100580922e-05, "loss": 0.9976, "step": 3101 }, { "epoch": 0.16, "learning_rate": 1.9158550575179525e-05, "loss": 1.0696, "step": 3102 }, { "epoch": 0.16, "learning_rate": 1.915788980182012e-05, "loss": 1.0791, "step": 3103 }, { "epoch": 0.16, "learning_rate": 1.9157228780520586e-05, "loss": 1.0195, "step": 3104 }, { "epoch": 0.16, "learning_rate": 1.9156567511298827e-05, "loss": 0.9377, "step": 3105 }, { "epoch": 0.16, "learning_rate": 1.9155905994172742e-05, "loss": 1.0604, "step": 3106 }, { "epoch": 0.16, "learning_rate": 1.9155244229160247e-05, "loss": 1.0253, "step": 3107 }, { "epoch": 0.16, "learning_rate": 1.9154582216279253e-05, "loss": 1.0161, "step": 3108 }, { "epoch": 0.16, "learning_rate": 1.9153919955547685e-05, "loss": 1.0164, "step": 3109 }, { "epoch": 0.16, "learning_rate": 1.9153257446983475e-05, "loss": 1.0403, "step": 3110 }, { "epoch": 0.16, "learning_rate": 1.9152594690604556e-05, "loss": 1.0241, "step": 3111 }, { "epoch": 0.16, "learning_rate": 1.9151931686428877e-05, "loss": 1.0625, "step": 3112 }, { "epoch": 0.16, "learning_rate": 1.9151268434474383e-05, "loss": 1.1559, "step": 3113 }, { "epoch": 0.16, "learning_rate": 1.915060493475903e-05, "loss": 0.9973, "step": 3114 }, { "epoch": 0.16, "learning_rate": 1.9149941187300784e-05, "loss": 0.9884, "step": 3115 }, { "epoch": 0.16, "learning_rate": 1.9149277192117616e-05, "loss": 0.9344, "step": 3116 }, { "epoch": 0.16, "learning_rate": 1.9148612949227504e-05, "loss": 1.0571, "step": 3117 }, { "epoch": 0.16, "learning_rate": 1.9147948458648427e-05, "loss": 1.1013, "step": 3118 }, { "epoch": 0.16, "learning_rate": 1.914728372039838e-05, "loss": 1.0282, "step": 3119 }, { "epoch": 0.16, "learning_rate": 1.9146618734495357e-05, "loss": 0.988, "step": 3120 }, { "epoch": 0.16, "learning_rate": 1.9145953500957363e-05, "loss": 0.8621, "step": 3121 }, { "epoch": 0.16, "learning_rate": 1.914528801980241e-05, "loss": 1.0596, "step": 3122 }, { "epoch": 0.16, "learning_rate": 1.914462229104851e-05, "loss": 0.957, "step": 3123 }, { "epoch": 0.16, "learning_rate": 1.9143956314713694e-05, "loss": 1.0544, "step": 3124 }, { "epoch": 0.16, "learning_rate": 1.914329009081599e-05, "loss": 0.9927, "step": 3125 }, { "epoch": 0.16, "learning_rate": 1.9142623619373434e-05, "loss": 1.0674, "step": 3126 }, { "epoch": 0.16, "learning_rate": 1.9141956900404067e-05, "loss": 1.0272, "step": 3127 }, { "epoch": 0.16, "learning_rate": 1.9141289933925947e-05, "loss": 1.0303, "step": 3128 }, { "epoch": 0.16, "learning_rate": 1.9140622719957125e-05, "loss": 0.9924, "step": 3129 }, { "epoch": 0.16, "learning_rate": 1.913995525851567e-05, "loss": 0.9758, "step": 3130 }, { "epoch": 0.16, "learning_rate": 1.9139287549619648e-05, "loss": 0.9483, "step": 3131 }, { "epoch": 0.16, "learning_rate": 1.9138619593287144e-05, "loss": 1.0232, "step": 3132 }, { "epoch": 0.16, "learning_rate": 1.913795138953623e-05, "loss": 1.0298, "step": 3133 }, { "epoch": 0.16, "learning_rate": 1.9137282938385008e-05, "loss": 1.0945, "step": 3134 }, { "epoch": 0.16, "learning_rate": 1.9136614239851573e-05, "loss": 1.0404, "step": 3135 }, { "epoch": 0.16, "learning_rate": 1.9135945293954027e-05, "loss": 1.0655, "step": 3136 }, { "epoch": 0.16, "learning_rate": 1.9135276100710477e-05, "loss": 0.9908, "step": 3137 }, { "epoch": 0.16, "learning_rate": 1.9134606660139052e-05, "loss": 0.9817, "step": 3138 }, { "epoch": 0.16, "learning_rate": 1.913393697225787e-05, "loss": 1.0078, "step": 3139 }, { "epoch": 0.16, "learning_rate": 1.913326703708506e-05, "loss": 1.0176, "step": 3140 }, { "epoch": 0.16, "learning_rate": 1.9132596854638764e-05, "loss": 1.0891, "step": 3141 }, { "epoch": 0.16, "learning_rate": 1.913192642493712e-05, "loss": 0.9686, "step": 3142 }, { "epoch": 0.16, "learning_rate": 1.913125574799829e-05, "loss": 1.0886, "step": 3143 }, { "epoch": 0.16, "learning_rate": 1.9130584823840422e-05, "loss": 1.0281, "step": 3144 }, { "epoch": 0.16, "learning_rate": 1.9129913652481683e-05, "loss": 1.0477, "step": 3145 }, { "epoch": 0.16, "learning_rate": 1.9129242233940245e-05, "loss": 1.0666, "step": 3146 }, { "epoch": 0.16, "learning_rate": 1.9128570568234292e-05, "loss": 1.0532, "step": 3147 }, { "epoch": 0.16, "learning_rate": 1.9127898655381997e-05, "loss": 1.0177, "step": 3148 }, { "epoch": 0.16, "learning_rate": 1.912722649540156e-05, "loss": 0.9944, "step": 3149 }, { "epoch": 0.16, "learning_rate": 1.9126554088311176e-05, "loss": 1.0556, "step": 3150 }, { "epoch": 0.16, "learning_rate": 1.912588143412905e-05, "loss": 0.9835, "step": 3151 }, { "epoch": 0.16, "learning_rate": 1.912520853287339e-05, "loss": 0.983, "step": 3152 }, { "epoch": 0.16, "learning_rate": 1.9124535384562423e-05, "loss": 1.0896, "step": 3153 }, { "epoch": 0.16, "learning_rate": 1.9123861989214365e-05, "loss": 0.943, "step": 3154 }, { "epoch": 0.16, "learning_rate": 1.912318834684745e-05, "loss": 1.0367, "step": 3155 }, { "epoch": 0.16, "learning_rate": 1.9122514457479917e-05, "loss": 1.0653, "step": 3156 }, { "epoch": 0.16, "learning_rate": 1.912184032113001e-05, "loss": 1.0079, "step": 3157 }, { "epoch": 0.16, "learning_rate": 1.9121165937815984e-05, "loss": 1.0381, "step": 3158 }, { "epoch": 0.16, "learning_rate": 1.9120491307556094e-05, "loss": 0.9996, "step": 3159 }, { "epoch": 0.16, "learning_rate": 1.9119816430368604e-05, "loss": 1.0322, "step": 3160 }, { "epoch": 0.16, "learning_rate": 1.9119141306271786e-05, "loss": 1.0529, "step": 3161 }, { "epoch": 0.16, "learning_rate": 1.9118465935283922e-05, "loss": 1.0338, "step": 3162 }, { "epoch": 0.16, "learning_rate": 1.911779031742329e-05, "loss": 1.0585, "step": 3163 }, { "epoch": 0.16, "learning_rate": 1.911711445270819e-05, "loss": 0.9961, "step": 3164 }, { "epoch": 0.16, "learning_rate": 1.9116438341156915e-05, "loss": 0.9625, "step": 3165 }, { "epoch": 0.16, "learning_rate": 1.911576198278777e-05, "loss": 1.0634, "step": 3166 }, { "epoch": 0.16, "learning_rate": 1.9115085377619067e-05, "loss": 0.9127, "step": 3167 }, { "epoch": 0.16, "learning_rate": 1.9114408525669124e-05, "loss": 1.0692, "step": 3168 }, { "epoch": 0.16, "learning_rate": 1.9113731426956272e-05, "loss": 1.0381, "step": 3169 }, { "epoch": 0.16, "learning_rate": 1.9113054081498833e-05, "loss": 0.9774, "step": 3170 }, { "epoch": 0.16, "learning_rate": 1.9112376489315153e-05, "loss": 1.0284, "step": 3171 }, { "epoch": 0.16, "learning_rate": 1.9111698650423576e-05, "loss": 1.0682, "step": 3172 }, { "epoch": 0.16, "learning_rate": 1.911102056484245e-05, "loss": 0.8789, "step": 3173 }, { "epoch": 0.16, "learning_rate": 1.9110342232590132e-05, "loss": 0.9629, "step": 3174 }, { "epoch": 0.16, "learning_rate": 1.9109663653684995e-05, "loss": 1.0515, "step": 3175 }, { "epoch": 0.16, "learning_rate": 1.9108984828145406e-05, "loss": 1.0561, "step": 3176 }, { "epoch": 0.16, "learning_rate": 1.910830575598974e-05, "loss": 0.9934, "step": 3177 }, { "epoch": 0.16, "learning_rate": 1.9107626437236392e-05, "loss": 1.0296, "step": 3178 }, { "epoch": 0.16, "learning_rate": 1.910694687190375e-05, "loss": 1.0479, "step": 3179 }, { "epoch": 0.16, "learning_rate": 1.91062670600102e-05, "loss": 1.0976, "step": 3180 }, { "epoch": 0.16, "learning_rate": 1.9105587001574167e-05, "loss": 1.0167, "step": 3181 }, { "epoch": 0.16, "learning_rate": 1.910490669661405e-05, "loss": 0.9776, "step": 3182 }, { "epoch": 0.16, "learning_rate": 1.910422614514827e-05, "loss": 0.9713, "step": 3183 }, { "epoch": 0.16, "learning_rate": 1.9103545347195255e-05, "loss": 1.0929, "step": 3184 }, { "epoch": 0.16, "learning_rate": 1.9102864302773436e-05, "loss": 0.9551, "step": 3185 }, { "epoch": 0.16, "learning_rate": 1.9102183011901248e-05, "loss": 0.9862, "step": 3186 }, { "epoch": 0.16, "learning_rate": 1.910150147459714e-05, "loss": 1.0338, "step": 3187 }, { "epoch": 0.16, "learning_rate": 1.9100819690879564e-05, "loss": 1.0855, "step": 3188 }, { "epoch": 0.16, "learning_rate": 1.9100137660766975e-05, "loss": 1.1003, "step": 3189 }, { "epoch": 0.16, "learning_rate": 1.9099455384277844e-05, "loss": 0.9735, "step": 3190 }, { "epoch": 0.16, "learning_rate": 1.9098772861430638e-05, "loss": 0.9808, "step": 3191 }, { "epoch": 0.16, "learning_rate": 1.9098090092243838e-05, "loss": 0.9783, "step": 3192 }, { "epoch": 0.16, "learning_rate": 1.9097407076735927e-05, "loss": 0.9736, "step": 3193 }, { "epoch": 0.16, "learning_rate": 1.9096723814925397e-05, "loss": 0.984, "step": 3194 }, { "epoch": 0.16, "learning_rate": 1.9096040306830747e-05, "loss": 0.9409, "step": 3195 }, { "epoch": 0.16, "learning_rate": 1.9095356552470487e-05, "loss": 0.9633, "step": 3196 }, { "epoch": 0.16, "learning_rate": 1.9094672551863126e-05, "loss": 1.0929, "step": 3197 }, { "epoch": 0.16, "learning_rate": 1.9093988305027177e-05, "loss": 0.9252, "step": 3198 }, { "epoch": 0.16, "learning_rate": 1.9093303811981173e-05, "loss": 0.954, "step": 3199 }, { "epoch": 0.16, "learning_rate": 1.909261907274364e-05, "loss": 1.0833, "step": 3200 }, { "epoch": 0.16, "learning_rate": 1.909193408733312e-05, "loss": 1.0521, "step": 3201 }, { "epoch": 0.16, "learning_rate": 1.909124885576816e-05, "loss": 1.0597, "step": 3202 }, { "epoch": 0.16, "learning_rate": 1.9090563378067307e-05, "loss": 0.9743, "step": 3203 }, { "epoch": 0.16, "learning_rate": 1.9089877654249125e-05, "loss": 1.0462, "step": 3204 }, { "epoch": 0.16, "learning_rate": 1.9089191684332174e-05, "loss": 0.9569, "step": 3205 }, { "epoch": 0.16, "learning_rate": 1.908850546833503e-05, "loss": 1.0168, "step": 3206 }, { "epoch": 0.16, "learning_rate": 1.908781900627627e-05, "loss": 0.9313, "step": 3207 }, { "epoch": 0.16, "learning_rate": 1.9087132298174476e-05, "loss": 0.9764, "step": 3208 }, { "epoch": 0.16, "learning_rate": 1.9086445344048246e-05, "loss": 0.9328, "step": 3209 }, { "epoch": 0.16, "learning_rate": 1.9085758143916174e-05, "loss": 0.9819, "step": 3210 }, { "epoch": 0.16, "learning_rate": 1.9085070697796868e-05, "loss": 1.0064, "step": 3211 }, { "epoch": 0.16, "learning_rate": 1.9084383005708938e-05, "loss": 1.013, "step": 3212 }, { "epoch": 0.16, "learning_rate": 1.9083695067671005e-05, "loss": 0.9323, "step": 3213 }, { "epoch": 0.16, "learning_rate": 1.9083006883701688e-05, "loss": 1.0623, "step": 3214 }, { "epoch": 0.16, "learning_rate": 1.9082318453819627e-05, "loss": 0.9579, "step": 3215 }, { "epoch": 0.16, "learning_rate": 1.9081629778043452e-05, "loss": 1.0927, "step": 3216 }, { "epoch": 0.16, "learning_rate": 1.908094085639182e-05, "loss": 0.9484, "step": 3217 }, { "epoch": 0.16, "learning_rate": 1.908025168888337e-05, "loss": 0.9514, "step": 3218 }, { "epoch": 0.16, "learning_rate": 1.9079562275536767e-05, "loss": 1.0662, "step": 3219 }, { "epoch": 0.16, "learning_rate": 1.9078872616370674e-05, "loss": 1.0508, "step": 3220 }, { "epoch": 0.16, "learning_rate": 1.907818271140377e-05, "loss": 0.8887, "step": 3221 }, { "epoch": 0.16, "learning_rate": 1.9077492560654723e-05, "loss": 0.9453, "step": 3222 }, { "epoch": 0.16, "learning_rate": 1.9076802164142224e-05, "loss": 1.0866, "step": 3223 }, { "epoch": 0.16, "learning_rate": 1.907611152188496e-05, "loss": 1.0099, "step": 3224 }, { "epoch": 0.16, "learning_rate": 1.9075420633901638e-05, "loss": 1.0165, "step": 3225 }, { "epoch": 0.16, "learning_rate": 1.9074729500210955e-05, "loss": 1.0286, "step": 3226 }, { "epoch": 0.16, "learning_rate": 1.9074038120831626e-05, "loss": 0.9963, "step": 3227 }, { "epoch": 0.16, "learning_rate": 1.907334649578237e-05, "loss": 1.0428, "step": 3228 }, { "epoch": 0.16, "learning_rate": 1.9072654625081908e-05, "loss": 1.0274, "step": 3229 }, { "epoch": 0.16, "learning_rate": 1.9071962508748977e-05, "loss": 0.9785, "step": 3230 }, { "epoch": 0.16, "learning_rate": 1.907127014680231e-05, "loss": 0.9338, "step": 3231 }, { "epoch": 0.16, "learning_rate": 1.9070577539260657e-05, "loss": 0.9377, "step": 3232 }, { "epoch": 0.16, "learning_rate": 1.906988468614277e-05, "loss": 1.0093, "step": 3233 }, { "epoch": 0.16, "learning_rate": 1.90691915874674e-05, "loss": 0.9041, "step": 3234 }, { "epoch": 0.16, "learning_rate": 1.906849824325332e-05, "loss": 0.9766, "step": 3235 }, { "epoch": 0.16, "learning_rate": 1.9067804653519297e-05, "loss": 1.0854, "step": 3236 }, { "epoch": 0.16, "learning_rate": 1.906711081828411e-05, "loss": 1.0301, "step": 3237 }, { "epoch": 0.16, "learning_rate": 1.9066416737566547e-05, "loss": 0.9926, "step": 3238 }, { "epoch": 0.16, "learning_rate": 1.9065722411385393e-05, "loss": 1.0, "step": 3239 }, { "epoch": 0.16, "learning_rate": 1.906502783975945e-05, "loss": 0.9963, "step": 3240 }, { "epoch": 0.16, "learning_rate": 1.9064333022707524e-05, "loss": 1.0429, "step": 3241 }, { "epoch": 0.16, "learning_rate": 1.9063637960248424e-05, "loss": 0.92, "step": 3242 }, { "epoch": 0.16, "learning_rate": 1.906294265240097e-05, "loss": 1.0251, "step": 3243 }, { "epoch": 0.16, "learning_rate": 1.9062247099183987e-05, "loss": 1.0046, "step": 3244 }, { "epoch": 0.16, "learning_rate": 1.9061551300616306e-05, "loss": 0.9931, "step": 3245 }, { "epoch": 0.16, "learning_rate": 1.9060855256716763e-05, "loss": 0.9974, "step": 3246 }, { "epoch": 0.16, "learning_rate": 1.9060158967504203e-05, "loss": 1.0284, "step": 3247 }, { "epoch": 0.17, "learning_rate": 1.9059462432997478e-05, "loss": 0.9595, "step": 3248 }, { "epoch": 0.17, "learning_rate": 1.9058765653215446e-05, "loss": 1.0643, "step": 3249 }, { "epoch": 0.17, "learning_rate": 1.9058068628176973e-05, "loss": 1.03, "step": 3250 }, { "epoch": 0.17, "learning_rate": 1.905737135790093e-05, "loss": 1.0369, "step": 3251 }, { "epoch": 0.17, "learning_rate": 1.9056673842406192e-05, "loss": 1.0083, "step": 3252 }, { "epoch": 0.17, "learning_rate": 1.9055976081711645e-05, "loss": 0.9626, "step": 3253 }, { "epoch": 0.17, "learning_rate": 1.905527807583618e-05, "loss": 0.9638, "step": 3254 }, { "epoch": 0.17, "learning_rate": 1.9054579824798696e-05, "loss": 1.0234, "step": 3255 }, { "epoch": 0.17, "learning_rate": 1.9053881328618096e-05, "loss": 1.0585, "step": 3256 }, { "epoch": 0.17, "learning_rate": 1.9053182587313293e-05, "loss": 1.1063, "step": 3257 }, { "epoch": 0.17, "learning_rate": 1.9052483600903202e-05, "loss": 1.0562, "step": 3258 }, { "epoch": 0.17, "learning_rate": 1.905178436940675e-05, "loss": 0.9448, "step": 3259 }, { "epoch": 0.17, "learning_rate": 1.9051084892842865e-05, "loss": 1.0544, "step": 3260 }, { "epoch": 0.17, "learning_rate": 1.9050385171230487e-05, "loss": 1.0446, "step": 3261 }, { "epoch": 0.17, "learning_rate": 1.904968520458856e-05, "loss": 1.0111, "step": 3262 }, { "epoch": 0.17, "learning_rate": 1.9048984992936032e-05, "loss": 1.0222, "step": 3263 }, { "epoch": 0.17, "learning_rate": 1.9048284536291865e-05, "loss": 1.037, "step": 3264 }, { "epoch": 0.17, "learning_rate": 1.9047583834675022e-05, "loss": 1.0728, "step": 3265 }, { "epoch": 0.17, "learning_rate": 1.9046882888104467e-05, "loss": 1.134, "step": 3266 }, { "epoch": 0.17, "learning_rate": 1.904618169659919e-05, "loss": 1.0476, "step": 3267 }, { "epoch": 0.17, "learning_rate": 1.904548026017816e-05, "loss": 0.9221, "step": 3268 }, { "epoch": 0.17, "learning_rate": 1.9044778578860382e-05, "loss": 1.0045, "step": 3269 }, { "epoch": 0.17, "learning_rate": 1.9044076652664843e-05, "loss": 0.9928, "step": 3270 }, { "epoch": 0.17, "learning_rate": 1.9043374481610554e-05, "loss": 1.0194, "step": 3271 }, { "epoch": 0.17, "learning_rate": 1.904267206571652e-05, "loss": 0.9967, "step": 3272 }, { "epoch": 0.17, "learning_rate": 1.904196940500176e-05, "loss": 1.0466, "step": 3273 }, { "epoch": 0.17, "learning_rate": 1.90412664994853e-05, "loss": 1.0156, "step": 3274 }, { "epoch": 0.17, "learning_rate": 1.904056334918617e-05, "loss": 0.9058, "step": 3275 }, { "epoch": 0.17, "learning_rate": 1.9039859954123402e-05, "loss": 1.0389, "step": 3276 }, { "epoch": 0.17, "learning_rate": 1.9039156314316046e-05, "loss": 0.9716, "step": 3277 }, { "epoch": 0.17, "learning_rate": 1.9038452429783148e-05, "loss": 0.8522, "step": 3278 }, { "epoch": 0.17, "learning_rate": 1.9037748300543767e-05, "loss": 1.0329, "step": 3279 }, { "epoch": 0.17, "learning_rate": 1.903704392661697e-05, "loss": 0.991, "step": 3280 }, { "epoch": 0.17, "learning_rate": 1.903633930802182e-05, "loss": 1.0197, "step": 3281 }, { "epoch": 0.17, "learning_rate": 1.9035634444777394e-05, "loss": 1.0023, "step": 3282 }, { "epoch": 0.17, "learning_rate": 1.9034929336902783e-05, "loss": 1.0694, "step": 3283 }, { "epoch": 0.17, "learning_rate": 1.9034223984417074e-05, "loss": 0.9799, "step": 3284 }, { "epoch": 0.17, "learning_rate": 1.9033518387339357e-05, "loss": 0.9364, "step": 3285 }, { "epoch": 0.17, "learning_rate": 1.9032812545688743e-05, "loss": 0.9729, "step": 3286 }, { "epoch": 0.17, "learning_rate": 1.9032106459484342e-05, "loss": 0.9992, "step": 3287 }, { "epoch": 0.17, "learning_rate": 1.9031400128745263e-05, "loss": 0.9992, "step": 3288 }, { "epoch": 0.17, "learning_rate": 1.903069355349064e-05, "loss": 1.1133, "step": 3289 }, { "epoch": 0.17, "learning_rate": 1.902998673373959e-05, "loss": 1.0869, "step": 3290 }, { "epoch": 0.17, "learning_rate": 1.902927966951126e-05, "loss": 1.0959, "step": 3291 }, { "epoch": 0.17, "learning_rate": 1.902857236082479e-05, "loss": 1.044, "step": 3292 }, { "epoch": 0.17, "learning_rate": 1.902786480769933e-05, "loss": 1.1391, "step": 3293 }, { "epoch": 0.17, "learning_rate": 1.9027157010154033e-05, "loss": 1.0684, "step": 3294 }, { "epoch": 0.17, "learning_rate": 1.9026448968208067e-05, "loss": 0.8937, "step": 3295 }, { "epoch": 0.17, "learning_rate": 1.9025740681880594e-05, "loss": 1.1016, "step": 3296 }, { "epoch": 0.17, "learning_rate": 1.90250321511908e-05, "loss": 0.9991, "step": 3297 }, { "epoch": 0.17, "learning_rate": 1.902432337615786e-05, "loss": 0.9636, "step": 3298 }, { "epoch": 0.17, "learning_rate": 1.902361435680096e-05, "loss": 0.9578, "step": 3299 }, { "epoch": 0.17, "learning_rate": 1.902290509313931e-05, "loss": 1.0634, "step": 3300 }, { "epoch": 0.17, "learning_rate": 1.90221955851921e-05, "loss": 0.9915, "step": 3301 }, { "epoch": 0.17, "learning_rate": 1.9021485832978546e-05, "loss": 0.9644, "step": 3302 }, { "epoch": 0.17, "learning_rate": 1.902077583651786e-05, "loss": 0.9215, "step": 3303 }, { "epoch": 0.17, "learning_rate": 1.9020065595829266e-05, "loss": 1.0779, "step": 3304 }, { "epoch": 0.17, "learning_rate": 1.9019355110931988e-05, "loss": 0.9744, "step": 3305 }, { "epoch": 0.17, "learning_rate": 1.901864438184527e-05, "loss": 0.9822, "step": 3306 }, { "epoch": 0.17, "learning_rate": 1.901793340858835e-05, "loss": 1.0731, "step": 3307 }, { "epoch": 0.17, "learning_rate": 1.9017222191180478e-05, "loss": 0.9151, "step": 3308 }, { "epoch": 0.17, "learning_rate": 1.901651072964091e-05, "loss": 0.9697, "step": 3309 }, { "epoch": 0.17, "learning_rate": 1.9015799023988904e-05, "loss": 0.947, "step": 3310 }, { "epoch": 0.17, "learning_rate": 1.9015087074243733e-05, "loss": 1.1337, "step": 3311 }, { "epoch": 0.17, "learning_rate": 1.901437488042467e-05, "loss": 0.9673, "step": 3312 }, { "epoch": 0.17, "learning_rate": 1.9013662442550998e-05, "loss": 1.0074, "step": 3313 }, { "epoch": 0.17, "learning_rate": 1.9012949760642007e-05, "loss": 0.9961, "step": 3314 }, { "epoch": 0.17, "learning_rate": 1.901223683471699e-05, "loss": 1.0015, "step": 3315 }, { "epoch": 0.17, "learning_rate": 1.9011523664795244e-05, "loss": 1.0398, "step": 3316 }, { "epoch": 0.17, "learning_rate": 1.9010810250896087e-05, "loss": 1.1054, "step": 3317 }, { "epoch": 0.17, "learning_rate": 1.901009659303883e-05, "loss": 0.998, "step": 3318 }, { "epoch": 0.17, "learning_rate": 1.900938269124279e-05, "loss": 0.985, "step": 3319 }, { "epoch": 0.17, "learning_rate": 1.9008668545527303e-05, "loss": 0.9088, "step": 3320 }, { "epoch": 0.17, "learning_rate": 1.9007954155911697e-05, "loss": 0.8756, "step": 3321 }, { "epoch": 0.17, "learning_rate": 1.9007239522415312e-05, "loss": 0.991, "step": 3322 }, { "epoch": 0.17, "learning_rate": 1.9006524645057505e-05, "loss": 1.0442, "step": 3323 }, { "epoch": 0.17, "learning_rate": 1.9005809523857624e-05, "loss": 1.034, "step": 3324 }, { "epoch": 0.17, "learning_rate": 1.9005094158835034e-05, "loss": 1.0542, "step": 3325 }, { "epoch": 0.17, "learning_rate": 1.9004378550009098e-05, "loss": 1.0016, "step": 3326 }, { "epoch": 0.17, "learning_rate": 1.9003662697399192e-05, "loss": 1.044, "step": 3327 }, { "epoch": 0.17, "learning_rate": 1.90029466010247e-05, "loss": 0.9611, "step": 3328 }, { "epoch": 0.17, "learning_rate": 1.9002230260905007e-05, "loss": 1.0387, "step": 3329 }, { "epoch": 0.17, "learning_rate": 1.9001513677059503e-05, "loss": 0.9368, "step": 3330 }, { "epoch": 0.17, "learning_rate": 1.9000796849507597e-05, "loss": 0.9615, "step": 3331 }, { "epoch": 0.17, "learning_rate": 1.9000079778268692e-05, "loss": 1.0705, "step": 3332 }, { "epoch": 0.17, "learning_rate": 1.89993624633622e-05, "loss": 0.9871, "step": 3333 }, { "epoch": 0.17, "learning_rate": 1.899864490480755e-05, "loss": 0.9825, "step": 3334 }, { "epoch": 0.17, "learning_rate": 1.8997927102624158e-05, "loss": 0.982, "step": 3335 }, { "epoch": 0.17, "learning_rate": 1.8997209056831462e-05, "loss": 0.9945, "step": 3336 }, { "epoch": 0.17, "learning_rate": 1.8996490767448902e-05, "loss": 0.9734, "step": 3337 }, { "epoch": 0.17, "learning_rate": 1.899577223449593e-05, "loss": 1.0572, "step": 3338 }, { "epoch": 0.17, "learning_rate": 1.8995053457991994e-05, "loss": 0.9843, "step": 3339 }, { "epoch": 0.17, "learning_rate": 1.8994334437956558e-05, "loss": 0.9954, "step": 3340 }, { "epoch": 0.17, "learning_rate": 1.8993615174409083e-05, "loss": 1.0055, "step": 3341 }, { "epoch": 0.17, "learning_rate": 1.899289566736905e-05, "loss": 1.0224, "step": 3342 }, { "epoch": 0.17, "learning_rate": 1.899217591685593e-05, "loss": 0.9838, "step": 3343 }, { "epoch": 0.17, "learning_rate": 1.899145592288921e-05, "loss": 1.0276, "step": 3344 }, { "epoch": 0.17, "learning_rate": 1.8990735685488392e-05, "loss": 1.0292, "step": 3345 }, { "epoch": 0.17, "learning_rate": 1.8990015204672973e-05, "loss": 1.0032, "step": 3346 }, { "epoch": 0.17, "learning_rate": 1.8989294480462456e-05, "loss": 0.9686, "step": 3347 }, { "epoch": 0.17, "learning_rate": 1.898857351287635e-05, "loss": 0.9615, "step": 3348 }, { "epoch": 0.17, "learning_rate": 1.8987852301934185e-05, "loss": 1.0349, "step": 3349 }, { "epoch": 0.17, "learning_rate": 1.8987130847655478e-05, "loss": 1.0844, "step": 3350 }, { "epoch": 0.17, "learning_rate": 1.8986409150059767e-05, "loss": 0.9227, "step": 3351 }, { "epoch": 0.17, "learning_rate": 1.8985687209166585e-05, "loss": 0.8954, "step": 3352 }, { "epoch": 0.17, "learning_rate": 1.8984965024995484e-05, "loss": 1.0839, "step": 3353 }, { "epoch": 0.17, "learning_rate": 1.898424259756602e-05, "loss": 1.0687, "step": 3354 }, { "epoch": 0.17, "learning_rate": 1.898351992689774e-05, "loss": 1.029, "step": 3355 }, { "epoch": 0.17, "learning_rate": 1.8982797013010215e-05, "loss": 1.0122, "step": 3356 }, { "epoch": 0.17, "learning_rate": 1.898207385592302e-05, "loss": 0.9814, "step": 3357 }, { "epoch": 0.17, "learning_rate": 1.898135045565573e-05, "loss": 1.0524, "step": 3358 }, { "epoch": 0.17, "learning_rate": 1.8980626812227933e-05, "loss": 1.041, "step": 3359 }, { "epoch": 0.17, "learning_rate": 1.897990292565922e-05, "loss": 1.1133, "step": 3360 }, { "epoch": 0.17, "learning_rate": 1.897917879596919e-05, "loss": 1.0599, "step": 3361 }, { "epoch": 0.17, "learning_rate": 1.8978454423177446e-05, "loss": 0.9031, "step": 3362 }, { "epoch": 0.17, "learning_rate": 1.89777298073036e-05, "loss": 0.9692, "step": 3363 }, { "epoch": 0.17, "learning_rate": 1.8977004948367272e-05, "loss": 1.0127, "step": 3364 }, { "epoch": 0.17, "learning_rate": 1.8976279846388087e-05, "loss": 1.0151, "step": 3365 }, { "epoch": 0.17, "learning_rate": 1.8975554501385672e-05, "loss": 1.1141, "step": 3366 }, { "epoch": 0.17, "learning_rate": 1.897482891337967e-05, "loss": 1.0785, "step": 3367 }, { "epoch": 0.17, "learning_rate": 1.8974103082389726e-05, "loss": 1.0211, "step": 3368 }, { "epoch": 0.17, "learning_rate": 1.8973377008435487e-05, "loss": 1.0415, "step": 3369 }, { "epoch": 0.17, "learning_rate": 1.8972650691536606e-05, "loss": 1.0255, "step": 3370 }, { "epoch": 0.17, "learning_rate": 1.8971924131712762e-05, "loss": 0.9644, "step": 3371 }, { "epoch": 0.17, "learning_rate": 1.8971197328983616e-05, "loss": 0.9617, "step": 3372 }, { "epoch": 0.17, "learning_rate": 1.8970470283368844e-05, "loss": 0.9424, "step": 3373 }, { "epoch": 0.17, "learning_rate": 1.8969742994888136e-05, "loss": 1.0043, "step": 3374 }, { "epoch": 0.17, "learning_rate": 1.8969015463561178e-05, "loss": 1.0259, "step": 3375 }, { "epoch": 0.17, "learning_rate": 1.896828768940767e-05, "loss": 1.0035, "step": 3376 }, { "epoch": 0.17, "learning_rate": 1.896755967244731e-05, "loss": 0.9786, "step": 3377 }, { "epoch": 0.17, "learning_rate": 1.8966831412699816e-05, "loss": 1.1263, "step": 3378 }, { "epoch": 0.17, "learning_rate": 1.89661029101849e-05, "loss": 1.0634, "step": 3379 }, { "epoch": 0.17, "learning_rate": 1.896537416492229e-05, "loss": 0.9829, "step": 3380 }, { "epoch": 0.17, "learning_rate": 1.896464517693171e-05, "loss": 0.9961, "step": 3381 }, { "epoch": 0.17, "learning_rate": 1.8963915946232903e-05, "loss": 1.0881, "step": 3382 }, { "epoch": 0.17, "learning_rate": 1.8963186472845606e-05, "loss": 0.9639, "step": 3383 }, { "epoch": 0.17, "learning_rate": 1.896245675678957e-05, "loss": 0.9919, "step": 3384 }, { "epoch": 0.17, "learning_rate": 1.8961726798084553e-05, "loss": 1.0162, "step": 3385 }, { "epoch": 0.17, "learning_rate": 1.896099659675032e-05, "loss": 1.02, "step": 3386 }, { "epoch": 0.17, "learning_rate": 1.8960266152806633e-05, "loss": 0.9775, "step": 3387 }, { "epoch": 0.17, "learning_rate": 1.895953546627328e-05, "loss": 0.9277, "step": 3388 }, { "epoch": 0.17, "learning_rate": 1.8958804537170033e-05, "loss": 1.097, "step": 3389 }, { "epoch": 0.17, "learning_rate": 1.8958073365516682e-05, "loss": 1.0201, "step": 3390 }, { "epoch": 0.17, "learning_rate": 1.8957341951333028e-05, "loss": 1.0692, "step": 3391 }, { "epoch": 0.17, "learning_rate": 1.895661029463887e-05, "loss": 1.0139, "step": 3392 }, { "epoch": 0.17, "learning_rate": 1.8955878395454017e-05, "loss": 0.9795, "step": 3393 }, { "epoch": 0.17, "learning_rate": 1.8955146253798284e-05, "loss": 1.0289, "step": 3394 }, { "epoch": 0.17, "learning_rate": 1.8954413869691497e-05, "loss": 1.014, "step": 3395 }, { "epoch": 0.17, "learning_rate": 1.8953681243153478e-05, "loss": 1.1007, "step": 3396 }, { "epoch": 0.17, "learning_rate": 1.8952948374204066e-05, "loss": 1.0307, "step": 3397 }, { "epoch": 0.17, "learning_rate": 1.89522152628631e-05, "loss": 0.9722, "step": 3398 }, { "epoch": 0.17, "learning_rate": 1.8951481909150435e-05, "loss": 0.9672, "step": 3399 }, { "epoch": 0.17, "learning_rate": 1.895074831308592e-05, "loss": 1.0912, "step": 3400 }, { "epoch": 0.17, "learning_rate": 1.895001447468941e-05, "loss": 1.0055, "step": 3401 }, { "epoch": 0.17, "learning_rate": 1.8949280393980787e-05, "loss": 0.9588, "step": 3402 }, { "epoch": 0.17, "learning_rate": 1.894854607097992e-05, "loss": 0.9109, "step": 3403 }, { "epoch": 0.17, "learning_rate": 1.8947811505706684e-05, "loss": 0.9598, "step": 3404 }, { "epoch": 0.17, "learning_rate": 1.8947076698180976e-05, "loss": 0.9763, "step": 3405 }, { "epoch": 0.17, "learning_rate": 1.8946341648422685e-05, "loss": 0.9453, "step": 3406 }, { "epoch": 0.17, "learning_rate": 1.894560635645171e-05, "loss": 0.9574, "step": 3407 }, { "epoch": 0.17, "learning_rate": 1.8944870822287957e-05, "loss": 1.0125, "step": 3408 }, { "epoch": 0.17, "learning_rate": 1.8944135045951347e-05, "loss": 1.0095, "step": 3409 }, { "epoch": 0.17, "learning_rate": 1.8943399027461794e-05, "loss": 0.9449, "step": 3410 }, { "epoch": 0.17, "learning_rate": 1.894266276683923e-05, "loss": 0.9854, "step": 3411 }, { "epoch": 0.17, "learning_rate": 1.8941926264103582e-05, "loss": 0.9604, "step": 3412 }, { "epoch": 0.17, "learning_rate": 1.89411895192748e-05, "loss": 1.0325, "step": 3413 }, { "epoch": 0.17, "learning_rate": 1.894045253237282e-05, "loss": 0.9946, "step": 3414 }, { "epoch": 0.17, "learning_rate": 1.8939715303417602e-05, "loss": 1.0952, "step": 3415 }, { "epoch": 0.17, "learning_rate": 1.8938977832429103e-05, "loss": 0.9922, "step": 3416 }, { "epoch": 0.17, "learning_rate": 1.8938240119427286e-05, "loss": 0.9728, "step": 3417 }, { "epoch": 0.17, "learning_rate": 1.893750216443213e-05, "loss": 1.0161, "step": 3418 }, { "epoch": 0.17, "learning_rate": 1.893676396746361e-05, "loss": 1.0594, "step": 3419 }, { "epoch": 0.17, "learning_rate": 1.8936025528541716e-05, "loss": 1.0801, "step": 3420 }, { "epoch": 0.17, "learning_rate": 1.8935286847686436e-05, "loss": 0.9978, "step": 3421 }, { "epoch": 0.17, "learning_rate": 1.893454792491777e-05, "loss": 1.0344, "step": 3422 }, { "epoch": 0.17, "learning_rate": 1.8933808760255726e-05, "loss": 1.0585, "step": 3423 }, { "epoch": 0.17, "learning_rate": 1.8933069353720317e-05, "loss": 0.9846, "step": 3424 }, { "epoch": 0.17, "learning_rate": 1.893232970533156e-05, "loss": 1.0397, "step": 3425 }, { "epoch": 0.17, "learning_rate": 1.8931589815109475e-05, "loss": 0.9315, "step": 3426 }, { "epoch": 0.17, "learning_rate": 1.89308496830741e-05, "loss": 1.0015, "step": 3427 }, { "epoch": 0.17, "learning_rate": 1.893010930924547e-05, "loss": 0.9508, "step": 3428 }, { "epoch": 0.17, "learning_rate": 1.8929368693643633e-05, "loss": 0.9633, "step": 3429 }, { "epoch": 0.17, "learning_rate": 1.8928627836288637e-05, "loss": 0.9171, "step": 3430 }, { "epoch": 0.17, "learning_rate": 1.8927886737200544e-05, "loss": 0.9923, "step": 3431 }, { "epoch": 0.17, "learning_rate": 1.8927145396399415e-05, "loss": 0.8803, "step": 3432 }, { "epoch": 0.17, "learning_rate": 1.8926403813905323e-05, "loss": 0.9902, "step": 3433 }, { "epoch": 0.17, "learning_rate": 1.8925661989738345e-05, "loss": 1.0264, "step": 3434 }, { "epoch": 0.17, "learning_rate": 1.8924919923918566e-05, "loss": 0.9432, "step": 3435 }, { "epoch": 0.17, "learning_rate": 1.892417761646607e-05, "loss": 1.0994, "step": 3436 }, { "epoch": 0.17, "learning_rate": 1.8923435067400964e-05, "loss": 0.8909, "step": 3437 }, { "epoch": 0.17, "learning_rate": 1.8922692276743346e-05, "loss": 1.0135, "step": 3438 }, { "epoch": 0.17, "learning_rate": 1.892194924451333e-05, "loss": 0.9979, "step": 3439 }, { "epoch": 0.17, "learning_rate": 1.8921205970731028e-05, "loss": 1.0891, "step": 3440 }, { "epoch": 0.17, "learning_rate": 1.8920462455416567e-05, "loss": 1.0104, "step": 3441 }, { "epoch": 0.17, "learning_rate": 1.891971869859007e-05, "loss": 1.0179, "step": 3442 }, { "epoch": 0.17, "learning_rate": 1.8918974700271687e-05, "loss": 0.938, "step": 3443 }, { "epoch": 0.17, "learning_rate": 1.8918230460481547e-05, "loss": 1.0132, "step": 3444 }, { "epoch": 0.18, "learning_rate": 1.891748597923981e-05, "loss": 0.8871, "step": 3445 }, { "epoch": 0.18, "learning_rate": 1.8916741256566625e-05, "loss": 1.0204, "step": 3446 }, { "epoch": 0.18, "learning_rate": 1.891599629248216e-05, "loss": 0.9693, "step": 3447 }, { "epoch": 0.18, "learning_rate": 1.8915251087006578e-05, "loss": 1.0231, "step": 3448 }, { "epoch": 0.18, "learning_rate": 1.8914505640160058e-05, "loss": 1.0385, "step": 3449 }, { "epoch": 0.18, "learning_rate": 1.8913759951962785e-05, "loss": 1.0449, "step": 3450 }, { "epoch": 0.18, "learning_rate": 1.8913014022434943e-05, "loss": 1.0747, "step": 3451 }, { "epoch": 0.18, "learning_rate": 1.8912267851596732e-05, "loss": 1.0767, "step": 3452 }, { "epoch": 0.18, "learning_rate": 1.8911521439468344e-05, "loss": 0.9629, "step": 3453 }, { "epoch": 0.18, "learning_rate": 1.891077478607e-05, "loss": 1.0808, "step": 3454 }, { "epoch": 0.18, "learning_rate": 1.8910027891421906e-05, "loss": 1.012, "step": 3455 }, { "epoch": 0.18, "learning_rate": 1.890928075554429e-05, "loss": 0.9781, "step": 3456 }, { "epoch": 0.18, "learning_rate": 1.890853337845737e-05, "loss": 0.9957, "step": 3457 }, { "epoch": 0.18, "learning_rate": 1.8907785760181392e-05, "loss": 0.9304, "step": 3458 }, { "epoch": 0.18, "learning_rate": 1.890703790073659e-05, "loss": 1.0419, "step": 3459 }, { "epoch": 0.18, "learning_rate": 1.8906289800143215e-05, "loss": 0.8863, "step": 3460 }, { "epoch": 0.18, "learning_rate": 1.8905541458421518e-05, "loss": 1.0642, "step": 3461 }, { "epoch": 0.18, "learning_rate": 1.8904792875591758e-05, "loss": 0.9856, "step": 3462 }, { "epoch": 0.18, "learning_rate": 1.8904044051674208e-05, "loss": 0.9995, "step": 3463 }, { "epoch": 0.18, "learning_rate": 1.8903294986689137e-05, "loss": 0.9785, "step": 3464 }, { "epoch": 0.18, "learning_rate": 1.890254568065683e-05, "loss": 0.9099, "step": 3465 }, { "epoch": 0.18, "learning_rate": 1.8901796133597562e-05, "loss": 0.9953, "step": 3466 }, { "epoch": 0.18, "learning_rate": 1.890104634553164e-05, "loss": 1.0658, "step": 3467 }, { "epoch": 0.18, "learning_rate": 1.890029631647936e-05, "loss": 1.0177, "step": 3468 }, { "epoch": 0.18, "learning_rate": 1.8899546046461024e-05, "loss": 0.9955, "step": 3469 }, { "epoch": 0.18, "learning_rate": 1.889879553549695e-05, "loss": 1.0335, "step": 3470 }, { "epoch": 0.18, "learning_rate": 1.889804478360745e-05, "loss": 0.9286, "step": 3471 }, { "epoch": 0.18, "learning_rate": 1.8897293790812857e-05, "loss": 0.9886, "step": 3472 }, { "epoch": 0.18, "learning_rate": 1.8896542557133498e-05, "loss": 1.0713, "step": 3473 }, { "epoch": 0.18, "learning_rate": 1.8895791082589715e-05, "loss": 1.0022, "step": 3474 }, { "epoch": 0.18, "learning_rate": 1.8895039367201855e-05, "loss": 0.9264, "step": 3475 }, { "epoch": 0.18, "learning_rate": 1.889428741099027e-05, "loss": 1.054, "step": 3476 }, { "epoch": 0.18, "learning_rate": 1.8893535213975312e-05, "loss": 1.0195, "step": 3477 }, { "epoch": 0.18, "learning_rate": 1.889278277617735e-05, "loss": 0.9699, "step": 3478 }, { "epoch": 0.18, "learning_rate": 1.889203009761676e-05, "loss": 0.9312, "step": 3479 }, { "epoch": 0.18, "learning_rate": 1.889127717831391e-05, "loss": 1.0842, "step": 3480 }, { "epoch": 0.18, "learning_rate": 1.889052401828919e-05, "loss": 0.9695, "step": 3481 }, { "epoch": 0.18, "learning_rate": 1.8889770617562995e-05, "loss": 1.0514, "step": 3482 }, { "epoch": 0.18, "learning_rate": 1.8889016976155716e-05, "loss": 0.9741, "step": 3483 }, { "epoch": 0.18, "learning_rate": 1.888826309408776e-05, "loss": 1.0768, "step": 3484 }, { "epoch": 0.18, "learning_rate": 1.888750897137954e-05, "loss": 0.9681, "step": 3485 }, { "epoch": 0.18, "learning_rate": 1.8886754608051462e-05, "loss": 0.9609, "step": 3486 }, { "epoch": 0.18, "learning_rate": 1.8886000004123965e-05, "loss": 0.922, "step": 3487 }, { "epoch": 0.18, "learning_rate": 1.888524515961747e-05, "loss": 0.9835, "step": 3488 }, { "epoch": 0.18, "learning_rate": 1.8884490074552416e-05, "loss": 1.0029, "step": 3489 }, { "epoch": 0.18, "learning_rate": 1.8883734748949246e-05, "loss": 1.0211, "step": 3490 }, { "epoch": 0.18, "learning_rate": 1.8882979182828407e-05, "loss": 0.9264, "step": 3491 }, { "epoch": 0.18, "learning_rate": 1.8882223376210358e-05, "loss": 0.9091, "step": 3492 }, { "epoch": 0.18, "learning_rate": 1.888146732911556e-05, "loss": 1.0338, "step": 3493 }, { "epoch": 0.18, "learning_rate": 1.8880711041564487e-05, "loss": 0.9125, "step": 3494 }, { "epoch": 0.18, "learning_rate": 1.8879954513577607e-05, "loss": 1.0369, "step": 3495 }, { "epoch": 0.18, "learning_rate": 1.8879197745175408e-05, "loss": 1.0405, "step": 3496 }, { "epoch": 0.18, "learning_rate": 1.8878440736378377e-05, "loss": 1.0018, "step": 3497 }, { "epoch": 0.18, "learning_rate": 1.887768348720701e-05, "loss": 1.0403, "step": 3498 }, { "epoch": 0.18, "learning_rate": 1.8876925997681806e-05, "loss": 1.0414, "step": 3499 }, { "epoch": 0.18, "learning_rate": 1.8876168267823273e-05, "loss": 0.9517, "step": 3500 }, { "epoch": 0.18, "learning_rate": 1.887541029765193e-05, "loss": 1.0796, "step": 3501 }, { "epoch": 0.18, "learning_rate": 1.8874652087188296e-05, "loss": 0.9993, "step": 3502 }, { "epoch": 0.18, "learning_rate": 1.88738936364529e-05, "loss": 1.1069, "step": 3503 }, { "epoch": 0.18, "learning_rate": 1.887313494546627e-05, "loss": 0.9888, "step": 3504 }, { "epoch": 0.18, "learning_rate": 1.8872376014248958e-05, "loss": 1.0729, "step": 3505 }, { "epoch": 0.18, "learning_rate": 1.88716168428215e-05, "loss": 1.042, "step": 3506 }, { "epoch": 0.18, "learning_rate": 1.8870857431204458e-05, "loss": 1.1208, "step": 3507 }, { "epoch": 0.18, "learning_rate": 1.8870097779418386e-05, "loss": 1.0126, "step": 3508 }, { "epoch": 0.18, "learning_rate": 1.8869337887483853e-05, "loss": 1.0526, "step": 3509 }, { "epoch": 0.18, "learning_rate": 1.8868577755421436e-05, "loss": 1.1138, "step": 3510 }, { "epoch": 0.18, "learning_rate": 1.886781738325171e-05, "loss": 0.9835, "step": 3511 }, { "epoch": 0.18, "learning_rate": 1.8867056770995264e-05, "loss": 1.043, "step": 3512 }, { "epoch": 0.18, "learning_rate": 1.886629591867269e-05, "loss": 1.0133, "step": 3513 }, { "epoch": 0.18, "learning_rate": 1.8865534826304585e-05, "loss": 1.0241, "step": 3514 }, { "epoch": 0.18, "learning_rate": 1.8864773493911558e-05, "loss": 0.954, "step": 3515 }, { "epoch": 0.18, "learning_rate": 1.886401192151422e-05, "loss": 1.0769, "step": 3516 }, { "epoch": 0.18, "learning_rate": 1.8863250109133188e-05, "loss": 0.9646, "step": 3517 }, { "epoch": 0.18, "learning_rate": 1.8862488056789092e-05, "loss": 0.9171, "step": 3518 }, { "epoch": 0.18, "learning_rate": 1.8861725764502557e-05, "loss": 0.9227, "step": 3519 }, { "epoch": 0.18, "learning_rate": 1.886096323229423e-05, "loss": 0.9618, "step": 3520 }, { "epoch": 0.18, "learning_rate": 1.8860200460184746e-05, "loss": 0.9463, "step": 3521 }, { "epoch": 0.18, "learning_rate": 1.8859437448194767e-05, "loss": 0.8728, "step": 3522 }, { "epoch": 0.18, "learning_rate": 1.8858674196344943e-05, "loss": 1.0004, "step": 3523 }, { "epoch": 0.18, "learning_rate": 1.8857910704655937e-05, "loss": 0.9531, "step": 3524 }, { "epoch": 0.18, "learning_rate": 1.885714697314842e-05, "loss": 1.0077, "step": 3525 }, { "epoch": 0.18, "learning_rate": 1.885638300184308e-05, "loss": 0.98, "step": 3526 }, { "epoch": 0.18, "learning_rate": 1.8855618790760587e-05, "loss": 1.0298, "step": 3527 }, { "epoch": 0.18, "learning_rate": 1.885485433992164e-05, "loss": 0.962, "step": 3528 }, { "epoch": 0.18, "learning_rate": 1.885408964934693e-05, "loss": 0.9868, "step": 3529 }, { "epoch": 0.18, "learning_rate": 1.8853324719057165e-05, "loss": 0.9838, "step": 3530 }, { "epoch": 0.18, "learning_rate": 1.885255954907305e-05, "loss": 1.0328, "step": 3531 }, { "epoch": 0.18, "learning_rate": 1.8851794139415306e-05, "loss": 0.9798, "step": 3532 }, { "epoch": 0.18, "learning_rate": 1.885102849010465e-05, "loss": 0.9628, "step": 3533 }, { "epoch": 0.18, "learning_rate": 1.8850262601161818e-05, "loss": 0.9044, "step": 3534 }, { "epoch": 0.18, "learning_rate": 1.884949647260754e-05, "loss": 1.059, "step": 3535 }, { "epoch": 0.18, "learning_rate": 1.8848730104462564e-05, "loss": 1.0336, "step": 3536 }, { "epoch": 0.18, "learning_rate": 1.8847963496747627e-05, "loss": 0.936, "step": 3537 }, { "epoch": 0.18, "learning_rate": 1.88471966494835e-05, "loss": 1.0609, "step": 3538 }, { "epoch": 0.18, "learning_rate": 1.8846429562690928e-05, "loss": 1.0234, "step": 3539 }, { "epoch": 0.18, "learning_rate": 1.8845662236390694e-05, "loss": 0.9955, "step": 3540 }, { "epoch": 0.18, "learning_rate": 1.8844894670603564e-05, "loss": 1.0507, "step": 3541 }, { "epoch": 0.18, "learning_rate": 1.8844126865350322e-05, "loss": 1.0062, "step": 3542 }, { "epoch": 0.18, "learning_rate": 1.8843358820651752e-05, "loss": 0.9485, "step": 3543 }, { "epoch": 0.18, "learning_rate": 1.8842590536528654e-05, "loss": 1.0689, "step": 3544 }, { "epoch": 0.18, "learning_rate": 1.8841822013001824e-05, "loss": 0.9881, "step": 3545 }, { "epoch": 0.18, "learning_rate": 1.8841053250092068e-05, "loss": 1.0937, "step": 3546 }, { "epoch": 0.18, "learning_rate": 1.8840284247820203e-05, "loss": 1.0715, "step": 3547 }, { "epoch": 0.18, "learning_rate": 1.8839515006207046e-05, "loss": 0.9973, "step": 3548 }, { "epoch": 0.18, "learning_rate": 1.8838745525273424e-05, "loss": 0.9792, "step": 3549 }, { "epoch": 0.18, "learning_rate": 1.8837975805040174e-05, "loss": 1.0706, "step": 3550 }, { "epoch": 0.18, "learning_rate": 1.883720584552813e-05, "loss": 0.9744, "step": 3551 }, { "epoch": 0.18, "learning_rate": 1.883643564675814e-05, "loss": 1.0667, "step": 3552 }, { "epoch": 0.18, "learning_rate": 1.8835665208751054e-05, "loss": 1.1048, "step": 3553 }, { "epoch": 0.18, "learning_rate": 1.8834894531527735e-05, "loss": 0.9892, "step": 3554 }, { "epoch": 0.18, "learning_rate": 1.8834123615109047e-05, "loss": 0.9774, "step": 3555 }, { "epoch": 0.18, "learning_rate": 1.883335245951586e-05, "loss": 1.0601, "step": 3556 }, { "epoch": 0.18, "learning_rate": 1.8832581064769053e-05, "loss": 0.8851, "step": 3557 }, { "epoch": 0.18, "learning_rate": 1.883180943088951e-05, "loss": 1.0119, "step": 3558 }, { "epoch": 0.18, "learning_rate": 1.8831037557898123e-05, "loss": 0.9319, "step": 3559 }, { "epoch": 0.18, "learning_rate": 1.883026544581579e-05, "loss": 0.9477, "step": 3560 }, { "epoch": 0.18, "learning_rate": 1.8829493094663417e-05, "loss": 0.9488, "step": 3561 }, { "epoch": 0.18, "learning_rate": 1.882872050446191e-05, "loss": 0.9644, "step": 3562 }, { "epoch": 0.18, "learning_rate": 1.8827947675232187e-05, "loss": 0.9325, "step": 3563 }, { "epoch": 0.18, "learning_rate": 1.8827174606995174e-05, "loss": 1.0977, "step": 3564 }, { "epoch": 0.18, "learning_rate": 1.88264012997718e-05, "loss": 1.0032, "step": 3565 }, { "epoch": 0.18, "learning_rate": 1.8825627753583006e-05, "loss": 0.9267, "step": 3566 }, { "epoch": 0.18, "learning_rate": 1.8824853968449724e-05, "loss": 0.9727, "step": 3567 }, { "epoch": 0.18, "learning_rate": 1.882407994439291e-05, "loss": 0.939, "step": 3568 }, { "epoch": 0.18, "learning_rate": 1.8823305681433524e-05, "loss": 1.0587, "step": 3569 }, { "epoch": 0.18, "learning_rate": 1.882253117959252e-05, "loss": 0.93, "step": 3570 }, { "epoch": 0.18, "learning_rate": 1.8821756438890873e-05, "loss": 0.9911, "step": 3571 }, { "epoch": 0.18, "learning_rate": 1.8820981459349555e-05, "loss": 0.9206, "step": 3572 }, { "epoch": 0.18, "learning_rate": 1.882020624098955e-05, "loss": 0.9629, "step": 3573 }, { "epoch": 0.18, "learning_rate": 1.881943078383184e-05, "loss": 1.0126, "step": 3574 }, { "epoch": 0.18, "learning_rate": 1.881865508789743e-05, "loss": 0.9907, "step": 3575 }, { "epoch": 0.18, "learning_rate": 1.8817879153207315e-05, "loss": 1.0359, "step": 3576 }, { "epoch": 0.18, "learning_rate": 1.88171029797825e-05, "loss": 1.0283, "step": 3577 }, { "epoch": 0.18, "learning_rate": 1.881632656764401e-05, "loss": 1.038, "step": 3578 }, { "epoch": 0.18, "learning_rate": 1.8815549916812853e-05, "loss": 0.9812, "step": 3579 }, { "epoch": 0.18, "learning_rate": 1.881477302731006e-05, "loss": 1.0062, "step": 3580 }, { "epoch": 0.18, "learning_rate": 1.8813995899156667e-05, "loss": 1.0362, "step": 3581 }, { "epoch": 0.18, "learning_rate": 1.8813218532373713e-05, "loss": 0.8748, "step": 3582 }, { "epoch": 0.18, "learning_rate": 1.881244092698224e-05, "loss": 0.9982, "step": 3583 }, { "epoch": 0.18, "learning_rate": 1.881166308300331e-05, "loss": 0.9131, "step": 3584 }, { "epoch": 0.18, "learning_rate": 1.8810885000457976e-05, "loss": 0.9647, "step": 3585 }, { "epoch": 0.18, "learning_rate": 1.88101066793673e-05, "loss": 1.0937, "step": 3586 }, { "epoch": 0.18, "learning_rate": 1.8809328119752362e-05, "loss": 1.0379, "step": 3587 }, { "epoch": 0.18, "learning_rate": 1.8808549321634238e-05, "loss": 1.0024, "step": 3588 }, { "epoch": 0.18, "learning_rate": 1.880777028503401e-05, "loss": 1.0642, "step": 3589 }, { "epoch": 0.18, "learning_rate": 1.8806991009972776e-05, "loss": 1.0234, "step": 3590 }, { "epoch": 0.18, "learning_rate": 1.8806211496471628e-05, "loss": 1.0186, "step": 3591 }, { "epoch": 0.18, "learning_rate": 1.8805431744551674e-05, "loss": 1.0258, "step": 3592 }, { "epoch": 0.18, "learning_rate": 1.8804651754234026e-05, "loss": 1.0335, "step": 3593 }, { "epoch": 0.18, "learning_rate": 1.8803871525539795e-05, "loss": 1.1185, "step": 3594 }, { "epoch": 0.18, "learning_rate": 1.880309105849011e-05, "loss": 1.0272, "step": 3595 }, { "epoch": 0.18, "learning_rate": 1.8802310353106103e-05, "loss": 1.0563, "step": 3596 }, { "epoch": 0.18, "learning_rate": 1.8801529409408905e-05, "loss": 1.0096, "step": 3597 }, { "epoch": 0.18, "learning_rate": 1.880074822741966e-05, "loss": 1.0363, "step": 3598 }, { "epoch": 0.18, "learning_rate": 1.8799966807159525e-05, "loss": 1.0443, "step": 3599 }, { "epoch": 0.18, "learning_rate": 1.879918514864965e-05, "loss": 1.0046, "step": 3600 }, { "epoch": 0.18, "learning_rate": 1.8798403251911198e-05, "loss": 0.9657, "step": 3601 }, { "epoch": 0.18, "learning_rate": 1.8797621116965338e-05, "loss": 0.9197, "step": 3602 }, { "epoch": 0.18, "learning_rate": 1.8796838743833245e-05, "loss": 1.0516, "step": 3603 }, { "epoch": 0.18, "learning_rate": 1.8796056132536107e-05, "loss": 0.9425, "step": 3604 }, { "epoch": 0.18, "learning_rate": 1.8795273283095103e-05, "loss": 1.0112, "step": 3605 }, { "epoch": 0.18, "learning_rate": 1.8794490195531434e-05, "loss": 1.0332, "step": 3606 }, { "epoch": 0.18, "learning_rate": 1.87937068698663e-05, "loss": 1.1415, "step": 3607 }, { "epoch": 0.18, "learning_rate": 1.8792923306120906e-05, "loss": 0.9705, "step": 3608 }, { "epoch": 0.18, "learning_rate": 1.8792139504316474e-05, "loss": 1.0899, "step": 3609 }, { "epoch": 0.18, "learning_rate": 1.8791355464474214e-05, "loss": 0.9861, "step": 3610 }, { "epoch": 0.18, "learning_rate": 1.8790571186615358e-05, "loss": 1.0597, "step": 3611 }, { "epoch": 0.18, "learning_rate": 1.878978667076114e-05, "loss": 0.9965, "step": 3612 }, { "epoch": 0.18, "learning_rate": 1.87890019169328e-05, "loss": 1.0588, "step": 3613 }, { "epoch": 0.18, "learning_rate": 1.8788216925151585e-05, "loss": 1.0007, "step": 3614 }, { "epoch": 0.18, "learning_rate": 1.8787431695438746e-05, "loss": 1.0076, "step": 3615 }, { "epoch": 0.18, "learning_rate": 1.8786646227815545e-05, "loss": 1.0692, "step": 3616 }, { "epoch": 0.18, "learning_rate": 1.878586052230324e-05, "loss": 0.986, "step": 3617 }, { "epoch": 0.18, "learning_rate": 1.8785074578923112e-05, "loss": 0.9643, "step": 3618 }, { "epoch": 0.18, "learning_rate": 1.878428839769644e-05, "loss": 1.0805, "step": 3619 }, { "epoch": 0.18, "learning_rate": 1.87835019786445e-05, "loss": 0.9074, "step": 3620 }, { "epoch": 0.18, "learning_rate": 1.8782715321788593e-05, "loss": 1.0385, "step": 3621 }, { "epoch": 0.18, "learning_rate": 1.8781928427150008e-05, "loss": 1.0293, "step": 3622 }, { "epoch": 0.18, "learning_rate": 1.8781141294750055e-05, "loss": 1.016, "step": 3623 }, { "epoch": 0.18, "learning_rate": 1.8780353924610047e-05, "loss": 1.1284, "step": 3624 }, { "epoch": 0.18, "learning_rate": 1.8779566316751294e-05, "loss": 0.9309, "step": 3625 }, { "epoch": 0.18, "learning_rate": 1.877877847119513e-05, "loss": 1.0504, "step": 3626 }, { "epoch": 0.18, "learning_rate": 1.8777990387962874e-05, "loss": 1.0152, "step": 3627 }, { "epoch": 0.18, "learning_rate": 1.8777202067075867e-05, "loss": 1.0206, "step": 3628 }, { "epoch": 0.18, "learning_rate": 1.8776413508555452e-05, "loss": 1.041, "step": 3629 }, { "epoch": 0.18, "learning_rate": 1.8775624712422982e-05, "loss": 0.9442, "step": 3630 }, { "epoch": 0.18, "learning_rate": 1.8774835678699807e-05, "loss": 0.9684, "step": 3631 }, { "epoch": 0.18, "learning_rate": 1.8774046407407292e-05, "loss": 0.9693, "step": 3632 }, { "epoch": 0.18, "learning_rate": 1.8773256898566806e-05, "loss": 0.9266, "step": 3633 }, { "epoch": 0.18, "learning_rate": 1.877246715219972e-05, "loss": 1.0755, "step": 3634 }, { "epoch": 0.18, "learning_rate": 1.8771677168327424e-05, "loss": 1.0103, "step": 3635 }, { "epoch": 0.18, "learning_rate": 1.87708869469713e-05, "loss": 1.0489, "step": 3636 }, { "epoch": 0.18, "learning_rate": 1.8770096488152742e-05, "loss": 1.0338, "step": 3637 }, { "epoch": 0.18, "learning_rate": 1.8769305791893153e-05, "loss": 1.0536, "step": 3638 }, { "epoch": 0.18, "learning_rate": 1.8768514858213938e-05, "loss": 1.065, "step": 3639 }, { "epoch": 0.18, "learning_rate": 1.8767723687136515e-05, "loss": 0.9435, "step": 3640 }, { "epoch": 0.18, "learning_rate": 1.87669322786823e-05, "loss": 0.9769, "step": 3641 }, { "epoch": 0.19, "learning_rate": 1.876614063287272e-05, "loss": 1.0192, "step": 3642 }, { "epoch": 0.19, "learning_rate": 1.876534874972921e-05, "loss": 1.0498, "step": 3643 }, { "epoch": 0.19, "learning_rate": 1.8764556629273207e-05, "loss": 1.036, "step": 3644 }, { "epoch": 0.19, "learning_rate": 1.876376427152616e-05, "loss": 1.0181, "step": 3645 }, { "epoch": 0.19, "learning_rate": 1.8762971676509515e-05, "loss": 1.0664, "step": 3646 }, { "epoch": 0.19, "learning_rate": 1.876217884424474e-05, "loss": 1.0118, "step": 3647 }, { "epoch": 0.19, "learning_rate": 1.876138577475329e-05, "loss": 1.0896, "step": 3648 }, { "epoch": 0.19, "learning_rate": 1.8760592468056647e-05, "loss": 1.0546, "step": 3649 }, { "epoch": 0.19, "learning_rate": 1.875979892417628e-05, "loss": 0.996, "step": 3650 }, { "epoch": 0.19, "learning_rate": 1.875900514313368e-05, "loss": 1.042, "step": 3651 }, { "epoch": 0.19, "learning_rate": 1.875821112495033e-05, "loss": 0.9886, "step": 3652 }, { "epoch": 0.19, "learning_rate": 1.8757416869647732e-05, "loss": 0.9958, "step": 3653 }, { "epoch": 0.19, "learning_rate": 1.8756622377247393e-05, "loss": 0.9822, "step": 3654 }, { "epoch": 0.19, "learning_rate": 1.8755827647770816e-05, "loss": 0.9096, "step": 3655 }, { "epoch": 0.19, "learning_rate": 1.8755032681239523e-05, "loss": 0.9938, "step": 3656 }, { "epoch": 0.19, "learning_rate": 1.8754237477675034e-05, "loss": 1.0417, "step": 3657 }, { "epoch": 0.19, "learning_rate": 1.8753442037098878e-05, "loss": 1.0526, "step": 3658 }, { "epoch": 0.19, "learning_rate": 1.8752646359532597e-05, "loss": 0.8842, "step": 3659 }, { "epoch": 0.19, "learning_rate": 1.8751850444997723e-05, "loss": 1.054, "step": 3660 }, { "epoch": 0.19, "learning_rate": 1.875105429351581e-05, "loss": 1.0777, "step": 3661 }, { "epoch": 0.19, "learning_rate": 1.8750257905108414e-05, "loss": 0.9482, "step": 3662 }, { "epoch": 0.19, "learning_rate": 1.874946127979709e-05, "loss": 1.0055, "step": 3663 }, { "epoch": 0.19, "learning_rate": 1.874866441760342e-05, "loss": 1.1067, "step": 3664 }, { "epoch": 0.19, "learning_rate": 1.8747867318548963e-05, "loss": 1.1658, "step": 3665 }, { "epoch": 0.19, "learning_rate": 1.8747069982655303e-05, "loss": 1.0771, "step": 3666 }, { "epoch": 0.19, "learning_rate": 1.874627240994403e-05, "loss": 1.094, "step": 3667 }, { "epoch": 0.19, "learning_rate": 1.874547460043674e-05, "loss": 0.9626, "step": 3668 }, { "epoch": 0.19, "learning_rate": 1.874467655415503e-05, "loss": 0.9357, "step": 3669 }, { "epoch": 0.19, "learning_rate": 1.8743878271120503e-05, "loss": 0.9866, "step": 3670 }, { "epoch": 0.19, "learning_rate": 1.8743079751354775e-05, "loss": 1.0387, "step": 3671 }, { "epoch": 0.19, "learning_rate": 1.8742280994879465e-05, "loss": 0.9813, "step": 3672 }, { "epoch": 0.19, "learning_rate": 1.8741482001716198e-05, "loss": 1.0504, "step": 3673 }, { "epoch": 0.19, "learning_rate": 1.874068277188661e-05, "loss": 1.0343, "step": 3674 }, { "epoch": 0.19, "learning_rate": 1.873988330541233e-05, "loss": 1.0178, "step": 3675 }, { "epoch": 0.19, "learning_rate": 1.873908360231501e-05, "loss": 1.0085, "step": 3676 }, { "epoch": 0.19, "learning_rate": 1.8738283662616297e-05, "loss": 0.8953, "step": 3677 }, { "epoch": 0.19, "learning_rate": 1.8737483486337856e-05, "loss": 0.9683, "step": 3678 }, { "epoch": 0.19, "learning_rate": 1.8736683073501338e-05, "loss": 1.0376, "step": 3679 }, { "epoch": 0.19, "learning_rate": 1.8735882424128427e-05, "loss": 1.047, "step": 3680 }, { "epoch": 0.19, "learning_rate": 1.8735081538240792e-05, "loss": 1.0654, "step": 3681 }, { "epoch": 0.19, "learning_rate": 1.873428041586012e-05, "loss": 0.9333, "step": 3682 }, { "epoch": 0.19, "learning_rate": 1.8733479057008094e-05, "loss": 0.9967, "step": 3683 }, { "epoch": 0.19, "learning_rate": 1.8732677461706416e-05, "loss": 1.0504, "step": 3684 }, { "epoch": 0.19, "learning_rate": 1.8731875629976785e-05, "loss": 1.0205, "step": 3685 }, { "epoch": 0.19, "learning_rate": 1.8731073561840915e-05, "loss": 1.0342, "step": 3686 }, { "epoch": 0.19, "learning_rate": 1.873027125732051e-05, "loss": 1.0825, "step": 3687 }, { "epoch": 0.19, "learning_rate": 1.8729468716437306e-05, "loss": 1.0784, "step": 3688 }, { "epoch": 0.19, "learning_rate": 1.8728665939213022e-05, "loss": 1.058, "step": 3689 }, { "epoch": 0.19, "learning_rate": 1.8727862925669392e-05, "loss": 0.939, "step": 3690 }, { "epoch": 0.19, "learning_rate": 1.8727059675828165e-05, "loss": 0.9203, "step": 3691 }, { "epoch": 0.19, "learning_rate": 1.8726256189711077e-05, "loss": 0.9456, "step": 3692 }, { "epoch": 0.19, "learning_rate": 1.872545246733989e-05, "loss": 0.9938, "step": 3693 }, { "epoch": 0.19, "learning_rate": 1.872464850873636e-05, "loss": 0.8456, "step": 3694 }, { "epoch": 0.19, "learning_rate": 1.872384431392225e-05, "loss": 1.0125, "step": 3695 }, { "epoch": 0.19, "learning_rate": 1.8723039882919338e-05, "loss": 0.9797, "step": 3696 }, { "epoch": 0.19, "learning_rate": 1.8722235215749406e-05, "loss": 1.0603, "step": 3697 }, { "epoch": 0.19, "learning_rate": 1.872143031243423e-05, "loss": 1.0039, "step": 3698 }, { "epoch": 0.19, "learning_rate": 1.872062517299561e-05, "loss": 1.0754, "step": 3699 }, { "epoch": 0.19, "learning_rate": 1.871981979745534e-05, "loss": 0.9471, "step": 3700 }, { "epoch": 0.19, "learning_rate": 1.871901418583523e-05, "loss": 0.9855, "step": 3701 }, { "epoch": 0.19, "learning_rate": 1.8718208338157082e-05, "loss": 0.9995, "step": 3702 }, { "epoch": 0.19, "learning_rate": 1.871740225444272e-05, "loss": 1.0438, "step": 3703 }, { "epoch": 0.19, "learning_rate": 1.871659593471397e-05, "loss": 1.0081, "step": 3704 }, { "epoch": 0.19, "learning_rate": 1.8715789378992656e-05, "loss": 0.9936, "step": 3705 }, { "epoch": 0.19, "learning_rate": 1.8714982587300618e-05, "loss": 1.1241, "step": 3706 }, { "epoch": 0.19, "learning_rate": 1.8714175559659696e-05, "loss": 0.9306, "step": 3707 }, { "epoch": 0.19, "learning_rate": 1.8713368296091746e-05, "loss": 1.031, "step": 3708 }, { "epoch": 0.19, "learning_rate": 1.871256079661862e-05, "loss": 0.9799, "step": 3709 }, { "epoch": 0.19, "learning_rate": 1.8711753061262174e-05, "loss": 1.0249, "step": 3710 }, { "epoch": 0.19, "learning_rate": 1.871094509004429e-05, "loss": 0.9645, "step": 3711 }, { "epoch": 0.19, "learning_rate": 1.871013688298683e-05, "loss": 1.0107, "step": 3712 }, { "epoch": 0.19, "learning_rate": 1.870932844011168e-05, "loss": 0.907, "step": 3713 }, { "epoch": 0.19, "learning_rate": 1.870851976144073e-05, "loss": 0.9755, "step": 3714 }, { "epoch": 0.19, "learning_rate": 1.8707710846995875e-05, "loss": 1.0396, "step": 3715 }, { "epoch": 0.19, "learning_rate": 1.870690169679901e-05, "loss": 1.0567, "step": 3716 }, { "epoch": 0.19, "learning_rate": 1.8706092310872047e-05, "loss": 1.014, "step": 3717 }, { "epoch": 0.19, "learning_rate": 1.87052826892369e-05, "loss": 0.9672, "step": 3718 }, { "epoch": 0.19, "learning_rate": 1.870447283191548e-05, "loss": 0.9154, "step": 3719 }, { "epoch": 0.19, "learning_rate": 1.870366273892972e-05, "loss": 0.958, "step": 3720 }, { "epoch": 0.19, "learning_rate": 1.8702852410301556e-05, "loss": 1.0326, "step": 3721 }, { "epoch": 0.19, "learning_rate": 1.8702041846052917e-05, "loss": 0.9092, "step": 3722 }, { "epoch": 0.19, "learning_rate": 1.8701231046205754e-05, "loss": 1.0417, "step": 3723 }, { "epoch": 0.19, "learning_rate": 1.8700420010782017e-05, "loss": 1.0356, "step": 3724 }, { "epoch": 0.19, "learning_rate": 1.8699608739803667e-05, "loss": 1.0924, "step": 3725 }, { "epoch": 0.19, "learning_rate": 1.869879723329267e-05, "loss": 1.0504, "step": 3726 }, { "epoch": 0.19, "learning_rate": 1.8697985491270983e-05, "loss": 0.9617, "step": 3727 }, { "epoch": 0.19, "learning_rate": 1.86971735137606e-05, "loss": 1.0561, "step": 3728 }, { "epoch": 0.19, "learning_rate": 1.8696361300783496e-05, "loss": 1.0181, "step": 3729 }, { "epoch": 0.19, "learning_rate": 1.869554885236166e-05, "loss": 1.028, "step": 3730 }, { "epoch": 0.19, "learning_rate": 1.869473616851709e-05, "loss": 1.0172, "step": 3731 }, { "epoch": 0.19, "learning_rate": 1.869392324927179e-05, "loss": 1.0109, "step": 3732 }, { "epoch": 0.19, "learning_rate": 1.869311009464777e-05, "loss": 1.0251, "step": 3733 }, { "epoch": 0.19, "learning_rate": 1.869229670466704e-05, "loss": 0.9786, "step": 3734 }, { "epoch": 0.19, "learning_rate": 1.8691483079351628e-05, "loss": 1.0518, "step": 3735 }, { "epoch": 0.19, "learning_rate": 1.8690669218723557e-05, "loss": 0.9809, "step": 3736 }, { "epoch": 0.19, "learning_rate": 1.8689855122804862e-05, "loss": 0.972, "step": 3737 }, { "epoch": 0.19, "learning_rate": 1.868904079161758e-05, "loss": 0.9834, "step": 3738 }, { "epoch": 0.19, "learning_rate": 1.868822622518377e-05, "loss": 0.8704, "step": 3739 }, { "epoch": 0.19, "learning_rate": 1.8687411423525476e-05, "loss": 1.0204, "step": 3740 }, { "epoch": 0.19, "learning_rate": 1.868659638666476e-05, "loss": 1.0255, "step": 3741 }, { "epoch": 0.19, "learning_rate": 1.868578111462369e-05, "loss": 0.9277, "step": 3742 }, { "epoch": 0.19, "learning_rate": 1.8684965607424337e-05, "loss": 1.0255, "step": 3743 }, { "epoch": 0.19, "learning_rate": 1.868414986508878e-05, "loss": 0.9891, "step": 3744 }, { "epoch": 0.19, "learning_rate": 1.8683333887639105e-05, "loss": 0.9691, "step": 3745 }, { "epoch": 0.19, "learning_rate": 1.86825176750974e-05, "loss": 1.0349, "step": 3746 }, { "epoch": 0.19, "learning_rate": 1.8681701227485772e-05, "loss": 0.947, "step": 3747 }, { "epoch": 0.19, "learning_rate": 1.8680884544826315e-05, "loss": 1.1154, "step": 3748 }, { "epoch": 0.19, "learning_rate": 1.868006762714115e-05, "loss": 0.9537, "step": 3749 }, { "epoch": 0.19, "learning_rate": 1.8679250474452382e-05, "loss": 0.9847, "step": 3750 }, { "epoch": 0.19, "learning_rate": 1.8678433086782145e-05, "loss": 1.0021, "step": 3751 }, { "epoch": 0.19, "learning_rate": 1.8677615464152564e-05, "loss": 0.9823, "step": 3752 }, { "epoch": 0.19, "learning_rate": 1.8676797606585778e-05, "loss": 1.0232, "step": 3753 }, { "epoch": 0.19, "learning_rate": 1.867597951410393e-05, "loss": 1.0095, "step": 3754 }, { "epoch": 0.19, "learning_rate": 1.8675161186729166e-05, "loss": 1.0607, "step": 3755 }, { "epoch": 0.19, "learning_rate": 1.8674342624483643e-05, "loss": 1.0555, "step": 3756 }, { "epoch": 0.19, "learning_rate": 1.8673523827389518e-05, "loss": 0.9626, "step": 3757 }, { "epoch": 0.19, "learning_rate": 1.867270479546897e-05, "loss": 0.8756, "step": 3758 }, { "epoch": 0.19, "learning_rate": 1.867188552874416e-05, "loss": 1.0407, "step": 3759 }, { "epoch": 0.19, "learning_rate": 1.8671066027237278e-05, "loss": 0.9311, "step": 3760 }, { "epoch": 0.19, "learning_rate": 1.8670246290970507e-05, "loss": 1.0034, "step": 3761 }, { "epoch": 0.19, "learning_rate": 1.8669426319966048e-05, "loss": 1.0476, "step": 3762 }, { "epoch": 0.19, "learning_rate": 1.866860611424609e-05, "loss": 0.9949, "step": 3763 }, { "epoch": 0.19, "learning_rate": 1.8667785673832846e-05, "loss": 1.0772, "step": 3764 }, { "epoch": 0.19, "learning_rate": 1.8666964998748526e-05, "loss": 1.0299, "step": 3765 }, { "epoch": 0.19, "learning_rate": 1.8666144089015347e-05, "loss": 1.0516, "step": 3766 }, { "epoch": 0.19, "learning_rate": 1.866532294465554e-05, "loss": 1.0345, "step": 3767 }, { "epoch": 0.19, "learning_rate": 1.8664501565691333e-05, "loss": 0.9924, "step": 3768 }, { "epoch": 0.19, "learning_rate": 1.8663679952144964e-05, "loss": 0.9516, "step": 3769 }, { "epoch": 0.19, "learning_rate": 1.8662858104038677e-05, "loss": 1.0452, "step": 3770 }, { "epoch": 0.19, "learning_rate": 1.8662036021394724e-05, "loss": 1.0167, "step": 3771 }, { "epoch": 0.19, "learning_rate": 1.866121370423536e-05, "loss": 0.9322, "step": 3772 }, { "epoch": 0.19, "learning_rate": 1.8660391152582855e-05, "loss": 1.0335, "step": 3773 }, { "epoch": 0.19, "learning_rate": 1.865956836645947e-05, "loss": 1.0504, "step": 3774 }, { "epoch": 0.19, "learning_rate": 1.8658745345887484e-05, "loss": 1.0019, "step": 3775 }, { "epoch": 0.19, "learning_rate": 1.8657922090889176e-05, "loss": 0.9609, "step": 3776 }, { "epoch": 0.19, "learning_rate": 1.8657098601486847e-05, "loss": 0.9999, "step": 3777 }, { "epoch": 0.19, "learning_rate": 1.865627487770278e-05, "loss": 1.0215, "step": 3778 }, { "epoch": 0.19, "learning_rate": 1.865545091955928e-05, "loss": 1.0063, "step": 3779 }, { "epoch": 0.19, "learning_rate": 1.8654626727078656e-05, "loss": 0.9756, "step": 3780 }, { "epoch": 0.19, "learning_rate": 1.865380230028322e-05, "loss": 0.9387, "step": 3781 }, { "epoch": 0.19, "learning_rate": 1.865297763919529e-05, "loss": 1.056, "step": 3782 }, { "epoch": 0.19, "learning_rate": 1.8652152743837205e-05, "loss": 1.0243, "step": 3783 }, { "epoch": 0.19, "learning_rate": 1.8651327614231284e-05, "loss": 1.0505, "step": 3784 }, { "epoch": 0.19, "learning_rate": 1.8650502250399872e-05, "loss": 0.9811, "step": 3785 }, { "epoch": 0.19, "learning_rate": 1.8649676652365318e-05, "loss": 1.0175, "step": 3786 }, { "epoch": 0.19, "learning_rate": 1.8648850820149968e-05, "loss": 1.0632, "step": 3787 }, { "epoch": 0.19, "learning_rate": 1.8648024753776182e-05, "loss": 1.0317, "step": 3788 }, { "epoch": 0.19, "learning_rate": 1.8647198453266332e-05, "loss": 1.1021, "step": 3789 }, { "epoch": 0.19, "learning_rate": 1.8646371918642782e-05, "loss": 0.9857, "step": 3790 }, { "epoch": 0.19, "learning_rate": 1.8645545149927907e-05, "loss": 1.0507, "step": 3791 }, { "epoch": 0.19, "learning_rate": 1.8644718147144104e-05, "loss": 0.9479, "step": 3792 }, { "epoch": 0.19, "learning_rate": 1.8643890910313748e-05, "loss": 1.0786, "step": 3793 }, { "epoch": 0.19, "learning_rate": 1.8643063439459243e-05, "loss": 0.9102, "step": 3794 }, { "epoch": 0.19, "learning_rate": 1.864223573460299e-05, "loss": 0.9156, "step": 3795 }, { "epoch": 0.19, "learning_rate": 1.86414077957674e-05, "loss": 1.0276, "step": 3796 }, { "epoch": 0.19, "learning_rate": 1.864057962297489e-05, "loss": 0.9301, "step": 3797 }, { "epoch": 0.19, "learning_rate": 1.8639751216247877e-05, "loss": 1.0116, "step": 3798 }, { "epoch": 0.19, "learning_rate": 1.863892257560879e-05, "loss": 0.9733, "step": 3799 }, { "epoch": 0.19, "learning_rate": 1.8638093701080068e-05, "loss": 0.9994, "step": 3800 }, { "epoch": 0.19, "learning_rate": 1.863726459268415e-05, "loss": 0.9309, "step": 3801 }, { "epoch": 0.19, "learning_rate": 1.8636435250443477e-05, "loss": 1.0767, "step": 3802 }, { "epoch": 0.19, "learning_rate": 1.8635605674380516e-05, "loss": 1.025, "step": 3803 }, { "epoch": 0.19, "learning_rate": 1.8634775864517715e-05, "loss": 0.9712, "step": 3804 }, { "epoch": 0.19, "learning_rate": 1.8633945820877543e-05, "loss": 1.0453, "step": 3805 }, { "epoch": 0.19, "learning_rate": 1.8633115543482476e-05, "loss": 0.9795, "step": 3806 }, { "epoch": 0.19, "learning_rate": 1.8632285032354992e-05, "loss": 1.0374, "step": 3807 }, { "epoch": 0.19, "learning_rate": 1.863145428751757e-05, "loss": 0.9943, "step": 3808 }, { "epoch": 0.19, "learning_rate": 1.863062330899271e-05, "loss": 1.1157, "step": 3809 }, { "epoch": 0.19, "learning_rate": 1.8629792096802903e-05, "loss": 1.1, "step": 3810 }, { "epoch": 0.19, "learning_rate": 1.862896065097066e-05, "loss": 0.9373, "step": 3811 }, { "epoch": 0.19, "learning_rate": 1.8628128971518485e-05, "loss": 1.0162, "step": 3812 }, { "epoch": 0.19, "learning_rate": 1.8627297058468898e-05, "loss": 1.0311, "step": 3813 }, { "epoch": 0.19, "learning_rate": 1.8626464911844426e-05, "loss": 1.0351, "step": 3814 }, { "epoch": 0.19, "learning_rate": 1.862563253166759e-05, "loss": 1.0013, "step": 3815 }, { "epoch": 0.19, "learning_rate": 1.862479991796093e-05, "loss": 1.0366, "step": 3816 }, { "epoch": 0.19, "learning_rate": 1.862396707074699e-05, "loss": 0.8697, "step": 3817 }, { "epoch": 0.19, "learning_rate": 1.8623133990048316e-05, "loss": 1.0426, "step": 3818 }, { "epoch": 0.19, "learning_rate": 1.8622300675887463e-05, "loss": 0.9755, "step": 3819 }, { "epoch": 0.19, "learning_rate": 1.8621467128286993e-05, "loss": 0.9675, "step": 3820 }, { "epoch": 0.19, "learning_rate": 1.8620633347269474e-05, "loss": 1.1135, "step": 3821 }, { "epoch": 0.19, "learning_rate": 1.8619799332857478e-05, "loss": 1.0142, "step": 3822 }, { "epoch": 0.19, "learning_rate": 1.8618965085073587e-05, "loss": 1.0868, "step": 3823 }, { "epoch": 0.19, "learning_rate": 1.8618130603940386e-05, "loss": 1.1031, "step": 3824 }, { "epoch": 0.19, "learning_rate": 1.8617295889480464e-05, "loss": 0.9559, "step": 3825 }, { "epoch": 0.19, "learning_rate": 1.861646094171643e-05, "loss": 0.8779, "step": 3826 }, { "epoch": 0.19, "learning_rate": 1.8615625760670878e-05, "loss": 1.0034, "step": 3827 }, { "epoch": 0.19, "learning_rate": 1.861479034636643e-05, "loss": 1.0098, "step": 3828 }, { "epoch": 0.19, "learning_rate": 1.8613954698825694e-05, "loss": 1.0112, "step": 3829 }, { "epoch": 0.19, "learning_rate": 1.8613118818071297e-05, "loss": 0.9705, "step": 3830 }, { "epoch": 0.19, "learning_rate": 1.8612282704125877e-05, "loss": 0.9586, "step": 3831 }, { "epoch": 0.19, "learning_rate": 1.8611446357012063e-05, "loss": 1.0142, "step": 3832 }, { "epoch": 0.19, "learning_rate": 1.8610609776752505e-05, "loss": 0.9594, "step": 3833 }, { "epoch": 0.19, "learning_rate": 1.8609772963369842e-05, "loss": 0.9833, "step": 3834 }, { "epoch": 0.19, "learning_rate": 1.860893591688674e-05, "loss": 1.0771, "step": 3835 }, { "epoch": 0.19, "learning_rate": 1.860809863732586e-05, "loss": 0.9636, "step": 3836 }, { "epoch": 0.19, "learning_rate": 1.860726112470986e-05, "loss": 1.0394, "step": 3837 }, { "epoch": 0.19, "learning_rate": 1.8606423379061428e-05, "loss": 1.123, "step": 3838 }, { "epoch": 0.2, "learning_rate": 1.8605585400403237e-05, "loss": 1.0362, "step": 3839 }, { "epoch": 0.2, "learning_rate": 1.8604747188757975e-05, "loss": 1.0538, "step": 3840 }, { "epoch": 0.2, "learning_rate": 1.8603908744148343e-05, "loss": 1.0339, "step": 3841 }, { "epoch": 0.2, "learning_rate": 1.8603070066597033e-05, "loss": 0.9811, "step": 3842 }, { "epoch": 0.2, "learning_rate": 1.8602231156126757e-05, "loss": 0.9677, "step": 3843 }, { "epoch": 0.2, "learning_rate": 1.860139201276022e-05, "loss": 1.019, "step": 3844 }, { "epoch": 0.2, "learning_rate": 1.8600552636520145e-05, "loss": 1.0551, "step": 3845 }, { "epoch": 0.2, "learning_rate": 1.8599713027429258e-05, "loss": 0.9431, "step": 3846 }, { "epoch": 0.2, "learning_rate": 1.8598873185510294e-05, "loss": 1.0368, "step": 3847 }, { "epoch": 0.2, "learning_rate": 1.8598033110785984e-05, "loss": 0.9875, "step": 3848 }, { "epoch": 0.2, "learning_rate": 1.8597192803279074e-05, "loss": 1.0814, "step": 3849 }, { "epoch": 0.2, "learning_rate": 1.8596352263012318e-05, "loss": 1.1008, "step": 3850 }, { "epoch": 0.2, "learning_rate": 1.859551149000847e-05, "loss": 1.0162, "step": 3851 }, { "epoch": 0.2, "learning_rate": 1.8594670484290293e-05, "loss": 0.8582, "step": 3852 }, { "epoch": 0.2, "learning_rate": 1.8593829245880557e-05, "loss": 0.9018, "step": 3853 }, { "epoch": 0.2, "learning_rate": 1.8592987774802037e-05, "loss": 0.8936, "step": 3854 }, { "epoch": 0.2, "learning_rate": 1.8592146071077516e-05, "loss": 1.0395, "step": 3855 }, { "epoch": 0.2, "learning_rate": 1.8591304134729776e-05, "loss": 0.9842, "step": 3856 }, { "epoch": 0.2, "learning_rate": 1.8590461965781624e-05, "loss": 0.9086, "step": 3857 }, { "epoch": 0.2, "learning_rate": 1.858961956425585e-05, "loss": 1.1106, "step": 3858 }, { "epoch": 0.2, "learning_rate": 1.8588776930175267e-05, "loss": 1.0166, "step": 3859 }, { "epoch": 0.2, "learning_rate": 1.8587934063562686e-05, "loss": 1.0084, "step": 3860 }, { "epoch": 0.2, "learning_rate": 1.8587090964440926e-05, "loss": 1.0651, "step": 3861 }, { "epoch": 0.2, "learning_rate": 1.8586247632832813e-05, "loss": 1.0114, "step": 3862 }, { "epoch": 0.2, "learning_rate": 1.8585404068761184e-05, "loss": 1.0955, "step": 3863 }, { "epoch": 0.2, "learning_rate": 1.8584560272248873e-05, "loss": 0.9576, "step": 3864 }, { "epoch": 0.2, "learning_rate": 1.8583716243318725e-05, "loss": 0.9585, "step": 3865 }, { "epoch": 0.2, "learning_rate": 1.8582871981993592e-05, "loss": 1.0322, "step": 3866 }, { "epoch": 0.2, "learning_rate": 1.858202748829633e-05, "loss": 0.9746, "step": 3867 }, { "epoch": 0.2, "learning_rate": 1.8581182762249806e-05, "loss": 1.0452, "step": 3868 }, { "epoch": 0.2, "learning_rate": 1.858033780387689e-05, "loss": 0.9572, "step": 3869 }, { "epoch": 0.2, "learning_rate": 1.8579492613200455e-05, "loss": 0.9954, "step": 3870 }, { "epoch": 0.2, "learning_rate": 1.8578647190243384e-05, "loss": 1.0645, "step": 3871 }, { "epoch": 0.2, "learning_rate": 1.8577801535028567e-05, "loss": 1.0576, "step": 3872 }, { "epoch": 0.2, "learning_rate": 1.8576955647578903e-05, "loss": 0.9323, "step": 3873 }, { "epoch": 0.2, "learning_rate": 1.8576109527917287e-05, "loss": 0.9494, "step": 3874 }, { "epoch": 0.2, "learning_rate": 1.857526317606663e-05, "loss": 0.893, "step": 3875 }, { "epoch": 0.2, "learning_rate": 1.8574416592049847e-05, "loss": 0.9904, "step": 3876 }, { "epoch": 0.2, "learning_rate": 1.8573569775889857e-05, "loss": 0.9161, "step": 3877 }, { "epoch": 0.2, "learning_rate": 1.857272272760959e-05, "loss": 0.9858, "step": 3878 }, { "epoch": 0.2, "learning_rate": 1.8571875447231972e-05, "loss": 0.9477, "step": 3879 }, { "epoch": 0.2, "learning_rate": 1.8571027934779942e-05, "loss": 0.9858, "step": 3880 }, { "epoch": 0.2, "learning_rate": 1.857018019027646e-05, "loss": 1.0679, "step": 3881 }, { "epoch": 0.2, "learning_rate": 1.8569332213744457e-05, "loss": 0.9591, "step": 3882 }, { "epoch": 0.2, "learning_rate": 1.8568484005206908e-05, "loss": 1.0133, "step": 3883 }, { "epoch": 0.2, "learning_rate": 1.856763556468677e-05, "loss": 1.0992, "step": 3884 }, { "epoch": 0.2, "learning_rate": 1.856678689220701e-05, "loss": 1.0128, "step": 3885 }, { "epoch": 0.2, "learning_rate": 1.8565937987790615e-05, "loss": 1.0544, "step": 3886 }, { "epoch": 0.2, "learning_rate": 1.856508885146056e-05, "loss": 0.9867, "step": 3887 }, { "epoch": 0.2, "learning_rate": 1.8564239483239836e-05, "loss": 1.0079, "step": 3888 }, { "epoch": 0.2, "learning_rate": 1.856338988315144e-05, "loss": 1.0743, "step": 3889 }, { "epoch": 0.2, "learning_rate": 1.856254005121837e-05, "loss": 0.9877, "step": 3890 }, { "epoch": 0.2, "learning_rate": 1.8561689987463646e-05, "loss": 1.0328, "step": 3891 }, { "epoch": 0.2, "learning_rate": 1.8560839691910268e-05, "loss": 0.9551, "step": 3892 }, { "epoch": 0.2, "learning_rate": 1.8559989164581265e-05, "loss": 1.0606, "step": 3893 }, { "epoch": 0.2, "learning_rate": 1.8559138405499663e-05, "loss": 1.0023, "step": 3894 }, { "epoch": 0.2, "learning_rate": 1.8558287414688496e-05, "loss": 1.0252, "step": 3895 }, { "epoch": 0.2, "learning_rate": 1.85574361921708e-05, "loss": 1.0805, "step": 3896 }, { "epoch": 0.2, "learning_rate": 1.8556584737969626e-05, "loss": 1.0399, "step": 3897 }, { "epoch": 0.2, "learning_rate": 1.8555733052108023e-05, "loss": 0.9577, "step": 3898 }, { "epoch": 0.2, "learning_rate": 1.8554881134609052e-05, "loss": 1.0785, "step": 3899 }, { "epoch": 0.2, "learning_rate": 1.8554028985495772e-05, "loss": 0.9811, "step": 3900 }, { "epoch": 0.2, "learning_rate": 1.855317660479126e-05, "loss": 0.8859, "step": 3901 }, { "epoch": 0.2, "learning_rate": 1.855232399251859e-05, "loss": 1.0201, "step": 3902 }, { "epoch": 0.2, "learning_rate": 1.855147114870085e-05, "loss": 0.9929, "step": 3903 }, { "epoch": 0.2, "learning_rate": 1.8550618073361125e-05, "loss": 1.0209, "step": 3904 }, { "epoch": 0.2, "learning_rate": 1.8549764766522508e-05, "loss": 0.9968, "step": 3905 }, { "epoch": 0.2, "learning_rate": 1.8548911228208113e-05, "loss": 1.0316, "step": 3906 }, { "epoch": 0.2, "learning_rate": 1.854805745844104e-05, "loss": 0.983, "step": 3907 }, { "epoch": 0.2, "learning_rate": 1.8547203457244406e-05, "loss": 0.9338, "step": 3908 }, { "epoch": 0.2, "learning_rate": 1.854634922464133e-05, "loss": 1.0226, "step": 3909 }, { "epoch": 0.2, "learning_rate": 1.8545494760654944e-05, "loss": 0.9368, "step": 3910 }, { "epoch": 0.2, "learning_rate": 1.8544640065308376e-05, "loss": 1.0698, "step": 3911 }, { "epoch": 0.2, "learning_rate": 1.8543785138624768e-05, "loss": 1.0927, "step": 3912 }, { "epoch": 0.2, "learning_rate": 1.8542929980627273e-05, "loss": 1.0587, "step": 3913 }, { "epoch": 0.2, "learning_rate": 1.8542074591339032e-05, "loss": 0.9031, "step": 3914 }, { "epoch": 0.2, "learning_rate": 1.854121897078321e-05, "loss": 0.9915, "step": 3915 }, { "epoch": 0.2, "learning_rate": 1.8540363118982977e-05, "loss": 0.9742, "step": 3916 }, { "epoch": 0.2, "learning_rate": 1.8539507035961495e-05, "loss": 1.0868, "step": 3917 }, { "epoch": 0.2, "learning_rate": 1.8538650721741947e-05, "loss": 0.9674, "step": 3918 }, { "epoch": 0.2, "learning_rate": 1.8537794176347514e-05, "loss": 1.0532, "step": 3919 }, { "epoch": 0.2, "learning_rate": 1.8536937399801384e-05, "loss": 1.0378, "step": 3920 }, { "epoch": 0.2, "learning_rate": 1.853608039212676e-05, "loss": 0.9977, "step": 3921 }, { "epoch": 0.2, "learning_rate": 1.8535223153346843e-05, "loss": 1.0419, "step": 3922 }, { "epoch": 0.2, "learning_rate": 1.8534365683484835e-05, "loss": 1.0374, "step": 3923 }, { "epoch": 0.2, "learning_rate": 1.853350798256396e-05, "loss": 1.0184, "step": 3924 }, { "epoch": 0.2, "learning_rate": 1.8532650050607433e-05, "loss": 0.9142, "step": 3925 }, { "epoch": 0.2, "learning_rate": 1.8531791887638483e-05, "loss": 1.0198, "step": 3926 }, { "epoch": 0.2, "learning_rate": 1.8530933493680345e-05, "loss": 1.0158, "step": 3927 }, { "epoch": 0.2, "learning_rate": 1.853007486875626e-05, "loss": 0.997, "step": 3928 }, { "epoch": 0.2, "learning_rate": 1.8529216012889476e-05, "loss": 0.9208, "step": 3929 }, { "epoch": 0.2, "learning_rate": 1.852835692610324e-05, "loss": 1.101, "step": 3930 }, { "epoch": 0.2, "learning_rate": 1.8527497608420812e-05, "loss": 1.0822, "step": 3931 }, { "epoch": 0.2, "learning_rate": 1.8526638059865463e-05, "loss": 0.9588, "step": 3932 }, { "epoch": 0.2, "learning_rate": 1.8525778280460457e-05, "loss": 0.9556, "step": 3933 }, { "epoch": 0.2, "learning_rate": 1.8524918270229072e-05, "loss": 1.0463, "step": 3934 }, { "epoch": 0.2, "learning_rate": 1.85240580291946e-05, "loss": 1.0253, "step": 3935 }, { "epoch": 0.2, "learning_rate": 1.8523197557380325e-05, "loss": 0.9636, "step": 3936 }, { "epoch": 0.2, "learning_rate": 1.8522336854809546e-05, "loss": 0.9949, "step": 3937 }, { "epoch": 0.2, "learning_rate": 1.8521475921505558e-05, "loss": 1.001, "step": 3938 }, { "epoch": 0.2, "learning_rate": 1.8520614757491677e-05, "loss": 0.9542, "step": 3939 }, { "epoch": 0.2, "learning_rate": 1.851975336279122e-05, "loss": 1.0023, "step": 3940 }, { "epoch": 0.2, "learning_rate": 1.85188917374275e-05, "loss": 1.0903, "step": 3941 }, { "epoch": 0.2, "learning_rate": 1.8518029881423854e-05, "loss": 0.9323, "step": 3942 }, { "epoch": 0.2, "learning_rate": 1.851716779480361e-05, "loss": 0.9186, "step": 3943 }, { "epoch": 0.2, "learning_rate": 1.851630547759011e-05, "loss": 0.9421, "step": 3944 }, { "epoch": 0.2, "learning_rate": 1.8515442929806698e-05, "loss": 0.9719, "step": 3945 }, { "epoch": 0.2, "learning_rate": 1.851458015147673e-05, "loss": 1.0581, "step": 3946 }, { "epoch": 0.2, "learning_rate": 1.8513717142623565e-05, "loss": 1.0516, "step": 3947 }, { "epoch": 0.2, "learning_rate": 1.8512853903270563e-05, "loss": 0.8664, "step": 3948 }, { "epoch": 0.2, "learning_rate": 1.8511990433441102e-05, "loss": 1.0345, "step": 3949 }, { "epoch": 0.2, "learning_rate": 1.851112673315855e-05, "loss": 0.9191, "step": 3950 }, { "epoch": 0.2, "learning_rate": 1.8510262802446303e-05, "loss": 1.0615, "step": 3951 }, { "epoch": 0.2, "learning_rate": 1.8509398641327744e-05, "loss": 1.0406, "step": 3952 }, { "epoch": 0.2, "learning_rate": 1.8508534249826263e-05, "loss": 1.0602, "step": 3953 }, { "epoch": 0.2, "learning_rate": 1.8507669627965278e-05, "loss": 0.9228, "step": 3954 }, { "epoch": 0.2, "learning_rate": 1.8506804775768187e-05, "loss": 1.0302, "step": 3955 }, { "epoch": 0.2, "learning_rate": 1.8505939693258407e-05, "loss": 0.9719, "step": 3956 }, { "epoch": 0.2, "learning_rate": 1.850507438045936e-05, "loss": 1.04, "step": 3957 }, { "epoch": 0.2, "learning_rate": 1.8504208837394468e-05, "loss": 0.9865, "step": 3958 }, { "epoch": 0.2, "learning_rate": 1.8503343064087174e-05, "loss": 1.0468, "step": 3959 }, { "epoch": 0.2, "learning_rate": 1.8502477060560913e-05, "loss": 0.9487, "step": 3960 }, { "epoch": 0.2, "learning_rate": 1.8501610826839133e-05, "loss": 0.9102, "step": 3961 }, { "epoch": 0.2, "learning_rate": 1.850074436294528e-05, "loss": 1.0016, "step": 3962 }, { "epoch": 0.2, "learning_rate": 1.8499877668902822e-05, "loss": 1.0238, "step": 3963 }, { "epoch": 0.2, "learning_rate": 1.8499010744735217e-05, "loss": 1.0253, "step": 3964 }, { "epoch": 0.2, "learning_rate": 1.8498143590465942e-05, "loss": 0.9592, "step": 3965 }, { "epoch": 0.2, "learning_rate": 1.8497276206118465e-05, "loss": 0.8663, "step": 3966 }, { "epoch": 0.2, "learning_rate": 1.8496408591716278e-05, "loss": 0.9958, "step": 3967 }, { "epoch": 0.2, "learning_rate": 1.8495540747282867e-05, "loss": 1.0026, "step": 3968 }, { "epoch": 0.2, "learning_rate": 1.849467267284173e-05, "loss": 1.0385, "step": 3969 }, { "epoch": 0.2, "learning_rate": 1.8493804368416367e-05, "loss": 0.8782, "step": 3970 }, { "epoch": 0.2, "learning_rate": 1.8492935834030286e-05, "loss": 1.0108, "step": 3971 }, { "epoch": 0.2, "learning_rate": 1.849206706970701e-05, "loss": 0.9726, "step": 3972 }, { "epoch": 0.2, "learning_rate": 1.8491198075470044e-05, "loss": 0.9051, "step": 3973 }, { "epoch": 0.2, "learning_rate": 1.849032885134293e-05, "loss": 0.9739, "step": 3974 }, { "epoch": 0.2, "learning_rate": 1.8489459397349194e-05, "loss": 1.0677, "step": 3975 }, { "epoch": 0.2, "learning_rate": 1.8488589713512373e-05, "loss": 1.0349, "step": 3976 }, { "epoch": 0.2, "learning_rate": 1.8487719799856022e-05, "loss": 1.0608, "step": 3977 }, { "epoch": 0.2, "learning_rate": 1.8486849656403684e-05, "loss": 0.9884, "step": 3978 }, { "epoch": 0.2, "learning_rate": 1.8485979283178925e-05, "loss": 1.02, "step": 3979 }, { "epoch": 0.2, "learning_rate": 1.84851086802053e-05, "loss": 0.8801, "step": 3980 }, { "epoch": 0.2, "learning_rate": 1.8484237847506387e-05, "loss": 0.9651, "step": 3981 }, { "epoch": 0.2, "learning_rate": 1.8483366785105762e-05, "loss": 0.9974, "step": 3982 }, { "epoch": 0.2, "learning_rate": 1.8482495493027007e-05, "loss": 1.0676, "step": 3983 }, { "epoch": 0.2, "learning_rate": 1.8481623971293713e-05, "loss": 1.031, "step": 3984 }, { "epoch": 0.2, "learning_rate": 1.848075221992947e-05, "loss": 1.0157, "step": 3985 }, { "epoch": 0.2, "learning_rate": 1.8479880238957882e-05, "loss": 0.9841, "step": 3986 }, { "epoch": 0.2, "learning_rate": 1.8479008028402562e-05, "loss": 0.9373, "step": 3987 }, { "epoch": 0.2, "learning_rate": 1.8478135588287117e-05, "loss": 0.9525, "step": 3988 }, { "epoch": 0.2, "learning_rate": 1.8477262918635176e-05, "loss": 0.9124, "step": 3989 }, { "epoch": 0.2, "learning_rate": 1.8476390019470358e-05, "loss": 0.9981, "step": 3990 }, { "epoch": 0.2, "learning_rate": 1.8475516890816297e-05, "loss": 1.0096, "step": 3991 }, { "epoch": 0.2, "learning_rate": 1.8474643532696634e-05, "loss": 1.0175, "step": 3992 }, { "epoch": 0.2, "learning_rate": 1.8473769945135017e-05, "loss": 1.1033, "step": 3993 }, { "epoch": 0.2, "learning_rate": 1.8472896128155092e-05, "loss": 1.1243, "step": 3994 }, { "epoch": 0.2, "learning_rate": 1.8472022081780515e-05, "loss": 1.074, "step": 3995 }, { "epoch": 0.2, "learning_rate": 1.8471147806034962e-05, "loss": 1.0442, "step": 3996 }, { "epoch": 0.2, "learning_rate": 1.8470273300942086e-05, "loss": 0.9164, "step": 3997 }, { "epoch": 0.2, "learning_rate": 1.8469398566525577e-05, "loss": 0.9379, "step": 3998 }, { "epoch": 0.2, "learning_rate": 1.846852360280911e-05, "loss": 0.9451, "step": 3999 }, { "epoch": 0.2, "learning_rate": 1.846764840981638e-05, "loss": 1.012, "step": 4000 }, { "epoch": 0.2, "learning_rate": 1.846677298757108e-05, "loss": 1.0792, "step": 4001 }, { "epoch": 0.2, "learning_rate": 1.84658973360969e-05, "loss": 0.9605, "step": 4002 }, { "epoch": 0.2, "learning_rate": 1.846502145541756e-05, "loss": 0.888, "step": 4003 }, { "epoch": 0.2, "learning_rate": 1.8464145345556776e-05, "loss": 0.9458, "step": 4004 }, { "epoch": 0.2, "learning_rate": 1.846326900653826e-05, "loss": 0.9918, "step": 4005 }, { "epoch": 0.2, "learning_rate": 1.8462392438385733e-05, "loss": 1.0371, "step": 4006 }, { "epoch": 0.2, "learning_rate": 1.846151564112294e-05, "loss": 1.0248, "step": 4007 }, { "epoch": 0.2, "learning_rate": 1.846063861477361e-05, "loss": 1.0052, "step": 4008 }, { "epoch": 0.2, "learning_rate": 1.8459761359361496e-05, "loss": 1.0352, "step": 4009 }, { "epoch": 0.2, "learning_rate": 1.845888387491034e-05, "loss": 1.0204, "step": 4010 }, { "epoch": 0.2, "learning_rate": 1.8458006161443905e-05, "loss": 1.0924, "step": 4011 }, { "epoch": 0.2, "learning_rate": 1.845712821898595e-05, "loss": 1.03, "step": 4012 }, { "epoch": 0.2, "learning_rate": 1.8456250047560247e-05, "loss": 1.0027, "step": 4013 }, { "epoch": 0.2, "learning_rate": 1.8455371647190567e-05, "loss": 1.0523, "step": 4014 }, { "epoch": 0.2, "learning_rate": 1.8454493017900698e-05, "loss": 0.9983, "step": 4015 }, { "epoch": 0.2, "learning_rate": 1.845361415971443e-05, "loss": 0.9724, "step": 4016 }, { "epoch": 0.2, "learning_rate": 1.8452735072655545e-05, "loss": 1.0219, "step": 4017 }, { "epoch": 0.2, "learning_rate": 1.8451855756747857e-05, "loss": 0.9713, "step": 4018 }, { "epoch": 0.2, "learning_rate": 1.845097621201516e-05, "loss": 1.0145, "step": 4019 }, { "epoch": 0.2, "learning_rate": 1.845009643848128e-05, "loss": 0.8663, "step": 4020 }, { "epoch": 0.2, "learning_rate": 1.8449216436170024e-05, "loss": 1.0351, "step": 4021 }, { "epoch": 0.2, "learning_rate": 1.8448336205105226e-05, "loss": 0.9758, "step": 4022 }, { "epoch": 0.2, "learning_rate": 1.844745574531071e-05, "loss": 1.0617, "step": 4023 }, { "epoch": 0.2, "learning_rate": 1.8446575056810322e-05, "loss": 1.0643, "step": 4024 }, { "epoch": 0.2, "learning_rate": 1.8445694139627895e-05, "loss": 1.0509, "step": 4025 }, { "epoch": 0.2, "learning_rate": 1.8444812993787287e-05, "loss": 0.9507, "step": 4026 }, { "epoch": 0.2, "learning_rate": 1.8443931619312354e-05, "loss": 1.0176, "step": 4027 }, { "epoch": 0.2, "learning_rate": 1.8443050016226952e-05, "loss": 0.9977, "step": 4028 }, { "epoch": 0.2, "learning_rate": 1.8442168184554957e-05, "loss": 0.9741, "step": 4029 }, { "epoch": 0.2, "learning_rate": 1.844128612432024e-05, "loss": 0.9841, "step": 4030 }, { "epoch": 0.2, "learning_rate": 1.8440403835546678e-05, "loss": 0.9717, "step": 4031 }, { "epoch": 0.2, "learning_rate": 1.8439521318258165e-05, "loss": 1.0015, "step": 4032 }, { "epoch": 0.2, "learning_rate": 1.8438638572478592e-05, "loss": 0.9731, "step": 4033 }, { "epoch": 0.2, "learning_rate": 1.8437755598231857e-05, "loss": 0.9778, "step": 4034 }, { "epoch": 0.2, "learning_rate": 1.843687239554187e-05, "loss": 0.8933, "step": 4035 }, { "epoch": 0.21, "learning_rate": 1.8435988964432532e-05, "loss": 1.0051, "step": 4036 }, { "epoch": 0.21, "learning_rate": 1.843510530492777e-05, "loss": 1.0452, "step": 4037 }, { "epoch": 0.21, "learning_rate": 1.8434221417051506e-05, "loss": 0.9581, "step": 4038 }, { "epoch": 0.21, "learning_rate": 1.8433337300827674e-05, "loss": 1.0468, "step": 4039 }, { "epoch": 0.21, "learning_rate": 1.8432452956280206e-05, "loss": 0.9583, "step": 4040 }, { "epoch": 0.21, "learning_rate": 1.8431568383433042e-05, "loss": 0.9949, "step": 4041 }, { "epoch": 0.21, "learning_rate": 1.843068358231014e-05, "loss": 0.9579, "step": 4042 }, { "epoch": 0.21, "learning_rate": 1.8429798552935443e-05, "loss": 0.9871, "step": 4043 }, { "epoch": 0.21, "learning_rate": 1.8428913295332922e-05, "loss": 0.8264, "step": 4044 }, { "epoch": 0.21, "learning_rate": 1.8428027809526543e-05, "loss": 0.9984, "step": 4045 }, { "epoch": 0.21, "learning_rate": 1.8427142095540275e-05, "loss": 0.9818, "step": 4046 }, { "epoch": 0.21, "learning_rate": 1.8426256153398104e-05, "loss": 0.9727, "step": 4047 }, { "epoch": 0.21, "learning_rate": 1.842536998312401e-05, "loss": 0.9868, "step": 4048 }, { "epoch": 0.21, "learning_rate": 1.8424483584741985e-05, "loss": 0.9975, "step": 4049 }, { "epoch": 0.21, "learning_rate": 1.8423596958276034e-05, "loss": 1.0079, "step": 4050 }, { "epoch": 0.21, "learning_rate": 1.8422710103750155e-05, "loss": 1.0868, "step": 4051 }, { "epoch": 0.21, "learning_rate": 1.842182302118836e-05, "loss": 0.9468, "step": 4052 }, { "epoch": 0.21, "learning_rate": 1.842093571061467e-05, "loss": 0.9319, "step": 4053 }, { "epoch": 0.21, "learning_rate": 1.8420048172053104e-05, "loss": 1.1055, "step": 4054 }, { "epoch": 0.21, "learning_rate": 1.841916040552769e-05, "loss": 1.0654, "step": 4055 }, { "epoch": 0.21, "learning_rate": 1.8418272411062466e-05, "loss": 1.0622, "step": 4056 }, { "epoch": 0.21, "learning_rate": 1.841738418868148e-05, "loss": 1.0117, "step": 4057 }, { "epoch": 0.21, "learning_rate": 1.8416495738408763e-05, "loss": 1.0482, "step": 4058 }, { "epoch": 0.21, "learning_rate": 1.841560706026838e-05, "loss": 1.0217, "step": 4059 }, { "epoch": 0.21, "learning_rate": 1.8414718154284393e-05, "loss": 0.9385, "step": 4060 }, { "epoch": 0.21, "learning_rate": 1.841382902048086e-05, "loss": 0.9507, "step": 4061 }, { "epoch": 0.21, "learning_rate": 1.8412939658881862e-05, "loss": 0.975, "step": 4062 }, { "epoch": 0.21, "learning_rate": 1.841205006951147e-05, "loss": 1.0066, "step": 4063 }, { "epoch": 0.21, "learning_rate": 1.8411160252393773e-05, "loss": 0.9548, "step": 4064 }, { "epoch": 0.21, "learning_rate": 1.8410270207552864e-05, "loss": 1.1129, "step": 4065 }, { "epoch": 0.21, "learning_rate": 1.8409379935012836e-05, "loss": 0.9281, "step": 4066 }, { "epoch": 0.21, "learning_rate": 1.840848943479779e-05, "loss": 0.9398, "step": 4067 }, { "epoch": 0.21, "learning_rate": 1.840759870693184e-05, "loss": 1.0119, "step": 4068 }, { "epoch": 0.21, "learning_rate": 1.84067077514391e-05, "loss": 1.0211, "step": 4069 }, { "epoch": 0.21, "learning_rate": 1.840581656834369e-05, "loss": 0.9519, "step": 4070 }, { "epoch": 0.21, "learning_rate": 1.840492515766974e-05, "loss": 1.0485, "step": 4071 }, { "epoch": 0.21, "learning_rate": 1.8404033519441385e-05, "loss": 1.0606, "step": 4072 }, { "epoch": 0.21, "learning_rate": 1.8403141653682765e-05, "loss": 0.9486, "step": 4073 }, { "epoch": 0.21, "learning_rate": 1.8402249560418023e-05, "loss": 0.8952, "step": 4074 }, { "epoch": 0.21, "learning_rate": 1.8401357239671315e-05, "loss": 0.9742, "step": 4075 }, { "epoch": 0.21, "learning_rate": 1.8400464691466794e-05, "loss": 0.9899, "step": 4076 }, { "epoch": 0.21, "learning_rate": 1.8399571915828634e-05, "loss": 0.9141, "step": 4077 }, { "epoch": 0.21, "learning_rate": 1.8398678912780997e-05, "loss": 1.1083, "step": 4078 }, { "epoch": 0.21, "learning_rate": 1.839778568234807e-05, "loss": 1.0069, "step": 4079 }, { "epoch": 0.21, "learning_rate": 1.8396892224554023e-05, "loss": 1.0057, "step": 4080 }, { "epoch": 0.21, "learning_rate": 1.8395998539423055e-05, "loss": 1.0346, "step": 4081 }, { "epoch": 0.21, "learning_rate": 1.8395104626979358e-05, "loss": 1.0406, "step": 4082 }, { "epoch": 0.21, "learning_rate": 1.839421048724714e-05, "loss": 1.0035, "step": 4083 }, { "epoch": 0.21, "learning_rate": 1.83933161202506e-05, "loss": 1.0046, "step": 4084 }, { "epoch": 0.21, "learning_rate": 1.8392421526013956e-05, "loss": 0.9161, "step": 4085 }, { "epoch": 0.21, "learning_rate": 1.8391526704561426e-05, "loss": 1.0662, "step": 4086 }, { "epoch": 0.21, "learning_rate": 1.839063165591724e-05, "loss": 0.92, "step": 4087 }, { "epoch": 0.21, "learning_rate": 1.8389736380105628e-05, "loss": 0.9258, "step": 4088 }, { "epoch": 0.21, "learning_rate": 1.8388840877150832e-05, "loss": 1.0225, "step": 4089 }, { "epoch": 0.21, "learning_rate": 1.8387945147077094e-05, "loss": 0.9963, "step": 4090 }, { "epoch": 0.21, "learning_rate": 1.8387049189908666e-05, "loss": 1.0532, "step": 4091 }, { "epoch": 0.21, "learning_rate": 1.8386153005669803e-05, "loss": 0.9151, "step": 4092 }, { "epoch": 0.21, "learning_rate": 1.8385256594384765e-05, "loss": 0.9438, "step": 4093 }, { "epoch": 0.21, "learning_rate": 1.8384359956077832e-05, "loss": 1.0155, "step": 4094 }, { "epoch": 0.21, "learning_rate": 1.838346309077327e-05, "loss": 1.1545, "step": 4095 }, { "epoch": 0.21, "learning_rate": 1.8382565998495367e-05, "loss": 1.0567, "step": 4096 }, { "epoch": 0.21, "learning_rate": 1.838166867926841e-05, "loss": 1.0672, "step": 4097 }, { "epoch": 0.21, "learning_rate": 1.8380771133116685e-05, "loss": 0.9987, "step": 4098 }, { "epoch": 0.21, "learning_rate": 1.83798733600645e-05, "loss": 0.9957, "step": 4099 }, { "epoch": 0.21, "learning_rate": 1.837897536013616e-05, "loss": 0.9937, "step": 4100 }, { "epoch": 0.21, "learning_rate": 1.8378077133355976e-05, "loss": 1.0679, "step": 4101 }, { "epoch": 0.21, "learning_rate": 1.8377178679748267e-05, "loss": 1.0103, "step": 4102 }, { "epoch": 0.21, "learning_rate": 1.837627999933736e-05, "loss": 0.9554, "step": 4103 }, { "epoch": 0.21, "learning_rate": 1.837538109214758e-05, "loss": 0.999, "step": 4104 }, { "epoch": 0.21, "learning_rate": 1.837448195820327e-05, "loss": 0.8707, "step": 4105 }, { "epoch": 0.21, "learning_rate": 1.837358259752877e-05, "loss": 1.0232, "step": 4106 }, { "epoch": 0.21, "learning_rate": 1.8372683010148432e-05, "loss": 1.0392, "step": 4107 }, { "epoch": 0.21, "learning_rate": 1.837178319608661e-05, "loss": 1.0331, "step": 4108 }, { "epoch": 0.21, "learning_rate": 1.837088315536766e-05, "loss": 0.9547, "step": 4109 }, { "epoch": 0.21, "learning_rate": 1.8369982888015955e-05, "loss": 0.9958, "step": 4110 }, { "epoch": 0.21, "learning_rate": 1.8369082394055872e-05, "loss": 1.1021, "step": 4111 }, { "epoch": 0.21, "learning_rate": 1.8368181673511786e-05, "loss": 0.9607, "step": 4112 }, { "epoch": 0.21, "learning_rate": 1.8367280726408084e-05, "loss": 1.0701, "step": 4113 }, { "epoch": 0.21, "learning_rate": 1.8366379552769158e-05, "loss": 0.9819, "step": 4114 }, { "epoch": 0.21, "learning_rate": 1.8365478152619408e-05, "loss": 0.9973, "step": 4115 }, { "epoch": 0.21, "learning_rate": 1.8364576525983233e-05, "loss": 0.9407, "step": 4116 }, { "epoch": 0.21, "learning_rate": 1.836367467288505e-05, "loss": 1.1398, "step": 4117 }, { "epoch": 0.21, "learning_rate": 1.8362772593349276e-05, "loss": 0.945, "step": 4118 }, { "epoch": 0.21, "learning_rate": 1.836187028740033e-05, "loss": 0.9676, "step": 4119 }, { "epoch": 0.21, "learning_rate": 1.836096775506264e-05, "loss": 0.974, "step": 4120 }, { "epoch": 0.21, "learning_rate": 1.836006499636065e-05, "loss": 0.9136, "step": 4121 }, { "epoch": 0.21, "learning_rate": 1.8359162011318787e-05, "loss": 0.9681, "step": 4122 }, { "epoch": 0.21, "learning_rate": 1.835825879996151e-05, "loss": 0.9507, "step": 4123 }, { "epoch": 0.21, "learning_rate": 1.8357355362313266e-05, "loss": 0.9669, "step": 4124 }, { "epoch": 0.21, "learning_rate": 1.835645169839852e-05, "loss": 1.0123, "step": 4125 }, { "epoch": 0.21, "learning_rate": 1.835554780824173e-05, "loss": 1.0016, "step": 4126 }, { "epoch": 0.21, "learning_rate": 1.8354643691867377e-05, "loss": 1.1105, "step": 4127 }, { "epoch": 0.21, "learning_rate": 1.8353739349299938e-05, "loss": 0.9695, "step": 4128 }, { "epoch": 0.21, "learning_rate": 1.8352834780563888e-05, "loss": 1.0945, "step": 4129 }, { "epoch": 0.21, "learning_rate": 1.8351929985683722e-05, "loss": 0.9553, "step": 4130 }, { "epoch": 0.21, "learning_rate": 1.835102496468394e-05, "loss": 0.868, "step": 4131 }, { "epoch": 0.21, "learning_rate": 1.8350119717589045e-05, "loss": 1.071, "step": 4132 }, { "epoch": 0.21, "learning_rate": 1.834921424442354e-05, "loss": 1.0007, "step": 4133 }, { "epoch": 0.21, "learning_rate": 1.834830854521194e-05, "loss": 0.9834, "step": 4134 }, { "epoch": 0.21, "learning_rate": 1.8347402619978768e-05, "loss": 0.9469, "step": 4135 }, { "epoch": 0.21, "learning_rate": 1.8346496468748555e-05, "loss": 0.9793, "step": 4136 }, { "epoch": 0.21, "learning_rate": 1.8345590091545825e-05, "loss": 0.9911, "step": 4137 }, { "epoch": 0.21, "learning_rate": 1.834468348839513e-05, "loss": 0.9943, "step": 4138 }, { "epoch": 0.21, "learning_rate": 1.8343776659320996e-05, "loss": 0.9789, "step": 4139 }, { "epoch": 0.21, "learning_rate": 1.8342869604347995e-05, "loss": 1.0206, "step": 4140 }, { "epoch": 0.21, "learning_rate": 1.8341962323500674e-05, "loss": 1.007, "step": 4141 }, { "epoch": 0.21, "learning_rate": 1.8341054816803595e-05, "loss": 1.0348, "step": 4142 }, { "epoch": 0.21, "learning_rate": 1.8340147084281336e-05, "loss": 0.975, "step": 4143 }, { "epoch": 0.21, "learning_rate": 1.8339239125958467e-05, "loss": 1.0925, "step": 4144 }, { "epoch": 0.21, "learning_rate": 1.833833094185957e-05, "loss": 0.986, "step": 4145 }, { "epoch": 0.21, "learning_rate": 1.833742253200923e-05, "loss": 0.9993, "step": 4146 }, { "epoch": 0.21, "learning_rate": 1.8336513896432046e-05, "loss": 0.9607, "step": 4147 }, { "epoch": 0.21, "learning_rate": 1.8335605035152624e-05, "loss": 0.9336, "step": 4148 }, { "epoch": 0.21, "learning_rate": 1.833469594819556e-05, "loss": 1.0549, "step": 4149 }, { "epoch": 0.21, "learning_rate": 1.8333786635585472e-05, "loss": 0.8829, "step": 4150 }, { "epoch": 0.21, "learning_rate": 1.833287709734697e-05, "loss": 0.9888, "step": 4151 }, { "epoch": 0.21, "learning_rate": 1.8331967333504698e-05, "loss": 1.0531, "step": 4152 }, { "epoch": 0.21, "learning_rate": 1.8331057344083266e-05, "loss": 1.0281, "step": 4153 }, { "epoch": 0.21, "learning_rate": 1.8330147129107324e-05, "loss": 1.004, "step": 4154 }, { "epoch": 0.21, "learning_rate": 1.8329236688601512e-05, "loss": 1.0461, "step": 4155 }, { "epoch": 0.21, "learning_rate": 1.8328326022590474e-05, "loss": 0.9507, "step": 4156 }, { "epoch": 0.21, "learning_rate": 1.832741513109887e-05, "loss": 1.0557, "step": 4157 }, { "epoch": 0.21, "learning_rate": 1.8326504014151365e-05, "loss": 0.9639, "step": 4158 }, { "epoch": 0.21, "learning_rate": 1.832559267177262e-05, "loss": 0.9579, "step": 4159 }, { "epoch": 0.21, "learning_rate": 1.8324681103987313e-05, "loss": 0.9689, "step": 4160 }, { "epoch": 0.21, "learning_rate": 1.832376931082012e-05, "loss": 1.0112, "step": 4161 }, { "epoch": 0.21, "learning_rate": 1.832285729229573e-05, "loss": 0.9575, "step": 4162 }, { "epoch": 0.21, "learning_rate": 1.832194504843883e-05, "loss": 1.0059, "step": 4163 }, { "epoch": 0.21, "learning_rate": 1.8321032579274127e-05, "loss": 0.9628, "step": 4164 }, { "epoch": 0.21, "learning_rate": 1.8320119884826317e-05, "loss": 0.9322, "step": 4165 }, { "epoch": 0.21, "learning_rate": 1.8319206965120113e-05, "loss": 1.0022, "step": 4166 }, { "epoch": 0.21, "learning_rate": 1.8318293820180232e-05, "loss": 1.0099, "step": 4167 }, { "epoch": 0.21, "learning_rate": 1.8317380450031397e-05, "loss": 0.9999, "step": 4168 }, { "epoch": 0.21, "learning_rate": 1.831646685469833e-05, "loss": 1.0087, "step": 4169 }, { "epoch": 0.21, "learning_rate": 1.831555303420578e-05, "loss": 0.9333, "step": 4170 }, { "epoch": 0.21, "learning_rate": 1.831463898857847e-05, "loss": 1.056, "step": 4171 }, { "epoch": 0.21, "learning_rate": 1.831372471784116e-05, "loss": 0.9382, "step": 4172 }, { "epoch": 0.21, "learning_rate": 1.8312810222018597e-05, "loss": 0.9558, "step": 4173 }, { "epoch": 0.21, "learning_rate": 1.831189550113554e-05, "loss": 1.0471, "step": 4174 }, { "epoch": 0.21, "learning_rate": 1.8310980555216758e-05, "loss": 0.991, "step": 4175 }, { "epoch": 0.21, "learning_rate": 1.831006538428702e-05, "loss": 0.9755, "step": 4176 }, { "epoch": 0.21, "learning_rate": 1.83091499883711e-05, "loss": 1.1094, "step": 4177 }, { "epoch": 0.21, "learning_rate": 1.8308234367493787e-05, "loss": 0.9767, "step": 4178 }, { "epoch": 0.21, "learning_rate": 1.8307318521679867e-05, "loss": 0.9734, "step": 4179 }, { "epoch": 0.21, "learning_rate": 1.8306402450954138e-05, "loss": 1.0145, "step": 4180 }, { "epoch": 0.21, "learning_rate": 1.8305486155341395e-05, "loss": 1.0076, "step": 4181 }, { "epoch": 0.21, "learning_rate": 1.8304569634866457e-05, "loss": 0.9747, "step": 4182 }, { "epoch": 0.21, "learning_rate": 1.8303652889554128e-05, "loss": 0.9374, "step": 4183 }, { "epoch": 0.21, "learning_rate": 1.830273591942923e-05, "loss": 1.0665, "step": 4184 }, { "epoch": 0.21, "learning_rate": 1.830181872451659e-05, "loss": 1.0274, "step": 4185 }, { "epoch": 0.21, "learning_rate": 1.8300901304841044e-05, "loss": 0.9309, "step": 4186 }, { "epoch": 0.21, "learning_rate": 1.8299983660427426e-05, "loss": 1.0804, "step": 4187 }, { "epoch": 0.21, "learning_rate": 1.8299065791300578e-05, "loss": 0.8957, "step": 4188 }, { "epoch": 0.21, "learning_rate": 1.8298147697485356e-05, "loss": 1.1127, "step": 4189 }, { "epoch": 0.21, "learning_rate": 1.8297229379006614e-05, "loss": 0.9885, "step": 4190 }, { "epoch": 0.21, "learning_rate": 1.8296310835889213e-05, "loss": 1.0942, "step": 4191 }, { "epoch": 0.21, "learning_rate": 1.8295392068158023e-05, "loss": 1.1199, "step": 4192 }, { "epoch": 0.21, "learning_rate": 1.8294473075837914e-05, "loss": 1.0724, "step": 4193 }, { "epoch": 0.21, "learning_rate": 1.8293553858953776e-05, "loss": 0.9824, "step": 4194 }, { "epoch": 0.21, "learning_rate": 1.829263441753049e-05, "loss": 0.9275, "step": 4195 }, { "epoch": 0.21, "learning_rate": 1.8291714751592952e-05, "loss": 0.9666, "step": 4196 }, { "epoch": 0.21, "learning_rate": 1.8290794861166056e-05, "loss": 0.9312, "step": 4197 }, { "epoch": 0.21, "learning_rate": 1.828987474627471e-05, "loss": 0.9611, "step": 4198 }, { "epoch": 0.21, "learning_rate": 1.8288954406943825e-05, "loss": 0.9688, "step": 4199 }, { "epoch": 0.21, "learning_rate": 1.828803384319832e-05, "loss": 1.0173, "step": 4200 }, { "epoch": 0.21, "learning_rate": 1.8287113055063114e-05, "loss": 0.9703, "step": 4201 }, { "epoch": 0.21, "learning_rate": 1.828619204256314e-05, "loss": 1.0414, "step": 4202 }, { "epoch": 0.21, "learning_rate": 1.828527080572333e-05, "loss": 0.9486, "step": 4203 }, { "epoch": 0.21, "learning_rate": 1.8284349344568633e-05, "loss": 0.9655, "step": 4204 }, { "epoch": 0.21, "learning_rate": 1.828342765912399e-05, "loss": 0.8983, "step": 4205 }, { "epoch": 0.21, "learning_rate": 1.8282505749414352e-05, "loss": 1.0131, "step": 4206 }, { "epoch": 0.21, "learning_rate": 1.8281583615464685e-05, "loss": 0.9665, "step": 4207 }, { "epoch": 0.21, "learning_rate": 1.8280661257299952e-05, "loss": 1.0147, "step": 4208 }, { "epoch": 0.21, "learning_rate": 1.827973867494513e-05, "loss": 1.0188, "step": 4209 }, { "epoch": 0.21, "learning_rate": 1.8278815868425184e-05, "loss": 1.0129, "step": 4210 }, { "epoch": 0.21, "learning_rate": 1.827789283776511e-05, "loss": 0.9627, "step": 4211 }, { "epoch": 0.21, "learning_rate": 1.8276969582989897e-05, "loss": 1.0361, "step": 4212 }, { "epoch": 0.21, "learning_rate": 1.8276046104124535e-05, "loss": 1.0059, "step": 4213 }, { "epoch": 0.21, "learning_rate": 1.8275122401194032e-05, "loss": 0.9778, "step": 4214 }, { "epoch": 0.21, "learning_rate": 1.8274198474223396e-05, "loss": 0.947, "step": 4215 }, { "epoch": 0.21, "learning_rate": 1.8273274323237634e-05, "loss": 1.0813, "step": 4216 }, { "epoch": 0.21, "learning_rate": 1.8272349948261775e-05, "loss": 0.9654, "step": 4217 }, { "epoch": 0.21, "learning_rate": 1.827142534932084e-05, "loss": 0.9649, "step": 4218 }, { "epoch": 0.21, "learning_rate": 1.8270500526439868e-05, "loss": 0.9541, "step": 4219 }, { "epoch": 0.21, "learning_rate": 1.826957547964389e-05, "loss": 1.0425, "step": 4220 }, { "epoch": 0.21, "learning_rate": 1.8268650208957956e-05, "loss": 0.9799, "step": 4221 }, { "epoch": 0.21, "learning_rate": 1.826772471440711e-05, "loss": 1.0219, "step": 4222 }, { "epoch": 0.21, "learning_rate": 1.826679899601642e-05, "loss": 0.9525, "step": 4223 }, { "epoch": 0.21, "learning_rate": 1.826587305381094e-05, "loss": 0.9331, "step": 4224 }, { "epoch": 0.21, "learning_rate": 1.8264946887815738e-05, "loss": 1.0547, "step": 4225 }, { "epoch": 0.21, "learning_rate": 1.82640204980559e-05, "loss": 0.9492, "step": 4226 }, { "epoch": 0.21, "learning_rate": 1.8263093884556492e-05, "loss": 1.0305, "step": 4227 }, { "epoch": 0.21, "learning_rate": 1.826216704734261e-05, "loss": 0.9533, "step": 4228 }, { "epoch": 0.21, "learning_rate": 1.8261239986439345e-05, "loss": 0.9584, "step": 4229 }, { "epoch": 0.21, "learning_rate": 1.82603127018718e-05, "loss": 1.0089, "step": 4230 }, { "epoch": 0.21, "learning_rate": 1.825938519366507e-05, "loss": 0.9794, "step": 4231 }, { "epoch": 0.21, "learning_rate": 1.8258457461844278e-05, "loss": 0.9808, "step": 4232 }, { "epoch": 0.22, "learning_rate": 1.825752950643454e-05, "loss": 0.9847, "step": 4233 }, { "epoch": 0.22, "learning_rate": 1.825660132746097e-05, "loss": 1.0118, "step": 4234 }, { "epoch": 0.22, "learning_rate": 1.8255672924948705e-05, "loss": 0.9593, "step": 4235 }, { "epoch": 0.22, "learning_rate": 1.825474429892288e-05, "loss": 0.9169, "step": 4236 }, { "epoch": 0.22, "learning_rate": 1.8253815449408637e-05, "loss": 1.0731, "step": 4237 }, { "epoch": 0.22, "learning_rate": 1.825288637643112e-05, "loss": 1.0547, "step": 4238 }, { "epoch": 0.22, "learning_rate": 1.8251957080015482e-05, "loss": 1.033, "step": 4239 }, { "epoch": 0.22, "learning_rate": 1.825102756018689e-05, "loss": 1.0638, "step": 4240 }, { "epoch": 0.22, "learning_rate": 1.8250097816970503e-05, "loss": 0.9453, "step": 4241 }, { "epoch": 0.22, "learning_rate": 1.8249167850391495e-05, "loss": 1.0105, "step": 4242 }, { "epoch": 0.22, "learning_rate": 1.8248237660475046e-05, "loss": 1.1346, "step": 4243 }, { "epoch": 0.22, "learning_rate": 1.8247307247246335e-05, "loss": 1.0121, "step": 4244 }, { "epoch": 0.22, "learning_rate": 1.8246376610730555e-05, "loss": 0.949, "step": 4245 }, { "epoch": 0.22, "learning_rate": 1.8245445750952902e-05, "loss": 1.0351, "step": 4246 }, { "epoch": 0.22, "learning_rate": 1.8244514667938577e-05, "loss": 0.982, "step": 4247 }, { "epoch": 0.22, "learning_rate": 1.8243583361712792e-05, "loss": 1.0804, "step": 4248 }, { "epoch": 0.22, "learning_rate": 1.8242651832300757e-05, "loss": 0.9839, "step": 4249 }, { "epoch": 0.22, "learning_rate": 1.824172007972769e-05, "loss": 1.06, "step": 4250 }, { "epoch": 0.22, "learning_rate": 1.8240788104018824e-05, "loss": 1.0463, "step": 4251 }, { "epoch": 0.22, "learning_rate": 1.8239855905199386e-05, "loss": 1.0258, "step": 4252 }, { "epoch": 0.22, "learning_rate": 1.8238923483294616e-05, "loss": 0.9684, "step": 4253 }, { "epoch": 0.22, "learning_rate": 1.823799083832976e-05, "loss": 1.0289, "step": 4254 }, { "epoch": 0.22, "learning_rate": 1.8237057970330063e-05, "loss": 0.9932, "step": 4255 }, { "epoch": 0.22, "learning_rate": 1.8236124879320788e-05, "loss": 0.992, "step": 4256 }, { "epoch": 0.22, "learning_rate": 1.8235191565327194e-05, "loss": 0.954, "step": 4257 }, { "epoch": 0.22, "learning_rate": 1.8234258028374548e-05, "loss": 1.1375, "step": 4258 }, { "epoch": 0.22, "learning_rate": 1.8233324268488126e-05, "loss": 0.988, "step": 4259 }, { "epoch": 0.22, "learning_rate": 1.823239028569321e-05, "loss": 1.0453, "step": 4260 }, { "epoch": 0.22, "learning_rate": 1.8231456080015085e-05, "loss": 0.9991, "step": 4261 }, { "epoch": 0.22, "learning_rate": 1.823052165147905e-05, "loss": 1.0584, "step": 4262 }, { "epoch": 0.22, "learning_rate": 1.822958700011039e-05, "loss": 1.0207, "step": 4263 }, { "epoch": 0.22, "learning_rate": 1.822865212593442e-05, "loss": 1.0055, "step": 4264 }, { "epoch": 0.22, "learning_rate": 1.822771702897645e-05, "loss": 0.9555, "step": 4265 }, { "epoch": 0.22, "learning_rate": 1.8226781709261793e-05, "loss": 1.069, "step": 4266 }, { "epoch": 0.22, "learning_rate": 1.8225846166815777e-05, "loss": 0.9541, "step": 4267 }, { "epoch": 0.22, "learning_rate": 1.8224910401663725e-05, "loss": 0.9648, "step": 4268 }, { "epoch": 0.22, "learning_rate": 1.8223974413830976e-05, "loss": 0.9968, "step": 4269 }, { "epoch": 0.22, "learning_rate": 1.822303820334287e-05, "loss": 0.9854, "step": 4270 }, { "epoch": 0.22, "learning_rate": 1.822210177022475e-05, "loss": 0.9556, "step": 4271 }, { "epoch": 0.22, "learning_rate": 1.8221165114501973e-05, "loss": 0.9837, "step": 4272 }, { "epoch": 0.22, "learning_rate": 1.82202282361999e-05, "loss": 1.0863, "step": 4273 }, { "epoch": 0.22, "learning_rate": 1.821929113534389e-05, "loss": 0.9154, "step": 4274 }, { "epoch": 0.22, "learning_rate": 1.821835381195932e-05, "loss": 1.0076, "step": 4275 }, { "epoch": 0.22, "learning_rate": 1.821741626607156e-05, "loss": 0.9825, "step": 4276 }, { "epoch": 0.22, "learning_rate": 1.8216478497706003e-05, "loss": 0.9963, "step": 4277 }, { "epoch": 0.22, "learning_rate": 1.8215540506888027e-05, "loss": 0.9753, "step": 4278 }, { "epoch": 0.22, "learning_rate": 1.8214602293643035e-05, "loss": 1.0643, "step": 4279 }, { "epoch": 0.22, "learning_rate": 1.8213663857996424e-05, "loss": 0.9579, "step": 4280 }, { "epoch": 0.22, "learning_rate": 1.8212725199973604e-05, "loss": 0.9738, "step": 4281 }, { "epoch": 0.22, "learning_rate": 1.8211786319599987e-05, "loss": 0.9461, "step": 4282 }, { "epoch": 0.22, "learning_rate": 1.8210847216900994e-05, "loss": 0.9912, "step": 4283 }, { "epoch": 0.22, "learning_rate": 1.8209907891902047e-05, "loss": 1.0093, "step": 4284 }, { "epoch": 0.22, "learning_rate": 1.8208968344628576e-05, "loss": 1.0315, "step": 4285 }, { "epoch": 0.22, "learning_rate": 1.820802857510602e-05, "loss": 0.942, "step": 4286 }, { "epoch": 0.22, "learning_rate": 1.8207088583359825e-05, "loss": 1.0564, "step": 4287 }, { "epoch": 0.22, "learning_rate": 1.820614836941544e-05, "loss": 1.0471, "step": 4288 }, { "epoch": 0.22, "learning_rate": 1.8205207933298315e-05, "loss": 0.9853, "step": 4289 }, { "epoch": 0.22, "learning_rate": 1.8204267275033916e-05, "loss": 1.126, "step": 4290 }, { "epoch": 0.22, "learning_rate": 1.820332639464771e-05, "loss": 0.9793, "step": 4291 }, { "epoch": 0.22, "learning_rate": 1.8202385292165173e-05, "loss": 0.996, "step": 4292 }, { "epoch": 0.22, "learning_rate": 1.8201443967611776e-05, "loss": 1.0111, "step": 4293 }, { "epoch": 0.22, "learning_rate": 1.8200502421013012e-05, "loss": 0.9865, "step": 4294 }, { "epoch": 0.22, "learning_rate": 1.8199560652394366e-05, "loss": 0.9861, "step": 4295 }, { "epoch": 0.22, "learning_rate": 1.819861866178134e-05, "loss": 0.9327, "step": 4296 }, { "epoch": 0.22, "learning_rate": 1.819767644919944e-05, "loss": 1.0286, "step": 4297 }, { "epoch": 0.22, "learning_rate": 1.819673401467417e-05, "loss": 0.991, "step": 4298 }, { "epoch": 0.22, "learning_rate": 1.8195791358231045e-05, "loss": 1.0193, "step": 4299 }, { "epoch": 0.22, "learning_rate": 1.8194848479895592e-05, "loss": 1.0373, "step": 4300 }, { "epoch": 0.22, "learning_rate": 1.819390537969333e-05, "loss": 0.9627, "step": 4301 }, { "epoch": 0.22, "learning_rate": 1.8192962057649804e-05, "loss": 0.9703, "step": 4302 }, { "epoch": 0.22, "learning_rate": 1.8192018513790542e-05, "loss": 1.0346, "step": 4303 }, { "epoch": 0.22, "learning_rate": 1.8191074748141097e-05, "loss": 1.0593, "step": 4304 }, { "epoch": 0.22, "learning_rate": 1.8190130760727015e-05, "loss": 0.9244, "step": 4305 }, { "epoch": 0.22, "learning_rate": 1.818918655157386e-05, "loss": 0.9608, "step": 4306 }, { "epoch": 0.22, "learning_rate": 1.818824212070719e-05, "loss": 0.9802, "step": 4307 }, { "epoch": 0.22, "learning_rate": 1.8187297468152575e-05, "loss": 1.0066, "step": 4308 }, { "epoch": 0.22, "learning_rate": 1.818635259393559e-05, "loss": 1.0381, "step": 4309 }, { "epoch": 0.22, "learning_rate": 1.8185407498081824e-05, "loss": 1.0409, "step": 4310 }, { "epoch": 0.22, "learning_rate": 1.8184462180616855e-05, "loss": 1.0134, "step": 4311 }, { "epoch": 0.22, "learning_rate": 1.8183516641566278e-05, "loss": 1.0107, "step": 4312 }, { "epoch": 0.22, "learning_rate": 1.8182570880955698e-05, "loss": 0.9557, "step": 4313 }, { "epoch": 0.22, "learning_rate": 1.8181624898810717e-05, "loss": 0.8752, "step": 4314 }, { "epoch": 0.22, "learning_rate": 1.8180678695156945e-05, "loss": 1.0258, "step": 4315 }, { "epoch": 0.22, "learning_rate": 1.817973227002e-05, "loss": 0.9153, "step": 4316 }, { "epoch": 0.22, "learning_rate": 1.8178785623425507e-05, "loss": 0.9926, "step": 4317 }, { "epoch": 0.22, "learning_rate": 1.8177838755399095e-05, "loss": 1.0172, "step": 4318 }, { "epoch": 0.22, "learning_rate": 1.81768916659664e-05, "loss": 1.0042, "step": 4319 }, { "epoch": 0.22, "learning_rate": 1.8175944355153065e-05, "loss": 0.9657, "step": 4320 }, { "epoch": 0.22, "learning_rate": 1.817499682298473e-05, "loss": 0.9479, "step": 4321 }, { "epoch": 0.22, "learning_rate": 1.8174049069487056e-05, "loss": 0.9634, "step": 4322 }, { "epoch": 0.22, "learning_rate": 1.81731010946857e-05, "loss": 0.9375, "step": 4323 }, { "epoch": 0.22, "learning_rate": 1.8172152898606324e-05, "loss": 0.9741, "step": 4324 }, { "epoch": 0.22, "learning_rate": 1.8171204481274607e-05, "loss": 0.9517, "step": 4325 }, { "epoch": 0.22, "learning_rate": 1.817025584271622e-05, "loss": 0.9638, "step": 4326 }, { "epoch": 0.22, "learning_rate": 1.816930698295685e-05, "loss": 1.0106, "step": 4327 }, { "epoch": 0.22, "learning_rate": 1.8168357902022187e-05, "loss": 0.9988, "step": 4328 }, { "epoch": 0.22, "learning_rate": 1.8167408599937922e-05, "loss": 0.9849, "step": 4329 }, { "epoch": 0.22, "learning_rate": 1.8166459076729758e-05, "loss": 1.0116, "step": 4330 }, { "epoch": 0.22, "learning_rate": 1.81655093324234e-05, "loss": 0.9357, "step": 4331 }, { "epoch": 0.22, "learning_rate": 1.816455936704457e-05, "loss": 0.9328, "step": 4332 }, { "epoch": 0.22, "learning_rate": 1.816360918061898e-05, "loss": 0.8677, "step": 4333 }, { "epoch": 0.22, "learning_rate": 1.8162658773172355e-05, "loss": 0.9871, "step": 4334 }, { "epoch": 0.22, "learning_rate": 1.816170814473043e-05, "loss": 1.0198, "step": 4335 }, { "epoch": 0.22, "learning_rate": 1.8160757295318938e-05, "loss": 1.0436, "step": 4336 }, { "epoch": 0.22, "learning_rate": 1.8159806224963624e-05, "loss": 1.0267, "step": 4337 }, { "epoch": 0.22, "learning_rate": 1.8158854933690243e-05, "loss": 0.9349, "step": 4338 }, { "epoch": 0.22, "learning_rate": 1.815790342152454e-05, "loss": 0.9836, "step": 4339 }, { "epoch": 0.22, "learning_rate": 1.8156951688492285e-05, "loss": 0.979, "step": 4340 }, { "epoch": 0.22, "learning_rate": 1.8155999734619238e-05, "loss": 0.9323, "step": 4341 }, { "epoch": 0.22, "learning_rate": 1.8155047559931177e-05, "loss": 0.9539, "step": 4342 }, { "epoch": 0.22, "learning_rate": 1.815409516445388e-05, "loss": 1.0812, "step": 4343 }, { "epoch": 0.22, "learning_rate": 1.815314254821313e-05, "loss": 0.9962, "step": 4344 }, { "epoch": 0.22, "learning_rate": 1.815218971123472e-05, "loss": 0.986, "step": 4345 }, { "epoch": 0.22, "learning_rate": 1.8151236653544447e-05, "loss": 0.8916, "step": 4346 }, { "epoch": 0.22, "learning_rate": 1.8150283375168112e-05, "loss": 0.9946, "step": 4347 }, { "epoch": 0.22, "learning_rate": 1.8149329876131533e-05, "loss": 1.0684, "step": 4348 }, { "epoch": 0.22, "learning_rate": 1.814837615646051e-05, "loss": 0.9802, "step": 4349 }, { "epoch": 0.22, "learning_rate": 1.8147422216180876e-05, "loss": 1.0392, "step": 4350 }, { "epoch": 0.22, "learning_rate": 1.814646805531845e-05, "loss": 0.9082, "step": 4351 }, { "epoch": 0.22, "learning_rate": 1.8145513673899074e-05, "loss": 1.0179, "step": 4352 }, { "epoch": 0.22, "learning_rate": 1.8144559071948578e-05, "loss": 0.9824, "step": 4353 }, { "epoch": 0.22, "learning_rate": 1.814360424949281e-05, "loss": 1.0268, "step": 4354 }, { "epoch": 0.22, "learning_rate": 1.8142649206557623e-05, "loss": 1.0305, "step": 4355 }, { "epoch": 0.22, "learning_rate": 1.814169394316887e-05, "loss": 1.0549, "step": 4356 }, { "epoch": 0.22, "learning_rate": 1.8140738459352416e-05, "loss": 0.9891, "step": 4357 }, { "epoch": 0.22, "learning_rate": 1.813978275513413e-05, "loss": 1.0009, "step": 4358 }, { "epoch": 0.22, "learning_rate": 1.8138826830539885e-05, "loss": 1.0157, "step": 4359 }, { "epoch": 0.22, "learning_rate": 1.8137870685595565e-05, "loss": 1.0285, "step": 4360 }, { "epoch": 0.22, "learning_rate": 1.813691432032705e-05, "loss": 1.0789, "step": 4361 }, { "epoch": 0.22, "learning_rate": 1.813595773476024e-05, "loss": 0.9305, "step": 4362 }, { "epoch": 0.22, "learning_rate": 1.8135000928921028e-05, "loss": 0.9237, "step": 4363 }, { "epoch": 0.22, "learning_rate": 1.8134043902835322e-05, "loss": 1.0552, "step": 4364 }, { "epoch": 0.22, "learning_rate": 1.8133086656529033e-05, "loss": 1.0067, "step": 4365 }, { "epoch": 0.22, "learning_rate": 1.8132129190028072e-05, "loss": 1.0221, "step": 4366 }, { "epoch": 0.22, "learning_rate": 1.813117150335837e-05, "loss": 1.038, "step": 4367 }, { "epoch": 0.22, "learning_rate": 1.8130213596545844e-05, "loss": 0.9722, "step": 4368 }, { "epoch": 0.22, "learning_rate": 1.812925546961644e-05, "loss": 0.9865, "step": 4369 }, { "epoch": 0.22, "learning_rate": 1.8128297122596092e-05, "loss": 0.9744, "step": 4370 }, { "epoch": 0.22, "learning_rate": 1.812733855551075e-05, "loss": 0.9199, "step": 4371 }, { "epoch": 0.22, "learning_rate": 1.8126379768386354e-05, "loss": 0.9442, "step": 4372 }, { "epoch": 0.22, "learning_rate": 1.8125420761248878e-05, "loss": 0.9272, "step": 4373 }, { "epoch": 0.22, "learning_rate": 1.812446153412428e-05, "loss": 1.0259, "step": 4374 }, { "epoch": 0.22, "learning_rate": 1.812350208703853e-05, "loss": 0.9643, "step": 4375 }, { "epoch": 0.22, "learning_rate": 1.8122542420017604e-05, "loss": 1.0375, "step": 4376 }, { "epoch": 0.22, "learning_rate": 1.812158253308748e-05, "loss": 1.0615, "step": 4377 }, { "epoch": 0.22, "learning_rate": 1.8120622426274153e-05, "loss": 0.8908, "step": 4378 }, { "epoch": 0.22, "learning_rate": 1.8119662099603612e-05, "loss": 0.9687, "step": 4379 }, { "epoch": 0.22, "learning_rate": 1.8118701553101858e-05, "loss": 0.9755, "step": 4380 }, { "epoch": 0.22, "learning_rate": 1.81177407867949e-05, "loss": 1.0164, "step": 4381 }, { "epoch": 0.22, "learning_rate": 1.8116779800708744e-05, "loss": 0.9028, "step": 4382 }, { "epoch": 0.22, "learning_rate": 1.811581859486941e-05, "loss": 0.911, "step": 4383 }, { "epoch": 0.22, "learning_rate": 1.8114857169302923e-05, "loss": 0.9094, "step": 4384 }, { "epoch": 0.22, "learning_rate": 1.8113895524035313e-05, "loss": 1.0074, "step": 4385 }, { "epoch": 0.22, "learning_rate": 1.8112933659092614e-05, "loss": 0.9573, "step": 4386 }, { "epoch": 0.22, "learning_rate": 1.8111971574500865e-05, "loss": 0.9449, "step": 4387 }, { "epoch": 0.22, "learning_rate": 1.8111009270286117e-05, "loss": 1.0068, "step": 4388 }, { "epoch": 0.22, "learning_rate": 1.8110046746474423e-05, "loss": 1.0761, "step": 4389 }, { "epoch": 0.22, "learning_rate": 1.8109084003091842e-05, "loss": 0.9437, "step": 4390 }, { "epoch": 0.22, "learning_rate": 1.810812104016444e-05, "loss": 1.1594, "step": 4391 }, { "epoch": 0.22, "learning_rate": 1.8107157857718286e-05, "loss": 0.9595, "step": 4392 }, { "epoch": 0.22, "learning_rate": 1.8106194455779457e-05, "loss": 0.9723, "step": 4393 }, { "epoch": 0.22, "learning_rate": 1.8105230834374043e-05, "loss": 0.8755, "step": 4394 }, { "epoch": 0.22, "learning_rate": 1.8104266993528118e-05, "loss": 0.947, "step": 4395 }, { "epoch": 0.22, "learning_rate": 1.8103302933267795e-05, "loss": 0.9549, "step": 4396 }, { "epoch": 0.22, "learning_rate": 1.8102338653619162e-05, "loss": 1.0834, "step": 4397 }, { "epoch": 0.22, "learning_rate": 1.8101374154608332e-05, "loss": 1.0544, "step": 4398 }, { "epoch": 0.22, "learning_rate": 1.810040943626141e-05, "loss": 0.8943, "step": 4399 }, { "epoch": 0.22, "learning_rate": 1.809944449860453e-05, "loss": 1.0264, "step": 4400 }, { "epoch": 0.22, "learning_rate": 1.8098479341663803e-05, "loss": 1.0773, "step": 4401 }, { "epoch": 0.22, "learning_rate": 1.8097513965465365e-05, "loss": 0.9625, "step": 4402 }, { "epoch": 0.22, "learning_rate": 1.809654837003535e-05, "loss": 1.0846, "step": 4403 }, { "epoch": 0.22, "learning_rate": 1.80955825553999e-05, "loss": 1.0723, "step": 4404 }, { "epoch": 0.22, "learning_rate": 1.8094616521585167e-05, "loss": 0.9694, "step": 4405 }, { "epoch": 0.22, "learning_rate": 1.8093650268617308e-05, "loss": 1.0126, "step": 4406 }, { "epoch": 0.22, "learning_rate": 1.8092683796522476e-05, "loss": 0.8742, "step": 4407 }, { "epoch": 0.22, "learning_rate": 1.8091717105326837e-05, "loss": 1.0173, "step": 4408 }, { "epoch": 0.22, "learning_rate": 1.809075019505657e-05, "loss": 1.0021, "step": 4409 }, { "epoch": 0.22, "learning_rate": 1.808978306573785e-05, "loss": 0.9655, "step": 4410 }, { "epoch": 0.22, "learning_rate": 1.808881571739686e-05, "loss": 0.9659, "step": 4411 }, { "epoch": 0.22, "learning_rate": 1.8087848150059786e-05, "loss": 1.0394, "step": 4412 }, { "epoch": 0.22, "learning_rate": 1.808688036375283e-05, "loss": 0.9059, "step": 4413 }, { "epoch": 0.22, "learning_rate": 1.8085912358502194e-05, "loss": 1.1204, "step": 4414 }, { "epoch": 0.22, "learning_rate": 1.8084944134334083e-05, "loss": 1.0506, "step": 4415 }, { "epoch": 0.22, "learning_rate": 1.8083975691274708e-05, "loss": 0.936, "step": 4416 }, { "epoch": 0.22, "learning_rate": 1.8083007029350297e-05, "loss": 0.9983, "step": 4417 }, { "epoch": 0.22, "learning_rate": 1.8082038148587066e-05, "loss": 0.9908, "step": 4418 }, { "epoch": 0.22, "learning_rate": 1.808106904901125e-05, "loss": 0.9864, "step": 4419 }, { "epoch": 0.22, "learning_rate": 1.8080099730649093e-05, "loss": 0.9442, "step": 4420 }, { "epoch": 0.22, "learning_rate": 1.807913019352683e-05, "loss": 0.9716, "step": 4421 }, { "epoch": 0.22, "learning_rate": 1.8078160437670712e-05, "loss": 0.987, "step": 4422 }, { "epoch": 0.22, "learning_rate": 1.8077190463106996e-05, "loss": 1.0299, "step": 4423 }, { "epoch": 0.22, "learning_rate": 1.807622026986194e-05, "loss": 0.9996, "step": 4424 }, { "epoch": 0.22, "learning_rate": 1.807524985796181e-05, "loss": 0.936, "step": 4425 }, { "epoch": 0.22, "learning_rate": 1.8074279227432883e-05, "loss": 0.915, "step": 4426 }, { "epoch": 0.22, "learning_rate": 1.807330837830144e-05, "loss": 0.97, "step": 4427 }, { "epoch": 0.22, "learning_rate": 1.807233731059376e-05, "loss": 1.0432, "step": 4428 }, { "epoch": 0.23, "learning_rate": 1.8071366024336132e-05, "loss": 0.9558, "step": 4429 }, { "epoch": 0.23, "learning_rate": 1.807039451955486e-05, "loss": 1.0169, "step": 4430 }, { "epoch": 0.23, "learning_rate": 1.806942279627624e-05, "loss": 0.944, "step": 4431 }, { "epoch": 0.23, "learning_rate": 1.8068450854526585e-05, "loss": 0.9952, "step": 4432 }, { "epoch": 0.23, "learning_rate": 1.8067478694332208e-05, "loss": 1.0807, "step": 4433 }, { "epoch": 0.23, "learning_rate": 1.806650631571943e-05, "loss": 0.9998, "step": 4434 }, { "epoch": 0.23, "learning_rate": 1.806553371871457e-05, "loss": 0.9526, "step": 4435 }, { "epoch": 0.23, "learning_rate": 1.806456090334397e-05, "loss": 1.0111, "step": 4436 }, { "epoch": 0.23, "learning_rate": 1.806358786963396e-05, "loss": 1.1018, "step": 4437 }, { "epoch": 0.23, "learning_rate": 1.8062614617610892e-05, "loss": 0.9554, "step": 4438 }, { "epoch": 0.23, "learning_rate": 1.806164114730111e-05, "loss": 0.9844, "step": 4439 }, { "epoch": 0.23, "learning_rate": 1.806066745873097e-05, "loss": 0.9709, "step": 4440 }, { "epoch": 0.23, "learning_rate": 1.805969355192683e-05, "loss": 1.1011, "step": 4441 }, { "epoch": 0.23, "learning_rate": 1.8058719426915067e-05, "loss": 1.0468, "step": 4442 }, { "epoch": 0.23, "learning_rate": 1.8057745083722046e-05, "loss": 0.9921, "step": 4443 }, { "epoch": 0.23, "learning_rate": 1.805677052237415e-05, "loss": 0.9691, "step": 4444 }, { "epoch": 0.23, "learning_rate": 1.805579574289777e-05, "loss": 0.9658, "step": 4445 }, { "epoch": 0.23, "learning_rate": 1.8054820745319283e-05, "loss": 1.0828, "step": 4446 }, { "epoch": 0.23, "learning_rate": 1.80538455296651e-05, "loss": 1.0442, "step": 4447 }, { "epoch": 0.23, "learning_rate": 1.8052870095961615e-05, "loss": 0.9304, "step": 4448 }, { "epoch": 0.23, "learning_rate": 1.805189444423524e-05, "loss": 0.9677, "step": 4449 }, { "epoch": 0.23, "learning_rate": 1.805091857451239e-05, "loss": 0.9294, "step": 4450 }, { "epoch": 0.23, "learning_rate": 1.8049942486819484e-05, "loss": 0.9969, "step": 4451 }, { "epoch": 0.23, "learning_rate": 1.8048966181182953e-05, "loss": 1.0325, "step": 4452 }, { "epoch": 0.23, "learning_rate": 1.8047989657629222e-05, "loss": 0.9797, "step": 4453 }, { "epoch": 0.23, "learning_rate": 1.8047012916184735e-05, "loss": 1.0535, "step": 4454 }, { "epoch": 0.23, "learning_rate": 1.8046035956875935e-05, "loss": 0.9984, "step": 4455 }, { "epoch": 0.23, "learning_rate": 1.804505877972927e-05, "loss": 0.9662, "step": 4456 }, { "epoch": 0.23, "learning_rate": 1.80440813847712e-05, "loss": 0.9566, "step": 4457 }, { "epoch": 0.23, "learning_rate": 1.8043103772028185e-05, "loss": 1.0151, "step": 4458 }, { "epoch": 0.23, "learning_rate": 1.8042125941526692e-05, "loss": 0.9973, "step": 4459 }, { "epoch": 0.23, "learning_rate": 1.8041147893293192e-05, "loss": 0.9097, "step": 4460 }, { "epoch": 0.23, "learning_rate": 1.8040169627354172e-05, "loss": 1.0281, "step": 4461 }, { "epoch": 0.23, "learning_rate": 1.803919114373611e-05, "loss": 0.9955, "step": 4462 }, { "epoch": 0.23, "learning_rate": 1.8038212442465504e-05, "loss": 0.9729, "step": 4463 }, { "epoch": 0.23, "learning_rate": 1.803723352356885e-05, "loss": 1.0187, "step": 4464 }, { "epoch": 0.23, "learning_rate": 1.8036254387072643e-05, "loss": 0.9446, "step": 4465 }, { "epoch": 0.23, "learning_rate": 1.80352750330034e-05, "loss": 1.0137, "step": 4466 }, { "epoch": 0.23, "learning_rate": 1.803429546138764e-05, "loss": 1.0043, "step": 4467 }, { "epoch": 0.23, "learning_rate": 1.8033315672251873e-05, "loss": 1.0222, "step": 4468 }, { "epoch": 0.23, "learning_rate": 1.803233566562263e-05, "loss": 1.0232, "step": 4469 }, { "epoch": 0.23, "learning_rate": 1.803135544152645e-05, "loss": 1.0167, "step": 4470 }, { "epoch": 0.23, "learning_rate": 1.803037499998986e-05, "loss": 1.026, "step": 4471 }, { "epoch": 0.23, "learning_rate": 1.8029394341039416e-05, "loss": 1.0133, "step": 4472 }, { "epoch": 0.23, "learning_rate": 1.802841346470166e-05, "loss": 0.979, "step": 4473 }, { "epoch": 0.23, "learning_rate": 1.8027432371003152e-05, "loss": 0.9634, "step": 4474 }, { "epoch": 0.23, "learning_rate": 1.802645105997045e-05, "loss": 0.9577, "step": 4475 }, { "epoch": 0.23, "learning_rate": 1.8025469531630127e-05, "loss": 0.8926, "step": 4476 }, { "epoch": 0.23, "learning_rate": 1.8024487786008754e-05, "loss": 0.967, "step": 4477 }, { "epoch": 0.23, "learning_rate": 1.8023505823132915e-05, "loss": 0.9415, "step": 4478 }, { "epoch": 0.23, "learning_rate": 1.8022523643029187e-05, "loss": 0.9921, "step": 4479 }, { "epoch": 0.23, "learning_rate": 1.8021541245724172e-05, "loss": 0.9956, "step": 4480 }, { "epoch": 0.23, "learning_rate": 1.802055863124446e-05, "loss": 0.9886, "step": 4481 }, { "epoch": 0.23, "learning_rate": 1.8019575799616654e-05, "loss": 0.9388, "step": 4482 }, { "epoch": 0.23, "learning_rate": 1.8018592750867368e-05, "loss": 0.9149, "step": 4483 }, { "epoch": 0.23, "learning_rate": 1.8017609485023212e-05, "loss": 1.0303, "step": 4484 }, { "epoch": 0.23, "learning_rate": 1.8016626002110814e-05, "loss": 0.9325, "step": 4485 }, { "epoch": 0.23, "learning_rate": 1.801564230215679e-05, "loss": 1.0655, "step": 4486 }, { "epoch": 0.23, "learning_rate": 1.8014658385187783e-05, "loss": 1.0076, "step": 4487 }, { "epoch": 0.23, "learning_rate": 1.8013674251230425e-05, "loss": 0.9295, "step": 4488 }, { "epoch": 0.23, "learning_rate": 1.801268990031136e-05, "loss": 1.0219, "step": 4489 }, { "epoch": 0.23, "learning_rate": 1.8011705332457246e-05, "loss": 0.9916, "step": 4490 }, { "epoch": 0.23, "learning_rate": 1.801072054769473e-05, "loss": 1.0232, "step": 4491 }, { "epoch": 0.23, "learning_rate": 1.8009735546050483e-05, "loss": 1.0333, "step": 4492 }, { "epoch": 0.23, "learning_rate": 1.8008750327551164e-05, "loss": 1.1091, "step": 4493 }, { "epoch": 0.23, "learning_rate": 1.8007764892223448e-05, "loss": 0.9451, "step": 4494 }, { "epoch": 0.23, "learning_rate": 1.8006779240094024e-05, "loss": 0.9586, "step": 4495 }, { "epoch": 0.23, "learning_rate": 1.8005793371189564e-05, "loss": 1.0023, "step": 4496 }, { "epoch": 0.23, "learning_rate": 1.8004807285536768e-05, "loss": 1.0318, "step": 4497 }, { "epoch": 0.23, "learning_rate": 1.8003820983162333e-05, "loss": 1.0371, "step": 4498 }, { "epoch": 0.23, "learning_rate": 1.800283446409296e-05, "loss": 1.078, "step": 4499 }, { "epoch": 0.23, "learning_rate": 1.800184772835536e-05, "loss": 1.0526, "step": 4500 }, { "epoch": 0.23, "learning_rate": 1.8000860775976244e-05, "loss": 0.9467, "step": 4501 }, { "epoch": 0.23, "learning_rate": 1.799987360698233e-05, "loss": 0.9596, "step": 4502 }, { "epoch": 0.23, "learning_rate": 1.7998886221400355e-05, "loss": 1.0065, "step": 4503 }, { "epoch": 0.23, "learning_rate": 1.7997898619257043e-05, "loss": 0.936, "step": 4504 }, { "epoch": 0.23, "learning_rate": 1.7996910800579137e-05, "loss": 0.9578, "step": 4505 }, { "epoch": 0.23, "learning_rate": 1.799592276539338e-05, "loss": 0.9808, "step": 4506 }, { "epoch": 0.23, "learning_rate": 1.799493451372652e-05, "loss": 0.9147, "step": 4507 }, { "epoch": 0.23, "learning_rate": 1.799394604560531e-05, "loss": 0.9824, "step": 4508 }, { "epoch": 0.23, "learning_rate": 1.799295736105652e-05, "loss": 1.0145, "step": 4509 }, { "epoch": 0.23, "learning_rate": 1.7991968460106914e-05, "loss": 0.9583, "step": 4510 }, { "epoch": 0.23, "learning_rate": 1.799097934278326e-05, "loss": 0.9936, "step": 4511 }, { "epoch": 0.23, "learning_rate": 1.7989990009112348e-05, "loss": 1.0486, "step": 4512 }, { "epoch": 0.23, "learning_rate": 1.7989000459120955e-05, "loss": 0.9504, "step": 4513 }, { "epoch": 0.23, "learning_rate": 1.798801069283587e-05, "loss": 0.9075, "step": 4514 }, { "epoch": 0.23, "learning_rate": 1.79870207102839e-05, "loss": 0.9254, "step": 4515 }, { "epoch": 0.23, "learning_rate": 1.798603051149184e-05, "loss": 0.9854, "step": 4516 }, { "epoch": 0.23, "learning_rate": 1.79850400964865e-05, "loss": 0.9269, "step": 4517 }, { "epoch": 0.23, "learning_rate": 1.7984049465294693e-05, "loss": 0.9435, "step": 4518 }, { "epoch": 0.23, "learning_rate": 1.7983058617943243e-05, "loss": 0.933, "step": 4519 }, { "epoch": 0.23, "learning_rate": 1.7982067554458977e-05, "loss": 0.9655, "step": 4520 }, { "epoch": 0.23, "learning_rate": 1.7981076274868718e-05, "loss": 0.9585, "step": 4521 }, { "epoch": 0.23, "learning_rate": 1.798008477919931e-05, "loss": 0.9859, "step": 4522 }, { "epoch": 0.23, "learning_rate": 1.7979093067477605e-05, "loss": 0.996, "step": 4523 }, { "epoch": 0.23, "learning_rate": 1.7978101139730437e-05, "loss": 0.991, "step": 4524 }, { "epoch": 0.23, "learning_rate": 1.797710899598467e-05, "loss": 1.0077, "step": 4525 }, { "epoch": 0.23, "learning_rate": 1.7976116636267168e-05, "loss": 0.9995, "step": 4526 }, { "epoch": 0.23, "learning_rate": 1.797512406060479e-05, "loss": 1.0873, "step": 4527 }, { "epoch": 0.23, "learning_rate": 1.7974131269024413e-05, "loss": 1.0305, "step": 4528 }, { "epoch": 0.23, "learning_rate": 1.7973138261552915e-05, "loss": 1.0312, "step": 4529 }, { "epoch": 0.23, "learning_rate": 1.7972145038217185e-05, "loss": 1.0083, "step": 4530 }, { "epoch": 0.23, "learning_rate": 1.7971151599044107e-05, "loss": 0.9851, "step": 4531 }, { "epoch": 0.23, "learning_rate": 1.797015794406058e-05, "loss": 1.0703, "step": 4532 }, { "epoch": 0.23, "learning_rate": 1.7969164073293506e-05, "loss": 1.0711, "step": 4533 }, { "epoch": 0.23, "learning_rate": 1.7968169986769796e-05, "loss": 1.031, "step": 4534 }, { "epoch": 0.23, "learning_rate": 1.796717568451636e-05, "loss": 0.9634, "step": 4535 }, { "epoch": 0.23, "learning_rate": 1.7966181166560114e-05, "loss": 0.9322, "step": 4536 }, { "epoch": 0.23, "learning_rate": 1.7965186432927996e-05, "loss": 0.968, "step": 4537 }, { "epoch": 0.23, "learning_rate": 1.7964191483646924e-05, "loss": 0.9183, "step": 4538 }, { "epoch": 0.23, "learning_rate": 1.796319631874384e-05, "loss": 0.9902, "step": 4539 }, { "epoch": 0.23, "learning_rate": 1.7962200938245692e-05, "loss": 0.9229, "step": 4540 }, { "epoch": 0.23, "learning_rate": 1.796120534217942e-05, "loss": 0.989, "step": 4541 }, { "epoch": 0.23, "learning_rate": 1.796020953057199e-05, "loss": 1.0311, "step": 4542 }, { "epoch": 0.23, "learning_rate": 1.795921350345035e-05, "loss": 0.9519, "step": 4543 }, { "epoch": 0.23, "learning_rate": 1.795821726084148e-05, "loss": 0.9648, "step": 4544 }, { "epoch": 0.23, "learning_rate": 1.7957220802772338e-05, "loss": 0.9056, "step": 4545 }, { "epoch": 0.23, "learning_rate": 1.7956224129269908e-05, "loss": 0.993, "step": 4546 }, { "epoch": 0.23, "learning_rate": 1.795522724036118e-05, "loss": 1.1003, "step": 4547 }, { "epoch": 0.23, "learning_rate": 1.7954230136073133e-05, "loss": 0.9623, "step": 4548 }, { "epoch": 0.23, "learning_rate": 1.7953232816432768e-05, "loss": 0.8955, "step": 4549 }, { "epoch": 0.23, "learning_rate": 1.7952235281467086e-05, "loss": 0.9381, "step": 4550 }, { "epoch": 0.23, "learning_rate": 1.7951237531203096e-05, "loss": 0.9273, "step": 4551 }, { "epoch": 0.23, "learning_rate": 1.795023956566781e-05, "loss": 1.0571, "step": 4552 }, { "epoch": 0.23, "learning_rate": 1.7949241384888244e-05, "loss": 0.9158, "step": 4553 }, { "epoch": 0.23, "learning_rate": 1.7948242988891427e-05, "loss": 0.8721, "step": 4554 }, { "epoch": 0.23, "learning_rate": 1.794724437770439e-05, "loss": 1.0521, "step": 4555 }, { "epoch": 0.23, "learning_rate": 1.7946245551354156e-05, "loss": 0.9703, "step": 4556 }, { "epoch": 0.23, "learning_rate": 1.7945246509867788e-05, "loss": 0.9996, "step": 4557 }, { "epoch": 0.23, "learning_rate": 1.794424725327232e-05, "loss": 0.9758, "step": 4558 }, { "epoch": 0.23, "learning_rate": 1.794324778159481e-05, "loss": 1.0669, "step": 4559 }, { "epoch": 0.23, "learning_rate": 1.7942248094862317e-05, "loss": 1.0375, "step": 4560 }, { "epoch": 0.23, "learning_rate": 1.7941248193101907e-05, "loss": 0.9737, "step": 4561 }, { "epoch": 0.23, "learning_rate": 1.7940248076340653e-05, "loss": 1.0612, "step": 4562 }, { "epoch": 0.23, "learning_rate": 1.7939247744605628e-05, "loss": 1.0327, "step": 4563 }, { "epoch": 0.23, "learning_rate": 1.7938247197923914e-05, "loss": 0.9486, "step": 4564 }, { "epoch": 0.23, "learning_rate": 1.793724643632261e-05, "loss": 0.9967, "step": 4565 }, { "epoch": 0.23, "learning_rate": 1.79362454598288e-05, "loss": 0.988, "step": 4566 }, { "epoch": 0.23, "learning_rate": 1.7935244268469584e-05, "loss": 0.882, "step": 4567 }, { "epoch": 0.23, "learning_rate": 1.7934242862272076e-05, "loss": 0.9827, "step": 4568 }, { "epoch": 0.23, "learning_rate": 1.7933241241263385e-05, "loss": 1.0033, "step": 4569 }, { "epoch": 0.23, "learning_rate": 1.7932239405470624e-05, "loss": 1.0049, "step": 4570 }, { "epoch": 0.23, "learning_rate": 1.793123735492092e-05, "loss": 0.9124, "step": 4571 }, { "epoch": 0.23, "learning_rate": 1.7930235089641404e-05, "loss": 0.9456, "step": 4572 }, { "epoch": 0.23, "learning_rate": 1.792923260965921e-05, "loss": 1.0486, "step": 4573 }, { "epoch": 0.23, "learning_rate": 1.792822991500148e-05, "loss": 0.9352, "step": 4574 }, { "epoch": 0.23, "learning_rate": 1.7927227005695362e-05, "loss": 1.0103, "step": 4575 }, { "epoch": 0.23, "learning_rate": 1.7926223881768004e-05, "loss": 1.0272, "step": 4576 }, { "epoch": 0.23, "learning_rate": 1.792522054324657e-05, "loss": 1.0002, "step": 4577 }, { "epoch": 0.23, "learning_rate": 1.7924216990158216e-05, "loss": 1.0275, "step": 4578 }, { "epoch": 0.23, "learning_rate": 1.792321322253012e-05, "loss": 0.9801, "step": 4579 }, { "epoch": 0.23, "learning_rate": 1.792220924038946e-05, "loss": 1.0345, "step": 4580 }, { "epoch": 0.23, "learning_rate": 1.7921205043763408e-05, "loss": 1.0379, "step": 4581 }, { "epoch": 0.23, "learning_rate": 1.792020063267916e-05, "loss": 1.0482, "step": 4582 }, { "epoch": 0.23, "learning_rate": 1.7919196007163905e-05, "loss": 1.0494, "step": 4583 }, { "epoch": 0.23, "learning_rate": 1.7918191167244842e-05, "loss": 1.0302, "step": 4584 }, { "epoch": 0.23, "learning_rate": 1.7917186112949177e-05, "loss": 0.9246, "step": 4585 }, { "epoch": 0.23, "learning_rate": 1.7916180844304127e-05, "loss": 0.9268, "step": 4586 }, { "epoch": 0.23, "learning_rate": 1.79151753613369e-05, "loss": 0.9933, "step": 4587 }, { "epoch": 0.23, "learning_rate": 1.791416966407472e-05, "loss": 0.9964, "step": 4588 }, { "epoch": 0.23, "learning_rate": 1.7913163752544815e-05, "loss": 1.0319, "step": 4589 }, { "epoch": 0.23, "learning_rate": 1.7912157626774423e-05, "loss": 1.0397, "step": 4590 }, { "epoch": 0.23, "learning_rate": 1.791115128679078e-05, "loss": 0.9809, "step": 4591 }, { "epoch": 0.23, "learning_rate": 1.7910144732621134e-05, "loss": 1.0073, "step": 4592 }, { "epoch": 0.23, "learning_rate": 1.7909137964292735e-05, "loss": 1.0373, "step": 4593 }, { "epoch": 0.23, "learning_rate": 1.7908130981832838e-05, "loss": 1.0177, "step": 4594 }, { "epoch": 0.23, "learning_rate": 1.790712378526871e-05, "loss": 0.8996, "step": 4595 }, { "epoch": 0.23, "learning_rate": 1.7906116374627614e-05, "loss": 0.9311, "step": 4596 }, { "epoch": 0.23, "learning_rate": 1.7905108749936836e-05, "loss": 0.9768, "step": 4597 }, { "epoch": 0.23, "learning_rate": 1.7904100911223645e-05, "loss": 0.9507, "step": 4598 }, { "epoch": 0.23, "learning_rate": 1.790309285851533e-05, "loss": 0.9889, "step": 4599 }, { "epoch": 0.23, "learning_rate": 1.7902084591839185e-05, "loss": 0.9418, "step": 4600 }, { "epoch": 0.23, "learning_rate": 1.7901076111222505e-05, "loss": 0.9818, "step": 4601 }, { "epoch": 0.23, "learning_rate": 1.7900067416692598e-05, "loss": 1.0026, "step": 4602 }, { "epoch": 0.23, "learning_rate": 1.789905850827677e-05, "loss": 1.0684, "step": 4603 }, { "epoch": 0.23, "learning_rate": 1.7898049386002338e-05, "loss": 0.9421, "step": 4604 }, { "epoch": 0.23, "learning_rate": 1.789704004989662e-05, "loss": 0.9498, "step": 4605 }, { "epoch": 0.23, "learning_rate": 1.789603049998695e-05, "loss": 0.9273, "step": 4606 }, { "epoch": 0.23, "learning_rate": 1.7895020736300647e-05, "loss": 1.0032, "step": 4607 }, { "epoch": 0.23, "learning_rate": 1.789401075886506e-05, "loss": 0.9277, "step": 4608 }, { "epoch": 0.23, "learning_rate": 1.789300056770753e-05, "loss": 1.0446, "step": 4609 }, { "epoch": 0.23, "learning_rate": 1.789199016285541e-05, "loss": 1.0146, "step": 4610 }, { "epoch": 0.23, "learning_rate": 1.789097954433605e-05, "loss": 1.021, "step": 4611 }, { "epoch": 0.23, "learning_rate": 1.7889968712176814e-05, "loss": 1.0292, "step": 4612 }, { "epoch": 0.23, "learning_rate": 1.788895766640507e-05, "loss": 1.0223, "step": 4613 }, { "epoch": 0.23, "learning_rate": 1.7887946407048193e-05, "loss": 1.0195, "step": 4614 }, { "epoch": 0.23, "learning_rate": 1.7886934934133555e-05, "loss": 1.0937, "step": 4615 }, { "epoch": 0.23, "learning_rate": 1.7885923247688545e-05, "loss": 1.0067, "step": 4616 }, { "epoch": 0.23, "learning_rate": 1.7884911347740556e-05, "loss": 0.9959, "step": 4617 }, { "epoch": 0.23, "learning_rate": 1.7883899234316978e-05, "loss": 1.1537, "step": 4618 }, { "epoch": 0.23, "learning_rate": 1.7882886907445217e-05, "loss": 1.0005, "step": 4619 }, { "epoch": 0.23, "learning_rate": 1.7881874367152683e-05, "loss": 0.9834, "step": 4620 }, { "epoch": 0.23, "learning_rate": 1.7880861613466784e-05, "loss": 0.9974, "step": 4621 }, { "epoch": 0.23, "learning_rate": 1.7879848646414938e-05, "loss": 0.9879, "step": 4622 }, { "epoch": 0.23, "learning_rate": 1.7878835466024576e-05, "loss": 0.9662, "step": 4623 }, { "epoch": 0.23, "learning_rate": 1.787782207232313e-05, "loss": 1.0535, "step": 4624 }, { "epoch": 0.23, "learning_rate": 1.7876808465338027e-05, "loss": 1.0062, "step": 4625 }, { "epoch": 0.24, "learning_rate": 1.7875794645096718e-05, "loss": 1.014, "step": 4626 }, { "epoch": 0.24, "learning_rate": 1.7874780611626644e-05, "loss": 0.936, "step": 4627 }, { "epoch": 0.24, "learning_rate": 1.7873766364955267e-05, "loss": 0.9084, "step": 4628 }, { "epoch": 0.24, "learning_rate": 1.787275190511004e-05, "loss": 0.9421, "step": 4629 }, { "epoch": 0.24, "learning_rate": 1.7871737232118437e-05, "loss": 1.0425, "step": 4630 }, { "epoch": 0.24, "learning_rate": 1.787072234600792e-05, "loss": 0.9165, "step": 4631 }, { "epoch": 0.24, "learning_rate": 1.786970724680597e-05, "loss": 0.9815, "step": 4632 }, { "epoch": 0.24, "learning_rate": 1.7868691934540066e-05, "loss": 0.8802, "step": 4633 }, { "epoch": 0.24, "learning_rate": 1.78676764092377e-05, "loss": 1.0446, "step": 4634 }, { "epoch": 0.24, "learning_rate": 1.7866660670926368e-05, "loss": 1.0596, "step": 4635 }, { "epoch": 0.24, "learning_rate": 1.786564471963357e-05, "loss": 1.0101, "step": 4636 }, { "epoch": 0.24, "learning_rate": 1.7864628555386803e-05, "loss": 0.9727, "step": 4637 }, { "epoch": 0.24, "learning_rate": 1.7863612178213586e-05, "loss": 0.9524, "step": 4638 }, { "epoch": 0.24, "learning_rate": 1.7862595588141444e-05, "loss": 1.0441, "step": 4639 }, { "epoch": 0.24, "learning_rate": 1.7861578785197884e-05, "loss": 1.0159, "step": 4640 }, { "epoch": 0.24, "learning_rate": 1.786056176941044e-05, "loss": 0.9284, "step": 4641 }, { "epoch": 0.24, "learning_rate": 1.7859544540806655e-05, "loss": 0.9337, "step": 4642 }, { "epoch": 0.24, "learning_rate": 1.7858527099414064e-05, "loss": 0.8811, "step": 4643 }, { "epoch": 0.24, "learning_rate": 1.7857509445260207e-05, "loss": 1.0446, "step": 4644 }, { "epoch": 0.24, "learning_rate": 1.7856491578372646e-05, "loss": 1.026, "step": 4645 }, { "epoch": 0.24, "learning_rate": 1.7855473498778934e-05, "loss": 1.0949, "step": 4646 }, { "epoch": 0.24, "learning_rate": 1.7854455206506633e-05, "loss": 1.0707, "step": 4647 }, { "epoch": 0.24, "learning_rate": 1.7853436701583312e-05, "loss": 0.9296, "step": 4648 }, { "epoch": 0.24, "learning_rate": 1.785241798403655e-05, "loss": 1.0163, "step": 4649 }, { "epoch": 0.24, "learning_rate": 1.785139905389393e-05, "loss": 1.0316, "step": 4650 }, { "epoch": 0.24, "learning_rate": 1.7850379911183025e-05, "loss": 1.0059, "step": 4651 }, { "epoch": 0.24, "learning_rate": 1.784936055593144e-05, "loss": 0.9413, "step": 4652 }, { "epoch": 0.24, "learning_rate": 1.7848340988166772e-05, "loss": 0.9467, "step": 4653 }, { "epoch": 0.24, "learning_rate": 1.784732120791662e-05, "loss": 0.9497, "step": 4654 }, { "epoch": 0.24, "learning_rate": 1.7846301215208593e-05, "loss": 0.9557, "step": 4655 }, { "epoch": 0.24, "learning_rate": 1.784528101007031e-05, "loss": 0.9487, "step": 4656 }, { "epoch": 0.24, "learning_rate": 1.784426059252939e-05, "loss": 0.9673, "step": 4657 }, { "epoch": 0.24, "learning_rate": 1.784323996261346e-05, "loss": 1.0117, "step": 4658 }, { "epoch": 0.24, "learning_rate": 1.7842219120350153e-05, "loss": 0.967, "step": 4659 }, { "epoch": 0.24, "learning_rate": 1.7841198065767107e-05, "loss": 0.9228, "step": 4660 }, { "epoch": 0.24, "learning_rate": 1.7840176798891965e-05, "loss": 1.0074, "step": 4661 }, { "epoch": 0.24, "learning_rate": 1.7839155319752377e-05, "loss": 0.9371, "step": 4662 }, { "epoch": 0.24, "learning_rate": 1.7838133628376e-05, "loss": 1.0626, "step": 4663 }, { "epoch": 0.24, "learning_rate": 1.7837111724790493e-05, "loss": 1.0289, "step": 4664 }, { "epoch": 0.24, "learning_rate": 1.7836089609023526e-05, "loss": 0.9857, "step": 4665 }, { "epoch": 0.24, "learning_rate": 1.7835067281102765e-05, "loss": 0.8766, "step": 4666 }, { "epoch": 0.24, "learning_rate": 1.78340447410559e-05, "loss": 0.9675, "step": 4667 }, { "epoch": 0.24, "learning_rate": 1.7833021988910604e-05, "loss": 1.0667, "step": 4668 }, { "epoch": 0.24, "learning_rate": 1.7831999024694573e-05, "loss": 0.9904, "step": 4669 }, { "epoch": 0.24, "learning_rate": 1.78309758484355e-05, "loss": 0.9385, "step": 4670 }, { "epoch": 0.24, "learning_rate": 1.782995246016109e-05, "loss": 0.9518, "step": 4671 }, { "epoch": 0.24, "learning_rate": 1.7828928859899044e-05, "loss": 0.9997, "step": 4672 }, { "epoch": 0.24, "learning_rate": 1.7827905047677082e-05, "loss": 0.8342, "step": 4673 }, { "epoch": 0.24, "learning_rate": 1.7826881023522916e-05, "loss": 1.0562, "step": 4674 }, { "epoch": 0.24, "learning_rate": 1.7825856787464272e-05, "loss": 0.9297, "step": 4675 }, { "epoch": 0.24, "learning_rate": 1.7824832339528885e-05, "loss": 0.9938, "step": 4676 }, { "epoch": 0.24, "learning_rate": 1.782380767974449e-05, "loss": 1.1018, "step": 4677 }, { "epoch": 0.24, "learning_rate": 1.782278280813882e-05, "loss": 1.028, "step": 4678 }, { "epoch": 0.24, "learning_rate": 1.7821757724739626e-05, "loss": 1.0171, "step": 4679 }, { "epoch": 0.24, "learning_rate": 1.782073242957467e-05, "loss": 0.9147, "step": 4680 }, { "epoch": 0.24, "learning_rate": 1.7819706922671703e-05, "loss": 0.9865, "step": 4681 }, { "epoch": 0.24, "learning_rate": 1.781868120405849e-05, "loss": 1.052, "step": 4682 }, { "epoch": 0.24, "learning_rate": 1.78176552737628e-05, "loss": 0.9793, "step": 4683 }, { "epoch": 0.24, "learning_rate": 1.7816629131812412e-05, "loss": 0.946, "step": 4684 }, { "epoch": 0.24, "learning_rate": 1.7815602778235107e-05, "loss": 1.0166, "step": 4685 }, { "epoch": 0.24, "learning_rate": 1.781457621305867e-05, "loss": 1.0261, "step": 4686 }, { "epoch": 0.24, "learning_rate": 1.78135494363109e-05, "loss": 0.9951, "step": 4687 }, { "epoch": 0.24, "learning_rate": 1.7812522448019588e-05, "loss": 0.9355, "step": 4688 }, { "epoch": 0.24, "learning_rate": 1.7811495248212546e-05, "loss": 0.9684, "step": 4689 }, { "epoch": 0.24, "learning_rate": 1.7810467836917576e-05, "loss": 0.8461, "step": 4690 }, { "epoch": 0.24, "learning_rate": 1.7809440214162505e-05, "loss": 1.068, "step": 4691 }, { "epoch": 0.24, "learning_rate": 1.7808412379975145e-05, "loss": 0.9886, "step": 4692 }, { "epoch": 0.24, "learning_rate": 1.780738433438333e-05, "loss": 0.9265, "step": 4693 }, { "epoch": 0.24, "learning_rate": 1.7806356077414887e-05, "loss": 0.9973, "step": 4694 }, { "epoch": 0.24, "learning_rate": 1.780532760909766e-05, "loss": 1.0072, "step": 4695 }, { "epoch": 0.24, "learning_rate": 1.7804298929459495e-05, "loss": 1.0447, "step": 4696 }, { "epoch": 0.24, "learning_rate": 1.7803270038528236e-05, "loss": 1.0831, "step": 4697 }, { "epoch": 0.24, "learning_rate": 1.780224093633174e-05, "loss": 1.0119, "step": 4698 }, { "epoch": 0.24, "learning_rate": 1.7801211622897877e-05, "loss": 1.1125, "step": 4699 }, { "epoch": 0.24, "learning_rate": 1.7800182098254507e-05, "loss": 1.0237, "step": 4700 }, { "epoch": 0.24, "learning_rate": 1.7799152362429507e-05, "loss": 1.0582, "step": 4701 }, { "epoch": 0.24, "learning_rate": 1.7798122415450753e-05, "loss": 0.9113, "step": 4702 }, { "epoch": 0.24, "learning_rate": 1.7797092257346134e-05, "loss": 0.9943, "step": 4703 }, { "epoch": 0.24, "learning_rate": 1.7796061888143533e-05, "loss": 1.0245, "step": 4704 }, { "epoch": 0.24, "learning_rate": 1.7795031307870855e-05, "loss": 0.9785, "step": 4705 }, { "epoch": 0.24, "learning_rate": 1.7794000516555995e-05, "loss": 0.9939, "step": 4706 }, { "epoch": 0.24, "learning_rate": 1.7792969514226864e-05, "loss": 1.0087, "step": 4707 }, { "epoch": 0.24, "learning_rate": 1.7791938300911374e-05, "loss": 1.0147, "step": 4708 }, { "epoch": 0.24, "learning_rate": 1.7790906876637442e-05, "loss": 0.9314, "step": 4709 }, { "epoch": 0.24, "learning_rate": 1.7789875241432998e-05, "loss": 1.0637, "step": 4710 }, { "epoch": 0.24, "learning_rate": 1.778884339532597e-05, "loss": 1.0048, "step": 4711 }, { "epoch": 0.24, "learning_rate": 1.7787811338344295e-05, "loss": 1.013, "step": 4712 }, { "epoch": 0.24, "learning_rate": 1.7786779070515913e-05, "loss": 0.976, "step": 4713 }, { "epoch": 0.24, "learning_rate": 1.778574659186877e-05, "loss": 1.0084, "step": 4714 }, { "epoch": 0.24, "learning_rate": 1.778471390243082e-05, "loss": 1.0421, "step": 4715 }, { "epoch": 0.24, "learning_rate": 1.7783681002230026e-05, "loss": 0.9407, "step": 4716 }, { "epoch": 0.24, "learning_rate": 1.778264789129435e-05, "loss": 1.0055, "step": 4717 }, { "epoch": 0.24, "learning_rate": 1.7781614569651766e-05, "loss": 1.0789, "step": 4718 }, { "epoch": 0.24, "learning_rate": 1.778058103733024e-05, "loss": 0.8906, "step": 4719 }, { "epoch": 0.24, "learning_rate": 1.7779547294357763e-05, "loss": 0.981, "step": 4720 }, { "epoch": 0.24, "learning_rate": 1.777851334076232e-05, "loss": 0.9546, "step": 4721 }, { "epoch": 0.24, "learning_rate": 1.7777479176571905e-05, "loss": 0.9929, "step": 4722 }, { "epoch": 0.24, "learning_rate": 1.7776444801814512e-05, "loss": 0.949, "step": 4723 }, { "epoch": 0.24, "learning_rate": 1.7775410216518155e-05, "loss": 1.0455, "step": 4724 }, { "epoch": 0.24, "learning_rate": 1.777437542071083e-05, "loss": 1.0507, "step": 4725 }, { "epoch": 0.24, "learning_rate": 1.7773340414420574e-05, "loss": 1.0583, "step": 4726 }, { "epoch": 0.24, "learning_rate": 1.777230519767539e-05, "loss": 1.0049, "step": 4727 }, { "epoch": 0.24, "learning_rate": 1.777126977050331e-05, "loss": 1.0345, "step": 4728 }, { "epoch": 0.24, "learning_rate": 1.777023413293237e-05, "loss": 0.9091, "step": 4729 }, { "epoch": 0.24, "learning_rate": 1.776919828499061e-05, "loss": 0.9696, "step": 4730 }, { "epoch": 0.24, "learning_rate": 1.776816222670607e-05, "loss": 0.9019, "step": 4731 }, { "epoch": 0.24, "learning_rate": 1.7767125958106803e-05, "loss": 1.0523, "step": 4732 }, { "epoch": 0.24, "learning_rate": 1.7766089479220866e-05, "loss": 0.9942, "step": 4733 }, { "epoch": 0.24, "learning_rate": 1.7765052790076316e-05, "loss": 0.9645, "step": 4734 }, { "epoch": 0.24, "learning_rate": 1.7764015890701225e-05, "loss": 1.0103, "step": 4735 }, { "epoch": 0.24, "learning_rate": 1.7762978781123663e-05, "loss": 0.9742, "step": 4736 }, { "epoch": 0.24, "learning_rate": 1.7761941461371707e-05, "loss": 1.0314, "step": 4737 }, { "epoch": 0.24, "learning_rate": 1.776090393147345e-05, "loss": 0.9785, "step": 4738 }, { "epoch": 0.24, "learning_rate": 1.775986619145697e-05, "loss": 0.9903, "step": 4739 }, { "epoch": 0.24, "learning_rate": 1.7758828241350373e-05, "loss": 1.0528, "step": 4740 }, { "epoch": 0.24, "learning_rate": 1.7757790081181756e-05, "loss": 0.9756, "step": 4741 }, { "epoch": 0.24, "learning_rate": 1.7756751710979224e-05, "loss": 0.9729, "step": 4742 }, { "epoch": 0.24, "learning_rate": 1.7755713130770895e-05, "loss": 1.0665, "step": 4743 }, { "epoch": 0.24, "learning_rate": 1.775467434058488e-05, "loss": 0.939, "step": 4744 }, { "epoch": 0.24, "learning_rate": 1.7753635340449312e-05, "loss": 0.8649, "step": 4745 }, { "epoch": 0.24, "learning_rate": 1.7752596130392312e-05, "loss": 0.9832, "step": 4746 }, { "epoch": 0.24, "learning_rate": 1.7751556710442024e-05, "loss": 0.9434, "step": 4747 }, { "epoch": 0.24, "learning_rate": 1.7750517080626585e-05, "loss": 0.9113, "step": 4748 }, { "epoch": 0.24, "learning_rate": 1.774947724097414e-05, "loss": 1.0174, "step": 4749 }, { "epoch": 0.24, "learning_rate": 1.774843719151284e-05, "loss": 0.9153, "step": 4750 }, { "epoch": 0.24, "learning_rate": 1.7747396932270853e-05, "loss": 1.0945, "step": 4751 }, { "epoch": 0.24, "learning_rate": 1.7746356463276332e-05, "loss": 1.067, "step": 4752 }, { "epoch": 0.24, "learning_rate": 1.7745315784557456e-05, "loss": 0.9426, "step": 4753 }, { "epoch": 0.24, "learning_rate": 1.7744274896142394e-05, "loss": 1.0445, "step": 4754 }, { "epoch": 0.24, "learning_rate": 1.7743233798059325e-05, "loss": 1.0326, "step": 4755 }, { "epoch": 0.24, "learning_rate": 1.7742192490336442e-05, "loss": 0.9342, "step": 4756 }, { "epoch": 0.24, "learning_rate": 1.7741150973001934e-05, "loss": 1.007, "step": 4757 }, { "epoch": 0.24, "learning_rate": 1.7740109246083997e-05, "loss": 1.0405, "step": 4758 }, { "epoch": 0.24, "learning_rate": 1.773906730961084e-05, "loss": 1.0869, "step": 4759 }, { "epoch": 0.24, "learning_rate": 1.773802516361067e-05, "loss": 1.0132, "step": 4760 }, { "epoch": 0.24, "learning_rate": 1.77369828081117e-05, "loss": 0.9369, "step": 4761 }, { "epoch": 0.24, "learning_rate": 1.7735940243142153e-05, "loss": 0.9902, "step": 4762 }, { "epoch": 0.24, "learning_rate": 1.773489746873025e-05, "loss": 0.9683, "step": 4763 }, { "epoch": 0.24, "learning_rate": 1.7733854484904228e-05, "loss": 1.0004, "step": 4764 }, { "epoch": 0.24, "learning_rate": 1.773281129169233e-05, "loss": 0.8761, "step": 4765 }, { "epoch": 0.24, "learning_rate": 1.773176788912279e-05, "loss": 0.9901, "step": 4766 }, { "epoch": 0.24, "learning_rate": 1.773072427722386e-05, "loss": 0.9602, "step": 4767 }, { "epoch": 0.24, "learning_rate": 1.7729680456023795e-05, "loss": 0.9695, "step": 4768 }, { "epoch": 0.24, "learning_rate": 1.7728636425550854e-05, "loss": 0.9756, "step": 4769 }, { "epoch": 0.24, "learning_rate": 1.7727592185833307e-05, "loss": 1.0269, "step": 4770 }, { "epoch": 0.24, "learning_rate": 1.7726547736899422e-05, "loss": 0.9514, "step": 4771 }, { "epoch": 0.24, "learning_rate": 1.7725503078777476e-05, "loss": 1.0092, "step": 4772 }, { "epoch": 0.24, "learning_rate": 1.7724458211495753e-05, "loss": 1.0634, "step": 4773 }, { "epoch": 0.24, "learning_rate": 1.7723413135082547e-05, "loss": 0.9428, "step": 4774 }, { "epoch": 0.24, "learning_rate": 1.7722367849566143e-05, "loss": 0.9623, "step": 4775 }, { "epoch": 0.24, "learning_rate": 1.7721322354974848e-05, "loss": 0.9658, "step": 4776 }, { "epoch": 0.24, "learning_rate": 1.7720276651336964e-05, "loss": 0.9162, "step": 4777 }, { "epoch": 0.24, "learning_rate": 1.77192307386808e-05, "loss": 1.0823, "step": 4778 }, { "epoch": 0.24, "learning_rate": 1.771818461703468e-05, "loss": 0.9194, "step": 4779 }, { "epoch": 0.24, "learning_rate": 1.7717138286426926e-05, "loss": 0.8657, "step": 4780 }, { "epoch": 0.24, "learning_rate": 1.7716091746885857e-05, "loss": 0.9874, "step": 4781 }, { "epoch": 0.24, "learning_rate": 1.7715044998439815e-05, "loss": 1.0464, "step": 4782 }, { "epoch": 0.24, "learning_rate": 1.771399804111714e-05, "loss": 0.9082, "step": 4783 }, { "epoch": 0.24, "learning_rate": 1.7712950874946174e-05, "loss": 1.0536, "step": 4784 }, { "epoch": 0.24, "learning_rate": 1.771190349995527e-05, "loss": 1.0317, "step": 4785 }, { "epoch": 0.24, "learning_rate": 1.771085591617278e-05, "loss": 0.9757, "step": 4786 }, { "epoch": 0.24, "learning_rate": 1.7709808123627074e-05, "loss": 1.0844, "step": 4787 }, { "epoch": 0.24, "learning_rate": 1.7708760122346516e-05, "loss": 1.0016, "step": 4788 }, { "epoch": 0.24, "learning_rate": 1.7707711912359476e-05, "loss": 1.0499, "step": 4789 }, { "epoch": 0.24, "learning_rate": 1.770666349369434e-05, "loss": 0.9196, "step": 4790 }, { "epoch": 0.24, "learning_rate": 1.7705614866379485e-05, "loss": 0.9792, "step": 4791 }, { "epoch": 0.24, "learning_rate": 1.7704566030443306e-05, "loss": 1.0674, "step": 4792 }, { "epoch": 0.24, "learning_rate": 1.77035169859142e-05, "loss": 0.9431, "step": 4793 }, { "epoch": 0.24, "learning_rate": 1.7702467732820565e-05, "loss": 0.8896, "step": 4794 }, { "epoch": 0.24, "learning_rate": 1.7701418271190814e-05, "loss": 1.0495, "step": 4795 }, { "epoch": 0.24, "learning_rate": 1.7700368601053356e-05, "loss": 1.0394, "step": 4796 }, { "epoch": 0.24, "learning_rate": 1.769931872243661e-05, "loss": 1.0103, "step": 4797 }, { "epoch": 0.24, "learning_rate": 1.7698268635369004e-05, "loss": 0.96, "step": 4798 }, { "epoch": 0.24, "learning_rate": 1.7697218339878963e-05, "loss": 0.9525, "step": 4799 }, { "epoch": 0.24, "learning_rate": 1.7696167835994927e-05, "loss": 0.9181, "step": 4800 }, { "epoch": 0.24, "learning_rate": 1.7695117123745333e-05, "loss": 0.9225, "step": 4801 }, { "epoch": 0.24, "learning_rate": 1.7694066203158627e-05, "loss": 0.9386, "step": 4802 }, { "epoch": 0.24, "learning_rate": 1.7693015074263267e-05, "loss": 0.9366, "step": 4803 }, { "epoch": 0.24, "learning_rate": 1.769196373708771e-05, "loss": 0.9727, "step": 4804 }, { "epoch": 0.24, "learning_rate": 1.7690912191660415e-05, "loss": 1.048, "step": 4805 }, { "epoch": 0.24, "learning_rate": 1.7689860438009857e-05, "loss": 0.932, "step": 4806 }, { "epoch": 0.24, "learning_rate": 1.768880847616451e-05, "loss": 1.0299, "step": 4807 }, { "epoch": 0.24, "learning_rate": 1.7687756306152852e-05, "loss": 0.9378, "step": 4808 }, { "epoch": 0.24, "learning_rate": 1.768670392800337e-05, "loss": 0.8443, "step": 4809 }, { "epoch": 0.24, "learning_rate": 1.7685651341744562e-05, "loss": 1.0184, "step": 4810 }, { "epoch": 0.24, "learning_rate": 1.7684598547404913e-05, "loss": 0.9311, "step": 4811 }, { "epoch": 0.24, "learning_rate": 1.7683545545012942e-05, "loss": 1.0225, "step": 4812 }, { "epoch": 0.24, "learning_rate": 1.7682492334597147e-05, "loss": 1.0344, "step": 4813 }, { "epoch": 0.24, "learning_rate": 1.7681438916186046e-05, "loss": 0.9847, "step": 4814 }, { "epoch": 0.24, "learning_rate": 1.7680385289808155e-05, "loss": 0.975, "step": 4815 }, { "epoch": 0.24, "learning_rate": 1.7679331455492008e-05, "loss": 0.9667, "step": 4816 }, { "epoch": 0.24, "learning_rate": 1.7678277413266132e-05, "loss": 0.9201, "step": 4817 }, { "epoch": 0.24, "learning_rate": 1.767722316315906e-05, "loss": 0.863, "step": 4818 }, { "epoch": 0.24, "learning_rate": 1.7676168705199345e-05, "loss": 0.9495, "step": 4819 }, { "epoch": 0.24, "learning_rate": 1.767511403941552e-05, "loss": 1.0343, "step": 4820 }, { "epoch": 0.24, "learning_rate": 1.7674059165836158e-05, "loss": 1.0315, "step": 4821 }, { "epoch": 0.24, "learning_rate": 1.76730040844898e-05, "loss": 0.9583, "step": 4822 }, { "epoch": 0.25, "learning_rate": 1.7671948795405026e-05, "loss": 0.9939, "step": 4823 }, { "epoch": 0.25, "learning_rate": 1.7670893298610398e-05, "loss": 0.8956, "step": 4824 }, { "epoch": 0.25, "learning_rate": 1.76698375941345e-05, "loss": 1.0724, "step": 4825 }, { "epoch": 0.25, "learning_rate": 1.76687816820059e-05, "loss": 1.0545, "step": 4826 }, { "epoch": 0.25, "learning_rate": 1.76677255622532e-05, "loss": 0.9968, "step": 4827 }, { "epoch": 0.25, "learning_rate": 1.766666923490499e-05, "loss": 0.9444, "step": 4828 }, { "epoch": 0.25, "learning_rate": 1.7665612699989866e-05, "loss": 1.001, "step": 4829 }, { "epoch": 0.25, "learning_rate": 1.7664555957536436e-05, "loss": 1.0185, "step": 4830 }, { "epoch": 0.25, "learning_rate": 1.7663499007573303e-05, "loss": 0.9965, "step": 4831 }, { "epoch": 0.25, "learning_rate": 1.766244185012909e-05, "loss": 0.9215, "step": 4832 }, { "epoch": 0.25, "learning_rate": 1.7661384485232415e-05, "loss": 0.9445, "step": 4833 }, { "epoch": 0.25, "learning_rate": 1.7660326912911906e-05, "loss": 1.0018, "step": 4834 }, { "epoch": 0.25, "learning_rate": 1.76592691331962e-05, "loss": 0.9491, "step": 4835 }, { "epoch": 0.25, "learning_rate": 1.7658211146113925e-05, "loss": 1.0135, "step": 4836 }, { "epoch": 0.25, "learning_rate": 1.7657152951693737e-05, "loss": 0.9286, "step": 4837 }, { "epoch": 0.25, "learning_rate": 1.7656094549964275e-05, "loss": 1.021, "step": 4838 }, { "epoch": 0.25, "learning_rate": 1.76550359409542e-05, "loss": 1.0306, "step": 4839 }, { "epoch": 0.25, "learning_rate": 1.765397712469217e-05, "loss": 0.9437, "step": 4840 }, { "epoch": 0.25, "learning_rate": 1.7652918101206852e-05, "loss": 0.9636, "step": 4841 }, { "epoch": 0.25, "learning_rate": 1.7651858870526922e-05, "loss": 1.0989, "step": 4842 }, { "epoch": 0.25, "learning_rate": 1.7650799432681053e-05, "loss": 0.9158, "step": 4843 }, { "epoch": 0.25, "learning_rate": 1.7649739787697925e-05, "loss": 0.942, "step": 4844 }, { "epoch": 0.25, "learning_rate": 1.7648679935606236e-05, "loss": 1.0078, "step": 4845 }, { "epoch": 0.25, "learning_rate": 1.7647619876434673e-05, "loss": 1.0571, "step": 4846 }, { "epoch": 0.25, "learning_rate": 1.7646559610211937e-05, "loss": 0.9971, "step": 4847 }, { "epoch": 0.25, "learning_rate": 1.764549913696674e-05, "loss": 1.0297, "step": 4848 }, { "epoch": 0.25, "learning_rate": 1.7644438456727782e-05, "loss": 0.9808, "step": 4849 }, { "epoch": 0.25, "learning_rate": 1.7643377569523786e-05, "loss": 1.0061, "step": 4850 }, { "epoch": 0.25, "learning_rate": 1.7642316475383478e-05, "loss": 0.9811, "step": 4851 }, { "epoch": 0.25, "learning_rate": 1.764125517433558e-05, "loss": 0.9334, "step": 4852 }, { "epoch": 0.25, "learning_rate": 1.7640193666408828e-05, "loss": 0.984, "step": 4853 }, { "epoch": 0.25, "learning_rate": 1.7639131951631962e-05, "loss": 0.9077, "step": 4854 }, { "epoch": 0.25, "learning_rate": 1.7638070030033726e-05, "loss": 0.9796, "step": 4855 }, { "epoch": 0.25, "learning_rate": 1.763700790164287e-05, "loss": 0.9901, "step": 4856 }, { "epoch": 0.25, "learning_rate": 1.7635945566488152e-05, "loss": 1.0878, "step": 4857 }, { "epoch": 0.25, "learning_rate": 1.7634883024598324e-05, "loss": 0.9796, "step": 4858 }, { "epoch": 0.25, "learning_rate": 1.763382027600217e-05, "loss": 1.0271, "step": 4859 }, { "epoch": 0.25, "learning_rate": 1.763275732072845e-05, "loss": 0.8558, "step": 4860 }, { "epoch": 0.25, "learning_rate": 1.7631694158805945e-05, "loss": 0.8795, "step": 4861 }, { "epoch": 0.25, "learning_rate": 1.763063079026344e-05, "loss": 1.0478, "step": 4862 }, { "epoch": 0.25, "learning_rate": 1.762956721512973e-05, "loss": 1.0125, "step": 4863 }, { "epoch": 0.25, "learning_rate": 1.76285034334336e-05, "loss": 0.9404, "step": 4864 }, { "epoch": 0.25, "learning_rate": 1.7627439445203854e-05, "loss": 0.9563, "step": 4865 }, { "epoch": 0.25, "learning_rate": 1.7626375250469306e-05, "loss": 1.0378, "step": 4866 }, { "epoch": 0.25, "learning_rate": 1.762531084925876e-05, "loss": 0.9659, "step": 4867 }, { "epoch": 0.25, "learning_rate": 1.762424624160103e-05, "loss": 0.9801, "step": 4868 }, { "epoch": 0.25, "learning_rate": 1.762318142752495e-05, "loss": 0.9179, "step": 4869 }, { "epoch": 0.25, "learning_rate": 1.762211640705934e-05, "loss": 0.966, "step": 4870 }, { "epoch": 0.25, "learning_rate": 1.7621051180233042e-05, "loss": 0.9211, "step": 4871 }, { "epoch": 0.25, "learning_rate": 1.761998574707489e-05, "loss": 0.9753, "step": 4872 }, { "epoch": 0.25, "learning_rate": 1.761892010761373e-05, "loss": 0.92, "step": 4873 }, { "epoch": 0.25, "learning_rate": 1.7617854261878413e-05, "loss": 0.9481, "step": 4874 }, { "epoch": 0.25, "learning_rate": 1.7616788209897798e-05, "loss": 0.9828, "step": 4875 }, { "epoch": 0.25, "learning_rate": 1.7615721951700743e-05, "loss": 0.9208, "step": 4876 }, { "epoch": 0.25, "learning_rate": 1.7614655487316123e-05, "loss": 1.0648, "step": 4877 }, { "epoch": 0.25, "learning_rate": 1.7613588816772808e-05, "loss": 0.978, "step": 4878 }, { "epoch": 0.25, "learning_rate": 1.761252194009967e-05, "loss": 0.9685, "step": 4879 }, { "epoch": 0.25, "learning_rate": 1.76114548573256e-05, "loss": 1.0528, "step": 4880 }, { "epoch": 0.25, "learning_rate": 1.7610387568479496e-05, "loss": 0.874, "step": 4881 }, { "epoch": 0.25, "learning_rate": 1.760932007359024e-05, "loss": 1.0265, "step": 4882 }, { "epoch": 0.25, "learning_rate": 1.7608252372686735e-05, "loss": 1.0718, "step": 4883 }, { "epoch": 0.25, "learning_rate": 1.7607184465797894e-05, "loss": 0.9769, "step": 4884 }, { "epoch": 0.25, "learning_rate": 1.7606116352952627e-05, "loss": 0.9634, "step": 4885 }, { "epoch": 0.25, "learning_rate": 1.7605048034179856e-05, "loss": 0.9106, "step": 4886 }, { "epoch": 0.25, "learning_rate": 1.7603979509508494e-05, "loss": 0.9752, "step": 4887 }, { "epoch": 0.25, "learning_rate": 1.7602910778967485e-05, "loss": 1.1028, "step": 4888 }, { "epoch": 0.25, "learning_rate": 1.7601841842585747e-05, "loss": 0.9738, "step": 4889 }, { "epoch": 0.25, "learning_rate": 1.7600772700392234e-05, "loss": 0.9548, "step": 4890 }, { "epoch": 0.25, "learning_rate": 1.7599703352415886e-05, "loss": 0.958, "step": 4891 }, { "epoch": 0.25, "learning_rate": 1.7598633798685653e-05, "loss": 0.9346, "step": 4892 }, { "epoch": 0.25, "learning_rate": 1.7597564039230498e-05, "loss": 1.0363, "step": 4893 }, { "epoch": 0.25, "learning_rate": 1.7596494074079374e-05, "loss": 0.9495, "step": 4894 }, { "epoch": 0.25, "learning_rate": 1.7595423903261258e-05, "loss": 1.0348, "step": 4895 }, { "epoch": 0.25, "learning_rate": 1.7594353526805125e-05, "loss": 0.9373, "step": 4896 }, { "epoch": 0.25, "learning_rate": 1.7593282944739943e-05, "loss": 0.9384, "step": 4897 }, { "epoch": 0.25, "learning_rate": 1.759221215709471e-05, "loss": 1.1584, "step": 4898 }, { "epoch": 0.25, "learning_rate": 1.7591141163898406e-05, "loss": 1.0193, "step": 4899 }, { "epoch": 0.25, "learning_rate": 1.759006996518003e-05, "loss": 0.8897, "step": 4900 }, { "epoch": 0.25, "learning_rate": 1.758899856096859e-05, "loss": 0.9839, "step": 4901 }, { "epoch": 0.25, "learning_rate": 1.7587926951293087e-05, "loss": 0.915, "step": 4902 }, { "epoch": 0.25, "learning_rate": 1.758685513618253e-05, "loss": 0.9083, "step": 4903 }, { "epoch": 0.25, "learning_rate": 1.7585783115665948e-05, "loss": 0.957, "step": 4904 }, { "epoch": 0.25, "learning_rate": 1.7584710889772355e-05, "loss": 0.948, "step": 4905 }, { "epoch": 0.25, "learning_rate": 1.758363845853079e-05, "loss": 0.9977, "step": 4906 }, { "epoch": 0.25, "learning_rate": 1.7582565821970273e-05, "loss": 0.9828, "step": 4907 }, { "epoch": 0.25, "learning_rate": 1.7581492980119857e-05, "loss": 0.9252, "step": 4908 }, { "epoch": 0.25, "learning_rate": 1.758041993300859e-05, "loss": 0.962, "step": 4909 }, { "epoch": 0.25, "learning_rate": 1.757934668066551e-05, "loss": 0.9905, "step": 4910 }, { "epoch": 0.25, "learning_rate": 1.757827322311969e-05, "loss": 0.9844, "step": 4911 }, { "epoch": 0.25, "learning_rate": 1.757719956040018e-05, "loss": 1.0477, "step": 4912 }, { "epoch": 0.25, "learning_rate": 1.7576125692536056e-05, "loss": 0.9307, "step": 4913 }, { "epoch": 0.25, "learning_rate": 1.7575051619556388e-05, "loss": 1.0378, "step": 4914 }, { "epoch": 0.25, "learning_rate": 1.7573977341490256e-05, "loss": 1.0812, "step": 4915 }, { "epoch": 0.25, "learning_rate": 1.7572902858366747e-05, "loss": 0.9112, "step": 4916 }, { "epoch": 0.25, "learning_rate": 1.7571828170214947e-05, "loss": 1.045, "step": 4917 }, { "epoch": 0.25, "learning_rate": 1.7570753277063958e-05, "loss": 0.9549, "step": 4918 }, { "epoch": 0.25, "learning_rate": 1.7569678178942878e-05, "loss": 0.9298, "step": 4919 }, { "epoch": 0.25, "learning_rate": 1.7568602875880814e-05, "loss": 0.9926, "step": 4920 }, { "epoch": 0.25, "learning_rate": 1.756752736790688e-05, "loss": 1.0125, "step": 4921 }, { "epoch": 0.25, "learning_rate": 1.7566451655050197e-05, "loss": 0.9369, "step": 4922 }, { "epoch": 0.25, "learning_rate": 1.756537573733988e-05, "loss": 0.8562, "step": 4923 }, { "epoch": 0.25, "learning_rate": 1.7564299614805066e-05, "loss": 0.8908, "step": 4924 }, { "epoch": 0.25, "learning_rate": 1.7563223287474893e-05, "loss": 1.036, "step": 4925 }, { "epoch": 0.25, "learning_rate": 1.756214675537849e-05, "loss": 1.0458, "step": 4926 }, { "epoch": 0.25, "learning_rate": 1.756107001854501e-05, "loss": 1.0375, "step": 4927 }, { "epoch": 0.25, "learning_rate": 1.7559993077003605e-05, "loss": 0.9983, "step": 4928 }, { "epoch": 0.25, "learning_rate": 1.7558915930783428e-05, "loss": 0.9422, "step": 4929 }, { "epoch": 0.25, "learning_rate": 1.7557838579913644e-05, "loss": 0.9433, "step": 4930 }, { "epoch": 0.25, "learning_rate": 1.7556761024423427e-05, "loss": 0.9109, "step": 4931 }, { "epoch": 0.25, "learning_rate": 1.755568326434194e-05, "loss": 1.0456, "step": 4932 }, { "epoch": 0.25, "learning_rate": 1.755460529969837e-05, "loss": 1.0032, "step": 4933 }, { "epoch": 0.25, "learning_rate": 1.7553527130521896e-05, "loss": 0.9952, "step": 4934 }, { "epoch": 0.25, "learning_rate": 1.7552448756841712e-05, "loss": 0.9567, "step": 4935 }, { "epoch": 0.25, "learning_rate": 1.7551370178687012e-05, "loss": 1.0399, "step": 4936 }, { "epoch": 0.25, "learning_rate": 1.7550291396087e-05, "loss": 0.9302, "step": 4937 }, { "epoch": 0.25, "learning_rate": 1.7549212409070883e-05, "loss": 0.9515, "step": 4938 }, { "epoch": 0.25, "learning_rate": 1.754813321766787e-05, "loss": 1.0422, "step": 4939 }, { "epoch": 0.25, "learning_rate": 1.7547053821907183e-05, "loss": 0.8949, "step": 4940 }, { "epoch": 0.25, "learning_rate": 1.7545974221818037e-05, "loss": 1.0353, "step": 4941 }, { "epoch": 0.25, "learning_rate": 1.7544894417429673e-05, "loss": 0.9821, "step": 4942 }, { "epoch": 0.25, "learning_rate": 1.754381440877132e-05, "loss": 1.0537, "step": 4943 }, { "epoch": 0.25, "learning_rate": 1.7542734195872215e-05, "loss": 0.968, "step": 4944 }, { "epoch": 0.25, "learning_rate": 1.7541653778761607e-05, "loss": 0.9606, "step": 4945 }, { "epoch": 0.25, "learning_rate": 1.754057315746875e-05, "loss": 0.9468, "step": 4946 }, { "epoch": 0.25, "learning_rate": 1.7539492332022894e-05, "loss": 1.0089, "step": 4947 }, { "epoch": 0.25, "learning_rate": 1.7538411302453305e-05, "loss": 0.9579, "step": 4948 }, { "epoch": 0.25, "learning_rate": 1.753733006878925e-05, "loss": 0.9564, "step": 4949 }, { "epoch": 0.25, "learning_rate": 1.753624863106e-05, "loss": 0.9223, "step": 4950 }, { "epoch": 0.25, "learning_rate": 1.753516698929484e-05, "loss": 0.979, "step": 4951 }, { "epoch": 0.25, "learning_rate": 1.7534085143523053e-05, "loss": 0.9849, "step": 4952 }, { "epoch": 0.25, "learning_rate": 1.753300309377392e-05, "loss": 0.9707, "step": 4953 }, { "epoch": 0.25, "learning_rate": 1.7531920840076745e-05, "loss": 1.0068, "step": 4954 }, { "epoch": 0.25, "learning_rate": 1.753083838246083e-05, "loss": 1.0296, "step": 4955 }, { "epoch": 0.25, "learning_rate": 1.7529755720955474e-05, "loss": 0.9855, "step": 4956 }, { "epoch": 0.25, "learning_rate": 1.7528672855589993e-05, "loss": 1.0616, "step": 4957 }, { "epoch": 0.25, "learning_rate": 1.7527589786393707e-05, "loss": 1.0469, "step": 4958 }, { "epoch": 0.25, "learning_rate": 1.7526506513395935e-05, "loss": 0.9427, "step": 4959 }, { "epoch": 0.25, "learning_rate": 1.7525423036626006e-05, "loss": 0.9122, "step": 4960 }, { "epoch": 0.25, "learning_rate": 1.7524339356113255e-05, "loss": 0.9493, "step": 4961 }, { "epoch": 0.25, "learning_rate": 1.752325547188702e-05, "loss": 0.8794, "step": 4962 }, { "epoch": 0.25, "learning_rate": 1.752217138397665e-05, "loss": 0.8718, "step": 4963 }, { "epoch": 0.25, "learning_rate": 1.752108709241149e-05, "loss": 1.0465, "step": 4964 }, { "epoch": 0.25, "learning_rate": 1.75200025972209e-05, "loss": 1.0042, "step": 4965 }, { "epoch": 0.25, "learning_rate": 1.7518917898434243e-05, "loss": 1.0267, "step": 4966 }, { "epoch": 0.25, "learning_rate": 1.751783299608088e-05, "loss": 1.0406, "step": 4967 }, { "epoch": 0.25, "learning_rate": 1.7516747890190187e-05, "loss": 1.058, "step": 4968 }, { "epoch": 0.25, "learning_rate": 1.751566258079154e-05, "loss": 0.9301, "step": 4969 }, { "epoch": 0.25, "learning_rate": 1.751457706791433e-05, "loss": 1.0083, "step": 4970 }, { "epoch": 0.25, "learning_rate": 1.751349135158794e-05, "loss": 0.9358, "step": 4971 }, { "epoch": 0.25, "learning_rate": 1.7512405431841763e-05, "loss": 0.919, "step": 4972 }, { "epoch": 0.25, "learning_rate": 1.7511319308705198e-05, "loss": 0.97, "step": 4973 }, { "epoch": 0.25, "learning_rate": 1.751023298220766e-05, "loss": 0.9918, "step": 4974 }, { "epoch": 0.25, "learning_rate": 1.7509146452378553e-05, "loss": 1.0359, "step": 4975 }, { "epoch": 0.25, "learning_rate": 1.7508059719247294e-05, "loss": 0.9857, "step": 4976 }, { "epoch": 0.25, "learning_rate": 1.7506972782843307e-05, "loss": 1.001, "step": 4977 }, { "epoch": 0.25, "learning_rate": 1.750588564319602e-05, "loss": 0.9596, "step": 4978 }, { "epoch": 0.25, "learning_rate": 1.750479830033486e-05, "loss": 0.9151, "step": 4979 }, { "epoch": 0.25, "learning_rate": 1.7503710754289277e-05, "loss": 1.0032, "step": 4980 }, { "epoch": 0.25, "learning_rate": 1.75026230050887e-05, "loss": 0.9406, "step": 4981 }, { "epoch": 0.25, "learning_rate": 1.7501535052762595e-05, "loss": 0.9546, "step": 4982 }, { "epoch": 0.25, "learning_rate": 1.7500446897340408e-05, "loss": 1.0813, "step": 4983 }, { "epoch": 0.25, "learning_rate": 1.74993585388516e-05, "loss": 1.0162, "step": 4984 }, { "epoch": 0.25, "learning_rate": 1.7498269977325636e-05, "loss": 0.8888, "step": 4985 }, { "epoch": 0.25, "learning_rate": 1.7497181212791994e-05, "loss": 0.9592, "step": 4986 }, { "epoch": 0.25, "learning_rate": 1.7496092245280147e-05, "loss": 1.0355, "step": 4987 }, { "epoch": 0.25, "learning_rate": 1.749500307481957e-05, "loss": 1.0643, "step": 4988 }, { "epoch": 0.25, "learning_rate": 1.7493913701439765e-05, "loss": 1.0662, "step": 4989 }, { "epoch": 0.25, "learning_rate": 1.749282412517022e-05, "loss": 0.9756, "step": 4990 }, { "epoch": 0.25, "learning_rate": 1.7491734346040434e-05, "loss": 1.0607, "step": 4991 }, { "epoch": 0.25, "learning_rate": 1.749064436407991e-05, "loss": 0.9722, "step": 4992 }, { "epoch": 0.25, "learning_rate": 1.748955417931816e-05, "loss": 0.9534, "step": 4993 }, { "epoch": 0.25, "learning_rate": 1.7488463791784697e-05, "loss": 0.9671, "step": 4994 }, { "epoch": 0.25, "learning_rate": 1.7487373201509044e-05, "loss": 1.0201, "step": 4995 }, { "epoch": 0.25, "learning_rate": 1.748628240852073e-05, "loss": 0.9587, "step": 4996 }, { "epoch": 0.25, "learning_rate": 1.748519141284928e-05, "loss": 0.9439, "step": 4997 }, { "epoch": 0.25, "learning_rate": 1.748410021452424e-05, "loss": 0.9224, "step": 4998 }, { "epoch": 0.25, "learning_rate": 1.748300881357515e-05, "loss": 0.9984, "step": 4999 }, { "epoch": 0.25, "learning_rate": 1.7481917210031556e-05, "loss": 0.9595, "step": 5000 }, { "epoch": 0.25, "learning_rate": 1.748082540392301e-05, "loss": 0.9387, "step": 5001 }, { "epoch": 0.25, "learning_rate": 1.747973339527908e-05, "loss": 1.0092, "step": 5002 }, { "epoch": 0.25, "learning_rate": 1.7478641184129328e-05, "loss": 0.9708, "step": 5003 }, { "epoch": 0.25, "learning_rate": 1.7477548770503315e-05, "loss": 1.0352, "step": 5004 }, { "epoch": 0.25, "learning_rate": 1.747645615443063e-05, "loss": 1.0591, "step": 5005 }, { "epoch": 0.25, "learning_rate": 1.747536333594085e-05, "loss": 1.0302, "step": 5006 }, { "epoch": 0.25, "learning_rate": 1.747427031506356e-05, "loss": 0.9919, "step": 5007 }, { "epoch": 0.25, "learning_rate": 1.747317709182835e-05, "loss": 1.0097, "step": 5008 }, { "epoch": 0.25, "learning_rate": 1.7472083666264822e-05, "loss": 0.985, "step": 5009 }, { "epoch": 0.25, "learning_rate": 1.747099003840258e-05, "loss": 0.9084, "step": 5010 }, { "epoch": 0.25, "learning_rate": 1.7469896208271232e-05, "loss": 1.0233, "step": 5011 }, { "epoch": 0.25, "learning_rate": 1.746880217590039e-05, "loss": 0.9413, "step": 5012 }, { "epoch": 0.25, "learning_rate": 1.7467707941319672e-05, "loss": 0.9365, "step": 5013 }, { "epoch": 0.25, "learning_rate": 1.746661350455871e-05, "loss": 0.9495, "step": 5014 }, { "epoch": 0.25, "learning_rate": 1.7465518865647133e-05, "loss": 0.9437, "step": 5015 }, { "epoch": 0.25, "learning_rate": 1.746442402461457e-05, "loss": 0.9502, "step": 5016 }, { "epoch": 0.25, "learning_rate": 1.7463328981490675e-05, "loss": 1.0256, "step": 5017 }, { "epoch": 0.25, "learning_rate": 1.7462233736305084e-05, "loss": 1.0372, "step": 5018 }, { "epoch": 0.25, "learning_rate": 1.7461138289087454e-05, "loss": 0.9507, "step": 5019 }, { "epoch": 0.26, "learning_rate": 1.7460042639867445e-05, "loss": 0.9871, "step": 5020 }, { "epoch": 0.26, "learning_rate": 1.7458946788674714e-05, "loss": 1.0015, "step": 5021 }, { "epoch": 0.26, "learning_rate": 1.745785073553894e-05, "loss": 0.9915, "step": 5022 }, { "epoch": 0.26, "learning_rate": 1.745675448048979e-05, "loss": 1.0164, "step": 5023 }, { "epoch": 0.26, "learning_rate": 1.7455658023556946e-05, "loss": 0.9676, "step": 5024 }, { "epoch": 0.26, "learning_rate": 1.745456136477009e-05, "loss": 0.9613, "step": 5025 }, { "epoch": 0.26, "learning_rate": 1.745346450415892e-05, "loss": 0.9766, "step": 5026 }, { "epoch": 0.26, "learning_rate": 1.7452367441753126e-05, "loss": 1.0392, "step": 5027 }, { "epoch": 0.26, "learning_rate": 1.745127017758241e-05, "loss": 0.9381, "step": 5028 }, { "epoch": 0.26, "learning_rate": 1.7450172711676484e-05, "loss": 0.9809, "step": 5029 }, { "epoch": 0.26, "learning_rate": 1.744907504406506e-05, "loss": 1.0296, "step": 5030 }, { "epoch": 0.26, "learning_rate": 1.7447977174777852e-05, "loss": 1.0344, "step": 5031 }, { "epoch": 0.26, "learning_rate": 1.7446879103844586e-05, "loss": 1.1053, "step": 5032 }, { "epoch": 0.26, "learning_rate": 1.744578083129499e-05, "loss": 0.9511, "step": 5033 }, { "epoch": 0.26, "learning_rate": 1.7444682357158808e-05, "loss": 0.9704, "step": 5034 }, { "epoch": 0.26, "learning_rate": 1.7443583681465762e-05, "loss": 0.9398, "step": 5035 }, { "epoch": 0.26, "learning_rate": 1.744248480424561e-05, "loss": 0.9766, "step": 5036 }, { "epoch": 0.26, "learning_rate": 1.74413857255281e-05, "loss": 0.943, "step": 5037 }, { "epoch": 0.26, "learning_rate": 1.744028644534299e-05, "loss": 1.0627, "step": 5038 }, { "epoch": 0.26, "learning_rate": 1.743918696372004e-05, "loss": 0.9384, "step": 5039 }, { "epoch": 0.26, "learning_rate": 1.743808728068902e-05, "loss": 1.0192, "step": 5040 }, { "epoch": 0.26, "learning_rate": 1.7436987396279696e-05, "loss": 1.0043, "step": 5041 }, { "epoch": 0.26, "learning_rate": 1.743588731052185e-05, "loss": 0.9807, "step": 5042 }, { "epoch": 0.26, "learning_rate": 1.7434787023445268e-05, "loss": 1.0643, "step": 5043 }, { "epoch": 0.26, "learning_rate": 1.7433686535079736e-05, "loss": 0.9529, "step": 5044 }, { "epoch": 0.26, "learning_rate": 1.7432585845455052e-05, "loss": 0.9971, "step": 5045 }, { "epoch": 0.26, "learning_rate": 1.7431484954601012e-05, "loss": 0.9205, "step": 5046 }, { "epoch": 0.26, "learning_rate": 1.7430383862547425e-05, "loss": 0.955, "step": 5047 }, { "epoch": 0.26, "learning_rate": 1.7429282569324102e-05, "loss": 0.9467, "step": 5048 }, { "epoch": 0.26, "learning_rate": 1.7428181074960856e-05, "loss": 0.9534, "step": 5049 }, { "epoch": 0.26, "learning_rate": 1.742707937948751e-05, "loss": 1.0201, "step": 5050 }, { "epoch": 0.26, "learning_rate": 1.7425977482933893e-05, "loss": 1.0222, "step": 5051 }, { "epoch": 0.26, "learning_rate": 1.742487538532983e-05, "loss": 0.9117, "step": 5052 }, { "epoch": 0.26, "learning_rate": 1.7423773086705173e-05, "loss": 0.9814, "step": 5053 }, { "epoch": 0.26, "learning_rate": 1.7422670587089754e-05, "loss": 1.0015, "step": 5054 }, { "epoch": 0.26, "learning_rate": 1.7421567886513426e-05, "loss": 0.9146, "step": 5055 }, { "epoch": 0.26, "learning_rate": 1.7420464985006044e-05, "loss": 0.9262, "step": 5056 }, { "epoch": 0.26, "learning_rate": 1.741936188259747e-05, "loss": 0.9842, "step": 5057 }, { "epoch": 0.26, "learning_rate": 1.7418258579317563e-05, "loss": 0.9844, "step": 5058 }, { "epoch": 0.26, "learning_rate": 1.74171550751962e-05, "loss": 0.9514, "step": 5059 }, { "epoch": 0.26, "learning_rate": 1.741605137026325e-05, "loss": 1.0544, "step": 5060 }, { "epoch": 0.26, "learning_rate": 1.7414947464548603e-05, "loss": 0.8782, "step": 5061 }, { "epoch": 0.26, "learning_rate": 1.7413843358082142e-05, "loss": 1.0008, "step": 5062 }, { "epoch": 0.26, "learning_rate": 1.741273905089376e-05, "loss": 1.0381, "step": 5063 }, { "epoch": 0.26, "learning_rate": 1.7411634543013355e-05, "loss": 1.0323, "step": 5064 }, { "epoch": 0.26, "learning_rate": 1.741052983447083e-05, "loss": 1.1572, "step": 5065 }, { "epoch": 0.26, "learning_rate": 1.7409424925296093e-05, "loss": 1.0059, "step": 5066 }, { "epoch": 0.26, "learning_rate": 1.7408319815519055e-05, "loss": 0.9733, "step": 5067 }, { "epoch": 0.26, "learning_rate": 1.7407214505169647e-05, "loss": 1.1176, "step": 5068 }, { "epoch": 0.26, "learning_rate": 1.7406108994277785e-05, "loss": 0.9829, "step": 5069 }, { "epoch": 0.26, "learning_rate": 1.74050032828734e-05, "loss": 0.9855, "step": 5070 }, { "epoch": 0.26, "learning_rate": 1.740389737098643e-05, "loss": 0.9424, "step": 5071 }, { "epoch": 0.26, "learning_rate": 1.7402791258646817e-05, "loss": 1.0798, "step": 5072 }, { "epoch": 0.26, "learning_rate": 1.7401684945884504e-05, "loss": 0.9447, "step": 5073 }, { "epoch": 0.26, "learning_rate": 1.7400578432729448e-05, "loss": 1.0287, "step": 5074 }, { "epoch": 0.26, "learning_rate": 1.7399471719211602e-05, "loss": 0.9878, "step": 5075 }, { "epoch": 0.26, "learning_rate": 1.7398364805360933e-05, "loss": 0.9817, "step": 5076 }, { "epoch": 0.26, "learning_rate": 1.739725769120741e-05, "loss": 0.9832, "step": 5077 }, { "epoch": 0.26, "learning_rate": 1.7396150376781005e-05, "loss": 1.0112, "step": 5078 }, { "epoch": 0.26, "learning_rate": 1.7395042862111696e-05, "loss": 0.985, "step": 5079 }, { "epoch": 0.26, "learning_rate": 1.739393514722947e-05, "loss": 0.9725, "step": 5080 }, { "epoch": 0.26, "learning_rate": 1.7392827232164312e-05, "loss": 0.9346, "step": 5081 }, { "epoch": 0.26, "learning_rate": 1.7391719116946227e-05, "loss": 0.9931, "step": 5082 }, { "epoch": 0.26, "learning_rate": 1.739061080160521e-05, "loss": 0.9252, "step": 5083 }, { "epoch": 0.26, "learning_rate": 1.7389502286171266e-05, "loss": 0.98, "step": 5084 }, { "epoch": 0.26, "learning_rate": 1.7388393570674417e-05, "loss": 1.097, "step": 5085 }, { "epoch": 0.26, "learning_rate": 1.738728465514466e-05, "loss": 0.9529, "step": 5086 }, { "epoch": 0.26, "learning_rate": 1.7386175539612044e-05, "loss": 0.9584, "step": 5087 }, { "epoch": 0.26, "learning_rate": 1.7385066224106575e-05, "loss": 1.0087, "step": 5088 }, { "epoch": 0.26, "learning_rate": 1.73839567086583e-05, "loss": 0.9274, "step": 5089 }, { "epoch": 0.26, "learning_rate": 1.738284699329725e-05, "loss": 0.9657, "step": 5090 }, { "epoch": 0.26, "learning_rate": 1.738173707805347e-05, "loss": 0.9836, "step": 5091 }, { "epoch": 0.26, "learning_rate": 1.7380626962957018e-05, "loss": 1.0331, "step": 5092 }, { "epoch": 0.26, "learning_rate": 1.737951664803794e-05, "loss": 0.9689, "step": 5093 }, { "epoch": 0.26, "learning_rate": 1.73784061333263e-05, "loss": 0.9879, "step": 5094 }, { "epoch": 0.26, "learning_rate": 1.737729541885216e-05, "loss": 0.8865, "step": 5095 }, { "epoch": 0.26, "learning_rate": 1.7376184504645602e-05, "loss": 0.9395, "step": 5096 }, { "epoch": 0.26, "learning_rate": 1.7375073390736697e-05, "loss": 0.9129, "step": 5097 }, { "epoch": 0.26, "learning_rate": 1.7373962077155522e-05, "loss": 0.9409, "step": 5098 }, { "epoch": 0.26, "learning_rate": 1.737285056393217e-05, "loss": 0.9333, "step": 5099 }, { "epoch": 0.26, "learning_rate": 1.7371738851096732e-05, "loss": 1.1169, "step": 5100 }, { "epoch": 0.26, "learning_rate": 1.737062693867931e-05, "loss": 0.9652, "step": 5101 }, { "epoch": 0.26, "learning_rate": 1.7369514826710006e-05, "loss": 1.0082, "step": 5102 }, { "epoch": 0.26, "learning_rate": 1.7368402515218924e-05, "loss": 0.9286, "step": 5103 }, { "epoch": 0.26, "learning_rate": 1.7367290004236186e-05, "loss": 0.9064, "step": 5104 }, { "epoch": 0.26, "learning_rate": 1.736617729379191e-05, "loss": 1.0204, "step": 5105 }, { "epoch": 0.26, "learning_rate": 1.7365064383916222e-05, "loss": 1.0806, "step": 5106 }, { "epoch": 0.26, "learning_rate": 1.7363951274639247e-05, "loss": 0.9619, "step": 5107 }, { "epoch": 0.26, "learning_rate": 1.736283796599113e-05, "loss": 0.9271, "step": 5108 }, { "epoch": 0.26, "learning_rate": 1.7361724458002006e-05, "loss": 1.0145, "step": 5109 }, { "epoch": 0.26, "learning_rate": 1.7360610750702027e-05, "loss": 0.9707, "step": 5110 }, { "epoch": 0.26, "learning_rate": 1.7359496844121344e-05, "loss": 1.0975, "step": 5111 }, { "epoch": 0.26, "learning_rate": 1.735838273829011e-05, "loss": 1.082, "step": 5112 }, { "epoch": 0.26, "learning_rate": 1.735726843323849e-05, "loss": 0.9561, "step": 5113 }, { "epoch": 0.26, "learning_rate": 1.7356153928996663e-05, "loss": 0.8693, "step": 5114 }, { "epoch": 0.26, "learning_rate": 1.7355039225594787e-05, "loss": 0.9665, "step": 5115 }, { "epoch": 0.26, "learning_rate": 1.7353924323063054e-05, "loss": 0.9787, "step": 5116 }, { "epoch": 0.26, "learning_rate": 1.735280922143164e-05, "loss": 1.0355, "step": 5117 }, { "epoch": 0.26, "learning_rate": 1.7351693920730742e-05, "loss": 0.9758, "step": 5118 }, { "epoch": 0.26, "learning_rate": 1.7350578420990553e-05, "loss": 0.9258, "step": 5119 }, { "epoch": 0.26, "learning_rate": 1.7349462722241274e-05, "loss": 0.8439, "step": 5120 }, { "epoch": 0.26, "learning_rate": 1.734834682451311e-05, "loss": 0.8371, "step": 5121 }, { "epoch": 0.26, "learning_rate": 1.734723072783627e-05, "loss": 0.9172, "step": 5122 }, { "epoch": 0.26, "learning_rate": 1.734611443224098e-05, "loss": 0.9224, "step": 5123 }, { "epoch": 0.26, "learning_rate": 1.7344997937757456e-05, "loss": 0.925, "step": 5124 }, { "epoch": 0.26, "learning_rate": 1.7343881244415924e-05, "loss": 1.0146, "step": 5125 }, { "epoch": 0.26, "learning_rate": 1.7342764352246628e-05, "loss": 1.0154, "step": 5126 }, { "epoch": 0.26, "learning_rate": 1.7341647261279793e-05, "loss": 1.02, "step": 5127 }, { "epoch": 0.26, "learning_rate": 1.734052997154567e-05, "loss": 0.9654, "step": 5128 }, { "epoch": 0.26, "learning_rate": 1.733941248307451e-05, "loss": 1.0468, "step": 5129 }, { "epoch": 0.26, "learning_rate": 1.7338294795896562e-05, "loss": 1.0496, "step": 5130 }, { "epoch": 0.26, "learning_rate": 1.7337176910042093e-05, "loss": 0.9241, "step": 5131 }, { "epoch": 0.26, "learning_rate": 1.7336058825541363e-05, "loss": 0.9437, "step": 5132 }, { "epoch": 0.26, "learning_rate": 1.7334940542424648e-05, "loss": 0.9791, "step": 5133 }, { "epoch": 0.26, "learning_rate": 1.733382206072222e-05, "loss": 1.0221, "step": 5134 }, { "epoch": 0.26, "learning_rate": 1.7332703380464363e-05, "loss": 0.9941, "step": 5135 }, { "epoch": 0.26, "learning_rate": 1.7331584501681363e-05, "loss": 0.953, "step": 5136 }, { "epoch": 0.26, "learning_rate": 1.7330465424403514e-05, "loss": 0.9552, "step": 5137 }, { "epoch": 0.26, "learning_rate": 1.732934614866111e-05, "loss": 0.9695, "step": 5138 }, { "epoch": 0.26, "learning_rate": 1.7328226674484457e-05, "loss": 1.0037, "step": 5139 }, { "epoch": 0.26, "learning_rate": 1.7327107001903864e-05, "loss": 0.939, "step": 5140 }, { "epoch": 0.26, "learning_rate": 1.7325987130949645e-05, "loss": 0.9659, "step": 5141 }, { "epoch": 0.26, "learning_rate": 1.732486706165212e-05, "loss": 0.9991, "step": 5142 }, { "epoch": 0.26, "learning_rate": 1.7323746794041608e-05, "loss": 0.8902, "step": 5143 }, { "epoch": 0.26, "learning_rate": 1.7322626328148445e-05, "loss": 1.0146, "step": 5144 }, { "epoch": 0.26, "learning_rate": 1.732150566400297e-05, "loss": 0.9613, "step": 5145 }, { "epoch": 0.26, "learning_rate": 1.7320384801635516e-05, "loss": 0.9437, "step": 5146 }, { "epoch": 0.26, "learning_rate": 1.7319263741076428e-05, "loss": 1.0673, "step": 5147 }, { "epoch": 0.26, "learning_rate": 1.7318142482356066e-05, "loss": 0.9603, "step": 5148 }, { "epoch": 0.26, "learning_rate": 1.731702102550478e-05, "loss": 0.9609, "step": 5149 }, { "epoch": 0.26, "learning_rate": 1.7315899370552932e-05, "loss": 0.9922, "step": 5150 }, { "epoch": 0.26, "learning_rate": 1.7314777517530893e-05, "loss": 1.0248, "step": 5151 }, { "epoch": 0.26, "learning_rate": 1.7313655466469037e-05, "loss": 0.8772, "step": 5152 }, { "epoch": 0.26, "learning_rate": 1.731253321739774e-05, "loss": 1.0023, "step": 5153 }, { "epoch": 0.26, "learning_rate": 1.7311410770347384e-05, "loss": 1.0307, "step": 5154 }, { "epoch": 0.26, "learning_rate": 1.7310288125348364e-05, "loss": 0.8898, "step": 5155 }, { "epoch": 0.26, "learning_rate": 1.730916528243107e-05, "loss": 0.9613, "step": 5156 }, { "epoch": 0.26, "learning_rate": 1.73080422416259e-05, "loss": 1.0285, "step": 5157 }, { "epoch": 0.26, "learning_rate": 1.7306919002963264e-05, "loss": 1.092, "step": 5158 }, { "epoch": 0.26, "learning_rate": 1.7305795566473564e-05, "loss": 1.0165, "step": 5159 }, { "epoch": 0.26, "learning_rate": 1.7304671932187227e-05, "loss": 0.9907, "step": 5160 }, { "epoch": 0.26, "learning_rate": 1.730354810013467e-05, "loss": 1.0862, "step": 5161 }, { "epoch": 0.26, "learning_rate": 1.7302424070346314e-05, "loss": 0.9291, "step": 5162 }, { "epoch": 0.26, "learning_rate": 1.73012998428526e-05, "loss": 0.9006, "step": 5163 }, { "epoch": 0.26, "learning_rate": 1.7300175417683956e-05, "loss": 0.9406, "step": 5164 }, { "epoch": 0.26, "learning_rate": 1.7299050794870834e-05, "loss": 0.9852, "step": 5165 }, { "epoch": 0.26, "learning_rate": 1.7297925974443675e-05, "loss": 1.0209, "step": 5166 }, { "epoch": 0.26, "learning_rate": 1.7296800956432934e-05, "loss": 0.928, "step": 5167 }, { "epoch": 0.26, "learning_rate": 1.7295675740869073e-05, "loss": 0.9735, "step": 5168 }, { "epoch": 0.26, "learning_rate": 1.7294550327782552e-05, "loss": 0.9549, "step": 5169 }, { "epoch": 0.26, "learning_rate": 1.7293424717203842e-05, "loss": 0.9819, "step": 5170 }, { "epoch": 0.26, "learning_rate": 1.7292298909163415e-05, "loss": 1.0549, "step": 5171 }, { "epoch": 0.26, "learning_rate": 1.7291172903691755e-05, "loss": 0.9956, "step": 5172 }, { "epoch": 0.26, "learning_rate": 1.7290046700819348e-05, "loss": 0.9453, "step": 5173 }, { "epoch": 0.26, "learning_rate": 1.7288920300576676e-05, "loss": 0.9093, "step": 5174 }, { "epoch": 0.26, "learning_rate": 1.7287793702994247e-05, "loss": 0.9717, "step": 5175 }, { "epoch": 0.26, "learning_rate": 1.7286666908102558e-05, "loss": 1.0168, "step": 5176 }, { "epoch": 0.26, "learning_rate": 1.7285539915932114e-05, "loss": 1.0454, "step": 5177 }, { "epoch": 0.26, "learning_rate": 1.7284412726513426e-05, "loss": 0.9596, "step": 5178 }, { "epoch": 0.26, "learning_rate": 1.7283285339877015e-05, "loss": 0.9436, "step": 5179 }, { "epoch": 0.26, "learning_rate": 1.72821577560534e-05, "loss": 0.9087, "step": 5180 }, { "epoch": 0.26, "learning_rate": 1.7281029975073113e-05, "loss": 0.9466, "step": 5181 }, { "epoch": 0.26, "learning_rate": 1.727990199696669e-05, "loss": 0.9396, "step": 5182 }, { "epoch": 0.26, "learning_rate": 1.7278773821764658e-05, "loss": 0.9436, "step": 5183 }, { "epoch": 0.26, "learning_rate": 1.7277645449497575e-05, "loss": 0.8936, "step": 5184 }, { "epoch": 0.26, "learning_rate": 1.727651688019598e-05, "loss": 0.9598, "step": 5185 }, { "epoch": 0.26, "learning_rate": 1.7275388113890432e-05, "loss": 0.941, "step": 5186 }, { "epoch": 0.26, "learning_rate": 1.7274259150611494e-05, "loss": 1.0247, "step": 5187 }, { "epoch": 0.26, "learning_rate": 1.7273129990389728e-05, "loss": 0.9628, "step": 5188 }, { "epoch": 0.26, "learning_rate": 1.7272000633255706e-05, "loss": 1.0701, "step": 5189 }, { "epoch": 0.26, "learning_rate": 1.727087107924e-05, "loss": 1.028, "step": 5190 }, { "epoch": 0.26, "learning_rate": 1.72697413283732e-05, "loss": 0.9288, "step": 5191 }, { "epoch": 0.26, "learning_rate": 1.7268611380685886e-05, "loss": 0.9845, "step": 5192 }, { "epoch": 0.26, "learning_rate": 1.726748123620865e-05, "loss": 1.0369, "step": 5193 }, { "epoch": 0.26, "learning_rate": 1.7266350894972096e-05, "loss": 0.9501, "step": 5194 }, { "epoch": 0.26, "learning_rate": 1.7265220357006816e-05, "loss": 0.9831, "step": 5195 }, { "epoch": 0.26, "learning_rate": 1.726408962234343e-05, "loss": 1.002, "step": 5196 }, { "epoch": 0.26, "learning_rate": 1.7262958691012547e-05, "loss": 0.926, "step": 5197 }, { "epoch": 0.26, "learning_rate": 1.726182756304478e-05, "loss": 0.9716, "step": 5198 }, { "epoch": 0.26, "learning_rate": 1.7260696238470757e-05, "loss": 0.9778, "step": 5199 }, { "epoch": 0.26, "learning_rate": 1.7259564717321115e-05, "loss": 0.9913, "step": 5200 }, { "epoch": 0.26, "learning_rate": 1.7258432999626476e-05, "loss": 0.997, "step": 5201 }, { "epoch": 0.26, "learning_rate": 1.7257301085417487e-05, "loss": 0.9584, "step": 5202 }, { "epoch": 0.26, "learning_rate": 1.7256168974724792e-05, "loss": 0.9057, "step": 5203 }, { "epoch": 0.26, "learning_rate": 1.725503666757904e-05, "loss": 0.8971, "step": 5204 }, { "epoch": 0.26, "learning_rate": 1.7253904164010893e-05, "loss": 0.9336, "step": 5205 }, { "epoch": 0.26, "learning_rate": 1.7252771464051008e-05, "loss": 0.9361, "step": 5206 }, { "epoch": 0.26, "learning_rate": 1.725163856773005e-05, "loss": 0.933, "step": 5207 }, { "epoch": 0.26, "learning_rate": 1.7250505475078693e-05, "loss": 0.9606, "step": 5208 }, { "epoch": 0.26, "learning_rate": 1.7249372186127614e-05, "loss": 1.016, "step": 5209 }, { "epoch": 0.26, "learning_rate": 1.7248238700907498e-05, "loss": 0.9906, "step": 5210 }, { "epoch": 0.26, "learning_rate": 1.724710501944903e-05, "loss": 0.9518, "step": 5211 }, { "epoch": 0.26, "learning_rate": 1.72459711417829e-05, "loss": 1.0607, "step": 5212 }, { "epoch": 0.26, "learning_rate": 1.7244837067939814e-05, "loss": 0.8874, "step": 5213 }, { "epoch": 0.26, "learning_rate": 1.7243702797950472e-05, "loss": 0.9293, "step": 5214 }, { "epoch": 0.26, "learning_rate": 1.7242568331845583e-05, "loss": 0.9198, "step": 5215 }, { "epoch": 0.26, "learning_rate": 1.7241433669655863e-05, "loss": 0.9548, "step": 5216 }, { "epoch": 0.27, "learning_rate": 1.724029881141203e-05, "loss": 0.9713, "step": 5217 }, { "epoch": 0.27, "learning_rate": 1.723916375714481e-05, "loss": 0.9023, "step": 5218 }, { "epoch": 0.27, "learning_rate": 1.7238028506884932e-05, "loss": 1.0575, "step": 5219 }, { "epoch": 0.27, "learning_rate": 1.7236893060663136e-05, "loss": 0.9965, "step": 5220 }, { "epoch": 0.27, "learning_rate": 1.7235757418510157e-05, "loss": 0.9949, "step": 5221 }, { "epoch": 0.27, "learning_rate": 1.7234621580456744e-05, "loss": 1.0279, "step": 5222 }, { "epoch": 0.27, "learning_rate": 1.723348554653365e-05, "loss": 1.0825, "step": 5223 }, { "epoch": 0.27, "learning_rate": 1.7232349316771628e-05, "loss": 0.9691, "step": 5224 }, { "epoch": 0.27, "learning_rate": 1.7231212891201444e-05, "loss": 0.9314, "step": 5225 }, { "epoch": 0.27, "learning_rate": 1.723007626985387e-05, "loss": 0.9658, "step": 5226 }, { "epoch": 0.27, "learning_rate": 1.7228939452759666e-05, "loss": 0.9728, "step": 5227 }, { "epoch": 0.27, "learning_rate": 1.7227802439949618e-05, "loss": 0.9288, "step": 5228 }, { "epoch": 0.27, "learning_rate": 1.7226665231454514e-05, "loss": 0.9243, "step": 5229 }, { "epoch": 0.27, "learning_rate": 1.7225527827305133e-05, "loss": 0.9585, "step": 5230 }, { "epoch": 0.27, "learning_rate": 1.7224390227532274e-05, "loss": 0.9845, "step": 5231 }, { "epoch": 0.27, "learning_rate": 1.7223252432166734e-05, "loss": 1.0128, "step": 5232 }, { "epoch": 0.27, "learning_rate": 1.722211444123932e-05, "loss": 1.0315, "step": 5233 }, { "epoch": 0.27, "learning_rate": 1.7220976254780844e-05, "loss": 0.902, "step": 5234 }, { "epoch": 0.27, "learning_rate": 1.7219837872822117e-05, "loss": 1.0841, "step": 5235 }, { "epoch": 0.27, "learning_rate": 1.721869929539396e-05, "loss": 0.9678, "step": 5236 }, { "epoch": 0.27, "learning_rate": 1.72175605225272e-05, "loss": 1.0153, "step": 5237 }, { "epoch": 0.27, "learning_rate": 1.7216421554252673e-05, "loss": 0.9177, "step": 5238 }, { "epoch": 0.27, "learning_rate": 1.7215282390601202e-05, "loss": 0.9827, "step": 5239 }, { "epoch": 0.27, "learning_rate": 1.7214143031603642e-05, "loss": 1.0007, "step": 5240 }, { "epoch": 0.27, "learning_rate": 1.7213003477290832e-05, "loss": 1.0124, "step": 5241 }, { "epoch": 0.27, "learning_rate": 1.7211863727693627e-05, "loss": 0.9742, "step": 5242 }, { "epoch": 0.27, "learning_rate": 1.7210723782842884e-05, "loss": 1.0281, "step": 5243 }, { "epoch": 0.27, "learning_rate": 1.720958364276947e-05, "loss": 0.982, "step": 5244 }, { "epoch": 0.27, "learning_rate": 1.7208443307504246e-05, "loss": 0.9787, "step": 5245 }, { "epoch": 0.27, "learning_rate": 1.720730277707809e-05, "loss": 0.9196, "step": 5246 }, { "epoch": 0.27, "learning_rate": 1.7206162051521878e-05, "loss": 0.9511, "step": 5247 }, { "epoch": 0.27, "learning_rate": 1.720502113086649e-05, "loss": 1.1207, "step": 5248 }, { "epoch": 0.27, "learning_rate": 1.720388001514283e-05, "loss": 0.9843, "step": 5249 }, { "epoch": 0.27, "learning_rate": 1.7202738704381782e-05, "loss": 0.9336, "step": 5250 }, { "epoch": 0.27, "learning_rate": 1.720159719861424e-05, "loss": 0.9359, "step": 5251 }, { "epoch": 0.27, "learning_rate": 1.720045549787112e-05, "loss": 1.0693, "step": 5252 }, { "epoch": 0.27, "learning_rate": 1.719931360218333e-05, "loss": 1.0023, "step": 5253 }, { "epoch": 0.27, "learning_rate": 1.719817151158178e-05, "loss": 1.0585, "step": 5254 }, { "epoch": 0.27, "learning_rate": 1.7197029226097394e-05, "loss": 0.991, "step": 5255 }, { "epoch": 0.27, "learning_rate": 1.7195886745761103e-05, "loss": 0.9146, "step": 5256 }, { "epoch": 0.27, "learning_rate": 1.7194744070603832e-05, "loss": 0.9279, "step": 5257 }, { "epoch": 0.27, "learning_rate": 1.719360120065652e-05, "loss": 0.8957, "step": 5258 }, { "epoch": 0.27, "learning_rate": 1.7192458135950106e-05, "loss": 0.9442, "step": 5259 }, { "epoch": 0.27, "learning_rate": 1.7191314876515543e-05, "loss": 0.8972, "step": 5260 }, { "epoch": 0.27, "learning_rate": 1.719017142238378e-05, "loss": 0.9706, "step": 5261 }, { "epoch": 0.27, "learning_rate": 1.7189027773585777e-05, "loss": 0.9726, "step": 5262 }, { "epoch": 0.27, "learning_rate": 1.7187883930152492e-05, "loss": 0.957, "step": 5263 }, { "epoch": 0.27, "learning_rate": 1.71867398921149e-05, "loss": 0.9255, "step": 5264 }, { "epoch": 0.27, "learning_rate": 1.718559565950397e-05, "loss": 0.9676, "step": 5265 }, { "epoch": 0.27, "learning_rate": 1.7184451232350686e-05, "loss": 0.9481, "step": 5266 }, { "epoch": 0.27, "learning_rate": 1.7183306610686023e-05, "loss": 0.9514, "step": 5267 }, { "epoch": 0.27, "learning_rate": 1.718216179454098e-05, "loss": 1.0712, "step": 5268 }, { "epoch": 0.27, "learning_rate": 1.7181016783946544e-05, "loss": 0.9669, "step": 5269 }, { "epoch": 0.27, "learning_rate": 1.7179871578933723e-05, "loss": 1.0427, "step": 5270 }, { "epoch": 0.27, "learning_rate": 1.7178726179533515e-05, "loss": 0.9153, "step": 5271 }, { "epoch": 0.27, "learning_rate": 1.7177580585776932e-05, "loss": 0.8602, "step": 5272 }, { "epoch": 0.27, "learning_rate": 1.7176434797694993e-05, "loss": 0.8981, "step": 5273 }, { "epoch": 0.27, "learning_rate": 1.7175288815318718e-05, "loss": 1.0009, "step": 5274 }, { "epoch": 0.27, "learning_rate": 1.7174142638679133e-05, "loss": 0.9671, "step": 5275 }, { "epoch": 0.27, "learning_rate": 1.7172996267807266e-05, "loss": 0.9781, "step": 5276 }, { "epoch": 0.27, "learning_rate": 1.7171849702734157e-05, "loss": 1.0082, "step": 5277 }, { "epoch": 0.27, "learning_rate": 1.7170702943490847e-05, "loss": 0.9665, "step": 5278 }, { "epoch": 0.27, "learning_rate": 1.7169555990108385e-05, "loss": 0.9231, "step": 5279 }, { "epoch": 0.27, "learning_rate": 1.716840884261782e-05, "loss": 0.9885, "step": 5280 }, { "epoch": 0.27, "learning_rate": 1.7167261501050215e-05, "loss": 0.8743, "step": 5281 }, { "epoch": 0.27, "learning_rate": 1.716611396543663e-05, "loss": 1.0537, "step": 5282 }, { "epoch": 0.27, "learning_rate": 1.716496623580813e-05, "loss": 0.9298, "step": 5283 }, { "epoch": 0.27, "learning_rate": 1.7163818312195798e-05, "loss": 0.9987, "step": 5284 }, { "epoch": 0.27, "learning_rate": 1.7162670194630704e-05, "loss": 0.9684, "step": 5285 }, { "epoch": 0.27, "learning_rate": 1.7161521883143936e-05, "loss": 0.96, "step": 5286 }, { "epoch": 0.27, "learning_rate": 1.716037337776658e-05, "loss": 0.8172, "step": 5287 }, { "epoch": 0.27, "learning_rate": 1.7159224678529734e-05, "loss": 0.9981, "step": 5288 }, { "epoch": 0.27, "learning_rate": 1.7158075785464497e-05, "loss": 0.9409, "step": 5289 }, { "epoch": 0.27, "learning_rate": 1.7156926698601973e-05, "loss": 0.9543, "step": 5290 }, { "epoch": 0.27, "learning_rate": 1.7155777417973273e-05, "loss": 0.9322, "step": 5291 }, { "epoch": 0.27, "learning_rate": 1.7154627943609513e-05, "loss": 0.9786, "step": 5292 }, { "epoch": 0.27, "learning_rate": 1.7153478275541814e-05, "loss": 0.9894, "step": 5293 }, { "epoch": 0.27, "learning_rate": 1.71523284138013e-05, "loss": 0.9677, "step": 5294 }, { "epoch": 0.27, "learning_rate": 1.7151178358419104e-05, "loss": 1.0807, "step": 5295 }, { "epoch": 0.27, "learning_rate": 1.7150028109426364e-05, "loss": 0.9166, "step": 5296 }, { "epoch": 0.27, "learning_rate": 1.7148877666854218e-05, "loss": 0.9887, "step": 5297 }, { "epoch": 0.27, "learning_rate": 1.7147727030733814e-05, "loss": 0.8857, "step": 5298 }, { "epoch": 0.27, "learning_rate": 1.714657620109631e-05, "loss": 1.0881, "step": 5299 }, { "epoch": 0.27, "learning_rate": 1.7145425177972856e-05, "loss": 1.0334, "step": 5300 }, { "epoch": 0.27, "learning_rate": 1.7144273961394617e-05, "loss": 0.9536, "step": 5301 }, { "epoch": 0.27, "learning_rate": 1.7143122551392755e-05, "loss": 0.925, "step": 5302 }, { "epoch": 0.27, "learning_rate": 1.714197094799846e-05, "loss": 1.0209, "step": 5303 }, { "epoch": 0.27, "learning_rate": 1.7140819151242895e-05, "loss": 1.0511, "step": 5304 }, { "epoch": 0.27, "learning_rate": 1.713966716115725e-05, "loss": 0.9494, "step": 5305 }, { "epoch": 0.27, "learning_rate": 1.713851497777271e-05, "loss": 0.9806, "step": 5306 }, { "epoch": 0.27, "learning_rate": 1.7137362601120474e-05, "loss": 0.9568, "step": 5307 }, { "epoch": 0.27, "learning_rate": 1.7136210031231742e-05, "loss": 0.9664, "step": 5308 }, { "epoch": 0.27, "learning_rate": 1.7135057268137712e-05, "loss": 0.9465, "step": 5309 }, { "epoch": 0.27, "learning_rate": 1.71339043118696e-05, "loss": 0.8785, "step": 5310 }, { "epoch": 0.27, "learning_rate": 1.7132751162458614e-05, "loss": 1.0425, "step": 5311 }, { "epoch": 0.27, "learning_rate": 1.7131597819935983e-05, "loss": 0.9826, "step": 5312 }, { "epoch": 0.27, "learning_rate": 1.713044428433293e-05, "loss": 1.0376, "step": 5313 }, { "epoch": 0.27, "learning_rate": 1.712929055568068e-05, "loss": 1.028, "step": 5314 }, { "epoch": 0.27, "learning_rate": 1.7128136634010474e-05, "loss": 1.0148, "step": 5315 }, { "epoch": 0.27, "learning_rate": 1.7126982519353554e-05, "loss": 0.9316, "step": 5316 }, { "epoch": 0.27, "learning_rate": 1.7125828211741162e-05, "loss": 0.9749, "step": 5317 }, { "epoch": 0.27, "learning_rate": 1.7124673711204555e-05, "loss": 0.9627, "step": 5318 }, { "epoch": 0.27, "learning_rate": 1.7123519017774988e-05, "loss": 1.1201, "step": 5319 }, { "epoch": 0.27, "learning_rate": 1.712236413148372e-05, "loss": 0.9599, "step": 5320 }, { "epoch": 0.27, "learning_rate": 1.712120905236202e-05, "loss": 0.8573, "step": 5321 }, { "epoch": 0.27, "learning_rate": 1.7120053780441166e-05, "loss": 0.9752, "step": 5322 }, { "epoch": 0.27, "learning_rate": 1.7118898315752427e-05, "loss": 1.0063, "step": 5323 }, { "epoch": 0.27, "learning_rate": 1.7117742658327094e-05, "loss": 0.998, "step": 5324 }, { "epoch": 0.27, "learning_rate": 1.7116586808196446e-05, "loss": 0.989, "step": 5325 }, { "epoch": 0.27, "learning_rate": 1.7115430765391786e-05, "loss": 0.9855, "step": 5326 }, { "epoch": 0.27, "learning_rate": 1.7114274529944407e-05, "loss": 0.9068, "step": 5327 }, { "epoch": 0.27, "learning_rate": 1.7113118101885612e-05, "loss": 0.9176, "step": 5328 }, { "epoch": 0.27, "learning_rate": 1.7111961481246714e-05, "loss": 0.9858, "step": 5329 }, { "epoch": 0.27, "learning_rate": 1.7110804668059026e-05, "loss": 0.8558, "step": 5330 }, { "epoch": 0.27, "learning_rate": 1.7109647662353864e-05, "loss": 1.0621, "step": 5331 }, { "epoch": 0.27, "learning_rate": 1.710849046416256e-05, "loss": 0.938, "step": 5332 }, { "epoch": 0.27, "learning_rate": 1.7107333073516434e-05, "loss": 0.9202, "step": 5333 }, { "epoch": 0.27, "learning_rate": 1.710617549044683e-05, "loss": 0.9176, "step": 5334 }, { "epoch": 0.27, "learning_rate": 1.7105017714985088e-05, "loss": 0.9425, "step": 5335 }, { "epoch": 0.27, "learning_rate": 1.7103859747162547e-05, "loss": 0.9556, "step": 5336 }, { "epoch": 0.27, "learning_rate": 1.710270158701056e-05, "loss": 1.0136, "step": 5337 }, { "epoch": 0.27, "learning_rate": 1.7101543234560486e-05, "loss": 0.9427, "step": 5338 }, { "epoch": 0.27, "learning_rate": 1.7100384689843683e-05, "loss": 0.9532, "step": 5339 }, { "epoch": 0.27, "learning_rate": 1.709922595289152e-05, "loss": 0.9533, "step": 5340 }, { "epoch": 0.27, "learning_rate": 1.709806702373537e-05, "loss": 0.9501, "step": 5341 }, { "epoch": 0.27, "learning_rate": 1.70969079024066e-05, "loss": 0.9621, "step": 5342 }, { "epoch": 0.27, "learning_rate": 1.7095748588936602e-05, "loss": 0.9678, "step": 5343 }, { "epoch": 0.27, "learning_rate": 1.7094589083356764e-05, "loss": 0.9572, "step": 5344 }, { "epoch": 0.27, "learning_rate": 1.7093429385698473e-05, "loss": 0.9879, "step": 5345 }, { "epoch": 0.27, "learning_rate": 1.7092269495993126e-05, "loss": 0.9221, "step": 5346 }, { "epoch": 0.27, "learning_rate": 1.709110941427213e-05, "loss": 0.8775, "step": 5347 }, { "epoch": 0.27, "learning_rate": 1.7089949140566887e-05, "loss": 0.9836, "step": 5348 }, { "epoch": 0.27, "learning_rate": 1.7088788674908817e-05, "loss": 0.9335, "step": 5349 }, { "epoch": 0.27, "learning_rate": 1.7087628017329337e-05, "loss": 1.0695, "step": 5350 }, { "epoch": 0.27, "learning_rate": 1.708646716785987e-05, "loss": 1.0599, "step": 5351 }, { "epoch": 0.27, "learning_rate": 1.708530612653184e-05, "loss": 0.9052, "step": 5352 }, { "epoch": 0.27, "learning_rate": 1.7084144893376686e-05, "loss": 0.9847, "step": 5353 }, { "epoch": 0.27, "learning_rate": 1.7082983468425848e-05, "loss": 0.941, "step": 5354 }, { "epoch": 0.27, "learning_rate": 1.708182185171077e-05, "loss": 1.0152, "step": 5355 }, { "epoch": 0.27, "learning_rate": 1.7080660043262893e-05, "loss": 0.9075, "step": 5356 }, { "epoch": 0.27, "learning_rate": 1.707949804311368e-05, "loss": 0.9198, "step": 5357 }, { "epoch": 0.27, "learning_rate": 1.707833585129459e-05, "loss": 1.0246, "step": 5358 }, { "epoch": 0.27, "learning_rate": 1.7077173467837092e-05, "loss": 1.0099, "step": 5359 }, { "epoch": 0.27, "learning_rate": 1.7076010892772653e-05, "loss": 0.9774, "step": 5360 }, { "epoch": 0.27, "learning_rate": 1.7074848126132743e-05, "loss": 0.9142, "step": 5361 }, { "epoch": 0.27, "learning_rate": 1.7073685167948847e-05, "loss": 0.9954, "step": 5362 }, { "epoch": 0.27, "learning_rate": 1.7072522018252452e-05, "loss": 1.0101, "step": 5363 }, { "epoch": 0.27, "learning_rate": 1.7071358677075043e-05, "loss": 0.9839, "step": 5364 }, { "epoch": 0.27, "learning_rate": 1.7070195144448124e-05, "loss": 1.0272, "step": 5365 }, { "epoch": 0.27, "learning_rate": 1.7069031420403197e-05, "loss": 0.9077, "step": 5366 }, { "epoch": 0.27, "learning_rate": 1.7067867504971762e-05, "loss": 0.9726, "step": 5367 }, { "epoch": 0.27, "learning_rate": 1.706670339818534e-05, "loss": 0.9508, "step": 5368 }, { "epoch": 0.27, "learning_rate": 1.7065539100075433e-05, "loss": 0.8691, "step": 5369 }, { "epoch": 0.27, "learning_rate": 1.7064374610673578e-05, "loss": 0.9526, "step": 5370 }, { "epoch": 0.27, "learning_rate": 1.7063209930011294e-05, "loss": 1.0673, "step": 5371 }, { "epoch": 0.27, "learning_rate": 1.7062045058120114e-05, "loss": 0.9763, "step": 5372 }, { "epoch": 0.27, "learning_rate": 1.7060879995031582e-05, "loss": 0.9493, "step": 5373 }, { "epoch": 0.27, "learning_rate": 1.7059714740777235e-05, "loss": 1.0229, "step": 5374 }, { "epoch": 0.27, "learning_rate": 1.705854929538862e-05, "loss": 0.9116, "step": 5375 }, { "epoch": 0.27, "learning_rate": 1.7057383658897294e-05, "loss": 0.8871, "step": 5376 }, { "epoch": 0.27, "learning_rate": 1.7056217831334812e-05, "loss": 0.9566, "step": 5377 }, { "epoch": 0.27, "learning_rate": 1.7055051812732742e-05, "loss": 1.0339, "step": 5378 }, { "epoch": 0.27, "learning_rate": 1.705388560312265e-05, "loss": 0.9593, "step": 5379 }, { "epoch": 0.27, "learning_rate": 1.7052719202536107e-05, "loss": 0.8992, "step": 5380 }, { "epoch": 0.27, "learning_rate": 1.70515526110047e-05, "loss": 1.0213, "step": 5381 }, { "epoch": 0.27, "learning_rate": 1.7050385828560007e-05, "loss": 1.0268, "step": 5382 }, { "epoch": 0.27, "learning_rate": 1.7049218855233617e-05, "loss": 1.0264, "step": 5383 }, { "epoch": 0.27, "learning_rate": 1.7048051691057125e-05, "loss": 0.9346, "step": 5384 }, { "epoch": 0.27, "learning_rate": 1.7046884336062132e-05, "loss": 0.964, "step": 5385 }, { "epoch": 0.27, "learning_rate": 1.7045716790280248e-05, "loss": 0.9489, "step": 5386 }, { "epoch": 0.27, "learning_rate": 1.7044549053743075e-05, "loss": 1.1294, "step": 5387 }, { "epoch": 0.27, "learning_rate": 1.7043381126482227e-05, "loss": 1.0121, "step": 5388 }, { "epoch": 0.27, "learning_rate": 1.7042213008529334e-05, "loss": 1.0242, "step": 5389 }, { "epoch": 0.27, "learning_rate": 1.7041044699916015e-05, "loss": 0.9213, "step": 5390 }, { "epoch": 0.27, "learning_rate": 1.7039876200673903e-05, "loss": 1.0014, "step": 5391 }, { "epoch": 0.27, "learning_rate": 1.703870751083463e-05, "loss": 0.9437, "step": 5392 }, { "epoch": 0.27, "learning_rate": 1.703753863042984e-05, "loss": 0.887, "step": 5393 }, { "epoch": 0.27, "learning_rate": 1.7036369559491178e-05, "loss": 0.9116, "step": 5394 }, { "epoch": 0.27, "learning_rate": 1.7035200298050297e-05, "loss": 0.9904, "step": 5395 }, { "epoch": 0.27, "learning_rate": 1.7034030846138855e-05, "loss": 0.9737, "step": 5396 }, { "epoch": 0.27, "learning_rate": 1.703286120378851e-05, "loss": 0.9873, "step": 5397 }, { "epoch": 0.27, "learning_rate": 1.703169137103093e-05, "loss": 0.9432, "step": 5398 }, { "epoch": 0.27, "learning_rate": 1.7030521347897787e-05, "loss": 0.9695, "step": 5399 }, { "epoch": 0.27, "learning_rate": 1.702935113442076e-05, "loss": 0.9657, "step": 5400 }, { "epoch": 0.27, "learning_rate": 1.702818073063153e-05, "loss": 1.013, "step": 5401 }, { "epoch": 0.27, "learning_rate": 1.702701013656178e-05, "loss": 1.0224, "step": 5402 }, { "epoch": 0.27, "learning_rate": 1.7025839352243214e-05, "loss": 0.9867, "step": 5403 }, { "epoch": 0.27, "learning_rate": 1.7024668377707515e-05, "loss": 1.0392, "step": 5404 }, { "epoch": 0.27, "learning_rate": 1.7023497212986396e-05, "loss": 1.0452, "step": 5405 }, { "epoch": 0.27, "learning_rate": 1.7022325858111565e-05, "loss": 0.9803, "step": 5406 }, { "epoch": 0.27, "learning_rate": 1.702115431311473e-05, "loss": 1.0244, "step": 5407 }, { "epoch": 0.27, "learning_rate": 1.7019982578027613e-05, "loss": 0.907, "step": 5408 }, { "epoch": 0.27, "learning_rate": 1.7018810652881935e-05, "loss": 0.9229, "step": 5409 }, { "epoch": 0.27, "learning_rate": 1.7017638537709426e-05, "loss": 1.0086, "step": 5410 }, { "epoch": 0.27, "learning_rate": 1.701646623254182e-05, "loss": 0.944, "step": 5411 }, { "epoch": 0.27, "learning_rate": 1.7015293737410855e-05, "loss": 1.0118, "step": 5412 }, { "epoch": 0.27, "learning_rate": 1.7014121052348277e-05, "loss": 0.9883, "step": 5413 }, { "epoch": 0.28, "learning_rate": 1.7012948177385835e-05, "loss": 0.9916, "step": 5414 }, { "epoch": 0.28, "learning_rate": 1.7011775112555283e-05, "loss": 0.954, "step": 5415 }, { "epoch": 0.28, "learning_rate": 1.7010601857888375e-05, "loss": 0.9319, "step": 5416 }, { "epoch": 0.28, "learning_rate": 1.7009428413416883e-05, "loss": 1.007, "step": 5417 }, { "epoch": 0.28, "learning_rate": 1.7008254779172576e-05, "loss": 1.0427, "step": 5418 }, { "epoch": 0.28, "learning_rate": 1.7007080955187224e-05, "loss": 1.0145, "step": 5419 }, { "epoch": 0.28, "learning_rate": 1.700590694149261e-05, "loss": 0.9495, "step": 5420 }, { "epoch": 0.28, "learning_rate": 1.7004732738120523e-05, "loss": 1.0266, "step": 5421 }, { "epoch": 0.28, "learning_rate": 1.7003558345102747e-05, "loss": 1.0116, "step": 5422 }, { "epoch": 0.28, "learning_rate": 1.700238376247108e-05, "loss": 1.0113, "step": 5423 }, { "epoch": 0.28, "learning_rate": 1.7001208990257322e-05, "loss": 0.9687, "step": 5424 }, { "epoch": 0.28, "learning_rate": 1.700003402849328e-05, "loss": 1.1125, "step": 5425 }, { "epoch": 0.28, "learning_rate": 1.6998858877210765e-05, "loss": 0.9306, "step": 5426 }, { "epoch": 0.28, "learning_rate": 1.699768353644159e-05, "loss": 0.937, "step": 5427 }, { "epoch": 0.28, "learning_rate": 1.6996508006217582e-05, "loss": 1.0828, "step": 5428 }, { "epoch": 0.28, "learning_rate": 1.699533228657056e-05, "loss": 1.0952, "step": 5429 }, { "epoch": 0.28, "learning_rate": 1.699415637753236e-05, "loss": 1.0344, "step": 5430 }, { "epoch": 0.28, "learning_rate": 1.6992980279134815e-05, "loss": 1.0142, "step": 5431 }, { "epoch": 0.28, "learning_rate": 1.6991803991409774e-05, "loss": 1.019, "step": 5432 }, { "epoch": 0.28, "learning_rate": 1.6990627514389075e-05, "loss": 0.9976, "step": 5433 }, { "epoch": 0.28, "learning_rate": 1.6989450848104573e-05, "loss": 0.9452, "step": 5434 }, { "epoch": 0.28, "learning_rate": 1.6988273992588126e-05, "loss": 0.8909, "step": 5435 }, { "epoch": 0.28, "learning_rate": 1.6987096947871594e-05, "loss": 0.9542, "step": 5436 }, { "epoch": 0.28, "learning_rate": 1.698591971398685e-05, "loss": 0.985, "step": 5437 }, { "epoch": 0.28, "learning_rate": 1.698474229096576e-05, "loss": 0.9937, "step": 5438 }, { "epoch": 0.28, "learning_rate": 1.6983564678840206e-05, "loss": 0.9548, "step": 5439 }, { "epoch": 0.28, "learning_rate": 1.6982386877642065e-05, "loss": 0.9344, "step": 5440 }, { "epoch": 0.28, "learning_rate": 1.698120888740323e-05, "loss": 0.9445, "step": 5441 }, { "epoch": 0.28, "learning_rate": 1.698003070815559e-05, "loss": 0.9919, "step": 5442 }, { "epoch": 0.28, "learning_rate": 1.6978852339931045e-05, "loss": 0.9782, "step": 5443 }, { "epoch": 0.28, "learning_rate": 1.6977673782761498e-05, "loss": 1.0383, "step": 5444 }, { "epoch": 0.28, "learning_rate": 1.6976495036678856e-05, "loss": 0.9741, "step": 5445 }, { "epoch": 0.28, "learning_rate": 1.6975316101715037e-05, "loss": 0.9245, "step": 5446 }, { "epoch": 0.28, "learning_rate": 1.6974136977901952e-05, "loss": 0.9171, "step": 5447 }, { "epoch": 0.28, "learning_rate": 1.697295766527153e-05, "loss": 0.8997, "step": 5448 }, { "epoch": 0.28, "learning_rate": 1.69717781638557e-05, "loss": 1.0173, "step": 5449 }, { "epoch": 0.28, "learning_rate": 1.697059847368639e-05, "loss": 0.9518, "step": 5450 }, { "epoch": 0.28, "learning_rate": 1.696941859479554e-05, "loss": 0.9541, "step": 5451 }, { "epoch": 0.28, "learning_rate": 1.69682385272151e-05, "loss": 0.9041, "step": 5452 }, { "epoch": 0.28, "learning_rate": 1.6967058270977018e-05, "loss": 1.0358, "step": 5453 }, { "epoch": 0.28, "learning_rate": 1.6965877826113243e-05, "loss": 1.0356, "step": 5454 }, { "epoch": 0.28, "learning_rate": 1.696469719265574e-05, "loss": 1.009, "step": 5455 }, { "epoch": 0.28, "learning_rate": 1.6963516370636462e-05, "loss": 0.9261, "step": 5456 }, { "epoch": 0.28, "learning_rate": 1.6962335360087395e-05, "loss": 1.0237, "step": 5457 }, { "epoch": 0.28, "learning_rate": 1.6961154161040503e-05, "loss": 0.9172, "step": 5458 }, { "epoch": 0.28, "learning_rate": 1.6959972773527767e-05, "loss": 0.9579, "step": 5459 }, { "epoch": 0.28, "learning_rate": 1.6958791197581175e-05, "loss": 0.9835, "step": 5460 }, { "epoch": 0.28, "learning_rate": 1.6957609433232714e-05, "loss": 0.9955, "step": 5461 }, { "epoch": 0.28, "learning_rate": 1.695642748051438e-05, "loss": 0.9503, "step": 5462 }, { "epoch": 0.28, "learning_rate": 1.695524533945817e-05, "loss": 0.9791, "step": 5463 }, { "epoch": 0.28, "learning_rate": 1.6954063010096097e-05, "loss": 0.954, "step": 5464 }, { "epoch": 0.28, "learning_rate": 1.6952880492460165e-05, "loss": 0.995, "step": 5465 }, { "epoch": 0.28, "learning_rate": 1.695169778658239e-05, "loss": 0.9997, "step": 5466 }, { "epoch": 0.28, "learning_rate": 1.6950514892494793e-05, "loss": 0.9799, "step": 5467 }, { "epoch": 0.28, "learning_rate": 1.69493318102294e-05, "loss": 1.0092, "step": 5468 }, { "epoch": 0.28, "learning_rate": 1.6948148539818244e-05, "loss": 1.006, "step": 5469 }, { "epoch": 0.28, "learning_rate": 1.6946965081293357e-05, "loss": 0.9957, "step": 5470 }, { "epoch": 0.28, "learning_rate": 1.6945781434686783e-05, "loss": 0.9485, "step": 5471 }, { "epoch": 0.28, "learning_rate": 1.6944597600030563e-05, "loss": 0.9657, "step": 5472 }, { "epoch": 0.28, "learning_rate": 1.6943413577356756e-05, "loss": 0.9565, "step": 5473 }, { "epoch": 0.28, "learning_rate": 1.694222936669741e-05, "loss": 0.9583, "step": 5474 }, { "epoch": 0.28, "learning_rate": 1.694104496808459e-05, "loss": 0.9615, "step": 5475 }, { "epoch": 0.28, "learning_rate": 1.6939860381550363e-05, "loss": 0.9389, "step": 5476 }, { "epoch": 0.28, "learning_rate": 1.69386756071268e-05, "loss": 0.9341, "step": 5477 }, { "epoch": 0.28, "learning_rate": 1.6937490644845977e-05, "loss": 1.0086, "step": 5478 }, { "epoch": 0.28, "learning_rate": 1.6936305494739974e-05, "loss": 1.1156, "step": 5479 }, { "epoch": 0.28, "learning_rate": 1.6935120156840884e-05, "loss": 0.9193, "step": 5480 }, { "epoch": 0.28, "learning_rate": 1.693393463118079e-05, "loss": 1.038, "step": 5481 }, { "epoch": 0.28, "learning_rate": 1.6932748917791792e-05, "loss": 0.9524, "step": 5482 }, { "epoch": 0.28, "learning_rate": 1.6931563016705998e-05, "loss": 0.9987, "step": 5483 }, { "epoch": 0.28, "learning_rate": 1.6930376927955504e-05, "loss": 1.0292, "step": 5484 }, { "epoch": 0.28, "learning_rate": 1.692919065157243e-05, "loss": 1.1224, "step": 5485 }, { "epoch": 0.28, "learning_rate": 1.6928004187588896e-05, "loss": 0.9329, "step": 5486 }, { "epoch": 0.28, "learning_rate": 1.6926817536037013e-05, "loss": 0.9723, "step": 5487 }, { "epoch": 0.28, "learning_rate": 1.692563069694892e-05, "loss": 0.9666, "step": 5488 }, { "epoch": 0.28, "learning_rate": 1.692444367035674e-05, "loss": 1.0743, "step": 5489 }, { "epoch": 0.28, "learning_rate": 1.6923256456292618e-05, "loss": 1.1152, "step": 5490 }, { "epoch": 0.28, "learning_rate": 1.692206905478869e-05, "loss": 0.9712, "step": 5491 }, { "epoch": 0.28, "learning_rate": 1.6920881465877107e-05, "loss": 0.9301, "step": 5492 }, { "epoch": 0.28, "learning_rate": 1.691969368959002e-05, "loss": 0.9641, "step": 5493 }, { "epoch": 0.28, "learning_rate": 1.6918505725959594e-05, "loss": 0.9378, "step": 5494 }, { "epoch": 0.28, "learning_rate": 1.6917317575017984e-05, "loss": 0.987, "step": 5495 }, { "epoch": 0.28, "learning_rate": 1.691612923679736e-05, "loss": 0.9516, "step": 5496 }, { "epoch": 0.28, "learning_rate": 1.6914940711329893e-05, "loss": 1.0307, "step": 5497 }, { "epoch": 0.28, "learning_rate": 1.6913751998647765e-05, "loss": 0.9283, "step": 5498 }, { "epoch": 0.28, "learning_rate": 1.6912563098783158e-05, "loss": 0.957, "step": 5499 }, { "epoch": 0.28, "learning_rate": 1.6911374011768257e-05, "loss": 0.9236, "step": 5500 }, { "epoch": 0.28, "learning_rate": 1.691018473763526e-05, "loss": 1.0772, "step": 5501 }, { "epoch": 0.28, "learning_rate": 1.690899527641636e-05, "loss": 1.0329, "step": 5502 }, { "epoch": 0.28, "learning_rate": 1.690780562814377e-05, "loss": 0.9672, "step": 5503 }, { "epoch": 0.28, "learning_rate": 1.6906615792849682e-05, "loss": 0.949, "step": 5504 }, { "epoch": 0.28, "learning_rate": 1.6905425770566327e-05, "loss": 0.9684, "step": 5505 }, { "epoch": 0.28, "learning_rate": 1.6904235561325914e-05, "loss": 1.0222, "step": 5506 }, { "epoch": 0.28, "learning_rate": 1.6903045165160668e-05, "loss": 1.0359, "step": 5507 }, { "epoch": 0.28, "learning_rate": 1.690185458210282e-05, "loss": 0.8892, "step": 5508 }, { "epoch": 0.28, "learning_rate": 1.6900663812184603e-05, "loss": 0.9118, "step": 5509 }, { "epoch": 0.28, "learning_rate": 1.689947285543825e-05, "loss": 0.9384, "step": 5510 }, { "epoch": 0.28, "learning_rate": 1.6898281711896015e-05, "loss": 0.9447, "step": 5511 }, { "epoch": 0.28, "learning_rate": 1.689709038159014e-05, "loss": 0.976, "step": 5512 }, { "epoch": 0.28, "learning_rate": 1.689589886455288e-05, "loss": 0.9847, "step": 5513 }, { "epoch": 0.28, "learning_rate": 1.6894707160816498e-05, "loss": 1.0807, "step": 5514 }, { "epoch": 0.28, "learning_rate": 1.6893515270413254e-05, "loss": 0.932, "step": 5515 }, { "epoch": 0.28, "learning_rate": 1.6892323193375417e-05, "loss": 0.9633, "step": 5516 }, { "epoch": 0.28, "learning_rate": 1.6891130929735263e-05, "loss": 0.954, "step": 5517 }, { "epoch": 0.28, "learning_rate": 1.688993847952507e-05, "loss": 0.8404, "step": 5518 }, { "epoch": 0.28, "learning_rate": 1.688874584277712e-05, "loss": 1.0775, "step": 5519 }, { "epoch": 0.28, "learning_rate": 1.6887553019523712e-05, "loss": 0.9602, "step": 5520 }, { "epoch": 0.28, "learning_rate": 1.688636000979713e-05, "loss": 1.088, "step": 5521 }, { "epoch": 0.28, "learning_rate": 1.6885166813629674e-05, "loss": 0.8404, "step": 5522 }, { "epoch": 0.28, "learning_rate": 1.6883973431053656e-05, "loss": 0.9568, "step": 5523 }, { "epoch": 0.28, "learning_rate": 1.6882779862101385e-05, "loss": 1.0153, "step": 5524 }, { "epoch": 0.28, "learning_rate": 1.6881586106805165e-05, "loss": 0.955, "step": 5525 }, { "epoch": 0.28, "learning_rate": 1.6880392165197324e-05, "loss": 0.9668, "step": 5526 }, { "epoch": 0.28, "learning_rate": 1.6879198037310186e-05, "loss": 0.9494, "step": 5527 }, { "epoch": 0.28, "learning_rate": 1.687800372317608e-05, "loss": 1.0001, "step": 5528 }, { "epoch": 0.28, "learning_rate": 1.6876809222827343e-05, "loss": 0.9605, "step": 5529 }, { "epoch": 0.28, "learning_rate": 1.6875614536296308e-05, "loss": 0.9956, "step": 5530 }, { "epoch": 0.28, "learning_rate": 1.687441966361533e-05, "loss": 0.9626, "step": 5531 }, { "epoch": 0.28, "learning_rate": 1.6873224604816753e-05, "loss": 0.954, "step": 5532 }, { "epoch": 0.28, "learning_rate": 1.687202935993293e-05, "loss": 0.9933, "step": 5533 }, { "epoch": 0.28, "learning_rate": 1.687083392899622e-05, "loss": 1.0387, "step": 5534 }, { "epoch": 0.28, "learning_rate": 1.6869638312038995e-05, "loss": 1.0187, "step": 5535 }, { "epoch": 0.28, "learning_rate": 1.6868442509093623e-05, "loss": 0.965, "step": 5536 }, { "epoch": 0.28, "learning_rate": 1.6867246520192477e-05, "loss": 1.0777, "step": 5537 }, { "epoch": 0.28, "learning_rate": 1.6866050345367938e-05, "loss": 0.9591, "step": 5538 }, { "epoch": 0.28, "learning_rate": 1.6864853984652386e-05, "loss": 0.9216, "step": 5539 }, { "epoch": 0.28, "learning_rate": 1.6863657438078222e-05, "loss": 1.0214, "step": 5540 }, { "epoch": 0.28, "learning_rate": 1.686246070567783e-05, "loss": 1.0114, "step": 5541 }, { "epoch": 0.28, "learning_rate": 1.686126378748362e-05, "loss": 0.9937, "step": 5542 }, { "epoch": 0.28, "learning_rate": 1.686006668352799e-05, "loss": 1.0025, "step": 5543 }, { "epoch": 0.28, "learning_rate": 1.6858869393843355e-05, "loss": 1.0339, "step": 5544 }, { "epoch": 0.28, "learning_rate": 1.6857671918462125e-05, "loss": 0.9792, "step": 5545 }, { "epoch": 0.28, "learning_rate": 1.6856474257416727e-05, "loss": 1.056, "step": 5546 }, { "epoch": 0.28, "learning_rate": 1.685527641073958e-05, "loss": 0.97, "step": 5547 }, { "epoch": 0.28, "learning_rate": 1.685407837846312e-05, "loss": 0.9891, "step": 5548 }, { "epoch": 0.28, "learning_rate": 1.685288016061978e-05, "loss": 0.9909, "step": 5549 }, { "epoch": 0.28, "learning_rate": 1.6851681757242e-05, "loss": 0.9751, "step": 5550 }, { "epoch": 0.28, "learning_rate": 1.685048316836223e-05, "loss": 0.9563, "step": 5551 }, { "epoch": 0.28, "learning_rate": 1.684928439401291e-05, "loss": 0.9738, "step": 5552 }, { "epoch": 0.28, "learning_rate": 1.6848085434226505e-05, "loss": 0.9988, "step": 5553 }, { "epoch": 0.28, "learning_rate": 1.6846886289035476e-05, "loss": 1.042, "step": 5554 }, { "epoch": 0.28, "learning_rate": 1.684568695847228e-05, "loss": 1.0042, "step": 5555 }, { "epoch": 0.28, "learning_rate": 1.6844487442569394e-05, "loss": 1.0083, "step": 5556 }, { "epoch": 0.28, "learning_rate": 1.6843287741359295e-05, "loss": 0.9709, "step": 5557 }, { "epoch": 0.28, "learning_rate": 1.6842087854874457e-05, "loss": 1.0242, "step": 5558 }, { "epoch": 0.28, "learning_rate": 1.6840887783147376e-05, "loss": 0.9104, "step": 5559 }, { "epoch": 0.28, "learning_rate": 1.6839687526210533e-05, "loss": 0.9506, "step": 5560 }, { "epoch": 0.28, "learning_rate": 1.6838487084096425e-05, "loss": 0.993, "step": 5561 }, { "epoch": 0.28, "learning_rate": 1.6837286456837558e-05, "loss": 1.0226, "step": 5562 }, { "epoch": 0.28, "learning_rate": 1.6836085644466433e-05, "loss": 1.0808, "step": 5563 }, { "epoch": 0.28, "learning_rate": 1.6834884647015563e-05, "loss": 0.9639, "step": 5564 }, { "epoch": 0.28, "learning_rate": 1.6833683464517457e-05, "loss": 0.9741, "step": 5565 }, { "epoch": 0.28, "learning_rate": 1.683248209700465e-05, "loss": 0.9261, "step": 5566 }, { "epoch": 0.28, "learning_rate": 1.6831280544509656e-05, "loss": 1.041, "step": 5567 }, { "epoch": 0.28, "learning_rate": 1.683007880706501e-05, "loss": 0.9265, "step": 5568 }, { "epoch": 0.28, "learning_rate": 1.6828876884703245e-05, "loss": 0.8978, "step": 5569 }, { "epoch": 0.28, "learning_rate": 1.6827674777456902e-05, "loss": 0.9271, "step": 5570 }, { "epoch": 0.28, "learning_rate": 1.6826472485358534e-05, "loss": 0.9348, "step": 5571 }, { "epoch": 0.28, "learning_rate": 1.6825270008440684e-05, "loss": 0.9966, "step": 5572 }, { "epoch": 0.28, "learning_rate": 1.6824067346735912e-05, "loss": 0.9515, "step": 5573 }, { "epoch": 0.28, "learning_rate": 1.6822864500276774e-05, "loss": 0.9523, "step": 5574 }, { "epoch": 0.28, "learning_rate": 1.6821661469095837e-05, "loss": 0.9486, "step": 5575 }, { "epoch": 0.28, "learning_rate": 1.6820458253225678e-05, "loss": 0.8894, "step": 5576 }, { "epoch": 0.28, "learning_rate": 1.6819254852698863e-05, "loss": 0.9691, "step": 5577 }, { "epoch": 0.28, "learning_rate": 1.681805126754798e-05, "loss": 1.0067, "step": 5578 }, { "epoch": 0.28, "learning_rate": 1.6816847497805617e-05, "loss": 0.9733, "step": 5579 }, { "epoch": 0.28, "learning_rate": 1.6815643543504355e-05, "loss": 0.996, "step": 5580 }, { "epoch": 0.28, "learning_rate": 1.6814439404676797e-05, "loss": 0.8874, "step": 5581 }, { "epoch": 0.28, "learning_rate": 1.6813235081355542e-05, "loss": 0.9074, "step": 5582 }, { "epoch": 0.28, "learning_rate": 1.6812030573573198e-05, "loss": 0.9078, "step": 5583 }, { "epoch": 0.28, "learning_rate": 1.6810825881362373e-05, "loss": 0.9486, "step": 5584 }, { "epoch": 0.28, "learning_rate": 1.680962100475568e-05, "loss": 0.9665, "step": 5585 }, { "epoch": 0.28, "learning_rate": 1.6808415943785744e-05, "loss": 0.9725, "step": 5586 }, { "epoch": 0.28, "learning_rate": 1.6807210698485193e-05, "loss": 1.0263, "step": 5587 }, { "epoch": 0.28, "learning_rate": 1.6806005268886654e-05, "loss": 0.9602, "step": 5588 }, { "epoch": 0.28, "learning_rate": 1.6804799655022765e-05, "loss": 1.0189, "step": 5589 }, { "epoch": 0.28, "learning_rate": 1.680359385692616e-05, "loss": 1.0544, "step": 5590 }, { "epoch": 0.28, "learning_rate": 1.6802387874629492e-05, "loss": 0.9367, "step": 5591 }, { "epoch": 0.28, "learning_rate": 1.680118170816541e-05, "loss": 0.9636, "step": 5592 }, { "epoch": 0.28, "learning_rate": 1.679997535756657e-05, "loss": 0.9896, "step": 5593 }, { "epoch": 0.28, "learning_rate": 1.6798768822865632e-05, "loss": 1.0107, "step": 5594 }, { "epoch": 0.28, "learning_rate": 1.6797562104095256e-05, "loss": 0.9568, "step": 5595 }, { "epoch": 0.28, "learning_rate": 1.6796355201288124e-05, "loss": 0.9906, "step": 5596 }, { "epoch": 0.28, "learning_rate": 1.6795148114476905e-05, "loss": 0.9767, "step": 5597 }, { "epoch": 0.28, "learning_rate": 1.679394084369428e-05, "loss": 1.0081, "step": 5598 }, { "epoch": 0.28, "learning_rate": 1.679273338897293e-05, "loss": 0.8947, "step": 5599 }, { "epoch": 0.28, "learning_rate": 1.6791525750345558e-05, "loss": 1.0307, "step": 5600 }, { "epoch": 0.28, "learning_rate": 1.6790317927844848e-05, "loss": 0.8968, "step": 5601 }, { "epoch": 0.28, "learning_rate": 1.6789109921503504e-05, "loss": 1.0211, "step": 5602 }, { "epoch": 0.28, "learning_rate": 1.6787901731354232e-05, "loss": 0.9626, "step": 5603 }, { "epoch": 0.28, "learning_rate": 1.6786693357429744e-05, "loss": 1.0763, "step": 5604 }, { "epoch": 0.28, "learning_rate": 1.6785484799762752e-05, "loss": 0.9511, "step": 5605 }, { "epoch": 0.28, "learning_rate": 1.6784276058385983e-05, "loss": 0.9653, "step": 5606 }, { "epoch": 0.28, "learning_rate": 1.6783067133332153e-05, "loss": 0.9161, "step": 5607 }, { "epoch": 0.28, "learning_rate": 1.6781858024633996e-05, "loss": 0.9759, "step": 5608 }, { "epoch": 0.28, "learning_rate": 1.678064873232425e-05, "loss": 1.0376, "step": 5609 }, { "epoch": 0.29, "learning_rate": 1.6779439256435654e-05, "loss": 0.9677, "step": 5610 }, { "epoch": 0.29, "learning_rate": 1.677822959700095e-05, "loss": 1.077, "step": 5611 }, { "epoch": 0.29, "learning_rate": 1.6777019754052895e-05, "loss": 0.9364, "step": 5612 }, { "epoch": 0.29, "learning_rate": 1.677580972762424e-05, "loss": 0.9188, "step": 5613 }, { "epoch": 0.29, "learning_rate": 1.6774599517747745e-05, "loss": 0.8796, "step": 5614 }, { "epoch": 0.29, "learning_rate": 1.6773389124456175e-05, "loss": 1.0343, "step": 5615 }, { "epoch": 0.29, "learning_rate": 1.6772178547782304e-05, "loss": 0.8789, "step": 5616 }, { "epoch": 0.29, "learning_rate": 1.67709677877589e-05, "loss": 0.9704, "step": 5617 }, { "epoch": 0.29, "learning_rate": 1.6769756844418746e-05, "loss": 1.0869, "step": 5618 }, { "epoch": 0.29, "learning_rate": 1.6768545717794632e-05, "loss": 0.917, "step": 5619 }, { "epoch": 0.29, "learning_rate": 1.676733440791934e-05, "loss": 0.9189, "step": 5620 }, { "epoch": 0.29, "learning_rate": 1.676612291482567e-05, "loss": 1.136, "step": 5621 }, { "epoch": 0.29, "learning_rate": 1.6764911238546422e-05, "loss": 1.0323, "step": 5622 }, { "epoch": 0.29, "learning_rate": 1.6763699379114394e-05, "loss": 0.9777, "step": 5623 }, { "epoch": 0.29, "learning_rate": 1.6762487336562408e-05, "loss": 1.0317, "step": 5624 }, { "epoch": 0.29, "learning_rate": 1.676127511092327e-05, "loss": 1.0368, "step": 5625 }, { "epoch": 0.29, "learning_rate": 1.6760062702229804e-05, "loss": 0.9984, "step": 5626 }, { "epoch": 0.29, "learning_rate": 1.6758850110514824e-05, "loss": 1.0244, "step": 5627 }, { "epoch": 0.29, "learning_rate": 1.675763733581118e-05, "loss": 0.9908, "step": 5628 }, { "epoch": 0.29, "learning_rate": 1.6756424378151686e-05, "loss": 0.9346, "step": 5629 }, { "epoch": 0.29, "learning_rate": 1.6755211237569193e-05, "loss": 0.9713, "step": 5630 }, { "epoch": 0.29, "learning_rate": 1.6753997914096545e-05, "loss": 0.9554, "step": 5631 }, { "epoch": 0.29, "learning_rate": 1.6752784407766583e-05, "loss": 0.9839, "step": 5632 }, { "epoch": 0.29, "learning_rate": 1.6751570718612177e-05, "loss": 1.0288, "step": 5633 }, { "epoch": 0.29, "learning_rate": 1.6750356846666173e-05, "loss": 0.8682, "step": 5634 }, { "epoch": 0.29, "learning_rate": 1.6749142791961433e-05, "loss": 0.9536, "step": 5635 }, { "epoch": 0.29, "learning_rate": 1.6747928554530837e-05, "loss": 0.9171, "step": 5636 }, { "epoch": 0.29, "learning_rate": 1.6746714134407258e-05, "loss": 0.9554, "step": 5637 }, { "epoch": 0.29, "learning_rate": 1.6745499531623565e-05, "loss": 1.0808, "step": 5638 }, { "epoch": 0.29, "learning_rate": 1.6744284746212653e-05, "loss": 0.9805, "step": 5639 }, { "epoch": 0.29, "learning_rate": 1.6743069778207405e-05, "loss": 0.9631, "step": 5640 }, { "epoch": 0.29, "learning_rate": 1.674185462764072e-05, "loss": 0.9487, "step": 5641 }, { "epoch": 0.29, "learning_rate": 1.6740639294545492e-05, "loss": 0.9565, "step": 5642 }, { "epoch": 0.29, "learning_rate": 1.6739423778954623e-05, "loss": 1.0098, "step": 5643 }, { "epoch": 0.29, "learning_rate": 1.6738208080901032e-05, "loss": 1.1138, "step": 5644 }, { "epoch": 0.29, "learning_rate": 1.673699220041762e-05, "loss": 0.9758, "step": 5645 }, { "epoch": 0.29, "learning_rate": 1.6735776137537315e-05, "loss": 1.0107, "step": 5646 }, { "epoch": 0.29, "learning_rate": 1.673455989229304e-05, "loss": 0.9439, "step": 5647 }, { "epoch": 0.29, "learning_rate": 1.6733343464717714e-05, "loss": 0.9032, "step": 5648 }, { "epoch": 0.29, "learning_rate": 1.6732126854844285e-05, "loss": 1.04, "step": 5649 }, { "epoch": 0.29, "learning_rate": 1.673091006270568e-05, "loss": 0.9546, "step": 5650 }, { "epoch": 0.29, "learning_rate": 1.6729693088334844e-05, "loss": 0.9243, "step": 5651 }, { "epoch": 0.29, "learning_rate": 1.672847593176473e-05, "loss": 0.893, "step": 5652 }, { "epoch": 0.29, "learning_rate": 1.672725859302829e-05, "loss": 0.9408, "step": 5653 }, { "epoch": 0.29, "learning_rate": 1.672604107215848e-05, "loss": 0.9639, "step": 5654 }, { "epoch": 0.29, "learning_rate": 1.672482336918826e-05, "loss": 0.9437, "step": 5655 }, { "epoch": 0.29, "learning_rate": 1.6723605484150606e-05, "loss": 0.8781, "step": 5656 }, { "epoch": 0.29, "learning_rate": 1.6722387417078487e-05, "loss": 1.1219, "step": 5657 }, { "epoch": 0.29, "learning_rate": 1.672116916800488e-05, "loss": 0.9571, "step": 5658 }, { "epoch": 0.29, "learning_rate": 1.6719950736962766e-05, "loss": 0.933, "step": 5659 }, { "epoch": 0.29, "learning_rate": 1.671873212398514e-05, "loss": 1.0166, "step": 5660 }, { "epoch": 0.29, "learning_rate": 1.6717513329104985e-05, "loss": 0.9146, "step": 5661 }, { "epoch": 0.29, "learning_rate": 1.6716294352355306e-05, "loss": 0.923, "step": 5662 }, { "epoch": 0.29, "learning_rate": 1.6715075193769107e-05, "loss": 1.0063, "step": 5663 }, { "epoch": 0.29, "learning_rate": 1.6713855853379385e-05, "loss": 1.0157, "step": 5664 }, { "epoch": 0.29, "learning_rate": 1.671263633121916e-05, "loss": 0.9727, "step": 5665 }, { "epoch": 0.29, "learning_rate": 1.6711416627321448e-05, "loss": 1.0319, "step": 5666 }, { "epoch": 0.29, "learning_rate": 1.6710196741719275e-05, "loss": 1.054, "step": 5667 }, { "epoch": 0.29, "learning_rate": 1.670897667444566e-05, "loss": 1.0203, "step": 5668 }, { "epoch": 0.29, "learning_rate": 1.670775642553364e-05, "loss": 0.9425, "step": 5669 }, { "epoch": 0.29, "learning_rate": 1.6706535995016253e-05, "loss": 0.9321, "step": 5670 }, { "epoch": 0.29, "learning_rate": 1.6705315382926538e-05, "loss": 0.9986, "step": 5671 }, { "epoch": 0.29, "learning_rate": 1.6704094589297544e-05, "loss": 1.1028, "step": 5672 }, { "epoch": 0.29, "learning_rate": 1.6702873614162318e-05, "loss": 1.0114, "step": 5673 }, { "epoch": 0.29, "learning_rate": 1.6701652457553925e-05, "loss": 0.9865, "step": 5674 }, { "epoch": 0.29, "learning_rate": 1.670043111950542e-05, "loss": 0.9583, "step": 5675 }, { "epoch": 0.29, "learning_rate": 1.6699209600049872e-05, "loss": 0.9121, "step": 5676 }, { "epoch": 0.29, "learning_rate": 1.6697987899220347e-05, "loss": 0.8978, "step": 5677 }, { "epoch": 0.29, "learning_rate": 1.669676601704993e-05, "loss": 0.9642, "step": 5678 }, { "epoch": 0.29, "learning_rate": 1.6695543953571698e-05, "loss": 0.9203, "step": 5679 }, { "epoch": 0.29, "learning_rate": 1.6694321708818734e-05, "loss": 0.9666, "step": 5680 }, { "epoch": 0.29, "learning_rate": 1.6693099282824133e-05, "loss": 0.9856, "step": 5681 }, { "epoch": 0.29, "learning_rate": 1.6691876675620993e-05, "loss": 1.0214, "step": 5682 }, { "epoch": 0.29, "learning_rate": 1.669065388724241e-05, "loss": 0.8836, "step": 5683 }, { "epoch": 0.29, "learning_rate": 1.668943091772149e-05, "loss": 1.0273, "step": 5684 }, { "epoch": 0.29, "learning_rate": 1.6688207767091343e-05, "loss": 0.9369, "step": 5685 }, { "epoch": 0.29, "learning_rate": 1.6686984435385093e-05, "loss": 1.0422, "step": 5686 }, { "epoch": 0.29, "learning_rate": 1.6685760922635848e-05, "loss": 0.9878, "step": 5687 }, { "epoch": 0.29, "learning_rate": 1.668453722887674e-05, "loss": 0.9624, "step": 5688 }, { "epoch": 0.29, "learning_rate": 1.6683313354140896e-05, "loss": 1.0649, "step": 5689 }, { "epoch": 0.29, "learning_rate": 1.668208929846146e-05, "loss": 1.0053, "step": 5690 }, { "epoch": 0.29, "learning_rate": 1.668086506187156e-05, "loss": 0.9431, "step": 5691 }, { "epoch": 0.29, "learning_rate": 1.6679640644404348e-05, "loss": 0.8794, "step": 5692 }, { "epoch": 0.29, "learning_rate": 1.6678416046092975e-05, "loss": 0.9619, "step": 5693 }, { "epoch": 0.29, "learning_rate": 1.667719126697059e-05, "loss": 1.0046, "step": 5694 }, { "epoch": 0.29, "learning_rate": 1.6675966307070356e-05, "loss": 1.0305, "step": 5695 }, { "epoch": 0.29, "learning_rate": 1.6674741166425437e-05, "loss": 0.9114, "step": 5696 }, { "epoch": 0.29, "learning_rate": 1.6673515845069e-05, "loss": 0.9336, "step": 5697 }, { "epoch": 0.29, "learning_rate": 1.6672290343034224e-05, "loss": 1.0306, "step": 5698 }, { "epoch": 0.29, "learning_rate": 1.6671064660354287e-05, "loss": 0.9702, "step": 5699 }, { "epoch": 0.29, "learning_rate": 1.6669838797062365e-05, "loss": 0.9301, "step": 5700 }, { "epoch": 0.29, "learning_rate": 1.6668612753191662e-05, "loss": 1.0191, "step": 5701 }, { "epoch": 0.29, "learning_rate": 1.666738652877536e-05, "loss": 0.9957, "step": 5702 }, { "epoch": 0.29, "learning_rate": 1.666616012384666e-05, "loss": 1.016, "step": 5703 }, { "epoch": 0.29, "learning_rate": 1.6664933538438767e-05, "loss": 0.9619, "step": 5704 }, { "epoch": 0.29, "learning_rate": 1.6663706772584895e-05, "loss": 0.9606, "step": 5705 }, { "epoch": 0.29, "learning_rate": 1.6662479826318243e-05, "loss": 0.8875, "step": 5706 }, { "epoch": 0.29, "learning_rate": 1.6661252699672045e-05, "loss": 0.9712, "step": 5707 }, { "epoch": 0.29, "learning_rate": 1.6660025392679516e-05, "loss": 1.0217, "step": 5708 }, { "epoch": 0.29, "learning_rate": 1.665879790537388e-05, "loss": 0.8995, "step": 5709 }, { "epoch": 0.29, "learning_rate": 1.665757023778838e-05, "loss": 0.9429, "step": 5710 }, { "epoch": 0.29, "learning_rate": 1.665634238995625e-05, "loss": 0.9144, "step": 5711 }, { "epoch": 0.29, "learning_rate": 1.6655114361910728e-05, "loss": 0.9467, "step": 5712 }, { "epoch": 0.29, "learning_rate": 1.6653886153685067e-05, "loss": 1.0856, "step": 5713 }, { "epoch": 0.29, "learning_rate": 1.6652657765312516e-05, "loss": 0.9165, "step": 5714 }, { "epoch": 0.29, "learning_rate": 1.6651429196826337e-05, "loss": 0.9489, "step": 5715 }, { "epoch": 0.29, "learning_rate": 1.6650200448259782e-05, "loss": 1.0636, "step": 5716 }, { "epoch": 0.29, "learning_rate": 1.6648971519646134e-05, "loss": 1.0327, "step": 5717 }, { "epoch": 0.29, "learning_rate": 1.6647742411018653e-05, "loss": 0.9096, "step": 5718 }, { "epoch": 0.29, "learning_rate": 1.6646513122410616e-05, "loss": 1.044, "step": 5719 }, { "epoch": 0.29, "learning_rate": 1.6645283653855308e-05, "loss": 1.0103, "step": 5720 }, { "epoch": 0.29, "learning_rate": 1.6644054005386018e-05, "loss": 0.9105, "step": 5721 }, { "epoch": 0.29, "learning_rate": 1.6642824177036034e-05, "loss": 1.0213, "step": 5722 }, { "epoch": 0.29, "learning_rate": 1.664159416883865e-05, "loss": 0.991, "step": 5723 }, { "epoch": 0.29, "learning_rate": 1.6640363980827174e-05, "loss": 0.9929, "step": 5724 }, { "epoch": 0.29, "learning_rate": 1.6639133613034907e-05, "loss": 0.9056, "step": 5725 }, { "epoch": 0.29, "learning_rate": 1.663790306549516e-05, "loss": 0.9929, "step": 5726 }, { "epoch": 0.29, "learning_rate": 1.6636672338241248e-05, "loss": 1.0026, "step": 5727 }, { "epoch": 0.29, "learning_rate": 1.6635441431306496e-05, "loss": 0.9501, "step": 5728 }, { "epoch": 0.29, "learning_rate": 1.6634210344724224e-05, "loss": 1.0248, "step": 5729 }, { "epoch": 0.29, "learning_rate": 1.6632979078527768e-05, "loss": 0.9241, "step": 5730 }, { "epoch": 0.29, "learning_rate": 1.6631747632750458e-05, "loss": 0.9475, "step": 5731 }, { "epoch": 0.29, "learning_rate": 1.663051600742564e-05, "loss": 1.0079, "step": 5732 }, { "epoch": 0.29, "learning_rate": 1.6629284202586652e-05, "loss": 0.9601, "step": 5733 }, { "epoch": 0.29, "learning_rate": 1.6628052218266848e-05, "loss": 0.9831, "step": 5734 }, { "epoch": 0.29, "learning_rate": 1.6626820054499583e-05, "loss": 1.1562, "step": 5735 }, { "epoch": 0.29, "learning_rate": 1.6625587711318215e-05, "loss": 0.9987, "step": 5736 }, { "epoch": 0.29, "learning_rate": 1.6624355188756108e-05, "loss": 1.0087, "step": 5737 }, { "epoch": 0.29, "learning_rate": 1.6623122486846634e-05, "loss": 0.8442, "step": 5738 }, { "epoch": 0.29, "learning_rate": 1.662188960562316e-05, "loss": 1.0264, "step": 5739 }, { "epoch": 0.29, "learning_rate": 1.6620656545119076e-05, "loss": 0.9587, "step": 5740 }, { "epoch": 0.29, "learning_rate": 1.6619423305367756e-05, "loss": 0.9484, "step": 5741 }, { "epoch": 0.29, "learning_rate": 1.6618189886402594e-05, "loss": 0.9896, "step": 5742 }, { "epoch": 0.29, "learning_rate": 1.6616956288256984e-05, "loss": 0.9594, "step": 5743 }, { "epoch": 0.29, "learning_rate": 1.661572251096432e-05, "loss": 0.9217, "step": 5744 }, { "epoch": 0.29, "learning_rate": 1.661448855455801e-05, "loss": 0.9692, "step": 5745 }, { "epoch": 0.29, "learning_rate": 1.6613254419071455e-05, "loss": 1.0289, "step": 5746 }, { "epoch": 0.29, "learning_rate": 1.6612020104538076e-05, "loss": 0.9264, "step": 5747 }, { "epoch": 0.29, "learning_rate": 1.6610785610991285e-05, "loss": 0.9879, "step": 5748 }, { "epoch": 0.29, "learning_rate": 1.660955093846451e-05, "loss": 0.9382, "step": 5749 }, { "epoch": 0.29, "learning_rate": 1.6608316086991173e-05, "loss": 0.9258, "step": 5750 }, { "epoch": 0.29, "learning_rate": 1.660708105660471e-05, "loss": 1.0289, "step": 5751 }, { "epoch": 0.29, "learning_rate": 1.6605845847338556e-05, "loss": 1.0598, "step": 5752 }, { "epoch": 0.29, "learning_rate": 1.6604610459226154e-05, "loss": 0.9012, "step": 5753 }, { "epoch": 0.29, "learning_rate": 1.660337489230095e-05, "loss": 1.0025, "step": 5754 }, { "epoch": 0.29, "learning_rate": 1.66021391465964e-05, "loss": 0.9902, "step": 5755 }, { "epoch": 0.29, "learning_rate": 1.660090322214595e-05, "loss": 1.0496, "step": 5756 }, { "epoch": 0.29, "learning_rate": 1.6599667118983073e-05, "loss": 0.9986, "step": 5757 }, { "epoch": 0.29, "learning_rate": 1.6598430837141228e-05, "loss": 1.0465, "step": 5758 }, { "epoch": 0.29, "learning_rate": 1.659719437665389e-05, "loss": 0.9825, "step": 5759 }, { "epoch": 0.29, "learning_rate": 1.6595957737554532e-05, "loss": 0.8544, "step": 5760 }, { "epoch": 0.29, "learning_rate": 1.6594720919876636e-05, "loss": 0.8518, "step": 5761 }, { "epoch": 0.29, "learning_rate": 1.6593483923653687e-05, "loss": 1.0092, "step": 5762 }, { "epoch": 0.29, "learning_rate": 1.6592246748919175e-05, "loss": 1.019, "step": 5763 }, { "epoch": 0.29, "learning_rate": 1.6591009395706595e-05, "loss": 0.9855, "step": 5764 }, { "epoch": 0.29, "learning_rate": 1.6589771864049452e-05, "loss": 1.0144, "step": 5765 }, { "epoch": 0.29, "learning_rate": 1.6588534153981244e-05, "loss": 0.9788, "step": 5766 }, { "epoch": 0.29, "learning_rate": 1.6587296265535483e-05, "loss": 0.9491, "step": 5767 }, { "epoch": 0.29, "learning_rate": 1.6586058198745685e-05, "loss": 0.9887, "step": 5768 }, { "epoch": 0.29, "learning_rate": 1.658481995364537e-05, "loss": 0.9525, "step": 5769 }, { "epoch": 0.29, "learning_rate": 1.658358153026806e-05, "loss": 0.9694, "step": 5770 }, { "epoch": 0.29, "learning_rate": 1.658234292864728e-05, "loss": 0.9278, "step": 5771 }, { "epoch": 0.29, "learning_rate": 1.6581104148816574e-05, "loss": 1.0032, "step": 5772 }, { "epoch": 0.29, "learning_rate": 1.657986519080947e-05, "loss": 0.9879, "step": 5773 }, { "epoch": 0.29, "learning_rate": 1.657862605465952e-05, "loss": 0.8922, "step": 5774 }, { "epoch": 0.29, "learning_rate": 1.6577386740400268e-05, "loss": 1.0844, "step": 5775 }, { "epoch": 0.29, "learning_rate": 1.6576147248065268e-05, "loss": 1.0002, "step": 5776 }, { "epoch": 0.29, "learning_rate": 1.6574907577688076e-05, "loss": 0.9385, "step": 5777 }, { "epoch": 0.29, "learning_rate": 1.6573667729302256e-05, "loss": 1.0386, "step": 5778 }, { "epoch": 0.29, "learning_rate": 1.6572427702941375e-05, "loss": 0.9545, "step": 5779 }, { "epoch": 0.29, "learning_rate": 1.657118749863901e-05, "loss": 1.0504, "step": 5780 }, { "epoch": 0.29, "learning_rate": 1.656994711642873e-05, "loss": 0.9595, "step": 5781 }, { "epoch": 0.29, "learning_rate": 1.6568706556344127e-05, "loss": 1.0156, "step": 5782 }, { "epoch": 0.29, "learning_rate": 1.6567465818418783e-05, "loss": 0.9099, "step": 5783 }, { "epoch": 0.29, "learning_rate": 1.6566224902686285e-05, "loss": 0.9145, "step": 5784 }, { "epoch": 0.29, "learning_rate": 1.6564983809180236e-05, "loss": 0.9987, "step": 5785 }, { "epoch": 0.29, "learning_rate": 1.6563742537934237e-05, "loss": 0.9963, "step": 5786 }, { "epoch": 0.29, "learning_rate": 1.6562501088981887e-05, "loss": 0.9745, "step": 5787 }, { "epoch": 0.29, "learning_rate": 1.6561259462356805e-05, "loss": 0.9399, "step": 5788 }, { "epoch": 0.29, "learning_rate": 1.6560017658092603e-05, "loss": 1.039, "step": 5789 }, { "epoch": 0.29, "learning_rate": 1.6558775676222905e-05, "loss": 1.0104, "step": 5790 }, { "epoch": 0.29, "learning_rate": 1.655753351678133e-05, "loss": 0.925, "step": 5791 }, { "epoch": 0.29, "learning_rate": 1.6556291179801515e-05, "loss": 0.9117, "step": 5792 }, { "epoch": 0.29, "learning_rate": 1.655504866531709e-05, "loss": 0.909, "step": 5793 }, { "epoch": 0.29, "learning_rate": 1.6553805973361692e-05, "loss": 1.0215, "step": 5794 }, { "epoch": 0.29, "learning_rate": 1.6552563103968975e-05, "loss": 1.028, "step": 5795 }, { "epoch": 0.29, "learning_rate": 1.6551320057172583e-05, "loss": 0.9852, "step": 5796 }, { "epoch": 0.29, "learning_rate": 1.6550076833006167e-05, "loss": 0.9845, "step": 5797 }, { "epoch": 0.29, "learning_rate": 1.654883343150339e-05, "loss": 0.9081, "step": 5798 }, { "epoch": 0.29, "learning_rate": 1.654758985269792e-05, "loss": 1.0626, "step": 5799 }, { "epoch": 0.29, "learning_rate": 1.6546346096623417e-05, "loss": 0.9464, "step": 5800 }, { "epoch": 0.29, "learning_rate": 1.6545102163313564e-05, "loss": 1.0527, "step": 5801 }, { "epoch": 0.29, "learning_rate": 1.6543858052802026e-05, "loss": 0.9782, "step": 5802 }, { "epoch": 0.29, "learning_rate": 1.6542613765122495e-05, "loss": 0.9219, "step": 5803 }, { "epoch": 0.29, "learning_rate": 1.654136930030866e-05, "loss": 0.9604, "step": 5804 }, { "epoch": 0.29, "learning_rate": 1.6540124658394207e-05, "loss": 0.9374, "step": 5805 }, { "epoch": 0.29, "learning_rate": 1.6538879839412837e-05, "loss": 0.9525, "step": 5806 }, { "epoch": 0.3, "learning_rate": 1.6537634843398256e-05, "loss": 0.9329, "step": 5807 }, { "epoch": 0.3, "learning_rate": 1.6536389670384165e-05, "loss": 0.9, "step": 5808 }, { "epoch": 0.3, "learning_rate": 1.653514432040428e-05, "loss": 1.0086, "step": 5809 }, { "epoch": 0.3, "learning_rate": 1.6533898793492312e-05, "loss": 1.039, "step": 5810 }, { "epoch": 0.3, "learning_rate": 1.6532653089681988e-05, "loss": 1.0845, "step": 5811 }, { "epoch": 0.3, "learning_rate": 1.6531407209007028e-05, "loss": 0.9393, "step": 5812 }, { "epoch": 0.3, "learning_rate": 1.6530161151501173e-05, "loss": 1.0696, "step": 5813 }, { "epoch": 0.3, "learning_rate": 1.6528914917198146e-05, "loss": 0.9276, "step": 5814 }, { "epoch": 0.3, "learning_rate": 1.6527668506131697e-05, "loss": 0.888, "step": 5815 }, { "epoch": 0.3, "learning_rate": 1.652642191833557e-05, "loss": 0.9388, "step": 5816 }, { "epoch": 0.3, "learning_rate": 1.6525175153843508e-05, "loss": 0.8995, "step": 5817 }, { "epoch": 0.3, "learning_rate": 1.6523928212689274e-05, "loss": 1.0812, "step": 5818 }, { "epoch": 0.3, "learning_rate": 1.6522681094906628e-05, "loss": 0.9617, "step": 5819 }, { "epoch": 0.3, "learning_rate": 1.6521433800529325e-05, "loss": 0.9527, "step": 5820 }, { "epoch": 0.3, "learning_rate": 1.6520186329591145e-05, "loss": 0.9791, "step": 5821 }, { "epoch": 0.3, "learning_rate": 1.6518938682125854e-05, "loss": 0.935, "step": 5822 }, { "epoch": 0.3, "learning_rate": 1.6517690858167235e-05, "loss": 0.9957, "step": 5823 }, { "epoch": 0.3, "learning_rate": 1.6516442857749068e-05, "loss": 1.0533, "step": 5824 }, { "epoch": 0.3, "learning_rate": 1.6515194680905145e-05, "loss": 0.9276, "step": 5825 }, { "epoch": 0.3, "learning_rate": 1.6513946327669258e-05, "loss": 1.0209, "step": 5826 }, { "epoch": 0.3, "learning_rate": 1.6512697798075202e-05, "loss": 0.8633, "step": 5827 }, { "epoch": 0.3, "learning_rate": 1.6511449092156785e-05, "loss": 0.9946, "step": 5828 }, { "epoch": 0.3, "learning_rate": 1.6510200209947812e-05, "loss": 0.8476, "step": 5829 }, { "epoch": 0.3, "learning_rate": 1.6508951151482095e-05, "loss": 0.9691, "step": 5830 }, { "epoch": 0.3, "learning_rate": 1.650770191679345e-05, "loss": 0.9625, "step": 5831 }, { "epoch": 0.3, "learning_rate": 1.6506452505915697e-05, "loss": 0.7745, "step": 5832 }, { "epoch": 0.3, "learning_rate": 1.6505202918882667e-05, "loss": 0.8988, "step": 5833 }, { "epoch": 0.3, "learning_rate": 1.6503953155728188e-05, "loss": 1.0542, "step": 5834 }, { "epoch": 0.3, "learning_rate": 1.6502703216486095e-05, "loss": 0.9554, "step": 5835 }, { "epoch": 0.3, "learning_rate": 1.650145310119023e-05, "loss": 0.9577, "step": 5836 }, { "epoch": 0.3, "learning_rate": 1.6500202809874446e-05, "loss": 0.903, "step": 5837 }, { "epoch": 0.3, "learning_rate": 1.649895234257258e-05, "loss": 1.049, "step": 5838 }, { "epoch": 0.3, "learning_rate": 1.6497701699318498e-05, "loss": 0.952, "step": 5839 }, { "epoch": 0.3, "learning_rate": 1.6496450880146054e-05, "loss": 0.9761, "step": 5840 }, { "epoch": 0.3, "learning_rate": 1.6495199885089116e-05, "loss": 0.9647, "step": 5841 }, { "epoch": 0.3, "learning_rate": 1.6493948714181548e-05, "loss": 1.0675, "step": 5842 }, { "epoch": 0.3, "learning_rate": 1.6492697367457225e-05, "loss": 0.9746, "step": 5843 }, { "epoch": 0.3, "learning_rate": 1.6491445844950035e-05, "loss": 0.9496, "step": 5844 }, { "epoch": 0.3, "learning_rate": 1.6490194146693852e-05, "loss": 0.9326, "step": 5845 }, { "epoch": 0.3, "learning_rate": 1.6488942272722568e-05, "loss": 0.935, "step": 5846 }, { "epoch": 0.3, "learning_rate": 1.6487690223070078e-05, "loss": 0.9079, "step": 5847 }, { "epoch": 0.3, "learning_rate": 1.6486437997770273e-05, "loss": 0.9425, "step": 5848 }, { "epoch": 0.3, "learning_rate": 1.6485185596857062e-05, "loss": 1.0237, "step": 5849 }, { "epoch": 0.3, "learning_rate": 1.6483933020364353e-05, "loss": 1.0016, "step": 5850 }, { "epoch": 0.3, "learning_rate": 1.648268026832605e-05, "loss": 0.9556, "step": 5851 }, { "epoch": 0.3, "learning_rate": 1.648142734077608e-05, "loss": 0.9705, "step": 5852 }, { "epoch": 0.3, "learning_rate": 1.648017423774836e-05, "loss": 0.9006, "step": 5853 }, { "epoch": 0.3, "learning_rate": 1.647892095927682e-05, "loss": 0.886, "step": 5854 }, { "epoch": 0.3, "learning_rate": 1.6477667505395384e-05, "loss": 0.8978, "step": 5855 }, { "epoch": 0.3, "learning_rate": 1.647641387613799e-05, "loss": 1.0041, "step": 5856 }, { "epoch": 0.3, "learning_rate": 1.647516007153859e-05, "loss": 1.0532, "step": 5857 }, { "epoch": 0.3, "learning_rate": 1.647390609163111e-05, "loss": 1.0232, "step": 5858 }, { "epoch": 0.3, "learning_rate": 1.6472651936449514e-05, "loss": 0.9173, "step": 5859 }, { "epoch": 0.3, "learning_rate": 1.6471397606027755e-05, "loss": 0.9785, "step": 5860 }, { "epoch": 0.3, "learning_rate": 1.6470143100399785e-05, "loss": 0.9252, "step": 5861 }, { "epoch": 0.3, "learning_rate": 1.646888841959958e-05, "loss": 1.0088, "step": 5862 }, { "epoch": 0.3, "learning_rate": 1.6467633563661105e-05, "loss": 1.0432, "step": 5863 }, { "epoch": 0.3, "learning_rate": 1.6466378532618327e-05, "loss": 0.9878, "step": 5864 }, { "epoch": 0.3, "learning_rate": 1.6465123326505233e-05, "loss": 0.9649, "step": 5865 }, { "epoch": 0.3, "learning_rate": 1.6463867945355803e-05, "loss": 1.0262, "step": 5866 }, { "epoch": 0.3, "learning_rate": 1.6462612389204023e-05, "loss": 0.9272, "step": 5867 }, { "epoch": 0.3, "learning_rate": 1.646135665808389e-05, "loss": 0.9282, "step": 5868 }, { "epoch": 0.3, "learning_rate": 1.64601007520294e-05, "loss": 0.9987, "step": 5869 }, { "epoch": 0.3, "learning_rate": 1.6458844671074554e-05, "loss": 0.9633, "step": 5870 }, { "epoch": 0.3, "learning_rate": 1.645758841525336e-05, "loss": 0.8894, "step": 5871 }, { "epoch": 0.3, "learning_rate": 1.6456331984599834e-05, "loss": 0.9823, "step": 5872 }, { "epoch": 0.3, "learning_rate": 1.6455075379147986e-05, "loss": 0.9919, "step": 5873 }, { "epoch": 0.3, "learning_rate": 1.6453818598931838e-05, "loss": 1.0433, "step": 5874 }, { "epoch": 0.3, "learning_rate": 1.645256164398542e-05, "loss": 0.8779, "step": 5875 }, { "epoch": 0.3, "learning_rate": 1.645130451434276e-05, "loss": 0.9356, "step": 5876 }, { "epoch": 0.3, "learning_rate": 1.6450047210037888e-05, "loss": 0.9448, "step": 5877 }, { "epoch": 0.3, "learning_rate": 1.644878973110486e-05, "loss": 1.0756, "step": 5878 }, { "epoch": 0.3, "learning_rate": 1.6447532077577707e-05, "loss": 0.915, "step": 5879 }, { "epoch": 0.3, "learning_rate": 1.644627424949048e-05, "loss": 0.9949, "step": 5880 }, { "epoch": 0.3, "learning_rate": 1.644501624687724e-05, "loss": 0.9842, "step": 5881 }, { "epoch": 0.3, "learning_rate": 1.644375806977204e-05, "loss": 0.9922, "step": 5882 }, { "epoch": 0.3, "learning_rate": 1.6442499718208944e-05, "loss": 0.9904, "step": 5883 }, { "epoch": 0.3, "learning_rate": 1.6441241192222025e-05, "loss": 0.9979, "step": 5884 }, { "epoch": 0.3, "learning_rate": 1.6439982491845354e-05, "loss": 0.919, "step": 5885 }, { "epoch": 0.3, "learning_rate": 1.6438723617113005e-05, "loss": 0.9544, "step": 5886 }, { "epoch": 0.3, "learning_rate": 1.643746456805907e-05, "loss": 0.9592, "step": 5887 }, { "epoch": 0.3, "learning_rate": 1.643620534471763e-05, "loss": 0.9771, "step": 5888 }, { "epoch": 0.3, "learning_rate": 1.6434945947122772e-05, "loss": 0.9902, "step": 5889 }, { "epoch": 0.3, "learning_rate": 1.6433686375308603e-05, "loss": 0.8529, "step": 5890 }, { "epoch": 0.3, "learning_rate": 1.6432426629309223e-05, "loss": 0.862, "step": 5891 }, { "epoch": 0.3, "learning_rate": 1.6431166709158728e-05, "loss": 0.9482, "step": 5892 }, { "epoch": 0.3, "learning_rate": 1.6429906614891242e-05, "loss": 1.0277, "step": 5893 }, { "epoch": 0.3, "learning_rate": 1.6428646346540874e-05, "loss": 0.9652, "step": 5894 }, { "epoch": 0.3, "learning_rate": 1.6427385904141745e-05, "loss": 0.8918, "step": 5895 }, { "epoch": 0.3, "learning_rate": 1.6426125287727982e-05, "loss": 1.0294, "step": 5896 }, { "epoch": 0.3, "learning_rate": 1.6424864497333715e-05, "loss": 1.041, "step": 5897 }, { "epoch": 0.3, "learning_rate": 1.6423603532993074e-05, "loss": 1.0557, "step": 5898 }, { "epoch": 0.3, "learning_rate": 1.6422342394740205e-05, "loss": 0.9162, "step": 5899 }, { "epoch": 0.3, "learning_rate": 1.6421081082609245e-05, "loss": 0.9435, "step": 5900 }, { "epoch": 0.3, "learning_rate": 1.641981959663435e-05, "loss": 0.9555, "step": 5901 }, { "epoch": 0.3, "learning_rate": 1.641855793684967e-05, "loss": 0.9593, "step": 5902 }, { "epoch": 0.3, "learning_rate": 1.6417296103289358e-05, "loss": 0.9911, "step": 5903 }, { "epoch": 0.3, "learning_rate": 1.6416034095987583e-05, "loss": 1.0491, "step": 5904 }, { "epoch": 0.3, "learning_rate": 1.641477191497851e-05, "loss": 1.0078, "step": 5905 }, { "epoch": 0.3, "learning_rate": 1.6413509560296315e-05, "loss": 1.0229, "step": 5906 }, { "epoch": 0.3, "learning_rate": 1.641224703197517e-05, "loss": 0.9366, "step": 5907 }, { "epoch": 0.3, "learning_rate": 1.6410984330049263e-05, "loss": 0.9527, "step": 5908 }, { "epoch": 0.3, "learning_rate": 1.640972145455277e-05, "loss": 1.0096, "step": 5909 }, { "epoch": 0.3, "learning_rate": 1.6408458405519892e-05, "loss": 0.8852, "step": 5910 }, { "epoch": 0.3, "learning_rate": 1.640719518298482e-05, "loss": 1.0544, "step": 5911 }, { "epoch": 0.3, "learning_rate": 1.6405931786981753e-05, "loss": 0.9562, "step": 5912 }, { "epoch": 0.3, "learning_rate": 1.64046682175449e-05, "loss": 0.962, "step": 5913 }, { "epoch": 0.3, "learning_rate": 1.640340447470847e-05, "loss": 0.9944, "step": 5914 }, { "epoch": 0.3, "learning_rate": 1.6402140558506677e-05, "loss": 0.9661, "step": 5915 }, { "epoch": 0.3, "learning_rate": 1.6400876468973735e-05, "loss": 0.9049, "step": 5916 }, { "epoch": 0.3, "learning_rate": 1.6399612206143876e-05, "loss": 0.9569, "step": 5917 }, { "epoch": 0.3, "learning_rate": 1.6398347770051324e-05, "loss": 1.0437, "step": 5918 }, { "epoch": 0.3, "learning_rate": 1.6397083160730315e-05, "loss": 0.947, "step": 5919 }, { "epoch": 0.3, "learning_rate": 1.6395818378215084e-05, "loss": 0.9562, "step": 5920 }, { "epoch": 0.3, "learning_rate": 1.6394553422539874e-05, "loss": 0.9745, "step": 5921 }, { "epoch": 0.3, "learning_rate": 1.6393288293738935e-05, "loss": 0.9703, "step": 5922 }, { "epoch": 0.3, "learning_rate": 1.639202299184652e-05, "loss": 0.9196, "step": 5923 }, { "epoch": 0.3, "learning_rate": 1.639075751689688e-05, "loss": 0.9552, "step": 5924 }, { "epoch": 0.3, "learning_rate": 1.638949186892428e-05, "loss": 0.9575, "step": 5925 }, { "epoch": 0.3, "learning_rate": 1.6388226047962984e-05, "loss": 0.8807, "step": 5926 }, { "epoch": 0.3, "learning_rate": 1.6386960054047266e-05, "loss": 1.0409, "step": 5927 }, { "epoch": 0.3, "learning_rate": 1.63856938872114e-05, "loss": 0.9786, "step": 5928 }, { "epoch": 0.3, "learning_rate": 1.6384427547489665e-05, "loss": 0.9471, "step": 5929 }, { "epoch": 0.3, "learning_rate": 1.6383161034916347e-05, "loss": 0.9548, "step": 5930 }, { "epoch": 0.3, "learning_rate": 1.6381894349525737e-05, "loss": 1.0266, "step": 5931 }, { "epoch": 0.3, "learning_rate": 1.6380627491352125e-05, "loss": 1.0258, "step": 5932 }, { "epoch": 0.3, "learning_rate": 1.637936046042981e-05, "loss": 0.9247, "step": 5933 }, { "epoch": 0.3, "learning_rate": 1.63780932567931e-05, "loss": 0.9607, "step": 5934 }, { "epoch": 0.3, "learning_rate": 1.6376825880476302e-05, "loss": 0.9789, "step": 5935 }, { "epoch": 0.3, "learning_rate": 1.6375558331513728e-05, "loss": 1.0108, "step": 5936 }, { "epoch": 0.3, "learning_rate": 1.637429060993969e-05, "loss": 0.8867, "step": 5937 }, { "epoch": 0.3, "learning_rate": 1.6373022715788523e-05, "loss": 0.9734, "step": 5938 }, { "epoch": 0.3, "learning_rate": 1.6371754649094543e-05, "loss": 0.9999, "step": 5939 }, { "epoch": 0.3, "learning_rate": 1.6370486409892084e-05, "loss": 0.9803, "step": 5940 }, { "epoch": 0.3, "learning_rate": 1.6369217998215483e-05, "loss": 1.038, "step": 5941 }, { "epoch": 0.3, "learning_rate": 1.636794941409908e-05, "loss": 0.8222, "step": 5942 }, { "epoch": 0.3, "learning_rate": 1.6366680657577226e-05, "loss": 1.0703, "step": 5943 }, { "epoch": 0.3, "learning_rate": 1.636541172868426e-05, "loss": 0.9383, "step": 5944 }, { "epoch": 0.3, "learning_rate": 1.636414262745455e-05, "loss": 0.9319, "step": 5945 }, { "epoch": 0.3, "learning_rate": 1.636287335392245e-05, "loss": 0.9283, "step": 5946 }, { "epoch": 0.3, "learning_rate": 1.636160390812232e-05, "loss": 0.9766, "step": 5947 }, { "epoch": 0.3, "learning_rate": 1.636033429008853e-05, "loss": 0.9374, "step": 5948 }, { "epoch": 0.3, "learning_rate": 1.6359064499855462e-05, "loss": 0.9085, "step": 5949 }, { "epoch": 0.3, "learning_rate": 1.6357794537457487e-05, "loss": 0.9711, "step": 5950 }, { "epoch": 0.3, "learning_rate": 1.6356524402928983e-05, "loss": 0.901, "step": 5951 }, { "epoch": 0.3, "learning_rate": 1.635525409630435e-05, "loss": 0.9813, "step": 5952 }, { "epoch": 0.3, "learning_rate": 1.635398361761797e-05, "loss": 0.9912, "step": 5953 }, { "epoch": 0.3, "learning_rate": 1.6352712966904247e-05, "loss": 0.8862, "step": 5954 }, { "epoch": 0.3, "learning_rate": 1.6351442144197577e-05, "loss": 0.9376, "step": 5955 }, { "epoch": 0.3, "learning_rate": 1.6350171149532366e-05, "loss": 0.9999, "step": 5956 }, { "epoch": 0.3, "learning_rate": 1.6348899982943032e-05, "loss": 0.9341, "step": 5957 }, { "epoch": 0.3, "learning_rate": 1.6347628644463984e-05, "loss": 0.9745, "step": 5958 }, { "epoch": 0.3, "learning_rate": 1.634635713412964e-05, "loss": 0.8368, "step": 5959 }, { "epoch": 0.3, "learning_rate": 1.6345085451974435e-05, "loss": 1.0509, "step": 5960 }, { "epoch": 0.3, "learning_rate": 1.6343813598032786e-05, "loss": 0.9517, "step": 5961 }, { "epoch": 0.3, "learning_rate": 1.634254157233914e-05, "loss": 1.051, "step": 5962 }, { "epoch": 0.3, "learning_rate": 1.634126937492792e-05, "loss": 0.8687, "step": 5963 }, { "epoch": 0.3, "learning_rate": 1.6339997005833588e-05, "loss": 0.9728, "step": 5964 }, { "epoch": 0.3, "learning_rate": 1.6338724465090578e-05, "loss": 0.8599, "step": 5965 }, { "epoch": 0.3, "learning_rate": 1.6337451752733345e-05, "loss": 0.8498, "step": 5966 }, { "epoch": 0.3, "learning_rate": 1.633617886879635e-05, "loss": 0.9355, "step": 5967 }, { "epoch": 0.3, "learning_rate": 1.6334905813314052e-05, "loss": 1.0321, "step": 5968 }, { "epoch": 0.3, "learning_rate": 1.633363258632092e-05, "loss": 1.0086, "step": 5969 }, { "epoch": 0.3, "learning_rate": 1.6332359187851424e-05, "loss": 0.9233, "step": 5970 }, { "epoch": 0.3, "learning_rate": 1.633108561794004e-05, "loss": 0.9958, "step": 5971 }, { "epoch": 0.3, "learning_rate": 1.6329811876621252e-05, "loss": 0.969, "step": 5972 }, { "epoch": 0.3, "learning_rate": 1.6328537963929538e-05, "loss": 0.8501, "step": 5973 }, { "epoch": 0.3, "learning_rate": 1.6327263879899392e-05, "loss": 0.955, "step": 5974 }, { "epoch": 0.3, "learning_rate": 1.632598962456531e-05, "loss": 0.9468, "step": 5975 }, { "epoch": 0.3, "learning_rate": 1.6324715197961785e-05, "loss": 0.9636, "step": 5976 }, { "epoch": 0.3, "learning_rate": 1.6323440600123327e-05, "loss": 1.0442, "step": 5977 }, { "epoch": 0.3, "learning_rate": 1.6322165831084444e-05, "loss": 0.901, "step": 5978 }, { "epoch": 0.3, "learning_rate": 1.632089089087965e-05, "loss": 0.8842, "step": 5979 }, { "epoch": 0.3, "learning_rate": 1.6319615779543456e-05, "loss": 1.019, "step": 5980 }, { "epoch": 0.3, "learning_rate": 1.6318340497110384e-05, "loss": 0.941, "step": 5981 }, { "epoch": 0.3, "learning_rate": 1.631706504361497e-05, "loss": 0.8907, "step": 5982 }, { "epoch": 0.3, "learning_rate": 1.6315789419091747e-05, "loss": 1.0056, "step": 5983 }, { "epoch": 0.3, "learning_rate": 1.6314513623575234e-05, "loss": 0.9541, "step": 5984 }, { "epoch": 0.3, "learning_rate": 1.6313237657099988e-05, "loss": 0.9298, "step": 5985 }, { "epoch": 0.3, "learning_rate": 1.631196151970055e-05, "loss": 1.0113, "step": 5986 }, { "epoch": 0.3, "learning_rate": 1.6310685211411467e-05, "loss": 0.8453, "step": 5987 }, { "epoch": 0.3, "learning_rate": 1.6309408732267296e-05, "loss": 0.9944, "step": 5988 }, { "epoch": 0.3, "learning_rate": 1.6308132082302598e-05, "loss": 1.0052, "step": 5989 }, { "epoch": 0.3, "learning_rate": 1.6306855261551934e-05, "loss": 1.073, "step": 5990 }, { "epoch": 0.3, "learning_rate": 1.6305578270049874e-05, "loss": 0.8934, "step": 5991 }, { "epoch": 0.3, "learning_rate": 1.6304301107830992e-05, "loss": 1.0059, "step": 5992 }, { "epoch": 0.3, "learning_rate": 1.6303023774929863e-05, "loss": 0.9678, "step": 5993 }, { "epoch": 0.3, "learning_rate": 1.630174627138107e-05, "loss": 0.91, "step": 5994 }, { "epoch": 0.3, "learning_rate": 1.6300468597219204e-05, "loss": 0.9185, "step": 5995 }, { "epoch": 0.3, "learning_rate": 1.6299190752478857e-05, "loss": 0.9495, "step": 5996 }, { "epoch": 0.3, "learning_rate": 1.6297912737194615e-05, "loss": 0.9924, "step": 5997 }, { "epoch": 0.3, "learning_rate": 1.629663455140109e-05, "loss": 1.0123, "step": 5998 }, { "epoch": 0.3, "learning_rate": 1.6295356195132884e-05, "loss": 1.0574, "step": 5999 }, { "epoch": 0.3, "learning_rate": 1.6294077668424606e-05, "loss": 0.9833, "step": 6000 }, { "epoch": 0.3, "learning_rate": 1.629279897131087e-05, "loss": 0.9854, "step": 6001 }, { "epoch": 0.3, "learning_rate": 1.62915201038263e-05, "loss": 0.9989, "step": 6002 }, { "epoch": 0.3, "learning_rate": 1.6290241066005514e-05, "loss": 1.0036, "step": 6003 }, { "epoch": 0.31, "learning_rate": 1.6288961857883144e-05, "loss": 0.9566, "step": 6004 }, { "epoch": 0.31, "learning_rate": 1.6287682479493826e-05, "loss": 0.9855, "step": 6005 }, { "epoch": 0.31, "learning_rate": 1.628640293087219e-05, "loss": 0.9818, "step": 6006 }, { "epoch": 0.31, "learning_rate": 1.6285123212052886e-05, "loss": 0.9948, "step": 6007 }, { "epoch": 0.31, "learning_rate": 1.6283843323070552e-05, "loss": 1.0151, "step": 6008 }, { "epoch": 0.31, "learning_rate": 1.6282563263959852e-05, "loss": 0.9711, "step": 6009 }, { "epoch": 0.31, "learning_rate": 1.6281283034755432e-05, "loss": 0.9429, "step": 6010 }, { "epoch": 0.31, "learning_rate": 1.628000263549196e-05, "loss": 0.9918, "step": 6011 }, { "epoch": 0.31, "learning_rate": 1.6278722066204096e-05, "loss": 0.9334, "step": 6012 }, { "epoch": 0.31, "learning_rate": 1.627744132692651e-05, "loss": 0.9048, "step": 6013 }, { "epoch": 0.31, "learning_rate": 1.6276160417693883e-05, "loss": 0.9979, "step": 6014 }, { "epoch": 0.31, "learning_rate": 1.6274879338540886e-05, "loss": 1.0247, "step": 6015 }, { "epoch": 0.31, "learning_rate": 1.6273598089502208e-05, "loss": 0.9707, "step": 6016 }, { "epoch": 0.31, "learning_rate": 1.627231667061254e-05, "loss": 1.0159, "step": 6017 }, { "epoch": 0.31, "learning_rate": 1.6271035081906568e-05, "loss": 0.9892, "step": 6018 }, { "epoch": 0.31, "learning_rate": 1.626975332341899e-05, "loss": 0.9904, "step": 6019 }, { "epoch": 0.31, "learning_rate": 1.626847139518452e-05, "loss": 0.9973, "step": 6020 }, { "epoch": 0.31, "learning_rate": 1.626718929723785e-05, "loss": 0.9588, "step": 6021 }, { "epoch": 0.31, "learning_rate": 1.6265907029613697e-05, "loss": 0.9338, "step": 6022 }, { "epoch": 0.31, "learning_rate": 1.626462459234678e-05, "loss": 0.9354, "step": 6023 }, { "epoch": 0.31, "learning_rate": 1.6263341985471815e-05, "loss": 0.9942, "step": 6024 }, { "epoch": 0.31, "learning_rate": 1.626205920902353e-05, "loss": 1.0514, "step": 6025 }, { "epoch": 0.31, "learning_rate": 1.6260776263036656e-05, "loss": 0.9753, "step": 6026 }, { "epoch": 0.31, "learning_rate": 1.6259493147545926e-05, "loss": 1.0405, "step": 6027 }, { "epoch": 0.31, "learning_rate": 1.6258209862586076e-05, "loss": 1.0065, "step": 6028 }, { "epoch": 0.31, "learning_rate": 1.6256926408191853e-05, "loss": 0.9904, "step": 6029 }, { "epoch": 0.31, "learning_rate": 1.6255642784398004e-05, "loss": 0.9529, "step": 6030 }, { "epoch": 0.31, "learning_rate": 1.6254358991239283e-05, "loss": 0.9827, "step": 6031 }, { "epoch": 0.31, "learning_rate": 1.6253075028750447e-05, "loss": 1.1058, "step": 6032 }, { "epoch": 0.31, "learning_rate": 1.6251790896966256e-05, "loss": 0.9982, "step": 6033 }, { "epoch": 0.31, "learning_rate": 1.6250506595921478e-05, "loss": 0.951, "step": 6034 }, { "epoch": 0.31, "learning_rate": 1.6249222125650884e-05, "loss": 1.0016, "step": 6035 }, { "epoch": 0.31, "learning_rate": 1.624793748618925e-05, "loss": 0.9509, "step": 6036 }, { "epoch": 0.31, "learning_rate": 1.6246652677571355e-05, "loss": 0.9272, "step": 6037 }, { "epoch": 0.31, "learning_rate": 1.6245367699831986e-05, "loss": 1.0279, "step": 6038 }, { "epoch": 0.31, "learning_rate": 1.624408255300593e-05, "loss": 0.9958, "step": 6039 }, { "epoch": 0.31, "learning_rate": 1.6242797237127983e-05, "loss": 0.976, "step": 6040 }, { "epoch": 0.31, "learning_rate": 1.6241511752232942e-05, "loss": 1.0099, "step": 6041 }, { "epoch": 0.31, "learning_rate": 1.624022609835561e-05, "loss": 0.879, "step": 6042 }, { "epoch": 0.31, "learning_rate": 1.6238940275530796e-05, "loss": 0.882, "step": 6043 }, { "epoch": 0.31, "learning_rate": 1.6237654283793313e-05, "loss": 0.987, "step": 6044 }, { "epoch": 0.31, "learning_rate": 1.6236368123177976e-05, "loss": 0.9675, "step": 6045 }, { "epoch": 0.31, "learning_rate": 1.6235081793719607e-05, "loss": 1.027, "step": 6046 }, { "epoch": 0.31, "learning_rate": 1.623379529545303e-05, "loss": 1.0419, "step": 6047 }, { "epoch": 0.31, "learning_rate": 1.623250862841308e-05, "loss": 1.002, "step": 6048 }, { "epoch": 0.31, "learning_rate": 1.623122179263459e-05, "loss": 0.9831, "step": 6049 }, { "epoch": 0.31, "learning_rate": 1.62299347881524e-05, "loss": 0.9524, "step": 6050 }, { "epoch": 0.31, "learning_rate": 1.622864761500135e-05, "loss": 0.9399, "step": 6051 }, { "epoch": 0.31, "learning_rate": 1.62273602732163e-05, "loss": 0.9144, "step": 6052 }, { "epoch": 0.31, "learning_rate": 1.6226072762832086e-05, "loss": 1.0011, "step": 6053 }, { "epoch": 0.31, "learning_rate": 1.6224785083883584e-05, "loss": 0.8798, "step": 6054 }, { "epoch": 0.31, "learning_rate": 1.6223497236405642e-05, "loss": 1.0551, "step": 6055 }, { "epoch": 0.31, "learning_rate": 1.6222209220433138e-05, "loss": 0.9967, "step": 6056 }, { "epoch": 0.31, "learning_rate": 1.622092103600094e-05, "loss": 0.9509, "step": 6057 }, { "epoch": 0.31, "learning_rate": 1.621963268314392e-05, "loss": 1.028, "step": 6058 }, { "epoch": 0.31, "learning_rate": 1.6218344161896965e-05, "loss": 0.9039, "step": 6059 }, { "epoch": 0.31, "learning_rate": 1.6217055472294954e-05, "loss": 0.9822, "step": 6060 }, { "epoch": 0.31, "learning_rate": 1.6215766614372785e-05, "loss": 1.0176, "step": 6061 }, { "epoch": 0.31, "learning_rate": 1.6214477588165346e-05, "loss": 0.8771, "step": 6062 }, { "epoch": 0.31, "learning_rate": 1.6213188393707534e-05, "loss": 1.0302, "step": 6063 }, { "epoch": 0.31, "learning_rate": 1.6211899031034268e-05, "loss": 1.0715, "step": 6064 }, { "epoch": 0.31, "learning_rate": 1.6210609500180435e-05, "loss": 1.0007, "step": 6065 }, { "epoch": 0.31, "learning_rate": 1.620931980118096e-05, "loss": 1.0457, "step": 6066 }, { "epoch": 0.31, "learning_rate": 1.6208029934070757e-05, "loss": 0.9008, "step": 6067 }, { "epoch": 0.31, "learning_rate": 1.6206739898884745e-05, "loss": 0.9924, "step": 6068 }, { "epoch": 0.31, "learning_rate": 1.620544969565786e-05, "loss": 0.9626, "step": 6069 }, { "epoch": 0.31, "learning_rate": 1.6204159324425024e-05, "loss": 0.9553, "step": 6070 }, { "epoch": 0.31, "learning_rate": 1.620286878522117e-05, "loss": 0.9921, "step": 6071 }, { "epoch": 0.31, "learning_rate": 1.620157807808125e-05, "loss": 0.9782, "step": 6072 }, { "epoch": 0.31, "learning_rate": 1.62002872030402e-05, "loss": 0.9363, "step": 6073 }, { "epoch": 0.31, "learning_rate": 1.6198996160132972e-05, "loss": 0.9188, "step": 6074 }, { "epoch": 0.31, "learning_rate": 1.6197704949394514e-05, "loss": 0.9525, "step": 6075 }, { "epoch": 0.31, "learning_rate": 1.619641357085979e-05, "loss": 1.061, "step": 6076 }, { "epoch": 0.31, "learning_rate": 1.619512202456376e-05, "loss": 0.9543, "step": 6077 }, { "epoch": 0.31, "learning_rate": 1.6193830310541397e-05, "loss": 0.8922, "step": 6078 }, { "epoch": 0.31, "learning_rate": 1.6192538428827663e-05, "loss": 0.8589, "step": 6079 }, { "epoch": 0.31, "learning_rate": 1.619124637945754e-05, "loss": 0.94, "step": 6080 }, { "epoch": 0.31, "learning_rate": 1.618995416246601e-05, "loss": 0.9329, "step": 6081 }, { "epoch": 0.31, "learning_rate": 1.618866177788806e-05, "loss": 0.9631, "step": 6082 }, { "epoch": 0.31, "learning_rate": 1.618736922575867e-05, "loss": 0.9349, "step": 6083 }, { "epoch": 0.31, "learning_rate": 1.6186076506112843e-05, "loss": 1.0461, "step": 6084 }, { "epoch": 0.31, "learning_rate": 1.6184783618985577e-05, "loss": 0.9673, "step": 6085 }, { "epoch": 0.31, "learning_rate": 1.6183490564411874e-05, "loss": 0.8775, "step": 6086 }, { "epoch": 0.31, "learning_rate": 1.6182197342426744e-05, "loss": 0.944, "step": 6087 }, { "epoch": 0.31, "learning_rate": 1.6180903953065197e-05, "loss": 0.9586, "step": 6088 }, { "epoch": 0.31, "learning_rate": 1.617961039636225e-05, "loss": 0.9719, "step": 6089 }, { "epoch": 0.31, "learning_rate": 1.6178316672352928e-05, "loss": 0.9505, "step": 6090 }, { "epoch": 0.31, "learning_rate": 1.6177022781072257e-05, "loss": 0.9638, "step": 6091 }, { "epoch": 0.31, "learning_rate": 1.6175728722555264e-05, "loss": 1.0376, "step": 6092 }, { "epoch": 0.31, "learning_rate": 1.6174434496836986e-05, "loss": 0.9607, "step": 6093 }, { "epoch": 0.31, "learning_rate": 1.6173140103952463e-05, "loss": 0.8677, "step": 6094 }, { "epoch": 0.31, "learning_rate": 1.6171845543936742e-05, "loss": 0.9406, "step": 6095 }, { "epoch": 0.31, "learning_rate": 1.6170550816824866e-05, "loss": 0.9828, "step": 6096 }, { "epoch": 0.31, "learning_rate": 1.6169255922651893e-05, "loss": 0.9251, "step": 6097 }, { "epoch": 0.31, "learning_rate": 1.616796086145288e-05, "loss": 1.0956, "step": 6098 }, { "epoch": 0.31, "learning_rate": 1.616666563326289e-05, "loss": 0.9445, "step": 6099 }, { "epoch": 0.31, "learning_rate": 1.616537023811698e-05, "loss": 0.9412, "step": 6100 }, { "epoch": 0.31, "learning_rate": 1.6164074676050242e-05, "loss": 0.8778, "step": 6101 }, { "epoch": 0.31, "learning_rate": 1.6162778947097734e-05, "loss": 0.9522, "step": 6102 }, { "epoch": 0.31, "learning_rate": 1.6161483051294544e-05, "loss": 0.9565, "step": 6103 }, { "epoch": 0.31, "learning_rate": 1.616018698867576e-05, "loss": 0.9485, "step": 6104 }, { "epoch": 0.31, "learning_rate": 1.615889075927646e-05, "loss": 1.031, "step": 6105 }, { "epoch": 0.31, "learning_rate": 1.615759436313175e-05, "loss": 0.9051, "step": 6106 }, { "epoch": 0.31, "learning_rate": 1.615629780027672e-05, "loss": 1.0463, "step": 6107 }, { "epoch": 0.31, "learning_rate": 1.615500107074648e-05, "loss": 0.9887, "step": 6108 }, { "epoch": 0.31, "learning_rate": 1.615370417457613e-05, "loss": 0.9244, "step": 6109 }, { "epoch": 0.31, "learning_rate": 1.615240711180079e-05, "loss": 0.9475, "step": 6110 }, { "epoch": 0.31, "learning_rate": 1.6151109882455577e-05, "loss": 1.0764, "step": 6111 }, { "epoch": 0.31, "learning_rate": 1.61498124865756e-05, "loss": 0.9138, "step": 6112 }, { "epoch": 0.31, "learning_rate": 1.6148514924195997e-05, "loss": 1.0185, "step": 6113 }, { "epoch": 0.31, "learning_rate": 1.6147217195351896e-05, "loss": 0.9321, "step": 6114 }, { "epoch": 0.31, "learning_rate": 1.6145919300078426e-05, "loss": 1.0158, "step": 6115 }, { "epoch": 0.31, "learning_rate": 1.614462123841073e-05, "loss": 0.906, "step": 6116 }, { "epoch": 0.31, "learning_rate": 1.6143323010383954e-05, "loss": 1.0052, "step": 6117 }, { "epoch": 0.31, "learning_rate": 1.614202461603324e-05, "loss": 0.8976, "step": 6118 }, { "epoch": 0.31, "learning_rate": 1.6140726055393746e-05, "loss": 1.0391, "step": 6119 }, { "epoch": 0.31, "learning_rate": 1.6139427328500628e-05, "loss": 0.9783, "step": 6120 }, { "epoch": 0.31, "learning_rate": 1.6138128435389044e-05, "loss": 0.9854, "step": 6121 }, { "epoch": 0.31, "learning_rate": 1.613682937609417e-05, "loss": 0.9864, "step": 6122 }, { "epoch": 0.31, "learning_rate": 1.6135530150651163e-05, "loss": 1.0057, "step": 6123 }, { "epoch": 0.31, "learning_rate": 1.6134230759095207e-05, "loss": 0.9713, "step": 6124 }, { "epoch": 0.31, "learning_rate": 1.6132931201461478e-05, "loss": 0.8866, "step": 6125 }, { "epoch": 0.31, "learning_rate": 1.6131631477785164e-05, "loss": 0.9587, "step": 6126 }, { "epoch": 0.31, "learning_rate": 1.613033158810145e-05, "loss": 0.9791, "step": 6127 }, { "epoch": 0.31, "learning_rate": 1.612903153244553e-05, "loss": 0.8972, "step": 6128 }, { "epoch": 0.31, "learning_rate": 1.6127731310852604e-05, "loss": 0.9049, "step": 6129 }, { "epoch": 0.31, "learning_rate": 1.612643092335787e-05, "loss": 1.0982, "step": 6130 }, { "epoch": 0.31, "learning_rate": 1.6125130369996535e-05, "loss": 1.0708, "step": 6131 }, { "epoch": 0.31, "learning_rate": 1.612382965080382e-05, "loss": 0.9662, "step": 6132 }, { "epoch": 0.31, "learning_rate": 1.6122528765814922e-05, "loss": 0.9421, "step": 6133 }, { "epoch": 0.31, "learning_rate": 1.6121227715065076e-05, "loss": 0.9225, "step": 6134 }, { "epoch": 0.31, "learning_rate": 1.6119926498589504e-05, "loss": 0.964, "step": 6135 }, { "epoch": 0.31, "learning_rate": 1.6118625116423432e-05, "loss": 1.0338, "step": 6136 }, { "epoch": 0.31, "learning_rate": 1.6117323568602096e-05, "loss": 0.9921, "step": 6137 }, { "epoch": 0.31, "learning_rate": 1.6116021855160727e-05, "loss": 1.0132, "step": 6138 }, { "epoch": 0.31, "learning_rate": 1.6114719976134577e-05, "loss": 0.9915, "step": 6139 }, { "epoch": 0.31, "learning_rate": 1.611341793155889e-05, "loss": 1.0356, "step": 6140 }, { "epoch": 0.31, "learning_rate": 1.6112115721468917e-05, "loss": 1.0524, "step": 6141 }, { "epoch": 0.31, "learning_rate": 1.6110813345899914e-05, "loss": 0.9146, "step": 6142 }, { "epoch": 0.31, "learning_rate": 1.610951080488714e-05, "loss": 0.9618, "step": 6143 }, { "epoch": 0.31, "learning_rate": 1.6108208098465857e-05, "loss": 1.0064, "step": 6144 }, { "epoch": 0.31, "learning_rate": 1.6106905226671346e-05, "loss": 0.9774, "step": 6145 }, { "epoch": 0.31, "learning_rate": 1.6105602189538868e-05, "loss": 0.9218, "step": 6146 }, { "epoch": 0.31, "learning_rate": 1.6104298987103708e-05, "loss": 0.9341, "step": 6147 }, { "epoch": 0.31, "learning_rate": 1.6102995619401145e-05, "loss": 0.9577, "step": 6148 }, { "epoch": 0.31, "learning_rate": 1.6101692086466472e-05, "loss": 0.9203, "step": 6149 }, { "epoch": 0.31, "learning_rate": 1.6100388388334977e-05, "loss": 0.9824, "step": 6150 }, { "epoch": 0.31, "learning_rate": 1.6099084525041954e-05, "loss": 0.9822, "step": 6151 }, { "epoch": 0.31, "learning_rate": 1.609778049662271e-05, "loss": 0.9959, "step": 6152 }, { "epoch": 0.31, "learning_rate": 1.6096476303112548e-05, "loss": 1.0276, "step": 6153 }, { "epoch": 0.31, "learning_rate": 1.609517194454677e-05, "loss": 1.1183, "step": 6154 }, { "epoch": 0.31, "learning_rate": 1.60938674209607e-05, "loss": 0.9423, "step": 6155 }, { "epoch": 0.31, "learning_rate": 1.6092562732389655e-05, "loss": 1.0006, "step": 6156 }, { "epoch": 0.31, "learning_rate": 1.609125787886895e-05, "loss": 0.9681, "step": 6157 }, { "epoch": 0.31, "learning_rate": 1.6089952860433925e-05, "loss": 0.9637, "step": 6158 }, { "epoch": 0.31, "learning_rate": 1.6088647677119902e-05, "loss": 0.9391, "step": 6159 }, { "epoch": 0.31, "learning_rate": 1.6087342328962222e-05, "loss": 0.9017, "step": 6160 }, { "epoch": 0.31, "learning_rate": 1.6086036815996223e-05, "loss": 0.9567, "step": 6161 }, { "epoch": 0.31, "learning_rate": 1.6084731138257255e-05, "loss": 0.9955, "step": 6162 }, { "epoch": 0.31, "learning_rate": 1.6083425295780664e-05, "loss": 0.975, "step": 6163 }, { "epoch": 0.31, "learning_rate": 1.6082119288601803e-05, "loss": 0.9801, "step": 6164 }, { "epoch": 0.31, "learning_rate": 1.6080813116756037e-05, "loss": 0.8976, "step": 6165 }, { "epoch": 0.31, "learning_rate": 1.607950678027872e-05, "loss": 0.9923, "step": 6166 }, { "epoch": 0.31, "learning_rate": 1.6078200279205232e-05, "loss": 0.9518, "step": 6167 }, { "epoch": 0.31, "learning_rate": 1.607689361357093e-05, "loss": 0.913, "step": 6168 }, { "epoch": 0.31, "learning_rate": 1.607558678341121e-05, "loss": 0.913, "step": 6169 }, { "epoch": 0.31, "learning_rate": 1.6074279788761433e-05, "loss": 0.8631, "step": 6170 }, { "epoch": 0.31, "learning_rate": 1.607297262965699e-05, "loss": 1.0234, "step": 6171 }, { "epoch": 0.31, "learning_rate": 1.6071665306133284e-05, "loss": 1.0256, "step": 6172 }, { "epoch": 0.31, "learning_rate": 1.6070357818225696e-05, "loss": 0.9784, "step": 6173 }, { "epoch": 0.31, "learning_rate": 1.6069050165969626e-05, "loss": 0.9873, "step": 6174 }, { "epoch": 0.31, "learning_rate": 1.606774234940048e-05, "loss": 0.9159, "step": 6175 }, { "epoch": 0.31, "learning_rate": 1.6066434368553667e-05, "loss": 0.9168, "step": 6176 }, { "epoch": 0.31, "learning_rate": 1.60651262234646e-05, "loss": 0.9139, "step": 6177 }, { "epoch": 0.31, "learning_rate": 1.606381791416869e-05, "loss": 1.0434, "step": 6178 }, { "epoch": 0.31, "learning_rate": 1.606250944070136e-05, "loss": 0.9595, "step": 6179 }, { "epoch": 0.31, "learning_rate": 1.606120080309804e-05, "loss": 0.889, "step": 6180 }, { "epoch": 0.31, "learning_rate": 1.605989200139416e-05, "loss": 1.1262, "step": 6181 }, { "epoch": 0.31, "learning_rate": 1.6058583035625144e-05, "loss": 1.0339, "step": 6182 }, { "epoch": 0.31, "learning_rate": 1.605727390582644e-05, "loss": 0.9064, "step": 6183 }, { "epoch": 0.31, "learning_rate": 1.6055964612033493e-05, "loss": 0.9251, "step": 6184 }, { "epoch": 0.31, "learning_rate": 1.6054655154281743e-05, "loss": 0.908, "step": 6185 }, { "epoch": 0.31, "learning_rate": 1.6053345532606647e-05, "loss": 0.9575, "step": 6186 }, { "epoch": 0.31, "learning_rate": 1.6052035747043663e-05, "loss": 1.0277, "step": 6187 }, { "epoch": 0.31, "learning_rate": 1.605072579762825e-05, "loss": 0.9341, "step": 6188 }, { "epoch": 0.31, "learning_rate": 1.6049415684395872e-05, "loss": 1.0154, "step": 6189 }, { "epoch": 0.31, "learning_rate": 1.6048105407382e-05, "loss": 1.027, "step": 6190 }, { "epoch": 0.31, "learning_rate": 1.604679496662211e-05, "loss": 0.9628, "step": 6191 }, { "epoch": 0.31, "learning_rate": 1.604548436215168e-05, "loss": 1.0148, "step": 6192 }, { "epoch": 0.31, "learning_rate": 1.6044173594006195e-05, "loss": 0.9619, "step": 6193 }, { "epoch": 0.31, "learning_rate": 1.6042862662221134e-05, "loss": 0.9497, "step": 6194 }, { "epoch": 0.31, "learning_rate": 1.6041551566832e-05, "loss": 1.0204, "step": 6195 }, { "epoch": 0.31, "learning_rate": 1.6040240307874286e-05, "loss": 0.9268, "step": 6196 }, { "epoch": 0.31, "learning_rate": 1.603892888538349e-05, "loss": 0.8606, "step": 6197 }, { "epoch": 0.31, "learning_rate": 1.603761729939512e-05, "loss": 0.9862, "step": 6198 }, { "epoch": 0.31, "learning_rate": 1.603630554994468e-05, "loss": 0.9644, "step": 6199 }, { "epoch": 0.31, "learning_rate": 1.60349936370677e-05, "loss": 0.8976, "step": 6200 }, { "epoch": 0.32, "learning_rate": 1.6033681560799677e-05, "loss": 0.9544, "step": 6201 }, { "epoch": 0.32, "learning_rate": 1.603236932117615e-05, "loss": 0.9971, "step": 6202 }, { "epoch": 0.32, "learning_rate": 1.6031056918232642e-05, "loss": 0.9536, "step": 6203 }, { "epoch": 0.32, "learning_rate": 1.6029744352004687e-05, "loss": 0.9714, "step": 6204 }, { "epoch": 0.32, "learning_rate": 1.6028431622527815e-05, "loss": 0.8241, "step": 6205 }, { "epoch": 0.32, "learning_rate": 1.602711872983757e-05, "loss": 0.9749, "step": 6206 }, { "epoch": 0.32, "learning_rate": 1.6025805673969502e-05, "loss": 0.9372, "step": 6207 }, { "epoch": 0.32, "learning_rate": 1.6024492454959153e-05, "loss": 0.9867, "step": 6208 }, { "epoch": 0.32, "learning_rate": 1.602317907284208e-05, "loss": 1.0244, "step": 6209 }, { "epoch": 0.32, "learning_rate": 1.6021865527653842e-05, "loss": 0.9979, "step": 6210 }, { "epoch": 0.32, "learning_rate": 1.6020551819430003e-05, "loss": 0.8994, "step": 6211 }, { "epoch": 0.32, "learning_rate": 1.601923794820613e-05, "loss": 0.9573, "step": 6212 }, { "epoch": 0.32, "learning_rate": 1.601792391401779e-05, "loss": 1.0469, "step": 6213 }, { "epoch": 0.32, "learning_rate": 1.6016609716900565e-05, "loss": 0.9548, "step": 6214 }, { "epoch": 0.32, "learning_rate": 1.6015295356890033e-05, "loss": 0.9391, "step": 6215 }, { "epoch": 0.32, "learning_rate": 1.6013980834021777e-05, "loss": 0.9681, "step": 6216 }, { "epoch": 0.32, "learning_rate": 1.601266614833139e-05, "loss": 0.9706, "step": 6217 }, { "epoch": 0.32, "learning_rate": 1.6011351299854464e-05, "loss": 0.9318, "step": 6218 }, { "epoch": 0.32, "learning_rate": 1.6010036288626595e-05, "loss": 0.97, "step": 6219 }, { "epoch": 0.32, "learning_rate": 1.600872111468339e-05, "loss": 0.9304, "step": 6220 }, { "epoch": 0.32, "learning_rate": 1.6007405778060453e-05, "loss": 0.9474, "step": 6221 }, { "epoch": 0.32, "learning_rate": 1.6006090278793395e-05, "loss": 0.8904, "step": 6222 }, { "epoch": 0.32, "learning_rate": 1.6004774616917835e-05, "loss": 1.0593, "step": 6223 }, { "epoch": 0.32, "learning_rate": 1.6003458792469387e-05, "loss": 0.9222, "step": 6224 }, { "epoch": 0.32, "learning_rate": 1.6002142805483686e-05, "loss": 0.9791, "step": 6225 }, { "epoch": 0.32, "learning_rate": 1.6000826655996347e-05, "loss": 0.9058, "step": 6226 }, { "epoch": 0.32, "learning_rate": 1.5999510344043017e-05, "loss": 0.8988, "step": 6227 }, { "epoch": 0.32, "learning_rate": 1.599819386965932e-05, "loss": 0.9822, "step": 6228 }, { "epoch": 0.32, "learning_rate": 1.599687723288091e-05, "loss": 0.9778, "step": 6229 }, { "epoch": 0.32, "learning_rate": 1.599556043374343e-05, "loss": 0.933, "step": 6230 }, { "epoch": 0.32, "learning_rate": 1.599424347228253e-05, "loss": 1.0687, "step": 6231 }, { "epoch": 0.32, "learning_rate": 1.5992926348533863e-05, "loss": 0.892, "step": 6232 }, { "epoch": 0.32, "learning_rate": 1.5991609062533094e-05, "loss": 0.9628, "step": 6233 }, { "epoch": 0.32, "learning_rate": 1.599029161431588e-05, "loss": 0.9123, "step": 6234 }, { "epoch": 0.32, "learning_rate": 1.5988974003917898e-05, "loss": 0.9883, "step": 6235 }, { "epoch": 0.32, "learning_rate": 1.5987656231374814e-05, "loss": 1.0114, "step": 6236 }, { "epoch": 0.32, "learning_rate": 1.598633829672231e-05, "loss": 0.8572, "step": 6237 }, { "epoch": 0.32, "learning_rate": 1.5985020199996065e-05, "loss": 0.866, "step": 6238 }, { "epoch": 0.32, "learning_rate": 1.5983701941231763e-05, "loss": 1.0117, "step": 6239 }, { "epoch": 0.32, "learning_rate": 1.59823835204651e-05, "loss": 0.9462, "step": 6240 }, { "epoch": 0.32, "learning_rate": 1.5981064937731768e-05, "loss": 0.9772, "step": 6241 }, { "epoch": 0.32, "learning_rate": 1.5979746193067463e-05, "loss": 1.0072, "step": 6242 }, { "epoch": 0.32, "learning_rate": 1.5978427286507895e-05, "loss": 1.0572, "step": 6243 }, { "epoch": 0.32, "learning_rate": 1.597710821808877e-05, "loss": 1.0107, "step": 6244 }, { "epoch": 0.32, "learning_rate": 1.5975788987845792e-05, "loss": 1.018, "step": 6245 }, { "epoch": 0.32, "learning_rate": 1.597446959581469e-05, "loss": 0.9822, "step": 6246 }, { "epoch": 0.32, "learning_rate": 1.597315004203118e-05, "loss": 0.861, "step": 6247 }, { "epoch": 0.32, "learning_rate": 1.5971830326530988e-05, "loss": 0.9992, "step": 6248 }, { "epoch": 0.32, "learning_rate": 1.5970510449349843e-05, "loss": 0.9346, "step": 6249 }, { "epoch": 0.32, "learning_rate": 1.596919041052348e-05, "loss": 0.9883, "step": 6250 }, { "epoch": 0.32, "learning_rate": 1.596787021008764e-05, "loss": 0.937, "step": 6251 }, { "epoch": 0.32, "learning_rate": 1.5966549848078058e-05, "loss": 0.9418, "step": 6252 }, { "epoch": 0.32, "learning_rate": 1.596522932453049e-05, "loss": 0.9829, "step": 6253 }, { "epoch": 0.32, "learning_rate": 1.596390863948068e-05, "loss": 1.0065, "step": 6254 }, { "epoch": 0.32, "learning_rate": 1.596258779296439e-05, "loss": 1.0047, "step": 6255 }, { "epoch": 0.32, "learning_rate": 1.5961266785017385e-05, "loss": 1.0306, "step": 6256 }, { "epoch": 0.32, "learning_rate": 1.595994561567542e-05, "loss": 0.88, "step": 6257 }, { "epoch": 0.32, "learning_rate": 1.595862428497427e-05, "loss": 0.8928, "step": 6258 }, { "epoch": 0.32, "learning_rate": 1.5957302792949707e-05, "loss": 1.0007, "step": 6259 }, { "epoch": 0.32, "learning_rate": 1.5955981139637504e-05, "loss": 0.9297, "step": 6260 }, { "epoch": 0.32, "learning_rate": 1.5954659325073453e-05, "loss": 1.0156, "step": 6261 }, { "epoch": 0.32, "learning_rate": 1.5953337349293338e-05, "loss": 0.9039, "step": 6262 }, { "epoch": 0.32, "learning_rate": 1.5952015212332942e-05, "loss": 0.9818, "step": 6263 }, { "epoch": 0.32, "learning_rate": 1.595069291422807e-05, "loss": 0.9155, "step": 6264 }, { "epoch": 0.32, "learning_rate": 1.594937045501452e-05, "loss": 0.9401, "step": 6265 }, { "epoch": 0.32, "learning_rate": 1.5948047834728092e-05, "loss": 0.9255, "step": 6266 }, { "epoch": 0.32, "learning_rate": 1.5946725053404597e-05, "loss": 0.9419, "step": 6267 }, { "epoch": 0.32, "learning_rate": 1.5945402111079847e-05, "loss": 1.0303, "step": 6268 }, { "epoch": 0.32, "learning_rate": 1.5944079007789662e-05, "loss": 1.0065, "step": 6269 }, { "epoch": 0.32, "learning_rate": 1.5942755743569864e-05, "loss": 0.9706, "step": 6270 }, { "epoch": 0.32, "learning_rate": 1.5941432318456277e-05, "loss": 0.8884, "step": 6271 }, { "epoch": 0.32, "learning_rate": 1.5940108732484726e-05, "loss": 1.0052, "step": 6272 }, { "epoch": 0.32, "learning_rate": 1.5938784985691055e-05, "loss": 0.9666, "step": 6273 }, { "epoch": 0.32, "learning_rate": 1.59374610781111e-05, "loss": 0.9064, "step": 6274 }, { "epoch": 0.32, "learning_rate": 1.59361370097807e-05, "loss": 1.0295, "step": 6275 }, { "epoch": 0.32, "learning_rate": 1.5934812780735704e-05, "loss": 0.9721, "step": 6276 }, { "epoch": 0.32, "learning_rate": 1.593348839101197e-05, "loss": 0.9902, "step": 6277 }, { "epoch": 0.32, "learning_rate": 1.5932163840645348e-05, "loss": 0.952, "step": 6278 }, { "epoch": 0.32, "learning_rate": 1.5930839129671705e-05, "loss": 0.8743, "step": 6279 }, { "epoch": 0.32, "learning_rate": 1.59295142581269e-05, "loss": 1.0469, "step": 6280 }, { "epoch": 0.32, "learning_rate": 1.5928189226046806e-05, "loss": 1.0963, "step": 6281 }, { "epoch": 0.32, "learning_rate": 1.5926864033467295e-05, "loss": 0.9255, "step": 6282 }, { "epoch": 0.32, "learning_rate": 1.5925538680424246e-05, "loss": 0.8829, "step": 6283 }, { "epoch": 0.32, "learning_rate": 1.5924213166953543e-05, "loss": 1.0558, "step": 6284 }, { "epoch": 0.32, "learning_rate": 1.592288749309107e-05, "loss": 0.9815, "step": 6285 }, { "epoch": 0.32, "learning_rate": 1.5921561658872716e-05, "loss": 0.9498, "step": 6286 }, { "epoch": 0.32, "learning_rate": 1.5920235664334388e-05, "loss": 0.8449, "step": 6287 }, { "epoch": 0.32, "learning_rate": 1.5918909509511974e-05, "loss": 1.0133, "step": 6288 }, { "epoch": 0.32, "learning_rate": 1.591758319444138e-05, "loss": 0.971, "step": 6289 }, { "epoch": 0.32, "learning_rate": 1.591625671915852e-05, "loss": 1.019, "step": 6290 }, { "epoch": 0.32, "learning_rate": 1.59149300836993e-05, "loss": 0.9009, "step": 6291 }, { "epoch": 0.32, "learning_rate": 1.5913603288099647e-05, "loss": 0.9387, "step": 6292 }, { "epoch": 0.32, "learning_rate": 1.5912276332395472e-05, "loss": 0.9772, "step": 6293 }, { "epoch": 0.32, "learning_rate": 1.5910949216622708e-05, "loss": 0.9891, "step": 6294 }, { "epoch": 0.32, "learning_rate": 1.590962194081728e-05, "loss": 0.9511, "step": 6295 }, { "epoch": 0.32, "learning_rate": 1.5908294505015126e-05, "loss": 0.9489, "step": 6296 }, { "epoch": 0.32, "learning_rate": 1.5906966909252185e-05, "loss": 0.831, "step": 6297 }, { "epoch": 0.32, "learning_rate": 1.59056391535644e-05, "loss": 1.0116, "step": 6298 }, { "epoch": 0.32, "learning_rate": 1.5904311237987716e-05, "loss": 0.9769, "step": 6299 }, { "epoch": 0.32, "learning_rate": 1.5902983162558088e-05, "loss": 1.0034, "step": 6300 }, { "epoch": 0.32, "learning_rate": 1.5901654927311472e-05, "loss": 0.9775, "step": 6301 }, { "epoch": 0.32, "learning_rate": 1.5900326532283824e-05, "loss": 0.9336, "step": 6302 }, { "epoch": 0.32, "learning_rate": 1.589899797751112e-05, "loss": 1.0258, "step": 6303 }, { "epoch": 0.32, "learning_rate": 1.5897669263029314e-05, "loss": 0.8719, "step": 6304 }, { "epoch": 0.32, "learning_rate": 1.5896340388874393e-05, "loss": 1.0148, "step": 6305 }, { "epoch": 0.32, "learning_rate": 1.5895011355082327e-05, "loss": 0.881, "step": 6306 }, { "epoch": 0.32, "learning_rate": 1.58936821616891e-05, "loss": 0.9716, "step": 6307 }, { "epoch": 0.32, "learning_rate": 1.58923528087307e-05, "loss": 0.8745, "step": 6308 }, { "epoch": 0.32, "learning_rate": 1.5891023296243116e-05, "loss": 0.9784, "step": 6309 }, { "epoch": 0.32, "learning_rate": 1.5889693624262342e-05, "loss": 1.027, "step": 6310 }, { "epoch": 0.32, "learning_rate": 1.588836379282438e-05, "loss": 0.9561, "step": 6311 }, { "epoch": 0.32, "learning_rate": 1.5887033801965234e-05, "loss": 0.9518, "step": 6312 }, { "epoch": 0.32, "learning_rate": 1.5885703651720913e-05, "loss": 0.9971, "step": 6313 }, { "epoch": 0.32, "learning_rate": 1.5884373342127425e-05, "loss": 0.9734, "step": 6314 }, { "epoch": 0.32, "learning_rate": 1.588304287322079e-05, "loss": 0.9224, "step": 6315 }, { "epoch": 0.32, "learning_rate": 1.5881712245037027e-05, "loss": 1.049, "step": 6316 }, { "epoch": 0.32, "learning_rate": 1.5880381457612165e-05, "loss": 0.9511, "step": 6317 }, { "epoch": 0.32, "learning_rate": 1.587905051098223e-05, "loss": 0.9855, "step": 6318 }, { "epoch": 0.32, "learning_rate": 1.5877719405183256e-05, "loss": 0.9335, "step": 6319 }, { "epoch": 0.32, "learning_rate": 1.5876388140251285e-05, "loss": 1.0235, "step": 6320 }, { "epoch": 0.32, "learning_rate": 1.5875056716222355e-05, "loss": 1.0581, "step": 6321 }, { "epoch": 0.32, "learning_rate": 1.5873725133132514e-05, "loss": 0.9882, "step": 6322 }, { "epoch": 0.32, "learning_rate": 1.5872393391017817e-05, "loss": 1.0095, "step": 6323 }, { "epoch": 0.32, "learning_rate": 1.5871061489914317e-05, "loss": 0.9073, "step": 6324 }, { "epoch": 0.32, "learning_rate": 1.586972942985807e-05, "loss": 0.9304, "step": 6325 }, { "epoch": 0.32, "learning_rate": 1.5868397210885146e-05, "loss": 0.8262, "step": 6326 }, { "epoch": 0.32, "learning_rate": 1.5867064833031612e-05, "loss": 0.9137, "step": 6327 }, { "epoch": 0.32, "learning_rate": 1.586573229633354e-05, "loss": 1.0267, "step": 6328 }, { "epoch": 0.32, "learning_rate": 1.5864399600827003e-05, "loss": 0.939, "step": 6329 }, { "epoch": 0.32, "learning_rate": 1.5863066746548088e-05, "loss": 1.0057, "step": 6330 }, { "epoch": 0.32, "learning_rate": 1.5861733733532878e-05, "loss": 0.9468, "step": 6331 }, { "epoch": 0.32, "learning_rate": 1.5860400561817467e-05, "loss": 1.0506, "step": 6332 }, { "epoch": 0.32, "learning_rate": 1.5859067231437945e-05, "loss": 0.978, "step": 6333 }, { "epoch": 0.32, "learning_rate": 1.5857733742430405e-05, "loss": 0.9239, "step": 6334 }, { "epoch": 0.32, "learning_rate": 1.5856400094830962e-05, "loss": 0.9847, "step": 6335 }, { "epoch": 0.32, "learning_rate": 1.5855066288675717e-05, "loss": 0.9899, "step": 6336 }, { "epoch": 0.32, "learning_rate": 1.5853732324000776e-05, "loss": 1.0211, "step": 6337 }, { "epoch": 0.32, "learning_rate": 1.5852398200842264e-05, "loss": 0.8768, "step": 6338 }, { "epoch": 0.32, "learning_rate": 1.58510639192363e-05, "loss": 0.9181, "step": 6339 }, { "epoch": 0.32, "learning_rate": 1.5849729479219e-05, "loss": 1.0025, "step": 6340 }, { "epoch": 0.32, "learning_rate": 1.5848394880826505e-05, "loss": 0.975, "step": 6341 }, { "epoch": 0.32, "learning_rate": 1.5847060124094938e-05, "loss": 0.9377, "step": 6342 }, { "epoch": 0.32, "learning_rate": 1.5845725209060435e-05, "loss": 0.9812, "step": 6343 }, { "epoch": 0.32, "learning_rate": 1.5844390135759144e-05, "loss": 0.9968, "step": 6344 }, { "epoch": 0.32, "learning_rate": 1.584305490422721e-05, "loss": 0.942, "step": 6345 }, { "epoch": 0.32, "learning_rate": 1.584171951450078e-05, "loss": 0.9979, "step": 6346 }, { "epoch": 0.32, "learning_rate": 1.5840383966616007e-05, "loss": 0.8821, "step": 6347 }, { "epoch": 0.32, "learning_rate": 1.5839048260609053e-05, "loss": 0.912, "step": 6348 }, { "epoch": 0.32, "learning_rate": 1.5837712396516077e-05, "loss": 0.982, "step": 6349 }, { "epoch": 0.32, "learning_rate": 1.5836376374373252e-05, "loss": 1.0028, "step": 6350 }, { "epoch": 0.32, "learning_rate": 1.5835040194216746e-05, "loss": 1.0115, "step": 6351 }, { "epoch": 0.32, "learning_rate": 1.5833703856082734e-05, "loss": 0.9825, "step": 6352 }, { "epoch": 0.32, "learning_rate": 1.5832367360007398e-05, "loss": 0.9268, "step": 6353 }, { "epoch": 0.32, "learning_rate": 1.583103070602692e-05, "loss": 1.0024, "step": 6354 }, { "epoch": 0.32, "learning_rate": 1.5829693894177486e-05, "loss": 1.0251, "step": 6355 }, { "epoch": 0.32, "learning_rate": 1.5828356924495295e-05, "loss": 1.0065, "step": 6356 }, { "epoch": 0.32, "learning_rate": 1.5827019797016544e-05, "loss": 1.0123, "step": 6357 }, { "epoch": 0.32, "learning_rate": 1.582568251177743e-05, "loss": 0.9529, "step": 6358 }, { "epoch": 0.32, "learning_rate": 1.5824345068814156e-05, "loss": 0.9052, "step": 6359 }, { "epoch": 0.32, "learning_rate": 1.582300746816294e-05, "loss": 0.9625, "step": 6360 }, { "epoch": 0.32, "learning_rate": 1.582166970985999e-05, "loss": 0.9846, "step": 6361 }, { "epoch": 0.32, "learning_rate": 1.582033179394153e-05, "loss": 1.1043, "step": 6362 }, { "epoch": 0.32, "learning_rate": 1.5818993720443775e-05, "loss": 1.0734, "step": 6363 }, { "epoch": 0.32, "learning_rate": 1.5817655489402955e-05, "loss": 0.9406, "step": 6364 }, { "epoch": 0.32, "learning_rate": 1.5816317100855305e-05, "loss": 1.0371, "step": 6365 }, { "epoch": 0.32, "learning_rate": 1.5814978554837057e-05, "loss": 1.0044, "step": 6366 }, { "epoch": 0.32, "learning_rate": 1.581363985138445e-05, "loss": 0.9832, "step": 6367 }, { "epoch": 0.32, "learning_rate": 1.581230099053373e-05, "loss": 0.8801, "step": 6368 }, { "epoch": 0.32, "learning_rate": 1.5810961972321146e-05, "loss": 0.9878, "step": 6369 }, { "epoch": 0.32, "learning_rate": 1.580962279678295e-05, "loss": 0.9184, "step": 6370 }, { "epoch": 0.32, "learning_rate": 1.5808283463955394e-05, "loss": 0.9756, "step": 6371 }, { "epoch": 0.32, "learning_rate": 1.5806943973874742e-05, "loss": 0.9137, "step": 6372 }, { "epoch": 0.32, "learning_rate": 1.5805604326577264e-05, "loss": 0.8854, "step": 6373 }, { "epoch": 0.32, "learning_rate": 1.5804264522099222e-05, "loss": 0.976, "step": 6374 }, { "epoch": 0.32, "learning_rate": 1.5802924560476897e-05, "loss": 0.9115, "step": 6375 }, { "epoch": 0.32, "learning_rate": 1.580158444174656e-05, "loss": 0.9058, "step": 6376 }, { "epoch": 0.32, "learning_rate": 1.5800244165944497e-05, "loss": 0.951, "step": 6377 }, { "epoch": 0.32, "learning_rate": 1.5798903733106996e-05, "loss": 0.9501, "step": 6378 }, { "epoch": 0.32, "learning_rate": 1.5797563143270346e-05, "loss": 0.8652, "step": 6379 }, { "epoch": 0.32, "learning_rate": 1.579622239647084e-05, "loss": 0.9573, "step": 6380 }, { "epoch": 0.32, "learning_rate": 1.579488149274478e-05, "loss": 1.0661, "step": 6381 }, { "epoch": 0.32, "learning_rate": 1.579354043212847e-05, "loss": 0.9425, "step": 6382 }, { "epoch": 0.32, "learning_rate": 1.579219921465822e-05, "loss": 0.9224, "step": 6383 }, { "epoch": 0.32, "learning_rate": 1.579085784037033e-05, "loss": 1.0215, "step": 6384 }, { "epoch": 0.32, "learning_rate": 1.578951630930113e-05, "loss": 0.9098, "step": 6385 }, { "epoch": 0.32, "learning_rate": 1.5788174621486936e-05, "loss": 0.9461, "step": 6386 }, { "epoch": 0.32, "learning_rate": 1.578683277696407e-05, "loss": 0.9287, "step": 6387 }, { "epoch": 0.32, "learning_rate": 1.5785490775768866e-05, "loss": 0.9282, "step": 6388 }, { "epoch": 0.32, "learning_rate": 1.5784148617937655e-05, "loss": 0.8907, "step": 6389 }, { "epoch": 0.32, "learning_rate": 1.578280630350677e-05, "loss": 0.9181, "step": 6390 }, { "epoch": 0.32, "learning_rate": 1.578146383251256e-05, "loss": 0.9429, "step": 6391 }, { "epoch": 0.32, "learning_rate": 1.5780121204991368e-05, "loss": 0.9616, "step": 6392 }, { "epoch": 0.32, "learning_rate": 1.577877842097954e-05, "loss": 0.895, "step": 6393 }, { "epoch": 0.32, "learning_rate": 1.5777435480513437e-05, "loss": 0.9267, "step": 6394 }, { "epoch": 0.32, "learning_rate": 1.577609238362942e-05, "loss": 1.0756, "step": 6395 }, { "epoch": 0.32, "learning_rate": 1.5774749130363836e-05, "loss": 0.9646, "step": 6396 }, { "epoch": 0.32, "learning_rate": 1.577340572075307e-05, "loss": 0.9268, "step": 6397 }, { "epoch": 0.33, "learning_rate": 1.5772062154833486e-05, "loss": 0.9682, "step": 6398 }, { "epoch": 0.33, "learning_rate": 1.5770718432641462e-05, "loss": 1.0463, "step": 6399 }, { "epoch": 0.33, "learning_rate": 1.5769374554213374e-05, "loss": 0.9581, "step": 6400 }, { "epoch": 0.33, "learning_rate": 1.5768030519585606e-05, "loss": 0.9673, "step": 6401 }, { "epoch": 0.33, "learning_rate": 1.576668632879455e-05, "loss": 0.9916, "step": 6402 }, { "epoch": 0.33, "learning_rate": 1.57653419818766e-05, "loss": 0.9621, "step": 6403 }, { "epoch": 0.33, "learning_rate": 1.5763997478868147e-05, "loss": 0.9734, "step": 6404 }, { "epoch": 0.33, "learning_rate": 1.5762652819805594e-05, "loss": 1.0068, "step": 6405 }, { "epoch": 0.33, "learning_rate": 1.5761308004725346e-05, "loss": 0.8449, "step": 6406 }, { "epoch": 0.33, "learning_rate": 1.5759963033663817e-05, "loss": 0.9834, "step": 6407 }, { "epoch": 0.33, "learning_rate": 1.575861790665741e-05, "loss": 0.8024, "step": 6408 }, { "epoch": 0.33, "learning_rate": 1.575727262374256e-05, "loss": 1.0438, "step": 6409 }, { "epoch": 0.33, "learning_rate": 1.575592718495567e-05, "loss": 1.0352, "step": 6410 }, { "epoch": 0.33, "learning_rate": 1.575458159033318e-05, "loss": 0.9778, "step": 6411 }, { "epoch": 0.33, "learning_rate": 1.5753235839911517e-05, "loss": 0.9214, "step": 6412 }, { "epoch": 0.33, "learning_rate": 1.5751889933727112e-05, "loss": 0.9261, "step": 6413 }, { "epoch": 0.33, "learning_rate": 1.5750543871816405e-05, "loss": 0.9509, "step": 6414 }, { "epoch": 0.33, "learning_rate": 1.5749197654215844e-05, "loss": 0.9096, "step": 6415 }, { "epoch": 0.33, "learning_rate": 1.5747851280961874e-05, "loss": 1.0418, "step": 6416 }, { "epoch": 0.33, "learning_rate": 1.5746504752090942e-05, "loss": 1.0225, "step": 6417 }, { "epoch": 0.33, "learning_rate": 1.574515806763951e-05, "loss": 0.9781, "step": 6418 }, { "epoch": 0.33, "learning_rate": 1.5743811227644035e-05, "loss": 0.9216, "step": 6419 }, { "epoch": 0.33, "learning_rate": 1.5742464232140982e-05, "loss": 0.8732, "step": 6420 }, { "epoch": 0.33, "learning_rate": 1.5741117081166818e-05, "loss": 0.9485, "step": 6421 }, { "epoch": 0.33, "learning_rate": 1.5739769774758018e-05, "loss": 1.0349, "step": 6422 }, { "epoch": 0.33, "learning_rate": 1.573842231295106e-05, "loss": 0.9083, "step": 6423 }, { "epoch": 0.33, "learning_rate": 1.573707469578242e-05, "loss": 0.9529, "step": 6424 }, { "epoch": 0.33, "learning_rate": 1.573572692328859e-05, "loss": 0.9206, "step": 6425 }, { "epoch": 0.33, "learning_rate": 1.5734378995506052e-05, "loss": 0.9842, "step": 6426 }, { "epoch": 0.33, "learning_rate": 1.5733030912471306e-05, "loss": 0.8783, "step": 6427 }, { "epoch": 0.33, "learning_rate": 1.5731682674220846e-05, "loss": 0.985, "step": 6428 }, { "epoch": 0.33, "learning_rate": 1.573033428079118e-05, "loss": 0.9521, "step": 6429 }, { "epoch": 0.33, "learning_rate": 1.5728985732218804e-05, "loss": 0.9998, "step": 6430 }, { "epoch": 0.33, "learning_rate": 1.5727637028540236e-05, "loss": 0.9928, "step": 6431 }, { "epoch": 0.33, "learning_rate": 1.572628816979199e-05, "loss": 0.9911, "step": 6432 }, { "epoch": 0.33, "learning_rate": 1.5724939156010583e-05, "loss": 1.0118, "step": 6433 }, { "epoch": 0.33, "learning_rate": 1.5723589987232543e-05, "loss": 0.926, "step": 6434 }, { "epoch": 0.33, "learning_rate": 1.5722240663494392e-05, "loss": 0.9292, "step": 6435 }, { "epoch": 0.33, "learning_rate": 1.572089118483266e-05, "loss": 0.9753, "step": 6436 }, { "epoch": 0.33, "learning_rate": 1.571954155128389e-05, "loss": 0.9595, "step": 6437 }, { "epoch": 0.33, "learning_rate": 1.5718191762884615e-05, "loss": 0.9688, "step": 6438 }, { "epoch": 0.33, "learning_rate": 1.571684181967138e-05, "loss": 0.9718, "step": 6439 }, { "epoch": 0.33, "learning_rate": 1.571549172168074e-05, "loss": 0.9464, "step": 6440 }, { "epoch": 0.33, "learning_rate": 1.571414146894924e-05, "loss": 0.8864, "step": 6441 }, { "epoch": 0.33, "learning_rate": 1.5712791061513435e-05, "loss": 0.8763, "step": 6442 }, { "epoch": 0.33, "learning_rate": 1.571144049940989e-05, "loss": 0.929, "step": 6443 }, { "epoch": 0.33, "learning_rate": 1.5710089782675178e-05, "loss": 0.9054, "step": 6444 }, { "epoch": 0.33, "learning_rate": 1.5708738911345857e-05, "loss": 0.9239, "step": 6445 }, { "epoch": 0.33, "learning_rate": 1.5707387885458498e-05, "loss": 0.9087, "step": 6446 }, { "epoch": 0.33, "learning_rate": 1.570603670504969e-05, "loss": 0.942, "step": 6447 }, { "epoch": 0.33, "learning_rate": 1.5704685370156008e-05, "loss": 0.9358, "step": 6448 }, { "epoch": 0.33, "learning_rate": 1.5703333880814037e-05, "loss": 0.9797, "step": 6449 }, { "epoch": 0.33, "learning_rate": 1.570198223706037e-05, "loss": 0.9815, "step": 6450 }, { "epoch": 0.33, "learning_rate": 1.57006304389316e-05, "loss": 0.9708, "step": 6451 }, { "epoch": 0.33, "learning_rate": 1.5699278486464327e-05, "loss": 1.0875, "step": 6452 }, { "epoch": 0.33, "learning_rate": 1.569792637969515e-05, "loss": 0.9989, "step": 6453 }, { "epoch": 0.33, "learning_rate": 1.5696574118660683e-05, "loss": 0.9658, "step": 6454 }, { "epoch": 0.33, "learning_rate": 1.569522170339753e-05, "loss": 0.9133, "step": 6455 }, { "epoch": 0.33, "learning_rate": 1.5693869133942307e-05, "loss": 1.0086, "step": 6456 }, { "epoch": 0.33, "learning_rate": 1.569251641033164e-05, "loss": 0.8327, "step": 6457 }, { "epoch": 0.33, "learning_rate": 1.5691163532602142e-05, "loss": 0.989, "step": 6458 }, { "epoch": 0.33, "learning_rate": 1.568981050079045e-05, "loss": 0.8897, "step": 6459 }, { "epoch": 0.33, "learning_rate": 1.568845731493319e-05, "loss": 0.9452, "step": 6460 }, { "epoch": 0.33, "learning_rate": 1.5687103975067e-05, "loss": 1.0466, "step": 6461 }, { "epoch": 0.33, "learning_rate": 1.568575048122852e-05, "loss": 0.8672, "step": 6462 }, { "epoch": 0.33, "learning_rate": 1.5684396833454397e-05, "loss": 0.9722, "step": 6463 }, { "epoch": 0.33, "learning_rate": 1.5683043031781276e-05, "loss": 0.8968, "step": 6464 }, { "epoch": 0.33, "learning_rate": 1.5681689076245812e-05, "loss": 0.9424, "step": 6465 }, { "epoch": 0.33, "learning_rate": 1.5680334966884663e-05, "loss": 1.0549, "step": 6466 }, { "epoch": 0.33, "learning_rate": 1.5678980703734484e-05, "loss": 1.0624, "step": 6467 }, { "epoch": 0.33, "learning_rate": 1.5677626286831948e-05, "loss": 0.982, "step": 6468 }, { "epoch": 0.33, "learning_rate": 1.5676271716213718e-05, "loss": 0.9331, "step": 6469 }, { "epoch": 0.33, "learning_rate": 1.5674916991916468e-05, "loss": 0.9965, "step": 6470 }, { "epoch": 0.33, "learning_rate": 1.5673562113976887e-05, "loss": 0.8582, "step": 6471 }, { "epoch": 0.33, "learning_rate": 1.5672207082431642e-05, "loss": 0.9038, "step": 6472 }, { "epoch": 0.33, "learning_rate": 1.5670851897317423e-05, "loss": 0.9022, "step": 6473 }, { "epoch": 0.33, "learning_rate": 1.5669496558670923e-05, "loss": 0.9264, "step": 6474 }, { "epoch": 0.33, "learning_rate": 1.566814106652884e-05, "loss": 0.8827, "step": 6475 }, { "epoch": 0.33, "learning_rate": 1.5666785420927862e-05, "loss": 0.8831, "step": 6476 }, { "epoch": 0.33, "learning_rate": 1.5665429621904705e-05, "loss": 1.011, "step": 6477 }, { "epoch": 0.33, "learning_rate": 1.5664073669496065e-05, "loss": 0.8933, "step": 6478 }, { "epoch": 0.33, "learning_rate": 1.5662717563738653e-05, "loss": 0.8261, "step": 6479 }, { "epoch": 0.33, "learning_rate": 1.5661361304669195e-05, "loss": 0.9846, "step": 6480 }, { "epoch": 0.33, "learning_rate": 1.56600048923244e-05, "loss": 0.9384, "step": 6481 }, { "epoch": 0.33, "learning_rate": 1.565864832674099e-05, "loss": 0.9162, "step": 6482 }, { "epoch": 0.33, "learning_rate": 1.56572916079557e-05, "loss": 0.9133, "step": 6483 }, { "epoch": 0.33, "learning_rate": 1.5655934736005258e-05, "loss": 0.8996, "step": 6484 }, { "epoch": 0.33, "learning_rate": 1.56545777109264e-05, "loss": 0.9849, "step": 6485 }, { "epoch": 0.33, "learning_rate": 1.565322053275587e-05, "loss": 0.9839, "step": 6486 }, { "epoch": 0.33, "learning_rate": 1.5651863201530403e-05, "loss": 0.9288, "step": 6487 }, { "epoch": 0.33, "learning_rate": 1.5650505717286757e-05, "loss": 0.916, "step": 6488 }, { "epoch": 0.33, "learning_rate": 1.5649148080061676e-05, "loss": 0.9309, "step": 6489 }, { "epoch": 0.33, "learning_rate": 1.5647790289891926e-05, "loss": 0.9768, "step": 6490 }, { "epoch": 0.33, "learning_rate": 1.5646432346814263e-05, "loss": 0.8581, "step": 6491 }, { "epoch": 0.33, "learning_rate": 1.564507425086545e-05, "loss": 0.9821, "step": 6492 }, { "epoch": 0.33, "learning_rate": 1.5643716002082255e-05, "loss": 0.8907, "step": 6493 }, { "epoch": 0.33, "learning_rate": 1.5642357600501457e-05, "loss": 1.0063, "step": 6494 }, { "epoch": 0.33, "learning_rate": 1.564099904615983e-05, "loss": 0.9449, "step": 6495 }, { "epoch": 0.33, "learning_rate": 1.5639640339094155e-05, "loss": 1.0107, "step": 6496 }, { "epoch": 0.33, "learning_rate": 1.5638281479341214e-05, "loss": 0.856, "step": 6497 }, { "epoch": 0.33, "learning_rate": 1.56369224669378e-05, "loss": 0.8777, "step": 6498 }, { "epoch": 0.33, "learning_rate": 1.5635563301920715e-05, "loss": 0.8992, "step": 6499 }, { "epoch": 0.33, "learning_rate": 1.5634203984326744e-05, "loss": 1.0684, "step": 6500 }, { "epoch": 0.33, "learning_rate": 1.5632844514192697e-05, "loss": 0.9733, "step": 6501 }, { "epoch": 0.33, "learning_rate": 1.5631484891555378e-05, "loss": 1.0544, "step": 6502 }, { "epoch": 0.33, "learning_rate": 1.5630125116451594e-05, "loss": 1.044, "step": 6503 }, { "epoch": 0.33, "learning_rate": 1.5628765188918166e-05, "loss": 0.9761, "step": 6504 }, { "epoch": 0.33, "learning_rate": 1.5627405108991903e-05, "loss": 0.9368, "step": 6505 }, { "epoch": 0.33, "learning_rate": 1.562604487670964e-05, "loss": 0.9422, "step": 6506 }, { "epoch": 0.33, "learning_rate": 1.5624684492108195e-05, "loss": 0.9473, "step": 6507 }, { "epoch": 0.33, "learning_rate": 1.5623323955224404e-05, "loss": 1.0445, "step": 6508 }, { "epoch": 0.33, "learning_rate": 1.5621963266095095e-05, "loss": 1.0229, "step": 6509 }, { "epoch": 0.33, "learning_rate": 1.5620602424757116e-05, "loss": 0.9613, "step": 6510 }, { "epoch": 0.33, "learning_rate": 1.5619241431247303e-05, "loss": 0.9142, "step": 6511 }, { "epoch": 0.33, "learning_rate": 1.561788028560251e-05, "loss": 0.9673, "step": 6512 }, { "epoch": 0.33, "learning_rate": 1.5616518987859582e-05, "loss": 0.9891, "step": 6513 }, { "epoch": 0.33, "learning_rate": 1.5615157538055377e-05, "loss": 0.9996, "step": 6514 }, { "epoch": 0.33, "learning_rate": 1.5613795936226763e-05, "loss": 0.9754, "step": 6515 }, { "epoch": 0.33, "learning_rate": 1.561243418241059e-05, "loss": 0.9969, "step": 6516 }, { "epoch": 0.33, "learning_rate": 1.5611072276643732e-05, "loss": 0.9572, "step": 6517 }, { "epoch": 0.33, "learning_rate": 1.5609710218963067e-05, "loss": 0.9554, "step": 6518 }, { "epoch": 0.33, "learning_rate": 1.5608348009405463e-05, "loss": 1.0519, "step": 6519 }, { "epoch": 0.33, "learning_rate": 1.56069856480078e-05, "loss": 0.9596, "step": 6520 }, { "epoch": 0.33, "learning_rate": 1.5605623134806972e-05, "loss": 0.9871, "step": 6521 }, { "epoch": 0.33, "learning_rate": 1.5604260469839858e-05, "loss": 0.9454, "step": 6522 }, { "epoch": 0.33, "learning_rate": 1.5602897653143356e-05, "loss": 0.9446, "step": 6523 }, { "epoch": 0.33, "learning_rate": 1.560153468475436e-05, "loss": 0.949, "step": 6524 }, { "epoch": 0.33, "learning_rate": 1.5600171564709775e-05, "loss": 0.8834, "step": 6525 }, { "epoch": 0.33, "learning_rate": 1.55988082930465e-05, "loss": 0.9702, "step": 6526 }, { "epoch": 0.33, "learning_rate": 1.5597444869801448e-05, "loss": 0.9261, "step": 6527 }, { "epoch": 0.33, "learning_rate": 1.559608129501153e-05, "loss": 1.0708, "step": 6528 }, { "epoch": 0.33, "learning_rate": 1.5594717568713664e-05, "loss": 0.9566, "step": 6529 }, { "epoch": 0.33, "learning_rate": 1.5593353690944775e-05, "loss": 0.9831, "step": 6530 }, { "epoch": 0.33, "learning_rate": 1.5591989661741785e-05, "loss": 0.9704, "step": 6531 }, { "epoch": 0.33, "learning_rate": 1.5590625481141624e-05, "loss": 0.8735, "step": 6532 }, { "epoch": 0.33, "learning_rate": 1.5589261149181228e-05, "loss": 0.9474, "step": 6533 }, { "epoch": 0.33, "learning_rate": 1.558789666589753e-05, "loss": 0.999, "step": 6534 }, { "epoch": 0.33, "learning_rate": 1.558653203132747e-05, "loss": 0.9362, "step": 6535 }, { "epoch": 0.33, "learning_rate": 1.5585167245508008e-05, "loss": 0.9945, "step": 6536 }, { "epoch": 0.33, "learning_rate": 1.5583802308476084e-05, "loss": 0.9183, "step": 6537 }, { "epoch": 0.33, "learning_rate": 1.5582437220268648e-05, "loss": 0.9991, "step": 6538 }, { "epoch": 0.33, "learning_rate": 1.558107198092267e-05, "loss": 0.9856, "step": 6539 }, { "epoch": 0.33, "learning_rate": 1.55797065904751e-05, "loss": 1.0151, "step": 6540 }, { "epoch": 0.33, "learning_rate": 1.557834104896291e-05, "loss": 0.938, "step": 6541 }, { "epoch": 0.33, "learning_rate": 1.5576975356423077e-05, "loss": 0.9542, "step": 6542 }, { "epoch": 0.33, "learning_rate": 1.5575609512892563e-05, "loss": 1.0007, "step": 6543 }, { "epoch": 0.33, "learning_rate": 1.557424351840836e-05, "loss": 0.9281, "step": 6544 }, { "epoch": 0.33, "learning_rate": 1.5572877373007437e-05, "loss": 0.991, "step": 6545 }, { "epoch": 0.33, "learning_rate": 1.5571511076726794e-05, "loss": 0.9964, "step": 6546 }, { "epoch": 0.33, "learning_rate": 1.5570144629603414e-05, "loss": 0.9693, "step": 6547 }, { "epoch": 0.33, "learning_rate": 1.5568778031674295e-05, "loss": 1.0056, "step": 6548 }, { "epoch": 0.33, "learning_rate": 1.5567411282976434e-05, "loss": 0.9967, "step": 6549 }, { "epoch": 0.33, "learning_rate": 1.5566044383546835e-05, "loss": 0.9587, "step": 6550 }, { "epoch": 0.33, "learning_rate": 1.556467733342251e-05, "loss": 0.9505, "step": 6551 }, { "epoch": 0.33, "learning_rate": 1.5563310132640465e-05, "loss": 1.0213, "step": 6552 }, { "epoch": 0.33, "learning_rate": 1.5561942781237714e-05, "loss": 1.0539, "step": 6553 }, { "epoch": 0.33, "learning_rate": 1.5560575279251285e-05, "loss": 1.0704, "step": 6554 }, { "epoch": 0.33, "learning_rate": 1.5559207626718195e-05, "loss": 0.9476, "step": 6555 }, { "epoch": 0.33, "learning_rate": 1.555783982367547e-05, "loss": 1.0277, "step": 6556 }, { "epoch": 0.33, "learning_rate": 1.5556471870160145e-05, "loss": 0.9797, "step": 6557 }, { "epoch": 0.33, "learning_rate": 1.5555103766209258e-05, "loss": 0.9087, "step": 6558 }, { "epoch": 0.33, "learning_rate": 1.5553735511859844e-05, "loss": 0.9375, "step": 6559 }, { "epoch": 0.33, "learning_rate": 1.5552367107148953e-05, "loss": 0.9146, "step": 6560 }, { "epoch": 0.33, "learning_rate": 1.555099855211363e-05, "loss": 0.9973, "step": 6561 }, { "epoch": 0.33, "learning_rate": 1.5549629846790923e-05, "loss": 0.9471, "step": 6562 }, { "epoch": 0.33, "learning_rate": 1.5548260991217893e-05, "loss": 0.9562, "step": 6563 }, { "epoch": 0.33, "learning_rate": 1.5546891985431606e-05, "loss": 1.0436, "step": 6564 }, { "epoch": 0.33, "learning_rate": 1.554552282946911e-05, "loss": 1.0559, "step": 6565 }, { "epoch": 0.33, "learning_rate": 1.554415352336749e-05, "loss": 1.037, "step": 6566 }, { "epoch": 0.33, "learning_rate": 1.5542784067163813e-05, "loss": 0.9424, "step": 6567 }, { "epoch": 0.33, "learning_rate": 1.554141446089515e-05, "loss": 0.9758, "step": 6568 }, { "epoch": 0.33, "learning_rate": 1.5540044704598588e-05, "loss": 0.8896, "step": 6569 }, { "epoch": 0.33, "learning_rate": 1.5538674798311212e-05, "loss": 0.9833, "step": 6570 }, { "epoch": 0.33, "learning_rate": 1.5537304742070104e-05, "loss": 0.9957, "step": 6571 }, { "epoch": 0.33, "learning_rate": 1.5535934535912363e-05, "loss": 0.9775, "step": 6572 }, { "epoch": 0.33, "learning_rate": 1.5534564179875083e-05, "loss": 0.8961, "step": 6573 }, { "epoch": 0.33, "learning_rate": 1.5533193673995368e-05, "loss": 1.0311, "step": 6574 }, { "epoch": 0.33, "learning_rate": 1.5531823018310324e-05, "loss": 0.9848, "step": 6575 }, { "epoch": 0.33, "learning_rate": 1.5530452212857054e-05, "loss": 1.0144, "step": 6576 }, { "epoch": 0.33, "learning_rate": 1.5529081257672672e-05, "loss": 0.9608, "step": 6577 }, { "epoch": 0.33, "learning_rate": 1.5527710152794303e-05, "loss": 0.9301, "step": 6578 }, { "epoch": 0.33, "learning_rate": 1.552633889825906e-05, "loss": 1.0346, "step": 6579 }, { "epoch": 0.33, "learning_rate": 1.5524967494104067e-05, "loss": 0.9525, "step": 6580 }, { "epoch": 0.33, "learning_rate": 1.5523595940366462e-05, "loss": 1.0037, "step": 6581 }, { "epoch": 0.33, "learning_rate": 1.552222423708337e-05, "loss": 0.9695, "step": 6582 }, { "epoch": 0.33, "learning_rate": 1.5520852384291934e-05, "loss": 0.8164, "step": 6583 }, { "epoch": 0.33, "learning_rate": 1.5519480382029296e-05, "loss": 0.9266, "step": 6584 }, { "epoch": 0.33, "learning_rate": 1.5518108230332594e-05, "loss": 0.9475, "step": 6585 }, { "epoch": 0.33, "learning_rate": 1.5516735929238985e-05, "loss": 0.8692, "step": 6586 }, { "epoch": 0.33, "learning_rate": 1.551536347878562e-05, "loss": 0.9495, "step": 6587 }, { "epoch": 0.33, "learning_rate": 1.5513990879009658e-05, "loss": 1.031, "step": 6588 }, { "epoch": 0.33, "learning_rate": 1.5512618129948255e-05, "loss": 0.9049, "step": 6589 }, { "epoch": 0.33, "learning_rate": 1.5511245231638586e-05, "loss": 0.9916, "step": 6590 }, { "epoch": 0.33, "learning_rate": 1.5509872184117812e-05, "loss": 0.9936, "step": 6591 }, { "epoch": 0.33, "learning_rate": 1.5508498987423115e-05, "loss": 0.9189, "step": 6592 }, { "epoch": 0.33, "learning_rate": 1.5507125641591666e-05, "loss": 0.9833, "step": 6593 }, { "epoch": 0.33, "learning_rate": 1.550575214666065e-05, "loss": 0.8931, "step": 6594 }, { "epoch": 0.34, "learning_rate": 1.550437850266725e-05, "loss": 0.9151, "step": 6595 }, { "epoch": 0.34, "learning_rate": 1.550300470964866e-05, "loss": 0.9659, "step": 6596 }, { "epoch": 0.34, "learning_rate": 1.5501630767642073e-05, "loss": 1.0219, "step": 6597 }, { "epoch": 0.34, "learning_rate": 1.5500256676684685e-05, "loss": 0.9451, "step": 6598 }, { "epoch": 0.34, "learning_rate": 1.54988824368137e-05, "loss": 1.0098, "step": 6599 }, { "epoch": 0.34, "learning_rate": 1.549750804806632e-05, "loss": 0.9722, "step": 6600 }, { "epoch": 0.34, "learning_rate": 1.5496133510479767e-05, "loss": 0.9252, "step": 6601 }, { "epoch": 0.34, "learning_rate": 1.549475882409124e-05, "loss": 1.0304, "step": 6602 }, { "epoch": 0.34, "learning_rate": 1.5493383988937966e-05, "loss": 1.0641, "step": 6603 }, { "epoch": 0.34, "learning_rate": 1.5492009005057164e-05, "loss": 0.9041, "step": 6604 }, { "epoch": 0.34, "learning_rate": 1.549063387248606e-05, "loss": 0.8586, "step": 6605 }, { "epoch": 0.34, "learning_rate": 1.5489258591261885e-05, "loss": 1.0569, "step": 6606 }, { "epoch": 0.34, "learning_rate": 1.5487883161421876e-05, "loss": 0.8321, "step": 6607 }, { "epoch": 0.34, "learning_rate": 1.548650758300327e-05, "loss": 0.9791, "step": 6608 }, { "epoch": 0.34, "learning_rate": 1.5485131856043304e-05, "loss": 0.9863, "step": 6609 }, { "epoch": 0.34, "learning_rate": 1.5483755980579234e-05, "loss": 0.8486, "step": 6610 }, { "epoch": 0.34, "learning_rate": 1.5482379956648303e-05, "loss": 0.8969, "step": 6611 }, { "epoch": 0.34, "learning_rate": 1.5481003784287767e-05, "loss": 1.0247, "step": 6612 }, { "epoch": 0.34, "learning_rate": 1.5479627463534885e-05, "loss": 0.9702, "step": 6613 }, { "epoch": 0.34, "learning_rate": 1.547825099442692e-05, "loss": 1.0403, "step": 6614 }, { "epoch": 0.34, "learning_rate": 1.5476874377001137e-05, "loss": 0.9274, "step": 6615 }, { "epoch": 0.34, "learning_rate": 1.5475497611294804e-05, "loss": 0.9769, "step": 6616 }, { "epoch": 0.34, "learning_rate": 1.5474120697345206e-05, "loss": 0.9959, "step": 6617 }, { "epoch": 0.34, "learning_rate": 1.547274363518961e-05, "loss": 0.9471, "step": 6618 }, { "epoch": 0.34, "learning_rate": 1.5471366424865302e-05, "loss": 0.9728, "step": 6619 }, { "epoch": 0.34, "learning_rate": 1.546998906640957e-05, "loss": 0.9237, "step": 6620 }, { "epoch": 0.34, "learning_rate": 1.5468611559859704e-05, "loss": 1.0118, "step": 6621 }, { "epoch": 0.34, "learning_rate": 1.5467233905252995e-05, "loss": 0.9968, "step": 6622 }, { "epoch": 0.34, "learning_rate": 1.546585610262675e-05, "loss": 1.0284, "step": 6623 }, { "epoch": 0.34, "learning_rate": 1.5464478152018264e-05, "loss": 0.9299, "step": 6624 }, { "epoch": 0.34, "learning_rate": 1.5463100053464843e-05, "loss": 0.966, "step": 6625 }, { "epoch": 0.34, "learning_rate": 1.5461721807003805e-05, "loss": 0.9192, "step": 6626 }, { "epoch": 0.34, "learning_rate": 1.5460343412672452e-05, "loss": 0.9862, "step": 6627 }, { "epoch": 0.34, "learning_rate": 1.545896487050812e-05, "loss": 1.0209, "step": 6628 }, { "epoch": 0.34, "learning_rate": 1.545758618054812e-05, "loss": 0.9401, "step": 6629 }, { "epoch": 0.34, "learning_rate": 1.5456207342829777e-05, "loss": 0.9944, "step": 6630 }, { "epoch": 0.34, "learning_rate": 1.5454828357390427e-05, "loss": 0.8569, "step": 6631 }, { "epoch": 0.34, "learning_rate": 1.54534492242674e-05, "loss": 1.025, "step": 6632 }, { "epoch": 0.34, "learning_rate": 1.545206994349804e-05, "loss": 0.9581, "step": 6633 }, { "epoch": 0.34, "learning_rate": 1.545069051511969e-05, "loss": 0.8641, "step": 6634 }, { "epoch": 0.34, "learning_rate": 1.544931093916969e-05, "loss": 0.9608, "step": 6635 }, { "epoch": 0.34, "learning_rate": 1.5447931215685393e-05, "loss": 1.001, "step": 6636 }, { "epoch": 0.34, "learning_rate": 1.5446551344704152e-05, "loss": 0.8775, "step": 6637 }, { "epoch": 0.34, "learning_rate": 1.5445171326263332e-05, "loss": 0.9104, "step": 6638 }, { "epoch": 0.34, "learning_rate": 1.544379116040029e-05, "loss": 0.9542, "step": 6639 }, { "epoch": 0.34, "learning_rate": 1.5442410847152396e-05, "loss": 0.9372, "step": 6640 }, { "epoch": 0.34, "learning_rate": 1.544103038655702e-05, "loss": 0.9402, "step": 6641 }, { "epoch": 0.34, "learning_rate": 1.5439649778651528e-05, "loss": 0.9416, "step": 6642 }, { "epoch": 0.34, "learning_rate": 1.5438269023473312e-05, "loss": 0.9581, "step": 6643 }, { "epoch": 0.34, "learning_rate": 1.5436888121059746e-05, "loss": 1.0604, "step": 6644 }, { "epoch": 0.34, "learning_rate": 1.543550707144822e-05, "loss": 1.0826, "step": 6645 }, { "epoch": 0.34, "learning_rate": 1.5434125874676122e-05, "loss": 1.0271, "step": 6646 }, { "epoch": 0.34, "learning_rate": 1.543274453078085e-05, "loss": 0.9111, "step": 6647 }, { "epoch": 0.34, "learning_rate": 1.5431363039799793e-05, "loss": 0.9388, "step": 6648 }, { "epoch": 0.34, "learning_rate": 1.5429981401770364e-05, "loss": 1.0351, "step": 6649 }, { "epoch": 0.34, "learning_rate": 1.5428599616729965e-05, "loss": 0.9595, "step": 6650 }, { "epoch": 0.34, "learning_rate": 1.542721768471601e-05, "loss": 0.9716, "step": 6651 }, { "epoch": 0.34, "learning_rate": 1.5425835605765905e-05, "loss": 1.0, "step": 6652 }, { "epoch": 0.34, "learning_rate": 1.5424453379917078e-05, "loss": 0.9668, "step": 6653 }, { "epoch": 0.34, "learning_rate": 1.5423071007206948e-05, "loss": 0.8718, "step": 6654 }, { "epoch": 0.34, "learning_rate": 1.5421688487672937e-05, "loss": 0.9152, "step": 6655 }, { "epoch": 0.34, "learning_rate": 1.5420305821352476e-05, "loss": 0.9728, "step": 6656 }, { "epoch": 0.34, "learning_rate": 1.5418923008283005e-05, "loss": 0.9513, "step": 6657 }, { "epoch": 0.34, "learning_rate": 1.541754004850196e-05, "loss": 0.9707, "step": 6658 }, { "epoch": 0.34, "learning_rate": 1.541615694204678e-05, "loss": 0.999, "step": 6659 }, { "epoch": 0.34, "learning_rate": 1.5414773688954915e-05, "loss": 0.8892, "step": 6660 }, { "epoch": 0.34, "learning_rate": 1.541339028926381e-05, "loss": 0.9999, "step": 6661 }, { "epoch": 0.34, "learning_rate": 1.5412006743010922e-05, "loss": 0.9513, "step": 6662 }, { "epoch": 0.34, "learning_rate": 1.5410623050233713e-05, "loss": 0.9935, "step": 6663 }, { "epoch": 0.34, "learning_rate": 1.540923921096964e-05, "loss": 0.9583, "step": 6664 }, { "epoch": 0.34, "learning_rate": 1.5407855225256173e-05, "loss": 1.0091, "step": 6665 }, { "epoch": 0.34, "learning_rate": 1.5406471093130774e-05, "loss": 0.8443, "step": 6666 }, { "epoch": 0.34, "learning_rate": 1.5405086814630926e-05, "loss": 0.9668, "step": 6667 }, { "epoch": 0.34, "learning_rate": 1.5403702389794098e-05, "loss": 0.9154, "step": 6668 }, { "epoch": 0.34, "learning_rate": 1.5402317818657784e-05, "loss": 0.9344, "step": 6669 }, { "epoch": 0.34, "learning_rate": 1.5400933101259462e-05, "loss": 1.0193, "step": 6670 }, { "epoch": 0.34, "learning_rate": 1.5399548237636616e-05, "loss": 0.945, "step": 6671 }, { "epoch": 0.34, "learning_rate": 1.539816322782675e-05, "loss": 0.9696, "step": 6672 }, { "epoch": 0.34, "learning_rate": 1.539677807186736e-05, "loss": 1.0009, "step": 6673 }, { "epoch": 0.34, "learning_rate": 1.5395392769795943e-05, "loss": 0.9258, "step": 6674 }, { "epoch": 0.34, "learning_rate": 1.539400732165001e-05, "loss": 0.9639, "step": 6675 }, { "epoch": 0.34, "learning_rate": 1.539262172746706e-05, "loss": 1.0964, "step": 6676 }, { "epoch": 0.34, "learning_rate": 1.5391235987284617e-05, "loss": 0.8719, "step": 6677 }, { "epoch": 0.34, "learning_rate": 1.53898501011402e-05, "loss": 0.8559, "step": 6678 }, { "epoch": 0.34, "learning_rate": 1.538846406907132e-05, "loss": 0.8724, "step": 6679 }, { "epoch": 0.34, "learning_rate": 1.5387077891115514e-05, "loss": 0.9279, "step": 6680 }, { "epoch": 0.34, "learning_rate": 1.5385691567310304e-05, "loss": 0.9648, "step": 6681 }, { "epoch": 0.34, "learning_rate": 1.538430509769322e-05, "loss": 0.9741, "step": 6682 }, { "epoch": 0.34, "learning_rate": 1.538291848230181e-05, "loss": 0.9456, "step": 6683 }, { "epoch": 0.34, "learning_rate": 1.5381531721173602e-05, "loss": 0.8499, "step": 6684 }, { "epoch": 0.34, "learning_rate": 1.5380144814346155e-05, "loss": 0.9382, "step": 6685 }, { "epoch": 0.34, "learning_rate": 1.537875776185701e-05, "loss": 0.9613, "step": 6686 }, { "epoch": 0.34, "learning_rate": 1.5377370563743717e-05, "loss": 0.8621, "step": 6687 }, { "epoch": 0.34, "learning_rate": 1.5375983220043842e-05, "loss": 0.9321, "step": 6688 }, { "epoch": 0.34, "learning_rate": 1.5374595730794938e-05, "loss": 1.0211, "step": 6689 }, { "epoch": 0.34, "learning_rate": 1.5373208096034575e-05, "loss": 0.9321, "step": 6690 }, { "epoch": 0.34, "learning_rate": 1.5371820315800316e-05, "loss": 0.9282, "step": 6691 }, { "epoch": 0.34, "learning_rate": 1.537043239012974e-05, "loss": 1.0966, "step": 6692 }, { "epoch": 0.34, "learning_rate": 1.536904431906042e-05, "loss": 0.9577, "step": 6693 }, { "epoch": 0.34, "learning_rate": 1.536765610262994e-05, "loss": 0.9184, "step": 6694 }, { "epoch": 0.34, "learning_rate": 1.536626774087588e-05, "loss": 0.9587, "step": 6695 }, { "epoch": 0.34, "learning_rate": 1.5364879233835827e-05, "loss": 0.9415, "step": 6696 }, { "epoch": 0.34, "learning_rate": 1.536349058154738e-05, "loss": 1.0211, "step": 6697 }, { "epoch": 0.34, "learning_rate": 1.536210178404813e-05, "loss": 1.02, "step": 6698 }, { "epoch": 0.34, "learning_rate": 1.5360712841375678e-05, "loss": 1.0052, "step": 6699 }, { "epoch": 0.34, "learning_rate": 1.535932375356763e-05, "loss": 0.975, "step": 6700 }, { "epoch": 0.34, "learning_rate": 1.5357934520661595e-05, "loss": 0.9475, "step": 6701 }, { "epoch": 0.34, "learning_rate": 1.535654514269518e-05, "loss": 1.0706, "step": 6702 }, { "epoch": 0.34, "learning_rate": 1.535515561970601e-05, "loss": 0.9235, "step": 6703 }, { "epoch": 0.34, "learning_rate": 1.5353765951731694e-05, "loss": 1.0251, "step": 6704 }, { "epoch": 0.34, "learning_rate": 1.5352376138809862e-05, "loss": 1.0087, "step": 6705 }, { "epoch": 0.34, "learning_rate": 1.535098618097814e-05, "loss": 0.9586, "step": 6706 }, { "epoch": 0.34, "learning_rate": 1.534959607827416e-05, "loss": 1.0078, "step": 6707 }, { "epoch": 0.34, "learning_rate": 1.534820583073556e-05, "loss": 0.9507, "step": 6708 }, { "epoch": 0.34, "learning_rate": 1.5346815438399975e-05, "loss": 1.0362, "step": 6709 }, { "epoch": 0.34, "learning_rate": 1.534542490130505e-05, "loss": 0.9568, "step": 6710 }, { "epoch": 0.34, "learning_rate": 1.5344034219488436e-05, "loss": 0.9694, "step": 6711 }, { "epoch": 0.34, "learning_rate": 1.534264339298778e-05, "loss": 0.9334, "step": 6712 }, { "epoch": 0.34, "learning_rate": 1.5341252421840738e-05, "loss": 0.9648, "step": 6713 }, { "epoch": 0.34, "learning_rate": 1.5339861306084966e-05, "loss": 0.9771, "step": 6714 }, { "epoch": 0.34, "learning_rate": 1.5338470045758136e-05, "loss": 0.8987, "step": 6715 }, { "epoch": 0.34, "learning_rate": 1.5337078640897907e-05, "loss": 1.0792, "step": 6716 }, { "epoch": 0.34, "learning_rate": 1.5335687091541948e-05, "loss": 0.9642, "step": 6717 }, { "epoch": 0.34, "learning_rate": 1.5334295397727942e-05, "loss": 0.9467, "step": 6718 }, { "epoch": 0.34, "learning_rate": 1.5332903559493564e-05, "loss": 0.8888, "step": 6719 }, { "epoch": 0.34, "learning_rate": 1.5331511576876492e-05, "loss": 0.9814, "step": 6720 }, { "epoch": 0.34, "learning_rate": 1.5330119449914418e-05, "loss": 1.0107, "step": 6721 }, { "epoch": 0.34, "learning_rate": 1.5328727178645034e-05, "loss": 0.8261, "step": 6722 }, { "epoch": 0.34, "learning_rate": 1.5327334763106027e-05, "loss": 0.924, "step": 6723 }, { "epoch": 0.34, "learning_rate": 1.53259422033351e-05, "loss": 0.8596, "step": 6724 }, { "epoch": 0.34, "learning_rate": 1.5324549499369954e-05, "loss": 0.9774, "step": 6725 }, { "epoch": 0.34, "learning_rate": 1.5323156651248297e-05, "loss": 0.9696, "step": 6726 }, { "epoch": 0.34, "learning_rate": 1.5321763659007837e-05, "loss": 1.0417, "step": 6727 }, { "epoch": 0.34, "learning_rate": 1.5320370522686286e-05, "loss": 0.8451, "step": 6728 }, { "epoch": 0.34, "learning_rate": 1.5318977242321365e-05, "loss": 0.9949, "step": 6729 }, { "epoch": 0.34, "learning_rate": 1.5317583817950793e-05, "loss": 0.9632, "step": 6730 }, { "epoch": 0.34, "learning_rate": 1.53161902496123e-05, "loss": 0.9807, "step": 6731 }, { "epoch": 0.34, "learning_rate": 1.5314796537343607e-05, "loss": 0.9313, "step": 6732 }, { "epoch": 0.34, "learning_rate": 1.5313402681182453e-05, "loss": 0.9543, "step": 6733 }, { "epoch": 0.34, "learning_rate": 1.5312008681166576e-05, "loss": 0.9173, "step": 6734 }, { "epoch": 0.34, "learning_rate": 1.531061453733371e-05, "loss": 1.0845, "step": 6735 }, { "epoch": 0.34, "learning_rate": 1.5309220249721614e-05, "loss": 0.9088, "step": 6736 }, { "epoch": 0.34, "learning_rate": 1.5307825818368025e-05, "loss": 1.0297, "step": 6737 }, { "epoch": 0.34, "learning_rate": 1.53064312433107e-05, "loss": 1.0454, "step": 6738 }, { "epoch": 0.34, "learning_rate": 1.5305036524587394e-05, "loss": 1.0157, "step": 6739 }, { "epoch": 0.34, "learning_rate": 1.5303641662235867e-05, "loss": 0.9877, "step": 6740 }, { "epoch": 0.34, "learning_rate": 1.5302246656293887e-05, "loss": 0.9277, "step": 6741 }, { "epoch": 0.34, "learning_rate": 1.5300851506799217e-05, "loss": 0.9348, "step": 6742 }, { "epoch": 0.34, "learning_rate": 1.5299456213789633e-05, "loss": 0.9716, "step": 6743 }, { "epoch": 0.34, "learning_rate": 1.529806077730291e-05, "loss": 0.8712, "step": 6744 }, { "epoch": 0.34, "learning_rate": 1.529666519737683e-05, "loss": 0.9134, "step": 6745 }, { "epoch": 0.34, "learning_rate": 1.5295269474049176e-05, "loss": 0.8645, "step": 6746 }, { "epoch": 0.34, "learning_rate": 1.529387360735773e-05, "loss": 0.9886, "step": 6747 }, { "epoch": 0.34, "learning_rate": 1.529247759734029e-05, "loss": 0.9307, "step": 6748 }, { "epoch": 0.34, "learning_rate": 1.5291081444034652e-05, "loss": 0.8756, "step": 6749 }, { "epoch": 0.34, "learning_rate": 1.5289685147478613e-05, "loss": 1.0175, "step": 6750 }, { "epoch": 0.34, "learning_rate": 1.5288288707709974e-05, "loss": 0.9364, "step": 6751 }, { "epoch": 0.34, "learning_rate": 1.5286892124766546e-05, "loss": 1.0561, "step": 6752 }, { "epoch": 0.34, "learning_rate": 1.5285495398686137e-05, "loss": 0.9764, "step": 6753 }, { "epoch": 0.34, "learning_rate": 1.5284098529506563e-05, "loss": 0.9374, "step": 6754 }, { "epoch": 0.34, "learning_rate": 1.5282701517265644e-05, "loss": 0.9934, "step": 6755 }, { "epoch": 0.34, "learning_rate": 1.5281304362001207e-05, "loss": 0.9974, "step": 6756 }, { "epoch": 0.34, "learning_rate": 1.5279907063751066e-05, "loss": 0.9141, "step": 6757 }, { "epoch": 0.34, "learning_rate": 1.5278509622553062e-05, "loss": 1.0078, "step": 6758 }, { "epoch": 0.34, "learning_rate": 1.5277112038445026e-05, "loss": 0.9828, "step": 6759 }, { "epoch": 0.34, "learning_rate": 1.5275714311464794e-05, "loss": 0.9828, "step": 6760 }, { "epoch": 0.34, "learning_rate": 1.527431644165021e-05, "loss": 0.9659, "step": 6761 }, { "epoch": 0.34, "learning_rate": 1.5272918429039125e-05, "loss": 0.9532, "step": 6762 }, { "epoch": 0.34, "learning_rate": 1.527152027366938e-05, "loss": 1.0345, "step": 6763 }, { "epoch": 0.34, "learning_rate": 1.527012197557883e-05, "loss": 0.9564, "step": 6764 }, { "epoch": 0.34, "learning_rate": 1.5268723534805335e-05, "loss": 0.915, "step": 6765 }, { "epoch": 0.34, "learning_rate": 1.5267324951386753e-05, "loss": 0.9173, "step": 6766 }, { "epoch": 0.34, "learning_rate": 1.526592622536096e-05, "loss": 0.9804, "step": 6767 }, { "epoch": 0.34, "learning_rate": 1.5264527356765815e-05, "loss": 0.9609, "step": 6768 }, { "epoch": 0.34, "learning_rate": 1.526312834563919e-05, "loss": 0.9669, "step": 6769 }, { "epoch": 0.34, "learning_rate": 1.5261729192018963e-05, "loss": 1.0233, "step": 6770 }, { "epoch": 0.34, "learning_rate": 1.5260329895943018e-05, "loss": 0.9153, "step": 6771 }, { "epoch": 0.34, "learning_rate": 1.5258930457449236e-05, "loss": 1.0531, "step": 6772 }, { "epoch": 0.34, "learning_rate": 1.525753087657551e-05, "loss": 1.0369, "step": 6773 }, { "epoch": 0.34, "learning_rate": 1.525613115335973e-05, "loss": 0.8611, "step": 6774 }, { "epoch": 0.34, "learning_rate": 1.5254731287839788e-05, "loss": 0.8994, "step": 6775 }, { "epoch": 0.34, "learning_rate": 1.5253331280053587e-05, "loss": 0.9646, "step": 6776 }, { "epoch": 0.34, "learning_rate": 1.5251931130039033e-05, "loss": 0.985, "step": 6777 }, { "epoch": 0.34, "learning_rate": 1.5250530837834028e-05, "loss": 1.0396, "step": 6778 }, { "epoch": 0.34, "learning_rate": 1.524913040347649e-05, "loss": 0.8862, "step": 6779 }, { "epoch": 0.34, "learning_rate": 1.5247729827004332e-05, "loss": 0.9288, "step": 6780 }, { "epoch": 0.34, "learning_rate": 1.5246329108455467e-05, "loss": 0.9453, "step": 6781 }, { "epoch": 0.34, "learning_rate": 1.5244928247867826e-05, "loss": 1.0164, "step": 6782 }, { "epoch": 0.34, "learning_rate": 1.5243527245279333e-05, "loss": 0.9857, "step": 6783 }, { "epoch": 0.34, "learning_rate": 1.5242126100727917e-05, "loss": 0.8892, "step": 6784 }, { "epoch": 0.34, "learning_rate": 1.5240724814251517e-05, "loss": 0.9097, "step": 6785 }, { "epoch": 0.34, "learning_rate": 1.5239323385888068e-05, "loss": 0.9886, "step": 6786 }, { "epoch": 0.34, "learning_rate": 1.5237921815675507e-05, "loss": 0.9448, "step": 6787 }, { "epoch": 0.34, "learning_rate": 1.5236520103651788e-05, "loss": 0.9888, "step": 6788 }, { "epoch": 0.34, "learning_rate": 1.5235118249854862e-05, "loss": 0.9764, "step": 6789 }, { "epoch": 0.34, "learning_rate": 1.5233716254322677e-05, "loss": 0.8922, "step": 6790 }, { "epoch": 0.35, "learning_rate": 1.5232314117093192e-05, "loss": 0.8829, "step": 6791 }, { "epoch": 0.35, "learning_rate": 1.523091183820437e-05, "loss": 0.8537, "step": 6792 }, { "epoch": 0.35, "learning_rate": 1.5229509417694174e-05, "loss": 1.0324, "step": 6793 }, { "epoch": 0.35, "learning_rate": 1.5228106855600576e-05, "loss": 0.9576, "step": 6794 }, { "epoch": 0.35, "learning_rate": 1.5226704151961546e-05, "loss": 0.9486, "step": 6795 }, { "epoch": 0.35, "learning_rate": 1.5225301306815063e-05, "loss": 0.9277, "step": 6796 }, { "epoch": 0.35, "learning_rate": 1.5223898320199106e-05, "loss": 0.9933, "step": 6797 }, { "epoch": 0.35, "learning_rate": 1.522249519215166e-05, "loss": 1.0112, "step": 6798 }, { "epoch": 0.35, "learning_rate": 1.522109192271071e-05, "loss": 1.0021, "step": 6799 }, { "epoch": 0.35, "learning_rate": 1.5219688511914255e-05, "loss": 0.9023, "step": 6800 }, { "epoch": 0.35, "learning_rate": 1.5218284959800284e-05, "loss": 0.8689, "step": 6801 }, { "epoch": 0.35, "learning_rate": 1.5216881266406802e-05, "loss": 0.9626, "step": 6802 }, { "epoch": 0.35, "learning_rate": 1.5215477431771808e-05, "loss": 0.9179, "step": 6803 }, { "epoch": 0.35, "learning_rate": 1.5214073455933314e-05, "loss": 1.0579, "step": 6804 }, { "epoch": 0.35, "learning_rate": 1.5212669338929327e-05, "loss": 0.9641, "step": 6805 }, { "epoch": 0.35, "learning_rate": 1.5211265080797862e-05, "loss": 0.9669, "step": 6806 }, { "epoch": 0.35, "learning_rate": 1.520986068157694e-05, "loss": 0.9505, "step": 6807 }, { "epoch": 0.35, "learning_rate": 1.5208456141304582e-05, "loss": 0.9323, "step": 6808 }, { "epoch": 0.35, "learning_rate": 1.5207051460018818e-05, "loss": 0.9545, "step": 6809 }, { "epoch": 0.35, "learning_rate": 1.5205646637757674e-05, "loss": 0.9427, "step": 6810 }, { "epoch": 0.35, "learning_rate": 1.5204241674559189e-05, "loss": 0.9517, "step": 6811 }, { "epoch": 0.35, "learning_rate": 1.520283657046139e-05, "loss": 1.0032, "step": 6812 }, { "epoch": 0.35, "learning_rate": 1.5201431325502332e-05, "loss": 1.0078, "step": 6813 }, { "epoch": 0.35, "learning_rate": 1.5200025939720053e-05, "loss": 0.8634, "step": 6814 }, { "epoch": 0.35, "learning_rate": 1.5198620413152606e-05, "loss": 0.8719, "step": 6815 }, { "epoch": 0.35, "learning_rate": 1.519721474583804e-05, "loss": 0.9662, "step": 6816 }, { "epoch": 0.35, "learning_rate": 1.5195808937814416e-05, "loss": 1.0681, "step": 6817 }, { "epoch": 0.35, "learning_rate": 1.5194402989119792e-05, "loss": 0.9398, "step": 6818 }, { "epoch": 0.35, "learning_rate": 1.5192996899792232e-05, "loss": 0.9715, "step": 6819 }, { "epoch": 0.35, "learning_rate": 1.5191590669869807e-05, "loss": 1.0731, "step": 6820 }, { "epoch": 0.35, "learning_rate": 1.5190184299390589e-05, "loss": 0.9708, "step": 6821 }, { "epoch": 0.35, "learning_rate": 1.518877778839265e-05, "loss": 0.9669, "step": 6822 }, { "epoch": 0.35, "learning_rate": 1.5187371136914076e-05, "loss": 0.9177, "step": 6823 }, { "epoch": 0.35, "learning_rate": 1.5185964344992946e-05, "loss": 0.9675, "step": 6824 }, { "epoch": 0.35, "learning_rate": 1.518455741266735e-05, "loss": 0.8848, "step": 6825 }, { "epoch": 0.35, "learning_rate": 1.5183150339975376e-05, "loss": 0.9097, "step": 6826 }, { "epoch": 0.35, "learning_rate": 1.5181743126955124e-05, "loss": 1.0549, "step": 6827 }, { "epoch": 0.35, "learning_rate": 1.5180335773644688e-05, "loss": 1.0002, "step": 6828 }, { "epoch": 0.35, "learning_rate": 1.517892828008217e-05, "loss": 0.9545, "step": 6829 }, { "epoch": 0.35, "learning_rate": 1.5177520646305683e-05, "loss": 0.9339, "step": 6830 }, { "epoch": 0.35, "learning_rate": 1.517611287235333e-05, "loss": 0.9902, "step": 6831 }, { "epoch": 0.35, "learning_rate": 1.5174704958263232e-05, "loss": 0.9981, "step": 6832 }, { "epoch": 0.35, "learning_rate": 1.5173296904073498e-05, "loss": 0.8671, "step": 6833 }, { "epoch": 0.35, "learning_rate": 1.5171888709822256e-05, "loss": 0.9908, "step": 6834 }, { "epoch": 0.35, "learning_rate": 1.517048037554763e-05, "loss": 0.9363, "step": 6835 }, { "epoch": 0.35, "learning_rate": 1.5169071901287748e-05, "loss": 0.918, "step": 6836 }, { "epoch": 0.35, "learning_rate": 1.5167663287080743e-05, "loss": 1.0477, "step": 6837 }, { "epoch": 0.35, "learning_rate": 1.5166254532964755e-05, "loss": 0.931, "step": 6838 }, { "epoch": 0.35, "learning_rate": 1.516484563897792e-05, "loss": 0.8972, "step": 6839 }, { "epoch": 0.35, "learning_rate": 1.5163436605158386e-05, "loss": 1.0502, "step": 6840 }, { "epoch": 0.35, "learning_rate": 1.5162027431544295e-05, "loss": 0.981, "step": 6841 }, { "epoch": 0.35, "learning_rate": 1.5160618118173806e-05, "loss": 1.0148, "step": 6842 }, { "epoch": 0.35, "learning_rate": 1.5159208665085072e-05, "loss": 0.9405, "step": 6843 }, { "epoch": 0.35, "learning_rate": 1.5157799072316254e-05, "loss": 0.9976, "step": 6844 }, { "epoch": 0.35, "learning_rate": 1.5156389339905514e-05, "loss": 0.9724, "step": 6845 }, { "epoch": 0.35, "learning_rate": 1.5154979467891018e-05, "loss": 0.9037, "step": 6846 }, { "epoch": 0.35, "learning_rate": 1.5153569456310933e-05, "loss": 0.9743, "step": 6847 }, { "epoch": 0.35, "learning_rate": 1.5152159305203445e-05, "loss": 1.0631, "step": 6848 }, { "epoch": 0.35, "learning_rate": 1.5150749014606718e-05, "loss": 1.0042, "step": 6849 }, { "epoch": 0.35, "learning_rate": 1.5149338584558946e-05, "loss": 0.8766, "step": 6850 }, { "epoch": 0.35, "learning_rate": 1.5147928015098309e-05, "loss": 0.9264, "step": 6851 }, { "epoch": 0.35, "learning_rate": 1.5146517306262998e-05, "loss": 0.9342, "step": 6852 }, { "epoch": 0.35, "learning_rate": 1.5145106458091205e-05, "loss": 0.9975, "step": 6853 }, { "epoch": 0.35, "learning_rate": 1.5143695470621133e-05, "loss": 0.9871, "step": 6854 }, { "epoch": 0.35, "learning_rate": 1.5142284343890974e-05, "loss": 0.9583, "step": 6855 }, { "epoch": 0.35, "learning_rate": 1.5140873077938941e-05, "loss": 0.9138, "step": 6856 }, { "epoch": 0.35, "learning_rate": 1.5139461672803235e-05, "loss": 1.0039, "step": 6857 }, { "epoch": 0.35, "learning_rate": 1.5138050128522074e-05, "loss": 0.9296, "step": 6858 }, { "epoch": 0.35, "learning_rate": 1.5136638445133672e-05, "loss": 0.9826, "step": 6859 }, { "epoch": 0.35, "learning_rate": 1.5135226622676249e-05, "loss": 0.9626, "step": 6860 }, { "epoch": 0.35, "learning_rate": 1.513381466118803e-05, "loss": 0.9044, "step": 6861 }, { "epoch": 0.35, "learning_rate": 1.513240256070724e-05, "loss": 0.8888, "step": 6862 }, { "epoch": 0.35, "learning_rate": 1.5130990321272111e-05, "loss": 1.0233, "step": 6863 }, { "epoch": 0.35, "learning_rate": 1.5129577942920879e-05, "loss": 0.9352, "step": 6864 }, { "epoch": 0.35, "learning_rate": 1.512816542569178e-05, "loss": 0.8271, "step": 6865 }, { "epoch": 0.35, "learning_rate": 1.512675276962306e-05, "loss": 1.0075, "step": 6866 }, { "epoch": 0.35, "learning_rate": 1.5125339974752962e-05, "loss": 0.8383, "step": 6867 }, { "epoch": 0.35, "learning_rate": 1.5123927041119738e-05, "loss": 0.9592, "step": 6868 }, { "epoch": 0.35, "learning_rate": 1.5122513968761638e-05, "loss": 0.9689, "step": 6869 }, { "epoch": 0.35, "learning_rate": 1.5121100757716924e-05, "loss": 0.9142, "step": 6870 }, { "epoch": 0.35, "learning_rate": 1.5119687408023856e-05, "loss": 1.0117, "step": 6871 }, { "epoch": 0.35, "learning_rate": 1.5118273919720697e-05, "loss": 0.9481, "step": 6872 }, { "epoch": 0.35, "learning_rate": 1.5116860292845719e-05, "loss": 0.9911, "step": 6873 }, { "epoch": 0.35, "learning_rate": 1.5115446527437193e-05, "loss": 0.8136, "step": 6874 }, { "epoch": 0.35, "learning_rate": 1.5114032623533393e-05, "loss": 0.8609, "step": 6875 }, { "epoch": 0.35, "learning_rate": 1.51126185811726e-05, "loss": 0.9458, "step": 6876 }, { "epoch": 0.35, "learning_rate": 1.5111204400393099e-05, "loss": 1.0594, "step": 6877 }, { "epoch": 0.35, "learning_rate": 1.5109790081233175e-05, "loss": 0.9782, "step": 6878 }, { "epoch": 0.35, "learning_rate": 1.5108375623731123e-05, "loss": 1.0309, "step": 6879 }, { "epoch": 0.35, "learning_rate": 1.5106961027925238e-05, "loss": 0.8531, "step": 6880 }, { "epoch": 0.35, "learning_rate": 1.5105546293853812e-05, "loss": 0.8602, "step": 6881 }, { "epoch": 0.35, "learning_rate": 1.5104131421555155e-05, "loss": 1.0101, "step": 6882 }, { "epoch": 0.35, "learning_rate": 1.5102716411067568e-05, "loss": 1.0219, "step": 6883 }, { "epoch": 0.35, "learning_rate": 1.5101301262429364e-05, "loss": 1.0133, "step": 6884 }, { "epoch": 0.35, "learning_rate": 1.5099885975678856e-05, "loss": 1.0339, "step": 6885 }, { "epoch": 0.35, "learning_rate": 1.5098470550854363e-05, "loss": 1.0388, "step": 6886 }, { "epoch": 0.35, "learning_rate": 1.5097054987994199e-05, "loss": 1.0521, "step": 6887 }, { "epoch": 0.35, "learning_rate": 1.5095639287136702e-05, "loss": 0.951, "step": 6888 }, { "epoch": 0.35, "learning_rate": 1.5094223448320185e-05, "loss": 1.0588, "step": 6889 }, { "epoch": 0.35, "learning_rate": 1.509280747158299e-05, "loss": 0.9758, "step": 6890 }, { "epoch": 0.35, "learning_rate": 1.5091391356963451e-05, "loss": 0.9084, "step": 6891 }, { "epoch": 0.35, "learning_rate": 1.5089975104499913e-05, "loss": 0.8964, "step": 6892 }, { "epoch": 0.35, "learning_rate": 1.5088558714230707e-05, "loss": 0.9594, "step": 6893 }, { "epoch": 0.35, "learning_rate": 1.5087142186194191e-05, "loss": 1.0332, "step": 6894 }, { "epoch": 0.35, "learning_rate": 1.5085725520428712e-05, "loss": 0.9852, "step": 6895 }, { "epoch": 0.35, "learning_rate": 1.5084308716972626e-05, "loss": 0.9413, "step": 6896 }, { "epoch": 0.35, "learning_rate": 1.508289177586429e-05, "loss": 1.0097, "step": 6897 }, { "epoch": 0.35, "learning_rate": 1.5081474697142068e-05, "loss": 1.0523, "step": 6898 }, { "epoch": 0.35, "learning_rate": 1.5080057480844323e-05, "loss": 0.9716, "step": 6899 }, { "epoch": 0.35, "learning_rate": 1.5078640127009427e-05, "loss": 0.9985, "step": 6900 }, { "epoch": 0.35, "learning_rate": 1.5077222635675754e-05, "loss": 0.8743, "step": 6901 }, { "epoch": 0.35, "learning_rate": 1.5075805006881678e-05, "loss": 0.9323, "step": 6902 }, { "epoch": 0.35, "learning_rate": 1.5074387240665584e-05, "loss": 1.0581, "step": 6903 }, { "epoch": 0.35, "learning_rate": 1.5072969337065855e-05, "loss": 0.9803, "step": 6904 }, { "epoch": 0.35, "learning_rate": 1.5071551296120874e-05, "loss": 1.0055, "step": 6905 }, { "epoch": 0.35, "learning_rate": 1.5070133117869041e-05, "loss": 1.0866, "step": 6906 }, { "epoch": 0.35, "learning_rate": 1.5068714802348748e-05, "loss": 1.0337, "step": 6907 }, { "epoch": 0.35, "learning_rate": 1.506729634959839e-05, "loss": 0.8706, "step": 6908 }, { "epoch": 0.35, "learning_rate": 1.5065877759656382e-05, "loss": 0.9459, "step": 6909 }, { "epoch": 0.35, "learning_rate": 1.5064459032561118e-05, "loss": 0.8954, "step": 6910 }, { "epoch": 0.35, "learning_rate": 1.5063040168351016e-05, "loss": 0.9015, "step": 6911 }, { "epoch": 0.35, "learning_rate": 1.5061621167064485e-05, "loss": 0.9722, "step": 6912 }, { "epoch": 0.35, "learning_rate": 1.5060202028739949e-05, "loss": 1.0102, "step": 6913 }, { "epoch": 0.35, "learning_rate": 1.5058782753415826e-05, "loss": 1.0464, "step": 6914 }, { "epoch": 0.35, "learning_rate": 1.5057363341130542e-05, "loss": 0.9745, "step": 6915 }, { "epoch": 0.35, "learning_rate": 1.5055943791922528e-05, "loss": 0.8777, "step": 6916 }, { "epoch": 0.35, "learning_rate": 1.5054524105830211e-05, "loss": 1.0515, "step": 6917 }, { "epoch": 0.35, "learning_rate": 1.5053104282892032e-05, "loss": 0.9406, "step": 6918 }, { "epoch": 0.35, "learning_rate": 1.5051684323146434e-05, "loss": 0.8851, "step": 6919 }, { "epoch": 0.35, "learning_rate": 1.5050264226631855e-05, "loss": 1.0009, "step": 6920 }, { "epoch": 0.35, "learning_rate": 1.5048843993386746e-05, "loss": 0.93, "step": 6921 }, { "epoch": 0.35, "learning_rate": 1.5047423623449557e-05, "loss": 0.9769, "step": 6922 }, { "epoch": 0.35, "learning_rate": 1.5046003116858745e-05, "loss": 0.9355, "step": 6923 }, { "epoch": 0.35, "learning_rate": 1.5044582473652764e-05, "loss": 0.9081, "step": 6924 }, { "epoch": 0.35, "learning_rate": 1.5043161693870083e-05, "loss": 0.9494, "step": 6925 }, { "epoch": 0.35, "learning_rate": 1.504174077754916e-05, "loss": 0.9507, "step": 6926 }, { "epoch": 0.35, "learning_rate": 1.5040319724728474e-05, "loss": 0.8516, "step": 6927 }, { "epoch": 0.35, "learning_rate": 1.5038898535446491e-05, "loss": 0.9351, "step": 6928 }, { "epoch": 0.35, "learning_rate": 1.5037477209741694e-05, "loss": 0.9951, "step": 6929 }, { "epoch": 0.35, "learning_rate": 1.5036055747652559e-05, "loss": 0.9311, "step": 6930 }, { "epoch": 0.35, "learning_rate": 1.503463414921757e-05, "loss": 0.9834, "step": 6931 }, { "epoch": 0.35, "learning_rate": 1.5033212414475222e-05, "loss": 0.9936, "step": 6932 }, { "epoch": 0.35, "learning_rate": 1.5031790543464003e-05, "loss": 0.873, "step": 6933 }, { "epoch": 0.35, "learning_rate": 1.5030368536222407e-05, "loss": 1.0092, "step": 6934 }, { "epoch": 0.35, "learning_rate": 1.5028946392788934e-05, "loss": 0.864, "step": 6935 }, { "epoch": 0.35, "learning_rate": 1.5027524113202088e-05, "loss": 0.8657, "step": 6936 }, { "epoch": 0.35, "learning_rate": 1.5026101697500374e-05, "loss": 1.0447, "step": 6937 }, { "epoch": 0.35, "learning_rate": 1.5024679145722308e-05, "loss": 0.9194, "step": 6938 }, { "epoch": 0.35, "learning_rate": 1.5023256457906401e-05, "loss": 0.9879, "step": 6939 }, { "epoch": 0.35, "learning_rate": 1.5021833634091162e-05, "loss": 0.9104, "step": 6940 }, { "epoch": 0.35, "learning_rate": 1.5020410674315127e-05, "loss": 0.9149, "step": 6941 }, { "epoch": 0.35, "learning_rate": 1.5018987578616813e-05, "loss": 0.932, "step": 6942 }, { "epoch": 0.35, "learning_rate": 1.5017564347034748e-05, "loss": 0.9606, "step": 6943 }, { "epoch": 0.35, "learning_rate": 1.5016140979607467e-05, "loss": 0.9737, "step": 6944 }, { "epoch": 0.35, "learning_rate": 1.5014717476373508e-05, "loss": 0.9634, "step": 6945 }, { "epoch": 0.35, "learning_rate": 1.5013293837371405e-05, "loss": 0.993, "step": 6946 }, { "epoch": 0.35, "learning_rate": 1.501187006263971e-05, "loss": 0.8629, "step": 6947 }, { "epoch": 0.35, "learning_rate": 1.5010446152216962e-05, "loss": 0.8924, "step": 6948 }, { "epoch": 0.35, "learning_rate": 1.5009022106141716e-05, "loss": 0.9923, "step": 6949 }, { "epoch": 0.35, "learning_rate": 1.5007597924452526e-05, "loss": 0.9988, "step": 6950 }, { "epoch": 0.35, "learning_rate": 1.5006173607187948e-05, "loss": 0.946, "step": 6951 }, { "epoch": 0.35, "learning_rate": 1.5004749154386547e-05, "loss": 1.0204, "step": 6952 }, { "epoch": 0.35, "learning_rate": 1.5003324566086887e-05, "loss": 0.8866, "step": 6953 }, { "epoch": 0.35, "learning_rate": 1.5001899842327539e-05, "loss": 0.9628, "step": 6954 }, { "epoch": 0.35, "learning_rate": 1.5000474983147068e-05, "loss": 1.1275, "step": 6955 }, { "epoch": 0.35, "learning_rate": 1.4999049988584064e-05, "loss": 0.8592, "step": 6956 }, { "epoch": 0.35, "learning_rate": 1.49976248586771e-05, "loss": 1.0023, "step": 6957 }, { "epoch": 0.35, "learning_rate": 1.4996199593464757e-05, "loss": 0.9327, "step": 6958 }, { "epoch": 0.35, "learning_rate": 1.4994774192985628e-05, "loss": 0.9488, "step": 6959 }, { "epoch": 0.35, "learning_rate": 1.49933486572783e-05, "loss": 0.9134, "step": 6960 }, { "epoch": 0.35, "learning_rate": 1.4991922986381368e-05, "loss": 0.9926, "step": 6961 }, { "epoch": 0.35, "learning_rate": 1.4990497180333435e-05, "loss": 0.9031, "step": 6962 }, { "epoch": 0.35, "learning_rate": 1.4989071239173101e-05, "loss": 1.0298, "step": 6963 }, { "epoch": 0.35, "learning_rate": 1.4987645162938966e-05, "loss": 0.9553, "step": 6964 }, { "epoch": 0.35, "learning_rate": 1.498621895166965e-05, "loss": 0.9985, "step": 6965 }, { "epoch": 0.35, "learning_rate": 1.4984792605403757e-05, "loss": 0.989, "step": 6966 }, { "epoch": 0.35, "learning_rate": 1.498336612417991e-05, "loss": 0.884, "step": 6967 }, { "epoch": 0.35, "learning_rate": 1.498193950803673e-05, "loss": 0.9391, "step": 6968 }, { "epoch": 0.35, "learning_rate": 1.4980512757012835e-05, "loss": 1.018, "step": 6969 }, { "epoch": 0.35, "learning_rate": 1.4979085871146853e-05, "loss": 0.9113, "step": 6970 }, { "epoch": 0.35, "learning_rate": 1.4977658850477421e-05, "loss": 0.9028, "step": 6971 }, { "epoch": 0.35, "learning_rate": 1.4976231695043173e-05, "loss": 0.9525, "step": 6972 }, { "epoch": 0.35, "learning_rate": 1.4974804404882743e-05, "loss": 0.9944, "step": 6973 }, { "epoch": 0.35, "learning_rate": 1.497337698003478e-05, "loss": 0.9588, "step": 6974 }, { "epoch": 0.35, "learning_rate": 1.4971949420537925e-05, "loss": 1.0033, "step": 6975 }, { "epoch": 0.35, "learning_rate": 1.4970521726430827e-05, "loss": 0.9656, "step": 6976 }, { "epoch": 0.35, "learning_rate": 1.4969093897752143e-05, "loss": 1.019, "step": 6977 }, { "epoch": 0.35, "learning_rate": 1.496766593454053e-05, "loss": 0.9623, "step": 6978 }, { "epoch": 0.35, "learning_rate": 1.4966237836834646e-05, "loss": 0.9987, "step": 6979 }, { "epoch": 0.35, "learning_rate": 1.4964809604673156e-05, "loss": 1.0543, "step": 6980 }, { "epoch": 0.35, "learning_rate": 1.496338123809473e-05, "loss": 1.0263, "step": 6981 }, { "epoch": 0.35, "learning_rate": 1.4961952737138034e-05, "loss": 0.9604, "step": 6982 }, { "epoch": 0.35, "learning_rate": 1.4960524101841747e-05, "loss": 1.0818, "step": 6983 }, { "epoch": 0.35, "learning_rate": 1.495909533224455e-05, "loss": 0.9495, "step": 6984 }, { "epoch": 0.35, "learning_rate": 1.4957666428385122e-05, "loss": 0.934, "step": 6985 }, { "epoch": 0.35, "learning_rate": 1.495623739030215e-05, "loss": 0.8902, "step": 6986 }, { "epoch": 0.35, "learning_rate": 1.4954808218034324e-05, "loss": 0.999, "step": 6987 }, { "epoch": 0.36, "learning_rate": 1.4953378911620337e-05, "loss": 1.0187, "step": 6988 }, { "epoch": 0.36, "learning_rate": 1.4951949471098886e-05, "loss": 0.9772, "step": 6989 }, { "epoch": 0.36, "learning_rate": 1.495051989650867e-05, "loss": 0.9889, "step": 6990 }, { "epoch": 0.36, "learning_rate": 1.4949090187888396e-05, "loss": 0.9857, "step": 6991 }, { "epoch": 0.36, "learning_rate": 1.494766034527677e-05, "loss": 0.9462, "step": 6992 }, { "epoch": 0.36, "learning_rate": 1.4946230368712507e-05, "loss": 0.9984, "step": 6993 }, { "epoch": 0.36, "learning_rate": 1.4944800258234316e-05, "loss": 0.9953, "step": 6994 }, { "epoch": 0.36, "learning_rate": 1.4943370013880918e-05, "loss": 1.0307, "step": 6995 }, { "epoch": 0.36, "learning_rate": 1.4941939635691036e-05, "loss": 0.9423, "step": 6996 }, { "epoch": 0.36, "learning_rate": 1.4940509123703399e-05, "loss": 0.9143, "step": 6997 }, { "epoch": 0.36, "learning_rate": 1.4939078477956735e-05, "loss": 0.877, "step": 6998 }, { "epoch": 0.36, "learning_rate": 1.4937647698489773e-05, "loss": 1.0096, "step": 6999 }, { "epoch": 0.36, "learning_rate": 1.4936216785341254e-05, "loss": 0.8775, "step": 7000 }, { "epoch": 0.36, "learning_rate": 1.4934785738549914e-05, "loss": 1.0439, "step": 7001 }, { "epoch": 0.36, "learning_rate": 1.4933354558154503e-05, "loss": 1.0302, "step": 7002 }, { "epoch": 0.36, "learning_rate": 1.4931923244193767e-05, "loss": 0.9897, "step": 7003 }, { "epoch": 0.36, "learning_rate": 1.4930491796706457e-05, "loss": 1.0059, "step": 7004 }, { "epoch": 0.36, "learning_rate": 1.4929060215731324e-05, "loss": 0.9041, "step": 7005 }, { "epoch": 0.36, "learning_rate": 1.4927628501307132e-05, "loss": 0.8987, "step": 7006 }, { "epoch": 0.36, "learning_rate": 1.492619665347264e-05, "loss": 0.8456, "step": 7007 }, { "epoch": 0.36, "learning_rate": 1.4924764672266614e-05, "loss": 0.9146, "step": 7008 }, { "epoch": 0.36, "learning_rate": 1.4923332557727825e-05, "loss": 1.0243, "step": 7009 }, { "epoch": 0.36, "learning_rate": 1.4921900309895046e-05, "loss": 1.0635, "step": 7010 }, { "epoch": 0.36, "learning_rate": 1.492046792880705e-05, "loss": 0.8803, "step": 7011 }, { "epoch": 0.36, "learning_rate": 1.4919035414502623e-05, "loss": 1.0242, "step": 7012 }, { "epoch": 0.36, "learning_rate": 1.4917602767020543e-05, "loss": 0.8592, "step": 7013 }, { "epoch": 0.36, "learning_rate": 1.49161699863996e-05, "loss": 0.9496, "step": 7014 }, { "epoch": 0.36, "learning_rate": 1.4914737072678587e-05, "loss": 0.9745, "step": 7015 }, { "epoch": 0.36, "learning_rate": 1.4913304025896295e-05, "loss": 0.9401, "step": 7016 }, { "epoch": 0.36, "learning_rate": 1.4911870846091526e-05, "loss": 0.9137, "step": 7017 }, { "epoch": 0.36, "learning_rate": 1.4910437533303079e-05, "loss": 0.9499, "step": 7018 }, { "epoch": 0.36, "learning_rate": 1.4909004087569758e-05, "loss": 0.9557, "step": 7019 }, { "epoch": 0.36, "learning_rate": 1.4907570508930374e-05, "loss": 0.9517, "step": 7020 }, { "epoch": 0.36, "learning_rate": 1.4906136797423743e-05, "loss": 1.0892, "step": 7021 }, { "epoch": 0.36, "learning_rate": 1.4904702953088674e-05, "loss": 0.9316, "step": 7022 }, { "epoch": 0.36, "learning_rate": 1.4903268975963994e-05, "loss": 0.9727, "step": 7023 }, { "epoch": 0.36, "learning_rate": 1.4901834866088522e-05, "loss": 0.9113, "step": 7024 }, { "epoch": 0.36, "learning_rate": 1.4900400623501085e-05, "loss": 1.0427, "step": 7025 }, { "epoch": 0.36, "learning_rate": 1.4898966248240515e-05, "loss": 0.9404, "step": 7026 }, { "epoch": 0.36, "learning_rate": 1.4897531740345648e-05, "loss": 0.9363, "step": 7027 }, { "epoch": 0.36, "learning_rate": 1.4896097099855318e-05, "loss": 0.9443, "step": 7028 }, { "epoch": 0.36, "learning_rate": 1.4894662326808367e-05, "loss": 1.0024, "step": 7029 }, { "epoch": 0.36, "learning_rate": 1.4893227421243644e-05, "loss": 1.0201, "step": 7030 }, { "epoch": 0.36, "learning_rate": 1.4891792383199992e-05, "loss": 1.0185, "step": 7031 }, { "epoch": 0.36, "learning_rate": 1.4890357212716266e-05, "loss": 0.9299, "step": 7032 }, { "epoch": 0.36, "learning_rate": 1.4888921909831324e-05, "loss": 0.9242, "step": 7033 }, { "epoch": 0.36, "learning_rate": 1.4887486474584022e-05, "loss": 0.9485, "step": 7034 }, { "epoch": 0.36, "learning_rate": 1.4886050907013218e-05, "loss": 0.9932, "step": 7035 }, { "epoch": 0.36, "learning_rate": 1.4884615207157788e-05, "loss": 0.9333, "step": 7036 }, { "epoch": 0.36, "learning_rate": 1.48831793750566e-05, "loss": 1.0, "step": 7037 }, { "epoch": 0.36, "learning_rate": 1.4881743410748524e-05, "loss": 0.9824, "step": 7038 }, { "epoch": 0.36, "learning_rate": 1.488030731427244e-05, "loss": 0.9347, "step": 7039 }, { "epoch": 0.36, "learning_rate": 1.4878871085667228e-05, "loss": 0.918, "step": 7040 }, { "epoch": 0.36, "learning_rate": 1.4877434724971768e-05, "loss": 1.0399, "step": 7041 }, { "epoch": 0.36, "learning_rate": 1.4875998232224956e-05, "loss": 0.9531, "step": 7042 }, { "epoch": 0.36, "learning_rate": 1.4874561607465678e-05, "loss": 0.9275, "step": 7043 }, { "epoch": 0.36, "learning_rate": 1.487312485073283e-05, "loss": 0.9092, "step": 7044 }, { "epoch": 0.36, "learning_rate": 1.4871687962065313e-05, "loss": 0.9474, "step": 7045 }, { "epoch": 0.36, "learning_rate": 1.487025094150203e-05, "loss": 0.9873, "step": 7046 }, { "epoch": 0.36, "learning_rate": 1.486881378908188e-05, "loss": 1.0096, "step": 7047 }, { "epoch": 0.36, "learning_rate": 1.4867376504843775e-05, "loss": 0.9297, "step": 7048 }, { "epoch": 0.36, "learning_rate": 1.4865939088826634e-05, "loss": 0.9291, "step": 7049 }, { "epoch": 0.36, "learning_rate": 1.4864501541069365e-05, "loss": 0.957, "step": 7050 }, { "epoch": 0.36, "learning_rate": 1.4863063861610896e-05, "loss": 0.8592, "step": 7051 }, { "epoch": 0.36, "learning_rate": 1.4861626050490148e-05, "loss": 0.9657, "step": 7052 }, { "epoch": 0.36, "learning_rate": 1.4860188107746041e-05, "loss": 0.9232, "step": 7053 }, { "epoch": 0.36, "learning_rate": 1.4858750033417516e-05, "loss": 1.0629, "step": 7054 }, { "epoch": 0.36, "learning_rate": 1.4857311827543503e-05, "loss": 0.9246, "step": 7055 }, { "epoch": 0.36, "learning_rate": 1.4855873490162937e-05, "loss": 1.0032, "step": 7056 }, { "epoch": 0.36, "learning_rate": 1.4854435021314766e-05, "loss": 0.9045, "step": 7057 }, { "epoch": 0.36, "learning_rate": 1.4852996421037932e-05, "loss": 0.9645, "step": 7058 }, { "epoch": 0.36, "learning_rate": 1.485155768937138e-05, "loss": 0.9218, "step": 7059 }, { "epoch": 0.36, "learning_rate": 1.4850118826354067e-05, "loss": 0.941, "step": 7060 }, { "epoch": 0.36, "learning_rate": 1.4848679832024948e-05, "loss": 0.9126, "step": 7061 }, { "epoch": 0.36, "learning_rate": 1.4847240706422978e-05, "loss": 0.9234, "step": 7062 }, { "epoch": 0.36, "learning_rate": 1.4845801449587125e-05, "loss": 0.8935, "step": 7063 }, { "epoch": 0.36, "learning_rate": 1.4844362061556355e-05, "loss": 1.0253, "step": 7064 }, { "epoch": 0.36, "learning_rate": 1.4842922542369633e-05, "loss": 1.0513, "step": 7065 }, { "epoch": 0.36, "learning_rate": 1.4841482892065938e-05, "loss": 0.9753, "step": 7066 }, { "epoch": 0.36, "learning_rate": 1.4840043110684241e-05, "loss": 0.956, "step": 7067 }, { "epoch": 0.36, "learning_rate": 1.483860319826353e-05, "loss": 0.9271, "step": 7068 }, { "epoch": 0.36, "learning_rate": 1.4837163154842784e-05, "loss": 0.9669, "step": 7069 }, { "epoch": 0.36, "learning_rate": 1.4835722980460992e-05, "loss": 0.9832, "step": 7070 }, { "epoch": 0.36, "learning_rate": 1.4834282675157145e-05, "loss": 1.1092, "step": 7071 }, { "epoch": 0.36, "learning_rate": 1.4832842238970237e-05, "loss": 0.9514, "step": 7072 }, { "epoch": 0.36, "learning_rate": 1.4831401671939268e-05, "loss": 0.9439, "step": 7073 }, { "epoch": 0.36, "learning_rate": 1.4829960974103238e-05, "loss": 0.857, "step": 7074 }, { "epoch": 0.36, "learning_rate": 1.4828520145501158e-05, "loss": 0.9333, "step": 7075 }, { "epoch": 0.36, "learning_rate": 1.4827079186172024e-05, "loss": 0.9568, "step": 7076 }, { "epoch": 0.36, "learning_rate": 1.4825638096154865e-05, "loss": 0.9199, "step": 7077 }, { "epoch": 0.36, "learning_rate": 1.4824196875488684e-05, "loss": 0.9328, "step": 7078 }, { "epoch": 0.36, "learning_rate": 1.4822755524212503e-05, "loss": 1.0103, "step": 7079 }, { "epoch": 0.36, "learning_rate": 1.482131404236535e-05, "loss": 1.0654, "step": 7080 }, { "epoch": 0.36, "learning_rate": 1.481987242998625e-05, "loss": 0.9134, "step": 7081 }, { "epoch": 0.36, "learning_rate": 1.4818430687114229e-05, "loss": 0.903, "step": 7082 }, { "epoch": 0.36, "learning_rate": 1.4816988813788327e-05, "loss": 0.9299, "step": 7083 }, { "epoch": 0.36, "learning_rate": 1.4815546810047577e-05, "loss": 1.0618, "step": 7084 }, { "epoch": 0.36, "learning_rate": 1.4814104675931017e-05, "loss": 0.9311, "step": 7085 }, { "epoch": 0.36, "learning_rate": 1.4812662411477697e-05, "loss": 0.9238, "step": 7086 }, { "epoch": 0.36, "learning_rate": 1.4811220016726664e-05, "loss": 0.9286, "step": 7087 }, { "epoch": 0.36, "learning_rate": 1.4809777491716965e-05, "loss": 0.938, "step": 7088 }, { "epoch": 0.36, "learning_rate": 1.4808334836487661e-05, "loss": 1.0035, "step": 7089 }, { "epoch": 0.36, "learning_rate": 1.4806892051077803e-05, "loss": 0.9607, "step": 7090 }, { "epoch": 0.36, "learning_rate": 1.4805449135526457e-05, "loss": 1.0172, "step": 7091 }, { "epoch": 0.36, "learning_rate": 1.4804006089872692e-05, "loss": 0.8881, "step": 7092 }, { "epoch": 0.36, "learning_rate": 1.4802562914155572e-05, "loss": 0.9376, "step": 7093 }, { "epoch": 0.36, "learning_rate": 1.4801119608414169e-05, "loss": 0.9941, "step": 7094 }, { "epoch": 0.36, "learning_rate": 1.4799676172687562e-05, "loss": 0.9634, "step": 7095 }, { "epoch": 0.36, "learning_rate": 1.479823260701483e-05, "loss": 0.9608, "step": 7096 }, { "epoch": 0.36, "learning_rate": 1.4796788911435052e-05, "loss": 0.847, "step": 7097 }, { "epoch": 0.36, "learning_rate": 1.4795345085987319e-05, "loss": 0.9799, "step": 7098 }, { "epoch": 0.36, "learning_rate": 1.4793901130710721e-05, "loss": 0.8785, "step": 7099 }, { "epoch": 0.36, "learning_rate": 1.4792457045644346e-05, "loss": 0.9767, "step": 7100 }, { "epoch": 0.36, "learning_rate": 1.4791012830827298e-05, "loss": 0.991, "step": 7101 }, { "epoch": 0.36, "learning_rate": 1.4789568486298678e-05, "loss": 0.9464, "step": 7102 }, { "epoch": 0.36, "learning_rate": 1.4788124012097583e-05, "loss": 0.8058, "step": 7103 }, { "epoch": 0.36, "learning_rate": 1.4786679408263124e-05, "loss": 0.8925, "step": 7104 }, { "epoch": 0.36, "learning_rate": 1.4785234674834417e-05, "loss": 0.9786, "step": 7105 }, { "epoch": 0.36, "learning_rate": 1.4783789811850569e-05, "loss": 1.0104, "step": 7106 }, { "epoch": 0.36, "learning_rate": 1.47823448193507e-05, "loss": 0.928, "step": 7107 }, { "epoch": 0.36, "learning_rate": 1.4780899697373937e-05, "loss": 0.922, "step": 7108 }, { "epoch": 0.36, "learning_rate": 1.4779454445959398e-05, "loss": 0.8797, "step": 7109 }, { "epoch": 0.36, "learning_rate": 1.4778009065146215e-05, "loss": 0.9342, "step": 7110 }, { "epoch": 0.36, "learning_rate": 1.4776563554973523e-05, "loss": 0.9955, "step": 7111 }, { "epoch": 0.36, "learning_rate": 1.4775117915480453e-05, "loss": 0.9717, "step": 7112 }, { "epoch": 0.36, "learning_rate": 1.4773672146706144e-05, "loss": 0.9613, "step": 7113 }, { "epoch": 0.36, "learning_rate": 1.477222624868974e-05, "loss": 0.9419, "step": 7114 }, { "epoch": 0.36, "learning_rate": 1.4770780221470389e-05, "loss": 0.9763, "step": 7115 }, { "epoch": 0.36, "learning_rate": 1.4769334065087239e-05, "loss": 0.957, "step": 7116 }, { "epoch": 0.36, "learning_rate": 1.4767887779579445e-05, "loss": 0.9899, "step": 7117 }, { "epoch": 0.36, "learning_rate": 1.4766441364986162e-05, "loss": 0.9944, "step": 7118 }, { "epoch": 0.36, "learning_rate": 1.4764994821346548e-05, "loss": 0.9304, "step": 7119 }, { "epoch": 0.36, "learning_rate": 1.4763548148699771e-05, "loss": 0.9325, "step": 7120 }, { "epoch": 0.36, "learning_rate": 1.4762101347084991e-05, "loss": 0.9302, "step": 7121 }, { "epoch": 0.36, "learning_rate": 1.476065441654139e-05, "loss": 1.0867, "step": 7122 }, { "epoch": 0.36, "learning_rate": 1.4759207357108133e-05, "loss": 0.9262, "step": 7123 }, { "epoch": 0.36, "learning_rate": 1.47577601688244e-05, "loss": 0.9785, "step": 7124 }, { "epoch": 0.36, "learning_rate": 1.4756312851729371e-05, "loss": 0.944, "step": 7125 }, { "epoch": 0.36, "learning_rate": 1.4754865405862233e-05, "loss": 0.9915, "step": 7126 }, { "epoch": 0.36, "learning_rate": 1.4753417831262173e-05, "loss": 0.9335, "step": 7127 }, { "epoch": 0.36, "learning_rate": 1.4751970127968384e-05, "loss": 1.03, "step": 7128 }, { "epoch": 0.36, "learning_rate": 1.4750522296020059e-05, "loss": 1.0748, "step": 7129 }, { "epoch": 0.36, "learning_rate": 1.4749074335456395e-05, "loss": 0.9869, "step": 7130 }, { "epoch": 0.36, "learning_rate": 1.4747626246316595e-05, "loss": 0.9459, "step": 7131 }, { "epoch": 0.36, "learning_rate": 1.4746178028639869e-05, "loss": 0.9402, "step": 7132 }, { "epoch": 0.36, "learning_rate": 1.474472968246542e-05, "loss": 0.9473, "step": 7133 }, { "epoch": 0.36, "learning_rate": 1.4743281207832466e-05, "loss": 1.0409, "step": 7134 }, { "epoch": 0.36, "learning_rate": 1.4741832604780216e-05, "loss": 0.9902, "step": 7135 }, { "epoch": 0.36, "learning_rate": 1.4740383873347895e-05, "loss": 0.9586, "step": 7136 }, { "epoch": 0.36, "learning_rate": 1.473893501357472e-05, "loss": 0.8898, "step": 7137 }, { "epoch": 0.36, "learning_rate": 1.4737486025499926e-05, "loss": 0.9496, "step": 7138 }, { "epoch": 0.36, "learning_rate": 1.4736036909162737e-05, "loss": 0.9837, "step": 7139 }, { "epoch": 0.36, "learning_rate": 1.4734587664602388e-05, "loss": 1.0052, "step": 7140 }, { "epoch": 0.36, "learning_rate": 1.4733138291858113e-05, "loss": 0.9709, "step": 7141 }, { "epoch": 0.36, "learning_rate": 1.4731688790969158e-05, "loss": 0.8676, "step": 7142 }, { "epoch": 0.36, "learning_rate": 1.473023916197476e-05, "loss": 0.9472, "step": 7143 }, { "epoch": 0.36, "learning_rate": 1.472878940491417e-05, "loss": 0.9196, "step": 7144 }, { "epoch": 0.36, "learning_rate": 1.4727339519826638e-05, "loss": 0.9098, "step": 7145 }, { "epoch": 0.36, "learning_rate": 1.472588950675142e-05, "loss": 1.0309, "step": 7146 }, { "epoch": 0.36, "learning_rate": 1.472443936572777e-05, "loss": 0.9232, "step": 7147 }, { "epoch": 0.36, "learning_rate": 1.472298909679495e-05, "loss": 0.8847, "step": 7148 }, { "epoch": 0.36, "learning_rate": 1.4721538699992226e-05, "loss": 0.9385, "step": 7149 }, { "epoch": 0.36, "learning_rate": 1.4720088175358861e-05, "loss": 0.9357, "step": 7150 }, { "epoch": 0.36, "learning_rate": 1.4718637522934136e-05, "loss": 0.9578, "step": 7151 }, { "epoch": 0.36, "learning_rate": 1.4717186742757318e-05, "loss": 0.8522, "step": 7152 }, { "epoch": 0.36, "learning_rate": 1.4715735834867688e-05, "loss": 0.9222, "step": 7153 }, { "epoch": 0.36, "learning_rate": 1.4714284799304528e-05, "loss": 0.9587, "step": 7154 }, { "epoch": 0.36, "learning_rate": 1.4712833636107123e-05, "loss": 0.8782, "step": 7155 }, { "epoch": 0.36, "learning_rate": 1.4711382345314757e-05, "loss": 0.8381, "step": 7156 }, { "epoch": 0.36, "learning_rate": 1.4709930926966732e-05, "loss": 1.0482, "step": 7157 }, { "epoch": 0.36, "learning_rate": 1.4708479381102336e-05, "loss": 0.9604, "step": 7158 }, { "epoch": 0.36, "learning_rate": 1.470702770776087e-05, "loss": 0.9565, "step": 7159 }, { "epoch": 0.36, "learning_rate": 1.4705575906981637e-05, "loss": 0.9457, "step": 7160 }, { "epoch": 0.36, "learning_rate": 1.4704123978803942e-05, "loss": 1.0224, "step": 7161 }, { "epoch": 0.36, "learning_rate": 1.4702671923267096e-05, "loss": 0.909, "step": 7162 }, { "epoch": 0.36, "learning_rate": 1.4701219740410411e-05, "loss": 0.9768, "step": 7163 }, { "epoch": 0.36, "learning_rate": 1.4699767430273202e-05, "loss": 0.9338, "step": 7164 }, { "epoch": 0.36, "learning_rate": 1.4698314992894792e-05, "loss": 0.8967, "step": 7165 }, { "epoch": 0.36, "learning_rate": 1.46968624283145e-05, "loss": 0.9722, "step": 7166 }, { "epoch": 0.36, "learning_rate": 1.4695409736571655e-05, "loss": 0.9442, "step": 7167 }, { "epoch": 0.36, "learning_rate": 1.4693956917705587e-05, "loss": 0.8971, "step": 7168 }, { "epoch": 0.36, "learning_rate": 1.469250397175563e-05, "loss": 0.9583, "step": 7169 }, { "epoch": 0.36, "learning_rate": 1.4691050898761123e-05, "loss": 0.8672, "step": 7170 }, { "epoch": 0.36, "learning_rate": 1.46895976987614e-05, "loss": 0.9309, "step": 7171 }, { "epoch": 0.36, "learning_rate": 1.4688144371795808e-05, "loss": 1.0526, "step": 7172 }, { "epoch": 0.36, "learning_rate": 1.4686690917903698e-05, "loss": 1.0492, "step": 7173 }, { "epoch": 0.36, "learning_rate": 1.4685237337124416e-05, "loss": 0.8916, "step": 7174 }, { "epoch": 0.36, "learning_rate": 1.4683783629497319e-05, "loss": 0.926, "step": 7175 }, { "epoch": 0.36, "learning_rate": 1.4682329795061763e-05, "loss": 0.9917, "step": 7176 }, { "epoch": 0.36, "learning_rate": 1.468087583385711e-05, "loss": 0.9049, "step": 7177 }, { "epoch": 0.36, "learning_rate": 1.4679421745922722e-05, "loss": 0.9235, "step": 7178 }, { "epoch": 0.36, "learning_rate": 1.467796753129797e-05, "loss": 0.9594, "step": 7179 }, { "epoch": 0.36, "learning_rate": 1.467651319002222e-05, "loss": 0.9399, "step": 7180 }, { "epoch": 0.36, "learning_rate": 1.4675058722134856e-05, "loss": 0.9424, "step": 7181 }, { "epoch": 0.36, "learning_rate": 1.4673604127675252e-05, "loss": 0.9756, "step": 7182 }, { "epoch": 0.36, "learning_rate": 1.4672149406682785e-05, "loss": 0.9556, "step": 7183 }, { "epoch": 0.36, "learning_rate": 1.4670694559196842e-05, "loss": 0.9642, "step": 7184 }, { "epoch": 0.37, "learning_rate": 1.4669239585256819e-05, "loss": 0.9134, "step": 7185 }, { "epoch": 0.37, "learning_rate": 1.4667784484902096e-05, "loss": 0.9109, "step": 7186 }, { "epoch": 0.37, "learning_rate": 1.4666329258172078e-05, "loss": 0.859, "step": 7187 }, { "epoch": 0.37, "learning_rate": 1.466487390510616e-05, "loss": 1.0045, "step": 7188 }, { "epoch": 0.37, "learning_rate": 1.4663418425743744e-05, "loss": 1.0238, "step": 7189 }, { "epoch": 0.37, "learning_rate": 1.466196282012423e-05, "loss": 0.8923, "step": 7190 }, { "epoch": 0.37, "learning_rate": 1.4660507088287039e-05, "loss": 0.9632, "step": 7191 }, { "epoch": 0.37, "learning_rate": 1.4659051230271574e-05, "loss": 0.9741, "step": 7192 }, { "epoch": 0.37, "learning_rate": 1.4657595246117256e-05, "loss": 0.9777, "step": 7193 }, { "epoch": 0.37, "learning_rate": 1.46561391358635e-05, "loss": 0.8748, "step": 7194 }, { "epoch": 0.37, "learning_rate": 1.4654682899549733e-05, "loss": 0.9644, "step": 7195 }, { "epoch": 0.37, "learning_rate": 1.4653226537215375e-05, "loss": 0.9358, "step": 7196 }, { "epoch": 0.37, "learning_rate": 1.4651770048899863e-05, "loss": 0.9039, "step": 7197 }, { "epoch": 0.37, "learning_rate": 1.4650313434642622e-05, "loss": 0.9335, "step": 7198 }, { "epoch": 0.37, "learning_rate": 1.4648856694483095e-05, "loss": 0.9183, "step": 7199 }, { "epoch": 0.37, "learning_rate": 1.464739982846072e-05, "loss": 0.9486, "step": 7200 }, { "epoch": 0.37, "learning_rate": 1.4645942836614936e-05, "loss": 0.9711, "step": 7201 }, { "epoch": 0.37, "learning_rate": 1.4644485718985194e-05, "loss": 0.9456, "step": 7202 }, { "epoch": 0.37, "learning_rate": 1.4643028475610943e-05, "loss": 0.9646, "step": 7203 }, { "epoch": 0.37, "learning_rate": 1.4641571106531637e-05, "loss": 0.8811, "step": 7204 }, { "epoch": 0.37, "learning_rate": 1.464011361178673e-05, "loss": 0.9695, "step": 7205 }, { "epoch": 0.37, "learning_rate": 1.4638655991415683e-05, "loss": 0.9587, "step": 7206 }, { "epoch": 0.37, "learning_rate": 1.463719824545796e-05, "loss": 1.0261, "step": 7207 }, { "epoch": 0.37, "learning_rate": 1.4635740373953029e-05, "loss": 0.9716, "step": 7208 }, { "epoch": 0.37, "learning_rate": 1.4634282376940358e-05, "loss": 0.8766, "step": 7209 }, { "epoch": 0.37, "learning_rate": 1.4632824254459426e-05, "loss": 0.9608, "step": 7210 }, { "epoch": 0.37, "learning_rate": 1.4631366006549703e-05, "loss": 1.1061, "step": 7211 }, { "epoch": 0.37, "learning_rate": 1.4629907633250672e-05, "loss": 1.0078, "step": 7212 }, { "epoch": 0.37, "learning_rate": 1.462844913460182e-05, "loss": 0.9477, "step": 7213 }, { "epoch": 0.37, "learning_rate": 1.4626990510642632e-05, "loss": 1.0009, "step": 7214 }, { "epoch": 0.37, "learning_rate": 1.4625531761412598e-05, "loss": 0.8618, "step": 7215 }, { "epoch": 0.37, "learning_rate": 1.462407288695121e-05, "loss": 0.8901, "step": 7216 }, { "epoch": 0.37, "learning_rate": 1.4622613887297971e-05, "loss": 0.9095, "step": 7217 }, { "epoch": 0.37, "learning_rate": 1.462115476249238e-05, "loss": 0.9677, "step": 7218 }, { "epoch": 0.37, "learning_rate": 1.4619695512573939e-05, "loss": 1.0017, "step": 7219 }, { "epoch": 0.37, "learning_rate": 1.4618236137582157e-05, "loss": 1.0034, "step": 7220 }, { "epoch": 0.37, "learning_rate": 1.4616776637556544e-05, "loss": 0.9383, "step": 7221 }, { "epoch": 0.37, "learning_rate": 1.4615317012536618e-05, "loss": 0.997, "step": 7222 }, { "epoch": 0.37, "learning_rate": 1.4613857262561892e-05, "loss": 0.9547, "step": 7223 }, { "epoch": 0.37, "learning_rate": 1.461239738767189e-05, "loss": 0.9663, "step": 7224 }, { "epoch": 0.37, "learning_rate": 1.4610937387906137e-05, "loss": 0.9974, "step": 7225 }, { "epoch": 0.37, "learning_rate": 1.460947726330416e-05, "loss": 0.8844, "step": 7226 }, { "epoch": 0.37, "learning_rate": 1.4608017013905487e-05, "loss": 0.9207, "step": 7227 }, { "epoch": 0.37, "learning_rate": 1.4606556639749658e-05, "loss": 0.9421, "step": 7228 }, { "epoch": 0.37, "learning_rate": 1.460509614087621e-05, "loss": 0.9373, "step": 7229 }, { "epoch": 0.37, "learning_rate": 1.460363551732468e-05, "loss": 0.9748, "step": 7230 }, { "epoch": 0.37, "learning_rate": 1.460217476913462e-05, "loss": 1.0853, "step": 7231 }, { "epoch": 0.37, "learning_rate": 1.4600713896345574e-05, "loss": 0.8748, "step": 7232 }, { "epoch": 0.37, "learning_rate": 1.4599252898997092e-05, "loss": 0.9277, "step": 7233 }, { "epoch": 0.37, "learning_rate": 1.459779177712873e-05, "loss": 1.0123, "step": 7234 }, { "epoch": 0.37, "learning_rate": 1.4596330530780052e-05, "loss": 0.9398, "step": 7235 }, { "epoch": 0.37, "learning_rate": 1.4594869159990612e-05, "loss": 0.8684, "step": 7236 }, { "epoch": 0.37, "learning_rate": 1.459340766479998e-05, "loss": 0.9526, "step": 7237 }, { "epoch": 0.37, "learning_rate": 1.4591946045247718e-05, "loss": 0.9985, "step": 7238 }, { "epoch": 0.37, "learning_rate": 1.4590484301373406e-05, "loss": 0.9234, "step": 7239 }, { "epoch": 0.37, "learning_rate": 1.4589022433216616e-05, "loss": 0.8033, "step": 7240 }, { "epoch": 0.37, "learning_rate": 1.4587560440816923e-05, "loss": 0.9685, "step": 7241 }, { "epoch": 0.37, "learning_rate": 1.4586098324213913e-05, "loss": 0.9389, "step": 7242 }, { "epoch": 0.37, "learning_rate": 1.458463608344717e-05, "loss": 1.0034, "step": 7243 }, { "epoch": 0.37, "learning_rate": 1.4583173718556285e-05, "loss": 0.9528, "step": 7244 }, { "epoch": 0.37, "learning_rate": 1.4581711229580843e-05, "loss": 1.0175, "step": 7245 }, { "epoch": 0.37, "learning_rate": 1.4580248616560449e-05, "loss": 1.0015, "step": 7246 }, { "epoch": 0.37, "learning_rate": 1.4578785879534697e-05, "loss": 0.9857, "step": 7247 }, { "epoch": 0.37, "learning_rate": 1.4577323018543184e-05, "loss": 0.8649, "step": 7248 }, { "epoch": 0.37, "learning_rate": 1.4575860033625522e-05, "loss": 0.9336, "step": 7249 }, { "epoch": 0.37, "learning_rate": 1.4574396924821319e-05, "loss": 0.9643, "step": 7250 }, { "epoch": 0.37, "learning_rate": 1.4572933692170184e-05, "loss": 0.9723, "step": 7251 }, { "epoch": 0.37, "learning_rate": 1.4571470335711738e-05, "loss": 0.9114, "step": 7252 }, { "epoch": 0.37, "learning_rate": 1.4570006855485595e-05, "loss": 0.9808, "step": 7253 }, { "epoch": 0.37, "learning_rate": 1.4568543251531377e-05, "loss": 0.9369, "step": 7254 }, { "epoch": 0.37, "learning_rate": 1.456707952388871e-05, "loss": 0.9054, "step": 7255 }, { "epoch": 0.37, "learning_rate": 1.4565615672597226e-05, "loss": 0.9701, "step": 7256 }, { "epoch": 0.37, "learning_rate": 1.4564151697696556e-05, "loss": 0.8669, "step": 7257 }, { "epoch": 0.37, "learning_rate": 1.4562687599226333e-05, "loss": 0.9853, "step": 7258 }, { "epoch": 0.37, "learning_rate": 1.45612233772262e-05, "loss": 0.9308, "step": 7259 }, { "epoch": 0.37, "learning_rate": 1.4559759031735792e-05, "loss": 0.9735, "step": 7260 }, { "epoch": 0.37, "learning_rate": 1.4558294562794761e-05, "loss": 0.9135, "step": 7261 }, { "epoch": 0.37, "learning_rate": 1.4556829970442758e-05, "loss": 0.882, "step": 7262 }, { "epoch": 0.37, "learning_rate": 1.4555365254719428e-05, "loss": 0.9331, "step": 7263 }, { "epoch": 0.37, "learning_rate": 1.4553900415664432e-05, "loss": 1.0627, "step": 7264 }, { "epoch": 0.37, "learning_rate": 1.4552435453317428e-05, "loss": 0.9147, "step": 7265 }, { "epoch": 0.37, "learning_rate": 1.4550970367718076e-05, "loss": 0.9353, "step": 7266 }, { "epoch": 0.37, "learning_rate": 1.4549505158906041e-05, "loss": 1.0762, "step": 7267 }, { "epoch": 0.37, "learning_rate": 1.4548039826920999e-05, "loss": 0.9538, "step": 7268 }, { "epoch": 0.37, "learning_rate": 1.4546574371802613e-05, "loss": 0.9935, "step": 7269 }, { "epoch": 0.37, "learning_rate": 1.4545108793590565e-05, "loss": 0.9222, "step": 7270 }, { "epoch": 0.37, "learning_rate": 1.4543643092324535e-05, "loss": 1.0161, "step": 7271 }, { "epoch": 0.37, "learning_rate": 1.4542177268044198e-05, "loss": 0.9668, "step": 7272 }, { "epoch": 0.37, "learning_rate": 1.4540711320789242e-05, "loss": 0.9886, "step": 7273 }, { "epoch": 0.37, "learning_rate": 1.4539245250599362e-05, "loss": 0.8438, "step": 7274 }, { "epoch": 0.37, "learning_rate": 1.4537779057514248e-05, "loss": 1.0257, "step": 7275 }, { "epoch": 0.37, "learning_rate": 1.453631274157359e-05, "loss": 0.8849, "step": 7276 }, { "epoch": 0.37, "learning_rate": 1.4534846302817092e-05, "loss": 0.9419, "step": 7277 }, { "epoch": 0.37, "learning_rate": 1.4533379741284455e-05, "loss": 1.0245, "step": 7278 }, { "epoch": 0.37, "learning_rate": 1.4531913057015383e-05, "loss": 1.0443, "step": 7279 }, { "epoch": 0.37, "learning_rate": 1.4530446250049588e-05, "loss": 1.0181, "step": 7280 }, { "epoch": 0.37, "learning_rate": 1.4528979320426784e-05, "loss": 0.9445, "step": 7281 }, { "epoch": 0.37, "learning_rate": 1.4527512268186681e-05, "loss": 1.008, "step": 7282 }, { "epoch": 0.37, "learning_rate": 1.4526045093368998e-05, "loss": 0.8799, "step": 7283 }, { "epoch": 0.37, "learning_rate": 1.4524577796013461e-05, "loss": 0.9357, "step": 7284 }, { "epoch": 0.37, "learning_rate": 1.4523110376159794e-05, "loss": 0.9584, "step": 7285 }, { "epoch": 0.37, "learning_rate": 1.4521642833847724e-05, "loss": 0.9653, "step": 7286 }, { "epoch": 0.37, "learning_rate": 1.4520175169116987e-05, "loss": 1.0918, "step": 7287 }, { "epoch": 0.37, "learning_rate": 1.4518707382007317e-05, "loss": 1.0538, "step": 7288 }, { "epoch": 0.37, "learning_rate": 1.4517239472558454e-05, "loss": 0.8266, "step": 7289 }, { "epoch": 0.37, "learning_rate": 1.4515771440810133e-05, "loss": 0.9156, "step": 7290 }, { "epoch": 0.37, "learning_rate": 1.4514303286802108e-05, "loss": 0.8953, "step": 7291 }, { "epoch": 0.37, "learning_rate": 1.4512835010574122e-05, "loss": 0.9587, "step": 7292 }, { "epoch": 0.37, "learning_rate": 1.4511366612165932e-05, "loss": 0.9659, "step": 7293 }, { "epoch": 0.37, "learning_rate": 1.4509898091617292e-05, "loss": 0.9639, "step": 7294 }, { "epoch": 0.37, "learning_rate": 1.4508429448967953e-05, "loss": 0.963, "step": 7295 }, { "epoch": 0.37, "learning_rate": 1.450696068425769e-05, "loss": 0.9515, "step": 7296 }, { "epoch": 0.37, "learning_rate": 1.4505491797526257e-05, "loss": 0.9576, "step": 7297 }, { "epoch": 0.37, "learning_rate": 1.4504022788813427e-05, "loss": 0.9701, "step": 7298 }, { "epoch": 0.37, "learning_rate": 1.4502553658158975e-05, "loss": 0.9522, "step": 7299 }, { "epoch": 0.37, "learning_rate": 1.4501084405602672e-05, "loss": 0.8027, "step": 7300 }, { "epoch": 0.37, "learning_rate": 1.4499615031184297e-05, "loss": 0.8986, "step": 7301 }, { "epoch": 0.37, "learning_rate": 1.4498145534943633e-05, "loss": 0.9942, "step": 7302 }, { "epoch": 0.37, "learning_rate": 1.4496675916920465e-05, "loss": 0.9775, "step": 7303 }, { "epoch": 0.37, "learning_rate": 1.4495206177154577e-05, "loss": 1.042, "step": 7304 }, { "epoch": 0.37, "learning_rate": 1.4493736315685768e-05, "loss": 0.9625, "step": 7305 }, { "epoch": 0.37, "learning_rate": 1.4492266332553828e-05, "loss": 0.9301, "step": 7306 }, { "epoch": 0.37, "learning_rate": 1.4490796227798557e-05, "loss": 1.0456, "step": 7307 }, { "epoch": 0.37, "learning_rate": 1.4489326001459754e-05, "loss": 0.9926, "step": 7308 }, { "epoch": 0.37, "learning_rate": 1.4487855653577228e-05, "loss": 0.8575, "step": 7309 }, { "epoch": 0.37, "learning_rate": 1.448638518419078e-05, "loss": 1.0192, "step": 7310 }, { "epoch": 0.37, "learning_rate": 1.4484914593340231e-05, "loss": 0.8885, "step": 7311 }, { "epoch": 0.37, "learning_rate": 1.4483443881065393e-05, "loss": 0.9793, "step": 7312 }, { "epoch": 0.37, "learning_rate": 1.4481973047406077e-05, "loss": 0.9478, "step": 7313 }, { "epoch": 0.37, "learning_rate": 1.4480502092402111e-05, "loss": 0.9619, "step": 7314 }, { "epoch": 0.37, "learning_rate": 1.4479031016093319e-05, "loss": 0.894, "step": 7315 }, { "epoch": 0.37, "learning_rate": 1.4477559818519526e-05, "loss": 0.955, "step": 7316 }, { "epoch": 0.37, "learning_rate": 1.4476088499720566e-05, "loss": 0.9089, "step": 7317 }, { "epoch": 0.37, "learning_rate": 1.4474617059736272e-05, "loss": 0.8918, "step": 7318 }, { "epoch": 0.37, "learning_rate": 1.447314549860648e-05, "loss": 0.9226, "step": 7319 }, { "epoch": 0.37, "learning_rate": 1.4471673816371032e-05, "loss": 0.9126, "step": 7320 }, { "epoch": 0.37, "learning_rate": 1.4470202013069776e-05, "loss": 0.9278, "step": 7321 }, { "epoch": 0.37, "learning_rate": 1.4468730088742552e-05, "loss": 0.9669, "step": 7322 }, { "epoch": 0.37, "learning_rate": 1.446725804342922e-05, "loss": 0.9409, "step": 7323 }, { "epoch": 0.37, "learning_rate": 1.446578587716963e-05, "loss": 0.9741, "step": 7324 }, { "epoch": 0.37, "learning_rate": 1.4464313590003634e-05, "loss": 0.9446, "step": 7325 }, { "epoch": 0.37, "learning_rate": 1.4462841181971099e-05, "loss": 0.988, "step": 7326 }, { "epoch": 0.37, "learning_rate": 1.4461368653111887e-05, "loss": 0.9065, "step": 7327 }, { "epoch": 0.37, "learning_rate": 1.4459896003465865e-05, "loss": 0.9601, "step": 7328 }, { "epoch": 0.37, "learning_rate": 1.4458423233072906e-05, "loss": 1.0262, "step": 7329 }, { "epoch": 0.37, "learning_rate": 1.445695034197288e-05, "loss": 0.8744, "step": 7330 }, { "epoch": 0.37, "learning_rate": 1.4455477330205665e-05, "loss": 0.9555, "step": 7331 }, { "epoch": 0.37, "learning_rate": 1.445400419781114e-05, "loss": 0.9566, "step": 7332 }, { "epoch": 0.37, "learning_rate": 1.445253094482919e-05, "loss": 0.9755, "step": 7333 }, { "epoch": 0.37, "learning_rate": 1.4451057571299705e-05, "loss": 1.0035, "step": 7334 }, { "epoch": 0.37, "learning_rate": 1.444958407726257e-05, "loss": 0.9825, "step": 7335 }, { "epoch": 0.37, "learning_rate": 1.4448110462757681e-05, "loss": 0.9357, "step": 7336 }, { "epoch": 0.37, "learning_rate": 1.4446636727824932e-05, "loss": 0.912, "step": 7337 }, { "epoch": 0.37, "learning_rate": 1.4445162872504224e-05, "loss": 0.9128, "step": 7338 }, { "epoch": 0.37, "learning_rate": 1.4443688896835463e-05, "loss": 1.0423, "step": 7339 }, { "epoch": 0.37, "learning_rate": 1.4442214800858549e-05, "loss": 0.9549, "step": 7340 }, { "epoch": 0.37, "learning_rate": 1.44407405846134e-05, "loss": 0.9554, "step": 7341 }, { "epoch": 0.37, "learning_rate": 1.4439266248139924e-05, "loss": 0.8829, "step": 7342 }, { "epoch": 0.37, "learning_rate": 1.4437791791478035e-05, "loss": 0.8973, "step": 7343 }, { "epoch": 0.37, "learning_rate": 1.4436317214667655e-05, "loss": 0.9034, "step": 7344 }, { "epoch": 0.37, "learning_rate": 1.4434842517748704e-05, "loss": 0.9703, "step": 7345 }, { "epoch": 0.37, "learning_rate": 1.4433367700761114e-05, "loss": 0.9377, "step": 7346 }, { "epoch": 0.37, "learning_rate": 1.443189276374481e-05, "loss": 1.0397, "step": 7347 }, { "epoch": 0.37, "learning_rate": 1.4430417706739724e-05, "loss": 1.028, "step": 7348 }, { "epoch": 0.37, "learning_rate": 1.442894252978579e-05, "loss": 0.9507, "step": 7349 }, { "epoch": 0.37, "learning_rate": 1.4427467232922951e-05, "loss": 0.9318, "step": 7350 }, { "epoch": 0.37, "learning_rate": 1.4425991816191145e-05, "loss": 0.9559, "step": 7351 }, { "epoch": 0.37, "learning_rate": 1.4424516279630322e-05, "loss": 1.0601, "step": 7352 }, { "epoch": 0.37, "learning_rate": 1.442304062328043e-05, "loss": 0.9518, "step": 7353 }, { "epoch": 0.37, "learning_rate": 1.4421564847181414e-05, "loss": 1.0797, "step": 7354 }, { "epoch": 0.37, "learning_rate": 1.4420088951373237e-05, "loss": 0.9999, "step": 7355 }, { "epoch": 0.37, "learning_rate": 1.4418612935895852e-05, "loss": 0.93, "step": 7356 }, { "epoch": 0.37, "learning_rate": 1.4417136800789224e-05, "loss": 0.9012, "step": 7357 }, { "epoch": 0.37, "learning_rate": 1.4415660546093317e-05, "loss": 0.8167, "step": 7358 }, { "epoch": 0.37, "learning_rate": 1.44141841718481e-05, "loss": 0.9982, "step": 7359 }, { "epoch": 0.37, "learning_rate": 1.441270767809354e-05, "loss": 0.9433, "step": 7360 }, { "epoch": 0.37, "learning_rate": 1.4411231064869617e-05, "loss": 1.026, "step": 7361 }, { "epoch": 0.37, "learning_rate": 1.4409754332216303e-05, "loss": 0.9454, "step": 7362 }, { "epoch": 0.37, "learning_rate": 1.4408277480173582e-05, "loss": 0.9023, "step": 7363 }, { "epoch": 0.37, "learning_rate": 1.4406800508781442e-05, "loss": 0.9975, "step": 7364 }, { "epoch": 0.37, "learning_rate": 1.4405323418079866e-05, "loss": 0.9272, "step": 7365 }, { "epoch": 0.37, "learning_rate": 1.4403846208108843e-05, "loss": 0.9729, "step": 7366 }, { "epoch": 0.37, "learning_rate": 1.4402368878908371e-05, "loss": 0.9517, "step": 7367 }, { "epoch": 0.37, "learning_rate": 1.4400891430518444e-05, "loss": 0.9978, "step": 7368 }, { "epoch": 0.37, "learning_rate": 1.4399413862979064e-05, "loss": 0.9512, "step": 7369 }, { "epoch": 0.37, "learning_rate": 1.4397936176330237e-05, "loss": 1.0463, "step": 7370 }, { "epoch": 0.37, "learning_rate": 1.4396458370611963e-05, "loss": 1.0325, "step": 7371 }, { "epoch": 0.37, "learning_rate": 1.4394980445864258e-05, "loss": 0.842, "step": 7372 }, { "epoch": 0.37, "learning_rate": 1.4393502402127132e-05, "loss": 0.9748, "step": 7373 }, { "epoch": 0.37, "learning_rate": 1.4392024239440603e-05, "loss": 0.9252, "step": 7374 }, { "epoch": 0.37, "learning_rate": 1.4390545957844691e-05, "loss": 0.8818, "step": 7375 }, { "epoch": 0.37, "learning_rate": 1.4389067557379422e-05, "loss": 0.9862, "step": 7376 }, { "epoch": 0.37, "learning_rate": 1.4387589038084815e-05, "loss": 0.9937, "step": 7377 }, { "epoch": 0.37, "learning_rate": 1.4386110400000901e-05, "loss": 0.8845, "step": 7378 }, { "epoch": 0.37, "learning_rate": 1.4384631643167715e-05, "loss": 0.9035, "step": 7379 }, { "epoch": 0.37, "learning_rate": 1.4383152767625294e-05, "loss": 0.9117, "step": 7380 }, { "epoch": 0.37, "learning_rate": 1.4381673773413676e-05, "loss": 0.9807, "step": 7381 }, { "epoch": 0.38, "learning_rate": 1.43801946605729e-05, "loss": 0.9551, "step": 7382 }, { "epoch": 0.38, "learning_rate": 1.4378715429143015e-05, "loss": 0.9272, "step": 7383 }, { "epoch": 0.38, "learning_rate": 1.4377236079164067e-05, "loss": 1.0186, "step": 7384 }, { "epoch": 0.38, "learning_rate": 1.437575661067611e-05, "loss": 0.8921, "step": 7385 }, { "epoch": 0.38, "learning_rate": 1.4374277023719199e-05, "loss": 0.903, "step": 7386 }, { "epoch": 0.38, "learning_rate": 1.4372797318333392e-05, "loss": 0.9719, "step": 7387 }, { "epoch": 0.38, "learning_rate": 1.4371317494558747e-05, "loss": 1.0207, "step": 7388 }, { "epoch": 0.38, "learning_rate": 1.4369837552435336e-05, "loss": 1.0506, "step": 7389 }, { "epoch": 0.38, "learning_rate": 1.4368357492003218e-05, "loss": 0.8871, "step": 7390 }, { "epoch": 0.38, "learning_rate": 1.436687731330247e-05, "loss": 0.9235, "step": 7391 }, { "epoch": 0.38, "learning_rate": 1.4365397016373165e-05, "loss": 0.9589, "step": 7392 }, { "epoch": 0.38, "learning_rate": 1.4363916601255379e-05, "loss": 1.0403, "step": 7393 }, { "epoch": 0.38, "learning_rate": 1.4362436067989197e-05, "loss": 0.8617, "step": 7394 }, { "epoch": 0.38, "learning_rate": 1.4360955416614699e-05, "loss": 0.9447, "step": 7395 }, { "epoch": 0.38, "learning_rate": 1.435947464717197e-05, "loss": 0.9408, "step": 7396 }, { "epoch": 0.38, "learning_rate": 1.4357993759701104e-05, "loss": 0.9753, "step": 7397 }, { "epoch": 0.38, "learning_rate": 1.4356512754242194e-05, "loss": 0.959, "step": 7398 }, { "epoch": 0.38, "learning_rate": 1.4355031630835337e-05, "loss": 0.9456, "step": 7399 }, { "epoch": 0.38, "learning_rate": 1.435355038952063e-05, "loss": 0.9848, "step": 7400 }, { "epoch": 0.38, "learning_rate": 1.435206903033818e-05, "loss": 0.9369, "step": 7401 }, { "epoch": 0.38, "learning_rate": 1.435058755332809e-05, "loss": 0.9219, "step": 7402 }, { "epoch": 0.38, "learning_rate": 1.4349105958530469e-05, "loss": 0.9577, "step": 7403 }, { "epoch": 0.38, "learning_rate": 1.4347624245985436e-05, "loss": 1.0103, "step": 7404 }, { "epoch": 0.38, "learning_rate": 1.43461424157331e-05, "loss": 0.9551, "step": 7405 }, { "epoch": 0.38, "learning_rate": 1.434466046781358e-05, "loss": 1.0154, "step": 7406 }, { "epoch": 0.38, "learning_rate": 1.4343178402267004e-05, "loss": 0.9573, "step": 7407 }, { "epoch": 0.38, "learning_rate": 1.434169621913349e-05, "loss": 0.9503, "step": 7408 }, { "epoch": 0.38, "learning_rate": 1.434021391845317e-05, "loss": 1.0579, "step": 7409 }, { "epoch": 0.38, "learning_rate": 1.4338731500266177e-05, "loss": 0.9101, "step": 7410 }, { "epoch": 0.38, "learning_rate": 1.4337248964612647e-05, "loss": 0.9564, "step": 7411 }, { "epoch": 0.38, "learning_rate": 1.433576631153271e-05, "loss": 0.9412, "step": 7412 }, { "epoch": 0.38, "learning_rate": 1.4334283541066516e-05, "loss": 0.9577, "step": 7413 }, { "epoch": 0.38, "learning_rate": 1.4332800653254209e-05, "loss": 0.8508, "step": 7414 }, { "epoch": 0.38, "learning_rate": 1.433131764813593e-05, "loss": 1.002, "step": 7415 }, { "epoch": 0.38, "learning_rate": 1.4329834525751835e-05, "loss": 0.9134, "step": 7416 }, { "epoch": 0.38, "learning_rate": 1.4328351286142081e-05, "loss": 0.9438, "step": 7417 }, { "epoch": 0.38, "learning_rate": 1.4326867929346814e-05, "loss": 0.998, "step": 7418 }, { "epoch": 0.38, "learning_rate": 1.4325384455406207e-05, "loss": 1.1011, "step": 7419 }, { "epoch": 0.38, "learning_rate": 1.4323900864360414e-05, "loss": 0.9165, "step": 7420 }, { "epoch": 0.38, "learning_rate": 1.4322417156249606e-05, "loss": 0.9395, "step": 7421 }, { "epoch": 0.38, "learning_rate": 1.4320933331113953e-05, "loss": 0.9087, "step": 7422 }, { "epoch": 0.38, "learning_rate": 1.431944938899363e-05, "loss": 1.0904, "step": 7423 }, { "epoch": 0.38, "learning_rate": 1.4317965329928806e-05, "loss": 0.9855, "step": 7424 }, { "epoch": 0.38, "learning_rate": 1.4316481153959665e-05, "loss": 0.9216, "step": 7425 }, { "epoch": 0.38, "learning_rate": 1.4314996861126391e-05, "loss": 0.973, "step": 7426 }, { "epoch": 0.38, "learning_rate": 1.4313512451469166e-05, "loss": 0.9275, "step": 7427 }, { "epoch": 0.38, "learning_rate": 1.431202792502818e-05, "loss": 0.9528, "step": 7428 }, { "epoch": 0.38, "learning_rate": 1.4310543281843632e-05, "loss": 0.8695, "step": 7429 }, { "epoch": 0.38, "learning_rate": 1.4309058521955704e-05, "loss": 1.0011, "step": 7430 }, { "epoch": 0.38, "learning_rate": 1.4307573645404602e-05, "loss": 0.9483, "step": 7431 }, { "epoch": 0.38, "learning_rate": 1.4306088652230527e-05, "loss": 0.9109, "step": 7432 }, { "epoch": 0.38, "learning_rate": 1.4304603542473684e-05, "loss": 0.8826, "step": 7433 }, { "epoch": 0.38, "learning_rate": 1.4303118316174278e-05, "loss": 1.0283, "step": 7434 }, { "epoch": 0.38, "learning_rate": 1.4301632973372526e-05, "loss": 1.0157, "step": 7435 }, { "epoch": 0.38, "learning_rate": 1.4300147514108634e-05, "loss": 0.981, "step": 7436 }, { "epoch": 0.38, "learning_rate": 1.4298661938422823e-05, "loss": 0.9308, "step": 7437 }, { "epoch": 0.38, "learning_rate": 1.4297176246355315e-05, "loss": 0.8687, "step": 7438 }, { "epoch": 0.38, "learning_rate": 1.429569043794633e-05, "loss": 0.8877, "step": 7439 }, { "epoch": 0.38, "learning_rate": 1.4294204513236098e-05, "loss": 0.8727, "step": 7440 }, { "epoch": 0.38, "learning_rate": 1.429271847226485e-05, "loss": 1.0685, "step": 7441 }, { "epoch": 0.38, "learning_rate": 1.4291232315072811e-05, "loss": 0.9842, "step": 7442 }, { "epoch": 0.38, "learning_rate": 1.4289746041700225e-05, "loss": 0.9251, "step": 7443 }, { "epoch": 0.38, "learning_rate": 1.428825965218733e-05, "loss": 0.8385, "step": 7444 }, { "epoch": 0.38, "learning_rate": 1.4286773146574367e-05, "loss": 0.9708, "step": 7445 }, { "epoch": 0.38, "learning_rate": 1.4285286524901581e-05, "loss": 0.9083, "step": 7446 }, { "epoch": 0.38, "learning_rate": 1.4283799787209222e-05, "loss": 1.0707, "step": 7447 }, { "epoch": 0.38, "learning_rate": 1.428231293353754e-05, "loss": 0.9369, "step": 7448 }, { "epoch": 0.38, "learning_rate": 1.4280825963926793e-05, "loss": 0.876, "step": 7449 }, { "epoch": 0.38, "learning_rate": 1.4279338878417237e-05, "loss": 0.8768, "step": 7450 }, { "epoch": 0.38, "learning_rate": 1.4277851677049132e-05, "loss": 0.8443, "step": 7451 }, { "epoch": 0.38, "learning_rate": 1.4276364359862748e-05, "loss": 1.0898, "step": 7452 }, { "epoch": 0.38, "learning_rate": 1.4274876926898346e-05, "loss": 0.8421, "step": 7453 }, { "epoch": 0.38, "learning_rate": 1.42733893781962e-05, "loss": 0.9074, "step": 7454 }, { "epoch": 0.38, "learning_rate": 1.4271901713796583e-05, "loss": 1.027, "step": 7455 }, { "epoch": 0.38, "learning_rate": 1.427041393373977e-05, "loss": 1.0112, "step": 7456 }, { "epoch": 0.38, "learning_rate": 1.4268926038066046e-05, "loss": 1.0405, "step": 7457 }, { "epoch": 0.38, "learning_rate": 1.4267438026815692e-05, "loss": 0.927, "step": 7458 }, { "epoch": 0.38, "learning_rate": 1.426594990002899e-05, "loss": 1.0753, "step": 7459 }, { "epoch": 0.38, "learning_rate": 1.4264461657746237e-05, "loss": 1.0075, "step": 7460 }, { "epoch": 0.38, "learning_rate": 1.4262973300007718e-05, "loss": 0.999, "step": 7461 }, { "epoch": 0.38, "learning_rate": 1.4261484826853736e-05, "loss": 0.9705, "step": 7462 }, { "epoch": 0.38, "learning_rate": 1.4259996238324583e-05, "loss": 1.0104, "step": 7463 }, { "epoch": 0.38, "learning_rate": 1.4258507534460568e-05, "loss": 1.0128, "step": 7464 }, { "epoch": 0.38, "learning_rate": 1.4257018715301989e-05, "loss": 0.9422, "step": 7465 }, { "epoch": 0.38, "learning_rate": 1.4255529780889156e-05, "loss": 0.8336, "step": 7466 }, { "epoch": 0.38, "learning_rate": 1.4254040731262384e-05, "loss": 0.9105, "step": 7467 }, { "epoch": 0.38, "learning_rate": 1.4252551566461983e-05, "loss": 0.8895, "step": 7468 }, { "epoch": 0.38, "learning_rate": 1.4251062286528277e-05, "loss": 0.946, "step": 7469 }, { "epoch": 0.38, "learning_rate": 1.4249572891501582e-05, "loss": 0.9464, "step": 7470 }, { "epoch": 0.38, "learning_rate": 1.4248083381422218e-05, "loss": 1.0587, "step": 7471 }, { "epoch": 0.38, "learning_rate": 1.424659375633052e-05, "loss": 0.9528, "step": 7472 }, { "epoch": 0.38, "learning_rate": 1.4245104016266812e-05, "loss": 0.9342, "step": 7473 }, { "epoch": 0.38, "learning_rate": 1.4243614161271426e-05, "loss": 1.0344, "step": 7474 }, { "epoch": 0.38, "learning_rate": 1.4242124191384705e-05, "loss": 0.9221, "step": 7475 }, { "epoch": 0.38, "learning_rate": 1.4240634106646986e-05, "loss": 0.9599, "step": 7476 }, { "epoch": 0.38, "learning_rate": 1.423914390709861e-05, "loss": 0.8693, "step": 7477 }, { "epoch": 0.38, "learning_rate": 1.423765359277992e-05, "loss": 0.9891, "step": 7478 }, { "epoch": 0.38, "learning_rate": 1.4236163163731269e-05, "loss": 0.9231, "step": 7479 }, { "epoch": 0.38, "learning_rate": 1.4234672619993006e-05, "loss": 0.9255, "step": 7480 }, { "epoch": 0.38, "learning_rate": 1.423318196160549e-05, "loss": 0.941, "step": 7481 }, { "epoch": 0.38, "learning_rate": 1.4231691188609075e-05, "loss": 0.9344, "step": 7482 }, { "epoch": 0.38, "learning_rate": 1.4230200301044123e-05, "loss": 1.0051, "step": 7483 }, { "epoch": 0.38, "learning_rate": 1.4228709298950998e-05, "loss": 0.9255, "step": 7484 }, { "epoch": 0.38, "learning_rate": 1.4227218182370069e-05, "loss": 0.9523, "step": 7485 }, { "epoch": 0.38, "learning_rate": 1.4225726951341703e-05, "loss": 0.9984, "step": 7486 }, { "epoch": 0.38, "learning_rate": 1.4224235605906278e-05, "loss": 0.941, "step": 7487 }, { "epoch": 0.38, "learning_rate": 1.4222744146104166e-05, "loss": 0.9318, "step": 7488 }, { "epoch": 0.38, "learning_rate": 1.422125257197575e-05, "loss": 0.8894, "step": 7489 }, { "epoch": 0.38, "learning_rate": 1.421976088356141e-05, "loss": 1.0197, "step": 7490 }, { "epoch": 0.38, "learning_rate": 1.4218269080901537e-05, "loss": 1.0424, "step": 7491 }, { "epoch": 0.38, "learning_rate": 1.4216777164036512e-05, "loss": 0.9592, "step": 7492 }, { "epoch": 0.38, "learning_rate": 1.4215285133006736e-05, "loss": 0.9369, "step": 7493 }, { "epoch": 0.38, "learning_rate": 1.4213792987852599e-05, "loss": 0.9075, "step": 7494 }, { "epoch": 0.38, "learning_rate": 1.4212300728614498e-05, "loss": 0.9532, "step": 7495 }, { "epoch": 0.38, "learning_rate": 1.4210808355332837e-05, "loss": 0.8649, "step": 7496 }, { "epoch": 0.38, "learning_rate": 1.4209315868048022e-05, "loss": 0.916, "step": 7497 }, { "epoch": 0.38, "learning_rate": 1.4207823266800453e-05, "loss": 0.9154, "step": 7498 }, { "epoch": 0.38, "learning_rate": 1.4206330551630551e-05, "loss": 0.9175, "step": 7499 }, { "epoch": 0.38, "learning_rate": 1.4204837722578724e-05, "loss": 1.0081, "step": 7500 }, { "epoch": 0.38, "learning_rate": 1.4203344779685387e-05, "loss": 0.9888, "step": 7501 }, { "epoch": 0.38, "learning_rate": 1.420185172299096e-05, "loss": 0.9061, "step": 7502 }, { "epoch": 0.38, "learning_rate": 1.4200358552535874e-05, "loss": 0.8831, "step": 7503 }, { "epoch": 0.38, "learning_rate": 1.4198865268360544e-05, "loss": 1.0029, "step": 7504 }, { "epoch": 0.38, "learning_rate": 1.4197371870505407e-05, "loss": 0.8427, "step": 7505 }, { "epoch": 0.38, "learning_rate": 1.4195878359010894e-05, "loss": 1.0108, "step": 7506 }, { "epoch": 0.38, "learning_rate": 1.4194384733917432e-05, "loss": 0.9238, "step": 7507 }, { "epoch": 0.38, "learning_rate": 1.419289099526547e-05, "loss": 0.9145, "step": 7508 }, { "epoch": 0.38, "learning_rate": 1.4191397143095445e-05, "loss": 1.1221, "step": 7509 }, { "epoch": 0.38, "learning_rate": 1.4189903177447801e-05, "loss": 0.9648, "step": 7510 }, { "epoch": 0.38, "learning_rate": 1.4188409098362988e-05, "loss": 0.9045, "step": 7511 }, { "epoch": 0.38, "learning_rate": 1.4186914905881452e-05, "loss": 0.9897, "step": 7512 }, { "epoch": 0.38, "learning_rate": 1.4185420600043652e-05, "loss": 0.9708, "step": 7513 }, { "epoch": 0.38, "learning_rate": 1.418392618089004e-05, "loss": 1.0367, "step": 7514 }, { "epoch": 0.38, "learning_rate": 1.4182431648461079e-05, "loss": 0.9707, "step": 7515 }, { "epoch": 0.38, "learning_rate": 1.4180937002797231e-05, "loss": 0.973, "step": 7516 }, { "epoch": 0.38, "learning_rate": 1.4179442243938962e-05, "loss": 0.9504, "step": 7517 }, { "epoch": 0.38, "learning_rate": 1.417794737192674e-05, "loss": 0.9827, "step": 7518 }, { "epoch": 0.38, "learning_rate": 1.4176452386801039e-05, "loss": 1.0956, "step": 7519 }, { "epoch": 0.38, "learning_rate": 1.4174957288602331e-05, "loss": 0.9399, "step": 7520 }, { "epoch": 0.38, "learning_rate": 1.4173462077371096e-05, "loss": 0.9317, "step": 7521 }, { "epoch": 0.38, "learning_rate": 1.4171966753147818e-05, "loss": 0.9204, "step": 7522 }, { "epoch": 0.38, "learning_rate": 1.417047131597298e-05, "loss": 0.9241, "step": 7523 }, { "epoch": 0.38, "learning_rate": 1.4168975765887065e-05, "loss": 1.0745, "step": 7524 }, { "epoch": 0.38, "learning_rate": 1.416748010293057e-05, "loss": 0.8311, "step": 7525 }, { "epoch": 0.38, "learning_rate": 1.416598432714398e-05, "loss": 0.9484, "step": 7526 }, { "epoch": 0.38, "learning_rate": 1.4164488438567799e-05, "loss": 0.9478, "step": 7527 }, { "epoch": 0.38, "learning_rate": 1.4162992437242523e-05, "loss": 0.8857, "step": 7528 }, { "epoch": 0.38, "learning_rate": 1.4161496323208659e-05, "loss": 0.9751, "step": 7529 }, { "epoch": 0.38, "learning_rate": 1.4160000096506707e-05, "loss": 0.9726, "step": 7530 }, { "epoch": 0.38, "learning_rate": 1.4158503757177182e-05, "loss": 1.0745, "step": 7531 }, { "epoch": 0.38, "learning_rate": 1.4157007305260585e-05, "loss": 0.8841, "step": 7532 }, { "epoch": 0.38, "learning_rate": 1.4155510740797444e-05, "loss": 1.0085, "step": 7533 }, { "epoch": 0.38, "learning_rate": 1.415401406382827e-05, "loss": 0.9601, "step": 7534 }, { "epoch": 0.38, "learning_rate": 1.4152517274393584e-05, "loss": 0.923, "step": 7535 }, { "epoch": 0.38, "learning_rate": 1.415102037253391e-05, "loss": 0.9967, "step": 7536 }, { "epoch": 0.38, "learning_rate": 1.4149523358289777e-05, "loss": 0.9701, "step": 7537 }, { "epoch": 0.38, "learning_rate": 1.414802623170171e-05, "loss": 0.9587, "step": 7538 }, { "epoch": 0.38, "learning_rate": 1.4146528992810249e-05, "loss": 0.9037, "step": 7539 }, { "epoch": 0.38, "learning_rate": 1.4145031641655929e-05, "loss": 0.9319, "step": 7540 }, { "epoch": 0.38, "learning_rate": 1.4143534178279286e-05, "loss": 0.9048, "step": 7541 }, { "epoch": 0.38, "learning_rate": 1.414203660272086e-05, "loss": 0.9148, "step": 7542 }, { "epoch": 0.38, "learning_rate": 1.4140538915021204e-05, "loss": 1.0233, "step": 7543 }, { "epoch": 0.38, "learning_rate": 1.413904111522086e-05, "loss": 0.9003, "step": 7544 }, { "epoch": 0.38, "learning_rate": 1.4137543203360382e-05, "loss": 0.998, "step": 7545 }, { "epoch": 0.38, "learning_rate": 1.4136045179480325e-05, "loss": 0.9349, "step": 7546 }, { "epoch": 0.38, "learning_rate": 1.4134547043621245e-05, "loss": 0.9492, "step": 7547 }, { "epoch": 0.38, "learning_rate": 1.4133048795823698e-05, "loss": 0.9689, "step": 7548 }, { "epoch": 0.38, "learning_rate": 1.4131550436128259e-05, "loss": 0.9574, "step": 7549 }, { "epoch": 0.38, "learning_rate": 1.4130051964575482e-05, "loss": 0.9646, "step": 7550 }, { "epoch": 0.38, "learning_rate": 1.4128553381205943e-05, "loss": 0.9651, "step": 7551 }, { "epoch": 0.38, "learning_rate": 1.4127054686060216e-05, "loss": 0.9562, "step": 7552 }, { "epoch": 0.38, "learning_rate": 1.4125555879178874e-05, "loss": 0.9237, "step": 7553 }, { "epoch": 0.38, "learning_rate": 1.4124056960602495e-05, "loss": 0.8719, "step": 7554 }, { "epoch": 0.38, "learning_rate": 1.4122557930371658e-05, "loss": 1.0063, "step": 7555 }, { "epoch": 0.38, "learning_rate": 1.4121058788526958e-05, "loss": 0.9039, "step": 7556 }, { "epoch": 0.38, "learning_rate": 1.411955953510897e-05, "loss": 0.825, "step": 7557 }, { "epoch": 0.38, "learning_rate": 1.4118060170158294e-05, "loss": 0.929, "step": 7558 }, { "epoch": 0.38, "learning_rate": 1.411656069371552e-05, "loss": 0.9915, "step": 7559 }, { "epoch": 0.38, "learning_rate": 1.4115061105821242e-05, "loss": 0.9173, "step": 7560 }, { "epoch": 0.38, "learning_rate": 1.4113561406516063e-05, "loss": 0.9049, "step": 7561 }, { "epoch": 0.38, "learning_rate": 1.411206159584059e-05, "loss": 0.8182, "step": 7562 }, { "epoch": 0.38, "learning_rate": 1.411056167383542e-05, "loss": 0.9865, "step": 7563 }, { "epoch": 0.38, "learning_rate": 1.4109061640541167e-05, "loss": 0.9313, "step": 7564 }, { "epoch": 0.38, "learning_rate": 1.4107561495998443e-05, "loss": 0.8912, "step": 7565 }, { "epoch": 0.38, "learning_rate": 1.410606124024786e-05, "loss": 0.9843, "step": 7566 }, { "epoch": 0.38, "learning_rate": 1.4104560873330035e-05, "loss": 0.8954, "step": 7567 }, { "epoch": 0.38, "learning_rate": 1.4103060395285595e-05, "loss": 0.9595, "step": 7568 }, { "epoch": 0.38, "learning_rate": 1.4101559806155158e-05, "loss": 0.9122, "step": 7569 }, { "epoch": 0.38, "learning_rate": 1.4100059105979355e-05, "loss": 0.9163, "step": 7570 }, { "epoch": 0.38, "learning_rate": 1.4098558294798814e-05, "loss": 0.9251, "step": 7571 }, { "epoch": 0.38, "learning_rate": 1.4097057372654166e-05, "loss": 0.9876, "step": 7572 }, { "epoch": 0.38, "learning_rate": 1.4095556339586048e-05, "loss": 0.9878, "step": 7573 }, { "epoch": 0.38, "learning_rate": 1.4094055195635102e-05, "loss": 0.9246, "step": 7574 }, { "epoch": 0.38, "learning_rate": 1.4092553940841964e-05, "loss": 0.925, "step": 7575 }, { "epoch": 0.38, "learning_rate": 1.4091052575247286e-05, "loss": 0.9178, "step": 7576 }, { "epoch": 0.38, "learning_rate": 1.4089551098891709e-05, "loss": 0.964, "step": 7577 }, { "epoch": 0.38, "learning_rate": 1.408804951181589e-05, "loss": 0.9689, "step": 7578 }, { "epoch": 0.39, "learning_rate": 1.4086547814060475e-05, "loss": 0.8877, "step": 7579 }, { "epoch": 0.39, "learning_rate": 1.408504600566613e-05, "loss": 0.9123, "step": 7580 }, { "epoch": 0.39, "learning_rate": 1.4083544086673509e-05, "loss": 0.9755, "step": 7581 }, { "epoch": 0.39, "learning_rate": 1.4082042057123277e-05, "loss": 0.9434, "step": 7582 }, { "epoch": 0.39, "learning_rate": 1.4080539917056099e-05, "loss": 1.0143, "step": 7583 }, { "epoch": 0.39, "learning_rate": 1.4079037666512643e-05, "loss": 0.9054, "step": 7584 }, { "epoch": 0.39, "learning_rate": 1.4077535305533582e-05, "loss": 0.9533, "step": 7585 }, { "epoch": 0.39, "learning_rate": 1.407603283415959e-05, "loss": 0.9233, "step": 7586 }, { "epoch": 0.39, "learning_rate": 1.4074530252431349e-05, "loss": 0.9894, "step": 7587 }, { "epoch": 0.39, "learning_rate": 1.4073027560389533e-05, "loss": 0.9582, "step": 7588 }, { "epoch": 0.39, "learning_rate": 1.4071524758074833e-05, "loss": 1.0251, "step": 7589 }, { "epoch": 0.39, "learning_rate": 1.407002184552793e-05, "loss": 1.0159, "step": 7590 }, { "epoch": 0.39, "learning_rate": 1.4068518822789514e-05, "loss": 1.0157, "step": 7591 }, { "epoch": 0.39, "learning_rate": 1.4067015689900279e-05, "loss": 1.0299, "step": 7592 }, { "epoch": 0.39, "learning_rate": 1.4065512446900924e-05, "loss": 0.9375, "step": 7593 }, { "epoch": 0.39, "learning_rate": 1.4064009093832146e-05, "loss": 1.002, "step": 7594 }, { "epoch": 0.39, "learning_rate": 1.4062505630734643e-05, "loss": 0.9794, "step": 7595 }, { "epoch": 0.39, "learning_rate": 1.4061002057649122e-05, "loss": 0.8963, "step": 7596 }, { "epoch": 0.39, "learning_rate": 1.405949837461629e-05, "loss": 0.8193, "step": 7597 }, { "epoch": 0.39, "learning_rate": 1.405799458167686e-05, "loss": 0.919, "step": 7598 }, { "epoch": 0.39, "learning_rate": 1.4056490678871545e-05, "loss": 1.0008, "step": 7599 }, { "epoch": 0.39, "learning_rate": 1.405498666624106e-05, "loss": 0.9083, "step": 7600 }, { "epoch": 0.39, "learning_rate": 1.4053482543826122e-05, "loss": 0.9314, "step": 7601 }, { "epoch": 0.39, "learning_rate": 1.405197831166746e-05, "loss": 0.9425, "step": 7602 }, { "epoch": 0.39, "learning_rate": 1.4050473969805795e-05, "loss": 0.9757, "step": 7603 }, { "epoch": 0.39, "learning_rate": 1.4048969518281856e-05, "loss": 0.9434, "step": 7604 }, { "epoch": 0.39, "learning_rate": 1.4047464957136379e-05, "loss": 0.9826, "step": 7605 }, { "epoch": 0.39, "learning_rate": 1.4045960286410093e-05, "loss": 0.9333, "step": 7606 }, { "epoch": 0.39, "learning_rate": 1.4044455506143732e-05, "loss": 1.0205, "step": 7607 }, { "epoch": 0.39, "learning_rate": 1.4042950616378048e-05, "loss": 0.8987, "step": 7608 }, { "epoch": 0.39, "learning_rate": 1.4041445617153773e-05, "loss": 0.9636, "step": 7609 }, { "epoch": 0.39, "learning_rate": 1.4039940508511658e-05, "loss": 1.0194, "step": 7610 }, { "epoch": 0.39, "learning_rate": 1.4038435290492453e-05, "loss": 0.8906, "step": 7611 }, { "epoch": 0.39, "learning_rate": 1.4036929963136909e-05, "loss": 0.8787, "step": 7612 }, { "epoch": 0.39, "learning_rate": 1.4035424526485782e-05, "loss": 0.9171, "step": 7613 }, { "epoch": 0.39, "learning_rate": 1.4033918980579828e-05, "loss": 0.9811, "step": 7614 }, { "epoch": 0.39, "learning_rate": 1.4032413325459808e-05, "loss": 0.8143, "step": 7615 }, { "epoch": 0.39, "learning_rate": 1.4030907561166489e-05, "loss": 0.8693, "step": 7616 }, { "epoch": 0.39, "learning_rate": 1.4029401687740639e-05, "loss": 0.9259, "step": 7617 }, { "epoch": 0.39, "learning_rate": 1.4027895705223023e-05, "loss": 0.9346, "step": 7618 }, { "epoch": 0.39, "learning_rate": 1.4026389613654411e-05, "loss": 0.919, "step": 7619 }, { "epoch": 0.39, "learning_rate": 1.4024883413075592e-05, "loss": 0.9146, "step": 7620 }, { "epoch": 0.39, "learning_rate": 1.4023377103527332e-05, "loss": 0.9062, "step": 7621 }, { "epoch": 0.39, "learning_rate": 1.4021870685050419e-05, "loss": 0.9053, "step": 7622 }, { "epoch": 0.39, "learning_rate": 1.4020364157685636e-05, "loss": 0.9941, "step": 7623 }, { "epoch": 0.39, "learning_rate": 1.401885752147377e-05, "loss": 0.899, "step": 7624 }, { "epoch": 0.39, "learning_rate": 1.4017350776455613e-05, "loss": 0.9923, "step": 7625 }, { "epoch": 0.39, "learning_rate": 1.4015843922671956e-05, "loss": 0.9901, "step": 7626 }, { "epoch": 0.39, "learning_rate": 1.4014336960163602e-05, "loss": 0.9589, "step": 7627 }, { "epoch": 0.39, "learning_rate": 1.4012829888971343e-05, "loss": 0.8915, "step": 7628 }, { "epoch": 0.39, "learning_rate": 1.4011322709135984e-05, "loss": 1.045, "step": 7629 }, { "epoch": 0.39, "learning_rate": 1.400981542069833e-05, "loss": 0.9153, "step": 7630 }, { "epoch": 0.39, "learning_rate": 1.4008308023699189e-05, "loss": 0.9546, "step": 7631 }, { "epoch": 0.39, "learning_rate": 1.4006800518179373e-05, "loss": 0.9368, "step": 7632 }, { "epoch": 0.39, "learning_rate": 1.4005292904179696e-05, "loss": 1.0312, "step": 7633 }, { "epoch": 0.39, "learning_rate": 1.4003785181740977e-05, "loss": 0.9242, "step": 7634 }, { "epoch": 0.39, "learning_rate": 1.4002277350904032e-05, "loss": 0.86, "step": 7635 }, { "epoch": 0.39, "learning_rate": 1.4000769411709687e-05, "loss": 0.9206, "step": 7636 }, { "epoch": 0.39, "learning_rate": 1.3999261364198766e-05, "loss": 0.945, "step": 7637 }, { "epoch": 0.39, "learning_rate": 1.3997753208412097e-05, "loss": 0.8611, "step": 7638 }, { "epoch": 0.39, "learning_rate": 1.3996244944390516e-05, "loss": 1.0498, "step": 7639 }, { "epoch": 0.39, "learning_rate": 1.3994736572174852e-05, "loss": 1.011, "step": 7640 }, { "epoch": 0.39, "learning_rate": 1.3993228091805949e-05, "loss": 0.9724, "step": 7641 }, { "epoch": 0.39, "learning_rate": 1.3991719503324642e-05, "loss": 0.9423, "step": 7642 }, { "epoch": 0.39, "learning_rate": 1.3990210806771773e-05, "loss": 0.9872, "step": 7643 }, { "epoch": 0.39, "learning_rate": 1.3988702002188195e-05, "loss": 0.9155, "step": 7644 }, { "epoch": 0.39, "learning_rate": 1.3987193089614754e-05, "loss": 0.9012, "step": 7645 }, { "epoch": 0.39, "learning_rate": 1.39856840690923e-05, "loss": 0.8862, "step": 7646 }, { "epoch": 0.39, "learning_rate": 1.3984174940661692e-05, "loss": 0.8823, "step": 7647 }, { "epoch": 0.39, "learning_rate": 1.3982665704363786e-05, "loss": 0.9567, "step": 7648 }, { "epoch": 0.39, "learning_rate": 1.3981156360239445e-05, "loss": 1.0249, "step": 7649 }, { "epoch": 0.39, "learning_rate": 1.3979646908329527e-05, "loss": 0.9997, "step": 7650 }, { "epoch": 0.39, "learning_rate": 1.3978137348674908e-05, "loss": 0.9488, "step": 7651 }, { "epoch": 0.39, "learning_rate": 1.3976627681316448e-05, "loss": 0.9465, "step": 7652 }, { "epoch": 0.39, "learning_rate": 1.3975117906295029e-05, "loss": 0.9432, "step": 7653 }, { "epoch": 0.39, "learning_rate": 1.397360802365152e-05, "loss": 1.0708, "step": 7654 }, { "epoch": 0.39, "learning_rate": 1.3972098033426799e-05, "loss": 0.9299, "step": 7655 }, { "epoch": 0.39, "learning_rate": 1.3970587935661751e-05, "loss": 0.8777, "step": 7656 }, { "epoch": 0.39, "learning_rate": 1.3969077730397258e-05, "loss": 0.9927, "step": 7657 }, { "epoch": 0.39, "learning_rate": 1.3967567417674211e-05, "loss": 1.004, "step": 7658 }, { "epoch": 0.39, "learning_rate": 1.3966056997533497e-05, "loss": 0.9718, "step": 7659 }, { "epoch": 0.39, "learning_rate": 1.3964546470016006e-05, "loss": 0.957, "step": 7660 }, { "epoch": 0.39, "learning_rate": 1.3963035835162642e-05, "loss": 0.9867, "step": 7661 }, { "epoch": 0.39, "learning_rate": 1.3961525093014294e-05, "loss": 0.885, "step": 7662 }, { "epoch": 0.39, "learning_rate": 1.3960014243611873e-05, "loss": 0.8911, "step": 7663 }, { "epoch": 0.39, "learning_rate": 1.3958503286996279e-05, "loss": 0.9352, "step": 7664 }, { "epoch": 0.39, "learning_rate": 1.3956992223208418e-05, "loss": 0.9366, "step": 7665 }, { "epoch": 0.39, "learning_rate": 1.3955481052289204e-05, "loss": 0.9609, "step": 7666 }, { "epoch": 0.39, "learning_rate": 1.395396977427955e-05, "loss": 0.9491, "step": 7667 }, { "epoch": 0.39, "learning_rate": 1.395245838922037e-05, "loss": 0.9145, "step": 7668 }, { "epoch": 0.39, "learning_rate": 1.395094689715258e-05, "loss": 0.965, "step": 7669 }, { "epoch": 0.39, "learning_rate": 1.3949435298117111e-05, "loss": 0.9313, "step": 7670 }, { "epoch": 0.39, "learning_rate": 1.3947923592154884e-05, "loss": 0.9318, "step": 7671 }, { "epoch": 0.39, "learning_rate": 1.3946411779306823e-05, "loss": 1.0147, "step": 7672 }, { "epoch": 0.39, "learning_rate": 1.3944899859613863e-05, "loss": 1.0378, "step": 7673 }, { "epoch": 0.39, "learning_rate": 1.3943387833116935e-05, "loss": 0.8912, "step": 7674 }, { "epoch": 0.39, "learning_rate": 1.3941875699856977e-05, "loss": 0.9708, "step": 7675 }, { "epoch": 0.39, "learning_rate": 1.3940363459874931e-05, "loss": 0.9332, "step": 7676 }, { "epoch": 0.39, "learning_rate": 1.3938851113211736e-05, "loss": 0.963, "step": 7677 }, { "epoch": 0.39, "learning_rate": 1.3937338659908335e-05, "loss": 0.952, "step": 7678 }, { "epoch": 0.39, "learning_rate": 1.3935826100005681e-05, "loss": 0.9156, "step": 7679 }, { "epoch": 0.39, "learning_rate": 1.3934313433544722e-05, "loss": 0.9421, "step": 7680 }, { "epoch": 0.39, "learning_rate": 1.393280066056641e-05, "loss": 0.986, "step": 7681 }, { "epoch": 0.39, "learning_rate": 1.3931287781111708e-05, "loss": 1.0011, "step": 7682 }, { "epoch": 0.39, "learning_rate": 1.392977479522157e-05, "loss": 1.0007, "step": 7683 }, { "epoch": 0.39, "learning_rate": 1.3928261702936963e-05, "loss": 0.9193, "step": 7684 }, { "epoch": 0.39, "learning_rate": 1.3926748504298847e-05, "loss": 0.9368, "step": 7685 }, { "epoch": 0.39, "learning_rate": 1.3925235199348192e-05, "loss": 0.9819, "step": 7686 }, { "epoch": 0.39, "learning_rate": 1.3923721788125972e-05, "loss": 0.8224, "step": 7687 }, { "epoch": 0.39, "learning_rate": 1.3922208270673162e-05, "loss": 0.8773, "step": 7688 }, { "epoch": 0.39, "learning_rate": 1.3920694647030732e-05, "loss": 1.044, "step": 7689 }, { "epoch": 0.39, "learning_rate": 1.3919180917239665e-05, "loss": 0.9126, "step": 7690 }, { "epoch": 0.39, "learning_rate": 1.3917667081340947e-05, "loss": 0.9496, "step": 7691 }, { "epoch": 0.39, "learning_rate": 1.3916153139375559e-05, "loss": 0.9762, "step": 7692 }, { "epoch": 0.39, "learning_rate": 1.3914639091384491e-05, "loss": 0.9886, "step": 7693 }, { "epoch": 0.39, "learning_rate": 1.3913124937408737e-05, "loss": 0.821, "step": 7694 }, { "epoch": 0.39, "learning_rate": 1.3911610677489289e-05, "loss": 0.9433, "step": 7695 }, { "epoch": 0.39, "learning_rate": 1.391009631166714e-05, "loss": 0.8119, "step": 7696 }, { "epoch": 0.39, "learning_rate": 1.3908581839983294e-05, "loss": 0.9528, "step": 7697 }, { "epoch": 0.39, "learning_rate": 1.3907067262478755e-05, "loss": 0.9155, "step": 7698 }, { "epoch": 0.39, "learning_rate": 1.3905552579194526e-05, "loss": 1.038, "step": 7699 }, { "epoch": 0.39, "learning_rate": 1.3904037790171616e-05, "loss": 0.9951, "step": 7700 }, { "epoch": 0.39, "learning_rate": 1.3902522895451037e-05, "loss": 0.848, "step": 7701 }, { "epoch": 0.39, "learning_rate": 1.3901007895073802e-05, "loss": 0.8894, "step": 7702 }, { "epoch": 0.39, "learning_rate": 1.3899492789080926e-05, "loss": 0.8924, "step": 7703 }, { "epoch": 0.39, "learning_rate": 1.3897977577513435e-05, "loss": 0.8406, "step": 7704 }, { "epoch": 0.39, "learning_rate": 1.3896462260412347e-05, "loss": 0.8814, "step": 7705 }, { "epoch": 0.39, "learning_rate": 1.389494683781869e-05, "loss": 0.9515, "step": 7706 }, { "epoch": 0.39, "learning_rate": 1.389343130977349e-05, "loss": 1.0786, "step": 7707 }, { "epoch": 0.39, "learning_rate": 1.389191567631778e-05, "loss": 0.876, "step": 7708 }, { "epoch": 0.39, "learning_rate": 1.3890399937492592e-05, "loss": 1.0298, "step": 7709 }, { "epoch": 0.39, "learning_rate": 1.3888884093338968e-05, "loss": 0.981, "step": 7710 }, { "epoch": 0.39, "learning_rate": 1.3887368143897941e-05, "loss": 0.9336, "step": 7711 }, { "epoch": 0.39, "learning_rate": 1.3885852089210561e-05, "loss": 1.0605, "step": 7712 }, { "epoch": 0.39, "learning_rate": 1.3884335929317869e-05, "loss": 0.9503, "step": 7713 }, { "epoch": 0.39, "learning_rate": 1.3882819664260912e-05, "loss": 0.9551, "step": 7714 }, { "epoch": 0.39, "learning_rate": 1.3881303294080744e-05, "loss": 0.8997, "step": 7715 }, { "epoch": 0.39, "learning_rate": 1.387978681881842e-05, "loss": 0.9362, "step": 7716 }, { "epoch": 0.39, "learning_rate": 1.3878270238514995e-05, "loss": 1.02, "step": 7717 }, { "epoch": 0.39, "learning_rate": 1.387675355321153e-05, "loss": 0.8571, "step": 7718 }, { "epoch": 0.39, "learning_rate": 1.3875236762949088e-05, "loss": 0.9371, "step": 7719 }, { "epoch": 0.39, "learning_rate": 1.3873719867768732e-05, "loss": 0.8166, "step": 7720 }, { "epoch": 0.39, "learning_rate": 1.387220286771153e-05, "loss": 1.0209, "step": 7721 }, { "epoch": 0.39, "learning_rate": 1.3870685762818558e-05, "loss": 0.876, "step": 7722 }, { "epoch": 0.39, "learning_rate": 1.3869168553130887e-05, "loss": 0.9177, "step": 7723 }, { "epoch": 0.39, "learning_rate": 1.3867651238689594e-05, "loss": 0.8878, "step": 7724 }, { "epoch": 0.39, "learning_rate": 1.3866133819535757e-05, "loss": 0.9137, "step": 7725 }, { "epoch": 0.39, "learning_rate": 1.3864616295710462e-05, "loss": 0.9445, "step": 7726 }, { "epoch": 0.39, "learning_rate": 1.386309866725479e-05, "loss": 0.9294, "step": 7727 }, { "epoch": 0.39, "learning_rate": 1.3861580934209832e-05, "loss": 1.0123, "step": 7728 }, { "epoch": 0.39, "learning_rate": 1.3860063096616681e-05, "loss": 0.9803, "step": 7729 }, { "epoch": 0.39, "learning_rate": 1.3858545154516427e-05, "loss": 1.0032, "step": 7730 }, { "epoch": 0.39, "learning_rate": 1.3857027107950167e-05, "loss": 0.9909, "step": 7731 }, { "epoch": 0.39, "learning_rate": 1.3855508956959004e-05, "loss": 0.9373, "step": 7732 }, { "epoch": 0.39, "learning_rate": 1.3853990701584037e-05, "loss": 0.9758, "step": 7733 }, { "epoch": 0.39, "learning_rate": 1.385247234186637e-05, "loss": 1.0209, "step": 7734 }, { "epoch": 0.39, "learning_rate": 1.3850953877847115e-05, "loss": 0.8924, "step": 7735 }, { "epoch": 0.39, "learning_rate": 1.3849435309567383e-05, "loss": 0.8819, "step": 7736 }, { "epoch": 0.39, "learning_rate": 1.3847916637068281e-05, "loss": 0.8923, "step": 7737 }, { "epoch": 0.39, "learning_rate": 1.3846397860390935e-05, "loss": 0.9343, "step": 7738 }, { "epoch": 0.39, "learning_rate": 1.3844878979576454e-05, "loss": 1.0174, "step": 7739 }, { "epoch": 0.39, "learning_rate": 1.3843359994665968e-05, "loss": 0.9567, "step": 7740 }, { "epoch": 0.39, "learning_rate": 1.3841840905700599e-05, "loss": 0.8716, "step": 7741 }, { "epoch": 0.39, "learning_rate": 1.3840321712721477e-05, "loss": 0.9762, "step": 7742 }, { "epoch": 0.39, "learning_rate": 1.3838802415769725e-05, "loss": 0.9682, "step": 7743 }, { "epoch": 0.39, "learning_rate": 1.3837283014886484e-05, "loss": 0.8394, "step": 7744 }, { "epoch": 0.39, "learning_rate": 1.383576351011289e-05, "loss": 0.991, "step": 7745 }, { "epoch": 0.39, "learning_rate": 1.3834243901490076e-05, "loss": 0.9015, "step": 7746 }, { "epoch": 0.39, "learning_rate": 1.3832724189059192e-05, "loss": 0.9418, "step": 7747 }, { "epoch": 0.39, "learning_rate": 1.3831204372861377e-05, "loss": 0.8919, "step": 7748 }, { "epoch": 0.39, "learning_rate": 1.3829684452937779e-05, "loss": 0.9636, "step": 7749 }, { "epoch": 0.39, "learning_rate": 1.382816442932955e-05, "loss": 1.0647, "step": 7750 }, { "epoch": 0.39, "learning_rate": 1.3826644302077839e-05, "loss": 0.9119, "step": 7751 }, { "epoch": 0.39, "learning_rate": 1.3825124071223805e-05, "loss": 0.9696, "step": 7752 }, { "epoch": 0.39, "learning_rate": 1.382360373680861e-05, "loss": 0.9576, "step": 7753 }, { "epoch": 0.39, "learning_rate": 1.382208329887341e-05, "loss": 1.0243, "step": 7754 }, { "epoch": 0.39, "learning_rate": 1.3820562757459371e-05, "loss": 1.0092, "step": 7755 }, { "epoch": 0.39, "learning_rate": 1.3819042112607661e-05, "loss": 0.8825, "step": 7756 }, { "epoch": 0.39, "learning_rate": 1.3817521364359447e-05, "loss": 0.9464, "step": 7757 }, { "epoch": 0.39, "learning_rate": 1.3816000512755902e-05, "loss": 0.8387, "step": 7758 }, { "epoch": 0.39, "learning_rate": 1.3814479557838207e-05, "loss": 0.8773, "step": 7759 }, { "epoch": 0.39, "learning_rate": 1.3812958499647537e-05, "loss": 1.0164, "step": 7760 }, { "epoch": 0.39, "learning_rate": 1.381143733822507e-05, "loss": 0.9543, "step": 7761 }, { "epoch": 0.39, "learning_rate": 1.3809916073611991e-05, "loss": 0.9507, "step": 7762 }, { "epoch": 0.39, "learning_rate": 1.380839470584949e-05, "loss": 0.9343, "step": 7763 }, { "epoch": 0.39, "learning_rate": 1.3806873234978753e-05, "loss": 0.9445, "step": 7764 }, { "epoch": 0.39, "learning_rate": 1.3805351661040973e-05, "loss": 0.9504, "step": 7765 }, { "epoch": 0.39, "learning_rate": 1.3803829984077348e-05, "loss": 0.9315, "step": 7766 }, { "epoch": 0.39, "learning_rate": 1.3802308204129069e-05, "loss": 0.9335, "step": 7767 }, { "epoch": 0.39, "learning_rate": 1.3800786321237342e-05, "loss": 0.951, "step": 7768 }, { "epoch": 0.39, "learning_rate": 1.3799264335443372e-05, "loss": 0.9301, "step": 7769 }, { "epoch": 0.39, "learning_rate": 1.379774224678836e-05, "loss": 1.0809, "step": 7770 }, { "epoch": 0.39, "learning_rate": 1.3796220055313516e-05, "loss": 1.0124, "step": 7771 }, { "epoch": 0.39, "learning_rate": 1.3794697761060056e-05, "loss": 0.9119, "step": 7772 }, { "epoch": 0.39, "learning_rate": 1.379317536406919e-05, "loss": 0.9185, "step": 7773 }, { "epoch": 0.39, "learning_rate": 1.3791652864382133e-05, "loss": 0.9391, "step": 7774 }, { "epoch": 0.39, "learning_rate": 1.3790130262040114e-05, "loss": 0.9732, "step": 7775 }, { "epoch": 0.4, "learning_rate": 1.3788607557084346e-05, "loss": 0.9334, "step": 7776 }, { "epoch": 0.4, "learning_rate": 1.378708474955606e-05, "loss": 0.9934, "step": 7777 }, { "epoch": 0.4, "learning_rate": 1.3785561839496489e-05, "loss": 0.9753, "step": 7778 }, { "epoch": 0.4, "learning_rate": 1.3784038826946853e-05, "loss": 0.9164, "step": 7779 }, { "epoch": 0.4, "learning_rate": 1.3782515711948392e-05, "loss": 0.8586, "step": 7780 }, { "epoch": 0.4, "learning_rate": 1.3780992494542345e-05, "loss": 0.968, "step": 7781 }, { "epoch": 0.4, "learning_rate": 1.3779469174769945e-05, "loss": 0.96, "step": 7782 }, { "epoch": 0.4, "learning_rate": 1.3777945752672444e-05, "loss": 0.9955, "step": 7783 }, { "epoch": 0.4, "learning_rate": 1.377642222829108e-05, "loss": 1.0071, "step": 7784 }, { "epoch": 0.4, "learning_rate": 1.3774898601667096e-05, "loss": 0.8733, "step": 7785 }, { "epoch": 0.4, "learning_rate": 1.3773374872841751e-05, "loss": 0.8961, "step": 7786 }, { "epoch": 0.4, "learning_rate": 1.37718510418563e-05, "loss": 1.1074, "step": 7787 }, { "epoch": 0.4, "learning_rate": 1.3770327108751991e-05, "loss": 0.9602, "step": 7788 }, { "epoch": 0.4, "learning_rate": 1.376880307357009e-05, "loss": 1.0069, "step": 7789 }, { "epoch": 0.4, "learning_rate": 1.3767278936351853e-05, "loss": 0.9919, "step": 7790 }, { "epoch": 0.4, "learning_rate": 1.3765754697138547e-05, "loss": 0.9141, "step": 7791 }, { "epoch": 0.4, "learning_rate": 1.376423035597144e-05, "loss": 0.9656, "step": 7792 }, { "epoch": 0.4, "learning_rate": 1.37627059128918e-05, "loss": 0.9406, "step": 7793 }, { "epoch": 0.4, "learning_rate": 1.37611813679409e-05, "loss": 0.8973, "step": 7794 }, { "epoch": 0.4, "learning_rate": 1.375965672116002e-05, "loss": 0.9294, "step": 7795 }, { "epoch": 0.4, "learning_rate": 1.375813197259043e-05, "loss": 0.9367, "step": 7796 }, { "epoch": 0.4, "learning_rate": 1.3756607122273416e-05, "loss": 0.9881, "step": 7797 }, { "epoch": 0.4, "learning_rate": 1.375508217025026e-05, "loss": 0.865, "step": 7798 }, { "epoch": 0.4, "learning_rate": 1.3753557116562251e-05, "loss": 0.8895, "step": 7799 }, { "epoch": 0.4, "learning_rate": 1.3752031961250676e-05, "loss": 0.9492, "step": 7800 }, { "epoch": 0.4, "learning_rate": 1.375050670435683e-05, "loss": 0.8399, "step": 7801 }, { "epoch": 0.4, "learning_rate": 1.3748981345922e-05, "loss": 0.9506, "step": 7802 }, { "epoch": 0.4, "learning_rate": 1.3747455885987493e-05, "loss": 0.9112, "step": 7803 }, { "epoch": 0.4, "learning_rate": 1.3745930324594603e-05, "loss": 0.9626, "step": 7804 }, { "epoch": 0.4, "learning_rate": 1.3744404661784634e-05, "loss": 1.0373, "step": 7805 }, { "epoch": 0.4, "learning_rate": 1.3742878897598895e-05, "loss": 0.8582, "step": 7806 }, { "epoch": 0.4, "learning_rate": 1.3741353032078693e-05, "loss": 1.0132, "step": 7807 }, { "epoch": 0.4, "learning_rate": 1.3739827065265334e-05, "loss": 0.8828, "step": 7808 }, { "epoch": 0.4, "learning_rate": 1.3738300997200137e-05, "loss": 0.9703, "step": 7809 }, { "epoch": 0.4, "learning_rate": 1.373677482792442e-05, "loss": 0.9638, "step": 7810 }, { "epoch": 0.4, "learning_rate": 1.3735248557479496e-05, "loss": 0.9506, "step": 7811 }, { "epoch": 0.4, "learning_rate": 1.3733722185906697e-05, "loss": 0.9942, "step": 7812 }, { "epoch": 0.4, "learning_rate": 1.373219571324734e-05, "loss": 0.916, "step": 7813 }, { "epoch": 0.4, "learning_rate": 1.3730669139542751e-05, "loss": 1.0576, "step": 7814 }, { "epoch": 0.4, "learning_rate": 1.372914246483427e-05, "loss": 0.9696, "step": 7815 }, { "epoch": 0.4, "learning_rate": 1.372761568916322e-05, "loss": 0.8778, "step": 7816 }, { "epoch": 0.4, "learning_rate": 1.3726088812570942e-05, "loss": 0.9573, "step": 7817 }, { "epoch": 0.4, "learning_rate": 1.3724561835098775e-05, "loss": 0.9656, "step": 7818 }, { "epoch": 0.4, "learning_rate": 1.372303475678806e-05, "loss": 0.9385, "step": 7819 }, { "epoch": 0.4, "learning_rate": 1.3721507577680137e-05, "loss": 0.9943, "step": 7820 }, { "epoch": 0.4, "learning_rate": 1.3719980297816358e-05, "loss": 0.8936, "step": 7821 }, { "epoch": 0.4, "learning_rate": 1.3718452917238068e-05, "loss": 0.8816, "step": 7822 }, { "epoch": 0.4, "learning_rate": 1.371692543598662e-05, "loss": 1.007, "step": 7823 }, { "epoch": 0.4, "learning_rate": 1.3715397854103376e-05, "loss": 0.8964, "step": 7824 }, { "epoch": 0.4, "learning_rate": 1.3713870171629684e-05, "loss": 0.9219, "step": 7825 }, { "epoch": 0.4, "learning_rate": 1.371234238860691e-05, "loss": 0.8914, "step": 7826 }, { "epoch": 0.4, "learning_rate": 1.3710814505076415e-05, "loss": 0.9865, "step": 7827 }, { "epoch": 0.4, "learning_rate": 1.3709286521079563e-05, "loss": 1.052, "step": 7828 }, { "epoch": 0.4, "learning_rate": 1.3707758436657725e-05, "loss": 0.8291, "step": 7829 }, { "epoch": 0.4, "learning_rate": 1.3706230251852275e-05, "loss": 0.9174, "step": 7830 }, { "epoch": 0.4, "learning_rate": 1.3704701966704585e-05, "loss": 0.9411, "step": 7831 }, { "epoch": 0.4, "learning_rate": 1.3703173581256027e-05, "loss": 0.9328, "step": 7832 }, { "epoch": 0.4, "learning_rate": 1.3701645095547984e-05, "loss": 0.9254, "step": 7833 }, { "epoch": 0.4, "learning_rate": 1.3700116509621842e-05, "loss": 0.9031, "step": 7834 }, { "epoch": 0.4, "learning_rate": 1.3698587823518978e-05, "loss": 0.8945, "step": 7835 }, { "epoch": 0.4, "learning_rate": 1.3697059037280789e-05, "loss": 0.8585, "step": 7836 }, { "epoch": 0.4, "learning_rate": 1.3695530150948657e-05, "loss": 0.8722, "step": 7837 }, { "epoch": 0.4, "learning_rate": 1.3694001164563977e-05, "loss": 0.9534, "step": 7838 }, { "epoch": 0.4, "learning_rate": 1.3692472078168144e-05, "loss": 0.9333, "step": 7839 }, { "epoch": 0.4, "learning_rate": 1.3690942891802562e-05, "loss": 0.8736, "step": 7840 }, { "epoch": 0.4, "learning_rate": 1.3689413605508623e-05, "loss": 0.8899, "step": 7841 }, { "epoch": 0.4, "learning_rate": 1.3687884219327741e-05, "loss": 0.9571, "step": 7842 }, { "epoch": 0.4, "learning_rate": 1.3686354733301318e-05, "loss": 0.9246, "step": 7843 }, { "epoch": 0.4, "learning_rate": 1.3684825147470757e-05, "loss": 0.9854, "step": 7844 }, { "epoch": 0.4, "learning_rate": 1.3683295461877477e-05, "loss": 0.859, "step": 7845 }, { "epoch": 0.4, "learning_rate": 1.3681765676562895e-05, "loss": 0.8684, "step": 7846 }, { "epoch": 0.4, "learning_rate": 1.368023579156842e-05, "loss": 0.9584, "step": 7847 }, { "epoch": 0.4, "learning_rate": 1.367870580693548e-05, "loss": 0.9495, "step": 7848 }, { "epoch": 0.4, "learning_rate": 1.3677175722705494e-05, "loss": 0.9642, "step": 7849 }, { "epoch": 0.4, "learning_rate": 1.3675645538919884e-05, "loss": 0.9665, "step": 7850 }, { "epoch": 0.4, "learning_rate": 1.3674115255620082e-05, "loss": 0.9798, "step": 7851 }, { "epoch": 0.4, "learning_rate": 1.3672584872847521e-05, "loss": 0.9458, "step": 7852 }, { "epoch": 0.4, "learning_rate": 1.367105439064363e-05, "loss": 0.9182, "step": 7853 }, { "epoch": 0.4, "learning_rate": 1.366952380904985e-05, "loss": 0.8887, "step": 7854 }, { "epoch": 0.4, "learning_rate": 1.3667993128107615e-05, "loss": 0.9965, "step": 7855 }, { "epoch": 0.4, "learning_rate": 1.3666462347858367e-05, "loss": 0.8823, "step": 7856 }, { "epoch": 0.4, "learning_rate": 1.3664931468343551e-05, "loss": 0.9636, "step": 7857 }, { "epoch": 0.4, "learning_rate": 1.3663400489604619e-05, "loss": 0.9994, "step": 7858 }, { "epoch": 0.4, "learning_rate": 1.3661869411683012e-05, "loss": 0.9184, "step": 7859 }, { "epoch": 0.4, "learning_rate": 1.3660338234620189e-05, "loss": 1.0281, "step": 7860 }, { "epoch": 0.4, "learning_rate": 1.3658806958457604e-05, "loss": 0.9304, "step": 7861 }, { "epoch": 0.4, "learning_rate": 1.3657275583236712e-05, "loss": 1.0197, "step": 7862 }, { "epoch": 0.4, "learning_rate": 1.3655744108998974e-05, "loss": 0.9386, "step": 7863 }, { "epoch": 0.4, "learning_rate": 1.365421253578585e-05, "loss": 0.9567, "step": 7864 }, { "epoch": 0.4, "learning_rate": 1.3652680863638816e-05, "loss": 0.9238, "step": 7865 }, { "epoch": 0.4, "learning_rate": 1.3651149092599332e-05, "loss": 0.8878, "step": 7866 }, { "epoch": 0.4, "learning_rate": 1.3649617222708869e-05, "loss": 0.8995, "step": 7867 }, { "epoch": 0.4, "learning_rate": 1.3648085254008906e-05, "loss": 0.8627, "step": 7868 }, { "epoch": 0.4, "learning_rate": 1.364655318654091e-05, "loss": 0.89, "step": 7869 }, { "epoch": 0.4, "learning_rate": 1.3645021020346368e-05, "loss": 0.9302, "step": 7870 }, { "epoch": 0.4, "learning_rate": 1.3643488755466762e-05, "loss": 1.0454, "step": 7871 }, { "epoch": 0.4, "learning_rate": 1.3641956391943577e-05, "loss": 0.9355, "step": 7872 }, { "epoch": 0.4, "learning_rate": 1.3640423929818292e-05, "loss": 0.8077, "step": 7873 }, { "epoch": 0.4, "learning_rate": 1.3638891369132406e-05, "loss": 0.97, "step": 7874 }, { "epoch": 0.4, "learning_rate": 1.3637358709927405e-05, "loss": 0.9067, "step": 7875 }, { "epoch": 0.4, "learning_rate": 1.3635825952244786e-05, "loss": 0.9491, "step": 7876 }, { "epoch": 0.4, "learning_rate": 1.363429309612605e-05, "loss": 0.9099, "step": 7877 }, { "epoch": 0.4, "learning_rate": 1.3632760141612696e-05, "loss": 0.9483, "step": 7878 }, { "epoch": 0.4, "learning_rate": 1.3631227088746222e-05, "loss": 1.093, "step": 7879 }, { "epoch": 0.4, "learning_rate": 1.362969393756814e-05, "loss": 0.9464, "step": 7880 }, { "epoch": 0.4, "learning_rate": 1.3628160688119958e-05, "loss": 0.9315, "step": 7881 }, { "epoch": 0.4, "learning_rate": 1.3626627340443182e-05, "loss": 0.9744, "step": 7882 }, { "epoch": 0.4, "learning_rate": 1.3625093894579333e-05, "loss": 0.8782, "step": 7883 }, { "epoch": 0.4, "learning_rate": 1.3623560350569922e-05, "loss": 0.9194, "step": 7884 }, { "epoch": 0.4, "learning_rate": 1.3622026708456469e-05, "loss": 1.0039, "step": 7885 }, { "epoch": 0.4, "learning_rate": 1.3620492968280497e-05, "loss": 0.9267, "step": 7886 }, { "epoch": 0.4, "learning_rate": 1.361895913008353e-05, "loss": 0.9483, "step": 7887 }, { "epoch": 0.4, "learning_rate": 1.3617425193907091e-05, "loss": 0.8945, "step": 7888 }, { "epoch": 0.4, "learning_rate": 1.361589115979272e-05, "loss": 0.9553, "step": 7889 }, { "epoch": 0.4, "learning_rate": 1.361435702778194e-05, "loss": 0.9976, "step": 7890 }, { "epoch": 0.4, "learning_rate": 1.3612822797916288e-05, "loss": 0.8795, "step": 7891 }, { "epoch": 0.4, "learning_rate": 1.3611288470237305e-05, "loss": 0.9708, "step": 7892 }, { "epoch": 0.4, "learning_rate": 1.3609754044786525e-05, "loss": 0.8762, "step": 7893 }, { "epoch": 0.4, "learning_rate": 1.3608219521605498e-05, "loss": 1.0124, "step": 7894 }, { "epoch": 0.4, "learning_rate": 1.3606684900735765e-05, "loss": 0.9565, "step": 7895 }, { "epoch": 0.4, "learning_rate": 1.3605150182218876e-05, "loss": 0.9334, "step": 7896 }, { "epoch": 0.4, "learning_rate": 1.3603615366096378e-05, "loss": 1.0639, "step": 7897 }, { "epoch": 0.4, "learning_rate": 1.3602080452409833e-05, "loss": 1.0172, "step": 7898 }, { "epoch": 0.4, "learning_rate": 1.3600545441200788e-05, "loss": 0.9133, "step": 7899 }, { "epoch": 0.4, "learning_rate": 1.3599010332510807e-05, "loss": 0.9559, "step": 7900 }, { "epoch": 0.4, "learning_rate": 1.3597475126381452e-05, "loss": 0.9376, "step": 7901 }, { "epoch": 0.4, "learning_rate": 1.3595939822854285e-05, "loss": 0.8956, "step": 7902 }, { "epoch": 0.4, "learning_rate": 1.3594404421970873e-05, "loss": 0.9834, "step": 7903 }, { "epoch": 0.4, "learning_rate": 1.3592868923772785e-05, "loss": 0.9318, "step": 7904 }, { "epoch": 0.4, "learning_rate": 1.3591333328301593e-05, "loss": 1.1003, "step": 7905 }, { "epoch": 0.4, "learning_rate": 1.3589797635598872e-05, "loss": 0.9618, "step": 7906 }, { "epoch": 0.4, "learning_rate": 1.3588261845706199e-05, "loss": 0.9968, "step": 7907 }, { "epoch": 0.4, "learning_rate": 1.3586725958665157e-05, "loss": 0.9139, "step": 7908 }, { "epoch": 0.4, "learning_rate": 1.3585189974517321e-05, "loss": 0.855, "step": 7909 }, { "epoch": 0.4, "learning_rate": 1.3583653893304283e-05, "loss": 0.8763, "step": 7910 }, { "epoch": 0.4, "learning_rate": 1.3582117715067628e-05, "loss": 0.9409, "step": 7911 }, { "epoch": 0.4, "learning_rate": 1.3580581439848947e-05, "loss": 0.8661, "step": 7912 }, { "epoch": 0.4, "learning_rate": 1.3579045067689836e-05, "loss": 1.0358, "step": 7913 }, { "epoch": 0.4, "learning_rate": 1.3577508598631886e-05, "loss": 0.9101, "step": 7914 }, { "epoch": 0.4, "learning_rate": 1.3575972032716695e-05, "loss": 0.9619, "step": 7915 }, { "epoch": 0.4, "learning_rate": 1.3574435369985866e-05, "loss": 0.8136, "step": 7916 }, { "epoch": 0.4, "learning_rate": 1.3572898610481002e-05, "loss": 0.8796, "step": 7917 }, { "epoch": 0.4, "learning_rate": 1.357136175424371e-05, "loss": 0.9726, "step": 7918 }, { "epoch": 0.4, "learning_rate": 1.3569824801315602e-05, "loss": 0.8876, "step": 7919 }, { "epoch": 0.4, "learning_rate": 1.3568287751738281e-05, "loss": 0.9236, "step": 7920 }, { "epoch": 0.4, "learning_rate": 1.3566750605553365e-05, "loss": 0.971, "step": 7921 }, { "epoch": 0.4, "learning_rate": 1.3565213362802472e-05, "loss": 0.9684, "step": 7922 }, { "epoch": 0.4, "learning_rate": 1.356367602352722e-05, "loss": 0.8876, "step": 7923 }, { "epoch": 0.4, "learning_rate": 1.3562138587769234e-05, "loss": 0.9852, "step": 7924 }, { "epoch": 0.4, "learning_rate": 1.3560601055570133e-05, "loss": 0.9891, "step": 7925 }, { "epoch": 0.4, "learning_rate": 1.3559063426971549e-05, "loss": 0.8817, "step": 7926 }, { "epoch": 0.4, "learning_rate": 1.3557525702015106e-05, "loss": 0.8881, "step": 7927 }, { "epoch": 0.4, "learning_rate": 1.3555987880742441e-05, "loss": 1.0194, "step": 7928 }, { "epoch": 0.4, "learning_rate": 1.3554449963195189e-05, "loss": 0.8933, "step": 7929 }, { "epoch": 0.4, "learning_rate": 1.3552911949414982e-05, "loss": 0.9849, "step": 7930 }, { "epoch": 0.4, "learning_rate": 1.3551373839443468e-05, "loss": 0.931, "step": 7931 }, { "epoch": 0.4, "learning_rate": 1.3549835633322284e-05, "loss": 0.9868, "step": 7932 }, { "epoch": 0.4, "learning_rate": 1.3548297331093077e-05, "loss": 1.0673, "step": 7933 }, { "epoch": 0.4, "learning_rate": 1.3546758932797492e-05, "loss": 0.9597, "step": 7934 }, { "epoch": 0.4, "learning_rate": 1.3545220438477185e-05, "loss": 0.9738, "step": 7935 }, { "epoch": 0.4, "learning_rate": 1.3543681848173806e-05, "loss": 0.9706, "step": 7936 }, { "epoch": 0.4, "learning_rate": 1.3542143161929013e-05, "loss": 1.0495, "step": 7937 }, { "epoch": 0.4, "learning_rate": 1.354060437978446e-05, "loss": 0.9891, "step": 7938 }, { "epoch": 0.4, "learning_rate": 1.3539065501781811e-05, "loss": 0.9311, "step": 7939 }, { "epoch": 0.4, "learning_rate": 1.353752652796273e-05, "loss": 0.868, "step": 7940 }, { "epoch": 0.4, "learning_rate": 1.3535987458368879e-05, "loss": 0.9671, "step": 7941 }, { "epoch": 0.4, "learning_rate": 1.3534448293041932e-05, "loss": 0.878, "step": 7942 }, { "epoch": 0.4, "learning_rate": 1.353290903202356e-05, "loss": 0.9294, "step": 7943 }, { "epoch": 0.4, "learning_rate": 1.353136967535543e-05, "loss": 0.8831, "step": 7944 }, { "epoch": 0.4, "learning_rate": 1.3529830223079228e-05, "loss": 1.0032, "step": 7945 }, { "epoch": 0.4, "learning_rate": 1.3528290675236624e-05, "loss": 1.0097, "step": 7946 }, { "epoch": 0.4, "learning_rate": 1.3526751031869303e-05, "loss": 0.908, "step": 7947 }, { "epoch": 0.4, "learning_rate": 1.3525211293018953e-05, "loss": 0.886, "step": 7948 }, { "epoch": 0.4, "learning_rate": 1.3523671458727258e-05, "loss": 0.9387, "step": 7949 }, { "epoch": 0.4, "learning_rate": 1.3522131529035907e-05, "loss": 0.9457, "step": 7950 }, { "epoch": 0.4, "learning_rate": 1.3520591503986592e-05, "loss": 1.0814, "step": 7951 }, { "epoch": 0.4, "learning_rate": 1.3519051383621005e-05, "loss": 0.9507, "step": 7952 }, { "epoch": 0.4, "learning_rate": 1.3517511167980847e-05, "loss": 1.0079, "step": 7953 }, { "epoch": 0.4, "learning_rate": 1.3515970857107818e-05, "loss": 0.9112, "step": 7954 }, { "epoch": 0.4, "learning_rate": 1.3514430451043619e-05, "loss": 0.9633, "step": 7955 }, { "epoch": 0.4, "learning_rate": 1.3512889949829952e-05, "loss": 0.9073, "step": 7956 }, { "epoch": 0.4, "learning_rate": 1.351134935350853e-05, "loss": 0.915, "step": 7957 }, { "epoch": 0.4, "learning_rate": 1.3509808662121056e-05, "loss": 1.0059, "step": 7958 }, { "epoch": 0.4, "learning_rate": 1.3508267875709248e-05, "loss": 1.0732, "step": 7959 }, { "epoch": 0.4, "learning_rate": 1.3506726994314823e-05, "loss": 0.9397, "step": 7960 }, { "epoch": 0.4, "learning_rate": 1.3505186017979495e-05, "loss": 0.8745, "step": 7961 }, { "epoch": 0.4, "learning_rate": 1.3503644946744982e-05, "loss": 0.9819, "step": 7962 }, { "epoch": 0.4, "learning_rate": 1.350210378065301e-05, "loss": 1.0643, "step": 7963 }, { "epoch": 0.4, "learning_rate": 1.3500562519745307e-05, "loss": 0.9768, "step": 7964 }, { "epoch": 0.4, "learning_rate": 1.3499021164063592e-05, "loss": 0.8628, "step": 7965 }, { "epoch": 0.4, "learning_rate": 1.3497479713649609e-05, "loss": 0.9467, "step": 7966 }, { "epoch": 0.4, "learning_rate": 1.349593816854508e-05, "loss": 0.8852, "step": 7967 }, { "epoch": 0.4, "learning_rate": 1.3494396528791742e-05, "loss": 0.893, "step": 7968 }, { "epoch": 0.4, "learning_rate": 1.3492854794431339e-05, "loss": 1.1088, "step": 7969 }, { "epoch": 0.4, "learning_rate": 1.349131296550561e-05, "loss": 0.9375, "step": 7970 }, { "epoch": 0.4, "learning_rate": 1.3489771042056292e-05, "loss": 1.0368, "step": 7971 }, { "epoch": 0.4, "learning_rate": 1.3488229024125142e-05, "loss": 0.936, "step": 7972 }, { "epoch": 0.41, "learning_rate": 1.3486686911753899e-05, "loss": 0.9239, "step": 7973 }, { "epoch": 0.41, "learning_rate": 1.3485144704984317e-05, "loss": 1.0471, "step": 7974 }, { "epoch": 0.41, "learning_rate": 1.348360240385815e-05, "loss": 0.8792, "step": 7975 }, { "epoch": 0.41, "learning_rate": 1.3482060008417154e-05, "loss": 0.8865, "step": 7976 }, { "epoch": 0.41, "learning_rate": 1.348051751870309e-05, "loss": 0.9867, "step": 7977 }, { "epoch": 0.41, "learning_rate": 1.3478974934757717e-05, "loss": 0.9712, "step": 7978 }, { "epoch": 0.41, "learning_rate": 1.3477432256622797e-05, "loss": 0.8941, "step": 7979 }, { "epoch": 0.41, "learning_rate": 1.34758894843401e-05, "loss": 0.9085, "step": 7980 }, { "epoch": 0.41, "learning_rate": 1.347434661795139e-05, "loss": 0.9973, "step": 7981 }, { "epoch": 0.41, "learning_rate": 1.3472803657498445e-05, "loss": 0.9164, "step": 7982 }, { "epoch": 0.41, "learning_rate": 1.3471260603023035e-05, "loss": 1.0542, "step": 7983 }, { "epoch": 0.41, "learning_rate": 1.3469717454566937e-05, "loss": 0.9434, "step": 7984 }, { "epoch": 0.41, "learning_rate": 1.3468174212171934e-05, "loss": 0.9503, "step": 7985 }, { "epoch": 0.41, "learning_rate": 1.3466630875879797e-05, "loss": 0.894, "step": 7986 }, { "epoch": 0.41, "learning_rate": 1.346508744573232e-05, "loss": 0.9091, "step": 7987 }, { "epoch": 0.41, "learning_rate": 1.3463543921771289e-05, "loss": 0.9613, "step": 7988 }, { "epoch": 0.41, "learning_rate": 1.346200030403849e-05, "loss": 0.8538, "step": 7989 }, { "epoch": 0.41, "learning_rate": 1.346045659257572e-05, "loss": 0.997, "step": 7990 }, { "epoch": 0.41, "learning_rate": 1.3458912787424766e-05, "loss": 0.8866, "step": 7991 }, { "epoch": 0.41, "learning_rate": 1.3457368888627427e-05, "loss": 0.8459, "step": 7992 }, { "epoch": 0.41, "learning_rate": 1.3455824896225502e-05, "loss": 1.0289, "step": 7993 }, { "epoch": 0.41, "learning_rate": 1.3454280810260799e-05, "loss": 0.9674, "step": 7994 }, { "epoch": 0.41, "learning_rate": 1.3452736630775113e-05, "loss": 0.91, "step": 7995 }, { "epoch": 0.41, "learning_rate": 1.3451192357810263e-05, "loss": 0.8877, "step": 7996 }, { "epoch": 0.41, "learning_rate": 1.3449647991408047e-05, "loss": 0.8212, "step": 7997 }, { "epoch": 0.41, "learning_rate": 1.3448103531610284e-05, "loss": 0.9478, "step": 7998 }, { "epoch": 0.41, "learning_rate": 1.3446558978458784e-05, "loss": 0.8326, "step": 7999 }, { "epoch": 0.41, "learning_rate": 1.344501433199537e-05, "loss": 1.0178, "step": 8000 }, { "epoch": 0.41, "learning_rate": 1.3443469592261853e-05, "loss": 1.0188, "step": 8001 }, { "epoch": 0.41, "learning_rate": 1.3441924759300065e-05, "loss": 0.9857, "step": 8002 }, { "epoch": 0.41, "learning_rate": 1.3440379833151824e-05, "loss": 0.806, "step": 8003 }, { "epoch": 0.41, "learning_rate": 1.3438834813858962e-05, "loss": 0.9439, "step": 8004 }, { "epoch": 0.41, "learning_rate": 1.34372897014633e-05, "loss": 1.0503, "step": 8005 }, { "epoch": 0.41, "learning_rate": 1.343574449600668e-05, "loss": 0.9896, "step": 8006 }, { "epoch": 0.41, "learning_rate": 1.3434199197530933e-05, "loss": 1.0474, "step": 8007 }, { "epoch": 0.41, "learning_rate": 1.34326538060779e-05, "loss": 0.9673, "step": 8008 }, { "epoch": 0.41, "learning_rate": 1.3431108321689413e-05, "loss": 0.8787, "step": 8009 }, { "epoch": 0.41, "learning_rate": 1.3429562744407318e-05, "loss": 0.8532, "step": 8010 }, { "epoch": 0.41, "learning_rate": 1.3428017074273462e-05, "loss": 0.8233, "step": 8011 }, { "epoch": 0.41, "learning_rate": 1.3426471311329689e-05, "loss": 0.9736, "step": 8012 }, { "epoch": 0.41, "learning_rate": 1.3424925455617854e-05, "loss": 0.9104, "step": 8013 }, { "epoch": 0.41, "learning_rate": 1.3423379507179805e-05, "loss": 1.0418, "step": 8014 }, { "epoch": 0.41, "learning_rate": 1.3421833466057397e-05, "loss": 0.8435, "step": 8015 }, { "epoch": 0.41, "learning_rate": 1.342028733229249e-05, "loss": 0.9291, "step": 8016 }, { "epoch": 0.41, "learning_rate": 1.341874110592694e-05, "loss": 0.9824, "step": 8017 }, { "epoch": 0.41, "learning_rate": 1.3417194787002614e-05, "loss": 0.9261, "step": 8018 }, { "epoch": 0.41, "learning_rate": 1.3415648375561373e-05, "loss": 0.9185, "step": 8019 }, { "epoch": 0.41, "learning_rate": 1.341410187164509e-05, "loss": 0.9198, "step": 8020 }, { "epoch": 0.41, "learning_rate": 1.3412555275295626e-05, "loss": 0.8724, "step": 8021 }, { "epoch": 0.41, "learning_rate": 1.3411008586554864e-05, "loss": 1.0109, "step": 8022 }, { "epoch": 0.41, "learning_rate": 1.3409461805464669e-05, "loss": 0.9809, "step": 8023 }, { "epoch": 0.41, "learning_rate": 1.3407914932066923e-05, "loss": 0.9296, "step": 8024 }, { "epoch": 0.41, "learning_rate": 1.3406367966403509e-05, "loss": 0.9998, "step": 8025 }, { "epoch": 0.41, "learning_rate": 1.3404820908516305e-05, "loss": 0.866, "step": 8026 }, { "epoch": 0.41, "learning_rate": 1.3403273758447198e-05, "loss": 0.9354, "step": 8027 }, { "epoch": 0.41, "learning_rate": 1.3401726516238074e-05, "loss": 0.9888, "step": 8028 }, { "epoch": 0.41, "learning_rate": 1.3400179181930822e-05, "loss": 0.8932, "step": 8029 }, { "epoch": 0.41, "learning_rate": 1.3398631755567337e-05, "loss": 0.9326, "step": 8030 }, { "epoch": 0.41, "learning_rate": 1.3397084237189515e-05, "loss": 0.8459, "step": 8031 }, { "epoch": 0.41, "learning_rate": 1.3395536626839253e-05, "loss": 0.9503, "step": 8032 }, { "epoch": 0.41, "learning_rate": 1.3393988924558445e-05, "loss": 0.8844, "step": 8033 }, { "epoch": 0.41, "learning_rate": 1.3392441130389e-05, "loss": 1.0224, "step": 8034 }, { "epoch": 0.41, "learning_rate": 1.339089324437282e-05, "loss": 0.9045, "step": 8035 }, { "epoch": 0.41, "learning_rate": 1.3389345266551813e-05, "loss": 0.9421, "step": 8036 }, { "epoch": 0.41, "learning_rate": 1.3387797196967889e-05, "loss": 0.9229, "step": 8037 }, { "epoch": 0.41, "learning_rate": 1.3386249035662962e-05, "loss": 1.0341, "step": 8038 }, { "epoch": 0.41, "learning_rate": 1.3384700782678944e-05, "loss": 0.8774, "step": 8039 }, { "epoch": 0.41, "learning_rate": 1.3383152438057749e-05, "loss": 0.9938, "step": 8040 }, { "epoch": 0.41, "learning_rate": 1.3381604001841306e-05, "loss": 0.9457, "step": 8041 }, { "epoch": 0.41, "learning_rate": 1.3380055474071531e-05, "loss": 0.9668, "step": 8042 }, { "epoch": 0.41, "learning_rate": 1.3378506854790352e-05, "loss": 0.9179, "step": 8043 }, { "epoch": 0.41, "learning_rate": 1.3376958144039693e-05, "loss": 0.9913, "step": 8044 }, { "epoch": 0.41, "learning_rate": 1.3375409341861485e-05, "loss": 0.9293, "step": 8045 }, { "epoch": 0.41, "learning_rate": 1.3373860448297659e-05, "loss": 0.9862, "step": 8046 }, { "epoch": 0.41, "learning_rate": 1.3372311463390152e-05, "loss": 0.8599, "step": 8047 }, { "epoch": 0.41, "learning_rate": 1.33707623871809e-05, "loss": 1.0388, "step": 8048 }, { "epoch": 0.41, "learning_rate": 1.3369213219711844e-05, "loss": 0.9027, "step": 8049 }, { "epoch": 0.41, "learning_rate": 1.3367663961024923e-05, "loss": 0.9137, "step": 8050 }, { "epoch": 0.41, "learning_rate": 1.3366114611162084e-05, "loss": 0.9909, "step": 8051 }, { "epoch": 0.41, "learning_rate": 1.3364565170165272e-05, "loss": 0.9149, "step": 8052 }, { "epoch": 0.41, "learning_rate": 1.3363015638076438e-05, "loss": 0.9678, "step": 8053 }, { "epoch": 0.41, "learning_rate": 1.3361466014937534e-05, "loss": 0.9212, "step": 8054 }, { "epoch": 0.41, "learning_rate": 1.3359916300790515e-05, "loss": 1.008, "step": 8055 }, { "epoch": 0.41, "learning_rate": 1.3358366495677336e-05, "loss": 0.9663, "step": 8056 }, { "epoch": 0.41, "learning_rate": 1.3356816599639954e-05, "loss": 0.9322, "step": 8057 }, { "epoch": 0.41, "learning_rate": 1.3355266612720337e-05, "loss": 0.8931, "step": 8058 }, { "epoch": 0.41, "learning_rate": 1.3353716534960445e-05, "loss": 1.0301, "step": 8059 }, { "epoch": 0.41, "learning_rate": 1.3352166366402246e-05, "loss": 1.0584, "step": 8060 }, { "epoch": 0.41, "learning_rate": 1.3350616107087713e-05, "loss": 0.9623, "step": 8061 }, { "epoch": 0.41, "learning_rate": 1.334906575705881e-05, "loss": 0.8647, "step": 8062 }, { "epoch": 0.41, "learning_rate": 1.3347515316357514e-05, "loss": 0.9292, "step": 8063 }, { "epoch": 0.41, "learning_rate": 1.3345964785025803e-05, "loss": 0.9335, "step": 8064 }, { "epoch": 0.41, "learning_rate": 1.3344414163105658e-05, "loss": 0.9492, "step": 8065 }, { "epoch": 0.41, "learning_rate": 1.3342863450639052e-05, "loss": 1.0057, "step": 8066 }, { "epoch": 0.41, "learning_rate": 1.3341312647667982e-05, "loss": 0.942, "step": 8067 }, { "epoch": 0.41, "learning_rate": 1.3339761754234422e-05, "loss": 0.9632, "step": 8068 }, { "epoch": 0.41, "learning_rate": 1.3338210770380366e-05, "loss": 1.0045, "step": 8069 }, { "epoch": 0.41, "learning_rate": 1.3336659696147804e-05, "loss": 0.9117, "step": 8070 }, { "epoch": 0.41, "learning_rate": 1.3335108531578734e-05, "loss": 0.9924, "step": 8071 }, { "epoch": 0.41, "learning_rate": 1.3333557276715145e-05, "loss": 0.944, "step": 8072 }, { "epoch": 0.41, "learning_rate": 1.3332005931599044e-05, "loss": 0.8884, "step": 8073 }, { "epoch": 0.41, "learning_rate": 1.3330454496272423e-05, "loss": 1.031, "step": 8074 }, { "epoch": 0.41, "learning_rate": 1.3328902970777292e-05, "loss": 0.985, "step": 8075 }, { "epoch": 0.41, "learning_rate": 1.3327351355155656e-05, "loss": 0.9921, "step": 8076 }, { "epoch": 0.41, "learning_rate": 1.3325799649449518e-05, "loss": 1.0126, "step": 8077 }, { "epoch": 0.41, "learning_rate": 1.3324247853700898e-05, "loss": 0.9727, "step": 8078 }, { "epoch": 0.41, "learning_rate": 1.3322695967951801e-05, "loss": 0.9088, "step": 8079 }, { "epoch": 0.41, "learning_rate": 1.3321143992244245e-05, "loss": 0.8433, "step": 8080 }, { "epoch": 0.41, "learning_rate": 1.331959192662025e-05, "loss": 0.9229, "step": 8081 }, { "epoch": 0.41, "learning_rate": 1.3318039771121836e-05, "loss": 0.9761, "step": 8082 }, { "epoch": 0.41, "learning_rate": 1.3316487525791023e-05, "loss": 0.888, "step": 8083 }, { "epoch": 0.41, "learning_rate": 1.3314935190669845e-05, "loss": 1.0526, "step": 8084 }, { "epoch": 0.41, "learning_rate": 1.3313382765800315e-05, "loss": 0.9232, "step": 8085 }, { "epoch": 0.41, "learning_rate": 1.3311830251224475e-05, "loss": 0.7977, "step": 8086 }, { "epoch": 0.41, "learning_rate": 1.3310277646984357e-05, "loss": 0.8475, "step": 8087 }, { "epoch": 0.41, "learning_rate": 1.3308724953121992e-05, "loss": 0.9551, "step": 8088 }, { "epoch": 0.41, "learning_rate": 1.3307172169679416e-05, "loss": 0.9564, "step": 8089 }, { "epoch": 0.41, "learning_rate": 1.3305619296698677e-05, "loss": 0.9088, "step": 8090 }, { "epoch": 0.41, "learning_rate": 1.330406633422181e-05, "loss": 0.8982, "step": 8091 }, { "epoch": 0.41, "learning_rate": 1.3302513282290862e-05, "loss": 0.9253, "step": 8092 }, { "epoch": 0.41, "learning_rate": 1.3300960140947882e-05, "loss": 0.9633, "step": 8093 }, { "epoch": 0.41, "learning_rate": 1.3299406910234917e-05, "loss": 0.9617, "step": 8094 }, { "epoch": 0.41, "learning_rate": 1.329785359019402e-05, "loss": 0.9443, "step": 8095 }, { "epoch": 0.41, "learning_rate": 1.329630018086725e-05, "loss": 0.9709, "step": 8096 }, { "epoch": 0.41, "learning_rate": 1.3294746682296653e-05, "loss": 0.947, "step": 8097 }, { "epoch": 0.41, "learning_rate": 1.3293193094524297e-05, "loss": 0.9781, "step": 8098 }, { "epoch": 0.41, "learning_rate": 1.3291639417592244e-05, "loss": 0.9684, "step": 8099 }, { "epoch": 0.41, "learning_rate": 1.3290085651542553e-05, "loss": 0.9468, "step": 8100 }, { "epoch": 0.41, "learning_rate": 1.3288531796417292e-05, "loss": 0.8757, "step": 8101 }, { "epoch": 0.41, "learning_rate": 1.3286977852258538e-05, "loss": 0.8833, "step": 8102 }, { "epoch": 0.41, "learning_rate": 1.3285423819108349e-05, "loss": 0.9926, "step": 8103 }, { "epoch": 0.41, "learning_rate": 1.3283869697008806e-05, "loss": 0.9206, "step": 8104 }, { "epoch": 0.41, "learning_rate": 1.3282315486001988e-05, "loss": 1.0141, "step": 8105 }, { "epoch": 0.41, "learning_rate": 1.3280761186129969e-05, "loss": 0.9253, "step": 8106 }, { "epoch": 0.41, "learning_rate": 1.3279206797434831e-05, "loss": 0.9684, "step": 8107 }, { "epoch": 0.41, "learning_rate": 1.3277652319958658e-05, "loss": 0.9186, "step": 8108 }, { "epoch": 0.41, "learning_rate": 1.3276097753743537e-05, "loss": 0.9121, "step": 8109 }, { "epoch": 0.41, "learning_rate": 1.3274543098831551e-05, "loss": 0.8752, "step": 8110 }, { "epoch": 0.41, "learning_rate": 1.3272988355264797e-05, "loss": 0.8683, "step": 8111 }, { "epoch": 0.41, "learning_rate": 1.3271433523085366e-05, "loss": 0.8896, "step": 8112 }, { "epoch": 0.41, "learning_rate": 1.3269878602335355e-05, "loss": 0.9401, "step": 8113 }, { "epoch": 0.41, "learning_rate": 1.3268323593056856e-05, "loss": 0.8733, "step": 8114 }, { "epoch": 0.41, "learning_rate": 1.3266768495291974e-05, "loss": 0.9626, "step": 8115 }, { "epoch": 0.41, "learning_rate": 1.3265213309082809e-05, "loss": 0.9143, "step": 8116 }, { "epoch": 0.41, "learning_rate": 1.326365803447147e-05, "loss": 0.945, "step": 8117 }, { "epoch": 0.41, "learning_rate": 1.3262102671500062e-05, "loss": 1.0082, "step": 8118 }, { "epoch": 0.41, "learning_rate": 1.3260547220210694e-05, "loss": 0.8651, "step": 8119 }, { "epoch": 0.41, "learning_rate": 1.3258991680645478e-05, "loss": 1.0001, "step": 8120 }, { "epoch": 0.41, "learning_rate": 1.3257436052846531e-05, "loss": 0.9495, "step": 8121 }, { "epoch": 0.41, "learning_rate": 1.3255880336855967e-05, "loss": 1.0085, "step": 8122 }, { "epoch": 0.41, "learning_rate": 1.3254324532715907e-05, "loss": 0.8925, "step": 8123 }, { "epoch": 0.41, "learning_rate": 1.3252768640468475e-05, "loss": 1.0068, "step": 8124 }, { "epoch": 0.41, "learning_rate": 1.3251212660155792e-05, "loss": 0.9996, "step": 8125 }, { "epoch": 0.41, "learning_rate": 1.3249656591819981e-05, "loss": 0.9296, "step": 8126 }, { "epoch": 0.41, "learning_rate": 1.324810043550318e-05, "loss": 0.9686, "step": 8127 }, { "epoch": 0.41, "learning_rate": 1.3246544191247512e-05, "loss": 1.0897, "step": 8128 }, { "epoch": 0.41, "learning_rate": 1.3244987859095115e-05, "loss": 0.8539, "step": 8129 }, { "epoch": 0.41, "learning_rate": 1.3243431439088128e-05, "loss": 0.8654, "step": 8130 }, { "epoch": 0.41, "learning_rate": 1.3241874931268683e-05, "loss": 0.9051, "step": 8131 }, { "epoch": 0.41, "learning_rate": 1.3240318335678922e-05, "loss": 1.0046, "step": 8132 }, { "epoch": 0.41, "learning_rate": 1.3238761652360992e-05, "loss": 1.0179, "step": 8133 }, { "epoch": 0.41, "learning_rate": 1.3237204881357033e-05, "loss": 0.8922, "step": 8134 }, { "epoch": 0.41, "learning_rate": 1.3235648022709196e-05, "loss": 0.8976, "step": 8135 }, { "epoch": 0.41, "learning_rate": 1.3234091076459634e-05, "loss": 0.9371, "step": 8136 }, { "epoch": 0.41, "learning_rate": 1.3232534042650496e-05, "loss": 1.0383, "step": 8137 }, { "epoch": 0.41, "learning_rate": 1.3230976921323937e-05, "loss": 0.9222, "step": 8138 }, { "epoch": 0.41, "learning_rate": 1.3229419712522116e-05, "loss": 0.8804, "step": 8139 }, { "epoch": 0.41, "learning_rate": 1.3227862416287192e-05, "loss": 0.8666, "step": 8140 }, { "epoch": 0.41, "learning_rate": 1.3226305032661326e-05, "loss": 0.938, "step": 8141 }, { "epoch": 0.41, "learning_rate": 1.3224747561686686e-05, "loss": 0.915, "step": 8142 }, { "epoch": 0.41, "learning_rate": 1.3223190003405435e-05, "loss": 0.9368, "step": 8143 }, { "epoch": 0.41, "learning_rate": 1.3221632357859745e-05, "loss": 0.9723, "step": 8144 }, { "epoch": 0.41, "learning_rate": 1.3220074625091783e-05, "loss": 0.9549, "step": 8145 }, { "epoch": 0.41, "learning_rate": 1.321851680514373e-05, "loss": 0.9904, "step": 8146 }, { "epoch": 0.41, "learning_rate": 1.3216958898057759e-05, "loss": 0.986, "step": 8147 }, { "epoch": 0.41, "learning_rate": 1.3215400903876048e-05, "loss": 0.9533, "step": 8148 }, { "epoch": 0.41, "learning_rate": 1.3213842822640779e-05, "loss": 0.9899, "step": 8149 }, { "epoch": 0.41, "learning_rate": 1.3212284654394131e-05, "loss": 1.0042, "step": 8150 }, { "epoch": 0.41, "learning_rate": 1.3210726399178298e-05, "loss": 1.0184, "step": 8151 }, { "epoch": 0.41, "learning_rate": 1.3209168057035461e-05, "loss": 0.9432, "step": 8152 }, { "epoch": 0.41, "learning_rate": 1.3207609628007813e-05, "loss": 0.9075, "step": 8153 }, { "epoch": 0.41, "learning_rate": 1.320605111213755e-05, "loss": 0.9695, "step": 8154 }, { "epoch": 0.41, "learning_rate": 1.3204492509466862e-05, "loss": 0.9173, "step": 8155 }, { "epoch": 0.41, "learning_rate": 1.3202933820037948e-05, "loss": 1.0519, "step": 8156 }, { "epoch": 0.41, "learning_rate": 1.320137504389301e-05, "loss": 0.8591, "step": 8157 }, { "epoch": 0.41, "learning_rate": 1.3199816181074247e-05, "loss": 0.9248, "step": 8158 }, { "epoch": 0.41, "learning_rate": 1.3198257231623866e-05, "loss": 0.7918, "step": 8159 }, { "epoch": 0.41, "learning_rate": 1.3196698195584075e-05, "loss": 0.9446, "step": 8160 }, { "epoch": 0.41, "learning_rate": 1.3195139072997081e-05, "loss": 1.0807, "step": 8161 }, { "epoch": 0.41, "learning_rate": 1.3193579863905092e-05, "loss": 0.9382, "step": 8162 }, { "epoch": 0.41, "learning_rate": 1.3192020568350327e-05, "loss": 0.9459, "step": 8163 }, { "epoch": 0.41, "learning_rate": 1.3190461186375006e-05, "loss": 1.0297, "step": 8164 }, { "epoch": 0.41, "learning_rate": 1.3188901718021336e-05, "loss": 0.8728, "step": 8165 }, { "epoch": 0.41, "learning_rate": 1.3187342163331552e-05, "loss": 0.9429, "step": 8166 }, { "epoch": 0.41, "learning_rate": 1.3185782522347865e-05, "loss": 0.8851, "step": 8167 }, { "epoch": 0.41, "learning_rate": 1.3184222795112506e-05, "loss": 0.997, "step": 8168 }, { "epoch": 0.42, "learning_rate": 1.3182662981667704e-05, "loss": 0.9775, "step": 8169 }, { "epoch": 0.42, "learning_rate": 1.3181103082055686e-05, "loss": 0.9901, "step": 8170 }, { "epoch": 0.42, "learning_rate": 1.3179543096318687e-05, "loss": 0.8676, "step": 8171 }, { "epoch": 0.42, "learning_rate": 1.3177983024498944e-05, "loss": 1.0048, "step": 8172 }, { "epoch": 0.42, "learning_rate": 1.317642286663869e-05, "loss": 0.8392, "step": 8173 }, { "epoch": 0.42, "learning_rate": 1.3174862622780164e-05, "loss": 0.9482, "step": 8174 }, { "epoch": 0.42, "learning_rate": 1.3173302292965613e-05, "loss": 0.9657, "step": 8175 }, { "epoch": 0.42, "learning_rate": 1.3171741877237278e-05, "loss": 0.863, "step": 8176 }, { "epoch": 0.42, "learning_rate": 1.3170181375637405e-05, "loss": 0.996, "step": 8177 }, { "epoch": 0.42, "learning_rate": 1.3168620788208248e-05, "loss": 0.8489, "step": 8178 }, { "epoch": 0.42, "learning_rate": 1.316706011499205e-05, "loss": 0.9717, "step": 8179 }, { "epoch": 0.42, "learning_rate": 1.3165499356031074e-05, "loss": 0.9301, "step": 8180 }, { "epoch": 0.42, "learning_rate": 1.3163938511367565e-05, "loss": 0.9345, "step": 8181 }, { "epoch": 0.42, "learning_rate": 1.3162377581043785e-05, "loss": 0.9332, "step": 8182 }, { "epoch": 0.42, "learning_rate": 1.3160816565102003e-05, "loss": 1.0388, "step": 8183 }, { "epoch": 0.42, "learning_rate": 1.3159255463584474e-05, "loss": 0.9385, "step": 8184 }, { "epoch": 0.42, "learning_rate": 1.3157694276533462e-05, "loss": 0.9878, "step": 8185 }, { "epoch": 0.42, "learning_rate": 1.315613300399124e-05, "loss": 0.8765, "step": 8186 }, { "epoch": 0.42, "learning_rate": 1.3154571646000072e-05, "loss": 0.9187, "step": 8187 }, { "epoch": 0.42, "learning_rate": 1.3153010202602232e-05, "loss": 0.9501, "step": 8188 }, { "epoch": 0.42, "learning_rate": 1.3151448673839999e-05, "loss": 1.0902, "step": 8189 }, { "epoch": 0.42, "learning_rate": 1.3149887059755646e-05, "loss": 0.9495, "step": 8190 }, { "epoch": 0.42, "learning_rate": 1.3148325360391449e-05, "loss": 0.9176, "step": 8191 }, { "epoch": 0.42, "learning_rate": 1.3146763575789695e-05, "loss": 1.0075, "step": 8192 }, { "epoch": 0.42, "learning_rate": 1.3145201705992663e-05, "loss": 0.9218, "step": 8193 }, { "epoch": 0.42, "learning_rate": 1.3143639751042643e-05, "loss": 0.9878, "step": 8194 }, { "epoch": 0.42, "learning_rate": 1.314207771098192e-05, "loss": 0.9289, "step": 8195 }, { "epoch": 0.42, "learning_rate": 1.3140515585852786e-05, "loss": 0.8717, "step": 8196 }, { "epoch": 0.42, "learning_rate": 1.3138953375697534e-05, "loss": 0.899, "step": 8197 }, { "epoch": 0.42, "learning_rate": 1.3137391080558462e-05, "loss": 1.0196, "step": 8198 }, { "epoch": 0.42, "learning_rate": 1.3135828700477857e-05, "loss": 0.8595, "step": 8199 }, { "epoch": 0.42, "learning_rate": 1.3134266235498032e-05, "loss": 0.9198, "step": 8200 }, { "epoch": 0.42, "learning_rate": 1.313270368566128e-05, "loss": 0.9047, "step": 8201 }, { "epoch": 0.42, "learning_rate": 1.3131141051009912e-05, "loss": 0.9553, "step": 8202 }, { "epoch": 0.42, "learning_rate": 1.312957833158623e-05, "loss": 0.9071, "step": 8203 }, { "epoch": 0.42, "learning_rate": 1.3128015527432545e-05, "loss": 0.9199, "step": 8204 }, { "epoch": 0.42, "learning_rate": 1.3126452638591167e-05, "loss": 0.9839, "step": 8205 }, { "epoch": 0.42, "learning_rate": 1.3124889665104407e-05, "loss": 0.951, "step": 8206 }, { "epoch": 0.42, "learning_rate": 1.3123326607014587e-05, "loss": 0.9288, "step": 8207 }, { "epoch": 0.42, "learning_rate": 1.3121763464364023e-05, "loss": 1.0041, "step": 8208 }, { "epoch": 0.42, "learning_rate": 1.3120200237195033e-05, "loss": 0.9565, "step": 8209 }, { "epoch": 0.42, "learning_rate": 1.311863692554994e-05, "loss": 0.9123, "step": 8210 }, { "epoch": 0.42, "learning_rate": 1.3117073529471074e-05, "loss": 1.0026, "step": 8211 }, { "epoch": 0.42, "learning_rate": 1.3115510049000755e-05, "loss": 1.0618, "step": 8212 }, { "epoch": 0.42, "learning_rate": 1.3113946484181316e-05, "loss": 0.9469, "step": 8213 }, { "epoch": 0.42, "learning_rate": 1.3112382835055092e-05, "loss": 0.8735, "step": 8214 }, { "epoch": 0.42, "learning_rate": 1.3110819101664411e-05, "loss": 0.932, "step": 8215 }, { "epoch": 0.42, "learning_rate": 1.3109255284051615e-05, "loss": 0.9326, "step": 8216 }, { "epoch": 0.42, "learning_rate": 1.3107691382259037e-05, "loss": 0.9184, "step": 8217 }, { "epoch": 0.42, "learning_rate": 1.3106127396329022e-05, "loss": 0.9123, "step": 8218 }, { "epoch": 0.42, "learning_rate": 1.3104563326303915e-05, "loss": 0.9128, "step": 8219 }, { "epoch": 0.42, "learning_rate": 1.310299917222606e-05, "loss": 1.0505, "step": 8220 }, { "epoch": 0.42, "learning_rate": 1.3101434934137798e-05, "loss": 0.7972, "step": 8221 }, { "epoch": 0.42, "learning_rate": 1.3099870612081489e-05, "loss": 0.9178, "step": 8222 }, { "epoch": 0.42, "learning_rate": 1.309830620609948e-05, "loss": 0.9245, "step": 8223 }, { "epoch": 0.42, "learning_rate": 1.3096741716234125e-05, "loss": 0.9156, "step": 8224 }, { "epoch": 0.42, "learning_rate": 1.3095177142527786e-05, "loss": 0.8771, "step": 8225 }, { "epoch": 0.42, "learning_rate": 1.309361248502282e-05, "loss": 0.9801, "step": 8226 }, { "epoch": 0.42, "learning_rate": 1.3092047743761582e-05, "loss": 1.0453, "step": 8227 }, { "epoch": 0.42, "learning_rate": 1.3090482918786442e-05, "loss": 1.0108, "step": 8228 }, { "epoch": 0.42, "learning_rate": 1.3088918010139769e-05, "loss": 0.9516, "step": 8229 }, { "epoch": 0.42, "learning_rate": 1.3087353017863924e-05, "loss": 0.95, "step": 8230 }, { "epoch": 0.42, "learning_rate": 1.3085787942001282e-05, "loss": 0.9161, "step": 8231 }, { "epoch": 0.42, "learning_rate": 1.3084222782594215e-05, "loss": 0.9759, "step": 8232 }, { "epoch": 0.42, "learning_rate": 1.3082657539685096e-05, "loss": 0.9736, "step": 8233 }, { "epoch": 0.42, "learning_rate": 1.3081092213316304e-05, "loss": 1.0208, "step": 8234 }, { "epoch": 0.42, "learning_rate": 1.307952680353022e-05, "loss": 0.9216, "step": 8235 }, { "epoch": 0.42, "learning_rate": 1.307796131036922e-05, "loss": 0.95, "step": 8236 }, { "epoch": 0.42, "learning_rate": 1.3076395733875697e-05, "loss": 1.0157, "step": 8237 }, { "epoch": 0.42, "learning_rate": 1.3074830074092033e-05, "loss": 0.8813, "step": 8238 }, { "epoch": 0.42, "learning_rate": 1.3073264331060614e-05, "loss": 0.9167, "step": 8239 }, { "epoch": 0.42, "learning_rate": 1.307169850482383e-05, "loss": 0.9675, "step": 8240 }, { "epoch": 0.42, "learning_rate": 1.3070132595424082e-05, "loss": 0.904, "step": 8241 }, { "epoch": 0.42, "learning_rate": 1.3068566602903757e-05, "loss": 0.9243, "step": 8242 }, { "epoch": 0.42, "learning_rate": 1.3067000527305257e-05, "loss": 0.8976, "step": 8243 }, { "epoch": 0.42, "learning_rate": 1.3065434368670982e-05, "loss": 0.9474, "step": 8244 }, { "epoch": 0.42, "learning_rate": 1.306386812704333e-05, "loss": 0.9583, "step": 8245 }, { "epoch": 0.42, "learning_rate": 1.306230180246471e-05, "loss": 0.9559, "step": 8246 }, { "epoch": 0.42, "learning_rate": 1.3060735394977524e-05, "loss": 1.0269, "step": 8247 }, { "epoch": 0.42, "learning_rate": 1.3059168904624189e-05, "loss": 0.9696, "step": 8248 }, { "epoch": 0.42, "learning_rate": 1.3057602331447108e-05, "loss": 1.0219, "step": 8249 }, { "epoch": 0.42, "learning_rate": 1.3056035675488696e-05, "loss": 0.9806, "step": 8250 }, { "epoch": 0.42, "learning_rate": 1.3054468936791371e-05, "loss": 0.9399, "step": 8251 }, { "epoch": 0.42, "learning_rate": 1.3052902115397548e-05, "loss": 0.9342, "step": 8252 }, { "epoch": 0.42, "learning_rate": 1.3051335211349647e-05, "loss": 0.9083, "step": 8253 }, { "epoch": 0.42, "learning_rate": 1.3049768224690093e-05, "loss": 0.8459, "step": 8254 }, { "epoch": 0.42, "learning_rate": 1.3048201155461311e-05, "loss": 0.9398, "step": 8255 }, { "epoch": 0.42, "learning_rate": 1.3046634003705723e-05, "loss": 0.8752, "step": 8256 }, { "epoch": 0.42, "learning_rate": 1.3045066769465761e-05, "loss": 1.0034, "step": 8257 }, { "epoch": 0.42, "learning_rate": 1.3043499452783856e-05, "loss": 0.8978, "step": 8258 }, { "epoch": 0.42, "learning_rate": 1.3041932053702442e-05, "loss": 0.9295, "step": 8259 }, { "epoch": 0.42, "learning_rate": 1.3040364572263954e-05, "loss": 0.8935, "step": 8260 }, { "epoch": 0.42, "learning_rate": 1.3038797008510831e-05, "loss": 0.926, "step": 8261 }, { "epoch": 0.42, "learning_rate": 1.3037229362485509e-05, "loss": 0.9096, "step": 8262 }, { "epoch": 0.42, "learning_rate": 1.3035661634230437e-05, "loss": 0.9531, "step": 8263 }, { "epoch": 0.42, "learning_rate": 1.3034093823788053e-05, "loss": 0.9866, "step": 8264 }, { "epoch": 0.42, "learning_rate": 1.3032525931200808e-05, "loss": 0.9127, "step": 8265 }, { "epoch": 0.42, "learning_rate": 1.303095795651115e-05, "loss": 0.8963, "step": 8266 }, { "epoch": 0.42, "learning_rate": 1.3029389899761528e-05, "loss": 0.9257, "step": 8267 }, { "epoch": 0.42, "learning_rate": 1.30278217609944e-05, "loss": 0.9015, "step": 8268 }, { "epoch": 0.42, "learning_rate": 1.3026253540252219e-05, "loss": 0.9686, "step": 8269 }, { "epoch": 0.42, "learning_rate": 1.302468523757744e-05, "loss": 0.9621, "step": 8270 }, { "epoch": 0.42, "learning_rate": 1.302311685301253e-05, "loss": 0.8824, "step": 8271 }, { "epoch": 0.42, "learning_rate": 1.3021548386599947e-05, "loss": 0.9711, "step": 8272 }, { "epoch": 0.42, "learning_rate": 1.3019979838382156e-05, "loss": 0.834, "step": 8273 }, { "epoch": 0.42, "learning_rate": 1.301841120840162e-05, "loss": 0.8678, "step": 8274 }, { "epoch": 0.42, "learning_rate": 1.3016842496700817e-05, "loss": 1.0095, "step": 8275 }, { "epoch": 0.42, "learning_rate": 1.301527370332221e-05, "loss": 0.9695, "step": 8276 }, { "epoch": 0.42, "learning_rate": 1.3013704828308276e-05, "loss": 0.9851, "step": 8277 }, { "epoch": 0.42, "learning_rate": 1.3012135871701492e-05, "loss": 0.9107, "step": 8278 }, { "epoch": 0.42, "learning_rate": 1.3010566833544331e-05, "loss": 0.9691, "step": 8279 }, { "epoch": 0.42, "learning_rate": 1.3008997713879277e-05, "loss": 0.9299, "step": 8280 }, { "epoch": 0.42, "learning_rate": 1.3007428512748811e-05, "loss": 0.9338, "step": 8281 }, { "epoch": 0.42, "learning_rate": 1.3005859230195421e-05, "loss": 0.9047, "step": 8282 }, { "epoch": 0.42, "learning_rate": 1.3004289866261586e-05, "loss": 0.8614, "step": 8283 }, { "epoch": 0.42, "learning_rate": 1.3002720420989801e-05, "loss": 0.8626, "step": 8284 }, { "epoch": 0.42, "learning_rate": 1.3001150894422557e-05, "loss": 0.9052, "step": 8285 }, { "epoch": 0.42, "learning_rate": 1.2999581286602342e-05, "loss": 0.9833, "step": 8286 }, { "epoch": 0.42, "learning_rate": 1.2998011597571656e-05, "loss": 0.9202, "step": 8287 }, { "epoch": 0.42, "learning_rate": 1.2996441827372996e-05, "loss": 0.9481, "step": 8288 }, { "epoch": 0.42, "learning_rate": 1.2994871976048861e-05, "loss": 1.0674, "step": 8289 }, { "epoch": 0.42, "learning_rate": 1.2993302043641755e-05, "loss": 0.9474, "step": 8290 }, { "epoch": 0.42, "learning_rate": 1.2991732030194182e-05, "loss": 0.8817, "step": 8291 }, { "epoch": 0.42, "learning_rate": 1.2990161935748644e-05, "loss": 0.8627, "step": 8292 }, { "epoch": 0.42, "learning_rate": 1.2988591760347653e-05, "loss": 0.9152, "step": 8293 }, { "epoch": 0.42, "learning_rate": 1.2987021504033722e-05, "loss": 0.9349, "step": 8294 }, { "epoch": 0.42, "learning_rate": 1.2985451166849359e-05, "loss": 0.961, "step": 8295 }, { "epoch": 0.42, "learning_rate": 1.2983880748837085e-05, "loss": 0.9659, "step": 8296 }, { "epoch": 0.42, "learning_rate": 1.2982310250039413e-05, "loss": 0.8876, "step": 8297 }, { "epoch": 0.42, "learning_rate": 1.2980739670498864e-05, "loss": 0.9866, "step": 8298 }, { "epoch": 0.42, "learning_rate": 1.2979169010257956e-05, "loss": 0.9598, "step": 8299 }, { "epoch": 0.42, "learning_rate": 1.2977598269359224e-05, "loss": 0.9204, "step": 8300 }, { "epoch": 0.42, "learning_rate": 1.297602744784518e-05, "loss": 0.9886, "step": 8301 }, { "epoch": 0.42, "learning_rate": 1.2974456545758365e-05, "loss": 0.9854, "step": 8302 }, { "epoch": 0.42, "learning_rate": 1.29728855631413e-05, "loss": 0.996, "step": 8303 }, { "epoch": 0.42, "learning_rate": 1.2971314500036521e-05, "loss": 0.8252, "step": 8304 }, { "epoch": 0.42, "learning_rate": 1.2969743356486564e-05, "loss": 1.004, "step": 8305 }, { "epoch": 0.42, "learning_rate": 1.2968172132533965e-05, "loss": 0.9611, "step": 8306 }, { "epoch": 0.42, "learning_rate": 1.2966600828221262e-05, "loss": 0.846, "step": 8307 }, { "epoch": 0.42, "learning_rate": 1.2965029443591002e-05, "loss": 1.1032, "step": 8308 }, { "epoch": 0.42, "learning_rate": 1.2963457978685723e-05, "loss": 0.9483, "step": 8309 }, { "epoch": 0.42, "learning_rate": 1.296188643354797e-05, "loss": 0.9075, "step": 8310 }, { "epoch": 0.42, "learning_rate": 1.2960314808220292e-05, "loss": 0.9255, "step": 8311 }, { "epoch": 0.42, "learning_rate": 1.2958743102745242e-05, "loss": 0.9431, "step": 8312 }, { "epoch": 0.42, "learning_rate": 1.2957171317165369e-05, "loss": 0.9253, "step": 8313 }, { "epoch": 0.42, "learning_rate": 1.295559945152323e-05, "loss": 0.9741, "step": 8314 }, { "epoch": 0.42, "learning_rate": 1.295402750586138e-05, "loss": 0.9285, "step": 8315 }, { "epoch": 0.42, "learning_rate": 1.2952455480222379e-05, "loss": 0.9454, "step": 8316 }, { "epoch": 0.42, "learning_rate": 1.2950883374648787e-05, "loss": 0.8969, "step": 8317 }, { "epoch": 0.42, "learning_rate": 1.2949311189183165e-05, "loss": 0.9603, "step": 8318 }, { "epoch": 0.42, "learning_rate": 1.2947738923868082e-05, "loss": 0.9404, "step": 8319 }, { "epoch": 0.42, "learning_rate": 1.2946166578746107e-05, "loss": 1.0299, "step": 8320 }, { "epoch": 0.42, "learning_rate": 1.2944594153859801e-05, "loss": 0.948, "step": 8321 }, { "epoch": 0.42, "learning_rate": 1.2943021649251745e-05, "loss": 0.9977, "step": 8322 }, { "epoch": 0.42, "learning_rate": 1.2941449064964507e-05, "loss": 0.9684, "step": 8323 }, { "epoch": 0.42, "learning_rate": 1.2939876401040663e-05, "loss": 0.9625, "step": 8324 }, { "epoch": 0.42, "learning_rate": 1.2938303657522794e-05, "loss": 0.9253, "step": 8325 }, { "epoch": 0.42, "learning_rate": 1.2936730834453484e-05, "loss": 0.9729, "step": 8326 }, { "epoch": 0.42, "learning_rate": 1.2935157931875306e-05, "loss": 0.8648, "step": 8327 }, { "epoch": 0.42, "learning_rate": 1.2933584949830853e-05, "loss": 0.82, "step": 8328 }, { "epoch": 0.42, "learning_rate": 1.2932011888362708e-05, "loss": 0.9992, "step": 8329 }, { "epoch": 0.42, "learning_rate": 1.2930438747513459e-05, "loss": 1.0015, "step": 8330 }, { "epoch": 0.42, "learning_rate": 1.29288655273257e-05, "loss": 0.9327, "step": 8331 }, { "epoch": 0.42, "learning_rate": 1.2927292227842022e-05, "loss": 0.9283, "step": 8332 }, { "epoch": 0.42, "learning_rate": 1.2925718849105022e-05, "loss": 0.8819, "step": 8333 }, { "epoch": 0.42, "learning_rate": 1.29241453911573e-05, "loss": 0.9692, "step": 8334 }, { "epoch": 0.42, "learning_rate": 1.2922571854041447e-05, "loss": 0.9958, "step": 8335 }, { "epoch": 0.42, "learning_rate": 1.2920998237800072e-05, "loss": 0.9498, "step": 8336 }, { "epoch": 0.42, "learning_rate": 1.2919424542475777e-05, "loss": 0.8871, "step": 8337 }, { "epoch": 0.42, "learning_rate": 1.2917850768111171e-05, "loss": 0.9167, "step": 8338 }, { "epoch": 0.42, "learning_rate": 1.291627691474886e-05, "loss": 0.9174, "step": 8339 }, { "epoch": 0.42, "learning_rate": 1.2914702982431453e-05, "loss": 0.8867, "step": 8340 }, { "epoch": 0.42, "learning_rate": 1.291312897120156e-05, "loss": 1.0456, "step": 8341 }, { "epoch": 0.42, "learning_rate": 1.2911554881101804e-05, "loss": 0.9006, "step": 8342 }, { "epoch": 0.42, "learning_rate": 1.2909980712174796e-05, "loss": 1.0092, "step": 8343 }, { "epoch": 0.42, "learning_rate": 1.2908406464463159e-05, "loss": 0.9266, "step": 8344 }, { "epoch": 0.42, "learning_rate": 1.2906832138009506e-05, "loss": 0.9849, "step": 8345 }, { "epoch": 0.42, "learning_rate": 1.290525773285647e-05, "loss": 0.8258, "step": 8346 }, { "epoch": 0.42, "learning_rate": 1.2903683249046668e-05, "loss": 1.067, "step": 8347 }, { "epoch": 0.42, "learning_rate": 1.2902108686622734e-05, "loss": 1.0041, "step": 8348 }, { "epoch": 0.42, "learning_rate": 1.2900534045627296e-05, "loss": 0.9672, "step": 8349 }, { "epoch": 0.42, "learning_rate": 1.2898959326102983e-05, "loss": 1.034, "step": 8350 }, { "epoch": 0.42, "learning_rate": 1.289738452809243e-05, "loss": 0.9526, "step": 8351 }, { "epoch": 0.42, "learning_rate": 1.2895809651638271e-05, "loss": 1.0405, "step": 8352 }, { "epoch": 0.42, "learning_rate": 1.289423469678315e-05, "loss": 0.9567, "step": 8353 }, { "epoch": 0.42, "learning_rate": 1.2892659663569703e-05, "loss": 0.8965, "step": 8354 }, { "epoch": 0.42, "learning_rate": 1.2891084552040573e-05, "loss": 0.9842, "step": 8355 }, { "epoch": 0.42, "learning_rate": 1.2889509362238406e-05, "loss": 0.8021, "step": 8356 }, { "epoch": 0.42, "learning_rate": 1.2887934094205844e-05, "loss": 0.9006, "step": 8357 }, { "epoch": 0.42, "learning_rate": 1.2886358747985542e-05, "loss": 0.9823, "step": 8358 }, { "epoch": 0.42, "learning_rate": 1.2884783323620146e-05, "loss": 0.9589, "step": 8359 }, { "epoch": 0.42, "learning_rate": 1.2883207821152313e-05, "loss": 1.1215, "step": 8360 }, { "epoch": 0.42, "learning_rate": 1.2881632240624696e-05, "loss": 0.9221, "step": 8361 }, { "epoch": 0.42, "learning_rate": 1.2880056582079949e-05, "loss": 0.8663, "step": 8362 }, { "epoch": 0.42, "learning_rate": 1.2878480845560735e-05, "loss": 0.9323, "step": 8363 }, { "epoch": 0.42, "learning_rate": 1.2876905031109712e-05, "loss": 0.9065, "step": 8364 }, { "epoch": 0.42, "learning_rate": 1.2875329138769553e-05, "loss": 0.9101, "step": 8365 }, { "epoch": 0.43, "learning_rate": 1.287375316858291e-05, "loss": 0.8669, "step": 8366 }, { "epoch": 0.43, "learning_rate": 1.2872177120592462e-05, "loss": 0.9097, "step": 8367 }, { "epoch": 0.43, "learning_rate": 1.2870600994840871e-05, "loss": 0.9869, "step": 8368 }, { "epoch": 0.43, "learning_rate": 1.2869024791370815e-05, "loss": 0.8451, "step": 8369 }, { "epoch": 0.43, "learning_rate": 1.2867448510224962e-05, "loss": 0.905, "step": 8370 }, { "epoch": 0.43, "learning_rate": 1.2865872151445993e-05, "loss": 0.855, "step": 8371 }, { "epoch": 0.43, "learning_rate": 1.2864295715076582e-05, "loss": 0.9809, "step": 8372 }, { "epoch": 0.43, "learning_rate": 1.2862719201159417e-05, "loss": 0.9899, "step": 8373 }, { "epoch": 0.43, "learning_rate": 1.2861142609737173e-05, "loss": 0.9589, "step": 8374 }, { "epoch": 0.43, "learning_rate": 1.2859565940852536e-05, "loss": 0.911, "step": 8375 }, { "epoch": 0.43, "learning_rate": 1.2857989194548192e-05, "loss": 0.9007, "step": 8376 }, { "epoch": 0.43, "learning_rate": 1.2856412370866836e-05, "loss": 0.9609, "step": 8377 }, { "epoch": 0.43, "learning_rate": 1.2854835469851148e-05, "loss": 0.9459, "step": 8378 }, { "epoch": 0.43, "learning_rate": 1.2853258491543832e-05, "loss": 0.8909, "step": 8379 }, { "epoch": 0.43, "learning_rate": 1.2851681435987578e-05, "loss": 0.9677, "step": 8380 }, { "epoch": 0.43, "learning_rate": 1.285010430322508e-05, "loss": 0.9409, "step": 8381 }, { "epoch": 0.43, "learning_rate": 1.2848527093299039e-05, "loss": 0.942, "step": 8382 }, { "epoch": 0.43, "learning_rate": 1.2846949806252163e-05, "loss": 0.9734, "step": 8383 }, { "epoch": 0.43, "learning_rate": 1.2845372442127144e-05, "loss": 0.9236, "step": 8384 }, { "epoch": 0.43, "learning_rate": 1.2843795000966699e-05, "loss": 0.9364, "step": 8385 }, { "epoch": 0.43, "learning_rate": 1.2842217482813525e-05, "loss": 0.9017, "step": 8386 }, { "epoch": 0.43, "learning_rate": 1.284063988771034e-05, "loss": 1.0458, "step": 8387 }, { "epoch": 0.43, "learning_rate": 1.2839062215699852e-05, "loss": 0.9589, "step": 8388 }, { "epoch": 0.43, "learning_rate": 1.283748446682477e-05, "loss": 0.9765, "step": 8389 }, { "epoch": 0.43, "learning_rate": 1.283590664112782e-05, "loss": 0.8728, "step": 8390 }, { "epoch": 0.43, "learning_rate": 1.2834328738651716e-05, "loss": 0.8797, "step": 8391 }, { "epoch": 0.43, "learning_rate": 1.2832750759439175e-05, "loss": 1.0234, "step": 8392 }, { "epoch": 0.43, "learning_rate": 1.2831172703532922e-05, "loss": 0.9827, "step": 8393 }, { "epoch": 0.43, "learning_rate": 1.2829594570975676e-05, "loss": 0.8857, "step": 8394 }, { "epoch": 0.43, "learning_rate": 1.282801636181017e-05, "loss": 0.9909, "step": 8395 }, { "epoch": 0.43, "learning_rate": 1.2826438076079131e-05, "loss": 0.9341, "step": 8396 }, { "epoch": 0.43, "learning_rate": 1.2824859713825288e-05, "loss": 0.9541, "step": 8397 }, { "epoch": 0.43, "learning_rate": 1.2823281275091372e-05, "loss": 0.934, "step": 8398 }, { "epoch": 0.43, "learning_rate": 1.282170275992012e-05, "loss": 0.9469, "step": 8399 }, { "epoch": 0.43, "learning_rate": 1.2820124168354266e-05, "loss": 0.8419, "step": 8400 }, { "epoch": 0.43, "learning_rate": 1.2818545500436553e-05, "loss": 0.8685, "step": 8401 }, { "epoch": 0.43, "learning_rate": 1.2816966756209719e-05, "loss": 0.9589, "step": 8402 }, { "epoch": 0.43, "learning_rate": 1.2815387935716508e-05, "loss": 0.9396, "step": 8403 }, { "epoch": 0.43, "learning_rate": 1.2813809038999662e-05, "loss": 0.9647, "step": 8404 }, { "epoch": 0.43, "learning_rate": 1.2812230066101932e-05, "loss": 0.884, "step": 8405 }, { "epoch": 0.43, "learning_rate": 1.2810651017066062e-05, "loss": 0.8246, "step": 8406 }, { "epoch": 0.43, "learning_rate": 1.2809071891934807e-05, "loss": 1.0141, "step": 8407 }, { "epoch": 0.43, "learning_rate": 1.2807492690750922e-05, "loss": 0.8552, "step": 8408 }, { "epoch": 0.43, "learning_rate": 1.2805913413557158e-05, "loss": 0.9693, "step": 8409 }, { "epoch": 0.43, "learning_rate": 1.2804334060396272e-05, "loss": 0.912, "step": 8410 }, { "epoch": 0.43, "learning_rate": 1.2802754631311024e-05, "loss": 0.9647, "step": 8411 }, { "epoch": 0.43, "learning_rate": 1.2801175126344178e-05, "loss": 0.9128, "step": 8412 }, { "epoch": 0.43, "learning_rate": 1.2799595545538495e-05, "loss": 0.877, "step": 8413 }, { "epoch": 0.43, "learning_rate": 1.2798015888936744e-05, "loss": 0.9044, "step": 8414 }, { "epoch": 0.43, "learning_rate": 1.2796436156581686e-05, "loss": 0.8928, "step": 8415 }, { "epoch": 0.43, "learning_rate": 1.2794856348516095e-05, "loss": 1.0251, "step": 8416 }, { "epoch": 0.43, "learning_rate": 1.2793276464782744e-05, "loss": 0.958, "step": 8417 }, { "epoch": 0.43, "learning_rate": 1.2791696505424403e-05, "loss": 0.9985, "step": 8418 }, { "epoch": 0.43, "learning_rate": 1.279011647048385e-05, "loss": 0.9138, "step": 8419 }, { "epoch": 0.43, "learning_rate": 1.2788536360003861e-05, "loss": 0.9511, "step": 8420 }, { "epoch": 0.43, "learning_rate": 1.2786956174027216e-05, "loss": 0.9275, "step": 8421 }, { "epoch": 0.43, "learning_rate": 1.2785375912596699e-05, "loss": 1.0411, "step": 8422 }, { "epoch": 0.43, "learning_rate": 1.2783795575755092e-05, "loss": 0.9626, "step": 8423 }, { "epoch": 0.43, "learning_rate": 1.2782215163545181e-05, "loss": 0.925, "step": 8424 }, { "epoch": 0.43, "learning_rate": 1.2780634676009755e-05, "loss": 0.8697, "step": 8425 }, { "epoch": 0.43, "learning_rate": 1.2779054113191605e-05, "loss": 0.9125, "step": 8426 }, { "epoch": 0.43, "learning_rate": 1.2777473475133522e-05, "loss": 0.9663, "step": 8427 }, { "epoch": 0.43, "learning_rate": 1.2775892761878296e-05, "loss": 0.8667, "step": 8428 }, { "epoch": 0.43, "learning_rate": 1.2774311973468728e-05, "loss": 0.8506, "step": 8429 }, { "epoch": 0.43, "learning_rate": 1.2772731109947616e-05, "loss": 1.0714, "step": 8430 }, { "epoch": 0.43, "learning_rate": 1.2771150171357757e-05, "loss": 0.9124, "step": 8431 }, { "epoch": 0.43, "learning_rate": 1.276956915774196e-05, "loss": 0.9032, "step": 8432 }, { "epoch": 0.43, "learning_rate": 1.2767988069143024e-05, "loss": 0.8862, "step": 8433 }, { "epoch": 0.43, "learning_rate": 1.2766406905603753e-05, "loss": 1.0225, "step": 8434 }, { "epoch": 0.43, "learning_rate": 1.2764825667166958e-05, "loss": 0.9287, "step": 8435 }, { "epoch": 0.43, "learning_rate": 1.276324435387545e-05, "loss": 1.0244, "step": 8436 }, { "epoch": 0.43, "learning_rate": 1.2761662965772041e-05, "loss": 0.9478, "step": 8437 }, { "epoch": 0.43, "learning_rate": 1.2760081502899549e-05, "loss": 0.9774, "step": 8438 }, { "epoch": 0.43, "learning_rate": 1.2758499965300784e-05, "loss": 1.0063, "step": 8439 }, { "epoch": 0.43, "learning_rate": 1.2756918353018566e-05, "loss": 0.8972, "step": 8440 }, { "epoch": 0.43, "learning_rate": 1.2755336666095717e-05, "loss": 0.9081, "step": 8441 }, { "epoch": 0.43, "learning_rate": 1.2753754904575062e-05, "loss": 0.964, "step": 8442 }, { "epoch": 0.43, "learning_rate": 1.275217306849942e-05, "loss": 0.9499, "step": 8443 }, { "epoch": 0.43, "learning_rate": 1.2750591157911622e-05, "loss": 0.8823, "step": 8444 }, { "epoch": 0.43, "learning_rate": 1.2749009172854494e-05, "loss": 0.8737, "step": 8445 }, { "epoch": 0.43, "learning_rate": 1.2747427113370867e-05, "loss": 0.9112, "step": 8446 }, { "epoch": 0.43, "learning_rate": 1.2745844979503572e-05, "loss": 0.9363, "step": 8447 }, { "epoch": 0.43, "learning_rate": 1.2744262771295447e-05, "loss": 0.9869, "step": 8448 }, { "epoch": 0.43, "learning_rate": 1.274268048878933e-05, "loss": 0.962, "step": 8449 }, { "epoch": 0.43, "learning_rate": 1.2741098132028054e-05, "loss": 0.9234, "step": 8450 }, { "epoch": 0.43, "learning_rate": 1.2739515701054466e-05, "loss": 0.9014, "step": 8451 }, { "epoch": 0.43, "learning_rate": 1.27379331959114e-05, "loss": 0.8681, "step": 8452 }, { "epoch": 0.43, "learning_rate": 1.2736350616641706e-05, "loss": 0.924, "step": 8453 }, { "epoch": 0.43, "learning_rate": 1.273476796328823e-05, "loss": 0.9683, "step": 8454 }, { "epoch": 0.43, "learning_rate": 1.2733185235893825e-05, "loss": 0.9766, "step": 8455 }, { "epoch": 0.43, "learning_rate": 1.2731602434501338e-05, "loss": 0.8934, "step": 8456 }, { "epoch": 0.43, "learning_rate": 1.2730019559153617e-05, "loss": 0.8979, "step": 8457 }, { "epoch": 0.43, "learning_rate": 1.2728436609893524e-05, "loss": 0.9645, "step": 8458 }, { "epoch": 0.43, "learning_rate": 1.272685358676391e-05, "loss": 0.9072, "step": 8459 }, { "epoch": 0.43, "learning_rate": 1.2725270489807637e-05, "loss": 0.9462, "step": 8460 }, { "epoch": 0.43, "learning_rate": 1.2723687319067566e-05, "loss": 0.937, "step": 8461 }, { "epoch": 0.43, "learning_rate": 1.272210407458656e-05, "loss": 0.9558, "step": 8462 }, { "epoch": 0.43, "learning_rate": 1.272052075640748e-05, "loss": 0.9768, "step": 8463 }, { "epoch": 0.43, "learning_rate": 1.2718937364573196e-05, "loss": 1.0325, "step": 8464 }, { "epoch": 0.43, "learning_rate": 1.2717353899126574e-05, "loss": 1.01, "step": 8465 }, { "epoch": 0.43, "learning_rate": 1.2715770360110486e-05, "loss": 0.9769, "step": 8466 }, { "epoch": 0.43, "learning_rate": 1.2714186747567807e-05, "loss": 0.9016, "step": 8467 }, { "epoch": 0.43, "learning_rate": 1.2712603061541408e-05, "loss": 0.8235, "step": 8468 }, { "epoch": 0.43, "learning_rate": 1.2711019302074168e-05, "loss": 0.8626, "step": 8469 }, { "epoch": 0.43, "learning_rate": 1.2709435469208966e-05, "loss": 0.9826, "step": 8470 }, { "epoch": 0.43, "learning_rate": 1.2707851562988678e-05, "loss": 1.0258, "step": 8471 }, { "epoch": 0.43, "learning_rate": 1.270626758345619e-05, "loss": 0.8975, "step": 8472 }, { "epoch": 0.43, "learning_rate": 1.2704683530654389e-05, "loss": 0.8801, "step": 8473 }, { "epoch": 0.43, "learning_rate": 1.2703099404626158e-05, "loss": 0.8576, "step": 8474 }, { "epoch": 0.43, "learning_rate": 1.2701515205414385e-05, "loss": 1.0281, "step": 8475 }, { "epoch": 0.43, "learning_rate": 1.2699930933061966e-05, "loss": 0.8684, "step": 8476 }, { "epoch": 0.43, "learning_rate": 1.2698346587611785e-05, "loss": 0.9756, "step": 8477 }, { "epoch": 0.43, "learning_rate": 1.2696762169106743e-05, "loss": 0.9501, "step": 8478 }, { "epoch": 0.43, "learning_rate": 1.2695177677589734e-05, "loss": 0.9204, "step": 8479 }, { "epoch": 0.43, "learning_rate": 1.2693593113103658e-05, "loss": 0.9074, "step": 8480 }, { "epoch": 0.43, "learning_rate": 1.2692008475691413e-05, "loss": 0.8753, "step": 8481 }, { "epoch": 0.43, "learning_rate": 1.2690423765395903e-05, "loss": 1.0255, "step": 8482 }, { "epoch": 0.43, "learning_rate": 1.2688838982260033e-05, "loss": 0.9323, "step": 8483 }, { "epoch": 0.43, "learning_rate": 1.2687254126326703e-05, "loss": 0.8869, "step": 8484 }, { "epoch": 0.43, "learning_rate": 1.2685669197638833e-05, "loss": 0.9135, "step": 8485 }, { "epoch": 0.43, "learning_rate": 1.2684084196239325e-05, "loss": 0.9692, "step": 8486 }, { "epoch": 0.43, "learning_rate": 1.268249912217109e-05, "loss": 0.9834, "step": 8487 }, { "epoch": 0.43, "learning_rate": 1.268091397547705e-05, "loss": 0.8814, "step": 8488 }, { "epoch": 0.43, "learning_rate": 1.2679328756200113e-05, "loss": 0.9443, "step": 8489 }, { "epoch": 0.43, "learning_rate": 1.2677743464383201e-05, "loss": 0.8814, "step": 8490 }, { "epoch": 0.43, "learning_rate": 1.2676158100069234e-05, "loss": 0.9471, "step": 8491 }, { "epoch": 0.43, "learning_rate": 1.2674572663301134e-05, "loss": 0.9186, "step": 8492 }, { "epoch": 0.43, "learning_rate": 1.2672987154121826e-05, "loss": 0.925, "step": 8493 }, { "epoch": 0.43, "learning_rate": 1.2671401572574231e-05, "loss": 0.8695, "step": 8494 }, { "epoch": 0.43, "learning_rate": 1.2669815918701282e-05, "loss": 0.9352, "step": 8495 }, { "epoch": 0.43, "learning_rate": 1.2668230192545907e-05, "loss": 0.9746, "step": 8496 }, { "epoch": 0.43, "learning_rate": 1.2666644394151043e-05, "loss": 0.92, "step": 8497 }, { "epoch": 0.43, "learning_rate": 1.2665058523559615e-05, "loss": 0.9341, "step": 8498 }, { "epoch": 0.43, "learning_rate": 1.2663472580814563e-05, "loss": 0.8747, "step": 8499 }, { "epoch": 0.43, "learning_rate": 1.2661886565958823e-05, "loss": 0.8308, "step": 8500 }, { "epoch": 0.43, "learning_rate": 1.266030047903534e-05, "loss": 1.0343, "step": 8501 }, { "epoch": 0.43, "learning_rate": 1.2658714320087049e-05, "loss": 0.9615, "step": 8502 }, { "epoch": 0.43, "learning_rate": 1.2657128089156899e-05, "loss": 0.866, "step": 8503 }, { "epoch": 0.43, "learning_rate": 1.2655541786287832e-05, "loss": 0.8592, "step": 8504 }, { "epoch": 0.43, "learning_rate": 1.2653955411522793e-05, "loss": 0.9923, "step": 8505 }, { "epoch": 0.43, "learning_rate": 1.2652368964904738e-05, "loss": 0.9314, "step": 8506 }, { "epoch": 0.43, "learning_rate": 1.2650782446476618e-05, "loss": 0.9367, "step": 8507 }, { "epoch": 0.43, "learning_rate": 1.2649195856281377e-05, "loss": 0.9899, "step": 8508 }, { "epoch": 0.43, "learning_rate": 1.2647609194361981e-05, "loss": 0.8702, "step": 8509 }, { "epoch": 0.43, "learning_rate": 1.2646022460761382e-05, "loss": 1.0143, "step": 8510 }, { "epoch": 0.43, "learning_rate": 1.264443565552254e-05, "loss": 0.9474, "step": 8511 }, { "epoch": 0.43, "learning_rate": 1.2642848778688415e-05, "loss": 0.9851, "step": 8512 }, { "epoch": 0.43, "learning_rate": 1.2641261830301973e-05, "loss": 1.0064, "step": 8513 }, { "epoch": 0.43, "learning_rate": 1.2639674810406175e-05, "loss": 0.9285, "step": 8514 }, { "epoch": 0.43, "learning_rate": 1.2638087719043992e-05, "loss": 0.9941, "step": 8515 }, { "epoch": 0.43, "learning_rate": 1.2636500556258392e-05, "loss": 0.8863, "step": 8516 }, { "epoch": 0.43, "learning_rate": 1.263491332209234e-05, "loss": 0.9526, "step": 8517 }, { "epoch": 0.43, "learning_rate": 1.2633326016588816e-05, "loss": 0.928, "step": 8518 }, { "epoch": 0.43, "learning_rate": 1.2631738639790792e-05, "loss": 0.915, "step": 8519 }, { "epoch": 0.43, "learning_rate": 1.2630151191741243e-05, "loss": 1.053, "step": 8520 }, { "epoch": 0.43, "learning_rate": 1.2628563672483147e-05, "loss": 0.8484, "step": 8521 }, { "epoch": 0.43, "learning_rate": 1.262697608205949e-05, "loss": 0.9314, "step": 8522 }, { "epoch": 0.43, "learning_rate": 1.2625388420513249e-05, "loss": 0.8936, "step": 8523 }, { "epoch": 0.43, "learning_rate": 1.2623800687887409e-05, "loss": 1.0372, "step": 8524 }, { "epoch": 0.43, "learning_rate": 1.2622212884224956e-05, "loss": 0.901, "step": 8525 }, { "epoch": 0.43, "learning_rate": 1.2620625009568881e-05, "loss": 1.0001, "step": 8526 }, { "epoch": 0.43, "learning_rate": 1.2619037063962172e-05, "loss": 0.923, "step": 8527 }, { "epoch": 0.43, "learning_rate": 1.2617449047447819e-05, "loss": 0.8669, "step": 8528 }, { "epoch": 0.43, "learning_rate": 1.261586096006882e-05, "loss": 0.9667, "step": 8529 }, { "epoch": 0.43, "learning_rate": 1.2614272801868165e-05, "loss": 0.933, "step": 8530 }, { "epoch": 0.43, "learning_rate": 1.2612684572888858e-05, "loss": 0.9895, "step": 8531 }, { "epoch": 0.43, "learning_rate": 1.2611096273173896e-05, "loss": 1.0873, "step": 8532 }, { "epoch": 0.43, "learning_rate": 1.2609507902766283e-05, "loss": 0.9671, "step": 8533 }, { "epoch": 0.43, "learning_rate": 1.2607919461709015e-05, "loss": 0.9567, "step": 8534 }, { "epoch": 0.43, "learning_rate": 1.2606330950045105e-05, "loss": 0.9914, "step": 8535 }, { "epoch": 0.43, "learning_rate": 1.2604742367817555e-05, "loss": 0.8885, "step": 8536 }, { "epoch": 0.43, "learning_rate": 1.2603153715069376e-05, "loss": 1.0322, "step": 8537 }, { "epoch": 0.43, "learning_rate": 1.2601564991843585e-05, "loss": 0.9207, "step": 8538 }, { "epoch": 0.43, "learning_rate": 1.2599976198183188e-05, "loss": 0.9346, "step": 8539 }, { "epoch": 0.43, "learning_rate": 1.25983873341312e-05, "loss": 0.9622, "step": 8540 }, { "epoch": 0.43, "learning_rate": 1.2596798399730639e-05, "loss": 0.9311, "step": 8541 }, { "epoch": 0.43, "learning_rate": 1.2595209395024524e-05, "loss": 0.9187, "step": 8542 }, { "epoch": 0.43, "learning_rate": 1.2593620320055875e-05, "loss": 0.8661, "step": 8543 }, { "epoch": 0.43, "learning_rate": 1.2592031174867718e-05, "loss": 0.9321, "step": 8544 }, { "epoch": 0.43, "learning_rate": 1.2590441959503074e-05, "loss": 0.9549, "step": 8545 }, { "epoch": 0.43, "learning_rate": 1.2588852674004968e-05, "loss": 0.8418, "step": 8546 }, { "epoch": 0.43, "learning_rate": 1.2587263318416432e-05, "loss": 0.8981, "step": 8547 }, { "epoch": 0.43, "learning_rate": 1.2585673892780492e-05, "loss": 1.0198, "step": 8548 }, { "epoch": 0.43, "learning_rate": 1.2584084397140184e-05, "loss": 0.8715, "step": 8549 }, { "epoch": 0.43, "learning_rate": 1.2582494831538542e-05, "loss": 0.9716, "step": 8550 }, { "epoch": 0.43, "learning_rate": 1.2580905196018598e-05, "loss": 0.9758, "step": 8551 }, { "epoch": 0.43, "learning_rate": 1.257931549062339e-05, "loss": 0.9174, "step": 8552 }, { "epoch": 0.43, "learning_rate": 1.2577725715395963e-05, "loss": 1.0001, "step": 8553 }, { "epoch": 0.43, "learning_rate": 1.2576135870379351e-05, "loss": 0.9374, "step": 8554 }, { "epoch": 0.43, "learning_rate": 1.2574545955616602e-05, "loss": 0.9339, "step": 8555 }, { "epoch": 0.43, "learning_rate": 1.2572955971150762e-05, "loss": 0.8924, "step": 8556 }, { "epoch": 0.43, "learning_rate": 1.2571365917024878e-05, "loss": 0.8749, "step": 8557 }, { "epoch": 0.43, "learning_rate": 1.2569775793281993e-05, "loss": 0.9244, "step": 8558 }, { "epoch": 0.43, "learning_rate": 1.2568185599965163e-05, "loss": 0.9297, "step": 8559 }, { "epoch": 0.43, "learning_rate": 1.2566595337117445e-05, "loss": 0.9679, "step": 8560 }, { "epoch": 0.43, "learning_rate": 1.2565005004781883e-05, "loss": 0.8747, "step": 8561 }, { "epoch": 0.43, "learning_rate": 1.2563414603001545e-05, "loss": 0.8446, "step": 8562 }, { "epoch": 0.44, "learning_rate": 1.2561824131819482e-05, "loss": 1.067, "step": 8563 }, { "epoch": 0.44, "learning_rate": 1.2560233591278754e-05, "loss": 0.9149, "step": 8564 }, { "epoch": 0.44, "learning_rate": 1.2558642981422424e-05, "loss": 0.9814, "step": 8565 }, { "epoch": 0.44, "learning_rate": 1.2557052302293563e-05, "loss": 0.831, "step": 8566 }, { "epoch": 0.44, "learning_rate": 1.2555461553935226e-05, "loss": 1.0215, "step": 8567 }, { "epoch": 0.44, "learning_rate": 1.255387073639049e-05, "loss": 0.889, "step": 8568 }, { "epoch": 0.44, "learning_rate": 1.2552279849702422e-05, "loss": 0.9727, "step": 8569 }, { "epoch": 0.44, "learning_rate": 1.2550688893914089e-05, "loss": 0.921, "step": 8570 }, { "epoch": 0.44, "learning_rate": 1.2549097869068567e-05, "loss": 0.9237, "step": 8571 }, { "epoch": 0.44, "learning_rate": 1.2547506775208937e-05, "loss": 0.9019, "step": 8572 }, { "epoch": 0.44, "learning_rate": 1.2545915612378265e-05, "loss": 0.8953, "step": 8573 }, { "epoch": 0.44, "learning_rate": 1.2544324380619644e-05, "loss": 0.8739, "step": 8574 }, { "epoch": 0.44, "learning_rate": 1.2542733079976144e-05, "loss": 0.9821, "step": 8575 }, { "epoch": 0.44, "learning_rate": 1.254114171049085e-05, "loss": 0.9807, "step": 8576 }, { "epoch": 0.44, "learning_rate": 1.2539550272206846e-05, "loss": 1.0082, "step": 8577 }, { "epoch": 0.44, "learning_rate": 1.2537958765167224e-05, "loss": 0.8859, "step": 8578 }, { "epoch": 0.44, "learning_rate": 1.2536367189415067e-05, "loss": 0.9094, "step": 8579 }, { "epoch": 0.44, "learning_rate": 1.2534775544993467e-05, "loss": 0.8981, "step": 8580 }, { "epoch": 0.44, "learning_rate": 1.2533183831945517e-05, "loss": 0.9202, "step": 8581 }, { "epoch": 0.44, "learning_rate": 1.2531592050314308e-05, "loss": 0.9584, "step": 8582 }, { "epoch": 0.44, "learning_rate": 1.2530000200142937e-05, "loss": 1.0544, "step": 8583 }, { "epoch": 0.44, "learning_rate": 1.2528408281474507e-05, "loss": 0.7915, "step": 8584 }, { "epoch": 0.44, "learning_rate": 1.2526816294352107e-05, "loss": 0.8546, "step": 8585 }, { "epoch": 0.44, "learning_rate": 1.2525224238818848e-05, "loss": 0.8823, "step": 8586 }, { "epoch": 0.44, "learning_rate": 1.252363211491783e-05, "loss": 0.9387, "step": 8587 }, { "epoch": 0.44, "learning_rate": 1.2522039922692152e-05, "loss": 0.9771, "step": 8588 }, { "epoch": 0.44, "learning_rate": 1.2520447662184928e-05, "loss": 0.9052, "step": 8589 }, { "epoch": 0.44, "learning_rate": 1.251885533343927e-05, "loss": 0.9252, "step": 8590 }, { "epoch": 0.44, "learning_rate": 1.251726293649828e-05, "loss": 0.9264, "step": 8591 }, { "epoch": 0.44, "learning_rate": 1.2515670471405074e-05, "loss": 0.953, "step": 8592 }, { "epoch": 0.44, "learning_rate": 1.2514077938202767e-05, "loss": 0.9252, "step": 8593 }, { "epoch": 0.44, "learning_rate": 1.2512485336934477e-05, "loss": 0.8757, "step": 8594 }, { "epoch": 0.44, "learning_rate": 1.2510892667643314e-05, "loss": 0.9542, "step": 8595 }, { "epoch": 0.44, "learning_rate": 1.2509299930372407e-05, "loss": 0.8885, "step": 8596 }, { "epoch": 0.44, "learning_rate": 1.2507707125164874e-05, "loss": 0.8213, "step": 8597 }, { "epoch": 0.44, "learning_rate": 1.2506114252063841e-05, "loss": 0.9533, "step": 8598 }, { "epoch": 0.44, "learning_rate": 1.2504521311112428e-05, "loss": 1.0419, "step": 8599 }, { "epoch": 0.44, "learning_rate": 1.2502928302353766e-05, "loss": 1.0562, "step": 8600 }, { "epoch": 0.44, "learning_rate": 1.2501335225830982e-05, "loss": 0.8743, "step": 8601 }, { "epoch": 0.44, "learning_rate": 1.2499742081587208e-05, "loss": 0.8907, "step": 8602 }, { "epoch": 0.44, "learning_rate": 1.2498148869665576e-05, "loss": 0.8331, "step": 8603 }, { "epoch": 0.44, "learning_rate": 1.2496555590109225e-05, "loss": 0.9646, "step": 8604 }, { "epoch": 0.44, "learning_rate": 1.2494962242961283e-05, "loss": 0.9442, "step": 8605 }, { "epoch": 0.44, "learning_rate": 1.2493368828264897e-05, "loss": 0.8594, "step": 8606 }, { "epoch": 0.44, "learning_rate": 1.2491775346063198e-05, "loss": 0.9462, "step": 8607 }, { "epoch": 0.44, "learning_rate": 1.2490181796399334e-05, "loss": 0.8969, "step": 8608 }, { "epoch": 0.44, "learning_rate": 1.248858817931645e-05, "loss": 0.9377, "step": 8609 }, { "epoch": 0.44, "learning_rate": 1.2486994494857688e-05, "loss": 0.86, "step": 8610 }, { "epoch": 0.44, "learning_rate": 1.2485400743066191e-05, "loss": 0.922, "step": 8611 }, { "epoch": 0.44, "learning_rate": 1.248380692398512e-05, "loss": 0.9549, "step": 8612 }, { "epoch": 0.44, "learning_rate": 1.2482213037657614e-05, "loss": 0.8268, "step": 8613 }, { "epoch": 0.44, "learning_rate": 1.2480619084126832e-05, "loss": 0.9381, "step": 8614 }, { "epoch": 0.44, "learning_rate": 1.247902506343593e-05, "loss": 1.0071, "step": 8615 }, { "epoch": 0.44, "learning_rate": 1.247743097562806e-05, "loss": 0.8549, "step": 8616 }, { "epoch": 0.44, "learning_rate": 1.2475836820746383e-05, "loss": 0.9632, "step": 8617 }, { "epoch": 0.44, "learning_rate": 1.2474242598834057e-05, "loss": 0.9293, "step": 8618 }, { "epoch": 0.44, "learning_rate": 1.2472648309934246e-05, "loss": 0.9204, "step": 8619 }, { "epoch": 0.44, "learning_rate": 1.247105395409011e-05, "loss": 0.9356, "step": 8620 }, { "epoch": 0.44, "learning_rate": 1.246945953134482e-05, "loss": 0.9326, "step": 8621 }, { "epoch": 0.44, "learning_rate": 1.2467865041741542e-05, "loss": 0.8494, "step": 8622 }, { "epoch": 0.44, "learning_rate": 1.246627048532344e-05, "loss": 0.9834, "step": 8623 }, { "epoch": 0.44, "learning_rate": 1.246467586213369e-05, "loss": 0.9769, "step": 8624 }, { "epoch": 0.44, "learning_rate": 1.2463081172215463e-05, "loss": 0.8967, "step": 8625 }, { "epoch": 0.44, "learning_rate": 1.2461486415611934e-05, "loss": 0.8655, "step": 8626 }, { "epoch": 0.44, "learning_rate": 1.245989159236628e-05, "loss": 0.9509, "step": 8627 }, { "epoch": 0.44, "learning_rate": 1.2458296702521677e-05, "loss": 0.9633, "step": 8628 }, { "epoch": 0.44, "learning_rate": 1.2456701746121308e-05, "loss": 0.9283, "step": 8629 }, { "epoch": 0.44, "learning_rate": 1.2455106723208348e-05, "loss": 0.9648, "step": 8630 }, { "epoch": 0.44, "learning_rate": 1.245351163382599e-05, "loss": 0.9096, "step": 8631 }, { "epoch": 0.44, "learning_rate": 1.2451916478017413e-05, "loss": 0.8504, "step": 8632 }, { "epoch": 0.44, "learning_rate": 1.2450321255825808e-05, "loss": 0.9278, "step": 8633 }, { "epoch": 0.44, "learning_rate": 1.2448725967294362e-05, "loss": 0.9132, "step": 8634 }, { "epoch": 0.44, "learning_rate": 1.2447130612466265e-05, "loss": 0.9212, "step": 8635 }, { "epoch": 0.44, "learning_rate": 1.2445535191384708e-05, "loss": 0.8816, "step": 8636 }, { "epoch": 0.44, "learning_rate": 1.244393970409289e-05, "loss": 0.9766, "step": 8637 }, { "epoch": 0.44, "learning_rate": 1.2442344150634003e-05, "loss": 0.8974, "step": 8638 }, { "epoch": 0.44, "learning_rate": 1.2440748531051247e-05, "loss": 0.8431, "step": 8639 }, { "epoch": 0.44, "learning_rate": 1.2439152845387823e-05, "loss": 0.9142, "step": 8640 }, { "epoch": 0.44, "learning_rate": 1.243755709368693e-05, "loss": 0.9683, "step": 8641 }, { "epoch": 0.44, "learning_rate": 1.243596127599177e-05, "loss": 0.9924, "step": 8642 }, { "epoch": 0.44, "learning_rate": 1.2434365392345553e-05, "loss": 0.8324, "step": 8643 }, { "epoch": 0.44, "learning_rate": 1.2432769442791478e-05, "loss": 0.9393, "step": 8644 }, { "epoch": 0.44, "learning_rate": 1.2431173427372764e-05, "loss": 0.937, "step": 8645 }, { "epoch": 0.44, "learning_rate": 1.2429577346132617e-05, "loss": 0.9878, "step": 8646 }, { "epoch": 0.44, "learning_rate": 1.2427981199114246e-05, "loss": 0.9907, "step": 8647 }, { "epoch": 0.44, "learning_rate": 1.2426384986360865e-05, "loss": 0.9159, "step": 8648 }, { "epoch": 0.44, "learning_rate": 1.2424788707915694e-05, "loss": 0.9816, "step": 8649 }, { "epoch": 0.44, "learning_rate": 1.2423192363821945e-05, "loss": 0.9183, "step": 8650 }, { "epoch": 0.44, "learning_rate": 1.2421595954122846e-05, "loss": 1.0097, "step": 8651 }, { "epoch": 0.44, "learning_rate": 1.2419999478861613e-05, "loss": 0.9646, "step": 8652 }, { "epoch": 0.44, "learning_rate": 1.2418402938081464e-05, "loss": 0.9394, "step": 8653 }, { "epoch": 0.44, "learning_rate": 1.241680633182563e-05, "loss": 0.9198, "step": 8654 }, { "epoch": 0.44, "learning_rate": 1.2415209660137339e-05, "loss": 0.9297, "step": 8655 }, { "epoch": 0.44, "learning_rate": 1.2413612923059813e-05, "loss": 1.0079, "step": 8656 }, { "epoch": 0.44, "learning_rate": 1.2412016120636285e-05, "loss": 0.8922, "step": 8657 }, { "epoch": 0.44, "learning_rate": 1.241041925290999e-05, "loss": 0.9236, "step": 8658 }, { "epoch": 0.44, "learning_rate": 1.2408822319924157e-05, "loss": 0.9333, "step": 8659 }, { "epoch": 0.44, "learning_rate": 1.2407225321722018e-05, "loss": 0.9446, "step": 8660 }, { "epoch": 0.44, "learning_rate": 1.240562825834682e-05, "loss": 0.86, "step": 8661 }, { "epoch": 0.44, "learning_rate": 1.2404031129841792e-05, "loss": 0.95, "step": 8662 }, { "epoch": 0.44, "learning_rate": 1.2402433936250183e-05, "loss": 0.9664, "step": 8663 }, { "epoch": 0.44, "learning_rate": 1.2400836677615226e-05, "loss": 0.9414, "step": 8664 }, { "epoch": 0.44, "learning_rate": 1.2399239353980178e-05, "loss": 0.9015, "step": 8665 }, { "epoch": 0.44, "learning_rate": 1.239764196538827e-05, "loss": 0.9551, "step": 8666 }, { "epoch": 0.44, "learning_rate": 1.2396044511882757e-05, "loss": 0.9735, "step": 8667 }, { "epoch": 0.44, "learning_rate": 1.2394446993506892e-05, "loss": 1.0145, "step": 8668 }, { "epoch": 0.44, "learning_rate": 1.2392849410303922e-05, "loss": 0.846, "step": 8669 }, { "epoch": 0.44, "learning_rate": 1.2391251762317097e-05, "loss": 1.0215, "step": 8670 }, { "epoch": 0.44, "learning_rate": 1.2389654049589676e-05, "loss": 0.9077, "step": 8671 }, { "epoch": 0.44, "learning_rate": 1.2388056272164914e-05, "loss": 0.9729, "step": 8672 }, { "epoch": 0.44, "learning_rate": 1.2386458430086069e-05, "loss": 0.9135, "step": 8673 }, { "epoch": 0.44, "learning_rate": 1.2384860523396401e-05, "loss": 0.8983, "step": 8674 }, { "epoch": 0.44, "learning_rate": 1.2383262552139172e-05, "loss": 0.8967, "step": 8675 }, { "epoch": 0.44, "learning_rate": 1.2381664516357643e-05, "loss": 0.9987, "step": 8676 }, { "epoch": 0.44, "learning_rate": 1.2380066416095085e-05, "loss": 0.8692, "step": 8677 }, { "epoch": 0.44, "learning_rate": 1.2378468251394757e-05, "loss": 0.9979, "step": 8678 }, { "epoch": 0.44, "learning_rate": 1.2376870022299934e-05, "loss": 0.8751, "step": 8679 }, { "epoch": 0.44, "learning_rate": 1.2375271728853884e-05, "loss": 0.9893, "step": 8680 }, { "epoch": 0.44, "learning_rate": 1.2373673371099879e-05, "loss": 0.9818, "step": 8681 }, { "epoch": 0.44, "learning_rate": 1.2372074949081192e-05, "loss": 0.9178, "step": 8682 }, { "epoch": 0.44, "learning_rate": 1.2370476462841099e-05, "loss": 0.8909, "step": 8683 }, { "epoch": 0.44, "learning_rate": 1.236887791242288e-05, "loss": 0.9825, "step": 8684 }, { "epoch": 0.44, "learning_rate": 1.2367279297869807e-05, "loss": 0.9028, "step": 8685 }, { "epoch": 0.44, "learning_rate": 1.236568061922517e-05, "loss": 0.9623, "step": 8686 }, { "epoch": 0.44, "learning_rate": 1.2364081876532246e-05, "loss": 0.9118, "step": 8687 }, { "epoch": 0.44, "learning_rate": 1.2362483069834318e-05, "loss": 0.9291, "step": 8688 }, { "epoch": 0.44, "learning_rate": 1.2360884199174682e-05, "loss": 1.0031, "step": 8689 }, { "epoch": 0.44, "learning_rate": 1.235928526459661e-05, "loss": 0.9665, "step": 8690 }, { "epoch": 0.44, "learning_rate": 1.23576862661434e-05, "loss": 0.9211, "step": 8691 }, { "epoch": 0.44, "learning_rate": 1.2356087203858346e-05, "loss": 1.0117, "step": 8692 }, { "epoch": 0.44, "learning_rate": 1.2354488077784739e-05, "loss": 0.949, "step": 8693 }, { "epoch": 0.44, "learning_rate": 1.2352888887965867e-05, "loss": 0.9535, "step": 8694 }, { "epoch": 0.44, "learning_rate": 1.2351289634445038e-05, "loss": 1.0419, "step": 8695 }, { "epoch": 0.44, "learning_rate": 1.2349690317265539e-05, "loss": 1.0295, "step": 8696 }, { "epoch": 0.44, "learning_rate": 1.2348090936470674e-05, "loss": 0.9666, "step": 8697 }, { "epoch": 0.44, "learning_rate": 1.234649149210375e-05, "loss": 0.9459, "step": 8698 }, { "epoch": 0.44, "learning_rate": 1.234489198420806e-05, "loss": 0.9644, "step": 8699 }, { "epoch": 0.44, "learning_rate": 1.2343292412826917e-05, "loss": 0.9134, "step": 8700 }, { "epoch": 0.44, "learning_rate": 1.2341692778003625e-05, "loss": 0.9403, "step": 8701 }, { "epoch": 0.44, "learning_rate": 1.2340093079781493e-05, "loss": 0.8882, "step": 8702 }, { "epoch": 0.44, "learning_rate": 1.2338493318203825e-05, "loss": 0.9617, "step": 8703 }, { "epoch": 0.44, "learning_rate": 1.2336893493313946e-05, "loss": 0.9498, "step": 8704 }, { "epoch": 0.44, "learning_rate": 1.2335293605155159e-05, "loss": 0.9633, "step": 8705 }, { "epoch": 0.44, "learning_rate": 1.2333693653770782e-05, "loss": 1.0443, "step": 8706 }, { "epoch": 0.44, "learning_rate": 1.233209363920413e-05, "loss": 0.9825, "step": 8707 }, { "epoch": 0.44, "learning_rate": 1.2330493561498526e-05, "loss": 1.0596, "step": 8708 }, { "epoch": 0.44, "learning_rate": 1.2328893420697288e-05, "loss": 0.8863, "step": 8709 }, { "epoch": 0.44, "learning_rate": 1.232729321684374e-05, "loss": 1.0294, "step": 8710 }, { "epoch": 0.44, "learning_rate": 1.2325692949981206e-05, "loss": 0.9249, "step": 8711 }, { "epoch": 0.44, "learning_rate": 1.2324092620153003e-05, "loss": 0.9547, "step": 8712 }, { "epoch": 0.44, "learning_rate": 1.2322492227402467e-05, "loss": 0.929, "step": 8713 }, { "epoch": 0.44, "learning_rate": 1.232089177177293e-05, "loss": 0.9264, "step": 8714 }, { "epoch": 0.44, "learning_rate": 1.2319291253307713e-05, "loss": 0.9864, "step": 8715 }, { "epoch": 0.44, "learning_rate": 1.2317690672050154e-05, "loss": 0.8557, "step": 8716 }, { "epoch": 0.44, "learning_rate": 1.2316090028043585e-05, "loss": 0.909, "step": 8717 }, { "epoch": 0.44, "learning_rate": 1.2314489321331342e-05, "loss": 0.9176, "step": 8718 }, { "epoch": 0.44, "learning_rate": 1.2312888551956763e-05, "loss": 0.9717, "step": 8719 }, { "epoch": 0.44, "learning_rate": 1.231128771996319e-05, "loss": 0.9578, "step": 8720 }, { "epoch": 0.44, "learning_rate": 1.2309686825393958e-05, "loss": 0.8003, "step": 8721 }, { "epoch": 0.44, "learning_rate": 1.2308085868292416e-05, "loss": 0.9244, "step": 8722 }, { "epoch": 0.44, "learning_rate": 1.2306484848701903e-05, "loss": 0.8595, "step": 8723 }, { "epoch": 0.44, "learning_rate": 1.2304883766665767e-05, "loss": 0.8957, "step": 8724 }, { "epoch": 0.44, "learning_rate": 1.2303282622227352e-05, "loss": 0.97, "step": 8725 }, { "epoch": 0.44, "learning_rate": 1.2301681415430015e-05, "loss": 0.9318, "step": 8726 }, { "epoch": 0.44, "learning_rate": 1.23000801463171e-05, "loss": 0.991, "step": 8727 }, { "epoch": 0.44, "learning_rate": 1.2298478814931963e-05, "loss": 0.9874, "step": 8728 }, { "epoch": 0.44, "learning_rate": 1.2296877421317958e-05, "loss": 0.9153, "step": 8729 }, { "epoch": 0.44, "learning_rate": 1.2295275965518439e-05, "loss": 1.0248, "step": 8730 }, { "epoch": 0.44, "learning_rate": 1.2293674447576764e-05, "loss": 0.9454, "step": 8731 }, { "epoch": 0.44, "learning_rate": 1.2292072867536295e-05, "loss": 0.9441, "step": 8732 }, { "epoch": 0.44, "learning_rate": 1.2290471225440393e-05, "loss": 0.9211, "step": 8733 }, { "epoch": 0.44, "learning_rate": 1.228886952133242e-05, "loss": 0.8735, "step": 8734 }, { "epoch": 0.44, "learning_rate": 1.2287267755255736e-05, "loss": 0.9884, "step": 8735 }, { "epoch": 0.44, "learning_rate": 1.2285665927253716e-05, "loss": 0.9318, "step": 8736 }, { "epoch": 0.44, "learning_rate": 1.2284064037369718e-05, "loss": 0.9516, "step": 8737 }, { "epoch": 0.44, "learning_rate": 1.2282462085647117e-05, "loss": 0.9584, "step": 8738 }, { "epoch": 0.44, "learning_rate": 1.2280860072129287e-05, "loss": 0.9484, "step": 8739 }, { "epoch": 0.44, "learning_rate": 1.2279257996859598e-05, "loss": 0.9018, "step": 8740 }, { "epoch": 0.44, "learning_rate": 1.2277655859881419e-05, "loss": 0.9023, "step": 8741 }, { "epoch": 0.44, "learning_rate": 1.2276053661238133e-05, "loss": 0.8117, "step": 8742 }, { "epoch": 0.44, "learning_rate": 1.2274451400973114e-05, "loss": 0.8997, "step": 8743 }, { "epoch": 0.44, "learning_rate": 1.2272849079129743e-05, "loss": 0.9559, "step": 8744 }, { "epoch": 0.44, "learning_rate": 1.2271246695751404e-05, "loss": 0.932, "step": 8745 }, { "epoch": 0.44, "learning_rate": 1.2269644250881476e-05, "loss": 1.0052, "step": 8746 }, { "epoch": 0.44, "learning_rate": 1.2268041744563344e-05, "loss": 1.0233, "step": 8747 }, { "epoch": 0.44, "learning_rate": 1.2266439176840396e-05, "loss": 0.8957, "step": 8748 }, { "epoch": 0.44, "learning_rate": 1.2264836547756019e-05, "loss": 0.9137, "step": 8749 }, { "epoch": 0.44, "learning_rate": 1.2263233857353598e-05, "loss": 0.95, "step": 8750 }, { "epoch": 0.44, "learning_rate": 1.2261631105676535e-05, "loss": 0.8893, "step": 8751 }, { "epoch": 0.44, "learning_rate": 1.226002829276821e-05, "loss": 0.9561, "step": 8752 }, { "epoch": 0.44, "learning_rate": 1.2258425418672024e-05, "loss": 0.9388, "step": 8753 }, { "epoch": 0.44, "learning_rate": 1.2256822483431376e-05, "loss": 0.8943, "step": 8754 }, { "epoch": 0.44, "learning_rate": 1.2255219487089659e-05, "loss": 0.9375, "step": 8755 }, { "epoch": 0.44, "learning_rate": 1.225361642969027e-05, "loss": 0.8849, "step": 8756 }, { "epoch": 0.44, "learning_rate": 1.225201331127662e-05, "loss": 0.9279, "step": 8757 }, { "epoch": 0.44, "learning_rate": 1.2250410131892101e-05, "loss": 1.0588, "step": 8758 }, { "epoch": 0.44, "learning_rate": 1.2248806891580123e-05, "loss": 0.822, "step": 8759 }, { "epoch": 0.45, "learning_rate": 1.2247203590384094e-05, "loss": 0.9487, "step": 8760 }, { "epoch": 0.45, "learning_rate": 1.2245600228347414e-05, "loss": 0.9071, "step": 8761 }, { "epoch": 0.45, "learning_rate": 1.2243996805513497e-05, "loss": 0.9288, "step": 8762 }, { "epoch": 0.45, "learning_rate": 1.2242393321925758e-05, "loss": 0.9499, "step": 8763 }, { "epoch": 0.45, "learning_rate": 1.2240789777627604e-05, "loss": 0.9529, "step": 8764 }, { "epoch": 0.45, "learning_rate": 1.223918617266245e-05, "loss": 0.9155, "step": 8765 }, { "epoch": 0.45, "learning_rate": 1.2237582507073713e-05, "loss": 0.9271, "step": 8766 }, { "epoch": 0.45, "learning_rate": 1.223597878090481e-05, "loss": 0.8774, "step": 8767 }, { "epoch": 0.45, "learning_rate": 1.2234374994199158e-05, "loss": 0.9514, "step": 8768 }, { "epoch": 0.45, "learning_rate": 1.2232771147000184e-05, "loss": 0.8946, "step": 8769 }, { "epoch": 0.45, "learning_rate": 1.2231167239351306e-05, "loss": 0.899, "step": 8770 }, { "epoch": 0.45, "learning_rate": 1.2229563271295948e-05, "loss": 0.9048, "step": 8771 }, { "epoch": 0.45, "learning_rate": 1.2227959242877536e-05, "loss": 0.8952, "step": 8772 }, { "epoch": 0.45, "learning_rate": 1.2226355154139497e-05, "loss": 0.966, "step": 8773 }, { "epoch": 0.45, "learning_rate": 1.2224751005125264e-05, "loss": 0.8237, "step": 8774 }, { "epoch": 0.45, "learning_rate": 1.2223146795878262e-05, "loss": 0.8706, "step": 8775 }, { "epoch": 0.45, "learning_rate": 1.222154252644193e-05, "loss": 0.8777, "step": 8776 }, { "epoch": 0.45, "learning_rate": 1.2219938196859692e-05, "loss": 0.9386, "step": 8777 }, { "epoch": 0.45, "learning_rate": 1.2218333807174992e-05, "loss": 0.9445, "step": 8778 }, { "epoch": 0.45, "learning_rate": 1.2216729357431263e-05, "loss": 0.9772, "step": 8779 }, { "epoch": 0.45, "learning_rate": 1.2215124847671947e-05, "loss": 0.8694, "step": 8780 }, { "epoch": 0.45, "learning_rate": 1.221352027794048e-05, "loss": 0.8721, "step": 8781 }, { "epoch": 0.45, "learning_rate": 1.2211915648280309e-05, "loss": 0.8122, "step": 8782 }, { "epoch": 0.45, "learning_rate": 1.2210310958734873e-05, "loss": 0.9069, "step": 8783 }, { "epoch": 0.45, "learning_rate": 1.2208706209347619e-05, "loss": 0.9345, "step": 8784 }, { "epoch": 0.45, "learning_rate": 1.2207101400161996e-05, "loss": 0.9385, "step": 8785 }, { "epoch": 0.45, "learning_rate": 1.220549653122145e-05, "loss": 0.9614, "step": 8786 }, { "epoch": 0.45, "learning_rate": 1.2203891602569434e-05, "loss": 0.9263, "step": 8787 }, { "epoch": 0.45, "learning_rate": 1.2202286614249394e-05, "loss": 0.965, "step": 8788 }, { "epoch": 0.45, "learning_rate": 1.2200681566304788e-05, "loss": 0.9104, "step": 8789 }, { "epoch": 0.45, "learning_rate": 1.219907645877907e-05, "loss": 0.9538, "step": 8790 }, { "epoch": 0.45, "learning_rate": 1.21974712917157e-05, "loss": 0.8491, "step": 8791 }, { "epoch": 0.45, "learning_rate": 1.2195866065158128e-05, "loss": 0.9569, "step": 8792 }, { "epoch": 0.45, "learning_rate": 1.219426077914982e-05, "loss": 0.9088, "step": 8793 }, { "epoch": 0.45, "learning_rate": 1.2192655433734237e-05, "loss": 1.0129, "step": 8794 }, { "epoch": 0.45, "learning_rate": 1.2191050028954838e-05, "loss": 0.957, "step": 8795 }, { "epoch": 0.45, "learning_rate": 1.2189444564855092e-05, "loss": 0.9909, "step": 8796 }, { "epoch": 0.45, "learning_rate": 1.2187839041478467e-05, "loss": 0.9357, "step": 8797 }, { "epoch": 0.45, "learning_rate": 1.2186233458868426e-05, "loss": 0.959, "step": 8798 }, { "epoch": 0.45, "learning_rate": 1.2184627817068437e-05, "loss": 0.9243, "step": 8799 }, { "epoch": 0.45, "learning_rate": 1.2183022116121978e-05, "loss": 0.9149, "step": 8800 }, { "epoch": 0.45, "learning_rate": 1.2181416356072516e-05, "loss": 0.9766, "step": 8801 }, { "epoch": 0.45, "learning_rate": 1.2179810536963525e-05, "loss": 0.8987, "step": 8802 }, { "epoch": 0.45, "learning_rate": 1.2178204658838483e-05, "loss": 0.9717, "step": 8803 }, { "epoch": 0.45, "learning_rate": 1.2176598721740873e-05, "loss": 0.9252, "step": 8804 }, { "epoch": 0.45, "learning_rate": 1.2174992725714162e-05, "loss": 0.867, "step": 8805 }, { "epoch": 0.45, "learning_rate": 1.2173386670801842e-05, "loss": 0.8967, "step": 8806 }, { "epoch": 0.45, "learning_rate": 1.2171780557047389e-05, "loss": 0.881, "step": 8807 }, { "epoch": 0.45, "learning_rate": 1.2170174384494284e-05, "loss": 0.9252, "step": 8808 }, { "epoch": 0.45, "learning_rate": 1.2168568153186018e-05, "loss": 0.8936, "step": 8809 }, { "epoch": 0.45, "learning_rate": 1.216696186316608e-05, "loss": 1.0112, "step": 8810 }, { "epoch": 0.45, "learning_rate": 1.2165355514477952e-05, "loss": 0.9741, "step": 8811 }, { "epoch": 0.45, "learning_rate": 1.2163749107165129e-05, "loss": 0.9088, "step": 8812 }, { "epoch": 0.45, "learning_rate": 1.21621426412711e-05, "loss": 1.0325, "step": 8813 }, { "epoch": 0.45, "learning_rate": 1.2160536116839357e-05, "loss": 0.7996, "step": 8814 }, { "epoch": 0.45, "learning_rate": 1.2158929533913399e-05, "loss": 0.9199, "step": 8815 }, { "epoch": 0.45, "learning_rate": 1.2157322892536723e-05, "loss": 0.9321, "step": 8816 }, { "epoch": 0.45, "learning_rate": 1.2155716192752822e-05, "loss": 0.8733, "step": 8817 }, { "epoch": 0.45, "learning_rate": 1.21541094346052e-05, "loss": 0.9594, "step": 8818 }, { "epoch": 0.45, "learning_rate": 1.2152502618137357e-05, "loss": 0.9634, "step": 8819 }, { "epoch": 0.45, "learning_rate": 1.2150895743392795e-05, "loss": 1.0093, "step": 8820 }, { "epoch": 0.45, "learning_rate": 1.2149288810415018e-05, "loss": 0.9017, "step": 8821 }, { "epoch": 0.45, "learning_rate": 1.2147681819247536e-05, "loss": 0.9611, "step": 8822 }, { "epoch": 0.45, "learning_rate": 1.2146074769933851e-05, "loss": 1.0259, "step": 8823 }, { "epoch": 0.45, "learning_rate": 1.2144467662517475e-05, "loss": 1.013, "step": 8824 }, { "epoch": 0.45, "learning_rate": 1.2142860497041921e-05, "loss": 0.9009, "step": 8825 }, { "epoch": 0.45, "learning_rate": 1.2141253273550698e-05, "loss": 0.9057, "step": 8826 }, { "epoch": 0.45, "learning_rate": 1.2139645992087319e-05, "loss": 0.8903, "step": 8827 }, { "epoch": 0.45, "learning_rate": 1.2138038652695304e-05, "loss": 0.9286, "step": 8828 }, { "epoch": 0.45, "learning_rate": 1.2136431255418165e-05, "loss": 0.9247, "step": 8829 }, { "epoch": 0.45, "learning_rate": 1.2134823800299423e-05, "loss": 0.8392, "step": 8830 }, { "epoch": 0.45, "learning_rate": 1.2133216287382598e-05, "loss": 0.8851, "step": 8831 }, { "epoch": 0.45, "learning_rate": 1.213160871671121e-05, "loss": 0.8726, "step": 8832 }, { "epoch": 0.45, "learning_rate": 1.2130001088328787e-05, "loss": 0.7975, "step": 8833 }, { "epoch": 0.45, "learning_rate": 1.212839340227885e-05, "loss": 0.9723, "step": 8834 }, { "epoch": 0.45, "learning_rate": 1.2126785658604925e-05, "loss": 0.9977, "step": 8835 }, { "epoch": 0.45, "learning_rate": 1.2125177857350538e-05, "loss": 0.9764, "step": 8836 }, { "epoch": 0.45, "learning_rate": 1.2123569998559226e-05, "loss": 0.9679, "step": 8837 }, { "epoch": 0.45, "learning_rate": 1.2121962082274514e-05, "loss": 0.9665, "step": 8838 }, { "epoch": 0.45, "learning_rate": 1.2120354108539938e-05, "loss": 0.9414, "step": 8839 }, { "epoch": 0.45, "learning_rate": 1.2118746077399029e-05, "loss": 0.9984, "step": 8840 }, { "epoch": 0.45, "learning_rate": 1.2117137988895324e-05, "loss": 0.9123, "step": 8841 }, { "epoch": 0.45, "learning_rate": 1.211552984307236e-05, "loss": 0.8469, "step": 8842 }, { "epoch": 0.45, "learning_rate": 1.2113921639973677e-05, "loss": 0.8402, "step": 8843 }, { "epoch": 0.45, "learning_rate": 1.2112313379642813e-05, "loss": 0.9767, "step": 8844 }, { "epoch": 0.45, "learning_rate": 1.2110705062123315e-05, "loss": 1.0374, "step": 8845 }, { "epoch": 0.45, "learning_rate": 1.210909668745872e-05, "loss": 0.9661, "step": 8846 }, { "epoch": 0.45, "learning_rate": 1.2107488255692578e-05, "loss": 1.018, "step": 8847 }, { "epoch": 0.45, "learning_rate": 1.210587976686843e-05, "loss": 0.9469, "step": 8848 }, { "epoch": 0.45, "learning_rate": 1.210427122102983e-05, "loss": 0.8626, "step": 8849 }, { "epoch": 0.45, "learning_rate": 1.2102662618220328e-05, "loss": 0.933, "step": 8850 }, { "epoch": 0.45, "learning_rate": 1.2101053958483472e-05, "loss": 0.9248, "step": 8851 }, { "epoch": 0.45, "learning_rate": 1.2099445241862812e-05, "loss": 1.0062, "step": 8852 }, { "epoch": 0.45, "learning_rate": 1.2097836468401911e-05, "loss": 0.8843, "step": 8853 }, { "epoch": 0.45, "learning_rate": 1.2096227638144314e-05, "loss": 0.8863, "step": 8854 }, { "epoch": 0.45, "learning_rate": 1.2094618751133585e-05, "loss": 0.9116, "step": 8855 }, { "epoch": 0.45, "learning_rate": 1.2093009807413285e-05, "loss": 0.8939, "step": 8856 }, { "epoch": 0.45, "learning_rate": 1.2091400807026969e-05, "loss": 0.9288, "step": 8857 }, { "epoch": 0.45, "learning_rate": 1.20897917500182e-05, "loss": 0.9257, "step": 8858 }, { "epoch": 0.45, "learning_rate": 1.2088182636430548e-05, "loss": 0.9962, "step": 8859 }, { "epoch": 0.45, "learning_rate": 1.2086573466307567e-05, "loss": 0.9623, "step": 8860 }, { "epoch": 0.45, "learning_rate": 1.2084964239692831e-05, "loss": 0.9903, "step": 8861 }, { "epoch": 0.45, "learning_rate": 1.2083354956629906e-05, "loss": 0.918, "step": 8862 }, { "epoch": 0.45, "learning_rate": 1.2081745617162366e-05, "loss": 0.9364, "step": 8863 }, { "epoch": 0.45, "learning_rate": 1.2080136221333773e-05, "loss": 0.8564, "step": 8864 }, { "epoch": 0.45, "learning_rate": 1.2078526769187708e-05, "loss": 0.9464, "step": 8865 }, { "epoch": 0.45, "learning_rate": 1.207691726076774e-05, "loss": 0.8932, "step": 8866 }, { "epoch": 0.45, "learning_rate": 1.2075307696117446e-05, "loss": 0.9413, "step": 8867 }, { "epoch": 0.45, "learning_rate": 1.207369807528041e-05, "loss": 0.9514, "step": 8868 }, { "epoch": 0.45, "learning_rate": 1.20720883983002e-05, "loss": 0.9425, "step": 8869 }, { "epoch": 0.45, "learning_rate": 1.2070478665220402e-05, "loss": 0.9139, "step": 8870 }, { "epoch": 0.45, "learning_rate": 1.2068868876084597e-05, "loss": 0.9979, "step": 8871 }, { "epoch": 0.45, "learning_rate": 1.2067259030936367e-05, "loss": 0.9178, "step": 8872 }, { "epoch": 0.45, "learning_rate": 1.20656491298193e-05, "loss": 0.9378, "step": 8873 }, { "epoch": 0.45, "learning_rate": 1.206403917277698e-05, "loss": 0.8886, "step": 8874 }, { "epoch": 0.45, "learning_rate": 1.2062429159852997e-05, "loss": 0.9001, "step": 8875 }, { "epoch": 0.45, "learning_rate": 1.2060819091090936e-05, "loss": 0.8764, "step": 8876 }, { "epoch": 0.45, "learning_rate": 1.205920896653439e-05, "loss": 0.8766, "step": 8877 }, { "epoch": 0.45, "learning_rate": 1.2057598786226953e-05, "loss": 0.9825, "step": 8878 }, { "epoch": 0.45, "learning_rate": 1.205598855021222e-05, "loss": 0.925, "step": 8879 }, { "epoch": 0.45, "learning_rate": 1.2054378258533784e-05, "loss": 0.9833, "step": 8880 }, { "epoch": 0.45, "learning_rate": 1.2052767911235242e-05, "loss": 0.8858, "step": 8881 }, { "epoch": 0.45, "learning_rate": 1.2051157508360191e-05, "loss": 0.9885, "step": 8882 }, { "epoch": 0.45, "learning_rate": 1.2049547049952232e-05, "loss": 0.9897, "step": 8883 }, { "epoch": 0.45, "learning_rate": 1.2047936536054972e-05, "loss": 0.9284, "step": 8884 }, { "epoch": 0.45, "learning_rate": 1.2046325966712005e-05, "loss": 0.9379, "step": 8885 }, { "epoch": 0.45, "learning_rate": 1.2044715341966941e-05, "loss": 0.9384, "step": 8886 }, { "epoch": 0.45, "learning_rate": 1.2043104661863386e-05, "loss": 0.8925, "step": 8887 }, { "epoch": 0.45, "learning_rate": 1.2041493926444943e-05, "loss": 0.9909, "step": 8888 }, { "epoch": 0.45, "learning_rate": 1.2039883135755224e-05, "loss": 0.9203, "step": 8889 }, { "epoch": 0.45, "learning_rate": 1.203827228983784e-05, "loss": 0.9218, "step": 8890 }, { "epoch": 0.45, "learning_rate": 1.2036661388736404e-05, "loss": 0.9567, "step": 8891 }, { "epoch": 0.45, "learning_rate": 1.2035050432494526e-05, "loss": 0.9865, "step": 8892 }, { "epoch": 0.45, "learning_rate": 1.2033439421155826e-05, "loss": 1.0328, "step": 8893 }, { "epoch": 0.45, "learning_rate": 1.2031828354763913e-05, "loss": 0.9106, "step": 8894 }, { "epoch": 0.45, "learning_rate": 1.2030217233362409e-05, "loss": 0.9468, "step": 8895 }, { "epoch": 0.45, "learning_rate": 1.2028606056994936e-05, "loss": 0.9979, "step": 8896 }, { "epoch": 0.45, "learning_rate": 1.2026994825705109e-05, "loss": 0.9144, "step": 8897 }, { "epoch": 0.45, "learning_rate": 1.2025383539536557e-05, "loss": 0.99, "step": 8898 }, { "epoch": 0.45, "learning_rate": 1.2023772198532902e-05, "loss": 0.9277, "step": 8899 }, { "epoch": 0.45, "learning_rate": 1.2022160802737761e-05, "loss": 0.8405, "step": 8900 }, { "epoch": 0.45, "learning_rate": 1.2020549352194772e-05, "loss": 1.0199, "step": 8901 }, { "epoch": 0.45, "learning_rate": 1.201893784694756e-05, "loss": 0.9149, "step": 8902 }, { "epoch": 0.45, "learning_rate": 1.201732628703975e-05, "loss": 0.844, "step": 8903 }, { "epoch": 0.45, "learning_rate": 1.2015714672514982e-05, "loss": 0.8849, "step": 8904 }, { "epoch": 0.45, "learning_rate": 1.201410300341688e-05, "loss": 0.8923, "step": 8905 }, { "epoch": 0.45, "learning_rate": 1.2012491279789085e-05, "loss": 0.8639, "step": 8906 }, { "epoch": 0.45, "learning_rate": 1.2010879501675226e-05, "loss": 0.8883, "step": 8907 }, { "epoch": 0.45, "learning_rate": 1.2009267669118945e-05, "loss": 0.8814, "step": 8908 }, { "epoch": 0.45, "learning_rate": 1.200765578216388e-05, "loss": 0.8653, "step": 8909 }, { "epoch": 0.45, "learning_rate": 1.2006043840853673e-05, "loss": 0.9586, "step": 8910 }, { "epoch": 0.45, "learning_rate": 1.200443184523196e-05, "loss": 0.9135, "step": 8911 }, { "epoch": 0.45, "learning_rate": 1.200281979534239e-05, "loss": 0.9714, "step": 8912 }, { "epoch": 0.45, "learning_rate": 1.2001207691228603e-05, "loss": 0.8736, "step": 8913 }, { "epoch": 0.45, "learning_rate": 1.1999595532934247e-05, "loss": 0.9741, "step": 8914 }, { "epoch": 0.45, "learning_rate": 1.199798332050297e-05, "loss": 0.9965, "step": 8915 }, { "epoch": 0.45, "learning_rate": 1.1996371053978421e-05, "loss": 0.8776, "step": 8916 }, { "epoch": 0.45, "learning_rate": 1.1994758733404247e-05, "loss": 0.8994, "step": 8917 }, { "epoch": 0.45, "learning_rate": 1.1993146358824107e-05, "loss": 0.9122, "step": 8918 }, { "epoch": 0.45, "learning_rate": 1.1991533930281646e-05, "loss": 0.849, "step": 8919 }, { "epoch": 0.45, "learning_rate": 1.1989921447820523e-05, "loss": 0.8527, "step": 8920 }, { "epoch": 0.45, "learning_rate": 1.1988308911484396e-05, "loss": 0.9594, "step": 8921 }, { "epoch": 0.45, "learning_rate": 1.198669632131692e-05, "loss": 0.9791, "step": 8922 }, { "epoch": 0.45, "learning_rate": 1.1985083677361753e-05, "loss": 1.0031, "step": 8923 }, { "epoch": 0.45, "learning_rate": 1.1983470979662561e-05, "loss": 0.9456, "step": 8924 }, { "epoch": 0.45, "learning_rate": 1.1981858228262999e-05, "loss": 1.0639, "step": 8925 }, { "epoch": 0.45, "learning_rate": 1.1980245423206733e-05, "loss": 0.9406, "step": 8926 }, { "epoch": 0.45, "learning_rate": 1.197863256453743e-05, "loss": 1.0353, "step": 8927 }, { "epoch": 0.45, "learning_rate": 1.197701965229876e-05, "loss": 0.8864, "step": 8928 }, { "epoch": 0.45, "learning_rate": 1.1975406686534382e-05, "loss": 1.0415, "step": 8929 }, { "epoch": 0.45, "learning_rate": 1.1973793667287971e-05, "loss": 0.8442, "step": 8930 }, { "epoch": 0.45, "learning_rate": 1.1972180594603196e-05, "loss": 0.9152, "step": 8931 }, { "epoch": 0.45, "learning_rate": 1.1970567468523728e-05, "loss": 0.8226, "step": 8932 }, { "epoch": 0.45, "learning_rate": 1.1968954289093244e-05, "loss": 0.9513, "step": 8933 }, { "epoch": 0.45, "learning_rate": 1.196734105635542e-05, "loss": 0.9626, "step": 8934 }, { "epoch": 0.45, "learning_rate": 1.1965727770353923e-05, "loss": 0.9069, "step": 8935 }, { "epoch": 0.45, "learning_rate": 1.1964114431132444e-05, "loss": 0.8905, "step": 8936 }, { "epoch": 0.45, "learning_rate": 1.1962501038734655e-05, "loss": 0.9608, "step": 8937 }, { "epoch": 0.45, "learning_rate": 1.1960887593204236e-05, "loss": 0.9351, "step": 8938 }, { "epoch": 0.45, "learning_rate": 1.1959274094584874e-05, "loss": 0.9863, "step": 8939 }, { "epoch": 0.45, "learning_rate": 1.1957660542920251e-05, "loss": 1.0153, "step": 8940 }, { "epoch": 0.45, "learning_rate": 1.1956046938254052e-05, "loss": 1.0292, "step": 8941 }, { "epoch": 0.45, "learning_rate": 1.1954433280629958e-05, "loss": 0.9877, "step": 8942 }, { "epoch": 0.45, "learning_rate": 1.1952819570091667e-05, "loss": 0.9243, "step": 8943 }, { "epoch": 0.45, "learning_rate": 1.1951205806682861e-05, "loss": 0.9511, "step": 8944 }, { "epoch": 0.45, "learning_rate": 1.1949591990447238e-05, "loss": 1.0253, "step": 8945 }, { "epoch": 0.45, "learning_rate": 1.1947978121428485e-05, "loss": 0.998, "step": 8946 }, { "epoch": 0.45, "learning_rate": 1.1946364199670294e-05, "loss": 0.9885, "step": 8947 }, { "epoch": 0.45, "learning_rate": 1.1944750225216363e-05, "loss": 0.9102, "step": 8948 }, { "epoch": 0.45, "learning_rate": 1.1943136198110392e-05, "loss": 0.9825, "step": 8949 }, { "epoch": 0.45, "learning_rate": 1.1941522118396073e-05, "loss": 0.8769, "step": 8950 }, { "epoch": 0.45, "learning_rate": 1.1939907986117109e-05, "loss": 0.9622, "step": 8951 }, { "epoch": 0.45, "learning_rate": 1.1938293801317202e-05, "loss": 0.9186, "step": 8952 }, { "epoch": 0.45, "learning_rate": 1.193667956404005e-05, "loss": 0.9354, "step": 8953 }, { "epoch": 0.45, "learning_rate": 1.1935065274329358e-05, "loss": 0.9626, "step": 8954 }, { "epoch": 0.45, "learning_rate": 1.1933450932228836e-05, "loss": 0.9373, "step": 8955 }, { "epoch": 0.45, "learning_rate": 1.1931836537782184e-05, "loss": 0.9866, "step": 8956 }, { "epoch": 0.46, "learning_rate": 1.1930222091033118e-05, "loss": 0.9461, "step": 8957 }, { "epoch": 0.46, "learning_rate": 1.1928607592025339e-05, "loss": 0.9143, "step": 8958 }, { "epoch": 0.46, "learning_rate": 1.1926993040802562e-05, "loss": 0.9908, "step": 8959 }, { "epoch": 0.46, "learning_rate": 1.1925378437408494e-05, "loss": 0.8934, "step": 8960 }, { "epoch": 0.46, "learning_rate": 1.1923763781886862e-05, "loss": 0.9361, "step": 8961 }, { "epoch": 0.46, "learning_rate": 1.1922149074281364e-05, "loss": 0.9276, "step": 8962 }, { "epoch": 0.46, "learning_rate": 1.192053431463573e-05, "loss": 0.8568, "step": 8963 }, { "epoch": 0.46, "learning_rate": 1.1918919502993674e-05, "loss": 0.9027, "step": 8964 }, { "epoch": 0.46, "learning_rate": 1.1917304639398908e-05, "loss": 0.9396, "step": 8965 }, { "epoch": 0.46, "learning_rate": 1.1915689723895162e-05, "loss": 0.8797, "step": 8966 }, { "epoch": 0.46, "learning_rate": 1.1914074756526154e-05, "loss": 0.8958, "step": 8967 }, { "epoch": 0.46, "learning_rate": 1.1912459737335609e-05, "loss": 0.9715, "step": 8968 }, { "epoch": 0.46, "learning_rate": 1.1910844666367254e-05, "loss": 0.8897, "step": 8969 }, { "epoch": 0.46, "learning_rate": 1.1909229543664811e-05, "loss": 0.907, "step": 8970 }, { "epoch": 0.46, "learning_rate": 1.1907614369272007e-05, "loss": 0.9619, "step": 8971 }, { "epoch": 0.46, "learning_rate": 1.1905999143232573e-05, "loss": 0.8932, "step": 8972 }, { "epoch": 0.46, "learning_rate": 1.1904383865590244e-05, "loss": 0.8924, "step": 8973 }, { "epoch": 0.46, "learning_rate": 1.1902768536388745e-05, "loss": 0.9025, "step": 8974 }, { "epoch": 0.46, "learning_rate": 1.1901153155671813e-05, "loss": 0.9351, "step": 8975 }, { "epoch": 0.46, "learning_rate": 1.1899537723483183e-05, "loss": 0.8899, "step": 8976 }, { "epoch": 0.46, "learning_rate": 1.1897922239866588e-05, "loss": 0.9442, "step": 8977 }, { "epoch": 0.46, "learning_rate": 1.189630670486577e-05, "loss": 0.905, "step": 8978 }, { "epoch": 0.46, "learning_rate": 1.1894691118524462e-05, "loss": 0.8552, "step": 8979 }, { "epoch": 0.46, "learning_rate": 1.1893075480886411e-05, "loss": 0.9888, "step": 8980 }, { "epoch": 0.46, "learning_rate": 1.1891459791995356e-05, "loss": 0.9408, "step": 8981 }, { "epoch": 0.46, "learning_rate": 1.1889844051895037e-05, "loss": 0.9301, "step": 8982 }, { "epoch": 0.46, "learning_rate": 1.1888228260629204e-05, "loss": 1.0324, "step": 8983 }, { "epoch": 0.46, "learning_rate": 1.1886612418241596e-05, "loss": 0.8924, "step": 8984 }, { "epoch": 0.46, "learning_rate": 1.1884996524775965e-05, "loss": 0.9124, "step": 8985 }, { "epoch": 0.46, "learning_rate": 1.188338058027606e-05, "loss": 1.002, "step": 8986 }, { "epoch": 0.46, "learning_rate": 1.188176458478563e-05, "loss": 0.8898, "step": 8987 }, { "epoch": 0.46, "learning_rate": 1.1880148538348423e-05, "loss": 0.903, "step": 8988 }, { "epoch": 0.46, "learning_rate": 1.1878532441008198e-05, "loss": 0.8903, "step": 8989 }, { "epoch": 0.46, "learning_rate": 1.1876916292808702e-05, "loss": 0.9798, "step": 8990 }, { "epoch": 0.46, "learning_rate": 1.1875300093793696e-05, "loss": 0.9645, "step": 8991 }, { "epoch": 0.46, "learning_rate": 1.1873683844006939e-05, "loss": 0.9605, "step": 8992 }, { "epoch": 0.46, "learning_rate": 1.1872067543492181e-05, "loss": 0.9515, "step": 8993 }, { "epoch": 0.46, "learning_rate": 1.1870451192293186e-05, "loss": 0.9813, "step": 8994 }, { "epoch": 0.46, "learning_rate": 1.1868834790453717e-05, "loss": 0.9093, "step": 8995 }, { "epoch": 0.46, "learning_rate": 1.1867218338017532e-05, "loss": 0.8995, "step": 8996 }, { "epoch": 0.46, "learning_rate": 1.1865601835028398e-05, "loss": 0.9377, "step": 8997 }, { "epoch": 0.46, "learning_rate": 1.186398528153008e-05, "loss": 0.8831, "step": 8998 }, { "epoch": 0.46, "learning_rate": 1.1862368677566343e-05, "loss": 0.8575, "step": 8999 }, { "epoch": 0.46, "learning_rate": 1.1860752023180954e-05, "loss": 1.0274, "step": 9000 }, { "epoch": 0.46, "learning_rate": 1.1859135318417686e-05, "loss": 0.8829, "step": 9001 }, { "epoch": 0.46, "learning_rate": 1.1857518563320306e-05, "loss": 0.8821, "step": 9002 }, { "epoch": 0.46, "learning_rate": 1.1855901757932586e-05, "loss": 1.0092, "step": 9003 }, { "epoch": 0.46, "learning_rate": 1.1854284902298304e-05, "loss": 0.8966, "step": 9004 }, { "epoch": 0.46, "learning_rate": 1.1852667996461228e-05, "loss": 0.8573, "step": 9005 }, { "epoch": 0.46, "learning_rate": 1.1851051040465137e-05, "loss": 0.9302, "step": 9006 }, { "epoch": 0.46, "learning_rate": 1.1849434034353813e-05, "loss": 0.9504, "step": 9007 }, { "epoch": 0.46, "learning_rate": 1.1847816978171024e-05, "loss": 0.9714, "step": 9008 }, { "epoch": 0.46, "learning_rate": 1.1846199871960557e-05, "loss": 0.8981, "step": 9009 }, { "epoch": 0.46, "learning_rate": 1.1844582715766197e-05, "loss": 0.849, "step": 9010 }, { "epoch": 0.46, "learning_rate": 1.184296550963172e-05, "loss": 0.9214, "step": 9011 }, { "epoch": 0.46, "learning_rate": 1.1841348253600913e-05, "loss": 0.9159, "step": 9012 }, { "epoch": 0.46, "learning_rate": 1.1839730947717558e-05, "loss": 0.9955, "step": 9013 }, { "epoch": 0.46, "learning_rate": 1.183811359202545e-05, "loss": 0.9597, "step": 9014 }, { "epoch": 0.46, "learning_rate": 1.183649618656837e-05, "loss": 1.0173, "step": 9015 }, { "epoch": 0.46, "learning_rate": 1.183487873139011e-05, "loss": 0.8793, "step": 9016 }, { "epoch": 0.46, "learning_rate": 1.183326122653446e-05, "loss": 0.9984, "step": 9017 }, { "epoch": 0.46, "learning_rate": 1.1831643672045212e-05, "loss": 0.9556, "step": 9018 }, { "epoch": 0.46, "learning_rate": 1.1830026067966159e-05, "loss": 0.8594, "step": 9019 }, { "epoch": 0.46, "learning_rate": 1.1828408414341101e-05, "loss": 0.8995, "step": 9020 }, { "epoch": 0.46, "learning_rate": 1.1826790711213827e-05, "loss": 0.9465, "step": 9021 }, { "epoch": 0.46, "learning_rate": 1.1825172958628142e-05, "loss": 0.8912, "step": 9022 }, { "epoch": 0.46, "learning_rate": 1.1823555156627837e-05, "loss": 0.9619, "step": 9023 }, { "epoch": 0.46, "learning_rate": 1.1821937305256719e-05, "loss": 1.0118, "step": 9024 }, { "epoch": 0.46, "learning_rate": 1.1820319404558582e-05, "loss": 0.9229, "step": 9025 }, { "epoch": 0.46, "learning_rate": 1.181870145457724e-05, "loss": 0.9648, "step": 9026 }, { "epoch": 0.46, "learning_rate": 1.1817083455356485e-05, "loss": 0.9644, "step": 9027 }, { "epoch": 0.46, "learning_rate": 1.1815465406940137e-05, "loss": 0.9807, "step": 9028 }, { "epoch": 0.46, "learning_rate": 1.1813847309371988e-05, "loss": 0.9576, "step": 9029 }, { "epoch": 0.46, "learning_rate": 1.1812229162695854e-05, "loss": 0.8281, "step": 9030 }, { "epoch": 0.46, "learning_rate": 1.1810610966955544e-05, "loss": 1.0387, "step": 9031 }, { "epoch": 0.46, "learning_rate": 1.180899272219487e-05, "loss": 0.9842, "step": 9032 }, { "epoch": 0.46, "learning_rate": 1.1807374428457639e-05, "loss": 0.8661, "step": 9033 }, { "epoch": 0.46, "learning_rate": 1.1805756085787672e-05, "loss": 0.9477, "step": 9034 }, { "epoch": 0.46, "learning_rate": 1.1804137694228778e-05, "loss": 0.9821, "step": 9035 }, { "epoch": 0.46, "learning_rate": 1.1802519253824776e-05, "loss": 0.8088, "step": 9036 }, { "epoch": 0.46, "learning_rate": 1.1800900764619483e-05, "loss": 0.921, "step": 9037 }, { "epoch": 0.46, "learning_rate": 1.179928222665672e-05, "loss": 0.9406, "step": 9038 }, { "epoch": 0.46, "learning_rate": 1.17976636399803e-05, "loss": 0.9647, "step": 9039 }, { "epoch": 0.46, "learning_rate": 1.1796045004634053e-05, "loss": 0.9141, "step": 9040 }, { "epoch": 0.46, "learning_rate": 1.1794426320661798e-05, "loss": 0.8781, "step": 9041 }, { "epoch": 0.46, "learning_rate": 1.1792807588107358e-05, "loss": 0.9738, "step": 9042 }, { "epoch": 0.46, "learning_rate": 1.179118880701456e-05, "loss": 0.9835, "step": 9043 }, { "epoch": 0.46, "learning_rate": 1.178956997742723e-05, "loss": 0.9023, "step": 9044 }, { "epoch": 0.46, "learning_rate": 1.1787951099389202e-05, "loss": 0.8571, "step": 9045 }, { "epoch": 0.46, "learning_rate": 1.1786332172944297e-05, "loss": 0.9157, "step": 9046 }, { "epoch": 0.46, "learning_rate": 1.1784713198136348e-05, "loss": 0.8821, "step": 9047 }, { "epoch": 0.46, "learning_rate": 1.1783094175009188e-05, "loss": 0.9436, "step": 9048 }, { "epoch": 0.46, "learning_rate": 1.178147510360665e-05, "loss": 0.8406, "step": 9049 }, { "epoch": 0.46, "learning_rate": 1.177985598397257e-05, "loss": 0.982, "step": 9050 }, { "epoch": 0.46, "learning_rate": 1.1778236816150784e-05, "loss": 0.9658, "step": 9051 }, { "epoch": 0.46, "learning_rate": 1.1776617600185127e-05, "loss": 0.9442, "step": 9052 }, { "epoch": 0.46, "learning_rate": 1.1774998336119436e-05, "loss": 0.8699, "step": 9053 }, { "epoch": 0.46, "learning_rate": 1.1773379023997557e-05, "loss": 0.9319, "step": 9054 }, { "epoch": 0.46, "learning_rate": 1.1771759663863327e-05, "loss": 0.9037, "step": 9055 }, { "epoch": 0.46, "learning_rate": 1.1770140255760588e-05, "loss": 0.9251, "step": 9056 }, { "epoch": 0.46, "learning_rate": 1.1768520799733185e-05, "loss": 0.957, "step": 9057 }, { "epoch": 0.46, "learning_rate": 1.1766901295824966e-05, "loss": 0.9386, "step": 9058 }, { "epoch": 0.46, "learning_rate": 1.176528174407977e-05, "loss": 0.8506, "step": 9059 }, { "epoch": 0.46, "learning_rate": 1.1763662144541453e-05, "loss": 0.9813, "step": 9060 }, { "epoch": 0.46, "learning_rate": 1.1762042497253856e-05, "loss": 0.9708, "step": 9061 }, { "epoch": 0.46, "learning_rate": 1.176042280226083e-05, "loss": 0.8683, "step": 9062 }, { "epoch": 0.46, "learning_rate": 1.1758803059606239e-05, "loss": 0.8336, "step": 9063 }, { "epoch": 0.46, "learning_rate": 1.175718326933392e-05, "loss": 0.9364, "step": 9064 }, { "epoch": 0.46, "learning_rate": 1.1755563431487735e-05, "loss": 0.8956, "step": 9065 }, { "epoch": 0.46, "learning_rate": 1.1753943546111539e-05, "loss": 0.9323, "step": 9066 }, { "epoch": 0.46, "learning_rate": 1.1752323613249183e-05, "loss": 0.916, "step": 9067 }, { "epoch": 0.46, "learning_rate": 1.1750703632944534e-05, "loss": 0.9589, "step": 9068 }, { "epoch": 0.46, "learning_rate": 1.1749083605241447e-05, "loss": 0.9239, "step": 9069 }, { "epoch": 0.46, "learning_rate": 1.1747463530183781e-05, "loss": 0.8924, "step": 9070 }, { "epoch": 0.46, "learning_rate": 1.1745843407815398e-05, "loss": 0.9263, "step": 9071 }, { "epoch": 0.46, "learning_rate": 1.1744223238180164e-05, "loss": 0.9263, "step": 9072 }, { "epoch": 0.46, "learning_rate": 1.1742603021321938e-05, "loss": 0.8702, "step": 9073 }, { "epoch": 0.46, "learning_rate": 1.1740982757284591e-05, "loss": 0.9749, "step": 9074 }, { "epoch": 0.46, "learning_rate": 1.173936244611199e-05, "loss": 0.8006, "step": 9075 }, { "epoch": 0.46, "learning_rate": 1.1737742087848e-05, "loss": 0.9093, "step": 9076 }, { "epoch": 0.46, "learning_rate": 1.1736121682536493e-05, "loss": 0.9116, "step": 9077 }, { "epoch": 0.46, "learning_rate": 1.1734501230221338e-05, "loss": 0.933, "step": 9078 }, { "epoch": 0.46, "learning_rate": 1.1732880730946405e-05, "loss": 0.8951, "step": 9079 }, { "epoch": 0.46, "learning_rate": 1.1731260184755573e-05, "loss": 0.8801, "step": 9080 }, { "epoch": 0.46, "learning_rate": 1.1729639591692712e-05, "loss": 0.9385, "step": 9081 }, { "epoch": 0.46, "learning_rate": 1.1728018951801701e-05, "loss": 0.8591, "step": 9082 }, { "epoch": 0.46, "learning_rate": 1.1726398265126413e-05, "loss": 0.8864, "step": 9083 }, { "epoch": 0.46, "learning_rate": 1.172477753171073e-05, "loss": 1.0307, "step": 9084 }, { "epoch": 0.46, "learning_rate": 1.172315675159853e-05, "loss": 0.9103, "step": 9085 }, { "epoch": 0.46, "learning_rate": 1.1721535924833693e-05, "loss": 0.909, "step": 9086 }, { "epoch": 0.46, "learning_rate": 1.1719915051460105e-05, "loss": 0.9695, "step": 9087 }, { "epoch": 0.46, "learning_rate": 1.1718294131521646e-05, "loss": 0.939, "step": 9088 }, { "epoch": 0.46, "learning_rate": 1.17166731650622e-05, "loss": 0.9194, "step": 9089 }, { "epoch": 0.46, "learning_rate": 1.1715052152125655e-05, "loss": 0.9615, "step": 9090 }, { "epoch": 0.46, "learning_rate": 1.1713431092755898e-05, "loss": 0.9258, "step": 9091 }, { "epoch": 0.46, "learning_rate": 1.1711809986996815e-05, "loss": 0.924, "step": 9092 }, { "epoch": 0.46, "learning_rate": 1.1710188834892303e-05, "loss": 0.9586, "step": 9093 }, { "epoch": 0.46, "learning_rate": 1.1708567636486245e-05, "loss": 0.8584, "step": 9094 }, { "epoch": 0.46, "learning_rate": 1.1706946391822534e-05, "loss": 0.869, "step": 9095 }, { "epoch": 0.46, "learning_rate": 1.1705325100945066e-05, "loss": 0.9464, "step": 9096 }, { "epoch": 0.46, "learning_rate": 1.1703703763897739e-05, "loss": 0.9064, "step": 9097 }, { "epoch": 0.46, "learning_rate": 1.170208238072444e-05, "loss": 0.9404, "step": 9098 }, { "epoch": 0.46, "learning_rate": 1.1700460951469075e-05, "loss": 0.8721, "step": 9099 }, { "epoch": 0.46, "learning_rate": 1.169883947617554e-05, "loss": 0.8607, "step": 9100 }, { "epoch": 0.46, "learning_rate": 1.169721795488773e-05, "loss": 0.928, "step": 9101 }, { "epoch": 0.46, "learning_rate": 1.169559638764955e-05, "loss": 0.9597, "step": 9102 }, { "epoch": 0.46, "learning_rate": 1.1693974774504903e-05, "loss": 0.8542, "step": 9103 }, { "epoch": 0.46, "learning_rate": 1.1692353115497692e-05, "loss": 0.8869, "step": 9104 }, { "epoch": 0.46, "learning_rate": 1.169073141067182e-05, "loss": 0.9016, "step": 9105 }, { "epoch": 0.46, "learning_rate": 1.1689109660071196e-05, "loss": 0.8776, "step": 9106 }, { "epoch": 0.46, "learning_rate": 1.1687487863739722e-05, "loss": 0.8617, "step": 9107 }, { "epoch": 0.46, "learning_rate": 1.1685866021721308e-05, "loss": 0.9001, "step": 9108 }, { "epoch": 0.46, "learning_rate": 1.168424413405987e-05, "loss": 0.9302, "step": 9109 }, { "epoch": 0.46, "learning_rate": 1.168262220079931e-05, "loss": 0.9214, "step": 9110 }, { "epoch": 0.46, "learning_rate": 1.1681000221983548e-05, "loss": 0.9145, "step": 9111 }, { "epoch": 0.46, "learning_rate": 1.167937819765649e-05, "loss": 0.8582, "step": 9112 }, { "epoch": 0.46, "learning_rate": 1.1677756127862054e-05, "loss": 0.9177, "step": 9113 }, { "epoch": 0.46, "learning_rate": 1.1676134012644155e-05, "loss": 0.9488, "step": 9114 }, { "epoch": 0.46, "learning_rate": 1.1674511852046713e-05, "loss": 0.8769, "step": 9115 }, { "epoch": 0.46, "learning_rate": 1.1672889646113643e-05, "loss": 0.9103, "step": 9116 }, { "epoch": 0.46, "learning_rate": 1.1671267394888868e-05, "loss": 0.8707, "step": 9117 }, { "epoch": 0.46, "learning_rate": 1.1669645098416305e-05, "loss": 0.9746, "step": 9118 }, { "epoch": 0.46, "learning_rate": 1.1668022756739878e-05, "loss": 1.0209, "step": 9119 }, { "epoch": 0.46, "learning_rate": 1.1666400369903507e-05, "loss": 1.0084, "step": 9120 }, { "epoch": 0.46, "learning_rate": 1.166477793795112e-05, "loss": 1.0326, "step": 9121 }, { "epoch": 0.46, "learning_rate": 1.1663155460926643e-05, "loss": 0.8942, "step": 9122 }, { "epoch": 0.46, "learning_rate": 1.1661532938874e-05, "loss": 0.9865, "step": 9123 }, { "epoch": 0.46, "learning_rate": 1.1659910371837122e-05, "loss": 1.0434, "step": 9124 }, { "epoch": 0.46, "learning_rate": 1.1658287759859936e-05, "loss": 0.9378, "step": 9125 }, { "epoch": 0.46, "learning_rate": 1.1656665102986371e-05, "loss": 0.9579, "step": 9126 }, { "epoch": 0.46, "learning_rate": 1.1655042401260362e-05, "loss": 0.8908, "step": 9127 }, { "epoch": 0.46, "learning_rate": 1.165341965472584e-05, "loss": 0.822, "step": 9128 }, { "epoch": 0.46, "learning_rate": 1.1651796863426745e-05, "loss": 0.9976, "step": 9129 }, { "epoch": 0.46, "learning_rate": 1.1650174027407002e-05, "loss": 0.928, "step": 9130 }, { "epoch": 0.46, "learning_rate": 1.1648551146710557e-05, "loss": 0.9953, "step": 9131 }, { "epoch": 0.46, "learning_rate": 1.164692822138134e-05, "loss": 0.826, "step": 9132 }, { "epoch": 0.46, "learning_rate": 1.1645305251463293e-05, "loss": 0.9595, "step": 9133 }, { "epoch": 0.46, "learning_rate": 1.164368223700036e-05, "loss": 0.8266, "step": 9134 }, { "epoch": 0.46, "learning_rate": 1.1642059178036477e-05, "loss": 0.9175, "step": 9135 }, { "epoch": 0.46, "learning_rate": 1.164043607461559e-05, "loss": 0.9277, "step": 9136 }, { "epoch": 0.46, "learning_rate": 1.1638812926781642e-05, "loss": 0.9582, "step": 9137 }, { "epoch": 0.46, "learning_rate": 1.1637189734578576e-05, "loss": 0.8911, "step": 9138 }, { "epoch": 0.46, "learning_rate": 1.1635566498050336e-05, "loss": 0.8745, "step": 9139 }, { "epoch": 0.46, "learning_rate": 1.1633943217240881e-05, "loss": 0.9183, "step": 9140 }, { "epoch": 0.46, "learning_rate": 1.1632319892194148e-05, "loss": 0.9127, "step": 9141 }, { "epoch": 0.46, "learning_rate": 1.163069652295409e-05, "loss": 0.9046, "step": 9142 }, { "epoch": 0.46, "learning_rate": 1.1629073109564662e-05, "loss": 1.0156, "step": 9143 }, { "epoch": 0.46, "learning_rate": 1.1627449652069809e-05, "loss": 0.9387, "step": 9144 }, { "epoch": 0.46, "learning_rate": 1.1625826150513488e-05, "loss": 0.988, "step": 9145 }, { "epoch": 0.46, "learning_rate": 1.1624202604939655e-05, "loss": 0.8385, "step": 9146 }, { "epoch": 0.46, "learning_rate": 1.1622579015392266e-05, "loss": 0.886, "step": 9147 }, { "epoch": 0.46, "learning_rate": 1.1620955381915273e-05, "loss": 0.9724, "step": 9148 }, { "epoch": 0.46, "learning_rate": 1.1619331704552641e-05, "loss": 0.825, "step": 9149 }, { "epoch": 0.46, "learning_rate": 1.1617707983348326e-05, "loss": 0.9097, "step": 9150 }, { "epoch": 0.46, "learning_rate": 1.1616084218346285e-05, "loss": 0.8752, "step": 9151 }, { "epoch": 0.46, "learning_rate": 1.1614460409590484e-05, "loss": 0.8758, "step": 9152 }, { "epoch": 0.46, "learning_rate": 1.161283655712489e-05, "loss": 0.8733, "step": 9153 }, { "epoch": 0.47, "learning_rate": 1.1611212660993456e-05, "loss": 0.9509, "step": 9154 }, { "epoch": 0.47, "learning_rate": 1.1609588721240154e-05, "loss": 0.8888, "step": 9155 }, { "epoch": 0.47, "learning_rate": 1.1607964737908954e-05, "loss": 0.874, "step": 9156 }, { "epoch": 0.47, "learning_rate": 1.1606340711043815e-05, "loss": 0.8732, "step": 9157 }, { "epoch": 0.47, "learning_rate": 1.1604716640688714e-05, "loss": 0.9174, "step": 9158 }, { "epoch": 0.47, "learning_rate": 1.1603092526887616e-05, "loss": 0.9426, "step": 9159 }, { "epoch": 0.47, "learning_rate": 1.1601468369684492e-05, "loss": 1.0115, "step": 9160 }, { "epoch": 0.47, "learning_rate": 1.1599844169123315e-05, "loss": 0.9305, "step": 9161 }, { "epoch": 0.47, "learning_rate": 1.1598219925248062e-05, "loss": 1.0056, "step": 9162 }, { "epoch": 0.47, "learning_rate": 1.1596595638102704e-05, "loss": 0.8858, "step": 9163 }, { "epoch": 0.47, "learning_rate": 1.1594971307731217e-05, "loss": 0.9277, "step": 9164 }, { "epoch": 0.47, "learning_rate": 1.1593346934177581e-05, "loss": 0.8566, "step": 9165 }, { "epoch": 0.47, "learning_rate": 1.1591722517485768e-05, "loss": 0.9371, "step": 9166 }, { "epoch": 0.47, "learning_rate": 1.1590098057699763e-05, "loss": 0.9199, "step": 9167 }, { "epoch": 0.47, "learning_rate": 1.1588473554863546e-05, "loss": 0.988, "step": 9168 }, { "epoch": 0.47, "learning_rate": 1.1586849009021097e-05, "loss": 0.9037, "step": 9169 }, { "epoch": 0.47, "learning_rate": 1.1585224420216399e-05, "loss": 0.9331, "step": 9170 }, { "epoch": 0.47, "learning_rate": 1.1583599788493437e-05, "loss": 0.8531, "step": 9171 }, { "epoch": 0.47, "learning_rate": 1.1581975113896196e-05, "loss": 0.9391, "step": 9172 }, { "epoch": 0.47, "learning_rate": 1.158035039646866e-05, "loss": 1.0095, "step": 9173 }, { "epoch": 0.47, "learning_rate": 1.1578725636254819e-05, "loss": 0.864, "step": 9174 }, { "epoch": 0.47, "learning_rate": 1.157710083329866e-05, "loss": 0.8423, "step": 9175 }, { "epoch": 0.47, "learning_rate": 1.1575475987644179e-05, "loss": 1.0452, "step": 9176 }, { "epoch": 0.47, "learning_rate": 1.1573851099335357e-05, "loss": 1.0134, "step": 9177 }, { "epoch": 0.47, "learning_rate": 1.1572226168416191e-05, "loss": 0.9727, "step": 9178 }, { "epoch": 0.47, "learning_rate": 1.1570601194930675e-05, "loss": 0.9553, "step": 9179 }, { "epoch": 0.47, "learning_rate": 1.1568976178922805e-05, "loss": 1.0162, "step": 9180 }, { "epoch": 0.47, "learning_rate": 1.156735112043657e-05, "loss": 0.9081, "step": 9181 }, { "epoch": 0.47, "learning_rate": 1.1565726019515972e-05, "loss": 0.9172, "step": 9182 }, { "epoch": 0.47, "learning_rate": 1.156410087620501e-05, "loss": 1.0276, "step": 9183 }, { "epoch": 0.47, "learning_rate": 1.156247569054768e-05, "loss": 0.9608, "step": 9184 }, { "epoch": 0.47, "learning_rate": 1.156085046258798e-05, "loss": 0.8771, "step": 9185 }, { "epoch": 0.47, "learning_rate": 1.1559225192369916e-05, "loss": 0.9447, "step": 9186 }, { "epoch": 0.47, "learning_rate": 1.155759987993749e-05, "loss": 0.9518, "step": 9187 }, { "epoch": 0.47, "learning_rate": 1.1555974525334707e-05, "loss": 0.9919, "step": 9188 }, { "epoch": 0.47, "learning_rate": 1.1554349128605566e-05, "loss": 0.8851, "step": 9189 }, { "epoch": 0.47, "learning_rate": 1.1552723689794076e-05, "loss": 0.9802, "step": 9190 }, { "epoch": 0.47, "learning_rate": 1.1551098208944244e-05, "loss": 0.9467, "step": 9191 }, { "epoch": 0.47, "learning_rate": 1.1549472686100079e-05, "loss": 0.9753, "step": 9192 }, { "epoch": 0.47, "learning_rate": 1.1547847121305589e-05, "loss": 0.9969, "step": 9193 }, { "epoch": 0.47, "learning_rate": 1.1546221514604788e-05, "loss": 1.012, "step": 9194 }, { "epoch": 0.47, "learning_rate": 1.1544595866041679e-05, "loss": 0.909, "step": 9195 }, { "epoch": 0.47, "learning_rate": 1.1542970175660284e-05, "loss": 1.015, "step": 9196 }, { "epoch": 0.47, "learning_rate": 1.1541344443504613e-05, "loss": 0.9435, "step": 9197 }, { "epoch": 0.47, "learning_rate": 1.1539718669618678e-05, "loss": 0.8717, "step": 9198 }, { "epoch": 0.47, "learning_rate": 1.15380928540465e-05, "loss": 0.9283, "step": 9199 }, { "epoch": 0.47, "learning_rate": 1.1536466996832097e-05, "loss": 0.9368, "step": 9200 }, { "epoch": 0.47, "learning_rate": 1.153484109801948e-05, "loss": 0.8405, "step": 9201 }, { "epoch": 0.47, "learning_rate": 1.1533215157652677e-05, "loss": 0.8927, "step": 9202 }, { "epoch": 0.47, "learning_rate": 1.1531589175775704e-05, "loss": 1.007, "step": 9203 }, { "epoch": 0.47, "learning_rate": 1.1529963152432581e-05, "loss": 0.9189, "step": 9204 }, { "epoch": 0.47, "learning_rate": 1.1528337087667336e-05, "loss": 0.9137, "step": 9205 }, { "epoch": 0.47, "learning_rate": 1.152671098152399e-05, "loss": 0.8605, "step": 9206 }, { "epoch": 0.47, "learning_rate": 1.1525084834046567e-05, "loss": 1.0575, "step": 9207 }, { "epoch": 0.47, "learning_rate": 1.1523458645279097e-05, "loss": 0.9496, "step": 9208 }, { "epoch": 0.47, "learning_rate": 1.1521832415265603e-05, "loss": 0.8568, "step": 9209 }, { "epoch": 0.47, "learning_rate": 1.1520206144050116e-05, "loss": 0.9655, "step": 9210 }, { "epoch": 0.47, "learning_rate": 1.1518579831676666e-05, "loss": 0.8328, "step": 9211 }, { "epoch": 0.47, "learning_rate": 1.1516953478189284e-05, "loss": 0.8945, "step": 9212 }, { "epoch": 0.47, "learning_rate": 1.1515327083631998e-05, "loss": 1.0392, "step": 9213 }, { "epoch": 0.47, "learning_rate": 1.1513700648048844e-05, "loss": 0.8489, "step": 9214 }, { "epoch": 0.47, "learning_rate": 1.1512074171483853e-05, "loss": 1.073, "step": 9215 }, { "epoch": 0.47, "learning_rate": 1.1510447653981065e-05, "loss": 0.959, "step": 9216 }, { "epoch": 0.47, "learning_rate": 1.1508821095584515e-05, "loss": 1.0385, "step": 9217 }, { "epoch": 0.47, "learning_rate": 1.1507194496338238e-05, "loss": 0.766, "step": 9218 }, { "epoch": 0.47, "learning_rate": 1.1505567856286272e-05, "loss": 0.9933, "step": 9219 }, { "epoch": 0.47, "learning_rate": 1.1503941175472659e-05, "loss": 0.9257, "step": 9220 }, { "epoch": 0.47, "learning_rate": 1.1502314453941441e-05, "loss": 0.9476, "step": 9221 }, { "epoch": 0.47, "learning_rate": 1.1500687691736654e-05, "loss": 0.8571, "step": 9222 }, { "epoch": 0.47, "learning_rate": 1.1499060888902347e-05, "loss": 0.9006, "step": 9223 }, { "epoch": 0.47, "learning_rate": 1.1497434045482561e-05, "loss": 0.8729, "step": 9224 }, { "epoch": 0.47, "learning_rate": 1.149580716152134e-05, "loss": 0.9279, "step": 9225 }, { "epoch": 0.47, "learning_rate": 1.149418023706273e-05, "loss": 0.894, "step": 9226 }, { "epoch": 0.47, "learning_rate": 1.1492553272150784e-05, "loss": 0.8642, "step": 9227 }, { "epoch": 0.47, "learning_rate": 1.1490926266829542e-05, "loss": 0.9174, "step": 9228 }, { "epoch": 0.47, "learning_rate": 1.1489299221143063e-05, "loss": 0.9138, "step": 9229 }, { "epoch": 0.47, "learning_rate": 1.1487672135135388e-05, "loss": 0.9401, "step": 9230 }, { "epoch": 0.47, "learning_rate": 1.1486045008850572e-05, "loss": 0.9498, "step": 9231 }, { "epoch": 0.47, "learning_rate": 1.148441784233267e-05, "loss": 0.8837, "step": 9232 }, { "epoch": 0.47, "learning_rate": 1.1482790635625733e-05, "loss": 0.9236, "step": 9233 }, { "epoch": 0.47, "learning_rate": 1.1481163388773818e-05, "loss": 0.9551, "step": 9234 }, { "epoch": 0.47, "learning_rate": 1.1479536101820981e-05, "loss": 0.9528, "step": 9235 }, { "epoch": 0.47, "learning_rate": 1.147790877481128e-05, "loss": 0.9203, "step": 9236 }, { "epoch": 0.47, "learning_rate": 1.1476281407788768e-05, "loss": 1.003, "step": 9237 }, { "epoch": 0.47, "learning_rate": 1.1474654000797505e-05, "loss": 0.9791, "step": 9238 }, { "epoch": 0.47, "learning_rate": 1.1473026553881559e-05, "loss": 0.8899, "step": 9239 }, { "epoch": 0.47, "learning_rate": 1.1471399067084983e-05, "loss": 0.8391, "step": 9240 }, { "epoch": 0.47, "learning_rate": 1.1469771540451844e-05, "loss": 0.9851, "step": 9241 }, { "epoch": 0.47, "learning_rate": 1.1468143974026207e-05, "loss": 0.9305, "step": 9242 }, { "epoch": 0.47, "learning_rate": 1.146651636785213e-05, "loss": 0.9837, "step": 9243 }, { "epoch": 0.47, "learning_rate": 1.1464888721973682e-05, "loss": 0.8844, "step": 9244 }, { "epoch": 0.47, "learning_rate": 1.1463261036434932e-05, "loss": 0.9638, "step": 9245 }, { "epoch": 0.47, "learning_rate": 1.1461633311279945e-05, "loss": 0.8578, "step": 9246 }, { "epoch": 0.47, "learning_rate": 1.1460005546552794e-05, "loss": 0.8986, "step": 9247 }, { "epoch": 0.47, "learning_rate": 1.1458377742297546e-05, "loss": 0.8852, "step": 9248 }, { "epoch": 0.47, "learning_rate": 1.145674989855827e-05, "loss": 0.8881, "step": 9249 }, { "epoch": 0.47, "learning_rate": 1.1455122015379039e-05, "loss": 0.94, "step": 9250 }, { "epoch": 0.47, "learning_rate": 1.1453494092803932e-05, "loss": 0.9288, "step": 9251 }, { "epoch": 0.47, "learning_rate": 1.1451866130877014e-05, "loss": 1.0013, "step": 9252 }, { "epoch": 0.47, "learning_rate": 1.145023812964237e-05, "loss": 0.8758, "step": 9253 }, { "epoch": 0.47, "learning_rate": 1.1448610089144068e-05, "loss": 0.9854, "step": 9254 }, { "epoch": 0.47, "learning_rate": 1.144698200942619e-05, "loss": 0.9548, "step": 9255 }, { "epoch": 0.47, "learning_rate": 1.1445353890532812e-05, "loss": 0.9104, "step": 9256 }, { "epoch": 0.47, "learning_rate": 1.1443725732508018e-05, "loss": 1.0892, "step": 9257 }, { "epoch": 0.47, "learning_rate": 1.1442097535395886e-05, "loss": 0.9944, "step": 9258 }, { "epoch": 0.47, "learning_rate": 1.1440469299240498e-05, "loss": 0.8928, "step": 9259 }, { "epoch": 0.47, "learning_rate": 1.1438841024085931e-05, "loss": 0.9716, "step": 9260 }, { "epoch": 0.47, "learning_rate": 1.143721270997628e-05, "loss": 0.9715, "step": 9261 }, { "epoch": 0.47, "learning_rate": 1.143558435695562e-05, "loss": 0.9224, "step": 9262 }, { "epoch": 0.47, "learning_rate": 1.1433955965068042e-05, "loss": 0.9004, "step": 9263 }, { "epoch": 0.47, "learning_rate": 1.1432327534357632e-05, "loss": 0.8898, "step": 9264 }, { "epoch": 0.47, "learning_rate": 1.143069906486848e-05, "loss": 0.9774, "step": 9265 }, { "epoch": 0.47, "learning_rate": 1.142907055664467e-05, "loss": 0.8629, "step": 9266 }, { "epoch": 0.47, "learning_rate": 1.1427442009730298e-05, "loss": 1.006, "step": 9267 }, { "epoch": 0.47, "learning_rate": 1.142581342416945e-05, "loss": 0.8936, "step": 9268 }, { "epoch": 0.47, "learning_rate": 1.142418480000622e-05, "loss": 0.9764, "step": 9269 }, { "epoch": 0.47, "learning_rate": 1.1422556137284704e-05, "loss": 0.9921, "step": 9270 }, { "epoch": 0.47, "learning_rate": 1.1420927436048993e-05, "loss": 0.9592, "step": 9271 }, { "epoch": 0.47, "learning_rate": 1.1419298696343183e-05, "loss": 0.957, "step": 9272 }, { "epoch": 0.47, "learning_rate": 1.1417669918211371e-05, "loss": 0.8259, "step": 9273 }, { "epoch": 0.47, "learning_rate": 1.1416041101697655e-05, "loss": 0.8781, "step": 9274 }, { "epoch": 0.47, "learning_rate": 1.141441224684613e-05, "loss": 0.9479, "step": 9275 }, { "epoch": 0.47, "learning_rate": 1.14127833537009e-05, "loss": 1.0345, "step": 9276 }, { "epoch": 0.47, "learning_rate": 1.1411154422306066e-05, "loss": 0.9421, "step": 9277 }, { "epoch": 0.47, "learning_rate": 1.1409525452705723e-05, "loss": 0.9307, "step": 9278 }, { "epoch": 0.47, "learning_rate": 1.1407896444943982e-05, "loss": 0.9271, "step": 9279 }, { "epoch": 0.47, "learning_rate": 1.140626739906494e-05, "loss": 1.0261, "step": 9280 }, { "epoch": 0.47, "learning_rate": 1.1404638315112706e-05, "loss": 0.9762, "step": 9281 }, { "epoch": 0.47, "learning_rate": 1.1403009193131383e-05, "loss": 0.9118, "step": 9282 }, { "epoch": 0.47, "learning_rate": 1.1401380033165081e-05, "loss": 0.9071, "step": 9283 }, { "epoch": 0.47, "learning_rate": 1.13997508352579e-05, "loss": 0.9393, "step": 9284 }, { "epoch": 0.47, "learning_rate": 1.1398121599453963e-05, "loss": 0.9312, "step": 9285 }, { "epoch": 0.47, "learning_rate": 1.1396492325797364e-05, "loss": 0.9325, "step": 9286 }, { "epoch": 0.47, "learning_rate": 1.1394863014332223e-05, "loss": 0.9455, "step": 9287 }, { "epoch": 0.47, "learning_rate": 1.139323366510265e-05, "loss": 0.9797, "step": 9288 }, { "epoch": 0.47, "learning_rate": 1.1391604278152759e-05, "loss": 0.9681, "step": 9289 }, { "epoch": 0.47, "learning_rate": 1.138997485352666e-05, "loss": 0.8936, "step": 9290 }, { "epoch": 0.47, "learning_rate": 1.1388345391268471e-05, "loss": 0.9125, "step": 9291 }, { "epoch": 0.47, "learning_rate": 1.138671589142231e-05, "loss": 0.8481, "step": 9292 }, { "epoch": 0.47, "learning_rate": 1.1385086354032291e-05, "loss": 0.9098, "step": 9293 }, { "epoch": 0.47, "learning_rate": 1.1383456779142531e-05, "loss": 0.9637, "step": 9294 }, { "epoch": 0.47, "learning_rate": 1.1381827166797154e-05, "loss": 0.9308, "step": 9295 }, { "epoch": 0.47, "learning_rate": 1.1380197517040275e-05, "loss": 0.9065, "step": 9296 }, { "epoch": 0.47, "learning_rate": 1.1378567829916013e-05, "loss": 0.8996, "step": 9297 }, { "epoch": 0.47, "learning_rate": 1.1376938105468497e-05, "loss": 0.856, "step": 9298 }, { "epoch": 0.47, "learning_rate": 1.1375308343741846e-05, "loss": 0.9788, "step": 9299 }, { "epoch": 0.47, "learning_rate": 1.1373678544780186e-05, "loss": 0.9431, "step": 9300 }, { "epoch": 0.47, "learning_rate": 1.1372048708627644e-05, "loss": 1.0115, "step": 9301 }, { "epoch": 0.47, "learning_rate": 1.1370418835328336e-05, "loss": 0.9677, "step": 9302 }, { "epoch": 0.47, "learning_rate": 1.1368788924926398e-05, "loss": 0.9557, "step": 9303 }, { "epoch": 0.47, "learning_rate": 1.1367158977465957e-05, "loss": 1.014, "step": 9304 }, { "epoch": 0.47, "learning_rate": 1.1365528992991142e-05, "loss": 0.888, "step": 9305 }, { "epoch": 0.47, "learning_rate": 1.1363898971546083e-05, "loss": 0.9719, "step": 9306 }, { "epoch": 0.47, "learning_rate": 1.136226891317491e-05, "loss": 0.9916, "step": 9307 }, { "epoch": 0.47, "learning_rate": 1.1360638817921755e-05, "loss": 0.9127, "step": 9308 }, { "epoch": 0.47, "learning_rate": 1.1359008685830749e-05, "loss": 0.8092, "step": 9309 }, { "epoch": 0.47, "learning_rate": 1.1357378516946031e-05, "loss": 1.0092, "step": 9310 }, { "epoch": 0.47, "learning_rate": 1.1355748311311734e-05, "loss": 0.9253, "step": 9311 }, { "epoch": 0.47, "learning_rate": 1.1354118068971993e-05, "loss": 0.9005, "step": 9312 }, { "epoch": 0.47, "learning_rate": 1.135248778997095e-05, "loss": 1.0024, "step": 9313 }, { "epoch": 0.47, "learning_rate": 1.1350857474352734e-05, "loss": 0.8657, "step": 9314 }, { "epoch": 0.47, "learning_rate": 1.1349227122161487e-05, "loss": 0.9327, "step": 9315 }, { "epoch": 0.47, "learning_rate": 1.1347596733441357e-05, "loss": 0.9458, "step": 9316 }, { "epoch": 0.47, "learning_rate": 1.1345966308236479e-05, "loss": 0.887, "step": 9317 }, { "epoch": 0.47, "learning_rate": 1.1344335846590992e-05, "loss": 0.8289, "step": 9318 }, { "epoch": 0.47, "learning_rate": 1.1342705348549046e-05, "loss": 0.9658, "step": 9319 }, { "epoch": 0.47, "learning_rate": 1.1341074814154775e-05, "loss": 0.8782, "step": 9320 }, { "epoch": 0.47, "learning_rate": 1.1339444243452333e-05, "loss": 0.9052, "step": 9321 }, { "epoch": 0.47, "learning_rate": 1.1337813636485863e-05, "loss": 0.9839, "step": 9322 }, { "epoch": 0.47, "learning_rate": 1.1336182993299516e-05, "loss": 0.9339, "step": 9323 }, { "epoch": 0.47, "learning_rate": 1.1334552313937434e-05, "loss": 0.946, "step": 9324 }, { "epoch": 0.47, "learning_rate": 1.1332921598443765e-05, "loss": 0.836, "step": 9325 }, { "epoch": 0.47, "learning_rate": 1.1331290846862665e-05, "loss": 0.9841, "step": 9326 }, { "epoch": 0.47, "learning_rate": 1.132966005923828e-05, "loss": 0.8972, "step": 9327 }, { "epoch": 0.47, "learning_rate": 1.1328029235614764e-05, "loss": 1.0131, "step": 9328 }, { "epoch": 0.47, "learning_rate": 1.1326398376036269e-05, "loss": 0.9761, "step": 9329 }, { "epoch": 0.47, "learning_rate": 1.1324767480546951e-05, "loss": 0.9321, "step": 9330 }, { "epoch": 0.47, "learning_rate": 1.1323136549190962e-05, "loss": 0.9617, "step": 9331 }, { "epoch": 0.47, "learning_rate": 1.1321505582012459e-05, "loss": 0.8903, "step": 9332 }, { "epoch": 0.47, "learning_rate": 1.1319874579055596e-05, "loss": 0.9357, "step": 9333 }, { "epoch": 0.47, "learning_rate": 1.1318243540364535e-05, "loss": 0.9387, "step": 9334 }, { "epoch": 0.47, "learning_rate": 1.1316612465983432e-05, "loss": 0.9443, "step": 9335 }, { "epoch": 0.47, "learning_rate": 1.1314981355956449e-05, "loss": 0.9023, "step": 9336 }, { "epoch": 0.47, "learning_rate": 1.1313350210327743e-05, "loss": 0.9393, "step": 9337 }, { "epoch": 0.47, "learning_rate": 1.1311719029141476e-05, "loss": 0.9114, "step": 9338 }, { "epoch": 0.47, "learning_rate": 1.1310087812441814e-05, "loss": 0.8683, "step": 9339 }, { "epoch": 0.47, "learning_rate": 1.1308456560272918e-05, "loss": 0.8918, "step": 9340 }, { "epoch": 0.47, "learning_rate": 1.1306825272678954e-05, "loss": 1.0161, "step": 9341 }, { "epoch": 0.47, "learning_rate": 1.1305193949704085e-05, "loss": 0.9133, "step": 9342 }, { "epoch": 0.47, "learning_rate": 1.1303562591392477e-05, "loss": 0.8643, "step": 9343 }, { "epoch": 0.47, "learning_rate": 1.1301931197788301e-05, "loss": 0.9327, "step": 9344 }, { "epoch": 0.47, "learning_rate": 1.1300299768935722e-05, "loss": 1.0156, "step": 9345 }, { "epoch": 0.47, "learning_rate": 1.1298668304878908e-05, "loss": 0.8446, "step": 9346 }, { "epoch": 0.47, "learning_rate": 1.1297036805662034e-05, "loss": 0.9242, "step": 9347 }, { "epoch": 0.47, "learning_rate": 1.1295405271329267e-05, "loss": 0.8974, "step": 9348 }, { "epoch": 0.47, "learning_rate": 1.1293773701924782e-05, "loss": 0.888, "step": 9349 }, { "epoch": 0.48, "learning_rate": 1.1292142097492752e-05, "loss": 0.8495, "step": 9350 }, { "epoch": 0.48, "learning_rate": 1.1290510458077344e-05, "loss": 0.9118, "step": 9351 }, { "epoch": 0.48, "learning_rate": 1.1288878783722741e-05, "loss": 0.9059, "step": 9352 }, { "epoch": 0.48, "learning_rate": 1.1287247074473118e-05, "loss": 0.96, "step": 9353 }, { "epoch": 0.48, "learning_rate": 1.1285615330372648e-05, "loss": 0.9081, "step": 9354 }, { "epoch": 0.48, "learning_rate": 1.1283983551465512e-05, "loss": 0.7759, "step": 9355 }, { "epoch": 0.48, "learning_rate": 1.1282351737795887e-05, "loss": 0.9745, "step": 9356 }, { "epoch": 0.48, "learning_rate": 1.1280719889407951e-05, "loss": 0.9504, "step": 9357 }, { "epoch": 0.48, "learning_rate": 1.1279088006345887e-05, "loss": 0.8537, "step": 9358 }, { "epoch": 0.48, "learning_rate": 1.127745608865388e-05, "loss": 0.9564, "step": 9359 }, { "epoch": 0.48, "learning_rate": 1.1275824136376106e-05, "loss": 0.9652, "step": 9360 }, { "epoch": 0.48, "learning_rate": 1.1274192149556748e-05, "loss": 0.8795, "step": 9361 }, { "epoch": 0.48, "learning_rate": 1.1272560128239995e-05, "loss": 0.8927, "step": 9362 }, { "epoch": 0.48, "learning_rate": 1.1270928072470034e-05, "loss": 0.9576, "step": 9363 }, { "epoch": 0.48, "learning_rate": 1.1269295982291045e-05, "loss": 1.0203, "step": 9364 }, { "epoch": 0.48, "learning_rate": 1.1267663857747217e-05, "loss": 0.8767, "step": 9365 }, { "epoch": 0.48, "learning_rate": 1.126603169888274e-05, "loss": 0.9753, "step": 9366 }, { "epoch": 0.48, "learning_rate": 1.1264399505741801e-05, "loss": 0.9484, "step": 9367 }, { "epoch": 0.48, "learning_rate": 1.1262767278368589e-05, "loss": 0.9737, "step": 9368 }, { "epoch": 0.48, "learning_rate": 1.12611350168073e-05, "loss": 0.9511, "step": 9369 }, { "epoch": 0.48, "learning_rate": 1.1259502721102118e-05, "loss": 0.9748, "step": 9370 }, { "epoch": 0.48, "learning_rate": 1.1257870391297243e-05, "loss": 0.9428, "step": 9371 }, { "epoch": 0.48, "learning_rate": 1.1256238027436867e-05, "loss": 0.9965, "step": 9372 }, { "epoch": 0.48, "learning_rate": 1.1254605629565179e-05, "loss": 0.8891, "step": 9373 }, { "epoch": 0.48, "learning_rate": 1.1252973197726377e-05, "loss": 0.9609, "step": 9374 }, { "epoch": 0.48, "learning_rate": 1.1251340731964664e-05, "loss": 0.8306, "step": 9375 }, { "epoch": 0.48, "learning_rate": 1.1249708232324228e-05, "loss": 0.9242, "step": 9376 }, { "epoch": 0.48, "learning_rate": 1.1248075698849275e-05, "loss": 0.8898, "step": 9377 }, { "epoch": 0.48, "learning_rate": 1.1246443131583999e-05, "loss": 0.8565, "step": 9378 }, { "epoch": 0.48, "learning_rate": 1.1244810530572599e-05, "loss": 0.9271, "step": 9379 }, { "epoch": 0.48, "learning_rate": 1.1243177895859279e-05, "loss": 0.9525, "step": 9380 }, { "epoch": 0.48, "learning_rate": 1.1241545227488242e-05, "loss": 0.9243, "step": 9381 }, { "epoch": 0.48, "learning_rate": 1.1239912525503687e-05, "loss": 0.8965, "step": 9382 }, { "epoch": 0.48, "learning_rate": 1.123827978994982e-05, "loss": 0.9752, "step": 9383 }, { "epoch": 0.48, "learning_rate": 1.1236647020870848e-05, "loss": 1.0201, "step": 9384 }, { "epoch": 0.48, "learning_rate": 1.1235014218310968e-05, "loss": 0.8898, "step": 9385 }, { "epoch": 0.48, "learning_rate": 1.1233381382314395e-05, "loss": 0.9261, "step": 9386 }, { "epoch": 0.48, "learning_rate": 1.1231748512925334e-05, "loss": 0.8774, "step": 9387 }, { "epoch": 0.48, "learning_rate": 1.123011561018799e-05, "loss": 0.905, "step": 9388 }, { "epoch": 0.48, "learning_rate": 1.1228482674146577e-05, "loss": 0.8364, "step": 9389 }, { "epoch": 0.48, "learning_rate": 1.1226849704845304e-05, "loss": 1.0283, "step": 9390 }, { "epoch": 0.48, "learning_rate": 1.1225216702328377e-05, "loss": 0.8089, "step": 9391 }, { "epoch": 0.48, "learning_rate": 1.1223583666640012e-05, "loss": 0.9067, "step": 9392 }, { "epoch": 0.48, "learning_rate": 1.1221950597824421e-05, "loss": 0.9304, "step": 9393 }, { "epoch": 0.48, "learning_rate": 1.1220317495925818e-05, "loss": 0.8228, "step": 9394 }, { "epoch": 0.48, "learning_rate": 1.121868436098842e-05, "loss": 0.9892, "step": 9395 }, { "epoch": 0.48, "learning_rate": 1.1217051193056434e-05, "loss": 0.8543, "step": 9396 }, { "epoch": 0.48, "learning_rate": 1.1215417992174087e-05, "loss": 0.9033, "step": 9397 }, { "epoch": 0.48, "learning_rate": 1.1213784758385586e-05, "loss": 0.8333, "step": 9398 }, { "epoch": 0.48, "learning_rate": 1.1212151491735157e-05, "loss": 0.9706, "step": 9399 }, { "epoch": 0.48, "learning_rate": 1.1210518192267017e-05, "loss": 1.014, "step": 9400 }, { "epoch": 0.48, "learning_rate": 1.1208884860025384e-05, "loss": 0.9329, "step": 9401 }, { "epoch": 0.48, "learning_rate": 1.1207251495054479e-05, "loss": 0.9778, "step": 9402 }, { "epoch": 0.48, "learning_rate": 1.1205618097398525e-05, "loss": 0.9656, "step": 9403 }, { "epoch": 0.48, "learning_rate": 1.1203984667101742e-05, "loss": 0.9254, "step": 9404 }, { "epoch": 0.48, "learning_rate": 1.1202351204208355e-05, "loss": 0.9011, "step": 9405 }, { "epoch": 0.48, "learning_rate": 1.120071770876259e-05, "loss": 0.9444, "step": 9406 }, { "epoch": 0.48, "learning_rate": 1.119908418080867e-05, "loss": 0.9982, "step": 9407 }, { "epoch": 0.48, "learning_rate": 1.1197450620390821e-05, "loss": 0.9462, "step": 9408 }, { "epoch": 0.48, "learning_rate": 1.1195817027553272e-05, "loss": 0.9671, "step": 9409 }, { "epoch": 0.48, "learning_rate": 1.1194183402340246e-05, "loss": 0.8575, "step": 9410 }, { "epoch": 0.48, "learning_rate": 1.1192549744795978e-05, "loss": 0.9155, "step": 9411 }, { "epoch": 0.48, "learning_rate": 1.1190916054964693e-05, "loss": 0.9636, "step": 9412 }, { "epoch": 0.48, "learning_rate": 1.1189282332890625e-05, "loss": 1.0035, "step": 9413 }, { "epoch": 0.48, "learning_rate": 1.1187648578618e-05, "loss": 0.8841, "step": 9414 }, { "epoch": 0.48, "learning_rate": 1.1186014792191056e-05, "loss": 0.8257, "step": 9415 }, { "epoch": 0.48, "learning_rate": 1.1184380973654022e-05, "loss": 0.9427, "step": 9416 }, { "epoch": 0.48, "learning_rate": 1.1182747123051132e-05, "loss": 0.9589, "step": 9417 }, { "epoch": 0.48, "learning_rate": 1.1181113240426624e-05, "loss": 0.8476, "step": 9418 }, { "epoch": 0.48, "learning_rate": 1.1179479325824731e-05, "loss": 0.909, "step": 9419 }, { "epoch": 0.48, "learning_rate": 1.117784537928969e-05, "loss": 0.9658, "step": 9420 }, { "epoch": 0.48, "learning_rate": 1.1176211400865737e-05, "loss": 0.9279, "step": 9421 }, { "epoch": 0.48, "learning_rate": 1.1174577390597114e-05, "loss": 0.9132, "step": 9422 }, { "epoch": 0.48, "learning_rate": 1.1172943348528056e-05, "loss": 0.8911, "step": 9423 }, { "epoch": 0.48, "learning_rate": 1.1171309274702809e-05, "loss": 0.9239, "step": 9424 }, { "epoch": 0.48, "learning_rate": 1.1169675169165604e-05, "loss": 0.9659, "step": 9425 }, { "epoch": 0.48, "learning_rate": 1.116804103196069e-05, "loss": 0.9092, "step": 9426 }, { "epoch": 0.48, "learning_rate": 1.1166406863132306e-05, "loss": 1.028, "step": 9427 }, { "epoch": 0.48, "learning_rate": 1.11647726627247e-05, "loss": 0.9368, "step": 9428 }, { "epoch": 0.48, "learning_rate": 1.1163138430782113e-05, "loss": 0.9044, "step": 9429 }, { "epoch": 0.48, "learning_rate": 1.1161504167348791e-05, "loss": 0.9989, "step": 9430 }, { "epoch": 0.48, "learning_rate": 1.115986987246898e-05, "loss": 0.9112, "step": 9431 }, { "epoch": 0.48, "learning_rate": 1.1158235546186924e-05, "loss": 0.9035, "step": 9432 }, { "epoch": 0.48, "learning_rate": 1.1156601188546871e-05, "loss": 0.8571, "step": 9433 }, { "epoch": 0.48, "learning_rate": 1.1154966799593076e-05, "loss": 0.9167, "step": 9434 }, { "epoch": 0.48, "learning_rate": 1.1153332379369782e-05, "loss": 0.9178, "step": 9435 }, { "epoch": 0.48, "learning_rate": 1.1151697927921242e-05, "loss": 0.9863, "step": 9436 }, { "epoch": 0.48, "learning_rate": 1.1150063445291706e-05, "loss": 0.9735, "step": 9437 }, { "epoch": 0.48, "learning_rate": 1.1148428931525422e-05, "loss": 0.9248, "step": 9438 }, { "epoch": 0.48, "learning_rate": 1.1146794386666649e-05, "loss": 1.0163, "step": 9439 }, { "epoch": 0.48, "learning_rate": 1.1145159810759638e-05, "loss": 0.9608, "step": 9440 }, { "epoch": 0.48, "learning_rate": 1.1143525203848642e-05, "loss": 0.9842, "step": 9441 }, { "epoch": 0.48, "learning_rate": 1.114189056597792e-05, "loss": 0.9492, "step": 9442 }, { "epoch": 0.48, "learning_rate": 1.1140255897191727e-05, "loss": 0.8941, "step": 9443 }, { "epoch": 0.48, "learning_rate": 1.1138621197534314e-05, "loss": 0.9432, "step": 9444 }, { "epoch": 0.48, "learning_rate": 1.1136986467049946e-05, "loss": 0.832, "step": 9445 }, { "epoch": 0.48, "learning_rate": 1.113535170578288e-05, "loss": 0.8345, "step": 9446 }, { "epoch": 0.48, "learning_rate": 1.1133716913777373e-05, "loss": 0.8448, "step": 9447 }, { "epoch": 0.48, "learning_rate": 1.1132082091077687e-05, "loss": 0.9486, "step": 9448 }, { "epoch": 0.48, "learning_rate": 1.1130447237728085e-05, "loss": 0.9359, "step": 9449 }, { "epoch": 0.48, "learning_rate": 1.1128812353772825e-05, "loss": 0.978, "step": 9450 }, { "epoch": 0.48, "learning_rate": 1.112717743925617e-05, "loss": 0.9767, "step": 9451 }, { "epoch": 0.48, "learning_rate": 1.1125542494222388e-05, "loss": 0.898, "step": 9452 }, { "epoch": 0.48, "learning_rate": 1.112390751871574e-05, "loss": 0.8917, "step": 9453 }, { "epoch": 0.48, "learning_rate": 1.112227251278049e-05, "loss": 0.8758, "step": 9454 }, { "epoch": 0.48, "learning_rate": 1.1120637476460908e-05, "loss": 0.906, "step": 9455 }, { "epoch": 0.48, "learning_rate": 1.1119002409801254e-05, "loss": 0.9907, "step": 9456 }, { "epoch": 0.48, "learning_rate": 1.1117367312845801e-05, "loss": 0.8425, "step": 9457 }, { "epoch": 0.48, "learning_rate": 1.111573218563882e-05, "loss": 0.9549, "step": 9458 }, { "epoch": 0.48, "learning_rate": 1.111409702822458e-05, "loss": 0.9207, "step": 9459 }, { "epoch": 0.48, "learning_rate": 1.1112461840647341e-05, "loss": 0.9398, "step": 9460 }, { "epoch": 0.48, "learning_rate": 1.1110826622951386e-05, "loss": 0.7956, "step": 9461 }, { "epoch": 0.48, "learning_rate": 1.1109191375180981e-05, "loss": 0.9447, "step": 9462 }, { "epoch": 0.48, "learning_rate": 1.1107556097380398e-05, "loss": 0.96, "step": 9463 }, { "epoch": 0.48, "learning_rate": 1.1105920789593911e-05, "loss": 0.9075, "step": 9464 }, { "epoch": 0.48, "learning_rate": 1.1104285451865801e-05, "loss": 0.9376, "step": 9465 }, { "epoch": 0.48, "learning_rate": 1.1102650084240335e-05, "loss": 0.9309, "step": 9466 }, { "epoch": 0.48, "learning_rate": 1.1101014686761787e-05, "loss": 0.8981, "step": 9467 }, { "epoch": 0.48, "learning_rate": 1.109937925947444e-05, "loss": 0.912, "step": 9468 }, { "epoch": 0.48, "learning_rate": 1.109774380242257e-05, "loss": 0.9792, "step": 9469 }, { "epoch": 0.48, "learning_rate": 1.1096108315650454e-05, "loss": 0.8837, "step": 9470 }, { "epoch": 0.48, "learning_rate": 1.1094472799202374e-05, "loss": 0.9079, "step": 9471 }, { "epoch": 0.48, "learning_rate": 1.1092837253122604e-05, "loss": 0.9348, "step": 9472 }, { "epoch": 0.48, "learning_rate": 1.1091201677455427e-05, "loss": 0.9497, "step": 9473 }, { "epoch": 0.48, "learning_rate": 1.108956607224513e-05, "loss": 0.9433, "step": 9474 }, { "epoch": 0.48, "learning_rate": 1.1087930437535987e-05, "loss": 0.9455, "step": 9475 }, { "epoch": 0.48, "learning_rate": 1.1086294773372287e-05, "loss": 0.9571, "step": 9476 }, { "epoch": 0.48, "learning_rate": 1.1084659079798315e-05, "loss": 0.9931, "step": 9477 }, { "epoch": 0.48, "learning_rate": 1.1083023356858347e-05, "loss": 0.8899, "step": 9478 }, { "epoch": 0.48, "learning_rate": 1.1081387604596674e-05, "loss": 0.9845, "step": 9479 }, { "epoch": 0.48, "learning_rate": 1.1079751823057583e-05, "loss": 0.9259, "step": 9480 }, { "epoch": 0.48, "learning_rate": 1.1078116012285361e-05, "loss": 0.9281, "step": 9481 }, { "epoch": 0.48, "learning_rate": 1.1076480172324293e-05, "loss": 0.9631, "step": 9482 }, { "epoch": 0.48, "learning_rate": 1.1074844303218674e-05, "loss": 0.8419, "step": 9483 }, { "epoch": 0.48, "learning_rate": 1.1073208405012786e-05, "loss": 0.8534, "step": 9484 }, { "epoch": 0.48, "learning_rate": 1.107157247775092e-05, "loss": 0.8631, "step": 9485 }, { "epoch": 0.48, "learning_rate": 1.1069936521477373e-05, "loss": 0.8496, "step": 9486 }, { "epoch": 0.48, "learning_rate": 1.1068300536236432e-05, "loss": 1.0171, "step": 9487 }, { "epoch": 0.48, "learning_rate": 1.106666452207239e-05, "loss": 0.9573, "step": 9488 }, { "epoch": 0.48, "learning_rate": 1.1065028479029542e-05, "loss": 0.8561, "step": 9489 }, { "epoch": 0.48, "learning_rate": 1.106339240715218e-05, "loss": 1.0061, "step": 9490 }, { "epoch": 0.48, "learning_rate": 1.1061756306484599e-05, "loss": 0.9301, "step": 9491 }, { "epoch": 0.48, "learning_rate": 1.1060120177071098e-05, "loss": 0.9742, "step": 9492 }, { "epoch": 0.48, "learning_rate": 1.1058484018955971e-05, "loss": 0.9663, "step": 9493 }, { "epoch": 0.48, "learning_rate": 1.1056847832183515e-05, "loss": 0.9937, "step": 9494 }, { "epoch": 0.48, "learning_rate": 1.1055211616798027e-05, "loss": 0.9612, "step": 9495 }, { "epoch": 0.48, "learning_rate": 1.1053575372843807e-05, "loss": 0.9163, "step": 9496 }, { "epoch": 0.48, "learning_rate": 1.1051939100365154e-05, "loss": 0.9316, "step": 9497 }, { "epoch": 0.48, "learning_rate": 1.1050302799406369e-05, "loss": 0.8475, "step": 9498 }, { "epoch": 0.48, "learning_rate": 1.1048666470011757e-05, "loss": 0.9041, "step": 9499 }, { "epoch": 0.48, "learning_rate": 1.1047030112225614e-05, "loss": 0.9427, "step": 9500 }, { "epoch": 0.48, "learning_rate": 1.1045393726092243e-05, "loss": 0.8977, "step": 9501 }, { "epoch": 0.48, "learning_rate": 1.1043757311655952e-05, "loss": 0.9189, "step": 9502 }, { "epoch": 0.48, "learning_rate": 1.104212086896104e-05, "loss": 0.9333, "step": 9503 }, { "epoch": 0.48, "learning_rate": 1.1040484398051814e-05, "loss": 0.9864, "step": 9504 }, { "epoch": 0.48, "learning_rate": 1.103884789897258e-05, "loss": 0.9282, "step": 9505 }, { "epoch": 0.48, "learning_rate": 1.1037211371767646e-05, "loss": 0.8318, "step": 9506 }, { "epoch": 0.48, "learning_rate": 1.1035574816481316e-05, "loss": 0.9194, "step": 9507 }, { "epoch": 0.48, "learning_rate": 1.1033938233157901e-05, "loss": 0.9189, "step": 9508 }, { "epoch": 0.48, "learning_rate": 1.1032301621841709e-05, "loss": 0.957, "step": 9509 }, { "epoch": 0.48, "learning_rate": 1.1030664982577046e-05, "loss": 0.9203, "step": 9510 }, { "epoch": 0.48, "learning_rate": 1.1029028315408228e-05, "loss": 1.0096, "step": 9511 }, { "epoch": 0.48, "learning_rate": 1.1027391620379562e-05, "loss": 0.9377, "step": 9512 }, { "epoch": 0.48, "learning_rate": 1.1025754897535358e-05, "loss": 0.9701, "step": 9513 }, { "epoch": 0.48, "learning_rate": 1.1024118146919938e-05, "loss": 0.9997, "step": 9514 }, { "epoch": 0.48, "learning_rate": 1.1022481368577603e-05, "loss": 0.882, "step": 9515 }, { "epoch": 0.48, "learning_rate": 1.1020844562552673e-05, "loss": 1.0491, "step": 9516 }, { "epoch": 0.48, "learning_rate": 1.1019207728889468e-05, "loss": 0.8924, "step": 9517 }, { "epoch": 0.48, "learning_rate": 1.1017570867632294e-05, "loss": 0.8593, "step": 9518 }, { "epoch": 0.48, "learning_rate": 1.1015933978825471e-05, "loss": 0.9579, "step": 9519 }, { "epoch": 0.48, "learning_rate": 1.1014297062513318e-05, "loss": 0.8205, "step": 9520 }, { "epoch": 0.48, "learning_rate": 1.1012660118740148e-05, "loss": 0.8758, "step": 9521 }, { "epoch": 0.48, "learning_rate": 1.1011023147550286e-05, "loss": 0.9121, "step": 9522 }, { "epoch": 0.48, "learning_rate": 1.1009386148988045e-05, "loss": 0.9155, "step": 9523 }, { "epoch": 0.48, "learning_rate": 1.1007749123097754e-05, "loss": 0.9218, "step": 9524 }, { "epoch": 0.48, "learning_rate": 1.1006112069923722e-05, "loss": 0.8604, "step": 9525 }, { "epoch": 0.48, "learning_rate": 1.1004474989510279e-05, "loss": 0.9651, "step": 9526 }, { "epoch": 0.48, "learning_rate": 1.1002837881901742e-05, "loss": 0.8752, "step": 9527 }, { "epoch": 0.48, "learning_rate": 1.1001200747142435e-05, "loss": 0.9602, "step": 9528 }, { "epoch": 0.48, "learning_rate": 1.0999563585276688e-05, "loss": 0.8651, "step": 9529 }, { "epoch": 0.48, "learning_rate": 1.0997926396348816e-05, "loss": 0.9424, "step": 9530 }, { "epoch": 0.48, "learning_rate": 1.099628918040315e-05, "loss": 0.9634, "step": 9531 }, { "epoch": 0.48, "learning_rate": 1.0994651937484014e-05, "loss": 1.0494, "step": 9532 }, { "epoch": 0.48, "learning_rate": 1.0993014667635738e-05, "loss": 0.9143, "step": 9533 }, { "epoch": 0.48, "learning_rate": 1.0991377370902641e-05, "loss": 0.9019, "step": 9534 }, { "epoch": 0.48, "learning_rate": 1.098974004732906e-05, "loss": 0.9081, "step": 9535 }, { "epoch": 0.48, "learning_rate": 1.0988102696959322e-05, "loss": 1.0018, "step": 9536 }, { "epoch": 0.48, "learning_rate": 1.0986465319837752e-05, "loss": 0.8971, "step": 9537 }, { "epoch": 0.48, "learning_rate": 1.098482791600868e-05, "loss": 0.8807, "step": 9538 }, { "epoch": 0.48, "learning_rate": 1.0983190485516446e-05, "loss": 0.8914, "step": 9539 }, { "epoch": 0.48, "learning_rate": 1.0981553028405372e-05, "loss": 0.9151, "step": 9540 }, { "epoch": 0.48, "learning_rate": 1.0979915544719794e-05, "loss": 1.0209, "step": 9541 }, { "epoch": 0.48, "learning_rate": 1.0978278034504049e-05, "loss": 0.9091, "step": 9542 }, { "epoch": 0.48, "learning_rate": 1.0976640497802463e-05, "loss": 0.9405, "step": 9543 }, { "epoch": 0.48, "learning_rate": 1.0975002934659373e-05, "loss": 0.9003, "step": 9544 }, { "epoch": 0.48, "learning_rate": 1.097336534511912e-05, "loss": 0.8982, "step": 9545 }, { "epoch": 0.48, "learning_rate": 1.0971727729226034e-05, "loss": 0.9247, "step": 9546 }, { "epoch": 0.49, "learning_rate": 1.0970090087024454e-05, "loss": 0.8887, "step": 9547 }, { "epoch": 0.49, "learning_rate": 1.0968452418558719e-05, "loss": 0.8831, "step": 9548 }, { "epoch": 0.49, "learning_rate": 1.0966814723873162e-05, "loss": 0.9039, "step": 9549 }, { "epoch": 0.49, "learning_rate": 1.0965177003012125e-05, "loss": 0.9538, "step": 9550 }, { "epoch": 0.49, "learning_rate": 1.0963539256019951e-05, "loss": 1.0048, "step": 9551 }, { "epoch": 0.49, "learning_rate": 1.0961901482940973e-05, "loss": 0.982, "step": 9552 }, { "epoch": 0.49, "learning_rate": 1.0960263683819537e-05, "loss": 0.866, "step": 9553 }, { "epoch": 0.49, "learning_rate": 1.0958625858699985e-05, "loss": 0.8934, "step": 9554 }, { "epoch": 0.49, "learning_rate": 1.0956988007626655e-05, "loss": 0.9026, "step": 9555 }, { "epoch": 0.49, "learning_rate": 1.0955350130643894e-05, "loss": 0.9171, "step": 9556 }, { "epoch": 0.49, "learning_rate": 1.0953712227796044e-05, "loss": 0.9143, "step": 9557 }, { "epoch": 0.49, "learning_rate": 1.0952074299127451e-05, "loss": 0.9073, "step": 9558 }, { "epoch": 0.49, "learning_rate": 1.095043634468246e-05, "loss": 0.9415, "step": 9559 }, { "epoch": 0.49, "learning_rate": 1.0948798364505415e-05, "loss": 0.9607, "step": 9560 }, { "epoch": 0.49, "learning_rate": 1.0947160358640665e-05, "loss": 0.9536, "step": 9561 }, { "epoch": 0.49, "learning_rate": 1.0945522327132555e-05, "loss": 0.9199, "step": 9562 }, { "epoch": 0.49, "learning_rate": 1.0943884270025433e-05, "loss": 0.9064, "step": 9563 }, { "epoch": 0.49, "learning_rate": 1.0942246187363648e-05, "loss": 0.9924, "step": 9564 }, { "epoch": 0.49, "learning_rate": 1.0940608079191553e-05, "loss": 0.8387, "step": 9565 }, { "epoch": 0.49, "learning_rate": 1.0938969945553492e-05, "loss": 0.9601, "step": 9566 }, { "epoch": 0.49, "learning_rate": 1.0937331786493821e-05, "loss": 0.908, "step": 9567 }, { "epoch": 0.49, "learning_rate": 1.0935693602056887e-05, "loss": 1.0068, "step": 9568 }, { "epoch": 0.49, "learning_rate": 1.0934055392287046e-05, "loss": 0.8933, "step": 9569 }, { "epoch": 0.49, "learning_rate": 1.093241715722865e-05, "loss": 0.9189, "step": 9570 }, { "epoch": 0.49, "learning_rate": 1.0930778896926051e-05, "loss": 0.8521, "step": 9571 }, { "epoch": 0.49, "learning_rate": 1.09291406114236e-05, "loss": 0.9431, "step": 9572 }, { "epoch": 0.49, "learning_rate": 1.092750230076566e-05, "loss": 0.983, "step": 9573 }, { "epoch": 0.49, "learning_rate": 1.092586396499658e-05, "loss": 0.9771, "step": 9574 }, { "epoch": 0.49, "learning_rate": 1.0924225604160717e-05, "loss": 1.0096, "step": 9575 }, { "epoch": 0.49, "learning_rate": 1.0922587218302433e-05, "loss": 0.9265, "step": 9576 }, { "epoch": 0.49, "learning_rate": 1.0920948807466079e-05, "loss": 0.9123, "step": 9577 }, { "epoch": 0.49, "learning_rate": 1.0919310371696014e-05, "loss": 0.9618, "step": 9578 }, { "epoch": 0.49, "learning_rate": 1.0917671911036601e-05, "loss": 0.8315, "step": 9579 }, { "epoch": 0.49, "learning_rate": 1.0916033425532196e-05, "loss": 0.9222, "step": 9580 }, { "epoch": 0.49, "learning_rate": 1.091439491522716e-05, "loss": 0.9678, "step": 9581 }, { "epoch": 0.49, "learning_rate": 1.0912756380165859e-05, "loss": 0.8942, "step": 9582 }, { "epoch": 0.49, "learning_rate": 1.0911117820392647e-05, "loss": 0.8098, "step": 9583 }, { "epoch": 0.49, "learning_rate": 1.0909479235951887e-05, "loss": 0.8892, "step": 9584 }, { "epoch": 0.49, "learning_rate": 1.090784062688795e-05, "loss": 0.855, "step": 9585 }, { "epoch": 0.49, "learning_rate": 1.090620199324519e-05, "loss": 0.8173, "step": 9586 }, { "epoch": 0.49, "learning_rate": 1.0904563335067973e-05, "loss": 0.9664, "step": 9587 }, { "epoch": 0.49, "learning_rate": 1.090292465240067e-05, "loss": 0.9477, "step": 9588 }, { "epoch": 0.49, "learning_rate": 1.0901285945287643e-05, "loss": 0.9376, "step": 9589 }, { "epoch": 0.49, "learning_rate": 1.0899647213773257e-05, "loss": 0.9396, "step": 9590 }, { "epoch": 0.49, "learning_rate": 1.089800845790188e-05, "loss": 0.9448, "step": 9591 }, { "epoch": 0.49, "learning_rate": 1.0896369677717878e-05, "loss": 0.907, "step": 9592 }, { "epoch": 0.49, "learning_rate": 1.089473087326562e-05, "loss": 1.0277, "step": 9593 }, { "epoch": 0.49, "learning_rate": 1.0893092044589477e-05, "loss": 0.8742, "step": 9594 }, { "epoch": 0.49, "learning_rate": 1.0891453191733818e-05, "loss": 1.0144, "step": 9595 }, { "epoch": 0.49, "learning_rate": 1.088981431474301e-05, "loss": 0.9084, "step": 9596 }, { "epoch": 0.49, "learning_rate": 1.0888175413661428e-05, "loss": 0.9083, "step": 9597 }, { "epoch": 0.49, "learning_rate": 1.0886536488533442e-05, "loss": 0.9571, "step": 9598 }, { "epoch": 0.49, "learning_rate": 1.088489753940342e-05, "loss": 0.8863, "step": 9599 }, { "epoch": 0.49, "learning_rate": 1.0883258566315742e-05, "loss": 0.9884, "step": 9600 }, { "epoch": 0.49, "learning_rate": 1.0881619569314779e-05, "loss": 0.9935, "step": 9601 }, { "epoch": 0.49, "learning_rate": 1.0879980548444902e-05, "loss": 0.9711, "step": 9602 }, { "epoch": 0.49, "learning_rate": 1.0878341503750486e-05, "loss": 0.7962, "step": 9603 }, { "epoch": 0.49, "learning_rate": 1.0876702435275913e-05, "loss": 0.9287, "step": 9604 }, { "epoch": 0.49, "learning_rate": 1.0875063343065551e-05, "loss": 0.9966, "step": 9605 }, { "epoch": 0.49, "learning_rate": 1.0873424227163783e-05, "loss": 0.8988, "step": 9606 }, { "epoch": 0.49, "learning_rate": 1.0871785087614981e-05, "loss": 1.0756, "step": 9607 }, { "epoch": 0.49, "learning_rate": 1.0870145924463525e-05, "loss": 0.9207, "step": 9608 }, { "epoch": 0.49, "learning_rate": 1.0868506737753792e-05, "loss": 0.9039, "step": 9609 }, { "epoch": 0.49, "learning_rate": 1.0866867527530167e-05, "loss": 0.8434, "step": 9610 }, { "epoch": 0.49, "learning_rate": 1.0865228293837024e-05, "loss": 0.8341, "step": 9611 }, { "epoch": 0.49, "learning_rate": 1.0863589036718747e-05, "loss": 0.9269, "step": 9612 }, { "epoch": 0.49, "learning_rate": 1.0861949756219717e-05, "loss": 1.0633, "step": 9613 }, { "epoch": 0.49, "learning_rate": 1.0860310452384308e-05, "loss": 0.8582, "step": 9614 }, { "epoch": 0.49, "learning_rate": 1.0858671125256912e-05, "loss": 0.8886, "step": 9615 }, { "epoch": 0.49, "learning_rate": 1.0857031774881913e-05, "loss": 0.9285, "step": 9616 }, { "epoch": 0.49, "learning_rate": 1.0855392401303684e-05, "loss": 0.9489, "step": 9617 }, { "epoch": 0.49, "learning_rate": 1.0853753004566621e-05, "loss": 0.9835, "step": 9618 }, { "epoch": 0.49, "learning_rate": 1.0852113584715103e-05, "loss": 0.9011, "step": 9619 }, { "epoch": 0.49, "learning_rate": 1.0850474141793513e-05, "loss": 0.9401, "step": 9620 }, { "epoch": 0.49, "learning_rate": 1.084883467584624e-05, "loss": 0.8697, "step": 9621 }, { "epoch": 0.49, "learning_rate": 1.0847195186917676e-05, "loss": 0.939, "step": 9622 }, { "epoch": 0.49, "learning_rate": 1.08455556750522e-05, "loss": 0.9447, "step": 9623 }, { "epoch": 0.49, "learning_rate": 1.0843916140294206e-05, "loss": 0.8846, "step": 9624 }, { "epoch": 0.49, "learning_rate": 1.084227658268808e-05, "loss": 0.9684, "step": 9625 }, { "epoch": 0.49, "learning_rate": 1.0840637002278209e-05, "loss": 0.9438, "step": 9626 }, { "epoch": 0.49, "learning_rate": 1.0838997399108986e-05, "loss": 0.8818, "step": 9627 }, { "epoch": 0.49, "learning_rate": 1.0837357773224803e-05, "loss": 0.9199, "step": 9628 }, { "epoch": 0.49, "learning_rate": 1.083571812467005e-05, "loss": 0.8979, "step": 9629 }, { "epoch": 0.49, "learning_rate": 1.0834078453489116e-05, "loss": 0.995, "step": 9630 }, { "epoch": 0.49, "learning_rate": 1.0832438759726399e-05, "loss": 0.9872, "step": 9631 }, { "epoch": 0.49, "learning_rate": 1.0830799043426283e-05, "loss": 1.0192, "step": 9632 }, { "epoch": 0.49, "learning_rate": 1.0829159304633167e-05, "loss": 0.8353, "step": 9633 }, { "epoch": 0.49, "learning_rate": 1.0827519543391447e-05, "loss": 1.0144, "step": 9634 }, { "epoch": 0.49, "learning_rate": 1.0825879759745516e-05, "loss": 0.9474, "step": 9635 }, { "epoch": 0.49, "learning_rate": 1.082423995373977e-05, "loss": 0.9417, "step": 9636 }, { "epoch": 0.49, "learning_rate": 1.08226001254186e-05, "loss": 0.989, "step": 9637 }, { "epoch": 0.49, "learning_rate": 1.0820960274826414e-05, "loss": 0.9771, "step": 9638 }, { "epoch": 0.49, "learning_rate": 1.0819320402007595e-05, "loss": 0.9654, "step": 9639 }, { "epoch": 0.49, "learning_rate": 1.081768050700655e-05, "loss": 0.9713, "step": 9640 }, { "epoch": 0.49, "learning_rate": 1.0816040589867675e-05, "loss": 0.8613, "step": 9641 }, { "epoch": 0.49, "learning_rate": 1.0814400650635373e-05, "loss": 0.8702, "step": 9642 }, { "epoch": 0.49, "learning_rate": 1.0812760689354034e-05, "loss": 0.9189, "step": 9643 }, { "epoch": 0.49, "learning_rate": 1.0811120706068068e-05, "loss": 0.9052, "step": 9644 }, { "epoch": 0.49, "learning_rate": 1.0809480700821871e-05, "loss": 0.9283, "step": 9645 }, { "epoch": 0.49, "learning_rate": 1.0807840673659844e-05, "loss": 0.8809, "step": 9646 }, { "epoch": 0.49, "learning_rate": 1.080620062462639e-05, "loss": 0.9222, "step": 9647 }, { "epoch": 0.49, "learning_rate": 1.0804560553765915e-05, "loss": 0.8423, "step": 9648 }, { "epoch": 0.49, "learning_rate": 1.0802920461122816e-05, "loss": 0.9298, "step": 9649 }, { "epoch": 0.49, "learning_rate": 1.0801280346741504e-05, "loss": 0.9557, "step": 9650 }, { "epoch": 0.49, "learning_rate": 1.0799640210666373e-05, "loss": 0.9693, "step": 9651 }, { "epoch": 0.49, "learning_rate": 1.0798000052941835e-05, "loss": 0.97, "step": 9652 }, { "epoch": 0.49, "learning_rate": 1.0796359873612299e-05, "loss": 0.9358, "step": 9653 }, { "epoch": 0.49, "learning_rate": 1.0794719672722165e-05, "loss": 0.8753, "step": 9654 }, { "epoch": 0.49, "learning_rate": 1.0793079450315838e-05, "loss": 0.8681, "step": 9655 }, { "epoch": 0.49, "learning_rate": 1.0791439206437733e-05, "loss": 0.9176, "step": 9656 }, { "epoch": 0.49, "learning_rate": 1.078979894113225e-05, "loss": 0.9059, "step": 9657 }, { "epoch": 0.49, "learning_rate": 1.07881586544438e-05, "loss": 0.8988, "step": 9658 }, { "epoch": 0.49, "learning_rate": 1.0786518346416799e-05, "loss": 0.8293, "step": 9659 }, { "epoch": 0.49, "learning_rate": 1.0784878017095648e-05, "loss": 0.9711, "step": 9660 }, { "epoch": 0.49, "learning_rate": 1.0783237666524758e-05, "loss": 0.8988, "step": 9661 }, { "epoch": 0.49, "learning_rate": 1.0781597294748543e-05, "loss": 0.9589, "step": 9662 }, { "epoch": 0.49, "learning_rate": 1.077995690181141e-05, "loss": 0.7501, "step": 9663 }, { "epoch": 0.49, "learning_rate": 1.0778316487757777e-05, "loss": 0.9389, "step": 9664 }, { "epoch": 0.49, "learning_rate": 1.0776676052632051e-05, "loss": 0.8746, "step": 9665 }, { "epoch": 0.49, "learning_rate": 1.0775035596478652e-05, "loss": 0.8928, "step": 9666 }, { "epoch": 0.49, "learning_rate": 1.0773395119341983e-05, "loss": 1.0291, "step": 9667 }, { "epoch": 0.49, "learning_rate": 1.0771754621266466e-05, "loss": 0.8976, "step": 9668 }, { "epoch": 0.49, "learning_rate": 1.0770114102296513e-05, "loss": 0.9879, "step": 9669 }, { "epoch": 0.49, "learning_rate": 1.076847356247654e-05, "loss": 0.918, "step": 9670 }, { "epoch": 0.49, "learning_rate": 1.0766833001850965e-05, "loss": 0.9145, "step": 9671 }, { "epoch": 0.49, "learning_rate": 1.0765192420464202e-05, "loss": 0.9042, "step": 9672 }, { "epoch": 0.49, "learning_rate": 1.0763551818360664e-05, "loss": 0.8853, "step": 9673 }, { "epoch": 0.49, "learning_rate": 1.0761911195584776e-05, "loss": 0.8917, "step": 9674 }, { "epoch": 0.49, "learning_rate": 1.0760270552180953e-05, "loss": 0.8746, "step": 9675 }, { "epoch": 0.49, "learning_rate": 1.0758629888193615e-05, "loss": 0.9507, "step": 9676 }, { "epoch": 0.49, "learning_rate": 1.0756989203667179e-05, "loss": 0.9321, "step": 9677 }, { "epoch": 0.49, "learning_rate": 1.0755348498646065e-05, "loss": 0.9461, "step": 9678 }, { "epoch": 0.49, "learning_rate": 1.0753707773174696e-05, "loss": 1.0414, "step": 9679 }, { "epoch": 0.49, "learning_rate": 1.0752067027297486e-05, "loss": 0.9524, "step": 9680 }, { "epoch": 0.49, "learning_rate": 1.0750426261058865e-05, "loss": 0.9027, "step": 9681 }, { "epoch": 0.49, "learning_rate": 1.0748785474503251e-05, "loss": 0.9237, "step": 9682 }, { "epoch": 0.49, "learning_rate": 1.0747144667675069e-05, "loss": 0.9796, "step": 9683 }, { "epoch": 0.49, "learning_rate": 1.0745503840618738e-05, "loss": 0.8839, "step": 9684 }, { "epoch": 0.49, "learning_rate": 1.0743862993378682e-05, "loss": 0.9404, "step": 9685 }, { "epoch": 0.49, "learning_rate": 1.074222212599933e-05, "loss": 0.8245, "step": 9686 }, { "epoch": 0.49, "learning_rate": 1.0740581238525101e-05, "loss": 0.9081, "step": 9687 }, { "epoch": 0.49, "learning_rate": 1.0738940331000422e-05, "loss": 1.0161, "step": 9688 }, { "epoch": 0.49, "learning_rate": 1.0737299403469723e-05, "loss": 0.9783, "step": 9689 }, { "epoch": 0.49, "learning_rate": 1.0735658455977429e-05, "loss": 0.8842, "step": 9690 }, { "epoch": 0.49, "learning_rate": 1.0734017488567961e-05, "loss": 0.8574, "step": 9691 }, { "epoch": 0.49, "learning_rate": 1.0732376501285748e-05, "loss": 0.948, "step": 9692 }, { "epoch": 0.49, "learning_rate": 1.0730735494175226e-05, "loss": 1.079, "step": 9693 }, { "epoch": 0.49, "learning_rate": 1.0729094467280813e-05, "loss": 0.8855, "step": 9694 }, { "epoch": 0.49, "learning_rate": 1.0727453420646948e-05, "loss": 0.9288, "step": 9695 }, { "epoch": 0.49, "learning_rate": 1.0725812354318052e-05, "loss": 0.9258, "step": 9696 }, { "epoch": 0.49, "learning_rate": 1.0724171268338561e-05, "loss": 0.9078, "step": 9697 }, { "epoch": 0.49, "learning_rate": 1.07225301627529e-05, "loss": 0.939, "step": 9698 }, { "epoch": 0.49, "learning_rate": 1.0720889037605506e-05, "loss": 0.9778, "step": 9699 }, { "epoch": 0.49, "learning_rate": 1.0719247892940807e-05, "loss": 0.8376, "step": 9700 }, { "epoch": 0.49, "learning_rate": 1.071760672880324e-05, "loss": 0.8568, "step": 9701 }, { "epoch": 0.49, "learning_rate": 1.0715965545237232e-05, "loss": 1.0131, "step": 9702 }, { "epoch": 0.49, "learning_rate": 1.071432434228722e-05, "loss": 0.8893, "step": 9703 }, { "epoch": 0.49, "learning_rate": 1.0712683119997631e-05, "loss": 0.9445, "step": 9704 }, { "epoch": 0.49, "learning_rate": 1.071104187841291e-05, "loss": 0.8727, "step": 9705 }, { "epoch": 0.49, "learning_rate": 1.0709400617577486e-05, "loss": 0.8777, "step": 9706 }, { "epoch": 0.49, "learning_rate": 1.0707759337535794e-05, "loss": 0.8993, "step": 9707 }, { "epoch": 0.49, "learning_rate": 1.070611803833227e-05, "loss": 0.9016, "step": 9708 }, { "epoch": 0.49, "learning_rate": 1.0704476720011354e-05, "loss": 0.8937, "step": 9709 }, { "epoch": 0.49, "learning_rate": 1.0702835382617479e-05, "loss": 0.785, "step": 9710 }, { "epoch": 0.49, "learning_rate": 1.070119402619508e-05, "loss": 0.9463, "step": 9711 }, { "epoch": 0.49, "learning_rate": 1.0699552650788605e-05, "loss": 0.9245, "step": 9712 }, { "epoch": 0.49, "learning_rate": 1.0697911256442484e-05, "loss": 0.9593, "step": 9713 }, { "epoch": 0.49, "learning_rate": 1.0696269843201156e-05, "loss": 0.905, "step": 9714 }, { "epoch": 0.49, "learning_rate": 1.0694628411109066e-05, "loss": 0.8341, "step": 9715 }, { "epoch": 0.49, "learning_rate": 1.069298696021065e-05, "loss": 0.8743, "step": 9716 }, { "epoch": 0.49, "learning_rate": 1.0691345490550346e-05, "loss": 0.8462, "step": 9717 }, { "epoch": 0.49, "learning_rate": 1.0689704002172604e-05, "loss": 0.9647, "step": 9718 }, { "epoch": 0.49, "learning_rate": 1.0688062495121855e-05, "loss": 0.9261, "step": 9719 }, { "epoch": 0.49, "learning_rate": 1.0686420969442548e-05, "loss": 0.9359, "step": 9720 }, { "epoch": 0.49, "learning_rate": 1.0684779425179125e-05, "loss": 0.916, "step": 9721 }, { "epoch": 0.49, "learning_rate": 1.0683137862376025e-05, "loss": 0.8922, "step": 9722 }, { "epoch": 0.49, "learning_rate": 1.0681496281077694e-05, "loss": 0.9089, "step": 9723 }, { "epoch": 0.49, "learning_rate": 1.0679854681328581e-05, "loss": 0.9071, "step": 9724 }, { "epoch": 0.49, "learning_rate": 1.0678213063173121e-05, "loss": 0.8982, "step": 9725 }, { "epoch": 0.49, "learning_rate": 1.0676571426655767e-05, "loss": 0.9844, "step": 9726 }, { "epoch": 0.49, "learning_rate": 1.067492977182096e-05, "loss": 0.8818, "step": 9727 }, { "epoch": 0.49, "learning_rate": 1.0673288098713146e-05, "loss": 0.8188, "step": 9728 }, { "epoch": 0.49, "learning_rate": 1.0671646407376773e-05, "loss": 0.9755, "step": 9729 }, { "epoch": 0.49, "learning_rate": 1.0670004697856292e-05, "loss": 0.8416, "step": 9730 }, { "epoch": 0.49, "learning_rate": 1.0668362970196143e-05, "loss": 0.8498, "step": 9731 }, { "epoch": 0.49, "learning_rate": 1.0666721224440776e-05, "loss": 0.9474, "step": 9732 }, { "epoch": 0.49, "learning_rate": 1.0665079460634646e-05, "loss": 0.9706, "step": 9733 }, { "epoch": 0.49, "learning_rate": 1.0663437678822195e-05, "loss": 0.9733, "step": 9734 }, { "epoch": 0.49, "learning_rate": 1.0661795879047874e-05, "loss": 0.9523, "step": 9735 }, { "epoch": 0.49, "learning_rate": 1.0660154061356137e-05, "loss": 0.889, "step": 9736 }, { "epoch": 0.49, "learning_rate": 1.0658512225791429e-05, "loss": 0.889, "step": 9737 }, { "epoch": 0.49, "learning_rate": 1.06568703723982e-05, "loss": 0.9286, "step": 9738 }, { "epoch": 0.49, "learning_rate": 1.0655228501220907e-05, "loss": 0.9152, "step": 9739 }, { "epoch": 0.49, "learning_rate": 1.0653586612304001e-05, "loss": 0.8083, "step": 9740 }, { "epoch": 0.49, "learning_rate": 1.065194470569193e-05, "loss": 0.9209, "step": 9741 }, { "epoch": 0.49, "learning_rate": 1.0650302781429152e-05, "loss": 0.9895, "step": 9742 }, { "epoch": 0.49, "learning_rate": 1.0648660839560118e-05, "loss": 1.0098, "step": 9743 }, { "epoch": 0.5, "learning_rate": 1.064701888012928e-05, "loss": 0.8945, "step": 9744 }, { "epoch": 0.5, "learning_rate": 1.0645376903181095e-05, "loss": 0.9346, "step": 9745 }, { "epoch": 0.5, "learning_rate": 1.0643734908760017e-05, "loss": 0.9218, "step": 9746 }, { "epoch": 0.5, "learning_rate": 1.06420928969105e-05, "loss": 0.8943, "step": 9747 }, { "epoch": 0.5, "learning_rate": 1.0640450867677003e-05, "loss": 0.8498, "step": 9748 }, { "epoch": 0.5, "learning_rate": 1.0638808821103984e-05, "loss": 0.8468, "step": 9749 }, { "epoch": 0.5, "learning_rate": 1.0637166757235889e-05, "loss": 0.9872, "step": 9750 }, { "epoch": 0.5, "learning_rate": 1.0635524676117183e-05, "loss": 0.8974, "step": 9751 }, { "epoch": 0.5, "learning_rate": 1.0633882577792325e-05, "loss": 0.9462, "step": 9752 }, { "epoch": 0.5, "learning_rate": 1.0632240462305767e-05, "loss": 0.9628, "step": 9753 }, { "epoch": 0.5, "learning_rate": 1.0630598329701975e-05, "loss": 0.8606, "step": 9754 }, { "epoch": 0.5, "learning_rate": 1.0628956180025405e-05, "loss": 0.9054, "step": 9755 }, { "epoch": 0.5, "learning_rate": 1.062731401332051e-05, "loss": 0.993, "step": 9756 }, { "epoch": 0.5, "learning_rate": 1.0625671829631761e-05, "loss": 1.0245, "step": 9757 }, { "epoch": 0.5, "learning_rate": 1.0624029629003612e-05, "loss": 0.8961, "step": 9758 }, { "epoch": 0.5, "learning_rate": 1.0622387411480521e-05, "loss": 0.995, "step": 9759 }, { "epoch": 0.5, "learning_rate": 1.0620745177106957e-05, "loss": 1.0056, "step": 9760 }, { "epoch": 0.5, "learning_rate": 1.0619102925927377e-05, "loss": 0.8853, "step": 9761 }, { "epoch": 0.5, "learning_rate": 1.0617460657986242e-05, "loss": 0.9422, "step": 9762 }, { "epoch": 0.5, "learning_rate": 1.0615818373328017e-05, "loss": 0.9332, "step": 9763 }, { "epoch": 0.5, "learning_rate": 1.0614176071997167e-05, "loss": 0.8743, "step": 9764 }, { "epoch": 0.5, "learning_rate": 1.061253375403815e-05, "loss": 0.9078, "step": 9765 }, { "epoch": 0.5, "learning_rate": 1.0610891419495434e-05, "loss": 0.9183, "step": 9766 }, { "epoch": 0.5, "learning_rate": 1.0609249068413488e-05, "loss": 0.9123, "step": 9767 }, { "epoch": 0.5, "learning_rate": 1.0607606700836766e-05, "loss": 0.967, "step": 9768 }, { "epoch": 0.5, "learning_rate": 1.0605964316809738e-05, "loss": 0.9146, "step": 9769 }, { "epoch": 0.5, "learning_rate": 1.0604321916376873e-05, "loss": 0.8261, "step": 9770 }, { "epoch": 0.5, "learning_rate": 1.0602679499582636e-05, "loss": 0.994, "step": 9771 }, { "epoch": 0.5, "learning_rate": 1.0601037066471493e-05, "loss": 0.8993, "step": 9772 }, { "epoch": 0.5, "learning_rate": 1.0599394617087908e-05, "loss": 0.8798, "step": 9773 }, { "epoch": 0.5, "learning_rate": 1.0597752151476355e-05, "loss": 0.9173, "step": 9774 }, { "epoch": 0.5, "learning_rate": 1.0596109669681295e-05, "loss": 0.8371, "step": 9775 }, { "epoch": 0.5, "learning_rate": 1.0594467171747202e-05, "loss": 0.9151, "step": 9776 }, { "epoch": 0.5, "learning_rate": 1.0592824657718543e-05, "loss": 0.8191, "step": 9777 }, { "epoch": 0.5, "learning_rate": 1.0591182127639788e-05, "loss": 0.9011, "step": 9778 }, { "epoch": 0.5, "learning_rate": 1.0589539581555404e-05, "loss": 0.9438, "step": 9779 }, { "epoch": 0.5, "learning_rate": 1.0587897019509865e-05, "loss": 1.0061, "step": 9780 }, { "epoch": 0.5, "learning_rate": 1.0586254441547636e-05, "loss": 0.888, "step": 9781 }, { "epoch": 0.5, "learning_rate": 1.0584611847713193e-05, "loss": 0.9483, "step": 9782 }, { "epoch": 0.5, "learning_rate": 1.0582969238051009e-05, "loss": 0.8848, "step": 9783 }, { "epoch": 0.5, "learning_rate": 1.0581326612605552e-05, "loss": 0.9138, "step": 9784 }, { "epoch": 0.5, "learning_rate": 1.0579683971421292e-05, "loss": 0.9408, "step": 9785 }, { "epoch": 0.5, "learning_rate": 1.057804131454271e-05, "loss": 0.8881, "step": 9786 }, { "epoch": 0.5, "learning_rate": 1.0576398642014273e-05, "loss": 0.8361, "step": 9787 }, { "epoch": 0.5, "learning_rate": 1.0574755953880454e-05, "loss": 0.8611, "step": 9788 }, { "epoch": 0.5, "learning_rate": 1.057311325018573e-05, "loss": 0.9019, "step": 9789 }, { "epoch": 0.5, "learning_rate": 1.0571470530974577e-05, "loss": 0.8766, "step": 9790 }, { "epoch": 0.5, "learning_rate": 1.0569827796291463e-05, "loss": 0.9335, "step": 9791 }, { "epoch": 0.5, "learning_rate": 1.0568185046180872e-05, "loss": 0.9543, "step": 9792 }, { "epoch": 0.5, "learning_rate": 1.0566542280687274e-05, "loss": 0.92, "step": 9793 }, { "epoch": 0.5, "learning_rate": 1.0564899499855144e-05, "loss": 1.0142, "step": 9794 }, { "epoch": 0.5, "learning_rate": 1.0563256703728963e-05, "loss": 1.0919, "step": 9795 }, { "epoch": 0.5, "learning_rate": 1.0561613892353206e-05, "loss": 1.0265, "step": 9796 }, { "epoch": 0.5, "learning_rate": 1.055997106577235e-05, "loss": 0.9343, "step": 9797 }, { "epoch": 0.5, "learning_rate": 1.0558328224030874e-05, "loss": 0.9298, "step": 9798 }, { "epoch": 0.5, "learning_rate": 1.0556685367173254e-05, "loss": 0.8639, "step": 9799 }, { "epoch": 0.5, "learning_rate": 1.055504249524397e-05, "loss": 0.9241, "step": 9800 }, { "epoch": 0.5, "learning_rate": 1.0553399608287502e-05, "loss": 0.9275, "step": 9801 }, { "epoch": 0.5, "learning_rate": 1.0551756706348331e-05, "loss": 0.8833, "step": 9802 }, { "epoch": 0.5, "learning_rate": 1.0550113789470927e-05, "loss": 0.8947, "step": 9803 }, { "epoch": 0.5, "learning_rate": 1.0548470857699784e-05, "loss": 0.9607, "step": 9804 }, { "epoch": 0.5, "learning_rate": 1.0546827911079374e-05, "loss": 0.9586, "step": 9805 }, { "epoch": 0.5, "learning_rate": 1.0545184949654177e-05, "loss": 0.8914, "step": 9806 }, { "epoch": 0.5, "learning_rate": 1.0543541973468683e-05, "loss": 0.8969, "step": 9807 }, { "epoch": 0.5, "learning_rate": 1.0541898982567366e-05, "loss": 0.8536, "step": 9808 }, { "epoch": 0.5, "learning_rate": 1.0540255976994708e-05, "loss": 0.9208, "step": 9809 }, { "epoch": 0.5, "learning_rate": 1.0538612956795196e-05, "loss": 0.9346, "step": 9810 }, { "epoch": 0.5, "learning_rate": 1.0536969922013315e-05, "loss": 0.9582, "step": 9811 }, { "epoch": 0.5, "learning_rate": 1.053532687269354e-05, "loss": 0.9653, "step": 9812 }, { "epoch": 0.5, "learning_rate": 1.053368380888036e-05, "loss": 1.0443, "step": 9813 }, { "epoch": 0.5, "learning_rate": 1.053204073061826e-05, "loss": 0.907, "step": 9814 }, { "epoch": 0.5, "learning_rate": 1.0530397637951724e-05, "loss": 0.9387, "step": 9815 }, { "epoch": 0.5, "learning_rate": 1.0528754530925234e-05, "loss": 0.8958, "step": 9816 }, { "epoch": 0.5, "learning_rate": 1.052711140958328e-05, "loss": 0.6929, "step": 9817 }, { "epoch": 0.5, "learning_rate": 1.052546827397034e-05, "loss": 0.9916, "step": 9818 }, { "epoch": 0.5, "learning_rate": 1.052382512413091e-05, "loss": 0.9182, "step": 9819 }, { "epoch": 0.5, "learning_rate": 1.0522181960109474e-05, "loss": 0.8748, "step": 9820 }, { "epoch": 0.5, "learning_rate": 1.0520538781950513e-05, "loss": 0.8687, "step": 9821 }, { "epoch": 0.5, "learning_rate": 1.0518895589698519e-05, "loss": 0.9821, "step": 9822 }, { "epoch": 0.5, "learning_rate": 1.0517252383397982e-05, "loss": 0.8746, "step": 9823 }, { "epoch": 0.5, "learning_rate": 1.0515609163093383e-05, "loss": 0.9875, "step": 9824 }, { "epoch": 0.5, "learning_rate": 1.0513965928829217e-05, "loss": 0.9014, "step": 9825 }, { "epoch": 0.5, "learning_rate": 1.0512322680649972e-05, "loss": 0.9882, "step": 9826 }, { "epoch": 0.5, "learning_rate": 1.0510679418600132e-05, "loss": 0.9765, "step": 9827 }, { "epoch": 0.5, "learning_rate": 1.050903614272419e-05, "loss": 0.9197, "step": 9828 }, { "epoch": 0.5, "learning_rate": 1.0507392853066638e-05, "loss": 0.8195, "step": 9829 }, { "epoch": 0.5, "learning_rate": 1.0505749549671962e-05, "loss": 0.8795, "step": 9830 }, { "epoch": 0.5, "learning_rate": 1.0504106232584659e-05, "loss": 0.9475, "step": 9831 }, { "epoch": 0.5, "learning_rate": 1.0502462901849216e-05, "loss": 1.0029, "step": 9832 }, { "epoch": 0.5, "learning_rate": 1.050081955751012e-05, "loss": 0.939, "step": 9833 }, { "epoch": 0.5, "learning_rate": 1.0499176199611866e-05, "loss": 0.9009, "step": 9834 }, { "epoch": 0.5, "learning_rate": 1.0497532828198953e-05, "loss": 0.9411, "step": 9835 }, { "epoch": 0.5, "learning_rate": 1.0495889443315866e-05, "loss": 0.9993, "step": 9836 }, { "epoch": 0.5, "learning_rate": 1.0494246045007102e-05, "loss": 0.918, "step": 9837 }, { "epoch": 0.5, "learning_rate": 1.049260263331715e-05, "loss": 0.9523, "step": 9838 }, { "epoch": 0.5, "learning_rate": 1.0490959208290504e-05, "loss": 0.9597, "step": 9839 }, { "epoch": 0.5, "learning_rate": 1.0489315769971659e-05, "loss": 0.9045, "step": 9840 }, { "epoch": 0.5, "learning_rate": 1.0487672318405116e-05, "loss": 0.9979, "step": 9841 }, { "epoch": 0.5, "learning_rate": 1.048602885363536e-05, "loss": 0.8804, "step": 9842 }, { "epoch": 0.5, "learning_rate": 1.0484385375706891e-05, "loss": 1.0563, "step": 9843 }, { "epoch": 0.5, "learning_rate": 1.0482741884664201e-05, "loss": 0.8896, "step": 9844 }, { "epoch": 0.5, "learning_rate": 1.0481098380551791e-05, "loss": 0.9298, "step": 9845 }, { "epoch": 0.5, "learning_rate": 1.047945486341415e-05, "loss": 0.8835, "step": 9846 }, { "epoch": 0.5, "learning_rate": 1.0477811333295782e-05, "loss": 1.0171, "step": 9847 }, { "epoch": 0.5, "learning_rate": 1.047616779024118e-05, "loss": 0.9378, "step": 9848 }, { "epoch": 0.5, "learning_rate": 1.0474524234294844e-05, "loss": 0.8787, "step": 9849 }, { "epoch": 0.5, "learning_rate": 1.0472880665501265e-05, "loss": 0.8956, "step": 9850 }, { "epoch": 0.5, "learning_rate": 1.0471237083904951e-05, "loss": 0.9212, "step": 9851 }, { "epoch": 0.5, "learning_rate": 1.046959348955039e-05, "loss": 0.9825, "step": 9852 }, { "epoch": 0.5, "learning_rate": 1.0467949882482082e-05, "loss": 0.9036, "step": 9853 }, { "epoch": 0.5, "learning_rate": 1.0466306262744534e-05, "loss": 0.9337, "step": 9854 }, { "epoch": 0.5, "learning_rate": 1.0464662630382239e-05, "loss": 0.9268, "step": 9855 }, { "epoch": 0.5, "learning_rate": 1.0463018985439695e-05, "loss": 0.8912, "step": 9856 }, { "epoch": 0.5, "learning_rate": 1.0461375327961407e-05, "loss": 0.9818, "step": 9857 }, { "epoch": 0.5, "learning_rate": 1.0459731657991872e-05, "loss": 0.9738, "step": 9858 }, { "epoch": 0.5, "learning_rate": 1.0458087975575591e-05, "loss": 0.9641, "step": 9859 }, { "epoch": 0.5, "learning_rate": 1.0456444280757067e-05, "loss": 0.9781, "step": 9860 }, { "epoch": 0.5, "learning_rate": 1.0454800573580798e-05, "loss": 0.9426, "step": 9861 }, { "epoch": 0.5, "learning_rate": 1.0453156854091288e-05, "loss": 0.9138, "step": 9862 }, { "epoch": 0.5, "learning_rate": 1.0451513122333042e-05, "loss": 0.859, "step": 9863 }, { "epoch": 0.5, "learning_rate": 1.0449869378350552e-05, "loss": 0.9171, "step": 9864 }, { "epoch": 0.5, "learning_rate": 1.044822562218833e-05, "loss": 0.9295, "step": 9865 }, { "epoch": 0.5, "learning_rate": 1.0446581853890877e-05, "loss": 0.965, "step": 9866 }, { "epoch": 0.5, "learning_rate": 1.0444938073502698e-05, "loss": 0.9759, "step": 9867 }, { "epoch": 0.5, "learning_rate": 1.0443294281068287e-05, "loss": 0.9528, "step": 9868 }, { "epoch": 0.5, "learning_rate": 1.0441650476632162e-05, "loss": 0.9668, "step": 9869 }, { "epoch": 0.5, "learning_rate": 1.0440006660238816e-05, "loss": 0.8779, "step": 9870 }, { "epoch": 0.5, "learning_rate": 1.0438362831932758e-05, "loss": 0.9364, "step": 9871 }, { "epoch": 0.5, "learning_rate": 1.0436718991758495e-05, "loss": 0.9438, "step": 9872 }, { "epoch": 0.5, "learning_rate": 1.0435075139760529e-05, "loss": 0.9564, "step": 9873 }, { "epoch": 0.5, "learning_rate": 1.0433431275983364e-05, "loss": 0.897, "step": 9874 }, { "epoch": 0.5, "learning_rate": 1.0431787400471511e-05, "loss": 0.8506, "step": 9875 }, { "epoch": 0.5, "learning_rate": 1.0430143513269473e-05, "loss": 0.9527, "step": 9876 }, { "epoch": 0.5, "learning_rate": 1.0428499614421754e-05, "loss": 0.9203, "step": 9877 }, { "epoch": 0.5, "learning_rate": 1.0426855703972869e-05, "loss": 0.9507, "step": 9878 }, { "epoch": 0.5, "learning_rate": 1.042521178196732e-05, "loss": 0.8726, "step": 9879 }, { "epoch": 0.5, "learning_rate": 1.042356784844961e-05, "loss": 0.9402, "step": 9880 }, { "epoch": 0.5, "learning_rate": 1.0421923903464253e-05, "loss": 0.8826, "step": 9881 }, { "epoch": 0.5, "learning_rate": 1.0420279947055756e-05, "loss": 0.8954, "step": 9882 }, { "epoch": 0.5, "learning_rate": 1.0418635979268625e-05, "loss": 1.0019, "step": 9883 }, { "epoch": 0.5, "learning_rate": 1.0416992000147372e-05, "loss": 0.9229, "step": 9884 }, { "epoch": 0.5, "learning_rate": 1.0415348009736503e-05, "loss": 0.9526, "step": 9885 }, { "epoch": 0.5, "learning_rate": 1.041370400808053e-05, "loss": 0.8146, "step": 9886 }, { "epoch": 0.5, "learning_rate": 1.0412059995223958e-05, "loss": 0.9807, "step": 9887 }, { "epoch": 0.5, "learning_rate": 1.0410415971211303e-05, "loss": 0.8983, "step": 9888 }, { "epoch": 0.5, "learning_rate": 1.0408771936087069e-05, "loss": 1.0013, "step": 9889 }, { "epoch": 0.5, "learning_rate": 1.0407127889895777e-05, "loss": 0.9741, "step": 9890 }, { "epoch": 0.5, "learning_rate": 1.0405483832681925e-05, "loss": 0.9811, "step": 9891 }, { "epoch": 0.5, "learning_rate": 1.0403839764490029e-05, "loss": 0.8809, "step": 9892 }, { "epoch": 0.5, "learning_rate": 1.0402195685364602e-05, "loss": 0.9268, "step": 9893 }, { "epoch": 0.5, "learning_rate": 1.0400551595350154e-05, "loss": 0.8696, "step": 9894 }, { "epoch": 0.5, "learning_rate": 1.03989074944912e-05, "loss": 0.8521, "step": 9895 }, { "epoch": 0.5, "learning_rate": 1.039726338283225e-05, "loss": 0.8596, "step": 9896 }, { "epoch": 0.5, "learning_rate": 1.0395619260417816e-05, "loss": 0.8906, "step": 9897 }, { "epoch": 0.5, "learning_rate": 1.0393975127292408e-05, "loss": 0.8838, "step": 9898 }, { "epoch": 0.5, "learning_rate": 1.0392330983500546e-05, "loss": 0.955, "step": 9899 }, { "epoch": 0.5, "learning_rate": 1.039068682908674e-05, "loss": 0.9926, "step": 9900 }, { "epoch": 0.5, "learning_rate": 1.0389042664095501e-05, "loss": 0.8556, "step": 9901 }, { "epoch": 0.5, "learning_rate": 1.038739848857135e-05, "loss": 0.915, "step": 9902 }, { "epoch": 0.5, "learning_rate": 1.0385754302558795e-05, "loss": 0.926, "step": 9903 }, { "epoch": 0.5, "learning_rate": 1.0384110106102352e-05, "loss": 0.8535, "step": 9904 }, { "epoch": 0.5, "learning_rate": 1.0382465899246537e-05, "loss": 0.9573, "step": 9905 }, { "epoch": 0.5, "learning_rate": 1.0380821682035864e-05, "loss": 0.8331, "step": 9906 }, { "epoch": 0.5, "learning_rate": 1.037917745451485e-05, "loss": 1.0287, "step": 9907 }, { "epoch": 0.5, "learning_rate": 1.037753321672801e-05, "loss": 0.9696, "step": 9908 }, { "epoch": 0.5, "learning_rate": 1.0375888968719863e-05, "loss": 0.9216, "step": 9909 }, { "epoch": 0.5, "learning_rate": 1.0374244710534918e-05, "loss": 0.9186, "step": 9910 }, { "epoch": 0.5, "learning_rate": 1.0372600442217694e-05, "loss": 0.9339, "step": 9911 }, { "epoch": 0.5, "learning_rate": 1.0370956163812712e-05, "loss": 0.9218, "step": 9912 }, { "epoch": 0.5, "learning_rate": 1.0369311875364486e-05, "loss": 0.943, "step": 9913 }, { "epoch": 0.5, "learning_rate": 1.0367667576917536e-05, "loss": 0.8444, "step": 9914 }, { "epoch": 0.5, "learning_rate": 1.0366023268516375e-05, "loss": 0.9045, "step": 9915 }, { "epoch": 0.5, "learning_rate": 1.0364378950205524e-05, "loss": 0.8478, "step": 9916 }, { "epoch": 0.5, "learning_rate": 1.03627346220295e-05, "loss": 0.8811, "step": 9917 }, { "epoch": 0.5, "learning_rate": 1.0361090284032821e-05, "loss": 0.9231, "step": 9918 }, { "epoch": 0.5, "learning_rate": 1.035944593626001e-05, "loss": 0.9071, "step": 9919 }, { "epoch": 0.5, "learning_rate": 1.0357801578755582e-05, "loss": 0.9032, "step": 9920 }, { "epoch": 0.5, "learning_rate": 1.0356157211564054e-05, "loss": 0.8924, "step": 9921 }, { "epoch": 0.5, "learning_rate": 1.0354512834729953e-05, "loss": 0.9373, "step": 9922 }, { "epoch": 0.5, "learning_rate": 1.0352868448297791e-05, "loss": 0.9534, "step": 9923 }, { "epoch": 0.5, "learning_rate": 1.035122405231209e-05, "loss": 0.9273, "step": 9924 }, { "epoch": 0.5, "learning_rate": 1.0349579646817377e-05, "loss": 0.9, "step": 9925 }, { "epoch": 0.5, "learning_rate": 1.0347935231858161e-05, "loss": 0.9541, "step": 9926 }, { "epoch": 0.5, "learning_rate": 1.034629080747897e-05, "loss": 0.9031, "step": 9927 }, { "epoch": 0.5, "learning_rate": 1.0344646373724328e-05, "loss": 0.943, "step": 9928 }, { "epoch": 0.5, "learning_rate": 1.0343001930638748e-05, "loss": 0.9464, "step": 9929 }, { "epoch": 0.5, "learning_rate": 1.0341357478266755e-05, "loss": 0.9321, "step": 9930 }, { "epoch": 0.5, "learning_rate": 1.0339713016652875e-05, "loss": 0.8607, "step": 9931 }, { "epoch": 0.5, "learning_rate": 1.0338068545841624e-05, "loss": 0.9082, "step": 9932 }, { "epoch": 0.5, "learning_rate": 1.0336424065877529e-05, "loss": 0.9501, "step": 9933 }, { "epoch": 0.5, "learning_rate": 1.0334779576805108e-05, "loss": 0.9738, "step": 9934 }, { "epoch": 0.5, "learning_rate": 1.0333135078668885e-05, "loss": 0.9456, "step": 9935 }, { "epoch": 0.5, "learning_rate": 1.0331490571513386e-05, "loss": 0.9037, "step": 9936 }, { "epoch": 0.5, "learning_rate": 1.0329846055383136e-05, "loss": 0.8963, "step": 9937 }, { "epoch": 0.5, "learning_rate": 1.032820153032265e-05, "loss": 0.8799, "step": 9938 }, { "epoch": 0.5, "learning_rate": 1.0326556996376459e-05, "loss": 0.8572, "step": 9939 }, { "epoch": 0.5, "learning_rate": 1.0324912453589084e-05, "loss": 0.9206, "step": 9940 }, { "epoch": 0.51, "learning_rate": 1.032326790200505e-05, "loss": 0.9106, "step": 9941 }, { "epoch": 0.51, "learning_rate": 1.0321623341668878e-05, "loss": 0.9305, "step": 9942 }, { "epoch": 0.51, "learning_rate": 1.0319978772625102e-05, "loss": 0.9382, "step": 9943 }, { "epoch": 0.51, "learning_rate": 1.031833419491824e-05, "loss": 0.8977, "step": 9944 }, { "epoch": 0.51, "learning_rate": 1.0316689608592816e-05, "loss": 0.9037, "step": 9945 }, { "epoch": 0.51, "learning_rate": 1.0315045013693362e-05, "loss": 0.877, "step": 9946 }, { "epoch": 0.51, "learning_rate": 1.0313400410264394e-05, "loss": 0.9648, "step": 9947 }, { "epoch": 0.51, "learning_rate": 1.0311755798350444e-05, "loss": 0.9424, "step": 9948 }, { "epoch": 0.51, "learning_rate": 1.031011117799604e-05, "loss": 0.9502, "step": 9949 }, { "epoch": 0.51, "learning_rate": 1.0308466549245706e-05, "loss": 0.877, "step": 9950 }, { "epoch": 0.51, "learning_rate": 1.0306821912143967e-05, "loss": 0.878, "step": 9951 }, { "epoch": 0.51, "learning_rate": 1.030517726673535e-05, "loss": 0.8334, "step": 9952 }, { "epoch": 0.51, "learning_rate": 1.0303532613064387e-05, "loss": 0.9297, "step": 9953 }, { "epoch": 0.51, "learning_rate": 1.0301887951175597e-05, "loss": 0.949, "step": 9954 }, { "epoch": 0.51, "learning_rate": 1.0300243281113516e-05, "loss": 0.9351, "step": 9955 }, { "epoch": 0.51, "learning_rate": 1.0298598602922668e-05, "loss": 0.9228, "step": 9956 }, { "epoch": 0.51, "learning_rate": 1.0296953916647574e-05, "loss": 0.9212, "step": 9957 }, { "epoch": 0.51, "learning_rate": 1.0295309222332771e-05, "loss": 0.9139, "step": 9958 }, { "epoch": 0.51, "learning_rate": 1.029366452002279e-05, "loss": 0.9061, "step": 9959 }, { "epoch": 0.51, "learning_rate": 1.029201980976215e-05, "loss": 0.8557, "step": 9960 }, { "epoch": 0.51, "learning_rate": 1.0290375091595386e-05, "loss": 0.9588, "step": 9961 }, { "epoch": 0.51, "learning_rate": 1.0288730365567025e-05, "loss": 0.9344, "step": 9962 }, { "epoch": 0.51, "learning_rate": 1.0287085631721596e-05, "loss": 0.9164, "step": 9963 }, { "epoch": 0.51, "learning_rate": 1.028544089010363e-05, "loss": 0.9944, "step": 9964 }, { "epoch": 0.51, "learning_rate": 1.0283796140757655e-05, "loss": 0.9204, "step": 9965 }, { "epoch": 0.51, "learning_rate": 1.0282151383728201e-05, "loss": 0.9186, "step": 9966 }, { "epoch": 0.51, "learning_rate": 1.02805066190598e-05, "loss": 0.8945, "step": 9967 }, { "epoch": 0.51, "learning_rate": 1.0278861846796982e-05, "loss": 0.9479, "step": 9968 }, { "epoch": 0.51, "learning_rate": 1.027721706698427e-05, "loss": 0.8789, "step": 9969 }, { "epoch": 0.51, "learning_rate": 1.0275572279666205e-05, "loss": 0.8524, "step": 9970 }, { "epoch": 0.51, "learning_rate": 1.0273927484887316e-05, "loss": 0.864, "step": 9971 }, { "epoch": 0.51, "learning_rate": 1.0272282682692127e-05, "loss": 0.942, "step": 9972 }, { "epoch": 0.51, "learning_rate": 1.0270637873125178e-05, "loss": 0.9067, "step": 9973 }, { "epoch": 0.51, "learning_rate": 1.0268993056230996e-05, "loss": 0.9624, "step": 9974 }, { "epoch": 0.51, "learning_rate": 1.026734823205411e-05, "loss": 0.8378, "step": 9975 }, { "epoch": 0.51, "learning_rate": 1.0265703400639056e-05, "loss": 0.9957, "step": 9976 }, { "epoch": 0.51, "learning_rate": 1.0264058562030368e-05, "loss": 0.9741, "step": 9977 }, { "epoch": 0.51, "learning_rate": 1.0262413716272572e-05, "loss": 0.9349, "step": 9978 }, { "epoch": 0.51, "learning_rate": 1.0260768863410207e-05, "loss": 0.9131, "step": 9979 }, { "epoch": 0.51, "learning_rate": 1.0259124003487801e-05, "loss": 0.9247, "step": 9980 }, { "epoch": 0.51, "learning_rate": 1.0257479136549889e-05, "loss": 0.9201, "step": 9981 }, { "epoch": 0.51, "learning_rate": 1.0255834262641e-05, "loss": 0.9203, "step": 9982 }, { "epoch": 0.51, "learning_rate": 1.0254189381805672e-05, "loss": 0.8243, "step": 9983 }, { "epoch": 0.51, "learning_rate": 1.0252544494088439e-05, "loss": 0.9002, "step": 9984 }, { "epoch": 0.51, "learning_rate": 1.0250899599533833e-05, "loss": 0.8784, "step": 9985 }, { "epoch": 0.51, "learning_rate": 1.0249254698186385e-05, "loss": 0.9173, "step": 9986 }, { "epoch": 0.51, "learning_rate": 1.0247609790090632e-05, "loss": 0.9879, "step": 9987 }, { "epoch": 0.51, "learning_rate": 1.0245964875291106e-05, "loss": 0.8396, "step": 9988 }, { "epoch": 0.51, "learning_rate": 1.0244319953832343e-05, "loss": 0.8899, "step": 9989 }, { "epoch": 0.51, "learning_rate": 1.024267502575888e-05, "loss": 0.8971, "step": 9990 }, { "epoch": 0.51, "learning_rate": 1.0241030091115247e-05, "loss": 0.9728, "step": 9991 }, { "epoch": 0.51, "learning_rate": 1.023938514994598e-05, "loss": 0.9087, "step": 9992 }, { "epoch": 0.51, "learning_rate": 1.0237740202295616e-05, "loss": 1.0005, "step": 9993 }, { "epoch": 0.51, "learning_rate": 1.0236095248208688e-05, "loss": 0.9366, "step": 9994 }, { "epoch": 0.51, "learning_rate": 1.0234450287729731e-05, "loss": 0.7918, "step": 9995 }, { "epoch": 0.51, "learning_rate": 1.0232805320903286e-05, "loss": 0.8439, "step": 9996 }, { "epoch": 0.51, "learning_rate": 1.0231160347773883e-05, "loss": 0.9231, "step": 9997 }, { "epoch": 0.51, "learning_rate": 1.0229515368386055e-05, "loss": 0.9047, "step": 9998 }, { "epoch": 0.51, "learning_rate": 1.0227870382784349e-05, "loss": 0.909, "step": 9999 }, { "epoch": 0.51, "learning_rate": 1.022622539101329e-05, "loss": 0.8964, "step": 10000 }, { "epoch": 0.51, "learning_rate": 1.0224580393117422e-05, "loss": 1.0377, "step": 10001 }, { "epoch": 0.51, "learning_rate": 1.022293538914128e-05, "loss": 0.9684, "step": 10002 }, { "epoch": 0.51, "learning_rate": 1.0221290379129398e-05, "loss": 0.8828, "step": 10003 }, { "epoch": 0.51, "learning_rate": 1.0219645363126314e-05, "loss": 0.8305, "step": 10004 }, { "epoch": 0.51, "learning_rate": 1.0218000341176566e-05, "loss": 0.9229, "step": 10005 }, { "epoch": 0.51, "learning_rate": 1.0216355313324688e-05, "loss": 0.9755, "step": 10006 }, { "epoch": 0.51, "learning_rate": 1.0214710279615223e-05, "loss": 0.9814, "step": 10007 }, { "epoch": 0.51, "learning_rate": 1.0213065240092705e-05, "loss": 0.9086, "step": 10008 }, { "epoch": 0.51, "learning_rate": 1.0211420194801674e-05, "loss": 0.9745, "step": 10009 }, { "epoch": 0.51, "learning_rate": 1.0209775143786662e-05, "loss": 0.923, "step": 10010 }, { "epoch": 0.51, "learning_rate": 1.0208130087092216e-05, "loss": 0.9091, "step": 10011 }, { "epoch": 0.51, "learning_rate": 1.0206485024762866e-05, "loss": 0.8982, "step": 10012 }, { "epoch": 0.51, "learning_rate": 1.0204839956843153e-05, "loss": 0.9341, "step": 10013 }, { "epoch": 0.51, "learning_rate": 1.0203194883377618e-05, "loss": 0.9948, "step": 10014 }, { "epoch": 0.51, "learning_rate": 1.02015498044108e-05, "loss": 0.8613, "step": 10015 }, { "epoch": 0.51, "learning_rate": 1.0199904719987232e-05, "loss": 0.9214, "step": 10016 }, { "epoch": 0.51, "learning_rate": 1.0198259630151455e-05, "loss": 0.9285, "step": 10017 }, { "epoch": 0.51, "learning_rate": 1.0196614534948013e-05, "loss": 0.942, "step": 10018 }, { "epoch": 0.51, "learning_rate": 1.019496943442144e-05, "loss": 1.0095, "step": 10019 }, { "epoch": 0.51, "learning_rate": 1.0193324328616278e-05, "loss": 0.9115, "step": 10020 }, { "epoch": 0.51, "learning_rate": 1.0191679217577065e-05, "loss": 0.9449, "step": 10021 }, { "epoch": 0.51, "learning_rate": 1.0190034101348341e-05, "loss": 0.9743, "step": 10022 }, { "epoch": 0.51, "learning_rate": 1.0188388979974646e-05, "loss": 0.9125, "step": 10023 }, { "epoch": 0.51, "learning_rate": 1.0186743853500521e-05, "loss": 0.943, "step": 10024 }, { "epoch": 0.51, "learning_rate": 1.01850987219705e-05, "loss": 0.8718, "step": 10025 }, { "epoch": 0.51, "learning_rate": 1.0183453585429132e-05, "loss": 0.9405, "step": 10026 }, { "epoch": 0.51, "learning_rate": 1.0181808443920955e-05, "loss": 0.8575, "step": 10027 }, { "epoch": 0.51, "learning_rate": 1.0180163297490505e-05, "loss": 0.9603, "step": 10028 }, { "epoch": 0.51, "learning_rate": 1.0178518146182325e-05, "loss": 0.9054, "step": 10029 }, { "epoch": 0.51, "learning_rate": 1.0176872990040958e-05, "loss": 0.924, "step": 10030 }, { "epoch": 0.51, "learning_rate": 1.0175227829110942e-05, "loss": 0.8618, "step": 10031 }, { "epoch": 0.51, "learning_rate": 1.017358266343682e-05, "loss": 0.9158, "step": 10032 }, { "epoch": 0.51, "learning_rate": 1.0171937493063133e-05, "loss": 0.9185, "step": 10033 }, { "epoch": 0.51, "learning_rate": 1.0170292318034418e-05, "loss": 0.8849, "step": 10034 }, { "epoch": 0.51, "learning_rate": 1.016864713839522e-05, "loss": 0.8525, "step": 10035 }, { "epoch": 0.51, "learning_rate": 1.0167001954190083e-05, "loss": 0.9904, "step": 10036 }, { "epoch": 0.51, "learning_rate": 1.0165356765463543e-05, "loss": 0.9016, "step": 10037 }, { "epoch": 0.51, "learning_rate": 1.0163711572260146e-05, "loss": 0.9008, "step": 10038 }, { "epoch": 0.51, "learning_rate": 1.0162066374624433e-05, "loss": 0.9309, "step": 10039 }, { "epoch": 0.51, "learning_rate": 1.0160421172600943e-05, "loss": 0.8958, "step": 10040 }, { "epoch": 0.51, "learning_rate": 1.015877596623422e-05, "loss": 0.8654, "step": 10041 }, { "epoch": 0.51, "learning_rate": 1.0157130755568812e-05, "loss": 0.999, "step": 10042 }, { "epoch": 0.51, "learning_rate": 1.015548554064925e-05, "loss": 0.8475, "step": 10043 }, { "epoch": 0.51, "learning_rate": 1.0153840321520085e-05, "loss": 0.8918, "step": 10044 }, { "epoch": 0.51, "learning_rate": 1.015219509822586e-05, "loss": 0.946, "step": 10045 }, { "epoch": 0.51, "learning_rate": 1.0150549870811108e-05, "loss": 0.9617, "step": 10046 }, { "epoch": 0.51, "learning_rate": 1.014890463932038e-05, "loss": 0.8928, "step": 10047 }, { "epoch": 0.51, "learning_rate": 1.0147259403798222e-05, "loss": 0.9159, "step": 10048 }, { "epoch": 0.51, "learning_rate": 1.0145614164289169e-05, "loss": 0.9076, "step": 10049 }, { "epoch": 0.51, "learning_rate": 1.0143968920837768e-05, "loss": 0.9576, "step": 10050 }, { "epoch": 0.51, "learning_rate": 1.0142323673488562e-05, "loss": 0.9682, "step": 10051 }, { "epoch": 0.51, "learning_rate": 1.0140678422286091e-05, "loss": 0.8895, "step": 10052 }, { "epoch": 0.51, "learning_rate": 1.0139033167274905e-05, "loss": 0.8971, "step": 10053 }, { "epoch": 0.51, "learning_rate": 1.0137387908499541e-05, "loss": 0.9113, "step": 10054 }, { "epoch": 0.51, "learning_rate": 1.013574264600455e-05, "loss": 1.0231, "step": 10055 }, { "epoch": 0.51, "learning_rate": 1.0134097379834468e-05, "loss": 1.004, "step": 10056 }, { "epoch": 0.51, "learning_rate": 1.013245211003384e-05, "loss": 0.8322, "step": 10057 }, { "epoch": 0.51, "learning_rate": 1.0130806836647218e-05, "loss": 0.9592, "step": 10058 }, { "epoch": 0.51, "learning_rate": 1.0129161559719134e-05, "loss": 0.9512, "step": 10059 }, { "epoch": 0.51, "learning_rate": 1.012751627929414e-05, "loss": 0.9083, "step": 10060 }, { "epoch": 0.51, "learning_rate": 1.0125870995416782e-05, "loss": 0.9653, "step": 10061 }, { "epoch": 0.51, "learning_rate": 1.01242257081316e-05, "loss": 0.886, "step": 10062 }, { "epoch": 0.51, "learning_rate": 1.0122580417483135e-05, "loss": 0.8506, "step": 10063 }, { "epoch": 0.51, "learning_rate": 1.0120935123515937e-05, "loss": 0.9971, "step": 10064 }, { "epoch": 0.51, "learning_rate": 1.011928982627455e-05, "loss": 0.8972, "step": 10065 }, { "epoch": 0.51, "learning_rate": 1.0117644525803517e-05, "loss": 0.8362, "step": 10066 }, { "epoch": 0.51, "learning_rate": 1.0115999222147386e-05, "loss": 0.9696, "step": 10067 }, { "epoch": 0.51, "learning_rate": 1.0114353915350699e-05, "loss": 0.9222, "step": 10068 }, { "epoch": 0.51, "learning_rate": 1.0112708605458e-05, "loss": 0.9003, "step": 10069 }, { "epoch": 0.51, "learning_rate": 1.0111063292513837e-05, "loss": 0.8764, "step": 10070 }, { "epoch": 0.51, "learning_rate": 1.010941797656275e-05, "loss": 0.9699, "step": 10071 }, { "epoch": 0.51, "learning_rate": 1.0107772657649289e-05, "loss": 0.9035, "step": 10072 }, { "epoch": 0.51, "learning_rate": 1.0106127335818001e-05, "loss": 0.969, "step": 10073 }, { "epoch": 0.51, "learning_rate": 1.0104482011113425e-05, "loss": 0.9699, "step": 10074 }, { "epoch": 0.51, "learning_rate": 1.0102836683580111e-05, "loss": 0.9706, "step": 10075 }, { "epoch": 0.51, "learning_rate": 1.0101191353262605e-05, "loss": 0.8258, "step": 10076 }, { "epoch": 0.51, "learning_rate": 1.0099546020205449e-05, "loss": 0.9549, "step": 10077 }, { "epoch": 0.51, "learning_rate": 1.0097900684453188e-05, "loss": 0.9711, "step": 10078 }, { "epoch": 0.51, "learning_rate": 1.0096255346050375e-05, "loss": 0.94, "step": 10079 }, { "epoch": 0.51, "learning_rate": 1.0094610005041547e-05, "loss": 0.9108, "step": 10080 }, { "epoch": 0.51, "learning_rate": 1.0092964661471254e-05, "loss": 0.9233, "step": 10081 }, { "epoch": 0.51, "learning_rate": 1.0091319315384046e-05, "loss": 0.9089, "step": 10082 }, { "epoch": 0.51, "learning_rate": 1.008967396682446e-05, "loss": 0.9913, "step": 10083 }, { "epoch": 0.51, "learning_rate": 1.0088028615837046e-05, "loss": 0.9405, "step": 10084 }, { "epoch": 0.51, "learning_rate": 1.0086383262466354e-05, "loss": 0.8702, "step": 10085 }, { "epoch": 0.51, "learning_rate": 1.0084737906756927e-05, "loss": 0.991, "step": 10086 }, { "epoch": 0.51, "learning_rate": 1.008309254875331e-05, "loss": 0.8805, "step": 10087 }, { "epoch": 0.51, "learning_rate": 1.008144718850005e-05, "loss": 0.8979, "step": 10088 }, { "epoch": 0.51, "learning_rate": 1.0079801826041698e-05, "loss": 0.8383, "step": 10089 }, { "epoch": 0.51, "learning_rate": 1.0078156461422793e-05, "loss": 0.8839, "step": 10090 }, { "epoch": 0.51, "learning_rate": 1.0076511094687886e-05, "loss": 0.9594, "step": 10091 }, { "epoch": 0.51, "learning_rate": 1.0074865725881522e-05, "loss": 0.8602, "step": 10092 }, { "epoch": 0.51, "learning_rate": 1.0073220355048248e-05, "loss": 0.8245, "step": 10093 }, { "epoch": 0.51, "learning_rate": 1.007157498223261e-05, "loss": 1.0154, "step": 10094 }, { "epoch": 0.51, "learning_rate": 1.006992960747916e-05, "loss": 0.7807, "step": 10095 }, { "epoch": 0.51, "learning_rate": 1.0068284230832432e-05, "loss": 0.8434, "step": 10096 }, { "epoch": 0.51, "learning_rate": 1.0066638852336988e-05, "loss": 0.8615, "step": 10097 }, { "epoch": 0.51, "learning_rate": 1.0064993472037367e-05, "loss": 0.9593, "step": 10098 }, { "epoch": 0.51, "learning_rate": 1.0063348089978115e-05, "loss": 0.861, "step": 10099 }, { "epoch": 0.51, "learning_rate": 1.006170270620378e-05, "loss": 0.95, "step": 10100 }, { "epoch": 0.51, "learning_rate": 1.0060057320758913e-05, "loss": 0.9552, "step": 10101 }, { "epoch": 0.51, "learning_rate": 1.0058411933688057e-05, "loss": 0.9263, "step": 10102 }, { "epoch": 0.51, "learning_rate": 1.005676654503576e-05, "loss": 0.953, "step": 10103 }, { "epoch": 0.51, "learning_rate": 1.005512115484657e-05, "loss": 0.8574, "step": 10104 }, { "epoch": 0.51, "learning_rate": 1.005347576316503e-05, "loss": 0.8922, "step": 10105 }, { "epoch": 0.51, "learning_rate": 1.005183037003569e-05, "loss": 0.8805, "step": 10106 }, { "epoch": 0.51, "learning_rate": 1.0050184975503104e-05, "loss": 0.87, "step": 10107 }, { "epoch": 0.51, "learning_rate": 1.0048539579611809e-05, "loss": 0.8309, "step": 10108 }, { "epoch": 0.51, "learning_rate": 1.0046894182406357e-05, "loss": 1.0093, "step": 10109 }, { "epoch": 0.51, "learning_rate": 1.0045248783931296e-05, "loss": 0.8995, "step": 10110 }, { "epoch": 0.51, "learning_rate": 1.0043603384231172e-05, "loss": 0.9581, "step": 10111 }, { "epoch": 0.51, "learning_rate": 1.004195798335053e-05, "loss": 0.8986, "step": 10112 }, { "epoch": 0.51, "learning_rate": 1.0040312581333925e-05, "loss": 1.0325, "step": 10113 }, { "epoch": 0.51, "learning_rate": 1.0038667178225896e-05, "loss": 0.8895, "step": 10114 }, { "epoch": 0.51, "learning_rate": 1.0037021774070996e-05, "loss": 0.8858, "step": 10115 }, { "epoch": 0.51, "learning_rate": 1.0035376368913772e-05, "loss": 1.0018, "step": 10116 }, { "epoch": 0.51, "learning_rate": 1.0033730962798767e-05, "loss": 0.9953, "step": 10117 }, { "epoch": 0.51, "learning_rate": 1.0032085555770535e-05, "loss": 1.0131, "step": 10118 }, { "epoch": 0.51, "learning_rate": 1.0030440147873619e-05, "loss": 0.8894, "step": 10119 }, { "epoch": 0.51, "learning_rate": 1.0028794739152568e-05, "loss": 0.8975, "step": 10120 }, { "epoch": 0.51, "learning_rate": 1.0027149329651933e-05, "loss": 0.821, "step": 10121 }, { "epoch": 0.51, "learning_rate": 1.0025503919416255e-05, "loss": 0.9379, "step": 10122 }, { "epoch": 0.51, "learning_rate": 1.0023858508490088e-05, "loss": 0.9196, "step": 10123 }, { "epoch": 0.51, "learning_rate": 1.0022213096917977e-05, "loss": 0.9047, "step": 10124 }, { "epoch": 0.51, "learning_rate": 1.0020567684744467e-05, "loss": 0.8374, "step": 10125 }, { "epoch": 0.51, "learning_rate": 1.0018922272014116e-05, "loss": 0.8331, "step": 10126 }, { "epoch": 0.51, "learning_rate": 1.001727685877146e-05, "loss": 0.8526, "step": 10127 }, { "epoch": 0.51, "learning_rate": 1.0015631445061049e-05, "loss": 0.9681, "step": 10128 }, { "epoch": 0.51, "learning_rate": 1.0013986030927436e-05, "loss": 0.9324, "step": 10129 }, { "epoch": 0.51, "learning_rate": 1.0012340616415166e-05, "loss": 0.9241, "step": 10130 }, { "epoch": 0.51, "learning_rate": 1.0010695201568785e-05, "loss": 0.9187, "step": 10131 }, { "epoch": 0.51, "learning_rate": 1.000904978643285e-05, "loss": 0.9335, "step": 10132 }, { "epoch": 0.51, "learning_rate": 1.0007404371051894e-05, "loss": 0.8121, "step": 10133 }, { "epoch": 0.51, "learning_rate": 1.0005758955470473e-05, "loss": 0.9722, "step": 10134 }, { "epoch": 0.51, "learning_rate": 1.0004113539733138e-05, "loss": 0.9117, "step": 10135 }, { "epoch": 0.51, "learning_rate": 1.000246812388443e-05, "loss": 0.9543, "step": 10136 }, { "epoch": 0.51, "learning_rate": 1.00008227079689e-05, "loss": 0.888, "step": 10137 }, { "epoch": 0.52, "learning_rate": 9.999177292031101e-06, "loss": 0.8431, "step": 10138 }, { "epoch": 0.52, "learning_rate": 9.99753187611557e-06, "loss": 0.9501, "step": 10139 }, { "epoch": 0.52, "learning_rate": 9.995886460266865e-06, "loss": 0.8808, "step": 10140 }, { "epoch": 0.52, "learning_rate": 9.994241044529529e-06, "loss": 0.8483, "step": 10141 }, { "epoch": 0.52, "learning_rate": 9.992595628948107e-06, "loss": 0.9721, "step": 10142 }, { "epoch": 0.52, "learning_rate": 9.990950213567155e-06, "loss": 0.9027, "step": 10143 }, { "epoch": 0.52, "learning_rate": 9.989304798431217e-06, "loss": 1.0759, "step": 10144 }, { "epoch": 0.52, "learning_rate": 9.987659383584834e-06, "loss": 0.9312, "step": 10145 }, { "epoch": 0.52, "learning_rate": 9.986013969072566e-06, "loss": 0.9436, "step": 10146 }, { "epoch": 0.52, "learning_rate": 9.984368554938953e-06, "loss": 0.9328, "step": 10147 }, { "epoch": 0.52, "learning_rate": 9.982723141228542e-06, "loss": 0.98, "step": 10148 }, { "epoch": 0.52, "learning_rate": 9.981077727985887e-06, "loss": 0.8148, "step": 10149 }, { "epoch": 0.52, "learning_rate": 9.979432315255535e-06, "loss": 0.9183, "step": 10150 }, { "epoch": 0.52, "learning_rate": 9.977786903082023e-06, "loss": 0.9827, "step": 10151 }, { "epoch": 0.52, "learning_rate": 9.976141491509914e-06, "loss": 0.7849, "step": 10152 }, { "epoch": 0.52, "learning_rate": 9.974496080583748e-06, "loss": 1.0312, "step": 10153 }, { "epoch": 0.52, "learning_rate": 9.972850670348069e-06, "loss": 0.8998, "step": 10154 }, { "epoch": 0.52, "learning_rate": 9.971205260847436e-06, "loss": 0.8787, "step": 10155 }, { "epoch": 0.52, "learning_rate": 9.969559852126386e-06, "loss": 0.9014, "step": 10156 }, { "epoch": 0.52, "learning_rate": 9.967914444229466e-06, "loss": 0.9478, "step": 10157 }, { "epoch": 0.52, "learning_rate": 9.966269037201235e-06, "loss": 0.9933, "step": 10158 }, { "epoch": 0.52, "learning_rate": 9.964623631086233e-06, "loss": 0.9303, "step": 10159 }, { "epoch": 0.52, "learning_rate": 9.962978225929004e-06, "loss": 0.9321, "step": 10160 }, { "epoch": 0.52, "learning_rate": 9.961332821774106e-06, "loss": 0.9423, "step": 10161 }, { "epoch": 0.52, "learning_rate": 9.95968741866608e-06, "loss": 0.932, "step": 10162 }, { "epoch": 0.52, "learning_rate": 9.95804201664947e-06, "loss": 0.9324, "step": 10163 }, { "epoch": 0.52, "learning_rate": 9.95639661576883e-06, "loss": 0.8443, "step": 10164 }, { "epoch": 0.52, "learning_rate": 9.954751216068707e-06, "loss": 0.8377, "step": 10165 }, { "epoch": 0.52, "learning_rate": 9.953105817593643e-06, "loss": 0.9669, "step": 10166 }, { "epoch": 0.52, "learning_rate": 9.951460420388195e-06, "loss": 0.8355, "step": 10167 }, { "epoch": 0.52, "learning_rate": 9.949815024496901e-06, "loss": 0.8972, "step": 10168 }, { "epoch": 0.52, "learning_rate": 9.948169629964308e-06, "loss": 0.968, "step": 10169 }, { "epoch": 0.52, "learning_rate": 9.946524236834972e-06, "loss": 0.9164, "step": 10170 }, { "epoch": 0.52, "learning_rate": 9.944878845153435e-06, "loss": 0.8791, "step": 10171 }, { "epoch": 0.52, "learning_rate": 9.94323345496424e-06, "loss": 0.8892, "step": 10172 }, { "epoch": 0.52, "learning_rate": 9.941588066311947e-06, "loss": 0.9309, "step": 10173 }, { "epoch": 0.52, "learning_rate": 9.939942679241092e-06, "loss": 0.9054, "step": 10174 }, { "epoch": 0.52, "learning_rate": 9.93829729379622e-06, "loss": 0.8886, "step": 10175 }, { "epoch": 0.52, "learning_rate": 9.936651910021887e-06, "loss": 0.9013, "step": 10176 }, { "epoch": 0.52, "learning_rate": 9.935006527962638e-06, "loss": 0.9355, "step": 10177 }, { "epoch": 0.52, "learning_rate": 9.933361147663012e-06, "loss": 0.9301, "step": 10178 }, { "epoch": 0.52, "learning_rate": 9.93171576916757e-06, "loss": 0.9166, "step": 10179 }, { "epoch": 0.52, "learning_rate": 9.930070392520848e-06, "loss": 1.0079, "step": 10180 }, { "epoch": 0.52, "learning_rate": 9.928425017767392e-06, "loss": 0.9999, "step": 10181 }, { "epoch": 0.52, "learning_rate": 9.926779644951756e-06, "loss": 0.9337, "step": 10182 }, { "epoch": 0.52, "learning_rate": 9.925134274118482e-06, "loss": 1.0095, "step": 10183 }, { "epoch": 0.52, "learning_rate": 9.923488905312115e-06, "loss": 0.919, "step": 10184 }, { "epoch": 0.52, "learning_rate": 9.921843538577212e-06, "loss": 0.9721, "step": 10185 }, { "epoch": 0.52, "learning_rate": 9.920198173958307e-06, "loss": 0.9529, "step": 10186 }, { "epoch": 0.52, "learning_rate": 9.91855281149995e-06, "loss": 0.897, "step": 10187 }, { "epoch": 0.52, "learning_rate": 9.916907451246691e-06, "loss": 0.9507, "step": 10188 }, { "epoch": 0.52, "learning_rate": 9.915262093243077e-06, "loss": 0.9265, "step": 10189 }, { "epoch": 0.52, "learning_rate": 9.913616737533647e-06, "loss": 0.9157, "step": 10190 }, { "epoch": 0.52, "learning_rate": 9.911971384162957e-06, "loss": 0.941, "step": 10191 }, { "epoch": 0.52, "learning_rate": 9.910326033175546e-06, "loss": 0.9324, "step": 10192 }, { "epoch": 0.52, "learning_rate": 9.908680684615958e-06, "loss": 0.8901, "step": 10193 }, { "epoch": 0.52, "learning_rate": 9.907035338528748e-06, "loss": 0.8868, "step": 10194 }, { "epoch": 0.52, "learning_rate": 9.905389994958458e-06, "loss": 0.8396, "step": 10195 }, { "epoch": 0.52, "learning_rate": 9.903744653949629e-06, "loss": 0.9297, "step": 10196 }, { "epoch": 0.52, "learning_rate": 9.902099315546815e-06, "loss": 0.8246, "step": 10197 }, { "epoch": 0.52, "learning_rate": 9.900453979794558e-06, "loss": 0.967, "step": 10198 }, { "epoch": 0.52, "learning_rate": 9.898808646737398e-06, "loss": 0.8854, "step": 10199 }, { "epoch": 0.52, "learning_rate": 9.89716331641989e-06, "loss": 0.9819, "step": 10200 }, { "epoch": 0.52, "learning_rate": 9.895517988886578e-06, "loss": 0.8271, "step": 10201 }, { "epoch": 0.52, "learning_rate": 9.893872664182002e-06, "loss": 0.9614, "step": 10202 }, { "epoch": 0.52, "learning_rate": 9.892227342350714e-06, "loss": 0.8782, "step": 10203 }, { "epoch": 0.52, "learning_rate": 9.890582023437255e-06, "loss": 0.8714, "step": 10204 }, { "epoch": 0.52, "learning_rate": 9.888936707486166e-06, "loss": 0.8477, "step": 10205 }, { "epoch": 0.52, "learning_rate": 9.887291394542004e-06, "loss": 0.8636, "step": 10206 }, { "epoch": 0.52, "learning_rate": 9.885646084649301e-06, "loss": 0.9018, "step": 10207 }, { "epoch": 0.52, "learning_rate": 9.884000777852618e-06, "loss": 0.8659, "step": 10208 }, { "epoch": 0.52, "learning_rate": 9.882355474196485e-06, "loss": 0.8907, "step": 10209 }, { "epoch": 0.52, "learning_rate": 9.88071017372545e-06, "loss": 0.9984, "step": 10210 }, { "epoch": 0.52, "learning_rate": 9.879064876484064e-06, "loss": 0.8592, "step": 10211 }, { "epoch": 0.52, "learning_rate": 9.877419582516869e-06, "loss": 0.8966, "step": 10212 }, { "epoch": 0.52, "learning_rate": 9.875774291868402e-06, "loss": 0.8118, "step": 10213 }, { "epoch": 0.52, "learning_rate": 9.874129004583221e-06, "loss": 1.0073, "step": 10214 }, { "epoch": 0.52, "learning_rate": 9.872483720705862e-06, "loss": 0.9381, "step": 10215 }, { "epoch": 0.52, "learning_rate": 9.870838440280866e-06, "loss": 0.882, "step": 10216 }, { "epoch": 0.52, "learning_rate": 9.869193163352785e-06, "loss": 0.8475, "step": 10217 }, { "epoch": 0.52, "learning_rate": 9.867547889966161e-06, "loss": 0.9781, "step": 10218 }, { "epoch": 0.52, "learning_rate": 9.865902620165534e-06, "loss": 0.9416, "step": 10219 }, { "epoch": 0.52, "learning_rate": 9.864257353995454e-06, "loss": 0.8151, "step": 10220 }, { "epoch": 0.52, "learning_rate": 9.862612091500462e-06, "loss": 0.8574, "step": 10221 }, { "epoch": 0.52, "learning_rate": 9.860966832725097e-06, "loss": 1.0109, "step": 10222 }, { "epoch": 0.52, "learning_rate": 9.85932157771391e-06, "loss": 0.972, "step": 10223 }, { "epoch": 0.52, "learning_rate": 9.857676326511443e-06, "loss": 0.8771, "step": 10224 }, { "epoch": 0.52, "learning_rate": 9.856031079162236e-06, "loss": 0.8795, "step": 10225 }, { "epoch": 0.52, "learning_rate": 9.854385835710835e-06, "loss": 0.854, "step": 10226 }, { "epoch": 0.52, "learning_rate": 9.852740596201783e-06, "loss": 0.8864, "step": 10227 }, { "epoch": 0.52, "learning_rate": 9.851095360679618e-06, "loss": 0.9699, "step": 10228 }, { "epoch": 0.52, "learning_rate": 9.849450129188895e-06, "loss": 0.8989, "step": 10229 }, { "epoch": 0.52, "learning_rate": 9.847804901774146e-06, "loss": 0.8987, "step": 10230 }, { "epoch": 0.52, "learning_rate": 9.846159678479918e-06, "loss": 0.9463, "step": 10231 }, { "epoch": 0.52, "learning_rate": 9.844514459350753e-06, "loss": 0.9608, "step": 10232 }, { "epoch": 0.52, "learning_rate": 9.842869244431195e-06, "loss": 0.9406, "step": 10233 }, { "epoch": 0.52, "learning_rate": 9.84122403376578e-06, "loss": 0.9184, "step": 10234 }, { "epoch": 0.52, "learning_rate": 9.839578827399059e-06, "loss": 0.8851, "step": 10235 }, { "epoch": 0.52, "learning_rate": 9.837933625375572e-06, "loss": 0.9046, "step": 10236 }, { "epoch": 0.52, "learning_rate": 9.836288427739856e-06, "loss": 0.9422, "step": 10237 }, { "epoch": 0.52, "learning_rate": 9.83464323453646e-06, "loss": 0.8788, "step": 10238 }, { "epoch": 0.52, "learning_rate": 9.832998045809922e-06, "loss": 0.9396, "step": 10239 }, { "epoch": 0.52, "learning_rate": 9.831352861604781e-06, "loss": 0.9714, "step": 10240 }, { "epoch": 0.52, "learning_rate": 9.829707681965585e-06, "loss": 0.9858, "step": 10241 }, { "epoch": 0.52, "learning_rate": 9.828062506936874e-06, "loss": 0.9583, "step": 10242 }, { "epoch": 0.52, "learning_rate": 9.826417336563183e-06, "loss": 0.813, "step": 10243 }, { "epoch": 0.52, "learning_rate": 9.824772170889061e-06, "loss": 0.9889, "step": 10244 }, { "epoch": 0.52, "learning_rate": 9.823127009959045e-06, "loss": 0.8352, "step": 10245 }, { "epoch": 0.52, "learning_rate": 9.821481853817675e-06, "loss": 1.0093, "step": 10246 }, { "epoch": 0.52, "learning_rate": 9.819836702509498e-06, "loss": 0.8423, "step": 10247 }, { "epoch": 0.52, "learning_rate": 9.81819155607905e-06, "loss": 1.0067, "step": 10248 }, { "epoch": 0.52, "learning_rate": 9.81654641457087e-06, "loss": 0.9622, "step": 10249 }, { "epoch": 0.52, "learning_rate": 9.814901278029501e-06, "loss": 0.8735, "step": 10250 }, { "epoch": 0.52, "learning_rate": 9.813256146499486e-06, "loss": 0.9044, "step": 10251 }, { "epoch": 0.52, "learning_rate": 9.811611020025356e-06, "loss": 0.9827, "step": 10252 }, { "epoch": 0.52, "learning_rate": 9.80996589865166e-06, "loss": 0.8573, "step": 10253 }, { "epoch": 0.52, "learning_rate": 9.808320782422938e-06, "loss": 0.8739, "step": 10254 }, { "epoch": 0.52, "learning_rate": 9.806675671383724e-06, "loss": 0.9016, "step": 10255 }, { "epoch": 0.52, "learning_rate": 9.805030565578563e-06, "loss": 0.9204, "step": 10256 }, { "epoch": 0.52, "learning_rate": 9.803385465051992e-06, "loss": 0.9148, "step": 10257 }, { "epoch": 0.52, "learning_rate": 9.801740369848545e-06, "loss": 0.8369, "step": 10258 }, { "epoch": 0.52, "learning_rate": 9.800095280012771e-06, "loss": 0.8395, "step": 10259 }, { "epoch": 0.52, "learning_rate": 9.798450195589206e-06, "loss": 0.9432, "step": 10260 }, { "epoch": 0.52, "learning_rate": 9.796805116622384e-06, "loss": 0.9124, "step": 10261 }, { "epoch": 0.52, "learning_rate": 9.79516004315685e-06, "loss": 0.9647, "step": 10262 }, { "epoch": 0.52, "learning_rate": 9.793514975237139e-06, "loss": 0.9525, "step": 10263 }, { "epoch": 0.52, "learning_rate": 9.791869912907786e-06, "loss": 0.9781, "step": 10264 }, { "epoch": 0.52, "learning_rate": 9.79022485621334e-06, "loss": 0.9028, "step": 10265 }, { "epoch": 0.52, "learning_rate": 9.788579805198331e-06, "loss": 0.9153, "step": 10266 }, { "epoch": 0.52, "learning_rate": 9.786934759907297e-06, "loss": 0.8717, "step": 10267 }, { "epoch": 0.52, "learning_rate": 9.78528972038478e-06, "loss": 0.941, "step": 10268 }, { "epoch": 0.52, "learning_rate": 9.783644686675317e-06, "loss": 0.8876, "step": 10269 }, { "epoch": 0.52, "learning_rate": 9.781999658823437e-06, "loss": 0.9113, "step": 10270 }, { "epoch": 0.52, "learning_rate": 9.78035463687369e-06, "loss": 0.9176, "step": 10271 }, { "epoch": 0.52, "learning_rate": 9.778709620870607e-06, "loss": 0.9219, "step": 10272 }, { "epoch": 0.52, "learning_rate": 9.777064610858725e-06, "loss": 0.9794, "step": 10273 }, { "epoch": 0.52, "learning_rate": 9.775419606882581e-06, "loss": 0.8493, "step": 10274 }, { "epoch": 0.52, "learning_rate": 9.773774608986714e-06, "loss": 0.9173, "step": 10275 }, { "epoch": 0.52, "learning_rate": 9.772129617215653e-06, "loss": 0.9343, "step": 10276 }, { "epoch": 0.52, "learning_rate": 9.770484631613946e-06, "loss": 0.8995, "step": 10277 }, { "epoch": 0.52, "learning_rate": 9.76883965222612e-06, "loss": 0.9492, "step": 10278 }, { "epoch": 0.52, "learning_rate": 9.767194679096718e-06, "loss": 0.9286, "step": 10279 }, { "epoch": 0.52, "learning_rate": 9.765549712270272e-06, "loss": 0.7996, "step": 10280 }, { "epoch": 0.52, "learning_rate": 9.763904751791313e-06, "loss": 0.8369, "step": 10281 }, { "epoch": 0.52, "learning_rate": 9.762259797704387e-06, "loss": 0.9026, "step": 10282 }, { "epoch": 0.52, "learning_rate": 9.760614850054023e-06, "loss": 0.9302, "step": 10283 }, { "epoch": 0.52, "learning_rate": 9.758969908884756e-06, "loss": 0.9788, "step": 10284 }, { "epoch": 0.52, "learning_rate": 9.757324974241124e-06, "loss": 0.8453, "step": 10285 }, { "epoch": 0.52, "learning_rate": 9.75568004616766e-06, "loss": 0.8984, "step": 10286 }, { "epoch": 0.52, "learning_rate": 9.754035124708895e-06, "loss": 0.9424, "step": 10287 }, { "epoch": 0.52, "learning_rate": 9.752390209909372e-06, "loss": 0.8666, "step": 10288 }, { "epoch": 0.52, "learning_rate": 9.750745301813619e-06, "loss": 0.8508, "step": 10289 }, { "epoch": 0.52, "learning_rate": 9.74910040046617e-06, "loss": 0.999, "step": 10290 }, { "epoch": 0.52, "learning_rate": 9.747455505911564e-06, "loss": 0.9699, "step": 10291 }, { "epoch": 0.52, "learning_rate": 9.745810618194332e-06, "loss": 0.9393, "step": 10292 }, { "epoch": 0.52, "learning_rate": 9.744165737359001e-06, "loss": 0.9336, "step": 10293 }, { "epoch": 0.52, "learning_rate": 9.742520863450116e-06, "loss": 0.8353, "step": 10294 }, { "epoch": 0.52, "learning_rate": 9.740875996512204e-06, "loss": 1.0135, "step": 10295 }, { "epoch": 0.52, "learning_rate": 9.739231136589797e-06, "loss": 0.9135, "step": 10296 }, { "epoch": 0.52, "learning_rate": 9.737586283727431e-06, "loss": 1.0544, "step": 10297 }, { "epoch": 0.52, "learning_rate": 9.735941437969637e-06, "loss": 0.9719, "step": 10298 }, { "epoch": 0.52, "learning_rate": 9.734296599360944e-06, "loss": 0.8915, "step": 10299 }, { "epoch": 0.52, "learning_rate": 9.732651767945893e-06, "loss": 0.9128, "step": 10300 }, { "epoch": 0.52, "learning_rate": 9.73100694376901e-06, "loss": 0.8933, "step": 10301 }, { "epoch": 0.52, "learning_rate": 9.729362126874825e-06, "loss": 0.9467, "step": 10302 }, { "epoch": 0.52, "learning_rate": 9.727717317307875e-06, "loss": 0.9475, "step": 10303 }, { "epoch": 0.52, "learning_rate": 9.72607251511269e-06, "loss": 0.9514, "step": 10304 }, { "epoch": 0.52, "learning_rate": 9.724427720333796e-06, "loss": 0.9731, "step": 10305 }, { "epoch": 0.52, "learning_rate": 9.722782933015731e-06, "loss": 0.9865, "step": 10306 }, { "epoch": 0.52, "learning_rate": 9.721138153203023e-06, "loss": 0.9771, "step": 10307 }, { "epoch": 0.52, "learning_rate": 9.719493380940203e-06, "loss": 0.947, "step": 10308 }, { "epoch": 0.52, "learning_rate": 9.717848616271802e-06, "loss": 0.9117, "step": 10309 }, { "epoch": 0.52, "learning_rate": 9.71620385924235e-06, "loss": 0.9885, "step": 10310 }, { "epoch": 0.52, "learning_rate": 9.714559109896372e-06, "loss": 0.947, "step": 10311 }, { "epoch": 0.52, "learning_rate": 9.712914368278406e-06, "loss": 0.9269, "step": 10312 }, { "epoch": 0.52, "learning_rate": 9.711269634432978e-06, "loss": 0.9556, "step": 10313 }, { "epoch": 0.52, "learning_rate": 9.709624908404616e-06, "loss": 0.993, "step": 10314 }, { "epoch": 0.52, "learning_rate": 9.707980190237853e-06, "loss": 0.9258, "step": 10315 }, { "epoch": 0.52, "learning_rate": 9.706335479977216e-06, "loss": 0.9952, "step": 10316 }, { "epoch": 0.52, "learning_rate": 9.704690777667229e-06, "loss": 0.9481, "step": 10317 }, { "epoch": 0.52, "learning_rate": 9.703046083352427e-06, "loss": 0.9234, "step": 10318 }, { "epoch": 0.52, "learning_rate": 9.701401397077339e-06, "loss": 0.9165, "step": 10319 }, { "epoch": 0.52, "learning_rate": 9.699756718886488e-06, "loss": 0.9109, "step": 10320 }, { "epoch": 0.52, "learning_rate": 9.698112048824406e-06, "loss": 0.9864, "step": 10321 }, { "epoch": 0.52, "learning_rate": 9.69646738693562e-06, "loss": 0.8517, "step": 10322 }, { "epoch": 0.52, "learning_rate": 9.694822733264651e-06, "loss": 0.7879, "step": 10323 }, { "epoch": 0.52, "learning_rate": 9.693178087856035e-06, "loss": 0.966, "step": 10324 }, { "epoch": 0.52, "learning_rate": 9.6915334507543e-06, "loss": 0.8246, "step": 10325 }, { "epoch": 0.52, "learning_rate": 9.689888822003963e-06, "loss": 0.8723, "step": 10326 }, { "epoch": 0.52, "learning_rate": 9.68824420164956e-06, "loss": 0.933, "step": 10327 }, { "epoch": 0.52, "learning_rate": 9.68659958973561e-06, "loss": 0.9125, "step": 10328 }, { "epoch": 0.52, "learning_rate": 9.684954986306643e-06, "loss": 1.0239, "step": 10329 }, { "epoch": 0.52, "learning_rate": 9.683310391407188e-06, "loss": 0.9003, "step": 10330 }, { "epoch": 0.52, "learning_rate": 9.681665805081765e-06, "loss": 0.9463, "step": 10331 }, { "epoch": 0.52, "learning_rate": 9.6800212273749e-06, "loss": 0.8974, "step": 10332 }, { "epoch": 0.52, "learning_rate": 9.678376658331123e-06, "loss": 1.0264, "step": 10333 }, { "epoch": 0.52, "learning_rate": 9.676732097994957e-06, "loss": 0.9189, "step": 10334 }, { "epoch": 0.53, "learning_rate": 9.67508754641092e-06, "loss": 0.8441, "step": 10335 }, { "epoch": 0.53, "learning_rate": 9.673443003623545e-06, "loss": 1.0041, "step": 10336 }, { "epoch": 0.53, "learning_rate": 9.671798469677354e-06, "loss": 0.8901, "step": 10337 }, { "epoch": 0.53, "learning_rate": 9.67015394461687e-06, "loss": 0.9359, "step": 10338 }, { "epoch": 0.53, "learning_rate": 9.668509428486616e-06, "loss": 0.8915, "step": 10339 }, { "epoch": 0.53, "learning_rate": 9.666864921331119e-06, "loss": 0.8438, "step": 10340 }, { "epoch": 0.53, "learning_rate": 9.665220423194894e-06, "loss": 1.0355, "step": 10341 }, { "epoch": 0.53, "learning_rate": 9.663575934122476e-06, "loss": 0.857, "step": 10342 }, { "epoch": 0.53, "learning_rate": 9.66193145415838e-06, "loss": 0.8493, "step": 10343 }, { "epoch": 0.53, "learning_rate": 9.66028698334713e-06, "loss": 0.9619, "step": 10344 }, { "epoch": 0.53, "learning_rate": 9.658642521733248e-06, "loss": 1.0197, "step": 10345 }, { "epoch": 0.53, "learning_rate": 9.656998069361253e-06, "loss": 0.8191, "step": 10346 }, { "epoch": 0.53, "learning_rate": 9.655353626275675e-06, "loss": 0.939, "step": 10347 }, { "epoch": 0.53, "learning_rate": 9.653709192521031e-06, "loss": 0.9912, "step": 10348 }, { "epoch": 0.53, "learning_rate": 9.65206476814184e-06, "loss": 0.9727, "step": 10349 }, { "epoch": 0.53, "learning_rate": 9.650420353182628e-06, "loss": 0.8751, "step": 10350 }, { "epoch": 0.53, "learning_rate": 9.648775947687914e-06, "loss": 0.9897, "step": 10351 }, { "epoch": 0.53, "learning_rate": 9.64713155170221e-06, "loss": 0.9603, "step": 10352 }, { "epoch": 0.53, "learning_rate": 9.64548716527005e-06, "loss": 0.8594, "step": 10353 }, { "epoch": 0.53, "learning_rate": 9.643842788435947e-06, "loss": 0.8765, "step": 10354 }, { "epoch": 0.53, "learning_rate": 9.642198421244421e-06, "loss": 0.9724, "step": 10355 }, { "epoch": 0.53, "learning_rate": 9.640554063739993e-06, "loss": 0.8857, "step": 10356 }, { "epoch": 0.53, "learning_rate": 9.638909715967182e-06, "loss": 0.958, "step": 10357 }, { "epoch": 0.53, "learning_rate": 9.6372653779705e-06, "loss": 0.87, "step": 10358 }, { "epoch": 0.53, "learning_rate": 9.635621049794479e-06, "loss": 0.9298, "step": 10359 }, { "epoch": 0.53, "learning_rate": 9.633976731483629e-06, "loss": 0.9702, "step": 10360 }, { "epoch": 0.53, "learning_rate": 9.632332423082468e-06, "loss": 0.961, "step": 10361 }, { "epoch": 0.53, "learning_rate": 9.630688124635516e-06, "loss": 0.8722, "step": 10362 }, { "epoch": 0.53, "learning_rate": 9.629043836187293e-06, "loss": 0.7929, "step": 10363 }, { "epoch": 0.53, "learning_rate": 9.627399557782306e-06, "loss": 0.9905, "step": 10364 }, { "epoch": 0.53, "learning_rate": 9.625755289465086e-06, "loss": 0.883, "step": 10365 }, { "epoch": 0.53, "learning_rate": 9.624111031280144e-06, "loss": 0.9055, "step": 10366 }, { "epoch": 0.53, "learning_rate": 9.622466783271992e-06, "loss": 0.9632, "step": 10367 }, { "epoch": 0.53, "learning_rate": 9.620822545485152e-06, "loss": 0.9831, "step": 10368 }, { "epoch": 0.53, "learning_rate": 9.61917831796414e-06, "loss": 0.9613, "step": 10369 }, { "epoch": 0.53, "learning_rate": 9.617534100753463e-06, "loss": 0.8942, "step": 10370 }, { "epoch": 0.53, "learning_rate": 9.615889893897651e-06, "loss": 1.0109, "step": 10371 }, { "epoch": 0.53, "learning_rate": 9.61424569744121e-06, "loss": 0.8686, "step": 10372 }, { "epoch": 0.53, "learning_rate": 9.612601511428654e-06, "loss": 0.8472, "step": 10373 }, { "epoch": 0.53, "learning_rate": 9.610957335904502e-06, "loss": 0.9052, "step": 10374 }, { "epoch": 0.53, "learning_rate": 9.609313170913266e-06, "loss": 0.8794, "step": 10375 }, { "epoch": 0.53, "learning_rate": 9.607669016499456e-06, "loss": 0.9808, "step": 10376 }, { "epoch": 0.53, "learning_rate": 9.606024872707593e-06, "loss": 0.851, "step": 10377 }, { "epoch": 0.53, "learning_rate": 9.604380739582189e-06, "loss": 0.9154, "step": 10378 }, { "epoch": 0.53, "learning_rate": 9.602736617167754e-06, "loss": 0.9629, "step": 10379 }, { "epoch": 0.53, "learning_rate": 9.601092505508804e-06, "loss": 0.923, "step": 10380 }, { "epoch": 0.53, "learning_rate": 9.59944840464985e-06, "loss": 0.9544, "step": 10381 }, { "epoch": 0.53, "learning_rate": 9.5978043146354e-06, "loss": 0.9724, "step": 10382 }, { "epoch": 0.53, "learning_rate": 9.596160235509974e-06, "loss": 0.9049, "step": 10383 }, { "epoch": 0.53, "learning_rate": 9.59451616731808e-06, "loss": 1.0042, "step": 10384 }, { "epoch": 0.53, "learning_rate": 9.592872110104228e-06, "loss": 0.8717, "step": 10385 }, { "epoch": 0.53, "learning_rate": 9.591228063912933e-06, "loss": 0.9265, "step": 10386 }, { "epoch": 0.53, "learning_rate": 9.589584028788702e-06, "loss": 0.9604, "step": 10387 }, { "epoch": 0.53, "learning_rate": 9.587940004776042e-06, "loss": 0.8942, "step": 10388 }, { "epoch": 0.53, "learning_rate": 9.586295991919472e-06, "loss": 0.8436, "step": 10389 }, { "epoch": 0.53, "learning_rate": 9.5846519902635e-06, "loss": 0.8758, "step": 10390 }, { "epoch": 0.53, "learning_rate": 9.583007999852631e-06, "loss": 0.9178, "step": 10391 }, { "epoch": 0.53, "learning_rate": 9.581364020731378e-06, "loss": 0.949, "step": 10392 }, { "epoch": 0.53, "learning_rate": 9.579720052944249e-06, "loss": 0.9627, "step": 10393 }, { "epoch": 0.53, "learning_rate": 9.578076096535749e-06, "loss": 0.9352, "step": 10394 }, { "epoch": 0.53, "learning_rate": 9.576432151550394e-06, "loss": 0.8927, "step": 10395 }, { "epoch": 0.53, "learning_rate": 9.574788218032685e-06, "loss": 0.9916, "step": 10396 }, { "epoch": 0.53, "learning_rate": 9.573144296027134e-06, "loss": 0.8959, "step": 10397 }, { "epoch": 0.53, "learning_rate": 9.571500385578247e-06, "loss": 1.0146, "step": 10398 }, { "epoch": 0.53, "learning_rate": 9.569856486730532e-06, "loss": 0.9319, "step": 10399 }, { "epoch": 0.53, "learning_rate": 9.56821259952849e-06, "loss": 0.9038, "step": 10400 }, { "epoch": 0.53, "learning_rate": 9.566568724016638e-06, "loss": 0.9124, "step": 10401 }, { "epoch": 0.53, "learning_rate": 9.564924860239476e-06, "loss": 0.8831, "step": 10402 }, { "epoch": 0.53, "learning_rate": 9.563281008241509e-06, "loss": 0.875, "step": 10403 }, { "epoch": 0.53, "learning_rate": 9.561637168067245e-06, "loss": 0.877, "step": 10404 }, { "epoch": 0.53, "learning_rate": 9.55999333976119e-06, "loss": 0.8849, "step": 10405 }, { "epoch": 0.53, "learning_rate": 9.558349523367842e-06, "loss": 0.9586, "step": 10406 }, { "epoch": 0.53, "learning_rate": 9.556705718931714e-06, "loss": 0.9501, "step": 10407 }, { "epoch": 0.53, "learning_rate": 9.555061926497309e-06, "loss": 0.9463, "step": 10408 }, { "epoch": 0.53, "learning_rate": 9.553418146109126e-06, "loss": 0.8643, "step": 10409 }, { "epoch": 0.53, "learning_rate": 9.551774377811673e-06, "loss": 0.9564, "step": 10410 }, { "epoch": 0.53, "learning_rate": 9.550130621649453e-06, "loss": 0.9159, "step": 10411 }, { "epoch": 0.53, "learning_rate": 9.548486877666963e-06, "loss": 0.9441, "step": 10412 }, { "epoch": 0.53, "learning_rate": 9.546843145908714e-06, "loss": 0.9681, "step": 10413 }, { "epoch": 0.53, "learning_rate": 9.545199426419205e-06, "loss": 0.9753, "step": 10414 }, { "epoch": 0.53, "learning_rate": 9.543555719242936e-06, "loss": 0.8715, "step": 10415 }, { "epoch": 0.53, "learning_rate": 9.541912024424412e-06, "loss": 0.8633, "step": 10416 }, { "epoch": 0.53, "learning_rate": 9.54026834200813e-06, "loss": 0.9663, "step": 10417 }, { "epoch": 0.53, "learning_rate": 9.538624672038594e-06, "loss": 0.9696, "step": 10418 }, { "epoch": 0.53, "learning_rate": 9.536981014560307e-06, "loss": 0.9089, "step": 10419 }, { "epoch": 0.53, "learning_rate": 9.535337369617764e-06, "loss": 0.9321, "step": 10420 }, { "epoch": 0.53, "learning_rate": 9.533693737255469e-06, "loss": 0.9861, "step": 10421 }, { "epoch": 0.53, "learning_rate": 9.53205011751792e-06, "loss": 0.9011, "step": 10422 }, { "epoch": 0.53, "learning_rate": 9.530406510449612e-06, "loss": 0.8729, "step": 10423 }, { "epoch": 0.53, "learning_rate": 9.528762916095054e-06, "loss": 0.8568, "step": 10424 }, { "epoch": 0.53, "learning_rate": 9.527119334498736e-06, "loss": 0.8418, "step": 10425 }, { "epoch": 0.53, "learning_rate": 9.525475765705157e-06, "loss": 0.9499, "step": 10426 }, { "epoch": 0.53, "learning_rate": 9.523832209758822e-06, "loss": 0.8576, "step": 10427 }, { "epoch": 0.53, "learning_rate": 9.522188666704221e-06, "loss": 0.8428, "step": 10428 }, { "epoch": 0.53, "learning_rate": 9.52054513658585e-06, "loss": 0.9155, "step": 10429 }, { "epoch": 0.53, "learning_rate": 9.518901619448212e-06, "loss": 0.8602, "step": 10430 }, { "epoch": 0.53, "learning_rate": 9.5172581153358e-06, "loss": 1.0339, "step": 10431 }, { "epoch": 0.53, "learning_rate": 9.515614624293112e-06, "loss": 0.9567, "step": 10432 }, { "epoch": 0.53, "learning_rate": 9.513971146364645e-06, "loss": 0.9701, "step": 10433 }, { "epoch": 0.53, "learning_rate": 9.51232768159489e-06, "loss": 0.8043, "step": 10434 }, { "epoch": 0.53, "learning_rate": 9.51068423002834e-06, "loss": 1.0101, "step": 10435 }, { "epoch": 0.53, "learning_rate": 9.509040791709498e-06, "loss": 0.8533, "step": 10436 }, { "epoch": 0.53, "learning_rate": 9.507397366682855e-06, "loss": 0.9012, "step": 10437 }, { "epoch": 0.53, "learning_rate": 9.505753954992901e-06, "loss": 0.9108, "step": 10438 }, { "epoch": 0.53, "learning_rate": 9.504110556684137e-06, "loss": 1.0181, "step": 10439 }, { "epoch": 0.53, "learning_rate": 9.502467171801052e-06, "loss": 0.9164, "step": 10440 }, { "epoch": 0.53, "learning_rate": 9.500823800388132e-06, "loss": 0.9798, "step": 10441 }, { "epoch": 0.53, "learning_rate": 9.499180442489884e-06, "loss": 0.9, "step": 10442 }, { "epoch": 0.53, "learning_rate": 9.497537098150791e-06, "loss": 0.8945, "step": 10443 }, { "epoch": 0.53, "learning_rate": 9.495893767415344e-06, "loss": 0.9145, "step": 10444 }, { "epoch": 0.53, "learning_rate": 9.49425045032804e-06, "loss": 0.8638, "step": 10445 }, { "epoch": 0.53, "learning_rate": 9.492607146933367e-06, "loss": 0.9246, "step": 10446 }, { "epoch": 0.53, "learning_rate": 9.490963857275812e-06, "loss": 0.895, "step": 10447 }, { "epoch": 0.53, "learning_rate": 9.489320581399871e-06, "loss": 0.942, "step": 10448 }, { "epoch": 0.53, "learning_rate": 9.487677319350033e-06, "loss": 0.9551, "step": 10449 }, { "epoch": 0.53, "learning_rate": 9.486034071170785e-06, "loss": 0.9591, "step": 10450 }, { "epoch": 0.53, "learning_rate": 9.484390836906619e-06, "loss": 0.8815, "step": 10451 }, { "epoch": 0.53, "learning_rate": 9.482747616602023e-06, "loss": 0.8994, "step": 10452 }, { "epoch": 0.53, "learning_rate": 9.48110441030148e-06, "loss": 0.9211, "step": 10453 }, { "epoch": 0.53, "learning_rate": 9.479461218049489e-06, "loss": 1.0575, "step": 10454 }, { "epoch": 0.53, "learning_rate": 9.47781803989053e-06, "loss": 0.8768, "step": 10455 }, { "epoch": 0.53, "learning_rate": 9.476174875869091e-06, "loss": 0.9899, "step": 10456 }, { "epoch": 0.53, "learning_rate": 9.474531726029663e-06, "loss": 0.8329, "step": 10457 }, { "epoch": 0.53, "learning_rate": 9.472888590416727e-06, "loss": 0.9269, "step": 10458 }, { "epoch": 0.53, "learning_rate": 9.471245469074767e-06, "loss": 0.9584, "step": 10459 }, { "epoch": 0.53, "learning_rate": 9.46960236204828e-06, "loss": 0.9448, "step": 10460 }, { "epoch": 0.53, "learning_rate": 9.467959269381744e-06, "loss": 0.902, "step": 10461 }, { "epoch": 0.53, "learning_rate": 9.466316191119642e-06, "loss": 0.8568, "step": 10462 }, { "epoch": 0.53, "learning_rate": 9.464673127306464e-06, "loss": 0.8805, "step": 10463 }, { "epoch": 0.53, "learning_rate": 9.46303007798669e-06, "loss": 0.8376, "step": 10464 }, { "epoch": 0.53, "learning_rate": 9.461387043204804e-06, "loss": 0.8981, "step": 10465 }, { "epoch": 0.53, "learning_rate": 9.459744023005293e-06, "loss": 0.9813, "step": 10466 }, { "epoch": 0.53, "learning_rate": 9.458101017432639e-06, "loss": 1.0006, "step": 10467 }, { "epoch": 0.53, "learning_rate": 9.456458026531322e-06, "loss": 0.8822, "step": 10468 }, { "epoch": 0.53, "learning_rate": 9.454815050345825e-06, "loss": 1.0121, "step": 10469 }, { "epoch": 0.53, "learning_rate": 9.453172088920632e-06, "loss": 0.8937, "step": 10470 }, { "epoch": 0.53, "learning_rate": 9.451529142300219e-06, "loss": 0.9025, "step": 10471 }, { "epoch": 0.53, "learning_rate": 9.449886210529074e-06, "loss": 0.8533, "step": 10472 }, { "epoch": 0.53, "learning_rate": 9.448243293651676e-06, "loss": 0.906, "step": 10473 }, { "epoch": 0.53, "learning_rate": 9.446600391712501e-06, "loss": 0.8961, "step": 10474 }, { "epoch": 0.53, "learning_rate": 9.444957504756033e-06, "loss": 0.9241, "step": 10475 }, { "epoch": 0.53, "learning_rate": 9.443314632826751e-06, "loss": 0.9888, "step": 10476 }, { "epoch": 0.53, "learning_rate": 9.441671775969128e-06, "loss": 0.8706, "step": 10477 }, { "epoch": 0.53, "learning_rate": 9.440028934227653e-06, "loss": 0.8954, "step": 10478 }, { "epoch": 0.53, "learning_rate": 9.438386107646798e-06, "loss": 0.8374, "step": 10479 }, { "epoch": 0.53, "learning_rate": 9.43674329627104e-06, "loss": 0.9851, "step": 10480 }, { "epoch": 0.53, "learning_rate": 9.43510050014486e-06, "loss": 0.8069, "step": 10481 }, { "epoch": 0.53, "learning_rate": 9.433457719312733e-06, "loss": 0.8702, "step": 10482 }, { "epoch": 0.53, "learning_rate": 9.431814953819131e-06, "loss": 0.9642, "step": 10483 }, { "epoch": 0.53, "learning_rate": 9.430172203708539e-06, "loss": 0.9688, "step": 10484 }, { "epoch": 0.53, "learning_rate": 9.428529469025425e-06, "loss": 0.9446, "step": 10485 }, { "epoch": 0.53, "learning_rate": 9.426886749814273e-06, "loss": 0.9459, "step": 10486 }, { "epoch": 0.53, "learning_rate": 9.42524404611955e-06, "loss": 0.775, "step": 10487 }, { "epoch": 0.53, "learning_rate": 9.423601357985729e-06, "loss": 0.8892, "step": 10488 }, { "epoch": 0.53, "learning_rate": 9.421958685457292e-06, "loss": 0.9144, "step": 10489 }, { "epoch": 0.53, "learning_rate": 9.42031602857871e-06, "loss": 0.8762, "step": 10490 }, { "epoch": 0.53, "learning_rate": 9.41867338739445e-06, "loss": 0.9177, "step": 10491 }, { "epoch": 0.53, "learning_rate": 9.417030761948994e-06, "loss": 0.8455, "step": 10492 }, { "epoch": 0.53, "learning_rate": 9.41538815228681e-06, "loss": 0.895, "step": 10493 }, { "epoch": 0.53, "learning_rate": 9.413745558452364e-06, "loss": 0.8972, "step": 10494 }, { "epoch": 0.53, "learning_rate": 9.412102980490138e-06, "loss": 0.9091, "step": 10495 }, { "epoch": 0.53, "learning_rate": 9.4104604184446e-06, "loss": 0.926, "step": 10496 }, { "epoch": 0.53, "learning_rate": 9.408817872360214e-06, "loss": 0.8576, "step": 10497 }, { "epoch": 0.53, "learning_rate": 9.407175342281459e-06, "loss": 0.9596, "step": 10498 }, { "epoch": 0.53, "learning_rate": 9.405532828252801e-06, "loss": 0.9479, "step": 10499 }, { "epoch": 0.53, "learning_rate": 9.403890330318705e-06, "loss": 0.8589, "step": 10500 }, { "epoch": 0.53, "learning_rate": 9.402247848523647e-06, "loss": 0.9851, "step": 10501 }, { "epoch": 0.53, "learning_rate": 9.400605382912094e-06, "loss": 0.7876, "step": 10502 }, { "epoch": 0.53, "learning_rate": 9.39896293352851e-06, "loss": 0.9469, "step": 10503 }, { "epoch": 0.53, "learning_rate": 9.397320500417367e-06, "loss": 0.9707, "step": 10504 }, { "epoch": 0.53, "learning_rate": 9.39567808362313e-06, "loss": 0.9044, "step": 10505 }, { "epoch": 0.53, "learning_rate": 9.394035683190262e-06, "loss": 0.9146, "step": 10506 }, { "epoch": 0.53, "learning_rate": 9.392393299163239e-06, "loss": 0.8165, "step": 10507 }, { "epoch": 0.53, "learning_rate": 9.390750931586519e-06, "loss": 0.922, "step": 10508 }, { "epoch": 0.53, "learning_rate": 9.389108580504567e-06, "loss": 0.9758, "step": 10509 }, { "epoch": 0.53, "learning_rate": 9.387466245961854e-06, "loss": 0.9068, "step": 10510 }, { "epoch": 0.53, "learning_rate": 9.38582392800284e-06, "loss": 0.8536, "step": 10511 }, { "epoch": 0.53, "learning_rate": 9.384181626671985e-06, "loss": 0.9409, "step": 10512 }, { "epoch": 0.53, "learning_rate": 9.382539342013761e-06, "loss": 0.9518, "step": 10513 }, { "epoch": 0.53, "learning_rate": 9.380897074072628e-06, "loss": 0.8744, "step": 10514 }, { "epoch": 0.53, "learning_rate": 9.379254822893046e-06, "loss": 0.9937, "step": 10515 }, { "epoch": 0.53, "learning_rate": 9.377612588519482e-06, "loss": 0.8936, "step": 10516 }, { "epoch": 0.53, "learning_rate": 9.375970370996395e-06, "loss": 0.8645, "step": 10517 }, { "epoch": 0.53, "learning_rate": 9.37432817036824e-06, "loss": 0.9682, "step": 10518 }, { "epoch": 0.53, "learning_rate": 9.372685986679491e-06, "loss": 1.0374, "step": 10519 }, { "epoch": 0.53, "learning_rate": 9.3710438199746e-06, "loss": 0.9402, "step": 10520 }, { "epoch": 0.53, "learning_rate": 9.369401670298027e-06, "loss": 0.8376, "step": 10521 }, { "epoch": 0.53, "learning_rate": 9.367759537694235e-06, "loss": 0.8624, "step": 10522 }, { "epoch": 0.53, "learning_rate": 9.36611742220768e-06, "loss": 0.8483, "step": 10523 }, { "epoch": 0.53, "learning_rate": 9.364475323882819e-06, "loss": 0.9896, "step": 10524 }, { "epoch": 0.53, "learning_rate": 9.362833242764115e-06, "loss": 0.8625, "step": 10525 }, { "epoch": 0.53, "learning_rate": 9.361191178896023e-06, "loss": 0.94, "step": 10526 }, { "epoch": 0.53, "learning_rate": 9.359549132322999e-06, "loss": 0.8806, "step": 10527 }, { "epoch": 0.53, "learning_rate": 9.357907103089502e-06, "loss": 0.9683, "step": 10528 }, { "epoch": 0.53, "learning_rate": 9.356265091239988e-06, "loss": 0.9682, "step": 10529 }, { "epoch": 0.53, "learning_rate": 9.354623096818906e-06, "loss": 0.9192, "step": 10530 }, { "epoch": 0.54, "learning_rate": 9.352981119870723e-06, "loss": 0.9269, "step": 10531 }, { "epoch": 0.54, "learning_rate": 9.351339160439887e-06, "loss": 0.9599, "step": 10532 }, { "epoch": 0.54, "learning_rate": 9.349697218570851e-06, "loss": 0.9379, "step": 10533 }, { "epoch": 0.54, "learning_rate": 9.348055294308074e-06, "loss": 0.9641, "step": 10534 }, { "epoch": 0.54, "learning_rate": 9.346413387696006e-06, "loss": 1.013, "step": 10535 }, { "epoch": 0.54, "learning_rate": 9.344771498779095e-06, "loss": 1.0207, "step": 10536 }, { "epoch": 0.54, "learning_rate": 9.343129627601803e-06, "loss": 0.8903, "step": 10537 }, { "epoch": 0.54, "learning_rate": 9.341487774208578e-06, "loss": 0.9053, "step": 10538 }, { "epoch": 0.54, "learning_rate": 9.339845938643868e-06, "loss": 0.7821, "step": 10539 }, { "epoch": 0.54, "learning_rate": 9.33820412095213e-06, "loss": 0.983, "step": 10540 }, { "epoch": 0.54, "learning_rate": 9.33656232117781e-06, "loss": 0.912, "step": 10541 }, { "epoch": 0.54, "learning_rate": 9.334920539365357e-06, "loss": 0.9438, "step": 10542 }, { "epoch": 0.54, "learning_rate": 9.333278775559225e-06, "loss": 0.9319, "step": 10543 }, { "epoch": 0.54, "learning_rate": 9.331637029803862e-06, "loss": 0.9099, "step": 10544 }, { "epoch": 0.54, "learning_rate": 9.329995302143713e-06, "loss": 0.852, "step": 10545 }, { "epoch": 0.54, "learning_rate": 9.32835359262323e-06, "loss": 0.9076, "step": 10546 }, { "epoch": 0.54, "learning_rate": 9.32671190128686e-06, "loss": 1.0117, "step": 10547 }, { "epoch": 0.54, "learning_rate": 9.325070228179044e-06, "loss": 0.8091, "step": 10548 }, { "epoch": 0.54, "learning_rate": 9.323428573344238e-06, "loss": 0.8223, "step": 10549 }, { "epoch": 0.54, "learning_rate": 9.321786936826882e-06, "loss": 0.9228, "step": 10550 }, { "epoch": 0.54, "learning_rate": 9.320145318671424e-06, "loss": 0.9044, "step": 10551 }, { "epoch": 0.54, "learning_rate": 9.318503718922308e-06, "loss": 0.8419, "step": 10552 }, { "epoch": 0.54, "learning_rate": 9.31686213762398e-06, "loss": 0.9319, "step": 10553 }, { "epoch": 0.54, "learning_rate": 9.315220574820878e-06, "loss": 0.8599, "step": 10554 }, { "epoch": 0.54, "learning_rate": 9.313579030557453e-06, "loss": 0.9693, "step": 10555 }, { "epoch": 0.54, "learning_rate": 9.311937504878146e-06, "loss": 0.8959, "step": 10556 }, { "epoch": 0.54, "learning_rate": 9.3102959978274e-06, "loss": 0.8886, "step": 10557 }, { "epoch": 0.54, "learning_rate": 9.308654509449655e-06, "loss": 0.8773, "step": 10558 }, { "epoch": 0.54, "learning_rate": 9.307013039789352e-06, "loss": 0.9675, "step": 10559 }, { "epoch": 0.54, "learning_rate": 9.305371588890937e-06, "loss": 0.8553, "step": 10560 }, { "epoch": 0.54, "learning_rate": 9.303730156798845e-06, "loss": 0.8845, "step": 10561 }, { "epoch": 0.54, "learning_rate": 9.30208874355752e-06, "loss": 1.0046, "step": 10562 }, { "epoch": 0.54, "learning_rate": 9.3004473492114e-06, "loss": 0.9432, "step": 10563 }, { "epoch": 0.54, "learning_rate": 9.298805973804921e-06, "loss": 0.9723, "step": 10564 }, { "epoch": 0.54, "learning_rate": 9.297164617382523e-06, "loss": 0.8442, "step": 10565 }, { "epoch": 0.54, "learning_rate": 9.295523279988649e-06, "loss": 0.9742, "step": 10566 }, { "epoch": 0.54, "learning_rate": 9.293881961667734e-06, "loss": 0.8841, "step": 10567 }, { "epoch": 0.54, "learning_rate": 9.292240662464209e-06, "loss": 0.9087, "step": 10568 }, { "epoch": 0.54, "learning_rate": 9.290599382422519e-06, "loss": 0.9508, "step": 10569 }, { "epoch": 0.54, "learning_rate": 9.288958121587094e-06, "loss": 0.829, "step": 10570 }, { "epoch": 0.54, "learning_rate": 9.28731688000237e-06, "loss": 0.9186, "step": 10571 }, { "epoch": 0.54, "learning_rate": 9.285675657712784e-06, "loss": 0.8806, "step": 10572 }, { "epoch": 0.54, "learning_rate": 9.284034454762773e-06, "loss": 0.985, "step": 10573 }, { "epoch": 0.54, "learning_rate": 9.282393271196763e-06, "loss": 0.8132, "step": 10574 }, { "epoch": 0.54, "learning_rate": 9.280752107059194e-06, "loss": 0.8032, "step": 10575 }, { "epoch": 0.54, "learning_rate": 9.279110962394498e-06, "loss": 1.0196, "step": 10576 }, { "epoch": 0.54, "learning_rate": 9.2774698372471e-06, "loss": 0.811, "step": 10577 }, { "epoch": 0.54, "learning_rate": 9.275828731661442e-06, "loss": 0.8302, "step": 10578 }, { "epoch": 0.54, "learning_rate": 9.274187645681951e-06, "loss": 0.8935, "step": 10579 }, { "epoch": 0.54, "learning_rate": 9.272546579353055e-06, "loss": 0.9307, "step": 10580 }, { "epoch": 0.54, "learning_rate": 9.270905532719188e-06, "loss": 0.9624, "step": 10581 }, { "epoch": 0.54, "learning_rate": 9.26926450582478e-06, "loss": 0.9918, "step": 10582 }, { "epoch": 0.54, "learning_rate": 9.267623498714252e-06, "loss": 0.9225, "step": 10583 }, { "epoch": 0.54, "learning_rate": 9.265982511432044e-06, "loss": 0.9493, "step": 10584 }, { "epoch": 0.54, "learning_rate": 9.264341544022576e-06, "loss": 0.9769, "step": 10585 }, { "epoch": 0.54, "learning_rate": 9.262700596530279e-06, "loss": 0.8963, "step": 10586 }, { "epoch": 0.54, "learning_rate": 9.26105966899958e-06, "loss": 0.9604, "step": 10587 }, { "epoch": 0.54, "learning_rate": 9.259418761474904e-06, "loss": 0.8371, "step": 10588 }, { "epoch": 0.54, "learning_rate": 9.257777874000672e-06, "loss": 0.9996, "step": 10589 }, { "epoch": 0.54, "learning_rate": 9.25613700662132e-06, "loss": 0.8826, "step": 10590 }, { "epoch": 0.54, "learning_rate": 9.254496159381267e-06, "loss": 0.8598, "step": 10591 }, { "epoch": 0.54, "learning_rate": 9.252855332324935e-06, "loss": 0.859, "step": 10592 }, { "epoch": 0.54, "learning_rate": 9.251214525496752e-06, "loss": 0.9182, "step": 10593 }, { "epoch": 0.54, "learning_rate": 9.249573738941138e-06, "loss": 1.0549, "step": 10594 }, { "epoch": 0.54, "learning_rate": 9.247932972702514e-06, "loss": 0.9114, "step": 10595 }, { "epoch": 0.54, "learning_rate": 9.24629222682531e-06, "loss": 0.9879, "step": 10596 }, { "epoch": 0.54, "learning_rate": 9.244651501353939e-06, "loss": 0.9046, "step": 10597 }, { "epoch": 0.54, "learning_rate": 9.243010796332825e-06, "loss": 0.8737, "step": 10598 }, { "epoch": 0.54, "learning_rate": 9.24137011180639e-06, "loss": 0.8749, "step": 10599 }, { "epoch": 0.54, "learning_rate": 9.23972944781905e-06, "loss": 0.9954, "step": 10600 }, { "epoch": 0.54, "learning_rate": 9.238088804415224e-06, "loss": 0.9134, "step": 10601 }, { "epoch": 0.54, "learning_rate": 9.236448181639337e-06, "loss": 0.894, "step": 10602 }, { "epoch": 0.54, "learning_rate": 9.234807579535803e-06, "loss": 0.8624, "step": 10603 }, { "epoch": 0.54, "learning_rate": 9.23316699814904e-06, "loss": 0.9339, "step": 10604 }, { "epoch": 0.54, "learning_rate": 9.231526437523464e-06, "loss": 1.004, "step": 10605 }, { "epoch": 0.54, "learning_rate": 9.229885897703492e-06, "loss": 1.066, "step": 10606 }, { "epoch": 0.54, "learning_rate": 9.228245378733537e-06, "loss": 0.8333, "step": 10607 }, { "epoch": 0.54, "learning_rate": 9.22660488065802e-06, "loss": 0.9125, "step": 10608 }, { "epoch": 0.54, "learning_rate": 9.224964403521355e-06, "loss": 0.8643, "step": 10609 }, { "epoch": 0.54, "learning_rate": 9.22332394736795e-06, "loss": 0.9941, "step": 10610 }, { "epoch": 0.54, "learning_rate": 9.221683512242226e-06, "loss": 0.9257, "step": 10611 }, { "epoch": 0.54, "learning_rate": 9.220043098188594e-06, "loss": 0.9434, "step": 10612 }, { "epoch": 0.54, "learning_rate": 9.21840270525146e-06, "loss": 0.933, "step": 10613 }, { "epoch": 0.54, "learning_rate": 9.216762333475246e-06, "loss": 0.9161, "step": 10614 }, { "epoch": 0.54, "learning_rate": 9.215121982904357e-06, "loss": 0.8844, "step": 10615 }, { "epoch": 0.54, "learning_rate": 9.213481653583204e-06, "loss": 0.9263, "step": 10616 }, { "epoch": 0.54, "learning_rate": 9.2118413455562e-06, "loss": 0.9898, "step": 10617 }, { "epoch": 0.54, "learning_rate": 9.210201058867755e-06, "loss": 0.9181, "step": 10618 }, { "epoch": 0.54, "learning_rate": 9.20856079356227e-06, "loss": 0.8906, "step": 10619 }, { "epoch": 0.54, "learning_rate": 9.206920549684164e-06, "loss": 0.9097, "step": 10620 }, { "epoch": 0.54, "learning_rate": 9.20528032727784e-06, "loss": 0.8404, "step": 10621 }, { "epoch": 0.54, "learning_rate": 9.203640126387704e-06, "loss": 0.9275, "step": 10622 }, { "epoch": 0.54, "learning_rate": 9.201999947058168e-06, "loss": 0.9128, "step": 10623 }, { "epoch": 0.54, "learning_rate": 9.200359789333629e-06, "loss": 0.8113, "step": 10624 }, { "epoch": 0.54, "learning_rate": 9.198719653258501e-06, "loss": 0.8501, "step": 10625 }, { "epoch": 0.54, "learning_rate": 9.197079538877187e-06, "loss": 0.9509, "step": 10626 }, { "epoch": 0.54, "learning_rate": 9.195439446234087e-06, "loss": 0.9115, "step": 10627 }, { "epoch": 0.54, "learning_rate": 9.193799375373612e-06, "loss": 0.8939, "step": 10628 }, { "epoch": 0.54, "learning_rate": 9.192159326340159e-06, "loss": 0.9559, "step": 10629 }, { "epoch": 0.54, "learning_rate": 9.19051929917813e-06, "loss": 0.9961, "step": 10630 }, { "epoch": 0.54, "learning_rate": 9.188879293931934e-06, "loss": 0.8452, "step": 10631 }, { "epoch": 0.54, "learning_rate": 9.187239310645968e-06, "loss": 0.9467, "step": 10632 }, { "epoch": 0.54, "learning_rate": 9.18559934936463e-06, "loss": 0.9809, "step": 10633 }, { "epoch": 0.54, "learning_rate": 9.183959410132326e-06, "loss": 0.8356, "step": 10634 }, { "epoch": 0.54, "learning_rate": 9.182319492993454e-06, "loss": 1.0077, "step": 10635 }, { "epoch": 0.54, "learning_rate": 9.180679597992405e-06, "loss": 0.8301, "step": 10636 }, { "epoch": 0.54, "learning_rate": 9.179039725173591e-06, "loss": 0.8743, "step": 10637 }, { "epoch": 0.54, "learning_rate": 9.177399874581401e-06, "loss": 0.8811, "step": 10638 }, { "epoch": 0.54, "learning_rate": 9.175760046260234e-06, "loss": 0.9263, "step": 10639 }, { "epoch": 0.54, "learning_rate": 9.174120240254487e-06, "loss": 0.851, "step": 10640 }, { "epoch": 0.54, "learning_rate": 9.172480456608556e-06, "loss": 0.9067, "step": 10641 }, { "epoch": 0.54, "learning_rate": 9.170840695366834e-06, "loss": 0.8721, "step": 10642 }, { "epoch": 0.54, "learning_rate": 9.16920095657372e-06, "loss": 0.8815, "step": 10643 }, { "epoch": 0.54, "learning_rate": 9.167561240273608e-06, "loss": 0.9033, "step": 10644 }, { "epoch": 0.54, "learning_rate": 9.165921546510886e-06, "loss": 0.872, "step": 10645 }, { "epoch": 0.54, "learning_rate": 9.164281875329954e-06, "loss": 0.808, "step": 10646 }, { "epoch": 0.54, "learning_rate": 9.1626422267752e-06, "loss": 0.9694, "step": 10647 }, { "epoch": 0.54, "learning_rate": 9.161002600891014e-06, "loss": 0.8803, "step": 10648 }, { "epoch": 0.54, "learning_rate": 9.159362997721794e-06, "loss": 0.8325, "step": 10649 }, { "epoch": 0.54, "learning_rate": 9.157723417311925e-06, "loss": 0.976, "step": 10650 }, { "epoch": 0.54, "learning_rate": 9.156083859705797e-06, "loss": 0.8714, "step": 10651 }, { "epoch": 0.54, "learning_rate": 9.154444324947804e-06, "loss": 0.8764, "step": 10652 }, { "epoch": 0.54, "learning_rate": 9.15280481308233e-06, "loss": 0.9713, "step": 10653 }, { "epoch": 0.54, "learning_rate": 9.15116532415376e-06, "loss": 0.881, "step": 10654 }, { "epoch": 0.54, "learning_rate": 9.14952585820649e-06, "loss": 0.8422, "step": 10655 }, { "epoch": 0.54, "learning_rate": 9.147886415284903e-06, "loss": 0.95, "step": 10656 }, { "epoch": 0.54, "learning_rate": 9.146246995433382e-06, "loss": 0.9253, "step": 10657 }, { "epoch": 0.54, "learning_rate": 9.144607598696319e-06, "loss": 0.8463, "step": 10658 }, { "epoch": 0.54, "learning_rate": 9.142968225118094e-06, "loss": 0.8762, "step": 10659 }, { "epoch": 0.54, "learning_rate": 9.141328874743088e-06, "loss": 0.9244, "step": 10660 }, { "epoch": 0.54, "learning_rate": 9.139689547615694e-06, "loss": 1.0004, "step": 10661 }, { "epoch": 0.54, "learning_rate": 9.13805024378029e-06, "loss": 0.9395, "step": 10662 }, { "epoch": 0.54, "learning_rate": 9.136410963281255e-06, "loss": 0.8418, "step": 10663 }, { "epoch": 0.54, "learning_rate": 9.134771706162979e-06, "loss": 0.9012, "step": 10664 }, { "epoch": 0.54, "learning_rate": 9.133132472469838e-06, "loss": 0.9144, "step": 10665 }, { "epoch": 0.54, "learning_rate": 9.131493262246208e-06, "loss": 0.8827, "step": 10666 }, { "epoch": 0.54, "learning_rate": 9.129854075536478e-06, "loss": 0.9166, "step": 10667 }, { "epoch": 0.54, "learning_rate": 9.128214912385024e-06, "loss": 0.9191, "step": 10668 }, { "epoch": 0.54, "learning_rate": 9.126575772836222e-06, "loss": 0.9409, "step": 10669 }, { "epoch": 0.54, "learning_rate": 9.124936656934452e-06, "loss": 1.0141, "step": 10670 }, { "epoch": 0.54, "learning_rate": 9.123297564724092e-06, "loss": 0.8889, "step": 10671 }, { "epoch": 0.54, "learning_rate": 9.121658496249514e-06, "loss": 0.9526, "step": 10672 }, { "epoch": 0.54, "learning_rate": 9.120019451555102e-06, "loss": 0.9302, "step": 10673 }, { "epoch": 0.54, "learning_rate": 9.118380430685226e-06, "loss": 0.9154, "step": 10674 }, { "epoch": 0.54, "learning_rate": 9.11674143368426e-06, "loss": 0.8531, "step": 10675 }, { "epoch": 0.54, "learning_rate": 9.115102460596582e-06, "loss": 0.9008, "step": 10676 }, { "epoch": 0.54, "learning_rate": 9.113463511466565e-06, "loss": 0.9217, "step": 10677 }, { "epoch": 0.54, "learning_rate": 9.111824586338573e-06, "loss": 0.9656, "step": 10678 }, { "epoch": 0.54, "learning_rate": 9.110185685256993e-06, "loss": 0.9481, "step": 10679 }, { "epoch": 0.54, "learning_rate": 9.108546808266186e-06, "loss": 0.8306, "step": 10680 }, { "epoch": 0.54, "learning_rate": 9.106907955410525e-06, "loss": 0.8992, "step": 10681 }, { "epoch": 0.54, "learning_rate": 9.105269126734383e-06, "loss": 1.0432, "step": 10682 }, { "epoch": 0.54, "learning_rate": 9.103630322282129e-06, "loss": 0.9551, "step": 10683 }, { "epoch": 0.54, "learning_rate": 9.101991542098122e-06, "loss": 0.8749, "step": 10684 }, { "epoch": 0.54, "learning_rate": 9.100352786226748e-06, "loss": 0.9067, "step": 10685 }, { "epoch": 0.54, "learning_rate": 9.098714054712362e-06, "loss": 0.9378, "step": 10686 }, { "epoch": 0.54, "learning_rate": 9.097075347599333e-06, "loss": 1.074, "step": 10687 }, { "epoch": 0.54, "learning_rate": 9.095436664932029e-06, "loss": 0.8947, "step": 10688 }, { "epoch": 0.54, "learning_rate": 9.093798006754815e-06, "loss": 0.9061, "step": 10689 }, { "epoch": 0.54, "learning_rate": 9.092159373112054e-06, "loss": 0.9616, "step": 10690 }, { "epoch": 0.54, "learning_rate": 9.090520764048115e-06, "loss": 0.852, "step": 10691 }, { "epoch": 0.54, "learning_rate": 9.088882179607358e-06, "loss": 0.8802, "step": 10692 }, { "epoch": 0.54, "learning_rate": 9.087243619834144e-06, "loss": 0.8555, "step": 10693 }, { "epoch": 0.54, "learning_rate": 9.085605084772843e-06, "loss": 0.9183, "step": 10694 }, { "epoch": 0.54, "learning_rate": 9.083966574467804e-06, "loss": 0.9558, "step": 10695 }, { "epoch": 0.54, "learning_rate": 9.0823280889634e-06, "loss": 0.9656, "step": 10696 }, { "epoch": 0.54, "learning_rate": 9.08068962830399e-06, "loss": 0.8818, "step": 10697 }, { "epoch": 0.54, "learning_rate": 9.079051192533925e-06, "loss": 0.875, "step": 10698 }, { "epoch": 0.54, "learning_rate": 9.077412781697572e-06, "loss": 0.9009, "step": 10699 }, { "epoch": 0.54, "learning_rate": 9.075774395839287e-06, "loss": 0.9053, "step": 10700 }, { "epoch": 0.54, "learning_rate": 9.074136035003422e-06, "loss": 0.9565, "step": 10701 }, { "epoch": 0.54, "learning_rate": 9.072497699234343e-06, "loss": 0.8847, "step": 10702 }, { "epoch": 0.54, "learning_rate": 9.070859388576401e-06, "loss": 0.9511, "step": 10703 }, { "epoch": 0.54, "learning_rate": 9.069221103073952e-06, "loss": 0.897, "step": 10704 }, { "epoch": 0.54, "learning_rate": 9.067582842771354e-06, "loss": 0.9203, "step": 10705 }, { "epoch": 0.54, "learning_rate": 9.065944607712957e-06, "loss": 1.0019, "step": 10706 }, { "epoch": 0.54, "learning_rate": 9.064306397943113e-06, "loss": 0.9218, "step": 10707 }, { "epoch": 0.54, "learning_rate": 9.06266821350618e-06, "loss": 1.006, "step": 10708 }, { "epoch": 0.54, "learning_rate": 9.061030054446511e-06, "loss": 0.9217, "step": 10709 }, { "epoch": 0.54, "learning_rate": 9.059391920808448e-06, "loss": 0.9008, "step": 10710 }, { "epoch": 0.54, "learning_rate": 9.057753812636353e-06, "loss": 0.9104, "step": 10711 }, { "epoch": 0.54, "learning_rate": 9.056115729974572e-06, "loss": 0.9168, "step": 10712 }, { "epoch": 0.54, "learning_rate": 9.054477672867447e-06, "loss": 0.8023, "step": 10713 }, { "epoch": 0.54, "learning_rate": 9.052839641359339e-06, "loss": 0.9401, "step": 10714 }, { "epoch": 0.54, "learning_rate": 9.051201635494588e-06, "loss": 0.8885, "step": 10715 }, { "epoch": 0.54, "learning_rate": 9.049563655317542e-06, "loss": 0.8825, "step": 10716 }, { "epoch": 0.54, "learning_rate": 9.047925700872552e-06, "loss": 0.9081, "step": 10717 }, { "epoch": 0.54, "learning_rate": 9.04628777220396e-06, "loss": 0.8684, "step": 10718 }, { "epoch": 0.54, "learning_rate": 9.044649869356108e-06, "loss": 0.923, "step": 10719 }, { "epoch": 0.54, "learning_rate": 9.043011992373347e-06, "loss": 0.8862, "step": 10720 }, { "epoch": 0.54, "learning_rate": 9.04137414130002e-06, "loss": 0.873, "step": 10721 }, { "epoch": 0.54, "learning_rate": 9.039736316180465e-06, "loss": 0.9415, "step": 10722 }, { "epoch": 0.54, "learning_rate": 9.03809851705903e-06, "loss": 1.0167, "step": 10723 }, { "epoch": 0.54, "learning_rate": 9.036460743980055e-06, "loss": 0.8628, "step": 10724 }, { "epoch": 0.54, "learning_rate": 9.034822996987876e-06, "loss": 0.9373, "step": 10725 }, { "epoch": 0.54, "learning_rate": 9.03318527612684e-06, "loss": 0.9423, "step": 10726 }, { "epoch": 0.54, "learning_rate": 9.031547581441286e-06, "loss": 0.8864, "step": 10727 }, { "epoch": 0.55, "learning_rate": 9.029909912975547e-06, "loss": 0.8627, "step": 10728 }, { "epoch": 0.55, "learning_rate": 9.02827227077397e-06, "loss": 0.9594, "step": 10729 }, { "epoch": 0.55, "learning_rate": 9.026634654880884e-06, "loss": 0.9565, "step": 10730 }, { "epoch": 0.55, "learning_rate": 9.024997065340626e-06, "loss": 0.8931, "step": 10731 }, { "epoch": 0.55, "learning_rate": 9.02335950219754e-06, "loss": 0.9015, "step": 10732 }, { "epoch": 0.55, "learning_rate": 9.021721965495956e-06, "loss": 0.835, "step": 10733 }, { "epoch": 0.55, "learning_rate": 9.020084455280207e-06, "loss": 0.9813, "step": 10734 }, { "epoch": 0.55, "learning_rate": 9.018446971594632e-06, "loss": 1.0306, "step": 10735 }, { "epoch": 0.55, "learning_rate": 9.01680951448356e-06, "loss": 0.889, "step": 10736 }, { "epoch": 0.55, "learning_rate": 9.01517208399132e-06, "loss": 1.0538, "step": 10737 }, { "epoch": 0.55, "learning_rate": 9.013534680162253e-06, "loss": 0.8674, "step": 10738 }, { "epoch": 0.55, "learning_rate": 9.011897303040683e-06, "loss": 0.877, "step": 10739 }, { "epoch": 0.55, "learning_rate": 9.010259952670943e-06, "loss": 0.8559, "step": 10740 }, { "epoch": 0.55, "learning_rate": 9.00862262909736e-06, "loss": 0.9154, "step": 10741 }, { "epoch": 0.55, "learning_rate": 9.006985332364269e-06, "loss": 0.8729, "step": 10742 }, { "epoch": 0.55, "learning_rate": 9.005348062515988e-06, "loss": 0.9682, "step": 10743 }, { "epoch": 0.55, "learning_rate": 9.003710819596853e-06, "loss": 0.913, "step": 10744 }, { "epoch": 0.55, "learning_rate": 9.002073603651189e-06, "loss": 0.9993, "step": 10745 }, { "epoch": 0.55, "learning_rate": 9.000436414723316e-06, "loss": 0.9198, "step": 10746 }, { "epoch": 0.55, "learning_rate": 8.998799252857568e-06, "loss": 0.9124, "step": 10747 }, { "epoch": 0.55, "learning_rate": 8.997162118098263e-06, "loss": 0.8794, "step": 10748 }, { "epoch": 0.55, "learning_rate": 8.995525010489724e-06, "loss": 0.9251, "step": 10749 }, { "epoch": 0.55, "learning_rate": 8.993887930076281e-06, "loss": 0.8837, "step": 10750 }, { "epoch": 0.55, "learning_rate": 8.992250876902251e-06, "loss": 0.8719, "step": 10751 }, { "epoch": 0.55, "learning_rate": 8.990613851011956e-06, "loss": 0.938, "step": 10752 }, { "epoch": 0.55, "learning_rate": 8.988976852449717e-06, "loss": 0.8236, "step": 10753 }, { "epoch": 0.55, "learning_rate": 8.987339881259856e-06, "loss": 0.9141, "step": 10754 }, { "epoch": 0.55, "learning_rate": 8.985702937486686e-06, "loss": 0.9077, "step": 10755 }, { "epoch": 0.55, "learning_rate": 8.984066021174532e-06, "loss": 0.9251, "step": 10756 }, { "epoch": 0.55, "learning_rate": 8.982429132367712e-06, "loss": 0.848, "step": 10757 }, { "epoch": 0.55, "learning_rate": 8.980792271110537e-06, "loss": 0.888, "step": 10758 }, { "epoch": 0.55, "learning_rate": 8.979155437447329e-06, "loss": 0.9181, "step": 10759 }, { "epoch": 0.55, "learning_rate": 8.977518631422402e-06, "loss": 0.9129, "step": 10760 }, { "epoch": 0.55, "learning_rate": 8.975881853080066e-06, "loss": 0.9546, "step": 10761 }, { "epoch": 0.55, "learning_rate": 8.974245102464643e-06, "loss": 0.9087, "step": 10762 }, { "epoch": 0.55, "learning_rate": 8.972608379620441e-06, "loss": 0.8013, "step": 10763 }, { "epoch": 0.55, "learning_rate": 8.970971684591776e-06, "loss": 0.8513, "step": 10764 }, { "epoch": 0.55, "learning_rate": 8.969335017422958e-06, "loss": 0.9556, "step": 10765 }, { "epoch": 0.55, "learning_rate": 8.967698378158293e-06, "loss": 0.85, "step": 10766 }, { "epoch": 0.55, "learning_rate": 8.9660617668421e-06, "loss": 0.9957, "step": 10767 }, { "epoch": 0.55, "learning_rate": 8.964425183518685e-06, "loss": 0.9727, "step": 10768 }, { "epoch": 0.55, "learning_rate": 8.962788628232355e-06, "loss": 0.9147, "step": 10769 }, { "epoch": 0.55, "learning_rate": 8.961152101027422e-06, "loss": 0.9459, "step": 10770 }, { "epoch": 0.55, "learning_rate": 8.95951560194819e-06, "loss": 0.9697, "step": 10771 }, { "epoch": 0.55, "learning_rate": 8.957879131038962e-06, "loss": 0.9268, "step": 10772 }, { "epoch": 0.55, "learning_rate": 8.956242688344051e-06, "loss": 0.7382, "step": 10773 }, { "epoch": 0.55, "learning_rate": 8.95460627390776e-06, "loss": 0.9521, "step": 10774 }, { "epoch": 0.55, "learning_rate": 8.952969887774389e-06, "loss": 0.9761, "step": 10775 }, { "epoch": 0.55, "learning_rate": 8.951333529988246e-06, "loss": 0.9006, "step": 10776 }, { "epoch": 0.55, "learning_rate": 8.949697200593633e-06, "loss": 0.9034, "step": 10777 }, { "epoch": 0.55, "learning_rate": 8.948060899634846e-06, "loss": 0.9346, "step": 10778 }, { "epoch": 0.55, "learning_rate": 8.946424627156196e-06, "loss": 0.8092, "step": 10779 }, { "epoch": 0.55, "learning_rate": 8.944788383201978e-06, "loss": 0.8471, "step": 10780 }, { "epoch": 0.55, "learning_rate": 8.94315216781649e-06, "loss": 0.9819, "step": 10781 }, { "epoch": 0.55, "learning_rate": 8.941515981044034e-06, "loss": 0.9112, "step": 10782 }, { "epoch": 0.55, "learning_rate": 8.939879822928907e-06, "loss": 0.8781, "step": 10783 }, { "epoch": 0.55, "learning_rate": 8.938243693515401e-06, "loss": 0.8025, "step": 10784 }, { "epoch": 0.55, "learning_rate": 8.936607592847822e-06, "loss": 0.8496, "step": 10785 }, { "epoch": 0.55, "learning_rate": 8.934971520970461e-06, "loss": 0.9375, "step": 10786 }, { "epoch": 0.55, "learning_rate": 8.933335477927612e-06, "loss": 0.9052, "step": 10787 }, { "epoch": 0.55, "learning_rate": 8.931699463763571e-06, "loss": 0.9696, "step": 10788 }, { "epoch": 0.55, "learning_rate": 8.930063478522632e-06, "loss": 0.9011, "step": 10789 }, { "epoch": 0.55, "learning_rate": 8.92842752224908e-06, "loss": 0.9299, "step": 10790 }, { "epoch": 0.55, "learning_rate": 8.926791594987217e-06, "loss": 0.9462, "step": 10791 }, { "epoch": 0.55, "learning_rate": 8.92515569678133e-06, "loss": 0.9741, "step": 10792 }, { "epoch": 0.55, "learning_rate": 8.923519827675707e-06, "loss": 0.8349, "step": 10793 }, { "epoch": 0.55, "learning_rate": 8.921883987714642e-06, "loss": 0.9079, "step": 10794 }, { "epoch": 0.55, "learning_rate": 8.92024817694242e-06, "loss": 0.8949, "step": 10795 }, { "epoch": 0.55, "learning_rate": 8.918612395403327e-06, "loss": 0.9271, "step": 10796 }, { "epoch": 0.55, "learning_rate": 8.916976643141657e-06, "loss": 0.9606, "step": 10797 }, { "epoch": 0.55, "learning_rate": 8.915340920201691e-06, "loss": 0.8179, "step": 10798 }, { "epoch": 0.55, "learning_rate": 8.913705226627715e-06, "loss": 0.9076, "step": 10799 }, { "epoch": 0.55, "learning_rate": 8.912069562464015e-06, "loss": 0.9691, "step": 10800 }, { "epoch": 0.55, "learning_rate": 8.910433927754874e-06, "loss": 0.9843, "step": 10801 }, { "epoch": 0.55, "learning_rate": 8.908798322544573e-06, "loss": 0.9493, "step": 10802 }, { "epoch": 0.55, "learning_rate": 8.907162746877398e-06, "loss": 0.8997, "step": 10803 }, { "epoch": 0.55, "learning_rate": 8.90552720079763e-06, "loss": 0.9684, "step": 10804 }, { "epoch": 0.55, "learning_rate": 8.903891684349546e-06, "loss": 0.86, "step": 10805 }, { "epoch": 0.55, "learning_rate": 8.902256197577432e-06, "loss": 0.8868, "step": 10806 }, { "epoch": 0.55, "learning_rate": 8.900620740525565e-06, "loss": 0.9197, "step": 10807 }, { "epoch": 0.55, "learning_rate": 8.898985313238215e-06, "loss": 0.9095, "step": 10808 }, { "epoch": 0.55, "learning_rate": 8.89734991575967e-06, "loss": 1.0027, "step": 10809 }, { "epoch": 0.55, "learning_rate": 8.895714548134204e-06, "loss": 0.818, "step": 10810 }, { "epoch": 0.55, "learning_rate": 8.894079210406089e-06, "loss": 0.95, "step": 10811 }, { "epoch": 0.55, "learning_rate": 8.892443902619605e-06, "loss": 0.9245, "step": 10812 }, { "epoch": 0.55, "learning_rate": 8.890808624819025e-06, "loss": 0.988, "step": 10813 }, { "epoch": 0.55, "learning_rate": 8.889173377048616e-06, "loss": 0.9081, "step": 10814 }, { "epoch": 0.55, "learning_rate": 8.88753815935266e-06, "loss": 0.9286, "step": 10815 }, { "epoch": 0.55, "learning_rate": 8.885902971775426e-06, "loss": 0.9089, "step": 10816 }, { "epoch": 0.55, "learning_rate": 8.884267814361181e-06, "loss": 0.99, "step": 10817 }, { "epoch": 0.55, "learning_rate": 8.8826326871542e-06, "loss": 0.8708, "step": 10818 }, { "epoch": 0.55, "learning_rate": 8.880997590198751e-06, "loss": 0.8351, "step": 10819 }, { "epoch": 0.55, "learning_rate": 8.879362523539097e-06, "loss": 0.9698, "step": 10820 }, { "epoch": 0.55, "learning_rate": 8.877727487219515e-06, "loss": 0.953, "step": 10821 }, { "epoch": 0.55, "learning_rate": 8.876092481284266e-06, "loss": 0.888, "step": 10822 }, { "epoch": 0.55, "learning_rate": 8.874457505777615e-06, "loss": 0.8362, "step": 10823 }, { "epoch": 0.55, "learning_rate": 8.872822560743834e-06, "loss": 0.8013, "step": 10824 }, { "epoch": 0.55, "learning_rate": 8.87118764622718e-06, "loss": 0.884, "step": 10825 }, { "epoch": 0.55, "learning_rate": 8.869552762271916e-06, "loss": 0.8524, "step": 10826 }, { "epoch": 0.55, "learning_rate": 8.867917908922314e-06, "loss": 0.9165, "step": 10827 }, { "epoch": 0.55, "learning_rate": 8.86628308622263e-06, "loss": 0.8906, "step": 10828 }, { "epoch": 0.55, "learning_rate": 8.864648294217122e-06, "loss": 1.0001, "step": 10829 }, { "epoch": 0.55, "learning_rate": 8.863013532950056e-06, "loss": 0.8732, "step": 10830 }, { "epoch": 0.55, "learning_rate": 8.86137880246569e-06, "loss": 0.9054, "step": 10831 }, { "epoch": 0.55, "learning_rate": 8.859744102808276e-06, "loss": 0.9553, "step": 10832 }, { "epoch": 0.55, "learning_rate": 8.858109434022081e-06, "loss": 0.9467, "step": 10833 }, { "epoch": 0.55, "learning_rate": 8.85647479615136e-06, "loss": 0.9135, "step": 10834 }, { "epoch": 0.55, "learning_rate": 8.854840189240365e-06, "loss": 0.8961, "step": 10835 }, { "epoch": 0.55, "learning_rate": 8.853205613333356e-06, "loss": 0.9582, "step": 10836 }, { "epoch": 0.55, "learning_rate": 8.851571068474578e-06, "loss": 0.9777, "step": 10837 }, { "epoch": 0.55, "learning_rate": 8.849936554708297e-06, "loss": 0.8909, "step": 10838 }, { "epoch": 0.55, "learning_rate": 8.848302072078762e-06, "loss": 0.8759, "step": 10839 }, { "epoch": 0.55, "learning_rate": 8.84666762063022e-06, "loss": 0.9635, "step": 10840 }, { "epoch": 0.55, "learning_rate": 8.845033200406926e-06, "loss": 0.9714, "step": 10841 }, { "epoch": 0.55, "learning_rate": 8.84339881145313e-06, "loss": 0.9519, "step": 10842 }, { "epoch": 0.55, "learning_rate": 8.841764453813076e-06, "loss": 0.8377, "step": 10843 }, { "epoch": 0.55, "learning_rate": 8.840130127531023e-06, "loss": 0.8451, "step": 10844 }, { "epoch": 0.55, "learning_rate": 8.838495832651212e-06, "loss": 0.9331, "step": 10845 }, { "epoch": 0.55, "learning_rate": 8.836861569217887e-06, "loss": 0.901, "step": 10846 }, { "epoch": 0.55, "learning_rate": 8.835227337275301e-06, "loss": 0.8346, "step": 10847 }, { "epoch": 0.55, "learning_rate": 8.833593136867696e-06, "loss": 0.8893, "step": 10848 }, { "epoch": 0.55, "learning_rate": 8.83195896803931e-06, "loss": 0.951, "step": 10849 }, { "epoch": 0.55, "learning_rate": 8.830324830834397e-06, "loss": 0.7893, "step": 10850 }, { "epoch": 0.55, "learning_rate": 8.828690725297196e-06, "loss": 0.8416, "step": 10851 }, { "epoch": 0.55, "learning_rate": 8.827056651471946e-06, "loss": 0.8509, "step": 10852 }, { "epoch": 0.55, "learning_rate": 8.825422609402889e-06, "loss": 0.9644, "step": 10853 }, { "epoch": 0.55, "learning_rate": 8.823788599134267e-06, "loss": 0.8758, "step": 10854 }, { "epoch": 0.55, "learning_rate": 8.822154620710312e-06, "loss": 0.9608, "step": 10855 }, { "epoch": 0.55, "learning_rate": 8.820520674175272e-06, "loss": 0.8483, "step": 10856 }, { "epoch": 0.55, "learning_rate": 8.81888675957338e-06, "loss": 0.9644, "step": 10857 }, { "epoch": 0.55, "learning_rate": 8.81725287694887e-06, "loss": 0.9259, "step": 10858 }, { "epoch": 0.55, "learning_rate": 8.815619026345982e-06, "loss": 0.8965, "step": 10859 }, { "epoch": 0.55, "learning_rate": 8.813985207808949e-06, "loss": 0.9437, "step": 10860 }, { "epoch": 0.55, "learning_rate": 8.812351421382001e-06, "loss": 0.9002, "step": 10861 }, { "epoch": 0.55, "learning_rate": 8.810717667109378e-06, "loss": 0.8907, "step": 10862 }, { "epoch": 0.55, "learning_rate": 8.80908394503531e-06, "loss": 0.9652, "step": 10863 }, { "epoch": 0.55, "learning_rate": 8.807450255204024e-06, "loss": 0.9672, "step": 10864 }, { "epoch": 0.55, "learning_rate": 8.805816597659756e-06, "loss": 0.9336, "step": 10865 }, { "epoch": 0.55, "learning_rate": 8.804182972446733e-06, "loss": 0.9854, "step": 10866 }, { "epoch": 0.55, "learning_rate": 8.80254937960918e-06, "loss": 0.8847, "step": 10867 }, { "epoch": 0.55, "learning_rate": 8.800915819191332e-06, "loss": 0.903, "step": 10868 }, { "epoch": 0.55, "learning_rate": 8.799282291237413e-06, "loss": 0.8889, "step": 10869 }, { "epoch": 0.55, "learning_rate": 8.797648795791647e-06, "loss": 0.8685, "step": 10870 }, { "epoch": 0.55, "learning_rate": 8.796015332898261e-06, "loss": 0.9573, "step": 10871 }, { "epoch": 0.55, "learning_rate": 8.794381902601481e-06, "loss": 0.9092, "step": 10872 }, { "epoch": 0.55, "learning_rate": 8.792748504945523e-06, "loss": 0.9853, "step": 10873 }, { "epoch": 0.55, "learning_rate": 8.79111513997462e-06, "loss": 0.9713, "step": 10874 }, { "epoch": 0.55, "learning_rate": 8.789481807732988e-06, "loss": 0.903, "step": 10875 }, { "epoch": 0.55, "learning_rate": 8.787848508264844e-06, "loss": 0.961, "step": 10876 }, { "epoch": 0.55, "learning_rate": 8.786215241614415e-06, "loss": 0.909, "step": 10877 }, { "epoch": 0.55, "learning_rate": 8.78458200782592e-06, "loss": 0.9106, "step": 10878 }, { "epoch": 0.55, "learning_rate": 8.782948806943567e-06, "loss": 0.9052, "step": 10879 }, { "epoch": 0.55, "learning_rate": 8.781315639011584e-06, "loss": 0.9412, "step": 10880 }, { "epoch": 0.55, "learning_rate": 8.779682504074185e-06, "loss": 0.9944, "step": 10881 }, { "epoch": 0.55, "learning_rate": 8.77804940217558e-06, "loss": 0.9097, "step": 10882 }, { "epoch": 0.55, "learning_rate": 8.776416333359991e-06, "loss": 0.8871, "step": 10883 }, { "epoch": 0.55, "learning_rate": 8.774783297671628e-06, "loss": 1.051, "step": 10884 }, { "epoch": 0.55, "learning_rate": 8.7731502951547e-06, "loss": 0.9444, "step": 10885 }, { "epoch": 0.55, "learning_rate": 8.771517325853425e-06, "loss": 0.9957, "step": 10886 }, { "epoch": 0.55, "learning_rate": 8.769884389812013e-06, "loss": 0.9159, "step": 10887 }, { "epoch": 0.55, "learning_rate": 8.76825148707467e-06, "loss": 0.8988, "step": 10888 }, { "epoch": 0.55, "learning_rate": 8.766618617685608e-06, "loss": 0.9625, "step": 10889 }, { "epoch": 0.55, "learning_rate": 8.764985781689037e-06, "loss": 0.8753, "step": 10890 }, { "epoch": 0.55, "learning_rate": 8.763352979129157e-06, "loss": 0.8997, "step": 10891 }, { "epoch": 0.55, "learning_rate": 8.761720210050182e-06, "loss": 0.9397, "step": 10892 }, { "epoch": 0.55, "learning_rate": 8.760087474496318e-06, "loss": 0.9984, "step": 10893 }, { "epoch": 0.55, "learning_rate": 8.758454772511761e-06, "loss": 0.9455, "step": 10894 }, { "epoch": 0.55, "learning_rate": 8.756822104140725e-06, "loss": 0.9293, "step": 10895 }, { "epoch": 0.55, "learning_rate": 8.755189469427405e-06, "loss": 0.9477, "step": 10896 }, { "epoch": 0.55, "learning_rate": 8.753556868416004e-06, "loss": 0.9199, "step": 10897 }, { "epoch": 0.55, "learning_rate": 8.751924301150729e-06, "loss": 0.8574, "step": 10898 }, { "epoch": 0.55, "learning_rate": 8.750291767675774e-06, "loss": 0.8378, "step": 10899 }, { "epoch": 0.55, "learning_rate": 8.748659268035339e-06, "loss": 0.875, "step": 10900 }, { "epoch": 0.55, "learning_rate": 8.747026802273625e-06, "loss": 0.9745, "step": 10901 }, { "epoch": 0.55, "learning_rate": 8.74539437043482e-06, "loss": 0.8994, "step": 10902 }, { "epoch": 0.55, "learning_rate": 8.743761972563136e-06, "loss": 0.8116, "step": 10903 }, { "epoch": 0.55, "learning_rate": 8.742129608702758e-06, "loss": 0.9551, "step": 10904 }, { "epoch": 0.55, "learning_rate": 8.740497278897882e-06, "loss": 0.8893, "step": 10905 }, { "epoch": 0.55, "learning_rate": 8.738864983192703e-06, "loss": 0.8265, "step": 10906 }, { "epoch": 0.55, "learning_rate": 8.737232721631413e-06, "loss": 0.9456, "step": 10907 }, { "epoch": 0.55, "learning_rate": 8.7356004942582e-06, "loss": 0.8827, "step": 10908 }, { "epoch": 0.55, "learning_rate": 8.733968301117262e-06, "loss": 0.8401, "step": 10909 }, { "epoch": 0.55, "learning_rate": 8.732336142252785e-06, "loss": 0.8985, "step": 10910 }, { "epoch": 0.55, "learning_rate": 8.730704017708957e-06, "loss": 0.8967, "step": 10911 }, { "epoch": 0.55, "learning_rate": 8.72907192752997e-06, "loss": 1.0352, "step": 10912 }, { "epoch": 0.55, "learning_rate": 8.727439871760006e-06, "loss": 0.9111, "step": 10913 }, { "epoch": 0.55, "learning_rate": 8.72580785044325e-06, "loss": 0.9255, "step": 10914 }, { "epoch": 0.55, "learning_rate": 8.724175863623897e-06, "loss": 0.9163, "step": 10915 }, { "epoch": 0.55, "learning_rate": 8.722543911346124e-06, "loss": 0.9312, "step": 10916 }, { "epoch": 0.55, "learning_rate": 8.720911993654113e-06, "loss": 0.9202, "step": 10917 }, { "epoch": 0.55, "learning_rate": 8.719280110592052e-06, "loss": 1.0235, "step": 10918 }, { "epoch": 0.55, "learning_rate": 8.71764826220412e-06, "loss": 0.7847, "step": 10919 }, { "epoch": 0.55, "learning_rate": 8.71601644853449e-06, "loss": 0.9503, "step": 10920 }, { "epoch": 0.55, "learning_rate": 8.714384669627353e-06, "loss": 0.81, "step": 10921 }, { "epoch": 0.55, "learning_rate": 8.712752925526887e-06, "loss": 0.8516, "step": 10922 }, { "epoch": 0.55, "learning_rate": 8.71112121627726e-06, "loss": 0.8677, "step": 10923 }, { "epoch": 0.55, "learning_rate": 8.709489541922659e-06, "loss": 0.8222, "step": 10924 }, { "epoch": 0.56, "learning_rate": 8.707857902507255e-06, "loss": 1.0053, "step": 10925 }, { "epoch": 0.56, "learning_rate": 8.70622629807522e-06, "loss": 0.8057, "step": 10926 }, { "epoch": 0.56, "learning_rate": 8.704594728670735e-06, "loss": 0.8558, "step": 10927 }, { "epoch": 0.56, "learning_rate": 8.70296319433797e-06, "loss": 0.8977, "step": 10928 }, { "epoch": 0.56, "learning_rate": 8.701331695121094e-06, "loss": 0.8577, "step": 10929 }, { "epoch": 0.56, "learning_rate": 8.699700231064283e-06, "loss": 0.8743, "step": 10930 }, { "epoch": 0.56, "learning_rate": 8.698068802211704e-06, "loss": 0.8971, "step": 10931 }, { "epoch": 0.56, "learning_rate": 8.696437408607525e-06, "loss": 0.9464, "step": 10932 }, { "epoch": 0.56, "learning_rate": 8.694806050295918e-06, "loss": 0.9356, "step": 10933 }, { "epoch": 0.56, "learning_rate": 8.69317472732105e-06, "loss": 0.827, "step": 10934 }, { "epoch": 0.56, "learning_rate": 8.691543439727084e-06, "loss": 0.9457, "step": 10935 }, { "epoch": 0.56, "learning_rate": 8.689912187558188e-06, "loss": 0.8896, "step": 10936 }, { "epoch": 0.56, "learning_rate": 8.688280970858527e-06, "loss": 1.0157, "step": 10937 }, { "epoch": 0.56, "learning_rate": 8.686649789672259e-06, "loss": 0.8739, "step": 10938 }, { "epoch": 0.56, "learning_rate": 8.685018644043553e-06, "loss": 0.9387, "step": 10939 }, { "epoch": 0.56, "learning_rate": 8.68338753401657e-06, "loss": 0.8083, "step": 10940 }, { "epoch": 0.56, "learning_rate": 8.681756459635466e-06, "loss": 0.9102, "step": 10941 }, { "epoch": 0.56, "learning_rate": 8.680125420944406e-06, "loss": 0.825, "step": 10942 }, { "epoch": 0.56, "learning_rate": 8.678494417987546e-06, "loss": 0.9017, "step": 10943 }, { "epoch": 0.56, "learning_rate": 8.67686345080904e-06, "loss": 0.9163, "step": 10944 }, { "epoch": 0.56, "learning_rate": 8.67523251945305e-06, "loss": 0.9493, "step": 10945 }, { "epoch": 0.56, "learning_rate": 8.673601623963733e-06, "loss": 0.8859, "step": 10946 }, { "epoch": 0.56, "learning_rate": 8.671970764385236e-06, "loss": 0.8467, "step": 10947 }, { "epoch": 0.56, "learning_rate": 8.670339940761723e-06, "loss": 0.9518, "step": 10948 }, { "epoch": 0.56, "learning_rate": 8.66870915313734e-06, "loss": 0.8964, "step": 10949 }, { "epoch": 0.56, "learning_rate": 8.667078401556235e-06, "loss": 0.8564, "step": 10950 }, { "epoch": 0.56, "learning_rate": 8.665447686062569e-06, "loss": 0.8828, "step": 10951 }, { "epoch": 0.56, "learning_rate": 8.663817006700488e-06, "loss": 0.9088, "step": 10952 }, { "epoch": 0.56, "learning_rate": 8.662186363514137e-06, "loss": 0.9177, "step": 10953 }, { "epoch": 0.56, "learning_rate": 8.660555756547668e-06, "loss": 0.9194, "step": 10954 }, { "epoch": 0.56, "learning_rate": 8.658925185845228e-06, "loss": 0.8585, "step": 10955 }, { "epoch": 0.56, "learning_rate": 8.657294651450959e-06, "loss": 0.8104, "step": 10956 }, { "epoch": 0.56, "learning_rate": 8.655664153409011e-06, "loss": 0.865, "step": 10957 }, { "epoch": 0.56, "learning_rate": 8.654033691763527e-06, "loss": 0.7664, "step": 10958 }, { "epoch": 0.56, "learning_rate": 8.652403266558645e-06, "loss": 0.8936, "step": 10959 }, { "epoch": 0.56, "learning_rate": 8.650772877838514e-06, "loss": 0.8886, "step": 10960 }, { "epoch": 0.56, "learning_rate": 8.649142525647271e-06, "loss": 0.8804, "step": 10961 }, { "epoch": 0.56, "learning_rate": 8.647512210029054e-06, "loss": 0.9805, "step": 10962 }, { "epoch": 0.56, "learning_rate": 8.645881931028008e-06, "loss": 0.9629, "step": 10963 }, { "epoch": 0.56, "learning_rate": 8.64425168868827e-06, "loss": 0.8363, "step": 10964 }, { "epoch": 0.56, "learning_rate": 8.64262148305397e-06, "loss": 0.9177, "step": 10965 }, { "epoch": 0.56, "learning_rate": 8.640991314169253e-06, "loss": 0.7996, "step": 10966 }, { "epoch": 0.56, "learning_rate": 8.639361182078252e-06, "loss": 0.909, "step": 10967 }, { "epoch": 0.56, "learning_rate": 8.637731086825094e-06, "loss": 0.8989, "step": 10968 }, { "epoch": 0.56, "learning_rate": 8.63610102845392e-06, "loss": 0.9313, "step": 10969 }, { "epoch": 0.56, "learning_rate": 8.634471007008861e-06, "loss": 0.9129, "step": 10970 }, { "epoch": 0.56, "learning_rate": 8.632841022534045e-06, "loss": 0.8765, "step": 10971 }, { "epoch": 0.56, "learning_rate": 8.631211075073604e-06, "loss": 0.9631, "step": 10972 }, { "epoch": 0.56, "learning_rate": 8.629581164671664e-06, "loss": 0.8892, "step": 10973 }, { "epoch": 0.56, "learning_rate": 8.627951291372361e-06, "loss": 0.8966, "step": 10974 }, { "epoch": 0.56, "learning_rate": 8.626321455219816e-06, "loss": 0.8783, "step": 10975 }, { "epoch": 0.56, "learning_rate": 8.624691656258154e-06, "loss": 0.9062, "step": 10976 }, { "epoch": 0.56, "learning_rate": 8.623061894531504e-06, "loss": 0.9167, "step": 10977 }, { "epoch": 0.56, "learning_rate": 8.62143217008399e-06, "loss": 0.971, "step": 10978 }, { "epoch": 0.56, "learning_rate": 8.619802482959727e-06, "loss": 0.9176, "step": 10979 }, { "epoch": 0.56, "learning_rate": 8.618172833202849e-06, "loss": 0.9162, "step": 10980 }, { "epoch": 0.56, "learning_rate": 8.61654322085747e-06, "loss": 0.8932, "step": 10981 }, { "epoch": 0.56, "learning_rate": 8.61491364596771e-06, "loss": 0.9118, "step": 10982 }, { "epoch": 0.56, "learning_rate": 8.613284108577691e-06, "loss": 0.9913, "step": 10983 }, { "epoch": 0.56, "learning_rate": 8.61165460873153e-06, "loss": 0.9465, "step": 10984 }, { "epoch": 0.56, "learning_rate": 8.610025146473339e-06, "loss": 0.814, "step": 10985 }, { "epoch": 0.56, "learning_rate": 8.608395721847244e-06, "loss": 0.9361, "step": 10986 }, { "epoch": 0.56, "learning_rate": 8.606766334897353e-06, "loss": 0.9309, "step": 10987 }, { "epoch": 0.56, "learning_rate": 8.605136985667779e-06, "loss": 0.8577, "step": 10988 }, { "epoch": 0.56, "learning_rate": 8.60350767420264e-06, "loss": 0.9121, "step": 10989 }, { "epoch": 0.56, "learning_rate": 8.601878400546044e-06, "loss": 0.9276, "step": 10990 }, { "epoch": 0.56, "learning_rate": 8.600249164742098e-06, "loss": 0.9167, "step": 10991 }, { "epoch": 0.56, "learning_rate": 8.598619966834922e-06, "loss": 0.9094, "step": 10992 }, { "epoch": 0.56, "learning_rate": 8.596990806868619e-06, "loss": 0.8186, "step": 10993 }, { "epoch": 0.56, "learning_rate": 8.595361684887296e-06, "loss": 0.8122, "step": 10994 }, { "epoch": 0.56, "learning_rate": 8.593732600935063e-06, "loss": 0.8953, "step": 10995 }, { "epoch": 0.56, "learning_rate": 8.592103555056023e-06, "loss": 0.8796, "step": 10996 }, { "epoch": 0.56, "learning_rate": 8.590474547294276e-06, "loss": 1.0166, "step": 10997 }, { "epoch": 0.56, "learning_rate": 8.588845577693937e-06, "loss": 0.8857, "step": 10998 }, { "epoch": 0.56, "learning_rate": 8.587216646299102e-06, "loss": 0.8889, "step": 10999 }, { "epoch": 0.56, "learning_rate": 8.58558775315387e-06, "loss": 0.8857, "step": 11000 }, { "epoch": 0.56, "learning_rate": 8.583958898302348e-06, "loss": 0.9262, "step": 11001 }, { "epoch": 0.56, "learning_rate": 8.582330081788634e-06, "loss": 0.8941, "step": 11002 }, { "epoch": 0.56, "learning_rate": 8.580701303656819e-06, "loss": 0.9319, "step": 11003 }, { "epoch": 0.56, "learning_rate": 8.57907256395101e-06, "loss": 0.8103, "step": 11004 }, { "epoch": 0.56, "learning_rate": 8.5774438627153e-06, "loss": 1.006, "step": 11005 }, { "epoch": 0.56, "learning_rate": 8.57581519999378e-06, "loss": 0.9422, "step": 11006 }, { "epoch": 0.56, "learning_rate": 8.574186575830553e-06, "loss": 0.9274, "step": 11007 }, { "epoch": 0.56, "learning_rate": 8.572557990269707e-06, "loss": 0.9432, "step": 11008 }, { "epoch": 0.56, "learning_rate": 8.570929443355331e-06, "loss": 1.0403, "step": 11009 }, { "epoch": 0.56, "learning_rate": 8.569300935131522e-06, "loss": 0.9139, "step": 11010 }, { "epoch": 0.56, "learning_rate": 8.567672465642369e-06, "loss": 0.8017, "step": 11011 }, { "epoch": 0.56, "learning_rate": 8.566044034931958e-06, "loss": 0.9677, "step": 11012 }, { "epoch": 0.56, "learning_rate": 8.564415643044383e-06, "loss": 0.8891, "step": 11013 }, { "epoch": 0.56, "learning_rate": 8.562787290023725e-06, "loss": 0.9525, "step": 11014 }, { "epoch": 0.56, "learning_rate": 8.561158975914067e-06, "loss": 0.8316, "step": 11015 }, { "epoch": 0.56, "learning_rate": 8.559530700759506e-06, "loss": 0.8126, "step": 11016 }, { "epoch": 0.56, "learning_rate": 8.557902464604118e-06, "loss": 0.8156, "step": 11017 }, { "epoch": 0.56, "learning_rate": 8.556274267491983e-06, "loss": 0.8375, "step": 11018 }, { "epoch": 0.56, "learning_rate": 8.55464610946719e-06, "loss": 0.9049, "step": 11019 }, { "epoch": 0.56, "learning_rate": 8.553017990573814e-06, "loss": 0.8334, "step": 11020 }, { "epoch": 0.56, "learning_rate": 8.551389910855934e-06, "loss": 0.9144, "step": 11021 }, { "epoch": 0.56, "learning_rate": 8.549761870357633e-06, "loss": 0.9761, "step": 11022 }, { "epoch": 0.56, "learning_rate": 8.548133869122988e-06, "loss": 0.9685, "step": 11023 }, { "epoch": 0.56, "learning_rate": 8.546505907196072e-06, "loss": 0.9369, "step": 11024 }, { "epoch": 0.56, "learning_rate": 8.544877984620963e-06, "loss": 0.8524, "step": 11025 }, { "epoch": 0.56, "learning_rate": 8.543250101441736e-06, "loss": 1.0204, "step": 11026 }, { "epoch": 0.56, "learning_rate": 8.541622257702457e-06, "loss": 0.8987, "step": 11027 }, { "epoch": 0.56, "learning_rate": 8.53999445344721e-06, "loss": 0.9568, "step": 11028 }, { "epoch": 0.56, "learning_rate": 8.538366688720056e-06, "loss": 0.8906, "step": 11029 }, { "epoch": 0.56, "learning_rate": 8.53673896356507e-06, "loss": 0.8912, "step": 11030 }, { "epoch": 0.56, "learning_rate": 8.535111278026321e-06, "loss": 0.9482, "step": 11031 }, { "epoch": 0.56, "learning_rate": 8.533483632147877e-06, "loss": 0.9023, "step": 11032 }, { "epoch": 0.56, "learning_rate": 8.531856025973798e-06, "loss": 0.8945, "step": 11033 }, { "epoch": 0.56, "learning_rate": 8.530228459548157e-06, "loss": 0.9415, "step": 11034 }, { "epoch": 0.56, "learning_rate": 8.52860093291502e-06, "loss": 0.8931, "step": 11035 }, { "epoch": 0.56, "learning_rate": 8.526973446118444e-06, "loss": 0.8608, "step": 11036 }, { "epoch": 0.56, "learning_rate": 8.525345999202497e-06, "loss": 0.886, "step": 11037 }, { "epoch": 0.56, "learning_rate": 8.523718592211239e-06, "loss": 0.9839, "step": 11038 }, { "epoch": 0.56, "learning_rate": 8.522091225188725e-06, "loss": 0.9559, "step": 11039 }, { "epoch": 0.56, "learning_rate": 8.52046389817902e-06, "loss": 0.9785, "step": 11040 }, { "epoch": 0.56, "learning_rate": 8.518836611226182e-06, "loss": 0.9068, "step": 11041 }, { "epoch": 0.56, "learning_rate": 8.517209364374268e-06, "loss": 0.8687, "step": 11042 }, { "epoch": 0.56, "learning_rate": 8.515582157667334e-06, "loss": 1.0239, "step": 11043 }, { "epoch": 0.56, "learning_rate": 8.513954991149428e-06, "loss": 0.8586, "step": 11044 }, { "epoch": 0.56, "learning_rate": 8.512327864864614e-06, "loss": 0.9225, "step": 11045 }, { "epoch": 0.56, "learning_rate": 8.510700778856942e-06, "loss": 0.9357, "step": 11046 }, { "epoch": 0.56, "learning_rate": 8.509073733170458e-06, "loss": 0.9873, "step": 11047 }, { "epoch": 0.56, "learning_rate": 8.50744672784922e-06, "loss": 0.8847, "step": 11048 }, { "epoch": 0.56, "learning_rate": 8.505819762937271e-06, "loss": 0.9819, "step": 11049 }, { "epoch": 0.56, "learning_rate": 8.50419283847866e-06, "loss": 0.8958, "step": 11050 }, { "epoch": 0.56, "learning_rate": 8.50256595451744e-06, "loss": 0.8344, "step": 11051 }, { "epoch": 0.56, "learning_rate": 8.500939111097656e-06, "loss": 0.9792, "step": 11052 }, { "epoch": 0.56, "learning_rate": 8.499312308263346e-06, "loss": 0.8901, "step": 11053 }, { "epoch": 0.56, "learning_rate": 8.497685546058562e-06, "loss": 0.9742, "step": 11054 }, { "epoch": 0.56, "learning_rate": 8.496058824527343e-06, "loss": 0.9009, "step": 11055 }, { "epoch": 0.56, "learning_rate": 8.494432143713727e-06, "loss": 0.9763, "step": 11056 }, { "epoch": 0.56, "learning_rate": 8.492805503661764e-06, "loss": 0.8723, "step": 11057 }, { "epoch": 0.56, "learning_rate": 8.491178904415488e-06, "loss": 0.9923, "step": 11058 }, { "epoch": 0.56, "learning_rate": 8.489552346018935e-06, "loss": 0.9172, "step": 11059 }, { "epoch": 0.56, "learning_rate": 8.487925828516149e-06, "loss": 0.8426, "step": 11060 }, { "epoch": 0.56, "learning_rate": 8.48629935195116e-06, "loss": 0.7931, "step": 11061 }, { "epoch": 0.56, "learning_rate": 8.484672916368003e-06, "loss": 0.9763, "step": 11062 }, { "epoch": 0.56, "learning_rate": 8.48304652181072e-06, "loss": 0.8425, "step": 11063 }, { "epoch": 0.56, "learning_rate": 8.481420168323337e-06, "loss": 0.922, "step": 11064 }, { "epoch": 0.56, "learning_rate": 8.479793855949886e-06, "loss": 0.9656, "step": 11065 }, { "epoch": 0.56, "learning_rate": 8.478167584734398e-06, "loss": 0.9033, "step": 11066 }, { "epoch": 0.56, "learning_rate": 8.476541354720908e-06, "loss": 0.9129, "step": 11067 }, { "epoch": 0.56, "learning_rate": 8.474915165953433e-06, "loss": 0.9162, "step": 11068 }, { "epoch": 0.56, "learning_rate": 8.473289018476012e-06, "loss": 0.8501, "step": 11069 }, { "epoch": 0.56, "learning_rate": 8.471662912332667e-06, "loss": 0.8699, "step": 11070 }, { "epoch": 0.56, "learning_rate": 8.47003684756742e-06, "loss": 0.9295, "step": 11071 }, { "epoch": 0.56, "learning_rate": 8.4684108242243e-06, "loss": 0.8526, "step": 11072 }, { "epoch": 0.56, "learning_rate": 8.466784842347328e-06, "loss": 0.8715, "step": 11073 }, { "epoch": 0.56, "learning_rate": 8.46515890198052e-06, "loss": 0.8989, "step": 11074 }, { "epoch": 0.56, "learning_rate": 8.463533003167906e-06, "loss": 0.8464, "step": 11075 }, { "epoch": 0.56, "learning_rate": 8.461907145953501e-06, "loss": 0.7972, "step": 11076 }, { "epoch": 0.56, "learning_rate": 8.460281330381324e-06, "loss": 0.9068, "step": 11077 }, { "epoch": 0.56, "learning_rate": 8.458655556495392e-06, "loss": 0.8513, "step": 11078 }, { "epoch": 0.56, "learning_rate": 8.457029824339721e-06, "loss": 0.8425, "step": 11079 }, { "epoch": 0.56, "learning_rate": 8.455404133958321e-06, "loss": 0.8875, "step": 11080 }, { "epoch": 0.56, "learning_rate": 8.453778485395217e-06, "loss": 0.8631, "step": 11081 }, { "epoch": 0.56, "learning_rate": 8.452152878694413e-06, "loss": 0.9027, "step": 11082 }, { "epoch": 0.56, "learning_rate": 8.450527313899923e-06, "loss": 0.9237, "step": 11083 }, { "epoch": 0.56, "learning_rate": 8.44890179105576e-06, "loss": 0.8756, "step": 11084 }, { "epoch": 0.56, "learning_rate": 8.447276310205928e-06, "loss": 0.8559, "step": 11085 }, { "epoch": 0.56, "learning_rate": 8.445650871394436e-06, "loss": 0.9556, "step": 11086 }, { "epoch": 0.56, "learning_rate": 8.444025474665296e-06, "loss": 0.8881, "step": 11087 }, { "epoch": 0.56, "learning_rate": 8.442400120062511e-06, "loss": 1.0316, "step": 11088 }, { "epoch": 0.56, "learning_rate": 8.440774807630084e-06, "loss": 0.7708, "step": 11089 }, { "epoch": 0.56, "learning_rate": 8.439149537412021e-06, "loss": 0.9244, "step": 11090 }, { "epoch": 0.56, "learning_rate": 8.437524309452326e-06, "loss": 0.9355, "step": 11091 }, { "epoch": 0.56, "learning_rate": 8.435899123794992e-06, "loss": 0.9146, "step": 11092 }, { "epoch": 0.56, "learning_rate": 8.43427398048403e-06, "loss": 0.9166, "step": 11093 }, { "epoch": 0.56, "learning_rate": 8.432648879563435e-06, "loss": 0.8605, "step": 11094 }, { "epoch": 0.56, "learning_rate": 8.4310238210772e-06, "loss": 0.9041, "step": 11095 }, { "epoch": 0.56, "learning_rate": 8.429398805069328e-06, "loss": 0.8644, "step": 11096 }, { "epoch": 0.56, "learning_rate": 8.427773831583814e-06, "loss": 0.8527, "step": 11097 }, { "epoch": 0.56, "learning_rate": 8.426148900664646e-06, "loss": 0.9473, "step": 11098 }, { "epoch": 0.56, "learning_rate": 8.424524012355827e-06, "loss": 0.9743, "step": 11099 }, { "epoch": 0.56, "learning_rate": 8.422899166701342e-06, "loss": 0.9337, "step": 11100 }, { "epoch": 0.56, "learning_rate": 8.421274363745183e-06, "loss": 0.9371, "step": 11101 }, { "epoch": 0.56, "learning_rate": 8.419649603531343e-06, "loss": 0.8207, "step": 11102 }, { "epoch": 0.56, "learning_rate": 8.41802488610381e-06, "loss": 0.8713, "step": 11103 }, { "epoch": 0.56, "learning_rate": 8.416400211506564e-06, "loss": 0.9446, "step": 11104 }, { "epoch": 0.56, "learning_rate": 8.414775579783603e-06, "loss": 0.8685, "step": 11105 }, { "epoch": 0.56, "learning_rate": 8.413150990978906e-06, "loss": 0.8509, "step": 11106 }, { "epoch": 0.56, "learning_rate": 8.411526445136456e-06, "loss": 0.8345, "step": 11107 }, { "epoch": 0.56, "learning_rate": 8.409901942300239e-06, "loss": 0.8879, "step": 11108 }, { "epoch": 0.56, "learning_rate": 8.408277482514235e-06, "loss": 0.8671, "step": 11109 }, { "epoch": 0.56, "learning_rate": 8.406653065822422e-06, "loss": 0.8762, "step": 11110 }, { "epoch": 0.56, "learning_rate": 8.405028692268786e-06, "loss": 0.9719, "step": 11111 }, { "epoch": 0.56, "learning_rate": 8.403404361897297e-06, "loss": 0.9174, "step": 11112 }, { "epoch": 0.56, "learning_rate": 8.40178007475194e-06, "loss": 0.8627, "step": 11113 }, { "epoch": 0.56, "learning_rate": 8.400155830876688e-06, "loss": 0.9005, "step": 11114 }, { "epoch": 0.56, "learning_rate": 8.398531630315508e-06, "loss": 0.8741, "step": 11115 }, { "epoch": 0.56, "learning_rate": 8.396907473112387e-06, "loss": 0.8039, "step": 11116 }, { "epoch": 0.56, "learning_rate": 8.395283359311289e-06, "loss": 0.9221, "step": 11117 }, { "epoch": 0.56, "learning_rate": 8.393659288956184e-06, "loss": 0.9511, "step": 11118 }, { "epoch": 0.56, "learning_rate": 8.39203526209105e-06, "loss": 0.9313, "step": 11119 }, { "epoch": 0.56, "learning_rate": 8.390411278759848e-06, "loss": 0.92, "step": 11120 }, { "epoch": 0.56, "learning_rate": 8.388787339006544e-06, "loss": 0.9092, "step": 11121 }, { "epoch": 0.57, "learning_rate": 8.387163442875116e-06, "loss": 1.0562, "step": 11122 }, { "epoch": 0.57, "learning_rate": 8.385539590409518e-06, "loss": 0.8553, "step": 11123 }, { "epoch": 0.57, "learning_rate": 8.383915781653717e-06, "loss": 0.9189, "step": 11124 }, { "epoch": 0.57, "learning_rate": 8.38229201665168e-06, "loss": 0.9012, "step": 11125 }, { "epoch": 0.57, "learning_rate": 8.380668295447362e-06, "loss": 1.0087, "step": 11126 }, { "epoch": 0.57, "learning_rate": 8.379044618084727e-06, "loss": 0.9661, "step": 11127 }, { "epoch": 0.57, "learning_rate": 8.377420984607735e-06, "loss": 0.9229, "step": 11128 }, { "epoch": 0.57, "learning_rate": 8.375797395060346e-06, "loss": 0.9289, "step": 11129 }, { "epoch": 0.57, "learning_rate": 8.374173849486513e-06, "loss": 0.9055, "step": 11130 }, { "epoch": 0.57, "learning_rate": 8.372550347930194e-06, "loss": 0.8831, "step": 11131 }, { "epoch": 0.57, "learning_rate": 8.370926890435343e-06, "loss": 0.9453, "step": 11132 }, { "epoch": 0.57, "learning_rate": 8.36930347704591e-06, "loss": 0.8503, "step": 11133 }, { "epoch": 0.57, "learning_rate": 8.367680107805854e-06, "loss": 0.8085, "step": 11134 }, { "epoch": 0.57, "learning_rate": 8.366056782759122e-06, "loss": 1.0461, "step": 11135 }, { "epoch": 0.57, "learning_rate": 8.364433501949662e-06, "loss": 0.911, "step": 11136 }, { "epoch": 0.57, "learning_rate": 8.362810265421428e-06, "loss": 0.8623, "step": 11137 }, { "epoch": 0.57, "learning_rate": 8.361187073218365e-06, "loss": 0.9423, "step": 11138 }, { "epoch": 0.57, "learning_rate": 8.35956392538441e-06, "loss": 0.9063, "step": 11139 }, { "epoch": 0.57, "learning_rate": 8.357940821963524e-06, "loss": 0.9301, "step": 11140 }, { "epoch": 0.57, "learning_rate": 8.356317762999644e-06, "loss": 0.885, "step": 11141 }, { "epoch": 0.57, "learning_rate": 8.354694748536709e-06, "loss": 0.9187, "step": 11142 }, { "epoch": 0.57, "learning_rate": 8.353071778618664e-06, "loss": 0.8461, "step": 11143 }, { "epoch": 0.57, "learning_rate": 8.351448853289448e-06, "loss": 0.9327, "step": 11144 }, { "epoch": 0.57, "learning_rate": 8.349825972592999e-06, "loss": 0.8551, "step": 11145 }, { "epoch": 0.57, "learning_rate": 8.348203136573259e-06, "loss": 0.8351, "step": 11146 }, { "epoch": 0.57, "learning_rate": 8.346580345274161e-06, "loss": 0.9664, "step": 11147 }, { "epoch": 0.57, "learning_rate": 8.344957598739638e-06, "loss": 1.0056, "step": 11148 }, { "epoch": 0.57, "learning_rate": 8.34333489701363e-06, "loss": 0.9371, "step": 11149 }, { "epoch": 0.57, "learning_rate": 8.34171224014007e-06, "loss": 0.8694, "step": 11150 }, { "epoch": 0.57, "learning_rate": 8.34008962816288e-06, "loss": 0.9448, "step": 11151 }, { "epoch": 0.57, "learning_rate": 8.338467061126002e-06, "loss": 0.9357, "step": 11152 }, { "epoch": 0.57, "learning_rate": 8.33684453907336e-06, "loss": 0.8589, "step": 11153 }, { "epoch": 0.57, "learning_rate": 8.335222062048881e-06, "loss": 0.9241, "step": 11154 }, { "epoch": 0.57, "learning_rate": 8.333599630096496e-06, "loss": 0.9073, "step": 11155 }, { "epoch": 0.57, "learning_rate": 8.331977243260127e-06, "loss": 0.8836, "step": 11156 }, { "epoch": 0.57, "learning_rate": 8.330354901583695e-06, "loss": 0.9641, "step": 11157 }, { "epoch": 0.57, "learning_rate": 8.328732605111133e-06, "loss": 0.9957, "step": 11158 }, { "epoch": 0.57, "learning_rate": 8.327110353886358e-06, "loss": 0.8806, "step": 11159 }, { "epoch": 0.57, "learning_rate": 8.325488147953289e-06, "loss": 0.9456, "step": 11160 }, { "epoch": 0.57, "learning_rate": 8.323865987355847e-06, "loss": 0.8393, "step": 11161 }, { "epoch": 0.57, "learning_rate": 8.32224387213795e-06, "loss": 0.9517, "step": 11162 }, { "epoch": 0.57, "learning_rate": 8.320621802343513e-06, "loss": 0.9317, "step": 11163 }, { "epoch": 0.57, "learning_rate": 8.318999778016457e-06, "loss": 0.8621, "step": 11164 }, { "epoch": 0.57, "learning_rate": 8.317377799200693e-06, "loss": 0.8461, "step": 11165 }, { "epoch": 0.57, "learning_rate": 8.315755865940133e-06, "loss": 0.8358, "step": 11166 }, { "epoch": 0.57, "learning_rate": 8.314133978278694e-06, "loss": 0.8718, "step": 11167 }, { "epoch": 0.57, "learning_rate": 8.312512136260283e-06, "loss": 0.85, "step": 11168 }, { "epoch": 0.57, "learning_rate": 8.310890339928807e-06, "loss": 0.9176, "step": 11169 }, { "epoch": 0.57, "learning_rate": 8.309268589328182e-06, "loss": 0.8509, "step": 11170 }, { "epoch": 0.57, "learning_rate": 8.307646884502311e-06, "loss": 0.8979, "step": 11171 }, { "epoch": 0.57, "learning_rate": 8.306025225495098e-06, "loss": 0.9943, "step": 11172 }, { "epoch": 0.57, "learning_rate": 8.304403612350453e-06, "loss": 0.9489, "step": 11173 }, { "epoch": 0.57, "learning_rate": 8.302782045112274e-06, "loss": 0.8069, "step": 11174 }, { "epoch": 0.57, "learning_rate": 8.301160523824463e-06, "loss": 0.9567, "step": 11175 }, { "epoch": 0.57, "learning_rate": 8.299539048530926e-06, "loss": 0.9405, "step": 11176 }, { "epoch": 0.57, "learning_rate": 8.297917619275561e-06, "loss": 0.8729, "step": 11177 }, { "epoch": 0.57, "learning_rate": 8.296296236102265e-06, "loss": 0.9125, "step": 11178 }, { "epoch": 0.57, "learning_rate": 8.294674899054937e-06, "loss": 0.903, "step": 11179 }, { "epoch": 0.57, "learning_rate": 8.293053608177466e-06, "loss": 0.9374, "step": 11180 }, { "epoch": 0.57, "learning_rate": 8.291432363513758e-06, "loss": 0.877, "step": 11181 }, { "epoch": 0.57, "learning_rate": 8.2898111651077e-06, "loss": 0.8814, "step": 11182 }, { "epoch": 0.57, "learning_rate": 8.288190013003185e-06, "loss": 1.005, "step": 11183 }, { "epoch": 0.57, "learning_rate": 8.286568907244105e-06, "loss": 0.8289, "step": 11184 }, { "epoch": 0.57, "learning_rate": 8.28494784787435e-06, "loss": 0.9064, "step": 11185 }, { "epoch": 0.57, "learning_rate": 8.283326834937802e-06, "loss": 0.9765, "step": 11186 }, { "epoch": 0.57, "learning_rate": 8.281705868478359e-06, "loss": 0.9217, "step": 11187 }, { "epoch": 0.57, "learning_rate": 8.280084948539899e-06, "loss": 0.8453, "step": 11188 }, { "epoch": 0.57, "learning_rate": 8.278464075166309e-06, "loss": 0.9236, "step": 11189 }, { "epoch": 0.57, "learning_rate": 8.276843248401474e-06, "loss": 0.8492, "step": 11190 }, { "epoch": 0.57, "learning_rate": 8.275222468289275e-06, "loss": 0.9116, "step": 11191 }, { "epoch": 0.57, "learning_rate": 8.273601734873587e-06, "loss": 0.8874, "step": 11192 }, { "epoch": 0.57, "learning_rate": 8.271981048198302e-06, "loss": 0.8997, "step": 11193 }, { "epoch": 0.57, "learning_rate": 8.270360408307291e-06, "loss": 0.9343, "step": 11194 }, { "epoch": 0.57, "learning_rate": 8.268739815244429e-06, "loss": 0.9257, "step": 11195 }, { "epoch": 0.57, "learning_rate": 8.267119269053596e-06, "loss": 0.8348, "step": 11196 }, { "epoch": 0.57, "learning_rate": 8.265498769778667e-06, "loss": 0.8728, "step": 11197 }, { "epoch": 0.57, "learning_rate": 8.263878317463507e-06, "loss": 0.8536, "step": 11198 }, { "epoch": 0.57, "learning_rate": 8.262257912152001e-06, "loss": 0.9489, "step": 11199 }, { "epoch": 0.57, "learning_rate": 8.260637553888012e-06, "loss": 0.9242, "step": 11200 }, { "epoch": 0.57, "learning_rate": 8.259017242715409e-06, "loss": 0.966, "step": 11201 }, { "epoch": 0.57, "learning_rate": 8.257396978678063e-06, "loss": 0.8044, "step": 11202 }, { "epoch": 0.57, "learning_rate": 8.255776761819841e-06, "loss": 0.8784, "step": 11203 }, { "epoch": 0.57, "learning_rate": 8.254156592184604e-06, "loss": 0.8723, "step": 11204 }, { "epoch": 0.57, "learning_rate": 8.25253646981622e-06, "loss": 0.863, "step": 11205 }, { "epoch": 0.57, "learning_rate": 8.250916394758556e-06, "loss": 0.9445, "step": 11206 }, { "epoch": 0.57, "learning_rate": 8.249296367055466e-06, "loss": 0.9447, "step": 11207 }, { "epoch": 0.57, "learning_rate": 8.247676386750819e-06, "loss": 1.0063, "step": 11208 }, { "epoch": 0.57, "learning_rate": 8.246056453888466e-06, "loss": 0.877, "step": 11209 }, { "epoch": 0.57, "learning_rate": 8.244436568512267e-06, "loss": 0.965, "step": 11210 }, { "epoch": 0.57, "learning_rate": 8.242816730666083e-06, "loss": 0.8812, "step": 11211 }, { "epoch": 0.57, "learning_rate": 8.241196940393765e-06, "loss": 0.8985, "step": 11212 }, { "epoch": 0.57, "learning_rate": 8.239577197739168e-06, "loss": 0.9003, "step": 11213 }, { "epoch": 0.57, "learning_rate": 8.237957502746147e-06, "loss": 1.0179, "step": 11214 }, { "epoch": 0.57, "learning_rate": 8.236337855458554e-06, "loss": 0.956, "step": 11215 }, { "epoch": 0.57, "learning_rate": 8.234718255920231e-06, "loss": 0.8704, "step": 11216 }, { "epoch": 0.57, "learning_rate": 8.233098704175037e-06, "loss": 0.8706, "step": 11217 }, { "epoch": 0.57, "learning_rate": 8.231479200266817e-06, "loss": 0.8507, "step": 11218 }, { "epoch": 0.57, "learning_rate": 8.229859744239414e-06, "loss": 0.8613, "step": 11219 }, { "epoch": 0.57, "learning_rate": 8.228240336136676e-06, "loss": 0.9366, "step": 11220 }, { "epoch": 0.57, "learning_rate": 8.226620976002446e-06, "loss": 0.8716, "step": 11221 }, { "epoch": 0.57, "learning_rate": 8.225001663880563e-06, "loss": 0.9424, "step": 11222 }, { "epoch": 0.57, "learning_rate": 8.223382399814875e-06, "loss": 0.798, "step": 11223 }, { "epoch": 0.57, "learning_rate": 8.22176318384922e-06, "loss": 0.9483, "step": 11224 }, { "epoch": 0.57, "learning_rate": 8.220144016027432e-06, "loss": 0.8508, "step": 11225 }, { "epoch": 0.57, "learning_rate": 8.218524896393352e-06, "loss": 0.9306, "step": 11226 }, { "epoch": 0.57, "learning_rate": 8.216905824990817e-06, "loss": 0.9498, "step": 11227 }, { "epoch": 0.57, "learning_rate": 8.215286801863654e-06, "loss": 0.8582, "step": 11228 }, { "epoch": 0.57, "learning_rate": 8.213667827055706e-06, "loss": 0.9063, "step": 11229 }, { "epoch": 0.57, "learning_rate": 8.212048900610802e-06, "loss": 0.9095, "step": 11230 }, { "epoch": 0.57, "learning_rate": 8.21043002257277e-06, "loss": 0.9545, "step": 11231 }, { "epoch": 0.57, "learning_rate": 8.208811192985442e-06, "loss": 0.9594, "step": 11232 }, { "epoch": 0.57, "learning_rate": 8.207192411892645e-06, "loss": 0.9751, "step": 11233 }, { "epoch": 0.57, "learning_rate": 8.205573679338204e-06, "loss": 0.7838, "step": 11234 }, { "epoch": 0.57, "learning_rate": 8.20395499536595e-06, "loss": 0.8991, "step": 11235 }, { "epoch": 0.57, "learning_rate": 8.202336360019702e-06, "loss": 0.9255, "step": 11236 }, { "epoch": 0.57, "learning_rate": 8.200717773343286e-06, "loss": 0.9318, "step": 11237 }, { "epoch": 0.57, "learning_rate": 8.199099235380522e-06, "loss": 0.8899, "step": 11238 }, { "epoch": 0.57, "learning_rate": 8.197480746175228e-06, "loss": 0.9176, "step": 11239 }, { "epoch": 0.57, "learning_rate": 8.195862305771223e-06, "loss": 0.9553, "step": 11240 }, { "epoch": 0.57, "learning_rate": 8.194243914212331e-06, "loss": 0.9446, "step": 11241 }, { "epoch": 0.57, "learning_rate": 8.192625571542363e-06, "loss": 0.9825, "step": 11242 }, { "epoch": 0.57, "learning_rate": 8.191007277805134e-06, "loss": 0.8333, "step": 11243 }, { "epoch": 0.57, "learning_rate": 8.18938903304446e-06, "loss": 0.9156, "step": 11244 }, { "epoch": 0.57, "learning_rate": 8.18777083730415e-06, "loss": 0.9003, "step": 11245 }, { "epoch": 0.57, "learning_rate": 8.186152690628014e-06, "loss": 0.8967, "step": 11246 }, { "epoch": 0.57, "learning_rate": 8.184534593059868e-06, "loss": 0.8551, "step": 11247 }, { "epoch": 0.57, "learning_rate": 8.182916544643516e-06, "loss": 0.9745, "step": 11248 }, { "epoch": 0.57, "learning_rate": 8.181298545422764e-06, "loss": 0.8873, "step": 11249 }, { "epoch": 0.57, "learning_rate": 8.17968059544142e-06, "loss": 0.9567, "step": 11250 }, { "epoch": 0.57, "learning_rate": 8.178062694743283e-06, "loss": 0.955, "step": 11251 }, { "epoch": 0.57, "learning_rate": 8.176444843372165e-06, "loss": 0.8991, "step": 11252 }, { "epoch": 0.57, "learning_rate": 8.174827041371863e-06, "loss": 0.9708, "step": 11253 }, { "epoch": 0.57, "learning_rate": 8.173209288786174e-06, "loss": 0.8202, "step": 11254 }, { "epoch": 0.57, "learning_rate": 8.171591585658902e-06, "loss": 0.9089, "step": 11255 }, { "epoch": 0.57, "learning_rate": 8.169973932033843e-06, "loss": 0.8451, "step": 11256 }, { "epoch": 0.57, "learning_rate": 8.16835632795479e-06, "loss": 0.8701, "step": 11257 }, { "epoch": 0.57, "learning_rate": 8.166738773465544e-06, "loss": 0.8802, "step": 11258 }, { "epoch": 0.57, "learning_rate": 8.165121268609894e-06, "loss": 0.8764, "step": 11259 }, { "epoch": 0.57, "learning_rate": 8.163503813431632e-06, "loss": 0.8906, "step": 11260 }, { "epoch": 0.57, "learning_rate": 8.161886407974553e-06, "loss": 0.8664, "step": 11261 }, { "epoch": 0.57, "learning_rate": 8.160269052282444e-06, "loss": 1.0394, "step": 11262 }, { "epoch": 0.57, "learning_rate": 8.158651746399088e-06, "loss": 0.9222, "step": 11263 }, { "epoch": 0.57, "learning_rate": 8.157034490368283e-06, "loss": 0.8666, "step": 11264 }, { "epoch": 0.57, "learning_rate": 8.155417284233806e-06, "loss": 0.8981, "step": 11265 }, { "epoch": 0.57, "learning_rate": 8.153800128039441e-06, "loss": 1.0253, "step": 11266 }, { "epoch": 0.57, "learning_rate": 8.152183021828979e-06, "loss": 0.913, "step": 11267 }, { "epoch": 0.57, "learning_rate": 8.150565965646193e-06, "loss": 0.9773, "step": 11268 }, { "epoch": 0.57, "learning_rate": 8.148948959534863e-06, "loss": 0.8783, "step": 11269 }, { "epoch": 0.57, "learning_rate": 8.147332003538773e-06, "loss": 0.8364, "step": 11270 }, { "epoch": 0.57, "learning_rate": 8.1457150977017e-06, "loss": 0.8954, "step": 11271 }, { "epoch": 0.57, "learning_rate": 8.144098242067414e-06, "loss": 0.9122, "step": 11272 }, { "epoch": 0.57, "learning_rate": 8.142481436679696e-06, "loss": 0.9096, "step": 11273 }, { "epoch": 0.57, "learning_rate": 8.140864681582319e-06, "loss": 0.869, "step": 11274 }, { "epoch": 0.57, "learning_rate": 8.139247976819046e-06, "loss": 0.7581, "step": 11275 }, { "epoch": 0.57, "learning_rate": 8.13763132243366e-06, "loss": 0.8533, "step": 11276 }, { "epoch": 0.57, "learning_rate": 8.136014718469924e-06, "loss": 0.8882, "step": 11277 }, { "epoch": 0.57, "learning_rate": 8.134398164971604e-06, "loss": 0.8738, "step": 11278 }, { "epoch": 0.57, "learning_rate": 8.132781661982472e-06, "loss": 0.8093, "step": 11279 }, { "epoch": 0.57, "learning_rate": 8.131165209546288e-06, "loss": 0.8558, "step": 11280 }, { "epoch": 0.57, "learning_rate": 8.129548807706815e-06, "loss": 0.8302, "step": 11281 }, { "epoch": 0.57, "learning_rate": 8.127932456507822e-06, "loss": 0.7902, "step": 11282 }, { "epoch": 0.57, "learning_rate": 8.126316155993066e-06, "loss": 0.9619, "step": 11283 }, { "epoch": 0.57, "learning_rate": 8.124699906206304e-06, "loss": 1.0168, "step": 11284 }, { "epoch": 0.57, "learning_rate": 8.1230837071913e-06, "loss": 0.9326, "step": 11285 }, { "epoch": 0.57, "learning_rate": 8.121467558991807e-06, "loss": 0.9652, "step": 11286 }, { "epoch": 0.57, "learning_rate": 8.119851461651578e-06, "loss": 0.9266, "step": 11287 }, { "epoch": 0.57, "learning_rate": 8.118235415214372e-06, "loss": 0.906, "step": 11288 }, { "epoch": 0.57, "learning_rate": 8.116619419723943e-06, "loss": 0.9035, "step": 11289 }, { "epoch": 0.57, "learning_rate": 8.115003475224036e-06, "loss": 0.9554, "step": 11290 }, { "epoch": 0.57, "learning_rate": 8.113387581758407e-06, "loss": 0.9421, "step": 11291 }, { "epoch": 0.57, "learning_rate": 8.111771739370802e-06, "loss": 0.9238, "step": 11292 }, { "epoch": 0.57, "learning_rate": 8.110155948104963e-06, "loss": 0.8058, "step": 11293 }, { "epoch": 0.57, "learning_rate": 8.108540208004647e-06, "loss": 0.9507, "step": 11294 }, { "epoch": 0.57, "learning_rate": 8.106924519113592e-06, "loss": 0.8317, "step": 11295 }, { "epoch": 0.57, "learning_rate": 8.105308881475538e-06, "loss": 0.9177, "step": 11296 }, { "epoch": 0.57, "learning_rate": 8.103693295134233e-06, "loss": 0.8748, "step": 11297 }, { "epoch": 0.57, "learning_rate": 8.102077760133415e-06, "loss": 0.8574, "step": 11298 }, { "epoch": 0.57, "learning_rate": 8.100462276516819e-06, "loss": 0.8858, "step": 11299 }, { "epoch": 0.57, "learning_rate": 8.098846844328188e-06, "loss": 0.8237, "step": 11300 }, { "epoch": 0.57, "learning_rate": 8.097231463611258e-06, "loss": 0.9655, "step": 11301 }, { "epoch": 0.57, "learning_rate": 8.09561613440976e-06, "loss": 0.9155, "step": 11302 }, { "epoch": 0.57, "learning_rate": 8.094000856767428e-06, "loss": 0.9626, "step": 11303 }, { "epoch": 0.57, "learning_rate": 8.092385630727998e-06, "loss": 0.9588, "step": 11304 }, { "epoch": 0.57, "learning_rate": 8.090770456335192e-06, "loss": 0.9379, "step": 11305 }, { "epoch": 0.57, "learning_rate": 8.08915533363275e-06, "loss": 1.002, "step": 11306 }, { "epoch": 0.57, "learning_rate": 8.087540262664393e-06, "loss": 0.9122, "step": 11307 }, { "epoch": 0.57, "learning_rate": 8.085925243473847e-06, "loss": 0.8326, "step": 11308 }, { "epoch": 0.57, "learning_rate": 8.08431027610484e-06, "loss": 0.8496, "step": 11309 }, { "epoch": 0.57, "learning_rate": 8.082695360601097e-06, "loss": 0.95, "step": 11310 }, { "epoch": 0.57, "learning_rate": 8.081080497006331e-06, "loss": 0.8385, "step": 11311 }, { "epoch": 0.57, "learning_rate": 8.079465685364274e-06, "loss": 0.8727, "step": 11312 }, { "epoch": 0.57, "learning_rate": 8.077850925718638e-06, "loss": 0.989, "step": 11313 }, { "epoch": 0.57, "learning_rate": 8.076236218113143e-06, "loss": 0.8751, "step": 11314 }, { "epoch": 0.57, "learning_rate": 8.074621562591507e-06, "loss": 0.8724, "step": 11315 }, { "epoch": 0.57, "learning_rate": 8.073006959197445e-06, "loss": 0.9523, "step": 11316 }, { "epoch": 0.57, "learning_rate": 8.071392407974663e-06, "loss": 0.8365, "step": 11317 }, { "epoch": 0.57, "learning_rate": 8.069777908966886e-06, "loss": 0.9138, "step": 11318 }, { "epoch": 0.58, "learning_rate": 8.068163462217814e-06, "loss": 0.8651, "step": 11319 }, { "epoch": 0.58, "learning_rate": 8.066549067771166e-06, "loss": 0.8809, "step": 11320 }, { "epoch": 0.58, "learning_rate": 8.064934725670644e-06, "loss": 0.9703, "step": 11321 }, { "epoch": 0.58, "learning_rate": 8.063320435959951e-06, "loss": 0.8263, "step": 11322 }, { "epoch": 0.58, "learning_rate": 8.0617061986828e-06, "loss": 0.9119, "step": 11323 }, { "epoch": 0.58, "learning_rate": 8.060092013882893e-06, "loss": 0.903, "step": 11324 }, { "epoch": 0.58, "learning_rate": 8.058477881603927e-06, "loss": 0.888, "step": 11325 }, { "epoch": 0.58, "learning_rate": 8.056863801889611e-06, "loss": 0.8335, "step": 11326 }, { "epoch": 0.58, "learning_rate": 8.05524977478364e-06, "loss": 0.9635, "step": 11327 }, { "epoch": 0.58, "learning_rate": 8.053635800329706e-06, "loss": 0.9322, "step": 11328 }, { "epoch": 0.58, "learning_rate": 8.052021878571519e-06, "loss": 0.8324, "step": 11329 }, { "epoch": 0.58, "learning_rate": 8.050408009552765e-06, "loss": 0.9251, "step": 11330 }, { "epoch": 0.58, "learning_rate": 8.048794193317137e-06, "loss": 0.8953, "step": 11331 }, { "epoch": 0.58, "learning_rate": 8.047180429908335e-06, "loss": 0.9368, "step": 11332 }, { "epoch": 0.58, "learning_rate": 8.045566719370044e-06, "loss": 0.9105, "step": 11333 }, { "epoch": 0.58, "learning_rate": 8.04395306174595e-06, "loss": 0.9091, "step": 11334 }, { "epoch": 0.58, "learning_rate": 8.04233945707975e-06, "loss": 0.8557, "step": 11335 }, { "epoch": 0.58, "learning_rate": 8.040725905415128e-06, "loss": 0.9685, "step": 11336 }, { "epoch": 0.58, "learning_rate": 8.039112406795766e-06, "loss": 0.926, "step": 11337 }, { "epoch": 0.58, "learning_rate": 8.037498961265349e-06, "loss": 0.8635, "step": 11338 }, { "epoch": 0.58, "learning_rate": 8.03588556886756e-06, "loss": 0.9021, "step": 11339 }, { "epoch": 0.58, "learning_rate": 8.034272229646075e-06, "loss": 0.7989, "step": 11340 }, { "epoch": 0.58, "learning_rate": 8.032658943644585e-06, "loss": 0.9224, "step": 11341 }, { "epoch": 0.58, "learning_rate": 8.031045710906758e-06, "loss": 0.8572, "step": 11342 }, { "epoch": 0.58, "learning_rate": 8.029432531476273e-06, "loss": 0.8604, "step": 11343 }, { "epoch": 0.58, "learning_rate": 8.027819405396809e-06, "loss": 0.9623, "step": 11344 }, { "epoch": 0.58, "learning_rate": 8.026206332712034e-06, "loss": 0.897, "step": 11345 }, { "epoch": 0.58, "learning_rate": 8.024593313465618e-06, "loss": 0.8951, "step": 11346 }, { "epoch": 0.58, "learning_rate": 8.022980347701244e-06, "loss": 0.8784, "step": 11347 }, { "epoch": 0.58, "learning_rate": 8.021367435462571e-06, "loss": 0.9222, "step": 11348 }, { "epoch": 0.58, "learning_rate": 8.019754576793267e-06, "loss": 0.9103, "step": 11349 }, { "epoch": 0.58, "learning_rate": 8.018141771737005e-06, "loss": 0.8636, "step": 11350 }, { "epoch": 0.58, "learning_rate": 8.016529020337445e-06, "loss": 0.9322, "step": 11351 }, { "epoch": 0.58, "learning_rate": 8.014916322638247e-06, "loss": 0.803, "step": 11352 }, { "epoch": 0.58, "learning_rate": 8.013303678683082e-06, "loss": 0.8836, "step": 11353 }, { "epoch": 0.58, "learning_rate": 8.011691088515607e-06, "loss": 1.0065, "step": 11354 }, { "epoch": 0.58, "learning_rate": 8.010078552179479e-06, "loss": 0.8496, "step": 11355 }, { "epoch": 0.58, "learning_rate": 8.008466069718357e-06, "loss": 0.9998, "step": 11356 }, { "epoch": 0.58, "learning_rate": 8.006853641175898e-06, "loss": 0.8386, "step": 11357 }, { "epoch": 0.58, "learning_rate": 8.005241266595753e-06, "loss": 0.8342, "step": 11358 }, { "epoch": 0.58, "learning_rate": 8.00362894602158e-06, "loss": 0.927, "step": 11359 }, { "epoch": 0.58, "learning_rate": 8.002016679497033e-06, "loss": 0.9171, "step": 11360 }, { "epoch": 0.58, "learning_rate": 8.000404467065754e-06, "loss": 0.8547, "step": 11361 }, { "epoch": 0.58, "learning_rate": 7.998792308771399e-06, "loss": 0.8938, "step": 11362 }, { "epoch": 0.58, "learning_rate": 7.997180204657614e-06, "loss": 0.9397, "step": 11363 }, { "epoch": 0.58, "learning_rate": 7.99556815476804e-06, "loss": 0.8602, "step": 11364 }, { "epoch": 0.58, "learning_rate": 7.99395615914633e-06, "loss": 0.9021, "step": 11365 }, { "epoch": 0.58, "learning_rate": 7.992344217836122e-06, "loss": 0.8403, "step": 11366 }, { "epoch": 0.58, "learning_rate": 7.990732330881056e-06, "loss": 0.8434, "step": 11367 }, { "epoch": 0.58, "learning_rate": 7.989120498324778e-06, "loss": 0.8682, "step": 11368 }, { "epoch": 0.58, "learning_rate": 7.987508720210922e-06, "loss": 0.9019, "step": 11369 }, { "epoch": 0.58, "learning_rate": 7.985896996583121e-06, "loss": 0.9459, "step": 11370 }, { "epoch": 0.58, "learning_rate": 7.984285327485021e-06, "loss": 0.9059, "step": 11371 }, { "epoch": 0.58, "learning_rate": 7.982673712960251e-06, "loss": 0.8884, "step": 11372 }, { "epoch": 0.58, "learning_rate": 7.981062153052443e-06, "loss": 0.8507, "step": 11373 }, { "epoch": 0.58, "learning_rate": 7.979450647805232e-06, "loss": 0.9367, "step": 11374 }, { "epoch": 0.58, "learning_rate": 7.977839197262242e-06, "loss": 0.9484, "step": 11375 }, { "epoch": 0.58, "learning_rate": 7.976227801467102e-06, "loss": 0.9111, "step": 11376 }, { "epoch": 0.58, "learning_rate": 7.974616460463446e-06, "loss": 0.9341, "step": 11377 }, { "epoch": 0.58, "learning_rate": 7.973005174294893e-06, "loss": 0.9418, "step": 11378 }, { "epoch": 0.58, "learning_rate": 7.971393943005067e-06, "loss": 0.8463, "step": 11379 }, { "epoch": 0.58, "learning_rate": 7.969782766637595e-06, "loss": 0.9193, "step": 11380 }, { "epoch": 0.58, "learning_rate": 7.968171645236092e-06, "loss": 1.0099, "step": 11381 }, { "epoch": 0.58, "learning_rate": 7.966560578844178e-06, "loss": 0.9128, "step": 11382 }, { "epoch": 0.58, "learning_rate": 7.964949567505476e-06, "loss": 0.868, "step": 11383 }, { "epoch": 0.58, "learning_rate": 7.963338611263601e-06, "loss": 0.8591, "step": 11384 }, { "epoch": 0.58, "learning_rate": 7.961727710162161e-06, "loss": 0.8922, "step": 11385 }, { "epoch": 0.58, "learning_rate": 7.96011686424478e-06, "loss": 0.9434, "step": 11386 }, { "epoch": 0.58, "learning_rate": 7.958506073555062e-06, "loss": 0.9099, "step": 11387 }, { "epoch": 0.58, "learning_rate": 7.956895338136618e-06, "loss": 0.9504, "step": 11388 }, { "epoch": 0.58, "learning_rate": 7.955284658033062e-06, "loss": 0.9059, "step": 11389 }, { "epoch": 0.58, "learning_rate": 7.953674033287997e-06, "loss": 0.8765, "step": 11390 }, { "epoch": 0.58, "learning_rate": 7.952063463945033e-06, "loss": 0.888, "step": 11391 }, { "epoch": 0.58, "learning_rate": 7.95045295004777e-06, "loss": 0.9406, "step": 11392 }, { "epoch": 0.58, "learning_rate": 7.94884249163981e-06, "loss": 0.9733, "step": 11393 }, { "epoch": 0.58, "learning_rate": 7.94723208876476e-06, "loss": 0.8648, "step": 11394 }, { "epoch": 0.58, "learning_rate": 7.94562174146622e-06, "loss": 0.7979, "step": 11395 }, { "epoch": 0.58, "learning_rate": 7.944011449787782e-06, "loss": 0.9474, "step": 11396 }, { "epoch": 0.58, "learning_rate": 7.942401213773048e-06, "loss": 0.9802, "step": 11397 }, { "epoch": 0.58, "learning_rate": 7.940791033465613e-06, "loss": 0.8664, "step": 11398 }, { "epoch": 0.58, "learning_rate": 7.939180908909065e-06, "loss": 0.9894, "step": 11399 }, { "epoch": 0.58, "learning_rate": 7.937570840147006e-06, "loss": 0.9435, "step": 11400 }, { "epoch": 0.58, "learning_rate": 7.935960827223022e-06, "loss": 0.9163, "step": 11401 }, { "epoch": 0.58, "learning_rate": 7.934350870180702e-06, "loss": 0.914, "step": 11402 }, { "epoch": 0.58, "learning_rate": 7.932740969063634e-06, "loss": 0.9276, "step": 11403 }, { "epoch": 0.58, "learning_rate": 7.931131123915407e-06, "loss": 0.9121, "step": 11404 }, { "epoch": 0.58, "learning_rate": 7.9295213347796e-06, "loss": 1.0217, "step": 11405 }, { "epoch": 0.58, "learning_rate": 7.927911601699803e-06, "loss": 1.0112, "step": 11406 }, { "epoch": 0.58, "learning_rate": 7.926301924719596e-06, "loss": 0.9878, "step": 11407 }, { "epoch": 0.58, "learning_rate": 7.924692303882554e-06, "loss": 0.7975, "step": 11408 }, { "epoch": 0.58, "learning_rate": 7.923082739232261e-06, "loss": 0.8706, "step": 11409 }, { "epoch": 0.58, "learning_rate": 7.921473230812297e-06, "loss": 0.9165, "step": 11410 }, { "epoch": 0.58, "learning_rate": 7.919863778666227e-06, "loss": 0.9716, "step": 11411 }, { "epoch": 0.58, "learning_rate": 7.918254382837638e-06, "loss": 0.9675, "step": 11412 }, { "epoch": 0.58, "learning_rate": 7.916645043370096e-06, "loss": 0.8556, "step": 11413 }, { "epoch": 0.58, "learning_rate": 7.91503576030717e-06, "loss": 0.9083, "step": 11414 }, { "epoch": 0.58, "learning_rate": 7.913426533692435e-06, "loss": 0.9742, "step": 11415 }, { "epoch": 0.58, "learning_rate": 7.91181736356946e-06, "loss": 0.8354, "step": 11416 }, { "epoch": 0.58, "learning_rate": 7.910208249981799e-06, "loss": 0.8822, "step": 11417 }, { "epoch": 0.58, "learning_rate": 7.908599192973033e-06, "loss": 0.8861, "step": 11418 }, { "epoch": 0.58, "learning_rate": 7.906990192586718e-06, "loss": 0.9229, "step": 11419 }, { "epoch": 0.58, "learning_rate": 7.905381248866415e-06, "loss": 0.9723, "step": 11420 }, { "epoch": 0.58, "learning_rate": 7.903772361855688e-06, "loss": 1.0198, "step": 11421 }, { "epoch": 0.58, "learning_rate": 7.902163531598095e-06, "loss": 0.9027, "step": 11422 }, { "epoch": 0.58, "learning_rate": 7.900554758137188e-06, "loss": 0.9455, "step": 11423 }, { "epoch": 0.58, "learning_rate": 7.898946041516531e-06, "loss": 0.8732, "step": 11424 }, { "epoch": 0.58, "learning_rate": 7.897337381779675e-06, "loss": 0.9725, "step": 11425 }, { "epoch": 0.58, "learning_rate": 7.895728778970168e-06, "loss": 0.9948, "step": 11426 }, { "epoch": 0.58, "learning_rate": 7.894120233131571e-06, "loss": 0.8065, "step": 11427 }, { "epoch": 0.58, "learning_rate": 7.892511744307427e-06, "loss": 0.8867, "step": 11428 }, { "epoch": 0.58, "learning_rate": 7.89090331254128e-06, "loss": 0.8745, "step": 11429 }, { "epoch": 0.58, "learning_rate": 7.889294937876688e-06, "loss": 0.9223, "step": 11430 }, { "epoch": 0.58, "learning_rate": 7.88768662035719e-06, "loss": 0.8683, "step": 11431 }, { "epoch": 0.58, "learning_rate": 7.886078360026325e-06, "loss": 0.8964, "step": 11432 }, { "epoch": 0.58, "learning_rate": 7.884470156927643e-06, "loss": 0.9349, "step": 11433 }, { "epoch": 0.58, "learning_rate": 7.882862011104681e-06, "loss": 0.9772, "step": 11434 }, { "epoch": 0.58, "learning_rate": 7.881253922600971e-06, "loss": 0.8992, "step": 11435 }, { "epoch": 0.58, "learning_rate": 7.879645891460066e-06, "loss": 0.9171, "step": 11436 }, { "epoch": 0.58, "learning_rate": 7.878037917725491e-06, "loss": 0.9561, "step": 11437 }, { "epoch": 0.58, "learning_rate": 7.876430001440776e-06, "loss": 0.8986, "step": 11438 }, { "epoch": 0.58, "learning_rate": 7.874822142649464e-06, "loss": 0.9235, "step": 11439 }, { "epoch": 0.58, "learning_rate": 7.87321434139508e-06, "loss": 0.9364, "step": 11440 }, { "epoch": 0.58, "learning_rate": 7.871606597721153e-06, "loss": 0.8325, "step": 11441 }, { "epoch": 0.58, "learning_rate": 7.869998911671216e-06, "loss": 0.9714, "step": 11442 }, { "epoch": 0.58, "learning_rate": 7.868391283288795e-06, "loss": 0.9647, "step": 11443 }, { "epoch": 0.58, "learning_rate": 7.866783712617405e-06, "loss": 0.9416, "step": 11444 }, { "epoch": 0.58, "learning_rate": 7.865176199700582e-06, "loss": 0.9166, "step": 11445 }, { "epoch": 0.58, "learning_rate": 7.86356874458184e-06, "loss": 0.9172, "step": 11446 }, { "epoch": 0.58, "learning_rate": 7.861961347304698e-06, "loss": 0.8818, "step": 11447 }, { "epoch": 0.58, "learning_rate": 7.860354007912683e-06, "loss": 0.7631, "step": 11448 }, { "epoch": 0.58, "learning_rate": 7.858746726449309e-06, "loss": 0.871, "step": 11449 }, { "epoch": 0.58, "learning_rate": 7.857139502958082e-06, "loss": 0.9221, "step": 11450 }, { "epoch": 0.58, "learning_rate": 7.855532337482528e-06, "loss": 0.92, "step": 11451 }, { "epoch": 0.58, "learning_rate": 7.853925230066152e-06, "loss": 0.8716, "step": 11452 }, { "epoch": 0.58, "learning_rate": 7.852318180752466e-06, "loss": 0.8507, "step": 11453 }, { "epoch": 0.58, "learning_rate": 7.850711189584984e-06, "loss": 0.9673, "step": 11454 }, { "epoch": 0.58, "learning_rate": 7.849104256607212e-06, "loss": 1.0026, "step": 11455 }, { "epoch": 0.58, "learning_rate": 7.847497381862647e-06, "loss": 0.843, "step": 11456 }, { "epoch": 0.58, "learning_rate": 7.845890565394804e-06, "loss": 0.8768, "step": 11457 }, { "epoch": 0.58, "learning_rate": 7.844283807247178e-06, "loss": 0.899, "step": 11458 }, { "epoch": 0.58, "learning_rate": 7.842677107463279e-06, "loss": 0.871, "step": 11459 }, { "epoch": 0.58, "learning_rate": 7.841070466086603e-06, "loss": 0.884, "step": 11460 }, { "epoch": 0.58, "learning_rate": 7.839463883160644e-06, "loss": 0.8403, "step": 11461 }, { "epoch": 0.58, "learning_rate": 7.837857358728905e-06, "loss": 1.018, "step": 11462 }, { "epoch": 0.58, "learning_rate": 7.836250892834876e-06, "loss": 0.7997, "step": 11463 }, { "epoch": 0.58, "learning_rate": 7.834644485522048e-06, "loss": 0.9104, "step": 11464 }, { "epoch": 0.58, "learning_rate": 7.833038136833922e-06, "loss": 0.9286, "step": 11465 }, { "epoch": 0.58, "learning_rate": 7.831431846813985e-06, "loss": 0.8412, "step": 11466 }, { "epoch": 0.58, "learning_rate": 7.829825615505716e-06, "loss": 0.9039, "step": 11467 }, { "epoch": 0.58, "learning_rate": 7.828219442952616e-06, "loss": 0.84, "step": 11468 }, { "epoch": 0.58, "learning_rate": 7.826613329198163e-06, "loss": 0.9124, "step": 11469 }, { "epoch": 0.58, "learning_rate": 7.825007274285836e-06, "loss": 0.8272, "step": 11470 }, { "epoch": 0.58, "learning_rate": 7.823401278259129e-06, "loss": 0.8002, "step": 11471 }, { "epoch": 0.58, "learning_rate": 7.821795341161518e-06, "loss": 0.8311, "step": 11472 }, { "epoch": 0.58, "learning_rate": 7.820189463036475e-06, "loss": 0.8771, "step": 11473 }, { "epoch": 0.58, "learning_rate": 7.818583643927488e-06, "loss": 0.9681, "step": 11474 }, { "epoch": 0.58, "learning_rate": 7.816977883878027e-06, "loss": 0.9178, "step": 11475 }, { "epoch": 0.58, "learning_rate": 7.815372182931562e-06, "loss": 0.8689, "step": 11476 }, { "epoch": 0.58, "learning_rate": 7.813766541131577e-06, "loss": 0.8925, "step": 11477 }, { "epoch": 0.58, "learning_rate": 7.812160958521538e-06, "loss": 0.9504, "step": 11478 }, { "epoch": 0.58, "learning_rate": 7.810555435144907e-06, "loss": 0.882, "step": 11479 }, { "epoch": 0.58, "learning_rate": 7.808949971045164e-06, "loss": 0.8474, "step": 11480 }, { "epoch": 0.58, "learning_rate": 7.807344566265768e-06, "loss": 0.9496, "step": 11481 }, { "epoch": 0.58, "learning_rate": 7.80573922085018e-06, "loss": 0.8916, "step": 11482 }, { "epoch": 0.58, "learning_rate": 7.804133934841874e-06, "loss": 0.9403, "step": 11483 }, { "epoch": 0.58, "learning_rate": 7.802528708284307e-06, "loss": 0.8815, "step": 11484 }, { "epoch": 0.58, "learning_rate": 7.80092354122093e-06, "loss": 0.8977, "step": 11485 }, { "epoch": 0.58, "learning_rate": 7.799318433695214e-06, "loss": 0.8943, "step": 11486 }, { "epoch": 0.58, "learning_rate": 7.79771338575061e-06, "loss": 0.8803, "step": 11487 }, { "epoch": 0.58, "learning_rate": 7.796108397430568e-06, "loss": 0.9904, "step": 11488 }, { "epoch": 0.58, "learning_rate": 7.794503468778551e-06, "loss": 0.9203, "step": 11489 }, { "epoch": 0.58, "learning_rate": 7.792898599838009e-06, "loss": 0.9363, "step": 11490 }, { "epoch": 0.58, "learning_rate": 7.791293790652383e-06, "loss": 0.9072, "step": 11491 }, { "epoch": 0.58, "learning_rate": 7.78968904126513e-06, "loss": 0.9258, "step": 11492 }, { "epoch": 0.58, "learning_rate": 7.788084351719696e-06, "loss": 0.9003, "step": 11493 }, { "epoch": 0.58, "learning_rate": 7.78647972205952e-06, "loss": 0.9061, "step": 11494 }, { "epoch": 0.58, "learning_rate": 7.784875152328056e-06, "loss": 0.8595, "step": 11495 }, { "epoch": 0.58, "learning_rate": 7.783270642568742e-06, "loss": 0.8993, "step": 11496 }, { "epoch": 0.58, "learning_rate": 7.78166619282501e-06, "loss": 0.8694, "step": 11497 }, { "epoch": 0.58, "learning_rate": 7.78006180314031e-06, "loss": 0.9285, "step": 11498 }, { "epoch": 0.58, "learning_rate": 7.778457473558076e-06, "loss": 0.9852, "step": 11499 }, { "epoch": 0.58, "learning_rate": 7.776853204121737e-06, "loss": 0.865, "step": 11500 }, { "epoch": 0.58, "learning_rate": 7.77524899487474e-06, "loss": 0.9331, "step": 11501 }, { "epoch": 0.58, "learning_rate": 7.773644845860506e-06, "loss": 1.0139, "step": 11502 }, { "epoch": 0.58, "learning_rate": 7.772040757122466e-06, "loss": 0.9482, "step": 11503 }, { "epoch": 0.58, "learning_rate": 7.770436728704055e-06, "loss": 0.7803, "step": 11504 }, { "epoch": 0.58, "learning_rate": 7.768832760648699e-06, "loss": 0.9415, "step": 11505 }, { "epoch": 0.58, "learning_rate": 7.767228852999816e-06, "loss": 0.8884, "step": 11506 }, { "epoch": 0.58, "learning_rate": 7.765625005800845e-06, "loss": 0.9692, "step": 11507 }, { "epoch": 0.58, "learning_rate": 7.764021219095197e-06, "loss": 0.9921, "step": 11508 }, { "epoch": 0.58, "learning_rate": 7.76241749292629e-06, "loss": 0.9109, "step": 11509 }, { "epoch": 0.58, "learning_rate": 7.760813827337555e-06, "loss": 0.9353, "step": 11510 }, { "epoch": 0.58, "learning_rate": 7.759210222372402e-06, "loss": 0.9735, "step": 11511 }, { "epoch": 0.58, "learning_rate": 7.757606678074243e-06, "loss": 0.9868, "step": 11512 }, { "epoch": 0.58, "learning_rate": 7.756003194486505e-06, "loss": 0.8674, "step": 11513 }, { "epoch": 0.58, "learning_rate": 7.754399771652592e-06, "loss": 0.9469, "step": 11514 }, { "epoch": 0.58, "learning_rate": 7.75279640961591e-06, "loss": 0.8986, "step": 11515 }, { "epoch": 0.59, "learning_rate": 7.75119310841988e-06, "loss": 0.842, "step": 11516 }, { "epoch": 0.59, "learning_rate": 7.749589868107902e-06, "loss": 0.8778, "step": 11517 }, { "epoch": 0.59, "learning_rate": 7.747986688723382e-06, "loss": 0.9156, "step": 11518 }, { "epoch": 0.59, "learning_rate": 7.746383570309732e-06, "loss": 0.9497, "step": 11519 }, { "epoch": 0.59, "learning_rate": 7.744780512910348e-06, "loss": 0.938, "step": 11520 }, { "epoch": 0.59, "learning_rate": 7.743177516568628e-06, "loss": 0.9049, "step": 11521 }, { "epoch": 0.59, "learning_rate": 7.74157458132798e-06, "loss": 0.8846, "step": 11522 }, { "epoch": 0.59, "learning_rate": 7.739971707231795e-06, "loss": 0.9534, "step": 11523 }, { "epoch": 0.59, "learning_rate": 7.738368894323469e-06, "loss": 0.9429, "step": 11524 }, { "epoch": 0.59, "learning_rate": 7.736766142646405e-06, "loss": 0.8706, "step": 11525 }, { "epoch": 0.59, "learning_rate": 7.735163452243988e-06, "loss": 0.8994, "step": 11526 }, { "epoch": 0.59, "learning_rate": 7.733560823159607e-06, "loss": 0.9694, "step": 11527 }, { "epoch": 0.59, "learning_rate": 7.73195825543666e-06, "loss": 0.9408, "step": 11528 }, { "epoch": 0.59, "learning_rate": 7.730355749118524e-06, "loss": 0.863, "step": 11529 }, { "epoch": 0.59, "learning_rate": 7.728753304248598e-06, "loss": 0.9151, "step": 11530 }, { "epoch": 0.59, "learning_rate": 7.72715092087026e-06, "loss": 0.8587, "step": 11531 }, { "epoch": 0.59, "learning_rate": 7.725548599026886e-06, "loss": 0.923, "step": 11532 }, { "epoch": 0.59, "learning_rate": 7.723946338761869e-06, "loss": 0.8923, "step": 11533 }, { "epoch": 0.59, "learning_rate": 7.722344140118585e-06, "loss": 0.9145, "step": 11534 }, { "epoch": 0.59, "learning_rate": 7.720742003140404e-06, "loss": 0.8291, "step": 11535 }, { "epoch": 0.59, "learning_rate": 7.719139927870716e-06, "loss": 0.7943, "step": 11536 }, { "epoch": 0.59, "learning_rate": 7.717537914352884e-06, "loss": 0.8157, "step": 11537 }, { "epoch": 0.59, "learning_rate": 7.71593596263028e-06, "loss": 0.9283, "step": 11538 }, { "epoch": 0.59, "learning_rate": 7.714334072746286e-06, "loss": 0.9451, "step": 11539 }, { "epoch": 0.59, "learning_rate": 7.712732244744265e-06, "loss": 0.8577, "step": 11540 }, { "epoch": 0.59, "learning_rate": 7.711130478667582e-06, "loss": 0.9191, "step": 11541 }, { "epoch": 0.59, "learning_rate": 7.709528774559608e-06, "loss": 0.918, "step": 11542 }, { "epoch": 0.59, "learning_rate": 7.707927132463708e-06, "loss": 0.9301, "step": 11543 }, { "epoch": 0.59, "learning_rate": 7.706325552423236e-06, "loss": 0.8755, "step": 11544 }, { "epoch": 0.59, "learning_rate": 7.704724034481563e-06, "loss": 0.8213, "step": 11545 }, { "epoch": 0.59, "learning_rate": 7.703122578682047e-06, "loss": 0.9218, "step": 11546 }, { "epoch": 0.59, "learning_rate": 7.701521185068037e-06, "loss": 0.8841, "step": 11547 }, { "epoch": 0.59, "learning_rate": 7.699919853682904e-06, "loss": 0.9369, "step": 11548 }, { "epoch": 0.59, "learning_rate": 7.69831858456999e-06, "loss": 0.922, "step": 11549 }, { "epoch": 0.59, "learning_rate": 7.696717377772648e-06, "loss": 0.9521, "step": 11550 }, { "epoch": 0.59, "learning_rate": 7.695116233334237e-06, "loss": 0.8185, "step": 11551 }, { "epoch": 0.59, "learning_rate": 7.6935151512981e-06, "loss": 1.0127, "step": 11552 }, { "epoch": 0.59, "learning_rate": 7.691914131707584e-06, "loss": 0.8105, "step": 11553 }, { "epoch": 0.59, "learning_rate": 7.690313174606044e-06, "loss": 0.9105, "step": 11554 }, { "epoch": 0.59, "learning_rate": 7.688712280036815e-06, "loss": 0.9988, "step": 11555 }, { "epoch": 0.59, "learning_rate": 7.687111448043236e-06, "loss": 0.8463, "step": 11556 }, { "epoch": 0.59, "learning_rate": 7.68551067866866e-06, "loss": 0.9191, "step": 11557 }, { "epoch": 0.59, "learning_rate": 7.68390997195642e-06, "loss": 0.8949, "step": 11558 }, { "epoch": 0.59, "learning_rate": 7.682309327949848e-06, "loss": 0.8699, "step": 11559 }, { "epoch": 0.59, "learning_rate": 7.680708746692292e-06, "loss": 0.9293, "step": 11560 }, { "epoch": 0.59, "learning_rate": 7.679108228227077e-06, "loss": 0.9294, "step": 11561 }, { "epoch": 0.59, "learning_rate": 7.677507772597533e-06, "loss": 0.9322, "step": 11562 }, { "epoch": 0.59, "learning_rate": 7.675907379847e-06, "loss": 0.9275, "step": 11563 }, { "epoch": 0.59, "learning_rate": 7.674307050018801e-06, "loss": 0.9036, "step": 11564 }, { "epoch": 0.59, "learning_rate": 7.672706783156263e-06, "loss": 0.8825, "step": 11565 }, { "epoch": 0.59, "learning_rate": 7.671106579302714e-06, "loss": 0.9951, "step": 11566 }, { "epoch": 0.59, "learning_rate": 7.669506438501477e-06, "loss": 0.8539, "step": 11567 }, { "epoch": 0.59, "learning_rate": 7.66790636079587e-06, "loss": 0.852, "step": 11568 }, { "epoch": 0.59, "learning_rate": 7.666306346229221e-06, "loss": 1.0132, "step": 11569 }, { "epoch": 0.59, "learning_rate": 7.664706394844845e-06, "loss": 0.8252, "step": 11570 }, { "epoch": 0.59, "learning_rate": 7.663106506686057e-06, "loss": 0.9131, "step": 11571 }, { "epoch": 0.59, "learning_rate": 7.661506681796176e-06, "loss": 0.9289, "step": 11572 }, { "epoch": 0.59, "learning_rate": 7.659906920218514e-06, "loss": 0.9222, "step": 11573 }, { "epoch": 0.59, "learning_rate": 7.658307221996377e-06, "loss": 0.9358, "step": 11574 }, { "epoch": 0.59, "learning_rate": 7.656707587173085e-06, "loss": 0.8912, "step": 11575 }, { "epoch": 0.59, "learning_rate": 7.655108015791943e-06, "loss": 0.9037, "step": 11576 }, { "epoch": 0.59, "learning_rate": 7.653508507896255e-06, "loss": 0.9603, "step": 11577 }, { "epoch": 0.59, "learning_rate": 7.651909063529328e-06, "loss": 0.8909, "step": 11578 }, { "epoch": 0.59, "learning_rate": 7.650309682734468e-06, "loss": 0.9613, "step": 11579 }, { "epoch": 0.59, "learning_rate": 7.648710365554965e-06, "loss": 0.955, "step": 11580 }, { "epoch": 0.59, "learning_rate": 7.647111112034134e-06, "loss": 0.9321, "step": 11581 }, { "epoch": 0.59, "learning_rate": 7.645511922215266e-06, "loss": 0.8833, "step": 11582 }, { "epoch": 0.59, "learning_rate": 7.643912796141655e-06, "loss": 0.8943, "step": 11583 }, { "epoch": 0.59, "learning_rate": 7.642313733856603e-06, "loss": 0.9558, "step": 11584 }, { "epoch": 0.59, "learning_rate": 7.640714735403395e-06, "loss": 0.7739, "step": 11585 }, { "epoch": 0.59, "learning_rate": 7.639115800825323e-06, "loss": 0.9, "step": 11586 }, { "epoch": 0.59, "learning_rate": 7.637516930165683e-06, "loss": 0.8271, "step": 11587 }, { "epoch": 0.59, "learning_rate": 7.635918123467758e-06, "loss": 0.8887, "step": 11588 }, { "epoch": 0.59, "learning_rate": 7.634319380774833e-06, "loss": 1.0148, "step": 11589 }, { "epoch": 0.59, "learning_rate": 7.632720702130196e-06, "loss": 0.9513, "step": 11590 }, { "epoch": 0.59, "learning_rate": 7.631122087577127e-06, "loss": 0.9998, "step": 11591 }, { "epoch": 0.59, "learning_rate": 7.629523537158903e-06, "loss": 1.0244, "step": 11592 }, { "epoch": 0.59, "learning_rate": 7.627925050918812e-06, "loss": 0.9635, "step": 11593 }, { "epoch": 0.59, "learning_rate": 7.626326628900126e-06, "loss": 0.8477, "step": 11594 }, { "epoch": 0.59, "learning_rate": 7.62472827114612e-06, "loss": 0.8831, "step": 11595 }, { "epoch": 0.59, "learning_rate": 7.62312997770007e-06, "loss": 0.8307, "step": 11596 }, { "epoch": 0.59, "learning_rate": 7.621531748605243e-06, "loss": 0.9284, "step": 11597 }, { "epoch": 0.59, "learning_rate": 7.619933583904918e-06, "loss": 0.9296, "step": 11598 }, { "epoch": 0.59, "learning_rate": 7.618335483642358e-06, "loss": 0.9181, "step": 11599 }, { "epoch": 0.59, "learning_rate": 7.6167374478608305e-06, "loss": 0.881, "step": 11600 }, { "epoch": 0.59, "learning_rate": 7.615139476603602e-06, "loss": 0.8654, "step": 11601 }, { "epoch": 0.59, "learning_rate": 7.613541569913936e-06, "loss": 0.8683, "step": 11602 }, { "epoch": 0.59, "learning_rate": 7.611943727835087e-06, "loss": 0.9063, "step": 11603 }, { "epoch": 0.59, "learning_rate": 7.610345950410326e-06, "loss": 0.8863, "step": 11604 }, { "epoch": 0.59, "learning_rate": 7.608748237682906e-06, "loss": 0.7896, "step": 11605 }, { "epoch": 0.59, "learning_rate": 7.6071505896960815e-06, "loss": 0.9399, "step": 11606 }, { "epoch": 0.59, "learning_rate": 7.605553006493112e-06, "loss": 0.9076, "step": 11607 }, { "epoch": 0.59, "learning_rate": 7.603955488117245e-06, "loss": 0.988, "step": 11608 }, { "epoch": 0.59, "learning_rate": 7.602358034611731e-06, "loss": 0.8871, "step": 11609 }, { "epoch": 0.59, "learning_rate": 7.600760646019827e-06, "loss": 0.9455, "step": 11610 }, { "epoch": 0.59, "learning_rate": 7.599163322384775e-06, "loss": 0.8247, "step": 11611 }, { "epoch": 0.59, "learning_rate": 7.597566063749821e-06, "loss": 0.8393, "step": 11612 }, { "epoch": 0.59, "learning_rate": 7.59596887015821e-06, "loss": 0.9852, "step": 11613 }, { "epoch": 0.59, "learning_rate": 7.594371741653186e-06, "loss": 0.7857, "step": 11614 }, { "epoch": 0.59, "learning_rate": 7.592774678277982e-06, "loss": 0.9609, "step": 11615 }, { "epoch": 0.59, "learning_rate": 7.591177680075847e-06, "loss": 0.9036, "step": 11616 }, { "epoch": 0.59, "learning_rate": 7.589580747090015e-06, "loss": 0.9361, "step": 11617 }, { "epoch": 0.59, "learning_rate": 7.5879838793637165e-06, "loss": 0.9813, "step": 11618 }, { "epoch": 0.59, "learning_rate": 7.5863870769401905e-06, "loss": 1.0034, "step": 11619 }, { "epoch": 0.59, "learning_rate": 7.584790339862666e-06, "loss": 0.7751, "step": 11620 }, { "epoch": 0.59, "learning_rate": 7.58319366817437e-06, "loss": 0.8239, "step": 11621 }, { "epoch": 0.59, "learning_rate": 7.5815970619185376e-06, "loss": 0.8771, "step": 11622 }, { "epoch": 0.59, "learning_rate": 7.580000521138393e-06, "loss": 0.9733, "step": 11623 }, { "epoch": 0.59, "learning_rate": 7.578404045877157e-06, "loss": 0.8715, "step": 11624 }, { "epoch": 0.59, "learning_rate": 7.576807636178056e-06, "loss": 0.8313, "step": 11625 }, { "epoch": 0.59, "learning_rate": 7.575211292084313e-06, "loss": 0.9044, "step": 11626 }, { "epoch": 0.59, "learning_rate": 7.573615013639137e-06, "loss": 0.8997, "step": 11627 }, { "epoch": 0.59, "learning_rate": 7.5720188008857586e-06, "loss": 0.7928, "step": 11628 }, { "epoch": 0.59, "learning_rate": 7.570422653867388e-06, "loss": 0.9189, "step": 11629 }, { "epoch": 0.59, "learning_rate": 7.568826572627237e-06, "loss": 0.9344, "step": 11630 }, { "epoch": 0.59, "learning_rate": 7.5672305572085225e-06, "loss": 0.9142, "step": 11631 }, { "epoch": 0.59, "learning_rate": 7.565634607654453e-06, "loss": 0.8165, "step": 11632 }, { "epoch": 0.59, "learning_rate": 7.564038724008232e-06, "loss": 0.8811, "step": 11633 }, { "epoch": 0.59, "learning_rate": 7.562442906313073e-06, "loss": 0.8749, "step": 11634 }, { "epoch": 0.59, "learning_rate": 7.560847154612181e-06, "loss": 0.8782, "step": 11635 }, { "epoch": 0.59, "learning_rate": 7.559251468948755e-06, "loss": 0.99, "step": 11636 }, { "epoch": 0.59, "learning_rate": 7.557655849366e-06, "loss": 0.9008, "step": 11637 }, { "epoch": 0.59, "learning_rate": 7.556060295907115e-06, "loss": 0.9211, "step": 11638 }, { "epoch": 0.59, "learning_rate": 7.554464808615294e-06, "loss": 0.9352, "step": 11639 }, { "epoch": 0.59, "learning_rate": 7.5528693875337385e-06, "loss": 0.8769, "step": 11640 }, { "epoch": 0.59, "learning_rate": 7.551274032705642e-06, "loss": 0.9845, "step": 11641 }, { "epoch": 0.59, "learning_rate": 7.549678744174194e-06, "loss": 0.841, "step": 11642 }, { "epoch": 0.59, "learning_rate": 7.54808352198259e-06, "loss": 0.8922, "step": 11643 }, { "epoch": 0.59, "learning_rate": 7.546488366174014e-06, "loss": 0.9491, "step": 11644 }, { "epoch": 0.59, "learning_rate": 7.5448932767916515e-06, "loss": 0.9195, "step": 11645 }, { "epoch": 0.59, "learning_rate": 7.543298253878697e-06, "loss": 0.8763, "step": 11646 }, { "epoch": 0.59, "learning_rate": 7.541703297478327e-06, "loss": 0.8294, "step": 11647 }, { "epoch": 0.59, "learning_rate": 7.5401084076337235e-06, "loss": 0.8856, "step": 11648 }, { "epoch": 0.59, "learning_rate": 7.53851358438807e-06, "loss": 0.8737, "step": 11649 }, { "epoch": 0.59, "learning_rate": 7.536918827784542e-06, "loss": 0.8316, "step": 11650 }, { "epoch": 0.59, "learning_rate": 7.5353241378663126e-06, "loss": 0.8965, "step": 11651 }, { "epoch": 0.59, "learning_rate": 7.533729514676563e-06, "loss": 0.8846, "step": 11652 }, { "epoch": 0.59, "learning_rate": 7.532134958258464e-06, "loss": 0.9012, "step": 11653 }, { "epoch": 0.59, "learning_rate": 7.530540468655182e-06, "loss": 0.9684, "step": 11654 }, { "epoch": 0.59, "learning_rate": 7.528946045909893e-06, "loss": 0.8333, "step": 11655 }, { "epoch": 0.59, "learning_rate": 7.52735169006576e-06, "loss": 0.802, "step": 11656 }, { "epoch": 0.59, "learning_rate": 7.525757401165945e-06, "loss": 0.8854, "step": 11657 }, { "epoch": 0.59, "learning_rate": 7.5241631792536206e-06, "loss": 0.8645, "step": 11658 }, { "epoch": 0.59, "learning_rate": 7.522569024371944e-06, "loss": 0.9202, "step": 11659 }, { "epoch": 0.59, "learning_rate": 7.520974936564074e-06, "loss": 0.8406, "step": 11660 }, { "epoch": 0.59, "learning_rate": 7.519380915873171e-06, "loss": 0.939, "step": 11661 }, { "epoch": 0.59, "learning_rate": 7.51778696234239e-06, "loss": 1.0123, "step": 11662 }, { "epoch": 0.59, "learning_rate": 7.516193076014884e-06, "loss": 0.8579, "step": 11663 }, { "epoch": 0.59, "learning_rate": 7.51459925693381e-06, "loss": 0.8918, "step": 11664 }, { "epoch": 0.59, "learning_rate": 7.513005505142317e-06, "loss": 0.9048, "step": 11665 }, { "epoch": 0.59, "learning_rate": 7.511411820683554e-06, "loss": 0.9483, "step": 11666 }, { "epoch": 0.59, "learning_rate": 7.50981820360067e-06, "loss": 0.9003, "step": 11667 }, { "epoch": 0.59, "learning_rate": 7.508224653936803e-06, "loss": 0.9247, "step": 11668 }, { "epoch": 0.59, "learning_rate": 7.506631171735107e-06, "loss": 0.9343, "step": 11669 }, { "epoch": 0.59, "learning_rate": 7.50503775703872e-06, "loss": 0.7972, "step": 11670 }, { "epoch": 0.59, "learning_rate": 7.5034444098907785e-06, "loss": 0.8338, "step": 11671 }, { "epoch": 0.59, "learning_rate": 7.501851130334425e-06, "loss": 0.9944, "step": 11672 }, { "epoch": 0.59, "learning_rate": 7.5002579184127965e-06, "loss": 0.8557, "step": 11673 }, { "epoch": 0.59, "learning_rate": 7.49866477416902e-06, "loss": 0.9246, "step": 11674 }, { "epoch": 0.59, "learning_rate": 7.497071697646237e-06, "loss": 0.8422, "step": 11675 }, { "epoch": 0.59, "learning_rate": 7.4954786888875764e-06, "loss": 0.8888, "step": 11676 }, { "epoch": 0.59, "learning_rate": 7.4938857479361624e-06, "loss": 0.9167, "step": 11677 }, { "epoch": 0.59, "learning_rate": 7.4922928748351275e-06, "loss": 0.8515, "step": 11678 }, { "epoch": 0.59, "learning_rate": 7.490700069627596e-06, "loss": 1.0033, "step": 11679 }, { "epoch": 0.59, "learning_rate": 7.489107332356685e-06, "loss": 0.976, "step": 11680 }, { "epoch": 0.59, "learning_rate": 7.487514663065527e-06, "loss": 0.9595, "step": 11681 }, { "epoch": 0.59, "learning_rate": 7.485922061797235e-06, "loss": 0.8296, "step": 11682 }, { "epoch": 0.59, "learning_rate": 7.484329528594927e-06, "loss": 0.8988, "step": 11683 }, { "epoch": 0.59, "learning_rate": 7.482737063501724e-06, "loss": 0.871, "step": 11684 }, { "epoch": 0.59, "learning_rate": 7.481144666560736e-06, "loss": 0.905, "step": 11685 }, { "epoch": 0.59, "learning_rate": 7.47955233781507e-06, "loss": 0.9042, "step": 11686 }, { "epoch": 0.59, "learning_rate": 7.47796007730785e-06, "loss": 0.9168, "step": 11687 }, { "epoch": 0.59, "learning_rate": 7.476367885082176e-06, "loss": 0.8419, "step": 11688 }, { "epoch": 0.59, "learning_rate": 7.4747757611811546e-06, "loss": 0.9359, "step": 11689 }, { "epoch": 0.59, "learning_rate": 7.473183705647895e-06, "loss": 0.9716, "step": 11690 }, { "epoch": 0.59, "learning_rate": 7.4715917185255e-06, "loss": 0.9363, "step": 11691 }, { "epoch": 0.59, "learning_rate": 7.469999799857063e-06, "loss": 0.9544, "step": 11692 }, { "epoch": 0.59, "learning_rate": 7.468407949685695e-06, "loss": 0.9611, "step": 11693 }, { "epoch": 0.59, "learning_rate": 7.466816168054487e-06, "loss": 0.9168, "step": 11694 }, { "epoch": 0.59, "learning_rate": 7.4652244550065344e-06, "loss": 0.8738, "step": 11695 }, { "epoch": 0.59, "learning_rate": 7.463632810584937e-06, "loss": 0.8411, "step": 11696 }, { "epoch": 0.59, "learning_rate": 7.462041234832781e-06, "loss": 0.925, "step": 11697 }, { "epoch": 0.59, "learning_rate": 7.460449727793154e-06, "loss": 0.8681, "step": 11698 }, { "epoch": 0.59, "learning_rate": 7.4588582895091545e-06, "loss": 0.9055, "step": 11699 }, { "epoch": 0.59, "learning_rate": 7.457266920023861e-06, "loss": 0.8268, "step": 11700 }, { "epoch": 0.59, "learning_rate": 7.45567561938036e-06, "loss": 0.9416, "step": 11701 }, { "epoch": 0.59, "learning_rate": 7.454084387621736e-06, "loss": 0.8559, "step": 11702 }, { "epoch": 0.59, "learning_rate": 7.452493224791069e-06, "loss": 0.9335, "step": 11703 }, { "epoch": 0.59, "learning_rate": 7.450902130931432e-06, "loss": 0.9046, "step": 11704 }, { "epoch": 0.59, "learning_rate": 7.449311106085915e-06, "loss": 0.8087, "step": 11705 }, { "epoch": 0.59, "learning_rate": 7.447720150297584e-06, "loss": 0.8597, "step": 11706 }, { "epoch": 0.59, "learning_rate": 7.446129263609512e-06, "loss": 0.9105, "step": 11707 }, { "epoch": 0.59, "learning_rate": 7.4445384460647755e-06, "loss": 0.9125, "step": 11708 }, { "epoch": 0.59, "learning_rate": 7.442947697706442e-06, "loss": 0.881, "step": 11709 }, { "epoch": 0.59, "learning_rate": 7.441357018577575e-06, "loss": 0.8699, "step": 11710 }, { "epoch": 0.59, "learning_rate": 7.43976640872125e-06, "loss": 0.8705, "step": 11711 }, { "epoch": 0.6, "learning_rate": 7.438175868180524e-06, "loss": 0.9113, "step": 11712 }, { "epoch": 0.6, "learning_rate": 7.436585396998459e-06, "loss": 1.0004, "step": 11713 }, { "epoch": 0.6, "learning_rate": 7.434994995218119e-06, "loss": 1.0056, "step": 11714 }, { "epoch": 0.6, "learning_rate": 7.433404662882561e-06, "loss": 0.8031, "step": 11715 }, { "epoch": 0.6, "learning_rate": 7.431814400034835e-06, "loss": 0.946, "step": 11716 }, { "epoch": 0.6, "learning_rate": 7.430224206718008e-06, "loss": 0.8845, "step": 11717 }, { "epoch": 0.6, "learning_rate": 7.428634082975128e-06, "loss": 0.8959, "step": 11718 }, { "epoch": 0.6, "learning_rate": 7.42704402884924e-06, "loss": 0.8868, "step": 11719 }, { "epoch": 0.6, "learning_rate": 7.4254540443834e-06, "loss": 0.845, "step": 11720 }, { "epoch": 0.6, "learning_rate": 7.423864129620654e-06, "loss": 0.8711, "step": 11721 }, { "epoch": 0.6, "learning_rate": 7.422274284604041e-06, "loss": 0.9709, "step": 11722 }, { "epoch": 0.6, "learning_rate": 7.4206845093766124e-06, "loss": 0.8731, "step": 11723 }, { "epoch": 0.6, "learning_rate": 7.4190948039814085e-06, "loss": 0.8057, "step": 11724 }, { "epoch": 0.6, "learning_rate": 7.417505168461463e-06, "loss": 0.9138, "step": 11725 }, { "epoch": 0.6, "learning_rate": 7.41591560285982e-06, "loss": 0.9171, "step": 11726 }, { "epoch": 0.6, "learning_rate": 7.414326107219513e-06, "loss": 0.8888, "step": 11727 }, { "epoch": 0.6, "learning_rate": 7.412736681583571e-06, "loss": 0.9612, "step": 11728 }, { "epoch": 0.6, "learning_rate": 7.411147325995036e-06, "loss": 0.9638, "step": 11729 }, { "epoch": 0.6, "learning_rate": 7.409558040496932e-06, "loss": 0.8651, "step": 11730 }, { "epoch": 0.6, "learning_rate": 7.407968825132285e-06, "loss": 0.7472, "step": 11731 }, { "epoch": 0.6, "learning_rate": 7.406379679944129e-06, "loss": 0.7878, "step": 11732 }, { "epoch": 0.6, "learning_rate": 7.404790604975482e-06, "loss": 0.8779, "step": 11733 }, { "epoch": 0.6, "learning_rate": 7.403201600269364e-06, "loss": 1.0046, "step": 11734 }, { "epoch": 0.6, "learning_rate": 7.401612665868805e-06, "loss": 0.9592, "step": 11735 }, { "epoch": 0.6, "learning_rate": 7.400023801816816e-06, "loss": 0.9103, "step": 11736 }, { "epoch": 0.6, "learning_rate": 7.398435008156418e-06, "loss": 0.8388, "step": 11737 }, { "epoch": 0.6, "learning_rate": 7.396846284930626e-06, "loss": 0.9045, "step": 11738 }, { "epoch": 0.6, "learning_rate": 7.395257632182446e-06, "loss": 0.9411, "step": 11739 }, { "epoch": 0.6, "learning_rate": 7.393669049954898e-06, "loss": 0.8232, "step": 11740 }, { "epoch": 0.6, "learning_rate": 7.392080538290989e-06, "loss": 0.9422, "step": 11741 }, { "epoch": 0.6, "learning_rate": 7.390492097233721e-06, "loss": 0.9328, "step": 11742 }, { "epoch": 0.6, "learning_rate": 7.388903726826107e-06, "loss": 0.9685, "step": 11743 }, { "epoch": 0.6, "learning_rate": 7.387315427111145e-06, "loss": 0.8878, "step": 11744 }, { "epoch": 0.6, "learning_rate": 7.385727198131835e-06, "loss": 0.8895, "step": 11745 }, { "epoch": 0.6, "learning_rate": 7.384139039931183e-06, "loss": 0.964, "step": 11746 }, { "epoch": 0.6, "learning_rate": 7.3825509525521835e-06, "loss": 0.9292, "step": 11747 }, { "epoch": 0.6, "learning_rate": 7.380962936037831e-06, "loss": 0.9249, "step": 11748 }, { "epoch": 0.6, "learning_rate": 7.379374990431123e-06, "loss": 1.0225, "step": 11749 }, { "epoch": 0.6, "learning_rate": 7.377787115775048e-06, "loss": 0.8997, "step": 11750 }, { "epoch": 0.6, "learning_rate": 7.376199312112593e-06, "loss": 0.9297, "step": 11751 }, { "epoch": 0.6, "learning_rate": 7.3746115794867545e-06, "loss": 0.8949, "step": 11752 }, { "epoch": 0.6, "learning_rate": 7.373023917940514e-06, "loss": 0.8936, "step": 11753 }, { "epoch": 0.6, "learning_rate": 7.371436327516854e-06, "loss": 0.9231, "step": 11754 }, { "epoch": 0.6, "learning_rate": 7.369848808258762e-06, "loss": 0.9433, "step": 11755 }, { "epoch": 0.6, "learning_rate": 7.3682613602092144e-06, "loss": 0.9402, "step": 11756 }, { "epoch": 0.6, "learning_rate": 7.366673983411186e-06, "loss": 0.8897, "step": 11757 }, { "epoch": 0.6, "learning_rate": 7.3650866779076626e-06, "loss": 0.9589, "step": 11758 }, { "epoch": 0.6, "learning_rate": 7.363499443741613e-06, "loss": 0.8842, "step": 11759 }, { "epoch": 0.6, "learning_rate": 7.36191228095601e-06, "loss": 0.9047, "step": 11760 }, { "epoch": 0.6, "learning_rate": 7.360325189593828e-06, "loss": 0.8867, "step": 11761 }, { "epoch": 0.6, "learning_rate": 7.358738169698032e-06, "loss": 0.9369, "step": 11762 }, { "epoch": 0.6, "learning_rate": 7.357151221311585e-06, "loss": 0.8058, "step": 11763 }, { "epoch": 0.6, "learning_rate": 7.355564344477463e-06, "loss": 0.9445, "step": 11764 }, { "epoch": 0.6, "learning_rate": 7.353977539238621e-06, "loss": 0.8995, "step": 11765 }, { "epoch": 0.6, "learning_rate": 7.352390805638021e-06, "loss": 1.0188, "step": 11766 }, { "epoch": 0.6, "learning_rate": 7.3508041437186264e-06, "loss": 0.8775, "step": 11767 }, { "epoch": 0.6, "learning_rate": 7.349217553523389e-06, "loss": 0.9103, "step": 11768 }, { "epoch": 0.6, "learning_rate": 7.347631035095263e-06, "loss": 0.9121, "step": 11769 }, { "epoch": 0.6, "learning_rate": 7.346044588477208e-06, "loss": 0.9436, "step": 11770 }, { "epoch": 0.6, "learning_rate": 7.344458213712173e-06, "loss": 0.8769, "step": 11771 }, { "epoch": 0.6, "learning_rate": 7.342871910843105e-06, "loss": 0.8821, "step": 11772 }, { "epoch": 0.6, "learning_rate": 7.341285679912955e-06, "loss": 0.9281, "step": 11773 }, { "epoch": 0.6, "learning_rate": 7.339699520964665e-06, "loss": 0.8317, "step": 11774 }, { "epoch": 0.6, "learning_rate": 7.338113434041177e-06, "loss": 0.8884, "step": 11775 }, { "epoch": 0.6, "learning_rate": 7.3365274191854405e-06, "loss": 0.8478, "step": 11776 }, { "epoch": 0.6, "learning_rate": 7.334941476440389e-06, "loss": 0.8766, "step": 11777 }, { "epoch": 0.6, "learning_rate": 7.333355605848962e-06, "loss": 0.9691, "step": 11778 }, { "epoch": 0.6, "learning_rate": 7.3317698074540945e-06, "loss": 0.9174, "step": 11779 }, { "epoch": 0.6, "learning_rate": 7.330184081298721e-06, "loss": 0.8338, "step": 11780 }, { "epoch": 0.6, "learning_rate": 7.32859842742577e-06, "loss": 0.8805, "step": 11781 }, { "epoch": 0.6, "learning_rate": 7.3270128458781786e-06, "loss": 0.8329, "step": 11782 }, { "epoch": 0.6, "learning_rate": 7.325427336698869e-06, "loss": 1.0373, "step": 11783 }, { "epoch": 0.6, "learning_rate": 7.323841899930768e-06, "loss": 0.9051, "step": 11784 }, { "epoch": 0.6, "learning_rate": 7.322256535616803e-06, "loss": 0.838, "step": 11785 }, { "epoch": 0.6, "learning_rate": 7.320671243799893e-06, "loss": 1.026, "step": 11786 }, { "epoch": 0.6, "learning_rate": 7.319086024522952e-06, "loss": 0.8427, "step": 11787 }, { "epoch": 0.6, "learning_rate": 7.317500877828911e-06, "loss": 0.8435, "step": 11788 }, { "epoch": 0.6, "learning_rate": 7.31591580376068e-06, "loss": 0.8344, "step": 11789 }, { "epoch": 0.6, "learning_rate": 7.31433080236117e-06, "loss": 0.9127, "step": 11790 }, { "epoch": 0.6, "learning_rate": 7.312745873673299e-06, "loss": 0.9325, "step": 11791 }, { "epoch": 0.6, "learning_rate": 7.311161017739974e-06, "loss": 0.9171, "step": 11792 }, { "epoch": 0.6, "learning_rate": 7.3095762346041e-06, "loss": 0.9685, "step": 11793 }, { "epoch": 0.6, "learning_rate": 7.3079915243085905e-06, "loss": 0.8512, "step": 11794 }, { "epoch": 0.6, "learning_rate": 7.306406886896347e-06, "loss": 0.8253, "step": 11795 }, { "epoch": 0.6, "learning_rate": 7.304822322410269e-06, "loss": 0.8526, "step": 11796 }, { "epoch": 0.6, "learning_rate": 7.3032378308932615e-06, "loss": 0.9615, "step": 11797 }, { "epoch": 0.6, "learning_rate": 7.30165341238822e-06, "loss": 0.7954, "step": 11798 }, { "epoch": 0.6, "learning_rate": 7.300069066938038e-06, "loss": 0.8531, "step": 11799 }, { "epoch": 0.6, "learning_rate": 7.298484794585616e-06, "loss": 0.9271, "step": 11800 }, { "epoch": 0.6, "learning_rate": 7.296900595373846e-06, "loss": 0.823, "step": 11801 }, { "epoch": 0.6, "learning_rate": 7.295316469345615e-06, "loss": 0.999, "step": 11802 }, { "epoch": 0.6, "learning_rate": 7.293732416543813e-06, "loss": 0.9055, "step": 11803 }, { "epoch": 0.6, "learning_rate": 7.2921484370113234e-06, "loss": 0.8017, "step": 11804 }, { "epoch": 0.6, "learning_rate": 7.290564530791038e-06, "loss": 0.8527, "step": 11805 }, { "epoch": 0.6, "learning_rate": 7.2889806979258356e-06, "loss": 0.9397, "step": 11806 }, { "epoch": 0.6, "learning_rate": 7.287396938458594e-06, "loss": 0.9329, "step": 11807 }, { "epoch": 0.6, "learning_rate": 7.285813252432197e-06, "loss": 0.9321, "step": 11808 }, { "epoch": 0.6, "learning_rate": 7.284229639889518e-06, "loss": 0.8291, "step": 11809 }, { "epoch": 0.6, "learning_rate": 7.282646100873428e-06, "loss": 0.8329, "step": 11810 }, { "epoch": 0.6, "learning_rate": 7.281062635426808e-06, "loss": 0.9157, "step": 11811 }, { "epoch": 0.6, "learning_rate": 7.279479243592525e-06, "loss": 0.8311, "step": 11812 }, { "epoch": 0.6, "learning_rate": 7.277895925413443e-06, "loss": 0.8829, "step": 11813 }, { "epoch": 0.6, "learning_rate": 7.276312680932437e-06, "loss": 0.8787, "step": 11814 }, { "epoch": 0.6, "learning_rate": 7.274729510192367e-06, "loss": 0.9094, "step": 11815 }, { "epoch": 0.6, "learning_rate": 7.2731464132360905e-06, "loss": 0.9518, "step": 11816 }, { "epoch": 0.6, "learning_rate": 7.27156339010648e-06, "loss": 0.8687, "step": 11817 }, { "epoch": 0.6, "learning_rate": 7.269980440846387e-06, "loss": 0.9197, "step": 11818 }, { "epoch": 0.6, "learning_rate": 7.268397565498666e-06, "loss": 0.9371, "step": 11819 }, { "epoch": 0.6, "learning_rate": 7.266814764106178e-06, "loss": 0.8655, "step": 11820 }, { "epoch": 0.6, "learning_rate": 7.265232036711771e-06, "loss": 0.7835, "step": 11821 }, { "epoch": 0.6, "learning_rate": 7.263649383358294e-06, "loss": 0.801, "step": 11822 }, { "epoch": 0.6, "learning_rate": 7.262066804088604e-06, "loss": 0.9397, "step": 11823 }, { "epoch": 0.6, "learning_rate": 7.26048429894554e-06, "loss": 0.9509, "step": 11824 }, { "epoch": 0.6, "learning_rate": 7.2589018679719485e-06, "loss": 0.9898, "step": 11825 }, { "epoch": 0.6, "learning_rate": 7.257319511210674e-06, "loss": 0.9586, "step": 11826 }, { "epoch": 0.6, "learning_rate": 7.255737228704556e-06, "loss": 0.8888, "step": 11827 }, { "epoch": 0.6, "learning_rate": 7.254155020496427e-06, "loss": 0.9518, "step": 11828 }, { "epoch": 0.6, "learning_rate": 7.252572886629137e-06, "loss": 0.8275, "step": 11829 }, { "epoch": 0.6, "learning_rate": 7.25099082714551e-06, "loss": 0.8776, "step": 11830 }, { "epoch": 0.6, "learning_rate": 7.249408842088381e-06, "loss": 0.9122, "step": 11831 }, { "epoch": 0.6, "learning_rate": 7.247826931500584e-06, "loss": 0.8494, "step": 11832 }, { "epoch": 0.6, "learning_rate": 7.2462450954249445e-06, "loss": 0.8171, "step": 11833 }, { "epoch": 0.6, "learning_rate": 7.244663333904282e-06, "loss": 0.8946, "step": 11834 }, { "epoch": 0.6, "learning_rate": 7.243081646981436e-06, "loss": 0.9757, "step": 11835 }, { "epoch": 0.6, "learning_rate": 7.2415000346992206e-06, "loss": 0.9968, "step": 11836 }, { "epoch": 0.6, "learning_rate": 7.239918497100454e-06, "loss": 0.9244, "step": 11837 }, { "epoch": 0.6, "learning_rate": 7.23833703422796e-06, "loss": 0.9618, "step": 11838 }, { "epoch": 0.6, "learning_rate": 7.236755646124554e-06, "loss": 0.9131, "step": 11839 }, { "epoch": 0.6, "learning_rate": 7.235174332833044e-06, "loss": 0.8548, "step": 11840 }, { "epoch": 0.6, "learning_rate": 7.233593094396251e-06, "loss": 0.8155, "step": 11841 }, { "epoch": 0.6, "learning_rate": 7.232011930856982e-06, "loss": 1.0038, "step": 11842 }, { "epoch": 0.6, "learning_rate": 7.230430842258043e-06, "loss": 0.872, "step": 11843 }, { "epoch": 0.6, "learning_rate": 7.228849828642243e-06, "loss": 0.8885, "step": 11844 }, { "epoch": 0.6, "learning_rate": 7.227268890052388e-06, "loss": 1.0315, "step": 11845 }, { "epoch": 0.6, "learning_rate": 7.225688026531272e-06, "loss": 0.9797, "step": 11846 }, { "epoch": 0.6, "learning_rate": 7.224107238121707e-06, "loss": 0.7972, "step": 11847 }, { "epoch": 0.6, "learning_rate": 7.222526524866483e-06, "loss": 0.9194, "step": 11848 }, { "epoch": 0.6, "learning_rate": 7.220945886808397e-06, "loss": 0.9192, "step": 11849 }, { "epoch": 0.6, "learning_rate": 7.219365323990248e-06, "loss": 0.8479, "step": 11850 }, { "epoch": 0.6, "learning_rate": 7.217784836454823e-06, "loss": 0.9324, "step": 11851 }, { "epoch": 0.6, "learning_rate": 7.216204424244911e-06, "loss": 0.897, "step": 11852 }, { "epoch": 0.6, "learning_rate": 7.214624087403304e-06, "loss": 0.8772, "step": 11853 }, { "epoch": 0.6, "learning_rate": 7.213043825972787e-06, "loss": 0.9515, "step": 11854 }, { "epoch": 0.6, "learning_rate": 7.211463639996143e-06, "loss": 0.8409, "step": 11855 }, { "epoch": 0.6, "learning_rate": 7.209883529516155e-06, "loss": 0.8731, "step": 11856 }, { "epoch": 0.6, "learning_rate": 7.208303494575602e-06, "loss": 0.9006, "step": 11857 }, { "epoch": 0.6, "learning_rate": 7.206723535217258e-06, "loss": 0.8611, "step": 11858 }, { "epoch": 0.6, "learning_rate": 7.2051436514839064e-06, "loss": 0.7813, "step": 11859 }, { "epoch": 0.6, "learning_rate": 7.2035638434183176e-06, "loss": 0.8393, "step": 11860 }, { "epoch": 0.6, "learning_rate": 7.20198411106326e-06, "loss": 0.877, "step": 11861 }, { "epoch": 0.6, "learning_rate": 7.200404454461507e-06, "loss": 0.9011, "step": 11862 }, { "epoch": 0.6, "learning_rate": 7.198824873655827e-06, "loss": 0.8227, "step": 11863 }, { "epoch": 0.6, "learning_rate": 7.197245368688976e-06, "loss": 0.9304, "step": 11864 }, { "epoch": 0.6, "learning_rate": 7.195665939603731e-06, "loss": 0.8027, "step": 11865 }, { "epoch": 0.6, "learning_rate": 7.194086586442848e-06, "loss": 0.9274, "step": 11866 }, { "epoch": 0.6, "learning_rate": 7.192507309249081e-06, "loss": 0.916, "step": 11867 }, { "epoch": 0.6, "learning_rate": 7.190928108065196e-06, "loss": 0.8743, "step": 11868 }, { "epoch": 0.6, "learning_rate": 7.189348982933941e-06, "loss": 0.9193, "step": 11869 }, { "epoch": 0.6, "learning_rate": 7.18776993389807e-06, "loss": 0.8149, "step": 11870 }, { "epoch": 0.6, "learning_rate": 7.186190961000341e-06, "loss": 0.8688, "step": 11871 }, { "epoch": 0.6, "learning_rate": 7.184612064283496e-06, "loss": 0.9348, "step": 11872 }, { "epoch": 0.6, "learning_rate": 7.183033243790283e-06, "loss": 0.8909, "step": 11873 }, { "epoch": 0.6, "learning_rate": 7.18145449956345e-06, "loss": 0.9595, "step": 11874 }, { "epoch": 0.6, "learning_rate": 7.179875831645733e-06, "loss": 0.9329, "step": 11875 }, { "epoch": 0.6, "learning_rate": 7.1782972400798825e-06, "loss": 0.9586, "step": 11876 }, { "epoch": 0.6, "learning_rate": 7.176718724908631e-06, "loss": 0.9114, "step": 11877 }, { "epoch": 0.6, "learning_rate": 7.175140286174716e-06, "loss": 0.8864, "step": 11878 }, { "epoch": 0.6, "learning_rate": 7.173561923920873e-06, "loss": 0.9485, "step": 11879 }, { "epoch": 0.6, "learning_rate": 7.171983638189835e-06, "loss": 0.8795, "step": 11880 }, { "epoch": 0.6, "learning_rate": 7.170405429024325e-06, "loss": 0.8203, "step": 11881 }, { "epoch": 0.6, "learning_rate": 7.1688272964670826e-06, "loss": 0.8581, "step": 11882 }, { "epoch": 0.6, "learning_rate": 7.1672492405608295e-06, "loss": 0.8695, "step": 11883 }, { "epoch": 0.6, "learning_rate": 7.165671261348287e-06, "loss": 0.8875, "step": 11884 }, { "epoch": 0.6, "learning_rate": 7.164093358872182e-06, "loss": 0.8466, "step": 11885 }, { "epoch": 0.6, "learning_rate": 7.162515533175232e-06, "loss": 0.8752, "step": 11886 }, { "epoch": 0.6, "learning_rate": 7.1609377843001505e-06, "loss": 0.8662, "step": 11887 }, { "epoch": 0.6, "learning_rate": 7.159360112289663e-06, "loss": 0.8906, "step": 11888 }, { "epoch": 0.6, "learning_rate": 7.157782517186478e-06, "loss": 0.8909, "step": 11889 }, { "epoch": 0.6, "learning_rate": 7.156204999033305e-06, "loss": 0.8158, "step": 11890 }, { "epoch": 0.6, "learning_rate": 7.154627557872857e-06, "loss": 0.9452, "step": 11891 }, { "epoch": 0.6, "learning_rate": 7.153050193747844e-06, "loss": 0.8532, "step": 11892 }, { "epoch": 0.6, "learning_rate": 7.15147290670096e-06, "loss": 0.9511, "step": 11893 }, { "epoch": 0.6, "learning_rate": 7.149895696774924e-06, "loss": 0.957, "step": 11894 }, { "epoch": 0.6, "learning_rate": 7.1483185640124275e-06, "loss": 0.9132, "step": 11895 }, { "epoch": 0.6, "learning_rate": 7.146741508456171e-06, "loss": 0.8925, "step": 11896 }, { "epoch": 0.6, "learning_rate": 7.145164530148854e-06, "loss": 0.8905, "step": 11897 }, { "epoch": 0.6, "learning_rate": 7.143587629133171e-06, "loss": 0.926, "step": 11898 }, { "epoch": 0.6, "learning_rate": 7.142010805451809e-06, "loss": 1.0657, "step": 11899 }, { "epoch": 0.6, "learning_rate": 7.140434059147468e-06, "loss": 0.8963, "step": 11900 }, { "epoch": 0.6, "learning_rate": 7.138857390262832e-06, "loss": 0.8357, "step": 11901 }, { "epoch": 0.6, "learning_rate": 7.1372807988405865e-06, "loss": 0.9476, "step": 11902 }, { "epoch": 0.6, "learning_rate": 7.135704284923419e-06, "loss": 0.8688, "step": 11903 }, { "epoch": 0.6, "learning_rate": 7.134127848554012e-06, "loss": 0.9847, "step": 11904 }, { "epoch": 0.6, "learning_rate": 7.132551489775039e-06, "loss": 0.8918, "step": 11905 }, { "epoch": 0.6, "learning_rate": 7.1309752086291895e-06, "loss": 0.9133, "step": 11906 }, { "epoch": 0.6, "learning_rate": 7.1293990051591325e-06, "loss": 0.9227, "step": 11907 }, { "epoch": 0.6, "learning_rate": 7.1278228794075425e-06, "loss": 0.8988, "step": 11908 }, { "epoch": 0.61, "learning_rate": 7.126246831417093e-06, "loss": 0.825, "step": 11909 }, { "epoch": 0.61, "learning_rate": 7.124670861230454e-06, "loss": 0.9328, "step": 11910 }, { "epoch": 0.61, "learning_rate": 7.1230949688902875e-06, "loss": 0.8572, "step": 11911 }, { "epoch": 0.61, "learning_rate": 7.121519154439268e-06, "loss": 0.9424, "step": 11912 }, { "epoch": 0.61, "learning_rate": 7.119943417920055e-06, "loss": 0.898, "step": 11913 }, { "epoch": 0.61, "learning_rate": 7.118367759375308e-06, "loss": 0.9642, "step": 11914 }, { "epoch": 0.61, "learning_rate": 7.1167921788476915e-06, "loss": 0.9803, "step": 11915 }, { "epoch": 0.61, "learning_rate": 7.115216676379857e-06, "loss": 0.9378, "step": 11916 }, { "epoch": 0.61, "learning_rate": 7.113641252014458e-06, "loss": 0.8734, "step": 11917 }, { "epoch": 0.61, "learning_rate": 7.1120659057941564e-06, "loss": 0.8145, "step": 11918 }, { "epoch": 0.61, "learning_rate": 7.1104906377615976e-06, "loss": 0.9166, "step": 11919 }, { "epoch": 0.61, "learning_rate": 7.108915447959429e-06, "loss": 0.9499, "step": 11920 }, { "epoch": 0.61, "learning_rate": 7.107340336430301e-06, "loss": 0.7956, "step": 11921 }, { "epoch": 0.61, "learning_rate": 7.105765303216854e-06, "loss": 0.9174, "step": 11922 }, { "epoch": 0.61, "learning_rate": 7.10419034836173e-06, "loss": 0.9461, "step": 11923 }, { "epoch": 0.61, "learning_rate": 7.102615471907575e-06, "loss": 0.897, "step": 11924 }, { "epoch": 0.61, "learning_rate": 7.101040673897023e-06, "loss": 0.9147, "step": 11925 }, { "epoch": 0.61, "learning_rate": 7.099465954372708e-06, "loss": 0.8228, "step": 11926 }, { "epoch": 0.61, "learning_rate": 7.09789131337727e-06, "loss": 0.8222, "step": 11927 }, { "epoch": 0.61, "learning_rate": 7.096316750953336e-06, "loss": 0.8808, "step": 11928 }, { "epoch": 0.61, "learning_rate": 7.094742267143533e-06, "loss": 0.9024, "step": 11929 }, { "epoch": 0.61, "learning_rate": 7.093167861990495e-06, "loss": 0.9467, "step": 11930 }, { "epoch": 0.61, "learning_rate": 7.091593535536847e-06, "loss": 0.9216, "step": 11931 }, { "epoch": 0.61, "learning_rate": 7.090019287825206e-06, "loss": 0.9749, "step": 11932 }, { "epoch": 0.61, "learning_rate": 7.0884451188982e-06, "loss": 0.9501, "step": 11933 }, { "epoch": 0.61, "learning_rate": 7.086871028798443e-06, "loss": 0.8575, "step": 11934 }, { "epoch": 0.61, "learning_rate": 7.085297017568551e-06, "loss": 0.9168, "step": 11935 }, { "epoch": 0.61, "learning_rate": 7.083723085251144e-06, "loss": 0.9909, "step": 11936 }, { "epoch": 0.61, "learning_rate": 7.082149231888833e-06, "loss": 0.9281, "step": 11937 }, { "epoch": 0.61, "learning_rate": 7.0805754575242234e-06, "loss": 0.9526, "step": 11938 }, { "epoch": 0.61, "learning_rate": 7.079001762199931e-06, "loss": 0.9096, "step": 11939 }, { "epoch": 0.61, "learning_rate": 7.0774281459585585e-06, "loss": 0.9512, "step": 11940 }, { "epoch": 0.61, "learning_rate": 7.075854608842705e-06, "loss": 0.959, "step": 11941 }, { "epoch": 0.61, "learning_rate": 7.074281150894981e-06, "loss": 0.8489, "step": 11942 }, { "epoch": 0.61, "learning_rate": 7.072707772157978e-06, "loss": 0.8998, "step": 11943 }, { "epoch": 0.61, "learning_rate": 7.071134472674302e-06, "loss": 0.9658, "step": 11944 }, { "epoch": 0.61, "learning_rate": 7.069561252486544e-06, "loss": 1.0112, "step": 11945 }, { "epoch": 0.61, "learning_rate": 7.067988111637294e-06, "loss": 0.9062, "step": 11946 }, { "epoch": 0.61, "learning_rate": 7.066415050169148e-06, "loss": 0.9361, "step": 11947 }, { "epoch": 0.61, "learning_rate": 7.064842068124695e-06, "loss": 0.871, "step": 11948 }, { "epoch": 0.61, "learning_rate": 7.063269165546518e-06, "loss": 0.7935, "step": 11949 }, { "epoch": 0.61, "learning_rate": 7.061696342477207e-06, "loss": 0.8827, "step": 11950 }, { "epoch": 0.61, "learning_rate": 7.060123598959341e-06, "loss": 0.9484, "step": 11951 }, { "epoch": 0.61, "learning_rate": 7.058550935035495e-06, "loss": 0.9678, "step": 11952 }, { "epoch": 0.61, "learning_rate": 7.056978350748258e-06, "loss": 0.8505, "step": 11953 }, { "epoch": 0.61, "learning_rate": 7.055405846140202e-06, "loss": 0.9344, "step": 11954 }, { "epoch": 0.61, "learning_rate": 7.053833421253897e-06, "loss": 0.9647, "step": 11955 }, { "epoch": 0.61, "learning_rate": 7.052261076131919e-06, "loss": 0.921, "step": 11956 }, { "epoch": 0.61, "learning_rate": 7.050688810816837e-06, "loss": 0.8157, "step": 11957 }, { "epoch": 0.61, "learning_rate": 7.049116625351213e-06, "loss": 0.9323, "step": 11958 }, { "epoch": 0.61, "learning_rate": 7.047544519777623e-06, "loss": 0.8806, "step": 11959 }, { "epoch": 0.61, "learning_rate": 7.045972494138622e-06, "loss": 0.8863, "step": 11960 }, { "epoch": 0.61, "learning_rate": 7.044400548476771e-06, "loss": 0.919, "step": 11961 }, { "epoch": 0.61, "learning_rate": 7.042828682834633e-06, "loss": 0.9447, "step": 11962 }, { "epoch": 0.61, "learning_rate": 7.041256897254763e-06, "loss": 0.8702, "step": 11963 }, { "epoch": 0.61, "learning_rate": 7.039685191779709e-06, "loss": 0.9061, "step": 11964 }, { "epoch": 0.61, "learning_rate": 7.038113566452034e-06, "loss": 0.8083, "step": 11965 }, { "epoch": 0.61, "learning_rate": 7.036542021314283e-06, "loss": 0.942, "step": 11966 }, { "epoch": 0.61, "learning_rate": 7.0349705564090025e-06, "loss": 0.9847, "step": 11967 }, { "epoch": 0.61, "learning_rate": 7.03339917177874e-06, "loss": 0.9463, "step": 11968 }, { "epoch": 0.61, "learning_rate": 7.031827867466039e-06, "loss": 0.8936, "step": 11969 }, { "epoch": 0.61, "learning_rate": 7.030256643513438e-06, "loss": 0.9319, "step": 11970 }, { "epoch": 0.61, "learning_rate": 7.0286854999634815e-06, "loss": 0.7755, "step": 11971 }, { "epoch": 0.61, "learning_rate": 7.0271144368587045e-06, "loss": 0.8099, "step": 11972 }, { "epoch": 0.61, "learning_rate": 7.025543454241639e-06, "loss": 0.9226, "step": 11973 }, { "epoch": 0.61, "learning_rate": 7.023972552154821e-06, "loss": 0.8918, "step": 11974 }, { "epoch": 0.61, "learning_rate": 7.022401730640782e-06, "loss": 1.0125, "step": 11975 }, { "epoch": 0.61, "learning_rate": 7.020830989742043e-06, "loss": 0.8746, "step": 11976 }, { "epoch": 0.61, "learning_rate": 7.019260329501139e-06, "loss": 0.8705, "step": 11977 }, { "epoch": 0.61, "learning_rate": 7.01768974996059e-06, "loss": 0.9329, "step": 11978 }, { "epoch": 0.61, "learning_rate": 7.0161192511629176e-06, "loss": 0.9167, "step": 11979 }, { "epoch": 0.61, "learning_rate": 7.014548833150643e-06, "loss": 0.8169, "step": 11980 }, { "epoch": 0.61, "learning_rate": 7.0129784959662825e-06, "loss": 0.8437, "step": 11981 }, { "epoch": 0.61, "learning_rate": 7.011408239652346e-06, "loss": 0.8355, "step": 11982 }, { "epoch": 0.61, "learning_rate": 7.009838064251359e-06, "loss": 0.973, "step": 11983 }, { "epoch": 0.61, "learning_rate": 7.008267969805822e-06, "loss": 0.8611, "step": 11984 }, { "epoch": 0.61, "learning_rate": 7.006697956358246e-06, "loss": 0.8891, "step": 11985 }, { "epoch": 0.61, "learning_rate": 7.005128023951141e-06, "loss": 0.9054, "step": 11986 }, { "epoch": 0.61, "learning_rate": 7.003558172627007e-06, "loss": 0.9037, "step": 11987 }, { "epoch": 0.61, "learning_rate": 7.001988402428345e-06, "loss": 0.9234, "step": 11988 }, { "epoch": 0.61, "learning_rate": 7.000418713397661e-06, "loss": 0.8521, "step": 11989 }, { "epoch": 0.61, "learning_rate": 6.998849105577447e-06, "loss": 0.7624, "step": 11990 }, { "epoch": 0.61, "learning_rate": 6.997279579010201e-06, "loss": 0.7965, "step": 11991 }, { "epoch": 0.61, "learning_rate": 6.995710133738417e-06, "loss": 0.9348, "step": 11992 }, { "epoch": 0.61, "learning_rate": 6.994140769804585e-06, "loss": 0.9575, "step": 11993 }, { "epoch": 0.61, "learning_rate": 6.9925714872511895e-06, "loss": 0.9647, "step": 11994 }, { "epoch": 0.61, "learning_rate": 6.991002286120725e-06, "loss": 0.8459, "step": 11995 }, { "epoch": 0.61, "learning_rate": 6.989433166455672e-06, "loss": 0.9113, "step": 11996 }, { "epoch": 0.61, "learning_rate": 6.987864128298511e-06, "loss": 0.8883, "step": 11997 }, { "epoch": 0.61, "learning_rate": 6.986295171691727e-06, "loss": 0.9187, "step": 11998 }, { "epoch": 0.61, "learning_rate": 6.984726296677795e-06, "loss": 0.8601, "step": 11999 }, { "epoch": 0.61, "learning_rate": 6.983157503299185e-06, "loss": 0.9176, "step": 12000 }, { "epoch": 0.61, "learning_rate": 6.981588791598381e-06, "loss": 0.8833, "step": 12001 }, { "epoch": 0.61, "learning_rate": 6.980020161617849e-06, "loss": 0.9038, "step": 12002 }, { "epoch": 0.61, "learning_rate": 6.978451613400057e-06, "loss": 0.9046, "step": 12003 }, { "epoch": 0.61, "learning_rate": 6.976883146987474e-06, "loss": 0.9662, "step": 12004 }, { "epoch": 0.61, "learning_rate": 6.9753147624225635e-06, "loss": 0.8905, "step": 12005 }, { "epoch": 0.61, "learning_rate": 6.9737464597477835e-06, "loss": 0.8944, "step": 12006 }, { "epoch": 0.61, "learning_rate": 6.9721782390056026e-06, "loss": 0.8407, "step": 12007 }, { "epoch": 0.61, "learning_rate": 6.970610100238475e-06, "loss": 0.8958, "step": 12008 }, { "epoch": 0.61, "learning_rate": 6.969042043488854e-06, "loss": 0.9882, "step": 12009 }, { "epoch": 0.61, "learning_rate": 6.967474068799196e-06, "loss": 0.8657, "step": 12010 }, { "epoch": 0.61, "learning_rate": 6.965906176211951e-06, "loss": 0.8475, "step": 12011 }, { "epoch": 0.61, "learning_rate": 6.9643383657695655e-06, "loss": 0.9446, "step": 12012 }, { "epoch": 0.61, "learning_rate": 6.9627706375144935e-06, "loss": 1.0033, "step": 12013 }, { "epoch": 0.61, "learning_rate": 6.961202991489171e-06, "loss": 0.9344, "step": 12014 }, { "epoch": 0.61, "learning_rate": 6.959635427736049e-06, "loss": 0.9401, "step": 12015 }, { "epoch": 0.61, "learning_rate": 6.958067946297562e-06, "loss": 0.9156, "step": 12016 }, { "epoch": 0.61, "learning_rate": 6.956500547216144e-06, "loss": 0.9697, "step": 12017 }, { "epoch": 0.61, "learning_rate": 6.9549332305342395e-06, "loss": 0.9543, "step": 12018 }, { "epoch": 0.61, "learning_rate": 6.9533659962942805e-06, "loss": 0.9272, "step": 12019 }, { "epoch": 0.61, "learning_rate": 6.951798844538692e-06, "loss": 0.7816, "step": 12020 }, { "epoch": 0.61, "learning_rate": 6.950231775309909e-06, "loss": 0.7464, "step": 12021 }, { "epoch": 0.61, "learning_rate": 6.948664788650357e-06, "loss": 0.9126, "step": 12022 }, { "epoch": 0.61, "learning_rate": 6.947097884602454e-06, "loss": 0.8607, "step": 12023 }, { "epoch": 0.61, "learning_rate": 6.945531063208632e-06, "loss": 0.9062, "step": 12024 }, { "epoch": 0.61, "learning_rate": 6.943964324511308e-06, "loss": 0.843, "step": 12025 }, { "epoch": 0.61, "learning_rate": 6.942397668552895e-06, "loss": 0.8732, "step": 12026 }, { "epoch": 0.61, "learning_rate": 6.940831095375815e-06, "loss": 0.9356, "step": 12027 }, { "epoch": 0.61, "learning_rate": 6.939264605022478e-06, "loss": 0.9856, "step": 12028 }, { "epoch": 0.61, "learning_rate": 6.937698197535291e-06, "loss": 0.9285, "step": 12029 }, { "epoch": 0.61, "learning_rate": 6.936131872956672e-06, "loss": 0.838, "step": 12030 }, { "epoch": 0.61, "learning_rate": 6.934565631329022e-06, "loss": 0.8267, "step": 12031 }, { "epoch": 0.61, "learning_rate": 6.932999472694746e-06, "loss": 0.8527, "step": 12032 }, { "epoch": 0.61, "learning_rate": 6.931433397096247e-06, "loss": 0.9355, "step": 12033 }, { "epoch": 0.61, "learning_rate": 6.929867404575924e-06, "loss": 0.8593, "step": 12034 }, { "epoch": 0.61, "learning_rate": 6.92830149517617e-06, "loss": 0.9592, "step": 12035 }, { "epoch": 0.61, "learning_rate": 6.92673566893939e-06, "loss": 0.8807, "step": 12036 }, { "epoch": 0.61, "learning_rate": 6.925169925907972e-06, "loss": 0.8484, "step": 12037 }, { "epoch": 0.61, "learning_rate": 6.923604266124304e-06, "loss": 0.8864, "step": 12038 }, { "epoch": 0.61, "learning_rate": 6.922038689630781e-06, "loss": 0.9611, "step": 12039 }, { "epoch": 0.61, "learning_rate": 6.9204731964697855e-06, "loss": 0.897, "step": 12040 }, { "epoch": 0.61, "learning_rate": 6.918907786683697e-06, "loss": 0.9074, "step": 12041 }, { "epoch": 0.61, "learning_rate": 6.917342460314907e-06, "loss": 0.9158, "step": 12042 }, { "epoch": 0.61, "learning_rate": 6.9157772174057895e-06, "loss": 0.8335, "step": 12043 }, { "epoch": 0.61, "learning_rate": 6.91421205799872e-06, "loss": 0.9005, "step": 12044 }, { "epoch": 0.61, "learning_rate": 6.91264698213608e-06, "loss": 0.8792, "step": 12045 }, { "epoch": 0.61, "learning_rate": 6.911081989860235e-06, "loss": 0.9049, "step": 12046 }, { "epoch": 0.61, "learning_rate": 6.9095170812135574e-06, "loss": 0.9467, "step": 12047 }, { "epoch": 0.61, "learning_rate": 6.9079522562384205e-06, "loss": 0.98, "step": 12048 }, { "epoch": 0.61, "learning_rate": 6.906387514977186e-06, "loss": 0.9448, "step": 12049 }, { "epoch": 0.61, "learning_rate": 6.904822857472216e-06, "loss": 0.8759, "step": 12050 }, { "epoch": 0.61, "learning_rate": 6.903258283765877e-06, "loss": 0.8695, "step": 12051 }, { "epoch": 0.61, "learning_rate": 6.901693793900526e-06, "loss": 0.9069, "step": 12052 }, { "epoch": 0.61, "learning_rate": 6.9001293879185126e-06, "loss": 0.8747, "step": 12053 }, { "epoch": 0.61, "learning_rate": 6.898565065862204e-06, "loss": 0.8517, "step": 12054 }, { "epoch": 0.61, "learning_rate": 6.897000827773945e-06, "loss": 0.8913, "step": 12055 }, { "epoch": 0.61, "learning_rate": 6.895436673696088e-06, "loss": 0.8715, "step": 12056 }, { "epoch": 0.61, "learning_rate": 6.893872603670979e-06, "loss": 0.9173, "step": 12057 }, { "epoch": 0.61, "learning_rate": 6.892308617740966e-06, "loss": 0.9507, "step": 12058 }, { "epoch": 0.61, "learning_rate": 6.890744715948388e-06, "loss": 0.9126, "step": 12059 }, { "epoch": 0.61, "learning_rate": 6.889180898335591e-06, "loss": 0.9995, "step": 12060 }, { "epoch": 0.61, "learning_rate": 6.8876171649449135e-06, "loss": 0.8846, "step": 12061 }, { "epoch": 0.61, "learning_rate": 6.8860535158186845e-06, "loss": 0.9128, "step": 12062 }, { "epoch": 0.61, "learning_rate": 6.884489950999249e-06, "loss": 0.9268, "step": 12063 }, { "epoch": 0.61, "learning_rate": 6.8829264705289326e-06, "loss": 0.8765, "step": 12064 }, { "epoch": 0.61, "learning_rate": 6.88136307445006e-06, "loss": 0.8715, "step": 12065 }, { "epoch": 0.61, "learning_rate": 6.87979976280497e-06, "loss": 0.9929, "step": 12066 }, { "epoch": 0.61, "learning_rate": 6.87823653563598e-06, "loss": 0.9038, "step": 12067 }, { "epoch": 0.61, "learning_rate": 6.876673392985414e-06, "loss": 0.9111, "step": 12068 }, { "epoch": 0.61, "learning_rate": 6.875110334895595e-06, "loss": 0.8898, "step": 12069 }, { "epoch": 0.61, "learning_rate": 6.873547361408839e-06, "loss": 0.8734, "step": 12070 }, { "epoch": 0.61, "learning_rate": 6.871984472567457e-06, "loss": 0.7752, "step": 12071 }, { "epoch": 0.61, "learning_rate": 6.870421668413773e-06, "loss": 0.8764, "step": 12072 }, { "epoch": 0.61, "learning_rate": 6.868858948990091e-06, "loss": 0.91, "step": 12073 }, { "epoch": 0.61, "learning_rate": 6.86729631433872e-06, "loss": 0.8539, "step": 12074 }, { "epoch": 0.61, "learning_rate": 6.8657337645019715e-06, "loss": 0.9013, "step": 12075 }, { "epoch": 0.61, "learning_rate": 6.864171299522146e-06, "loss": 0.894, "step": 12076 }, { "epoch": 0.61, "learning_rate": 6.862608919441542e-06, "loss": 0.8678, "step": 12077 }, { "epoch": 0.61, "learning_rate": 6.861046624302468e-06, "loss": 0.9541, "step": 12078 }, { "epoch": 0.61, "learning_rate": 6.8594844141472174e-06, "loss": 0.8768, "step": 12079 }, { "epoch": 0.61, "learning_rate": 6.8579222890180815e-06, "loss": 0.8537, "step": 12080 }, { "epoch": 0.61, "learning_rate": 6.8563602489573614e-06, "loss": 0.996, "step": 12081 }, { "epoch": 0.61, "learning_rate": 6.854798294007337e-06, "loss": 0.9062, "step": 12082 }, { "epoch": 0.61, "learning_rate": 6.853236424210308e-06, "loss": 0.8544, "step": 12083 }, { "epoch": 0.61, "learning_rate": 6.851674639608553e-06, "loss": 0.798, "step": 12084 }, { "epoch": 0.61, "learning_rate": 6.850112940244357e-06, "loss": 0.9405, "step": 12085 }, { "epoch": 0.61, "learning_rate": 6.848551326160003e-06, "loss": 0.7367, "step": 12086 }, { "epoch": 0.61, "learning_rate": 6.846989797397769e-06, "loss": 0.9435, "step": 12087 }, { "epoch": 0.61, "learning_rate": 6.845428353999928e-06, "loss": 0.8835, "step": 12088 }, { "epoch": 0.61, "learning_rate": 6.843866996008763e-06, "loss": 0.8853, "step": 12089 }, { "epoch": 0.61, "learning_rate": 6.84230572346654e-06, "loss": 0.8968, "step": 12090 }, { "epoch": 0.61, "learning_rate": 6.840744536415529e-06, "loss": 0.8848, "step": 12091 }, { "epoch": 0.61, "learning_rate": 6.839183434898e-06, "loss": 0.9713, "step": 12092 }, { "epoch": 0.61, "learning_rate": 6.837622418956216e-06, "loss": 0.9231, "step": 12093 }, { "epoch": 0.61, "learning_rate": 6.836061488632437e-06, "loss": 0.9035, "step": 12094 }, { "epoch": 0.61, "learning_rate": 6.834500643968931e-06, "loss": 0.9033, "step": 12095 }, { "epoch": 0.61, "learning_rate": 6.832939885007952e-06, "loss": 0.9178, "step": 12096 }, { "epoch": 0.61, "learning_rate": 6.831379211791755e-06, "loss": 0.9019, "step": 12097 }, { "epoch": 0.61, "learning_rate": 6.829818624362596e-06, "loss": 0.8743, "step": 12098 }, { "epoch": 0.61, "learning_rate": 6.828258122762725e-06, "loss": 0.8677, "step": 12099 }, { "epoch": 0.61, "learning_rate": 6.8266977070343875e-06, "loss": 0.8608, "step": 12100 }, { "epoch": 0.61, "learning_rate": 6.825137377219837e-06, "loss": 0.9152, "step": 12101 }, { "epoch": 0.61, "learning_rate": 6.823577133361314e-06, "loss": 0.8545, "step": 12102 }, { "epoch": 0.61, "learning_rate": 6.822016975501059e-06, "loss": 0.8959, "step": 12103 }, { "epoch": 0.61, "learning_rate": 6.820456903681315e-06, "loss": 0.8572, "step": 12104 }, { "epoch": 0.61, "learning_rate": 6.818896917944318e-06, "loss": 0.8759, "step": 12105 }, { "epoch": 0.62, "learning_rate": 6.817337018332297e-06, "loss": 0.8579, "step": 12106 }, { "epoch": 0.62, "learning_rate": 6.815777204887496e-06, "loss": 0.8077, "step": 12107 }, { "epoch": 0.62, "learning_rate": 6.814217477652139e-06, "loss": 0.7947, "step": 12108 }, { "epoch": 0.62, "learning_rate": 6.812657836668453e-06, "loss": 0.9525, "step": 12109 }, { "epoch": 0.62, "learning_rate": 6.811098281978665e-06, "loss": 0.9184, "step": 12110 }, { "epoch": 0.62, "learning_rate": 6.809538813624999e-06, "loss": 0.8767, "step": 12111 }, { "epoch": 0.62, "learning_rate": 6.807979431649672e-06, "loss": 0.9383, "step": 12112 }, { "epoch": 0.62, "learning_rate": 6.80642013609491e-06, "loss": 0.9005, "step": 12113 }, { "epoch": 0.62, "learning_rate": 6.804860927002926e-06, "loss": 0.9097, "step": 12114 }, { "epoch": 0.62, "learning_rate": 6.803301804415928e-06, "loss": 0.8737, "step": 12115 }, { "epoch": 0.62, "learning_rate": 6.801742768376136e-06, "loss": 0.815, "step": 12116 }, { "epoch": 0.62, "learning_rate": 6.800183818925756e-06, "loss": 0.8972, "step": 12117 }, { "epoch": 0.62, "learning_rate": 6.798624956106992e-06, "loss": 0.9362, "step": 12118 }, { "epoch": 0.62, "learning_rate": 6.797066179962054e-06, "loss": 0.9259, "step": 12119 }, { "epoch": 0.62, "learning_rate": 6.795507490533142e-06, "loss": 0.8077, "step": 12120 }, { "epoch": 0.62, "learning_rate": 6.793948887862453e-06, "loss": 1.0284, "step": 12121 }, { "epoch": 0.62, "learning_rate": 6.792390371992189e-06, "loss": 0.873, "step": 12122 }, { "epoch": 0.62, "learning_rate": 6.790831942964543e-06, "loss": 0.9267, "step": 12123 }, { "epoch": 0.62, "learning_rate": 6.7892736008217044e-06, "loss": 0.8903, "step": 12124 }, { "epoch": 0.62, "learning_rate": 6.7877153456058694e-06, "loss": 0.8883, "step": 12125 }, { "epoch": 0.62, "learning_rate": 6.786157177359226e-06, "loss": 0.9341, "step": 12126 }, { "epoch": 0.62, "learning_rate": 6.784599096123955e-06, "loss": 0.9439, "step": 12127 }, { "epoch": 0.62, "learning_rate": 6.7830411019422446e-06, "loss": 0.9405, "step": 12128 }, { "epoch": 0.62, "learning_rate": 6.781483194856273e-06, "loss": 0.8978, "step": 12129 }, { "epoch": 0.62, "learning_rate": 6.779925374908216e-06, "loss": 0.8704, "step": 12130 }, { "epoch": 0.62, "learning_rate": 6.778367642140258e-06, "loss": 0.8657, "step": 12131 }, { "epoch": 0.62, "learning_rate": 6.776809996594569e-06, "loss": 0.869, "step": 12132 }, { "epoch": 0.62, "learning_rate": 6.775252438313318e-06, "loss": 0.9524, "step": 12133 }, { "epoch": 0.62, "learning_rate": 6.773694967338678e-06, "loss": 0.9988, "step": 12134 }, { "epoch": 0.62, "learning_rate": 6.772137583712813e-06, "loss": 0.8915, "step": 12135 }, { "epoch": 0.62, "learning_rate": 6.770580287477886e-06, "loss": 0.887, "step": 12136 }, { "epoch": 0.62, "learning_rate": 6.769023078676067e-06, "loss": 0.8905, "step": 12137 }, { "epoch": 0.62, "learning_rate": 6.767465957349508e-06, "loss": 0.9515, "step": 12138 }, { "epoch": 0.62, "learning_rate": 6.76590892354037e-06, "loss": 0.9292, "step": 12139 }, { "epoch": 0.62, "learning_rate": 6.7643519772908065e-06, "loss": 0.9291, "step": 12140 }, { "epoch": 0.62, "learning_rate": 6.762795118642973e-06, "loss": 0.8438, "step": 12141 }, { "epoch": 0.62, "learning_rate": 6.761238347639012e-06, "loss": 0.8615, "step": 12142 }, { "epoch": 0.62, "learning_rate": 6.759681664321081e-06, "loss": 0.9604, "step": 12143 }, { "epoch": 0.62, "learning_rate": 6.758125068731322e-06, "loss": 0.8928, "step": 12144 }, { "epoch": 0.62, "learning_rate": 6.756568560911876e-06, "loss": 0.9475, "step": 12145 }, { "epoch": 0.62, "learning_rate": 6.755012140904886e-06, "loss": 0.8181, "step": 12146 }, { "epoch": 0.62, "learning_rate": 6.753455808752492e-06, "loss": 0.8772, "step": 12147 }, { "epoch": 0.62, "learning_rate": 6.751899564496822e-06, "loss": 0.9554, "step": 12148 }, { "epoch": 0.62, "learning_rate": 6.75034340818002e-06, "loss": 0.9365, "step": 12149 }, { "epoch": 0.62, "learning_rate": 6.748787339844214e-06, "loss": 0.9561, "step": 12150 }, { "epoch": 0.62, "learning_rate": 6.747231359531529e-06, "loss": 0.9229, "step": 12151 }, { "epoch": 0.62, "learning_rate": 6.745675467284096e-06, "loss": 0.8164, "step": 12152 }, { "epoch": 0.62, "learning_rate": 6.744119663144033e-06, "loss": 0.9136, "step": 12153 }, { "epoch": 0.62, "learning_rate": 6.742563947153471e-06, "loss": 0.928, "step": 12154 }, { "epoch": 0.62, "learning_rate": 6.741008319354525e-06, "loss": 0.8892, "step": 12155 }, { "epoch": 0.62, "learning_rate": 6.739452779789308e-06, "loss": 0.96, "step": 12156 }, { "epoch": 0.62, "learning_rate": 6.737897328499942e-06, "loss": 0.9583, "step": 12157 }, { "epoch": 0.62, "learning_rate": 6.736341965528535e-06, "loss": 0.8781, "step": 12158 }, { "epoch": 0.62, "learning_rate": 6.73478669091719e-06, "loss": 0.8251, "step": 12159 }, { "epoch": 0.62, "learning_rate": 6.733231504708029e-06, "loss": 0.8658, "step": 12160 }, { "epoch": 0.62, "learning_rate": 6.731676406943147e-06, "loss": 1.0092, "step": 12161 }, { "epoch": 0.62, "learning_rate": 6.730121397664649e-06, "loss": 0.9438, "step": 12162 }, { "epoch": 0.62, "learning_rate": 6.728566476914636e-06, "loss": 0.9082, "step": 12163 }, { "epoch": 0.62, "learning_rate": 6.7270116447352056e-06, "loss": 0.8032, "step": 12164 }, { "epoch": 0.62, "learning_rate": 6.725456901168448e-06, "loss": 0.8548, "step": 12165 }, { "epoch": 0.62, "learning_rate": 6.723902246256466e-06, "loss": 0.9404, "step": 12166 }, { "epoch": 0.62, "learning_rate": 6.722347680041344e-06, "loss": 0.9258, "step": 12167 }, { "epoch": 0.62, "learning_rate": 6.72079320256517e-06, "loss": 0.8629, "step": 12168 }, { "epoch": 0.62, "learning_rate": 6.719238813870033e-06, "loss": 0.945, "step": 12169 }, { "epoch": 0.62, "learning_rate": 6.717684513998015e-06, "loss": 0.953, "step": 12170 }, { "epoch": 0.62, "learning_rate": 6.716130302991193e-06, "loss": 0.8653, "step": 12171 }, { "epoch": 0.62, "learning_rate": 6.714576180891653e-06, "loss": 0.8746, "step": 12172 }, { "epoch": 0.62, "learning_rate": 6.713022147741468e-06, "loss": 0.7843, "step": 12173 }, { "epoch": 0.62, "learning_rate": 6.711468203582709e-06, "loss": 0.8666, "step": 12174 }, { "epoch": 0.62, "learning_rate": 6.709914348457451e-06, "loss": 0.9184, "step": 12175 }, { "epoch": 0.62, "learning_rate": 6.708360582407762e-06, "loss": 0.9316, "step": 12176 }, { "epoch": 0.62, "learning_rate": 6.7068069054757046e-06, "loss": 0.9518, "step": 12177 }, { "epoch": 0.62, "learning_rate": 6.70525331770335e-06, "loss": 0.9469, "step": 12178 }, { "epoch": 0.62, "learning_rate": 6.7036998191327565e-06, "loss": 0.8172, "step": 12179 }, { "epoch": 0.62, "learning_rate": 6.7021464098059805e-06, "loss": 0.9007, "step": 12180 }, { "epoch": 0.62, "learning_rate": 6.700593089765086e-06, "loss": 0.9701, "step": 12181 }, { "epoch": 0.62, "learning_rate": 6.699039859052122e-06, "loss": 0.968, "step": 12182 }, { "epoch": 0.62, "learning_rate": 6.697486717709138e-06, "loss": 0.8881, "step": 12183 }, { "epoch": 0.62, "learning_rate": 6.695933665778191e-06, "loss": 0.9272, "step": 12184 }, { "epoch": 0.62, "learning_rate": 6.694380703301327e-06, "loss": 0.8417, "step": 12185 }, { "epoch": 0.62, "learning_rate": 6.692827830320584e-06, "loss": 0.906, "step": 12186 }, { "epoch": 0.62, "learning_rate": 6.691275046878012e-06, "loss": 0.8521, "step": 12187 }, { "epoch": 0.62, "learning_rate": 6.689722353015648e-06, "loss": 0.8897, "step": 12188 }, { "epoch": 0.62, "learning_rate": 6.6881697487755244e-06, "loss": 0.9311, "step": 12189 }, { "epoch": 0.62, "learning_rate": 6.686617234199686e-06, "loss": 0.8892, "step": 12190 }, { "epoch": 0.62, "learning_rate": 6.685064809330162e-06, "loss": 0.9058, "step": 12191 }, { "epoch": 0.62, "learning_rate": 6.683512474208977e-06, "loss": 0.8636, "step": 12192 }, { "epoch": 0.62, "learning_rate": 6.681960228878168e-06, "loss": 0.871, "step": 12193 }, { "epoch": 0.62, "learning_rate": 6.680408073379754e-06, "loss": 0.9313, "step": 12194 }, { "epoch": 0.62, "learning_rate": 6.6788560077557566e-06, "loss": 0.9007, "step": 12195 }, { "epoch": 0.62, "learning_rate": 6.677304032048202e-06, "loss": 0.8835, "step": 12196 }, { "epoch": 0.62, "learning_rate": 6.675752146299106e-06, "loss": 0.8473, "step": 12197 }, { "epoch": 0.62, "learning_rate": 6.674200350550483e-06, "loss": 0.9261, "step": 12198 }, { "epoch": 0.62, "learning_rate": 6.672648644844348e-06, "loss": 0.8474, "step": 12199 }, { "epoch": 0.62, "learning_rate": 6.671097029222712e-06, "loss": 0.8715, "step": 12200 }, { "epoch": 0.62, "learning_rate": 6.6695455037275766e-06, "loss": 0.8233, "step": 12201 }, { "epoch": 0.62, "learning_rate": 6.66799406840096e-06, "loss": 0.9659, "step": 12202 }, { "epoch": 0.62, "learning_rate": 6.6664427232848575e-06, "loss": 0.9076, "step": 12203 }, { "epoch": 0.62, "learning_rate": 6.664891468421269e-06, "loss": 0.9496, "step": 12204 }, { "epoch": 0.62, "learning_rate": 6.6633403038521986e-06, "loss": 0.9179, "step": 12205 }, { "epoch": 0.62, "learning_rate": 6.6617892296196394e-06, "loss": 0.9468, "step": 12206 }, { "epoch": 0.62, "learning_rate": 6.6602382457655805e-06, "loss": 0.846, "step": 12207 }, { "epoch": 0.62, "learning_rate": 6.658687352332023e-06, "loss": 0.9144, "step": 12208 }, { "epoch": 0.62, "learning_rate": 6.65713654936095e-06, "loss": 0.8196, "step": 12209 }, { "epoch": 0.62, "learning_rate": 6.6555858368943475e-06, "loss": 0.8968, "step": 12210 }, { "epoch": 0.62, "learning_rate": 6.654035214974199e-06, "loss": 0.8295, "step": 12211 }, { "epoch": 0.62, "learning_rate": 6.6524846836424904e-06, "loss": 0.7776, "step": 12212 }, { "epoch": 0.62, "learning_rate": 6.650934242941193e-06, "loss": 0.8933, "step": 12213 }, { "epoch": 0.62, "learning_rate": 6.649383892912291e-06, "loss": 1.0466, "step": 12214 }, { "epoch": 0.62, "learning_rate": 6.647833633597756e-06, "loss": 0.8467, "step": 12215 }, { "epoch": 0.62, "learning_rate": 6.6462834650395554e-06, "loss": 0.858, "step": 12216 }, { "epoch": 0.62, "learning_rate": 6.644733387279667e-06, "loss": 0.9057, "step": 12217 }, { "epoch": 0.62, "learning_rate": 6.64318340036005e-06, "loss": 0.9971, "step": 12218 }, { "epoch": 0.62, "learning_rate": 6.6416335043226675e-06, "loss": 0.9191, "step": 12219 }, { "epoch": 0.62, "learning_rate": 6.6400836992094895e-06, "loss": 0.8817, "step": 12220 }, { "epoch": 0.62, "learning_rate": 6.638533985062468e-06, "loss": 0.8121, "step": 12221 }, { "epoch": 0.62, "learning_rate": 6.636984361923564e-06, "loss": 0.9197, "step": 12222 }, { "epoch": 0.62, "learning_rate": 6.635434829834732e-06, "loss": 0.8149, "step": 12223 }, { "epoch": 0.62, "learning_rate": 6.633885388837917e-06, "loss": 1.0049, "step": 12224 }, { "epoch": 0.62, "learning_rate": 6.6323360389750794e-06, "loss": 0.8867, "step": 12225 }, { "epoch": 0.62, "learning_rate": 6.630786780288159e-06, "loss": 0.9425, "step": 12226 }, { "epoch": 0.62, "learning_rate": 6.629237612819101e-06, "loss": 0.8412, "step": 12227 }, { "epoch": 0.62, "learning_rate": 6.62768853660985e-06, "loss": 0.9125, "step": 12228 }, { "epoch": 0.62, "learning_rate": 6.6261395517023444e-06, "loss": 0.9108, "step": 12229 }, { "epoch": 0.62, "learning_rate": 6.624590658138515e-06, "loss": 0.9075, "step": 12230 }, { "epoch": 0.62, "learning_rate": 6.623041855960308e-06, "loss": 0.9399, "step": 12231 }, { "epoch": 0.62, "learning_rate": 6.62149314520965e-06, "loss": 0.8918, "step": 12232 }, { "epoch": 0.62, "learning_rate": 6.619944525928469e-06, "loss": 0.917, "step": 12233 }, { "epoch": 0.62, "learning_rate": 6.618395998158695e-06, "loss": 0.9472, "step": 12234 }, { "epoch": 0.62, "learning_rate": 6.616847561942252e-06, "loss": 0.8725, "step": 12235 }, { "epoch": 0.62, "learning_rate": 6.6152992173210585e-06, "loss": 0.9373, "step": 12236 }, { "epoch": 0.62, "learning_rate": 6.61375096433704e-06, "loss": 0.9819, "step": 12237 }, { "epoch": 0.62, "learning_rate": 6.612202803032113e-06, "loss": 0.9352, "step": 12238 }, { "epoch": 0.62, "learning_rate": 6.610654733448189e-06, "loss": 0.957, "step": 12239 }, { "epoch": 0.62, "learning_rate": 6.609106755627184e-06, "loss": 0.8739, "step": 12240 }, { "epoch": 0.62, "learning_rate": 6.607558869611004e-06, "loss": 0.7882, "step": 12241 }, { "epoch": 0.62, "learning_rate": 6.606011075441556e-06, "loss": 0.9024, "step": 12242 }, { "epoch": 0.62, "learning_rate": 6.60446337316075e-06, "loss": 0.8528, "step": 12243 }, { "epoch": 0.62, "learning_rate": 6.602915762810487e-06, "loss": 0.8452, "step": 12244 }, { "epoch": 0.62, "learning_rate": 6.601368244432663e-06, "loss": 0.8716, "step": 12245 }, { "epoch": 0.62, "learning_rate": 6.59982081806918e-06, "loss": 0.9086, "step": 12246 }, { "epoch": 0.62, "learning_rate": 6.598273483761931e-06, "loss": 0.9671, "step": 12247 }, { "epoch": 0.62, "learning_rate": 6.596726241552805e-06, "loss": 0.9194, "step": 12248 }, { "epoch": 0.62, "learning_rate": 6.595179091483697e-06, "loss": 0.9582, "step": 12249 }, { "epoch": 0.62, "learning_rate": 6.5936320335964945e-06, "loss": 0.8389, "step": 12250 }, { "epoch": 0.62, "learning_rate": 6.592085067933077e-06, "loss": 0.9449, "step": 12251 }, { "epoch": 0.62, "learning_rate": 6.590538194535334e-06, "loss": 0.9073, "step": 12252 }, { "epoch": 0.62, "learning_rate": 6.588991413445141e-06, "loss": 0.9208, "step": 12253 }, { "epoch": 0.62, "learning_rate": 6.587444724704374e-06, "loss": 0.8253, "step": 12254 }, { "epoch": 0.62, "learning_rate": 6.585898128354913e-06, "loss": 0.905, "step": 12255 }, { "epoch": 0.62, "learning_rate": 6.584351624438628e-06, "loss": 0.9084, "step": 12256 }, { "epoch": 0.62, "learning_rate": 6.582805212997388e-06, "loss": 0.8715, "step": 12257 }, { "epoch": 0.62, "learning_rate": 6.581258894073062e-06, "loss": 0.9821, "step": 12258 }, { "epoch": 0.62, "learning_rate": 6.579712667707514e-06, "loss": 1.0029, "step": 12259 }, { "epoch": 0.62, "learning_rate": 6.578166533942602e-06, "loss": 0.8835, "step": 12260 }, { "epoch": 0.62, "learning_rate": 6.576620492820197e-06, "loss": 0.9772, "step": 12261 }, { "epoch": 0.62, "learning_rate": 6.575074544382148e-06, "loss": 0.9013, "step": 12262 }, { "epoch": 0.62, "learning_rate": 6.573528688670311e-06, "loss": 0.9368, "step": 12263 }, { "epoch": 0.62, "learning_rate": 6.571982925726541e-06, "loss": 0.8005, "step": 12264 }, { "epoch": 0.62, "learning_rate": 6.570437255592685e-06, "loss": 0.9339, "step": 12265 }, { "epoch": 0.62, "learning_rate": 6.56889167831059e-06, "loss": 1.0369, "step": 12266 }, { "epoch": 0.62, "learning_rate": 6.567346193922104e-06, "loss": 0.8566, "step": 12267 }, { "epoch": 0.62, "learning_rate": 6.565800802469069e-06, "loss": 0.9053, "step": 12268 }, { "epoch": 0.62, "learning_rate": 6.564255503993321e-06, "loss": 0.8621, "step": 12269 }, { "epoch": 0.62, "learning_rate": 6.562710298536702e-06, "loss": 0.9112, "step": 12270 }, { "epoch": 0.62, "learning_rate": 6.561165186141045e-06, "loss": 0.892, "step": 12271 }, { "epoch": 0.62, "learning_rate": 6.559620166848178e-06, "loss": 0.8616, "step": 12272 }, { "epoch": 0.62, "learning_rate": 6.558075240699938e-06, "loss": 0.9417, "step": 12273 }, { "epoch": 0.62, "learning_rate": 6.556530407738148e-06, "loss": 0.877, "step": 12274 }, { "epoch": 0.62, "learning_rate": 6.554985668004634e-06, "loss": 0.9462, "step": 12275 }, { "epoch": 0.62, "learning_rate": 6.55344102154122e-06, "loss": 1.006, "step": 12276 }, { "epoch": 0.62, "learning_rate": 6.551896468389723e-06, "loss": 1.0095, "step": 12277 }, { "epoch": 0.62, "learning_rate": 6.550352008591954e-06, "loss": 0.9055, "step": 12278 }, { "epoch": 0.62, "learning_rate": 6.548807642189741e-06, "loss": 0.9767, "step": 12279 }, { "epoch": 0.62, "learning_rate": 6.547263369224888e-06, "loss": 0.9539, "step": 12280 }, { "epoch": 0.62, "learning_rate": 6.545719189739205e-06, "loss": 0.8838, "step": 12281 }, { "epoch": 0.62, "learning_rate": 6.5441751037745015e-06, "loss": 0.8546, "step": 12282 }, { "epoch": 0.62, "learning_rate": 6.542631111372579e-06, "loss": 0.847, "step": 12283 }, { "epoch": 0.62, "learning_rate": 6.541087212575239e-06, "loss": 0.901, "step": 12284 }, { "epoch": 0.62, "learning_rate": 6.539543407424285e-06, "loss": 0.8899, "step": 12285 }, { "epoch": 0.62, "learning_rate": 6.537999695961512e-06, "loss": 0.9306, "step": 12286 }, { "epoch": 0.62, "learning_rate": 6.536456078228713e-06, "loss": 0.8864, "step": 12287 }, { "epoch": 0.62, "learning_rate": 6.5349125542676805e-06, "loss": 0.8796, "step": 12288 }, { "epoch": 0.62, "learning_rate": 6.533369124120206e-06, "loss": 0.8089, "step": 12289 }, { "epoch": 0.62, "learning_rate": 6.53182578782807e-06, "loss": 0.8447, "step": 12290 }, { "epoch": 0.62, "learning_rate": 6.530282545433064e-06, "loss": 0.83, "step": 12291 }, { "epoch": 0.62, "learning_rate": 6.528739396976966e-06, "loss": 0.8082, "step": 12292 }, { "epoch": 0.62, "learning_rate": 6.527196342501556e-06, "loss": 0.8726, "step": 12293 }, { "epoch": 0.62, "learning_rate": 6.525653382048613e-06, "loss": 0.8438, "step": 12294 }, { "epoch": 0.62, "learning_rate": 6.524110515659901e-06, "loss": 0.9458, "step": 12295 }, { "epoch": 0.62, "learning_rate": 6.522567743377204e-06, "loss": 0.9016, "step": 12296 }, { "epoch": 0.62, "learning_rate": 6.521025065242287e-06, "loss": 0.8018, "step": 12297 }, { "epoch": 0.62, "learning_rate": 6.519482481296912e-06, "loss": 0.8982, "step": 12298 }, { "epoch": 0.62, "learning_rate": 6.5179399915828466e-06, "loss": 0.9863, "step": 12299 }, { "epoch": 0.62, "learning_rate": 6.516397596141853e-06, "loss": 0.8524, "step": 12300 }, { "epoch": 0.62, "learning_rate": 6.514855295015683e-06, "loss": 0.7698, "step": 12301 }, { "epoch": 0.62, "learning_rate": 6.513313088246103e-06, "loss": 0.8876, "step": 12302 }, { "epoch": 0.63, "learning_rate": 6.511770975874862e-06, "loss": 0.9117, "step": 12303 }, { "epoch": 0.63, "learning_rate": 6.510228957943707e-06, "loss": 0.9065, "step": 12304 }, { "epoch": 0.63, "learning_rate": 6.508687034494394e-06, "loss": 0.8897, "step": 12305 }, { "epoch": 0.63, "learning_rate": 6.5071452055686635e-06, "loss": 0.9465, "step": 12306 }, { "epoch": 0.63, "learning_rate": 6.505603471208257e-06, "loss": 0.9207, "step": 12307 }, { "epoch": 0.63, "learning_rate": 6.504061831454923e-06, "loss": 0.9084, "step": 12308 }, { "epoch": 0.63, "learning_rate": 6.502520286350396e-06, "loss": 0.8261, "step": 12309 }, { "epoch": 0.63, "learning_rate": 6.500978835936408e-06, "loss": 0.9528, "step": 12310 }, { "epoch": 0.63, "learning_rate": 6.499437480254698e-06, "loss": 0.8167, "step": 12311 }, { "epoch": 0.63, "learning_rate": 6.497896219346993e-06, "loss": 0.9164, "step": 12312 }, { "epoch": 0.63, "learning_rate": 6.496355053255019e-06, "loss": 0.9573, "step": 12313 }, { "epoch": 0.63, "learning_rate": 6.494813982020509e-06, "loss": 0.9029, "step": 12314 }, { "epoch": 0.63, "learning_rate": 6.49327300568518e-06, "loss": 0.9271, "step": 12315 }, { "epoch": 0.63, "learning_rate": 6.491732124290752e-06, "loss": 0.9566, "step": 12316 }, { "epoch": 0.63, "learning_rate": 6.490191337878946e-06, "loss": 0.8962, "step": 12317 }, { "epoch": 0.63, "learning_rate": 6.488650646491476e-06, "loss": 0.9259, "step": 12318 }, { "epoch": 0.63, "learning_rate": 6.487110050170049e-06, "loss": 0.9562, "step": 12319 }, { "epoch": 0.63, "learning_rate": 6.485569548956384e-06, "loss": 0.9351, "step": 12320 }, { "epoch": 0.63, "learning_rate": 6.484029142892185e-06, "loss": 0.8999, "step": 12321 }, { "epoch": 0.63, "learning_rate": 6.4824888320191534e-06, "loss": 0.8923, "step": 12322 }, { "epoch": 0.63, "learning_rate": 6.480948616378997e-06, "loss": 0.8935, "step": 12323 }, { "epoch": 0.63, "learning_rate": 6.479408496013413e-06, "loss": 1.0378, "step": 12324 }, { "epoch": 0.63, "learning_rate": 6.477868470964095e-06, "loss": 0.9366, "step": 12325 }, { "epoch": 0.63, "learning_rate": 6.476328541272743e-06, "loss": 0.9637, "step": 12326 }, { "epoch": 0.63, "learning_rate": 6.474788706981048e-06, "loss": 0.9738, "step": 12327 }, { "epoch": 0.63, "learning_rate": 6.473248968130696e-06, "loss": 0.905, "step": 12328 }, { "epoch": 0.63, "learning_rate": 6.471709324763379e-06, "loss": 0.9544, "step": 12329 }, { "epoch": 0.63, "learning_rate": 6.470169776920777e-06, "loss": 0.9259, "step": 12330 }, { "epoch": 0.63, "learning_rate": 6.468630324644569e-06, "loss": 0.8506, "step": 12331 }, { "epoch": 0.63, "learning_rate": 6.467090967976443e-06, "loss": 0.9451, "step": 12332 }, { "epoch": 0.63, "learning_rate": 6.465551706958068e-06, "loss": 0.9425, "step": 12333 }, { "epoch": 0.63, "learning_rate": 6.464012541631121e-06, "loss": 0.8585, "step": 12334 }, { "epoch": 0.63, "learning_rate": 6.462473472037273e-06, "loss": 0.9235, "step": 12335 }, { "epoch": 0.63, "learning_rate": 6.460934498218192e-06, "loss": 0.8809, "step": 12336 }, { "epoch": 0.63, "learning_rate": 6.45939562021554e-06, "loss": 0.9022, "step": 12337 }, { "epoch": 0.63, "learning_rate": 6.45785683807099e-06, "loss": 0.955, "step": 12338 }, { "epoch": 0.63, "learning_rate": 6.456318151826195e-06, "loss": 0.8809, "step": 12339 }, { "epoch": 0.63, "learning_rate": 6.454779561522817e-06, "loss": 0.8177, "step": 12340 }, { "epoch": 0.63, "learning_rate": 6.453241067202511e-06, "loss": 0.8483, "step": 12341 }, { "epoch": 0.63, "learning_rate": 6.4517026689069295e-06, "loss": 0.9046, "step": 12342 }, { "epoch": 0.63, "learning_rate": 6.450164366677718e-06, "loss": 0.9138, "step": 12343 }, { "epoch": 0.63, "learning_rate": 6.448626160556536e-06, "loss": 0.8922, "step": 12344 }, { "epoch": 0.63, "learning_rate": 6.447088050585021e-06, "loss": 0.9414, "step": 12345 }, { "epoch": 0.63, "learning_rate": 6.445550036804815e-06, "loss": 0.8845, "step": 12346 }, { "epoch": 0.63, "learning_rate": 6.4440121192575625e-06, "loss": 0.8472, "step": 12347 }, { "epoch": 0.63, "learning_rate": 6.4424742979848986e-06, "loss": 0.9296, "step": 12348 }, { "epoch": 0.63, "learning_rate": 6.4409365730284536e-06, "loss": 0.8826, "step": 12349 }, { "epoch": 0.63, "learning_rate": 6.4393989444298685e-06, "loss": 0.9177, "step": 12350 }, { "epoch": 0.63, "learning_rate": 6.43786141223077e-06, "loss": 0.9015, "step": 12351 }, { "epoch": 0.63, "learning_rate": 6.43632397647278e-06, "loss": 0.8202, "step": 12352 }, { "epoch": 0.63, "learning_rate": 6.43478663719753e-06, "loss": 0.9156, "step": 12353 }, { "epoch": 0.63, "learning_rate": 6.433249394446639e-06, "loss": 0.9819, "step": 12354 }, { "epoch": 0.63, "learning_rate": 6.4317122482617215e-06, "loss": 0.861, "step": 12355 }, { "epoch": 0.63, "learning_rate": 6.430175198684403e-06, "loss": 0.9229, "step": 12356 }, { "epoch": 0.63, "learning_rate": 6.428638245756292e-06, "loss": 0.9358, "step": 12357 }, { "epoch": 0.63, "learning_rate": 6.427101389519e-06, "loss": 0.8689, "step": 12358 }, { "epoch": 0.63, "learning_rate": 6.425564630014138e-06, "loss": 0.8807, "step": 12359 }, { "epoch": 0.63, "learning_rate": 6.424027967283306e-06, "loss": 0.8665, "step": 12360 }, { "epoch": 0.63, "learning_rate": 6.4224914013681175e-06, "loss": 0.8863, "step": 12361 }, { "epoch": 0.63, "learning_rate": 6.420954932310167e-06, "loss": 0.9576, "step": 12362 }, { "epoch": 0.63, "learning_rate": 6.4194185601510525e-06, "loss": 0.8914, "step": 12363 }, { "epoch": 0.63, "learning_rate": 6.417882284932373e-06, "loss": 0.922, "step": 12364 }, { "epoch": 0.63, "learning_rate": 6.41634610669572e-06, "loss": 0.9099, "step": 12365 }, { "epoch": 0.63, "learning_rate": 6.414810025482679e-06, "loss": 0.8818, "step": 12366 }, { "epoch": 0.63, "learning_rate": 6.413274041334847e-06, "loss": 0.8871, "step": 12367 }, { "epoch": 0.63, "learning_rate": 6.411738154293803e-06, "loss": 0.9138, "step": 12368 }, { "epoch": 0.63, "learning_rate": 6.41020236440113e-06, "loss": 1.0161, "step": 12369 }, { "epoch": 0.63, "learning_rate": 6.40866667169841e-06, "loss": 0.8279, "step": 12370 }, { "epoch": 0.63, "learning_rate": 6.40713107622722e-06, "loss": 0.8577, "step": 12371 }, { "epoch": 0.63, "learning_rate": 6.4055955780291294e-06, "loss": 0.911, "step": 12372 }, { "epoch": 0.63, "learning_rate": 6.404060177145717e-06, "loss": 0.9289, "step": 12373 }, { "epoch": 0.63, "learning_rate": 6.40252487361855e-06, "loss": 0.886, "step": 12374 }, { "epoch": 0.63, "learning_rate": 6.400989667489192e-06, "loss": 0.9075, "step": 12375 }, { "epoch": 0.63, "learning_rate": 6.399454558799214e-06, "loss": 0.847, "step": 12376 }, { "epoch": 0.63, "learning_rate": 6.397919547590172e-06, "loss": 0.863, "step": 12377 }, { "epoch": 0.63, "learning_rate": 6.396384633903621e-06, "loss": 0.868, "step": 12378 }, { "epoch": 0.63, "learning_rate": 6.394849817781128e-06, "loss": 0.9205, "step": 12379 }, { "epoch": 0.63, "learning_rate": 6.393315099264239e-06, "loss": 0.9297, "step": 12380 }, { "epoch": 0.63, "learning_rate": 6.391780478394505e-06, "loss": 0.8948, "step": 12381 }, { "epoch": 0.63, "learning_rate": 6.390245955213478e-06, "loss": 0.8432, "step": 12382 }, { "epoch": 0.63, "learning_rate": 6.388711529762701e-06, "loss": 0.896, "step": 12383 }, { "epoch": 0.63, "learning_rate": 6.387177202083712e-06, "loss": 0.8257, "step": 12384 }, { "epoch": 0.63, "learning_rate": 6.385642972218063e-06, "loss": 0.878, "step": 12385 }, { "epoch": 0.63, "learning_rate": 6.384108840207284e-06, "loss": 0.9282, "step": 12386 }, { "epoch": 0.63, "learning_rate": 6.3825748060929085e-06, "loss": 0.8916, "step": 12387 }, { "epoch": 0.63, "learning_rate": 6.381040869916474e-06, "loss": 0.8658, "step": 12388 }, { "epoch": 0.63, "learning_rate": 6.379507031719507e-06, "loss": 0.7805, "step": 12389 }, { "epoch": 0.63, "learning_rate": 6.377973291543532e-06, "loss": 0.9987, "step": 12390 }, { "epoch": 0.63, "learning_rate": 6.376439649430081e-06, "loss": 0.8763, "step": 12391 }, { "epoch": 0.63, "learning_rate": 6.374906105420671e-06, "loss": 0.8422, "step": 12392 }, { "epoch": 0.63, "learning_rate": 6.373372659556818e-06, "loss": 0.8593, "step": 12393 }, { "epoch": 0.63, "learning_rate": 6.371839311880046e-06, "loss": 0.8759, "step": 12394 }, { "epoch": 0.63, "learning_rate": 6.370306062431864e-06, "loss": 0.912, "step": 12395 }, { "epoch": 0.63, "learning_rate": 6.3687729112537774e-06, "loss": 0.8992, "step": 12396 }, { "epoch": 0.63, "learning_rate": 6.3672398583873085e-06, "loss": 0.9067, "step": 12397 }, { "epoch": 0.63, "learning_rate": 6.365706903873954e-06, "loss": 0.9287, "step": 12398 }, { "epoch": 0.63, "learning_rate": 6.364174047755216e-06, "loss": 1.0158, "step": 12399 }, { "epoch": 0.63, "learning_rate": 6.3626412900725995e-06, "loss": 0.8521, "step": 12400 }, { "epoch": 0.63, "learning_rate": 6.361108630867601e-06, "loss": 0.8611, "step": 12401 }, { "epoch": 0.63, "learning_rate": 6.35957607018171e-06, "loss": 0.8735, "step": 12402 }, { "epoch": 0.63, "learning_rate": 6.358043608056428e-06, "loss": 0.9208, "step": 12403 }, { "epoch": 0.63, "learning_rate": 6.35651124453324e-06, "loss": 0.8694, "step": 12404 }, { "epoch": 0.63, "learning_rate": 6.354978979653632e-06, "loss": 0.9246, "step": 12405 }, { "epoch": 0.63, "learning_rate": 6.353446813459092e-06, "loss": 0.7727, "step": 12406 }, { "epoch": 0.63, "learning_rate": 6.351914745991101e-06, "loss": 0.8961, "step": 12407 }, { "epoch": 0.63, "learning_rate": 6.350382777291133e-06, "loss": 0.821, "step": 12408 }, { "epoch": 0.63, "learning_rate": 6.348850907400671e-06, "loss": 0.9296, "step": 12409 }, { "epoch": 0.63, "learning_rate": 6.347319136361188e-06, "loss": 0.8671, "step": 12410 }, { "epoch": 0.63, "learning_rate": 6.345787464214149e-06, "loss": 0.9362, "step": 12411 }, { "epoch": 0.63, "learning_rate": 6.34425589100103e-06, "loss": 0.9937, "step": 12412 }, { "epoch": 0.63, "learning_rate": 6.342724416763293e-06, "loss": 0.9386, "step": 12413 }, { "epoch": 0.63, "learning_rate": 6.3411930415423976e-06, "loss": 0.9072, "step": 12414 }, { "epoch": 0.63, "learning_rate": 6.339661765379812e-06, "loss": 0.9173, "step": 12415 }, { "epoch": 0.63, "learning_rate": 6.338130588316989e-06, "loss": 0.8494, "step": 12416 }, { "epoch": 0.63, "learning_rate": 6.336599510395384e-06, "loss": 0.9131, "step": 12417 }, { "epoch": 0.63, "learning_rate": 6.33506853165645e-06, "loss": 0.897, "step": 12418 }, { "epoch": 0.63, "learning_rate": 6.333537652141637e-06, "loss": 0.8596, "step": 12419 }, { "epoch": 0.63, "learning_rate": 6.332006871892388e-06, "loss": 0.8654, "step": 12420 }, { "epoch": 0.63, "learning_rate": 6.330476190950153e-06, "loss": 0.9186, "step": 12421 }, { "epoch": 0.63, "learning_rate": 6.328945609356372e-06, "loss": 0.8713, "step": 12422 }, { "epoch": 0.63, "learning_rate": 6.3274151271524804e-06, "loss": 0.9048, "step": 12423 }, { "epoch": 0.63, "learning_rate": 6.32588474437992e-06, "loss": 0.8495, "step": 12424 }, { "epoch": 0.63, "learning_rate": 6.324354461080121e-06, "loss": 0.8418, "step": 12425 }, { "epoch": 0.63, "learning_rate": 6.3228242772945105e-06, "loss": 0.881, "step": 12426 }, { "epoch": 0.63, "learning_rate": 6.321294193064523e-06, "loss": 0.9028, "step": 12427 }, { "epoch": 0.63, "learning_rate": 6.3197642084315826e-06, "loss": 0.8355, "step": 12428 }, { "epoch": 0.63, "learning_rate": 6.318234323437109e-06, "loss": 0.8517, "step": 12429 }, { "epoch": 0.63, "learning_rate": 6.316704538122524e-06, "loss": 0.9116, "step": 12430 }, { "epoch": 0.63, "learning_rate": 6.315174852529243e-06, "loss": 0.899, "step": 12431 }, { "epoch": 0.63, "learning_rate": 6.313645266698687e-06, "loss": 0.9766, "step": 12432 }, { "epoch": 0.63, "learning_rate": 6.312115780672261e-06, "loss": 0.9054, "step": 12433 }, { "epoch": 0.63, "learning_rate": 6.310586394491376e-06, "loss": 0.8853, "step": 12434 }, { "epoch": 0.63, "learning_rate": 6.309057108197442e-06, "loss": 0.7766, "step": 12435 }, { "epoch": 0.63, "learning_rate": 6.3075279218318585e-06, "loss": 0.8575, "step": 12436 }, { "epoch": 0.63, "learning_rate": 6.305998835436025e-06, "loss": 0.959, "step": 12437 }, { "epoch": 0.63, "learning_rate": 6.304469849051347e-06, "loss": 0.8832, "step": 12438 }, { "epoch": 0.63, "learning_rate": 6.302940962719215e-06, "loss": 0.8497, "step": 12439 }, { "epoch": 0.63, "learning_rate": 6.301412176481022e-06, "loss": 0.8959, "step": 12440 }, { "epoch": 0.63, "learning_rate": 6.29988349037816e-06, "loss": 0.9491, "step": 12441 }, { "epoch": 0.63, "learning_rate": 6.298354904452018e-06, "loss": 0.8766, "step": 12442 }, { "epoch": 0.63, "learning_rate": 6.296826418743973e-06, "loss": 0.941, "step": 12443 }, { "epoch": 0.63, "learning_rate": 6.2952980332954185e-06, "loss": 0.8956, "step": 12444 }, { "epoch": 0.63, "learning_rate": 6.2937697481477265e-06, "loss": 0.9183, "step": 12445 }, { "epoch": 0.63, "learning_rate": 6.292241563342274e-06, "loss": 0.9048, "step": 12446 }, { "epoch": 0.63, "learning_rate": 6.290713478920439e-06, "loss": 0.9023, "step": 12447 }, { "epoch": 0.63, "learning_rate": 6.2891854949235894e-06, "loss": 0.87, "step": 12448 }, { "epoch": 0.63, "learning_rate": 6.287657611393092e-06, "loss": 0.9041, "step": 12449 }, { "epoch": 0.63, "learning_rate": 6.286129828370318e-06, "loss": 0.9331, "step": 12450 }, { "epoch": 0.63, "learning_rate": 6.284602145896628e-06, "loss": 0.9777, "step": 12451 }, { "epoch": 0.63, "learning_rate": 6.283074564013379e-06, "loss": 0.8111, "step": 12452 }, { "epoch": 0.63, "learning_rate": 6.281547082761934e-06, "loss": 0.9548, "step": 12453 }, { "epoch": 0.63, "learning_rate": 6.280019702183647e-06, "loss": 0.8924, "step": 12454 }, { "epoch": 0.63, "learning_rate": 6.278492422319864e-06, "loss": 0.9839, "step": 12455 }, { "epoch": 0.63, "learning_rate": 6.276965243211943e-06, "loss": 0.9007, "step": 12456 }, { "epoch": 0.63, "learning_rate": 6.275438164901226e-06, "loss": 0.9312, "step": 12457 }, { "epoch": 0.63, "learning_rate": 6.273911187429058e-06, "loss": 0.8881, "step": 12458 }, { "epoch": 0.63, "learning_rate": 6.2723843108367825e-06, "loss": 0.8963, "step": 12459 }, { "epoch": 0.63, "learning_rate": 6.270857535165735e-06, "loss": 0.8937, "step": 12460 }, { "epoch": 0.63, "learning_rate": 6.269330860457248e-06, "loss": 0.848, "step": 12461 }, { "epoch": 0.63, "learning_rate": 6.267804286752663e-06, "loss": 0.8813, "step": 12462 }, { "epoch": 0.63, "learning_rate": 6.2662778140933064e-06, "loss": 0.8136, "step": 12463 }, { "epoch": 0.63, "learning_rate": 6.264751442520504e-06, "loss": 0.9659, "step": 12464 }, { "epoch": 0.63, "learning_rate": 6.263225172075583e-06, "loss": 0.9551, "step": 12465 }, { "epoch": 0.63, "learning_rate": 6.261699002799866e-06, "loss": 0.9284, "step": 12466 }, { "epoch": 0.63, "learning_rate": 6.260172934734667e-06, "loss": 0.886, "step": 12467 }, { "epoch": 0.63, "learning_rate": 6.258646967921311e-06, "loss": 0.8901, "step": 12468 }, { "epoch": 0.63, "learning_rate": 6.257121102401107e-06, "loss": 0.9308, "step": 12469 }, { "epoch": 0.63, "learning_rate": 6.255595338215366e-06, "loss": 0.9227, "step": 12470 }, { "epoch": 0.63, "learning_rate": 6.2540696754053995e-06, "loss": 0.9268, "step": 12471 }, { "epoch": 0.63, "learning_rate": 6.252544114012511e-06, "loss": 0.8833, "step": 12472 }, { "epoch": 0.63, "learning_rate": 6.251018654078001e-06, "loss": 0.9714, "step": 12473 }, { "epoch": 0.63, "learning_rate": 6.249493295643174e-06, "loss": 0.8931, "step": 12474 }, { "epoch": 0.63, "learning_rate": 6.247968038749326e-06, "loss": 0.8412, "step": 12475 }, { "epoch": 0.63, "learning_rate": 6.246442883437751e-06, "loss": 0.8695, "step": 12476 }, { "epoch": 0.63, "learning_rate": 6.244917829749742e-06, "loss": 0.8418, "step": 12477 }, { "epoch": 0.63, "learning_rate": 6.243392877726589e-06, "loss": 0.9244, "step": 12478 }, { "epoch": 0.63, "learning_rate": 6.241868027409572e-06, "loss": 0.9893, "step": 12479 }, { "epoch": 0.63, "learning_rate": 6.240343278839985e-06, "loss": 0.8985, "step": 12480 }, { "epoch": 0.63, "learning_rate": 6.238818632059103e-06, "loss": 0.8653, "step": 12481 }, { "epoch": 0.63, "learning_rate": 6.237294087108203e-06, "loss": 0.9662, "step": 12482 }, { "epoch": 0.63, "learning_rate": 6.2357696440285646e-06, "loss": 0.7871, "step": 12483 }, { "epoch": 0.63, "learning_rate": 6.234245302861457e-06, "loss": 1.0104, "step": 12484 }, { "epoch": 0.63, "learning_rate": 6.232721063648148e-06, "loss": 1.0209, "step": 12485 }, { "epoch": 0.63, "learning_rate": 6.231196926429913e-06, "loss": 0.9307, "step": 12486 }, { "epoch": 0.63, "learning_rate": 6.229672891248011e-06, "loss": 0.9343, "step": 12487 }, { "epoch": 0.63, "learning_rate": 6.228148958143703e-06, "loss": 0.8698, "step": 12488 }, { "epoch": 0.63, "learning_rate": 6.22662512715825e-06, "loss": 0.8973, "step": 12489 }, { "epoch": 0.63, "learning_rate": 6.225101398332907e-06, "loss": 0.9334, "step": 12490 }, { "epoch": 0.63, "learning_rate": 6.223577771708924e-06, "loss": 0.8876, "step": 12491 }, { "epoch": 0.63, "learning_rate": 6.222054247327559e-06, "loss": 0.9012, "step": 12492 }, { "epoch": 0.63, "learning_rate": 6.220530825230055e-06, "loss": 0.979, "step": 12493 }, { "epoch": 0.63, "learning_rate": 6.219007505457657e-06, "loss": 0.9093, "step": 12494 }, { "epoch": 0.63, "learning_rate": 6.21748428805161e-06, "loss": 0.8396, "step": 12495 }, { "epoch": 0.63, "learning_rate": 6.215961173053151e-06, "loss": 1.0001, "step": 12496 }, { "epoch": 0.63, "learning_rate": 6.214438160503514e-06, "loss": 0.8761, "step": 12497 }, { "epoch": 0.63, "learning_rate": 6.2129152504439404e-06, "loss": 0.8401, "step": 12498 }, { "epoch": 0.63, "learning_rate": 6.211392442915654e-06, "loss": 0.9361, "step": 12499 }, { "epoch": 0.64, "learning_rate": 6.20986973795989e-06, "loss": 0.884, "step": 12500 }, { "epoch": 0.64, "learning_rate": 6.208347135617869e-06, "loss": 0.9124, "step": 12501 }, { "epoch": 0.64, "learning_rate": 6.206824635930812e-06, "loss": 0.9249, "step": 12502 }, { "epoch": 0.64, "learning_rate": 6.205302238939948e-06, "loss": 0.8718, "step": 12503 }, { "epoch": 0.64, "learning_rate": 6.203779944686486e-06, "loss": 0.9009, "step": 12504 }, { "epoch": 0.64, "learning_rate": 6.2022577532116425e-06, "loss": 0.928, "step": 12505 }, { "epoch": 0.64, "learning_rate": 6.2007356645566315e-06, "loss": 0.9393, "step": 12506 }, { "epoch": 0.64, "learning_rate": 6.199213678762661e-06, "loss": 0.9474, "step": 12507 }, { "epoch": 0.64, "learning_rate": 6.19769179587093e-06, "loss": 0.9455, "step": 12508 }, { "epoch": 0.64, "learning_rate": 6.1961700159226555e-06, "loss": 0.8137, "step": 12509 }, { "epoch": 0.64, "learning_rate": 6.194648338959028e-06, "loss": 0.9164, "step": 12510 }, { "epoch": 0.64, "learning_rate": 6.193126765021249e-06, "loss": 0.8067, "step": 12511 }, { "epoch": 0.64, "learning_rate": 6.191605294150514e-06, "loss": 0.8855, "step": 12512 }, { "epoch": 0.64, "learning_rate": 6.190083926388012e-06, "loss": 0.9096, "step": 12513 }, { "epoch": 0.64, "learning_rate": 6.1885626617749304e-06, "loss": 0.8461, "step": 12514 }, { "epoch": 0.64, "learning_rate": 6.187041500352465e-06, "loss": 0.9041, "step": 12515 }, { "epoch": 0.64, "learning_rate": 6.185520442161794e-06, "loss": 0.8795, "step": 12516 }, { "epoch": 0.64, "learning_rate": 6.183999487244097e-06, "loss": 0.898, "step": 12517 }, { "epoch": 0.64, "learning_rate": 6.182478635640555e-06, "loss": 0.8991, "step": 12518 }, { "epoch": 0.64, "learning_rate": 6.180957887392344e-06, "loss": 0.9056, "step": 12519 }, { "epoch": 0.64, "learning_rate": 6.179437242540629e-06, "loss": 0.9295, "step": 12520 }, { "epoch": 0.64, "learning_rate": 6.177916701126592e-06, "loss": 0.7988, "step": 12521 }, { "epoch": 0.64, "learning_rate": 6.176396263191393e-06, "loss": 0.942, "step": 12522 }, { "epoch": 0.64, "learning_rate": 6.174875928776194e-06, "loss": 0.843, "step": 12523 }, { "epoch": 0.64, "learning_rate": 6.173355697922163e-06, "loss": 0.7793, "step": 12524 }, { "epoch": 0.64, "learning_rate": 6.1718355706704545e-06, "loss": 0.9195, "step": 12525 }, { "epoch": 0.64, "learning_rate": 6.170315547062222e-06, "loss": 0.9133, "step": 12526 }, { "epoch": 0.64, "learning_rate": 6.168795627138626e-06, "loss": 0.9044, "step": 12527 }, { "epoch": 0.64, "learning_rate": 6.167275810940812e-06, "loss": 0.8466, "step": 12528 }, { "epoch": 0.64, "learning_rate": 6.165756098509924e-06, "loss": 0.921, "step": 12529 }, { "epoch": 0.64, "learning_rate": 6.1642364898871144e-06, "loss": 0.8025, "step": 12530 }, { "epoch": 0.64, "learning_rate": 6.16271698511352e-06, "loss": 0.8919, "step": 12531 }, { "epoch": 0.64, "learning_rate": 6.161197584230276e-06, "loss": 0.8662, "step": 12532 }, { "epoch": 0.64, "learning_rate": 6.1596782872785275e-06, "loss": 0.9259, "step": 12533 }, { "epoch": 0.64, "learning_rate": 6.158159094299405e-06, "loss": 0.8696, "step": 12534 }, { "epoch": 0.64, "learning_rate": 6.156640005334034e-06, "loss": 0.8971, "step": 12535 }, { "epoch": 0.64, "learning_rate": 6.155121020423549e-06, "loss": 0.9062, "step": 12536 }, { "epoch": 0.64, "learning_rate": 6.153602139609071e-06, "loss": 0.9301, "step": 12537 }, { "epoch": 0.64, "learning_rate": 6.152083362931719e-06, "loss": 0.8975, "step": 12538 }, { "epoch": 0.64, "learning_rate": 6.15056469043262e-06, "loss": 0.8575, "step": 12539 }, { "epoch": 0.64, "learning_rate": 6.149046122152886e-06, "loss": 0.9255, "step": 12540 }, { "epoch": 0.64, "learning_rate": 6.147527658133631e-06, "loss": 0.8287, "step": 12541 }, { "epoch": 0.64, "learning_rate": 6.146009298415967e-06, "loss": 0.8802, "step": 12542 }, { "epoch": 0.64, "learning_rate": 6.144491043041e-06, "loss": 0.9365, "step": 12543 }, { "epoch": 0.64, "learning_rate": 6.1429728920498324e-06, "loss": 0.9176, "step": 12544 }, { "epoch": 0.64, "learning_rate": 6.1414548454835745e-06, "loss": 0.9901, "step": 12545 }, { "epoch": 0.64, "learning_rate": 6.139936903383322e-06, "loss": 0.871, "step": 12546 }, { "epoch": 0.64, "learning_rate": 6.138419065790169e-06, "loss": 0.9909, "step": 12547 }, { "epoch": 0.64, "learning_rate": 6.1369013327452135e-06, "loss": 0.8404, "step": 12548 }, { "epoch": 0.64, "learning_rate": 6.135383704289544e-06, "loss": 0.9009, "step": 12549 }, { "epoch": 0.64, "learning_rate": 6.133866180464244e-06, "loss": 0.8656, "step": 12550 }, { "epoch": 0.64, "learning_rate": 6.1323487613104095e-06, "loss": 0.8477, "step": 12551 }, { "epoch": 0.64, "learning_rate": 6.1308314468691165e-06, "loss": 0.8889, "step": 12552 }, { "epoch": 0.64, "learning_rate": 6.129314237181443e-06, "loss": 0.8731, "step": 12553 }, { "epoch": 0.64, "learning_rate": 6.1277971322884725e-06, "loss": 0.8451, "step": 12554 }, { "epoch": 0.64, "learning_rate": 6.126280132231273e-06, "loss": 0.84, "step": 12555 }, { "epoch": 0.64, "learning_rate": 6.124763237050915e-06, "loss": 0.8698, "step": 12556 }, { "epoch": 0.64, "learning_rate": 6.1232464467884735e-06, "loss": 0.9056, "step": 12557 }, { "epoch": 0.64, "learning_rate": 6.121729761485009e-06, "loss": 0.9636, "step": 12558 }, { "epoch": 0.64, "learning_rate": 6.120213181181582e-06, "loss": 0.9395, "step": 12559 }, { "epoch": 0.64, "learning_rate": 6.118696705919259e-06, "loss": 0.9179, "step": 12560 }, { "epoch": 0.64, "learning_rate": 6.117180335739094e-06, "loss": 0.9434, "step": 12561 }, { "epoch": 0.64, "learning_rate": 6.115664070682134e-06, "loss": 0.8187, "step": 12562 }, { "epoch": 0.64, "learning_rate": 6.114147910789443e-06, "loss": 0.9291, "step": 12563 }, { "epoch": 0.64, "learning_rate": 6.1126318561020625e-06, "loss": 0.9231, "step": 12564 }, { "epoch": 0.64, "learning_rate": 6.111115906661036e-06, "loss": 0.9583, "step": 12565 }, { "epoch": 0.64, "learning_rate": 6.10960006250741e-06, "loss": 0.8274, "step": 12566 }, { "epoch": 0.64, "learning_rate": 6.108084323682227e-06, "loss": 0.901, "step": 12567 }, { "epoch": 0.64, "learning_rate": 6.106568690226513e-06, "loss": 0.9101, "step": 12568 }, { "epoch": 0.64, "learning_rate": 6.105053162181314e-06, "loss": 0.8744, "step": 12569 }, { "epoch": 0.64, "learning_rate": 6.103537739587655e-06, "loss": 0.8921, "step": 12570 }, { "epoch": 0.64, "learning_rate": 6.1020224224865675e-06, "loss": 0.8864, "step": 12571 }, { "epoch": 0.64, "learning_rate": 6.1005072109190765e-06, "loss": 0.8397, "step": 12572 }, { "epoch": 0.64, "learning_rate": 6.098992104926201e-06, "loss": 0.9233, "step": 12573 }, { "epoch": 0.64, "learning_rate": 6.097477104548966e-06, "loss": 0.9215, "step": 12574 }, { "epoch": 0.64, "learning_rate": 6.0959622098283866e-06, "loss": 0.9571, "step": 12575 }, { "epoch": 0.64, "learning_rate": 6.094447420805476e-06, "loss": 0.8239, "step": 12576 }, { "epoch": 0.64, "learning_rate": 6.0929327375212465e-06, "loss": 0.9098, "step": 12577 }, { "epoch": 0.64, "learning_rate": 6.091418160016709e-06, "loss": 0.9485, "step": 12578 }, { "epoch": 0.64, "learning_rate": 6.089903688332861e-06, "loss": 0.8633, "step": 12579 }, { "epoch": 0.64, "learning_rate": 6.088389322510715e-06, "loss": 0.9746, "step": 12580 }, { "epoch": 0.64, "learning_rate": 6.086875062591265e-06, "loss": 0.8456, "step": 12581 }, { "epoch": 0.64, "learning_rate": 6.085360908615509e-06, "loss": 0.9951, "step": 12582 }, { "epoch": 0.64, "learning_rate": 6.083846860624443e-06, "loss": 0.8575, "step": 12583 }, { "epoch": 0.64, "learning_rate": 6.082332918659057e-06, "loss": 0.858, "step": 12584 }, { "epoch": 0.64, "learning_rate": 6.080819082760334e-06, "loss": 0.8676, "step": 12585 }, { "epoch": 0.64, "learning_rate": 6.07930535296927e-06, "loss": 0.7989, "step": 12586 }, { "epoch": 0.64, "learning_rate": 6.0777917293268425e-06, "loss": 0.8818, "step": 12587 }, { "epoch": 0.64, "learning_rate": 6.076278211874028e-06, "loss": 0.9528, "step": 12588 }, { "epoch": 0.64, "learning_rate": 6.074764800651809e-06, "loss": 0.9434, "step": 12589 }, { "epoch": 0.64, "learning_rate": 6.073251495701158e-06, "loss": 0.9306, "step": 12590 }, { "epoch": 0.64, "learning_rate": 6.071738297063039e-06, "loss": 0.8661, "step": 12591 }, { "epoch": 0.64, "learning_rate": 6.07022520477843e-06, "loss": 0.8781, "step": 12592 }, { "epoch": 0.64, "learning_rate": 6.068712218888295e-06, "loss": 0.9125, "step": 12593 }, { "epoch": 0.64, "learning_rate": 6.067199339433591e-06, "loss": 0.8536, "step": 12594 }, { "epoch": 0.64, "learning_rate": 6.065686566455282e-06, "loss": 0.9863, "step": 12595 }, { "epoch": 0.64, "learning_rate": 6.064173899994323e-06, "loss": 0.9383, "step": 12596 }, { "epoch": 0.64, "learning_rate": 6.062661340091667e-06, "loss": 0.9641, "step": 12597 }, { "epoch": 0.64, "learning_rate": 6.061148886788268e-06, "loss": 0.9731, "step": 12598 }, { "epoch": 0.64, "learning_rate": 6.059636540125072e-06, "loss": 0.8552, "step": 12599 }, { "epoch": 0.64, "learning_rate": 6.058124300143023e-06, "loss": 0.8617, "step": 12600 }, { "epoch": 0.64, "learning_rate": 6.056612166883067e-06, "loss": 0.7679, "step": 12601 }, { "epoch": 0.64, "learning_rate": 6.055100140386142e-06, "loss": 0.9061, "step": 12602 }, { "epoch": 0.64, "learning_rate": 6.053588220693179e-06, "loss": 0.9154, "step": 12603 }, { "epoch": 0.64, "learning_rate": 6.05207640784512e-06, "loss": 0.8174, "step": 12604 }, { "epoch": 0.64, "learning_rate": 6.050564701882892e-06, "loss": 0.9433, "step": 12605 }, { "epoch": 0.64, "learning_rate": 6.04905310284742e-06, "loss": 0.8492, "step": 12606 }, { "epoch": 0.64, "learning_rate": 6.047541610779635e-06, "loss": 0.8774, "step": 12607 }, { "epoch": 0.64, "learning_rate": 6.046030225720456e-06, "loss": 0.8403, "step": 12608 }, { "epoch": 0.64, "learning_rate": 6.044518947710797e-06, "loss": 0.795, "step": 12609 }, { "epoch": 0.64, "learning_rate": 6.043007776791583e-06, "loss": 0.8631, "step": 12610 }, { "epoch": 0.64, "learning_rate": 6.041496713003725e-06, "loss": 0.8651, "step": 12611 }, { "epoch": 0.64, "learning_rate": 6.039985756388129e-06, "loss": 0.9159, "step": 12612 }, { "epoch": 0.64, "learning_rate": 6.038474906985708e-06, "loss": 0.8891, "step": 12613 }, { "epoch": 0.64, "learning_rate": 6.036964164837362e-06, "loss": 0.9098, "step": 12614 }, { "epoch": 0.64, "learning_rate": 6.0354535299839935e-06, "loss": 0.9448, "step": 12615 }, { "epoch": 0.64, "learning_rate": 6.033943002466506e-06, "loss": 0.9155, "step": 12616 }, { "epoch": 0.64, "learning_rate": 6.032432582325792e-06, "loss": 0.9208, "step": 12617 }, { "epoch": 0.64, "learning_rate": 6.030922269602742e-06, "loss": 0.7719, "step": 12618 }, { "epoch": 0.64, "learning_rate": 6.0294120643382516e-06, "loss": 0.8842, "step": 12619 }, { "epoch": 0.64, "learning_rate": 6.027901966573207e-06, "loss": 0.8438, "step": 12620 }, { "epoch": 0.64, "learning_rate": 6.026391976348484e-06, "loss": 0.8785, "step": 12621 }, { "epoch": 0.64, "learning_rate": 6.024882093704974e-06, "loss": 0.903, "step": 12622 }, { "epoch": 0.64, "learning_rate": 6.0233723186835534e-06, "loss": 0.9124, "step": 12623 }, { "epoch": 0.64, "learning_rate": 6.0218626513250955e-06, "loss": 0.8779, "step": 12624 }, { "epoch": 0.64, "learning_rate": 6.020353091670474e-06, "loss": 0.8684, "step": 12625 }, { "epoch": 0.64, "learning_rate": 6.018843639760561e-06, "loss": 0.958, "step": 12626 }, { "epoch": 0.64, "learning_rate": 6.0173342956362145e-06, "loss": 0.855, "step": 12627 }, { "epoch": 0.64, "learning_rate": 6.01582505933831e-06, "loss": 0.9459, "step": 12628 }, { "epoch": 0.64, "learning_rate": 6.0143159309077035e-06, "loss": 0.9781, "step": 12629 }, { "epoch": 0.64, "learning_rate": 6.012806910385249e-06, "loss": 0.9785, "step": 12630 }, { "epoch": 0.64, "learning_rate": 6.011297997811809e-06, "loss": 1.0079, "step": 12631 }, { "epoch": 0.64, "learning_rate": 6.009789193228232e-06, "loss": 0.9521, "step": 12632 }, { "epoch": 0.64, "learning_rate": 6.008280496675362e-06, "loss": 0.8962, "step": 12633 }, { "epoch": 0.64, "learning_rate": 6.006771908194055e-06, "loss": 0.9158, "step": 12634 }, { "epoch": 0.64, "learning_rate": 6.005263427825152e-06, "loss": 0.8645, "step": 12635 }, { "epoch": 0.64, "learning_rate": 6.003755055609488e-06, "loss": 0.9106, "step": 12636 }, { "epoch": 0.64, "learning_rate": 6.002246791587906e-06, "loss": 0.8954, "step": 12637 }, { "epoch": 0.64, "learning_rate": 6.000738635801235e-06, "loss": 0.8856, "step": 12638 }, { "epoch": 0.64, "learning_rate": 5.999230588290314e-06, "loss": 0.8345, "step": 12639 }, { "epoch": 0.64, "learning_rate": 5.997722649095969e-06, "loss": 0.9204, "step": 12640 }, { "epoch": 0.64, "learning_rate": 5.996214818259025e-06, "loss": 0.8247, "step": 12641 }, { "epoch": 0.64, "learning_rate": 5.994707095820304e-06, "loss": 0.9752, "step": 12642 }, { "epoch": 0.64, "learning_rate": 5.99319948182063e-06, "loss": 0.9303, "step": 12643 }, { "epoch": 0.64, "learning_rate": 5.9916919763008105e-06, "loss": 0.9367, "step": 12644 }, { "epoch": 0.64, "learning_rate": 5.990184579301672e-06, "loss": 0.8892, "step": 12645 }, { "epoch": 0.64, "learning_rate": 5.98867729086402e-06, "loss": 0.9486, "step": 12646 }, { "epoch": 0.64, "learning_rate": 5.98717011102866e-06, "loss": 0.8396, "step": 12647 }, { "epoch": 0.64, "learning_rate": 5.985663039836401e-06, "loss": 0.9291, "step": 12648 }, { "epoch": 0.64, "learning_rate": 5.984156077328045e-06, "loss": 0.9095, "step": 12649 }, { "epoch": 0.64, "learning_rate": 5.982649223544388e-06, "loss": 0.9365, "step": 12650 }, { "epoch": 0.64, "learning_rate": 5.9811424785262314e-06, "loss": 0.8732, "step": 12651 }, { "epoch": 0.64, "learning_rate": 5.9796358423143675e-06, "loss": 0.9064, "step": 12652 }, { "epoch": 0.64, "learning_rate": 5.978129314949583e-06, "loss": 0.9019, "step": 12653 }, { "epoch": 0.64, "learning_rate": 5.97662289647267e-06, "loss": 0.9459, "step": 12654 }, { "epoch": 0.64, "learning_rate": 5.975116586924413e-06, "loss": 1.028, "step": 12655 }, { "epoch": 0.64, "learning_rate": 5.973610386345588e-06, "loss": 0.9421, "step": 12656 }, { "epoch": 0.64, "learning_rate": 5.972104294776981e-06, "loss": 0.8972, "step": 12657 }, { "epoch": 0.64, "learning_rate": 5.970598312259366e-06, "loss": 0.9432, "step": 12658 }, { "epoch": 0.64, "learning_rate": 5.969092438833513e-06, "loss": 0.8682, "step": 12659 }, { "epoch": 0.64, "learning_rate": 5.967586674540194e-06, "loss": 0.9072, "step": 12660 }, { "epoch": 0.64, "learning_rate": 5.966081019420177e-06, "loss": 0.8374, "step": 12661 }, { "epoch": 0.64, "learning_rate": 5.96457547351422e-06, "loss": 0.9976, "step": 12662 }, { "epoch": 0.64, "learning_rate": 5.9630700368630926e-06, "loss": 0.7863, "step": 12663 }, { "epoch": 0.64, "learning_rate": 5.96156470950755e-06, "loss": 0.8672, "step": 12664 }, { "epoch": 0.64, "learning_rate": 5.960059491488344e-06, "loss": 0.8109, "step": 12665 }, { "epoch": 0.64, "learning_rate": 5.95855438284623e-06, "loss": 0.8708, "step": 12666 }, { "epoch": 0.64, "learning_rate": 5.957049383621958e-06, "loss": 1.0086, "step": 12667 }, { "epoch": 0.64, "learning_rate": 5.955544493856267e-06, "loss": 0.7975, "step": 12668 }, { "epoch": 0.64, "learning_rate": 5.95403971358991e-06, "loss": 0.9004, "step": 12669 }, { "epoch": 0.64, "learning_rate": 5.952535042863624e-06, "loss": 0.9259, "step": 12670 }, { "epoch": 0.64, "learning_rate": 5.951030481718143e-06, "loss": 0.8842, "step": 12671 }, { "epoch": 0.64, "learning_rate": 5.949526030194206e-06, "loss": 0.8789, "step": 12672 }, { "epoch": 0.64, "learning_rate": 5.948021688332544e-06, "loss": 0.8596, "step": 12673 }, { "epoch": 0.64, "learning_rate": 5.946517456173877e-06, "loss": 0.913, "step": 12674 }, { "epoch": 0.64, "learning_rate": 5.945013333758943e-06, "loss": 0.9283, "step": 12675 }, { "epoch": 0.64, "learning_rate": 5.943509321128459e-06, "loss": 0.9327, "step": 12676 }, { "epoch": 0.64, "learning_rate": 5.942005418323142e-06, "loss": 0.7912, "step": 12677 }, { "epoch": 0.64, "learning_rate": 5.940501625383712e-06, "loss": 1.0263, "step": 12678 }, { "epoch": 0.64, "learning_rate": 5.938997942350884e-06, "loss": 0.8876, "step": 12679 }, { "epoch": 0.64, "learning_rate": 5.93749436926536e-06, "loss": 0.9081, "step": 12680 }, { "epoch": 0.64, "learning_rate": 5.935990906167858e-06, "loss": 0.9736, "step": 12681 }, { "epoch": 0.64, "learning_rate": 5.934487553099079e-06, "loss": 0.8412, "step": 12682 }, { "epoch": 0.64, "learning_rate": 5.932984310099721e-06, "loss": 0.9517, "step": 12683 }, { "epoch": 0.64, "learning_rate": 5.93148117721049e-06, "loss": 0.9491, "step": 12684 }, { "epoch": 0.64, "learning_rate": 5.929978154472076e-06, "loss": 1.0024, "step": 12685 }, { "epoch": 0.64, "learning_rate": 5.928475241925169e-06, "loss": 0.8447, "step": 12686 }, { "epoch": 0.64, "learning_rate": 5.9269724396104675e-06, "loss": 0.9674, "step": 12687 }, { "epoch": 0.64, "learning_rate": 5.925469747568654e-06, "loss": 0.8301, "step": 12688 }, { "epoch": 0.64, "learning_rate": 5.92396716584041e-06, "loss": 0.9623, "step": 12689 }, { "epoch": 0.64, "learning_rate": 5.922464694466421e-06, "loss": 0.8417, "step": 12690 }, { "epoch": 0.64, "learning_rate": 5.920962333487361e-06, "loss": 0.9274, "step": 12691 }, { "epoch": 0.64, "learning_rate": 5.919460082943904e-06, "loss": 0.8289, "step": 12692 }, { "epoch": 0.64, "learning_rate": 5.9179579428767266e-06, "loss": 0.8134, "step": 12693 }, { "epoch": 0.64, "learning_rate": 5.916455913326495e-06, "loss": 0.8687, "step": 12694 }, { "epoch": 0.64, "learning_rate": 5.914953994333873e-06, "loss": 0.9537, "step": 12695 }, { "epoch": 0.64, "learning_rate": 5.913452185939528e-06, "loss": 0.8339, "step": 12696 }, { "epoch": 0.65, "learning_rate": 5.9119504881841164e-06, "loss": 0.8851, "step": 12697 }, { "epoch": 0.65, "learning_rate": 5.910448901108292e-06, "loss": 0.7796, "step": 12698 }, { "epoch": 0.65, "learning_rate": 5.908947424752718e-06, "loss": 0.9822, "step": 12699 }, { "epoch": 0.65, "learning_rate": 5.9074460591580384e-06, "loss": 0.8834, "step": 12700 }, { "epoch": 0.65, "learning_rate": 5.905944804364902e-06, "loss": 0.8182, "step": 12701 }, { "epoch": 0.65, "learning_rate": 5.9044436604139545e-06, "loss": 0.8283, "step": 12702 }, { "epoch": 0.65, "learning_rate": 5.902942627345839e-06, "loss": 0.8873, "step": 12703 }, { "epoch": 0.65, "learning_rate": 5.901441705201189e-06, "loss": 1.0149, "step": 12704 }, { "epoch": 0.65, "learning_rate": 5.8999408940206485e-06, "loss": 0.8947, "step": 12705 }, { "epoch": 0.65, "learning_rate": 5.898440193844844e-06, "loss": 0.9295, "step": 12706 }, { "epoch": 0.65, "learning_rate": 5.896939604714408e-06, "loss": 0.9357, "step": 12707 }, { "epoch": 0.65, "learning_rate": 5.895439126669968e-06, "loss": 0.9384, "step": 12708 }, { "epoch": 0.65, "learning_rate": 5.8939387597521425e-06, "loss": 0.8408, "step": 12709 }, { "epoch": 0.65, "learning_rate": 5.892438504001561e-06, "loss": 0.8851, "step": 12710 }, { "epoch": 0.65, "learning_rate": 5.890938359458837e-06, "loss": 0.9266, "step": 12711 }, { "epoch": 0.65, "learning_rate": 5.889438326164583e-06, "loss": 0.9201, "step": 12712 }, { "epoch": 0.65, "learning_rate": 5.887938404159415e-06, "loss": 0.9073, "step": 12713 }, { "epoch": 0.65, "learning_rate": 5.8864385934839385e-06, "loss": 0.7832, "step": 12714 }, { "epoch": 0.65, "learning_rate": 5.884938894178758e-06, "loss": 0.843, "step": 12715 }, { "epoch": 0.65, "learning_rate": 5.883439306284483e-06, "loss": 0.9001, "step": 12716 }, { "epoch": 0.65, "learning_rate": 5.881939829841708e-06, "loss": 1.0, "step": 12717 }, { "epoch": 0.65, "learning_rate": 5.880440464891031e-06, "loss": 0.9134, "step": 12718 }, { "epoch": 0.65, "learning_rate": 5.878941211473046e-06, "loss": 0.7512, "step": 12719 }, { "epoch": 0.65, "learning_rate": 5.877442069628343e-06, "loss": 0.8817, "step": 12720 }, { "epoch": 0.65, "learning_rate": 5.875943039397507e-06, "loss": 0.9695, "step": 12721 }, { "epoch": 0.65, "learning_rate": 5.874444120821128e-06, "loss": 0.9135, "step": 12722 }, { "epoch": 0.65, "learning_rate": 5.872945313939786e-06, "loss": 0.8373, "step": 12723 }, { "epoch": 0.65, "learning_rate": 5.871446618794057e-06, "loss": 0.8206, "step": 12724 }, { "epoch": 0.65, "learning_rate": 5.8699480354245196e-06, "loss": 0.9172, "step": 12725 }, { "epoch": 0.65, "learning_rate": 5.868449563871747e-06, "loss": 0.9125, "step": 12726 }, { "epoch": 0.65, "learning_rate": 5.866951204176301e-06, "loss": 0.9287, "step": 12727 }, { "epoch": 0.65, "learning_rate": 5.8654529563787584e-06, "loss": 0.8839, "step": 12728 }, { "epoch": 0.65, "learning_rate": 5.8639548205196785e-06, "loss": 0.8889, "step": 12729 }, { "epoch": 0.65, "learning_rate": 5.86245679663962e-06, "loss": 0.9218, "step": 12730 }, { "epoch": 0.65, "learning_rate": 5.860958884779142e-06, "loss": 0.939, "step": 12731 }, { "epoch": 0.65, "learning_rate": 5.859461084978799e-06, "loss": 0.8975, "step": 12732 }, { "epoch": 0.65, "learning_rate": 5.8579633972791395e-06, "loss": 0.9343, "step": 12733 }, { "epoch": 0.65, "learning_rate": 5.856465821720717e-06, "loss": 0.8516, "step": 12734 }, { "epoch": 0.65, "learning_rate": 5.8549683583440746e-06, "loss": 0.8515, "step": 12735 }, { "epoch": 0.65, "learning_rate": 5.85347100718975e-06, "loss": 0.8856, "step": 12736 }, { "epoch": 0.65, "learning_rate": 5.851973768298292e-06, "loss": 0.8812, "step": 12737 }, { "epoch": 0.65, "learning_rate": 5.850476641710227e-06, "loss": 0.8543, "step": 12738 }, { "epoch": 0.65, "learning_rate": 5.848979627466092e-06, "loss": 0.9178, "step": 12739 }, { "epoch": 0.65, "learning_rate": 5.847482725606419e-06, "loss": 0.9241, "step": 12740 }, { "epoch": 0.65, "learning_rate": 5.8459859361717345e-06, "loss": 0.9232, "step": 12741 }, { "epoch": 0.65, "learning_rate": 5.844489259202557e-06, "loss": 0.9162, "step": 12742 }, { "epoch": 0.65, "learning_rate": 5.8429926947394155e-06, "loss": 0.8312, "step": 12743 }, { "epoch": 0.65, "learning_rate": 5.8414962428228235e-06, "loss": 0.8235, "step": 12744 }, { "epoch": 0.65, "learning_rate": 5.839999903493292e-06, "loss": 0.8971, "step": 12745 }, { "epoch": 0.65, "learning_rate": 5.8385036767913425e-06, "loss": 0.8764, "step": 12746 }, { "epoch": 0.65, "learning_rate": 5.83700756275748e-06, "loss": 0.8983, "step": 12747 }, { "epoch": 0.65, "learning_rate": 5.835511561432201e-06, "loss": 0.8676, "step": 12748 }, { "epoch": 0.65, "learning_rate": 5.834015672856023e-06, "loss": 0.8825, "step": 12749 }, { "epoch": 0.65, "learning_rate": 5.832519897069434e-06, "loss": 0.9003, "step": 12750 }, { "epoch": 0.65, "learning_rate": 5.831024234112937e-06, "loss": 0.8142, "step": 12751 }, { "epoch": 0.65, "learning_rate": 5.829528684027025e-06, "loss": 0.8614, "step": 12752 }, { "epoch": 0.65, "learning_rate": 5.828033246852184e-06, "loss": 0.9218, "step": 12753 }, { "epoch": 0.65, "learning_rate": 5.8265379226289035e-06, "loss": 0.9279, "step": 12754 }, { "epoch": 0.65, "learning_rate": 5.825042711397669e-06, "loss": 0.8815, "step": 12755 }, { "epoch": 0.65, "learning_rate": 5.8235476131989655e-06, "loss": 0.8589, "step": 12756 }, { "epoch": 0.65, "learning_rate": 5.8220526280732635e-06, "loss": 0.9103, "step": 12757 }, { "epoch": 0.65, "learning_rate": 5.82055775606104e-06, "loss": 0.9653, "step": 12758 }, { "epoch": 0.65, "learning_rate": 5.819062997202775e-06, "loss": 0.86, "step": 12759 }, { "epoch": 0.65, "learning_rate": 5.817568351538921e-06, "loss": 0.859, "step": 12760 }, { "epoch": 0.65, "learning_rate": 5.8160738191099645e-06, "loss": 0.889, "step": 12761 }, { "epoch": 0.65, "learning_rate": 5.8145793999563516e-06, "loss": 0.7852, "step": 12762 }, { "epoch": 0.65, "learning_rate": 5.813085094118549e-06, "loss": 0.8684, "step": 12763 }, { "epoch": 0.65, "learning_rate": 5.811590901637017e-06, "loss": 0.9468, "step": 12764 }, { "epoch": 0.65, "learning_rate": 5.810096822552202e-06, "loss": 0.9192, "step": 12765 }, { "epoch": 0.65, "learning_rate": 5.8086028569045574e-06, "loss": 0.8213, "step": 12766 }, { "epoch": 0.65, "learning_rate": 5.807109004734532e-06, "loss": 0.859, "step": 12767 }, { "epoch": 0.65, "learning_rate": 5.805615266082571e-06, "loss": 0.7878, "step": 12768 }, { "epoch": 0.65, "learning_rate": 5.804121640989113e-06, "loss": 0.8537, "step": 12769 }, { "epoch": 0.65, "learning_rate": 5.802628129494596e-06, "loss": 0.8231, "step": 12770 }, { "epoch": 0.65, "learning_rate": 5.801134731639461e-06, "loss": 0.8591, "step": 12771 }, { "epoch": 0.65, "learning_rate": 5.799641447464128e-06, "loss": 0.874, "step": 12772 }, { "epoch": 0.65, "learning_rate": 5.798148277009041e-06, "loss": 0.8907, "step": 12773 }, { "epoch": 0.65, "learning_rate": 5.79665522031462e-06, "loss": 0.8864, "step": 12774 }, { "epoch": 0.65, "learning_rate": 5.795162277421279e-06, "loss": 0.9625, "step": 12775 }, { "epoch": 0.65, "learning_rate": 5.793669448369455e-06, "loss": 0.837, "step": 12776 }, { "epoch": 0.65, "learning_rate": 5.792176733199545e-06, "loss": 0.8815, "step": 12777 }, { "epoch": 0.65, "learning_rate": 5.7906841319519825e-06, "loss": 0.8267, "step": 12778 }, { "epoch": 0.65, "learning_rate": 5.789191644667165e-06, "loss": 0.9155, "step": 12779 }, { "epoch": 0.65, "learning_rate": 5.7876992713855026e-06, "loss": 0.8622, "step": 12780 }, { "epoch": 0.65, "learning_rate": 5.786207012147404e-06, "loss": 0.913, "step": 12781 }, { "epoch": 0.65, "learning_rate": 5.784714866993267e-06, "loss": 0.8535, "step": 12782 }, { "epoch": 0.65, "learning_rate": 5.783222835963487e-06, "loss": 0.9053, "step": 12783 }, { "epoch": 0.65, "learning_rate": 5.7817309190984634e-06, "loss": 0.9151, "step": 12784 }, { "epoch": 0.65, "learning_rate": 5.780239116438592e-06, "loss": 0.8852, "step": 12785 }, { "epoch": 0.65, "learning_rate": 5.778747428024253e-06, "loss": 0.812, "step": 12786 }, { "epoch": 0.65, "learning_rate": 5.777255853895836e-06, "loss": 0.9982, "step": 12787 }, { "epoch": 0.65, "learning_rate": 5.775764394093728e-06, "loss": 0.9482, "step": 12788 }, { "epoch": 0.65, "learning_rate": 5.774273048658297e-06, "loss": 0.9717, "step": 12789 }, { "epoch": 0.65, "learning_rate": 5.772781817629935e-06, "loss": 0.9848, "step": 12790 }, { "epoch": 0.65, "learning_rate": 5.7712907010490036e-06, "loss": 0.8456, "step": 12791 }, { "epoch": 0.65, "learning_rate": 5.769799698955878e-06, "loss": 0.842, "step": 12792 }, { "epoch": 0.65, "learning_rate": 5.768308811390928e-06, "loss": 0.9099, "step": 12793 }, { "epoch": 0.65, "learning_rate": 5.766818038394512e-06, "loss": 0.8028, "step": 12794 }, { "epoch": 0.65, "learning_rate": 5.765327380006994e-06, "loss": 0.9723, "step": 12795 }, { "epoch": 0.65, "learning_rate": 5.763836836268731e-06, "loss": 0.9368, "step": 12796 }, { "epoch": 0.65, "learning_rate": 5.762346407220083e-06, "loss": 0.8571, "step": 12797 }, { "epoch": 0.65, "learning_rate": 5.760856092901394e-06, "loss": 0.9557, "step": 12798 }, { "epoch": 0.65, "learning_rate": 5.759365893353016e-06, "loss": 0.8889, "step": 12799 }, { "epoch": 0.65, "learning_rate": 5.757875808615299e-06, "loss": 0.8911, "step": 12800 }, { "epoch": 0.65, "learning_rate": 5.756385838728573e-06, "loss": 0.8654, "step": 12801 }, { "epoch": 0.65, "learning_rate": 5.754895983733193e-06, "loss": 0.925, "step": 12802 }, { "epoch": 0.65, "learning_rate": 5.7534062436694835e-06, "loss": 0.8993, "step": 12803 }, { "epoch": 0.65, "learning_rate": 5.751916618577783e-06, "loss": 0.9369, "step": 12804 }, { "epoch": 0.65, "learning_rate": 5.750427108498424e-06, "loss": 0.9641, "step": 12805 }, { "epoch": 0.65, "learning_rate": 5.748937713471726e-06, "loss": 0.9207, "step": 12806 }, { "epoch": 0.65, "learning_rate": 5.747448433538017e-06, "loss": 0.9117, "step": 12807 }, { "epoch": 0.65, "learning_rate": 5.745959268737616e-06, "loss": 0.828, "step": 12808 }, { "epoch": 0.65, "learning_rate": 5.744470219110847e-06, "loss": 0.8045, "step": 12809 }, { "epoch": 0.65, "learning_rate": 5.742981284698015e-06, "loss": 0.9227, "step": 12810 }, { "epoch": 0.65, "learning_rate": 5.7414924655394356e-06, "loss": 0.9548, "step": 12811 }, { "epoch": 0.65, "learning_rate": 5.740003761675421e-06, "loss": 0.9103, "step": 12812 }, { "epoch": 0.65, "learning_rate": 5.738515173146264e-06, "loss": 0.892, "step": 12813 }, { "epoch": 0.65, "learning_rate": 5.737026699992283e-06, "loss": 0.8412, "step": 12814 }, { "epoch": 0.65, "learning_rate": 5.735538342253769e-06, "loss": 0.8719, "step": 12815 }, { "epoch": 0.65, "learning_rate": 5.734050099971009e-06, "loss": 0.9293, "step": 12816 }, { "epoch": 0.65, "learning_rate": 5.732561973184312e-06, "loss": 0.8903, "step": 12817 }, { "epoch": 0.65, "learning_rate": 5.731073961933955e-06, "loss": 0.8871, "step": 12818 }, { "epoch": 0.65, "learning_rate": 5.72958606626023e-06, "loss": 0.9133, "step": 12819 }, { "epoch": 0.65, "learning_rate": 5.7280982862034185e-06, "loss": 0.9152, "step": 12820 }, { "epoch": 0.65, "learning_rate": 5.726610621803803e-06, "loss": 0.8836, "step": 12821 }, { "epoch": 0.65, "learning_rate": 5.725123073101657e-06, "loss": 0.8497, "step": 12822 }, { "epoch": 0.65, "learning_rate": 5.723635640137254e-06, "loss": 0.7959, "step": 12823 }, { "epoch": 0.65, "learning_rate": 5.722148322950871e-06, "loss": 0.8824, "step": 12824 }, { "epoch": 0.65, "learning_rate": 5.720661121582762e-06, "loss": 0.9419, "step": 12825 }, { "epoch": 0.65, "learning_rate": 5.7191740360732095e-06, "loss": 0.9137, "step": 12826 }, { "epoch": 0.65, "learning_rate": 5.717687066462465e-06, "loss": 0.9672, "step": 12827 }, { "epoch": 0.65, "learning_rate": 5.716200212790779e-06, "loss": 0.9203, "step": 12828 }, { "epoch": 0.65, "learning_rate": 5.714713475098423e-06, "loss": 0.953, "step": 12829 }, { "epoch": 0.65, "learning_rate": 5.713226853425637e-06, "loss": 0.8548, "step": 12830 }, { "epoch": 0.65, "learning_rate": 5.711740347812672e-06, "loss": 0.8129, "step": 12831 }, { "epoch": 0.65, "learning_rate": 5.710253958299776e-06, "loss": 0.8889, "step": 12832 }, { "epoch": 0.65, "learning_rate": 5.708767684927192e-06, "loss": 0.9023, "step": 12833 }, { "epoch": 0.65, "learning_rate": 5.707281527735156e-06, "loss": 0.8378, "step": 12834 }, { "epoch": 0.65, "learning_rate": 5.705795486763903e-06, "loss": 0.8775, "step": 12835 }, { "epoch": 0.65, "learning_rate": 5.704309562053675e-06, "loss": 0.8604, "step": 12836 }, { "epoch": 0.65, "learning_rate": 5.702823753644686e-06, "loss": 0.9021, "step": 12837 }, { "epoch": 0.65, "learning_rate": 5.70133806157718e-06, "loss": 0.9168, "step": 12838 }, { "epoch": 0.65, "learning_rate": 5.699852485891373e-06, "loss": 0.8007, "step": 12839 }, { "epoch": 0.65, "learning_rate": 5.698367026627477e-06, "loss": 0.9087, "step": 12840 }, { "epoch": 0.65, "learning_rate": 5.696881683825724e-06, "loss": 0.9212, "step": 12841 }, { "epoch": 0.65, "learning_rate": 5.695396457526319e-06, "loss": 1.0094, "step": 12842 }, { "epoch": 0.65, "learning_rate": 5.693911347769474e-06, "loss": 0.8743, "step": 12843 }, { "epoch": 0.65, "learning_rate": 5.692426354595403e-06, "loss": 0.9218, "step": 12844 }, { "epoch": 0.65, "learning_rate": 5.690941478044301e-06, "loss": 0.8353, "step": 12845 }, { "epoch": 0.65, "learning_rate": 5.689456718156374e-06, "loss": 0.9146, "step": 12846 }, { "epoch": 0.65, "learning_rate": 5.687972074971819e-06, "loss": 0.9291, "step": 12847 }, { "epoch": 0.65, "learning_rate": 5.6864875485308345e-06, "loss": 0.8287, "step": 12848 }, { "epoch": 0.65, "learning_rate": 5.685003138873609e-06, "loss": 0.8735, "step": 12849 }, { "epoch": 0.65, "learning_rate": 5.683518846040336e-06, "loss": 0.7959, "step": 12850 }, { "epoch": 0.65, "learning_rate": 5.682034670071197e-06, "loss": 0.9127, "step": 12851 }, { "epoch": 0.65, "learning_rate": 5.680550611006372e-06, "loss": 0.9238, "step": 12852 }, { "epoch": 0.65, "learning_rate": 5.679066668886051e-06, "loss": 0.7848, "step": 12853 }, { "epoch": 0.65, "learning_rate": 5.677582843750392e-06, "loss": 0.864, "step": 12854 }, { "epoch": 0.65, "learning_rate": 5.6760991356395875e-06, "loss": 0.9153, "step": 12855 }, { "epoch": 0.65, "learning_rate": 5.6746155445938e-06, "loss": 0.799, "step": 12856 }, { "epoch": 0.65, "learning_rate": 5.673132070653185e-06, "loss": 0.7567, "step": 12857 }, { "epoch": 0.65, "learning_rate": 5.671648713857925e-06, "loss": 0.9854, "step": 12858 }, { "epoch": 0.65, "learning_rate": 5.670165474248166e-06, "loss": 0.8887, "step": 12859 }, { "epoch": 0.65, "learning_rate": 5.66868235186407e-06, "loss": 0.754, "step": 12860 }, { "epoch": 0.65, "learning_rate": 5.667199346745793e-06, "loss": 0.8755, "step": 12861 }, { "epoch": 0.65, "learning_rate": 5.665716458933485e-06, "loss": 0.9161, "step": 12862 }, { "epoch": 0.65, "learning_rate": 5.664233688467291e-06, "loss": 0.9333, "step": 12863 }, { "epoch": 0.65, "learning_rate": 5.662751035387357e-06, "loss": 0.8956, "step": 12864 }, { "epoch": 0.65, "learning_rate": 5.6612684997338275e-06, "loss": 0.7917, "step": 12865 }, { "epoch": 0.65, "learning_rate": 5.6597860815468295e-06, "loss": 0.8125, "step": 12866 }, { "epoch": 0.65, "learning_rate": 5.658303780866514e-06, "loss": 0.8411, "step": 12867 }, { "epoch": 0.65, "learning_rate": 5.656821597733005e-06, "loss": 0.9123, "step": 12868 }, { "epoch": 0.65, "learning_rate": 5.6553395321864215e-06, "loss": 0.9494, "step": 12869 }, { "epoch": 0.65, "learning_rate": 5.653857584266905e-06, "loss": 0.8713, "step": 12870 }, { "epoch": 0.65, "learning_rate": 5.6523757540145675e-06, "loss": 0.8236, "step": 12871 }, { "epoch": 0.65, "learning_rate": 5.65089404146953e-06, "loss": 0.8753, "step": 12872 }, { "epoch": 0.65, "learning_rate": 5.649412446671915e-06, "loss": 0.877, "step": 12873 }, { "epoch": 0.65, "learning_rate": 5.647930969661823e-06, "loss": 0.8688, "step": 12874 }, { "epoch": 0.65, "learning_rate": 5.646449610479372e-06, "loss": 0.8598, "step": 12875 }, { "epoch": 0.65, "learning_rate": 5.644968369164665e-06, "loss": 0.8997, "step": 12876 }, { "epoch": 0.65, "learning_rate": 5.64348724575781e-06, "loss": 0.9032, "step": 12877 }, { "epoch": 0.65, "learning_rate": 5.642006240298896e-06, "loss": 0.8706, "step": 12878 }, { "epoch": 0.65, "learning_rate": 5.640525352828032e-06, "loss": 0.876, "step": 12879 }, { "epoch": 0.65, "learning_rate": 5.6390445833853095e-06, "loss": 0.8541, "step": 12880 }, { "epoch": 0.65, "learning_rate": 5.637563932010805e-06, "loss": 0.8542, "step": 12881 }, { "epoch": 0.65, "learning_rate": 5.636083398744623e-06, "loss": 0.961, "step": 12882 }, { "epoch": 0.65, "learning_rate": 5.634602983626838e-06, "loss": 0.9028, "step": 12883 }, { "epoch": 0.65, "learning_rate": 5.633122686697532e-06, "loss": 0.8747, "step": 12884 }, { "epoch": 0.65, "learning_rate": 5.631642507996786e-06, "loss": 0.8367, "step": 12885 }, { "epoch": 0.65, "learning_rate": 5.630162447564669e-06, "loss": 0.8824, "step": 12886 }, { "epoch": 0.65, "learning_rate": 5.6286825054412546e-06, "loss": 0.8614, "step": 12887 }, { "epoch": 0.65, "learning_rate": 5.627202681666611e-06, "loss": 0.899, "step": 12888 }, { "epoch": 0.65, "learning_rate": 5.6257229762808055e-06, "loss": 0.8862, "step": 12889 }, { "epoch": 0.65, "learning_rate": 5.624243389323889e-06, "loss": 0.9601, "step": 12890 }, { "epoch": 0.65, "learning_rate": 5.6227639208359345e-06, "loss": 0.871, "step": 12891 }, { "epoch": 0.65, "learning_rate": 5.621284570856991e-06, "loss": 0.869, "step": 12892 }, { "epoch": 0.65, "learning_rate": 5.6198053394271e-06, "loss": 0.8822, "step": 12893 }, { "epoch": 0.66, "learning_rate": 5.618326226586328e-06, "loss": 0.9748, "step": 12894 }, { "epoch": 0.66, "learning_rate": 5.616847232374707e-06, "loss": 0.882, "step": 12895 }, { "epoch": 0.66, "learning_rate": 5.615368356832285e-06, "loss": 0.9265, "step": 12896 }, { "epoch": 0.66, "learning_rate": 5.613889599999103e-06, "loss": 0.8555, "step": 12897 }, { "epoch": 0.66, "learning_rate": 5.612410961915191e-06, "loss": 0.9147, "step": 12898 }, { "epoch": 0.66, "learning_rate": 5.610932442620583e-06, "loss": 0.8928, "step": 12899 }, { "epoch": 0.66, "learning_rate": 5.609454042155309e-06, "loss": 0.8675, "step": 12900 }, { "epoch": 0.66, "learning_rate": 5.6079757605594005e-06, "loss": 0.9359, "step": 12901 }, { "epoch": 0.66, "learning_rate": 5.6064975978728675e-06, "loss": 0.9278, "step": 12902 }, { "epoch": 0.66, "learning_rate": 5.605019554135746e-06, "loss": 0.9413, "step": 12903 }, { "epoch": 0.66, "learning_rate": 5.603541629388043e-06, "loss": 0.8898, "step": 12904 }, { "epoch": 0.66, "learning_rate": 5.602063823669766e-06, "loss": 0.8153, "step": 12905 }, { "epoch": 0.66, "learning_rate": 5.6005861370209406e-06, "loss": 0.9238, "step": 12906 }, { "epoch": 0.66, "learning_rate": 5.59910856948156e-06, "loss": 0.9315, "step": 12907 }, { "epoch": 0.66, "learning_rate": 5.597631121091631e-06, "loss": 0.9328, "step": 12908 }, { "epoch": 0.66, "learning_rate": 5.596153791891161e-06, "loss": 0.9349, "step": 12909 }, { "epoch": 0.66, "learning_rate": 5.59467658192014e-06, "loss": 0.8127, "step": 12910 }, { "epoch": 0.66, "learning_rate": 5.593199491218561e-06, "loss": 0.8548, "step": 12911 }, { "epoch": 0.66, "learning_rate": 5.591722519826418e-06, "loss": 0.8613, "step": 12912 }, { "epoch": 0.66, "learning_rate": 5.590245667783701e-06, "loss": 0.8635, "step": 12913 }, { "epoch": 0.66, "learning_rate": 5.588768935130388e-06, "loss": 0.8893, "step": 12914 }, { "epoch": 0.66, "learning_rate": 5.587292321906464e-06, "loss": 0.9183, "step": 12915 }, { "epoch": 0.66, "learning_rate": 5.585815828151904e-06, "loss": 0.8647, "step": 12916 }, { "epoch": 0.66, "learning_rate": 5.584339453906684e-06, "loss": 0.9371, "step": 12917 }, { "epoch": 0.66, "learning_rate": 5.5828631992107795e-06, "loss": 0.8868, "step": 12918 }, { "epoch": 0.66, "learning_rate": 5.581387064104147e-06, "loss": 0.879, "step": 12919 }, { "epoch": 0.66, "learning_rate": 5.5799110486267646e-06, "loss": 0.9195, "step": 12920 }, { "epoch": 0.66, "learning_rate": 5.578435152818591e-06, "loss": 0.978, "step": 12921 }, { "epoch": 0.66, "learning_rate": 5.576959376719572e-06, "loss": 0.8368, "step": 12922 }, { "epoch": 0.66, "learning_rate": 5.575483720369681e-06, "loss": 0.9128, "step": 12923 }, { "epoch": 0.66, "learning_rate": 5.5740081838088555e-06, "loss": 0.7583, "step": 12924 }, { "epoch": 0.66, "learning_rate": 5.57253276707705e-06, "loss": 0.8611, "step": 12925 }, { "epoch": 0.66, "learning_rate": 5.571057470214214e-06, "loss": 0.8676, "step": 12926 }, { "epoch": 0.66, "learning_rate": 5.5695822932602806e-06, "loss": 0.8528, "step": 12927 }, { "epoch": 0.66, "learning_rate": 5.568107236255194e-06, "loss": 0.9185, "step": 12928 }, { "epoch": 0.66, "learning_rate": 5.566632299238888e-06, "loss": 1.0152, "step": 12929 }, { "epoch": 0.66, "learning_rate": 5.565157482251299e-06, "loss": 0.8943, "step": 12930 }, { "epoch": 0.66, "learning_rate": 5.563682785332345e-06, "loss": 0.8642, "step": 12931 }, { "epoch": 0.66, "learning_rate": 5.562208208521969e-06, "loss": 0.9148, "step": 12932 }, { "epoch": 0.66, "learning_rate": 5.560733751860083e-06, "loss": 0.9295, "step": 12933 }, { "epoch": 0.66, "learning_rate": 5.5592594153865995e-06, "loss": 0.9192, "step": 12934 }, { "epoch": 0.66, "learning_rate": 5.5577851991414526e-06, "loss": 0.8855, "step": 12935 }, { "epoch": 0.66, "learning_rate": 5.55631110316454e-06, "loss": 0.879, "step": 12936 }, { "epoch": 0.66, "learning_rate": 5.5548371274957754e-06, "loss": 0.8516, "step": 12937 }, { "epoch": 0.66, "learning_rate": 5.553363272175072e-06, "loss": 0.8662, "step": 12938 }, { "epoch": 0.66, "learning_rate": 5.551889537242323e-06, "loss": 0.8741, "step": 12939 }, { "epoch": 0.66, "learning_rate": 5.550415922737433e-06, "loss": 0.9742, "step": 12940 }, { "epoch": 0.66, "learning_rate": 5.548942428700297e-06, "loss": 0.8994, "step": 12941 }, { "epoch": 0.66, "learning_rate": 5.547469055170813e-06, "loss": 0.858, "step": 12942 }, { "epoch": 0.66, "learning_rate": 5.5459958021888635e-06, "loss": 0.8123, "step": 12943 }, { "epoch": 0.66, "learning_rate": 5.544522669794339e-06, "loss": 0.8892, "step": 12944 }, { "epoch": 0.66, "learning_rate": 5.543049658027127e-06, "loss": 0.9301, "step": 12945 }, { "epoch": 0.66, "learning_rate": 5.541576766927095e-06, "loss": 0.9519, "step": 12946 }, { "epoch": 0.66, "learning_rate": 5.5401039965341375e-06, "loss": 0.8773, "step": 12947 }, { "epoch": 0.66, "learning_rate": 5.538631346888115e-06, "loss": 0.8891, "step": 12948 }, { "epoch": 0.66, "learning_rate": 5.537158818028903e-06, "loss": 0.9705, "step": 12949 }, { "epoch": 0.66, "learning_rate": 5.53568640999637e-06, "loss": 0.8416, "step": 12950 }, { "epoch": 0.66, "learning_rate": 5.534214122830375e-06, "loss": 0.8983, "step": 12951 }, { "epoch": 0.66, "learning_rate": 5.532741956570782e-06, "loss": 0.8068, "step": 12952 }, { "epoch": 0.66, "learning_rate": 5.531269911257446e-06, "loss": 0.8477, "step": 12953 }, { "epoch": 0.66, "learning_rate": 5.529797986930229e-06, "loss": 0.9351, "step": 12954 }, { "epoch": 0.66, "learning_rate": 5.52832618362897e-06, "loss": 0.9156, "step": 12955 }, { "epoch": 0.66, "learning_rate": 5.5268545013935235e-06, "loss": 0.9373, "step": 12956 }, { "epoch": 0.66, "learning_rate": 5.525382940263735e-06, "loss": 0.8416, "step": 12957 }, { "epoch": 0.66, "learning_rate": 5.523911500279435e-06, "loss": 0.8945, "step": 12958 }, { "epoch": 0.66, "learning_rate": 5.5224401814804776e-06, "loss": 0.7257, "step": 12959 }, { "epoch": 0.66, "learning_rate": 5.5209689839066835e-06, "loss": 0.9314, "step": 12960 }, { "epoch": 0.66, "learning_rate": 5.5194979075978905e-06, "loss": 0.8572, "step": 12961 }, { "epoch": 0.66, "learning_rate": 5.518026952593927e-06, "loss": 0.8951, "step": 12962 }, { "epoch": 0.66, "learning_rate": 5.516556118934612e-06, "loss": 0.9975, "step": 12963 }, { "epoch": 0.66, "learning_rate": 5.515085406659771e-06, "loss": 0.9674, "step": 12964 }, { "epoch": 0.66, "learning_rate": 5.513614815809218e-06, "loss": 0.9299, "step": 12965 }, { "epoch": 0.66, "learning_rate": 5.5121443464227786e-06, "loss": 0.962, "step": 12966 }, { "epoch": 0.66, "learning_rate": 5.51067399854025e-06, "loss": 0.8982, "step": 12967 }, { "epoch": 0.66, "learning_rate": 5.5092037722014476e-06, "loss": 0.9789, "step": 12968 }, { "epoch": 0.66, "learning_rate": 5.507733667446178e-06, "loss": 0.7603, "step": 12969 }, { "epoch": 0.66, "learning_rate": 5.506263684314233e-06, "loss": 0.9997, "step": 12970 }, { "epoch": 0.66, "learning_rate": 5.504793822845427e-06, "loss": 0.888, "step": 12971 }, { "epoch": 0.66, "learning_rate": 5.5033240830795395e-06, "loss": 0.9592, "step": 12972 }, { "epoch": 0.66, "learning_rate": 5.5018544650563684e-06, "loss": 0.8355, "step": 12973 }, { "epoch": 0.66, "learning_rate": 5.5003849688157075e-06, "loss": 0.909, "step": 12974 }, { "epoch": 0.66, "learning_rate": 5.498915594397332e-06, "loss": 0.9061, "step": 12975 }, { "epoch": 0.66, "learning_rate": 5.497446341841027e-06, "loss": 0.8387, "step": 12976 }, { "epoch": 0.66, "learning_rate": 5.4959772111865726e-06, "loss": 0.8875, "step": 12977 }, { "epoch": 0.66, "learning_rate": 5.494508202473748e-06, "loss": 0.955, "step": 12978 }, { "epoch": 0.66, "learning_rate": 5.4930393157423144e-06, "loss": 0.8965, "step": 12979 }, { "epoch": 0.66, "learning_rate": 5.491570551032048e-06, "loss": 0.8593, "step": 12980 }, { "epoch": 0.66, "learning_rate": 5.4901019083827165e-06, "loss": 1.0072, "step": 12981 }, { "epoch": 0.66, "learning_rate": 5.488633387834068e-06, "loss": 0.8209, "step": 12982 }, { "epoch": 0.66, "learning_rate": 5.48716498942588e-06, "loss": 0.824, "step": 12983 }, { "epoch": 0.66, "learning_rate": 5.485696713197894e-06, "loss": 0.7786, "step": 12984 }, { "epoch": 0.66, "learning_rate": 5.484228559189868e-06, "loss": 0.8973, "step": 12985 }, { "epoch": 0.66, "learning_rate": 5.482760527441553e-06, "loss": 0.8822, "step": 12986 }, { "epoch": 0.66, "learning_rate": 5.481292617992682e-06, "loss": 0.8883, "step": 12987 }, { "epoch": 0.66, "learning_rate": 5.479824830883014e-06, "loss": 0.8763, "step": 12988 }, { "epoch": 0.66, "learning_rate": 5.478357166152276e-06, "loss": 0.9025, "step": 12989 }, { "epoch": 0.66, "learning_rate": 5.476889623840207e-06, "loss": 1.0093, "step": 12990 }, { "epoch": 0.66, "learning_rate": 5.475422203986542e-06, "loss": 0.8767, "step": 12991 }, { "epoch": 0.66, "learning_rate": 5.473954906631005e-06, "loss": 0.9804, "step": 12992 }, { "epoch": 0.66, "learning_rate": 5.472487731813323e-06, "loss": 0.938, "step": 12993 }, { "epoch": 0.66, "learning_rate": 5.471020679573218e-06, "loss": 0.9221, "step": 12994 }, { "epoch": 0.66, "learning_rate": 5.469553749950413e-06, "loss": 0.9522, "step": 12995 }, { "epoch": 0.66, "learning_rate": 5.468086942984619e-06, "loss": 0.9175, "step": 12996 }, { "epoch": 0.66, "learning_rate": 5.466620258715547e-06, "loss": 0.8495, "step": 12997 }, { "epoch": 0.66, "learning_rate": 5.465153697182913e-06, "loss": 0.9447, "step": 12998 }, { "epoch": 0.66, "learning_rate": 5.46368725842641e-06, "loss": 0.8622, "step": 12999 }, { "epoch": 0.66, "learning_rate": 5.462220942485756e-06, "loss": 0.9124, "step": 13000 }, { "epoch": 0.66, "learning_rate": 5.4607547494006385e-06, "loss": 0.942, "step": 13001 }, { "epoch": 0.66, "learning_rate": 5.459288679210757e-06, "loss": 0.9021, "step": 13002 }, { "epoch": 0.66, "learning_rate": 5.457822731955807e-06, "loss": 0.9204, "step": 13003 }, { "epoch": 0.66, "learning_rate": 5.456356907675471e-06, "loss": 0.8523, "step": 13004 }, { "epoch": 0.66, "learning_rate": 5.454891206409436e-06, "loss": 0.8839, "step": 13005 }, { "epoch": 0.66, "learning_rate": 5.453425628197387e-06, "loss": 0.8967, "step": 13006 }, { "epoch": 0.66, "learning_rate": 5.451960173079007e-06, "loss": 0.8324, "step": 13007 }, { "epoch": 0.66, "learning_rate": 5.4504948410939605e-06, "loss": 0.8315, "step": 13008 }, { "epoch": 0.66, "learning_rate": 5.449029632281928e-06, "loss": 0.8701, "step": 13009 }, { "epoch": 0.66, "learning_rate": 5.447564546682577e-06, "loss": 0.8918, "step": 13010 }, { "epoch": 0.66, "learning_rate": 5.4460995843355695e-06, "loss": 0.9033, "step": 13011 }, { "epoch": 0.66, "learning_rate": 5.444634745280575e-06, "loss": 0.8824, "step": 13012 }, { "epoch": 0.66, "learning_rate": 5.4431700295572455e-06, "loss": 0.9345, "step": 13013 }, { "epoch": 0.66, "learning_rate": 5.441705437205238e-06, "loss": 0.8951, "step": 13014 }, { "epoch": 0.66, "learning_rate": 5.440240968264212e-06, "loss": 0.9255, "step": 13015 }, { "epoch": 0.66, "learning_rate": 5.438776622773805e-06, "loss": 0.8565, "step": 13016 }, { "epoch": 0.66, "learning_rate": 5.4373124007736686e-06, "loss": 0.8194, "step": 13017 }, { "epoch": 0.66, "learning_rate": 5.435848302303446e-06, "loss": 0.9908, "step": 13018 }, { "epoch": 0.66, "learning_rate": 5.434384327402777e-06, "loss": 0.9798, "step": 13019 }, { "epoch": 0.66, "learning_rate": 5.432920476111292e-06, "loss": 0.8568, "step": 13020 }, { "epoch": 0.66, "learning_rate": 5.431456748468627e-06, "loss": 0.9071, "step": 13021 }, { "epoch": 0.66, "learning_rate": 5.429993144514412e-06, "loss": 0.9114, "step": 13022 }, { "epoch": 0.66, "learning_rate": 5.428529664288264e-06, "loss": 0.8876, "step": 13023 }, { "epoch": 0.66, "learning_rate": 5.427066307829818e-06, "loss": 0.8012, "step": 13024 }, { "epoch": 0.66, "learning_rate": 5.425603075178684e-06, "loss": 0.8499, "step": 13025 }, { "epoch": 0.66, "learning_rate": 5.42413996637448e-06, "loss": 0.8707, "step": 13026 }, { "epoch": 0.66, "learning_rate": 5.42267698145682e-06, "loss": 0.8898, "step": 13027 }, { "epoch": 0.66, "learning_rate": 5.421214120465309e-06, "loss": 0.9006, "step": 13028 }, { "epoch": 0.66, "learning_rate": 5.419751383439553e-06, "loss": 0.8987, "step": 13029 }, { "epoch": 0.66, "learning_rate": 5.418288770419157e-06, "loss": 0.8644, "step": 13030 }, { "epoch": 0.66, "learning_rate": 5.416826281443721e-06, "loss": 0.8305, "step": 13031 }, { "epoch": 0.66, "learning_rate": 5.415363916552833e-06, "loss": 0.9138, "step": 13032 }, { "epoch": 0.66, "learning_rate": 5.413901675786088e-06, "loss": 0.8897, "step": 13033 }, { "epoch": 0.66, "learning_rate": 5.412439559183083e-06, "loss": 1.0099, "step": 13034 }, { "epoch": 0.66, "learning_rate": 5.4109775667833866e-06, "loss": 0.902, "step": 13035 }, { "epoch": 0.66, "learning_rate": 5.409515698626597e-06, "loss": 1.0153, "step": 13036 }, { "epoch": 0.66, "learning_rate": 5.408053954752283e-06, "loss": 0.8182, "step": 13037 }, { "epoch": 0.66, "learning_rate": 5.406592335200024e-06, "loss": 0.9273, "step": 13038 }, { "epoch": 0.66, "learning_rate": 5.405130840009393e-06, "loss": 0.9698, "step": 13039 }, { "epoch": 0.66, "learning_rate": 5.403669469219952e-06, "loss": 0.9159, "step": 13040 }, { "epoch": 0.66, "learning_rate": 5.40220822287127e-06, "loss": 0.8568, "step": 13041 }, { "epoch": 0.66, "learning_rate": 5.40074710100291e-06, "loss": 0.9071, "step": 13042 }, { "epoch": 0.66, "learning_rate": 5.399286103654432e-06, "loss": 0.8642, "step": 13043 }, { "epoch": 0.66, "learning_rate": 5.3978252308653835e-06, "loss": 0.9033, "step": 13044 }, { "epoch": 0.66, "learning_rate": 5.396364482675322e-06, "loss": 0.9177, "step": 13045 }, { "epoch": 0.66, "learning_rate": 5.394903859123797e-06, "loss": 0.9223, "step": 13046 }, { "epoch": 0.66, "learning_rate": 5.393443360250344e-06, "loss": 0.8264, "step": 13047 }, { "epoch": 0.66, "learning_rate": 5.391982986094517e-06, "loss": 0.8745, "step": 13048 }, { "epoch": 0.66, "learning_rate": 5.390522736695845e-06, "loss": 0.8679, "step": 13049 }, { "epoch": 0.66, "learning_rate": 5.389062612093865e-06, "loss": 0.9922, "step": 13050 }, { "epoch": 0.66, "learning_rate": 5.387602612328115e-06, "loss": 0.9056, "step": 13051 }, { "epoch": 0.66, "learning_rate": 5.386142737438111e-06, "loss": 0.8914, "step": 13052 }, { "epoch": 0.66, "learning_rate": 5.384682987463385e-06, "loss": 0.8484, "step": 13053 }, { "epoch": 0.66, "learning_rate": 5.383223362443457e-06, "loss": 0.9228, "step": 13054 }, { "epoch": 0.66, "learning_rate": 5.381763862417843e-06, "loss": 0.8402, "step": 13055 }, { "epoch": 0.66, "learning_rate": 5.380304487426064e-06, "loss": 1.0069, "step": 13056 }, { "epoch": 0.66, "learning_rate": 5.378845237507624e-06, "loss": 0.9583, "step": 13057 }, { "epoch": 0.66, "learning_rate": 5.37738611270203e-06, "loss": 0.9095, "step": 13058 }, { "epoch": 0.66, "learning_rate": 5.375927113048789e-06, "loss": 0.8761, "step": 13059 }, { "epoch": 0.66, "learning_rate": 5.374468238587407e-06, "loss": 0.8452, "step": 13060 }, { "epoch": 0.66, "learning_rate": 5.373009489357373e-06, "loss": 0.8806, "step": 13061 }, { "epoch": 0.66, "learning_rate": 5.371550865398182e-06, "loss": 0.8851, "step": 13062 }, { "epoch": 0.66, "learning_rate": 5.370092366749332e-06, "loss": 0.8558, "step": 13063 }, { "epoch": 0.66, "learning_rate": 5.3686339934502985e-06, "loss": 0.8923, "step": 13064 }, { "epoch": 0.66, "learning_rate": 5.367175745540579e-06, "loss": 0.8702, "step": 13065 }, { "epoch": 0.66, "learning_rate": 5.365717623059643e-06, "loss": 0.8496, "step": 13066 }, { "epoch": 0.66, "learning_rate": 5.3642596260469725e-06, "loss": 1.0006, "step": 13067 }, { "epoch": 0.66, "learning_rate": 5.362801754542044e-06, "loss": 0.8304, "step": 13068 }, { "epoch": 0.66, "learning_rate": 5.361344008584322e-06, "loss": 0.863, "step": 13069 }, { "epoch": 0.66, "learning_rate": 5.359886388213273e-06, "loss": 0.8698, "step": 13070 }, { "epoch": 0.66, "learning_rate": 5.358428893468366e-06, "loss": 0.9189, "step": 13071 }, { "epoch": 0.66, "learning_rate": 5.356971524389061e-06, "loss": 0.8353, "step": 13072 }, { "epoch": 0.66, "learning_rate": 5.355514281014809e-06, "loss": 0.8603, "step": 13073 }, { "epoch": 0.66, "learning_rate": 5.354057163385066e-06, "loss": 0.784, "step": 13074 }, { "epoch": 0.66, "learning_rate": 5.3526001715392865e-06, "loss": 0.9272, "step": 13075 }, { "epoch": 0.66, "learning_rate": 5.3511433055169046e-06, "loss": 0.903, "step": 13076 }, { "epoch": 0.66, "learning_rate": 5.34968656535738e-06, "loss": 0.9472, "step": 13077 }, { "epoch": 0.66, "learning_rate": 5.348229951100141e-06, "loss": 0.786, "step": 13078 }, { "epoch": 0.66, "learning_rate": 5.346773462784626e-06, "loss": 0.8153, "step": 13079 }, { "epoch": 0.66, "learning_rate": 5.3453171004502735e-06, "loss": 0.9584, "step": 13080 }, { "epoch": 0.66, "learning_rate": 5.343860864136503e-06, "loss": 0.9238, "step": 13081 }, { "epoch": 0.66, "learning_rate": 5.342404753882746e-06, "loss": 0.9315, "step": 13082 }, { "epoch": 0.66, "learning_rate": 5.3409487697284265e-06, "loss": 0.7799, "step": 13083 }, { "epoch": 0.66, "learning_rate": 5.339492911712966e-06, "loss": 0.8265, "step": 13084 }, { "epoch": 0.66, "learning_rate": 5.338037179875771e-06, "loss": 0.8821, "step": 13085 }, { "epoch": 0.66, "learning_rate": 5.336581574256261e-06, "loss": 0.9774, "step": 13086 }, { "epoch": 0.66, "learning_rate": 5.335126094893847e-06, "loss": 0.9364, "step": 13087 }, { "epoch": 0.66, "learning_rate": 5.333670741827922e-06, "loss": 0.929, "step": 13088 }, { "epoch": 0.66, "learning_rate": 5.332215515097906e-06, "loss": 0.8937, "step": 13089 }, { "epoch": 0.67, "learning_rate": 5.330760414743185e-06, "loss": 0.8451, "step": 13090 }, { "epoch": 0.67, "learning_rate": 5.329305440803158e-06, "loss": 0.8721, "step": 13091 }, { "epoch": 0.67, "learning_rate": 5.32785059331722e-06, "loss": 0.8823, "step": 13092 }, { "epoch": 0.67, "learning_rate": 5.326395872324752e-06, "loss": 0.889, "step": 13093 }, { "epoch": 0.67, "learning_rate": 5.324941277865145e-06, "loss": 0.8404, "step": 13094 }, { "epoch": 0.67, "learning_rate": 5.323486809977778e-06, "loss": 0.911, "step": 13095 }, { "epoch": 0.67, "learning_rate": 5.322032468702037e-06, "loss": 0.8693, "step": 13096 }, { "epoch": 0.67, "learning_rate": 5.320578254077282e-06, "loss": 0.8495, "step": 13097 }, { "epoch": 0.67, "learning_rate": 5.319124166142895e-06, "loss": 0.8433, "step": 13098 }, { "epoch": 0.67, "learning_rate": 5.317670204938243e-06, "loss": 0.8532, "step": 13099 }, { "epoch": 0.67, "learning_rate": 5.316216370502683e-06, "loss": 0.8745, "step": 13100 }, { "epoch": 0.67, "learning_rate": 5.314762662875588e-06, "loss": 0.8692, "step": 13101 }, { "epoch": 0.67, "learning_rate": 5.313309082096305e-06, "loss": 0.8891, "step": 13102 }, { "epoch": 0.67, "learning_rate": 5.311855628204192e-06, "loss": 0.8601, "step": 13103 }, { "epoch": 0.67, "learning_rate": 5.3104023012386056e-06, "loss": 0.8646, "step": 13104 }, { "epoch": 0.67, "learning_rate": 5.308949101238883e-06, "loss": 0.9437, "step": 13105 }, { "epoch": 0.67, "learning_rate": 5.307496028244372e-06, "loss": 0.8094, "step": 13106 }, { "epoch": 0.67, "learning_rate": 5.306043082294414e-06, "loss": 0.7898, "step": 13107 }, { "epoch": 0.67, "learning_rate": 5.30459026342835e-06, "loss": 0.8669, "step": 13108 }, { "epoch": 0.67, "learning_rate": 5.303137571685504e-06, "loss": 0.9143, "step": 13109 }, { "epoch": 0.67, "learning_rate": 5.3016850071052124e-06, "loss": 0.951, "step": 13110 }, { "epoch": 0.67, "learning_rate": 5.300232569726805e-06, "loss": 0.9056, "step": 13111 }, { "epoch": 0.67, "learning_rate": 5.298780259589591e-06, "loss": 0.8568, "step": 13112 }, { "epoch": 0.67, "learning_rate": 5.297328076732909e-06, "loss": 0.8394, "step": 13113 }, { "epoch": 0.67, "learning_rate": 5.29587602119606e-06, "loss": 0.8698, "step": 13114 }, { "epoch": 0.67, "learning_rate": 5.294424093018364e-06, "loss": 0.8422, "step": 13115 }, { "epoch": 0.67, "learning_rate": 5.292972292239135e-06, "loss": 0.9283, "step": 13116 }, { "epoch": 0.67, "learning_rate": 5.291520618897668e-06, "loss": 0.8472, "step": 13117 }, { "epoch": 0.67, "learning_rate": 5.29006907303327e-06, "loss": 0.9213, "step": 13118 }, { "epoch": 0.67, "learning_rate": 5.288617654685243e-06, "loss": 0.9381, "step": 13119 }, { "epoch": 0.67, "learning_rate": 5.287166363892884e-06, "loss": 0.782, "step": 13120 }, { "epoch": 0.67, "learning_rate": 5.285715200695476e-06, "loss": 0.8952, "step": 13121 }, { "epoch": 0.67, "learning_rate": 5.284264165132315e-06, "loss": 0.9145, "step": 13122 }, { "epoch": 0.67, "learning_rate": 5.282813257242682e-06, "loss": 0.8273, "step": 13123 }, { "epoch": 0.67, "learning_rate": 5.2813624770658645e-06, "loss": 0.8942, "step": 13124 }, { "epoch": 0.67, "learning_rate": 5.279911824641139e-06, "loss": 0.9573, "step": 13125 }, { "epoch": 0.67, "learning_rate": 5.278461300007778e-06, "loss": 0.9328, "step": 13126 }, { "epoch": 0.67, "learning_rate": 5.277010903205053e-06, "loss": 0.9088, "step": 13127 }, { "epoch": 0.67, "learning_rate": 5.2755606342722365e-06, "loss": 0.9493, "step": 13128 }, { "epoch": 0.67, "learning_rate": 5.274110493248581e-06, "loss": 0.8622, "step": 13129 }, { "epoch": 0.67, "learning_rate": 5.272660480173364e-06, "loss": 0.8877, "step": 13130 }, { "epoch": 0.67, "learning_rate": 5.2712105950858316e-06, "loss": 0.8244, "step": 13131 }, { "epoch": 0.67, "learning_rate": 5.269760838025241e-06, "loss": 0.9435, "step": 13132 }, { "epoch": 0.67, "learning_rate": 5.268311209030848e-06, "loss": 0.9779, "step": 13133 }, { "epoch": 0.67, "learning_rate": 5.266861708141888e-06, "loss": 0.8516, "step": 13134 }, { "epoch": 0.67, "learning_rate": 5.265412335397615e-06, "loss": 0.7674, "step": 13135 }, { "epoch": 0.67, "learning_rate": 5.263963090837263e-06, "loss": 0.8848, "step": 13136 }, { "epoch": 0.67, "learning_rate": 5.262513974500077e-06, "loss": 0.7953, "step": 13137 }, { "epoch": 0.67, "learning_rate": 5.261064986425282e-06, "loss": 0.7948, "step": 13138 }, { "epoch": 0.67, "learning_rate": 5.259616126652109e-06, "loss": 0.9363, "step": 13139 }, { "epoch": 0.67, "learning_rate": 5.258167395219791e-06, "loss": 0.9066, "step": 13140 }, { "epoch": 0.67, "learning_rate": 5.256718792167537e-06, "loss": 0.8354, "step": 13141 }, { "epoch": 0.67, "learning_rate": 5.255270317534585e-06, "loss": 1.0004, "step": 13142 }, { "epoch": 0.67, "learning_rate": 5.2538219713601344e-06, "loss": 0.9094, "step": 13143 }, { "epoch": 0.67, "learning_rate": 5.252373753683406e-06, "loss": 0.8385, "step": 13144 }, { "epoch": 0.67, "learning_rate": 5.250925664543611e-06, "loss": 0.9336, "step": 13145 }, { "epoch": 0.67, "learning_rate": 5.249477703979946e-06, "loss": 0.8475, "step": 13146 }, { "epoch": 0.67, "learning_rate": 5.248029872031619e-06, "loss": 0.8274, "step": 13147 }, { "epoch": 0.67, "learning_rate": 5.246582168737828e-06, "loss": 0.8793, "step": 13148 }, { "epoch": 0.67, "learning_rate": 5.245134594137771e-06, "loss": 0.8526, "step": 13149 }, { "epoch": 0.67, "learning_rate": 5.243687148270632e-06, "loss": 0.8384, "step": 13150 }, { "epoch": 0.67, "learning_rate": 5.242239831175603e-06, "loss": 0.8278, "step": 13151 }, { "epoch": 0.67, "learning_rate": 5.240792642891873e-06, "loss": 0.8825, "step": 13152 }, { "epoch": 0.67, "learning_rate": 5.2393455834586106e-06, "loss": 0.8398, "step": 13153 }, { "epoch": 0.67, "learning_rate": 5.23789865291501e-06, "loss": 0.8673, "step": 13154 }, { "epoch": 0.67, "learning_rate": 5.236451851300233e-06, "loss": 0.8746, "step": 13155 }, { "epoch": 0.67, "learning_rate": 5.2350051786534535e-06, "loss": 0.8999, "step": 13156 }, { "epoch": 0.67, "learning_rate": 5.233558635013842e-06, "loss": 0.8934, "step": 13157 }, { "epoch": 0.67, "learning_rate": 5.232112220420558e-06, "loss": 0.8903, "step": 13158 }, { "epoch": 0.67, "learning_rate": 5.230665934912762e-06, "loss": 0.9208, "step": 13159 }, { "epoch": 0.67, "learning_rate": 5.229219778529611e-06, "loss": 0.9139, "step": 13160 }, { "epoch": 0.67, "learning_rate": 5.2277737513102634e-06, "loss": 0.9301, "step": 13161 }, { "epoch": 0.67, "learning_rate": 5.22632785329386e-06, "loss": 0.8105, "step": 13162 }, { "epoch": 0.67, "learning_rate": 5.224882084519552e-06, "loss": 0.864, "step": 13163 }, { "epoch": 0.67, "learning_rate": 5.223436445026484e-06, "loss": 0.7258, "step": 13164 }, { "epoch": 0.67, "learning_rate": 5.221990934853784e-06, "loss": 0.8381, "step": 13165 }, { "epoch": 0.67, "learning_rate": 5.220545554040606e-06, "loss": 0.9512, "step": 13166 }, { "epoch": 0.67, "learning_rate": 5.219100302626067e-06, "loss": 0.8257, "step": 13167 }, { "epoch": 0.67, "learning_rate": 5.2176551806493e-06, "loss": 0.8615, "step": 13168 }, { "epoch": 0.67, "learning_rate": 5.216210188149436e-06, "loss": 0.9078, "step": 13169 }, { "epoch": 0.67, "learning_rate": 5.214765325165587e-06, "loss": 0.7985, "step": 13170 }, { "epoch": 0.67, "learning_rate": 5.213320591736877e-06, "loss": 0.8904, "step": 13171 }, { "epoch": 0.67, "learning_rate": 5.21187598790242e-06, "loss": 0.9203, "step": 13172 }, { "epoch": 0.67, "learning_rate": 5.2104315137013284e-06, "loss": 0.8332, "step": 13173 }, { "epoch": 0.67, "learning_rate": 5.208987169172703e-06, "loss": 0.8116, "step": 13174 }, { "epoch": 0.67, "learning_rate": 5.207542954355656e-06, "loss": 0.8834, "step": 13175 }, { "epoch": 0.67, "learning_rate": 5.206098869289286e-06, "loss": 0.8723, "step": 13176 }, { "epoch": 0.67, "learning_rate": 5.204654914012683e-06, "loss": 0.9367, "step": 13177 }, { "epoch": 0.67, "learning_rate": 5.203211088564952e-06, "loss": 0.8535, "step": 13178 }, { "epoch": 0.67, "learning_rate": 5.201767392985175e-06, "loss": 0.8483, "step": 13179 }, { "epoch": 0.67, "learning_rate": 5.20032382731244e-06, "loss": 0.9065, "step": 13180 }, { "epoch": 0.67, "learning_rate": 5.198880391585835e-06, "loss": 0.8642, "step": 13181 }, { "epoch": 0.67, "learning_rate": 5.197437085844433e-06, "loss": 0.9166, "step": 13182 }, { "epoch": 0.67, "learning_rate": 5.195993910127311e-06, "loss": 0.8627, "step": 13183 }, { "epoch": 0.67, "learning_rate": 5.194550864473542e-06, "loss": 0.8563, "step": 13184 }, { "epoch": 0.67, "learning_rate": 5.193107948922201e-06, "loss": 0.8231, "step": 13185 }, { "epoch": 0.67, "learning_rate": 5.1916651635123445e-06, "loss": 0.9306, "step": 13186 }, { "epoch": 0.67, "learning_rate": 5.190222508283037e-06, "loss": 0.9529, "step": 13187 }, { "epoch": 0.67, "learning_rate": 5.188779983273342e-06, "loss": 0.8138, "step": 13188 }, { "epoch": 0.67, "learning_rate": 5.187337588522303e-06, "loss": 0.8201, "step": 13189 }, { "epoch": 0.67, "learning_rate": 5.185895324068986e-06, "loss": 0.873, "step": 13190 }, { "epoch": 0.67, "learning_rate": 5.184453189952427e-06, "loss": 0.9013, "step": 13191 }, { "epoch": 0.67, "learning_rate": 5.183011186211675e-06, "loss": 0.9258, "step": 13192 }, { "epoch": 0.67, "learning_rate": 5.181569312885775e-06, "loss": 0.8983, "step": 13193 }, { "epoch": 0.67, "learning_rate": 5.180127570013751e-06, "loss": 0.9251, "step": 13194 }, { "epoch": 0.67, "learning_rate": 5.178685957634653e-06, "loss": 0.9234, "step": 13195 }, { "epoch": 0.67, "learning_rate": 5.177244475787497e-06, "loss": 0.911, "step": 13196 }, { "epoch": 0.67, "learning_rate": 5.175803124511319e-06, "loss": 0.8864, "step": 13197 }, { "epoch": 0.67, "learning_rate": 5.174361903845142e-06, "loss": 0.8105, "step": 13198 }, { "epoch": 0.67, "learning_rate": 5.172920813827977e-06, "loss": 0.8908, "step": 13199 }, { "epoch": 0.67, "learning_rate": 5.171479854498847e-06, "loss": 0.9606, "step": 13200 }, { "epoch": 0.67, "learning_rate": 5.170039025896762e-06, "loss": 0.9174, "step": 13201 }, { "epoch": 0.67, "learning_rate": 5.1685983280607355e-06, "loss": 0.9603, "step": 13202 }, { "epoch": 0.67, "learning_rate": 5.167157761029766e-06, "loss": 0.956, "step": 13203 }, { "epoch": 0.67, "learning_rate": 5.165717324842857e-06, "loss": 0.8965, "step": 13204 }, { "epoch": 0.67, "learning_rate": 5.164277019539013e-06, "loss": 0.9103, "step": 13205 }, { "epoch": 0.67, "learning_rate": 5.162836845157217e-06, "loss": 0.8503, "step": 13206 }, { "epoch": 0.67, "learning_rate": 5.161396801736473e-06, "loss": 0.9064, "step": 13207 }, { "epoch": 0.67, "learning_rate": 5.159956889315761e-06, "loss": 0.8443, "step": 13208 }, { "epoch": 0.67, "learning_rate": 5.158517107934065e-06, "loss": 0.9166, "step": 13209 }, { "epoch": 0.67, "learning_rate": 5.15707745763037e-06, "loss": 0.9417, "step": 13210 }, { "epoch": 0.67, "learning_rate": 5.155637938443649e-06, "loss": 0.9661, "step": 13211 }, { "epoch": 0.67, "learning_rate": 5.154198550412877e-06, "loss": 0.9323, "step": 13212 }, { "epoch": 0.67, "learning_rate": 5.152759293577022e-06, "loss": 0.8756, "step": 13213 }, { "epoch": 0.67, "learning_rate": 5.151320167975057e-06, "loss": 0.8469, "step": 13214 }, { "epoch": 0.67, "learning_rate": 5.149881173645935e-06, "loss": 0.878, "step": 13215 }, { "epoch": 0.67, "learning_rate": 5.1484423106286205e-06, "loss": 0.8963, "step": 13216 }, { "epoch": 0.67, "learning_rate": 5.147003578962073e-06, "loss": 0.8657, "step": 13217 }, { "epoch": 0.67, "learning_rate": 5.145564978685234e-06, "loss": 0.8512, "step": 13218 }, { "epoch": 0.67, "learning_rate": 5.144126509837064e-06, "loss": 0.8367, "step": 13219 }, { "epoch": 0.67, "learning_rate": 5.1426881724565006e-06, "loss": 0.8943, "step": 13220 }, { "epoch": 0.67, "learning_rate": 5.141249966582485e-06, "loss": 0.8916, "step": 13221 }, { "epoch": 0.67, "learning_rate": 5.139811892253963e-06, "loss": 0.8434, "step": 13222 }, { "epoch": 0.67, "learning_rate": 5.138373949509858e-06, "loss": 0.8366, "step": 13223 }, { "epoch": 0.67, "learning_rate": 5.136936138389106e-06, "loss": 0.8965, "step": 13224 }, { "epoch": 0.67, "learning_rate": 5.135498458930635e-06, "loss": 0.8388, "step": 13225 }, { "epoch": 0.67, "learning_rate": 5.134060911173371e-06, "loss": 0.8688, "step": 13226 }, { "epoch": 0.67, "learning_rate": 5.1326234951562284e-06, "loss": 0.8235, "step": 13227 }, { "epoch": 0.67, "learning_rate": 5.131186210918124e-06, "loss": 0.8773, "step": 13228 }, { "epoch": 0.67, "learning_rate": 5.129749058497978e-06, "loss": 0.7855, "step": 13229 }, { "epoch": 0.67, "learning_rate": 5.128312037934687e-06, "loss": 0.8876, "step": 13230 }, { "epoch": 0.67, "learning_rate": 5.126875149267172e-06, "loss": 0.8923, "step": 13231 }, { "epoch": 0.67, "learning_rate": 5.125438392534324e-06, "loss": 0.8323, "step": 13232 }, { "epoch": 0.67, "learning_rate": 5.124001767775047e-06, "loss": 0.9008, "step": 13233 }, { "epoch": 0.67, "learning_rate": 5.122565275028235e-06, "loss": 0.9133, "step": 13234 }, { "epoch": 0.67, "learning_rate": 5.121128914332777e-06, "loss": 0.8539, "step": 13235 }, { "epoch": 0.67, "learning_rate": 5.119692685727563e-06, "loss": 0.918, "step": 13236 }, { "epoch": 0.67, "learning_rate": 5.118256589251477e-06, "loss": 0.8419, "step": 13237 }, { "epoch": 0.67, "learning_rate": 5.116820624943403e-06, "loss": 0.931, "step": 13238 }, { "epoch": 0.67, "learning_rate": 5.1153847928422126e-06, "loss": 0.8555, "step": 13239 }, { "epoch": 0.67, "learning_rate": 5.113949092986783e-06, "loss": 0.9051, "step": 13240 }, { "epoch": 0.67, "learning_rate": 5.112513525415985e-06, "loss": 0.8794, "step": 13241 }, { "epoch": 0.67, "learning_rate": 5.111078090168677e-06, "loss": 0.9036, "step": 13242 }, { "epoch": 0.67, "learning_rate": 5.109642787283737e-06, "loss": 0.8565, "step": 13243 }, { "epoch": 0.67, "learning_rate": 5.108207616800011e-06, "loss": 0.9068, "step": 13244 }, { "epoch": 0.67, "learning_rate": 5.106772578756358e-06, "loss": 0.8809, "step": 13245 }, { "epoch": 0.67, "learning_rate": 5.105337673191637e-06, "loss": 1.0453, "step": 13246 }, { "epoch": 0.67, "learning_rate": 5.103902900144686e-06, "loss": 0.8517, "step": 13247 }, { "epoch": 0.67, "learning_rate": 5.102468259654356e-06, "loss": 0.8813, "step": 13248 }, { "epoch": 0.67, "learning_rate": 5.101033751759485e-06, "loss": 0.9181, "step": 13249 }, { "epoch": 0.67, "learning_rate": 5.099599376498919e-06, "loss": 0.8887, "step": 13250 }, { "epoch": 0.67, "learning_rate": 5.098165133911482e-06, "loss": 0.8511, "step": 13251 }, { "epoch": 0.67, "learning_rate": 5.0967310240360105e-06, "loss": 0.9442, "step": 13252 }, { "epoch": 0.67, "learning_rate": 5.095297046911332e-06, "loss": 0.8622, "step": 13253 }, { "epoch": 0.67, "learning_rate": 5.093863202576259e-06, "loss": 0.8444, "step": 13254 }, { "epoch": 0.67, "learning_rate": 5.09242949106963e-06, "loss": 0.79, "step": 13255 }, { "epoch": 0.67, "learning_rate": 5.090995912430246e-06, "loss": 0.9182, "step": 13256 }, { "epoch": 0.67, "learning_rate": 5.089562466696925e-06, "loss": 0.9258, "step": 13257 }, { "epoch": 0.67, "learning_rate": 5.088129153908479e-06, "loss": 0.7666, "step": 13258 }, { "epoch": 0.67, "learning_rate": 5.086695974103708e-06, "loss": 0.9233, "step": 13259 }, { "epoch": 0.67, "learning_rate": 5.085262927321415e-06, "loss": 0.8303, "step": 13260 }, { "epoch": 0.67, "learning_rate": 5.0838300136004005e-06, "loss": 0.8381, "step": 13261 }, { "epoch": 0.67, "learning_rate": 5.082397232979457e-06, "loss": 0.906, "step": 13262 }, { "epoch": 0.67, "learning_rate": 5.080964585497381e-06, "loss": 0.8481, "step": 13263 }, { "epoch": 0.67, "learning_rate": 5.079532071192952e-06, "loss": 0.7838, "step": 13264 }, { "epoch": 0.67, "learning_rate": 5.078099690104956e-06, "loss": 0.853, "step": 13265 }, { "epoch": 0.67, "learning_rate": 5.0766674422721765e-06, "loss": 0.8692, "step": 13266 }, { "epoch": 0.67, "learning_rate": 5.07523532773339e-06, "loss": 0.904, "step": 13267 }, { "epoch": 0.67, "learning_rate": 5.073803346527364e-06, "loss": 0.9204, "step": 13268 }, { "epoch": 0.67, "learning_rate": 5.072371498692871e-06, "loss": 0.8651, "step": 13269 }, { "epoch": 0.67, "learning_rate": 5.07093978426868e-06, "loss": 0.8413, "step": 13270 }, { "epoch": 0.67, "learning_rate": 5.069508203293545e-06, "loss": 0.8911, "step": 13271 }, { "epoch": 0.67, "learning_rate": 5.068076755806236e-06, "loss": 0.8088, "step": 13272 }, { "epoch": 0.67, "learning_rate": 5.066645441845498e-06, "loss": 0.8779, "step": 13273 }, { "epoch": 0.67, "learning_rate": 5.065214261450086e-06, "loss": 0.9009, "step": 13274 }, { "epoch": 0.67, "learning_rate": 5.063783214658752e-06, "loss": 0.8453, "step": 13275 }, { "epoch": 0.67, "learning_rate": 5.062352301510232e-06, "loss": 1.0199, "step": 13276 }, { "epoch": 0.67, "learning_rate": 5.06092152204327e-06, "loss": 0.9799, "step": 13277 }, { "epoch": 0.67, "learning_rate": 5.059490876296602e-06, "loss": 0.8139, "step": 13278 }, { "epoch": 0.67, "learning_rate": 5.058060364308965e-06, "loss": 0.939, "step": 13279 }, { "epoch": 0.67, "learning_rate": 5.0566299861190845e-06, "loss": 0.9688, "step": 13280 }, { "epoch": 0.67, "learning_rate": 5.055199741765687e-06, "loss": 0.9046, "step": 13281 }, { "epoch": 0.67, "learning_rate": 5.053769631287499e-06, "loss": 0.8069, "step": 13282 }, { "epoch": 0.67, "learning_rate": 5.05233965472323e-06, "loss": 0.9254, "step": 13283 }, { "epoch": 0.67, "learning_rate": 5.050909812111607e-06, "loss": 0.7679, "step": 13284 }, { "epoch": 0.67, "learning_rate": 5.0494801034913325e-06, "loss": 0.8576, "step": 13285 }, { "epoch": 0.67, "learning_rate": 5.048050528901116e-06, "loss": 0.8661, "step": 13286 }, { "epoch": 0.68, "learning_rate": 5.0466210883796685e-06, "loss": 0.7674, "step": 13287 }, { "epoch": 0.68, "learning_rate": 5.0451917819656795e-06, "loss": 0.925, "step": 13288 }, { "epoch": 0.68, "learning_rate": 5.043762609697852e-06, "loss": 0.9092, "step": 13289 }, { "epoch": 0.68, "learning_rate": 5.042333571614879e-06, "loss": 0.9208, "step": 13290 }, { "epoch": 0.68, "learning_rate": 5.040904667755454e-06, "loss": 0.7539, "step": 13291 }, { "epoch": 0.68, "learning_rate": 5.039475898158255e-06, "loss": 1.0231, "step": 13292 }, { "epoch": 0.68, "learning_rate": 5.038047262861969e-06, "loss": 0.9143, "step": 13293 }, { "epoch": 0.68, "learning_rate": 5.036618761905277e-06, "loss": 0.8561, "step": 13294 }, { "epoch": 0.68, "learning_rate": 5.035190395326844e-06, "loss": 0.945, "step": 13295 }, { "epoch": 0.68, "learning_rate": 5.033762163165358e-06, "loss": 0.8431, "step": 13296 }, { "epoch": 0.68, "learning_rate": 5.0323340654594725e-06, "loss": 0.8961, "step": 13297 }, { "epoch": 0.68, "learning_rate": 5.030906102247857e-06, "loss": 0.823, "step": 13298 }, { "epoch": 0.68, "learning_rate": 5.0294782735691775e-06, "loss": 0.7878, "step": 13299 }, { "epoch": 0.68, "learning_rate": 5.028050579462081e-06, "loss": 0.9048, "step": 13300 }, { "epoch": 0.68, "learning_rate": 5.026623019965223e-06, "loss": 0.8701, "step": 13301 }, { "epoch": 0.68, "learning_rate": 5.025195595117258e-06, "loss": 0.8888, "step": 13302 }, { "epoch": 0.68, "learning_rate": 5.023768304956833e-06, "loss": 0.8957, "step": 13303 }, { "epoch": 0.68, "learning_rate": 5.022341149522582e-06, "loss": 0.9775, "step": 13304 }, { "epoch": 0.68, "learning_rate": 5.0209141288531494e-06, "loss": 0.9409, "step": 13305 }, { "epoch": 0.68, "learning_rate": 5.019487242987173e-06, "loss": 0.8736, "step": 13306 }, { "epoch": 0.68, "learning_rate": 5.018060491963272e-06, "loss": 0.8314, "step": 13307 }, { "epoch": 0.68, "learning_rate": 5.016633875820092e-06, "loss": 0.9036, "step": 13308 }, { "epoch": 0.68, "learning_rate": 5.0152073945962445e-06, "loss": 0.874, "step": 13309 }, { "epoch": 0.68, "learning_rate": 5.013781048330353e-06, "loss": 0.9177, "step": 13310 }, { "epoch": 0.68, "learning_rate": 5.012354837061037e-06, "loss": 0.899, "step": 13311 }, { "epoch": 0.68, "learning_rate": 5.010928760826906e-06, "loss": 0.9784, "step": 13312 }, { "epoch": 0.68, "learning_rate": 5.009502819666568e-06, "loss": 0.8258, "step": 13313 }, { "epoch": 0.68, "learning_rate": 5.008077013618633e-06, "loss": 0.8854, "step": 13314 }, { "epoch": 0.68, "learning_rate": 5.0066513427217065e-06, "loss": 0.8628, "step": 13315 }, { "epoch": 0.68, "learning_rate": 5.005225807014378e-06, "loss": 0.9305, "step": 13316 }, { "epoch": 0.68, "learning_rate": 5.003800406535246e-06, "loss": 0.9037, "step": 13317 }, { "epoch": 0.68, "learning_rate": 5.002375141322906e-06, "loss": 0.8898, "step": 13318 }, { "epoch": 0.68, "learning_rate": 5.000950011415936e-06, "loss": 0.9047, "step": 13319 }, { "epoch": 0.68, "learning_rate": 4.999525016852933e-06, "loss": 0.8615, "step": 13320 }, { "epoch": 0.68, "learning_rate": 4.9981001576724655e-06, "loss": 0.9217, "step": 13321 }, { "epoch": 0.68, "learning_rate": 4.996675433913115e-06, "loss": 0.9359, "step": 13322 }, { "epoch": 0.68, "learning_rate": 4.995250845613458e-06, "loss": 0.8136, "step": 13323 }, { "epoch": 0.68, "learning_rate": 4.993826392812056e-06, "loss": 1.0459, "step": 13324 }, { "epoch": 0.68, "learning_rate": 4.992402075547478e-06, "loss": 0.8997, "step": 13325 }, { "epoch": 0.68, "learning_rate": 4.990977893858286e-06, "loss": 0.9123, "step": 13326 }, { "epoch": 0.68, "learning_rate": 4.989553847783043e-06, "loss": 0.8676, "step": 13327 }, { "epoch": 0.68, "learning_rate": 4.988129937360294e-06, "loss": 0.8546, "step": 13328 }, { "epoch": 0.68, "learning_rate": 4.986706162628596e-06, "loss": 0.8625, "step": 13329 }, { "epoch": 0.68, "learning_rate": 4.9852825236264975e-06, "loss": 0.8256, "step": 13330 }, { "epoch": 0.68, "learning_rate": 4.983859020392532e-06, "loss": 0.9005, "step": 13331 }, { "epoch": 0.68, "learning_rate": 4.982435652965255e-06, "loss": 0.9265, "step": 13332 }, { "epoch": 0.68, "learning_rate": 4.98101242138319e-06, "loss": 0.8116, "step": 13333 }, { "epoch": 0.68, "learning_rate": 4.979589325684875e-06, "loss": 0.8318, "step": 13334 }, { "epoch": 0.68, "learning_rate": 4.97816636590884e-06, "loss": 0.9419, "step": 13335 }, { "epoch": 0.68, "learning_rate": 4.976743542093601e-06, "loss": 0.9256, "step": 13336 }, { "epoch": 0.68, "learning_rate": 4.975320854277694e-06, "loss": 0.9265, "step": 13337 }, { "epoch": 0.68, "learning_rate": 4.973898302499625e-06, "loss": 0.8021, "step": 13338 }, { "epoch": 0.68, "learning_rate": 4.972475886797912e-06, "loss": 0.8652, "step": 13339 }, { "epoch": 0.68, "learning_rate": 4.971053607211069e-06, "loss": 0.8408, "step": 13340 }, { "epoch": 0.68, "learning_rate": 4.969631463777597e-06, "loss": 0.9314, "step": 13341 }, { "epoch": 0.68, "learning_rate": 4.968209456536e-06, "loss": 0.9328, "step": 13342 }, { "epoch": 0.68, "learning_rate": 4.966787585524778e-06, "loss": 0.8495, "step": 13343 }, { "epoch": 0.68, "learning_rate": 4.965365850782432e-06, "loss": 0.9226, "step": 13344 }, { "epoch": 0.68, "learning_rate": 4.9639442523474445e-06, "loss": 0.7705, "step": 13345 }, { "epoch": 0.68, "learning_rate": 4.96252279025831e-06, "loss": 0.9244, "step": 13346 }, { "epoch": 0.68, "learning_rate": 4.961101464553514e-06, "loss": 0.9171, "step": 13347 }, { "epoch": 0.68, "learning_rate": 4.959680275271526e-06, "loss": 0.9299, "step": 13348 }, { "epoch": 0.68, "learning_rate": 4.958259222450842e-06, "loss": 0.836, "step": 13349 }, { "epoch": 0.68, "learning_rate": 4.956838306129921e-06, "loss": 0.9116, "step": 13350 }, { "epoch": 0.68, "learning_rate": 4.955417526347237e-06, "loss": 0.8069, "step": 13351 }, { "epoch": 0.68, "learning_rate": 4.953996883141261e-06, "loss": 0.9053, "step": 13352 }, { "epoch": 0.68, "learning_rate": 4.9525763765504466e-06, "loss": 0.8086, "step": 13353 }, { "epoch": 0.68, "learning_rate": 4.951156006613257e-06, "loss": 1.0073, "step": 13354 }, { "epoch": 0.68, "learning_rate": 4.949735773368146e-06, "loss": 0.8816, "step": 13355 }, { "epoch": 0.68, "learning_rate": 4.9483156768535714e-06, "loss": 0.8529, "step": 13356 }, { "epoch": 0.68, "learning_rate": 4.94689571710797e-06, "loss": 0.921, "step": 13357 }, { "epoch": 0.68, "learning_rate": 4.9454758941697915e-06, "loss": 0.8588, "step": 13358 }, { "epoch": 0.68, "learning_rate": 4.94405620807748e-06, "loss": 0.8296, "step": 13359 }, { "epoch": 0.68, "learning_rate": 4.942636658869459e-06, "loss": 0.8616, "step": 13360 }, { "epoch": 0.68, "learning_rate": 4.941217246584178e-06, "loss": 0.9213, "step": 13361 }, { "epoch": 0.68, "learning_rate": 4.9397979712600544e-06, "loss": 0.8547, "step": 13362 }, { "epoch": 0.68, "learning_rate": 4.938378832935515e-06, "loss": 0.9304, "step": 13363 }, { "epoch": 0.68, "learning_rate": 4.93695983164899e-06, "loss": 0.8779, "step": 13364 }, { "epoch": 0.68, "learning_rate": 4.9355409674388855e-06, "loss": 0.9527, "step": 13365 }, { "epoch": 0.68, "learning_rate": 4.934122240343622e-06, "loss": 0.9467, "step": 13366 }, { "epoch": 0.68, "learning_rate": 4.932703650401609e-06, "loss": 0.8918, "step": 13367 }, { "epoch": 0.68, "learning_rate": 4.9312851976512575e-06, "loss": 0.9246, "step": 13368 }, { "epoch": 0.68, "learning_rate": 4.929866882130962e-06, "loss": 0.9315, "step": 13369 }, { "epoch": 0.68, "learning_rate": 4.928448703879127e-06, "loss": 0.9313, "step": 13370 }, { "epoch": 0.68, "learning_rate": 4.927030662934151e-06, "loss": 0.8385, "step": 13371 }, { "epoch": 0.68, "learning_rate": 4.925612759334416e-06, "loss": 0.9822, "step": 13372 }, { "epoch": 0.68, "learning_rate": 4.924194993118324e-06, "loss": 0.9212, "step": 13373 }, { "epoch": 0.68, "learning_rate": 4.922777364324249e-06, "loss": 0.8763, "step": 13374 }, { "epoch": 0.68, "learning_rate": 4.921359872990573e-06, "loss": 0.8553, "step": 13375 }, { "epoch": 0.68, "learning_rate": 4.91994251915568e-06, "loss": 0.8978, "step": 13376 }, { "epoch": 0.68, "learning_rate": 4.918525302857936e-06, "loss": 0.934, "step": 13377 }, { "epoch": 0.68, "learning_rate": 4.9171082241357126e-06, "loss": 0.8549, "step": 13378 }, { "epoch": 0.68, "learning_rate": 4.915691283027376e-06, "loss": 0.8506, "step": 13379 }, { "epoch": 0.68, "learning_rate": 4.914274479571292e-06, "loss": 0.8962, "step": 13380 }, { "epoch": 0.68, "learning_rate": 4.912857813805812e-06, "loss": 0.8527, "step": 13381 }, { "epoch": 0.68, "learning_rate": 4.911441285769295e-06, "loss": 0.8485, "step": 13382 }, { "epoch": 0.68, "learning_rate": 4.910024895500095e-06, "loss": 0.8762, "step": 13383 }, { "epoch": 0.68, "learning_rate": 4.908608643036548e-06, "loss": 0.8489, "step": 13384 }, { "epoch": 0.68, "learning_rate": 4.907192528417013e-06, "loss": 0.886, "step": 13385 }, { "epoch": 0.68, "learning_rate": 4.9057765516798175e-06, "loss": 0.8698, "step": 13386 }, { "epoch": 0.68, "learning_rate": 4.904360712863303e-06, "loss": 0.8325, "step": 13387 }, { "epoch": 0.68, "learning_rate": 4.902945012005804e-06, "loss": 0.8289, "step": 13388 }, { "epoch": 0.68, "learning_rate": 4.901529449145642e-06, "loss": 0.8838, "step": 13389 }, { "epoch": 0.68, "learning_rate": 4.9001140243211466e-06, "loss": 0.7514, "step": 13390 }, { "epoch": 0.68, "learning_rate": 4.898698737570638e-06, "loss": 0.7968, "step": 13391 }, { "epoch": 0.68, "learning_rate": 4.897283588932437e-06, "loss": 0.8829, "step": 13392 }, { "epoch": 0.68, "learning_rate": 4.895868578444849e-06, "loss": 0.9513, "step": 13393 }, { "epoch": 0.68, "learning_rate": 4.89445370614619e-06, "loss": 0.7783, "step": 13394 }, { "epoch": 0.68, "learning_rate": 4.893038972074769e-06, "loss": 0.8049, "step": 13395 }, { "epoch": 0.68, "learning_rate": 4.891624376268877e-06, "loss": 0.8241, "step": 13396 }, { "epoch": 0.68, "learning_rate": 4.890209918766828e-06, "loss": 0.9606, "step": 13397 }, { "epoch": 0.68, "learning_rate": 4.8887955996069045e-06, "loss": 0.8809, "step": 13398 }, { "epoch": 0.68, "learning_rate": 4.887381418827403e-06, "loss": 0.928, "step": 13399 }, { "epoch": 0.68, "learning_rate": 4.885967376466614e-06, "loss": 0.9076, "step": 13400 }, { "epoch": 0.68, "learning_rate": 4.884553472562809e-06, "loss": 0.8712, "step": 13401 }, { "epoch": 0.68, "learning_rate": 4.883139707154284e-06, "loss": 0.8799, "step": 13402 }, { "epoch": 0.68, "learning_rate": 4.881726080279304e-06, "loss": 0.8777, "step": 13403 }, { "epoch": 0.68, "learning_rate": 4.8803125919761455e-06, "loss": 0.9524, "step": 13404 }, { "epoch": 0.68, "learning_rate": 4.87889924228308e-06, "loss": 0.9348, "step": 13405 }, { "epoch": 0.68, "learning_rate": 4.8774860312383645e-06, "loss": 0.9058, "step": 13406 }, { "epoch": 0.68, "learning_rate": 4.876072958880266e-06, "loss": 0.9052, "step": 13407 }, { "epoch": 0.68, "learning_rate": 4.87466002524704e-06, "loss": 0.8777, "step": 13408 }, { "epoch": 0.68, "learning_rate": 4.8732472303769455e-06, "loss": 0.8667, "step": 13409 }, { "epoch": 0.68, "learning_rate": 4.8718345743082225e-06, "loss": 0.9198, "step": 13410 }, { "epoch": 0.68, "learning_rate": 4.870422057079124e-06, "loss": 0.8379, "step": 13411 }, { "epoch": 0.68, "learning_rate": 4.869009678727894e-06, "loss": 0.9965, "step": 13412 }, { "epoch": 0.68, "learning_rate": 4.8675974392927595e-06, "loss": 0.8523, "step": 13413 }, { "epoch": 0.68, "learning_rate": 4.866185338811973e-06, "loss": 0.8169, "step": 13414 }, { "epoch": 0.68, "learning_rate": 4.864773377323752e-06, "loss": 0.9168, "step": 13415 }, { "epoch": 0.68, "learning_rate": 4.863361554866328e-06, "loss": 0.9288, "step": 13416 }, { "epoch": 0.68, "learning_rate": 4.861949871477929e-06, "loss": 0.9184, "step": 13417 }, { "epoch": 0.68, "learning_rate": 4.8605383271967676e-06, "loss": 0.8659, "step": 13418 }, { "epoch": 0.68, "learning_rate": 4.859126922061062e-06, "loss": 0.8904, "step": 13419 }, { "epoch": 0.68, "learning_rate": 4.857715656109027e-06, "loss": 0.9069, "step": 13420 }, { "epoch": 0.68, "learning_rate": 4.856304529378872e-06, "loss": 0.8979, "step": 13421 }, { "epoch": 0.68, "learning_rate": 4.854893541908796e-06, "loss": 0.8587, "step": 13422 }, { "epoch": 0.68, "learning_rate": 4.853482693737004e-06, "loss": 0.7843, "step": 13423 }, { "epoch": 0.68, "learning_rate": 4.852071984901696e-06, "loss": 0.894, "step": 13424 }, { "epoch": 0.68, "learning_rate": 4.850661415441055e-06, "loss": 0.9702, "step": 13425 }, { "epoch": 0.68, "learning_rate": 4.849250985393285e-06, "loss": 0.8454, "step": 13426 }, { "epoch": 0.68, "learning_rate": 4.847840694796559e-06, "loss": 0.8462, "step": 13427 }, { "epoch": 0.68, "learning_rate": 4.846430543689067e-06, "loss": 0.8911, "step": 13428 }, { "epoch": 0.68, "learning_rate": 4.845020532108988e-06, "loss": 0.8749, "step": 13429 }, { "epoch": 0.68, "learning_rate": 4.84361066009449e-06, "loss": 0.8792, "step": 13430 }, { "epoch": 0.68, "learning_rate": 4.8422009276837475e-06, "loss": 0.901, "step": 13431 }, { "epoch": 0.68, "learning_rate": 4.840791334914927e-06, "loss": 0.9742, "step": 13432 }, { "epoch": 0.68, "learning_rate": 4.839381881826196e-06, "loss": 0.9218, "step": 13433 }, { "epoch": 0.68, "learning_rate": 4.837972568455707e-06, "loss": 0.8157, "step": 13434 }, { "epoch": 0.68, "learning_rate": 4.836563394841618e-06, "loss": 0.8162, "step": 13435 }, { "epoch": 0.68, "learning_rate": 4.835154361022086e-06, "loss": 0.9098, "step": 13436 }, { "epoch": 0.68, "learning_rate": 4.833745467035247e-06, "loss": 0.7964, "step": 13437 }, { "epoch": 0.68, "learning_rate": 4.83233671291926e-06, "loss": 0.8734, "step": 13438 }, { "epoch": 0.68, "learning_rate": 4.8309280987122555e-06, "loss": 0.8808, "step": 13439 }, { "epoch": 0.68, "learning_rate": 4.8295196244523725e-06, "loss": 0.8328, "step": 13440 }, { "epoch": 0.68, "learning_rate": 4.8281112901777495e-06, "loss": 0.9301, "step": 13441 }, { "epoch": 0.68, "learning_rate": 4.826703095926506e-06, "loss": 0.9035, "step": 13442 }, { "epoch": 0.68, "learning_rate": 4.8252950417367724e-06, "loss": 0.9396, "step": 13443 }, { "epoch": 0.68, "learning_rate": 4.823887127646672e-06, "loss": 0.9796, "step": 13444 }, { "epoch": 0.68, "learning_rate": 4.822479353694321e-06, "loss": 0.9651, "step": 13445 }, { "epoch": 0.68, "learning_rate": 4.821071719917831e-06, "loss": 0.8668, "step": 13446 }, { "epoch": 0.68, "learning_rate": 4.819664226355315e-06, "loss": 0.8809, "step": 13447 }, { "epoch": 0.68, "learning_rate": 4.818256873044882e-06, "loss": 0.8737, "step": 13448 }, { "epoch": 0.68, "learning_rate": 4.816849660024624e-06, "loss": 0.8585, "step": 13449 }, { "epoch": 0.68, "learning_rate": 4.815442587332654e-06, "loss": 0.9341, "step": 13450 }, { "epoch": 0.68, "learning_rate": 4.814035655007055e-06, "loss": 0.9012, "step": 13451 }, { "epoch": 0.68, "learning_rate": 4.812628863085925e-06, "loss": 0.8842, "step": 13452 }, { "epoch": 0.68, "learning_rate": 4.811222211607352e-06, "loss": 0.9199, "step": 13453 }, { "epoch": 0.68, "learning_rate": 4.809815700609415e-06, "loss": 0.7428, "step": 13454 }, { "epoch": 0.68, "learning_rate": 4.808409330130196e-06, "loss": 0.7877, "step": 13455 }, { "epoch": 0.68, "learning_rate": 4.80700310020777e-06, "loss": 0.7151, "step": 13456 }, { "epoch": 0.68, "learning_rate": 4.805597010880214e-06, "loss": 0.9088, "step": 13457 }, { "epoch": 0.68, "learning_rate": 4.804191062185588e-06, "loss": 0.8447, "step": 13458 }, { "epoch": 0.68, "learning_rate": 4.802785254161962e-06, "loss": 0.9861, "step": 13459 }, { "epoch": 0.68, "learning_rate": 4.8013795868474e-06, "loss": 0.8843, "step": 13460 }, { "epoch": 0.68, "learning_rate": 4.799974060279947e-06, "loss": 0.9073, "step": 13461 }, { "epoch": 0.68, "learning_rate": 4.7985686744976714e-06, "loss": 0.8646, "step": 13462 }, { "epoch": 0.68, "learning_rate": 4.79716342953861e-06, "loss": 0.8534, "step": 13463 }, { "epoch": 0.68, "learning_rate": 4.795758325440816e-06, "loss": 0.8004, "step": 13464 }, { "epoch": 0.68, "learning_rate": 4.7943533622423296e-06, "loss": 0.9253, "step": 13465 }, { "epoch": 0.68, "learning_rate": 4.792948539981186e-06, "loss": 0.8293, "step": 13466 }, { "epoch": 0.68, "learning_rate": 4.7915438586954186e-06, "loss": 0.9186, "step": 13467 }, { "epoch": 0.68, "learning_rate": 4.790139318423061e-06, "loss": 0.8943, "step": 13468 }, { "epoch": 0.68, "learning_rate": 4.7887349192021424e-06, "loss": 0.8816, "step": 13469 }, { "epoch": 0.68, "learning_rate": 4.787330661070678e-06, "loss": 0.8302, "step": 13470 }, { "epoch": 0.68, "learning_rate": 4.78592654406669e-06, "loss": 0.8482, "step": 13471 }, { "epoch": 0.68, "learning_rate": 4.7845225682281934e-06, "loss": 0.9254, "step": 13472 }, { "epoch": 0.68, "learning_rate": 4.783118733593199e-06, "loss": 0.8651, "step": 13473 }, { "epoch": 0.68, "learning_rate": 4.781715040199718e-06, "loss": 0.8734, "step": 13474 }, { "epoch": 0.68, "learning_rate": 4.7803114880857485e-06, "loss": 0.9495, "step": 13475 }, { "epoch": 0.68, "learning_rate": 4.77890807728929e-06, "loss": 0.8668, "step": 13476 }, { "epoch": 0.68, "learning_rate": 4.777504807848346e-06, "loss": 0.8351, "step": 13477 }, { "epoch": 0.68, "learning_rate": 4.776101679800894e-06, "loss": 0.8744, "step": 13478 }, { "epoch": 0.68, "learning_rate": 4.774698693184939e-06, "loss": 0.8487, "step": 13479 }, { "epoch": 0.68, "learning_rate": 4.773295848038455e-06, "loss": 0.8985, "step": 13480 }, { "epoch": 0.68, "learning_rate": 4.771893144399424e-06, "loss": 0.8103, "step": 13481 }, { "epoch": 0.68, "learning_rate": 4.770490582305829e-06, "loss": 0.9325, "step": 13482 }, { "epoch": 0.68, "learning_rate": 4.769088161795632e-06, "loss": 0.8142, "step": 13483 }, { "epoch": 0.69, "learning_rate": 4.767685882906809e-06, "loss": 0.8036, "step": 13484 }, { "epoch": 0.69, "learning_rate": 4.766283745677323e-06, "loss": 1.063, "step": 13485 }, { "epoch": 0.69, "learning_rate": 4.764881750145142e-06, "loss": 0.835, "step": 13486 }, { "epoch": 0.69, "learning_rate": 4.763479896348213e-06, "loss": 0.8041, "step": 13487 }, { "epoch": 0.69, "learning_rate": 4.762078184324494e-06, "loss": 0.8904, "step": 13488 }, { "epoch": 0.69, "learning_rate": 4.76067661411194e-06, "loss": 0.9769, "step": 13489 }, { "epoch": 0.69, "learning_rate": 4.759275185748485e-06, "loss": 0.8729, "step": 13490 }, { "epoch": 0.69, "learning_rate": 4.757873899272085e-06, "loss": 0.8975, "step": 13491 }, { "epoch": 0.69, "learning_rate": 4.75647275472067e-06, "loss": 0.8965, "step": 13492 }, { "epoch": 0.69, "learning_rate": 4.755071752132176e-06, "loss": 0.8551, "step": 13493 }, { "epoch": 0.69, "learning_rate": 4.7536708915445374e-06, "loss": 0.9081, "step": 13494 }, { "epoch": 0.69, "learning_rate": 4.752270172995674e-06, "loss": 0.9548, "step": 13495 }, { "epoch": 0.69, "learning_rate": 4.750869596523513e-06, "loss": 0.8225, "step": 13496 }, { "epoch": 0.69, "learning_rate": 4.749469162165972e-06, "loss": 0.7993, "step": 13497 }, { "epoch": 0.69, "learning_rate": 4.748068869960972e-06, "loss": 0.8857, "step": 13498 }, { "epoch": 0.69, "learning_rate": 4.746668719946416e-06, "loss": 0.8633, "step": 13499 }, { "epoch": 0.69, "learning_rate": 4.745268712160215e-06, "loss": 0.9372, "step": 13500 }, { "epoch": 0.69, "learning_rate": 4.743868846640276e-06, "loss": 0.8912, "step": 13501 }, { "epoch": 0.69, "learning_rate": 4.74246912342449e-06, "loss": 0.8087, "step": 13502 }, { "epoch": 0.69, "learning_rate": 4.7410695425507655e-06, "loss": 0.8921, "step": 13503 }, { "epoch": 0.69, "learning_rate": 4.739670104056984e-06, "loss": 0.9245, "step": 13504 }, { "epoch": 0.69, "learning_rate": 4.738270807981038e-06, "loss": 0.8922, "step": 13505 }, { "epoch": 0.69, "learning_rate": 4.736871654360816e-06, "loss": 0.8845, "step": 13506 }, { "epoch": 0.69, "learning_rate": 4.735472643234192e-06, "loss": 0.9643, "step": 13507 }, { "epoch": 0.69, "learning_rate": 4.734073774639043e-06, "loss": 0.8563, "step": 13508 }, { "epoch": 0.69, "learning_rate": 4.732675048613245e-06, "loss": 0.8703, "step": 13509 }, { "epoch": 0.69, "learning_rate": 4.73127646519467e-06, "loss": 0.997, "step": 13510 }, { "epoch": 0.69, "learning_rate": 4.729878024421174e-06, "loss": 0.9442, "step": 13511 }, { "epoch": 0.69, "learning_rate": 4.728479726330625e-06, "loss": 0.8633, "step": 13512 }, { "epoch": 0.69, "learning_rate": 4.727081570960882e-06, "loss": 0.9073, "step": 13513 }, { "epoch": 0.69, "learning_rate": 4.725683558349788e-06, "loss": 0.818, "step": 13514 }, { "epoch": 0.69, "learning_rate": 4.724285688535208e-06, "loss": 0.9231, "step": 13515 }, { "epoch": 0.69, "learning_rate": 4.722887961554977e-06, "loss": 0.8357, "step": 13516 }, { "epoch": 0.69, "learning_rate": 4.721490377446939e-06, "loss": 0.9616, "step": 13517 }, { "epoch": 0.69, "learning_rate": 4.720092936248938e-06, "loss": 0.9531, "step": 13518 }, { "epoch": 0.69, "learning_rate": 4.718695637998799e-06, "loss": 0.8286, "step": 13519 }, { "epoch": 0.69, "learning_rate": 4.717298482734356e-06, "loss": 0.9374, "step": 13520 }, { "epoch": 0.69, "learning_rate": 4.715901470493437e-06, "loss": 0.8708, "step": 13521 }, { "epoch": 0.69, "learning_rate": 4.714504601313868e-06, "loss": 0.8645, "step": 13522 }, { "epoch": 0.69, "learning_rate": 4.713107875233459e-06, "loss": 0.962, "step": 13523 }, { "epoch": 0.69, "learning_rate": 4.71171129229003e-06, "loss": 0.8411, "step": 13524 }, { "epoch": 0.69, "learning_rate": 4.710314852521395e-06, "loss": 0.8759, "step": 13525 }, { "epoch": 0.69, "learning_rate": 4.7089185559653495e-06, "loss": 0.8315, "step": 13526 }, { "epoch": 0.69, "learning_rate": 4.707522402659713e-06, "loss": 0.8827, "step": 13527 }, { "epoch": 0.69, "learning_rate": 4.706126392642273e-06, "loss": 0.8488, "step": 13528 }, { "epoch": 0.69, "learning_rate": 4.704730525950829e-06, "loss": 0.8788, "step": 13529 }, { "epoch": 0.69, "learning_rate": 4.703334802623175e-06, "loss": 0.7917, "step": 13530 }, { "epoch": 0.69, "learning_rate": 4.701939222697093e-06, "loss": 0.8019, "step": 13531 }, { "epoch": 0.69, "learning_rate": 4.7005437862103695e-06, "loss": 0.8519, "step": 13532 }, { "epoch": 0.69, "learning_rate": 4.699148493200786e-06, "loss": 0.9125, "step": 13533 }, { "epoch": 0.69, "learning_rate": 4.69775334370612e-06, "loss": 0.8468, "step": 13534 }, { "epoch": 0.69, "learning_rate": 4.696358337764138e-06, "loss": 0.8301, "step": 13535 }, { "epoch": 0.69, "learning_rate": 4.69496347541261e-06, "loss": 0.8988, "step": 13536 }, { "epoch": 0.69, "learning_rate": 4.693568756689306e-06, "loss": 0.8987, "step": 13537 }, { "epoch": 0.69, "learning_rate": 4.6921741816319755e-06, "loss": 0.8324, "step": 13538 }, { "epoch": 0.69, "learning_rate": 4.690779750278389e-06, "loss": 0.9057, "step": 13539 }, { "epoch": 0.69, "learning_rate": 4.689385462666288e-06, "loss": 0.8329, "step": 13540 }, { "epoch": 0.69, "learning_rate": 4.687991318833427e-06, "loss": 0.8881, "step": 13541 }, { "epoch": 0.69, "learning_rate": 4.686597318817552e-06, "loss": 0.8441, "step": 13542 }, { "epoch": 0.69, "learning_rate": 4.685203462656394e-06, "loss": 0.9005, "step": 13543 }, { "epoch": 0.69, "learning_rate": 4.683809750387706e-06, "loss": 0.7733, "step": 13544 }, { "epoch": 0.69, "learning_rate": 4.68241618204921e-06, "loss": 0.9252, "step": 13545 }, { "epoch": 0.69, "learning_rate": 4.681022757678637e-06, "loss": 0.8891, "step": 13546 }, { "epoch": 0.69, "learning_rate": 4.679629477313718e-06, "loss": 0.9854, "step": 13547 }, { "epoch": 0.69, "learning_rate": 4.6782363409921674e-06, "loss": 0.9253, "step": 13548 }, { "epoch": 0.69, "learning_rate": 4.676843348751705e-06, "loss": 0.8683, "step": 13549 }, { "epoch": 0.69, "learning_rate": 4.675450500630045e-06, "loss": 0.8142, "step": 13550 }, { "epoch": 0.69, "learning_rate": 4.674057796664903e-06, "loss": 0.8687, "step": 13551 }, { "epoch": 0.69, "learning_rate": 4.672665236893975e-06, "loss": 0.8752, "step": 13552 }, { "epoch": 0.69, "learning_rate": 4.671272821354969e-06, "loss": 0.9296, "step": 13553 }, { "epoch": 0.69, "learning_rate": 4.669880550085585e-06, "loss": 0.9307, "step": 13554 }, { "epoch": 0.69, "learning_rate": 4.668488423123507e-06, "loss": 0.8691, "step": 13555 }, { "epoch": 0.69, "learning_rate": 4.66709644050644e-06, "loss": 0.8865, "step": 13556 }, { "epoch": 0.69, "learning_rate": 4.66570460227206e-06, "loss": 0.8126, "step": 13557 }, { "epoch": 0.69, "learning_rate": 4.664312908458052e-06, "loss": 0.8205, "step": 13558 }, { "epoch": 0.69, "learning_rate": 4.6629213591021e-06, "loss": 0.9051, "step": 13559 }, { "epoch": 0.69, "learning_rate": 4.661529954241868e-06, "loss": 0.9334, "step": 13560 }, { "epoch": 0.69, "learning_rate": 4.6601386939150354e-06, "loss": 0.9293, "step": 13561 }, { "epoch": 0.69, "learning_rate": 4.6587475781592655e-06, "loss": 0.871, "step": 13562 }, { "epoch": 0.69, "learning_rate": 4.657356607012225e-06, "loss": 0.7779, "step": 13563 }, { "epoch": 0.69, "learning_rate": 4.655965780511568e-06, "loss": 0.8456, "step": 13564 }, { "epoch": 0.69, "learning_rate": 4.65457509869495e-06, "loss": 0.7828, "step": 13565 }, { "epoch": 0.69, "learning_rate": 4.65318456160003e-06, "loss": 0.9008, "step": 13566 }, { "epoch": 0.69, "learning_rate": 4.651794169264441e-06, "loss": 0.8171, "step": 13567 }, { "epoch": 0.69, "learning_rate": 4.6504039217258426e-06, "loss": 0.9333, "step": 13568 }, { "epoch": 0.69, "learning_rate": 4.649013819021861e-06, "loss": 0.8619, "step": 13569 }, { "epoch": 0.69, "learning_rate": 4.647623861190139e-06, "loss": 0.8188, "step": 13570 }, { "epoch": 0.69, "learning_rate": 4.64623404826831e-06, "loss": 0.8858, "step": 13571 }, { "epoch": 0.69, "learning_rate": 4.644844380293995e-06, "loss": 0.8254, "step": 13572 }, { "epoch": 0.69, "learning_rate": 4.64345485730482e-06, "loss": 0.8224, "step": 13573 }, { "epoch": 0.69, "learning_rate": 4.642065479338406e-06, "loss": 0.8854, "step": 13574 }, { "epoch": 0.69, "learning_rate": 4.640676246432373e-06, "loss": 0.9163, "step": 13575 }, { "epoch": 0.69, "learning_rate": 4.639287158624325e-06, "loss": 0.8418, "step": 13576 }, { "epoch": 0.69, "learning_rate": 4.637898215951874e-06, "loss": 0.8653, "step": 13577 }, { "epoch": 0.69, "learning_rate": 4.636509418452626e-06, "loss": 0.8523, "step": 13578 }, { "epoch": 0.69, "learning_rate": 4.635120766164174e-06, "loss": 0.96, "step": 13579 }, { "epoch": 0.69, "learning_rate": 4.6337322591241255e-06, "loss": 0.8874, "step": 13580 }, { "epoch": 0.69, "learning_rate": 4.632343897370063e-06, "loss": 0.9262, "step": 13581 }, { "epoch": 0.69, "learning_rate": 4.630955680939579e-06, "loss": 0.9098, "step": 13582 }, { "epoch": 0.69, "learning_rate": 4.629567609870262e-06, "loss": 0.8831, "step": 13583 }, { "epoch": 0.69, "learning_rate": 4.628179684199685e-06, "loss": 0.9534, "step": 13584 }, { "epoch": 0.69, "learning_rate": 4.626791903965426e-06, "loss": 0.8992, "step": 13585 }, { "epoch": 0.69, "learning_rate": 4.6254042692050615e-06, "loss": 0.8249, "step": 13586 }, { "epoch": 0.69, "learning_rate": 4.6240167799561605e-06, "loss": 0.8461, "step": 13587 }, { "epoch": 0.69, "learning_rate": 4.622629436256283e-06, "loss": 0.8413, "step": 13588 }, { "epoch": 0.69, "learning_rate": 4.621242238142992e-06, "loss": 0.8626, "step": 13589 }, { "epoch": 0.69, "learning_rate": 4.6198551856538484e-06, "loss": 0.9289, "step": 13590 }, { "epoch": 0.69, "learning_rate": 4.618468278826394e-06, "loss": 0.9039, "step": 13591 }, { "epoch": 0.69, "learning_rate": 4.617081517698194e-06, "loss": 0.8765, "step": 13592 }, { "epoch": 0.69, "learning_rate": 4.615694902306781e-06, "loss": 0.8653, "step": 13593 }, { "epoch": 0.69, "learning_rate": 4.614308432689699e-06, "loss": 0.81, "step": 13594 }, { "epoch": 0.69, "learning_rate": 4.61292210888449e-06, "loss": 0.8855, "step": 13595 }, { "epoch": 0.69, "learning_rate": 4.611535930928681e-06, "loss": 0.8028, "step": 13596 }, { "epoch": 0.69, "learning_rate": 4.6101498988598025e-06, "loss": 0.8992, "step": 13597 }, { "epoch": 0.69, "learning_rate": 4.608764012715382e-06, "loss": 0.9142, "step": 13598 }, { "epoch": 0.69, "learning_rate": 4.607378272532944e-06, "loss": 0.8815, "step": 13599 }, { "epoch": 0.69, "learning_rate": 4.605992678349998e-06, "loss": 0.8212, "step": 13600 }, { "epoch": 0.69, "learning_rate": 4.6046072302040605e-06, "loss": 0.7851, "step": 13601 }, { "epoch": 0.69, "learning_rate": 4.603221928132646e-06, "loss": 0.9863, "step": 13602 }, { "epoch": 0.69, "learning_rate": 4.601836772173249e-06, "loss": 0.8178, "step": 13603 }, { "epoch": 0.69, "learning_rate": 4.600451762363386e-06, "loss": 0.9205, "step": 13604 }, { "epoch": 0.69, "learning_rate": 4.599066898740543e-06, "loss": 0.8409, "step": 13605 }, { "epoch": 0.69, "learning_rate": 4.597682181342218e-06, "loss": 0.8844, "step": 13606 }, { "epoch": 0.69, "learning_rate": 4.5962976102059035e-06, "loss": 0.9883, "step": 13607 }, { "epoch": 0.69, "learning_rate": 4.594913185369077e-06, "loss": 0.8084, "step": 13608 }, { "epoch": 0.69, "learning_rate": 4.5935289068692275e-06, "loss": 0.9102, "step": 13609 }, { "epoch": 0.69, "learning_rate": 4.59214477474383e-06, "loss": 0.8334, "step": 13610 }, { "epoch": 0.69, "learning_rate": 4.5907607890303595e-06, "loss": 0.8609, "step": 13611 }, { "epoch": 0.69, "learning_rate": 4.5893769497662884e-06, "loss": 0.8019, "step": 13612 }, { "epoch": 0.69, "learning_rate": 4.587993256989079e-06, "loss": 0.9003, "step": 13613 }, { "epoch": 0.69, "learning_rate": 4.58660971073619e-06, "loss": 0.9534, "step": 13614 }, { "epoch": 0.69, "learning_rate": 4.585226311045087e-06, "loss": 0.8683, "step": 13615 }, { "epoch": 0.69, "learning_rate": 4.583843057953224e-06, "loss": 0.8584, "step": 13616 }, { "epoch": 0.69, "learning_rate": 4.5824599514980425e-06, "loss": 0.917, "step": 13617 }, { "epoch": 0.69, "learning_rate": 4.581076991716996e-06, "loss": 0.8811, "step": 13618 }, { "epoch": 0.69, "learning_rate": 4.579694178647527e-06, "loss": 0.9211, "step": 13619 }, { "epoch": 0.69, "learning_rate": 4.578311512327065e-06, "loss": 0.9074, "step": 13620 }, { "epoch": 0.69, "learning_rate": 4.576928992793057e-06, "loss": 0.8248, "step": 13621 }, { "epoch": 0.69, "learning_rate": 4.5755466200829235e-06, "loss": 0.9101, "step": 13622 }, { "epoch": 0.69, "learning_rate": 4.574164394234094e-06, "loss": 0.9213, "step": 13623 }, { "epoch": 0.69, "learning_rate": 4.572782315283995e-06, "loss": 0.9098, "step": 13624 }, { "epoch": 0.69, "learning_rate": 4.571400383270037e-06, "loss": 0.8363, "step": 13625 }, { "epoch": 0.69, "learning_rate": 4.570018598229638e-06, "loss": 0.9243, "step": 13626 }, { "epoch": 0.69, "learning_rate": 4.568636960200208e-06, "loss": 0.9328, "step": 13627 }, { "epoch": 0.69, "learning_rate": 4.5672554692191575e-06, "loss": 0.9188, "step": 13628 }, { "epoch": 0.69, "learning_rate": 4.5658741253238815e-06, "loss": 0.8228, "step": 13629 }, { "epoch": 0.69, "learning_rate": 4.5644929285517825e-06, "loss": 0.9493, "step": 13630 }, { "epoch": 0.69, "learning_rate": 4.563111878940259e-06, "loss": 0.9172, "step": 13631 }, { "epoch": 0.69, "learning_rate": 4.561730976526687e-06, "loss": 0.9094, "step": 13632 }, { "epoch": 0.69, "learning_rate": 4.560350221348473e-06, "loss": 0.8263, "step": 13633 }, { "epoch": 0.69, "learning_rate": 4.558969613442984e-06, "loss": 0.8303, "step": 13634 }, { "epoch": 0.69, "learning_rate": 4.5575891528476045e-06, "loss": 0.8664, "step": 13635 }, { "epoch": 0.69, "learning_rate": 4.556208839599713e-06, "loss": 0.8267, "step": 13636 }, { "epoch": 0.69, "learning_rate": 4.55482867373667e-06, "loss": 0.8457, "step": 13637 }, { "epoch": 0.69, "learning_rate": 4.553448655295849e-06, "loss": 0.9376, "step": 13638 }, { "epoch": 0.69, "learning_rate": 4.55206878431461e-06, "loss": 0.9401, "step": 13639 }, { "epoch": 0.69, "learning_rate": 4.550689060830317e-06, "loss": 0.8795, "step": 13640 }, { "epoch": 0.69, "learning_rate": 4.549309484880315e-06, "loss": 0.863, "step": 13641 }, { "epoch": 0.69, "learning_rate": 4.547930056501961e-06, "loss": 0.9025, "step": 13642 }, { "epoch": 0.69, "learning_rate": 4.5465507757326035e-06, "loss": 0.8321, "step": 13643 }, { "epoch": 0.69, "learning_rate": 4.5451716426095736e-06, "loss": 0.878, "step": 13644 }, { "epoch": 0.69, "learning_rate": 4.543792657170228e-06, "loss": 0.8792, "step": 13645 }, { "epoch": 0.69, "learning_rate": 4.542413819451884e-06, "loss": 0.834, "step": 13646 }, { "epoch": 0.69, "learning_rate": 4.541035129491882e-06, "loss": 0.927, "step": 13647 }, { "epoch": 0.69, "learning_rate": 4.539656587327548e-06, "loss": 0.892, "step": 13648 }, { "epoch": 0.69, "learning_rate": 4.5382781929962e-06, "loss": 0.8464, "step": 13649 }, { "epoch": 0.69, "learning_rate": 4.536899946535158e-06, "loss": 0.8121, "step": 13650 }, { "epoch": 0.69, "learning_rate": 4.5355218479817385e-06, "loss": 0.9163, "step": 13651 }, { "epoch": 0.69, "learning_rate": 4.534143897373255e-06, "loss": 0.9239, "step": 13652 }, { "epoch": 0.69, "learning_rate": 4.532766094747006e-06, "loss": 0.8829, "step": 13653 }, { "epoch": 0.69, "learning_rate": 4.5313884401403e-06, "loss": 0.7949, "step": 13654 }, { "epoch": 0.69, "learning_rate": 4.530010933590435e-06, "loss": 0.8112, "step": 13655 }, { "epoch": 0.69, "learning_rate": 4.528633575134699e-06, "loss": 0.9771, "step": 13656 }, { "epoch": 0.69, "learning_rate": 4.527256364810394e-06, "loss": 0.9258, "step": 13657 }, { "epoch": 0.69, "learning_rate": 4.525879302654798e-06, "loss": 0.9592, "step": 13658 }, { "epoch": 0.69, "learning_rate": 4.524502388705194e-06, "loss": 0.8175, "step": 13659 }, { "epoch": 0.69, "learning_rate": 4.523125622998868e-06, "loss": 0.8811, "step": 13660 }, { "epoch": 0.69, "learning_rate": 4.521749005573083e-06, "loss": 0.9185, "step": 13661 }, { "epoch": 0.69, "learning_rate": 4.520372536465116e-06, "loss": 0.8671, "step": 13662 }, { "epoch": 0.69, "learning_rate": 4.518996215712233e-06, "loss": 0.8221, "step": 13663 }, { "epoch": 0.69, "learning_rate": 4.5176200433517005e-06, "loss": 0.9507, "step": 13664 }, { "epoch": 0.69, "learning_rate": 4.516244019420769e-06, "loss": 0.8469, "step": 13665 }, { "epoch": 0.69, "learning_rate": 4.514868143956696e-06, "loss": 0.9376, "step": 13666 }, { "epoch": 0.69, "learning_rate": 4.5134924169967355e-06, "loss": 0.9193, "step": 13667 }, { "epoch": 0.69, "learning_rate": 4.5121168385781235e-06, "loss": 0.87, "step": 13668 }, { "epoch": 0.69, "learning_rate": 4.5107414087381165e-06, "loss": 0.9006, "step": 13669 }, { "epoch": 0.69, "learning_rate": 4.509366127513942e-06, "loss": 0.8825, "step": 13670 }, { "epoch": 0.69, "learning_rate": 4.507990994942839e-06, "loss": 0.8401, "step": 13671 }, { "epoch": 0.69, "learning_rate": 4.506616011062039e-06, "loss": 0.8527, "step": 13672 }, { "epoch": 0.69, "learning_rate": 4.5052411759087635e-06, "loss": 0.8128, "step": 13673 }, { "epoch": 0.69, "learning_rate": 4.503866489520237e-06, "loss": 0.7997, "step": 13674 }, { "epoch": 0.69, "learning_rate": 4.5024919519336774e-06, "loss": 0.8107, "step": 13675 }, { "epoch": 0.69, "learning_rate": 4.5011175631863035e-06, "loss": 1.0313, "step": 13676 }, { "epoch": 0.69, "learning_rate": 4.499743323315317e-06, "loss": 0.8947, "step": 13677 }, { "epoch": 0.69, "learning_rate": 4.498369232357928e-06, "loss": 0.8707, "step": 13678 }, { "epoch": 0.69, "learning_rate": 4.49699529035134e-06, "loss": 0.8684, "step": 13679 }, { "epoch": 0.69, "learning_rate": 4.4956214973327494e-06, "loss": 0.8536, "step": 13680 }, { "epoch": 0.7, "learning_rate": 4.494247853339354e-06, "loss": 0.8964, "step": 13681 }, { "epoch": 0.7, "learning_rate": 4.492874358408337e-06, "loss": 0.9017, "step": 13682 }, { "epoch": 0.7, "learning_rate": 4.491501012576886e-06, "loss": 0.8661, "step": 13683 }, { "epoch": 0.7, "learning_rate": 4.49012781588219e-06, "loss": 0.905, "step": 13684 }, { "epoch": 0.7, "learning_rate": 4.488754768361413e-06, "loss": 0.7963, "step": 13685 }, { "epoch": 0.7, "learning_rate": 4.4873818700517445e-06, "loss": 0.9435, "step": 13686 }, { "epoch": 0.7, "learning_rate": 4.486009120990344e-06, "loss": 0.8846, "step": 13687 }, { "epoch": 0.7, "learning_rate": 4.48463652121438e-06, "loss": 0.9184, "step": 13688 }, { "epoch": 0.7, "learning_rate": 4.483264070761018e-06, "loss": 1.0117, "step": 13689 }, { "epoch": 0.7, "learning_rate": 4.481891769667408e-06, "loss": 0.9207, "step": 13690 }, { "epoch": 0.7, "learning_rate": 4.4805196179707066e-06, "loss": 0.916, "step": 13691 }, { "epoch": 0.7, "learning_rate": 4.479147615708066e-06, "loss": 0.7942, "step": 13692 }, { "epoch": 0.7, "learning_rate": 4.477775762916633e-06, "loss": 0.8422, "step": 13693 }, { "epoch": 0.7, "learning_rate": 4.476404059633541e-06, "loss": 0.9105, "step": 13694 }, { "epoch": 0.7, "learning_rate": 4.475032505895935e-06, "loss": 0.8668, "step": 13695 }, { "epoch": 0.7, "learning_rate": 4.4736611017409474e-06, "loss": 0.8245, "step": 13696 }, { "epoch": 0.7, "learning_rate": 4.4722898472057e-06, "loss": 0.7922, "step": 13697 }, { "epoch": 0.7, "learning_rate": 4.470918742327329e-06, "loss": 0.8555, "step": 13698 }, { "epoch": 0.7, "learning_rate": 4.4695477871429495e-06, "loss": 0.8521, "step": 13699 }, { "epoch": 0.7, "learning_rate": 4.468176981689678e-06, "loss": 0.8387, "step": 13700 }, { "epoch": 0.7, "learning_rate": 4.4668063260046346e-06, "loss": 0.9544, "step": 13701 }, { "epoch": 0.7, "learning_rate": 4.465435820124918e-06, "loss": 0.9275, "step": 13702 }, { "epoch": 0.7, "learning_rate": 4.464065464087639e-06, "loss": 0.8015, "step": 13703 }, { "epoch": 0.7, "learning_rate": 4.462695257929898e-06, "loss": 0.888, "step": 13704 }, { "epoch": 0.7, "learning_rate": 4.461325201688794e-06, "loss": 0.9294, "step": 13705 }, { "epoch": 0.7, "learning_rate": 4.459955295401415e-06, "loss": 0.8686, "step": 13706 }, { "epoch": 0.7, "learning_rate": 4.458585539104852e-06, "loss": 0.9056, "step": 13707 }, { "epoch": 0.7, "learning_rate": 4.457215932836193e-06, "loss": 0.8759, "step": 13708 }, { "epoch": 0.7, "learning_rate": 4.455846476632509e-06, "loss": 0.9532, "step": 13709 }, { "epoch": 0.7, "learning_rate": 4.45447717053089e-06, "loss": 0.7984, "step": 13710 }, { "epoch": 0.7, "learning_rate": 4.453108014568398e-06, "loss": 0.8924, "step": 13711 }, { "epoch": 0.7, "learning_rate": 4.4517390087821055e-06, "loss": 0.8682, "step": 13712 }, { "epoch": 0.7, "learning_rate": 4.45037015320908e-06, "loss": 0.8646, "step": 13713 }, { "epoch": 0.7, "learning_rate": 4.449001447886374e-06, "loss": 0.9368, "step": 13714 }, { "epoch": 0.7, "learning_rate": 4.4476328928510495e-06, "loss": 0.9008, "step": 13715 }, { "epoch": 0.7, "learning_rate": 4.446264488140155e-06, "loss": 0.8858, "step": 13716 }, { "epoch": 0.7, "learning_rate": 4.4448962337907465e-06, "loss": 0.808, "step": 13717 }, { "epoch": 0.7, "learning_rate": 4.443528129839857e-06, "loss": 0.85, "step": 13718 }, { "epoch": 0.7, "learning_rate": 4.442160176324532e-06, "loss": 0.8093, "step": 13719 }, { "epoch": 0.7, "learning_rate": 4.440792373281811e-06, "loss": 0.8822, "step": 13720 }, { "epoch": 0.7, "learning_rate": 4.439424720748716e-06, "loss": 0.9615, "step": 13721 }, { "epoch": 0.7, "learning_rate": 4.438057218762287e-06, "loss": 0.8553, "step": 13722 }, { "epoch": 0.7, "learning_rate": 4.436689867359537e-06, "loss": 0.8225, "step": 13723 }, { "epoch": 0.7, "learning_rate": 4.4353226665774905e-06, "loss": 0.9028, "step": 13724 }, { "epoch": 0.7, "learning_rate": 4.433955616453167e-06, "loss": 0.8718, "step": 13725 }, { "epoch": 0.7, "learning_rate": 4.432588717023568e-06, "loss": 0.8376, "step": 13726 }, { "epoch": 0.7, "learning_rate": 4.431221968325707e-06, "loss": 0.803, "step": 13727 }, { "epoch": 0.7, "learning_rate": 4.429855370396587e-06, "loss": 0.847, "step": 13728 }, { "epoch": 0.7, "learning_rate": 4.428488923273211e-06, "loss": 0.9012, "step": 13729 }, { "epoch": 0.7, "learning_rate": 4.427122626992565e-06, "loss": 0.8489, "step": 13730 }, { "epoch": 0.7, "learning_rate": 4.425756481591644e-06, "loss": 0.8338, "step": 13731 }, { "epoch": 0.7, "learning_rate": 4.4243904871074405e-06, "loss": 0.966, "step": 13732 }, { "epoch": 0.7, "learning_rate": 4.4230246435769255e-06, "loss": 0.8979, "step": 13733 }, { "epoch": 0.7, "learning_rate": 4.421658951037091e-06, "loss": 0.8844, "step": 13734 }, { "epoch": 0.7, "learning_rate": 4.420293409524903e-06, "loss": 0.9708, "step": 13735 }, { "epoch": 0.7, "learning_rate": 4.418928019077334e-06, "loss": 0.8295, "step": 13736 }, { "epoch": 0.7, "learning_rate": 4.417562779731355e-06, "loss": 0.8285, "step": 13737 }, { "epoch": 0.7, "learning_rate": 4.416197691523922e-06, "loss": 0.9089, "step": 13738 }, { "epoch": 0.7, "learning_rate": 4.414832754491994e-06, "loss": 0.7541, "step": 13739 }, { "epoch": 0.7, "learning_rate": 4.413467968672526e-06, "loss": 0.9171, "step": 13740 }, { "epoch": 0.7, "learning_rate": 4.412103334102476e-06, "loss": 0.8572, "step": 13741 }, { "epoch": 0.7, "learning_rate": 4.410738850818778e-06, "loss": 0.9091, "step": 13742 }, { "epoch": 0.7, "learning_rate": 4.409374518858378e-06, "loss": 0.8832, "step": 13743 }, { "epoch": 0.7, "learning_rate": 4.40801033825822e-06, "loss": 0.8494, "step": 13744 }, { "epoch": 0.7, "learning_rate": 4.4066463090552245e-06, "loss": 0.8957, "step": 13745 }, { "epoch": 0.7, "learning_rate": 4.405282431286337e-06, "loss": 0.9549, "step": 13746 }, { "epoch": 0.7, "learning_rate": 4.403918704988473e-06, "loss": 0.9516, "step": 13747 }, { "epoch": 0.7, "learning_rate": 4.402555130198554e-06, "loss": 0.7887, "step": 13748 }, { "epoch": 0.7, "learning_rate": 4.401191706953505e-06, "loss": 0.8512, "step": 13749 }, { "epoch": 0.7, "learning_rate": 4.399828435290225e-06, "loss": 0.8559, "step": 13750 }, { "epoch": 0.7, "learning_rate": 4.398465315245641e-06, "loss": 0.9151, "step": 13751 }, { "epoch": 0.7, "learning_rate": 4.397102346856645e-06, "loss": 0.8414, "step": 13752 }, { "epoch": 0.7, "learning_rate": 4.395739530160141e-06, "loss": 0.9653, "step": 13753 }, { "epoch": 0.7, "learning_rate": 4.39437686519303e-06, "loss": 0.8488, "step": 13754 }, { "epoch": 0.7, "learning_rate": 4.393014351992199e-06, "loss": 0.7809, "step": 13755 }, { "epoch": 0.7, "learning_rate": 4.3916519905945396e-06, "loss": 0.8127, "step": 13756 }, { "epoch": 0.7, "learning_rate": 4.390289781036935e-06, "loss": 0.8004, "step": 13757 }, { "epoch": 0.7, "learning_rate": 4.38892772335627e-06, "loss": 0.9223, "step": 13758 }, { "epoch": 0.7, "learning_rate": 4.387565817589413e-06, "loss": 0.8836, "step": 13759 }, { "epoch": 0.7, "learning_rate": 4.386204063773242e-06, "loss": 0.9598, "step": 13760 }, { "epoch": 0.7, "learning_rate": 4.384842461944625e-06, "loss": 0.8426, "step": 13761 }, { "epoch": 0.7, "learning_rate": 4.383481012140418e-06, "loss": 0.9396, "step": 13762 }, { "epoch": 0.7, "learning_rate": 4.382119714397495e-06, "loss": 0.8359, "step": 13763 }, { "epoch": 0.7, "learning_rate": 4.380758568752699e-06, "loss": 0.9005, "step": 13764 }, { "epoch": 0.7, "learning_rate": 4.379397575242886e-06, "loss": 0.8997, "step": 13765 }, { "epoch": 0.7, "learning_rate": 4.3780367339049094e-06, "loss": 0.9172, "step": 13766 }, { "epoch": 0.7, "learning_rate": 4.376676044775601e-06, "loss": 0.9044, "step": 13767 }, { "epoch": 0.7, "learning_rate": 4.375315507891808e-06, "loss": 0.8541, "step": 13768 }, { "epoch": 0.7, "learning_rate": 4.373955123290362e-06, "loss": 0.8786, "step": 13769 }, { "epoch": 0.7, "learning_rate": 4.3725948910081e-06, "loss": 0.8, "step": 13770 }, { "epoch": 0.7, "learning_rate": 4.371234811081839e-06, "loss": 0.8984, "step": 13771 }, { "epoch": 0.7, "learning_rate": 4.369874883548408e-06, "loss": 0.8464, "step": 13772 }, { "epoch": 0.7, "learning_rate": 4.368515108444629e-06, "loss": 0.8704, "step": 13773 }, { "epoch": 0.7, "learning_rate": 4.3671554858073025e-06, "loss": 0.9359, "step": 13774 }, { "epoch": 0.7, "learning_rate": 4.3657960156732585e-06, "loss": 0.8477, "step": 13775 }, { "epoch": 0.7, "learning_rate": 4.364436698079287e-06, "loss": 0.9225, "step": 13776 }, { "epoch": 0.7, "learning_rate": 4.363077533062198e-06, "loss": 0.8932, "step": 13777 }, { "epoch": 0.7, "learning_rate": 4.3617185206587896e-06, "loss": 0.8418, "step": 13778 }, { "epoch": 0.7, "learning_rate": 4.3603596609058505e-06, "loss": 0.8599, "step": 13779 }, { "epoch": 0.7, "learning_rate": 4.359000953840173e-06, "loss": 0.9003, "step": 13780 }, { "epoch": 0.7, "learning_rate": 4.357642399498545e-06, "loss": 0.9216, "step": 13781 }, { "epoch": 0.7, "learning_rate": 4.356283997917749e-06, "loss": 1.0129, "step": 13782 }, { "epoch": 0.7, "learning_rate": 4.354925749134554e-06, "loss": 0.9015, "step": 13783 }, { "epoch": 0.7, "learning_rate": 4.3535676531857395e-06, "loss": 0.8991, "step": 13784 }, { "epoch": 0.7, "learning_rate": 4.3522097101080775e-06, "loss": 0.9274, "step": 13785 }, { "epoch": 0.7, "learning_rate": 4.350851919938322e-06, "loss": 0.851, "step": 13786 }, { "epoch": 0.7, "learning_rate": 4.349494282713246e-06, "loss": 0.9724, "step": 13787 }, { "epoch": 0.7, "learning_rate": 4.348136798469597e-06, "loss": 0.825, "step": 13788 }, { "epoch": 0.7, "learning_rate": 4.346779467244133e-06, "loss": 0.8424, "step": 13789 }, { "epoch": 0.7, "learning_rate": 4.345422289073603e-06, "loss": 0.9206, "step": 13790 }, { "epoch": 0.7, "learning_rate": 4.344065263994746e-06, "loss": 0.8244, "step": 13791 }, { "epoch": 0.7, "learning_rate": 4.342708392044302e-06, "loss": 0.9155, "step": 13792 }, { "epoch": 0.7, "learning_rate": 4.3413516732590125e-06, "loss": 0.8849, "step": 13793 }, { "epoch": 0.7, "learning_rate": 4.339995107675607e-06, "loss": 0.896, "step": 13794 }, { "epoch": 0.7, "learning_rate": 4.33863869533081e-06, "loss": 0.9326, "step": 13795 }, { "epoch": 0.7, "learning_rate": 4.337282436261347e-06, "loss": 0.8522, "step": 13796 }, { "epoch": 0.7, "learning_rate": 4.335926330503941e-06, "loss": 0.8457, "step": 13797 }, { "epoch": 0.7, "learning_rate": 4.334570378095296e-06, "loss": 0.834, "step": 13798 }, { "epoch": 0.7, "learning_rate": 4.333214579072138e-06, "loss": 0.8525, "step": 13799 }, { "epoch": 0.7, "learning_rate": 4.331858933471163e-06, "loss": 0.8692, "step": 13800 }, { "epoch": 0.7, "learning_rate": 4.330503441329076e-06, "loss": 0.839, "step": 13801 }, { "epoch": 0.7, "learning_rate": 4.32914810268258e-06, "loss": 0.899, "step": 13802 }, { "epoch": 0.7, "learning_rate": 4.327792917568363e-06, "loss": 0.8394, "step": 13803 }, { "epoch": 0.7, "learning_rate": 4.326437886023118e-06, "loss": 0.8761, "step": 13804 }, { "epoch": 0.7, "learning_rate": 4.325083008083531e-06, "loss": 0.8877, "step": 13805 }, { "epoch": 0.7, "learning_rate": 4.323728283786287e-06, "loss": 0.8905, "step": 13806 }, { "epoch": 0.7, "learning_rate": 4.3223737131680566e-06, "loss": 0.8003, "step": 13807 }, { "epoch": 0.7, "learning_rate": 4.321019296265518e-06, "loss": 0.9885, "step": 13808 }, { "epoch": 0.7, "learning_rate": 4.319665033115345e-06, "loss": 0.8291, "step": 13809 }, { "epoch": 0.7, "learning_rate": 4.3183109237541885e-06, "loss": 0.8913, "step": 13810 }, { "epoch": 0.7, "learning_rate": 4.316956968218728e-06, "loss": 0.9312, "step": 13811 }, { "epoch": 0.7, "learning_rate": 4.315603166545606e-06, "loss": 0.8701, "step": 13812 }, { "epoch": 0.7, "learning_rate": 4.3142495187714805e-06, "loss": 0.8982, "step": 13813 }, { "epoch": 0.7, "learning_rate": 4.312896024933004e-06, "loss": 0.8575, "step": 13814 }, { "epoch": 0.7, "learning_rate": 4.311542685066814e-06, "loss": 0.9358, "step": 13815 }, { "epoch": 0.7, "learning_rate": 4.310189499209554e-06, "loss": 0.9649, "step": 13816 }, { "epoch": 0.7, "learning_rate": 4.30883646739786e-06, "loss": 0.8871, "step": 13817 }, { "epoch": 0.7, "learning_rate": 4.307483589668363e-06, "loss": 0.9078, "step": 13818 }, { "epoch": 0.7, "learning_rate": 4.306130866057695e-06, "loss": 0.9214, "step": 13819 }, { "epoch": 0.7, "learning_rate": 4.304778296602472e-06, "loss": 0.9056, "step": 13820 }, { "epoch": 0.7, "learning_rate": 4.303425881339318e-06, "loss": 0.7362, "step": 13821 }, { "epoch": 0.7, "learning_rate": 4.3020736203048475e-06, "loss": 0.8284, "step": 13822 }, { "epoch": 0.7, "learning_rate": 4.300721513535676e-06, "loss": 0.8596, "step": 13823 }, { "epoch": 0.7, "learning_rate": 4.299369561068402e-06, "loss": 0.8513, "step": 13824 }, { "epoch": 0.7, "learning_rate": 4.2980177629396315e-06, "loss": 0.9261, "step": 13825 }, { "epoch": 0.7, "learning_rate": 4.296666119185967e-06, "loss": 0.8317, "step": 13826 }, { "epoch": 0.7, "learning_rate": 4.295314629843993e-06, "loss": 0.9109, "step": 13827 }, { "epoch": 0.7, "learning_rate": 4.293963294950313e-06, "loss": 0.8986, "step": 13828 }, { "epoch": 0.7, "learning_rate": 4.292612114541502e-06, "loss": 0.9415, "step": 13829 }, { "epoch": 0.7, "learning_rate": 4.291261088654146e-06, "loss": 0.8356, "step": 13830 }, { "epoch": 0.7, "learning_rate": 4.289910217324827e-06, "loss": 0.9451, "step": 13831 }, { "epoch": 0.7, "learning_rate": 4.288559500590109e-06, "loss": 0.8405, "step": 13832 }, { "epoch": 0.7, "learning_rate": 4.2872089384865665e-06, "loss": 0.8079, "step": 13833 }, { "epoch": 0.7, "learning_rate": 4.285858531050765e-06, "loss": 0.8333, "step": 13834 }, { "epoch": 0.7, "learning_rate": 4.284508278319267e-06, "loss": 0.8138, "step": 13835 }, { "epoch": 0.7, "learning_rate": 4.283158180328623e-06, "loss": 0.8348, "step": 13836 }, { "epoch": 0.7, "learning_rate": 4.281808237115389e-06, "loss": 0.9066, "step": 13837 }, { "epoch": 0.7, "learning_rate": 4.280458448716117e-06, "loss": 0.9315, "step": 13838 }, { "epoch": 0.7, "learning_rate": 4.2791088151673395e-06, "loss": 0.9289, "step": 13839 }, { "epoch": 0.7, "learning_rate": 4.277759336505612e-06, "loss": 0.9211, "step": 13840 }, { "epoch": 0.7, "learning_rate": 4.27641001276746e-06, "loss": 0.849, "step": 13841 }, { "epoch": 0.7, "learning_rate": 4.275060843989416e-06, "loss": 0.8721, "step": 13842 }, { "epoch": 0.7, "learning_rate": 4.273711830208013e-06, "loss": 0.897, "step": 13843 }, { "epoch": 0.7, "learning_rate": 4.272362971459767e-06, "loss": 0.8939, "step": 13844 }, { "epoch": 0.7, "learning_rate": 4.271014267781199e-06, "loss": 0.9258, "step": 13845 }, { "epoch": 0.7, "learning_rate": 4.269665719208824e-06, "loss": 0.8331, "step": 13846 }, { "epoch": 0.7, "learning_rate": 4.268317325779158e-06, "loss": 0.8571, "step": 13847 }, { "epoch": 0.7, "learning_rate": 4.266969087528697e-06, "loss": 0.8917, "step": 13848 }, { "epoch": 0.7, "learning_rate": 4.26562100449395e-06, "loss": 0.8929, "step": 13849 }, { "epoch": 0.7, "learning_rate": 4.264273076711416e-06, "loss": 0.9156, "step": 13850 }, { "epoch": 0.7, "learning_rate": 4.26292530421758e-06, "loss": 0.9212, "step": 13851 }, { "epoch": 0.7, "learning_rate": 4.261577687048944e-06, "loss": 0.84, "step": 13852 }, { "epoch": 0.7, "learning_rate": 4.260230225241984e-06, "loss": 0.9406, "step": 13853 }, { "epoch": 0.7, "learning_rate": 4.258882918833184e-06, "loss": 0.8454, "step": 13854 }, { "epoch": 0.7, "learning_rate": 4.257535767859024e-06, "loss": 0.9482, "step": 13855 }, { "epoch": 0.7, "learning_rate": 4.256188772355969e-06, "loss": 0.8997, "step": 13856 }, { "epoch": 0.7, "learning_rate": 4.254841932360494e-06, "loss": 0.9369, "step": 13857 }, { "epoch": 0.7, "learning_rate": 4.25349524790906e-06, "loss": 0.8932, "step": 13858 }, { "epoch": 0.7, "learning_rate": 4.252148719038133e-06, "loss": 0.926, "step": 13859 }, { "epoch": 0.7, "learning_rate": 4.25080234578416e-06, "loss": 0.8315, "step": 13860 }, { "epoch": 0.7, "learning_rate": 4.249456128183597e-06, "loss": 0.8094, "step": 13861 }, { "epoch": 0.7, "learning_rate": 4.248110066272894e-06, "loss": 0.8329, "step": 13862 }, { "epoch": 0.7, "learning_rate": 4.246764160088485e-06, "loss": 0.8901, "step": 13863 }, { "epoch": 0.7, "learning_rate": 4.245418409666823e-06, "loss": 0.894, "step": 13864 }, { "epoch": 0.7, "learning_rate": 4.24407281504433e-06, "loss": 0.8946, "step": 13865 }, { "epoch": 0.7, "learning_rate": 4.242727376257443e-06, "loss": 0.8221, "step": 13866 }, { "epoch": 0.7, "learning_rate": 4.241382093342591e-06, "loss": 0.7972, "step": 13867 }, { "epoch": 0.7, "learning_rate": 4.240036966336187e-06, "loss": 0.9625, "step": 13868 }, { "epoch": 0.7, "learning_rate": 4.238691995274655e-06, "loss": 0.8853, "step": 13869 }, { "epoch": 0.7, "learning_rate": 4.237347180194407e-06, "loss": 0.844, "step": 13870 }, { "epoch": 0.7, "learning_rate": 4.2360025211318575e-06, "loss": 0.9322, "step": 13871 }, { "epoch": 0.7, "learning_rate": 4.2346580181234045e-06, "loss": 0.9068, "step": 13872 }, { "epoch": 0.7, "learning_rate": 4.233313671205451e-06, "loss": 0.8579, "step": 13873 }, { "epoch": 0.7, "learning_rate": 4.231969480414397e-06, "loss": 0.8733, "step": 13874 }, { "epoch": 0.7, "learning_rate": 4.230625445786627e-06, "loss": 0.8207, "step": 13875 }, { "epoch": 0.7, "learning_rate": 4.2292815673585415e-06, "loss": 0.9087, "step": 13876 }, { "epoch": 0.7, "learning_rate": 4.227937845166515e-06, "loss": 0.8646, "step": 13877 }, { "epoch": 0.71, "learning_rate": 4.226594279246929e-06, "loss": 0.9331, "step": 13878 }, { "epoch": 0.71, "learning_rate": 4.225250869636166e-06, "loss": 0.8308, "step": 13879 }, { "epoch": 0.71, "learning_rate": 4.223907616370587e-06, "loss": 0.9272, "step": 13880 }, { "epoch": 0.71, "learning_rate": 4.222564519486564e-06, "loss": 0.8209, "step": 13881 }, { "epoch": 0.71, "learning_rate": 4.22122157902046e-06, "loss": 0.8165, "step": 13882 }, { "epoch": 0.71, "learning_rate": 4.2198787950086375e-06, "loss": 0.9837, "step": 13883 }, { "epoch": 0.71, "learning_rate": 4.218536167487444e-06, "loss": 0.907, "step": 13884 }, { "epoch": 0.71, "learning_rate": 4.2171936964932316e-06, "loss": 0.8327, "step": 13885 }, { "epoch": 0.71, "learning_rate": 4.2158513820623516e-06, "loss": 0.8836, "step": 13886 }, { "epoch": 0.71, "learning_rate": 4.214509224231134e-06, "loss": 0.9299, "step": 13887 }, { "epoch": 0.71, "learning_rate": 4.213167223035931e-06, "loss": 0.9132, "step": 13888 }, { "epoch": 0.71, "learning_rate": 4.211825378513066e-06, "loss": 0.8848, "step": 13889 }, { "epoch": 0.71, "learning_rate": 4.21048369069887e-06, "loss": 0.9388, "step": 13890 }, { "epoch": 0.71, "learning_rate": 4.209142159629672e-06, "loss": 0.9175, "step": 13891 }, { "epoch": 0.71, "learning_rate": 4.207800785341782e-06, "loss": 0.8237, "step": 13892 }, { "epoch": 0.71, "learning_rate": 4.2064595678715315e-06, "loss": 0.8125, "step": 13893 }, { "epoch": 0.71, "learning_rate": 4.2051185072552204e-06, "loss": 0.9696, "step": 13894 }, { "epoch": 0.71, "learning_rate": 4.20377760352916e-06, "loss": 0.9013, "step": 13895 }, { "epoch": 0.71, "learning_rate": 4.202436856729658e-06, "loss": 0.9038, "step": 13896 }, { "epoch": 0.71, "learning_rate": 4.201096266893006e-06, "loss": 0.7702, "step": 13897 }, { "epoch": 0.71, "learning_rate": 4.199755834055503e-06, "loss": 0.8839, "step": 13898 }, { "epoch": 0.71, "learning_rate": 4.19841555825344e-06, "loss": 0.9729, "step": 13899 }, { "epoch": 0.71, "learning_rate": 4.197075439523108e-06, "loss": 0.9963, "step": 13900 }, { "epoch": 0.71, "learning_rate": 4.19573547790078e-06, "loss": 0.9451, "step": 13901 }, { "epoch": 0.71, "learning_rate": 4.194395673422739e-06, "loss": 0.9034, "step": 13902 }, { "epoch": 0.71, "learning_rate": 4.193056026125262e-06, "loss": 0.93, "step": 13903 }, { "epoch": 0.71, "learning_rate": 4.191716536044606e-06, "loss": 0.9001, "step": 13904 }, { "epoch": 0.71, "learning_rate": 4.190377203217056e-06, "loss": 0.8684, "step": 13905 }, { "epoch": 0.71, "learning_rate": 4.1890380276788554e-06, "loss": 0.8473, "step": 13906 }, { "epoch": 0.71, "learning_rate": 4.1876990094662695e-06, "loss": 0.8621, "step": 13907 }, { "epoch": 0.71, "learning_rate": 4.1863601486155525e-06, "loss": 0.8969, "step": 13908 }, { "epoch": 0.71, "learning_rate": 4.185021445162946e-06, "loss": 0.9569, "step": 13909 }, { "epoch": 0.71, "learning_rate": 4.183682899144696e-06, "loss": 0.8244, "step": 13910 }, { "epoch": 0.71, "learning_rate": 4.182344510597045e-06, "loss": 0.7988, "step": 13911 }, { "epoch": 0.71, "learning_rate": 4.18100627955623e-06, "loss": 0.8652, "step": 13912 }, { "epoch": 0.71, "learning_rate": 4.179668206058475e-06, "loss": 0.8543, "step": 13913 }, { "epoch": 0.71, "learning_rate": 4.178330290140011e-06, "loss": 0.8471, "step": 13914 }, { "epoch": 0.71, "learning_rate": 4.176992531837065e-06, "loss": 0.8856, "step": 13915 }, { "epoch": 0.71, "learning_rate": 4.175654931185843e-06, "loss": 0.9741, "step": 13916 }, { "epoch": 0.71, "learning_rate": 4.174317488222575e-06, "loss": 0.9579, "step": 13917 }, { "epoch": 0.71, "learning_rate": 4.172980202983459e-06, "loss": 0.9308, "step": 13918 }, { "epoch": 0.71, "learning_rate": 4.171643075504704e-06, "loss": 0.8953, "step": 13919 }, { "epoch": 0.71, "learning_rate": 4.170306105822516e-06, "loss": 0.8318, "step": 13920 }, { "epoch": 0.71, "learning_rate": 4.168969293973084e-06, "loss": 0.8513, "step": 13921 }, { "epoch": 0.71, "learning_rate": 4.167632639992605e-06, "loss": 0.8589, "step": 13922 }, { "epoch": 0.71, "learning_rate": 4.166296143917267e-06, "loss": 0.897, "step": 13923 }, { "epoch": 0.71, "learning_rate": 4.164959805783257e-06, "loss": 0.9015, "step": 13924 }, { "epoch": 0.71, "learning_rate": 4.16362362562675e-06, "loss": 0.8542, "step": 13925 }, { "epoch": 0.71, "learning_rate": 4.1622876034839234e-06, "loss": 0.8654, "step": 13926 }, { "epoch": 0.71, "learning_rate": 4.160951739390952e-06, "loss": 0.9748, "step": 13927 }, { "epoch": 0.71, "learning_rate": 4.159616033383994e-06, "loss": 0.8677, "step": 13928 }, { "epoch": 0.71, "learning_rate": 4.158280485499225e-06, "loss": 0.9068, "step": 13929 }, { "epoch": 0.71, "learning_rate": 4.156945095772793e-06, "loss": 0.9443, "step": 13930 }, { "epoch": 0.71, "learning_rate": 4.155609864240857e-06, "loss": 0.8008, "step": 13931 }, { "epoch": 0.71, "learning_rate": 4.154274790939568e-06, "loss": 0.9239, "step": 13932 }, { "epoch": 0.71, "learning_rate": 4.152939875905068e-06, "loss": 0.817, "step": 13933 }, { "epoch": 0.71, "learning_rate": 4.151605119173498e-06, "loss": 0.9706, "step": 13934 }, { "epoch": 0.71, "learning_rate": 4.150270520780998e-06, "loss": 0.9468, "step": 13935 }, { "epoch": 0.71, "learning_rate": 4.148936080763703e-06, "loss": 0.8461, "step": 13936 }, { "epoch": 0.71, "learning_rate": 4.147601799157736e-06, "loss": 0.8962, "step": 13937 }, { "epoch": 0.71, "learning_rate": 4.146267675999224e-06, "loss": 0.8252, "step": 13938 }, { "epoch": 0.71, "learning_rate": 4.144933711324289e-06, "loss": 0.8136, "step": 13939 }, { "epoch": 0.71, "learning_rate": 4.143599905169039e-06, "loss": 0.8655, "step": 13940 }, { "epoch": 0.71, "learning_rate": 4.1422662575695975e-06, "loss": 0.8698, "step": 13941 }, { "epoch": 0.71, "learning_rate": 4.1409327685620605e-06, "loss": 0.9135, "step": 13942 }, { "epoch": 0.71, "learning_rate": 4.139599438182536e-06, "loss": 0.8416, "step": 13943 }, { "epoch": 0.71, "learning_rate": 4.138266266467124e-06, "loss": 0.8354, "step": 13944 }, { "epoch": 0.71, "learning_rate": 4.1369332534519145e-06, "loss": 0.91, "step": 13945 }, { "epoch": 0.71, "learning_rate": 4.135600399172999e-06, "loss": 0.9246, "step": 13946 }, { "epoch": 0.71, "learning_rate": 4.134267703666464e-06, "loss": 0.995, "step": 13947 }, { "epoch": 0.71, "learning_rate": 4.132935166968392e-06, "loss": 0.8919, "step": 13948 }, { "epoch": 0.71, "learning_rate": 4.131602789114857e-06, "loss": 0.9102, "step": 13949 }, { "epoch": 0.71, "learning_rate": 4.130270570141931e-06, "loss": 0.992, "step": 13950 }, { "epoch": 0.71, "learning_rate": 4.128938510085689e-06, "loss": 0.828, "step": 13951 }, { "epoch": 0.71, "learning_rate": 4.127606608982183e-06, "loss": 0.88, "step": 13952 }, { "epoch": 0.71, "learning_rate": 4.126274866867488e-06, "loss": 0.8177, "step": 13953 }, { "epoch": 0.71, "learning_rate": 4.124943283777648e-06, "loss": 0.8807, "step": 13954 }, { "epoch": 0.71, "learning_rate": 4.1236118597487175e-06, "loss": 0.805, "step": 13955 }, { "epoch": 0.71, "learning_rate": 4.122280594816747e-06, "loss": 0.9786, "step": 13956 }, { "epoch": 0.71, "learning_rate": 4.120949489017772e-06, "loss": 0.9465, "step": 13957 }, { "epoch": 0.71, "learning_rate": 4.119618542387839e-06, "loss": 0.9469, "step": 13958 }, { "epoch": 0.71, "learning_rate": 4.118287754962974e-06, "loss": 0.9058, "step": 13959 }, { "epoch": 0.71, "learning_rate": 4.116957126779212e-06, "loss": 0.8477, "step": 13960 }, { "epoch": 0.71, "learning_rate": 4.115626657872579e-06, "loss": 0.858, "step": 13961 }, { "epoch": 0.71, "learning_rate": 4.114296348279091e-06, "loss": 0.9023, "step": 13962 }, { "epoch": 0.71, "learning_rate": 4.112966198034767e-06, "loss": 0.9565, "step": 13963 }, { "epoch": 0.71, "learning_rate": 4.11163620717562e-06, "loss": 0.8053, "step": 13964 }, { "epoch": 0.71, "learning_rate": 4.110306375737662e-06, "loss": 0.9564, "step": 13965 }, { "epoch": 0.71, "learning_rate": 4.108976703756888e-06, "loss": 0.8689, "step": 13966 }, { "epoch": 0.71, "learning_rate": 4.107647191269304e-06, "loss": 0.8932, "step": 13967 }, { "epoch": 0.71, "learning_rate": 4.106317838310906e-06, "loss": 0.8774, "step": 13968 }, { "epoch": 0.71, "learning_rate": 4.104988644917674e-06, "loss": 0.9908, "step": 13969 }, { "epoch": 0.71, "learning_rate": 4.103659611125611e-06, "loss": 0.7781, "step": 13970 }, { "epoch": 0.71, "learning_rate": 4.102330736970686e-06, "loss": 0.8497, "step": 13971 }, { "epoch": 0.71, "learning_rate": 4.101002022488884e-06, "loss": 0.9597, "step": 13972 }, { "epoch": 0.71, "learning_rate": 4.099673467716178e-06, "loss": 0.9231, "step": 13973 }, { "epoch": 0.71, "learning_rate": 4.098345072688532e-06, "loss": 0.8813, "step": 13974 }, { "epoch": 0.71, "learning_rate": 4.097016837441914e-06, "loss": 0.8677, "step": 13975 }, { "epoch": 0.71, "learning_rate": 4.0956887620122855e-06, "loss": 0.9048, "step": 13976 }, { "epoch": 0.71, "learning_rate": 4.094360846435605e-06, "loss": 0.8857, "step": 13977 }, { "epoch": 0.71, "learning_rate": 4.093033090747819e-06, "loss": 0.9223, "step": 13978 }, { "epoch": 0.71, "learning_rate": 4.0917054949848765e-06, "loss": 0.8788, "step": 13979 }, { "epoch": 0.71, "learning_rate": 4.090378059182726e-06, "loss": 0.9087, "step": 13980 }, { "epoch": 0.71, "learning_rate": 4.089050783377295e-06, "loss": 0.8327, "step": 13981 }, { "epoch": 0.71, "learning_rate": 4.087723667604532e-06, "loss": 0.8616, "step": 13982 }, { "epoch": 0.71, "learning_rate": 4.086396711900358e-06, "loss": 0.9125, "step": 13983 }, { "epoch": 0.71, "learning_rate": 4.0850699163007e-06, "loss": 0.8093, "step": 13984 }, { "epoch": 0.71, "learning_rate": 4.0837432808414856e-06, "loss": 0.9236, "step": 13985 }, { "epoch": 0.71, "learning_rate": 4.082416805558623e-06, "loss": 0.8741, "step": 13986 }, { "epoch": 0.71, "learning_rate": 4.08109049048803e-06, "loss": 0.8497, "step": 13987 }, { "epoch": 0.71, "learning_rate": 4.079764335665615e-06, "loss": 0.8378, "step": 13988 }, { "epoch": 0.71, "learning_rate": 4.078438341127285e-06, "loss": 0.8882, "step": 13989 }, { "epoch": 0.71, "learning_rate": 4.077112506908935e-06, "loss": 0.991, "step": 13990 }, { "epoch": 0.71, "learning_rate": 4.07578683304646e-06, "loss": 0.8334, "step": 13991 }, { "epoch": 0.71, "learning_rate": 4.0744613195757585e-06, "loss": 0.8754, "step": 13992 }, { "epoch": 0.71, "learning_rate": 4.073135966532705e-06, "loss": 0.8898, "step": 13993 }, { "epoch": 0.71, "learning_rate": 4.071810773953197e-06, "loss": 0.7697, "step": 13994 }, { "epoch": 0.71, "learning_rate": 4.070485741873101e-06, "loss": 0.8363, "step": 13995 }, { "epoch": 0.71, "learning_rate": 4.069160870328296e-06, "loss": 0.8156, "step": 13996 }, { "epoch": 0.71, "learning_rate": 4.067836159354654e-06, "loss": 0.7702, "step": 13997 }, { "epoch": 0.71, "learning_rate": 4.066511608988033e-06, "loss": 0.8145, "step": 13998 }, { "epoch": 0.71, "learning_rate": 4.065187219264297e-06, "loss": 0.9044, "step": 13999 }, { "epoch": 0.71, "learning_rate": 4.063862990219303e-06, "loss": 0.8745, "step": 14000 }, { "epoch": 0.71, "learning_rate": 4.0625389218889066e-06, "loss": 0.9282, "step": 14001 }, { "epoch": 0.71, "learning_rate": 4.061215014308948e-06, "loss": 0.8569, "step": 14002 }, { "epoch": 0.71, "learning_rate": 4.059891267515274e-06, "loss": 0.8973, "step": 14003 }, { "epoch": 0.71, "learning_rate": 4.05856768154373e-06, "loss": 0.8669, "step": 14004 }, { "epoch": 0.71, "learning_rate": 4.057244256430136e-06, "loss": 0.9789, "step": 14005 }, { "epoch": 0.71, "learning_rate": 4.055920992210338e-06, "loss": 0.945, "step": 14006 }, { "epoch": 0.71, "learning_rate": 4.054597888920153e-06, "loss": 0.8677, "step": 14007 }, { "epoch": 0.71, "learning_rate": 4.053274946595403e-06, "loss": 0.9077, "step": 14008 }, { "epoch": 0.71, "learning_rate": 4.0519521652719115e-06, "loss": 0.831, "step": 14009 }, { "epoch": 0.71, "learning_rate": 4.050629544985484e-06, "loss": 0.9122, "step": 14010 }, { "epoch": 0.71, "learning_rate": 4.0493070857719305e-06, "loss": 0.8548, "step": 14011 }, { "epoch": 0.71, "learning_rate": 4.047984787667059e-06, "loss": 0.9242, "step": 14012 }, { "epoch": 0.71, "learning_rate": 4.046662650706667e-06, "loss": 0.8065, "step": 14013 }, { "epoch": 0.71, "learning_rate": 4.045340674926549e-06, "loss": 0.9258, "step": 14014 }, { "epoch": 0.71, "learning_rate": 4.044018860362496e-06, "loss": 0.9067, "step": 14015 }, { "epoch": 0.71, "learning_rate": 4.0426972070503e-06, "loss": 0.929, "step": 14016 }, { "epoch": 0.71, "learning_rate": 4.041375715025731e-06, "loss": 0.9683, "step": 14017 }, { "epoch": 0.71, "learning_rate": 4.040054384324582e-06, "loss": 0.8829, "step": 14018 }, { "epoch": 0.71, "learning_rate": 4.038733214982618e-06, "loss": 0.8179, "step": 14019 }, { "epoch": 0.71, "learning_rate": 4.037412207035608e-06, "loss": 0.8137, "step": 14020 }, { "epoch": 0.71, "learning_rate": 4.036091360519323e-06, "loss": 0.7917, "step": 14021 }, { "epoch": 0.71, "learning_rate": 4.034770675469515e-06, "loss": 0.8648, "step": 14022 }, { "epoch": 0.71, "learning_rate": 4.033450151921945e-06, "loss": 0.8702, "step": 14023 }, { "epoch": 0.71, "learning_rate": 4.032129789912365e-06, "loss": 0.8226, "step": 14024 }, { "epoch": 0.71, "learning_rate": 4.030809589476524e-06, "loss": 0.8192, "step": 14025 }, { "epoch": 0.71, "learning_rate": 4.029489550650159e-06, "loss": 0.9439, "step": 14026 }, { "epoch": 0.71, "learning_rate": 4.028169673469013e-06, "loss": 0.8984, "step": 14027 }, { "epoch": 0.71, "learning_rate": 4.02684995796882e-06, "loss": 0.8075, "step": 14028 }, { "epoch": 0.71, "learning_rate": 4.025530404185308e-06, "loss": 0.8423, "step": 14029 }, { "epoch": 0.71, "learning_rate": 4.024211012154209e-06, "loss": 0.7994, "step": 14030 }, { "epoch": 0.71, "learning_rate": 4.0228917819112344e-06, "loss": 0.9924, "step": 14031 }, { "epoch": 0.71, "learning_rate": 4.021572713492107e-06, "loss": 0.8636, "step": 14032 }, { "epoch": 0.71, "learning_rate": 4.020253806932541e-06, "loss": 0.755, "step": 14033 }, { "epoch": 0.71, "learning_rate": 4.018935062268233e-06, "loss": 0.8084, "step": 14034 }, { "epoch": 0.71, "learning_rate": 4.017616479534903e-06, "loss": 0.8947, "step": 14035 }, { "epoch": 0.71, "learning_rate": 4.016298058768238e-06, "loss": 0.8749, "step": 14036 }, { "epoch": 0.71, "learning_rate": 4.014979800003937e-06, "loss": 0.8178, "step": 14037 }, { "epoch": 0.71, "learning_rate": 4.013661703277694e-06, "loss": 0.9568, "step": 14038 }, { "epoch": 0.71, "learning_rate": 4.0123437686251875e-06, "loss": 0.8294, "step": 14039 }, { "epoch": 0.71, "learning_rate": 4.011025996082104e-06, "loss": 0.9292, "step": 14040 }, { "epoch": 0.71, "learning_rate": 4.009708385684119e-06, "loss": 0.9715, "step": 14041 }, { "epoch": 0.71, "learning_rate": 4.008390937466911e-06, "loss": 0.8998, "step": 14042 }, { "epoch": 0.71, "learning_rate": 4.007073651466139e-06, "loss": 0.897, "step": 14043 }, { "epoch": 0.71, "learning_rate": 4.005756527717473e-06, "loss": 0.7547, "step": 14044 }, { "epoch": 0.71, "learning_rate": 4.004439566256574e-06, "loss": 0.8984, "step": 14045 }, { "epoch": 0.71, "learning_rate": 4.0031227671190885e-06, "loss": 0.8541, "step": 14046 }, { "epoch": 0.71, "learning_rate": 4.00180613034068e-06, "loss": 0.8588, "step": 14047 }, { "epoch": 0.71, "learning_rate": 4.0004896559569875e-06, "loss": 0.9008, "step": 14048 }, { "epoch": 0.71, "learning_rate": 3.999173344003653e-06, "loss": 0.7702, "step": 14049 }, { "epoch": 0.71, "learning_rate": 3.997857194516319e-06, "loss": 0.8963, "step": 14050 }, { "epoch": 0.71, "learning_rate": 3.9965412075306135e-06, "loss": 0.894, "step": 14051 }, { "epoch": 0.71, "learning_rate": 3.995225383082168e-06, "loss": 0.8528, "step": 14052 }, { "epoch": 0.71, "learning_rate": 3.993909721206605e-06, "loss": 0.929, "step": 14053 }, { "epoch": 0.71, "learning_rate": 3.992594221939551e-06, "loss": 0.9626, "step": 14054 }, { "epoch": 0.71, "learning_rate": 3.991278885316613e-06, "loss": 0.9519, "step": 14055 }, { "epoch": 0.71, "learning_rate": 3.989963711373406e-06, "loss": 0.9066, "step": 14056 }, { "epoch": 0.71, "learning_rate": 3.988648700145541e-06, "loss": 0.9138, "step": 14057 }, { "epoch": 0.71, "learning_rate": 3.987333851668611e-06, "loss": 0.876, "step": 14058 }, { "epoch": 0.71, "learning_rate": 3.986019165978225e-06, "loss": 0.9366, "step": 14059 }, { "epoch": 0.71, "learning_rate": 3.9847046431099704e-06, "loss": 0.789, "step": 14060 }, { "epoch": 0.71, "learning_rate": 3.983390283099436e-06, "loss": 0.9153, "step": 14061 }, { "epoch": 0.71, "learning_rate": 3.982076085982214e-06, "loss": 0.892, "step": 14062 }, { "epoch": 0.71, "learning_rate": 3.980762051793875e-06, "loss": 0.8423, "step": 14063 }, { "epoch": 0.71, "learning_rate": 3.979448180569999e-06, "loss": 0.8326, "step": 14064 }, { "epoch": 0.71, "learning_rate": 3.9781344723461576e-06, "loss": 0.8245, "step": 14065 }, { "epoch": 0.71, "learning_rate": 3.976820927157924e-06, "loss": 0.8195, "step": 14066 }, { "epoch": 0.71, "learning_rate": 3.975507545040851e-06, "loss": 0.9031, "step": 14067 }, { "epoch": 0.71, "learning_rate": 3.974194326030502e-06, "loss": 0.8697, "step": 14068 }, { "epoch": 0.71, "learning_rate": 3.972881270162433e-06, "loss": 0.8856, "step": 14069 }, { "epoch": 0.71, "learning_rate": 3.971568377472185e-06, "loss": 0.9129, "step": 14070 }, { "epoch": 0.71, "learning_rate": 3.970255647995317e-06, "loss": 0.8455, "step": 14071 }, { "epoch": 0.71, "learning_rate": 3.968943081767358e-06, "loss": 0.9155, "step": 14072 }, { "epoch": 0.71, "learning_rate": 3.967630678823848e-06, "loss": 0.8737, "step": 14073 }, { "epoch": 0.71, "learning_rate": 3.966318439200324e-06, "loss": 0.9442, "step": 14074 }, { "epoch": 0.72, "learning_rate": 3.965006362932306e-06, "loss": 0.9211, "step": 14075 }, { "epoch": 0.72, "learning_rate": 3.963694450055319e-06, "loss": 0.9285, "step": 14076 }, { "epoch": 0.72, "learning_rate": 3.962382700604882e-06, "loss": 0.8414, "step": 14077 }, { "epoch": 0.72, "learning_rate": 3.961071114616515e-06, "loss": 0.9526, "step": 14078 }, { "epoch": 0.72, "learning_rate": 3.959759692125717e-06, "loss": 0.8823, "step": 14079 }, { "epoch": 0.72, "learning_rate": 3.958448433168002e-06, "loss": 0.8969, "step": 14080 }, { "epoch": 0.72, "learning_rate": 3.957137337778869e-06, "loss": 0.9157, "step": 14081 }, { "epoch": 0.72, "learning_rate": 3.955826405993807e-06, "loss": 0.8314, "step": 14082 }, { "epoch": 0.72, "learning_rate": 3.954515637848322e-06, "loss": 0.8874, "step": 14083 }, { "epoch": 0.72, "learning_rate": 3.953205033377891e-06, "loss": 0.8982, "step": 14084 }, { "epoch": 0.72, "learning_rate": 3.9518945926179996e-06, "loss": 0.909, "step": 14085 }, { "epoch": 0.72, "learning_rate": 3.950584315604131e-06, "loss": 0.8474, "step": 14086 }, { "epoch": 0.72, "learning_rate": 3.949274202371753e-06, "loss": 0.8909, "step": 14087 }, { "epoch": 0.72, "learning_rate": 3.947964252956339e-06, "loss": 0.8925, "step": 14088 }, { "epoch": 0.72, "learning_rate": 3.946654467393353e-06, "loss": 0.8746, "step": 14089 }, { "epoch": 0.72, "learning_rate": 3.945344845718261e-06, "loss": 0.8916, "step": 14090 }, { "epoch": 0.72, "learning_rate": 3.944035387966512e-06, "loss": 0.8977, "step": 14091 }, { "epoch": 0.72, "learning_rate": 3.942726094173561e-06, "loss": 0.8138, "step": 14092 }, { "epoch": 0.72, "learning_rate": 3.941416964374861e-06, "loss": 0.9149, "step": 14093 }, { "epoch": 0.72, "learning_rate": 3.940107998605845e-06, "loss": 0.8245, "step": 14094 }, { "epoch": 0.72, "learning_rate": 3.9387991969019625e-06, "loss": 0.9032, "step": 14095 }, { "epoch": 0.72, "learning_rate": 3.93749055929864e-06, "loss": 0.8618, "step": 14096 }, { "epoch": 0.72, "learning_rate": 3.936182085831312e-06, "loss": 0.8885, "step": 14097 }, { "epoch": 0.72, "learning_rate": 3.934873776535405e-06, "loss": 0.9403, "step": 14098 }, { "epoch": 0.72, "learning_rate": 3.933565631446331e-06, "loss": 0.8995, "step": 14099 }, { "epoch": 0.72, "learning_rate": 3.9322576505995205e-06, "loss": 0.9933, "step": 14100 }, { "epoch": 0.72, "learning_rate": 3.9309498340303755e-06, "loss": 0.8066, "step": 14101 }, { "epoch": 0.72, "learning_rate": 3.929642181774306e-06, "loss": 0.9223, "step": 14102 }, { "epoch": 0.72, "learning_rate": 3.92833469386672e-06, "loss": 0.7438, "step": 14103 }, { "epoch": 0.72, "learning_rate": 3.927027370343008e-06, "loss": 0.826, "step": 14104 }, { "epoch": 0.72, "learning_rate": 3.9257202112385695e-06, "loss": 0.8155, "step": 14105 }, { "epoch": 0.72, "learning_rate": 3.9244132165887935e-06, "loss": 0.8974, "step": 14106 }, { "epoch": 0.72, "learning_rate": 3.9231063864290695e-06, "loss": 0.8262, "step": 14107 }, { "epoch": 0.72, "learning_rate": 3.921799720794772e-06, "loss": 0.8757, "step": 14108 }, { "epoch": 0.72, "learning_rate": 3.92049321972128e-06, "loss": 0.8713, "step": 14109 }, { "epoch": 0.72, "learning_rate": 3.919186883243968e-06, "loss": 0.9911, "step": 14110 }, { "epoch": 0.72, "learning_rate": 3.917880711398196e-06, "loss": 0.8778, "step": 14111 }, { "epoch": 0.72, "learning_rate": 3.9165747042193405e-06, "loss": 0.8323, "step": 14112 }, { "epoch": 0.72, "learning_rate": 3.915268861742748e-06, "loss": 0.8049, "step": 14113 }, { "epoch": 0.72, "learning_rate": 3.913963184003778e-06, "loss": 0.9344, "step": 14114 }, { "epoch": 0.72, "learning_rate": 3.912657671037783e-06, "loss": 0.8773, "step": 14115 }, { "epoch": 0.72, "learning_rate": 3.911352322880102e-06, "loss": 0.9137, "step": 14116 }, { "epoch": 0.72, "learning_rate": 3.910047139566079e-06, "loss": 0.9319, "step": 14117 }, { "epoch": 0.72, "learning_rate": 3.90874212113105e-06, "loss": 0.8269, "step": 14118 }, { "epoch": 0.72, "learning_rate": 3.9074372676103514e-06, "loss": 0.9302, "step": 14119 }, { "epoch": 0.72, "learning_rate": 3.906132579039303e-06, "loss": 0.8944, "step": 14120 }, { "epoch": 0.72, "learning_rate": 3.904828055453231e-06, "loss": 0.8298, "step": 14121 }, { "epoch": 0.72, "learning_rate": 3.903523696887459e-06, "loss": 0.8043, "step": 14122 }, { "epoch": 0.72, "learning_rate": 3.90221950337729e-06, "loss": 0.7998, "step": 14123 }, { "epoch": 0.72, "learning_rate": 3.900915474958047e-06, "loss": 0.9005, "step": 14124 }, { "epoch": 0.72, "learning_rate": 3.899611611665025e-06, "loss": 0.8489, "step": 14125 }, { "epoch": 0.72, "learning_rate": 3.898307913533529e-06, "loss": 0.7762, "step": 14126 }, { "epoch": 0.72, "learning_rate": 3.897004380598857e-06, "loss": 0.9301, "step": 14127 }, { "epoch": 0.72, "learning_rate": 3.895701012896296e-06, "loss": 0.9206, "step": 14128 }, { "epoch": 0.72, "learning_rate": 3.894397810461135e-06, "loss": 0.8506, "step": 14129 }, { "epoch": 0.72, "learning_rate": 3.893094773328657e-06, "loss": 0.8772, "step": 14130 }, { "epoch": 0.72, "learning_rate": 3.891791901534144e-06, "loss": 0.8138, "step": 14131 }, { "epoch": 0.72, "learning_rate": 3.890489195112864e-06, "loss": 0.8539, "step": 14132 }, { "epoch": 0.72, "learning_rate": 3.889186654100089e-06, "loss": 0.8669, "step": 14133 }, { "epoch": 0.72, "learning_rate": 3.8878842785310876e-06, "loss": 0.8461, "step": 14134 }, { "epoch": 0.72, "learning_rate": 3.886582068441109e-06, "loss": 0.7515, "step": 14135 }, { "epoch": 0.72, "learning_rate": 3.8852800238654245e-06, "loss": 0.9399, "step": 14136 }, { "epoch": 0.72, "learning_rate": 3.883978144839273e-06, "loss": 0.871, "step": 14137 }, { "epoch": 0.72, "learning_rate": 3.882676431397907e-06, "loss": 0.8965, "step": 14138 }, { "epoch": 0.72, "learning_rate": 3.881374883576572e-06, "loss": 0.842, "step": 14139 }, { "epoch": 0.72, "learning_rate": 3.8800735014105e-06, "loss": 0.9017, "step": 14140 }, { "epoch": 0.72, "learning_rate": 3.878772284934925e-06, "loss": 0.9414, "step": 14141 }, { "epoch": 0.72, "learning_rate": 3.877471234185079e-06, "loss": 0.7885, "step": 14142 }, { "epoch": 0.72, "learning_rate": 3.876170349196187e-06, "loss": 0.8267, "step": 14143 }, { "epoch": 0.72, "learning_rate": 3.874869630003466e-06, "loss": 0.894, "step": 14144 }, { "epoch": 0.72, "learning_rate": 3.873569076642133e-06, "loss": 0.9307, "step": 14145 }, { "epoch": 0.72, "learning_rate": 3.872268689147403e-06, "loss": 0.9657, "step": 14146 }, { "epoch": 0.72, "learning_rate": 3.87096846755447e-06, "loss": 0.8605, "step": 14147 }, { "epoch": 0.72, "learning_rate": 3.869668411898554e-06, "loss": 0.9004, "step": 14148 }, { "epoch": 0.72, "learning_rate": 3.868368522214842e-06, "loss": 0.9045, "step": 14149 }, { "epoch": 0.72, "learning_rate": 3.8670687985385235e-06, "loss": 0.9561, "step": 14150 }, { "epoch": 0.72, "learning_rate": 3.865769240904798e-06, "loss": 0.8745, "step": 14151 }, { "epoch": 0.72, "learning_rate": 3.864469849348841e-06, "loss": 0.9214, "step": 14152 }, { "epoch": 0.72, "learning_rate": 3.863170623905835e-06, "loss": 0.859, "step": 14153 }, { "epoch": 0.72, "learning_rate": 3.861871564610955e-06, "loss": 0.8529, "step": 14154 }, { "epoch": 0.72, "learning_rate": 3.860572671499376e-06, "loss": 0.9879, "step": 14155 }, { "epoch": 0.72, "learning_rate": 3.859273944606256e-06, "loss": 0.97, "step": 14156 }, { "epoch": 0.72, "learning_rate": 3.857975383966761e-06, "loss": 0.8605, "step": 14157 }, { "epoch": 0.72, "learning_rate": 3.8566769896160514e-06, "loss": 0.9131, "step": 14158 }, { "epoch": 0.72, "learning_rate": 3.8553787615892706e-06, "loss": 0.8238, "step": 14159 }, { "epoch": 0.72, "learning_rate": 3.854080699921577e-06, "loss": 0.8607, "step": 14160 }, { "epoch": 0.72, "learning_rate": 3.852782804648111e-06, "loss": 0.796, "step": 14161 }, { "epoch": 0.72, "learning_rate": 3.851485075804003e-06, "loss": 0.8033, "step": 14162 }, { "epoch": 0.72, "learning_rate": 3.850187513424402e-06, "loss": 0.7893, "step": 14163 }, { "epoch": 0.72, "learning_rate": 3.84889011754443e-06, "loss": 0.8699, "step": 14164 }, { "epoch": 0.72, "learning_rate": 3.847592888199211e-06, "loss": 0.799, "step": 14165 }, { "epoch": 0.72, "learning_rate": 3.84629582542387e-06, "loss": 0.989, "step": 14166 }, { "epoch": 0.72, "learning_rate": 3.844998929253523e-06, "loss": 0.8531, "step": 14167 }, { "epoch": 0.72, "learning_rate": 3.843702199723283e-06, "loss": 0.8592, "step": 14168 }, { "epoch": 0.72, "learning_rate": 3.842405636868255e-06, "loss": 0.8552, "step": 14169 }, { "epoch": 0.72, "learning_rate": 3.8411092407235415e-06, "loss": 0.7748, "step": 14170 }, { "epoch": 0.72, "learning_rate": 3.839813011324244e-06, "loss": 0.886, "step": 14171 }, { "epoch": 0.72, "learning_rate": 3.838516948705459e-06, "loss": 0.9481, "step": 14172 }, { "epoch": 0.72, "learning_rate": 3.837221052902268e-06, "loss": 0.8962, "step": 14173 }, { "epoch": 0.72, "learning_rate": 3.835925323949761e-06, "loss": 0.8811, "step": 14174 }, { "epoch": 0.72, "learning_rate": 3.8346297618830195e-06, "loss": 0.9221, "step": 14175 }, { "epoch": 0.72, "learning_rate": 3.833334366737112e-06, "loss": 0.9765, "step": 14176 }, { "epoch": 0.72, "learning_rate": 3.832039138547123e-06, "loss": 0.9185, "step": 14177 }, { "epoch": 0.72, "learning_rate": 3.830744077348109e-06, "loss": 0.8746, "step": 14178 }, { "epoch": 0.72, "learning_rate": 3.8294491831751346e-06, "loss": 1.0248, "step": 14179 }, { "epoch": 0.72, "learning_rate": 3.828154456063262e-06, "loss": 0.915, "step": 14180 }, { "epoch": 0.72, "learning_rate": 3.826859896047539e-06, "loss": 0.8726, "step": 14181 }, { "epoch": 0.72, "learning_rate": 3.825565503163016e-06, "loss": 0.8517, "step": 14182 }, { "epoch": 0.72, "learning_rate": 3.824271277444736e-06, "loss": 0.8885, "step": 14183 }, { "epoch": 0.72, "learning_rate": 3.822977218927747e-06, "loss": 0.7914, "step": 14184 }, { "epoch": 0.72, "learning_rate": 3.821683327647073e-06, "loss": 0.8626, "step": 14185 }, { "epoch": 0.72, "learning_rate": 3.820389603637749e-06, "loss": 0.8839, "step": 14186 }, { "epoch": 0.72, "learning_rate": 3.819096046934807e-06, "loss": 0.871, "step": 14187 }, { "epoch": 0.72, "learning_rate": 3.8178026575732566e-06, "loss": 0.9039, "step": 14188 }, { "epoch": 0.72, "learning_rate": 3.816509435588128e-06, "loss": 0.8216, "step": 14189 }, { "epoch": 0.72, "learning_rate": 3.815216381014428e-06, "loss": 0.9254, "step": 14190 }, { "epoch": 0.72, "learning_rate": 3.8139234938871573e-06, "loss": 0.9244, "step": 14191 }, { "epoch": 0.72, "learning_rate": 3.812630774241334e-06, "loss": 0.9175, "step": 14192 }, { "epoch": 0.72, "learning_rate": 3.8113382221119453e-06, "loss": 0.8673, "step": 14193 }, { "epoch": 0.72, "learning_rate": 3.81004583753399e-06, "loss": 0.864, "step": 14194 }, { "epoch": 0.72, "learning_rate": 3.808753620542459e-06, "loss": 0.8488, "step": 14195 }, { "epoch": 0.72, "learning_rate": 3.8074615711723395e-06, "loss": 0.9228, "step": 14196 }, { "epoch": 0.72, "learning_rate": 3.8061696894586062e-06, "loss": 0.8535, "step": 14197 }, { "epoch": 0.72, "learning_rate": 3.8048779754362398e-06, "loss": 0.926, "step": 14198 }, { "epoch": 0.72, "learning_rate": 3.8035864291402137e-06, "loss": 0.9162, "step": 14199 }, { "epoch": 0.72, "learning_rate": 3.8022950506054855e-06, "loss": 0.9413, "step": 14200 }, { "epoch": 0.72, "learning_rate": 3.8010038398670324e-06, "loss": 0.8155, "step": 14201 }, { "epoch": 0.72, "learning_rate": 3.799712796959805e-06, "loss": 0.8801, "step": 14202 }, { "epoch": 0.72, "learning_rate": 3.79842192191875e-06, "loss": 0.8577, "step": 14203 }, { "epoch": 0.72, "learning_rate": 3.79713121477883e-06, "loss": 0.9286, "step": 14204 }, { "epoch": 0.72, "learning_rate": 3.79584067557498e-06, "loss": 0.8948, "step": 14205 }, { "epoch": 0.72, "learning_rate": 3.7945503043421426e-06, "loss": 0.9543, "step": 14206 }, { "epoch": 0.72, "learning_rate": 3.7932601011152547e-06, "loss": 0.8621, "step": 14207 }, { "epoch": 0.72, "learning_rate": 3.791970065929249e-06, "loss": 1.0064, "step": 14208 }, { "epoch": 0.72, "learning_rate": 3.7906801988190444e-06, "loss": 0.8458, "step": 14209 }, { "epoch": 0.72, "learning_rate": 3.7893904998195684e-06, "loss": 0.8608, "step": 14210 }, { "epoch": 0.72, "learning_rate": 3.788100968965741e-06, "loss": 0.8821, "step": 14211 }, { "epoch": 0.72, "learning_rate": 3.786811606292463e-06, "loss": 0.8794, "step": 14212 }, { "epoch": 0.72, "learning_rate": 3.785522411834658e-06, "loss": 0.7591, "step": 14213 }, { "epoch": 0.72, "learning_rate": 3.784233385627222e-06, "loss": 0.8654, "step": 14214 }, { "epoch": 0.72, "learning_rate": 3.7829445277050457e-06, "loss": 0.8761, "step": 14215 }, { "epoch": 0.72, "learning_rate": 3.78165583810304e-06, "loss": 0.9372, "step": 14216 }, { "epoch": 0.72, "learning_rate": 3.7803673168560838e-06, "loss": 0.9001, "step": 14217 }, { "epoch": 0.72, "learning_rate": 3.7790789639990643e-06, "loss": 0.7674, "step": 14218 }, { "epoch": 0.72, "learning_rate": 3.777790779566867e-06, "loss": 0.7742, "step": 14219 }, { "epoch": 0.72, "learning_rate": 3.7765027635943606e-06, "loss": 0.8468, "step": 14220 }, { "epoch": 0.72, "learning_rate": 3.775214916116422e-06, "loss": 0.8996, "step": 14221 }, { "epoch": 0.72, "learning_rate": 3.773927237167916e-06, "loss": 0.8733, "step": 14222 }, { "epoch": 0.72, "learning_rate": 3.7726397267837088e-06, "loss": 0.8351, "step": 14223 }, { "epoch": 0.72, "learning_rate": 3.77135238499865e-06, "loss": 0.9498, "step": 14224 }, { "epoch": 0.72, "learning_rate": 3.7700652118476045e-06, "loss": 0.8906, "step": 14225 }, { "epoch": 0.72, "learning_rate": 3.7687782073654156e-06, "loss": 0.8531, "step": 14226 }, { "epoch": 0.72, "learning_rate": 3.767491371586921e-06, "loss": 0.8674, "step": 14227 }, { "epoch": 0.72, "learning_rate": 3.766204704546973e-06, "loss": 0.8741, "step": 14228 }, { "epoch": 0.72, "learning_rate": 3.764918206280397e-06, "loss": 0.775, "step": 14229 }, { "epoch": 0.72, "learning_rate": 3.763631876822026e-06, "loss": 0.8941, "step": 14230 }, { "epoch": 0.72, "learning_rate": 3.762345716206691e-06, "loss": 0.894, "step": 14231 }, { "epoch": 0.72, "learning_rate": 3.761059724469207e-06, "loss": 0.937, "step": 14232 }, { "epoch": 0.72, "learning_rate": 3.7597739016443922e-06, "loss": 0.8383, "step": 14233 }, { "epoch": 0.72, "learning_rate": 3.7584882477670603e-06, "loss": 0.8991, "step": 14234 }, { "epoch": 0.72, "learning_rate": 3.7572027628720186e-06, "loss": 0.8919, "step": 14235 }, { "epoch": 0.72, "learning_rate": 3.7559174469940697e-06, "loss": 0.8373, "step": 14236 }, { "epoch": 0.72, "learning_rate": 3.7546323001680173e-06, "loss": 0.9171, "step": 14237 }, { "epoch": 0.72, "learning_rate": 3.7533473224286466e-06, "loss": 0.8855, "step": 14238 }, { "epoch": 0.72, "learning_rate": 3.7520625138107515e-06, "loss": 0.9313, "step": 14239 }, { "epoch": 0.72, "learning_rate": 3.7507778743491196e-06, "loss": 0.9323, "step": 14240 }, { "epoch": 0.72, "learning_rate": 3.7494934040785213e-06, "loss": 1.0047, "step": 14241 }, { "epoch": 0.72, "learning_rate": 3.7482091030337463e-06, "loss": 0.8858, "step": 14242 }, { "epoch": 0.72, "learning_rate": 3.746924971249558e-06, "loss": 0.8917, "step": 14243 }, { "epoch": 0.72, "learning_rate": 3.7456410087607176e-06, "loss": 0.8957, "step": 14244 }, { "epoch": 0.72, "learning_rate": 3.744357215601998e-06, "loss": 0.8977, "step": 14245 }, { "epoch": 0.72, "learning_rate": 3.7430735918081497e-06, "loss": 0.9372, "step": 14246 }, { "epoch": 0.72, "learning_rate": 3.7417901374139255e-06, "loss": 0.927, "step": 14247 }, { "epoch": 0.72, "learning_rate": 3.740506852454079e-06, "loss": 0.9057, "step": 14248 }, { "epoch": 0.72, "learning_rate": 3.7392237369633465e-06, "loss": 0.9557, "step": 14249 }, { "epoch": 0.72, "learning_rate": 3.7379407909764708e-06, "loss": 0.8998, "step": 14250 }, { "epoch": 0.72, "learning_rate": 3.736658014528186e-06, "loss": 0.9241, "step": 14251 }, { "epoch": 0.72, "learning_rate": 3.7353754076532243e-06, "loss": 0.9509, "step": 14252 }, { "epoch": 0.72, "learning_rate": 3.7340929703863017e-06, "loss": 0.7543, "step": 14253 }, { "epoch": 0.72, "learning_rate": 3.7328107027621528e-06, "loss": 0.8704, "step": 14254 }, { "epoch": 0.72, "learning_rate": 3.7315286048154862e-06, "loss": 0.8386, "step": 14255 }, { "epoch": 0.72, "learning_rate": 3.7302466765810076e-06, "loss": 0.8093, "step": 14256 }, { "epoch": 0.72, "learning_rate": 3.728964918093435e-06, "loss": 0.8704, "step": 14257 }, { "epoch": 0.72, "learning_rate": 3.7276833293874624e-06, "loss": 0.9257, "step": 14258 }, { "epoch": 0.72, "learning_rate": 3.726401910497791e-06, "loss": 0.8801, "step": 14259 }, { "epoch": 0.72, "learning_rate": 3.725120661459116e-06, "loss": 0.8289, "step": 14260 }, { "epoch": 0.72, "learning_rate": 3.7238395823061203e-06, "loss": 0.9212, "step": 14261 }, { "epoch": 0.72, "learning_rate": 3.7225586730734908e-06, "loss": 0.9446, "step": 14262 }, { "epoch": 0.72, "learning_rate": 3.7212779337959058e-06, "loss": 0.911, "step": 14263 }, { "epoch": 0.72, "learning_rate": 3.7199973645080446e-06, "loss": 0.8748, "step": 14264 }, { "epoch": 0.72, "learning_rate": 3.7187169652445665e-06, "loss": 0.88, "step": 14265 }, { "epoch": 0.72, "learning_rate": 3.7174367360401507e-06, "loss": 0.8578, "step": 14266 }, { "epoch": 0.72, "learning_rate": 3.716156676929451e-06, "loss": 0.8953, "step": 14267 }, { "epoch": 0.72, "learning_rate": 3.714876787947117e-06, "loss": 0.8303, "step": 14268 }, { "epoch": 0.72, "learning_rate": 3.713597069127813e-06, "loss": 0.8565, "step": 14269 }, { "epoch": 0.72, "learning_rate": 3.712317520506179e-06, "loss": 0.9136, "step": 14270 }, { "epoch": 0.73, "learning_rate": 3.711038142116856e-06, "loss": 0.8437, "step": 14271 }, { "epoch": 0.73, "learning_rate": 3.7097589339944896e-06, "loss": 0.8051, "step": 14272 }, { "epoch": 0.73, "learning_rate": 3.7084798961737032e-06, "loss": 0.9349, "step": 14273 }, { "epoch": 0.73, "learning_rate": 3.7072010286891314e-06, "loss": 0.9006, "step": 14274 }, { "epoch": 0.73, "learning_rate": 3.7059223315753956e-06, "loss": 0.9387, "step": 14275 }, { "epoch": 0.73, "learning_rate": 3.7046438048671196e-06, "loss": 0.8997, "step": 14276 }, { "epoch": 0.73, "learning_rate": 3.703365448598909e-06, "loss": 0.8729, "step": 14277 }, { "epoch": 0.73, "learning_rate": 3.7020872628053863e-06, "loss": 0.8589, "step": 14278 }, { "epoch": 0.73, "learning_rate": 3.7008092475211497e-06, "loss": 0.9081, "step": 14279 }, { "epoch": 0.73, "learning_rate": 3.699531402780795e-06, "loss": 0.9468, "step": 14280 }, { "epoch": 0.73, "learning_rate": 3.69825372861893e-06, "loss": 0.8818, "step": 14281 }, { "epoch": 0.73, "learning_rate": 3.696976225070139e-06, "loss": 0.8935, "step": 14282 }, { "epoch": 0.73, "learning_rate": 3.6956988921690096e-06, "loss": 0.9102, "step": 14283 }, { "epoch": 0.73, "learning_rate": 3.6944217299501297e-06, "loss": 0.894, "step": 14284 }, { "epoch": 0.73, "learning_rate": 3.693144738448069e-06, "loss": 0.8666, "step": 14285 }, { "epoch": 0.73, "learning_rate": 3.6918679176974047e-06, "loss": 0.8739, "step": 14286 }, { "epoch": 0.73, "learning_rate": 3.6905912677327047e-06, "loss": 0.9582, "step": 14287 }, { "epoch": 0.73, "learning_rate": 3.6893147885885374e-06, "loss": 0.9042, "step": 14288 }, { "epoch": 0.73, "learning_rate": 3.688038480299454e-06, "loss": 0.9138, "step": 14289 }, { "epoch": 0.73, "learning_rate": 3.6867623429000144e-06, "loss": 0.859, "step": 14290 }, { "epoch": 0.73, "learning_rate": 3.6854863764247708e-06, "loss": 0.9548, "step": 14291 }, { "epoch": 0.73, "learning_rate": 3.6842105809082574e-06, "loss": 0.9142, "step": 14292 }, { "epoch": 0.73, "learning_rate": 3.682934956385031e-06, "loss": 0.8936, "step": 14293 }, { "epoch": 0.73, "learning_rate": 3.6816595028896164e-06, "loss": 0.8863, "step": 14294 }, { "epoch": 0.73, "learning_rate": 3.680384220456549e-06, "loss": 0.8425, "step": 14295 }, { "epoch": 0.73, "learning_rate": 3.679109109120358e-06, "loss": 0.8965, "step": 14296 }, { "epoch": 0.73, "learning_rate": 3.6778341689155596e-06, "loss": 0.8904, "step": 14297 }, { "epoch": 0.73, "learning_rate": 3.6765593998766747e-06, "loss": 0.8986, "step": 14298 }, { "epoch": 0.73, "learning_rate": 3.675284802038217e-06, "loss": 0.8779, "step": 14299 }, { "epoch": 0.73, "learning_rate": 3.674010375434697e-06, "loss": 0.9162, "step": 14300 }, { "epoch": 0.73, "learning_rate": 3.672736120100613e-06, "loss": 0.8719, "step": 14301 }, { "epoch": 0.73, "learning_rate": 3.6714620360704666e-06, "loss": 0.813, "step": 14302 }, { "epoch": 0.73, "learning_rate": 3.6701881233787553e-06, "loss": 0.941, "step": 14303 }, { "epoch": 0.73, "learning_rate": 3.6689143820599606e-06, "loss": 0.9434, "step": 14304 }, { "epoch": 0.73, "learning_rate": 3.6676408121485794e-06, "loss": 0.8152, "step": 14305 }, { "epoch": 0.73, "learning_rate": 3.6663674136790795e-06, "loss": 0.851, "step": 14306 }, { "epoch": 0.73, "learning_rate": 3.665094186685949e-06, "loss": 0.8682, "step": 14307 }, { "epoch": 0.73, "learning_rate": 3.663821131203655e-06, "loss": 0.8304, "step": 14308 }, { "epoch": 0.73, "learning_rate": 3.6625482472666564e-06, "loss": 0.9009, "step": 14309 }, { "epoch": 0.73, "learning_rate": 3.6612755349094266e-06, "loss": 0.8231, "step": 14310 }, { "epoch": 0.73, "learning_rate": 3.6600029941664163e-06, "loss": 0.9565, "step": 14311 }, { "epoch": 0.73, "learning_rate": 3.6587306250720777e-06, "loss": 0.9127, "step": 14312 }, { "epoch": 0.73, "learning_rate": 3.657458427660866e-06, "loss": 0.7928, "step": 14313 }, { "epoch": 0.73, "learning_rate": 3.6561864019672146e-06, "loss": 0.8756, "step": 14314 }, { "epoch": 0.73, "learning_rate": 3.6549145480255675e-06, "loss": 0.9142, "step": 14315 }, { "epoch": 0.73, "learning_rate": 3.6536428658703594e-06, "loss": 0.8439, "step": 14316 }, { "epoch": 0.73, "learning_rate": 3.6523713555360206e-06, "loss": 0.8628, "step": 14317 }, { "epoch": 0.73, "learning_rate": 3.6511000170569723e-06, "loss": 0.8991, "step": 14318 }, { "epoch": 0.73, "learning_rate": 3.649828850467635e-06, "loss": 0.9228, "step": 14319 }, { "epoch": 0.73, "learning_rate": 3.6485578558024293e-06, "loss": 0.8956, "step": 14320 }, { "epoch": 0.73, "learning_rate": 3.6472870330957555e-06, "loss": 0.9048, "step": 14321 }, { "epoch": 0.73, "learning_rate": 3.646016382382033e-06, "loss": 0.7621, "step": 14322 }, { "epoch": 0.73, "learning_rate": 3.6447459036956535e-06, "loss": 0.9909, "step": 14323 }, { "epoch": 0.73, "learning_rate": 3.643475597071018e-06, "loss": 0.9069, "step": 14324 }, { "epoch": 0.73, "learning_rate": 3.6422054625425205e-06, "loss": 0.8723, "step": 14325 }, { "epoch": 0.73, "learning_rate": 3.6409355001445425e-06, "loss": 0.8356, "step": 14326 }, { "epoch": 0.73, "learning_rate": 3.6396657099114706e-06, "loss": 0.7951, "step": 14327 }, { "epoch": 0.73, "learning_rate": 3.638396091877683e-06, "loss": 0.9359, "step": 14328 }, { "epoch": 0.73, "learning_rate": 3.6371266460775558e-06, "loss": 0.8608, "step": 14329 }, { "epoch": 0.73, "learning_rate": 3.6358573725454525e-06, "loss": 0.941, "step": 14330 }, { "epoch": 0.73, "learning_rate": 3.634588271315739e-06, "loss": 0.9187, "step": 14331 }, { "epoch": 0.73, "learning_rate": 3.6333193424227796e-06, "loss": 0.7983, "step": 14332 }, { "epoch": 0.73, "learning_rate": 3.632050585900919e-06, "loss": 0.9738, "step": 14333 }, { "epoch": 0.73, "learning_rate": 3.6307820017845197e-06, "loss": 0.8563, "step": 14334 }, { "epoch": 0.73, "learning_rate": 3.6295135901079193e-06, "loss": 0.9281, "step": 14335 }, { "epoch": 0.73, "learning_rate": 3.6282453509054594e-06, "loss": 0.9289, "step": 14336 }, { "epoch": 0.73, "learning_rate": 3.6269772842114816e-06, "loss": 0.9761, "step": 14337 }, { "epoch": 0.73, "learning_rate": 3.6257093900603103e-06, "loss": 0.9445, "step": 14338 }, { "epoch": 0.73, "learning_rate": 3.6244416684862748e-06, "loss": 0.8569, "step": 14339 }, { "epoch": 0.73, "learning_rate": 3.623174119523699e-06, "loss": 0.858, "step": 14340 }, { "epoch": 0.73, "learning_rate": 3.6219067432069035e-06, "loss": 0.8497, "step": 14341 }, { "epoch": 0.73, "learning_rate": 3.620639539570192e-06, "loss": 0.7833, "step": 14342 }, { "epoch": 0.73, "learning_rate": 3.619372508647878e-06, "loss": 0.8831, "step": 14343 }, { "epoch": 0.73, "learning_rate": 3.6181056504742695e-06, "loss": 0.8826, "step": 14344 }, { "epoch": 0.73, "learning_rate": 3.6168389650836534e-06, "loss": 0.8633, "step": 14345 }, { "epoch": 0.73, "learning_rate": 3.6155724525103375e-06, "loss": 0.8877, "step": 14346 }, { "epoch": 0.73, "learning_rate": 3.6143061127886025e-06, "loss": 0.8851, "step": 14347 }, { "epoch": 0.73, "learning_rate": 3.6130399459527354e-06, "loss": 0.7914, "step": 14348 }, { "epoch": 0.73, "learning_rate": 3.6117739520370186e-06, "loss": 0.7994, "step": 14349 }, { "epoch": 0.73, "learning_rate": 3.610508131075724e-06, "loss": 0.9903, "step": 14350 }, { "epoch": 0.73, "learning_rate": 3.6092424831031226e-06, "loss": 0.8736, "step": 14351 }, { "epoch": 0.73, "learning_rate": 3.6079770081534824e-06, "loss": 0.8994, "step": 14352 }, { "epoch": 0.73, "learning_rate": 3.606711706261068e-06, "loss": 0.8418, "step": 14353 }, { "epoch": 0.73, "learning_rate": 3.605446577460128e-06, "loss": 0.9084, "step": 14354 }, { "epoch": 0.73, "learning_rate": 3.6041816217849178e-06, "loss": 0.913, "step": 14355 }, { "epoch": 0.73, "learning_rate": 3.602916839269691e-06, "loss": 0.953, "step": 14356 }, { "epoch": 0.73, "learning_rate": 3.601652229948677e-06, "loss": 0.8149, "step": 14357 }, { "epoch": 0.73, "learning_rate": 3.6003877938561283e-06, "loss": 0.7753, "step": 14358 }, { "epoch": 0.73, "learning_rate": 3.5991235310262672e-06, "loss": 0.9157, "step": 14359 }, { "epoch": 0.73, "learning_rate": 3.5978594414933277e-06, "loss": 0.7759, "step": 14360 }, { "epoch": 0.73, "learning_rate": 3.5965955252915362e-06, "loss": 0.9078, "step": 14361 }, { "epoch": 0.73, "learning_rate": 3.5953317824551036e-06, "loss": 0.8202, "step": 14362 }, { "epoch": 0.73, "learning_rate": 3.594068213018249e-06, "loss": 0.8423, "step": 14363 }, { "epoch": 0.73, "learning_rate": 3.5928048170151832e-06, "loss": 0.8337, "step": 14364 }, { "epoch": 0.73, "learning_rate": 3.5915415944801137e-06, "loss": 0.9401, "step": 14365 }, { "epoch": 0.73, "learning_rate": 3.590278545447233e-06, "loss": 0.8331, "step": 14366 }, { "epoch": 0.73, "learning_rate": 3.589015669950742e-06, "loss": 0.8416, "step": 14367 }, { "epoch": 0.73, "learning_rate": 3.5877529680248336e-06, "loss": 0.9055, "step": 14368 }, { "epoch": 0.73, "learning_rate": 3.5864904397036858e-06, "loss": 0.8741, "step": 14369 }, { "epoch": 0.73, "learning_rate": 3.5852280850214915e-06, "loss": 0.839, "step": 14370 }, { "epoch": 0.73, "learning_rate": 3.58396590401242e-06, "loss": 0.787, "step": 14371 }, { "epoch": 0.73, "learning_rate": 3.5827038967106443e-06, "loss": 0.9301, "step": 14372 }, { "epoch": 0.73, "learning_rate": 3.5814420631503364e-06, "loss": 0.8405, "step": 14373 }, { "epoch": 0.73, "learning_rate": 3.5801804033656508e-06, "loss": 0.8829, "step": 14374 }, { "epoch": 0.73, "learning_rate": 3.578918917390756e-06, "loss": 0.8719, "step": 14375 }, { "epoch": 0.73, "learning_rate": 3.5776576052597968e-06, "loss": 0.8956, "step": 14376 }, { "epoch": 0.73, "learning_rate": 3.576396467006925e-06, "loss": 0.8745, "step": 14377 }, { "epoch": 0.73, "learning_rate": 3.575135502666288e-06, "loss": 0.842, "step": 14378 }, { "epoch": 0.73, "learning_rate": 3.5738747122720187e-06, "loss": 0.8227, "step": 14379 }, { "epoch": 0.73, "learning_rate": 3.5726140958582556e-06, "loss": 0.8474, "step": 14380 }, { "epoch": 0.73, "learning_rate": 3.5713536534591263e-06, "loss": 0.9209, "step": 14381 }, { "epoch": 0.73, "learning_rate": 3.5700933851087603e-06, "loss": 0.8951, "step": 14382 }, { "epoch": 0.73, "learning_rate": 3.5688332908412726e-06, "loss": 0.8551, "step": 14383 }, { "epoch": 0.73, "learning_rate": 3.5675733706907813e-06, "loss": 0.8916, "step": 14384 }, { "epoch": 0.73, "learning_rate": 3.566313624691401e-06, "loss": 0.8527, "step": 14385 }, { "epoch": 0.73, "learning_rate": 3.565054052877227e-06, "loss": 0.923, "step": 14386 }, { "epoch": 0.73, "learning_rate": 3.5637946552823757e-06, "loss": 0.8723, "step": 14387 }, { "epoch": 0.73, "learning_rate": 3.562535431940932e-06, "loss": 0.9128, "step": 14388 }, { "epoch": 0.73, "learning_rate": 3.561276382886993e-06, "loss": 0.8384, "step": 14389 }, { "epoch": 0.73, "learning_rate": 3.5600175081546506e-06, "loss": 0.852, "step": 14390 }, { "epoch": 0.73, "learning_rate": 3.5587588077779776e-06, "loss": 0.8443, "step": 14391 }, { "epoch": 0.73, "learning_rate": 3.5575002817910575e-06, "loss": 0.964, "step": 14392 }, { "epoch": 0.73, "learning_rate": 3.5562419302279626e-06, "loss": 0.8273, "step": 14393 }, { "epoch": 0.73, "learning_rate": 3.554983753122765e-06, "loss": 0.8453, "step": 14394 }, { "epoch": 0.73, "learning_rate": 3.5537257505095225e-06, "loss": 0.8656, "step": 14395 }, { "epoch": 0.73, "learning_rate": 3.552467922422297e-06, "loss": 0.8329, "step": 14396 }, { "epoch": 0.73, "learning_rate": 3.5512102688951465e-06, "loss": 0.9312, "step": 14397 }, { "epoch": 0.73, "learning_rate": 3.5499527899621088e-06, "loss": 0.8208, "step": 14398 }, { "epoch": 0.73, "learning_rate": 3.5486954856572455e-06, "loss": 0.9822, "step": 14399 }, { "epoch": 0.73, "learning_rate": 3.547438356014584e-06, "loss": 0.8409, "step": 14400 }, { "epoch": 0.73, "learning_rate": 3.546181401068164e-06, "loss": 0.9649, "step": 14401 }, { "epoch": 0.73, "learning_rate": 3.5449246208520204e-06, "loss": 0.8365, "step": 14402 }, { "epoch": 0.73, "learning_rate": 3.543668015400171e-06, "loss": 0.9245, "step": 14403 }, { "epoch": 0.73, "learning_rate": 3.542411584746641e-06, "loss": 0.9417, "step": 14404 }, { "epoch": 0.73, "learning_rate": 3.5411553289254473e-06, "loss": 0.8088, "step": 14405 }, { "epoch": 0.73, "learning_rate": 3.539899247970604e-06, "loss": 0.8056, "step": 14406 }, { "epoch": 0.73, "learning_rate": 3.538643341916114e-06, "loss": 0.7067, "step": 14407 }, { "epoch": 0.73, "learning_rate": 3.5373876107959793e-06, "loss": 0.8727, "step": 14408 }, { "epoch": 0.73, "learning_rate": 3.5361320546442036e-06, "loss": 0.8293, "step": 14409 }, { "epoch": 0.73, "learning_rate": 3.5348766734947683e-06, "loss": 0.8842, "step": 14410 }, { "epoch": 0.73, "learning_rate": 3.533621467381676e-06, "loss": 0.9314, "step": 14411 }, { "epoch": 0.73, "learning_rate": 3.532366436338899e-06, "loss": 0.8648, "step": 14412 }, { "epoch": 0.73, "learning_rate": 3.5311115804004205e-06, "loss": 0.9877, "step": 14413 }, { "epoch": 0.73, "learning_rate": 3.529856899600216e-06, "loss": 0.7801, "step": 14414 }, { "epoch": 0.73, "learning_rate": 3.528602393972249e-06, "loss": 0.8614, "step": 14415 }, { "epoch": 0.73, "learning_rate": 3.5273480635504876e-06, "loss": 0.9354, "step": 14416 }, { "epoch": 0.73, "learning_rate": 3.5260939083688904e-06, "loss": 0.873, "step": 14417 }, { "epoch": 0.73, "learning_rate": 3.5248399284614175e-06, "loss": 0.8471, "step": 14418 }, { "epoch": 0.73, "learning_rate": 3.5235861238620095e-06, "loss": 0.928, "step": 14419 }, { "epoch": 0.73, "learning_rate": 3.5223324946046177e-06, "loss": 0.8945, "step": 14420 }, { "epoch": 0.73, "learning_rate": 3.5210790407231855e-06, "loss": 0.9744, "step": 14421 }, { "epoch": 0.73, "learning_rate": 3.5198257622516385e-06, "loss": 0.9106, "step": 14422 }, { "epoch": 0.73, "learning_rate": 3.5185726592239213e-06, "loss": 0.91, "step": 14423 }, { "epoch": 0.73, "learning_rate": 3.5173197316739505e-06, "loss": 0.8787, "step": 14424 }, { "epoch": 0.73, "learning_rate": 3.5160669796356506e-06, "loss": 0.8915, "step": 14425 }, { "epoch": 0.73, "learning_rate": 3.514814403142942e-06, "loss": 0.9469, "step": 14426 }, { "epoch": 0.73, "learning_rate": 3.5135620022297303e-06, "loss": 0.8528, "step": 14427 }, { "epoch": 0.73, "learning_rate": 3.5123097769299263e-06, "loss": 1.0623, "step": 14428 }, { "epoch": 0.73, "learning_rate": 3.5110577272774336e-06, "loss": 0.9011, "step": 14429 }, { "epoch": 0.73, "learning_rate": 3.509805853306152e-06, "loss": 0.8648, "step": 14430 }, { "epoch": 0.73, "learning_rate": 3.508554155049969e-06, "loss": 0.8419, "step": 14431 }, { "epoch": 0.73, "learning_rate": 3.507302632542775e-06, "loss": 0.7927, "step": 14432 }, { "epoch": 0.73, "learning_rate": 3.5060512858184583e-06, "loss": 0.906, "step": 14433 }, { "epoch": 0.73, "learning_rate": 3.504800114910888e-06, "loss": 0.7954, "step": 14434 }, { "epoch": 0.73, "learning_rate": 3.5035491198539494e-06, "loss": 0.8669, "step": 14435 }, { "epoch": 0.73, "learning_rate": 3.5022983006815046e-06, "loss": 0.7959, "step": 14436 }, { "epoch": 0.73, "learning_rate": 3.50104765742742e-06, "loss": 0.8782, "step": 14437 }, { "epoch": 0.73, "learning_rate": 3.4997971901255588e-06, "loss": 0.85, "step": 14438 }, { "epoch": 0.73, "learning_rate": 3.4985468988097704e-06, "loss": 0.7991, "step": 14439 }, { "epoch": 0.73, "learning_rate": 3.497296783513907e-06, "loss": 0.8548, "step": 14440 }, { "epoch": 0.73, "learning_rate": 3.4960468442718156e-06, "loss": 0.9019, "step": 14441 }, { "epoch": 0.73, "learning_rate": 3.4947970811173393e-06, "loss": 0.8666, "step": 14442 }, { "epoch": 0.73, "learning_rate": 3.493547494084307e-06, "loss": 0.8116, "step": 14443 }, { "epoch": 0.73, "learning_rate": 3.4922980832065544e-06, "loss": 0.9398, "step": 14444 }, { "epoch": 0.73, "learning_rate": 3.4910488485179084e-06, "loss": 0.9015, "step": 14445 }, { "epoch": 0.73, "learning_rate": 3.4897997900521886e-06, "loss": 0.9509, "step": 14446 }, { "epoch": 0.73, "learning_rate": 3.488550907843217e-06, "loss": 0.841, "step": 14447 }, { "epoch": 0.73, "learning_rate": 3.487302201924798e-06, "loss": 0.8305, "step": 14448 }, { "epoch": 0.73, "learning_rate": 3.4860536723307435e-06, "loss": 0.8483, "step": 14449 }, { "epoch": 0.73, "learning_rate": 3.4848053190948592e-06, "loss": 0.8839, "step": 14450 }, { "epoch": 0.73, "learning_rate": 3.483557142250933e-06, "loss": 0.8577, "step": 14451 }, { "epoch": 0.73, "learning_rate": 3.4823091418327695e-06, "loss": 0.8095, "step": 14452 }, { "epoch": 0.73, "learning_rate": 3.481061317874149e-06, "loss": 0.7663, "step": 14453 }, { "epoch": 0.73, "learning_rate": 3.479813670408858e-06, "loss": 0.8802, "step": 14454 }, { "epoch": 0.73, "learning_rate": 3.4785661994706775e-06, "loss": 0.8643, "step": 14455 }, { "epoch": 0.73, "learning_rate": 3.4773189050933765e-06, "loss": 0.8731, "step": 14456 }, { "epoch": 0.73, "learning_rate": 3.4760717873107263e-06, "loss": 0.9531, "step": 14457 }, { "epoch": 0.73, "learning_rate": 3.4748248461564913e-06, "loss": 0.9378, "step": 14458 }, { "epoch": 0.73, "learning_rate": 3.4735780816644348e-06, "loss": 0.8479, "step": 14459 }, { "epoch": 0.73, "learning_rate": 3.472331493868305e-06, "loss": 0.8898, "step": 14460 }, { "epoch": 0.73, "learning_rate": 3.471085082801855e-06, "loss": 0.9113, "step": 14461 }, { "epoch": 0.73, "learning_rate": 3.469838848498833e-06, "loss": 0.8387, "step": 14462 }, { "epoch": 0.73, "learning_rate": 3.468592790992971e-06, "loss": 0.9351, "step": 14463 }, { "epoch": 0.73, "learning_rate": 3.467346910318016e-06, "loss": 0.8619, "step": 14464 }, { "epoch": 0.73, "learning_rate": 3.466101206507689e-06, "loss": 0.8272, "step": 14465 }, { "epoch": 0.73, "learning_rate": 3.4648556795957224e-06, "loss": 0.8973, "step": 14466 }, { "epoch": 0.73, "learning_rate": 3.4636103296158373e-06, "loss": 0.918, "step": 14467 }, { "epoch": 0.74, "learning_rate": 3.4623651566017457e-06, "loss": 0.7656, "step": 14468 }, { "epoch": 0.74, "learning_rate": 3.4611201605871625e-06, "loss": 0.8722, "step": 14469 }, { "epoch": 0.74, "learning_rate": 3.459875341605793e-06, "loss": 0.8895, "step": 14470 }, { "epoch": 0.74, "learning_rate": 3.4586306996913444e-06, "loss": 0.881, "step": 14471 }, { "epoch": 0.74, "learning_rate": 3.4573862348775066e-06, "loss": 0.9281, "step": 14472 }, { "epoch": 0.74, "learning_rate": 3.4561419471979762e-06, "loss": 0.8663, "step": 14473 }, { "epoch": 0.74, "learning_rate": 3.454897836686444e-06, "loss": 0.9023, "step": 14474 }, { "epoch": 0.74, "learning_rate": 3.4536539033765814e-06, "loss": 0.8627, "step": 14475 }, { "epoch": 0.74, "learning_rate": 3.4524101473020823e-06, "loss": 0.8157, "step": 14476 }, { "epoch": 0.74, "learning_rate": 3.451166568496609e-06, "loss": 0.9003, "step": 14477 }, { "epoch": 0.74, "learning_rate": 3.4499231669938337e-06, "loss": 0.9449, "step": 14478 }, { "epoch": 0.74, "learning_rate": 3.448679942827422e-06, "loss": 0.8626, "step": 14479 }, { "epoch": 0.74, "learning_rate": 3.4474368960310288e-06, "loss": 0.8756, "step": 14480 }, { "epoch": 0.74, "learning_rate": 3.44619402663831e-06, "loss": 0.82, "step": 14481 }, { "epoch": 0.74, "learning_rate": 3.4449513346829143e-06, "loss": 0.8629, "step": 14482 }, { "epoch": 0.74, "learning_rate": 3.4437088201984916e-06, "loss": 0.8733, "step": 14483 }, { "epoch": 0.74, "learning_rate": 3.442466483218674e-06, "loss": 0.9038, "step": 14484 }, { "epoch": 0.74, "learning_rate": 3.441224323777098e-06, "loss": 0.7653, "step": 14485 }, { "epoch": 0.74, "learning_rate": 3.439982341907401e-06, "loss": 0.83, "step": 14486 }, { "epoch": 0.74, "learning_rate": 3.4387405376431947e-06, "loss": 0.8779, "step": 14487 }, { "epoch": 0.74, "learning_rate": 3.4374989110181143e-06, "loss": 0.9004, "step": 14488 }, { "epoch": 0.74, "learning_rate": 3.4362574620657663e-06, "loss": 0.7999, "step": 14489 }, { "epoch": 0.74, "learning_rate": 3.435016190819764e-06, "loss": 0.8831, "step": 14490 }, { "epoch": 0.74, "learning_rate": 3.4337750973137175e-06, "loss": 0.915, "step": 14491 }, { "epoch": 0.74, "learning_rate": 3.432534181581221e-06, "loss": 0.8494, "step": 14492 }, { "epoch": 0.74, "learning_rate": 3.431293443655873e-06, "loss": 0.9149, "step": 14493 }, { "epoch": 0.74, "learning_rate": 3.4300528835712677e-06, "loss": 0.7818, "step": 14494 }, { "epoch": 0.74, "learning_rate": 3.428812501360994e-06, "loss": 0.8533, "step": 14495 }, { "epoch": 0.74, "learning_rate": 3.4275722970586268e-06, "loss": 0.8602, "step": 14496 }, { "epoch": 0.74, "learning_rate": 3.4263322706977477e-06, "loss": 0.9244, "step": 14497 }, { "epoch": 0.74, "learning_rate": 3.4250924223119308e-06, "loss": 0.9151, "step": 14498 }, { "epoch": 0.74, "learning_rate": 3.4238527519347353e-06, "loss": 0.8027, "step": 14499 }, { "epoch": 0.74, "learning_rate": 3.4226132595997374e-06, "loss": 0.9328, "step": 14500 }, { "epoch": 0.74, "learning_rate": 3.4213739453404837e-06, "loss": 0.9046, "step": 14501 }, { "epoch": 0.74, "learning_rate": 3.4201348091905316e-06, "loss": 0.9159, "step": 14502 }, { "epoch": 0.74, "learning_rate": 3.4188958511834325e-06, "loss": 0.7728, "step": 14503 }, { "epoch": 0.74, "learning_rate": 3.4176570713527224e-06, "loss": 0.938, "step": 14504 }, { "epoch": 0.74, "learning_rate": 3.416418469731945e-06, "loss": 0.8448, "step": 14505 }, { "epoch": 0.74, "learning_rate": 3.4151800463546325e-06, "loss": 0.9686, "step": 14506 }, { "epoch": 0.74, "learning_rate": 3.4139418012543178e-06, "loss": 0.8305, "step": 14507 }, { "epoch": 0.74, "learning_rate": 3.41270373446452e-06, "loss": 0.9344, "step": 14508 }, { "epoch": 0.74, "learning_rate": 3.4114658460187587e-06, "loss": 0.8748, "step": 14509 }, { "epoch": 0.74, "learning_rate": 3.410228135950554e-06, "loss": 0.989, "step": 14510 }, { "epoch": 0.74, "learning_rate": 3.408990604293404e-06, "loss": 0.8784, "step": 14511 }, { "epoch": 0.74, "learning_rate": 3.4077532510808287e-06, "loss": 0.8586, "step": 14512 }, { "epoch": 0.74, "learning_rate": 3.406516076346317e-06, "loss": 0.815, "step": 14513 }, { "epoch": 0.74, "learning_rate": 3.405279080123367e-06, "loss": 0.8618, "step": 14514 }, { "epoch": 0.74, "learning_rate": 3.404042262445474e-06, "loss": 0.8485, "step": 14515 }, { "epoch": 0.74, "learning_rate": 3.4028056233461117e-06, "loss": 0.8648, "step": 14516 }, { "epoch": 0.74, "learning_rate": 3.4015691628587753e-06, "loss": 0.894, "step": 14517 }, { "epoch": 0.74, "learning_rate": 3.40033288101693e-06, "loss": 0.8894, "step": 14518 }, { "epoch": 0.74, "learning_rate": 3.399096777854051e-06, "loss": 0.7569, "step": 14519 }, { "epoch": 0.74, "learning_rate": 3.3978608534036073e-06, "loss": 0.8355, "step": 14520 }, { "epoch": 0.74, "learning_rate": 3.3966251076990533e-06, "loss": 0.9679, "step": 14521 }, { "epoch": 0.74, "learning_rate": 3.395389540773848e-06, "loss": 0.924, "step": 14522 }, { "epoch": 0.74, "learning_rate": 3.3941541526614452e-06, "loss": 0.9066, "step": 14523 }, { "epoch": 0.74, "learning_rate": 3.3929189433952946e-06, "loss": 0.9266, "step": 14524 }, { "epoch": 0.74, "learning_rate": 3.3916839130088295e-06, "loss": 0.8976, "step": 14525 }, { "epoch": 0.74, "learning_rate": 3.3904490615354923e-06, "loss": 0.9044, "step": 14526 }, { "epoch": 0.74, "learning_rate": 3.3892143890087194e-06, "loss": 0.8481, "step": 14527 }, { "epoch": 0.74, "learning_rate": 3.3879798954619248e-06, "loss": 0.88, "step": 14528 }, { "epoch": 0.74, "learning_rate": 3.3867455809285475e-06, "loss": 0.8342, "step": 14529 }, { "epoch": 0.74, "learning_rate": 3.3855114454419944e-06, "loss": 0.9438, "step": 14530 }, { "epoch": 0.74, "learning_rate": 3.384277489035682e-06, "loss": 0.9122, "step": 14531 }, { "epoch": 0.74, "learning_rate": 3.38304371174302e-06, "loss": 0.8745, "step": 14532 }, { "epoch": 0.74, "learning_rate": 3.381810113597408e-06, "loss": 0.9348, "step": 14533 }, { "epoch": 0.74, "learning_rate": 3.380576694632245e-06, "loss": 0.9125, "step": 14534 }, { "epoch": 0.74, "learning_rate": 3.3793434548809256e-06, "loss": 0.9009, "step": 14535 }, { "epoch": 0.74, "learning_rate": 3.378110394376842e-06, "loss": 0.871, "step": 14536 }, { "epoch": 0.74, "learning_rate": 3.3768775131533704e-06, "loss": 0.9175, "step": 14537 }, { "epoch": 0.74, "learning_rate": 3.3756448112438946e-06, "loss": 0.9144, "step": 14538 }, { "epoch": 0.74, "learning_rate": 3.3744122886817907e-06, "loss": 0.9185, "step": 14539 }, { "epoch": 0.74, "learning_rate": 3.373179945500418e-06, "loss": 0.913, "step": 14540 }, { "epoch": 0.74, "learning_rate": 3.3719477817331547e-06, "loss": 0.905, "step": 14541 }, { "epoch": 0.74, "learning_rate": 3.3707157974133497e-06, "loss": 0.9013, "step": 14542 }, { "epoch": 0.74, "learning_rate": 3.369483992574363e-06, "loss": 0.9034, "step": 14543 }, { "epoch": 0.74, "learning_rate": 3.368252367249545e-06, "loss": 0.8273, "step": 14544 }, { "epoch": 0.74, "learning_rate": 3.367020921472235e-06, "loss": 0.9404, "step": 14545 }, { "epoch": 0.74, "learning_rate": 3.3657896552757773e-06, "loss": 0.8768, "step": 14546 }, { "epoch": 0.74, "learning_rate": 3.3645585686935065e-06, "loss": 0.7946, "step": 14547 }, { "epoch": 0.74, "learning_rate": 3.363327661758755e-06, "loss": 0.8427, "step": 14548 }, { "epoch": 0.74, "learning_rate": 3.362096934504845e-06, "loss": 0.8276, "step": 14549 }, { "epoch": 0.74, "learning_rate": 3.360866386965097e-06, "loss": 0.8494, "step": 14550 }, { "epoch": 0.74, "learning_rate": 3.3596360191728306e-06, "loss": 0.8993, "step": 14551 }, { "epoch": 0.74, "learning_rate": 3.3584058311613487e-06, "loss": 0.9116, "step": 14552 }, { "epoch": 0.74, "learning_rate": 3.3571758229639696e-06, "loss": 0.9268, "step": 14553 }, { "epoch": 0.74, "learning_rate": 3.355945994613984e-06, "loss": 0.9177, "step": 14554 }, { "epoch": 0.74, "learning_rate": 3.354716346144692e-06, "loss": 0.9032, "step": 14555 }, { "epoch": 0.74, "learning_rate": 3.353486877589388e-06, "loss": 0.8554, "step": 14556 }, { "epoch": 0.74, "learning_rate": 3.3522575889813524e-06, "loss": 0.8424, "step": 14557 }, { "epoch": 0.74, "learning_rate": 3.3510284803538694e-06, "loss": 0.8729, "step": 14558 }, { "epoch": 0.74, "learning_rate": 3.3497995517402173e-06, "loss": 0.9509, "step": 14559 }, { "epoch": 0.74, "learning_rate": 3.3485708031736698e-06, "loss": 0.8106, "step": 14560 }, { "epoch": 0.74, "learning_rate": 3.3473422346874873e-06, "loss": 0.8293, "step": 14561 }, { "epoch": 0.74, "learning_rate": 3.3461138463149367e-06, "loss": 0.8001, "step": 14562 }, { "epoch": 0.74, "learning_rate": 3.344885638089278e-06, "loss": 0.9372, "step": 14563 }, { "epoch": 0.74, "learning_rate": 3.3436576100437535e-06, "loss": 0.859, "step": 14564 }, { "epoch": 0.74, "learning_rate": 3.3424297622116233e-06, "loss": 0.7751, "step": 14565 }, { "epoch": 0.74, "learning_rate": 3.3412020946261213e-06, "loss": 1.0048, "step": 14566 }, { "epoch": 0.74, "learning_rate": 3.3399746073204886e-06, "loss": 0.8534, "step": 14567 }, { "epoch": 0.74, "learning_rate": 3.33874730032796e-06, "loss": 0.9311, "step": 14568 }, { "epoch": 0.74, "learning_rate": 3.3375201736817586e-06, "loss": 0.8064, "step": 14569 }, { "epoch": 0.74, "learning_rate": 3.3362932274151103e-06, "loss": 0.7484, "step": 14570 }, { "epoch": 0.74, "learning_rate": 3.3350664615612337e-06, "loss": 0.9805, "step": 14571 }, { "epoch": 0.74, "learning_rate": 3.3338398761533443e-06, "loss": 0.9156, "step": 14572 }, { "epoch": 0.74, "learning_rate": 3.332613471224645e-06, "loss": 0.9455, "step": 14573 }, { "epoch": 0.74, "learning_rate": 3.331387246808342e-06, "loss": 0.7833, "step": 14574 }, { "epoch": 0.74, "learning_rate": 3.3301612029376383e-06, "loss": 0.8898, "step": 14575 }, { "epoch": 0.74, "learning_rate": 3.328935339645717e-06, "loss": 0.8961, "step": 14576 }, { "epoch": 0.74, "learning_rate": 3.327709656965781e-06, "loss": 0.8884, "step": 14577 }, { "epoch": 0.74, "learning_rate": 3.3264841549310024e-06, "loss": 0.8752, "step": 14578 }, { "epoch": 0.74, "learning_rate": 3.325258833574566e-06, "loss": 0.8884, "step": 14579 }, { "epoch": 0.74, "learning_rate": 3.32403369292965e-06, "loss": 0.919, "step": 14580 }, { "epoch": 0.74, "learning_rate": 3.322808733029411e-06, "loss": 0.9501, "step": 14581 }, { "epoch": 0.74, "learning_rate": 3.321583953907028e-06, "loss": 0.8699, "step": 14582 }, { "epoch": 0.74, "learning_rate": 3.3203593555956527e-06, "loss": 0.905, "step": 14583 }, { "epoch": 0.74, "learning_rate": 3.31913493812844e-06, "loss": 0.8796, "step": 14584 }, { "epoch": 0.74, "learning_rate": 3.3179107015385438e-06, "loss": 0.9079, "step": 14585 }, { "epoch": 0.74, "learning_rate": 3.316686645859104e-06, "loss": 0.7895, "step": 14586 }, { "epoch": 0.74, "learning_rate": 3.3154627711232624e-06, "loss": 0.7878, "step": 14587 }, { "epoch": 0.74, "learning_rate": 3.314239077364154e-06, "loss": 0.8253, "step": 14588 }, { "epoch": 0.74, "learning_rate": 3.3130155646149133e-06, "loss": 0.8148, "step": 14589 }, { "epoch": 0.74, "learning_rate": 3.3117922329086584e-06, "loss": 0.9107, "step": 14590 }, { "epoch": 0.74, "learning_rate": 3.310569082278512e-06, "loss": 0.7791, "step": 14591 }, { "epoch": 0.74, "learning_rate": 3.3093461127575954e-06, "loss": 0.8156, "step": 14592 }, { "epoch": 0.74, "learning_rate": 3.308123324379008e-06, "loss": 0.9118, "step": 14593 }, { "epoch": 0.74, "learning_rate": 3.3069007171758673e-06, "loss": 0.866, "step": 14594 }, { "epoch": 0.74, "learning_rate": 3.305678291181267e-06, "loss": 0.8711, "step": 14595 }, { "epoch": 0.74, "learning_rate": 3.3044560464283038e-06, "loss": 0.9664, "step": 14596 }, { "epoch": 0.74, "learning_rate": 3.3032339829500726e-06, "loss": 0.8135, "step": 14597 }, { "epoch": 0.74, "learning_rate": 3.3020121007796533e-06, "loss": 0.8287, "step": 14598 }, { "epoch": 0.74, "learning_rate": 3.300790399950131e-06, "loss": 0.9457, "step": 14599 }, { "epoch": 0.74, "learning_rate": 3.299568880494581e-06, "loss": 0.8852, "step": 14600 }, { "epoch": 0.74, "learning_rate": 3.2983475424460786e-06, "loss": 0.8658, "step": 14601 }, { "epoch": 0.74, "learning_rate": 3.297126385837682e-06, "loss": 0.9357, "step": 14602 }, { "epoch": 0.74, "learning_rate": 3.2959054107024592e-06, "loss": 0.9404, "step": 14603 }, { "epoch": 0.74, "learning_rate": 3.2946846170734657e-06, "loss": 0.9153, "step": 14604 }, { "epoch": 0.74, "learning_rate": 3.2934640049837472e-06, "loss": 0.9481, "step": 14605 }, { "epoch": 0.74, "learning_rate": 3.292243574466362e-06, "loss": 0.7855, "step": 14606 }, { "epoch": 0.74, "learning_rate": 3.291023325554342e-06, "loss": 0.8886, "step": 14607 }, { "epoch": 0.74, "learning_rate": 3.2898032582807283e-06, "loss": 0.9074, "step": 14608 }, { "epoch": 0.74, "learning_rate": 3.2885833726785543e-06, "loss": 0.8392, "step": 14609 }, { "epoch": 0.74, "learning_rate": 3.287363668780843e-06, "loss": 0.9404, "step": 14610 }, { "epoch": 0.74, "learning_rate": 3.2861441466206177e-06, "loss": 0.8414, "step": 14611 }, { "epoch": 0.74, "learning_rate": 3.284924806230897e-06, "loss": 0.7451, "step": 14612 }, { "epoch": 0.74, "learning_rate": 3.283705647644697e-06, "loss": 0.7466, "step": 14613 }, { "epoch": 0.74, "learning_rate": 3.2824866708950166e-06, "loss": 0.8645, "step": 14614 }, { "epoch": 0.74, "learning_rate": 3.281267876014863e-06, "loss": 0.7372, "step": 14615 }, { "epoch": 0.74, "learning_rate": 3.2800492630372373e-06, "loss": 0.8747, "step": 14616 }, { "epoch": 0.74, "learning_rate": 3.2788308319951214e-06, "loss": 0.8926, "step": 14617 }, { "epoch": 0.74, "learning_rate": 3.2776125829215167e-06, "loss": 0.8906, "step": 14618 }, { "epoch": 0.74, "learning_rate": 3.276394515849396e-06, "loss": 0.7755, "step": 14619 }, { "epoch": 0.74, "learning_rate": 3.2751766308117406e-06, "loss": 0.7924, "step": 14620 }, { "epoch": 0.74, "learning_rate": 3.2739589278415252e-06, "loss": 0.8158, "step": 14621 }, { "epoch": 0.74, "learning_rate": 3.2727414069717146e-06, "loss": 0.8293, "step": 14622 }, { "epoch": 0.74, "learning_rate": 3.271524068235271e-06, "loss": 0.8602, "step": 14623 }, { "epoch": 0.74, "learning_rate": 3.2703069116651564e-06, "loss": 0.9499, "step": 14624 }, { "epoch": 0.74, "learning_rate": 3.2690899372943253e-06, "loss": 0.885, "step": 14625 }, { "epoch": 0.74, "learning_rate": 3.2678731451557198e-06, "loss": 0.8895, "step": 14626 }, { "epoch": 0.74, "learning_rate": 3.2666565352822856e-06, "loss": 0.7961, "step": 14627 }, { "epoch": 0.74, "learning_rate": 3.2654401077069663e-06, "loss": 0.8957, "step": 14628 }, { "epoch": 0.74, "learning_rate": 3.264223862462684e-06, "loss": 0.7481, "step": 14629 }, { "epoch": 0.74, "learning_rate": 3.2630077995823817e-06, "loss": 0.8268, "step": 14630 }, { "epoch": 0.74, "learning_rate": 3.261791919098971e-06, "loss": 0.9117, "step": 14631 }, { "epoch": 0.74, "learning_rate": 3.260576221045376e-06, "loss": 0.8857, "step": 14632 }, { "epoch": 0.74, "learning_rate": 3.259360705454513e-06, "loss": 0.8947, "step": 14633 }, { "epoch": 0.74, "learning_rate": 3.258145372359284e-06, "loss": 0.8558, "step": 14634 }, { "epoch": 0.74, "learning_rate": 3.2569302217925958e-06, "loss": 0.8956, "step": 14635 }, { "epoch": 0.74, "learning_rate": 3.2557152537873482e-06, "loss": 0.8835, "step": 14636 }, { "epoch": 0.74, "learning_rate": 3.254500468376438e-06, "loss": 0.8713, "step": 14637 }, { "epoch": 0.74, "learning_rate": 3.253285865592748e-06, "loss": 0.9062, "step": 14638 }, { "epoch": 0.74, "learning_rate": 3.252071445469165e-06, "loss": 0.8501, "step": 14639 }, { "epoch": 0.74, "learning_rate": 3.2508572080385716e-06, "loss": 0.8583, "step": 14640 }, { "epoch": 0.74, "learning_rate": 3.2496431533338325e-06, "loss": 0.8533, "step": 14641 }, { "epoch": 0.74, "learning_rate": 3.248429281387829e-06, "loss": 0.8976, "step": 14642 }, { "epoch": 0.74, "learning_rate": 3.247215592233417e-06, "loss": 0.8796, "step": 14643 }, { "epoch": 0.74, "learning_rate": 3.246002085903458e-06, "loss": 0.7933, "step": 14644 }, { "epoch": 0.74, "learning_rate": 3.2447887624308106e-06, "loss": 0.9305, "step": 14645 }, { "epoch": 0.74, "learning_rate": 3.243575621848317e-06, "loss": 0.8429, "step": 14646 }, { "epoch": 0.74, "learning_rate": 3.242362664188825e-06, "loss": 0.8825, "step": 14647 }, { "epoch": 0.74, "learning_rate": 3.2411498894851743e-06, "loss": 0.8309, "step": 14648 }, { "epoch": 0.74, "learning_rate": 3.2399372977702026e-06, "loss": 0.8778, "step": 14649 }, { "epoch": 0.74, "learning_rate": 3.2387248890767344e-06, "loss": 0.8869, "step": 14650 }, { "epoch": 0.74, "learning_rate": 3.2375126634375952e-06, "loss": 0.8762, "step": 14651 }, { "epoch": 0.74, "learning_rate": 3.2363006208856064e-06, "loss": 0.8969, "step": 14652 }, { "epoch": 0.74, "learning_rate": 3.235088761453581e-06, "loss": 0.8088, "step": 14653 }, { "epoch": 0.74, "learning_rate": 3.233877085174335e-06, "loss": 0.8528, "step": 14654 }, { "epoch": 0.74, "learning_rate": 3.232665592080664e-06, "loss": 0.9095, "step": 14655 }, { "epoch": 0.74, "learning_rate": 3.2314542822053727e-06, "loss": 0.927, "step": 14656 }, { "epoch": 0.74, "learning_rate": 3.2302431555812585e-06, "loss": 0.8871, "step": 14657 }, { "epoch": 0.74, "learning_rate": 3.2290322122411023e-06, "loss": 0.8412, "step": 14658 }, { "epoch": 0.74, "learning_rate": 3.227821452217702e-06, "loss": 0.8606, "step": 14659 }, { "epoch": 0.74, "learning_rate": 3.2266108755438277e-06, "loss": 0.8807, "step": 14660 }, { "epoch": 0.74, "learning_rate": 3.2254004822522565e-06, "loss": 0.8875, "step": 14661 }, { "epoch": 0.74, "learning_rate": 3.2241902723757636e-06, "loss": 0.9773, "step": 14662 }, { "epoch": 0.74, "learning_rate": 3.222980245947107e-06, "loss": 0.8495, "step": 14663 }, { "epoch": 0.74, "learning_rate": 3.2217704029990493e-06, "loss": 0.883, "step": 14664 }, { "epoch": 0.75, "learning_rate": 3.2205607435643473e-06, "loss": 0.9089, "step": 14665 }, { "epoch": 0.75, "learning_rate": 3.219351267675753e-06, "loss": 0.8826, "step": 14666 }, { "epoch": 0.75, "learning_rate": 3.2181419753660072e-06, "loss": 0.7891, "step": 14667 }, { "epoch": 0.75, "learning_rate": 3.2169328666678512e-06, "loss": 0.9344, "step": 14668 }, { "epoch": 0.75, "learning_rate": 3.215723941614024e-06, "loss": 0.9808, "step": 14669 }, { "epoch": 0.75, "learning_rate": 3.214515200237247e-06, "loss": 0.7761, "step": 14670 }, { "epoch": 0.75, "learning_rate": 3.2133066425702585e-06, "loss": 0.8354, "step": 14671 }, { "epoch": 0.75, "learning_rate": 3.2120982686457692e-06, "loss": 0.8987, "step": 14672 }, { "epoch": 0.75, "learning_rate": 3.210890078496497e-06, "loss": 0.7743, "step": 14673 }, { "epoch": 0.75, "learning_rate": 3.209682072155156e-06, "loss": 0.8365, "step": 14674 }, { "epoch": 0.75, "learning_rate": 3.208474249654445e-06, "loss": 0.8195, "step": 14675 }, { "epoch": 0.75, "learning_rate": 3.207266611027069e-06, "loss": 0.8956, "step": 14676 }, { "epoch": 0.75, "learning_rate": 3.206059156305723e-06, "loss": 0.8724, "step": 14677 }, { "epoch": 0.75, "learning_rate": 3.204851885523099e-06, "loss": 0.8456, "step": 14678 }, { "epoch": 0.75, "learning_rate": 3.203644798711878e-06, "loss": 0.8245, "step": 14679 }, { "epoch": 0.75, "learning_rate": 3.2024378959047442e-06, "loss": 0.9189, "step": 14680 }, { "epoch": 0.75, "learning_rate": 3.201231177134374e-06, "loss": 0.8983, "step": 14681 }, { "epoch": 0.75, "learning_rate": 3.2000246424334315e-06, "loss": 0.8391, "step": 14682 }, { "epoch": 0.75, "learning_rate": 3.1988182918345923e-06, "loss": 0.8316, "step": 14683 }, { "epoch": 0.75, "learning_rate": 3.1976121253705095e-06, "loss": 0.8151, "step": 14684 }, { "epoch": 0.75, "learning_rate": 3.1964061430738413e-06, "loss": 0.7917, "step": 14685 }, { "epoch": 0.75, "learning_rate": 3.1952003449772417e-06, "loss": 0.874, "step": 14686 }, { "epoch": 0.75, "learning_rate": 3.193994731113349e-06, "loss": 0.9054, "step": 14687 }, { "epoch": 0.75, "learning_rate": 3.192789301514808e-06, "loss": 0.8412, "step": 14688 }, { "epoch": 0.75, "learning_rate": 3.191584056214255e-06, "loss": 0.9477, "step": 14689 }, { "epoch": 0.75, "learning_rate": 3.190378995244323e-06, "loss": 0.9403, "step": 14690 }, { "epoch": 0.75, "learning_rate": 3.1891741186376314e-06, "loss": 0.8704, "step": 14691 }, { "epoch": 0.75, "learning_rate": 3.1879694264268047e-06, "loss": 0.9192, "step": 14692 }, { "epoch": 0.75, "learning_rate": 3.1867649186444618e-06, "loss": 0.8934, "step": 14693 }, { "epoch": 0.75, "learning_rate": 3.1855605953232027e-06, "loss": 0.8952, "step": 14694 }, { "epoch": 0.75, "learning_rate": 3.1843564564956477e-06, "loss": 0.8966, "step": 14695 }, { "epoch": 0.75, "learning_rate": 3.183152502194388e-06, "loss": 0.8745, "step": 14696 }, { "epoch": 0.75, "learning_rate": 3.1819487324520193e-06, "loss": 0.8678, "step": 14697 }, { "epoch": 0.75, "learning_rate": 3.1807451473011396e-06, "loss": 0.8973, "step": 14698 }, { "epoch": 0.75, "learning_rate": 3.1795417467743273e-06, "loss": 0.8409, "step": 14699 }, { "epoch": 0.75, "learning_rate": 3.178338530904165e-06, "loss": 0.8203, "step": 14700 }, { "epoch": 0.75, "learning_rate": 3.1771354997232282e-06, "loss": 0.8606, "step": 14701 }, { "epoch": 0.75, "learning_rate": 3.1759326532640944e-06, "loss": 0.907, "step": 14702 }, { "epoch": 0.75, "learning_rate": 3.1747299915593186e-06, "loss": 0.8619, "step": 14703 }, { "epoch": 0.75, "learning_rate": 3.173527514641468e-06, "loss": 0.8919, "step": 14704 }, { "epoch": 0.75, "learning_rate": 3.1723252225431e-06, "loss": 0.8571, "step": 14705 }, { "epoch": 0.75, "learning_rate": 3.171123115296756e-06, "loss": 0.8591, "step": 14706 }, { "epoch": 0.75, "learning_rate": 3.1699211929349936e-06, "loss": 0.8973, "step": 14707 }, { "epoch": 0.75, "learning_rate": 3.1687194554903467e-06, "loss": 0.9415, "step": 14708 }, { "epoch": 0.75, "learning_rate": 3.167517902995352e-06, "loss": 0.8963, "step": 14709 }, { "epoch": 0.75, "learning_rate": 3.166316535482544e-06, "loss": 0.9556, "step": 14710 }, { "epoch": 0.75, "learning_rate": 3.1651153529844426e-06, "loss": 0.9083, "step": 14711 }, { "epoch": 0.75, "learning_rate": 3.1639143555335706e-06, "loss": 0.9122, "step": 14712 }, { "epoch": 0.75, "learning_rate": 3.1627135431624443e-06, "loss": 0.8659, "step": 14713 }, { "epoch": 0.75, "learning_rate": 3.1615129159035784e-06, "loss": 0.8701, "step": 14714 }, { "epoch": 0.75, "learning_rate": 3.160312473789472e-06, "loss": 0.8192, "step": 14715 }, { "epoch": 0.75, "learning_rate": 3.159112216852628e-06, "loss": 0.8199, "step": 14716 }, { "epoch": 0.75, "learning_rate": 3.1579121451255457e-06, "loss": 0.9384, "step": 14717 }, { "epoch": 0.75, "learning_rate": 3.156712258640705e-06, "loss": 0.8477, "step": 14718 }, { "epoch": 0.75, "learning_rate": 3.1555125574306077e-06, "loss": 0.8862, "step": 14719 }, { "epoch": 0.75, "learning_rate": 3.1543130415277212e-06, "loss": 0.9296, "step": 14720 }, { "epoch": 0.75, "learning_rate": 3.1531137109645273e-06, "loss": 0.9037, "step": 14721 }, { "epoch": 0.75, "learning_rate": 3.1519145657734974e-06, "loss": 0.8866, "step": 14722 }, { "epoch": 0.75, "learning_rate": 3.150715605987088e-06, "loss": 0.8552, "step": 14723 }, { "epoch": 0.75, "learning_rate": 3.1495168316377733e-06, "loss": 0.8607, "step": 14724 }, { "epoch": 0.75, "learning_rate": 3.1483182427579995e-06, "loss": 0.8085, "step": 14725 }, { "epoch": 0.75, "learning_rate": 3.1471198393802195e-06, "loss": 0.91, "step": 14726 }, { "epoch": 0.75, "learning_rate": 3.1459216215368815e-06, "loss": 0.8453, "step": 14727 }, { "epoch": 0.75, "learning_rate": 3.144723589260421e-06, "loss": 0.8085, "step": 14728 }, { "epoch": 0.75, "learning_rate": 3.1435257425832754e-06, "loss": 0.9693, "step": 14729 }, { "epoch": 0.75, "learning_rate": 3.1423280815378753e-06, "loss": 0.9414, "step": 14730 }, { "epoch": 0.75, "learning_rate": 3.1411306061566504e-06, "loss": 0.8828, "step": 14731 }, { "epoch": 0.75, "learning_rate": 3.1399333164720137e-06, "loss": 0.8694, "step": 14732 }, { "epoch": 0.75, "learning_rate": 3.1387362125163832e-06, "loss": 0.8863, "step": 14733 }, { "epoch": 0.75, "learning_rate": 3.1375392943221737e-06, "loss": 0.9151, "step": 14734 }, { "epoch": 0.75, "learning_rate": 3.1363425619217803e-06, "loss": 0.8409, "step": 14735 }, { "epoch": 0.75, "learning_rate": 3.1351460153476156e-06, "loss": 0.9128, "step": 14736 }, { "epoch": 0.75, "learning_rate": 3.133949654632067e-06, "loss": 0.9217, "step": 14737 }, { "epoch": 0.75, "learning_rate": 3.1327534798075254e-06, "loss": 0.9167, "step": 14738 }, { "epoch": 0.75, "learning_rate": 3.1315574909063807e-06, "loss": 0.8572, "step": 14739 }, { "epoch": 0.75, "learning_rate": 3.1303616879610056e-06, "loss": 0.9, "step": 14740 }, { "epoch": 0.75, "learning_rate": 3.1291660710037798e-06, "loss": 0.9029, "step": 14741 }, { "epoch": 0.75, "learning_rate": 3.127970640067073e-06, "loss": 0.797, "step": 14742 }, { "epoch": 0.75, "learning_rate": 3.1267753951832523e-06, "loss": 0.8428, "step": 14743 }, { "epoch": 0.75, "learning_rate": 3.125580336384673e-06, "loss": 0.9348, "step": 14744 }, { "epoch": 0.75, "learning_rate": 3.124385463703691e-06, "loss": 0.9838, "step": 14745 }, { "epoch": 0.75, "learning_rate": 3.1231907771726612e-06, "loss": 0.8256, "step": 14746 }, { "epoch": 0.75, "learning_rate": 3.1219962768239187e-06, "loss": 1.0, "step": 14747 }, { "epoch": 0.75, "learning_rate": 3.1208019626898155e-06, "loss": 0.7583, "step": 14748 }, { "epoch": 0.75, "learning_rate": 3.1196078348026772e-06, "loss": 0.8748, "step": 14749 }, { "epoch": 0.75, "learning_rate": 3.118413893194837e-06, "loss": 0.8963, "step": 14750 }, { "epoch": 0.75, "learning_rate": 3.1172201378986224e-06, "loss": 0.9146, "step": 14751 }, { "epoch": 0.75, "learning_rate": 3.116026568946345e-06, "loss": 0.8521, "step": 14752 }, { "epoch": 0.75, "learning_rate": 3.1148331863703263e-06, "loss": 0.8557, "step": 14753 }, { "epoch": 0.75, "learning_rate": 3.1136399902028726e-06, "loss": 0.9305, "step": 14754 }, { "epoch": 0.75, "learning_rate": 3.1124469804762936e-06, "loss": 0.8285, "step": 14755 }, { "epoch": 0.75, "learning_rate": 3.1112541572228817e-06, "loss": 0.8605, "step": 14756 }, { "epoch": 0.75, "learning_rate": 3.110061520474934e-06, "loss": 0.9529, "step": 14757 }, { "epoch": 0.75, "learning_rate": 3.1088690702647438e-06, "loss": 1.0479, "step": 14758 }, { "epoch": 0.75, "learning_rate": 3.1076768066245854e-06, "loss": 0.8839, "step": 14759 }, { "epoch": 0.75, "learning_rate": 3.1064847295867505e-06, "loss": 0.863, "step": 14760 }, { "epoch": 0.75, "learning_rate": 3.1052928391835047e-06, "loss": 0.8945, "step": 14761 }, { "epoch": 0.75, "learning_rate": 3.1041011354471195e-06, "loss": 0.8415, "step": 14762 }, { "epoch": 0.75, "learning_rate": 3.1029096184098626e-06, "loss": 0.8186, "step": 14763 }, { "epoch": 0.75, "learning_rate": 3.1017182881039874e-06, "loss": 0.8024, "step": 14764 }, { "epoch": 0.75, "learning_rate": 3.10052714456175e-06, "loss": 0.8632, "step": 14765 }, { "epoch": 0.75, "learning_rate": 3.0993361878154003e-06, "loss": 0.8609, "step": 14766 }, { "epoch": 0.75, "learning_rate": 3.0981454178971836e-06, "loss": 0.855, "step": 14767 }, { "epoch": 0.75, "learning_rate": 3.0969548348393343e-06, "loss": 0.9592, "step": 14768 }, { "epoch": 0.75, "learning_rate": 3.0957644386740883e-06, "loss": 0.8886, "step": 14769 }, { "epoch": 0.75, "learning_rate": 3.094574229433678e-06, "loss": 0.8793, "step": 14770 }, { "epoch": 0.75, "learning_rate": 3.0933842071503173e-06, "loss": 0.8793, "step": 14771 }, { "epoch": 0.75, "learning_rate": 3.0921943718562375e-06, "loss": 0.7778, "step": 14772 }, { "epoch": 0.75, "learning_rate": 3.0910047235836417e-06, "loss": 0.7763, "step": 14773 }, { "epoch": 0.75, "learning_rate": 3.0898152623647435e-06, "loss": 0.9835, "step": 14774 }, { "epoch": 0.75, "learning_rate": 3.088625988231748e-06, "loss": 0.8714, "step": 14775 }, { "epoch": 0.75, "learning_rate": 3.087436901216847e-06, "loss": 0.8911, "step": 14776 }, { "epoch": 0.75, "learning_rate": 3.086248001352239e-06, "loss": 0.9459, "step": 14777 }, { "epoch": 0.75, "learning_rate": 3.085059288670109e-06, "loss": 0.8002, "step": 14778 }, { "epoch": 0.75, "learning_rate": 3.0838707632026456e-06, "loss": 0.8513, "step": 14779 }, { "epoch": 0.75, "learning_rate": 3.0826824249820207e-06, "loss": 0.8846, "step": 14780 }, { "epoch": 0.75, "learning_rate": 3.0814942740404084e-06, "loss": 0.8744, "step": 14781 }, { "epoch": 0.75, "learning_rate": 3.0803063104099828e-06, "loss": 0.8796, "step": 14782 }, { "epoch": 0.75, "learning_rate": 3.0791185341228935e-06, "loss": 0.9444, "step": 14783 }, { "epoch": 0.75, "learning_rate": 3.077930945211315e-06, "loss": 0.9834, "step": 14784 }, { "epoch": 0.75, "learning_rate": 3.0767435437073866e-06, "loss": 0.8248, "step": 14785 }, { "epoch": 0.75, "learning_rate": 3.0755563296432624e-06, "loss": 0.8753, "step": 14786 }, { "epoch": 0.75, "learning_rate": 3.0743693030510857e-06, "loss": 0.9562, "step": 14787 }, { "epoch": 0.75, "learning_rate": 3.0731824639629893e-06, "loss": 0.8911, "step": 14788 }, { "epoch": 0.75, "learning_rate": 3.0719958124111083e-06, "loss": 0.8573, "step": 14789 }, { "epoch": 0.75, "learning_rate": 3.070809348427569e-06, "loss": 0.8835, "step": 14790 }, { "epoch": 0.75, "learning_rate": 3.069623072044495e-06, "loss": 0.9216, "step": 14791 }, { "epoch": 0.75, "learning_rate": 3.0684369832940074e-06, "loss": 0.9362, "step": 14792 }, { "epoch": 0.75, "learning_rate": 3.0672510822082092e-06, "loss": 0.8592, "step": 14793 }, { "epoch": 0.75, "learning_rate": 3.066065368819212e-06, "loss": 0.9406, "step": 14794 }, { "epoch": 0.75, "learning_rate": 3.064879843159119e-06, "loss": 0.8617, "step": 14795 }, { "epoch": 0.75, "learning_rate": 3.0636945052600275e-06, "loss": 0.8414, "step": 14796 }, { "epoch": 0.75, "learning_rate": 3.0625093551540254e-06, "loss": 0.8857, "step": 14797 }, { "epoch": 0.75, "learning_rate": 3.061324392873202e-06, "loss": 0.8234, "step": 14798 }, { "epoch": 0.75, "learning_rate": 3.060139618449641e-06, "loss": 0.9238, "step": 14799 }, { "epoch": 0.75, "learning_rate": 3.0589550319154104e-06, "loss": 0.9206, "step": 14800 }, { "epoch": 0.75, "learning_rate": 3.0577706333025935e-06, "loss": 0.849, "step": 14801 }, { "epoch": 0.75, "learning_rate": 3.056586422643247e-06, "loss": 0.911, "step": 14802 }, { "epoch": 0.75, "learning_rate": 3.055402399969437e-06, "loss": 1.0098, "step": 14803 }, { "epoch": 0.75, "learning_rate": 3.0542185653132216e-06, "loss": 0.848, "step": 14804 }, { "epoch": 0.75, "learning_rate": 3.0530349187066454e-06, "loss": 0.9058, "step": 14805 }, { "epoch": 0.75, "learning_rate": 3.0518514601817585e-06, "loss": 0.8941, "step": 14806 }, { "epoch": 0.75, "learning_rate": 3.0506681897706005e-06, "loss": 0.8964, "step": 14807 }, { "epoch": 0.75, "learning_rate": 3.04948510750521e-06, "loss": 0.8996, "step": 14808 }, { "epoch": 0.75, "learning_rate": 3.048302213417613e-06, "loss": 0.8842, "step": 14809 }, { "epoch": 0.75, "learning_rate": 3.0471195075398384e-06, "loss": 0.953, "step": 14810 }, { "epoch": 0.75, "learning_rate": 3.045936989903908e-06, "loss": 0.9463, "step": 14811 }, { "epoch": 0.75, "learning_rate": 3.0447546605418287e-06, "loss": 0.9308, "step": 14812 }, { "epoch": 0.75, "learning_rate": 3.043572519485624e-06, "loss": 0.8515, "step": 14813 }, { "epoch": 0.75, "learning_rate": 3.042390566767289e-06, "loss": 0.7667, "step": 14814 }, { "epoch": 0.75, "learning_rate": 3.041208802418827e-06, "loss": 0.8411, "step": 14815 }, { "epoch": 0.75, "learning_rate": 3.0400272264722365e-06, "loss": 0.9506, "step": 14816 }, { "epoch": 0.75, "learning_rate": 3.0388458389595e-06, "loss": 0.8763, "step": 14817 }, { "epoch": 0.75, "learning_rate": 3.0376646399126076e-06, "loss": 0.9047, "step": 14818 }, { "epoch": 0.75, "learning_rate": 3.036483629363537e-06, "loss": 0.9274, "step": 14819 }, { "epoch": 0.75, "learning_rate": 3.035302807344267e-06, "loss": 0.7893, "step": 14820 }, { "epoch": 0.75, "learning_rate": 3.03412217388676e-06, "loss": 0.7916, "step": 14821 }, { "epoch": 0.75, "learning_rate": 3.0329417290229845e-06, "loss": 0.875, "step": 14822 }, { "epoch": 0.75, "learning_rate": 3.031761472784902e-06, "loss": 0.9381, "step": 14823 }, { "epoch": 0.75, "learning_rate": 3.030581405204458e-06, "loss": 0.8841, "step": 14824 }, { "epoch": 0.75, "learning_rate": 3.0294015263136135e-06, "loss": 0.9859, "step": 14825 }, { "epoch": 0.75, "learning_rate": 3.028221836144304e-06, "loss": 0.9877, "step": 14826 }, { "epoch": 0.75, "learning_rate": 3.0270423347284704e-06, "loss": 0.8781, "step": 14827 }, { "epoch": 0.75, "learning_rate": 3.02586302209805e-06, "loss": 0.8704, "step": 14828 }, { "epoch": 0.75, "learning_rate": 3.024683898284966e-06, "loss": 0.8998, "step": 14829 }, { "epoch": 0.75, "learning_rate": 3.0235049633211443e-06, "loss": 0.9151, "step": 14830 }, { "epoch": 0.75, "learning_rate": 3.0223262172385024e-06, "loss": 0.8343, "step": 14831 }, { "epoch": 0.75, "learning_rate": 3.021147660068958e-06, "loss": 0.9698, "step": 14832 }, { "epoch": 0.75, "learning_rate": 3.0199692918444135e-06, "loss": 0.8734, "step": 14833 }, { "epoch": 0.75, "learning_rate": 3.018791112596773e-06, "loss": 0.8623, "step": 14834 }, { "epoch": 0.75, "learning_rate": 3.0176131223579397e-06, "loss": 0.9503, "step": 14835 }, { "epoch": 0.75, "learning_rate": 3.016435321159795e-06, "loss": 0.9474, "step": 14836 }, { "epoch": 0.75, "learning_rate": 3.0152577090342417e-06, "loss": 0.902, "step": 14837 }, { "epoch": 0.75, "learning_rate": 3.0140802860131514e-06, "loss": 0.8559, "step": 14838 }, { "epoch": 0.75, "learning_rate": 3.0129030521284043e-06, "loss": 0.8921, "step": 14839 }, { "epoch": 0.75, "learning_rate": 3.011726007411877e-06, "loss": 0.8673, "step": 14840 }, { "epoch": 0.75, "learning_rate": 3.0105491518954298e-06, "loss": 0.8777, "step": 14841 }, { "epoch": 0.75, "learning_rate": 3.0093724856109276e-06, "loss": 0.8539, "step": 14842 }, { "epoch": 0.75, "learning_rate": 3.0081960085902284e-06, "loss": 0.8535, "step": 14843 }, { "epoch": 0.75, "learning_rate": 3.0070197208651875e-06, "loss": 0.7876, "step": 14844 }, { "epoch": 0.75, "learning_rate": 3.0058436224676436e-06, "loss": 0.8373, "step": 14845 }, { "epoch": 0.75, "learning_rate": 3.004667713429442e-06, "loss": 0.8765, "step": 14846 }, { "epoch": 0.75, "learning_rate": 3.0034919937824236e-06, "loss": 0.8022, "step": 14847 }, { "epoch": 0.75, "learning_rate": 3.002316463558409e-06, "loss": 0.953, "step": 14848 }, { "epoch": 0.75, "learning_rate": 3.0011411227892384e-06, "loss": 0.8659, "step": 14849 }, { "epoch": 0.75, "learning_rate": 2.9999659715067218e-06, "loss": 0.892, "step": 14850 }, { "epoch": 0.75, "learning_rate": 2.9987910097426788e-06, "loss": 0.8424, "step": 14851 }, { "epoch": 0.75, "learning_rate": 2.997616237528924e-06, "loss": 0.8327, "step": 14852 }, { "epoch": 0.75, "learning_rate": 2.9964416548972564e-06, "loss": 0.9046, "step": 14853 }, { "epoch": 0.75, "learning_rate": 2.99526726187948e-06, "loss": 0.9008, "step": 14854 }, { "epoch": 0.75, "learning_rate": 2.9940930585073904e-06, "loss": 0.8992, "step": 14855 }, { "epoch": 0.75, "learning_rate": 2.99291904481278e-06, "loss": 0.9343, "step": 14856 }, { "epoch": 0.75, "learning_rate": 2.991745220827429e-06, "loss": 0.8276, "step": 14857 }, { "epoch": 0.75, "learning_rate": 2.9905715865831185e-06, "loss": 0.8258, "step": 14858 }, { "epoch": 0.75, "learning_rate": 2.9893981421116258e-06, "loss": 0.9028, "step": 14859 }, { "epoch": 0.75, "learning_rate": 2.9882248874447205e-06, "loss": 0.9452, "step": 14860 }, { "epoch": 0.75, "learning_rate": 2.987051822614169e-06, "loss": 0.8849, "step": 14861 }, { "epoch": 0.76, "learning_rate": 2.9858789476517246e-06, "loss": 0.8235, "step": 14862 }, { "epoch": 0.76, "learning_rate": 2.9847062625891453e-06, "loss": 0.9335, "step": 14863 }, { "epoch": 0.76, "learning_rate": 2.9835337674581844e-06, "loss": 0.9653, "step": 14864 }, { "epoch": 0.76, "learning_rate": 2.982361462290575e-06, "loss": 0.8793, "step": 14865 }, { "epoch": 0.76, "learning_rate": 2.9811893471180687e-06, "loss": 0.8596, "step": 14866 }, { "epoch": 0.76, "learning_rate": 2.98001742197239e-06, "loss": 0.8482, "step": 14867 }, { "epoch": 0.76, "learning_rate": 2.978845686885271e-06, "loss": 0.8954, "step": 14868 }, { "epoch": 0.76, "learning_rate": 2.9776741418884393e-06, "loss": 0.8927, "step": 14869 }, { "epoch": 0.76, "learning_rate": 2.9765027870136053e-06, "loss": 0.8598, "step": 14870 }, { "epoch": 0.76, "learning_rate": 2.9753316222924854e-06, "loss": 0.9041, "step": 14871 }, { "epoch": 0.76, "learning_rate": 2.9741606477567887e-06, "loss": 0.8796, "step": 14872 }, { "epoch": 0.76, "learning_rate": 2.9729898634382213e-06, "loss": 0.7735, "step": 14873 }, { "epoch": 0.76, "learning_rate": 2.9718192693684733e-06, "loss": 0.9225, "step": 14874 }, { "epoch": 0.76, "learning_rate": 2.970648865579241e-06, "loss": 0.8582, "step": 14875 }, { "epoch": 0.76, "learning_rate": 2.9694786521022155e-06, "loss": 0.9393, "step": 14876 }, { "epoch": 0.76, "learning_rate": 2.96830862896907e-06, "loss": 0.845, "step": 14877 }, { "epoch": 0.76, "learning_rate": 2.9671387962114928e-06, "loss": 0.8803, "step": 14878 }, { "epoch": 0.76, "learning_rate": 2.965969153861147e-06, "loss": 0.9003, "step": 14879 }, { "epoch": 0.76, "learning_rate": 2.9647997019497023e-06, "loss": 0.952, "step": 14880 }, { "epoch": 0.76, "learning_rate": 2.9636304405088247e-06, "loss": 0.8669, "step": 14881 }, { "epoch": 0.76, "learning_rate": 2.962461369570164e-06, "loss": 0.8705, "step": 14882 }, { "epoch": 0.76, "learning_rate": 2.961292489165374e-06, "loss": 0.9012, "step": 14883 }, { "epoch": 0.76, "learning_rate": 2.960123799326101e-06, "loss": 0.835, "step": 14884 }, { "epoch": 0.76, "learning_rate": 2.95895530008399e-06, "loss": 0.9549, "step": 14885 }, { "epoch": 0.76, "learning_rate": 2.9577869914706693e-06, "loss": 0.851, "step": 14886 }, { "epoch": 0.76, "learning_rate": 2.956618873517774e-06, "loss": 0.865, "step": 14887 }, { "epoch": 0.76, "learning_rate": 2.9554509462569314e-06, "loss": 0.9521, "step": 14888 }, { "epoch": 0.76, "learning_rate": 2.9542832097197538e-06, "loss": 0.8803, "step": 14889 }, { "epoch": 0.76, "learning_rate": 2.9531156639378687e-06, "loss": 0.8723, "step": 14890 }, { "epoch": 0.76, "learning_rate": 2.951948308942877e-06, "loss": 0.9238, "step": 14891 }, { "epoch": 0.76, "learning_rate": 2.950781144766386e-06, "loss": 0.8411, "step": 14892 }, { "epoch": 0.76, "learning_rate": 2.9496141714399984e-06, "loss": 0.918, "step": 14893 }, { "epoch": 0.76, "learning_rate": 2.9484473889953035e-06, "loss": 0.9073, "step": 14894 }, { "epoch": 0.76, "learning_rate": 2.947280797463894e-06, "loss": 0.8934, "step": 14895 }, { "epoch": 0.76, "learning_rate": 2.946114396877352e-06, "loss": 0.8065, "step": 14896 }, { "epoch": 0.76, "learning_rate": 2.9449481872672613e-06, "loss": 0.9681, "step": 14897 }, { "epoch": 0.76, "learning_rate": 2.94378216866519e-06, "loss": 0.8611, "step": 14898 }, { "epoch": 0.76, "learning_rate": 2.9426163411027086e-06, "loss": 0.8927, "step": 14899 }, { "epoch": 0.76, "learning_rate": 2.941450704611385e-06, "loss": 0.8604, "step": 14900 }, { "epoch": 0.76, "learning_rate": 2.940285259222767e-06, "loss": 0.8842, "step": 14901 }, { "epoch": 0.76, "learning_rate": 2.939120004968422e-06, "loss": 0.7755, "step": 14902 }, { "epoch": 0.76, "learning_rate": 2.937954941879886e-06, "loss": 0.8977, "step": 14903 }, { "epoch": 0.76, "learning_rate": 2.9367900699887075e-06, "loss": 0.8374, "step": 14904 }, { "epoch": 0.76, "learning_rate": 2.9356253893264266e-06, "loss": 0.8997, "step": 14905 }, { "epoch": 0.76, "learning_rate": 2.9344608999245684e-06, "loss": 0.798, "step": 14906 }, { "epoch": 0.76, "learning_rate": 2.9332966018146646e-06, "loss": 0.9907, "step": 14907 }, { "epoch": 0.76, "learning_rate": 2.932132495028237e-06, "loss": 0.8701, "step": 14908 }, { "epoch": 0.76, "learning_rate": 2.930968579596806e-06, "loss": 0.9, "step": 14909 }, { "epoch": 0.76, "learning_rate": 2.9298048555518752e-06, "loss": 0.7755, "step": 14910 }, { "epoch": 0.76, "learning_rate": 2.928641322924958e-06, "loss": 0.8453, "step": 14911 }, { "epoch": 0.76, "learning_rate": 2.9274779817475553e-06, "loss": 0.772, "step": 14912 }, { "epoch": 0.76, "learning_rate": 2.926314832051156e-06, "loss": 0.9032, "step": 14913 }, { "epoch": 0.76, "learning_rate": 2.9251518738672624e-06, "loss": 0.8326, "step": 14914 }, { "epoch": 0.76, "learning_rate": 2.923989107227352e-06, "loss": 0.8978, "step": 14915 }, { "epoch": 0.76, "learning_rate": 2.922826532162908e-06, "loss": 0.9098, "step": 14916 }, { "epoch": 0.76, "learning_rate": 2.9216641487054098e-06, "loss": 0.8378, "step": 14917 }, { "epoch": 0.76, "learning_rate": 2.9205019568863214e-06, "loss": 0.8661, "step": 14918 }, { "epoch": 0.76, "learning_rate": 2.9193399567371095e-06, "loss": 0.8363, "step": 14919 }, { "epoch": 0.76, "learning_rate": 2.9181781482892344e-06, "loss": 0.9283, "step": 14920 }, { "epoch": 0.76, "learning_rate": 2.917016531574156e-06, "loss": 0.855, "step": 14921 }, { "epoch": 0.76, "learning_rate": 2.9158551066233153e-06, "loss": 0.9132, "step": 14922 }, { "epoch": 0.76, "learning_rate": 2.914693873468162e-06, "loss": 0.8757, "step": 14923 }, { "epoch": 0.76, "learning_rate": 2.913532832140136e-06, "loss": 0.8801, "step": 14924 }, { "epoch": 0.76, "learning_rate": 2.912371982670663e-06, "loss": 0.8909, "step": 14925 }, { "epoch": 0.76, "learning_rate": 2.9112113250911844e-06, "loss": 0.8151, "step": 14926 }, { "epoch": 0.76, "learning_rate": 2.9100508594331134e-06, "loss": 0.871, "step": 14927 }, { "epoch": 0.76, "learning_rate": 2.9088905857278726e-06, "loss": 0.8732, "step": 14928 }, { "epoch": 0.76, "learning_rate": 2.907730504006878e-06, "loss": 0.8326, "step": 14929 }, { "epoch": 0.76, "learning_rate": 2.9065706143015295e-06, "loss": 0.7833, "step": 14930 }, { "epoch": 0.76, "learning_rate": 2.9054109166432397e-06, "loss": 0.7517, "step": 14931 }, { "epoch": 0.76, "learning_rate": 2.9042514110633978e-06, "loss": 0.8914, "step": 14932 }, { "epoch": 0.76, "learning_rate": 2.9030920975934006e-06, "loss": 0.8961, "step": 14933 }, { "epoch": 0.76, "learning_rate": 2.9019329762646363e-06, "loss": 0.9443, "step": 14934 }, { "epoch": 0.76, "learning_rate": 2.900774047108483e-06, "loss": 0.853, "step": 14935 }, { "epoch": 0.76, "learning_rate": 2.8996153101563195e-06, "loss": 0.9068, "step": 14936 }, { "epoch": 0.76, "learning_rate": 2.898456765439517e-06, "loss": 0.848, "step": 14937 }, { "epoch": 0.76, "learning_rate": 2.8972984129894445e-06, "loss": 0.9804, "step": 14938 }, { "epoch": 0.76, "learning_rate": 2.896140252837458e-06, "loss": 0.8543, "step": 14939 }, { "epoch": 0.76, "learning_rate": 2.894982285014917e-06, "loss": 0.8836, "step": 14940 }, { "epoch": 0.76, "learning_rate": 2.8938245095531735e-06, "loss": 0.7868, "step": 14941 }, { "epoch": 0.76, "learning_rate": 2.892666926483565e-06, "loss": 0.9462, "step": 14942 }, { "epoch": 0.76, "learning_rate": 2.8915095358374447e-06, "loss": 0.8394, "step": 14943 }, { "epoch": 0.76, "learning_rate": 2.890352337646137e-06, "loss": 0.9037, "step": 14944 }, { "epoch": 0.76, "learning_rate": 2.8891953319409773e-06, "loss": 0.944, "step": 14945 }, { "epoch": 0.76, "learning_rate": 2.88803851875329e-06, "loss": 0.8257, "step": 14946 }, { "epoch": 0.76, "learning_rate": 2.886881898114391e-06, "loss": 0.8653, "step": 14947 }, { "epoch": 0.76, "learning_rate": 2.885725470055596e-06, "loss": 0.8535, "step": 14948 }, { "epoch": 0.76, "learning_rate": 2.8845692346082153e-06, "loss": 0.883, "step": 14949 }, { "epoch": 0.76, "learning_rate": 2.8834131918035567e-06, "loss": 0.8026, "step": 14950 }, { "epoch": 0.76, "learning_rate": 2.8822573416729106e-06, "loss": 0.8785, "step": 14951 }, { "epoch": 0.76, "learning_rate": 2.8811016842475738e-06, "loss": 0.9296, "step": 14952 }, { "epoch": 0.76, "learning_rate": 2.879946219558838e-06, "loss": 0.8767, "step": 14953 }, { "epoch": 0.76, "learning_rate": 2.8787909476379773e-06, "loss": 0.8334, "step": 14954 }, { "epoch": 0.76, "learning_rate": 2.877635868516282e-06, "loss": 0.8313, "step": 14955 }, { "epoch": 0.76, "learning_rate": 2.876480982225014e-06, "loss": 0.8395, "step": 14956 }, { "epoch": 0.76, "learning_rate": 2.8753262887954447e-06, "loss": 0.8676, "step": 14957 }, { "epoch": 0.76, "learning_rate": 2.8741717882588407e-06, "loss": 0.8139, "step": 14958 }, { "epoch": 0.76, "learning_rate": 2.8730174806464493e-06, "loss": 0.8914, "step": 14959 }, { "epoch": 0.76, "learning_rate": 2.871863365989528e-06, "loss": 0.8524, "step": 14960 }, { "epoch": 0.76, "learning_rate": 2.870709444319322e-06, "loss": 0.8772, "step": 14961 }, { "epoch": 0.76, "learning_rate": 2.869555715667076e-06, "loss": 0.8293, "step": 14962 }, { "epoch": 0.76, "learning_rate": 2.8684021800640194e-06, "loss": 0.8712, "step": 14963 }, { "epoch": 0.76, "learning_rate": 2.8672488375413865e-06, "loss": 0.8899, "step": 14964 }, { "epoch": 0.76, "learning_rate": 2.866095688130406e-06, "loss": 0.8498, "step": 14965 }, { "epoch": 0.76, "learning_rate": 2.8649427318622893e-06, "loss": 0.8402, "step": 14966 }, { "epoch": 0.76, "learning_rate": 2.8637899687682615e-06, "loss": 0.8745, "step": 14967 }, { "epoch": 0.76, "learning_rate": 2.8626373988795262e-06, "loss": 0.8531, "step": 14968 }, { "epoch": 0.76, "learning_rate": 2.8614850222272896e-06, "loss": 0.9375, "step": 14969 }, { "epoch": 0.76, "learning_rate": 2.8603328388427544e-06, "loss": 0.8333, "step": 14970 }, { "epoch": 0.76, "learning_rate": 2.859180848757108e-06, "loss": 0.8614, "step": 14971 }, { "epoch": 0.76, "learning_rate": 2.8580290520015432e-06, "loss": 0.8128, "step": 14972 }, { "epoch": 0.76, "learning_rate": 2.8568774486072426e-06, "loss": 0.9293, "step": 14973 }, { "epoch": 0.76, "learning_rate": 2.85572603860539e-06, "loss": 0.8674, "step": 14974 }, { "epoch": 0.76, "learning_rate": 2.8545748220271496e-06, "loss": 0.8154, "step": 14975 }, { "epoch": 0.76, "learning_rate": 2.8534237989036937e-06, "loss": 0.84, "step": 14976 }, { "epoch": 0.76, "learning_rate": 2.8522729692661897e-06, "loss": 0.8406, "step": 14977 }, { "epoch": 0.76, "learning_rate": 2.8511223331457825e-06, "loss": 0.8683, "step": 14978 }, { "epoch": 0.76, "learning_rate": 2.8499718905736396e-06, "loss": 0.8481, "step": 14979 }, { "epoch": 0.76, "learning_rate": 2.848821641580898e-06, "loss": 0.886, "step": 14980 }, { "epoch": 0.76, "learning_rate": 2.8476715861987016e-06, "loss": 0.8723, "step": 14981 }, { "epoch": 0.76, "learning_rate": 2.84652172445819e-06, "loss": 0.8531, "step": 14982 }, { "epoch": 0.76, "learning_rate": 2.8453720563904896e-06, "loss": 0.8803, "step": 14983 }, { "epoch": 0.76, "learning_rate": 2.8442225820267287e-06, "loss": 0.8818, "step": 14984 }, { "epoch": 0.76, "learning_rate": 2.8430733013980284e-06, "loss": 0.876, "step": 14985 }, { "epoch": 0.76, "learning_rate": 2.8419242145355076e-06, "loss": 0.8763, "step": 14986 }, { "epoch": 0.76, "learning_rate": 2.8407753214702694e-06, "loss": 0.8571, "step": 14987 }, { "epoch": 0.76, "learning_rate": 2.8396266222334225e-06, "loss": 0.8297, "step": 14988 }, { "epoch": 0.76, "learning_rate": 2.8384781168560693e-06, "loss": 0.886, "step": 14989 }, { "epoch": 0.76, "learning_rate": 2.8373298053692966e-06, "loss": 0.9229, "step": 14990 }, { "epoch": 0.76, "learning_rate": 2.8361816878042046e-06, "loss": 0.7872, "step": 14991 }, { "epoch": 0.76, "learning_rate": 2.835033764191869e-06, "loss": 0.9418, "step": 14992 }, { "epoch": 0.76, "learning_rate": 2.8338860345633714e-06, "loss": 0.9132, "step": 14993 }, { "epoch": 0.76, "learning_rate": 2.8327384989497887e-06, "loss": 0.8372, "step": 14994 }, { "epoch": 0.76, "learning_rate": 2.831591157382182e-06, "loss": 0.9114, "step": 14995 }, { "epoch": 0.76, "learning_rate": 2.8304440098916173e-06, "loss": 1.0332, "step": 14996 }, { "epoch": 0.76, "learning_rate": 2.829297056509155e-06, "loss": 0.8629, "step": 14997 }, { "epoch": 0.76, "learning_rate": 2.828150297265845e-06, "loss": 0.925, "step": 14998 }, { "epoch": 0.76, "learning_rate": 2.827003732192739e-06, "loss": 0.9752, "step": 14999 }, { "epoch": 0.76, "learning_rate": 2.825857361320872e-06, "loss": 0.9099, "step": 15000 }, { "epoch": 0.76, "learning_rate": 2.8247111846812836e-06, "loss": 0.8676, "step": 15001 }, { "epoch": 0.76, "learning_rate": 2.8235652023050065e-06, "loss": 0.9337, "step": 15002 }, { "epoch": 0.76, "learning_rate": 2.8224194142230697e-06, "loss": 0.8787, "step": 15003 }, { "epoch": 0.76, "learning_rate": 2.821273820466488e-06, "loss": 0.9115, "step": 15004 }, { "epoch": 0.76, "learning_rate": 2.8201284210662795e-06, "loss": 0.8784, "step": 15005 }, { "epoch": 0.76, "learning_rate": 2.8189832160534593e-06, "loss": 0.9277, "step": 15006 }, { "epoch": 0.76, "learning_rate": 2.8178382054590215e-06, "loss": 0.7544, "step": 15007 }, { "epoch": 0.76, "learning_rate": 2.816693389313979e-06, "loss": 0.9327, "step": 15008 }, { "epoch": 0.76, "learning_rate": 2.8155487676493176e-06, "loss": 0.9089, "step": 15009 }, { "epoch": 0.76, "learning_rate": 2.81440434049603e-06, "loss": 0.8527, "step": 15010 }, { "epoch": 0.76, "learning_rate": 2.813260107885103e-06, "loss": 0.9186, "step": 15011 }, { "epoch": 0.76, "learning_rate": 2.8121160698475093e-06, "loss": 0.8499, "step": 15012 }, { "epoch": 0.76, "learning_rate": 2.810972226414226e-06, "loss": 0.9301, "step": 15013 }, { "epoch": 0.76, "learning_rate": 2.8098285776162215e-06, "loss": 0.8412, "step": 15014 }, { "epoch": 0.76, "learning_rate": 2.8086851234844603e-06, "loss": 0.8348, "step": 15015 }, { "epoch": 0.76, "learning_rate": 2.8075418640498962e-06, "loss": 0.8915, "step": 15016 }, { "epoch": 0.76, "learning_rate": 2.8063987993434836e-06, "loss": 0.8554, "step": 15017 }, { "epoch": 0.76, "learning_rate": 2.8052559293961744e-06, "loss": 0.9554, "step": 15018 }, { "epoch": 0.76, "learning_rate": 2.8041132542388993e-06, "loss": 0.8206, "step": 15019 }, { "epoch": 0.76, "learning_rate": 2.802970773902608e-06, "loss": 0.9077, "step": 15020 }, { "epoch": 0.76, "learning_rate": 2.801828488418222e-06, "loss": 0.8662, "step": 15021 }, { "epoch": 0.76, "learning_rate": 2.800686397816673e-06, "loss": 0.8452, "step": 15022 }, { "epoch": 0.76, "learning_rate": 2.7995445021288823e-06, "loss": 0.8782, "step": 15023 }, { "epoch": 0.76, "learning_rate": 2.7984028013857613e-06, "loss": 0.8367, "step": 15024 }, { "epoch": 0.76, "learning_rate": 2.797261295618222e-06, "loss": 0.8065, "step": 15025 }, { "epoch": 0.76, "learning_rate": 2.7961199848571705e-06, "loss": 0.8871, "step": 15026 }, { "epoch": 0.76, "learning_rate": 2.7949788691335087e-06, "loss": 0.8773, "step": 15027 }, { "epoch": 0.76, "learning_rate": 2.7938379484781254e-06, "loss": 0.8488, "step": 15028 }, { "epoch": 0.76, "learning_rate": 2.792697222921913e-06, "loss": 0.8721, "step": 15029 }, { "epoch": 0.76, "learning_rate": 2.791556692495759e-06, "loss": 0.9152, "step": 15030 }, { "epoch": 0.76, "learning_rate": 2.7904163572305322e-06, "loss": 0.9398, "step": 15031 }, { "epoch": 0.76, "learning_rate": 2.7892762171571176e-06, "loss": 0.8623, "step": 15032 }, { "epoch": 0.76, "learning_rate": 2.7881362723063743e-06, "loss": 0.7928, "step": 15033 }, { "epoch": 0.76, "learning_rate": 2.786996522709169e-06, "loss": 0.8867, "step": 15034 }, { "epoch": 0.76, "learning_rate": 2.785856968396362e-06, "loss": 0.894, "step": 15035 }, { "epoch": 0.76, "learning_rate": 2.7847176093987992e-06, "loss": 0.908, "step": 15036 }, { "epoch": 0.76, "learning_rate": 2.783578445747331e-06, "loss": 0.8841, "step": 15037 }, { "epoch": 0.76, "learning_rate": 2.782439477472799e-06, "loss": 0.8538, "step": 15038 }, { "epoch": 0.76, "learning_rate": 2.7813007046060414e-06, "loss": 0.914, "step": 15039 }, { "epoch": 0.76, "learning_rate": 2.7801621271778846e-06, "loss": 0.8097, "step": 15040 }, { "epoch": 0.76, "learning_rate": 2.779023745219157e-06, "loss": 0.8781, "step": 15041 }, { "epoch": 0.76, "learning_rate": 2.777885558760682e-06, "loss": 0.8645, "step": 15042 }, { "epoch": 0.76, "learning_rate": 2.7767475678332656e-06, "loss": 0.9081, "step": 15043 }, { "epoch": 0.76, "learning_rate": 2.77560977246773e-06, "loss": 0.8735, "step": 15044 }, { "epoch": 0.76, "learning_rate": 2.774472172694871e-06, "loss": 0.9063, "step": 15045 }, { "epoch": 0.76, "learning_rate": 2.7733347685454893e-06, "loss": 0.8955, "step": 15046 }, { "epoch": 0.76, "learning_rate": 2.772197560050385e-06, "loss": 0.8326, "step": 15047 }, { "epoch": 0.76, "learning_rate": 2.7710605472403373e-06, "loss": 0.8072, "step": 15048 }, { "epoch": 0.76, "learning_rate": 2.7699237301461346e-06, "loss": 0.9442, "step": 15049 }, { "epoch": 0.76, "learning_rate": 2.768787108798555e-06, "loss": 0.9097, "step": 15050 }, { "epoch": 0.76, "learning_rate": 2.7676506832283746e-06, "loss": 0.8597, "step": 15051 }, { "epoch": 0.76, "learning_rate": 2.7665144534663534e-06, "loss": 0.8601, "step": 15052 }, { "epoch": 0.76, "learning_rate": 2.7653784195432585e-06, "loss": 0.8608, "step": 15053 }, { "epoch": 0.76, "learning_rate": 2.7642425814898475e-06, "loss": 0.8594, "step": 15054 }, { "epoch": 0.76, "learning_rate": 2.763106939336866e-06, "loss": 0.9259, "step": 15055 }, { "epoch": 0.76, "learning_rate": 2.7619714931150697e-06, "loss": 0.8444, "step": 15056 }, { "epoch": 0.76, "learning_rate": 2.7608362428551914e-06, "loss": 0.8555, "step": 15057 }, { "epoch": 0.76, "learning_rate": 2.759701188587971e-06, "loss": 0.8588, "step": 15058 }, { "epoch": 0.77, "learning_rate": 2.75856633034414e-06, "loss": 0.9295, "step": 15059 }, { "epoch": 0.77, "learning_rate": 2.757431668154419e-06, "loss": 0.8394, "step": 15060 }, { "epoch": 0.77, "learning_rate": 2.7562972020495294e-06, "loss": 0.8673, "step": 15061 }, { "epoch": 0.77, "learning_rate": 2.7551629320601856e-06, "loss": 0.8153, "step": 15062 }, { "epoch": 0.77, "learning_rate": 2.7540288582171014e-06, "loss": 0.9589, "step": 15063 }, { "epoch": 0.77, "learning_rate": 2.752894980550974e-06, "loss": 0.847, "step": 15064 }, { "epoch": 0.77, "learning_rate": 2.751761299092505e-06, "loss": 0.8337, "step": 15065 }, { "epoch": 0.77, "learning_rate": 2.7506278138723897e-06, "loss": 0.8834, "step": 15066 }, { "epoch": 0.77, "learning_rate": 2.7494945249213077e-06, "loss": 0.894, "step": 15067 }, { "epoch": 0.77, "learning_rate": 2.7483614322699537e-06, "loss": 0.8924, "step": 15068 }, { "epoch": 0.77, "learning_rate": 2.747228535948996e-06, "loss": 0.8629, "step": 15069 }, { "epoch": 0.77, "learning_rate": 2.746095835989109e-06, "loss": 0.7678, "step": 15070 }, { "epoch": 0.77, "learning_rate": 2.744963332420962e-06, "loss": 0.8063, "step": 15071 }, { "epoch": 0.77, "learning_rate": 2.7438310252752087e-06, "loss": 0.9229, "step": 15072 }, { "epoch": 0.77, "learning_rate": 2.7426989145825165e-06, "loss": 0.8565, "step": 15073 }, { "epoch": 0.77, "learning_rate": 2.7415670003735274e-06, "loss": 0.9176, "step": 15074 }, { "epoch": 0.77, "learning_rate": 2.7404352826788884e-06, "loss": 0.8743, "step": 15075 }, { "epoch": 0.77, "learning_rate": 2.739303761529244e-06, "loss": 0.8621, "step": 15076 }, { "epoch": 0.77, "learning_rate": 2.738172436955223e-06, "loss": 0.8649, "step": 15077 }, { "epoch": 0.77, "learning_rate": 2.737041308987457e-06, "loss": 0.885, "step": 15078 }, { "epoch": 0.77, "learning_rate": 2.7359103776565697e-06, "loss": 0.8344, "step": 15079 }, { "epoch": 0.77, "learning_rate": 2.7347796429931848e-06, "loss": 0.9302, "step": 15080 }, { "epoch": 0.77, "learning_rate": 2.7336491050279077e-06, "loss": 0.9304, "step": 15081 }, { "epoch": 0.77, "learning_rate": 2.73251876379135e-06, "loss": 0.9745, "step": 15082 }, { "epoch": 0.77, "learning_rate": 2.731388619314118e-06, "loss": 0.9086, "step": 15083 }, { "epoch": 0.77, "learning_rate": 2.7302586716267998e-06, "loss": 0.8984, "step": 15084 }, { "epoch": 0.77, "learning_rate": 2.7291289207600005e-06, "loss": 0.978, "step": 15085 }, { "epoch": 0.77, "learning_rate": 2.727999366744296e-06, "loss": 0.8471, "step": 15086 }, { "epoch": 0.77, "learning_rate": 2.726870009610273e-06, "loss": 0.8986, "step": 15087 }, { "epoch": 0.77, "learning_rate": 2.7257408493885085e-06, "loss": 0.9088, "step": 15088 }, { "epoch": 0.77, "learning_rate": 2.7246118861095695e-06, "loss": 0.8292, "step": 15089 }, { "epoch": 0.77, "learning_rate": 2.7234831198040224e-06, "loss": 0.8769, "step": 15090 }, { "epoch": 0.77, "learning_rate": 2.7223545505024274e-06, "loss": 0.9861, "step": 15091 }, { "epoch": 0.77, "learning_rate": 2.7212261782353445e-06, "loss": 0.9118, "step": 15092 }, { "epoch": 0.77, "learning_rate": 2.7200980030333156e-06, "loss": 0.9522, "step": 15093 }, { "epoch": 0.77, "learning_rate": 2.718970024926887e-06, "loss": 0.803, "step": 15094 }, { "epoch": 0.77, "learning_rate": 2.717842243946602e-06, "loss": 0.7767, "step": 15095 }, { "epoch": 0.77, "learning_rate": 2.7167146601229854e-06, "loss": 0.8922, "step": 15096 }, { "epoch": 0.77, "learning_rate": 2.715587273486576e-06, "loss": 0.8296, "step": 15097 }, { "epoch": 0.77, "learning_rate": 2.7144600840678893e-06, "loss": 0.9903, "step": 15098 }, { "epoch": 0.77, "learning_rate": 2.713333091897443e-06, "loss": 0.8237, "step": 15099 }, { "epoch": 0.77, "learning_rate": 2.712206297005755e-06, "loss": 0.8631, "step": 15100 }, { "epoch": 0.77, "learning_rate": 2.7110796994233245e-06, "loss": 0.871, "step": 15101 }, { "epoch": 0.77, "learning_rate": 2.7099532991806556e-06, "loss": 0.9106, "step": 15102 }, { "epoch": 0.77, "learning_rate": 2.7088270963082466e-06, "loss": 0.9243, "step": 15103 }, { "epoch": 0.77, "learning_rate": 2.7077010908365885e-06, "loss": 0.8657, "step": 15104 }, { "epoch": 0.77, "learning_rate": 2.7065752827961622e-06, "loss": 0.8518, "step": 15105 }, { "epoch": 0.77, "learning_rate": 2.7054496722174505e-06, "loss": 0.9163, "step": 15106 }, { "epoch": 0.77, "learning_rate": 2.7043242591309315e-06, "loss": 0.8715, "step": 15107 }, { "epoch": 0.77, "learning_rate": 2.703199043567065e-06, "loss": 0.8291, "step": 15108 }, { "epoch": 0.77, "learning_rate": 2.702074025556327e-06, "loss": 0.9074, "step": 15109 }, { "epoch": 0.77, "learning_rate": 2.7009492051291686e-06, "loss": 0.9326, "step": 15110 }, { "epoch": 0.77, "learning_rate": 2.699824582316043e-06, "loss": 0.8378, "step": 15111 }, { "epoch": 0.77, "learning_rate": 2.6987001571474046e-06, "loss": 0.9123, "step": 15112 }, { "epoch": 0.77, "learning_rate": 2.6975759296536876e-06, "loss": 0.8477, "step": 15113 }, { "epoch": 0.77, "learning_rate": 2.696451899865333e-06, "loss": 0.9109, "step": 15114 }, { "epoch": 0.77, "learning_rate": 2.6953280678127735e-06, "loss": 0.9046, "step": 15115 }, { "epoch": 0.77, "learning_rate": 2.6942044335264385e-06, "loss": 0.8849, "step": 15116 }, { "epoch": 0.77, "learning_rate": 2.693080997036741e-06, "loss": 0.9654, "step": 15117 }, { "epoch": 0.77, "learning_rate": 2.691957758374103e-06, "loss": 0.818, "step": 15118 }, { "epoch": 0.77, "learning_rate": 2.6908347175689365e-06, "loss": 0.8337, "step": 15119 }, { "epoch": 0.77, "learning_rate": 2.689711874651637e-06, "loss": 0.8603, "step": 15120 }, { "epoch": 0.77, "learning_rate": 2.6885892296526175e-06, "loss": 0.8836, "step": 15121 }, { "epoch": 0.77, "learning_rate": 2.6874667826022617e-06, "loss": 0.8848, "step": 15122 }, { "epoch": 0.77, "learning_rate": 2.6863445335309644e-06, "loss": 0.8609, "step": 15123 }, { "epoch": 0.77, "learning_rate": 2.68522248246911e-06, "loss": 0.7862, "step": 15124 }, { "epoch": 0.77, "learning_rate": 2.6841006294470715e-06, "loss": 0.8154, "step": 15125 }, { "epoch": 0.77, "learning_rate": 2.682978974495225e-06, "loss": 0.8972, "step": 15126 }, { "epoch": 0.77, "learning_rate": 2.681857517643938e-06, "loss": 0.8389, "step": 15127 }, { "epoch": 0.77, "learning_rate": 2.6807362589235764e-06, "loss": 0.8573, "step": 15128 }, { "epoch": 0.77, "learning_rate": 2.67961519836449e-06, "loss": 0.8715, "step": 15129 }, { "epoch": 0.77, "learning_rate": 2.6784943359970337e-06, "loss": 0.8617, "step": 15130 }, { "epoch": 0.77, "learning_rate": 2.6773736718515577e-06, "loss": 0.8925, "step": 15131 }, { "epoch": 0.77, "learning_rate": 2.676253205958391e-06, "loss": 0.9323, "step": 15132 }, { "epoch": 0.77, "learning_rate": 2.675132938347884e-06, "loss": 0.9478, "step": 15133 }, { "epoch": 0.77, "learning_rate": 2.6740128690503575e-06, "loss": 0.9005, "step": 15134 }, { "epoch": 0.77, "learning_rate": 2.672892998096137e-06, "loss": 0.7782, "step": 15135 }, { "epoch": 0.77, "learning_rate": 2.6717733255155463e-06, "loss": 0.8908, "step": 15136 }, { "epoch": 0.77, "learning_rate": 2.670653851338891e-06, "loss": 0.8455, "step": 15137 }, { "epoch": 0.77, "learning_rate": 2.669534575596491e-06, "loss": 0.8081, "step": 15138 }, { "epoch": 0.77, "learning_rate": 2.6684154983186396e-06, "loss": 0.8815, "step": 15139 }, { "epoch": 0.77, "learning_rate": 2.6672966195356386e-06, "loss": 0.9148, "step": 15140 }, { "epoch": 0.77, "learning_rate": 2.6661779392777833e-06, "loss": 0.8939, "step": 15141 }, { "epoch": 0.77, "learning_rate": 2.6650594575753542e-06, "loss": 0.918, "step": 15142 }, { "epoch": 0.77, "learning_rate": 2.6639411744586375e-06, "loss": 0.8232, "step": 15143 }, { "epoch": 0.77, "learning_rate": 2.662823089957908e-06, "loss": 0.7867, "step": 15144 }, { "epoch": 0.77, "learning_rate": 2.6617052041034398e-06, "loss": 0.9286, "step": 15145 }, { "epoch": 0.77, "learning_rate": 2.6605875169254936e-06, "loss": 0.9352, "step": 15146 }, { "epoch": 0.77, "learning_rate": 2.6594700284543318e-06, "loss": 0.8302, "step": 15147 }, { "epoch": 0.77, "learning_rate": 2.6583527387202123e-06, "loss": 0.9432, "step": 15148 }, { "epoch": 0.77, "learning_rate": 2.6572356477533744e-06, "loss": 0.8851, "step": 15149 }, { "epoch": 0.77, "learning_rate": 2.656118755584076e-06, "loss": 0.8085, "step": 15150 }, { "epoch": 0.77, "learning_rate": 2.6550020622425467e-06, "loss": 0.9426, "step": 15151 }, { "epoch": 0.77, "learning_rate": 2.6538855677590215e-06, "loss": 0.9659, "step": 15152 }, { "epoch": 0.77, "learning_rate": 2.652769272163732e-06, "loss": 0.8339, "step": 15153 }, { "epoch": 0.77, "learning_rate": 2.6516531754868945e-06, "loss": 0.913, "step": 15154 }, { "epoch": 0.77, "learning_rate": 2.6505372777587292e-06, "loss": 0.8881, "step": 15155 }, { "epoch": 0.77, "learning_rate": 2.649421579009448e-06, "loss": 0.8905, "step": 15156 }, { "epoch": 0.77, "learning_rate": 2.6483060792692606e-06, "loss": 0.8875, "step": 15157 }, { "epoch": 0.77, "learning_rate": 2.6471907785683604e-06, "loss": 0.8485, "step": 15158 }, { "epoch": 0.77, "learning_rate": 2.646075676936948e-06, "loss": 0.906, "step": 15159 }, { "epoch": 0.77, "learning_rate": 2.6449607744052154e-06, "loss": 0.8888, "step": 15160 }, { "epoch": 0.77, "learning_rate": 2.6438460710033385e-06, "loss": 0.9181, "step": 15161 }, { "epoch": 0.77, "learning_rate": 2.6427315667615094e-06, "loss": 0.8448, "step": 15162 }, { "epoch": 0.77, "learning_rate": 2.6416172617098925e-06, "loss": 0.9322, "step": 15163 }, { "epoch": 0.77, "learning_rate": 2.6405031558786587e-06, "loss": 0.929, "step": 15164 }, { "epoch": 0.77, "learning_rate": 2.6393892492979756e-06, "loss": 0.7963, "step": 15165 }, { "epoch": 0.77, "learning_rate": 2.6382755419979945e-06, "loss": 0.834, "step": 15166 }, { "epoch": 0.77, "learning_rate": 2.637162034008871e-06, "loss": 0.9302, "step": 15167 }, { "epoch": 0.77, "learning_rate": 2.6360487253607525e-06, "loss": 0.851, "step": 15168 }, { "epoch": 0.77, "learning_rate": 2.6349356160837826e-06, "loss": 0.819, "step": 15169 }, { "epoch": 0.77, "learning_rate": 2.6338227062080924e-06, "loss": 0.8739, "step": 15170 }, { "epoch": 0.77, "learning_rate": 2.6327099957638147e-06, "loss": 0.7964, "step": 15171 }, { "epoch": 0.77, "learning_rate": 2.631597484781079e-06, "loss": 0.8455, "step": 15172 }, { "epoch": 0.77, "learning_rate": 2.630485173289996e-06, "loss": 0.953, "step": 15173 }, { "epoch": 0.77, "learning_rate": 2.6293730613206924e-06, "loss": 0.8521, "step": 15174 }, { "epoch": 0.77, "learning_rate": 2.6282611489032683e-06, "loss": 0.8944, "step": 15175 }, { "epoch": 0.77, "learning_rate": 2.6271494360678317e-06, "loss": 0.8178, "step": 15176 }, { "epoch": 0.77, "learning_rate": 2.626037922844482e-06, "loss": 0.7328, "step": 15177 }, { "epoch": 0.77, "learning_rate": 2.6249266092633073e-06, "loss": 0.975, "step": 15178 }, { "epoch": 0.77, "learning_rate": 2.6238154953543983e-06, "loss": 0.8615, "step": 15179 }, { "epoch": 0.77, "learning_rate": 2.6227045811478368e-06, "loss": 0.8302, "step": 15180 }, { "epoch": 0.77, "learning_rate": 2.621593866673704e-06, "loss": 0.8595, "step": 15181 }, { "epoch": 0.77, "learning_rate": 2.620483351962063e-06, "loss": 0.7845, "step": 15182 }, { "epoch": 0.77, "learning_rate": 2.619373037042985e-06, "loss": 0.9756, "step": 15183 }, { "epoch": 0.77, "learning_rate": 2.6182629219465327e-06, "loss": 0.8143, "step": 15184 }, { "epoch": 0.77, "learning_rate": 2.6171530067027516e-06, "loss": 0.9227, "step": 15185 }, { "epoch": 0.77, "learning_rate": 2.6160432913417046e-06, "loss": 0.8666, "step": 15186 }, { "epoch": 0.77, "learning_rate": 2.6149337758934266e-06, "loss": 0.9102, "step": 15187 }, { "epoch": 0.77, "learning_rate": 2.6138244603879594e-06, "loss": 0.9203, "step": 15188 }, { "epoch": 0.77, "learning_rate": 2.6127153448553387e-06, "loss": 0.7638, "step": 15189 }, { "epoch": 0.77, "learning_rate": 2.611606429325588e-06, "loss": 0.8499, "step": 15190 }, { "epoch": 0.77, "learning_rate": 2.6104977138287335e-06, "loss": 0.8954, "step": 15191 }, { "epoch": 0.77, "learning_rate": 2.609389198394792e-06, "loss": 0.9103, "step": 15192 }, { "epoch": 0.77, "learning_rate": 2.6082808830537763e-06, "loss": 0.8945, "step": 15193 }, { "epoch": 0.77, "learning_rate": 2.6071727678356897e-06, "loss": 0.8691, "step": 15194 }, { "epoch": 0.77, "learning_rate": 2.6060648527705348e-06, "loss": 0.8041, "step": 15195 }, { "epoch": 0.77, "learning_rate": 2.60495713788831e-06, "loss": 0.9286, "step": 15196 }, { "epoch": 0.77, "learning_rate": 2.603849623218997e-06, "loss": 0.8745, "step": 15197 }, { "epoch": 0.77, "learning_rate": 2.602742308792593e-06, "loss": 0.9877, "step": 15198 }, { "epoch": 0.77, "learning_rate": 2.6016351946390683e-06, "loss": 0.8164, "step": 15199 }, { "epoch": 0.77, "learning_rate": 2.6005282807883992e-06, "loss": 0.8643, "step": 15200 }, { "epoch": 0.77, "learning_rate": 2.5994215672705568e-06, "loss": 0.9459, "step": 15201 }, { "epoch": 0.77, "learning_rate": 2.598315054115499e-06, "loss": 0.9513, "step": 15202 }, { "epoch": 0.77, "learning_rate": 2.5972087413531866e-06, "loss": 0.8882, "step": 15203 }, { "epoch": 0.77, "learning_rate": 2.5961026290135717e-06, "loss": 0.8051, "step": 15204 }, { "epoch": 0.77, "learning_rate": 2.5949967171266043e-06, "loss": 0.8686, "step": 15205 }, { "epoch": 0.77, "learning_rate": 2.593891005722219e-06, "loss": 0.963, "step": 15206 }, { "epoch": 0.77, "learning_rate": 2.5927854948303554e-06, "loss": 0.862, "step": 15207 }, { "epoch": 0.77, "learning_rate": 2.5916801844809434e-06, "loss": 0.9275, "step": 15208 }, { "epoch": 0.77, "learning_rate": 2.590575074703909e-06, "loss": 0.7664, "step": 15209 }, { "epoch": 0.77, "learning_rate": 2.589470165529174e-06, "loss": 0.902, "step": 15210 }, { "epoch": 0.77, "learning_rate": 2.588365456986648e-06, "loss": 0.8639, "step": 15211 }, { "epoch": 0.77, "learning_rate": 2.5872609491062415e-06, "loss": 0.8216, "step": 15212 }, { "epoch": 0.77, "learning_rate": 2.586156641917862e-06, "loss": 0.8277, "step": 15213 }, { "epoch": 0.77, "learning_rate": 2.5850525354513965e-06, "loss": 0.8971, "step": 15214 }, { "epoch": 0.77, "learning_rate": 2.583948629736751e-06, "loss": 0.9333, "step": 15215 }, { "epoch": 0.77, "learning_rate": 2.5828449248038033e-06, "loss": 0.8744, "step": 15216 }, { "epoch": 0.77, "learning_rate": 2.581741420682439e-06, "loss": 0.9329, "step": 15217 }, { "epoch": 0.77, "learning_rate": 2.580638117402535e-06, "loss": 0.899, "step": 15218 }, { "epoch": 0.77, "learning_rate": 2.579535014993958e-06, "loss": 0.8873, "step": 15219 }, { "epoch": 0.77, "learning_rate": 2.578432113486575e-06, "loss": 0.9479, "step": 15220 }, { "epoch": 0.77, "learning_rate": 2.5773294129102477e-06, "loss": 0.8789, "step": 15221 }, { "epoch": 0.77, "learning_rate": 2.5762269132948315e-06, "loss": 0.8873, "step": 15222 }, { "epoch": 0.77, "learning_rate": 2.5751246146701715e-06, "loss": 0.8524, "step": 15223 }, { "epoch": 0.77, "learning_rate": 2.5740225170661114e-06, "loss": 0.8779, "step": 15224 }, { "epoch": 0.77, "learning_rate": 2.572920620512496e-06, "loss": 0.7631, "step": 15225 }, { "epoch": 0.77, "learning_rate": 2.5718189250391457e-06, "loss": 0.7651, "step": 15226 }, { "epoch": 0.77, "learning_rate": 2.5707174306759007e-06, "loss": 0.9236, "step": 15227 }, { "epoch": 0.77, "learning_rate": 2.569616137452575e-06, "loss": 0.9197, "step": 15228 }, { "epoch": 0.77, "learning_rate": 2.568515045398987e-06, "loss": 0.8034, "step": 15229 }, { "epoch": 0.77, "learning_rate": 2.5674141545449506e-06, "loss": 0.8715, "step": 15230 }, { "epoch": 0.77, "learning_rate": 2.566313464920265e-06, "loss": 0.8404, "step": 15231 }, { "epoch": 0.77, "learning_rate": 2.565212976554734e-06, "loss": 0.8328, "step": 15232 }, { "epoch": 0.77, "learning_rate": 2.564112689478151e-06, "loss": 0.8019, "step": 15233 }, { "epoch": 0.77, "learning_rate": 2.5630126037203096e-06, "loss": 0.9954, "step": 15234 }, { "epoch": 0.77, "learning_rate": 2.561912719310986e-06, "loss": 0.884, "step": 15235 }, { "epoch": 0.77, "learning_rate": 2.560813036279962e-06, "loss": 0.8129, "step": 15236 }, { "epoch": 0.77, "learning_rate": 2.559713554657014e-06, "loss": 0.9018, "step": 15237 }, { "epoch": 0.77, "learning_rate": 2.5586142744719e-06, "loss": 0.9287, "step": 15238 }, { "epoch": 0.77, "learning_rate": 2.557515195754392e-06, "loss": 0.864, "step": 15239 }, { "epoch": 0.77, "learning_rate": 2.5564163185342392e-06, "loss": 0.8624, "step": 15240 }, { "epoch": 0.77, "learning_rate": 2.5553176428411962e-06, "loss": 0.8256, "step": 15241 }, { "epoch": 0.77, "learning_rate": 2.5542191687050098e-06, "loss": 0.7829, "step": 15242 }, { "epoch": 0.77, "learning_rate": 2.5531208961554155e-06, "loss": 0.7857, "step": 15243 }, { "epoch": 0.77, "learning_rate": 2.552022825222149e-06, "loss": 0.9282, "step": 15244 }, { "epoch": 0.77, "learning_rate": 2.5509249559349414e-06, "loss": 0.9232, "step": 15245 }, { "epoch": 0.77, "learning_rate": 2.5498272883235187e-06, "loss": 0.9337, "step": 15246 }, { "epoch": 0.77, "learning_rate": 2.5487298224175917e-06, "loss": 0.9528, "step": 15247 }, { "epoch": 0.77, "learning_rate": 2.5476325582468776e-06, "loss": 0.8841, "step": 15248 }, { "epoch": 0.77, "learning_rate": 2.546535495841086e-06, "loss": 0.8452, "step": 15249 }, { "epoch": 0.77, "learning_rate": 2.545438635229911e-06, "loss": 0.8129, "step": 15250 }, { "epoch": 0.77, "learning_rate": 2.544341976443059e-06, "loss": 0.8338, "step": 15251 }, { "epoch": 0.77, "learning_rate": 2.5432455195102134e-06, "loss": 0.8347, "step": 15252 }, { "epoch": 0.77, "learning_rate": 2.5421492644610614e-06, "loss": 0.8301, "step": 15253 }, { "epoch": 0.77, "learning_rate": 2.541053211325286e-06, "loss": 0.8767, "step": 15254 }, { "epoch": 0.77, "learning_rate": 2.539957360132558e-06, "loss": 0.819, "step": 15255 }, { "epoch": 0.78, "learning_rate": 2.538861710912547e-06, "loss": 0.8335, "step": 15256 }, { "epoch": 0.78, "learning_rate": 2.537766263694916e-06, "loss": 0.8, "step": 15257 }, { "epoch": 0.78, "learning_rate": 2.536671018509328e-06, "loss": 0.9187, "step": 15258 }, { "epoch": 0.78, "learning_rate": 2.5355759753854293e-06, "loss": 0.8257, "step": 15259 }, { "epoch": 0.78, "learning_rate": 2.5344811343528684e-06, "loss": 0.8577, "step": 15260 }, { "epoch": 0.78, "learning_rate": 2.5333864954412923e-06, "loss": 0.9412, "step": 15261 }, { "epoch": 0.78, "learning_rate": 2.5322920586803255e-06, "loss": 0.985, "step": 15262 }, { "epoch": 0.78, "learning_rate": 2.5311978240996138e-06, "loss": 0.874, "step": 15263 }, { "epoch": 0.78, "learning_rate": 2.5301037917287706e-06, "loss": 0.8609, "step": 15264 }, { "epoch": 0.78, "learning_rate": 2.5290099615974207e-06, "loss": 0.8817, "step": 15265 }, { "epoch": 0.78, "learning_rate": 2.527916333735181e-06, "loss": 0.8569, "step": 15266 }, { "epoch": 0.78, "learning_rate": 2.526822908171653e-06, "loss": 0.8959, "step": 15267 }, { "epoch": 0.78, "learning_rate": 2.5257296849364444e-06, "loss": 0.8224, "step": 15268 }, { "epoch": 0.78, "learning_rate": 2.5246366640591525e-06, "loss": 0.9128, "step": 15269 }, { "epoch": 0.78, "learning_rate": 2.5235438455693715e-06, "loss": 0.8867, "step": 15270 }, { "epoch": 0.78, "learning_rate": 2.5224512294966853e-06, "loss": 0.9561, "step": 15271 }, { "epoch": 0.78, "learning_rate": 2.521358815870676e-06, "loss": 0.9001, "step": 15272 }, { "epoch": 0.78, "learning_rate": 2.5202666047209235e-06, "loss": 0.8376, "step": 15273 }, { "epoch": 0.78, "learning_rate": 2.5191745960769874e-06, "loss": 0.8286, "step": 15274 }, { "epoch": 0.78, "learning_rate": 2.5180827899684477e-06, "loss": 0.8259, "step": 15275 }, { "epoch": 0.78, "learning_rate": 2.516991186424853e-06, "loss": 0.8536, "step": 15276 }, { "epoch": 0.78, "learning_rate": 2.5158997854757605e-06, "loss": 0.9552, "step": 15277 }, { "epoch": 0.78, "learning_rate": 2.514808587150721e-06, "loss": 0.8153, "step": 15278 }, { "epoch": 0.78, "learning_rate": 2.5137175914792713e-06, "loss": 0.887, "step": 15279 }, { "epoch": 0.78, "learning_rate": 2.512626798490957e-06, "loss": 0.8106, "step": 15280 }, { "epoch": 0.78, "learning_rate": 2.511536208215304e-06, "loss": 0.8546, "step": 15281 }, { "epoch": 0.78, "learning_rate": 2.5104458206818415e-06, "loss": 0.9369, "step": 15282 }, { "epoch": 0.78, "learning_rate": 2.5093556359200923e-06, "loss": 0.8573, "step": 15283 }, { "epoch": 0.78, "learning_rate": 2.508265653959567e-06, "loss": 0.9417, "step": 15284 }, { "epoch": 0.78, "learning_rate": 2.507175874829779e-06, "loss": 0.8353, "step": 15285 }, { "epoch": 0.78, "learning_rate": 2.506086298560233e-06, "loss": 0.9293, "step": 15286 }, { "epoch": 0.78, "learning_rate": 2.5049969251804298e-06, "loss": 0.9227, "step": 15287 }, { "epoch": 0.78, "learning_rate": 2.5039077547198576e-06, "loss": 0.8185, "step": 15288 }, { "epoch": 0.78, "learning_rate": 2.502818787208008e-06, "loss": 0.872, "step": 15289 }, { "epoch": 0.78, "learning_rate": 2.501730022674367e-06, "loss": 0.9031, "step": 15290 }, { "epoch": 0.78, "learning_rate": 2.500641461148401e-06, "loss": 0.8806, "step": 15291 }, { "epoch": 0.78, "learning_rate": 2.4995531026595952e-06, "loss": 0.9232, "step": 15292 }, { "epoch": 0.78, "learning_rate": 2.4984649472374067e-06, "loss": 0.8609, "step": 15293 }, { "epoch": 0.78, "learning_rate": 2.497376994911298e-06, "loss": 0.8317, "step": 15294 }, { "epoch": 0.78, "learning_rate": 2.4962892457107292e-06, "loss": 0.7697, "step": 15295 }, { "epoch": 0.78, "learning_rate": 2.4952016996651417e-06, "loss": 0.8394, "step": 15296 }, { "epoch": 0.78, "learning_rate": 2.4941143568039837e-06, "loss": 0.8014, "step": 15297 }, { "epoch": 0.78, "learning_rate": 2.493027217156694e-06, "loss": 0.8319, "step": 15298 }, { "epoch": 0.78, "learning_rate": 2.491940280752708e-06, "loss": 0.9515, "step": 15299 }, { "epoch": 0.78, "learning_rate": 2.490853547621449e-06, "loss": 0.9235, "step": 15300 }, { "epoch": 0.78, "learning_rate": 2.4897670177923404e-06, "loss": 0.917, "step": 15301 }, { "epoch": 0.78, "learning_rate": 2.4886806912948034e-06, "loss": 0.9094, "step": 15302 }, { "epoch": 0.78, "learning_rate": 2.487594568158239e-06, "loss": 0.8849, "step": 15303 }, { "epoch": 0.78, "learning_rate": 2.486508648412065e-06, "loss": 0.9811, "step": 15304 }, { "epoch": 0.78, "learning_rate": 2.485422932085673e-06, "loss": 0.822, "step": 15305 }, { "epoch": 0.78, "learning_rate": 2.484337419208459e-06, "loss": 0.7662, "step": 15306 }, { "epoch": 0.78, "learning_rate": 2.483252109809817e-06, "loss": 0.8736, "step": 15307 }, { "epoch": 0.78, "learning_rate": 2.4821670039191235e-06, "loss": 0.909, "step": 15308 }, { "epoch": 0.78, "learning_rate": 2.4810821015657604e-06, "loss": 0.8525, "step": 15309 }, { "epoch": 0.78, "learning_rate": 2.4799974027791006e-06, "loss": 0.8011, "step": 15310 }, { "epoch": 0.78, "learning_rate": 2.4789129075885134e-06, "loss": 0.9179, "step": 15311 }, { "epoch": 0.78, "learning_rate": 2.4778286160233534e-06, "loss": 0.8207, "step": 15312 }, { "epoch": 0.78, "learning_rate": 2.4767445281129807e-06, "loss": 0.8932, "step": 15313 }, { "epoch": 0.78, "learning_rate": 2.4756606438867494e-06, "loss": 0.974, "step": 15314 }, { "epoch": 0.78, "learning_rate": 2.4745769633739947e-06, "loss": 0.9357, "step": 15315 }, { "epoch": 0.78, "learning_rate": 2.4734934866040683e-06, "loss": 0.844, "step": 15316 }, { "epoch": 0.78, "learning_rate": 2.4724102136062955e-06, "loss": 0.7764, "step": 15317 }, { "epoch": 0.78, "learning_rate": 2.4713271444100073e-06, "loss": 0.9427, "step": 15318 }, { "epoch": 0.78, "learning_rate": 2.4702442790445292e-06, "loss": 0.8136, "step": 15319 }, { "epoch": 0.78, "learning_rate": 2.469161617539174e-06, "loss": 0.9073, "step": 15320 }, { "epoch": 0.78, "learning_rate": 2.4680791599232558e-06, "loss": 0.9175, "step": 15321 }, { "epoch": 0.78, "learning_rate": 2.46699690622608e-06, "loss": 0.8483, "step": 15322 }, { "epoch": 0.78, "learning_rate": 2.465914856476952e-06, "loss": 0.9563, "step": 15323 }, { "epoch": 0.78, "learning_rate": 2.464833010705161e-06, "loss": 0.9197, "step": 15324 }, { "epoch": 0.78, "learning_rate": 2.4637513689399993e-06, "loss": 0.9144, "step": 15325 }, { "epoch": 0.78, "learning_rate": 2.4626699312107537e-06, "loss": 0.8239, "step": 15326 }, { "epoch": 0.78, "learning_rate": 2.461588697546696e-06, "loss": 0.8173, "step": 15327 }, { "epoch": 0.78, "learning_rate": 2.4605076679771088e-06, "loss": 0.9003, "step": 15328 }, { "epoch": 0.78, "learning_rate": 2.459426842531253e-06, "loss": 0.8024, "step": 15329 }, { "epoch": 0.78, "learning_rate": 2.4583462212383924e-06, "loss": 0.8101, "step": 15330 }, { "epoch": 0.78, "learning_rate": 2.4572658041277874e-06, "loss": 0.9227, "step": 15331 }, { "epoch": 0.78, "learning_rate": 2.4561855912286826e-06, "loss": 0.7948, "step": 15332 }, { "epoch": 0.78, "learning_rate": 2.4551055825703273e-06, "loss": 0.9216, "step": 15333 }, { "epoch": 0.78, "learning_rate": 2.454025778181961e-06, "loss": 0.84, "step": 15334 }, { "epoch": 0.78, "learning_rate": 2.452946178092821e-06, "loss": 0.8137, "step": 15335 }, { "epoch": 0.78, "learning_rate": 2.451866782332132e-06, "loss": 0.9099, "step": 15336 }, { "epoch": 0.78, "learning_rate": 2.450787590929118e-06, "loss": 0.7748, "step": 15337 }, { "epoch": 0.78, "learning_rate": 2.4497086039130024e-06, "loss": 0.9301, "step": 15338 }, { "epoch": 0.78, "learning_rate": 2.4486298213129865e-06, "loss": 0.8229, "step": 15339 }, { "epoch": 0.78, "learning_rate": 2.4475512431582905e-06, "loss": 0.8389, "step": 15340 }, { "epoch": 0.78, "learning_rate": 2.446472869478107e-06, "loss": 0.8167, "step": 15341 }, { "epoch": 0.78, "learning_rate": 2.4453947003016333e-06, "loss": 0.8674, "step": 15342 }, { "epoch": 0.78, "learning_rate": 2.4443167356580644e-06, "loss": 0.875, "step": 15343 }, { "epoch": 0.78, "learning_rate": 2.4432389755765772e-06, "loss": 0.8766, "step": 15344 }, { "epoch": 0.78, "learning_rate": 2.442161420086356e-06, "loss": 0.812, "step": 15345 }, { "epoch": 0.78, "learning_rate": 2.4410840692165727e-06, "loss": 0.98, "step": 15346 }, { "epoch": 0.78, "learning_rate": 2.4400069229963973e-06, "loss": 0.9642, "step": 15347 }, { "epoch": 0.78, "learning_rate": 2.4389299814549925e-06, "loss": 0.9007, "step": 15348 }, { "epoch": 0.78, "learning_rate": 2.437853244621513e-06, "loss": 0.8848, "step": 15349 }, { "epoch": 0.78, "learning_rate": 2.436776712525111e-06, "loss": 0.8799, "step": 15350 }, { "epoch": 0.78, "learning_rate": 2.435700385194932e-06, "loss": 0.8162, "step": 15351 }, { "epoch": 0.78, "learning_rate": 2.4346242626601204e-06, "loss": 0.9652, "step": 15352 }, { "epoch": 0.78, "learning_rate": 2.4335483449498053e-06, "loss": 0.8523, "step": 15353 }, { "epoch": 0.78, "learning_rate": 2.4324726320931192e-06, "loss": 0.8766, "step": 15354 }, { "epoch": 0.78, "learning_rate": 2.431397124119188e-06, "loss": 0.8651, "step": 15355 }, { "epoch": 0.78, "learning_rate": 2.430321821057121e-06, "loss": 0.9054, "step": 15356 }, { "epoch": 0.78, "learning_rate": 2.429246722936043e-06, "loss": 0.8824, "step": 15357 }, { "epoch": 0.78, "learning_rate": 2.428171829785053e-06, "loss": 0.9223, "step": 15358 }, { "epoch": 0.78, "learning_rate": 2.4270971416332547e-06, "loss": 0.7724, "step": 15359 }, { "epoch": 0.78, "learning_rate": 2.426022658509747e-06, "loss": 0.9465, "step": 15360 }, { "epoch": 0.78, "learning_rate": 2.424948380443615e-06, "loss": 0.9442, "step": 15361 }, { "epoch": 0.78, "learning_rate": 2.423874307463946e-06, "loss": 0.8165, "step": 15362 }, { "epoch": 0.78, "learning_rate": 2.422800439599821e-06, "loss": 0.9103, "step": 15363 }, { "epoch": 0.78, "learning_rate": 2.4217267768803143e-06, "loss": 0.8688, "step": 15364 }, { "epoch": 0.78, "learning_rate": 2.4206533193344896e-06, "loss": 0.9261, "step": 15365 }, { "epoch": 0.78, "learning_rate": 2.4195800669914127e-06, "loss": 0.8394, "step": 15366 }, { "epoch": 0.78, "learning_rate": 2.4185070198801443e-06, "loss": 0.8814, "step": 15367 }, { "epoch": 0.78, "learning_rate": 2.417434178029725e-06, "loss": 0.9384, "step": 15368 }, { "epoch": 0.78, "learning_rate": 2.4163615414692156e-06, "loss": 0.8538, "step": 15369 }, { "epoch": 0.78, "learning_rate": 2.4152891102276444e-06, "loss": 0.9526, "step": 15370 }, { "epoch": 0.78, "learning_rate": 2.414216884334052e-06, "loss": 0.9436, "step": 15371 }, { "epoch": 0.78, "learning_rate": 2.4131448638174694e-06, "loss": 0.8266, "step": 15372 }, { "epoch": 0.78, "learning_rate": 2.4120730487069153e-06, "loss": 0.9033, "step": 15373 }, { "epoch": 0.78, "learning_rate": 2.4110014390314106e-06, "loss": 0.8765, "step": 15374 }, { "epoch": 0.78, "learning_rate": 2.409930034819967e-06, "loss": 0.9167, "step": 15375 }, { "epoch": 0.78, "learning_rate": 2.408858836101596e-06, "loss": 0.8761, "step": 15376 }, { "epoch": 0.78, "learning_rate": 2.4077878429052938e-06, "loss": 0.8268, "step": 15377 }, { "epoch": 0.78, "learning_rate": 2.4067170552600572e-06, "loss": 0.973, "step": 15378 }, { "epoch": 0.78, "learning_rate": 2.4056464731948805e-06, "loss": 0.8739, "step": 15379 }, { "epoch": 0.78, "learning_rate": 2.40457609673874e-06, "loss": 0.9294, "step": 15380 }, { "epoch": 0.78, "learning_rate": 2.4035059259206263e-06, "loss": 0.9826, "step": 15381 }, { "epoch": 0.78, "learning_rate": 2.402435960769506e-06, "loss": 0.8831, "step": 15382 }, { "epoch": 0.78, "learning_rate": 2.401366201314348e-06, "loss": 0.8635, "step": 15383 }, { "epoch": 0.78, "learning_rate": 2.400296647584118e-06, "loss": 0.8275, "step": 15384 }, { "epoch": 0.78, "learning_rate": 2.3992272996077693e-06, "loss": 0.9557, "step": 15385 }, { "epoch": 0.78, "learning_rate": 2.398158157414254e-06, "loss": 0.9237, "step": 15386 }, { "epoch": 0.78, "learning_rate": 2.3970892210325194e-06, "loss": 0.9335, "step": 15387 }, { "epoch": 0.78, "learning_rate": 2.3960204904915074e-06, "loss": 0.869, "step": 15388 }, { "epoch": 0.78, "learning_rate": 2.394951965820148e-06, "loss": 0.9732, "step": 15389 }, { "epoch": 0.78, "learning_rate": 2.3938836470473734e-06, "loss": 0.8448, "step": 15390 }, { "epoch": 0.78, "learning_rate": 2.3928155342021097e-06, "loss": 0.8892, "step": 15391 }, { "epoch": 0.78, "learning_rate": 2.391747627313266e-06, "loss": 0.9616, "step": 15392 }, { "epoch": 0.78, "learning_rate": 2.390679926409767e-06, "loss": 0.8888, "step": 15393 }, { "epoch": 0.78, "learning_rate": 2.389612431520509e-06, "loss": 0.7974, "step": 15394 }, { "epoch": 0.78, "learning_rate": 2.388545142674399e-06, "loss": 0.7666, "step": 15395 }, { "epoch": 0.78, "learning_rate": 2.3874780599003335e-06, "loss": 0.8339, "step": 15396 }, { "epoch": 0.78, "learning_rate": 2.3864111832271985e-06, "loss": 0.8285, "step": 15397 }, { "epoch": 0.78, "learning_rate": 2.385344512683879e-06, "loss": 0.8732, "step": 15398 }, { "epoch": 0.78, "learning_rate": 2.3842780482992565e-06, "loss": 0.7856, "step": 15399 }, { "epoch": 0.78, "learning_rate": 2.3832117901022056e-06, "loss": 0.9775, "step": 15400 }, { "epoch": 0.78, "learning_rate": 2.3821457381215896e-06, "loss": 0.8749, "step": 15401 }, { "epoch": 0.78, "learning_rate": 2.381079892386273e-06, "loss": 0.8202, "step": 15402 }, { "epoch": 0.78, "learning_rate": 2.3800142529251147e-06, "loss": 0.9122, "step": 15403 }, { "epoch": 0.78, "learning_rate": 2.3789488197669586e-06, "loss": 0.9149, "step": 15404 }, { "epoch": 0.78, "learning_rate": 2.377883592940661e-06, "loss": 0.8746, "step": 15405 }, { "epoch": 0.78, "learning_rate": 2.3768185724750515e-06, "loss": 0.8697, "step": 15406 }, { "epoch": 0.78, "learning_rate": 2.37575375839897e-06, "loss": 0.9603, "step": 15407 }, { "epoch": 0.78, "learning_rate": 2.3746891507412463e-06, "loss": 0.8372, "step": 15408 }, { "epoch": 0.78, "learning_rate": 2.3736247495306974e-06, "loss": 0.8651, "step": 15409 }, { "epoch": 0.78, "learning_rate": 2.372560554796146e-06, "loss": 0.8615, "step": 15410 }, { "epoch": 0.78, "learning_rate": 2.3714965665664013e-06, "loss": 0.8258, "step": 15411 }, { "epoch": 0.78, "learning_rate": 2.370432784870275e-06, "loss": 0.8623, "step": 15412 }, { "epoch": 0.78, "learning_rate": 2.3693692097365605e-06, "loss": 0.9647, "step": 15413 }, { "epoch": 0.78, "learning_rate": 2.3683058411940563e-06, "loss": 0.9312, "step": 15414 }, { "epoch": 0.78, "learning_rate": 2.367242679271552e-06, "loss": 0.8373, "step": 15415 }, { "epoch": 0.78, "learning_rate": 2.366179723997831e-06, "loss": 0.9452, "step": 15416 }, { "epoch": 0.78, "learning_rate": 2.365116975401677e-06, "loss": 0.8605, "step": 15417 }, { "epoch": 0.78, "learning_rate": 2.3640544335118533e-06, "loss": 0.8959, "step": 15418 }, { "epoch": 0.78, "learning_rate": 2.362992098357132e-06, "loss": 1.0002, "step": 15419 }, { "epoch": 0.78, "learning_rate": 2.361929969966278e-06, "loss": 0.9014, "step": 15420 }, { "epoch": 0.78, "learning_rate": 2.3608680483680378e-06, "loss": 0.9107, "step": 15421 }, { "epoch": 0.78, "learning_rate": 2.3598063335911738e-06, "loss": 0.8629, "step": 15422 }, { "epoch": 0.78, "learning_rate": 2.358744825664421e-06, "loss": 0.9255, "step": 15423 }, { "epoch": 0.78, "learning_rate": 2.357683524616523e-06, "loss": 0.8464, "step": 15424 }, { "epoch": 0.78, "learning_rate": 2.3566224304762153e-06, "loss": 0.8665, "step": 15425 }, { "epoch": 0.78, "learning_rate": 2.355561543272221e-06, "loss": 0.8896, "step": 15426 }, { "epoch": 0.78, "learning_rate": 2.3545008630332646e-06, "loss": 0.854, "step": 15427 }, { "epoch": 0.78, "learning_rate": 2.353440389788063e-06, "loss": 0.8043, "step": 15428 }, { "epoch": 0.78, "learning_rate": 2.352380123565331e-06, "loss": 0.9013, "step": 15429 }, { "epoch": 0.78, "learning_rate": 2.351320064393767e-06, "loss": 0.9309, "step": 15430 }, { "epoch": 0.78, "learning_rate": 2.3502602123020755e-06, "loss": 0.7538, "step": 15431 }, { "epoch": 0.78, "learning_rate": 2.349200567318953e-06, "loss": 0.8478, "step": 15432 }, { "epoch": 0.78, "learning_rate": 2.3481411294730794e-06, "loss": 0.8661, "step": 15433 }, { "epoch": 0.78, "learning_rate": 2.34708189879315e-06, "loss": 0.8588, "step": 15434 }, { "epoch": 0.78, "learning_rate": 2.346022875307832e-06, "loss": 0.8468, "step": 15435 }, { "epoch": 0.78, "learning_rate": 2.344964059045802e-06, "loss": 0.8719, "step": 15436 }, { "epoch": 0.78, "learning_rate": 2.3439054500357285e-06, "loss": 0.9217, "step": 15437 }, { "epoch": 0.78, "learning_rate": 2.342847048306267e-06, "loss": 0.8306, "step": 15438 }, { "epoch": 0.78, "learning_rate": 2.3417888538860755e-06, "loss": 0.9641, "step": 15439 }, { "epoch": 0.78, "learning_rate": 2.340730866803802e-06, "loss": 0.9537, "step": 15440 }, { "epoch": 0.78, "learning_rate": 2.3396730870880958e-06, "loss": 0.8105, "step": 15441 }, { "epoch": 0.78, "learning_rate": 2.338615514767586e-06, "loss": 0.9034, "step": 15442 }, { "epoch": 0.78, "learning_rate": 2.337558149870912e-06, "loss": 0.9082, "step": 15443 }, { "epoch": 0.78, "learning_rate": 2.3365009924267002e-06, "loss": 0.9229, "step": 15444 }, { "epoch": 0.78, "learning_rate": 2.335444042463566e-06, "loss": 0.9122, "step": 15445 }, { "epoch": 0.78, "learning_rate": 2.3343873000101357e-06, "loss": 0.8848, "step": 15446 }, { "epoch": 0.78, "learning_rate": 2.33333076509501e-06, "loss": 0.7936, "step": 15447 }, { "epoch": 0.78, "learning_rate": 2.3322744377467976e-06, "loss": 0.8817, "step": 15448 }, { "epoch": 0.78, "learning_rate": 2.3312183179940995e-06, "loss": 0.8175, "step": 15449 }, { "epoch": 0.78, "learning_rate": 2.330162405865504e-06, "loss": 0.8538, "step": 15450 }, { "epoch": 0.78, "learning_rate": 2.3291067013896016e-06, "loss": 0.7978, "step": 15451 }, { "epoch": 0.79, "learning_rate": 2.328051204594973e-06, "loss": 0.8737, "step": 15452 }, { "epoch": 0.79, "learning_rate": 2.3269959155101997e-06, "loss": 0.874, "step": 15453 }, { "epoch": 0.79, "learning_rate": 2.3259408341638455e-06, "loss": 0.801, "step": 15454 }, { "epoch": 0.79, "learning_rate": 2.324885960584479e-06, "loss": 0.9703, "step": 15455 }, { "epoch": 0.79, "learning_rate": 2.323831294800661e-06, "loss": 0.8042, "step": 15456 }, { "epoch": 0.79, "learning_rate": 2.322776836840939e-06, "loss": 0.9105, "step": 15457 }, { "epoch": 0.79, "learning_rate": 2.321722586733872e-06, "loss": 0.9318, "step": 15458 }, { "epoch": 0.79, "learning_rate": 2.3206685445079935e-06, "loss": 0.9051, "step": 15459 }, { "epoch": 0.79, "learning_rate": 2.319614710191844e-06, "loss": 0.8357, "step": 15460 }, { "epoch": 0.79, "learning_rate": 2.3185610838139594e-06, "loss": 0.8753, "step": 15461 }, { "epoch": 0.79, "learning_rate": 2.317507665402856e-06, "loss": 0.905, "step": 15462 }, { "epoch": 0.79, "learning_rate": 2.3164544549870605e-06, "loss": 0.8469, "step": 15463 }, { "epoch": 0.79, "learning_rate": 2.315401452595085e-06, "loss": 0.8041, "step": 15464 }, { "epoch": 0.79, "learning_rate": 2.3143486582554432e-06, "loss": 0.8902, "step": 15465 }, { "epoch": 0.79, "learning_rate": 2.3132960719966314e-06, "loss": 0.8389, "step": 15466 }, { "epoch": 0.79, "learning_rate": 2.3122436938471503e-06, "loss": 0.8441, "step": 15467 }, { "epoch": 0.79, "learning_rate": 2.3111915238354954e-06, "loss": 0.8258, "step": 15468 }, { "epoch": 0.79, "learning_rate": 2.3101395619901434e-06, "loss": 0.8682, "step": 15469 }, { "epoch": 0.79, "learning_rate": 2.3090878083395872e-06, "loss": 0.8493, "step": 15470 }, { "epoch": 0.79, "learning_rate": 2.3080362629122932e-06, "loss": 0.8389, "step": 15471 }, { "epoch": 0.79, "learning_rate": 2.306984925736734e-06, "loss": 0.897, "step": 15472 }, { "epoch": 0.79, "learning_rate": 2.3059337968413765e-06, "loss": 0.8799, "step": 15473 }, { "epoch": 0.79, "learning_rate": 2.3048828762546713e-06, "loss": 0.9062, "step": 15474 }, { "epoch": 0.79, "learning_rate": 2.3038321640050763e-06, "loss": 0.8878, "step": 15475 }, { "epoch": 0.79, "learning_rate": 2.3027816601210374e-06, "loss": 0.7497, "step": 15476 }, { "epoch": 0.79, "learning_rate": 2.301731364630999e-06, "loss": 0.8915, "step": 15477 }, { "epoch": 0.79, "learning_rate": 2.3006812775633904e-06, "loss": 0.8396, "step": 15478 }, { "epoch": 0.79, "learning_rate": 2.2996313989466446e-06, "loss": 0.7646, "step": 15479 }, { "epoch": 0.79, "learning_rate": 2.2985817288091885e-06, "loss": 0.7936, "step": 15480 }, { "epoch": 0.79, "learning_rate": 2.297532267179433e-06, "loss": 0.9479, "step": 15481 }, { "epoch": 0.79, "learning_rate": 2.296483014085803e-06, "loss": 0.8592, "step": 15482 }, { "epoch": 0.79, "learning_rate": 2.2954339695566963e-06, "loss": 0.8688, "step": 15483 }, { "epoch": 0.79, "learning_rate": 2.294385133620518e-06, "loss": 0.856, "step": 15484 }, { "epoch": 0.79, "learning_rate": 2.293336506305667e-06, "loss": 0.8244, "step": 15485 }, { "epoch": 0.79, "learning_rate": 2.292288087640524e-06, "loss": 0.8785, "step": 15486 }, { "epoch": 0.79, "learning_rate": 2.2912398776534884e-06, "loss": 0.8707, "step": 15487 }, { "epoch": 0.79, "learning_rate": 2.290191876372927e-06, "loss": 0.856, "step": 15488 }, { "epoch": 0.79, "learning_rate": 2.2891440838272185e-06, "loss": 0.8913, "step": 15489 }, { "epoch": 0.79, "learning_rate": 2.2880965000447333e-06, "loss": 0.7833, "step": 15490 }, { "epoch": 0.79, "learning_rate": 2.2870491250538273e-06, "loss": 0.8591, "step": 15491 }, { "epoch": 0.79, "learning_rate": 2.2860019588828607e-06, "loss": 0.8682, "step": 15492 }, { "epoch": 0.79, "learning_rate": 2.284955001560184e-06, "loss": 0.8686, "step": 15493 }, { "epoch": 0.79, "learning_rate": 2.2839082531141453e-06, "loss": 0.8834, "step": 15494 }, { "epoch": 0.79, "learning_rate": 2.282861713573078e-06, "loss": 0.9114, "step": 15495 }, { "epoch": 0.79, "learning_rate": 2.28181538296532e-06, "loss": 0.9174, "step": 15496 }, { "epoch": 0.79, "learning_rate": 2.2807692613192023e-06, "loss": 0.7948, "step": 15497 }, { "epoch": 0.79, "learning_rate": 2.279723348663038e-06, "loss": 0.8188, "step": 15498 }, { "epoch": 0.79, "learning_rate": 2.2786776450251556e-06, "loss": 0.936, "step": 15499 }, { "epoch": 0.79, "learning_rate": 2.2776321504338583e-06, "loss": 0.9644, "step": 15500 }, { "epoch": 0.79, "learning_rate": 2.276586864917455e-06, "loss": 0.8786, "step": 15501 }, { "epoch": 0.79, "learning_rate": 2.2755417885042476e-06, "loss": 0.8459, "step": 15502 }, { "epoch": 0.79, "learning_rate": 2.2744969212225267e-06, "loss": 0.9031, "step": 15503 }, { "epoch": 0.79, "learning_rate": 2.2734522631005806e-06, "loss": 0.8442, "step": 15504 }, { "epoch": 0.79, "learning_rate": 2.272407814166695e-06, "loss": 0.796, "step": 15505 }, { "epoch": 0.79, "learning_rate": 2.271363574449149e-06, "loss": 0.8558, "step": 15506 }, { "epoch": 0.79, "learning_rate": 2.2703195439762084e-06, "loss": 0.7661, "step": 15507 }, { "epoch": 0.79, "learning_rate": 2.269275722776143e-06, "loss": 0.8371, "step": 15508 }, { "epoch": 0.79, "learning_rate": 2.2682321108772152e-06, "loss": 0.7673, "step": 15509 }, { "epoch": 0.79, "learning_rate": 2.2671887083076705e-06, "loss": 0.8183, "step": 15510 }, { "epoch": 0.79, "learning_rate": 2.2661455150957713e-06, "loss": 0.8387, "step": 15511 }, { "epoch": 0.79, "learning_rate": 2.2651025312697504e-06, "loss": 0.9691, "step": 15512 }, { "epoch": 0.79, "learning_rate": 2.2640597568578506e-06, "loss": 0.8268, "step": 15513 }, { "epoch": 0.79, "learning_rate": 2.2630171918883038e-06, "loss": 0.8177, "step": 15514 }, { "epoch": 0.79, "learning_rate": 2.2619748363893335e-06, "loss": 0.8763, "step": 15515 }, { "epoch": 0.79, "learning_rate": 2.2609326903891604e-06, "loss": 0.8118, "step": 15516 }, { "epoch": 0.79, "learning_rate": 2.259890753916002e-06, "loss": 0.8984, "step": 15517 }, { "epoch": 0.79, "learning_rate": 2.258849026998069e-06, "loss": 0.8889, "step": 15518 }, { "epoch": 0.79, "learning_rate": 2.257807509663561e-06, "loss": 0.86, "step": 15519 }, { "epoch": 0.79, "learning_rate": 2.2567662019406765e-06, "loss": 0.9871, "step": 15520 }, { "epoch": 0.79, "learning_rate": 2.255725103857611e-06, "loss": 0.8716, "step": 15521 }, { "epoch": 0.79, "learning_rate": 2.2546842154425453e-06, "loss": 0.8298, "step": 15522 }, { "epoch": 0.79, "learning_rate": 2.253643536723669e-06, "loss": 0.8511, "step": 15523 }, { "epoch": 0.79, "learning_rate": 2.2526030677291523e-06, "loss": 0.8535, "step": 15524 }, { "epoch": 0.79, "learning_rate": 2.2515628084871587e-06, "loss": 0.8808, "step": 15525 }, { "epoch": 0.79, "learning_rate": 2.250522759025865e-06, "loss": 0.8321, "step": 15526 }, { "epoch": 0.79, "learning_rate": 2.249482919373419e-06, "loss": 0.8614, "step": 15527 }, { "epoch": 0.79, "learning_rate": 2.2484432895579776e-06, "loss": 0.7895, "step": 15528 }, { "epoch": 0.79, "learning_rate": 2.247403869607687e-06, "loss": 0.9284, "step": 15529 }, { "epoch": 0.79, "learning_rate": 2.2463646595506917e-06, "loss": 0.8191, "step": 15530 }, { "epoch": 0.79, "learning_rate": 2.2453256594151206e-06, "loss": 1.0022, "step": 15531 }, { "epoch": 0.79, "learning_rate": 2.2442868692291075e-06, "loss": 0.9977, "step": 15532 }, { "epoch": 0.79, "learning_rate": 2.2432482890207787e-06, "loss": 0.8367, "step": 15533 }, { "epoch": 0.79, "learning_rate": 2.2422099188182445e-06, "loss": 0.8962, "step": 15534 }, { "epoch": 0.79, "learning_rate": 2.2411717586496286e-06, "loss": 0.7784, "step": 15535 }, { "epoch": 0.79, "learning_rate": 2.2401338085430326e-06, "loss": 0.8925, "step": 15536 }, { "epoch": 0.79, "learning_rate": 2.2390960685265517e-06, "loss": 0.8465, "step": 15537 }, { "epoch": 0.79, "learning_rate": 2.238058538628294e-06, "loss": 0.9248, "step": 15538 }, { "epoch": 0.79, "learning_rate": 2.2370212188763407e-06, "loss": 0.897, "step": 15539 }, { "epoch": 0.79, "learning_rate": 2.235984109298778e-06, "loss": 0.7914, "step": 15540 }, { "epoch": 0.79, "learning_rate": 2.2349472099236856e-06, "loss": 0.9444, "step": 15541 }, { "epoch": 0.79, "learning_rate": 2.233910520779139e-06, "loss": 0.8829, "step": 15542 }, { "epoch": 0.79, "learning_rate": 2.2328740418932005e-06, "loss": 0.9252, "step": 15543 }, { "epoch": 0.79, "learning_rate": 2.2318377732939324e-06, "loss": 0.8649, "step": 15544 }, { "epoch": 0.79, "learning_rate": 2.230801715009395e-06, "loss": 0.7864, "step": 15545 }, { "epoch": 0.79, "learning_rate": 2.2297658670676303e-06, "loss": 0.9571, "step": 15546 }, { "epoch": 0.79, "learning_rate": 2.228730229496694e-06, "loss": 0.9234, "step": 15547 }, { "epoch": 0.79, "learning_rate": 2.227694802324618e-06, "loss": 0.8686, "step": 15548 }, { "epoch": 0.79, "learning_rate": 2.2266595855794303e-06, "loss": 0.9586, "step": 15549 }, { "epoch": 0.79, "learning_rate": 2.2256245792891697e-06, "loss": 0.8889, "step": 15550 }, { "epoch": 0.79, "learning_rate": 2.2245897834818497e-06, "loss": 0.8803, "step": 15551 }, { "epoch": 0.79, "learning_rate": 2.2235551981854897e-06, "loss": 0.8242, "step": 15552 }, { "epoch": 0.79, "learning_rate": 2.2225208234280982e-06, "loss": 0.826, "step": 15553 }, { "epoch": 0.79, "learning_rate": 2.2214866592376815e-06, "loss": 0.8402, "step": 15554 }, { "epoch": 0.79, "learning_rate": 2.2204527056422407e-06, "loss": 0.9143, "step": 15555 }, { "epoch": 0.79, "learning_rate": 2.2194189626697628e-06, "loss": 0.8658, "step": 15556 }, { "epoch": 0.79, "learning_rate": 2.2183854303482386e-06, "loss": 0.9301, "step": 15557 }, { "epoch": 0.79, "learning_rate": 2.2173521087056505e-06, "loss": 0.7693, "step": 15558 }, { "epoch": 0.79, "learning_rate": 2.2163189977699763e-06, "loss": 0.7404, "step": 15559 }, { "epoch": 0.79, "learning_rate": 2.2152860975691816e-06, "loss": 0.8548, "step": 15560 }, { "epoch": 0.79, "learning_rate": 2.2142534081312326e-06, "loss": 0.8589, "step": 15561 }, { "epoch": 0.79, "learning_rate": 2.2132209294840923e-06, "loss": 0.8129, "step": 15562 }, { "epoch": 0.79, "learning_rate": 2.212188661655705e-06, "loss": 0.9478, "step": 15563 }, { "epoch": 0.79, "learning_rate": 2.2111566046740305e-06, "loss": 0.9545, "step": 15564 }, { "epoch": 0.79, "learning_rate": 2.2101247585670047e-06, "loss": 0.7799, "step": 15565 }, { "epoch": 0.79, "learning_rate": 2.2090931233625567e-06, "loss": 0.7363, "step": 15566 }, { "epoch": 0.79, "learning_rate": 2.2080616990886295e-06, "loss": 0.7973, "step": 15567 }, { "epoch": 0.79, "learning_rate": 2.2070304857731396e-06, "loss": 0.8653, "step": 15568 }, { "epoch": 0.79, "learning_rate": 2.2059994834440067e-06, "loss": 0.8687, "step": 15569 }, { "epoch": 0.79, "learning_rate": 2.2049686921291467e-06, "loss": 0.821, "step": 15570 }, { "epoch": 0.79, "learning_rate": 2.2039381118564685e-06, "loss": 0.9222, "step": 15571 }, { "epoch": 0.79, "learning_rate": 2.2029077426538692e-06, "loss": 0.8541, "step": 15572 }, { "epoch": 0.79, "learning_rate": 2.2018775845492467e-06, "loss": 0.9622, "step": 15573 }, { "epoch": 0.79, "learning_rate": 2.200847637570496e-06, "loss": 0.8475, "step": 15574 }, { "epoch": 0.79, "learning_rate": 2.199817901745491e-06, "loss": 0.9137, "step": 15575 }, { "epoch": 0.79, "learning_rate": 2.1987883771021235e-06, "loss": 0.8783, "step": 15576 }, { "epoch": 0.79, "learning_rate": 2.197759063668261e-06, "loss": 0.863, "step": 15577 }, { "epoch": 0.79, "learning_rate": 2.1967299614717654e-06, "loss": 0.8134, "step": 15578 }, { "epoch": 0.79, "learning_rate": 2.1957010705405102e-06, "loss": 0.8661, "step": 15579 }, { "epoch": 0.79, "learning_rate": 2.1946723909023416e-06, "loss": 0.9662, "step": 15580 }, { "epoch": 0.79, "learning_rate": 2.1936439225851138e-06, "loss": 0.8651, "step": 15581 }, { "epoch": 0.79, "learning_rate": 2.192615665616672e-06, "loss": 0.785, "step": 15582 }, { "epoch": 0.79, "learning_rate": 2.1915876200248578e-06, "loss": 0.8479, "step": 15583 }, { "epoch": 0.79, "learning_rate": 2.190559785837497e-06, "loss": 0.7975, "step": 15584 }, { "epoch": 0.79, "learning_rate": 2.189532163082423e-06, "loss": 0.7899, "step": 15585 }, { "epoch": 0.79, "learning_rate": 2.188504751787458e-06, "loss": 0.9531, "step": 15586 }, { "epoch": 0.79, "learning_rate": 2.1874775519804116e-06, "loss": 0.8095, "step": 15587 }, { "epoch": 0.79, "learning_rate": 2.1864505636891033e-06, "loss": 0.9456, "step": 15588 }, { "epoch": 0.79, "learning_rate": 2.185423786941332e-06, "loss": 0.8725, "step": 15589 }, { "epoch": 0.79, "learning_rate": 2.1843972217648934e-06, "loss": 0.8998, "step": 15590 }, { "epoch": 0.79, "learning_rate": 2.1833708681875898e-06, "loss": 0.8242, "step": 15591 }, { "epoch": 0.79, "learning_rate": 2.182344726237201e-06, "loss": 0.837, "step": 15592 }, { "epoch": 0.79, "learning_rate": 2.1813187959415126e-06, "loss": 0.8711, "step": 15593 }, { "epoch": 0.79, "learning_rate": 2.1802930773283005e-06, "loss": 0.8584, "step": 15594 }, { "epoch": 0.79, "learning_rate": 2.179267570425332e-06, "loss": 0.8328, "step": 15595 }, { "epoch": 0.79, "learning_rate": 2.1782422752603727e-06, "loss": 0.8369, "step": 15596 }, { "epoch": 0.79, "learning_rate": 2.177217191861183e-06, "loss": 0.8941, "step": 15597 }, { "epoch": 0.79, "learning_rate": 2.176192320255517e-06, "loss": 0.8134, "step": 15598 }, { "epoch": 0.79, "learning_rate": 2.175167660471116e-06, "loss": 0.8472, "step": 15599 }, { "epoch": 0.79, "learning_rate": 2.17414321253573e-06, "loss": 0.889, "step": 15600 }, { "epoch": 0.79, "learning_rate": 2.17311897647709e-06, "loss": 0.8911, "step": 15601 }, { "epoch": 0.79, "learning_rate": 2.172094952322922e-06, "loss": 0.8055, "step": 15602 }, { "epoch": 0.79, "learning_rate": 2.1710711401009597e-06, "loss": 0.9593, "step": 15603 }, { "epoch": 0.79, "learning_rate": 2.1700475398389143e-06, "loss": 0.8868, "step": 15604 }, { "epoch": 0.79, "learning_rate": 2.1690241515645017e-06, "loss": 1.0156, "step": 15605 }, { "epoch": 0.79, "learning_rate": 2.168000975305431e-06, "loss": 0.8552, "step": 15606 }, { "epoch": 0.79, "learning_rate": 2.166978011089399e-06, "loss": 0.8576, "step": 15607 }, { "epoch": 0.79, "learning_rate": 2.1659552589441036e-06, "loss": 0.9254, "step": 15608 }, { "epoch": 0.79, "learning_rate": 2.1649327188972346e-06, "loss": 0.8074, "step": 15609 }, { "epoch": 0.79, "learning_rate": 2.163910390976479e-06, "loss": 0.8762, "step": 15610 }, { "epoch": 0.79, "learning_rate": 2.1628882752095072e-06, "loss": 0.8752, "step": 15611 }, { "epoch": 0.79, "learning_rate": 2.161866371624003e-06, "loss": 0.8565, "step": 15612 }, { "epoch": 0.79, "learning_rate": 2.160844680247628e-06, "loss": 0.9836, "step": 15613 }, { "epoch": 0.79, "learning_rate": 2.1598232011080368e-06, "loss": 0.855, "step": 15614 }, { "epoch": 0.79, "learning_rate": 2.158801934232897e-06, "loss": 0.8621, "step": 15615 }, { "epoch": 0.79, "learning_rate": 2.15778087964985e-06, "loss": 0.8815, "step": 15616 }, { "epoch": 0.79, "learning_rate": 2.1567600373865427e-06, "loss": 0.8253, "step": 15617 }, { "epoch": 0.79, "learning_rate": 2.155739407470614e-06, "loss": 0.983, "step": 15618 }, { "epoch": 0.79, "learning_rate": 2.1547189899296937e-06, "loss": 0.8197, "step": 15619 }, { "epoch": 0.79, "learning_rate": 2.153698784791409e-06, "loss": 0.8357, "step": 15620 }, { "epoch": 0.79, "learning_rate": 2.1526787920833835e-06, "loss": 0.8501, "step": 15621 }, { "epoch": 0.79, "learning_rate": 2.1516590118332324e-06, "loss": 0.9267, "step": 15622 }, { "epoch": 0.79, "learning_rate": 2.1506394440685586e-06, "loss": 0.8736, "step": 15623 }, { "epoch": 0.79, "learning_rate": 2.149620088816976e-06, "loss": 0.884, "step": 15624 }, { "epoch": 0.79, "learning_rate": 2.1486009461060743e-06, "loss": 0.8584, "step": 15625 }, { "epoch": 0.79, "learning_rate": 2.147582015963449e-06, "loss": 0.8564, "step": 15626 }, { "epoch": 0.79, "learning_rate": 2.1465632984166894e-06, "loss": 0.8405, "step": 15627 }, { "epoch": 0.79, "learning_rate": 2.1455447934933673e-06, "loss": 0.9337, "step": 15628 }, { "epoch": 0.79, "learning_rate": 2.1445265012210683e-06, "loss": 0.8832, "step": 15629 }, { "epoch": 0.79, "learning_rate": 2.1435084216273573e-06, "loss": 0.7617, "step": 15630 }, { "epoch": 0.79, "learning_rate": 2.1424905547397923e-06, "loss": 0.804, "step": 15631 }, { "epoch": 0.79, "learning_rate": 2.1414729005859402e-06, "loss": 0.9108, "step": 15632 }, { "epoch": 0.79, "learning_rate": 2.140455459193346e-06, "loss": 0.8403, "step": 15633 }, { "epoch": 0.79, "learning_rate": 2.139438230589559e-06, "loss": 0.8787, "step": 15634 }, { "epoch": 0.79, "learning_rate": 2.138421214802121e-06, "loss": 0.8988, "step": 15635 }, { "epoch": 0.79, "learning_rate": 2.1374044118585624e-06, "loss": 0.9245, "step": 15636 }, { "epoch": 0.79, "learning_rate": 2.136387821786413e-06, "loss": 0.8493, "step": 15637 }, { "epoch": 0.79, "learning_rate": 2.135371444613199e-06, "loss": 0.8602, "step": 15638 }, { "epoch": 0.79, "learning_rate": 2.134355280366437e-06, "loss": 0.9206, "step": 15639 }, { "epoch": 0.79, "learning_rate": 2.133339329073633e-06, "loss": 0.9177, "step": 15640 }, { "epoch": 0.79, "learning_rate": 2.132323590762302e-06, "loss": 0.8265, "step": 15641 }, { "epoch": 0.79, "learning_rate": 2.1313080654599382e-06, "loss": 0.924, "step": 15642 }, { "epoch": 0.79, "learning_rate": 2.1302927531940333e-06, "loss": 0.8458, "step": 15643 }, { "epoch": 0.79, "learning_rate": 2.1292776539920846e-06, "loss": 0.8314, "step": 15644 }, { "epoch": 0.79, "learning_rate": 2.128262767881566e-06, "loss": 0.8264, "step": 15645 }, { "epoch": 0.79, "learning_rate": 2.127248094889959e-06, "loss": 0.806, "step": 15646 }, { "epoch": 0.79, "learning_rate": 2.1262336350447354e-06, "loss": 0.825, "step": 15647 }, { "epoch": 0.79, "learning_rate": 2.1252193883733573e-06, "loss": 0.8146, "step": 15648 }, { "epoch": 0.8, "learning_rate": 2.1242053549032858e-06, "loss": 0.8842, "step": 15649 }, { "epoch": 0.8, "learning_rate": 2.123191534661976e-06, "loss": 0.8556, "step": 15650 }, { "epoch": 0.8, "learning_rate": 2.122177927676876e-06, "loss": 0.8776, "step": 15651 }, { "epoch": 0.8, "learning_rate": 2.1211645339754238e-06, "loss": 1.0158, "step": 15652 }, { "epoch": 0.8, "learning_rate": 2.120151353585064e-06, "loss": 0.8366, "step": 15653 }, { "epoch": 0.8, "learning_rate": 2.1191383865332228e-06, "loss": 0.8941, "step": 15654 }, { "epoch": 0.8, "learning_rate": 2.11812563284732e-06, "loss": 0.8077, "step": 15655 }, { "epoch": 0.8, "learning_rate": 2.1171130925547846e-06, "loss": 0.9153, "step": 15656 }, { "epoch": 0.8, "learning_rate": 2.1161007656830233e-06, "loss": 0.8761, "step": 15657 }, { "epoch": 0.8, "learning_rate": 2.115088652259446e-06, "loss": 0.7518, "step": 15658 }, { "epoch": 0.8, "learning_rate": 2.114076752311458e-06, "loss": 0.83, "step": 15659 }, { "epoch": 0.8, "learning_rate": 2.1130650658664486e-06, "loss": 0.9128, "step": 15660 }, { "epoch": 0.8, "learning_rate": 2.1120535929518116e-06, "loss": 0.848, "step": 15661 }, { "epoch": 0.8, "learning_rate": 2.1110423335949313e-06, "loss": 0.8616, "step": 15662 }, { "epoch": 0.8, "learning_rate": 2.1100312878231888e-06, "loss": 0.8552, "step": 15663 }, { "epoch": 0.8, "learning_rate": 2.109020455663954e-06, "loss": 0.9118, "step": 15664 }, { "epoch": 0.8, "learning_rate": 2.1080098371445936e-06, "loss": 0.8738, "step": 15665 }, { "epoch": 0.8, "learning_rate": 2.1069994322924724e-06, "loss": 0.8386, "step": 15666 }, { "epoch": 0.8, "learning_rate": 2.1059892411349405e-06, "loss": 0.9091, "step": 15667 }, { "epoch": 0.8, "learning_rate": 2.104979263699356e-06, "loss": 0.8024, "step": 15668 }, { "epoch": 0.8, "learning_rate": 2.1039695000130557e-06, "loss": 0.8762, "step": 15669 }, { "epoch": 0.8, "learning_rate": 2.1029599501033793e-06, "loss": 0.822, "step": 15670 }, { "epoch": 0.8, "learning_rate": 2.101950613997664e-06, "loss": 0.888, "step": 15671 }, { "epoch": 0.8, "learning_rate": 2.1009414917232305e-06, "loss": 0.8163, "step": 15672 }, { "epoch": 0.8, "learning_rate": 2.0999325833074024e-06, "loss": 0.7973, "step": 15673 }, { "epoch": 0.8, "learning_rate": 2.0989238887774943e-06, "loss": 0.9752, "step": 15674 }, { "epoch": 0.8, "learning_rate": 2.0979154081608188e-06, "loss": 0.9315, "step": 15675 }, { "epoch": 0.8, "learning_rate": 2.0969071414846733e-06, "loss": 0.8909, "step": 15676 }, { "epoch": 0.8, "learning_rate": 2.0958990887763584e-06, "loss": 0.8853, "step": 15677 }, { "epoch": 0.8, "learning_rate": 2.0948912500631703e-06, "loss": 0.7532, "step": 15678 }, { "epoch": 0.8, "learning_rate": 2.0938836253723847e-06, "loss": 0.9786, "step": 15679 }, { "epoch": 0.8, "learning_rate": 2.092876214731294e-06, "loss": 0.952, "step": 15680 }, { "epoch": 0.8, "learning_rate": 2.0918690181671652e-06, "loss": 0.8425, "step": 15681 }, { "epoch": 0.8, "learning_rate": 2.0908620357072685e-06, "loss": 0.8926, "step": 15682 }, { "epoch": 0.8, "learning_rate": 2.0898552673788708e-06, "loss": 0.9332, "step": 15683 }, { "epoch": 0.8, "learning_rate": 2.088848713209223e-06, "loss": 0.8981, "step": 15684 }, { "epoch": 0.8, "learning_rate": 2.0878423732255793e-06, "loss": 0.8687, "step": 15685 }, { "epoch": 0.8, "learning_rate": 2.086836247455186e-06, "loss": 0.7744, "step": 15686 }, { "epoch": 0.8, "learning_rate": 2.085830335925284e-06, "loss": 0.8497, "step": 15687 }, { "epoch": 0.8, "learning_rate": 2.0848246386631046e-06, "loss": 0.7983, "step": 15688 }, { "epoch": 0.8, "learning_rate": 2.0838191556958752e-06, "loss": 0.9874, "step": 15689 }, { "epoch": 0.8, "learning_rate": 2.082813887050824e-06, "loss": 0.8962, "step": 15690 }, { "epoch": 0.8, "learning_rate": 2.081808832755158e-06, "loss": 0.8635, "step": 15691 }, { "epoch": 0.8, "learning_rate": 2.080803992836099e-06, "loss": 0.8637, "step": 15692 }, { "epoch": 0.8, "learning_rate": 2.079799367320843e-06, "loss": 0.876, "step": 15693 }, { "epoch": 0.8, "learning_rate": 2.078794956236594e-06, "loss": 0.889, "step": 15694 }, { "epoch": 0.8, "learning_rate": 2.077790759610545e-06, "loss": 0.8788, "step": 15695 }, { "epoch": 0.8, "learning_rate": 2.076786777469879e-06, "loss": 0.8448, "step": 15696 }, { "epoch": 0.8, "learning_rate": 2.075783009841786e-06, "loss": 0.8183, "step": 15697 }, { "epoch": 0.8, "learning_rate": 2.074779456753434e-06, "loss": 0.8109, "step": 15698 }, { "epoch": 0.8, "learning_rate": 2.073776118231997e-06, "loss": 0.82, "step": 15699 }, { "epoch": 0.8, "learning_rate": 2.0727729943046426e-06, "loss": 0.7927, "step": 15700 }, { "epoch": 0.8, "learning_rate": 2.071770084998521e-06, "loss": 0.8876, "step": 15701 }, { "epoch": 0.8, "learning_rate": 2.07076739034079e-06, "loss": 0.8656, "step": 15702 }, { "epoch": 0.8, "learning_rate": 2.069764910358596e-06, "loss": 0.7584, "step": 15703 }, { "epoch": 0.8, "learning_rate": 2.068762645079083e-06, "loss": 0.8207, "step": 15704 }, { "epoch": 0.8, "learning_rate": 2.06776059452938e-06, "loss": 0.8849, "step": 15705 }, { "epoch": 0.8, "learning_rate": 2.06675875873662e-06, "loss": 0.7849, "step": 15706 }, { "epoch": 0.8, "learning_rate": 2.0657571377279283e-06, "loss": 0.927, "step": 15707 }, { "epoch": 0.8, "learning_rate": 2.0647557315304156e-06, "loss": 0.853, "step": 15708 }, { "epoch": 0.8, "learning_rate": 2.0637545401712043e-06, "loss": 0.898, "step": 15709 }, { "epoch": 0.8, "learning_rate": 2.062753563677393e-06, "loss": 0.9011, "step": 15710 }, { "epoch": 0.8, "learning_rate": 2.0617528020760847e-06, "loss": 0.8077, "step": 15711 }, { "epoch": 0.8, "learning_rate": 2.060752255394376e-06, "loss": 0.8661, "step": 15712 }, { "epoch": 0.8, "learning_rate": 2.059751923659351e-06, "loss": 0.8022, "step": 15713 }, { "epoch": 0.8, "learning_rate": 2.0587518068980938e-06, "loss": 0.9335, "step": 15714 }, { "epoch": 0.8, "learning_rate": 2.057751905137684e-06, "loss": 0.8323, "step": 15715 }, { "epoch": 0.8, "learning_rate": 2.0567522184051935e-06, "loss": 0.8485, "step": 15716 }, { "epoch": 0.8, "learning_rate": 2.055752746727683e-06, "loss": 0.8597, "step": 15717 }, { "epoch": 0.8, "learning_rate": 2.0547534901322153e-06, "loss": 0.8734, "step": 15718 }, { "epoch": 0.8, "learning_rate": 2.053754448645846e-06, "loss": 0.8508, "step": 15719 }, { "epoch": 0.8, "learning_rate": 2.0527556222956147e-06, "loss": 0.9117, "step": 15720 }, { "epoch": 0.8, "learning_rate": 2.051757011108576e-06, "loss": 0.8178, "step": 15721 }, { "epoch": 0.8, "learning_rate": 2.0507586151117573e-06, "loss": 0.9263, "step": 15722 }, { "epoch": 0.8, "learning_rate": 2.049760434332192e-06, "loss": 0.7727, "step": 15723 }, { "epoch": 0.8, "learning_rate": 2.0487624687969064e-06, "loss": 0.8822, "step": 15724 }, { "epoch": 0.8, "learning_rate": 2.0477647185329153e-06, "loss": 0.8946, "step": 15725 }, { "epoch": 0.8, "learning_rate": 2.0467671835672344e-06, "loss": 0.878, "step": 15726 }, { "epoch": 0.8, "learning_rate": 2.0457698639268696e-06, "loss": 0.9176, "step": 15727 }, { "epoch": 0.8, "learning_rate": 2.0447727596388266e-06, "loss": 0.8315, "step": 15728 }, { "epoch": 0.8, "learning_rate": 2.0437758707300947e-06, "loss": 0.8491, "step": 15729 }, { "epoch": 0.8, "learning_rate": 2.0427791972276655e-06, "loss": 0.8806, "step": 15730 }, { "epoch": 0.8, "learning_rate": 2.0417827391585276e-06, "loss": 0.8723, "step": 15731 }, { "epoch": 0.8, "learning_rate": 2.0407864965496483e-06, "loss": 0.9111, "step": 15732 }, { "epoch": 0.8, "learning_rate": 2.0397904694280134e-06, "loss": 0.8056, "step": 15733 }, { "epoch": 0.8, "learning_rate": 2.0387946578205787e-06, "loss": 0.9148, "step": 15734 }, { "epoch": 0.8, "learning_rate": 2.0377990617543096e-06, "loss": 0.8384, "step": 15735 }, { "epoch": 0.8, "learning_rate": 2.036803681256162e-06, "loss": 0.9825, "step": 15736 }, { "epoch": 0.8, "learning_rate": 2.0358085163530794e-06, "loss": 0.8354, "step": 15737 }, { "epoch": 0.8, "learning_rate": 2.0348135670720094e-06, "loss": 0.86, "step": 15738 }, { "epoch": 0.8, "learning_rate": 2.0338188334398866e-06, "loss": 0.8186, "step": 15739 }, { "epoch": 0.8, "learning_rate": 2.0328243154836457e-06, "loss": 0.8638, "step": 15740 }, { "epoch": 0.8, "learning_rate": 2.0318300132302084e-06, "loss": 0.9044, "step": 15741 }, { "epoch": 0.8, "learning_rate": 2.030835926706496e-06, "loss": 0.9541, "step": 15742 }, { "epoch": 0.8, "learning_rate": 2.029842055939425e-06, "loss": 0.7808, "step": 15743 }, { "epoch": 0.8, "learning_rate": 2.028848400955895e-06, "loss": 0.8864, "step": 15744 }, { "epoch": 0.8, "learning_rate": 2.027854961782819e-06, "loss": 0.8753, "step": 15745 }, { "epoch": 0.8, "learning_rate": 2.026861738447087e-06, "loss": 0.8448, "step": 15746 }, { "epoch": 0.8, "learning_rate": 2.02586873097559e-06, "loss": 0.8714, "step": 15747 }, { "epoch": 0.8, "learning_rate": 2.0248759393952154e-06, "loss": 0.8344, "step": 15748 }, { "epoch": 0.8, "learning_rate": 2.023883363732837e-06, "loss": 0.8813, "step": 15749 }, { "epoch": 0.8, "learning_rate": 2.0228910040153314e-06, "loss": 0.7541, "step": 15750 }, { "epoch": 0.8, "learning_rate": 2.0218988602695646e-06, "loss": 0.9201, "step": 15751 }, { "epoch": 0.8, "learning_rate": 2.0209069325224005e-06, "loss": 0.7906, "step": 15752 }, { "epoch": 0.8, "learning_rate": 2.01991522080069e-06, "loss": 0.8234, "step": 15753 }, { "epoch": 0.8, "learning_rate": 2.0189237251312844e-06, "loss": 0.8912, "step": 15754 }, { "epoch": 0.8, "learning_rate": 2.017932445541031e-06, "loss": 0.88, "step": 15755 }, { "epoch": 0.8, "learning_rate": 2.0169413820567585e-06, "loss": 0.8117, "step": 15756 }, { "epoch": 0.8, "learning_rate": 2.0159505347053097e-06, "loss": 0.8761, "step": 15757 }, { "epoch": 0.8, "learning_rate": 2.0149599035135037e-06, "loss": 0.8343, "step": 15758 }, { "epoch": 0.8, "learning_rate": 2.013969488508163e-06, "loss": 0.9131, "step": 15759 }, { "epoch": 0.8, "learning_rate": 2.012979289716104e-06, "loss": 0.8814, "step": 15760 }, { "epoch": 0.8, "learning_rate": 2.0119893071641307e-06, "loss": 0.9012, "step": 15761 }, { "epoch": 0.8, "learning_rate": 2.010999540879048e-06, "loss": 0.8774, "step": 15762 }, { "epoch": 0.8, "learning_rate": 2.010009990887655e-06, "loss": 0.8417, "step": 15763 }, { "epoch": 0.8, "learning_rate": 2.0090206572167392e-06, "loss": 0.8374, "step": 15764 }, { "epoch": 0.8, "learning_rate": 2.0080315398930906e-06, "loss": 0.7288, "step": 15765 }, { "epoch": 0.8, "learning_rate": 2.0070426389434815e-06, "loss": 0.8682, "step": 15766 }, { "epoch": 0.8, "learning_rate": 2.00605395439469e-06, "loss": 0.7518, "step": 15767 }, { "epoch": 0.8, "learning_rate": 2.0050654862734833e-06, "loss": 0.9272, "step": 15768 }, { "epoch": 0.8, "learning_rate": 2.004077234606625e-06, "loss": 0.9148, "step": 15769 }, { "epoch": 0.8, "learning_rate": 2.0030891994208655e-06, "loss": 0.8043, "step": 15770 }, { "epoch": 0.8, "learning_rate": 2.002101380742958e-06, "loss": 0.8923, "step": 15771 }, { "epoch": 0.8, "learning_rate": 2.0011137785996493e-06, "loss": 0.9041, "step": 15772 }, { "epoch": 0.8, "learning_rate": 2.0001263930176696e-06, "loss": 0.8839, "step": 15773 }, { "epoch": 0.8, "learning_rate": 1.999139224023762e-06, "loss": 0.8459, "step": 15774 }, { "epoch": 0.8, "learning_rate": 1.9981522716446446e-06, "loss": 0.8737, "step": 15775 }, { "epoch": 0.8, "learning_rate": 1.9971655359070407e-06, "loss": 0.7296, "step": 15776 }, { "epoch": 0.8, "learning_rate": 1.996179016837669e-06, "loss": 0.969, "step": 15777 }, { "epoch": 0.8, "learning_rate": 1.995192714463232e-06, "loss": 0.9423, "step": 15778 }, { "epoch": 0.8, "learning_rate": 1.9942066288104367e-06, "loss": 0.8336, "step": 15779 }, { "epoch": 0.8, "learning_rate": 1.9932207599059782e-06, "loss": 0.9637, "step": 15780 }, { "epoch": 0.8, "learning_rate": 1.9922351077765523e-06, "loss": 0.8214, "step": 15781 }, { "epoch": 0.8, "learning_rate": 1.9912496724488397e-06, "loss": 0.8918, "step": 15782 }, { "epoch": 0.8, "learning_rate": 1.9902644539495196e-06, "loss": 0.9163, "step": 15783 }, { "epoch": 0.8, "learning_rate": 1.9892794523052717e-06, "loss": 0.8635, "step": 15784 }, { "epoch": 0.8, "learning_rate": 1.988294667542754e-06, "loss": 0.8731, "step": 15785 }, { "epoch": 0.8, "learning_rate": 1.98731009968864e-06, "loss": 0.874, "step": 15786 }, { "epoch": 0.8, "learning_rate": 1.9863257487695775e-06, "loss": 0.7608, "step": 15787 }, { "epoch": 0.8, "learning_rate": 1.9853416148122197e-06, "loss": 0.8815, "step": 15788 }, { "epoch": 0.8, "learning_rate": 1.9843576978432134e-06, "loss": 0.8578, "step": 15789 }, { "epoch": 0.8, "learning_rate": 1.983373997889191e-06, "loss": 0.8923, "step": 15790 }, { "epoch": 0.8, "learning_rate": 1.98239051497679e-06, "loss": 0.961, "step": 15791 }, { "epoch": 0.8, "learning_rate": 1.981407249132634e-06, "loss": 0.8599, "step": 15792 }, { "epoch": 0.8, "learning_rate": 1.9804242003833496e-06, "loss": 0.8828, "step": 15793 }, { "epoch": 0.8, "learning_rate": 1.9794413687555447e-06, "loss": 0.7779, "step": 15794 }, { "epoch": 0.8, "learning_rate": 1.9784587542758315e-06, "loss": 0.8303, "step": 15795 }, { "epoch": 0.8, "learning_rate": 1.9774763569708157e-06, "loss": 0.9483, "step": 15796 }, { "epoch": 0.8, "learning_rate": 1.9764941768670874e-06, "loss": 0.8526, "step": 15797 }, { "epoch": 0.8, "learning_rate": 1.9755122139912476e-06, "loss": 0.9346, "step": 15798 }, { "epoch": 0.8, "learning_rate": 1.974530468369875e-06, "loss": 0.8137, "step": 15799 }, { "epoch": 0.8, "learning_rate": 1.9735489400295516e-06, "loss": 0.8422, "step": 15800 }, { "epoch": 0.8, "learning_rate": 1.9725676289968543e-06, "loss": 0.9415, "step": 15801 }, { "epoch": 0.8, "learning_rate": 1.9715865352983443e-06, "loss": 0.8548, "step": 15802 }, { "epoch": 0.8, "learning_rate": 1.970605658960588e-06, "loss": 0.7621, "step": 15803 }, { "epoch": 0.8, "learning_rate": 1.96962500001014e-06, "loss": 0.8278, "step": 15804 }, { "epoch": 0.8, "learning_rate": 1.968644558473556e-06, "loss": 0.947, "step": 15805 }, { "epoch": 0.8, "learning_rate": 1.9676643343773716e-06, "loss": 0.8952, "step": 15806 }, { "epoch": 0.8, "learning_rate": 1.9666843277481297e-06, "loss": 0.9608, "step": 15807 }, { "epoch": 0.8, "learning_rate": 1.9657045386123665e-06, "loss": 0.9165, "step": 15808 }, { "epoch": 0.8, "learning_rate": 1.9647249669965984e-06, "loss": 0.993, "step": 15809 }, { "epoch": 0.8, "learning_rate": 1.963745612927359e-06, "loss": 0.8344, "step": 15810 }, { "epoch": 0.8, "learning_rate": 1.962766476431155e-06, "loss": 0.842, "step": 15811 }, { "epoch": 0.8, "learning_rate": 1.961787557534497e-06, "loss": 0.927, "step": 15812 }, { "epoch": 0.8, "learning_rate": 1.9608088562638917e-06, "loss": 0.9484, "step": 15813 }, { "epoch": 0.8, "learning_rate": 1.959830372645831e-06, "loss": 0.9529, "step": 15814 }, { "epoch": 0.8, "learning_rate": 1.958852106706809e-06, "loss": 0.9114, "step": 15815 }, { "epoch": 0.8, "learning_rate": 1.9578740584733115e-06, "loss": 0.9254, "step": 15816 }, { "epoch": 0.8, "learning_rate": 1.9568962279718194e-06, "loss": 0.9535, "step": 15817 }, { "epoch": 0.8, "learning_rate": 1.9559186152288036e-06, "loss": 0.7287, "step": 15818 }, { "epoch": 0.8, "learning_rate": 1.9549412202707318e-06, "loss": 0.8539, "step": 15819 }, { "epoch": 0.8, "learning_rate": 1.9539640431240704e-06, "loss": 0.8228, "step": 15820 }, { "epoch": 0.8, "learning_rate": 1.9529870838152663e-06, "loss": 0.8517, "step": 15821 }, { "epoch": 0.8, "learning_rate": 1.9520103423707816e-06, "loss": 0.9034, "step": 15822 }, { "epoch": 0.8, "learning_rate": 1.9510338188170506e-06, "loss": 0.8185, "step": 15823 }, { "epoch": 0.8, "learning_rate": 1.950057513180517e-06, "loss": 0.8294, "step": 15824 }, { "epoch": 0.8, "learning_rate": 1.9490814254876135e-06, "loss": 0.8768, "step": 15825 }, { "epoch": 0.8, "learning_rate": 1.9481055557647623e-06, "loss": 0.9126, "step": 15826 }, { "epoch": 0.8, "learning_rate": 1.947129904038387e-06, "loss": 0.8365, "step": 15827 }, { "epoch": 0.8, "learning_rate": 1.9461544703349023e-06, "loss": 0.8494, "step": 15828 }, { "epoch": 0.8, "learning_rate": 1.945179254680718e-06, "loss": 0.9069, "step": 15829 }, { "epoch": 0.8, "learning_rate": 1.944204257102235e-06, "loss": 0.8644, "step": 15830 }, { "epoch": 0.8, "learning_rate": 1.9432294776258497e-06, "loss": 0.9092, "step": 15831 }, { "epoch": 0.8, "learning_rate": 1.9422549162779547e-06, "loss": 0.7769, "step": 15832 }, { "epoch": 0.8, "learning_rate": 1.941280573084935e-06, "loss": 0.8267, "step": 15833 }, { "epoch": 0.8, "learning_rate": 1.940306448073174e-06, "loss": 0.8642, "step": 15834 }, { "epoch": 0.8, "learning_rate": 1.9393325412690367e-06, "loss": 0.8116, "step": 15835 }, { "epoch": 0.8, "learning_rate": 1.938358852698895e-06, "loss": 0.929, "step": 15836 }, { "epoch": 0.8, "learning_rate": 1.9373853823891142e-06, "loss": 0.8893, "step": 15837 }, { "epoch": 0.8, "learning_rate": 1.9364121303660397e-06, "loss": 0.9025, "step": 15838 }, { "epoch": 0.8, "learning_rate": 1.9354390966560343e-06, "loss": 0.7951, "step": 15839 }, { "epoch": 0.8, "learning_rate": 1.934466281285432e-06, "loss": 0.9281, "step": 15840 }, { "epoch": 0.8, "learning_rate": 1.933493684280574e-06, "loss": 0.9403, "step": 15841 }, { "epoch": 0.8, "learning_rate": 1.932521305667796e-06, "loss": 0.907, "step": 15842 }, { "epoch": 0.8, "learning_rate": 1.931549145473417e-06, "loss": 0.9593, "step": 15843 }, { "epoch": 0.8, "learning_rate": 1.9305772037237615e-06, "loss": 0.874, "step": 15844 }, { "epoch": 0.8, "learning_rate": 1.9296054804451426e-06, "loss": 0.8443, "step": 15845 }, { "epoch": 0.81, "learning_rate": 1.9286339756638707e-06, "loss": 0.8625, "step": 15846 }, { "epoch": 0.81, "learning_rate": 1.9276626894062457e-06, "loss": 0.8571, "step": 15847 }, { "epoch": 0.81, "learning_rate": 1.9266916216985643e-06, "loss": 0.861, "step": 15848 }, { "epoch": 0.81, "learning_rate": 1.92572077256712e-06, "loss": 0.817, "step": 15849 }, { "epoch": 0.81, "learning_rate": 1.924750142038191e-06, "loss": 0.8994, "step": 15850 }, { "epoch": 0.81, "learning_rate": 1.9237797301380656e-06, "loss": 0.8009, "step": 15851 }, { "epoch": 0.81, "learning_rate": 1.922809536893009e-06, "loss": 0.9545, "step": 15852 }, { "epoch": 0.81, "learning_rate": 1.9218395623292905e-06, "loss": 0.8696, "step": 15853 }, { "epoch": 0.81, "learning_rate": 1.9208698064731734e-06, "loss": 0.9164, "step": 15854 }, { "epoch": 0.81, "learning_rate": 1.9199002693509093e-06, "loss": 0.869, "step": 15855 }, { "epoch": 0.81, "learning_rate": 1.9189309509887477e-06, "loss": 0.7618, "step": 15856 }, { "epoch": 0.81, "learning_rate": 1.917961851412934e-06, "loss": 0.8733, "step": 15857 }, { "epoch": 0.81, "learning_rate": 1.916992970649707e-06, "loss": 0.8639, "step": 15858 }, { "epoch": 0.81, "learning_rate": 1.9160243087252928e-06, "loss": 0.8483, "step": 15859 }, { "epoch": 0.81, "learning_rate": 1.9150558656659203e-06, "loss": 0.8868, "step": 15860 }, { "epoch": 0.81, "learning_rate": 1.9140876414978106e-06, "loss": 0.8428, "step": 15861 }, { "epoch": 0.81, "learning_rate": 1.9131196362471706e-06, "loss": 0.822, "step": 15862 }, { "epoch": 0.81, "learning_rate": 1.912151849940217e-06, "loss": 0.8885, "step": 15863 }, { "epoch": 0.81, "learning_rate": 1.911184282603146e-06, "loss": 0.8332, "step": 15864 }, { "epoch": 0.81, "learning_rate": 1.9102169342621535e-06, "loss": 0.9369, "step": 15865 }, { "epoch": 0.81, "learning_rate": 1.909249804943434e-06, "loss": 0.8497, "step": 15866 }, { "epoch": 0.81, "learning_rate": 1.9082828946731646e-06, "loss": 0.7677, "step": 15867 }, { "epoch": 0.81, "learning_rate": 1.9073162034775273e-06, "loss": 0.9284, "step": 15868 }, { "epoch": 0.81, "learning_rate": 1.9063497313826939e-06, "loss": 0.921, "step": 15869 }, { "epoch": 0.81, "learning_rate": 1.905383478414834e-06, "loss": 0.9494, "step": 15870 }, { "epoch": 0.81, "learning_rate": 1.9044174446001007e-06, "loss": 0.9439, "step": 15871 }, { "epoch": 0.81, "learning_rate": 1.9034516299646522e-06, "loss": 0.8729, "step": 15872 }, { "epoch": 0.81, "learning_rate": 1.9024860345346397e-06, "loss": 0.917, "step": 15873 }, { "epoch": 0.81, "learning_rate": 1.9015206583361979e-06, "loss": 0.9458, "step": 15874 }, { "epoch": 0.81, "learning_rate": 1.9005555013954734e-06, "loss": 0.8605, "step": 15875 }, { "epoch": 0.81, "learning_rate": 1.8995905637385881e-06, "loss": 0.8983, "step": 15876 }, { "epoch": 0.81, "learning_rate": 1.8986258453916705e-06, "loss": 0.8844, "step": 15877 }, { "epoch": 0.81, "learning_rate": 1.8976613463808423e-06, "loss": 0.8741, "step": 15878 }, { "epoch": 0.81, "learning_rate": 1.8966970667322093e-06, "loss": 0.9557, "step": 15879 }, { "epoch": 0.81, "learning_rate": 1.8957330064718827e-06, "loss": 0.9133, "step": 15880 }, { "epoch": 0.81, "learning_rate": 1.8947691656259625e-06, "loss": 0.8307, "step": 15881 }, { "epoch": 0.81, "learning_rate": 1.8938055442205462e-06, "loss": 0.8459, "step": 15882 }, { "epoch": 0.81, "learning_rate": 1.8928421422817177e-06, "loss": 0.8318, "step": 15883 }, { "epoch": 0.81, "learning_rate": 1.8918789598355635e-06, "loss": 0.9039, "step": 15884 }, { "epoch": 0.81, "learning_rate": 1.8909159969081624e-06, "loss": 0.8995, "step": 15885 }, { "epoch": 0.81, "learning_rate": 1.8899532535255772e-06, "loss": 0.905, "step": 15886 }, { "epoch": 0.81, "learning_rate": 1.8889907297138855e-06, "loss": 0.907, "step": 15887 }, { "epoch": 0.81, "learning_rate": 1.8880284254991377e-06, "loss": 0.9083, "step": 15888 }, { "epoch": 0.81, "learning_rate": 1.8870663409073897e-06, "loss": 0.952, "step": 15889 }, { "epoch": 0.81, "learning_rate": 1.8861044759646918e-06, "loss": 0.8372, "step": 15890 }, { "epoch": 0.81, "learning_rate": 1.8851428306970798e-06, "loss": 0.8438, "step": 15891 }, { "epoch": 0.81, "learning_rate": 1.8841814051305918e-06, "loss": 0.8546, "step": 15892 }, { "epoch": 0.81, "learning_rate": 1.8832201992912592e-06, "loss": 0.8864, "step": 15893 }, { "epoch": 0.81, "learning_rate": 1.8822592132051053e-06, "loss": 0.8689, "step": 15894 }, { "epoch": 0.81, "learning_rate": 1.8812984468981444e-06, "loss": 0.8319, "step": 15895 }, { "epoch": 0.81, "learning_rate": 1.8803379003963907e-06, "loss": 0.9426, "step": 15896 }, { "epoch": 0.81, "learning_rate": 1.8793775737258514e-06, "loss": 0.894, "step": 15897 }, { "epoch": 0.81, "learning_rate": 1.87841746691252e-06, "loss": 0.9521, "step": 15898 }, { "epoch": 0.81, "learning_rate": 1.8774575799824002e-06, "loss": 0.8881, "step": 15899 }, { "epoch": 0.81, "learning_rate": 1.8764979129614724e-06, "loss": 0.7437, "step": 15900 }, { "epoch": 0.81, "learning_rate": 1.8755384658757202e-06, "loss": 0.8023, "step": 15901 }, { "epoch": 0.81, "learning_rate": 1.8745792387511241e-06, "loss": 0.8766, "step": 15902 }, { "epoch": 0.81, "learning_rate": 1.8736202316136442e-06, "loss": 0.9022, "step": 15903 }, { "epoch": 0.81, "learning_rate": 1.8726614444892566e-06, "loss": 0.8273, "step": 15904 }, { "epoch": 0.81, "learning_rate": 1.8717028774039104e-06, "loss": 0.8694, "step": 15905 }, { "epoch": 0.81, "learning_rate": 1.8707445303835603e-06, "loss": 0.9319, "step": 15906 }, { "epoch": 0.81, "learning_rate": 1.8697864034541568e-06, "loss": 0.8237, "step": 15907 }, { "epoch": 0.81, "learning_rate": 1.8688284966416336e-06, "loss": 0.9121, "step": 15908 }, { "epoch": 0.81, "learning_rate": 1.8678708099719288e-06, "loss": 0.8691, "step": 15909 }, { "epoch": 0.81, "learning_rate": 1.8669133434709686e-06, "loss": 0.9169, "step": 15910 }, { "epoch": 0.81, "learning_rate": 1.8659560971646807e-06, "loss": 0.7938, "step": 15911 }, { "epoch": 0.81, "learning_rate": 1.8649990710789734e-06, "loss": 0.8128, "step": 15912 }, { "epoch": 0.81, "learning_rate": 1.8640422652397628e-06, "loss": 0.9254, "step": 15913 }, { "epoch": 0.81, "learning_rate": 1.8630856796729535e-06, "loss": 0.8786, "step": 15914 }, { "epoch": 0.81, "learning_rate": 1.8621293144044372e-06, "loss": 0.7557, "step": 15915 }, { "epoch": 0.81, "learning_rate": 1.8611731694601177e-06, "loss": 0.7768, "step": 15916 }, { "epoch": 0.81, "learning_rate": 1.8602172448658717e-06, "loss": 0.809, "step": 15917 }, { "epoch": 0.81, "learning_rate": 1.8592615406475844e-06, "loss": 0.8284, "step": 15918 }, { "epoch": 0.81, "learning_rate": 1.8583060568311329e-06, "loss": 0.8947, "step": 15919 }, { "epoch": 0.81, "learning_rate": 1.8573507934423796e-06, "loss": 0.8374, "step": 15920 }, { "epoch": 0.81, "learning_rate": 1.8563957505071906e-06, "loss": 0.8546, "step": 15921 }, { "epoch": 0.81, "learning_rate": 1.855440928051423e-06, "loss": 0.9449, "step": 15922 }, { "epoch": 0.81, "learning_rate": 1.8544863261009294e-06, "loss": 0.7867, "step": 15923 }, { "epoch": 0.81, "learning_rate": 1.8535319446815502e-06, "loss": 0.9185, "step": 15924 }, { "epoch": 0.81, "learning_rate": 1.8525777838191262e-06, "loss": 0.8194, "step": 15925 }, { "epoch": 0.81, "learning_rate": 1.8516238435394928e-06, "loss": 0.9432, "step": 15926 }, { "epoch": 0.81, "learning_rate": 1.85067012386847e-06, "loss": 0.8625, "step": 15927 }, { "epoch": 0.81, "learning_rate": 1.8497166248318876e-06, "loss": 0.8808, "step": 15928 }, { "epoch": 0.81, "learning_rate": 1.8487633464555542e-06, "loss": 0.8643, "step": 15929 }, { "epoch": 0.81, "learning_rate": 1.8478102887652815e-06, "loss": 0.7979, "step": 15930 }, { "epoch": 0.81, "learning_rate": 1.8468574517868743e-06, "loss": 0.9289, "step": 15931 }, { "epoch": 0.81, "learning_rate": 1.845904835546124e-06, "loss": 0.8541, "step": 15932 }, { "epoch": 0.81, "learning_rate": 1.8449524400688256e-06, "loss": 0.9283, "step": 15933 }, { "epoch": 0.81, "learning_rate": 1.8440002653807642e-06, "loss": 0.8632, "step": 15934 }, { "epoch": 0.81, "learning_rate": 1.8430483115077202e-06, "loss": 0.7893, "step": 15935 }, { "epoch": 0.81, "learning_rate": 1.8420965784754629e-06, "loss": 0.9139, "step": 15936 }, { "epoch": 0.81, "learning_rate": 1.8411450663097608e-06, "loss": 0.9164, "step": 15937 }, { "epoch": 0.81, "learning_rate": 1.8401937750363785e-06, "loss": 0.8917, "step": 15938 }, { "epoch": 0.81, "learning_rate": 1.8392427046810624e-06, "loss": 0.8934, "step": 15939 }, { "epoch": 0.81, "learning_rate": 1.8382918552695738e-06, "loss": 0.8569, "step": 15940 }, { "epoch": 0.81, "learning_rate": 1.837341226827647e-06, "loss": 0.9362, "step": 15941 }, { "epoch": 0.81, "learning_rate": 1.836390819381022e-06, "loss": 0.9333, "step": 15942 }, { "epoch": 0.81, "learning_rate": 1.835440632955433e-06, "loss": 0.8017, "step": 15943 }, { "epoch": 0.81, "learning_rate": 1.8344906675766006e-06, "loss": 0.839, "step": 15944 }, { "epoch": 0.81, "learning_rate": 1.833540923270245e-06, "loss": 0.8012, "step": 15945 }, { "epoch": 0.81, "learning_rate": 1.8325914000620803e-06, "loss": 0.8924, "step": 15946 }, { "epoch": 0.81, "learning_rate": 1.831642097977817e-06, "loss": 0.8855, "step": 15947 }, { "epoch": 0.81, "learning_rate": 1.8306930170431503e-06, "loss": 0.7824, "step": 15948 }, { "epoch": 0.81, "learning_rate": 1.8297441572837793e-06, "loss": 0.8137, "step": 15949 }, { "epoch": 0.81, "learning_rate": 1.8287955187253959e-06, "loss": 0.8351, "step": 15950 }, { "epoch": 0.81, "learning_rate": 1.8278471013936738e-06, "loss": 0.9423, "step": 15951 }, { "epoch": 0.81, "learning_rate": 1.8268989053143028e-06, "loss": 0.7929, "step": 15952 }, { "epoch": 0.81, "learning_rate": 1.8259509305129463e-06, "loss": 0.8797, "step": 15953 }, { "epoch": 0.81, "learning_rate": 1.825003177015271e-06, "loss": 0.98, "step": 15954 }, { "epoch": 0.81, "learning_rate": 1.8240556448469405e-06, "loss": 0.9177, "step": 15955 }, { "epoch": 0.81, "learning_rate": 1.823108334033602e-06, "loss": 0.9125, "step": 15956 }, { "epoch": 0.81, "learning_rate": 1.8221612446009063e-06, "loss": 0.8535, "step": 15957 }, { "epoch": 0.81, "learning_rate": 1.8212143765744938e-06, "loss": 0.8316, "step": 15958 }, { "epoch": 0.81, "learning_rate": 1.820267729980003e-06, "loss": 0.8761, "step": 15959 }, { "epoch": 0.81, "learning_rate": 1.8193213048430581e-06, "loss": 0.871, "step": 15960 }, { "epoch": 0.81, "learning_rate": 1.8183751011892858e-06, "loss": 0.8918, "step": 15961 }, { "epoch": 0.81, "learning_rate": 1.817429119044306e-06, "loss": 0.9113, "step": 15962 }, { "epoch": 0.81, "learning_rate": 1.8164833584337216e-06, "loss": 0.9101, "step": 15963 }, { "epoch": 0.81, "learning_rate": 1.8155378193831485e-06, "loss": 0.888, "step": 15964 }, { "epoch": 0.81, "learning_rate": 1.8145925019181788e-06, "loss": 0.9288, "step": 15965 }, { "epoch": 0.81, "learning_rate": 1.8136474060644093e-06, "loss": 0.9342, "step": 15966 }, { "epoch": 0.81, "learning_rate": 1.812702531847429e-06, "loss": 0.918, "step": 15967 }, { "epoch": 0.81, "learning_rate": 1.8117578792928137e-06, "loss": 0.815, "step": 15968 }, { "epoch": 0.81, "learning_rate": 1.810813448426143e-06, "loss": 0.8723, "step": 15969 }, { "epoch": 0.81, "learning_rate": 1.8098692392729855e-06, "loss": 0.7709, "step": 15970 }, { "epoch": 0.81, "learning_rate": 1.8089252518589051e-06, "loss": 0.9896, "step": 15971 }, { "epoch": 0.81, "learning_rate": 1.8079814862094603e-06, "loss": 0.8818, "step": 15972 }, { "epoch": 0.81, "learning_rate": 1.8070379423501994e-06, "loss": 0.8214, "step": 15973 }, { "epoch": 0.81, "learning_rate": 1.8060946203066699e-06, "loss": 0.9008, "step": 15974 }, { "epoch": 0.81, "learning_rate": 1.8051515201044101e-06, "loss": 0.8051, "step": 15975 }, { "epoch": 0.81, "learning_rate": 1.8042086417689574e-06, "loss": 0.8114, "step": 15976 }, { "epoch": 0.81, "learning_rate": 1.8032659853258327e-06, "loss": 0.8129, "step": 15977 }, { "epoch": 0.81, "learning_rate": 1.8023235508005621e-06, "loss": 0.8514, "step": 15978 }, { "epoch": 0.81, "learning_rate": 1.8013813382186619e-06, "loss": 0.9129, "step": 15979 }, { "epoch": 0.81, "learning_rate": 1.8004393476056337e-06, "loss": 0.9554, "step": 15980 }, { "epoch": 0.81, "learning_rate": 1.7994975789869918e-06, "loss": 0.859, "step": 15981 }, { "epoch": 0.81, "learning_rate": 1.7985560323882257e-06, "loss": 0.8185, "step": 15982 }, { "epoch": 0.81, "learning_rate": 1.7976147078348294e-06, "loss": 0.8392, "step": 15983 }, { "epoch": 0.81, "learning_rate": 1.7966736053522905e-06, "loss": 0.8969, "step": 15984 }, { "epoch": 0.81, "learning_rate": 1.7957327249660838e-06, "loss": 0.8715, "step": 15985 }, { "epoch": 0.81, "learning_rate": 1.794792066701685e-06, "loss": 0.895, "step": 15986 }, { "epoch": 0.81, "learning_rate": 1.7938516305845622e-06, "loss": 0.9647, "step": 15987 }, { "epoch": 0.81, "learning_rate": 1.7929114166401773e-06, "loss": 0.8543, "step": 15988 }, { "epoch": 0.81, "learning_rate": 1.7919714248939813e-06, "loss": 0.8255, "step": 15989 }, { "epoch": 0.81, "learning_rate": 1.791031655371427e-06, "loss": 0.9185, "step": 15990 }, { "epoch": 0.81, "learning_rate": 1.7900921080979593e-06, "loss": 0.812, "step": 15991 }, { "epoch": 0.81, "learning_rate": 1.7891527830990085e-06, "loss": 0.9176, "step": 15992 }, { "epoch": 0.81, "learning_rate": 1.788213680400015e-06, "loss": 0.8305, "step": 15993 }, { "epoch": 0.81, "learning_rate": 1.7872748000263972e-06, "loss": 0.8128, "step": 15994 }, { "epoch": 0.81, "learning_rate": 1.7863361420035764e-06, "loss": 0.8796, "step": 15995 }, { "epoch": 0.81, "learning_rate": 1.7853977063569683e-06, "loss": 0.9653, "step": 15996 }, { "epoch": 0.81, "learning_rate": 1.7844594931119752e-06, "loss": 0.9459, "step": 15997 }, { "epoch": 0.81, "learning_rate": 1.7835215022940011e-06, "loss": 0.9567, "step": 15998 }, { "epoch": 0.81, "learning_rate": 1.7825837339284403e-06, "loss": 0.9356, "step": 15999 }, { "epoch": 0.81, "learning_rate": 1.7816461880406843e-06, "loss": 0.8213, "step": 16000 }, { "epoch": 0.81, "learning_rate": 1.780708864656112e-06, "loss": 0.8694, "step": 16001 }, { "epoch": 0.81, "learning_rate": 1.7797717638001023e-06, "loss": 0.8391, "step": 16002 }, { "epoch": 0.81, "learning_rate": 1.7788348854980298e-06, "loss": 0.8875, "step": 16003 }, { "epoch": 0.81, "learning_rate": 1.77789822977525e-06, "loss": 0.9355, "step": 16004 }, { "epoch": 0.81, "learning_rate": 1.7769617966571339e-06, "loss": 0.8497, "step": 16005 }, { "epoch": 0.81, "learning_rate": 1.7760255861690257e-06, "loss": 0.8629, "step": 16006 }, { "epoch": 0.81, "learning_rate": 1.775089598336276e-06, "loss": 0.8607, "step": 16007 }, { "epoch": 0.81, "learning_rate": 1.7741538331842257e-06, "loss": 0.773, "step": 16008 }, { "epoch": 0.81, "learning_rate": 1.773218290738208e-06, "loss": 0.7802, "step": 16009 }, { "epoch": 0.81, "learning_rate": 1.7722829710235512e-06, "loss": 0.883, "step": 16010 }, { "epoch": 0.81, "learning_rate": 1.77134787406558e-06, "loss": 0.8706, "step": 16011 }, { "epoch": 0.81, "learning_rate": 1.7704129998896125e-06, "loss": 0.843, "step": 16012 }, { "epoch": 0.81, "learning_rate": 1.7694783485209542e-06, "loss": 0.865, "step": 16013 }, { "epoch": 0.81, "learning_rate": 1.768543919984914e-06, "loss": 0.8902, "step": 16014 }, { "epoch": 0.81, "learning_rate": 1.7676097143067927e-06, "loss": 0.8255, "step": 16015 }, { "epoch": 0.81, "learning_rate": 1.7666757315118744e-06, "loss": 0.8406, "step": 16016 }, { "epoch": 0.81, "learning_rate": 1.7657419716254553e-06, "loss": 0.8756, "step": 16017 }, { "epoch": 0.81, "learning_rate": 1.76480843467281e-06, "loss": 0.8891, "step": 16018 }, { "epoch": 0.81, "learning_rate": 1.7638751206792149e-06, "loss": 0.8466, "step": 16019 }, { "epoch": 0.81, "learning_rate": 1.7629420296699406e-06, "loss": 0.8434, "step": 16020 }, { "epoch": 0.81, "learning_rate": 1.762009161670245e-06, "loss": 0.8507, "step": 16021 }, { "epoch": 0.81, "learning_rate": 1.7610765167053856e-06, "loss": 0.9603, "step": 16022 }, { "epoch": 0.81, "learning_rate": 1.7601440948006155e-06, "loss": 0.9026, "step": 16023 }, { "epoch": 0.81, "learning_rate": 1.75921189598118e-06, "loss": 0.7837, "step": 16024 }, { "epoch": 0.81, "learning_rate": 1.7582799202723111e-06, "loss": 0.8394, "step": 16025 }, { "epoch": 0.81, "learning_rate": 1.7573481676992454e-06, "loss": 0.8094, "step": 16026 }, { "epoch": 0.81, "learning_rate": 1.7564166382872118e-06, "loss": 0.905, "step": 16027 }, { "epoch": 0.81, "learning_rate": 1.7554853320614219e-06, "loss": 0.8447, "step": 16028 }, { "epoch": 0.81, "learning_rate": 1.7545542490471002e-06, "loss": 0.8898, "step": 16029 }, { "epoch": 0.81, "learning_rate": 1.7536233892694465e-06, "loss": 0.8824, "step": 16030 }, { "epoch": 0.81, "learning_rate": 1.7526927527536674e-06, "loss": 0.8589, "step": 16031 }, { "epoch": 0.81, "learning_rate": 1.7517623395249594e-06, "loss": 0.8483, "step": 16032 }, { "epoch": 0.81, "learning_rate": 1.750832149608508e-06, "loss": 0.7991, "step": 16033 }, { "epoch": 0.81, "learning_rate": 1.7499021830294993e-06, "loss": 0.7544, "step": 16034 }, { "epoch": 0.81, "learning_rate": 1.7489724398131124e-06, "loss": 0.9968, "step": 16035 }, { "epoch": 0.81, "learning_rate": 1.7480429199845206e-06, "loss": 0.8543, "step": 16036 }, { "epoch": 0.81, "learning_rate": 1.7471136235688846e-06, "loss": 0.9225, "step": 16037 }, { "epoch": 0.81, "learning_rate": 1.7461845505913666e-06, "loss": 0.8424, "step": 16038 }, { "epoch": 0.81, "learning_rate": 1.745255701077121e-06, "loss": 0.9026, "step": 16039 }, { "epoch": 0.81, "learning_rate": 1.7443270750512942e-06, "loss": 0.8419, "step": 16040 }, { "epoch": 0.81, "learning_rate": 1.7433986725390318e-06, "loss": 0.7819, "step": 16041 }, { "epoch": 0.81, "learning_rate": 1.7424704935654636e-06, "loss": 0.8072, "step": 16042 }, { "epoch": 0.82, "learning_rate": 1.7415425381557216e-06, "loss": 0.9135, "step": 16043 }, { "epoch": 0.82, "learning_rate": 1.7406148063349304e-06, "loss": 0.8512, "step": 16044 }, { "epoch": 0.82, "learning_rate": 1.739687298128202e-06, "loss": 0.796, "step": 16045 }, { "epoch": 0.82, "learning_rate": 1.7387600135606563e-06, "loss": 0.8629, "step": 16046 }, { "epoch": 0.82, "learning_rate": 1.7378329526573922e-06, "loss": 0.779, "step": 16047 }, { "epoch": 0.82, "learning_rate": 1.7369061154435095e-06, "loss": 0.8317, "step": 16048 }, { "epoch": 0.82, "learning_rate": 1.7359795019441062e-06, "loss": 0.8784, "step": 16049 }, { "epoch": 0.82, "learning_rate": 1.7350531121842628e-06, "loss": 0.8781, "step": 16050 }, { "epoch": 0.82, "learning_rate": 1.734126946189062e-06, "loss": 0.855, "step": 16051 }, { "epoch": 0.82, "learning_rate": 1.7332010039835811e-06, "loss": 0.8713, "step": 16052 }, { "epoch": 0.82, "learning_rate": 1.7322752855928904e-06, "loss": 0.8862, "step": 16053 }, { "epoch": 0.82, "learning_rate": 1.7313497910420473e-06, "loss": 0.9034, "step": 16054 }, { "epoch": 0.82, "learning_rate": 1.7304245203561121e-06, "loss": 0.8731, "step": 16055 }, { "epoch": 0.82, "learning_rate": 1.7294994735601368e-06, "loss": 0.8055, "step": 16056 }, { "epoch": 0.82, "learning_rate": 1.728574650679159e-06, "loss": 0.9359, "step": 16057 }, { "epoch": 0.82, "learning_rate": 1.7276500517382277e-06, "loss": 0.871, "step": 16058 }, { "epoch": 0.82, "learning_rate": 1.7267256767623674e-06, "loss": 0.9318, "step": 16059 }, { "epoch": 0.82, "learning_rate": 1.725801525776607e-06, "loss": 0.8456, "step": 16060 }, { "epoch": 0.82, "learning_rate": 1.7248775988059707e-06, "loss": 0.7958, "step": 16061 }, { "epoch": 0.82, "learning_rate": 1.7239538958754664e-06, "loss": 0.851, "step": 16062 }, { "epoch": 0.82, "learning_rate": 1.7230304170101053e-06, "loss": 0.7796, "step": 16063 }, { "epoch": 0.82, "learning_rate": 1.7221071622348896e-06, "loss": 0.8116, "step": 16064 }, { "epoch": 0.82, "learning_rate": 1.7211841315748178e-06, "loss": 0.8576, "step": 16065 }, { "epoch": 0.82, "learning_rate": 1.7202613250548762e-06, "loss": 0.7961, "step": 16066 }, { "epoch": 0.82, "learning_rate": 1.7193387427000485e-06, "loss": 0.9019, "step": 16067 }, { "epoch": 0.82, "learning_rate": 1.7184163845353185e-06, "loss": 0.8326, "step": 16068 }, { "epoch": 0.82, "learning_rate": 1.7174942505856485e-06, "loss": 0.8223, "step": 16069 }, { "epoch": 0.82, "learning_rate": 1.7165723408760149e-06, "loss": 0.8675, "step": 16070 }, { "epoch": 0.82, "learning_rate": 1.715650655431369e-06, "loss": 0.9259, "step": 16071 }, { "epoch": 0.82, "learning_rate": 1.7147291942766686e-06, "loss": 0.8109, "step": 16072 }, { "epoch": 0.82, "learning_rate": 1.7138079574368627e-06, "loss": 0.8859, "step": 16073 }, { "epoch": 0.82, "learning_rate": 1.712886944936888e-06, "loss": 0.7751, "step": 16074 }, { "epoch": 0.82, "learning_rate": 1.711966156801682e-06, "loss": 0.7818, "step": 16075 }, { "epoch": 0.82, "learning_rate": 1.7110455930561753e-06, "loss": 0.8427, "step": 16076 }, { "epoch": 0.82, "learning_rate": 1.7101252537252933e-06, "loss": 0.8287, "step": 16077 }, { "epoch": 0.82, "learning_rate": 1.709205138833947e-06, "loss": 0.8148, "step": 16078 }, { "epoch": 0.82, "learning_rate": 1.708285248407051e-06, "loss": 0.7944, "step": 16079 }, { "epoch": 0.82, "learning_rate": 1.7073655824695123e-06, "loss": 0.8734, "step": 16080 }, { "epoch": 0.82, "learning_rate": 1.7064461410462229e-06, "loss": 0.9462, "step": 16081 }, { "epoch": 0.82, "learning_rate": 1.7055269241620864e-06, "loss": 0.7692, "step": 16082 }, { "epoch": 0.82, "learning_rate": 1.7046079318419806e-06, "loss": 0.8611, "step": 16083 }, { "epoch": 0.82, "learning_rate": 1.70368916411079e-06, "loss": 0.7771, "step": 16084 }, { "epoch": 0.82, "learning_rate": 1.7027706209933903e-06, "loss": 0.8668, "step": 16085 }, { "epoch": 0.82, "learning_rate": 1.7018523025146472e-06, "loss": 0.901, "step": 16086 }, { "epoch": 0.82, "learning_rate": 1.700934208699423e-06, "loss": 0.9715, "step": 16087 }, { "epoch": 0.82, "learning_rate": 1.7000163395725766e-06, "loss": 0.7844, "step": 16088 }, { "epoch": 0.82, "learning_rate": 1.6990986951589595e-06, "loss": 0.8182, "step": 16089 }, { "epoch": 0.82, "learning_rate": 1.6981812754834115e-06, "loss": 0.8989, "step": 16090 }, { "epoch": 0.82, "learning_rate": 1.697264080570773e-06, "loss": 0.8551, "step": 16091 }, { "epoch": 0.82, "learning_rate": 1.6963471104458783e-06, "loss": 0.8234, "step": 16092 }, { "epoch": 0.82, "learning_rate": 1.6954303651335468e-06, "loss": 0.8318, "step": 16093 }, { "epoch": 0.82, "learning_rate": 1.6945138446586074e-06, "loss": 0.8414, "step": 16094 }, { "epoch": 0.82, "learning_rate": 1.6935975490458667e-06, "loss": 0.7867, "step": 16095 }, { "epoch": 0.82, "learning_rate": 1.692681478320135e-06, "loss": 0.7837, "step": 16096 }, { "epoch": 0.82, "learning_rate": 1.691765632506217e-06, "loss": 0.847, "step": 16097 }, { "epoch": 0.82, "learning_rate": 1.6908500116289017e-06, "loss": 0.8255, "step": 16098 }, { "epoch": 0.82, "learning_rate": 1.6899346157129826e-06, "loss": 0.7983, "step": 16099 }, { "epoch": 0.82, "learning_rate": 1.6890194447832432e-06, "loss": 0.8731, "step": 16100 }, { "epoch": 0.82, "learning_rate": 1.6881044988644623e-06, "loss": 0.8716, "step": 16101 }, { "epoch": 0.82, "learning_rate": 1.6871897779814062e-06, "loss": 0.8421, "step": 16102 }, { "epoch": 0.82, "learning_rate": 1.6862752821588424e-06, "loss": 0.8757, "step": 16103 }, { "epoch": 0.82, "learning_rate": 1.6853610114215335e-06, "loss": 0.8396, "step": 16104 }, { "epoch": 0.82, "learning_rate": 1.6844469657942242e-06, "loss": 0.8652, "step": 16105 }, { "epoch": 0.82, "learning_rate": 1.68353314530167e-06, "loss": 0.8116, "step": 16106 }, { "epoch": 0.82, "learning_rate": 1.682619549968606e-06, "loss": 0.8229, "step": 16107 }, { "epoch": 0.82, "learning_rate": 1.6817061798197697e-06, "loss": 0.9661, "step": 16108 }, { "epoch": 0.82, "learning_rate": 1.6807930348798896e-06, "loss": 0.8476, "step": 16109 }, { "epoch": 0.82, "learning_rate": 1.6798801151736832e-06, "loss": 0.9572, "step": 16110 }, { "epoch": 0.82, "learning_rate": 1.6789674207258754e-06, "loss": 0.8811, "step": 16111 }, { "epoch": 0.82, "learning_rate": 1.6780549515611689e-06, "loss": 0.8799, "step": 16112 }, { "epoch": 0.82, "learning_rate": 1.6771427077042713e-06, "loss": 0.8939, "step": 16113 }, { "epoch": 0.82, "learning_rate": 1.6762306891798829e-06, "loss": 0.873, "step": 16114 }, { "epoch": 0.82, "learning_rate": 1.6753188960126898e-06, "loss": 0.8132, "step": 16115 }, { "epoch": 0.82, "learning_rate": 1.674407328227381e-06, "loss": 0.9638, "step": 16116 }, { "epoch": 0.82, "learning_rate": 1.6734959858486354e-06, "loss": 0.8981, "step": 16117 }, { "epoch": 0.82, "learning_rate": 1.6725848689011304e-06, "loss": 0.8496, "step": 16118 }, { "epoch": 0.82, "learning_rate": 1.6716739774095281e-06, "loss": 0.9121, "step": 16119 }, { "epoch": 0.82, "learning_rate": 1.670763311398491e-06, "loss": 0.8893, "step": 16120 }, { "epoch": 0.82, "learning_rate": 1.6698528708926798e-06, "loss": 0.8716, "step": 16121 }, { "epoch": 0.82, "learning_rate": 1.6689426559167332e-06, "loss": 0.8772, "step": 16122 }, { "epoch": 0.82, "learning_rate": 1.6680326664953062e-06, "loss": 0.7579, "step": 16123 }, { "epoch": 0.82, "learning_rate": 1.6671229026530278e-06, "loss": 0.7804, "step": 16124 }, { "epoch": 0.82, "learning_rate": 1.6662133644145318e-06, "loss": 0.8664, "step": 16125 }, { "epoch": 0.82, "learning_rate": 1.6653040518044439e-06, "loss": 0.8484, "step": 16126 }, { "epoch": 0.82, "learning_rate": 1.66439496484738e-06, "loss": 0.9809, "step": 16127 }, { "epoch": 0.82, "learning_rate": 1.663486103567954e-06, "loss": 0.8713, "step": 16128 }, { "epoch": 0.82, "learning_rate": 1.6625774679907713e-06, "loss": 0.8391, "step": 16129 }, { "epoch": 0.82, "learning_rate": 1.661669058140436e-06, "loss": 0.8816, "step": 16130 }, { "epoch": 0.82, "learning_rate": 1.6607608740415383e-06, "loss": 1.0226, "step": 16131 }, { "epoch": 0.82, "learning_rate": 1.6598529157186672e-06, "loss": 0.8386, "step": 16132 }, { "epoch": 0.82, "learning_rate": 1.6589451831964064e-06, "loss": 0.84, "step": 16133 }, { "epoch": 0.82, "learning_rate": 1.6580376764993278e-06, "loss": 0.8693, "step": 16134 }, { "epoch": 0.82, "learning_rate": 1.6571303956520068e-06, "loss": 0.8806, "step": 16135 }, { "epoch": 0.82, "learning_rate": 1.6562233406790029e-06, "loss": 0.8449, "step": 16136 }, { "epoch": 0.82, "learning_rate": 1.6553165116048753e-06, "loss": 0.8799, "step": 16137 }, { "epoch": 0.82, "learning_rate": 1.6544099084541765e-06, "loss": 0.8779, "step": 16138 }, { "epoch": 0.82, "learning_rate": 1.653503531251449e-06, "loss": 0.8456, "step": 16139 }, { "epoch": 0.82, "learning_rate": 1.6525973800212325e-06, "loss": 0.8976, "step": 16140 }, { "epoch": 0.82, "learning_rate": 1.6516914547880614e-06, "loss": 0.8922, "step": 16141 }, { "epoch": 0.82, "learning_rate": 1.650785755576465e-06, "loss": 0.8662, "step": 16142 }, { "epoch": 0.82, "learning_rate": 1.6498802824109593e-06, "loss": 0.7209, "step": 16143 }, { "epoch": 0.82, "learning_rate": 1.6489750353160606e-06, "loss": 0.8768, "step": 16144 }, { "epoch": 0.82, "learning_rate": 1.6480700143162798e-06, "loss": 0.8442, "step": 16145 }, { "epoch": 0.82, "learning_rate": 1.6471652194361131e-06, "loss": 0.8597, "step": 16146 }, { "epoch": 0.82, "learning_rate": 1.6462606507000666e-06, "loss": 0.863, "step": 16147 }, { "epoch": 0.82, "learning_rate": 1.645356308132623e-06, "loss": 0.8867, "step": 16148 }, { "epoch": 0.82, "learning_rate": 1.644452191758269e-06, "loss": 0.8689, "step": 16149 }, { "epoch": 0.82, "learning_rate": 1.6435483016014842e-06, "loss": 0.9306, "step": 16150 }, { "epoch": 0.82, "learning_rate": 1.6426446376867355e-06, "loss": 0.7572, "step": 16151 }, { "epoch": 0.82, "learning_rate": 1.6417412000384924e-06, "loss": 0.9356, "step": 16152 }, { "epoch": 0.82, "learning_rate": 1.6408379886812142e-06, "loss": 0.8163, "step": 16153 }, { "epoch": 0.82, "learning_rate": 1.6399350036393557e-06, "loss": 0.8457, "step": 16154 }, { "epoch": 0.82, "learning_rate": 1.6390322449373607e-06, "loss": 0.8319, "step": 16155 }, { "epoch": 0.82, "learning_rate": 1.638129712599672e-06, "loss": 0.8976, "step": 16156 }, { "epoch": 0.82, "learning_rate": 1.6372274066507276e-06, "loss": 0.834, "step": 16157 }, { "epoch": 0.82, "learning_rate": 1.6363253271149493e-06, "loss": 0.8834, "step": 16158 }, { "epoch": 0.82, "learning_rate": 1.6354234740167686e-06, "loss": 0.9557, "step": 16159 }, { "epoch": 0.82, "learning_rate": 1.6345218473805958e-06, "loss": 0.911, "step": 16160 }, { "epoch": 0.82, "learning_rate": 1.6336204472308447e-06, "loss": 0.9639, "step": 16161 }, { "epoch": 0.82, "learning_rate": 1.6327192735919206e-06, "loss": 0.8962, "step": 16162 }, { "epoch": 0.82, "learning_rate": 1.631818326488217e-06, "loss": 0.8196, "step": 16163 }, { "epoch": 0.82, "learning_rate": 1.63091760594413e-06, "loss": 0.8745, "step": 16164 }, { "epoch": 0.82, "learning_rate": 1.6300171119840447e-06, "loss": 0.8059, "step": 16165 }, { "epoch": 0.82, "learning_rate": 1.6291168446323435e-06, "loss": 0.9043, "step": 16166 }, { "epoch": 0.82, "learning_rate": 1.628216803913396e-06, "loss": 0.8045, "step": 16167 }, { "epoch": 0.82, "learning_rate": 1.6273169898515707e-06, "loss": 0.8492, "step": 16168 }, { "epoch": 0.82, "learning_rate": 1.626417402471233e-06, "loss": 0.8842, "step": 16169 }, { "epoch": 0.82, "learning_rate": 1.6255180417967309e-06, "loss": 0.8648, "step": 16170 }, { "epoch": 0.82, "learning_rate": 1.6246189078524221e-06, "loss": 0.8622, "step": 16171 }, { "epoch": 0.82, "learning_rate": 1.6237200006626442e-06, "loss": 0.7654, "step": 16172 }, { "epoch": 0.82, "learning_rate": 1.622821320251734e-06, "loss": 0.8733, "step": 16173 }, { "epoch": 0.82, "learning_rate": 1.621922866644028e-06, "loss": 0.9401, "step": 16174 }, { "epoch": 0.82, "learning_rate": 1.6210246398638429e-06, "loss": 0.8325, "step": 16175 }, { "epoch": 0.82, "learning_rate": 1.6201266399355031e-06, "loss": 0.9514, "step": 16176 }, { "epoch": 0.82, "learning_rate": 1.619228866883318e-06, "loss": 0.849, "step": 16177 }, { "epoch": 0.82, "learning_rate": 1.6183313207315944e-06, "loss": 0.8693, "step": 16178 }, { "epoch": 0.82, "learning_rate": 1.6174340015046353e-06, "loss": 0.8781, "step": 16179 }, { "epoch": 0.82, "learning_rate": 1.6165369092267302e-06, "loss": 0.807, "step": 16180 }, { "epoch": 0.82, "learning_rate": 1.6156400439221698e-06, "loss": 0.9126, "step": 16181 }, { "epoch": 0.82, "learning_rate": 1.6147434056152333e-06, "loss": 0.7977, "step": 16182 }, { "epoch": 0.82, "learning_rate": 1.6138469943302015e-06, "loss": 0.8522, "step": 16183 }, { "epoch": 0.82, "learning_rate": 1.6129508100913372e-06, "loss": 0.8589, "step": 16184 }, { "epoch": 0.82, "learning_rate": 1.6120548529229075e-06, "loss": 0.9211, "step": 16185 }, { "epoch": 0.82, "learning_rate": 1.6111591228491707e-06, "loss": 0.8183, "step": 16186 }, { "epoch": 0.82, "learning_rate": 1.61026361989437e-06, "loss": 0.81, "step": 16187 }, { "epoch": 0.82, "learning_rate": 1.6093683440827612e-06, "loss": 0.8922, "step": 16188 }, { "epoch": 0.82, "learning_rate": 1.6084732954385752e-06, "loss": 0.8655, "step": 16189 }, { "epoch": 0.82, "learning_rate": 1.6075784739860467e-06, "loss": 0.8709, "step": 16190 }, { "epoch": 0.82, "learning_rate": 1.6066838797494044e-06, "loss": 0.7791, "step": 16191 }, { "epoch": 0.82, "learning_rate": 1.6057895127528645e-06, "loss": 0.9096, "step": 16192 }, { "epoch": 0.82, "learning_rate": 1.604895373020642e-06, "loss": 0.9254, "step": 16193 }, { "epoch": 0.82, "learning_rate": 1.6040014605769461e-06, "loss": 0.9001, "step": 16194 }, { "epoch": 0.82, "learning_rate": 1.60310777544598e-06, "loss": 0.7753, "step": 16195 }, { "epoch": 0.82, "learning_rate": 1.6022143176519344e-06, "loss": 0.9839, "step": 16196 }, { "epoch": 0.82, "learning_rate": 1.6013210872190033e-06, "loss": 0.9342, "step": 16197 }, { "epoch": 0.82, "learning_rate": 1.6004280841713694e-06, "loss": 0.8953, "step": 16198 }, { "epoch": 0.82, "learning_rate": 1.5995353085332043e-06, "loss": 0.8044, "step": 16199 }, { "epoch": 0.82, "learning_rate": 1.598642760328688e-06, "loss": 0.9196, "step": 16200 }, { "epoch": 0.82, "learning_rate": 1.5977504395819776e-06, "loss": 0.8157, "step": 16201 }, { "epoch": 0.82, "learning_rate": 1.596858346317236e-06, "loss": 0.8413, "step": 16202 }, { "epoch": 0.82, "learning_rate": 1.5959664805586161e-06, "loss": 0.8241, "step": 16203 }, { "epoch": 0.82, "learning_rate": 1.5950748423302598e-06, "loss": 0.8725, "step": 16204 }, { "epoch": 0.82, "learning_rate": 1.594183431656311e-06, "loss": 0.9169, "step": 16205 }, { "epoch": 0.82, "learning_rate": 1.5932922485609026e-06, "loss": 0.8222, "step": 16206 }, { "epoch": 0.82, "learning_rate": 1.5924012930681643e-06, "loss": 0.8666, "step": 16207 }, { "epoch": 0.82, "learning_rate": 1.5915105652022133e-06, "loss": 0.8906, "step": 16208 }, { "epoch": 0.82, "learning_rate": 1.5906200649871684e-06, "loss": 0.8411, "step": 16209 }, { "epoch": 0.82, "learning_rate": 1.5897297924471411e-06, "loss": 0.8752, "step": 16210 }, { "epoch": 0.82, "learning_rate": 1.5888397476062268e-06, "loss": 0.8327, "step": 16211 }, { "epoch": 0.82, "learning_rate": 1.5879499304885325e-06, "loss": 0.9212, "step": 16212 }, { "epoch": 0.82, "learning_rate": 1.5870603411181406e-06, "loss": 0.8507, "step": 16213 }, { "epoch": 0.82, "learning_rate": 1.5861709795191404e-06, "loss": 0.9471, "step": 16214 }, { "epoch": 0.82, "learning_rate": 1.5852818457156115e-06, "loss": 0.9644, "step": 16215 }, { "epoch": 0.82, "learning_rate": 1.5843929397316215e-06, "loss": 0.831, "step": 16216 }, { "epoch": 0.82, "learning_rate": 1.5835042615912399e-06, "loss": 0.8441, "step": 16217 }, { "epoch": 0.82, "learning_rate": 1.5826158113185252e-06, "loss": 0.9125, "step": 16218 }, { "epoch": 0.82, "learning_rate": 1.581727588937535e-06, "loss": 0.9081, "step": 16219 }, { "epoch": 0.82, "learning_rate": 1.5808395944723109e-06, "loss": 0.8635, "step": 16220 }, { "epoch": 0.82, "learning_rate": 1.5799518279468973e-06, "loss": 0.9013, "step": 16221 }, { "epoch": 0.82, "learning_rate": 1.5790642893853324e-06, "loss": 0.7306, "step": 16222 }, { "epoch": 0.82, "learning_rate": 1.5781769788116385e-06, "loss": 0.8637, "step": 16223 }, { "epoch": 0.82, "learning_rate": 1.5772898962498473e-06, "loss": 0.8838, "step": 16224 }, { "epoch": 0.82, "learning_rate": 1.5764030417239684e-06, "loss": 0.9852, "step": 16225 }, { "epoch": 0.82, "learning_rate": 1.5755164152580161e-06, "loss": 0.8109, "step": 16226 }, { "epoch": 0.82, "learning_rate": 1.5746300168759953e-06, "loss": 0.8379, "step": 16227 }, { "epoch": 0.82, "learning_rate": 1.5737438466019006e-06, "loss": 0.7842, "step": 16228 }, { "epoch": 0.82, "learning_rate": 1.572857904459727e-06, "loss": 0.8572, "step": 16229 }, { "epoch": 0.82, "learning_rate": 1.5719721904734587e-06, "loss": 0.8239, "step": 16230 }, { "epoch": 0.82, "learning_rate": 1.5710867046670797e-06, "loss": 0.9061, "step": 16231 }, { "epoch": 0.82, "learning_rate": 1.5702014470645588e-06, "loss": 0.8739, "step": 16232 }, { "epoch": 0.82, "learning_rate": 1.5693164176898646e-06, "loss": 0.9787, "step": 16233 }, { "epoch": 0.82, "learning_rate": 1.568431616566961e-06, "loss": 0.8007, "step": 16234 }, { "epoch": 0.82, "learning_rate": 1.567547043719796e-06, "loss": 0.7986, "step": 16235 }, { "epoch": 0.82, "learning_rate": 1.566662699172329e-06, "loss": 0.9097, "step": 16236 }, { "epoch": 0.82, "learning_rate": 1.5657785829484939e-06, "loss": 0.8172, "step": 16237 }, { "epoch": 0.82, "learning_rate": 1.5648946950722299e-06, "loss": 0.9096, "step": 16238 }, { "epoch": 0.82, "learning_rate": 1.5640110355674709e-06, "loss": 0.8292, "step": 16239 }, { "epoch": 0.83, "learning_rate": 1.5631276044581356e-06, "loss": 0.8094, "step": 16240 }, { "epoch": 0.83, "learning_rate": 1.5622444017681438e-06, "loss": 0.9641, "step": 16241 }, { "epoch": 0.83, "learning_rate": 1.5613614275214084e-06, "loss": 0.8489, "step": 16242 }, { "epoch": 0.83, "learning_rate": 1.560478681741836e-06, "loss": 0.974, "step": 16243 }, { "epoch": 0.83, "learning_rate": 1.559596164453323e-06, "loss": 0.8913, "step": 16244 }, { "epoch": 0.83, "learning_rate": 1.5587138756797637e-06, "loss": 0.8491, "step": 16245 }, { "epoch": 0.83, "learning_rate": 1.5578318154450466e-06, "loss": 0.8774, "step": 16246 }, { "epoch": 0.83, "learning_rate": 1.5569499837730474e-06, "loss": 0.9046, "step": 16247 }, { "epoch": 0.83, "learning_rate": 1.5560683806876498e-06, "loss": 0.9121, "step": 16248 }, { "epoch": 0.83, "learning_rate": 1.555187006212714e-06, "loss": 0.8925, "step": 16249 }, { "epoch": 0.83, "learning_rate": 1.5543058603721062e-06, "loss": 0.8799, "step": 16250 }, { "epoch": 0.83, "learning_rate": 1.5534249431896841e-06, "loss": 0.897, "step": 16251 }, { "epoch": 0.83, "learning_rate": 1.55254425468929e-06, "loss": 0.8118, "step": 16252 }, { "epoch": 0.83, "learning_rate": 1.5516637948947777e-06, "loss": 0.9039, "step": 16253 }, { "epoch": 0.83, "learning_rate": 1.5507835638299773e-06, "loss": 0.8424, "step": 16254 }, { "epoch": 0.83, "learning_rate": 1.549903561518723e-06, "loss": 0.9293, "step": 16255 }, { "epoch": 0.83, "learning_rate": 1.5490237879848413e-06, "loss": 0.8638, "step": 16256 }, { "epoch": 0.83, "learning_rate": 1.5481442432521477e-06, "loss": 0.783, "step": 16257 }, { "epoch": 0.83, "learning_rate": 1.5472649273444562e-06, "loss": 0.8086, "step": 16258 }, { "epoch": 0.83, "learning_rate": 1.5463858402855746e-06, "loss": 0.9115, "step": 16259 }, { "epoch": 0.83, "learning_rate": 1.545506982099304e-06, "loss": 0.8098, "step": 16260 }, { "epoch": 0.83, "learning_rate": 1.5446283528094352e-06, "loss": 0.8739, "step": 16261 }, { "epoch": 0.83, "learning_rate": 1.543749952439757e-06, "loss": 0.912, "step": 16262 }, { "epoch": 0.83, "learning_rate": 1.5428717810140547e-06, "loss": 0.8127, "step": 16263 }, { "epoch": 0.83, "learning_rate": 1.5419938385560974e-06, "loss": 0.9173, "step": 16264 }, { "epoch": 0.83, "learning_rate": 1.5411161250896621e-06, "loss": 0.8136, "step": 16265 }, { "epoch": 0.83, "learning_rate": 1.5402386406385062e-06, "loss": 0.8273, "step": 16266 }, { "epoch": 0.83, "learning_rate": 1.5393613852263889e-06, "loss": 0.8121, "step": 16267 }, { "epoch": 0.83, "learning_rate": 1.5384843588770626e-06, "loss": 0.8793, "step": 16268 }, { "epoch": 0.83, "learning_rate": 1.5376075616142682e-06, "loss": 0.9239, "step": 16269 }, { "epoch": 0.83, "learning_rate": 1.5367309934617447e-06, "loss": 0.8331, "step": 16270 }, { "epoch": 0.83, "learning_rate": 1.5358546544432263e-06, "loss": 0.8282, "step": 16271 }, { "epoch": 0.83, "learning_rate": 1.5349785445824406e-06, "loss": 0.8151, "step": 16272 }, { "epoch": 0.83, "learning_rate": 1.534102663903101e-06, "loss": 0.8448, "step": 16273 }, { "epoch": 0.83, "learning_rate": 1.5332270124289262e-06, "loss": 0.9845, "step": 16274 }, { "epoch": 0.83, "learning_rate": 1.532351590183624e-06, "loss": 0.9296, "step": 16275 }, { "epoch": 0.83, "learning_rate": 1.5314763971908885e-06, "loss": 0.8824, "step": 16276 }, { "epoch": 0.83, "learning_rate": 1.5306014334744247e-06, "loss": 0.8748, "step": 16277 }, { "epoch": 0.83, "learning_rate": 1.529726699057914e-06, "loss": 0.9058, "step": 16278 }, { "epoch": 0.83, "learning_rate": 1.5288521939650414e-06, "loss": 0.8648, "step": 16279 }, { "epoch": 0.83, "learning_rate": 1.5279779182194855e-06, "loss": 0.8926, "step": 16280 }, { "epoch": 0.83, "learning_rate": 1.5271038718449117e-06, "loss": 0.9454, "step": 16281 }, { "epoch": 0.83, "learning_rate": 1.5262300548649855e-06, "loss": 0.856, "step": 16282 }, { "epoch": 0.83, "learning_rate": 1.525356467303365e-06, "loss": 0.9628, "step": 16283 }, { "epoch": 0.83, "learning_rate": 1.5244831091837054e-06, "loss": 0.885, "step": 16284 }, { "epoch": 0.83, "learning_rate": 1.5236099805296445e-06, "loss": 0.9608, "step": 16285 }, { "epoch": 0.83, "learning_rate": 1.5227370813648268e-06, "loss": 0.9238, "step": 16286 }, { "epoch": 0.83, "learning_rate": 1.5218644117128845e-06, "loss": 0.7648, "step": 16287 }, { "epoch": 0.83, "learning_rate": 1.5209919715974385e-06, "loss": 0.9696, "step": 16288 }, { "epoch": 0.83, "learning_rate": 1.5201197610421203e-06, "loss": 0.8712, "step": 16289 }, { "epoch": 0.83, "learning_rate": 1.519247780070534e-06, "loss": 0.9304, "step": 16290 }, { "epoch": 0.83, "learning_rate": 1.518376028706291e-06, "loss": 0.9034, "step": 16291 }, { "epoch": 0.83, "learning_rate": 1.5175045069729966e-06, "loss": 0.9464, "step": 16292 }, { "epoch": 0.83, "learning_rate": 1.5166332148942398e-06, "loss": 0.8744, "step": 16293 }, { "epoch": 0.83, "learning_rate": 1.5157621524936138e-06, "loss": 0.9467, "step": 16294 }, { "epoch": 0.83, "learning_rate": 1.5148913197947001e-06, "loss": 0.8155, "step": 16295 }, { "epoch": 0.83, "learning_rate": 1.5140207168210796e-06, "loss": 0.9642, "step": 16296 }, { "epoch": 0.83, "learning_rate": 1.513150343596318e-06, "loss": 0.8177, "step": 16297 }, { "epoch": 0.83, "learning_rate": 1.5122802001439806e-06, "loss": 1.0025, "step": 16298 }, { "epoch": 0.83, "learning_rate": 1.511410286487629e-06, "loss": 0.8792, "step": 16299 }, { "epoch": 0.83, "learning_rate": 1.5105406026508085e-06, "loss": 0.9464, "step": 16300 }, { "epoch": 0.83, "learning_rate": 1.5096711486570736e-06, "loss": 0.8696, "step": 16301 }, { "epoch": 0.83, "learning_rate": 1.5088019245299568e-06, "loss": 0.888, "step": 16302 }, { "epoch": 0.83, "learning_rate": 1.5079329302929936e-06, "loss": 0.8971, "step": 16303 }, { "epoch": 0.83, "learning_rate": 1.507064165969715e-06, "loss": 0.865, "step": 16304 }, { "epoch": 0.83, "learning_rate": 1.5061956315836345e-06, "loss": 0.8724, "step": 16305 }, { "epoch": 0.83, "learning_rate": 1.5053273271582714e-06, "loss": 0.8643, "step": 16306 }, { "epoch": 0.83, "learning_rate": 1.5044592527171343e-06, "loss": 0.7968, "step": 16307 }, { "epoch": 0.83, "learning_rate": 1.5035914082837256e-06, "loss": 0.9469, "step": 16308 }, { "epoch": 0.83, "learning_rate": 1.5027237938815375e-06, "loss": 0.9149, "step": 16309 }, { "epoch": 0.83, "learning_rate": 1.5018564095340627e-06, "loss": 0.8533, "step": 16310 }, { "epoch": 0.83, "learning_rate": 1.5009892552647865e-06, "loss": 0.8166, "step": 16311 }, { "epoch": 0.83, "learning_rate": 1.500122331097179e-06, "loss": 0.8799, "step": 16312 }, { "epoch": 0.83, "learning_rate": 1.4992556370547218e-06, "loss": 0.92, "step": 16313 }, { "epoch": 0.83, "learning_rate": 1.4983891731608714e-06, "loss": 0.9973, "step": 16314 }, { "epoch": 0.83, "learning_rate": 1.497522939439089e-06, "loss": 0.9219, "step": 16315 }, { "epoch": 0.83, "learning_rate": 1.4966569359128292e-06, "loss": 1.0137, "step": 16316 }, { "epoch": 0.83, "learning_rate": 1.495791162605531e-06, "loss": 0.8606, "step": 16317 }, { "epoch": 0.83, "learning_rate": 1.4949256195406448e-06, "loss": 0.8, "step": 16318 }, { "epoch": 0.83, "learning_rate": 1.494060306741596e-06, "loss": 0.9624, "step": 16319 }, { "epoch": 0.83, "learning_rate": 1.493195224231815e-06, "loss": 0.8688, "step": 16320 }, { "epoch": 0.83, "learning_rate": 1.492330372034725e-06, "loss": 0.861, "step": 16321 }, { "epoch": 0.83, "learning_rate": 1.491465750173736e-06, "loss": 0.8031, "step": 16322 }, { "epoch": 0.83, "learning_rate": 1.49060135867226e-06, "loss": 0.9264, "step": 16323 }, { "epoch": 0.83, "learning_rate": 1.489737197553699e-06, "loss": 0.859, "step": 16324 }, { "epoch": 0.83, "learning_rate": 1.488873266841452e-06, "loss": 0.887, "step": 16325 }, { "epoch": 0.83, "learning_rate": 1.4880095665589024e-06, "loss": 0.8364, "step": 16326 }, { "epoch": 0.83, "learning_rate": 1.4871460967294394e-06, "loss": 0.8516, "step": 16327 }, { "epoch": 0.83, "learning_rate": 1.48628285737644e-06, "loss": 0.7477, "step": 16328 }, { "epoch": 0.83, "learning_rate": 1.4854198485232696e-06, "loss": 0.7004, "step": 16329 }, { "epoch": 0.83, "learning_rate": 1.4845570701933031e-06, "loss": 0.9331, "step": 16330 }, { "epoch": 0.83, "learning_rate": 1.483694522409892e-06, "loss": 0.8951, "step": 16331 }, { "epoch": 0.83, "learning_rate": 1.4828322051963906e-06, "loss": 0.8055, "step": 16332 }, { "epoch": 0.83, "learning_rate": 1.4819701185761482e-06, "loss": 0.8674, "step": 16333 }, { "epoch": 0.83, "learning_rate": 1.4811082625725005e-06, "loss": 0.8308, "step": 16334 }, { "epoch": 0.83, "learning_rate": 1.4802466372087832e-06, "loss": 0.7996, "step": 16335 }, { "epoch": 0.83, "learning_rate": 1.4793852425083233e-06, "loss": 0.972, "step": 16336 }, { "epoch": 0.83, "learning_rate": 1.4785240784944455e-06, "loss": 0.8761, "step": 16337 }, { "epoch": 0.83, "learning_rate": 1.47766314519046e-06, "loss": 0.7506, "step": 16338 }, { "epoch": 0.83, "learning_rate": 1.476802442619677e-06, "loss": 0.8716, "step": 16339 }, { "epoch": 0.83, "learning_rate": 1.4759419708054025e-06, "loss": 0.8425, "step": 16340 }, { "epoch": 0.83, "learning_rate": 1.4750817297709264e-06, "loss": 0.8466, "step": 16341 }, { "epoch": 0.83, "learning_rate": 1.474221719539546e-06, "loss": 0.8935, "step": 16342 }, { "epoch": 0.83, "learning_rate": 1.4733619401345401e-06, "loss": 0.8604, "step": 16343 }, { "epoch": 0.83, "learning_rate": 1.472502391579188e-06, "loss": 0.8583, "step": 16344 }, { "epoch": 0.83, "learning_rate": 1.4716430738967636e-06, "loss": 0.8617, "step": 16345 }, { "epoch": 0.83, "learning_rate": 1.4707839871105278e-06, "loss": 0.872, "step": 16346 }, { "epoch": 0.83, "learning_rate": 1.4699251312437402e-06, "loss": 0.9109, "step": 16347 }, { "epoch": 0.83, "learning_rate": 1.4690665063196553e-06, "loss": 0.8365, "step": 16348 }, { "epoch": 0.83, "learning_rate": 1.46820811236152e-06, "loss": 0.8516, "step": 16349 }, { "epoch": 0.83, "learning_rate": 1.4673499493925703e-06, "loss": 0.7999, "step": 16350 }, { "epoch": 0.83, "learning_rate": 1.466492017436044e-06, "loss": 0.8115, "step": 16351 }, { "epoch": 0.83, "learning_rate": 1.4656343165151688e-06, "loss": 0.9283, "step": 16352 }, { "epoch": 0.83, "learning_rate": 1.4647768466531598e-06, "loss": 0.848, "step": 16353 }, { "epoch": 0.83, "learning_rate": 1.463919607873242e-06, "loss": 0.8865, "step": 16354 }, { "epoch": 0.83, "learning_rate": 1.463062600198617e-06, "loss": 0.8339, "step": 16355 }, { "epoch": 0.83, "learning_rate": 1.4622058236524894e-06, "loss": 0.9668, "step": 16356 }, { "epoch": 0.83, "learning_rate": 1.4613492782580574e-06, "loss": 0.8544, "step": 16357 }, { "epoch": 0.83, "learning_rate": 1.460492964038508e-06, "loss": 0.8439, "step": 16358 }, { "epoch": 0.83, "learning_rate": 1.4596368810170258e-06, "loss": 0.8701, "step": 16359 }, { "epoch": 0.83, "learning_rate": 1.4587810292167881e-06, "loss": 0.8719, "step": 16360 }, { "epoch": 0.83, "learning_rate": 1.4579254086609696e-06, "loss": 0.9037, "step": 16361 }, { "epoch": 0.83, "learning_rate": 1.4570700193727305e-06, "loss": 0.9451, "step": 16362 }, { "epoch": 0.83, "learning_rate": 1.4562148613752326e-06, "loss": 0.8613, "step": 16363 }, { "epoch": 0.83, "learning_rate": 1.455359934691628e-06, "loss": 0.8455, "step": 16364 }, { "epoch": 0.83, "learning_rate": 1.4545052393450586e-06, "loss": 0.7199, "step": 16365 }, { "epoch": 0.83, "learning_rate": 1.4536507753586725e-06, "loss": 0.8536, "step": 16366 }, { "epoch": 0.83, "learning_rate": 1.452796542755598e-06, "loss": 0.9125, "step": 16367 }, { "epoch": 0.83, "learning_rate": 1.4519425415589616e-06, "loss": 0.8264, "step": 16368 }, { "epoch": 0.83, "learning_rate": 1.4510887717918898e-06, "loss": 0.8544, "step": 16369 }, { "epoch": 0.83, "learning_rate": 1.4502352334774927e-06, "loss": 0.8761, "step": 16370 }, { "epoch": 0.83, "learning_rate": 1.4493819266388788e-06, "loss": 0.821, "step": 16371 }, { "epoch": 0.83, "learning_rate": 1.4485288512991535e-06, "loss": 0.8067, "step": 16372 }, { "epoch": 0.83, "learning_rate": 1.4476760074814133e-06, "loss": 0.7906, "step": 16373 }, { "epoch": 0.83, "learning_rate": 1.446823395208743e-06, "loss": 0.8791, "step": 16374 }, { "epoch": 0.83, "learning_rate": 1.4459710145042306e-06, "loss": 0.9091, "step": 16375 }, { "epoch": 0.83, "learning_rate": 1.4451188653909543e-06, "loss": 0.8862, "step": 16376 }, { "epoch": 0.83, "learning_rate": 1.444266947891978e-06, "loss": 0.7516, "step": 16377 }, { "epoch": 0.83, "learning_rate": 1.4434152620303766e-06, "loss": 0.8712, "step": 16378 }, { "epoch": 0.83, "learning_rate": 1.4425638078292015e-06, "loss": 0.8839, "step": 16379 }, { "epoch": 0.83, "learning_rate": 1.4417125853115054e-06, "loss": 0.9488, "step": 16380 }, { "epoch": 0.83, "learning_rate": 1.440861594500339e-06, "loss": 0.9238, "step": 16381 }, { "epoch": 0.83, "learning_rate": 1.440010835418737e-06, "loss": 0.8095, "step": 16382 }, { "epoch": 0.83, "learning_rate": 1.4391603080897342e-06, "loss": 0.8651, "step": 16383 }, { "epoch": 0.83, "learning_rate": 1.438310012536358e-06, "loss": 0.8779, "step": 16384 }, { "epoch": 0.83, "learning_rate": 1.437459948781631e-06, "loss": 0.8704, "step": 16385 }, { "epoch": 0.83, "learning_rate": 1.4366101168485635e-06, "loss": 0.9525, "step": 16386 }, { "epoch": 0.83, "learning_rate": 1.4357605167601674e-06, "loss": 0.7886, "step": 16387 }, { "epoch": 0.83, "learning_rate": 1.434911148539444e-06, "loss": 0.8527, "step": 16388 }, { "epoch": 0.83, "learning_rate": 1.4340620122093874e-06, "loss": 0.8919, "step": 16389 }, { "epoch": 0.83, "learning_rate": 1.433213107792991e-06, "loss": 0.8242, "step": 16390 }, { "epoch": 0.83, "learning_rate": 1.4323644353132337e-06, "loss": 0.9168, "step": 16391 }, { "epoch": 0.83, "learning_rate": 1.4315159947930934e-06, "loss": 0.9513, "step": 16392 }, { "epoch": 0.83, "learning_rate": 1.430667786255544e-06, "loss": 0.9598, "step": 16393 }, { "epoch": 0.83, "learning_rate": 1.4298198097235428e-06, "loss": 0.8832, "step": 16394 }, { "epoch": 0.83, "learning_rate": 1.4289720652200566e-06, "loss": 0.9372, "step": 16395 }, { "epoch": 0.83, "learning_rate": 1.4281245527680309e-06, "loss": 0.8472, "step": 16396 }, { "epoch": 0.83, "learning_rate": 1.4272772723904127e-06, "loss": 0.929, "step": 16397 }, { "epoch": 0.83, "learning_rate": 1.4264302241101447e-06, "loss": 0.8593, "step": 16398 }, { "epoch": 0.83, "learning_rate": 1.4255834079501541e-06, "loss": 0.8027, "step": 16399 }, { "epoch": 0.83, "learning_rate": 1.4247368239333703e-06, "loss": 0.8233, "step": 16400 }, { "epoch": 0.83, "learning_rate": 1.423890472082714e-06, "loss": 0.8529, "step": 16401 }, { "epoch": 0.83, "learning_rate": 1.4230443524211002e-06, "loss": 0.8552, "step": 16402 }, { "epoch": 0.83, "learning_rate": 1.4221984649714338e-06, "loss": 0.8956, "step": 16403 }, { "epoch": 0.83, "learning_rate": 1.4213528097566176e-06, "loss": 0.8539, "step": 16404 }, { "epoch": 0.83, "learning_rate": 1.42050738679955e-06, "loss": 0.8542, "step": 16405 }, { "epoch": 0.83, "learning_rate": 1.4196621961231116e-06, "loss": 0.8079, "step": 16406 }, { "epoch": 0.83, "learning_rate": 1.4188172377501964e-06, "loss": 0.8124, "step": 16407 }, { "epoch": 0.83, "learning_rate": 1.4179725117036713e-06, "loss": 0.7387, "step": 16408 }, { "epoch": 0.83, "learning_rate": 1.4171280180064107e-06, "loss": 0.9094, "step": 16409 }, { "epoch": 0.83, "learning_rate": 1.4162837566812792e-06, "loss": 0.8738, "step": 16410 }, { "epoch": 0.83, "learning_rate": 1.41543972775113e-06, "loss": 0.8364, "step": 16411 }, { "epoch": 0.83, "learning_rate": 1.414595931238818e-06, "loss": 1.0172, "step": 16412 }, { "epoch": 0.83, "learning_rate": 1.413752367167186e-06, "loss": 0.8713, "step": 16413 }, { "epoch": 0.83, "learning_rate": 1.4129090355590768e-06, "loss": 0.8851, "step": 16414 }, { "epoch": 0.83, "learning_rate": 1.4120659364373169e-06, "loss": 0.899, "step": 16415 }, { "epoch": 0.83, "learning_rate": 1.4112230698247342e-06, "loss": 0.8359, "step": 16416 }, { "epoch": 0.83, "learning_rate": 1.410380435744152e-06, "loss": 0.8838, "step": 16417 }, { "epoch": 0.83, "learning_rate": 1.4095380342183773e-06, "loss": 0.8406, "step": 16418 }, { "epoch": 0.83, "learning_rate": 1.4086958652702244e-06, "loss": 0.835, "step": 16419 }, { "epoch": 0.83, "learning_rate": 1.407853928922488e-06, "loss": 0.8291, "step": 16420 }, { "epoch": 0.83, "learning_rate": 1.4070122251979656e-06, "loss": 0.8953, "step": 16421 }, { "epoch": 0.83, "learning_rate": 1.4061707541194469e-06, "loss": 0.8646, "step": 16422 }, { "epoch": 0.83, "learning_rate": 1.4053295157097091e-06, "loss": 0.8069, "step": 16423 }, { "epoch": 0.83, "learning_rate": 1.4044885099915317e-06, "loss": 0.8711, "step": 16424 }, { "epoch": 0.83, "learning_rate": 1.4036477369876833e-06, "loss": 0.8346, "step": 16425 }, { "epoch": 0.83, "learning_rate": 1.4028071967209277e-06, "loss": 0.9245, "step": 16426 }, { "epoch": 0.83, "learning_rate": 1.4019668892140193e-06, "loss": 0.7612, "step": 16427 }, { "epoch": 0.83, "learning_rate": 1.4011268144897094e-06, "loss": 0.913, "step": 16428 }, { "epoch": 0.83, "learning_rate": 1.4002869725707447e-06, "loss": 0.8343, "step": 16429 }, { "epoch": 0.83, "learning_rate": 1.3994473634798555e-06, "loss": 0.9345, "step": 16430 }, { "epoch": 0.83, "learning_rate": 1.398607987239784e-06, "loss": 0.9406, "step": 16431 }, { "epoch": 0.83, "learning_rate": 1.3977688438732483e-06, "loss": 0.8796, "step": 16432 }, { "epoch": 0.83, "learning_rate": 1.3969299334029685e-06, "loss": 0.8673, "step": 16433 }, { "epoch": 0.83, "learning_rate": 1.3960912558516603e-06, "loss": 0.8534, "step": 16434 }, { "epoch": 0.83, "learning_rate": 1.3952528112420249e-06, "loss": 0.8947, "step": 16435 }, { "epoch": 0.83, "learning_rate": 1.394414599596765e-06, "loss": 0.8331, "step": 16436 }, { "epoch": 0.84, "learning_rate": 1.3935766209385747e-06, "loss": 0.8695, "step": 16437 }, { "epoch": 0.84, "learning_rate": 1.3927388752901428e-06, "loss": 0.807, "step": 16438 }, { "epoch": 0.84, "learning_rate": 1.3919013626741461e-06, "loss": 0.8343, "step": 16439 }, { "epoch": 0.84, "learning_rate": 1.3910640831132616e-06, "loss": 0.8712, "step": 16440 }, { "epoch": 0.84, "learning_rate": 1.3902270366301606e-06, "loss": 0.8396, "step": 16441 }, { "epoch": 0.84, "learning_rate": 1.3893902232474975e-06, "loss": 0.8313, "step": 16442 }, { "epoch": 0.84, "learning_rate": 1.3885536429879376e-06, "loss": 0.8918, "step": 16443 }, { "epoch": 0.84, "learning_rate": 1.3877172958741247e-06, "loss": 0.7648, "step": 16444 }, { "epoch": 0.84, "learning_rate": 1.386881181928702e-06, "loss": 0.8335, "step": 16445 }, { "epoch": 0.84, "learning_rate": 1.3860453011743103e-06, "loss": 0.8859, "step": 16446 }, { "epoch": 0.84, "learning_rate": 1.3852096536335758e-06, "loss": 0.8214, "step": 16447 }, { "epoch": 0.84, "learning_rate": 1.3843742393291238e-06, "loss": 0.8642, "step": 16448 }, { "epoch": 0.84, "learning_rate": 1.3835390582835729e-06, "loss": 0.857, "step": 16449 }, { "epoch": 0.84, "learning_rate": 1.3827041105195382e-06, "loss": 0.9569, "step": 16450 }, { "epoch": 0.84, "learning_rate": 1.3818693960596186e-06, "loss": 0.9263, "step": 16451 }, { "epoch": 0.84, "learning_rate": 1.3810349149264158e-06, "loss": 0.8369, "step": 16452 }, { "epoch": 0.84, "learning_rate": 1.380200667142525e-06, "loss": 0.87, "step": 16453 }, { "epoch": 0.84, "learning_rate": 1.3793666527305272e-06, "loss": 0.8697, "step": 16454 }, { "epoch": 0.84, "learning_rate": 1.3785328717130085e-06, "loss": 0.8635, "step": 16455 }, { "epoch": 0.84, "learning_rate": 1.3776993241125392e-06, "loss": 0.8864, "step": 16456 }, { "epoch": 0.84, "learning_rate": 1.3768660099516861e-06, "loss": 0.8955, "step": 16457 }, { "epoch": 0.84, "learning_rate": 1.376032929253014e-06, "loss": 0.8461, "step": 16458 }, { "epoch": 0.84, "learning_rate": 1.375200082039071e-06, "loss": 0.8141, "step": 16459 }, { "epoch": 0.84, "learning_rate": 1.3743674683324137e-06, "loss": 0.9604, "step": 16460 }, { "epoch": 0.84, "learning_rate": 1.3735350881555776e-06, "loss": 0.9246, "step": 16461 }, { "epoch": 0.84, "learning_rate": 1.3727029415311021e-06, "loss": 0.8882, "step": 16462 }, { "epoch": 0.84, "learning_rate": 1.3718710284815174e-06, "loss": 0.9209, "step": 16463 }, { "epoch": 0.84, "learning_rate": 1.3710393490293427e-06, "loss": 0.9364, "step": 16464 }, { "epoch": 0.84, "learning_rate": 1.3702079031970982e-06, "loss": 0.8373, "step": 16465 }, { "epoch": 0.84, "learning_rate": 1.3693766910072925e-06, "loss": 0.8297, "step": 16466 }, { "epoch": 0.84, "learning_rate": 1.368545712482433e-06, "loss": 0.8136, "step": 16467 }, { "epoch": 0.84, "learning_rate": 1.367714967645013e-06, "loss": 0.793, "step": 16468 }, { "epoch": 0.84, "learning_rate": 1.3668844565175254e-06, "loss": 0.7775, "step": 16469 }, { "epoch": 0.84, "learning_rate": 1.366054179122459e-06, "loss": 0.9154, "step": 16470 }, { "epoch": 0.84, "learning_rate": 1.365224135482286e-06, "loss": 0.8922, "step": 16471 }, { "epoch": 0.84, "learning_rate": 1.364394325619487e-06, "loss": 0.8808, "step": 16472 }, { "epoch": 0.84, "learning_rate": 1.363564749556522e-06, "loss": 0.8687, "step": 16473 }, { "epoch": 0.84, "learning_rate": 1.3627354073158528e-06, "loss": 0.8193, "step": 16474 }, { "epoch": 0.84, "learning_rate": 1.361906298919935e-06, "loss": 0.8866, "step": 16475 }, { "epoch": 0.84, "learning_rate": 1.3610774243912117e-06, "loss": 0.8254, "step": 16476 }, { "epoch": 0.84, "learning_rate": 1.360248783752126e-06, "loss": 0.8326, "step": 16477 }, { "epoch": 0.84, "learning_rate": 1.3594203770251134e-06, "loss": 0.9283, "step": 16478 }, { "epoch": 0.84, "learning_rate": 1.3585922042326028e-06, "loss": 0.9503, "step": 16479 }, { "epoch": 0.84, "learning_rate": 1.3577642653970124e-06, "loss": 0.9777, "step": 16480 }, { "epoch": 0.84, "learning_rate": 1.3569365605407604e-06, "loss": 0.7792, "step": 16481 }, { "epoch": 0.84, "learning_rate": 1.3561090896862573e-06, "loss": 0.7474, "step": 16482 }, { "epoch": 0.84, "learning_rate": 1.3552818528558997e-06, "loss": 0.9378, "step": 16483 }, { "epoch": 0.84, "learning_rate": 1.3544548500720933e-06, "loss": 0.9414, "step": 16484 }, { "epoch": 0.84, "learning_rate": 1.353628081357221e-06, "loss": 0.8102, "step": 16485 }, { "epoch": 0.84, "learning_rate": 1.3528015467336698e-06, "loss": 0.9214, "step": 16486 }, { "epoch": 0.84, "learning_rate": 1.3519752462238188e-06, "loss": 0.9396, "step": 16487 }, { "epoch": 0.84, "learning_rate": 1.3511491798500353e-06, "loss": 0.8569, "step": 16488 }, { "epoch": 0.84, "learning_rate": 1.3503233476346855e-06, "loss": 0.9332, "step": 16489 }, { "epoch": 0.84, "learning_rate": 1.3494977496001293e-06, "loss": 0.7606, "step": 16490 }, { "epoch": 0.84, "learning_rate": 1.34867238576872e-06, "loss": 0.8258, "step": 16491 }, { "epoch": 0.84, "learning_rate": 1.3478472561627997e-06, "loss": 0.8911, "step": 16492 }, { "epoch": 0.84, "learning_rate": 1.3470223608047094e-06, "loss": 0.8594, "step": 16493 }, { "epoch": 0.84, "learning_rate": 1.3461976997167846e-06, "loss": 0.8955, "step": 16494 }, { "epoch": 0.84, "learning_rate": 1.3453732729213464e-06, "loss": 0.8142, "step": 16495 }, { "epoch": 0.84, "learning_rate": 1.3445490804407235e-06, "loss": 0.8298, "step": 16496 }, { "epoch": 0.84, "learning_rate": 1.3437251222972226e-06, "loss": 0.8712, "step": 16497 }, { "epoch": 0.84, "learning_rate": 1.3429013985131545e-06, "loss": 0.8036, "step": 16498 }, { "epoch": 0.84, "learning_rate": 1.3420779091108237e-06, "loss": 0.9036, "step": 16499 }, { "epoch": 0.84, "learning_rate": 1.3412546541125192e-06, "loss": 0.8677, "step": 16500 }, { "epoch": 0.84, "learning_rate": 1.3404316335405332e-06, "loss": 0.8202, "step": 16501 }, { "epoch": 0.84, "learning_rate": 1.3396088474171475e-06, "loss": 0.8134, "step": 16502 }, { "epoch": 0.84, "learning_rate": 1.3387862957646403e-06, "loss": 0.836, "step": 16503 }, { "epoch": 0.84, "learning_rate": 1.3379639786052778e-06, "loss": 0.8763, "step": 16504 }, { "epoch": 0.84, "learning_rate": 1.3371418959613236e-06, "loss": 0.873, "step": 16505 }, { "epoch": 0.84, "learning_rate": 1.3363200478550397e-06, "loss": 0.8544, "step": 16506 }, { "epoch": 0.84, "learning_rate": 1.3354984343086675e-06, "loss": 0.9461, "step": 16507 }, { "epoch": 0.84, "learning_rate": 1.334677055344462e-06, "loss": 0.8508, "step": 16508 }, { "epoch": 0.84, "learning_rate": 1.3338559109846538e-06, "loss": 0.92, "step": 16509 }, { "epoch": 0.84, "learning_rate": 1.333035001251477e-06, "loss": 0.8886, "step": 16510 }, { "epoch": 0.84, "learning_rate": 1.3322143261671584e-06, "loss": 0.8754, "step": 16511 }, { "epoch": 0.84, "learning_rate": 1.3313938857539133e-06, "loss": 0.8833, "step": 16512 }, { "epoch": 0.84, "learning_rate": 1.3305736800339553e-06, "loss": 0.8675, "step": 16513 }, { "epoch": 0.84, "learning_rate": 1.329753709029492e-06, "loss": 0.8884, "step": 16514 }, { "epoch": 0.84, "learning_rate": 1.3289339727627249e-06, "loss": 0.9052, "step": 16515 }, { "epoch": 0.84, "learning_rate": 1.3281144712558424e-06, "loss": 0.7197, "step": 16516 }, { "epoch": 0.84, "learning_rate": 1.327295204531035e-06, "loss": 0.848, "step": 16517 }, { "epoch": 0.84, "learning_rate": 1.3264761726104847e-06, "loss": 0.8249, "step": 16518 }, { "epoch": 0.84, "learning_rate": 1.3256573755163593e-06, "loss": 0.8928, "step": 16519 }, { "epoch": 0.84, "learning_rate": 1.3248388132708368e-06, "loss": 0.7879, "step": 16520 }, { "epoch": 0.84, "learning_rate": 1.3240204858960715e-06, "loss": 0.8472, "step": 16521 }, { "epoch": 0.84, "learning_rate": 1.3232023934142213e-06, "loss": 0.9196, "step": 16522 }, { "epoch": 0.84, "learning_rate": 1.3223845358474363e-06, "loss": 0.9539, "step": 16523 }, { "epoch": 0.84, "learning_rate": 1.3215669132178566e-06, "loss": 0.8673, "step": 16524 }, { "epoch": 0.84, "learning_rate": 1.3207495255476188e-06, "loss": 0.9012, "step": 16525 }, { "epoch": 0.84, "learning_rate": 1.3199323728588543e-06, "loss": 0.821, "step": 16526 }, { "epoch": 0.84, "learning_rate": 1.319115455173685e-06, "loss": 0.8891, "step": 16527 }, { "epoch": 0.84, "learning_rate": 1.3182987725142315e-06, "loss": 0.883, "step": 16528 }, { "epoch": 0.84, "learning_rate": 1.3174823249026003e-06, "loss": 0.8879, "step": 16529 }, { "epoch": 0.84, "learning_rate": 1.316666112360897e-06, "loss": 0.8879, "step": 16530 }, { "epoch": 0.84, "learning_rate": 1.3158501349112219e-06, "loss": 0.8959, "step": 16531 }, { "epoch": 0.84, "learning_rate": 1.315034392575666e-06, "loss": 0.8401, "step": 16532 }, { "epoch": 0.84, "learning_rate": 1.3142188853763116e-06, "loss": 0.8534, "step": 16533 }, { "epoch": 0.84, "learning_rate": 1.31340361333524e-06, "loss": 0.8518, "step": 16534 }, { "epoch": 0.84, "learning_rate": 1.3125885764745271e-06, "loss": 0.8584, "step": 16535 }, { "epoch": 0.84, "learning_rate": 1.3117737748162306e-06, "loss": 0.8919, "step": 16536 }, { "epoch": 0.84, "learning_rate": 1.3109592083824196e-06, "loss": 0.8982, "step": 16537 }, { "epoch": 0.84, "learning_rate": 1.310144877195142e-06, "loss": 0.8802, "step": 16538 }, { "epoch": 0.84, "learning_rate": 1.3093307812764466e-06, "loss": 0.9098, "step": 16539 }, { "epoch": 0.84, "learning_rate": 1.308516920648376e-06, "loss": 0.7529, "step": 16540 }, { "epoch": 0.84, "learning_rate": 1.3077032953329604e-06, "loss": 0.9275, "step": 16541 }, { "epoch": 0.84, "learning_rate": 1.306889905352231e-06, "loss": 0.8241, "step": 16542 }, { "epoch": 0.84, "learning_rate": 1.306076750728208e-06, "loss": 0.8727, "step": 16543 }, { "epoch": 0.84, "learning_rate": 1.3052638314829102e-06, "loss": 0.8832, "step": 16544 }, { "epoch": 0.84, "learning_rate": 1.3044511476383404e-06, "loss": 0.851, "step": 16545 }, { "epoch": 0.84, "learning_rate": 1.303638699216505e-06, "loss": 0.9094, "step": 16546 }, { "epoch": 0.84, "learning_rate": 1.3028264862394024e-06, "loss": 0.878, "step": 16547 }, { "epoch": 0.84, "learning_rate": 1.3020145087290149e-06, "loss": 0.9009, "step": 16548 }, { "epoch": 0.84, "learning_rate": 1.3012027667073346e-06, "loss": 0.888, "step": 16549 }, { "epoch": 0.84, "learning_rate": 1.300391260196333e-06, "loss": 0.8754, "step": 16550 }, { "epoch": 0.84, "learning_rate": 1.2995799892179818e-06, "loss": 0.8876, "step": 16551 }, { "epoch": 0.84, "learning_rate": 1.2987689537942483e-06, "loss": 0.8919, "step": 16552 }, { "epoch": 0.84, "learning_rate": 1.2979581539470854e-06, "loss": 0.9977, "step": 16553 }, { "epoch": 0.84, "learning_rate": 1.2971475896984475e-06, "loss": 0.8524, "step": 16554 }, { "epoch": 0.84, "learning_rate": 1.2963372610702806e-06, "loss": 0.9153, "step": 16555 }, { "epoch": 0.84, "learning_rate": 1.2955271680845227e-06, "loss": 0.806, "step": 16556 }, { "epoch": 0.84, "learning_rate": 1.294717310763105e-06, "loss": 0.868, "step": 16557 }, { "epoch": 0.84, "learning_rate": 1.2939076891279545e-06, "loss": 0.7772, "step": 16558 }, { "epoch": 0.84, "learning_rate": 1.2930983032009925e-06, "loss": 0.7909, "step": 16559 }, { "epoch": 0.84, "learning_rate": 1.292289153004126e-06, "loss": 0.8609, "step": 16560 }, { "epoch": 0.84, "learning_rate": 1.2914802385592706e-06, "loss": 0.9245, "step": 16561 }, { "epoch": 0.84, "learning_rate": 1.2906715598883212e-06, "loss": 0.7876, "step": 16562 }, { "epoch": 0.84, "learning_rate": 1.2898631170131725e-06, "loss": 0.861, "step": 16563 }, { "epoch": 0.84, "learning_rate": 1.2890549099557158e-06, "loss": 0.8032, "step": 16564 }, { "epoch": 0.84, "learning_rate": 1.2882469387378271e-06, "loss": 0.871, "step": 16565 }, { "epoch": 0.84, "learning_rate": 1.2874392033813844e-06, "loss": 0.8979, "step": 16566 }, { "epoch": 0.84, "learning_rate": 1.2866317039082554e-06, "loss": 0.8448, "step": 16567 }, { "epoch": 0.84, "learning_rate": 1.2858244403403052e-06, "loss": 0.8254, "step": 16568 }, { "epoch": 0.84, "learning_rate": 1.2850174126993852e-06, "loss": 0.8762, "step": 16569 }, { "epoch": 0.84, "learning_rate": 1.2842106210073468e-06, "loss": 0.8087, "step": 16570 }, { "epoch": 0.84, "learning_rate": 1.2834040652860347e-06, "loss": 0.9124, "step": 16571 }, { "epoch": 0.84, "learning_rate": 1.282597745557279e-06, "loss": 0.8585, "step": 16572 }, { "epoch": 0.84, "learning_rate": 1.2817916618429194e-06, "loss": 0.8467, "step": 16573 }, { "epoch": 0.84, "learning_rate": 1.2809858141647734e-06, "loss": 0.8877, "step": 16574 }, { "epoch": 0.84, "learning_rate": 1.2801802025446598e-06, "loss": 0.8896, "step": 16575 }, { "epoch": 0.84, "learning_rate": 1.2793748270043925e-06, "loss": 0.8539, "step": 16576 }, { "epoch": 0.84, "learning_rate": 1.2785696875657716e-06, "loss": 0.8361, "step": 16577 }, { "epoch": 0.84, "learning_rate": 1.2777647842505968e-06, "loss": 0.8506, "step": 16578 }, { "epoch": 0.84, "learning_rate": 1.2769601170806622e-06, "loss": 0.9045, "step": 16579 }, { "epoch": 0.84, "learning_rate": 1.2761556860777525e-06, "loss": 0.8403, "step": 16580 }, { "epoch": 0.84, "learning_rate": 1.2753514912636445e-06, "loss": 0.8303, "step": 16581 }, { "epoch": 0.84, "learning_rate": 1.2745475326601132e-06, "loss": 0.8734, "step": 16582 }, { "epoch": 0.84, "learning_rate": 1.2737438102889256e-06, "loss": 0.8167, "step": 16583 }, { "epoch": 0.84, "learning_rate": 1.2729403241718364e-06, "loss": 0.8955, "step": 16584 }, { "epoch": 0.84, "learning_rate": 1.2721370743306072e-06, "loss": 0.8384, "step": 16585 }, { "epoch": 0.84, "learning_rate": 1.2713340607869795e-06, "loss": 0.9578, "step": 16586 }, { "epoch": 0.84, "learning_rate": 1.2705312835626948e-06, "loss": 0.9765, "step": 16587 }, { "epoch": 0.84, "learning_rate": 1.2697287426794902e-06, "loss": 0.8128, "step": 16588 }, { "epoch": 0.84, "learning_rate": 1.2689264381590904e-06, "loss": 1.0186, "step": 16589 }, { "epoch": 0.84, "learning_rate": 1.268124370023217e-06, "loss": 0.8766, "step": 16590 }, { "epoch": 0.84, "learning_rate": 1.2673225382935873e-06, "loss": 0.8606, "step": 16591 }, { "epoch": 0.84, "learning_rate": 1.2665209429919112e-06, "loss": 0.7545, "step": 16592 }, { "epoch": 0.84, "learning_rate": 1.2657195841398861e-06, "loss": 0.8278, "step": 16593 }, { "epoch": 0.84, "learning_rate": 1.2649184617592104e-06, "loss": 0.942, "step": 16594 }, { "epoch": 0.84, "learning_rate": 1.264117575871575e-06, "loss": 0.8202, "step": 16595 }, { "epoch": 0.84, "learning_rate": 1.263316926498661e-06, "loss": 0.8406, "step": 16596 }, { "epoch": 0.84, "learning_rate": 1.2625165136621486e-06, "loss": 0.8004, "step": 16597 }, { "epoch": 0.84, "learning_rate": 1.2617163373837038e-06, "loss": 0.988, "step": 16598 }, { "epoch": 0.84, "learning_rate": 1.2609163976849925e-06, "loss": 0.8577, "step": 16599 }, { "epoch": 0.84, "learning_rate": 1.2601166945876742e-06, "loss": 0.8781, "step": 16600 }, { "epoch": 0.84, "learning_rate": 1.2593172281133925e-06, "loss": 0.8306, "step": 16601 }, { "epoch": 0.84, "learning_rate": 1.2585179982838026e-06, "loss": 0.933, "step": 16602 }, { "epoch": 0.84, "learning_rate": 1.2577190051205356e-06, "loss": 0.8758, "step": 16603 }, { "epoch": 0.84, "learning_rate": 1.2569202486452258e-06, "loss": 0.8378, "step": 16604 }, { "epoch": 0.84, "learning_rate": 1.2561217288795003e-06, "loss": 0.8984, "step": 16605 }, { "epoch": 0.84, "learning_rate": 1.2553234458449737e-06, "loss": 0.9282, "step": 16606 }, { "epoch": 0.84, "learning_rate": 1.2545253995632612e-06, "loss": 0.8708, "step": 16607 }, { "epoch": 0.84, "learning_rate": 1.2537275900559686e-06, "loss": 0.8481, "step": 16608 }, { "epoch": 0.84, "learning_rate": 1.2529300173447e-06, "loss": 0.8978, "step": 16609 }, { "epoch": 0.84, "learning_rate": 1.2521326814510416e-06, "loss": 0.9093, "step": 16610 }, { "epoch": 0.84, "learning_rate": 1.2513355823965845e-06, "loss": 0.9199, "step": 16611 }, { "epoch": 0.84, "learning_rate": 1.2505387202029095e-06, "loss": 0.7353, "step": 16612 }, { "epoch": 0.84, "learning_rate": 1.2497420948915872e-06, "loss": 0.8832, "step": 16613 }, { "epoch": 0.84, "learning_rate": 1.2489457064841914e-06, "loss": 0.8519, "step": 16614 }, { "epoch": 0.84, "learning_rate": 1.2481495550022792e-06, "loss": 0.8134, "step": 16615 }, { "epoch": 0.84, "learning_rate": 1.2473536404674058e-06, "loss": 0.7789, "step": 16616 }, { "epoch": 0.84, "learning_rate": 1.2465579629011226e-06, "loss": 0.7644, "step": 16617 }, { "epoch": 0.84, "learning_rate": 1.245762522324967e-06, "loss": 0.8766, "step": 16618 }, { "epoch": 0.84, "learning_rate": 1.2449673187604782e-06, "loss": 0.9564, "step": 16619 }, { "epoch": 0.84, "learning_rate": 1.2441723522291848e-06, "loss": 0.9407, "step": 16620 }, { "epoch": 0.84, "learning_rate": 1.2433776227526105e-06, "loss": 0.8932, "step": 16621 }, { "epoch": 0.84, "learning_rate": 1.2425831303522695e-06, "loss": 0.8657, "step": 16622 }, { "epoch": 0.84, "learning_rate": 1.241788875049673e-06, "loss": 0.7891, "step": 16623 }, { "epoch": 0.84, "learning_rate": 1.2409948568663265e-06, "loss": 0.7747, "step": 16624 }, { "epoch": 0.84, "learning_rate": 1.2402010758237215e-06, "loss": 0.8994, "step": 16625 }, { "epoch": 0.84, "learning_rate": 1.2394075319433574e-06, "loss": 0.8758, "step": 16626 }, { "epoch": 0.84, "learning_rate": 1.2386142252467103e-06, "loss": 0.9102, "step": 16627 }, { "epoch": 0.84, "learning_rate": 1.2378211557552633e-06, "loss": 0.8658, "step": 16628 }, { "epoch": 0.84, "learning_rate": 1.2370283234904868e-06, "loss": 0.8925, "step": 16629 }, { "epoch": 0.84, "learning_rate": 1.2362357284738445e-06, "loss": 0.9173, "step": 16630 }, { "epoch": 0.84, "learning_rate": 1.2354433707267954e-06, "loss": 0.9641, "step": 16631 }, { "epoch": 0.84, "learning_rate": 1.234651250270792e-06, "loss": 0.8885, "step": 16632 }, { "epoch": 0.85, "learning_rate": 1.2338593671272837e-06, "loss": 0.9023, "step": 16633 }, { "epoch": 0.85, "learning_rate": 1.2330677213177034e-06, "loss": 0.8658, "step": 16634 }, { "epoch": 0.85, "learning_rate": 1.2322763128634873e-06, "loss": 0.8859, "step": 16635 }, { "epoch": 0.85, "learning_rate": 1.2314851417860641e-06, "loss": 0.8661, "step": 16636 }, { "epoch": 0.85, "learning_rate": 1.2306942081068474e-06, "loss": 0.8832, "step": 16637 }, { "epoch": 0.85, "learning_rate": 1.2299035118472602e-06, "loss": 0.8295, "step": 16638 }, { "epoch": 0.85, "learning_rate": 1.2291130530287022e-06, "loss": 0.8991, "step": 16639 }, { "epoch": 0.85, "learning_rate": 1.228322831672576e-06, "loss": 0.8815, "step": 16640 }, { "epoch": 0.85, "learning_rate": 1.22753284780028e-06, "loss": 0.913, "step": 16641 }, { "epoch": 0.85, "learning_rate": 1.2267431014331966e-06, "loss": 0.8636, "step": 16642 }, { "epoch": 0.85, "learning_rate": 1.225953592592709e-06, "loss": 0.9935, "step": 16643 }, { "epoch": 0.85, "learning_rate": 1.225164321300194e-06, "loss": 0.8373, "step": 16644 }, { "epoch": 0.85, "learning_rate": 1.224375287577021e-06, "loss": 0.8399, "step": 16645 }, { "epoch": 0.85, "learning_rate": 1.2235864914445483e-06, "loss": 0.8585, "step": 16646 }, { "epoch": 0.85, "learning_rate": 1.2227979329241346e-06, "loss": 0.8829, "step": 16647 }, { "epoch": 0.85, "learning_rate": 1.2220096120371306e-06, "loss": 0.9019, "step": 16648 }, { "epoch": 0.85, "learning_rate": 1.2212215288048723e-06, "loss": 0.9338, "step": 16649 }, { "epoch": 0.85, "learning_rate": 1.2204336832487063e-06, "loss": 0.8834, "step": 16650 }, { "epoch": 0.85, "learning_rate": 1.2196460753899552e-06, "loss": 0.8234, "step": 16651 }, { "epoch": 0.85, "learning_rate": 1.2188587052499445e-06, "loss": 0.7968, "step": 16652 }, { "epoch": 0.85, "learning_rate": 1.2180715728499948e-06, "loss": 0.9332, "step": 16653 }, { "epoch": 0.85, "learning_rate": 1.2172846782114123e-06, "loss": 0.8461, "step": 16654 }, { "epoch": 0.85, "learning_rate": 1.2164980213555023e-06, "loss": 0.9319, "step": 16655 }, { "epoch": 0.85, "learning_rate": 1.215711602303563e-06, "loss": 0.8047, "step": 16656 }, { "epoch": 0.85, "learning_rate": 1.21492542107689e-06, "loss": 0.9735, "step": 16657 }, { "epoch": 0.85, "learning_rate": 1.2141394776967607e-06, "loss": 0.9637, "step": 16658 }, { "epoch": 0.85, "learning_rate": 1.2133537721844591e-06, "loss": 0.8285, "step": 16659 }, { "epoch": 0.85, "learning_rate": 1.2125683045612568e-06, "loss": 0.8796, "step": 16660 }, { "epoch": 0.85, "learning_rate": 1.2117830748484149e-06, "loss": 0.922, "step": 16661 }, { "epoch": 0.85, "learning_rate": 1.2109980830672007e-06, "loss": 0.8062, "step": 16662 }, { "epoch": 0.85, "learning_rate": 1.2102133292388608e-06, "loss": 0.887, "step": 16663 }, { "epoch": 0.85, "learning_rate": 1.2094288133846432e-06, "loss": 0.8781, "step": 16664 }, { "epoch": 0.85, "learning_rate": 1.2086445355257902e-06, "loss": 0.8883, "step": 16665 }, { "epoch": 0.85, "learning_rate": 1.2078604956835283e-06, "loss": 0.8777, "step": 16666 }, { "epoch": 0.85, "learning_rate": 1.2070766938790945e-06, "loss": 0.8456, "step": 16667 }, { "epoch": 0.85, "learning_rate": 1.206293130133701e-06, "loss": 0.9821, "step": 16668 }, { "epoch": 0.85, "learning_rate": 1.2055098044685665e-06, "loss": 0.9067, "step": 16669 }, { "epoch": 0.85, "learning_rate": 1.2047267169048982e-06, "loss": 0.8866, "step": 16670 }, { "epoch": 0.85, "learning_rate": 1.2039438674638958e-06, "loss": 0.7953, "step": 16671 }, { "epoch": 0.85, "learning_rate": 1.203161256166754e-06, "loss": 0.7892, "step": 16672 }, { "epoch": 0.85, "learning_rate": 1.2023788830346628e-06, "loss": 0.9016, "step": 16673 }, { "epoch": 0.85, "learning_rate": 1.2015967480888048e-06, "loss": 0.891, "step": 16674 }, { "epoch": 0.85, "learning_rate": 1.2008148513503525e-06, "loss": 0.927, "step": 16675 }, { "epoch": 0.85, "learning_rate": 1.2000331928404763e-06, "loss": 0.9553, "step": 16676 }, { "epoch": 0.85, "learning_rate": 1.1992517725803409e-06, "loss": 0.8387, "step": 16677 }, { "epoch": 0.85, "learning_rate": 1.1984705905910965e-06, "loss": 0.8434, "step": 16678 }, { "epoch": 0.85, "learning_rate": 1.1976896468939015e-06, "loss": 0.8338, "step": 16679 }, { "epoch": 0.85, "learning_rate": 1.196908941509891e-06, "loss": 0.8893, "step": 16680 }, { "epoch": 0.85, "learning_rate": 1.1961284744602064e-06, "loss": 0.9092, "step": 16681 }, { "epoch": 0.85, "learning_rate": 1.1953482457659783e-06, "loss": 0.8224, "step": 16682 }, { "epoch": 0.85, "learning_rate": 1.1945682554483272e-06, "loss": 0.8967, "step": 16683 }, { "epoch": 0.85, "learning_rate": 1.1937885035283724e-06, "loss": 0.8278, "step": 16684 }, { "epoch": 0.85, "learning_rate": 1.1930089900272245e-06, "loss": 0.8184, "step": 16685 }, { "epoch": 0.85, "learning_rate": 1.19222971496599e-06, "loss": 0.8281, "step": 16686 }, { "epoch": 0.85, "learning_rate": 1.1914506783657642e-06, "loss": 0.7938, "step": 16687 }, { "epoch": 0.85, "learning_rate": 1.1906718802476391e-06, "loss": 0.9644, "step": 16688 }, { "epoch": 0.85, "learning_rate": 1.1898933206327024e-06, "loss": 0.9073, "step": 16689 }, { "epoch": 0.85, "learning_rate": 1.189114999542027e-06, "loss": 0.8776, "step": 16690 }, { "epoch": 0.85, "learning_rate": 1.1883369169966929e-06, "loss": 0.9003, "step": 16691 }, { "epoch": 0.85, "learning_rate": 1.18755907301776e-06, "loss": 0.88, "step": 16692 }, { "epoch": 0.85, "learning_rate": 1.186781467626289e-06, "loss": 0.8333, "step": 16693 }, { "epoch": 0.85, "learning_rate": 1.1860041008433365e-06, "loss": 0.8375, "step": 16694 }, { "epoch": 0.85, "learning_rate": 1.1852269726899423e-06, "loss": 0.8613, "step": 16695 }, { "epoch": 0.85, "learning_rate": 1.1844500831871497e-06, "loss": 0.935, "step": 16696 }, { "epoch": 0.85, "learning_rate": 1.1836734323559929e-06, "loss": 0.8536, "step": 16697 }, { "epoch": 0.85, "learning_rate": 1.1828970202174993e-06, "loss": 0.8366, "step": 16698 }, { "epoch": 0.85, "learning_rate": 1.182120846792687e-06, "loss": 0.8779, "step": 16699 }, { "epoch": 0.85, "learning_rate": 1.1813449121025711e-06, "loss": 0.831, "step": 16700 }, { "epoch": 0.85, "learning_rate": 1.1805692161681603e-06, "loss": 0.8018, "step": 16701 }, { "epoch": 0.85, "learning_rate": 1.1797937590104524e-06, "loss": 0.9067, "step": 16702 }, { "epoch": 0.85, "learning_rate": 1.1790185406504474e-06, "loss": 0.8378, "step": 16703 }, { "epoch": 0.85, "learning_rate": 1.1782435611091291e-06, "loss": 0.8622, "step": 16704 }, { "epoch": 0.85, "learning_rate": 1.1774688204074802e-06, "loss": 0.8085, "step": 16705 }, { "epoch": 0.85, "learning_rate": 1.1766943185664791e-06, "loss": 0.9478, "step": 16706 }, { "epoch": 0.85, "learning_rate": 1.1759200556070904e-06, "loss": 0.9235, "step": 16707 }, { "epoch": 0.85, "learning_rate": 1.1751460315502783e-06, "loss": 0.7877, "step": 16708 }, { "epoch": 0.85, "learning_rate": 1.1743722464169971e-06, "loss": 0.8989, "step": 16709 }, { "epoch": 0.85, "learning_rate": 1.1735987002282013e-06, "loss": 1.0226, "step": 16710 }, { "epoch": 0.85, "learning_rate": 1.1728253930048272e-06, "loss": 0.9683, "step": 16711 }, { "epoch": 0.85, "learning_rate": 1.172052324767814e-06, "loss": 0.828, "step": 16712 }, { "epoch": 0.85, "learning_rate": 1.1712794955380946e-06, "loss": 0.8916, "step": 16713 }, { "epoch": 0.85, "learning_rate": 1.1705069053365846e-06, "loss": 0.8541, "step": 16714 }, { "epoch": 0.85, "learning_rate": 1.1697345541842108e-06, "loss": 0.8604, "step": 16715 }, { "epoch": 0.85, "learning_rate": 1.1689624421018787e-06, "loss": 0.8538, "step": 16716 }, { "epoch": 0.85, "learning_rate": 1.168190569110491e-06, "loss": 0.8959, "step": 16717 }, { "epoch": 0.85, "learning_rate": 1.1674189352309505e-06, "loss": 0.8154, "step": 16718 }, { "epoch": 0.85, "learning_rate": 1.1666475404841427e-06, "loss": 0.8927, "step": 16719 }, { "epoch": 0.85, "learning_rate": 1.1658763848909548e-06, "loss": 0.8895, "step": 16720 }, { "epoch": 0.85, "learning_rate": 1.1651054684722662e-06, "loss": 0.9548, "step": 16721 }, { "epoch": 0.85, "learning_rate": 1.1643347912489478e-06, "loss": 0.8963, "step": 16722 }, { "epoch": 0.85, "learning_rate": 1.1635643532418638e-06, "loss": 0.938, "step": 16723 }, { "epoch": 0.85, "learning_rate": 1.1627941544718724e-06, "loss": 0.8643, "step": 16724 }, { "epoch": 0.85, "learning_rate": 1.16202419495983e-06, "loss": 0.783, "step": 16725 }, { "epoch": 0.85, "learning_rate": 1.1612544747265764e-06, "loss": 0.9353, "step": 16726 }, { "epoch": 0.85, "learning_rate": 1.1604849937929573e-06, "loss": 0.924, "step": 16727 }, { "epoch": 0.85, "learning_rate": 1.1597157521798008e-06, "loss": 0.8309, "step": 16728 }, { "epoch": 0.85, "learning_rate": 1.1589467499079343e-06, "loss": 0.9991, "step": 16729 }, { "epoch": 0.85, "learning_rate": 1.1581779869981813e-06, "loss": 0.8715, "step": 16730 }, { "epoch": 0.85, "learning_rate": 1.1574094634713495e-06, "loss": 0.853, "step": 16731 }, { "epoch": 0.85, "learning_rate": 1.156641179348249e-06, "loss": 0.9419, "step": 16732 }, { "epoch": 0.85, "learning_rate": 1.15587313464968e-06, "loss": 0.7719, "step": 16733 }, { "epoch": 0.85, "learning_rate": 1.1551053293964365e-06, "loss": 0.9136, "step": 16734 }, { "epoch": 0.85, "learning_rate": 1.154337763609309e-06, "loss": 0.8481, "step": 16735 }, { "epoch": 0.85, "learning_rate": 1.1535704373090716e-06, "loss": 0.8062, "step": 16736 }, { "epoch": 0.85, "learning_rate": 1.1528033505165036e-06, "loss": 0.8382, "step": 16737 }, { "epoch": 0.85, "learning_rate": 1.1520365032523727e-06, "loss": 0.8896, "step": 16738 }, { "epoch": 0.85, "learning_rate": 1.1512698955374412e-06, "loss": 0.8167, "step": 16739 }, { "epoch": 0.85, "learning_rate": 1.1505035273924613e-06, "loss": 0.8637, "step": 16740 }, { "epoch": 0.85, "learning_rate": 1.1497373988381843e-06, "loss": 0.9197, "step": 16741 }, { "epoch": 0.85, "learning_rate": 1.1489715098953524e-06, "loss": 0.9381, "step": 16742 }, { "epoch": 0.85, "learning_rate": 1.1482058605846958e-06, "loss": 0.8927, "step": 16743 }, { "epoch": 0.85, "learning_rate": 1.1474404509269522e-06, "loss": 0.9132, "step": 16744 }, { "epoch": 0.85, "learning_rate": 1.1466752809428384e-06, "loss": 0.8559, "step": 16745 }, { "epoch": 0.85, "learning_rate": 1.145910350653071e-06, "loss": 0.842, "step": 16746 }, { "epoch": 0.85, "learning_rate": 1.145145660078364e-06, "loss": 0.9146, "step": 16747 }, { "epoch": 0.85, "learning_rate": 1.1443812092394146e-06, "loss": 0.8682, "step": 16748 }, { "epoch": 0.85, "learning_rate": 1.1436169981569234e-06, "loss": 0.8756, "step": 16749 }, { "epoch": 0.85, "learning_rate": 1.1428530268515791e-06, "loss": 0.8923, "step": 16750 }, { "epoch": 0.85, "learning_rate": 1.1420892953440676e-06, "loss": 0.9193, "step": 16751 }, { "epoch": 0.85, "learning_rate": 1.1413258036550623e-06, "loss": 0.9048, "step": 16752 }, { "epoch": 0.85, "learning_rate": 1.1405625518052354e-06, "loss": 0.8453, "step": 16753 }, { "epoch": 0.85, "learning_rate": 1.139799539815255e-06, "loss": 0.8513, "step": 16754 }, { "epoch": 0.85, "learning_rate": 1.1390367677057712e-06, "loss": 0.9268, "step": 16755 }, { "epoch": 0.85, "learning_rate": 1.138274235497443e-06, "loss": 0.8417, "step": 16756 }, { "epoch": 0.85, "learning_rate": 1.1375119432109094e-06, "loss": 0.8121, "step": 16757 }, { "epoch": 0.85, "learning_rate": 1.1367498908668118e-06, "loss": 0.8119, "step": 16758 }, { "epoch": 0.85, "learning_rate": 1.1359880784857835e-06, "loss": 0.8565, "step": 16759 }, { "epoch": 0.85, "learning_rate": 1.1352265060884448e-06, "loss": 0.9032, "step": 16760 }, { "epoch": 0.85, "learning_rate": 1.134465173695417e-06, "loss": 0.8872, "step": 16761 }, { "epoch": 0.85, "learning_rate": 1.1337040813273125e-06, "loss": 0.8974, "step": 16762 }, { "epoch": 0.85, "learning_rate": 1.1329432290047394e-06, "loss": 0.8781, "step": 16763 }, { "epoch": 0.85, "learning_rate": 1.132182616748292e-06, "loss": 0.8452, "step": 16764 }, { "epoch": 0.85, "learning_rate": 1.1314222445785662e-06, "loss": 0.8998, "step": 16765 }, { "epoch": 0.85, "learning_rate": 1.13066211251615e-06, "loss": 0.8846, "step": 16766 }, { "epoch": 0.85, "learning_rate": 1.1299022205816158e-06, "loss": 0.9003, "step": 16767 }, { "epoch": 0.85, "learning_rate": 1.129142568795547e-06, "loss": 0.7765, "step": 16768 }, { "epoch": 0.85, "learning_rate": 1.128383157178502e-06, "loss": 0.8114, "step": 16769 }, { "epoch": 0.85, "learning_rate": 1.1276239857510451e-06, "loss": 0.8475, "step": 16770 }, { "epoch": 0.85, "learning_rate": 1.1268650545337312e-06, "loss": 0.9991, "step": 16771 }, { "epoch": 0.85, "learning_rate": 1.1261063635471047e-06, "loss": 0.8853, "step": 16772 }, { "epoch": 0.85, "learning_rate": 1.1253479128117062e-06, "loss": 0.8286, "step": 16773 }, { "epoch": 0.85, "learning_rate": 1.1245897023480712e-06, "loss": 0.8063, "step": 16774 }, { "epoch": 0.85, "learning_rate": 1.1238317321767288e-06, "loss": 0.9705, "step": 16775 }, { "epoch": 0.85, "learning_rate": 1.1230740023181975e-06, "loss": 0.847, "step": 16776 }, { "epoch": 0.85, "learning_rate": 1.1223165127929935e-06, "loss": 0.9031, "step": 16777 }, { "epoch": 0.85, "learning_rate": 1.1215592636216266e-06, "loss": 0.8907, "step": 16778 }, { "epoch": 0.85, "learning_rate": 1.1208022548245923e-06, "loss": 0.8039, "step": 16779 }, { "epoch": 0.85, "learning_rate": 1.1200454864223943e-06, "loss": 0.959, "step": 16780 }, { "epoch": 0.85, "learning_rate": 1.119288958435516e-06, "loss": 0.8334, "step": 16781 }, { "epoch": 0.85, "learning_rate": 1.1185326708844401e-06, "loss": 0.7655, "step": 16782 }, { "epoch": 0.85, "learning_rate": 1.117776623789645e-06, "loss": 0.9129, "step": 16783 }, { "epoch": 0.85, "learning_rate": 1.1170208171715956e-06, "loss": 0.9078, "step": 16784 }, { "epoch": 0.85, "learning_rate": 1.1162652510507577e-06, "loss": 0.8949, "step": 16785 }, { "epoch": 0.85, "learning_rate": 1.1155099254475865e-06, "loss": 0.8815, "step": 16786 }, { "epoch": 0.85, "learning_rate": 1.1147548403825326e-06, "loss": 0.8606, "step": 16787 }, { "epoch": 0.85, "learning_rate": 1.1139999958760372e-06, "loss": 0.8609, "step": 16788 }, { "epoch": 0.85, "learning_rate": 1.1132453919485376e-06, "loss": 0.8379, "step": 16789 }, { "epoch": 0.85, "learning_rate": 1.112491028620467e-06, "loss": 0.8055, "step": 16790 }, { "epoch": 0.85, "learning_rate": 1.1117369059122418e-06, "loss": 0.8331, "step": 16791 }, { "epoch": 0.85, "learning_rate": 1.1109830238442864e-06, "loss": 0.8657, "step": 16792 }, { "epoch": 0.85, "learning_rate": 1.110229382437008e-06, "loss": 0.9274, "step": 16793 }, { "epoch": 0.85, "learning_rate": 1.1094759817108103e-06, "loss": 0.8909, "step": 16794 }, { "epoch": 0.85, "learning_rate": 1.1087228216860935e-06, "loss": 0.8607, "step": 16795 }, { "epoch": 0.85, "learning_rate": 1.1079699023832457e-06, "loss": 0.783, "step": 16796 }, { "epoch": 0.85, "learning_rate": 1.107217223822652e-06, "loss": 0.8226, "step": 16797 }, { "epoch": 0.85, "learning_rate": 1.1064647860246902e-06, "loss": 0.8098, "step": 16798 }, { "epoch": 0.85, "learning_rate": 1.1057125890097342e-06, "loss": 0.8587, "step": 16799 }, { "epoch": 0.85, "learning_rate": 1.1049606327981455e-06, "loss": 0.8294, "step": 16800 }, { "epoch": 0.85, "learning_rate": 1.1042089174102843e-06, "loss": 0.8915, "step": 16801 }, { "epoch": 0.85, "learning_rate": 1.1034574428665046e-06, "loss": 0.7775, "step": 16802 }, { "epoch": 0.85, "learning_rate": 1.1027062091871443e-06, "loss": 0.8209, "step": 16803 }, { "epoch": 0.85, "learning_rate": 1.101955216392553e-06, "loss": 0.8216, "step": 16804 }, { "epoch": 0.85, "learning_rate": 1.101204464503054e-06, "loss": 0.8666, "step": 16805 }, { "epoch": 0.85, "learning_rate": 1.1004539535389768e-06, "loss": 0.7833, "step": 16806 }, { "epoch": 0.85, "learning_rate": 1.0997036835206431e-06, "loss": 0.7592, "step": 16807 }, { "epoch": 0.85, "learning_rate": 1.0989536544683588e-06, "loss": 0.8275, "step": 16808 }, { "epoch": 0.85, "learning_rate": 1.0982038664024374e-06, "loss": 0.863, "step": 16809 }, { "epoch": 0.85, "learning_rate": 1.0974543193431753e-06, "loss": 0.8457, "step": 16810 }, { "epoch": 0.85, "learning_rate": 1.0967050133108648e-06, "loss": 0.902, "step": 16811 }, { "epoch": 0.85, "learning_rate": 1.0959559483257954e-06, "loss": 0.8405, "step": 16812 }, { "epoch": 0.85, "learning_rate": 1.095207124408244e-06, "loss": 0.8586, "step": 16813 }, { "epoch": 0.85, "learning_rate": 1.094458541578486e-06, "loss": 0.9611, "step": 16814 }, { "epoch": 0.85, "learning_rate": 1.0937101998567879e-06, "loss": 0.8295, "step": 16815 }, { "epoch": 0.85, "learning_rate": 1.0929620992634126e-06, "loss": 0.8448, "step": 16816 }, { "epoch": 0.85, "learning_rate": 1.0922142398186097e-06, "loss": 0.8567, "step": 16817 }, { "epoch": 0.85, "learning_rate": 1.0914666215426307e-06, "loss": 0.8561, "step": 16818 }, { "epoch": 0.85, "learning_rate": 1.0907192444557157e-06, "loss": 0.8739, "step": 16819 }, { "epoch": 0.85, "learning_rate": 1.0899721085780945e-06, "loss": 0.7967, "step": 16820 }, { "epoch": 0.85, "learning_rate": 1.089225213930003e-06, "loss": 0.8654, "step": 16821 }, { "epoch": 0.85, "learning_rate": 1.088478560531656e-06, "loss": 0.7926, "step": 16822 }, { "epoch": 0.85, "learning_rate": 1.087732148403272e-06, "loss": 0.821, "step": 16823 }, { "epoch": 0.85, "learning_rate": 1.0869859775650593e-06, "loss": 0.7984, "step": 16824 }, { "epoch": 0.85, "learning_rate": 1.086240048037217e-06, "loss": 0.8747, "step": 16825 }, { "epoch": 0.85, "learning_rate": 1.0854943598399414e-06, "loss": 0.8694, "step": 16826 }, { "epoch": 0.85, "learning_rate": 1.084748912993423e-06, "loss": 0.8913, "step": 16827 }, { "epoch": 0.85, "learning_rate": 1.0840037075178432e-06, "loss": 0.7825, "step": 16828 }, { "epoch": 0.85, "learning_rate": 1.0832587434333764e-06, "loss": 0.8151, "step": 16829 }, { "epoch": 0.86, "learning_rate": 1.0825140207601914e-06, "loss": 0.8673, "step": 16830 }, { "epoch": 0.86, "learning_rate": 1.0817695395184546e-06, "loss": 0.8374, "step": 16831 }, { "epoch": 0.86, "learning_rate": 1.0810252997283144e-06, "loss": 0.8445, "step": 16832 }, { "epoch": 0.86, "learning_rate": 1.08028130140993e-06, "loss": 0.8677, "step": 16833 }, { "epoch": 0.86, "learning_rate": 1.0795375445834365e-06, "loss": 0.9195, "step": 16834 }, { "epoch": 0.86, "learning_rate": 1.0787940292689747e-06, "loss": 0.8955, "step": 16835 }, { "epoch": 0.86, "learning_rate": 1.0780507554866747e-06, "loss": 0.8809, "step": 16836 }, { "epoch": 0.86, "learning_rate": 1.0773077232566554e-06, "loss": 0.7623, "step": 16837 }, { "epoch": 0.86, "learning_rate": 1.0765649325990378e-06, "loss": 0.8617, "step": 16838 }, { "epoch": 0.86, "learning_rate": 1.0758223835339299e-06, "loss": 0.88, "step": 16839 }, { "epoch": 0.86, "learning_rate": 1.0750800760814384e-06, "loss": 0.8096, "step": 16840 }, { "epoch": 0.86, "learning_rate": 1.0743380102616574e-06, "loss": 0.9149, "step": 16841 }, { "epoch": 0.86, "learning_rate": 1.0735961860946774e-06, "loss": 0.8404, "step": 16842 }, { "epoch": 0.86, "learning_rate": 1.0728546036005871e-06, "loss": 0.9811, "step": 16843 }, { "epoch": 0.86, "learning_rate": 1.0721132627994556e-06, "loss": 0.8371, "step": 16844 }, { "epoch": 0.86, "learning_rate": 1.0713721637113638e-06, "loss": 0.8795, "step": 16845 }, { "epoch": 0.86, "learning_rate": 1.070631306356369e-06, "loss": 0.9539, "step": 16846 }, { "epoch": 0.86, "learning_rate": 1.0698906907545304e-06, "loss": 0.8211, "step": 16847 }, { "epoch": 0.86, "learning_rate": 1.0691503169259032e-06, "loss": 0.8534, "step": 16848 }, { "epoch": 0.86, "learning_rate": 1.0684101848905282e-06, "loss": 0.8847, "step": 16849 }, { "epoch": 0.86, "learning_rate": 1.0676702946684447e-06, "loss": 0.8574, "step": 16850 }, { "epoch": 0.86, "learning_rate": 1.0669306462796847e-06, "loss": 0.9257, "step": 16851 }, { "epoch": 0.86, "learning_rate": 1.0661912397442742e-06, "loss": 0.8201, "step": 16852 }, { "epoch": 0.86, "learning_rate": 1.0654520750822305e-06, "loss": 0.8624, "step": 16853 }, { "epoch": 0.86, "learning_rate": 1.0647131523135657e-06, "loss": 0.8345, "step": 16854 }, { "epoch": 0.86, "learning_rate": 1.0639744714582877e-06, "loss": 0.9689, "step": 16855 }, { "epoch": 0.86, "learning_rate": 1.0632360325363911e-06, "loss": 0.838, "step": 16856 }, { "epoch": 0.86, "learning_rate": 1.0624978355678728e-06, "loss": 0.8945, "step": 16857 }, { "epoch": 0.86, "learning_rate": 1.061759880572717e-06, "loss": 0.9294, "step": 16858 }, { "epoch": 0.86, "learning_rate": 1.061022167570901e-06, "loss": 0.9249, "step": 16859 }, { "epoch": 0.86, "learning_rate": 1.060284696582402e-06, "loss": 0.9068, "step": 16860 }, { "epoch": 0.86, "learning_rate": 1.0595474676271823e-06, "loss": 0.9354, "step": 16861 }, { "epoch": 0.86, "learning_rate": 1.0588104807252032e-06, "loss": 0.9276, "step": 16862 }, { "epoch": 0.86, "learning_rate": 1.0580737358964167e-06, "loss": 0.8443, "step": 16863 }, { "epoch": 0.86, "learning_rate": 1.0573372331607735e-06, "loss": 0.822, "step": 16864 }, { "epoch": 0.86, "learning_rate": 1.0566009725382076e-06, "loss": 0.8598, "step": 16865 }, { "epoch": 0.86, "learning_rate": 1.0558649540486553e-06, "loss": 0.7944, "step": 16866 }, { "epoch": 0.86, "learning_rate": 1.0551291777120465e-06, "loss": 0.8837, "step": 16867 }, { "epoch": 0.86, "learning_rate": 1.0543936435482937e-06, "loss": 0.9954, "step": 16868 }, { "epoch": 0.86, "learning_rate": 1.0536583515773192e-06, "loss": 0.8651, "step": 16869 }, { "epoch": 0.86, "learning_rate": 1.0529233018190278e-06, "loss": 0.8071, "step": 16870 }, { "epoch": 0.86, "learning_rate": 1.0521884942933146e-06, "loss": 0.907, "step": 16871 }, { "epoch": 0.86, "learning_rate": 1.051453929020082e-06, "loss": 0.8619, "step": 16872 }, { "epoch": 0.86, "learning_rate": 1.0507196060192104e-06, "loss": 0.9374, "step": 16873 }, { "epoch": 0.86, "learning_rate": 1.0499855253105873e-06, "loss": 0.8705, "step": 16874 }, { "epoch": 0.86, "learning_rate": 1.0492516869140823e-06, "loss": 0.833, "step": 16875 }, { "epoch": 0.86, "learning_rate": 1.0485180908495651e-06, "loss": 0.8702, "step": 16876 }, { "epoch": 0.86, "learning_rate": 1.0477847371369e-06, "loss": 0.7945, "step": 16877 }, { "epoch": 0.86, "learning_rate": 1.0470516257959351e-06, "loss": 0.8672, "step": 16878 }, { "epoch": 0.86, "learning_rate": 1.0463187568465227e-06, "loss": 0.8543, "step": 16879 }, { "epoch": 0.86, "learning_rate": 1.0455861303085047e-06, "loss": 0.921, "step": 16880 }, { "epoch": 0.86, "learning_rate": 1.0448537462017172e-06, "loss": 0.7962, "step": 16881 }, { "epoch": 0.86, "learning_rate": 1.0441216045459846e-06, "loss": 0.9429, "step": 16882 }, { "epoch": 0.86, "learning_rate": 1.043389705361132e-06, "loss": 0.8466, "step": 16883 }, { "epoch": 0.86, "learning_rate": 1.0426580486669746e-06, "loss": 0.8578, "step": 16884 }, { "epoch": 0.86, "learning_rate": 1.0419266344833178e-06, "loss": 0.8803, "step": 16885 }, { "epoch": 0.86, "learning_rate": 1.0411954628299702e-06, "loss": 0.9066, "step": 16886 }, { "epoch": 0.86, "learning_rate": 1.0404645337267227e-06, "loss": 0.846, "step": 16887 }, { "epoch": 0.86, "learning_rate": 1.039733847193366e-06, "loss": 0.9004, "step": 16888 }, { "epoch": 0.86, "learning_rate": 1.0390034032496832e-06, "loss": 0.7952, "step": 16889 }, { "epoch": 0.86, "learning_rate": 1.0382732019154484e-06, "loss": 0.8441, "step": 16890 }, { "epoch": 0.86, "learning_rate": 1.0375432432104314e-06, "loss": 0.8183, "step": 16891 }, { "epoch": 0.86, "learning_rate": 1.0368135271543965e-06, "loss": 0.9233, "step": 16892 }, { "epoch": 0.86, "learning_rate": 1.0360840537671014e-06, "loss": 0.8353, "step": 16893 }, { "epoch": 0.86, "learning_rate": 1.0353548230682907e-06, "loss": 0.893, "step": 16894 }, { "epoch": 0.86, "learning_rate": 1.0346258350777117e-06, "loss": 0.842, "step": 16895 }, { "epoch": 0.86, "learning_rate": 1.0338970898151014e-06, "loss": 0.9081, "step": 16896 }, { "epoch": 0.86, "learning_rate": 1.033168587300184e-06, "loss": 0.7938, "step": 16897 }, { "epoch": 0.86, "learning_rate": 1.0324403275526906e-06, "loss": 0.8743, "step": 16898 }, { "epoch": 0.86, "learning_rate": 1.0317123105923333e-06, "loss": 0.8579, "step": 16899 }, { "epoch": 0.86, "learning_rate": 1.0309845364388238e-06, "loss": 0.9038, "step": 16900 }, { "epoch": 0.86, "learning_rate": 1.0302570051118678e-06, "loss": 0.9081, "step": 16901 }, { "epoch": 0.86, "learning_rate": 1.029529716631158e-06, "loss": 0.8693, "step": 16902 }, { "epoch": 0.86, "learning_rate": 1.0288026710163868e-06, "loss": 0.804, "step": 16903 }, { "epoch": 0.86, "learning_rate": 1.0280758682872393e-06, "loss": 0.9052, "step": 16904 }, { "epoch": 0.86, "learning_rate": 1.0273493084633933e-06, "loss": 0.9058, "step": 16905 }, { "epoch": 0.86, "learning_rate": 1.0266229915645175e-06, "loss": 0.8468, "step": 16906 }, { "epoch": 0.86, "learning_rate": 1.0258969176102785e-06, "loss": 0.9182, "step": 16907 }, { "epoch": 0.86, "learning_rate": 1.0251710866203323e-06, "loss": 0.918, "step": 16908 }, { "epoch": 0.86, "learning_rate": 1.0244454986143282e-06, "loss": 0.8899, "step": 16909 }, { "epoch": 0.86, "learning_rate": 1.0237201536119167e-06, "loss": 0.9135, "step": 16910 }, { "epoch": 0.86, "learning_rate": 1.0229950516327314e-06, "loss": 0.8059, "step": 16911 }, { "epoch": 0.86, "learning_rate": 1.0222701926964006e-06, "loss": 0.8376, "step": 16912 }, { "epoch": 0.86, "learning_rate": 1.021545576822557e-06, "loss": 0.8021, "step": 16913 }, { "epoch": 0.86, "learning_rate": 1.020821204030813e-06, "loss": 0.7717, "step": 16914 }, { "epoch": 0.86, "learning_rate": 1.0200970743407812e-06, "loss": 0.9666, "step": 16915 }, { "epoch": 0.86, "learning_rate": 1.019373187772067e-06, "loss": 0.9323, "step": 16916 }, { "epoch": 0.86, "learning_rate": 1.0186495443442713e-06, "loss": 0.8805, "step": 16917 }, { "epoch": 0.86, "learning_rate": 1.0179261440769817e-06, "loss": 0.793, "step": 16918 }, { "epoch": 0.86, "learning_rate": 1.017202986989786e-06, "loss": 0.7917, "step": 16919 }, { "epoch": 0.86, "learning_rate": 1.016480073102264e-06, "loss": 0.8725, "step": 16920 }, { "epoch": 0.86, "learning_rate": 1.0157574024339834e-06, "loss": 0.8512, "step": 16921 }, { "epoch": 0.86, "learning_rate": 1.0150349750045153e-06, "loss": 0.8648, "step": 16922 }, { "epoch": 0.86, "learning_rate": 1.014312790833417e-06, "loss": 0.8767, "step": 16923 }, { "epoch": 0.86, "learning_rate": 1.0135908499402348e-06, "loss": 0.8791, "step": 16924 }, { "epoch": 0.86, "learning_rate": 1.0128691523445245e-06, "loss": 0.8364, "step": 16925 }, { "epoch": 0.86, "learning_rate": 1.0121476980658175e-06, "loss": 0.9042, "step": 16926 }, { "epoch": 0.86, "learning_rate": 1.01142648712365e-06, "loss": 0.8464, "step": 16927 }, { "epoch": 0.86, "learning_rate": 1.010705519537547e-06, "loss": 0.8838, "step": 16928 }, { "epoch": 0.86, "learning_rate": 1.0099847953270304e-06, "loss": 0.8177, "step": 16929 }, { "epoch": 0.86, "learning_rate": 1.0092643145116087e-06, "loss": 0.8999, "step": 16930 }, { "epoch": 0.86, "learning_rate": 1.00854407711079e-06, "loss": 0.8424, "step": 16931 }, { "epoch": 0.86, "learning_rate": 1.0078240831440766e-06, "loss": 0.8401, "step": 16932 }, { "epoch": 0.86, "learning_rate": 1.007104332630955e-06, "loss": 0.8521, "step": 16933 }, { "epoch": 0.86, "learning_rate": 1.0063848255909204e-06, "loss": 0.8341, "step": 16934 }, { "epoch": 0.86, "learning_rate": 1.0056655620434475e-06, "loss": 0.8632, "step": 16935 }, { "epoch": 0.86, "learning_rate": 1.0049465420080074e-06, "loss": 0.8209, "step": 16936 }, { "epoch": 0.86, "learning_rate": 1.0042277655040721e-06, "loss": 0.9331, "step": 16937 }, { "epoch": 0.86, "learning_rate": 1.0035092325510987e-06, "loss": 0.8682, "step": 16938 }, { "epoch": 0.86, "learning_rate": 1.00279094316854e-06, "loss": 0.9747, "step": 16939 }, { "epoch": 0.86, "learning_rate": 1.0020728973758475e-06, "loss": 0.9302, "step": 16940 }, { "epoch": 0.86, "learning_rate": 1.0013550951924555e-06, "loss": 0.8562, "step": 16941 }, { "epoch": 0.86, "learning_rate": 1.0006375366378008e-06, "loss": 0.8659, "step": 16942 }, { "epoch": 0.86, "learning_rate": 9.999202217313097e-07, "loss": 0.8488, "step": 16943 }, { "epoch": 0.86, "learning_rate": 9.992031504924039e-07, "loss": 0.8733, "step": 16944 }, { "epoch": 0.86, "learning_rate": 9.984863229404973e-07, "loss": 0.9111, "step": 16945 }, { "epoch": 0.86, "learning_rate": 9.977697390949969e-07, "loss": 0.8338, "step": 16946 }, { "epoch": 0.86, "learning_rate": 9.970533989753028e-07, "loss": 0.8975, "step": 16947 }, { "epoch": 0.86, "learning_rate": 9.963373026008084e-07, "loss": 0.8077, "step": 16948 }, { "epoch": 0.86, "learning_rate": 9.956214499909055e-07, "loss": 0.8797, "step": 16949 }, { "epoch": 0.86, "learning_rate": 9.94905841164967e-07, "loss": 0.8222, "step": 16950 }, { "epoch": 0.86, "learning_rate": 9.94190476142377e-07, "loss": 0.8638, "step": 16951 }, { "epoch": 0.86, "learning_rate": 9.93475354942497e-07, "loss": 0.8591, "step": 16952 }, { "epoch": 0.86, "learning_rate": 9.927604775846867e-07, "loss": 0.8657, "step": 16953 }, { "epoch": 0.86, "learning_rate": 9.920458440883075e-07, "loss": 0.8943, "step": 16954 }, { "epoch": 0.86, "learning_rate": 9.913314544727015e-07, "loss": 0.74, "step": 16955 }, { "epoch": 0.86, "learning_rate": 9.906173087572112e-07, "loss": 0.8438, "step": 16956 }, { "epoch": 0.86, "learning_rate": 9.899034069611724e-07, "loss": 0.8786, "step": 16957 }, { "epoch": 0.86, "learning_rate": 9.891897491039148e-07, "loss": 0.9102, "step": 16958 }, { "epoch": 0.86, "learning_rate": 9.884763352047566e-07, "loss": 0.8518, "step": 16959 }, { "epoch": 0.86, "learning_rate": 9.877631652830134e-07, "loss": 0.8298, "step": 16960 }, { "epoch": 0.86, "learning_rate": 9.870502393579962e-07, "loss": 0.8588, "step": 16961 }, { "epoch": 0.86, "learning_rate": 9.86337557449002e-07, "loss": 0.8889, "step": 16962 }, { "epoch": 0.86, "learning_rate": 9.856251195753307e-07, "loss": 0.911, "step": 16963 }, { "epoch": 0.86, "learning_rate": 9.849129257562707e-07, "loss": 0.8797, "step": 16964 }, { "epoch": 0.86, "learning_rate": 9.84200976011096e-07, "loss": 0.8027, "step": 16965 }, { "epoch": 0.86, "learning_rate": 9.834892703590927e-07, "loss": 0.8559, "step": 16966 }, { "epoch": 0.86, "learning_rate": 9.827778088195227e-07, "loss": 0.825, "step": 16967 }, { "epoch": 0.86, "learning_rate": 9.8206659141165e-07, "loss": 0.875, "step": 16968 }, { "epoch": 0.86, "learning_rate": 9.813556181547324e-07, "loss": 0.8659, "step": 16969 }, { "epoch": 0.86, "learning_rate": 9.806448890680132e-07, "loss": 0.8725, "step": 16970 }, { "epoch": 0.86, "learning_rate": 9.799344041707392e-07, "loss": 0.8523, "step": 16971 }, { "epoch": 0.86, "learning_rate": 9.792241634821442e-07, "loss": 0.9444, "step": 16972 }, { "epoch": 0.86, "learning_rate": 9.785141670214582e-07, "loss": 0.8942, "step": 16973 }, { "epoch": 0.86, "learning_rate": 9.778044148079001e-07, "loss": 0.8211, "step": 16974 }, { "epoch": 0.86, "learning_rate": 9.770949068606927e-07, "loss": 0.8054, "step": 16975 }, { "epoch": 0.86, "learning_rate": 9.763856431990415e-07, "loss": 0.8453, "step": 16976 }, { "epoch": 0.86, "learning_rate": 9.756766238421446e-07, "loss": 0.7874, "step": 16977 }, { "epoch": 0.86, "learning_rate": 9.749678488092051e-07, "loss": 0.8104, "step": 16978 }, { "epoch": 0.86, "learning_rate": 9.742593181194083e-07, "loss": 0.9232, "step": 16979 }, { "epoch": 0.86, "learning_rate": 9.73551031791936e-07, "loss": 0.833, "step": 16980 }, { "epoch": 0.86, "learning_rate": 9.728429898459701e-07, "loss": 0.9083, "step": 16981 }, { "epoch": 0.86, "learning_rate": 9.721351923006727e-07, "loss": 0.8683, "step": 16982 }, { "epoch": 0.86, "learning_rate": 9.714276391752108e-07, "loss": 0.7531, "step": 16983 }, { "epoch": 0.86, "learning_rate": 9.707203304887403e-07, "loss": 0.7946, "step": 16984 }, { "epoch": 0.86, "learning_rate": 9.700132662604123e-07, "loss": 0.8625, "step": 16985 }, { "epoch": 0.86, "learning_rate": 9.693064465093638e-07, "loss": 0.8508, "step": 16986 }, { "epoch": 0.86, "learning_rate": 9.685998712547384e-07, "loss": 0.9113, "step": 16987 }, { "epoch": 0.86, "learning_rate": 9.678935405156643e-07, "loss": 0.8775, "step": 16988 }, { "epoch": 0.86, "learning_rate": 9.67187454311258e-07, "loss": 0.8651, "step": 16989 }, { "epoch": 0.86, "learning_rate": 9.664816126606459e-07, "loss": 0.8985, "step": 16990 }, { "epoch": 0.86, "learning_rate": 9.657760155829309e-07, "loss": 0.9024, "step": 16991 }, { "epoch": 0.86, "learning_rate": 9.650706630972185e-07, "loss": 0.8138, "step": 16992 }, { "epoch": 0.86, "learning_rate": 9.643655552226084e-07, "loss": 0.8492, "step": 16993 }, { "epoch": 0.86, "learning_rate": 9.636606919781844e-07, "loss": 0.8468, "step": 16994 }, { "epoch": 0.86, "learning_rate": 9.629560733830346e-07, "loss": 0.9821, "step": 16995 }, { "epoch": 0.86, "learning_rate": 9.622516994562337e-07, "loss": 0.9014, "step": 16996 }, { "epoch": 0.86, "learning_rate": 9.615475702168542e-07, "loss": 0.9005, "step": 16997 }, { "epoch": 0.86, "learning_rate": 9.608436856839553e-07, "loss": 0.8707, "step": 16998 }, { "epoch": 0.86, "learning_rate": 9.601400458765997e-07, "loss": 0.764, "step": 16999 }, { "epoch": 0.86, "learning_rate": 9.594366508138352e-07, "loss": 0.826, "step": 17000 }, { "epoch": 0.86, "learning_rate": 9.587335005147015e-07, "loss": 0.9233, "step": 17001 }, { "epoch": 0.86, "learning_rate": 9.580305949982417e-07, "loss": 0.8787, "step": 17002 }, { "epoch": 0.86, "learning_rate": 9.57327934283483e-07, "loss": 0.9504, "step": 17003 }, { "epoch": 0.86, "learning_rate": 9.566255183894491e-07, "loss": 0.9297, "step": 17004 }, { "epoch": 0.86, "learning_rate": 9.559233473351592e-07, "loss": 0.8181, "step": 17005 }, { "epoch": 0.86, "learning_rate": 9.552214211396216e-07, "loss": 0.8065, "step": 17006 }, { "epoch": 0.86, "learning_rate": 9.54519739821841e-07, "loss": 0.9096, "step": 17007 }, { "epoch": 0.86, "learning_rate": 9.538183034008152e-07, "loss": 0.8876, "step": 17008 }, { "epoch": 0.86, "learning_rate": 9.531171118955352e-07, "loss": 0.8578, "step": 17009 }, { "epoch": 0.86, "learning_rate": 9.52416165324983e-07, "loss": 0.8763, "step": 17010 }, { "epoch": 0.86, "learning_rate": 9.517154637081371e-07, "loss": 0.8565, "step": 17011 }, { "epoch": 0.86, "learning_rate": 9.51015007063969e-07, "loss": 0.9486, "step": 17012 }, { "epoch": 0.86, "learning_rate": 9.503147954114423e-07, "loss": 0.7791, "step": 17013 }, { "epoch": 0.86, "learning_rate": 9.49614828769515e-07, "loss": 0.8212, "step": 17014 }, { "epoch": 0.86, "learning_rate": 9.489151071571346e-07, "loss": 0.8723, "step": 17015 }, { "epoch": 0.86, "learning_rate": 9.482156305932522e-07, "loss": 0.8534, "step": 17016 }, { "epoch": 0.86, "learning_rate": 9.475163990968006e-07, "loss": 0.8508, "step": 17017 }, { "epoch": 0.86, "learning_rate": 9.46817412686708e-07, "loss": 0.8548, "step": 17018 }, { "epoch": 0.86, "learning_rate": 9.461186713819048e-07, "loss": 0.8124, "step": 17019 }, { "epoch": 0.86, "learning_rate": 9.45420175201306e-07, "loss": 0.8166, "step": 17020 }, { "epoch": 0.86, "learning_rate": 9.447219241638206e-07, "loss": 0.8014, "step": 17021 }, { "epoch": 0.86, "learning_rate": 9.440239182883581e-07, "loss": 0.8661, "step": 17022 }, { "epoch": 0.86, "learning_rate": 9.433261575938113e-07, "loss": 0.8169, "step": 17023 }, { "epoch": 0.86, "learning_rate": 9.426286420990727e-07, "loss": 0.9229, "step": 17024 }, { "epoch": 0.86, "learning_rate": 9.419313718230283e-07, "loss": 0.929, "step": 17025 }, { "epoch": 0.86, "learning_rate": 9.412343467845564e-07, "loss": 0.932, "step": 17026 }, { "epoch": 0.87, "learning_rate": 9.405375670025229e-07, "loss": 0.8381, "step": 17027 }, { "epoch": 0.87, "learning_rate": 9.398410324957996e-07, "loss": 0.8078, "step": 17028 }, { "epoch": 0.87, "learning_rate": 9.391447432832423e-07, "loss": 0.8693, "step": 17029 }, { "epoch": 0.87, "learning_rate": 9.384486993836961e-07, "loss": 0.7986, "step": 17030 }, { "epoch": 0.87, "learning_rate": 9.377529008160147e-07, "loss": 0.8179, "step": 17031 }, { "epoch": 0.87, "learning_rate": 9.370573475990308e-07, "loss": 0.8954, "step": 17032 }, { "epoch": 0.87, "learning_rate": 9.36362039751576e-07, "loss": 0.7947, "step": 17033 }, { "epoch": 0.87, "learning_rate": 9.356669772924787e-07, "loss": 0.9207, "step": 17034 }, { "epoch": 0.87, "learning_rate": 9.349721602405526e-07, "loss": 0.9648, "step": 17035 }, { "epoch": 0.87, "learning_rate": 9.342775886146093e-07, "loss": 0.8282, "step": 17036 }, { "epoch": 0.87, "learning_rate": 9.335832624334574e-07, "loss": 0.884, "step": 17037 }, { "epoch": 0.87, "learning_rate": 9.328891817158925e-07, "loss": 0.9599, "step": 17038 }, { "epoch": 0.87, "learning_rate": 9.321953464807054e-07, "loss": 0.8488, "step": 17039 }, { "epoch": 0.87, "learning_rate": 9.315017567466822e-07, "loss": 0.8553, "step": 17040 }, { "epoch": 0.87, "learning_rate": 9.308084125326022e-07, "loss": 0.892, "step": 17041 }, { "epoch": 0.87, "learning_rate": 9.301153138572328e-07, "loss": 0.8571, "step": 17042 }, { "epoch": 0.87, "learning_rate": 9.294224607393442e-07, "loss": 0.8868, "step": 17043 }, { "epoch": 0.87, "learning_rate": 9.287298531976907e-07, "loss": 0.808, "step": 17044 }, { "epoch": 0.87, "learning_rate": 9.280374912510248e-07, "loss": 0.8748, "step": 17045 }, { "epoch": 0.87, "learning_rate": 9.273453749180949e-07, "loss": 0.8208, "step": 17046 }, { "epoch": 0.87, "learning_rate": 9.266535042176338e-07, "loss": 0.8832, "step": 17047 }, { "epoch": 0.87, "learning_rate": 9.259618791683767e-07, "loss": 0.8651, "step": 17048 }, { "epoch": 0.87, "learning_rate": 9.252704997890472e-07, "loss": 0.8427, "step": 17049 }, { "epoch": 0.87, "learning_rate": 9.24579366098366e-07, "loss": 0.8777, "step": 17050 }, { "epoch": 0.87, "learning_rate": 9.238884781150404e-07, "loss": 0.856, "step": 17051 }, { "epoch": 0.87, "learning_rate": 9.231978358577787e-07, "loss": 0.9001, "step": 17052 }, { "epoch": 0.87, "learning_rate": 9.225074393452804e-07, "loss": 0.781, "step": 17053 }, { "epoch": 0.87, "learning_rate": 9.218172885962329e-07, "loss": 0.8521, "step": 17054 }, { "epoch": 0.87, "learning_rate": 9.211273836293255e-07, "loss": 0.9204, "step": 17055 }, { "epoch": 0.87, "learning_rate": 9.204377244632345e-07, "loss": 0.8745, "step": 17056 }, { "epoch": 0.87, "learning_rate": 9.197483111166328e-07, "loss": 0.7826, "step": 17057 }, { "epoch": 0.87, "learning_rate": 9.190591436081853e-07, "loss": 0.8914, "step": 17058 }, { "epoch": 0.87, "learning_rate": 9.183702219565494e-07, "loss": 0.8428, "step": 17059 }, { "epoch": 0.87, "learning_rate": 9.176815461803767e-07, "loss": 0.8863, "step": 17060 }, { "epoch": 0.87, "learning_rate": 9.169931162983137e-07, "loss": 0.8246, "step": 17061 }, { "epoch": 0.87, "learning_rate": 9.163049323290007e-07, "loss": 0.864, "step": 17062 }, { "epoch": 0.87, "learning_rate": 9.156169942910653e-07, "loss": 0.831, "step": 17063 }, { "epoch": 0.87, "learning_rate": 9.149293022031347e-07, "loss": 0.7562, "step": 17064 }, { "epoch": 0.87, "learning_rate": 9.142418560838295e-07, "loss": 0.8045, "step": 17065 }, { "epoch": 0.87, "learning_rate": 9.13554655951756e-07, "loss": 0.9109, "step": 17066 }, { "epoch": 0.87, "learning_rate": 9.128677018255261e-07, "loss": 0.7951, "step": 17067 }, { "epoch": 0.87, "learning_rate": 9.121809937237335e-07, "loss": 0.8996, "step": 17068 }, { "epoch": 0.87, "learning_rate": 9.114945316649715e-07, "loss": 0.8342, "step": 17069 }, { "epoch": 0.87, "learning_rate": 9.108083156678282e-07, "loss": 0.8766, "step": 17070 }, { "epoch": 0.87, "learning_rate": 9.101223457508779e-07, "loss": 0.835, "step": 17071 }, { "epoch": 0.87, "learning_rate": 9.094366219326934e-07, "loss": 0.7772, "step": 17072 }, { "epoch": 0.87, "learning_rate": 9.087511442318419e-07, "loss": 0.8239, "step": 17073 }, { "epoch": 0.87, "learning_rate": 9.080659126668822e-07, "loss": 0.8333, "step": 17074 }, { "epoch": 0.87, "learning_rate": 9.073809272563627e-07, "loss": 0.8458, "step": 17075 }, { "epoch": 0.87, "learning_rate": 9.066961880188308e-07, "loss": 0.8259, "step": 17076 }, { "epoch": 0.87, "learning_rate": 9.06011694972827e-07, "loss": 0.8716, "step": 17077 }, { "epoch": 0.87, "learning_rate": 9.053274481368779e-07, "loss": 0.9424, "step": 17078 }, { "epoch": 0.87, "learning_rate": 9.046434475295152e-07, "loss": 0.8773, "step": 17079 }, { "epoch": 0.87, "learning_rate": 9.039596931692529e-07, "loss": 0.9045, "step": 17080 }, { "epoch": 0.87, "learning_rate": 9.032761850746053e-07, "loss": 0.9183, "step": 17081 }, { "epoch": 0.87, "learning_rate": 9.025929232640784e-07, "loss": 0.9218, "step": 17082 }, { "epoch": 0.87, "learning_rate": 9.019099077561644e-07, "loss": 0.9377, "step": 17083 }, { "epoch": 0.87, "learning_rate": 9.01227138569365e-07, "loss": 0.8495, "step": 17084 }, { "epoch": 0.87, "learning_rate": 9.005446157221576e-07, "loss": 0.8292, "step": 17085 }, { "epoch": 0.87, "learning_rate": 8.998623392330241e-07, "loss": 0.8395, "step": 17086 }, { "epoch": 0.87, "learning_rate": 8.991803091204387e-07, "loss": 0.8781, "step": 17087 }, { "epoch": 0.87, "learning_rate": 8.98498525402861e-07, "loss": 0.8359, "step": 17088 }, { "epoch": 0.87, "learning_rate": 8.97816988098753e-07, "loss": 0.8632, "step": 17089 }, { "epoch": 0.87, "learning_rate": 8.971356972265655e-07, "loss": 0.938, "step": 17090 }, { "epoch": 0.87, "learning_rate": 8.964546528047469e-07, "loss": 0.8903, "step": 17091 }, { "epoch": 0.87, "learning_rate": 8.957738548517314e-07, "loss": 0.9131, "step": 17092 }, { "epoch": 0.87, "learning_rate": 8.950933033859521e-07, "loss": 0.765, "step": 17093 }, { "epoch": 0.87, "learning_rate": 8.944129984258365e-07, "loss": 0.7547, "step": 17094 }, { "epoch": 0.87, "learning_rate": 8.937329399897976e-07, "loss": 0.8404, "step": 17095 }, { "epoch": 0.87, "learning_rate": 8.930531280962551e-07, "loss": 0.8567, "step": 17096 }, { "epoch": 0.87, "learning_rate": 8.923735627636087e-07, "loss": 0.8257, "step": 17097 }, { "epoch": 0.87, "learning_rate": 8.916942440102583e-07, "loss": 0.8762, "step": 17098 }, { "epoch": 0.87, "learning_rate": 8.910151718545967e-07, "loss": 0.9061, "step": 17099 }, { "epoch": 0.87, "learning_rate": 8.903363463150072e-07, "loss": 0.7874, "step": 17100 }, { "epoch": 0.87, "learning_rate": 8.896577674098683e-07, "loss": 0.8889, "step": 17101 }, { "epoch": 0.87, "learning_rate": 8.889794351575531e-07, "loss": 0.8508, "step": 17102 }, { "epoch": 0.87, "learning_rate": 8.883013495764281e-07, "loss": 0.9131, "step": 17103 }, { "epoch": 0.87, "learning_rate": 8.876235106848485e-07, "loss": 0.943, "step": 17104 }, { "epoch": 0.87, "learning_rate": 8.869459185011675e-07, "loss": 0.8711, "step": 17105 }, { "epoch": 0.87, "learning_rate": 8.862685730437315e-07, "loss": 0.7961, "step": 17106 }, { "epoch": 0.87, "learning_rate": 8.855914743308747e-07, "loss": 0.8803, "step": 17107 }, { "epoch": 0.87, "learning_rate": 8.849146223809346e-07, "loss": 0.8593, "step": 17108 }, { "epoch": 0.87, "learning_rate": 8.842380172122322e-07, "loss": 0.8892, "step": 17109 }, { "epoch": 0.87, "learning_rate": 8.835616588430873e-07, "loss": 0.8605, "step": 17110 }, { "epoch": 0.87, "learning_rate": 8.828855472918129e-07, "loss": 0.8239, "step": 17111 }, { "epoch": 0.87, "learning_rate": 8.822096825767102e-07, "loss": 0.8116, "step": 17112 }, { "epoch": 0.87, "learning_rate": 8.815340647160809e-07, "loss": 0.9237, "step": 17113 }, { "epoch": 0.87, "learning_rate": 8.808586937282149e-07, "loss": 0.8186, "step": 17114 }, { "epoch": 0.87, "learning_rate": 8.801835696313988e-07, "loss": 0.8179, "step": 17115 }, { "epoch": 0.87, "learning_rate": 8.795086924439089e-07, "loss": 0.8926, "step": 17116 }, { "epoch": 0.87, "learning_rate": 8.788340621840174e-07, "loss": 0.884, "step": 17117 }, { "epoch": 0.87, "learning_rate": 8.78159678869992e-07, "loss": 0.8091, "step": 17118 }, { "epoch": 0.87, "learning_rate": 8.774855425200835e-07, "loss": 0.9193, "step": 17119 }, { "epoch": 0.87, "learning_rate": 8.768116531525528e-07, "loss": 0.8468, "step": 17120 }, { "epoch": 0.87, "learning_rate": 8.761380107856387e-07, "loss": 0.9066, "step": 17121 }, { "epoch": 0.87, "learning_rate": 8.754646154375801e-07, "loss": 0.8631, "step": 17122 }, { "epoch": 0.87, "learning_rate": 8.747914671266122e-07, "loss": 0.9068, "step": 17123 }, { "epoch": 0.87, "learning_rate": 8.741185658709539e-07, "loss": 0.7836, "step": 17124 }, { "epoch": 0.87, "learning_rate": 8.734459116888272e-07, "loss": 0.918, "step": 17125 }, { "epoch": 0.87, "learning_rate": 8.72773504598442e-07, "loss": 0.9499, "step": 17126 }, { "epoch": 0.87, "learning_rate": 8.721013446180049e-07, "loss": 0.9267, "step": 17127 }, { "epoch": 0.87, "learning_rate": 8.714294317657123e-07, "loss": 0.8236, "step": 17128 }, { "epoch": 0.87, "learning_rate": 8.707577660597544e-07, "loss": 0.9428, "step": 17129 }, { "epoch": 0.87, "learning_rate": 8.700863475183196e-07, "loss": 0.8502, "step": 17130 }, { "epoch": 0.87, "learning_rate": 8.694151761595793e-07, "loss": 0.8583, "step": 17131 }, { "epoch": 0.87, "learning_rate": 8.687442520017131e-07, "loss": 0.7975, "step": 17132 }, { "epoch": 0.87, "learning_rate": 8.680735750628799e-07, "loss": 0.8852, "step": 17133 }, { "epoch": 0.87, "learning_rate": 8.674031453612386e-07, "loss": 0.865, "step": 17134 }, { "epoch": 0.87, "learning_rate": 8.667329629149424e-07, "loss": 0.7804, "step": 17135 }, { "epoch": 0.87, "learning_rate": 8.660630277421334e-07, "loss": 0.7834, "step": 17136 }, { "epoch": 0.87, "learning_rate": 8.653933398609493e-07, "loss": 0.9109, "step": 17137 }, { "epoch": 0.87, "learning_rate": 8.647238992895224e-07, "loss": 0.9089, "step": 17138 }, { "epoch": 0.87, "learning_rate": 8.64054706045978e-07, "loss": 0.818, "step": 17139 }, { "epoch": 0.87, "learning_rate": 8.633857601484307e-07, "loss": 0.8355, "step": 17140 }, { "epoch": 0.87, "learning_rate": 8.627170616149938e-07, "loss": 0.9001, "step": 17141 }, { "epoch": 0.87, "learning_rate": 8.620486104637737e-07, "loss": 0.8705, "step": 17142 }, { "epoch": 0.87, "learning_rate": 8.613804067128606e-07, "loss": 0.8501, "step": 17143 }, { "epoch": 0.87, "learning_rate": 8.607124503803544e-07, "loss": 0.8084, "step": 17144 }, { "epoch": 0.87, "learning_rate": 8.600447414843338e-07, "loss": 0.8618, "step": 17145 }, { "epoch": 0.87, "learning_rate": 8.593772800428767e-07, "loss": 0.8873, "step": 17146 }, { "epoch": 0.87, "learning_rate": 8.587100660740577e-07, "loss": 0.7547, "step": 17147 }, { "epoch": 0.87, "learning_rate": 8.580430995959355e-07, "loss": 0.9171, "step": 17148 }, { "epoch": 0.87, "learning_rate": 8.5737638062657e-07, "loss": 0.8777, "step": 17149 }, { "epoch": 0.87, "learning_rate": 8.567099091840125e-07, "loss": 0.8252, "step": 17150 }, { "epoch": 0.87, "learning_rate": 8.560436852863064e-07, "loss": 0.7654, "step": 17151 }, { "epoch": 0.87, "learning_rate": 8.553777089514914e-07, "loss": 0.8552, "step": 17152 }, { "epoch": 0.87, "learning_rate": 8.547119801975934e-07, "loss": 0.7952, "step": 17153 }, { "epoch": 0.87, "learning_rate": 8.540464990426389e-07, "loss": 0.9241, "step": 17154 }, { "epoch": 0.87, "learning_rate": 8.533812655046447e-07, "loss": 0.8427, "step": 17155 }, { "epoch": 0.87, "learning_rate": 8.52716279601623e-07, "loss": 0.8324, "step": 17156 }, { "epoch": 0.87, "learning_rate": 8.52051541351575e-07, "loss": 0.8679, "step": 17157 }, { "epoch": 0.87, "learning_rate": 8.513870507724986e-07, "loss": 0.8622, "step": 17158 }, { "epoch": 0.87, "learning_rate": 8.50722807882387e-07, "loss": 0.8482, "step": 17159 }, { "epoch": 0.87, "learning_rate": 8.500588126992171e-07, "loss": 0.9276, "step": 17160 }, { "epoch": 0.87, "learning_rate": 8.493950652409732e-07, "loss": 0.9196, "step": 17161 }, { "epoch": 0.87, "learning_rate": 8.487315655256212e-07, "loss": 0.8775, "step": 17162 }, { "epoch": 0.87, "learning_rate": 8.480683135711254e-07, "loss": 0.9101, "step": 17163 }, { "epoch": 0.87, "learning_rate": 8.474053093954449e-07, "loss": 0.8745, "step": 17164 }, { "epoch": 0.87, "learning_rate": 8.467425530165263e-07, "loss": 0.8827, "step": 17165 }, { "epoch": 0.87, "learning_rate": 8.460800444523153e-07, "loss": 0.9245, "step": 17166 }, { "epoch": 0.87, "learning_rate": 8.454177837207478e-07, "loss": 0.8427, "step": 17167 }, { "epoch": 0.87, "learning_rate": 8.447557708397558e-07, "loss": 0.8802, "step": 17168 }, { "epoch": 0.87, "learning_rate": 8.440940058272596e-07, "loss": 0.8718, "step": 17169 }, { "epoch": 0.87, "learning_rate": 8.434324887011769e-07, "loss": 0.7925, "step": 17170 }, { "epoch": 0.87, "learning_rate": 8.427712194794191e-07, "loss": 0.8799, "step": 17171 }, { "epoch": 0.87, "learning_rate": 8.421101981798841e-07, "loss": 0.797, "step": 17172 }, { "epoch": 0.87, "learning_rate": 8.414494248204763e-07, "loss": 0.9125, "step": 17173 }, { "epoch": 0.87, "learning_rate": 8.407888994190805e-07, "loss": 0.9144, "step": 17174 }, { "epoch": 0.87, "learning_rate": 8.401286219935789e-07, "loss": 0.9077, "step": 17175 }, { "epoch": 0.87, "learning_rate": 8.394685925618529e-07, "loss": 0.8753, "step": 17176 }, { "epoch": 0.87, "learning_rate": 8.388088111417658e-07, "loss": 0.9069, "step": 17177 }, { "epoch": 0.87, "learning_rate": 8.381492777511846e-07, "loss": 0.8881, "step": 17178 }, { "epoch": 0.87, "learning_rate": 8.374899924079627e-07, "loss": 0.8546, "step": 17179 }, { "epoch": 0.87, "learning_rate": 8.368309551299536e-07, "loss": 0.8585, "step": 17180 }, { "epoch": 0.87, "learning_rate": 8.361721659349964e-07, "loss": 0.9578, "step": 17181 }, { "epoch": 0.87, "learning_rate": 8.355136248409279e-07, "loss": 0.8264, "step": 17182 }, { "epoch": 0.87, "learning_rate": 8.348553318655795e-07, "loss": 0.8546, "step": 17183 }, { "epoch": 0.87, "learning_rate": 8.341972870267678e-07, "loss": 0.9056, "step": 17184 }, { "epoch": 0.87, "learning_rate": 8.335394903423167e-07, "loss": 0.9248, "step": 17185 }, { "epoch": 0.87, "learning_rate": 8.328819418300305e-07, "loss": 0.9056, "step": 17186 }, { "epoch": 0.87, "learning_rate": 8.322246415077117e-07, "loss": 0.8533, "step": 17187 }, { "epoch": 0.87, "learning_rate": 8.315675893931596e-07, "loss": 0.8332, "step": 17188 }, { "epoch": 0.87, "learning_rate": 8.309107855041587e-07, "loss": 0.8931, "step": 17189 }, { "epoch": 0.87, "learning_rate": 8.302542298584926e-07, "loss": 1.0267, "step": 17190 }, { "epoch": 0.87, "learning_rate": 8.295979224739381e-07, "loss": 0.7949, "step": 17191 }, { "epoch": 0.87, "learning_rate": 8.289418633682645e-07, "loss": 0.9008, "step": 17192 }, { "epoch": 0.87, "learning_rate": 8.282860525592318e-07, "loss": 0.8501, "step": 17193 }, { "epoch": 0.87, "learning_rate": 8.276304900645948e-07, "loss": 0.8894, "step": 17194 }, { "epoch": 0.87, "learning_rate": 8.269751759021072e-07, "loss": 0.8773, "step": 17195 }, { "epoch": 0.87, "learning_rate": 8.263201100895024e-07, "loss": 0.8629, "step": 17196 }, { "epoch": 0.87, "learning_rate": 8.256652926445252e-07, "loss": 0.7987, "step": 17197 }, { "epoch": 0.87, "learning_rate": 8.25010723584897e-07, "loss": 0.9204, "step": 17198 }, { "epoch": 0.87, "learning_rate": 8.243564029283424e-07, "loss": 0.8696, "step": 17199 }, { "epoch": 0.87, "learning_rate": 8.237023306925774e-07, "loss": 0.798, "step": 17200 }, { "epoch": 0.87, "learning_rate": 8.230485068953076e-07, "loss": 0.9073, "step": 17201 }, { "epoch": 0.87, "learning_rate": 8.223949315542357e-07, "loss": 0.877, "step": 17202 }, { "epoch": 0.87, "learning_rate": 8.217416046870563e-07, "loss": 0.8978, "step": 17203 }, { "epoch": 0.87, "learning_rate": 8.21088526311461e-07, "loss": 0.9493, "step": 17204 }, { "epoch": 0.87, "learning_rate": 8.204356964451255e-07, "loss": 0.8537, "step": 17205 }, { "epoch": 0.87, "learning_rate": 8.197831151057278e-07, "loss": 0.7611, "step": 17206 }, { "epoch": 0.87, "learning_rate": 8.191307823109373e-07, "loss": 0.9106, "step": 17207 }, { "epoch": 0.87, "learning_rate": 8.184786980784088e-07, "loss": 0.8714, "step": 17208 }, { "epoch": 0.87, "learning_rate": 8.178268624258056e-07, "loss": 0.7878, "step": 17209 }, { "epoch": 0.87, "learning_rate": 8.171752753707684e-07, "loss": 0.9002, "step": 17210 }, { "epoch": 0.87, "learning_rate": 8.165239369309419e-07, "loss": 0.923, "step": 17211 }, { "epoch": 0.87, "learning_rate": 8.158728471239608e-07, "loss": 0.8043, "step": 17212 }, { "epoch": 0.87, "learning_rate": 8.152220059674487e-07, "loss": 0.8077, "step": 17213 }, { "epoch": 0.87, "learning_rate": 8.145714134790295e-07, "loss": 0.9027, "step": 17214 }, { "epoch": 0.87, "learning_rate": 8.139210696763167e-07, "loss": 0.8384, "step": 17215 }, { "epoch": 0.87, "learning_rate": 8.132709745769196e-07, "loss": 0.8013, "step": 17216 }, { "epoch": 0.87, "learning_rate": 8.126211281984353e-07, "loss": 0.8132, "step": 17217 }, { "epoch": 0.87, "learning_rate": 8.119715305584585e-07, "loss": 0.8004, "step": 17218 }, { "epoch": 0.87, "learning_rate": 8.113221816745798e-07, "loss": 0.815, "step": 17219 }, { "epoch": 0.87, "learning_rate": 8.106730815643737e-07, "loss": 0.8347, "step": 17220 }, { "epoch": 0.87, "learning_rate": 8.100242302454209e-07, "loss": 0.8289, "step": 17221 }, { "epoch": 0.87, "learning_rate": 8.093756277352816e-07, "loss": 0.7554, "step": 17222 }, { "epoch": 0.87, "learning_rate": 8.087272740515206e-07, "loss": 0.8052, "step": 17223 }, { "epoch": 0.88, "learning_rate": 8.080791692116908e-07, "loss": 0.7845, "step": 17224 }, { "epoch": 0.88, "learning_rate": 8.074313132333345e-07, "loss": 0.8528, "step": 17225 }, { "epoch": 0.88, "learning_rate": 8.067837061339989e-07, "loss": 0.8588, "step": 17226 }, { "epoch": 0.88, "learning_rate": 8.061363479312123e-07, "loss": 0.9073, "step": 17227 }, { "epoch": 0.88, "learning_rate": 8.054892386425028e-07, "loss": 0.847, "step": 17228 }, { "epoch": 0.88, "learning_rate": 8.048423782853909e-07, "loss": 0.8679, "step": 17229 }, { "epoch": 0.88, "learning_rate": 8.04195766877387e-07, "loss": 0.9555, "step": 17230 }, { "epoch": 0.88, "learning_rate": 8.035494044360003e-07, "loss": 0.8744, "step": 17231 }, { "epoch": 0.88, "learning_rate": 8.029032909787282e-07, "loss": 0.8091, "step": 17232 }, { "epoch": 0.88, "learning_rate": 8.022574265230665e-07, "loss": 0.8459, "step": 17233 }, { "epoch": 0.88, "learning_rate": 8.016118110864979e-07, "loss": 0.8105, "step": 17234 }, { "epoch": 0.88, "learning_rate": 8.009664446865029e-07, "loss": 0.869, "step": 17235 }, { "epoch": 0.88, "learning_rate": 8.003213273405563e-07, "loss": 0.8477, "step": 17236 }, { "epoch": 0.88, "learning_rate": 7.996764590661177e-07, "loss": 0.9498, "step": 17237 }, { "epoch": 0.88, "learning_rate": 7.990318398806552e-07, "loss": 0.8243, "step": 17238 }, { "epoch": 0.88, "learning_rate": 7.983874698016147e-07, "loss": 0.8477, "step": 17239 }, { "epoch": 0.88, "learning_rate": 7.977433488464436e-07, "loss": 0.9019, "step": 17240 }, { "epoch": 0.88, "learning_rate": 7.970994770325824e-07, "loss": 0.9172, "step": 17241 }, { "epoch": 0.88, "learning_rate": 7.964558543774603e-07, "loss": 0.8138, "step": 17242 }, { "epoch": 0.88, "learning_rate": 7.958124808985035e-07, "loss": 0.9379, "step": 17243 }, { "epoch": 0.88, "learning_rate": 7.951693566131325e-07, "loss": 0.8808, "step": 17244 }, { "epoch": 0.88, "learning_rate": 7.945264815387588e-07, "loss": 0.8322, "step": 17245 }, { "epoch": 0.88, "learning_rate": 7.938838556927853e-07, "loss": 0.8635, "step": 17246 }, { "epoch": 0.88, "learning_rate": 7.932414790926113e-07, "loss": 0.833, "step": 17247 }, { "epoch": 0.88, "learning_rate": 7.925993517556319e-07, "loss": 0.8835, "step": 17248 }, { "epoch": 0.88, "learning_rate": 7.919574736992241e-07, "loss": 0.8799, "step": 17249 }, { "epoch": 0.88, "learning_rate": 7.913158449407766e-07, "loss": 0.8133, "step": 17250 }, { "epoch": 0.88, "learning_rate": 7.906744654976517e-07, "loss": 0.9092, "step": 17251 }, { "epoch": 0.88, "learning_rate": 7.900333353872192e-07, "loss": 0.8365, "step": 17252 }, { "epoch": 0.88, "learning_rate": 7.893924546268361e-07, "loss": 0.86, "step": 17253 }, { "epoch": 0.88, "learning_rate": 7.88751823233852e-07, "loss": 0.8565, "step": 17254 }, { "epoch": 0.88, "learning_rate": 7.881114412256119e-07, "loss": 0.9125, "step": 17255 }, { "epoch": 0.88, "learning_rate": 7.874713086194541e-07, "loss": 0.8584, "step": 17256 }, { "epoch": 0.88, "learning_rate": 7.868314254327103e-07, "loss": 0.8873, "step": 17257 }, { "epoch": 0.88, "learning_rate": 7.861917916827022e-07, "loss": 0.8983, "step": 17258 }, { "epoch": 0.88, "learning_rate": 7.855524073867493e-07, "loss": 0.9022, "step": 17259 }, { "epoch": 0.88, "learning_rate": 7.849132725621633e-07, "loss": 0.8263, "step": 17260 }, { "epoch": 0.88, "learning_rate": 7.842743872262426e-07, "loss": 0.851, "step": 17261 }, { "epoch": 0.88, "learning_rate": 7.836357513962911e-07, "loss": 0.8645, "step": 17262 }, { "epoch": 0.88, "learning_rate": 7.82997365089595e-07, "loss": 0.8435, "step": 17263 }, { "epoch": 0.88, "learning_rate": 7.823592283234383e-07, "loss": 0.9534, "step": 17264 }, { "epoch": 0.88, "learning_rate": 7.817213411151015e-07, "loss": 0.8323, "step": 17265 }, { "epoch": 0.88, "learning_rate": 7.810837034818497e-07, "loss": 0.848, "step": 17266 }, { "epoch": 0.88, "learning_rate": 7.80446315440948e-07, "loss": 0.7949, "step": 17267 }, { "epoch": 0.88, "learning_rate": 7.798091770096538e-07, "loss": 0.897, "step": 17268 }, { "epoch": 0.88, "learning_rate": 7.791722882052189e-07, "loss": 0.925, "step": 17269 }, { "epoch": 0.88, "learning_rate": 7.785356490448814e-07, "loss": 0.8358, "step": 17270 }, { "epoch": 0.88, "learning_rate": 7.778992595458811e-07, "loss": 0.9649, "step": 17271 }, { "epoch": 0.88, "learning_rate": 7.772631197254488e-07, "loss": 0.8645, "step": 17272 }, { "epoch": 0.88, "learning_rate": 7.766272296008015e-07, "loss": 0.8472, "step": 17273 }, { "epoch": 0.88, "learning_rate": 7.759915891891623e-07, "loss": 0.7827, "step": 17274 }, { "epoch": 0.88, "learning_rate": 7.753561985077351e-07, "loss": 0.8906, "step": 17275 }, { "epoch": 0.88, "learning_rate": 7.747210575737252e-07, "loss": 0.8794, "step": 17276 }, { "epoch": 0.88, "learning_rate": 7.740861664043297e-07, "loss": 0.9282, "step": 17277 }, { "epoch": 0.88, "learning_rate": 7.73451525016734e-07, "loss": 0.8979, "step": 17278 }, { "epoch": 0.88, "learning_rate": 7.728171334281209e-07, "loss": 0.9032, "step": 17279 }, { "epoch": 0.88, "learning_rate": 7.721829916556678e-07, "loss": 0.8726, "step": 17280 }, { "epoch": 0.88, "learning_rate": 7.715490997165442e-07, "loss": 0.8792, "step": 17281 }, { "epoch": 0.88, "learning_rate": 7.709154576279088e-07, "loss": 0.9662, "step": 17282 }, { "epoch": 0.88, "learning_rate": 7.702820654069188e-07, "loss": 0.8771, "step": 17283 }, { "epoch": 0.88, "learning_rate": 7.696489230707227e-07, "loss": 0.851, "step": 17284 }, { "epoch": 0.88, "learning_rate": 7.690160306364592e-07, "loss": 0.7971, "step": 17285 }, { "epoch": 0.88, "learning_rate": 7.683833881212688e-07, "loss": 0.8879, "step": 17286 }, { "epoch": 0.88, "learning_rate": 7.677509955422758e-07, "loss": 0.8262, "step": 17287 }, { "epoch": 0.88, "learning_rate": 7.671188529166007e-07, "loss": 0.9266, "step": 17288 }, { "epoch": 0.88, "learning_rate": 7.664869602613623e-07, "loss": 0.7917, "step": 17289 }, { "epoch": 0.88, "learning_rate": 7.658553175936623e-07, "loss": 0.8856, "step": 17290 }, { "epoch": 0.88, "learning_rate": 7.652239249306082e-07, "loss": 0.9061, "step": 17291 }, { "epoch": 0.88, "learning_rate": 7.645927822892895e-07, "loss": 0.7874, "step": 17292 }, { "epoch": 0.88, "learning_rate": 7.639618896867962e-07, "loss": 0.8553, "step": 17293 }, { "epoch": 0.88, "learning_rate": 7.633312471402087e-07, "loss": 0.8889, "step": 17294 }, { "epoch": 0.88, "learning_rate": 7.627008546666004e-07, "loss": 0.8373, "step": 17295 }, { "epoch": 0.88, "learning_rate": 7.620707122830384e-07, "loss": 0.8542, "step": 17296 }, { "epoch": 0.88, "learning_rate": 7.614408200065838e-07, "loss": 0.8727, "step": 17297 }, { "epoch": 0.88, "learning_rate": 7.608111778542904e-07, "loss": 0.8285, "step": 17298 }, { "epoch": 0.88, "learning_rate": 7.601817858432037e-07, "loss": 0.8176, "step": 17299 }, { "epoch": 0.88, "learning_rate": 7.595526439903656e-07, "loss": 0.8984, "step": 17300 }, { "epoch": 0.88, "learning_rate": 7.58923752312809e-07, "loss": 0.8455, "step": 17301 }, { "epoch": 0.88, "learning_rate": 7.582951108275572e-07, "loss": 0.8108, "step": 17302 }, { "epoch": 0.88, "learning_rate": 7.576667195516363e-07, "loss": 0.7928, "step": 17303 }, { "epoch": 0.88, "learning_rate": 7.57038578502054e-07, "loss": 0.8768, "step": 17304 }, { "epoch": 0.88, "learning_rate": 7.564106876958188e-07, "loss": 0.8273, "step": 17305 }, { "epoch": 0.88, "learning_rate": 7.557830471499306e-07, "loss": 0.75, "step": 17306 }, { "epoch": 0.88, "learning_rate": 7.551556568813812e-07, "loss": 0.846, "step": 17307 }, { "epoch": 0.88, "learning_rate": 7.545285169071547e-07, "loss": 0.8246, "step": 17308 }, { "epoch": 0.88, "learning_rate": 7.539016272442334e-07, "loss": 0.8921, "step": 17309 }, { "epoch": 0.88, "learning_rate": 7.532749879095902e-07, "loss": 0.8385, "step": 17310 }, { "epoch": 0.88, "learning_rate": 7.526485989201871e-07, "loss": 0.9156, "step": 17311 }, { "epoch": 0.88, "learning_rate": 7.520224602929849e-07, "loss": 0.8294, "step": 17312 }, { "epoch": 0.88, "learning_rate": 7.513965720449368e-07, "loss": 0.8339, "step": 17313 }, { "epoch": 0.88, "learning_rate": 7.507709341929836e-07, "loss": 0.8867, "step": 17314 }, { "epoch": 0.88, "learning_rate": 7.501455467540708e-07, "loss": 0.881, "step": 17315 }, { "epoch": 0.88, "learning_rate": 7.495204097451247e-07, "loss": 0.9019, "step": 17316 }, { "epoch": 0.88, "learning_rate": 7.488955231830708e-07, "loss": 0.9271, "step": 17317 }, { "epoch": 0.88, "learning_rate": 7.482708870848299e-07, "loss": 0.8191, "step": 17318 }, { "epoch": 0.88, "learning_rate": 7.476465014673107e-07, "loss": 0.9173, "step": 17319 }, { "epoch": 0.88, "learning_rate": 7.470223663474185e-07, "loss": 0.8784, "step": 17320 }, { "epoch": 0.88, "learning_rate": 7.463984817420511e-07, "loss": 0.8718, "step": 17321 }, { "epoch": 0.88, "learning_rate": 7.457748476681015e-07, "loss": 0.7719, "step": 17322 }, { "epoch": 0.88, "learning_rate": 7.451514641424506e-07, "loss": 0.8204, "step": 17323 }, { "epoch": 0.88, "learning_rate": 7.445283311819785e-07, "loss": 0.8023, "step": 17324 }, { "epoch": 0.88, "learning_rate": 7.439054488035547e-07, "loss": 0.7925, "step": 17325 }, { "epoch": 0.88, "learning_rate": 7.432828170240402e-07, "loss": 0.8487, "step": 17326 }, { "epoch": 0.88, "learning_rate": 7.426604358602995e-07, "loss": 0.8733, "step": 17327 }, { "epoch": 0.88, "learning_rate": 7.420383053291757e-07, "loss": 0.8347, "step": 17328 }, { "epoch": 0.88, "learning_rate": 7.414164254475153e-07, "loss": 0.8355, "step": 17329 }, { "epoch": 0.88, "learning_rate": 7.407947962321572e-07, "loss": 0.863, "step": 17330 }, { "epoch": 0.88, "learning_rate": 7.401734176999265e-07, "loss": 0.8951, "step": 17331 }, { "epoch": 0.88, "learning_rate": 7.395522898676488e-07, "loss": 0.798, "step": 17332 }, { "epoch": 0.88, "learning_rate": 7.389314127521407e-07, "loss": 0.9663, "step": 17333 }, { "epoch": 0.88, "learning_rate": 7.383107863702133e-07, "loss": 0.8191, "step": 17334 }, { "epoch": 0.88, "learning_rate": 7.376904107386662e-07, "loss": 0.854, "step": 17335 }, { "epoch": 0.88, "learning_rate": 7.370702858742962e-07, "loss": 0.9041, "step": 17336 }, { "epoch": 0.88, "learning_rate": 7.364504117938964e-07, "loss": 0.8514, "step": 17337 }, { "epoch": 0.88, "learning_rate": 7.358307885142413e-07, "loss": 0.8418, "step": 17338 }, { "epoch": 0.88, "learning_rate": 7.352114160521151e-07, "loss": 0.9638, "step": 17339 }, { "epoch": 0.88, "learning_rate": 7.345922944242812e-07, "loss": 0.7926, "step": 17340 }, { "epoch": 0.88, "learning_rate": 7.339734236475038e-07, "loss": 0.8892, "step": 17341 }, { "epoch": 0.88, "learning_rate": 7.333548037385385e-07, "loss": 0.8306, "step": 17342 }, { "epoch": 0.88, "learning_rate": 7.327364347141319e-07, "loss": 0.7502, "step": 17343 }, { "epoch": 0.88, "learning_rate": 7.321183165910261e-07, "loss": 0.8578, "step": 17344 }, { "epoch": 0.88, "learning_rate": 7.315004493859567e-07, "loss": 0.874, "step": 17345 }, { "epoch": 0.88, "learning_rate": 7.308828331156537e-07, "loss": 0.8063, "step": 17346 }, { "epoch": 0.88, "learning_rate": 7.302654677968346e-07, "loss": 0.9365, "step": 17347 }, { "epoch": 0.88, "learning_rate": 7.296483534462162e-07, "loss": 0.9528, "step": 17348 }, { "epoch": 0.88, "learning_rate": 7.290314900805073e-07, "loss": 0.9102, "step": 17349 }, { "epoch": 0.88, "learning_rate": 7.284148777164035e-07, "loss": 0.8456, "step": 17350 }, { "epoch": 0.88, "learning_rate": 7.277985163706069e-07, "loss": 0.9071, "step": 17351 }, { "epoch": 0.88, "learning_rate": 7.271824060597976e-07, "loss": 0.8596, "step": 17352 }, { "epoch": 0.88, "learning_rate": 7.2656654680066e-07, "loss": 0.8893, "step": 17353 }, { "epoch": 0.88, "learning_rate": 7.259509386098684e-07, "loss": 0.8321, "step": 17354 }, { "epoch": 0.88, "learning_rate": 7.253355815040875e-07, "loss": 0.9417, "step": 17355 }, { "epoch": 0.88, "learning_rate": 7.24720475499977e-07, "loss": 0.8165, "step": 17356 }, { "epoch": 0.88, "learning_rate": 7.241056206141927e-07, "loss": 0.794, "step": 17357 }, { "epoch": 0.88, "learning_rate": 7.234910168633802e-07, "loss": 0.8453, "step": 17358 }, { "epoch": 0.88, "learning_rate": 7.228766642641782e-07, "loss": 0.825, "step": 17359 }, { "epoch": 0.88, "learning_rate": 7.222625628332203e-07, "loss": 0.8629, "step": 17360 }, { "epoch": 0.88, "learning_rate": 7.21648712587133e-07, "loss": 0.9156, "step": 17361 }, { "epoch": 0.88, "learning_rate": 7.210351135425353e-07, "loss": 0.7962, "step": 17362 }, { "epoch": 0.88, "learning_rate": 7.204217657160417e-07, "loss": 0.8467, "step": 17363 }, { "epoch": 0.88, "learning_rate": 7.198086691242545e-07, "loss": 0.9377, "step": 17364 }, { "epoch": 0.88, "learning_rate": 7.191958237837737e-07, "loss": 0.8332, "step": 17365 }, { "epoch": 0.88, "learning_rate": 7.185832297111939e-07, "loss": 0.8868, "step": 17366 }, { "epoch": 0.88, "learning_rate": 7.17970886923095e-07, "loss": 0.8544, "step": 17367 }, { "epoch": 0.88, "learning_rate": 7.173587954360628e-07, "loss": 0.9122, "step": 17368 }, { "epoch": 0.88, "learning_rate": 7.167469552666628e-07, "loss": 0.7513, "step": 17369 }, { "epoch": 0.88, "learning_rate": 7.16135366431463e-07, "loss": 0.9263, "step": 17370 }, { "epoch": 0.88, "learning_rate": 7.155240289470222e-07, "loss": 0.7892, "step": 17371 }, { "epoch": 0.88, "learning_rate": 7.149129428298895e-07, "loss": 0.9122, "step": 17372 }, { "epoch": 0.88, "learning_rate": 7.143021080966095e-07, "loss": 0.8726, "step": 17373 }, { "epoch": 0.88, "learning_rate": 7.13691524763721e-07, "loss": 0.9102, "step": 17374 }, { "epoch": 0.88, "learning_rate": 7.130811928477566e-07, "loss": 0.8758, "step": 17375 }, { "epoch": 0.88, "learning_rate": 7.124711123652361e-07, "loss": 0.9487, "step": 17376 }, { "epoch": 0.88, "learning_rate": 7.118612833326799e-07, "loss": 0.8674, "step": 17377 }, { "epoch": 0.88, "learning_rate": 7.11251705766599e-07, "loss": 0.8418, "step": 17378 }, { "epoch": 0.88, "learning_rate": 7.106423796834915e-07, "loss": 0.9269, "step": 17379 }, { "epoch": 0.88, "learning_rate": 7.100333050998632e-07, "loss": 0.9318, "step": 17380 }, { "epoch": 0.88, "learning_rate": 7.094244820321971e-07, "loss": 0.8442, "step": 17381 }, { "epoch": 0.88, "learning_rate": 7.088159104969783e-07, "loss": 0.9277, "step": 17382 }, { "epoch": 0.88, "learning_rate": 7.082075905106855e-07, "loss": 0.9216, "step": 17383 }, { "epoch": 0.88, "learning_rate": 7.075995220897858e-07, "loss": 0.9201, "step": 17384 }, { "epoch": 0.88, "learning_rate": 7.069917052507424e-07, "loss": 0.9462, "step": 17385 }, { "epoch": 0.88, "learning_rate": 7.063841400100102e-07, "loss": 0.8192, "step": 17386 }, { "epoch": 0.88, "learning_rate": 7.057768263840425e-07, "loss": 0.8741, "step": 17387 }, { "epoch": 0.88, "learning_rate": 7.051697643892774e-07, "loss": 0.9254, "step": 17388 }, { "epoch": 0.88, "learning_rate": 7.045629540421528e-07, "loss": 0.8963, "step": 17389 }, { "epoch": 0.88, "learning_rate": 7.039563953590966e-07, "loss": 0.8482, "step": 17390 }, { "epoch": 0.88, "learning_rate": 7.03350088356528e-07, "loss": 0.8662, "step": 17391 }, { "epoch": 0.88, "learning_rate": 7.027440330508694e-07, "loss": 0.8184, "step": 17392 }, { "epoch": 0.88, "learning_rate": 7.021382294585211e-07, "loss": 0.8466, "step": 17393 }, { "epoch": 0.88, "learning_rate": 7.015326775958897e-07, "loss": 0.7818, "step": 17394 }, { "epoch": 0.88, "learning_rate": 7.00927377479369e-07, "loss": 0.752, "step": 17395 }, { "epoch": 0.88, "learning_rate": 7.003223291253447e-07, "loss": 0.9178, "step": 17396 }, { "epoch": 0.88, "learning_rate": 6.997175325501993e-07, "loss": 0.8235, "step": 17397 }, { "epoch": 0.88, "learning_rate": 6.991129877703073e-07, "loss": 0.8794, "step": 17398 }, { "epoch": 0.88, "learning_rate": 6.98508694802037e-07, "loss": 0.9243, "step": 17399 }, { "epoch": 0.88, "learning_rate": 6.979046536617463e-07, "loss": 0.8441, "step": 17400 }, { "epoch": 0.88, "learning_rate": 6.973008643657897e-07, "loss": 0.8654, "step": 17401 }, { "epoch": 0.88, "learning_rate": 6.966973269305178e-07, "loss": 0.8015, "step": 17402 }, { "epoch": 0.88, "learning_rate": 6.960940413722639e-07, "loss": 0.8018, "step": 17403 }, { "epoch": 0.88, "learning_rate": 6.954910077073685e-07, "loss": 0.9581, "step": 17404 }, { "epoch": 0.88, "learning_rate": 6.948882259521528e-07, "loss": 0.9124, "step": 17405 }, { "epoch": 0.88, "learning_rate": 6.942856961229394e-07, "loss": 0.8393, "step": 17406 }, { "epoch": 0.88, "learning_rate": 6.936834182360408e-07, "loss": 0.854, "step": 17407 }, { "epoch": 0.88, "learning_rate": 6.930813923077617e-07, "loss": 0.7679, "step": 17408 }, { "epoch": 0.88, "learning_rate": 6.924796183544025e-07, "loss": 0.8008, "step": 17409 }, { "epoch": 0.88, "learning_rate": 6.918780963922544e-07, "loss": 0.8409, "step": 17410 }, { "epoch": 0.88, "learning_rate": 6.912768264376058e-07, "loss": 1.0114, "step": 17411 }, { "epoch": 0.88, "learning_rate": 6.906758085067322e-07, "loss": 0.9319, "step": 17412 }, { "epoch": 0.88, "learning_rate": 6.900750426159064e-07, "loss": 0.8791, "step": 17413 }, { "epoch": 0.88, "learning_rate": 6.894745287813954e-07, "loss": 0.8513, "step": 17414 }, { "epoch": 0.88, "learning_rate": 6.888742670194526e-07, "loss": 0.9568, "step": 17415 }, { "epoch": 0.88, "learning_rate": 6.882742573463352e-07, "loss": 0.8087, "step": 17416 }, { "epoch": 0.88, "learning_rate": 6.876744997782847e-07, "loss": 0.8913, "step": 17417 }, { "epoch": 0.88, "learning_rate": 6.870749943315402e-07, "loss": 0.8663, "step": 17418 }, { "epoch": 0.88, "learning_rate": 6.864757410223321e-07, "loss": 0.8982, "step": 17419 }, { "epoch": 0.88, "learning_rate": 6.858767398668842e-07, "loss": 0.7545, "step": 17420 }, { "epoch": 0.89, "learning_rate": 6.852779908814133e-07, "loss": 0.9033, "step": 17421 }, { "epoch": 0.89, "learning_rate": 6.846794940821311e-07, "loss": 0.863, "step": 17422 }, { "epoch": 0.89, "learning_rate": 6.840812494852422e-07, "loss": 0.8974, "step": 17423 }, { "epoch": 0.89, "learning_rate": 6.834832571069417e-07, "loss": 0.8248, "step": 17424 }, { "epoch": 0.89, "learning_rate": 6.828855169634185e-07, "loss": 0.8131, "step": 17425 }, { "epoch": 0.89, "learning_rate": 6.8228802907086e-07, "loss": 0.8371, "step": 17426 }, { "epoch": 0.89, "learning_rate": 6.816907934454353e-07, "loss": 0.8703, "step": 17427 }, { "epoch": 0.89, "learning_rate": 6.810938101033227e-07, "loss": 0.8903, "step": 17428 }, { "epoch": 0.89, "learning_rate": 6.804970790606791e-07, "loss": 0.9042, "step": 17429 }, { "epoch": 0.89, "learning_rate": 6.799006003336616e-07, "loss": 0.9321, "step": 17430 }, { "epoch": 0.89, "learning_rate": 6.793043739384209e-07, "loss": 0.824, "step": 17431 }, { "epoch": 0.89, "learning_rate": 6.787083998910948e-07, "loss": 0.8573, "step": 17432 }, { "epoch": 0.89, "learning_rate": 6.781126782078251e-07, "loss": 0.8538, "step": 17433 }, { "epoch": 0.89, "learning_rate": 6.775172089047355e-07, "loss": 0.9365, "step": 17434 }, { "epoch": 0.89, "learning_rate": 6.769219919979486e-07, "loss": 0.8564, "step": 17435 }, { "epoch": 0.89, "learning_rate": 6.763270275035827e-07, "loss": 0.8909, "step": 17436 }, { "epoch": 0.89, "learning_rate": 6.757323154377393e-07, "loss": 0.828, "step": 17437 }, { "epoch": 0.89, "learning_rate": 6.751378558165245e-07, "loss": 0.8505, "step": 17438 }, { "epoch": 0.89, "learning_rate": 6.74543648656032e-07, "loss": 0.8083, "step": 17439 }, { "epoch": 0.89, "learning_rate": 6.739496939723489e-07, "loss": 0.9192, "step": 17440 }, { "epoch": 0.89, "learning_rate": 6.733559917815547e-07, "loss": 0.8152, "step": 17441 }, { "epoch": 0.89, "learning_rate": 6.727625420997242e-07, "loss": 0.9037, "step": 17442 }, { "epoch": 0.89, "learning_rate": 6.721693449429256e-07, "loss": 0.8534, "step": 17443 }, { "epoch": 0.89, "learning_rate": 6.715764003272152e-07, "loss": 0.8798, "step": 17444 }, { "epoch": 0.89, "learning_rate": 6.709837082686522e-07, "loss": 0.7984, "step": 17445 }, { "epoch": 0.89, "learning_rate": 6.703912687832792e-07, "loss": 0.9598, "step": 17446 }, { "epoch": 0.89, "learning_rate": 6.697990818871358e-07, "loss": 0.8392, "step": 17447 }, { "epoch": 0.89, "learning_rate": 6.69207147596257e-07, "loss": 0.8964, "step": 17448 }, { "epoch": 0.89, "learning_rate": 6.686154659266664e-07, "loss": 0.8861, "step": 17449 }, { "epoch": 0.89, "learning_rate": 6.680240368943847e-07, "loss": 0.8978, "step": 17450 }, { "epoch": 0.89, "learning_rate": 6.674328605154234e-07, "loss": 0.7997, "step": 17451 }, { "epoch": 0.89, "learning_rate": 6.6684193680579e-07, "loss": 0.8328, "step": 17452 }, { "epoch": 0.89, "learning_rate": 6.662512657814791e-07, "loss": 0.8687, "step": 17453 }, { "epoch": 0.89, "learning_rate": 6.656608474584858e-07, "loss": 0.7246, "step": 17454 }, { "epoch": 0.89, "learning_rate": 6.650706818527953e-07, "loss": 0.7675, "step": 17455 }, { "epoch": 0.89, "learning_rate": 6.644807689803812e-07, "loss": 0.818, "step": 17456 }, { "epoch": 0.89, "learning_rate": 6.638911088572209e-07, "loss": 0.7941, "step": 17457 }, { "epoch": 0.89, "learning_rate": 6.633017014992749e-07, "loss": 0.8918, "step": 17458 }, { "epoch": 0.89, "learning_rate": 6.627125469225026e-07, "loss": 0.8138, "step": 17459 }, { "epoch": 0.89, "learning_rate": 6.621236451428547e-07, "loss": 0.7514, "step": 17460 }, { "epoch": 0.89, "learning_rate": 6.615349961762729e-07, "loss": 0.9427, "step": 17461 }, { "epoch": 0.89, "learning_rate": 6.609466000386955e-07, "loss": 0.8044, "step": 17462 }, { "epoch": 0.89, "learning_rate": 6.60358456746053e-07, "loss": 0.7936, "step": 17463 }, { "epoch": 0.89, "learning_rate": 6.597705663142706e-07, "loss": 0.8096, "step": 17464 }, { "epoch": 0.89, "learning_rate": 6.591829287592621e-07, "loss": 0.8273, "step": 17465 }, { "epoch": 0.89, "learning_rate": 6.58595544096936e-07, "loss": 0.8393, "step": 17466 }, { "epoch": 0.89, "learning_rate": 6.580084123432007e-07, "loss": 0.8361, "step": 17467 }, { "epoch": 0.89, "learning_rate": 6.574215335139445e-07, "loss": 0.8764, "step": 17468 }, { "epoch": 0.89, "learning_rate": 6.568349076250635e-07, "loss": 0.8155, "step": 17469 }, { "epoch": 0.89, "learning_rate": 6.562485346924352e-07, "loss": 0.8039, "step": 17470 }, { "epoch": 0.89, "learning_rate": 6.556624147319368e-07, "loss": 0.8896, "step": 17471 }, { "epoch": 0.89, "learning_rate": 6.55076547759439e-07, "loss": 0.7837, "step": 17472 }, { "epoch": 0.89, "learning_rate": 6.544909337908001e-07, "loss": 1.0091, "step": 17473 }, { "epoch": 0.89, "learning_rate": 6.539055728418764e-07, "loss": 0.9215, "step": 17474 }, { "epoch": 0.89, "learning_rate": 6.533204649285152e-07, "loss": 0.8927, "step": 17475 }, { "epoch": 0.89, "learning_rate": 6.527356100665605e-07, "loss": 0.8279, "step": 17476 }, { "epoch": 0.89, "learning_rate": 6.521510082718418e-07, "loss": 0.808, "step": 17477 }, { "epoch": 0.89, "learning_rate": 6.515666595601899e-07, "loss": 0.8329, "step": 17478 }, { "epoch": 0.89, "learning_rate": 6.509825639474265e-07, "loss": 0.9197, "step": 17479 }, { "epoch": 0.89, "learning_rate": 6.503987214493612e-07, "loss": 0.8196, "step": 17480 }, { "epoch": 0.89, "learning_rate": 6.498151320818047e-07, "loss": 0.8439, "step": 17481 }, { "epoch": 0.89, "learning_rate": 6.492317958605554e-07, "loss": 0.8166, "step": 17482 }, { "epoch": 0.89, "learning_rate": 6.486487128014063e-07, "loss": 0.8204, "step": 17483 }, { "epoch": 0.89, "learning_rate": 6.480658829201459e-07, "loss": 0.9258, "step": 17484 }, { "epoch": 0.89, "learning_rate": 6.474833062325514e-07, "loss": 0.9131, "step": 17485 }, { "epoch": 0.89, "learning_rate": 6.469009827543948e-07, "loss": 0.8872, "step": 17486 }, { "epoch": 0.89, "learning_rate": 6.463189125014436e-07, "loss": 0.8236, "step": 17487 }, { "epoch": 0.89, "learning_rate": 6.457370954894582e-07, "loss": 0.8382, "step": 17488 }, { "epoch": 0.89, "learning_rate": 6.451555317341862e-07, "loss": 0.7982, "step": 17489 }, { "epoch": 0.89, "learning_rate": 6.445742212513761e-07, "loss": 0.8701, "step": 17490 }, { "epoch": 0.89, "learning_rate": 6.439931640567676e-07, "loss": 0.8624, "step": 17491 }, { "epoch": 0.89, "learning_rate": 6.434123601660869e-07, "loss": 0.9054, "step": 17492 }, { "epoch": 0.89, "learning_rate": 6.428318095950648e-07, "loss": 0.8217, "step": 17493 }, { "epoch": 0.89, "learning_rate": 6.422515123594141e-07, "loss": 0.7964, "step": 17494 }, { "epoch": 0.89, "learning_rate": 6.416714684748481e-07, "loss": 0.9116, "step": 17495 }, { "epoch": 0.89, "learning_rate": 6.410916779570731e-07, "loss": 0.7868, "step": 17496 }, { "epoch": 0.89, "learning_rate": 6.405121408217795e-07, "loss": 0.8828, "step": 17497 }, { "epoch": 0.89, "learning_rate": 6.399328570846653e-07, "loss": 0.7701, "step": 17498 }, { "epoch": 0.89, "learning_rate": 6.393538267614096e-07, "loss": 0.8819, "step": 17499 }, { "epoch": 0.89, "learning_rate": 6.387750498676903e-07, "loss": 0.8072, "step": 17500 }, { "epoch": 0.89, "learning_rate": 6.381965264191781e-07, "loss": 0.8438, "step": 17501 }, { "epoch": 0.89, "learning_rate": 6.376182564315326e-07, "loss": 0.7403, "step": 17502 }, { "epoch": 0.89, "learning_rate": 6.370402399204134e-07, "loss": 0.8819, "step": 17503 }, { "epoch": 0.89, "learning_rate": 6.364624769014671e-07, "loss": 0.8757, "step": 17504 }, { "epoch": 0.89, "learning_rate": 6.358849673903388e-07, "loss": 0.8297, "step": 17505 }, { "epoch": 0.89, "learning_rate": 6.353077114026618e-07, "loss": 0.9146, "step": 17506 }, { "epoch": 0.89, "learning_rate": 6.347307089540644e-07, "loss": 0.8422, "step": 17507 }, { "epoch": 0.89, "learning_rate": 6.34153960060171e-07, "loss": 0.8704, "step": 17508 }, { "epoch": 0.89, "learning_rate": 6.335774647365922e-07, "loss": 0.9431, "step": 17509 }, { "epoch": 0.89, "learning_rate": 6.330012229989413e-07, "loss": 0.8102, "step": 17510 }, { "epoch": 0.89, "learning_rate": 6.324252348628146e-07, "loss": 0.815, "step": 17511 }, { "epoch": 0.89, "learning_rate": 6.318495003438086e-07, "loss": 0.8438, "step": 17512 }, { "epoch": 0.89, "learning_rate": 6.312740194575118e-07, "loss": 0.9142, "step": 17513 }, { "epoch": 0.89, "learning_rate": 6.306987922195018e-07, "loss": 0.8498, "step": 17514 }, { "epoch": 0.89, "learning_rate": 6.30123818645354e-07, "loss": 0.8477, "step": 17515 }, { "epoch": 0.89, "learning_rate": 6.295490987506348e-07, "loss": 0.9245, "step": 17516 }, { "epoch": 0.89, "learning_rate": 6.289746325509061e-07, "loss": 0.8499, "step": 17517 }, { "epoch": 0.89, "learning_rate": 6.284004200617177e-07, "loss": 0.8843, "step": 17518 }, { "epoch": 0.89, "learning_rate": 6.278264612986173e-07, "loss": 0.8551, "step": 17519 }, { "epoch": 0.89, "learning_rate": 6.272527562771457e-07, "loss": 0.8596, "step": 17520 }, { "epoch": 0.89, "learning_rate": 6.266793050128306e-07, "loss": 0.8189, "step": 17521 }, { "epoch": 0.89, "learning_rate": 6.261061075212049e-07, "loss": 0.8548, "step": 17522 }, { "epoch": 0.89, "learning_rate": 6.255331638177809e-07, "loss": 0.8091, "step": 17523 }, { "epoch": 0.89, "learning_rate": 6.249604739180726e-07, "loss": 0.8476, "step": 17524 }, { "epoch": 0.89, "learning_rate": 6.243880378375877e-07, "loss": 0.9035, "step": 17525 }, { "epoch": 0.89, "learning_rate": 6.238158555918195e-07, "loss": 0.8497, "step": 17526 }, { "epoch": 0.89, "learning_rate": 6.23243927196262e-07, "loss": 0.8743, "step": 17527 }, { "epoch": 0.89, "learning_rate": 6.226722526663997e-07, "loss": 0.7784, "step": 17528 }, { "epoch": 0.89, "learning_rate": 6.221008320177102e-07, "loss": 0.9196, "step": 17529 }, { "epoch": 0.89, "learning_rate": 6.21529665265661e-07, "loss": 0.8581, "step": 17530 }, { "epoch": 0.89, "learning_rate": 6.209587524257199e-07, "loss": 0.991, "step": 17531 }, { "epoch": 0.89, "learning_rate": 6.203880935133421e-07, "loss": 0.8185, "step": 17532 }, { "epoch": 0.89, "learning_rate": 6.198176885439755e-07, "loss": 0.8415, "step": 17533 }, { "epoch": 0.89, "learning_rate": 6.192475375330686e-07, "loss": 0.8045, "step": 17534 }, { "epoch": 0.89, "learning_rate": 6.186776404960515e-07, "loss": 0.8645, "step": 17535 }, { "epoch": 0.89, "learning_rate": 6.181079974483573e-07, "loss": 0.9009, "step": 17536 }, { "epoch": 0.89, "learning_rate": 6.175386084054091e-07, "loss": 0.7828, "step": 17537 }, { "epoch": 0.89, "learning_rate": 6.169694733826181e-07, "loss": 0.7895, "step": 17538 }, { "epoch": 0.89, "learning_rate": 6.164005923953975e-07, "loss": 0.7748, "step": 17539 }, { "epoch": 0.89, "learning_rate": 6.158319654591471e-07, "loss": 0.8758, "step": 17540 }, { "epoch": 0.89, "learning_rate": 6.152635925892636e-07, "loss": 0.8232, "step": 17541 }, { "epoch": 0.89, "learning_rate": 6.146954738011313e-07, "loss": 0.8308, "step": 17542 }, { "epoch": 0.89, "learning_rate": 6.141276091101345e-07, "loss": 0.8434, "step": 17543 }, { "epoch": 0.89, "learning_rate": 6.135599985316487e-07, "loss": 0.8667, "step": 17544 }, { "epoch": 0.89, "learning_rate": 6.129926420810372e-07, "loss": 0.8643, "step": 17545 }, { "epoch": 0.89, "learning_rate": 6.124255397736645e-07, "loss": 1.0076, "step": 17546 }, { "epoch": 0.89, "learning_rate": 6.118586916248826e-07, "loss": 0.8805, "step": 17547 }, { "epoch": 0.89, "learning_rate": 6.112920976500369e-07, "loss": 0.7814, "step": 17548 }, { "epoch": 0.89, "learning_rate": 6.107257578644721e-07, "loss": 0.8133, "step": 17549 }, { "epoch": 0.89, "learning_rate": 6.101596722835157e-07, "loss": 0.8268, "step": 17550 }, { "epoch": 0.89, "learning_rate": 6.095938409224955e-07, "loss": 0.7806, "step": 17551 }, { "epoch": 0.89, "learning_rate": 6.090282637967315e-07, "loss": 0.8624, "step": 17552 }, { "epoch": 0.89, "learning_rate": 6.084629409215381e-07, "loss": 0.8775, "step": 17553 }, { "epoch": 0.89, "learning_rate": 6.078978723122176e-07, "loss": 0.883, "step": 17554 }, { "epoch": 0.89, "learning_rate": 6.073330579840697e-07, "loss": 0.7397, "step": 17555 }, { "epoch": 0.89, "learning_rate": 6.067684979523869e-07, "loss": 0.8795, "step": 17556 }, { "epoch": 0.89, "learning_rate": 6.062041922324513e-07, "loss": 0.9462, "step": 17557 }, { "epoch": 0.89, "learning_rate": 6.056401408395452e-07, "loss": 0.8275, "step": 17558 }, { "epoch": 0.89, "learning_rate": 6.050763437889373e-07, "loss": 0.9051, "step": 17559 }, { "epoch": 0.89, "learning_rate": 6.045128010958912e-07, "loss": 0.8174, "step": 17560 }, { "epoch": 0.89, "learning_rate": 6.039495127756656e-07, "loss": 0.8957, "step": 17561 }, { "epoch": 0.89, "learning_rate": 6.033864788435095e-07, "loss": 0.7962, "step": 17562 }, { "epoch": 0.89, "learning_rate": 6.028236993146674e-07, "loss": 0.8602, "step": 17563 }, { "epoch": 0.89, "learning_rate": 6.02261174204376e-07, "loss": 0.8667, "step": 17564 }, { "epoch": 0.89, "learning_rate": 6.016989035278664e-07, "loss": 0.8347, "step": 17565 }, { "epoch": 0.89, "learning_rate": 6.011368873003587e-07, "loss": 0.9648, "step": 17566 }, { "epoch": 0.89, "learning_rate": 6.005751255370695e-07, "loss": 0.9681, "step": 17567 }, { "epoch": 0.89, "learning_rate": 6.00013618253209e-07, "loss": 0.9278, "step": 17568 }, { "epoch": 0.89, "learning_rate": 5.994523654639783e-07, "loss": 0.8436, "step": 17569 }, { "epoch": 0.89, "learning_rate": 5.988913671845764e-07, "loss": 0.9484, "step": 17570 }, { "epoch": 0.89, "learning_rate": 5.983306234301856e-07, "loss": 0.8408, "step": 17571 }, { "epoch": 0.89, "learning_rate": 5.977701342159914e-07, "loss": 0.8978, "step": 17572 }, { "epoch": 0.89, "learning_rate": 5.972098995571696e-07, "loss": 0.7819, "step": 17573 }, { "epoch": 0.89, "learning_rate": 5.966499194688824e-07, "loss": 0.8344, "step": 17574 }, { "epoch": 0.89, "learning_rate": 5.960901939662967e-07, "loss": 0.9005, "step": 17575 }, { "epoch": 0.89, "learning_rate": 5.955307230645635e-07, "loss": 0.9333, "step": 17576 }, { "epoch": 0.89, "learning_rate": 5.949715067788297e-07, "loss": 0.8114, "step": 17577 }, { "epoch": 0.89, "learning_rate": 5.944125451242378e-07, "loss": 0.8366, "step": 17578 }, { "epoch": 0.89, "learning_rate": 5.938538381159187e-07, "loss": 0.907, "step": 17579 }, { "epoch": 0.89, "learning_rate": 5.932953857689994e-07, "loss": 0.8214, "step": 17580 }, { "epoch": 0.89, "learning_rate": 5.927371880985988e-07, "loss": 0.8936, "step": 17581 }, { "epoch": 0.89, "learning_rate": 5.921792451198316e-07, "loss": 0.9056, "step": 17582 }, { "epoch": 0.89, "learning_rate": 5.916215568478023e-07, "loss": 0.9366, "step": 17583 }, { "epoch": 0.89, "learning_rate": 5.910641232976078e-07, "loss": 0.8779, "step": 17584 }, { "epoch": 0.89, "learning_rate": 5.905069444843448e-07, "loss": 0.7972, "step": 17585 }, { "epoch": 0.89, "learning_rate": 5.899500204230912e-07, "loss": 0.9214, "step": 17586 }, { "epoch": 0.89, "learning_rate": 5.89393351128933e-07, "loss": 0.864, "step": 17587 }, { "epoch": 0.89, "learning_rate": 5.888369366169367e-07, "loss": 0.9744, "step": 17588 }, { "epoch": 0.89, "learning_rate": 5.88280776902167e-07, "loss": 0.8436, "step": 17589 }, { "epoch": 0.89, "learning_rate": 5.877248719996831e-07, "loss": 0.8033, "step": 17590 }, { "epoch": 0.89, "learning_rate": 5.871692219245329e-07, "loss": 0.8275, "step": 17591 }, { "epoch": 0.89, "learning_rate": 5.866138266917621e-07, "loss": 0.8379, "step": 17592 }, { "epoch": 0.89, "learning_rate": 5.860586863164075e-07, "loss": 0.9222, "step": 17593 }, { "epoch": 0.89, "learning_rate": 5.855038008134983e-07, "loss": 0.9241, "step": 17594 }, { "epoch": 0.89, "learning_rate": 5.84949170198057e-07, "loss": 0.9417, "step": 17595 }, { "epoch": 0.89, "learning_rate": 5.843947944851002e-07, "loss": 0.7989, "step": 17596 }, { "epoch": 0.89, "learning_rate": 5.838406736896374e-07, "loss": 0.8564, "step": 17597 }, { "epoch": 0.89, "learning_rate": 5.832868078266685e-07, "loss": 0.8922, "step": 17598 }, { "epoch": 0.89, "learning_rate": 5.827331969111927e-07, "loss": 0.9373, "step": 17599 }, { "epoch": 0.89, "learning_rate": 5.821798409581958e-07, "loss": 0.8002, "step": 17600 }, { "epoch": 0.89, "learning_rate": 5.816267399826592e-07, "loss": 0.7783, "step": 17601 }, { "epoch": 0.89, "learning_rate": 5.810738939995608e-07, "loss": 0.8222, "step": 17602 }, { "epoch": 0.89, "learning_rate": 5.805213030238632e-07, "loss": 0.8403, "step": 17603 }, { "epoch": 0.89, "learning_rate": 5.799689670705299e-07, "loss": 0.8088, "step": 17604 }, { "epoch": 0.89, "learning_rate": 5.794168861545157e-07, "loss": 0.7845, "step": 17605 }, { "epoch": 0.89, "learning_rate": 5.788650602907675e-07, "loss": 0.9063, "step": 17606 }, { "epoch": 0.89, "learning_rate": 5.783134894942222e-07, "loss": 0.7367, "step": 17607 }, { "epoch": 0.89, "learning_rate": 5.777621737798156e-07, "loss": 0.9069, "step": 17608 }, { "epoch": 0.89, "learning_rate": 5.772111131624758e-07, "loss": 0.8217, "step": 17609 }, { "epoch": 0.89, "learning_rate": 5.766603076571164e-07, "loss": 0.8182, "step": 17610 }, { "epoch": 0.89, "learning_rate": 5.761097572786579e-07, "loss": 0.9186, "step": 17611 }, { "epoch": 0.89, "learning_rate": 5.75559462041998e-07, "loss": 0.8242, "step": 17612 }, { "epoch": 0.89, "learning_rate": 5.750094219620406e-07, "loss": 0.803, "step": 17613 }, { "epoch": 0.89, "learning_rate": 5.744596370536759e-07, "loss": 0.8126, "step": 17614 }, { "epoch": 0.89, "learning_rate": 5.739101073317876e-07, "loss": 0.8981, "step": 17615 }, { "epoch": 0.89, "learning_rate": 5.73360832811255e-07, "loss": 0.8198, "step": 17616 }, { "epoch": 0.89, "learning_rate": 5.728118135069483e-07, "loss": 0.8578, "step": 17617 }, { "epoch": 0.9, "learning_rate": 5.722630494337333e-07, "loss": 0.8774, "step": 17618 }, { "epoch": 0.9, "learning_rate": 5.717145406064639e-07, "loss": 0.7616, "step": 17619 }, { "epoch": 0.9, "learning_rate": 5.711662870399936e-07, "loss": 0.8832, "step": 17620 }, { "epoch": 0.9, "learning_rate": 5.706182887491651e-07, "loss": 0.8969, "step": 17621 }, { "epoch": 0.9, "learning_rate": 5.70070545748812e-07, "loss": 0.9056, "step": 17622 }, { "epoch": 0.9, "learning_rate": 5.695230580537691e-07, "loss": 0.7709, "step": 17623 }, { "epoch": 0.9, "learning_rate": 5.689758256788547e-07, "loss": 0.8471, "step": 17624 }, { "epoch": 0.9, "learning_rate": 5.684288486388856e-07, "loss": 0.9733, "step": 17625 }, { "epoch": 0.9, "learning_rate": 5.678821269486734e-07, "loss": 0.9188, "step": 17626 }, { "epoch": 0.9, "learning_rate": 5.673356606230151e-07, "loss": 0.818, "step": 17627 }, { "epoch": 0.9, "learning_rate": 5.66789449676709e-07, "loss": 0.7997, "step": 17628 }, { "epoch": 0.9, "learning_rate": 5.66243494124542e-07, "loss": 0.869, "step": 17629 }, { "epoch": 0.9, "learning_rate": 5.656977939812969e-07, "loss": 0.8847, "step": 17630 }, { "epoch": 0.9, "learning_rate": 5.651523492617462e-07, "loss": 0.8702, "step": 17631 }, { "epoch": 0.9, "learning_rate": 5.646071599806568e-07, "loss": 0.7846, "step": 17632 }, { "epoch": 0.9, "learning_rate": 5.640622261527917e-07, "loss": 0.8505, "step": 17633 }, { "epoch": 0.9, "learning_rate": 5.635175477929e-07, "loss": 0.9053, "step": 17634 }, { "epoch": 0.9, "learning_rate": 5.629731249157344e-07, "loss": 0.8985, "step": 17635 }, { "epoch": 0.9, "learning_rate": 5.624289575360286e-07, "loss": 0.815, "step": 17636 }, { "epoch": 0.9, "learning_rate": 5.618850456685188e-07, "loss": 0.8043, "step": 17637 }, { "epoch": 0.9, "learning_rate": 5.613413893279318e-07, "loss": 0.7486, "step": 17638 }, { "epoch": 0.9, "learning_rate": 5.607979885289816e-07, "loss": 0.8627, "step": 17639 }, { "epoch": 0.9, "learning_rate": 5.602548432863863e-07, "loss": 0.8966, "step": 17640 }, { "epoch": 0.9, "learning_rate": 5.597119536148466e-07, "loss": 0.8312, "step": 17641 }, { "epoch": 0.9, "learning_rate": 5.591693195290615e-07, "loss": 0.8425, "step": 17642 }, { "epoch": 0.9, "learning_rate": 5.58626941043725e-07, "loss": 0.7696, "step": 17643 }, { "epoch": 0.9, "learning_rate": 5.580848181735176e-07, "loss": 0.7544, "step": 17644 }, { "epoch": 0.9, "learning_rate": 5.575429509331177e-07, "loss": 0.871, "step": 17645 }, { "epoch": 0.9, "learning_rate": 5.570013393371964e-07, "loss": 0.8437, "step": 17646 }, { "epoch": 0.9, "learning_rate": 5.564599834004203e-07, "loss": 0.9632, "step": 17647 }, { "epoch": 0.9, "learning_rate": 5.559188831374396e-07, "loss": 0.9291, "step": 17648 }, { "epoch": 0.9, "learning_rate": 5.553780385629082e-07, "loss": 0.8668, "step": 17649 }, { "epoch": 0.9, "learning_rate": 5.548374496914699e-07, "loss": 0.873, "step": 17650 }, { "epoch": 0.9, "learning_rate": 5.542971165377564e-07, "loss": 0.9167, "step": 17651 }, { "epoch": 0.9, "learning_rate": 5.537570391164004e-07, "loss": 0.9641, "step": 17652 }, { "epoch": 0.9, "learning_rate": 5.532172174420225e-07, "loss": 0.8634, "step": 17653 }, { "epoch": 0.9, "learning_rate": 5.526776515292376e-07, "loss": 0.8404, "step": 17654 }, { "epoch": 0.9, "learning_rate": 5.521383413926551e-07, "loss": 0.939, "step": 17655 }, { "epoch": 0.9, "learning_rate": 5.515992870468734e-07, "loss": 0.8944, "step": 17656 }, { "epoch": 0.9, "learning_rate": 5.510604885064896e-07, "loss": 0.8817, "step": 17657 }, { "epoch": 0.9, "learning_rate": 5.505219457860899e-07, "loss": 0.8821, "step": 17658 }, { "epoch": 0.9, "learning_rate": 5.499836589002561e-07, "loss": 0.951, "step": 17659 }, { "epoch": 0.9, "learning_rate": 5.494456278635596e-07, "loss": 0.7943, "step": 17660 }, { "epoch": 0.9, "learning_rate": 5.489078526905678e-07, "loss": 0.874, "step": 17661 }, { "epoch": 0.9, "learning_rate": 5.483703333958435e-07, "loss": 0.8721, "step": 17662 }, { "epoch": 0.9, "learning_rate": 5.478330699939327e-07, "loss": 0.861, "step": 17663 }, { "epoch": 0.9, "learning_rate": 5.472960624993883e-07, "loss": 0.8426, "step": 17664 }, { "epoch": 0.9, "learning_rate": 5.467593109267444e-07, "loss": 0.9111, "step": 17665 }, { "epoch": 0.9, "learning_rate": 5.462228152905347e-07, "loss": 0.8732, "step": 17666 }, { "epoch": 0.9, "learning_rate": 5.456865756052853e-07, "loss": 0.8342, "step": 17667 }, { "epoch": 0.9, "learning_rate": 5.451505918855127e-07, "loss": 0.9105, "step": 17668 }, { "epoch": 0.9, "learning_rate": 5.446148641457283e-07, "loss": 0.85, "step": 17669 }, { "epoch": 0.9, "learning_rate": 5.440793924004373e-07, "loss": 0.8876, "step": 17670 }, { "epoch": 0.9, "learning_rate": 5.43544176664137e-07, "loss": 0.9199, "step": 17671 }, { "epoch": 0.9, "learning_rate": 5.430092169513168e-07, "loss": 0.7878, "step": 17672 }, { "epoch": 0.9, "learning_rate": 5.424745132764598e-07, "loss": 0.8504, "step": 17673 }, { "epoch": 0.9, "learning_rate": 5.419400656540452e-07, "loss": 0.8335, "step": 17674 }, { "epoch": 0.9, "learning_rate": 5.414058740985373e-07, "loss": 0.8263, "step": 17675 }, { "epoch": 0.9, "learning_rate": 5.408719386244054e-07, "loss": 0.7872, "step": 17676 }, { "epoch": 0.9, "learning_rate": 5.403382592461004e-07, "loss": 0.9863, "step": 17677 }, { "epoch": 0.9, "learning_rate": 5.398048359780727e-07, "loss": 0.8607, "step": 17678 }, { "epoch": 0.9, "learning_rate": 5.392716688347665e-07, "loss": 0.8362, "step": 17679 }, { "epoch": 0.9, "learning_rate": 5.387387578306125e-07, "loss": 0.8511, "step": 17680 }, { "epoch": 0.9, "learning_rate": 5.382061029800401e-07, "loss": 0.8227, "step": 17681 }, { "epoch": 0.9, "learning_rate": 5.376737042974712e-07, "loss": 0.8974, "step": 17682 }, { "epoch": 0.9, "learning_rate": 5.371415617973208e-07, "loss": 0.9125, "step": 17683 }, { "epoch": 0.9, "learning_rate": 5.366096754939942e-07, "loss": 0.9435, "step": 17684 }, { "epoch": 0.9, "learning_rate": 5.360780454018922e-07, "loss": 0.8113, "step": 17685 }, { "epoch": 0.9, "learning_rate": 5.355466715354096e-07, "loss": 0.9174, "step": 17686 }, { "epoch": 0.9, "learning_rate": 5.350155539089286e-07, "loss": 0.9834, "step": 17687 }, { "epoch": 0.9, "learning_rate": 5.344846925368341e-07, "loss": 0.897, "step": 17688 }, { "epoch": 0.9, "learning_rate": 5.339540874334959e-07, "loss": 0.9057, "step": 17689 }, { "epoch": 0.9, "learning_rate": 5.334237386132779e-07, "loss": 0.829, "step": 17690 }, { "epoch": 0.9, "learning_rate": 5.328936460905432e-07, "loss": 0.8326, "step": 17691 }, { "epoch": 0.9, "learning_rate": 5.32363809879639e-07, "loss": 0.8625, "step": 17692 }, { "epoch": 0.9, "learning_rate": 5.31834229994912e-07, "loss": 0.8185, "step": 17693 }, { "epoch": 0.9, "learning_rate": 5.313049064506993e-07, "loss": 0.9379, "step": 17694 }, { "epoch": 0.9, "learning_rate": 5.30775839261335e-07, "loss": 0.8474, "step": 17695 }, { "epoch": 0.9, "learning_rate": 5.302470284411377e-07, "loss": 0.8716, "step": 17696 }, { "epoch": 0.9, "learning_rate": 5.297184740044281e-07, "loss": 0.7788, "step": 17697 }, { "epoch": 0.9, "learning_rate": 5.29190175965516e-07, "loss": 0.9082, "step": 17698 }, { "epoch": 0.9, "learning_rate": 5.28662134338701e-07, "loss": 0.9017, "step": 17699 }, { "epoch": 0.9, "learning_rate": 5.28134349138284e-07, "loss": 0.8313, "step": 17700 }, { "epoch": 0.9, "learning_rate": 5.276068203785523e-07, "loss": 0.9365, "step": 17701 }, { "epoch": 0.9, "learning_rate": 5.270795480737878e-07, "loss": 0.8419, "step": 17702 }, { "epoch": 0.9, "learning_rate": 5.26552532238267e-07, "loss": 0.8666, "step": 17703 }, { "epoch": 0.9, "learning_rate": 5.260257728862561e-07, "loss": 0.8484, "step": 17704 }, { "epoch": 0.9, "learning_rate": 5.254992700320183e-07, "loss": 0.9156, "step": 17705 }, { "epoch": 0.9, "learning_rate": 5.249730236898077e-07, "loss": 0.9012, "step": 17706 }, { "epoch": 0.9, "learning_rate": 5.244470338738705e-07, "loss": 0.8629, "step": 17707 }, { "epoch": 0.9, "learning_rate": 5.239213005984512e-07, "loss": 0.8693, "step": 17708 }, { "epoch": 0.9, "learning_rate": 5.233958238777803e-07, "loss": 0.9641, "step": 17709 }, { "epoch": 0.9, "learning_rate": 5.228706037260845e-07, "loss": 0.7739, "step": 17710 }, { "epoch": 0.9, "learning_rate": 5.223456401575833e-07, "loss": 0.7875, "step": 17711 }, { "epoch": 0.9, "learning_rate": 5.218209331864932e-07, "loss": 0.8099, "step": 17712 }, { "epoch": 0.9, "learning_rate": 5.212964828270151e-07, "loss": 0.9943, "step": 17713 }, { "epoch": 0.9, "learning_rate": 5.207722890933509e-07, "loss": 0.8285, "step": 17714 }, { "epoch": 0.9, "learning_rate": 5.202483519996926e-07, "loss": 0.9009, "step": 17715 }, { "epoch": 0.9, "learning_rate": 5.197246715602211e-07, "loss": 0.8442, "step": 17716 }, { "epoch": 0.9, "learning_rate": 5.192012477891217e-07, "loss": 0.8386, "step": 17717 }, { "epoch": 0.9, "learning_rate": 5.186780807005609e-07, "loss": 0.8507, "step": 17718 }, { "epoch": 0.9, "learning_rate": 5.181551703087028e-07, "loss": 0.7586, "step": 17719 }, { "epoch": 0.9, "learning_rate": 5.176325166277075e-07, "loss": 0.8879, "step": 17720 }, { "epoch": 0.9, "learning_rate": 5.17110119671721e-07, "loss": 0.8761, "step": 17721 }, { "epoch": 0.9, "learning_rate": 5.1658797945489e-07, "loss": 0.8538, "step": 17722 }, { "epoch": 0.9, "learning_rate": 5.160660959913488e-07, "loss": 0.8676, "step": 17723 }, { "epoch": 0.9, "learning_rate": 5.155444692952305e-07, "loss": 0.82, "step": 17724 }, { "epoch": 0.9, "learning_rate": 5.150230993806537e-07, "loss": 0.8372, "step": 17725 }, { "epoch": 0.9, "learning_rate": 5.145019862617339e-07, "loss": 0.8457, "step": 17726 }, { "epoch": 0.9, "learning_rate": 5.139811299525843e-07, "loss": 0.8697, "step": 17727 }, { "epoch": 0.9, "learning_rate": 5.134605304672991e-07, "loss": 0.8261, "step": 17728 }, { "epoch": 0.9, "learning_rate": 5.129401878199803e-07, "loss": 0.8176, "step": 17729 }, { "epoch": 0.9, "learning_rate": 5.12420102024711e-07, "loss": 0.8946, "step": 17730 }, { "epoch": 0.9, "learning_rate": 5.119002730955735e-07, "loss": 0.9052, "step": 17731 }, { "epoch": 0.9, "learning_rate": 5.113807010466432e-07, "loss": 0.8479, "step": 17732 }, { "epoch": 0.9, "learning_rate": 5.108613858919842e-07, "loss": 0.837, "step": 17733 }, { "epoch": 0.9, "learning_rate": 5.103423276456565e-07, "loss": 0.8663, "step": 17734 }, { "epoch": 0.9, "learning_rate": 5.098235263217155e-07, "loss": 0.8629, "step": 17735 }, { "epoch": 0.9, "learning_rate": 5.093049819342056e-07, "loss": 0.8788, "step": 17736 }, { "epoch": 0.9, "learning_rate": 5.087866944971654e-07, "loss": 0.8816, "step": 17737 }, { "epoch": 0.9, "learning_rate": 5.082686640246281e-07, "loss": 0.9197, "step": 17738 }, { "epoch": 0.9, "learning_rate": 5.077508905306184e-07, "loss": 0.875, "step": 17739 }, { "epoch": 0.9, "learning_rate": 5.072333740291524e-07, "loss": 0.7718, "step": 17740 }, { "epoch": 0.9, "learning_rate": 5.067161145342459e-07, "loss": 0.8658, "step": 17741 }, { "epoch": 0.9, "learning_rate": 5.061991120598997e-07, "loss": 0.8955, "step": 17742 }, { "epoch": 0.9, "learning_rate": 5.056823666201116e-07, "loss": 0.9167, "step": 17743 }, { "epoch": 0.9, "learning_rate": 5.051658782288737e-07, "loss": 0.884, "step": 17744 }, { "epoch": 0.9, "learning_rate": 5.04649646900166e-07, "loss": 0.9227, "step": 17745 }, { "epoch": 0.9, "learning_rate": 5.041336726479673e-07, "loss": 0.9144, "step": 17746 }, { "epoch": 0.9, "learning_rate": 5.036179554862464e-07, "loss": 0.8613, "step": 17747 }, { "epoch": 0.9, "learning_rate": 5.031024954289676e-07, "loss": 0.8893, "step": 17748 }, { "epoch": 0.9, "learning_rate": 5.025872924900832e-07, "loss": 0.8087, "step": 17749 }, { "epoch": 0.9, "learning_rate": 5.020723466835442e-07, "loss": 0.9229, "step": 17750 }, { "epoch": 0.9, "learning_rate": 5.015576580232917e-07, "loss": 0.7828, "step": 17751 }, { "epoch": 0.9, "learning_rate": 5.010432265232578e-07, "loss": 0.8534, "step": 17752 }, { "epoch": 0.9, "learning_rate": 5.00529052197376e-07, "loss": 0.92, "step": 17753 }, { "epoch": 0.9, "learning_rate": 5.000151350595605e-07, "loss": 0.8096, "step": 17754 }, { "epoch": 0.9, "learning_rate": 4.995014751237293e-07, "loss": 0.851, "step": 17755 }, { "epoch": 0.9, "learning_rate": 4.989880724037888e-07, "loss": 0.8782, "step": 17756 }, { "epoch": 0.9, "learning_rate": 4.984749269136368e-07, "loss": 0.8606, "step": 17757 }, { "epoch": 0.9, "learning_rate": 4.979620386671669e-07, "loss": 0.8444, "step": 17758 }, { "epoch": 0.9, "learning_rate": 4.974494076782654e-07, "loss": 0.9555, "step": 17759 }, { "epoch": 0.9, "learning_rate": 4.969370339608138e-07, "loss": 0.9105, "step": 17760 }, { "epoch": 0.9, "learning_rate": 4.964249175286784e-07, "loss": 0.9211, "step": 17761 }, { "epoch": 0.9, "learning_rate": 4.959130583957283e-07, "loss": 0.8968, "step": 17762 }, { "epoch": 0.9, "learning_rate": 4.954014565758225e-07, "loss": 0.8646, "step": 17763 }, { "epoch": 0.9, "learning_rate": 4.948901120828065e-07, "loss": 0.8863, "step": 17764 }, { "epoch": 0.9, "learning_rate": 4.943790249305314e-07, "loss": 0.8185, "step": 17765 }, { "epoch": 0.9, "learning_rate": 4.938681951328295e-07, "loss": 0.8953, "step": 17766 }, { "epoch": 0.9, "learning_rate": 4.933576227035319e-07, "loss": 0.8121, "step": 17767 }, { "epoch": 0.9, "learning_rate": 4.928473076564643e-07, "loss": 0.9514, "step": 17768 }, { "epoch": 0.9, "learning_rate": 4.923372500054391e-07, "loss": 0.7878, "step": 17769 }, { "epoch": 0.9, "learning_rate": 4.918274497642683e-07, "loss": 0.8519, "step": 17770 }, { "epoch": 0.9, "learning_rate": 4.913179069467533e-07, "loss": 0.9098, "step": 17771 }, { "epoch": 0.9, "learning_rate": 4.908086215666907e-07, "loss": 0.8071, "step": 17772 }, { "epoch": 0.9, "learning_rate": 4.902995936378663e-07, "loss": 0.8111, "step": 17773 }, { "epoch": 0.9, "learning_rate": 4.897908231740634e-07, "loss": 0.804, "step": 17774 }, { "epoch": 0.9, "learning_rate": 4.892823101890543e-07, "loss": 0.8354, "step": 17775 }, { "epoch": 0.9, "learning_rate": 4.887740546966091e-07, "loss": 0.8077, "step": 17776 }, { "epoch": 0.9, "learning_rate": 4.882660567104891e-07, "loss": 0.9306, "step": 17777 }, { "epoch": 0.9, "learning_rate": 4.877583162444444e-07, "loss": 0.8646, "step": 17778 }, { "epoch": 0.9, "learning_rate": 4.872508333122228e-07, "loss": 0.9354, "step": 17779 }, { "epoch": 0.9, "learning_rate": 4.867436079275656e-07, "loss": 0.8918, "step": 17780 }, { "epoch": 0.9, "learning_rate": 4.862366401042006e-07, "loss": 0.8263, "step": 17781 }, { "epoch": 0.9, "learning_rate": 4.857299298558594e-07, "loss": 0.9047, "step": 17782 }, { "epoch": 0.9, "learning_rate": 4.852234771962572e-07, "loss": 0.9736, "step": 17783 }, { "epoch": 0.9, "learning_rate": 4.847172821391066e-07, "loss": 0.8051, "step": 17784 }, { "epoch": 0.9, "learning_rate": 4.842113446981134e-07, "loss": 0.9304, "step": 17785 }, { "epoch": 0.9, "learning_rate": 4.837056648869721e-07, "loss": 0.8796, "step": 17786 }, { "epoch": 0.9, "learning_rate": 4.83200242719376e-07, "loss": 0.8343, "step": 17787 }, { "epoch": 0.9, "learning_rate": 4.826950782090078e-07, "loss": 0.8124, "step": 17788 }, { "epoch": 0.9, "learning_rate": 4.821901713695465e-07, "loss": 0.8289, "step": 17789 }, { "epoch": 0.9, "learning_rate": 4.816855222146577e-07, "loss": 0.8936, "step": 17790 }, { "epoch": 0.9, "learning_rate": 4.811811307580083e-07, "loss": 0.8812, "step": 17791 }, { "epoch": 0.9, "learning_rate": 4.80676997013253e-07, "loss": 0.788, "step": 17792 }, { "epoch": 0.9, "learning_rate": 4.801731209940375e-07, "loss": 0.8615, "step": 17793 }, { "epoch": 0.9, "learning_rate": 4.796695027140097e-07, "loss": 0.8265, "step": 17794 }, { "epoch": 0.9, "learning_rate": 4.791661421867999e-07, "loss": 0.9273, "step": 17795 }, { "epoch": 0.9, "learning_rate": 4.786630394260361e-07, "loss": 0.8421, "step": 17796 }, { "epoch": 0.9, "learning_rate": 4.78160194445344e-07, "loss": 0.8798, "step": 17797 }, { "epoch": 0.9, "learning_rate": 4.776576072583317e-07, "loss": 0.9083, "step": 17798 }, { "epoch": 0.9, "learning_rate": 4.771552778786082e-07, "loss": 0.8763, "step": 17799 }, { "epoch": 0.9, "learning_rate": 4.7665320631977485e-07, "loss": 0.9398, "step": 17800 }, { "epoch": 0.9, "learning_rate": 4.761513925954253e-07, "loss": 0.9575, "step": 17801 }, { "epoch": 0.9, "learning_rate": 4.75649836719142e-07, "loss": 0.8635, "step": 17802 }, { "epoch": 0.9, "learning_rate": 4.751485387045074e-07, "loss": 0.8059, "step": 17803 }, { "epoch": 0.9, "learning_rate": 4.7464749856509285e-07, "loss": 0.9058, "step": 17804 }, { "epoch": 0.9, "learning_rate": 4.7414671631446087e-07, "loss": 0.8602, "step": 17805 }, { "epoch": 0.9, "learning_rate": 4.73646191966175e-07, "loss": 0.9268, "step": 17806 }, { "epoch": 0.9, "learning_rate": 4.731459255337811e-07, "loss": 0.8699, "step": 17807 }, { "epoch": 0.9, "learning_rate": 4.726459170308262e-07, "loss": 0.825, "step": 17808 }, { "epoch": 0.9, "learning_rate": 4.721461664708482e-07, "loss": 0.9144, "step": 17809 }, { "epoch": 0.9, "learning_rate": 4.7164667386737415e-07, "loss": 0.7956, "step": 17810 }, { "epoch": 0.9, "learning_rate": 4.711474392339288e-07, "loss": 0.761, "step": 17811 }, { "epoch": 0.9, "learning_rate": 4.70648462584028e-07, "loss": 0.8763, "step": 17812 }, { "epoch": 0.9, "learning_rate": 4.7014974393118417e-07, "loss": 0.907, "step": 17813 }, { "epoch": 0.9, "learning_rate": 4.6965128328889554e-07, "loss": 0.8466, "step": 17814 }, { "epoch": 0.91, "learning_rate": 4.6915308067065903e-07, "loss": 0.8192, "step": 17815 }, { "epoch": 0.91, "learning_rate": 4.6865513608996385e-07, "loss": 0.9251, "step": 17816 }, { "epoch": 0.91, "learning_rate": 4.68157449560287e-07, "loss": 0.8953, "step": 17817 }, { "epoch": 0.91, "learning_rate": 4.676600210951099e-07, "loss": 0.8999, "step": 17818 }, { "epoch": 0.91, "learning_rate": 4.6716285070789405e-07, "loss": 0.8312, "step": 17819 }, { "epoch": 0.91, "learning_rate": 4.6666593841210193e-07, "loss": 0.9088, "step": 17820 }, { "epoch": 0.91, "learning_rate": 4.6616928422118846e-07, "loss": 0.8809, "step": 17821 }, { "epoch": 0.91, "learning_rate": 4.656728881485972e-07, "loss": 0.8446, "step": 17822 }, { "epoch": 0.91, "learning_rate": 4.6517675020776866e-07, "loss": 0.9021, "step": 17823 }, { "epoch": 0.91, "learning_rate": 4.646808704121353e-07, "loss": 0.8935, "step": 17824 }, { "epoch": 0.91, "learning_rate": 4.641852487751242e-07, "loss": 0.9227, "step": 17825 }, { "epoch": 0.91, "learning_rate": 4.636898853101512e-07, "loss": 0.8838, "step": 17826 }, { "epoch": 0.91, "learning_rate": 4.63194780030628e-07, "loss": 1.0021, "step": 17827 }, { "epoch": 0.91, "learning_rate": 4.626999329499604e-07, "loss": 0.8669, "step": 17828 }, { "epoch": 0.91, "learning_rate": 4.622053440815444e-07, "loss": 0.8659, "step": 17829 }, { "epoch": 0.91, "learning_rate": 4.6171101343877253e-07, "loss": 0.8631, "step": 17830 }, { "epoch": 0.91, "learning_rate": 4.6121694103502647e-07, "loss": 0.8667, "step": 17831 }, { "epoch": 0.91, "learning_rate": 4.60723126883682e-07, "loss": 0.8616, "step": 17832 }, { "epoch": 0.91, "learning_rate": 4.6022957099811304e-07, "loss": 0.8017, "step": 17833 }, { "epoch": 0.91, "learning_rate": 4.597362733916766e-07, "loss": 0.8576, "step": 17834 }, { "epoch": 0.91, "learning_rate": 4.5924323407772973e-07, "loss": 0.7467, "step": 17835 }, { "epoch": 0.91, "learning_rate": 4.5875045306962294e-07, "loss": 0.894, "step": 17836 }, { "epoch": 0.91, "learning_rate": 4.582579303806978e-07, "loss": 0.9072, "step": 17837 }, { "epoch": 0.91, "learning_rate": 4.5776566602428576e-07, "loss": 0.8872, "step": 17838 }, { "epoch": 0.91, "learning_rate": 4.572736600137162e-07, "loss": 0.8843, "step": 17839 }, { "epoch": 0.91, "learning_rate": 4.567819123623118e-07, "loss": 0.968, "step": 17840 }, { "epoch": 0.91, "learning_rate": 4.5629042308337977e-07, "loss": 0.9286, "step": 17841 }, { "epoch": 0.91, "learning_rate": 4.5579919219023494e-07, "loss": 0.876, "step": 17842 }, { "epoch": 0.91, "learning_rate": 4.553082196961711e-07, "loss": 0.7862, "step": 17843 }, { "epoch": 0.91, "learning_rate": 4.548175056144821e-07, "loss": 0.7856, "step": 17844 }, { "epoch": 0.91, "learning_rate": 4.543270499584551e-07, "loss": 0.9174, "step": 17845 }, { "epoch": 0.91, "learning_rate": 4.53836852741365e-07, "loss": 0.8073, "step": 17846 }, { "epoch": 0.91, "learning_rate": 4.533469139764879e-07, "loss": 0.9232, "step": 17847 }, { "epoch": 0.91, "learning_rate": 4.5285723367708533e-07, "loss": 0.9674, "step": 17848 }, { "epoch": 0.91, "learning_rate": 4.5236781185641677e-07, "loss": 0.8149, "step": 17849 }, { "epoch": 0.91, "learning_rate": 4.5187864852773156e-07, "loss": 0.8917, "step": 17850 }, { "epoch": 0.91, "learning_rate": 4.513897437042736e-07, "loss": 0.8738, "step": 17851 }, { "epoch": 0.91, "learning_rate": 4.509010973992778e-07, "loss": 0.8629, "step": 17852 }, { "epoch": 0.91, "learning_rate": 4.5041270962597694e-07, "loss": 0.9053, "step": 17853 }, { "epoch": 0.91, "learning_rate": 4.499245803975927e-07, "loss": 0.8912, "step": 17854 }, { "epoch": 0.91, "learning_rate": 4.494367097273389e-07, "loss": 0.8813, "step": 17855 }, { "epoch": 0.91, "learning_rate": 4.48949097628425e-07, "loss": 0.8575, "step": 17856 }, { "epoch": 0.91, "learning_rate": 4.4846174411405373e-07, "loss": 0.895, "step": 17857 }, { "epoch": 0.91, "learning_rate": 4.4797464919741573e-07, "loss": 0.9111, "step": 17858 }, { "epoch": 0.91, "learning_rate": 4.474878128917048e-07, "loss": 0.8315, "step": 17859 }, { "epoch": 0.91, "learning_rate": 4.470012352100972e-07, "loss": 0.7475, "step": 17860 }, { "epoch": 0.91, "learning_rate": 4.4651491616576783e-07, "loss": 0.8643, "step": 17861 }, { "epoch": 0.91, "learning_rate": 4.460288557718828e-07, "loss": 0.8386, "step": 17862 }, { "epoch": 0.91, "learning_rate": 4.4554305404160167e-07, "loss": 0.8679, "step": 17863 }, { "epoch": 0.91, "learning_rate": 4.4505751098807723e-07, "loss": 0.8483, "step": 17864 }, { "epoch": 0.91, "learning_rate": 4.4457222662445454e-07, "loss": 0.8823, "step": 17865 }, { "epoch": 0.91, "learning_rate": 4.440872009638753e-07, "loss": 0.901, "step": 17866 }, { "epoch": 0.91, "learning_rate": 4.436024340194656e-07, "loss": 0.8139, "step": 17867 }, { "epoch": 0.91, "learning_rate": 4.4311792580435276e-07, "loss": 0.881, "step": 17868 }, { "epoch": 0.91, "learning_rate": 4.426336763316563e-07, "loss": 0.9007, "step": 17869 }, { "epoch": 0.91, "learning_rate": 4.421496856144825e-07, "loss": 0.875, "step": 17870 }, { "epoch": 0.91, "learning_rate": 4.416659536659396e-07, "loss": 0.7812, "step": 17871 }, { "epoch": 0.91, "learning_rate": 4.411824804991194e-07, "loss": 0.8786, "step": 17872 }, { "epoch": 0.91, "learning_rate": 4.4069926612711477e-07, "loss": 0.8032, "step": 17873 }, { "epoch": 0.91, "learning_rate": 4.4021631056300863e-07, "loss": 0.8066, "step": 17874 }, { "epoch": 0.91, "learning_rate": 4.397336138198738e-07, "loss": 0.8299, "step": 17875 }, { "epoch": 0.91, "learning_rate": 4.392511759107798e-07, "loss": 0.8848, "step": 17876 }, { "epoch": 0.91, "learning_rate": 4.387689968487874e-07, "loss": 0.9364, "step": 17877 }, { "epoch": 0.91, "learning_rate": 4.382870766469549e-07, "loss": 0.9135, "step": 17878 }, { "epoch": 0.91, "learning_rate": 4.378054153183242e-07, "loss": 0.8714, "step": 17879 }, { "epoch": 0.91, "learning_rate": 4.3732401287593927e-07, "loss": 0.7224, "step": 17880 }, { "epoch": 0.91, "learning_rate": 4.3684286933283415e-07, "loss": 0.8197, "step": 17881 }, { "epoch": 0.91, "learning_rate": 4.3636198470203173e-07, "loss": 0.8861, "step": 17882 }, { "epoch": 0.91, "learning_rate": 4.3588135899655603e-07, "loss": 0.8518, "step": 17883 }, { "epoch": 0.91, "learning_rate": 4.354009922294156e-07, "loss": 0.8662, "step": 17884 }, { "epoch": 0.91, "learning_rate": 4.349208844136177e-07, "loss": 0.8438, "step": 17885 }, { "epoch": 0.91, "learning_rate": 4.344410355621609e-07, "loss": 0.9139, "step": 17886 }, { "epoch": 0.91, "learning_rate": 4.3396144568803587e-07, "loss": 0.8828, "step": 17887 }, { "epoch": 0.91, "learning_rate": 4.3348211480422674e-07, "loss": 0.8817, "step": 17888 }, { "epoch": 0.91, "learning_rate": 4.3300304292371087e-07, "loss": 0.9045, "step": 17889 }, { "epoch": 0.91, "learning_rate": 4.325242300594612e-07, "loss": 0.7571, "step": 17890 }, { "epoch": 0.91, "learning_rate": 4.3204567622443736e-07, "loss": 0.7776, "step": 17891 }, { "epoch": 0.91, "learning_rate": 4.315673814315968e-07, "loss": 0.7827, "step": 17892 }, { "epoch": 0.91, "learning_rate": 4.310893456938914e-07, "loss": 0.8811, "step": 17893 }, { "epoch": 0.91, "learning_rate": 4.306115690242585e-07, "loss": 0.8087, "step": 17894 }, { "epoch": 0.91, "learning_rate": 4.3013405143564003e-07, "loss": 0.9445, "step": 17895 }, { "epoch": 0.91, "learning_rate": 4.2965679294095784e-07, "loss": 0.8119, "step": 17896 }, { "epoch": 0.91, "learning_rate": 4.291797935531361e-07, "loss": 0.9611, "step": 17897 }, { "epoch": 0.91, "learning_rate": 4.287030532850911e-07, "loss": 0.874, "step": 17898 }, { "epoch": 0.91, "learning_rate": 4.2822657214972585e-07, "loss": 0.8748, "step": 17899 }, { "epoch": 0.91, "learning_rate": 4.277503501599434e-07, "loss": 0.8879, "step": 17900 }, { "epoch": 0.91, "learning_rate": 4.2727438732863557e-07, "loss": 0.8752, "step": 17901 }, { "epoch": 0.91, "learning_rate": 4.267986836686899e-07, "loss": 0.8938, "step": 17902 }, { "epoch": 0.91, "learning_rate": 4.2632323919298393e-07, "loss": 0.9097, "step": 17903 }, { "epoch": 0.91, "learning_rate": 4.2584805391439054e-07, "loss": 0.8737, "step": 17904 }, { "epoch": 0.91, "learning_rate": 4.2537312784577734e-07, "loss": 0.8858, "step": 17905 }, { "epoch": 0.91, "learning_rate": 4.248984609999962e-07, "loss": 0.7888, "step": 17906 }, { "epoch": 0.91, "learning_rate": 4.2442405338990464e-07, "loss": 0.8859, "step": 17907 }, { "epoch": 0.91, "learning_rate": 4.2394990502834353e-07, "loss": 0.8066, "step": 17908 }, { "epoch": 0.91, "learning_rate": 4.2347601592815147e-07, "loss": 0.9167, "step": 17909 }, { "epoch": 0.91, "learning_rate": 4.2300238610215703e-07, "loss": 0.8498, "step": 17910 }, { "epoch": 0.91, "learning_rate": 4.2252901556318446e-07, "loss": 0.8076, "step": 17911 }, { "epoch": 0.91, "learning_rate": 4.22055904324048e-07, "loss": 0.8791, "step": 17912 }, { "epoch": 0.91, "learning_rate": 4.2158305239755836e-07, "loss": 0.973, "step": 17913 }, { "epoch": 0.91, "learning_rate": 4.2111045979651655e-07, "loss": 0.8327, "step": 17914 }, { "epoch": 0.91, "learning_rate": 4.206381265337189e-07, "loss": 0.8201, "step": 17915 }, { "epoch": 0.91, "learning_rate": 4.20166052621952e-07, "loss": 0.8004, "step": 17916 }, { "epoch": 0.91, "learning_rate": 4.196942380739966e-07, "loss": 0.7904, "step": 17917 }, { "epoch": 0.91, "learning_rate": 4.192226829026269e-07, "loss": 0.7904, "step": 17918 }, { "epoch": 0.91, "learning_rate": 4.1875138712061057e-07, "loss": 0.8375, "step": 17919 }, { "epoch": 0.91, "learning_rate": 4.182803507407063e-07, "loss": 0.7844, "step": 17920 }, { "epoch": 0.91, "learning_rate": 4.1780957377566713e-07, "loss": 0.8392, "step": 17921 }, { "epoch": 0.91, "learning_rate": 4.173390562382407e-07, "loss": 0.7832, "step": 17922 }, { "epoch": 0.91, "learning_rate": 4.168687981411623e-07, "loss": 0.8657, "step": 17923 }, { "epoch": 0.91, "learning_rate": 4.1639879949716634e-07, "loss": 0.914, "step": 17924 }, { "epoch": 0.91, "learning_rate": 4.1592906031897696e-07, "loss": 0.8431, "step": 17925 }, { "epoch": 0.91, "learning_rate": 4.154595806193107e-07, "loss": 0.9574, "step": 17926 }, { "epoch": 0.91, "learning_rate": 4.149903604108807e-07, "loss": 0.8228, "step": 17927 }, { "epoch": 0.91, "learning_rate": 4.145213997063879e-07, "loss": 0.8448, "step": 17928 }, { "epoch": 0.91, "learning_rate": 4.1405269851853004e-07, "loss": 0.9021, "step": 17929 }, { "epoch": 0.91, "learning_rate": 4.135842568599968e-07, "loss": 0.8519, "step": 17930 }, { "epoch": 0.91, "learning_rate": 4.131160747434715e-07, "loss": 0.8247, "step": 17931 }, { "epoch": 0.91, "learning_rate": 4.126481521816284e-07, "loss": 0.9057, "step": 17932 }, { "epoch": 0.91, "learning_rate": 4.1218048918713616e-07, "loss": 0.913, "step": 17933 }, { "epoch": 0.91, "learning_rate": 4.11713085772657e-07, "loss": 0.9039, "step": 17934 }, { "epoch": 0.91, "learning_rate": 4.11245941950843e-07, "loss": 0.9049, "step": 17935 }, { "epoch": 0.91, "learning_rate": 4.107790577343462e-07, "loss": 0.8605, "step": 17936 }, { "epoch": 0.91, "learning_rate": 4.1031243313580327e-07, "loss": 0.8693, "step": 17937 }, { "epoch": 0.91, "learning_rate": 4.098460681678473e-07, "loss": 0.8315, "step": 17938 }, { "epoch": 0.91, "learning_rate": 4.093799628431083e-07, "loss": 0.8928, "step": 17939 }, { "epoch": 0.91, "learning_rate": 4.0891411717420173e-07, "loss": 0.8374, "step": 17940 }, { "epoch": 0.91, "learning_rate": 4.0844853117374075e-07, "loss": 0.8649, "step": 17941 }, { "epoch": 0.91, "learning_rate": 4.0798320485433194e-07, "loss": 0.8117, "step": 17942 }, { "epoch": 0.91, "learning_rate": 4.0751813822857423e-07, "loss": 0.876, "step": 17943 }, { "epoch": 0.91, "learning_rate": 4.070533313090552e-07, "loss": 0.8756, "step": 17944 }, { "epoch": 0.91, "learning_rate": 4.0658878410836153e-07, "loss": 0.9016, "step": 17945 }, { "epoch": 0.91, "learning_rate": 4.0612449663907094e-07, "loss": 0.9169, "step": 17946 }, { "epoch": 0.91, "learning_rate": 4.0566046891374997e-07, "loss": 0.7693, "step": 17947 }, { "epoch": 0.91, "learning_rate": 4.0519670094496753e-07, "loss": 0.7754, "step": 17948 }, { "epoch": 0.91, "learning_rate": 4.047331927452747e-07, "loss": 0.9216, "step": 17949 }, { "epoch": 0.91, "learning_rate": 4.0426994432722134e-07, "loss": 0.8501, "step": 17950 }, { "epoch": 0.91, "learning_rate": 4.0380695570335306e-07, "loss": 0.8505, "step": 17951 }, { "epoch": 0.91, "learning_rate": 4.033442268861987e-07, "loss": 0.869, "step": 17952 }, { "epoch": 0.91, "learning_rate": 4.028817578882904e-07, "loss": 0.8537, "step": 17953 }, { "epoch": 0.91, "learning_rate": 4.024195487221483e-07, "loss": 0.8853, "step": 17954 }, { "epoch": 0.91, "learning_rate": 4.0195759940028665e-07, "loss": 0.9204, "step": 17955 }, { "epoch": 0.91, "learning_rate": 4.0149590993521005e-07, "loss": 0.8588, "step": 17956 }, { "epoch": 0.91, "learning_rate": 4.010344803394206e-07, "loss": 0.9534, "step": 17957 }, { "epoch": 0.91, "learning_rate": 4.005733106254106e-07, "loss": 0.8897, "step": 17958 }, { "epoch": 0.91, "learning_rate": 4.001124008056623e-07, "loss": 0.8677, "step": 17959 }, { "epoch": 0.91, "learning_rate": 3.996517508926612e-07, "loss": 0.8579, "step": 17960 }, { "epoch": 0.91, "learning_rate": 3.9919136089887287e-07, "loss": 0.8402, "step": 17961 }, { "epoch": 0.91, "learning_rate": 3.9873123083676303e-07, "loss": 0.8636, "step": 17962 }, { "epoch": 0.91, "learning_rate": 3.982713607187927e-07, "loss": 0.8982, "step": 17963 }, { "epoch": 0.91, "learning_rate": 3.978117505574086e-07, "loss": 0.8626, "step": 17964 }, { "epoch": 0.91, "learning_rate": 3.973524003650553e-07, "loss": 0.8702, "step": 17965 }, { "epoch": 0.91, "learning_rate": 3.968933101541683e-07, "loss": 0.8164, "step": 17966 }, { "epoch": 0.91, "learning_rate": 3.9643447993717997e-07, "loss": 0.8755, "step": 17967 }, { "epoch": 0.91, "learning_rate": 3.9597590972651034e-07, "loss": 0.8784, "step": 17968 }, { "epoch": 0.91, "learning_rate": 3.9551759953457394e-07, "loss": 0.7988, "step": 17969 }, { "epoch": 0.91, "learning_rate": 3.9505954937378187e-07, "loss": 0.8457, "step": 17970 }, { "epoch": 0.91, "learning_rate": 3.94601759256531e-07, "loss": 0.8757, "step": 17971 }, { "epoch": 0.91, "learning_rate": 3.9414422919522134e-07, "loss": 0.8558, "step": 17972 }, { "epoch": 0.91, "learning_rate": 3.936869592022341e-07, "loss": 0.7702, "step": 17973 }, { "epoch": 0.91, "learning_rate": 3.9322994928995384e-07, "loss": 0.8976, "step": 17974 }, { "epoch": 0.91, "learning_rate": 3.9277319947075174e-07, "loss": 0.8826, "step": 17975 }, { "epoch": 0.91, "learning_rate": 3.9231670975699354e-07, "loss": 0.9024, "step": 17976 }, { "epoch": 0.91, "learning_rate": 3.918604801610382e-07, "loss": 0.9827, "step": 17977 }, { "epoch": 0.91, "learning_rate": 3.914045106952391e-07, "loss": 0.8756, "step": 17978 }, { "epoch": 0.91, "learning_rate": 3.909488013719398e-07, "loss": 0.7619, "step": 17979 }, { "epoch": 0.91, "learning_rate": 3.9049335220347815e-07, "loss": 0.8861, "step": 17980 }, { "epoch": 0.91, "learning_rate": 3.9003816320218546e-07, "loss": 0.927, "step": 17981 }, { "epoch": 0.91, "learning_rate": 3.8958323438038625e-07, "loss": 0.9294, "step": 17982 }, { "epoch": 0.91, "learning_rate": 3.8912856575039403e-07, "loss": 0.9375, "step": 17983 }, { "epoch": 0.91, "learning_rate": 3.886741573245223e-07, "loss": 0.8433, "step": 17984 }, { "epoch": 0.91, "learning_rate": 3.882200091150723e-07, "loss": 0.981, "step": 17985 }, { "epoch": 0.91, "learning_rate": 3.877661211343386e-07, "loss": 0.8727, "step": 17986 }, { "epoch": 0.91, "learning_rate": 3.873124933946115e-07, "loss": 0.8961, "step": 17987 }, { "epoch": 0.91, "learning_rate": 3.8685912590817e-07, "loss": 0.9213, "step": 17988 }, { "epoch": 0.91, "learning_rate": 3.86406018687292e-07, "loss": 0.8287, "step": 17989 }, { "epoch": 0.91, "learning_rate": 3.8595317174424105e-07, "loss": 0.8938, "step": 17990 }, { "epoch": 0.91, "learning_rate": 3.855005850912807e-07, "loss": 0.8756, "step": 17991 }, { "epoch": 0.91, "learning_rate": 3.8504825874066345e-07, "loss": 0.777, "step": 17992 }, { "epoch": 0.91, "learning_rate": 3.845961927046349e-07, "loss": 0.8461, "step": 17993 }, { "epoch": 0.91, "learning_rate": 3.841443869954331e-07, "loss": 0.9523, "step": 17994 }, { "epoch": 0.91, "learning_rate": 3.8369284162529163e-07, "loss": 0.9507, "step": 17995 }, { "epoch": 0.91, "learning_rate": 3.832415566064374e-07, "loss": 0.8562, "step": 17996 }, { "epoch": 0.91, "learning_rate": 3.8279053195108604e-07, "loss": 0.783, "step": 17997 }, { "epoch": 0.91, "learning_rate": 3.8233976767144796e-07, "loss": 0.9324, "step": 17998 }, { "epoch": 0.91, "learning_rate": 3.8188926377973e-07, "loss": 0.8308, "step": 17999 }, { "epoch": 0.91, "learning_rate": 3.814390202881246e-07, "loss": 0.931, "step": 18000 }, { "epoch": 0.91, "learning_rate": 3.809890372088265e-07, "loss": 0.7722, "step": 18001 }, { "epoch": 0.91, "learning_rate": 3.805393145540159e-07, "loss": 0.969, "step": 18002 }, { "epoch": 0.91, "learning_rate": 3.800898523358687e-07, "loss": 0.843, "step": 18003 }, { "epoch": 0.91, "learning_rate": 3.796406505665551e-07, "loss": 0.827, "step": 18004 }, { "epoch": 0.91, "learning_rate": 3.7919170925823536e-07, "loss": 0.8556, "step": 18005 }, { "epoch": 0.91, "learning_rate": 3.787430284230631e-07, "loss": 0.8465, "step": 18006 }, { "epoch": 0.91, "learning_rate": 3.7829460807318753e-07, "loss": 0.8318, "step": 18007 }, { "epoch": 0.91, "learning_rate": 3.7784644822075e-07, "loss": 0.8302, "step": 18008 }, { "epoch": 0.91, "learning_rate": 3.7739854887788195e-07, "loss": 0.8362, "step": 18009 }, { "epoch": 0.91, "learning_rate": 3.7695091005671035e-07, "loss": 0.8553, "step": 18010 }, { "epoch": 0.92, "learning_rate": 3.765035317693555e-07, "loss": 0.8702, "step": 18011 }, { "epoch": 0.92, "learning_rate": 3.760564140279266e-07, "loss": 0.8169, "step": 18012 }, { "epoch": 0.92, "learning_rate": 3.756095568445339e-07, "loss": 0.8978, "step": 18013 }, { "epoch": 0.92, "learning_rate": 3.751629602312723e-07, "loss": 0.8644, "step": 18014 }, { "epoch": 0.92, "learning_rate": 3.74716624200232e-07, "loss": 0.9283, "step": 18015 }, { "epoch": 0.92, "learning_rate": 3.742705487635012e-07, "loss": 0.9166, "step": 18016 }, { "epoch": 0.92, "learning_rate": 3.7382473393315244e-07, "loss": 0.8434, "step": 18017 }, { "epoch": 0.92, "learning_rate": 3.7337917972125825e-07, "loss": 0.893, "step": 18018 }, { "epoch": 0.92, "learning_rate": 3.7293388613988014e-07, "loss": 0.871, "step": 18019 }, { "epoch": 0.92, "learning_rate": 3.724888532010762e-07, "loss": 0.7683, "step": 18020 }, { "epoch": 0.92, "learning_rate": 3.720440809168924e-07, "loss": 0.846, "step": 18021 }, { "epoch": 0.92, "learning_rate": 3.7159956929937124e-07, "loss": 0.8715, "step": 18022 }, { "epoch": 0.92, "learning_rate": 3.7115531836054985e-07, "loss": 0.8533, "step": 18023 }, { "epoch": 0.92, "learning_rate": 3.707113281124508e-07, "loss": 0.8493, "step": 18024 }, { "epoch": 0.92, "learning_rate": 3.7026759856710117e-07, "loss": 0.88, "step": 18025 }, { "epoch": 0.92, "learning_rate": 3.69824129736509e-07, "loss": 0.8939, "step": 18026 }, { "epoch": 0.92, "learning_rate": 3.693809216326827e-07, "loss": 0.8254, "step": 18027 }, { "epoch": 0.92, "learning_rate": 3.689379742676236e-07, "loss": 0.8858, "step": 18028 }, { "epoch": 0.92, "learning_rate": 3.68495287653321e-07, "loss": 0.8568, "step": 18029 }, { "epoch": 0.92, "learning_rate": 3.680528618017609e-07, "loss": 0.8324, "step": 18030 }, { "epoch": 0.92, "learning_rate": 3.6761069672492264e-07, "loss": 0.876, "step": 18031 }, { "epoch": 0.92, "learning_rate": 3.6716879243477776e-07, "loss": 0.8249, "step": 18032 }, { "epoch": 0.92, "learning_rate": 3.6672714894328774e-07, "loss": 0.8176, "step": 18033 }, { "epoch": 0.92, "learning_rate": 3.6628576626241195e-07, "loss": 0.9346, "step": 18034 }, { "epoch": 0.92, "learning_rate": 3.6584464440409974e-07, "loss": 0.9073, "step": 18035 }, { "epoch": 0.92, "learning_rate": 3.6540378338029147e-07, "loss": 0.8008, "step": 18036 }, { "epoch": 0.92, "learning_rate": 3.649631832029288e-07, "loss": 0.8307, "step": 18037 }, { "epoch": 0.92, "learning_rate": 3.6452284388393545e-07, "loss": 0.8437, "step": 18038 }, { "epoch": 0.92, "learning_rate": 3.6408276543523416e-07, "loss": 0.954, "step": 18039 }, { "epoch": 0.92, "learning_rate": 3.6364294786874197e-07, "loss": 0.8074, "step": 18040 }, { "epoch": 0.92, "learning_rate": 3.632033911963628e-07, "loss": 0.8615, "step": 18041 }, { "epoch": 0.92, "learning_rate": 3.6276409543000044e-07, "loss": 0.8421, "step": 18042 }, { "epoch": 0.92, "learning_rate": 3.6232506058154536e-07, "loss": 0.85, "step": 18043 }, { "epoch": 0.92, "learning_rate": 3.6188628666288803e-07, "loss": 0.8702, "step": 18044 }, { "epoch": 0.92, "learning_rate": 3.614477736859034e-07, "loss": 0.841, "step": 18045 }, { "epoch": 0.92, "learning_rate": 3.610095216624665e-07, "loss": 0.8846, "step": 18046 }, { "epoch": 0.92, "learning_rate": 3.605715306044422e-07, "loss": 0.8973, "step": 18047 }, { "epoch": 0.92, "learning_rate": 3.6013380052368543e-07, "loss": 0.8325, "step": 18048 }, { "epoch": 0.92, "learning_rate": 3.596963314320534e-07, "loss": 0.7409, "step": 18049 }, { "epoch": 0.92, "learning_rate": 3.5925912334138445e-07, "loss": 0.8711, "step": 18050 }, { "epoch": 0.92, "learning_rate": 3.5882217626351913e-07, "loss": 0.7551, "step": 18051 }, { "epoch": 0.92, "learning_rate": 3.583854902102857e-07, "loss": 0.885, "step": 18052 }, { "epoch": 0.92, "learning_rate": 3.579490651935058e-07, "loss": 0.8218, "step": 18053 }, { "epoch": 0.92, "learning_rate": 3.575129012249989e-07, "loss": 0.9175, "step": 18054 }, { "epoch": 0.92, "learning_rate": 3.570769983165701e-07, "loss": 0.8204, "step": 18055 }, { "epoch": 0.92, "learning_rate": 3.56641356480022e-07, "loss": 0.7844, "step": 18056 }, { "epoch": 0.92, "learning_rate": 3.562059757271508e-07, "loss": 0.7973, "step": 18057 }, { "epoch": 0.92, "learning_rate": 3.5577085606974257e-07, "loss": 0.9988, "step": 18058 }, { "epoch": 0.92, "learning_rate": 3.553359975195758e-07, "loss": 0.9318, "step": 18059 }, { "epoch": 0.92, "learning_rate": 3.549014000884277e-07, "loss": 0.8304, "step": 18060 }, { "epoch": 0.92, "learning_rate": 3.5446706378806205e-07, "loss": 0.8581, "step": 18061 }, { "epoch": 0.92, "learning_rate": 3.540329886302385e-07, "loss": 0.8421, "step": 18062 }, { "epoch": 0.92, "learning_rate": 3.535991746267087e-07, "loss": 0.8784, "step": 18063 }, { "epoch": 0.92, "learning_rate": 3.5316562178921985e-07, "loss": 0.8455, "step": 18064 }, { "epoch": 0.92, "learning_rate": 3.5273233012950603e-07, "loss": 0.8662, "step": 18065 }, { "epoch": 0.92, "learning_rate": 3.5229929965930334e-07, "loss": 0.8271, "step": 18066 }, { "epoch": 0.92, "learning_rate": 3.5186653039033015e-07, "loss": 0.8808, "step": 18067 }, { "epoch": 0.92, "learning_rate": 3.514340223343071e-07, "loss": 0.7707, "step": 18068 }, { "epoch": 0.92, "learning_rate": 3.510017755029438e-07, "loss": 0.8446, "step": 18069 }, { "epoch": 0.92, "learning_rate": 3.5056978990793967e-07, "loss": 0.822, "step": 18070 }, { "epoch": 0.92, "learning_rate": 3.5013806556099316e-07, "loss": 0.8542, "step": 18071 }, { "epoch": 0.92, "learning_rate": 3.497066024737905e-07, "loss": 0.816, "step": 18072 }, { "epoch": 0.92, "learning_rate": 3.492754006580168e-07, "loss": 0.9007, "step": 18073 }, { "epoch": 0.92, "learning_rate": 3.488444601253427e-07, "loss": 0.8996, "step": 18074 }, { "epoch": 0.92, "learning_rate": 3.484137808874366e-07, "loss": 0.7538, "step": 18075 }, { "epoch": 0.92, "learning_rate": 3.4798336295596037e-07, "loss": 0.8048, "step": 18076 }, { "epoch": 0.92, "learning_rate": 3.475532063425635e-07, "loss": 0.8578, "step": 18077 }, { "epoch": 0.92, "learning_rate": 3.4712331105889674e-07, "loss": 0.9139, "step": 18078 }, { "epoch": 0.92, "learning_rate": 3.4669367711659407e-07, "loss": 0.8443, "step": 18079 }, { "epoch": 0.92, "learning_rate": 3.4626430452729064e-07, "loss": 0.89, "step": 18080 }, { "epoch": 0.92, "learning_rate": 3.4583519330261164e-07, "loss": 0.7933, "step": 18081 }, { "epoch": 0.92, "learning_rate": 3.454063434541721e-07, "loss": 0.7976, "step": 18082 }, { "epoch": 0.92, "learning_rate": 3.44977754993584e-07, "loss": 0.931, "step": 18083 }, { "epoch": 0.92, "learning_rate": 3.445494279324513e-07, "loss": 0.7707, "step": 18084 }, { "epoch": 0.92, "learning_rate": 3.4412136228237135e-07, "loss": 0.9105, "step": 18085 }, { "epoch": 0.92, "learning_rate": 3.4369355805493164e-07, "loss": 0.8521, "step": 18086 }, { "epoch": 0.92, "learning_rate": 3.432660152617151e-07, "loss": 0.8309, "step": 18087 }, { "epoch": 0.92, "learning_rate": 3.42838733914298e-07, "loss": 0.9095, "step": 18088 }, { "epoch": 0.92, "learning_rate": 3.4241171402424667e-07, "loss": 0.8371, "step": 18089 }, { "epoch": 0.92, "learning_rate": 3.4198495560312406e-07, "loss": 0.8839, "step": 18090 }, { "epoch": 0.92, "learning_rate": 3.4155845866248317e-07, "loss": 0.7934, "step": 18091 }, { "epoch": 0.92, "learning_rate": 3.4113222321387143e-07, "loss": 0.7641, "step": 18092 }, { "epoch": 0.92, "learning_rate": 3.4070624926882955e-07, "loss": 0.8919, "step": 18093 }, { "epoch": 0.92, "learning_rate": 3.402805368388873e-07, "loss": 0.8982, "step": 18094 }, { "epoch": 0.92, "learning_rate": 3.3985508593557313e-07, "loss": 0.8777, "step": 18095 }, { "epoch": 0.92, "learning_rate": 3.394298965704046e-07, "loss": 0.8152, "step": 18096 }, { "epoch": 0.92, "learning_rate": 3.3900496875489575e-07, "loss": 0.9115, "step": 18097 }, { "epoch": 0.92, "learning_rate": 3.385803025005463e-07, "loss": 0.831, "step": 18098 }, { "epoch": 0.92, "learning_rate": 3.3815589781885703e-07, "loss": 0.8627, "step": 18099 }, { "epoch": 0.92, "learning_rate": 3.377317547213188e-07, "loss": 0.8208, "step": 18100 }, { "epoch": 0.92, "learning_rate": 3.373078732194102e-07, "loss": 0.847, "step": 18101 }, { "epoch": 0.92, "learning_rate": 3.3688425332461305e-07, "loss": 0.8461, "step": 18102 }, { "epoch": 0.92, "learning_rate": 3.3646089504839385e-07, "loss": 0.8016, "step": 18103 }, { "epoch": 0.92, "learning_rate": 3.360377984022145e-07, "loss": 0.8934, "step": 18104 }, { "epoch": 0.92, "learning_rate": 3.356149633975303e-07, "loss": 0.8653, "step": 18105 }, { "epoch": 0.92, "learning_rate": 3.3519239004578876e-07, "loss": 0.8372, "step": 18106 }, { "epoch": 0.92, "learning_rate": 3.3477007835842955e-07, "loss": 0.8794, "step": 18107 }, { "epoch": 0.92, "learning_rate": 3.343480283468892e-07, "loss": 0.8253, "step": 18108 }, { "epoch": 0.92, "learning_rate": 3.339262400225918e-07, "loss": 0.7915, "step": 18109 }, { "epoch": 0.92, "learning_rate": 3.3350471339695713e-07, "loss": 0.8891, "step": 18110 }, { "epoch": 0.92, "learning_rate": 3.330834484813983e-07, "loss": 0.7859, "step": 18111 }, { "epoch": 0.92, "learning_rate": 3.326624452873217e-07, "loss": 0.8369, "step": 18112 }, { "epoch": 0.92, "learning_rate": 3.3224170382612166e-07, "loss": 0.8321, "step": 18113 }, { "epoch": 0.92, "learning_rate": 3.3182122410919336e-07, "loss": 0.8342, "step": 18114 }, { "epoch": 0.92, "learning_rate": 3.314010061479189e-07, "loss": 0.8726, "step": 18115 }, { "epoch": 0.92, "learning_rate": 3.3098104995367583e-07, "loss": 0.8414, "step": 18116 }, { "epoch": 0.92, "learning_rate": 3.3056135553783395e-07, "loss": 0.8262, "step": 18117 }, { "epoch": 0.92, "learning_rate": 3.3014192291175526e-07, "loss": 0.8456, "step": 18118 }, { "epoch": 0.92, "learning_rate": 3.2972275208679625e-07, "loss": 0.8147, "step": 18119 }, { "epoch": 0.92, "learning_rate": 3.2930384307430563e-07, "loss": 0.8912, "step": 18120 }, { "epoch": 0.92, "learning_rate": 3.2888519588562537e-07, "loss": 0.8524, "step": 18121 }, { "epoch": 0.92, "learning_rate": 3.2846681053208873e-07, "loss": 0.848, "step": 18122 }, { "epoch": 0.92, "learning_rate": 3.280486870250221e-07, "loss": 0.8857, "step": 18123 }, { "epoch": 0.92, "learning_rate": 3.2763082537574874e-07, "loss": 0.9145, "step": 18124 }, { "epoch": 0.92, "learning_rate": 3.2721322559557843e-07, "loss": 0.9198, "step": 18125 }, { "epoch": 0.92, "learning_rate": 3.2679588769582105e-07, "loss": 0.8453, "step": 18126 }, { "epoch": 0.92, "learning_rate": 3.263788116877731e-07, "loss": 0.808, "step": 18127 }, { "epoch": 0.92, "learning_rate": 3.259619975827255e-07, "loss": 0.8593, "step": 18128 }, { "epoch": 0.92, "learning_rate": 3.255454453919671e-07, "loss": 0.9273, "step": 18129 }, { "epoch": 0.92, "learning_rate": 3.251291551267699e-07, "loss": 0.8034, "step": 18130 }, { "epoch": 0.92, "learning_rate": 3.247131267984094e-07, "loss": 0.8928, "step": 18131 }, { "epoch": 0.92, "learning_rate": 3.242973604181465e-07, "loss": 0.8121, "step": 18132 }, { "epoch": 0.92, "learning_rate": 3.2388185599723896e-07, "loss": 0.8524, "step": 18133 }, { "epoch": 0.92, "learning_rate": 3.2346661354693555e-07, "loss": 0.7173, "step": 18134 }, { "epoch": 0.92, "learning_rate": 3.230516330784783e-07, "loss": 0.8394, "step": 18135 }, { "epoch": 0.92, "learning_rate": 3.2263691460310277e-07, "loss": 0.9291, "step": 18136 }, { "epoch": 0.92, "learning_rate": 3.2222245813203656e-07, "loss": 0.8065, "step": 18137 }, { "epoch": 0.92, "learning_rate": 3.2180826367650186e-07, "loss": 0.8752, "step": 18138 }, { "epoch": 0.92, "learning_rate": 3.213943312477108e-07, "loss": 0.9664, "step": 18139 }, { "epoch": 0.92, "learning_rate": 3.209806608568711e-07, "loss": 0.8409, "step": 18140 }, { "epoch": 0.92, "learning_rate": 3.205672525151837e-07, "loss": 0.8255, "step": 18141 }, { "epoch": 0.92, "learning_rate": 3.2015410623383645e-07, "loss": 0.9106, "step": 18142 }, { "epoch": 0.92, "learning_rate": 3.1974122202402145e-07, "loss": 0.8622, "step": 18143 }, { "epoch": 0.92, "learning_rate": 3.193285998969131e-07, "loss": 0.9399, "step": 18144 }, { "epoch": 0.92, "learning_rate": 3.1891623986368245e-07, "loss": 0.7801, "step": 18145 }, { "epoch": 0.92, "learning_rate": 3.1850414193549615e-07, "loss": 0.8667, "step": 18146 }, { "epoch": 0.92, "learning_rate": 3.180923061235086e-07, "loss": 0.9507, "step": 18147 }, { "epoch": 0.92, "learning_rate": 3.17680732438872e-07, "loss": 0.9304, "step": 18148 }, { "epoch": 0.92, "learning_rate": 3.172694208927274e-07, "loss": 0.7884, "step": 18149 }, { "epoch": 0.92, "learning_rate": 3.1685837149621367e-07, "loss": 0.8609, "step": 18150 }, { "epoch": 0.92, "learning_rate": 3.164475842604564e-07, "loss": 0.8423, "step": 18151 }, { "epoch": 0.92, "learning_rate": 3.160370591965778e-07, "loss": 0.9004, "step": 18152 }, { "epoch": 0.92, "learning_rate": 3.1562679631569336e-07, "loss": 0.7482, "step": 18153 }, { "epoch": 0.92, "learning_rate": 3.1521679562890874e-07, "loss": 0.9322, "step": 18154 }, { "epoch": 0.92, "learning_rate": 3.1480705714732716e-07, "loss": 0.8402, "step": 18155 }, { "epoch": 0.92, "learning_rate": 3.143975808820399e-07, "loss": 0.8603, "step": 18156 }, { "epoch": 0.92, "learning_rate": 3.1398836684413347e-07, "loss": 0.8486, "step": 18157 }, { "epoch": 0.92, "learning_rate": 3.1357941504468695e-07, "loss": 0.9315, "step": 18158 }, { "epoch": 0.92, "learning_rate": 3.1317072549477246e-07, "loss": 0.7785, "step": 18159 }, { "epoch": 0.92, "learning_rate": 3.1276229820545343e-07, "loss": 0.8939, "step": 18160 }, { "epoch": 0.92, "learning_rate": 3.123541331877888e-07, "loss": 0.9165, "step": 18161 }, { "epoch": 0.92, "learning_rate": 3.119462304528309e-07, "loss": 0.8334, "step": 18162 }, { "epoch": 0.92, "learning_rate": 3.115385900116208e-07, "loss": 0.8696, "step": 18163 }, { "epoch": 0.92, "learning_rate": 3.1113121187519526e-07, "loss": 0.9203, "step": 18164 }, { "epoch": 0.92, "learning_rate": 3.107240960545843e-07, "loss": 0.8727, "step": 18165 }, { "epoch": 0.92, "learning_rate": 3.103172425608092e-07, "loss": 0.8597, "step": 18166 }, { "epoch": 0.92, "learning_rate": 3.0991065140488665e-07, "loss": 0.9159, "step": 18167 }, { "epoch": 0.92, "learning_rate": 3.095043225978234e-07, "loss": 0.8535, "step": 18168 }, { "epoch": 0.92, "learning_rate": 3.090982561506206e-07, "loss": 0.8723, "step": 18169 }, { "epoch": 0.92, "learning_rate": 3.0869245207427287e-07, "loss": 0.8842, "step": 18170 }, { "epoch": 0.92, "learning_rate": 3.082869103797659e-07, "loss": 0.8032, "step": 18171 }, { "epoch": 0.92, "learning_rate": 3.0788163107807856e-07, "loss": 0.8975, "step": 18172 }, { "epoch": 0.92, "learning_rate": 3.074766141801855e-07, "loss": 0.8405, "step": 18173 }, { "epoch": 0.92, "learning_rate": 3.0707185969705124e-07, "loss": 0.8221, "step": 18174 }, { "epoch": 0.92, "learning_rate": 3.0666736763963366e-07, "loss": 1.0212, "step": 18175 }, { "epoch": 0.92, "learning_rate": 3.0626313801888406e-07, "loss": 0.8568, "step": 18176 }, { "epoch": 0.92, "learning_rate": 3.058591708457481e-07, "loss": 0.8035, "step": 18177 }, { "epoch": 0.92, "learning_rate": 3.054554661311582e-07, "loss": 0.8651, "step": 18178 }, { "epoch": 0.92, "learning_rate": 3.050520238860499e-07, "loss": 0.8785, "step": 18179 }, { "epoch": 0.92, "learning_rate": 3.0464884412134245e-07, "loss": 0.8254, "step": 18180 }, { "epoch": 0.92, "learning_rate": 3.042459268479525e-07, "loss": 0.914, "step": 18181 }, { "epoch": 0.92, "learning_rate": 3.038432720767903e-07, "loss": 0.8077, "step": 18182 }, { "epoch": 0.92, "learning_rate": 3.0344087981875383e-07, "loss": 0.7588, "step": 18183 }, { "epoch": 0.92, "learning_rate": 3.0303875008473983e-07, "loss": 0.9537, "step": 18184 }, { "epoch": 0.92, "learning_rate": 3.0263688288563407e-07, "loss": 0.8476, "step": 18185 }, { "epoch": 0.92, "learning_rate": 3.02235278232319e-07, "loss": 0.9202, "step": 18186 }, { "epoch": 0.92, "learning_rate": 3.018339361356648e-07, "loss": 0.8831, "step": 18187 }, { "epoch": 0.92, "learning_rate": 3.0143285660653944e-07, "loss": 0.882, "step": 18188 }, { "epoch": 0.92, "learning_rate": 3.0103203965580083e-07, "loss": 0.7562, "step": 18189 }, { "epoch": 0.92, "learning_rate": 3.0063148529430043e-07, "loss": 0.9617, "step": 18190 }, { "epoch": 0.92, "learning_rate": 3.0023119353288385e-07, "loss": 0.8741, "step": 18191 }, { "epoch": 0.92, "learning_rate": 2.9983116438238815e-07, "loss": 0.8788, "step": 18192 }, { "epoch": 0.92, "learning_rate": 2.994313978536423e-07, "loss": 0.8219, "step": 18193 }, { "epoch": 0.92, "learning_rate": 2.990318939574721e-07, "loss": 0.9249, "step": 18194 }, { "epoch": 0.92, "learning_rate": 2.9863265270469124e-07, "loss": 0.8258, "step": 18195 }, { "epoch": 0.92, "learning_rate": 2.982336741061109e-07, "loss": 0.837, "step": 18196 }, { "epoch": 0.92, "learning_rate": 2.978349581725315e-07, "loss": 0.9112, "step": 18197 }, { "epoch": 0.92, "learning_rate": 2.9743650491474764e-07, "loss": 0.801, "step": 18198 }, { "epoch": 0.92, "learning_rate": 2.9703831434354957e-07, "loss": 0.8702, "step": 18199 }, { "epoch": 0.92, "learning_rate": 2.9664038646971424e-07, "loss": 0.8145, "step": 18200 }, { "epoch": 0.92, "learning_rate": 2.962427213040175e-07, "loss": 0.9052, "step": 18201 }, { "epoch": 0.92, "learning_rate": 2.9584531885722524e-07, "loss": 0.9432, "step": 18202 }, { "epoch": 0.92, "learning_rate": 2.954481791400976e-07, "loss": 0.8443, "step": 18203 }, { "epoch": 0.92, "learning_rate": 2.9505130216338386e-07, "loss": 0.7422, "step": 18204 }, { "epoch": 0.92, "learning_rate": 2.946546879378309e-07, "loss": 0.8705, "step": 18205 }, { "epoch": 0.92, "learning_rate": 2.9425833647417803e-07, "loss": 0.818, "step": 18206 }, { "epoch": 0.92, "learning_rate": 2.938622477831532e-07, "loss": 0.8954, "step": 18207 }, { "epoch": 0.93, "learning_rate": 2.9346642187548233e-07, "loss": 0.8986, "step": 18208 }, { "epoch": 0.93, "learning_rate": 2.9307085876188025e-07, "loss": 0.9148, "step": 18209 }, { "epoch": 0.93, "learning_rate": 2.9267555845305826e-07, "loss": 0.7813, "step": 18210 }, { "epoch": 0.93, "learning_rate": 2.922805209597179e-07, "loss": 0.7696, "step": 18211 }, { "epoch": 0.93, "learning_rate": 2.9188574629255393e-07, "loss": 0.8864, "step": 18212 }, { "epoch": 0.93, "learning_rate": 2.9149123446225447e-07, "loss": 0.7992, "step": 18213 }, { "epoch": 0.93, "learning_rate": 2.910969854794998e-07, "loss": 0.9359, "step": 18214 }, { "epoch": 0.93, "learning_rate": 2.9070299935496705e-07, "loss": 0.8579, "step": 18215 }, { "epoch": 0.93, "learning_rate": 2.903092760993198e-07, "loss": 0.8229, "step": 18216 }, { "epoch": 0.93, "learning_rate": 2.899158157232174e-07, "loss": 0.8643, "step": 18217 }, { "epoch": 0.93, "learning_rate": 2.8952261823731564e-07, "loss": 0.8719, "step": 18218 }, { "epoch": 0.93, "learning_rate": 2.8912968365225503e-07, "loss": 0.8913, "step": 18219 }, { "epoch": 0.93, "learning_rate": 2.887370119786792e-07, "loss": 0.8508, "step": 18220 }, { "epoch": 0.93, "learning_rate": 2.8834460322721634e-07, "loss": 0.9355, "step": 18221 }, { "epoch": 0.93, "learning_rate": 2.879524574084902e-07, "loss": 0.8292, "step": 18222 }, { "epoch": 0.93, "learning_rate": 2.8756057453312003e-07, "loss": 0.8481, "step": 18223 }, { "epoch": 0.93, "learning_rate": 2.871689546117129e-07, "loss": 0.9714, "step": 18224 }, { "epoch": 0.93, "learning_rate": 2.8677759765487256e-07, "loss": 0.7986, "step": 18225 }, { "epoch": 0.93, "learning_rate": 2.8638650367319387e-07, "loss": 0.8023, "step": 18226 }, { "epoch": 0.93, "learning_rate": 2.8599567267726834e-07, "loss": 0.8106, "step": 18227 }, { "epoch": 0.93, "learning_rate": 2.856051046776731e-07, "loss": 0.9395, "step": 18228 }, { "epoch": 0.93, "learning_rate": 2.852147996849852e-07, "loss": 0.9165, "step": 18229 }, { "epoch": 0.93, "learning_rate": 2.848247577097718e-07, "loss": 0.8625, "step": 18230 }, { "epoch": 0.93, "learning_rate": 2.8443497876259e-07, "loss": 0.8902, "step": 18231 }, { "epoch": 0.93, "learning_rate": 2.840454628539957e-07, "loss": 0.8586, "step": 18232 }, { "epoch": 0.93, "learning_rate": 2.8365620999453283e-07, "loss": 0.8335, "step": 18233 }, { "epoch": 0.93, "learning_rate": 2.832672201947417e-07, "loss": 0.8851, "step": 18234 }, { "epoch": 0.93, "learning_rate": 2.8287849346515295e-07, "loss": 0.9057, "step": 18235 }, { "epoch": 0.93, "learning_rate": 2.8249002981629024e-07, "loss": 0.8213, "step": 18236 }, { "epoch": 0.93, "learning_rate": 2.821018292586708e-07, "loss": 0.8799, "step": 18237 }, { "epoch": 0.93, "learning_rate": 2.8171389180280505e-07, "loss": 0.7473, "step": 18238 }, { "epoch": 0.93, "learning_rate": 2.81326217459198e-07, "loss": 0.8868, "step": 18239 }, { "epoch": 0.93, "learning_rate": 2.8093880623834224e-07, "loss": 0.836, "step": 18240 }, { "epoch": 0.93, "learning_rate": 2.805516581507284e-07, "loss": 0.8687, "step": 18241 }, { "epoch": 0.93, "learning_rate": 2.801647732068391e-07, "loss": 0.8459, "step": 18242 }, { "epoch": 0.93, "learning_rate": 2.7977815141714493e-07, "loss": 0.8234, "step": 18243 }, { "epoch": 0.93, "learning_rate": 2.793917927921175e-07, "loss": 0.935, "step": 18244 }, { "epoch": 0.93, "learning_rate": 2.7900569734221615e-07, "loss": 0.8625, "step": 18245 }, { "epoch": 0.93, "learning_rate": 2.786198650778904e-07, "loss": 0.8944, "step": 18246 }, { "epoch": 0.93, "learning_rate": 2.7823429600959184e-07, "loss": 0.9099, "step": 18247 }, { "epoch": 0.93, "learning_rate": 2.778489901477555e-07, "loss": 0.8365, "step": 18248 }, { "epoch": 0.93, "learning_rate": 2.77463947502814e-07, "loss": 0.8417, "step": 18249 }, { "epoch": 0.93, "learning_rate": 2.7707916808519144e-07, "loss": 0.9008, "step": 18250 }, { "epoch": 0.93, "learning_rate": 2.766946519053082e-07, "loss": 0.8717, "step": 18251 }, { "epoch": 0.93, "learning_rate": 2.763103989735705e-07, "loss": 0.8945, "step": 18252 }, { "epoch": 0.93, "learning_rate": 2.759264093003844e-07, "loss": 0.8761, "step": 18253 }, { "epoch": 0.93, "learning_rate": 2.7554268289614607e-07, "loss": 0.8255, "step": 18254 }, { "epoch": 0.93, "learning_rate": 2.751592197712405e-07, "loss": 0.7721, "step": 18255 }, { "epoch": 0.93, "learning_rate": 2.7477601993605604e-07, "loss": 0.9471, "step": 18256 }, { "epoch": 0.93, "learning_rate": 2.743930834009634e-07, "loss": 0.8726, "step": 18257 }, { "epoch": 0.93, "learning_rate": 2.7401041017632855e-07, "loss": 0.792, "step": 18258 }, { "epoch": 0.93, "learning_rate": 2.736280002725167e-07, "loss": 0.9196, "step": 18259 }, { "epoch": 0.93, "learning_rate": 2.732458536998772e-07, "loss": 0.8857, "step": 18260 }, { "epoch": 0.93, "learning_rate": 2.7286397046875854e-07, "loss": 0.7298, "step": 18261 }, { "epoch": 0.93, "learning_rate": 2.72482350589498e-07, "loss": 0.8087, "step": 18262 }, { "epoch": 0.93, "learning_rate": 2.7210099407242954e-07, "loss": 0.8279, "step": 18263 }, { "epoch": 0.93, "learning_rate": 2.71719900927877e-07, "loss": 0.8958, "step": 18264 }, { "epoch": 0.93, "learning_rate": 2.7133907116615786e-07, "loss": 0.8079, "step": 18265 }, { "epoch": 0.93, "learning_rate": 2.709585047975827e-07, "loss": 0.9097, "step": 18266 }, { "epoch": 0.93, "learning_rate": 2.705782018324554e-07, "loss": 0.8076, "step": 18267 }, { "epoch": 0.93, "learning_rate": 2.701981622810723e-07, "loss": 0.8124, "step": 18268 }, { "epoch": 0.93, "learning_rate": 2.698183861537218e-07, "loss": 0.8225, "step": 18269 }, { "epoch": 0.93, "learning_rate": 2.6943887346068565e-07, "loss": 0.7689, "step": 18270 }, { "epoch": 0.93, "learning_rate": 2.6905962421224117e-07, "loss": 0.9236, "step": 18271 }, { "epoch": 0.93, "learning_rate": 2.686806384186524e-07, "loss": 0.8114, "step": 18272 }, { "epoch": 0.93, "learning_rate": 2.683019160901845e-07, "loss": 0.8776, "step": 18273 }, { "epoch": 0.93, "learning_rate": 2.6792345723708704e-07, "loss": 0.8156, "step": 18274 }, { "epoch": 0.93, "learning_rate": 2.675452618696073e-07, "loss": 0.8848, "step": 18275 }, { "epoch": 0.93, "learning_rate": 2.671673299979871e-07, "loss": 0.878, "step": 18276 }, { "epoch": 0.93, "learning_rate": 2.66789661632455e-07, "loss": 0.8359, "step": 18277 }, { "epoch": 0.93, "learning_rate": 2.664122567832372e-07, "loss": 0.8254, "step": 18278 }, { "epoch": 0.93, "learning_rate": 2.660351154605523e-07, "loss": 0.8221, "step": 18279 }, { "epoch": 0.93, "learning_rate": 2.6565823767461084e-07, "loss": 0.8007, "step": 18280 }, { "epoch": 0.93, "learning_rate": 2.6528162343561593e-07, "loss": 0.816, "step": 18281 }, { "epoch": 0.93, "learning_rate": 2.649052727537638e-07, "loss": 0.8778, "step": 18282 }, { "epoch": 0.93, "learning_rate": 2.6452918563924513e-07, "loss": 0.8378, "step": 18283 }, { "epoch": 0.93, "learning_rate": 2.6415336210223853e-07, "loss": 0.8244, "step": 18284 }, { "epoch": 0.93, "learning_rate": 2.637778021529247e-07, "loss": 0.8288, "step": 18285 }, { "epoch": 0.93, "learning_rate": 2.634025058014666e-07, "loss": 0.8092, "step": 18286 }, { "epoch": 0.93, "learning_rate": 2.6302747305802624e-07, "loss": 0.8246, "step": 18287 }, { "epoch": 0.93, "learning_rate": 2.6265270393275977e-07, "loss": 0.8204, "step": 18288 }, { "epoch": 0.93, "learning_rate": 2.6227819843580916e-07, "loss": 0.9011, "step": 18289 }, { "epoch": 0.93, "learning_rate": 2.6190395657731737e-07, "loss": 0.8615, "step": 18290 }, { "epoch": 0.93, "learning_rate": 2.615299783674152e-07, "loss": 0.9155, "step": 18291 }, { "epoch": 0.93, "learning_rate": 2.6115626381622905e-07, "loss": 0.9087, "step": 18292 }, { "epoch": 0.93, "learning_rate": 2.607828129338752e-07, "loss": 0.8517, "step": 18293 }, { "epoch": 0.93, "learning_rate": 2.604096257304645e-07, "loss": 0.847, "step": 18294 }, { "epoch": 0.93, "learning_rate": 2.60036702216101e-07, "loss": 0.9006, "step": 18295 }, { "epoch": 0.93, "learning_rate": 2.5966404240088114e-07, "loss": 0.9936, "step": 18296 }, { "epoch": 0.93, "learning_rate": 2.5929164629489466e-07, "loss": 0.9041, "step": 18297 }, { "epoch": 0.93, "learning_rate": 2.5891951390822454e-07, "loss": 0.8587, "step": 18298 }, { "epoch": 0.93, "learning_rate": 2.585476452509428e-07, "loss": 0.8827, "step": 18299 }, { "epoch": 0.93, "learning_rate": 2.5817604033312237e-07, "loss": 0.8994, "step": 18300 }, { "epoch": 0.93, "learning_rate": 2.5780469916481863e-07, "loss": 0.8513, "step": 18301 }, { "epoch": 0.93, "learning_rate": 2.574336217560891e-07, "loss": 0.9518, "step": 18302 }, { "epoch": 0.93, "learning_rate": 2.570628081169779e-07, "loss": 0.8245, "step": 18303 }, { "epoch": 0.93, "learning_rate": 2.5669225825752707e-07, "loss": 0.7749, "step": 18304 }, { "epoch": 0.93, "learning_rate": 2.563219721877663e-07, "loss": 0.789, "step": 18305 }, { "epoch": 0.93, "learning_rate": 2.55951949917721e-07, "loss": 0.8599, "step": 18306 }, { "epoch": 0.93, "learning_rate": 2.555821914574119e-07, "loss": 0.8231, "step": 18307 }, { "epoch": 0.93, "learning_rate": 2.552126968168445e-07, "loss": 0.9694, "step": 18308 }, { "epoch": 0.93, "learning_rate": 2.548434660060284e-07, "loss": 0.9465, "step": 18309 }, { "epoch": 0.93, "learning_rate": 2.544744990349579e-07, "loss": 0.9836, "step": 18310 }, { "epoch": 0.93, "learning_rate": 2.541057959136184e-07, "loss": 0.872, "step": 18311 }, { "epoch": 0.93, "learning_rate": 2.537373566519996e-07, "loss": 0.8803, "step": 18312 }, { "epoch": 0.93, "learning_rate": 2.533691812600703e-07, "loss": 0.9339, "step": 18313 }, { "epoch": 0.93, "learning_rate": 2.5300126974780124e-07, "loss": 0.7892, "step": 18314 }, { "epoch": 0.93, "learning_rate": 2.5263362212515353e-07, "loss": 0.9324, "step": 18315 }, { "epoch": 0.93, "learning_rate": 2.522662384020791e-07, "loss": 0.8837, "step": 18316 }, { "epoch": 0.93, "learning_rate": 2.5189911858852665e-07, "loss": 0.854, "step": 18317 }, { "epoch": 0.93, "learning_rate": 2.5153226269443274e-07, "loss": 0.8542, "step": 18318 }, { "epoch": 0.93, "learning_rate": 2.511656707297339e-07, "loss": 0.8436, "step": 18319 }, { "epoch": 0.93, "learning_rate": 2.507993427043487e-07, "loss": 0.8774, "step": 18320 }, { "epoch": 0.93, "learning_rate": 2.5043327862820265e-07, "loss": 0.7996, "step": 18321 }, { "epoch": 0.93, "learning_rate": 2.500674785112023e-07, "loss": 0.8441, "step": 18322 }, { "epoch": 0.93, "learning_rate": 2.497019423632496e-07, "loss": 0.8527, "step": 18323 }, { "epoch": 0.93, "learning_rate": 2.493366701942457e-07, "loss": 0.8611, "step": 18324 }, { "epoch": 0.93, "learning_rate": 2.4897166201407587e-07, "loss": 0.8112, "step": 18325 }, { "epoch": 0.93, "learning_rate": 2.486069178326245e-07, "loss": 0.8032, "step": 18326 }, { "epoch": 0.93, "learning_rate": 2.48242437659767e-07, "loss": 0.7904, "step": 18327 }, { "epoch": 0.93, "learning_rate": 2.478782215053699e-07, "loss": 0.8673, "step": 18328 }, { "epoch": 0.93, "learning_rate": 2.4751426937929426e-07, "loss": 0.8866, "step": 18329 }, { "epoch": 0.93, "learning_rate": 2.4715058129139323e-07, "loss": 0.9084, "step": 18330 }, { "epoch": 0.93, "learning_rate": 2.467871572515146e-07, "loss": 0.7671, "step": 18331 }, { "epoch": 0.93, "learning_rate": 2.464239972694982e-07, "loss": 1.0066, "step": 18332 }, { "epoch": 0.93, "learning_rate": 2.4606110135517504e-07, "loss": 0.7356, "step": 18333 }, { "epoch": 0.93, "learning_rate": 2.456984695183684e-07, "loss": 0.8309, "step": 18334 }, { "epoch": 0.93, "learning_rate": 2.4533610176889934e-07, "loss": 0.8434, "step": 18335 }, { "epoch": 0.93, "learning_rate": 2.449739981165766e-07, "loss": 0.8626, "step": 18336 }, { "epoch": 0.93, "learning_rate": 2.4461215857120356e-07, "loss": 0.8908, "step": 18337 }, { "epoch": 0.93, "learning_rate": 2.4425058314257897e-07, "loss": 0.8445, "step": 18338 }, { "epoch": 0.93, "learning_rate": 2.438892718404906e-07, "loss": 0.9647, "step": 18339 }, { "epoch": 0.93, "learning_rate": 2.435282246747195e-07, "loss": 0.8564, "step": 18340 }, { "epoch": 0.93, "learning_rate": 2.4316744165504224e-07, "loss": 0.8461, "step": 18341 }, { "epoch": 0.93, "learning_rate": 2.4280692279122554e-07, "loss": 0.8314, "step": 18342 }, { "epoch": 0.93, "learning_rate": 2.424466680930304e-07, "loss": 0.8082, "step": 18343 }, { "epoch": 0.93, "learning_rate": 2.4208667757021245e-07, "loss": 0.848, "step": 18344 }, { "epoch": 0.93, "learning_rate": 2.4172695123251376e-07, "loss": 0.8341, "step": 18345 }, { "epoch": 0.93, "learning_rate": 2.4136748908967664e-07, "loss": 0.8544, "step": 18346 }, { "epoch": 0.93, "learning_rate": 2.410082911514322e-07, "loss": 0.8952, "step": 18347 }, { "epoch": 0.93, "learning_rate": 2.4064935742750597e-07, "loss": 0.83, "step": 18348 }, { "epoch": 0.93, "learning_rate": 2.4029068792761457e-07, "loss": 0.829, "step": 18349 }, { "epoch": 0.93, "learning_rate": 2.399322826614703e-07, "loss": 0.8654, "step": 18350 }, { "epoch": 0.93, "learning_rate": 2.3957414163877535e-07, "loss": 0.8702, "step": 18351 }, { "epoch": 0.93, "learning_rate": 2.392162648692253e-07, "loss": 0.8521, "step": 18352 }, { "epoch": 0.93, "learning_rate": 2.388586523625114e-07, "loss": 0.8012, "step": 18353 }, { "epoch": 0.93, "learning_rate": 2.385013041283135e-07, "loss": 0.856, "step": 18354 }, { "epoch": 0.93, "learning_rate": 2.3814422017630734e-07, "loss": 0.8872, "step": 18355 }, { "epoch": 0.93, "learning_rate": 2.377874005161618e-07, "loss": 0.8232, "step": 18356 }, { "epoch": 0.93, "learning_rate": 2.374308451575358e-07, "loss": 0.7869, "step": 18357 }, { "epoch": 0.93, "learning_rate": 2.3707455411008272e-07, "loss": 0.9092, "step": 18358 }, { "epoch": 0.93, "learning_rate": 2.3671852738344824e-07, "loss": 0.7821, "step": 18359 }, { "epoch": 0.93, "learning_rate": 2.363627649872746e-07, "loss": 0.9053, "step": 18360 }, { "epoch": 0.93, "learning_rate": 2.3600726693118857e-07, "loss": 0.8814, "step": 18361 }, { "epoch": 0.93, "learning_rate": 2.356520332248202e-07, "loss": 0.8509, "step": 18362 }, { "epoch": 0.93, "learning_rate": 2.3529706387778406e-07, "loss": 0.9635, "step": 18363 }, { "epoch": 0.93, "learning_rate": 2.349423588996902e-07, "loss": 0.8096, "step": 18364 }, { "epoch": 0.93, "learning_rate": 2.3458791830014316e-07, "loss": 0.782, "step": 18365 }, { "epoch": 0.93, "learning_rate": 2.342337420887386e-07, "loss": 0.9345, "step": 18366 }, { "epoch": 0.93, "learning_rate": 2.3387983027506555e-07, "loss": 0.7871, "step": 18367 }, { "epoch": 0.93, "learning_rate": 2.3352618286870632e-07, "loss": 0.8731, "step": 18368 }, { "epoch": 0.93, "learning_rate": 2.3317279987923437e-07, "loss": 0.8224, "step": 18369 }, { "epoch": 0.93, "learning_rate": 2.3281968131621757e-07, "loss": 0.8672, "step": 18370 }, { "epoch": 0.93, "learning_rate": 2.324668271892172e-07, "loss": 0.804, "step": 18371 }, { "epoch": 0.93, "learning_rate": 2.321142375077856e-07, "loss": 0.8646, "step": 18372 }, { "epoch": 0.93, "learning_rate": 2.317619122814674e-07, "loss": 0.7627, "step": 18373 }, { "epoch": 0.93, "learning_rate": 2.314098515198049e-07, "loss": 0.8074, "step": 18374 }, { "epoch": 0.93, "learning_rate": 2.3105805523232715e-07, "loss": 0.8536, "step": 18375 }, { "epoch": 0.93, "learning_rate": 2.3070652342855772e-07, "loss": 0.9995, "step": 18376 }, { "epoch": 0.93, "learning_rate": 2.3035525611801668e-07, "loss": 0.9412, "step": 18377 }, { "epoch": 0.93, "learning_rate": 2.3000425331021316e-07, "loss": 0.8196, "step": 18378 }, { "epoch": 0.93, "learning_rate": 2.296535150146495e-07, "loss": 0.8363, "step": 18379 }, { "epoch": 0.93, "learning_rate": 2.293030412408226e-07, "loss": 0.7981, "step": 18380 }, { "epoch": 0.93, "learning_rate": 2.2895283199822038e-07, "loss": 0.8904, "step": 18381 }, { "epoch": 0.93, "learning_rate": 2.286028872963242e-07, "loss": 0.8729, "step": 18382 }, { "epoch": 0.93, "learning_rate": 2.2825320714460862e-07, "loss": 0.7891, "step": 18383 }, { "epoch": 0.93, "learning_rate": 2.2790379155254284e-07, "loss": 0.8645, "step": 18384 }, { "epoch": 0.93, "learning_rate": 2.2755464052958365e-07, "loss": 0.8282, "step": 18385 }, { "epoch": 0.93, "learning_rate": 2.2720575408518686e-07, "loss": 0.8345, "step": 18386 }, { "epoch": 0.93, "learning_rate": 2.2685713222879602e-07, "loss": 0.9307, "step": 18387 }, { "epoch": 0.93, "learning_rate": 2.2650877496984913e-07, "loss": 0.825, "step": 18388 }, { "epoch": 0.93, "learning_rate": 2.2616068231778198e-07, "loss": 0.8295, "step": 18389 }, { "epoch": 0.93, "learning_rate": 2.2581285428201373e-07, "loss": 0.8405, "step": 18390 }, { "epoch": 0.93, "learning_rate": 2.2546529087196344e-07, "loss": 0.8029, "step": 18391 }, { "epoch": 0.93, "learning_rate": 2.2511799209704254e-07, "loss": 0.8737, "step": 18392 }, { "epoch": 0.93, "learning_rate": 2.2477095796665128e-07, "loss": 0.8091, "step": 18393 }, { "epoch": 0.93, "learning_rate": 2.2442418849018653e-07, "loss": 0.8054, "step": 18394 }, { "epoch": 0.93, "learning_rate": 2.2407768367703642e-07, "loss": 0.8852, "step": 18395 }, { "epoch": 0.93, "learning_rate": 2.2373144353658227e-07, "loss": 1.0172, "step": 18396 }, { "epoch": 0.93, "learning_rate": 2.2338546807819773e-07, "loss": 0.8452, "step": 18397 }, { "epoch": 0.93, "learning_rate": 2.2303975731125082e-07, "loss": 0.7463, "step": 18398 }, { "epoch": 0.93, "learning_rate": 2.2269431124510077e-07, "loss": 0.9326, "step": 18399 }, { "epoch": 0.93, "learning_rate": 2.2234912988909895e-07, "loss": 0.8264, "step": 18400 }, { "epoch": 0.93, "learning_rate": 2.2200421325259347e-07, "loss": 0.805, "step": 18401 }, { "epoch": 0.93, "learning_rate": 2.2165956134491905e-07, "loss": 0.9089, "step": 18402 }, { "epoch": 0.93, "learning_rate": 2.2131517417540937e-07, "loss": 0.8557, "step": 18403 }, { "epoch": 0.93, "learning_rate": 2.2097105175338917e-07, "loss": 0.9266, "step": 18404 }, { "epoch": 0.94, "learning_rate": 2.2062719408817102e-07, "loss": 0.9814, "step": 18405 }, { "epoch": 0.94, "learning_rate": 2.2028360118906966e-07, "loss": 0.7433, "step": 18406 }, { "epoch": 0.94, "learning_rate": 2.1994027306538325e-07, "loss": 0.9066, "step": 18407 }, { "epoch": 0.94, "learning_rate": 2.1959720972640875e-07, "loss": 0.8502, "step": 18408 }, { "epoch": 0.94, "learning_rate": 2.1925441118143654e-07, "loss": 0.8474, "step": 18409 }, { "epoch": 0.94, "learning_rate": 2.1891187743974253e-07, "loss": 0.9294, "step": 18410 }, { "epoch": 0.94, "learning_rate": 2.185696085106048e-07, "loss": 0.8764, "step": 18411 }, { "epoch": 0.94, "learning_rate": 2.1822760440328716e-07, "loss": 0.8781, "step": 18412 }, { "epoch": 0.94, "learning_rate": 2.1788586512705212e-07, "loss": 0.8751, "step": 18413 }, { "epoch": 0.94, "learning_rate": 2.1754439069114785e-07, "loss": 0.9337, "step": 18414 }, { "epoch": 0.94, "learning_rate": 2.1720318110482254e-07, "loss": 0.9054, "step": 18415 }, { "epoch": 0.94, "learning_rate": 2.168622363773132e-07, "loss": 0.9375, "step": 18416 }, { "epoch": 0.94, "learning_rate": 2.16521556517848e-07, "loss": 0.849, "step": 18417 }, { "epoch": 0.94, "learning_rate": 2.1618114153565628e-07, "loss": 0.9245, "step": 18418 }, { "epoch": 0.94, "learning_rate": 2.158409914399484e-07, "loss": 0.8279, "step": 18419 }, { "epoch": 0.94, "learning_rate": 2.1550110623993702e-07, "loss": 0.8247, "step": 18420 }, { "epoch": 0.94, "learning_rate": 2.1516148594482477e-07, "loss": 0.8702, "step": 18421 }, { "epoch": 0.94, "learning_rate": 2.148221305638032e-07, "loss": 0.8286, "step": 18422 }, { "epoch": 0.94, "learning_rate": 2.144830401060627e-07, "loss": 0.843, "step": 18423 }, { "epoch": 0.94, "learning_rate": 2.141442145807815e-07, "loss": 0.9028, "step": 18424 }, { "epoch": 0.94, "learning_rate": 2.138056539971367e-07, "loss": 0.8749, "step": 18425 }, { "epoch": 0.94, "learning_rate": 2.13467358364291e-07, "loss": 0.8463, "step": 18426 }, { "epoch": 0.94, "learning_rate": 2.1312932769140372e-07, "loss": 0.9015, "step": 18427 }, { "epoch": 0.94, "learning_rate": 2.1279156198762862e-07, "loss": 0.8744, "step": 18428 }, { "epoch": 0.94, "learning_rate": 2.1245406126210732e-07, "loss": 0.9025, "step": 18429 }, { "epoch": 0.94, "learning_rate": 2.1211682552398138e-07, "loss": 0.8152, "step": 18430 }, { "epoch": 0.94, "learning_rate": 2.1177985478237795e-07, "loss": 0.8897, "step": 18431 }, { "epoch": 0.94, "learning_rate": 2.1144314904642194e-07, "loss": 0.8822, "step": 18432 }, { "epoch": 0.94, "learning_rate": 2.111067083252283e-07, "loss": 0.8661, "step": 18433 }, { "epoch": 0.94, "learning_rate": 2.1077053262790526e-07, "loss": 0.9129, "step": 18434 }, { "epoch": 0.94, "learning_rate": 2.1043462196355447e-07, "loss": 0.8306, "step": 18435 }, { "epoch": 0.94, "learning_rate": 2.1009897634127196e-07, "loss": 0.8841, "step": 18436 }, { "epoch": 0.94, "learning_rate": 2.0976359577014495e-07, "loss": 0.7956, "step": 18437 }, { "epoch": 0.94, "learning_rate": 2.0942848025925167e-07, "loss": 0.8185, "step": 18438 }, { "epoch": 0.94, "learning_rate": 2.090936298176649e-07, "loss": 0.8779, "step": 18439 }, { "epoch": 0.94, "learning_rate": 2.0875904445445295e-07, "loss": 0.886, "step": 18440 }, { "epoch": 0.94, "learning_rate": 2.0842472417867188e-07, "loss": 0.9501, "step": 18441 }, { "epoch": 0.94, "learning_rate": 2.0809066899937446e-07, "loss": 0.8682, "step": 18442 }, { "epoch": 0.94, "learning_rate": 2.0775687892560458e-07, "loss": 0.7596, "step": 18443 }, { "epoch": 0.94, "learning_rate": 2.0742335396639834e-07, "loss": 0.7933, "step": 18444 }, { "epoch": 0.94, "learning_rate": 2.070900941307874e-07, "loss": 0.799, "step": 18445 }, { "epoch": 0.94, "learning_rate": 2.0675709942779231e-07, "loss": 0.7886, "step": 18446 }, { "epoch": 0.94, "learning_rate": 2.0642436986643033e-07, "loss": 0.8629, "step": 18447 }, { "epoch": 0.94, "learning_rate": 2.0609190545570755e-07, "loss": 0.8405, "step": 18448 }, { "epoch": 0.94, "learning_rate": 2.0575970620462903e-07, "loss": 0.8642, "step": 18449 }, { "epoch": 0.94, "learning_rate": 2.0542777212218424e-07, "loss": 0.8582, "step": 18450 }, { "epoch": 0.94, "learning_rate": 2.050961032173626e-07, "loss": 0.8108, "step": 18451 }, { "epoch": 0.94, "learning_rate": 2.047646994991437e-07, "loss": 0.9045, "step": 18452 }, { "epoch": 0.94, "learning_rate": 2.044335609764969e-07, "loss": 0.8837, "step": 18453 }, { "epoch": 0.94, "learning_rate": 2.0410268765839182e-07, "loss": 0.8632, "step": 18454 }, { "epoch": 0.94, "learning_rate": 2.0377207955378342e-07, "loss": 0.7886, "step": 18455 }, { "epoch": 0.94, "learning_rate": 2.034417366716246e-07, "loss": 0.8197, "step": 18456 }, { "epoch": 0.94, "learning_rate": 2.0311165902085817e-07, "loss": 0.8193, "step": 18457 }, { "epoch": 0.94, "learning_rate": 2.027818466104192e-07, "loss": 0.7633, "step": 18458 }, { "epoch": 0.94, "learning_rate": 2.024522994492384e-07, "loss": 0.8583, "step": 18459 }, { "epoch": 0.94, "learning_rate": 2.0212301754623852e-07, "loss": 0.7495, "step": 18460 }, { "epoch": 0.94, "learning_rate": 2.0179400091033475e-07, "loss": 0.794, "step": 18461 }, { "epoch": 0.94, "learning_rate": 2.0146524955043212e-07, "loss": 0.7621, "step": 18462 }, { "epoch": 0.94, "learning_rate": 2.0113676347543353e-07, "loss": 0.9597, "step": 18463 }, { "epoch": 0.94, "learning_rate": 2.00808542694233e-07, "loss": 0.7772, "step": 18464 }, { "epoch": 0.94, "learning_rate": 2.004805872157134e-07, "loss": 0.8314, "step": 18465 }, { "epoch": 0.94, "learning_rate": 2.0015289704875874e-07, "loss": 0.836, "step": 18466 }, { "epoch": 0.94, "learning_rate": 1.9982547220223636e-07, "loss": 0.7841, "step": 18467 }, { "epoch": 0.94, "learning_rate": 1.9949831268501362e-07, "loss": 0.8128, "step": 18468 }, { "epoch": 0.94, "learning_rate": 1.991714185059468e-07, "loss": 0.9506, "step": 18469 }, { "epoch": 0.94, "learning_rate": 1.9884478967388542e-07, "loss": 0.795, "step": 18470 }, { "epoch": 0.94, "learning_rate": 1.985184261976758e-07, "loss": 0.8495, "step": 18471 }, { "epoch": 0.94, "learning_rate": 1.9819232808615085e-07, "loss": 0.9202, "step": 18472 }, { "epoch": 0.94, "learning_rate": 1.9786649534814018e-07, "loss": 0.7501, "step": 18473 }, { "epoch": 0.94, "learning_rate": 1.975409279924667e-07, "loss": 0.8976, "step": 18474 }, { "epoch": 0.94, "learning_rate": 1.972156260279423e-07, "loss": 0.8221, "step": 18475 }, { "epoch": 0.94, "learning_rate": 1.9689058946337547e-07, "loss": 0.8628, "step": 18476 }, { "epoch": 0.94, "learning_rate": 1.9656581830756583e-07, "loss": 0.9216, "step": 18477 }, { "epoch": 0.94, "learning_rate": 1.9624131256930856e-07, "loss": 0.9089, "step": 18478 }, { "epoch": 0.94, "learning_rate": 1.9591707225738554e-07, "loss": 0.7847, "step": 18479 }, { "epoch": 0.94, "learning_rate": 1.9559309738057752e-07, "loss": 0.8683, "step": 18480 }, { "epoch": 0.94, "learning_rate": 1.9526938794765639e-07, "loss": 0.9097, "step": 18481 }, { "epoch": 0.94, "learning_rate": 1.9494594396738398e-07, "loss": 0.7925, "step": 18482 }, { "epoch": 0.94, "learning_rate": 1.9462276544851888e-07, "loss": 0.8629, "step": 18483 }, { "epoch": 0.94, "learning_rate": 1.9429985239980964e-07, "loss": 0.8628, "step": 18484 }, { "epoch": 0.94, "learning_rate": 1.9397720483000037e-07, "loss": 0.8871, "step": 18485 }, { "epoch": 0.94, "learning_rate": 1.936548227478252e-07, "loss": 0.7458, "step": 18486 }, { "epoch": 0.94, "learning_rate": 1.9333270616201273e-07, "loss": 0.8292, "step": 18487 }, { "epoch": 0.94, "learning_rate": 1.930108550812826e-07, "loss": 0.9307, "step": 18488 }, { "epoch": 0.94, "learning_rate": 1.9268926951435118e-07, "loss": 0.7889, "step": 18489 }, { "epoch": 0.94, "learning_rate": 1.9236794946992376e-07, "loss": 0.8962, "step": 18490 }, { "epoch": 0.94, "learning_rate": 1.920468949566978e-07, "loss": 0.8434, "step": 18491 }, { "epoch": 0.94, "learning_rate": 1.917261059833686e-07, "loss": 0.8583, "step": 18492 }, { "epoch": 0.94, "learning_rate": 1.914055825586203e-07, "loss": 0.8527, "step": 18493 }, { "epoch": 0.94, "learning_rate": 1.9108532469112928e-07, "loss": 0.865, "step": 18494 }, { "epoch": 0.94, "learning_rate": 1.9076533238956751e-07, "loss": 0.8878, "step": 18495 }, { "epoch": 0.94, "learning_rate": 1.9044560566259918e-07, "loss": 0.9019, "step": 18496 }, { "epoch": 0.94, "learning_rate": 1.9012614451887844e-07, "loss": 0.8592, "step": 18497 }, { "epoch": 0.94, "learning_rate": 1.8980694896705622e-07, "loss": 0.8563, "step": 18498 }, { "epoch": 0.94, "learning_rate": 1.8948801901577218e-07, "loss": 0.7687, "step": 18499 }, { "epoch": 0.94, "learning_rate": 1.8916935467366394e-07, "loss": 0.8646, "step": 18500 }, { "epoch": 0.94, "learning_rate": 1.8885095594935677e-07, "loss": 0.8479, "step": 18501 }, { "epoch": 0.94, "learning_rate": 1.8853282285147267e-07, "loss": 0.879, "step": 18502 }, { "epoch": 0.94, "learning_rate": 1.8821495538862254e-07, "loss": 0.8331, "step": 18503 }, { "epoch": 0.94, "learning_rate": 1.8789735356941397e-07, "loss": 0.7964, "step": 18504 }, { "epoch": 0.94, "learning_rate": 1.8758001740244669e-07, "loss": 0.9626, "step": 18505 }, { "epoch": 0.94, "learning_rate": 1.872629468963083e-07, "loss": 0.7271, "step": 18506 }, { "epoch": 0.94, "learning_rate": 1.869461420595875e-07, "loss": 0.8058, "step": 18507 }, { "epoch": 0.94, "learning_rate": 1.866296029008585e-07, "loss": 0.8455, "step": 18508 }, { "epoch": 0.94, "learning_rate": 1.863133294286934e-07, "loss": 0.8719, "step": 18509 }, { "epoch": 0.94, "learning_rate": 1.8599732165165418e-07, "loss": 0.9221, "step": 18510 }, { "epoch": 0.94, "learning_rate": 1.8568157957829514e-07, "loss": 0.7758, "step": 18511 }, { "epoch": 0.94, "learning_rate": 1.853661032171661e-07, "loss": 0.923, "step": 18512 }, { "epoch": 0.94, "learning_rate": 1.8505089257680798e-07, "loss": 0.9125, "step": 18513 }, { "epoch": 0.94, "learning_rate": 1.847359476657551e-07, "loss": 0.7921, "step": 18514 }, { "epoch": 0.94, "learning_rate": 1.8442126849253282e-07, "loss": 0.8926, "step": 18515 }, { "epoch": 0.94, "learning_rate": 1.8410685506566218e-07, "loss": 0.9442, "step": 18516 }, { "epoch": 0.94, "learning_rate": 1.837927073936563e-07, "loss": 0.8329, "step": 18517 }, { "epoch": 0.94, "learning_rate": 1.8347882548501728e-07, "loss": 0.8346, "step": 18518 }, { "epoch": 0.94, "learning_rate": 1.8316520934824722e-07, "loss": 0.8824, "step": 18519 }, { "epoch": 0.94, "learning_rate": 1.8285185899183266e-07, "loss": 0.868, "step": 18520 }, { "epoch": 0.94, "learning_rate": 1.8253877442426016e-07, "loss": 0.9391, "step": 18521 }, { "epoch": 0.94, "learning_rate": 1.8222595565400624e-07, "loss": 0.8568, "step": 18522 }, { "epoch": 0.94, "learning_rate": 1.819134026895386e-07, "loss": 0.7511, "step": 18523 }, { "epoch": 0.94, "learning_rate": 1.8160111553932047e-07, "loss": 0.8653, "step": 18524 }, { "epoch": 0.94, "learning_rate": 1.8128909421180506e-07, "loss": 0.8318, "step": 18525 }, { "epoch": 0.94, "learning_rate": 1.8097733871544233e-07, "loss": 0.8626, "step": 18526 }, { "epoch": 0.94, "learning_rate": 1.8066584905867103e-07, "loss": 0.793, "step": 18527 }, { "epoch": 0.94, "learning_rate": 1.8035462524992442e-07, "loss": 0.9209, "step": 18528 }, { "epoch": 0.94, "learning_rate": 1.8004366729763023e-07, "loss": 0.775, "step": 18529 }, { "epoch": 0.94, "learning_rate": 1.7973297521020393e-07, "loss": 0.8222, "step": 18530 }, { "epoch": 0.94, "learning_rate": 1.7942254899606215e-07, "loss": 0.8633, "step": 18531 }, { "epoch": 0.94, "learning_rate": 1.7911238866360481e-07, "loss": 0.8726, "step": 18532 }, { "epoch": 0.94, "learning_rate": 1.7880249422123186e-07, "loss": 0.7994, "step": 18533 }, { "epoch": 0.94, "learning_rate": 1.7849286567733214e-07, "loss": 0.8549, "step": 18534 }, { "epoch": 0.94, "learning_rate": 1.7818350304028896e-07, "loss": 0.8203, "step": 18535 }, { "epoch": 0.94, "learning_rate": 1.7787440631847675e-07, "loss": 0.9043, "step": 18536 }, { "epoch": 0.94, "learning_rate": 1.7756557552026655e-07, "loss": 0.8163, "step": 18537 }, { "epoch": 0.94, "learning_rate": 1.7725701065401724e-07, "loss": 0.8005, "step": 18538 }, { "epoch": 0.94, "learning_rate": 1.7694871172808436e-07, "loss": 0.7359, "step": 18539 }, { "epoch": 0.94, "learning_rate": 1.7664067875081347e-07, "loss": 0.9198, "step": 18540 }, { "epoch": 0.94, "learning_rate": 1.7633291173054568e-07, "loss": 0.8836, "step": 18541 }, { "epoch": 0.94, "learning_rate": 1.7602541067561208e-07, "loss": 0.8469, "step": 18542 }, { "epoch": 0.94, "learning_rate": 1.7571817559433934e-07, "loss": 0.8385, "step": 18543 }, { "epoch": 0.94, "learning_rate": 1.7541120649504419e-07, "loss": 0.8341, "step": 18544 }, { "epoch": 0.94, "learning_rate": 1.751045033860388e-07, "loss": 0.8707, "step": 18545 }, { "epoch": 0.94, "learning_rate": 1.7479806627562545e-07, "loss": 0.8708, "step": 18546 }, { "epoch": 0.94, "learning_rate": 1.7449189517210086e-07, "loss": 0.9124, "step": 18547 }, { "epoch": 0.94, "learning_rate": 1.7418599008375614e-07, "loss": 0.8983, "step": 18548 }, { "epoch": 0.94, "learning_rate": 1.7388035101887025e-07, "loss": 0.7557, "step": 18549 }, { "epoch": 0.94, "learning_rate": 1.7357497798572097e-07, "loss": 0.9344, "step": 18550 }, { "epoch": 0.94, "learning_rate": 1.732698709925751e-07, "loss": 0.8267, "step": 18551 }, { "epoch": 0.94, "learning_rate": 1.729650300476915e-07, "loss": 0.8629, "step": 18552 }, { "epoch": 0.94, "learning_rate": 1.7266045515932472e-07, "loss": 0.9038, "step": 18553 }, { "epoch": 0.94, "learning_rate": 1.723561463357204e-07, "loss": 0.8536, "step": 18554 }, { "epoch": 0.94, "learning_rate": 1.720521035851186e-07, "loss": 0.9273, "step": 18555 }, { "epoch": 0.94, "learning_rate": 1.7174832691574829e-07, "loss": 0.7883, "step": 18556 }, { "epoch": 0.94, "learning_rate": 1.7144481633583732e-07, "loss": 0.8934, "step": 18557 }, { "epoch": 0.94, "learning_rate": 1.7114157185360136e-07, "loss": 0.7489, "step": 18558 }, { "epoch": 0.94, "learning_rate": 1.708385934772483e-07, "loss": 0.8443, "step": 18559 }, { "epoch": 0.94, "learning_rate": 1.7053588121498377e-07, "loss": 0.8175, "step": 18560 }, { "epoch": 0.94, "learning_rate": 1.7023343507500345e-07, "loss": 0.7842, "step": 18561 }, { "epoch": 0.94, "learning_rate": 1.6993125506549414e-07, "loss": 0.8228, "step": 18562 }, { "epoch": 0.94, "learning_rate": 1.6962934119463815e-07, "loss": 0.8354, "step": 18563 }, { "epoch": 0.94, "learning_rate": 1.6932769347060784e-07, "loss": 0.8696, "step": 18564 }, { "epoch": 0.94, "learning_rate": 1.6902631190157227e-07, "loss": 0.8682, "step": 18565 }, { "epoch": 0.94, "learning_rate": 1.687251964956893e-07, "loss": 0.8927, "step": 18566 }, { "epoch": 0.94, "learning_rate": 1.6842434726111356e-07, "loss": 0.8841, "step": 18567 }, { "epoch": 0.94, "learning_rate": 1.6812376420598742e-07, "loss": 0.8616, "step": 18568 }, { "epoch": 0.94, "learning_rate": 1.678234473384499e-07, "loss": 0.9155, "step": 18569 }, { "epoch": 0.94, "learning_rate": 1.675233966666323e-07, "loss": 0.7787, "step": 18570 }, { "epoch": 0.94, "learning_rate": 1.67223612198657e-07, "loss": 0.8345, "step": 18571 }, { "epoch": 0.94, "learning_rate": 1.6692409394264196e-07, "loss": 0.7721, "step": 18572 }, { "epoch": 0.94, "learning_rate": 1.6662484190669515e-07, "loss": 0.8451, "step": 18573 }, { "epoch": 0.94, "learning_rate": 1.6632585609891893e-07, "loss": 0.8154, "step": 18574 }, { "epoch": 0.94, "learning_rate": 1.6602713652740909e-07, "loss": 0.8761, "step": 18575 }, { "epoch": 0.94, "learning_rate": 1.6572868320025138e-07, "loss": 0.8351, "step": 18576 }, { "epoch": 0.94, "learning_rate": 1.65430496125526e-07, "loss": 0.883, "step": 18577 }, { "epoch": 0.94, "learning_rate": 1.6513257531130644e-07, "loss": 0.8338, "step": 18578 }, { "epoch": 0.94, "learning_rate": 1.6483492076566078e-07, "loss": 0.9186, "step": 18579 }, { "epoch": 0.94, "learning_rate": 1.6453753249664473e-07, "loss": 0.8231, "step": 18580 }, { "epoch": 0.94, "learning_rate": 1.6424041051231078e-07, "loss": 0.8428, "step": 18581 }, { "epoch": 0.94, "learning_rate": 1.6394355482070356e-07, "loss": 0.8018, "step": 18582 }, { "epoch": 0.94, "learning_rate": 1.636469654298589e-07, "loss": 0.8426, "step": 18583 }, { "epoch": 0.94, "learning_rate": 1.633506423478093e-07, "loss": 0.8826, "step": 18584 }, { "epoch": 0.94, "learning_rate": 1.630545855825738e-07, "loss": 0.8082, "step": 18585 }, { "epoch": 0.94, "learning_rate": 1.6275879514217052e-07, "loss": 0.9291, "step": 18586 }, { "epoch": 0.94, "learning_rate": 1.6246327103460746e-07, "loss": 0.9049, "step": 18587 }, { "epoch": 0.94, "learning_rate": 1.6216801326788488e-07, "loss": 0.7511, "step": 18588 }, { "epoch": 0.94, "learning_rate": 1.6187302184999643e-07, "loss": 0.8273, "step": 18589 }, { "epoch": 0.94, "learning_rate": 1.615782967889279e-07, "loss": 0.9337, "step": 18590 }, { "epoch": 0.94, "learning_rate": 1.612838380926618e-07, "loss": 0.8986, "step": 18591 }, { "epoch": 0.94, "learning_rate": 1.6098964576916732e-07, "loss": 0.8721, "step": 18592 }, { "epoch": 0.94, "learning_rate": 1.606957198264103e-07, "loss": 0.8596, "step": 18593 }, { "epoch": 0.94, "learning_rate": 1.6040206027234885e-07, "loss": 0.7847, "step": 18594 }, { "epoch": 0.94, "learning_rate": 1.60108667114931e-07, "loss": 0.7816, "step": 18595 }, { "epoch": 0.94, "learning_rate": 1.598155403621049e-07, "loss": 0.8673, "step": 18596 }, { "epoch": 0.94, "learning_rate": 1.5952268002180194e-07, "loss": 0.8696, "step": 18597 }, { "epoch": 0.94, "learning_rate": 1.5923008610195357e-07, "loss": 0.8132, "step": 18598 }, { "epoch": 0.94, "learning_rate": 1.5893775861048233e-07, "loss": 0.8649, "step": 18599 }, { "epoch": 0.94, "learning_rate": 1.5864569755529968e-07, "loss": 0.8254, "step": 18600 }, { "epoch": 0.94, "learning_rate": 1.5835390294431484e-07, "loss": 0.8102, "step": 18601 }, { "epoch": 0.95, "learning_rate": 1.5806237478542709e-07, "loss": 0.9067, "step": 18602 }, { "epoch": 0.95, "learning_rate": 1.5777111308653003e-07, "loss": 0.8432, "step": 18603 }, { "epoch": 0.95, "learning_rate": 1.5748011785550855e-07, "loss": 0.7957, "step": 18604 }, { "epoch": 0.95, "learning_rate": 1.5718938910024074e-07, "loss": 0.9267, "step": 18605 }, { "epoch": 0.95, "learning_rate": 1.5689892682859918e-07, "loss": 0.7491, "step": 18606 }, { "epoch": 0.95, "learning_rate": 1.5660873104844544e-07, "loss": 0.8788, "step": 18607 }, { "epoch": 0.95, "learning_rate": 1.563188017676398e-07, "loss": 0.8445, "step": 18608 }, { "epoch": 0.95, "learning_rate": 1.5602913899402826e-07, "loss": 0.9309, "step": 18609 }, { "epoch": 0.95, "learning_rate": 1.5573974273545455e-07, "loss": 0.8914, "step": 18610 }, { "epoch": 0.95, "learning_rate": 1.554506129997546e-07, "loss": 0.8846, "step": 18611 }, { "epoch": 0.95, "learning_rate": 1.551617497947544e-07, "loss": 0.9059, "step": 18612 }, { "epoch": 0.95, "learning_rate": 1.5487315312827767e-07, "loss": 0.9438, "step": 18613 }, { "epoch": 0.95, "learning_rate": 1.5458482300813482e-07, "loss": 0.8725, "step": 18614 }, { "epoch": 0.95, "learning_rate": 1.5429675944213296e-07, "loss": 0.8804, "step": 18615 }, { "epoch": 0.95, "learning_rate": 1.540089624380725e-07, "loss": 0.8366, "step": 18616 }, { "epoch": 0.95, "learning_rate": 1.5372143200374278e-07, "loss": 0.8435, "step": 18617 }, { "epoch": 0.95, "learning_rate": 1.534341681469298e-07, "loss": 0.8054, "step": 18618 }, { "epoch": 0.95, "learning_rate": 1.5314717087541175e-07, "loss": 0.8708, "step": 18619 }, { "epoch": 0.95, "learning_rate": 1.528604401969569e-07, "loss": 0.8938, "step": 18620 }, { "epoch": 0.95, "learning_rate": 1.5257397611933013e-07, "loss": 0.846, "step": 18621 }, { "epoch": 0.95, "learning_rate": 1.522877786502852e-07, "loss": 0.8548, "step": 18622 }, { "epoch": 0.95, "learning_rate": 1.5200184779757153e-07, "loss": 0.8148, "step": 18623 }, { "epoch": 0.95, "learning_rate": 1.517161835689296e-07, "loss": 0.8345, "step": 18624 }, { "epoch": 0.95, "learning_rate": 1.5143078597209648e-07, "loss": 0.8883, "step": 18625 }, { "epoch": 0.95, "learning_rate": 1.5114565501479495e-07, "loss": 0.8957, "step": 18626 }, { "epoch": 0.95, "learning_rate": 1.5086079070474658e-07, "loss": 0.8151, "step": 18627 }, { "epoch": 0.95, "learning_rate": 1.505761930496652e-07, "loss": 0.9203, "step": 18628 }, { "epoch": 0.95, "learning_rate": 1.5029186205725245e-07, "loss": 0.8273, "step": 18629 }, { "epoch": 0.95, "learning_rate": 1.500077977352088e-07, "loss": 0.8865, "step": 18630 }, { "epoch": 0.95, "learning_rate": 1.4972400009122478e-07, "loss": 0.8286, "step": 18631 }, { "epoch": 0.95, "learning_rate": 1.4944046913298428e-07, "loss": 0.9934, "step": 18632 }, { "epoch": 0.95, "learning_rate": 1.491572048681622e-07, "loss": 0.8498, "step": 18633 }, { "epoch": 0.95, "learning_rate": 1.4887420730442803e-07, "loss": 0.9119, "step": 18634 }, { "epoch": 0.95, "learning_rate": 1.4859147644944449e-07, "loss": 0.8618, "step": 18635 }, { "epoch": 0.95, "learning_rate": 1.4830901231086548e-07, "loss": 0.8543, "step": 18636 }, { "epoch": 0.95, "learning_rate": 1.480268148963382e-07, "loss": 0.8458, "step": 18637 }, { "epoch": 0.95, "learning_rate": 1.477448842135043e-07, "loss": 0.8992, "step": 18638 }, { "epoch": 0.95, "learning_rate": 1.4746322026999438e-07, "loss": 0.9218, "step": 18639 }, { "epoch": 0.95, "learning_rate": 1.4718182307343677e-07, "loss": 0.8049, "step": 18640 }, { "epoch": 0.95, "learning_rate": 1.4690069263144757e-07, "loss": 0.8412, "step": 18641 }, { "epoch": 0.95, "learning_rate": 1.4661982895163962e-07, "loss": 0.7721, "step": 18642 }, { "epoch": 0.95, "learning_rate": 1.463392320416168e-07, "loss": 0.8925, "step": 18643 }, { "epoch": 0.95, "learning_rate": 1.460589019089753e-07, "loss": 0.8678, "step": 18644 }, { "epoch": 0.95, "learning_rate": 1.457788385613057e-07, "loss": 0.7733, "step": 18645 }, { "epoch": 0.95, "learning_rate": 1.4549904200618968e-07, "loss": 0.8521, "step": 18646 }, { "epoch": 0.95, "learning_rate": 1.4521951225120345e-07, "loss": 0.7408, "step": 18647 }, { "epoch": 0.95, "learning_rate": 1.4494024930391315e-07, "loss": 0.8118, "step": 18648 }, { "epoch": 0.95, "learning_rate": 1.446612531718816e-07, "loss": 0.8209, "step": 18649 }, { "epoch": 0.95, "learning_rate": 1.443825238626617e-07, "loss": 0.9144, "step": 18650 }, { "epoch": 0.95, "learning_rate": 1.441040613837985e-07, "loss": 0.8148, "step": 18651 }, { "epoch": 0.95, "learning_rate": 1.4382586574283264e-07, "loss": 0.7822, "step": 18652 }, { "epoch": 0.95, "learning_rate": 1.4354793694729474e-07, "loss": 0.8957, "step": 18653 }, { "epoch": 0.95, "learning_rate": 1.432702750047099e-07, "loss": 0.8681, "step": 18654 }, { "epoch": 0.95, "learning_rate": 1.4299287992259549e-07, "loss": 0.8546, "step": 18655 }, { "epoch": 0.95, "learning_rate": 1.427157517084632e-07, "loss": 0.9334, "step": 18656 }, { "epoch": 0.95, "learning_rate": 1.4243889036981373e-07, "loss": 0.8311, "step": 18657 }, { "epoch": 0.95, "learning_rate": 1.4216229591414333e-07, "loss": 1.0159, "step": 18658 }, { "epoch": 0.95, "learning_rate": 1.4188596834894152e-07, "loss": 0.866, "step": 18659 }, { "epoch": 0.95, "learning_rate": 1.4160990768168792e-07, "loss": 0.8678, "step": 18660 }, { "epoch": 0.95, "learning_rate": 1.4133411391985875e-07, "loss": 0.9348, "step": 18661 }, { "epoch": 0.95, "learning_rate": 1.4105858707091913e-07, "loss": 0.9263, "step": 18662 }, { "epoch": 0.95, "learning_rate": 1.4078332714232869e-07, "loss": 0.7451, "step": 18663 }, { "epoch": 0.95, "learning_rate": 1.4050833414154143e-07, "loss": 0.8599, "step": 18664 }, { "epoch": 0.95, "learning_rate": 1.402336080760003e-07, "loss": 0.9801, "step": 18665 }, { "epoch": 0.95, "learning_rate": 1.399591489531449e-07, "loss": 0.8933, "step": 18666 }, { "epoch": 0.95, "learning_rate": 1.3968495678040483e-07, "loss": 0.8642, "step": 18667 }, { "epoch": 0.95, "learning_rate": 1.3941103156520418e-07, "loss": 0.8532, "step": 18668 }, { "epoch": 0.95, "learning_rate": 1.3913737331495924e-07, "loss": 0.8553, "step": 18669 }, { "epoch": 0.95, "learning_rate": 1.388639820370785e-07, "loss": 0.787, "step": 18670 }, { "epoch": 0.95, "learning_rate": 1.3859085773896386e-07, "loss": 0.9627, "step": 18671 }, { "epoch": 0.95, "learning_rate": 1.3831800042800936e-07, "loss": 0.8999, "step": 18672 }, { "epoch": 0.95, "learning_rate": 1.3804541011160466e-07, "loss": 0.8604, "step": 18673 }, { "epoch": 0.95, "learning_rate": 1.377730867971272e-07, "loss": 0.903, "step": 18674 }, { "epoch": 0.95, "learning_rate": 1.3750103049195108e-07, "loss": 0.8362, "step": 18675 }, { "epoch": 0.95, "learning_rate": 1.3722924120344261e-07, "loss": 0.9117, "step": 18676 }, { "epoch": 0.95, "learning_rate": 1.3695771893895816e-07, "loss": 0.7824, "step": 18677 }, { "epoch": 0.95, "learning_rate": 1.366864637058496e-07, "loss": 0.9653, "step": 18678 }, { "epoch": 0.95, "learning_rate": 1.3641547551146218e-07, "loss": 0.8543, "step": 18679 }, { "epoch": 0.95, "learning_rate": 1.3614475436313114e-07, "loss": 0.7971, "step": 18680 }, { "epoch": 0.95, "learning_rate": 1.3587430026818727e-07, "loss": 0.8549, "step": 18681 }, { "epoch": 0.95, "learning_rate": 1.356041132339514e-07, "loss": 0.924, "step": 18682 }, { "epoch": 0.95, "learning_rate": 1.3533419326773989e-07, "loss": 0.8997, "step": 18683 }, { "epoch": 0.95, "learning_rate": 1.3506454037685912e-07, "loss": 0.8682, "step": 18684 }, { "epoch": 0.95, "learning_rate": 1.3479515456861103e-07, "loss": 0.9355, "step": 18685 }, { "epoch": 0.95, "learning_rate": 1.3452603585028867e-07, "loss": 0.807, "step": 18686 }, { "epoch": 0.95, "learning_rate": 1.3425718422917733e-07, "loss": 0.8715, "step": 18687 }, { "epoch": 0.95, "learning_rate": 1.3398859971255672e-07, "loss": 0.8774, "step": 18688 }, { "epoch": 0.95, "learning_rate": 1.3372028230769663e-07, "loss": 0.906, "step": 18689 }, { "epoch": 0.95, "learning_rate": 1.3345223202186564e-07, "loss": 0.898, "step": 18690 }, { "epoch": 0.95, "learning_rate": 1.3318444886231686e-07, "loss": 0.8723, "step": 18691 }, { "epoch": 0.95, "learning_rate": 1.3291693283630114e-07, "loss": 0.9594, "step": 18692 }, { "epoch": 0.95, "learning_rate": 1.326496839510627e-07, "loss": 0.8747, "step": 18693 }, { "epoch": 0.95, "learning_rate": 1.3238270221383575e-07, "loss": 0.8204, "step": 18694 }, { "epoch": 0.95, "learning_rate": 1.3211598763184896e-07, "loss": 0.9204, "step": 18695 }, { "epoch": 0.95, "learning_rate": 1.318495402123232e-07, "loss": 0.9114, "step": 18696 }, { "epoch": 0.95, "learning_rate": 1.3158335996247163e-07, "loss": 0.7904, "step": 18697 }, { "epoch": 0.95, "learning_rate": 1.3131744688950176e-07, "loss": 0.9971, "step": 18698 }, { "epoch": 0.95, "learning_rate": 1.3105180100061343e-07, "loss": 0.8649, "step": 18699 }, { "epoch": 0.95, "learning_rate": 1.3078642230299754e-07, "loss": 0.7603, "step": 18700 }, { "epoch": 0.95, "learning_rate": 1.3052131080383723e-07, "loss": 0.8879, "step": 18701 }, { "epoch": 0.95, "learning_rate": 1.3025646651031453e-07, "loss": 0.8627, "step": 18702 }, { "epoch": 0.95, "learning_rate": 1.2999188942959596e-07, "loss": 0.8778, "step": 18703 }, { "epoch": 0.95, "learning_rate": 1.2972757956884686e-07, "loss": 0.8509, "step": 18704 }, { "epoch": 0.95, "learning_rate": 1.2946353693522261e-07, "loss": 0.9801, "step": 18705 }, { "epoch": 0.95, "learning_rate": 1.29199761535872e-07, "loss": 0.8961, "step": 18706 }, { "epoch": 0.95, "learning_rate": 1.289362533779348e-07, "loss": 0.8499, "step": 18707 }, { "epoch": 0.95, "learning_rate": 1.2867301246854757e-07, "loss": 0.8435, "step": 18708 }, { "epoch": 0.95, "learning_rate": 1.2841003881483572e-07, "loss": 0.8653, "step": 18709 }, { "epoch": 0.95, "learning_rate": 1.281473324239202e-07, "loss": 0.8313, "step": 18710 }, { "epoch": 0.95, "learning_rate": 1.2788489330291197e-07, "loss": 0.8403, "step": 18711 }, { "epoch": 0.95, "learning_rate": 1.2762272145891763e-07, "loss": 0.815, "step": 18712 }, { "epoch": 0.95, "learning_rate": 1.2736081689903367e-07, "loss": 0.8638, "step": 18713 }, { "epoch": 0.95, "learning_rate": 1.2709917963035446e-07, "loss": 0.8656, "step": 18714 }, { "epoch": 0.95, "learning_rate": 1.2683780965995874e-07, "loss": 0.8969, "step": 18715 }, { "epoch": 0.95, "learning_rate": 1.265767069949253e-07, "loss": 0.9468, "step": 18716 }, { "epoch": 0.95, "learning_rate": 1.2631587164232405e-07, "loss": 0.7773, "step": 18717 }, { "epoch": 0.95, "learning_rate": 1.2605530360921602e-07, "loss": 0.8401, "step": 18718 }, { "epoch": 0.95, "learning_rate": 1.2579500290265445e-07, "loss": 0.8716, "step": 18719 }, { "epoch": 0.95, "learning_rate": 1.2553496952968812e-07, "loss": 0.9097, "step": 18720 }, { "epoch": 0.95, "learning_rate": 1.25275203497357e-07, "loss": 0.8782, "step": 18721 }, { "epoch": 0.95, "learning_rate": 1.250157048126943e-07, "loss": 0.8434, "step": 18722 }, { "epoch": 0.95, "learning_rate": 1.2475647348272445e-07, "loss": 0.8272, "step": 18723 }, { "epoch": 0.95, "learning_rate": 1.2449750951446738e-07, "loss": 0.82, "step": 18724 }, { "epoch": 0.95, "learning_rate": 1.24238812914933e-07, "loss": 0.8319, "step": 18725 }, { "epoch": 0.95, "learning_rate": 1.2398038369112575e-07, "loss": 0.8872, "step": 18726 }, { "epoch": 0.95, "learning_rate": 1.2372222185004223e-07, "loss": 0.8442, "step": 18727 }, { "epoch": 0.95, "learning_rate": 1.2346432739867244e-07, "loss": 0.8467, "step": 18728 }, { "epoch": 0.95, "learning_rate": 1.232067003439985e-07, "loss": 0.8396, "step": 18729 }, { "epoch": 0.95, "learning_rate": 1.229493406929949e-07, "loss": 0.951, "step": 18730 }, { "epoch": 0.95, "learning_rate": 1.226922484526294e-07, "loss": 1.0067, "step": 18731 }, { "epoch": 0.95, "learning_rate": 1.2243542362986193e-07, "loss": 0.8218, "step": 18732 }, { "epoch": 0.95, "learning_rate": 1.221788662316481e-07, "loss": 0.9727, "step": 18733 }, { "epoch": 0.95, "learning_rate": 1.219225762649312e-07, "loss": 0.8451, "step": 18734 }, { "epoch": 0.95, "learning_rate": 1.2166655373665126e-07, "loss": 0.9388, "step": 18735 }, { "epoch": 0.95, "learning_rate": 1.214107986537394e-07, "loss": 0.8301, "step": 18736 }, { "epoch": 0.95, "learning_rate": 1.2115531102312005e-07, "loss": 0.8119, "step": 18737 }, { "epoch": 0.95, "learning_rate": 1.2090009085171107e-07, "loss": 0.8576, "step": 18738 }, { "epoch": 0.95, "learning_rate": 1.2064513814642132e-07, "loss": 0.8497, "step": 18739 }, { "epoch": 0.95, "learning_rate": 1.203904529141531e-07, "loss": 0.8958, "step": 18740 }, { "epoch": 0.95, "learning_rate": 1.2013603516180305e-07, "loss": 0.8087, "step": 18741 }, { "epoch": 0.95, "learning_rate": 1.1988188489625797e-07, "loss": 0.7435, "step": 18742 }, { "epoch": 0.95, "learning_rate": 1.1962800212439896e-07, "loss": 0.8134, "step": 18743 }, { "epoch": 0.95, "learning_rate": 1.1937438685310055e-07, "loss": 0.8032, "step": 18744 }, { "epoch": 0.95, "learning_rate": 1.1912103908922945e-07, "loss": 0.8516, "step": 18745 }, { "epoch": 0.95, "learning_rate": 1.1886795883964242e-07, "loss": 0.9234, "step": 18746 }, { "epoch": 0.95, "learning_rate": 1.1861514611119284e-07, "loss": 0.8335, "step": 18747 }, { "epoch": 0.95, "learning_rate": 1.1836260091072526e-07, "loss": 0.8156, "step": 18748 }, { "epoch": 0.95, "learning_rate": 1.1811032324507643e-07, "loss": 0.8498, "step": 18749 }, { "epoch": 0.95, "learning_rate": 1.1785831312107865e-07, "loss": 0.8275, "step": 18750 }, { "epoch": 0.95, "learning_rate": 1.1760657054555202e-07, "loss": 0.8631, "step": 18751 }, { "epoch": 0.95, "learning_rate": 1.1735509552531332e-07, "loss": 0.8583, "step": 18752 }, { "epoch": 0.95, "learning_rate": 1.1710388806717266e-07, "loss": 0.8783, "step": 18753 }, { "epoch": 0.95, "learning_rate": 1.168529481779268e-07, "loss": 0.739, "step": 18754 }, { "epoch": 0.95, "learning_rate": 1.1660227586437589e-07, "loss": 0.8734, "step": 18755 }, { "epoch": 0.95, "learning_rate": 1.1635187113330115e-07, "loss": 0.8906, "step": 18756 }, { "epoch": 0.95, "learning_rate": 1.1610173399148383e-07, "loss": 0.8363, "step": 18757 }, { "epoch": 0.95, "learning_rate": 1.1585186444569851e-07, "loss": 0.893, "step": 18758 }, { "epoch": 0.95, "learning_rate": 1.1560226250270645e-07, "loss": 0.8666, "step": 18759 }, { "epoch": 0.95, "learning_rate": 1.1535292816926669e-07, "loss": 0.8109, "step": 18760 }, { "epoch": 0.95, "learning_rate": 1.1510386145213049e-07, "loss": 0.8022, "step": 18761 }, { "epoch": 0.95, "learning_rate": 1.1485506235804133e-07, "loss": 0.8619, "step": 18762 }, { "epoch": 0.95, "learning_rate": 1.1460653089373386e-07, "loss": 0.9429, "step": 18763 }, { "epoch": 0.95, "learning_rate": 1.1435826706593711e-07, "loss": 0.8616, "step": 18764 }, { "epoch": 0.95, "learning_rate": 1.141102708813735e-07, "loss": 0.8298, "step": 18765 }, { "epoch": 0.95, "learning_rate": 1.1386254234675431e-07, "loss": 0.865, "step": 18766 }, { "epoch": 0.95, "learning_rate": 1.1361508146879085e-07, "loss": 0.8305, "step": 18767 }, { "epoch": 0.95, "learning_rate": 1.1336788825417999e-07, "loss": 0.8379, "step": 18768 }, { "epoch": 0.95, "learning_rate": 1.1312096270961525e-07, "loss": 0.9604, "step": 18769 }, { "epoch": 0.95, "learning_rate": 1.1287430484178241e-07, "loss": 0.8613, "step": 18770 }, { "epoch": 0.95, "learning_rate": 1.1262791465735723e-07, "loss": 0.7793, "step": 18771 }, { "epoch": 0.95, "learning_rate": 1.1238179216301326e-07, "loss": 0.802, "step": 18772 }, { "epoch": 0.95, "learning_rate": 1.1213593736541184e-07, "loss": 0.8483, "step": 18773 }, { "epoch": 0.95, "learning_rate": 1.1189035027121098e-07, "loss": 0.9813, "step": 18774 }, { "epoch": 0.95, "learning_rate": 1.1164503088705758e-07, "loss": 0.7816, "step": 18775 }, { "epoch": 0.95, "learning_rate": 1.1139997921959522e-07, "loss": 0.8289, "step": 18776 }, { "epoch": 0.95, "learning_rate": 1.1115519527545748e-07, "loss": 0.7931, "step": 18777 }, { "epoch": 0.95, "learning_rate": 1.1091067906127129e-07, "loss": 0.9777, "step": 18778 }, { "epoch": 0.95, "learning_rate": 1.1066643058365911e-07, "loss": 0.9324, "step": 18779 }, { "epoch": 0.95, "learning_rate": 1.1042244984923012e-07, "loss": 0.8276, "step": 18780 }, { "epoch": 0.95, "learning_rate": 1.1017873686459236e-07, "loss": 0.8136, "step": 18781 }, { "epoch": 0.95, "learning_rate": 1.0993529163634386e-07, "loss": 0.8418, "step": 18782 }, { "epoch": 0.95, "learning_rate": 1.0969211417107383e-07, "loss": 0.9047, "step": 18783 }, { "epoch": 0.95, "learning_rate": 1.0944920447536811e-07, "loss": 0.8869, "step": 18784 }, { "epoch": 0.95, "learning_rate": 1.0920656255580252e-07, "loss": 0.9588, "step": 18785 }, { "epoch": 0.95, "learning_rate": 1.0896418841894741e-07, "loss": 0.8165, "step": 18786 }, { "epoch": 0.95, "learning_rate": 1.0872208207136192e-07, "loss": 0.8339, "step": 18787 }, { "epoch": 0.95, "learning_rate": 1.0848024351960307e-07, "loss": 0.8986, "step": 18788 }, { "epoch": 0.95, "learning_rate": 1.0823867277021894e-07, "loss": 0.753, "step": 18789 }, { "epoch": 0.95, "learning_rate": 1.0799736982974651e-07, "loss": 0.7686, "step": 18790 }, { "epoch": 0.95, "learning_rate": 1.0775633470472279e-07, "loss": 0.8235, "step": 18791 }, { "epoch": 0.95, "learning_rate": 1.0751556740167147e-07, "loss": 0.9398, "step": 18792 }, { "epoch": 0.95, "learning_rate": 1.0727506792711063e-07, "loss": 0.9232, "step": 18793 }, { "epoch": 0.95, "learning_rate": 1.0703483628755284e-07, "loss": 0.8317, "step": 18794 }, { "epoch": 0.95, "learning_rate": 1.0679487248950182e-07, "loss": 0.8488, "step": 18795 }, { "epoch": 0.95, "learning_rate": 1.0655517653945347e-07, "loss": 0.8321, "step": 18796 }, { "epoch": 0.95, "learning_rate": 1.0631574844389813e-07, "loss": 0.8716, "step": 18797 }, { "epoch": 0.95, "learning_rate": 1.0607658820931843e-07, "loss": 0.9589, "step": 18798 }, { "epoch": 0.96, "learning_rate": 1.0583769584218917e-07, "loss": 0.8533, "step": 18799 }, { "epoch": 0.96, "learning_rate": 1.0559907134897629e-07, "loss": 0.8487, "step": 18800 }, { "epoch": 0.96, "learning_rate": 1.053607147361424e-07, "loss": 0.8262, "step": 18801 }, { "epoch": 0.96, "learning_rate": 1.0512262601014011e-07, "loss": 0.7549, "step": 18802 }, { "epoch": 0.96, "learning_rate": 1.0488480517741539e-07, "loss": 0.8231, "step": 18803 }, { "epoch": 0.96, "learning_rate": 1.0464725224440753e-07, "loss": 0.8936, "step": 18804 }, { "epoch": 0.96, "learning_rate": 1.0440996721754582e-07, "loss": 0.9011, "step": 18805 }, { "epoch": 0.96, "learning_rate": 1.0417295010325846e-07, "loss": 0.9028, "step": 18806 }, { "epoch": 0.96, "learning_rate": 1.0393620090795809e-07, "loss": 0.9029, "step": 18807 }, { "epoch": 0.96, "learning_rate": 1.0369971963805737e-07, "loss": 0.8942, "step": 18808 }, { "epoch": 0.96, "learning_rate": 1.0346350629995783e-07, "loss": 0.9688, "step": 18809 }, { "epoch": 0.96, "learning_rate": 1.0322756090005548e-07, "loss": 0.8339, "step": 18810 }, { "epoch": 0.96, "learning_rate": 1.0299188344473743e-07, "loss": 0.8377, "step": 18811 }, { "epoch": 0.96, "learning_rate": 1.0275647394038301e-07, "loss": 0.8361, "step": 18812 }, { "epoch": 0.96, "learning_rate": 1.0252133239336936e-07, "loss": 0.9308, "step": 18813 }, { "epoch": 0.96, "learning_rate": 1.0228645881005917e-07, "loss": 0.8981, "step": 18814 }, { "epoch": 0.96, "learning_rate": 1.0205185319681287e-07, "loss": 0.8805, "step": 18815 }, { "epoch": 0.96, "learning_rate": 1.0181751555998209e-07, "loss": 0.9563, "step": 18816 }, { "epoch": 0.96, "learning_rate": 1.0158344590591173e-07, "loss": 0.8714, "step": 18817 }, { "epoch": 0.96, "learning_rate": 1.0134964424093896e-07, "loss": 0.8354, "step": 18818 }, { "epoch": 0.96, "learning_rate": 1.0111611057139092e-07, "loss": 0.9816, "step": 18819 }, { "epoch": 0.96, "learning_rate": 1.0088284490359478e-07, "loss": 0.8256, "step": 18820 }, { "epoch": 0.96, "learning_rate": 1.0064984724386328e-07, "loss": 0.8334, "step": 18821 }, { "epoch": 0.96, "learning_rate": 1.004171175985047e-07, "loss": 0.9164, "step": 18822 }, { "epoch": 0.96, "learning_rate": 1.0018465597382065e-07, "loss": 0.9571, "step": 18823 }, { "epoch": 0.96, "learning_rate": 9.995246237610501e-08, "loss": 0.8073, "step": 18824 }, { "epoch": 0.96, "learning_rate": 9.972053681164273e-08, "loss": 0.8395, "step": 18825 }, { "epoch": 0.96, "learning_rate": 9.948887928671325e-08, "loss": 0.8183, "step": 18826 }, { "epoch": 0.96, "learning_rate": 9.925748980759042e-08, "loss": 0.9123, "step": 18827 }, { "epoch": 0.96, "learning_rate": 9.902636838053703e-08, "loss": 0.9232, "step": 18828 }, { "epoch": 0.96, "learning_rate": 9.879551501181139e-08, "loss": 0.8502, "step": 18829 }, { "epoch": 0.96, "learning_rate": 9.856492970766296e-08, "loss": 0.9295, "step": 18830 }, { "epoch": 0.96, "learning_rate": 9.83346124743334e-08, "loss": 0.7814, "step": 18831 }, { "epoch": 0.96, "learning_rate": 9.810456331806106e-08, "loss": 0.8104, "step": 18832 }, { "epoch": 0.96, "learning_rate": 9.787478224507208e-08, "loss": 0.8682, "step": 18833 }, { "epoch": 0.96, "learning_rate": 9.764526926158923e-08, "loss": 0.8801, "step": 18834 }, { "epoch": 0.96, "learning_rate": 9.741602437382536e-08, "loss": 0.9535, "step": 18835 }, { "epoch": 0.96, "learning_rate": 9.718704758798659e-08, "loss": 0.8718, "step": 18836 }, { "epoch": 0.96, "learning_rate": 9.69583389102724e-08, "loss": 0.9059, "step": 18837 }, { "epoch": 0.96, "learning_rate": 9.672989834687563e-08, "loss": 0.8617, "step": 18838 }, { "epoch": 0.96, "learning_rate": 9.650172590398132e-08, "loss": 0.894, "step": 18839 }, { "epoch": 0.96, "learning_rate": 9.627382158776566e-08, "loss": 0.8362, "step": 18840 }, { "epoch": 0.96, "learning_rate": 9.604618540439925e-08, "loss": 0.9052, "step": 18841 }, { "epoch": 0.96, "learning_rate": 9.581881736004606e-08, "loss": 0.9022, "step": 18842 }, { "epoch": 0.96, "learning_rate": 9.559171746086004e-08, "loss": 0.845, "step": 18843 }, { "epoch": 0.96, "learning_rate": 9.536488571299185e-08, "loss": 0.8826, "step": 18844 }, { "epoch": 0.96, "learning_rate": 9.513832212258101e-08, "loss": 0.8911, "step": 18845 }, { "epoch": 0.96, "learning_rate": 9.49120266957626e-08, "loss": 0.9603, "step": 18846 }, { "epoch": 0.96, "learning_rate": 9.468599943866285e-08, "loss": 0.8405, "step": 18847 }, { "epoch": 0.96, "learning_rate": 9.446024035740131e-08, "loss": 0.9005, "step": 18848 }, { "epoch": 0.96, "learning_rate": 9.423474945808975e-08, "loss": 0.8174, "step": 18849 }, { "epoch": 0.96, "learning_rate": 9.400952674683328e-08, "loss": 0.9455, "step": 18850 }, { "epoch": 0.96, "learning_rate": 9.378457222973037e-08, "loss": 0.8521, "step": 18851 }, { "epoch": 0.96, "learning_rate": 9.355988591287057e-08, "loss": 0.8752, "step": 18852 }, { "epoch": 0.96, "learning_rate": 9.333546780233683e-08, "loss": 0.921, "step": 18853 }, { "epoch": 0.96, "learning_rate": 9.311131790420647e-08, "loss": 0.8831, "step": 18854 }, { "epoch": 0.96, "learning_rate": 9.288743622454577e-08, "loss": 0.861, "step": 18855 }, { "epoch": 0.96, "learning_rate": 9.266382276941766e-08, "loss": 0.8405, "step": 18856 }, { "epoch": 0.96, "learning_rate": 9.244047754487617e-08, "loss": 0.7887, "step": 18857 }, { "epoch": 0.96, "learning_rate": 9.22174005569687e-08, "loss": 0.7542, "step": 18858 }, { "epoch": 0.96, "learning_rate": 9.199459181173376e-08, "loss": 0.867, "step": 18859 }, { "epoch": 0.96, "learning_rate": 9.177205131520316e-08, "loss": 0.9588, "step": 18860 }, { "epoch": 0.96, "learning_rate": 9.154977907340323e-08, "loss": 0.8203, "step": 18861 }, { "epoch": 0.96, "learning_rate": 9.132777509235135e-08, "loss": 0.9074, "step": 18862 }, { "epoch": 0.96, "learning_rate": 9.110603937805828e-08, "loss": 0.8807, "step": 18863 }, { "epoch": 0.96, "learning_rate": 9.088457193652588e-08, "loss": 0.8239, "step": 18864 }, { "epoch": 0.96, "learning_rate": 9.06633727737527e-08, "loss": 0.8023, "step": 18865 }, { "epoch": 0.96, "learning_rate": 9.044244189572504e-08, "loss": 0.8139, "step": 18866 }, { "epoch": 0.96, "learning_rate": 9.022177930842479e-08, "loss": 0.9139, "step": 18867 }, { "epoch": 0.96, "learning_rate": 9.000138501782829e-08, "loss": 0.808, "step": 18868 }, { "epoch": 0.96, "learning_rate": 8.978125902989964e-08, "loss": 0.8374, "step": 18869 }, { "epoch": 0.96, "learning_rate": 8.956140135059966e-08, "loss": 0.9236, "step": 18870 }, { "epoch": 0.96, "learning_rate": 8.934181198588244e-08, "loss": 0.8525, "step": 18871 }, { "epoch": 0.96, "learning_rate": 8.91224909416899e-08, "loss": 0.8121, "step": 18872 }, { "epoch": 0.96, "learning_rate": 8.890343822396174e-08, "loss": 0.8638, "step": 18873 }, { "epoch": 0.96, "learning_rate": 8.868465383862878e-08, "loss": 0.8309, "step": 18874 }, { "epoch": 0.96, "learning_rate": 8.846613779161406e-08, "loss": 0.7907, "step": 18875 }, { "epoch": 0.96, "learning_rate": 8.824789008883394e-08, "loss": 0.8603, "step": 18876 }, { "epoch": 0.96, "learning_rate": 8.802991073619594e-08, "loss": 0.9343, "step": 18877 }, { "epoch": 0.96, "learning_rate": 8.78121997396042e-08, "loss": 0.948, "step": 18878 }, { "epoch": 0.96, "learning_rate": 8.759475710494958e-08, "loss": 0.9885, "step": 18879 }, { "epoch": 0.96, "learning_rate": 8.737758283812181e-08, "loss": 0.7817, "step": 18880 }, { "epoch": 0.96, "learning_rate": 8.716067694499952e-08, "loss": 0.8137, "step": 18881 }, { "epoch": 0.96, "learning_rate": 8.694403943145579e-08, "loss": 0.8683, "step": 18882 }, { "epoch": 0.96, "learning_rate": 8.672767030335593e-08, "loss": 0.8139, "step": 18883 }, { "epoch": 0.96, "learning_rate": 8.651156956655637e-08, "loss": 0.9653, "step": 18884 }, { "epoch": 0.96, "learning_rate": 8.62957372269102e-08, "loss": 0.8842, "step": 18885 }, { "epoch": 0.96, "learning_rate": 8.60801732902583e-08, "loss": 0.7959, "step": 18886 }, { "epoch": 0.96, "learning_rate": 8.586487776243934e-08, "loss": 0.9628, "step": 18887 }, { "epoch": 0.96, "learning_rate": 8.564985064928089e-08, "loss": 0.8484, "step": 18888 }, { "epoch": 0.96, "learning_rate": 8.543509195660382e-08, "loss": 0.9738, "step": 18889 }, { "epoch": 0.96, "learning_rate": 8.52206016902235e-08, "loss": 0.8862, "step": 18890 }, { "epoch": 0.96, "learning_rate": 8.50063798559475e-08, "loss": 0.8463, "step": 18891 }, { "epoch": 0.96, "learning_rate": 8.479242645957453e-08, "loss": 0.8651, "step": 18892 }, { "epoch": 0.96, "learning_rate": 8.457874150689771e-08, "loss": 0.7876, "step": 18893 }, { "epoch": 0.96, "learning_rate": 8.436532500370242e-08, "loss": 0.926, "step": 18894 }, { "epoch": 0.96, "learning_rate": 8.415217695576739e-08, "loss": 0.82, "step": 18895 }, { "epoch": 0.96, "learning_rate": 8.39392973688613e-08, "loss": 0.8989, "step": 18896 }, { "epoch": 0.96, "learning_rate": 8.372668624874957e-08, "loss": 0.8082, "step": 18897 }, { "epoch": 0.96, "learning_rate": 8.351434360118759e-08, "loss": 0.885, "step": 18898 }, { "epoch": 0.96, "learning_rate": 8.330226943192409e-08, "loss": 0.8287, "step": 18899 }, { "epoch": 0.96, "learning_rate": 8.309046374670115e-08, "loss": 0.8661, "step": 18900 }, { "epoch": 0.96, "learning_rate": 8.287892655125307e-08, "loss": 0.9558, "step": 18901 }, { "epoch": 0.96, "learning_rate": 8.266765785130748e-08, "loss": 0.9226, "step": 18902 }, { "epoch": 0.96, "learning_rate": 8.245665765258315e-08, "loss": 0.8605, "step": 18903 }, { "epoch": 0.96, "learning_rate": 8.22459259607944e-08, "loss": 0.7694, "step": 18904 }, { "epoch": 0.96, "learning_rate": 8.203546278164554e-08, "loss": 0.8098, "step": 18905 }, { "epoch": 0.96, "learning_rate": 8.182526812083424e-08, "loss": 0.8089, "step": 18906 }, { "epoch": 0.96, "learning_rate": 8.161534198405152e-08, "loss": 0.9265, "step": 18907 }, { "epoch": 0.96, "learning_rate": 8.14056843769806e-08, "loss": 0.8824, "step": 18908 }, { "epoch": 0.96, "learning_rate": 8.119629530529916e-08, "loss": 0.8934, "step": 18909 }, { "epoch": 0.96, "learning_rate": 8.098717477467488e-08, "loss": 0.9548, "step": 18910 }, { "epoch": 0.96, "learning_rate": 8.077832279076991e-08, "loss": 0.8514, "step": 18911 }, { "epoch": 0.96, "learning_rate": 8.056973935923862e-08, "loss": 0.8452, "step": 18912 }, { "epoch": 0.96, "learning_rate": 8.03614244857287e-08, "loss": 0.7891, "step": 18913 }, { "epoch": 0.96, "learning_rate": 8.015337817587898e-08, "loss": 0.787, "step": 18914 }, { "epoch": 0.96, "learning_rate": 7.994560043532162e-08, "loss": 0.834, "step": 18915 }, { "epoch": 0.96, "learning_rate": 7.973809126968545e-08, "loss": 0.8044, "step": 18916 }, { "epoch": 0.96, "learning_rate": 7.953085068458377e-08, "loss": 0.8821, "step": 18917 }, { "epoch": 0.96, "learning_rate": 7.932387868563096e-08, "loss": 0.802, "step": 18918 }, { "epoch": 0.96, "learning_rate": 7.911717527842921e-08, "loss": 0.8692, "step": 18919 }, { "epoch": 0.96, "learning_rate": 7.891074046857517e-08, "loss": 0.835, "step": 18920 }, { "epoch": 0.96, "learning_rate": 7.870457426165767e-08, "loss": 0.8752, "step": 18921 }, { "epoch": 0.96, "learning_rate": 7.849867666325783e-08, "loss": 0.9097, "step": 18922 }, { "epoch": 0.96, "learning_rate": 7.829304767895007e-08, "loss": 0.8527, "step": 18923 }, { "epoch": 0.96, "learning_rate": 7.808768731430438e-08, "loss": 0.9065, "step": 18924 }, { "epoch": 0.96, "learning_rate": 7.788259557487632e-08, "loss": 0.8547, "step": 18925 }, { "epoch": 0.96, "learning_rate": 7.767777246622144e-08, "loss": 0.8301, "step": 18926 }, { "epoch": 0.96, "learning_rate": 7.747321799388419e-08, "loss": 0.8519, "step": 18927 }, { "epoch": 0.96, "learning_rate": 7.726893216340348e-08, "loss": 0.793, "step": 18928 }, { "epoch": 0.96, "learning_rate": 7.706491498030932e-08, "loss": 0.8507, "step": 18929 }, { "epoch": 0.96, "learning_rate": 7.68611664501251e-08, "loss": 0.8694, "step": 18930 }, { "epoch": 0.96, "learning_rate": 7.665768657836747e-08, "loss": 0.813, "step": 18931 }, { "epoch": 0.96, "learning_rate": 7.645447537054428e-08, "loss": 0.8421, "step": 18932 }, { "epoch": 0.96, "learning_rate": 7.625153283216002e-08, "loss": 0.8376, "step": 18933 }, { "epoch": 0.96, "learning_rate": 7.604885896870695e-08, "loss": 0.8492, "step": 18934 }, { "epoch": 0.96, "learning_rate": 7.58464537856729e-08, "loss": 0.8562, "step": 18935 }, { "epoch": 0.96, "learning_rate": 7.564431728853793e-08, "loss": 0.9999, "step": 18936 }, { "epoch": 0.96, "learning_rate": 7.544244948277324e-08, "loss": 0.8251, "step": 18937 }, { "epoch": 0.96, "learning_rate": 7.524085037384554e-08, "loss": 0.896, "step": 18938 }, { "epoch": 0.96, "learning_rate": 7.503951996721381e-08, "loss": 0.8728, "step": 18939 }, { "epoch": 0.96, "learning_rate": 7.483845826832703e-08, "loss": 0.8613, "step": 18940 }, { "epoch": 0.96, "learning_rate": 7.463766528262972e-08, "loss": 0.796, "step": 18941 }, { "epoch": 0.96, "learning_rate": 7.443714101555755e-08, "loss": 0.8086, "step": 18942 }, { "epoch": 0.96, "learning_rate": 7.423688547253948e-08, "loss": 0.9568, "step": 18943 }, { "epoch": 0.96, "learning_rate": 7.403689865899788e-08, "loss": 0.9145, "step": 18944 }, { "epoch": 0.96, "learning_rate": 7.383718058034728e-08, "loss": 0.9099, "step": 18945 }, { "epoch": 0.96, "learning_rate": 7.363773124199447e-08, "loss": 0.8796, "step": 18946 }, { "epoch": 0.96, "learning_rate": 7.343855064933847e-08, "loss": 0.7789, "step": 18947 }, { "epoch": 0.96, "learning_rate": 7.323963880777385e-08, "loss": 0.8964, "step": 18948 }, { "epoch": 0.96, "learning_rate": 7.304099572268297e-08, "loss": 0.9186, "step": 18949 }, { "epoch": 0.96, "learning_rate": 7.28426213994471e-08, "loss": 0.9144, "step": 18950 }, { "epoch": 0.96, "learning_rate": 7.264451584343524e-08, "loss": 0.7913, "step": 18951 }, { "epoch": 0.96, "learning_rate": 7.244667906001202e-08, "loss": 0.9409, "step": 18952 }, { "epoch": 0.96, "learning_rate": 7.224911105453092e-08, "loss": 0.8609, "step": 18953 }, { "epoch": 0.96, "learning_rate": 7.205181183234433e-08, "loss": 0.9298, "step": 18954 }, { "epoch": 0.96, "learning_rate": 7.185478139879243e-08, "loss": 0.871, "step": 18955 }, { "epoch": 0.96, "learning_rate": 7.165801975920871e-08, "loss": 0.8386, "step": 18956 }, { "epoch": 0.96, "learning_rate": 7.146152691892227e-08, "loss": 0.9148, "step": 18957 }, { "epoch": 0.96, "learning_rate": 7.126530288325106e-08, "loss": 0.8798, "step": 18958 }, { "epoch": 0.96, "learning_rate": 7.106934765750861e-08, "loss": 0.9482, "step": 18959 }, { "epoch": 0.96, "learning_rate": 7.087366124700067e-08, "loss": 0.8064, "step": 18960 }, { "epoch": 0.96, "learning_rate": 7.067824365702302e-08, "loss": 0.8624, "step": 18961 }, { "epoch": 0.96, "learning_rate": 7.04830948928692e-08, "loss": 0.9689, "step": 18962 }, { "epoch": 0.96, "learning_rate": 7.028821495982163e-08, "loss": 0.8754, "step": 18963 }, { "epoch": 0.96, "learning_rate": 7.0093603863155e-08, "loss": 0.8546, "step": 18964 }, { "epoch": 0.96, "learning_rate": 6.989926160814176e-08, "loss": 0.9159, "step": 18965 }, { "epoch": 0.96, "learning_rate": 6.97051882000388e-08, "loss": 0.9279, "step": 18966 }, { "epoch": 0.96, "learning_rate": 6.951138364410415e-08, "loss": 0.8672, "step": 18967 }, { "epoch": 0.96, "learning_rate": 6.931784794558249e-08, "loss": 0.8289, "step": 18968 }, { "epoch": 0.96, "learning_rate": 6.912458110971632e-08, "loss": 0.8543, "step": 18969 }, { "epoch": 0.96, "learning_rate": 6.893158314173587e-08, "loss": 0.9304, "step": 18970 }, { "epoch": 0.96, "learning_rate": 6.873885404686586e-08, "loss": 0.7804, "step": 18971 }, { "epoch": 0.96, "learning_rate": 6.854639383032658e-08, "loss": 0.8022, "step": 18972 }, { "epoch": 0.96, "learning_rate": 6.835420249732716e-08, "loss": 0.8829, "step": 18973 }, { "epoch": 0.96, "learning_rate": 6.816228005307124e-08, "loss": 0.8818, "step": 18974 }, { "epoch": 0.96, "learning_rate": 6.797062650275466e-08, "loss": 0.884, "step": 18975 }, { "epoch": 0.96, "learning_rate": 6.777924185156659e-08, "loss": 0.8806, "step": 18976 }, { "epoch": 0.96, "learning_rate": 6.758812610468846e-08, "loss": 0.8358, "step": 18977 }, { "epoch": 0.96, "learning_rate": 6.7397279267295e-08, "loss": 0.8389, "step": 18978 }, { "epoch": 0.96, "learning_rate": 6.720670134455209e-08, "loss": 0.8511, "step": 18979 }, { "epoch": 0.96, "learning_rate": 6.701639234162116e-08, "loss": 0.9115, "step": 18980 }, { "epoch": 0.96, "learning_rate": 6.682635226365253e-08, "loss": 0.9478, "step": 18981 }, { "epoch": 0.96, "learning_rate": 6.663658111579318e-08, "loss": 0.8512, "step": 18982 }, { "epoch": 0.96, "learning_rate": 6.644707890317903e-08, "loss": 0.8083, "step": 18983 }, { "epoch": 0.96, "learning_rate": 6.625784563094262e-08, "loss": 0.7794, "step": 18984 }, { "epoch": 0.96, "learning_rate": 6.606888130420541e-08, "loss": 0.8263, "step": 18985 }, { "epoch": 0.96, "learning_rate": 6.588018592808554e-08, "loss": 0.927, "step": 18986 }, { "epoch": 0.96, "learning_rate": 6.569175950769002e-08, "loss": 0.8676, "step": 18987 }, { "epoch": 0.96, "learning_rate": 6.550360204812145e-08, "loss": 0.8666, "step": 18988 }, { "epoch": 0.96, "learning_rate": 6.531571355447242e-08, "loss": 0.8892, "step": 18989 }, { "epoch": 0.96, "learning_rate": 6.512809403183107e-08, "loss": 0.8671, "step": 18990 }, { "epoch": 0.96, "learning_rate": 6.494074348527668e-08, "loss": 0.8767, "step": 18991 }, { "epoch": 0.96, "learning_rate": 6.475366191988186e-08, "loss": 0.729, "step": 18992 }, { "epoch": 0.96, "learning_rate": 6.456684934071145e-08, "loss": 0.9508, "step": 18993 }, { "epoch": 0.96, "learning_rate": 6.438030575282362e-08, "loss": 0.8681, "step": 18994 }, { "epoch": 0.96, "learning_rate": 6.419403116126654e-08, "loss": 0.8255, "step": 18995 }, { "epoch": 0.97, "learning_rate": 6.400802557108732e-08, "loss": 0.9553, "step": 18996 }, { "epoch": 0.97, "learning_rate": 6.382228898731746e-08, "loss": 0.7695, "step": 18997 }, { "epoch": 0.97, "learning_rate": 6.363682141498962e-08, "loss": 0.7956, "step": 18998 }, { "epoch": 0.97, "learning_rate": 6.34516228591231e-08, "loss": 0.8503, "step": 18999 }, { "epoch": 0.97, "learning_rate": 6.326669332473167e-08, "loss": 0.7619, "step": 19000 }, { "epoch": 0.97, "learning_rate": 6.308203281682356e-08, "loss": 0.8509, "step": 19001 }, { "epoch": 0.97, "learning_rate": 6.289764134039699e-08, "loss": 0.8883, "step": 19002 }, { "epoch": 0.97, "learning_rate": 6.271351890044463e-08, "loss": 0.8045, "step": 19003 }, { "epoch": 0.97, "learning_rate": 6.252966550195028e-08, "loss": 0.9088, "step": 19004 }, { "epoch": 0.97, "learning_rate": 6.234608114989438e-08, "loss": 0.7863, "step": 19005 }, { "epoch": 0.97, "learning_rate": 6.216276584924408e-08, "loss": 0.8894, "step": 19006 }, { "epoch": 0.97, "learning_rate": 6.19797196049643e-08, "loss": 0.8076, "step": 19007 }, { "epoch": 0.97, "learning_rate": 6.179694242200995e-08, "loss": 0.858, "step": 19008 }, { "epoch": 0.97, "learning_rate": 6.161443430533043e-08, "loss": 0.9069, "step": 19009 }, { "epoch": 0.97, "learning_rate": 6.143219525986621e-08, "loss": 0.8819, "step": 19010 }, { "epoch": 0.97, "learning_rate": 6.125022529055225e-08, "loss": 0.8062, "step": 19011 }, { "epoch": 0.97, "learning_rate": 6.106852440231347e-08, "loss": 0.8483, "step": 19012 }, { "epoch": 0.97, "learning_rate": 6.088709260007153e-08, "loss": 0.8771, "step": 19013 }, { "epoch": 0.97, "learning_rate": 6.07059298887358e-08, "loss": 0.8578, "step": 19014 }, { "epoch": 0.97, "learning_rate": 6.052503627321238e-08, "loss": 0.8888, "step": 19015 }, { "epoch": 0.97, "learning_rate": 6.034441175839956e-08, "loss": 0.87, "step": 19016 }, { "epoch": 0.97, "learning_rate": 6.016405634918676e-08, "loss": 0.9877, "step": 19017 }, { "epoch": 0.97, "learning_rate": 5.998397005045786e-08, "loss": 0.9407, "step": 19018 }, { "epoch": 0.97, "learning_rate": 5.980415286708562e-08, "loss": 0.8294, "step": 19019 }, { "epoch": 0.97, "learning_rate": 5.962460480394172e-08, "loss": 0.829, "step": 19020 }, { "epoch": 0.97, "learning_rate": 5.9445325865885586e-08, "loss": 0.7554, "step": 19021 }, { "epoch": 0.97, "learning_rate": 5.9266316057772224e-08, "loss": 0.7634, "step": 19022 }, { "epoch": 0.97, "learning_rate": 5.908757538444554e-08, "loss": 0.9247, "step": 19023 }, { "epoch": 0.97, "learning_rate": 5.890910385074722e-08, "loss": 0.8473, "step": 19024 }, { "epoch": 0.97, "learning_rate": 5.873090146150895e-08, "loss": 0.9131, "step": 19025 }, { "epoch": 0.97, "learning_rate": 5.855296822155354e-08, "loss": 0.8402, "step": 19026 }, { "epoch": 0.97, "learning_rate": 5.837530413570158e-08, "loss": 0.8453, "step": 19027 }, { "epoch": 0.97, "learning_rate": 5.819790920876034e-08, "loss": 0.8989, "step": 19028 }, { "epoch": 0.97, "learning_rate": 5.8020783445532634e-08, "loss": 0.9055, "step": 19029 }, { "epoch": 0.97, "learning_rate": 5.7843926850815745e-08, "loss": 0.8738, "step": 19030 }, { "epoch": 0.97, "learning_rate": 5.766733942939695e-08, "loss": 0.763, "step": 19031 }, { "epoch": 0.97, "learning_rate": 5.749102118605576e-08, "loss": 0.9415, "step": 19032 }, { "epoch": 0.97, "learning_rate": 5.731497212556836e-08, "loss": 0.9409, "step": 19033 }, { "epoch": 0.97, "learning_rate": 5.7139192252700926e-08, "loss": 0.8762, "step": 19034 }, { "epoch": 0.97, "learning_rate": 5.696368157220966e-08, "loss": 0.8562, "step": 19035 }, { "epoch": 0.97, "learning_rate": 5.678844008884854e-08, "loss": 0.8821, "step": 19036 }, { "epoch": 0.97, "learning_rate": 5.6613467807362654e-08, "loss": 0.828, "step": 19037 }, { "epoch": 0.97, "learning_rate": 5.6438764732487104e-08, "loss": 0.8, "step": 19038 }, { "epoch": 0.97, "learning_rate": 5.626433086895366e-08, "loss": 0.8036, "step": 19039 }, { "epoch": 0.97, "learning_rate": 5.6090166221483e-08, "loss": 0.9355, "step": 19040 }, { "epoch": 0.97, "learning_rate": 5.591627079479245e-08, "loss": 0.8488, "step": 19041 }, { "epoch": 0.97, "learning_rate": 5.574264459358936e-08, "loss": 0.9197, "step": 19042 }, { "epoch": 0.97, "learning_rate": 5.5569287622574406e-08, "loss": 0.875, "step": 19043 }, { "epoch": 0.97, "learning_rate": 5.5396199886440516e-08, "loss": 0.8839, "step": 19044 }, { "epoch": 0.97, "learning_rate": 5.522338138987393e-08, "loss": 0.8629, "step": 19045 }, { "epoch": 0.97, "learning_rate": 5.505083213755425e-08, "loss": 0.9232, "step": 19046 }, { "epoch": 0.97, "learning_rate": 5.487855213415327e-08, "loss": 0.8451, "step": 19047 }, { "epoch": 0.97, "learning_rate": 5.470654138433396e-08, "loss": 0.8548, "step": 19048 }, { "epoch": 0.97, "learning_rate": 5.4534799892754784e-08, "loss": 0.771, "step": 19049 }, { "epoch": 0.97, "learning_rate": 5.436332766406316e-08, "loss": 0.8411, "step": 19050 }, { "epoch": 0.97, "learning_rate": 5.419212470290425e-08, "loss": 0.8097, "step": 19051 }, { "epoch": 0.97, "learning_rate": 5.402119101391212e-08, "loss": 0.8496, "step": 19052 }, { "epoch": 0.97, "learning_rate": 5.38505266017153e-08, "loss": 0.871, "step": 19053 }, { "epoch": 0.97, "learning_rate": 5.368013147093232e-08, "loss": 0.8343, "step": 19054 }, { "epoch": 0.97, "learning_rate": 5.351000562617836e-08, "loss": 0.8681, "step": 19055 }, { "epoch": 0.97, "learning_rate": 5.3340149072058646e-08, "loss": 0.8953, "step": 19056 }, { "epoch": 0.97, "learning_rate": 5.317056181317171e-08, "loss": 0.8081, "step": 19057 }, { "epoch": 0.97, "learning_rate": 5.300124385410943e-08, "loss": 0.8947, "step": 19058 }, { "epoch": 0.97, "learning_rate": 5.283219519945593e-08, "loss": 0.8162, "step": 19059 }, { "epoch": 0.97, "learning_rate": 5.2663415853787535e-08, "loss": 0.8019, "step": 19060 }, { "epoch": 0.97, "learning_rate": 5.249490582167393e-08, "loss": 0.8925, "step": 19061 }, { "epoch": 0.97, "learning_rate": 5.232666510767703e-08, "loss": 0.8651, "step": 19062 }, { "epoch": 0.97, "learning_rate": 5.215869371635318e-08, "loss": 0.8857, "step": 19063 }, { "epoch": 0.97, "learning_rate": 5.199099165224875e-08, "loss": 0.8859, "step": 19064 }, { "epoch": 0.97, "learning_rate": 5.1823558919904534e-08, "loss": 0.8376, "step": 19065 }, { "epoch": 0.97, "learning_rate": 5.16563955238536e-08, "loss": 0.8286, "step": 19066 }, { "epoch": 0.97, "learning_rate": 5.148950146862009e-08, "loss": 0.9045, "step": 19067 }, { "epoch": 0.97, "learning_rate": 5.132287675872483e-08, "loss": 0.8989, "step": 19068 }, { "epoch": 0.97, "learning_rate": 5.115652139867866e-08, "loss": 0.8087, "step": 19069 }, { "epoch": 0.97, "learning_rate": 5.0990435392984647e-08, "loss": 0.878, "step": 19070 }, { "epoch": 0.97, "learning_rate": 5.08246187461392e-08, "loss": 0.9312, "step": 19071 }, { "epoch": 0.97, "learning_rate": 5.065907146263205e-08, "loss": 0.9072, "step": 19072 }, { "epoch": 0.97, "learning_rate": 5.049379354694628e-08, "loss": 0.8519, "step": 19073 }, { "epoch": 0.97, "learning_rate": 5.032878500355498e-08, "loss": 0.8917, "step": 19074 }, { "epoch": 0.97, "learning_rate": 5.016404583692569e-08, "loss": 0.8611, "step": 19075 }, { "epoch": 0.97, "learning_rate": 4.999957605152039e-08, "loss": 0.8061, "step": 19076 }, { "epoch": 0.97, "learning_rate": 4.983537565178886e-08, "loss": 0.7982, "step": 19077 }, { "epoch": 0.97, "learning_rate": 4.967144464217977e-08, "loss": 0.8834, "step": 19078 }, { "epoch": 0.97, "learning_rate": 4.9507783027129555e-08, "loss": 0.8763, "step": 19079 }, { "epoch": 0.97, "learning_rate": 4.934439081107023e-08, "loss": 0.8686, "step": 19080 }, { "epoch": 0.97, "learning_rate": 4.918126799842382e-08, "loss": 0.8752, "step": 19081 }, { "epoch": 0.97, "learning_rate": 4.90184145936079e-08, "loss": 0.8501, "step": 19082 }, { "epoch": 0.97, "learning_rate": 4.8855830601032275e-08, "loss": 0.8805, "step": 19083 }, { "epoch": 0.97, "learning_rate": 4.8693516025096756e-08, "loss": 0.8479, "step": 19084 }, { "epoch": 0.97, "learning_rate": 4.8531470870197827e-08, "loss": 0.9283, "step": 19085 }, { "epoch": 0.97, "learning_rate": 4.8369695140720873e-08, "loss": 0.9055, "step": 19086 }, { "epoch": 0.97, "learning_rate": 4.820818884104794e-08, "loss": 0.7896, "step": 19087 }, { "epoch": 0.97, "learning_rate": 4.804695197554887e-08, "loss": 0.8515, "step": 19088 }, { "epoch": 0.97, "learning_rate": 4.788598454859128e-08, "loss": 0.7909, "step": 19089 }, { "epoch": 0.97, "learning_rate": 4.772528656453279e-08, "loss": 0.8459, "step": 19090 }, { "epoch": 0.97, "learning_rate": 4.756485802772437e-08, "loss": 0.9873, "step": 19091 }, { "epoch": 0.97, "learning_rate": 4.740469894250699e-08, "loss": 0.9092, "step": 19092 }, { "epoch": 0.97, "learning_rate": 4.724480931322051e-08, "loss": 0.8195, "step": 19093 }, { "epoch": 0.97, "learning_rate": 4.708518914419036e-08, "loss": 0.8662, "step": 19094 }, { "epoch": 0.97, "learning_rate": 4.6925838439740854e-08, "loss": 0.7626, "step": 19095 }, { "epoch": 0.97, "learning_rate": 4.676675720418411e-08, "loss": 0.8939, "step": 19096 }, { "epoch": 0.97, "learning_rate": 4.66079454418289e-08, "loss": 0.8659, "step": 19097 }, { "epoch": 0.97, "learning_rate": 4.644940315697288e-08, "loss": 0.9557, "step": 19098 }, { "epoch": 0.97, "learning_rate": 4.629113035391153e-08, "loss": 0.8315, "step": 19099 }, { "epoch": 0.97, "learning_rate": 4.613312703692585e-08, "loss": 0.8108, "step": 19100 }, { "epoch": 0.97, "learning_rate": 4.597539321029687e-08, "loss": 0.8584, "step": 19101 }, { "epoch": 0.97, "learning_rate": 4.58179288782934e-08, "loss": 0.7887, "step": 19102 }, { "epoch": 0.97, "learning_rate": 4.566073404517868e-08, "loss": 0.9125, "step": 19103 }, { "epoch": 0.97, "learning_rate": 4.5503808715209316e-08, "loss": 0.9495, "step": 19104 }, { "epoch": 0.97, "learning_rate": 4.534715289263414e-08, "loss": 0.907, "step": 19105 }, { "epoch": 0.97, "learning_rate": 4.519076658169308e-08, "loss": 0.8468, "step": 19106 }, { "epoch": 0.97, "learning_rate": 4.503464978662164e-08, "loss": 0.9271, "step": 19107 }, { "epoch": 0.97, "learning_rate": 4.487880251164534e-08, "loss": 0.8685, "step": 19108 }, { "epoch": 0.97, "learning_rate": 4.4723224760984116e-08, "loss": 0.8526, "step": 19109 }, { "epoch": 0.97, "learning_rate": 4.456791653885017e-08, "loss": 0.8428, "step": 19110 }, { "epoch": 0.97, "learning_rate": 4.4412877849447923e-08, "loss": 0.8387, "step": 19111 }, { "epoch": 0.97, "learning_rate": 4.425810869697511e-08, "loss": 0.868, "step": 19112 }, { "epoch": 0.97, "learning_rate": 4.4103609085621723e-08, "loss": 0.872, "step": 19113 }, { "epoch": 0.97, "learning_rate": 4.394937901957219e-08, "loss": 0.8339, "step": 19114 }, { "epoch": 0.97, "learning_rate": 4.3795418502998734e-08, "loss": 0.8972, "step": 19115 }, { "epoch": 0.97, "learning_rate": 4.3641727540073564e-08, "loss": 0.9398, "step": 19116 }, { "epoch": 0.97, "learning_rate": 4.3488306134954474e-08, "loss": 0.7564, "step": 19117 }, { "epoch": 0.97, "learning_rate": 4.333515429179702e-08, "loss": 0.8687, "step": 19118 }, { "epoch": 0.97, "learning_rate": 4.318227201474789e-08, "loss": 0.8555, "step": 19119 }, { "epoch": 0.97, "learning_rate": 4.302965930794489e-08, "loss": 0.8861, "step": 19120 }, { "epoch": 0.97, "learning_rate": 4.2877316175521375e-08, "loss": 0.9518, "step": 19121 }, { "epoch": 0.97, "learning_rate": 4.2725242621599607e-08, "loss": 0.7834, "step": 19122 }, { "epoch": 0.97, "learning_rate": 4.257343865029961e-08, "loss": 0.8529, "step": 19123 }, { "epoch": 0.97, "learning_rate": 4.242190426572812e-08, "loss": 0.8469, "step": 19124 }, { "epoch": 0.97, "learning_rate": 4.227063947199073e-08, "loss": 0.8923, "step": 19125 }, { "epoch": 0.97, "learning_rate": 4.211964427318194e-08, "loss": 0.8407, "step": 19126 }, { "epoch": 0.97, "learning_rate": 4.1968918673388484e-08, "loss": 0.8246, "step": 19127 }, { "epoch": 0.97, "learning_rate": 4.181846267669265e-08, "loss": 0.8355, "step": 19128 }, { "epoch": 0.97, "learning_rate": 4.166827628716674e-08, "loss": 0.8345, "step": 19129 }, { "epoch": 0.97, "learning_rate": 4.15183595088775e-08, "loss": 0.8817, "step": 19130 }, { "epoch": 0.97, "learning_rate": 4.136871234588502e-08, "loss": 0.9042, "step": 19131 }, { "epoch": 0.97, "learning_rate": 4.121933480223828e-08, "loss": 0.8914, "step": 19132 }, { "epoch": 0.97, "learning_rate": 4.107022688198292e-08, "loss": 0.9515, "step": 19133 }, { "epoch": 0.97, "learning_rate": 4.092138858915573e-08, "loss": 0.9443, "step": 19134 }, { "epoch": 0.97, "learning_rate": 4.07728199277857e-08, "loss": 0.821, "step": 19135 }, { "epoch": 0.97, "learning_rate": 4.062452090189628e-08, "loss": 0.7169, "step": 19136 }, { "epoch": 0.97, "learning_rate": 4.0476491515502034e-08, "loss": 0.8516, "step": 19137 }, { "epoch": 0.97, "learning_rate": 4.032873177261087e-08, "loss": 0.819, "step": 19138 }, { "epoch": 0.97, "learning_rate": 4.018124167722293e-08, "loss": 0.8336, "step": 19139 }, { "epoch": 0.97, "learning_rate": 4.0034021233331666e-08, "loss": 0.7352, "step": 19140 }, { "epoch": 0.97, "learning_rate": 3.988707044492279e-08, "loss": 0.8954, "step": 19141 }, { "epoch": 0.97, "learning_rate": 3.974038931597424e-08, "loss": 0.8237, "step": 19142 }, { "epoch": 0.97, "learning_rate": 3.9593977850458375e-08, "loss": 0.8767, "step": 19143 }, { "epoch": 0.97, "learning_rate": 3.9447836052337594e-08, "loss": 0.8078, "step": 19144 }, { "epoch": 0.97, "learning_rate": 3.9301963925569844e-08, "loss": 0.8646, "step": 19145 }, { "epoch": 0.97, "learning_rate": 3.9156361474104176e-08, "loss": 0.8018, "step": 19146 }, { "epoch": 0.97, "learning_rate": 3.9011028701883006e-08, "loss": 0.8828, "step": 19147 }, { "epoch": 0.97, "learning_rate": 3.886596561283984e-08, "loss": 0.8829, "step": 19148 }, { "epoch": 0.97, "learning_rate": 3.8721172210902656e-08, "loss": 0.8907, "step": 19149 }, { "epoch": 0.97, "learning_rate": 3.8576648499992765e-08, "loss": 0.8209, "step": 19150 }, { "epoch": 0.97, "learning_rate": 3.8432394484021476e-08, "loss": 0.8289, "step": 19151 }, { "epoch": 0.97, "learning_rate": 3.828841016689455e-08, "loss": 0.8523, "step": 19152 }, { "epoch": 0.97, "learning_rate": 3.81446955525111e-08, "loss": 0.8936, "step": 19153 }, { "epoch": 0.97, "learning_rate": 3.8001250644761345e-08, "loss": 0.9198, "step": 19154 }, { "epoch": 0.97, "learning_rate": 3.785807544752884e-08, "loss": 0.8881, "step": 19155 }, { "epoch": 0.97, "learning_rate": 3.771516996468938e-08, "loss": 0.775, "step": 19156 }, { "epoch": 0.97, "learning_rate": 3.75725342001132e-08, "loss": 0.8213, "step": 19157 }, { "epoch": 0.97, "learning_rate": 3.743016815766165e-08, "loss": 0.7866, "step": 19158 }, { "epoch": 0.97, "learning_rate": 3.7288071841189435e-08, "loss": 0.8605, "step": 19159 }, { "epoch": 0.97, "learning_rate": 3.7146245254542357e-08, "loss": 0.9033, "step": 19160 }, { "epoch": 0.97, "learning_rate": 3.7004688401560684e-08, "loss": 0.9745, "step": 19161 }, { "epoch": 0.97, "learning_rate": 3.6863401286078015e-08, "loss": 0.8831, "step": 19162 }, { "epoch": 0.97, "learning_rate": 3.6722383911917957e-08, "loss": 0.9105, "step": 19163 }, { "epoch": 0.97, "learning_rate": 3.658163628290079e-08, "loss": 0.8209, "step": 19164 }, { "epoch": 0.97, "learning_rate": 3.6441158402834575e-08, "loss": 0.9117, "step": 19165 }, { "epoch": 0.97, "learning_rate": 3.630095027552405e-08, "loss": 0.9296, "step": 19166 }, { "epoch": 0.97, "learning_rate": 3.6161011904763956e-08, "loss": 0.9614, "step": 19167 }, { "epoch": 0.97, "learning_rate": 3.602134329434459e-08, "loss": 0.7942, "step": 19168 }, { "epoch": 0.97, "learning_rate": 3.588194444804627e-08, "loss": 0.8439, "step": 19169 }, { "epoch": 0.97, "learning_rate": 3.574281536964264e-08, "loss": 0.9099, "step": 19170 }, { "epoch": 0.97, "learning_rate": 3.560395606290179e-08, "loss": 0.8293, "step": 19171 }, { "epoch": 0.97, "learning_rate": 3.5465366531582947e-08, "loss": 0.8038, "step": 19172 }, { "epoch": 0.97, "learning_rate": 3.5327046779437545e-08, "loss": 0.8874, "step": 19173 }, { "epoch": 0.97, "learning_rate": 3.518899681021038e-08, "loss": 0.8131, "step": 19174 }, { "epoch": 0.97, "learning_rate": 3.505121662763955e-08, "loss": 0.9412, "step": 19175 }, { "epoch": 0.97, "learning_rate": 3.491370623545542e-08, "loss": 0.8932, "step": 19176 }, { "epoch": 0.97, "learning_rate": 3.477646563738057e-08, "loss": 0.7909, "step": 19177 }, { "epoch": 0.97, "learning_rate": 3.4639494837130916e-08, "loss": 0.8606, "step": 19178 }, { "epoch": 0.97, "learning_rate": 3.45027938384146e-08, "loss": 0.8401, "step": 19179 }, { "epoch": 0.97, "learning_rate": 3.4366362644931986e-08, "loss": 0.8607, "step": 19180 }, { "epoch": 0.97, "learning_rate": 3.423020126037902e-08, "loss": 0.9476, "step": 19181 }, { "epoch": 0.97, "learning_rate": 3.409430968843941e-08, "loss": 0.8423, "step": 19182 }, { "epoch": 0.97, "learning_rate": 3.3958687932793554e-08, "loss": 0.846, "step": 19183 }, { "epoch": 0.97, "learning_rate": 3.382333599711407e-08, "loss": 0.8373, "step": 19184 }, { "epoch": 0.97, "learning_rate": 3.368825388506358e-08, "loss": 0.9351, "step": 19185 }, { "epoch": 0.97, "learning_rate": 3.355344160030027e-08, "loss": 0.9141, "step": 19186 }, { "epoch": 0.97, "learning_rate": 3.341889914647456e-08, "loss": 0.9101, "step": 19187 }, { "epoch": 0.97, "learning_rate": 3.328462652722797e-08, "loss": 0.9326, "step": 19188 }, { "epoch": 0.97, "learning_rate": 3.31506237461976e-08, "loss": 0.8674, "step": 19189 }, { "epoch": 0.97, "learning_rate": 3.301689080700832e-08, "loss": 0.8214, "step": 19190 }, { "epoch": 0.97, "learning_rate": 3.288342771328501e-08, "loss": 0.8147, "step": 19191 }, { "epoch": 0.98, "learning_rate": 3.275023446863701e-08, "loss": 0.8445, "step": 19192 }, { "epoch": 0.98, "learning_rate": 3.261731107667254e-08, "loss": 0.9155, "step": 19193 }, { "epoch": 0.98, "learning_rate": 3.248465754098984e-08, "loss": 0.79, "step": 19194 }, { "epoch": 0.98, "learning_rate": 3.235227386518047e-08, "loss": 0.8416, "step": 19195 }, { "epoch": 0.98, "learning_rate": 3.2220160052828245e-08, "loss": 0.8822, "step": 19196 }, { "epoch": 0.98, "learning_rate": 3.208831610751029e-08, "loss": 0.8551, "step": 19197 }, { "epoch": 0.98, "learning_rate": 3.1956742032797086e-08, "loss": 0.8906, "step": 19198 }, { "epoch": 0.98, "learning_rate": 3.182543783224912e-08, "loss": 0.8558, "step": 19199 }, { "epoch": 0.98, "learning_rate": 3.1694403509422435e-08, "loss": 0.894, "step": 19200 }, { "epoch": 0.98, "learning_rate": 3.156363906786419e-08, "loss": 0.851, "step": 19201 }, { "epoch": 0.98, "learning_rate": 3.14331445111149e-08, "loss": 0.9518, "step": 19202 }, { "epoch": 0.98, "learning_rate": 3.130291984270728e-08, "loss": 0.8624, "step": 19203 }, { "epoch": 0.98, "learning_rate": 3.1172965066167404e-08, "loss": 0.8352, "step": 19204 }, { "epoch": 0.98, "learning_rate": 3.104328018501357e-08, "loss": 0.8452, "step": 19205 }, { "epoch": 0.98, "learning_rate": 3.091386520275741e-08, "loss": 0.881, "step": 19206 }, { "epoch": 0.98, "learning_rate": 3.078472012290057e-08, "loss": 0.7987, "step": 19207 }, { "epoch": 0.98, "learning_rate": 3.065584494894247e-08, "loss": 0.7608, "step": 19208 }, { "epoch": 0.98, "learning_rate": 3.052723968437033e-08, "loss": 0.9135, "step": 19209 }, { "epoch": 0.98, "learning_rate": 3.039890433266579e-08, "loss": 0.8096, "step": 19210 }, { "epoch": 0.98, "learning_rate": 3.0270838897304976e-08, "loss": 0.8606, "step": 19211 }, { "epoch": 0.98, "learning_rate": 3.0143043381754e-08, "loss": 0.8483, "step": 19212 }, { "epoch": 0.98, "learning_rate": 3.001551778947342e-08, "loss": 0.8517, "step": 19213 }, { "epoch": 0.98, "learning_rate": 2.988826212391494e-08, "loss": 0.8012, "step": 19214 }, { "epoch": 0.98, "learning_rate": 2.9761276388524662e-08, "loss": 0.7893, "step": 19215 }, { "epoch": 0.98, "learning_rate": 2.9634560586739857e-08, "loss": 0.8356, "step": 19216 }, { "epoch": 0.98, "learning_rate": 2.950811472199222e-08, "loss": 0.7671, "step": 19217 }, { "epoch": 0.98, "learning_rate": 2.938193879770457e-08, "loss": 0.9292, "step": 19218 }, { "epoch": 0.98, "learning_rate": 2.9256032817291947e-08, "loss": 0.8747, "step": 19219 }, { "epoch": 0.98, "learning_rate": 2.913039678416607e-08, "loss": 0.8927, "step": 19220 }, { "epoch": 0.98, "learning_rate": 2.9005030701725335e-08, "loss": 0.988, "step": 19221 }, { "epoch": 0.98, "learning_rate": 2.8879934573364798e-08, "loss": 0.8792, "step": 19222 }, { "epoch": 0.98, "learning_rate": 2.8755108402471755e-08, "loss": 0.9226, "step": 19223 }, { "epoch": 0.98, "learning_rate": 2.863055219242683e-08, "loss": 0.8614, "step": 19224 }, { "epoch": 0.98, "learning_rate": 2.8506265946599555e-08, "loss": 0.7048, "step": 19225 }, { "epoch": 0.98, "learning_rate": 2.8382249668357232e-08, "loss": 0.8221, "step": 19226 }, { "epoch": 0.98, "learning_rate": 2.8258503361056068e-08, "loss": 0.8194, "step": 19227 }, { "epoch": 0.98, "learning_rate": 2.8135027028046712e-08, "loss": 0.8339, "step": 19228 }, { "epoch": 0.98, "learning_rate": 2.801182067267205e-08, "loss": 1.0151, "step": 19229 }, { "epoch": 0.98, "learning_rate": 2.7888884298268304e-08, "loss": 0.8315, "step": 19230 }, { "epoch": 0.98, "learning_rate": 2.7766217908163916e-08, "loss": 0.8654, "step": 19231 }, { "epoch": 0.98, "learning_rate": 2.7643821505679568e-08, "loss": 0.8037, "step": 19232 }, { "epoch": 0.98, "learning_rate": 2.7521695094128166e-08, "loss": 0.8607, "step": 19233 }, { "epoch": 0.98, "learning_rate": 2.7399838676818168e-08, "loss": 0.9055, "step": 19234 }, { "epoch": 0.98, "learning_rate": 2.7278252257046943e-08, "loss": 0.8418, "step": 19235 }, { "epoch": 0.98, "learning_rate": 2.71569358381063e-08, "loss": 0.9207, "step": 19236 }, { "epoch": 0.98, "learning_rate": 2.7035889423282503e-08, "loss": 0.8439, "step": 19237 }, { "epoch": 0.98, "learning_rate": 2.6915113015850702e-08, "loss": 0.8557, "step": 19238 }, { "epoch": 0.98, "learning_rate": 2.679460661908162e-08, "loss": 0.8753, "step": 19239 }, { "epoch": 0.98, "learning_rate": 2.66743702362382e-08, "loss": 0.8701, "step": 19240 }, { "epoch": 0.98, "learning_rate": 2.655440387057562e-08, "loss": 0.9393, "step": 19241 }, { "epoch": 0.98, "learning_rate": 2.6434707525341273e-08, "loss": 0.7909, "step": 19242 }, { "epoch": 0.98, "learning_rate": 2.631528120377702e-08, "loss": 0.8153, "step": 19243 }, { "epoch": 0.98, "learning_rate": 2.6196124909114717e-08, "loss": 0.882, "step": 19244 }, { "epoch": 0.98, "learning_rate": 2.607723864458067e-08, "loss": 0.8507, "step": 19245 }, { "epoch": 0.98, "learning_rate": 2.595862241339453e-08, "loss": 0.787, "step": 19246 }, { "epoch": 0.98, "learning_rate": 2.5840276218767058e-08, "loss": 0.9214, "step": 19247 }, { "epoch": 0.98, "learning_rate": 2.5722200063902357e-08, "loss": 0.8594, "step": 19248 }, { "epoch": 0.98, "learning_rate": 2.5604393951997874e-08, "loss": 0.8037, "step": 19249 }, { "epoch": 0.98, "learning_rate": 2.5486857886241056e-08, "loss": 0.8387, "step": 19250 }, { "epoch": 0.98, "learning_rate": 2.5369591869816023e-08, "loss": 0.9434, "step": 19251 }, { "epoch": 0.98, "learning_rate": 2.5252595905898013e-08, "loss": 0.8002, "step": 19252 }, { "epoch": 0.98, "learning_rate": 2.5135869997653384e-08, "loss": 0.9206, "step": 19253 }, { "epoch": 0.98, "learning_rate": 2.5019414148241827e-08, "loss": 0.7645, "step": 19254 }, { "epoch": 0.98, "learning_rate": 2.4903228360817488e-08, "loss": 0.8419, "step": 19255 }, { "epoch": 0.98, "learning_rate": 2.478731263852674e-08, "loss": 0.9841, "step": 19256 }, { "epoch": 0.98, "learning_rate": 2.467166698450485e-08, "loss": 0.8279, "step": 19257 }, { "epoch": 0.98, "learning_rate": 2.455629140188598e-08, "loss": 0.944, "step": 19258 }, { "epoch": 0.98, "learning_rate": 2.4441185893792074e-08, "loss": 0.7849, "step": 19259 }, { "epoch": 0.98, "learning_rate": 2.432635046333953e-08, "loss": 0.8659, "step": 19260 }, { "epoch": 0.98, "learning_rate": 2.4211785113638085e-08, "loss": 0.7867, "step": 19261 }, { "epoch": 0.98, "learning_rate": 2.4097489847789703e-08, "loss": 0.8675, "step": 19262 }, { "epoch": 0.98, "learning_rate": 2.398346466888746e-08, "loss": 0.8406, "step": 19263 }, { "epoch": 0.98, "learning_rate": 2.3869709580019996e-08, "loss": 0.8338, "step": 19264 }, { "epoch": 0.98, "learning_rate": 2.3756224584265963e-08, "loss": 0.9191, "step": 19265 }, { "epoch": 0.98, "learning_rate": 2.3643009684697347e-08, "loss": 0.8428, "step": 19266 }, { "epoch": 0.98, "learning_rate": 2.3530064884381697e-08, "loss": 0.8487, "step": 19267 }, { "epoch": 0.98, "learning_rate": 2.3417390186374346e-08, "loss": 0.9432, "step": 19268 }, { "epoch": 0.98, "learning_rate": 2.3304985593727292e-08, "loss": 0.8745, "step": 19269 }, { "epoch": 0.98, "learning_rate": 2.319285110948366e-08, "loss": 0.8207, "step": 19270 }, { "epoch": 0.98, "learning_rate": 2.30809867366788e-08, "loss": 0.8378, "step": 19271 }, { "epoch": 0.98, "learning_rate": 2.29693924783414e-08, "loss": 0.8215, "step": 19272 }, { "epoch": 0.98, "learning_rate": 2.28580683374946e-08, "loss": 0.9442, "step": 19273 }, { "epoch": 0.98, "learning_rate": 2.2747014317149318e-08, "loss": 0.7793, "step": 19274 }, { "epoch": 0.98, "learning_rate": 2.2636230420313156e-08, "loss": 0.8868, "step": 19275 }, { "epoch": 0.98, "learning_rate": 2.252571664998704e-08, "loss": 0.8641, "step": 19276 }, { "epoch": 0.98, "learning_rate": 2.2415473009161913e-08, "loss": 0.7573, "step": 19277 }, { "epoch": 0.98, "learning_rate": 2.2305499500822058e-08, "loss": 0.7927, "step": 19278 }, { "epoch": 0.98, "learning_rate": 2.219579612794509e-08, "loss": 0.9557, "step": 19279 }, { "epoch": 0.98, "learning_rate": 2.208636289350197e-08, "loss": 0.8793, "step": 19280 }, { "epoch": 0.98, "learning_rate": 2.1977199800453652e-08, "loss": 0.8943, "step": 19281 }, { "epoch": 0.98, "learning_rate": 2.186830685175889e-08, "loss": 0.8004, "step": 19282 }, { "epoch": 0.98, "learning_rate": 2.1759684050363106e-08, "loss": 0.728, "step": 19283 }, { "epoch": 0.98, "learning_rate": 2.1651331399207276e-08, "loss": 0.7554, "step": 19284 }, { "epoch": 0.98, "learning_rate": 2.154324890122572e-08, "loss": 0.8001, "step": 19285 }, { "epoch": 0.98, "learning_rate": 2.143543655934499e-08, "loss": 0.8459, "step": 19286 }, { "epoch": 0.98, "learning_rate": 2.132789437648275e-08, "loss": 0.8935, "step": 19287 }, { "epoch": 0.98, "learning_rate": 2.1220622355551113e-08, "loss": 0.9157, "step": 19288 }, { "epoch": 0.98, "learning_rate": 2.1113620499455533e-08, "loss": 0.9408, "step": 19289 }, { "epoch": 0.98, "learning_rate": 2.1006888811091475e-08, "loss": 0.8391, "step": 19290 }, { "epoch": 0.98, "learning_rate": 2.0900427293348847e-08, "loss": 0.8553, "step": 19291 }, { "epoch": 0.98, "learning_rate": 2.0794235949110896e-08, "loss": 0.8004, "step": 19292 }, { "epoch": 0.98, "learning_rate": 2.0688314781250883e-08, "loss": 0.8004, "step": 19293 }, { "epoch": 0.98, "learning_rate": 2.0582663792638736e-08, "loss": 0.8924, "step": 19294 }, { "epoch": 0.98, "learning_rate": 2.0477282986133273e-08, "loss": 0.9233, "step": 19295 }, { "epoch": 0.98, "learning_rate": 2.0372172364587773e-08, "loss": 0.9121, "step": 19296 }, { "epoch": 0.98, "learning_rate": 2.0267331930848844e-08, "loss": 0.8402, "step": 19297 }, { "epoch": 0.98, "learning_rate": 2.0162761687753106e-08, "loss": 0.8563, "step": 19298 }, { "epoch": 0.98, "learning_rate": 2.0058461638133852e-08, "loss": 0.8395, "step": 19299 }, { "epoch": 0.98, "learning_rate": 1.9954431784814377e-08, "loss": 0.8525, "step": 19300 }, { "epoch": 0.98, "learning_rate": 1.9850672130610206e-08, "loss": 0.8657, "step": 19301 }, { "epoch": 0.98, "learning_rate": 1.9747182678330202e-08, "loss": 0.8988, "step": 19302 }, { "epoch": 0.98, "learning_rate": 1.964396343077768e-08, "loss": 0.7911, "step": 19303 }, { "epoch": 0.98, "learning_rate": 1.954101439074707e-08, "loss": 0.8487, "step": 19304 }, { "epoch": 0.98, "learning_rate": 1.9438335561023924e-08, "loss": 0.8582, "step": 19305 }, { "epoch": 0.98, "learning_rate": 1.933592694439046e-08, "loss": 0.878, "step": 19306 }, { "epoch": 0.98, "learning_rate": 1.9233788543617792e-08, "loss": 0.8222, "step": 19307 }, { "epoch": 0.98, "learning_rate": 1.91319203614726e-08, "loss": 0.7831, "step": 19308 }, { "epoch": 0.98, "learning_rate": 1.9030322400710455e-08, "loss": 0.89, "step": 19309 }, { "epoch": 0.98, "learning_rate": 1.89289946640836e-08, "loss": 0.8408, "step": 19310 }, { "epoch": 0.98, "learning_rate": 1.8827937154335396e-08, "loss": 0.8675, "step": 19311 }, { "epoch": 0.98, "learning_rate": 1.8727149874202542e-08, "loss": 0.9635, "step": 19312 }, { "epoch": 0.98, "learning_rate": 1.862663282641175e-08, "loss": 0.8348, "step": 19313 }, { "epoch": 0.98, "learning_rate": 1.8526386013686392e-08, "loss": 0.8486, "step": 19314 }, { "epoch": 0.98, "learning_rate": 1.8426409438738745e-08, "loss": 0.8526, "step": 19315 }, { "epoch": 0.98, "learning_rate": 1.8326703104276645e-08, "loss": 0.9154, "step": 19316 }, { "epoch": 0.98, "learning_rate": 1.822726701300015e-08, "loss": 0.8011, "step": 19317 }, { "epoch": 0.98, "learning_rate": 1.812810116760044e-08, "loss": 0.9151, "step": 19318 }, { "epoch": 0.98, "learning_rate": 1.8029205570762042e-08, "loss": 0.817, "step": 19319 }, { "epoch": 0.98, "learning_rate": 1.7930580225162807e-08, "loss": 0.9645, "step": 19320 }, { "epoch": 0.98, "learning_rate": 1.7832225133473936e-08, "loss": 0.8802, "step": 19321 }, { "epoch": 0.98, "learning_rate": 1.7734140298355517e-08, "loss": 0.8548, "step": 19322 }, { "epoch": 0.98, "learning_rate": 1.763632572246654e-08, "loss": 0.8347, "step": 19323 }, { "epoch": 0.98, "learning_rate": 1.7538781408453775e-08, "loss": 0.9081, "step": 19324 }, { "epoch": 0.98, "learning_rate": 1.7441507358957333e-08, "loss": 0.8872, "step": 19325 }, { "epoch": 0.98, "learning_rate": 1.734450357661177e-08, "loss": 0.9156, "step": 19326 }, { "epoch": 0.98, "learning_rate": 1.7247770064042768e-08, "loss": 0.877, "step": 19327 }, { "epoch": 0.98, "learning_rate": 1.715130682387045e-08, "loss": 0.8618, "step": 19328 }, { "epoch": 0.98, "learning_rate": 1.7055113858703843e-08, "loss": 0.8256, "step": 19329 }, { "epoch": 0.98, "learning_rate": 1.6959191171150856e-08, "loss": 0.8929, "step": 19330 }, { "epoch": 0.98, "learning_rate": 1.6863538763806087e-08, "loss": 0.8806, "step": 19331 }, { "epoch": 0.98, "learning_rate": 1.676815663925968e-08, "loss": 0.8431, "step": 19332 }, { "epoch": 0.98, "learning_rate": 1.6673044800094018e-08, "loss": 0.7864, "step": 19333 }, { "epoch": 0.98, "learning_rate": 1.6578203248883707e-08, "loss": 0.7836, "step": 19334 }, { "epoch": 0.98, "learning_rate": 1.6483631988197803e-08, "loss": 0.8888, "step": 19335 }, { "epoch": 0.98, "learning_rate": 1.6389331020595366e-08, "loss": 0.8734, "step": 19336 }, { "epoch": 0.98, "learning_rate": 1.6295300348629918e-08, "loss": 0.8579, "step": 19337 }, { "epoch": 0.98, "learning_rate": 1.62015399748483e-08, "loss": 0.8403, "step": 19338 }, { "epoch": 0.98, "learning_rate": 1.610804990178738e-08, "loss": 0.8273, "step": 19339 }, { "epoch": 0.98, "learning_rate": 1.6014830131979575e-08, "loss": 0.8283, "step": 19340 }, { "epoch": 0.98, "learning_rate": 1.592188066794731e-08, "loss": 0.7767, "step": 19341 }, { "epoch": 0.98, "learning_rate": 1.582920151220857e-08, "loss": 0.8076, "step": 19342 }, { "epoch": 0.98, "learning_rate": 1.573679266727135e-08, "loss": 0.904, "step": 19343 }, { "epoch": 0.98, "learning_rate": 1.5644654135639202e-08, "loss": 0.8588, "step": 19344 }, { "epoch": 0.98, "learning_rate": 1.555278591980458e-08, "loss": 0.8412, "step": 19345 }, { "epoch": 0.98, "learning_rate": 1.5461188022255492e-08, "loss": 0.8489, "step": 19346 }, { "epoch": 0.98, "learning_rate": 1.5369860445472173e-08, "loss": 0.8221, "step": 19347 }, { "epoch": 0.98, "learning_rate": 1.5278803191927095e-08, "loss": 0.8525, "step": 19348 }, { "epoch": 0.98, "learning_rate": 1.5188016264086058e-08, "loss": 0.9098, "step": 19349 }, { "epoch": 0.98, "learning_rate": 1.509749966440599e-08, "loss": 0.8731, "step": 19350 }, { "epoch": 0.98, "learning_rate": 1.5007253395338263e-08, "loss": 0.795, "step": 19351 }, { "epoch": 0.98, "learning_rate": 1.4917277459325363e-08, "loss": 0.8856, "step": 19352 }, { "epoch": 0.98, "learning_rate": 1.4827571858804235e-08, "loss": 0.92, "step": 19353 }, { "epoch": 0.98, "learning_rate": 1.4738136596202935e-08, "loss": 0.828, "step": 19354 }, { "epoch": 0.98, "learning_rate": 1.4648971673943968e-08, "loss": 0.8936, "step": 19355 }, { "epoch": 0.98, "learning_rate": 1.456007709443985e-08, "loss": 0.7904, "step": 19356 }, { "epoch": 0.98, "learning_rate": 1.4471452860098657e-08, "loss": 0.7993, "step": 19357 }, { "epoch": 0.98, "learning_rate": 1.4383098973317356e-08, "loss": 0.852, "step": 19358 }, { "epoch": 0.98, "learning_rate": 1.429501543649181e-08, "loss": 0.76, "step": 19359 }, { "epoch": 0.98, "learning_rate": 1.420720225200345e-08, "loss": 0.8367, "step": 19360 }, { "epoch": 0.98, "learning_rate": 1.411965942223148e-08, "loss": 0.8717, "step": 19361 }, { "epoch": 0.98, "learning_rate": 1.4032386949546228e-08, "loss": 0.8429, "step": 19362 }, { "epoch": 0.98, "learning_rate": 1.394538483630914e-08, "loss": 0.8436, "step": 19363 }, { "epoch": 0.98, "learning_rate": 1.3858653084876106e-08, "loss": 0.8547, "step": 19364 }, { "epoch": 0.98, "learning_rate": 1.377219169759636e-08, "loss": 0.9308, "step": 19365 }, { "epoch": 0.98, "learning_rate": 1.368600067681025e-08, "loss": 0.7923, "step": 19366 }, { "epoch": 0.98, "learning_rate": 1.360008002485036e-08, "loss": 0.8895, "step": 19367 }, { "epoch": 0.98, "learning_rate": 1.351442974404371e-08, "loss": 0.895, "step": 19368 }, { "epoch": 0.98, "learning_rate": 1.3429049836709563e-08, "loss": 0.9029, "step": 19369 }, { "epoch": 0.98, "learning_rate": 1.3343940305158288e-08, "loss": 0.8923, "step": 19370 }, { "epoch": 0.98, "learning_rate": 1.325910115169471e-08, "loss": 0.8475, "step": 19371 }, { "epoch": 0.98, "learning_rate": 1.317453237861699e-08, "loss": 0.7473, "step": 19372 }, { "epoch": 0.98, "learning_rate": 1.3090233988212187e-08, "loss": 0.9021, "step": 19373 }, { "epoch": 0.98, "learning_rate": 1.3006205982765141e-08, "loss": 0.8301, "step": 19374 }, { "epoch": 0.98, "learning_rate": 1.2922448364549589e-08, "loss": 0.864, "step": 19375 }, { "epoch": 0.98, "learning_rate": 1.2838961135833716e-08, "loss": 0.8502, "step": 19376 }, { "epoch": 0.98, "learning_rate": 1.2755744298876826e-08, "loss": 0.8454, "step": 19377 }, { "epoch": 0.98, "learning_rate": 1.2672797855932673e-08, "loss": 0.9409, "step": 19378 }, { "epoch": 0.98, "learning_rate": 1.2590121809247235e-08, "loss": 0.9143, "step": 19379 }, { "epoch": 0.98, "learning_rate": 1.2507716161058725e-08, "loss": 0.8002, "step": 19380 }, { "epoch": 0.98, "learning_rate": 1.2425580913596469e-08, "loss": 0.9137, "step": 19381 }, { "epoch": 0.98, "learning_rate": 1.2343716069087574e-08, "loss": 0.8869, "step": 19382 }, { "epoch": 0.98, "learning_rate": 1.2262121629744717e-08, "loss": 0.9009, "step": 19383 }, { "epoch": 0.98, "learning_rate": 1.2180797597780569e-08, "loss": 0.9391, "step": 19384 }, { "epoch": 0.98, "learning_rate": 1.2099743975393374e-08, "loss": 0.7872, "step": 19385 }, { "epoch": 0.98, "learning_rate": 1.2018960764780263e-08, "loss": 0.9224, "step": 19386 }, { "epoch": 0.98, "learning_rate": 1.1938447968127265e-08, "loss": 0.8419, "step": 19387 }, { "epoch": 0.98, "learning_rate": 1.1858205587614857e-08, "loss": 0.7837, "step": 19388 }, { "epoch": 0.99, "learning_rate": 1.1778233625414636e-08, "loss": 0.8928, "step": 19389 }, { "epoch": 0.99, "learning_rate": 1.1698532083691538e-08, "loss": 0.913, "step": 19390 }, { "epoch": 0.99, "learning_rate": 1.1619100964604945e-08, "loss": 0.9077, "step": 19391 }, { "epoch": 0.99, "learning_rate": 1.1539940270303141e-08, "loss": 0.9916, "step": 19392 }, { "epoch": 0.99, "learning_rate": 1.1461050002932183e-08, "loss": 0.8752, "step": 19393 }, { "epoch": 0.99, "learning_rate": 1.1382430164624814e-08, "loss": 0.8573, "step": 19394 }, { "epoch": 0.99, "learning_rate": 1.130408075751155e-08, "loss": 0.817, "step": 19395 }, { "epoch": 0.99, "learning_rate": 1.1226001783714024e-08, "loss": 0.8469, "step": 19396 }, { "epoch": 0.99, "learning_rate": 1.1148193245343886e-08, "loss": 0.852, "step": 19397 }, { "epoch": 0.99, "learning_rate": 1.1070655144510555e-08, "loss": 0.7538, "step": 19398 }, { "epoch": 0.99, "learning_rate": 1.0993387483310136e-08, "loss": 0.8592, "step": 19399 }, { "epoch": 0.99, "learning_rate": 1.0916390263837618e-08, "loss": 0.8462, "step": 19400 }, { "epoch": 0.99, "learning_rate": 1.0839663488175777e-08, "loss": 0.8783, "step": 19401 }, { "epoch": 0.99, "learning_rate": 1.0763207158402955e-08, "loss": 0.8295, "step": 19402 }, { "epoch": 0.99, "learning_rate": 1.0687021276588605e-08, "loss": 0.8705, "step": 19403 }, { "epoch": 0.99, "learning_rate": 1.0611105844794411e-08, "loss": 0.9017, "step": 19404 }, { "epoch": 0.99, "learning_rate": 1.0535460865077618e-08, "loss": 0.911, "step": 19405 }, { "epoch": 0.99, "learning_rate": 1.0460086339484365e-08, "loss": 0.9293, "step": 19406 }, { "epoch": 0.99, "learning_rate": 1.0384982270057464e-08, "loss": 0.8504, "step": 19407 }, { "epoch": 0.99, "learning_rate": 1.0310148658827512e-08, "loss": 0.8543, "step": 19408 }, { "epoch": 0.99, "learning_rate": 1.0235585507822887e-08, "loss": 0.9117, "step": 19409 }, { "epoch": 0.99, "learning_rate": 1.0161292819061973e-08, "loss": 0.9341, "step": 19410 }, { "epoch": 0.99, "learning_rate": 1.0087270594555388e-08, "loss": 0.9241, "step": 19411 }, { "epoch": 0.99, "learning_rate": 1.001351883630708e-08, "loss": 0.91, "step": 19412 }, { "epoch": 0.99, "learning_rate": 9.940037546314341e-09, "loss": 0.7971, "step": 19413 }, { "epoch": 0.99, "learning_rate": 9.866826726566691e-09, "loss": 0.8672, "step": 19414 }, { "epoch": 0.99, "learning_rate": 9.79388637904699e-09, "loss": 0.7178, "step": 19415 }, { "epoch": 0.99, "learning_rate": 9.721216505728104e-09, "loss": 0.9097, "step": 19416 }, { "epoch": 0.99, "learning_rate": 9.648817108578457e-09, "loss": 0.8874, "step": 19417 }, { "epoch": 0.99, "learning_rate": 9.576688189557593e-09, "loss": 0.9403, "step": 19418 }, { "epoch": 0.99, "learning_rate": 9.504829750619505e-09, "loss": 0.7937, "step": 19419 }, { "epoch": 0.99, "learning_rate": 9.433241793709303e-09, "loss": 0.8983, "step": 19420 }, { "epoch": 0.99, "learning_rate": 9.361924320764327e-09, "loss": 0.7986, "step": 19421 }, { "epoch": 0.99, "learning_rate": 9.290877333717474e-09, "loss": 0.833, "step": 19422 }, { "epoch": 0.99, "learning_rate": 9.220100834489431e-09, "loss": 0.8629, "step": 19423 }, { "epoch": 0.99, "learning_rate": 9.149594824998664e-09, "loss": 0.7802, "step": 19424 }, { "epoch": 0.99, "learning_rate": 9.079359307152535e-09, "loss": 0.9078, "step": 19425 }, { "epoch": 0.99, "learning_rate": 9.009394282852857e-09, "loss": 0.8722, "step": 19426 }, { "epoch": 0.99, "learning_rate": 8.93969975399589e-09, "loss": 0.8327, "step": 19427 }, { "epoch": 0.99, "learning_rate": 8.870275722465682e-09, "loss": 0.8358, "step": 19428 }, { "epoch": 0.99, "learning_rate": 8.801122190142952e-09, "loss": 0.8419, "step": 19429 }, { "epoch": 0.99, "learning_rate": 8.732239158900646e-09, "loss": 0.8533, "step": 19430 }, { "epoch": 0.99, "learning_rate": 8.663626630602828e-09, "loss": 0.7669, "step": 19431 }, { "epoch": 0.99, "learning_rate": 8.595284607106902e-09, "loss": 0.8562, "step": 19432 }, { "epoch": 0.99, "learning_rate": 8.527213090264719e-09, "loss": 0.8908, "step": 19433 }, { "epoch": 0.99, "learning_rate": 8.459412081917028e-09, "loss": 0.7479, "step": 19434 }, { "epoch": 0.99, "learning_rate": 8.39188158390125e-09, "loss": 0.7747, "step": 19435 }, { "epoch": 0.99, "learning_rate": 8.324621598044813e-09, "loss": 0.9247, "step": 19436 }, { "epoch": 0.99, "learning_rate": 8.25763212616848e-09, "loss": 0.8423, "step": 19437 }, { "epoch": 0.99, "learning_rate": 8.190913170087467e-09, "loss": 0.8736, "step": 19438 }, { "epoch": 0.99, "learning_rate": 8.124464731605886e-09, "loss": 0.9062, "step": 19439 }, { "epoch": 0.99, "learning_rate": 8.05828681252452e-09, "loss": 0.8306, "step": 19440 }, { "epoch": 0.99, "learning_rate": 7.992379414634155e-09, "loss": 0.8159, "step": 19441 }, { "epoch": 0.99, "learning_rate": 7.926742539720035e-09, "loss": 0.824, "step": 19442 }, { "epoch": 0.99, "learning_rate": 7.861376189558512e-09, "loss": 0.7681, "step": 19443 }, { "epoch": 0.99, "learning_rate": 7.796280365919284e-09, "loss": 0.8343, "step": 19444 }, { "epoch": 0.99, "learning_rate": 7.731455070564276e-09, "loss": 0.8056, "step": 19445 }, { "epoch": 0.99, "learning_rate": 7.666900305249859e-09, "loss": 0.8424, "step": 19446 }, { "epoch": 0.99, "learning_rate": 7.602616071723523e-09, "loss": 0.8824, "step": 19447 }, { "epoch": 0.99, "learning_rate": 7.538602371724991e-09, "loss": 0.7922, "step": 19448 }, { "epoch": 0.99, "learning_rate": 7.474859206987318e-09, "loss": 0.9086, "step": 19449 }, { "epoch": 0.99, "learning_rate": 7.411386579235791e-09, "loss": 0.9175, "step": 19450 }, { "epoch": 0.99, "learning_rate": 7.348184490191257e-09, "loss": 0.8317, "step": 19451 }, { "epoch": 0.99, "learning_rate": 7.285252941562348e-09, "loss": 0.8702, "step": 19452 }, { "epoch": 0.99, "learning_rate": 7.2225919350543685e-09, "loss": 0.9285, "step": 19453 }, { "epoch": 0.99, "learning_rate": 7.160201472362627e-09, "loss": 0.8907, "step": 19454 }, { "epoch": 0.99, "learning_rate": 7.098081555176883e-09, "loss": 0.8384, "step": 19455 }, { "epoch": 0.99, "learning_rate": 7.036232185180236e-09, "loss": 0.85, "step": 19456 }, { "epoch": 0.99, "learning_rate": 6.974653364044681e-09, "loss": 0.83, "step": 19457 }, { "epoch": 0.99, "learning_rate": 6.913345093438883e-09, "loss": 0.8376, "step": 19458 }, { "epoch": 0.99, "learning_rate": 6.8523073750226266e-09, "loss": 0.8583, "step": 19459 }, { "epoch": 0.99, "learning_rate": 6.791540210447922e-09, "loss": 0.8295, "step": 19460 }, { "epoch": 0.99, "learning_rate": 6.7310436013612314e-09, "loss": 0.8511, "step": 19461 }, { "epoch": 0.99, "learning_rate": 6.670817549399022e-09, "loss": 0.8719, "step": 19462 }, { "epoch": 0.99, "learning_rate": 6.610862056192213e-09, "loss": 0.8863, "step": 19463 }, { "epoch": 0.99, "learning_rate": 6.55117712336395e-09, "loss": 0.8115, "step": 19464 }, { "epoch": 0.99, "learning_rate": 6.491762752530717e-09, "loss": 0.9036, "step": 19465 }, { "epoch": 0.99, "learning_rate": 6.4326189453001175e-09, "loss": 0.8713, "step": 19466 }, { "epoch": 0.99, "learning_rate": 6.373745703275314e-09, "loss": 0.8659, "step": 19467 }, { "epoch": 0.99, "learning_rate": 6.315143028048365e-09, "loss": 0.8068, "step": 19468 }, { "epoch": 0.99, "learning_rate": 6.25681092120578e-09, "loss": 0.8339, "step": 19469 }, { "epoch": 0.99, "learning_rate": 6.1987493843285175e-09, "loss": 0.9671, "step": 19470 }, { "epoch": 0.99, "learning_rate": 6.1409584189864315e-09, "loss": 0.8137, "step": 19471 }, { "epoch": 0.99, "learning_rate": 6.083438026746047e-09, "loss": 0.9209, "step": 19472 }, { "epoch": 0.99, "learning_rate": 6.0261882091638965e-09, "loss": 0.7909, "step": 19473 }, { "epoch": 0.99, "learning_rate": 5.9692089677898525e-09, "loss": 0.8132, "step": 19474 }, { "epoch": 0.99, "learning_rate": 5.9125003041671234e-09, "loss": 0.9123, "step": 19475 }, { "epoch": 0.99, "learning_rate": 5.8560622198300385e-09, "loss": 0.9254, "step": 19476 }, { "epoch": 0.99, "learning_rate": 5.799894716307375e-09, "loss": 0.8503, "step": 19477 }, { "epoch": 0.99, "learning_rate": 5.743997795119027e-09, "loss": 0.8727, "step": 19478 }, { "epoch": 0.99, "learning_rate": 5.68837145777934e-09, "loss": 0.8979, "step": 19479 }, { "epoch": 0.99, "learning_rate": 5.6330157057937764e-09, "loss": 0.8557, "step": 19480 }, { "epoch": 0.99, "learning_rate": 5.5779305406611364e-09, "loss": 0.7917, "step": 19481 }, { "epoch": 0.99, "learning_rate": 5.52311596387245e-09, "loss": 0.8963, "step": 19482 }, { "epoch": 0.99, "learning_rate": 5.468571976913195e-09, "loss": 0.9437, "step": 19483 }, { "epoch": 0.99, "learning_rate": 5.414298581257749e-09, "loss": 0.8989, "step": 19484 }, { "epoch": 0.99, "learning_rate": 5.360295778378266e-09, "loss": 0.8426, "step": 19485 }, { "epoch": 0.99, "learning_rate": 5.30656356973358e-09, "loss": 0.9257, "step": 19486 }, { "epoch": 0.99, "learning_rate": 5.253101956781415e-09, "loss": 0.9016, "step": 19487 }, { "epoch": 0.99, "learning_rate": 5.199910940967279e-09, "loss": 0.8716, "step": 19488 }, { "epoch": 0.99, "learning_rate": 5.146990523732243e-09, "loss": 0.8833, "step": 19489 }, { "epoch": 0.99, "learning_rate": 5.094340706508494e-09, "loss": 0.8142, "step": 19490 }, { "epoch": 0.99, "learning_rate": 5.041961490721559e-09, "loss": 0.9395, "step": 19491 }, { "epoch": 0.99, "learning_rate": 4.989852877790302e-09, "loss": 0.8683, "step": 19492 }, { "epoch": 0.99, "learning_rate": 4.938014869124708e-09, "loss": 0.7805, "step": 19493 }, { "epoch": 0.99, "learning_rate": 4.886447466129207e-09, "loss": 0.8596, "step": 19494 }, { "epoch": 0.99, "learning_rate": 4.835150670198241e-09, "loss": 0.93, "step": 19495 }, { "epoch": 0.99, "learning_rate": 4.784124482721808e-09, "loss": 0.8701, "step": 19496 }, { "epoch": 0.99, "learning_rate": 4.733368905081026e-09, "loss": 0.8967, "step": 19497 }, { "epoch": 0.99, "learning_rate": 4.6828839386492406e-09, "loss": 0.7736, "step": 19498 }, { "epoch": 0.99, "learning_rate": 4.632669584795357e-09, "loss": 0.8486, "step": 19499 }, { "epoch": 0.99, "learning_rate": 4.5827258448771785e-09, "loss": 0.8376, "step": 19500 }, { "epoch": 0.99, "learning_rate": 4.5330527202480656e-09, "loss": 0.8366, "step": 19501 }, { "epoch": 0.99, "learning_rate": 4.4836502122524995e-09, "loss": 0.7765, "step": 19502 }, { "epoch": 0.99, "learning_rate": 4.4345183222260775e-09, "loss": 0.8397, "step": 19503 }, { "epoch": 0.99, "learning_rate": 4.385657051502179e-09, "loss": 0.8694, "step": 19504 }, { "epoch": 0.99, "learning_rate": 4.3370664014019685e-09, "loss": 0.8618, "step": 19505 }, { "epoch": 0.99, "learning_rate": 4.288746373239949e-09, "loss": 0.823, "step": 19506 }, { "epoch": 0.99, "learning_rate": 4.2406969683261855e-09, "loss": 0.9354, "step": 19507 }, { "epoch": 0.99, "learning_rate": 4.192918187961859e-09, "loss": 0.9544, "step": 19508 }, { "epoch": 0.99, "learning_rate": 4.145410033438158e-09, "loss": 0.8436, "step": 19509 }, { "epoch": 0.99, "learning_rate": 4.098172506042941e-09, "loss": 0.9216, "step": 19510 }, { "epoch": 0.99, "learning_rate": 4.051205607055187e-09, "loss": 0.8449, "step": 19511 }, { "epoch": 0.99, "learning_rate": 4.0045093377460985e-09, "loss": 0.8255, "step": 19512 }, { "epoch": 0.99, "learning_rate": 3.958083699381332e-09, "loss": 0.8767, "step": 19513 }, { "epoch": 0.99, "learning_rate": 3.911928693214328e-09, "loss": 0.7903, "step": 19514 }, { "epoch": 0.99, "learning_rate": 3.86604432049853e-09, "loss": 0.8015, "step": 19515 }, { "epoch": 0.99, "learning_rate": 3.820430582474055e-09, "loss": 0.8136, "step": 19516 }, { "epoch": 0.99, "learning_rate": 3.775087480375472e-09, "loss": 0.7319, "step": 19517 }, { "epoch": 0.99, "learning_rate": 3.730015015431798e-09, "loss": 0.8474, "step": 19518 }, { "epoch": 0.99, "learning_rate": 3.6852131888631683e-09, "loss": 0.8228, "step": 19519 }, { "epoch": 0.99, "learning_rate": 3.6406820018819456e-09, "loss": 0.8141, "step": 19520 }, { "epoch": 0.99, "learning_rate": 3.5964214556938327e-09, "loss": 0.8599, "step": 19521 }, { "epoch": 0.99, "learning_rate": 3.5524315514978702e-09, "loss": 0.7897, "step": 19522 }, { "epoch": 0.99, "learning_rate": 3.5087122904831074e-09, "loss": 0.8975, "step": 19523 }, { "epoch": 0.99, "learning_rate": 3.465263673835262e-09, "loss": 0.7742, "step": 19524 }, { "epoch": 0.99, "learning_rate": 3.4220857027300604e-09, "loss": 0.8078, "step": 19525 }, { "epoch": 0.99, "learning_rate": 3.3791783783365672e-09, "loss": 0.8322, "step": 19526 }, { "epoch": 0.99, "learning_rate": 3.336541701816076e-09, "loss": 0.8367, "step": 19527 }, { "epoch": 0.99, "learning_rate": 3.2941756743232188e-09, "loss": 0.9341, "step": 19528 }, { "epoch": 0.99, "learning_rate": 3.2520802970037457e-09, "loss": 0.8274, "step": 19529 }, { "epoch": 0.99, "learning_rate": 3.210255571000076e-09, "loss": 0.9468, "step": 19530 }, { "epoch": 0.99, "learning_rate": 3.1687014974424168e-09, "loss": 0.8028, "step": 19531 }, { "epoch": 0.99, "learning_rate": 3.1274180774565343e-09, "loss": 0.8097, "step": 19532 }, { "epoch": 0.99, "learning_rate": 3.086405312159313e-09, "loss": 0.8444, "step": 19533 }, { "epoch": 0.99, "learning_rate": 3.0456632026620858e-09, "loss": 0.8984, "step": 19534 }, { "epoch": 0.99, "learning_rate": 3.0051917500684146e-09, "loss": 0.8601, "step": 19535 }, { "epoch": 0.99, "learning_rate": 2.9649909554718692e-09, "loss": 0.8286, "step": 19536 }, { "epoch": 0.99, "learning_rate": 2.9250608199626885e-09, "loss": 0.8818, "step": 19537 }, { "epoch": 0.99, "learning_rate": 2.8854013446222296e-09, "loss": 0.8901, "step": 19538 }, { "epoch": 0.99, "learning_rate": 2.846012530522968e-09, "loss": 0.8479, "step": 19539 }, { "epoch": 0.99, "learning_rate": 2.806894378730718e-09, "loss": 0.8461, "step": 19540 }, { "epoch": 0.99, "learning_rate": 2.7680468903068523e-09, "loss": 0.96, "step": 19541 }, { "epoch": 0.99, "learning_rate": 2.7294700663016426e-09, "loss": 0.9244, "step": 19542 }, { "epoch": 0.99, "learning_rate": 2.6911639077598083e-09, "loss": 0.9399, "step": 19543 }, { "epoch": 0.99, "learning_rate": 2.653128415718298e-09, "loss": 0.8713, "step": 19544 }, { "epoch": 0.99, "learning_rate": 2.6153635912073983e-09, "loss": 0.9054, "step": 19545 }, { "epoch": 0.99, "learning_rate": 2.5778694352485144e-09, "loss": 0.8679, "step": 19546 }, { "epoch": 0.99, "learning_rate": 2.540645948857501e-09, "loss": 0.8502, "step": 19547 }, { "epoch": 0.99, "learning_rate": 2.5036931330424395e-09, "loss": 0.9339, "step": 19548 }, { "epoch": 0.99, "learning_rate": 2.4670109888025318e-09, "loss": 0.9393, "step": 19549 }, { "epoch": 0.99, "learning_rate": 2.430599517132537e-09, "loss": 0.8477, "step": 19550 }, { "epoch": 0.99, "learning_rate": 2.394458719017223e-09, "loss": 0.9016, "step": 19551 }, { "epoch": 0.99, "learning_rate": 2.3585885954358066e-09, "loss": 0.7741, "step": 19552 }, { "epoch": 0.99, "learning_rate": 2.322989147358623e-09, "loss": 0.8417, "step": 19553 }, { "epoch": 0.99, "learning_rate": 2.2876603757493455e-09, "loss": 0.9738, "step": 19554 }, { "epoch": 0.99, "learning_rate": 2.252602281564986e-09, "loss": 0.8328, "step": 19555 }, { "epoch": 0.99, "learning_rate": 2.217814865754786e-09, "loss": 0.8198, "step": 19556 }, { "epoch": 0.99, "learning_rate": 2.1832981292613242e-09, "loss": 0.8728, "step": 19557 }, { "epoch": 0.99, "learning_rate": 2.149052073017188e-09, "loss": 0.9125, "step": 19558 }, { "epoch": 0.99, "learning_rate": 2.1150766979516347e-09, "loss": 0.7933, "step": 19559 }, { "epoch": 0.99, "learning_rate": 2.0813720049828178e-09, "loss": 0.7632, "step": 19560 }, { "epoch": 0.99, "learning_rate": 2.0479379950233415e-09, "loss": 0.9203, "step": 19561 }, { "epoch": 0.99, "learning_rate": 2.014774668979147e-09, "loss": 0.8008, "step": 19562 }, { "epoch": 0.99, "learning_rate": 1.981882027748405e-09, "loss": 0.8026, "step": 19563 }, { "epoch": 0.99, "learning_rate": 1.9492600722215148e-09, "loss": 0.8679, "step": 19564 }, { "epoch": 0.99, "learning_rate": 1.916908803279993e-09, "loss": 0.873, "step": 19565 }, { "epoch": 0.99, "learning_rate": 1.8848282218020265e-09, "loss": 0.8617, "step": 19566 }, { "epoch": 0.99, "learning_rate": 1.853018328655809e-09, "loss": 0.8777, "step": 19567 }, { "epoch": 0.99, "learning_rate": 1.8214791247006536e-09, "loss": 0.8346, "step": 19568 }, { "epoch": 0.99, "learning_rate": 1.790210610792542e-09, "loss": 0.8015, "step": 19569 }, { "epoch": 0.99, "learning_rate": 1.7592127877763542e-09, "loss": 0.9028, "step": 19570 }, { "epoch": 0.99, "learning_rate": 1.728485656492529e-09, "loss": 0.8968, "step": 19571 }, { "epoch": 0.99, "learning_rate": 1.6980292177737334e-09, "loss": 0.8707, "step": 19572 }, { "epoch": 0.99, "learning_rate": 1.6678434724426429e-09, "loss": 0.7627, "step": 19573 }, { "epoch": 0.99, "learning_rate": 1.6379284213163816e-09, "loss": 0.8481, "step": 19574 }, { "epoch": 0.99, "learning_rate": 1.6082840652065224e-09, "loss": 0.7832, "step": 19575 }, { "epoch": 0.99, "learning_rate": 1.5789104049157567e-09, "loss": 0.8316, "step": 19576 }, { "epoch": 0.99, "learning_rate": 1.5498074412367835e-09, "loss": 0.9411, "step": 19577 }, { "epoch": 0.99, "learning_rate": 1.5209751749611924e-09, "loss": 0.8926, "step": 19578 }, { "epoch": 0.99, "learning_rate": 1.4924136068661387e-09, "loss": 0.8215, "step": 19579 }, { "epoch": 0.99, "learning_rate": 1.4641227377265587e-09, "loss": 0.8222, "step": 19580 }, { "epoch": 0.99, "learning_rate": 1.4361025683096164e-09, "loss": 0.8127, "step": 19581 }, { "epoch": 0.99, "learning_rate": 1.4083530993713735e-09, "loss": 0.8654, "step": 19582 }, { "epoch": 0.99, "learning_rate": 1.3808743316645613e-09, "loss": 0.8961, "step": 19583 }, { "epoch": 0.99, "learning_rate": 1.353666265933029e-09, "loss": 0.9277, "step": 19584 }, { "epoch": 0.99, "learning_rate": 1.326728902912855e-09, "loss": 0.7535, "step": 19585 }, { "epoch": 1.0, "learning_rate": 1.3000622433334554e-09, "loss": 0.8513, "step": 19586 }, { "epoch": 1.0, "learning_rate": 1.2736662879175854e-09, "loss": 0.8156, "step": 19587 }, { "epoch": 1.0, "learning_rate": 1.2475410373791187e-09, "loss": 0.9025, "step": 19588 }, { "epoch": 1.0, "learning_rate": 1.2216864924263772e-09, "loss": 0.7966, "step": 19589 }, { "epoch": 1.0, "learning_rate": 1.1961026537588016e-09, "loss": 0.8569, "step": 19590 }, { "epoch": 1.0, "learning_rate": 1.170789522068061e-09, "loss": 0.8185, "step": 19591 }, { "epoch": 1.0, "learning_rate": 1.1457470980402729e-09, "loss": 0.8433, "step": 19592 }, { "epoch": 1.0, "learning_rate": 1.1209753823537839e-09, "loss": 0.8336, "step": 19593 }, { "epoch": 1.0, "learning_rate": 1.0964743756780583e-09, "loss": 0.8942, "step": 19594 }, { "epoch": 1.0, "learning_rate": 1.0722440786781197e-09, "loss": 0.7836, "step": 19595 }, { "epoch": 1.0, "learning_rate": 1.0482844920089996e-09, "loss": 0.8317, "step": 19596 }, { "epoch": 1.0, "learning_rate": 1.0245956163201788e-09, "loss": 0.9286, "step": 19597 }, { "epoch": 1.0, "learning_rate": 1.0011774522511452e-09, "loss": 0.7998, "step": 19598 }, { "epoch": 1.0, "learning_rate": 9.780300004380573e-10, "loss": 0.8973, "step": 19599 }, { "epoch": 1.0, "learning_rate": 9.5515326150597e-10, "loss": 0.9242, "step": 19600 }, { "epoch": 1.0, "learning_rate": 9.325472360754984e-10, "loss": 0.8152, "step": 19601 }, { "epoch": 1.0, "learning_rate": 9.102119247583752e-10, "loss": 0.773, "step": 19602 }, { "epoch": 1.0, "learning_rate": 8.881473281585617e-10, "loss": 0.7936, "step": 19603 }, { "epoch": 1.0, "learning_rate": 8.663534468733581e-10, "loss": 0.8126, "step": 19604 }, { "epoch": 1.0, "learning_rate": 8.448302814945131e-10, "loss": 0.9277, "step": 19605 }, { "epoch": 1.0, "learning_rate": 8.235778326026734e-10, "loss": 0.7857, "step": 19606 }, { "epoch": 1.0, "learning_rate": 8.025961007751548e-10, "loss": 0.8847, "step": 19607 }, { "epoch": 1.0, "learning_rate": 7.818850865781713e-10, "loss": 0.8652, "step": 19608 }, { "epoch": 1.0, "learning_rate": 7.614447905734956e-10, "loss": 0.8108, "step": 19609 }, { "epoch": 1.0, "learning_rate": 7.412752133151291e-10, "loss": 0.7903, "step": 19610 }, { "epoch": 1.0, "learning_rate": 7.213763553481912e-10, "loss": 0.7597, "step": 19611 }, { "epoch": 1.0, "learning_rate": 7.017482172111401e-10, "loss": 0.8896, "step": 19612 }, { "epoch": 1.0, "learning_rate": 6.823907994357726e-10, "loss": 0.8302, "step": 19613 }, { "epoch": 1.0, "learning_rate": 6.63304102546114e-10, "loss": 0.8288, "step": 19614 }, { "epoch": 1.0, "learning_rate": 6.444881270595283e-10, "loss": 0.8389, "step": 19615 }, { "epoch": 1.0, "learning_rate": 6.259428734844975e-10, "loss": 0.8371, "step": 19616 }, { "epoch": 1.0, "learning_rate": 6.076683423239526e-10, "loss": 0.889, "step": 19617 }, { "epoch": 1.0, "learning_rate": 5.896645340719431e-10, "loss": 0.7789, "step": 19618 }, { "epoch": 1.0, "learning_rate": 5.71931449216967e-10, "loss": 0.8698, "step": 19619 }, { "epoch": 1.0, "learning_rate": 5.544690882375303e-10, "loss": 0.9373, "step": 19620 }, { "epoch": 1.0, "learning_rate": 5.372774516076984e-10, "loss": 0.8046, "step": 19621 }, { "epoch": 1.0, "learning_rate": 5.203565397915444e-10, "loss": 0.9084, "step": 19622 }, { "epoch": 1.0, "learning_rate": 5.037063532498109e-10, "loss": 0.8849, "step": 19623 }, { "epoch": 1.0, "learning_rate": 4.873268924299179e-10, "loss": 0.7821, "step": 19624 }, { "epoch": 1.0, "learning_rate": 4.712181577781749e-10, "loss": 0.8307, "step": 19625 }, { "epoch": 1.0, "learning_rate": 4.553801497297894e-10, "loss": 0.9447, "step": 19626 }, { "epoch": 1.0, "learning_rate": 4.398128687121972e-10, "loss": 0.9111, "step": 19627 }, { "epoch": 1.0, "learning_rate": 4.2451631514839333e-10, "loss": 0.8454, "step": 19628 }, { "epoch": 1.0, "learning_rate": 4.094904894524909e-10, "loss": 0.9701, "step": 19629 }, { "epoch": 1.0, "learning_rate": 3.9473539203083166e-10, "loss": 0.8961, "step": 19630 }, { "epoch": 1.0, "learning_rate": 3.802510232819856e-10, "loss": 0.8692, "step": 19631 }, { "epoch": 1.0, "learning_rate": 3.660373836000819e-10, "loss": 0.8554, "step": 19632 }, { "epoch": 1.0, "learning_rate": 3.520944733681475e-10, "loss": 0.7814, "step": 19633 }, { "epoch": 1.0, "learning_rate": 3.384222929658787e-10, "loss": 0.7443, "step": 19634 }, { "epoch": 1.0, "learning_rate": 3.250208427607593e-10, "loss": 0.923, "step": 19635 }, { "epoch": 1.0, "learning_rate": 3.1189012311694245e-10, "loss": 0.885, "step": 19636 }, { "epoch": 1.0, "learning_rate": 2.9903013439080974e-10, "loss": 0.8447, "step": 19637 }, { "epoch": 1.0, "learning_rate": 2.864408769287508e-10, "loss": 0.8142, "step": 19638 }, { "epoch": 1.0, "learning_rate": 2.7412235107271424e-10, "loss": 0.8874, "step": 19639 }, { "epoch": 1.0, "learning_rate": 2.62074557155767e-10, "loss": 0.8477, "step": 19640 }, { "epoch": 1.0, "learning_rate": 2.502974955043147e-10, "loss": 0.8595, "step": 19641 }, { "epoch": 1.0, "learning_rate": 2.387911664381015e-10, "loss": 0.7501, "step": 19642 }, { "epoch": 1.0, "learning_rate": 2.275555702668797e-10, "loss": 0.856, "step": 19643 }, { "epoch": 1.0, "learning_rate": 2.165907072948503e-10, "loss": 0.8611, "step": 19644 }, { "epoch": 1.0, "learning_rate": 2.058965778206634e-10, "loss": 0.8224, "step": 19645 }, { "epoch": 1.0, "learning_rate": 1.9547318213297695e-10, "loss": 0.8715, "step": 19646 }, { "epoch": 1.0, "learning_rate": 1.8532052051378758e-10, "loss": 0.8645, "step": 19647 }, { "epoch": 1.0, "learning_rate": 1.7543859323732037e-10, "loss": 0.8186, "step": 19648 }, { "epoch": 1.0, "learning_rate": 1.6582740057335955e-10, "loss": 0.9552, "step": 19649 }, { "epoch": 1.0, "learning_rate": 1.5648694277947685e-10, "loss": 0.7833, "step": 19650 }, { "epoch": 1.0, "learning_rate": 1.4741722010991332e-10, "loss": 0.9253, "step": 19651 }, { "epoch": 1.0, "learning_rate": 1.386182328100283e-10, "loss": 0.8889, "step": 19652 }, { "epoch": 1.0, "learning_rate": 1.3008998111851967e-10, "loss": 0.8759, "step": 19653 }, { "epoch": 1.0, "learning_rate": 1.2183246526520364e-10, "loss": 0.9138, "step": 19654 }, { "epoch": 1.0, "learning_rate": 1.1384568547434527e-10, "loss": 0.8824, "step": 19655 }, { "epoch": 1.0, "learning_rate": 1.0612964196243803e-10, "loss": 0.9122, "step": 19656 }, { "epoch": 1.0, "learning_rate": 9.868433493709362e-11, "loss": 0.8592, "step": 19657 }, { "epoch": 1.0, "learning_rate": 9.150976460148287e-11, "loss": 0.8344, "step": 19658 }, { "epoch": 1.0, "learning_rate": 8.46059311498948e-11, "loss": 0.8378, "step": 19659 }, { "epoch": 1.0, "learning_rate": 7.797283476773665e-11, "loss": 0.9199, "step": 19660 }, { "epoch": 1.0, "learning_rate": 7.161047563486457e-11, "loss": 0.8677, "step": 19661 }, { "epoch": 1.0, "learning_rate": 6.551885392447333e-11, "loss": 0.913, "step": 19662 }, { "epoch": 1.0, "learning_rate": 5.969796980198617e-11, "loss": 0.8747, "step": 19663 }, { "epoch": 1.0, "learning_rate": 5.4147823422834313e-11, "loss": 0.8894, "step": 19664 }, { "epoch": 1.0, "learning_rate": 4.8868414939118316e-11, "loss": 0.8212, "step": 19665 }, { "epoch": 1.0, "learning_rate": 4.3859744492946723e-11, "loss": 0.7402, "step": 19666 }, { "epoch": 1.0, "learning_rate": 3.912181222087696e-11, "loss": 0.7861, "step": 19667 }, { "epoch": 1.0, "learning_rate": 3.465461824947447e-11, "loss": 0.8705, "step": 19668 }, { "epoch": 1.0, "learning_rate": 3.045816270086377e-11, "loss": 0.9416, "step": 19669 }, { "epoch": 1.0, "learning_rate": 2.6532445689397833e-11, "loss": 0.858, "step": 19670 }, { "epoch": 1.0, "learning_rate": 2.287746731943763e-11, "loss": 0.8393, "step": 19671 }, { "epoch": 1.0, "learning_rate": 1.9493227690903227e-11, "loss": 0.8311, "step": 19672 }, { "epoch": 1.0, "learning_rate": 1.6379726895943138e-11, "loss": 0.7805, "step": 19673 }, { "epoch": 1.0, "learning_rate": 1.3536965017824088e-11, "loss": 0.83, "step": 19674 }, { "epoch": 1.0, "learning_rate": 1.0964942135371915e-11, "loss": 0.7232, "step": 19675 }, { "epoch": 1.0, "learning_rate": 8.663658315199997e-12, "loss": 0.8425, "step": 19676 }, { "epoch": 1.0, "learning_rate": 6.633113621701271e-12, "loss": 0.7876, "step": 19677 }, { "epoch": 1.0, "learning_rate": 4.873308110386888e-12, "loss": 0.8324, "step": 19678 }, { "epoch": 1.0, "learning_rate": 3.3842418267759915e-12, "loss": 0.9495, "step": 19679 }, { "epoch": 1.0, "learning_rate": 2.1659148130570572e-12, "loss": 0.9106, "step": 19680 }, { "epoch": 1.0, "learning_rate": 1.218327101426553e-12, "loss": 0.8223, "step": 19681 }, { "epoch": 1.0, "learning_rate": 5.414787185298309e-13, "loss": 0.8712, "step": 19682 }, { "epoch": 1.0, "learning_rate": 1.3536967991001348e-13, "loss": 0.8605, "step": 19683 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 0.7451, "step": 19684 }, { "epoch": 1.0, "step": 19684, "total_flos": 3.757636157437064e+19, "train_loss": 0.9381690918248593, "train_runtime": 185106.1952, "train_samples_per_second": 13.611, "train_steps_per_second": 0.106 } ], "max_steps": 19684, "num_train_epochs": 1, "total_flos": 3.757636157437064e+19, "trial_name": null, "trial_params": null }