diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,114978 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.9999608533959679, + "eval_steps": 500, + "global_step": 19158, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 6.956521739130435e-10, + "loss": 0.2284, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 1.391304347826087e-09, + "loss": 0.1883, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 2.0869565217391305e-09, + "loss": 0.2199, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 2.782608695652174e-09, + "loss": 0.2431, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 3.4782608695652172e-09, + "loss": 0.206, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 4.173913043478261e-09, + "loss": 0.1651, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 4.869565217391304e-09, + "loss": 0.2045, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 5.565217391304348e-09, + "loss": 0.1929, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 6.2608695652173915e-09, + "loss": 0.2078, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 6.9565217391304344e-09, + "loss": 0.1997, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 7.652173913043477e-09, + "loss": 0.1959, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 8.347826086956522e-09, + "loss": 0.2109, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 9.043478260869565e-09, + "loss": 0.2497, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 9.739130434782608e-09, + "loss": 0.2017, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 1.043478260869565e-08, + "loss": 0.2039, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 1.1130434782608695e-08, + "loss": 0.1839, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 1.1826086956521738e-08, + "loss": 0.202, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 1.2521739130434783e-08, + "loss": 0.1803, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 1.3217391304347824e-08, + "loss": 0.2131, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 1.3913043478260869e-08, + "loss": 0.2312, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 1.4608695652173913e-08, + "loss": 0.2067, + "step": 21 + }, + { + "epoch": 0.0, + "learning_rate": 1.5304347826086955e-08, + "loss": 0.2229, + "step": 22 + }, + { + "epoch": 0.0, + "learning_rate": 1.6e-08, + "loss": 0.2166, + "step": 23 + }, + { + "epoch": 0.0, + "learning_rate": 1.6695652173913044e-08, + "loss": 0.1903, + "step": 24 + }, + { + "epoch": 0.0, + "learning_rate": 1.7391304347826087e-08, + "loss": 0.1634, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 1.808695652173913e-08, + "loss": 0.1794, + "step": 26 + }, + { + "epoch": 0.0, + "learning_rate": 1.8782608695652173e-08, + "loss": 0.1999, + "step": 27 + }, + { + "epoch": 0.0, + "learning_rate": 1.9478260869565216e-08, + "loss": 0.2145, + "step": 28 + }, + { + "epoch": 0.0, + "learning_rate": 2.0173913043478262e-08, + "loss": 0.1853, + "step": 29 + }, + { + "epoch": 0.0, + "learning_rate": 2.08695652173913e-08, + "loss": 0.1729, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 2.1565217391304348e-08, + "loss": 0.1867, + "step": 31 + }, + { + "epoch": 0.0, + "learning_rate": 2.226086956521739e-08, + "loss": 0.1901, + "step": 32 + }, + { + "epoch": 0.0, + "learning_rate": 2.2956521739130434e-08, + "loss": 0.1979, + "step": 33 + }, + { + "epoch": 0.0, + "learning_rate": 2.3652173913043477e-08, + "loss": 0.1817, + "step": 34 + }, + { + "epoch": 0.0, + "learning_rate": 2.4347826086956523e-08, + "loss": 0.2083, + "step": 35 + }, + { + "epoch": 0.0, + "learning_rate": 2.5043478260869566e-08, + "loss": 0.1781, + "step": 36 + }, + { + "epoch": 0.0, + "learning_rate": 2.573913043478261e-08, + "loss": 0.2254, + "step": 37 + }, + { + "epoch": 0.0, + "learning_rate": 2.643478260869565e-08, + "loss": 0.1958, + "step": 38 + }, + { + "epoch": 0.0, + "learning_rate": 2.7130434782608698e-08, + "loss": 0.2498, + "step": 39 + }, + { + "epoch": 0.0, + "learning_rate": 2.7826086956521738e-08, + "loss": 0.1859, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 2.852173913043478e-08, + "loss": 0.1958, + "step": 41 + }, + { + "epoch": 0.0, + "learning_rate": 2.9217391304347827e-08, + "loss": 0.1669, + "step": 42 + }, + { + "epoch": 0.0, + "learning_rate": 2.991304347826087e-08, + "loss": 0.2056, + "step": 43 + }, + { + "epoch": 0.0, + "learning_rate": 3.060869565217391e-08, + "loss": 0.2458, + "step": 44 + }, + { + "epoch": 0.0, + "learning_rate": 3.1304347826086956e-08, + "loss": 0.1747, + "step": 45 + }, + { + "epoch": 0.0, + "learning_rate": 3.2e-08, + "loss": 0.2189, + "step": 46 + }, + { + "epoch": 0.0, + "learning_rate": 3.269565217391304e-08, + "loss": 0.1708, + "step": 47 + }, + { + "epoch": 0.0, + "learning_rate": 3.339130434782609e-08, + "loss": 0.1888, + "step": 48 + }, + { + "epoch": 0.0, + "learning_rate": 3.408695652173913e-08, + "loss": 0.1986, + "step": 49 + }, + { + "epoch": 0.0, + "learning_rate": 3.4782608695652174e-08, + "loss": 0.2243, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 3.5478260869565213e-08, + "loss": 0.164, + "step": 51 + }, + { + "epoch": 0.0, + "learning_rate": 3.617391304347826e-08, + "loss": 0.1986, + "step": 52 + }, + { + "epoch": 0.0, + "learning_rate": 3.68695652173913e-08, + "loss": 0.2044, + "step": 53 + }, + { + "epoch": 0.0, + "learning_rate": 3.7565217391304346e-08, + "loss": 0.1732, + "step": 54 + }, + { + "epoch": 0.0, + "learning_rate": 3.826086956521739e-08, + "loss": 0.1822, + "step": 55 + }, + { + "epoch": 0.0, + "learning_rate": 3.895652173913043e-08, + "loss": 0.1796, + "step": 56 + }, + { + "epoch": 0.0, + "learning_rate": 3.965217391304347e-08, + "loss": 0.2011, + "step": 57 + }, + { + "epoch": 0.0, + "learning_rate": 4.0347826086956524e-08, + "loss": 0.2046, + "step": 58 + }, + { + "epoch": 0.0, + "learning_rate": 4.1043478260869564e-08, + "loss": 0.1951, + "step": 59 + }, + { + "epoch": 0.0, + "learning_rate": 4.17391304347826e-08, + "loss": 0.2242, + "step": 60 + }, + { + "epoch": 0.0, + "learning_rate": 4.243478260869565e-08, + "loss": 0.1967, + "step": 61 + }, + { + "epoch": 0.0, + "learning_rate": 4.3130434782608696e-08, + "loss": 0.202, + "step": 62 + }, + { + "epoch": 0.0, + "learning_rate": 4.3826086956521735e-08, + "loss": 0.1669, + "step": 63 + }, + { + "epoch": 0.0, + "learning_rate": 4.452173913043478e-08, + "loss": 0.192, + "step": 64 + }, + { + "epoch": 0.0, + "learning_rate": 4.521739130434782e-08, + "loss": 0.1674, + "step": 65 + }, + { + "epoch": 0.0, + "learning_rate": 4.591304347826087e-08, + "loss": 0.2355, + "step": 66 + }, + { + "epoch": 0.0, + "learning_rate": 4.6608695652173914e-08, + "loss": 0.2632, + "step": 67 + }, + { + "epoch": 0.0, + "learning_rate": 4.7304347826086953e-08, + "loss": 0.2088, + "step": 68 + }, + { + "epoch": 0.0, + "learning_rate": 4.799999999999999e-08, + "loss": 0.2004, + "step": 69 + }, + { + "epoch": 0.0, + "learning_rate": 4.8695652173913046e-08, + "loss": 0.2404, + "step": 70 + }, + { + "epoch": 0.0, + "learning_rate": 4.9391304347826086e-08, + "loss": 0.1771, + "step": 71 + }, + { + "epoch": 0.0, + "learning_rate": 5.008695652173913e-08, + "loss": 0.1829, + "step": 72 + }, + { + "epoch": 0.0, + "learning_rate": 5.078260869565217e-08, + "loss": 0.17, + "step": 73 + }, + { + "epoch": 0.0, + "learning_rate": 5.147826086956522e-08, + "loss": 0.1925, + "step": 74 + }, + { + "epoch": 0.0, + "learning_rate": 5.217391304347826e-08, + "loss": 0.2334, + "step": 75 + }, + { + "epoch": 0.0, + "learning_rate": 5.28695652173913e-08, + "loss": 0.1993, + "step": 76 + }, + { + "epoch": 0.0, + "learning_rate": 5.356521739130434e-08, + "loss": 0.1785, + "step": 77 + }, + { + "epoch": 0.0, + "learning_rate": 5.4260869565217396e-08, + "loss": 0.2214, + "step": 78 + }, + { + "epoch": 0.0, + "learning_rate": 5.4956521739130436e-08, + "loss": 0.1722, + "step": 79 + }, + { + "epoch": 0.0, + "learning_rate": 5.5652173913043475e-08, + "loss": 0.2024, + "step": 80 + }, + { + "epoch": 0.0, + "learning_rate": 5.6347826086956515e-08, + "loss": 0.2265, + "step": 81 + }, + { + "epoch": 0.0, + "learning_rate": 5.704347826086956e-08, + "loss": 0.1864, + "step": 82 + }, + { + "epoch": 0.0, + "learning_rate": 5.77391304347826e-08, + "loss": 0.1782, + "step": 83 + }, + { + "epoch": 0.0, + "learning_rate": 5.8434782608695654e-08, + "loss": 0.1822, + "step": 84 + }, + { + "epoch": 0.0, + "learning_rate": 5.9130434782608693e-08, + "loss": 0.1757, + "step": 85 + }, + { + "epoch": 0.0, + "learning_rate": 5.982608695652174e-08, + "loss": 0.2312, + "step": 86 + }, + { + "epoch": 0.0, + "learning_rate": 6.052173913043477e-08, + "loss": 0.2224, + "step": 87 + }, + { + "epoch": 0.0, + "learning_rate": 6.121739130434782e-08, + "loss": 0.1866, + "step": 88 + }, + { + "epoch": 0.0, + "learning_rate": 6.191304347826087e-08, + "loss": 0.1605, + "step": 89 + }, + { + "epoch": 0.0, + "learning_rate": 6.260869565217391e-08, + "loss": 0.1832, + "step": 90 + }, + { + "epoch": 0.0, + "learning_rate": 6.330434782608696e-08, + "loss": 0.1847, + "step": 91 + }, + { + "epoch": 0.0, + "learning_rate": 6.4e-08, + "loss": 0.2278, + "step": 92 + }, + { + "epoch": 0.0, + "learning_rate": 6.469565217391304e-08, + "loss": 0.1611, + "step": 93 + }, + { + "epoch": 0.0, + "learning_rate": 6.539130434782608e-08, + "loss": 0.212, + "step": 94 + }, + { + "epoch": 0.0, + "learning_rate": 6.608695652173913e-08, + "loss": 0.2058, + "step": 95 + }, + { + "epoch": 0.01, + "learning_rate": 6.678260869565218e-08, + "loss": 0.1543, + "step": 96 + }, + { + "epoch": 0.01, + "learning_rate": 6.747826086956522e-08, + "loss": 0.1611, + "step": 97 + }, + { + "epoch": 0.01, + "learning_rate": 6.817391304347826e-08, + "loss": 0.213, + "step": 98 + }, + { + "epoch": 0.01, + "learning_rate": 6.88695652173913e-08, + "loss": 0.2213, + "step": 99 + }, + { + "epoch": 0.01, + "learning_rate": 6.956521739130435e-08, + "loss": 0.1981, + "step": 100 + }, + { + "epoch": 0.01, + "learning_rate": 7.026086956521738e-08, + "loss": 0.1474, + "step": 101 + }, + { + "epoch": 0.01, + "learning_rate": 7.095652173913043e-08, + "loss": 0.1816, + "step": 102 + }, + { + "epoch": 0.01, + "learning_rate": 7.165217391304347e-08, + "loss": 0.2076, + "step": 103 + }, + { + "epoch": 0.01, + "learning_rate": 7.234782608695652e-08, + "loss": 0.187, + "step": 104 + }, + { + "epoch": 0.01, + "learning_rate": 7.304347826086957e-08, + "loss": 0.2154, + "step": 105 + }, + { + "epoch": 0.01, + "learning_rate": 7.37391304347826e-08, + "loss": 0.165, + "step": 106 + }, + { + "epoch": 0.01, + "learning_rate": 7.443478260869564e-08, + "loss": 0.1721, + "step": 107 + }, + { + "epoch": 0.01, + "learning_rate": 7.513043478260869e-08, + "loss": 0.1772, + "step": 108 + }, + { + "epoch": 0.01, + "learning_rate": 7.582608695652174e-08, + "loss": 0.1519, + "step": 109 + }, + { + "epoch": 0.01, + "learning_rate": 7.652173913043478e-08, + "loss": 0.1536, + "step": 110 + }, + { + "epoch": 0.01, + "learning_rate": 7.721739130434782e-08, + "loss": 0.1785, + "step": 111 + }, + { + "epoch": 0.01, + "learning_rate": 7.791304347826086e-08, + "loss": 0.1701, + "step": 112 + }, + { + "epoch": 0.01, + "learning_rate": 7.860869565217391e-08, + "loss": 0.1365, + "step": 113 + }, + { + "epoch": 0.01, + "learning_rate": 7.930434782608694e-08, + "loss": 0.1524, + "step": 114 + }, + { + "epoch": 0.01, + "learning_rate": 8e-08, + "loss": 0.1685, + "step": 115 + }, + { + "epoch": 0.01, + "learning_rate": 8.069565217391305e-08, + "loss": 0.1662, + "step": 116 + }, + { + "epoch": 0.01, + "learning_rate": 8.139130434782608e-08, + "loss": 0.1482, + "step": 117 + }, + { + "epoch": 0.01, + "learning_rate": 8.208695652173913e-08, + "loss": 0.1655, + "step": 118 + }, + { + "epoch": 0.01, + "learning_rate": 8.278260869565217e-08, + "loss": 0.1992, + "step": 119 + }, + { + "epoch": 0.01, + "learning_rate": 8.34782608695652e-08, + "loss": 0.1637, + "step": 120 + }, + { + "epoch": 0.01, + "learning_rate": 8.417391304347827e-08, + "loss": 0.1131, + "step": 121 + }, + { + "epoch": 0.01, + "learning_rate": 8.48695652173913e-08, + "loss": 0.1546, + "step": 122 + }, + { + "epoch": 0.01, + "learning_rate": 8.556521739130435e-08, + "loss": 0.1669, + "step": 123 + }, + { + "epoch": 0.01, + "learning_rate": 8.626086956521739e-08, + "loss": 0.174, + "step": 124 + }, + { + "epoch": 0.01, + "learning_rate": 8.695652173913042e-08, + "loss": 0.1837, + "step": 125 + }, + { + "epoch": 0.01, + "learning_rate": 8.765217391304347e-08, + "loss": 0.1625, + "step": 126 + }, + { + "epoch": 0.01, + "learning_rate": 8.834782608695652e-08, + "loss": 0.1651, + "step": 127 + }, + { + "epoch": 0.01, + "learning_rate": 8.904347826086956e-08, + "loss": 0.1621, + "step": 128 + }, + { + "epoch": 0.01, + "learning_rate": 8.973913043478261e-08, + "loss": 0.1644, + "step": 129 + }, + { + "epoch": 0.01, + "learning_rate": 9.043478260869564e-08, + "loss": 0.1773, + "step": 130 + }, + { + "epoch": 0.01, + "learning_rate": 9.113043478260869e-08, + "loss": 0.1617, + "step": 131 + }, + { + "epoch": 0.01, + "learning_rate": 9.182608695652174e-08, + "loss": 0.1726, + "step": 132 + }, + { + "epoch": 0.01, + "learning_rate": 9.252173913043478e-08, + "loss": 0.1543, + "step": 133 + }, + { + "epoch": 0.01, + "learning_rate": 9.321739130434783e-08, + "loss": 0.1708, + "step": 134 + }, + { + "epoch": 0.01, + "learning_rate": 9.391304347826086e-08, + "loss": 0.1801, + "step": 135 + }, + { + "epoch": 0.01, + "learning_rate": 9.460869565217391e-08, + "loss": 0.1636, + "step": 136 + }, + { + "epoch": 0.01, + "learning_rate": 9.530434782608695e-08, + "loss": 0.136, + "step": 137 + }, + { + "epoch": 0.01, + "learning_rate": 9.599999999999999e-08, + "loss": 0.1706, + "step": 138 + }, + { + "epoch": 0.01, + "learning_rate": 9.669565217391303e-08, + "loss": 0.136, + "step": 139 + }, + { + "epoch": 0.01, + "learning_rate": 9.739130434782609e-08, + "loss": 0.1745, + "step": 140 + }, + { + "epoch": 0.01, + "learning_rate": 9.808695652173912e-08, + "loss": 0.1614, + "step": 141 + }, + { + "epoch": 0.01, + "learning_rate": 9.878260869565217e-08, + "loss": 0.1598, + "step": 142 + }, + { + "epoch": 0.01, + "learning_rate": 9.947826086956522e-08, + "loss": 0.1531, + "step": 143 + }, + { + "epoch": 0.01, + "learning_rate": 1.0017391304347826e-07, + "loss": 0.1743, + "step": 144 + }, + { + "epoch": 0.01, + "learning_rate": 1.008695652173913e-07, + "loss": 0.1828, + "step": 145 + }, + { + "epoch": 0.01, + "learning_rate": 1.0156521739130434e-07, + "loss": 0.2055, + "step": 146 + }, + { + "epoch": 0.01, + "learning_rate": 1.0226086956521738e-07, + "loss": 0.1682, + "step": 147 + }, + { + "epoch": 0.01, + "learning_rate": 1.0295652173913044e-07, + "loss": 0.1405, + "step": 148 + }, + { + "epoch": 0.01, + "learning_rate": 1.0365217391304348e-07, + "loss": 0.186, + "step": 149 + }, + { + "epoch": 0.01, + "learning_rate": 1.0434782608695651e-07, + "loss": 0.196, + "step": 150 + }, + { + "epoch": 0.01, + "learning_rate": 1.0504347826086956e-07, + "loss": 0.2395, + "step": 151 + }, + { + "epoch": 0.01, + "learning_rate": 1.057391304347826e-07, + "loss": 0.1693, + "step": 152 + }, + { + "epoch": 0.01, + "learning_rate": 1.0643478260869565e-07, + "loss": 0.1429, + "step": 153 + }, + { + "epoch": 0.01, + "learning_rate": 1.0713043478260869e-07, + "loss": 0.1611, + "step": 154 + }, + { + "epoch": 0.01, + "learning_rate": 1.0782608695652173e-07, + "loss": 0.1797, + "step": 155 + }, + { + "epoch": 0.01, + "learning_rate": 1.0852173913043479e-07, + "loss": 0.1532, + "step": 156 + }, + { + "epoch": 0.01, + "learning_rate": 1.0921739130434781e-07, + "loss": 0.137, + "step": 157 + }, + { + "epoch": 0.01, + "learning_rate": 1.0991304347826087e-07, + "loss": 0.1513, + "step": 158 + }, + { + "epoch": 0.01, + "learning_rate": 1.106086956521739e-07, + "loss": 0.1704, + "step": 159 + }, + { + "epoch": 0.01, + "learning_rate": 1.1130434782608695e-07, + "loss": 0.1723, + "step": 160 + }, + { + "epoch": 0.01, + "learning_rate": 1.1200000000000001e-07, + "loss": 0.1473, + "step": 161 + }, + { + "epoch": 0.01, + "learning_rate": 1.1269565217391303e-07, + "loss": 0.1563, + "step": 162 + }, + { + "epoch": 0.01, + "learning_rate": 1.1339130434782609e-07, + "loss": 0.1713, + "step": 163 + }, + { + "epoch": 0.01, + "learning_rate": 1.1408695652173912e-07, + "loss": 0.1831, + "step": 164 + }, + { + "epoch": 0.01, + "learning_rate": 1.1478260869565217e-07, + "loss": 0.1708, + "step": 165 + }, + { + "epoch": 0.01, + "learning_rate": 1.154782608695652e-07, + "loss": 0.1451, + "step": 166 + }, + { + "epoch": 0.01, + "learning_rate": 1.1617391304347826e-07, + "loss": 0.1552, + "step": 167 + }, + { + "epoch": 0.01, + "learning_rate": 1.1686956521739131e-07, + "loss": 0.1339, + "step": 168 + }, + { + "epoch": 0.01, + "learning_rate": 1.1756521739130434e-07, + "loss": 0.1572, + "step": 169 + }, + { + "epoch": 0.01, + "learning_rate": 1.1826086956521739e-07, + "loss": 0.1679, + "step": 170 + }, + { + "epoch": 0.01, + "learning_rate": 1.1895652173913042e-07, + "loss": 0.1476, + "step": 171 + }, + { + "epoch": 0.01, + "learning_rate": 1.1965217391304348e-07, + "loss": 0.1718, + "step": 172 + }, + { + "epoch": 0.01, + "learning_rate": 1.203478260869565e-07, + "loss": 0.2379, + "step": 173 + }, + { + "epoch": 0.01, + "learning_rate": 1.2104347826086955e-07, + "loss": 0.1348, + "step": 174 + }, + { + "epoch": 0.01, + "learning_rate": 1.2173913043478262e-07, + "loss": 0.1353, + "step": 175 + }, + { + "epoch": 0.01, + "learning_rate": 1.2243478260869564e-07, + "loss": 0.1328, + "step": 176 + }, + { + "epoch": 0.01, + "learning_rate": 1.2313043478260868e-07, + "loss": 0.157, + "step": 177 + }, + { + "epoch": 0.01, + "learning_rate": 1.2382608695652173e-07, + "loss": 0.1543, + "step": 178 + }, + { + "epoch": 0.01, + "learning_rate": 1.2452173913043478e-07, + "loss": 0.1547, + "step": 179 + }, + { + "epoch": 0.01, + "learning_rate": 1.2521739130434782e-07, + "loss": 0.1462, + "step": 180 + }, + { + "epoch": 0.01, + "learning_rate": 1.2591304347826087e-07, + "loss": 0.156, + "step": 181 + }, + { + "epoch": 0.01, + "learning_rate": 1.2660869565217392e-07, + "loss": 0.1298, + "step": 182 + }, + { + "epoch": 0.01, + "learning_rate": 1.2730434782608694e-07, + "loss": 0.1679, + "step": 183 + }, + { + "epoch": 0.01, + "learning_rate": 1.28e-07, + "loss": 0.1628, + "step": 184 + }, + { + "epoch": 0.01, + "learning_rate": 1.2869565217391303e-07, + "loss": 0.1912, + "step": 185 + }, + { + "epoch": 0.01, + "learning_rate": 1.2939130434782607e-07, + "loss": 0.1908, + "step": 186 + }, + { + "epoch": 0.01, + "learning_rate": 1.3008695652173912e-07, + "loss": 0.1685, + "step": 187 + }, + { + "epoch": 0.01, + "learning_rate": 1.3078260869565217e-07, + "loss": 0.1424, + "step": 188 + }, + { + "epoch": 0.01, + "learning_rate": 1.314782608695652e-07, + "loss": 0.1776, + "step": 189 + }, + { + "epoch": 0.01, + "learning_rate": 1.3217391304347826e-07, + "loss": 0.165, + "step": 190 + }, + { + "epoch": 0.01, + "learning_rate": 1.328695652173913e-07, + "loss": 0.1562, + "step": 191 + }, + { + "epoch": 0.01, + "learning_rate": 1.3356521739130435e-07, + "loss": 0.1716, + "step": 192 + }, + { + "epoch": 0.01, + "learning_rate": 1.3426086956521737e-07, + "loss": 0.1769, + "step": 193 + }, + { + "epoch": 0.01, + "learning_rate": 1.3495652173913044e-07, + "loss": 0.1703, + "step": 194 + }, + { + "epoch": 0.01, + "learning_rate": 1.3565217391304346e-07, + "loss": 0.1627, + "step": 195 + }, + { + "epoch": 0.01, + "learning_rate": 1.363478260869565e-07, + "loss": 0.1555, + "step": 196 + }, + { + "epoch": 0.01, + "learning_rate": 1.3704347826086956e-07, + "loss": 0.1168, + "step": 197 + }, + { + "epoch": 0.01, + "learning_rate": 1.377391304347826e-07, + "loss": 0.1698, + "step": 198 + }, + { + "epoch": 0.01, + "learning_rate": 1.3843478260869565e-07, + "loss": 0.1598, + "step": 199 + }, + { + "epoch": 0.01, + "learning_rate": 1.391304347826087e-07, + "loss": 0.1486, + "step": 200 + }, + { + "epoch": 0.01, + "learning_rate": 1.3982608695652174e-07, + "loss": 0.1836, + "step": 201 + }, + { + "epoch": 0.01, + "learning_rate": 1.4052173913043476e-07, + "loss": 0.1412, + "step": 202 + }, + { + "epoch": 0.01, + "learning_rate": 1.4121739130434783e-07, + "loss": 0.1788, + "step": 203 + }, + { + "epoch": 0.01, + "learning_rate": 1.4191304347826085e-07, + "loss": 0.1577, + "step": 204 + }, + { + "epoch": 0.01, + "learning_rate": 1.426086956521739e-07, + "loss": 0.1502, + "step": 205 + }, + { + "epoch": 0.01, + "learning_rate": 1.4330434782608695e-07, + "loss": 0.1819, + "step": 206 + }, + { + "epoch": 0.01, + "learning_rate": 1.44e-07, + "loss": 0.1581, + "step": 207 + }, + { + "epoch": 0.01, + "learning_rate": 1.4469565217391304e-07, + "loss": 0.1663, + "step": 208 + }, + { + "epoch": 0.01, + "learning_rate": 1.4539130434782606e-07, + "loss": 0.1956, + "step": 209 + }, + { + "epoch": 0.01, + "learning_rate": 1.4608695652173913e-07, + "loss": 0.1732, + "step": 210 + }, + { + "epoch": 0.01, + "learning_rate": 1.4678260869565218e-07, + "loss": 0.1525, + "step": 211 + }, + { + "epoch": 0.01, + "learning_rate": 1.474782608695652e-07, + "loss": 0.1964, + "step": 212 + }, + { + "epoch": 0.01, + "learning_rate": 1.4817391304347827e-07, + "loss": 0.1601, + "step": 213 + }, + { + "epoch": 0.01, + "learning_rate": 1.488695652173913e-07, + "loss": 0.1683, + "step": 214 + }, + { + "epoch": 0.01, + "learning_rate": 1.4956521739130434e-07, + "loss": 0.1451, + "step": 215 + }, + { + "epoch": 0.01, + "learning_rate": 1.5026086956521738e-07, + "loss": 0.1472, + "step": 216 + }, + { + "epoch": 0.01, + "learning_rate": 1.5095652173913043e-07, + "loss": 0.1635, + "step": 217 + }, + { + "epoch": 0.01, + "learning_rate": 1.5165217391304347e-07, + "loss": 0.1393, + "step": 218 + }, + { + "epoch": 0.01, + "learning_rate": 1.5234782608695652e-07, + "loss": 0.143, + "step": 219 + }, + { + "epoch": 0.01, + "learning_rate": 1.5304347826086957e-07, + "loss": 0.1458, + "step": 220 + }, + { + "epoch": 0.01, + "learning_rate": 1.537391304347826e-07, + "loss": 0.1622, + "step": 221 + }, + { + "epoch": 0.01, + "learning_rate": 1.5443478260869563e-07, + "loss": 0.1584, + "step": 222 + }, + { + "epoch": 0.01, + "learning_rate": 1.551304347826087e-07, + "loss": 0.1512, + "step": 223 + }, + { + "epoch": 0.01, + "learning_rate": 1.5582608695652173e-07, + "loss": 0.1979, + "step": 224 + }, + { + "epoch": 0.01, + "learning_rate": 1.5652173913043477e-07, + "loss": 0.1729, + "step": 225 + }, + { + "epoch": 0.01, + "learning_rate": 1.5721739130434782e-07, + "loss": 0.1906, + "step": 226 + }, + { + "epoch": 0.01, + "learning_rate": 1.5791304347826086e-07, + "loss": 0.1928, + "step": 227 + }, + { + "epoch": 0.01, + "learning_rate": 1.5860869565217388e-07, + "loss": 0.1399, + "step": 228 + }, + { + "epoch": 0.01, + "learning_rate": 1.5930434782608696e-07, + "loss": 0.1537, + "step": 229 + }, + { + "epoch": 0.01, + "learning_rate": 1.6e-07, + "loss": 0.1801, + "step": 230 + }, + { + "epoch": 0.01, + "learning_rate": 1.6069565217391302e-07, + "loss": 0.1746, + "step": 231 + }, + { + "epoch": 0.01, + "learning_rate": 1.613913043478261e-07, + "loss": 0.1502, + "step": 232 + }, + { + "epoch": 0.01, + "learning_rate": 1.6208695652173912e-07, + "loss": 0.1684, + "step": 233 + }, + { + "epoch": 0.01, + "learning_rate": 1.6278260869565216e-07, + "loss": 0.1614, + "step": 234 + }, + { + "epoch": 0.01, + "learning_rate": 1.634782608695652e-07, + "loss": 0.1567, + "step": 235 + }, + { + "epoch": 0.01, + "learning_rate": 1.6417391304347825e-07, + "loss": 0.1727, + "step": 236 + }, + { + "epoch": 0.01, + "learning_rate": 1.648695652173913e-07, + "loss": 0.1728, + "step": 237 + }, + { + "epoch": 0.01, + "learning_rate": 1.6556521739130435e-07, + "loss": 0.1778, + "step": 238 + }, + { + "epoch": 0.01, + "learning_rate": 1.662608695652174e-07, + "loss": 0.2088, + "step": 239 + }, + { + "epoch": 0.01, + "learning_rate": 1.669565217391304e-07, + "loss": 0.1565, + "step": 240 + }, + { + "epoch": 0.01, + "learning_rate": 1.6765217391304346e-07, + "loss": 0.162, + "step": 241 + }, + { + "epoch": 0.01, + "learning_rate": 1.6834782608695653e-07, + "loss": 0.151, + "step": 242 + }, + { + "epoch": 0.01, + "learning_rate": 1.6904347826086955e-07, + "loss": 0.127, + "step": 243 + }, + { + "epoch": 0.01, + "learning_rate": 1.697391304347826e-07, + "loss": 0.1463, + "step": 244 + }, + { + "epoch": 0.01, + "learning_rate": 1.7043478260869564e-07, + "loss": 0.141, + "step": 245 + }, + { + "epoch": 0.01, + "learning_rate": 1.711304347826087e-07, + "loss": 0.1981, + "step": 246 + }, + { + "epoch": 0.01, + "learning_rate": 1.718260869565217e-07, + "loss": 0.1935, + "step": 247 + }, + { + "epoch": 0.01, + "learning_rate": 1.7252173913043478e-07, + "loss": 0.1416, + "step": 248 + }, + { + "epoch": 0.01, + "learning_rate": 1.7321739130434783e-07, + "loss": 0.1895, + "step": 249 + }, + { + "epoch": 0.01, + "learning_rate": 1.7391304347826085e-07, + "loss": 0.137, + "step": 250 + }, + { + "epoch": 0.01, + "learning_rate": 1.7460869565217392e-07, + "loss": 0.1227, + "step": 251 + }, + { + "epoch": 0.01, + "learning_rate": 1.7530434782608694e-07, + "loss": 0.1483, + "step": 252 + }, + { + "epoch": 0.01, + "learning_rate": 1.76e-07, + "loss": 0.1864, + "step": 253 + }, + { + "epoch": 0.01, + "learning_rate": 1.7669565217391303e-07, + "loss": 0.2167, + "step": 254 + }, + { + "epoch": 0.01, + "learning_rate": 1.7739130434782608e-07, + "loss": 0.2198, + "step": 255 + }, + { + "epoch": 0.01, + "learning_rate": 1.7808695652173913e-07, + "loss": 0.1428, + "step": 256 + }, + { + "epoch": 0.01, + "learning_rate": 1.7878260869565215e-07, + "loss": 0.1902, + "step": 257 + }, + { + "epoch": 0.01, + "learning_rate": 1.7947826086956522e-07, + "loss": 0.1522, + "step": 258 + }, + { + "epoch": 0.01, + "learning_rate": 1.8017391304347824e-07, + "loss": 0.1727, + "step": 259 + }, + { + "epoch": 0.01, + "learning_rate": 1.8086956521739129e-07, + "loss": 0.1123, + "step": 260 + }, + { + "epoch": 0.01, + "learning_rate": 1.8156521739130436e-07, + "loss": 0.1421, + "step": 261 + }, + { + "epoch": 0.01, + "learning_rate": 1.8226086956521738e-07, + "loss": 0.197, + "step": 262 + }, + { + "epoch": 0.01, + "learning_rate": 1.8295652173913042e-07, + "loss": 0.2031, + "step": 263 + }, + { + "epoch": 0.01, + "learning_rate": 1.8365217391304347e-07, + "loss": 0.1955, + "step": 264 + }, + { + "epoch": 0.01, + "learning_rate": 1.8434782608695652e-07, + "loss": 0.1613, + "step": 265 + }, + { + "epoch": 0.01, + "learning_rate": 1.8504347826086956e-07, + "loss": 0.1763, + "step": 266 + }, + { + "epoch": 0.01, + "learning_rate": 1.857391304347826e-07, + "loss": 0.1856, + "step": 267 + }, + { + "epoch": 0.01, + "learning_rate": 1.8643478260869566e-07, + "loss": 0.1552, + "step": 268 + }, + { + "epoch": 0.01, + "learning_rate": 1.8713043478260867e-07, + "loss": 0.1435, + "step": 269 + }, + { + "epoch": 0.01, + "learning_rate": 1.8782608695652172e-07, + "loss": 0.1467, + "step": 270 + }, + { + "epoch": 0.01, + "learning_rate": 1.8852173913043477e-07, + "loss": 0.1626, + "step": 271 + }, + { + "epoch": 0.01, + "learning_rate": 1.8921739130434781e-07, + "loss": 0.1639, + "step": 272 + }, + { + "epoch": 0.01, + "learning_rate": 1.8991304347826086e-07, + "loss": 0.1423, + "step": 273 + }, + { + "epoch": 0.01, + "learning_rate": 1.906086956521739e-07, + "loss": 0.1735, + "step": 274 + }, + { + "epoch": 0.01, + "learning_rate": 1.9130434782608695e-07, + "loss": 0.1721, + "step": 275 + }, + { + "epoch": 0.01, + "learning_rate": 1.9199999999999997e-07, + "loss": 0.1347, + "step": 276 + }, + { + "epoch": 0.01, + "learning_rate": 1.9269565217391304e-07, + "loss": 0.1576, + "step": 277 + }, + { + "epoch": 0.01, + "learning_rate": 1.9339130434782606e-07, + "loss": 0.1389, + "step": 278 + }, + { + "epoch": 0.01, + "learning_rate": 1.940869565217391e-07, + "loss": 0.165, + "step": 279 + }, + { + "epoch": 0.01, + "learning_rate": 1.9478260869565218e-07, + "loss": 0.1805, + "step": 280 + }, + { + "epoch": 0.01, + "learning_rate": 1.954782608695652e-07, + "loss": 0.1958, + "step": 281 + }, + { + "epoch": 0.01, + "learning_rate": 1.9617391304347825e-07, + "loss": 0.1399, + "step": 282 + }, + { + "epoch": 0.01, + "learning_rate": 1.968695652173913e-07, + "loss": 0.1412, + "step": 283 + }, + { + "epoch": 0.01, + "learning_rate": 1.9756521739130434e-07, + "loss": 0.1853, + "step": 284 + }, + { + "epoch": 0.01, + "learning_rate": 1.982608695652174e-07, + "loss": 0.1467, + "step": 285 + }, + { + "epoch": 0.01, + "learning_rate": 1.9895652173913043e-07, + "loss": 0.1558, + "step": 286 + }, + { + "epoch": 0.01, + "learning_rate": 1.9965217391304348e-07, + "loss": 0.1717, + "step": 287 + }, + { + "epoch": 0.02, + "learning_rate": 2.0034782608695653e-07, + "loss": 0.1407, + "step": 288 + }, + { + "epoch": 0.02, + "learning_rate": 2.0104347826086955e-07, + "loss": 0.1499, + "step": 289 + }, + { + "epoch": 0.02, + "learning_rate": 2.017391304347826e-07, + "loss": 0.1775, + "step": 290 + }, + { + "epoch": 0.02, + "learning_rate": 2.0243478260869567e-07, + "loss": 0.1299, + "step": 291 + }, + { + "epoch": 0.02, + "learning_rate": 2.0313043478260869e-07, + "loss": 0.1474, + "step": 292 + }, + { + "epoch": 0.02, + "learning_rate": 2.0382608695652173e-07, + "loss": 0.1743, + "step": 293 + }, + { + "epoch": 0.02, + "learning_rate": 2.0452173913043475e-07, + "loss": 0.1661, + "step": 294 + }, + { + "epoch": 0.02, + "learning_rate": 2.0521739130434782e-07, + "loss": 0.1164, + "step": 295 + }, + { + "epoch": 0.02, + "learning_rate": 2.0591304347826087e-07, + "loss": 0.1523, + "step": 296 + }, + { + "epoch": 0.02, + "learning_rate": 2.066086956521739e-07, + "loss": 0.21, + "step": 297 + }, + { + "epoch": 0.02, + "learning_rate": 2.0730434782608696e-07, + "loss": 0.147, + "step": 298 + }, + { + "epoch": 0.02, + "learning_rate": 2.08e-07, + "loss": 0.148, + "step": 299 + }, + { + "epoch": 0.02, + "learning_rate": 2.0869565217391303e-07, + "loss": 0.1781, + "step": 300 + }, + { + "epoch": 0.02, + "learning_rate": 2.0939130434782605e-07, + "loss": 0.1769, + "step": 301 + }, + { + "epoch": 0.02, + "learning_rate": 2.1008695652173912e-07, + "loss": 0.1659, + "step": 302 + }, + { + "epoch": 0.02, + "learning_rate": 2.1078260869565217e-07, + "loss": 0.1729, + "step": 303 + }, + { + "epoch": 0.02, + "learning_rate": 2.114782608695652e-07, + "loss": 0.122, + "step": 304 + }, + { + "epoch": 0.02, + "learning_rate": 2.1217391304347826e-07, + "loss": 0.1893, + "step": 305 + }, + { + "epoch": 0.02, + "learning_rate": 2.128695652173913e-07, + "loss": 0.1469, + "step": 306 + }, + { + "epoch": 0.02, + "learning_rate": 2.1356521739130433e-07, + "loss": 0.1468, + "step": 307 + }, + { + "epoch": 0.02, + "learning_rate": 2.1426086956521737e-07, + "loss": 0.1614, + "step": 308 + }, + { + "epoch": 0.02, + "learning_rate": 2.1495652173913045e-07, + "loss": 0.155, + "step": 309 + }, + { + "epoch": 0.02, + "learning_rate": 2.1565217391304347e-07, + "loss": 0.1754, + "step": 310 + }, + { + "epoch": 0.02, + "learning_rate": 2.1634782608695649e-07, + "loss": 0.1868, + "step": 311 + }, + { + "epoch": 0.02, + "learning_rate": 2.1704347826086958e-07, + "loss": 0.1573, + "step": 312 + }, + { + "epoch": 0.02, + "learning_rate": 2.177391304347826e-07, + "loss": 0.1189, + "step": 313 + }, + { + "epoch": 0.02, + "learning_rate": 2.1843478260869562e-07, + "loss": 0.1606, + "step": 314 + }, + { + "epoch": 0.02, + "learning_rate": 2.1913043478260867e-07, + "loss": 0.1137, + "step": 315 + }, + { + "epoch": 0.02, + "learning_rate": 2.1982608695652174e-07, + "loss": 0.136, + "step": 316 + }, + { + "epoch": 0.02, + "learning_rate": 2.2052173913043476e-07, + "loss": 0.1817, + "step": 317 + }, + { + "epoch": 0.02, + "learning_rate": 2.212173913043478e-07, + "loss": 0.1303, + "step": 318 + }, + { + "epoch": 0.02, + "learning_rate": 2.2191304347826088e-07, + "loss": 0.1832, + "step": 319 + }, + { + "epoch": 0.02, + "learning_rate": 2.226086956521739e-07, + "loss": 0.1422, + "step": 320 + }, + { + "epoch": 0.02, + "learning_rate": 2.2330434782608695e-07, + "loss": 0.1357, + "step": 321 + }, + { + "epoch": 0.02, + "learning_rate": 2.2400000000000002e-07, + "loss": 0.169, + "step": 322 + }, + { + "epoch": 0.02, + "learning_rate": 2.2469565217391304e-07, + "loss": 0.1226, + "step": 323 + }, + { + "epoch": 0.02, + "learning_rate": 2.2539130434782606e-07, + "loss": 0.1262, + "step": 324 + }, + { + "epoch": 0.02, + "learning_rate": 2.260869565217391e-07, + "loss": 0.1707, + "step": 325 + }, + { + "epoch": 0.02, + "learning_rate": 2.2678260869565218e-07, + "loss": 0.1648, + "step": 326 + }, + { + "epoch": 0.02, + "learning_rate": 2.274782608695652e-07, + "loss": 0.1402, + "step": 327 + }, + { + "epoch": 0.02, + "learning_rate": 2.2817391304347825e-07, + "loss": 0.1645, + "step": 328 + }, + { + "epoch": 0.02, + "learning_rate": 2.2886956521739132e-07, + "loss": 0.1663, + "step": 329 + }, + { + "epoch": 0.02, + "learning_rate": 2.2956521739130434e-07, + "loss": 0.1445, + "step": 330 + }, + { + "epoch": 0.02, + "learning_rate": 2.3026086956521738e-07, + "loss": 0.1705, + "step": 331 + }, + { + "epoch": 0.02, + "learning_rate": 2.309565217391304e-07, + "loss": 0.1289, + "step": 332 + }, + { + "epoch": 0.02, + "learning_rate": 2.3165217391304348e-07, + "loss": 0.1499, + "step": 333 + }, + { + "epoch": 0.02, + "learning_rate": 2.3234782608695652e-07, + "loss": 0.1764, + "step": 334 + }, + { + "epoch": 0.02, + "learning_rate": 2.3304347826086954e-07, + "loss": 0.1594, + "step": 335 + }, + { + "epoch": 0.02, + "learning_rate": 2.3373913043478262e-07, + "loss": 0.1802, + "step": 336 + }, + { + "epoch": 0.02, + "learning_rate": 2.3443478260869563e-07, + "loss": 0.1616, + "step": 337 + }, + { + "epoch": 0.02, + "learning_rate": 2.3513043478260868e-07, + "loss": 0.157, + "step": 338 + }, + { + "epoch": 0.02, + "learning_rate": 2.358260869565217e-07, + "loss": 0.1871, + "step": 339 + }, + { + "epoch": 0.02, + "learning_rate": 2.3652173913043477e-07, + "loss": 0.1941, + "step": 340 + }, + { + "epoch": 0.02, + "learning_rate": 2.3721739130434782e-07, + "loss": 0.1432, + "step": 341 + }, + { + "epoch": 0.02, + "learning_rate": 2.3791304347826084e-07, + "loss": 0.1536, + "step": 342 + }, + { + "epoch": 0.02, + "learning_rate": 2.3860869565217394e-07, + "loss": 0.1531, + "step": 343 + }, + { + "epoch": 0.02, + "learning_rate": 2.3930434782608696e-07, + "loss": 0.1753, + "step": 344 + }, + { + "epoch": 0.02, + "learning_rate": 2.4e-07, + "loss": 0.1853, + "step": 345 + }, + { + "epoch": 0.02, + "learning_rate": 2.40695652173913e-07, + "loss": 0.1299, + "step": 346 + }, + { + "epoch": 0.02, + "learning_rate": 2.4139130434782607e-07, + "loss": 0.1316, + "step": 347 + }, + { + "epoch": 0.02, + "learning_rate": 2.420869565217391e-07, + "loss": 0.1714, + "step": 348 + }, + { + "epoch": 0.02, + "learning_rate": 2.4278260869565216e-07, + "loss": 0.1318, + "step": 349 + }, + { + "epoch": 0.02, + "learning_rate": 2.4347826086956524e-07, + "loss": 0.1367, + "step": 350 + }, + { + "epoch": 0.02, + "learning_rate": 2.4417391304347826e-07, + "loss": 0.1508, + "step": 351 + }, + { + "epoch": 0.02, + "learning_rate": 2.448695652173913e-07, + "loss": 0.1973, + "step": 352 + }, + { + "epoch": 0.02, + "learning_rate": 2.4556521739130435e-07, + "loss": 0.1318, + "step": 353 + }, + { + "epoch": 0.02, + "learning_rate": 2.4626086956521737e-07, + "loss": 0.1149, + "step": 354 + }, + { + "epoch": 0.02, + "learning_rate": 2.4695652173913044e-07, + "loss": 0.1541, + "step": 355 + }, + { + "epoch": 0.02, + "learning_rate": 2.4765217391304346e-07, + "loss": 0.1696, + "step": 356 + }, + { + "epoch": 0.02, + "learning_rate": 2.4834782608695653e-07, + "loss": 0.1421, + "step": 357 + }, + { + "epoch": 0.02, + "learning_rate": 2.4904347826086955e-07, + "loss": 0.1584, + "step": 358 + }, + { + "epoch": 0.02, + "learning_rate": 2.4973913043478257e-07, + "loss": 0.1774, + "step": 359 + }, + { + "epoch": 0.02, + "learning_rate": 2.5043478260869565e-07, + "loss": 0.1758, + "step": 360 + }, + { + "epoch": 0.02, + "learning_rate": 2.5113043478260867e-07, + "loss": 0.1909, + "step": 361 + }, + { + "epoch": 0.02, + "learning_rate": 2.5182608695652174e-07, + "loss": 0.1694, + "step": 362 + }, + { + "epoch": 0.02, + "learning_rate": 2.5252173913043476e-07, + "loss": 0.1706, + "step": 363 + }, + { + "epoch": 0.02, + "learning_rate": 2.5321739130434783e-07, + "loss": 0.1298, + "step": 364 + }, + { + "epoch": 0.02, + "learning_rate": 2.5391304347826085e-07, + "loss": 0.1621, + "step": 365 + }, + { + "epoch": 0.02, + "learning_rate": 2.5460869565217387e-07, + "loss": 0.1231, + "step": 366 + }, + { + "epoch": 0.02, + "learning_rate": 2.5530434782608694e-07, + "loss": 0.1637, + "step": 367 + }, + { + "epoch": 0.02, + "learning_rate": 2.56e-07, + "loss": 0.1746, + "step": 368 + }, + { + "epoch": 0.02, + "learning_rate": 2.5669565217391304e-07, + "loss": 0.1615, + "step": 369 + }, + { + "epoch": 0.02, + "learning_rate": 2.5739130434782606e-07, + "loss": 0.1659, + "step": 370 + }, + { + "epoch": 0.02, + "learning_rate": 2.5808695652173913e-07, + "loss": 0.1596, + "step": 371 + }, + { + "epoch": 0.02, + "learning_rate": 2.5878260869565215e-07, + "loss": 0.1721, + "step": 372 + }, + { + "epoch": 0.02, + "learning_rate": 2.5947826086956517e-07, + "loss": 0.1596, + "step": 373 + }, + { + "epoch": 0.02, + "learning_rate": 2.6017391304347824e-07, + "loss": 0.1754, + "step": 374 + }, + { + "epoch": 0.02, + "learning_rate": 2.608695652173913e-07, + "loss": 0.1686, + "step": 375 + }, + { + "epoch": 0.02, + "learning_rate": 2.6156521739130433e-07, + "loss": 0.1583, + "step": 376 + }, + { + "epoch": 0.02, + "learning_rate": 2.6226086956521735e-07, + "loss": 0.1625, + "step": 377 + }, + { + "epoch": 0.02, + "learning_rate": 2.629565217391304e-07, + "loss": 0.1187, + "step": 378 + }, + { + "epoch": 0.02, + "learning_rate": 2.6365217391304345e-07, + "loss": 0.1809, + "step": 379 + }, + { + "epoch": 0.02, + "learning_rate": 2.643478260869565e-07, + "loss": 0.1602, + "step": 380 + }, + { + "epoch": 0.02, + "learning_rate": 2.650434782608696e-07, + "loss": 0.1192, + "step": 381 + }, + { + "epoch": 0.02, + "learning_rate": 2.657391304347826e-07, + "loss": 0.1915, + "step": 382 + }, + { + "epoch": 0.02, + "learning_rate": 2.6643478260869563e-07, + "loss": 0.2071, + "step": 383 + }, + { + "epoch": 0.02, + "learning_rate": 2.671304347826087e-07, + "loss": 0.186, + "step": 384 + }, + { + "epoch": 0.02, + "learning_rate": 2.678260869565217e-07, + "loss": 0.1944, + "step": 385 + }, + { + "epoch": 0.02, + "learning_rate": 2.6852173913043474e-07, + "loss": 0.1502, + "step": 386 + }, + { + "epoch": 0.02, + "learning_rate": 2.692173913043478e-07, + "loss": 0.1581, + "step": 387 + }, + { + "epoch": 0.02, + "learning_rate": 2.699130434782609e-07, + "loss": 0.1365, + "step": 388 + }, + { + "epoch": 0.02, + "learning_rate": 2.706086956521739e-07, + "loss": 0.1597, + "step": 389 + }, + { + "epoch": 0.02, + "learning_rate": 2.7130434782608693e-07, + "loss": 0.1456, + "step": 390 + }, + { + "epoch": 0.02, + "learning_rate": 2.72e-07, + "loss": 0.137, + "step": 391 + }, + { + "epoch": 0.02, + "learning_rate": 2.72695652173913e-07, + "loss": 0.132, + "step": 392 + }, + { + "epoch": 0.02, + "learning_rate": 2.733913043478261e-07, + "loss": 0.1383, + "step": 393 + }, + { + "epoch": 0.02, + "learning_rate": 2.740869565217391e-07, + "loss": 0.1902, + "step": 394 + }, + { + "epoch": 0.02, + "learning_rate": 2.747826086956522e-07, + "loss": 0.1576, + "step": 395 + }, + { + "epoch": 0.02, + "learning_rate": 2.754782608695652e-07, + "loss": 0.1722, + "step": 396 + }, + { + "epoch": 0.02, + "learning_rate": 2.761739130434782e-07, + "loss": 0.1272, + "step": 397 + }, + { + "epoch": 0.02, + "learning_rate": 2.768695652173913e-07, + "loss": 0.1652, + "step": 398 + }, + { + "epoch": 0.02, + "learning_rate": 2.775652173913043e-07, + "loss": 0.1871, + "step": 399 + }, + { + "epoch": 0.02, + "learning_rate": 2.782608695652174e-07, + "loss": 0.1394, + "step": 400 + }, + { + "epoch": 0.02, + "learning_rate": 2.789565217391304e-07, + "loss": 0.179, + "step": 401 + }, + { + "epoch": 0.02, + "learning_rate": 2.796521739130435e-07, + "loss": 0.1357, + "step": 402 + }, + { + "epoch": 0.02, + "learning_rate": 2.803478260869565e-07, + "loss": 0.1732, + "step": 403 + }, + { + "epoch": 0.02, + "learning_rate": 2.810434782608695e-07, + "loss": 0.1992, + "step": 404 + }, + { + "epoch": 0.02, + "learning_rate": 2.817391304347826e-07, + "loss": 0.156, + "step": 405 + }, + { + "epoch": 0.02, + "learning_rate": 2.8243478260869567e-07, + "loss": 0.1517, + "step": 406 + }, + { + "epoch": 0.02, + "learning_rate": 2.831304347826087e-07, + "loss": 0.1551, + "step": 407 + }, + { + "epoch": 0.02, + "learning_rate": 2.838260869565217e-07, + "loss": 0.1679, + "step": 408 + }, + { + "epoch": 0.02, + "learning_rate": 2.845217391304348e-07, + "loss": 0.1374, + "step": 409 + }, + { + "epoch": 0.02, + "learning_rate": 2.852173913043478e-07, + "loss": 0.1405, + "step": 410 + }, + { + "epoch": 0.02, + "learning_rate": 2.859130434782608e-07, + "loss": 0.1529, + "step": 411 + }, + { + "epoch": 0.02, + "learning_rate": 2.866086956521739e-07, + "loss": 0.1812, + "step": 412 + }, + { + "epoch": 0.02, + "learning_rate": 2.8730434782608697e-07, + "loss": 0.154, + "step": 413 + }, + { + "epoch": 0.02, + "learning_rate": 2.88e-07, + "loss": 0.1592, + "step": 414 + }, + { + "epoch": 0.02, + "learning_rate": 2.8869565217391306e-07, + "loss": 0.1682, + "step": 415 + }, + { + "epoch": 0.02, + "learning_rate": 2.893913043478261e-07, + "loss": 0.1892, + "step": 416 + }, + { + "epoch": 0.02, + "learning_rate": 2.900869565217391e-07, + "loss": 0.1438, + "step": 417 + }, + { + "epoch": 0.02, + "learning_rate": 2.907826086956521e-07, + "loss": 0.1654, + "step": 418 + }, + { + "epoch": 0.02, + "learning_rate": 2.9147826086956524e-07, + "loss": 0.157, + "step": 419 + }, + { + "epoch": 0.02, + "learning_rate": 2.9217391304347826e-07, + "loss": 0.1498, + "step": 420 + }, + { + "epoch": 0.02, + "learning_rate": 2.928695652173913e-07, + "loss": 0.1399, + "step": 421 + }, + { + "epoch": 0.02, + "learning_rate": 2.9356521739130435e-07, + "loss": 0.172, + "step": 422 + }, + { + "epoch": 0.02, + "learning_rate": 2.942608695652174e-07, + "loss": 0.1538, + "step": 423 + }, + { + "epoch": 0.02, + "learning_rate": 2.949565217391304e-07, + "loss": 0.1914, + "step": 424 + }, + { + "epoch": 0.02, + "learning_rate": 2.9565217391304347e-07, + "loss": 0.1631, + "step": 425 + }, + { + "epoch": 0.02, + "learning_rate": 2.9634782608695654e-07, + "loss": 0.1606, + "step": 426 + }, + { + "epoch": 0.02, + "learning_rate": 2.9704347826086956e-07, + "loss": 0.143, + "step": 427 + }, + { + "epoch": 0.02, + "learning_rate": 2.977391304347826e-07, + "loss": 0.1599, + "step": 428 + }, + { + "epoch": 0.02, + "learning_rate": 2.9843478260869565e-07, + "loss": 0.1918, + "step": 429 + }, + { + "epoch": 0.02, + "learning_rate": 2.9913043478260867e-07, + "loss": 0.1817, + "step": 430 + }, + { + "epoch": 0.02, + "learning_rate": 2.998260869565217e-07, + "loss": 0.1788, + "step": 431 + }, + { + "epoch": 0.02, + "learning_rate": 3.0052173913043476e-07, + "loss": 0.1722, + "step": 432 + }, + { + "epoch": 0.02, + "learning_rate": 3.0121739130434784e-07, + "loss": 0.1391, + "step": 433 + }, + { + "epoch": 0.02, + "learning_rate": 3.0191304347826086e-07, + "loss": 0.1803, + "step": 434 + }, + { + "epoch": 0.02, + "learning_rate": 3.026086956521739e-07, + "loss": 0.1628, + "step": 435 + }, + { + "epoch": 0.02, + "learning_rate": 3.0330434782608695e-07, + "loss": 0.1373, + "step": 436 + }, + { + "epoch": 0.02, + "learning_rate": 3.0399999999999997e-07, + "loss": 0.1341, + "step": 437 + }, + { + "epoch": 0.02, + "learning_rate": 3.0469565217391304e-07, + "loss": 0.1714, + "step": 438 + }, + { + "epoch": 0.02, + "learning_rate": 3.053913043478261e-07, + "loss": 0.1843, + "step": 439 + }, + { + "epoch": 0.02, + "learning_rate": 3.0608695652173913e-07, + "loss": 0.1908, + "step": 440 + }, + { + "epoch": 0.02, + "learning_rate": 3.0678260869565215e-07, + "loss": 0.1706, + "step": 441 + }, + { + "epoch": 0.02, + "learning_rate": 3.074782608695652e-07, + "loss": 0.1585, + "step": 442 + }, + { + "epoch": 0.02, + "learning_rate": 3.0817391304347825e-07, + "loss": 0.1477, + "step": 443 + }, + { + "epoch": 0.02, + "learning_rate": 3.0886956521739127e-07, + "loss": 0.14, + "step": 444 + }, + { + "epoch": 0.02, + "learning_rate": 3.0956521739130434e-07, + "loss": 0.1297, + "step": 445 + }, + { + "epoch": 0.02, + "learning_rate": 3.102608695652174e-07, + "loss": 0.1971, + "step": 446 + }, + { + "epoch": 0.02, + "learning_rate": 3.1095652173913043e-07, + "loss": 0.155, + "step": 447 + }, + { + "epoch": 0.02, + "learning_rate": 3.1165217391304345e-07, + "loss": 0.1585, + "step": 448 + }, + { + "epoch": 0.02, + "learning_rate": 3.1234782608695647e-07, + "loss": 0.1617, + "step": 449 + }, + { + "epoch": 0.02, + "learning_rate": 3.1304347826086954e-07, + "loss": 0.176, + "step": 450 + }, + { + "epoch": 0.02, + "learning_rate": 3.137391304347826e-07, + "loss": 0.1936, + "step": 451 + }, + { + "epoch": 0.02, + "learning_rate": 3.1443478260869564e-07, + "loss": 0.1623, + "step": 452 + }, + { + "epoch": 0.02, + "learning_rate": 3.151304347826087e-07, + "loss": 0.1636, + "step": 453 + }, + { + "epoch": 0.02, + "learning_rate": 3.1582608695652173e-07, + "loss": 0.1344, + "step": 454 + }, + { + "epoch": 0.02, + "learning_rate": 3.1652173913043475e-07, + "loss": 0.1709, + "step": 455 + }, + { + "epoch": 0.02, + "learning_rate": 3.1721739130434777e-07, + "loss": 0.1989, + "step": 456 + }, + { + "epoch": 0.02, + "learning_rate": 3.1791304347826084e-07, + "loss": 0.1504, + "step": 457 + }, + { + "epoch": 0.02, + "learning_rate": 3.186086956521739e-07, + "loss": 0.1441, + "step": 458 + }, + { + "epoch": 0.02, + "learning_rate": 3.1930434782608693e-07, + "loss": 0.1684, + "step": 459 + }, + { + "epoch": 0.02, + "learning_rate": 3.2e-07, + "loss": 0.1678, + "step": 460 + }, + { + "epoch": 0.02, + "learning_rate": 3.2069565217391303e-07, + "loss": 0.162, + "step": 461 + }, + { + "epoch": 0.02, + "learning_rate": 3.2139130434782605e-07, + "loss": 0.1476, + "step": 462 + }, + { + "epoch": 0.02, + "learning_rate": 3.220869565217391e-07, + "loss": 0.1459, + "step": 463 + }, + { + "epoch": 0.02, + "learning_rate": 3.227826086956522e-07, + "loss": 0.1296, + "step": 464 + }, + { + "epoch": 0.02, + "learning_rate": 3.234782608695652e-07, + "loss": 0.1274, + "step": 465 + }, + { + "epoch": 0.02, + "learning_rate": 3.2417391304347823e-07, + "loss": 0.1501, + "step": 466 + }, + { + "epoch": 0.02, + "learning_rate": 3.248695652173913e-07, + "loss": 0.1575, + "step": 467 + }, + { + "epoch": 0.02, + "learning_rate": 3.255652173913043e-07, + "loss": 0.1697, + "step": 468 + }, + { + "epoch": 0.02, + "learning_rate": 3.2626086956521734e-07, + "loss": 0.1744, + "step": 469 + }, + { + "epoch": 0.02, + "learning_rate": 3.269565217391304e-07, + "loss": 0.1918, + "step": 470 + }, + { + "epoch": 0.02, + "learning_rate": 3.276521739130435e-07, + "loss": 0.2023, + "step": 471 + }, + { + "epoch": 0.02, + "learning_rate": 3.283478260869565e-07, + "loss": 0.1402, + "step": 472 + }, + { + "epoch": 0.02, + "learning_rate": 3.2904347826086953e-07, + "loss": 0.1591, + "step": 473 + }, + { + "epoch": 0.02, + "learning_rate": 3.297391304347826e-07, + "loss": 0.1541, + "step": 474 + }, + { + "epoch": 0.02, + "learning_rate": 3.304347826086956e-07, + "loss": 0.1525, + "step": 475 + }, + { + "epoch": 0.02, + "learning_rate": 3.311304347826087e-07, + "loss": 0.1485, + "step": 476 + }, + { + "epoch": 0.02, + "learning_rate": 3.3182608695652177e-07, + "loss": 0.1449, + "step": 477 + }, + { + "epoch": 0.02, + "learning_rate": 3.325217391304348e-07, + "loss": 0.1662, + "step": 478 + }, + { + "epoch": 0.03, + "learning_rate": 3.332173913043478e-07, + "loss": 0.1598, + "step": 479 + }, + { + "epoch": 0.03, + "learning_rate": 3.339130434782608e-07, + "loss": 0.1824, + "step": 480 + }, + { + "epoch": 0.03, + "learning_rate": 3.346086956521739e-07, + "loss": 0.1505, + "step": 481 + }, + { + "epoch": 0.03, + "learning_rate": 3.353043478260869e-07, + "loss": 0.1611, + "step": 482 + }, + { + "epoch": 0.03, + "learning_rate": 3.36e-07, + "loss": 0.1255, + "step": 483 + }, + { + "epoch": 0.03, + "learning_rate": 3.3669565217391306e-07, + "loss": 0.1598, + "step": 484 + }, + { + "epoch": 0.03, + "learning_rate": 3.373913043478261e-07, + "loss": 0.1497, + "step": 485 + }, + { + "epoch": 0.03, + "learning_rate": 3.380869565217391e-07, + "loss": 0.154, + "step": 486 + }, + { + "epoch": 0.03, + "learning_rate": 3.387826086956521e-07, + "loss": 0.1585, + "step": 487 + }, + { + "epoch": 0.03, + "learning_rate": 3.394782608695652e-07, + "loss": 0.1295, + "step": 488 + }, + { + "epoch": 0.03, + "learning_rate": 3.4017391304347827e-07, + "loss": 0.124, + "step": 489 + }, + { + "epoch": 0.03, + "learning_rate": 3.408695652173913e-07, + "loss": 0.1509, + "step": 490 + }, + { + "epoch": 0.03, + "learning_rate": 3.4156521739130436e-07, + "loss": 0.1846, + "step": 491 + }, + { + "epoch": 0.03, + "learning_rate": 3.422608695652174e-07, + "loss": 0.178, + "step": 492 + }, + { + "epoch": 0.03, + "learning_rate": 3.429565217391304e-07, + "loss": 0.1966, + "step": 493 + }, + { + "epoch": 0.03, + "learning_rate": 3.436521739130434e-07, + "loss": 0.166, + "step": 494 + }, + { + "epoch": 0.03, + "learning_rate": 3.443478260869565e-07, + "loss": 0.175, + "step": 495 + }, + { + "epoch": 0.03, + "learning_rate": 3.4504347826086957e-07, + "loss": 0.1677, + "step": 496 + }, + { + "epoch": 0.03, + "learning_rate": 3.457391304347826e-07, + "loss": 0.154, + "step": 497 + }, + { + "epoch": 0.03, + "learning_rate": 3.4643478260869566e-07, + "loss": 0.1715, + "step": 498 + }, + { + "epoch": 0.03, + "learning_rate": 3.471304347826087e-07, + "loss": 0.1594, + "step": 499 + }, + { + "epoch": 0.03, + "learning_rate": 3.478260869565217e-07, + "loss": 0.1449, + "step": 500 + }, + { + "epoch": 0.03, + "learning_rate": 3.4852173913043477e-07, + "loss": 0.1466, + "step": 501 + }, + { + "epoch": 0.03, + "learning_rate": 3.4921739130434784e-07, + "loss": 0.1697, + "step": 502 + }, + { + "epoch": 0.03, + "learning_rate": 3.4991304347826086e-07, + "loss": 0.1394, + "step": 503 + }, + { + "epoch": 0.03, + "learning_rate": 3.506086956521739e-07, + "loss": 0.1364, + "step": 504 + }, + { + "epoch": 0.03, + "learning_rate": 3.5130434782608696e-07, + "loss": 0.1712, + "step": 505 + }, + { + "epoch": 0.03, + "learning_rate": 3.52e-07, + "loss": 0.1417, + "step": 506 + }, + { + "epoch": 0.03, + "learning_rate": 3.52695652173913e-07, + "loss": 0.1602, + "step": 507 + }, + { + "epoch": 0.03, + "learning_rate": 3.5339130434782607e-07, + "loss": 0.1645, + "step": 508 + }, + { + "epoch": 0.03, + "learning_rate": 3.5408695652173914e-07, + "loss": 0.1672, + "step": 509 + }, + { + "epoch": 0.03, + "learning_rate": 3.5478260869565216e-07, + "loss": 0.2001, + "step": 510 + }, + { + "epoch": 0.03, + "learning_rate": 3.554782608695652e-07, + "loss": 0.2011, + "step": 511 + }, + { + "epoch": 0.03, + "learning_rate": 3.5617391304347825e-07, + "loss": 0.1602, + "step": 512 + }, + { + "epoch": 0.03, + "learning_rate": 3.5686956521739127e-07, + "loss": 0.1542, + "step": 513 + }, + { + "epoch": 0.03, + "learning_rate": 3.575652173913043e-07, + "loss": 0.1862, + "step": 514 + }, + { + "epoch": 0.03, + "learning_rate": 3.582608695652174e-07, + "loss": 0.1824, + "step": 515 + }, + { + "epoch": 0.03, + "learning_rate": 3.5895652173913044e-07, + "loss": 0.1795, + "step": 516 + }, + { + "epoch": 0.03, + "learning_rate": 3.5965217391304346e-07, + "loss": 0.1383, + "step": 517 + }, + { + "epoch": 0.03, + "learning_rate": 3.603478260869565e-07, + "loss": 0.1752, + "step": 518 + }, + { + "epoch": 0.03, + "learning_rate": 3.6104347826086955e-07, + "loss": 0.1457, + "step": 519 + }, + { + "epoch": 0.03, + "learning_rate": 3.6173913043478257e-07, + "loss": 0.1607, + "step": 520 + }, + { + "epoch": 0.03, + "learning_rate": 3.6243478260869564e-07, + "loss": 0.1927, + "step": 521 + }, + { + "epoch": 0.03, + "learning_rate": 3.631304347826087e-07, + "loss": 0.1799, + "step": 522 + }, + { + "epoch": 0.03, + "learning_rate": 3.6382608695652174e-07, + "loss": 0.1972, + "step": 523 + }, + { + "epoch": 0.03, + "learning_rate": 3.6452173913043476e-07, + "loss": 0.162, + "step": 524 + }, + { + "epoch": 0.03, + "learning_rate": 3.652173913043478e-07, + "loss": 0.136, + "step": 525 + }, + { + "epoch": 0.03, + "learning_rate": 3.6591304347826085e-07, + "loss": 0.1826, + "step": 526 + }, + { + "epoch": 0.03, + "learning_rate": 3.6660869565217387e-07, + "loss": 0.1412, + "step": 527 + }, + { + "epoch": 0.03, + "learning_rate": 3.6730434782608694e-07, + "loss": 0.1557, + "step": 528 + }, + { + "epoch": 0.03, + "learning_rate": 3.68e-07, + "loss": 0.1595, + "step": 529 + }, + { + "epoch": 0.03, + "learning_rate": 3.6869565217391303e-07, + "loss": 0.1534, + "step": 530 + }, + { + "epoch": 0.03, + "learning_rate": 3.6939130434782605e-07, + "loss": 0.1487, + "step": 531 + }, + { + "epoch": 0.03, + "learning_rate": 3.700869565217391e-07, + "loss": 0.1503, + "step": 532 + }, + { + "epoch": 0.03, + "learning_rate": 3.7078260869565215e-07, + "loss": 0.1495, + "step": 533 + }, + { + "epoch": 0.03, + "learning_rate": 3.714782608695652e-07, + "loss": 0.1314, + "step": 534 + }, + { + "epoch": 0.03, + "learning_rate": 3.7217391304347824e-07, + "loss": 0.1882, + "step": 535 + }, + { + "epoch": 0.03, + "learning_rate": 3.728695652173913e-07, + "loss": 0.136, + "step": 536 + }, + { + "epoch": 0.03, + "learning_rate": 3.7356521739130433e-07, + "loss": 0.1792, + "step": 537 + }, + { + "epoch": 0.03, + "learning_rate": 3.7426086956521735e-07, + "loss": 0.1811, + "step": 538 + }, + { + "epoch": 0.03, + "learning_rate": 3.749565217391304e-07, + "loss": 0.1562, + "step": 539 + }, + { + "epoch": 0.03, + "learning_rate": 3.7565217391304344e-07, + "loss": 0.1422, + "step": 540 + }, + { + "epoch": 0.03, + "learning_rate": 3.763478260869565e-07, + "loss": 0.1315, + "step": 541 + }, + { + "epoch": 0.03, + "learning_rate": 3.7704347826086953e-07, + "loss": 0.158, + "step": 542 + }, + { + "epoch": 0.03, + "learning_rate": 3.777391304347826e-07, + "loss": 0.1648, + "step": 543 + }, + { + "epoch": 0.03, + "learning_rate": 3.7843478260869563e-07, + "loss": 0.1728, + "step": 544 + }, + { + "epoch": 0.03, + "learning_rate": 3.7913043478260865e-07, + "loss": 0.166, + "step": 545 + }, + { + "epoch": 0.03, + "learning_rate": 3.798260869565217e-07, + "loss": 0.1411, + "step": 546 + }, + { + "epoch": 0.03, + "learning_rate": 3.805217391304348e-07, + "loss": 0.1438, + "step": 547 + }, + { + "epoch": 0.03, + "learning_rate": 3.812173913043478e-07, + "loss": 0.1854, + "step": 548 + }, + { + "epoch": 0.03, + "learning_rate": 3.8191304347826083e-07, + "loss": 0.1725, + "step": 549 + }, + { + "epoch": 0.03, + "learning_rate": 3.826086956521739e-07, + "loss": 0.1264, + "step": 550 + }, + { + "epoch": 0.03, + "learning_rate": 3.833043478260869e-07, + "loss": 0.1643, + "step": 551 + }, + { + "epoch": 0.03, + "learning_rate": 3.8399999999999994e-07, + "loss": 0.1727, + "step": 552 + }, + { + "epoch": 0.03, + "learning_rate": 3.84695652173913e-07, + "loss": 0.148, + "step": 553 + }, + { + "epoch": 0.03, + "learning_rate": 3.853913043478261e-07, + "loss": 0.1706, + "step": 554 + }, + { + "epoch": 0.03, + "learning_rate": 3.860869565217391e-07, + "loss": 0.1975, + "step": 555 + }, + { + "epoch": 0.03, + "learning_rate": 3.8678260869565213e-07, + "loss": 0.1395, + "step": 556 + }, + { + "epoch": 0.03, + "learning_rate": 3.874782608695652e-07, + "loss": 0.1718, + "step": 557 + }, + { + "epoch": 0.03, + "learning_rate": 3.881739130434782e-07, + "loss": 0.1972, + "step": 558 + }, + { + "epoch": 0.03, + "learning_rate": 3.888695652173913e-07, + "loss": 0.154, + "step": 559 + }, + { + "epoch": 0.03, + "learning_rate": 3.8956521739130437e-07, + "loss": 0.1465, + "step": 560 + }, + { + "epoch": 0.03, + "learning_rate": 3.902608695652174e-07, + "loss": 0.1485, + "step": 561 + }, + { + "epoch": 0.03, + "learning_rate": 3.909565217391304e-07, + "loss": 0.1405, + "step": 562 + }, + { + "epoch": 0.03, + "learning_rate": 3.916521739130435e-07, + "loss": 0.1716, + "step": 563 + }, + { + "epoch": 0.03, + "learning_rate": 3.923478260869565e-07, + "loss": 0.1848, + "step": 564 + }, + { + "epoch": 0.03, + "learning_rate": 3.930434782608695e-07, + "loss": 0.1596, + "step": 565 + }, + { + "epoch": 0.03, + "learning_rate": 3.937391304347826e-07, + "loss": 0.177, + "step": 566 + }, + { + "epoch": 0.03, + "learning_rate": 3.9443478260869566e-07, + "loss": 0.1251, + "step": 567 + }, + { + "epoch": 0.03, + "learning_rate": 3.951304347826087e-07, + "loss": 0.1826, + "step": 568 + }, + { + "epoch": 0.03, + "learning_rate": 3.958260869565217e-07, + "loss": 0.1361, + "step": 569 + }, + { + "epoch": 0.03, + "learning_rate": 3.965217391304348e-07, + "loss": 0.162, + "step": 570 + }, + { + "epoch": 0.03, + "learning_rate": 3.972173913043478e-07, + "loss": 0.1507, + "step": 571 + }, + { + "epoch": 0.03, + "learning_rate": 3.9791304347826087e-07, + "loss": 0.1288, + "step": 572 + }, + { + "epoch": 0.03, + "learning_rate": 3.986086956521739e-07, + "loss": 0.1396, + "step": 573 + }, + { + "epoch": 0.03, + "learning_rate": 3.9930434782608696e-07, + "loss": 0.1731, + "step": 574 + }, + { + "epoch": 0.03, + "learning_rate": 4e-07, + "loss": 0.1653, + "step": 575 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999999714196137e-07, + "loss": 0.1552, + "step": 576 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999998856784564e-07, + "loss": 0.1656, + "step": 577 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999997427765305e-07, + "loss": 0.1517, + "step": 578 + }, + { + "epoch": 0.03, + "learning_rate": 3.99999954271384e-07, + "loss": 0.1883, + "step": 579 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999992854903896e-07, + "loss": 0.1748, + "step": 580 + }, + { + "epoch": 0.03, + "learning_rate": 3.999998971106188e-07, + "loss": 0.2036, + "step": 581 + }, + { + "epoch": 0.03, + "learning_rate": 3.999998599561244e-07, + "loss": 0.1411, + "step": 582 + }, + { + "epoch": 0.03, + "learning_rate": 3.999998170855568e-07, + "loss": 0.1563, + "step": 583 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999976849891716e-07, + "loss": 0.1345, + "step": 584 + }, + { + "epoch": 0.03, + "learning_rate": 3.99999714196207e-07, + "loss": 0.2119, + "step": 585 + }, + { + "epoch": 0.03, + "learning_rate": 3.999996541774278e-07, + "loss": 0.1388, + "step": 586 + }, + { + "epoch": 0.03, + "learning_rate": 3.999995884425812e-07, + "loss": 0.1482, + "step": 587 + }, + { + "epoch": 0.03, + "learning_rate": 3.999995169916692e-07, + "loss": 0.1418, + "step": 588 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999943982469374e-07, + "loss": 0.1606, + "step": 589 + }, + { + "epoch": 0.03, + "learning_rate": 3.999993569416572e-07, + "loss": 0.1445, + "step": 590 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999926834256175e-07, + "loss": 0.1384, + "step": 591 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999917402741e-07, + "loss": 0.1544, + "step": 592 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999907399620467e-07, + "loss": 0.152, + "step": 593 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999896824894864e-07, + "loss": 0.1824, + "step": 594 + }, + { + "epoch": 0.03, + "learning_rate": 3.999988567856448e-07, + "loss": 0.1438, + "step": 595 + }, + { + "epoch": 0.03, + "learning_rate": 3.999987396062965e-07, + "loss": 0.144, + "step": 596 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999861671090696e-07, + "loss": 0.1418, + "step": 597 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999848809947984e-07, + "loss": 0.1308, + "step": 598 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999835377201864e-07, + "loss": 0.1731, + "step": 599 + }, + { + "epoch": 0.03, + "learning_rate": 3.999982137285273e-07, + "loss": 0.1219, + "step": 600 + }, + { + "epoch": 0.03, + "learning_rate": 3.999980679690098e-07, + "loss": 0.1481, + "step": 601 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999791649347027e-07, + "loss": 0.1448, + "step": 602 + }, + { + "epoch": 0.03, + "learning_rate": 3.999977593019132e-07, + "loss": 0.1311, + "step": 603 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999759639434286e-07, + "loss": 0.1777, + "step": 604 + }, + { + "epoch": 0.03, + "learning_rate": 3.99997427770764e-07, + "loss": 0.122, + "step": 605 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999725343118146e-07, + "loss": 0.1626, + "step": 606 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999707337560025e-07, + "loss": 0.1497, + "step": 607 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999688760402546e-07, + "loss": 0.1535, + "step": 608 + }, + { + "epoch": 0.03, + "learning_rate": 3.999966961164625e-07, + "loss": 0.1657, + "step": 609 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999649891291664e-07, + "loss": 0.1802, + "step": 610 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999629599339366e-07, + "loss": 0.1427, + "step": 611 + }, + { + "epoch": 0.03, + "learning_rate": 3.999960873578993e-07, + "loss": 0.195, + "step": 612 + }, + { + "epoch": 0.03, + "learning_rate": 3.999958730064396e-07, + "loss": 0.1532, + "step": 613 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999565293902066e-07, + "loss": 0.1631, + "step": 614 + }, + { + "epoch": 0.03, + "learning_rate": 3.999954271556487e-07, + "loss": 0.1958, + "step": 615 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999519565633033e-07, + "loss": 0.1643, + "step": 616 + }, + { + "epoch": 0.03, + "learning_rate": 3.99994958441072e-07, + "loss": 0.1341, + "step": 617 + }, + { + "epoch": 0.03, + "learning_rate": 3.999947155098805e-07, + "loss": 0.1424, + "step": 618 + }, + { + "epoch": 0.03, + "learning_rate": 3.999944668627629e-07, + "loss": 0.157, + "step": 619 + }, + { + "epoch": 0.03, + "learning_rate": 3.999942124997262e-07, + "loss": 0.2005, + "step": 620 + }, + { + "epoch": 0.03, + "learning_rate": 3.999939524207777e-07, + "loss": 0.1575, + "step": 621 + }, + { + "epoch": 0.03, + "learning_rate": 3.999936866259248e-07, + "loss": 0.1783, + "step": 622 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999341511517517e-07, + "loss": 0.1437, + "step": 623 + }, + { + "epoch": 0.03, + "learning_rate": 3.999931378885366e-07, + "loss": 0.1434, + "step": 624 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999285494601685e-07, + "loss": 0.203, + "step": 625 + }, + { + "epoch": 0.03, + "learning_rate": 3.999925662876241e-07, + "loss": 0.1857, + "step": 626 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999227191336664e-07, + "loss": 0.1913, + "step": 627 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999197182325284e-07, + "loss": 0.1644, + "step": 628 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999166601729126e-07, + "loss": 0.1554, + "step": 629 + }, + { + "epoch": 0.03, + "learning_rate": 3.999913544954907e-07, + "loss": 0.1738, + "step": 630 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999103725786e-07, + "loss": 0.1749, + "step": 631 + }, + { + "epoch": 0.03, + "learning_rate": 3.999907143044082e-07, + "loss": 0.1339, + "step": 632 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999038563514463e-07, + "loss": 0.1428, + "step": 633 + }, + { + "epoch": 0.03, + "learning_rate": 3.9999005125007864e-07, + "loss": 0.1529, + "step": 634 + }, + { + "epoch": 0.03, + "learning_rate": 3.999897111492197e-07, + "loss": 0.2042, + "step": 635 + }, + { + "epoch": 0.03, + "learning_rate": 3.9998936533257774e-07, + "loss": 0.1649, + "step": 636 + }, + { + "epoch": 0.03, + "learning_rate": 3.999890138001625e-07, + "loss": 0.1956, + "step": 637 + }, + { + "epoch": 0.03, + "learning_rate": 3.9998865655198397e-07, + "loss": 0.1375, + "step": 638 + }, + { + "epoch": 0.03, + "learning_rate": 3.9998829358805244e-07, + "loss": 0.1541, + "step": 639 + }, + { + "epoch": 0.03, + "learning_rate": 3.999879249083783e-07, + "loss": 0.1613, + "step": 640 + }, + { + "epoch": 0.03, + "learning_rate": 3.999875505129721e-07, + "loss": 0.1226, + "step": 641 + }, + { + "epoch": 0.03, + "learning_rate": 3.999871704018444e-07, + "loss": 0.1458, + "step": 642 + }, + { + "epoch": 0.03, + "learning_rate": 3.9998678457500626e-07, + "loss": 0.1633, + "step": 643 + }, + { + "epoch": 0.03, + "learning_rate": 3.999863930324686e-07, + "loss": 0.1498, + "step": 644 + }, + { + "epoch": 0.03, + "learning_rate": 3.9998599577424264e-07, + "loss": 0.1352, + "step": 645 + }, + { + "epoch": 0.03, + "learning_rate": 3.999855928003397e-07, + "loss": 0.1629, + "step": 646 + }, + { + "epoch": 0.03, + "learning_rate": 3.999851841107713e-07, + "loss": 0.1699, + "step": 647 + }, + { + "epoch": 0.03, + "learning_rate": 3.999847697055491e-07, + "loss": 0.1435, + "step": 648 + }, + { + "epoch": 0.03, + "learning_rate": 3.9998434958468504e-07, + "loss": 0.1698, + "step": 649 + }, + { + "epoch": 0.03, + "learning_rate": 3.9998392374819103e-07, + "loss": 0.159, + "step": 650 + }, + { + "epoch": 0.03, + "learning_rate": 3.999834921960793e-07, + "loss": 0.1533, + "step": 651 + }, + { + "epoch": 0.03, + "learning_rate": 3.9998305492836217e-07, + "loss": 0.1816, + "step": 652 + }, + { + "epoch": 0.03, + "learning_rate": 3.999826119450521e-07, + "loss": 0.1731, + "step": 653 + }, + { + "epoch": 0.03, + "learning_rate": 3.999821632461618e-07, + "loss": 0.1312, + "step": 654 + }, + { + "epoch": 0.03, + "learning_rate": 3.999817088317041e-07, + "loss": 0.1493, + "step": 655 + }, + { + "epoch": 0.03, + "learning_rate": 3.9998124870169193e-07, + "loss": 0.1454, + "step": 656 + }, + { + "epoch": 0.03, + "learning_rate": 3.9998078285613844e-07, + "loss": 0.1361, + "step": 657 + }, + { + "epoch": 0.03, + "learning_rate": 3.999803112950571e-07, + "loss": 0.1449, + "step": 658 + }, + { + "epoch": 0.03, + "learning_rate": 3.9997983401846113e-07, + "loss": 0.1928, + "step": 659 + }, + { + "epoch": 0.03, + "learning_rate": 3.999793510263643e-07, + "loss": 0.183, + "step": 660 + }, + { + "epoch": 0.03, + "learning_rate": 3.999788623187805e-07, + "loss": 0.1687, + "step": 661 + }, + { + "epoch": 0.03, + "learning_rate": 3.9997836789572354e-07, + "loss": 0.1381, + "step": 662 + }, + { + "epoch": 0.03, + "learning_rate": 3.9997786775720765e-07, + "loss": 0.1582, + "step": 663 + }, + { + "epoch": 0.03, + "learning_rate": 3.9997736190324716e-07, + "loss": 0.1346, + "step": 664 + }, + { + "epoch": 0.03, + "learning_rate": 3.999768503338564e-07, + "loss": 0.18, + "step": 665 + }, + { + "epoch": 0.03, + "learning_rate": 3.9997633304905014e-07, + "loss": 0.1811, + "step": 666 + }, + { + "epoch": 0.03, + "learning_rate": 3.99975810048843e-07, + "loss": 0.1763, + "step": 667 + }, + { + "epoch": 0.03, + "learning_rate": 3.9997528133324997e-07, + "loss": 0.1722, + "step": 668 + }, + { + "epoch": 0.03, + "learning_rate": 3.999747469022863e-07, + "loss": 0.1337, + "step": 669 + }, + { + "epoch": 0.03, + "learning_rate": 3.999742067559671e-07, + "loss": 0.2022, + "step": 670 + }, + { + "epoch": 0.04, + "learning_rate": 3.999736608943079e-07, + "loss": 0.1595, + "step": 671 + }, + { + "epoch": 0.04, + "learning_rate": 3.9997310931732427e-07, + "loss": 0.1926, + "step": 672 + }, + { + "epoch": 0.04, + "learning_rate": 3.9997255202503204e-07, + "loss": 0.1366, + "step": 673 + }, + { + "epoch": 0.04, + "learning_rate": 3.9997198901744704e-07, + "loss": 0.1288, + "step": 674 + }, + { + "epoch": 0.04, + "learning_rate": 3.9997142029458537e-07, + "loss": 0.1697, + "step": 675 + }, + { + "epoch": 0.04, + "learning_rate": 3.9997084585646333e-07, + "loss": 0.1366, + "step": 676 + }, + { + "epoch": 0.04, + "learning_rate": 3.999702657030973e-07, + "loss": 0.1871, + "step": 677 + }, + { + "epoch": 0.04, + "learning_rate": 3.999696798345039e-07, + "loss": 0.1263, + "step": 678 + }, + { + "epoch": 0.04, + "learning_rate": 3.999690882506998e-07, + "loss": 0.1465, + "step": 679 + }, + { + "epoch": 0.04, + "learning_rate": 3.99968490951702e-07, + "loss": 0.1729, + "step": 680 + }, + { + "epoch": 0.04, + "learning_rate": 3.999678879375276e-07, + "loss": 0.1597, + "step": 681 + }, + { + "epoch": 0.04, + "learning_rate": 3.999672792081937e-07, + "loss": 0.1923, + "step": 682 + }, + { + "epoch": 0.04, + "learning_rate": 3.9996666476371775e-07, + "loss": 0.1581, + "step": 683 + }, + { + "epoch": 0.04, + "learning_rate": 3.999660446041173e-07, + "loss": 0.1377, + "step": 684 + }, + { + "epoch": 0.04, + "learning_rate": 3.999654187294102e-07, + "loss": 0.158, + "step": 685 + }, + { + "epoch": 0.04, + "learning_rate": 3.9996478713961415e-07, + "loss": 0.1395, + "step": 686 + }, + { + "epoch": 0.04, + "learning_rate": 3.999641498347474e-07, + "loss": 0.1435, + "step": 687 + }, + { + "epoch": 0.04, + "learning_rate": 3.999635068148279e-07, + "loss": 0.2051, + "step": 688 + }, + { + "epoch": 0.04, + "learning_rate": 3.999628580798743e-07, + "loss": 0.1457, + "step": 689 + }, + { + "epoch": 0.04, + "learning_rate": 3.99962203629905e-07, + "loss": 0.1683, + "step": 690 + }, + { + "epoch": 0.04, + "learning_rate": 3.999615434649387e-07, + "loss": 0.1355, + "step": 691 + }, + { + "epoch": 0.04, + "learning_rate": 3.9996087758499425e-07, + "loss": 0.1809, + "step": 692 + }, + { + "epoch": 0.04, + "learning_rate": 3.999602059900908e-07, + "loss": 0.1509, + "step": 693 + }, + { + "epoch": 0.04, + "learning_rate": 3.9995952868024747e-07, + "loss": 0.1526, + "step": 694 + }, + { + "epoch": 0.04, + "learning_rate": 3.9995884565548363e-07, + "loss": 0.1564, + "step": 695 + }, + { + "epoch": 0.04, + "learning_rate": 3.999581569158187e-07, + "loss": 0.144, + "step": 696 + }, + { + "epoch": 0.04, + "learning_rate": 3.999574624612726e-07, + "loss": 0.1041, + "step": 697 + }, + { + "epoch": 0.04, + "learning_rate": 3.999567622918649e-07, + "loss": 0.1519, + "step": 698 + }, + { + "epoch": 0.04, + "learning_rate": 3.999560564076158e-07, + "loss": 0.1572, + "step": 699 + }, + { + "epoch": 0.04, + "learning_rate": 3.9995534480854535e-07, + "loss": 0.1493, + "step": 700 + }, + { + "epoch": 0.04, + "learning_rate": 3.9995462749467405e-07, + "loss": 0.1625, + "step": 701 + }, + { + "epoch": 0.04, + "learning_rate": 3.9995390446602223e-07, + "loss": 0.1627, + "step": 702 + }, + { + "epoch": 0.04, + "learning_rate": 3.9995317572261064e-07, + "loss": 0.1801, + "step": 703 + }, + { + "epoch": 0.04, + "learning_rate": 3.999524412644602e-07, + "loss": 0.1622, + "step": 704 + }, + { + "epoch": 0.04, + "learning_rate": 3.999517010915917e-07, + "loss": 0.1572, + "step": 705 + }, + { + "epoch": 0.04, + "learning_rate": 3.999509552040264e-07, + "loss": 0.1771, + "step": 706 + }, + { + "epoch": 0.04, + "learning_rate": 3.9995020360178565e-07, + "loss": 0.1899, + "step": 707 + }, + { + "epoch": 0.04, + "learning_rate": 3.999494462848909e-07, + "loss": 0.1711, + "step": 708 + }, + { + "epoch": 0.04, + "learning_rate": 3.9994868325336377e-07, + "loss": 0.2309, + "step": 709 + }, + { + "epoch": 0.04, + "learning_rate": 3.99947914507226e-07, + "loss": 0.1223, + "step": 710 + }, + { + "epoch": 0.04, + "learning_rate": 3.9994714004649976e-07, + "loss": 0.1573, + "step": 711 + }, + { + "epoch": 0.04, + "learning_rate": 3.9994635987120706e-07, + "loss": 0.1432, + "step": 712 + }, + { + "epoch": 0.04, + "learning_rate": 3.9994557398137016e-07, + "loss": 0.1625, + "step": 713 + }, + { + "epoch": 0.04, + "learning_rate": 3.999447823770116e-07, + "loss": 0.1626, + "step": 714 + }, + { + "epoch": 0.04, + "learning_rate": 3.999439850581539e-07, + "loss": 0.147, + "step": 715 + }, + { + "epoch": 0.04, + "learning_rate": 3.9994318202482005e-07, + "loss": 0.1526, + "step": 716 + }, + { + "epoch": 0.04, + "learning_rate": 3.9994237327703276e-07, + "loss": 0.1639, + "step": 717 + }, + { + "epoch": 0.04, + "learning_rate": 3.999415588148153e-07, + "loss": 0.155, + "step": 718 + }, + { + "epoch": 0.04, + "learning_rate": 3.9994073863819094e-07, + "loss": 0.1452, + "step": 719 + }, + { + "epoch": 0.04, + "learning_rate": 3.99939912747183e-07, + "loss": 0.1507, + "step": 720 + }, + { + "epoch": 0.04, + "learning_rate": 3.999390811418152e-07, + "loss": 0.1396, + "step": 721 + }, + { + "epoch": 0.04, + "learning_rate": 3.9993824382211133e-07, + "loss": 0.1528, + "step": 722 + }, + { + "epoch": 0.04, + "learning_rate": 3.999374007880952e-07, + "loss": 0.1838, + "step": 723 + }, + { + "epoch": 0.04, + "learning_rate": 3.99936552039791e-07, + "loss": 0.157, + "step": 724 + }, + { + "epoch": 0.04, + "learning_rate": 3.9993569757722296e-07, + "loss": 0.1557, + "step": 725 + }, + { + "epoch": 0.04, + "learning_rate": 3.999348374004155e-07, + "loss": 0.1536, + "step": 726 + }, + { + "epoch": 0.04, + "learning_rate": 3.9993397150939324e-07, + "loss": 0.1697, + "step": 727 + }, + { + "epoch": 0.04, + "learning_rate": 3.999330999041808e-07, + "loss": 0.1172, + "step": 728 + }, + { + "epoch": 0.04, + "learning_rate": 3.9993222258480327e-07, + "loss": 0.1962, + "step": 729 + }, + { + "epoch": 0.04, + "learning_rate": 3.999313395512856e-07, + "loss": 0.1389, + "step": 730 + }, + { + "epoch": 0.04, + "learning_rate": 3.9993045080365305e-07, + "loss": 0.1548, + "step": 731 + }, + { + "epoch": 0.04, + "learning_rate": 3.999295563419311e-07, + "loss": 0.2159, + "step": 732 + }, + { + "epoch": 0.04, + "learning_rate": 3.999286561661452e-07, + "loss": 0.1838, + "step": 733 + }, + { + "epoch": 0.04, + "learning_rate": 3.9992775027632115e-07, + "loss": 0.1685, + "step": 734 + }, + { + "epoch": 0.04, + "learning_rate": 3.999268386724848e-07, + "loss": 0.1472, + "step": 735 + }, + { + "epoch": 0.04, + "learning_rate": 3.999259213546622e-07, + "loss": 0.1457, + "step": 736 + }, + { + "epoch": 0.04, + "learning_rate": 3.9992499832287956e-07, + "loss": 0.1324, + "step": 737 + }, + { + "epoch": 0.04, + "learning_rate": 3.9992406957716337e-07, + "loss": 0.1677, + "step": 738 + }, + { + "epoch": 0.04, + "learning_rate": 3.9992313511754005e-07, + "loss": 0.1535, + "step": 739 + }, + { + "epoch": 0.04, + "learning_rate": 3.999221949440364e-07, + "loss": 0.1509, + "step": 740 + }, + { + "epoch": 0.04, + "learning_rate": 3.999212490566792e-07, + "loss": 0.1505, + "step": 741 + }, + { + "epoch": 0.04, + "learning_rate": 3.9992029745549553e-07, + "loss": 0.1551, + "step": 742 + }, + { + "epoch": 0.04, + "learning_rate": 3.999193401405126e-07, + "loss": 0.1503, + "step": 743 + }, + { + "epoch": 0.04, + "learning_rate": 3.999183771117577e-07, + "loss": 0.2017, + "step": 744 + }, + { + "epoch": 0.04, + "learning_rate": 3.999174083692584e-07, + "loss": 0.1876, + "step": 745 + }, + { + "epoch": 0.04, + "learning_rate": 3.9991643391304246e-07, + "loss": 0.1447, + "step": 746 + }, + { + "epoch": 0.04, + "learning_rate": 3.9991545374313765e-07, + "loss": 0.1676, + "step": 747 + }, + { + "epoch": 0.04, + "learning_rate": 3.99914467859572e-07, + "loss": 0.1727, + "step": 748 + }, + { + "epoch": 0.04, + "learning_rate": 3.9991347626237364e-07, + "loss": 0.1852, + "step": 749 + }, + { + "epoch": 0.04, + "learning_rate": 3.99912478951571e-07, + "loss": 0.1923, + "step": 750 + }, + { + "epoch": 0.04, + "learning_rate": 3.999114759271925e-07, + "loss": 0.1531, + "step": 751 + }, + { + "epoch": 0.04, + "learning_rate": 3.9991046718926685e-07, + "loss": 0.1641, + "step": 752 + }, + { + "epoch": 0.04, + "learning_rate": 3.999094527378229e-07, + "loss": 0.2365, + "step": 753 + }, + { + "epoch": 0.04, + "learning_rate": 3.999084325728896e-07, + "loss": 0.1495, + "step": 754 + }, + { + "epoch": 0.04, + "learning_rate": 3.999074066944961e-07, + "loss": 0.1734, + "step": 755 + }, + { + "epoch": 0.04, + "learning_rate": 3.9990637510267174e-07, + "loss": 0.1843, + "step": 756 + }, + { + "epoch": 0.04, + "learning_rate": 3.9990533779744606e-07, + "loss": 0.1597, + "step": 757 + }, + { + "epoch": 0.04, + "learning_rate": 3.999042947788487e-07, + "loss": 0.1536, + "step": 758 + }, + { + "epoch": 0.04, + "learning_rate": 3.999032460469093e-07, + "loss": 0.1992, + "step": 759 + }, + { + "epoch": 0.04, + "learning_rate": 3.9990219160165807e-07, + "loss": 0.1756, + "step": 760 + }, + { + "epoch": 0.04, + "learning_rate": 3.999011314431249e-07, + "loss": 0.151, + "step": 761 + }, + { + "epoch": 0.04, + "learning_rate": 3.9990006557134034e-07, + "loss": 0.1483, + "step": 762 + }, + { + "epoch": 0.04, + "learning_rate": 3.998989939863347e-07, + "loss": 0.1346, + "step": 763 + }, + { + "epoch": 0.04, + "learning_rate": 3.998979166881386e-07, + "loss": 0.1518, + "step": 764 + }, + { + "epoch": 0.04, + "learning_rate": 3.99896833676783e-07, + "loss": 0.1541, + "step": 765 + }, + { + "epoch": 0.04, + "learning_rate": 3.998957449522986e-07, + "loss": 0.1499, + "step": 766 + }, + { + "epoch": 0.04, + "learning_rate": 3.9989465051471667e-07, + "loss": 0.1476, + "step": 767 + }, + { + "epoch": 0.04, + "learning_rate": 3.998935503640685e-07, + "loss": 0.1324, + "step": 768 + }, + { + "epoch": 0.04, + "learning_rate": 3.9989244450038544e-07, + "loss": 0.176, + "step": 769 + }, + { + "epoch": 0.04, + "learning_rate": 3.998913329236992e-07, + "loss": 0.1712, + "step": 770 + }, + { + "epoch": 0.04, + "learning_rate": 3.998902156340415e-07, + "loss": 0.1701, + "step": 771 + }, + { + "epoch": 0.04, + "learning_rate": 3.998890926314442e-07, + "loss": 0.1713, + "step": 772 + }, + { + "epoch": 0.04, + "learning_rate": 3.9988796391593954e-07, + "loss": 0.1463, + "step": 773 + }, + { + "epoch": 0.04, + "learning_rate": 3.998868294875597e-07, + "loss": 0.173, + "step": 774 + }, + { + "epoch": 0.04, + "learning_rate": 3.9988568934633705e-07, + "loss": 0.1666, + "step": 775 + }, + { + "epoch": 0.04, + "learning_rate": 3.9988454349230436e-07, + "loss": 0.1713, + "step": 776 + }, + { + "epoch": 0.04, + "learning_rate": 3.9988339192549416e-07, + "loss": 0.1564, + "step": 777 + }, + { + "epoch": 0.04, + "learning_rate": 3.9988223464593945e-07, + "loss": 0.1616, + "step": 778 + }, + { + "epoch": 0.04, + "learning_rate": 3.9988107165367335e-07, + "loss": 0.1849, + "step": 779 + }, + { + "epoch": 0.04, + "learning_rate": 3.99879902948729e-07, + "loss": 0.1123, + "step": 780 + }, + { + "epoch": 0.04, + "learning_rate": 3.9987872853113995e-07, + "loss": 0.1084, + "step": 781 + }, + { + "epoch": 0.04, + "learning_rate": 3.998775484009396e-07, + "loss": 0.1534, + "step": 782 + }, + { + "epoch": 0.04, + "learning_rate": 3.998763625581618e-07, + "loss": 0.1371, + "step": 783 + }, + { + "epoch": 0.04, + "learning_rate": 3.998751710028403e-07, + "loss": 0.2053, + "step": 784 + }, + { + "epoch": 0.04, + "learning_rate": 3.998739737350094e-07, + "loss": 0.1528, + "step": 785 + }, + { + "epoch": 0.04, + "learning_rate": 3.998727707547031e-07, + "loss": 0.1551, + "step": 786 + }, + { + "epoch": 0.04, + "learning_rate": 3.998715620619559e-07, + "loss": 0.1448, + "step": 787 + }, + { + "epoch": 0.04, + "learning_rate": 3.998703476568022e-07, + "loss": 0.1383, + "step": 788 + }, + { + "epoch": 0.04, + "learning_rate": 3.9986912753927687e-07, + "loss": 0.2011, + "step": 789 + }, + { + "epoch": 0.04, + "learning_rate": 3.998679017094147e-07, + "loss": 0.1774, + "step": 790 + }, + { + "epoch": 0.04, + "learning_rate": 3.998666701672508e-07, + "loss": 0.1711, + "step": 791 + }, + { + "epoch": 0.04, + "learning_rate": 3.998654329128203e-07, + "loss": 0.1549, + "step": 792 + }, + { + "epoch": 0.04, + "learning_rate": 3.998641899461585e-07, + "loss": 0.129, + "step": 793 + }, + { + "epoch": 0.04, + "learning_rate": 3.9986294126730106e-07, + "loss": 0.1834, + "step": 794 + }, + { + "epoch": 0.04, + "learning_rate": 3.998616868762836e-07, + "loss": 0.1294, + "step": 795 + }, + { + "epoch": 0.04, + "learning_rate": 3.998604267731419e-07, + "loss": 0.1227, + "step": 796 + }, + { + "epoch": 0.04, + "learning_rate": 3.9985916095791216e-07, + "loss": 0.1552, + "step": 797 + }, + { + "epoch": 0.04, + "learning_rate": 3.998578894306304e-07, + "loss": 0.1345, + "step": 798 + }, + { + "epoch": 0.04, + "learning_rate": 3.998566121913329e-07, + "loss": 0.1565, + "step": 799 + }, + { + "epoch": 0.04, + "learning_rate": 3.998553292400564e-07, + "loss": 0.1587, + "step": 800 + }, + { + "epoch": 0.04, + "learning_rate": 3.998540405768374e-07, + "loss": 0.1206, + "step": 801 + }, + { + "epoch": 0.04, + "learning_rate": 3.998527462017128e-07, + "loss": 0.1518, + "step": 802 + }, + { + "epoch": 0.04, + "learning_rate": 3.998514461147195e-07, + "loss": 0.1726, + "step": 803 + }, + { + "epoch": 0.04, + "learning_rate": 3.998501403158948e-07, + "loss": 0.1479, + "step": 804 + }, + { + "epoch": 0.04, + "learning_rate": 3.9984882880527587e-07, + "loss": 0.1281, + "step": 805 + }, + { + "epoch": 0.04, + "learning_rate": 3.9984751158290033e-07, + "loss": 0.1458, + "step": 806 + }, + { + "epoch": 0.04, + "learning_rate": 3.998461886488057e-07, + "loss": 0.1802, + "step": 807 + }, + { + "epoch": 0.04, + "learning_rate": 3.998448600030299e-07, + "loss": 0.1721, + "step": 808 + }, + { + "epoch": 0.04, + "learning_rate": 3.998435256456107e-07, + "loss": 0.1245, + "step": 809 + }, + { + "epoch": 0.04, + "learning_rate": 3.998421855765866e-07, + "loss": 0.1531, + "step": 810 + }, + { + "epoch": 0.04, + "learning_rate": 3.998408397959956e-07, + "loss": 0.1389, + "step": 811 + }, + { + "epoch": 0.04, + "learning_rate": 3.998394883038762e-07, + "loss": 0.1756, + "step": 812 + }, + { + "epoch": 0.04, + "learning_rate": 3.9983813110026716e-07, + "loss": 0.1713, + "step": 813 + }, + { + "epoch": 0.04, + "learning_rate": 3.9983676818520713e-07, + "loss": 0.1247, + "step": 814 + }, + { + "epoch": 0.04, + "learning_rate": 3.9983539955873517e-07, + "loss": 0.1588, + "step": 815 + }, + { + "epoch": 0.04, + "learning_rate": 3.9983402522089035e-07, + "loss": 0.153, + "step": 816 + }, + { + "epoch": 0.04, + "learning_rate": 3.998326451717119e-07, + "loss": 0.1566, + "step": 817 + }, + { + "epoch": 0.04, + "learning_rate": 3.998312594112394e-07, + "loss": 0.1714, + "step": 818 + }, + { + "epoch": 0.04, + "learning_rate": 3.998298679395122e-07, + "loss": 0.1818, + "step": 819 + }, + { + "epoch": 0.04, + "learning_rate": 3.9982847075657037e-07, + "loss": 0.1787, + "step": 820 + }, + { + "epoch": 0.04, + "learning_rate": 3.9982706786245366e-07, + "loss": 0.1753, + "step": 821 + }, + { + "epoch": 0.04, + "learning_rate": 3.998256592572022e-07, + "loss": 0.1685, + "step": 822 + }, + { + "epoch": 0.04, + "learning_rate": 3.998242449408563e-07, + "loss": 0.1262, + "step": 823 + }, + { + "epoch": 0.04, + "learning_rate": 3.9982282491345633e-07, + "loss": 0.2134, + "step": 824 + }, + { + "epoch": 0.04, + "learning_rate": 3.998213991750428e-07, + "loss": 0.1481, + "step": 825 + }, + { + "epoch": 0.04, + "learning_rate": 3.9981996772565665e-07, + "loss": 0.1515, + "step": 826 + }, + { + "epoch": 0.04, + "learning_rate": 3.9981853056533865e-07, + "loss": 0.1647, + "step": 827 + }, + { + "epoch": 0.04, + "learning_rate": 3.998170876941299e-07, + "loss": 0.1535, + "step": 828 + }, + { + "epoch": 0.04, + "learning_rate": 3.998156391120716e-07, + "loss": 0.1368, + "step": 829 + }, + { + "epoch": 0.04, + "learning_rate": 3.998141848192053e-07, + "loss": 0.1937, + "step": 830 + }, + { + "epoch": 0.04, + "learning_rate": 3.9981272481557236e-07, + "loss": 0.1416, + "step": 831 + }, + { + "epoch": 0.04, + "learning_rate": 3.9981125910121466e-07, + "loss": 0.127, + "step": 832 + }, + { + "epoch": 0.04, + "learning_rate": 3.99809787676174e-07, + "loss": 0.1685, + "step": 833 + }, + { + "epoch": 0.04, + "learning_rate": 3.9980831054049256e-07, + "loss": 0.1479, + "step": 834 + }, + { + "epoch": 0.04, + "learning_rate": 3.9980682769421244e-07, + "loss": 0.1434, + "step": 835 + }, + { + "epoch": 0.04, + "learning_rate": 3.99805339137376e-07, + "loss": 0.13, + "step": 836 + }, + { + "epoch": 0.04, + "learning_rate": 3.9980384487002586e-07, + "loss": 0.1514, + "step": 837 + }, + { + "epoch": 0.04, + "learning_rate": 3.998023448922047e-07, + "loss": 0.1406, + "step": 838 + }, + { + "epoch": 0.04, + "learning_rate": 3.998008392039554e-07, + "loss": 0.2055, + "step": 839 + }, + { + "epoch": 0.04, + "learning_rate": 3.99799327805321e-07, + "loss": 0.1788, + "step": 840 + }, + { + "epoch": 0.04, + "learning_rate": 3.9979781069634464e-07, + "loss": 0.1962, + "step": 841 + }, + { + "epoch": 0.04, + "learning_rate": 3.9979628787706974e-07, + "loss": 0.1745, + "step": 842 + }, + { + "epoch": 0.04, + "learning_rate": 3.9979475934753986e-07, + "loss": 0.1659, + "step": 843 + }, + { + "epoch": 0.04, + "learning_rate": 3.9979322510779856e-07, + "loss": 0.1377, + "step": 844 + }, + { + "epoch": 0.04, + "learning_rate": 3.9979168515788973e-07, + "loss": 0.2241, + "step": 845 + }, + { + "epoch": 0.04, + "learning_rate": 3.9979013949785747e-07, + "loss": 0.1546, + "step": 846 + }, + { + "epoch": 0.04, + "learning_rate": 3.997885881277459e-07, + "loss": 0.151, + "step": 847 + }, + { + "epoch": 0.04, + "learning_rate": 3.997870310475993e-07, + "loss": 0.2115, + "step": 848 + }, + { + "epoch": 0.04, + "learning_rate": 3.9978546825746226e-07, + "loss": 0.1568, + "step": 849 + }, + { + "epoch": 0.04, + "learning_rate": 3.997838997573795e-07, + "loss": 0.1524, + "step": 850 + }, + { + "epoch": 0.04, + "learning_rate": 3.997823255473956e-07, + "loss": 0.1415, + "step": 851 + }, + { + "epoch": 0.04, + "learning_rate": 3.9978074562755586e-07, + "loss": 0.163, + "step": 852 + }, + { + "epoch": 0.04, + "learning_rate": 3.9977915999790515e-07, + "loss": 0.1309, + "step": 853 + }, + { + "epoch": 0.04, + "learning_rate": 3.99777568658489e-07, + "loss": 0.1612, + "step": 854 + }, + { + "epoch": 0.04, + "learning_rate": 3.997759716093528e-07, + "loss": 0.1677, + "step": 855 + }, + { + "epoch": 0.04, + "learning_rate": 3.997743688505422e-07, + "loss": 0.1811, + "step": 856 + }, + { + "epoch": 0.04, + "learning_rate": 3.9977276038210307e-07, + "loss": 0.1727, + "step": 857 + }, + { + "epoch": 0.04, + "learning_rate": 3.997711462040813e-07, + "loss": 0.1345, + "step": 858 + }, + { + "epoch": 0.04, + "learning_rate": 3.99769526316523e-07, + "loss": 0.1922, + "step": 859 + }, + { + "epoch": 0.04, + "learning_rate": 3.997679007194746e-07, + "loss": 0.1618, + "step": 860 + }, + { + "epoch": 0.04, + "learning_rate": 3.9976626941298246e-07, + "loss": 0.1656, + "step": 861 + }, + { + "epoch": 0.04, + "learning_rate": 3.997646323970932e-07, + "loss": 0.1758, + "step": 862 + }, + { + "epoch": 0.05, + "learning_rate": 3.9976298967185363e-07, + "loss": 0.1412, + "step": 863 + }, + { + "epoch": 0.05, + "learning_rate": 3.9976134123731075e-07, + "loss": 0.1481, + "step": 864 + }, + { + "epoch": 0.05, + "learning_rate": 3.9975968709351154e-07, + "loss": 0.1361, + "step": 865 + }, + { + "epoch": 0.05, + "learning_rate": 3.997580272405034e-07, + "loss": 0.1419, + "step": 866 + }, + { + "epoch": 0.05, + "learning_rate": 3.997563616783337e-07, + "loss": 0.1239, + "step": 867 + }, + { + "epoch": 0.05, + "learning_rate": 3.997546904070501e-07, + "loss": 0.1562, + "step": 868 + }, + { + "epoch": 0.05, + "learning_rate": 3.997530134267003e-07, + "loss": 0.1787, + "step": 869 + }, + { + "epoch": 0.05, + "learning_rate": 3.997513307373323e-07, + "loss": 0.168, + "step": 870 + }, + { + "epoch": 0.05, + "learning_rate": 3.9974964233899417e-07, + "loss": 0.1159, + "step": 871 + }, + { + "epoch": 0.05, + "learning_rate": 3.997479482317341e-07, + "loss": 0.1634, + "step": 872 + }, + { + "epoch": 0.05, + "learning_rate": 3.9974624841560064e-07, + "loss": 0.2077, + "step": 873 + }, + { + "epoch": 0.05, + "learning_rate": 3.997445428906422e-07, + "loss": 0.2213, + "step": 874 + }, + { + "epoch": 0.05, + "learning_rate": 3.9974283165690767e-07, + "loss": 0.153, + "step": 875 + }, + { + "epoch": 0.05, + "learning_rate": 3.997411147144459e-07, + "loss": 0.1495, + "step": 876 + }, + { + "epoch": 0.05, + "learning_rate": 3.997393920633059e-07, + "loss": 0.2021, + "step": 877 + }, + { + "epoch": 0.05, + "learning_rate": 3.99737663703537e-07, + "loss": 0.1824, + "step": 878 + }, + { + "epoch": 0.05, + "learning_rate": 3.9973592963518864e-07, + "loss": 0.1633, + "step": 879 + }, + { + "epoch": 0.05, + "learning_rate": 3.997341898583102e-07, + "loss": 0.1547, + "step": 880 + }, + { + "epoch": 0.05, + "learning_rate": 3.9973244437295154e-07, + "loss": 0.1761, + "step": 881 + }, + { + "epoch": 0.05, + "learning_rate": 3.997306931791625e-07, + "loss": 0.1526, + "step": 882 + }, + { + "epoch": 0.05, + "learning_rate": 3.997289362769932e-07, + "loss": 0.1656, + "step": 883 + }, + { + "epoch": 0.05, + "learning_rate": 3.997271736664938e-07, + "loss": 0.1657, + "step": 884 + }, + { + "epoch": 0.05, + "learning_rate": 3.9972540534771465e-07, + "loss": 0.1276, + "step": 885 + }, + { + "epoch": 0.05, + "learning_rate": 3.997236313207063e-07, + "loss": 0.143, + "step": 886 + }, + { + "epoch": 0.05, + "learning_rate": 3.997218515855195e-07, + "loss": 0.1946, + "step": 887 + }, + { + "epoch": 0.05, + "learning_rate": 3.9972006614220504e-07, + "loss": 0.1311, + "step": 888 + }, + { + "epoch": 0.05, + "learning_rate": 3.99718274990814e-07, + "loss": 0.1632, + "step": 889 + }, + { + "epoch": 0.05, + "learning_rate": 3.997164781313976e-07, + "loss": 0.1603, + "step": 890 + }, + { + "epoch": 0.05, + "learning_rate": 3.997146755640071e-07, + "loss": 0.1311, + "step": 891 + }, + { + "epoch": 0.05, + "learning_rate": 3.997128672886941e-07, + "loss": 0.1775, + "step": 892 + }, + { + "epoch": 0.05, + "learning_rate": 3.997110533055102e-07, + "loss": 0.2015, + "step": 893 + }, + { + "epoch": 0.05, + "learning_rate": 3.997092336145074e-07, + "loss": 0.1834, + "step": 894 + }, + { + "epoch": 0.05, + "learning_rate": 3.9970740821573754e-07, + "loss": 0.1465, + "step": 895 + }, + { + "epoch": 0.05, + "learning_rate": 3.997055771092529e-07, + "loss": 0.1606, + "step": 896 + }, + { + "epoch": 0.05, + "learning_rate": 3.997037402951058e-07, + "loss": 0.1394, + "step": 897 + }, + { + "epoch": 0.05, + "learning_rate": 3.9970189777334866e-07, + "loss": 0.1322, + "step": 898 + }, + { + "epoch": 0.05, + "learning_rate": 3.997000495440342e-07, + "loss": 0.1781, + "step": 899 + }, + { + "epoch": 0.05, + "learning_rate": 3.996981956072152e-07, + "loss": 0.1583, + "step": 900 + }, + { + "epoch": 0.05, + "learning_rate": 3.996963359629447e-07, + "loss": 0.1559, + "step": 901 + }, + { + "epoch": 0.05, + "learning_rate": 3.9969447061127585e-07, + "loss": 0.1678, + "step": 902 + }, + { + "epoch": 0.05, + "learning_rate": 3.996925995522619e-07, + "loss": 0.1788, + "step": 903 + }, + { + "epoch": 0.05, + "learning_rate": 3.996907227859564e-07, + "loss": 0.1529, + "step": 904 + }, + { + "epoch": 0.05, + "learning_rate": 3.9968884031241294e-07, + "loss": 0.166, + "step": 905 + }, + { + "epoch": 0.05, + "learning_rate": 3.996869521316853e-07, + "loss": 0.1351, + "step": 906 + }, + { + "epoch": 0.05, + "learning_rate": 3.9968505824382756e-07, + "loss": 0.1687, + "step": 907 + }, + { + "epoch": 0.05, + "learning_rate": 3.996831586488937e-07, + "loss": 0.1659, + "step": 908 + }, + { + "epoch": 0.05, + "learning_rate": 3.996812533469381e-07, + "loss": 0.165, + "step": 909 + }, + { + "epoch": 0.05, + "learning_rate": 3.9967934233801526e-07, + "loss": 0.1595, + "step": 910 + }, + { + "epoch": 0.05, + "learning_rate": 3.9967742562217965e-07, + "loss": 0.1534, + "step": 911 + }, + { + "epoch": 0.05, + "learning_rate": 3.996755031994861e-07, + "loss": 0.1675, + "step": 912 + }, + { + "epoch": 0.05, + "learning_rate": 3.996735750699897e-07, + "loss": 0.178, + "step": 913 + }, + { + "epoch": 0.05, + "learning_rate": 3.996716412337454e-07, + "loss": 0.1606, + "step": 914 + }, + { + "epoch": 0.05, + "learning_rate": 3.9966970169080847e-07, + "loss": 0.1412, + "step": 915 + }, + { + "epoch": 0.05, + "learning_rate": 3.9966775644123443e-07, + "loss": 0.1245, + "step": 916 + }, + { + "epoch": 0.05, + "learning_rate": 3.996658054850788e-07, + "loss": 0.1301, + "step": 917 + }, + { + "epoch": 0.05, + "learning_rate": 3.9966384882239737e-07, + "loss": 0.1506, + "step": 918 + }, + { + "epoch": 0.05, + "learning_rate": 3.996618864532461e-07, + "loss": 0.1588, + "step": 919 + }, + { + "epoch": 0.05, + "learning_rate": 3.99659918377681e-07, + "loss": 0.1442, + "step": 920 + }, + { + "epoch": 0.05, + "learning_rate": 3.9965794459575837e-07, + "loss": 0.1548, + "step": 921 + }, + { + "epoch": 0.05, + "learning_rate": 3.996559651075346e-07, + "loss": 0.1294, + "step": 922 + }, + { + "epoch": 0.05, + "learning_rate": 3.996539799130663e-07, + "loss": 0.1787, + "step": 923 + }, + { + "epoch": 0.05, + "learning_rate": 3.9965198901241017e-07, + "loss": 0.1588, + "step": 924 + }, + { + "epoch": 0.05, + "learning_rate": 3.996499924056231e-07, + "loss": 0.1828, + "step": 925 + }, + { + "epoch": 0.05, + "learning_rate": 3.996479900927622e-07, + "loss": 0.1318, + "step": 926 + }, + { + "epoch": 0.05, + "learning_rate": 3.996459820738846e-07, + "loss": 0.1321, + "step": 927 + }, + { + "epoch": 0.05, + "learning_rate": 3.9964396834904787e-07, + "loss": 0.1578, + "step": 928 + }, + { + "epoch": 0.05, + "learning_rate": 3.996419489183094e-07, + "loss": 0.1448, + "step": 929 + }, + { + "epoch": 0.05, + "learning_rate": 3.99639923781727e-07, + "loss": 0.1457, + "step": 930 + }, + { + "epoch": 0.05, + "learning_rate": 3.996378929393584e-07, + "loss": 0.1594, + "step": 931 + }, + { + "epoch": 0.05, + "learning_rate": 3.9963585639126183e-07, + "loss": 0.1552, + "step": 932 + }, + { + "epoch": 0.05, + "learning_rate": 3.996338141374954e-07, + "loss": 0.1455, + "step": 933 + }, + { + "epoch": 0.05, + "learning_rate": 3.9963176617811757e-07, + "loss": 0.1701, + "step": 934 + }, + { + "epoch": 0.05, + "learning_rate": 3.996297125131867e-07, + "loss": 0.1987, + "step": 935 + }, + { + "epoch": 0.05, + "learning_rate": 3.9962765314276163e-07, + "loss": 0.1603, + "step": 936 + }, + { + "epoch": 0.05, + "learning_rate": 3.9962558806690117e-07, + "loss": 0.139, + "step": 937 + }, + { + "epoch": 0.05, + "learning_rate": 3.996235172856643e-07, + "loss": 0.1913, + "step": 938 + }, + { + "epoch": 0.05, + "learning_rate": 3.996214407991103e-07, + "loss": 0.1985, + "step": 939 + }, + { + "epoch": 0.05, + "learning_rate": 3.9961935860729846e-07, + "loss": 0.1691, + "step": 940 + }, + { + "epoch": 0.05, + "learning_rate": 3.9961727071028824e-07, + "loss": 0.1721, + "step": 941 + }, + { + "epoch": 0.05, + "learning_rate": 3.996151771081393e-07, + "loss": 0.1668, + "step": 942 + }, + { + "epoch": 0.05, + "learning_rate": 3.996130778009117e-07, + "loss": 0.1527, + "step": 943 + }, + { + "epoch": 0.05, + "learning_rate": 3.9961097278866517e-07, + "loss": 0.1687, + "step": 944 + }, + { + "epoch": 0.05, + "learning_rate": 3.9960886207146e-07, + "loss": 0.1749, + "step": 945 + }, + { + "epoch": 0.05, + "learning_rate": 3.9960674564935654e-07, + "loss": 0.1762, + "step": 946 + }, + { + "epoch": 0.05, + "learning_rate": 3.996046235224152e-07, + "loss": 0.1685, + "step": 947 + }, + { + "epoch": 0.05, + "learning_rate": 3.9960249569069666e-07, + "loss": 0.2005, + "step": 948 + }, + { + "epoch": 0.05, + "learning_rate": 3.9960036215426174e-07, + "loss": 0.1699, + "step": 949 + }, + { + "epoch": 0.05, + "learning_rate": 3.995982229131714e-07, + "loss": 0.1504, + "step": 950 + }, + { + "epoch": 0.05, + "learning_rate": 3.9959607796748687e-07, + "loss": 0.1571, + "step": 951 + }, + { + "epoch": 0.05, + "learning_rate": 3.995939273172693e-07, + "loss": 0.1561, + "step": 952 + }, + { + "epoch": 0.05, + "learning_rate": 3.9959177096258027e-07, + "loss": 0.1858, + "step": 953 + }, + { + "epoch": 0.05, + "learning_rate": 3.9958960890348134e-07, + "loss": 0.1392, + "step": 954 + }, + { + "epoch": 0.05, + "learning_rate": 3.995874411400344e-07, + "loss": 0.1274, + "step": 955 + }, + { + "epoch": 0.05, + "learning_rate": 3.9958526767230125e-07, + "loss": 0.2138, + "step": 956 + }, + { + "epoch": 0.05, + "learning_rate": 3.995830885003442e-07, + "loss": 0.2352, + "step": 957 + }, + { + "epoch": 0.05, + "learning_rate": 3.995809036242254e-07, + "loss": 0.1234, + "step": 958 + }, + { + "epoch": 0.05, + "learning_rate": 3.995787130440073e-07, + "loss": 0.1561, + "step": 959 + }, + { + "epoch": 0.05, + "learning_rate": 3.9957651675975255e-07, + "loss": 0.1784, + "step": 960 + }, + { + "epoch": 0.05, + "learning_rate": 3.9957431477152393e-07, + "loss": 0.1575, + "step": 961 + }, + { + "epoch": 0.05, + "learning_rate": 3.9957210707938434e-07, + "loss": 0.1371, + "step": 962 + }, + { + "epoch": 0.05, + "learning_rate": 3.995698936833969e-07, + "loss": 0.1902, + "step": 963 + }, + { + "epoch": 0.05, + "learning_rate": 3.995676745836248e-07, + "loss": 0.1721, + "step": 964 + }, + { + "epoch": 0.05, + "learning_rate": 3.995654497801316e-07, + "loss": 0.1437, + "step": 965 + }, + { + "epoch": 0.05, + "learning_rate": 3.9956321927298076e-07, + "loss": 0.1369, + "step": 966 + }, + { + "epoch": 0.05, + "learning_rate": 3.995609830622361e-07, + "loss": 0.1332, + "step": 967 + }, + { + "epoch": 0.05, + "learning_rate": 3.9955874114796146e-07, + "loss": 0.1422, + "step": 968 + }, + { + "epoch": 0.05, + "learning_rate": 3.9955649353022097e-07, + "loss": 0.1636, + "step": 969 + }, + { + "epoch": 0.05, + "learning_rate": 3.995542402090789e-07, + "loss": 0.1732, + "step": 970 + }, + { + "epoch": 0.05, + "learning_rate": 3.995519811845996e-07, + "loss": 0.1584, + "step": 971 + }, + { + "epoch": 0.05, + "learning_rate": 3.995497164568476e-07, + "loss": 0.1249, + "step": 972 + }, + { + "epoch": 0.05, + "learning_rate": 3.9954744602588766e-07, + "loss": 0.1947, + "step": 973 + }, + { + "epoch": 0.05, + "learning_rate": 3.9954516989178477e-07, + "loss": 0.1573, + "step": 974 + }, + { + "epoch": 0.05, + "learning_rate": 3.995428880546038e-07, + "loss": 0.1601, + "step": 975 + }, + { + "epoch": 0.05, + "learning_rate": 3.995406005144101e-07, + "loss": 0.1812, + "step": 976 + }, + { + "epoch": 0.05, + "learning_rate": 3.99538307271269e-07, + "loss": 0.1544, + "step": 977 + }, + { + "epoch": 0.05, + "learning_rate": 3.9953600832524605e-07, + "loss": 0.1622, + "step": 978 + }, + { + "epoch": 0.05, + "learning_rate": 3.99533703676407e-07, + "loss": 0.1556, + "step": 979 + }, + { + "epoch": 0.05, + "learning_rate": 3.995313933248176e-07, + "loss": 0.1562, + "step": 980 + }, + { + "epoch": 0.05, + "learning_rate": 3.99529077270544e-07, + "loss": 0.1716, + "step": 981 + }, + { + "epoch": 0.05, + "learning_rate": 3.9952675551365233e-07, + "loss": 0.1443, + "step": 982 + }, + { + "epoch": 0.05, + "learning_rate": 3.9952442805420893e-07, + "loss": 0.1807, + "step": 983 + }, + { + "epoch": 0.05, + "learning_rate": 3.995220948922804e-07, + "loss": 0.1214, + "step": 984 + }, + { + "epoch": 0.05, + "learning_rate": 3.9951975602793333e-07, + "loss": 0.1621, + "step": 985 + }, + { + "epoch": 0.05, + "learning_rate": 3.9951741146123464e-07, + "loss": 0.1397, + "step": 986 + }, + { + "epoch": 0.05, + "learning_rate": 3.9951506119225133e-07, + "loss": 0.173, + "step": 987 + }, + { + "epoch": 0.05, + "learning_rate": 3.9951270522105053e-07, + "loss": 0.2137, + "step": 988 + }, + { + "epoch": 0.05, + "learning_rate": 3.9951034354769953e-07, + "loss": 0.1696, + "step": 989 + }, + { + "epoch": 0.05, + "learning_rate": 3.9950797617226593e-07, + "loss": 0.156, + "step": 990 + }, + { + "epoch": 0.05, + "learning_rate": 3.9950560309481744e-07, + "loss": 0.148, + "step": 991 + }, + { + "epoch": 0.05, + "learning_rate": 3.9950322431542167e-07, + "loss": 0.1769, + "step": 992 + }, + { + "epoch": 0.05, + "learning_rate": 3.9950083983414685e-07, + "loss": 0.126, + "step": 993 + }, + { + "epoch": 0.05, + "learning_rate": 3.9949844965106095e-07, + "loss": 0.1714, + "step": 994 + }, + { + "epoch": 0.05, + "learning_rate": 3.9949605376623233e-07, + "loss": 0.2019, + "step": 995 + }, + { + "epoch": 0.05, + "learning_rate": 3.994936521797295e-07, + "loss": 0.1625, + "step": 996 + }, + { + "epoch": 0.05, + "learning_rate": 3.994912448916211e-07, + "loss": 0.1612, + "step": 997 + }, + { + "epoch": 0.05, + "learning_rate": 3.994888319019758e-07, + "loss": 0.1897, + "step": 998 + }, + { + "epoch": 0.05, + "learning_rate": 3.9948641321086287e-07, + "loss": 0.1732, + "step": 999 + }, + { + "epoch": 0.05, + "learning_rate": 3.9948398881835113e-07, + "loss": 0.1442, + "step": 1000 + }, + { + "epoch": 0.05, + "learning_rate": 3.9948155872450996e-07, + "loss": 0.1793, + "step": 1001 + }, + { + "epoch": 0.05, + "learning_rate": 3.994791229294089e-07, + "loss": 0.1574, + "step": 1002 + }, + { + "epoch": 0.05, + "learning_rate": 3.9947668143311744e-07, + "loss": 0.1818, + "step": 1003 + }, + { + "epoch": 0.05, + "learning_rate": 3.994742342357054e-07, + "loss": 0.1389, + "step": 1004 + }, + { + "epoch": 0.05, + "learning_rate": 3.994717813372428e-07, + "loss": 0.156, + "step": 1005 + }, + { + "epoch": 0.05, + "learning_rate": 3.994693227377997e-07, + "loss": 0.1802, + "step": 1006 + }, + { + "epoch": 0.05, + "learning_rate": 3.9946685843744625e-07, + "loss": 0.1327, + "step": 1007 + }, + { + "epoch": 0.05, + "learning_rate": 3.9946438843625303e-07, + "loss": 0.1753, + "step": 1008 + }, + { + "epoch": 0.05, + "learning_rate": 3.9946191273429064e-07, + "loss": 0.168, + "step": 1009 + }, + { + "epoch": 0.05, + "learning_rate": 3.9945943133162976e-07, + "loss": 0.1884, + "step": 1010 + }, + { + "epoch": 0.05, + "learning_rate": 3.994569442283414e-07, + "loss": 0.1663, + "step": 1011 + }, + { + "epoch": 0.05, + "learning_rate": 3.9945445142449643e-07, + "loss": 0.1753, + "step": 1012 + }, + { + "epoch": 0.05, + "learning_rate": 3.9945195292016633e-07, + "loss": 0.1523, + "step": 1013 + }, + { + "epoch": 0.05, + "learning_rate": 3.994494487154224e-07, + "loss": 0.1409, + "step": 1014 + }, + { + "epoch": 0.05, + "learning_rate": 3.994469388103362e-07, + "loss": 0.1904, + "step": 1015 + }, + { + "epoch": 0.05, + "learning_rate": 3.994444232049796e-07, + "loss": 0.1282, + "step": 1016 + }, + { + "epoch": 0.05, + "learning_rate": 3.9944190189942426e-07, + "loss": 0.19, + "step": 1017 + }, + { + "epoch": 0.05, + "learning_rate": 3.9943937489374246e-07, + "loss": 0.1413, + "step": 1018 + }, + { + "epoch": 0.05, + "learning_rate": 3.994368421880063e-07, + "loss": 0.1521, + "step": 1019 + }, + { + "epoch": 0.05, + "learning_rate": 3.994343037822882e-07, + "loss": 0.1666, + "step": 1020 + }, + { + "epoch": 0.05, + "learning_rate": 3.9943175967666074e-07, + "loss": 0.1184, + "step": 1021 + }, + { + "epoch": 0.05, + "learning_rate": 3.9942920987119653e-07, + "loss": 0.1626, + "step": 1022 + }, + { + "epoch": 0.05, + "learning_rate": 3.994266543659686e-07, + "loss": 0.1241, + "step": 1023 + }, + { + "epoch": 0.05, + "learning_rate": 3.9942409316104985e-07, + "loss": 0.1392, + "step": 1024 + }, + { + "epoch": 0.05, + "learning_rate": 3.9942152625651354e-07, + "loss": 0.1608, + "step": 1025 + }, + { + "epoch": 0.05, + "learning_rate": 3.9941895365243303e-07, + "loss": 0.1751, + "step": 1026 + }, + { + "epoch": 0.05, + "learning_rate": 3.994163753488818e-07, + "loss": 0.1622, + "step": 1027 + }, + { + "epoch": 0.05, + "learning_rate": 3.9941379134593364e-07, + "loss": 0.1869, + "step": 1028 + }, + { + "epoch": 0.05, + "learning_rate": 3.994112016436623e-07, + "loss": 0.1548, + "step": 1029 + }, + { + "epoch": 0.05, + "learning_rate": 3.994086062421418e-07, + "loss": 0.1345, + "step": 1030 + }, + { + "epoch": 0.05, + "learning_rate": 3.9940600514144647e-07, + "loss": 0.1558, + "step": 1031 + }, + { + "epoch": 0.05, + "learning_rate": 3.9940339834165046e-07, + "loss": 0.1526, + "step": 1032 + }, + { + "epoch": 0.05, + "learning_rate": 3.9940078584282836e-07, + "loss": 0.182, + "step": 1033 + }, + { + "epoch": 0.05, + "learning_rate": 3.9939816764505483e-07, + "loss": 0.1625, + "step": 1034 + }, + { + "epoch": 0.05, + "learning_rate": 3.9939554374840466e-07, + "loss": 0.138, + "step": 1035 + }, + { + "epoch": 0.05, + "learning_rate": 3.99392914152953e-07, + "loss": 0.162, + "step": 1036 + }, + { + "epoch": 0.05, + "learning_rate": 3.993902788587747e-07, + "loss": 0.1399, + "step": 1037 + }, + { + "epoch": 0.05, + "learning_rate": 3.9938763786594543e-07, + "loss": 0.1689, + "step": 1038 + }, + { + "epoch": 0.05, + "learning_rate": 3.993849911745403e-07, + "loss": 0.216, + "step": 1039 + }, + { + "epoch": 0.05, + "learning_rate": 3.993823387846353e-07, + "loss": 0.1727, + "step": 1040 + }, + { + "epoch": 0.05, + "learning_rate": 3.9937968069630605e-07, + "loss": 0.1511, + "step": 1041 + }, + { + "epoch": 0.05, + "learning_rate": 3.9937701690962857e-07, + "loss": 0.1142, + "step": 1042 + }, + { + "epoch": 0.05, + "learning_rate": 3.99374347424679e-07, + "loss": 0.1296, + "step": 1043 + }, + { + "epoch": 0.05, + "learning_rate": 3.9937167224153355e-07, + "loss": 0.1539, + "step": 1044 + }, + { + "epoch": 0.05, + "learning_rate": 3.9936899136026884e-07, + "loss": 0.1312, + "step": 1045 + }, + { + "epoch": 0.05, + "learning_rate": 3.9936630478096125e-07, + "loss": 0.1787, + "step": 1046 + }, + { + "epoch": 0.05, + "learning_rate": 3.993636125036878e-07, + "loss": 0.1582, + "step": 1047 + }, + { + "epoch": 0.05, + "learning_rate": 3.993609145285253e-07, + "loss": 0.1247, + "step": 1048 + }, + { + "epoch": 0.05, + "learning_rate": 3.99358210855551e-07, + "loss": 0.1455, + "step": 1049 + }, + { + "epoch": 0.05, + "learning_rate": 3.9935550148484196e-07, + "loss": 0.162, + "step": 1050 + }, + { + "epoch": 0.05, + "learning_rate": 3.9935278641647576e-07, + "loss": 0.1555, + "step": 1051 + }, + { + "epoch": 0.05, + "learning_rate": 3.9935006565053e-07, + "loss": 0.1729, + "step": 1052 + }, + { + "epoch": 0.05, + "learning_rate": 3.993473391870824e-07, + "loss": 0.1309, + "step": 1053 + }, + { + "epoch": 0.06, + "learning_rate": 3.9934460702621086e-07, + "loss": 0.1509, + "step": 1054 + }, + { + "epoch": 0.06, + "learning_rate": 3.993418691679935e-07, + "loss": 0.1459, + "step": 1055 + }, + { + "epoch": 0.06, + "learning_rate": 3.993391256125086e-07, + "loss": 0.1743, + "step": 1056 + }, + { + "epoch": 0.06, + "learning_rate": 3.993363763598345e-07, + "loss": 0.1478, + "step": 1057 + }, + { + "epoch": 0.06, + "learning_rate": 3.9933362141004984e-07, + "loss": 0.1937, + "step": 1058 + }, + { + "epoch": 0.06, + "learning_rate": 3.993308607632333e-07, + "loss": 0.1733, + "step": 1059 + }, + { + "epoch": 0.06, + "learning_rate": 3.9932809441946383e-07, + "loss": 0.1595, + "step": 1060 + }, + { + "epoch": 0.06, + "learning_rate": 3.993253223788205e-07, + "loss": 0.156, + "step": 1061 + }, + { + "epoch": 0.06, + "learning_rate": 3.993225446413825e-07, + "loss": 0.1492, + "step": 1062 + }, + { + "epoch": 0.06, + "learning_rate": 3.993197612072292e-07, + "loss": 0.1602, + "step": 1063 + }, + { + "epoch": 0.06, + "learning_rate": 3.9931697207644015e-07, + "loss": 0.1443, + "step": 1064 + }, + { + "epoch": 0.06, + "learning_rate": 3.993141772490951e-07, + "loss": 0.1361, + "step": 1065 + }, + { + "epoch": 0.06, + "learning_rate": 3.9931137672527393e-07, + "loss": 0.1585, + "step": 1066 + }, + { + "epoch": 0.06, + "learning_rate": 3.9930857050505673e-07, + "loss": 0.1538, + "step": 1067 + }, + { + "epoch": 0.06, + "learning_rate": 3.993057585885236e-07, + "loss": 0.1893, + "step": 1068 + }, + { + "epoch": 0.06, + "learning_rate": 3.9930294097575496e-07, + "loss": 0.1491, + "step": 1069 + }, + { + "epoch": 0.06, + "learning_rate": 3.9930011766683133e-07, + "loss": 0.1528, + "step": 1070 + }, + { + "epoch": 0.06, + "learning_rate": 3.992972886618334e-07, + "loss": 0.1319, + "step": 1071 + }, + { + "epoch": 0.06, + "learning_rate": 3.99294453960842e-07, + "loss": 0.1685, + "step": 1072 + }, + { + "epoch": 0.06, + "learning_rate": 3.992916135639382e-07, + "loss": 0.1155, + "step": 1073 + }, + { + "epoch": 0.06, + "learning_rate": 3.9928876747120316e-07, + "loss": 0.1356, + "step": 1074 + }, + { + "epoch": 0.06, + "learning_rate": 3.9928591568271817e-07, + "loss": 0.205, + "step": 1075 + }, + { + "epoch": 0.06, + "learning_rate": 3.9928305819856473e-07, + "loss": 0.1406, + "step": 1076 + }, + { + "epoch": 0.06, + "learning_rate": 3.992801950188247e-07, + "loss": 0.1366, + "step": 1077 + }, + { + "epoch": 0.06, + "learning_rate": 3.992773261435797e-07, + "loss": 0.1355, + "step": 1078 + }, + { + "epoch": 0.06, + "learning_rate": 3.9927445157291175e-07, + "loss": 0.1365, + "step": 1079 + }, + { + "epoch": 0.06, + "learning_rate": 3.992715713069031e-07, + "loss": 0.1991, + "step": 1080 + }, + { + "epoch": 0.06, + "learning_rate": 3.9926868534563597e-07, + "loss": 0.1759, + "step": 1081 + }, + { + "epoch": 0.06, + "learning_rate": 3.992657936891929e-07, + "loss": 0.1526, + "step": 1082 + }, + { + "epoch": 0.06, + "learning_rate": 3.992628963376565e-07, + "loss": 0.1834, + "step": 1083 + }, + { + "epoch": 0.06, + "learning_rate": 3.992599932911097e-07, + "loss": 0.1735, + "step": 1084 + }, + { + "epoch": 0.06, + "learning_rate": 3.9925708454963523e-07, + "loss": 0.1663, + "step": 1085 + }, + { + "epoch": 0.06, + "learning_rate": 3.9925417011331646e-07, + "loss": 0.1637, + "step": 1086 + }, + { + "epoch": 0.06, + "learning_rate": 3.9925124998223657e-07, + "loss": 0.1724, + "step": 1087 + }, + { + "epoch": 0.06, + "learning_rate": 3.99248324156479e-07, + "loss": 0.1901, + "step": 1088 + }, + { + "epoch": 0.06, + "learning_rate": 3.992453926361275e-07, + "loss": 0.1368, + "step": 1089 + }, + { + "epoch": 0.06, + "learning_rate": 3.992424554212657e-07, + "loss": 0.1342, + "step": 1090 + }, + { + "epoch": 0.06, + "learning_rate": 3.992395125119776e-07, + "loss": 0.1336, + "step": 1091 + }, + { + "epoch": 0.06, + "learning_rate": 3.9923656390834727e-07, + "loss": 0.1992, + "step": 1092 + }, + { + "epoch": 0.06, + "learning_rate": 3.992336096104591e-07, + "loss": 0.1559, + "step": 1093 + }, + { + "epoch": 0.06, + "learning_rate": 3.9923064961839745e-07, + "loss": 0.1416, + "step": 1094 + }, + { + "epoch": 0.06, + "learning_rate": 3.992276839322469e-07, + "loss": 0.1525, + "step": 1095 + }, + { + "epoch": 0.06, + "learning_rate": 3.992247125520922e-07, + "loss": 0.1661, + "step": 1096 + }, + { + "epoch": 0.06, + "learning_rate": 3.992217354780184e-07, + "loss": 0.1507, + "step": 1097 + }, + { + "epoch": 0.06, + "learning_rate": 3.992187527101104e-07, + "loss": 0.1386, + "step": 1098 + }, + { + "epoch": 0.06, + "learning_rate": 3.9921576424845365e-07, + "loss": 0.1838, + "step": 1099 + }, + { + "epoch": 0.06, + "learning_rate": 3.992127700931334e-07, + "loss": 0.1551, + "step": 1100 + }, + { + "epoch": 0.06, + "learning_rate": 3.992097702442352e-07, + "loss": 0.1749, + "step": 1101 + }, + { + "epoch": 0.06, + "learning_rate": 3.99206764701845e-07, + "loss": 0.1527, + "step": 1102 + }, + { + "epoch": 0.06, + "learning_rate": 3.9920375346604847e-07, + "loss": 0.1441, + "step": 1103 + }, + { + "epoch": 0.06, + "learning_rate": 3.992007365369318e-07, + "loss": 0.1184, + "step": 1104 + }, + { + "epoch": 0.06, + "learning_rate": 3.9919771391458117e-07, + "loss": 0.1573, + "step": 1105 + }, + { + "epoch": 0.06, + "learning_rate": 3.9919468559908297e-07, + "loss": 0.181, + "step": 1106 + }, + { + "epoch": 0.06, + "learning_rate": 3.9919165159052376e-07, + "loss": 0.1492, + "step": 1107 + }, + { + "epoch": 0.06, + "learning_rate": 3.9918861188899025e-07, + "loss": 0.1509, + "step": 1108 + }, + { + "epoch": 0.06, + "learning_rate": 3.991855664945694e-07, + "loss": 0.1608, + "step": 1109 + }, + { + "epoch": 0.06, + "learning_rate": 3.9918251540734807e-07, + "loss": 0.1789, + "step": 1110 + }, + { + "epoch": 0.06, + "learning_rate": 3.9917945862741355e-07, + "loss": 0.1498, + "step": 1111 + }, + { + "epoch": 0.06, + "learning_rate": 3.991763961548533e-07, + "loss": 0.1191, + "step": 1112 + }, + { + "epoch": 0.06, + "learning_rate": 3.9917332798975466e-07, + "loss": 0.1447, + "step": 1113 + }, + { + "epoch": 0.06, + "learning_rate": 3.9917025413220545e-07, + "loss": 0.1616, + "step": 1114 + }, + { + "epoch": 0.06, + "learning_rate": 3.991671745822935e-07, + "loss": 0.1482, + "step": 1115 + }, + { + "epoch": 0.06, + "learning_rate": 3.991640893401068e-07, + "loss": 0.1158, + "step": 1116 + }, + { + "epoch": 0.06, + "learning_rate": 3.9916099840573355e-07, + "loss": 0.1259, + "step": 1117 + }, + { + "epoch": 0.06, + "learning_rate": 3.991579017792621e-07, + "loss": 0.1501, + "step": 1118 + }, + { + "epoch": 0.06, + "learning_rate": 3.991547994607809e-07, + "loss": 0.1739, + "step": 1119 + }, + { + "epoch": 0.06, + "learning_rate": 3.991516914503787e-07, + "loss": 0.1451, + "step": 1120 + }, + { + "epoch": 0.06, + "learning_rate": 3.9914857774814423e-07, + "loss": 0.1621, + "step": 1121 + }, + { + "epoch": 0.06, + "learning_rate": 3.9914545835416656e-07, + "loss": 0.1807, + "step": 1122 + }, + { + "epoch": 0.06, + "learning_rate": 3.9914233326853475e-07, + "loss": 0.1531, + "step": 1123 + }, + { + "epoch": 0.06, + "learning_rate": 3.9913920249133827e-07, + "loss": 0.1571, + "step": 1124 + }, + { + "epoch": 0.06, + "learning_rate": 3.991360660226664e-07, + "loss": 0.1199, + "step": 1125 + }, + { + "epoch": 0.06, + "learning_rate": 3.99132923862609e-07, + "loss": 0.1701, + "step": 1126 + }, + { + "epoch": 0.06, + "learning_rate": 3.9912977601125573e-07, + "loss": 0.1365, + "step": 1127 + }, + { + "epoch": 0.06, + "learning_rate": 3.9912662246869655e-07, + "loss": 0.1447, + "step": 1128 + }, + { + "epoch": 0.06, + "learning_rate": 3.991234632350217e-07, + "loss": 0.1747, + "step": 1129 + }, + { + "epoch": 0.06, + "learning_rate": 3.9912029831032133e-07, + "loss": 0.1208, + "step": 1130 + }, + { + "epoch": 0.06, + "learning_rate": 3.9911712769468597e-07, + "loss": 0.1376, + "step": 1131 + }, + { + "epoch": 0.06, + "learning_rate": 3.991139513882063e-07, + "loss": 0.1471, + "step": 1132 + }, + { + "epoch": 0.06, + "learning_rate": 3.99110769390973e-07, + "loss": 0.1875, + "step": 1133 + }, + { + "epoch": 0.06, + "learning_rate": 3.9910758170307704e-07, + "loss": 0.1582, + "step": 1134 + }, + { + "epoch": 0.06, + "learning_rate": 3.991043883246095e-07, + "loss": 0.1518, + "step": 1135 + }, + { + "epoch": 0.06, + "learning_rate": 3.991011892556617e-07, + "loss": 0.1529, + "step": 1136 + }, + { + "epoch": 0.06, + "learning_rate": 3.9909798449632506e-07, + "loss": 0.1267, + "step": 1137 + }, + { + "epoch": 0.06, + "learning_rate": 3.9909477404669117e-07, + "loss": 0.1564, + "step": 1138 + }, + { + "epoch": 0.06, + "learning_rate": 3.990915579068518e-07, + "loss": 0.135, + "step": 1139 + }, + { + "epoch": 0.06, + "learning_rate": 3.990883360768988e-07, + "loss": 0.1585, + "step": 1140 + }, + { + "epoch": 0.06, + "learning_rate": 3.990851085569243e-07, + "loss": 0.1689, + "step": 1141 + }, + { + "epoch": 0.06, + "learning_rate": 3.990818753470206e-07, + "loss": 0.1451, + "step": 1142 + }, + { + "epoch": 0.06, + "learning_rate": 3.9907863644727996e-07, + "loss": 0.1809, + "step": 1143 + }, + { + "epoch": 0.06, + "learning_rate": 3.990753918577951e-07, + "loss": 0.149, + "step": 1144 + }, + { + "epoch": 0.06, + "learning_rate": 3.990721415786587e-07, + "loss": 0.1162, + "step": 1145 + }, + { + "epoch": 0.06, + "learning_rate": 3.990688856099636e-07, + "loss": 0.1659, + "step": 1146 + }, + { + "epoch": 0.06, + "learning_rate": 3.990656239518029e-07, + "loss": 0.1186, + "step": 1147 + }, + { + "epoch": 0.06, + "learning_rate": 3.990623566042699e-07, + "loss": 0.1612, + "step": 1148 + }, + { + "epoch": 0.06, + "learning_rate": 3.9905908356745785e-07, + "loss": 0.1614, + "step": 1149 + }, + { + "epoch": 0.06, + "learning_rate": 3.9905580484146026e-07, + "loss": 0.1402, + "step": 1150 + }, + { + "epoch": 0.06, + "learning_rate": 3.9905252042637103e-07, + "loss": 0.1899, + "step": 1151 + }, + { + "epoch": 0.06, + "learning_rate": 3.9904923032228394e-07, + "loss": 0.1419, + "step": 1152 + }, + { + "epoch": 0.06, + "learning_rate": 3.99045934529293e-07, + "loss": 0.1799, + "step": 1153 + }, + { + "epoch": 0.06, + "learning_rate": 3.9904263304749237e-07, + "loss": 0.1453, + "step": 1154 + }, + { + "epoch": 0.06, + "learning_rate": 3.990393258769764e-07, + "loss": 0.1948, + "step": 1155 + }, + { + "epoch": 0.06, + "learning_rate": 3.990360130178398e-07, + "loss": 0.1586, + "step": 1156 + }, + { + "epoch": 0.06, + "learning_rate": 3.9903269447017695e-07, + "loss": 0.1882, + "step": 1157 + }, + { + "epoch": 0.06, + "learning_rate": 3.9902937023408294e-07, + "loss": 0.1744, + "step": 1158 + }, + { + "epoch": 0.06, + "learning_rate": 3.990260403096527e-07, + "loss": 0.1646, + "step": 1159 + }, + { + "epoch": 0.06, + "learning_rate": 3.9902270469698143e-07, + "loss": 0.1533, + "step": 1160 + }, + { + "epoch": 0.06, + "learning_rate": 3.9901936339616435e-07, + "loss": 0.172, + "step": 1161 + }, + { + "epoch": 0.06, + "learning_rate": 3.990160164072971e-07, + "loss": 0.1405, + "step": 1162 + }, + { + "epoch": 0.06, + "learning_rate": 3.9901266373047523e-07, + "loss": 0.1854, + "step": 1163 + }, + { + "epoch": 0.06, + "learning_rate": 3.990093053657946e-07, + "loss": 0.1883, + "step": 1164 + }, + { + "epoch": 0.06, + "learning_rate": 3.990059413133513e-07, + "loss": 0.1616, + "step": 1165 + }, + { + "epoch": 0.06, + "learning_rate": 3.9900257157324127e-07, + "loss": 0.1466, + "step": 1166 + }, + { + "epoch": 0.06, + "learning_rate": 3.989991961455609e-07, + "loss": 0.1485, + "step": 1167 + }, + { + "epoch": 0.06, + "learning_rate": 3.9899581503040676e-07, + "loss": 0.1641, + "step": 1168 + }, + { + "epoch": 0.06, + "learning_rate": 3.989924282278754e-07, + "loss": 0.1269, + "step": 1169 + }, + { + "epoch": 0.06, + "learning_rate": 3.989890357380636e-07, + "loss": 0.1675, + "step": 1170 + }, + { + "epoch": 0.06, + "learning_rate": 3.9898563756106833e-07, + "loss": 0.1439, + "step": 1171 + }, + { + "epoch": 0.06, + "learning_rate": 3.989822336969867e-07, + "loss": 0.1329, + "step": 1172 + }, + { + "epoch": 0.06, + "learning_rate": 3.9897882414591605e-07, + "loss": 0.2138, + "step": 1173 + }, + { + "epoch": 0.06, + "learning_rate": 3.989754089079538e-07, + "loss": 0.2044, + "step": 1174 + }, + { + "epoch": 0.06, + "learning_rate": 3.989719879831975e-07, + "loss": 0.1459, + "step": 1175 + }, + { + "epoch": 0.06, + "learning_rate": 3.989685613717451e-07, + "loss": 0.1338, + "step": 1176 + }, + { + "epoch": 0.06, + "learning_rate": 3.9896512907369424e-07, + "loss": 0.1394, + "step": 1177 + }, + { + "epoch": 0.06, + "learning_rate": 3.989616910891433e-07, + "loss": 0.1557, + "step": 1178 + }, + { + "epoch": 0.06, + "learning_rate": 3.989582474181904e-07, + "loss": 0.1612, + "step": 1179 + }, + { + "epoch": 0.06, + "learning_rate": 3.989547980609339e-07, + "loss": 0.1432, + "step": 1180 + }, + { + "epoch": 0.06, + "learning_rate": 3.9895134301747255e-07, + "loss": 0.1512, + "step": 1181 + }, + { + "epoch": 0.06, + "learning_rate": 3.98947882287905e-07, + "loss": 0.1249, + "step": 1182 + }, + { + "epoch": 0.06, + "learning_rate": 3.9894441587233014e-07, + "loss": 0.132, + "step": 1183 + }, + { + "epoch": 0.06, + "learning_rate": 3.9894094377084713e-07, + "loss": 0.1549, + "step": 1184 + }, + { + "epoch": 0.06, + "learning_rate": 3.989374659835551e-07, + "loss": 0.1209, + "step": 1185 + }, + { + "epoch": 0.06, + "learning_rate": 3.9893398251055353e-07, + "loss": 0.1486, + "step": 1186 + }, + { + "epoch": 0.06, + "learning_rate": 3.9893049335194195e-07, + "loss": 0.1471, + "step": 1187 + }, + { + "epoch": 0.06, + "learning_rate": 3.9892699850782e-07, + "loss": 0.1945, + "step": 1188 + }, + { + "epoch": 0.06, + "learning_rate": 3.9892349797828775e-07, + "loss": 0.1709, + "step": 1189 + }, + { + "epoch": 0.06, + "learning_rate": 3.989199917634451e-07, + "loss": 0.1317, + "step": 1190 + }, + { + "epoch": 0.06, + "learning_rate": 3.9891647986339224e-07, + "loss": 0.1482, + "step": 1191 + }, + { + "epoch": 0.06, + "learning_rate": 3.9891296227822964e-07, + "loss": 0.1404, + "step": 1192 + }, + { + "epoch": 0.06, + "learning_rate": 3.989094390080578e-07, + "loss": 0.1582, + "step": 1193 + }, + { + "epoch": 0.06, + "learning_rate": 3.9890591005297736e-07, + "loss": 0.1613, + "step": 1194 + }, + { + "epoch": 0.06, + "learning_rate": 3.989023754130893e-07, + "loss": 0.1599, + "step": 1195 + }, + { + "epoch": 0.06, + "learning_rate": 3.9889883508849447e-07, + "loss": 0.1665, + "step": 1196 + }, + { + "epoch": 0.06, + "learning_rate": 3.9889528907929416e-07, + "loss": 0.1533, + "step": 1197 + }, + { + "epoch": 0.06, + "learning_rate": 3.988917373855898e-07, + "loss": 0.1467, + "step": 1198 + }, + { + "epoch": 0.06, + "learning_rate": 3.9888818000748275e-07, + "loss": 0.145, + "step": 1199 + }, + { + "epoch": 0.06, + "learning_rate": 3.9888461694507473e-07, + "loss": 0.1617, + "step": 1200 + }, + { + "epoch": 0.06, + "learning_rate": 3.988810481984676e-07, + "loss": 0.1638, + "step": 1201 + }, + { + "epoch": 0.06, + "learning_rate": 3.988774737677633e-07, + "loss": 0.2183, + "step": 1202 + }, + { + "epoch": 0.06, + "learning_rate": 3.9887389365306405e-07, + "loss": 0.1408, + "step": 1203 + }, + { + "epoch": 0.06, + "learning_rate": 3.988703078544722e-07, + "loss": 0.1696, + "step": 1204 + }, + { + "epoch": 0.06, + "learning_rate": 3.988667163720901e-07, + "loss": 0.1607, + "step": 1205 + }, + { + "epoch": 0.06, + "learning_rate": 3.988631192060205e-07, + "loss": 0.1728, + "step": 1206 + }, + { + "epoch": 0.06, + "learning_rate": 3.9885951635636614e-07, + "loss": 0.1425, + "step": 1207 + }, + { + "epoch": 0.06, + "learning_rate": 3.988559078232301e-07, + "loss": 0.1567, + "step": 1208 + }, + { + "epoch": 0.06, + "learning_rate": 3.988522936067154e-07, + "loss": 0.125, + "step": 1209 + }, + { + "epoch": 0.06, + "learning_rate": 3.9884867370692544e-07, + "loss": 0.1259, + "step": 1210 + }, + { + "epoch": 0.06, + "learning_rate": 3.9884504812396363e-07, + "loss": 0.181, + "step": 1211 + }, + { + "epoch": 0.06, + "learning_rate": 3.988414168579335e-07, + "loss": 0.204, + "step": 1212 + }, + { + "epoch": 0.06, + "learning_rate": 3.9883777990893897e-07, + "loss": 0.1914, + "step": 1213 + }, + { + "epoch": 0.06, + "learning_rate": 3.9883413727708396e-07, + "loss": 0.1211, + "step": 1214 + }, + { + "epoch": 0.06, + "learning_rate": 3.9883048896247253e-07, + "loss": 0.1338, + "step": 1215 + }, + { + "epoch": 0.06, + "learning_rate": 3.9882683496520895e-07, + "loss": 0.1364, + "step": 1216 + }, + { + "epoch": 0.06, + "learning_rate": 3.9882317528539764e-07, + "loss": 0.1593, + "step": 1217 + }, + { + "epoch": 0.06, + "learning_rate": 3.988195099231433e-07, + "loss": 0.1577, + "step": 1218 + }, + { + "epoch": 0.06, + "learning_rate": 3.988158388785506e-07, + "loss": 0.1853, + "step": 1219 + }, + { + "epoch": 0.06, + "learning_rate": 3.988121621517244e-07, + "loss": 0.1741, + "step": 1220 + }, + { + "epoch": 0.06, + "learning_rate": 3.9880847974276996e-07, + "loss": 0.1247, + "step": 1221 + }, + { + "epoch": 0.06, + "learning_rate": 3.9880479165179234e-07, + "loss": 0.1532, + "step": 1222 + }, + { + "epoch": 0.06, + "learning_rate": 3.9880109787889706e-07, + "loss": 0.1262, + "step": 1223 + }, + { + "epoch": 0.06, + "learning_rate": 3.9879739842418964e-07, + "loss": 0.1464, + "step": 1224 + }, + { + "epoch": 0.06, + "learning_rate": 3.9879369328777583e-07, + "loss": 0.1516, + "step": 1225 + }, + { + "epoch": 0.06, + "learning_rate": 3.9878998246976153e-07, + "loss": 0.1686, + "step": 1226 + }, + { + "epoch": 0.06, + "learning_rate": 3.9878626597025283e-07, + "loss": 0.1455, + "step": 1227 + }, + { + "epoch": 0.06, + "learning_rate": 3.9878254378935587e-07, + "loss": 0.1627, + "step": 1228 + }, + { + "epoch": 0.06, + "learning_rate": 3.987788159271771e-07, + "loss": 0.1438, + "step": 1229 + }, + { + "epoch": 0.06, + "learning_rate": 3.98775082383823e-07, + "loss": 0.1661, + "step": 1230 + }, + { + "epoch": 0.06, + "learning_rate": 3.9877134315940034e-07, + "loss": 0.1785, + "step": 1231 + }, + { + "epoch": 0.06, + "learning_rate": 3.987675982540159e-07, + "loss": 0.1348, + "step": 1232 + }, + { + "epoch": 0.06, + "learning_rate": 3.987638476677769e-07, + "loss": 0.169, + "step": 1233 + }, + { + "epoch": 0.06, + "learning_rate": 3.987600914007903e-07, + "loss": 0.174, + "step": 1234 + }, + { + "epoch": 0.06, + "learning_rate": 3.987563294531635e-07, + "loss": 0.109, + "step": 1235 + }, + { + "epoch": 0.06, + "learning_rate": 3.987525618250042e-07, + "loss": 0.2232, + "step": 1236 + }, + { + "epoch": 0.06, + "learning_rate": 3.987487885164199e-07, + "loss": 0.1482, + "step": 1237 + }, + { + "epoch": 0.06, + "learning_rate": 3.987450095275185e-07, + "loss": 0.1451, + "step": 1238 + }, + { + "epoch": 0.06, + "learning_rate": 3.9874122485840805e-07, + "loss": 0.1453, + "step": 1239 + }, + { + "epoch": 0.06, + "learning_rate": 3.987374345091966e-07, + "loss": 0.1646, + "step": 1240 + }, + { + "epoch": 0.06, + "learning_rate": 3.987336384799926e-07, + "loss": 0.1582, + "step": 1241 + }, + { + "epoch": 0.06, + "learning_rate": 3.9872983677090444e-07, + "loss": 0.1816, + "step": 1242 + }, + { + "epoch": 0.06, + "learning_rate": 3.987260293820409e-07, + "loss": 0.1562, + "step": 1243 + }, + { + "epoch": 0.06, + "learning_rate": 3.9872221631351067e-07, + "loss": 0.1912, + "step": 1244 + }, + { + "epoch": 0.06, + "learning_rate": 3.9871839756542284e-07, + "loss": 0.1296, + "step": 1245 + }, + { + "epoch": 0.07, + "learning_rate": 3.987145731378864e-07, + "loss": 0.1392, + "step": 1246 + }, + { + "epoch": 0.07, + "learning_rate": 3.987107430310108e-07, + "loss": 0.1242, + "step": 1247 + }, + { + "epoch": 0.07, + "learning_rate": 3.9870690724490545e-07, + "loss": 0.1919, + "step": 1248 + }, + { + "epoch": 0.07, + "learning_rate": 3.9870306577968e-07, + "loss": 0.1569, + "step": 1249 + }, + { + "epoch": 0.07, + "learning_rate": 3.986992186354443e-07, + "loss": 0.1173, + "step": 1250 + }, + { + "epoch": 0.07, + "learning_rate": 3.986953658123081e-07, + "loss": 0.1416, + "step": 1251 + }, + { + "epoch": 0.07, + "learning_rate": 3.9869150731038166e-07, + "loss": 0.1664, + "step": 1252 + }, + { + "epoch": 0.07, + "learning_rate": 3.9868764312977525e-07, + "loss": 0.1273, + "step": 1253 + }, + { + "epoch": 0.07, + "learning_rate": 3.986837732705993e-07, + "loss": 0.1176, + "step": 1254 + }, + { + "epoch": 0.07, + "learning_rate": 3.9867989773296447e-07, + "loss": 0.2068, + "step": 1255 + }, + { + "epoch": 0.07, + "learning_rate": 3.986760165169814e-07, + "loss": 0.1309, + "step": 1256 + }, + { + "epoch": 0.07, + "learning_rate": 3.9867212962276107e-07, + "loss": 0.1589, + "step": 1257 + }, + { + "epoch": 0.07, + "learning_rate": 3.986682370504146e-07, + "loss": 0.1368, + "step": 1258 + }, + { + "epoch": 0.07, + "learning_rate": 3.9866433880005323e-07, + "loss": 0.1347, + "step": 1259 + }, + { + "epoch": 0.07, + "learning_rate": 3.9866043487178837e-07, + "loss": 0.181, + "step": 1260 + }, + { + "epoch": 0.07, + "learning_rate": 3.986565252657316e-07, + "loss": 0.1611, + "step": 1261 + }, + { + "epoch": 0.07, + "learning_rate": 3.9865260998199465e-07, + "loss": 0.1809, + "step": 1262 + }, + { + "epoch": 0.07, + "learning_rate": 3.986486890206894e-07, + "loss": 0.1302, + "step": 1263 + }, + { + "epoch": 0.07, + "learning_rate": 3.986447623819279e-07, + "loss": 0.1682, + "step": 1264 + }, + { + "epoch": 0.07, + "learning_rate": 3.986408300658224e-07, + "loss": 0.2017, + "step": 1265 + }, + { + "epoch": 0.07, + "learning_rate": 3.9863689207248535e-07, + "loss": 0.16, + "step": 1266 + }, + { + "epoch": 0.07, + "learning_rate": 3.986329484020293e-07, + "loss": 0.1285, + "step": 1267 + }, + { + "epoch": 0.07, + "learning_rate": 3.9862899905456675e-07, + "loss": 0.1817, + "step": 1268 + }, + { + "epoch": 0.07, + "learning_rate": 3.9862504403021084e-07, + "loss": 0.1295, + "step": 1269 + }, + { + "epoch": 0.07, + "learning_rate": 3.9862108332907443e-07, + "loss": 0.1528, + "step": 1270 + }, + { + "epoch": 0.07, + "learning_rate": 3.986171169512708e-07, + "loss": 0.1416, + "step": 1271 + }, + { + "epoch": 0.07, + "learning_rate": 3.9861314489691336e-07, + "loss": 0.1612, + "step": 1272 + }, + { + "epoch": 0.07, + "learning_rate": 3.986091671661155e-07, + "loss": 0.1529, + "step": 1273 + }, + { + "epoch": 0.07, + "learning_rate": 3.98605183758991e-07, + "loss": 0.1678, + "step": 1274 + }, + { + "epoch": 0.07, + "learning_rate": 3.9860119467565364e-07, + "loss": 0.1554, + "step": 1275 + }, + { + "epoch": 0.07, + "learning_rate": 3.985971999162175e-07, + "loss": 0.163, + "step": 1276 + }, + { + "epoch": 0.07, + "learning_rate": 3.9859319948079674e-07, + "loss": 0.2005, + "step": 1277 + }, + { + "epoch": 0.07, + "learning_rate": 3.985891933695057e-07, + "loss": 0.166, + "step": 1278 + }, + { + "epoch": 0.07, + "learning_rate": 3.985851815824587e-07, + "loss": 0.1549, + "step": 1279 + }, + { + "epoch": 0.07, + "learning_rate": 3.985811641197707e-07, + "loss": 0.1778, + "step": 1280 + }, + { + "epoch": 0.07, + "learning_rate": 3.9857714098155635e-07, + "loss": 0.2084, + "step": 1281 + }, + { + "epoch": 0.07, + "learning_rate": 3.985731121679306e-07, + "loss": 0.1337, + "step": 1282 + }, + { + "epoch": 0.07, + "learning_rate": 3.9856907767900865e-07, + "loss": 0.1354, + "step": 1283 + }, + { + "epoch": 0.07, + "learning_rate": 3.985650375149058e-07, + "loss": 0.1402, + "step": 1284 + }, + { + "epoch": 0.07, + "learning_rate": 3.9856099167573757e-07, + "loss": 0.1308, + "step": 1285 + }, + { + "epoch": 0.07, + "learning_rate": 3.985569401616195e-07, + "loss": 0.1337, + "step": 1286 + }, + { + "epoch": 0.07, + "learning_rate": 3.9855288297266747e-07, + "loss": 0.1645, + "step": 1287 + }, + { + "epoch": 0.07, + "learning_rate": 3.985488201089974e-07, + "loss": 0.1359, + "step": 1288 + }, + { + "epoch": 0.07, + "learning_rate": 3.9854475157072534e-07, + "loss": 0.1841, + "step": 1289 + }, + { + "epoch": 0.07, + "learning_rate": 3.985406773579677e-07, + "loss": 0.1614, + "step": 1290 + }, + { + "epoch": 0.07, + "learning_rate": 3.9853659747084085e-07, + "loss": 0.1451, + "step": 1291 + }, + { + "epoch": 0.07, + "learning_rate": 3.9853251190946135e-07, + "loss": 0.179, + "step": 1292 + }, + { + "epoch": 0.07, + "learning_rate": 3.9852842067394605e-07, + "loss": 0.1332, + "step": 1293 + }, + { + "epoch": 0.07, + "learning_rate": 3.9852432376441183e-07, + "loss": 0.1583, + "step": 1294 + }, + { + "epoch": 0.07, + "learning_rate": 3.9852022118097585e-07, + "loss": 0.1576, + "step": 1295 + }, + { + "epoch": 0.07, + "learning_rate": 3.9851611292375526e-07, + "loss": 0.1639, + "step": 1296 + }, + { + "epoch": 0.07, + "learning_rate": 3.985119989928676e-07, + "loss": 0.1538, + "step": 1297 + }, + { + "epoch": 0.07, + "learning_rate": 3.9850787938843027e-07, + "loss": 0.1591, + "step": 1298 + }, + { + "epoch": 0.07, + "learning_rate": 3.985037541105612e-07, + "loss": 0.1403, + "step": 1299 + }, + { + "epoch": 0.07, + "learning_rate": 3.9849962315937813e-07, + "loss": 0.1677, + "step": 1300 + }, + { + "epoch": 0.07, + "learning_rate": 3.9849548653499926e-07, + "loss": 0.1412, + "step": 1301 + }, + { + "epoch": 0.07, + "learning_rate": 3.9849134423754274e-07, + "loss": 0.147, + "step": 1302 + }, + { + "epoch": 0.07, + "learning_rate": 3.98487196267127e-07, + "loss": 0.1194, + "step": 1303 + }, + { + "epoch": 0.07, + "learning_rate": 3.9848304262387055e-07, + "loss": 0.1299, + "step": 1304 + }, + { + "epoch": 0.07, + "learning_rate": 3.984788833078921e-07, + "loss": 0.1387, + "step": 1305 + }, + { + "epoch": 0.07, + "learning_rate": 3.9847471831931057e-07, + "loss": 0.1615, + "step": 1306 + }, + { + "epoch": 0.07, + "learning_rate": 3.98470547658245e-07, + "loss": 0.1348, + "step": 1307 + }, + { + "epoch": 0.07, + "learning_rate": 3.984663713248145e-07, + "loss": 0.2092, + "step": 1308 + }, + { + "epoch": 0.07, + "learning_rate": 3.9846218931913854e-07, + "loss": 0.153, + "step": 1309 + }, + { + "epoch": 0.07, + "learning_rate": 3.9845800164133654e-07, + "loss": 0.1301, + "step": 1310 + }, + { + "epoch": 0.07, + "learning_rate": 3.984538082915283e-07, + "loss": 0.1541, + "step": 1311 + }, + { + "epoch": 0.07, + "learning_rate": 3.9844960926983355e-07, + "loss": 0.1797, + "step": 1312 + }, + { + "epoch": 0.07, + "learning_rate": 3.984454045763724e-07, + "loss": 0.1679, + "step": 1313 + }, + { + "epoch": 0.07, + "learning_rate": 3.9844119421126495e-07, + "loss": 0.1594, + "step": 1314 + }, + { + "epoch": 0.07, + "learning_rate": 3.984369781746316e-07, + "loss": 0.1499, + "step": 1315 + }, + { + "epoch": 0.07, + "learning_rate": 3.9843275646659275e-07, + "loss": 0.1616, + "step": 1316 + }, + { + "epoch": 0.07, + "learning_rate": 3.984285290872692e-07, + "loss": 0.1334, + "step": 1317 + }, + { + "epoch": 0.07, + "learning_rate": 3.984242960367816e-07, + "loss": 0.1313, + "step": 1318 + }, + { + "epoch": 0.07, + "learning_rate": 3.984200573152511e-07, + "loss": 0.1427, + "step": 1319 + }, + { + "epoch": 0.07, + "learning_rate": 3.984158129227987e-07, + "loss": 0.1548, + "step": 1320 + }, + { + "epoch": 0.07, + "learning_rate": 3.9841156285954583e-07, + "loss": 0.1299, + "step": 1321 + }, + { + "epoch": 0.07, + "learning_rate": 3.984073071256139e-07, + "loss": 0.1345, + "step": 1322 + }, + { + "epoch": 0.07, + "learning_rate": 3.9840304572112447e-07, + "loss": 0.166, + "step": 1323 + }, + { + "epoch": 0.07, + "learning_rate": 3.9839877864619946e-07, + "loss": 0.146, + "step": 1324 + }, + { + "epoch": 0.07, + "learning_rate": 3.983945059009607e-07, + "loss": 0.163, + "step": 1325 + }, + { + "epoch": 0.07, + "learning_rate": 3.9839022748553044e-07, + "loss": 0.1668, + "step": 1326 + }, + { + "epoch": 0.07, + "learning_rate": 3.983859434000309e-07, + "loss": 0.1406, + "step": 1327 + }, + { + "epoch": 0.07, + "learning_rate": 3.983816536445845e-07, + "loss": 0.1537, + "step": 1328 + }, + { + "epoch": 0.07, + "learning_rate": 3.983773582193138e-07, + "loss": 0.1634, + "step": 1329 + }, + { + "epoch": 0.07, + "learning_rate": 3.983730571243417e-07, + "loss": 0.1656, + "step": 1330 + }, + { + "epoch": 0.07, + "learning_rate": 3.9836875035979096e-07, + "loss": 0.1976, + "step": 1331 + }, + { + "epoch": 0.07, + "learning_rate": 3.983644379257848e-07, + "loss": 0.1525, + "step": 1332 + }, + { + "epoch": 0.07, + "learning_rate": 3.983601198224464e-07, + "loss": 0.1408, + "step": 1333 + }, + { + "epoch": 0.07, + "learning_rate": 3.983557960498993e-07, + "loss": 0.15, + "step": 1334 + }, + { + "epoch": 0.07, + "learning_rate": 3.9835146660826695e-07, + "loss": 0.1593, + "step": 1335 + }, + { + "epoch": 0.07, + "learning_rate": 3.9834713149767303e-07, + "loss": 0.1394, + "step": 1336 + }, + { + "epoch": 0.07, + "learning_rate": 3.983427907182416e-07, + "loss": 0.181, + "step": 1337 + }, + { + "epoch": 0.07, + "learning_rate": 3.9833844427009654e-07, + "loss": 0.1262, + "step": 1338 + }, + { + "epoch": 0.07, + "learning_rate": 3.9833409215336226e-07, + "loss": 0.1596, + "step": 1339 + }, + { + "epoch": 0.07, + "learning_rate": 3.9832973436816306e-07, + "loss": 0.1344, + "step": 1340 + }, + { + "epoch": 0.07, + "learning_rate": 3.9832537091462346e-07, + "loss": 0.1591, + "step": 1341 + }, + { + "epoch": 0.07, + "learning_rate": 3.983210017928683e-07, + "loss": 0.1634, + "step": 1342 + }, + { + "epoch": 0.07, + "learning_rate": 3.983166270030222e-07, + "loss": 0.1557, + "step": 1343 + }, + { + "epoch": 0.07, + "learning_rate": 3.983122465452104e-07, + "loss": 0.1313, + "step": 1344 + }, + { + "epoch": 0.07, + "learning_rate": 3.98307860419558e-07, + "loss": 0.2033, + "step": 1345 + }, + { + "epoch": 0.07, + "learning_rate": 3.983034686261905e-07, + "loss": 0.1516, + "step": 1346 + }, + { + "epoch": 0.07, + "learning_rate": 3.9829907116523324e-07, + "loss": 0.1614, + "step": 1347 + }, + { + "epoch": 0.07, + "learning_rate": 3.98294668036812e-07, + "loss": 0.1521, + "step": 1348 + }, + { + "epoch": 0.07, + "learning_rate": 3.982902592410526e-07, + "loss": 0.1487, + "step": 1349 + }, + { + "epoch": 0.07, + "learning_rate": 3.98285844778081e-07, + "loss": 0.1651, + "step": 1350 + }, + { + "epoch": 0.07, + "learning_rate": 3.982814246480235e-07, + "loss": 0.1686, + "step": 1351 + }, + { + "epoch": 0.07, + "learning_rate": 3.982769988510062e-07, + "loss": 0.1533, + "step": 1352 + }, + { + "epoch": 0.07, + "learning_rate": 3.982725673871558e-07, + "loss": 0.158, + "step": 1353 + }, + { + "epoch": 0.07, + "learning_rate": 3.982681302565989e-07, + "loss": 0.1367, + "step": 1354 + }, + { + "epoch": 0.07, + "learning_rate": 3.9826368745946233e-07, + "loss": 0.1572, + "step": 1355 + }, + { + "epoch": 0.07, + "learning_rate": 3.9825923899587295e-07, + "loss": 0.1955, + "step": 1356 + }, + { + "epoch": 0.07, + "learning_rate": 3.98254784865958e-07, + "loss": 0.1671, + "step": 1357 + }, + { + "epoch": 0.07, + "learning_rate": 3.982503250698448e-07, + "loss": 0.1355, + "step": 1358 + }, + { + "epoch": 0.07, + "learning_rate": 3.9824585960766077e-07, + "loss": 0.1434, + "step": 1359 + }, + { + "epoch": 0.07, + "learning_rate": 3.9824138847953357e-07, + "loss": 0.1565, + "step": 1360 + }, + { + "epoch": 0.07, + "learning_rate": 3.982369116855909e-07, + "loss": 0.1798, + "step": 1361 + }, + { + "epoch": 0.07, + "learning_rate": 3.9823242922596087e-07, + "loss": 0.1753, + "step": 1362 + }, + { + "epoch": 0.07, + "learning_rate": 3.9822794110077133e-07, + "loss": 0.1449, + "step": 1363 + }, + { + "epoch": 0.07, + "learning_rate": 3.982234473101508e-07, + "loss": 0.1518, + "step": 1364 + }, + { + "epoch": 0.07, + "learning_rate": 3.9821894785422767e-07, + "loss": 0.1073, + "step": 1365 + }, + { + "epoch": 0.07, + "learning_rate": 3.9821444273313045e-07, + "loss": 0.1802, + "step": 1366 + }, + { + "epoch": 0.07, + "learning_rate": 3.9820993194698784e-07, + "loss": 0.1485, + "step": 1367 + }, + { + "epoch": 0.07, + "learning_rate": 3.98205415495929e-07, + "loss": 0.1492, + "step": 1368 + }, + { + "epoch": 0.07, + "learning_rate": 3.982008933800828e-07, + "loss": 0.1428, + "step": 1369 + }, + { + "epoch": 0.07, + "learning_rate": 3.9819636559957857e-07, + "loss": 0.1897, + "step": 1370 + }, + { + "epoch": 0.07, + "learning_rate": 3.981918321545457e-07, + "loss": 0.1254, + "step": 1371 + }, + { + "epoch": 0.07, + "learning_rate": 3.9818729304511374e-07, + "loss": 0.1384, + "step": 1372 + }, + { + "epoch": 0.07, + "learning_rate": 3.9818274827141244e-07, + "loss": 0.1374, + "step": 1373 + }, + { + "epoch": 0.07, + "learning_rate": 3.981781978335717e-07, + "loss": 0.1655, + "step": 1374 + }, + { + "epoch": 0.07, + "learning_rate": 3.981736417317216e-07, + "loss": 0.1555, + "step": 1375 + }, + { + "epoch": 0.07, + "learning_rate": 3.9816907996599223e-07, + "loss": 0.1294, + "step": 1376 + }, + { + "epoch": 0.07, + "learning_rate": 3.9816451253651413e-07, + "loss": 0.1526, + "step": 1377 + }, + { + "epoch": 0.07, + "learning_rate": 3.9815993944341773e-07, + "loss": 0.1294, + "step": 1378 + }, + { + "epoch": 0.07, + "learning_rate": 3.9815536068683373e-07, + "loss": 0.1403, + "step": 1379 + }, + { + "epoch": 0.07, + "learning_rate": 3.981507762668931e-07, + "loss": 0.1443, + "step": 1380 + }, + { + "epoch": 0.07, + "learning_rate": 3.981461861837268e-07, + "loss": 0.1357, + "step": 1381 + }, + { + "epoch": 0.07, + "learning_rate": 3.9814159043746593e-07, + "loss": 0.1185, + "step": 1382 + }, + { + "epoch": 0.07, + "learning_rate": 3.981369890282419e-07, + "loss": 0.1494, + "step": 1383 + }, + { + "epoch": 0.07, + "learning_rate": 3.981323819561863e-07, + "loss": 0.1326, + "step": 1384 + }, + { + "epoch": 0.07, + "learning_rate": 3.9812776922143076e-07, + "loss": 0.152, + "step": 1385 + }, + { + "epoch": 0.07, + "learning_rate": 3.98123150824107e-07, + "loss": 0.1528, + "step": 1386 + }, + { + "epoch": 0.07, + "learning_rate": 3.9811852676434715e-07, + "loss": 0.1182, + "step": 1387 + }, + { + "epoch": 0.07, + "learning_rate": 3.9811389704228337e-07, + "loss": 0.1557, + "step": 1388 + }, + { + "epoch": 0.07, + "learning_rate": 3.9810926165804786e-07, + "loss": 0.1203, + "step": 1389 + }, + { + "epoch": 0.07, + "learning_rate": 3.9810462061177323e-07, + "loss": 0.173, + "step": 1390 + }, + { + "epoch": 0.07, + "learning_rate": 3.98099973903592e-07, + "loss": 0.1434, + "step": 1391 + }, + { + "epoch": 0.07, + "learning_rate": 3.9809532153363713e-07, + "loss": 0.1641, + "step": 1392 + }, + { + "epoch": 0.07, + "learning_rate": 3.980906635020414e-07, + "loss": 0.207, + "step": 1393 + }, + { + "epoch": 0.07, + "learning_rate": 3.9808599980893815e-07, + "loss": 0.1686, + "step": 1394 + }, + { + "epoch": 0.07, + "learning_rate": 3.980813304544605e-07, + "loss": 0.1603, + "step": 1395 + }, + { + "epoch": 0.07, + "learning_rate": 3.980766554387419e-07, + "loss": 0.1406, + "step": 1396 + }, + { + "epoch": 0.07, + "learning_rate": 3.9807197476191607e-07, + "loss": 0.196, + "step": 1397 + }, + { + "epoch": 0.07, + "learning_rate": 3.9806728842411673e-07, + "loss": 0.119, + "step": 1398 + }, + { + "epoch": 0.07, + "learning_rate": 3.980625964254779e-07, + "loss": 0.1969, + "step": 1399 + }, + { + "epoch": 0.07, + "learning_rate": 3.980578987661335e-07, + "loss": 0.173, + "step": 1400 + }, + { + "epoch": 0.07, + "learning_rate": 3.9805319544621795e-07, + "loss": 0.124, + "step": 1401 + }, + { + "epoch": 0.07, + "learning_rate": 3.9804848646586564e-07, + "loss": 0.1888, + "step": 1402 + }, + { + "epoch": 0.07, + "learning_rate": 3.9804377182521105e-07, + "loss": 0.1245, + "step": 1403 + }, + { + "epoch": 0.07, + "learning_rate": 3.980390515243891e-07, + "loss": 0.1396, + "step": 1404 + }, + { + "epoch": 0.07, + "learning_rate": 3.980343255635345e-07, + "loss": 0.1485, + "step": 1405 + }, + { + "epoch": 0.07, + "learning_rate": 3.980295939427825e-07, + "loss": 0.1582, + "step": 1406 + }, + { + "epoch": 0.07, + "learning_rate": 3.980248566622682e-07, + "loss": 0.1427, + "step": 1407 + }, + { + "epoch": 0.07, + "learning_rate": 3.9802011372212714e-07, + "loss": 0.1477, + "step": 1408 + }, + { + "epoch": 0.07, + "learning_rate": 3.9801536512249476e-07, + "loss": 0.1771, + "step": 1409 + }, + { + "epoch": 0.07, + "learning_rate": 3.980106108635068e-07, + "loss": 0.1466, + "step": 1410 + }, + { + "epoch": 0.07, + "learning_rate": 3.9800585094529913e-07, + "loss": 0.1572, + "step": 1411 + }, + { + "epoch": 0.07, + "learning_rate": 3.9800108536800777e-07, + "loss": 0.166, + "step": 1412 + }, + { + "epoch": 0.07, + "learning_rate": 3.97996314131769e-07, + "loss": 0.1573, + "step": 1413 + }, + { + "epoch": 0.07, + "learning_rate": 3.9799153723671913e-07, + "loss": 0.1534, + "step": 1414 + }, + { + "epoch": 0.07, + "learning_rate": 3.979867546829947e-07, + "loss": 0.1546, + "step": 1415 + }, + { + "epoch": 0.07, + "learning_rate": 3.9798196647073237e-07, + "loss": 0.1758, + "step": 1416 + }, + { + "epoch": 0.07, + "learning_rate": 3.97977172600069e-07, + "loss": 0.147, + "step": 1417 + }, + { + "epoch": 0.07, + "learning_rate": 3.9797237307114165e-07, + "loss": 0.1547, + "step": 1418 + }, + { + "epoch": 0.07, + "learning_rate": 3.9796756788408743e-07, + "loss": 0.1735, + "step": 1419 + }, + { + "epoch": 0.07, + "learning_rate": 3.9796275703904366e-07, + "loss": 0.1728, + "step": 1420 + }, + { + "epoch": 0.07, + "learning_rate": 3.9795794053614797e-07, + "loss": 0.1821, + "step": 1421 + }, + { + "epoch": 0.07, + "learning_rate": 3.9795311837553785e-07, + "loss": 0.1591, + "step": 1422 + }, + { + "epoch": 0.07, + "learning_rate": 3.979482905573512e-07, + "loss": 0.152, + "step": 1423 + }, + { + "epoch": 0.07, + "learning_rate": 3.9794345708172594e-07, + "loss": 0.1506, + "step": 1424 + }, + { + "epoch": 0.07, + "learning_rate": 3.9793861794880034e-07, + "loss": 0.1707, + "step": 1425 + }, + { + "epoch": 0.07, + "learning_rate": 3.979337731587126e-07, + "loss": 0.1594, + "step": 1426 + }, + { + "epoch": 0.07, + "learning_rate": 3.979289227116012e-07, + "loss": 0.1401, + "step": 1427 + }, + { + "epoch": 0.07, + "learning_rate": 3.979240666076048e-07, + "loss": 0.1592, + "step": 1428 + }, + { + "epoch": 0.07, + "learning_rate": 3.9791920484686214e-07, + "loss": 0.1419, + "step": 1429 + }, + { + "epoch": 0.07, + "learning_rate": 3.9791433742951225e-07, + "loss": 0.1125, + "step": 1430 + }, + { + "epoch": 0.07, + "learning_rate": 3.979094643556942e-07, + "loss": 0.1911, + "step": 1431 + }, + { + "epoch": 0.07, + "learning_rate": 3.9790458562554716e-07, + "loss": 0.1481, + "step": 1432 + }, + { + "epoch": 0.07, + "learning_rate": 3.9789970123921074e-07, + "loss": 0.1525, + "step": 1433 + }, + { + "epoch": 0.07, + "learning_rate": 3.978948111968244e-07, + "loss": 0.1557, + "step": 1434 + }, + { + "epoch": 0.07, + "learning_rate": 3.9788991549852805e-07, + "loss": 0.1725, + "step": 1435 + }, + { + "epoch": 0.07, + "learning_rate": 3.978850141444615e-07, + "loss": 0.1312, + "step": 1436 + }, + { + "epoch": 0.08, + "learning_rate": 3.978801071347648e-07, + "loss": 0.1641, + "step": 1437 + }, + { + "epoch": 0.08, + "learning_rate": 3.9787519446957833e-07, + "loss": 0.152, + "step": 1438 + }, + { + "epoch": 0.08, + "learning_rate": 3.978702761490424e-07, + "loss": 0.1845, + "step": 1439 + }, + { + "epoch": 0.08, + "learning_rate": 3.978653521732975e-07, + "loss": 0.1581, + "step": 1440 + }, + { + "epoch": 0.08, + "learning_rate": 3.9786042254248457e-07, + "loss": 0.1533, + "step": 1441 + }, + { + "epoch": 0.08, + "learning_rate": 3.978554872567443e-07, + "loss": 0.1244, + "step": 1442 + }, + { + "epoch": 0.08, + "learning_rate": 3.9785054631621783e-07, + "loss": 0.1839, + "step": 1443 + }, + { + "epoch": 0.08, + "learning_rate": 3.9784559972104643e-07, + "loss": 0.1783, + "step": 1444 + }, + { + "epoch": 0.08, + "learning_rate": 3.9784064747137133e-07, + "loss": 0.1768, + "step": 1445 + }, + { + "epoch": 0.08, + "learning_rate": 3.978356895673342e-07, + "loss": 0.1449, + "step": 1446 + }, + { + "epoch": 0.08, + "learning_rate": 3.978307260090767e-07, + "loss": 0.1519, + "step": 1447 + }, + { + "epoch": 0.08, + "learning_rate": 3.9782575679674065e-07, + "loss": 0.1967, + "step": 1448 + }, + { + "epoch": 0.08, + "learning_rate": 3.978207819304681e-07, + "loss": 0.1662, + "step": 1449 + }, + { + "epoch": 0.08, + "learning_rate": 3.9781580141040124e-07, + "loss": 0.1757, + "step": 1450 + }, + { + "epoch": 0.08, + "learning_rate": 3.978108152366824e-07, + "loss": 0.1475, + "step": 1451 + }, + { + "epoch": 0.08, + "learning_rate": 3.9780582340945417e-07, + "loss": 0.1352, + "step": 1452 + }, + { + "epoch": 0.08, + "learning_rate": 3.9780082592885914e-07, + "loss": 0.1274, + "step": 1453 + }, + { + "epoch": 0.08, + "learning_rate": 3.9779582279504007e-07, + "loss": 0.1676, + "step": 1454 + }, + { + "epoch": 0.08, + "learning_rate": 3.977908140081401e-07, + "loss": 0.1501, + "step": 1455 + }, + { + "epoch": 0.08, + "learning_rate": 3.977857995683022e-07, + "loss": 0.1437, + "step": 1456 + }, + { + "epoch": 0.08, + "learning_rate": 3.977807794756699e-07, + "loss": 0.167, + "step": 1457 + }, + { + "epoch": 0.08, + "learning_rate": 3.9777575373038656e-07, + "loss": 0.16, + "step": 1458 + }, + { + "epoch": 0.08, + "learning_rate": 3.977707223325958e-07, + "loss": 0.127, + "step": 1459 + }, + { + "epoch": 0.08, + "learning_rate": 3.977656852824414e-07, + "loss": 0.1659, + "step": 1460 + }, + { + "epoch": 0.08, + "learning_rate": 3.9776064258006747e-07, + "loss": 0.1551, + "step": 1461 + }, + { + "epoch": 0.08, + "learning_rate": 3.9775559422561796e-07, + "loss": 0.1433, + "step": 1462 + }, + { + "epoch": 0.08, + "learning_rate": 3.977505402192373e-07, + "loss": 0.1294, + "step": 1463 + }, + { + "epoch": 0.08, + "learning_rate": 3.977454805610698e-07, + "loss": 0.144, + "step": 1464 + }, + { + "epoch": 0.08, + "learning_rate": 3.9774041525126013e-07, + "loss": 0.1209, + "step": 1465 + }, + { + "epoch": 0.08, + "learning_rate": 3.977353442899531e-07, + "loss": 0.1456, + "step": 1466 + }, + { + "epoch": 0.08, + "learning_rate": 3.9773026767729354e-07, + "loss": 0.1387, + "step": 1467 + }, + { + "epoch": 0.08, + "learning_rate": 3.977251854134266e-07, + "loss": 0.1446, + "step": 1468 + }, + { + "epoch": 0.08, + "learning_rate": 3.977200974984976e-07, + "loss": 0.1527, + "step": 1469 + }, + { + "epoch": 0.08, + "learning_rate": 3.977150039326518e-07, + "loss": 0.1325, + "step": 1470 + }, + { + "epoch": 0.08, + "learning_rate": 3.977099047160349e-07, + "loss": 0.1523, + "step": 1471 + }, + { + "epoch": 0.08, + "learning_rate": 3.9770479984879267e-07, + "loss": 0.1151, + "step": 1472 + }, + { + "epoch": 0.08, + "learning_rate": 3.976996893310709e-07, + "loss": 0.1347, + "step": 1473 + }, + { + "epoch": 0.08, + "learning_rate": 3.976945731630156e-07, + "loss": 0.1693, + "step": 1474 + }, + { + "epoch": 0.08, + "learning_rate": 3.9768945134477314e-07, + "loss": 0.1293, + "step": 1475 + }, + { + "epoch": 0.08, + "learning_rate": 3.9768432387648985e-07, + "loss": 0.1252, + "step": 1476 + }, + { + "epoch": 0.08, + "learning_rate": 3.9767919075831225e-07, + "loss": 0.1108, + "step": 1477 + }, + { + "epoch": 0.08, + "learning_rate": 3.976740519903871e-07, + "loss": 0.1476, + "step": 1478 + }, + { + "epoch": 0.08, + "learning_rate": 3.976689075728612e-07, + "loss": 0.1632, + "step": 1479 + }, + { + "epoch": 0.08, + "learning_rate": 3.9766375750588164e-07, + "loss": 0.1477, + "step": 1480 + }, + { + "epoch": 0.08, + "learning_rate": 3.9765860178959557e-07, + "loss": 0.1389, + "step": 1481 + }, + { + "epoch": 0.08, + "learning_rate": 3.976534404241504e-07, + "loss": 0.1674, + "step": 1482 + }, + { + "epoch": 0.08, + "learning_rate": 3.9764827340969353e-07, + "loss": 0.1596, + "step": 1483 + }, + { + "epoch": 0.08, + "learning_rate": 3.976431007463727e-07, + "loss": 0.1527, + "step": 1484 + }, + { + "epoch": 0.08, + "learning_rate": 3.976379224343358e-07, + "loss": 0.1398, + "step": 1485 + }, + { + "epoch": 0.08, + "learning_rate": 3.976327384737308e-07, + "loss": 0.1497, + "step": 1486 + }, + { + "epoch": 0.08, + "learning_rate": 3.976275488647058e-07, + "loss": 0.1376, + "step": 1487 + }, + { + "epoch": 0.08, + "learning_rate": 3.976223536074092e-07, + "loss": 0.1562, + "step": 1488 + }, + { + "epoch": 0.08, + "learning_rate": 3.9761715270198944e-07, + "loss": 0.1268, + "step": 1489 + }, + { + "epoch": 0.08, + "learning_rate": 3.9761194614859514e-07, + "loss": 0.1294, + "step": 1490 + }, + { + "epoch": 0.08, + "learning_rate": 3.976067339473751e-07, + "loss": 0.1478, + "step": 1491 + }, + { + "epoch": 0.08, + "learning_rate": 3.976015160984784e-07, + "loss": 0.2009, + "step": 1492 + }, + { + "epoch": 0.08, + "learning_rate": 3.97596292602054e-07, + "loss": 0.1356, + "step": 1493 + }, + { + "epoch": 0.08, + "learning_rate": 3.9759106345825133e-07, + "loss": 0.1296, + "step": 1494 + }, + { + "epoch": 0.08, + "learning_rate": 3.9758582866721977e-07, + "loss": 0.1737, + "step": 1495 + }, + { + "epoch": 0.08, + "learning_rate": 3.97580588229109e-07, + "loss": 0.151, + "step": 1496 + }, + { + "epoch": 0.08, + "learning_rate": 3.975753421440687e-07, + "loss": 0.1523, + "step": 1497 + }, + { + "epoch": 0.08, + "learning_rate": 3.9757009041224886e-07, + "loss": 0.1694, + "step": 1498 + }, + { + "epoch": 0.08, + "learning_rate": 3.9756483303379956e-07, + "loss": 0.1354, + "step": 1499 + }, + { + "epoch": 0.08, + "learning_rate": 3.975595700088711e-07, + "loss": 0.1465, + "step": 1500 + }, + { + "epoch": 0.08, + "learning_rate": 3.9755430133761383e-07, + "loss": 0.1501, + "step": 1501 + }, + { + "epoch": 0.08, + "learning_rate": 3.975490270201783e-07, + "loss": 0.1547, + "step": 1502 + }, + { + "epoch": 0.08, + "learning_rate": 3.975437470567154e-07, + "loss": 0.1236, + "step": 1503 + }, + { + "epoch": 0.08, + "learning_rate": 3.97538461447376e-07, + "loss": 0.1537, + "step": 1504 + }, + { + "epoch": 0.08, + "learning_rate": 3.9753317019231105e-07, + "loss": 0.1979, + "step": 1505 + }, + { + "epoch": 0.08, + "learning_rate": 3.9752787329167177e-07, + "loss": 0.1345, + "step": 1506 + }, + { + "epoch": 0.08, + "learning_rate": 3.9752257074560973e-07, + "loss": 0.1621, + "step": 1507 + }, + { + "epoch": 0.08, + "learning_rate": 3.975172625542763e-07, + "loss": 0.1633, + "step": 1508 + }, + { + "epoch": 0.08, + "learning_rate": 3.975119487178233e-07, + "loss": 0.1453, + "step": 1509 + }, + { + "epoch": 0.08, + "learning_rate": 3.975066292364025e-07, + "loss": 0.131, + "step": 1510 + }, + { + "epoch": 0.08, + "learning_rate": 3.9750130411016603e-07, + "loss": 0.1277, + "step": 1511 + }, + { + "epoch": 0.08, + "learning_rate": 3.974959733392661e-07, + "loss": 0.1744, + "step": 1512 + }, + { + "epoch": 0.08, + "learning_rate": 3.974906369238549e-07, + "loss": 0.1652, + "step": 1513 + }, + { + "epoch": 0.08, + "learning_rate": 3.974852948640851e-07, + "loss": 0.1569, + "step": 1514 + }, + { + "epoch": 0.08, + "learning_rate": 3.974799471601094e-07, + "loss": 0.1568, + "step": 1515 + }, + { + "epoch": 0.08, + "learning_rate": 3.974745938120805e-07, + "loss": 0.1486, + "step": 1516 + }, + { + "epoch": 0.08, + "learning_rate": 3.974692348201515e-07, + "loss": 0.1906, + "step": 1517 + }, + { + "epoch": 0.08, + "learning_rate": 3.9746387018447554e-07, + "loss": 0.1969, + "step": 1518 + }, + { + "epoch": 0.08, + "learning_rate": 3.97458499905206e-07, + "loss": 0.1277, + "step": 1519 + }, + { + "epoch": 0.08, + "learning_rate": 3.9745312398249623e-07, + "loss": 0.125, + "step": 1520 + }, + { + "epoch": 0.08, + "learning_rate": 3.9744774241649993e-07, + "loss": 0.1305, + "step": 1521 + }, + { + "epoch": 0.08, + "learning_rate": 3.97442355207371e-07, + "loss": 0.1849, + "step": 1522 + }, + { + "epoch": 0.08, + "learning_rate": 3.974369623552633e-07, + "loss": 0.1545, + "step": 1523 + }, + { + "epoch": 0.08, + "learning_rate": 3.974315638603309e-07, + "loss": 0.1709, + "step": 1524 + }, + { + "epoch": 0.08, + "learning_rate": 3.974261597227284e-07, + "loss": 0.1613, + "step": 1525 + }, + { + "epoch": 0.08, + "learning_rate": 3.974207499426099e-07, + "loss": 0.2016, + "step": 1526 + }, + { + "epoch": 0.08, + "learning_rate": 3.974153345201302e-07, + "loss": 0.1189, + "step": 1527 + }, + { + "epoch": 0.08, + "learning_rate": 3.97409913455444e-07, + "loss": 0.1632, + "step": 1528 + }, + { + "epoch": 0.08, + "learning_rate": 3.9740448674870626e-07, + "loss": 0.175, + "step": 1529 + }, + { + "epoch": 0.08, + "learning_rate": 3.9739905440007214e-07, + "loss": 0.1396, + "step": 1530 + }, + { + "epoch": 0.08, + "learning_rate": 3.973936164096968e-07, + "loss": 0.1417, + "step": 1531 + }, + { + "epoch": 0.08, + "learning_rate": 3.9738817277773567e-07, + "loss": 0.192, + "step": 1532 + }, + { + "epoch": 0.08, + "learning_rate": 3.973827235043444e-07, + "loss": 0.1781, + "step": 1533 + }, + { + "epoch": 0.08, + "learning_rate": 3.9737726858967873e-07, + "loss": 0.145, + "step": 1534 + }, + { + "epoch": 0.08, + "learning_rate": 3.973718080338945e-07, + "loss": 0.1705, + "step": 1535 + }, + { + "epoch": 0.08, + "learning_rate": 3.973663418371478e-07, + "loss": 0.1629, + "step": 1536 + }, + { + "epoch": 0.08, + "learning_rate": 3.9736086999959484e-07, + "loss": 0.1576, + "step": 1537 + }, + { + "epoch": 0.08, + "learning_rate": 3.9735539252139205e-07, + "loss": 0.1397, + "step": 1538 + }, + { + "epoch": 0.08, + "learning_rate": 3.973499094026959e-07, + "loss": 0.1367, + "step": 1539 + }, + { + "epoch": 0.08, + "learning_rate": 3.973444206436632e-07, + "loss": 0.1371, + "step": 1540 + }, + { + "epoch": 0.08, + "learning_rate": 3.973389262444508e-07, + "loss": 0.1719, + "step": 1541 + }, + { + "epoch": 0.08, + "learning_rate": 3.973334262052157e-07, + "loss": 0.1306, + "step": 1542 + }, + { + "epoch": 0.08, + "learning_rate": 3.973279205261151e-07, + "loss": 0.1935, + "step": 1543 + }, + { + "epoch": 0.08, + "learning_rate": 3.973224092073063e-07, + "loss": 0.1392, + "step": 1544 + }, + { + "epoch": 0.08, + "learning_rate": 3.973168922489469e-07, + "loss": 0.1452, + "step": 1545 + }, + { + "epoch": 0.08, + "learning_rate": 3.973113696511946e-07, + "loss": 0.1748, + "step": 1546 + }, + { + "epoch": 0.08, + "learning_rate": 3.9730584141420716e-07, + "loss": 0.1504, + "step": 1547 + }, + { + "epoch": 0.08, + "learning_rate": 3.973003075381426e-07, + "loss": 0.1805, + "step": 1548 + }, + { + "epoch": 0.08, + "learning_rate": 3.9729476802315906e-07, + "loss": 0.153, + "step": 1549 + }, + { + "epoch": 0.08, + "learning_rate": 3.9728922286941486e-07, + "loss": 0.1396, + "step": 1550 + }, + { + "epoch": 0.08, + "learning_rate": 3.9728367207706856e-07, + "loss": 0.1702, + "step": 1551 + }, + { + "epoch": 0.08, + "learning_rate": 3.9727811564627874e-07, + "loss": 0.1463, + "step": 1552 + }, + { + "epoch": 0.08, + "learning_rate": 3.972725535772042e-07, + "loss": 0.138, + "step": 1553 + }, + { + "epoch": 0.08, + "learning_rate": 3.97266985870004e-07, + "loss": 0.1318, + "step": 1554 + }, + { + "epoch": 0.08, + "learning_rate": 3.9726141252483706e-07, + "loss": 0.1468, + "step": 1555 + }, + { + "epoch": 0.08, + "learning_rate": 3.972558335418629e-07, + "loss": 0.1247, + "step": 1556 + }, + { + "epoch": 0.08, + "learning_rate": 3.9725024892124086e-07, + "loss": 0.1581, + "step": 1557 + }, + { + "epoch": 0.08, + "learning_rate": 3.972446586631306e-07, + "loss": 0.1279, + "step": 1558 + }, + { + "epoch": 0.08, + "learning_rate": 3.972390627676917e-07, + "loss": 0.1782, + "step": 1559 + }, + { + "epoch": 0.08, + "learning_rate": 3.972334612350844e-07, + "loss": 0.1508, + "step": 1560 + }, + { + "epoch": 0.08, + "learning_rate": 3.9722785406546857e-07, + "loss": 0.1622, + "step": 1561 + }, + { + "epoch": 0.08, + "learning_rate": 3.972222412590045e-07, + "loss": 0.1462, + "step": 1562 + }, + { + "epoch": 0.08, + "learning_rate": 3.972166228158527e-07, + "loss": 0.1506, + "step": 1563 + }, + { + "epoch": 0.08, + "learning_rate": 3.9721099873617366e-07, + "loss": 0.1656, + "step": 1564 + }, + { + "epoch": 0.08, + "learning_rate": 3.972053690201282e-07, + "loss": 0.1288, + "step": 1565 + }, + { + "epoch": 0.08, + "learning_rate": 3.971997336678771e-07, + "loss": 0.1559, + "step": 1566 + }, + { + "epoch": 0.08, + "learning_rate": 3.9719409267958146e-07, + "loss": 0.1449, + "step": 1567 + }, + { + "epoch": 0.08, + "learning_rate": 3.971884460554026e-07, + "loss": 0.1549, + "step": 1568 + }, + { + "epoch": 0.08, + "learning_rate": 3.9718279379550177e-07, + "loss": 0.1537, + "step": 1569 + }, + { + "epoch": 0.08, + "learning_rate": 3.971771359000406e-07, + "loss": 0.1658, + "step": 1570 + }, + { + "epoch": 0.08, + "learning_rate": 3.9717147236918073e-07, + "loss": 0.1655, + "step": 1571 + }, + { + "epoch": 0.08, + "learning_rate": 3.9716580320308417e-07, + "loss": 0.1418, + "step": 1572 + }, + { + "epoch": 0.08, + "learning_rate": 3.9716012840191277e-07, + "loss": 0.1612, + "step": 1573 + }, + { + "epoch": 0.08, + "learning_rate": 3.9715444796582875e-07, + "loss": 0.1642, + "step": 1574 + }, + { + "epoch": 0.08, + "learning_rate": 3.9714876189499453e-07, + "loss": 0.1636, + "step": 1575 + }, + { + "epoch": 0.08, + "learning_rate": 3.971430701895726e-07, + "loss": 0.1408, + "step": 1576 + }, + { + "epoch": 0.08, + "learning_rate": 3.9713737284972563e-07, + "loss": 0.1683, + "step": 1577 + }, + { + "epoch": 0.08, + "learning_rate": 3.971316698756165e-07, + "loss": 0.154, + "step": 1578 + }, + { + "epoch": 0.08, + "learning_rate": 3.97125961267408e-07, + "loss": 0.1349, + "step": 1579 + }, + { + "epoch": 0.08, + "learning_rate": 3.971202470252635e-07, + "loss": 0.1925, + "step": 1580 + }, + { + "epoch": 0.08, + "learning_rate": 3.9711452714934625e-07, + "loss": 0.166, + "step": 1581 + }, + { + "epoch": 0.08, + "learning_rate": 3.971088016398197e-07, + "loss": 0.124, + "step": 1582 + }, + { + "epoch": 0.08, + "learning_rate": 3.9710307049684755e-07, + "loss": 0.1581, + "step": 1583 + }, + { + "epoch": 0.08, + "learning_rate": 3.9709733372059347e-07, + "loss": 0.111, + "step": 1584 + }, + { + "epoch": 0.08, + "learning_rate": 3.9709159131122163e-07, + "loss": 0.1238, + "step": 1585 + }, + { + "epoch": 0.08, + "learning_rate": 3.9708584326889593e-07, + "loss": 0.1319, + "step": 1586 + }, + { + "epoch": 0.08, + "learning_rate": 3.970800895937808e-07, + "loss": 0.1556, + "step": 1587 + }, + { + "epoch": 0.08, + "learning_rate": 3.9707433028604057e-07, + "loss": 0.1581, + "step": 1588 + }, + { + "epoch": 0.08, + "learning_rate": 3.9706856534584e-07, + "loss": 0.1739, + "step": 1589 + }, + { + "epoch": 0.08, + "learning_rate": 3.9706279477334363e-07, + "loss": 0.131, + "step": 1590 + }, + { + "epoch": 0.08, + "learning_rate": 3.970570185687166e-07, + "loss": 0.1668, + "step": 1591 + }, + { + "epoch": 0.08, + "learning_rate": 3.9705123673212383e-07, + "loss": 0.1472, + "step": 1592 + }, + { + "epoch": 0.08, + "learning_rate": 3.970454492637307e-07, + "loss": 0.1609, + "step": 1593 + }, + { + "epoch": 0.08, + "learning_rate": 3.9703965616370255e-07, + "loss": 0.1602, + "step": 1594 + }, + { + "epoch": 0.08, + "learning_rate": 3.97033857432205e-07, + "loss": 0.1457, + "step": 1595 + }, + { + "epoch": 0.08, + "learning_rate": 3.9702805306940374e-07, + "loss": 0.1346, + "step": 1596 + }, + { + "epoch": 0.08, + "learning_rate": 3.9702224307546463e-07, + "loss": 0.1511, + "step": 1597 + }, + { + "epoch": 0.08, + "learning_rate": 3.9701642745055377e-07, + "loss": 0.1398, + "step": 1598 + }, + { + "epoch": 0.08, + "learning_rate": 3.9701060619483733e-07, + "loss": 0.1755, + "step": 1599 + }, + { + "epoch": 0.08, + "learning_rate": 3.970047793084817e-07, + "loss": 0.1492, + "step": 1600 + }, + { + "epoch": 0.08, + "learning_rate": 3.9699894679165347e-07, + "loss": 0.1387, + "step": 1601 + }, + { + "epoch": 0.08, + "learning_rate": 3.969931086445193e-07, + "loss": 0.1841, + "step": 1602 + }, + { + "epoch": 0.08, + "learning_rate": 3.96987264867246e-07, + "loss": 0.165, + "step": 1603 + }, + { + "epoch": 0.08, + "learning_rate": 3.969814154600006e-07, + "loss": 0.1638, + "step": 1604 + }, + { + "epoch": 0.08, + "learning_rate": 3.969755604229504e-07, + "loss": 0.1397, + "step": 1605 + }, + { + "epoch": 0.08, + "learning_rate": 3.969696997562626e-07, + "loss": 0.1504, + "step": 1606 + }, + { + "epoch": 0.08, + "learning_rate": 3.969638334601047e-07, + "loss": 0.1417, + "step": 1607 + }, + { + "epoch": 0.08, + "learning_rate": 3.9695796153464443e-07, + "loss": 0.1087, + "step": 1608 + }, + { + "epoch": 0.08, + "learning_rate": 3.969520839800496e-07, + "loss": 0.1397, + "step": 1609 + }, + { + "epoch": 0.08, + "learning_rate": 3.9694620079648816e-07, + "loss": 0.1359, + "step": 1610 + }, + { + "epoch": 0.08, + "learning_rate": 3.9694031198412827e-07, + "loss": 0.1572, + "step": 1611 + }, + { + "epoch": 0.08, + "learning_rate": 3.9693441754313825e-07, + "loss": 0.1621, + "step": 1612 + }, + { + "epoch": 0.08, + "learning_rate": 3.9692851747368657e-07, + "loss": 0.1215, + "step": 1613 + }, + { + "epoch": 0.08, + "learning_rate": 3.969226117759418e-07, + "loss": 0.2317, + "step": 1614 + }, + { + "epoch": 0.08, + "learning_rate": 3.969167004500728e-07, + "loss": 0.1664, + "step": 1615 + }, + { + "epoch": 0.08, + "learning_rate": 3.9691078349624845e-07, + "loss": 0.1453, + "step": 1616 + }, + { + "epoch": 0.08, + "learning_rate": 3.969048609146379e-07, + "loss": 0.1759, + "step": 1617 + }, + { + "epoch": 0.08, + "learning_rate": 3.968989327054104e-07, + "loss": 0.162, + "step": 1618 + }, + { + "epoch": 0.08, + "learning_rate": 3.9689299886873544e-07, + "loss": 0.1217, + "step": 1619 + }, + { + "epoch": 0.08, + "learning_rate": 3.968870594047825e-07, + "loss": 0.134, + "step": 1620 + }, + { + "epoch": 0.08, + "learning_rate": 3.968811143137214e-07, + "loss": 0.2002, + "step": 1621 + }, + { + "epoch": 0.08, + "learning_rate": 3.9687516359572216e-07, + "loss": 0.1575, + "step": 1622 + }, + { + "epoch": 0.08, + "learning_rate": 3.9686920725095465e-07, + "loss": 0.1601, + "step": 1623 + }, + { + "epoch": 0.08, + "learning_rate": 3.968632452795892e-07, + "loss": 0.2088, + "step": 1624 + }, + { + "epoch": 0.08, + "learning_rate": 3.968572776817963e-07, + "loss": 0.212, + "step": 1625 + }, + { + "epoch": 0.08, + "learning_rate": 3.9685130445774627e-07, + "loss": 0.1257, + "step": 1626 + }, + { + "epoch": 0.08, + "learning_rate": 3.9684532560760997e-07, + "loss": 0.1545, + "step": 1627 + }, + { + "epoch": 0.08, + "learning_rate": 3.968393411315584e-07, + "loss": 0.1432, + "step": 1628 + }, + { + "epoch": 0.09, + "learning_rate": 3.968333510297624e-07, + "loss": 0.1323, + "step": 1629 + }, + { + "epoch": 0.09, + "learning_rate": 3.968273553023932e-07, + "loss": 0.141, + "step": 1630 + }, + { + "epoch": 0.09, + "learning_rate": 3.9682135394962226e-07, + "loss": 0.1812, + "step": 1631 + }, + { + "epoch": 0.09, + "learning_rate": 3.9681534697162105e-07, + "loss": 0.1805, + "step": 1632 + }, + { + "epoch": 0.09, + "learning_rate": 3.9680933436856117e-07, + "loss": 0.1686, + "step": 1633 + }, + { + "epoch": 0.09, + "learning_rate": 3.9680331614061463e-07, + "loss": 0.1554, + "step": 1634 + }, + { + "epoch": 0.09, + "learning_rate": 3.9679729228795333e-07, + "loss": 0.1559, + "step": 1635 + }, + { + "epoch": 0.09, + "learning_rate": 3.967912628107494e-07, + "loss": 0.1466, + "step": 1636 + }, + { + "epoch": 0.09, + "learning_rate": 3.9678522770917515e-07, + "loss": 0.1457, + "step": 1637 + }, + { + "epoch": 0.09, + "learning_rate": 3.9677918698340324e-07, + "loss": 0.1465, + "step": 1638 + }, + { + "epoch": 0.09, + "learning_rate": 3.967731406336062e-07, + "loss": 0.1679, + "step": 1639 + }, + { + "epoch": 0.09, + "learning_rate": 3.9676708865995677e-07, + "loss": 0.1497, + "step": 1640 + }, + { + "epoch": 0.09, + "learning_rate": 3.967610310626281e-07, + "loss": 0.1321, + "step": 1641 + }, + { + "epoch": 0.09, + "learning_rate": 3.9675496784179306e-07, + "loss": 0.1908, + "step": 1642 + }, + { + "epoch": 0.09, + "learning_rate": 3.967488989976251e-07, + "loss": 0.1248, + "step": 1643 + }, + { + "epoch": 0.09, + "learning_rate": 3.9674282453029775e-07, + "loss": 0.149, + "step": 1644 + }, + { + "epoch": 0.09, + "learning_rate": 3.967367444399845e-07, + "loss": 0.1239, + "step": 1645 + }, + { + "epoch": 0.09, + "learning_rate": 3.9673065872685907e-07, + "loss": 0.1533, + "step": 1646 + }, + { + "epoch": 0.09, + "learning_rate": 3.9672456739109556e-07, + "loss": 0.1188, + "step": 1647 + }, + { + "epoch": 0.09, + "learning_rate": 3.967184704328679e-07, + "loss": 0.1425, + "step": 1648 + }, + { + "epoch": 0.09, + "learning_rate": 3.967123678523505e-07, + "loss": 0.1812, + "step": 1649 + }, + { + "epoch": 0.09, + "learning_rate": 3.9670625964971757e-07, + "loss": 0.1682, + "step": 1650 + }, + { + "epoch": 0.09, + "learning_rate": 3.967001458251439e-07, + "loss": 0.136, + "step": 1651 + }, + { + "epoch": 0.09, + "learning_rate": 3.9669402637880407e-07, + "loss": 0.131, + "step": 1652 + }, + { + "epoch": 0.09, + "learning_rate": 3.9668790131087305e-07, + "loss": 0.1737, + "step": 1653 + }, + { + "epoch": 0.09, + "learning_rate": 3.9668177062152593e-07, + "loss": 0.1651, + "step": 1654 + }, + { + "epoch": 0.09, + "learning_rate": 3.966756343109378e-07, + "loss": 0.1483, + "step": 1655 + }, + { + "epoch": 0.09, + "learning_rate": 3.966694923792842e-07, + "loss": 0.123, + "step": 1656 + }, + { + "epoch": 0.09, + "learning_rate": 3.966633448267405e-07, + "loss": 0.1131, + "step": 1657 + }, + { + "epoch": 0.09, + "learning_rate": 3.9665719165348256e-07, + "loss": 0.1483, + "step": 1658 + }, + { + "epoch": 0.09, + "learning_rate": 3.966510328596861e-07, + "loss": 0.2329, + "step": 1659 + }, + { + "epoch": 0.09, + "learning_rate": 3.9664486844552727e-07, + "loss": 0.1816, + "step": 1660 + }, + { + "epoch": 0.09, + "learning_rate": 3.966386984111821e-07, + "loss": 0.1591, + "step": 1661 + }, + { + "epoch": 0.09, + "learning_rate": 3.9663252275682713e-07, + "loss": 0.138, + "step": 1662 + }, + { + "epoch": 0.09, + "learning_rate": 3.9662634148263867e-07, + "loss": 0.1506, + "step": 1663 + }, + { + "epoch": 0.09, + "learning_rate": 3.966201545887935e-07, + "loss": 0.1253, + "step": 1664 + }, + { + "epoch": 0.09, + "learning_rate": 3.966139620754684e-07, + "loss": 0.1637, + "step": 1665 + }, + { + "epoch": 0.09, + "learning_rate": 3.9660776394284036e-07, + "loss": 0.164, + "step": 1666 + }, + { + "epoch": 0.09, + "learning_rate": 3.9660156019108654e-07, + "loss": 0.1558, + "step": 1667 + }, + { + "epoch": 0.09, + "learning_rate": 3.9659535082038424e-07, + "loss": 0.1586, + "step": 1668 + }, + { + "epoch": 0.09, + "learning_rate": 3.9658913583091086e-07, + "loss": 0.1235, + "step": 1669 + }, + { + "epoch": 0.09, + "learning_rate": 3.9658291522284416e-07, + "loss": 0.1203, + "step": 1670 + }, + { + "epoch": 0.09, + "learning_rate": 3.9657668899636186e-07, + "loss": 0.1677, + "step": 1671 + }, + { + "epoch": 0.09, + "learning_rate": 3.9657045715164185e-07, + "loss": 0.1672, + "step": 1672 + }, + { + "epoch": 0.09, + "learning_rate": 3.965642196888623e-07, + "loss": 0.1418, + "step": 1673 + }, + { + "epoch": 0.09, + "learning_rate": 3.9655797660820144e-07, + "loss": 0.1504, + "step": 1674 + }, + { + "epoch": 0.09, + "learning_rate": 3.965517279098378e-07, + "loss": 0.1342, + "step": 1675 + }, + { + "epoch": 0.09, + "learning_rate": 3.9654547359394986e-07, + "loss": 0.1398, + "step": 1676 + }, + { + "epoch": 0.09, + "learning_rate": 3.965392136607164e-07, + "loss": 0.1279, + "step": 1677 + }, + { + "epoch": 0.09, + "learning_rate": 3.9653294811031637e-07, + "loss": 0.1448, + "step": 1678 + }, + { + "epoch": 0.09, + "learning_rate": 3.965266769429288e-07, + "loss": 0.1493, + "step": 1679 + }, + { + "epoch": 0.09, + "learning_rate": 3.9652040015873294e-07, + "loss": 0.1674, + "step": 1680 + }, + { + "epoch": 0.09, + "learning_rate": 3.965141177579082e-07, + "loss": 0.1645, + "step": 1681 + }, + { + "epoch": 0.09, + "learning_rate": 3.9650782974063413e-07, + "loss": 0.2383, + "step": 1682 + }, + { + "epoch": 0.09, + "learning_rate": 3.9650153610709037e-07, + "loss": 0.1678, + "step": 1683 + }, + { + "epoch": 0.09, + "learning_rate": 3.964952368574568e-07, + "loss": 0.1533, + "step": 1684 + }, + { + "epoch": 0.09, + "learning_rate": 3.964889319919136e-07, + "loss": 0.1783, + "step": 1685 + }, + { + "epoch": 0.09, + "learning_rate": 3.9648262151064085e-07, + "loss": 0.1309, + "step": 1686 + }, + { + "epoch": 0.09, + "learning_rate": 3.9647630541381896e-07, + "loss": 0.1498, + "step": 1687 + }, + { + "epoch": 0.09, + "learning_rate": 3.9646998370162835e-07, + "loss": 0.1523, + "step": 1688 + }, + { + "epoch": 0.09, + "learning_rate": 3.9646365637424983e-07, + "loss": 0.146, + "step": 1689 + }, + { + "epoch": 0.09, + "learning_rate": 3.964573234318641e-07, + "loss": 0.1333, + "step": 1690 + }, + { + "epoch": 0.09, + "learning_rate": 3.964509848746522e-07, + "loss": 0.1462, + "step": 1691 + }, + { + "epoch": 0.09, + "learning_rate": 3.964446407027954e-07, + "loss": 0.1209, + "step": 1692 + }, + { + "epoch": 0.09, + "learning_rate": 3.964382909164749e-07, + "loss": 0.1445, + "step": 1693 + }, + { + "epoch": 0.09, + "learning_rate": 3.964319355158722e-07, + "loss": 0.1409, + "step": 1694 + }, + { + "epoch": 0.09, + "learning_rate": 3.96425574501169e-07, + "loss": 0.1641, + "step": 1695 + }, + { + "epoch": 0.09, + "learning_rate": 3.96419207872547e-07, + "loss": 0.1329, + "step": 1696 + }, + { + "epoch": 0.09, + "learning_rate": 3.964128356301883e-07, + "loss": 0.1168, + "step": 1697 + }, + { + "epoch": 0.09, + "learning_rate": 3.9640645777427483e-07, + "loss": 0.2213, + "step": 1698 + }, + { + "epoch": 0.09, + "learning_rate": 3.96400074304989e-07, + "loss": 0.1406, + "step": 1699 + }, + { + "epoch": 0.09, + "learning_rate": 3.963936852225132e-07, + "loss": 0.1563, + "step": 1700 + }, + { + "epoch": 0.09, + "learning_rate": 3.9638729052703013e-07, + "loss": 0.1424, + "step": 1701 + }, + { + "epoch": 0.09, + "learning_rate": 3.9638089021872244e-07, + "loss": 0.1649, + "step": 1702 + }, + { + "epoch": 0.09, + "learning_rate": 3.9637448429777317e-07, + "loss": 0.1826, + "step": 1703 + }, + { + "epoch": 0.09, + "learning_rate": 3.963680727643653e-07, + "loss": 0.1251, + "step": 1704 + }, + { + "epoch": 0.09, + "learning_rate": 3.9636165561868203e-07, + "loss": 0.1538, + "step": 1705 + }, + { + "epoch": 0.09, + "learning_rate": 3.963552328609069e-07, + "loss": 0.122, + "step": 1706 + }, + { + "epoch": 0.09, + "learning_rate": 3.963488044912234e-07, + "loss": 0.1599, + "step": 1707 + }, + { + "epoch": 0.09, + "learning_rate": 3.963423705098153e-07, + "loss": 0.1247, + "step": 1708 + }, + { + "epoch": 0.09, + "learning_rate": 3.963359309168664e-07, + "loss": 0.1411, + "step": 1709 + }, + { + "epoch": 0.09, + "learning_rate": 3.963294857125608e-07, + "loss": 0.1545, + "step": 1710 + }, + { + "epoch": 0.09, + "learning_rate": 3.9632303489708273e-07, + "loss": 0.1622, + "step": 1711 + }, + { + "epoch": 0.09, + "learning_rate": 3.9631657847061655e-07, + "loss": 0.1606, + "step": 1712 + }, + { + "epoch": 0.09, + "learning_rate": 3.9631011643334674e-07, + "loss": 0.1701, + "step": 1713 + }, + { + "epoch": 0.09, + "learning_rate": 3.9630364878545807e-07, + "loss": 0.1179, + "step": 1714 + }, + { + "epoch": 0.09, + "learning_rate": 3.9629717552713533e-07, + "loss": 0.1668, + "step": 1715 + }, + { + "epoch": 0.09, + "learning_rate": 3.9629069665856345e-07, + "loss": 0.1271, + "step": 1716 + }, + { + "epoch": 0.09, + "learning_rate": 3.9628421217992783e-07, + "loss": 0.1636, + "step": 1717 + }, + { + "epoch": 0.09, + "learning_rate": 3.962777220914136e-07, + "loss": 0.1943, + "step": 1718 + }, + { + "epoch": 0.09, + "learning_rate": 3.962712263932062e-07, + "loss": 0.1542, + "step": 1719 + }, + { + "epoch": 0.09, + "learning_rate": 3.962647250854914e-07, + "loss": 0.1743, + "step": 1720 + }, + { + "epoch": 0.09, + "learning_rate": 3.9625821816845513e-07, + "loss": 0.1624, + "step": 1721 + }, + { + "epoch": 0.09, + "learning_rate": 3.9625170564228307e-07, + "loss": 0.1549, + "step": 1722 + }, + { + "epoch": 0.09, + "learning_rate": 3.962451875071616e-07, + "loss": 0.1643, + "step": 1723 + }, + { + "epoch": 0.09, + "learning_rate": 3.9623866376327693e-07, + "loss": 0.1371, + "step": 1724 + }, + { + "epoch": 0.09, + "learning_rate": 3.962321344108154e-07, + "loss": 0.1237, + "step": 1725 + }, + { + "epoch": 0.09, + "learning_rate": 3.962255994499637e-07, + "loss": 0.1491, + "step": 1726 + }, + { + "epoch": 0.09, + "learning_rate": 3.962190588809087e-07, + "loss": 0.1584, + "step": 1727 + }, + { + "epoch": 0.09, + "learning_rate": 3.9621251270383724e-07, + "loss": 0.1461, + "step": 1728 + }, + { + "epoch": 0.09, + "learning_rate": 3.962059609189364e-07, + "loss": 0.1166, + "step": 1729 + }, + { + "epoch": 0.09, + "learning_rate": 3.9619940352639346e-07, + "loss": 0.1401, + "step": 1730 + }, + { + "epoch": 0.09, + "learning_rate": 3.9619284052639586e-07, + "loss": 0.1576, + "step": 1731 + }, + { + "epoch": 0.09, + "learning_rate": 3.961862719191311e-07, + "loss": 0.1447, + "step": 1732 + }, + { + "epoch": 0.09, + "learning_rate": 3.9617969770478695e-07, + "loss": 0.1353, + "step": 1733 + }, + { + "epoch": 0.09, + "learning_rate": 3.961731178835513e-07, + "loss": 0.1406, + "step": 1734 + }, + { + "epoch": 0.09, + "learning_rate": 3.9616653245561225e-07, + "loss": 0.1178, + "step": 1735 + }, + { + "epoch": 0.09, + "learning_rate": 3.9615994142115793e-07, + "loss": 0.1591, + "step": 1736 + }, + { + "epoch": 0.09, + "learning_rate": 3.9615334478037677e-07, + "loss": 0.1718, + "step": 1737 + }, + { + "epoch": 0.09, + "learning_rate": 3.961467425334573e-07, + "loss": 0.1151, + "step": 1738 + }, + { + "epoch": 0.09, + "learning_rate": 3.961401346805882e-07, + "loss": 0.2115, + "step": 1739 + }, + { + "epoch": 0.09, + "learning_rate": 3.961335212219584e-07, + "loss": 0.1502, + "step": 1740 + }, + { + "epoch": 0.09, + "learning_rate": 3.961269021577568e-07, + "loss": 0.1861, + "step": 1741 + }, + { + "epoch": 0.09, + "learning_rate": 3.961202774881726e-07, + "loss": 0.1399, + "step": 1742 + }, + { + "epoch": 0.09, + "learning_rate": 3.9611364721339514e-07, + "loss": 0.134, + "step": 1743 + }, + { + "epoch": 0.09, + "learning_rate": 3.9610701133361396e-07, + "loss": 0.1383, + "step": 1744 + }, + { + "epoch": 0.09, + "learning_rate": 3.9610036984901873e-07, + "loss": 0.1946, + "step": 1745 + }, + { + "epoch": 0.09, + "learning_rate": 3.9609372275979924e-07, + "loss": 0.1671, + "step": 1746 + }, + { + "epoch": 0.09, + "learning_rate": 3.960870700661454e-07, + "loss": 0.1764, + "step": 1747 + }, + { + "epoch": 0.09, + "learning_rate": 3.960804117682474e-07, + "loss": 0.1451, + "step": 1748 + }, + { + "epoch": 0.09, + "learning_rate": 3.9607374786629555e-07, + "loss": 0.1097, + "step": 1749 + }, + { + "epoch": 0.09, + "learning_rate": 3.9606707836048024e-07, + "loss": 0.1216, + "step": 1750 + }, + { + "epoch": 0.09, + "learning_rate": 3.9606040325099227e-07, + "loss": 0.127, + "step": 1751 + }, + { + "epoch": 0.09, + "learning_rate": 3.960537225380222e-07, + "loss": 0.187, + "step": 1752 + }, + { + "epoch": 0.09, + "learning_rate": 3.9604703622176106e-07, + "loss": 0.1307, + "step": 1753 + }, + { + "epoch": 0.09, + "learning_rate": 3.9604034430239994e-07, + "loss": 0.1726, + "step": 1754 + }, + { + "epoch": 0.09, + "learning_rate": 3.9603364678013013e-07, + "loss": 0.1702, + "step": 1755 + }, + { + "epoch": 0.09, + "learning_rate": 3.96026943655143e-07, + "loss": 0.1855, + "step": 1756 + }, + { + "epoch": 0.09, + "learning_rate": 3.960202349276301e-07, + "loss": 0.1614, + "step": 1757 + }, + { + "epoch": 0.09, + "learning_rate": 3.960135205977833e-07, + "loss": 0.2133, + "step": 1758 + }, + { + "epoch": 0.09, + "learning_rate": 3.960068006657944e-07, + "loss": 0.1271, + "step": 1759 + }, + { + "epoch": 0.09, + "learning_rate": 3.960000751318554e-07, + "loss": 0.1414, + "step": 1760 + }, + { + "epoch": 0.09, + "learning_rate": 3.9599334399615865e-07, + "loss": 0.1704, + "step": 1761 + }, + { + "epoch": 0.09, + "learning_rate": 3.9598660725889643e-07, + "loss": 0.1454, + "step": 1762 + }, + { + "epoch": 0.09, + "learning_rate": 3.9597986492026134e-07, + "loss": 0.1442, + "step": 1763 + }, + { + "epoch": 0.09, + "learning_rate": 3.959731169804461e-07, + "loss": 0.1765, + "step": 1764 + }, + { + "epoch": 0.09, + "learning_rate": 3.9596636343964346e-07, + "loss": 0.1604, + "step": 1765 + }, + { + "epoch": 0.09, + "learning_rate": 3.959596042980465e-07, + "loss": 0.16, + "step": 1766 + }, + { + "epoch": 0.09, + "learning_rate": 3.959528395558484e-07, + "loss": 0.1473, + "step": 1767 + }, + { + "epoch": 0.09, + "learning_rate": 3.9594606921324253e-07, + "loss": 0.1374, + "step": 1768 + }, + { + "epoch": 0.09, + "learning_rate": 3.959392932704223e-07, + "loss": 0.156, + "step": 1769 + }, + { + "epoch": 0.09, + "learning_rate": 3.9593251172758153e-07, + "loss": 0.2003, + "step": 1770 + }, + { + "epoch": 0.09, + "learning_rate": 3.959257245849139e-07, + "loss": 0.159, + "step": 1771 + }, + { + "epoch": 0.09, + "learning_rate": 3.959189318426134e-07, + "loss": 0.1647, + "step": 1772 + }, + { + "epoch": 0.09, + "learning_rate": 3.959121335008742e-07, + "loss": 0.1329, + "step": 1773 + }, + { + "epoch": 0.09, + "learning_rate": 3.959053295598906e-07, + "loss": 0.1689, + "step": 1774 + }, + { + "epoch": 0.09, + "learning_rate": 3.9589852001985707e-07, + "loss": 0.1663, + "step": 1775 + }, + { + "epoch": 0.09, + "learning_rate": 3.958917048809682e-07, + "loss": 0.1575, + "step": 1776 + }, + { + "epoch": 0.09, + "learning_rate": 3.958848841434188e-07, + "loss": 0.1739, + "step": 1777 + }, + { + "epoch": 0.09, + "learning_rate": 3.9587805780740373e-07, + "loss": 0.1332, + "step": 1778 + }, + { + "epoch": 0.09, + "learning_rate": 3.9587122587311823e-07, + "loss": 0.128, + "step": 1779 + }, + { + "epoch": 0.09, + "learning_rate": 3.958643883407574e-07, + "loss": 0.1677, + "step": 1780 + }, + { + "epoch": 0.09, + "learning_rate": 3.958575452105168e-07, + "loss": 0.1164, + "step": 1781 + }, + { + "epoch": 0.09, + "learning_rate": 3.958506964825919e-07, + "loss": 0.1342, + "step": 1782 + }, + { + "epoch": 0.09, + "learning_rate": 3.9584384215717855e-07, + "loss": 0.1731, + "step": 1783 + }, + { + "epoch": 0.09, + "learning_rate": 3.958369822344726e-07, + "loss": 0.1413, + "step": 1784 + }, + { + "epoch": 0.09, + "learning_rate": 3.9583011671467004e-07, + "loss": 0.1679, + "step": 1785 + }, + { + "epoch": 0.09, + "learning_rate": 3.958232455979671e-07, + "loss": 0.1651, + "step": 1786 + }, + { + "epoch": 0.09, + "learning_rate": 3.9581636888456033e-07, + "loss": 0.1114, + "step": 1787 + }, + { + "epoch": 0.09, + "learning_rate": 3.958094865746461e-07, + "loss": 0.1474, + "step": 1788 + }, + { + "epoch": 0.09, + "learning_rate": 3.958025986684211e-07, + "loss": 0.1735, + "step": 1789 + }, + { + "epoch": 0.09, + "learning_rate": 3.9579570516608227e-07, + "loss": 0.1418, + "step": 1790 + }, + { + "epoch": 0.09, + "learning_rate": 3.957888060678266e-07, + "loss": 0.1386, + "step": 1791 + }, + { + "epoch": 0.09, + "learning_rate": 3.957819013738513e-07, + "loss": 0.1314, + "step": 1792 + }, + { + "epoch": 0.09, + "learning_rate": 3.957749910843537e-07, + "loss": 0.1473, + "step": 1793 + }, + { + "epoch": 0.09, + "learning_rate": 3.957680751995312e-07, + "loss": 0.1653, + "step": 1794 + }, + { + "epoch": 0.09, + "learning_rate": 3.957611537195816e-07, + "loss": 0.2166, + "step": 1795 + }, + { + "epoch": 0.09, + "learning_rate": 3.957542266447026e-07, + "loss": 0.1325, + "step": 1796 + }, + { + "epoch": 0.09, + "learning_rate": 3.9574729397509227e-07, + "loss": 0.1252, + "step": 1797 + }, + { + "epoch": 0.09, + "learning_rate": 3.957403557109487e-07, + "loss": 0.1911, + "step": 1798 + }, + { + "epoch": 0.09, + "learning_rate": 3.957334118524702e-07, + "loss": 0.1362, + "step": 1799 + }, + { + "epoch": 0.09, + "learning_rate": 3.9572646239985523e-07, + "loss": 0.1508, + "step": 1800 + }, + { + "epoch": 0.09, + "learning_rate": 3.9571950735330243e-07, + "loss": 0.1826, + "step": 1801 + }, + { + "epoch": 0.09, + "learning_rate": 3.9571254671301054e-07, + "loss": 0.1602, + "step": 1802 + }, + { + "epoch": 0.09, + "learning_rate": 3.9570558047917855e-07, + "loss": 0.1283, + "step": 1803 + }, + { + "epoch": 0.09, + "learning_rate": 3.9569860865200547e-07, + "loss": 0.1396, + "step": 1804 + }, + { + "epoch": 0.09, + "learning_rate": 3.956916312316906e-07, + "loss": 0.1506, + "step": 1805 + }, + { + "epoch": 0.09, + "learning_rate": 3.9568464821843337e-07, + "loss": 0.1515, + "step": 1806 + }, + { + "epoch": 0.09, + "learning_rate": 3.956776596124334e-07, + "loss": 0.1639, + "step": 1807 + }, + { + "epoch": 0.09, + "learning_rate": 3.9567066541389034e-07, + "loss": 0.1716, + "step": 1808 + }, + { + "epoch": 0.09, + "learning_rate": 3.956636656230041e-07, + "loss": 0.1806, + "step": 1809 + }, + { + "epoch": 0.09, + "learning_rate": 3.956566602399748e-07, + "loss": 0.1849, + "step": 1810 + }, + { + "epoch": 0.09, + "learning_rate": 3.956496492650026e-07, + "loss": 0.1763, + "step": 1811 + }, + { + "epoch": 0.09, + "learning_rate": 3.956426326982879e-07, + "loss": 0.1472, + "step": 1812 + }, + { + "epoch": 0.09, + "learning_rate": 3.956356105400313e-07, + "loss": 0.1829, + "step": 1813 + }, + { + "epoch": 0.09, + "learning_rate": 3.9562858279043335e-07, + "loss": 0.1594, + "step": 1814 + }, + { + "epoch": 0.09, + "learning_rate": 3.9562154944969503e-07, + "loss": 0.1149, + "step": 1815 + }, + { + "epoch": 0.09, + "learning_rate": 3.956145105180172e-07, + "loss": 0.1864, + "step": 1816 + }, + { + "epoch": 0.09, + "learning_rate": 3.9560746599560125e-07, + "loss": 0.1628, + "step": 1817 + }, + { + "epoch": 0.09, + "learning_rate": 3.956004158826484e-07, + "loss": 0.1631, + "step": 1818 + }, + { + "epoch": 0.09, + "learning_rate": 3.9559336017936015e-07, + "loss": 0.1757, + "step": 1819 + }, + { + "epoch": 0.09, + "learning_rate": 3.9558629888593815e-07, + "loss": 0.1442, + "step": 1820 + }, + { + "epoch": 0.1, + "learning_rate": 3.955792320025842e-07, + "loss": 0.1291, + "step": 1821 + }, + { + "epoch": 0.1, + "learning_rate": 3.955721595295003e-07, + "loss": 0.156, + "step": 1822 + }, + { + "epoch": 0.1, + "learning_rate": 3.955650814668886e-07, + "loss": 0.1564, + "step": 1823 + }, + { + "epoch": 0.1, + "learning_rate": 3.9555799781495143e-07, + "loss": 0.1377, + "step": 1824 + }, + { + "epoch": 0.1, + "learning_rate": 3.955509085738911e-07, + "loss": 0.1656, + "step": 1825 + }, + { + "epoch": 0.1, + "learning_rate": 3.9554381374391036e-07, + "loss": 0.1456, + "step": 1826 + }, + { + "epoch": 0.1, + "learning_rate": 3.9553671332521194e-07, + "loss": 0.1818, + "step": 1827 + }, + { + "epoch": 0.1, + "learning_rate": 3.955296073179987e-07, + "loss": 0.1642, + "step": 1828 + }, + { + "epoch": 0.1, + "learning_rate": 3.955224957224739e-07, + "loss": 0.1448, + "step": 1829 + }, + { + "epoch": 0.1, + "learning_rate": 3.955153785388407e-07, + "loss": 0.1416, + "step": 1830 + }, + { + "epoch": 0.1, + "learning_rate": 3.955082557673024e-07, + "loss": 0.1534, + "step": 1831 + }, + { + "epoch": 0.1, + "learning_rate": 3.955011274080627e-07, + "loss": 0.1392, + "step": 1832 + }, + { + "epoch": 0.1, + "learning_rate": 3.9549399346132536e-07, + "loss": 0.1738, + "step": 1833 + }, + { + "epoch": 0.1, + "learning_rate": 3.954868539272942e-07, + "loss": 0.1797, + "step": 1834 + }, + { + "epoch": 0.1, + "learning_rate": 3.954797088061733e-07, + "loss": 0.1306, + "step": 1835 + }, + { + "epoch": 0.1, + "learning_rate": 3.954725580981668e-07, + "loss": 0.1605, + "step": 1836 + }, + { + "epoch": 0.1, + "learning_rate": 3.954654018034792e-07, + "loss": 0.152, + "step": 1837 + }, + { + "epoch": 0.1, + "learning_rate": 3.9545823992231495e-07, + "loss": 0.1769, + "step": 1838 + }, + { + "epoch": 0.1, + "learning_rate": 3.954510724548787e-07, + "loss": 0.1237, + "step": 1839 + }, + { + "epoch": 0.1, + "learning_rate": 3.954438994013754e-07, + "loss": 0.1517, + "step": 1840 + }, + { + "epoch": 0.1, + "learning_rate": 3.9543672076201e-07, + "loss": 0.1531, + "step": 1841 + }, + { + "epoch": 0.1, + "learning_rate": 3.9542953653698765e-07, + "loss": 0.1419, + "step": 1842 + }, + { + "epoch": 0.1, + "learning_rate": 3.9542234672651373e-07, + "loss": 0.1297, + "step": 1843 + }, + { + "epoch": 0.1, + "learning_rate": 3.954151513307937e-07, + "loss": 0.1776, + "step": 1844 + }, + { + "epoch": 0.1, + "learning_rate": 3.954079503500332e-07, + "loss": 0.1454, + "step": 1845 + }, + { + "epoch": 0.1, + "learning_rate": 3.95400743784438e-07, + "loss": 0.1481, + "step": 1846 + }, + { + "epoch": 0.1, + "learning_rate": 3.953935316342141e-07, + "loss": 0.129, + "step": 1847 + }, + { + "epoch": 0.1, + "learning_rate": 3.953863138995677e-07, + "loss": 0.1746, + "step": 1848 + }, + { + "epoch": 0.1, + "learning_rate": 3.95379090580705e-07, + "loss": 0.1502, + "step": 1849 + }, + { + "epoch": 0.1, + "learning_rate": 3.9537186167783245e-07, + "loss": 0.1722, + "step": 1850 + }, + { + "epoch": 0.1, + "learning_rate": 3.953646271911567e-07, + "loss": 0.1442, + "step": 1851 + }, + { + "epoch": 0.1, + "learning_rate": 3.953573871208845e-07, + "loss": 0.1704, + "step": 1852 + }, + { + "epoch": 0.1, + "learning_rate": 3.953501414672227e-07, + "loss": 0.1399, + "step": 1853 + }, + { + "epoch": 0.1, + "learning_rate": 3.953428902303785e-07, + "loss": 0.1507, + "step": 1854 + }, + { + "epoch": 0.1, + "learning_rate": 3.9533563341055907e-07, + "loss": 0.1884, + "step": 1855 + }, + { + "epoch": 0.1, + "learning_rate": 3.9532837100797185e-07, + "loss": 0.1343, + "step": 1856 + }, + { + "epoch": 0.1, + "learning_rate": 3.9532110302282436e-07, + "loss": 0.153, + "step": 1857 + }, + { + "epoch": 0.1, + "learning_rate": 3.953138294553244e-07, + "loss": 0.1277, + "step": 1858 + }, + { + "epoch": 0.1, + "learning_rate": 3.953065503056797e-07, + "loss": 0.1508, + "step": 1859 + }, + { + "epoch": 0.1, + "learning_rate": 3.952992655740984e-07, + "loss": 0.1427, + "step": 1860 + }, + { + "epoch": 0.1, + "learning_rate": 3.9529197526078874e-07, + "loss": 0.148, + "step": 1861 + }, + { + "epoch": 0.1, + "learning_rate": 3.9528467936595906e-07, + "loss": 0.1433, + "step": 1862 + }, + { + "epoch": 0.1, + "learning_rate": 3.9527737788981784e-07, + "loss": 0.1418, + "step": 1863 + }, + { + "epoch": 0.1, + "learning_rate": 3.9527007083257373e-07, + "loss": 0.1721, + "step": 1864 + }, + { + "epoch": 0.1, + "learning_rate": 3.9526275819443573e-07, + "loss": 0.1512, + "step": 1865 + }, + { + "epoch": 0.1, + "learning_rate": 3.952554399756126e-07, + "loss": 0.1559, + "step": 1866 + }, + { + "epoch": 0.1, + "learning_rate": 3.9524811617631366e-07, + "loss": 0.1397, + "step": 1867 + }, + { + "epoch": 0.1, + "learning_rate": 3.952407867967482e-07, + "loss": 0.1471, + "step": 1868 + }, + { + "epoch": 0.1, + "learning_rate": 3.952334518371256e-07, + "loss": 0.1481, + "step": 1869 + }, + { + "epoch": 0.1, + "learning_rate": 3.952261112976557e-07, + "loss": 0.1344, + "step": 1870 + }, + { + "epoch": 0.1, + "learning_rate": 3.9521876517854807e-07, + "loss": 0.1615, + "step": 1871 + }, + { + "epoch": 0.1, + "learning_rate": 3.952114134800129e-07, + "loss": 0.1879, + "step": 1872 + }, + { + "epoch": 0.1, + "learning_rate": 3.952040562022601e-07, + "loss": 0.1561, + "step": 1873 + }, + { + "epoch": 0.1, + "learning_rate": 3.951966933455e-07, + "loss": 0.2035, + "step": 1874 + }, + { + "epoch": 0.1, + "learning_rate": 3.95189324909943e-07, + "loss": 0.165, + "step": 1875 + }, + { + "epoch": 0.1, + "learning_rate": 3.951819508957998e-07, + "loss": 0.1394, + "step": 1876 + }, + { + "epoch": 0.1, + "learning_rate": 3.951745713032811e-07, + "loss": 0.1628, + "step": 1877 + }, + { + "epoch": 0.1, + "learning_rate": 3.9516718613259784e-07, + "loss": 0.1535, + "step": 1878 + }, + { + "epoch": 0.1, + "learning_rate": 3.9515979538396104e-07, + "loss": 0.1881, + "step": 1879 + }, + { + "epoch": 0.1, + "learning_rate": 3.9515239905758194e-07, + "loss": 0.1307, + "step": 1880 + }, + { + "epoch": 0.1, + "learning_rate": 3.951449971536719e-07, + "loss": 0.1651, + "step": 1881 + }, + { + "epoch": 0.1, + "learning_rate": 3.951375896724426e-07, + "loss": 0.1268, + "step": 1882 + }, + { + "epoch": 0.1, + "learning_rate": 3.951301766141055e-07, + "loss": 0.1466, + "step": 1883 + }, + { + "epoch": 0.1, + "learning_rate": 3.951227579788728e-07, + "loss": 0.1794, + "step": 1884 + }, + { + "epoch": 0.1, + "learning_rate": 3.9511533376695627e-07, + "loss": 0.1363, + "step": 1885 + }, + { + "epoch": 0.1, + "learning_rate": 3.951079039785681e-07, + "loss": 0.1826, + "step": 1886 + }, + { + "epoch": 0.1, + "learning_rate": 3.9510046861392084e-07, + "loss": 0.1609, + "step": 1887 + }, + { + "epoch": 0.1, + "learning_rate": 3.9509302767322686e-07, + "loss": 0.1276, + "step": 1888 + }, + { + "epoch": 0.1, + "learning_rate": 3.950855811566988e-07, + "loss": 0.1558, + "step": 1889 + }, + { + "epoch": 0.1, + "learning_rate": 3.950781290645495e-07, + "loss": 0.1584, + "step": 1890 + }, + { + "epoch": 0.1, + "learning_rate": 3.9507067139699196e-07, + "loss": 0.1577, + "step": 1891 + }, + { + "epoch": 0.1, + "learning_rate": 3.950632081542394e-07, + "loss": 0.1822, + "step": 1892 + }, + { + "epoch": 0.1, + "learning_rate": 3.95055739336505e-07, + "loss": 0.1464, + "step": 1893 + }, + { + "epoch": 0.1, + "learning_rate": 3.9504826494400227e-07, + "loss": 0.1363, + "step": 1894 + }, + { + "epoch": 0.1, + "learning_rate": 3.950407849769449e-07, + "loss": 0.1895, + "step": 1895 + }, + { + "epoch": 0.1, + "learning_rate": 3.9503329943554647e-07, + "loss": 0.1451, + "step": 1896 + }, + { + "epoch": 0.1, + "learning_rate": 3.950258083200211e-07, + "loss": 0.1483, + "step": 1897 + }, + { + "epoch": 0.1, + "learning_rate": 3.950183116305829e-07, + "loss": 0.1401, + "step": 1898 + }, + { + "epoch": 0.1, + "learning_rate": 3.9501080936744603e-07, + "loss": 0.1698, + "step": 1899 + }, + { + "epoch": 0.1, + "learning_rate": 3.950033015308249e-07, + "loss": 0.1671, + "step": 1900 + }, + { + "epoch": 0.1, + "learning_rate": 3.949957881209342e-07, + "loss": 0.1637, + "step": 1901 + }, + { + "epoch": 0.1, + "learning_rate": 3.9498826913798864e-07, + "loss": 0.1412, + "step": 1902 + }, + { + "epoch": 0.1, + "learning_rate": 3.9498074458220297e-07, + "loss": 0.1673, + "step": 1903 + }, + { + "epoch": 0.1, + "learning_rate": 3.9497321445379235e-07, + "loss": 0.1405, + "step": 1904 + }, + { + "epoch": 0.1, + "learning_rate": 3.949656787529721e-07, + "loss": 0.1275, + "step": 1905 + }, + { + "epoch": 0.1, + "learning_rate": 3.949581374799574e-07, + "loss": 0.1507, + "step": 1906 + }, + { + "epoch": 0.1, + "learning_rate": 3.949505906349639e-07, + "loss": 0.1407, + "step": 1907 + }, + { + "epoch": 0.1, + "learning_rate": 3.9494303821820725e-07, + "loss": 0.1523, + "step": 1908 + }, + { + "epoch": 0.1, + "learning_rate": 3.949354802299033e-07, + "loss": 0.1367, + "step": 1909 + }, + { + "epoch": 0.1, + "learning_rate": 3.949279166702681e-07, + "loss": 0.1347, + "step": 1910 + }, + { + "epoch": 0.1, + "learning_rate": 3.949203475395178e-07, + "loss": 0.1332, + "step": 1911 + }, + { + "epoch": 0.1, + "learning_rate": 3.9491277283786866e-07, + "loss": 0.1712, + "step": 1912 + }, + { + "epoch": 0.1, + "learning_rate": 3.949051925655373e-07, + "loss": 0.1227, + "step": 1913 + }, + { + "epoch": 0.1, + "learning_rate": 3.9489760672274025e-07, + "loss": 0.1661, + "step": 1914 + }, + { + "epoch": 0.1, + "learning_rate": 3.9489001530969437e-07, + "loss": 0.1595, + "step": 1915 + }, + { + "epoch": 0.1, + "learning_rate": 3.948824183266167e-07, + "loss": 0.1458, + "step": 1916 + }, + { + "epoch": 0.1, + "learning_rate": 3.9487481577372414e-07, + "loss": 0.164, + "step": 1917 + }, + { + "epoch": 0.1, + "learning_rate": 3.948672076512342e-07, + "loss": 0.189, + "step": 1918 + }, + { + "epoch": 0.1, + "learning_rate": 3.948595939593642e-07, + "loss": 0.149, + "step": 1919 + }, + { + "epoch": 0.1, + "learning_rate": 3.9485197469833185e-07, + "loss": 0.1232, + "step": 1920 + }, + { + "epoch": 0.1, + "learning_rate": 3.948443498683548e-07, + "loss": 0.1469, + "step": 1921 + }, + { + "epoch": 0.1, + "learning_rate": 3.9483671946965093e-07, + "loss": 0.1874, + "step": 1922 + }, + { + "epoch": 0.1, + "learning_rate": 3.9482908350243856e-07, + "loss": 0.1611, + "step": 1923 + }, + { + "epoch": 0.1, + "learning_rate": 3.9482144196693567e-07, + "loss": 0.1662, + "step": 1924 + }, + { + "epoch": 0.1, + "learning_rate": 3.948137948633608e-07, + "loss": 0.1561, + "step": 1925 + }, + { + "epoch": 0.1, + "learning_rate": 3.9480614219193243e-07, + "loss": 0.1406, + "step": 1926 + }, + { + "epoch": 0.1, + "learning_rate": 3.9479848395286936e-07, + "loss": 0.1725, + "step": 1927 + }, + { + "epoch": 0.1, + "learning_rate": 3.9479082014639034e-07, + "loss": 0.1555, + "step": 1928 + }, + { + "epoch": 0.1, + "learning_rate": 3.947831507727145e-07, + "loss": 0.1509, + "step": 1929 + }, + { + "epoch": 0.1, + "learning_rate": 3.947754758320611e-07, + "loss": 0.1695, + "step": 1930 + }, + { + "epoch": 0.1, + "learning_rate": 3.947677953246494e-07, + "loss": 0.129, + "step": 1931 + }, + { + "epoch": 0.1, + "learning_rate": 3.947601092506989e-07, + "loss": 0.1247, + "step": 1932 + }, + { + "epoch": 0.1, + "learning_rate": 3.947524176104292e-07, + "loss": 0.1756, + "step": 1933 + }, + { + "epoch": 0.1, + "learning_rate": 3.947447204040603e-07, + "loss": 0.1242, + "step": 1934 + }, + { + "epoch": 0.1, + "learning_rate": 3.947370176318122e-07, + "loss": 0.1523, + "step": 1935 + }, + { + "epoch": 0.1, + "learning_rate": 3.9472930929390484e-07, + "loss": 0.1668, + "step": 1936 + }, + { + "epoch": 0.1, + "learning_rate": 3.947215953905587e-07, + "loss": 0.1366, + "step": 1937 + }, + { + "epoch": 0.1, + "learning_rate": 3.947138759219942e-07, + "loss": 0.2027, + "step": 1938 + }, + { + "epoch": 0.1, + "learning_rate": 3.9470615088843187e-07, + "loss": 0.1637, + "step": 1939 + }, + { + "epoch": 0.1, + "learning_rate": 3.946984202900927e-07, + "loss": 0.1646, + "step": 1940 + }, + { + "epoch": 0.1, + "learning_rate": 3.946906841271974e-07, + "loss": 0.1306, + "step": 1941 + }, + { + "epoch": 0.1, + "learning_rate": 3.946829423999673e-07, + "loss": 0.1388, + "step": 1942 + }, + { + "epoch": 0.1, + "learning_rate": 3.946751951086234e-07, + "loss": 0.159, + "step": 1943 + }, + { + "epoch": 0.1, + "learning_rate": 3.9466744225338743e-07, + "loss": 0.1463, + "step": 1944 + }, + { + "epoch": 0.1, + "learning_rate": 3.9465968383448074e-07, + "loss": 0.1684, + "step": 1945 + }, + { + "epoch": 0.1, + "learning_rate": 3.946519198521252e-07, + "loss": 0.1756, + "step": 1946 + }, + { + "epoch": 0.1, + "learning_rate": 3.9464415030654256e-07, + "loss": 0.1594, + "step": 1947 + }, + { + "epoch": 0.1, + "learning_rate": 3.94636375197955e-07, + "loss": 0.1468, + "step": 1948 + }, + { + "epoch": 0.1, + "learning_rate": 3.9462859452658473e-07, + "loss": 0.1244, + "step": 1949 + }, + { + "epoch": 0.1, + "learning_rate": 3.946208082926541e-07, + "loss": 0.1637, + "step": 1950 + }, + { + "epoch": 0.1, + "learning_rate": 3.946130164963856e-07, + "loss": 0.1535, + "step": 1951 + }, + { + "epoch": 0.1, + "learning_rate": 3.94605219138002e-07, + "loss": 0.1559, + "step": 1952 + }, + { + "epoch": 0.1, + "learning_rate": 3.9459741621772606e-07, + "loss": 0.1386, + "step": 1953 + }, + { + "epoch": 0.1, + "learning_rate": 3.945896077357808e-07, + "loss": 0.1614, + "step": 1954 + }, + { + "epoch": 0.1, + "learning_rate": 3.9458179369238953e-07, + "loss": 0.1344, + "step": 1955 + }, + { + "epoch": 0.1, + "learning_rate": 3.9457397408777547e-07, + "loss": 0.1535, + "step": 1956 + }, + { + "epoch": 0.1, + "learning_rate": 3.945661489221621e-07, + "loss": 0.1352, + "step": 1957 + }, + { + "epoch": 0.1, + "learning_rate": 3.945583181957731e-07, + "loss": 0.1391, + "step": 1958 + }, + { + "epoch": 0.1, + "learning_rate": 3.945504819088322e-07, + "loss": 0.1425, + "step": 1959 + }, + { + "epoch": 0.1, + "learning_rate": 3.9454264006156345e-07, + "loss": 0.1301, + "step": 1960 + }, + { + "epoch": 0.1, + "learning_rate": 3.9453479265419096e-07, + "loss": 0.1937, + "step": 1961 + }, + { + "epoch": 0.1, + "learning_rate": 3.94526939686939e-07, + "loss": 0.1814, + "step": 1962 + }, + { + "epoch": 0.1, + "learning_rate": 3.94519081160032e-07, + "loss": 0.1631, + "step": 1963 + }, + { + "epoch": 0.1, + "learning_rate": 3.945112170736946e-07, + "loss": 0.1499, + "step": 1964 + }, + { + "epoch": 0.1, + "learning_rate": 3.945033474281515e-07, + "loss": 0.1494, + "step": 1965 + }, + { + "epoch": 0.1, + "learning_rate": 3.944954722236276e-07, + "loss": 0.1173, + "step": 1966 + }, + { + "epoch": 0.1, + "learning_rate": 3.9448759146034804e-07, + "loss": 0.1843, + "step": 1967 + }, + { + "epoch": 0.1, + "learning_rate": 3.9447970513853807e-07, + "loss": 0.1688, + "step": 1968 + }, + { + "epoch": 0.1, + "learning_rate": 3.9447181325842305e-07, + "loss": 0.1145, + "step": 1969 + }, + { + "epoch": 0.1, + "learning_rate": 3.9446391582022855e-07, + "loss": 0.1711, + "step": 1970 + }, + { + "epoch": 0.1, + "learning_rate": 3.944560128241802e-07, + "loss": 0.1297, + "step": 1971 + }, + { + "epoch": 0.1, + "learning_rate": 3.94448104270504e-07, + "loss": 0.1352, + "step": 1972 + }, + { + "epoch": 0.1, + "learning_rate": 3.9444019015942587e-07, + "loss": 0.1683, + "step": 1973 + }, + { + "epoch": 0.1, + "learning_rate": 3.9443227049117206e-07, + "loss": 0.1909, + "step": 1974 + }, + { + "epoch": 0.1, + "learning_rate": 3.944243452659689e-07, + "loss": 0.1753, + "step": 1975 + }, + { + "epoch": 0.1, + "learning_rate": 3.9441641448404293e-07, + "loss": 0.1609, + "step": 1976 + }, + { + "epoch": 0.1, + "learning_rate": 3.944084781456207e-07, + "loss": 0.1414, + "step": 1977 + }, + { + "epoch": 0.1, + "learning_rate": 3.9440053625092925e-07, + "loss": 0.1993, + "step": 1978 + }, + { + "epoch": 0.1, + "learning_rate": 3.943925888001953e-07, + "loss": 0.1258, + "step": 1979 + }, + { + "epoch": 0.1, + "learning_rate": 3.943846357936462e-07, + "loss": 0.1818, + "step": 1980 + }, + { + "epoch": 0.1, + "learning_rate": 3.943766772315091e-07, + "loss": 0.1678, + "step": 1981 + }, + { + "epoch": 0.1, + "learning_rate": 3.9436871311401163e-07, + "loss": 0.1285, + "step": 1982 + }, + { + "epoch": 0.1, + "learning_rate": 3.943607434413812e-07, + "loss": 0.1503, + "step": 1983 + }, + { + "epoch": 0.1, + "learning_rate": 3.943527682138457e-07, + "loss": 0.1383, + "step": 1984 + }, + { + "epoch": 0.1, + "learning_rate": 3.943447874316331e-07, + "loss": 0.1815, + "step": 1985 + }, + { + "epoch": 0.1, + "learning_rate": 3.943368010949714e-07, + "loss": 0.148, + "step": 1986 + }, + { + "epoch": 0.1, + "learning_rate": 3.9432880920408897e-07, + "loss": 0.1462, + "step": 1987 + }, + { + "epoch": 0.1, + "learning_rate": 3.9432081175921414e-07, + "loss": 0.1577, + "step": 1988 + }, + { + "epoch": 0.1, + "learning_rate": 3.9431280876057546e-07, + "loss": 0.1322, + "step": 1989 + }, + { + "epoch": 0.1, + "learning_rate": 3.9430480020840173e-07, + "loss": 0.1429, + "step": 1990 + }, + { + "epoch": 0.1, + "learning_rate": 3.942967861029218e-07, + "loss": 0.1693, + "step": 1991 + }, + { + "epoch": 0.1, + "learning_rate": 3.942887664443647e-07, + "loss": 0.2039, + "step": 1992 + }, + { + "epoch": 0.1, + "learning_rate": 3.942807412329596e-07, + "loss": 0.1518, + "step": 1993 + }, + { + "epoch": 0.1, + "learning_rate": 3.94272710468936e-07, + "loss": 0.1182, + "step": 1994 + }, + { + "epoch": 0.1, + "learning_rate": 3.942646741525233e-07, + "loss": 0.1472, + "step": 1995 + }, + { + "epoch": 0.1, + "learning_rate": 3.942566322839513e-07, + "loss": 0.1427, + "step": 1996 + }, + { + "epoch": 0.1, + "learning_rate": 3.9424858486344964e-07, + "loss": 0.1488, + "step": 1997 + }, + { + "epoch": 0.1, + "learning_rate": 3.9424053189124847e-07, + "loss": 0.1539, + "step": 1998 + }, + { + "epoch": 0.1, + "learning_rate": 3.9423247336757786e-07, + "loss": 0.1368, + "step": 1999 + }, + { + "epoch": 0.1, + "learning_rate": 3.9422440929266825e-07, + "loss": 0.1752, + "step": 2000 + }, + { + "epoch": 0.1, + "learning_rate": 3.942163396667501e-07, + "loss": 0.1344, + "step": 2001 + }, + { + "epoch": 0.1, + "learning_rate": 3.942082644900539e-07, + "loss": 0.1577, + "step": 2002 + }, + { + "epoch": 0.1, + "learning_rate": 3.9420018376281055e-07, + "loss": 0.1747, + "step": 2003 + }, + { + "epoch": 0.1, + "learning_rate": 3.9419209748525104e-07, + "loss": 0.1721, + "step": 2004 + }, + { + "epoch": 0.1, + "learning_rate": 3.9418400565760633e-07, + "loss": 0.1913, + "step": 2005 + }, + { + "epoch": 0.1, + "learning_rate": 3.941759082801078e-07, + "loss": 0.1603, + "step": 2006 + }, + { + "epoch": 0.1, + "learning_rate": 3.9416780535298687e-07, + "loss": 0.1562, + "step": 2007 + }, + { + "epoch": 0.1, + "learning_rate": 3.9415969687647506e-07, + "loss": 0.1694, + "step": 2008 + }, + { + "epoch": 0.1, + "learning_rate": 3.9415158285080424e-07, + "loss": 0.1842, + "step": 2009 + }, + { + "epoch": 0.1, + "learning_rate": 3.941434632762062e-07, + "loss": 0.1564, + "step": 2010 + }, + { + "epoch": 0.1, + "learning_rate": 3.9413533815291304e-07, + "loss": 0.1355, + "step": 2011 + }, + { + "epoch": 0.11, + "learning_rate": 3.941272074811569e-07, + "loss": 0.1547, + "step": 2012 + }, + { + "epoch": 0.11, + "learning_rate": 3.941190712611703e-07, + "loss": 0.1389, + "step": 2013 + }, + { + "epoch": 0.11, + "learning_rate": 3.9411092949318563e-07, + "loss": 0.1528, + "step": 2014 + }, + { + "epoch": 0.11, + "learning_rate": 3.9410278217743577e-07, + "loss": 0.1788, + "step": 2015 + }, + { + "epoch": 0.11, + "learning_rate": 3.940946293141534e-07, + "loss": 0.1615, + "step": 2016 + }, + { + "epoch": 0.11, + "learning_rate": 3.9408647090357167e-07, + "loss": 0.1638, + "step": 2017 + }, + { + "epoch": 0.11, + "learning_rate": 3.9407830694592356e-07, + "loss": 0.1389, + "step": 2018 + }, + { + "epoch": 0.11, + "learning_rate": 3.9407013744144255e-07, + "loss": 0.1798, + "step": 2019 + }, + { + "epoch": 0.11, + "learning_rate": 3.9406196239036217e-07, + "loss": 0.172, + "step": 2020 + }, + { + "epoch": 0.11, + "learning_rate": 3.9405378179291587e-07, + "loss": 0.1333, + "step": 2021 + }, + { + "epoch": 0.11, + "learning_rate": 3.940455956493377e-07, + "loss": 0.14, + "step": 2022 + }, + { + "epoch": 0.11, + "learning_rate": 3.9403740395986144e-07, + "loss": 0.1356, + "step": 2023 + }, + { + "epoch": 0.11, + "learning_rate": 3.9402920672472125e-07, + "loss": 0.1259, + "step": 2024 + }, + { + "epoch": 0.11, + "learning_rate": 3.9402100394415143e-07, + "loss": 0.1705, + "step": 2025 + }, + { + "epoch": 0.11, + "learning_rate": 3.9401279561838646e-07, + "loss": 0.1581, + "step": 2026 + }, + { + "epoch": 0.11, + "learning_rate": 3.9400458174766084e-07, + "loss": 0.147, + "step": 2027 + }, + { + "epoch": 0.11, + "learning_rate": 3.939963623322094e-07, + "loss": 0.1284, + "step": 2028 + }, + { + "epoch": 0.11, + "learning_rate": 3.939881373722671e-07, + "loss": 0.1767, + "step": 2029 + }, + { + "epoch": 0.11, + "learning_rate": 3.939799068680688e-07, + "loss": 0.1857, + "step": 2030 + }, + { + "epoch": 0.11, + "learning_rate": 3.9397167081985e-07, + "loss": 0.1436, + "step": 2031 + }, + { + "epoch": 0.11, + "learning_rate": 3.93963429227846e-07, + "loss": 0.1294, + "step": 2032 + }, + { + "epoch": 0.11, + "learning_rate": 3.9395518209229224e-07, + "loss": 0.1551, + "step": 2033 + }, + { + "epoch": 0.11, + "learning_rate": 3.939469294134245e-07, + "loss": 0.1353, + "step": 2034 + }, + { + "epoch": 0.11, + "learning_rate": 3.939386711914786e-07, + "loss": 0.1334, + "step": 2035 + }, + { + "epoch": 0.11, + "learning_rate": 3.939304074266907e-07, + "loss": 0.1593, + "step": 2036 + }, + { + "epoch": 0.11, + "learning_rate": 3.9392213811929686e-07, + "loss": 0.1914, + "step": 2037 + }, + { + "epoch": 0.11, + "learning_rate": 3.9391386326953345e-07, + "loss": 0.1404, + "step": 2038 + }, + { + "epoch": 0.11, + "learning_rate": 3.9390558287763696e-07, + "loss": 0.1833, + "step": 2039 + }, + { + "epoch": 0.11, + "learning_rate": 3.9389729694384403e-07, + "loss": 0.1189, + "step": 2040 + }, + { + "epoch": 0.11, + "learning_rate": 3.9388900546839157e-07, + "loss": 0.132, + "step": 2041 + }, + { + "epoch": 0.11, + "learning_rate": 3.938807084515164e-07, + "loss": 0.155, + "step": 2042 + }, + { + "epoch": 0.11, + "learning_rate": 3.938724058934558e-07, + "loss": 0.1376, + "step": 2043 + }, + { + "epoch": 0.11, + "learning_rate": 3.9386409779444707e-07, + "loss": 0.1592, + "step": 2044 + }, + { + "epoch": 0.11, + "learning_rate": 3.9385578415472747e-07, + "loss": 0.1741, + "step": 2045 + }, + { + "epoch": 0.11, + "learning_rate": 3.9384746497453476e-07, + "loss": 0.1588, + "step": 2046 + }, + { + "epoch": 0.11, + "learning_rate": 3.9383914025410664e-07, + "loss": 0.208, + "step": 2047 + }, + { + "epoch": 0.11, + "learning_rate": 3.938308099936811e-07, + "loss": 0.1596, + "step": 2048 + }, + { + "epoch": 0.11, + "learning_rate": 3.9382247419349614e-07, + "loss": 0.1526, + "step": 2049 + }, + { + "epoch": 0.11, + "learning_rate": 3.9381413285379006e-07, + "loss": 0.1388, + "step": 2050 + }, + { + "epoch": 0.11, + "learning_rate": 3.9380578597480126e-07, + "loss": 0.175, + "step": 2051 + }, + { + "epoch": 0.11, + "learning_rate": 3.937974335567683e-07, + "loss": 0.1234, + "step": 2052 + }, + { + "epoch": 0.11, + "learning_rate": 3.9378907559992984e-07, + "loss": 0.1581, + "step": 2053 + }, + { + "epoch": 0.11, + "learning_rate": 3.9378071210452483e-07, + "loss": 0.1337, + "step": 2054 + }, + { + "epoch": 0.11, + "learning_rate": 3.937723430707922e-07, + "loss": 0.1729, + "step": 2055 + }, + { + "epoch": 0.11, + "learning_rate": 3.937639684989713e-07, + "loss": 0.1406, + "step": 2056 + }, + { + "epoch": 0.11, + "learning_rate": 3.9375558838930125e-07, + "loss": 0.1412, + "step": 2057 + }, + { + "epoch": 0.11, + "learning_rate": 3.937472027420218e-07, + "loss": 0.1524, + "step": 2058 + }, + { + "epoch": 0.11, + "learning_rate": 3.937388115573724e-07, + "loss": 0.1403, + "step": 2059 + }, + { + "epoch": 0.11, + "learning_rate": 3.93730414835593e-07, + "loss": 0.1405, + "step": 2060 + }, + { + "epoch": 0.11, + "learning_rate": 3.9372201257692367e-07, + "loss": 0.1219, + "step": 2061 + }, + { + "epoch": 0.11, + "learning_rate": 3.937136047816043e-07, + "loss": 0.1614, + "step": 2062 + }, + { + "epoch": 0.11, + "learning_rate": 3.937051914498754e-07, + "loss": 0.1393, + "step": 2063 + }, + { + "epoch": 0.11, + "learning_rate": 3.936967725819773e-07, + "loss": 0.1255, + "step": 2064 + }, + { + "epoch": 0.11, + "learning_rate": 3.936883481781507e-07, + "loss": 0.2181, + "step": 2065 + }, + { + "epoch": 0.11, + "learning_rate": 3.936799182386363e-07, + "loss": 0.1486, + "step": 2066 + }, + { + "epoch": 0.11, + "learning_rate": 3.9367148276367514e-07, + "loss": 0.1454, + "step": 2067 + }, + { + "epoch": 0.11, + "learning_rate": 3.936630417535082e-07, + "loss": 0.1716, + "step": 2068 + }, + { + "epoch": 0.11, + "learning_rate": 3.9365459520837675e-07, + "loss": 0.1672, + "step": 2069 + }, + { + "epoch": 0.11, + "learning_rate": 3.936461431285222e-07, + "loss": 0.1385, + "step": 2070 + }, + { + "epoch": 0.11, + "learning_rate": 3.9363768551418615e-07, + "loss": 0.1511, + "step": 2071 + }, + { + "epoch": 0.11, + "learning_rate": 3.9362922236561033e-07, + "loss": 0.1518, + "step": 2072 + }, + { + "epoch": 0.11, + "learning_rate": 3.9362075368303655e-07, + "loss": 0.1357, + "step": 2073 + }, + { + "epoch": 0.11, + "learning_rate": 3.9361227946670684e-07, + "loss": 0.1431, + "step": 2074 + }, + { + "epoch": 0.11, + "learning_rate": 3.936037997168635e-07, + "loss": 0.1163, + "step": 2075 + }, + { + "epoch": 0.11, + "learning_rate": 3.935953144337488e-07, + "loss": 0.1358, + "step": 2076 + }, + { + "epoch": 0.11, + "learning_rate": 3.9358682361760527e-07, + "loss": 0.1477, + "step": 2077 + }, + { + "epoch": 0.11, + "learning_rate": 3.9357832726867565e-07, + "loss": 0.1637, + "step": 2078 + }, + { + "epoch": 0.11, + "learning_rate": 3.935698253872027e-07, + "loss": 0.1778, + "step": 2079 + }, + { + "epoch": 0.11, + "learning_rate": 3.935613179734293e-07, + "loss": 0.1826, + "step": 2080 + }, + { + "epoch": 0.11, + "learning_rate": 3.935528050275988e-07, + "loss": 0.1787, + "step": 2081 + }, + { + "epoch": 0.11, + "learning_rate": 3.935442865499544e-07, + "loss": 0.1229, + "step": 2082 + }, + { + "epoch": 0.11, + "learning_rate": 3.935357625407396e-07, + "loss": 0.2088, + "step": 2083 + }, + { + "epoch": 0.11, + "learning_rate": 3.9352723300019797e-07, + "loss": 0.1363, + "step": 2084 + }, + { + "epoch": 0.11, + "learning_rate": 3.935186979285733e-07, + "loss": 0.1292, + "step": 2085 + }, + { + "epoch": 0.11, + "learning_rate": 3.9351015732610956e-07, + "loss": 0.1481, + "step": 2086 + }, + { + "epoch": 0.11, + "learning_rate": 3.935016111930508e-07, + "loss": 0.1662, + "step": 2087 + }, + { + "epoch": 0.11, + "learning_rate": 3.9349305952964133e-07, + "loss": 0.1488, + "step": 2088 + }, + { + "epoch": 0.11, + "learning_rate": 3.934845023361255e-07, + "loss": 0.1837, + "step": 2089 + }, + { + "epoch": 0.11, + "learning_rate": 3.9347593961274786e-07, + "loss": 0.1534, + "step": 2090 + }, + { + "epoch": 0.11, + "learning_rate": 3.9346737135975324e-07, + "loss": 0.144, + "step": 2091 + }, + { + "epoch": 0.11, + "learning_rate": 3.934587975773865e-07, + "loss": 0.1731, + "step": 2092 + }, + { + "epoch": 0.11, + "learning_rate": 3.9345021826589255e-07, + "loss": 0.129, + "step": 2093 + }, + { + "epoch": 0.11, + "learning_rate": 3.934416334255167e-07, + "loss": 0.1505, + "step": 2094 + }, + { + "epoch": 0.11, + "learning_rate": 3.934330430565043e-07, + "loss": 0.1644, + "step": 2095 + }, + { + "epoch": 0.11, + "learning_rate": 3.934244471591008e-07, + "loss": 0.1494, + "step": 2096 + }, + { + "epoch": 0.11, + "learning_rate": 3.9341584573355203e-07, + "loss": 0.1186, + "step": 2097 + }, + { + "epoch": 0.11, + "learning_rate": 3.934072387801037e-07, + "loss": 0.1498, + "step": 2098 + }, + { + "epoch": 0.11, + "learning_rate": 3.933986262990018e-07, + "loss": 0.1391, + "step": 2099 + }, + { + "epoch": 0.11, + "learning_rate": 3.9339000829049247e-07, + "loss": 0.147, + "step": 2100 + }, + { + "epoch": 0.11, + "learning_rate": 3.933813847548221e-07, + "loss": 0.1524, + "step": 2101 + }, + { + "epoch": 0.11, + "learning_rate": 3.933727556922371e-07, + "loss": 0.166, + "step": 2102 + }, + { + "epoch": 0.11, + "learning_rate": 3.933641211029841e-07, + "loss": 0.1392, + "step": 2103 + }, + { + "epoch": 0.11, + "learning_rate": 3.933554809873098e-07, + "loss": 0.1642, + "step": 2104 + }, + { + "epoch": 0.11, + "learning_rate": 3.933468353454613e-07, + "loss": 0.1628, + "step": 2105 + }, + { + "epoch": 0.11, + "learning_rate": 3.9333818417768557e-07, + "loss": 0.1399, + "step": 2106 + }, + { + "epoch": 0.11, + "learning_rate": 3.933295274842299e-07, + "loss": 0.1489, + "step": 2107 + }, + { + "epoch": 0.11, + "learning_rate": 3.933208652653417e-07, + "loss": 0.1283, + "step": 2108 + }, + { + "epoch": 0.11, + "learning_rate": 3.933121975212685e-07, + "loss": 0.1599, + "step": 2109 + }, + { + "epoch": 0.11, + "learning_rate": 3.9330352425225814e-07, + "loss": 0.1536, + "step": 2110 + }, + { + "epoch": 0.11, + "learning_rate": 3.932948454585584e-07, + "loss": 0.1366, + "step": 2111 + }, + { + "epoch": 0.11, + "learning_rate": 3.9328616114041736e-07, + "loss": 0.1584, + "step": 2112 + }, + { + "epoch": 0.11, + "learning_rate": 3.9327747129808325e-07, + "loss": 0.135, + "step": 2113 + }, + { + "epoch": 0.11, + "learning_rate": 3.9326877593180435e-07, + "loss": 0.1677, + "step": 2114 + }, + { + "epoch": 0.11, + "learning_rate": 3.9326007504182926e-07, + "loss": 0.1517, + "step": 2115 + }, + { + "epoch": 0.11, + "learning_rate": 3.932513686284066e-07, + "loss": 0.1505, + "step": 2116 + }, + { + "epoch": 0.11, + "learning_rate": 3.932426566917852e-07, + "loss": 0.1566, + "step": 2117 + }, + { + "epoch": 0.11, + "learning_rate": 3.932339392322141e-07, + "loss": 0.1294, + "step": 2118 + }, + { + "epoch": 0.11, + "learning_rate": 3.932252162499425e-07, + "loss": 0.1626, + "step": 2119 + }, + { + "epoch": 0.11, + "learning_rate": 3.9321648774521947e-07, + "loss": 0.1611, + "step": 2120 + }, + { + "epoch": 0.11, + "learning_rate": 3.9320775371829475e-07, + "loss": 0.1425, + "step": 2121 + }, + { + "epoch": 0.11, + "learning_rate": 3.931990141694178e-07, + "loss": 0.1851, + "step": 2122 + }, + { + "epoch": 0.11, + "learning_rate": 3.931902690988384e-07, + "loss": 0.1394, + "step": 2123 + }, + { + "epoch": 0.11, + "learning_rate": 3.931815185068066e-07, + "loss": 0.1573, + "step": 2124 + }, + { + "epoch": 0.11, + "learning_rate": 3.9317276239357236e-07, + "loss": 0.1461, + "step": 2125 + }, + { + "epoch": 0.11, + "learning_rate": 3.93164000759386e-07, + "loss": 0.1376, + "step": 2126 + }, + { + "epoch": 0.11, + "learning_rate": 3.93155233604498e-07, + "loss": 0.1621, + "step": 2127 + }, + { + "epoch": 0.11, + "learning_rate": 3.9314646092915885e-07, + "loss": 0.1327, + "step": 2128 + }, + { + "epoch": 0.11, + "learning_rate": 3.9313768273361925e-07, + "loss": 0.1733, + "step": 2129 + }, + { + "epoch": 0.11, + "learning_rate": 3.931288990181301e-07, + "loss": 0.1213, + "step": 2130 + }, + { + "epoch": 0.11, + "learning_rate": 3.931201097829425e-07, + "loss": 0.1475, + "step": 2131 + }, + { + "epoch": 0.11, + "learning_rate": 3.9311131502830764e-07, + "loss": 0.1572, + "step": 2132 + }, + { + "epoch": 0.11, + "learning_rate": 3.9310251475447676e-07, + "loss": 0.1639, + "step": 2133 + }, + { + "epoch": 0.11, + "learning_rate": 3.930937089617015e-07, + "loss": 0.1286, + "step": 2134 + }, + { + "epoch": 0.11, + "learning_rate": 3.930848976502335e-07, + "loss": 0.1556, + "step": 2135 + }, + { + "epoch": 0.11, + "learning_rate": 3.930760808203246e-07, + "loss": 0.189, + "step": 2136 + }, + { + "epoch": 0.11, + "learning_rate": 3.930672584722268e-07, + "loss": 0.1423, + "step": 2137 + }, + { + "epoch": 0.11, + "learning_rate": 3.930584306061922e-07, + "loss": 0.1384, + "step": 2138 + }, + { + "epoch": 0.11, + "learning_rate": 3.930495972224731e-07, + "loss": 0.1323, + "step": 2139 + }, + { + "epoch": 0.11, + "learning_rate": 3.93040758321322e-07, + "loss": 0.1425, + "step": 2140 + }, + { + "epoch": 0.11, + "learning_rate": 3.930319139029915e-07, + "loss": 0.1929, + "step": 2141 + }, + { + "epoch": 0.11, + "learning_rate": 3.9302306396773435e-07, + "loss": 0.1278, + "step": 2142 + }, + { + "epoch": 0.11, + "learning_rate": 3.930142085158036e-07, + "loss": 0.1527, + "step": 2143 + }, + { + "epoch": 0.11, + "learning_rate": 3.930053475474522e-07, + "loss": 0.1623, + "step": 2144 + }, + { + "epoch": 0.11, + "learning_rate": 3.929964810629335e-07, + "loss": 0.171, + "step": 2145 + }, + { + "epoch": 0.11, + "learning_rate": 3.9298760906250085e-07, + "loss": 0.1304, + "step": 2146 + }, + { + "epoch": 0.11, + "learning_rate": 3.929787315464078e-07, + "loss": 0.1665, + "step": 2147 + }, + { + "epoch": 0.11, + "learning_rate": 3.929698485149082e-07, + "loss": 0.1472, + "step": 2148 + }, + { + "epoch": 0.11, + "learning_rate": 3.929609599682557e-07, + "loss": 0.1329, + "step": 2149 + }, + { + "epoch": 0.11, + "learning_rate": 3.929520659067045e-07, + "loss": 0.1476, + "step": 2150 + }, + { + "epoch": 0.11, + "learning_rate": 3.929431663305089e-07, + "loss": 0.174, + "step": 2151 + }, + { + "epoch": 0.11, + "learning_rate": 3.92934261239923e-07, + "loss": 0.1316, + "step": 2152 + }, + { + "epoch": 0.11, + "learning_rate": 3.929253506352014e-07, + "loss": 0.1659, + "step": 2153 + }, + { + "epoch": 0.11, + "learning_rate": 3.929164345165989e-07, + "loss": 0.1082, + "step": 2154 + }, + { + "epoch": 0.11, + "learning_rate": 3.9290751288437023e-07, + "loss": 0.1461, + "step": 2155 + }, + { + "epoch": 0.11, + "learning_rate": 3.9289858573877036e-07, + "loss": 0.1239, + "step": 2156 + }, + { + "epoch": 0.11, + "learning_rate": 3.928896530800544e-07, + "loss": 0.1706, + "step": 2157 + }, + { + "epoch": 0.11, + "learning_rate": 3.9288071490847773e-07, + "loss": 0.1407, + "step": 2158 + }, + { + "epoch": 0.11, + "learning_rate": 3.9287177122429577e-07, + "loss": 0.156, + "step": 2159 + }, + { + "epoch": 0.11, + "learning_rate": 3.928628220277641e-07, + "loss": 0.1419, + "step": 2160 + }, + { + "epoch": 0.11, + "learning_rate": 3.9285386731913853e-07, + "loss": 0.1715, + "step": 2161 + }, + { + "epoch": 0.11, + "learning_rate": 3.92844907098675e-07, + "loss": 0.1485, + "step": 2162 + }, + { + "epoch": 0.11, + "learning_rate": 3.9283594136662954e-07, + "loss": 0.1846, + "step": 2163 + }, + { + "epoch": 0.11, + "learning_rate": 3.928269701232585e-07, + "loss": 0.1849, + "step": 2164 + }, + { + "epoch": 0.11, + "learning_rate": 3.928179933688182e-07, + "loss": 0.1242, + "step": 2165 + }, + { + "epoch": 0.11, + "learning_rate": 3.928090111035652e-07, + "loss": 0.1836, + "step": 2166 + }, + { + "epoch": 0.11, + "learning_rate": 3.9280002332775617e-07, + "loss": 0.1733, + "step": 2167 + }, + { + "epoch": 0.11, + "learning_rate": 3.927910300416481e-07, + "loss": 0.1542, + "step": 2168 + }, + { + "epoch": 0.11, + "learning_rate": 3.9278203124549795e-07, + "loss": 0.135, + "step": 2169 + }, + { + "epoch": 0.11, + "learning_rate": 3.9277302693956294e-07, + "loss": 0.1511, + "step": 2170 + }, + { + "epoch": 0.11, + "learning_rate": 3.9276401712410043e-07, + "loss": 0.1678, + "step": 2171 + }, + { + "epoch": 0.11, + "learning_rate": 3.9275500179936786e-07, + "loss": 0.1175, + "step": 2172 + }, + { + "epoch": 0.11, + "learning_rate": 3.9274598096562296e-07, + "loss": 0.1555, + "step": 2173 + }, + { + "epoch": 0.11, + "learning_rate": 3.9273695462312346e-07, + "loss": 0.1647, + "step": 2174 + }, + { + "epoch": 0.11, + "learning_rate": 3.9272792277212747e-07, + "loss": 0.192, + "step": 2175 + }, + { + "epoch": 0.11, + "learning_rate": 3.92718885412893e-07, + "loss": 0.1565, + "step": 2176 + }, + { + "epoch": 0.11, + "learning_rate": 3.927098425456784e-07, + "loss": 0.1313, + "step": 2177 + }, + { + "epoch": 0.11, + "learning_rate": 3.927007941707421e-07, + "loss": 0.1149, + "step": 2178 + }, + { + "epoch": 0.11, + "learning_rate": 3.9269174028834274e-07, + "loss": 0.1267, + "step": 2179 + }, + { + "epoch": 0.11, + "learning_rate": 3.92682680898739e-07, + "loss": 0.1474, + "step": 2180 + }, + { + "epoch": 0.11, + "learning_rate": 3.9267361600218987e-07, + "loss": 0.1217, + "step": 2181 + }, + { + "epoch": 0.11, + "learning_rate": 3.926645455989544e-07, + "loss": 0.1155, + "step": 2182 + }, + { + "epoch": 0.11, + "learning_rate": 3.926554696892919e-07, + "loss": 0.1613, + "step": 2183 + }, + { + "epoch": 0.11, + "learning_rate": 3.9264638827346166e-07, + "loss": 0.1493, + "step": 2184 + }, + { + "epoch": 0.11, + "learning_rate": 3.9263730135172334e-07, + "loss": 0.151, + "step": 2185 + }, + { + "epoch": 0.11, + "learning_rate": 3.926282089243365e-07, + "loss": 0.1679, + "step": 2186 + }, + { + "epoch": 0.11, + "learning_rate": 3.9261911099156115e-07, + "loss": 0.139, + "step": 2187 + }, + { + "epoch": 0.11, + "learning_rate": 3.9261000755365725e-07, + "loss": 0.1066, + "step": 2188 + }, + { + "epoch": 0.11, + "learning_rate": 3.926008986108849e-07, + "loss": 0.1518, + "step": 2189 + }, + { + "epoch": 0.11, + "learning_rate": 3.925917841635046e-07, + "loss": 0.1448, + "step": 2190 + }, + { + "epoch": 0.11, + "learning_rate": 3.925826642117767e-07, + "loss": 0.123, + "step": 2191 + }, + { + "epoch": 0.11, + "learning_rate": 3.925735387559619e-07, + "loss": 0.1476, + "step": 2192 + }, + { + "epoch": 0.11, + "learning_rate": 3.925644077963211e-07, + "loss": 0.1265, + "step": 2193 + }, + { + "epoch": 0.11, + "learning_rate": 3.925552713331151e-07, + "loss": 0.1553, + "step": 2194 + }, + { + "epoch": 0.11, + "learning_rate": 3.9254612936660517e-07, + "loss": 0.1546, + "step": 2195 + }, + { + "epoch": 0.11, + "learning_rate": 3.925369818970525e-07, + "loss": 0.1609, + "step": 2196 + }, + { + "epoch": 0.11, + "learning_rate": 3.925278289247186e-07, + "loss": 0.1489, + "step": 2197 + }, + { + "epoch": 0.11, + "learning_rate": 3.9251867044986504e-07, + "loss": 0.2105, + "step": 2198 + }, + { + "epoch": 0.11, + "learning_rate": 3.925095064727535e-07, + "loss": 0.1551, + "step": 2199 + }, + { + "epoch": 0.11, + "learning_rate": 3.92500336993646e-07, + "loss": 0.1354, + "step": 2200 + }, + { + "epoch": 0.11, + "learning_rate": 3.924911620128045e-07, + "loss": 0.1529, + "step": 2201 + }, + { + "epoch": 0.11, + "learning_rate": 3.9248198153049135e-07, + "loss": 0.1398, + "step": 2202 + }, + { + "epoch": 0.11, + "learning_rate": 3.924727955469688e-07, + "loss": 0.1665, + "step": 2203 + }, + { + "epoch": 0.12, + "learning_rate": 3.9246360406249944e-07, + "loss": 0.157, + "step": 2204 + }, + { + "epoch": 0.12, + "learning_rate": 3.92454407077346e-07, + "loss": 0.1684, + "step": 2205 + }, + { + "epoch": 0.12, + "learning_rate": 3.9244520459177133e-07, + "loss": 0.1308, + "step": 2206 + }, + { + "epoch": 0.12, + "learning_rate": 3.9243599660603843e-07, + "loss": 0.1342, + "step": 2207 + }, + { + "epoch": 0.12, + "learning_rate": 3.924267831204105e-07, + "loss": 0.166, + "step": 2208 + }, + { + "epoch": 0.12, + "learning_rate": 3.9241756413515074e-07, + "loss": 0.1536, + "step": 2209 + }, + { + "epoch": 0.12, + "learning_rate": 3.924083396505227e-07, + "loss": 0.1545, + "step": 2210 + }, + { + "epoch": 0.12, + "learning_rate": 3.923991096667901e-07, + "loss": 0.1407, + "step": 2211 + }, + { + "epoch": 0.12, + "learning_rate": 3.9238987418421664e-07, + "loss": 0.1204, + "step": 2212 + }, + { + "epoch": 0.12, + "learning_rate": 3.923806332030663e-07, + "loss": 0.1478, + "step": 2213 + }, + { + "epoch": 0.12, + "learning_rate": 3.923713867236032e-07, + "loss": 0.1672, + "step": 2214 + }, + { + "epoch": 0.12, + "learning_rate": 3.9236213474609167e-07, + "loss": 0.1327, + "step": 2215 + }, + { + "epoch": 0.12, + "learning_rate": 3.92352877270796e-07, + "loss": 0.1242, + "step": 2216 + }, + { + "epoch": 0.12, + "learning_rate": 3.923436142979809e-07, + "loss": 0.1247, + "step": 2217 + }, + { + "epoch": 0.12, + "learning_rate": 3.9233434582791097e-07, + "loss": 0.1403, + "step": 2218 + }, + { + "epoch": 0.12, + "learning_rate": 3.9232507186085126e-07, + "loss": 0.125, + "step": 2219 + }, + { + "epoch": 0.12, + "learning_rate": 3.923157923970667e-07, + "loss": 0.1597, + "step": 2220 + }, + { + "epoch": 0.12, + "learning_rate": 3.923065074368226e-07, + "loss": 0.1707, + "step": 2221 + }, + { + "epoch": 0.12, + "learning_rate": 3.9229721698038425e-07, + "loss": 0.1661, + "step": 2222 + }, + { + "epoch": 0.12, + "learning_rate": 3.9228792102801715e-07, + "loss": 0.1565, + "step": 2223 + }, + { + "epoch": 0.12, + "learning_rate": 3.9227861957998716e-07, + "loss": 0.1368, + "step": 2224 + }, + { + "epoch": 0.12, + "learning_rate": 3.922693126365599e-07, + "loss": 0.2002, + "step": 2225 + }, + { + "epoch": 0.12, + "learning_rate": 3.922600001980015e-07, + "loss": 0.1407, + "step": 2226 + }, + { + "epoch": 0.12, + "learning_rate": 3.9225068226457807e-07, + "loss": 0.1691, + "step": 2227 + }, + { + "epoch": 0.12, + "learning_rate": 3.922413588365559e-07, + "loss": 0.1741, + "step": 2228 + }, + { + "epoch": 0.12, + "learning_rate": 3.922320299142015e-07, + "loss": 0.1516, + "step": 2229 + }, + { + "epoch": 0.12, + "learning_rate": 3.922226954977815e-07, + "loss": 0.1925, + "step": 2230 + }, + { + "epoch": 0.12, + "learning_rate": 3.9221335558756265e-07, + "loss": 0.1487, + "step": 2231 + }, + { + "epoch": 0.12, + "learning_rate": 3.922040101838119e-07, + "loss": 0.1499, + "step": 2232 + }, + { + "epoch": 0.12, + "learning_rate": 3.921946592867963e-07, + "loss": 0.1305, + "step": 2233 + }, + { + "epoch": 0.12, + "learning_rate": 3.9218530289678313e-07, + "loss": 0.155, + "step": 2234 + }, + { + "epoch": 0.12, + "learning_rate": 3.9217594101403984e-07, + "loss": 0.1288, + "step": 2235 + }, + { + "epoch": 0.12, + "learning_rate": 3.92166573638834e-07, + "loss": 0.166, + "step": 2236 + }, + { + "epoch": 0.12, + "learning_rate": 3.9215720077143324e-07, + "loss": 0.1179, + "step": 2237 + }, + { + "epoch": 0.12, + "learning_rate": 3.921478224121056e-07, + "loss": 0.1372, + "step": 2238 + }, + { + "epoch": 0.12, + "learning_rate": 3.9213843856111895e-07, + "loss": 0.1304, + "step": 2239 + }, + { + "epoch": 0.12, + "learning_rate": 3.921290492187415e-07, + "loss": 0.1308, + "step": 2240 + }, + { + "epoch": 0.12, + "learning_rate": 3.921196543852417e-07, + "loss": 0.1744, + "step": 2241 + }, + { + "epoch": 0.12, + "learning_rate": 3.9211025406088807e-07, + "loss": 0.1287, + "step": 2242 + }, + { + "epoch": 0.12, + "learning_rate": 3.921008482459492e-07, + "loss": 0.1228, + "step": 2243 + }, + { + "epoch": 0.12, + "learning_rate": 3.920914369406939e-07, + "loss": 0.1564, + "step": 2244 + }, + { + "epoch": 0.12, + "learning_rate": 3.9208202014539116e-07, + "loss": 0.1715, + "step": 2245 + }, + { + "epoch": 0.12, + "learning_rate": 3.920725978603102e-07, + "loss": 0.1602, + "step": 2246 + }, + { + "epoch": 0.12, + "learning_rate": 3.9206317008572015e-07, + "loss": 0.153, + "step": 2247 + }, + { + "epoch": 0.12, + "learning_rate": 3.9205373682189065e-07, + "loss": 0.1284, + "step": 2248 + }, + { + "epoch": 0.12, + "learning_rate": 3.920442980690911e-07, + "loss": 0.1554, + "step": 2249 + }, + { + "epoch": 0.12, + "learning_rate": 3.9203485382759145e-07, + "loss": 0.1588, + "step": 2250 + }, + { + "epoch": 0.12, + "learning_rate": 3.920254040976616e-07, + "loss": 0.1227, + "step": 2251 + }, + { + "epoch": 0.12, + "learning_rate": 3.9201594887957146e-07, + "loss": 0.1459, + "step": 2252 + }, + { + "epoch": 0.12, + "learning_rate": 3.9200648817359145e-07, + "loss": 0.1183, + "step": 2253 + }, + { + "epoch": 0.12, + "learning_rate": 3.9199702197999177e-07, + "loss": 0.1653, + "step": 2254 + }, + { + "epoch": 0.12, + "learning_rate": 3.9198755029904325e-07, + "loss": 0.1547, + "step": 2255 + }, + { + "epoch": 0.12, + "learning_rate": 3.919780731310163e-07, + "loss": 0.1628, + "step": 2256 + }, + { + "epoch": 0.12, + "learning_rate": 3.9196859047618197e-07, + "loss": 0.1558, + "step": 2257 + }, + { + "epoch": 0.12, + "learning_rate": 3.919591023348112e-07, + "loss": 0.1233, + "step": 2258 + }, + { + "epoch": 0.12, + "learning_rate": 3.919496087071752e-07, + "loss": 0.1549, + "step": 2259 + }, + { + "epoch": 0.12, + "learning_rate": 3.919401095935453e-07, + "loss": 0.1671, + "step": 2260 + }, + { + "epoch": 0.12, + "learning_rate": 3.919306049941929e-07, + "loss": 0.1428, + "step": 2261 + }, + { + "epoch": 0.12, + "learning_rate": 3.9192109490938974e-07, + "loss": 0.1388, + "step": 2262 + }, + { + "epoch": 0.12, + "learning_rate": 3.919115793394076e-07, + "loss": 0.152, + "step": 2263 + }, + { + "epoch": 0.12, + "learning_rate": 3.919020582845185e-07, + "loss": 0.14, + "step": 2264 + }, + { + "epoch": 0.12, + "learning_rate": 3.918925317449944e-07, + "loss": 0.1367, + "step": 2265 + }, + { + "epoch": 0.12, + "learning_rate": 3.9188299972110775e-07, + "loss": 0.1344, + "step": 2266 + }, + { + "epoch": 0.12, + "learning_rate": 3.918734622131309e-07, + "loss": 0.1707, + "step": 2267 + }, + { + "epoch": 0.12, + "learning_rate": 3.918639192213363e-07, + "loss": 0.1426, + "step": 2268 + }, + { + "epoch": 0.12, + "learning_rate": 3.9185437074599693e-07, + "loss": 0.1465, + "step": 2269 + }, + { + "epoch": 0.12, + "learning_rate": 3.9184481678738557e-07, + "loss": 0.1173, + "step": 2270 + }, + { + "epoch": 0.12, + "learning_rate": 3.9183525734577525e-07, + "loss": 0.1417, + "step": 2271 + }, + { + "epoch": 0.12, + "learning_rate": 3.9182569242143926e-07, + "loss": 0.1532, + "step": 2272 + }, + { + "epoch": 0.12, + "learning_rate": 3.9181612201465096e-07, + "loss": 0.1143, + "step": 2273 + }, + { + "epoch": 0.12, + "learning_rate": 3.9180654612568373e-07, + "loss": 0.1597, + "step": 2274 + }, + { + "epoch": 0.12, + "learning_rate": 3.9179696475481144e-07, + "loss": 0.1562, + "step": 2275 + }, + { + "epoch": 0.12, + "learning_rate": 3.9178737790230783e-07, + "loss": 0.1691, + "step": 2276 + }, + { + "epoch": 0.12, + "learning_rate": 3.91777785568447e-07, + "loss": 0.1369, + "step": 2277 + }, + { + "epoch": 0.12, + "learning_rate": 3.9176818775350294e-07, + "loss": 0.1195, + "step": 2278 + }, + { + "epoch": 0.12, + "learning_rate": 3.9175858445775006e-07, + "loss": 0.1459, + "step": 2279 + }, + { + "epoch": 0.12, + "learning_rate": 3.917489756814628e-07, + "loss": 0.1304, + "step": 2280 + }, + { + "epoch": 0.12, + "learning_rate": 3.917393614249158e-07, + "loss": 0.1636, + "step": 2281 + }, + { + "epoch": 0.12, + "learning_rate": 3.917297416883839e-07, + "loss": 0.1164, + "step": 2282 + }, + { + "epoch": 0.12, + "learning_rate": 3.917201164721419e-07, + "loss": 0.1518, + "step": 2283 + }, + { + "epoch": 0.12, + "learning_rate": 3.9171048577646496e-07, + "loss": 0.1446, + "step": 2284 + }, + { + "epoch": 0.12, + "learning_rate": 3.917008496016283e-07, + "loss": 0.1523, + "step": 2285 + }, + { + "epoch": 0.12, + "learning_rate": 3.9169120794790744e-07, + "loss": 0.1856, + "step": 2286 + }, + { + "epoch": 0.12, + "learning_rate": 3.9168156081557776e-07, + "loss": 0.1171, + "step": 2287 + }, + { + "epoch": 0.12, + "learning_rate": 3.9167190820491516e-07, + "loss": 0.1304, + "step": 2288 + }, + { + "epoch": 0.12, + "learning_rate": 3.9166225011619535e-07, + "loss": 0.1413, + "step": 2289 + }, + { + "epoch": 0.12, + "learning_rate": 3.916525865496945e-07, + "loss": 0.1629, + "step": 2290 + }, + { + "epoch": 0.12, + "learning_rate": 3.9164291750568875e-07, + "loss": 0.1335, + "step": 2291 + }, + { + "epoch": 0.12, + "learning_rate": 3.9163324298445444e-07, + "loss": 0.1724, + "step": 2292 + }, + { + "epoch": 0.12, + "learning_rate": 3.91623562986268e-07, + "loss": 0.1562, + "step": 2293 + }, + { + "epoch": 0.12, + "learning_rate": 3.9161387751140623e-07, + "loss": 0.1385, + "step": 2294 + }, + { + "epoch": 0.12, + "learning_rate": 3.9160418656014583e-07, + "loss": 0.167, + "step": 2295 + }, + { + "epoch": 0.12, + "learning_rate": 3.9159449013276385e-07, + "loss": 0.1528, + "step": 2296 + }, + { + "epoch": 0.12, + "learning_rate": 3.9158478822953727e-07, + "loss": 0.1761, + "step": 2297 + }, + { + "epoch": 0.12, + "learning_rate": 3.915750808507436e-07, + "loss": 0.1743, + "step": 2298 + }, + { + "epoch": 0.12, + "learning_rate": 3.9156536799666013e-07, + "loss": 0.1436, + "step": 2299 + }, + { + "epoch": 0.12, + "learning_rate": 3.915556496675645e-07, + "loss": 0.1644, + "step": 2300 + }, + { + "epoch": 0.12, + "learning_rate": 3.9154592586373445e-07, + "loss": 0.1304, + "step": 2301 + }, + { + "epoch": 0.12, + "learning_rate": 3.915361965854479e-07, + "loss": 0.1398, + "step": 2302 + }, + { + "epoch": 0.12, + "learning_rate": 3.915264618329829e-07, + "loss": 0.1497, + "step": 2303 + }, + { + "epoch": 0.12, + "learning_rate": 3.9151672160661765e-07, + "loss": 0.147, + "step": 2304 + }, + { + "epoch": 0.12, + "learning_rate": 3.915069759066306e-07, + "loss": 0.1416, + "step": 2305 + }, + { + "epoch": 0.12, + "learning_rate": 3.9149722473330027e-07, + "loss": 0.1372, + "step": 2306 + }, + { + "epoch": 0.12, + "learning_rate": 3.9148746808690526e-07, + "loss": 0.1551, + "step": 2307 + }, + { + "epoch": 0.12, + "learning_rate": 3.9147770596772455e-07, + "loss": 0.1516, + "step": 2308 + }, + { + "epoch": 0.12, + "learning_rate": 3.914679383760371e-07, + "loss": 0.1458, + "step": 2309 + }, + { + "epoch": 0.12, + "learning_rate": 3.91458165312122e-07, + "loss": 0.1291, + "step": 2310 + }, + { + "epoch": 0.12, + "learning_rate": 3.914483867762587e-07, + "loss": 0.1856, + "step": 2311 + }, + { + "epoch": 0.12, + "learning_rate": 3.914386027687265e-07, + "loss": 0.1454, + "step": 2312 + }, + { + "epoch": 0.12, + "learning_rate": 3.914288132898052e-07, + "loss": 0.1465, + "step": 2313 + }, + { + "epoch": 0.12, + "learning_rate": 3.9141901833977454e-07, + "loss": 0.1273, + "step": 2314 + }, + { + "epoch": 0.12, + "learning_rate": 3.9140921791891433e-07, + "loss": 0.1557, + "step": 2315 + }, + { + "epoch": 0.12, + "learning_rate": 3.913994120275049e-07, + "loss": 0.1534, + "step": 2316 + }, + { + "epoch": 0.12, + "learning_rate": 3.9138960066582633e-07, + "loss": 0.1306, + "step": 2317 + }, + { + "epoch": 0.12, + "learning_rate": 3.913797838341591e-07, + "loss": 0.1496, + "step": 2318 + }, + { + "epoch": 0.12, + "learning_rate": 3.913699615327838e-07, + "loss": 0.1673, + "step": 2319 + }, + { + "epoch": 0.12, + "learning_rate": 3.913601337619811e-07, + "loss": 0.1768, + "step": 2320 + }, + { + "epoch": 0.12, + "learning_rate": 3.9135030052203185e-07, + "loss": 0.1287, + "step": 2321 + }, + { + "epoch": 0.12, + "learning_rate": 3.9134046181321716e-07, + "loss": 0.1397, + "step": 2322 + }, + { + "epoch": 0.12, + "learning_rate": 3.9133061763581825e-07, + "loss": 0.1567, + "step": 2323 + }, + { + "epoch": 0.12, + "learning_rate": 3.913207679901164e-07, + "loss": 0.1291, + "step": 2324 + }, + { + "epoch": 0.12, + "learning_rate": 3.9131091287639323e-07, + "loss": 0.1512, + "step": 2325 + }, + { + "epoch": 0.12, + "learning_rate": 3.913010522949302e-07, + "loss": 0.1811, + "step": 2326 + }, + { + "epoch": 0.12, + "learning_rate": 3.9129118624600925e-07, + "loss": 0.1823, + "step": 2327 + }, + { + "epoch": 0.12, + "learning_rate": 3.912813147299125e-07, + "loss": 0.1011, + "step": 2328 + }, + { + "epoch": 0.12, + "learning_rate": 3.9127143774692176e-07, + "loss": 0.2226, + "step": 2329 + }, + { + "epoch": 0.12, + "learning_rate": 3.912615552973196e-07, + "loss": 0.1573, + "step": 2330 + }, + { + "epoch": 0.12, + "learning_rate": 3.912516673813883e-07, + "loss": 0.1351, + "step": 2331 + }, + { + "epoch": 0.12, + "learning_rate": 3.9124177399941044e-07, + "loss": 0.1442, + "step": 2332 + }, + { + "epoch": 0.12, + "learning_rate": 3.9123187515166894e-07, + "loss": 0.1471, + "step": 2333 + }, + { + "epoch": 0.12, + "learning_rate": 3.9122197083844666e-07, + "loss": 0.1494, + "step": 2334 + }, + { + "epoch": 0.12, + "learning_rate": 3.912120610600266e-07, + "loss": 0.1224, + "step": 2335 + }, + { + "epoch": 0.12, + "learning_rate": 3.91202145816692e-07, + "loss": 0.1218, + "step": 2336 + }, + { + "epoch": 0.12, + "learning_rate": 3.911922251087263e-07, + "loss": 0.1335, + "step": 2337 + }, + { + "epoch": 0.12, + "learning_rate": 3.9118229893641293e-07, + "loss": 0.1353, + "step": 2338 + }, + { + "epoch": 0.12, + "learning_rate": 3.9117236730003573e-07, + "loss": 0.1661, + "step": 2339 + }, + { + "epoch": 0.12, + "learning_rate": 3.9116243019987843e-07, + "loss": 0.1244, + "step": 2340 + }, + { + "epoch": 0.12, + "learning_rate": 3.911524876362251e-07, + "loss": 0.144, + "step": 2341 + }, + { + "epoch": 0.12, + "learning_rate": 3.9114253960935984e-07, + "loss": 0.1265, + "step": 2342 + }, + { + "epoch": 0.12, + "learning_rate": 3.9113258611956706e-07, + "loss": 0.1707, + "step": 2343 + }, + { + "epoch": 0.12, + "learning_rate": 3.9112262716713117e-07, + "loss": 0.1642, + "step": 2344 + }, + { + "epoch": 0.12, + "learning_rate": 3.911126627523368e-07, + "loss": 0.1389, + "step": 2345 + }, + { + "epoch": 0.12, + "learning_rate": 3.9110269287546877e-07, + "loss": 0.1317, + "step": 2346 + }, + { + "epoch": 0.12, + "learning_rate": 3.91092717536812e-07, + "loss": 0.1372, + "step": 2347 + }, + { + "epoch": 0.12, + "learning_rate": 3.910827367366516e-07, + "loss": 0.15, + "step": 2348 + }, + { + "epoch": 0.12, + "learning_rate": 3.9107275047527277e-07, + "loss": 0.1335, + "step": 2349 + }, + { + "epoch": 0.12, + "learning_rate": 3.9106275875296105e-07, + "loss": 0.14, + "step": 2350 + }, + { + "epoch": 0.12, + "learning_rate": 3.9105276157000183e-07, + "loss": 0.1405, + "step": 2351 + }, + { + "epoch": 0.12, + "learning_rate": 3.9104275892668103e-07, + "loss": 0.1455, + "step": 2352 + }, + { + "epoch": 0.12, + "learning_rate": 3.9103275082328436e-07, + "loss": 0.1578, + "step": 2353 + }, + { + "epoch": 0.12, + "learning_rate": 3.91022737260098e-07, + "loss": 0.1677, + "step": 2354 + }, + { + "epoch": 0.12, + "learning_rate": 3.91012718237408e-07, + "loss": 0.142, + "step": 2355 + }, + { + "epoch": 0.12, + "learning_rate": 3.910026937555008e-07, + "loss": 0.1896, + "step": 2356 + }, + { + "epoch": 0.12, + "learning_rate": 3.9099266381466285e-07, + "loss": 0.1393, + "step": 2357 + }, + { + "epoch": 0.12, + "learning_rate": 3.9098262841518085e-07, + "loss": 0.1472, + "step": 2358 + }, + { + "epoch": 0.12, + "learning_rate": 3.909725875573416e-07, + "loss": 0.1479, + "step": 2359 + }, + { + "epoch": 0.12, + "learning_rate": 3.909625412414321e-07, + "loss": 0.1391, + "step": 2360 + }, + { + "epoch": 0.12, + "learning_rate": 3.9095248946773945e-07, + "loss": 0.1602, + "step": 2361 + }, + { + "epoch": 0.12, + "learning_rate": 3.909424322365509e-07, + "loss": 0.1569, + "step": 2362 + }, + { + "epoch": 0.12, + "learning_rate": 3.909323695481539e-07, + "loss": 0.1715, + "step": 2363 + }, + { + "epoch": 0.12, + "learning_rate": 3.9092230140283616e-07, + "loss": 0.1823, + "step": 2364 + }, + { + "epoch": 0.12, + "learning_rate": 3.909122278008853e-07, + "loss": 0.1466, + "step": 2365 + }, + { + "epoch": 0.12, + "learning_rate": 3.9090214874258923e-07, + "loss": 0.1352, + "step": 2366 + }, + { + "epoch": 0.12, + "learning_rate": 3.9089206422823615e-07, + "loss": 0.1541, + "step": 2367 + }, + { + "epoch": 0.12, + "learning_rate": 3.908819742581141e-07, + "loss": 0.1555, + "step": 2368 + }, + { + "epoch": 0.12, + "learning_rate": 3.9087187883251156e-07, + "loss": 0.1596, + "step": 2369 + }, + { + "epoch": 0.12, + "learning_rate": 3.9086177795171704e-07, + "loss": 0.179, + "step": 2370 + }, + { + "epoch": 0.12, + "learning_rate": 3.908516716160193e-07, + "loss": 0.1595, + "step": 2371 + }, + { + "epoch": 0.12, + "learning_rate": 3.90841559825707e-07, + "loss": 0.1343, + "step": 2372 + }, + { + "epoch": 0.12, + "learning_rate": 3.908314425810693e-07, + "loss": 0.1662, + "step": 2373 + }, + { + "epoch": 0.12, + "learning_rate": 3.9082131988239527e-07, + "loss": 0.1262, + "step": 2374 + }, + { + "epoch": 0.12, + "learning_rate": 3.9081119172997423e-07, + "loss": 0.156, + "step": 2375 + }, + { + "epoch": 0.12, + "learning_rate": 3.9080105812409573e-07, + "loss": 0.1418, + "step": 2376 + }, + { + "epoch": 0.12, + "learning_rate": 3.9079091906504934e-07, + "loss": 0.1307, + "step": 2377 + }, + { + "epoch": 0.12, + "learning_rate": 3.907807745531248e-07, + "loss": 0.1825, + "step": 2378 + }, + { + "epoch": 0.12, + "learning_rate": 3.9077062458861213e-07, + "loss": 0.1683, + "step": 2379 + }, + { + "epoch": 0.12, + "learning_rate": 3.9076046917180125e-07, + "loss": 0.1747, + "step": 2380 + }, + { + "epoch": 0.12, + "learning_rate": 3.907503083029826e-07, + "loss": 0.1618, + "step": 2381 + }, + { + "epoch": 0.12, + "learning_rate": 3.907401419824465e-07, + "loss": 0.1374, + "step": 2382 + }, + { + "epoch": 0.12, + "learning_rate": 3.907299702104835e-07, + "loss": 0.1421, + "step": 2383 + }, + { + "epoch": 0.12, + "learning_rate": 3.9071979298738433e-07, + "loss": 0.1022, + "step": 2384 + }, + { + "epoch": 0.12, + "learning_rate": 3.9070961031343983e-07, + "loss": 0.1351, + "step": 2385 + }, + { + "epoch": 0.12, + "learning_rate": 3.9069942218894107e-07, + "loss": 0.138, + "step": 2386 + }, + { + "epoch": 0.12, + "learning_rate": 3.906892286141792e-07, + "loss": 0.1317, + "step": 2387 + }, + { + "epoch": 0.12, + "learning_rate": 3.9067902958944554e-07, + "loss": 0.145, + "step": 2388 + }, + { + "epoch": 0.12, + "learning_rate": 3.9066882511503164e-07, + "loss": 0.1388, + "step": 2389 + }, + { + "epoch": 0.12, + "learning_rate": 3.9065861519122905e-07, + "loss": 0.1363, + "step": 2390 + }, + { + "epoch": 0.12, + "learning_rate": 3.906483998183297e-07, + "loss": 0.1352, + "step": 2391 + }, + { + "epoch": 0.12, + "learning_rate": 3.9063817899662544e-07, + "loss": 0.1556, + "step": 2392 + }, + { + "epoch": 0.12, + "learning_rate": 3.9062795272640845e-07, + "loss": 0.1616, + "step": 2393 + }, + { + "epoch": 0.12, + "learning_rate": 3.9061772100797103e-07, + "loss": 0.1559, + "step": 2394 + }, + { + "epoch": 0.13, + "learning_rate": 3.906074838416055e-07, + "loss": 0.1478, + "step": 2395 + }, + { + "epoch": 0.13, + "learning_rate": 3.9059724122760454e-07, + "loss": 0.1582, + "step": 2396 + }, + { + "epoch": 0.13, + "learning_rate": 3.9058699316626077e-07, + "loss": 0.1362, + "step": 2397 + }, + { + "epoch": 0.13, + "learning_rate": 3.905767396578672e-07, + "loss": 0.1549, + "step": 2398 + }, + { + "epoch": 0.13, + "learning_rate": 3.9056648070271693e-07, + "loss": 0.153, + "step": 2399 + }, + { + "epoch": 0.13, + "learning_rate": 3.9055621630110296e-07, + "loss": 0.1478, + "step": 2400 + }, + { + "epoch": 0.13, + "learning_rate": 3.905459464533188e-07, + "loss": 0.1532, + "step": 2401 + }, + { + "epoch": 0.13, + "learning_rate": 3.905356711596579e-07, + "loss": 0.1478, + "step": 2402 + }, + { + "epoch": 0.13, + "learning_rate": 3.9052539042041405e-07, + "loss": 0.1333, + "step": 2403 + }, + { + "epoch": 0.13, + "learning_rate": 3.905151042358809e-07, + "loss": 0.1376, + "step": 2404 + }, + { + "epoch": 0.13, + "learning_rate": 3.905048126063526e-07, + "loss": 0.145, + "step": 2405 + }, + { + "epoch": 0.13, + "learning_rate": 3.9049451553212316e-07, + "loss": 0.1567, + "step": 2406 + }, + { + "epoch": 0.13, + "learning_rate": 3.9048421301348696e-07, + "loss": 0.1423, + "step": 2407 + }, + { + "epoch": 0.13, + "learning_rate": 3.904739050507384e-07, + "loss": 0.1442, + "step": 2408 + }, + { + "epoch": 0.13, + "learning_rate": 3.904635916441721e-07, + "loss": 0.1256, + "step": 2409 + }, + { + "epoch": 0.13, + "learning_rate": 3.9045327279408287e-07, + "loss": 0.1411, + "step": 2410 + }, + { + "epoch": 0.13, + "learning_rate": 3.9044294850076556e-07, + "loss": 0.1424, + "step": 2411 + }, + { + "epoch": 0.13, + "learning_rate": 3.904326187645152e-07, + "loss": 0.1764, + "step": 2412 + }, + { + "epoch": 0.13, + "learning_rate": 3.9042228358562714e-07, + "loss": 0.1646, + "step": 2413 + }, + { + "epoch": 0.13, + "learning_rate": 3.904119429643967e-07, + "loss": 0.1669, + "step": 2414 + }, + { + "epoch": 0.13, + "learning_rate": 3.904015969011194e-07, + "loss": 0.1218, + "step": 2415 + }, + { + "epoch": 0.13, + "learning_rate": 3.90391245396091e-07, + "loss": 0.1271, + "step": 2416 + }, + { + "epoch": 0.13, + "learning_rate": 3.903808884496073e-07, + "loss": 0.1335, + "step": 2417 + }, + { + "epoch": 0.13, + "learning_rate": 3.9037052606196424e-07, + "loss": 0.1711, + "step": 2418 + }, + { + "epoch": 0.13, + "learning_rate": 3.9036015823345816e-07, + "loss": 0.1236, + "step": 2419 + }, + { + "epoch": 0.13, + "learning_rate": 3.9034978496438524e-07, + "loss": 0.1312, + "step": 2420 + }, + { + "epoch": 0.13, + "learning_rate": 3.9033940625504194e-07, + "loss": 0.1259, + "step": 2421 + }, + { + "epoch": 0.13, + "learning_rate": 3.90329022105725e-07, + "loss": 0.1726, + "step": 2422 + }, + { + "epoch": 0.13, + "learning_rate": 3.9031863251673105e-07, + "loss": 0.1588, + "step": 2423 + }, + { + "epoch": 0.13, + "learning_rate": 3.9030823748835713e-07, + "loss": 0.1496, + "step": 2424 + }, + { + "epoch": 0.13, + "learning_rate": 3.9029783702090036e-07, + "loss": 0.1501, + "step": 2425 + }, + { + "epoch": 0.13, + "learning_rate": 3.902874311146579e-07, + "loss": 0.1481, + "step": 2426 + }, + { + "epoch": 0.13, + "learning_rate": 3.902770197699273e-07, + "loss": 0.1527, + "step": 2427 + }, + { + "epoch": 0.13, + "learning_rate": 3.902666029870059e-07, + "loss": 0.1521, + "step": 2428 + }, + { + "epoch": 0.13, + "learning_rate": 3.902561807661916e-07, + "loss": 0.131, + "step": 2429 + }, + { + "epoch": 0.13, + "learning_rate": 3.902457531077822e-07, + "loss": 0.1343, + "step": 2430 + }, + { + "epoch": 0.13, + "learning_rate": 3.9023532001207567e-07, + "loss": 0.1686, + "step": 2431 + }, + { + "epoch": 0.13, + "learning_rate": 3.9022488147937034e-07, + "loss": 0.1405, + "step": 2432 + }, + { + "epoch": 0.13, + "learning_rate": 3.9021443750996444e-07, + "loss": 0.1331, + "step": 2433 + }, + { + "epoch": 0.13, + "learning_rate": 3.902039881041565e-07, + "loss": 0.1741, + "step": 2434 + }, + { + "epoch": 0.13, + "learning_rate": 3.901935332622451e-07, + "loss": 0.148, + "step": 2435 + }, + { + "epoch": 0.13, + "learning_rate": 3.901830729845292e-07, + "loss": 0.1626, + "step": 2436 + }, + { + "epoch": 0.13, + "learning_rate": 3.9017260727130755e-07, + "loss": 0.1548, + "step": 2437 + }, + { + "epoch": 0.13, + "learning_rate": 3.901621361228794e-07, + "loss": 0.1624, + "step": 2438 + }, + { + "epoch": 0.13, + "learning_rate": 3.90151659539544e-07, + "loss": 0.1645, + "step": 2439 + }, + { + "epoch": 0.13, + "learning_rate": 3.901411775216008e-07, + "loss": 0.1415, + "step": 2440 + }, + { + "epoch": 0.13, + "learning_rate": 3.9013069006934926e-07, + "loss": 0.1309, + "step": 2441 + }, + { + "epoch": 0.13, + "learning_rate": 3.901201971830892e-07, + "loss": 0.1485, + "step": 2442 + }, + { + "epoch": 0.13, + "learning_rate": 3.901096988631206e-07, + "loss": 0.1533, + "step": 2443 + }, + { + "epoch": 0.13, + "learning_rate": 3.9009919510974336e-07, + "loss": 0.1149, + "step": 2444 + }, + { + "epoch": 0.13, + "learning_rate": 3.9008868592325783e-07, + "loss": 0.1273, + "step": 2445 + }, + { + "epoch": 0.13, + "learning_rate": 3.9007817130396417e-07, + "loss": 0.1295, + "step": 2446 + }, + { + "epoch": 0.13, + "learning_rate": 3.9006765125216303e-07, + "loss": 0.1803, + "step": 2447 + }, + { + "epoch": 0.13, + "learning_rate": 3.90057125768155e-07, + "loss": 0.1769, + "step": 2448 + }, + { + "epoch": 0.13, + "learning_rate": 3.90046594852241e-07, + "loss": 0.1646, + "step": 2449 + }, + { + "epoch": 0.13, + "learning_rate": 3.9003605850472196e-07, + "loss": 0.157, + "step": 2450 + }, + { + "epoch": 0.13, + "learning_rate": 3.9002551672589897e-07, + "loss": 0.1615, + "step": 2451 + }, + { + "epoch": 0.13, + "learning_rate": 3.9001496951607336e-07, + "loss": 0.1516, + "step": 2452 + }, + { + "epoch": 0.13, + "learning_rate": 3.9000441687554657e-07, + "loss": 0.1066, + "step": 2453 + }, + { + "epoch": 0.13, + "learning_rate": 3.899938588046202e-07, + "loss": 0.1229, + "step": 2454 + }, + { + "epoch": 0.13, + "learning_rate": 3.8998329530359603e-07, + "loss": 0.1387, + "step": 2455 + }, + { + "epoch": 0.13, + "learning_rate": 3.899727263727759e-07, + "loss": 0.1253, + "step": 2456 + }, + { + "epoch": 0.13, + "learning_rate": 3.899621520124619e-07, + "loss": 0.1977, + "step": 2457 + }, + { + "epoch": 0.13, + "learning_rate": 3.8995157222295634e-07, + "loss": 0.1541, + "step": 2458 + }, + { + "epoch": 0.13, + "learning_rate": 3.899409870045614e-07, + "loss": 0.1813, + "step": 2459 + }, + { + "epoch": 0.13, + "learning_rate": 3.8993039635757983e-07, + "loss": 0.1711, + "step": 2460 + }, + { + "epoch": 0.13, + "learning_rate": 3.899198002823142e-07, + "loss": 0.1613, + "step": 2461 + }, + { + "epoch": 0.13, + "learning_rate": 3.899091987790673e-07, + "loss": 0.1315, + "step": 2462 + }, + { + "epoch": 0.13, + "learning_rate": 3.898985918481422e-07, + "loss": 0.1425, + "step": 2463 + }, + { + "epoch": 0.13, + "learning_rate": 3.8988797948984205e-07, + "loss": 0.1532, + "step": 2464 + }, + { + "epoch": 0.13, + "learning_rate": 3.8987736170447016e-07, + "loss": 0.1358, + "step": 2465 + }, + { + "epoch": 0.13, + "learning_rate": 3.8986673849232997e-07, + "loss": 0.1373, + "step": 2466 + }, + { + "epoch": 0.13, + "learning_rate": 3.898561098537251e-07, + "loss": 0.152, + "step": 2467 + }, + { + "epoch": 0.13, + "learning_rate": 3.898454757889593e-07, + "loss": 0.1334, + "step": 2468 + }, + { + "epoch": 0.13, + "learning_rate": 3.8983483629833656e-07, + "loss": 0.1767, + "step": 2469 + }, + { + "epoch": 0.13, + "learning_rate": 3.8982419138216087e-07, + "loss": 0.1373, + "step": 2470 + }, + { + "epoch": 0.13, + "learning_rate": 3.898135410407365e-07, + "loss": 0.1614, + "step": 2471 + }, + { + "epoch": 0.13, + "learning_rate": 3.898028852743679e-07, + "loss": 0.1481, + "step": 2472 + }, + { + "epoch": 0.13, + "learning_rate": 3.897922240833595e-07, + "loss": 0.1525, + "step": 2473 + }, + { + "epoch": 0.13, + "learning_rate": 3.8978155746801617e-07, + "loss": 0.1446, + "step": 2474 + }, + { + "epoch": 0.13, + "learning_rate": 3.8977088542864257e-07, + "loss": 0.126, + "step": 2475 + }, + { + "epoch": 0.13, + "learning_rate": 3.8976020796554384e-07, + "loss": 0.1665, + "step": 2476 + }, + { + "epoch": 0.13, + "learning_rate": 3.897495250790251e-07, + "loss": 0.1253, + "step": 2477 + }, + { + "epoch": 0.13, + "learning_rate": 3.8973883676939174e-07, + "loss": 0.1899, + "step": 2478 + }, + { + "epoch": 0.13, + "learning_rate": 3.8972814303694903e-07, + "loss": 0.1283, + "step": 2479 + }, + { + "epoch": 0.13, + "learning_rate": 3.8971744388200293e-07, + "loss": 0.1636, + "step": 2480 + }, + { + "epoch": 0.13, + "learning_rate": 3.8970673930485895e-07, + "loss": 0.1478, + "step": 2481 + }, + { + "epoch": 0.13, + "learning_rate": 3.8969602930582315e-07, + "loss": 0.1614, + "step": 2482 + }, + { + "epoch": 0.13, + "learning_rate": 3.896853138852016e-07, + "loss": 0.1733, + "step": 2483 + }, + { + "epoch": 0.13, + "learning_rate": 3.8967459304330054e-07, + "loss": 0.1385, + "step": 2484 + }, + { + "epoch": 0.13, + "learning_rate": 3.8966386678042636e-07, + "loss": 0.1479, + "step": 2485 + }, + { + "epoch": 0.13, + "learning_rate": 3.896531350968857e-07, + "loss": 0.1544, + "step": 2486 + }, + { + "epoch": 0.13, + "learning_rate": 3.896423979929852e-07, + "loss": 0.1234, + "step": 2487 + }, + { + "epoch": 0.13, + "learning_rate": 3.896316554690318e-07, + "loss": 0.1325, + "step": 2488 + }, + { + "epoch": 0.13, + "learning_rate": 3.896209075253324e-07, + "loss": 0.1342, + "step": 2489 + }, + { + "epoch": 0.13, + "learning_rate": 3.8961015416219435e-07, + "loss": 0.1489, + "step": 2490 + }, + { + "epoch": 0.13, + "learning_rate": 3.8959939537992485e-07, + "loss": 0.1418, + "step": 2491 + }, + { + "epoch": 0.13, + "learning_rate": 3.8958863117883147e-07, + "loss": 0.1392, + "step": 2492 + }, + { + "epoch": 0.13, + "learning_rate": 3.895778615592218e-07, + "loss": 0.1456, + "step": 2493 + }, + { + "epoch": 0.13, + "learning_rate": 3.895670865214036e-07, + "loss": 0.1502, + "step": 2494 + }, + { + "epoch": 0.13, + "learning_rate": 3.89556306065685e-07, + "loss": 0.1179, + "step": 2495 + }, + { + "epoch": 0.13, + "learning_rate": 3.8954552019237393e-07, + "loss": 0.215, + "step": 2496 + }, + { + "epoch": 0.13, + "learning_rate": 3.895347289017787e-07, + "loss": 0.1282, + "step": 2497 + }, + { + "epoch": 0.13, + "learning_rate": 3.8952393219420785e-07, + "loss": 0.1302, + "step": 2498 + }, + { + "epoch": 0.13, + "learning_rate": 3.895131300699698e-07, + "loss": 0.1858, + "step": 2499 + }, + { + "epoch": 0.13, + "learning_rate": 3.895023225293733e-07, + "loss": 0.1624, + "step": 2500 + }, + { + "epoch": 0.13, + "learning_rate": 3.8949150957272725e-07, + "loss": 0.1787, + "step": 2501 + }, + { + "epoch": 0.13, + "learning_rate": 3.894806912003408e-07, + "loss": 0.151, + "step": 2502 + }, + { + "epoch": 0.13, + "learning_rate": 3.89469867412523e-07, + "loss": 0.147, + "step": 2503 + }, + { + "epoch": 0.13, + "learning_rate": 3.894590382095832e-07, + "loss": 0.1403, + "step": 2504 + }, + { + "epoch": 0.13, + "learning_rate": 3.89448203591831e-07, + "loss": 0.1479, + "step": 2505 + }, + { + "epoch": 0.13, + "learning_rate": 3.8943736355957595e-07, + "loss": 0.1739, + "step": 2506 + }, + { + "epoch": 0.13, + "learning_rate": 3.89426518113128e-07, + "loss": 0.1827, + "step": 2507 + }, + { + "epoch": 0.13, + "learning_rate": 3.8941566725279704e-07, + "loss": 0.1692, + "step": 2508 + }, + { + "epoch": 0.13, + "learning_rate": 3.894048109788931e-07, + "loss": 0.16, + "step": 2509 + }, + { + "epoch": 0.13, + "learning_rate": 3.893939492917266e-07, + "loss": 0.1463, + "step": 2510 + }, + { + "epoch": 0.13, + "learning_rate": 3.893830821916079e-07, + "loss": 0.132, + "step": 2511 + }, + { + "epoch": 0.13, + "learning_rate": 3.8937220967884764e-07, + "loss": 0.185, + "step": 2512 + }, + { + "epoch": 0.13, + "learning_rate": 3.893613317537565e-07, + "loss": 0.1301, + "step": 2513 + }, + { + "epoch": 0.13, + "learning_rate": 3.893504484166454e-07, + "loss": 0.1559, + "step": 2514 + }, + { + "epoch": 0.13, + "learning_rate": 3.8933955966782535e-07, + "loss": 0.1651, + "step": 2515 + }, + { + "epoch": 0.13, + "learning_rate": 3.8932866550760766e-07, + "loss": 0.1782, + "step": 2516 + }, + { + "epoch": 0.13, + "learning_rate": 3.893177659363036e-07, + "loss": 0.1891, + "step": 2517 + }, + { + "epoch": 0.13, + "learning_rate": 3.893068609542247e-07, + "loss": 0.2101, + "step": 2518 + }, + { + "epoch": 0.13, + "learning_rate": 3.892959505616826e-07, + "loss": 0.1484, + "step": 2519 + }, + { + "epoch": 0.13, + "learning_rate": 3.892850347589892e-07, + "loss": 0.167, + "step": 2520 + }, + { + "epoch": 0.13, + "learning_rate": 3.892741135464564e-07, + "loss": 0.1301, + "step": 2521 + }, + { + "epoch": 0.13, + "learning_rate": 3.8926318692439633e-07, + "loss": 0.1294, + "step": 2522 + }, + { + "epoch": 0.13, + "learning_rate": 3.8925225489312143e-07, + "loss": 0.1307, + "step": 2523 + }, + { + "epoch": 0.13, + "learning_rate": 3.8924131745294396e-07, + "loss": 0.1809, + "step": 2524 + }, + { + "epoch": 0.13, + "learning_rate": 3.892303746041766e-07, + "loss": 0.128, + "step": 2525 + }, + { + "epoch": 0.13, + "learning_rate": 3.8921942634713207e-07, + "loss": 0.1214, + "step": 2526 + }, + { + "epoch": 0.13, + "learning_rate": 3.892084726821233e-07, + "loss": 0.1255, + "step": 2527 + }, + { + "epoch": 0.13, + "learning_rate": 3.891975136094633e-07, + "loss": 0.1709, + "step": 2528 + }, + { + "epoch": 0.13, + "learning_rate": 3.8918654912946537e-07, + "loss": 0.183, + "step": 2529 + }, + { + "epoch": 0.13, + "learning_rate": 3.891755792424428e-07, + "loss": 0.1312, + "step": 2530 + }, + { + "epoch": 0.13, + "learning_rate": 3.8916460394870913e-07, + "loss": 0.137, + "step": 2531 + }, + { + "epoch": 0.13, + "learning_rate": 3.891536232485781e-07, + "loss": 0.1903, + "step": 2532 + }, + { + "epoch": 0.13, + "learning_rate": 3.8914263714236346e-07, + "loss": 0.1077, + "step": 2533 + }, + { + "epoch": 0.13, + "learning_rate": 3.891316456303793e-07, + "loss": 0.1548, + "step": 2534 + }, + { + "epoch": 0.13, + "learning_rate": 3.8912064871293964e-07, + "loss": 0.1444, + "step": 2535 + }, + { + "epoch": 0.13, + "learning_rate": 3.8910964639035885e-07, + "loss": 0.116, + "step": 2536 + }, + { + "epoch": 0.13, + "learning_rate": 3.890986386629514e-07, + "loss": 0.1821, + "step": 2537 + }, + { + "epoch": 0.13, + "learning_rate": 3.890876255310318e-07, + "loss": 0.1297, + "step": 2538 + }, + { + "epoch": 0.13, + "learning_rate": 3.8907660699491485e-07, + "loss": 0.1387, + "step": 2539 + }, + { + "epoch": 0.13, + "learning_rate": 3.890655830549155e-07, + "loss": 0.1428, + "step": 2540 + }, + { + "epoch": 0.13, + "learning_rate": 3.890545537113488e-07, + "loss": 0.1401, + "step": 2541 + }, + { + "epoch": 0.13, + "learning_rate": 3.8904351896453e-07, + "loss": 0.1671, + "step": 2542 + }, + { + "epoch": 0.13, + "learning_rate": 3.8903247881477446e-07, + "loss": 0.1454, + "step": 2543 + }, + { + "epoch": 0.13, + "learning_rate": 3.8902143326239766e-07, + "loss": 0.1374, + "step": 2544 + }, + { + "epoch": 0.13, + "learning_rate": 3.890103823077154e-07, + "loss": 0.1607, + "step": 2545 + }, + { + "epoch": 0.13, + "learning_rate": 3.8899932595104334e-07, + "loss": 0.1015, + "step": 2546 + }, + { + "epoch": 0.13, + "learning_rate": 3.8898826419269766e-07, + "loss": 0.1415, + "step": 2547 + }, + { + "epoch": 0.13, + "learning_rate": 3.889771970329944e-07, + "loss": 0.137, + "step": 2548 + }, + { + "epoch": 0.13, + "learning_rate": 3.889661244722499e-07, + "loss": 0.173, + "step": 2549 + }, + { + "epoch": 0.13, + "learning_rate": 3.889550465107806e-07, + "loss": 0.1592, + "step": 2550 + }, + { + "epoch": 0.13, + "learning_rate": 3.8894396314890317e-07, + "loss": 0.1625, + "step": 2551 + }, + { + "epoch": 0.13, + "learning_rate": 3.8893287438693426e-07, + "loss": 0.1165, + "step": 2552 + }, + { + "epoch": 0.13, + "learning_rate": 3.88921780225191e-07, + "loss": 0.1545, + "step": 2553 + }, + { + "epoch": 0.13, + "learning_rate": 3.889106806639902e-07, + "loss": 0.1604, + "step": 2554 + }, + { + "epoch": 0.13, + "learning_rate": 3.8889957570364926e-07, + "loss": 0.1185, + "step": 2555 + }, + { + "epoch": 0.13, + "learning_rate": 3.888884653444855e-07, + "loss": 0.1524, + "step": 2556 + }, + { + "epoch": 0.13, + "learning_rate": 3.8887734958681653e-07, + "loss": 0.1405, + "step": 2557 + }, + { + "epoch": 0.13, + "learning_rate": 3.8886622843096e-07, + "loss": 0.1502, + "step": 2558 + }, + { + "epoch": 0.13, + "learning_rate": 3.888551018772337e-07, + "loss": 0.1422, + "step": 2559 + }, + { + "epoch": 0.13, + "learning_rate": 3.888439699259557e-07, + "loss": 0.1801, + "step": 2560 + }, + { + "epoch": 0.13, + "learning_rate": 3.8883283257744414e-07, + "loss": 0.1239, + "step": 2561 + }, + { + "epoch": 0.13, + "learning_rate": 3.888216898320174e-07, + "loss": 0.1602, + "step": 2562 + }, + { + "epoch": 0.13, + "learning_rate": 3.888105416899938e-07, + "loss": 0.162, + "step": 2563 + }, + { + "epoch": 0.13, + "learning_rate": 3.8879938815169205e-07, + "loss": 0.1608, + "step": 2564 + }, + { + "epoch": 0.13, + "learning_rate": 3.887882292174309e-07, + "loss": 0.1401, + "step": 2565 + }, + { + "epoch": 0.13, + "learning_rate": 3.8877706488752923e-07, + "loss": 0.1473, + "step": 2566 + }, + { + "epoch": 0.13, + "learning_rate": 3.887658951623062e-07, + "loss": 0.1801, + "step": 2567 + }, + { + "epoch": 0.13, + "learning_rate": 3.88754720042081e-07, + "loss": 0.1373, + "step": 2568 + }, + { + "epoch": 0.13, + "learning_rate": 3.887435395271731e-07, + "loss": 0.1392, + "step": 2569 + }, + { + "epoch": 0.13, + "learning_rate": 3.8873235361790193e-07, + "loss": 0.1306, + "step": 2570 + }, + { + "epoch": 0.13, + "learning_rate": 3.887211623145872e-07, + "loss": 0.2035, + "step": 2571 + }, + { + "epoch": 0.13, + "learning_rate": 3.887099656175489e-07, + "loss": 0.1228, + "step": 2572 + }, + { + "epoch": 0.13, + "learning_rate": 3.886987635271069e-07, + "loss": 0.1419, + "step": 2573 + }, + { + "epoch": 0.13, + "learning_rate": 3.886875560435813e-07, + "loss": 0.1618, + "step": 2574 + }, + { + "epoch": 0.13, + "learning_rate": 3.886763431672926e-07, + "loss": 0.1409, + "step": 2575 + }, + { + "epoch": 0.13, + "learning_rate": 3.886651248985612e-07, + "loss": 0.1816, + "step": 2576 + }, + { + "epoch": 0.13, + "learning_rate": 3.886539012377076e-07, + "loss": 0.1436, + "step": 2577 + }, + { + "epoch": 0.13, + "learning_rate": 3.8864267218505274e-07, + "loss": 0.1707, + "step": 2578 + }, + { + "epoch": 0.13, + "learning_rate": 3.8863143774091743e-07, + "loss": 0.1314, + "step": 2579 + }, + { + "epoch": 0.13, + "learning_rate": 3.8862019790562287e-07, + "loss": 0.1329, + "step": 2580 + }, + { + "epoch": 0.13, + "learning_rate": 3.8860895267949025e-07, + "loss": 0.1733, + "step": 2581 + }, + { + "epoch": 0.13, + "learning_rate": 3.885977020628409e-07, + "loss": 0.1648, + "step": 2582 + }, + { + "epoch": 0.13, + "learning_rate": 3.8858644605599643e-07, + "loss": 0.1624, + "step": 2583 + }, + { + "epoch": 0.13, + "learning_rate": 3.8857518465927857e-07, + "loss": 0.168, + "step": 2584 + }, + { + "epoch": 0.13, + "learning_rate": 3.885639178730091e-07, + "loss": 0.1333, + "step": 2585 + }, + { + "epoch": 0.13, + "learning_rate": 3.8855264569751007e-07, + "loss": 0.1747, + "step": 2586 + }, + { + "epoch": 0.14, + "learning_rate": 3.8854136813310365e-07, + "loss": 0.1349, + "step": 2587 + }, + { + "epoch": 0.14, + "learning_rate": 3.885300851801121e-07, + "loss": 0.1153, + "step": 2588 + }, + { + "epoch": 0.14, + "learning_rate": 3.88518796838858e-07, + "loss": 0.1608, + "step": 2589 + }, + { + "epoch": 0.14, + "learning_rate": 3.8850750310966384e-07, + "loss": 0.1268, + "step": 2590 + }, + { + "epoch": 0.14, + "learning_rate": 3.884962039928525e-07, + "loss": 0.1177, + "step": 2591 + }, + { + "epoch": 0.14, + "learning_rate": 3.884848994887469e-07, + "loss": 0.1661, + "step": 2592 + }, + { + "epoch": 0.14, + "learning_rate": 3.8847358959767013e-07, + "loss": 0.1348, + "step": 2593 + }, + { + "epoch": 0.14, + "learning_rate": 3.8846227431994543e-07, + "loss": 0.1474, + "step": 2594 + }, + { + "epoch": 0.14, + "learning_rate": 3.8845095365589613e-07, + "loss": 0.1548, + "step": 2595 + }, + { + "epoch": 0.14, + "learning_rate": 3.8843962760584574e-07, + "loss": 0.1232, + "step": 2596 + }, + { + "epoch": 0.14, + "learning_rate": 3.8842829617011815e-07, + "loss": 0.14, + "step": 2597 + }, + { + "epoch": 0.14, + "learning_rate": 3.8841695934903714e-07, + "loss": 0.1722, + "step": 2598 + }, + { + "epoch": 0.14, + "learning_rate": 3.8840561714292665e-07, + "loss": 0.1558, + "step": 2599 + }, + { + "epoch": 0.14, + "learning_rate": 3.8839426955211087e-07, + "loss": 0.1176, + "step": 2600 + }, + { + "epoch": 0.14, + "learning_rate": 3.8838291657691417e-07, + "loss": 0.1756, + "step": 2601 + }, + { + "epoch": 0.14, + "learning_rate": 3.8837155821766096e-07, + "loss": 0.1408, + "step": 2602 + }, + { + "epoch": 0.14, + "learning_rate": 3.883601944746759e-07, + "loss": 0.1531, + "step": 2603 + }, + { + "epoch": 0.14, + "learning_rate": 3.883488253482838e-07, + "loss": 0.1468, + "step": 2604 + }, + { + "epoch": 0.14, + "learning_rate": 3.883374508388096e-07, + "loss": 0.1464, + "step": 2605 + }, + { + "epoch": 0.14, + "learning_rate": 3.883260709465783e-07, + "loss": 0.1916, + "step": 2606 + }, + { + "epoch": 0.14, + "learning_rate": 3.8831468567191515e-07, + "loss": 0.2256, + "step": 2607 + }, + { + "epoch": 0.14, + "learning_rate": 3.8830329501514563e-07, + "loss": 0.153, + "step": 2608 + }, + { + "epoch": 0.14, + "learning_rate": 3.8829189897659526e-07, + "loss": 0.1941, + "step": 2609 + }, + { + "epoch": 0.14, + "learning_rate": 3.882804975565897e-07, + "loss": 0.1737, + "step": 2610 + }, + { + "epoch": 0.14, + "learning_rate": 3.882690907554548e-07, + "loss": 0.1583, + "step": 2611 + }, + { + "epoch": 0.14, + "learning_rate": 3.882576785735167e-07, + "loss": 0.1321, + "step": 2612 + }, + { + "epoch": 0.14, + "learning_rate": 3.882462610111014e-07, + "loss": 0.1115, + "step": 2613 + }, + { + "epoch": 0.14, + "learning_rate": 3.8823483806853534e-07, + "loss": 0.1324, + "step": 2614 + }, + { + "epoch": 0.14, + "learning_rate": 3.882234097461449e-07, + "loss": 0.1379, + "step": 2615 + }, + { + "epoch": 0.14, + "learning_rate": 3.8821197604425676e-07, + "loss": 0.1451, + "step": 2616 + }, + { + "epoch": 0.14, + "learning_rate": 3.8820053696319773e-07, + "loss": 0.1348, + "step": 2617 + }, + { + "epoch": 0.14, + "learning_rate": 3.881890925032946e-07, + "loss": 0.1716, + "step": 2618 + }, + { + "epoch": 0.14, + "learning_rate": 3.881776426648747e-07, + "loss": 0.1234, + "step": 2619 + }, + { + "epoch": 0.14, + "learning_rate": 3.8816618744826503e-07, + "loss": 0.1449, + "step": 2620 + }, + { + "epoch": 0.14, + "learning_rate": 3.881547268537931e-07, + "loss": 0.1521, + "step": 2621 + }, + { + "epoch": 0.14, + "learning_rate": 3.8814326088178644e-07, + "loss": 0.1389, + "step": 2622 + }, + { + "epoch": 0.14, + "learning_rate": 3.8813178953257276e-07, + "loss": 0.1211, + "step": 2623 + }, + { + "epoch": 0.14, + "learning_rate": 3.881203128064799e-07, + "loss": 0.151, + "step": 2624 + }, + { + "epoch": 0.14, + "learning_rate": 3.8810883070383593e-07, + "loss": 0.1424, + "step": 2625 + }, + { + "epoch": 0.14, + "learning_rate": 3.880973432249689e-07, + "loss": 0.1597, + "step": 2626 + }, + { + "epoch": 0.14, + "learning_rate": 3.880858503702072e-07, + "loss": 0.1429, + "step": 2627 + }, + { + "epoch": 0.14, + "learning_rate": 3.880743521398793e-07, + "loss": 0.1788, + "step": 2628 + }, + { + "epoch": 0.14, + "learning_rate": 3.8806284853431384e-07, + "loss": 0.1543, + "step": 2629 + }, + { + "epoch": 0.14, + "learning_rate": 3.880513395538395e-07, + "loss": 0.1339, + "step": 2630 + }, + { + "epoch": 0.14, + "learning_rate": 3.880398251987853e-07, + "loss": 0.1168, + "step": 2631 + }, + { + "epoch": 0.14, + "learning_rate": 3.8802830546948035e-07, + "loss": 0.147, + "step": 2632 + }, + { + "epoch": 0.14, + "learning_rate": 3.8801678036625384e-07, + "loss": 0.1284, + "step": 2633 + }, + { + "epoch": 0.14, + "learning_rate": 3.880052498894351e-07, + "loss": 0.2179, + "step": 2634 + }, + { + "epoch": 0.14, + "learning_rate": 3.8799371403935373e-07, + "loss": 0.1097, + "step": 2635 + }, + { + "epoch": 0.14, + "learning_rate": 3.8798217281633955e-07, + "loss": 0.1208, + "step": 2636 + }, + { + "epoch": 0.14, + "learning_rate": 3.879706262207223e-07, + "loss": 0.1508, + "step": 2637 + }, + { + "epoch": 0.14, + "learning_rate": 3.8795907425283183e-07, + "loss": 0.1432, + "step": 2638 + }, + { + "epoch": 0.14, + "learning_rate": 3.8794751691299857e-07, + "loss": 0.125, + "step": 2639 + }, + { + "epoch": 0.14, + "learning_rate": 3.8793595420155277e-07, + "loss": 0.1725, + "step": 2640 + }, + { + "epoch": 0.14, + "learning_rate": 3.879243861188247e-07, + "loss": 0.1389, + "step": 2641 + }, + { + "epoch": 0.14, + "learning_rate": 3.879128126651453e-07, + "loss": 0.1518, + "step": 2642 + }, + { + "epoch": 0.14, + "learning_rate": 3.8790123384084505e-07, + "loss": 0.129, + "step": 2643 + }, + { + "epoch": 0.14, + "learning_rate": 3.878896496462551e-07, + "loss": 0.1404, + "step": 2644 + }, + { + "epoch": 0.14, + "learning_rate": 3.8787806008170634e-07, + "loss": 0.1908, + "step": 2645 + }, + { + "epoch": 0.14, + "learning_rate": 3.8786646514753014e-07, + "loss": 0.1898, + "step": 2646 + }, + { + "epoch": 0.14, + "learning_rate": 3.878548648440578e-07, + "loss": 0.1635, + "step": 2647 + }, + { + "epoch": 0.14, + "learning_rate": 3.878432591716209e-07, + "loss": 0.1298, + "step": 2648 + }, + { + "epoch": 0.14, + "learning_rate": 3.878316481305512e-07, + "loss": 0.1824, + "step": 2649 + }, + { + "epoch": 0.14, + "learning_rate": 3.8782003172118047e-07, + "loss": 0.1662, + "step": 2650 + }, + { + "epoch": 0.14, + "learning_rate": 3.878084099438407e-07, + "loss": 0.1588, + "step": 2651 + }, + { + "epoch": 0.14, + "learning_rate": 3.877967827988641e-07, + "loss": 0.1593, + "step": 2652 + }, + { + "epoch": 0.14, + "learning_rate": 3.8778515028658295e-07, + "loss": 0.1845, + "step": 2653 + }, + { + "epoch": 0.14, + "learning_rate": 3.877735124073297e-07, + "loss": 0.1381, + "step": 2654 + }, + { + "epoch": 0.14, + "learning_rate": 3.8776186916143694e-07, + "loss": 0.1417, + "step": 2655 + }, + { + "epoch": 0.14, + "learning_rate": 3.877502205492375e-07, + "loss": 0.1242, + "step": 2656 + }, + { + "epoch": 0.14, + "learning_rate": 3.877385665710643e-07, + "loss": 0.1552, + "step": 2657 + }, + { + "epoch": 0.14, + "learning_rate": 3.877269072272504e-07, + "loss": 0.1873, + "step": 2658 + }, + { + "epoch": 0.14, + "learning_rate": 3.87715242518129e-07, + "loss": 0.1828, + "step": 2659 + }, + { + "epoch": 0.14, + "learning_rate": 3.877035724440335e-07, + "loss": 0.1749, + "step": 2660 + }, + { + "epoch": 0.14, + "learning_rate": 3.8769189700529745e-07, + "loss": 0.1378, + "step": 2661 + }, + { + "epoch": 0.14, + "learning_rate": 3.876802162022545e-07, + "loss": 0.1201, + "step": 2662 + }, + { + "epoch": 0.14, + "learning_rate": 3.8766853003523857e-07, + "loss": 0.1631, + "step": 2663 + }, + { + "epoch": 0.14, + "learning_rate": 3.876568385045836e-07, + "loss": 0.1495, + "step": 2664 + }, + { + "epoch": 0.14, + "learning_rate": 3.876451416106237e-07, + "loss": 0.1515, + "step": 2665 + }, + { + "epoch": 0.14, + "learning_rate": 3.8763343935369324e-07, + "loss": 0.1797, + "step": 2666 + }, + { + "epoch": 0.14, + "learning_rate": 3.8762173173412664e-07, + "loss": 0.1398, + "step": 2667 + }, + { + "epoch": 0.14, + "learning_rate": 3.876100187522585e-07, + "loss": 0.1384, + "step": 2668 + }, + { + "epoch": 0.14, + "learning_rate": 3.8759830040842365e-07, + "loss": 0.16, + "step": 2669 + }, + { + "epoch": 0.14, + "learning_rate": 3.875865767029569e-07, + "loss": 0.1338, + "step": 2670 + }, + { + "epoch": 0.14, + "learning_rate": 3.8757484763619345e-07, + "loss": 0.1906, + "step": 2671 + }, + { + "epoch": 0.14, + "learning_rate": 3.875631132084684e-07, + "loss": 0.1734, + "step": 2672 + }, + { + "epoch": 0.14, + "learning_rate": 3.875513734201171e-07, + "loss": 0.1501, + "step": 2673 + }, + { + "epoch": 0.14, + "learning_rate": 3.8753962827147524e-07, + "loss": 0.137, + "step": 2674 + }, + { + "epoch": 0.14, + "learning_rate": 3.875278777628783e-07, + "loss": 0.1313, + "step": 2675 + }, + { + "epoch": 0.14, + "learning_rate": 3.875161218946624e-07, + "loss": 0.1311, + "step": 2676 + }, + { + "epoch": 0.14, + "learning_rate": 3.8750436066716315e-07, + "loss": 0.1306, + "step": 2677 + }, + { + "epoch": 0.14, + "learning_rate": 3.8749259408071697e-07, + "loss": 0.1639, + "step": 2678 + }, + { + "epoch": 0.14, + "learning_rate": 3.874808221356601e-07, + "loss": 0.1468, + "step": 2679 + }, + { + "epoch": 0.14, + "learning_rate": 3.8746904483232897e-07, + "loss": 0.1526, + "step": 2680 + }, + { + "epoch": 0.14, + "learning_rate": 3.8745726217106013e-07, + "loss": 0.1707, + "step": 2681 + }, + { + "epoch": 0.14, + "learning_rate": 3.874454741521903e-07, + "loss": 0.1178, + "step": 2682 + }, + { + "epoch": 0.14, + "learning_rate": 3.8743368077605665e-07, + "loss": 0.1873, + "step": 2683 + }, + { + "epoch": 0.14, + "learning_rate": 3.874218820429959e-07, + "loss": 0.1867, + "step": 2684 + }, + { + "epoch": 0.14, + "learning_rate": 3.874100779533454e-07, + "loss": 0.1826, + "step": 2685 + }, + { + "epoch": 0.14, + "learning_rate": 3.873982685074426e-07, + "loss": 0.1418, + "step": 2686 + }, + { + "epoch": 0.14, + "learning_rate": 3.8738645370562495e-07, + "loss": 0.1497, + "step": 2687 + }, + { + "epoch": 0.14, + "learning_rate": 3.8737463354823004e-07, + "loss": 0.1412, + "step": 2688 + }, + { + "epoch": 0.14, + "learning_rate": 3.873628080355958e-07, + "loss": 0.1459, + "step": 2689 + }, + { + "epoch": 0.14, + "learning_rate": 3.8735097716806016e-07, + "loss": 0.1866, + "step": 2690 + }, + { + "epoch": 0.14, + "learning_rate": 3.873391409459613e-07, + "loss": 0.1395, + "step": 2691 + }, + { + "epoch": 0.14, + "learning_rate": 3.8732729936963746e-07, + "loss": 0.1664, + "step": 2692 + }, + { + "epoch": 0.14, + "learning_rate": 3.8731545243942706e-07, + "loss": 0.119, + "step": 2693 + }, + { + "epoch": 0.14, + "learning_rate": 3.873036001556688e-07, + "loss": 0.1798, + "step": 2694 + }, + { + "epoch": 0.14, + "learning_rate": 3.8729174251870126e-07, + "loss": 0.1409, + "step": 2695 + }, + { + "epoch": 0.14, + "learning_rate": 3.8727987952886344e-07, + "loss": 0.1479, + "step": 2696 + }, + { + "epoch": 0.14, + "learning_rate": 3.8726801118649435e-07, + "loss": 0.1172, + "step": 2697 + }, + { + "epoch": 0.14, + "learning_rate": 3.8725613749193323e-07, + "loss": 0.1761, + "step": 2698 + }, + { + "epoch": 0.14, + "learning_rate": 3.8724425844551943e-07, + "loss": 0.1631, + "step": 2699 + }, + { + "epoch": 0.14, + "learning_rate": 3.872323740475924e-07, + "loss": 0.1439, + "step": 2700 + }, + { + "epoch": 0.14, + "learning_rate": 3.872204842984919e-07, + "loss": 0.1441, + "step": 2701 + }, + { + "epoch": 0.14, + "learning_rate": 3.872085891985576e-07, + "loss": 0.1741, + "step": 2702 + }, + { + "epoch": 0.14, + "learning_rate": 3.871966887481296e-07, + "loss": 0.1845, + "step": 2703 + }, + { + "epoch": 0.14, + "learning_rate": 3.87184782947548e-07, + "loss": 0.1438, + "step": 2704 + }, + { + "epoch": 0.14, + "learning_rate": 3.8717287179715295e-07, + "loss": 0.1507, + "step": 2705 + }, + { + "epoch": 0.14, + "learning_rate": 3.8716095529728506e-07, + "loss": 0.174, + "step": 2706 + }, + { + "epoch": 0.14, + "learning_rate": 3.871490334482848e-07, + "loss": 0.1234, + "step": 2707 + }, + { + "epoch": 0.14, + "learning_rate": 3.871371062504929e-07, + "loss": 0.2021, + "step": 2708 + }, + { + "epoch": 0.14, + "learning_rate": 3.8712517370425026e-07, + "loss": 0.1469, + "step": 2709 + }, + { + "epoch": 0.14, + "learning_rate": 3.871132358098979e-07, + "loss": 0.1169, + "step": 2710 + }, + { + "epoch": 0.14, + "learning_rate": 3.8710129256777703e-07, + "loss": 0.1381, + "step": 2711 + }, + { + "epoch": 0.14, + "learning_rate": 3.87089343978229e-07, + "loss": 0.1599, + "step": 2712 + }, + { + "epoch": 0.14, + "learning_rate": 3.870773900415953e-07, + "loss": 0.1715, + "step": 2713 + }, + { + "epoch": 0.14, + "learning_rate": 3.870654307582176e-07, + "loss": 0.116, + "step": 2714 + }, + { + "epoch": 0.14, + "learning_rate": 3.8705346612843766e-07, + "loss": 0.1491, + "step": 2715 + }, + { + "epoch": 0.14, + "learning_rate": 3.8704149615259746e-07, + "loss": 0.1349, + "step": 2716 + }, + { + "epoch": 0.14, + "learning_rate": 3.8702952083103906e-07, + "loss": 0.1247, + "step": 2717 + }, + { + "epoch": 0.14, + "learning_rate": 3.8701754016410475e-07, + "loss": 0.1567, + "step": 2718 + }, + { + "epoch": 0.14, + "learning_rate": 3.8700555415213695e-07, + "loss": 0.1816, + "step": 2719 + }, + { + "epoch": 0.14, + "learning_rate": 3.8699356279547824e-07, + "loss": 0.1572, + "step": 2720 + }, + { + "epoch": 0.14, + "learning_rate": 3.869815660944713e-07, + "loss": 0.1313, + "step": 2721 + }, + { + "epoch": 0.14, + "learning_rate": 3.869695640494591e-07, + "loss": 0.1624, + "step": 2722 + }, + { + "epoch": 0.14, + "learning_rate": 3.869575566607844e-07, + "loss": 0.1859, + "step": 2723 + }, + { + "epoch": 0.14, + "learning_rate": 3.869455439287908e-07, + "loss": 0.1566, + "step": 2724 + }, + { + "epoch": 0.14, + "learning_rate": 3.8693352585382114e-07, + "loss": 0.158, + "step": 2725 + }, + { + "epoch": 0.14, + "learning_rate": 3.869215024362193e-07, + "loss": 0.1731, + "step": 2726 + }, + { + "epoch": 0.14, + "learning_rate": 3.869094736763287e-07, + "loss": 0.1332, + "step": 2727 + }, + { + "epoch": 0.14, + "learning_rate": 3.8689743957449325e-07, + "loss": 0.1891, + "step": 2728 + }, + { + "epoch": 0.14, + "learning_rate": 3.8688540013105684e-07, + "loss": 0.1268, + "step": 2729 + }, + { + "epoch": 0.14, + "learning_rate": 3.8687335534636344e-07, + "loss": 0.1369, + "step": 2730 + }, + { + "epoch": 0.14, + "learning_rate": 3.8686130522075747e-07, + "loss": 0.1224, + "step": 2731 + }, + { + "epoch": 0.14, + "learning_rate": 3.8684924975458325e-07, + "loss": 0.1384, + "step": 2732 + }, + { + "epoch": 0.14, + "learning_rate": 3.8683718894818536e-07, + "loss": 0.1398, + "step": 2733 + }, + { + "epoch": 0.14, + "learning_rate": 3.8682512280190845e-07, + "loss": 0.1577, + "step": 2734 + }, + { + "epoch": 0.14, + "learning_rate": 3.8681305131609746e-07, + "loss": 0.1521, + "step": 2735 + }, + { + "epoch": 0.14, + "learning_rate": 3.868009744910973e-07, + "loss": 0.1348, + "step": 2736 + }, + { + "epoch": 0.14, + "learning_rate": 3.867888923272532e-07, + "loss": 0.1629, + "step": 2737 + }, + { + "epoch": 0.14, + "learning_rate": 3.8677680482491043e-07, + "loss": 0.1763, + "step": 2738 + }, + { + "epoch": 0.14, + "learning_rate": 3.867647119844145e-07, + "loss": 0.1434, + "step": 2739 + }, + { + "epoch": 0.14, + "learning_rate": 3.86752613806111e-07, + "loss": 0.1308, + "step": 2740 + }, + { + "epoch": 0.14, + "learning_rate": 3.8674051029034567e-07, + "loss": 0.1238, + "step": 2741 + }, + { + "epoch": 0.14, + "learning_rate": 3.867284014374645e-07, + "loss": 0.1505, + "step": 2742 + }, + { + "epoch": 0.14, + "learning_rate": 3.867162872478135e-07, + "loss": 0.1728, + "step": 2743 + }, + { + "epoch": 0.14, + "learning_rate": 3.86704167721739e-07, + "loss": 0.1539, + "step": 2744 + }, + { + "epoch": 0.14, + "learning_rate": 3.866920428595872e-07, + "loss": 0.1496, + "step": 2745 + }, + { + "epoch": 0.14, + "learning_rate": 3.8667991266170486e-07, + "loss": 0.1331, + "step": 2746 + }, + { + "epoch": 0.14, + "learning_rate": 3.8666777712843845e-07, + "loss": 0.1921, + "step": 2747 + }, + { + "epoch": 0.14, + "learning_rate": 3.86655636260135e-07, + "loss": 0.1665, + "step": 2748 + }, + { + "epoch": 0.14, + "learning_rate": 3.8664349005714135e-07, + "loss": 0.1186, + "step": 2749 + }, + { + "epoch": 0.14, + "learning_rate": 3.8663133851980474e-07, + "loss": 0.1333, + "step": 2750 + }, + { + "epoch": 0.14, + "learning_rate": 3.866191816484724e-07, + "loss": 0.1342, + "step": 2751 + }, + { + "epoch": 0.14, + "learning_rate": 3.866070194434918e-07, + "loss": 0.1177, + "step": 2752 + }, + { + "epoch": 0.14, + "learning_rate": 3.8659485190521055e-07, + "loss": 0.1824, + "step": 2753 + }, + { + "epoch": 0.14, + "learning_rate": 3.8658267903397644e-07, + "loss": 0.1669, + "step": 2754 + }, + { + "epoch": 0.14, + "learning_rate": 3.865705008301373e-07, + "loss": 0.1402, + "step": 2755 + }, + { + "epoch": 0.14, + "learning_rate": 3.865583172940412e-07, + "loss": 0.183, + "step": 2756 + }, + { + "epoch": 0.14, + "learning_rate": 3.865461284260364e-07, + "loss": 0.1281, + "step": 2757 + }, + { + "epoch": 0.14, + "learning_rate": 3.8653393422647124e-07, + "loss": 0.1922, + "step": 2758 + }, + { + "epoch": 0.14, + "learning_rate": 3.8652173469569425e-07, + "loss": 0.2043, + "step": 2759 + }, + { + "epoch": 0.14, + "learning_rate": 3.865095298340541e-07, + "loss": 0.1867, + "step": 2760 + }, + { + "epoch": 0.14, + "learning_rate": 3.8649731964189945e-07, + "loss": 0.1701, + "step": 2761 + }, + { + "epoch": 0.14, + "learning_rate": 3.8648510411957955e-07, + "loss": 0.1981, + "step": 2762 + }, + { + "epoch": 0.14, + "learning_rate": 3.8647288326744337e-07, + "loss": 0.175, + "step": 2763 + }, + { + "epoch": 0.14, + "learning_rate": 3.8646065708584015e-07, + "loss": 0.1527, + "step": 2764 + }, + { + "epoch": 0.14, + "learning_rate": 3.8644842557511936e-07, + "loss": 0.139, + "step": 2765 + }, + { + "epoch": 0.14, + "learning_rate": 3.8643618873563065e-07, + "loss": 0.1492, + "step": 2766 + }, + { + "epoch": 0.14, + "learning_rate": 3.864239465677237e-07, + "loss": 0.1498, + "step": 2767 + }, + { + "epoch": 0.14, + "learning_rate": 3.864116990717483e-07, + "loss": 0.2008, + "step": 2768 + }, + { + "epoch": 0.14, + "learning_rate": 3.863994462480546e-07, + "loss": 0.1262, + "step": 2769 + }, + { + "epoch": 0.14, + "learning_rate": 3.8638718809699287e-07, + "loss": 0.129, + "step": 2770 + }, + { + "epoch": 0.14, + "learning_rate": 3.863749246189132e-07, + "loss": 0.141, + "step": 2771 + }, + { + "epoch": 0.14, + "learning_rate": 3.8636265581416634e-07, + "loss": 0.1534, + "step": 2772 + }, + { + "epoch": 0.14, + "learning_rate": 3.863503816831028e-07, + "loss": 0.1838, + "step": 2773 + }, + { + "epoch": 0.14, + "learning_rate": 3.863381022260734e-07, + "loss": 0.1496, + "step": 2774 + }, + { + "epoch": 0.14, + "learning_rate": 3.8632581744342917e-07, + "loss": 0.1652, + "step": 2775 + }, + { + "epoch": 0.14, + "learning_rate": 3.863135273355211e-07, + "loss": 0.1516, + "step": 2776 + }, + { + "epoch": 0.14, + "learning_rate": 3.863012319027005e-07, + "loss": 0.1559, + "step": 2777 + }, + { + "epoch": 0.14, + "learning_rate": 3.8628893114531876e-07, + "loss": 0.1655, + "step": 2778 + }, + { + "epoch": 0.15, + "learning_rate": 3.862766250637275e-07, + "loss": 0.1512, + "step": 2779 + }, + { + "epoch": 0.15, + "learning_rate": 3.862643136582783e-07, + "loss": 0.1825, + "step": 2780 + }, + { + "epoch": 0.15, + "learning_rate": 3.862519969293232e-07, + "loss": 0.1741, + "step": 2781 + }, + { + "epoch": 0.15, + "learning_rate": 3.8623967487721403e-07, + "loss": 0.1244, + "step": 2782 + }, + { + "epoch": 0.15, + "learning_rate": 3.8622734750230313e-07, + "loss": 0.1308, + "step": 2783 + }, + { + "epoch": 0.15, + "learning_rate": 3.8621501480494276e-07, + "loss": 0.1558, + "step": 2784 + }, + { + "epoch": 0.15, + "learning_rate": 3.8620267678548533e-07, + "loss": 0.1161, + "step": 2785 + }, + { + "epoch": 0.15, + "learning_rate": 3.861903334442836e-07, + "loss": 0.1455, + "step": 2786 + }, + { + "epoch": 0.15, + "learning_rate": 3.8617798478169015e-07, + "loss": 0.2036, + "step": 2787 + }, + { + "epoch": 0.15, + "learning_rate": 3.8616563079805814e-07, + "loss": 0.1442, + "step": 2788 + }, + { + "epoch": 0.15, + "learning_rate": 3.861532714937405e-07, + "loss": 0.1662, + "step": 2789 + }, + { + "epoch": 0.15, + "learning_rate": 3.861409068690904e-07, + "loss": 0.1485, + "step": 2790 + }, + { + "epoch": 0.15, + "learning_rate": 3.861285369244614e-07, + "loss": 0.1297, + "step": 2791 + }, + { + "epoch": 0.15, + "learning_rate": 3.86116161660207e-07, + "loss": 0.1243, + "step": 2792 + }, + { + "epoch": 0.15, + "learning_rate": 3.861037810766808e-07, + "loss": 0.1753, + "step": 2793 + }, + { + "epoch": 0.15, + "learning_rate": 3.8609139517423665e-07, + "loss": 0.1446, + "step": 2794 + }, + { + "epoch": 0.15, + "learning_rate": 3.8607900395322865e-07, + "loss": 0.1217, + "step": 2795 + }, + { + "epoch": 0.15, + "learning_rate": 3.8606660741401085e-07, + "loss": 0.1707, + "step": 2796 + }, + { + "epoch": 0.15, + "learning_rate": 3.860542055569376e-07, + "loss": 0.1942, + "step": 2797 + }, + { + "epoch": 0.15, + "learning_rate": 3.8604179838236337e-07, + "loss": 0.1575, + "step": 2798 + }, + { + "epoch": 0.15, + "learning_rate": 3.8602938589064267e-07, + "loss": 0.1833, + "step": 2799 + }, + { + "epoch": 0.15, + "learning_rate": 3.860169680821303e-07, + "loss": 0.1447, + "step": 2800 + }, + { + "epoch": 0.15, + "learning_rate": 3.860045449571812e-07, + "loss": 0.1316, + "step": 2801 + }, + { + "epoch": 0.15, + "learning_rate": 3.859921165161504e-07, + "loss": 0.1401, + "step": 2802 + }, + { + "epoch": 0.15, + "learning_rate": 3.859796827593931e-07, + "loss": 0.1321, + "step": 2803 + }, + { + "epoch": 0.15, + "learning_rate": 3.859672436872647e-07, + "loss": 0.1177, + "step": 2804 + }, + { + "epoch": 0.15, + "learning_rate": 3.8595479930012064e-07, + "loss": 0.1359, + "step": 2805 + }, + { + "epoch": 0.15, + "learning_rate": 3.859423495983167e-07, + "loss": 0.1511, + "step": 2806 + }, + { + "epoch": 0.15, + "learning_rate": 3.859298945822085e-07, + "loss": 0.1862, + "step": 2807 + }, + { + "epoch": 0.15, + "learning_rate": 3.8591743425215225e-07, + "loss": 0.152, + "step": 2808 + }, + { + "epoch": 0.15, + "learning_rate": 3.859049686085039e-07, + "loss": 0.1282, + "step": 2809 + }, + { + "epoch": 0.15, + "learning_rate": 3.8589249765161983e-07, + "loss": 0.1663, + "step": 2810 + }, + { + "epoch": 0.15, + "learning_rate": 3.858800213818564e-07, + "loss": 0.1394, + "step": 2811 + }, + { + "epoch": 0.15, + "learning_rate": 3.8586753979957016e-07, + "loss": 0.1398, + "step": 2812 + }, + { + "epoch": 0.15, + "learning_rate": 3.8585505290511795e-07, + "loss": 0.164, + "step": 2813 + }, + { + "epoch": 0.15, + "learning_rate": 3.8584256069885654e-07, + "loss": 0.1489, + "step": 2814 + }, + { + "epoch": 0.15, + "learning_rate": 3.8583006318114305e-07, + "loss": 0.1303, + "step": 2815 + }, + { + "epoch": 0.15, + "learning_rate": 3.8581756035233456e-07, + "loss": 0.1769, + "step": 2816 + }, + { + "epoch": 0.15, + "learning_rate": 3.858050522127884e-07, + "loss": 0.1374, + "step": 2817 + }, + { + "epoch": 0.15, + "learning_rate": 3.8579253876286224e-07, + "loss": 0.1612, + "step": 2818 + }, + { + "epoch": 0.15, + "learning_rate": 3.8578002000291355e-07, + "loss": 0.1784, + "step": 2819 + }, + { + "epoch": 0.15, + "learning_rate": 3.8576749593330014e-07, + "loss": 0.112, + "step": 2820 + }, + { + "epoch": 0.15, + "learning_rate": 3.857549665543801e-07, + "loss": 0.151, + "step": 2821 + }, + { + "epoch": 0.15, + "learning_rate": 3.8574243186651125e-07, + "loss": 0.1378, + "step": 2822 + }, + { + "epoch": 0.15, + "learning_rate": 3.857298918700521e-07, + "loss": 0.1894, + "step": 2823 + }, + { + "epoch": 0.15, + "learning_rate": 3.857173465653609e-07, + "loss": 0.1558, + "step": 2824 + }, + { + "epoch": 0.15, + "learning_rate": 3.8570479595279623e-07, + "loss": 0.1425, + "step": 2825 + }, + { + "epoch": 0.15, + "learning_rate": 3.8569224003271684e-07, + "loss": 0.1257, + "step": 2826 + }, + { + "epoch": 0.15, + "learning_rate": 3.8567967880548153e-07, + "loss": 0.1085, + "step": 2827 + }, + { + "epoch": 0.15, + "learning_rate": 3.8566711227144933e-07, + "loss": 0.1376, + "step": 2828 + }, + { + "epoch": 0.15, + "learning_rate": 3.856545404309794e-07, + "loss": 0.1462, + "step": 2829 + }, + { + "epoch": 0.15, + "learning_rate": 3.85641963284431e-07, + "loss": 0.1635, + "step": 2830 + }, + { + "epoch": 0.15, + "learning_rate": 3.856293808321636e-07, + "loss": 0.1599, + "step": 2831 + }, + { + "epoch": 0.15, + "learning_rate": 3.856167930745369e-07, + "loss": 0.1572, + "step": 2832 + }, + { + "epoch": 0.15, + "learning_rate": 3.856042000119106e-07, + "loss": 0.1629, + "step": 2833 + }, + { + "epoch": 0.15, + "learning_rate": 3.855916016446446e-07, + "loss": 0.1761, + "step": 2834 + }, + { + "epoch": 0.15, + "learning_rate": 3.8557899797309893e-07, + "loss": 0.1513, + "step": 2835 + }, + { + "epoch": 0.15, + "learning_rate": 3.855663889976339e-07, + "loss": 0.1542, + "step": 2836 + }, + { + "epoch": 0.15, + "learning_rate": 3.8555377471860977e-07, + "loss": 0.1522, + "step": 2837 + }, + { + "epoch": 0.15, + "learning_rate": 3.855411551363872e-07, + "loss": 0.1761, + "step": 2838 + }, + { + "epoch": 0.15, + "learning_rate": 3.855285302513268e-07, + "loss": 0.1588, + "step": 2839 + }, + { + "epoch": 0.15, + "learning_rate": 3.855159000637893e-07, + "loss": 0.1774, + "step": 2840 + }, + { + "epoch": 0.15, + "learning_rate": 3.8550326457413586e-07, + "loss": 0.1437, + "step": 2841 + }, + { + "epoch": 0.15, + "learning_rate": 3.8549062378272745e-07, + "loss": 0.1317, + "step": 2842 + }, + { + "epoch": 0.15, + "learning_rate": 3.854779776899254e-07, + "loss": 0.1426, + "step": 2843 + }, + { + "epoch": 0.15, + "learning_rate": 3.854653262960911e-07, + "loss": 0.1392, + "step": 2844 + }, + { + "epoch": 0.15, + "learning_rate": 3.8545266960158635e-07, + "loss": 0.1444, + "step": 2845 + }, + { + "epoch": 0.15, + "learning_rate": 3.854400076067726e-07, + "loss": 0.1772, + "step": 2846 + }, + { + "epoch": 0.15, + "learning_rate": 3.854273403120118e-07, + "loss": 0.1507, + "step": 2847 + }, + { + "epoch": 0.15, + "learning_rate": 3.854146677176661e-07, + "loss": 0.1681, + "step": 2848 + }, + { + "epoch": 0.15, + "learning_rate": 3.854019898240976e-07, + "loss": 0.1812, + "step": 2849 + }, + { + "epoch": 0.15, + "learning_rate": 3.8538930663166864e-07, + "loss": 0.1394, + "step": 2850 + }, + { + "epoch": 0.15, + "learning_rate": 3.8537661814074176e-07, + "loss": 0.1627, + "step": 2851 + }, + { + "epoch": 0.15, + "learning_rate": 3.853639243516795e-07, + "loss": 0.1725, + "step": 2852 + }, + { + "epoch": 0.15, + "learning_rate": 3.8535122526484485e-07, + "loss": 0.1391, + "step": 2853 + }, + { + "epoch": 0.15, + "learning_rate": 3.8533852088060056e-07, + "loss": 0.1437, + "step": 2854 + }, + { + "epoch": 0.15, + "learning_rate": 3.8532581119930973e-07, + "loss": 0.1468, + "step": 2855 + }, + { + "epoch": 0.15, + "learning_rate": 3.853130962213358e-07, + "loss": 0.1605, + "step": 2856 + }, + { + "epoch": 0.15, + "learning_rate": 3.85300375947042e-07, + "loss": 0.1636, + "step": 2857 + }, + { + "epoch": 0.15, + "learning_rate": 3.852876503767919e-07, + "loss": 0.1258, + "step": 2858 + }, + { + "epoch": 0.15, + "learning_rate": 3.8527491951094924e-07, + "loss": 0.1726, + "step": 2859 + }, + { + "epoch": 0.15, + "learning_rate": 3.8526218334987784e-07, + "loss": 0.1236, + "step": 2860 + }, + { + "epoch": 0.15, + "learning_rate": 3.852494418939417e-07, + "loss": 0.1496, + "step": 2861 + }, + { + "epoch": 0.15, + "learning_rate": 3.8523669514350507e-07, + "loss": 0.1883, + "step": 2862 + }, + { + "epoch": 0.15, + "learning_rate": 3.852239430989321e-07, + "loss": 0.1434, + "step": 2863 + }, + { + "epoch": 0.15, + "learning_rate": 3.852111857605874e-07, + "loss": 0.154, + "step": 2864 + }, + { + "epoch": 0.15, + "learning_rate": 3.8519842312883544e-07, + "loss": 0.1885, + "step": 2865 + }, + { + "epoch": 0.15, + "learning_rate": 3.851856552040411e-07, + "loss": 0.1541, + "step": 2866 + }, + { + "epoch": 0.15, + "learning_rate": 3.8517288198656926e-07, + "loss": 0.1335, + "step": 2867 + }, + { + "epoch": 0.15, + "learning_rate": 3.8516010347678494e-07, + "loss": 0.18, + "step": 2868 + }, + { + "epoch": 0.15, + "learning_rate": 3.8514731967505344e-07, + "loss": 0.1437, + "step": 2869 + }, + { + "epoch": 0.15, + "learning_rate": 3.8513453058174e-07, + "loss": 0.1557, + "step": 2870 + }, + { + "epoch": 0.15, + "learning_rate": 3.851217361972102e-07, + "loss": 0.1545, + "step": 2871 + }, + { + "epoch": 0.15, + "learning_rate": 3.851089365218298e-07, + "loss": 0.1502, + "step": 2872 + }, + { + "epoch": 0.15, + "learning_rate": 3.8509613155596444e-07, + "loss": 0.1402, + "step": 2873 + }, + { + "epoch": 0.15, + "learning_rate": 3.8508332129998025e-07, + "loss": 0.1631, + "step": 2874 + }, + { + "epoch": 0.15, + "learning_rate": 3.8507050575424327e-07, + "loss": 0.163, + "step": 2875 + }, + { + "epoch": 0.15, + "learning_rate": 3.8505768491911974e-07, + "loss": 0.1542, + "step": 2876 + }, + { + "epoch": 0.15, + "learning_rate": 3.8504485879497616e-07, + "loss": 0.1354, + "step": 2877 + }, + { + "epoch": 0.15, + "learning_rate": 3.850320273821792e-07, + "loss": 0.1889, + "step": 2878 + }, + { + "epoch": 0.15, + "learning_rate": 3.850191906810953e-07, + "loss": 0.1552, + "step": 2879 + }, + { + "epoch": 0.15, + "learning_rate": 3.850063486920916e-07, + "loss": 0.1418, + "step": 2880 + }, + { + "epoch": 0.15, + "learning_rate": 3.84993501415535e-07, + "loss": 0.1636, + "step": 2881 + }, + { + "epoch": 0.15, + "learning_rate": 3.849806488517927e-07, + "loss": 0.1614, + "step": 2882 + }, + { + "epoch": 0.15, + "learning_rate": 3.8496779100123213e-07, + "loss": 0.1327, + "step": 2883 + }, + { + "epoch": 0.15, + "learning_rate": 3.8495492786422056e-07, + "loss": 0.1255, + "step": 2884 + }, + { + "epoch": 0.15, + "learning_rate": 3.849420594411259e-07, + "loss": 0.1516, + "step": 2885 + }, + { + "epoch": 0.15, + "learning_rate": 3.849291857323157e-07, + "loss": 0.1669, + "step": 2886 + }, + { + "epoch": 0.15, + "learning_rate": 3.8491630673815795e-07, + "loss": 0.1474, + "step": 2887 + }, + { + "epoch": 0.15, + "learning_rate": 3.849034224590208e-07, + "loss": 0.1797, + "step": 2888 + }, + { + "epoch": 0.15, + "learning_rate": 3.8489053289527245e-07, + "loss": 0.1412, + "step": 2889 + }, + { + "epoch": 0.15, + "learning_rate": 3.848776380472813e-07, + "loss": 0.1555, + "step": 2890 + }, + { + "epoch": 0.15, + "learning_rate": 3.8486473791541593e-07, + "loss": 0.1543, + "step": 2891 + }, + { + "epoch": 0.15, + "learning_rate": 3.84851832500045e-07, + "loss": 0.1305, + "step": 2892 + }, + { + "epoch": 0.15, + "learning_rate": 3.8483892180153727e-07, + "loss": 0.1362, + "step": 2893 + }, + { + "epoch": 0.15, + "learning_rate": 3.848260058202618e-07, + "loss": 0.1731, + "step": 2894 + }, + { + "epoch": 0.15, + "learning_rate": 3.8481308455658776e-07, + "loss": 0.1287, + "step": 2895 + }, + { + "epoch": 0.15, + "learning_rate": 3.8480015801088437e-07, + "loss": 0.1726, + "step": 2896 + }, + { + "epoch": 0.15, + "learning_rate": 3.8478722618352117e-07, + "loss": 0.132, + "step": 2897 + }, + { + "epoch": 0.15, + "learning_rate": 3.847742890748678e-07, + "loss": 0.1356, + "step": 2898 + }, + { + "epoch": 0.15, + "learning_rate": 3.847613466852937e-07, + "loss": 0.1392, + "step": 2899 + }, + { + "epoch": 0.15, + "learning_rate": 3.8474839901516916e-07, + "loss": 0.1793, + "step": 2900 + }, + { + "epoch": 0.15, + "learning_rate": 3.84735446064864e-07, + "loss": 0.1447, + "step": 2901 + }, + { + "epoch": 0.15, + "learning_rate": 3.847224878347485e-07, + "loss": 0.1427, + "step": 2902 + }, + { + "epoch": 0.15, + "learning_rate": 3.8470952432519297e-07, + "loss": 0.1711, + "step": 2903 + }, + { + "epoch": 0.15, + "learning_rate": 3.8469655553656794e-07, + "loss": 0.1899, + "step": 2904 + }, + { + "epoch": 0.15, + "learning_rate": 3.84683581469244e-07, + "loss": 0.1533, + "step": 2905 + }, + { + "epoch": 0.15, + "learning_rate": 3.8467060212359204e-07, + "loss": 0.1599, + "step": 2906 + }, + { + "epoch": 0.15, + "learning_rate": 3.84657617499983e-07, + "loss": 0.1267, + "step": 2907 + }, + { + "epoch": 0.15, + "learning_rate": 3.8464462759878796e-07, + "loss": 0.1479, + "step": 2908 + }, + { + "epoch": 0.15, + "learning_rate": 3.846316324203782e-07, + "loss": 0.1267, + "step": 2909 + }, + { + "epoch": 0.15, + "learning_rate": 3.8461863196512503e-07, + "loss": 0.1116, + "step": 2910 + }, + { + "epoch": 0.15, + "learning_rate": 3.846056262334002e-07, + "loss": 0.1484, + "step": 2911 + }, + { + "epoch": 0.15, + "learning_rate": 3.8459261522557524e-07, + "loss": 0.1406, + "step": 2912 + }, + { + "epoch": 0.15, + "learning_rate": 3.845795989420221e-07, + "loss": 0.1672, + "step": 2913 + }, + { + "epoch": 0.15, + "learning_rate": 3.8456657738311277e-07, + "loss": 0.1552, + "step": 2914 + }, + { + "epoch": 0.15, + "learning_rate": 3.8455355054921934e-07, + "loss": 0.1743, + "step": 2915 + }, + { + "epoch": 0.15, + "learning_rate": 3.845405184407143e-07, + "loss": 0.1901, + "step": 2916 + }, + { + "epoch": 0.15, + "learning_rate": 3.8452748105796997e-07, + "loss": 0.1524, + "step": 2917 + }, + { + "epoch": 0.15, + "learning_rate": 3.84514438401359e-07, + "loss": 0.1587, + "step": 2918 + }, + { + "epoch": 0.15, + "learning_rate": 3.845013904712541e-07, + "loss": 0.1886, + "step": 2919 + }, + { + "epoch": 0.15, + "learning_rate": 3.8448833726802825e-07, + "loss": 0.1851, + "step": 2920 + }, + { + "epoch": 0.15, + "learning_rate": 3.844752787920545e-07, + "loss": 0.1562, + "step": 2921 + }, + { + "epoch": 0.15, + "learning_rate": 3.844622150437061e-07, + "loss": 0.1567, + "step": 2922 + }, + { + "epoch": 0.15, + "learning_rate": 3.8444914602335636e-07, + "loss": 0.1465, + "step": 2923 + }, + { + "epoch": 0.15, + "learning_rate": 3.8443607173137887e-07, + "loss": 0.1543, + "step": 2924 + }, + { + "epoch": 0.15, + "learning_rate": 3.8442299216814716e-07, + "loss": 0.173, + "step": 2925 + }, + { + "epoch": 0.15, + "learning_rate": 3.8440990733403524e-07, + "loss": 0.1513, + "step": 2926 + }, + { + "epoch": 0.15, + "learning_rate": 3.8439681722941694e-07, + "loss": 0.1301, + "step": 2927 + }, + { + "epoch": 0.15, + "learning_rate": 3.843837218546664e-07, + "loss": 0.1338, + "step": 2928 + }, + { + "epoch": 0.15, + "learning_rate": 3.843706212101579e-07, + "loss": 0.1508, + "step": 2929 + }, + { + "epoch": 0.15, + "learning_rate": 3.843575152962659e-07, + "loss": 0.1956, + "step": 2930 + }, + { + "epoch": 0.15, + "learning_rate": 3.8434440411336493e-07, + "loss": 0.1465, + "step": 2931 + }, + { + "epoch": 0.15, + "learning_rate": 3.8433128766182973e-07, + "loss": 0.1177, + "step": 2932 + }, + { + "epoch": 0.15, + "learning_rate": 3.8431816594203516e-07, + "loss": 0.1166, + "step": 2933 + }, + { + "epoch": 0.15, + "learning_rate": 3.8430503895435624e-07, + "loss": 0.1326, + "step": 2934 + }, + { + "epoch": 0.15, + "learning_rate": 3.8429190669916825e-07, + "loss": 0.1576, + "step": 2935 + }, + { + "epoch": 0.15, + "learning_rate": 3.842787691768463e-07, + "loss": 0.1505, + "step": 2936 + }, + { + "epoch": 0.15, + "learning_rate": 3.8426562638776607e-07, + "loss": 0.1505, + "step": 2937 + }, + { + "epoch": 0.15, + "learning_rate": 3.842524783323031e-07, + "loss": 0.1843, + "step": 2938 + }, + { + "epoch": 0.15, + "learning_rate": 3.842393250108331e-07, + "loss": 0.1583, + "step": 2939 + }, + { + "epoch": 0.15, + "learning_rate": 3.8422616642373216e-07, + "loss": 0.1294, + "step": 2940 + }, + { + "epoch": 0.15, + "learning_rate": 3.8421300257137614e-07, + "loss": 0.1377, + "step": 2941 + }, + { + "epoch": 0.15, + "learning_rate": 3.841998334541415e-07, + "loss": 0.1468, + "step": 2942 + }, + { + "epoch": 0.15, + "learning_rate": 3.841866590724045e-07, + "loss": 0.211, + "step": 2943 + }, + { + "epoch": 0.15, + "learning_rate": 3.8417347942654165e-07, + "loss": 0.1482, + "step": 2944 + }, + { + "epoch": 0.15, + "learning_rate": 3.8416029451692973e-07, + "loss": 0.1487, + "step": 2945 + }, + { + "epoch": 0.15, + "learning_rate": 3.841471043439454e-07, + "loss": 0.1676, + "step": 2946 + }, + { + "epoch": 0.15, + "learning_rate": 3.8413390890796577e-07, + "loss": 0.1489, + "step": 2947 + }, + { + "epoch": 0.15, + "learning_rate": 3.84120708209368e-07, + "loss": 0.1228, + "step": 2948 + }, + { + "epoch": 0.15, + "learning_rate": 3.8410750224852927e-07, + "loss": 0.1635, + "step": 2949 + }, + { + "epoch": 0.15, + "learning_rate": 3.8409429102582704e-07, + "loss": 0.1709, + "step": 2950 + }, + { + "epoch": 0.15, + "learning_rate": 3.8408107454163894e-07, + "loss": 0.1229, + "step": 2951 + }, + { + "epoch": 0.15, + "learning_rate": 3.840678527963426e-07, + "loss": 0.1463, + "step": 2952 + }, + { + "epoch": 0.15, + "learning_rate": 3.8405462579031604e-07, + "loss": 0.1362, + "step": 2953 + }, + { + "epoch": 0.15, + "learning_rate": 3.8404139352393717e-07, + "loss": 0.1697, + "step": 2954 + }, + { + "epoch": 0.15, + "learning_rate": 3.840281559975842e-07, + "loss": 0.149, + "step": 2955 + }, + { + "epoch": 0.15, + "learning_rate": 3.840149132116356e-07, + "loss": 0.1319, + "step": 2956 + }, + { + "epoch": 0.15, + "learning_rate": 3.8400166516646966e-07, + "loss": 0.163, + "step": 2957 + }, + { + "epoch": 0.15, + "learning_rate": 3.839884118624652e-07, + "loss": 0.1794, + "step": 2958 + }, + { + "epoch": 0.15, + "learning_rate": 3.839751533000008e-07, + "loss": 0.1494, + "step": 2959 + }, + { + "epoch": 0.15, + "learning_rate": 3.839618894794555e-07, + "loss": 0.1801, + "step": 2960 + }, + { + "epoch": 0.15, + "learning_rate": 3.839486204012084e-07, + "loss": 0.2265, + "step": 2961 + }, + { + "epoch": 0.15, + "learning_rate": 3.839353460656387e-07, + "loss": 0.1731, + "step": 2962 + }, + { + "epoch": 0.15, + "learning_rate": 3.8392206647312586e-07, + "loss": 0.1473, + "step": 2963 + }, + { + "epoch": 0.15, + "learning_rate": 3.839087816240493e-07, + "loss": 0.1702, + "step": 2964 + }, + { + "epoch": 0.15, + "learning_rate": 3.838954915187888e-07, + "loss": 0.149, + "step": 2965 + }, + { + "epoch": 0.15, + "learning_rate": 3.838821961577241e-07, + "loss": 0.1468, + "step": 2966 + }, + { + "epoch": 0.15, + "learning_rate": 3.8386889554123534e-07, + "loss": 0.134, + "step": 2967 + }, + { + "epoch": 0.15, + "learning_rate": 3.838555896697025e-07, + "loss": 0.1361, + "step": 2968 + }, + { + "epoch": 0.15, + "learning_rate": 3.838422785435059e-07, + "loss": 0.1268, + "step": 2969 + }, + { + "epoch": 0.16, + "learning_rate": 3.838289621630261e-07, + "loss": 0.1231, + "step": 2970 + }, + { + "epoch": 0.16, + "learning_rate": 3.838156405286436e-07, + "loss": 0.1239, + "step": 2971 + }, + { + "epoch": 0.16, + "learning_rate": 3.838023136407391e-07, + "loss": 0.1505, + "step": 2972 + }, + { + "epoch": 0.16, + "learning_rate": 3.837889814996935e-07, + "loss": 0.169, + "step": 2973 + }, + { + "epoch": 0.16, + "learning_rate": 3.837756441058879e-07, + "loss": 0.1842, + "step": 2974 + }, + { + "epoch": 0.16, + "learning_rate": 3.8376230145970344e-07, + "loss": 0.201, + "step": 2975 + }, + { + "epoch": 0.16, + "learning_rate": 3.8374895356152137e-07, + "loss": 0.1482, + "step": 2976 + }, + { + "epoch": 0.16, + "learning_rate": 3.8373560041172344e-07, + "loss": 0.1637, + "step": 2977 + }, + { + "epoch": 0.16, + "learning_rate": 3.8372224201069096e-07, + "loss": 0.1426, + "step": 2978 + }, + { + "epoch": 0.16, + "learning_rate": 3.8370887835880595e-07, + "loss": 0.1875, + "step": 2979 + }, + { + "epoch": 0.16, + "learning_rate": 3.8369550945645027e-07, + "loss": 0.1374, + "step": 2980 + }, + { + "epoch": 0.16, + "learning_rate": 3.836821353040061e-07, + "loss": 0.1975, + "step": 2981 + }, + { + "epoch": 0.16, + "learning_rate": 3.836687559018555e-07, + "loss": 0.124, + "step": 2982 + }, + { + "epoch": 0.16, + "learning_rate": 3.8365537125038095e-07, + "loss": 0.2005, + "step": 2983 + }, + { + "epoch": 0.16, + "learning_rate": 3.83641981349965e-07, + "loss": 0.1364, + "step": 2984 + }, + { + "epoch": 0.16, + "learning_rate": 3.8362858620099034e-07, + "loss": 0.1511, + "step": 2985 + }, + { + "epoch": 0.16, + "learning_rate": 3.836151858038398e-07, + "loss": 0.1999, + "step": 2986 + }, + { + "epoch": 0.16, + "learning_rate": 3.836017801588964e-07, + "loss": 0.1364, + "step": 2987 + }, + { + "epoch": 0.16, + "learning_rate": 3.835883692665432e-07, + "loss": 0.1578, + "step": 2988 + }, + { + "epoch": 0.16, + "learning_rate": 3.8357495312716353e-07, + "loss": 0.1738, + "step": 2989 + }, + { + "epoch": 0.16, + "learning_rate": 3.835615317411409e-07, + "loss": 0.1684, + "step": 2990 + }, + { + "epoch": 0.16, + "learning_rate": 3.835481051088587e-07, + "loss": 0.1491, + "step": 2991 + }, + { + "epoch": 0.16, + "learning_rate": 3.835346732307009e-07, + "loss": 0.1557, + "step": 2992 + }, + { + "epoch": 0.16, + "learning_rate": 3.835212361070513e-07, + "loss": 0.1241, + "step": 2993 + }, + { + "epoch": 0.16, + "learning_rate": 3.835077937382939e-07, + "loss": 0.1165, + "step": 2994 + }, + { + "epoch": 0.16, + "learning_rate": 3.834943461248129e-07, + "loss": 0.1855, + "step": 2995 + }, + { + "epoch": 0.16, + "learning_rate": 3.8348089326699267e-07, + "loss": 0.1328, + "step": 2996 + }, + { + "epoch": 0.16, + "learning_rate": 3.8346743516521763e-07, + "loss": 0.1553, + "step": 2997 + }, + { + "epoch": 0.16, + "learning_rate": 3.8345397181987253e-07, + "loss": 0.1396, + "step": 2998 + }, + { + "epoch": 0.16, + "learning_rate": 3.8344050323134203e-07, + "loss": 0.1599, + "step": 2999 + }, + { + "epoch": 0.16, + "learning_rate": 3.834270294000112e-07, + "loss": 0.1625, + "step": 3000 + }, + { + "epoch": 0.16, + "learning_rate": 3.834135503262651e-07, + "loss": 0.1643, + "step": 3001 + }, + { + "epoch": 0.16, + "learning_rate": 3.834000660104888e-07, + "loss": 0.1279, + "step": 3002 + }, + { + "epoch": 0.16, + "learning_rate": 3.833865764530679e-07, + "loss": 0.137, + "step": 3003 + }, + { + "epoch": 0.16, + "learning_rate": 3.8337308165438783e-07, + "loss": 0.125, + "step": 3004 + }, + { + "epoch": 0.16, + "learning_rate": 3.833595816148343e-07, + "loss": 0.1406, + "step": 3005 + }, + { + "epoch": 0.16, + "learning_rate": 3.833460763347931e-07, + "loss": 0.1538, + "step": 3006 + }, + { + "epoch": 0.16, + "learning_rate": 3.8333256581465026e-07, + "loss": 0.1691, + "step": 3007 + }, + { + "epoch": 0.16, + "learning_rate": 3.83319050054792e-07, + "loss": 0.118, + "step": 3008 + }, + { + "epoch": 0.16, + "learning_rate": 3.833055290556045e-07, + "loss": 0.1697, + "step": 3009 + }, + { + "epoch": 0.16, + "learning_rate": 3.8329200281747414e-07, + "loss": 0.1715, + "step": 3010 + }, + { + "epoch": 0.16, + "learning_rate": 3.8327847134078765e-07, + "loss": 0.1733, + "step": 3011 + }, + { + "epoch": 0.16, + "learning_rate": 3.832649346259317e-07, + "loss": 0.1276, + "step": 3012 + }, + { + "epoch": 0.16, + "learning_rate": 3.8325139267329316e-07, + "loss": 0.1472, + "step": 3013 + }, + { + "epoch": 0.16, + "learning_rate": 3.8323784548325903e-07, + "loss": 0.1826, + "step": 3014 + }, + { + "epoch": 0.16, + "learning_rate": 3.832242930562166e-07, + "loss": 0.1346, + "step": 3015 + }, + { + "epoch": 0.16, + "learning_rate": 3.832107353925531e-07, + "loss": 0.126, + "step": 3016 + }, + { + "epoch": 0.16, + "learning_rate": 3.8319717249265607e-07, + "loss": 0.1521, + "step": 3017 + }, + { + "epoch": 0.16, + "learning_rate": 3.831836043569131e-07, + "loss": 0.1228, + "step": 3018 + }, + { + "epoch": 0.16, + "learning_rate": 3.8317003098571205e-07, + "loss": 0.1502, + "step": 3019 + }, + { + "epoch": 0.16, + "learning_rate": 3.831564523794408e-07, + "loss": 0.1534, + "step": 3020 + }, + { + "epoch": 0.16, + "learning_rate": 3.8314286853848743e-07, + "loss": 0.177, + "step": 3021 + }, + { + "epoch": 0.16, + "learning_rate": 3.831292794632402e-07, + "loss": 0.205, + "step": 3022 + }, + { + "epoch": 0.16, + "learning_rate": 3.831156851540874e-07, + "loss": 0.1622, + "step": 3023 + }, + { + "epoch": 0.16, + "learning_rate": 3.8310208561141763e-07, + "loss": 0.2051, + "step": 3024 + }, + { + "epoch": 0.16, + "learning_rate": 3.830884808356196e-07, + "loss": 0.1888, + "step": 3025 + }, + { + "epoch": 0.16, + "learning_rate": 3.8307487082708213e-07, + "loss": 0.1262, + "step": 3026 + }, + { + "epoch": 0.16, + "learning_rate": 3.8306125558619417e-07, + "loss": 0.1526, + "step": 3027 + }, + { + "epoch": 0.16, + "learning_rate": 3.8304763511334484e-07, + "loss": 0.1852, + "step": 3028 + }, + { + "epoch": 0.16, + "learning_rate": 3.8303400940892347e-07, + "loss": 0.1383, + "step": 3029 + }, + { + "epoch": 0.16, + "learning_rate": 3.830203784733194e-07, + "loss": 0.1255, + "step": 3030 + }, + { + "epoch": 0.16, + "learning_rate": 3.8300674230692226e-07, + "loss": 0.1987, + "step": 3031 + }, + { + "epoch": 0.16, + "learning_rate": 3.8299310091012185e-07, + "loss": 0.1637, + "step": 3032 + }, + { + "epoch": 0.16, + "learning_rate": 3.8297945428330794e-07, + "loss": 0.1488, + "step": 3033 + }, + { + "epoch": 0.16, + "learning_rate": 3.829658024268705e-07, + "loss": 0.123, + "step": 3034 + }, + { + "epoch": 0.16, + "learning_rate": 3.8295214534119993e-07, + "loss": 0.1596, + "step": 3035 + }, + { + "epoch": 0.16, + "learning_rate": 3.829384830266863e-07, + "loss": 0.1537, + "step": 3036 + }, + { + "epoch": 0.16, + "learning_rate": 3.8292481548372033e-07, + "loss": 0.135, + "step": 3037 + }, + { + "epoch": 0.16, + "learning_rate": 3.829111427126924e-07, + "loss": 0.1282, + "step": 3038 + }, + { + "epoch": 0.16, + "learning_rate": 3.828974647139935e-07, + "loss": 0.1231, + "step": 3039 + }, + { + "epoch": 0.16, + "learning_rate": 3.828837814880144e-07, + "loss": 0.196, + "step": 3040 + }, + { + "epoch": 0.16, + "learning_rate": 3.828700930351462e-07, + "loss": 0.1562, + "step": 3041 + }, + { + "epoch": 0.16, + "learning_rate": 3.828563993557802e-07, + "loss": 0.1062, + "step": 3042 + }, + { + "epoch": 0.16, + "learning_rate": 3.828427004503077e-07, + "loss": 0.1523, + "step": 3043 + }, + { + "epoch": 0.16, + "learning_rate": 3.828289963191202e-07, + "loss": 0.1534, + "step": 3044 + }, + { + "epoch": 0.16, + "learning_rate": 3.828152869626094e-07, + "loss": 0.1201, + "step": 3045 + }, + { + "epoch": 0.16, + "learning_rate": 3.828015723811672e-07, + "loss": 0.135, + "step": 3046 + }, + { + "epoch": 0.16, + "learning_rate": 3.8278785257518544e-07, + "loss": 0.1662, + "step": 3047 + }, + { + "epoch": 0.16, + "learning_rate": 3.827741275450563e-07, + "loss": 0.1269, + "step": 3048 + }, + { + "epoch": 0.16, + "learning_rate": 3.8276039729117205e-07, + "loss": 0.1188, + "step": 3049 + }, + { + "epoch": 0.16, + "learning_rate": 3.827466618139251e-07, + "loss": 0.1327, + "step": 3050 + }, + { + "epoch": 0.16, + "learning_rate": 3.82732921113708e-07, + "loss": 0.1272, + "step": 3051 + }, + { + "epoch": 0.16, + "learning_rate": 3.8271917519091344e-07, + "loss": 0.1515, + "step": 3052 + }, + { + "epoch": 0.16, + "learning_rate": 3.827054240459343e-07, + "loss": 0.173, + "step": 3053 + }, + { + "epoch": 0.16, + "learning_rate": 3.826916676791637e-07, + "loss": 0.1806, + "step": 3054 + }, + { + "epoch": 0.16, + "learning_rate": 3.8267790609099464e-07, + "loss": 0.1467, + "step": 3055 + }, + { + "epoch": 0.16, + "learning_rate": 3.826641392818205e-07, + "loss": 0.1582, + "step": 3056 + }, + { + "epoch": 0.16, + "learning_rate": 3.8265036725203477e-07, + "loss": 0.147, + "step": 3057 + }, + { + "epoch": 0.16, + "learning_rate": 3.8263659000203103e-07, + "loss": 0.1344, + "step": 3058 + }, + { + "epoch": 0.16, + "learning_rate": 3.8262280753220307e-07, + "loss": 0.1163, + "step": 3059 + }, + { + "epoch": 0.16, + "learning_rate": 3.8260901984294474e-07, + "loss": 0.1183, + "step": 3060 + }, + { + "epoch": 0.16, + "learning_rate": 3.8259522693465013e-07, + "loss": 0.1441, + "step": 3061 + }, + { + "epoch": 0.16, + "learning_rate": 3.825814288077135e-07, + "loss": 0.1269, + "step": 3062 + }, + { + "epoch": 0.16, + "learning_rate": 3.8256762546252904e-07, + "loss": 0.1455, + "step": 3063 + }, + { + "epoch": 0.16, + "learning_rate": 3.8255381689949146e-07, + "loss": 0.1536, + "step": 3064 + }, + { + "epoch": 0.16, + "learning_rate": 3.8254000311899526e-07, + "loss": 0.1586, + "step": 3065 + }, + { + "epoch": 0.16, + "learning_rate": 3.8252618412143535e-07, + "loss": 0.1505, + "step": 3066 + }, + { + "epoch": 0.16, + "learning_rate": 3.8251235990720663e-07, + "loss": 0.1515, + "step": 3067 + }, + { + "epoch": 0.16, + "learning_rate": 3.8249853047670417e-07, + "loss": 0.1578, + "step": 3068 + }, + { + "epoch": 0.16, + "learning_rate": 3.824846958303233e-07, + "loss": 0.1602, + "step": 3069 + }, + { + "epoch": 0.16, + "learning_rate": 3.8247085596845934e-07, + "loss": 0.1769, + "step": 3070 + }, + { + "epoch": 0.16, + "learning_rate": 3.824570108915079e-07, + "loss": 0.1624, + "step": 3071 + }, + { + "epoch": 0.16, + "learning_rate": 3.824431605998646e-07, + "loss": 0.1295, + "step": 3072 + }, + { + "epoch": 0.16, + "learning_rate": 3.8242930509392546e-07, + "loss": 0.1276, + "step": 3073 + }, + { + "epoch": 0.16, + "learning_rate": 3.824154443740863e-07, + "loss": 0.1761, + "step": 3074 + }, + { + "epoch": 0.16, + "learning_rate": 3.824015784407432e-07, + "loss": 0.1723, + "step": 3075 + }, + { + "epoch": 0.16, + "learning_rate": 3.823877072942927e-07, + "loss": 0.1385, + "step": 3076 + }, + { + "epoch": 0.16, + "learning_rate": 3.823738309351311e-07, + "loss": 0.1393, + "step": 3077 + }, + { + "epoch": 0.16, + "learning_rate": 3.82359949363655e-07, + "loss": 0.1413, + "step": 3078 + }, + { + "epoch": 0.16, + "learning_rate": 3.823460625802612e-07, + "loss": 0.1289, + "step": 3079 + }, + { + "epoch": 0.16, + "learning_rate": 3.8233217058534645e-07, + "loss": 0.166, + "step": 3080 + }, + { + "epoch": 0.16, + "learning_rate": 3.823182733793079e-07, + "loss": 0.1239, + "step": 3081 + }, + { + "epoch": 0.16, + "learning_rate": 3.823043709625427e-07, + "loss": 0.1547, + "step": 3082 + }, + { + "epoch": 0.16, + "learning_rate": 3.822904633354483e-07, + "loss": 0.147, + "step": 3083 + }, + { + "epoch": 0.16, + "learning_rate": 3.82276550498422e-07, + "loss": 0.1618, + "step": 3084 + }, + { + "epoch": 0.16, + "learning_rate": 3.822626324518615e-07, + "loss": 0.1705, + "step": 3085 + }, + { + "epoch": 0.16, + "learning_rate": 3.822487091961647e-07, + "loss": 0.1372, + "step": 3086 + }, + { + "epoch": 0.16, + "learning_rate": 3.8223478073172935e-07, + "loss": 0.1319, + "step": 3087 + }, + { + "epoch": 0.16, + "learning_rate": 3.822208470589537e-07, + "loss": 0.164, + "step": 3088 + }, + { + "epoch": 0.16, + "learning_rate": 3.822069081782358e-07, + "loss": 0.1222, + "step": 3089 + }, + { + "epoch": 0.16, + "learning_rate": 3.821929640899742e-07, + "loss": 0.1087, + "step": 3090 + }, + { + "epoch": 0.16, + "learning_rate": 3.8217901479456727e-07, + "loss": 0.1346, + "step": 3091 + }, + { + "epoch": 0.16, + "learning_rate": 3.8216506029241384e-07, + "loss": 0.1146, + "step": 3092 + }, + { + "epoch": 0.16, + "learning_rate": 3.8215110058391263e-07, + "loss": 0.1396, + "step": 3093 + }, + { + "epoch": 0.16, + "learning_rate": 3.8213713566946264e-07, + "loss": 0.1116, + "step": 3094 + }, + { + "epoch": 0.16, + "learning_rate": 3.821231655494631e-07, + "loss": 0.1527, + "step": 3095 + }, + { + "epoch": 0.16, + "learning_rate": 3.8210919022431306e-07, + "loss": 0.1241, + "step": 3096 + }, + { + "epoch": 0.16, + "learning_rate": 3.820952096944121e-07, + "loss": 0.1564, + "step": 3097 + }, + { + "epoch": 0.16, + "learning_rate": 3.820812239601598e-07, + "loss": 0.1387, + "step": 3098 + }, + { + "epoch": 0.16, + "learning_rate": 3.8206723302195573e-07, + "loss": 0.164, + "step": 3099 + }, + { + "epoch": 0.16, + "learning_rate": 3.8205323688019995e-07, + "loss": 0.1653, + "step": 3100 + }, + { + "epoch": 0.16, + "learning_rate": 3.8203923553529235e-07, + "loss": 0.1404, + "step": 3101 + }, + { + "epoch": 0.16, + "learning_rate": 3.820252289876331e-07, + "loss": 0.1635, + "step": 3102 + }, + { + "epoch": 0.16, + "learning_rate": 3.8201121723762253e-07, + "loss": 0.1403, + "step": 3103 + }, + { + "epoch": 0.16, + "learning_rate": 3.8199720028566113e-07, + "loss": 0.1515, + "step": 3104 + }, + { + "epoch": 0.16, + "learning_rate": 3.819831781321495e-07, + "loss": 0.163, + "step": 3105 + }, + { + "epoch": 0.16, + "learning_rate": 3.819691507774884e-07, + "loss": 0.1254, + "step": 3106 + }, + { + "epoch": 0.16, + "learning_rate": 3.8195511822207874e-07, + "loss": 0.1484, + "step": 3107 + }, + { + "epoch": 0.16, + "learning_rate": 3.8194108046632143e-07, + "loss": 0.1417, + "step": 3108 + }, + { + "epoch": 0.16, + "learning_rate": 3.819270375106179e-07, + "loss": 0.1349, + "step": 3109 + }, + { + "epoch": 0.16, + "learning_rate": 3.8191298935536944e-07, + "loss": 0.1653, + "step": 3110 + }, + { + "epoch": 0.16, + "learning_rate": 3.818989360009775e-07, + "loss": 0.1565, + "step": 3111 + }, + { + "epoch": 0.16, + "learning_rate": 3.8188487744784367e-07, + "loss": 0.1249, + "step": 3112 + }, + { + "epoch": 0.16, + "learning_rate": 3.8187081369636993e-07, + "loss": 0.1717, + "step": 3113 + }, + { + "epoch": 0.16, + "learning_rate": 3.818567447469581e-07, + "loss": 0.149, + "step": 3114 + }, + { + "epoch": 0.16, + "learning_rate": 3.818426706000102e-07, + "loss": 0.1602, + "step": 3115 + }, + { + "epoch": 0.16, + "learning_rate": 3.818285912559287e-07, + "loss": 0.1536, + "step": 3116 + }, + { + "epoch": 0.16, + "learning_rate": 3.8181450671511587e-07, + "loss": 0.16, + "step": 3117 + }, + { + "epoch": 0.16, + "learning_rate": 3.818004169779742e-07, + "loss": 0.1925, + "step": 3118 + }, + { + "epoch": 0.16, + "learning_rate": 3.8178632204490644e-07, + "loss": 0.1276, + "step": 3119 + }, + { + "epoch": 0.16, + "learning_rate": 3.817722219163154e-07, + "loss": 0.1668, + "step": 3120 + }, + { + "epoch": 0.16, + "learning_rate": 3.8175811659260415e-07, + "loss": 0.1482, + "step": 3121 + }, + { + "epoch": 0.16, + "learning_rate": 3.817440060741757e-07, + "loss": 0.2208, + "step": 3122 + }, + { + "epoch": 0.16, + "learning_rate": 3.817298903614334e-07, + "loss": 0.1563, + "step": 3123 + }, + { + "epoch": 0.16, + "learning_rate": 3.817157694547807e-07, + "loss": 0.1678, + "step": 3124 + }, + { + "epoch": 0.16, + "learning_rate": 3.8170164335462116e-07, + "loss": 0.1584, + "step": 3125 + }, + { + "epoch": 0.16, + "learning_rate": 3.816875120613585e-07, + "loss": 0.1396, + "step": 3126 + }, + { + "epoch": 0.16, + "learning_rate": 3.816733755753966e-07, + "loss": 0.1554, + "step": 3127 + }, + { + "epoch": 0.16, + "learning_rate": 3.8165923389713946e-07, + "loss": 0.1393, + "step": 3128 + }, + { + "epoch": 0.16, + "learning_rate": 3.8164508702699134e-07, + "loss": 0.1485, + "step": 3129 + }, + { + "epoch": 0.16, + "learning_rate": 3.816309349653565e-07, + "loss": 0.1249, + "step": 3130 + }, + { + "epoch": 0.16, + "learning_rate": 3.816167777126394e-07, + "loss": 0.1779, + "step": 3131 + }, + { + "epoch": 0.16, + "learning_rate": 3.816026152692447e-07, + "loss": 0.1411, + "step": 3132 + }, + { + "epoch": 0.16, + "learning_rate": 3.815884476355772e-07, + "loss": 0.1839, + "step": 3133 + }, + { + "epoch": 0.16, + "learning_rate": 3.8157427481204174e-07, + "loss": 0.1839, + "step": 3134 + }, + { + "epoch": 0.16, + "learning_rate": 3.815600967990433e-07, + "loss": 0.1398, + "step": 3135 + }, + { + "epoch": 0.16, + "learning_rate": 3.8154591359698736e-07, + "loss": 0.1404, + "step": 3136 + }, + { + "epoch": 0.16, + "learning_rate": 3.81531725206279e-07, + "loss": 0.1511, + "step": 3137 + }, + { + "epoch": 0.16, + "learning_rate": 3.8151753162732396e-07, + "loss": 0.1634, + "step": 3138 + }, + { + "epoch": 0.16, + "learning_rate": 3.8150333286052776e-07, + "loss": 0.163, + "step": 3139 + }, + { + "epoch": 0.16, + "learning_rate": 3.814891289062962e-07, + "loss": 0.1249, + "step": 3140 + }, + { + "epoch": 0.16, + "learning_rate": 3.8147491976503533e-07, + "loss": 0.1308, + "step": 3141 + }, + { + "epoch": 0.16, + "learning_rate": 3.8146070543715116e-07, + "loss": 0.1343, + "step": 3142 + }, + { + "epoch": 0.16, + "learning_rate": 3.8144648592305e-07, + "loss": 0.1326, + "step": 3143 + }, + { + "epoch": 0.16, + "learning_rate": 3.814322612231382e-07, + "loss": 0.1428, + "step": 3144 + }, + { + "epoch": 0.16, + "learning_rate": 3.8141803133782235e-07, + "loss": 0.1402, + "step": 3145 + }, + { + "epoch": 0.16, + "learning_rate": 3.814037962675091e-07, + "loss": 0.1584, + "step": 3146 + }, + { + "epoch": 0.16, + "learning_rate": 3.8138955601260537e-07, + "loss": 0.1303, + "step": 3147 + }, + { + "epoch": 0.16, + "learning_rate": 3.813753105735181e-07, + "loss": 0.1264, + "step": 3148 + }, + { + "epoch": 0.16, + "learning_rate": 3.8136105995065446e-07, + "loss": 0.1526, + "step": 3149 + }, + { + "epoch": 0.16, + "learning_rate": 3.813468041444216e-07, + "loss": 0.1227, + "step": 3150 + }, + { + "epoch": 0.16, + "learning_rate": 3.813325431552272e-07, + "loss": 0.142, + "step": 3151 + }, + { + "epoch": 0.16, + "learning_rate": 3.8131827698347865e-07, + "loss": 0.1406, + "step": 3152 + }, + { + "epoch": 0.16, + "learning_rate": 3.813040056295838e-07, + "loss": 0.1218, + "step": 3153 + }, + { + "epoch": 0.16, + "learning_rate": 3.8128972909395037e-07, + "loss": 0.1403, + "step": 3154 + }, + { + "epoch": 0.16, + "learning_rate": 3.812754473769866e-07, + "loss": 0.1332, + "step": 3155 + }, + { + "epoch": 0.16, + "learning_rate": 3.8126116047910056e-07, + "loss": 0.137, + "step": 3156 + }, + { + "epoch": 0.16, + "learning_rate": 3.8124686840070055e-07, + "loss": 0.1446, + "step": 3157 + }, + { + "epoch": 0.16, + "learning_rate": 3.812325711421951e-07, + "loss": 0.159, + "step": 3158 + }, + { + "epoch": 0.16, + "learning_rate": 3.8121826870399273e-07, + "loss": 0.1859, + "step": 3159 + }, + { + "epoch": 0.16, + "learning_rate": 3.8120396108650236e-07, + "loss": 0.1324, + "step": 3160 + }, + { + "epoch": 0.16, + "learning_rate": 3.8118964829013275e-07, + "loss": 0.1636, + "step": 3161 + }, + { + "epoch": 0.17, + "learning_rate": 3.8117533031529317e-07, + "loss": 0.1567, + "step": 3162 + }, + { + "epoch": 0.17, + "learning_rate": 3.8116100716239263e-07, + "loss": 0.1336, + "step": 3163 + }, + { + "epoch": 0.17, + "learning_rate": 3.811466788318405e-07, + "loss": 0.1372, + "step": 3164 + }, + { + "epoch": 0.17, + "learning_rate": 3.811323453240465e-07, + "loss": 0.1297, + "step": 3165 + }, + { + "epoch": 0.17, + "learning_rate": 3.8111800663942e-07, + "loss": 0.1392, + "step": 3166 + }, + { + "epoch": 0.17, + "learning_rate": 3.811036627783711e-07, + "loss": 0.1917, + "step": 3167 + }, + { + "epoch": 0.17, + "learning_rate": 3.8108931374130955e-07, + "loss": 0.1569, + "step": 3168 + }, + { + "epoch": 0.17, + "learning_rate": 3.810749595286455e-07, + "loss": 0.1326, + "step": 3169 + }, + { + "epoch": 0.17, + "learning_rate": 3.810606001407892e-07, + "loss": 0.2104, + "step": 3170 + }, + { + "epoch": 0.17, + "learning_rate": 3.8104623557815106e-07, + "loss": 0.173, + "step": 3171 + }, + { + "epoch": 0.17, + "learning_rate": 3.810318658411416e-07, + "loss": 0.1535, + "step": 3172 + }, + { + "epoch": 0.17, + "learning_rate": 3.8101749093017157e-07, + "loss": 0.1491, + "step": 3173 + }, + { + "epoch": 0.17, + "learning_rate": 3.810031108456517e-07, + "loss": 0.1577, + "step": 3174 + }, + { + "epoch": 0.17, + "learning_rate": 3.809887255879931e-07, + "loss": 0.1235, + "step": 3175 + }, + { + "epoch": 0.17, + "learning_rate": 3.8097433515760684e-07, + "loss": 0.1406, + "step": 3176 + }, + { + "epoch": 0.17, + "learning_rate": 3.809599395549042e-07, + "loss": 0.1539, + "step": 3177 + }, + { + "epoch": 0.17, + "learning_rate": 3.8094553878029664e-07, + "loss": 0.1617, + "step": 3178 + }, + { + "epoch": 0.17, + "learning_rate": 3.8093113283419573e-07, + "loss": 0.1259, + "step": 3179 + }, + { + "epoch": 0.17, + "learning_rate": 3.809167217170132e-07, + "loss": 0.1403, + "step": 3180 + }, + { + "epoch": 0.17, + "learning_rate": 3.809023054291609e-07, + "loss": 0.1389, + "step": 3181 + }, + { + "epoch": 0.17, + "learning_rate": 3.808878839710509e-07, + "loss": 0.1519, + "step": 3182 + }, + { + "epoch": 0.17, + "learning_rate": 3.808734573430954e-07, + "loss": 0.1544, + "step": 3183 + }, + { + "epoch": 0.17, + "learning_rate": 3.808590255457066e-07, + "loss": 0.1357, + "step": 3184 + }, + { + "epoch": 0.17, + "learning_rate": 3.8084458857929704e-07, + "loss": 0.1658, + "step": 3185 + }, + { + "epoch": 0.17, + "learning_rate": 3.8083014644427937e-07, + "loss": 0.1746, + "step": 3186 + }, + { + "epoch": 0.17, + "learning_rate": 3.808156991410662e-07, + "loss": 0.1412, + "step": 3187 + }, + { + "epoch": 0.17, + "learning_rate": 3.808012466700706e-07, + "loss": 0.1933, + "step": 3188 + }, + { + "epoch": 0.17, + "learning_rate": 3.807867890317056e-07, + "loss": 0.1486, + "step": 3189 + }, + { + "epoch": 0.17, + "learning_rate": 3.807723262263844e-07, + "loss": 0.1479, + "step": 3190 + }, + { + "epoch": 0.17, + "learning_rate": 3.807578582545203e-07, + "loss": 0.1327, + "step": 3191 + }, + { + "epoch": 0.17, + "learning_rate": 3.8074338511652686e-07, + "loss": 0.1734, + "step": 3192 + }, + { + "epoch": 0.17, + "learning_rate": 3.8072890681281763e-07, + "loss": 0.1486, + "step": 3193 + }, + { + "epoch": 0.17, + "learning_rate": 3.807144233438065e-07, + "loss": 0.1628, + "step": 3194 + }, + { + "epoch": 0.17, + "learning_rate": 3.806999347099074e-07, + "loss": 0.1419, + "step": 3195 + }, + { + "epoch": 0.17, + "learning_rate": 3.8068544091153443e-07, + "loss": 0.1957, + "step": 3196 + }, + { + "epoch": 0.17, + "learning_rate": 3.8067094194910175e-07, + "loss": 0.1136, + "step": 3197 + }, + { + "epoch": 0.17, + "learning_rate": 3.8065643782302387e-07, + "loss": 0.1473, + "step": 3198 + }, + { + "epoch": 0.17, + "learning_rate": 3.806419285337152e-07, + "loss": 0.152, + "step": 3199 + }, + { + "epoch": 0.17, + "learning_rate": 3.8062741408159053e-07, + "loss": 0.1735, + "step": 3200 + }, + { + "epoch": 0.17, + "learning_rate": 3.806128944670646e-07, + "loss": 0.1632, + "step": 3201 + }, + { + "epoch": 0.17, + "learning_rate": 3.8059836969055247e-07, + "loss": 0.1566, + "step": 3202 + }, + { + "epoch": 0.17, + "learning_rate": 3.8058383975246917e-07, + "loss": 0.1508, + "step": 3203 + }, + { + "epoch": 0.17, + "learning_rate": 3.8056930465323005e-07, + "loss": 0.1317, + "step": 3204 + }, + { + "epoch": 0.17, + "learning_rate": 3.805547643932504e-07, + "loss": 0.1802, + "step": 3205 + }, + { + "epoch": 0.17, + "learning_rate": 3.8054021897294597e-07, + "loss": 0.161, + "step": 3206 + }, + { + "epoch": 0.17, + "learning_rate": 3.805256683927324e-07, + "loss": 0.1368, + "step": 3207 + }, + { + "epoch": 0.17, + "learning_rate": 3.8051111265302547e-07, + "loss": 0.1475, + "step": 3208 + }, + { + "epoch": 0.17, + "learning_rate": 3.804965517542413e-07, + "loss": 0.1879, + "step": 3209 + }, + { + "epoch": 0.17, + "learning_rate": 3.80481985696796e-07, + "loss": 0.1313, + "step": 3210 + }, + { + "epoch": 0.17, + "learning_rate": 3.804674144811059e-07, + "loss": 0.1169, + "step": 3211 + }, + { + "epoch": 0.17, + "learning_rate": 3.804528381075874e-07, + "loss": 0.1135, + "step": 3212 + }, + { + "epoch": 0.17, + "learning_rate": 3.8043825657665713e-07, + "loss": 0.161, + "step": 3213 + }, + { + "epoch": 0.17, + "learning_rate": 3.804236698887318e-07, + "loss": 0.1246, + "step": 3214 + }, + { + "epoch": 0.17, + "learning_rate": 3.8040907804422835e-07, + "loss": 0.1365, + "step": 3215 + }, + { + "epoch": 0.17, + "learning_rate": 3.8039448104356384e-07, + "loss": 0.1214, + "step": 3216 + }, + { + "epoch": 0.17, + "learning_rate": 3.803798788871554e-07, + "loss": 0.1557, + "step": 3217 + }, + { + "epoch": 0.17, + "learning_rate": 3.8036527157542035e-07, + "loss": 0.1533, + "step": 3218 + }, + { + "epoch": 0.17, + "learning_rate": 3.803506591087763e-07, + "loss": 0.1496, + "step": 3219 + }, + { + "epoch": 0.17, + "learning_rate": 3.8033604148764075e-07, + "loss": 0.1329, + "step": 3220 + }, + { + "epoch": 0.17, + "learning_rate": 3.803214187124315e-07, + "loss": 0.1495, + "step": 3221 + }, + { + "epoch": 0.17, + "learning_rate": 3.8030679078356646e-07, + "loss": 0.1469, + "step": 3222 + }, + { + "epoch": 0.17, + "learning_rate": 3.802921577014638e-07, + "loss": 0.1328, + "step": 3223 + }, + { + "epoch": 0.17, + "learning_rate": 3.8027751946654163e-07, + "loss": 0.1558, + "step": 3224 + }, + { + "epoch": 0.17, + "learning_rate": 3.8026287607921833e-07, + "loss": 0.1609, + "step": 3225 + }, + { + "epoch": 0.17, + "learning_rate": 3.802482275399125e-07, + "loss": 0.1469, + "step": 3226 + }, + { + "epoch": 0.17, + "learning_rate": 3.802335738490427e-07, + "loss": 0.1734, + "step": 3227 + }, + { + "epoch": 0.17, + "learning_rate": 3.8021891500702787e-07, + "loss": 0.1363, + "step": 3228 + }, + { + "epoch": 0.17, + "learning_rate": 3.802042510142868e-07, + "loss": 0.148, + "step": 3229 + }, + { + "epoch": 0.17, + "learning_rate": 3.8018958187123864e-07, + "loss": 0.1184, + "step": 3230 + }, + { + "epoch": 0.17, + "learning_rate": 3.8017490757830275e-07, + "loss": 0.136, + "step": 3231 + }, + { + "epoch": 0.17, + "learning_rate": 3.801602281358984e-07, + "loss": 0.2018, + "step": 3232 + }, + { + "epoch": 0.17, + "learning_rate": 3.8014554354444525e-07, + "loss": 0.1839, + "step": 3233 + }, + { + "epoch": 0.17, + "learning_rate": 3.801308538043629e-07, + "loss": 0.1636, + "step": 3234 + }, + { + "epoch": 0.17, + "learning_rate": 3.8011615891607114e-07, + "loss": 0.1424, + "step": 3235 + }, + { + "epoch": 0.17, + "learning_rate": 3.8010145887999006e-07, + "loss": 0.1386, + "step": 3236 + }, + { + "epoch": 0.17, + "learning_rate": 3.8008675369653974e-07, + "loss": 0.107, + "step": 3237 + }, + { + "epoch": 0.17, + "learning_rate": 3.800720433661405e-07, + "loss": 0.1311, + "step": 3238 + }, + { + "epoch": 0.17, + "learning_rate": 3.800573278892128e-07, + "loss": 0.1733, + "step": 3239 + }, + { + "epoch": 0.17, + "learning_rate": 3.800426072661771e-07, + "loss": 0.1899, + "step": 3240 + }, + { + "epoch": 0.17, + "learning_rate": 3.800278814974542e-07, + "loss": 0.1747, + "step": 3241 + }, + { + "epoch": 0.17, + "learning_rate": 3.800131505834649e-07, + "loss": 0.1363, + "step": 3242 + }, + { + "epoch": 0.17, + "learning_rate": 3.799984145246303e-07, + "loss": 0.1498, + "step": 3243 + }, + { + "epoch": 0.17, + "learning_rate": 3.7998367332137153e-07, + "loss": 0.1883, + "step": 3244 + }, + { + "epoch": 0.17, + "learning_rate": 3.799689269741099e-07, + "loss": 0.1643, + "step": 3245 + }, + { + "epoch": 0.17, + "learning_rate": 3.799541754832669e-07, + "loss": 0.1819, + "step": 3246 + }, + { + "epoch": 0.17, + "learning_rate": 3.7993941884926404e-07, + "loss": 0.1299, + "step": 3247 + }, + { + "epoch": 0.17, + "learning_rate": 3.7992465707252316e-07, + "loss": 0.1703, + "step": 3248 + }, + { + "epoch": 0.17, + "learning_rate": 3.799098901534661e-07, + "loss": 0.1389, + "step": 3249 + }, + { + "epoch": 0.17, + "learning_rate": 3.7989511809251493e-07, + "loss": 0.1544, + "step": 3250 + }, + { + "epoch": 0.17, + "learning_rate": 3.7988034089009176e-07, + "loss": 0.1673, + "step": 3251 + }, + { + "epoch": 0.17, + "learning_rate": 3.798655585466191e-07, + "loss": 0.1963, + "step": 3252 + }, + { + "epoch": 0.17, + "learning_rate": 3.7985077106251934e-07, + "loss": 0.1601, + "step": 3253 + }, + { + "epoch": 0.17, + "learning_rate": 3.7983597843821516e-07, + "loss": 0.1646, + "step": 3254 + }, + { + "epoch": 0.17, + "learning_rate": 3.798211806741292e-07, + "loss": 0.1677, + "step": 3255 + }, + { + "epoch": 0.17, + "learning_rate": 3.798063777706845e-07, + "loss": 0.157, + "step": 3256 + }, + { + "epoch": 0.17, + "learning_rate": 3.7979156972830414e-07, + "loss": 0.122, + "step": 3257 + }, + { + "epoch": 0.17, + "learning_rate": 3.797767565474113e-07, + "loss": 0.1436, + "step": 3258 + }, + { + "epoch": 0.17, + "learning_rate": 3.797619382284293e-07, + "loss": 0.1458, + "step": 3259 + }, + { + "epoch": 0.17, + "learning_rate": 3.797471147717818e-07, + "loss": 0.1486, + "step": 3260 + }, + { + "epoch": 0.17, + "learning_rate": 3.797322861778923e-07, + "loss": 0.1363, + "step": 3261 + }, + { + "epoch": 0.17, + "learning_rate": 3.7971745244718473e-07, + "loss": 0.1484, + "step": 3262 + }, + { + "epoch": 0.17, + "learning_rate": 3.79702613580083e-07, + "loss": 0.1376, + "step": 3263 + }, + { + "epoch": 0.17, + "learning_rate": 3.7968776957701117e-07, + "loss": 0.171, + "step": 3264 + }, + { + "epoch": 0.17, + "learning_rate": 3.796729204383935e-07, + "loss": 0.1372, + "step": 3265 + }, + { + "epoch": 0.17, + "learning_rate": 3.796580661646544e-07, + "loss": 0.186, + "step": 3266 + }, + { + "epoch": 0.17, + "learning_rate": 3.7964320675621843e-07, + "loss": 0.1841, + "step": 3267 + }, + { + "epoch": 0.17, + "learning_rate": 3.796283422135103e-07, + "loss": 0.1429, + "step": 3268 + }, + { + "epoch": 0.17, + "learning_rate": 3.7961347253695477e-07, + "loss": 0.1729, + "step": 3269 + }, + { + "epoch": 0.17, + "learning_rate": 3.795985977269769e-07, + "loss": 0.1645, + "step": 3270 + }, + { + "epoch": 0.17, + "learning_rate": 3.795837177840017e-07, + "loss": 0.1653, + "step": 3271 + }, + { + "epoch": 0.17, + "learning_rate": 3.7956883270845455e-07, + "loss": 0.1435, + "step": 3272 + }, + { + "epoch": 0.17, + "learning_rate": 3.7955394250076084e-07, + "loss": 0.1631, + "step": 3273 + }, + { + "epoch": 0.17, + "learning_rate": 3.795390471613461e-07, + "loss": 0.1452, + "step": 3274 + }, + { + "epoch": 0.17, + "learning_rate": 3.795241466906362e-07, + "loss": 0.135, + "step": 3275 + }, + { + "epoch": 0.17, + "learning_rate": 3.795092410890568e-07, + "loss": 0.1754, + "step": 3276 + }, + { + "epoch": 0.17, + "learning_rate": 3.79494330357034e-07, + "loss": 0.1362, + "step": 3277 + }, + { + "epoch": 0.17, + "learning_rate": 3.7947941449499393e-07, + "loss": 0.1482, + "step": 3278 + }, + { + "epoch": 0.17, + "learning_rate": 3.7946449350336296e-07, + "loss": 0.127, + "step": 3279 + }, + { + "epoch": 0.17, + "learning_rate": 3.7944956738256745e-07, + "loss": 0.1765, + "step": 3280 + }, + { + "epoch": 0.17, + "learning_rate": 3.794346361330341e-07, + "loss": 0.1469, + "step": 3281 + }, + { + "epoch": 0.17, + "learning_rate": 3.7941969975518944e-07, + "loss": 0.1092, + "step": 3282 + }, + { + "epoch": 0.17, + "learning_rate": 3.794047582494606e-07, + "loss": 0.1603, + "step": 3283 + }, + { + "epoch": 0.17, + "learning_rate": 3.7938981161627445e-07, + "loss": 0.156, + "step": 3284 + }, + { + "epoch": 0.17, + "learning_rate": 3.793748598560583e-07, + "loss": 0.1622, + "step": 3285 + }, + { + "epoch": 0.17, + "learning_rate": 3.7935990296923936e-07, + "loss": 0.1307, + "step": 3286 + }, + { + "epoch": 0.17, + "learning_rate": 3.7934494095624523e-07, + "loss": 0.1823, + "step": 3287 + }, + { + "epoch": 0.17, + "learning_rate": 3.7932997381750337e-07, + "loss": 0.1555, + "step": 3288 + }, + { + "epoch": 0.17, + "learning_rate": 3.793150015534416e-07, + "loss": 0.1566, + "step": 3289 + }, + { + "epoch": 0.17, + "learning_rate": 3.793000241644879e-07, + "loss": 0.1342, + "step": 3290 + }, + { + "epoch": 0.17, + "learning_rate": 3.7928504165107035e-07, + "loss": 0.2074, + "step": 3291 + }, + { + "epoch": 0.17, + "learning_rate": 3.792700540136171e-07, + "loss": 0.1577, + "step": 3292 + }, + { + "epoch": 0.17, + "learning_rate": 3.792550612525564e-07, + "loss": 0.1152, + "step": 3293 + }, + { + "epoch": 0.17, + "learning_rate": 3.7924006336831694e-07, + "loss": 0.1632, + "step": 3294 + }, + { + "epoch": 0.17, + "learning_rate": 3.792250603613272e-07, + "loss": 0.145, + "step": 3295 + }, + { + "epoch": 0.17, + "learning_rate": 3.7921005223201614e-07, + "loss": 0.1402, + "step": 3296 + }, + { + "epoch": 0.17, + "learning_rate": 3.7919503898081254e-07, + "loss": 0.139, + "step": 3297 + }, + { + "epoch": 0.17, + "learning_rate": 3.791800206081456e-07, + "loss": 0.1063, + "step": 3298 + }, + { + "epoch": 0.17, + "learning_rate": 3.791649971144444e-07, + "loss": 0.1416, + "step": 3299 + }, + { + "epoch": 0.17, + "learning_rate": 3.7914996850013854e-07, + "loss": 0.0973, + "step": 3300 + }, + { + "epoch": 0.17, + "learning_rate": 3.7913493476565737e-07, + "loss": 0.1642, + "step": 3301 + }, + { + "epoch": 0.17, + "learning_rate": 3.7911989591143055e-07, + "loss": 0.1437, + "step": 3302 + }, + { + "epoch": 0.17, + "learning_rate": 3.7910485193788805e-07, + "loss": 0.1275, + "step": 3303 + }, + { + "epoch": 0.17, + "learning_rate": 3.790898028454597e-07, + "loss": 0.1288, + "step": 3304 + }, + { + "epoch": 0.17, + "learning_rate": 3.790747486345757e-07, + "loss": 0.1631, + "step": 3305 + }, + { + "epoch": 0.17, + "learning_rate": 3.7905968930566623e-07, + "loss": 0.1561, + "step": 3306 + }, + { + "epoch": 0.17, + "learning_rate": 3.7904462485916164e-07, + "loss": 0.1797, + "step": 3307 + }, + { + "epoch": 0.17, + "learning_rate": 3.7902955529549263e-07, + "loss": 0.1234, + "step": 3308 + }, + { + "epoch": 0.17, + "learning_rate": 3.7901448061508976e-07, + "loss": 0.1379, + "step": 3309 + }, + { + "epoch": 0.17, + "learning_rate": 3.7899940081838395e-07, + "loss": 0.1557, + "step": 3310 + }, + { + "epoch": 0.17, + "learning_rate": 3.789843159058062e-07, + "loss": 0.1179, + "step": 3311 + }, + { + "epoch": 0.17, + "learning_rate": 3.789692258777876e-07, + "loss": 0.151, + "step": 3312 + }, + { + "epoch": 0.17, + "learning_rate": 3.789541307347594e-07, + "loss": 0.1389, + "step": 3313 + }, + { + "epoch": 0.17, + "learning_rate": 3.7893903047715304e-07, + "loss": 0.1323, + "step": 3314 + }, + { + "epoch": 0.17, + "learning_rate": 3.789239251054002e-07, + "loss": 0.1828, + "step": 3315 + }, + { + "epoch": 0.17, + "learning_rate": 3.7890881461993244e-07, + "loss": 0.1238, + "step": 3316 + }, + { + "epoch": 0.17, + "learning_rate": 3.7889369902118175e-07, + "loss": 0.1392, + "step": 3317 + }, + { + "epoch": 0.17, + "learning_rate": 3.7887857830958005e-07, + "loss": 0.1383, + "step": 3318 + }, + { + "epoch": 0.17, + "learning_rate": 3.788634524855595e-07, + "loss": 0.1548, + "step": 3319 + }, + { + "epoch": 0.17, + "learning_rate": 3.7884832154955247e-07, + "loss": 0.1391, + "step": 3320 + }, + { + "epoch": 0.17, + "learning_rate": 3.7883318550199135e-07, + "loss": 0.1594, + "step": 3321 + }, + { + "epoch": 0.17, + "learning_rate": 3.788180443433087e-07, + "loss": 0.1246, + "step": 3322 + }, + { + "epoch": 0.17, + "learning_rate": 3.7880289807393735e-07, + "loss": 0.2048, + "step": 3323 + }, + { + "epoch": 0.17, + "learning_rate": 3.7878774669431015e-07, + "loss": 0.1628, + "step": 3324 + }, + { + "epoch": 0.17, + "learning_rate": 3.787725902048602e-07, + "loss": 0.1551, + "step": 3325 + }, + { + "epoch": 0.17, + "learning_rate": 3.7875742860602055e-07, + "loss": 0.1395, + "step": 3326 + }, + { + "epoch": 0.17, + "learning_rate": 3.787422618982246e-07, + "loss": 0.1301, + "step": 3327 + }, + { + "epoch": 0.17, + "learning_rate": 3.787270900819058e-07, + "loss": 0.1523, + "step": 3328 + }, + { + "epoch": 0.17, + "learning_rate": 3.787119131574977e-07, + "loss": 0.1474, + "step": 3329 + }, + { + "epoch": 0.17, + "learning_rate": 3.7869673112543417e-07, + "loss": 0.1315, + "step": 3330 + }, + { + "epoch": 0.17, + "learning_rate": 3.7868154398614915e-07, + "loss": 0.1325, + "step": 3331 + }, + { + "epoch": 0.17, + "learning_rate": 3.7866635174007655e-07, + "loss": 0.1137, + "step": 3332 + }, + { + "epoch": 0.17, + "learning_rate": 3.7865115438765074e-07, + "loss": 0.1229, + "step": 3333 + }, + { + "epoch": 0.17, + "learning_rate": 3.786359519293058e-07, + "loss": 0.1132, + "step": 3334 + }, + { + "epoch": 0.17, + "learning_rate": 3.7862074436547654e-07, + "loss": 0.1708, + "step": 3335 + }, + { + "epoch": 0.17, + "learning_rate": 3.786055316965974e-07, + "loss": 0.1707, + "step": 3336 + }, + { + "epoch": 0.17, + "learning_rate": 3.785903139231032e-07, + "loss": 0.1594, + "step": 3337 + }, + { + "epoch": 0.17, + "learning_rate": 3.7857509104542894e-07, + "loss": 0.1413, + "step": 3338 + }, + { + "epoch": 0.17, + "learning_rate": 3.785598630640096e-07, + "loss": 0.1456, + "step": 3339 + }, + { + "epoch": 0.17, + "learning_rate": 3.785446299792805e-07, + "loss": 0.1414, + "step": 3340 + }, + { + "epoch": 0.17, + "learning_rate": 3.7852939179167686e-07, + "loss": 0.1383, + "step": 3341 + }, + { + "epoch": 0.17, + "learning_rate": 3.7851414850163433e-07, + "loss": 0.1396, + "step": 3342 + }, + { + "epoch": 0.17, + "learning_rate": 3.784989001095885e-07, + "loss": 0.1463, + "step": 3343 + }, + { + "epoch": 0.17, + "learning_rate": 3.7848364661597525e-07, + "loss": 0.15, + "step": 3344 + }, + { + "epoch": 0.17, + "learning_rate": 3.784683880212304e-07, + "loss": 0.1854, + "step": 3345 + }, + { + "epoch": 0.17, + "learning_rate": 3.784531243257902e-07, + "loss": 0.1051, + "step": 3346 + }, + { + "epoch": 0.17, + "learning_rate": 3.7843785553009074e-07, + "loss": 0.1548, + "step": 3347 + }, + { + "epoch": 0.17, + "learning_rate": 3.784225816345685e-07, + "loss": 0.1314, + "step": 3348 + }, + { + "epoch": 0.17, + "learning_rate": 3.7840730263966003e-07, + "loss": 0.1553, + "step": 3349 + }, + { + "epoch": 0.17, + "learning_rate": 3.7839201854580204e-07, + "loss": 0.1351, + "step": 3350 + }, + { + "epoch": 0.17, + "learning_rate": 3.7837672935343124e-07, + "loss": 0.1336, + "step": 3351 + }, + { + "epoch": 0.17, + "learning_rate": 3.783614350629846e-07, + "loss": 0.146, + "step": 3352 + }, + { + "epoch": 0.18, + "learning_rate": 3.783461356748994e-07, + "loss": 0.1536, + "step": 3353 + }, + { + "epoch": 0.18, + "learning_rate": 3.7833083118961264e-07, + "loss": 0.1608, + "step": 3354 + }, + { + "epoch": 0.18, + "learning_rate": 3.7831552160756204e-07, + "loss": 0.1283, + "step": 3355 + }, + { + "epoch": 0.18, + "learning_rate": 3.783002069291849e-07, + "loss": 0.1413, + "step": 3356 + }, + { + "epoch": 0.18, + "learning_rate": 3.7828488715491906e-07, + "loss": 0.147, + "step": 3357 + }, + { + "epoch": 0.18, + "learning_rate": 3.7826956228520225e-07, + "loss": 0.1603, + "step": 3358 + }, + { + "epoch": 0.18, + "learning_rate": 3.7825423232047263e-07, + "loss": 0.1042, + "step": 3359 + }, + { + "epoch": 0.18, + "learning_rate": 3.7823889726116816e-07, + "loss": 0.1717, + "step": 3360 + }, + { + "epoch": 0.18, + "learning_rate": 3.7822355710772723e-07, + "loss": 0.1643, + "step": 3361 + }, + { + "epoch": 0.18, + "learning_rate": 3.7820821186058825e-07, + "loss": 0.1132, + "step": 3362 + }, + { + "epoch": 0.18, + "learning_rate": 3.781928615201898e-07, + "loss": 0.1376, + "step": 3363 + }, + { + "epoch": 0.18, + "learning_rate": 3.781775060869705e-07, + "loss": 0.1528, + "step": 3364 + }, + { + "epoch": 0.18, + "learning_rate": 3.781621455613694e-07, + "loss": 0.1414, + "step": 3365 + }, + { + "epoch": 0.18, + "learning_rate": 3.781467799438253e-07, + "loss": 0.1395, + "step": 3366 + }, + { + "epoch": 0.18, + "learning_rate": 3.7813140923477746e-07, + "loss": 0.1554, + "step": 3367 + }, + { + "epoch": 0.18, + "learning_rate": 3.7811603343466524e-07, + "loss": 0.1742, + "step": 3368 + }, + { + "epoch": 0.18, + "learning_rate": 3.78100652543928e-07, + "loss": 0.156, + "step": 3369 + }, + { + "epoch": 0.18, + "learning_rate": 3.7808526656300533e-07, + "loss": 0.166, + "step": 3370 + }, + { + "epoch": 0.18, + "learning_rate": 3.78069875492337e-07, + "loss": 0.1467, + "step": 3371 + }, + { + "epoch": 0.18, + "learning_rate": 3.7805447933236294e-07, + "loss": 0.1354, + "step": 3372 + }, + { + "epoch": 0.18, + "learning_rate": 3.780390780835231e-07, + "loss": 0.1813, + "step": 3373 + }, + { + "epoch": 0.18, + "learning_rate": 3.7802367174625765e-07, + "loss": 0.1329, + "step": 3374 + }, + { + "epoch": 0.18, + "learning_rate": 3.78008260321007e-07, + "loss": 0.124, + "step": 3375 + }, + { + "epoch": 0.18, + "learning_rate": 3.779928438082115e-07, + "loss": 0.1338, + "step": 3376 + }, + { + "epoch": 0.18, + "learning_rate": 3.7797742220831184e-07, + "loss": 0.1138, + "step": 3377 + }, + { + "epoch": 0.18, + "learning_rate": 3.779619955217487e-07, + "loss": 0.1802, + "step": 3378 + }, + { + "epoch": 0.18, + "learning_rate": 3.779465637489631e-07, + "loss": 0.1305, + "step": 3379 + }, + { + "epoch": 0.18, + "learning_rate": 3.779311268903959e-07, + "loss": 0.1452, + "step": 3380 + }, + { + "epoch": 0.18, + "learning_rate": 3.7791568494648854e-07, + "loss": 0.1867, + "step": 3381 + }, + { + "epoch": 0.18, + "learning_rate": 3.7790023791768214e-07, + "loss": 0.1243, + "step": 3382 + }, + { + "epoch": 0.18, + "learning_rate": 3.778847858044183e-07, + "loss": 0.152, + "step": 3383 + }, + { + "epoch": 0.18, + "learning_rate": 3.778693286071386e-07, + "loss": 0.1322, + "step": 3384 + }, + { + "epoch": 0.18, + "learning_rate": 3.778538663262848e-07, + "loss": 0.1603, + "step": 3385 + }, + { + "epoch": 0.18, + "learning_rate": 3.7783839896229884e-07, + "loss": 0.1571, + "step": 3386 + }, + { + "epoch": 0.18, + "learning_rate": 3.778229265156228e-07, + "loss": 0.1497, + "step": 3387 + }, + { + "epoch": 0.18, + "learning_rate": 3.7780744898669886e-07, + "loss": 0.1559, + "step": 3388 + }, + { + "epoch": 0.18, + "learning_rate": 3.7779196637596937e-07, + "loss": 0.1523, + "step": 3389 + }, + { + "epoch": 0.18, + "learning_rate": 3.7777647868387695e-07, + "loss": 0.1505, + "step": 3390 + }, + { + "epoch": 0.18, + "learning_rate": 3.7776098591086407e-07, + "loss": 0.1371, + "step": 3391 + }, + { + "epoch": 0.18, + "learning_rate": 3.7774548805737356e-07, + "loss": 0.1607, + "step": 3392 + }, + { + "epoch": 0.18, + "learning_rate": 3.777299851238484e-07, + "loss": 0.1449, + "step": 3393 + }, + { + "epoch": 0.18, + "learning_rate": 3.777144771107317e-07, + "loss": 0.1603, + "step": 3394 + }, + { + "epoch": 0.18, + "learning_rate": 3.7769896401846664e-07, + "loss": 0.129, + "step": 3395 + }, + { + "epoch": 0.18, + "learning_rate": 3.776834458474965e-07, + "loss": 0.1603, + "step": 3396 + }, + { + "epoch": 0.18, + "learning_rate": 3.77667922598265e-07, + "loss": 0.1692, + "step": 3397 + }, + { + "epoch": 0.18, + "learning_rate": 3.776523942712156e-07, + "loss": 0.1572, + "step": 3398 + }, + { + "epoch": 0.18, + "learning_rate": 3.7763686086679224e-07, + "loss": 0.1345, + "step": 3399 + }, + { + "epoch": 0.18, + "learning_rate": 3.776213223854388e-07, + "loss": 0.1588, + "step": 3400 + }, + { + "epoch": 0.18, + "learning_rate": 3.776057788275994e-07, + "loss": 0.1858, + "step": 3401 + }, + { + "epoch": 0.18, + "learning_rate": 3.775902301937183e-07, + "loss": 0.1449, + "step": 3402 + }, + { + "epoch": 0.18, + "learning_rate": 3.7757467648423977e-07, + "loss": 0.1533, + "step": 3403 + }, + { + "epoch": 0.18, + "learning_rate": 3.7755911769960853e-07, + "loss": 0.1087, + "step": 3404 + }, + { + "epoch": 0.18, + "learning_rate": 3.775435538402692e-07, + "loss": 0.1654, + "step": 3405 + }, + { + "epoch": 0.18, + "learning_rate": 3.7752798490666646e-07, + "loss": 0.1748, + "step": 3406 + }, + { + "epoch": 0.18, + "learning_rate": 3.775124108992454e-07, + "loss": 0.1748, + "step": 3407 + }, + { + "epoch": 0.18, + "learning_rate": 3.774968318184512e-07, + "loss": 0.1413, + "step": 3408 + }, + { + "epoch": 0.18, + "learning_rate": 3.7748124766472894e-07, + "loss": 0.1266, + "step": 3409 + }, + { + "epoch": 0.18, + "learning_rate": 3.774656584385241e-07, + "loss": 0.1148, + "step": 3410 + }, + { + "epoch": 0.18, + "learning_rate": 3.7745006414028226e-07, + "loss": 0.1591, + "step": 3411 + }, + { + "epoch": 0.18, + "learning_rate": 3.774344647704491e-07, + "loss": 0.1427, + "step": 3412 + }, + { + "epoch": 0.18, + "learning_rate": 3.774188603294704e-07, + "loss": 0.1605, + "step": 3413 + }, + { + "epoch": 0.18, + "learning_rate": 3.7740325081779217e-07, + "loss": 0.1597, + "step": 3414 + }, + { + "epoch": 0.18, + "learning_rate": 3.7738763623586063e-07, + "loss": 0.2008, + "step": 3415 + }, + { + "epoch": 0.18, + "learning_rate": 3.7737201658412187e-07, + "loss": 0.165, + "step": 3416 + }, + { + "epoch": 0.18, + "learning_rate": 3.7735639186302246e-07, + "loss": 0.1217, + "step": 3417 + }, + { + "epoch": 0.18, + "learning_rate": 3.773407620730089e-07, + "loss": 0.1361, + "step": 3418 + }, + { + "epoch": 0.18, + "learning_rate": 3.7732512721452794e-07, + "loss": 0.1683, + "step": 3419 + }, + { + "epoch": 0.18, + "learning_rate": 3.7730948728802633e-07, + "loss": 0.1593, + "step": 3420 + }, + { + "epoch": 0.18, + "learning_rate": 3.772938422939511e-07, + "loss": 0.1283, + "step": 3421 + }, + { + "epoch": 0.18, + "learning_rate": 3.7727819223274947e-07, + "loss": 0.1763, + "step": 3422 + }, + { + "epoch": 0.18, + "learning_rate": 3.7726253710486863e-07, + "loss": 0.1449, + "step": 3423 + }, + { + "epoch": 0.18, + "learning_rate": 3.7724687691075615e-07, + "loss": 0.1407, + "step": 3424 + }, + { + "epoch": 0.18, + "learning_rate": 3.772312116508594e-07, + "loss": 0.1792, + "step": 3425 + }, + { + "epoch": 0.18, + "learning_rate": 3.772155413256262e-07, + "loss": 0.165, + "step": 3426 + }, + { + "epoch": 0.18, + "learning_rate": 3.771998659355045e-07, + "loss": 0.1405, + "step": 3427 + }, + { + "epoch": 0.18, + "learning_rate": 3.7718418548094216e-07, + "loss": 0.1527, + "step": 3428 + }, + { + "epoch": 0.18, + "learning_rate": 3.771684999623874e-07, + "loss": 0.1417, + "step": 3429 + }, + { + "epoch": 0.18, + "learning_rate": 3.771528093802885e-07, + "loss": 0.1483, + "step": 3430 + }, + { + "epoch": 0.18, + "learning_rate": 3.7713711373509395e-07, + "loss": 0.1671, + "step": 3431 + }, + { + "epoch": 0.18, + "learning_rate": 3.7712141302725235e-07, + "loss": 0.1545, + "step": 3432 + }, + { + "epoch": 0.18, + "learning_rate": 3.771057072572123e-07, + "loss": 0.1319, + "step": 3433 + }, + { + "epoch": 0.18, + "learning_rate": 3.770899964254228e-07, + "loss": 0.1932, + "step": 3434 + }, + { + "epoch": 0.18, + "learning_rate": 3.770742805323329e-07, + "loss": 0.1652, + "step": 3435 + }, + { + "epoch": 0.18, + "learning_rate": 3.7705855957839166e-07, + "loss": 0.1315, + "step": 3436 + }, + { + "epoch": 0.18, + "learning_rate": 3.770428335640484e-07, + "loss": 0.1651, + "step": 3437 + }, + { + "epoch": 0.18, + "learning_rate": 3.7702710248975265e-07, + "loss": 0.1669, + "step": 3438 + }, + { + "epoch": 0.18, + "learning_rate": 3.77011366355954e-07, + "loss": 0.1965, + "step": 3439 + }, + { + "epoch": 0.18, + "learning_rate": 3.7699562516310214e-07, + "loss": 0.1468, + "step": 3440 + }, + { + "epoch": 0.18, + "learning_rate": 3.7697987891164696e-07, + "loss": 0.1372, + "step": 3441 + }, + { + "epoch": 0.18, + "learning_rate": 3.769641276020385e-07, + "loss": 0.1511, + "step": 3442 + }, + { + "epoch": 0.18, + "learning_rate": 3.76948371234727e-07, + "loss": 0.1447, + "step": 3443 + }, + { + "epoch": 0.18, + "learning_rate": 3.7693260981016273e-07, + "loss": 0.1508, + "step": 3444 + }, + { + "epoch": 0.18, + "learning_rate": 3.769168433287962e-07, + "loss": 0.2029, + "step": 3445 + }, + { + "epoch": 0.18, + "learning_rate": 3.769010717910779e-07, + "loss": 0.1464, + "step": 3446 + }, + { + "epoch": 0.18, + "learning_rate": 3.768852951974588e-07, + "loss": 0.1229, + "step": 3447 + }, + { + "epoch": 0.18, + "learning_rate": 3.768695135483896e-07, + "loss": 0.1556, + "step": 3448 + }, + { + "epoch": 0.18, + "learning_rate": 3.768537268443215e-07, + "loss": 0.1582, + "step": 3449 + }, + { + "epoch": 0.18, + "learning_rate": 3.7683793508570555e-07, + "loss": 0.1935, + "step": 3450 + }, + { + "epoch": 0.18, + "learning_rate": 3.7682213827299314e-07, + "loss": 0.1417, + "step": 3451 + }, + { + "epoch": 0.18, + "learning_rate": 3.768063364066358e-07, + "loss": 0.1388, + "step": 3452 + }, + { + "epoch": 0.18, + "learning_rate": 3.767905294870851e-07, + "loss": 0.1703, + "step": 3453 + }, + { + "epoch": 0.18, + "learning_rate": 3.767747175147928e-07, + "loss": 0.144, + "step": 3454 + }, + { + "epoch": 0.18, + "learning_rate": 3.767589004902108e-07, + "loss": 0.1302, + "step": 3455 + }, + { + "epoch": 0.18, + "learning_rate": 3.767430784137913e-07, + "loss": 0.1411, + "step": 3456 + }, + { + "epoch": 0.18, + "learning_rate": 3.767272512859863e-07, + "loss": 0.1635, + "step": 3457 + }, + { + "epoch": 0.18, + "learning_rate": 3.7671141910724824e-07, + "loss": 0.1498, + "step": 3458 + }, + { + "epoch": 0.18, + "learning_rate": 3.7669558187802963e-07, + "loss": 0.153, + "step": 3459 + }, + { + "epoch": 0.18, + "learning_rate": 3.76679739598783e-07, + "loss": 0.1444, + "step": 3460 + }, + { + "epoch": 0.18, + "learning_rate": 3.766638922699613e-07, + "loss": 0.194, + "step": 3461 + }, + { + "epoch": 0.18, + "learning_rate": 3.766480398920174e-07, + "loss": 0.1493, + "step": 3462 + }, + { + "epoch": 0.18, + "learning_rate": 3.7663218246540424e-07, + "loss": 0.1796, + "step": 3463 + }, + { + "epoch": 0.18, + "learning_rate": 3.766163199905751e-07, + "loss": 0.1225, + "step": 3464 + }, + { + "epoch": 0.18, + "learning_rate": 3.766004524679834e-07, + "loss": 0.1391, + "step": 3465 + }, + { + "epoch": 0.18, + "learning_rate": 3.765845798980826e-07, + "loss": 0.1344, + "step": 3466 + }, + { + "epoch": 0.18, + "learning_rate": 3.7656870228132626e-07, + "loss": 0.1393, + "step": 3467 + }, + { + "epoch": 0.18, + "learning_rate": 3.765528196181683e-07, + "loss": 0.1581, + "step": 3468 + }, + { + "epoch": 0.18, + "learning_rate": 3.7653693190906265e-07, + "loss": 0.1389, + "step": 3469 + }, + { + "epoch": 0.18, + "learning_rate": 3.765210391544633e-07, + "loss": 0.1398, + "step": 3470 + }, + { + "epoch": 0.18, + "learning_rate": 3.765051413548245e-07, + "loss": 0.1381, + "step": 3471 + }, + { + "epoch": 0.18, + "learning_rate": 3.764892385106006e-07, + "loss": 0.1371, + "step": 3472 + }, + { + "epoch": 0.18, + "learning_rate": 3.764733306222462e-07, + "loss": 0.1487, + "step": 3473 + }, + { + "epoch": 0.18, + "learning_rate": 3.7645741769021583e-07, + "loss": 0.1492, + "step": 3474 + }, + { + "epoch": 0.18, + "learning_rate": 3.764414997149643e-07, + "loss": 0.1484, + "step": 3475 + }, + { + "epoch": 0.18, + "learning_rate": 3.764255766969467e-07, + "loss": 0.168, + "step": 3476 + }, + { + "epoch": 0.18, + "learning_rate": 3.7640964863661796e-07, + "loss": 0.1719, + "step": 3477 + }, + { + "epoch": 0.18, + "learning_rate": 3.7639371553443335e-07, + "loss": 0.1387, + "step": 3478 + }, + { + "epoch": 0.18, + "learning_rate": 3.7637777739084823e-07, + "loss": 0.1586, + "step": 3479 + }, + { + "epoch": 0.18, + "learning_rate": 3.763618342063182e-07, + "loss": 0.1379, + "step": 3480 + }, + { + "epoch": 0.18, + "learning_rate": 3.7634588598129885e-07, + "loss": 0.1574, + "step": 3481 + }, + { + "epoch": 0.18, + "learning_rate": 3.76329932716246e-07, + "loss": 0.1423, + "step": 3482 + }, + { + "epoch": 0.18, + "learning_rate": 3.763139744116156e-07, + "loss": 0.1453, + "step": 3483 + }, + { + "epoch": 0.18, + "learning_rate": 3.7629801106786374e-07, + "loss": 0.2019, + "step": 3484 + }, + { + "epoch": 0.18, + "learning_rate": 3.7628204268544667e-07, + "loss": 0.1508, + "step": 3485 + }, + { + "epoch": 0.18, + "learning_rate": 3.762660692648208e-07, + "loss": 0.132, + "step": 3486 + }, + { + "epoch": 0.18, + "learning_rate": 3.762500908064426e-07, + "loss": 0.1897, + "step": 3487 + }, + { + "epoch": 0.18, + "learning_rate": 3.7623410731076876e-07, + "loss": 0.1573, + "step": 3488 + }, + { + "epoch": 0.18, + "learning_rate": 3.762181187782561e-07, + "loss": 0.1114, + "step": 3489 + }, + { + "epoch": 0.18, + "learning_rate": 3.762021252093616e-07, + "loss": 0.1343, + "step": 3490 + }, + { + "epoch": 0.18, + "learning_rate": 3.7618612660454225e-07, + "loss": 0.149, + "step": 3491 + }, + { + "epoch": 0.18, + "learning_rate": 3.761701229642555e-07, + "loss": 0.1446, + "step": 3492 + }, + { + "epoch": 0.18, + "learning_rate": 3.7615411428895855e-07, + "loss": 0.1643, + "step": 3493 + }, + { + "epoch": 0.18, + "learning_rate": 3.761381005791091e-07, + "loss": 0.1682, + "step": 3494 + }, + { + "epoch": 0.18, + "learning_rate": 3.7612208183516463e-07, + "loss": 0.1405, + "step": 3495 + }, + { + "epoch": 0.18, + "learning_rate": 3.7610605805758314e-07, + "loss": 0.1339, + "step": 3496 + }, + { + "epoch": 0.18, + "learning_rate": 3.760900292468225e-07, + "loss": 0.1844, + "step": 3497 + }, + { + "epoch": 0.18, + "learning_rate": 3.760739954033409e-07, + "loss": 0.1211, + "step": 3498 + }, + { + "epoch": 0.18, + "learning_rate": 3.760579565275965e-07, + "loss": 0.1137, + "step": 3499 + }, + { + "epoch": 0.18, + "learning_rate": 3.7604191262004773e-07, + "loss": 0.1485, + "step": 3500 + }, + { + "epoch": 0.18, + "learning_rate": 3.760258636811532e-07, + "loss": 0.1605, + "step": 3501 + }, + { + "epoch": 0.18, + "learning_rate": 3.7600980971137145e-07, + "loss": 0.1609, + "step": 3502 + }, + { + "epoch": 0.18, + "learning_rate": 3.7599375071116136e-07, + "loss": 0.1388, + "step": 3503 + }, + { + "epoch": 0.18, + "learning_rate": 3.75977686680982e-07, + "loss": 0.1595, + "step": 3504 + }, + { + "epoch": 0.18, + "learning_rate": 3.7596161762129245e-07, + "loss": 0.1644, + "step": 3505 + }, + { + "epoch": 0.18, + "learning_rate": 3.759455435325519e-07, + "loss": 0.1169, + "step": 3506 + }, + { + "epoch": 0.18, + "learning_rate": 3.7592946441521975e-07, + "loss": 0.1829, + "step": 3507 + }, + { + "epoch": 0.18, + "learning_rate": 3.759133802697557e-07, + "loss": 0.1416, + "step": 3508 + }, + { + "epoch": 0.18, + "learning_rate": 3.758972910966193e-07, + "loss": 0.1333, + "step": 3509 + }, + { + "epoch": 0.18, + "learning_rate": 3.7588119689627034e-07, + "loss": 0.1433, + "step": 3510 + }, + { + "epoch": 0.18, + "learning_rate": 3.7586509766916894e-07, + "loss": 0.1255, + "step": 3511 + }, + { + "epoch": 0.18, + "learning_rate": 3.7584899341577514e-07, + "loss": 0.1288, + "step": 3512 + }, + { + "epoch": 0.18, + "learning_rate": 3.7583288413654926e-07, + "loss": 0.1519, + "step": 3513 + }, + { + "epoch": 0.18, + "learning_rate": 3.758167698319516e-07, + "loss": 0.172, + "step": 3514 + }, + { + "epoch": 0.18, + "learning_rate": 3.7580065050244284e-07, + "loss": 0.1344, + "step": 3515 + }, + { + "epoch": 0.18, + "learning_rate": 3.757845261484836e-07, + "loss": 0.1416, + "step": 3516 + }, + { + "epoch": 0.18, + "learning_rate": 3.757683967705348e-07, + "loss": 0.13, + "step": 3517 + }, + { + "epoch": 0.18, + "learning_rate": 3.7575226236905724e-07, + "loss": 0.1371, + "step": 3518 + }, + { + "epoch": 0.18, + "learning_rate": 3.7573612294451227e-07, + "loss": 0.1262, + "step": 3519 + }, + { + "epoch": 0.18, + "learning_rate": 3.757199784973611e-07, + "loss": 0.1197, + "step": 3520 + }, + { + "epoch": 0.18, + "learning_rate": 3.7570382902806504e-07, + "loss": 0.1295, + "step": 3521 + }, + { + "epoch": 0.18, + "learning_rate": 3.7568767453708573e-07, + "loss": 0.1674, + "step": 3522 + }, + { + "epoch": 0.18, + "learning_rate": 3.7567151502488484e-07, + "loss": 0.1018, + "step": 3523 + }, + { + "epoch": 0.18, + "learning_rate": 3.756553504919243e-07, + "loss": 0.1534, + "step": 3524 + }, + { + "epoch": 0.18, + "learning_rate": 3.7563918093866596e-07, + "loss": 0.1667, + "step": 3525 + }, + { + "epoch": 0.18, + "learning_rate": 3.756230063655721e-07, + "loss": 0.1498, + "step": 3526 + }, + { + "epoch": 0.18, + "learning_rate": 3.7560682677310485e-07, + "loss": 0.1542, + "step": 3527 + }, + { + "epoch": 0.18, + "learning_rate": 3.7559064216172674e-07, + "loss": 0.1559, + "step": 3528 + }, + { + "epoch": 0.18, + "learning_rate": 3.7557445253190034e-07, + "loss": 0.1553, + "step": 3529 + }, + { + "epoch": 0.18, + "learning_rate": 3.755582578840882e-07, + "loss": 0.1547, + "step": 3530 + }, + { + "epoch": 0.18, + "learning_rate": 3.7554205821875334e-07, + "loss": 0.1404, + "step": 3531 + }, + { + "epoch": 0.18, + "learning_rate": 3.7552585353635865e-07, + "loss": 0.1625, + "step": 3532 + }, + { + "epoch": 0.18, + "learning_rate": 3.7550964383736733e-07, + "loss": 0.1484, + "step": 3533 + }, + { + "epoch": 0.18, + "learning_rate": 3.7549342912224266e-07, + "loss": 0.1479, + "step": 3534 + }, + { + "epoch": 0.18, + "learning_rate": 3.7547720939144803e-07, + "loss": 0.156, + "step": 3535 + }, + { + "epoch": 0.18, + "learning_rate": 3.7546098464544697e-07, + "loss": 0.1224, + "step": 3536 + }, + { + "epoch": 0.18, + "learning_rate": 3.754447548847033e-07, + "loss": 0.1227, + "step": 3537 + }, + { + "epoch": 0.18, + "learning_rate": 3.7542852010968076e-07, + "loss": 0.1777, + "step": 3538 + }, + { + "epoch": 0.18, + "learning_rate": 3.7541228032084344e-07, + "loss": 0.1726, + "step": 3539 + }, + { + "epoch": 0.18, + "learning_rate": 3.753960355186554e-07, + "loss": 0.1462, + "step": 3540 + }, + { + "epoch": 0.18, + "learning_rate": 3.753797857035809e-07, + "loss": 0.1536, + "step": 3541 + }, + { + "epoch": 0.18, + "learning_rate": 3.753635308760845e-07, + "loss": 0.1403, + "step": 3542 + }, + { + "epoch": 0.18, + "learning_rate": 3.7534727103663066e-07, + "loss": 0.1634, + "step": 3543 + }, + { + "epoch": 0.18, + "learning_rate": 3.7533100618568414e-07, + "loss": 0.1168, + "step": 3544 + }, + { + "epoch": 0.19, + "learning_rate": 3.7531473632370975e-07, + "loss": 0.1422, + "step": 3545 + }, + { + "epoch": 0.19, + "learning_rate": 3.7529846145117253e-07, + "loss": 0.1875, + "step": 3546 + }, + { + "epoch": 0.19, + "learning_rate": 3.7528218156853765e-07, + "loss": 0.1281, + "step": 3547 + }, + { + "epoch": 0.19, + "learning_rate": 3.752658966762703e-07, + "loss": 0.1503, + "step": 3548 + }, + { + "epoch": 0.19, + "learning_rate": 3.7524960677483596e-07, + "loss": 0.152, + "step": 3549 + }, + { + "epoch": 0.19, + "learning_rate": 3.7523331186470025e-07, + "loss": 0.1335, + "step": 3550 + }, + { + "epoch": 0.19, + "learning_rate": 3.752170119463288e-07, + "loss": 0.1191, + "step": 3551 + }, + { + "epoch": 0.19, + "learning_rate": 3.752007070201875e-07, + "loss": 0.1619, + "step": 3552 + }, + { + "epoch": 0.19, + "learning_rate": 3.7518439708674244e-07, + "loss": 0.1942, + "step": 3553 + }, + { + "epoch": 0.19, + "learning_rate": 3.751680821464596e-07, + "loss": 0.1435, + "step": 3554 + }, + { + "epoch": 0.19, + "learning_rate": 3.751517621998054e-07, + "loss": 0.1287, + "step": 3555 + }, + { + "epoch": 0.19, + "learning_rate": 3.751354372472462e-07, + "loss": 0.1444, + "step": 3556 + }, + { + "epoch": 0.19, + "learning_rate": 3.751191072892486e-07, + "loss": 0.1502, + "step": 3557 + }, + { + "epoch": 0.19, + "learning_rate": 3.7510277232627933e-07, + "loss": 0.1219, + "step": 3558 + }, + { + "epoch": 0.19, + "learning_rate": 3.7508643235880515e-07, + "loss": 0.1406, + "step": 3559 + }, + { + "epoch": 0.19, + "learning_rate": 3.750700873872932e-07, + "loss": 0.1336, + "step": 3560 + }, + { + "epoch": 0.19, + "learning_rate": 3.750537374122106e-07, + "loss": 0.1273, + "step": 3561 + }, + { + "epoch": 0.19, + "learning_rate": 3.750373824340246e-07, + "loss": 0.1655, + "step": 3562 + }, + { + "epoch": 0.19, + "learning_rate": 3.750210224532026e-07, + "loss": 0.1656, + "step": 3563 + }, + { + "epoch": 0.19, + "learning_rate": 3.7500465747021225e-07, + "loss": 0.1615, + "step": 3564 + }, + { + "epoch": 0.19, + "learning_rate": 3.7498828748552117e-07, + "loss": 0.1099, + "step": 3565 + }, + { + "epoch": 0.19, + "learning_rate": 3.749719124995973e-07, + "loss": 0.1507, + "step": 3566 + }, + { + "epoch": 0.19, + "learning_rate": 3.749555325129087e-07, + "loss": 0.1298, + "step": 3567 + }, + { + "epoch": 0.19, + "learning_rate": 3.7493914752592336e-07, + "loss": 0.1772, + "step": 3568 + }, + { + "epoch": 0.19, + "learning_rate": 3.7492275753910965e-07, + "loss": 0.157, + "step": 3569 + }, + { + "epoch": 0.19, + "learning_rate": 3.7490636255293607e-07, + "loss": 0.1822, + "step": 3570 + }, + { + "epoch": 0.19, + "learning_rate": 3.7488996256787107e-07, + "loss": 0.1441, + "step": 3571 + }, + { + "epoch": 0.19, + "learning_rate": 3.7487355758438344e-07, + "loss": 0.1582, + "step": 3572 + }, + { + "epoch": 0.19, + "learning_rate": 3.74857147602942e-07, + "loss": 0.1224, + "step": 3573 + }, + { + "epoch": 0.19, + "learning_rate": 3.7484073262401584e-07, + "loss": 0.134, + "step": 3574 + }, + { + "epoch": 0.19, + "learning_rate": 3.74824312648074e-07, + "loss": 0.1294, + "step": 3575 + }, + { + "epoch": 0.19, + "learning_rate": 3.748078876755858e-07, + "loss": 0.1293, + "step": 3576 + }, + { + "epoch": 0.19, + "learning_rate": 3.7479145770702076e-07, + "loss": 0.1569, + "step": 3577 + }, + { + "epoch": 0.19, + "learning_rate": 3.7477502274284834e-07, + "loss": 0.1262, + "step": 3578 + }, + { + "epoch": 0.19, + "learning_rate": 3.747585827835383e-07, + "loss": 0.1294, + "step": 3579 + }, + { + "epoch": 0.19, + "learning_rate": 3.747421378295605e-07, + "loss": 0.1593, + "step": 3580 + }, + { + "epoch": 0.19, + "learning_rate": 3.7472568788138494e-07, + "loss": 0.155, + "step": 3581 + }, + { + "epoch": 0.19, + "learning_rate": 3.747092329394818e-07, + "loss": 0.1437, + "step": 3582 + }, + { + "epoch": 0.19, + "learning_rate": 3.746927730043213e-07, + "loss": 0.1142, + "step": 3583 + }, + { + "epoch": 0.19, + "learning_rate": 3.746763080763739e-07, + "loss": 0.1451, + "step": 3584 + }, + { + "epoch": 0.19, + "learning_rate": 3.746598381561102e-07, + "loss": 0.1266, + "step": 3585 + }, + { + "epoch": 0.19, + "learning_rate": 3.746433632440009e-07, + "loss": 0.1391, + "step": 3586 + }, + { + "epoch": 0.19, + "learning_rate": 3.7462688334051686e-07, + "loss": 0.1395, + "step": 3587 + }, + { + "epoch": 0.19, + "learning_rate": 3.746103984461291e-07, + "loss": 0.1378, + "step": 3588 + }, + { + "epoch": 0.19, + "learning_rate": 3.745939085613087e-07, + "loss": 0.1552, + "step": 3589 + }, + { + "epoch": 0.19, + "learning_rate": 3.745774136865271e-07, + "loss": 0.126, + "step": 3590 + }, + { + "epoch": 0.19, + "learning_rate": 3.745609138222555e-07, + "loss": 0.1455, + "step": 3591 + }, + { + "epoch": 0.19, + "learning_rate": 3.7454440896896564e-07, + "loss": 0.1379, + "step": 3592 + }, + { + "epoch": 0.19, + "learning_rate": 3.745278991271292e-07, + "loss": 0.1463, + "step": 3593 + }, + { + "epoch": 0.19, + "learning_rate": 3.74511384297218e-07, + "loss": 0.1177, + "step": 3594 + }, + { + "epoch": 0.19, + "learning_rate": 3.744948644797041e-07, + "loss": 0.1547, + "step": 3595 + }, + { + "epoch": 0.19, + "learning_rate": 3.744783396750596e-07, + "loss": 0.1561, + "step": 3596 + }, + { + "epoch": 0.19, + "learning_rate": 3.7446180988375683e-07, + "loss": 0.1834, + "step": 3597 + }, + { + "epoch": 0.19, + "learning_rate": 3.744452751062681e-07, + "loss": 0.1494, + "step": 3598 + }, + { + "epoch": 0.19, + "learning_rate": 3.7442873534306615e-07, + "loss": 0.1389, + "step": 3599 + }, + { + "epoch": 0.19, + "learning_rate": 3.7441219059462357e-07, + "loss": 0.1394, + "step": 3600 + }, + { + "epoch": 0.19, + "learning_rate": 3.7439564086141327e-07, + "loss": 0.1519, + "step": 3601 + }, + { + "epoch": 0.19, + "learning_rate": 3.743790861439082e-07, + "loss": 0.2126, + "step": 3602 + }, + { + "epoch": 0.19, + "learning_rate": 3.743625264425816e-07, + "loss": 0.1647, + "step": 3603 + }, + { + "epoch": 0.19, + "learning_rate": 3.743459617579066e-07, + "loss": 0.158, + "step": 3604 + }, + { + "epoch": 0.19, + "learning_rate": 3.7432939209035677e-07, + "loss": 0.1298, + "step": 3605 + }, + { + "epoch": 0.19, + "learning_rate": 3.743128174404056e-07, + "loss": 0.1359, + "step": 3606 + }, + { + "epoch": 0.19, + "learning_rate": 3.742962378085268e-07, + "loss": 0.1692, + "step": 3607 + }, + { + "epoch": 0.19, + "learning_rate": 3.742796531951942e-07, + "loss": 0.1152, + "step": 3608 + }, + { + "epoch": 0.19, + "learning_rate": 3.742630636008819e-07, + "loss": 0.1518, + "step": 3609 + }, + { + "epoch": 0.19, + "learning_rate": 3.7424646902606395e-07, + "loss": 0.1732, + "step": 3610 + }, + { + "epoch": 0.19, + "learning_rate": 3.7422986947121463e-07, + "loss": 0.116, + "step": 3611 + }, + { + "epoch": 0.19, + "learning_rate": 3.7421326493680843e-07, + "loss": 0.1396, + "step": 3612 + }, + { + "epoch": 0.19, + "learning_rate": 3.741966554233198e-07, + "loss": 0.1078, + "step": 3613 + }, + { + "epoch": 0.19, + "learning_rate": 3.7418004093122354e-07, + "loss": 0.1342, + "step": 3614 + }, + { + "epoch": 0.19, + "learning_rate": 3.7416342146099454e-07, + "loss": 0.1314, + "step": 3615 + }, + { + "epoch": 0.19, + "learning_rate": 3.741467970131076e-07, + "loss": 0.1874, + "step": 3616 + }, + { + "epoch": 0.19, + "learning_rate": 3.74130167588038e-07, + "loss": 0.1397, + "step": 3617 + }, + { + "epoch": 0.19, + "learning_rate": 3.74113533186261e-07, + "loss": 0.1374, + "step": 3618 + }, + { + "epoch": 0.19, + "learning_rate": 3.74096893808252e-07, + "loss": 0.1297, + "step": 3619 + }, + { + "epoch": 0.19, + "learning_rate": 3.7408024945448667e-07, + "loss": 0.1448, + "step": 3620 + }, + { + "epoch": 0.19, + "learning_rate": 3.7406360012544054e-07, + "loss": 0.1489, + "step": 3621 + }, + { + "epoch": 0.19, + "learning_rate": 3.740469458215895e-07, + "loss": 0.1468, + "step": 3622 + }, + { + "epoch": 0.19, + "learning_rate": 3.7403028654340954e-07, + "loss": 0.1306, + "step": 3623 + }, + { + "epoch": 0.19, + "learning_rate": 3.7401362229137685e-07, + "loss": 0.1288, + "step": 3624 + }, + { + "epoch": 0.19, + "learning_rate": 3.739969530659676e-07, + "loss": 0.1433, + "step": 3625 + }, + { + "epoch": 0.19, + "learning_rate": 3.739802788676583e-07, + "loss": 0.1571, + "step": 3626 + }, + { + "epoch": 0.19, + "learning_rate": 3.739635996969255e-07, + "loss": 0.1549, + "step": 3627 + }, + { + "epoch": 0.19, + "learning_rate": 3.739469155542459e-07, + "loss": 0.138, + "step": 3628 + }, + { + "epoch": 0.19, + "learning_rate": 3.739302264400963e-07, + "loss": 0.1416, + "step": 3629 + }, + { + "epoch": 0.19, + "learning_rate": 3.7391353235495354e-07, + "loss": 0.1345, + "step": 3630 + }, + { + "epoch": 0.19, + "learning_rate": 3.7389683329929505e-07, + "loss": 0.1341, + "step": 3631 + }, + { + "epoch": 0.19, + "learning_rate": 3.738801292735978e-07, + "loss": 0.1557, + "step": 3632 + }, + { + "epoch": 0.19, + "learning_rate": 3.738634202783394e-07, + "loss": 0.1627, + "step": 3633 + }, + { + "epoch": 0.19, + "learning_rate": 3.7384670631399733e-07, + "loss": 0.1343, + "step": 3634 + }, + { + "epoch": 0.19, + "learning_rate": 3.7382998738104927e-07, + "loss": 0.1686, + "step": 3635 + }, + { + "epoch": 0.19, + "learning_rate": 3.738132634799731e-07, + "loss": 0.1373, + "step": 3636 + }, + { + "epoch": 0.19, + "learning_rate": 3.737965346112467e-07, + "loss": 0.1711, + "step": 3637 + }, + { + "epoch": 0.19, + "learning_rate": 3.7377980077534825e-07, + "loss": 0.1499, + "step": 3638 + }, + { + "epoch": 0.19, + "learning_rate": 3.737630619727561e-07, + "loss": 0.2119, + "step": 3639 + }, + { + "epoch": 0.19, + "learning_rate": 3.7374631820394844e-07, + "loss": 0.1678, + "step": 3640 + }, + { + "epoch": 0.19, + "learning_rate": 3.7372956946940403e-07, + "loss": 0.1366, + "step": 3641 + }, + { + "epoch": 0.19, + "learning_rate": 3.737128157696014e-07, + "loss": 0.1246, + "step": 3642 + }, + { + "epoch": 0.19, + "learning_rate": 3.7369605710501944e-07, + "loss": 0.1108, + "step": 3643 + }, + { + "epoch": 0.19, + "learning_rate": 3.7367929347613715e-07, + "loss": 0.1755, + "step": 3644 + }, + { + "epoch": 0.19, + "learning_rate": 3.736625248834336e-07, + "loss": 0.0998, + "step": 3645 + }, + { + "epoch": 0.19, + "learning_rate": 3.73645751327388e-07, + "loss": 0.1648, + "step": 3646 + }, + { + "epoch": 0.19, + "learning_rate": 3.7362897280847984e-07, + "loss": 0.1714, + "step": 3647 + }, + { + "epoch": 0.19, + "learning_rate": 3.7361218932718857e-07, + "loss": 0.1402, + "step": 3648 + }, + { + "epoch": 0.19, + "learning_rate": 3.7359540088399397e-07, + "loss": 0.1475, + "step": 3649 + }, + { + "epoch": 0.19, + "learning_rate": 3.7357860747937577e-07, + "loss": 0.146, + "step": 3650 + }, + { + "epoch": 0.19, + "learning_rate": 3.7356180911381393e-07, + "loss": 0.11, + "step": 3651 + }, + { + "epoch": 0.19, + "learning_rate": 3.735450057877886e-07, + "loss": 0.1533, + "step": 3652 + }, + { + "epoch": 0.19, + "learning_rate": 3.7352819750178e-07, + "loss": 0.1609, + "step": 3653 + }, + { + "epoch": 0.19, + "learning_rate": 3.7351138425626855e-07, + "loss": 0.1265, + "step": 3654 + }, + { + "epoch": 0.19, + "learning_rate": 3.7349456605173475e-07, + "loss": 0.1359, + "step": 3655 + }, + { + "epoch": 0.19, + "learning_rate": 3.734777428886593e-07, + "loss": 0.1522, + "step": 3656 + }, + { + "epoch": 0.19, + "learning_rate": 3.734609147675229e-07, + "loss": 0.1735, + "step": 3657 + }, + { + "epoch": 0.19, + "learning_rate": 3.7344408168880663e-07, + "loss": 0.1595, + "step": 3658 + }, + { + "epoch": 0.19, + "learning_rate": 3.7342724365299166e-07, + "loss": 0.177, + "step": 3659 + }, + { + "epoch": 0.19, + "learning_rate": 3.73410400660559e-07, + "loss": 0.1576, + "step": 3660 + }, + { + "epoch": 0.19, + "learning_rate": 3.733935527119902e-07, + "loss": 0.1139, + "step": 3661 + }, + { + "epoch": 0.19, + "learning_rate": 3.733766998077667e-07, + "loss": 0.1514, + "step": 3662 + }, + { + "epoch": 0.19, + "learning_rate": 3.733598419483702e-07, + "loss": 0.1572, + "step": 3663 + }, + { + "epoch": 0.19, + "learning_rate": 3.733429791342825e-07, + "loss": 0.1486, + "step": 3664 + }, + { + "epoch": 0.19, + "learning_rate": 3.733261113659855e-07, + "loss": 0.1461, + "step": 3665 + }, + { + "epoch": 0.19, + "learning_rate": 3.7330923864396134e-07, + "loss": 0.14, + "step": 3666 + }, + { + "epoch": 0.19, + "learning_rate": 3.7329236096869225e-07, + "loss": 0.1458, + "step": 3667 + }, + { + "epoch": 0.19, + "learning_rate": 3.732754783406606e-07, + "loss": 0.1557, + "step": 3668 + }, + { + "epoch": 0.19, + "learning_rate": 3.7325859076034883e-07, + "loss": 0.1046, + "step": 3669 + }, + { + "epoch": 0.19, + "learning_rate": 3.7324169822823974e-07, + "loss": 0.1398, + "step": 3670 + }, + { + "epoch": 0.19, + "learning_rate": 3.7322480074481597e-07, + "loss": 0.1546, + "step": 3671 + }, + { + "epoch": 0.19, + "learning_rate": 3.7320789831056054e-07, + "loss": 0.1599, + "step": 3672 + }, + { + "epoch": 0.19, + "learning_rate": 3.731909909259565e-07, + "loss": 0.1727, + "step": 3673 + }, + { + "epoch": 0.19, + "learning_rate": 3.731740785914871e-07, + "loss": 0.1543, + "step": 3674 + }, + { + "epoch": 0.19, + "learning_rate": 3.731571613076357e-07, + "loss": 0.1552, + "step": 3675 + }, + { + "epoch": 0.19, + "learning_rate": 3.7314023907488574e-07, + "loss": 0.123, + "step": 3676 + }, + { + "epoch": 0.19, + "learning_rate": 3.7312331189372097e-07, + "loss": 0.1547, + "step": 3677 + }, + { + "epoch": 0.19, + "learning_rate": 3.731063797646251e-07, + "loss": 0.1625, + "step": 3678 + }, + { + "epoch": 0.19, + "learning_rate": 3.7308944268808203e-07, + "loss": 0.2003, + "step": 3679 + }, + { + "epoch": 0.19, + "learning_rate": 3.730725006645759e-07, + "loss": 0.1468, + "step": 3680 + }, + { + "epoch": 0.19, + "learning_rate": 3.730555536945909e-07, + "loss": 0.1368, + "step": 3681 + }, + { + "epoch": 0.19, + "learning_rate": 3.730386017786113e-07, + "loss": 0.176, + "step": 3682 + }, + { + "epoch": 0.19, + "learning_rate": 3.730216449171217e-07, + "loss": 0.1079, + "step": 3683 + }, + { + "epoch": 0.19, + "learning_rate": 3.730046831106067e-07, + "loss": 0.1729, + "step": 3684 + }, + { + "epoch": 0.19, + "learning_rate": 3.7298771635955107e-07, + "loss": 0.1793, + "step": 3685 + }, + { + "epoch": 0.19, + "learning_rate": 3.7297074466443973e-07, + "loss": 0.1285, + "step": 3686 + }, + { + "epoch": 0.19, + "learning_rate": 3.7295376802575766e-07, + "loss": 0.1687, + "step": 3687 + }, + { + "epoch": 0.19, + "learning_rate": 3.7293678644399026e-07, + "loss": 0.1611, + "step": 3688 + }, + { + "epoch": 0.19, + "learning_rate": 3.729197999196227e-07, + "loss": 0.1648, + "step": 3689 + }, + { + "epoch": 0.19, + "learning_rate": 3.729028084531404e-07, + "loss": 0.1269, + "step": 3690 + }, + { + "epoch": 0.19, + "learning_rate": 3.728858120450292e-07, + "loss": 0.1641, + "step": 3691 + }, + { + "epoch": 0.19, + "learning_rate": 3.7286881069577473e-07, + "loss": 0.1409, + "step": 3692 + }, + { + "epoch": 0.19, + "learning_rate": 3.7285180440586297e-07, + "loss": 0.1395, + "step": 3693 + }, + { + "epoch": 0.19, + "learning_rate": 3.7283479317577983e-07, + "loss": 0.1642, + "step": 3694 + }, + { + "epoch": 0.19, + "learning_rate": 3.7281777700601167e-07, + "loss": 0.1485, + "step": 3695 + }, + { + "epoch": 0.19, + "learning_rate": 3.728007558970446e-07, + "loss": 0.1718, + "step": 3696 + }, + { + "epoch": 0.19, + "learning_rate": 3.7278372984936526e-07, + "loss": 0.1483, + "step": 3697 + }, + { + "epoch": 0.19, + "learning_rate": 3.7276669886346026e-07, + "loss": 0.1345, + "step": 3698 + }, + { + "epoch": 0.19, + "learning_rate": 3.727496629398163e-07, + "loss": 0.1083, + "step": 3699 + }, + { + "epoch": 0.19, + "learning_rate": 3.7273262207892035e-07, + "loss": 0.1493, + "step": 3700 + }, + { + "epoch": 0.19, + "learning_rate": 3.727155762812592e-07, + "loss": 0.1243, + "step": 3701 + }, + { + "epoch": 0.19, + "learning_rate": 3.726985255473204e-07, + "loss": 0.1522, + "step": 3702 + }, + { + "epoch": 0.19, + "learning_rate": 3.7268146987759094e-07, + "loss": 0.1629, + "step": 3703 + }, + { + "epoch": 0.19, + "learning_rate": 3.726644092725585e-07, + "loss": 0.1593, + "step": 3704 + }, + { + "epoch": 0.19, + "learning_rate": 3.7264734373271054e-07, + "loss": 0.1877, + "step": 3705 + }, + { + "epoch": 0.19, + "learning_rate": 3.7263027325853484e-07, + "loss": 0.1195, + "step": 3706 + }, + { + "epoch": 0.19, + "learning_rate": 3.726131978505193e-07, + "loss": 0.1474, + "step": 3707 + }, + { + "epoch": 0.19, + "learning_rate": 3.7259611750915197e-07, + "loss": 0.1819, + "step": 3708 + }, + { + "epoch": 0.19, + "learning_rate": 3.725790322349209e-07, + "loss": 0.1251, + "step": 3709 + }, + { + "epoch": 0.19, + "learning_rate": 3.725619420283145e-07, + "loss": 0.1493, + "step": 3710 + }, + { + "epoch": 0.19, + "learning_rate": 3.7254484688982116e-07, + "loss": 0.1443, + "step": 3711 + }, + { + "epoch": 0.19, + "learning_rate": 3.725277468199295e-07, + "loss": 0.151, + "step": 3712 + }, + { + "epoch": 0.19, + "learning_rate": 3.725106418191283e-07, + "loss": 0.181, + "step": 3713 + }, + { + "epoch": 0.19, + "learning_rate": 3.724935318879063e-07, + "loss": 0.1511, + "step": 3714 + }, + { + "epoch": 0.19, + "learning_rate": 3.724764170267525e-07, + "loss": 0.153, + "step": 3715 + }, + { + "epoch": 0.19, + "learning_rate": 3.724592972361562e-07, + "loss": 0.1664, + "step": 3716 + }, + { + "epoch": 0.19, + "learning_rate": 3.724421725166066e-07, + "loss": 0.154, + "step": 3717 + }, + { + "epoch": 0.19, + "learning_rate": 3.7242504286859313e-07, + "loss": 0.1381, + "step": 3718 + }, + { + "epoch": 0.19, + "learning_rate": 3.724079082926054e-07, + "loss": 0.1605, + "step": 3719 + }, + { + "epoch": 0.19, + "learning_rate": 3.723907687891331e-07, + "loss": 0.1417, + "step": 3720 + }, + { + "epoch": 0.19, + "learning_rate": 3.7237362435866605e-07, + "loss": 0.1504, + "step": 3721 + }, + { + "epoch": 0.19, + "learning_rate": 3.723564750016943e-07, + "loss": 0.1427, + "step": 3722 + }, + { + "epoch": 0.19, + "learning_rate": 3.723393207187079e-07, + "loss": 0.1309, + "step": 3723 + }, + { + "epoch": 0.19, + "learning_rate": 3.7232216151019726e-07, + "loss": 0.1394, + "step": 3724 + }, + { + "epoch": 0.19, + "learning_rate": 3.723049973766526e-07, + "loss": 0.1414, + "step": 3725 + }, + { + "epoch": 0.19, + "learning_rate": 3.722878283185647e-07, + "loss": 0.1506, + "step": 3726 + }, + { + "epoch": 0.19, + "learning_rate": 3.722706543364242e-07, + "loss": 0.1357, + "step": 3727 + }, + { + "epoch": 0.19, + "learning_rate": 3.722534754307218e-07, + "loss": 0.1454, + "step": 3728 + }, + { + "epoch": 0.19, + "learning_rate": 3.722362916019486e-07, + "loss": 0.1169, + "step": 3729 + }, + { + "epoch": 0.19, + "learning_rate": 3.722191028505957e-07, + "loss": 0.1426, + "step": 3730 + }, + { + "epoch": 0.19, + "learning_rate": 3.7220190917715435e-07, + "loss": 0.1436, + "step": 3731 + }, + { + "epoch": 0.19, + "learning_rate": 3.72184710582116e-07, + "loss": 0.1471, + "step": 3732 + }, + { + "epoch": 0.19, + "learning_rate": 3.721675070659721e-07, + "loss": 0.1914, + "step": 3733 + }, + { + "epoch": 0.19, + "learning_rate": 3.721502986292144e-07, + "loss": 0.1532, + "step": 3734 + }, + { + "epoch": 0.19, + "learning_rate": 3.7213308527233476e-07, + "loss": 0.1388, + "step": 3735 + }, + { + "epoch": 0.2, + "learning_rate": 3.7211586699582504e-07, + "loss": 0.145, + "step": 3736 + }, + { + "epoch": 0.2, + "learning_rate": 3.720986438001774e-07, + "loss": 0.156, + "step": 3737 + }, + { + "epoch": 0.2, + "learning_rate": 3.7208141568588406e-07, + "loss": 0.1636, + "step": 3738 + }, + { + "epoch": 0.2, + "learning_rate": 3.7206418265343747e-07, + "loss": 0.145, + "step": 3739 + }, + { + "epoch": 0.2, + "learning_rate": 3.7204694470333013e-07, + "loss": 0.1508, + "step": 3740 + }, + { + "epoch": 0.2, + "learning_rate": 3.720297018360546e-07, + "loss": 0.1469, + "step": 3741 + }, + { + "epoch": 0.2, + "learning_rate": 3.720124540521039e-07, + "loss": 0.1267, + "step": 3742 + }, + { + "epoch": 0.2, + "learning_rate": 3.719952013519708e-07, + "loss": 0.2155, + "step": 3743 + }, + { + "epoch": 0.2, + "learning_rate": 3.719779437361485e-07, + "loss": 0.1269, + "step": 3744 + }, + { + "epoch": 0.2, + "learning_rate": 3.7196068120513014e-07, + "loss": 0.1419, + "step": 3745 + }, + { + "epoch": 0.2, + "learning_rate": 3.719434137594091e-07, + "loss": 0.1626, + "step": 3746 + }, + { + "epoch": 0.2, + "learning_rate": 3.7192614139947895e-07, + "loss": 0.1531, + "step": 3747 + }, + { + "epoch": 0.2, + "learning_rate": 3.7190886412583326e-07, + "loss": 0.1603, + "step": 3748 + }, + { + "epoch": 0.2, + "learning_rate": 3.7189158193896593e-07, + "loss": 0.1235, + "step": 3749 + }, + { + "epoch": 0.2, + "learning_rate": 3.718742948393708e-07, + "loss": 0.1627, + "step": 3750 + }, + { + "epoch": 0.2, + "learning_rate": 3.7185700282754195e-07, + "loss": 0.1427, + "step": 3751 + }, + { + "epoch": 0.2, + "learning_rate": 3.718397059039737e-07, + "loss": 0.1491, + "step": 3752 + }, + { + "epoch": 0.2, + "learning_rate": 3.718224040691603e-07, + "loss": 0.1748, + "step": 3753 + }, + { + "epoch": 0.2, + "learning_rate": 3.7180509732359617e-07, + "loss": 0.1542, + "step": 3754 + }, + { + "epoch": 0.2, + "learning_rate": 3.717877856677761e-07, + "loss": 0.1518, + "step": 3755 + }, + { + "epoch": 0.2, + "learning_rate": 3.7177046910219485e-07, + "loss": 0.143, + "step": 3756 + }, + { + "epoch": 0.2, + "learning_rate": 3.7175314762734723e-07, + "loss": 0.1398, + "step": 3757 + }, + { + "epoch": 0.2, + "learning_rate": 3.7173582124372833e-07, + "loss": 0.1579, + "step": 3758 + }, + { + "epoch": 0.2, + "learning_rate": 3.717184899518334e-07, + "loss": 0.1266, + "step": 3759 + }, + { + "epoch": 0.2, + "learning_rate": 3.7170115375215766e-07, + "loss": 0.1733, + "step": 3760 + }, + { + "epoch": 0.2, + "learning_rate": 3.716838126451967e-07, + "loss": 0.1607, + "step": 3761 + }, + { + "epoch": 0.2, + "learning_rate": 3.7166646663144616e-07, + "loss": 0.1472, + "step": 3762 + }, + { + "epoch": 0.2, + "learning_rate": 3.716491157114017e-07, + "loss": 0.1864, + "step": 3763 + }, + { + "epoch": 0.2, + "learning_rate": 3.716317598855592e-07, + "loss": 0.1393, + "step": 3764 + }, + { + "epoch": 0.2, + "learning_rate": 3.716143991544148e-07, + "loss": 0.1209, + "step": 3765 + }, + { + "epoch": 0.2, + "learning_rate": 3.7159703351846463e-07, + "loss": 0.143, + "step": 3766 + }, + { + "epoch": 0.2, + "learning_rate": 3.7157966297820493e-07, + "loss": 0.1485, + "step": 3767 + }, + { + "epoch": 0.2, + "learning_rate": 3.715622875341323e-07, + "loss": 0.1478, + "step": 3768 + }, + { + "epoch": 0.2, + "learning_rate": 3.7154490718674325e-07, + "loss": 0.1952, + "step": 3769 + }, + { + "epoch": 0.2, + "learning_rate": 3.7152752193653447e-07, + "loss": 0.1434, + "step": 3770 + }, + { + "epoch": 0.2, + "learning_rate": 3.7151013178400297e-07, + "loss": 0.1518, + "step": 3771 + }, + { + "epoch": 0.2, + "learning_rate": 3.7149273672964564e-07, + "loss": 0.1267, + "step": 3772 + }, + { + "epoch": 0.2, + "learning_rate": 3.7147533677395974e-07, + "loss": 0.1356, + "step": 3773 + }, + { + "epoch": 0.2, + "learning_rate": 3.714579319174425e-07, + "loss": 0.1285, + "step": 3774 + }, + { + "epoch": 0.2, + "learning_rate": 3.7144052216059134e-07, + "loss": 0.1375, + "step": 3775 + }, + { + "epoch": 0.2, + "learning_rate": 3.714231075039039e-07, + "loss": 0.1549, + "step": 3776 + }, + { + "epoch": 0.2, + "learning_rate": 3.7140568794787787e-07, + "loss": 0.1412, + "step": 3777 + }, + { + "epoch": 0.2, + "learning_rate": 3.713882634930111e-07, + "loss": 0.1422, + "step": 3778 + }, + { + "epoch": 0.2, + "learning_rate": 3.713708341398017e-07, + "loss": 0.1332, + "step": 3779 + }, + { + "epoch": 0.2, + "learning_rate": 3.7135339988874757e-07, + "loss": 0.1571, + "step": 3780 + }, + { + "epoch": 0.2, + "learning_rate": 3.713359607403472e-07, + "loss": 0.1638, + "step": 3781 + }, + { + "epoch": 0.2, + "learning_rate": 3.713185166950989e-07, + "loss": 0.1478, + "step": 3782 + }, + { + "epoch": 0.2, + "learning_rate": 3.7130106775350124e-07, + "loss": 0.1477, + "step": 3783 + }, + { + "epoch": 0.2, + "learning_rate": 3.7128361391605295e-07, + "loss": 0.1568, + "step": 3784 + }, + { + "epoch": 0.2, + "learning_rate": 3.7126615518325287e-07, + "loss": 0.14, + "step": 3785 + }, + { + "epoch": 0.2, + "learning_rate": 3.712486915555999e-07, + "loss": 0.1302, + "step": 3786 + }, + { + "epoch": 0.2, + "learning_rate": 3.712312230335933e-07, + "loss": 0.1387, + "step": 3787 + }, + { + "epoch": 0.2, + "learning_rate": 3.712137496177322e-07, + "loss": 0.1634, + "step": 3788 + }, + { + "epoch": 0.2, + "learning_rate": 3.7119627130851604e-07, + "loss": 0.1484, + "step": 3789 + }, + { + "epoch": 0.2, + "learning_rate": 3.7117878810644437e-07, + "loss": 0.1383, + "step": 3790 + }, + { + "epoch": 0.2, + "learning_rate": 3.7116130001201686e-07, + "loss": 0.149, + "step": 3791 + }, + { + "epoch": 0.2, + "learning_rate": 3.7114380702573325e-07, + "loss": 0.1603, + "step": 3792 + }, + { + "epoch": 0.2, + "learning_rate": 3.711263091480936e-07, + "loss": 0.1384, + "step": 3793 + }, + { + "epoch": 0.2, + "learning_rate": 3.7110880637959803e-07, + "loss": 0.1304, + "step": 3794 + }, + { + "epoch": 0.2, + "learning_rate": 3.710912987207467e-07, + "loss": 0.1409, + "step": 3795 + }, + { + "epoch": 0.2, + "learning_rate": 3.7107378617204e-07, + "loss": 0.1748, + "step": 3796 + }, + { + "epoch": 0.2, + "learning_rate": 3.7105626873397846e-07, + "loss": 0.1449, + "step": 3797 + }, + { + "epoch": 0.2, + "learning_rate": 3.710387464070627e-07, + "loss": 0.129, + "step": 3798 + }, + { + "epoch": 0.2, + "learning_rate": 3.710212191917935e-07, + "loss": 0.1353, + "step": 3799 + }, + { + "epoch": 0.2, + "learning_rate": 3.710036870886719e-07, + "loss": 0.1703, + "step": 3800 + }, + { + "epoch": 0.2, + "learning_rate": 3.7098615009819894e-07, + "loss": 0.1503, + "step": 3801 + }, + { + "epoch": 0.2, + "learning_rate": 3.709686082208757e-07, + "loss": 0.1854, + "step": 3802 + }, + { + "epoch": 0.2, + "learning_rate": 3.7095106145720367e-07, + "loss": 0.1375, + "step": 3803 + }, + { + "epoch": 0.2, + "learning_rate": 3.7093350980768437e-07, + "loss": 0.141, + "step": 3804 + }, + { + "epoch": 0.2, + "learning_rate": 3.709159532728193e-07, + "loss": 0.1543, + "step": 3805 + }, + { + "epoch": 0.2, + "learning_rate": 3.7089839185311036e-07, + "loss": 0.1519, + "step": 3806 + }, + { + "epoch": 0.2, + "learning_rate": 3.708808255490594e-07, + "loss": 0.1288, + "step": 3807 + }, + { + "epoch": 0.2, + "learning_rate": 3.708632543611684e-07, + "loss": 0.1288, + "step": 3808 + }, + { + "epoch": 0.2, + "learning_rate": 3.7084567828993976e-07, + "loss": 0.1192, + "step": 3809 + }, + { + "epoch": 0.2, + "learning_rate": 3.708280973358756e-07, + "loss": 0.1222, + "step": 3810 + }, + { + "epoch": 0.2, + "learning_rate": 3.7081051149947846e-07, + "loss": 0.15, + "step": 3811 + }, + { + "epoch": 0.2, + "learning_rate": 3.70792920781251e-07, + "loss": 0.1569, + "step": 3812 + }, + { + "epoch": 0.2, + "learning_rate": 3.7077532518169595e-07, + "loss": 0.1346, + "step": 3813 + }, + { + "epoch": 0.2, + "learning_rate": 3.7075772470131614e-07, + "loss": 0.1609, + "step": 3814 + }, + { + "epoch": 0.2, + "learning_rate": 3.7074011934061466e-07, + "loss": 0.1562, + "step": 3815 + }, + { + "epoch": 0.2, + "learning_rate": 3.7072250910009465e-07, + "loss": 0.1987, + "step": 3816 + }, + { + "epoch": 0.2, + "learning_rate": 3.707048939802594e-07, + "loss": 0.1378, + "step": 3817 + }, + { + "epoch": 0.2, + "learning_rate": 3.706872739816124e-07, + "loss": 0.165, + "step": 3818 + }, + { + "epoch": 0.2, + "learning_rate": 3.706696491046572e-07, + "loss": 0.1292, + "step": 3819 + }, + { + "epoch": 0.2, + "learning_rate": 3.7065201934989763e-07, + "loss": 0.2017, + "step": 3820 + }, + { + "epoch": 0.2, + "learning_rate": 3.7063438471783734e-07, + "loss": 0.1399, + "step": 3821 + }, + { + "epoch": 0.2, + "learning_rate": 3.706167452089805e-07, + "loss": 0.1613, + "step": 3822 + }, + { + "epoch": 0.2, + "learning_rate": 3.7059910082383125e-07, + "loss": 0.1794, + "step": 3823 + }, + { + "epoch": 0.2, + "learning_rate": 3.705814515628939e-07, + "loss": 0.1278, + "step": 3824 + }, + { + "epoch": 0.2, + "learning_rate": 3.705637974266727e-07, + "loss": 0.212, + "step": 3825 + }, + { + "epoch": 0.2, + "learning_rate": 3.705461384156724e-07, + "loss": 0.1445, + "step": 3826 + }, + { + "epoch": 0.2, + "learning_rate": 3.7052847453039757e-07, + "loss": 0.1344, + "step": 3827 + }, + { + "epoch": 0.2, + "learning_rate": 3.7051080577135316e-07, + "loss": 0.1298, + "step": 3828 + }, + { + "epoch": 0.2, + "learning_rate": 3.7049313213904407e-07, + "loss": 0.1514, + "step": 3829 + }, + { + "epoch": 0.2, + "learning_rate": 3.704754536339754e-07, + "loss": 0.1577, + "step": 3830 + }, + { + "epoch": 0.2, + "learning_rate": 3.704577702566525e-07, + "loss": 0.1134, + "step": 3831 + }, + { + "epoch": 0.2, + "learning_rate": 3.704400820075808e-07, + "loss": 0.1398, + "step": 3832 + }, + { + "epoch": 0.2, + "learning_rate": 3.704223888872656e-07, + "loss": 0.1626, + "step": 3833 + }, + { + "epoch": 0.2, + "learning_rate": 3.704046908962129e-07, + "loss": 0.1299, + "step": 3834 + }, + { + "epoch": 0.2, + "learning_rate": 3.7038698803492825e-07, + "loss": 0.1575, + "step": 3835 + }, + { + "epoch": 0.2, + "learning_rate": 3.7036928030391773e-07, + "loss": 0.1534, + "step": 3836 + }, + { + "epoch": 0.2, + "learning_rate": 3.703515677036874e-07, + "loss": 0.1683, + "step": 3837 + }, + { + "epoch": 0.2, + "learning_rate": 3.703338502347435e-07, + "loss": 0.1942, + "step": 3838 + }, + { + "epoch": 0.2, + "learning_rate": 3.703161278975925e-07, + "loss": 0.1317, + "step": 3839 + }, + { + "epoch": 0.2, + "learning_rate": 3.702984006927407e-07, + "loss": 0.122, + "step": 3840 + }, + { + "epoch": 0.2, + "learning_rate": 3.702806686206949e-07, + "loss": 0.155, + "step": 3841 + }, + { + "epoch": 0.2, + "learning_rate": 3.702629316819618e-07, + "loss": 0.1554, + "step": 3842 + }, + { + "epoch": 0.2, + "learning_rate": 3.7024518987704845e-07, + "loss": 0.1412, + "step": 3843 + }, + { + "epoch": 0.2, + "learning_rate": 3.702274432064619e-07, + "loss": 0.1309, + "step": 3844 + }, + { + "epoch": 0.2, + "learning_rate": 3.7020969167070924e-07, + "loss": 0.1301, + "step": 3845 + }, + { + "epoch": 0.2, + "learning_rate": 3.701919352702979e-07, + "loss": 0.1159, + "step": 3846 + }, + { + "epoch": 0.2, + "learning_rate": 3.701741740057353e-07, + "loss": 0.1629, + "step": 3847 + }, + { + "epoch": 0.2, + "learning_rate": 3.7015640787752917e-07, + "loss": 0.1365, + "step": 3848 + }, + { + "epoch": 0.2, + "learning_rate": 3.701386368861872e-07, + "loss": 0.165, + "step": 3849 + }, + { + "epoch": 0.2, + "learning_rate": 3.7012086103221736e-07, + "loss": 0.141, + "step": 3850 + }, + { + "epoch": 0.2, + "learning_rate": 3.7010308031612753e-07, + "loss": 0.1461, + "step": 3851 + }, + { + "epoch": 0.2, + "learning_rate": 3.700852947384261e-07, + "loss": 0.1306, + "step": 3852 + }, + { + "epoch": 0.2, + "learning_rate": 3.700675042996212e-07, + "loss": 0.1509, + "step": 3853 + }, + { + "epoch": 0.2, + "learning_rate": 3.7004970900022146e-07, + "loss": 0.1511, + "step": 3854 + }, + { + "epoch": 0.2, + "learning_rate": 3.700319088407354e-07, + "loss": 0.1346, + "step": 3855 + }, + { + "epoch": 0.2, + "learning_rate": 3.7001410382167164e-07, + "loss": 0.1491, + "step": 3856 + }, + { + "epoch": 0.2, + "learning_rate": 3.6999629394353923e-07, + "loss": 0.1222, + "step": 3857 + }, + { + "epoch": 0.2, + "learning_rate": 3.6997847920684706e-07, + "loss": 0.1323, + "step": 3858 + }, + { + "epoch": 0.2, + "learning_rate": 3.699606596121044e-07, + "loss": 0.1264, + "step": 3859 + }, + { + "epoch": 0.2, + "learning_rate": 3.6994283515982046e-07, + "loss": 0.1108, + "step": 3860 + }, + { + "epoch": 0.2, + "learning_rate": 3.699250058505047e-07, + "loss": 0.1153, + "step": 3861 + }, + { + "epoch": 0.2, + "learning_rate": 3.6990717168466665e-07, + "loss": 0.1387, + "step": 3862 + }, + { + "epoch": 0.2, + "learning_rate": 3.6988933266281604e-07, + "loss": 0.1744, + "step": 3863 + }, + { + "epoch": 0.2, + "learning_rate": 3.698714887854627e-07, + "loss": 0.1207, + "step": 3864 + }, + { + "epoch": 0.2, + "learning_rate": 3.698536400531167e-07, + "loss": 0.1973, + "step": 3865 + }, + { + "epoch": 0.2, + "learning_rate": 3.6983578646628803e-07, + "loss": 0.1761, + "step": 3866 + }, + { + "epoch": 0.2, + "learning_rate": 3.69817928025487e-07, + "loss": 0.1161, + "step": 3867 + }, + { + "epoch": 0.2, + "learning_rate": 3.698000647312241e-07, + "loss": 0.1911, + "step": 3868 + }, + { + "epoch": 0.2, + "learning_rate": 3.6978219658400983e-07, + "loss": 0.2243, + "step": 3869 + }, + { + "epoch": 0.2, + "learning_rate": 3.6976432358435475e-07, + "loss": 0.1454, + "step": 3870 + }, + { + "epoch": 0.2, + "learning_rate": 3.6974644573276985e-07, + "loss": 0.1659, + "step": 3871 + }, + { + "epoch": 0.2, + "learning_rate": 3.697285630297659e-07, + "loss": 0.1685, + "step": 3872 + }, + { + "epoch": 0.2, + "learning_rate": 3.6971067547585417e-07, + "loss": 0.1409, + "step": 3873 + }, + { + "epoch": 0.2, + "learning_rate": 3.6969278307154586e-07, + "loss": 0.1357, + "step": 3874 + }, + { + "epoch": 0.2, + "learning_rate": 3.696748858173522e-07, + "loss": 0.2187, + "step": 3875 + }, + { + "epoch": 0.2, + "learning_rate": 3.696569837137849e-07, + "loss": 0.1289, + "step": 3876 + }, + { + "epoch": 0.2, + "learning_rate": 3.696390767613554e-07, + "loss": 0.1264, + "step": 3877 + }, + { + "epoch": 0.2, + "learning_rate": 3.6962116496057563e-07, + "loss": 0.1221, + "step": 3878 + }, + { + "epoch": 0.2, + "learning_rate": 3.6960324831195753e-07, + "loss": 0.1322, + "step": 3879 + }, + { + "epoch": 0.2, + "learning_rate": 3.695853268160131e-07, + "loss": 0.1608, + "step": 3880 + }, + { + "epoch": 0.2, + "learning_rate": 3.695674004732546e-07, + "loss": 0.1795, + "step": 3881 + }, + { + "epoch": 0.2, + "learning_rate": 3.695494692841943e-07, + "loss": 0.1564, + "step": 3882 + }, + { + "epoch": 0.2, + "learning_rate": 3.695315332493447e-07, + "loss": 0.1344, + "step": 3883 + }, + { + "epoch": 0.2, + "learning_rate": 3.695135923692184e-07, + "loss": 0.166, + "step": 3884 + }, + { + "epoch": 0.2, + "learning_rate": 3.6949564664432833e-07, + "loss": 0.1939, + "step": 3885 + }, + { + "epoch": 0.2, + "learning_rate": 3.6947769607518715e-07, + "loss": 0.1527, + "step": 3886 + }, + { + "epoch": 0.2, + "learning_rate": 3.6945974066230796e-07, + "loss": 0.1685, + "step": 3887 + }, + { + "epoch": 0.2, + "learning_rate": 3.69441780406204e-07, + "loss": 0.1316, + "step": 3888 + }, + { + "epoch": 0.2, + "learning_rate": 3.694238153073886e-07, + "loss": 0.1426, + "step": 3889 + }, + { + "epoch": 0.2, + "learning_rate": 3.694058453663751e-07, + "loss": 0.1579, + "step": 3890 + }, + { + "epoch": 0.2, + "learning_rate": 3.693878705836771e-07, + "loss": 0.1391, + "step": 3891 + }, + { + "epoch": 0.2, + "learning_rate": 3.693698909598084e-07, + "loss": 0.1819, + "step": 3892 + }, + { + "epoch": 0.2, + "learning_rate": 3.693519064952829e-07, + "loss": 0.1418, + "step": 3893 + }, + { + "epoch": 0.2, + "learning_rate": 3.6933391719061445e-07, + "loss": 0.1295, + "step": 3894 + }, + { + "epoch": 0.2, + "learning_rate": 3.6931592304631736e-07, + "loss": 0.1421, + "step": 3895 + }, + { + "epoch": 0.2, + "learning_rate": 3.6929792406290574e-07, + "loss": 0.1339, + "step": 3896 + }, + { + "epoch": 0.2, + "learning_rate": 3.692799202408941e-07, + "loss": 0.1602, + "step": 3897 + }, + { + "epoch": 0.2, + "learning_rate": 3.6926191158079705e-07, + "loss": 0.1521, + "step": 3898 + }, + { + "epoch": 0.2, + "learning_rate": 3.692438980831292e-07, + "loss": 0.1274, + "step": 3899 + }, + { + "epoch": 0.2, + "learning_rate": 3.692258797484054e-07, + "loss": 0.1522, + "step": 3900 + }, + { + "epoch": 0.2, + "learning_rate": 3.6920785657714063e-07, + "loss": 0.147, + "step": 3901 + }, + { + "epoch": 0.2, + "learning_rate": 3.6918982856985004e-07, + "loss": 0.1306, + "step": 3902 + }, + { + "epoch": 0.2, + "learning_rate": 3.691717957270488e-07, + "loss": 0.1401, + "step": 3903 + }, + { + "epoch": 0.2, + "learning_rate": 3.691537580492523e-07, + "loss": 0.1168, + "step": 3904 + }, + { + "epoch": 0.2, + "learning_rate": 3.691357155369761e-07, + "loss": 0.1946, + "step": 3905 + }, + { + "epoch": 0.2, + "learning_rate": 3.691176681907359e-07, + "loss": 0.1819, + "step": 3906 + }, + { + "epoch": 0.2, + "learning_rate": 3.690996160110474e-07, + "loss": 0.1433, + "step": 3907 + }, + { + "epoch": 0.2, + "learning_rate": 3.690815589984267e-07, + "loss": 0.1477, + "step": 3908 + }, + { + "epoch": 0.2, + "learning_rate": 3.690634971533897e-07, + "loss": 0.1732, + "step": 3909 + }, + { + "epoch": 0.2, + "learning_rate": 3.690454304764527e-07, + "loss": 0.1724, + "step": 3910 + }, + { + "epoch": 0.2, + "learning_rate": 3.69027358968132e-07, + "loss": 0.1313, + "step": 3911 + }, + { + "epoch": 0.2, + "learning_rate": 3.690092826289442e-07, + "loss": 0.1403, + "step": 3912 + }, + { + "epoch": 0.2, + "learning_rate": 3.689912014594058e-07, + "loss": 0.125, + "step": 3913 + }, + { + "epoch": 0.2, + "learning_rate": 3.689731154600336e-07, + "loss": 0.1276, + "step": 3914 + }, + { + "epoch": 0.2, + "learning_rate": 3.6895502463134463e-07, + "loss": 0.1535, + "step": 3915 + }, + { + "epoch": 0.2, + "learning_rate": 3.6893692897385574e-07, + "loss": 0.1443, + "step": 3916 + }, + { + "epoch": 0.2, + "learning_rate": 3.6891882848808424e-07, + "loss": 0.1317, + "step": 3917 + }, + { + "epoch": 0.2, + "learning_rate": 3.6890072317454745e-07, + "loss": 0.1415, + "step": 3918 + }, + { + "epoch": 0.2, + "learning_rate": 3.6888261303376274e-07, + "loss": 0.1701, + "step": 3919 + }, + { + "epoch": 0.2, + "learning_rate": 3.688644980662477e-07, + "loss": 0.1789, + "step": 3920 + }, + { + "epoch": 0.2, + "learning_rate": 3.688463782725202e-07, + "loss": 0.1505, + "step": 3921 + }, + { + "epoch": 0.2, + "learning_rate": 3.6882825365309805e-07, + "loss": 0.1665, + "step": 3922 + }, + { + "epoch": 0.2, + "learning_rate": 3.688101242084992e-07, + "loss": 0.1183, + "step": 3923 + }, + { + "epoch": 0.2, + "learning_rate": 3.687919899392418e-07, + "loss": 0.1203, + "step": 3924 + }, + { + "epoch": 0.2, + "learning_rate": 3.687738508458442e-07, + "loss": 0.1684, + "step": 3925 + }, + { + "epoch": 0.2, + "learning_rate": 3.687557069288248e-07, + "loss": 0.1508, + "step": 3926 + }, + { + "epoch": 0.2, + "learning_rate": 3.6873755818870215e-07, + "loss": 0.1254, + "step": 3927 + }, + { + "epoch": 0.21, + "learning_rate": 3.687194046259949e-07, + "loss": 0.1344, + "step": 3928 + }, + { + "epoch": 0.21, + "learning_rate": 3.68701246241222e-07, + "loss": 0.1295, + "step": 3929 + }, + { + "epoch": 0.21, + "learning_rate": 3.686830830349023e-07, + "loss": 0.1435, + "step": 3930 + }, + { + "epoch": 0.21, + "learning_rate": 3.68664915007555e-07, + "loss": 0.1354, + "step": 3931 + }, + { + "epoch": 0.21, + "learning_rate": 3.686467421596993e-07, + "loss": 0.1245, + "step": 3932 + }, + { + "epoch": 0.21, + "learning_rate": 3.686285644918547e-07, + "loss": 0.1453, + "step": 3933 + }, + { + "epoch": 0.21, + "learning_rate": 3.686103820045405e-07, + "loss": 0.148, + "step": 3934 + }, + { + "epoch": 0.21, + "learning_rate": 3.6859219469827656e-07, + "loss": 0.1529, + "step": 3935 + }, + { + "epoch": 0.21, + "learning_rate": 3.685740025735826e-07, + "loss": 0.1411, + "step": 3936 + }, + { + "epoch": 0.21, + "learning_rate": 3.6855580563097855e-07, + "loss": 0.1811, + "step": 3937 + }, + { + "epoch": 0.21, + "learning_rate": 3.685376038709845e-07, + "loss": 0.1702, + "step": 3938 + }, + { + "epoch": 0.21, + "learning_rate": 3.685193972941207e-07, + "loss": 0.1082, + "step": 3939 + }, + { + "epoch": 0.21, + "learning_rate": 3.685011859009075e-07, + "loss": 0.1569, + "step": 3940 + }, + { + "epoch": 0.21, + "learning_rate": 3.6848296969186524e-07, + "loss": 0.1663, + "step": 3941 + }, + { + "epoch": 0.21, + "learning_rate": 3.6846474866751477e-07, + "loss": 0.1208, + "step": 3942 + }, + { + "epoch": 0.21, + "learning_rate": 3.684465228283767e-07, + "loss": 0.1765, + "step": 3943 + }, + { + "epoch": 0.21, + "learning_rate": 3.68428292174972e-07, + "loss": 0.1645, + "step": 3944 + }, + { + "epoch": 0.21, + "learning_rate": 3.684100567078217e-07, + "loss": 0.1574, + "step": 3945 + }, + { + "epoch": 0.21, + "learning_rate": 3.6839181642744697e-07, + "loss": 0.1548, + "step": 3946 + }, + { + "epoch": 0.21, + "learning_rate": 3.683735713343691e-07, + "loss": 0.1459, + "step": 3947 + }, + { + "epoch": 0.21, + "learning_rate": 3.6835532142910956e-07, + "loss": 0.1236, + "step": 3948 + }, + { + "epoch": 0.21, + "learning_rate": 3.6833706671219e-07, + "loss": 0.133, + "step": 3949 + }, + { + "epoch": 0.21, + "learning_rate": 3.68318807184132e-07, + "loss": 0.1365, + "step": 3950 + }, + { + "epoch": 0.21, + "learning_rate": 3.683005428454575e-07, + "loss": 0.1425, + "step": 3951 + }, + { + "epoch": 0.21, + "learning_rate": 3.682822736966886e-07, + "loss": 0.1444, + "step": 3952 + }, + { + "epoch": 0.21, + "learning_rate": 3.6826399973834734e-07, + "loss": 0.1385, + "step": 3953 + }, + { + "epoch": 0.21, + "learning_rate": 3.68245720970956e-07, + "loss": 0.1494, + "step": 3954 + }, + { + "epoch": 0.21, + "learning_rate": 3.6822743739503703e-07, + "loss": 0.1382, + "step": 3955 + }, + { + "epoch": 0.21, + "learning_rate": 3.682091490111129e-07, + "loss": 0.1556, + "step": 3956 + }, + { + "epoch": 0.21, + "learning_rate": 3.681908558197064e-07, + "loss": 0.1412, + "step": 3957 + }, + { + "epoch": 0.21, + "learning_rate": 3.6817255782134027e-07, + "loss": 0.1425, + "step": 3958 + }, + { + "epoch": 0.21, + "learning_rate": 3.681542550165376e-07, + "loss": 0.1276, + "step": 3959 + }, + { + "epoch": 0.21, + "learning_rate": 3.681359474058213e-07, + "loss": 0.1511, + "step": 3960 + }, + { + "epoch": 0.21, + "learning_rate": 3.6811763498971473e-07, + "loss": 0.133, + "step": 3961 + }, + { + "epoch": 0.21, + "learning_rate": 3.680993177687413e-07, + "loss": 0.1543, + "step": 3962 + }, + { + "epoch": 0.21, + "learning_rate": 3.6808099574342447e-07, + "loss": 0.1698, + "step": 3963 + }, + { + "epoch": 0.21, + "learning_rate": 3.6806266891428786e-07, + "loss": 0.1736, + "step": 3964 + }, + { + "epoch": 0.21, + "learning_rate": 3.680443372818554e-07, + "loss": 0.1939, + "step": 3965 + }, + { + "epoch": 0.21, + "learning_rate": 3.6802600084665077e-07, + "loss": 0.1474, + "step": 3966 + }, + { + "epoch": 0.21, + "learning_rate": 3.680076596091982e-07, + "loss": 0.1395, + "step": 3967 + }, + { + "epoch": 0.21, + "learning_rate": 3.67989313570022e-07, + "loss": 0.1401, + "step": 3968 + }, + { + "epoch": 0.21, + "learning_rate": 3.679709627296462e-07, + "loss": 0.1464, + "step": 3969 + }, + { + "epoch": 0.21, + "learning_rate": 3.679526070885955e-07, + "loss": 0.1467, + "step": 3970 + }, + { + "epoch": 0.21, + "learning_rate": 3.6793424664739446e-07, + "loss": 0.1818, + "step": 3971 + }, + { + "epoch": 0.21, + "learning_rate": 3.6791588140656784e-07, + "loss": 0.131, + "step": 3972 + }, + { + "epoch": 0.21, + "learning_rate": 3.6789751136664046e-07, + "loss": 0.1348, + "step": 3973 + }, + { + "epoch": 0.21, + "learning_rate": 3.678791365281374e-07, + "loss": 0.1239, + "step": 3974 + }, + { + "epoch": 0.21, + "learning_rate": 3.678607568915838e-07, + "loss": 0.1678, + "step": 3975 + }, + { + "epoch": 0.21, + "learning_rate": 3.6784237245750507e-07, + "loss": 0.1659, + "step": 3976 + }, + { + "epoch": 0.21, + "learning_rate": 3.6782398322642643e-07, + "loss": 0.1635, + "step": 3977 + }, + { + "epoch": 0.21, + "learning_rate": 3.678055891988736e-07, + "loss": 0.1328, + "step": 3978 + }, + { + "epoch": 0.21, + "learning_rate": 3.677871903753723e-07, + "loss": 0.1749, + "step": 3979 + }, + { + "epoch": 0.21, + "learning_rate": 3.6776878675644823e-07, + "loss": 0.1463, + "step": 3980 + }, + { + "epoch": 0.21, + "learning_rate": 3.6775037834262757e-07, + "loss": 0.1325, + "step": 3981 + }, + { + "epoch": 0.21, + "learning_rate": 3.677319651344362e-07, + "loss": 0.1454, + "step": 3982 + }, + { + "epoch": 0.21, + "learning_rate": 3.677135471324006e-07, + "loss": 0.149, + "step": 3983 + }, + { + "epoch": 0.21, + "learning_rate": 3.676951243370471e-07, + "loss": 0.1619, + "step": 3984 + }, + { + "epoch": 0.21, + "learning_rate": 3.6767669674890217e-07, + "loss": 0.1804, + "step": 3985 + }, + { + "epoch": 0.21, + "learning_rate": 3.676582643684925e-07, + "loss": 0.1574, + "step": 3986 + }, + { + "epoch": 0.21, + "learning_rate": 3.67639827196345e-07, + "loss": 0.1781, + "step": 3987 + }, + { + "epoch": 0.21, + "learning_rate": 3.6762138523298647e-07, + "loss": 0.1455, + "step": 3988 + }, + { + "epoch": 0.21, + "learning_rate": 3.676029384789441e-07, + "loss": 0.1327, + "step": 3989 + }, + { + "epoch": 0.21, + "learning_rate": 3.6758448693474494e-07, + "loss": 0.1518, + "step": 3990 + }, + { + "epoch": 0.21, + "learning_rate": 3.675660306009165e-07, + "loss": 0.1576, + "step": 3991 + }, + { + "epoch": 0.21, + "learning_rate": 3.675475694779862e-07, + "loss": 0.1352, + "step": 3992 + }, + { + "epoch": 0.21, + "learning_rate": 3.6752910356648175e-07, + "loss": 0.1342, + "step": 3993 + }, + { + "epoch": 0.21, + "learning_rate": 3.6751063286693076e-07, + "loss": 0.1678, + "step": 3994 + }, + { + "epoch": 0.21, + "learning_rate": 3.6749215737986125e-07, + "loss": 0.1541, + "step": 3995 + }, + { + "epoch": 0.21, + "learning_rate": 3.6747367710580124e-07, + "loss": 0.1589, + "step": 3996 + }, + { + "epoch": 0.21, + "learning_rate": 3.6745519204527885e-07, + "loss": 0.1557, + "step": 3997 + }, + { + "epoch": 0.21, + "learning_rate": 3.6743670219882247e-07, + "loss": 0.1213, + "step": 3998 + }, + { + "epoch": 0.21, + "learning_rate": 3.674182075669605e-07, + "loss": 0.1549, + "step": 3999 + }, + { + "epoch": 0.21, + "learning_rate": 3.673997081502215e-07, + "loss": 0.1621, + "step": 4000 + }, + { + "epoch": 0.21, + "learning_rate": 3.673812039491342e-07, + "loss": 0.1439, + "step": 4001 + }, + { + "epoch": 0.21, + "learning_rate": 3.6736269496422753e-07, + "loss": 0.1447, + "step": 4002 + }, + { + "epoch": 0.21, + "learning_rate": 3.673441811960304e-07, + "loss": 0.1685, + "step": 4003 + }, + { + "epoch": 0.21, + "learning_rate": 3.67325662645072e-07, + "loss": 0.1386, + "step": 4004 + }, + { + "epoch": 0.21, + "learning_rate": 3.673071393118816e-07, + "loss": 0.1698, + "step": 4005 + }, + { + "epoch": 0.21, + "learning_rate": 3.6728861119698846e-07, + "loss": 0.173, + "step": 4006 + }, + { + "epoch": 0.21, + "learning_rate": 3.6727007830092227e-07, + "loss": 0.1325, + "step": 4007 + }, + { + "epoch": 0.21, + "learning_rate": 3.672515406242127e-07, + "loss": 0.1606, + "step": 4008 + }, + { + "epoch": 0.21, + "learning_rate": 3.6723299816738955e-07, + "loss": 0.1351, + "step": 4009 + }, + { + "epoch": 0.21, + "learning_rate": 3.672144509309827e-07, + "loss": 0.1191, + "step": 4010 + }, + { + "epoch": 0.21, + "learning_rate": 3.6719589891552234e-07, + "loss": 0.1276, + "step": 4011 + }, + { + "epoch": 0.21, + "learning_rate": 3.671773421215386e-07, + "loss": 0.1371, + "step": 4012 + }, + { + "epoch": 0.21, + "learning_rate": 3.671587805495619e-07, + "loss": 0.1638, + "step": 4013 + }, + { + "epoch": 0.21, + "learning_rate": 3.6714021420012273e-07, + "loss": 0.1354, + "step": 4014 + }, + { + "epoch": 0.21, + "learning_rate": 3.6712164307375174e-07, + "loss": 0.1286, + "step": 4015 + }, + { + "epoch": 0.21, + "learning_rate": 3.671030671709796e-07, + "loss": 0.142, + "step": 4016 + }, + { + "epoch": 0.21, + "learning_rate": 3.6708448649233737e-07, + "loss": 0.1352, + "step": 4017 + }, + { + "epoch": 0.21, + "learning_rate": 3.6706590103835595e-07, + "loss": 0.146, + "step": 4018 + }, + { + "epoch": 0.21, + "learning_rate": 3.670473108095667e-07, + "loss": 0.209, + "step": 4019 + }, + { + "epoch": 0.21, + "learning_rate": 3.670287158065007e-07, + "loss": 0.1285, + "step": 4020 + }, + { + "epoch": 0.21, + "learning_rate": 3.6701011602968956e-07, + "loss": 0.1186, + "step": 4021 + }, + { + "epoch": 0.21, + "learning_rate": 3.6699151147966487e-07, + "loss": 0.1521, + "step": 4022 + }, + { + "epoch": 0.21, + "learning_rate": 3.669729021569583e-07, + "loss": 0.1367, + "step": 4023 + }, + { + "epoch": 0.21, + "learning_rate": 3.669542880621017e-07, + "loss": 0.1768, + "step": 4024 + }, + { + "epoch": 0.21, + "learning_rate": 3.6693566919562706e-07, + "loss": 0.1407, + "step": 4025 + }, + { + "epoch": 0.21, + "learning_rate": 3.669170455580666e-07, + "loss": 0.1623, + "step": 4026 + }, + { + "epoch": 0.21, + "learning_rate": 3.6689841714995253e-07, + "loss": 0.2211, + "step": 4027 + }, + { + "epoch": 0.21, + "learning_rate": 3.668797839718173e-07, + "loss": 0.1763, + "step": 4028 + }, + { + "epoch": 0.21, + "learning_rate": 3.668611460241934e-07, + "loss": 0.1476, + "step": 4029 + }, + { + "epoch": 0.21, + "learning_rate": 3.6684250330761356e-07, + "loss": 0.1429, + "step": 4030 + }, + { + "epoch": 0.21, + "learning_rate": 3.668238558226105e-07, + "loss": 0.1604, + "step": 4031 + }, + { + "epoch": 0.21, + "learning_rate": 3.668052035697173e-07, + "loss": 0.1494, + "step": 4032 + }, + { + "epoch": 0.21, + "learning_rate": 3.667865465494669e-07, + "loss": 0.1393, + "step": 4033 + }, + { + "epoch": 0.21, + "learning_rate": 3.6676788476239275e-07, + "loss": 0.1554, + "step": 4034 + }, + { + "epoch": 0.21, + "learning_rate": 3.66749218209028e-07, + "loss": 0.1755, + "step": 4035 + }, + { + "epoch": 0.21, + "learning_rate": 3.6673054688990623e-07, + "loss": 0.1482, + "step": 4036 + }, + { + "epoch": 0.21, + "learning_rate": 3.667118708055611e-07, + "loss": 0.1122, + "step": 4037 + }, + { + "epoch": 0.21, + "learning_rate": 3.666931899565263e-07, + "loss": 0.1397, + "step": 4038 + }, + { + "epoch": 0.21, + "learning_rate": 3.6667450434333577e-07, + "loss": 0.1737, + "step": 4039 + }, + { + "epoch": 0.21, + "learning_rate": 3.666558139665236e-07, + "loss": 0.1527, + "step": 4040 + }, + { + "epoch": 0.21, + "learning_rate": 3.666371188266239e-07, + "loss": 0.1424, + "step": 4041 + }, + { + "epoch": 0.21, + "learning_rate": 3.6661841892417104e-07, + "loss": 0.1496, + "step": 4042 + }, + { + "epoch": 0.21, + "learning_rate": 3.6659971425969944e-07, + "loss": 0.1368, + "step": 4043 + }, + { + "epoch": 0.21, + "learning_rate": 3.6658100483374367e-07, + "loss": 0.1146, + "step": 4044 + }, + { + "epoch": 0.21, + "learning_rate": 3.665622906468385e-07, + "loss": 0.1645, + "step": 4045 + }, + { + "epoch": 0.21, + "learning_rate": 3.6654357169951875e-07, + "loss": 0.1242, + "step": 4046 + }, + { + "epoch": 0.21, + "learning_rate": 3.6652484799231946e-07, + "loss": 0.1755, + "step": 4047 + }, + { + "epoch": 0.21, + "learning_rate": 3.6650611952577566e-07, + "loss": 0.1506, + "step": 4048 + }, + { + "epoch": 0.21, + "learning_rate": 3.664873863004228e-07, + "loss": 0.1302, + "step": 4049 + }, + { + "epoch": 0.21, + "learning_rate": 3.6646864831679603e-07, + "loss": 0.163, + "step": 4050 + }, + { + "epoch": 0.21, + "learning_rate": 3.664499055754311e-07, + "loss": 0.1234, + "step": 4051 + }, + { + "epoch": 0.21, + "learning_rate": 3.6643115807686363e-07, + "loss": 0.1519, + "step": 4052 + }, + { + "epoch": 0.21, + "learning_rate": 3.664124058216294e-07, + "loss": 0.1378, + "step": 4053 + }, + { + "epoch": 0.21, + "learning_rate": 3.663936488102643e-07, + "loss": 0.164, + "step": 4054 + }, + { + "epoch": 0.21, + "learning_rate": 3.6637488704330457e-07, + "loss": 0.1381, + "step": 4055 + }, + { + "epoch": 0.21, + "learning_rate": 3.663561205212863e-07, + "loss": 0.1531, + "step": 4056 + }, + { + "epoch": 0.21, + "learning_rate": 3.6633734924474594e-07, + "loss": 0.1651, + "step": 4057 + }, + { + "epoch": 0.21, + "learning_rate": 3.6631857321421985e-07, + "loss": 0.1583, + "step": 4058 + }, + { + "epoch": 0.21, + "learning_rate": 3.6629979243024486e-07, + "loss": 0.1577, + "step": 4059 + }, + { + "epoch": 0.21, + "learning_rate": 3.6628100689335746e-07, + "loss": 0.1575, + "step": 4060 + }, + { + "epoch": 0.21, + "learning_rate": 3.662622166040948e-07, + "loss": 0.1851, + "step": 4061 + }, + { + "epoch": 0.21, + "learning_rate": 3.662434215629937e-07, + "loss": 0.1471, + "step": 4062 + }, + { + "epoch": 0.21, + "learning_rate": 3.662246217705915e-07, + "loss": 0.1554, + "step": 4063 + }, + { + "epoch": 0.21, + "learning_rate": 3.6620581722742545e-07, + "loss": 0.155, + "step": 4064 + }, + { + "epoch": 0.21, + "learning_rate": 3.6618700793403294e-07, + "loss": 0.1321, + "step": 4065 + }, + { + "epoch": 0.21, + "learning_rate": 3.661681938909516e-07, + "loss": 0.1586, + "step": 4066 + }, + { + "epoch": 0.21, + "learning_rate": 3.6614937509871914e-07, + "loss": 0.1835, + "step": 4067 + }, + { + "epoch": 0.21, + "learning_rate": 3.6613055155787337e-07, + "loss": 0.1363, + "step": 4068 + }, + { + "epoch": 0.21, + "learning_rate": 3.6611172326895233e-07, + "loss": 0.1418, + "step": 4069 + }, + { + "epoch": 0.21, + "learning_rate": 3.660928902324942e-07, + "loss": 0.1554, + "step": 4070 + }, + { + "epoch": 0.21, + "learning_rate": 3.66074052449037e-07, + "loss": 0.1302, + "step": 4071 + }, + { + "epoch": 0.21, + "learning_rate": 3.660552099191194e-07, + "loss": 0.1449, + "step": 4072 + }, + { + "epoch": 0.21, + "learning_rate": 3.660363626432797e-07, + "loss": 0.1201, + "step": 4073 + }, + { + "epoch": 0.21, + "learning_rate": 3.660175106220567e-07, + "loss": 0.1405, + "step": 4074 + }, + { + "epoch": 0.21, + "learning_rate": 3.6599865385598916e-07, + "loss": 0.1294, + "step": 4075 + }, + { + "epoch": 0.21, + "learning_rate": 3.6597979234561595e-07, + "loss": 0.1519, + "step": 4076 + }, + { + "epoch": 0.21, + "learning_rate": 3.659609260914763e-07, + "loss": 0.15, + "step": 4077 + }, + { + "epoch": 0.21, + "learning_rate": 3.659420550941092e-07, + "loss": 0.118, + "step": 4078 + }, + { + "epoch": 0.21, + "learning_rate": 3.659231793540542e-07, + "loss": 0.1363, + "step": 4079 + }, + { + "epoch": 0.21, + "learning_rate": 3.6590429887185063e-07, + "loss": 0.1635, + "step": 4080 + }, + { + "epoch": 0.21, + "learning_rate": 3.658854136480382e-07, + "loss": 0.1328, + "step": 4081 + }, + { + "epoch": 0.21, + "learning_rate": 3.658665236831566e-07, + "loss": 0.145, + "step": 4082 + }, + { + "epoch": 0.21, + "learning_rate": 3.6584762897774573e-07, + "loss": 0.1757, + "step": 4083 + }, + { + "epoch": 0.21, + "learning_rate": 3.658287295323455e-07, + "loss": 0.1427, + "step": 4084 + }, + { + "epoch": 0.21, + "learning_rate": 3.658098253474963e-07, + "loss": 0.1264, + "step": 4085 + }, + { + "epoch": 0.21, + "learning_rate": 3.6579091642373817e-07, + "loss": 0.1694, + "step": 4086 + }, + { + "epoch": 0.21, + "learning_rate": 3.657720027616118e-07, + "loss": 0.1377, + "step": 4087 + }, + { + "epoch": 0.21, + "learning_rate": 3.657530843616574e-07, + "loss": 0.1375, + "step": 4088 + }, + { + "epoch": 0.21, + "learning_rate": 3.6573416122441597e-07, + "loss": 0.168, + "step": 4089 + }, + { + "epoch": 0.21, + "learning_rate": 3.6571523335042826e-07, + "loss": 0.1495, + "step": 4090 + }, + { + "epoch": 0.21, + "learning_rate": 3.6569630074023514e-07, + "loss": 0.169, + "step": 4091 + }, + { + "epoch": 0.21, + "learning_rate": 3.656773633943778e-07, + "loss": 0.1763, + "step": 4092 + }, + { + "epoch": 0.21, + "learning_rate": 3.656584213133975e-07, + "loss": 0.1648, + "step": 4093 + }, + { + "epoch": 0.21, + "learning_rate": 3.656394744978355e-07, + "loss": 0.1479, + "step": 4094 + }, + { + "epoch": 0.21, + "learning_rate": 3.6562052294823343e-07, + "loss": 0.1331, + "step": 4095 + }, + { + "epoch": 0.21, + "learning_rate": 3.6560156666513284e-07, + "loss": 0.1512, + "step": 4096 + }, + { + "epoch": 0.21, + "learning_rate": 3.6558260564907557e-07, + "loss": 0.1492, + "step": 4097 + }, + { + "epoch": 0.21, + "learning_rate": 3.655636399006035e-07, + "loss": 0.1611, + "step": 4098 + }, + { + "epoch": 0.21, + "learning_rate": 3.655446694202587e-07, + "loss": 0.1569, + "step": 4099 + }, + { + "epoch": 0.21, + "learning_rate": 3.655256942085833e-07, + "loss": 0.1435, + "step": 4100 + }, + { + "epoch": 0.21, + "learning_rate": 3.6550671426611974e-07, + "loss": 0.112, + "step": 4101 + }, + { + "epoch": 0.21, + "learning_rate": 3.654877295934103e-07, + "loss": 0.1718, + "step": 4102 + }, + { + "epoch": 0.21, + "learning_rate": 3.654687401909977e-07, + "loss": 0.1317, + "step": 4103 + }, + { + "epoch": 0.21, + "learning_rate": 3.6544974605942467e-07, + "loss": 0.1691, + "step": 4104 + }, + { + "epoch": 0.21, + "learning_rate": 3.65430747199234e-07, + "loss": 0.1578, + "step": 4105 + }, + { + "epoch": 0.21, + "learning_rate": 3.6541174361096874e-07, + "loss": 0.1984, + "step": 4106 + }, + { + "epoch": 0.21, + "learning_rate": 3.653927352951719e-07, + "loss": 0.145, + "step": 4107 + }, + { + "epoch": 0.21, + "learning_rate": 3.653737222523869e-07, + "loss": 0.1745, + "step": 4108 + }, + { + "epoch": 0.21, + "learning_rate": 3.6535470448315706e-07, + "loss": 0.1593, + "step": 4109 + }, + { + "epoch": 0.21, + "learning_rate": 3.6533568198802594e-07, + "loss": 0.1702, + "step": 4110 + }, + { + "epoch": 0.21, + "learning_rate": 3.653166547675372e-07, + "loss": 0.1805, + "step": 4111 + }, + { + "epoch": 0.21, + "learning_rate": 3.652976228222347e-07, + "loss": 0.1402, + "step": 4112 + }, + { + "epoch": 0.21, + "learning_rate": 3.652785861526623e-07, + "loss": 0.1118, + "step": 4113 + }, + { + "epoch": 0.21, + "learning_rate": 3.652595447593641e-07, + "loss": 0.1404, + "step": 4114 + }, + { + "epoch": 0.21, + "learning_rate": 3.652404986428843e-07, + "loss": 0.1462, + "step": 4115 + }, + { + "epoch": 0.21, + "learning_rate": 3.6522144780376724e-07, + "loss": 0.1677, + "step": 4116 + }, + { + "epoch": 0.21, + "learning_rate": 3.652023922425575e-07, + "loss": 0.1498, + "step": 4117 + }, + { + "epoch": 0.21, + "learning_rate": 3.651833319597995e-07, + "loss": 0.1709, + "step": 4118 + }, + { + "epoch": 0.21, + "learning_rate": 3.651642669560382e-07, + "loss": 0.1312, + "step": 4119 + }, + { + "epoch": 0.22, + "learning_rate": 3.651451972318183e-07, + "loss": 0.1313, + "step": 4120 + }, + { + "epoch": 0.22, + "learning_rate": 3.65126122787685e-07, + "loss": 0.1327, + "step": 4121 + }, + { + "epoch": 0.22, + "learning_rate": 3.6510704362418333e-07, + "loss": 0.1783, + "step": 4122 + }, + { + "epoch": 0.22, + "learning_rate": 3.650879597418586e-07, + "loss": 0.1397, + "step": 4123 + }, + { + "epoch": 0.22, + "learning_rate": 3.650688711412563e-07, + "loss": 0.1289, + "step": 4124 + }, + { + "epoch": 0.22, + "learning_rate": 3.650497778229219e-07, + "loss": 0.1686, + "step": 4125 + }, + { + "epoch": 0.22, + "learning_rate": 3.650306797874011e-07, + "loss": 0.152, + "step": 4126 + }, + { + "epoch": 0.22, + "learning_rate": 3.650115770352398e-07, + "loss": 0.1557, + "step": 4127 + }, + { + "epoch": 0.22, + "learning_rate": 3.6499246956698395e-07, + "loss": 0.1859, + "step": 4128 + }, + { + "epoch": 0.22, + "learning_rate": 3.649733573831796e-07, + "loss": 0.1246, + "step": 4129 + }, + { + "epoch": 0.22, + "learning_rate": 3.64954240484373e-07, + "loss": 0.187, + "step": 4130 + }, + { + "epoch": 0.22, + "learning_rate": 3.6493511887111056e-07, + "loss": 0.1541, + "step": 4131 + }, + { + "epoch": 0.22, + "learning_rate": 3.6491599254393876e-07, + "loss": 0.1867, + "step": 4132 + }, + { + "epoch": 0.22, + "learning_rate": 3.648968615034042e-07, + "loss": 0.1331, + "step": 4133 + }, + { + "epoch": 0.22, + "learning_rate": 3.648777257500537e-07, + "loss": 0.1147, + "step": 4134 + }, + { + "epoch": 0.22, + "learning_rate": 3.6485858528443414e-07, + "loss": 0.1235, + "step": 4135 + }, + { + "epoch": 0.22, + "learning_rate": 3.6483944010709256e-07, + "loss": 0.1717, + "step": 4136 + }, + { + "epoch": 0.22, + "learning_rate": 3.6482029021857616e-07, + "loss": 0.1425, + "step": 4137 + }, + { + "epoch": 0.22, + "learning_rate": 3.648011356194322e-07, + "loss": 0.1878, + "step": 4138 + }, + { + "epoch": 0.22, + "learning_rate": 3.6478197631020817e-07, + "loss": 0.1693, + "step": 4139 + }, + { + "epoch": 0.22, + "learning_rate": 3.6476281229145173e-07, + "loss": 0.1139, + "step": 4140 + }, + { + "epoch": 0.22, + "learning_rate": 3.6474364356371043e-07, + "loss": 0.1955, + "step": 4141 + }, + { + "epoch": 0.22, + "learning_rate": 3.647244701275322e-07, + "loss": 0.1741, + "step": 4142 + }, + { + "epoch": 0.22, + "learning_rate": 3.647052919834651e-07, + "loss": 0.1567, + "step": 4143 + }, + { + "epoch": 0.22, + "learning_rate": 3.646861091320571e-07, + "loss": 0.139, + "step": 4144 + }, + { + "epoch": 0.22, + "learning_rate": 3.6466692157385656e-07, + "loss": 0.1297, + "step": 4145 + }, + { + "epoch": 0.22, + "learning_rate": 3.646477293094118e-07, + "loss": 0.1627, + "step": 4146 + }, + { + "epoch": 0.22, + "learning_rate": 3.646285323392714e-07, + "loss": 0.1559, + "step": 4147 + }, + { + "epoch": 0.22, + "learning_rate": 3.64609330663984e-07, + "loss": 0.1638, + "step": 4148 + }, + { + "epoch": 0.22, + "learning_rate": 3.6459012428409833e-07, + "loss": 0.1493, + "step": 4149 + }, + { + "epoch": 0.22, + "learning_rate": 3.6457091320016343e-07, + "loss": 0.1423, + "step": 4150 + }, + { + "epoch": 0.22, + "learning_rate": 3.6455169741272827e-07, + "loss": 0.1325, + "step": 4151 + }, + { + "epoch": 0.22, + "learning_rate": 3.6453247692234215e-07, + "loss": 0.143, + "step": 4152 + }, + { + "epoch": 0.22, + "learning_rate": 3.645132517295542e-07, + "loss": 0.1676, + "step": 4153 + }, + { + "epoch": 0.22, + "learning_rate": 3.6449402183491405e-07, + "loss": 0.1386, + "step": 4154 + }, + { + "epoch": 0.22, + "learning_rate": 3.6447478723897127e-07, + "loss": 0.1516, + "step": 4155 + }, + { + "epoch": 0.22, + "learning_rate": 3.644555479422756e-07, + "loss": 0.1784, + "step": 4156 + }, + { + "epoch": 0.22, + "learning_rate": 3.6443630394537684e-07, + "loss": 0.181, + "step": 4157 + }, + { + "epoch": 0.22, + "learning_rate": 3.64417055248825e-07, + "loss": 0.1273, + "step": 4158 + }, + { + "epoch": 0.22, + "learning_rate": 3.6439780185317036e-07, + "loss": 0.1787, + "step": 4159 + }, + { + "epoch": 0.22, + "learning_rate": 3.6437854375896306e-07, + "loss": 0.128, + "step": 4160 + }, + { + "epoch": 0.22, + "learning_rate": 3.643592809667535e-07, + "loss": 0.1561, + "step": 4161 + }, + { + "epoch": 0.22, + "learning_rate": 3.643400134770922e-07, + "loss": 0.1648, + "step": 4162 + }, + { + "epoch": 0.22, + "learning_rate": 3.643207412905299e-07, + "loss": 0.1504, + "step": 4163 + }, + { + "epoch": 0.22, + "learning_rate": 3.6430146440761736e-07, + "loss": 0.1262, + "step": 4164 + }, + { + "epoch": 0.22, + "learning_rate": 3.6428218282890555e-07, + "loss": 0.1461, + "step": 4165 + }, + { + "epoch": 0.22, + "learning_rate": 3.642628965549456e-07, + "loss": 0.1287, + "step": 4166 + }, + { + "epoch": 0.22, + "learning_rate": 3.6424360558628856e-07, + "loss": 0.1728, + "step": 4167 + }, + { + "epoch": 0.22, + "learning_rate": 3.642243099234859e-07, + "loss": 0.1652, + "step": 4168 + }, + { + "epoch": 0.22, + "learning_rate": 3.642050095670891e-07, + "loss": 0.2245, + "step": 4169 + }, + { + "epoch": 0.22, + "learning_rate": 3.641857045176497e-07, + "loss": 0.1597, + "step": 4170 + }, + { + "epoch": 0.22, + "learning_rate": 3.641663947757195e-07, + "loss": 0.1419, + "step": 4171 + }, + { + "epoch": 0.22, + "learning_rate": 3.6414708034185036e-07, + "loss": 0.1412, + "step": 4172 + }, + { + "epoch": 0.22, + "learning_rate": 3.641277612165943e-07, + "loss": 0.1496, + "step": 4173 + }, + { + "epoch": 0.22, + "learning_rate": 3.641084374005035e-07, + "loss": 0.1701, + "step": 4174 + }, + { + "epoch": 0.22, + "learning_rate": 3.640891088941301e-07, + "loss": 0.1428, + "step": 4175 + }, + { + "epoch": 0.22, + "learning_rate": 3.640697756980267e-07, + "loss": 0.1677, + "step": 4176 + }, + { + "epoch": 0.22, + "learning_rate": 3.6405043781274577e-07, + "loss": 0.1271, + "step": 4177 + }, + { + "epoch": 0.22, + "learning_rate": 3.6403109523884e-07, + "loss": 0.1747, + "step": 4178 + }, + { + "epoch": 0.22, + "learning_rate": 3.640117479768622e-07, + "loss": 0.1095, + "step": 4179 + }, + { + "epoch": 0.22, + "learning_rate": 3.639923960273653e-07, + "loss": 0.1264, + "step": 4180 + }, + { + "epoch": 0.22, + "learning_rate": 3.6397303939090243e-07, + "loss": 0.1639, + "step": 4181 + }, + { + "epoch": 0.22, + "learning_rate": 3.6395367806802683e-07, + "loss": 0.163, + "step": 4182 + }, + { + "epoch": 0.22, + "learning_rate": 3.639343120592918e-07, + "loss": 0.1436, + "step": 4183 + }, + { + "epoch": 0.22, + "learning_rate": 3.6391494136525084e-07, + "loss": 0.2036, + "step": 4184 + }, + { + "epoch": 0.22, + "learning_rate": 3.6389556598645756e-07, + "loss": 0.1632, + "step": 4185 + }, + { + "epoch": 0.22, + "learning_rate": 3.6387618592346574e-07, + "loss": 0.1193, + "step": 4186 + }, + { + "epoch": 0.22, + "learning_rate": 3.6385680117682926e-07, + "loss": 0.1531, + "step": 4187 + }, + { + "epoch": 0.22, + "learning_rate": 3.638374117471022e-07, + "loss": 0.1639, + "step": 4188 + }, + { + "epoch": 0.22, + "learning_rate": 3.638180176348386e-07, + "loss": 0.1152, + "step": 4189 + }, + { + "epoch": 0.22, + "learning_rate": 3.6379861884059283e-07, + "loss": 0.1756, + "step": 4190 + }, + { + "epoch": 0.22, + "learning_rate": 3.6377921536491927e-07, + "loss": 0.1236, + "step": 4191 + }, + { + "epoch": 0.22, + "learning_rate": 3.637598072083726e-07, + "loss": 0.142, + "step": 4192 + }, + { + "epoch": 0.22, + "learning_rate": 3.6374039437150733e-07, + "loss": 0.1194, + "step": 4193 + }, + { + "epoch": 0.22, + "learning_rate": 3.6372097685487845e-07, + "loss": 0.1337, + "step": 4194 + }, + { + "epoch": 0.22, + "learning_rate": 3.6370155465904085e-07, + "loss": 0.1652, + "step": 4195 + }, + { + "epoch": 0.22, + "learning_rate": 3.636821277845496e-07, + "loss": 0.1284, + "step": 4196 + }, + { + "epoch": 0.22, + "learning_rate": 3.636626962319599e-07, + "loss": 0.2113, + "step": 4197 + }, + { + "epoch": 0.22, + "learning_rate": 3.636432600018272e-07, + "loss": 0.1443, + "step": 4198 + }, + { + "epoch": 0.22, + "learning_rate": 3.63623819094707e-07, + "loss": 0.1407, + "step": 4199 + }, + { + "epoch": 0.22, + "learning_rate": 3.636043735111548e-07, + "loss": 0.1488, + "step": 4200 + }, + { + "epoch": 0.22, + "learning_rate": 3.6358492325172654e-07, + "loss": 0.1355, + "step": 4201 + }, + { + "epoch": 0.22, + "learning_rate": 3.6356546831697793e-07, + "loss": 0.1579, + "step": 4202 + }, + { + "epoch": 0.22, + "learning_rate": 3.635460087074651e-07, + "loss": 0.1519, + "step": 4203 + }, + { + "epoch": 0.22, + "learning_rate": 3.635265444237443e-07, + "loss": 0.1465, + "step": 4204 + }, + { + "epoch": 0.22, + "learning_rate": 3.635070754663717e-07, + "loss": 0.0915, + "step": 4205 + }, + { + "epoch": 0.22, + "learning_rate": 3.634876018359037e-07, + "loss": 0.1335, + "step": 4206 + }, + { + "epoch": 0.22, + "learning_rate": 3.63468123532897e-07, + "loss": 0.1392, + "step": 4207 + }, + { + "epoch": 0.22, + "learning_rate": 3.634486405579082e-07, + "loss": 0.1184, + "step": 4208 + }, + { + "epoch": 0.22, + "learning_rate": 3.6342915291149413e-07, + "loss": 0.138, + "step": 4209 + }, + { + "epoch": 0.22, + "learning_rate": 3.634096605942118e-07, + "loss": 0.1651, + "step": 4210 + }, + { + "epoch": 0.22, + "learning_rate": 3.633901636066183e-07, + "loss": 0.1505, + "step": 4211 + }, + { + "epoch": 0.22, + "learning_rate": 3.633706619492708e-07, + "loss": 0.1636, + "step": 4212 + }, + { + "epoch": 0.22, + "learning_rate": 3.633511556227268e-07, + "loss": 0.1248, + "step": 4213 + }, + { + "epoch": 0.22, + "learning_rate": 3.633316446275437e-07, + "loss": 0.1318, + "step": 4214 + }, + { + "epoch": 0.22, + "learning_rate": 3.6331212896427907e-07, + "loss": 0.1742, + "step": 4215 + }, + { + "epoch": 0.22, + "learning_rate": 3.6329260863349076e-07, + "loss": 0.1771, + "step": 4216 + }, + { + "epoch": 0.22, + "learning_rate": 3.6327308363573667e-07, + "loss": 0.1371, + "step": 4217 + }, + { + "epoch": 0.22, + "learning_rate": 3.6325355397157485e-07, + "loss": 0.1443, + "step": 4218 + }, + { + "epoch": 0.22, + "learning_rate": 3.6323401964156337e-07, + "loss": 0.1703, + "step": 4219 + }, + { + "epoch": 0.22, + "learning_rate": 3.632144806462607e-07, + "loss": 0.1331, + "step": 4220 + }, + { + "epoch": 0.22, + "learning_rate": 3.631949369862251e-07, + "loss": 0.1438, + "step": 4221 + }, + { + "epoch": 0.22, + "learning_rate": 3.631753886620152e-07, + "loss": 0.1625, + "step": 4222 + }, + { + "epoch": 0.22, + "learning_rate": 3.6315583567418974e-07, + "loss": 0.1023, + "step": 4223 + }, + { + "epoch": 0.22, + "learning_rate": 3.6313627802330744e-07, + "loss": 0.12, + "step": 4224 + }, + { + "epoch": 0.22, + "learning_rate": 3.631167157099274e-07, + "loss": 0.1333, + "step": 4225 + }, + { + "epoch": 0.22, + "learning_rate": 3.6309714873460864e-07, + "loss": 0.1532, + "step": 4226 + }, + { + "epoch": 0.22, + "learning_rate": 3.6307757709791044e-07, + "loss": 0.1161, + "step": 4227 + }, + { + "epoch": 0.22, + "learning_rate": 3.6305800080039205e-07, + "loss": 0.1996, + "step": 4228 + }, + { + "epoch": 0.22, + "learning_rate": 3.6303841984261306e-07, + "loss": 0.1456, + "step": 4229 + }, + { + "epoch": 0.22, + "learning_rate": 3.6301883422513313e-07, + "loss": 0.1119, + "step": 4230 + }, + { + "epoch": 0.22, + "learning_rate": 3.6299924394851196e-07, + "loss": 0.1241, + "step": 4231 + }, + { + "epoch": 0.22, + "learning_rate": 3.6297964901330957e-07, + "loss": 0.1615, + "step": 4232 + }, + { + "epoch": 0.22, + "learning_rate": 3.6296004942008576e-07, + "loss": 0.1695, + "step": 4233 + }, + { + "epoch": 0.22, + "learning_rate": 3.6294044516940094e-07, + "loss": 0.1688, + "step": 4234 + }, + { + "epoch": 0.22, + "learning_rate": 3.629208362618152e-07, + "loss": 0.1628, + "step": 4235 + }, + { + "epoch": 0.22, + "learning_rate": 3.629012226978891e-07, + "loss": 0.1742, + "step": 4236 + }, + { + "epoch": 0.22, + "learning_rate": 3.628816044781832e-07, + "loss": 0.1294, + "step": 4237 + }, + { + "epoch": 0.22, + "learning_rate": 3.628619816032582e-07, + "loss": 0.1381, + "step": 4238 + }, + { + "epoch": 0.22, + "learning_rate": 3.628423540736748e-07, + "loss": 0.1176, + "step": 4239 + }, + { + "epoch": 0.22, + "learning_rate": 3.628227218899941e-07, + "loss": 0.147, + "step": 4240 + }, + { + "epoch": 0.22, + "learning_rate": 3.6280308505277715e-07, + "loss": 0.1243, + "step": 4241 + }, + { + "epoch": 0.22, + "learning_rate": 3.6278344356258517e-07, + "loss": 0.1518, + "step": 4242 + }, + { + "epoch": 0.22, + "learning_rate": 3.627637974199796e-07, + "loss": 0.1554, + "step": 4243 + }, + { + "epoch": 0.22, + "learning_rate": 3.6274414662552174e-07, + "loss": 0.1625, + "step": 4244 + }, + { + "epoch": 0.22, + "learning_rate": 3.6272449117977347e-07, + "loss": 0.139, + "step": 4245 + }, + { + "epoch": 0.22, + "learning_rate": 3.627048310832964e-07, + "loss": 0.151, + "step": 4246 + }, + { + "epoch": 0.22, + "learning_rate": 3.6268516633665236e-07, + "loss": 0.2001, + "step": 4247 + }, + { + "epoch": 0.22, + "learning_rate": 3.6266549694040353e-07, + "loss": 0.1318, + "step": 4248 + }, + { + "epoch": 0.22, + "learning_rate": 3.62645822895112e-07, + "loss": 0.1722, + "step": 4249 + }, + { + "epoch": 0.22, + "learning_rate": 3.6262614420134005e-07, + "loss": 0.159, + "step": 4250 + }, + { + "epoch": 0.22, + "learning_rate": 3.6260646085965015e-07, + "loss": 0.1285, + "step": 4251 + }, + { + "epoch": 0.22, + "learning_rate": 3.6258677287060477e-07, + "loss": 0.1619, + "step": 4252 + }, + { + "epoch": 0.22, + "learning_rate": 3.6256708023476666e-07, + "loss": 0.1399, + "step": 4253 + }, + { + "epoch": 0.22, + "learning_rate": 3.6254738295269867e-07, + "loss": 0.1232, + "step": 4254 + }, + { + "epoch": 0.22, + "learning_rate": 3.6252768102496376e-07, + "loss": 0.1084, + "step": 4255 + }, + { + "epoch": 0.22, + "learning_rate": 3.6250797445212495e-07, + "loss": 0.1524, + "step": 4256 + }, + { + "epoch": 0.22, + "learning_rate": 3.6248826323474545e-07, + "loss": 0.1517, + "step": 4257 + }, + { + "epoch": 0.22, + "learning_rate": 3.624685473733887e-07, + "loss": 0.1065, + "step": 4258 + }, + { + "epoch": 0.22, + "learning_rate": 3.624488268686182e-07, + "loss": 0.1134, + "step": 4259 + }, + { + "epoch": 0.22, + "learning_rate": 3.6242910172099743e-07, + "loss": 0.1269, + "step": 4260 + }, + { + "epoch": 0.22, + "learning_rate": 3.6240937193109024e-07, + "loss": 0.1689, + "step": 4261 + }, + { + "epoch": 0.22, + "learning_rate": 3.6238963749946056e-07, + "loss": 0.1693, + "step": 4262 + }, + { + "epoch": 0.22, + "learning_rate": 3.6236989842667234e-07, + "loss": 0.1381, + "step": 4263 + }, + { + "epoch": 0.22, + "learning_rate": 3.623501547132897e-07, + "loss": 0.1546, + "step": 4264 + }, + { + "epoch": 0.22, + "learning_rate": 3.6233040635987696e-07, + "loss": 0.1182, + "step": 4265 + }, + { + "epoch": 0.22, + "learning_rate": 3.6231065336699863e-07, + "loss": 0.1373, + "step": 4266 + }, + { + "epoch": 0.22, + "learning_rate": 3.622908957352191e-07, + "loss": 0.1133, + "step": 4267 + }, + { + "epoch": 0.22, + "learning_rate": 3.622711334651031e-07, + "loss": 0.1342, + "step": 4268 + }, + { + "epoch": 0.22, + "learning_rate": 3.622513665572155e-07, + "loss": 0.1657, + "step": 4269 + }, + { + "epoch": 0.22, + "learning_rate": 3.622315950121212e-07, + "loss": 0.1557, + "step": 4270 + }, + { + "epoch": 0.22, + "learning_rate": 3.622118188303853e-07, + "loss": 0.1282, + "step": 4271 + }, + { + "epoch": 0.22, + "learning_rate": 3.621920380125729e-07, + "loss": 0.1382, + "step": 4272 + }, + { + "epoch": 0.22, + "learning_rate": 3.6217225255924957e-07, + "loss": 0.1803, + "step": 4273 + }, + { + "epoch": 0.22, + "learning_rate": 3.6215246247098063e-07, + "loss": 0.1631, + "step": 4274 + }, + { + "epoch": 0.22, + "learning_rate": 3.6213266774833165e-07, + "loss": 0.1291, + "step": 4275 + }, + { + "epoch": 0.22, + "learning_rate": 3.621128683918685e-07, + "loss": 0.1826, + "step": 4276 + }, + { + "epoch": 0.22, + "learning_rate": 3.62093064402157e-07, + "loss": 0.1216, + "step": 4277 + }, + { + "epoch": 0.22, + "learning_rate": 3.620732557797631e-07, + "loss": 0.1433, + "step": 4278 + }, + { + "epoch": 0.22, + "learning_rate": 3.62053442525253e-07, + "loss": 0.1393, + "step": 4279 + }, + { + "epoch": 0.22, + "learning_rate": 3.620336246391929e-07, + "loss": 0.1906, + "step": 4280 + }, + { + "epoch": 0.22, + "learning_rate": 3.620138021221494e-07, + "loss": 0.1397, + "step": 4281 + }, + { + "epoch": 0.22, + "learning_rate": 3.6199397497468874e-07, + "loss": 0.1336, + "step": 4282 + }, + { + "epoch": 0.22, + "learning_rate": 3.619741431973778e-07, + "loss": 0.1317, + "step": 4283 + }, + { + "epoch": 0.22, + "learning_rate": 3.6195430679078333e-07, + "loss": 0.1628, + "step": 4284 + }, + { + "epoch": 0.22, + "learning_rate": 3.6193446575547224e-07, + "loss": 0.1069, + "step": 4285 + }, + { + "epoch": 0.22, + "learning_rate": 3.6191462009201157e-07, + "loss": 0.1668, + "step": 4286 + }, + { + "epoch": 0.22, + "learning_rate": 3.618947698009686e-07, + "loss": 0.1516, + "step": 4287 + }, + { + "epoch": 0.22, + "learning_rate": 3.6187491488291065e-07, + "loss": 0.1229, + "step": 4288 + }, + { + "epoch": 0.22, + "learning_rate": 3.61855055338405e-07, + "loss": 0.1573, + "step": 4289 + }, + { + "epoch": 0.22, + "learning_rate": 3.618351911680195e-07, + "loss": 0.1264, + "step": 4290 + }, + { + "epoch": 0.22, + "learning_rate": 3.618153223723218e-07, + "loss": 0.139, + "step": 4291 + }, + { + "epoch": 0.22, + "learning_rate": 3.617954489518796e-07, + "loss": 0.1567, + "step": 4292 + }, + { + "epoch": 0.22, + "learning_rate": 3.617755709072611e-07, + "loss": 0.1409, + "step": 4293 + }, + { + "epoch": 0.22, + "learning_rate": 3.6175568823903435e-07, + "loss": 0.1186, + "step": 4294 + }, + { + "epoch": 0.22, + "learning_rate": 3.6173580094776747e-07, + "loss": 0.0928, + "step": 4295 + }, + { + "epoch": 0.22, + "learning_rate": 3.6171590903402904e-07, + "loss": 0.155, + "step": 4296 + }, + { + "epoch": 0.22, + "learning_rate": 3.616960124983875e-07, + "loss": 0.1414, + "step": 4297 + }, + { + "epoch": 0.22, + "learning_rate": 3.6167611134141147e-07, + "loss": 0.1493, + "step": 4298 + }, + { + "epoch": 0.22, + "learning_rate": 3.6165620556366985e-07, + "loss": 0.105, + "step": 4299 + }, + { + "epoch": 0.22, + "learning_rate": 3.616362951657314e-07, + "loss": 0.1467, + "step": 4300 + }, + { + "epoch": 0.22, + "learning_rate": 3.616163801481652e-07, + "loss": 0.1632, + "step": 4301 + }, + { + "epoch": 0.22, + "learning_rate": 3.6159646051154054e-07, + "loss": 0.1393, + "step": 4302 + }, + { + "epoch": 0.22, + "learning_rate": 3.615765362564266e-07, + "loss": 0.1812, + "step": 4303 + }, + { + "epoch": 0.22, + "learning_rate": 3.6155660738339294e-07, + "loss": 0.1156, + "step": 4304 + }, + { + "epoch": 0.22, + "learning_rate": 3.61536673893009e-07, + "loss": 0.15, + "step": 4305 + }, + { + "epoch": 0.22, + "learning_rate": 3.6151673578584465e-07, + "loss": 0.15, + "step": 4306 + }, + { + "epoch": 0.22, + "learning_rate": 3.6149679306246954e-07, + "loss": 0.1229, + "step": 4307 + }, + { + "epoch": 0.22, + "learning_rate": 3.6147684572345377e-07, + "loss": 0.1338, + "step": 4308 + }, + { + "epoch": 0.22, + "learning_rate": 3.6145689376936745e-07, + "loss": 0.163, + "step": 4309 + }, + { + "epoch": 0.22, + "learning_rate": 3.6143693720078074e-07, + "loss": 0.1416, + "step": 4310 + }, + { + "epoch": 0.23, + "learning_rate": 3.6141697601826405e-07, + "loss": 0.1559, + "step": 4311 + }, + { + "epoch": 0.23, + "learning_rate": 3.613970102223879e-07, + "loss": 0.1396, + "step": 4312 + }, + { + "epoch": 0.23, + "learning_rate": 3.613770398137229e-07, + "loss": 0.1716, + "step": 4313 + }, + { + "epoch": 0.23, + "learning_rate": 3.613570647928398e-07, + "loss": 0.1386, + "step": 4314 + }, + { + "epoch": 0.23, + "learning_rate": 3.6133708516030945e-07, + "loss": 0.1692, + "step": 4315 + }, + { + "epoch": 0.23, + "learning_rate": 3.6131710091670297e-07, + "loss": 0.1224, + "step": 4316 + }, + { + "epoch": 0.23, + "learning_rate": 3.6129711206259144e-07, + "loss": 0.1152, + "step": 4317 + }, + { + "epoch": 0.23, + "learning_rate": 3.612771185985462e-07, + "loss": 0.1563, + "step": 4318 + }, + { + "epoch": 0.23, + "learning_rate": 3.6125712052513864e-07, + "loss": 0.1363, + "step": 4319 + }, + { + "epoch": 0.23, + "learning_rate": 3.612371178429403e-07, + "loss": 0.1248, + "step": 4320 + }, + { + "epoch": 0.23, + "learning_rate": 3.6121711055252293e-07, + "loss": 0.1704, + "step": 4321 + }, + { + "epoch": 0.23, + "learning_rate": 3.611970986544583e-07, + "loss": 0.171, + "step": 4322 + }, + { + "epoch": 0.23, + "learning_rate": 3.6117708214931837e-07, + "loss": 0.1279, + "step": 4323 + }, + { + "epoch": 0.23, + "learning_rate": 3.611570610376752e-07, + "loss": 0.1456, + "step": 4324 + }, + { + "epoch": 0.23, + "learning_rate": 3.61137035320101e-07, + "loss": 0.1462, + "step": 4325 + }, + { + "epoch": 0.23, + "learning_rate": 3.6111700499716814e-07, + "loss": 0.1906, + "step": 4326 + }, + { + "epoch": 0.23, + "learning_rate": 3.6109697006944907e-07, + "loss": 0.147, + "step": 4327 + }, + { + "epoch": 0.23, + "learning_rate": 3.610769305375164e-07, + "loss": 0.1504, + "step": 4328 + }, + { + "epoch": 0.23, + "learning_rate": 3.610568864019429e-07, + "loss": 0.1423, + "step": 4329 + }, + { + "epoch": 0.23, + "learning_rate": 3.6103683766330136e-07, + "loss": 0.1711, + "step": 4330 + }, + { + "epoch": 0.23, + "learning_rate": 3.6101678432216487e-07, + "loss": 0.1348, + "step": 4331 + }, + { + "epoch": 0.23, + "learning_rate": 3.6099672637910653e-07, + "loss": 0.1575, + "step": 4332 + }, + { + "epoch": 0.23, + "learning_rate": 3.609766638346996e-07, + "loss": 0.13, + "step": 4333 + }, + { + "epoch": 0.23, + "learning_rate": 3.609565966895175e-07, + "loss": 0.1614, + "step": 4334 + }, + { + "epoch": 0.23, + "learning_rate": 3.6093652494413367e-07, + "loss": 0.1554, + "step": 4335 + }, + { + "epoch": 0.23, + "learning_rate": 3.6091644859912187e-07, + "loss": 0.1321, + "step": 4336 + }, + { + "epoch": 0.23, + "learning_rate": 3.608963676550558e-07, + "loss": 0.1265, + "step": 4337 + }, + { + "epoch": 0.23, + "learning_rate": 3.6087628211250953e-07, + "loss": 0.1568, + "step": 4338 + }, + { + "epoch": 0.23, + "learning_rate": 3.6085619197205693e-07, + "loss": 0.1128, + "step": 4339 + }, + { + "epoch": 0.23, + "learning_rate": 3.608360972342723e-07, + "loss": 0.1434, + "step": 4340 + }, + { + "epoch": 0.23, + "learning_rate": 3.608159978997299e-07, + "loss": 0.132, + "step": 4341 + }, + { + "epoch": 0.23, + "learning_rate": 3.607958939690042e-07, + "loss": 0.1544, + "step": 4342 + }, + { + "epoch": 0.23, + "learning_rate": 3.6077578544266973e-07, + "loss": 0.1391, + "step": 4343 + }, + { + "epoch": 0.23, + "learning_rate": 3.6075567232130133e-07, + "loss": 0.1268, + "step": 4344 + }, + { + "epoch": 0.23, + "learning_rate": 3.607355546054737e-07, + "loss": 0.1355, + "step": 4345 + }, + { + "epoch": 0.23, + "learning_rate": 3.607154322957619e-07, + "loss": 0.1271, + "step": 4346 + }, + { + "epoch": 0.23, + "learning_rate": 3.606953053927411e-07, + "loss": 0.1644, + "step": 4347 + }, + { + "epoch": 0.23, + "learning_rate": 3.606751738969863e-07, + "loss": 0.1527, + "step": 4348 + }, + { + "epoch": 0.23, + "learning_rate": 3.6065503780907306e-07, + "loss": 0.1336, + "step": 4349 + }, + { + "epoch": 0.23, + "learning_rate": 3.6063489712957677e-07, + "loss": 0.1322, + "step": 4350 + }, + { + "epoch": 0.23, + "learning_rate": 3.606147518590731e-07, + "loss": 0.1682, + "step": 4351 + }, + { + "epoch": 0.23, + "learning_rate": 3.6059460199813784e-07, + "loss": 0.1456, + "step": 4352 + }, + { + "epoch": 0.23, + "learning_rate": 3.605744475473469e-07, + "loss": 0.1871, + "step": 4353 + }, + { + "epoch": 0.23, + "learning_rate": 3.6055428850727625e-07, + "loss": 0.1431, + "step": 4354 + }, + { + "epoch": 0.23, + "learning_rate": 3.6053412487850206e-07, + "loss": 0.1305, + "step": 4355 + }, + { + "epoch": 0.23, + "learning_rate": 3.605139566616006e-07, + "loss": 0.1519, + "step": 4356 + }, + { + "epoch": 0.23, + "learning_rate": 3.604937838571483e-07, + "loss": 0.1334, + "step": 4357 + }, + { + "epoch": 0.23, + "learning_rate": 3.6047360646572164e-07, + "loss": 0.1553, + "step": 4358 + }, + { + "epoch": 0.23, + "learning_rate": 3.604534244878974e-07, + "loss": 0.1589, + "step": 4359 + }, + { + "epoch": 0.23, + "learning_rate": 3.604332379242524e-07, + "loss": 0.1435, + "step": 4360 + }, + { + "epoch": 0.23, + "learning_rate": 3.6041304677536345e-07, + "loss": 0.1583, + "step": 4361 + }, + { + "epoch": 0.23, + "learning_rate": 3.603928510418077e-07, + "loss": 0.1549, + "step": 4362 + }, + { + "epoch": 0.23, + "learning_rate": 3.6037265072416237e-07, + "loss": 0.1402, + "step": 4363 + }, + { + "epoch": 0.23, + "learning_rate": 3.603524458230048e-07, + "loss": 0.1197, + "step": 4364 + }, + { + "epoch": 0.23, + "learning_rate": 3.603322363389124e-07, + "loss": 0.1263, + "step": 4365 + }, + { + "epoch": 0.23, + "learning_rate": 3.6031202227246276e-07, + "loss": 0.1371, + "step": 4366 + }, + { + "epoch": 0.23, + "learning_rate": 3.6029180362423367e-07, + "loss": 0.1971, + "step": 4367 + }, + { + "epoch": 0.23, + "learning_rate": 3.6027158039480295e-07, + "loss": 0.1326, + "step": 4368 + }, + { + "epoch": 0.23, + "learning_rate": 3.6025135258474866e-07, + "loss": 0.1584, + "step": 4369 + }, + { + "epoch": 0.23, + "learning_rate": 3.6023112019464873e-07, + "loss": 0.1399, + "step": 4370 + }, + { + "epoch": 0.23, + "learning_rate": 3.602108832250816e-07, + "loss": 0.1546, + "step": 4371 + }, + { + "epoch": 0.23, + "learning_rate": 3.601906416766256e-07, + "loss": 0.1414, + "step": 4372 + }, + { + "epoch": 0.23, + "learning_rate": 3.6017039554985915e-07, + "loss": 0.1294, + "step": 4373 + }, + { + "epoch": 0.23, + "learning_rate": 3.60150144845361e-07, + "loss": 0.1463, + "step": 4374 + }, + { + "epoch": 0.23, + "learning_rate": 3.6012988956370985e-07, + "loss": 0.157, + "step": 4375 + }, + { + "epoch": 0.23, + "learning_rate": 3.6010962970548473e-07, + "loss": 0.1385, + "step": 4376 + }, + { + "epoch": 0.23, + "learning_rate": 3.600893652712645e-07, + "loss": 0.1253, + "step": 4377 + }, + { + "epoch": 0.23, + "learning_rate": 3.6006909626162844e-07, + "loss": 0.1413, + "step": 4378 + }, + { + "epoch": 0.23, + "learning_rate": 3.600488226771558e-07, + "loss": 0.1602, + "step": 4379 + }, + { + "epoch": 0.23, + "learning_rate": 3.6002854451842603e-07, + "loss": 0.1083, + "step": 4380 + }, + { + "epoch": 0.23, + "learning_rate": 3.600082617860187e-07, + "loss": 0.1335, + "step": 4381 + }, + { + "epoch": 0.23, + "learning_rate": 3.599879744805134e-07, + "loss": 0.1731, + "step": 4382 + }, + { + "epoch": 0.23, + "learning_rate": 3.5996768260249016e-07, + "loss": 0.1425, + "step": 4383 + }, + { + "epoch": 0.23, + "learning_rate": 3.599473861525287e-07, + "loss": 0.1287, + "step": 4384 + }, + { + "epoch": 0.23, + "learning_rate": 3.5992708513120925e-07, + "loss": 0.1437, + "step": 4385 + }, + { + "epoch": 0.23, + "learning_rate": 3.5990677953911197e-07, + "loss": 0.1505, + "step": 4386 + }, + { + "epoch": 0.23, + "learning_rate": 3.5988646937681714e-07, + "loss": 0.1493, + "step": 4387 + }, + { + "epoch": 0.23, + "learning_rate": 3.598661546449054e-07, + "loss": 0.1392, + "step": 4388 + }, + { + "epoch": 0.23, + "learning_rate": 3.5984583534395717e-07, + "loss": 0.1471, + "step": 4389 + }, + { + "epoch": 0.23, + "learning_rate": 3.598255114745532e-07, + "loss": 0.1329, + "step": 4390 + }, + { + "epoch": 0.23, + "learning_rate": 3.598051830372745e-07, + "loss": 0.1358, + "step": 4391 + }, + { + "epoch": 0.23, + "learning_rate": 3.5978485003270196e-07, + "loss": 0.1497, + "step": 4392 + }, + { + "epoch": 0.23, + "learning_rate": 3.5976451246141675e-07, + "loss": 0.1515, + "step": 4393 + }, + { + "epoch": 0.23, + "learning_rate": 3.5974417032400006e-07, + "loss": 0.1714, + "step": 4394 + }, + { + "epoch": 0.23, + "learning_rate": 3.5972382362103327e-07, + "loss": 0.1347, + "step": 4395 + }, + { + "epoch": 0.23, + "learning_rate": 3.5970347235309804e-07, + "loss": 0.1682, + "step": 4396 + }, + { + "epoch": 0.23, + "learning_rate": 3.596831165207759e-07, + "loss": 0.157, + "step": 4397 + }, + { + "epoch": 0.23, + "learning_rate": 3.596627561246486e-07, + "loss": 0.1139, + "step": 4398 + }, + { + "epoch": 0.23, + "learning_rate": 3.596423911652981e-07, + "loss": 0.1182, + "step": 4399 + }, + { + "epoch": 0.23, + "learning_rate": 3.596220216433065e-07, + "loss": 0.1878, + "step": 4400 + }, + { + "epoch": 0.23, + "learning_rate": 3.596016475592558e-07, + "loss": 0.1202, + "step": 4401 + }, + { + "epoch": 0.23, + "learning_rate": 3.595812689137284e-07, + "loss": 0.1546, + "step": 4402 + }, + { + "epoch": 0.23, + "learning_rate": 3.595608857073068e-07, + "loss": 0.1295, + "step": 4403 + }, + { + "epoch": 0.23, + "learning_rate": 3.5954049794057355e-07, + "loss": 0.1548, + "step": 4404 + }, + { + "epoch": 0.23, + "learning_rate": 3.5952010561411116e-07, + "loss": 0.142, + "step": 4405 + }, + { + "epoch": 0.23, + "learning_rate": 3.5949970872850255e-07, + "loss": 0.1564, + "step": 4406 + }, + { + "epoch": 0.23, + "learning_rate": 3.5947930728433076e-07, + "loss": 0.1347, + "step": 4407 + }, + { + "epoch": 0.23, + "learning_rate": 3.594589012821788e-07, + "loss": 0.1938, + "step": 4408 + }, + { + "epoch": 0.23, + "learning_rate": 3.5943849072262993e-07, + "loss": 0.1452, + "step": 4409 + }, + { + "epoch": 0.23, + "learning_rate": 3.5941807560626736e-07, + "loss": 0.1468, + "step": 4410 + }, + { + "epoch": 0.23, + "learning_rate": 3.5939765593367464e-07, + "loss": 0.1597, + "step": 4411 + }, + { + "epoch": 0.23, + "learning_rate": 3.5937723170543545e-07, + "loss": 0.1512, + "step": 4412 + }, + { + "epoch": 0.23, + "learning_rate": 3.593568029221334e-07, + "loss": 0.1192, + "step": 4413 + }, + { + "epoch": 0.23, + "learning_rate": 3.5933636958435246e-07, + "loss": 0.1239, + "step": 4414 + }, + { + "epoch": 0.23, + "learning_rate": 3.593159316926765e-07, + "loss": 0.1613, + "step": 4415 + }, + { + "epoch": 0.23, + "learning_rate": 3.5929548924768973e-07, + "loss": 0.1371, + "step": 4416 + }, + { + "epoch": 0.23, + "learning_rate": 3.5927504224997646e-07, + "loss": 0.1535, + "step": 4417 + }, + { + "epoch": 0.23, + "learning_rate": 3.5925459070012087e-07, + "loss": 0.0996, + "step": 4418 + }, + { + "epoch": 0.23, + "learning_rate": 3.592341345987077e-07, + "loss": 0.1251, + "step": 4419 + }, + { + "epoch": 0.23, + "learning_rate": 3.592136739463214e-07, + "loss": 0.1638, + "step": 4420 + }, + { + "epoch": 0.23, + "learning_rate": 3.5919320874354693e-07, + "loss": 0.1162, + "step": 4421 + }, + { + "epoch": 0.23, + "learning_rate": 3.591727389909691e-07, + "loss": 0.1459, + "step": 4422 + }, + { + "epoch": 0.23, + "learning_rate": 3.591522646891729e-07, + "loss": 0.1902, + "step": 4423 + }, + { + "epoch": 0.23, + "learning_rate": 3.591317858387435e-07, + "loss": 0.1513, + "step": 4424 + }, + { + "epoch": 0.23, + "learning_rate": 3.591113024402663e-07, + "loss": 0.1726, + "step": 4425 + }, + { + "epoch": 0.23, + "learning_rate": 3.590908144943266e-07, + "loss": 0.1291, + "step": 4426 + }, + { + "epoch": 0.23, + "learning_rate": 3.5907032200151e-07, + "loss": 0.142, + "step": 4427 + }, + { + "epoch": 0.23, + "learning_rate": 3.5904982496240225e-07, + "loss": 0.1614, + "step": 4428 + }, + { + "epoch": 0.23, + "learning_rate": 3.5902932337758906e-07, + "loss": 0.1244, + "step": 4429 + }, + { + "epoch": 0.23, + "learning_rate": 3.590088172476564e-07, + "loss": 0.1299, + "step": 4430 + }, + { + "epoch": 0.23, + "learning_rate": 3.589883065731904e-07, + "loss": 0.1255, + "step": 4431 + }, + { + "epoch": 0.23, + "learning_rate": 3.5896779135477723e-07, + "loss": 0.1391, + "step": 4432 + }, + { + "epoch": 0.23, + "learning_rate": 3.589472715930032e-07, + "loss": 0.1323, + "step": 4433 + }, + { + "epoch": 0.23, + "learning_rate": 3.589267472884548e-07, + "loss": 0.1428, + "step": 4434 + }, + { + "epoch": 0.23, + "learning_rate": 3.5890621844171854e-07, + "loss": 0.1446, + "step": 4435 + }, + { + "epoch": 0.23, + "learning_rate": 3.5888568505338126e-07, + "loss": 0.1575, + "step": 4436 + }, + { + "epoch": 0.23, + "learning_rate": 3.5886514712402977e-07, + "loss": 0.178, + "step": 4437 + }, + { + "epoch": 0.23, + "learning_rate": 3.5884460465425103e-07, + "loss": 0.1216, + "step": 4438 + }, + { + "epoch": 0.23, + "learning_rate": 3.5882405764463223e-07, + "loss": 0.1432, + "step": 4439 + }, + { + "epoch": 0.23, + "learning_rate": 3.588035060957605e-07, + "loss": 0.1694, + "step": 4440 + }, + { + "epoch": 0.23, + "learning_rate": 3.5878295000822326e-07, + "loss": 0.1422, + "step": 4441 + }, + { + "epoch": 0.23, + "learning_rate": 3.58762389382608e-07, + "loss": 0.1625, + "step": 4442 + }, + { + "epoch": 0.23, + "learning_rate": 3.5874182421950244e-07, + "loss": 0.1293, + "step": 4443 + }, + { + "epoch": 0.23, + "learning_rate": 3.5872125451949417e-07, + "loss": 0.1295, + "step": 4444 + }, + { + "epoch": 0.23, + "learning_rate": 3.587006802831712e-07, + "loss": 0.1177, + "step": 4445 + }, + { + "epoch": 0.23, + "learning_rate": 3.5868010151112157e-07, + "loss": 0.1751, + "step": 4446 + }, + { + "epoch": 0.23, + "learning_rate": 3.586595182039333e-07, + "loss": 0.1325, + "step": 4447 + }, + { + "epoch": 0.23, + "learning_rate": 3.5863893036219486e-07, + "loss": 0.1572, + "step": 4448 + }, + { + "epoch": 0.23, + "learning_rate": 3.5861833798649445e-07, + "loss": 0.1496, + "step": 4449 + }, + { + "epoch": 0.23, + "learning_rate": 3.585977410774207e-07, + "loss": 0.125, + "step": 4450 + }, + { + "epoch": 0.23, + "learning_rate": 3.585771396355624e-07, + "loss": 0.1313, + "step": 4451 + }, + { + "epoch": 0.23, + "learning_rate": 3.5855653366150814e-07, + "loss": 0.1524, + "step": 4452 + }, + { + "epoch": 0.23, + "learning_rate": 3.5853592315584693e-07, + "loss": 0.1471, + "step": 4453 + }, + { + "epoch": 0.23, + "learning_rate": 3.585153081191679e-07, + "loss": 0.1395, + "step": 4454 + }, + { + "epoch": 0.23, + "learning_rate": 3.584946885520601e-07, + "loss": 0.1592, + "step": 4455 + }, + { + "epoch": 0.23, + "learning_rate": 3.5847406445511295e-07, + "loss": 0.1484, + "step": 4456 + }, + { + "epoch": 0.23, + "learning_rate": 3.5845343582891587e-07, + "loss": 0.1238, + "step": 4457 + }, + { + "epoch": 0.23, + "learning_rate": 3.5843280267405846e-07, + "loss": 0.1423, + "step": 4458 + }, + { + "epoch": 0.23, + "learning_rate": 3.584121649911303e-07, + "loss": 0.1274, + "step": 4459 + }, + { + "epoch": 0.23, + "learning_rate": 3.583915227807214e-07, + "loss": 0.1743, + "step": 4460 + }, + { + "epoch": 0.23, + "learning_rate": 3.5837087604342156e-07, + "loss": 0.1357, + "step": 4461 + }, + { + "epoch": 0.23, + "learning_rate": 3.5835022477982093e-07, + "loss": 0.1666, + "step": 4462 + }, + { + "epoch": 0.23, + "learning_rate": 3.583295689905098e-07, + "loss": 0.1256, + "step": 4463 + }, + { + "epoch": 0.23, + "learning_rate": 3.583089086760785e-07, + "loss": 0.1465, + "step": 4464 + }, + { + "epoch": 0.23, + "learning_rate": 3.5828824383711747e-07, + "loss": 0.1716, + "step": 4465 + }, + { + "epoch": 0.23, + "learning_rate": 3.5826757447421725e-07, + "loss": 0.1406, + "step": 4466 + }, + { + "epoch": 0.23, + "learning_rate": 3.582469005879687e-07, + "loss": 0.1206, + "step": 4467 + }, + { + "epoch": 0.23, + "learning_rate": 3.5822622217896266e-07, + "loss": 0.1157, + "step": 4468 + }, + { + "epoch": 0.23, + "learning_rate": 3.5820553924779006e-07, + "loss": 0.1543, + "step": 4469 + }, + { + "epoch": 0.23, + "learning_rate": 3.581848517950421e-07, + "loss": 0.1857, + "step": 4470 + }, + { + "epoch": 0.23, + "learning_rate": 3.5816415982131006e-07, + "loss": 0.1348, + "step": 4471 + }, + { + "epoch": 0.23, + "learning_rate": 3.581434633271852e-07, + "loss": 0.1242, + "step": 4472 + }, + { + "epoch": 0.23, + "learning_rate": 3.5812276231325914e-07, + "loss": 0.1686, + "step": 4473 + }, + { + "epoch": 0.23, + "learning_rate": 3.581020567801235e-07, + "loss": 0.1589, + "step": 4474 + }, + { + "epoch": 0.23, + "learning_rate": 3.5808134672837e-07, + "loss": 0.1175, + "step": 4475 + }, + { + "epoch": 0.23, + "learning_rate": 3.580606321585906e-07, + "loss": 0.138, + "step": 4476 + }, + { + "epoch": 0.23, + "learning_rate": 3.580399130713774e-07, + "loss": 0.1494, + "step": 4477 + }, + { + "epoch": 0.23, + "learning_rate": 3.5801918946732237e-07, + "loss": 0.1509, + "step": 4478 + }, + { + "epoch": 0.23, + "learning_rate": 3.5799846134701794e-07, + "loss": 0.1478, + "step": 4479 + }, + { + "epoch": 0.23, + "learning_rate": 3.5797772871105647e-07, + "loss": 0.1266, + "step": 4480 + }, + { + "epoch": 0.23, + "learning_rate": 3.5795699156003055e-07, + "loss": 0.1307, + "step": 4481 + }, + { + "epoch": 0.23, + "learning_rate": 3.579362498945328e-07, + "loss": 0.13, + "step": 4482 + }, + { + "epoch": 0.23, + "learning_rate": 3.579155037151561e-07, + "loss": 0.1659, + "step": 4483 + }, + { + "epoch": 0.23, + "learning_rate": 3.5789475302249335e-07, + "loss": 0.1725, + "step": 4484 + }, + { + "epoch": 0.23, + "learning_rate": 3.5787399781713756e-07, + "loss": 0.1354, + "step": 4485 + }, + { + "epoch": 0.23, + "learning_rate": 3.57853238099682e-07, + "loss": 0.1328, + "step": 4486 + }, + { + "epoch": 0.23, + "learning_rate": 3.5783247387071995e-07, + "loss": 0.121, + "step": 4487 + }, + { + "epoch": 0.23, + "learning_rate": 3.5781170513084493e-07, + "loss": 0.1514, + "step": 4488 + }, + { + "epoch": 0.23, + "learning_rate": 3.577909318806504e-07, + "loss": 0.1506, + "step": 4489 + }, + { + "epoch": 0.23, + "learning_rate": 3.577701541207301e-07, + "loss": 0.1173, + "step": 4490 + }, + { + "epoch": 0.23, + "learning_rate": 3.577493718516779e-07, + "loss": 0.1199, + "step": 4491 + }, + { + "epoch": 0.23, + "learning_rate": 3.5772858507408777e-07, + "loss": 0.183, + "step": 4492 + }, + { + "epoch": 0.23, + "learning_rate": 3.5770779378855377e-07, + "loss": 0.1581, + "step": 4493 + }, + { + "epoch": 0.23, + "learning_rate": 3.5768699799567016e-07, + "loss": 0.1234, + "step": 4494 + }, + { + "epoch": 0.23, + "learning_rate": 3.5766619769603124e-07, + "loss": 0.1207, + "step": 4495 + }, + { + "epoch": 0.23, + "learning_rate": 3.5764539289023156e-07, + "loss": 0.1219, + "step": 4496 + }, + { + "epoch": 0.23, + "learning_rate": 3.576245835788657e-07, + "loss": 0.1604, + "step": 4497 + }, + { + "epoch": 0.23, + "learning_rate": 3.576037697625284e-07, + "loss": 0.1501, + "step": 4498 + }, + { + "epoch": 0.23, + "learning_rate": 3.575829514418145e-07, + "loss": 0.1461, + "step": 4499 + }, + { + "epoch": 0.23, + "learning_rate": 3.5756212861731896e-07, + "loss": 0.143, + "step": 4500 + }, + { + "epoch": 0.23, + "learning_rate": 3.57541301289637e-07, + "loss": 0.1346, + "step": 4501 + }, + { + "epoch": 0.23, + "learning_rate": 3.5752046945936386e-07, + "loss": 0.18, + "step": 4502 + }, + { + "epoch": 0.24, + "learning_rate": 3.574996331270949e-07, + "loss": 0.1382, + "step": 4503 + }, + { + "epoch": 0.24, + "learning_rate": 3.574787922934256e-07, + "loss": 0.1828, + "step": 4504 + }, + { + "epoch": 0.24, + "learning_rate": 3.5745794695895166e-07, + "loss": 0.1717, + "step": 4505 + }, + { + "epoch": 0.24, + "learning_rate": 3.5743709712426874e-07, + "loss": 0.1374, + "step": 4506 + }, + { + "epoch": 0.24, + "learning_rate": 3.5741624278997284e-07, + "loss": 0.1284, + "step": 4507 + }, + { + "epoch": 0.24, + "learning_rate": 3.5739538395665997e-07, + "loss": 0.1706, + "step": 4508 + }, + { + "epoch": 0.24, + "learning_rate": 3.5737452062492626e-07, + "loss": 0.1425, + "step": 4509 + }, + { + "epoch": 0.24, + "learning_rate": 3.57353652795368e-07, + "loss": 0.1523, + "step": 4510 + }, + { + "epoch": 0.24, + "learning_rate": 3.5733278046858155e-07, + "loss": 0.1575, + "step": 4511 + }, + { + "epoch": 0.24, + "learning_rate": 3.5731190364516354e-07, + "loss": 0.1639, + "step": 4512 + }, + { + "epoch": 0.24, + "learning_rate": 3.5729102232571063e-07, + "loss": 0.1354, + "step": 4513 + }, + { + "epoch": 0.24, + "learning_rate": 3.5727013651081956e-07, + "loss": 0.1384, + "step": 4514 + }, + { + "epoch": 0.24, + "learning_rate": 3.5724924620108723e-07, + "loss": 0.123, + "step": 4515 + }, + { + "epoch": 0.24, + "learning_rate": 3.572283513971108e-07, + "loss": 0.1417, + "step": 4516 + }, + { + "epoch": 0.24, + "learning_rate": 3.5720745209948733e-07, + "loss": 0.1403, + "step": 4517 + }, + { + "epoch": 0.24, + "learning_rate": 3.571865483088142e-07, + "loss": 0.1353, + "step": 4518 + }, + { + "epoch": 0.24, + "learning_rate": 3.571656400256889e-07, + "loss": 0.1235, + "step": 4519 + }, + { + "epoch": 0.24, + "learning_rate": 3.5714472725070886e-07, + "loss": 0.1421, + "step": 4520 + }, + { + "epoch": 0.24, + "learning_rate": 3.571238099844719e-07, + "loss": 0.1728, + "step": 4521 + }, + { + "epoch": 0.24, + "learning_rate": 3.571028882275758e-07, + "loss": 0.1745, + "step": 4522 + }, + { + "epoch": 0.24, + "learning_rate": 3.5708196198061845e-07, + "loss": 0.1203, + "step": 4523 + }, + { + "epoch": 0.24, + "learning_rate": 3.570610312441981e-07, + "loss": 0.1283, + "step": 4524 + }, + { + "epoch": 0.24, + "learning_rate": 3.570400960189127e-07, + "loss": 0.1792, + "step": 4525 + }, + { + "epoch": 0.24, + "learning_rate": 3.5701915630536084e-07, + "loss": 0.1187, + "step": 4526 + }, + { + "epoch": 0.24, + "learning_rate": 3.5699821210414085e-07, + "loss": 0.1213, + "step": 4527 + }, + { + "epoch": 0.24, + "learning_rate": 3.569772634158513e-07, + "loss": 0.1718, + "step": 4528 + }, + { + "epoch": 0.24, + "learning_rate": 3.5695631024109104e-07, + "loss": 0.1315, + "step": 4529 + }, + { + "epoch": 0.24, + "learning_rate": 3.5693535258045875e-07, + "loss": 0.151, + "step": 4530 + }, + { + "epoch": 0.24, + "learning_rate": 3.5691439043455355e-07, + "loss": 0.1235, + "step": 4531 + }, + { + "epoch": 0.24, + "learning_rate": 3.5689342380397455e-07, + "loss": 0.1289, + "step": 4532 + }, + { + "epoch": 0.24, + "learning_rate": 3.5687245268932086e-07, + "loss": 0.1954, + "step": 4533 + }, + { + "epoch": 0.24, + "learning_rate": 3.568514770911919e-07, + "loss": 0.1441, + "step": 4534 + }, + { + "epoch": 0.24, + "learning_rate": 3.5683049701018727e-07, + "loss": 0.1715, + "step": 4535 + }, + { + "epoch": 0.24, + "learning_rate": 3.5680951244690645e-07, + "loss": 0.1357, + "step": 4536 + }, + { + "epoch": 0.24, + "learning_rate": 3.567885234019492e-07, + "loss": 0.1673, + "step": 4537 + }, + { + "epoch": 0.24, + "learning_rate": 3.567675298759154e-07, + "loss": 0.1379, + "step": 4538 + }, + { + "epoch": 0.24, + "learning_rate": 3.5674653186940513e-07, + "loss": 0.1511, + "step": 4539 + }, + { + "epoch": 0.24, + "learning_rate": 3.567255293830184e-07, + "loss": 0.1576, + "step": 4540 + }, + { + "epoch": 0.24, + "learning_rate": 3.5670452241735564e-07, + "loss": 0.1588, + "step": 4541 + }, + { + "epoch": 0.24, + "learning_rate": 3.566835109730171e-07, + "loss": 0.1833, + "step": 4542 + }, + { + "epoch": 0.24, + "learning_rate": 3.566624950506033e-07, + "loss": 0.1573, + "step": 4543 + }, + { + "epoch": 0.24, + "learning_rate": 3.566414746507149e-07, + "loss": 0.1512, + "step": 4544 + }, + { + "epoch": 0.24, + "learning_rate": 3.5662044977395276e-07, + "loss": 0.1291, + "step": 4545 + }, + { + "epoch": 0.24, + "learning_rate": 3.5659942042091763e-07, + "loss": 0.1154, + "step": 4546 + }, + { + "epoch": 0.24, + "learning_rate": 3.5657838659221066e-07, + "loss": 0.1315, + "step": 4547 + }, + { + "epoch": 0.24, + "learning_rate": 3.5655734828843296e-07, + "loss": 0.1392, + "step": 4548 + }, + { + "epoch": 0.24, + "learning_rate": 3.5653630551018576e-07, + "loss": 0.1704, + "step": 4549 + }, + { + "epoch": 0.24, + "learning_rate": 3.565152582580706e-07, + "loss": 0.12, + "step": 4550 + }, + { + "epoch": 0.24, + "learning_rate": 3.564942065326888e-07, + "loss": 0.1394, + "step": 4551 + }, + { + "epoch": 0.24, + "learning_rate": 3.564731503346423e-07, + "loss": 0.1331, + "step": 4552 + }, + { + "epoch": 0.24, + "learning_rate": 3.5645208966453266e-07, + "loss": 0.1401, + "step": 4553 + }, + { + "epoch": 0.24, + "learning_rate": 3.5643102452296193e-07, + "loss": 0.1294, + "step": 4554 + }, + { + "epoch": 0.24, + "learning_rate": 3.5640995491053214e-07, + "loss": 0.1171, + "step": 4555 + }, + { + "epoch": 0.24, + "learning_rate": 3.563888808278455e-07, + "loss": 0.1186, + "step": 4556 + }, + { + "epoch": 0.24, + "learning_rate": 3.5636780227550427e-07, + "loss": 0.1383, + "step": 4557 + }, + { + "epoch": 0.24, + "learning_rate": 3.5634671925411083e-07, + "loss": 0.1236, + "step": 4558 + }, + { + "epoch": 0.24, + "learning_rate": 3.563256317642678e-07, + "loss": 0.1223, + "step": 4559 + }, + { + "epoch": 0.24, + "learning_rate": 3.563045398065779e-07, + "loss": 0.1685, + "step": 4560 + }, + { + "epoch": 0.24, + "learning_rate": 3.5628344338164394e-07, + "loss": 0.1544, + "step": 4561 + }, + { + "epoch": 0.24, + "learning_rate": 3.562623424900688e-07, + "loss": 0.1362, + "step": 4562 + }, + { + "epoch": 0.24, + "learning_rate": 3.562412371324556e-07, + "loss": 0.1399, + "step": 4563 + }, + { + "epoch": 0.24, + "learning_rate": 3.562201273094075e-07, + "loss": 0.1325, + "step": 4564 + }, + { + "epoch": 0.24, + "learning_rate": 3.5619901302152786e-07, + "loss": 0.1552, + "step": 4565 + }, + { + "epoch": 0.24, + "learning_rate": 3.561778942694201e-07, + "loss": 0.1675, + "step": 4566 + }, + { + "epoch": 0.24, + "learning_rate": 3.5615677105368793e-07, + "loss": 0.1361, + "step": 4567 + }, + { + "epoch": 0.24, + "learning_rate": 3.561356433749349e-07, + "loss": 0.1531, + "step": 4568 + }, + { + "epoch": 0.24, + "learning_rate": 3.561145112337649e-07, + "loss": 0.1245, + "step": 4569 + }, + { + "epoch": 0.24, + "learning_rate": 3.560933746307819e-07, + "loss": 0.175, + "step": 4570 + }, + { + "epoch": 0.24, + "learning_rate": 3.5607223356659006e-07, + "loss": 0.1233, + "step": 4571 + }, + { + "epoch": 0.24, + "learning_rate": 3.5605108804179344e-07, + "loss": 0.1207, + "step": 4572 + }, + { + "epoch": 0.24, + "learning_rate": 3.560299380569965e-07, + "loss": 0.1396, + "step": 4573 + }, + { + "epoch": 0.24, + "learning_rate": 3.560087836128037e-07, + "loss": 0.1458, + "step": 4574 + }, + { + "epoch": 0.24, + "learning_rate": 3.559876247098197e-07, + "loss": 0.129, + "step": 4575 + }, + { + "epoch": 0.24, + "learning_rate": 3.5596646134864914e-07, + "loss": 0.1176, + "step": 4576 + }, + { + "epoch": 0.24, + "learning_rate": 3.5594529352989686e-07, + "loss": 0.1489, + "step": 4577 + }, + { + "epoch": 0.24, + "learning_rate": 3.5592412125416795e-07, + "loss": 0.1416, + "step": 4578 + }, + { + "epoch": 0.24, + "learning_rate": 3.559029445220674e-07, + "loss": 0.1664, + "step": 4579 + }, + { + "epoch": 0.24, + "learning_rate": 3.558817633342006e-07, + "loss": 0.1222, + "step": 4580 + }, + { + "epoch": 0.24, + "learning_rate": 3.558605776911728e-07, + "loss": 0.1483, + "step": 4581 + }, + { + "epoch": 0.24, + "learning_rate": 3.558393875935894e-07, + "loss": 0.1544, + "step": 4582 + }, + { + "epoch": 0.24, + "learning_rate": 3.5581819304205627e-07, + "loss": 0.1376, + "step": 4583 + }, + { + "epoch": 0.24, + "learning_rate": 3.55796994037179e-07, + "loss": 0.163, + "step": 4584 + }, + { + "epoch": 0.24, + "learning_rate": 3.557757905795635e-07, + "loss": 0.1598, + "step": 4585 + }, + { + "epoch": 0.24, + "learning_rate": 3.5575458266981575e-07, + "loss": 0.1515, + "step": 4586 + }, + { + "epoch": 0.24, + "learning_rate": 3.557333703085419e-07, + "loss": 0.136, + "step": 4587 + }, + { + "epoch": 0.24, + "learning_rate": 3.5571215349634826e-07, + "loss": 0.1265, + "step": 4588 + }, + { + "epoch": 0.24, + "learning_rate": 3.556909322338411e-07, + "loss": 0.1526, + "step": 4589 + }, + { + "epoch": 0.24, + "learning_rate": 3.5566970652162703e-07, + "loss": 0.1588, + "step": 4590 + }, + { + "epoch": 0.24, + "learning_rate": 3.5564847636031264e-07, + "loss": 0.1412, + "step": 4591 + }, + { + "epoch": 0.24, + "learning_rate": 3.556272417505047e-07, + "loss": 0.122, + "step": 4592 + }, + { + "epoch": 0.24, + "learning_rate": 3.5560600269281016e-07, + "loss": 0.144, + "step": 4593 + }, + { + "epoch": 0.24, + "learning_rate": 3.555847591878359e-07, + "loss": 0.1191, + "step": 4594 + }, + { + "epoch": 0.24, + "learning_rate": 3.555635112361893e-07, + "loss": 0.1796, + "step": 4595 + }, + { + "epoch": 0.24, + "learning_rate": 3.555422588384774e-07, + "loss": 0.1118, + "step": 4596 + }, + { + "epoch": 0.24, + "learning_rate": 3.555210019953077e-07, + "loss": 0.2043, + "step": 4597 + }, + { + "epoch": 0.24, + "learning_rate": 3.554997407072877e-07, + "loss": 0.141, + "step": 4598 + }, + { + "epoch": 0.24, + "learning_rate": 3.554784749750251e-07, + "loss": 0.1476, + "step": 4599 + }, + { + "epoch": 0.24, + "learning_rate": 3.554572047991277e-07, + "loss": 0.1298, + "step": 4600 + }, + { + "epoch": 0.24, + "learning_rate": 3.5543593018020337e-07, + "loss": 0.1463, + "step": 4601 + }, + { + "epoch": 0.24, + "learning_rate": 3.5541465111886007e-07, + "loss": 0.126, + "step": 4602 + }, + { + "epoch": 0.24, + "learning_rate": 3.553933676157062e-07, + "loss": 0.1526, + "step": 4603 + }, + { + "epoch": 0.24, + "learning_rate": 3.553720796713498e-07, + "loss": 0.1021, + "step": 4604 + }, + { + "epoch": 0.24, + "learning_rate": 3.553507872863994e-07, + "loss": 0.1578, + "step": 4605 + }, + { + "epoch": 0.24, + "learning_rate": 3.553294904614635e-07, + "loss": 0.1607, + "step": 4606 + }, + { + "epoch": 0.24, + "learning_rate": 3.553081891971508e-07, + "loss": 0.1164, + "step": 4607 + }, + { + "epoch": 0.24, + "learning_rate": 3.5528688349407016e-07, + "loss": 0.1477, + "step": 4608 + }, + { + "epoch": 0.24, + "learning_rate": 3.552655733528303e-07, + "loss": 0.1537, + "step": 4609 + }, + { + "epoch": 0.24, + "learning_rate": 3.5524425877404056e-07, + "loss": 0.1417, + "step": 4610 + }, + { + "epoch": 0.24, + "learning_rate": 3.5522293975830995e-07, + "loss": 0.1236, + "step": 4611 + }, + { + "epoch": 0.24, + "learning_rate": 3.5520161630624777e-07, + "loss": 0.1282, + "step": 4612 + }, + { + "epoch": 0.24, + "learning_rate": 3.551802884184635e-07, + "loss": 0.1162, + "step": 4613 + }, + { + "epoch": 0.24, + "learning_rate": 3.5515895609556663e-07, + "loss": 0.147, + "step": 4614 + }, + { + "epoch": 0.24, + "learning_rate": 3.5513761933816693e-07, + "loss": 0.1352, + "step": 4615 + }, + { + "epoch": 0.24, + "learning_rate": 3.551162781468742e-07, + "loss": 0.1557, + "step": 4616 + }, + { + "epoch": 0.24, + "learning_rate": 3.5509493252229834e-07, + "loss": 0.1267, + "step": 4617 + }, + { + "epoch": 0.24, + "learning_rate": 3.550735824650495e-07, + "loss": 0.1374, + "step": 4618 + }, + { + "epoch": 0.24, + "learning_rate": 3.550522279757377e-07, + "loss": 0.178, + "step": 4619 + }, + { + "epoch": 0.24, + "learning_rate": 3.5503086905497344e-07, + "loss": 0.1529, + "step": 4620 + }, + { + "epoch": 0.24, + "learning_rate": 3.5500950570336706e-07, + "loss": 0.1152, + "step": 4621 + }, + { + "epoch": 0.24, + "learning_rate": 3.5498813792152917e-07, + "loss": 0.1445, + "step": 4622 + }, + { + "epoch": 0.24, + "learning_rate": 3.549667657100705e-07, + "loss": 0.1477, + "step": 4623 + }, + { + "epoch": 0.24, + "learning_rate": 3.5494538906960185e-07, + "loss": 0.1995, + "step": 4624 + }, + { + "epoch": 0.24, + "learning_rate": 3.549240080007341e-07, + "loss": 0.1712, + "step": 4625 + }, + { + "epoch": 0.24, + "learning_rate": 3.5490262250407843e-07, + "loss": 0.1439, + "step": 4626 + }, + { + "epoch": 0.24, + "learning_rate": 3.5488123258024603e-07, + "loss": 0.1483, + "step": 4627 + }, + { + "epoch": 0.24, + "learning_rate": 3.548598382298482e-07, + "loss": 0.1221, + "step": 4628 + }, + { + "epoch": 0.24, + "learning_rate": 3.5483843945349643e-07, + "loss": 0.1106, + "step": 4629 + }, + { + "epoch": 0.24, + "learning_rate": 3.5481703625180224e-07, + "loss": 0.1273, + "step": 4630 + }, + { + "epoch": 0.24, + "learning_rate": 3.5479562862537744e-07, + "loss": 0.1902, + "step": 4631 + }, + { + "epoch": 0.24, + "learning_rate": 3.5477421657483376e-07, + "loss": 0.1544, + "step": 4632 + }, + { + "epoch": 0.24, + "learning_rate": 3.547528001007833e-07, + "loss": 0.1499, + "step": 4633 + }, + { + "epoch": 0.24, + "learning_rate": 3.54731379203838e-07, + "loss": 0.1462, + "step": 4634 + }, + { + "epoch": 0.24, + "learning_rate": 3.5470995388461015e-07, + "loss": 0.1876, + "step": 4635 + }, + { + "epoch": 0.24, + "learning_rate": 3.546885241437121e-07, + "loss": 0.1602, + "step": 4636 + }, + { + "epoch": 0.24, + "learning_rate": 3.546670899817563e-07, + "loss": 0.1704, + "step": 4637 + }, + { + "epoch": 0.24, + "learning_rate": 3.546456513993554e-07, + "loss": 0.1416, + "step": 4638 + }, + { + "epoch": 0.24, + "learning_rate": 3.5462420839712213e-07, + "loss": 0.1585, + "step": 4639 + }, + { + "epoch": 0.24, + "learning_rate": 3.546027609756692e-07, + "loss": 0.2172, + "step": 4640 + }, + { + "epoch": 0.24, + "learning_rate": 3.5458130913560974e-07, + "loss": 0.1033, + "step": 4641 + }, + { + "epoch": 0.24, + "learning_rate": 3.5455985287755674e-07, + "loss": 0.1137, + "step": 4642 + }, + { + "epoch": 0.24, + "learning_rate": 3.5453839220212354e-07, + "loss": 0.1536, + "step": 4643 + }, + { + "epoch": 0.24, + "learning_rate": 3.545169271099234e-07, + "loss": 0.175, + "step": 4644 + }, + { + "epoch": 0.24, + "learning_rate": 3.5449545760156983e-07, + "loss": 0.144, + "step": 4645 + }, + { + "epoch": 0.24, + "learning_rate": 3.5447398367767646e-07, + "loss": 0.1452, + "step": 4646 + }, + { + "epoch": 0.24, + "learning_rate": 3.54452505338857e-07, + "loss": 0.1552, + "step": 4647 + }, + { + "epoch": 0.24, + "learning_rate": 3.544310225857253e-07, + "loss": 0.1023, + "step": 4648 + }, + { + "epoch": 0.24, + "learning_rate": 3.544095354188953e-07, + "loss": 0.1874, + "step": 4649 + }, + { + "epoch": 0.24, + "learning_rate": 3.543880438389813e-07, + "loss": 0.1313, + "step": 4650 + }, + { + "epoch": 0.24, + "learning_rate": 3.5436654784659734e-07, + "loss": 0.1517, + "step": 4651 + }, + { + "epoch": 0.24, + "learning_rate": 3.543450474423578e-07, + "loss": 0.1354, + "step": 4652 + }, + { + "epoch": 0.24, + "learning_rate": 3.543235426268773e-07, + "loss": 0.1575, + "step": 4653 + }, + { + "epoch": 0.24, + "learning_rate": 3.5430203340077036e-07, + "loss": 0.1533, + "step": 4654 + }, + { + "epoch": 0.24, + "learning_rate": 3.542805197646517e-07, + "loss": 0.1746, + "step": 4655 + }, + { + "epoch": 0.24, + "learning_rate": 3.5425900171913624e-07, + "loss": 0.193, + "step": 4656 + }, + { + "epoch": 0.24, + "learning_rate": 3.54237479264839e-07, + "loss": 0.1291, + "step": 4657 + }, + { + "epoch": 0.24, + "learning_rate": 3.54215952402375e-07, + "loss": 0.1132, + "step": 4658 + }, + { + "epoch": 0.24, + "learning_rate": 3.5419442113235956e-07, + "loss": 0.1158, + "step": 4659 + }, + { + "epoch": 0.24, + "learning_rate": 3.5417288545540804e-07, + "loss": 0.1368, + "step": 4660 + }, + { + "epoch": 0.24, + "learning_rate": 3.54151345372136e-07, + "loss": 0.125, + "step": 4661 + }, + { + "epoch": 0.24, + "learning_rate": 3.5412980088315893e-07, + "loss": 0.1309, + "step": 4662 + }, + { + "epoch": 0.24, + "learning_rate": 3.541082519890927e-07, + "loss": 0.1443, + "step": 4663 + }, + { + "epoch": 0.24, + "learning_rate": 3.540866986905531e-07, + "loss": 0.156, + "step": 4664 + }, + { + "epoch": 0.24, + "learning_rate": 3.540651409881562e-07, + "loss": 0.1058, + "step": 4665 + }, + { + "epoch": 0.24, + "learning_rate": 3.5404357888251804e-07, + "loss": 0.145, + "step": 4666 + }, + { + "epoch": 0.24, + "learning_rate": 3.54022012374255e-07, + "loss": 0.1625, + "step": 4667 + }, + { + "epoch": 0.24, + "learning_rate": 3.540004414639833e-07, + "loss": 0.1446, + "step": 4668 + }, + { + "epoch": 0.24, + "learning_rate": 3.5397886615231957e-07, + "loss": 0.1459, + "step": 4669 + }, + { + "epoch": 0.24, + "learning_rate": 3.5395728643988044e-07, + "loss": 0.1622, + "step": 4670 + }, + { + "epoch": 0.24, + "learning_rate": 3.539357023272826e-07, + "loss": 0.1795, + "step": 4671 + }, + { + "epoch": 0.24, + "learning_rate": 3.53914113815143e-07, + "loss": 0.1461, + "step": 4672 + }, + { + "epoch": 0.24, + "learning_rate": 3.538925209040786e-07, + "loss": 0.1286, + "step": 4673 + }, + { + "epoch": 0.24, + "learning_rate": 3.5387092359470654e-07, + "loss": 0.1225, + "step": 4674 + }, + { + "epoch": 0.24, + "learning_rate": 3.5384932188764403e-07, + "loss": 0.1954, + "step": 4675 + }, + { + "epoch": 0.24, + "learning_rate": 3.5382771578350856e-07, + "loss": 0.1328, + "step": 4676 + }, + { + "epoch": 0.24, + "learning_rate": 3.5380610528291755e-07, + "loss": 0.1253, + "step": 4677 + }, + { + "epoch": 0.24, + "learning_rate": 3.5378449038648876e-07, + "loss": 0.1164, + "step": 4678 + }, + { + "epoch": 0.24, + "learning_rate": 3.5376287109483974e-07, + "loss": 0.109, + "step": 4679 + }, + { + "epoch": 0.24, + "learning_rate": 3.5374124740858864e-07, + "loss": 0.1397, + "step": 4680 + }, + { + "epoch": 0.24, + "learning_rate": 3.537196193283533e-07, + "loss": 0.1694, + "step": 4681 + }, + { + "epoch": 0.24, + "learning_rate": 3.5369798685475185e-07, + "loss": 0.1465, + "step": 4682 + }, + { + "epoch": 0.24, + "learning_rate": 3.536763499884026e-07, + "loss": 0.1395, + "step": 4683 + }, + { + "epoch": 0.24, + "learning_rate": 3.53654708729924e-07, + "loss": 0.1651, + "step": 4684 + }, + { + "epoch": 0.24, + "learning_rate": 3.5363306307993446e-07, + "loss": 0.157, + "step": 4685 + }, + { + "epoch": 0.24, + "learning_rate": 3.5361141303905266e-07, + "loss": 0.148, + "step": 4686 + }, + { + "epoch": 0.24, + "learning_rate": 3.535897586078974e-07, + "loss": 0.1381, + "step": 4687 + }, + { + "epoch": 0.24, + "learning_rate": 3.535680997870876e-07, + "loss": 0.1692, + "step": 4688 + }, + { + "epoch": 0.24, + "learning_rate": 3.5354643657724213e-07, + "loss": 0.147, + "step": 4689 + }, + { + "epoch": 0.24, + "learning_rate": 3.535247689789803e-07, + "loss": 0.171, + "step": 4690 + }, + { + "epoch": 0.24, + "learning_rate": 3.5350309699292123e-07, + "loss": 0.1338, + "step": 4691 + }, + { + "epoch": 0.24, + "learning_rate": 3.5348142061968446e-07, + "loss": 0.1427, + "step": 4692 + }, + { + "epoch": 0.24, + "learning_rate": 3.5345973985988937e-07, + "loss": 0.153, + "step": 4693 + }, + { + "epoch": 0.25, + "learning_rate": 3.534380547141557e-07, + "loss": 0.149, + "step": 4694 + }, + { + "epoch": 0.25, + "learning_rate": 3.534163651831033e-07, + "loss": 0.1527, + "step": 4695 + }, + { + "epoch": 0.25, + "learning_rate": 3.5339467126735183e-07, + "loss": 0.1083, + "step": 4696 + }, + { + "epoch": 0.25, + "learning_rate": 3.533729729675215e-07, + "loss": 0.1269, + "step": 4697 + }, + { + "epoch": 0.25, + "learning_rate": 3.5335127028423236e-07, + "loss": 0.1526, + "step": 4698 + }, + { + "epoch": 0.25, + "learning_rate": 3.5332956321810474e-07, + "loss": 0.1374, + "step": 4699 + }, + { + "epoch": 0.25, + "learning_rate": 3.53307851769759e-07, + "loss": 0.1252, + "step": 4700 + }, + { + "epoch": 0.25, + "learning_rate": 3.532861359398157e-07, + "loss": 0.1208, + "step": 4701 + }, + { + "epoch": 0.25, + "learning_rate": 3.532644157288954e-07, + "loss": 0.1417, + "step": 4702 + }, + { + "epoch": 0.25, + "learning_rate": 3.5324269113761895e-07, + "loss": 0.133, + "step": 4703 + }, + { + "epoch": 0.25, + "learning_rate": 3.532209621666073e-07, + "loss": 0.1332, + "step": 4704 + }, + { + "epoch": 0.25, + "learning_rate": 3.5319922881648133e-07, + "loss": 0.1229, + "step": 4705 + }, + { + "epoch": 0.25, + "learning_rate": 3.5317749108786225e-07, + "loss": 0.1514, + "step": 4706 + }, + { + "epoch": 0.25, + "learning_rate": 3.531557489813714e-07, + "loss": 0.1276, + "step": 4707 + }, + { + "epoch": 0.25, + "learning_rate": 3.531340024976301e-07, + "loss": 0.122, + "step": 4708 + }, + { + "epoch": 0.25, + "learning_rate": 3.5311225163725987e-07, + "loss": 0.1369, + "step": 4709 + }, + { + "epoch": 0.25, + "learning_rate": 3.530904964008824e-07, + "loss": 0.153, + "step": 4710 + }, + { + "epoch": 0.25, + "learning_rate": 3.530687367891194e-07, + "loss": 0.126, + "step": 4711 + }, + { + "epoch": 0.25, + "learning_rate": 3.5304697280259286e-07, + "loss": 0.1501, + "step": 4712 + }, + { + "epoch": 0.25, + "learning_rate": 3.530252044419248e-07, + "loss": 0.154, + "step": 4713 + }, + { + "epoch": 0.25, + "learning_rate": 3.530034317077372e-07, + "loss": 0.1264, + "step": 4714 + }, + { + "epoch": 0.25, + "learning_rate": 3.5298165460065255e-07, + "loss": 0.1414, + "step": 4715 + }, + { + "epoch": 0.25, + "learning_rate": 3.5295987312129313e-07, + "loss": 0.133, + "step": 4716 + }, + { + "epoch": 0.25, + "learning_rate": 3.529380872702815e-07, + "loss": 0.1469, + "step": 4717 + }, + { + "epoch": 0.25, + "learning_rate": 3.529162970482403e-07, + "loss": 0.1164, + "step": 4718 + }, + { + "epoch": 0.25, + "learning_rate": 3.528945024557923e-07, + "loss": 0.1406, + "step": 4719 + }, + { + "epoch": 0.25, + "learning_rate": 3.5287270349356035e-07, + "loss": 0.1372, + "step": 4720 + }, + { + "epoch": 0.25, + "learning_rate": 3.5285090016216755e-07, + "loss": 0.1743, + "step": 4721 + }, + { + "epoch": 0.25, + "learning_rate": 3.52829092462237e-07, + "loss": 0.1394, + "step": 4722 + }, + { + "epoch": 0.25, + "learning_rate": 3.52807280394392e-07, + "loss": 0.1404, + "step": 4723 + }, + { + "epoch": 0.25, + "learning_rate": 3.5278546395925595e-07, + "loss": 0.116, + "step": 4724 + }, + { + "epoch": 0.25, + "learning_rate": 3.527636431574524e-07, + "loss": 0.1396, + "step": 4725 + }, + { + "epoch": 0.25, + "learning_rate": 3.527418179896049e-07, + "loss": 0.1382, + "step": 4726 + }, + { + "epoch": 0.25, + "learning_rate": 3.5271998845633726e-07, + "loss": 0.1676, + "step": 4727 + }, + { + "epoch": 0.25, + "learning_rate": 3.5269815455827344e-07, + "loss": 0.1669, + "step": 4728 + }, + { + "epoch": 0.25, + "learning_rate": 3.5267631629603736e-07, + "loss": 0.1119, + "step": 4729 + }, + { + "epoch": 0.25, + "learning_rate": 3.526544736702533e-07, + "loss": 0.1579, + "step": 4730 + }, + { + "epoch": 0.25, + "learning_rate": 3.526326266815454e-07, + "loss": 0.1035, + "step": 4731 + }, + { + "epoch": 0.25, + "learning_rate": 3.5261077533053813e-07, + "loss": 0.1345, + "step": 4732 + }, + { + "epoch": 0.25, + "learning_rate": 3.525889196178559e-07, + "loss": 0.1535, + "step": 4733 + }, + { + "epoch": 0.25, + "learning_rate": 3.5256705954412355e-07, + "loss": 0.106, + "step": 4734 + }, + { + "epoch": 0.25, + "learning_rate": 3.525451951099657e-07, + "loss": 0.1397, + "step": 4735 + }, + { + "epoch": 0.25, + "learning_rate": 3.525233263160073e-07, + "loss": 0.1702, + "step": 4736 + }, + { + "epoch": 0.25, + "learning_rate": 3.5250145316287336e-07, + "loss": 0.1502, + "step": 4737 + }, + { + "epoch": 0.25, + "learning_rate": 3.52479575651189e-07, + "loss": 0.1624, + "step": 4738 + }, + { + "epoch": 0.25, + "learning_rate": 3.524576937815795e-07, + "loss": 0.1543, + "step": 4739 + }, + { + "epoch": 0.25, + "learning_rate": 3.5243580755467024e-07, + "loss": 0.1438, + "step": 4740 + }, + { + "epoch": 0.25, + "learning_rate": 3.5241391697108683e-07, + "loss": 0.1707, + "step": 4741 + }, + { + "epoch": 0.25, + "learning_rate": 3.5239202203145475e-07, + "loss": 0.1873, + "step": 4742 + }, + { + "epoch": 0.25, + "learning_rate": 3.523701227363999e-07, + "loss": 0.1506, + "step": 4743 + }, + { + "epoch": 0.25, + "learning_rate": 3.523482190865481e-07, + "loss": 0.1473, + "step": 4744 + }, + { + "epoch": 0.25, + "learning_rate": 3.5232631108252535e-07, + "loss": 0.1873, + "step": 4745 + }, + { + "epoch": 0.25, + "learning_rate": 3.5230439872495794e-07, + "loss": 0.1219, + "step": 4746 + }, + { + "epoch": 0.25, + "learning_rate": 3.522824820144719e-07, + "loss": 0.1228, + "step": 4747 + }, + { + "epoch": 0.25, + "learning_rate": 3.522605609516938e-07, + "loss": 0.1466, + "step": 4748 + }, + { + "epoch": 0.25, + "learning_rate": 3.522386355372501e-07, + "loss": 0.1322, + "step": 4749 + }, + { + "epoch": 0.25, + "learning_rate": 3.5221670577176737e-07, + "loss": 0.1588, + "step": 4750 + }, + { + "epoch": 0.25, + "learning_rate": 3.5219477165587246e-07, + "loss": 0.1462, + "step": 4751 + }, + { + "epoch": 0.25, + "learning_rate": 3.521728331901922e-07, + "loss": 0.1878, + "step": 4752 + }, + { + "epoch": 0.25, + "learning_rate": 3.5215089037535364e-07, + "loss": 0.1386, + "step": 4753 + }, + { + "epoch": 0.25, + "learning_rate": 3.521289432119839e-07, + "loss": 0.1533, + "step": 4754 + }, + { + "epoch": 0.25, + "learning_rate": 3.5210699170071027e-07, + "loss": 0.1461, + "step": 4755 + }, + { + "epoch": 0.25, + "learning_rate": 3.520850358421601e-07, + "loss": 0.1797, + "step": 4756 + }, + { + "epoch": 0.25, + "learning_rate": 3.520630756369608e-07, + "loss": 0.1648, + "step": 4757 + }, + { + "epoch": 0.25, + "learning_rate": 3.5204111108574014e-07, + "loss": 0.1405, + "step": 4758 + }, + { + "epoch": 0.25, + "learning_rate": 3.520191421891259e-07, + "loss": 0.1383, + "step": 4759 + }, + { + "epoch": 0.25, + "learning_rate": 3.519971689477458e-07, + "loss": 0.1407, + "step": 4760 + }, + { + "epoch": 0.25, + "learning_rate": 3.5197519136222795e-07, + "loss": 0.1255, + "step": 4761 + }, + { + "epoch": 0.25, + "learning_rate": 3.519532094332005e-07, + "loss": 0.1148, + "step": 4762 + }, + { + "epoch": 0.25, + "learning_rate": 3.5193122316129163e-07, + "loss": 0.1412, + "step": 4763 + }, + { + "epoch": 0.25, + "learning_rate": 3.519092325471298e-07, + "loss": 0.1437, + "step": 4764 + }, + { + "epoch": 0.25, + "learning_rate": 3.5188723759134345e-07, + "loss": 0.1524, + "step": 4765 + }, + { + "epoch": 0.25, + "learning_rate": 3.518652382945612e-07, + "loss": 0.1489, + "step": 4766 + }, + { + "epoch": 0.25, + "learning_rate": 3.518432346574118e-07, + "loss": 0.1537, + "step": 4767 + }, + { + "epoch": 0.25, + "learning_rate": 3.5182122668052415e-07, + "loss": 0.1391, + "step": 4768 + }, + { + "epoch": 0.25, + "learning_rate": 3.517992143645272e-07, + "loss": 0.1552, + "step": 4769 + }, + { + "epoch": 0.25, + "learning_rate": 3.5177719771005013e-07, + "loss": 0.1659, + "step": 4770 + }, + { + "epoch": 0.25, + "learning_rate": 3.517551767177222e-07, + "loss": 0.1455, + "step": 4771 + }, + { + "epoch": 0.25, + "learning_rate": 3.517331513881727e-07, + "loss": 0.1129, + "step": 4772 + }, + { + "epoch": 0.25, + "learning_rate": 3.517111217220311e-07, + "loss": 0.1887, + "step": 4773 + }, + { + "epoch": 0.25, + "learning_rate": 3.5168908771992715e-07, + "loss": 0.1861, + "step": 4774 + }, + { + "epoch": 0.25, + "learning_rate": 3.516670493824905e-07, + "loss": 0.1353, + "step": 4775 + }, + { + "epoch": 0.25, + "learning_rate": 3.51645006710351e-07, + "loss": 0.2008, + "step": 4776 + }, + { + "epoch": 0.25, + "learning_rate": 3.516229597041387e-07, + "loss": 0.1584, + "step": 4777 + }, + { + "epoch": 0.25, + "learning_rate": 3.516009083644837e-07, + "loss": 0.1849, + "step": 4778 + }, + { + "epoch": 0.25, + "learning_rate": 3.515788526920161e-07, + "loss": 0.153, + "step": 4779 + }, + { + "epoch": 0.25, + "learning_rate": 3.515567926873665e-07, + "loss": 0.1239, + "step": 4780 + }, + { + "epoch": 0.25, + "learning_rate": 3.515347283511652e-07, + "loss": 0.1638, + "step": 4781 + }, + { + "epoch": 0.25, + "learning_rate": 3.515126596840429e-07, + "loss": 0.1319, + "step": 4782 + }, + { + "epoch": 0.25, + "learning_rate": 3.5149058668663025e-07, + "loss": 0.1332, + "step": 4783 + }, + { + "epoch": 0.25, + "learning_rate": 3.5146850935955823e-07, + "loss": 0.1588, + "step": 4784 + }, + { + "epoch": 0.25, + "learning_rate": 3.514464277034577e-07, + "loss": 0.1594, + "step": 4785 + }, + { + "epoch": 0.25, + "learning_rate": 3.514243417189598e-07, + "loss": 0.1726, + "step": 4786 + }, + { + "epoch": 0.25, + "learning_rate": 3.5140225140669577e-07, + "loss": 0.1614, + "step": 4787 + }, + { + "epoch": 0.25, + "learning_rate": 3.5138015676729686e-07, + "loss": 0.1329, + "step": 4788 + }, + { + "epoch": 0.25, + "learning_rate": 3.513580578013947e-07, + "loss": 0.1156, + "step": 4789 + }, + { + "epoch": 0.25, + "learning_rate": 3.513359545096208e-07, + "loss": 0.1504, + "step": 4790 + }, + { + "epoch": 0.25, + "learning_rate": 3.5131384689260694e-07, + "loss": 0.1665, + "step": 4791 + }, + { + "epoch": 0.25, + "learning_rate": 3.512917349509849e-07, + "loss": 0.1864, + "step": 4792 + }, + { + "epoch": 0.25, + "learning_rate": 3.5126961868538665e-07, + "loss": 0.1351, + "step": 4793 + }, + { + "epoch": 0.25, + "learning_rate": 3.512474980964444e-07, + "loss": 0.1573, + "step": 4794 + }, + { + "epoch": 0.25, + "learning_rate": 3.5122537318479014e-07, + "loss": 0.1244, + "step": 4795 + }, + { + "epoch": 0.25, + "learning_rate": 3.512032439510564e-07, + "loss": 0.1529, + "step": 4796 + }, + { + "epoch": 0.25, + "learning_rate": 3.511811103958755e-07, + "loss": 0.1528, + "step": 4797 + }, + { + "epoch": 0.25, + "learning_rate": 3.511589725198802e-07, + "loss": 0.1549, + "step": 4798 + }, + { + "epoch": 0.25, + "learning_rate": 3.511368303237031e-07, + "loss": 0.1364, + "step": 4799 + }, + { + "epoch": 0.25, + "learning_rate": 3.51114683807977e-07, + "loss": 0.1534, + "step": 4800 + }, + { + "epoch": 0.25, + "learning_rate": 3.510925329733349e-07, + "loss": 0.1435, + "step": 4801 + }, + { + "epoch": 0.25, + "learning_rate": 3.5107037782040993e-07, + "loss": 0.1216, + "step": 4802 + }, + { + "epoch": 0.25, + "learning_rate": 3.510482183498352e-07, + "loss": 0.121, + "step": 4803 + }, + { + "epoch": 0.25, + "learning_rate": 3.510260545622441e-07, + "loss": 0.12, + "step": 4804 + }, + { + "epoch": 0.25, + "learning_rate": 3.5100388645827e-07, + "loss": 0.118, + "step": 4805 + }, + { + "epoch": 0.25, + "learning_rate": 3.509817140385466e-07, + "loss": 0.1521, + "step": 4806 + }, + { + "epoch": 0.25, + "learning_rate": 3.5095953730370756e-07, + "loss": 0.1344, + "step": 4807 + }, + { + "epoch": 0.25, + "learning_rate": 3.509373562543866e-07, + "loss": 0.1486, + "step": 4808 + }, + { + "epoch": 0.25, + "learning_rate": 3.5091517089121776e-07, + "loss": 0.1469, + "step": 4809 + }, + { + "epoch": 0.25, + "learning_rate": 3.5089298121483503e-07, + "loss": 0.1738, + "step": 4810 + }, + { + "epoch": 0.25, + "learning_rate": 3.5087078722587273e-07, + "loss": 0.1611, + "step": 4811 + }, + { + "epoch": 0.25, + "learning_rate": 3.50848588924965e-07, + "loss": 0.1295, + "step": 4812 + }, + { + "epoch": 0.25, + "learning_rate": 3.508263863127464e-07, + "loss": 0.1243, + "step": 4813 + }, + { + "epoch": 0.25, + "learning_rate": 3.5080417938985153e-07, + "loss": 0.1701, + "step": 4814 + }, + { + "epoch": 0.25, + "learning_rate": 3.507819681569149e-07, + "loss": 0.1447, + "step": 4815 + }, + { + "epoch": 0.25, + "learning_rate": 3.507597526145714e-07, + "loss": 0.1468, + "step": 4816 + }, + { + "epoch": 0.25, + "learning_rate": 3.50737532763456e-07, + "loss": 0.132, + "step": 4817 + }, + { + "epoch": 0.25, + "learning_rate": 3.507153086042038e-07, + "loss": 0.1196, + "step": 4818 + }, + { + "epoch": 0.25, + "learning_rate": 3.5069308013744976e-07, + "loss": 0.1615, + "step": 4819 + }, + { + "epoch": 0.25, + "learning_rate": 3.5067084736382945e-07, + "loss": 0.1797, + "step": 4820 + }, + { + "epoch": 0.25, + "learning_rate": 3.506486102839781e-07, + "loss": 0.1155, + "step": 4821 + }, + { + "epoch": 0.25, + "learning_rate": 3.506263688985313e-07, + "loss": 0.1454, + "step": 4822 + }, + { + "epoch": 0.25, + "learning_rate": 3.5060412320812473e-07, + "loss": 0.1699, + "step": 4823 + }, + { + "epoch": 0.25, + "learning_rate": 3.505818732133942e-07, + "loss": 0.124, + "step": 4824 + }, + { + "epoch": 0.25, + "learning_rate": 3.505596189149756e-07, + "loss": 0.1364, + "step": 4825 + }, + { + "epoch": 0.25, + "learning_rate": 3.505373603135049e-07, + "loss": 0.1431, + "step": 4826 + }, + { + "epoch": 0.25, + "learning_rate": 3.5051509740961846e-07, + "loss": 0.1677, + "step": 4827 + }, + { + "epoch": 0.25, + "learning_rate": 3.5049283020395236e-07, + "loss": 0.1529, + "step": 4828 + }, + { + "epoch": 0.25, + "learning_rate": 3.5047055869714306e-07, + "loss": 0.1438, + "step": 4829 + }, + { + "epoch": 0.25, + "learning_rate": 3.504482828898271e-07, + "loss": 0.1543, + "step": 4830 + }, + { + "epoch": 0.25, + "learning_rate": 3.504260027826412e-07, + "loss": 0.1092, + "step": 4831 + }, + { + "epoch": 0.25, + "learning_rate": 3.50403718376222e-07, + "loss": 0.1277, + "step": 4832 + }, + { + "epoch": 0.25, + "learning_rate": 3.5038142967120647e-07, + "loss": 0.1526, + "step": 4833 + }, + { + "epoch": 0.25, + "learning_rate": 3.5035913666823166e-07, + "loss": 0.13, + "step": 4834 + }, + { + "epoch": 0.25, + "learning_rate": 3.5033683936793467e-07, + "loss": 0.1583, + "step": 4835 + }, + { + "epoch": 0.25, + "learning_rate": 3.503145377709528e-07, + "loss": 0.1615, + "step": 4836 + }, + { + "epoch": 0.25, + "learning_rate": 3.502922318779234e-07, + "loss": 0.1365, + "step": 4837 + }, + { + "epoch": 0.25, + "learning_rate": 3.502699216894839e-07, + "loss": 0.165, + "step": 4838 + }, + { + "epoch": 0.25, + "learning_rate": 3.502476072062722e-07, + "loss": 0.162, + "step": 4839 + }, + { + "epoch": 0.25, + "learning_rate": 3.502252884289258e-07, + "loss": 0.1547, + "step": 4840 + }, + { + "epoch": 0.25, + "learning_rate": 3.5020296535808266e-07, + "loss": 0.1324, + "step": 4841 + }, + { + "epoch": 0.25, + "learning_rate": 3.501806379943808e-07, + "loss": 0.1498, + "step": 4842 + }, + { + "epoch": 0.25, + "learning_rate": 3.501583063384583e-07, + "loss": 0.136, + "step": 4843 + }, + { + "epoch": 0.25, + "learning_rate": 3.5013597039095343e-07, + "loss": 0.1139, + "step": 4844 + }, + { + "epoch": 0.25, + "learning_rate": 3.501136301525046e-07, + "loss": 0.1271, + "step": 4845 + }, + { + "epoch": 0.25, + "learning_rate": 3.500912856237503e-07, + "loss": 0.144, + "step": 4846 + }, + { + "epoch": 0.25, + "learning_rate": 3.5006893680532903e-07, + "loss": 0.1527, + "step": 4847 + }, + { + "epoch": 0.25, + "learning_rate": 3.500465836978797e-07, + "loss": 0.2016, + "step": 4848 + }, + { + "epoch": 0.25, + "learning_rate": 3.5002422630204103e-07, + "loss": 0.1258, + "step": 4849 + }, + { + "epoch": 0.25, + "learning_rate": 3.500018646184521e-07, + "loss": 0.1213, + "step": 4850 + }, + { + "epoch": 0.25, + "learning_rate": 3.49979498647752e-07, + "loss": 0.1536, + "step": 4851 + }, + { + "epoch": 0.25, + "learning_rate": 3.4995712839057976e-07, + "loss": 0.1317, + "step": 4852 + }, + { + "epoch": 0.25, + "learning_rate": 3.4993475384757504e-07, + "loss": 0.1326, + "step": 4853 + }, + { + "epoch": 0.25, + "learning_rate": 3.499123750193772e-07, + "loss": 0.1723, + "step": 4854 + }, + { + "epoch": 0.25, + "learning_rate": 3.498899919066257e-07, + "loss": 0.1402, + "step": 4855 + }, + { + "epoch": 0.25, + "learning_rate": 3.498676045099605e-07, + "loss": 0.1088, + "step": 4856 + }, + { + "epoch": 0.25, + "learning_rate": 3.4984521283002126e-07, + "loss": 0.1512, + "step": 4857 + }, + { + "epoch": 0.25, + "learning_rate": 3.4982281686744797e-07, + "loss": 0.2071, + "step": 4858 + }, + { + "epoch": 0.25, + "learning_rate": 3.498004166228807e-07, + "loss": 0.142, + "step": 4859 + }, + { + "epoch": 0.25, + "learning_rate": 3.4977801209695974e-07, + "loss": 0.1143, + "step": 4860 + }, + { + "epoch": 0.25, + "learning_rate": 3.497556032903254e-07, + "loss": 0.1508, + "step": 4861 + }, + { + "epoch": 0.25, + "learning_rate": 3.49733190203618e-07, + "loss": 0.1236, + "step": 4862 + }, + { + "epoch": 0.25, + "learning_rate": 3.4971077283747833e-07, + "loss": 0.1273, + "step": 4863 + }, + { + "epoch": 0.25, + "learning_rate": 3.4968835119254695e-07, + "loss": 0.1838, + "step": 4864 + }, + { + "epoch": 0.25, + "learning_rate": 3.4966592526946465e-07, + "loss": 0.1525, + "step": 4865 + }, + { + "epoch": 0.25, + "learning_rate": 3.4964349506887254e-07, + "loss": 0.132, + "step": 4866 + }, + { + "epoch": 0.25, + "learning_rate": 3.496210605914115e-07, + "loss": 0.141, + "step": 4867 + }, + { + "epoch": 0.25, + "learning_rate": 3.4959862183772277e-07, + "loss": 0.1514, + "step": 4868 + }, + { + "epoch": 0.25, + "learning_rate": 3.4957617880844773e-07, + "loss": 0.1531, + "step": 4869 + }, + { + "epoch": 0.25, + "learning_rate": 3.495537315042277e-07, + "loss": 0.1619, + "step": 4870 + }, + { + "epoch": 0.25, + "learning_rate": 3.495312799257044e-07, + "loss": 0.1579, + "step": 4871 + }, + { + "epoch": 0.25, + "learning_rate": 3.4950882407351926e-07, + "loss": 0.1709, + "step": 4872 + }, + { + "epoch": 0.25, + "learning_rate": 3.494863639483143e-07, + "loss": 0.1418, + "step": 4873 + }, + { + "epoch": 0.25, + "learning_rate": 3.494638995507313e-07, + "loss": 0.152, + "step": 4874 + }, + { + "epoch": 0.25, + "learning_rate": 3.4944143088141236e-07, + "loss": 0.1512, + "step": 4875 + }, + { + "epoch": 0.25, + "learning_rate": 3.4941895794099965e-07, + "loss": 0.1226, + "step": 4876 + }, + { + "epoch": 0.25, + "learning_rate": 3.4939648073013545e-07, + "loss": 0.1312, + "step": 4877 + }, + { + "epoch": 0.25, + "learning_rate": 3.4937399924946217e-07, + "loss": 0.1273, + "step": 4878 + }, + { + "epoch": 0.25, + "learning_rate": 3.493515134996223e-07, + "loss": 0.1349, + "step": 4879 + }, + { + "epoch": 0.25, + "learning_rate": 3.493290234812585e-07, + "loss": 0.1431, + "step": 4880 + }, + { + "epoch": 0.25, + "learning_rate": 3.4930652919501356e-07, + "loss": 0.1333, + "step": 4881 + }, + { + "epoch": 0.25, + "learning_rate": 3.492840306415305e-07, + "loss": 0.1467, + "step": 4882 + }, + { + "epoch": 0.25, + "learning_rate": 3.4926152782145205e-07, + "loss": 0.1469, + "step": 4883 + }, + { + "epoch": 0.25, + "learning_rate": 3.492390207354216e-07, + "loss": 0.1392, + "step": 4884 + }, + { + "epoch": 0.25, + "learning_rate": 3.492165093840823e-07, + "loss": 0.1398, + "step": 4885 + }, + { + "epoch": 0.26, + "learning_rate": 3.4919399376807756e-07, + "loss": 0.1313, + "step": 4886 + }, + { + "epoch": 0.26, + "learning_rate": 3.491714738880509e-07, + "loss": 0.1347, + "step": 4887 + }, + { + "epoch": 0.26, + "learning_rate": 3.4914894974464596e-07, + "loss": 0.1586, + "step": 4888 + }, + { + "epoch": 0.26, + "learning_rate": 3.4912642133850633e-07, + "loss": 0.1814, + "step": 4889 + }, + { + "epoch": 0.26, + "learning_rate": 3.491038886702761e-07, + "loss": 0.1499, + "step": 4890 + }, + { + "epoch": 0.26, + "learning_rate": 3.4908135174059916e-07, + "loss": 0.1773, + "step": 4891 + }, + { + "epoch": 0.26, + "learning_rate": 3.490588105501197e-07, + "loss": 0.1183, + "step": 4892 + }, + { + "epoch": 0.26, + "learning_rate": 3.490362650994818e-07, + "loss": 0.1455, + "step": 4893 + }, + { + "epoch": 0.26, + "learning_rate": 3.4901371538932996e-07, + "loss": 0.1385, + "step": 4894 + }, + { + "epoch": 0.26, + "learning_rate": 3.4899116142030863e-07, + "loss": 0.1575, + "step": 4895 + }, + { + "epoch": 0.26, + "learning_rate": 3.489686031930623e-07, + "loss": 0.1731, + "step": 4896 + }, + { + "epoch": 0.26, + "learning_rate": 3.489460407082359e-07, + "loss": 0.1462, + "step": 4897 + }, + { + "epoch": 0.26, + "learning_rate": 3.4892347396647404e-07, + "loss": 0.177, + "step": 4898 + }, + { + "epoch": 0.26, + "learning_rate": 3.489009029684219e-07, + "loss": 0.1317, + "step": 4899 + }, + { + "epoch": 0.26, + "learning_rate": 3.4887832771472445e-07, + "loss": 0.1575, + "step": 4900 + }, + { + "epoch": 0.26, + "learning_rate": 3.488557482060269e-07, + "loss": 0.1553, + "step": 4901 + }, + { + "epoch": 0.26, + "learning_rate": 3.488331644429746e-07, + "loss": 0.1633, + "step": 4902 + }, + { + "epoch": 0.26, + "learning_rate": 3.4881057642621304e-07, + "loss": 0.1148, + "step": 4903 + }, + { + "epoch": 0.26, + "learning_rate": 3.4878798415638773e-07, + "loss": 0.1166, + "step": 4904 + }, + { + "epoch": 0.26, + "learning_rate": 3.4876538763414436e-07, + "loss": 0.1366, + "step": 4905 + }, + { + "epoch": 0.26, + "learning_rate": 3.4874278686012885e-07, + "loss": 0.1748, + "step": 4906 + }, + { + "epoch": 0.26, + "learning_rate": 3.48720181834987e-07, + "loss": 0.1442, + "step": 4907 + }, + { + "epoch": 0.26, + "learning_rate": 3.48697572559365e-07, + "loss": 0.1302, + "step": 4908 + }, + { + "epoch": 0.26, + "learning_rate": 3.486749590339089e-07, + "loss": 0.1308, + "step": 4909 + }, + { + "epoch": 0.26, + "learning_rate": 3.4865234125926515e-07, + "loss": 0.1491, + "step": 4910 + }, + { + "epoch": 0.26, + "learning_rate": 3.4862971923608016e-07, + "loss": 0.1306, + "step": 4911 + }, + { + "epoch": 0.26, + "learning_rate": 3.4860709296500027e-07, + "loss": 0.1372, + "step": 4912 + }, + { + "epoch": 0.26, + "learning_rate": 3.4858446244667234e-07, + "loss": 0.1684, + "step": 4913 + }, + { + "epoch": 0.26, + "learning_rate": 3.4856182768174315e-07, + "loss": 0.1408, + "step": 4914 + }, + { + "epoch": 0.26, + "learning_rate": 3.485391886708596e-07, + "loss": 0.1445, + "step": 4915 + }, + { + "epoch": 0.26, + "learning_rate": 3.4851654541466865e-07, + "loss": 0.1629, + "step": 4916 + }, + { + "epoch": 0.26, + "learning_rate": 3.484938979138175e-07, + "loss": 0.1144, + "step": 4917 + }, + { + "epoch": 0.26, + "learning_rate": 3.4847124616895344e-07, + "loss": 0.1902, + "step": 4918 + }, + { + "epoch": 0.26, + "learning_rate": 3.4844859018072384e-07, + "loss": 0.1705, + "step": 4919 + }, + { + "epoch": 0.26, + "learning_rate": 3.484259299497762e-07, + "loss": 0.1586, + "step": 4920 + }, + { + "epoch": 0.26, + "learning_rate": 3.484032654767582e-07, + "loss": 0.1583, + "step": 4921 + }, + { + "epoch": 0.26, + "learning_rate": 3.4838059676231763e-07, + "loss": 0.1436, + "step": 4922 + }, + { + "epoch": 0.26, + "learning_rate": 3.483579238071023e-07, + "loss": 0.1577, + "step": 4923 + }, + { + "epoch": 0.26, + "learning_rate": 3.483352466117602e-07, + "loss": 0.1056, + "step": 4924 + }, + { + "epoch": 0.26, + "learning_rate": 3.483125651769395e-07, + "loss": 0.132, + "step": 4925 + }, + { + "epoch": 0.26, + "learning_rate": 3.4828987950328844e-07, + "loss": 0.1506, + "step": 4926 + }, + { + "epoch": 0.26, + "learning_rate": 3.4826718959145537e-07, + "loss": 0.1396, + "step": 4927 + }, + { + "epoch": 0.26, + "learning_rate": 3.482444954420888e-07, + "loss": 0.124, + "step": 4928 + }, + { + "epoch": 0.26, + "learning_rate": 3.4822179705583735e-07, + "loss": 0.1495, + "step": 4929 + }, + { + "epoch": 0.26, + "learning_rate": 3.481990944333497e-07, + "loss": 0.1186, + "step": 4930 + }, + { + "epoch": 0.26, + "learning_rate": 3.4817638757527466e-07, + "loss": 0.1249, + "step": 4931 + }, + { + "epoch": 0.26, + "learning_rate": 3.4815367648226133e-07, + "loss": 0.1466, + "step": 4932 + }, + { + "epoch": 0.26, + "learning_rate": 3.481309611549587e-07, + "loss": 0.1455, + "step": 4933 + }, + { + "epoch": 0.26, + "learning_rate": 3.4810824159401604e-07, + "loss": 0.1558, + "step": 4934 + }, + { + "epoch": 0.26, + "learning_rate": 3.4808551780008263e-07, + "loss": 0.1618, + "step": 4935 + }, + { + "epoch": 0.26, + "learning_rate": 3.4806278977380797e-07, + "loss": 0.1394, + "step": 4936 + }, + { + "epoch": 0.26, + "learning_rate": 3.480400575158416e-07, + "loss": 0.1298, + "step": 4937 + }, + { + "epoch": 0.26, + "learning_rate": 3.480173210268333e-07, + "loss": 0.1667, + "step": 4938 + }, + { + "epoch": 0.26, + "learning_rate": 3.4799458030743275e-07, + "loss": 0.1311, + "step": 4939 + }, + { + "epoch": 0.26, + "learning_rate": 3.4797183535829006e-07, + "loss": 0.1243, + "step": 4940 + }, + { + "epoch": 0.26, + "learning_rate": 3.4794908618005507e-07, + "loss": 0.1384, + "step": 4941 + }, + { + "epoch": 0.26, + "learning_rate": 3.479263327733782e-07, + "loss": 0.1626, + "step": 4942 + }, + { + "epoch": 0.26, + "learning_rate": 3.479035751389095e-07, + "loss": 0.142, + "step": 4943 + }, + { + "epoch": 0.26, + "learning_rate": 3.478808132772997e-07, + "loss": 0.1416, + "step": 4944 + }, + { + "epoch": 0.26, + "learning_rate": 3.47858047189199e-07, + "loss": 0.1327, + "step": 4945 + }, + { + "epoch": 0.26, + "learning_rate": 3.4783527687525835e-07, + "loss": 0.1282, + "step": 4946 + }, + { + "epoch": 0.26, + "learning_rate": 3.4781250233612836e-07, + "loss": 0.1231, + "step": 4947 + }, + { + "epoch": 0.26, + "learning_rate": 3.4778972357246004e-07, + "loss": 0.1257, + "step": 4948 + }, + { + "epoch": 0.26, + "learning_rate": 3.4776694058490427e-07, + "loss": 0.1488, + "step": 4949 + }, + { + "epoch": 0.26, + "learning_rate": 3.4774415337411237e-07, + "loss": 0.135, + "step": 4950 + }, + { + "epoch": 0.26, + "learning_rate": 3.4772136194073555e-07, + "loss": 0.1359, + "step": 4951 + }, + { + "epoch": 0.26, + "learning_rate": 3.476985662854251e-07, + "loss": 0.1526, + "step": 4952 + }, + { + "epoch": 0.26, + "learning_rate": 3.476757664088327e-07, + "loss": 0.1139, + "step": 4953 + }, + { + "epoch": 0.26, + "learning_rate": 3.4765296231160983e-07, + "loss": 0.1282, + "step": 4954 + }, + { + "epoch": 0.26, + "learning_rate": 3.476301539944083e-07, + "loss": 0.1643, + "step": 4955 + }, + { + "epoch": 0.26, + "learning_rate": 3.4760734145788e-07, + "loss": 0.1338, + "step": 4956 + }, + { + "epoch": 0.26, + "learning_rate": 3.4758452470267687e-07, + "loss": 0.1226, + "step": 4957 + }, + { + "epoch": 0.26, + "learning_rate": 3.4756170372945107e-07, + "loss": 0.1362, + "step": 4958 + }, + { + "epoch": 0.26, + "learning_rate": 3.4753887853885484e-07, + "loss": 0.1593, + "step": 4959 + }, + { + "epoch": 0.26, + "learning_rate": 3.4751604913154045e-07, + "loss": 0.1209, + "step": 4960 + }, + { + "epoch": 0.26, + "learning_rate": 3.474932155081605e-07, + "loss": 0.1702, + "step": 4961 + }, + { + "epoch": 0.26, + "learning_rate": 3.4747037766936743e-07, + "loss": 0.1607, + "step": 4962 + }, + { + "epoch": 0.26, + "learning_rate": 3.4744753561581403e-07, + "loss": 0.1009, + "step": 4963 + }, + { + "epoch": 0.26, + "learning_rate": 3.474246893481532e-07, + "loss": 0.1325, + "step": 4964 + }, + { + "epoch": 0.26, + "learning_rate": 3.4740183886703785e-07, + "loss": 0.1891, + "step": 4965 + }, + { + "epoch": 0.26, + "learning_rate": 3.47378984173121e-07, + "loss": 0.1345, + "step": 4966 + }, + { + "epoch": 0.26, + "learning_rate": 3.473561252670559e-07, + "loss": 0.1096, + "step": 4967 + }, + { + "epoch": 0.26, + "learning_rate": 3.4733326214949586e-07, + "loss": 0.1485, + "step": 4968 + }, + { + "epoch": 0.26, + "learning_rate": 3.4731039482109437e-07, + "loss": 0.1744, + "step": 4969 + }, + { + "epoch": 0.26, + "learning_rate": 3.472875232825048e-07, + "loss": 0.1413, + "step": 4970 + }, + { + "epoch": 0.26, + "learning_rate": 3.4726464753438104e-07, + "loss": 0.1017, + "step": 4971 + }, + { + "epoch": 0.26, + "learning_rate": 3.472417675773768e-07, + "loss": 0.1368, + "step": 4972 + }, + { + "epoch": 0.26, + "learning_rate": 3.47218883412146e-07, + "loss": 0.1201, + "step": 4973 + }, + { + "epoch": 0.26, + "learning_rate": 3.4719599503934263e-07, + "loss": 0.1315, + "step": 4974 + }, + { + "epoch": 0.26, + "learning_rate": 3.4717310245962097e-07, + "loss": 0.1279, + "step": 4975 + }, + { + "epoch": 0.26, + "learning_rate": 3.471502056736352e-07, + "loss": 0.1172, + "step": 4976 + }, + { + "epoch": 0.26, + "learning_rate": 3.4712730468203974e-07, + "loss": 0.1513, + "step": 4977 + }, + { + "epoch": 0.26, + "learning_rate": 3.471043994854891e-07, + "loss": 0.1391, + "step": 4978 + }, + { + "epoch": 0.26, + "learning_rate": 3.4708149008463793e-07, + "loss": 0.1867, + "step": 4979 + }, + { + "epoch": 0.26, + "learning_rate": 3.47058576480141e-07, + "loss": 0.1594, + "step": 4980 + }, + { + "epoch": 0.26, + "learning_rate": 3.4703565867265326e-07, + "loss": 0.1273, + "step": 4981 + }, + { + "epoch": 0.26, + "learning_rate": 3.4701273666282955e-07, + "loss": 0.1471, + "step": 4982 + }, + { + "epoch": 0.26, + "learning_rate": 3.469898104513251e-07, + "loss": 0.1193, + "step": 4983 + }, + { + "epoch": 0.26, + "learning_rate": 3.4696688003879506e-07, + "loss": 0.1061, + "step": 4984 + }, + { + "epoch": 0.26, + "learning_rate": 3.4694394542589497e-07, + "loss": 0.1352, + "step": 4985 + }, + { + "epoch": 0.26, + "learning_rate": 3.469210066132801e-07, + "loss": 0.1445, + "step": 4986 + }, + { + "epoch": 0.26, + "learning_rate": 3.468980636016062e-07, + "loss": 0.1552, + "step": 4987 + }, + { + "epoch": 0.26, + "learning_rate": 3.468751163915289e-07, + "loss": 0.1343, + "step": 4988 + }, + { + "epoch": 0.26, + "learning_rate": 3.4685216498370407e-07, + "loss": 0.1411, + "step": 4989 + }, + { + "epoch": 0.26, + "learning_rate": 3.468292093787877e-07, + "loss": 0.1386, + "step": 4990 + }, + { + "epoch": 0.26, + "learning_rate": 3.468062495774358e-07, + "loss": 0.1242, + "step": 4991 + }, + { + "epoch": 0.26, + "learning_rate": 3.467832855803047e-07, + "loss": 0.1432, + "step": 4992 + }, + { + "epoch": 0.26, + "learning_rate": 3.4676031738805057e-07, + "loss": 0.1539, + "step": 4993 + }, + { + "epoch": 0.26, + "learning_rate": 3.4673734500132994e-07, + "loss": 0.1514, + "step": 4994 + }, + { + "epoch": 0.26, + "learning_rate": 3.467143684207993e-07, + "loss": 0.1614, + "step": 4995 + }, + { + "epoch": 0.26, + "learning_rate": 3.4669138764711543e-07, + "loss": 0.157, + "step": 4996 + }, + { + "epoch": 0.26, + "learning_rate": 3.466684026809351e-07, + "loss": 0.1452, + "step": 4997 + }, + { + "epoch": 0.26, + "learning_rate": 3.466454135229151e-07, + "loss": 0.1497, + "step": 4998 + }, + { + "epoch": 0.26, + "learning_rate": 3.4662242017371265e-07, + "loss": 0.1453, + "step": 4999 + }, + { + "epoch": 0.26, + "learning_rate": 3.4659942263398485e-07, + "loss": 0.1397, + "step": 5000 + }, + { + "epoch": 0.26, + "learning_rate": 3.4657642090438884e-07, + "loss": 0.1555, + "step": 5001 + }, + { + "epoch": 0.26, + "learning_rate": 3.465534149855822e-07, + "loss": 0.2007, + "step": 5002 + }, + { + "epoch": 0.26, + "learning_rate": 3.465304048782224e-07, + "loss": 0.1271, + "step": 5003 + }, + { + "epoch": 0.26, + "learning_rate": 3.4650739058296704e-07, + "loss": 0.1612, + "step": 5004 + }, + { + "epoch": 0.26, + "learning_rate": 3.4648437210047393e-07, + "loss": 0.1382, + "step": 5005 + }, + { + "epoch": 0.26, + "learning_rate": 3.464613494314009e-07, + "loss": 0.1339, + "step": 5006 + }, + { + "epoch": 0.26, + "learning_rate": 3.4643832257640595e-07, + "loss": 0.1394, + "step": 5007 + }, + { + "epoch": 0.26, + "learning_rate": 3.464152915361472e-07, + "loss": 0.1664, + "step": 5008 + }, + { + "epoch": 0.26, + "learning_rate": 3.463922563112829e-07, + "loss": 0.1447, + "step": 5009 + }, + { + "epoch": 0.26, + "learning_rate": 3.463692169024714e-07, + "loss": 0.1422, + "step": 5010 + }, + { + "epoch": 0.26, + "learning_rate": 3.4634617331037126e-07, + "loss": 0.1756, + "step": 5011 + }, + { + "epoch": 0.26, + "learning_rate": 3.463231255356409e-07, + "loss": 0.1441, + "step": 5012 + }, + { + "epoch": 0.26, + "learning_rate": 3.463000735789392e-07, + "loss": 0.1103, + "step": 5013 + }, + { + "epoch": 0.26, + "learning_rate": 3.4627701744092483e-07, + "loss": 0.1723, + "step": 5014 + }, + { + "epoch": 0.26, + "learning_rate": 3.462539571222569e-07, + "loss": 0.1134, + "step": 5015 + }, + { + "epoch": 0.26, + "learning_rate": 3.462308926235944e-07, + "loss": 0.1487, + "step": 5016 + }, + { + "epoch": 0.26, + "learning_rate": 3.462078239455965e-07, + "loss": 0.134, + "step": 5017 + }, + { + "epoch": 0.26, + "learning_rate": 3.4618475108892255e-07, + "loss": 0.1402, + "step": 5018 + }, + { + "epoch": 0.26, + "learning_rate": 3.4616167405423203e-07, + "loss": 0.1884, + "step": 5019 + }, + { + "epoch": 0.26, + "learning_rate": 3.461385928421844e-07, + "loss": 0.1469, + "step": 5020 + }, + { + "epoch": 0.26, + "learning_rate": 3.4611550745343943e-07, + "loss": 0.1219, + "step": 5021 + }, + { + "epoch": 0.26, + "learning_rate": 3.460924178886568e-07, + "loss": 0.171, + "step": 5022 + }, + { + "epoch": 0.26, + "learning_rate": 3.4606932414849653e-07, + "loss": 0.1339, + "step": 5023 + }, + { + "epoch": 0.26, + "learning_rate": 3.4604622623361854e-07, + "loss": 0.1256, + "step": 5024 + }, + { + "epoch": 0.26, + "learning_rate": 3.46023124144683e-07, + "loss": 0.1908, + "step": 5025 + }, + { + "epoch": 0.26, + "learning_rate": 3.460000178823502e-07, + "loss": 0.1246, + "step": 5026 + }, + { + "epoch": 0.26, + "learning_rate": 3.4597690744728055e-07, + "loss": 0.1311, + "step": 5027 + }, + { + "epoch": 0.26, + "learning_rate": 3.4595379284013454e-07, + "loss": 0.12, + "step": 5028 + }, + { + "epoch": 0.26, + "learning_rate": 3.4593067406157277e-07, + "loss": 0.1374, + "step": 5029 + }, + { + "epoch": 0.26, + "learning_rate": 3.45907551112256e-07, + "loss": 0.1473, + "step": 5030 + }, + { + "epoch": 0.26, + "learning_rate": 3.4588442399284514e-07, + "loss": 0.1454, + "step": 5031 + }, + { + "epoch": 0.26, + "learning_rate": 3.458612927040011e-07, + "loss": 0.1143, + "step": 5032 + }, + { + "epoch": 0.26, + "learning_rate": 3.4583815724638496e-07, + "loss": 0.1646, + "step": 5033 + }, + { + "epoch": 0.26, + "learning_rate": 3.4581501762065803e-07, + "loss": 0.1665, + "step": 5034 + }, + { + "epoch": 0.26, + "learning_rate": 3.457918738274816e-07, + "loss": 0.1478, + "step": 5035 + }, + { + "epoch": 0.26, + "learning_rate": 3.4576872586751715e-07, + "loss": 0.1433, + "step": 5036 + }, + { + "epoch": 0.26, + "learning_rate": 3.457455737414262e-07, + "loss": 0.1173, + "step": 5037 + }, + { + "epoch": 0.26, + "learning_rate": 3.457224174498705e-07, + "loss": 0.1132, + "step": 5038 + }, + { + "epoch": 0.26, + "learning_rate": 3.456992569935119e-07, + "loss": 0.1345, + "step": 5039 + }, + { + "epoch": 0.26, + "learning_rate": 3.4567609237301226e-07, + "loss": 0.1652, + "step": 5040 + }, + { + "epoch": 0.26, + "learning_rate": 3.456529235890337e-07, + "loss": 0.1598, + "step": 5041 + }, + { + "epoch": 0.26, + "learning_rate": 3.456297506422383e-07, + "loss": 0.1345, + "step": 5042 + }, + { + "epoch": 0.26, + "learning_rate": 3.4560657353328844e-07, + "loss": 0.1566, + "step": 5043 + }, + { + "epoch": 0.26, + "learning_rate": 3.4558339226284647e-07, + "loss": 0.1296, + "step": 5044 + }, + { + "epoch": 0.26, + "learning_rate": 3.4556020683157495e-07, + "loss": 0.1376, + "step": 5045 + }, + { + "epoch": 0.26, + "learning_rate": 3.4553701724013653e-07, + "loss": 0.186, + "step": 5046 + }, + { + "epoch": 0.26, + "learning_rate": 3.45513823489194e-07, + "loss": 0.1438, + "step": 5047 + }, + { + "epoch": 0.26, + "learning_rate": 3.454906255794102e-07, + "loss": 0.1467, + "step": 5048 + }, + { + "epoch": 0.26, + "learning_rate": 3.454674235114482e-07, + "loss": 0.1174, + "step": 5049 + }, + { + "epoch": 0.26, + "learning_rate": 3.4544421728597103e-07, + "loss": 0.127, + "step": 5050 + }, + { + "epoch": 0.26, + "learning_rate": 3.45421006903642e-07, + "loss": 0.1241, + "step": 5051 + }, + { + "epoch": 0.26, + "learning_rate": 3.4539779236512445e-07, + "loss": 0.115, + "step": 5052 + }, + { + "epoch": 0.26, + "learning_rate": 3.453745736710819e-07, + "loss": 0.1356, + "step": 5053 + }, + { + "epoch": 0.26, + "learning_rate": 3.4535135082217786e-07, + "loss": 0.1563, + "step": 5054 + }, + { + "epoch": 0.26, + "learning_rate": 3.453281238190761e-07, + "loss": 0.1394, + "step": 5055 + }, + { + "epoch": 0.26, + "learning_rate": 3.4530489266244045e-07, + "loss": 0.1462, + "step": 5056 + }, + { + "epoch": 0.26, + "learning_rate": 3.4528165735293497e-07, + "loss": 0.1157, + "step": 5057 + }, + { + "epoch": 0.26, + "learning_rate": 3.452584178912235e-07, + "loss": 0.2005, + "step": 5058 + }, + { + "epoch": 0.26, + "learning_rate": 3.4523517427797043e-07, + "loss": 0.1389, + "step": 5059 + }, + { + "epoch": 0.26, + "learning_rate": 3.4521192651384005e-07, + "loss": 0.1286, + "step": 5060 + }, + { + "epoch": 0.26, + "learning_rate": 3.451886745994967e-07, + "loss": 0.1388, + "step": 5061 + }, + { + "epoch": 0.26, + "learning_rate": 3.451654185356051e-07, + "loss": 0.1151, + "step": 5062 + }, + { + "epoch": 0.26, + "learning_rate": 3.4514215832282964e-07, + "loss": 0.1392, + "step": 5063 + }, + { + "epoch": 0.26, + "learning_rate": 3.4511889396183536e-07, + "loss": 0.1348, + "step": 5064 + }, + { + "epoch": 0.26, + "learning_rate": 3.450956254532871e-07, + "loss": 0.1153, + "step": 5065 + }, + { + "epoch": 0.26, + "learning_rate": 3.4507235279784973e-07, + "loss": 0.1564, + "step": 5066 + }, + { + "epoch": 0.26, + "learning_rate": 3.450490759961886e-07, + "loss": 0.1566, + "step": 5067 + }, + { + "epoch": 0.26, + "learning_rate": 3.450257950489689e-07, + "loss": 0.1296, + "step": 5068 + }, + { + "epoch": 0.26, + "learning_rate": 3.4500250995685593e-07, + "loss": 0.1398, + "step": 5069 + }, + { + "epoch": 0.26, + "learning_rate": 3.449792207205153e-07, + "loss": 0.1802, + "step": 5070 + }, + { + "epoch": 0.26, + "learning_rate": 3.4495592734061247e-07, + "loss": 0.1546, + "step": 5071 + }, + { + "epoch": 0.26, + "learning_rate": 3.449326298178134e-07, + "loss": 0.1248, + "step": 5072 + }, + { + "epoch": 0.26, + "learning_rate": 3.4490932815278376e-07, + "loss": 0.1335, + "step": 5073 + }, + { + "epoch": 0.26, + "learning_rate": 3.448860223461896e-07, + "loss": 0.1592, + "step": 5074 + }, + { + "epoch": 0.26, + "learning_rate": 3.44862712398697e-07, + "loss": 0.134, + "step": 5075 + }, + { + "epoch": 0.26, + "learning_rate": 3.4483939831097216e-07, + "loss": 0.1278, + "step": 5076 + }, + { + "epoch": 0.26, + "learning_rate": 3.4481608008368136e-07, + "loss": 0.1676, + "step": 5077 + }, + { + "epoch": 0.27, + "learning_rate": 3.447927577174911e-07, + "loss": 0.1178, + "step": 5078 + }, + { + "epoch": 0.27, + "learning_rate": 3.4476943121306795e-07, + "loss": 0.157, + "step": 5079 + }, + { + "epoch": 0.27, + "learning_rate": 3.4474610057107856e-07, + "loss": 0.1243, + "step": 5080 + }, + { + "epoch": 0.27, + "learning_rate": 3.4472276579218974e-07, + "loss": 0.1437, + "step": 5081 + }, + { + "epoch": 0.27, + "learning_rate": 3.446994268770684e-07, + "loss": 0.1507, + "step": 5082 + }, + { + "epoch": 0.27, + "learning_rate": 3.4467608382638156e-07, + "loss": 0.1332, + "step": 5083 + }, + { + "epoch": 0.27, + "learning_rate": 3.446527366407965e-07, + "loss": 0.14, + "step": 5084 + }, + { + "epoch": 0.27, + "learning_rate": 3.4462938532098025e-07, + "loss": 0.125, + "step": 5085 + }, + { + "epoch": 0.27, + "learning_rate": 3.446060298676004e-07, + "loss": 0.1471, + "step": 5086 + }, + { + "epoch": 0.27, + "learning_rate": 3.445826702813244e-07, + "loss": 0.1491, + "step": 5087 + }, + { + "epoch": 0.27, + "learning_rate": 3.4455930656281986e-07, + "loss": 0.1068, + "step": 5088 + }, + { + "epoch": 0.27, + "learning_rate": 3.445359387127545e-07, + "loss": 0.1233, + "step": 5089 + }, + { + "epoch": 0.27, + "learning_rate": 3.4451256673179623e-07, + "loss": 0.1556, + "step": 5090 + }, + { + "epoch": 0.27, + "learning_rate": 3.44489190620613e-07, + "loss": 0.156, + "step": 5091 + }, + { + "epoch": 0.27, + "learning_rate": 3.44465810379873e-07, + "loss": 0.1481, + "step": 5092 + }, + { + "epoch": 0.27, + "learning_rate": 3.444424260102443e-07, + "loss": 0.1436, + "step": 5093 + }, + { + "epoch": 0.27, + "learning_rate": 3.444190375123953e-07, + "loss": 0.1338, + "step": 5094 + }, + { + "epoch": 0.27, + "learning_rate": 3.4439564488699444e-07, + "loss": 0.1372, + "step": 5095 + }, + { + "epoch": 0.27, + "learning_rate": 3.443722481347104e-07, + "loss": 0.1188, + "step": 5096 + }, + { + "epoch": 0.27, + "learning_rate": 3.4434884725621167e-07, + "loss": 0.1194, + "step": 5097 + }, + { + "epoch": 0.27, + "learning_rate": 3.4432544225216726e-07, + "loss": 0.1425, + "step": 5098 + }, + { + "epoch": 0.27, + "learning_rate": 3.4430203312324594e-07, + "loss": 0.1657, + "step": 5099 + }, + { + "epoch": 0.27, + "learning_rate": 3.442786198701168e-07, + "loss": 0.1691, + "step": 5100 + }, + { + "epoch": 0.27, + "learning_rate": 3.4425520249344905e-07, + "loss": 0.1326, + "step": 5101 + }, + { + "epoch": 0.27, + "learning_rate": 3.442317809939119e-07, + "loss": 0.1671, + "step": 5102 + }, + { + "epoch": 0.27, + "learning_rate": 3.4420835537217473e-07, + "loss": 0.1656, + "step": 5103 + }, + { + "epoch": 0.27, + "learning_rate": 3.441849256289071e-07, + "loss": 0.1296, + "step": 5104 + }, + { + "epoch": 0.27, + "learning_rate": 3.441614917647787e-07, + "loss": 0.1258, + "step": 5105 + }, + { + "epoch": 0.27, + "learning_rate": 3.4413805378045916e-07, + "loss": 0.1362, + "step": 5106 + }, + { + "epoch": 0.27, + "learning_rate": 3.441146116766184e-07, + "loss": 0.1643, + "step": 5107 + }, + { + "epoch": 0.27, + "learning_rate": 3.4409116545392644e-07, + "loss": 0.1551, + "step": 5108 + }, + { + "epoch": 0.27, + "learning_rate": 3.440677151130534e-07, + "loss": 0.138, + "step": 5109 + }, + { + "epoch": 0.27, + "learning_rate": 3.4404426065466933e-07, + "loss": 0.1383, + "step": 5110 + }, + { + "epoch": 0.27, + "learning_rate": 3.440208020794447e-07, + "loss": 0.1453, + "step": 5111 + }, + { + "epoch": 0.27, + "learning_rate": 3.4399733938805e-07, + "loss": 0.1091, + "step": 5112 + }, + { + "epoch": 0.27, + "learning_rate": 3.439738725811557e-07, + "loss": 0.1358, + "step": 5113 + }, + { + "epoch": 0.27, + "learning_rate": 3.4395040165943254e-07, + "loss": 0.1273, + "step": 5114 + }, + { + "epoch": 0.27, + "learning_rate": 3.439269266235514e-07, + "loss": 0.1487, + "step": 5115 + }, + { + "epoch": 0.27, + "learning_rate": 3.439034474741831e-07, + "loss": 0.1631, + "step": 5116 + }, + { + "epoch": 0.27, + "learning_rate": 3.438799642119987e-07, + "loss": 0.2114, + "step": 5117 + }, + { + "epoch": 0.27, + "learning_rate": 3.438564768376694e-07, + "loss": 0.1527, + "step": 5118 + }, + { + "epoch": 0.27, + "learning_rate": 3.438329853518665e-07, + "loss": 0.1701, + "step": 5119 + }, + { + "epoch": 0.27, + "learning_rate": 3.438094897552613e-07, + "loss": 0.1387, + "step": 5120 + }, + { + "epoch": 0.27, + "learning_rate": 3.4378599004852535e-07, + "loss": 0.1215, + "step": 5121 + }, + { + "epoch": 0.27, + "learning_rate": 3.437624862323303e-07, + "loss": 0.1423, + "step": 5122 + }, + { + "epoch": 0.27, + "learning_rate": 3.4373897830734795e-07, + "loss": 0.1515, + "step": 5123 + }, + { + "epoch": 0.27, + "learning_rate": 3.4371546627425e-07, + "loss": 0.1302, + "step": 5124 + }, + { + "epoch": 0.27, + "learning_rate": 3.436919501337087e-07, + "loss": 0.1469, + "step": 5125 + }, + { + "epoch": 0.27, + "learning_rate": 3.4366842988639586e-07, + "loss": 0.1308, + "step": 5126 + }, + { + "epoch": 0.27, + "learning_rate": 3.4364490553298387e-07, + "loss": 0.1441, + "step": 5127 + }, + { + "epoch": 0.27, + "learning_rate": 3.436213770741451e-07, + "loss": 0.1564, + "step": 5128 + }, + { + "epoch": 0.27, + "learning_rate": 3.435978445105519e-07, + "loss": 0.1088, + "step": 5129 + }, + { + "epoch": 0.27, + "learning_rate": 3.4357430784287677e-07, + "loss": 0.1265, + "step": 5130 + }, + { + "epoch": 0.27, + "learning_rate": 3.4355076707179256e-07, + "loss": 0.1526, + "step": 5131 + }, + { + "epoch": 0.27, + "learning_rate": 3.435272221979721e-07, + "loss": 0.1433, + "step": 5132 + }, + { + "epoch": 0.27, + "learning_rate": 3.4350367322208813e-07, + "loss": 0.1295, + "step": 5133 + }, + { + "epoch": 0.27, + "learning_rate": 3.434801201448138e-07, + "loss": 0.1471, + "step": 5134 + }, + { + "epoch": 0.27, + "learning_rate": 3.4345656296682225e-07, + "loss": 0.2007, + "step": 5135 + }, + { + "epoch": 0.27, + "learning_rate": 3.434330016887867e-07, + "loss": 0.1759, + "step": 5136 + }, + { + "epoch": 0.27, + "learning_rate": 3.4340943631138064e-07, + "loss": 0.1322, + "step": 5137 + }, + { + "epoch": 0.27, + "learning_rate": 3.4338586683527755e-07, + "loss": 0.1501, + "step": 5138 + }, + { + "epoch": 0.27, + "learning_rate": 3.4336229326115104e-07, + "loss": 0.1372, + "step": 5139 + }, + { + "epoch": 0.27, + "learning_rate": 3.4333871558967474e-07, + "loss": 0.1161, + "step": 5140 + }, + { + "epoch": 0.27, + "learning_rate": 3.433151338215227e-07, + "loss": 0.1406, + "step": 5141 + }, + { + "epoch": 0.27, + "learning_rate": 3.4329154795736883e-07, + "loss": 0.1338, + "step": 5142 + }, + { + "epoch": 0.27, + "learning_rate": 3.432679579978871e-07, + "loss": 0.142, + "step": 5143 + }, + { + "epoch": 0.27, + "learning_rate": 3.4324436394375193e-07, + "loss": 0.103, + "step": 5144 + }, + { + "epoch": 0.27, + "learning_rate": 3.432207657956375e-07, + "loss": 0.1301, + "step": 5145 + }, + { + "epoch": 0.27, + "learning_rate": 3.431971635542183e-07, + "loss": 0.1561, + "step": 5146 + }, + { + "epoch": 0.27, + "learning_rate": 3.431735572201689e-07, + "loss": 0.1309, + "step": 5147 + }, + { + "epoch": 0.27, + "learning_rate": 3.4314994679416397e-07, + "loss": 0.1341, + "step": 5148 + }, + { + "epoch": 0.27, + "learning_rate": 3.4312633227687825e-07, + "loss": 0.1757, + "step": 5149 + }, + { + "epoch": 0.27, + "learning_rate": 3.4310271366898677e-07, + "loss": 0.1709, + "step": 5150 + }, + { + "epoch": 0.27, + "learning_rate": 3.4307909097116436e-07, + "loss": 0.1578, + "step": 5151 + }, + { + "epoch": 0.27, + "learning_rate": 3.430554641840865e-07, + "loss": 0.1372, + "step": 5152 + }, + { + "epoch": 0.27, + "learning_rate": 3.4303183330842804e-07, + "loss": 0.1609, + "step": 5153 + }, + { + "epoch": 0.27, + "learning_rate": 3.4300819834486473e-07, + "loss": 0.1588, + "step": 5154 + }, + { + "epoch": 0.27, + "learning_rate": 3.4298455929407177e-07, + "loss": 0.1495, + "step": 5155 + }, + { + "epoch": 0.27, + "learning_rate": 3.42960916156725e-07, + "loss": 0.1282, + "step": 5156 + }, + { + "epoch": 0.27, + "learning_rate": 3.4293726893350004e-07, + "loss": 0.1479, + "step": 5157 + }, + { + "epoch": 0.27, + "learning_rate": 3.429136176250727e-07, + "loss": 0.16, + "step": 5158 + }, + { + "epoch": 0.27, + "learning_rate": 3.428899622321191e-07, + "loss": 0.1045, + "step": 5159 + }, + { + "epoch": 0.27, + "learning_rate": 3.4286630275531514e-07, + "loss": 0.1164, + "step": 5160 + }, + { + "epoch": 0.27, + "learning_rate": 3.428426391953371e-07, + "loss": 0.1345, + "step": 5161 + }, + { + "epoch": 0.27, + "learning_rate": 3.4281897155286134e-07, + "loss": 0.1375, + "step": 5162 + }, + { + "epoch": 0.27, + "learning_rate": 3.427952998285642e-07, + "loss": 0.1219, + "step": 5163 + }, + { + "epoch": 0.27, + "learning_rate": 3.4277162402312226e-07, + "loss": 0.1188, + "step": 5164 + }, + { + "epoch": 0.27, + "learning_rate": 3.427479441372122e-07, + "loss": 0.1465, + "step": 5165 + }, + { + "epoch": 0.27, + "learning_rate": 3.427242601715108e-07, + "loss": 0.1547, + "step": 5166 + }, + { + "epoch": 0.27, + "learning_rate": 3.4270057212669493e-07, + "loss": 0.1136, + "step": 5167 + }, + { + "epoch": 0.27, + "learning_rate": 3.4267688000344166e-07, + "loss": 0.1638, + "step": 5168 + }, + { + "epoch": 0.27, + "learning_rate": 3.42653183802428e-07, + "loss": 0.1509, + "step": 5169 + }, + { + "epoch": 0.27, + "learning_rate": 3.4262948352433136e-07, + "loss": 0.1014, + "step": 5170 + }, + { + "epoch": 0.27, + "learning_rate": 3.42605779169829e-07, + "loss": 0.1218, + "step": 5171 + }, + { + "epoch": 0.27, + "learning_rate": 3.425820707395984e-07, + "loss": 0.1688, + "step": 5172 + }, + { + "epoch": 0.27, + "learning_rate": 3.425583582343172e-07, + "loss": 0.1728, + "step": 5173 + }, + { + "epoch": 0.27, + "learning_rate": 3.4253464165466307e-07, + "loss": 0.1449, + "step": 5174 + }, + { + "epoch": 0.27, + "learning_rate": 3.425109210013139e-07, + "loss": 0.1389, + "step": 5175 + }, + { + "epoch": 0.27, + "learning_rate": 3.424871962749475e-07, + "loss": 0.1671, + "step": 5176 + }, + { + "epoch": 0.27, + "learning_rate": 3.424634674762421e-07, + "loss": 0.1145, + "step": 5177 + }, + { + "epoch": 0.27, + "learning_rate": 3.4243973460587576e-07, + "loss": 0.1635, + "step": 5178 + }, + { + "epoch": 0.27, + "learning_rate": 3.424159976645269e-07, + "loss": 0.1362, + "step": 5179 + }, + { + "epoch": 0.27, + "learning_rate": 3.423922566528738e-07, + "loss": 0.1078, + "step": 5180 + }, + { + "epoch": 0.27, + "learning_rate": 3.4236851157159495e-07, + "loss": 0.1479, + "step": 5181 + }, + { + "epoch": 0.27, + "learning_rate": 3.4234476242136914e-07, + "loss": 0.1418, + "step": 5182 + }, + { + "epoch": 0.27, + "learning_rate": 3.423210092028751e-07, + "loss": 0.1311, + "step": 5183 + }, + { + "epoch": 0.27, + "learning_rate": 3.422972519167917e-07, + "loss": 0.146, + "step": 5184 + }, + { + "epoch": 0.27, + "learning_rate": 3.4227349056379785e-07, + "loss": 0.1211, + "step": 5185 + }, + { + "epoch": 0.27, + "learning_rate": 3.422497251445727e-07, + "loss": 0.1501, + "step": 5186 + }, + { + "epoch": 0.27, + "learning_rate": 3.422259556597955e-07, + "loss": 0.1612, + "step": 5187 + }, + { + "epoch": 0.27, + "learning_rate": 3.422021821101457e-07, + "loss": 0.1431, + "step": 5188 + }, + { + "epoch": 0.27, + "learning_rate": 3.421784044963025e-07, + "loss": 0.1527, + "step": 5189 + }, + { + "epoch": 0.27, + "learning_rate": 3.4215462281894567e-07, + "loss": 0.1411, + "step": 5190 + }, + { + "epoch": 0.27, + "learning_rate": 3.421308370787548e-07, + "loss": 0.1449, + "step": 5191 + }, + { + "epoch": 0.27, + "learning_rate": 3.4210704727640985e-07, + "loss": 0.1326, + "step": 5192 + }, + { + "epoch": 0.27, + "learning_rate": 3.420832534125905e-07, + "loss": 0.1637, + "step": 5193 + }, + { + "epoch": 0.27, + "learning_rate": 3.4205945548797697e-07, + "loss": 0.1556, + "step": 5194 + }, + { + "epoch": 0.27, + "learning_rate": 3.4203565350324934e-07, + "loss": 0.1412, + "step": 5195 + }, + { + "epoch": 0.27, + "learning_rate": 3.42011847459088e-07, + "loss": 0.1216, + "step": 5196 + }, + { + "epoch": 0.27, + "learning_rate": 3.419880373561731e-07, + "loss": 0.1254, + "step": 5197 + }, + { + "epoch": 0.27, + "learning_rate": 3.4196422319518534e-07, + "loss": 0.1416, + "step": 5198 + }, + { + "epoch": 0.27, + "learning_rate": 3.4194040497680526e-07, + "loss": 0.1928, + "step": 5199 + }, + { + "epoch": 0.27, + "learning_rate": 3.419165827017136e-07, + "loss": 0.1252, + "step": 5200 + }, + { + "epoch": 0.27, + "learning_rate": 3.418927563705913e-07, + "loss": 0.1445, + "step": 5201 + }, + { + "epoch": 0.27, + "learning_rate": 3.418689259841192e-07, + "loss": 0.1344, + "step": 5202 + }, + { + "epoch": 0.27, + "learning_rate": 3.4184509154297844e-07, + "loss": 0.1678, + "step": 5203 + }, + { + "epoch": 0.27, + "learning_rate": 3.4182125304785013e-07, + "loss": 0.1279, + "step": 5204 + }, + { + "epoch": 0.27, + "learning_rate": 3.417974104994158e-07, + "loss": 0.1379, + "step": 5205 + }, + { + "epoch": 0.27, + "learning_rate": 3.417735638983566e-07, + "loss": 0.1491, + "step": 5206 + }, + { + "epoch": 0.27, + "learning_rate": 3.4174971324535433e-07, + "loss": 0.1564, + "step": 5207 + }, + { + "epoch": 0.27, + "learning_rate": 3.4172585854109046e-07, + "loss": 0.132, + "step": 5208 + }, + { + "epoch": 0.27, + "learning_rate": 3.417019997862469e-07, + "loss": 0.1285, + "step": 5209 + }, + { + "epoch": 0.27, + "learning_rate": 3.416781369815054e-07, + "loss": 0.1388, + "step": 5210 + }, + { + "epoch": 0.27, + "learning_rate": 3.416542701275481e-07, + "loss": 0.126, + "step": 5211 + }, + { + "epoch": 0.27, + "learning_rate": 3.4163039922505707e-07, + "loss": 0.1678, + "step": 5212 + }, + { + "epoch": 0.27, + "learning_rate": 3.4160652427471453e-07, + "loss": 0.1594, + "step": 5213 + }, + { + "epoch": 0.27, + "learning_rate": 3.4158264527720284e-07, + "loss": 0.128, + "step": 5214 + }, + { + "epoch": 0.27, + "learning_rate": 3.415587622332046e-07, + "loss": 0.1355, + "step": 5215 + }, + { + "epoch": 0.27, + "learning_rate": 3.415348751434022e-07, + "loss": 0.171, + "step": 5216 + }, + { + "epoch": 0.27, + "learning_rate": 3.415109840084785e-07, + "loss": 0.191, + "step": 5217 + }, + { + "epoch": 0.27, + "learning_rate": 3.414870888291162e-07, + "loss": 0.1511, + "step": 5218 + }, + { + "epoch": 0.27, + "learning_rate": 3.4146318960599825e-07, + "loss": 0.1999, + "step": 5219 + }, + { + "epoch": 0.27, + "learning_rate": 3.414392863398078e-07, + "loss": 0.1535, + "step": 5220 + }, + { + "epoch": 0.27, + "learning_rate": 3.4141537903122795e-07, + "loss": 0.1364, + "step": 5221 + }, + { + "epoch": 0.27, + "learning_rate": 3.4139146768094185e-07, + "loss": 0.1782, + "step": 5222 + }, + { + "epoch": 0.27, + "learning_rate": 3.4136755228963313e-07, + "loss": 0.1484, + "step": 5223 + }, + { + "epoch": 0.27, + "learning_rate": 3.413436328579852e-07, + "loss": 0.1597, + "step": 5224 + }, + { + "epoch": 0.27, + "learning_rate": 3.413197093866816e-07, + "loss": 0.1395, + "step": 5225 + }, + { + "epoch": 0.27, + "learning_rate": 3.412957818764062e-07, + "loss": 0.1722, + "step": 5226 + }, + { + "epoch": 0.27, + "learning_rate": 3.4127185032784286e-07, + "loss": 0.1208, + "step": 5227 + }, + { + "epoch": 0.27, + "learning_rate": 3.412479147416754e-07, + "loss": 0.1283, + "step": 5228 + }, + { + "epoch": 0.27, + "learning_rate": 3.412239751185881e-07, + "loss": 0.1228, + "step": 5229 + }, + { + "epoch": 0.27, + "learning_rate": 3.4120003145926504e-07, + "loss": 0.1518, + "step": 5230 + }, + { + "epoch": 0.27, + "learning_rate": 3.4117608376439056e-07, + "loss": 0.1309, + "step": 5231 + }, + { + "epoch": 0.27, + "learning_rate": 3.411521320346491e-07, + "loss": 0.1343, + "step": 5232 + }, + { + "epoch": 0.27, + "learning_rate": 3.411281762707252e-07, + "loss": 0.1453, + "step": 5233 + }, + { + "epoch": 0.27, + "learning_rate": 3.411042164733036e-07, + "loss": 0.1276, + "step": 5234 + }, + { + "epoch": 0.27, + "learning_rate": 3.4108025264306894e-07, + "loss": 0.178, + "step": 5235 + }, + { + "epoch": 0.27, + "learning_rate": 3.410562847807063e-07, + "loss": 0.124, + "step": 5236 + }, + { + "epoch": 0.27, + "learning_rate": 3.410323128869005e-07, + "loss": 0.1216, + "step": 5237 + }, + { + "epoch": 0.27, + "learning_rate": 3.4100833696233685e-07, + "loss": 0.2179, + "step": 5238 + }, + { + "epoch": 0.27, + "learning_rate": 3.409843570077004e-07, + "loss": 0.117, + "step": 5239 + }, + { + "epoch": 0.27, + "learning_rate": 3.409603730236766e-07, + "loss": 0.1846, + "step": 5240 + }, + { + "epoch": 0.27, + "learning_rate": 3.4093638501095097e-07, + "loss": 0.1845, + "step": 5241 + }, + { + "epoch": 0.27, + "learning_rate": 3.4091239297020906e-07, + "loss": 0.136, + "step": 5242 + }, + { + "epoch": 0.27, + "learning_rate": 3.4088839690213655e-07, + "loss": 0.1547, + "step": 5243 + }, + { + "epoch": 0.27, + "learning_rate": 3.4086439680741924e-07, + "loss": 0.1191, + "step": 5244 + }, + { + "epoch": 0.27, + "learning_rate": 3.408403926867431e-07, + "loss": 0.1746, + "step": 5245 + }, + { + "epoch": 0.27, + "learning_rate": 3.408163845407942e-07, + "loss": 0.0928, + "step": 5246 + }, + { + "epoch": 0.27, + "learning_rate": 3.407923723702586e-07, + "loss": 0.1574, + "step": 5247 + }, + { + "epoch": 0.27, + "learning_rate": 3.4076835617582267e-07, + "loss": 0.1692, + "step": 5248 + }, + { + "epoch": 0.27, + "learning_rate": 3.4074433595817285e-07, + "loss": 0.1736, + "step": 5249 + }, + { + "epoch": 0.27, + "learning_rate": 3.407203117179955e-07, + "loss": 0.1203, + "step": 5250 + }, + { + "epoch": 0.27, + "learning_rate": 3.406962834559773e-07, + "loss": 0.1141, + "step": 5251 + }, + { + "epoch": 0.27, + "learning_rate": 3.406722511728051e-07, + "loss": 0.1384, + "step": 5252 + }, + { + "epoch": 0.27, + "learning_rate": 3.4064821486916555e-07, + "loss": 0.122, + "step": 5253 + }, + { + "epoch": 0.27, + "learning_rate": 3.406241745457458e-07, + "loss": 0.1549, + "step": 5254 + }, + { + "epoch": 0.27, + "learning_rate": 3.4060013020323286e-07, + "loss": 0.168, + "step": 5255 + }, + { + "epoch": 0.27, + "learning_rate": 3.405760818423139e-07, + "loss": 0.142, + "step": 5256 + }, + { + "epoch": 0.27, + "learning_rate": 3.4055202946367623e-07, + "loss": 0.1097, + "step": 5257 + }, + { + "epoch": 0.27, + "learning_rate": 3.4052797306800727e-07, + "loss": 0.1235, + "step": 5258 + }, + { + "epoch": 0.27, + "learning_rate": 3.405039126559947e-07, + "loss": 0.1481, + "step": 5259 + }, + { + "epoch": 0.27, + "learning_rate": 3.40479848228326e-07, + "loss": 0.1422, + "step": 5260 + }, + { + "epoch": 0.27, + "learning_rate": 3.40455779785689e-07, + "loss": 0.1789, + "step": 5261 + }, + { + "epoch": 0.27, + "learning_rate": 3.404317073287716e-07, + "loss": 0.1434, + "step": 5262 + }, + { + "epoch": 0.27, + "learning_rate": 3.404076308582618e-07, + "loss": 0.1266, + "step": 5263 + }, + { + "epoch": 0.27, + "learning_rate": 3.403835503748477e-07, + "loss": 0.1294, + "step": 5264 + }, + { + "epoch": 0.27, + "learning_rate": 3.4035946587921753e-07, + "loss": 0.169, + "step": 5265 + }, + { + "epoch": 0.27, + "learning_rate": 3.4033537737205965e-07, + "loss": 0.1395, + "step": 5266 + }, + { + "epoch": 0.27, + "learning_rate": 3.403112848540625e-07, + "loss": 0.1509, + "step": 5267 + }, + { + "epoch": 0.27, + "learning_rate": 3.4028718832591467e-07, + "loss": 0.1063, + "step": 5268 + }, + { + "epoch": 0.28, + "learning_rate": 3.402630877883048e-07, + "loss": 0.1837, + "step": 5269 + }, + { + "epoch": 0.28, + "learning_rate": 3.402389832419218e-07, + "loss": 0.1339, + "step": 5270 + }, + { + "epoch": 0.28, + "learning_rate": 3.402148746874545e-07, + "loss": 0.135, + "step": 5271 + }, + { + "epoch": 0.28, + "learning_rate": 3.4019076212559195e-07, + "loss": 0.1422, + "step": 5272 + }, + { + "epoch": 0.28, + "learning_rate": 3.4016664555702333e-07, + "loss": 0.1255, + "step": 5273 + }, + { + "epoch": 0.28, + "learning_rate": 3.4014252498243786e-07, + "loss": 0.1508, + "step": 5274 + }, + { + "epoch": 0.28, + "learning_rate": 3.4011840040252485e-07, + "loss": 0.1937, + "step": 5275 + }, + { + "epoch": 0.28, + "learning_rate": 3.400942718179739e-07, + "loss": 0.1737, + "step": 5276 + }, + { + "epoch": 0.28, + "learning_rate": 3.4007013922947465e-07, + "loss": 0.1445, + "step": 5277 + }, + { + "epoch": 0.28, + "learning_rate": 3.400460026377167e-07, + "loss": 0.1455, + "step": 5278 + }, + { + "epoch": 0.28, + "learning_rate": 3.400218620433899e-07, + "loss": 0.1671, + "step": 5279 + }, + { + "epoch": 0.28, + "learning_rate": 3.3999771744718427e-07, + "loss": 0.1311, + "step": 5280 + }, + { + "epoch": 0.28, + "learning_rate": 3.3997356884978985e-07, + "loss": 0.1317, + "step": 5281 + }, + { + "epoch": 0.28, + "learning_rate": 3.3994941625189677e-07, + "loss": 0.1492, + "step": 5282 + }, + { + "epoch": 0.28, + "learning_rate": 3.399252596541953e-07, + "loss": 0.1398, + "step": 5283 + }, + { + "epoch": 0.28, + "learning_rate": 3.3990109905737593e-07, + "loss": 0.149, + "step": 5284 + }, + { + "epoch": 0.28, + "learning_rate": 3.398769344621292e-07, + "loss": 0.1127, + "step": 5285 + }, + { + "epoch": 0.28, + "learning_rate": 3.3985276586914563e-07, + "loss": 0.1385, + "step": 5286 + }, + { + "epoch": 0.28, + "learning_rate": 3.39828593279116e-07, + "loss": 0.1363, + "step": 5287 + }, + { + "epoch": 0.28, + "learning_rate": 3.398044166927312e-07, + "loss": 0.1558, + "step": 5288 + }, + { + "epoch": 0.28, + "learning_rate": 3.3978023611068227e-07, + "loss": 0.12, + "step": 5289 + }, + { + "epoch": 0.28, + "learning_rate": 3.397560515336602e-07, + "loss": 0.1717, + "step": 5290 + }, + { + "epoch": 0.28, + "learning_rate": 3.397318629623562e-07, + "loss": 0.1632, + "step": 5291 + }, + { + "epoch": 0.28, + "learning_rate": 3.397076703974616e-07, + "loss": 0.1333, + "step": 5292 + }, + { + "epoch": 0.28, + "learning_rate": 3.3968347383966787e-07, + "loss": 0.1451, + "step": 5293 + }, + { + "epoch": 0.28, + "learning_rate": 3.396592732896666e-07, + "loss": 0.1769, + "step": 5294 + }, + { + "epoch": 0.28, + "learning_rate": 3.396350687481493e-07, + "loss": 0.1419, + "step": 5295 + }, + { + "epoch": 0.28, + "learning_rate": 3.396108602158078e-07, + "loss": 0.1059, + "step": 5296 + }, + { + "epoch": 0.28, + "learning_rate": 3.395866476933341e-07, + "loss": 0.1405, + "step": 5297 + }, + { + "epoch": 0.28, + "learning_rate": 3.395624311814201e-07, + "loss": 0.1255, + "step": 5298 + }, + { + "epoch": 0.28, + "learning_rate": 3.3953821068075794e-07, + "loss": 0.1284, + "step": 5299 + }, + { + "epoch": 0.28, + "learning_rate": 3.3951398619203985e-07, + "loss": 0.1384, + "step": 5300 + }, + { + "epoch": 0.28, + "learning_rate": 3.3948975771595814e-07, + "loss": 0.1535, + "step": 5301 + }, + { + "epoch": 0.28, + "learning_rate": 3.3946552525320533e-07, + "loss": 0.1637, + "step": 5302 + }, + { + "epoch": 0.28, + "learning_rate": 3.3944128880447403e-07, + "loss": 0.1262, + "step": 5303 + }, + { + "epoch": 0.28, + "learning_rate": 3.3941704837045684e-07, + "loss": 0.1275, + "step": 5304 + }, + { + "epoch": 0.28, + "learning_rate": 3.393928039518465e-07, + "loss": 0.1616, + "step": 5305 + }, + { + "epoch": 0.28, + "learning_rate": 3.393685555493361e-07, + "loss": 0.139, + "step": 5306 + }, + { + "epoch": 0.28, + "learning_rate": 3.3934430316361856e-07, + "loss": 0.1133, + "step": 5307 + }, + { + "epoch": 0.28, + "learning_rate": 3.39320046795387e-07, + "loss": 0.1236, + "step": 5308 + }, + { + "epoch": 0.28, + "learning_rate": 3.392957864453348e-07, + "loss": 0.1604, + "step": 5309 + }, + { + "epoch": 0.28, + "learning_rate": 3.392715221141552e-07, + "loss": 0.1618, + "step": 5310 + }, + { + "epoch": 0.28, + "learning_rate": 3.3924725380254176e-07, + "loss": 0.1028, + "step": 5311 + }, + { + "epoch": 0.28, + "learning_rate": 3.3922298151118804e-07, + "loss": 0.1382, + "step": 5312 + }, + { + "epoch": 0.28, + "learning_rate": 3.3919870524078773e-07, + "loss": 0.1175, + "step": 5313 + }, + { + "epoch": 0.28, + "learning_rate": 3.3917442499203476e-07, + "loss": 0.1498, + "step": 5314 + }, + { + "epoch": 0.28, + "learning_rate": 3.3915014076562296e-07, + "loss": 0.1567, + "step": 5315 + }, + { + "epoch": 0.28, + "learning_rate": 3.3912585256224637e-07, + "loss": 0.1415, + "step": 5316 + }, + { + "epoch": 0.28, + "learning_rate": 3.3910156038259923e-07, + "loss": 0.1267, + "step": 5317 + }, + { + "epoch": 0.28, + "learning_rate": 3.390772642273758e-07, + "loss": 0.1733, + "step": 5318 + }, + { + "epoch": 0.28, + "learning_rate": 3.3905296409727047e-07, + "loss": 0.1287, + "step": 5319 + }, + { + "epoch": 0.28, + "learning_rate": 3.3902865999297783e-07, + "loss": 0.124, + "step": 5320 + }, + { + "epoch": 0.28, + "learning_rate": 3.390043519151923e-07, + "loss": 0.155, + "step": 5321 + }, + { + "epoch": 0.28, + "learning_rate": 3.389800398646088e-07, + "loss": 0.1858, + "step": 5322 + }, + { + "epoch": 0.28, + "learning_rate": 3.389557238419221e-07, + "loss": 0.1506, + "step": 5323 + }, + { + "epoch": 0.28, + "learning_rate": 3.389314038478272e-07, + "loss": 0.1261, + "step": 5324 + }, + { + "epoch": 0.28, + "learning_rate": 3.3890707988301904e-07, + "loss": 0.1443, + "step": 5325 + }, + { + "epoch": 0.28, + "learning_rate": 3.3888275194819293e-07, + "loss": 0.1401, + "step": 5326 + }, + { + "epoch": 0.28, + "learning_rate": 3.388584200440443e-07, + "loss": 0.1485, + "step": 5327 + }, + { + "epoch": 0.28, + "learning_rate": 3.388340841712683e-07, + "loss": 0.1191, + "step": 5328 + }, + { + "epoch": 0.28, + "learning_rate": 3.3880974433056056e-07, + "loss": 0.1622, + "step": 5329 + }, + { + "epoch": 0.28, + "learning_rate": 3.3878540052261676e-07, + "loss": 0.1545, + "step": 5330 + }, + { + "epoch": 0.28, + "learning_rate": 3.3876105274813264e-07, + "loss": 0.2054, + "step": 5331 + }, + { + "epoch": 0.28, + "learning_rate": 3.387367010078041e-07, + "loss": 0.1604, + "step": 5332 + }, + { + "epoch": 0.28, + "learning_rate": 3.387123453023271e-07, + "loss": 0.1603, + "step": 5333 + }, + { + "epoch": 0.28, + "learning_rate": 3.386879856323977e-07, + "loss": 0.1621, + "step": 5334 + }, + { + "epoch": 0.28, + "learning_rate": 3.386636219987121e-07, + "loss": 0.1561, + "step": 5335 + }, + { + "epoch": 0.28, + "learning_rate": 3.386392544019667e-07, + "loss": 0.1745, + "step": 5336 + }, + { + "epoch": 0.28, + "learning_rate": 3.3861488284285783e-07, + "loss": 0.1469, + "step": 5337 + }, + { + "epoch": 0.28, + "learning_rate": 3.3859050732208216e-07, + "loss": 0.1271, + "step": 5338 + }, + { + "epoch": 0.28, + "learning_rate": 3.385661278403363e-07, + "loss": 0.1307, + "step": 5339 + }, + { + "epoch": 0.28, + "learning_rate": 3.3854174439831696e-07, + "loss": 0.1558, + "step": 5340 + }, + { + "epoch": 0.28, + "learning_rate": 3.385173569967211e-07, + "loss": 0.1442, + "step": 5341 + }, + { + "epoch": 0.28, + "learning_rate": 3.3849296563624575e-07, + "loss": 0.1222, + "step": 5342 + }, + { + "epoch": 0.28, + "learning_rate": 3.384685703175879e-07, + "loss": 0.2163, + "step": 5343 + }, + { + "epoch": 0.28, + "learning_rate": 3.384441710414449e-07, + "loss": 0.1473, + "step": 5344 + }, + { + "epoch": 0.28, + "learning_rate": 3.38419767808514e-07, + "loss": 0.1517, + "step": 5345 + }, + { + "epoch": 0.28, + "learning_rate": 3.3839536061949276e-07, + "loss": 0.1345, + "step": 5346 + }, + { + "epoch": 0.28, + "learning_rate": 3.3837094947507863e-07, + "loss": 0.1373, + "step": 5347 + }, + { + "epoch": 0.28, + "learning_rate": 3.3834653437596937e-07, + "loss": 0.162, + "step": 5348 + }, + { + "epoch": 0.28, + "learning_rate": 3.383221153228628e-07, + "loss": 0.1949, + "step": 5349 + }, + { + "epoch": 0.28, + "learning_rate": 3.382976923164567e-07, + "loss": 0.1328, + "step": 5350 + }, + { + "epoch": 0.28, + "learning_rate": 3.3827326535744924e-07, + "loss": 0.1565, + "step": 5351 + }, + { + "epoch": 0.28, + "learning_rate": 3.3824883444653847e-07, + "loss": 0.1589, + "step": 5352 + }, + { + "epoch": 0.28, + "learning_rate": 3.3822439958442255e-07, + "loss": 0.1446, + "step": 5353 + }, + { + "epoch": 0.28, + "learning_rate": 3.381999607718e-07, + "loss": 0.1415, + "step": 5354 + }, + { + "epoch": 0.28, + "learning_rate": 3.381755180093692e-07, + "loss": 0.1639, + "step": 5355 + }, + { + "epoch": 0.28, + "learning_rate": 3.3815107129782876e-07, + "loss": 0.1286, + "step": 5356 + }, + { + "epoch": 0.28, + "learning_rate": 3.381266206378774e-07, + "loss": 0.0996, + "step": 5357 + }, + { + "epoch": 0.28, + "learning_rate": 3.381021660302139e-07, + "loss": 0.1247, + "step": 5358 + }, + { + "epoch": 0.28, + "learning_rate": 3.380777074755372e-07, + "loss": 0.1334, + "step": 5359 + }, + { + "epoch": 0.28, + "learning_rate": 3.3805324497454624e-07, + "loss": 0.1353, + "step": 5360 + }, + { + "epoch": 0.28, + "learning_rate": 3.3802877852794026e-07, + "loss": 0.1612, + "step": 5361 + }, + { + "epoch": 0.28, + "learning_rate": 3.380043081364185e-07, + "loss": 0.1422, + "step": 5362 + }, + { + "epoch": 0.28, + "learning_rate": 3.379798338006804e-07, + "loss": 0.1395, + "step": 5363 + }, + { + "epoch": 0.28, + "learning_rate": 3.3795535552142536e-07, + "loss": 0.1882, + "step": 5364 + }, + { + "epoch": 0.28, + "learning_rate": 3.37930873299353e-07, + "loss": 0.1492, + "step": 5365 + }, + { + "epoch": 0.28, + "learning_rate": 3.37906387135163e-07, + "loss": 0.1545, + "step": 5366 + }, + { + "epoch": 0.28, + "learning_rate": 3.378818970295553e-07, + "loss": 0.1203, + "step": 5367 + }, + { + "epoch": 0.28, + "learning_rate": 3.378574029832297e-07, + "loss": 0.167, + "step": 5368 + }, + { + "epoch": 0.28, + "learning_rate": 3.3783290499688626e-07, + "loss": 0.1526, + "step": 5369 + }, + { + "epoch": 0.28, + "learning_rate": 3.3780840307122526e-07, + "loss": 0.1586, + "step": 5370 + }, + { + "epoch": 0.28, + "learning_rate": 3.377838972069469e-07, + "loss": 0.1436, + "step": 5371 + }, + { + "epoch": 0.28, + "learning_rate": 3.377593874047515e-07, + "loss": 0.1377, + "step": 5372 + }, + { + "epoch": 0.28, + "learning_rate": 3.3773487366533966e-07, + "loss": 0.1284, + "step": 5373 + }, + { + "epoch": 0.28, + "learning_rate": 3.37710355989412e-07, + "loss": 0.1854, + "step": 5374 + }, + { + "epoch": 0.28, + "learning_rate": 3.3768583437766915e-07, + "loss": 0.1084, + "step": 5375 + }, + { + "epoch": 0.28, + "learning_rate": 3.3766130883081206e-07, + "loss": 0.1476, + "step": 5376 + }, + { + "epoch": 0.28, + "learning_rate": 3.3763677934954155e-07, + "loss": 0.1279, + "step": 5377 + }, + { + "epoch": 0.28, + "learning_rate": 3.376122459345588e-07, + "loss": 0.1861, + "step": 5378 + }, + { + "epoch": 0.28, + "learning_rate": 3.375877085865649e-07, + "loss": 0.1398, + "step": 5379 + }, + { + "epoch": 0.28, + "learning_rate": 3.375631673062612e-07, + "loss": 0.1265, + "step": 5380 + }, + { + "epoch": 0.28, + "learning_rate": 3.3753862209434906e-07, + "loss": 0.1262, + "step": 5381 + }, + { + "epoch": 0.28, + "learning_rate": 3.3751407295153004e-07, + "loss": 0.1621, + "step": 5382 + }, + { + "epoch": 0.28, + "learning_rate": 3.374895198785057e-07, + "loss": 0.1326, + "step": 5383 + }, + { + "epoch": 0.28, + "learning_rate": 3.374649628759778e-07, + "loss": 0.161, + "step": 5384 + }, + { + "epoch": 0.28, + "learning_rate": 3.3744040194464826e-07, + "loss": 0.1556, + "step": 5385 + }, + { + "epoch": 0.28, + "learning_rate": 3.374158370852189e-07, + "loss": 0.1311, + "step": 5386 + }, + { + "epoch": 0.28, + "learning_rate": 3.373912682983919e-07, + "loss": 0.133, + "step": 5387 + }, + { + "epoch": 0.28, + "learning_rate": 3.373666955848694e-07, + "loss": 0.1552, + "step": 5388 + }, + { + "epoch": 0.28, + "learning_rate": 3.3734211894535376e-07, + "loss": 0.1306, + "step": 5389 + }, + { + "epoch": 0.28, + "learning_rate": 3.3731753838054725e-07, + "loss": 0.1303, + "step": 5390 + }, + { + "epoch": 0.28, + "learning_rate": 3.372929538911525e-07, + "loss": 0.1861, + "step": 5391 + }, + { + "epoch": 0.28, + "learning_rate": 3.372683654778722e-07, + "loss": 0.1584, + "step": 5392 + }, + { + "epoch": 0.28, + "learning_rate": 3.3724377314140897e-07, + "loss": 0.1245, + "step": 5393 + }, + { + "epoch": 0.28, + "learning_rate": 3.372191768824657e-07, + "loss": 0.1068, + "step": 5394 + }, + { + "epoch": 0.28, + "learning_rate": 3.3719457670174546e-07, + "loss": 0.1429, + "step": 5395 + }, + { + "epoch": 0.28, + "learning_rate": 3.371699725999512e-07, + "loss": 0.1814, + "step": 5396 + }, + { + "epoch": 0.28, + "learning_rate": 3.371453645777862e-07, + "loss": 0.1422, + "step": 5397 + }, + { + "epoch": 0.28, + "learning_rate": 3.3712075263595373e-07, + "loss": 0.1639, + "step": 5398 + }, + { + "epoch": 0.28, + "learning_rate": 3.3709613677515715e-07, + "loss": 0.1026, + "step": 5399 + }, + { + "epoch": 0.28, + "learning_rate": 3.3707151699610014e-07, + "loss": 0.1596, + "step": 5400 + }, + { + "epoch": 0.28, + "learning_rate": 3.3704689329948626e-07, + "loss": 0.1448, + "step": 5401 + }, + { + "epoch": 0.28, + "learning_rate": 3.370222656860192e-07, + "loss": 0.1425, + "step": 5402 + }, + { + "epoch": 0.28, + "learning_rate": 3.3699763415640295e-07, + "loss": 0.1749, + "step": 5403 + }, + { + "epoch": 0.28, + "learning_rate": 3.3697299871134134e-07, + "loss": 0.1403, + "step": 5404 + }, + { + "epoch": 0.28, + "learning_rate": 3.3694835935153867e-07, + "loss": 0.149, + "step": 5405 + }, + { + "epoch": 0.28, + "learning_rate": 3.36923716077699e-07, + "loss": 0.132, + "step": 5406 + }, + { + "epoch": 0.28, + "learning_rate": 3.368990688905266e-07, + "loss": 0.1953, + "step": 5407 + }, + { + "epoch": 0.28, + "learning_rate": 3.3687441779072596e-07, + "loss": 0.1299, + "step": 5408 + }, + { + "epoch": 0.28, + "learning_rate": 3.368497627790017e-07, + "loss": 0.1469, + "step": 5409 + }, + { + "epoch": 0.28, + "learning_rate": 3.3682510385605835e-07, + "loss": 0.1488, + "step": 5410 + }, + { + "epoch": 0.28, + "learning_rate": 3.368004410226007e-07, + "loss": 0.1203, + "step": 5411 + }, + { + "epoch": 0.28, + "learning_rate": 3.367757742793336e-07, + "loss": 0.141, + "step": 5412 + }, + { + "epoch": 0.28, + "learning_rate": 3.367511036269621e-07, + "loss": 0.13, + "step": 5413 + }, + { + "epoch": 0.28, + "learning_rate": 3.367264290661914e-07, + "loss": 0.1804, + "step": 5414 + }, + { + "epoch": 0.28, + "learning_rate": 3.367017505977264e-07, + "loss": 0.1263, + "step": 5415 + }, + { + "epoch": 0.28, + "learning_rate": 3.366770682222727e-07, + "loss": 0.1118, + "step": 5416 + }, + { + "epoch": 0.28, + "learning_rate": 3.3665238194053553e-07, + "loss": 0.1369, + "step": 5417 + }, + { + "epoch": 0.28, + "learning_rate": 3.3662769175322056e-07, + "loss": 0.1511, + "step": 5418 + }, + { + "epoch": 0.28, + "learning_rate": 3.3660299766103346e-07, + "loss": 0.1575, + "step": 5419 + }, + { + "epoch": 0.28, + "learning_rate": 3.365782996646799e-07, + "loss": 0.1463, + "step": 5420 + }, + { + "epoch": 0.28, + "learning_rate": 3.365535977648659e-07, + "loss": 0.1463, + "step": 5421 + }, + { + "epoch": 0.28, + "learning_rate": 3.3652889196229725e-07, + "loss": 0.1485, + "step": 5422 + }, + { + "epoch": 0.28, + "learning_rate": 3.365041822576803e-07, + "loss": 0.1532, + "step": 5423 + }, + { + "epoch": 0.28, + "learning_rate": 3.36479468651721e-07, + "loss": 0.1495, + "step": 5424 + }, + { + "epoch": 0.28, + "learning_rate": 3.364547511451258e-07, + "loss": 0.1631, + "step": 5425 + }, + { + "epoch": 0.28, + "learning_rate": 3.364300297386012e-07, + "loss": 0.1415, + "step": 5426 + }, + { + "epoch": 0.28, + "learning_rate": 3.3640530443285363e-07, + "loss": 0.1845, + "step": 5427 + }, + { + "epoch": 0.28, + "learning_rate": 3.3638057522858986e-07, + "loss": 0.1684, + "step": 5428 + }, + { + "epoch": 0.28, + "learning_rate": 3.363558421265166e-07, + "loss": 0.1655, + "step": 5429 + }, + { + "epoch": 0.28, + "learning_rate": 3.363311051273407e-07, + "loss": 0.1222, + "step": 5430 + }, + { + "epoch": 0.28, + "learning_rate": 3.3630636423176914e-07, + "loss": 0.1236, + "step": 5431 + }, + { + "epoch": 0.28, + "learning_rate": 3.3628161944050916e-07, + "loss": 0.218, + "step": 5432 + }, + { + "epoch": 0.28, + "learning_rate": 3.3625687075426785e-07, + "loss": 0.1469, + "step": 5433 + }, + { + "epoch": 0.28, + "learning_rate": 3.362321181737526e-07, + "loss": 0.1313, + "step": 5434 + }, + { + "epoch": 0.28, + "learning_rate": 3.3620736169967073e-07, + "loss": 0.1427, + "step": 5435 + }, + { + "epoch": 0.28, + "learning_rate": 3.3618260133273e-07, + "loss": 0.1435, + "step": 5436 + }, + { + "epoch": 0.28, + "learning_rate": 3.3615783707363787e-07, + "loss": 0.1242, + "step": 5437 + }, + { + "epoch": 0.28, + "learning_rate": 3.361330689231022e-07, + "loss": 0.1252, + "step": 5438 + }, + { + "epoch": 0.28, + "learning_rate": 3.3610829688183094e-07, + "loss": 0.1042, + "step": 5439 + }, + { + "epoch": 0.28, + "learning_rate": 3.3608352095053194e-07, + "loss": 0.213, + "step": 5440 + }, + { + "epoch": 0.28, + "learning_rate": 3.360587411299134e-07, + "loss": 0.148, + "step": 5441 + }, + { + "epoch": 0.28, + "learning_rate": 3.360339574206835e-07, + "loss": 0.122, + "step": 5442 + }, + { + "epoch": 0.28, + "learning_rate": 3.360091698235506e-07, + "loss": 0.1211, + "step": 5443 + }, + { + "epoch": 0.28, + "learning_rate": 3.3598437833922316e-07, + "loss": 0.1273, + "step": 5444 + }, + { + "epoch": 0.28, + "learning_rate": 3.359595829684096e-07, + "loss": 0.1628, + "step": 5445 + }, + { + "epoch": 0.28, + "learning_rate": 3.3593478371181875e-07, + "loss": 0.164, + "step": 5446 + }, + { + "epoch": 0.28, + "learning_rate": 3.359099805701593e-07, + "loss": 0.1337, + "step": 5447 + }, + { + "epoch": 0.28, + "learning_rate": 3.358851735441401e-07, + "loss": 0.1537, + "step": 5448 + }, + { + "epoch": 0.28, + "learning_rate": 3.3586036263447017e-07, + "loss": 0.1301, + "step": 5449 + }, + { + "epoch": 0.28, + "learning_rate": 3.358355478418587e-07, + "loss": 0.1583, + "step": 5450 + }, + { + "epoch": 0.28, + "learning_rate": 3.3581072916701477e-07, + "loss": 0.1345, + "step": 5451 + }, + { + "epoch": 0.28, + "learning_rate": 3.357859066106478e-07, + "loss": 0.1368, + "step": 5452 + }, + { + "epoch": 0.28, + "learning_rate": 3.357610801734672e-07, + "loss": 0.1103, + "step": 5453 + }, + { + "epoch": 0.28, + "learning_rate": 3.357362498561825e-07, + "loss": 0.1527, + "step": 5454 + }, + { + "epoch": 0.28, + "learning_rate": 3.357114156595034e-07, + "loss": 0.1662, + "step": 5455 + }, + { + "epoch": 0.28, + "learning_rate": 3.3568657758413963e-07, + "loss": 0.1092, + "step": 5456 + }, + { + "epoch": 0.28, + "learning_rate": 3.3566173563080113e-07, + "loss": 0.1424, + "step": 5457 + }, + { + "epoch": 0.28, + "learning_rate": 3.3563688980019785e-07, + "loss": 0.104, + "step": 5458 + }, + { + "epoch": 0.28, + "learning_rate": 3.3561204009303986e-07, + "loss": 0.1782, + "step": 5459 + }, + { + "epoch": 0.28, + "learning_rate": 3.355871865100374e-07, + "loss": 0.1958, + "step": 5460 + }, + { + "epoch": 0.29, + "learning_rate": 3.3556232905190094e-07, + "loss": 0.1405, + "step": 5461 + }, + { + "epoch": 0.29, + "learning_rate": 3.3553746771934063e-07, + "loss": 0.1142, + "step": 5462 + }, + { + "epoch": 0.29, + "learning_rate": 3.355126025130672e-07, + "loss": 0.1397, + "step": 5463 + }, + { + "epoch": 0.29, + "learning_rate": 3.354877334337913e-07, + "loss": 0.1766, + "step": 5464 + }, + { + "epoch": 0.29, + "learning_rate": 3.354628604822237e-07, + "loss": 0.1236, + "step": 5465 + }, + { + "epoch": 0.29, + "learning_rate": 3.354379836590752e-07, + "loss": 0.1601, + "step": 5466 + }, + { + "epoch": 0.29, + "learning_rate": 3.3541310296505686e-07, + "loss": 0.1912, + "step": 5467 + }, + { + "epoch": 0.29, + "learning_rate": 3.3538821840087976e-07, + "loss": 0.1271, + "step": 5468 + }, + { + "epoch": 0.29, + "learning_rate": 3.353633299672552e-07, + "loss": 0.1689, + "step": 5469 + }, + { + "epoch": 0.29, + "learning_rate": 3.353384376648943e-07, + "loss": 0.1449, + "step": 5470 + }, + { + "epoch": 0.29, + "learning_rate": 3.353135414945087e-07, + "loss": 0.1448, + "step": 5471 + }, + { + "epoch": 0.29, + "learning_rate": 3.3528864145680977e-07, + "loss": 0.1217, + "step": 5472 + }, + { + "epoch": 0.29, + "learning_rate": 3.352637375525092e-07, + "loss": 0.166, + "step": 5473 + }, + { + "epoch": 0.29, + "learning_rate": 3.3523882978231896e-07, + "loss": 0.1599, + "step": 5474 + }, + { + "epoch": 0.29, + "learning_rate": 3.352139181469506e-07, + "loss": 0.1336, + "step": 5475 + }, + { + "epoch": 0.29, + "learning_rate": 3.3518900264711634e-07, + "loss": 0.1744, + "step": 5476 + }, + { + "epoch": 0.29, + "learning_rate": 3.3516408328352817e-07, + "loss": 0.163, + "step": 5477 + }, + { + "epoch": 0.29, + "learning_rate": 3.351391600568983e-07, + "loss": 0.1541, + "step": 5478 + }, + { + "epoch": 0.29, + "learning_rate": 3.3511423296793916e-07, + "loss": 0.141, + "step": 5479 + }, + { + "epoch": 0.29, + "learning_rate": 3.35089302017363e-07, + "loss": 0.1579, + "step": 5480 + }, + { + "epoch": 0.29, + "learning_rate": 3.350643672058824e-07, + "loss": 0.1528, + "step": 5481 + }, + { + "epoch": 0.29, + "learning_rate": 3.3503942853421013e-07, + "loss": 0.1267, + "step": 5482 + }, + { + "epoch": 0.29, + "learning_rate": 3.350144860030588e-07, + "loss": 0.1226, + "step": 5483 + }, + { + "epoch": 0.29, + "learning_rate": 3.349895396131414e-07, + "loss": 0.1157, + "step": 5484 + }, + { + "epoch": 0.29, + "learning_rate": 3.3496458936517083e-07, + "loss": 0.099, + "step": 5485 + }, + { + "epoch": 0.29, + "learning_rate": 3.349396352598602e-07, + "loss": 0.1376, + "step": 5486 + }, + { + "epoch": 0.29, + "learning_rate": 3.3491467729792266e-07, + "loss": 0.1336, + "step": 5487 + }, + { + "epoch": 0.29, + "learning_rate": 3.348897154800716e-07, + "loss": 0.1396, + "step": 5488 + }, + { + "epoch": 0.29, + "learning_rate": 3.3486474980702037e-07, + "loss": 0.1469, + "step": 5489 + }, + { + "epoch": 0.29, + "learning_rate": 3.3483978027948254e-07, + "loss": 0.153, + "step": 5490 + }, + { + "epoch": 0.29, + "learning_rate": 3.348148068981717e-07, + "loss": 0.1564, + "step": 5491 + }, + { + "epoch": 0.29, + "learning_rate": 3.347898296638017e-07, + "loss": 0.1311, + "step": 5492 + }, + { + "epoch": 0.29, + "learning_rate": 3.3476484857708627e-07, + "loss": 0.1789, + "step": 5493 + }, + { + "epoch": 0.29, + "learning_rate": 3.347398636387395e-07, + "loss": 0.1956, + "step": 5494 + }, + { + "epoch": 0.29, + "learning_rate": 3.347148748494755e-07, + "loss": 0.1462, + "step": 5495 + }, + { + "epoch": 0.29, + "learning_rate": 3.346898822100082e-07, + "loss": 0.1528, + "step": 5496 + }, + { + "epoch": 0.29, + "learning_rate": 3.346648857210521e-07, + "loss": 0.136, + "step": 5497 + }, + { + "epoch": 0.29, + "learning_rate": 3.3463988538332163e-07, + "loss": 0.1386, + "step": 5498 + }, + { + "epoch": 0.29, + "learning_rate": 3.3461488119753123e-07, + "loss": 0.134, + "step": 5499 + }, + { + "epoch": 0.29, + "learning_rate": 3.345898731643956e-07, + "loss": 0.1398, + "step": 5500 + }, + { + "epoch": 0.29, + "learning_rate": 3.3456486128462944e-07, + "loss": 0.1677, + "step": 5501 + }, + { + "epoch": 0.29, + "learning_rate": 3.3453984555894756e-07, + "loss": 0.1536, + "step": 5502 + }, + { + "epoch": 0.29, + "learning_rate": 3.345148259880649e-07, + "loss": 0.1141, + "step": 5503 + }, + { + "epoch": 0.29, + "learning_rate": 3.3448980257269665e-07, + "loss": 0.1267, + "step": 5504 + }, + { + "epoch": 0.29, + "learning_rate": 3.34464775313558e-07, + "loss": 0.1506, + "step": 5505 + }, + { + "epoch": 0.29, + "learning_rate": 3.3443974421136404e-07, + "loss": 0.1347, + "step": 5506 + }, + { + "epoch": 0.29, + "learning_rate": 3.3441470926683033e-07, + "loss": 0.1617, + "step": 5507 + }, + { + "epoch": 0.29, + "learning_rate": 3.343896704806724e-07, + "loss": 0.1553, + "step": 5508 + }, + { + "epoch": 0.29, + "learning_rate": 3.3436462785360575e-07, + "loss": 0.1418, + "step": 5509 + }, + { + "epoch": 0.29, + "learning_rate": 3.343395813863462e-07, + "loss": 0.136, + "step": 5510 + }, + { + "epoch": 0.29, + "learning_rate": 3.3431453107960946e-07, + "loss": 0.131, + "step": 5511 + }, + { + "epoch": 0.29, + "learning_rate": 3.342894769341116e-07, + "loss": 0.1301, + "step": 5512 + }, + { + "epoch": 0.29, + "learning_rate": 3.3426441895056876e-07, + "loss": 0.1449, + "step": 5513 + }, + { + "epoch": 0.29, + "learning_rate": 3.342393571296969e-07, + "loss": 0.1181, + "step": 5514 + }, + { + "epoch": 0.29, + "learning_rate": 3.342142914722124e-07, + "loss": 0.1335, + "step": 5515 + }, + { + "epoch": 0.29, + "learning_rate": 3.341892219788317e-07, + "loss": 0.1393, + "step": 5516 + }, + { + "epoch": 0.29, + "learning_rate": 3.341641486502712e-07, + "loss": 0.1379, + "step": 5517 + }, + { + "epoch": 0.29, + "learning_rate": 3.341390714872475e-07, + "loss": 0.1339, + "step": 5518 + }, + { + "epoch": 0.29, + "learning_rate": 3.3411399049047733e-07, + "loss": 0.1432, + "step": 5519 + }, + { + "epoch": 0.29, + "learning_rate": 3.340889056606776e-07, + "loss": 0.1358, + "step": 5520 + }, + { + "epoch": 0.29, + "learning_rate": 3.3406381699856517e-07, + "loss": 0.1456, + "step": 5521 + }, + { + "epoch": 0.29, + "learning_rate": 3.340387245048571e-07, + "loss": 0.1059, + "step": 5522 + }, + { + "epoch": 0.29, + "learning_rate": 3.340136281802705e-07, + "loss": 0.1438, + "step": 5523 + }, + { + "epoch": 0.29, + "learning_rate": 3.3398852802552274e-07, + "loss": 0.1407, + "step": 5524 + }, + { + "epoch": 0.29, + "learning_rate": 3.339634240413311e-07, + "loss": 0.1606, + "step": 5525 + }, + { + "epoch": 0.29, + "learning_rate": 3.3393831622841304e-07, + "loss": 0.1551, + "step": 5526 + }, + { + "epoch": 0.29, + "learning_rate": 3.3391320458748626e-07, + "loss": 0.1505, + "step": 5527 + }, + { + "epoch": 0.29, + "learning_rate": 3.3388808911926836e-07, + "loss": 0.1346, + "step": 5528 + }, + { + "epoch": 0.29, + "learning_rate": 3.338629698244772e-07, + "loss": 0.126, + "step": 5529 + }, + { + "epoch": 0.29, + "learning_rate": 3.338378467038307e-07, + "loss": 0.118, + "step": 5530 + }, + { + "epoch": 0.29, + "learning_rate": 3.3381271975804685e-07, + "loss": 0.1434, + "step": 5531 + }, + { + "epoch": 0.29, + "learning_rate": 3.3378758898784377e-07, + "loss": 0.1282, + "step": 5532 + }, + { + "epoch": 0.29, + "learning_rate": 3.3376245439393983e-07, + "loss": 0.1597, + "step": 5533 + }, + { + "epoch": 0.29, + "learning_rate": 3.337373159770533e-07, + "loss": 0.1138, + "step": 5534 + }, + { + "epoch": 0.29, + "learning_rate": 3.3371217373790263e-07, + "loss": 0.1412, + "step": 5535 + }, + { + "epoch": 0.29, + "learning_rate": 3.336870276772064e-07, + "loss": 0.1396, + "step": 5536 + }, + { + "epoch": 0.29, + "learning_rate": 3.3366187779568334e-07, + "loss": 0.1329, + "step": 5537 + }, + { + "epoch": 0.29, + "learning_rate": 3.3363672409405215e-07, + "loss": 0.1751, + "step": 5538 + }, + { + "epoch": 0.29, + "learning_rate": 3.336115665730319e-07, + "loss": 0.1882, + "step": 5539 + }, + { + "epoch": 0.29, + "learning_rate": 3.335864052333414e-07, + "loss": 0.1778, + "step": 5540 + }, + { + "epoch": 0.29, + "learning_rate": 3.3356124007569995e-07, + "loss": 0.1409, + "step": 5541 + }, + { + "epoch": 0.29, + "learning_rate": 3.335360711008266e-07, + "loss": 0.144, + "step": 5542 + }, + { + "epoch": 0.29, + "learning_rate": 3.3351089830944094e-07, + "loss": 0.1879, + "step": 5543 + }, + { + "epoch": 0.29, + "learning_rate": 3.3348572170226215e-07, + "loss": 0.1336, + "step": 5544 + }, + { + "epoch": 0.29, + "learning_rate": 3.334605412800099e-07, + "loss": 0.1435, + "step": 5545 + }, + { + "epoch": 0.29, + "learning_rate": 3.3343535704340387e-07, + "loss": 0.1339, + "step": 5546 + }, + { + "epoch": 0.29, + "learning_rate": 3.3341016899316386e-07, + "loss": 0.1755, + "step": 5547 + }, + { + "epoch": 0.29, + "learning_rate": 3.3338497713000975e-07, + "loss": 0.1382, + "step": 5548 + }, + { + "epoch": 0.29, + "learning_rate": 3.333597814546614e-07, + "loss": 0.1349, + "step": 5549 + }, + { + "epoch": 0.29, + "learning_rate": 3.3333458196783904e-07, + "loss": 0.158, + "step": 5550 + }, + { + "epoch": 0.29, + "learning_rate": 3.333093786702629e-07, + "loss": 0.1237, + "step": 5551 + }, + { + "epoch": 0.29, + "learning_rate": 3.3328417156265325e-07, + "loss": 0.156, + "step": 5552 + }, + { + "epoch": 0.29, + "learning_rate": 3.3325896064573053e-07, + "loss": 0.1048, + "step": 5553 + }, + { + "epoch": 0.29, + "learning_rate": 3.332337459202152e-07, + "loss": 0.1179, + "step": 5554 + }, + { + "epoch": 0.29, + "learning_rate": 3.332085273868281e-07, + "loss": 0.1412, + "step": 5555 + }, + { + "epoch": 0.29, + "learning_rate": 3.3318330504628976e-07, + "loss": 0.1354, + "step": 5556 + }, + { + "epoch": 0.29, + "learning_rate": 3.3315807889932115e-07, + "loss": 0.1393, + "step": 5557 + }, + { + "epoch": 0.29, + "learning_rate": 3.331328489466433e-07, + "loss": 0.1409, + "step": 5558 + }, + { + "epoch": 0.29, + "learning_rate": 3.331076151889772e-07, + "loss": 0.1312, + "step": 5559 + }, + { + "epoch": 0.29, + "learning_rate": 3.330823776270441e-07, + "loss": 0.1675, + "step": 5560 + }, + { + "epoch": 0.29, + "learning_rate": 3.3305713626156524e-07, + "loss": 0.1715, + "step": 5561 + }, + { + "epoch": 0.29, + "learning_rate": 3.330318910932621e-07, + "loss": 0.1161, + "step": 5562 + }, + { + "epoch": 0.29, + "learning_rate": 3.330066421228561e-07, + "loss": 0.1762, + "step": 5563 + }, + { + "epoch": 0.29, + "learning_rate": 3.3298138935106897e-07, + "loss": 0.1356, + "step": 5564 + }, + { + "epoch": 0.29, + "learning_rate": 3.3295613277862237e-07, + "loss": 0.1158, + "step": 5565 + }, + { + "epoch": 0.29, + "learning_rate": 3.3293087240623816e-07, + "loss": 0.1395, + "step": 5566 + }, + { + "epoch": 0.29, + "learning_rate": 3.3290560823463833e-07, + "loss": 0.1394, + "step": 5567 + }, + { + "epoch": 0.29, + "learning_rate": 3.328803402645449e-07, + "loss": 0.1613, + "step": 5568 + }, + { + "epoch": 0.29, + "learning_rate": 3.3285506849668004e-07, + "loss": 0.1106, + "step": 5569 + }, + { + "epoch": 0.29, + "learning_rate": 3.3282979293176603e-07, + "loss": 0.1294, + "step": 5570 + }, + { + "epoch": 0.29, + "learning_rate": 3.3280451357052526e-07, + "loss": 0.124, + "step": 5571 + }, + { + "epoch": 0.29, + "learning_rate": 3.3277923041368026e-07, + "loss": 0.1614, + "step": 5572 + }, + { + "epoch": 0.29, + "learning_rate": 3.327539434619536e-07, + "loss": 0.1591, + "step": 5573 + }, + { + "epoch": 0.29, + "learning_rate": 3.327286527160679e-07, + "loss": 0.1129, + "step": 5574 + }, + { + "epoch": 0.29, + "learning_rate": 3.3270335817674614e-07, + "loss": 0.1386, + "step": 5575 + }, + { + "epoch": 0.29, + "learning_rate": 3.3267805984471116e-07, + "loss": 0.1331, + "step": 5576 + }, + { + "epoch": 0.29, + "learning_rate": 3.32652757720686e-07, + "loss": 0.1539, + "step": 5577 + }, + { + "epoch": 0.29, + "learning_rate": 3.326274518053938e-07, + "loss": 0.1042, + "step": 5578 + }, + { + "epoch": 0.29, + "learning_rate": 3.3260214209955786e-07, + "loss": 0.1248, + "step": 5579 + }, + { + "epoch": 0.29, + "learning_rate": 3.325768286039015e-07, + "loss": 0.1853, + "step": 5580 + }, + { + "epoch": 0.29, + "learning_rate": 3.325515113191482e-07, + "loss": 0.1343, + "step": 5581 + }, + { + "epoch": 0.29, + "learning_rate": 3.325261902460216e-07, + "loss": 0.1445, + "step": 5582 + }, + { + "epoch": 0.29, + "learning_rate": 3.3250086538524526e-07, + "loss": 0.1432, + "step": 5583 + }, + { + "epoch": 0.29, + "learning_rate": 3.32475536737543e-07, + "loss": 0.1381, + "step": 5584 + }, + { + "epoch": 0.29, + "learning_rate": 3.3245020430363885e-07, + "loss": 0.104, + "step": 5585 + }, + { + "epoch": 0.29, + "learning_rate": 3.3242486808425663e-07, + "loss": 0.1342, + "step": 5586 + }, + { + "epoch": 0.29, + "learning_rate": 3.323995280801206e-07, + "loss": 0.1544, + "step": 5587 + }, + { + "epoch": 0.29, + "learning_rate": 3.3237418429195493e-07, + "loss": 0.1488, + "step": 5588 + }, + { + "epoch": 0.29, + "learning_rate": 3.32348836720484e-07, + "loss": 0.1528, + "step": 5589 + }, + { + "epoch": 0.29, + "learning_rate": 3.3232348536643227e-07, + "loss": 0.1332, + "step": 5590 + }, + { + "epoch": 0.29, + "learning_rate": 3.3229813023052417e-07, + "loss": 0.1548, + "step": 5591 + }, + { + "epoch": 0.29, + "learning_rate": 3.322727713134845e-07, + "loss": 0.113, + "step": 5592 + }, + { + "epoch": 0.29, + "learning_rate": 3.322474086160379e-07, + "loss": 0.1326, + "step": 5593 + }, + { + "epoch": 0.29, + "learning_rate": 3.322220421389093e-07, + "loss": 0.1362, + "step": 5594 + }, + { + "epoch": 0.29, + "learning_rate": 3.3219667188282376e-07, + "loss": 0.1234, + "step": 5595 + }, + { + "epoch": 0.29, + "learning_rate": 3.3217129784850624e-07, + "loss": 0.1231, + "step": 5596 + }, + { + "epoch": 0.29, + "learning_rate": 3.32145920036682e-07, + "loss": 0.1497, + "step": 5597 + }, + { + "epoch": 0.29, + "learning_rate": 3.3212053844807644e-07, + "loss": 0.1389, + "step": 5598 + }, + { + "epoch": 0.29, + "learning_rate": 3.320951530834148e-07, + "loss": 0.1187, + "step": 5599 + }, + { + "epoch": 0.29, + "learning_rate": 3.320697639434227e-07, + "loss": 0.1398, + "step": 5600 + }, + { + "epoch": 0.29, + "learning_rate": 3.3204437102882583e-07, + "loss": 0.1722, + "step": 5601 + }, + { + "epoch": 0.29, + "learning_rate": 3.320189743403498e-07, + "loss": 0.1206, + "step": 5602 + }, + { + "epoch": 0.29, + "learning_rate": 3.3199357387872054e-07, + "loss": 0.1557, + "step": 5603 + }, + { + "epoch": 0.29, + "learning_rate": 3.3196816964466397e-07, + "loss": 0.203, + "step": 5604 + }, + { + "epoch": 0.29, + "learning_rate": 3.319427616389062e-07, + "loss": 0.134, + "step": 5605 + }, + { + "epoch": 0.29, + "learning_rate": 3.3191734986217333e-07, + "loss": 0.1568, + "step": 5606 + }, + { + "epoch": 0.29, + "learning_rate": 3.318919343151918e-07, + "loss": 0.1028, + "step": 5607 + }, + { + "epoch": 0.29, + "learning_rate": 3.318665149986877e-07, + "loss": 0.1291, + "step": 5608 + }, + { + "epoch": 0.29, + "learning_rate": 3.3184109191338773e-07, + "loss": 0.1395, + "step": 5609 + }, + { + "epoch": 0.29, + "learning_rate": 3.3181566506001854e-07, + "loss": 0.133, + "step": 5610 + }, + { + "epoch": 0.29, + "learning_rate": 3.317902344393067e-07, + "loss": 0.1478, + "step": 5611 + }, + { + "epoch": 0.29, + "learning_rate": 3.31764800051979e-07, + "loss": 0.1162, + "step": 5612 + }, + { + "epoch": 0.29, + "learning_rate": 3.317393618987626e-07, + "loss": 0.1398, + "step": 5613 + }, + { + "epoch": 0.29, + "learning_rate": 3.3171391998038433e-07, + "loss": 0.1447, + "step": 5614 + }, + { + "epoch": 0.29, + "learning_rate": 3.3168847429757135e-07, + "loss": 0.1443, + "step": 5615 + }, + { + "epoch": 0.29, + "learning_rate": 3.3166302485105096e-07, + "loss": 0.1409, + "step": 5616 + }, + { + "epoch": 0.29, + "learning_rate": 3.316375716415505e-07, + "loss": 0.1261, + "step": 5617 + }, + { + "epoch": 0.29, + "learning_rate": 3.3161211466979737e-07, + "loss": 0.1463, + "step": 5618 + }, + { + "epoch": 0.29, + "learning_rate": 3.3158665393651923e-07, + "loss": 0.1801, + "step": 5619 + }, + { + "epoch": 0.29, + "learning_rate": 3.315611894424438e-07, + "loss": 0.1239, + "step": 5620 + }, + { + "epoch": 0.29, + "learning_rate": 3.315357211882987e-07, + "loss": 0.1479, + "step": 5621 + }, + { + "epoch": 0.29, + "learning_rate": 3.315102491748119e-07, + "loss": 0.1506, + "step": 5622 + }, + { + "epoch": 0.29, + "learning_rate": 3.3148477340271144e-07, + "loss": 0.114, + "step": 5623 + }, + { + "epoch": 0.29, + "learning_rate": 3.314592938727254e-07, + "loss": 0.1345, + "step": 5624 + }, + { + "epoch": 0.29, + "learning_rate": 3.3143381058558204e-07, + "loss": 0.112, + "step": 5625 + }, + { + "epoch": 0.29, + "learning_rate": 3.314083235420095e-07, + "loss": 0.1623, + "step": 5626 + }, + { + "epoch": 0.29, + "learning_rate": 3.313828327427364e-07, + "loss": 0.1707, + "step": 5627 + }, + { + "epoch": 0.29, + "learning_rate": 3.313573381884913e-07, + "loss": 0.1424, + "step": 5628 + }, + { + "epoch": 0.29, + "learning_rate": 3.3133183988000274e-07, + "loss": 0.1608, + "step": 5629 + }, + { + "epoch": 0.29, + "learning_rate": 3.3130633781799943e-07, + "loss": 0.1441, + "step": 5630 + }, + { + "epoch": 0.29, + "learning_rate": 3.312808320032104e-07, + "loss": 0.1687, + "step": 5631 + }, + { + "epoch": 0.29, + "learning_rate": 3.312553224363644e-07, + "loss": 0.1202, + "step": 5632 + }, + { + "epoch": 0.29, + "learning_rate": 3.312298091181907e-07, + "loss": 0.1309, + "step": 5633 + }, + { + "epoch": 0.29, + "learning_rate": 3.312042920494184e-07, + "loss": 0.1527, + "step": 5634 + }, + { + "epoch": 0.29, + "learning_rate": 3.311787712307767e-07, + "loss": 0.153, + "step": 5635 + }, + { + "epoch": 0.29, + "learning_rate": 3.311532466629951e-07, + "loss": 0.1111, + "step": 5636 + }, + { + "epoch": 0.29, + "learning_rate": 3.3112771834680314e-07, + "loss": 0.1462, + "step": 5637 + }, + { + "epoch": 0.29, + "learning_rate": 3.311021862829303e-07, + "loss": 0.1293, + "step": 5638 + }, + { + "epoch": 0.29, + "learning_rate": 3.310766504721064e-07, + "loss": 0.1264, + "step": 5639 + }, + { + "epoch": 0.29, + "learning_rate": 3.3105111091506123e-07, + "loss": 0.1329, + "step": 5640 + }, + { + "epoch": 0.29, + "learning_rate": 3.3102556761252476e-07, + "loss": 0.1348, + "step": 5641 + }, + { + "epoch": 0.29, + "learning_rate": 3.310000205652269e-07, + "loss": 0.1435, + "step": 5642 + }, + { + "epoch": 0.29, + "learning_rate": 3.309744697738979e-07, + "loss": 0.122, + "step": 5643 + }, + { + "epoch": 0.29, + "learning_rate": 3.3094891523926803e-07, + "loss": 0.1111, + "step": 5644 + }, + { + "epoch": 0.29, + "learning_rate": 3.3092335696206764e-07, + "loss": 0.1482, + "step": 5645 + }, + { + "epoch": 0.29, + "learning_rate": 3.3089779494302706e-07, + "loss": 0.1608, + "step": 5646 + }, + { + "epoch": 0.29, + "learning_rate": 3.308722291828771e-07, + "loss": 0.171, + "step": 5647 + }, + { + "epoch": 0.29, + "learning_rate": 3.3084665968234823e-07, + "loss": 0.129, + "step": 5648 + }, + { + "epoch": 0.29, + "learning_rate": 3.308210864421713e-07, + "loss": 0.1249, + "step": 5649 + }, + { + "epoch": 0.29, + "learning_rate": 3.3079550946307724e-07, + "loss": 0.1095, + "step": 5650 + }, + { + "epoch": 0.29, + "learning_rate": 3.3076992874579705e-07, + "loss": 0.1386, + "step": 5651 + }, + { + "epoch": 0.3, + "learning_rate": 3.3074434429106174e-07, + "loss": 0.1218, + "step": 5652 + }, + { + "epoch": 0.3, + "learning_rate": 3.307187560996027e-07, + "loss": 0.1378, + "step": 5653 + }, + { + "epoch": 0.3, + "learning_rate": 3.306931641721511e-07, + "loss": 0.1583, + "step": 5654 + }, + { + "epoch": 0.3, + "learning_rate": 3.306675685094384e-07, + "loss": 0.1596, + "step": 5655 + }, + { + "epoch": 0.3, + "learning_rate": 3.306419691121962e-07, + "loss": 0.1497, + "step": 5656 + }, + { + "epoch": 0.3, + "learning_rate": 3.3061636598115604e-07, + "loss": 0.1578, + "step": 5657 + }, + { + "epoch": 0.3, + "learning_rate": 3.3059075911704973e-07, + "loss": 0.1365, + "step": 5658 + }, + { + "epoch": 0.3, + "learning_rate": 3.305651485206091e-07, + "loss": 0.1467, + "step": 5659 + }, + { + "epoch": 0.3, + "learning_rate": 3.305395341925661e-07, + "loss": 0.1221, + "step": 5660 + }, + { + "epoch": 0.3, + "learning_rate": 3.305139161336529e-07, + "loss": 0.1413, + "step": 5661 + }, + { + "epoch": 0.3, + "learning_rate": 3.3048829434460153e-07, + "loss": 0.1558, + "step": 5662 + }, + { + "epoch": 0.3, + "learning_rate": 3.3046266882614437e-07, + "loss": 0.1203, + "step": 5663 + }, + { + "epoch": 0.3, + "learning_rate": 3.3043703957901377e-07, + "loss": 0.1277, + "step": 5664 + }, + { + "epoch": 0.3, + "learning_rate": 3.304114066039422e-07, + "loss": 0.1309, + "step": 5665 + }, + { + "epoch": 0.3, + "learning_rate": 3.3038576990166227e-07, + "loss": 0.1828, + "step": 5666 + }, + { + "epoch": 0.3, + "learning_rate": 3.3036012947290675e-07, + "loss": 0.1545, + "step": 5667 + }, + { + "epoch": 0.3, + "learning_rate": 3.303344853184084e-07, + "loss": 0.1394, + "step": 5668 + }, + { + "epoch": 0.3, + "learning_rate": 3.3030883743890014e-07, + "loss": 0.1681, + "step": 5669 + }, + { + "epoch": 0.3, + "learning_rate": 3.3028318583511496e-07, + "loss": 0.1363, + "step": 5670 + }, + { + "epoch": 0.3, + "learning_rate": 3.3025753050778607e-07, + "loss": 0.1461, + "step": 5671 + }, + { + "epoch": 0.3, + "learning_rate": 3.302318714576467e-07, + "loss": 0.1347, + "step": 5672 + }, + { + "epoch": 0.3, + "learning_rate": 3.3020620868543007e-07, + "loss": 0.1222, + "step": 5673 + }, + { + "epoch": 0.3, + "learning_rate": 3.301805421918698e-07, + "loss": 0.1566, + "step": 5674 + }, + { + "epoch": 0.3, + "learning_rate": 3.3015487197769944e-07, + "loss": 0.1113, + "step": 5675 + }, + { + "epoch": 0.3, + "learning_rate": 3.3012919804365254e-07, + "loss": 0.1671, + "step": 5676 + }, + { + "epoch": 0.3, + "learning_rate": 3.301035203904629e-07, + "loss": 0.1211, + "step": 5677 + }, + { + "epoch": 0.3, + "learning_rate": 3.3007783901886443e-07, + "loss": 0.1741, + "step": 5678 + }, + { + "epoch": 0.3, + "learning_rate": 3.300521539295911e-07, + "loss": 0.1689, + "step": 5679 + }, + { + "epoch": 0.3, + "learning_rate": 3.30026465123377e-07, + "loss": 0.2228, + "step": 5680 + }, + { + "epoch": 0.3, + "learning_rate": 3.3000077260095634e-07, + "loss": 0.1499, + "step": 5681 + }, + { + "epoch": 0.3, + "learning_rate": 3.2997507636306344e-07, + "loss": 0.144, + "step": 5682 + }, + { + "epoch": 0.3, + "learning_rate": 3.2994937641043265e-07, + "loss": 0.1321, + "step": 5683 + }, + { + "epoch": 0.3, + "learning_rate": 3.299236727437986e-07, + "loss": 0.1521, + "step": 5684 + }, + { + "epoch": 0.3, + "learning_rate": 3.298979653638957e-07, + "loss": 0.1376, + "step": 5685 + }, + { + "epoch": 0.3, + "learning_rate": 3.2987225427145886e-07, + "loss": 0.1434, + "step": 5686 + }, + { + "epoch": 0.3, + "learning_rate": 3.298465394672229e-07, + "loss": 0.1539, + "step": 5687 + }, + { + "epoch": 0.3, + "learning_rate": 3.2982082095192266e-07, + "loss": 0.1379, + "step": 5688 + }, + { + "epoch": 0.3, + "learning_rate": 3.297950987262933e-07, + "loss": 0.136, + "step": 5689 + }, + { + "epoch": 0.3, + "learning_rate": 3.2976937279106987e-07, + "loss": 0.1538, + "step": 5690 + }, + { + "epoch": 0.3, + "learning_rate": 3.2974364314698766e-07, + "loss": 0.1547, + "step": 5691 + }, + { + "epoch": 0.3, + "learning_rate": 3.2971790979478203e-07, + "loss": 0.1669, + "step": 5692 + }, + { + "epoch": 0.3, + "learning_rate": 3.296921727351885e-07, + "loss": 0.1394, + "step": 5693 + }, + { + "epoch": 0.3, + "learning_rate": 3.2966643196894257e-07, + "loss": 0.1574, + "step": 5694 + }, + { + "epoch": 0.3, + "learning_rate": 3.2964068749678e-07, + "loss": 0.1609, + "step": 5695 + }, + { + "epoch": 0.3, + "learning_rate": 3.2961493931943653e-07, + "loss": 0.1489, + "step": 5696 + }, + { + "epoch": 0.3, + "learning_rate": 3.2958918743764804e-07, + "loss": 0.1599, + "step": 5697 + }, + { + "epoch": 0.3, + "learning_rate": 3.2956343185215056e-07, + "loss": 0.1282, + "step": 5698 + }, + { + "epoch": 0.3, + "learning_rate": 3.2953767256368015e-07, + "loss": 0.1165, + "step": 5699 + }, + { + "epoch": 0.3, + "learning_rate": 3.2951190957297304e-07, + "loss": 0.1298, + "step": 5700 + }, + { + "epoch": 0.3, + "learning_rate": 3.294861428807656e-07, + "loss": 0.1376, + "step": 5701 + }, + { + "epoch": 0.3, + "learning_rate": 3.294603724877942e-07, + "loss": 0.1643, + "step": 5702 + }, + { + "epoch": 0.3, + "learning_rate": 3.294345983947954e-07, + "loss": 0.1412, + "step": 5703 + }, + { + "epoch": 0.3, + "learning_rate": 3.2940882060250575e-07, + "loss": 0.1335, + "step": 5704 + }, + { + "epoch": 0.3, + "learning_rate": 3.2938303911166214e-07, + "loss": 0.1275, + "step": 5705 + }, + { + "epoch": 0.3, + "learning_rate": 3.293572539230013e-07, + "loss": 0.1416, + "step": 5706 + }, + { + "epoch": 0.3, + "learning_rate": 3.2933146503726015e-07, + "loss": 0.1683, + "step": 5707 + }, + { + "epoch": 0.3, + "learning_rate": 3.2930567245517586e-07, + "loss": 0.1361, + "step": 5708 + }, + { + "epoch": 0.3, + "learning_rate": 3.292798761774855e-07, + "loss": 0.1222, + "step": 5709 + }, + { + "epoch": 0.3, + "learning_rate": 3.2925407620492636e-07, + "loss": 0.174, + "step": 5710 + }, + { + "epoch": 0.3, + "learning_rate": 3.292282725382359e-07, + "loss": 0.1106, + "step": 5711 + }, + { + "epoch": 0.3, + "learning_rate": 3.2920246517815147e-07, + "loss": 0.1426, + "step": 5712 + }, + { + "epoch": 0.3, + "learning_rate": 3.291766541254106e-07, + "loss": 0.1519, + "step": 5713 + }, + { + "epoch": 0.3, + "learning_rate": 3.2915083938075124e-07, + "loss": 0.1728, + "step": 5714 + }, + { + "epoch": 0.3, + "learning_rate": 3.29125020944911e-07, + "loss": 0.1059, + "step": 5715 + }, + { + "epoch": 0.3, + "learning_rate": 3.290991988186278e-07, + "loss": 0.1407, + "step": 5716 + }, + { + "epoch": 0.3, + "learning_rate": 3.290733730026397e-07, + "loss": 0.1513, + "step": 5717 + }, + { + "epoch": 0.3, + "learning_rate": 3.290475434976847e-07, + "loss": 0.1302, + "step": 5718 + }, + { + "epoch": 0.3, + "learning_rate": 3.2902171030450117e-07, + "loss": 0.126, + "step": 5719 + }, + { + "epoch": 0.3, + "learning_rate": 3.289958734238273e-07, + "loss": 0.122, + "step": 5720 + }, + { + "epoch": 0.3, + "learning_rate": 3.289700328564016e-07, + "loss": 0.1216, + "step": 5721 + }, + { + "epoch": 0.3, + "learning_rate": 3.289441886029625e-07, + "loss": 0.174, + "step": 5722 + }, + { + "epoch": 0.3, + "learning_rate": 3.289183406642488e-07, + "loss": 0.1444, + "step": 5723 + }, + { + "epoch": 0.3, + "learning_rate": 3.2889248904099905e-07, + "loss": 0.1398, + "step": 5724 + }, + { + "epoch": 0.3, + "learning_rate": 3.288666337339523e-07, + "loss": 0.1188, + "step": 5725 + }, + { + "epoch": 0.3, + "learning_rate": 3.2884077474384737e-07, + "loss": 0.1793, + "step": 5726 + }, + { + "epoch": 0.3, + "learning_rate": 3.288149120714234e-07, + "loss": 0.1736, + "step": 5727 + }, + { + "epoch": 0.3, + "learning_rate": 3.287890457174195e-07, + "loss": 0.1287, + "step": 5728 + }, + { + "epoch": 0.3, + "learning_rate": 3.287631756825749e-07, + "loss": 0.1164, + "step": 5729 + }, + { + "epoch": 0.3, + "learning_rate": 3.287373019676291e-07, + "loss": 0.1695, + "step": 5730 + }, + { + "epoch": 0.3, + "learning_rate": 3.287114245733215e-07, + "loss": 0.1369, + "step": 5731 + }, + { + "epoch": 0.3, + "learning_rate": 3.2868554350039175e-07, + "loss": 0.1519, + "step": 5732 + }, + { + "epoch": 0.3, + "learning_rate": 3.2865965874957933e-07, + "loss": 0.1246, + "step": 5733 + }, + { + "epoch": 0.3, + "learning_rate": 3.2863377032162436e-07, + "loss": 0.1332, + "step": 5734 + }, + { + "epoch": 0.3, + "learning_rate": 3.286078782172665e-07, + "loss": 0.1142, + "step": 5735 + }, + { + "epoch": 0.3, + "learning_rate": 3.285819824372458e-07, + "loss": 0.1466, + "step": 5736 + }, + { + "epoch": 0.3, + "learning_rate": 3.2855608298230247e-07, + "loss": 0.1323, + "step": 5737 + }, + { + "epoch": 0.3, + "learning_rate": 3.2853017985317664e-07, + "loss": 0.1453, + "step": 5738 + }, + { + "epoch": 0.3, + "learning_rate": 3.2850427305060863e-07, + "loss": 0.1299, + "step": 5739 + }, + { + "epoch": 0.3, + "learning_rate": 3.2847836257533895e-07, + "loss": 0.109, + "step": 5740 + }, + { + "epoch": 0.3, + "learning_rate": 3.28452448428108e-07, + "loss": 0.1496, + "step": 5741 + }, + { + "epoch": 0.3, + "learning_rate": 3.284265306096565e-07, + "loss": 0.1393, + "step": 5742 + }, + { + "epoch": 0.3, + "learning_rate": 3.284006091207252e-07, + "loss": 0.141, + "step": 5743 + }, + { + "epoch": 0.3, + "learning_rate": 3.283746839620549e-07, + "loss": 0.129, + "step": 5744 + }, + { + "epoch": 0.3, + "learning_rate": 3.2834875513438663e-07, + "loss": 0.1257, + "step": 5745 + }, + { + "epoch": 0.3, + "learning_rate": 3.2832282263846134e-07, + "loss": 0.1606, + "step": 5746 + }, + { + "epoch": 0.3, + "learning_rate": 3.2829688647502023e-07, + "loss": 0.1749, + "step": 5747 + }, + { + "epoch": 0.3, + "learning_rate": 3.2827094664480463e-07, + "loss": 0.1368, + "step": 5748 + }, + { + "epoch": 0.3, + "learning_rate": 3.2824500314855584e-07, + "loss": 0.1566, + "step": 5749 + }, + { + "epoch": 0.3, + "learning_rate": 3.2821905598701533e-07, + "loss": 0.1154, + "step": 5750 + }, + { + "epoch": 0.3, + "learning_rate": 3.281931051609247e-07, + "loss": 0.0986, + "step": 5751 + }, + { + "epoch": 0.3, + "learning_rate": 3.2816715067102563e-07, + "loss": 0.1568, + "step": 5752 + }, + { + "epoch": 0.3, + "learning_rate": 3.281411925180599e-07, + "loss": 0.1135, + "step": 5753 + }, + { + "epoch": 0.3, + "learning_rate": 3.2811523070276954e-07, + "loss": 0.1209, + "step": 5754 + }, + { + "epoch": 0.3, + "learning_rate": 3.280892652258963e-07, + "loss": 0.116, + "step": 5755 + }, + { + "epoch": 0.3, + "learning_rate": 3.2806329608818246e-07, + "loss": 0.1611, + "step": 5756 + }, + { + "epoch": 0.3, + "learning_rate": 3.280373232903702e-07, + "loss": 0.1324, + "step": 5757 + }, + { + "epoch": 0.3, + "learning_rate": 3.2801134683320174e-07, + "loss": 0.1636, + "step": 5758 + }, + { + "epoch": 0.3, + "learning_rate": 3.279853667174197e-07, + "loss": 0.139, + "step": 5759 + }, + { + "epoch": 0.3, + "learning_rate": 3.2795938294376635e-07, + "loss": 0.158, + "step": 5760 + }, + { + "epoch": 0.3, + "learning_rate": 3.2793339551298447e-07, + "loss": 0.1346, + "step": 5761 + }, + { + "epoch": 0.3, + "learning_rate": 3.279074044258168e-07, + "loss": 0.1324, + "step": 5762 + }, + { + "epoch": 0.3, + "learning_rate": 3.278814096830061e-07, + "loss": 0.1581, + "step": 5763 + }, + { + "epoch": 0.3, + "learning_rate": 3.278554112852953e-07, + "loss": 0.1389, + "step": 5764 + }, + { + "epoch": 0.3, + "learning_rate": 3.278294092334276e-07, + "loss": 0.1469, + "step": 5765 + }, + { + "epoch": 0.3, + "learning_rate": 3.278034035281459e-07, + "loss": 0.1182, + "step": 5766 + }, + { + "epoch": 0.3, + "learning_rate": 3.277773941701937e-07, + "loss": 0.1292, + "step": 5767 + }, + { + "epoch": 0.3, + "learning_rate": 3.2775138116031415e-07, + "loss": 0.1357, + "step": 5768 + }, + { + "epoch": 0.3, + "learning_rate": 3.277253644992509e-07, + "loss": 0.1379, + "step": 5769 + }, + { + "epoch": 0.3, + "learning_rate": 3.276993441877474e-07, + "loss": 0.1658, + "step": 5770 + }, + { + "epoch": 0.3, + "learning_rate": 3.276733202265474e-07, + "loss": 0.1601, + "step": 5771 + }, + { + "epoch": 0.3, + "learning_rate": 3.2764729261639453e-07, + "loss": 0.1377, + "step": 5772 + }, + { + "epoch": 0.3, + "learning_rate": 3.2762126135803277e-07, + "loss": 0.1499, + "step": 5773 + }, + { + "epoch": 0.3, + "learning_rate": 3.275952264522061e-07, + "loss": 0.1501, + "step": 5774 + }, + { + "epoch": 0.3, + "learning_rate": 3.2756918789965866e-07, + "loss": 0.1307, + "step": 5775 + }, + { + "epoch": 0.3, + "learning_rate": 3.2754314570113455e-07, + "loss": 0.1467, + "step": 5776 + }, + { + "epoch": 0.3, + "learning_rate": 3.275170998573781e-07, + "loss": 0.1603, + "step": 5777 + }, + { + "epoch": 0.3, + "learning_rate": 3.2749105036913365e-07, + "loss": 0.151, + "step": 5778 + }, + { + "epoch": 0.3, + "learning_rate": 3.2746499723714584e-07, + "loss": 0.1653, + "step": 5779 + }, + { + "epoch": 0.3, + "learning_rate": 3.2743894046215917e-07, + "loss": 0.1309, + "step": 5780 + }, + { + "epoch": 0.3, + "learning_rate": 3.2741288004491843e-07, + "loss": 0.1085, + "step": 5781 + }, + { + "epoch": 0.3, + "learning_rate": 3.273868159861683e-07, + "loss": 0.1608, + "step": 5782 + }, + { + "epoch": 0.3, + "learning_rate": 3.273607482866539e-07, + "loss": 0.1878, + "step": 5783 + }, + { + "epoch": 0.3, + "learning_rate": 3.2733467694712003e-07, + "loss": 0.1302, + "step": 5784 + }, + { + "epoch": 0.3, + "learning_rate": 3.27308601968312e-07, + "loss": 0.1296, + "step": 5785 + }, + { + "epoch": 0.3, + "learning_rate": 3.2728252335097503e-07, + "loss": 0.1293, + "step": 5786 + }, + { + "epoch": 0.3, + "learning_rate": 3.2725644109585436e-07, + "loss": 0.1171, + "step": 5787 + }, + { + "epoch": 0.3, + "learning_rate": 3.2723035520369545e-07, + "loss": 0.2253, + "step": 5788 + }, + { + "epoch": 0.3, + "learning_rate": 3.272042656752439e-07, + "loss": 0.1296, + "step": 5789 + }, + { + "epoch": 0.3, + "learning_rate": 3.2717817251124535e-07, + "loss": 0.1783, + "step": 5790 + }, + { + "epoch": 0.3, + "learning_rate": 3.271520757124455e-07, + "loss": 0.1439, + "step": 5791 + }, + { + "epoch": 0.3, + "learning_rate": 3.271259752795902e-07, + "loss": 0.1249, + "step": 5792 + }, + { + "epoch": 0.3, + "learning_rate": 3.2709987121342553e-07, + "loss": 0.163, + "step": 5793 + }, + { + "epoch": 0.3, + "learning_rate": 3.270737635146974e-07, + "loss": 0.1668, + "step": 5794 + }, + { + "epoch": 0.3, + "learning_rate": 3.2704765218415223e-07, + "loss": 0.152, + "step": 5795 + }, + { + "epoch": 0.3, + "learning_rate": 3.27021537222536e-07, + "loss": 0.1385, + "step": 5796 + }, + { + "epoch": 0.3, + "learning_rate": 3.269954186305951e-07, + "loss": 0.1108, + "step": 5797 + }, + { + "epoch": 0.3, + "learning_rate": 3.269692964090762e-07, + "loss": 0.118, + "step": 5798 + }, + { + "epoch": 0.3, + "learning_rate": 3.269431705587258e-07, + "loss": 0.1537, + "step": 5799 + }, + { + "epoch": 0.3, + "learning_rate": 3.269170410802906e-07, + "loss": 0.1103, + "step": 5800 + }, + { + "epoch": 0.3, + "learning_rate": 3.2689090797451733e-07, + "loss": 0.1331, + "step": 5801 + }, + { + "epoch": 0.3, + "learning_rate": 3.268647712421529e-07, + "loss": 0.129, + "step": 5802 + }, + { + "epoch": 0.3, + "learning_rate": 3.268386308839444e-07, + "loss": 0.1484, + "step": 5803 + }, + { + "epoch": 0.3, + "learning_rate": 3.268124869006389e-07, + "loss": 0.1696, + "step": 5804 + }, + { + "epoch": 0.3, + "learning_rate": 3.2678633929298344e-07, + "loss": 0.177, + "step": 5805 + }, + { + "epoch": 0.3, + "learning_rate": 3.2676018806172556e-07, + "loss": 0.1504, + "step": 5806 + }, + { + "epoch": 0.3, + "learning_rate": 3.2673403320761257e-07, + "loss": 0.1537, + "step": 5807 + }, + { + "epoch": 0.3, + "learning_rate": 3.2670787473139194e-07, + "loss": 0.1431, + "step": 5808 + }, + { + "epoch": 0.3, + "learning_rate": 3.2668171263381133e-07, + "loss": 0.1514, + "step": 5809 + }, + { + "epoch": 0.3, + "learning_rate": 3.2665554691561845e-07, + "loss": 0.1402, + "step": 5810 + }, + { + "epoch": 0.3, + "learning_rate": 3.266293775775612e-07, + "loss": 0.1451, + "step": 5811 + }, + { + "epoch": 0.3, + "learning_rate": 3.266032046203874e-07, + "loss": 0.1581, + "step": 5812 + }, + { + "epoch": 0.3, + "learning_rate": 3.265770280448453e-07, + "loss": 0.1467, + "step": 5813 + }, + { + "epoch": 0.3, + "learning_rate": 3.2655084785168265e-07, + "loss": 0.1484, + "step": 5814 + }, + { + "epoch": 0.3, + "learning_rate": 3.2652466404164805e-07, + "loss": 0.1333, + "step": 5815 + }, + { + "epoch": 0.3, + "learning_rate": 3.2649847661548966e-07, + "loss": 0.1122, + "step": 5816 + }, + { + "epoch": 0.3, + "learning_rate": 3.26472285573956e-07, + "loss": 0.1422, + "step": 5817 + }, + { + "epoch": 0.3, + "learning_rate": 3.264460909177956e-07, + "loss": 0.1347, + "step": 5818 + }, + { + "epoch": 0.3, + "learning_rate": 3.26419892647757e-07, + "loss": 0.1044, + "step": 5819 + }, + { + "epoch": 0.3, + "learning_rate": 3.263936907645892e-07, + "loss": 0.1443, + "step": 5820 + }, + { + "epoch": 0.3, + "learning_rate": 3.263674852690409e-07, + "loss": 0.1136, + "step": 5821 + }, + { + "epoch": 0.3, + "learning_rate": 3.263412761618611e-07, + "loss": 0.1191, + "step": 5822 + }, + { + "epoch": 0.3, + "learning_rate": 3.2631506344379885e-07, + "loss": 0.1368, + "step": 5823 + }, + { + "epoch": 0.3, + "learning_rate": 3.2628884711560326e-07, + "loss": 0.1647, + "step": 5824 + }, + { + "epoch": 0.3, + "learning_rate": 3.2626262717802373e-07, + "loss": 0.1328, + "step": 5825 + }, + { + "epoch": 0.3, + "learning_rate": 3.262364036318095e-07, + "loss": 0.1397, + "step": 5826 + }, + { + "epoch": 0.3, + "learning_rate": 3.2621017647771023e-07, + "loss": 0.1473, + "step": 5827 + }, + { + "epoch": 0.3, + "learning_rate": 3.261839457164753e-07, + "loss": 0.1488, + "step": 5828 + }, + { + "epoch": 0.3, + "learning_rate": 3.2615771134885455e-07, + "loss": 0.1446, + "step": 5829 + }, + { + "epoch": 0.3, + "learning_rate": 3.2613147337559767e-07, + "loss": 0.1172, + "step": 5830 + }, + { + "epoch": 0.3, + "learning_rate": 3.2610523179745464e-07, + "loss": 0.1165, + "step": 5831 + }, + { + "epoch": 0.3, + "learning_rate": 3.260789866151753e-07, + "loss": 0.1839, + "step": 5832 + }, + { + "epoch": 0.3, + "learning_rate": 3.2605273782950997e-07, + "loss": 0.1694, + "step": 5833 + }, + { + "epoch": 0.3, + "learning_rate": 3.260264854412087e-07, + "loss": 0.1496, + "step": 5834 + }, + { + "epoch": 0.3, + "learning_rate": 3.2600022945102177e-07, + "loss": 0.1516, + "step": 5835 + }, + { + "epoch": 0.3, + "learning_rate": 3.259739698596997e-07, + "loss": 0.1546, + "step": 5836 + }, + { + "epoch": 0.3, + "learning_rate": 3.259477066679929e-07, + "loss": 0.1658, + "step": 5837 + }, + { + "epoch": 0.3, + "learning_rate": 3.25921439876652e-07, + "loss": 0.1647, + "step": 5838 + }, + { + "epoch": 0.3, + "learning_rate": 3.258951694864278e-07, + "loss": 0.1298, + "step": 5839 + }, + { + "epoch": 0.3, + "learning_rate": 3.2586889549807105e-07, + "loss": 0.1492, + "step": 5840 + }, + { + "epoch": 0.3, + "learning_rate": 3.2584261791233265e-07, + "loss": 0.123, + "step": 5841 + }, + { + "epoch": 0.3, + "learning_rate": 3.2581633672996366e-07, + "loss": 0.1649, + "step": 5842 + }, + { + "epoch": 0.3, + "learning_rate": 3.257900519517152e-07, + "loss": 0.1285, + "step": 5843 + }, + { + "epoch": 0.31, + "learning_rate": 3.257637635783385e-07, + "loss": 0.1574, + "step": 5844 + }, + { + "epoch": 0.31, + "learning_rate": 3.2573747161058485e-07, + "loss": 0.1987, + "step": 5845 + }, + { + "epoch": 0.31, + "learning_rate": 3.257111760492058e-07, + "loss": 0.1448, + "step": 5846 + }, + { + "epoch": 0.31, + "learning_rate": 3.256848768949527e-07, + "loss": 0.1608, + "step": 5847 + }, + { + "epoch": 0.31, + "learning_rate": 3.256585741485774e-07, + "loss": 0.1674, + "step": 5848 + }, + { + "epoch": 0.31, + "learning_rate": 3.256322678108315e-07, + "loss": 0.1369, + "step": 5849 + }, + { + "epoch": 0.31, + "learning_rate": 3.2560595788246684e-07, + "loss": 0.1305, + "step": 5850 + }, + { + "epoch": 0.31, + "learning_rate": 3.255796443642355e-07, + "loss": 0.1242, + "step": 5851 + }, + { + "epoch": 0.31, + "learning_rate": 3.255533272568894e-07, + "loss": 0.143, + "step": 5852 + }, + { + "epoch": 0.31, + "learning_rate": 3.255270065611808e-07, + "loss": 0.1552, + "step": 5853 + }, + { + "epoch": 0.31, + "learning_rate": 3.255006822778618e-07, + "loss": 0.1919, + "step": 5854 + }, + { + "epoch": 0.31, + "learning_rate": 3.254743544076849e-07, + "loss": 0.1373, + "step": 5855 + }, + { + "epoch": 0.31, + "learning_rate": 3.2544802295140246e-07, + "loss": 0.1091, + "step": 5856 + }, + { + "epoch": 0.31, + "learning_rate": 3.2542168790976716e-07, + "loss": 0.1415, + "step": 5857 + }, + { + "epoch": 0.31, + "learning_rate": 3.253953492835316e-07, + "loss": 0.1831, + "step": 5858 + }, + { + "epoch": 0.31, + "learning_rate": 3.253690070734485e-07, + "loss": 0.1539, + "step": 5859 + }, + { + "epoch": 0.31, + "learning_rate": 3.2534266128027086e-07, + "loss": 0.1388, + "step": 5860 + }, + { + "epoch": 0.31, + "learning_rate": 3.253163119047515e-07, + "loss": 0.1493, + "step": 5861 + }, + { + "epoch": 0.31, + "learning_rate": 3.2528995894764355e-07, + "loss": 0.1531, + "step": 5862 + }, + { + "epoch": 0.31, + "learning_rate": 3.2526360240970034e-07, + "loss": 0.123, + "step": 5863 + }, + { + "epoch": 0.31, + "learning_rate": 3.2523724229167486e-07, + "loss": 0.1561, + "step": 5864 + }, + { + "epoch": 0.31, + "learning_rate": 3.2521087859432077e-07, + "loss": 0.1789, + "step": 5865 + }, + { + "epoch": 0.31, + "learning_rate": 3.2518451131839133e-07, + "loss": 0.1481, + "step": 5866 + }, + { + "epoch": 0.31, + "learning_rate": 3.251581404646403e-07, + "loss": 0.1275, + "step": 5867 + }, + { + "epoch": 0.31, + "learning_rate": 3.251317660338213e-07, + "loss": 0.1406, + "step": 5868 + }, + { + "epoch": 0.31, + "learning_rate": 3.2510538802668813e-07, + "loss": 0.1287, + "step": 5869 + }, + { + "epoch": 0.31, + "learning_rate": 3.250790064439946e-07, + "loss": 0.1176, + "step": 5870 + }, + { + "epoch": 0.31, + "learning_rate": 3.2505262128649486e-07, + "loss": 0.1424, + "step": 5871 + }, + { + "epoch": 0.31, + "learning_rate": 3.2502623255494294e-07, + "loss": 0.1343, + "step": 5872 + }, + { + "epoch": 0.31, + "learning_rate": 3.249998402500929e-07, + "loss": 0.1016, + "step": 5873 + }, + { + "epoch": 0.31, + "learning_rate": 3.249734443726993e-07, + "loss": 0.1434, + "step": 5874 + }, + { + "epoch": 0.31, + "learning_rate": 3.2494704492351635e-07, + "loss": 0.1613, + "step": 5875 + }, + { + "epoch": 0.31, + "learning_rate": 3.249206419032986e-07, + "loss": 0.1461, + "step": 5876 + }, + { + "epoch": 0.31, + "learning_rate": 3.248942353128008e-07, + "loss": 0.1664, + "step": 5877 + }, + { + "epoch": 0.31, + "learning_rate": 3.2486782515277745e-07, + "loss": 0.1732, + "step": 5878 + }, + { + "epoch": 0.31, + "learning_rate": 3.248414114239835e-07, + "loss": 0.1295, + "step": 5879 + }, + { + "epoch": 0.31, + "learning_rate": 3.2481499412717376e-07, + "loss": 0.1269, + "step": 5880 + }, + { + "epoch": 0.31, + "learning_rate": 3.247885732631033e-07, + "loss": 0.1588, + "step": 5881 + }, + { + "epoch": 0.31, + "learning_rate": 3.247621488325273e-07, + "loss": 0.1486, + "step": 5882 + }, + { + "epoch": 0.31, + "learning_rate": 3.2473572083620085e-07, + "loss": 0.1202, + "step": 5883 + }, + { + "epoch": 0.31, + "learning_rate": 3.247092892748794e-07, + "loss": 0.1772, + "step": 5884 + }, + { + "epoch": 0.31, + "learning_rate": 3.2468285414931835e-07, + "loss": 0.1132, + "step": 5885 + }, + { + "epoch": 0.31, + "learning_rate": 3.2465641546027317e-07, + "loss": 0.1958, + "step": 5886 + }, + { + "epoch": 0.31, + "learning_rate": 3.2462997320849946e-07, + "loss": 0.1231, + "step": 5887 + }, + { + "epoch": 0.31, + "learning_rate": 3.2460352739475305e-07, + "loss": 0.1304, + "step": 5888 + }, + { + "epoch": 0.31, + "learning_rate": 3.245770780197897e-07, + "loss": 0.1519, + "step": 5889 + }, + { + "epoch": 0.31, + "learning_rate": 3.2455062508436546e-07, + "loss": 0.1541, + "step": 5890 + }, + { + "epoch": 0.31, + "learning_rate": 3.2452416858923617e-07, + "loss": 0.1319, + "step": 5891 + }, + { + "epoch": 0.31, + "learning_rate": 3.2449770853515814e-07, + "loss": 0.1184, + "step": 5892 + }, + { + "epoch": 0.31, + "learning_rate": 3.244712449228874e-07, + "loss": 0.1619, + "step": 5893 + }, + { + "epoch": 0.31, + "learning_rate": 3.244447777531806e-07, + "loss": 0.1213, + "step": 5894 + }, + { + "epoch": 0.31, + "learning_rate": 3.244183070267939e-07, + "loss": 0.1521, + "step": 5895 + }, + { + "epoch": 0.31, + "learning_rate": 3.2439183274448403e-07, + "loss": 0.1368, + "step": 5896 + }, + { + "epoch": 0.31, + "learning_rate": 3.243653549070075e-07, + "loss": 0.1458, + "step": 5897 + }, + { + "epoch": 0.31, + "learning_rate": 3.2433887351512123e-07, + "loss": 0.1666, + "step": 5898 + }, + { + "epoch": 0.31, + "learning_rate": 3.2431238856958186e-07, + "loss": 0.1421, + "step": 5899 + }, + { + "epoch": 0.31, + "learning_rate": 3.242859000711464e-07, + "loss": 0.1273, + "step": 5900 + }, + { + "epoch": 0.31, + "learning_rate": 3.24259408020572e-07, + "loss": 0.1118, + "step": 5901 + }, + { + "epoch": 0.31, + "learning_rate": 3.242329124186157e-07, + "loss": 0.1311, + "step": 5902 + }, + { + "epoch": 0.31, + "learning_rate": 3.2420641326603486e-07, + "loss": 0.1393, + "step": 5903 + }, + { + "epoch": 0.31, + "learning_rate": 3.2417991056358674e-07, + "loss": 0.1778, + "step": 5904 + }, + { + "epoch": 0.31, + "learning_rate": 3.2415340431202885e-07, + "loss": 0.1315, + "step": 5905 + }, + { + "epoch": 0.31, + "learning_rate": 3.241268945121187e-07, + "loss": 0.1338, + "step": 5906 + }, + { + "epoch": 0.31, + "learning_rate": 3.241003811646141e-07, + "loss": 0.0997, + "step": 5907 + }, + { + "epoch": 0.31, + "learning_rate": 3.240738642702726e-07, + "loss": 0.1473, + "step": 5908 + }, + { + "epoch": 0.31, + "learning_rate": 3.240473438298522e-07, + "loss": 0.1279, + "step": 5909 + }, + { + "epoch": 0.31, + "learning_rate": 3.240208198441108e-07, + "loss": 0.1245, + "step": 5910 + }, + { + "epoch": 0.31, + "learning_rate": 3.239942923138065e-07, + "loss": 0.1332, + "step": 5911 + }, + { + "epoch": 0.31, + "learning_rate": 3.239677612396975e-07, + "loss": 0.1251, + "step": 5912 + }, + { + "epoch": 0.31, + "learning_rate": 3.23941226622542e-07, + "loss": 0.1415, + "step": 5913 + }, + { + "epoch": 0.31, + "learning_rate": 3.2391468846309833e-07, + "loss": 0.1404, + "step": 5914 + }, + { + "epoch": 0.31, + "learning_rate": 3.2388814676212517e-07, + "loss": 0.1834, + "step": 5915 + }, + { + "epoch": 0.31, + "learning_rate": 3.238616015203808e-07, + "loss": 0.1302, + "step": 5916 + }, + { + "epoch": 0.31, + "learning_rate": 3.238350527386242e-07, + "loss": 0.1131, + "step": 5917 + }, + { + "epoch": 0.31, + "learning_rate": 3.238085004176138e-07, + "loss": 0.1354, + "step": 5918 + }, + { + "epoch": 0.31, + "learning_rate": 3.237819445581088e-07, + "loss": 0.137, + "step": 5919 + }, + { + "epoch": 0.31, + "learning_rate": 3.23755385160868e-07, + "loss": 0.1285, + "step": 5920 + }, + { + "epoch": 0.31, + "learning_rate": 3.237288222266505e-07, + "loss": 0.1357, + "step": 5921 + }, + { + "epoch": 0.31, + "learning_rate": 3.237022557562155e-07, + "loss": 0.1264, + "step": 5922 + }, + { + "epoch": 0.31, + "learning_rate": 3.2367568575032234e-07, + "loss": 0.1711, + "step": 5923 + }, + { + "epoch": 0.31, + "learning_rate": 3.236491122097303e-07, + "loss": 0.1466, + "step": 5924 + }, + { + "epoch": 0.31, + "learning_rate": 3.2362253513519887e-07, + "loss": 0.1384, + "step": 5925 + }, + { + "epoch": 0.31, + "learning_rate": 3.235959545274877e-07, + "loss": 0.137, + "step": 5926 + }, + { + "epoch": 0.31, + "learning_rate": 3.235693703873564e-07, + "loss": 0.1297, + "step": 5927 + }, + { + "epoch": 0.31, + "learning_rate": 3.2354278271556484e-07, + "loss": 0.1138, + "step": 5928 + }, + { + "epoch": 0.31, + "learning_rate": 3.235161915128728e-07, + "loss": 0.1051, + "step": 5929 + }, + { + "epoch": 0.31, + "learning_rate": 3.2348959678004033e-07, + "loss": 0.1454, + "step": 5930 + }, + { + "epoch": 0.31, + "learning_rate": 3.234629985178276e-07, + "loss": 0.1731, + "step": 5931 + }, + { + "epoch": 0.31, + "learning_rate": 3.234363967269946e-07, + "loss": 0.1558, + "step": 5932 + }, + { + "epoch": 0.31, + "learning_rate": 3.2340979140830185e-07, + "loss": 0.1541, + "step": 5933 + }, + { + "epoch": 0.31, + "learning_rate": 3.233831825625095e-07, + "loss": 0.1491, + "step": 5934 + }, + { + "epoch": 0.31, + "learning_rate": 3.2335657019037823e-07, + "loss": 0.1365, + "step": 5935 + }, + { + "epoch": 0.31, + "learning_rate": 3.233299542926685e-07, + "loss": 0.1338, + "step": 5936 + }, + { + "epoch": 0.31, + "learning_rate": 3.2330333487014113e-07, + "loss": 0.1765, + "step": 5937 + }, + { + "epoch": 0.31, + "learning_rate": 3.2327671192355676e-07, + "loss": 0.1229, + "step": 5938 + }, + { + "epoch": 0.31, + "learning_rate": 3.232500854536765e-07, + "loss": 0.1492, + "step": 5939 + }, + { + "epoch": 0.31, + "learning_rate": 3.2322345546126115e-07, + "loss": 0.1166, + "step": 5940 + }, + { + "epoch": 0.31, + "learning_rate": 3.2319682194707187e-07, + "loss": 0.1991, + "step": 5941 + }, + { + "epoch": 0.31, + "learning_rate": 3.231701849118699e-07, + "loss": 0.1913, + "step": 5942 + }, + { + "epoch": 0.31, + "learning_rate": 3.231435443564165e-07, + "loss": 0.1297, + "step": 5943 + }, + { + "epoch": 0.31, + "learning_rate": 3.23116900281473e-07, + "loss": 0.1451, + "step": 5944 + }, + { + "epoch": 0.31, + "learning_rate": 3.2309025268780105e-07, + "loss": 0.1297, + "step": 5945 + }, + { + "epoch": 0.31, + "learning_rate": 3.230636015761621e-07, + "loss": 0.175, + "step": 5946 + }, + { + "epoch": 0.31, + "learning_rate": 3.230369469473179e-07, + "loss": 0.1514, + "step": 5947 + }, + { + "epoch": 0.31, + "learning_rate": 3.230102888020303e-07, + "loss": 0.1511, + "step": 5948 + }, + { + "epoch": 0.31, + "learning_rate": 3.229836271410611e-07, + "loss": 0.1577, + "step": 5949 + }, + { + "epoch": 0.31, + "learning_rate": 3.229569619651724e-07, + "loss": 0.1318, + "step": 5950 + }, + { + "epoch": 0.31, + "learning_rate": 3.2293029327512625e-07, + "loss": 0.1532, + "step": 5951 + }, + { + "epoch": 0.31, + "learning_rate": 3.229036210716849e-07, + "loss": 0.1472, + "step": 5952 + }, + { + "epoch": 0.31, + "learning_rate": 3.228769453556105e-07, + "loss": 0.1592, + "step": 5953 + }, + { + "epoch": 0.31, + "learning_rate": 3.2285026612766573e-07, + "loss": 0.1253, + "step": 5954 + }, + { + "epoch": 0.31, + "learning_rate": 3.2282358338861284e-07, + "loss": 0.122, + "step": 5955 + }, + { + "epoch": 0.31, + "learning_rate": 3.227968971392145e-07, + "loss": 0.1662, + "step": 5956 + }, + { + "epoch": 0.31, + "learning_rate": 3.2277020738023344e-07, + "loss": 0.1638, + "step": 5957 + }, + { + "epoch": 0.31, + "learning_rate": 3.2274351411243254e-07, + "loss": 0.1336, + "step": 5958 + }, + { + "epoch": 0.31, + "learning_rate": 3.2271681733657454e-07, + "loss": 0.1385, + "step": 5959 + }, + { + "epoch": 0.31, + "learning_rate": 3.2269011705342257e-07, + "loss": 0.1553, + "step": 5960 + }, + { + "epoch": 0.31, + "learning_rate": 3.226634132637397e-07, + "loss": 0.1657, + "step": 5961 + }, + { + "epoch": 0.31, + "learning_rate": 3.226367059682891e-07, + "loss": 0.1311, + "step": 5962 + }, + { + "epoch": 0.31, + "learning_rate": 3.226099951678341e-07, + "loss": 0.1223, + "step": 5963 + }, + { + "epoch": 0.31, + "learning_rate": 3.2258328086313807e-07, + "loss": 0.1315, + "step": 5964 + }, + { + "epoch": 0.31, + "learning_rate": 3.2255656305496463e-07, + "loss": 0.1254, + "step": 5965 + }, + { + "epoch": 0.31, + "learning_rate": 3.2252984174407723e-07, + "loss": 0.1565, + "step": 5966 + }, + { + "epoch": 0.31, + "learning_rate": 3.2250311693123973e-07, + "loss": 0.1574, + "step": 5967 + }, + { + "epoch": 0.31, + "learning_rate": 3.224763886172158e-07, + "loss": 0.1294, + "step": 5968 + }, + { + "epoch": 0.31, + "learning_rate": 3.2244965680276944e-07, + "loss": 0.1772, + "step": 5969 + }, + { + "epoch": 0.31, + "learning_rate": 3.2242292148866454e-07, + "loss": 0.1274, + "step": 5970 + }, + { + "epoch": 0.31, + "learning_rate": 3.223961826756654e-07, + "loss": 0.1368, + "step": 5971 + }, + { + "epoch": 0.31, + "learning_rate": 3.2236944036453606e-07, + "loss": 0.1804, + "step": 5972 + }, + { + "epoch": 0.31, + "learning_rate": 3.2234269455604086e-07, + "loss": 0.1648, + "step": 5973 + }, + { + "epoch": 0.31, + "learning_rate": 3.223159452509442e-07, + "loss": 0.1662, + "step": 5974 + }, + { + "epoch": 0.31, + "learning_rate": 3.2228919245001065e-07, + "loss": 0.0928, + "step": 5975 + }, + { + "epoch": 0.31, + "learning_rate": 3.2226243615400475e-07, + "loss": 0.1268, + "step": 5976 + }, + { + "epoch": 0.31, + "learning_rate": 3.222356763636912e-07, + "loss": 0.1627, + "step": 5977 + }, + { + "epoch": 0.31, + "learning_rate": 3.2220891307983487e-07, + "loss": 0.2017, + "step": 5978 + }, + { + "epoch": 0.31, + "learning_rate": 3.221821463032007e-07, + "loss": 0.0968, + "step": 5979 + }, + { + "epoch": 0.31, + "learning_rate": 3.221553760345535e-07, + "loss": 0.1647, + "step": 5980 + }, + { + "epoch": 0.31, + "learning_rate": 3.221286022746585e-07, + "loss": 0.1286, + "step": 5981 + }, + { + "epoch": 0.31, + "learning_rate": 3.2210182502428095e-07, + "loss": 0.1392, + "step": 5982 + }, + { + "epoch": 0.31, + "learning_rate": 3.2207504428418607e-07, + "loss": 0.1315, + "step": 5983 + }, + { + "epoch": 0.31, + "learning_rate": 3.2204826005513935e-07, + "loss": 0.1407, + "step": 5984 + }, + { + "epoch": 0.31, + "learning_rate": 3.220214723379062e-07, + "loss": 0.1963, + "step": 5985 + }, + { + "epoch": 0.31, + "learning_rate": 3.219946811332523e-07, + "loss": 0.1252, + "step": 5986 + }, + { + "epoch": 0.31, + "learning_rate": 3.219678864419433e-07, + "loss": 0.1475, + "step": 5987 + }, + { + "epoch": 0.31, + "learning_rate": 3.21941088264745e-07, + "loss": 0.147, + "step": 5988 + }, + { + "epoch": 0.31, + "learning_rate": 3.219142866024233e-07, + "loss": 0.1385, + "step": 5989 + }, + { + "epoch": 0.31, + "learning_rate": 3.2188748145574433e-07, + "loss": 0.1559, + "step": 5990 + }, + { + "epoch": 0.31, + "learning_rate": 3.21860672825474e-07, + "loss": 0.1369, + "step": 5991 + }, + { + "epoch": 0.31, + "learning_rate": 3.218338607123787e-07, + "loss": 0.0991, + "step": 5992 + }, + { + "epoch": 0.31, + "learning_rate": 3.2180704511722446e-07, + "loss": 0.1531, + "step": 5993 + }, + { + "epoch": 0.31, + "learning_rate": 3.2178022604077795e-07, + "loss": 0.1159, + "step": 5994 + }, + { + "epoch": 0.31, + "learning_rate": 3.2175340348380557e-07, + "loss": 0.1182, + "step": 5995 + }, + { + "epoch": 0.31, + "learning_rate": 3.2172657744707387e-07, + "loss": 0.1755, + "step": 5996 + }, + { + "epoch": 0.31, + "learning_rate": 3.2169974793134964e-07, + "loss": 0.1432, + "step": 5997 + }, + { + "epoch": 0.31, + "learning_rate": 3.2167291493739964e-07, + "loss": 0.1573, + "step": 5998 + }, + { + "epoch": 0.31, + "learning_rate": 3.216460784659907e-07, + "loss": 0.1624, + "step": 5999 + }, + { + "epoch": 0.31, + "learning_rate": 3.2161923851788995e-07, + "loss": 0.1547, + "step": 6000 + }, + { + "epoch": 0.31, + "learning_rate": 3.215923950938644e-07, + "loss": 0.1427, + "step": 6001 + }, + { + "epoch": 0.31, + "learning_rate": 3.2156554819468124e-07, + "loss": 0.1522, + "step": 6002 + }, + { + "epoch": 0.31, + "learning_rate": 3.215386978211078e-07, + "loss": 0.2147, + "step": 6003 + }, + { + "epoch": 0.31, + "learning_rate": 3.2151184397391143e-07, + "loss": 0.1106, + "step": 6004 + }, + { + "epoch": 0.31, + "learning_rate": 3.2148498665385964e-07, + "loss": 0.1296, + "step": 6005 + }, + { + "epoch": 0.31, + "learning_rate": 3.2145812586172007e-07, + "loss": 0.1208, + "step": 6006 + }, + { + "epoch": 0.31, + "learning_rate": 3.2143126159826033e-07, + "loss": 0.1192, + "step": 6007 + }, + { + "epoch": 0.31, + "learning_rate": 3.214043938642483e-07, + "loss": 0.1479, + "step": 6008 + }, + { + "epoch": 0.31, + "learning_rate": 3.2137752266045186e-07, + "loss": 0.1254, + "step": 6009 + }, + { + "epoch": 0.31, + "learning_rate": 3.213506479876389e-07, + "loss": 0.1921, + "step": 6010 + }, + { + "epoch": 0.31, + "learning_rate": 3.213237698465776e-07, + "loss": 0.1288, + "step": 6011 + }, + { + "epoch": 0.31, + "learning_rate": 3.212968882380361e-07, + "loss": 0.1074, + "step": 6012 + }, + { + "epoch": 0.31, + "learning_rate": 3.212700031627827e-07, + "loss": 0.1404, + "step": 6013 + }, + { + "epoch": 0.31, + "learning_rate": 3.2124311462158587e-07, + "loss": 0.1359, + "step": 6014 + }, + { + "epoch": 0.31, + "learning_rate": 3.2121622261521397e-07, + "loss": 0.1787, + "step": 6015 + }, + { + "epoch": 0.31, + "learning_rate": 3.211893271444356e-07, + "loss": 0.1523, + "step": 6016 + }, + { + "epoch": 0.31, + "learning_rate": 3.211624282100195e-07, + "loss": 0.1532, + "step": 6017 + }, + { + "epoch": 0.31, + "learning_rate": 3.2113552581273446e-07, + "loss": 0.1583, + "step": 6018 + }, + { + "epoch": 0.31, + "learning_rate": 3.2110861995334936e-07, + "loss": 0.159, + "step": 6019 + }, + { + "epoch": 0.31, + "learning_rate": 3.2108171063263305e-07, + "loss": 0.1517, + "step": 6020 + }, + { + "epoch": 0.31, + "learning_rate": 3.2105479785135474e-07, + "loss": 0.1641, + "step": 6021 + }, + { + "epoch": 0.31, + "learning_rate": 3.2102788161028364e-07, + "loss": 0.131, + "step": 6022 + }, + { + "epoch": 0.31, + "learning_rate": 3.210009619101889e-07, + "loss": 0.1516, + "step": 6023 + }, + { + "epoch": 0.31, + "learning_rate": 3.209740387518399e-07, + "loss": 0.1644, + "step": 6024 + }, + { + "epoch": 0.31, + "learning_rate": 3.209471121360063e-07, + "loss": 0.1751, + "step": 6025 + }, + { + "epoch": 0.31, + "learning_rate": 3.2092018206345745e-07, + "loss": 0.1383, + "step": 6026 + }, + { + "epoch": 0.31, + "learning_rate": 3.2089324853496323e-07, + "loss": 0.1448, + "step": 6027 + }, + { + "epoch": 0.31, + "learning_rate": 3.208663115512932e-07, + "loss": 0.1588, + "step": 6028 + }, + { + "epoch": 0.31, + "learning_rate": 3.2083937111321734e-07, + "loss": 0.1553, + "step": 6029 + }, + { + "epoch": 0.31, + "learning_rate": 3.2081242722150556e-07, + "loss": 0.162, + "step": 6030 + }, + { + "epoch": 0.31, + "learning_rate": 3.207854798769281e-07, + "loss": 0.1392, + "step": 6031 + }, + { + "epoch": 0.31, + "learning_rate": 3.2075852908025487e-07, + "loss": 0.1638, + "step": 6032 + }, + { + "epoch": 0.31, + "learning_rate": 3.2073157483225633e-07, + "loss": 0.1249, + "step": 6033 + }, + { + "epoch": 0.31, + "learning_rate": 3.2070461713370275e-07, + "loss": 0.168, + "step": 6034 + }, + { + "epoch": 0.31, + "learning_rate": 3.2067765598536465e-07, + "loss": 0.169, + "step": 6035 + }, + { + "epoch": 0.32, + "learning_rate": 3.206506913880125e-07, + "loss": 0.1102, + "step": 6036 + }, + { + "epoch": 0.32, + "learning_rate": 3.20623723342417e-07, + "loss": 0.1264, + "step": 6037 + }, + { + "epoch": 0.32, + "learning_rate": 3.20596751849349e-07, + "loss": 0.1217, + "step": 6038 + }, + { + "epoch": 0.32, + "learning_rate": 3.205697769095793e-07, + "loss": 0.1235, + "step": 6039 + }, + { + "epoch": 0.32, + "learning_rate": 3.2054279852387873e-07, + "loss": 0.1308, + "step": 6040 + }, + { + "epoch": 0.32, + "learning_rate": 3.205158166930185e-07, + "loss": 0.1304, + "step": 6041 + }, + { + "epoch": 0.32, + "learning_rate": 3.2048883141776965e-07, + "loss": 0.1696, + "step": 6042 + }, + { + "epoch": 0.32, + "learning_rate": 3.2046184269890357e-07, + "loss": 0.1148, + "step": 6043 + }, + { + "epoch": 0.32, + "learning_rate": 3.204348505371915e-07, + "loss": 0.1298, + "step": 6044 + }, + { + "epoch": 0.32, + "learning_rate": 3.204078549334049e-07, + "loss": 0.1372, + "step": 6045 + }, + { + "epoch": 0.32, + "learning_rate": 3.2038085588831526e-07, + "loss": 0.1353, + "step": 6046 + }, + { + "epoch": 0.32, + "learning_rate": 3.203538534026943e-07, + "loss": 0.1538, + "step": 6047 + }, + { + "epoch": 0.32, + "learning_rate": 3.203268474773138e-07, + "loss": 0.1233, + "step": 6048 + }, + { + "epoch": 0.32, + "learning_rate": 3.2029983811294557e-07, + "loss": 0.1327, + "step": 6049 + }, + { + "epoch": 0.32, + "learning_rate": 3.202728253103615e-07, + "loss": 0.1303, + "step": 6050 + }, + { + "epoch": 0.32, + "learning_rate": 3.202458090703336e-07, + "loss": 0.1338, + "step": 6051 + }, + { + "epoch": 0.32, + "learning_rate": 3.2021878939363415e-07, + "loss": 0.1611, + "step": 6052 + }, + { + "epoch": 0.32, + "learning_rate": 3.2019176628103527e-07, + "loss": 0.1428, + "step": 6053 + }, + { + "epoch": 0.32, + "learning_rate": 3.2016473973330926e-07, + "loss": 0.1579, + "step": 6054 + }, + { + "epoch": 0.32, + "learning_rate": 3.201377097512286e-07, + "loss": 0.1745, + "step": 6055 + }, + { + "epoch": 0.32, + "learning_rate": 3.2011067633556586e-07, + "loss": 0.1383, + "step": 6056 + }, + { + "epoch": 0.32, + "learning_rate": 3.2008363948709366e-07, + "loss": 0.1542, + "step": 6057 + }, + { + "epoch": 0.32, + "learning_rate": 3.200565992065846e-07, + "loss": 0.1318, + "step": 6058 + }, + { + "epoch": 0.32, + "learning_rate": 3.200295554948116e-07, + "loss": 0.1734, + "step": 6059 + }, + { + "epoch": 0.32, + "learning_rate": 3.200025083525476e-07, + "loss": 0.1481, + "step": 6060 + }, + { + "epoch": 0.32, + "learning_rate": 3.199754577805656e-07, + "loss": 0.1344, + "step": 6061 + }, + { + "epoch": 0.32, + "learning_rate": 3.1994840377963865e-07, + "loss": 0.1129, + "step": 6062 + }, + { + "epoch": 0.32, + "learning_rate": 3.1992134635054007e-07, + "loss": 0.1294, + "step": 6063 + }, + { + "epoch": 0.32, + "learning_rate": 3.19894285494043e-07, + "loss": 0.1557, + "step": 6064 + }, + { + "epoch": 0.32, + "learning_rate": 3.198672212109211e-07, + "loss": 0.1246, + "step": 6065 + }, + { + "epoch": 0.32, + "learning_rate": 3.198401535019477e-07, + "loss": 0.1248, + "step": 6066 + }, + { + "epoch": 0.32, + "learning_rate": 3.198130823678965e-07, + "loss": 0.1313, + "step": 6067 + }, + { + "epoch": 0.32, + "learning_rate": 3.1978600780954105e-07, + "loss": 0.1696, + "step": 6068 + }, + { + "epoch": 0.32, + "learning_rate": 3.1975892982765535e-07, + "loss": 0.1467, + "step": 6069 + }, + { + "epoch": 0.32, + "learning_rate": 3.197318484230131e-07, + "loss": 0.1211, + "step": 6070 + }, + { + "epoch": 0.32, + "learning_rate": 3.1970476359638844e-07, + "loss": 0.1755, + "step": 6071 + }, + { + "epoch": 0.32, + "learning_rate": 3.1967767534855543e-07, + "loss": 0.117, + "step": 6072 + }, + { + "epoch": 0.32, + "learning_rate": 3.196505836802883e-07, + "loss": 0.1186, + "step": 6073 + }, + { + "epoch": 0.32, + "learning_rate": 3.196234885923613e-07, + "loss": 0.1082, + "step": 6074 + }, + { + "epoch": 0.32, + "learning_rate": 3.1959639008554875e-07, + "loss": 0.1286, + "step": 6075 + }, + { + "epoch": 0.32, + "learning_rate": 3.195692881606253e-07, + "loss": 0.1337, + "step": 6076 + }, + { + "epoch": 0.32, + "learning_rate": 3.1954218281836537e-07, + "loss": 0.1486, + "step": 6077 + }, + { + "epoch": 0.32, + "learning_rate": 3.1951507405954373e-07, + "loss": 0.1413, + "step": 6078 + }, + { + "epoch": 0.32, + "learning_rate": 3.194879618849351e-07, + "loss": 0.1334, + "step": 6079 + }, + { + "epoch": 0.32, + "learning_rate": 3.1946084629531436e-07, + "loss": 0.1478, + "step": 6080 + }, + { + "epoch": 0.32, + "learning_rate": 3.1943372729145664e-07, + "loss": 0.1661, + "step": 6081 + }, + { + "epoch": 0.32, + "learning_rate": 3.194066048741368e-07, + "loss": 0.1138, + "step": 6082 + }, + { + "epoch": 0.32, + "learning_rate": 3.193794790441301e-07, + "loss": 0.1401, + "step": 6083 + }, + { + "epoch": 0.32, + "learning_rate": 3.1935234980221184e-07, + "loss": 0.1583, + "step": 6084 + }, + { + "epoch": 0.32, + "learning_rate": 3.193252171491573e-07, + "loss": 0.1432, + "step": 6085 + }, + { + "epoch": 0.32, + "learning_rate": 3.1929808108574206e-07, + "loss": 0.1276, + "step": 6086 + }, + { + "epoch": 0.32, + "learning_rate": 3.1927094161274157e-07, + "loss": 0.1265, + "step": 6087 + }, + { + "epoch": 0.32, + "learning_rate": 3.192437987309315e-07, + "loss": 0.1316, + "step": 6088 + }, + { + "epoch": 0.32, + "learning_rate": 3.1921665244108764e-07, + "loss": 0.1493, + "step": 6089 + }, + { + "epoch": 0.32, + "learning_rate": 3.191895027439858e-07, + "loss": 0.108, + "step": 6090 + }, + { + "epoch": 0.32, + "learning_rate": 3.1916234964040207e-07, + "loss": 0.1343, + "step": 6091 + }, + { + "epoch": 0.32, + "learning_rate": 3.191351931311123e-07, + "loss": 0.1392, + "step": 6092 + }, + { + "epoch": 0.32, + "learning_rate": 3.191080332168927e-07, + "loss": 0.1341, + "step": 6093 + }, + { + "epoch": 0.32, + "learning_rate": 3.190808698985196e-07, + "loss": 0.1158, + "step": 6094 + }, + { + "epoch": 0.32, + "learning_rate": 3.1905370317676924e-07, + "loss": 0.167, + "step": 6095 + }, + { + "epoch": 0.32, + "learning_rate": 3.1902653305241804e-07, + "loss": 0.2291, + "step": 6096 + }, + { + "epoch": 0.32, + "learning_rate": 3.189993595262426e-07, + "loss": 0.1325, + "step": 6097 + }, + { + "epoch": 0.32, + "learning_rate": 3.1897218259901956e-07, + "loss": 0.1267, + "step": 6098 + }, + { + "epoch": 0.32, + "learning_rate": 3.1894500227152563e-07, + "loss": 0.1519, + "step": 6099 + }, + { + "epoch": 0.32, + "learning_rate": 3.189178185445376e-07, + "loss": 0.1427, + "step": 6100 + }, + { + "epoch": 0.32, + "learning_rate": 3.1889063141883236e-07, + "loss": 0.1616, + "step": 6101 + }, + { + "epoch": 0.32, + "learning_rate": 3.18863440895187e-07, + "loss": 0.128, + "step": 6102 + }, + { + "epoch": 0.32, + "learning_rate": 3.1883624697437864e-07, + "loss": 0.1484, + "step": 6103 + }, + { + "epoch": 0.32, + "learning_rate": 3.1880904965718446e-07, + "loss": 0.1425, + "step": 6104 + }, + { + "epoch": 0.32, + "learning_rate": 3.187818489443818e-07, + "loss": 0.1501, + "step": 6105 + }, + { + "epoch": 0.32, + "learning_rate": 3.18754644836748e-07, + "loss": 0.1506, + "step": 6106 + }, + { + "epoch": 0.32, + "learning_rate": 3.187274373350606e-07, + "loss": 0.1181, + "step": 6107 + }, + { + "epoch": 0.32, + "learning_rate": 3.1870022644009724e-07, + "loss": 0.1548, + "step": 6108 + }, + { + "epoch": 0.32, + "learning_rate": 3.1867301215263555e-07, + "loss": 0.1468, + "step": 6109 + }, + { + "epoch": 0.32, + "learning_rate": 3.186457944734534e-07, + "loss": 0.1645, + "step": 6110 + }, + { + "epoch": 0.32, + "learning_rate": 3.1861857340332854e-07, + "loss": 0.1337, + "step": 6111 + }, + { + "epoch": 0.32, + "learning_rate": 3.185913489430392e-07, + "loss": 0.1251, + "step": 6112 + }, + { + "epoch": 0.32, + "learning_rate": 3.1856412109336324e-07, + "loss": 0.1542, + "step": 6113 + }, + { + "epoch": 0.32, + "learning_rate": 3.1853688985507895e-07, + "loss": 0.1328, + "step": 6114 + }, + { + "epoch": 0.32, + "learning_rate": 3.1850965522896463e-07, + "loss": 0.1211, + "step": 6115 + }, + { + "epoch": 0.32, + "learning_rate": 3.1848241721579854e-07, + "loss": 0.1215, + "step": 6116 + }, + { + "epoch": 0.32, + "learning_rate": 3.184551758163593e-07, + "loss": 0.1438, + "step": 6117 + }, + { + "epoch": 0.32, + "learning_rate": 3.1842793103142533e-07, + "loss": 0.148, + "step": 6118 + }, + { + "epoch": 0.32, + "learning_rate": 3.1840068286177543e-07, + "loss": 0.1721, + "step": 6119 + }, + { + "epoch": 0.32, + "learning_rate": 3.1837343130818835e-07, + "loss": 0.1464, + "step": 6120 + }, + { + "epoch": 0.32, + "learning_rate": 3.1834617637144283e-07, + "loss": 0.1463, + "step": 6121 + }, + { + "epoch": 0.32, + "learning_rate": 3.183189180523179e-07, + "loss": 0.124, + "step": 6122 + }, + { + "epoch": 0.32, + "learning_rate": 3.182916563515927e-07, + "loss": 0.1532, + "step": 6123 + }, + { + "epoch": 0.32, + "learning_rate": 3.1826439127004616e-07, + "loss": 0.1525, + "step": 6124 + }, + { + "epoch": 0.32, + "learning_rate": 3.1823712280845776e-07, + "loss": 0.1418, + "step": 6125 + }, + { + "epoch": 0.32, + "learning_rate": 3.1820985096760674e-07, + "loss": 0.1618, + "step": 6126 + }, + { + "epoch": 0.32, + "learning_rate": 3.1818257574827254e-07, + "loss": 0.1272, + "step": 6127 + }, + { + "epoch": 0.32, + "learning_rate": 3.181552971512347e-07, + "loss": 0.1471, + "step": 6128 + }, + { + "epoch": 0.32, + "learning_rate": 3.181280151772729e-07, + "loss": 0.1597, + "step": 6129 + }, + { + "epoch": 0.32, + "learning_rate": 3.181007298271667e-07, + "loss": 0.1258, + "step": 6130 + }, + { + "epoch": 0.32, + "learning_rate": 3.180734411016961e-07, + "loss": 0.1276, + "step": 6131 + }, + { + "epoch": 0.32, + "learning_rate": 3.18046149001641e-07, + "loss": 0.1495, + "step": 6132 + }, + { + "epoch": 0.32, + "learning_rate": 3.1801885352778137e-07, + "loss": 0.1362, + "step": 6133 + }, + { + "epoch": 0.32, + "learning_rate": 3.1799155468089736e-07, + "loss": 0.1103, + "step": 6134 + }, + { + "epoch": 0.32, + "learning_rate": 3.1796425246176916e-07, + "loss": 0.1201, + "step": 6135 + }, + { + "epoch": 0.32, + "learning_rate": 3.179369468711771e-07, + "loss": 0.1644, + "step": 6136 + }, + { + "epoch": 0.32, + "learning_rate": 3.1790963790990156e-07, + "loss": 0.1762, + "step": 6137 + }, + { + "epoch": 0.32, + "learning_rate": 3.178823255787231e-07, + "loss": 0.1484, + "step": 6138 + }, + { + "epoch": 0.32, + "learning_rate": 3.178550098784222e-07, + "loss": 0.1839, + "step": 6139 + }, + { + "epoch": 0.32, + "learning_rate": 3.178276908097796e-07, + "loss": 0.1602, + "step": 6140 + }, + { + "epoch": 0.32, + "learning_rate": 3.178003683735762e-07, + "loss": 0.1364, + "step": 6141 + }, + { + "epoch": 0.32, + "learning_rate": 3.177730425705928e-07, + "loss": 0.1777, + "step": 6142 + }, + { + "epoch": 0.32, + "learning_rate": 3.1774571340161025e-07, + "loss": 0.125, + "step": 6143 + }, + { + "epoch": 0.32, + "learning_rate": 3.177183808674099e-07, + "loss": 0.1418, + "step": 6144 + }, + { + "epoch": 0.32, + "learning_rate": 3.1769104496877266e-07, + "loss": 0.1263, + "step": 6145 + }, + { + "epoch": 0.32, + "learning_rate": 3.1766370570648e-07, + "loss": 0.1615, + "step": 6146 + }, + { + "epoch": 0.32, + "learning_rate": 3.176363630813132e-07, + "loss": 0.1999, + "step": 6147 + }, + { + "epoch": 0.32, + "learning_rate": 3.176090170940537e-07, + "loss": 0.1258, + "step": 6148 + }, + { + "epoch": 0.32, + "learning_rate": 3.175816677454831e-07, + "loss": 0.1562, + "step": 6149 + }, + { + "epoch": 0.32, + "learning_rate": 3.175543150363831e-07, + "loss": 0.1172, + "step": 6150 + }, + { + "epoch": 0.32, + "learning_rate": 3.175269589675353e-07, + "loss": 0.1371, + "step": 6151 + }, + { + "epoch": 0.32, + "learning_rate": 3.1749959953972167e-07, + "loss": 0.1535, + "step": 6152 + }, + { + "epoch": 0.32, + "learning_rate": 3.174722367537241e-07, + "loss": 0.1226, + "step": 6153 + }, + { + "epoch": 0.32, + "learning_rate": 3.1744487061032477e-07, + "loss": 0.1508, + "step": 6154 + }, + { + "epoch": 0.32, + "learning_rate": 3.1741750111030564e-07, + "loss": 0.1336, + "step": 6155 + }, + { + "epoch": 0.32, + "learning_rate": 3.1739012825444896e-07, + "loss": 0.1536, + "step": 6156 + }, + { + "epoch": 0.32, + "learning_rate": 3.173627520435372e-07, + "loss": 0.1322, + "step": 6157 + }, + { + "epoch": 0.32, + "learning_rate": 3.1733537247835254e-07, + "loss": 0.1074, + "step": 6158 + }, + { + "epoch": 0.32, + "learning_rate": 3.1730798955967775e-07, + "loss": 0.131, + "step": 6159 + }, + { + "epoch": 0.32, + "learning_rate": 3.1728060328829524e-07, + "loss": 0.1262, + "step": 6160 + }, + { + "epoch": 0.32, + "learning_rate": 3.172532136649879e-07, + "loss": 0.1387, + "step": 6161 + }, + { + "epoch": 0.32, + "learning_rate": 3.172258206905384e-07, + "loss": 0.1352, + "step": 6162 + }, + { + "epoch": 0.32, + "learning_rate": 3.171984243657297e-07, + "loss": 0.1458, + "step": 6163 + }, + { + "epoch": 0.32, + "learning_rate": 3.1717102469134483e-07, + "loss": 0.1237, + "step": 6164 + }, + { + "epoch": 0.32, + "learning_rate": 3.171436216681668e-07, + "loss": 0.1089, + "step": 6165 + }, + { + "epoch": 0.32, + "learning_rate": 3.1711621529697886e-07, + "loss": 0.1841, + "step": 6166 + }, + { + "epoch": 0.32, + "learning_rate": 3.170888055785642e-07, + "loss": 0.172, + "step": 6167 + }, + { + "epoch": 0.32, + "learning_rate": 3.170613925137064e-07, + "loss": 0.1718, + "step": 6168 + }, + { + "epoch": 0.32, + "learning_rate": 3.1703397610318873e-07, + "loss": 0.1675, + "step": 6169 + }, + { + "epoch": 0.32, + "learning_rate": 3.1700655634779486e-07, + "loss": 0.123, + "step": 6170 + }, + { + "epoch": 0.32, + "learning_rate": 3.1697913324830845e-07, + "loss": 0.1047, + "step": 6171 + }, + { + "epoch": 0.32, + "learning_rate": 3.1695170680551323e-07, + "loss": 0.144, + "step": 6172 + }, + { + "epoch": 0.32, + "learning_rate": 3.169242770201931e-07, + "loss": 0.1467, + "step": 6173 + }, + { + "epoch": 0.32, + "learning_rate": 3.16896843893132e-07, + "loss": 0.1402, + "step": 6174 + }, + { + "epoch": 0.32, + "learning_rate": 3.1686940742511395e-07, + "loss": 0.1551, + "step": 6175 + }, + { + "epoch": 0.32, + "learning_rate": 3.168419676169231e-07, + "loss": 0.1511, + "step": 6176 + }, + { + "epoch": 0.32, + "learning_rate": 3.168145244693438e-07, + "loss": 0.1321, + "step": 6177 + }, + { + "epoch": 0.32, + "learning_rate": 3.1678707798316025e-07, + "loss": 0.1193, + "step": 6178 + }, + { + "epoch": 0.32, + "learning_rate": 3.167596281591569e-07, + "loss": 0.1459, + "step": 6179 + }, + { + "epoch": 0.32, + "learning_rate": 3.167321749981183e-07, + "loss": 0.1386, + "step": 6180 + }, + { + "epoch": 0.32, + "learning_rate": 3.1670471850082913e-07, + "loss": 0.1274, + "step": 6181 + }, + { + "epoch": 0.32, + "learning_rate": 3.16677258668074e-07, + "loss": 0.2117, + "step": 6182 + }, + { + "epoch": 0.32, + "learning_rate": 3.166497955006378e-07, + "loss": 0.1319, + "step": 6183 + }, + { + "epoch": 0.32, + "learning_rate": 3.166223289993054e-07, + "loss": 0.1542, + "step": 6184 + }, + { + "epoch": 0.32, + "learning_rate": 3.1659485916486184e-07, + "loss": 0.1522, + "step": 6185 + }, + { + "epoch": 0.32, + "learning_rate": 3.165673859980922e-07, + "loss": 0.1636, + "step": 6186 + }, + { + "epoch": 0.32, + "learning_rate": 3.1653990949978167e-07, + "loss": 0.145, + "step": 6187 + }, + { + "epoch": 0.32, + "learning_rate": 3.1651242967071546e-07, + "loss": 0.1276, + "step": 6188 + }, + { + "epoch": 0.32, + "learning_rate": 3.164849465116791e-07, + "loss": 0.1567, + "step": 6189 + }, + { + "epoch": 0.32, + "learning_rate": 3.16457460023458e-07, + "loss": 0.1328, + "step": 6190 + }, + { + "epoch": 0.32, + "learning_rate": 3.1642997020683775e-07, + "loss": 0.1288, + "step": 6191 + }, + { + "epoch": 0.32, + "learning_rate": 3.1640247706260393e-07, + "loss": 0.1416, + "step": 6192 + }, + { + "epoch": 0.32, + "learning_rate": 3.1637498059154243e-07, + "loss": 0.1261, + "step": 6193 + }, + { + "epoch": 0.32, + "learning_rate": 3.163474807944391e-07, + "loss": 0.1648, + "step": 6194 + }, + { + "epoch": 0.32, + "learning_rate": 3.1631997767207976e-07, + "loss": 0.1319, + "step": 6195 + }, + { + "epoch": 0.32, + "learning_rate": 3.162924712252506e-07, + "loss": 0.1854, + "step": 6196 + }, + { + "epoch": 0.32, + "learning_rate": 3.1626496145473774e-07, + "loss": 0.1498, + "step": 6197 + }, + { + "epoch": 0.32, + "learning_rate": 3.162374483613274e-07, + "loss": 0.1389, + "step": 6198 + }, + { + "epoch": 0.32, + "learning_rate": 3.162099319458059e-07, + "loss": 0.1474, + "step": 6199 + }, + { + "epoch": 0.32, + "learning_rate": 3.1618241220895967e-07, + "loss": 0.094, + "step": 6200 + }, + { + "epoch": 0.32, + "learning_rate": 3.1615488915157526e-07, + "loss": 0.1629, + "step": 6201 + }, + { + "epoch": 0.32, + "learning_rate": 3.161273627744393e-07, + "loss": 0.141, + "step": 6202 + }, + { + "epoch": 0.32, + "learning_rate": 3.160998330783385e-07, + "loss": 0.1412, + "step": 6203 + }, + { + "epoch": 0.32, + "learning_rate": 3.160723000640596e-07, + "loss": 0.1458, + "step": 6204 + }, + { + "epoch": 0.32, + "learning_rate": 3.160447637323895e-07, + "loss": 0.1739, + "step": 6205 + }, + { + "epoch": 0.32, + "learning_rate": 3.1601722408411533e-07, + "loss": 0.1568, + "step": 6206 + }, + { + "epoch": 0.32, + "learning_rate": 3.159896811200241e-07, + "loss": 0.1159, + "step": 6207 + }, + { + "epoch": 0.32, + "learning_rate": 3.15962134840903e-07, + "loss": 0.1324, + "step": 6208 + }, + { + "epoch": 0.32, + "learning_rate": 3.1593458524753935e-07, + "loss": 0.1401, + "step": 6209 + }, + { + "epoch": 0.32, + "learning_rate": 3.159070323407205e-07, + "loss": 0.1543, + "step": 6210 + }, + { + "epoch": 0.32, + "learning_rate": 3.1587947612123387e-07, + "loss": 0.1427, + "step": 6211 + }, + { + "epoch": 0.32, + "learning_rate": 3.158519165898671e-07, + "loss": 0.1396, + "step": 6212 + }, + { + "epoch": 0.32, + "learning_rate": 3.1582435374740784e-07, + "loss": 0.1255, + "step": 6213 + }, + { + "epoch": 0.32, + "learning_rate": 3.157967875946438e-07, + "loss": 0.164, + "step": 6214 + }, + { + "epoch": 0.32, + "learning_rate": 3.15769218132363e-07, + "loss": 0.1254, + "step": 6215 + }, + { + "epoch": 0.32, + "learning_rate": 3.1574164536135313e-07, + "loss": 0.1468, + "step": 6216 + }, + { + "epoch": 0.32, + "learning_rate": 3.157140692824024e-07, + "loss": 0.1534, + "step": 6217 + }, + { + "epoch": 0.32, + "learning_rate": 3.156864898962988e-07, + "loss": 0.1416, + "step": 6218 + }, + { + "epoch": 0.32, + "learning_rate": 3.156589072038308e-07, + "loss": 0.1371, + "step": 6219 + }, + { + "epoch": 0.32, + "learning_rate": 3.1563132120578655e-07, + "loss": 0.1383, + "step": 6220 + }, + { + "epoch": 0.32, + "learning_rate": 3.1560373190295446e-07, + "loss": 0.1285, + "step": 6221 + }, + { + "epoch": 0.32, + "learning_rate": 3.1557613929612303e-07, + "loss": 0.123, + "step": 6222 + }, + { + "epoch": 0.32, + "learning_rate": 3.15548543386081e-07, + "loss": 0.1335, + "step": 6223 + }, + { + "epoch": 0.32, + "learning_rate": 3.15520944173617e-07, + "loss": 0.1889, + "step": 6224 + }, + { + "epoch": 0.32, + "learning_rate": 3.1549334165951984e-07, + "loss": 0.1306, + "step": 6225 + }, + { + "epoch": 0.32, + "learning_rate": 3.1546573584457834e-07, + "loss": 0.1465, + "step": 6226 + }, + { + "epoch": 0.33, + "learning_rate": 3.1543812672958156e-07, + "loss": 0.13, + "step": 6227 + }, + { + "epoch": 0.33, + "learning_rate": 3.1541051431531855e-07, + "loss": 0.153, + "step": 6228 + }, + { + "epoch": 0.33, + "learning_rate": 3.153828986025785e-07, + "loss": 0.1118, + "step": 6229 + }, + { + "epoch": 0.33, + "learning_rate": 3.1535527959215064e-07, + "loss": 0.161, + "step": 6230 + }, + { + "epoch": 0.33, + "learning_rate": 3.1532765728482433e-07, + "loss": 0.1554, + "step": 6231 + }, + { + "epoch": 0.33, + "learning_rate": 3.153000316813891e-07, + "loss": 0.1221, + "step": 6232 + }, + { + "epoch": 0.33, + "learning_rate": 3.152724027826345e-07, + "loss": 0.1224, + "step": 6233 + }, + { + "epoch": 0.33, + "learning_rate": 3.1524477058935e-07, + "loss": 0.1308, + "step": 6234 + }, + { + "epoch": 0.33, + "learning_rate": 3.1521713510232557e-07, + "loss": 0.1779, + "step": 6235 + }, + { + "epoch": 0.33, + "learning_rate": 3.151894963223509e-07, + "loss": 0.1404, + "step": 6236 + }, + { + "epoch": 0.33, + "learning_rate": 3.15161854250216e-07, + "loss": 0.1397, + "step": 6237 + }, + { + "epoch": 0.33, + "learning_rate": 3.151342088867108e-07, + "loss": 0.1056, + "step": 6238 + }, + { + "epoch": 0.33, + "learning_rate": 3.151065602326255e-07, + "loss": 0.1388, + "step": 6239 + }, + { + "epoch": 0.33, + "learning_rate": 3.1507890828875025e-07, + "loss": 0.1246, + "step": 6240 + }, + { + "epoch": 0.33, + "learning_rate": 3.150512530558754e-07, + "loss": 0.1247, + "step": 6241 + }, + { + "epoch": 0.33, + "learning_rate": 3.150235945347913e-07, + "loss": 0.1429, + "step": 6242 + }, + { + "epoch": 0.33, + "learning_rate": 3.1499593272628847e-07, + "loss": 0.1636, + "step": 6243 + }, + { + "epoch": 0.33, + "learning_rate": 3.149682676311575e-07, + "loss": 0.1066, + "step": 6244 + }, + { + "epoch": 0.33, + "learning_rate": 3.14940599250189e-07, + "loss": 0.1254, + "step": 6245 + }, + { + "epoch": 0.33, + "learning_rate": 3.149129275841739e-07, + "loss": 0.1412, + "step": 6246 + }, + { + "epoch": 0.33, + "learning_rate": 3.148852526339029e-07, + "loss": 0.1556, + "step": 6247 + }, + { + "epoch": 0.33, + "learning_rate": 3.14857574400167e-07, + "loss": 0.181, + "step": 6248 + }, + { + "epoch": 0.33, + "learning_rate": 3.148298928837574e-07, + "loss": 0.1241, + "step": 6249 + }, + { + "epoch": 0.33, + "learning_rate": 3.1480220808546504e-07, + "loss": 0.1412, + "step": 6250 + }, + { + "epoch": 0.33, + "learning_rate": 3.1477452000608126e-07, + "loss": 0.1431, + "step": 6251 + }, + { + "epoch": 0.33, + "learning_rate": 3.1474682864639743e-07, + "loss": 0.124, + "step": 6252 + }, + { + "epoch": 0.33, + "learning_rate": 3.147191340072049e-07, + "loss": 0.1236, + "step": 6253 + }, + { + "epoch": 0.33, + "learning_rate": 3.146914360892953e-07, + "loss": 0.1876, + "step": 6254 + }, + { + "epoch": 0.33, + "learning_rate": 3.146637348934601e-07, + "loss": 0.1527, + "step": 6255 + }, + { + "epoch": 0.33, + "learning_rate": 3.1463603042049115e-07, + "loss": 0.1431, + "step": 6256 + }, + { + "epoch": 0.33, + "learning_rate": 3.1460832267118016e-07, + "loss": 0.1244, + "step": 6257 + }, + { + "epoch": 0.33, + "learning_rate": 3.145806116463191e-07, + "loss": 0.1223, + "step": 6258 + }, + { + "epoch": 0.33, + "learning_rate": 3.145528973466999e-07, + "loss": 0.1114, + "step": 6259 + }, + { + "epoch": 0.33, + "learning_rate": 3.145251797731147e-07, + "loss": 0.2034, + "step": 6260 + }, + { + "epoch": 0.33, + "learning_rate": 3.144974589263556e-07, + "loss": 0.1141, + "step": 6261 + }, + { + "epoch": 0.33, + "learning_rate": 3.1446973480721496e-07, + "loss": 0.1703, + "step": 6262 + }, + { + "epoch": 0.33, + "learning_rate": 3.144420074164851e-07, + "loss": 0.1344, + "step": 6263 + }, + { + "epoch": 0.33, + "learning_rate": 3.1441427675495853e-07, + "loss": 0.1653, + "step": 6264 + }, + { + "epoch": 0.33, + "learning_rate": 3.143865428234277e-07, + "loss": 0.1403, + "step": 6265 + }, + { + "epoch": 0.33, + "learning_rate": 3.1435880562268534e-07, + "loss": 0.146, + "step": 6266 + }, + { + "epoch": 0.33, + "learning_rate": 3.143310651535242e-07, + "loss": 0.1538, + "step": 6267 + }, + { + "epoch": 0.33, + "learning_rate": 3.14303321416737e-07, + "loss": 0.1166, + "step": 6268 + }, + { + "epoch": 0.33, + "learning_rate": 3.1427557441311683e-07, + "loss": 0.1609, + "step": 6269 + }, + { + "epoch": 0.33, + "learning_rate": 3.142478241434566e-07, + "loss": 0.1126, + "step": 6270 + }, + { + "epoch": 0.33, + "learning_rate": 3.1422007060854945e-07, + "loss": 0.1549, + "step": 6271 + }, + { + "epoch": 0.33, + "learning_rate": 3.141923138091886e-07, + "loss": 0.1307, + "step": 6272 + }, + { + "epoch": 0.33, + "learning_rate": 3.1416455374616726e-07, + "loss": 0.1763, + "step": 6273 + }, + { + "epoch": 0.33, + "learning_rate": 3.14136790420279e-07, + "loss": 0.1737, + "step": 6274 + }, + { + "epoch": 0.33, + "learning_rate": 3.141090238323172e-07, + "loss": 0.1816, + "step": 6275 + }, + { + "epoch": 0.33, + "learning_rate": 3.1408125398307545e-07, + "loss": 0.1538, + "step": 6276 + }, + { + "epoch": 0.33, + "learning_rate": 3.1405348087334736e-07, + "loss": 0.1399, + "step": 6277 + }, + { + "epoch": 0.33, + "learning_rate": 3.1402570450392684e-07, + "loss": 0.1561, + "step": 6278 + }, + { + "epoch": 0.33, + "learning_rate": 3.139979248756076e-07, + "loss": 0.1924, + "step": 6279 + }, + { + "epoch": 0.33, + "learning_rate": 3.139701419891837e-07, + "loss": 0.1062, + "step": 6280 + }, + { + "epoch": 0.33, + "learning_rate": 3.1394235584544914e-07, + "loss": 0.1193, + "step": 6281 + }, + { + "epoch": 0.33, + "learning_rate": 3.13914566445198e-07, + "loss": 0.1237, + "step": 6282 + }, + { + "epoch": 0.33, + "learning_rate": 3.138867737892246e-07, + "loss": 0.1205, + "step": 6283 + }, + { + "epoch": 0.33, + "learning_rate": 3.1385897787832327e-07, + "loss": 0.162, + "step": 6284 + }, + { + "epoch": 0.33, + "learning_rate": 3.1383117871328844e-07, + "loss": 0.1481, + "step": 6285 + }, + { + "epoch": 0.33, + "learning_rate": 3.138033762949145e-07, + "loss": 0.1516, + "step": 6286 + }, + { + "epoch": 0.33, + "learning_rate": 3.1377557062399617e-07, + "loss": 0.1517, + "step": 6287 + }, + { + "epoch": 0.33, + "learning_rate": 3.137477617013282e-07, + "loss": 0.1029, + "step": 6288 + }, + { + "epoch": 0.33, + "learning_rate": 3.1371994952770513e-07, + "loss": 0.1593, + "step": 6289 + }, + { + "epoch": 0.33, + "learning_rate": 3.1369213410392206e-07, + "loss": 0.1514, + "step": 6290 + }, + { + "epoch": 0.33, + "learning_rate": 3.136643154307739e-07, + "loss": 0.1101, + "step": 6291 + }, + { + "epoch": 0.33, + "learning_rate": 3.1363649350905577e-07, + "loss": 0.1181, + "step": 6292 + }, + { + "epoch": 0.33, + "learning_rate": 3.1360866833956275e-07, + "loss": 0.1594, + "step": 6293 + }, + { + "epoch": 0.33, + "learning_rate": 3.1358083992309016e-07, + "loss": 0.1971, + "step": 6294 + }, + { + "epoch": 0.33, + "learning_rate": 3.1355300826043325e-07, + "loss": 0.1674, + "step": 6295 + }, + { + "epoch": 0.33, + "learning_rate": 3.1352517335238766e-07, + "loss": 0.1394, + "step": 6296 + }, + { + "epoch": 0.33, + "learning_rate": 3.1349733519974865e-07, + "loss": 0.1358, + "step": 6297 + }, + { + "epoch": 0.33, + "learning_rate": 3.134694938033121e-07, + "loss": 0.1717, + "step": 6298 + }, + { + "epoch": 0.33, + "learning_rate": 3.1344164916387354e-07, + "loss": 0.1587, + "step": 6299 + }, + { + "epoch": 0.33, + "learning_rate": 3.1341380128222886e-07, + "loss": 0.1843, + "step": 6300 + }, + { + "epoch": 0.33, + "learning_rate": 3.1338595015917396e-07, + "loss": 0.1394, + "step": 6301 + }, + { + "epoch": 0.33, + "learning_rate": 3.133580957955049e-07, + "loss": 0.1292, + "step": 6302 + }, + { + "epoch": 0.33, + "learning_rate": 3.133302381920176e-07, + "loss": 0.1545, + "step": 6303 + }, + { + "epoch": 0.33, + "learning_rate": 3.1330237734950837e-07, + "loss": 0.145, + "step": 6304 + }, + { + "epoch": 0.33, + "learning_rate": 3.132745132687735e-07, + "loss": 0.1464, + "step": 6305 + }, + { + "epoch": 0.33, + "learning_rate": 3.132466459506093e-07, + "loss": 0.1653, + "step": 6306 + }, + { + "epoch": 0.33, + "learning_rate": 3.1321877539581214e-07, + "loss": 0.1337, + "step": 6307 + }, + { + "epoch": 0.33, + "learning_rate": 3.131909016051788e-07, + "loss": 0.1286, + "step": 6308 + }, + { + "epoch": 0.33, + "learning_rate": 3.1316302457950573e-07, + "loss": 0.1624, + "step": 6309 + }, + { + "epoch": 0.33, + "learning_rate": 3.131351443195898e-07, + "loss": 0.1409, + "step": 6310 + }, + { + "epoch": 0.33, + "learning_rate": 3.1310726082622774e-07, + "loss": 0.1414, + "step": 6311 + }, + { + "epoch": 0.33, + "learning_rate": 3.1307937410021644e-07, + "loss": 0.1416, + "step": 6312 + }, + { + "epoch": 0.33, + "learning_rate": 3.1305148414235304e-07, + "loss": 0.1435, + "step": 6313 + }, + { + "epoch": 0.33, + "learning_rate": 3.1302359095343456e-07, + "loss": 0.1329, + "step": 6314 + }, + { + "epoch": 0.33, + "learning_rate": 3.1299569453425826e-07, + "loss": 0.1337, + "step": 6315 + }, + { + "epoch": 0.33, + "learning_rate": 3.129677948856213e-07, + "loss": 0.1377, + "step": 6316 + }, + { + "epoch": 0.33, + "learning_rate": 3.129398920083212e-07, + "loss": 0.1797, + "step": 6317 + }, + { + "epoch": 0.33, + "learning_rate": 3.1291198590315546e-07, + "loss": 0.1184, + "step": 6318 + }, + { + "epoch": 0.33, + "learning_rate": 3.128840765709215e-07, + "loss": 0.1693, + "step": 6319 + }, + { + "epoch": 0.33, + "learning_rate": 3.1285616401241704e-07, + "loss": 0.224, + "step": 6320 + }, + { + "epoch": 0.33, + "learning_rate": 3.1282824822843984e-07, + "loss": 0.1525, + "step": 6321 + }, + { + "epoch": 0.33, + "learning_rate": 3.128003292197878e-07, + "loss": 0.1746, + "step": 6322 + }, + { + "epoch": 0.33, + "learning_rate": 3.1277240698725883e-07, + "loss": 0.1866, + "step": 6323 + }, + { + "epoch": 0.33, + "learning_rate": 3.1274448153165085e-07, + "loss": 0.1242, + "step": 6324 + }, + { + "epoch": 0.33, + "learning_rate": 3.127165528537621e-07, + "loss": 0.1333, + "step": 6325 + }, + { + "epoch": 0.33, + "learning_rate": 3.1268862095439077e-07, + "loss": 0.1807, + "step": 6326 + }, + { + "epoch": 0.33, + "learning_rate": 3.126606858343352e-07, + "loss": 0.1146, + "step": 6327 + }, + { + "epoch": 0.33, + "learning_rate": 3.126327474943936e-07, + "loss": 0.1362, + "step": 6328 + }, + { + "epoch": 0.33, + "learning_rate": 3.1260480593536466e-07, + "loss": 0.1462, + "step": 6329 + }, + { + "epoch": 0.33, + "learning_rate": 3.125768611580469e-07, + "loss": 0.1318, + "step": 6330 + }, + { + "epoch": 0.33, + "learning_rate": 3.1254891316323903e-07, + "loss": 0.1238, + "step": 6331 + }, + { + "epoch": 0.33, + "learning_rate": 3.125209619517397e-07, + "loss": 0.1942, + "step": 6332 + }, + { + "epoch": 0.33, + "learning_rate": 3.124930075243479e-07, + "loss": 0.1322, + "step": 6333 + }, + { + "epoch": 0.33, + "learning_rate": 3.124650498818625e-07, + "loss": 0.1283, + "step": 6334 + }, + { + "epoch": 0.33, + "learning_rate": 3.124370890250826e-07, + "loss": 0.1494, + "step": 6335 + }, + { + "epoch": 0.33, + "learning_rate": 3.124091249548072e-07, + "loss": 0.1469, + "step": 6336 + }, + { + "epoch": 0.33, + "learning_rate": 3.123811576718357e-07, + "loss": 0.1307, + "step": 6337 + }, + { + "epoch": 0.33, + "learning_rate": 3.1235318717696724e-07, + "loss": 0.1417, + "step": 6338 + }, + { + "epoch": 0.33, + "learning_rate": 3.123252134710014e-07, + "loss": 0.1452, + "step": 6339 + }, + { + "epoch": 0.33, + "learning_rate": 3.122972365547376e-07, + "loss": 0.1417, + "step": 6340 + }, + { + "epoch": 0.33, + "learning_rate": 3.1226925642897537e-07, + "loss": 0.1229, + "step": 6341 + }, + { + "epoch": 0.33, + "learning_rate": 3.122412730945145e-07, + "loss": 0.1227, + "step": 6342 + }, + { + "epoch": 0.33, + "learning_rate": 3.1221328655215465e-07, + "loss": 0.1433, + "step": 6343 + }, + { + "epoch": 0.33, + "learning_rate": 3.1218529680269583e-07, + "loss": 0.1544, + "step": 6344 + }, + { + "epoch": 0.33, + "learning_rate": 3.1215730384693784e-07, + "loss": 0.1261, + "step": 6345 + }, + { + "epoch": 0.33, + "learning_rate": 3.1212930768568085e-07, + "loss": 0.1475, + "step": 6346 + }, + { + "epoch": 0.33, + "learning_rate": 3.12101308319725e-07, + "loss": 0.126, + "step": 6347 + }, + { + "epoch": 0.33, + "learning_rate": 3.120733057498704e-07, + "loss": 0.1451, + "step": 6348 + }, + { + "epoch": 0.33, + "learning_rate": 3.120452999769175e-07, + "loss": 0.1231, + "step": 6349 + }, + { + "epoch": 0.33, + "learning_rate": 3.120172910016667e-07, + "loss": 0.172, + "step": 6350 + }, + { + "epoch": 0.33, + "learning_rate": 3.1198927882491846e-07, + "loss": 0.1649, + "step": 6351 + }, + { + "epoch": 0.33, + "learning_rate": 3.119612634474734e-07, + "loss": 0.1232, + "step": 6352 + }, + { + "epoch": 0.33, + "learning_rate": 3.119332448701322e-07, + "loss": 0.1102, + "step": 6353 + }, + { + "epoch": 0.33, + "learning_rate": 3.1190522309369564e-07, + "loss": 0.145, + "step": 6354 + }, + { + "epoch": 0.33, + "learning_rate": 3.118771981189646e-07, + "loss": 0.1376, + "step": 6355 + }, + { + "epoch": 0.33, + "learning_rate": 3.118491699467401e-07, + "loss": 0.1465, + "step": 6356 + }, + { + "epoch": 0.33, + "learning_rate": 3.118211385778231e-07, + "loss": 0.161, + "step": 6357 + }, + { + "epoch": 0.33, + "learning_rate": 3.117931040130148e-07, + "loss": 0.1461, + "step": 6358 + }, + { + "epoch": 0.33, + "learning_rate": 3.117650662531164e-07, + "loss": 0.1521, + "step": 6359 + }, + { + "epoch": 0.33, + "learning_rate": 3.117370252989294e-07, + "loss": 0.1554, + "step": 6360 + }, + { + "epoch": 0.33, + "learning_rate": 3.117089811512549e-07, + "loss": 0.1611, + "step": 6361 + }, + { + "epoch": 0.33, + "learning_rate": 3.116809338108947e-07, + "loss": 0.1458, + "step": 6362 + }, + { + "epoch": 0.33, + "learning_rate": 3.1165288327865027e-07, + "loss": 0.1252, + "step": 6363 + }, + { + "epoch": 0.33, + "learning_rate": 3.116248295553233e-07, + "loss": 0.1817, + "step": 6364 + }, + { + "epoch": 0.33, + "learning_rate": 3.115967726417157e-07, + "loss": 0.1788, + "step": 6365 + }, + { + "epoch": 0.33, + "learning_rate": 3.1156871253862917e-07, + "loss": 0.1604, + "step": 6366 + }, + { + "epoch": 0.33, + "learning_rate": 3.1154064924686576e-07, + "loss": 0.1853, + "step": 6367 + }, + { + "epoch": 0.33, + "learning_rate": 3.1151258276722763e-07, + "loss": 0.1388, + "step": 6368 + }, + { + "epoch": 0.33, + "learning_rate": 3.114845131005168e-07, + "loss": 0.1093, + "step": 6369 + }, + { + "epoch": 0.33, + "learning_rate": 3.1145644024753555e-07, + "loss": 0.1228, + "step": 6370 + }, + { + "epoch": 0.33, + "learning_rate": 3.1142836420908614e-07, + "loss": 0.1648, + "step": 6371 + }, + { + "epoch": 0.33, + "learning_rate": 3.114002849859711e-07, + "loss": 0.1851, + "step": 6372 + }, + { + "epoch": 0.33, + "learning_rate": 3.1137220257899293e-07, + "loss": 0.1143, + "step": 6373 + }, + { + "epoch": 0.33, + "learning_rate": 3.113441169889542e-07, + "loss": 0.1287, + "step": 6374 + }, + { + "epoch": 0.33, + "learning_rate": 3.113160282166577e-07, + "loss": 0.1408, + "step": 6375 + }, + { + "epoch": 0.33, + "learning_rate": 3.11287936262906e-07, + "loss": 0.153, + "step": 6376 + }, + { + "epoch": 0.33, + "learning_rate": 3.112598411285022e-07, + "loss": 0.1288, + "step": 6377 + }, + { + "epoch": 0.33, + "learning_rate": 3.112317428142492e-07, + "loss": 0.1429, + "step": 6378 + }, + { + "epoch": 0.33, + "learning_rate": 3.1120364132095005e-07, + "loss": 0.1727, + "step": 6379 + }, + { + "epoch": 0.33, + "learning_rate": 3.111755366494079e-07, + "loss": 0.1343, + "step": 6380 + }, + { + "epoch": 0.33, + "learning_rate": 3.11147428800426e-07, + "loss": 0.1556, + "step": 6381 + }, + { + "epoch": 0.33, + "learning_rate": 3.1111931777480767e-07, + "loss": 0.1354, + "step": 6382 + }, + { + "epoch": 0.33, + "learning_rate": 3.110912035733563e-07, + "loss": 0.1146, + "step": 6383 + }, + { + "epoch": 0.33, + "learning_rate": 3.110630861968755e-07, + "loss": 0.1664, + "step": 6384 + }, + { + "epoch": 0.33, + "learning_rate": 3.110349656461688e-07, + "loss": 0.1159, + "step": 6385 + }, + { + "epoch": 0.33, + "learning_rate": 3.110068419220399e-07, + "loss": 0.1474, + "step": 6386 + }, + { + "epoch": 0.33, + "learning_rate": 3.109787150252926e-07, + "loss": 0.1329, + "step": 6387 + }, + { + "epoch": 0.33, + "learning_rate": 3.109505849567308e-07, + "loss": 0.1503, + "step": 6388 + }, + { + "epoch": 0.33, + "learning_rate": 3.109224517171585e-07, + "loss": 0.2025, + "step": 6389 + }, + { + "epoch": 0.33, + "learning_rate": 3.108943153073796e-07, + "loss": 0.1347, + "step": 6390 + }, + { + "epoch": 0.33, + "learning_rate": 3.108661757281985e-07, + "loss": 0.1647, + "step": 6391 + }, + { + "epoch": 0.33, + "learning_rate": 3.108380329804192e-07, + "loss": 0.1348, + "step": 6392 + }, + { + "epoch": 0.33, + "learning_rate": 3.1080988706484617e-07, + "loss": 0.1459, + "step": 6393 + }, + { + "epoch": 0.33, + "learning_rate": 3.1078173798228376e-07, + "loss": 0.1378, + "step": 6394 + }, + { + "epoch": 0.33, + "learning_rate": 3.1075358573353655e-07, + "loss": 0.1491, + "step": 6395 + }, + { + "epoch": 0.33, + "learning_rate": 3.1072543031940905e-07, + "loss": 0.1008, + "step": 6396 + }, + { + "epoch": 0.33, + "learning_rate": 3.1069727174070606e-07, + "loss": 0.165, + "step": 6397 + }, + { + "epoch": 0.33, + "learning_rate": 3.1066910999823226e-07, + "loss": 0.1324, + "step": 6398 + }, + { + "epoch": 0.33, + "learning_rate": 3.1064094509279264e-07, + "loss": 0.1103, + "step": 6399 + }, + { + "epoch": 0.33, + "learning_rate": 3.1061277702519205e-07, + "loss": 0.1622, + "step": 6400 + }, + { + "epoch": 0.33, + "learning_rate": 3.1058460579623563e-07, + "loss": 0.1439, + "step": 6401 + }, + { + "epoch": 0.33, + "learning_rate": 3.105564314067284e-07, + "loss": 0.1553, + "step": 6402 + }, + { + "epoch": 0.33, + "learning_rate": 3.1052825385747573e-07, + "loss": 0.186, + "step": 6403 + }, + { + "epoch": 0.33, + "learning_rate": 3.105000731492829e-07, + "loss": 0.1442, + "step": 6404 + }, + { + "epoch": 0.33, + "learning_rate": 3.1047188928295535e-07, + "loss": 0.1312, + "step": 6405 + }, + { + "epoch": 0.33, + "learning_rate": 3.1044370225929854e-07, + "loss": 0.1928, + "step": 6406 + }, + { + "epoch": 0.33, + "learning_rate": 3.104155120791181e-07, + "loss": 0.132, + "step": 6407 + }, + { + "epoch": 0.33, + "learning_rate": 3.1038731874321966e-07, + "loss": 0.1682, + "step": 6408 + }, + { + "epoch": 0.33, + "learning_rate": 3.1035912225240907e-07, + "loss": 0.1288, + "step": 6409 + }, + { + "epoch": 0.33, + "learning_rate": 3.1033092260749215e-07, + "loss": 0.1268, + "step": 6410 + }, + { + "epoch": 0.33, + "learning_rate": 3.1030271980927485e-07, + "loss": 0.161, + "step": 6411 + }, + { + "epoch": 0.33, + "learning_rate": 3.102745138585633e-07, + "loss": 0.1383, + "step": 6412 + }, + { + "epoch": 0.33, + "learning_rate": 3.1024630475616347e-07, + "loss": 0.1538, + "step": 6413 + }, + { + "epoch": 0.33, + "learning_rate": 3.1021809250288176e-07, + "loss": 0.1181, + "step": 6414 + }, + { + "epoch": 0.33, + "learning_rate": 3.1018987709952444e-07, + "loss": 0.1328, + "step": 6415 + }, + { + "epoch": 0.33, + "learning_rate": 3.101616585468979e-07, + "loss": 0.1331, + "step": 6416 + }, + { + "epoch": 0.33, + "learning_rate": 3.1013343684580857e-07, + "loss": 0.1754, + "step": 6417 + }, + { + "epoch": 0.33, + "learning_rate": 3.101052119970631e-07, + "loss": 0.1323, + "step": 6418 + }, + { + "epoch": 0.34, + "learning_rate": 3.1007698400146817e-07, + "loss": 0.1539, + "step": 6419 + }, + { + "epoch": 0.34, + "learning_rate": 3.100487528598306e-07, + "loss": 0.1316, + "step": 6420 + }, + { + "epoch": 0.34, + "learning_rate": 3.1002051857295717e-07, + "loss": 0.1218, + "step": 6421 + }, + { + "epoch": 0.34, + "learning_rate": 3.0999228114165477e-07, + "loss": 0.121, + "step": 6422 + }, + { + "epoch": 0.34, + "learning_rate": 3.099640405667306e-07, + "loss": 0.1533, + "step": 6423 + }, + { + "epoch": 0.34, + "learning_rate": 3.099357968489917e-07, + "loss": 0.149, + "step": 6424 + }, + { + "epoch": 0.34, + "learning_rate": 3.099075499892453e-07, + "loss": 0.1183, + "step": 6425 + }, + { + "epoch": 0.34, + "learning_rate": 3.098792999882986e-07, + "loss": 0.1115, + "step": 6426 + }, + { + "epoch": 0.34, + "learning_rate": 3.098510468469591e-07, + "loss": 0.154, + "step": 6427 + }, + { + "epoch": 0.34, + "learning_rate": 3.0982279056603433e-07, + "loss": 0.1575, + "step": 6428 + }, + { + "epoch": 0.34, + "learning_rate": 3.097945311463318e-07, + "loss": 0.1661, + "step": 6429 + }, + { + "epoch": 0.34, + "learning_rate": 3.097662685886592e-07, + "loss": 0.1412, + "step": 6430 + }, + { + "epoch": 0.34, + "learning_rate": 3.0973800289382424e-07, + "loss": 0.158, + "step": 6431 + }, + { + "epoch": 0.34, + "learning_rate": 3.0970973406263483e-07, + "loss": 0.1843, + "step": 6432 + }, + { + "epoch": 0.34, + "learning_rate": 3.096814620958988e-07, + "loss": 0.1276, + "step": 6433 + }, + { + "epoch": 0.34, + "learning_rate": 3.0965318699442434e-07, + "loss": 0.1564, + "step": 6434 + }, + { + "epoch": 0.34, + "learning_rate": 3.096249087590194e-07, + "loss": 0.1321, + "step": 6435 + }, + { + "epoch": 0.34, + "learning_rate": 3.0959662739049226e-07, + "loss": 0.1378, + "step": 6436 + }, + { + "epoch": 0.34, + "learning_rate": 3.095683428896512e-07, + "loss": 0.1413, + "step": 6437 + }, + { + "epoch": 0.34, + "learning_rate": 3.095400552573046e-07, + "loss": 0.1458, + "step": 6438 + }, + { + "epoch": 0.34, + "learning_rate": 3.0951176449426095e-07, + "loss": 0.1357, + "step": 6439 + }, + { + "epoch": 0.34, + "learning_rate": 3.094834706013288e-07, + "loss": 0.1324, + "step": 6440 + }, + { + "epoch": 0.34, + "learning_rate": 3.094551735793168e-07, + "loss": 0.1528, + "step": 6441 + }, + { + "epoch": 0.34, + "learning_rate": 3.0942687342903365e-07, + "loss": 0.1758, + "step": 6442 + }, + { + "epoch": 0.34, + "learning_rate": 3.093985701512882e-07, + "loss": 0.171, + "step": 6443 + }, + { + "epoch": 0.34, + "learning_rate": 3.093702637468894e-07, + "loss": 0.1273, + "step": 6444 + }, + { + "epoch": 0.34, + "learning_rate": 3.0934195421664624e-07, + "loss": 0.1328, + "step": 6445 + }, + { + "epoch": 0.34, + "learning_rate": 3.093136415613679e-07, + "loss": 0.1448, + "step": 6446 + }, + { + "epoch": 0.34, + "learning_rate": 3.0928532578186336e-07, + "loss": 0.1349, + "step": 6447 + }, + { + "epoch": 0.34, + "learning_rate": 3.0925700687894205e-07, + "loss": 0.1137, + "step": 6448 + }, + { + "epoch": 0.34, + "learning_rate": 3.0922868485341334e-07, + "loss": 0.1503, + "step": 6449 + }, + { + "epoch": 0.34, + "learning_rate": 3.0920035970608663e-07, + "loss": 0.1529, + "step": 6450 + }, + { + "epoch": 0.34, + "learning_rate": 3.091720314377715e-07, + "loss": 0.1427, + "step": 6451 + }, + { + "epoch": 0.34, + "learning_rate": 3.091437000492775e-07, + "loss": 0.143, + "step": 6452 + }, + { + "epoch": 0.34, + "learning_rate": 3.0911536554141454e-07, + "loss": 0.1233, + "step": 6453 + }, + { + "epoch": 0.34, + "learning_rate": 3.0908702791499224e-07, + "loss": 0.1715, + "step": 6454 + }, + { + "epoch": 0.34, + "learning_rate": 3.090586871708206e-07, + "loss": 0.1448, + "step": 6455 + }, + { + "epoch": 0.34, + "learning_rate": 3.090303433097096e-07, + "loss": 0.1046, + "step": 6456 + }, + { + "epoch": 0.34, + "learning_rate": 3.090019963324693e-07, + "loss": 0.1439, + "step": 6457 + }, + { + "epoch": 0.34, + "learning_rate": 3.089736462399098e-07, + "loss": 0.1773, + "step": 6458 + }, + { + "epoch": 0.34, + "learning_rate": 3.0894529303284137e-07, + "loss": 0.1323, + "step": 6459 + }, + { + "epoch": 0.34, + "learning_rate": 3.0891693671207454e-07, + "loss": 0.1537, + "step": 6460 + }, + { + "epoch": 0.34, + "learning_rate": 3.088885772784195e-07, + "loss": 0.1301, + "step": 6461 + }, + { + "epoch": 0.34, + "learning_rate": 3.0886021473268695e-07, + "loss": 0.1371, + "step": 6462 + }, + { + "epoch": 0.34, + "learning_rate": 3.088318490756874e-07, + "loss": 0.1512, + "step": 6463 + }, + { + "epoch": 0.34, + "learning_rate": 3.088034803082317e-07, + "loss": 0.1559, + "step": 6464 + }, + { + "epoch": 0.34, + "learning_rate": 3.0877510843113043e-07, + "loss": 0.1529, + "step": 6465 + }, + { + "epoch": 0.34, + "learning_rate": 3.087467334451946e-07, + "loss": 0.133, + "step": 6466 + }, + { + "epoch": 0.34, + "learning_rate": 3.087183553512351e-07, + "loss": 0.1747, + "step": 6467 + }, + { + "epoch": 0.34, + "learning_rate": 3.086899741500631e-07, + "loss": 0.1804, + "step": 6468 + }, + { + "epoch": 0.34, + "learning_rate": 3.086615898424897e-07, + "loss": 0.124, + "step": 6469 + }, + { + "epoch": 0.34, + "learning_rate": 3.08633202429326e-07, + "loss": 0.131, + "step": 6470 + }, + { + "epoch": 0.34, + "learning_rate": 3.0860481191138355e-07, + "loss": 0.1116, + "step": 6471 + }, + { + "epoch": 0.34, + "learning_rate": 3.085764182894736e-07, + "loss": 0.1102, + "step": 6472 + }, + { + "epoch": 0.34, + "learning_rate": 3.085480215644077e-07, + "loss": 0.169, + "step": 6473 + }, + { + "epoch": 0.34, + "learning_rate": 3.085196217369975e-07, + "loss": 0.1699, + "step": 6474 + }, + { + "epoch": 0.34, + "learning_rate": 3.0849121880805456e-07, + "loss": 0.1374, + "step": 6475 + }, + { + "epoch": 0.34, + "learning_rate": 3.084628127783907e-07, + "loss": 0.1645, + "step": 6476 + }, + { + "epoch": 0.34, + "learning_rate": 3.0843440364881776e-07, + "loss": 0.1475, + "step": 6477 + }, + { + "epoch": 0.34, + "learning_rate": 3.084059914201478e-07, + "loss": 0.1696, + "step": 6478 + }, + { + "epoch": 0.34, + "learning_rate": 3.0837757609319264e-07, + "loss": 0.1262, + "step": 6479 + }, + { + "epoch": 0.34, + "learning_rate": 3.0834915766876453e-07, + "loss": 0.1469, + "step": 6480 + }, + { + "epoch": 0.34, + "learning_rate": 3.0832073614767577e-07, + "loss": 0.1195, + "step": 6481 + }, + { + "epoch": 0.34, + "learning_rate": 3.082923115307385e-07, + "loss": 0.1206, + "step": 6482 + }, + { + "epoch": 0.34, + "learning_rate": 3.082638838187651e-07, + "loss": 0.148, + "step": 6483 + }, + { + "epoch": 0.34, + "learning_rate": 3.0823545301256815e-07, + "loss": 0.1024, + "step": 6484 + }, + { + "epoch": 0.34, + "learning_rate": 3.082070191129602e-07, + "loss": 0.1428, + "step": 6485 + }, + { + "epoch": 0.34, + "learning_rate": 3.0817858212075385e-07, + "loss": 0.1424, + "step": 6486 + }, + { + "epoch": 0.34, + "learning_rate": 3.0815014203676187e-07, + "loss": 0.1405, + "step": 6487 + }, + { + "epoch": 0.34, + "learning_rate": 3.081216988617971e-07, + "loss": 0.1506, + "step": 6488 + }, + { + "epoch": 0.34, + "learning_rate": 3.080932525966724e-07, + "loss": 0.1499, + "step": 6489 + }, + { + "epoch": 0.34, + "learning_rate": 3.080648032422008e-07, + "loss": 0.1313, + "step": 6490 + }, + { + "epoch": 0.34, + "learning_rate": 3.0803635079919543e-07, + "loss": 0.1255, + "step": 6491 + }, + { + "epoch": 0.34, + "learning_rate": 3.080078952684694e-07, + "loss": 0.1156, + "step": 6492 + }, + { + "epoch": 0.34, + "learning_rate": 3.079794366508361e-07, + "loss": 0.14, + "step": 6493 + }, + { + "epoch": 0.34, + "learning_rate": 3.079509749471088e-07, + "loss": 0.1364, + "step": 6494 + }, + { + "epoch": 0.34, + "learning_rate": 3.0792251015810094e-07, + "loss": 0.1388, + "step": 6495 + }, + { + "epoch": 0.34, + "learning_rate": 3.0789404228462614e-07, + "loss": 0.1617, + "step": 6496 + }, + { + "epoch": 0.34, + "learning_rate": 3.0786557132749787e-07, + "loss": 0.1435, + "step": 6497 + }, + { + "epoch": 0.34, + "learning_rate": 3.0783709728753003e-07, + "loss": 0.1575, + "step": 6498 + }, + { + "epoch": 0.34, + "learning_rate": 3.0780862016553626e-07, + "loss": 0.1456, + "step": 6499 + }, + { + "epoch": 0.34, + "learning_rate": 3.077801399623305e-07, + "loss": 0.1394, + "step": 6500 + }, + { + "epoch": 0.34, + "learning_rate": 3.0775165667872666e-07, + "loss": 0.1374, + "step": 6501 + }, + { + "epoch": 0.34, + "learning_rate": 3.07723170315539e-07, + "loss": 0.1276, + "step": 6502 + }, + { + "epoch": 0.34, + "learning_rate": 3.0769468087358156e-07, + "loss": 0.1401, + "step": 6503 + }, + { + "epoch": 0.34, + "learning_rate": 3.076661883536685e-07, + "loss": 0.1511, + "step": 6504 + }, + { + "epoch": 0.34, + "learning_rate": 3.0763769275661415e-07, + "loss": 0.1363, + "step": 6505 + }, + { + "epoch": 0.34, + "learning_rate": 3.076091940832331e-07, + "loss": 0.1367, + "step": 6506 + }, + { + "epoch": 0.34, + "learning_rate": 3.075806923343397e-07, + "loss": 0.16, + "step": 6507 + }, + { + "epoch": 0.34, + "learning_rate": 3.075521875107486e-07, + "loss": 0.176, + "step": 6508 + }, + { + "epoch": 0.34, + "learning_rate": 3.0752367961327445e-07, + "loss": 0.1459, + "step": 6509 + }, + { + "epoch": 0.34, + "learning_rate": 3.07495168642732e-07, + "loss": 0.158, + "step": 6510 + }, + { + "epoch": 0.34, + "learning_rate": 3.0746665459993615e-07, + "loss": 0.1298, + "step": 6511 + }, + { + "epoch": 0.34, + "learning_rate": 3.074381374857019e-07, + "loss": 0.1333, + "step": 6512 + }, + { + "epoch": 0.34, + "learning_rate": 3.074096173008441e-07, + "loss": 0.1054, + "step": 6513 + }, + { + "epoch": 0.34, + "learning_rate": 3.07381094046178e-07, + "loss": 0.1467, + "step": 6514 + }, + { + "epoch": 0.34, + "learning_rate": 3.0735256772251884e-07, + "loss": 0.1299, + "step": 6515 + }, + { + "epoch": 0.34, + "learning_rate": 3.0732403833068183e-07, + "loss": 0.1559, + "step": 6516 + }, + { + "epoch": 0.34, + "learning_rate": 3.072955058714823e-07, + "loss": 0.1373, + "step": 6517 + }, + { + "epoch": 0.34, + "learning_rate": 3.072669703457358e-07, + "loss": 0.1379, + "step": 6518 + }, + { + "epoch": 0.34, + "learning_rate": 3.0723843175425803e-07, + "loss": 0.1967, + "step": 6519 + }, + { + "epoch": 0.34, + "learning_rate": 3.072098900978643e-07, + "loss": 0.1474, + "step": 6520 + }, + { + "epoch": 0.34, + "learning_rate": 3.0718134537737063e-07, + "loss": 0.1361, + "step": 6521 + }, + { + "epoch": 0.34, + "learning_rate": 3.071527975935926e-07, + "loss": 0.1227, + "step": 6522 + }, + { + "epoch": 0.34, + "learning_rate": 3.071242467473464e-07, + "loss": 0.1469, + "step": 6523 + }, + { + "epoch": 0.34, + "learning_rate": 3.0709569283944785e-07, + "loss": 0.1403, + "step": 6524 + }, + { + "epoch": 0.34, + "learning_rate": 3.0706713587071307e-07, + "loss": 0.1398, + "step": 6525 + }, + { + "epoch": 0.34, + "learning_rate": 3.070385758419581e-07, + "loss": 0.1553, + "step": 6526 + }, + { + "epoch": 0.34, + "learning_rate": 3.070100127539994e-07, + "loss": 0.1088, + "step": 6527 + }, + { + "epoch": 0.34, + "learning_rate": 3.0698144660765325e-07, + "loss": 0.1596, + "step": 6528 + }, + { + "epoch": 0.34, + "learning_rate": 3.0695287740373596e-07, + "loss": 0.1491, + "step": 6529 + }, + { + "epoch": 0.34, + "learning_rate": 3.069243051430642e-07, + "loss": 0.1256, + "step": 6530 + }, + { + "epoch": 0.34, + "learning_rate": 3.0689572982645454e-07, + "loss": 0.1264, + "step": 6531 + }, + { + "epoch": 0.34, + "learning_rate": 3.068671514547237e-07, + "loss": 0.1328, + "step": 6532 + }, + { + "epoch": 0.34, + "learning_rate": 3.0683857002868833e-07, + "loss": 0.1395, + "step": 6533 + }, + { + "epoch": 0.34, + "learning_rate": 3.0680998554916537e-07, + "loss": 0.1814, + "step": 6534 + }, + { + "epoch": 0.34, + "learning_rate": 3.0678139801697183e-07, + "loss": 0.139, + "step": 6535 + }, + { + "epoch": 0.34, + "learning_rate": 3.067528074329247e-07, + "loss": 0.1464, + "step": 6536 + }, + { + "epoch": 0.34, + "learning_rate": 3.067242137978412e-07, + "loss": 0.1223, + "step": 6537 + }, + { + "epoch": 0.34, + "learning_rate": 3.0669561711253833e-07, + "loss": 0.1534, + "step": 6538 + }, + { + "epoch": 0.34, + "learning_rate": 3.0666701737783367e-07, + "loss": 0.1308, + "step": 6539 + }, + { + "epoch": 0.34, + "learning_rate": 3.066384145945444e-07, + "loss": 0.1709, + "step": 6540 + }, + { + "epoch": 0.34, + "learning_rate": 3.0660980876348814e-07, + "loss": 0.1277, + "step": 6541 + }, + { + "epoch": 0.34, + "learning_rate": 3.0658119988548225e-07, + "loss": 0.1718, + "step": 6542 + }, + { + "epoch": 0.34, + "learning_rate": 3.0655258796134467e-07, + "loss": 0.1988, + "step": 6543 + }, + { + "epoch": 0.34, + "learning_rate": 3.0652397299189287e-07, + "loss": 0.1089, + "step": 6544 + }, + { + "epoch": 0.34, + "learning_rate": 3.0649535497794494e-07, + "loss": 0.1641, + "step": 6545 + }, + { + "epoch": 0.34, + "learning_rate": 3.0646673392031853e-07, + "loss": 0.1237, + "step": 6546 + }, + { + "epoch": 0.34, + "learning_rate": 3.064381098198318e-07, + "loss": 0.1258, + "step": 6547 + }, + { + "epoch": 0.34, + "learning_rate": 3.064094826773028e-07, + "loss": 0.1229, + "step": 6548 + }, + { + "epoch": 0.34, + "learning_rate": 3.063808524935498e-07, + "loss": 0.1329, + "step": 6549 + }, + { + "epoch": 0.34, + "learning_rate": 3.063522192693909e-07, + "loss": 0.1248, + "step": 6550 + }, + { + "epoch": 0.34, + "learning_rate": 3.0632358300564445e-07, + "loss": 0.1183, + "step": 6551 + }, + { + "epoch": 0.34, + "learning_rate": 3.0629494370312905e-07, + "loss": 0.1687, + "step": 6552 + }, + { + "epoch": 0.34, + "learning_rate": 3.062663013626631e-07, + "loss": 0.1093, + "step": 6553 + }, + { + "epoch": 0.34, + "learning_rate": 3.062376559850652e-07, + "loss": 0.1481, + "step": 6554 + }, + { + "epoch": 0.34, + "learning_rate": 3.0620900757115414e-07, + "loss": 0.1379, + "step": 6555 + }, + { + "epoch": 0.34, + "learning_rate": 3.0618035612174857e-07, + "loss": 0.1936, + "step": 6556 + }, + { + "epoch": 0.34, + "learning_rate": 3.061517016376674e-07, + "loss": 0.1429, + "step": 6557 + }, + { + "epoch": 0.34, + "learning_rate": 3.061230441197297e-07, + "loss": 0.174, + "step": 6558 + }, + { + "epoch": 0.34, + "learning_rate": 3.060943835687545e-07, + "loss": 0.1325, + "step": 6559 + }, + { + "epoch": 0.34, + "learning_rate": 3.0606571998556076e-07, + "loss": 0.1358, + "step": 6560 + }, + { + "epoch": 0.34, + "learning_rate": 3.060370533709678e-07, + "loss": 0.1401, + "step": 6561 + }, + { + "epoch": 0.34, + "learning_rate": 3.06008383725795e-07, + "loss": 0.1198, + "step": 6562 + }, + { + "epoch": 0.34, + "learning_rate": 3.059797110508616e-07, + "loss": 0.1366, + "step": 6563 + }, + { + "epoch": 0.34, + "learning_rate": 3.0595103534698705e-07, + "loss": 0.1178, + "step": 6564 + }, + { + "epoch": 0.34, + "learning_rate": 3.0592235661499113e-07, + "loss": 0.1594, + "step": 6565 + }, + { + "epoch": 0.34, + "learning_rate": 3.0589367485569336e-07, + "loss": 0.146, + "step": 6566 + }, + { + "epoch": 0.34, + "learning_rate": 3.058649900699135e-07, + "loss": 0.1707, + "step": 6567 + }, + { + "epoch": 0.34, + "learning_rate": 3.058363022584713e-07, + "loss": 0.1566, + "step": 6568 + }, + { + "epoch": 0.34, + "learning_rate": 3.058076114221867e-07, + "loss": 0.1495, + "step": 6569 + }, + { + "epoch": 0.34, + "learning_rate": 3.057789175618798e-07, + "loss": 0.1569, + "step": 6570 + }, + { + "epoch": 0.34, + "learning_rate": 3.057502206783706e-07, + "loss": 0.1262, + "step": 6571 + }, + { + "epoch": 0.34, + "learning_rate": 3.057215207724792e-07, + "loss": 0.1508, + "step": 6572 + }, + { + "epoch": 0.34, + "learning_rate": 3.05692817845026e-07, + "loss": 0.171, + "step": 6573 + }, + { + "epoch": 0.34, + "learning_rate": 3.056641118968312e-07, + "loss": 0.1538, + "step": 6574 + }, + { + "epoch": 0.34, + "learning_rate": 3.056354029287154e-07, + "loss": 0.1502, + "step": 6575 + }, + { + "epoch": 0.34, + "learning_rate": 3.0560669094149883e-07, + "loss": 0.168, + "step": 6576 + }, + { + "epoch": 0.34, + "learning_rate": 3.055779759360024e-07, + "loss": 0.1055, + "step": 6577 + }, + { + "epoch": 0.34, + "learning_rate": 3.055492579130466e-07, + "loss": 0.1354, + "step": 6578 + }, + { + "epoch": 0.34, + "learning_rate": 3.055205368734523e-07, + "loss": 0.1128, + "step": 6579 + }, + { + "epoch": 0.34, + "learning_rate": 3.054918128180403e-07, + "loss": 0.1413, + "step": 6580 + }, + { + "epoch": 0.34, + "learning_rate": 3.054630857476316e-07, + "loss": 0.1317, + "step": 6581 + }, + { + "epoch": 0.34, + "learning_rate": 3.054343556630471e-07, + "loss": 0.1549, + "step": 6582 + }, + { + "epoch": 0.34, + "learning_rate": 3.0540562256510807e-07, + "loss": 0.1674, + "step": 6583 + }, + { + "epoch": 0.34, + "learning_rate": 3.0537688645463575e-07, + "loss": 0.1563, + "step": 6584 + }, + { + "epoch": 0.34, + "learning_rate": 3.053481473324512e-07, + "loss": 0.1132, + "step": 6585 + }, + { + "epoch": 0.34, + "learning_rate": 3.05319405199376e-07, + "loss": 0.1577, + "step": 6586 + }, + { + "epoch": 0.34, + "learning_rate": 3.052906600562315e-07, + "loss": 0.1535, + "step": 6587 + }, + { + "epoch": 0.34, + "learning_rate": 3.052619119038394e-07, + "loss": 0.1434, + "step": 6588 + }, + { + "epoch": 0.34, + "learning_rate": 3.052331607430211e-07, + "loss": 0.1423, + "step": 6589 + }, + { + "epoch": 0.34, + "learning_rate": 3.0520440657459846e-07, + "loss": 0.1354, + "step": 6590 + }, + { + "epoch": 0.34, + "learning_rate": 3.0517564939939327e-07, + "loss": 0.1585, + "step": 6591 + }, + { + "epoch": 0.34, + "learning_rate": 3.0514688921822744e-07, + "loss": 0.1516, + "step": 6592 + }, + { + "epoch": 0.34, + "learning_rate": 3.051181260319229e-07, + "loss": 0.1029, + "step": 6593 + }, + { + "epoch": 0.34, + "learning_rate": 3.0508935984130176e-07, + "loss": 0.1034, + "step": 6594 + }, + { + "epoch": 0.34, + "learning_rate": 3.0506059064718614e-07, + "loss": 0.1273, + "step": 6595 + }, + { + "epoch": 0.34, + "learning_rate": 3.0503181845039834e-07, + "loss": 0.2144, + "step": 6596 + }, + { + "epoch": 0.34, + "learning_rate": 3.0500304325176055e-07, + "loss": 0.1898, + "step": 6597 + }, + { + "epoch": 0.34, + "learning_rate": 3.0497426505209525e-07, + "loss": 0.1548, + "step": 6598 + }, + { + "epoch": 0.34, + "learning_rate": 3.0494548385222496e-07, + "loss": 0.1649, + "step": 6599 + }, + { + "epoch": 0.34, + "learning_rate": 3.0491669965297224e-07, + "loss": 0.1132, + "step": 6600 + }, + { + "epoch": 0.34, + "learning_rate": 3.048879124551597e-07, + "loss": 0.1297, + "step": 6601 + }, + { + "epoch": 0.34, + "learning_rate": 3.0485912225961013e-07, + "loss": 0.1395, + "step": 6602 + }, + { + "epoch": 0.34, + "learning_rate": 3.0483032906714635e-07, + "loss": 0.152, + "step": 6603 + }, + { + "epoch": 0.34, + "learning_rate": 3.0480153287859136e-07, + "loss": 0.1491, + "step": 6604 + }, + { + "epoch": 0.34, + "learning_rate": 3.0477273369476814e-07, + "loss": 0.147, + "step": 6605 + }, + { + "epoch": 0.34, + "learning_rate": 3.0474393151649965e-07, + "loss": 0.1548, + "step": 6606 + }, + { + "epoch": 0.34, + "learning_rate": 3.047151263446092e-07, + "loss": 0.1161, + "step": 6607 + }, + { + "epoch": 0.34, + "learning_rate": 3.0468631817991996e-07, + "loss": 0.1414, + "step": 6608 + }, + { + "epoch": 0.34, + "learning_rate": 3.0465750702325543e-07, + "loss": 0.1484, + "step": 6609 + }, + { + "epoch": 0.35, + "learning_rate": 3.046286928754389e-07, + "loss": 0.1415, + "step": 6610 + }, + { + "epoch": 0.35, + "learning_rate": 3.0459987573729394e-07, + "loss": 0.1312, + "step": 6611 + }, + { + "epoch": 0.35, + "learning_rate": 3.045710556096441e-07, + "loss": 0.1492, + "step": 6612 + }, + { + "epoch": 0.35, + "learning_rate": 3.045422324933132e-07, + "loss": 0.1409, + "step": 6613 + }, + { + "epoch": 0.35, + "learning_rate": 3.0451340638912486e-07, + "loss": 0.1012, + "step": 6614 + }, + { + "epoch": 0.35, + "learning_rate": 3.0448457729790315e-07, + "loss": 0.1337, + "step": 6615 + }, + { + "epoch": 0.35, + "learning_rate": 3.044557452204717e-07, + "loss": 0.1268, + "step": 6616 + }, + { + "epoch": 0.35, + "learning_rate": 3.0442691015765485e-07, + "loss": 0.1297, + "step": 6617 + }, + { + "epoch": 0.35, + "learning_rate": 3.043980721102766e-07, + "loss": 0.1384, + "step": 6618 + }, + { + "epoch": 0.35, + "learning_rate": 3.0436923107916114e-07, + "loss": 0.1662, + "step": 6619 + }, + { + "epoch": 0.35, + "learning_rate": 3.043403870651327e-07, + "loss": 0.1351, + "step": 6620 + }, + { + "epoch": 0.35, + "learning_rate": 3.043115400690158e-07, + "loss": 0.1224, + "step": 6621 + }, + { + "epoch": 0.35, + "learning_rate": 3.0428269009163476e-07, + "loss": 0.1335, + "step": 6622 + }, + { + "epoch": 0.35, + "learning_rate": 3.042538371338142e-07, + "loss": 0.1537, + "step": 6623 + }, + { + "epoch": 0.35, + "learning_rate": 3.0422498119637874e-07, + "loss": 0.1147, + "step": 6624 + }, + { + "epoch": 0.35, + "learning_rate": 3.041961222801531e-07, + "loss": 0.141, + "step": 6625 + }, + { + "epoch": 0.35, + "learning_rate": 3.04167260385962e-07, + "loss": 0.1668, + "step": 6626 + }, + { + "epoch": 0.35, + "learning_rate": 3.0413839551463045e-07, + "loss": 0.132, + "step": 6627 + }, + { + "epoch": 0.35, + "learning_rate": 3.041095276669833e-07, + "loss": 0.1853, + "step": 6628 + }, + { + "epoch": 0.35, + "learning_rate": 3.0408065684384567e-07, + "loss": 0.1621, + "step": 6629 + }, + { + "epoch": 0.35, + "learning_rate": 3.0405178304604273e-07, + "loss": 0.1175, + "step": 6630 + }, + { + "epoch": 0.35, + "learning_rate": 3.0402290627439966e-07, + "loss": 0.185, + "step": 6631 + }, + { + "epoch": 0.35, + "learning_rate": 3.039940265297417e-07, + "loss": 0.1297, + "step": 6632 + }, + { + "epoch": 0.35, + "learning_rate": 3.039651438128944e-07, + "loss": 0.1635, + "step": 6633 + }, + { + "epoch": 0.35, + "learning_rate": 3.039362581246831e-07, + "loss": 0.1209, + "step": 6634 + }, + { + "epoch": 0.35, + "learning_rate": 3.0390736946593346e-07, + "loss": 0.1479, + "step": 6635 + }, + { + "epoch": 0.35, + "learning_rate": 3.03878477837471e-07, + "loss": 0.1428, + "step": 6636 + }, + { + "epoch": 0.35, + "learning_rate": 3.0384958324012165e-07, + "loss": 0.1648, + "step": 6637 + }, + { + "epoch": 0.35, + "learning_rate": 3.0382068567471106e-07, + "loss": 0.1297, + "step": 6638 + }, + { + "epoch": 0.35, + "learning_rate": 3.037917851420652e-07, + "loss": 0.1346, + "step": 6639 + }, + { + "epoch": 0.35, + "learning_rate": 3.037628816430101e-07, + "loss": 0.1433, + "step": 6640 + }, + { + "epoch": 0.35, + "learning_rate": 3.037339751783717e-07, + "loss": 0.1415, + "step": 6641 + }, + { + "epoch": 0.35, + "learning_rate": 3.0370506574897634e-07, + "loss": 0.1858, + "step": 6642 + }, + { + "epoch": 0.35, + "learning_rate": 3.0367615335565013e-07, + "loss": 0.1347, + "step": 6643 + }, + { + "epoch": 0.35, + "learning_rate": 3.0364723799921944e-07, + "loss": 0.1984, + "step": 6644 + }, + { + "epoch": 0.35, + "learning_rate": 3.0361831968051067e-07, + "loss": 0.1519, + "step": 6645 + }, + { + "epoch": 0.35, + "learning_rate": 3.035893984003503e-07, + "loss": 0.157, + "step": 6646 + }, + { + "epoch": 0.35, + "learning_rate": 3.035604741595649e-07, + "loss": 0.1283, + "step": 6647 + }, + { + "epoch": 0.35, + "learning_rate": 3.0353154695898123e-07, + "loss": 0.1426, + "step": 6648 + }, + { + "epoch": 0.35, + "learning_rate": 3.0350261679942604e-07, + "loss": 0.1581, + "step": 6649 + }, + { + "epoch": 0.35, + "learning_rate": 3.0347368368172597e-07, + "loss": 0.1674, + "step": 6650 + }, + { + "epoch": 0.35, + "learning_rate": 3.034447476067082e-07, + "loss": 0.1749, + "step": 6651 + }, + { + "epoch": 0.35, + "learning_rate": 3.034158085751995e-07, + "loss": 0.1309, + "step": 6652 + }, + { + "epoch": 0.35, + "learning_rate": 3.033868665880272e-07, + "loss": 0.135, + "step": 6653 + }, + { + "epoch": 0.35, + "learning_rate": 3.0335792164601817e-07, + "loss": 0.1401, + "step": 6654 + }, + { + "epoch": 0.35, + "learning_rate": 3.0332897374999996e-07, + "loss": 0.123, + "step": 6655 + }, + { + "epoch": 0.35, + "learning_rate": 3.033000229007998e-07, + "loss": 0.098, + "step": 6656 + }, + { + "epoch": 0.35, + "learning_rate": 3.0327106909924505e-07, + "loss": 0.1615, + "step": 6657 + }, + { + "epoch": 0.35, + "learning_rate": 3.032421123461633e-07, + "loss": 0.1098, + "step": 6658 + }, + { + "epoch": 0.35, + "learning_rate": 3.0321315264238203e-07, + "loss": 0.1273, + "step": 6659 + }, + { + "epoch": 0.35, + "learning_rate": 3.0318418998872917e-07, + "loss": 0.1264, + "step": 6660 + }, + { + "epoch": 0.35, + "learning_rate": 3.0315522438603226e-07, + "loss": 0.1254, + "step": 6661 + }, + { + "epoch": 0.35, + "learning_rate": 3.031262558351193e-07, + "loss": 0.1717, + "step": 6662 + }, + { + "epoch": 0.35, + "learning_rate": 3.03097284336818e-07, + "loss": 0.1202, + "step": 6663 + }, + { + "epoch": 0.35, + "learning_rate": 3.0306830989195654e-07, + "loss": 0.1184, + "step": 6664 + }, + { + "epoch": 0.35, + "learning_rate": 3.0303933250136304e-07, + "loss": 0.133, + "step": 6665 + }, + { + "epoch": 0.35, + "learning_rate": 3.0301035216586564e-07, + "loss": 0.1099, + "step": 6666 + }, + { + "epoch": 0.35, + "learning_rate": 3.0298136888629257e-07, + "loss": 0.1564, + "step": 6667 + }, + { + "epoch": 0.35, + "learning_rate": 3.0295238266347223e-07, + "loss": 0.1968, + "step": 6668 + }, + { + "epoch": 0.35, + "learning_rate": 3.029233934982331e-07, + "loss": 0.1483, + "step": 6669 + }, + { + "epoch": 0.35, + "learning_rate": 3.028944013914036e-07, + "loss": 0.1394, + "step": 6670 + }, + { + "epoch": 0.35, + "learning_rate": 3.0286540634381244e-07, + "loss": 0.1217, + "step": 6671 + }, + { + "epoch": 0.35, + "learning_rate": 3.028364083562882e-07, + "loss": 0.1182, + "step": 6672 + }, + { + "epoch": 0.35, + "learning_rate": 3.0280740742965975e-07, + "loss": 0.1365, + "step": 6673 + }, + { + "epoch": 0.35, + "learning_rate": 3.0277840356475593e-07, + "loss": 0.1469, + "step": 6674 + }, + { + "epoch": 0.35, + "learning_rate": 3.027493967624056e-07, + "loss": 0.1475, + "step": 6675 + }, + { + "epoch": 0.35, + "learning_rate": 3.027203870234378e-07, + "loss": 0.1396, + "step": 6676 + }, + { + "epoch": 0.35, + "learning_rate": 3.026913743486818e-07, + "loss": 0.1293, + "step": 6677 + }, + { + "epoch": 0.35, + "learning_rate": 3.0266235873896665e-07, + "loss": 0.1416, + "step": 6678 + }, + { + "epoch": 0.35, + "learning_rate": 3.0263334019512155e-07, + "loss": 0.1451, + "step": 6679 + }, + { + "epoch": 0.35, + "learning_rate": 3.026043187179761e-07, + "loss": 0.1743, + "step": 6680 + }, + { + "epoch": 0.35, + "learning_rate": 3.0257529430835946e-07, + "loss": 0.1214, + "step": 6681 + }, + { + "epoch": 0.35, + "learning_rate": 3.0254626696710154e-07, + "loss": 0.1366, + "step": 6682 + }, + { + "epoch": 0.35, + "learning_rate": 3.025172366950316e-07, + "loss": 0.1804, + "step": 6683 + }, + { + "epoch": 0.35, + "learning_rate": 3.0248820349297937e-07, + "loss": 0.1715, + "step": 6684 + }, + { + "epoch": 0.35, + "learning_rate": 3.0245916736177483e-07, + "loss": 0.1525, + "step": 6685 + }, + { + "epoch": 0.35, + "learning_rate": 3.024301283022477e-07, + "loss": 0.1379, + "step": 6686 + }, + { + "epoch": 0.35, + "learning_rate": 3.02401086315228e-07, + "loss": 0.1658, + "step": 6687 + }, + { + "epoch": 0.35, + "learning_rate": 3.023720414015456e-07, + "loss": 0.1369, + "step": 6688 + }, + { + "epoch": 0.35, + "learning_rate": 3.023429935620309e-07, + "loss": 0.1578, + "step": 6689 + }, + { + "epoch": 0.35, + "learning_rate": 3.0231394279751383e-07, + "loss": 0.1323, + "step": 6690 + }, + { + "epoch": 0.35, + "learning_rate": 3.022848891088249e-07, + "loss": 0.1462, + "step": 6691 + }, + { + "epoch": 0.35, + "learning_rate": 3.022558324967942e-07, + "loss": 0.1288, + "step": 6692 + }, + { + "epoch": 0.35, + "learning_rate": 3.022267729622524e-07, + "loss": 0.1486, + "step": 6693 + }, + { + "epoch": 0.35, + "learning_rate": 3.0219771050603e-07, + "loss": 0.1221, + "step": 6694 + }, + { + "epoch": 0.35, + "learning_rate": 3.021686451289575e-07, + "loss": 0.1159, + "step": 6695 + }, + { + "epoch": 0.35, + "learning_rate": 3.0213957683186574e-07, + "loss": 0.1253, + "step": 6696 + }, + { + "epoch": 0.35, + "learning_rate": 3.0211050561558546e-07, + "loss": 0.163, + "step": 6697 + }, + { + "epoch": 0.35, + "learning_rate": 3.0208143148094744e-07, + "loss": 0.1464, + "step": 6698 + }, + { + "epoch": 0.35, + "learning_rate": 3.020523544287828e-07, + "loss": 0.1467, + "step": 6699 + }, + { + "epoch": 0.35, + "learning_rate": 3.020232744599224e-07, + "loss": 0.1327, + "step": 6700 + }, + { + "epoch": 0.35, + "learning_rate": 3.019941915751974e-07, + "loss": 0.1157, + "step": 6701 + }, + { + "epoch": 0.35, + "learning_rate": 3.019651057754391e-07, + "loss": 0.1427, + "step": 6702 + }, + { + "epoch": 0.35, + "learning_rate": 3.019360170614787e-07, + "loss": 0.1433, + "step": 6703 + }, + { + "epoch": 0.35, + "learning_rate": 3.019069254341476e-07, + "loss": 0.1189, + "step": 6704 + }, + { + "epoch": 0.35, + "learning_rate": 3.018778308942772e-07, + "loss": 0.1434, + "step": 6705 + }, + { + "epoch": 0.35, + "learning_rate": 3.0184873344269905e-07, + "loss": 0.1246, + "step": 6706 + }, + { + "epoch": 0.35, + "learning_rate": 3.018196330802448e-07, + "loss": 0.1369, + "step": 6707 + }, + { + "epoch": 0.35, + "learning_rate": 3.017905298077461e-07, + "loss": 0.1631, + "step": 6708 + }, + { + "epoch": 0.35, + "learning_rate": 3.017614236260348e-07, + "loss": 0.1248, + "step": 6709 + }, + { + "epoch": 0.35, + "learning_rate": 3.017323145359427e-07, + "loss": 0.1167, + "step": 6710 + }, + { + "epoch": 0.35, + "learning_rate": 3.0170320253830186e-07, + "loss": 0.131, + "step": 6711 + }, + { + "epoch": 0.35, + "learning_rate": 3.016740876339441e-07, + "loss": 0.1497, + "step": 6712 + }, + { + "epoch": 0.35, + "learning_rate": 3.016449698237018e-07, + "loss": 0.1525, + "step": 6713 + }, + { + "epoch": 0.35, + "learning_rate": 3.0161584910840693e-07, + "loss": 0.1282, + "step": 6714 + }, + { + "epoch": 0.35, + "learning_rate": 3.015867254888919e-07, + "loss": 0.1082, + "step": 6715 + }, + { + "epoch": 0.35, + "learning_rate": 3.0155759896598907e-07, + "loss": 0.1729, + "step": 6716 + }, + { + "epoch": 0.35, + "learning_rate": 3.0152846954053077e-07, + "loss": 0.1676, + "step": 6717 + }, + { + "epoch": 0.35, + "learning_rate": 3.014993372133497e-07, + "loss": 0.1319, + "step": 6718 + }, + { + "epoch": 0.35, + "learning_rate": 3.014702019852783e-07, + "loss": 0.1217, + "step": 6719 + }, + { + "epoch": 0.35, + "learning_rate": 3.0144106385714945e-07, + "loss": 0.157, + "step": 6720 + }, + { + "epoch": 0.35, + "learning_rate": 3.014119228297958e-07, + "loss": 0.1467, + "step": 6721 + }, + { + "epoch": 0.35, + "learning_rate": 3.0138277890405026e-07, + "loss": 0.1551, + "step": 6722 + }, + { + "epoch": 0.35, + "learning_rate": 3.013536320807457e-07, + "loss": 0.1301, + "step": 6723 + }, + { + "epoch": 0.35, + "learning_rate": 3.013244823607153e-07, + "loss": 0.1351, + "step": 6724 + }, + { + "epoch": 0.35, + "learning_rate": 3.0129532974479197e-07, + "loss": 0.1327, + "step": 6725 + }, + { + "epoch": 0.35, + "learning_rate": 3.012661742338091e-07, + "loss": 0.1161, + "step": 6726 + }, + { + "epoch": 0.35, + "learning_rate": 3.0123701582859986e-07, + "loss": 0.1532, + "step": 6727 + }, + { + "epoch": 0.35, + "learning_rate": 3.012078545299976e-07, + "loss": 0.12, + "step": 6728 + }, + { + "epoch": 0.35, + "learning_rate": 3.0117869033883585e-07, + "loss": 0.1442, + "step": 6729 + }, + { + "epoch": 0.35, + "learning_rate": 3.01149523255948e-07, + "loss": 0.1706, + "step": 6730 + }, + { + "epoch": 0.35, + "learning_rate": 3.0112035328216774e-07, + "loss": 0.1125, + "step": 6731 + }, + { + "epoch": 0.35, + "learning_rate": 3.0109118041832876e-07, + "loss": 0.1358, + "step": 6732 + }, + { + "epoch": 0.35, + "learning_rate": 3.010620046652648e-07, + "loss": 0.1298, + "step": 6733 + }, + { + "epoch": 0.35, + "learning_rate": 3.010328260238098e-07, + "loss": 0.0858, + "step": 6734 + }, + { + "epoch": 0.35, + "learning_rate": 3.0100364449479757e-07, + "loss": 0.124, + "step": 6735 + }, + { + "epoch": 0.35, + "learning_rate": 3.009744600790622e-07, + "loss": 0.1406, + "step": 6736 + }, + { + "epoch": 0.35, + "learning_rate": 3.0094527277743776e-07, + "loss": 0.1591, + "step": 6737 + }, + { + "epoch": 0.35, + "learning_rate": 3.009160825907585e-07, + "loss": 0.151, + "step": 6738 + }, + { + "epoch": 0.35, + "learning_rate": 3.0088688951985856e-07, + "loss": 0.14, + "step": 6739 + }, + { + "epoch": 0.35, + "learning_rate": 3.008576935655725e-07, + "loss": 0.1566, + "step": 6740 + }, + { + "epoch": 0.35, + "learning_rate": 3.008284947287345e-07, + "loss": 0.145, + "step": 6741 + }, + { + "epoch": 0.35, + "learning_rate": 3.0079929301017924e-07, + "loss": 0.1384, + "step": 6742 + }, + { + "epoch": 0.35, + "learning_rate": 3.007700884107414e-07, + "loss": 0.1237, + "step": 6743 + }, + { + "epoch": 0.35, + "learning_rate": 3.0074088093125533e-07, + "loss": 0.1601, + "step": 6744 + }, + { + "epoch": 0.35, + "learning_rate": 3.0071167057255607e-07, + "loss": 0.1182, + "step": 6745 + }, + { + "epoch": 0.35, + "learning_rate": 3.006824573354784e-07, + "loss": 0.16, + "step": 6746 + }, + { + "epoch": 0.35, + "learning_rate": 3.0065324122085726e-07, + "loss": 0.1317, + "step": 6747 + }, + { + "epoch": 0.35, + "learning_rate": 3.0062402222952754e-07, + "loss": 0.1519, + "step": 6748 + }, + { + "epoch": 0.35, + "learning_rate": 3.005948003623245e-07, + "loss": 0.1669, + "step": 6749 + }, + { + "epoch": 0.35, + "learning_rate": 3.005655756200832e-07, + "loss": 0.1179, + "step": 6750 + }, + { + "epoch": 0.35, + "learning_rate": 3.005363480036389e-07, + "loss": 0.1126, + "step": 6751 + }, + { + "epoch": 0.35, + "learning_rate": 3.0050711751382703e-07, + "loss": 0.1344, + "step": 6752 + }, + { + "epoch": 0.35, + "learning_rate": 3.004778841514829e-07, + "loss": 0.1675, + "step": 6753 + }, + { + "epoch": 0.35, + "learning_rate": 3.0044864791744207e-07, + "loss": 0.115, + "step": 6754 + }, + { + "epoch": 0.35, + "learning_rate": 3.0041940881254013e-07, + "loss": 0.1816, + "step": 6755 + }, + { + "epoch": 0.35, + "learning_rate": 3.0039016683761264e-07, + "loss": 0.1592, + "step": 6756 + }, + { + "epoch": 0.35, + "learning_rate": 3.003609219934955e-07, + "loss": 0.1315, + "step": 6757 + }, + { + "epoch": 0.35, + "learning_rate": 3.0033167428102444e-07, + "loss": 0.1032, + "step": 6758 + }, + { + "epoch": 0.35, + "learning_rate": 3.0030242370103545e-07, + "loss": 0.1247, + "step": 6759 + }, + { + "epoch": 0.35, + "learning_rate": 3.002731702543644e-07, + "loss": 0.1639, + "step": 6760 + }, + { + "epoch": 0.35, + "learning_rate": 3.002439139418475e-07, + "loss": 0.1521, + "step": 6761 + }, + { + "epoch": 0.35, + "learning_rate": 3.0021465476432066e-07, + "loss": 0.1312, + "step": 6762 + }, + { + "epoch": 0.35, + "learning_rate": 3.0018539272262054e-07, + "loss": 0.1241, + "step": 6763 + }, + { + "epoch": 0.35, + "learning_rate": 3.001561278175831e-07, + "loss": 0.157, + "step": 6764 + }, + { + "epoch": 0.35, + "learning_rate": 3.001268600500449e-07, + "loss": 0.1756, + "step": 6765 + }, + { + "epoch": 0.35, + "learning_rate": 3.000975894208423e-07, + "loss": 0.146, + "step": 6766 + }, + { + "epoch": 0.35, + "learning_rate": 3.0006831593081206e-07, + "loss": 0.1344, + "step": 6767 + }, + { + "epoch": 0.35, + "learning_rate": 3.000390395807907e-07, + "loss": 0.0945, + "step": 6768 + }, + { + "epoch": 0.35, + "learning_rate": 3.0000976037161493e-07, + "loss": 0.1394, + "step": 6769 + }, + { + "epoch": 0.35, + "learning_rate": 2.9998047830412157e-07, + "loss": 0.1542, + "step": 6770 + }, + { + "epoch": 0.35, + "learning_rate": 2.999511933791476e-07, + "loss": 0.1285, + "step": 6771 + }, + { + "epoch": 0.35, + "learning_rate": 2.9992190559753e-07, + "loss": 0.1544, + "step": 6772 + }, + { + "epoch": 0.35, + "learning_rate": 2.998926149601057e-07, + "loss": 0.1191, + "step": 6773 + }, + { + "epoch": 0.35, + "learning_rate": 2.9986332146771186e-07, + "loss": 0.125, + "step": 6774 + }, + { + "epoch": 0.35, + "learning_rate": 2.9983402512118574e-07, + "loss": 0.1194, + "step": 6775 + }, + { + "epoch": 0.35, + "learning_rate": 2.9980472592136476e-07, + "loss": 0.1153, + "step": 6776 + }, + { + "epoch": 0.35, + "learning_rate": 2.9977542386908605e-07, + "loss": 0.1542, + "step": 6777 + }, + { + "epoch": 0.35, + "learning_rate": 2.997461189651873e-07, + "loss": 0.1683, + "step": 6778 + }, + { + "epoch": 0.35, + "learning_rate": 2.9971681121050586e-07, + "loss": 0.114, + "step": 6779 + }, + { + "epoch": 0.35, + "learning_rate": 2.996875006058796e-07, + "loss": 0.1353, + "step": 6780 + }, + { + "epoch": 0.35, + "learning_rate": 2.9965818715214603e-07, + "loss": 0.165, + "step": 6781 + }, + { + "epoch": 0.35, + "learning_rate": 2.9962887085014296e-07, + "loss": 0.1516, + "step": 6782 + }, + { + "epoch": 0.35, + "learning_rate": 2.995995517007083e-07, + "loss": 0.1049, + "step": 6783 + }, + { + "epoch": 0.35, + "learning_rate": 2.9957022970468e-07, + "loss": 0.1461, + "step": 6784 + }, + { + "epoch": 0.35, + "learning_rate": 2.995409048628962e-07, + "loss": 0.1272, + "step": 6785 + }, + { + "epoch": 0.35, + "learning_rate": 2.9951157717619483e-07, + "loss": 0.1449, + "step": 6786 + }, + { + "epoch": 0.35, + "learning_rate": 2.994822466454142e-07, + "loss": 0.1259, + "step": 6787 + }, + { + "epoch": 0.35, + "learning_rate": 2.994529132713925e-07, + "loss": 0.1216, + "step": 6788 + }, + { + "epoch": 0.35, + "learning_rate": 2.9942357705496817e-07, + "loss": 0.1456, + "step": 6789 + }, + { + "epoch": 0.35, + "learning_rate": 2.993942379969797e-07, + "loss": 0.1859, + "step": 6790 + }, + { + "epoch": 0.35, + "learning_rate": 2.9936489609826543e-07, + "loss": 0.117, + "step": 6791 + }, + { + "epoch": 0.35, + "learning_rate": 2.9933555135966415e-07, + "loss": 0.1097, + "step": 6792 + }, + { + "epoch": 0.35, + "learning_rate": 2.993062037820144e-07, + "loss": 0.1235, + "step": 6793 + }, + { + "epoch": 0.35, + "learning_rate": 2.9927685336615507e-07, + "loss": 0.1388, + "step": 6794 + }, + { + "epoch": 0.35, + "learning_rate": 2.992475001129249e-07, + "loss": 0.1071, + "step": 6795 + }, + { + "epoch": 0.35, + "learning_rate": 2.992181440231629e-07, + "loss": 0.1332, + "step": 6796 + }, + { + "epoch": 0.35, + "learning_rate": 2.99188785097708e-07, + "loss": 0.138, + "step": 6797 + }, + { + "epoch": 0.35, + "learning_rate": 2.991594233373994e-07, + "loss": 0.13, + "step": 6798 + }, + { + "epoch": 0.35, + "learning_rate": 2.991300587430762e-07, + "loss": 0.1564, + "step": 6799 + }, + { + "epoch": 0.35, + "learning_rate": 2.991006913155775e-07, + "loss": 0.1207, + "step": 6800 + }, + { + "epoch": 0.35, + "learning_rate": 2.9907132105574294e-07, + "loss": 0.1363, + "step": 6801 + }, + { + "epoch": 0.36, + "learning_rate": 2.9904194796441173e-07, + "loss": 0.128, + "step": 6802 + }, + { + "epoch": 0.36, + "learning_rate": 2.990125720424234e-07, + "loss": 0.1731, + "step": 6803 + }, + { + "epoch": 0.36, + "learning_rate": 2.9898319329061757e-07, + "loss": 0.1638, + "step": 6804 + }, + { + "epoch": 0.36, + "learning_rate": 2.9895381170983385e-07, + "loss": 0.1219, + "step": 6805 + }, + { + "epoch": 0.36, + "learning_rate": 2.9892442730091205e-07, + "loss": 0.1304, + "step": 6806 + }, + { + "epoch": 0.36, + "learning_rate": 2.9889504006469184e-07, + "loss": 0.1246, + "step": 6807 + }, + { + "epoch": 0.36, + "learning_rate": 2.9886565000201326e-07, + "loss": 0.1192, + "step": 6808 + }, + { + "epoch": 0.36, + "learning_rate": 2.9883625711371615e-07, + "loss": 0.1459, + "step": 6809 + }, + { + "epoch": 0.36, + "learning_rate": 2.9880686140064077e-07, + "loss": 0.1245, + "step": 6810 + }, + { + "epoch": 0.36, + "learning_rate": 2.987774628636271e-07, + "loss": 0.132, + "step": 6811 + }, + { + "epoch": 0.36, + "learning_rate": 2.987480615035154e-07, + "loss": 0.1331, + "step": 6812 + }, + { + "epoch": 0.36, + "learning_rate": 2.98718657321146e-07, + "loss": 0.1325, + "step": 6813 + }, + { + "epoch": 0.36, + "learning_rate": 2.986892503173593e-07, + "loss": 0.1436, + "step": 6814 + }, + { + "epoch": 0.36, + "learning_rate": 2.986598404929956e-07, + "loss": 0.1367, + "step": 6815 + }, + { + "epoch": 0.36, + "learning_rate": 2.9863042784889566e-07, + "loss": 0.1401, + "step": 6816 + }, + { + "epoch": 0.36, + "learning_rate": 2.9860101238590006e-07, + "loss": 0.1108, + "step": 6817 + }, + { + "epoch": 0.36, + "learning_rate": 2.9857159410484936e-07, + "loss": 0.138, + "step": 6818 + }, + { + "epoch": 0.36, + "learning_rate": 2.9854217300658454e-07, + "loss": 0.1325, + "step": 6819 + }, + { + "epoch": 0.36, + "learning_rate": 2.9851274909194634e-07, + "loss": 0.1355, + "step": 6820 + }, + { + "epoch": 0.36, + "learning_rate": 2.984833223617757e-07, + "loss": 0.1217, + "step": 6821 + }, + { + "epoch": 0.36, + "learning_rate": 2.9845389281691374e-07, + "loss": 0.1423, + "step": 6822 + }, + { + "epoch": 0.36, + "learning_rate": 2.984244604582016e-07, + "loss": 0.1139, + "step": 6823 + }, + { + "epoch": 0.36, + "learning_rate": 2.983950252864802e-07, + "loss": 0.1275, + "step": 6824 + }, + { + "epoch": 0.36, + "learning_rate": 2.983655873025911e-07, + "loss": 0.1657, + "step": 6825 + }, + { + "epoch": 0.36, + "learning_rate": 2.983361465073755e-07, + "loss": 0.1478, + "step": 6826 + }, + { + "epoch": 0.36, + "learning_rate": 2.983067029016749e-07, + "loss": 0.1277, + "step": 6827 + }, + { + "epoch": 0.36, + "learning_rate": 2.982772564863308e-07, + "loss": 0.1513, + "step": 6828 + }, + { + "epoch": 0.36, + "learning_rate": 2.9824780726218475e-07, + "loss": 0.1386, + "step": 6829 + }, + { + "epoch": 0.36, + "learning_rate": 2.982183552300784e-07, + "loss": 0.1356, + "step": 6830 + }, + { + "epoch": 0.36, + "learning_rate": 2.981889003908536e-07, + "loss": 0.117, + "step": 6831 + }, + { + "epoch": 0.36, + "learning_rate": 2.981594427453521e-07, + "loss": 0.1269, + "step": 6832 + }, + { + "epoch": 0.36, + "learning_rate": 2.981299822944158e-07, + "loss": 0.1392, + "step": 6833 + }, + { + "epoch": 0.36, + "learning_rate": 2.981005190388868e-07, + "loss": 0.1316, + "step": 6834 + }, + { + "epoch": 0.36, + "learning_rate": 2.98071052979607e-07, + "loss": 0.1281, + "step": 6835 + }, + { + "epoch": 0.36, + "learning_rate": 2.9804158411741874e-07, + "loss": 0.1125, + "step": 6836 + }, + { + "epoch": 0.36, + "learning_rate": 2.9801211245316415e-07, + "loss": 0.128, + "step": 6837 + }, + { + "epoch": 0.36, + "learning_rate": 2.979826379876855e-07, + "loss": 0.1628, + "step": 6838 + }, + { + "epoch": 0.36, + "learning_rate": 2.979531607218252e-07, + "loss": 0.1514, + "step": 6839 + }, + { + "epoch": 0.36, + "learning_rate": 2.9792368065642583e-07, + "loss": 0.1234, + "step": 6840 + }, + { + "epoch": 0.36, + "learning_rate": 2.9789419779232987e-07, + "loss": 0.1183, + "step": 6841 + }, + { + "epoch": 0.36, + "learning_rate": 2.978647121303799e-07, + "loss": 0.1465, + "step": 6842 + }, + { + "epoch": 0.36, + "learning_rate": 2.9783522367141867e-07, + "loss": 0.1282, + "step": 6843 + }, + { + "epoch": 0.36, + "learning_rate": 2.97805732416289e-07, + "loss": 0.1352, + "step": 6844 + }, + { + "epoch": 0.36, + "learning_rate": 2.977762383658337e-07, + "loss": 0.1363, + "step": 6845 + }, + { + "epoch": 0.36, + "learning_rate": 2.9774674152089584e-07, + "loss": 0.1381, + "step": 6846 + }, + { + "epoch": 0.36, + "learning_rate": 2.9771724188231837e-07, + "loss": 0.1259, + "step": 6847 + }, + { + "epoch": 0.36, + "learning_rate": 2.976877394509443e-07, + "loss": 0.1235, + "step": 6848 + }, + { + "epoch": 0.36, + "learning_rate": 2.9765823422761706e-07, + "loss": 0.1428, + "step": 6849 + }, + { + "epoch": 0.36, + "learning_rate": 2.976287262131797e-07, + "loss": 0.1176, + "step": 6850 + }, + { + "epoch": 0.36, + "learning_rate": 2.9759921540847563e-07, + "loss": 0.1372, + "step": 6851 + }, + { + "epoch": 0.36, + "learning_rate": 2.975697018143484e-07, + "loss": 0.1302, + "step": 6852 + }, + { + "epoch": 0.36, + "learning_rate": 2.9754018543164136e-07, + "loss": 0.124, + "step": 6853 + }, + { + "epoch": 0.36, + "learning_rate": 2.975106662611982e-07, + "loss": 0.1654, + "step": 6854 + }, + { + "epoch": 0.36, + "learning_rate": 2.9748114430386254e-07, + "loss": 0.1305, + "step": 6855 + }, + { + "epoch": 0.36, + "learning_rate": 2.9745161956047816e-07, + "loss": 0.1409, + "step": 6856 + }, + { + "epoch": 0.36, + "learning_rate": 2.974220920318888e-07, + "loss": 0.1389, + "step": 6857 + }, + { + "epoch": 0.36, + "learning_rate": 2.9739256171893855e-07, + "loss": 0.1453, + "step": 6858 + }, + { + "epoch": 0.36, + "learning_rate": 2.973630286224712e-07, + "loss": 0.1536, + "step": 6859 + }, + { + "epoch": 0.36, + "learning_rate": 2.97333492743331e-07, + "loss": 0.155, + "step": 6860 + }, + { + "epoch": 0.36, + "learning_rate": 2.9730395408236193e-07, + "loss": 0.1207, + "step": 6861 + }, + { + "epoch": 0.36, + "learning_rate": 2.972744126404083e-07, + "loss": 0.1461, + "step": 6862 + }, + { + "epoch": 0.36, + "learning_rate": 2.972448684183144e-07, + "loss": 0.1419, + "step": 6863 + }, + { + "epoch": 0.36, + "learning_rate": 2.9721532141692463e-07, + "loss": 0.1254, + "step": 6864 + }, + { + "epoch": 0.36, + "learning_rate": 2.9718577163708343e-07, + "loss": 0.125, + "step": 6865 + }, + { + "epoch": 0.36, + "learning_rate": 2.9715621907963543e-07, + "loss": 0.1475, + "step": 6866 + }, + { + "epoch": 0.36, + "learning_rate": 2.971266637454251e-07, + "loss": 0.1336, + "step": 6867 + }, + { + "epoch": 0.36, + "learning_rate": 2.970971056352973e-07, + "loss": 0.1899, + "step": 6868 + }, + { + "epoch": 0.36, + "learning_rate": 2.9706754475009663e-07, + "loss": 0.1436, + "step": 6869 + }, + { + "epoch": 0.36, + "learning_rate": 2.9703798109066815e-07, + "loss": 0.1428, + "step": 6870 + }, + { + "epoch": 0.36, + "learning_rate": 2.9700841465785666e-07, + "loss": 0.1685, + "step": 6871 + }, + { + "epoch": 0.36, + "learning_rate": 2.9697884545250723e-07, + "loss": 0.1227, + "step": 6872 + }, + { + "epoch": 0.36, + "learning_rate": 2.9694927347546504e-07, + "loss": 0.1567, + "step": 6873 + }, + { + "epoch": 0.36, + "learning_rate": 2.969196987275751e-07, + "loss": 0.1287, + "step": 6874 + }, + { + "epoch": 0.36, + "learning_rate": 2.968901212096828e-07, + "loss": 0.1505, + "step": 6875 + }, + { + "epoch": 0.36, + "learning_rate": 2.968605409226334e-07, + "loss": 0.1414, + "step": 6876 + }, + { + "epoch": 0.36, + "learning_rate": 2.968309578672724e-07, + "loss": 0.1176, + "step": 6877 + }, + { + "epoch": 0.36, + "learning_rate": 2.9680137204444515e-07, + "loss": 0.1372, + "step": 6878 + }, + { + "epoch": 0.36, + "learning_rate": 2.9677178345499744e-07, + "loss": 0.1417, + "step": 6879 + }, + { + "epoch": 0.36, + "learning_rate": 2.967421920997747e-07, + "loss": 0.1754, + "step": 6880 + }, + { + "epoch": 0.36, + "learning_rate": 2.9671259797962284e-07, + "loss": 0.1274, + "step": 6881 + }, + { + "epoch": 0.36, + "learning_rate": 2.966830010953875e-07, + "loss": 0.1505, + "step": 6882 + }, + { + "epoch": 0.36, + "learning_rate": 2.966534014479148e-07, + "loss": 0.1436, + "step": 6883 + }, + { + "epoch": 0.36, + "learning_rate": 2.9662379903805045e-07, + "loss": 0.1582, + "step": 6884 + }, + { + "epoch": 0.36, + "learning_rate": 2.9659419386664065e-07, + "loss": 0.105, + "step": 6885 + }, + { + "epoch": 0.36, + "learning_rate": 2.965645859345315e-07, + "loss": 0.1525, + "step": 6886 + }, + { + "epoch": 0.36, + "learning_rate": 2.965349752425692e-07, + "loss": 0.1389, + "step": 6887 + }, + { + "epoch": 0.36, + "learning_rate": 2.9650536179160015e-07, + "loss": 0.1264, + "step": 6888 + }, + { + "epoch": 0.36, + "learning_rate": 2.964757455824705e-07, + "loss": 0.212, + "step": 6889 + }, + { + "epoch": 0.36, + "learning_rate": 2.9644612661602685e-07, + "loss": 0.1477, + "step": 6890 + }, + { + "epoch": 0.36, + "learning_rate": 2.964165048931156e-07, + "loss": 0.1569, + "step": 6891 + }, + { + "epoch": 0.36, + "learning_rate": 2.963868804145835e-07, + "loss": 0.141, + "step": 6892 + }, + { + "epoch": 0.36, + "learning_rate": 2.9635725318127704e-07, + "loss": 0.1251, + "step": 6893 + }, + { + "epoch": 0.36, + "learning_rate": 2.9632762319404314e-07, + "loss": 0.1172, + "step": 6894 + }, + { + "epoch": 0.36, + "learning_rate": 2.9629799045372864e-07, + "loss": 0.1909, + "step": 6895 + }, + { + "epoch": 0.36, + "learning_rate": 2.962683549611804e-07, + "loss": 0.1477, + "step": 6896 + }, + { + "epoch": 0.36, + "learning_rate": 2.962387167172454e-07, + "loss": 0.1479, + "step": 6897 + }, + { + "epoch": 0.36, + "learning_rate": 2.9620907572277063e-07, + "loss": 0.1405, + "step": 6898 + }, + { + "epoch": 0.36, + "learning_rate": 2.9617943197860337e-07, + "loss": 0.1532, + "step": 6899 + }, + { + "epoch": 0.36, + "learning_rate": 2.961497854855909e-07, + "loss": 0.1648, + "step": 6900 + }, + { + "epoch": 0.36, + "learning_rate": 2.9612013624458034e-07, + "loss": 0.1494, + "step": 6901 + }, + { + "epoch": 0.36, + "learning_rate": 2.960904842564192e-07, + "loss": 0.1354, + "step": 6902 + }, + { + "epoch": 0.36, + "learning_rate": 2.9606082952195495e-07, + "loss": 0.1721, + "step": 6903 + }, + { + "epoch": 0.36, + "learning_rate": 2.960311720420351e-07, + "loss": 0.0987, + "step": 6904 + }, + { + "epoch": 0.36, + "learning_rate": 2.9600151181750737e-07, + "loss": 0.1339, + "step": 6905 + }, + { + "epoch": 0.36, + "learning_rate": 2.959718488492193e-07, + "loss": 0.1429, + "step": 6906 + }, + { + "epoch": 0.36, + "learning_rate": 2.9594218313801864e-07, + "loss": 0.131, + "step": 6907 + }, + { + "epoch": 0.36, + "learning_rate": 2.959125146847535e-07, + "loss": 0.1388, + "step": 6908 + }, + { + "epoch": 0.36, + "learning_rate": 2.958828434902716e-07, + "loss": 0.1392, + "step": 6909 + }, + { + "epoch": 0.36, + "learning_rate": 2.9585316955542105e-07, + "loss": 0.1541, + "step": 6910 + }, + { + "epoch": 0.36, + "learning_rate": 2.9582349288104987e-07, + "loss": 0.1307, + "step": 6911 + }, + { + "epoch": 0.36, + "learning_rate": 2.9579381346800624e-07, + "loss": 0.1303, + "step": 6912 + }, + { + "epoch": 0.36, + "learning_rate": 2.9576413131713856e-07, + "loss": 0.1386, + "step": 6913 + }, + { + "epoch": 0.36, + "learning_rate": 2.9573444642929493e-07, + "loss": 0.1218, + "step": 6914 + }, + { + "epoch": 0.36, + "learning_rate": 2.957047588053239e-07, + "loss": 0.1445, + "step": 6915 + }, + { + "epoch": 0.36, + "learning_rate": 2.9567506844607393e-07, + "loss": 0.1681, + "step": 6916 + }, + { + "epoch": 0.36, + "learning_rate": 2.9564537535239353e-07, + "loss": 0.1488, + "step": 6917 + }, + { + "epoch": 0.36, + "learning_rate": 2.9561567952513146e-07, + "loss": 0.1636, + "step": 6918 + }, + { + "epoch": 0.36, + "learning_rate": 2.955859809651362e-07, + "loss": 0.113, + "step": 6919 + }, + { + "epoch": 0.36, + "learning_rate": 2.9555627967325687e-07, + "loss": 0.1251, + "step": 6920 + }, + { + "epoch": 0.36, + "learning_rate": 2.9552657565034207e-07, + "loss": 0.1353, + "step": 6921 + }, + { + "epoch": 0.36, + "learning_rate": 2.9549686889724097e-07, + "loss": 0.1244, + "step": 6922 + }, + { + "epoch": 0.36, + "learning_rate": 2.954671594148024e-07, + "loss": 0.1564, + "step": 6923 + }, + { + "epoch": 0.36, + "learning_rate": 2.954374472038756e-07, + "loss": 0.1261, + "step": 6924 + }, + { + "epoch": 0.36, + "learning_rate": 2.9540773226530967e-07, + "loss": 0.1665, + "step": 6925 + }, + { + "epoch": 0.36, + "learning_rate": 2.95378014599954e-07, + "loss": 0.1419, + "step": 6926 + }, + { + "epoch": 0.36, + "learning_rate": 2.9534829420865784e-07, + "loss": 0.145, + "step": 6927 + }, + { + "epoch": 0.36, + "learning_rate": 2.9531857109227056e-07, + "loss": 0.1396, + "step": 6928 + }, + { + "epoch": 0.36, + "learning_rate": 2.952888452516418e-07, + "loss": 0.1191, + "step": 6929 + }, + { + "epoch": 0.36, + "learning_rate": 2.95259116687621e-07, + "loss": 0.1393, + "step": 6930 + }, + { + "epoch": 0.36, + "learning_rate": 2.952293854010579e-07, + "loss": 0.1316, + "step": 6931 + }, + { + "epoch": 0.36, + "learning_rate": 2.9519965139280216e-07, + "loss": 0.1344, + "step": 6932 + }, + { + "epoch": 0.36, + "learning_rate": 2.9516991466370365e-07, + "loss": 0.1347, + "step": 6933 + }, + { + "epoch": 0.36, + "learning_rate": 2.9514017521461224e-07, + "loss": 0.1684, + "step": 6934 + }, + { + "epoch": 0.36, + "learning_rate": 2.9511043304637795e-07, + "loss": 0.1529, + "step": 6935 + }, + { + "epoch": 0.36, + "learning_rate": 2.950806881598507e-07, + "loss": 0.1243, + "step": 6936 + }, + { + "epoch": 0.36, + "learning_rate": 2.9505094055588073e-07, + "loss": 0.1691, + "step": 6937 + }, + { + "epoch": 0.36, + "learning_rate": 2.9502119023531807e-07, + "loss": 0.1388, + "step": 6938 + }, + { + "epoch": 0.36, + "learning_rate": 2.9499143719901324e-07, + "loss": 0.1233, + "step": 6939 + }, + { + "epoch": 0.36, + "learning_rate": 2.9496168144781647e-07, + "loss": 0.145, + "step": 6940 + }, + { + "epoch": 0.36, + "learning_rate": 2.9493192298257806e-07, + "loss": 0.1465, + "step": 6941 + }, + { + "epoch": 0.36, + "learning_rate": 2.949021618041487e-07, + "loss": 0.1463, + "step": 6942 + }, + { + "epoch": 0.36, + "learning_rate": 2.9487239791337903e-07, + "loss": 0.108, + "step": 6943 + }, + { + "epoch": 0.36, + "learning_rate": 2.948426313111195e-07, + "loss": 0.1236, + "step": 6944 + }, + { + "epoch": 0.36, + "learning_rate": 2.9481286199822095e-07, + "loss": 0.1242, + "step": 6945 + }, + { + "epoch": 0.36, + "learning_rate": 2.9478308997553423e-07, + "loss": 0.1533, + "step": 6946 + }, + { + "epoch": 0.36, + "learning_rate": 2.9475331524391015e-07, + "loss": 0.1707, + "step": 6947 + }, + { + "epoch": 0.36, + "learning_rate": 2.947235378041998e-07, + "loss": 0.1602, + "step": 6948 + }, + { + "epoch": 0.36, + "learning_rate": 2.946937576572542e-07, + "loss": 0.1166, + "step": 6949 + }, + { + "epoch": 0.36, + "learning_rate": 2.946639748039244e-07, + "loss": 0.1545, + "step": 6950 + }, + { + "epoch": 0.36, + "learning_rate": 2.9463418924506165e-07, + "loss": 0.152, + "step": 6951 + }, + { + "epoch": 0.36, + "learning_rate": 2.9460440098151725e-07, + "loss": 0.1637, + "step": 6952 + }, + { + "epoch": 0.36, + "learning_rate": 2.945746100141426e-07, + "loss": 0.1388, + "step": 6953 + }, + { + "epoch": 0.36, + "learning_rate": 2.94544816343789e-07, + "loss": 0.1133, + "step": 6954 + }, + { + "epoch": 0.36, + "learning_rate": 2.9451501997130814e-07, + "loss": 0.1378, + "step": 6955 + }, + { + "epoch": 0.36, + "learning_rate": 2.9448522089755143e-07, + "loss": 0.1097, + "step": 6956 + }, + { + "epoch": 0.36, + "learning_rate": 2.9445541912337073e-07, + "loss": 0.1124, + "step": 6957 + }, + { + "epoch": 0.36, + "learning_rate": 2.944256146496176e-07, + "loss": 0.1376, + "step": 6958 + }, + { + "epoch": 0.36, + "learning_rate": 2.94395807477144e-07, + "loss": 0.1547, + "step": 6959 + }, + { + "epoch": 0.36, + "learning_rate": 2.9436599760680174e-07, + "loss": 0.1369, + "step": 6960 + }, + { + "epoch": 0.36, + "learning_rate": 2.943361850394429e-07, + "loss": 0.1311, + "step": 6961 + }, + { + "epoch": 0.36, + "learning_rate": 2.9430636977591945e-07, + "loss": 0.14, + "step": 6962 + }, + { + "epoch": 0.36, + "learning_rate": 2.942765518170835e-07, + "loss": 0.1105, + "step": 6963 + }, + { + "epoch": 0.36, + "learning_rate": 2.9424673116378735e-07, + "loss": 0.1177, + "step": 6964 + }, + { + "epoch": 0.36, + "learning_rate": 2.942169078168833e-07, + "loss": 0.1661, + "step": 6965 + }, + { + "epoch": 0.36, + "learning_rate": 2.941870817772235e-07, + "loss": 0.165, + "step": 6966 + }, + { + "epoch": 0.36, + "learning_rate": 2.9415725304566066e-07, + "loss": 0.1356, + "step": 6967 + }, + { + "epoch": 0.36, + "learning_rate": 2.9412742162304717e-07, + "loss": 0.1126, + "step": 6968 + }, + { + "epoch": 0.36, + "learning_rate": 2.9409758751023566e-07, + "loss": 0.1418, + "step": 6969 + }, + { + "epoch": 0.36, + "learning_rate": 2.9406775070807865e-07, + "loss": 0.1509, + "step": 6970 + }, + { + "epoch": 0.36, + "learning_rate": 2.940379112174291e-07, + "loss": 0.1262, + "step": 6971 + }, + { + "epoch": 0.36, + "learning_rate": 2.940080690391397e-07, + "loss": 0.1554, + "step": 6972 + }, + { + "epoch": 0.36, + "learning_rate": 2.939782241740635e-07, + "loss": 0.1163, + "step": 6973 + }, + { + "epoch": 0.36, + "learning_rate": 2.939483766230533e-07, + "loss": 0.1694, + "step": 6974 + }, + { + "epoch": 0.36, + "learning_rate": 2.9391852638696216e-07, + "loss": 0.1471, + "step": 6975 + }, + { + "epoch": 0.36, + "learning_rate": 2.9388867346664336e-07, + "loss": 0.084, + "step": 6976 + }, + { + "epoch": 0.36, + "learning_rate": 2.9385881786295003e-07, + "loss": 0.1689, + "step": 6977 + }, + { + "epoch": 0.36, + "learning_rate": 2.938289595767355e-07, + "loss": 0.1691, + "step": 6978 + }, + { + "epoch": 0.36, + "learning_rate": 2.93799098608853e-07, + "loss": 0.1658, + "step": 6979 + }, + { + "epoch": 0.36, + "learning_rate": 2.9376923496015604e-07, + "loss": 0.113, + "step": 6980 + }, + { + "epoch": 0.36, + "learning_rate": 2.937393686314982e-07, + "loss": 0.1703, + "step": 6981 + }, + { + "epoch": 0.36, + "learning_rate": 2.9370949962373297e-07, + "loss": 0.1373, + "step": 6982 + }, + { + "epoch": 0.36, + "learning_rate": 2.9367962793771417e-07, + "loss": 0.1194, + "step": 6983 + }, + { + "epoch": 0.36, + "learning_rate": 2.936497535742953e-07, + "loss": 0.1717, + "step": 6984 + }, + { + "epoch": 0.36, + "learning_rate": 2.9361987653433044e-07, + "loss": 0.132, + "step": 6985 + }, + { + "epoch": 0.36, + "learning_rate": 2.935899968186733e-07, + "loss": 0.1485, + "step": 6986 + }, + { + "epoch": 0.36, + "learning_rate": 2.9356011442817794e-07, + "loss": 0.1311, + "step": 6987 + }, + { + "epoch": 0.36, + "learning_rate": 2.9353022936369836e-07, + "loss": 0.1494, + "step": 6988 + }, + { + "epoch": 0.36, + "learning_rate": 2.9350034162608874e-07, + "loss": 0.143, + "step": 6989 + }, + { + "epoch": 0.36, + "learning_rate": 2.934704512162033e-07, + "loss": 0.1177, + "step": 6990 + }, + { + "epoch": 0.36, + "learning_rate": 2.9344055813489627e-07, + "loss": 0.1732, + "step": 6991 + }, + { + "epoch": 0.36, + "learning_rate": 2.9341066238302194e-07, + "loss": 0.1241, + "step": 6992 + }, + { + "epoch": 0.37, + "learning_rate": 2.933807639614349e-07, + "loss": 0.1511, + "step": 6993 + }, + { + "epoch": 0.37, + "learning_rate": 2.9335086287098955e-07, + "loss": 0.1454, + "step": 6994 + }, + { + "epoch": 0.37, + "learning_rate": 2.933209591125405e-07, + "loss": 0.1576, + "step": 6995 + }, + { + "epoch": 0.37, + "learning_rate": 2.932910526869424e-07, + "loss": 0.139, + "step": 6996 + }, + { + "epoch": 0.37, + "learning_rate": 2.9326114359505e-07, + "loss": 0.2404, + "step": 6997 + }, + { + "epoch": 0.37, + "learning_rate": 2.9323123183771813e-07, + "loss": 0.1033, + "step": 6998 + }, + { + "epoch": 0.37, + "learning_rate": 2.932013174158017e-07, + "loss": 0.1501, + "step": 6999 + }, + { + "epoch": 0.37, + "learning_rate": 2.931714003301556e-07, + "loss": 0.1427, + "step": 7000 + }, + { + "epoch": 0.37, + "learning_rate": 2.931414805816349e-07, + "loss": 0.1045, + "step": 7001 + }, + { + "epoch": 0.37, + "learning_rate": 2.931115581710947e-07, + "loss": 0.094, + "step": 7002 + }, + { + "epoch": 0.37, + "learning_rate": 2.9308163309939034e-07, + "loss": 0.1404, + "step": 7003 + }, + { + "epoch": 0.37, + "learning_rate": 2.9305170536737686e-07, + "loss": 0.1343, + "step": 7004 + }, + { + "epoch": 0.37, + "learning_rate": 2.9302177497590974e-07, + "loss": 0.1289, + "step": 7005 + }, + { + "epoch": 0.37, + "learning_rate": 2.929918419258444e-07, + "loss": 0.1054, + "step": 7006 + }, + { + "epoch": 0.37, + "learning_rate": 2.929619062180363e-07, + "loss": 0.1585, + "step": 7007 + }, + { + "epoch": 0.37, + "learning_rate": 2.929319678533411e-07, + "loss": 0.1283, + "step": 7008 + }, + { + "epoch": 0.37, + "learning_rate": 2.9290202683261435e-07, + "loss": 0.1486, + "step": 7009 + }, + { + "epoch": 0.37, + "learning_rate": 2.928720831567117e-07, + "loss": 0.1561, + "step": 7010 + }, + { + "epoch": 0.37, + "learning_rate": 2.9284213682648926e-07, + "loss": 0.1714, + "step": 7011 + }, + { + "epoch": 0.37, + "learning_rate": 2.928121878428026e-07, + "loss": 0.1027, + "step": 7012 + }, + { + "epoch": 0.37, + "learning_rate": 2.927822362065077e-07, + "loss": 0.1486, + "step": 7013 + }, + { + "epoch": 0.37, + "learning_rate": 2.9275228191846084e-07, + "loss": 0.1291, + "step": 7014 + }, + { + "epoch": 0.37, + "learning_rate": 2.927223249795178e-07, + "loss": 0.1678, + "step": 7015 + }, + { + "epoch": 0.37, + "learning_rate": 2.9269236539053503e-07, + "loss": 0.1057, + "step": 7016 + }, + { + "epoch": 0.37, + "learning_rate": 2.926624031523686e-07, + "loss": 0.1172, + "step": 7017 + }, + { + "epoch": 0.37, + "learning_rate": 2.92632438265875e-07, + "loss": 0.1233, + "step": 7018 + }, + { + "epoch": 0.37, + "learning_rate": 2.926024707319105e-07, + "loss": 0.1506, + "step": 7019 + }, + { + "epoch": 0.37, + "learning_rate": 2.925725005513316e-07, + "loss": 0.1375, + "step": 7020 + }, + { + "epoch": 0.37, + "learning_rate": 2.92542527724995e-07, + "loss": 0.1497, + "step": 7021 + }, + { + "epoch": 0.37, + "learning_rate": 2.925125522537571e-07, + "loss": 0.1386, + "step": 7022 + }, + { + "epoch": 0.37, + "learning_rate": 2.9248257413847485e-07, + "loss": 0.135, + "step": 7023 + }, + { + "epoch": 0.37, + "learning_rate": 2.9245259338000485e-07, + "loss": 0.1502, + "step": 7024 + }, + { + "epoch": 0.37, + "learning_rate": 2.924226099792042e-07, + "loss": 0.1998, + "step": 7025 + }, + { + "epoch": 0.37, + "learning_rate": 2.9239262393692953e-07, + "loss": 0.1125, + "step": 7026 + }, + { + "epoch": 0.37, + "learning_rate": 2.9236263525403803e-07, + "loss": 0.1184, + "step": 7027 + }, + { + "epoch": 0.37, + "learning_rate": 2.9233264393138676e-07, + "loss": 0.1679, + "step": 7028 + }, + { + "epoch": 0.37, + "learning_rate": 2.923026499698329e-07, + "loss": 0.1703, + "step": 7029 + }, + { + "epoch": 0.37, + "learning_rate": 2.922726533702337e-07, + "loss": 0.1599, + "step": 7030 + }, + { + "epoch": 0.37, + "learning_rate": 2.9224265413344636e-07, + "loss": 0.1614, + "step": 7031 + }, + { + "epoch": 0.37, + "learning_rate": 2.922126522603284e-07, + "loss": 0.1272, + "step": 7032 + }, + { + "epoch": 0.37, + "learning_rate": 2.921826477517372e-07, + "loss": 0.1371, + "step": 7033 + }, + { + "epoch": 0.37, + "learning_rate": 2.9215264060853047e-07, + "loss": 0.1499, + "step": 7034 + }, + { + "epoch": 0.37, + "learning_rate": 2.9212263083156555e-07, + "loss": 0.1515, + "step": 7035 + }, + { + "epoch": 0.37, + "learning_rate": 2.9209261842170037e-07, + "loss": 0.1476, + "step": 7036 + }, + { + "epoch": 0.37, + "learning_rate": 2.920626033797926e-07, + "loss": 0.15, + "step": 7037 + }, + { + "epoch": 0.37, + "learning_rate": 2.9203258570670003e-07, + "loss": 0.1009, + "step": 7038 + }, + { + "epoch": 0.37, + "learning_rate": 2.9200256540328065e-07, + "loss": 0.1279, + "step": 7039 + }, + { + "epoch": 0.37, + "learning_rate": 2.919725424703924e-07, + "loss": 0.1623, + "step": 7040 + }, + { + "epoch": 0.37, + "learning_rate": 2.919425169088935e-07, + "loss": 0.1298, + "step": 7041 + }, + { + "epoch": 0.37, + "learning_rate": 2.919124887196419e-07, + "loss": 0.1299, + "step": 7042 + }, + { + "epoch": 0.37, + "learning_rate": 2.918824579034959e-07, + "loss": 0.1475, + "step": 7043 + }, + { + "epoch": 0.37, + "learning_rate": 2.9185242446131375e-07, + "loss": 0.1257, + "step": 7044 + }, + { + "epoch": 0.37, + "learning_rate": 2.9182238839395387e-07, + "loss": 0.128, + "step": 7045 + }, + { + "epoch": 0.37, + "learning_rate": 2.917923497022747e-07, + "loss": 0.1131, + "step": 7046 + }, + { + "epoch": 0.37, + "learning_rate": 2.9176230838713474e-07, + "loss": 0.1263, + "step": 7047 + }, + { + "epoch": 0.37, + "learning_rate": 2.917322644493925e-07, + "loss": 0.1374, + "step": 7048 + }, + { + "epoch": 0.37, + "learning_rate": 2.9170221788990684e-07, + "loss": 0.1464, + "step": 7049 + }, + { + "epoch": 0.37, + "learning_rate": 2.9167216870953636e-07, + "loss": 0.1398, + "step": 7050 + }, + { + "epoch": 0.37, + "learning_rate": 2.916421169091399e-07, + "loss": 0.1506, + "step": 7051 + }, + { + "epoch": 0.37, + "learning_rate": 2.916120624895764e-07, + "loss": 0.1553, + "step": 7052 + }, + { + "epoch": 0.37, + "learning_rate": 2.9158200545170466e-07, + "loss": 0.146, + "step": 7053 + }, + { + "epoch": 0.37, + "learning_rate": 2.91551945796384e-07, + "loss": 0.1594, + "step": 7054 + }, + { + "epoch": 0.37, + "learning_rate": 2.9152188352447333e-07, + "loss": 0.1487, + "step": 7055 + }, + { + "epoch": 0.37, + "learning_rate": 2.914918186368319e-07, + "loss": 0.1155, + "step": 7056 + }, + { + "epoch": 0.37, + "learning_rate": 2.9146175113431893e-07, + "loss": 0.1369, + "step": 7057 + }, + { + "epoch": 0.37, + "learning_rate": 2.9143168101779385e-07, + "loss": 0.1324, + "step": 7058 + }, + { + "epoch": 0.37, + "learning_rate": 2.9140160828811607e-07, + "loss": 0.1699, + "step": 7059 + }, + { + "epoch": 0.37, + "learning_rate": 2.9137153294614495e-07, + "loss": 0.1656, + "step": 7060 + }, + { + "epoch": 0.37, + "learning_rate": 2.913414549927402e-07, + "loss": 0.1477, + "step": 7061 + }, + { + "epoch": 0.37, + "learning_rate": 2.913113744287614e-07, + "loss": 0.136, + "step": 7062 + }, + { + "epoch": 0.37, + "learning_rate": 2.9128129125506827e-07, + "loss": 0.1834, + "step": 7063 + }, + { + "epoch": 0.37, + "learning_rate": 2.9125120547252055e-07, + "loss": 0.1646, + "step": 7064 + }, + { + "epoch": 0.37, + "learning_rate": 2.9122111708197824e-07, + "loss": 0.1293, + "step": 7065 + }, + { + "epoch": 0.37, + "learning_rate": 2.911910260843011e-07, + "loss": 0.1374, + "step": 7066 + }, + { + "epoch": 0.37, + "learning_rate": 2.911609324803493e-07, + "loss": 0.1318, + "step": 7067 + }, + { + "epoch": 0.37, + "learning_rate": 2.9113083627098284e-07, + "loss": 0.1486, + "step": 7068 + }, + { + "epoch": 0.37, + "learning_rate": 2.9110073745706187e-07, + "loss": 0.1443, + "step": 7069 + }, + { + "epoch": 0.37, + "learning_rate": 2.9107063603944664e-07, + "loss": 0.1093, + "step": 7070 + }, + { + "epoch": 0.37, + "learning_rate": 2.9104053201899753e-07, + "loss": 0.1654, + "step": 7071 + }, + { + "epoch": 0.37, + "learning_rate": 2.9101042539657486e-07, + "loss": 0.1197, + "step": 7072 + }, + { + "epoch": 0.37, + "learning_rate": 2.909803161730391e-07, + "loss": 0.1302, + "step": 7073 + }, + { + "epoch": 0.37, + "learning_rate": 2.9095020434925074e-07, + "loss": 0.1535, + "step": 7074 + }, + { + "epoch": 0.37, + "learning_rate": 2.9092008992607046e-07, + "loss": 0.128, + "step": 7075 + }, + { + "epoch": 0.37, + "learning_rate": 2.908899729043589e-07, + "loss": 0.1593, + "step": 7076 + }, + { + "epoch": 0.37, + "learning_rate": 2.9085985328497687e-07, + "loss": 0.1109, + "step": 7077 + }, + { + "epoch": 0.37, + "learning_rate": 2.908297310687851e-07, + "loss": 0.1866, + "step": 7078 + }, + { + "epoch": 0.37, + "learning_rate": 2.907996062566446e-07, + "loss": 0.1661, + "step": 7079 + }, + { + "epoch": 0.37, + "learning_rate": 2.907694788494163e-07, + "loss": 0.125, + "step": 7080 + }, + { + "epoch": 0.37, + "learning_rate": 2.9073934884796123e-07, + "loss": 0.104, + "step": 7081 + }, + { + "epoch": 0.37, + "learning_rate": 2.907092162531405e-07, + "loss": 0.145, + "step": 7082 + }, + { + "epoch": 0.37, + "learning_rate": 2.906790810658154e-07, + "loss": 0.1422, + "step": 7083 + }, + { + "epoch": 0.37, + "learning_rate": 2.9064894328684713e-07, + "loss": 0.1698, + "step": 7084 + }, + { + "epoch": 0.37, + "learning_rate": 2.906188029170971e-07, + "loss": 0.1541, + "step": 7085 + }, + { + "epoch": 0.37, + "learning_rate": 2.905886599574267e-07, + "loss": 0.1478, + "step": 7086 + }, + { + "epoch": 0.37, + "learning_rate": 2.905585144086974e-07, + "loss": 0.1485, + "step": 7087 + }, + { + "epoch": 0.37, + "learning_rate": 2.9052836627177084e-07, + "loss": 0.1274, + "step": 7088 + }, + { + "epoch": 0.37, + "learning_rate": 2.9049821554750865e-07, + "loss": 0.131, + "step": 7089 + }, + { + "epoch": 0.37, + "learning_rate": 2.904680622367725e-07, + "loss": 0.1216, + "step": 7090 + }, + { + "epoch": 0.37, + "learning_rate": 2.9043790634042416e-07, + "loss": 0.1119, + "step": 7091 + }, + { + "epoch": 0.37, + "learning_rate": 2.9040774785932556e-07, + "loss": 0.1587, + "step": 7092 + }, + { + "epoch": 0.37, + "learning_rate": 2.9037758679433857e-07, + "loss": 0.1452, + "step": 7093 + }, + { + "epoch": 0.37, + "learning_rate": 2.9034742314632534e-07, + "loss": 0.1181, + "step": 7094 + }, + { + "epoch": 0.37, + "learning_rate": 2.903172569161479e-07, + "loss": 0.1139, + "step": 7095 + }, + { + "epoch": 0.37, + "learning_rate": 2.902870881046683e-07, + "loss": 0.1432, + "step": 7096 + }, + { + "epoch": 0.37, + "learning_rate": 2.9025691671274894e-07, + "loss": 0.1872, + "step": 7097 + }, + { + "epoch": 0.37, + "learning_rate": 2.9022674274125204e-07, + "loss": 0.1585, + "step": 7098 + }, + { + "epoch": 0.37, + "learning_rate": 2.9019656619104005e-07, + "loss": 0.1071, + "step": 7099 + }, + { + "epoch": 0.37, + "learning_rate": 2.901663870629753e-07, + "loss": 0.1124, + "step": 7100 + }, + { + "epoch": 0.37, + "learning_rate": 2.901362053579204e-07, + "loss": 0.1386, + "step": 7101 + }, + { + "epoch": 0.37, + "learning_rate": 2.90106021076738e-07, + "loss": 0.1509, + "step": 7102 + }, + { + "epoch": 0.37, + "learning_rate": 2.9007583422029076e-07, + "loss": 0.1484, + "step": 7103 + }, + { + "epoch": 0.37, + "learning_rate": 2.9004564478944133e-07, + "loss": 0.1193, + "step": 7104 + }, + { + "epoch": 0.37, + "learning_rate": 2.9001545278505265e-07, + "loss": 0.1596, + "step": 7105 + }, + { + "epoch": 0.37, + "learning_rate": 2.899852582079876e-07, + "loss": 0.1514, + "step": 7106 + }, + { + "epoch": 0.37, + "learning_rate": 2.8995506105910904e-07, + "loss": 0.1568, + "step": 7107 + }, + { + "epoch": 0.37, + "learning_rate": 2.8992486133928015e-07, + "loss": 0.1325, + "step": 7108 + }, + { + "epoch": 0.37, + "learning_rate": 2.8989465904936404e-07, + "loss": 0.1751, + "step": 7109 + }, + { + "epoch": 0.37, + "learning_rate": 2.8986445419022395e-07, + "loss": 0.1191, + "step": 7110 + }, + { + "epoch": 0.37, + "learning_rate": 2.8983424676272293e-07, + "loss": 0.1201, + "step": 7111 + }, + { + "epoch": 0.37, + "learning_rate": 2.8980403676772455e-07, + "loss": 0.1219, + "step": 7112 + }, + { + "epoch": 0.37, + "learning_rate": 2.8977382420609204e-07, + "loss": 0.1082, + "step": 7113 + }, + { + "epoch": 0.37, + "learning_rate": 2.8974360907868906e-07, + "loss": 0.1409, + "step": 7114 + }, + { + "epoch": 0.37, + "learning_rate": 2.8971339138637904e-07, + "loss": 0.1555, + "step": 7115 + }, + { + "epoch": 0.37, + "learning_rate": 2.8968317113002564e-07, + "loss": 0.137, + "step": 7116 + }, + { + "epoch": 0.37, + "learning_rate": 2.896529483104927e-07, + "loss": 0.1518, + "step": 7117 + }, + { + "epoch": 0.37, + "learning_rate": 2.896227229286437e-07, + "loss": 0.1512, + "step": 7118 + }, + { + "epoch": 0.37, + "learning_rate": 2.8959249498534293e-07, + "loss": 0.1483, + "step": 7119 + }, + { + "epoch": 0.37, + "learning_rate": 2.895622644814539e-07, + "loss": 0.1573, + "step": 7120 + }, + { + "epoch": 0.37, + "learning_rate": 2.895320314178408e-07, + "loss": 0.1402, + "step": 7121 + }, + { + "epoch": 0.37, + "learning_rate": 2.8950179579536763e-07, + "loss": 0.1395, + "step": 7122 + }, + { + "epoch": 0.37, + "learning_rate": 2.894715576148987e-07, + "loss": 0.1312, + "step": 7123 + }, + { + "epoch": 0.37, + "learning_rate": 2.8944131687729806e-07, + "loss": 0.1567, + "step": 7124 + }, + { + "epoch": 0.37, + "learning_rate": 2.8941107358343e-07, + "loss": 0.1138, + "step": 7125 + }, + { + "epoch": 0.37, + "learning_rate": 2.8938082773415903e-07, + "loss": 0.1134, + "step": 7126 + }, + { + "epoch": 0.37, + "learning_rate": 2.8935057933034943e-07, + "loss": 0.1828, + "step": 7127 + }, + { + "epoch": 0.37, + "learning_rate": 2.893203283728659e-07, + "loss": 0.156, + "step": 7128 + }, + { + "epoch": 0.37, + "learning_rate": 2.8929007486257274e-07, + "loss": 0.1419, + "step": 7129 + }, + { + "epoch": 0.37, + "learning_rate": 2.8925981880033487e-07, + "loss": 0.1464, + "step": 7130 + }, + { + "epoch": 0.37, + "learning_rate": 2.892295601870169e-07, + "loss": 0.156, + "step": 7131 + }, + { + "epoch": 0.37, + "learning_rate": 2.8919929902348364e-07, + "loss": 0.1151, + "step": 7132 + }, + { + "epoch": 0.37, + "learning_rate": 2.891690353106e-07, + "loss": 0.1497, + "step": 7133 + }, + { + "epoch": 0.37, + "learning_rate": 2.8913876904923087e-07, + "loss": 0.1334, + "step": 7134 + }, + { + "epoch": 0.37, + "learning_rate": 2.8910850024024135e-07, + "loss": 0.1538, + "step": 7135 + }, + { + "epoch": 0.37, + "learning_rate": 2.890782288844965e-07, + "loss": 0.1174, + "step": 7136 + }, + { + "epoch": 0.37, + "learning_rate": 2.890479549828615e-07, + "loss": 0.1767, + "step": 7137 + }, + { + "epoch": 0.37, + "learning_rate": 2.8901767853620147e-07, + "loss": 0.1323, + "step": 7138 + }, + { + "epoch": 0.37, + "learning_rate": 2.8898739954538195e-07, + "loss": 0.108, + "step": 7139 + }, + { + "epoch": 0.37, + "learning_rate": 2.889571180112681e-07, + "loss": 0.1039, + "step": 7140 + }, + { + "epoch": 0.37, + "learning_rate": 2.889268339347255e-07, + "loss": 0.1658, + "step": 7141 + }, + { + "epoch": 0.37, + "learning_rate": 2.888965473166197e-07, + "loss": 0.1152, + "step": 7142 + }, + { + "epoch": 0.37, + "learning_rate": 2.888662581578162e-07, + "loss": 0.1717, + "step": 7143 + }, + { + "epoch": 0.37, + "learning_rate": 2.8883596645918076e-07, + "loss": 0.1381, + "step": 7144 + }, + { + "epoch": 0.37, + "learning_rate": 2.8880567222157913e-07, + "loss": 0.1336, + "step": 7145 + }, + { + "epoch": 0.37, + "learning_rate": 2.887753754458771e-07, + "loss": 0.1445, + "step": 7146 + }, + { + "epoch": 0.37, + "learning_rate": 2.887450761329405e-07, + "loss": 0.1332, + "step": 7147 + }, + { + "epoch": 0.37, + "learning_rate": 2.8871477428363544e-07, + "loss": 0.1646, + "step": 7148 + }, + { + "epoch": 0.37, + "learning_rate": 2.8868446989882793e-07, + "loss": 0.1204, + "step": 7149 + }, + { + "epoch": 0.37, + "learning_rate": 2.886541629793839e-07, + "loss": 0.1399, + "step": 7150 + }, + { + "epoch": 0.37, + "learning_rate": 2.8862385352616977e-07, + "loss": 0.1145, + "step": 7151 + }, + { + "epoch": 0.37, + "learning_rate": 2.8859354154005164e-07, + "loss": 0.1245, + "step": 7152 + }, + { + "epoch": 0.37, + "learning_rate": 2.8856322702189595e-07, + "loss": 0.1063, + "step": 7153 + }, + { + "epoch": 0.37, + "learning_rate": 2.8853290997256897e-07, + "loss": 0.1434, + "step": 7154 + }, + { + "epoch": 0.37, + "learning_rate": 2.8850259039293726e-07, + "loss": 0.1084, + "step": 7155 + }, + { + "epoch": 0.37, + "learning_rate": 2.884722682838674e-07, + "loss": 0.1286, + "step": 7156 + }, + { + "epoch": 0.37, + "learning_rate": 2.8844194364622596e-07, + "loss": 0.1438, + "step": 7157 + }, + { + "epoch": 0.37, + "learning_rate": 2.8841161648087957e-07, + "loss": 0.0989, + "step": 7158 + }, + { + "epoch": 0.37, + "learning_rate": 2.883812867886951e-07, + "loss": 0.1006, + "step": 7159 + }, + { + "epoch": 0.37, + "learning_rate": 2.883509545705393e-07, + "loss": 0.133, + "step": 7160 + }, + { + "epoch": 0.37, + "learning_rate": 2.8832061982727913e-07, + "loss": 0.1365, + "step": 7161 + }, + { + "epoch": 0.37, + "learning_rate": 2.8829028255978154e-07, + "loss": 0.1329, + "step": 7162 + }, + { + "epoch": 0.37, + "learning_rate": 2.882599427689136e-07, + "loss": 0.1547, + "step": 7163 + }, + { + "epoch": 0.37, + "learning_rate": 2.8822960045554243e-07, + "loss": 0.1317, + "step": 7164 + }, + { + "epoch": 0.37, + "learning_rate": 2.8819925562053517e-07, + "loss": 0.1511, + "step": 7165 + }, + { + "epoch": 0.37, + "learning_rate": 2.8816890826475923e-07, + "loss": 0.1313, + "step": 7166 + }, + { + "epoch": 0.37, + "learning_rate": 2.8813855838908177e-07, + "loss": 0.1323, + "step": 7167 + }, + { + "epoch": 0.37, + "learning_rate": 2.881082059943703e-07, + "loss": 0.1439, + "step": 7168 + }, + { + "epoch": 0.37, + "learning_rate": 2.880778510814923e-07, + "loss": 0.1351, + "step": 7169 + }, + { + "epoch": 0.37, + "learning_rate": 2.8804749365131534e-07, + "loss": 0.191, + "step": 7170 + }, + { + "epoch": 0.37, + "learning_rate": 2.8801713370470704e-07, + "loss": 0.1217, + "step": 7171 + }, + { + "epoch": 0.37, + "learning_rate": 2.8798677124253503e-07, + "loss": 0.1409, + "step": 7172 + }, + { + "epoch": 0.37, + "learning_rate": 2.8795640626566715e-07, + "loss": 0.1769, + "step": 7173 + }, + { + "epoch": 0.37, + "learning_rate": 2.879260387749712e-07, + "loss": 0.1385, + "step": 7174 + }, + { + "epoch": 0.37, + "learning_rate": 2.878956687713152e-07, + "loss": 0.1479, + "step": 7175 + }, + { + "epoch": 0.37, + "learning_rate": 2.8786529625556697e-07, + "loss": 0.1771, + "step": 7176 + }, + { + "epoch": 0.37, + "learning_rate": 2.8783492122859467e-07, + "loss": 0.1504, + "step": 7177 + }, + { + "epoch": 0.37, + "learning_rate": 2.878045436912664e-07, + "loss": 0.1628, + "step": 7178 + }, + { + "epoch": 0.37, + "learning_rate": 2.877741636444504e-07, + "loss": 0.1332, + "step": 7179 + }, + { + "epoch": 0.37, + "learning_rate": 2.877437810890149e-07, + "loss": 0.1504, + "step": 7180 + }, + { + "epoch": 0.37, + "learning_rate": 2.8771339602582827e-07, + "loss": 0.2007, + "step": 7181 + }, + { + "epoch": 0.37, + "learning_rate": 2.8768300845575894e-07, + "loss": 0.1503, + "step": 7182 + }, + { + "epoch": 0.37, + "learning_rate": 2.876526183796754e-07, + "loss": 0.1361, + "step": 7183 + }, + { + "epoch": 0.37, + "learning_rate": 2.8762222579844613e-07, + "loss": 0.1435, + "step": 7184 + }, + { + "epoch": 0.38, + "learning_rate": 2.8759183071293987e-07, + "loss": 0.1293, + "step": 7185 + }, + { + "epoch": 0.38, + "learning_rate": 2.8756143312402524e-07, + "loss": 0.1203, + "step": 7186 + }, + { + "epoch": 0.38, + "learning_rate": 2.8753103303257103e-07, + "loss": 0.1582, + "step": 7187 + }, + { + "epoch": 0.38, + "learning_rate": 2.875006304394462e-07, + "loss": 0.1308, + "step": 7188 + }, + { + "epoch": 0.38, + "learning_rate": 2.874702253455195e-07, + "loss": 0.1561, + "step": 7189 + }, + { + "epoch": 0.38, + "learning_rate": 2.8743981775165996e-07, + "loss": 0.1369, + "step": 7190 + }, + { + "epoch": 0.38, + "learning_rate": 2.874094076587368e-07, + "loss": 0.1286, + "step": 7191 + }, + { + "epoch": 0.38, + "learning_rate": 2.8737899506761894e-07, + "loss": 0.1342, + "step": 7192 + }, + { + "epoch": 0.38, + "learning_rate": 2.8734857997917567e-07, + "loss": 0.1692, + "step": 7193 + }, + { + "epoch": 0.38, + "learning_rate": 2.873181623942762e-07, + "loss": 0.149, + "step": 7194 + }, + { + "epoch": 0.38, + "learning_rate": 2.8728774231379007e-07, + "loss": 0.1574, + "step": 7195 + }, + { + "epoch": 0.38, + "learning_rate": 2.872573197385865e-07, + "loss": 0.1442, + "step": 7196 + }, + { + "epoch": 0.38, + "learning_rate": 2.8722689466953504e-07, + "loss": 0.1747, + "step": 7197 + }, + { + "epoch": 0.38, + "learning_rate": 2.871964671075053e-07, + "loss": 0.1193, + "step": 7198 + }, + { + "epoch": 0.38, + "learning_rate": 2.871660370533668e-07, + "loss": 0.1262, + "step": 7199 + }, + { + "epoch": 0.38, + "learning_rate": 2.871356045079894e-07, + "loss": 0.1657, + "step": 7200 + }, + { + "epoch": 0.38, + "learning_rate": 2.871051694722427e-07, + "loss": 0.1059, + "step": 7201 + }, + { + "epoch": 0.38, + "learning_rate": 2.870747319469967e-07, + "loss": 0.1254, + "step": 7202 + }, + { + "epoch": 0.38, + "learning_rate": 2.8704429193312115e-07, + "loss": 0.1323, + "step": 7203 + }, + { + "epoch": 0.38, + "learning_rate": 2.8701384943148627e-07, + "loss": 0.1182, + "step": 7204 + }, + { + "epoch": 0.38, + "learning_rate": 2.869834044429619e-07, + "loss": 0.1148, + "step": 7205 + }, + { + "epoch": 0.38, + "learning_rate": 2.8695295696841826e-07, + "loss": 0.2012, + "step": 7206 + }, + { + "epoch": 0.38, + "learning_rate": 2.869225070087255e-07, + "loss": 0.1717, + "step": 7207 + }, + { + "epoch": 0.38, + "learning_rate": 2.8689205456475396e-07, + "loss": 0.1232, + "step": 7208 + }, + { + "epoch": 0.38, + "learning_rate": 2.86861599637374e-07, + "loss": 0.1205, + "step": 7209 + }, + { + "epoch": 0.38, + "learning_rate": 2.86831142227456e-07, + "loss": 0.1332, + "step": 7210 + }, + { + "epoch": 0.38, + "learning_rate": 2.868006823358704e-07, + "loss": 0.1269, + "step": 7211 + }, + { + "epoch": 0.38, + "learning_rate": 2.8677021996348777e-07, + "loss": 0.1451, + "step": 7212 + }, + { + "epoch": 0.38, + "learning_rate": 2.8673975511117884e-07, + "loss": 0.1367, + "step": 7213 + }, + { + "epoch": 0.38, + "learning_rate": 2.867092877798141e-07, + "loss": 0.1617, + "step": 7214 + }, + { + "epoch": 0.38, + "learning_rate": 2.8667881797026456e-07, + "loss": 0.1609, + "step": 7215 + }, + { + "epoch": 0.38, + "learning_rate": 2.866483456834008e-07, + "loss": 0.1324, + "step": 7216 + }, + { + "epoch": 0.38, + "learning_rate": 2.86617870920094e-07, + "loss": 0.1255, + "step": 7217 + }, + { + "epoch": 0.38, + "learning_rate": 2.86587393681215e-07, + "loss": 0.1223, + "step": 7218 + }, + { + "epoch": 0.38, + "learning_rate": 2.865569139676348e-07, + "loss": 0.193, + "step": 7219 + }, + { + "epoch": 0.38, + "learning_rate": 2.8652643178022464e-07, + "loss": 0.1338, + "step": 7220 + }, + { + "epoch": 0.38, + "learning_rate": 2.8649594711985563e-07, + "loss": 0.1382, + "step": 7221 + }, + { + "epoch": 0.38, + "learning_rate": 2.864654599873991e-07, + "loss": 0.1189, + "step": 7222 + }, + { + "epoch": 0.38, + "learning_rate": 2.864349703837263e-07, + "loss": 0.1578, + "step": 7223 + }, + { + "epoch": 0.38, + "learning_rate": 2.8640447830970867e-07, + "loss": 0.1213, + "step": 7224 + }, + { + "epoch": 0.38, + "learning_rate": 2.863739837662177e-07, + "loss": 0.1103, + "step": 7225 + }, + { + "epoch": 0.38, + "learning_rate": 2.863434867541249e-07, + "loss": 0.1313, + "step": 7226 + }, + { + "epoch": 0.38, + "learning_rate": 2.86312987274302e-07, + "loss": 0.1453, + "step": 7227 + }, + { + "epoch": 0.38, + "learning_rate": 2.8628248532762055e-07, + "loss": 0.1372, + "step": 7228 + }, + { + "epoch": 0.38, + "learning_rate": 2.862519809149524e-07, + "loss": 0.138, + "step": 7229 + }, + { + "epoch": 0.38, + "learning_rate": 2.862214740371693e-07, + "loss": 0.1557, + "step": 7230 + }, + { + "epoch": 0.38, + "learning_rate": 2.861909646951432e-07, + "loss": 0.1393, + "step": 7231 + }, + { + "epoch": 0.38, + "learning_rate": 2.861604528897461e-07, + "loss": 0.1311, + "step": 7232 + }, + { + "epoch": 0.38, + "learning_rate": 2.861299386218499e-07, + "loss": 0.1291, + "step": 7233 + }, + { + "epoch": 0.38, + "learning_rate": 2.8609942189232683e-07, + "loss": 0.1258, + "step": 7234 + }, + { + "epoch": 0.38, + "learning_rate": 2.860689027020491e-07, + "loss": 0.1315, + "step": 7235 + }, + { + "epoch": 0.38, + "learning_rate": 2.860383810518889e-07, + "loss": 0.1267, + "step": 7236 + }, + { + "epoch": 0.38, + "learning_rate": 2.860078569427185e-07, + "loss": 0.1526, + "step": 7237 + }, + { + "epoch": 0.38, + "learning_rate": 2.859773303754104e-07, + "loss": 0.1267, + "step": 7238 + }, + { + "epoch": 0.38, + "learning_rate": 2.85946801350837e-07, + "loss": 0.1143, + "step": 7239 + }, + { + "epoch": 0.38, + "learning_rate": 2.859162698698708e-07, + "loss": 0.1437, + "step": 7240 + }, + { + "epoch": 0.38, + "learning_rate": 2.858857359333844e-07, + "loss": 0.143, + "step": 7241 + }, + { + "epoch": 0.38, + "learning_rate": 2.8585519954225064e-07, + "loss": 0.1524, + "step": 7242 + }, + { + "epoch": 0.38, + "learning_rate": 2.8582466069734207e-07, + "loss": 0.1283, + "step": 7243 + }, + { + "epoch": 0.38, + "learning_rate": 2.857941193995315e-07, + "loss": 0.1377, + "step": 7244 + }, + { + "epoch": 0.38, + "learning_rate": 2.85763575649692e-07, + "loss": 0.1286, + "step": 7245 + }, + { + "epoch": 0.38, + "learning_rate": 2.857330294486963e-07, + "loss": 0.1562, + "step": 7246 + }, + { + "epoch": 0.38, + "learning_rate": 2.8570248079741764e-07, + "loss": 0.1834, + "step": 7247 + }, + { + "epoch": 0.38, + "learning_rate": 2.856719296967289e-07, + "loss": 0.1419, + "step": 7248 + }, + { + "epoch": 0.38, + "learning_rate": 2.8564137614750335e-07, + "loss": 0.1258, + "step": 7249 + }, + { + "epoch": 0.38, + "learning_rate": 2.856108201506142e-07, + "loss": 0.1676, + "step": 7250 + }, + { + "epoch": 0.38, + "learning_rate": 2.8558026170693474e-07, + "loss": 0.0972, + "step": 7251 + }, + { + "epoch": 0.38, + "learning_rate": 2.8554970081733844e-07, + "loss": 0.1701, + "step": 7252 + }, + { + "epoch": 0.38, + "learning_rate": 2.855191374826986e-07, + "loss": 0.1589, + "step": 7253 + }, + { + "epoch": 0.38, + "learning_rate": 2.8548857170388883e-07, + "loss": 0.1509, + "step": 7254 + }, + { + "epoch": 0.38, + "learning_rate": 2.854580034817826e-07, + "loss": 0.1562, + "step": 7255 + }, + { + "epoch": 0.38, + "learning_rate": 2.8542743281725374e-07, + "loss": 0.1156, + "step": 7256 + }, + { + "epoch": 0.38, + "learning_rate": 2.8539685971117584e-07, + "loss": 0.1071, + "step": 7257 + }, + { + "epoch": 0.38, + "learning_rate": 2.8536628416442274e-07, + "loss": 0.1393, + "step": 7258 + }, + { + "epoch": 0.38, + "learning_rate": 2.8533570617786826e-07, + "loss": 0.1605, + "step": 7259 + }, + { + "epoch": 0.38, + "learning_rate": 2.853051257523864e-07, + "loss": 0.1628, + "step": 7260 + }, + { + "epoch": 0.38, + "learning_rate": 2.8527454288885104e-07, + "loss": 0.1258, + "step": 7261 + }, + { + "epoch": 0.38, + "learning_rate": 2.852439575881363e-07, + "loss": 0.1133, + "step": 7262 + }, + { + "epoch": 0.38, + "learning_rate": 2.852133698511165e-07, + "loss": 0.1582, + "step": 7263 + }, + { + "epoch": 0.38, + "learning_rate": 2.8518277967866555e-07, + "loss": 0.1507, + "step": 7264 + }, + { + "epoch": 0.38, + "learning_rate": 2.8515218707165796e-07, + "loss": 0.1253, + "step": 7265 + }, + { + "epoch": 0.38, + "learning_rate": 2.8512159203096795e-07, + "loss": 0.1323, + "step": 7266 + }, + { + "epoch": 0.38, + "learning_rate": 2.8509099455747e-07, + "loss": 0.1449, + "step": 7267 + }, + { + "epoch": 0.38, + "learning_rate": 2.850603946520385e-07, + "loss": 0.1523, + "step": 7268 + }, + { + "epoch": 0.38, + "learning_rate": 2.850297923155482e-07, + "loss": 0.1602, + "step": 7269 + }, + { + "epoch": 0.38, + "learning_rate": 2.849991875488735e-07, + "loss": 0.1317, + "step": 7270 + }, + { + "epoch": 0.38, + "learning_rate": 2.849685803528893e-07, + "loss": 0.1163, + "step": 7271 + }, + { + "epoch": 0.38, + "learning_rate": 2.849379707284702e-07, + "loss": 0.1219, + "step": 7272 + }, + { + "epoch": 0.38, + "learning_rate": 2.849073586764912e-07, + "loss": 0.1387, + "step": 7273 + }, + { + "epoch": 0.38, + "learning_rate": 2.84876744197827e-07, + "loss": 0.1719, + "step": 7274 + }, + { + "epoch": 0.38, + "learning_rate": 2.8484612729335275e-07, + "loss": 0.1475, + "step": 7275 + }, + { + "epoch": 0.38, + "learning_rate": 2.8481550796394344e-07, + "loss": 0.1575, + "step": 7276 + }, + { + "epoch": 0.38, + "learning_rate": 2.8478488621047416e-07, + "loss": 0.141, + "step": 7277 + }, + { + "epoch": 0.38, + "learning_rate": 2.847542620338201e-07, + "loss": 0.1105, + "step": 7278 + }, + { + "epoch": 0.38, + "learning_rate": 2.8472363543485647e-07, + "loss": 0.1229, + "step": 7279 + }, + { + "epoch": 0.38, + "learning_rate": 2.8469300641445866e-07, + "loss": 0.1401, + "step": 7280 + }, + { + "epoch": 0.38, + "learning_rate": 2.8466237497350205e-07, + "loss": 0.134, + "step": 7281 + }, + { + "epoch": 0.38, + "learning_rate": 2.846317411128621e-07, + "loss": 0.1481, + "step": 7282 + }, + { + "epoch": 0.38, + "learning_rate": 2.846011048334143e-07, + "loss": 0.1497, + "step": 7283 + }, + { + "epoch": 0.38, + "learning_rate": 2.845704661360342e-07, + "loss": 0.1477, + "step": 7284 + }, + { + "epoch": 0.38, + "learning_rate": 2.8453982502159757e-07, + "loss": 0.1223, + "step": 7285 + }, + { + "epoch": 0.38, + "learning_rate": 2.8450918149098015e-07, + "loss": 0.1292, + "step": 7286 + }, + { + "epoch": 0.38, + "learning_rate": 2.8447853554505765e-07, + "loss": 0.1187, + "step": 7287 + }, + { + "epoch": 0.38, + "learning_rate": 2.8444788718470596e-07, + "loss": 0.1873, + "step": 7288 + }, + { + "epoch": 0.38, + "learning_rate": 2.8441723641080117e-07, + "loss": 0.1337, + "step": 7289 + }, + { + "epoch": 0.38, + "learning_rate": 2.8438658322421906e-07, + "loss": 0.1575, + "step": 7290 + }, + { + "epoch": 0.38, + "learning_rate": 2.843559276258359e-07, + "loss": 0.1176, + "step": 7291 + }, + { + "epoch": 0.38, + "learning_rate": 2.843252696165277e-07, + "loss": 0.1816, + "step": 7292 + }, + { + "epoch": 0.38, + "learning_rate": 2.8429460919717074e-07, + "loss": 0.1177, + "step": 7293 + }, + { + "epoch": 0.38, + "learning_rate": 2.842639463686414e-07, + "loss": 0.139, + "step": 7294 + }, + { + "epoch": 0.38, + "learning_rate": 2.8423328113181584e-07, + "loss": 0.151, + "step": 7295 + }, + { + "epoch": 0.38, + "learning_rate": 2.8420261348757066e-07, + "loss": 0.1596, + "step": 7296 + }, + { + "epoch": 0.38, + "learning_rate": 2.841719434367822e-07, + "loss": 0.1264, + "step": 7297 + }, + { + "epoch": 0.38, + "learning_rate": 2.841412709803272e-07, + "loss": 0.1397, + "step": 7298 + }, + { + "epoch": 0.38, + "learning_rate": 2.8411059611908215e-07, + "loss": 0.1638, + "step": 7299 + }, + { + "epoch": 0.38, + "learning_rate": 2.840799188539238e-07, + "loss": 0.1163, + "step": 7300 + }, + { + "epoch": 0.38, + "learning_rate": 2.8404923918572897e-07, + "loss": 0.1586, + "step": 7301 + }, + { + "epoch": 0.38, + "learning_rate": 2.840185571153743e-07, + "loss": 0.1217, + "step": 7302 + }, + { + "epoch": 0.38, + "learning_rate": 2.8398787264373694e-07, + "loss": 0.1187, + "step": 7303 + }, + { + "epoch": 0.38, + "learning_rate": 2.839571857716937e-07, + "loss": 0.1264, + "step": 7304 + }, + { + "epoch": 0.38, + "learning_rate": 2.8392649650012173e-07, + "loss": 0.1448, + "step": 7305 + }, + { + "epoch": 0.38, + "learning_rate": 2.8389580482989807e-07, + "loss": 0.1638, + "step": 7306 + }, + { + "epoch": 0.38, + "learning_rate": 2.838651107618999e-07, + "loss": 0.1139, + "step": 7307 + }, + { + "epoch": 0.38, + "learning_rate": 2.838344142970045e-07, + "loss": 0.1522, + "step": 7308 + }, + { + "epoch": 0.38, + "learning_rate": 2.838037154360892e-07, + "loss": 0.1746, + "step": 7309 + }, + { + "epoch": 0.38, + "learning_rate": 2.8377301418003133e-07, + "loss": 0.1324, + "step": 7310 + }, + { + "epoch": 0.38, + "learning_rate": 2.8374231052970835e-07, + "loss": 0.1587, + "step": 7311 + }, + { + "epoch": 0.38, + "learning_rate": 2.837116044859979e-07, + "loss": 0.1269, + "step": 7312 + }, + { + "epoch": 0.38, + "learning_rate": 2.836808960497774e-07, + "loss": 0.1304, + "step": 7313 + }, + { + "epoch": 0.38, + "learning_rate": 2.8365018522192466e-07, + "loss": 0.151, + "step": 7314 + }, + { + "epoch": 0.38, + "learning_rate": 2.8361947200331723e-07, + "loss": 0.1442, + "step": 7315 + }, + { + "epoch": 0.38, + "learning_rate": 2.835887563948331e-07, + "loss": 0.1227, + "step": 7316 + }, + { + "epoch": 0.38, + "learning_rate": 2.8355803839735e-07, + "loss": 0.1288, + "step": 7317 + }, + { + "epoch": 0.38, + "learning_rate": 2.835273180117459e-07, + "loss": 0.1705, + "step": 7318 + }, + { + "epoch": 0.38, + "learning_rate": 2.834965952388988e-07, + "loss": 0.1572, + "step": 7319 + }, + { + "epoch": 0.38, + "learning_rate": 2.8346587007968683e-07, + "loss": 0.1081, + "step": 7320 + }, + { + "epoch": 0.38, + "learning_rate": 2.8343514253498803e-07, + "loss": 0.1569, + "step": 7321 + }, + { + "epoch": 0.38, + "learning_rate": 2.834044126056806e-07, + "loss": 0.149, + "step": 7322 + }, + { + "epoch": 0.38, + "learning_rate": 2.83373680292643e-07, + "loss": 0.1704, + "step": 7323 + }, + { + "epoch": 0.38, + "learning_rate": 2.833429455967533e-07, + "loss": 0.1213, + "step": 7324 + }, + { + "epoch": 0.38, + "learning_rate": 2.8331220851889014e-07, + "loss": 0.1436, + "step": 7325 + }, + { + "epoch": 0.38, + "learning_rate": 2.832814690599318e-07, + "loss": 0.1555, + "step": 7326 + }, + { + "epoch": 0.38, + "learning_rate": 2.83250727220757e-07, + "loss": 0.1422, + "step": 7327 + }, + { + "epoch": 0.38, + "learning_rate": 2.832199830022442e-07, + "loss": 0.1468, + "step": 7328 + }, + { + "epoch": 0.38, + "learning_rate": 2.8318923640527225e-07, + "loss": 0.1171, + "step": 7329 + }, + { + "epoch": 0.38, + "learning_rate": 2.8315848743071977e-07, + "loss": 0.1437, + "step": 7330 + }, + { + "epoch": 0.38, + "learning_rate": 2.831277360794656e-07, + "loss": 0.1345, + "step": 7331 + }, + { + "epoch": 0.38, + "learning_rate": 2.830969823523887e-07, + "loss": 0.1629, + "step": 7332 + }, + { + "epoch": 0.38, + "learning_rate": 2.830662262503679e-07, + "loss": 0.136, + "step": 7333 + }, + { + "epoch": 0.38, + "learning_rate": 2.8303546777428234e-07, + "loss": 0.1716, + "step": 7334 + }, + { + "epoch": 0.38, + "learning_rate": 2.83004706925011e-07, + "loss": 0.1448, + "step": 7335 + }, + { + "epoch": 0.38, + "learning_rate": 2.8297394370343313e-07, + "loss": 0.1344, + "step": 7336 + }, + { + "epoch": 0.38, + "learning_rate": 2.829431781104279e-07, + "loss": 0.1615, + "step": 7337 + }, + { + "epoch": 0.38, + "learning_rate": 2.829124101468746e-07, + "loss": 0.1446, + "step": 7338 + }, + { + "epoch": 0.38, + "learning_rate": 2.8288163981365267e-07, + "loss": 0.1558, + "step": 7339 + }, + { + "epoch": 0.38, + "learning_rate": 2.828508671116414e-07, + "loss": 0.159, + "step": 7340 + }, + { + "epoch": 0.38, + "learning_rate": 2.8282009204172045e-07, + "loss": 0.1954, + "step": 7341 + }, + { + "epoch": 0.38, + "learning_rate": 2.8278931460476925e-07, + "loss": 0.1503, + "step": 7342 + }, + { + "epoch": 0.38, + "learning_rate": 2.827585348016675e-07, + "loss": 0.1226, + "step": 7343 + }, + { + "epoch": 0.38, + "learning_rate": 2.827277526332948e-07, + "loss": 0.1581, + "step": 7344 + }, + { + "epoch": 0.38, + "learning_rate": 2.8269696810053113e-07, + "loss": 0.146, + "step": 7345 + }, + { + "epoch": 0.38, + "learning_rate": 2.826661812042561e-07, + "loss": 0.1258, + "step": 7346 + }, + { + "epoch": 0.38, + "learning_rate": 2.826353919453497e-07, + "loss": 0.1799, + "step": 7347 + }, + { + "epoch": 0.38, + "learning_rate": 2.8260460032469194e-07, + "loss": 0.1608, + "step": 7348 + }, + { + "epoch": 0.38, + "learning_rate": 2.825738063431627e-07, + "loss": 0.1706, + "step": 7349 + }, + { + "epoch": 0.38, + "learning_rate": 2.825430100016424e-07, + "loss": 0.1588, + "step": 7350 + }, + { + "epoch": 0.38, + "learning_rate": 2.8251221130101087e-07, + "loss": 0.1529, + "step": 7351 + }, + { + "epoch": 0.38, + "learning_rate": 2.824814102421485e-07, + "loss": 0.2024, + "step": 7352 + }, + { + "epoch": 0.38, + "learning_rate": 2.824506068259356e-07, + "loss": 0.1394, + "step": 7353 + }, + { + "epoch": 0.38, + "learning_rate": 2.824198010532525e-07, + "loss": 0.1499, + "step": 7354 + }, + { + "epoch": 0.38, + "learning_rate": 2.8238899292497975e-07, + "loss": 0.1381, + "step": 7355 + }, + { + "epoch": 0.38, + "learning_rate": 2.8235818244199764e-07, + "loss": 0.1096, + "step": 7356 + }, + { + "epoch": 0.38, + "learning_rate": 2.82327369605187e-07, + "loss": 0.1372, + "step": 7357 + }, + { + "epoch": 0.38, + "learning_rate": 2.822965544154283e-07, + "loss": 0.1393, + "step": 7358 + }, + { + "epoch": 0.38, + "learning_rate": 2.8226573687360235e-07, + "loss": 0.1644, + "step": 7359 + }, + { + "epoch": 0.38, + "learning_rate": 2.822349169805898e-07, + "loss": 0.1262, + "step": 7360 + }, + { + "epoch": 0.38, + "learning_rate": 2.8220409473727167e-07, + "loss": 0.1426, + "step": 7361 + }, + { + "epoch": 0.38, + "learning_rate": 2.8217327014452875e-07, + "loss": 0.1488, + "step": 7362 + }, + { + "epoch": 0.38, + "learning_rate": 2.8214244320324204e-07, + "loss": 0.1444, + "step": 7363 + }, + { + "epoch": 0.38, + "learning_rate": 2.821116139142926e-07, + "loss": 0.1143, + "step": 7364 + }, + { + "epoch": 0.38, + "learning_rate": 2.820807822785615e-07, + "loss": 0.1519, + "step": 7365 + }, + { + "epoch": 0.38, + "learning_rate": 2.8204994829693e-07, + "loss": 0.137, + "step": 7366 + }, + { + "epoch": 0.38, + "learning_rate": 2.820191119702793e-07, + "loss": 0.1266, + "step": 7367 + }, + { + "epoch": 0.38, + "learning_rate": 2.819882732994907e-07, + "loss": 0.151, + "step": 7368 + }, + { + "epoch": 0.38, + "learning_rate": 2.8195743228544557e-07, + "loss": 0.1671, + "step": 7369 + }, + { + "epoch": 0.38, + "learning_rate": 2.819265889290254e-07, + "loss": 0.1166, + "step": 7370 + }, + { + "epoch": 0.38, + "learning_rate": 2.8189574323111174e-07, + "loss": 0.1577, + "step": 7371 + }, + { + "epoch": 0.38, + "learning_rate": 2.8186489519258605e-07, + "loss": 0.1498, + "step": 7372 + }, + { + "epoch": 0.38, + "learning_rate": 2.818340448143301e-07, + "loss": 0.1558, + "step": 7373 + }, + { + "epoch": 0.38, + "learning_rate": 2.8180319209722555e-07, + "loss": 0.1222, + "step": 7374 + }, + { + "epoch": 0.38, + "learning_rate": 2.817723370421542e-07, + "loss": 0.1515, + "step": 7375 + }, + { + "epoch": 0.38, + "learning_rate": 2.817414796499979e-07, + "loss": 0.1769, + "step": 7376 + }, + { + "epoch": 0.39, + "learning_rate": 2.817106199216385e-07, + "loss": 0.11, + "step": 7377 + }, + { + "epoch": 0.39, + "learning_rate": 2.816797578579581e-07, + "loss": 0.1485, + "step": 7378 + }, + { + "epoch": 0.39, + "learning_rate": 2.816488934598387e-07, + "loss": 0.1349, + "step": 7379 + }, + { + "epoch": 0.39, + "learning_rate": 2.816180267281624e-07, + "loss": 0.1242, + "step": 7380 + }, + { + "epoch": 0.39, + "learning_rate": 2.8158715766381134e-07, + "loss": 0.1183, + "step": 7381 + }, + { + "epoch": 0.39, + "learning_rate": 2.815562862676678e-07, + "loss": 0.1291, + "step": 7382 + }, + { + "epoch": 0.39, + "learning_rate": 2.815254125406142e-07, + "loss": 0.1331, + "step": 7383 + }, + { + "epoch": 0.39, + "learning_rate": 2.814945364835328e-07, + "loss": 0.1613, + "step": 7384 + }, + { + "epoch": 0.39, + "learning_rate": 2.814636580973061e-07, + "loss": 0.1422, + "step": 7385 + }, + { + "epoch": 0.39, + "learning_rate": 2.8143277738281656e-07, + "loss": 0.1301, + "step": 7386 + }, + { + "epoch": 0.39, + "learning_rate": 2.814018943409468e-07, + "loss": 0.1446, + "step": 7387 + }, + { + "epoch": 0.39, + "learning_rate": 2.813710089725795e-07, + "loss": 0.1528, + "step": 7388 + }, + { + "epoch": 0.39, + "learning_rate": 2.8134012127859736e-07, + "loss": 0.1422, + "step": 7389 + }, + { + "epoch": 0.39, + "learning_rate": 2.813092312598832e-07, + "loss": 0.1882, + "step": 7390 + }, + { + "epoch": 0.39, + "learning_rate": 2.8127833891731974e-07, + "loss": 0.1525, + "step": 7391 + }, + { + "epoch": 0.39, + "learning_rate": 2.8124744425179e-07, + "loss": 0.1468, + "step": 7392 + }, + { + "epoch": 0.39, + "learning_rate": 2.8121654726417697e-07, + "loss": 0.1449, + "step": 7393 + }, + { + "epoch": 0.39, + "learning_rate": 2.811856479553636e-07, + "loss": 0.1632, + "step": 7394 + }, + { + "epoch": 0.39, + "learning_rate": 2.8115474632623316e-07, + "loss": 0.1731, + "step": 7395 + }, + { + "epoch": 0.39, + "learning_rate": 2.811238423776686e-07, + "loss": 0.1107, + "step": 7396 + }, + { + "epoch": 0.39, + "learning_rate": 2.810929361105535e-07, + "loss": 0.1652, + "step": 7397 + }, + { + "epoch": 0.39, + "learning_rate": 2.8106202752577087e-07, + "loss": 0.1437, + "step": 7398 + }, + { + "epoch": 0.39, + "learning_rate": 2.8103111662420425e-07, + "loss": 0.1481, + "step": 7399 + }, + { + "epoch": 0.39, + "learning_rate": 2.8100020340673695e-07, + "loss": 0.132, + "step": 7400 + }, + { + "epoch": 0.39, + "learning_rate": 2.8096928787425263e-07, + "loss": 0.1151, + "step": 7401 + }, + { + "epoch": 0.39, + "learning_rate": 2.809383700276348e-07, + "loss": 0.112, + "step": 7402 + }, + { + "epoch": 0.39, + "learning_rate": 2.809074498677671e-07, + "loss": 0.1404, + "step": 7403 + }, + { + "epoch": 0.39, + "learning_rate": 2.808765273955332e-07, + "loss": 0.1262, + "step": 7404 + }, + { + "epoch": 0.39, + "learning_rate": 2.80845602611817e-07, + "loss": 0.1437, + "step": 7405 + }, + { + "epoch": 0.39, + "learning_rate": 2.8081467551750225e-07, + "loss": 0.1543, + "step": 7406 + }, + { + "epoch": 0.39, + "learning_rate": 2.8078374611347286e-07, + "loss": 0.15, + "step": 7407 + }, + { + "epoch": 0.39, + "learning_rate": 2.807528144006129e-07, + "loss": 0.1324, + "step": 7408 + }, + { + "epoch": 0.39, + "learning_rate": 2.807218803798062e-07, + "loss": 0.139, + "step": 7409 + }, + { + "epoch": 0.39, + "learning_rate": 2.806909440519371e-07, + "loss": 0.1482, + "step": 7410 + }, + { + "epoch": 0.39, + "learning_rate": 2.806600054178896e-07, + "loss": 0.1477, + "step": 7411 + }, + { + "epoch": 0.39, + "learning_rate": 2.8062906447854803e-07, + "loss": 0.1185, + "step": 7412 + }, + { + "epoch": 0.39, + "learning_rate": 2.8059812123479666e-07, + "loss": 0.1308, + "step": 7413 + }, + { + "epoch": 0.39, + "learning_rate": 2.8056717568751986e-07, + "loss": 0.1398, + "step": 7414 + }, + { + "epoch": 0.39, + "learning_rate": 2.8053622783760216e-07, + "loss": 0.1675, + "step": 7415 + }, + { + "epoch": 0.39, + "learning_rate": 2.805052776859279e-07, + "loss": 0.0962, + "step": 7416 + }, + { + "epoch": 0.39, + "learning_rate": 2.804743252333818e-07, + "loss": 0.1341, + "step": 7417 + }, + { + "epoch": 0.39, + "learning_rate": 2.804433704808483e-07, + "loss": 0.1269, + "step": 7418 + }, + { + "epoch": 0.39, + "learning_rate": 2.8041241342921227e-07, + "loss": 0.1577, + "step": 7419 + }, + { + "epoch": 0.39, + "learning_rate": 2.8038145407935847e-07, + "loss": 0.1411, + "step": 7420 + }, + { + "epoch": 0.39, + "learning_rate": 2.803504924321716e-07, + "loss": 0.1396, + "step": 7421 + }, + { + "epoch": 0.39, + "learning_rate": 2.8031952848853673e-07, + "loss": 0.1244, + "step": 7422 + }, + { + "epoch": 0.39, + "learning_rate": 2.802885622493387e-07, + "loss": 0.1674, + "step": 7423 + }, + { + "epoch": 0.39, + "learning_rate": 2.802575937154626e-07, + "loss": 0.1157, + "step": 7424 + }, + { + "epoch": 0.39, + "learning_rate": 2.802266228877934e-07, + "loss": 0.1241, + "step": 7425 + }, + { + "epoch": 0.39, + "learning_rate": 2.8019564976721637e-07, + "loss": 0.1635, + "step": 7426 + }, + { + "epoch": 0.39, + "learning_rate": 2.8016467435461675e-07, + "loss": 0.1351, + "step": 7427 + }, + { + "epoch": 0.39, + "learning_rate": 2.801336966508797e-07, + "loss": 0.1315, + "step": 7428 + }, + { + "epoch": 0.39, + "learning_rate": 2.801027166568908e-07, + "loss": 0.13, + "step": 7429 + }, + { + "epoch": 0.39, + "learning_rate": 2.8007173437353527e-07, + "loss": 0.1267, + "step": 7430 + }, + { + "epoch": 0.39, + "learning_rate": 2.8004074980169866e-07, + "loss": 0.124, + "step": 7431 + }, + { + "epoch": 0.39, + "learning_rate": 2.800097629422665e-07, + "loss": 0.1346, + "step": 7432 + }, + { + "epoch": 0.39, + "learning_rate": 2.799787737961245e-07, + "loss": 0.1785, + "step": 7433 + }, + { + "epoch": 0.39, + "learning_rate": 2.7994778236415817e-07, + "loss": 0.1221, + "step": 7434 + }, + { + "epoch": 0.39, + "learning_rate": 2.7991678864725343e-07, + "loss": 0.1472, + "step": 7435 + }, + { + "epoch": 0.39, + "learning_rate": 2.79885792646296e-07, + "loss": 0.1653, + "step": 7436 + }, + { + "epoch": 0.39, + "learning_rate": 2.798547943621718e-07, + "loss": 0.1245, + "step": 7437 + }, + { + "epoch": 0.39, + "learning_rate": 2.7982379379576667e-07, + "loss": 0.1553, + "step": 7438 + }, + { + "epoch": 0.39, + "learning_rate": 2.7979279094796677e-07, + "loss": 0.1695, + "step": 7439 + }, + { + "epoch": 0.39, + "learning_rate": 2.797617858196581e-07, + "loss": 0.1423, + "step": 7440 + }, + { + "epoch": 0.39, + "learning_rate": 2.797307784117268e-07, + "loss": 0.1236, + "step": 7441 + }, + { + "epoch": 0.39, + "learning_rate": 2.796997687250591e-07, + "loss": 0.1232, + "step": 7442 + }, + { + "epoch": 0.39, + "learning_rate": 2.7966875676054115e-07, + "loss": 0.1478, + "step": 7443 + }, + { + "epoch": 0.39, + "learning_rate": 2.796377425190594e-07, + "loss": 0.1214, + "step": 7444 + }, + { + "epoch": 0.39, + "learning_rate": 2.796067260015003e-07, + "loss": 0.1155, + "step": 7445 + }, + { + "epoch": 0.39, + "learning_rate": 2.7957570720875017e-07, + "loss": 0.1944, + "step": 7446 + }, + { + "epoch": 0.39, + "learning_rate": 2.795446861416956e-07, + "loss": 0.131, + "step": 7447 + }, + { + "epoch": 0.39, + "learning_rate": 2.795136628012233e-07, + "loss": 0.1534, + "step": 7448 + }, + { + "epoch": 0.39, + "learning_rate": 2.794826371882197e-07, + "loss": 0.1387, + "step": 7449 + }, + { + "epoch": 0.39, + "learning_rate": 2.794516093035717e-07, + "loss": 0.1239, + "step": 7450 + }, + { + "epoch": 0.39, + "learning_rate": 2.7942057914816605e-07, + "loss": 0.1249, + "step": 7451 + }, + { + "epoch": 0.39, + "learning_rate": 2.793895467228895e-07, + "loss": 0.1339, + "step": 7452 + }, + { + "epoch": 0.39, + "learning_rate": 2.7935851202862917e-07, + "loss": 0.1178, + "step": 7453 + }, + { + "epoch": 0.39, + "learning_rate": 2.7932747506627186e-07, + "loss": 0.1256, + "step": 7454 + }, + { + "epoch": 0.39, + "learning_rate": 2.7929643583670476e-07, + "loss": 0.1018, + "step": 7455 + }, + { + "epoch": 0.39, + "learning_rate": 2.792653943408148e-07, + "loss": 0.1347, + "step": 7456 + }, + { + "epoch": 0.39, + "learning_rate": 2.792343505794894e-07, + "loss": 0.142, + "step": 7457 + }, + { + "epoch": 0.39, + "learning_rate": 2.792033045536156e-07, + "loss": 0.1387, + "step": 7458 + }, + { + "epoch": 0.39, + "learning_rate": 2.7917225626408074e-07, + "loss": 0.126, + "step": 7459 + }, + { + "epoch": 0.39, + "learning_rate": 2.791412057117723e-07, + "loss": 0.1142, + "step": 7460 + }, + { + "epoch": 0.39, + "learning_rate": 2.7911015289757756e-07, + "loss": 0.1062, + "step": 7461 + }, + { + "epoch": 0.39, + "learning_rate": 2.790790978223843e-07, + "loss": 0.1347, + "step": 7462 + }, + { + "epoch": 0.39, + "learning_rate": 2.7904804048707966e-07, + "loss": 0.1427, + "step": 7463 + }, + { + "epoch": 0.39, + "learning_rate": 2.790169808925517e-07, + "loss": 0.1251, + "step": 7464 + }, + { + "epoch": 0.39, + "learning_rate": 2.789859190396878e-07, + "loss": 0.131, + "step": 7465 + }, + { + "epoch": 0.39, + "learning_rate": 2.789548549293759e-07, + "loss": 0.1226, + "step": 7466 + }, + { + "epoch": 0.39, + "learning_rate": 2.7892378856250377e-07, + "loss": 0.1359, + "step": 7467 + }, + { + "epoch": 0.39, + "learning_rate": 2.7889271993995923e-07, + "loss": 0.1277, + "step": 7468 + }, + { + "epoch": 0.39, + "learning_rate": 2.788616490626304e-07, + "loss": 0.1523, + "step": 7469 + }, + { + "epoch": 0.39, + "learning_rate": 2.7883057593140514e-07, + "loss": 0.131, + "step": 7470 + }, + { + "epoch": 0.39, + "learning_rate": 2.787995005471716e-07, + "loss": 0.1397, + "step": 7471 + }, + { + "epoch": 0.39, + "learning_rate": 2.787684229108179e-07, + "loss": 0.1301, + "step": 7472 + }, + { + "epoch": 0.39, + "learning_rate": 2.787373430232323e-07, + "loss": 0.1347, + "step": 7473 + }, + { + "epoch": 0.39, + "learning_rate": 2.78706260885303e-07, + "loss": 0.1042, + "step": 7474 + }, + { + "epoch": 0.39, + "learning_rate": 2.7867517649791834e-07, + "loss": 0.1442, + "step": 7475 + }, + { + "epoch": 0.39, + "learning_rate": 2.7864408986196685e-07, + "loss": 0.1845, + "step": 7476 + }, + { + "epoch": 0.39, + "learning_rate": 2.786130009783369e-07, + "loss": 0.1354, + "step": 7477 + }, + { + "epoch": 0.39, + "learning_rate": 2.78581909847917e-07, + "loss": 0.1675, + "step": 7478 + }, + { + "epoch": 0.39, + "learning_rate": 2.785508164715958e-07, + "loss": 0.148, + "step": 7479 + }, + { + "epoch": 0.39, + "learning_rate": 2.7851972085026197e-07, + "loss": 0.1498, + "step": 7480 + }, + { + "epoch": 0.39, + "learning_rate": 2.784886229848041e-07, + "loss": 0.1534, + "step": 7481 + }, + { + "epoch": 0.39, + "learning_rate": 2.784575228761112e-07, + "loss": 0.1293, + "step": 7482 + }, + { + "epoch": 0.39, + "learning_rate": 2.78426420525072e-07, + "loss": 0.1319, + "step": 7483 + }, + { + "epoch": 0.39, + "learning_rate": 2.783953159325754e-07, + "loss": 0.1316, + "step": 7484 + }, + { + "epoch": 0.39, + "learning_rate": 2.7836420909951036e-07, + "loss": 0.1383, + "step": 7485 + }, + { + "epoch": 0.39, + "learning_rate": 2.78333100026766e-07, + "loss": 0.1447, + "step": 7486 + }, + { + "epoch": 0.39, + "learning_rate": 2.7830198871523145e-07, + "loss": 0.1417, + "step": 7487 + }, + { + "epoch": 0.39, + "learning_rate": 2.7827087516579575e-07, + "loss": 0.1471, + "step": 7488 + }, + { + "epoch": 0.39, + "learning_rate": 2.782397593793483e-07, + "loss": 0.1491, + "step": 7489 + }, + { + "epoch": 0.39, + "learning_rate": 2.7820864135677827e-07, + "loss": 0.1419, + "step": 7490 + }, + { + "epoch": 0.39, + "learning_rate": 2.7817752109897514e-07, + "loss": 0.1048, + "step": 7491 + }, + { + "epoch": 0.39, + "learning_rate": 2.7814639860682827e-07, + "loss": 0.1453, + "step": 7492 + }, + { + "epoch": 0.39, + "learning_rate": 2.7811527388122715e-07, + "loss": 0.1678, + "step": 7493 + }, + { + "epoch": 0.39, + "learning_rate": 2.7808414692306127e-07, + "loss": 0.1206, + "step": 7494 + }, + { + "epoch": 0.39, + "learning_rate": 2.780530177332204e-07, + "loss": 0.1327, + "step": 7495 + }, + { + "epoch": 0.39, + "learning_rate": 2.7802188631259413e-07, + "loss": 0.1513, + "step": 7496 + }, + { + "epoch": 0.39, + "learning_rate": 2.7799075266207223e-07, + "loss": 0.1235, + "step": 7497 + }, + { + "epoch": 0.39, + "learning_rate": 2.7795961678254454e-07, + "loss": 0.1316, + "step": 7498 + }, + { + "epoch": 0.39, + "learning_rate": 2.7792847867490086e-07, + "loss": 0.1053, + "step": 7499 + }, + { + "epoch": 0.39, + "learning_rate": 2.7789733834003126e-07, + "loss": 0.1205, + "step": 7500 + }, + { + "epoch": 0.39, + "learning_rate": 2.778661957788255e-07, + "loss": 0.1178, + "step": 7501 + }, + { + "epoch": 0.39, + "learning_rate": 2.7783505099217394e-07, + "loss": 0.1197, + "step": 7502 + }, + { + "epoch": 0.39, + "learning_rate": 2.778039039809665e-07, + "loss": 0.1418, + "step": 7503 + }, + { + "epoch": 0.39, + "learning_rate": 2.7777275474609353e-07, + "loss": 0.1577, + "step": 7504 + }, + { + "epoch": 0.39, + "learning_rate": 2.777416032884452e-07, + "loss": 0.1187, + "step": 7505 + }, + { + "epoch": 0.39, + "learning_rate": 2.7771044960891177e-07, + "loss": 0.1313, + "step": 7506 + }, + { + "epoch": 0.39, + "learning_rate": 2.776792937083837e-07, + "loss": 0.1637, + "step": 7507 + }, + { + "epoch": 0.39, + "learning_rate": 2.7764813558775146e-07, + "loss": 0.1431, + "step": 7508 + }, + { + "epoch": 0.39, + "learning_rate": 2.776169752479056e-07, + "loss": 0.1835, + "step": 7509 + }, + { + "epoch": 0.39, + "learning_rate": 2.775858126897365e-07, + "loss": 0.1167, + "step": 7510 + }, + { + "epoch": 0.39, + "learning_rate": 2.7755464791413497e-07, + "loss": 0.1607, + "step": 7511 + }, + { + "epoch": 0.39, + "learning_rate": 2.775234809219916e-07, + "loss": 0.1561, + "step": 7512 + }, + { + "epoch": 0.39, + "learning_rate": 2.774923117141973e-07, + "loss": 0.1355, + "step": 7513 + }, + { + "epoch": 0.39, + "learning_rate": 2.774611402916428e-07, + "loss": 0.1534, + "step": 7514 + }, + { + "epoch": 0.39, + "learning_rate": 2.7742996665521905e-07, + "loss": 0.1293, + "step": 7515 + }, + { + "epoch": 0.39, + "learning_rate": 2.773987908058169e-07, + "loss": 0.1404, + "step": 7516 + }, + { + "epoch": 0.39, + "learning_rate": 2.7736761274432747e-07, + "loss": 0.1479, + "step": 7517 + }, + { + "epoch": 0.39, + "learning_rate": 2.7733643247164176e-07, + "loss": 0.1547, + "step": 7518 + }, + { + "epoch": 0.39, + "learning_rate": 2.7730524998865094e-07, + "loss": 0.12, + "step": 7519 + }, + { + "epoch": 0.39, + "learning_rate": 2.772740652962463e-07, + "loss": 0.1201, + "step": 7520 + }, + { + "epoch": 0.39, + "learning_rate": 2.7724287839531893e-07, + "loss": 0.1321, + "step": 7521 + }, + { + "epoch": 0.39, + "learning_rate": 2.7721168928676043e-07, + "loss": 0.1457, + "step": 7522 + }, + { + "epoch": 0.39, + "learning_rate": 2.7718049797146193e-07, + "loss": 0.1667, + "step": 7523 + }, + { + "epoch": 0.39, + "learning_rate": 2.7714930445031504e-07, + "loss": 0.1676, + "step": 7524 + }, + { + "epoch": 0.39, + "learning_rate": 2.7711810872421127e-07, + "loss": 0.1384, + "step": 7525 + }, + { + "epoch": 0.39, + "learning_rate": 2.770869107940421e-07, + "loss": 0.1091, + "step": 7526 + }, + { + "epoch": 0.39, + "learning_rate": 2.770557106606994e-07, + "loss": 0.1608, + "step": 7527 + }, + { + "epoch": 0.39, + "learning_rate": 2.7702450832507464e-07, + "loss": 0.1322, + "step": 7528 + }, + { + "epoch": 0.39, + "learning_rate": 2.769933037880597e-07, + "loss": 0.1403, + "step": 7529 + }, + { + "epoch": 0.39, + "learning_rate": 2.769620970505465e-07, + "loss": 0.1388, + "step": 7530 + }, + { + "epoch": 0.39, + "learning_rate": 2.7693088811342674e-07, + "loss": 0.117, + "step": 7531 + }, + { + "epoch": 0.39, + "learning_rate": 2.7689967697759253e-07, + "loss": 0.1263, + "step": 7532 + }, + { + "epoch": 0.39, + "learning_rate": 2.768684636439359e-07, + "loss": 0.1307, + "step": 7533 + }, + { + "epoch": 0.39, + "learning_rate": 2.768372481133489e-07, + "loss": 0.1616, + "step": 7534 + }, + { + "epoch": 0.39, + "learning_rate": 2.768060303867237e-07, + "loss": 0.1308, + "step": 7535 + }, + { + "epoch": 0.39, + "learning_rate": 2.7677481046495247e-07, + "loss": 0.1677, + "step": 7536 + }, + { + "epoch": 0.39, + "learning_rate": 2.7674358834892746e-07, + "loss": 0.1177, + "step": 7537 + }, + { + "epoch": 0.39, + "learning_rate": 2.767123640395412e-07, + "loss": 0.1291, + "step": 7538 + }, + { + "epoch": 0.39, + "learning_rate": 2.766811375376859e-07, + "loss": 0.1194, + "step": 7539 + }, + { + "epoch": 0.39, + "learning_rate": 2.76649908844254e-07, + "loss": 0.1365, + "step": 7540 + }, + { + "epoch": 0.39, + "learning_rate": 2.766186779601383e-07, + "loss": 0.1528, + "step": 7541 + }, + { + "epoch": 0.39, + "learning_rate": 2.7658744488623105e-07, + "loss": 0.1884, + "step": 7542 + }, + { + "epoch": 0.39, + "learning_rate": 2.7655620962342514e-07, + "loss": 0.1271, + "step": 7543 + }, + { + "epoch": 0.39, + "learning_rate": 2.7652497217261317e-07, + "loss": 0.1482, + "step": 7544 + }, + { + "epoch": 0.39, + "learning_rate": 2.7649373253468795e-07, + "loss": 0.1291, + "step": 7545 + }, + { + "epoch": 0.39, + "learning_rate": 2.7646249071054233e-07, + "loss": 0.1135, + "step": 7546 + }, + { + "epoch": 0.39, + "learning_rate": 2.764312467010693e-07, + "loss": 0.1165, + "step": 7547 + }, + { + "epoch": 0.39, + "learning_rate": 2.764000005071616e-07, + "loss": 0.1369, + "step": 7548 + }, + { + "epoch": 0.39, + "learning_rate": 2.7636875212971246e-07, + "loss": 0.1404, + "step": 7549 + }, + { + "epoch": 0.39, + "learning_rate": 2.7633750156961487e-07, + "loss": 0.1618, + "step": 7550 + }, + { + "epoch": 0.39, + "learning_rate": 2.7630624882776206e-07, + "loss": 0.1366, + "step": 7551 + }, + { + "epoch": 0.39, + "learning_rate": 2.7627499390504717e-07, + "loss": 0.1797, + "step": 7552 + }, + { + "epoch": 0.39, + "learning_rate": 2.762437368023635e-07, + "loss": 0.1391, + "step": 7553 + }, + { + "epoch": 0.39, + "learning_rate": 2.762124775206044e-07, + "loss": 0.1433, + "step": 7554 + }, + { + "epoch": 0.39, + "learning_rate": 2.761812160606633e-07, + "loss": 0.1435, + "step": 7555 + }, + { + "epoch": 0.39, + "learning_rate": 2.7614995242343366e-07, + "loss": 0.1413, + "step": 7556 + }, + { + "epoch": 0.39, + "learning_rate": 2.761186866098089e-07, + "loss": 0.1582, + "step": 7557 + }, + { + "epoch": 0.39, + "learning_rate": 2.7608741862068275e-07, + "loss": 0.1425, + "step": 7558 + }, + { + "epoch": 0.39, + "learning_rate": 2.7605614845694876e-07, + "loss": 0.1208, + "step": 7559 + }, + { + "epoch": 0.39, + "learning_rate": 2.760248761195007e-07, + "loss": 0.1132, + "step": 7560 + }, + { + "epoch": 0.39, + "learning_rate": 2.759936016092324e-07, + "loss": 0.2069, + "step": 7561 + }, + { + "epoch": 0.39, + "learning_rate": 2.7596232492703754e-07, + "loss": 0.193, + "step": 7562 + }, + { + "epoch": 0.39, + "learning_rate": 2.759310460738101e-07, + "loss": 0.1642, + "step": 7563 + }, + { + "epoch": 0.39, + "learning_rate": 2.7589976505044417e-07, + "loss": 0.1772, + "step": 7564 + }, + { + "epoch": 0.39, + "learning_rate": 2.7586848185783356e-07, + "loss": 0.1168, + "step": 7565 + }, + { + "epoch": 0.39, + "learning_rate": 2.7583719649687245e-07, + "loss": 0.1454, + "step": 7566 + }, + { + "epoch": 0.39, + "learning_rate": 2.758059089684551e-07, + "loss": 0.1331, + "step": 7567 + }, + { + "epoch": 0.4, + "learning_rate": 2.7577461927347545e-07, + "loss": 0.1291, + "step": 7568 + }, + { + "epoch": 0.4, + "learning_rate": 2.75743327412828e-07, + "loss": 0.1317, + "step": 7569 + }, + { + "epoch": 0.4, + "learning_rate": 2.75712033387407e-07, + "loss": 0.1403, + "step": 7570 + }, + { + "epoch": 0.4, + "learning_rate": 2.756807371981069e-07, + "loss": 0.1508, + "step": 7571 + }, + { + "epoch": 0.4, + "learning_rate": 2.7564943884582204e-07, + "loss": 0.1129, + "step": 7572 + }, + { + "epoch": 0.4, + "learning_rate": 2.756181383314471e-07, + "loss": 0.1295, + "step": 7573 + }, + { + "epoch": 0.4, + "learning_rate": 2.755868356558765e-07, + "loss": 0.1353, + "step": 7574 + }, + { + "epoch": 0.4, + "learning_rate": 2.75555530820005e-07, + "loss": 0.1734, + "step": 7575 + }, + { + "epoch": 0.4, + "learning_rate": 2.755242238247272e-07, + "loss": 0.1907, + "step": 7576 + }, + { + "epoch": 0.4, + "learning_rate": 2.7549291467093805e-07, + "loss": 0.126, + "step": 7577 + }, + { + "epoch": 0.4, + "learning_rate": 2.7546160335953213e-07, + "loss": 0.1419, + "step": 7578 + }, + { + "epoch": 0.4, + "learning_rate": 2.754302898914045e-07, + "loss": 0.1141, + "step": 7579 + }, + { + "epoch": 0.4, + "learning_rate": 2.753989742674501e-07, + "loss": 0.162, + "step": 7580 + }, + { + "epoch": 0.4, + "learning_rate": 2.753676564885639e-07, + "loss": 0.1179, + "step": 7581 + }, + { + "epoch": 0.4, + "learning_rate": 2.75336336555641e-07, + "loss": 0.1062, + "step": 7582 + }, + { + "epoch": 0.4, + "learning_rate": 2.7530501446957644e-07, + "loss": 0.1363, + "step": 7583 + }, + { + "epoch": 0.4, + "learning_rate": 2.752736902312655e-07, + "loss": 0.1377, + "step": 7584 + }, + { + "epoch": 0.4, + "learning_rate": 2.7524236384160354e-07, + "loss": 0.1437, + "step": 7585 + }, + { + "epoch": 0.4, + "learning_rate": 2.7521103530148574e-07, + "loss": 0.1338, + "step": 7586 + }, + { + "epoch": 0.4, + "learning_rate": 2.7517970461180744e-07, + "loss": 0.1637, + "step": 7587 + }, + { + "epoch": 0.4, + "learning_rate": 2.751483717734642e-07, + "loss": 0.1121, + "step": 7588 + }, + { + "epoch": 0.4, + "learning_rate": 2.7511703678735147e-07, + "loss": 0.137, + "step": 7589 + }, + { + "epoch": 0.4, + "learning_rate": 2.750856996543649e-07, + "loss": 0.1162, + "step": 7590 + }, + { + "epoch": 0.4, + "learning_rate": 2.750543603754e-07, + "loss": 0.1298, + "step": 7591 + }, + { + "epoch": 0.4, + "learning_rate": 2.750230189513525e-07, + "loss": 0.1961, + "step": 7592 + }, + { + "epoch": 0.4, + "learning_rate": 2.7499167538311813e-07, + "loss": 0.1266, + "step": 7593 + }, + { + "epoch": 0.4, + "learning_rate": 2.749603296715928e-07, + "loss": 0.1391, + "step": 7594 + }, + { + "epoch": 0.4, + "learning_rate": 2.749289818176723e-07, + "loss": 0.1362, + "step": 7595 + }, + { + "epoch": 0.4, + "learning_rate": 2.7489763182225246e-07, + "loss": 0.1684, + "step": 7596 + }, + { + "epoch": 0.4, + "learning_rate": 2.748662796862295e-07, + "loss": 0.0976, + "step": 7597 + }, + { + "epoch": 0.4, + "learning_rate": 2.7483492541049933e-07, + "loss": 0.1275, + "step": 7598 + }, + { + "epoch": 0.4, + "learning_rate": 2.7480356899595815e-07, + "loss": 0.1529, + "step": 7599 + }, + { + "epoch": 0.4, + "learning_rate": 2.74772210443502e-07, + "loss": 0.1293, + "step": 7600 + }, + { + "epoch": 0.4, + "learning_rate": 2.747408497540273e-07, + "loss": 0.1426, + "step": 7601 + }, + { + "epoch": 0.4, + "learning_rate": 2.747094869284302e-07, + "loss": 0.1013, + "step": 7602 + }, + { + "epoch": 0.4, + "learning_rate": 2.7467812196760717e-07, + "loss": 0.1373, + "step": 7603 + }, + { + "epoch": 0.4, + "learning_rate": 2.7464675487245454e-07, + "loss": 0.1198, + "step": 7604 + }, + { + "epoch": 0.4, + "learning_rate": 2.7461538564386886e-07, + "loss": 0.1403, + "step": 7605 + }, + { + "epoch": 0.4, + "learning_rate": 2.7458401428274666e-07, + "loss": 0.1378, + "step": 7606 + }, + { + "epoch": 0.4, + "learning_rate": 2.745526407899845e-07, + "loss": 0.141, + "step": 7607 + }, + { + "epoch": 0.4, + "learning_rate": 2.7452126516647917e-07, + "loss": 0.1195, + "step": 7608 + }, + { + "epoch": 0.4, + "learning_rate": 2.744898874131273e-07, + "loss": 0.1655, + "step": 7609 + }, + { + "epoch": 0.4, + "learning_rate": 2.744585075308256e-07, + "loss": 0.116, + "step": 7610 + }, + { + "epoch": 0.4, + "learning_rate": 2.7442712552047104e-07, + "loss": 0.1507, + "step": 7611 + }, + { + "epoch": 0.4, + "learning_rate": 2.743957413829606e-07, + "loss": 0.1344, + "step": 7612 + }, + { + "epoch": 0.4, + "learning_rate": 2.7436435511919103e-07, + "loss": 0.1605, + "step": 7613 + }, + { + "epoch": 0.4, + "learning_rate": 2.7433296673005956e-07, + "loss": 0.1387, + "step": 7614 + }, + { + "epoch": 0.4, + "learning_rate": 2.7430157621646315e-07, + "loss": 0.1435, + "step": 7615 + }, + { + "epoch": 0.4, + "learning_rate": 2.742701835792991e-07, + "loss": 0.1317, + "step": 7616 + }, + { + "epoch": 0.4, + "learning_rate": 2.742387888194644e-07, + "loss": 0.1193, + "step": 7617 + }, + { + "epoch": 0.4, + "learning_rate": 2.7420739193785656e-07, + "loss": 0.1566, + "step": 7618 + }, + { + "epoch": 0.4, + "learning_rate": 2.7417599293537277e-07, + "loss": 0.1209, + "step": 7619 + }, + { + "epoch": 0.4, + "learning_rate": 2.7414459181291046e-07, + "loss": 0.1485, + "step": 7620 + }, + { + "epoch": 0.4, + "learning_rate": 2.741131885713671e-07, + "loss": 0.1296, + "step": 7621 + }, + { + "epoch": 0.4, + "learning_rate": 2.740817832116402e-07, + "loss": 0.1454, + "step": 7622 + }, + { + "epoch": 0.4, + "learning_rate": 2.7405037573462737e-07, + "loss": 0.1235, + "step": 7623 + }, + { + "epoch": 0.4, + "learning_rate": 2.740189661412262e-07, + "loss": 0.1586, + "step": 7624 + }, + { + "epoch": 0.4, + "learning_rate": 2.739875544323344e-07, + "loss": 0.1273, + "step": 7625 + }, + { + "epoch": 0.4, + "learning_rate": 2.739561406088497e-07, + "loss": 0.1466, + "step": 7626 + }, + { + "epoch": 0.4, + "learning_rate": 2.7392472467166996e-07, + "loss": 0.1504, + "step": 7627 + }, + { + "epoch": 0.4, + "learning_rate": 2.7389330662169307e-07, + "loss": 0.1566, + "step": 7628 + }, + { + "epoch": 0.4, + "learning_rate": 2.7386188645981703e-07, + "loss": 0.1682, + "step": 7629 + }, + { + "epoch": 0.4, + "learning_rate": 2.7383046418693964e-07, + "loss": 0.1427, + "step": 7630 + }, + { + "epoch": 0.4, + "learning_rate": 2.7379903980395914e-07, + "loss": 0.1304, + "step": 7631 + }, + { + "epoch": 0.4, + "learning_rate": 2.737676133117736e-07, + "loss": 0.1622, + "step": 7632 + }, + { + "epoch": 0.4, + "learning_rate": 2.7373618471128114e-07, + "loss": 0.1363, + "step": 7633 + }, + { + "epoch": 0.4, + "learning_rate": 2.7370475400338004e-07, + "loss": 0.1625, + "step": 7634 + }, + { + "epoch": 0.4, + "learning_rate": 2.7367332118896873e-07, + "loss": 0.1517, + "step": 7635 + }, + { + "epoch": 0.4, + "learning_rate": 2.7364188626894537e-07, + "loss": 0.1377, + "step": 7636 + }, + { + "epoch": 0.4, + "learning_rate": 2.736104492442085e-07, + "loss": 0.1484, + "step": 7637 + }, + { + "epoch": 0.4, + "learning_rate": 2.735790101156565e-07, + "loss": 0.16, + "step": 7638 + }, + { + "epoch": 0.4, + "learning_rate": 2.7354756888418813e-07, + "loss": 0.1445, + "step": 7639 + }, + { + "epoch": 0.4, + "learning_rate": 2.7351612555070176e-07, + "loss": 0.1623, + "step": 7640 + }, + { + "epoch": 0.4, + "learning_rate": 2.7348468011609616e-07, + "loss": 0.1511, + "step": 7641 + }, + { + "epoch": 0.4, + "learning_rate": 2.7345323258127004e-07, + "loss": 0.1598, + "step": 7642 + }, + { + "epoch": 0.4, + "learning_rate": 2.7342178294712215e-07, + "loss": 0.1333, + "step": 7643 + }, + { + "epoch": 0.4, + "learning_rate": 2.733903312145514e-07, + "loss": 0.1332, + "step": 7644 + }, + { + "epoch": 0.4, + "learning_rate": 2.7335887738445657e-07, + "loss": 0.1454, + "step": 7645 + }, + { + "epoch": 0.4, + "learning_rate": 2.7332742145773686e-07, + "loss": 0.2228, + "step": 7646 + }, + { + "epoch": 0.4, + "learning_rate": 2.73295963435291e-07, + "loss": 0.1169, + "step": 7647 + }, + { + "epoch": 0.4, + "learning_rate": 2.732645033180183e-07, + "loss": 0.1387, + "step": 7648 + }, + { + "epoch": 0.4, + "learning_rate": 2.7323304110681767e-07, + "loss": 0.1656, + "step": 7649 + }, + { + "epoch": 0.4, + "learning_rate": 2.732015768025886e-07, + "loss": 0.1312, + "step": 7650 + }, + { + "epoch": 0.4, + "learning_rate": 2.7317011040623015e-07, + "loss": 0.1166, + "step": 7651 + }, + { + "epoch": 0.4, + "learning_rate": 2.7313864191864166e-07, + "loss": 0.1546, + "step": 7652 + }, + { + "epoch": 0.4, + "learning_rate": 2.731071713407226e-07, + "loss": 0.1813, + "step": 7653 + }, + { + "epoch": 0.4, + "learning_rate": 2.7307569867337237e-07, + "loss": 0.1276, + "step": 7654 + }, + { + "epoch": 0.4, + "learning_rate": 2.7304422391749045e-07, + "loss": 0.139, + "step": 7655 + }, + { + "epoch": 0.4, + "learning_rate": 2.730127470739764e-07, + "loss": 0.1689, + "step": 7656 + }, + { + "epoch": 0.4, + "learning_rate": 2.729812681437299e-07, + "loss": 0.1251, + "step": 7657 + }, + { + "epoch": 0.4, + "learning_rate": 2.729497871276505e-07, + "loss": 0.0954, + "step": 7658 + }, + { + "epoch": 0.4, + "learning_rate": 2.729183040266381e-07, + "loss": 0.143, + "step": 7659 + }, + { + "epoch": 0.4, + "learning_rate": 2.728868188415924e-07, + "loss": 0.1367, + "step": 7660 + }, + { + "epoch": 0.4, + "learning_rate": 2.728553315734133e-07, + "loss": 0.1438, + "step": 7661 + }, + { + "epoch": 0.4, + "learning_rate": 2.7282384222300063e-07, + "loss": 0.1648, + "step": 7662 + }, + { + "epoch": 0.4, + "learning_rate": 2.7279235079125445e-07, + "loss": 0.164, + "step": 7663 + }, + { + "epoch": 0.4, + "learning_rate": 2.7276085727907486e-07, + "loss": 0.142, + "step": 7664 + }, + { + "epoch": 0.4, + "learning_rate": 2.727293616873618e-07, + "loss": 0.1112, + "step": 7665 + }, + { + "epoch": 0.4, + "learning_rate": 2.7269786401701555e-07, + "loss": 0.1606, + "step": 7666 + }, + { + "epoch": 0.4, + "learning_rate": 2.7266636426893625e-07, + "loss": 0.1413, + "step": 7667 + }, + { + "epoch": 0.4, + "learning_rate": 2.7263486244402423e-07, + "loss": 0.1356, + "step": 7668 + }, + { + "epoch": 0.4, + "learning_rate": 2.726033585431798e-07, + "loss": 0.1531, + "step": 7669 + }, + { + "epoch": 0.4, + "learning_rate": 2.7257185256730334e-07, + "loss": 0.1312, + "step": 7670 + }, + { + "epoch": 0.4, + "learning_rate": 2.725403445172953e-07, + "loss": 0.1286, + "step": 7671 + }, + { + "epoch": 0.4, + "learning_rate": 2.7250883439405625e-07, + "loss": 0.1246, + "step": 7672 + }, + { + "epoch": 0.4, + "learning_rate": 2.7247732219848675e-07, + "loss": 0.1749, + "step": 7673 + }, + { + "epoch": 0.4, + "learning_rate": 2.724458079314873e-07, + "loss": 0.133, + "step": 7674 + }, + { + "epoch": 0.4, + "learning_rate": 2.7241429159395876e-07, + "loss": 0.2082, + "step": 7675 + }, + { + "epoch": 0.4, + "learning_rate": 2.7238277318680185e-07, + "loss": 0.1143, + "step": 7676 + }, + { + "epoch": 0.4, + "learning_rate": 2.723512527109173e-07, + "loss": 0.1665, + "step": 7677 + }, + { + "epoch": 0.4, + "learning_rate": 2.72319730167206e-07, + "loss": 0.1358, + "step": 7678 + }, + { + "epoch": 0.4, + "learning_rate": 2.722882055565689e-07, + "loss": 0.1179, + "step": 7679 + }, + { + "epoch": 0.4, + "learning_rate": 2.72256678879907e-07, + "loss": 0.1601, + "step": 7680 + }, + { + "epoch": 0.4, + "learning_rate": 2.722251501381213e-07, + "loss": 0.1406, + "step": 7681 + }, + { + "epoch": 0.4, + "learning_rate": 2.7219361933211294e-07, + "loss": 0.1471, + "step": 7682 + }, + { + "epoch": 0.4, + "learning_rate": 2.7216208646278303e-07, + "loss": 0.1182, + "step": 7683 + }, + { + "epoch": 0.4, + "learning_rate": 2.7213055153103294e-07, + "loss": 0.1322, + "step": 7684 + }, + { + "epoch": 0.4, + "learning_rate": 2.720990145377637e-07, + "loss": 0.1314, + "step": 7685 + }, + { + "epoch": 0.4, + "learning_rate": 2.720674754838769e-07, + "loss": 0.135, + "step": 7686 + }, + { + "epoch": 0.4, + "learning_rate": 2.720359343702738e-07, + "loss": 0.1424, + "step": 7687 + }, + { + "epoch": 0.4, + "learning_rate": 2.720043911978559e-07, + "loss": 0.1804, + "step": 7688 + }, + { + "epoch": 0.4, + "learning_rate": 2.719728459675247e-07, + "loss": 0.172, + "step": 7689 + }, + { + "epoch": 0.4, + "learning_rate": 2.7194129868018173e-07, + "loss": 0.1511, + "step": 7690 + }, + { + "epoch": 0.4, + "learning_rate": 2.719097493367287e-07, + "loss": 0.1116, + "step": 7691 + }, + { + "epoch": 0.4, + "learning_rate": 2.718781979380672e-07, + "loss": 0.1474, + "step": 7692 + }, + { + "epoch": 0.4, + "learning_rate": 2.718466444850992e-07, + "loss": 0.159, + "step": 7693 + }, + { + "epoch": 0.4, + "learning_rate": 2.7181508897872627e-07, + "loss": 0.1557, + "step": 7694 + }, + { + "epoch": 0.4, + "learning_rate": 2.717835314198504e-07, + "loss": 0.1077, + "step": 7695 + }, + { + "epoch": 0.4, + "learning_rate": 2.717519718093735e-07, + "loss": 0.1662, + "step": 7696 + }, + { + "epoch": 0.4, + "learning_rate": 2.717204101481975e-07, + "loss": 0.1329, + "step": 7697 + }, + { + "epoch": 0.4, + "learning_rate": 2.716888464372246e-07, + "loss": 0.1034, + "step": 7698 + }, + { + "epoch": 0.4, + "learning_rate": 2.716572806773567e-07, + "loss": 0.1489, + "step": 7699 + }, + { + "epoch": 0.4, + "learning_rate": 2.716257128694961e-07, + "loss": 0.1756, + "step": 7700 + }, + { + "epoch": 0.4, + "learning_rate": 2.7159414301454494e-07, + "loss": 0.1462, + "step": 7701 + }, + { + "epoch": 0.4, + "learning_rate": 2.715625711134056e-07, + "loss": 0.1253, + "step": 7702 + }, + { + "epoch": 0.4, + "learning_rate": 2.715309971669803e-07, + "loss": 0.1405, + "step": 7703 + }, + { + "epoch": 0.4, + "learning_rate": 2.7149942117617145e-07, + "loss": 0.1343, + "step": 7704 + }, + { + "epoch": 0.4, + "learning_rate": 2.7146784314188156e-07, + "loss": 0.126, + "step": 7705 + }, + { + "epoch": 0.4, + "learning_rate": 2.7143626306501316e-07, + "loss": 0.151, + "step": 7706 + }, + { + "epoch": 0.4, + "learning_rate": 2.714046809464688e-07, + "loss": 0.1055, + "step": 7707 + }, + { + "epoch": 0.4, + "learning_rate": 2.71373096787151e-07, + "loss": 0.1273, + "step": 7708 + }, + { + "epoch": 0.4, + "learning_rate": 2.7134151058796264e-07, + "loss": 0.1447, + "step": 7709 + }, + { + "epoch": 0.4, + "learning_rate": 2.7130992234980634e-07, + "loss": 0.1601, + "step": 7710 + }, + { + "epoch": 0.4, + "learning_rate": 2.7127833207358494e-07, + "loss": 0.181, + "step": 7711 + }, + { + "epoch": 0.4, + "learning_rate": 2.7124673976020124e-07, + "loss": 0.172, + "step": 7712 + }, + { + "epoch": 0.4, + "learning_rate": 2.712151454105582e-07, + "loss": 0.1482, + "step": 7713 + }, + { + "epoch": 0.4, + "learning_rate": 2.7118354902555893e-07, + "loss": 0.1275, + "step": 7714 + }, + { + "epoch": 0.4, + "learning_rate": 2.7115195060610625e-07, + "loss": 0.1612, + "step": 7715 + }, + { + "epoch": 0.4, + "learning_rate": 2.711203501531034e-07, + "loss": 0.1201, + "step": 7716 + }, + { + "epoch": 0.4, + "learning_rate": 2.710887476674534e-07, + "loss": 0.1645, + "step": 7717 + }, + { + "epoch": 0.4, + "learning_rate": 2.710571431500597e-07, + "loss": 0.1132, + "step": 7718 + }, + { + "epoch": 0.4, + "learning_rate": 2.710255366018253e-07, + "loss": 0.1512, + "step": 7719 + }, + { + "epoch": 0.4, + "learning_rate": 2.709939280236537e-07, + "loss": 0.1203, + "step": 7720 + }, + { + "epoch": 0.4, + "learning_rate": 2.7096231741644817e-07, + "loss": 0.1778, + "step": 7721 + }, + { + "epoch": 0.4, + "learning_rate": 2.709307047811123e-07, + "loss": 0.1387, + "step": 7722 + }, + { + "epoch": 0.4, + "learning_rate": 2.7089909011854947e-07, + "loss": 0.1401, + "step": 7723 + }, + { + "epoch": 0.4, + "learning_rate": 2.7086747342966327e-07, + "loss": 0.1354, + "step": 7724 + }, + { + "epoch": 0.4, + "learning_rate": 2.708358547153573e-07, + "loss": 0.1517, + "step": 7725 + }, + { + "epoch": 0.4, + "learning_rate": 2.7080423397653525e-07, + "loss": 0.1064, + "step": 7726 + }, + { + "epoch": 0.4, + "learning_rate": 2.7077261121410095e-07, + "loss": 0.111, + "step": 7727 + }, + { + "epoch": 0.4, + "learning_rate": 2.7074098642895795e-07, + "loss": 0.1576, + "step": 7728 + }, + { + "epoch": 0.4, + "learning_rate": 2.7070935962201036e-07, + "loss": 0.1105, + "step": 7729 + }, + { + "epoch": 0.4, + "learning_rate": 2.7067773079416195e-07, + "loss": 0.1337, + "step": 7730 + }, + { + "epoch": 0.4, + "learning_rate": 2.706460999463167e-07, + "loss": 0.1447, + "step": 7731 + }, + { + "epoch": 0.4, + "learning_rate": 2.7061446707937865e-07, + "loss": 0.1213, + "step": 7732 + }, + { + "epoch": 0.4, + "learning_rate": 2.705828321942519e-07, + "loss": 0.1415, + "step": 7733 + }, + { + "epoch": 0.4, + "learning_rate": 2.705511952918405e-07, + "loss": 0.1197, + "step": 7734 + }, + { + "epoch": 0.4, + "learning_rate": 2.7051955637304876e-07, + "loss": 0.1326, + "step": 7735 + }, + { + "epoch": 0.4, + "learning_rate": 2.7048791543878084e-07, + "loss": 0.1561, + "step": 7736 + }, + { + "epoch": 0.4, + "learning_rate": 2.704562724899411e-07, + "loss": 0.1372, + "step": 7737 + }, + { + "epoch": 0.4, + "learning_rate": 2.704246275274339e-07, + "loss": 0.1693, + "step": 7738 + }, + { + "epoch": 0.4, + "learning_rate": 2.703929805521636e-07, + "loss": 0.1569, + "step": 7739 + }, + { + "epoch": 0.4, + "learning_rate": 2.703613315650348e-07, + "loss": 0.1205, + "step": 7740 + }, + { + "epoch": 0.4, + "learning_rate": 2.70329680566952e-07, + "loss": 0.1299, + "step": 7741 + }, + { + "epoch": 0.4, + "learning_rate": 2.7029802755881974e-07, + "loss": 0.1375, + "step": 7742 + }, + { + "epoch": 0.4, + "learning_rate": 2.702663725415427e-07, + "loss": 0.1315, + "step": 7743 + }, + { + "epoch": 0.4, + "learning_rate": 2.7023471551602566e-07, + "loss": 0.1555, + "step": 7744 + }, + { + "epoch": 0.4, + "learning_rate": 2.702030564831733e-07, + "loss": 0.1331, + "step": 7745 + }, + { + "epoch": 0.4, + "learning_rate": 2.7017139544389044e-07, + "loss": 0.1798, + "step": 7746 + }, + { + "epoch": 0.4, + "learning_rate": 2.701397323990821e-07, + "loss": 0.1253, + "step": 7747 + }, + { + "epoch": 0.4, + "learning_rate": 2.7010806734965303e-07, + "loss": 0.14, + "step": 7748 + }, + { + "epoch": 0.4, + "learning_rate": 2.700764002965085e-07, + "loss": 0.1196, + "step": 7749 + }, + { + "epoch": 0.4, + "learning_rate": 2.7004473124055324e-07, + "loss": 0.1777, + "step": 7750 + }, + { + "epoch": 0.4, + "learning_rate": 2.700130601826926e-07, + "loss": 0.1634, + "step": 7751 + }, + { + "epoch": 0.4, + "learning_rate": 2.6998138712383165e-07, + "loss": 0.1326, + "step": 7752 + }, + { + "epoch": 0.4, + "learning_rate": 2.6994971206487565e-07, + "loss": 0.1602, + "step": 7753 + }, + { + "epoch": 0.4, + "learning_rate": 2.699180350067298e-07, + "loss": 0.1299, + "step": 7754 + }, + { + "epoch": 0.4, + "learning_rate": 2.698863559502996e-07, + "loss": 0.1021, + "step": 7755 + }, + { + "epoch": 0.4, + "learning_rate": 2.6985467489649037e-07, + "loss": 0.1457, + "step": 7756 + }, + { + "epoch": 0.4, + "learning_rate": 2.698229918462076e-07, + "loss": 0.1264, + "step": 7757 + }, + { + "epoch": 0.4, + "learning_rate": 2.697913068003566e-07, + "loss": 0.1423, + "step": 7758 + }, + { + "epoch": 0.4, + "learning_rate": 2.6975961975984324e-07, + "loss": 0.1165, + "step": 7759 + }, + { + "epoch": 0.41, + "learning_rate": 2.69727930725573e-07, + "loss": 0.1653, + "step": 7760 + }, + { + "epoch": 0.41, + "learning_rate": 2.6969623969845156e-07, + "loss": 0.153, + "step": 7761 + }, + { + "epoch": 0.41, + "learning_rate": 2.696645466793847e-07, + "loss": 0.1467, + "step": 7762 + }, + { + "epoch": 0.41, + "learning_rate": 2.6963285166927824e-07, + "loss": 0.1593, + "step": 7763 + }, + { + "epoch": 0.41, + "learning_rate": 2.6960115466903796e-07, + "loss": 0.1642, + "step": 7764 + }, + { + "epoch": 0.41, + "learning_rate": 2.695694556795698e-07, + "loss": 0.1456, + "step": 7765 + }, + { + "epoch": 0.41, + "learning_rate": 2.695377547017798e-07, + "loss": 0.1043, + "step": 7766 + }, + { + "epoch": 0.41, + "learning_rate": 2.6950605173657387e-07, + "loss": 0.1547, + "step": 7767 + }, + { + "epoch": 0.41, + "learning_rate": 2.6947434678485817e-07, + "loss": 0.1549, + "step": 7768 + }, + { + "epoch": 0.41, + "learning_rate": 2.694426398475388e-07, + "loss": 0.1363, + "step": 7769 + }, + { + "epoch": 0.41, + "learning_rate": 2.69410930925522e-07, + "loss": 0.1358, + "step": 7770 + }, + { + "epoch": 0.41, + "learning_rate": 2.69379220019714e-07, + "loss": 0.1498, + "step": 7771 + }, + { + "epoch": 0.41, + "learning_rate": 2.6934750713102104e-07, + "loss": 0.1219, + "step": 7772 + }, + { + "epoch": 0.41, + "learning_rate": 2.693157922603496e-07, + "loss": 0.2034, + "step": 7773 + }, + { + "epoch": 0.41, + "learning_rate": 2.6928407540860615e-07, + "loss": 0.1057, + "step": 7774 + }, + { + "epoch": 0.41, + "learning_rate": 2.6925235657669696e-07, + "loss": 0.1326, + "step": 7775 + }, + { + "epoch": 0.41, + "learning_rate": 2.6922063576552874e-07, + "loss": 0.1512, + "step": 7776 + }, + { + "epoch": 0.41, + "learning_rate": 2.6918891297600797e-07, + "loss": 0.1441, + "step": 7777 + }, + { + "epoch": 0.41, + "learning_rate": 2.691571882090414e-07, + "loss": 0.1328, + "step": 7778 + }, + { + "epoch": 0.41, + "learning_rate": 2.691254614655357e-07, + "loss": 0.1428, + "step": 7779 + }, + { + "epoch": 0.41, + "learning_rate": 2.690937327463976e-07, + "loss": 0.1675, + "step": 7780 + }, + { + "epoch": 0.41, + "learning_rate": 2.6906200205253395e-07, + "loss": 0.1423, + "step": 7781 + }, + { + "epoch": 0.41, + "learning_rate": 2.690302693848517e-07, + "loss": 0.1161, + "step": 7782 + }, + { + "epoch": 0.41, + "learning_rate": 2.689985347442576e-07, + "loss": 0.109, + "step": 7783 + }, + { + "epoch": 0.41, + "learning_rate": 2.6896679813165876e-07, + "loss": 0.104, + "step": 7784 + }, + { + "epoch": 0.41, + "learning_rate": 2.6893505954796223e-07, + "loss": 0.1584, + "step": 7785 + }, + { + "epoch": 0.41, + "learning_rate": 2.6890331899407506e-07, + "loss": 0.1346, + "step": 7786 + }, + { + "epoch": 0.41, + "learning_rate": 2.688715764709045e-07, + "loss": 0.1139, + "step": 7787 + }, + { + "epoch": 0.41, + "learning_rate": 2.6883983197935764e-07, + "loss": 0.1618, + "step": 7788 + }, + { + "epoch": 0.41, + "learning_rate": 2.688080855203418e-07, + "loss": 0.1226, + "step": 7789 + }, + { + "epoch": 0.41, + "learning_rate": 2.687763370947643e-07, + "loss": 0.1554, + "step": 7790 + }, + { + "epoch": 0.41, + "learning_rate": 2.687445867035326e-07, + "loss": 0.1614, + "step": 7791 + }, + { + "epoch": 0.41, + "learning_rate": 2.68712834347554e-07, + "loss": 0.1806, + "step": 7792 + }, + { + "epoch": 0.41, + "learning_rate": 2.686810800277361e-07, + "loss": 0.1403, + "step": 7793 + }, + { + "epoch": 0.41, + "learning_rate": 2.6864932374498643e-07, + "loss": 0.1672, + "step": 7794 + }, + { + "epoch": 0.41, + "learning_rate": 2.686175655002126e-07, + "loss": 0.1855, + "step": 7795 + }, + { + "epoch": 0.41, + "learning_rate": 2.6858580529432227e-07, + "loss": 0.1336, + "step": 7796 + }, + { + "epoch": 0.41, + "learning_rate": 2.685540431282231e-07, + "loss": 0.1889, + "step": 7797 + }, + { + "epoch": 0.41, + "learning_rate": 2.6852227900282293e-07, + "loss": 0.126, + "step": 7798 + }, + { + "epoch": 0.41, + "learning_rate": 2.684905129190296e-07, + "loss": 0.1456, + "step": 7799 + }, + { + "epoch": 0.41, + "learning_rate": 2.684587448777509e-07, + "loss": 0.1272, + "step": 7800 + }, + { + "epoch": 0.41, + "learning_rate": 2.684269748798949e-07, + "loss": 0.1426, + "step": 7801 + }, + { + "epoch": 0.41, + "learning_rate": 2.6839520292636955e-07, + "loss": 0.1751, + "step": 7802 + }, + { + "epoch": 0.41, + "learning_rate": 2.6836342901808294e-07, + "loss": 0.1701, + "step": 7803 + }, + { + "epoch": 0.41, + "learning_rate": 2.6833165315594304e-07, + "loss": 0.1255, + "step": 7804 + }, + { + "epoch": 0.41, + "learning_rate": 2.6829987534085814e-07, + "loss": 0.119, + "step": 7805 + }, + { + "epoch": 0.41, + "learning_rate": 2.682680955737364e-07, + "loss": 0.1409, + "step": 7806 + }, + { + "epoch": 0.41, + "learning_rate": 2.682363138554862e-07, + "loss": 0.106, + "step": 7807 + }, + { + "epoch": 0.41, + "learning_rate": 2.682045301870158e-07, + "loss": 0.1325, + "step": 7808 + }, + { + "epoch": 0.41, + "learning_rate": 2.6817274456923356e-07, + "loss": 0.1171, + "step": 7809 + }, + { + "epoch": 0.41, + "learning_rate": 2.6814095700304804e-07, + "loss": 0.1629, + "step": 7810 + }, + { + "epoch": 0.41, + "learning_rate": 2.681091674893676e-07, + "loss": 0.1188, + "step": 7811 + }, + { + "epoch": 0.41, + "learning_rate": 2.680773760291009e-07, + "loss": 0.1339, + "step": 7812 + }, + { + "epoch": 0.41, + "learning_rate": 2.680455826231565e-07, + "loss": 0.1191, + "step": 7813 + }, + { + "epoch": 0.41, + "learning_rate": 2.6801378727244307e-07, + "loss": 0.1465, + "step": 7814 + }, + { + "epoch": 0.41, + "learning_rate": 2.679819899778693e-07, + "loss": 0.1285, + "step": 7815 + }, + { + "epoch": 0.41, + "learning_rate": 2.679501907403441e-07, + "loss": 0.1589, + "step": 7816 + }, + { + "epoch": 0.41, + "learning_rate": 2.6791838956077616e-07, + "loss": 0.1632, + "step": 7817 + }, + { + "epoch": 0.41, + "learning_rate": 2.6788658644007443e-07, + "loss": 0.1185, + "step": 7818 + }, + { + "epoch": 0.41, + "learning_rate": 2.6785478137914787e-07, + "loss": 0.1577, + "step": 7819 + }, + { + "epoch": 0.41, + "learning_rate": 2.6782297437890543e-07, + "loss": 0.1107, + "step": 7820 + }, + { + "epoch": 0.41, + "learning_rate": 2.6779116544025626e-07, + "loss": 0.1417, + "step": 7821 + }, + { + "epoch": 0.41, + "learning_rate": 2.677593545641094e-07, + "loss": 0.1168, + "step": 7822 + }, + { + "epoch": 0.41, + "learning_rate": 2.6772754175137397e-07, + "loss": 0.1258, + "step": 7823 + }, + { + "epoch": 0.41, + "learning_rate": 2.6769572700295933e-07, + "loss": 0.1302, + "step": 7824 + }, + { + "epoch": 0.41, + "learning_rate": 2.676639103197746e-07, + "loss": 0.1536, + "step": 7825 + }, + { + "epoch": 0.41, + "learning_rate": 2.676320917027292e-07, + "loss": 0.1198, + "step": 7826 + }, + { + "epoch": 0.41, + "learning_rate": 2.676002711527326e-07, + "loss": 0.1553, + "step": 7827 + }, + { + "epoch": 0.41, + "learning_rate": 2.6756844867069407e-07, + "loss": 0.1731, + "step": 7828 + }, + { + "epoch": 0.41, + "learning_rate": 2.675366242575232e-07, + "loss": 0.1497, + "step": 7829 + }, + { + "epoch": 0.41, + "learning_rate": 2.675047979141296e-07, + "loss": 0.124, + "step": 7830 + }, + { + "epoch": 0.41, + "learning_rate": 2.674729696414227e-07, + "loss": 0.1517, + "step": 7831 + }, + { + "epoch": 0.41, + "learning_rate": 2.674411394403124e-07, + "loss": 0.1665, + "step": 7832 + }, + { + "epoch": 0.41, + "learning_rate": 2.674093073117082e-07, + "loss": 0.1642, + "step": 7833 + }, + { + "epoch": 0.41, + "learning_rate": 2.673774732565201e-07, + "loss": 0.155, + "step": 7834 + }, + { + "epoch": 0.41, + "learning_rate": 2.673456372756577e-07, + "loss": 0.1195, + "step": 7835 + }, + { + "epoch": 0.41, + "learning_rate": 2.67313799370031e-07, + "loss": 0.1087, + "step": 7836 + }, + { + "epoch": 0.41, + "learning_rate": 2.6728195954054996e-07, + "loss": 0.1913, + "step": 7837 + }, + { + "epoch": 0.41, + "learning_rate": 2.672501177881245e-07, + "loss": 0.1568, + "step": 7838 + }, + { + "epoch": 0.41, + "learning_rate": 2.672182741136648e-07, + "loss": 0.1365, + "step": 7839 + }, + { + "epoch": 0.41, + "learning_rate": 2.671864285180808e-07, + "loss": 0.1476, + "step": 7840 + }, + { + "epoch": 0.41, + "learning_rate": 2.6715458100228275e-07, + "loss": 0.1189, + "step": 7841 + }, + { + "epoch": 0.41, + "learning_rate": 2.671227315671808e-07, + "loss": 0.2024, + "step": 7842 + }, + { + "epoch": 0.41, + "learning_rate": 2.670908802136854e-07, + "loss": 0.1137, + "step": 7843 + }, + { + "epoch": 0.41, + "learning_rate": 2.670590269427066e-07, + "loss": 0.1647, + "step": 7844 + }, + { + "epoch": 0.41, + "learning_rate": 2.67027171755155e-07, + "loss": 0.1627, + "step": 7845 + }, + { + "epoch": 0.41, + "learning_rate": 2.669953146519409e-07, + "loss": 0.1499, + "step": 7846 + }, + { + "epoch": 0.41, + "learning_rate": 2.66963455633975e-07, + "loss": 0.1211, + "step": 7847 + }, + { + "epoch": 0.41, + "learning_rate": 2.6693159470216753e-07, + "loss": 0.1171, + "step": 7848 + }, + { + "epoch": 0.41, + "learning_rate": 2.668997318574293e-07, + "loss": 0.1073, + "step": 7849 + }, + { + "epoch": 0.41, + "learning_rate": 2.668678671006709e-07, + "loss": 0.1415, + "step": 7850 + }, + { + "epoch": 0.41, + "learning_rate": 2.6683600043280305e-07, + "loss": 0.1294, + "step": 7851 + }, + { + "epoch": 0.41, + "learning_rate": 2.6680413185473654e-07, + "loss": 0.1587, + "step": 7852 + }, + { + "epoch": 0.41, + "learning_rate": 2.667722613673821e-07, + "loss": 0.1264, + "step": 7853 + }, + { + "epoch": 0.41, + "learning_rate": 2.6674038897165065e-07, + "loss": 0.116, + "step": 7854 + }, + { + "epoch": 0.41, + "learning_rate": 2.6670851466845314e-07, + "loss": 0.1422, + "step": 7855 + }, + { + "epoch": 0.41, + "learning_rate": 2.666766384587005e-07, + "loss": 0.1532, + "step": 7856 + }, + { + "epoch": 0.41, + "learning_rate": 2.6664476034330385e-07, + "loss": 0.1014, + "step": 7857 + }, + { + "epoch": 0.41, + "learning_rate": 2.666128803231742e-07, + "loss": 0.1613, + "step": 7858 + }, + { + "epoch": 0.41, + "learning_rate": 2.665809983992227e-07, + "loss": 0.1469, + "step": 7859 + }, + { + "epoch": 0.41, + "learning_rate": 2.665491145723606e-07, + "loss": 0.1373, + "step": 7860 + }, + { + "epoch": 0.41, + "learning_rate": 2.6651722884349906e-07, + "loss": 0.1151, + "step": 7861 + }, + { + "epoch": 0.41, + "learning_rate": 2.6648534121354943e-07, + "loss": 0.1242, + "step": 7862 + }, + { + "epoch": 0.41, + "learning_rate": 2.6645345168342317e-07, + "loss": 0.1566, + "step": 7863 + }, + { + "epoch": 0.41, + "learning_rate": 2.6642156025403155e-07, + "loss": 0.1157, + "step": 7864 + }, + { + "epoch": 0.41, + "learning_rate": 2.6638966692628606e-07, + "loss": 0.1404, + "step": 7865 + }, + { + "epoch": 0.41, + "learning_rate": 2.6635777170109835e-07, + "loss": 0.1544, + "step": 7866 + }, + { + "epoch": 0.41, + "learning_rate": 2.663258745793798e-07, + "loss": 0.1461, + "step": 7867 + }, + { + "epoch": 0.41, + "learning_rate": 2.662939755620422e-07, + "loss": 0.1639, + "step": 7868 + }, + { + "epoch": 0.41, + "learning_rate": 2.662620746499972e-07, + "loss": 0.1236, + "step": 7869 + }, + { + "epoch": 0.41, + "learning_rate": 2.662301718441566e-07, + "loss": 0.1387, + "step": 7870 + }, + { + "epoch": 0.41, + "learning_rate": 2.66198267145432e-07, + "loss": 0.1148, + "step": 7871 + }, + { + "epoch": 0.41, + "learning_rate": 2.661663605547354e-07, + "loss": 0.1144, + "step": 7872 + }, + { + "epoch": 0.41, + "learning_rate": 2.6613445207297874e-07, + "loss": 0.1527, + "step": 7873 + }, + { + "epoch": 0.41, + "learning_rate": 2.661025417010738e-07, + "loss": 0.1642, + "step": 7874 + }, + { + "epoch": 0.41, + "learning_rate": 2.660706294399328e-07, + "loss": 0.1261, + "step": 7875 + }, + { + "epoch": 0.41, + "learning_rate": 2.660387152904676e-07, + "loss": 0.1342, + "step": 7876 + }, + { + "epoch": 0.41, + "learning_rate": 2.6600679925359057e-07, + "loss": 0.1075, + "step": 7877 + }, + { + "epoch": 0.41, + "learning_rate": 2.6597488133021364e-07, + "loss": 0.137, + "step": 7878 + }, + { + "epoch": 0.41, + "learning_rate": 2.6594296152124915e-07, + "loss": 0.1388, + "step": 7879 + }, + { + "epoch": 0.41, + "learning_rate": 2.659110398276093e-07, + "loss": 0.1272, + "step": 7880 + }, + { + "epoch": 0.41, + "learning_rate": 2.658791162502066e-07, + "loss": 0.1263, + "step": 7881 + }, + { + "epoch": 0.41, + "learning_rate": 2.658471907899533e-07, + "loss": 0.1446, + "step": 7882 + }, + { + "epoch": 0.41, + "learning_rate": 2.6581526344776184e-07, + "loss": 0.1769, + "step": 7883 + }, + { + "epoch": 0.41, + "learning_rate": 2.657833342245447e-07, + "loss": 0.1653, + "step": 7884 + }, + { + "epoch": 0.41, + "learning_rate": 2.657514031212145e-07, + "loss": 0.1529, + "step": 7885 + }, + { + "epoch": 0.41, + "learning_rate": 2.6571947013868387e-07, + "loss": 0.1798, + "step": 7886 + }, + { + "epoch": 0.41, + "learning_rate": 2.6568753527786534e-07, + "loss": 0.1303, + "step": 7887 + }, + { + "epoch": 0.41, + "learning_rate": 2.656555985396717e-07, + "loss": 0.119, + "step": 7888 + }, + { + "epoch": 0.41, + "learning_rate": 2.656236599250158e-07, + "loss": 0.1329, + "step": 7889 + }, + { + "epoch": 0.41, + "learning_rate": 2.655917194348103e-07, + "loss": 0.129, + "step": 7890 + }, + { + "epoch": 0.41, + "learning_rate": 2.655597770699681e-07, + "loss": 0.1531, + "step": 7891 + }, + { + "epoch": 0.41, + "learning_rate": 2.655278328314022e-07, + "loss": 0.1426, + "step": 7892 + }, + { + "epoch": 0.41, + "learning_rate": 2.6549588672002554e-07, + "loss": 0.1394, + "step": 7893 + }, + { + "epoch": 0.41, + "learning_rate": 2.6546393873675115e-07, + "loss": 0.1637, + "step": 7894 + }, + { + "epoch": 0.41, + "learning_rate": 2.6543198888249216e-07, + "loss": 0.1386, + "step": 7895 + }, + { + "epoch": 0.41, + "learning_rate": 2.654000371581616e-07, + "loss": 0.1482, + "step": 7896 + }, + { + "epoch": 0.41, + "learning_rate": 2.653680835646728e-07, + "loss": 0.1054, + "step": 7897 + }, + { + "epoch": 0.41, + "learning_rate": 2.6533612810293886e-07, + "loss": 0.117, + "step": 7898 + }, + { + "epoch": 0.41, + "learning_rate": 2.6530417077387324e-07, + "loss": 0.1491, + "step": 7899 + }, + { + "epoch": 0.41, + "learning_rate": 2.6527221157838916e-07, + "loss": 0.1339, + "step": 7900 + }, + { + "epoch": 0.41, + "learning_rate": 2.652402505174001e-07, + "loss": 0.1482, + "step": 7901 + }, + { + "epoch": 0.41, + "learning_rate": 2.6520828759181944e-07, + "loss": 0.1477, + "step": 7902 + }, + { + "epoch": 0.41, + "learning_rate": 2.651763228025608e-07, + "loss": 0.1385, + "step": 7903 + }, + { + "epoch": 0.41, + "learning_rate": 2.6514435615053774e-07, + "loss": 0.1336, + "step": 7904 + }, + { + "epoch": 0.41, + "learning_rate": 2.651123876366637e-07, + "loss": 0.1335, + "step": 7905 + }, + { + "epoch": 0.41, + "learning_rate": 2.650804172618526e-07, + "loss": 0.1348, + "step": 7906 + }, + { + "epoch": 0.41, + "learning_rate": 2.65048445027018e-07, + "loss": 0.1707, + "step": 7907 + }, + { + "epoch": 0.41, + "learning_rate": 2.650164709330738e-07, + "loss": 0.1113, + "step": 7908 + }, + { + "epoch": 0.41, + "learning_rate": 2.6498449498093364e-07, + "loss": 0.1177, + "step": 7909 + }, + { + "epoch": 0.41, + "learning_rate": 2.649525171715116e-07, + "loss": 0.1315, + "step": 7910 + }, + { + "epoch": 0.41, + "learning_rate": 2.6492053750572156e-07, + "loss": 0.1757, + "step": 7911 + }, + { + "epoch": 0.41, + "learning_rate": 2.648885559844774e-07, + "loss": 0.1306, + "step": 7912 + }, + { + "epoch": 0.41, + "learning_rate": 2.6485657260869334e-07, + "loss": 0.2079, + "step": 7913 + }, + { + "epoch": 0.41, + "learning_rate": 2.648245873792834e-07, + "loss": 0.1495, + "step": 7914 + }, + { + "epoch": 0.41, + "learning_rate": 2.6479260029716175e-07, + "loss": 0.1587, + "step": 7915 + }, + { + "epoch": 0.41, + "learning_rate": 2.6476061136324255e-07, + "loss": 0.1408, + "step": 7916 + }, + { + "epoch": 0.41, + "learning_rate": 2.6472862057844006e-07, + "loss": 0.1249, + "step": 7917 + }, + { + "epoch": 0.41, + "learning_rate": 2.646966279436686e-07, + "loss": 0.1207, + "step": 7918 + }, + { + "epoch": 0.41, + "learning_rate": 2.6466463345984255e-07, + "loss": 0.1327, + "step": 7919 + }, + { + "epoch": 0.41, + "learning_rate": 2.6463263712787634e-07, + "loss": 0.1568, + "step": 7920 + }, + { + "epoch": 0.41, + "learning_rate": 2.646006389486843e-07, + "loss": 0.1492, + "step": 7921 + }, + { + "epoch": 0.41, + "learning_rate": 2.6456863892318113e-07, + "loss": 0.1459, + "step": 7922 + }, + { + "epoch": 0.41, + "learning_rate": 2.6453663705228135e-07, + "loss": 0.11, + "step": 7923 + }, + { + "epoch": 0.41, + "learning_rate": 2.6450463333689967e-07, + "loss": 0.1313, + "step": 7924 + }, + { + "epoch": 0.41, + "learning_rate": 2.644726277779505e-07, + "loss": 0.1232, + "step": 7925 + }, + { + "epoch": 0.41, + "learning_rate": 2.6444062037634884e-07, + "loss": 0.1164, + "step": 7926 + }, + { + "epoch": 0.41, + "learning_rate": 2.644086111330093e-07, + "loss": 0.1292, + "step": 7927 + }, + { + "epoch": 0.41, + "learning_rate": 2.643766000488468e-07, + "loss": 0.1484, + "step": 7928 + }, + { + "epoch": 0.41, + "learning_rate": 2.643445871247763e-07, + "loss": 0.1396, + "step": 7929 + }, + { + "epoch": 0.41, + "learning_rate": 2.643125723617126e-07, + "loss": 0.1674, + "step": 7930 + }, + { + "epoch": 0.41, + "learning_rate": 2.642805557605707e-07, + "loss": 0.1195, + "step": 7931 + }, + { + "epoch": 0.41, + "learning_rate": 2.642485373222658e-07, + "loss": 0.1019, + "step": 7932 + }, + { + "epoch": 0.41, + "learning_rate": 2.642165170477129e-07, + "loss": 0.1587, + "step": 7933 + }, + { + "epoch": 0.41, + "learning_rate": 2.6418449493782707e-07, + "loss": 0.1213, + "step": 7934 + }, + { + "epoch": 0.41, + "learning_rate": 2.6415247099352365e-07, + "loss": 0.1685, + "step": 7935 + }, + { + "epoch": 0.41, + "learning_rate": 2.641204452157178e-07, + "loss": 0.1294, + "step": 7936 + }, + { + "epoch": 0.41, + "learning_rate": 2.64088417605325e-07, + "loss": 0.1469, + "step": 7937 + }, + { + "epoch": 0.41, + "learning_rate": 2.6405638816326033e-07, + "loss": 0.1339, + "step": 7938 + }, + { + "epoch": 0.41, + "learning_rate": 2.6402435689043945e-07, + "loss": 0.1418, + "step": 7939 + }, + { + "epoch": 0.41, + "learning_rate": 2.639923237877776e-07, + "loss": 0.1534, + "step": 7940 + }, + { + "epoch": 0.41, + "learning_rate": 2.639602888561906e-07, + "loss": 0.1559, + "step": 7941 + }, + { + "epoch": 0.41, + "learning_rate": 2.639282520965938e-07, + "loss": 0.1714, + "step": 7942 + }, + { + "epoch": 0.41, + "learning_rate": 2.638962135099029e-07, + "loss": 0.1456, + "step": 7943 + }, + { + "epoch": 0.41, + "learning_rate": 2.638641730970335e-07, + "loss": 0.1354, + "step": 7944 + }, + { + "epoch": 0.41, + "learning_rate": 2.638321308589013e-07, + "loss": 0.1323, + "step": 7945 + }, + { + "epoch": 0.41, + "learning_rate": 2.6380008679642237e-07, + "loss": 0.1187, + "step": 7946 + }, + { + "epoch": 0.41, + "learning_rate": 2.6376804091051216e-07, + "loss": 0.1217, + "step": 7947 + }, + { + "epoch": 0.41, + "learning_rate": 2.637359932020868e-07, + "loss": 0.1343, + "step": 7948 + }, + { + "epoch": 0.41, + "learning_rate": 2.637039436720621e-07, + "loss": 0.1437, + "step": 7949 + }, + { + "epoch": 0.41, + "learning_rate": 2.6367189232135415e-07, + "loss": 0.12, + "step": 7950 + }, + { + "epoch": 0.42, + "learning_rate": 2.636398391508789e-07, + "loss": 0.0973, + "step": 7951 + }, + { + "epoch": 0.42, + "learning_rate": 2.636077841615525e-07, + "loss": 0.1339, + "step": 7952 + }, + { + "epoch": 0.42, + "learning_rate": 2.6357572735429106e-07, + "loss": 0.1695, + "step": 7953 + }, + { + "epoch": 0.42, + "learning_rate": 2.635436687300108e-07, + "loss": 0.1681, + "step": 7954 + }, + { + "epoch": 0.42, + "learning_rate": 2.63511608289628e-07, + "loss": 0.1648, + "step": 7955 + }, + { + "epoch": 0.42, + "learning_rate": 2.6347954603405885e-07, + "loss": 0.168, + "step": 7956 + }, + { + "epoch": 0.42, + "learning_rate": 2.6344748196421976e-07, + "loss": 0.1504, + "step": 7957 + }, + { + "epoch": 0.42, + "learning_rate": 2.634154160810272e-07, + "loss": 0.1476, + "step": 7958 + }, + { + "epoch": 0.42, + "learning_rate": 2.633833483853975e-07, + "loss": 0.1077, + "step": 7959 + }, + { + "epoch": 0.42, + "learning_rate": 2.633512788782473e-07, + "loss": 0.1223, + "step": 7960 + }, + { + "epoch": 0.42, + "learning_rate": 2.63319207560493e-07, + "loss": 0.1479, + "step": 7961 + }, + { + "epoch": 0.42, + "learning_rate": 2.632871344330514e-07, + "loss": 0.1381, + "step": 7962 + }, + { + "epoch": 0.42, + "learning_rate": 2.632550594968391e-07, + "loss": 0.1253, + "step": 7963 + }, + { + "epoch": 0.42, + "learning_rate": 2.632229827527727e-07, + "loss": 0.1359, + "step": 7964 + }, + { + "epoch": 0.42, + "learning_rate": 2.6319090420176903e-07, + "loss": 0.1137, + "step": 7965 + }, + { + "epoch": 0.42, + "learning_rate": 2.63158823844745e-07, + "loss": 0.1498, + "step": 7966 + }, + { + "epoch": 0.42, + "learning_rate": 2.631267416826174e-07, + "loss": 0.1342, + "step": 7967 + }, + { + "epoch": 0.42, + "learning_rate": 2.6309465771630306e-07, + "loss": 0.1529, + "step": 7968 + }, + { + "epoch": 0.42, + "learning_rate": 2.6306257194671913e-07, + "loss": 0.1484, + "step": 7969 + }, + { + "epoch": 0.42, + "learning_rate": 2.6303048437478244e-07, + "loss": 0.1295, + "step": 7970 + }, + { + "epoch": 0.42, + "learning_rate": 2.629983950014103e-07, + "loss": 0.1469, + "step": 7971 + }, + { + "epoch": 0.42, + "learning_rate": 2.629663038275197e-07, + "loss": 0.14, + "step": 7972 + }, + { + "epoch": 0.42, + "learning_rate": 2.6293421085402776e-07, + "loss": 0.1645, + "step": 7973 + }, + { + "epoch": 0.42, + "learning_rate": 2.629021160818518e-07, + "loss": 0.1311, + "step": 7974 + }, + { + "epoch": 0.42, + "learning_rate": 2.628700195119091e-07, + "loss": 0.1162, + "step": 7975 + }, + { + "epoch": 0.42, + "learning_rate": 2.62837921145117e-07, + "loss": 0.1283, + "step": 7976 + }, + { + "epoch": 0.42, + "learning_rate": 2.628058209823928e-07, + "loss": 0.1107, + "step": 7977 + }, + { + "epoch": 0.42, + "learning_rate": 2.62773719024654e-07, + "loss": 0.1227, + "step": 7978 + }, + { + "epoch": 0.42, + "learning_rate": 2.627416152728181e-07, + "loss": 0.1492, + "step": 7979 + }, + { + "epoch": 0.42, + "learning_rate": 2.627095097278026e-07, + "loss": 0.1099, + "step": 7980 + }, + { + "epoch": 0.42, + "learning_rate": 2.6267740239052514e-07, + "loss": 0.121, + "step": 7981 + }, + { + "epoch": 0.42, + "learning_rate": 2.626452932619033e-07, + "loss": 0.118, + "step": 7982 + }, + { + "epoch": 0.42, + "learning_rate": 2.6261318234285474e-07, + "loss": 0.1093, + "step": 7983 + }, + { + "epoch": 0.42, + "learning_rate": 2.6258106963429737e-07, + "loss": 0.1529, + "step": 7984 + }, + { + "epoch": 0.42, + "learning_rate": 2.6254895513714877e-07, + "loss": 0.1037, + "step": 7985 + }, + { + "epoch": 0.42, + "learning_rate": 2.625168388523269e-07, + "loss": 0.1195, + "step": 7986 + }, + { + "epoch": 0.42, + "learning_rate": 2.624847207807497e-07, + "loss": 0.1197, + "step": 7987 + }, + { + "epoch": 0.42, + "learning_rate": 2.624526009233349e-07, + "loss": 0.1439, + "step": 7988 + }, + { + "epoch": 0.42, + "learning_rate": 2.624204792810009e-07, + "loss": 0.148, + "step": 7989 + }, + { + "epoch": 0.42, + "learning_rate": 2.623883558546653e-07, + "loss": 0.1144, + "step": 7990 + }, + { + "epoch": 0.42, + "learning_rate": 2.623562306452466e-07, + "loss": 0.1393, + "step": 7991 + }, + { + "epoch": 0.42, + "learning_rate": 2.6232410365366257e-07, + "loss": 0.1437, + "step": 7992 + }, + { + "epoch": 0.42, + "learning_rate": 2.622919748808317e-07, + "loss": 0.1376, + "step": 7993 + }, + { + "epoch": 0.42, + "learning_rate": 2.622598443276721e-07, + "loss": 0.1367, + "step": 7994 + }, + { + "epoch": 0.42, + "learning_rate": 2.6222771199510214e-07, + "loss": 0.1302, + "step": 7995 + }, + { + "epoch": 0.42, + "learning_rate": 2.6219557788404014e-07, + "loss": 0.143, + "step": 7996 + }, + { + "epoch": 0.42, + "learning_rate": 2.621634419954045e-07, + "loss": 0.1507, + "step": 7997 + }, + { + "epoch": 0.42, + "learning_rate": 2.621313043301137e-07, + "loss": 0.1715, + "step": 7998 + }, + { + "epoch": 0.42, + "learning_rate": 2.620991648890862e-07, + "loss": 0.1481, + "step": 7999 + }, + { + "epoch": 0.42, + "learning_rate": 2.6206702367324065e-07, + "loss": 0.1025, + "step": 8000 + }, + { + "epoch": 0.42, + "learning_rate": 2.6203488068349565e-07, + "loss": 0.1636, + "step": 8001 + }, + { + "epoch": 0.42, + "learning_rate": 2.6200273592076977e-07, + "loss": 0.1527, + "step": 8002 + }, + { + "epoch": 0.42, + "learning_rate": 2.6197058938598167e-07, + "loss": 0.1555, + "step": 8003 + }, + { + "epoch": 0.42, + "learning_rate": 2.619384410800503e-07, + "loss": 0.1051, + "step": 8004 + }, + { + "epoch": 0.42, + "learning_rate": 2.619062910038943e-07, + "loss": 0.1412, + "step": 8005 + }, + { + "epoch": 0.42, + "learning_rate": 2.618741391584327e-07, + "loss": 0.1341, + "step": 8006 + }, + { + "epoch": 0.42, + "learning_rate": 2.618419855445843e-07, + "loss": 0.1578, + "step": 8007 + }, + { + "epoch": 0.42, + "learning_rate": 2.61809830163268e-07, + "loss": 0.1533, + "step": 8008 + }, + { + "epoch": 0.42, + "learning_rate": 2.617776730154029e-07, + "loss": 0.1102, + "step": 8009 + }, + { + "epoch": 0.42, + "learning_rate": 2.617455141019081e-07, + "loss": 0.1323, + "step": 8010 + }, + { + "epoch": 0.42, + "learning_rate": 2.6171335342370273e-07, + "loss": 0.1248, + "step": 8011 + }, + { + "epoch": 0.42, + "learning_rate": 2.616811909817058e-07, + "loss": 0.1081, + "step": 8012 + }, + { + "epoch": 0.42, + "learning_rate": 2.6164902677683666e-07, + "loss": 0.1247, + "step": 8013 + }, + { + "epoch": 0.42, + "learning_rate": 2.616168608100145e-07, + "loss": 0.1411, + "step": 8014 + }, + { + "epoch": 0.42, + "learning_rate": 2.615846930821587e-07, + "loss": 0.1108, + "step": 8015 + }, + { + "epoch": 0.42, + "learning_rate": 2.6155252359418864e-07, + "loss": 0.1794, + "step": 8016 + }, + { + "epoch": 0.42, + "learning_rate": 2.615203523470236e-07, + "loss": 0.1606, + "step": 8017 + }, + { + "epoch": 0.42, + "learning_rate": 2.614881793415832e-07, + "loss": 0.1261, + "step": 8018 + }, + { + "epoch": 0.42, + "learning_rate": 2.6145600457878695e-07, + "loss": 0.1355, + "step": 8019 + }, + { + "epoch": 0.42, + "learning_rate": 2.614238280595543e-07, + "loss": 0.1181, + "step": 8020 + }, + { + "epoch": 0.42, + "learning_rate": 2.613916497848049e-07, + "loss": 0.1221, + "step": 8021 + }, + { + "epoch": 0.42, + "learning_rate": 2.613594697554585e-07, + "loss": 0.1946, + "step": 8022 + }, + { + "epoch": 0.42, + "learning_rate": 2.613272879724348e-07, + "loss": 0.1676, + "step": 8023 + }, + { + "epoch": 0.42, + "learning_rate": 2.6129510443665344e-07, + "loss": 0.1361, + "step": 8024 + }, + { + "epoch": 0.42, + "learning_rate": 2.612629191490344e-07, + "loss": 0.119, + "step": 8025 + }, + { + "epoch": 0.42, + "learning_rate": 2.6123073211049743e-07, + "loss": 0.1454, + "step": 8026 + }, + { + "epoch": 0.42, + "learning_rate": 2.611985433219626e-07, + "loss": 0.127, + "step": 8027 + }, + { + "epoch": 0.42, + "learning_rate": 2.611663527843497e-07, + "loss": 0.1572, + "step": 8028 + }, + { + "epoch": 0.42, + "learning_rate": 2.6113416049857885e-07, + "loss": 0.1787, + "step": 8029 + }, + { + "epoch": 0.42, + "learning_rate": 2.6110196646557003e-07, + "loss": 0.1045, + "step": 8030 + }, + { + "epoch": 0.42, + "learning_rate": 2.610697706862435e-07, + "loss": 0.1389, + "step": 8031 + }, + { + "epoch": 0.42, + "learning_rate": 2.6103757316151933e-07, + "loss": 0.1444, + "step": 8032 + }, + { + "epoch": 0.42, + "learning_rate": 2.610053738923177e-07, + "loss": 0.1485, + "step": 8033 + }, + { + "epoch": 0.42, + "learning_rate": 2.60973172879559e-07, + "loss": 0.1578, + "step": 8034 + }, + { + "epoch": 0.42, + "learning_rate": 2.6094097012416336e-07, + "loss": 0.147, + "step": 8035 + }, + { + "epoch": 0.42, + "learning_rate": 2.609087656270515e-07, + "loss": 0.1086, + "step": 8036 + }, + { + "epoch": 0.42, + "learning_rate": 2.608765593891434e-07, + "loss": 0.1332, + "step": 8037 + }, + { + "epoch": 0.42, + "learning_rate": 2.608443514113598e-07, + "loss": 0.1284, + "step": 8038 + }, + { + "epoch": 0.42, + "learning_rate": 2.608121416946211e-07, + "loss": 0.1272, + "step": 8039 + }, + { + "epoch": 0.42, + "learning_rate": 2.6077993023984807e-07, + "loss": 0.169, + "step": 8040 + }, + { + "epoch": 0.42, + "learning_rate": 2.60747717047961e-07, + "loss": 0.1384, + "step": 8041 + }, + { + "epoch": 0.42, + "learning_rate": 2.607155021198808e-07, + "loss": 0.1414, + "step": 8042 + }, + { + "epoch": 0.42, + "learning_rate": 2.6068328545652813e-07, + "loss": 0.1432, + "step": 8043 + }, + { + "epoch": 0.42, + "learning_rate": 2.606510670588237e-07, + "loss": 0.1345, + "step": 8044 + }, + { + "epoch": 0.42, + "learning_rate": 2.606188469276884e-07, + "loss": 0.12, + "step": 8045 + }, + { + "epoch": 0.42, + "learning_rate": 2.6058662506404296e-07, + "loss": 0.2109, + "step": 8046 + }, + { + "epoch": 0.42, + "learning_rate": 2.605544014688085e-07, + "loss": 0.1643, + "step": 8047 + }, + { + "epoch": 0.42, + "learning_rate": 2.6052217614290575e-07, + "loss": 0.1451, + "step": 8048 + }, + { + "epoch": 0.42, + "learning_rate": 2.60489949087256e-07, + "loss": 0.1538, + "step": 8049 + }, + { + "epoch": 0.42, + "learning_rate": 2.6045772030278e-07, + "loss": 0.1163, + "step": 8050 + }, + { + "epoch": 0.42, + "learning_rate": 2.604254897903991e-07, + "loss": 0.1448, + "step": 8051 + }, + { + "epoch": 0.42, + "learning_rate": 2.6039325755103435e-07, + "loss": 0.1432, + "step": 8052 + }, + { + "epoch": 0.42, + "learning_rate": 2.6036102358560696e-07, + "loss": 0.1707, + "step": 8053 + }, + { + "epoch": 0.42, + "learning_rate": 2.6032878789503827e-07, + "loss": 0.1344, + "step": 8054 + }, + { + "epoch": 0.42, + "learning_rate": 2.602965504802495e-07, + "loss": 0.1601, + "step": 8055 + }, + { + "epoch": 0.42, + "learning_rate": 2.6026431134216203e-07, + "loss": 0.135, + "step": 8056 + }, + { + "epoch": 0.42, + "learning_rate": 2.6023207048169735e-07, + "loss": 0.1181, + "step": 8057 + }, + { + "epoch": 0.42, + "learning_rate": 2.601998278997768e-07, + "loss": 0.162, + "step": 8058 + }, + { + "epoch": 0.42, + "learning_rate": 2.6016758359732186e-07, + "loss": 0.1311, + "step": 8059 + }, + { + "epoch": 0.42, + "learning_rate": 2.6013533757525424e-07, + "loss": 0.1385, + "step": 8060 + }, + { + "epoch": 0.42, + "learning_rate": 2.6010308983449547e-07, + "loss": 0.1533, + "step": 8061 + }, + { + "epoch": 0.42, + "learning_rate": 2.600708403759671e-07, + "loss": 0.1521, + "step": 8062 + }, + { + "epoch": 0.42, + "learning_rate": 2.6003858920059096e-07, + "loss": 0.1452, + "step": 8063 + }, + { + "epoch": 0.42, + "learning_rate": 2.6000633630928874e-07, + "loss": 0.1354, + "step": 8064 + }, + { + "epoch": 0.42, + "learning_rate": 2.599740817029823e-07, + "loss": 0.1318, + "step": 8065 + }, + { + "epoch": 0.42, + "learning_rate": 2.599418253825935e-07, + "loss": 0.1243, + "step": 8066 + }, + { + "epoch": 0.42, + "learning_rate": 2.5990956734904414e-07, + "loss": 0.1336, + "step": 8067 + }, + { + "epoch": 0.42, + "learning_rate": 2.5987730760325617e-07, + "loss": 0.1278, + "step": 8068 + }, + { + "epoch": 0.42, + "learning_rate": 2.598450461461517e-07, + "loss": 0.1246, + "step": 8069 + }, + { + "epoch": 0.42, + "learning_rate": 2.5981278297865274e-07, + "loss": 0.1124, + "step": 8070 + }, + { + "epoch": 0.42, + "learning_rate": 2.5978051810168126e-07, + "loss": 0.1675, + "step": 8071 + }, + { + "epoch": 0.42, + "learning_rate": 2.597482515161595e-07, + "loss": 0.1538, + "step": 8072 + }, + { + "epoch": 0.42, + "learning_rate": 2.5971598322300973e-07, + "loss": 0.1192, + "step": 8073 + }, + { + "epoch": 0.42, + "learning_rate": 2.596837132231541e-07, + "loss": 0.154, + "step": 8074 + }, + { + "epoch": 0.42, + "learning_rate": 2.5965144151751485e-07, + "loss": 0.1373, + "step": 8075 + }, + { + "epoch": 0.42, + "learning_rate": 2.596191681070144e-07, + "loss": 0.1574, + "step": 8076 + }, + { + "epoch": 0.42, + "learning_rate": 2.595868929925751e-07, + "loss": 0.1149, + "step": 8077 + }, + { + "epoch": 0.42, + "learning_rate": 2.595546161751194e-07, + "loss": 0.2228, + "step": 8078 + }, + { + "epoch": 0.42, + "learning_rate": 2.5952233765556977e-07, + "loss": 0.1401, + "step": 8079 + }, + { + "epoch": 0.42, + "learning_rate": 2.5949005743484866e-07, + "loss": 0.1454, + "step": 8080 + }, + { + "epoch": 0.42, + "learning_rate": 2.594577755138789e-07, + "loss": 0.115, + "step": 8081 + }, + { + "epoch": 0.42, + "learning_rate": 2.594254918935829e-07, + "loss": 0.123, + "step": 8082 + }, + { + "epoch": 0.42, + "learning_rate": 2.5939320657488346e-07, + "loss": 0.1431, + "step": 8083 + }, + { + "epoch": 0.42, + "learning_rate": 2.5936091955870316e-07, + "loss": 0.1548, + "step": 8084 + }, + { + "epoch": 0.42, + "learning_rate": 2.593286308459649e-07, + "loss": 0.1481, + "step": 8085 + }, + { + "epoch": 0.42, + "learning_rate": 2.592963404375914e-07, + "loss": 0.1291, + "step": 8086 + }, + { + "epoch": 0.42, + "learning_rate": 2.592640483345057e-07, + "loss": 0.1283, + "step": 8087 + }, + { + "epoch": 0.42, + "learning_rate": 2.5923175453763064e-07, + "loss": 0.1453, + "step": 8088 + }, + { + "epoch": 0.42, + "learning_rate": 2.59199459047889e-07, + "loss": 0.1461, + "step": 8089 + }, + { + "epoch": 0.42, + "learning_rate": 2.5916716186620414e-07, + "loss": 0.168, + "step": 8090 + }, + { + "epoch": 0.42, + "learning_rate": 2.5913486299349895e-07, + "loss": 0.1123, + "step": 8091 + }, + { + "epoch": 0.42, + "learning_rate": 2.5910256243069647e-07, + "loss": 0.1528, + "step": 8092 + }, + { + "epoch": 0.42, + "learning_rate": 2.590702601787199e-07, + "loss": 0.1382, + "step": 8093 + }, + { + "epoch": 0.42, + "learning_rate": 2.5903795623849263e-07, + "loss": 0.1323, + "step": 8094 + }, + { + "epoch": 0.42, + "learning_rate": 2.5900565061093764e-07, + "loss": 0.1793, + "step": 8095 + }, + { + "epoch": 0.42, + "learning_rate": 2.589733432969785e-07, + "loss": 0.1556, + "step": 8096 + }, + { + "epoch": 0.42, + "learning_rate": 2.5894103429753833e-07, + "loss": 0.1387, + "step": 8097 + }, + { + "epoch": 0.42, + "learning_rate": 2.589087236135407e-07, + "loss": 0.1469, + "step": 8098 + }, + { + "epoch": 0.42, + "learning_rate": 2.58876411245909e-07, + "loss": 0.1268, + "step": 8099 + }, + { + "epoch": 0.42, + "learning_rate": 2.588440971955667e-07, + "loss": 0.1402, + "step": 8100 + }, + { + "epoch": 0.42, + "learning_rate": 2.588117814634375e-07, + "loss": 0.1245, + "step": 8101 + }, + { + "epoch": 0.42, + "learning_rate": 2.587794640504448e-07, + "loss": 0.161, + "step": 8102 + }, + { + "epoch": 0.42, + "learning_rate": 2.5874714495751233e-07, + "loss": 0.1136, + "step": 8103 + }, + { + "epoch": 0.42, + "learning_rate": 2.587148241855638e-07, + "loss": 0.1409, + "step": 8104 + }, + { + "epoch": 0.42, + "learning_rate": 2.586825017355229e-07, + "loss": 0.1448, + "step": 8105 + }, + { + "epoch": 0.42, + "learning_rate": 2.5865017760831345e-07, + "loss": 0.1163, + "step": 8106 + }, + { + "epoch": 0.42, + "learning_rate": 2.586178518048593e-07, + "loss": 0.1303, + "step": 8107 + }, + { + "epoch": 0.42, + "learning_rate": 2.585855243260843e-07, + "loss": 0.1791, + "step": 8108 + }, + { + "epoch": 0.42, + "learning_rate": 2.5855319517291245e-07, + "loss": 0.1391, + "step": 8109 + }, + { + "epoch": 0.42, + "learning_rate": 2.585208643462676e-07, + "loss": 0.1695, + "step": 8110 + }, + { + "epoch": 0.42, + "learning_rate": 2.5848853184707387e-07, + "loss": 0.1361, + "step": 8111 + }, + { + "epoch": 0.42, + "learning_rate": 2.5845619767625535e-07, + "loss": 0.1268, + "step": 8112 + }, + { + "epoch": 0.42, + "learning_rate": 2.584238618347362e-07, + "loss": 0.1693, + "step": 8113 + }, + { + "epoch": 0.42, + "learning_rate": 2.5839152432344043e-07, + "loss": 0.135, + "step": 8114 + }, + { + "epoch": 0.42, + "learning_rate": 2.5835918514329235e-07, + "loss": 0.1373, + "step": 8115 + }, + { + "epoch": 0.42, + "learning_rate": 2.583268442952163e-07, + "loss": 0.1452, + "step": 8116 + }, + { + "epoch": 0.42, + "learning_rate": 2.5829450178013644e-07, + "loss": 0.1474, + "step": 8117 + }, + { + "epoch": 0.42, + "learning_rate": 2.582621575989772e-07, + "loss": 0.1398, + "step": 8118 + }, + { + "epoch": 0.42, + "learning_rate": 2.582298117526631e-07, + "loss": 0.1281, + "step": 8119 + }, + { + "epoch": 0.42, + "learning_rate": 2.581974642421184e-07, + "loss": 0.1409, + "step": 8120 + }, + { + "epoch": 0.42, + "learning_rate": 2.5816511506826784e-07, + "loss": 0.1413, + "step": 8121 + }, + { + "epoch": 0.42, + "learning_rate": 2.5813276423203575e-07, + "loss": 0.1433, + "step": 8122 + }, + { + "epoch": 0.42, + "learning_rate": 2.581004117343469e-07, + "loss": 0.1537, + "step": 8123 + }, + { + "epoch": 0.42, + "learning_rate": 2.5806805757612574e-07, + "loss": 0.1445, + "step": 8124 + }, + { + "epoch": 0.42, + "learning_rate": 2.5803570175829706e-07, + "loss": 0.1495, + "step": 8125 + }, + { + "epoch": 0.42, + "learning_rate": 2.580033442817857e-07, + "loss": 0.1322, + "step": 8126 + }, + { + "epoch": 0.42, + "learning_rate": 2.5797098514751635e-07, + "loss": 0.1398, + "step": 8127 + }, + { + "epoch": 0.42, + "learning_rate": 2.579386243564139e-07, + "loss": 0.119, + "step": 8128 + }, + { + "epoch": 0.42, + "learning_rate": 2.5790626190940316e-07, + "loss": 0.1253, + "step": 8129 + }, + { + "epoch": 0.42, + "learning_rate": 2.5787389780740914e-07, + "loss": 0.1172, + "step": 8130 + }, + { + "epoch": 0.42, + "learning_rate": 2.578415320513567e-07, + "loss": 0.132, + "step": 8131 + }, + { + "epoch": 0.42, + "learning_rate": 2.57809164642171e-07, + "loss": 0.127, + "step": 8132 + }, + { + "epoch": 0.42, + "learning_rate": 2.57776795580777e-07, + "loss": 0.1231, + "step": 8133 + }, + { + "epoch": 0.42, + "learning_rate": 2.5774442486809996e-07, + "loss": 0.1144, + "step": 8134 + }, + { + "epoch": 0.42, + "learning_rate": 2.577120525050649e-07, + "loss": 0.1333, + "step": 8135 + }, + { + "epoch": 0.42, + "learning_rate": 2.5767967849259716e-07, + "loss": 0.1599, + "step": 8136 + }, + { + "epoch": 0.42, + "learning_rate": 2.576473028316219e-07, + "loss": 0.1189, + "step": 8137 + }, + { + "epoch": 0.42, + "learning_rate": 2.576149255230645e-07, + "loss": 0.1197, + "step": 8138 + }, + { + "epoch": 0.42, + "learning_rate": 2.575825465678503e-07, + "loss": 0.1707, + "step": 8139 + }, + { + "epoch": 0.42, + "learning_rate": 2.575501659669046e-07, + "loss": 0.1347, + "step": 8140 + }, + { + "epoch": 0.42, + "learning_rate": 2.5751778372115305e-07, + "loss": 0.1585, + "step": 8141 + }, + { + "epoch": 0.42, + "learning_rate": 2.5748539983152097e-07, + "loss": 0.1078, + "step": 8142 + }, + { + "epoch": 0.43, + "learning_rate": 2.57453014298934e-07, + "loss": 0.1363, + "step": 8143 + }, + { + "epoch": 0.43, + "learning_rate": 2.574206271243177e-07, + "loss": 0.1672, + "step": 8144 + }, + { + "epoch": 0.43, + "learning_rate": 2.573882383085976e-07, + "loss": 0.1506, + "step": 8145 + }, + { + "epoch": 0.43, + "learning_rate": 2.573558478526997e-07, + "loss": 0.1356, + "step": 8146 + }, + { + "epoch": 0.43, + "learning_rate": 2.5732345575754944e-07, + "loss": 0.1396, + "step": 8147 + }, + { + "epoch": 0.43, + "learning_rate": 2.5729106202407266e-07, + "loss": 0.148, + "step": 8148 + }, + { + "epoch": 0.43, + "learning_rate": 2.572586666531952e-07, + "loss": 0.112, + "step": 8149 + }, + { + "epoch": 0.43, + "learning_rate": 2.57226269645843e-07, + "loss": 0.1227, + "step": 8150 + }, + { + "epoch": 0.43, + "learning_rate": 2.5719387100294185e-07, + "loss": 0.1422, + "step": 8151 + }, + { + "epoch": 0.43, + "learning_rate": 2.5716147072541786e-07, + "loss": 0.1336, + "step": 8152 + }, + { + "epoch": 0.43, + "learning_rate": 2.571290688141969e-07, + "loss": 0.1251, + "step": 8153 + }, + { + "epoch": 0.43, + "learning_rate": 2.570966652702051e-07, + "loss": 0.1193, + "step": 8154 + }, + { + "epoch": 0.43, + "learning_rate": 2.570642600943686e-07, + "loss": 0.136, + "step": 8155 + }, + { + "epoch": 0.43, + "learning_rate": 2.5703185328761355e-07, + "loss": 0.1576, + "step": 8156 + }, + { + "epoch": 0.43, + "learning_rate": 2.5699944485086607e-07, + "loss": 0.1424, + "step": 8157 + }, + { + "epoch": 0.43, + "learning_rate": 2.5696703478505247e-07, + "loss": 0.1239, + "step": 8158 + }, + { + "epoch": 0.43, + "learning_rate": 2.5693462309109903e-07, + "loss": 0.1172, + "step": 8159 + }, + { + "epoch": 0.43, + "learning_rate": 2.5690220976993203e-07, + "loss": 0.1431, + "step": 8160 + }, + { + "epoch": 0.43, + "learning_rate": 2.56869794822478e-07, + "loss": 0.1183, + "step": 8161 + }, + { + "epoch": 0.43, + "learning_rate": 2.568373782496632e-07, + "loss": 0.173, + "step": 8162 + }, + { + "epoch": 0.43, + "learning_rate": 2.568049600524142e-07, + "loss": 0.1741, + "step": 8163 + }, + { + "epoch": 0.43, + "learning_rate": 2.5677254023165757e-07, + "loss": 0.1529, + "step": 8164 + }, + { + "epoch": 0.43, + "learning_rate": 2.5674011878831976e-07, + "loss": 0.1291, + "step": 8165 + }, + { + "epoch": 0.43, + "learning_rate": 2.5670769572332747e-07, + "loss": 0.1697, + "step": 8166 + }, + { + "epoch": 0.43, + "learning_rate": 2.5667527103760735e-07, + "loss": 0.1424, + "step": 8167 + }, + { + "epoch": 0.43, + "learning_rate": 2.566428447320861e-07, + "loss": 0.1323, + "step": 8168 + }, + { + "epoch": 0.43, + "learning_rate": 2.566104168076905e-07, + "loss": 0.1019, + "step": 8169 + }, + { + "epoch": 0.43, + "learning_rate": 2.5657798726534735e-07, + "loss": 0.133, + "step": 8170 + }, + { + "epoch": 0.43, + "learning_rate": 2.565455561059834e-07, + "loss": 0.1218, + "step": 8171 + }, + { + "epoch": 0.43, + "learning_rate": 2.5651312333052566e-07, + "loss": 0.1366, + "step": 8172 + }, + { + "epoch": 0.43, + "learning_rate": 2.564806889399011e-07, + "loss": 0.1321, + "step": 8173 + }, + { + "epoch": 0.43, + "learning_rate": 2.564482529350366e-07, + "loss": 0.1694, + "step": 8174 + }, + { + "epoch": 0.43, + "learning_rate": 2.564158153168592e-07, + "loss": 0.1336, + "step": 8175 + }, + { + "epoch": 0.43, + "learning_rate": 2.563833760862961e-07, + "loss": 0.121, + "step": 8176 + }, + { + "epoch": 0.43, + "learning_rate": 2.5635093524427434e-07, + "loss": 0.128, + "step": 8177 + }, + { + "epoch": 0.43, + "learning_rate": 2.5631849279172105e-07, + "loss": 0.1172, + "step": 8178 + }, + { + "epoch": 0.43, + "learning_rate": 2.5628604872956354e-07, + "loss": 0.1443, + "step": 8179 + }, + { + "epoch": 0.43, + "learning_rate": 2.5625360305872895e-07, + "loss": 0.1854, + "step": 8180 + }, + { + "epoch": 0.43, + "learning_rate": 2.562211557801447e-07, + "loss": 0.1682, + "step": 8181 + }, + { + "epoch": 0.43, + "learning_rate": 2.561887068947382e-07, + "loss": 0.1145, + "step": 8182 + }, + { + "epoch": 0.43, + "learning_rate": 2.5615625640343667e-07, + "loss": 0.1599, + "step": 8183 + }, + { + "epoch": 0.43, + "learning_rate": 2.5612380430716774e-07, + "loss": 0.1389, + "step": 8184 + }, + { + "epoch": 0.43, + "learning_rate": 2.5609135060685865e-07, + "loss": 0.1421, + "step": 8185 + }, + { + "epoch": 0.43, + "learning_rate": 2.560588953034373e-07, + "loss": 0.1623, + "step": 8186 + }, + { + "epoch": 0.43, + "learning_rate": 2.5602643839783093e-07, + "loss": 0.1305, + "step": 8187 + }, + { + "epoch": 0.43, + "learning_rate": 2.559939798909674e-07, + "loss": 0.136, + "step": 8188 + }, + { + "epoch": 0.43, + "learning_rate": 2.5596151978377433e-07, + "loss": 0.1544, + "step": 8189 + }, + { + "epoch": 0.43, + "learning_rate": 2.559290580771794e-07, + "loss": 0.1298, + "step": 8190 + }, + { + "epoch": 0.43, + "learning_rate": 2.558965947721104e-07, + "loss": 0.1323, + "step": 8191 + }, + { + "epoch": 0.43, + "learning_rate": 2.558641298694951e-07, + "loss": 0.146, + "step": 8192 + }, + { + "epoch": 0.43, + "learning_rate": 2.5583166337026143e-07, + "loss": 0.1365, + "step": 8193 + }, + { + "epoch": 0.43, + "learning_rate": 2.5579919527533726e-07, + "loss": 0.1526, + "step": 8194 + }, + { + "epoch": 0.43, + "learning_rate": 2.557667255856506e-07, + "loss": 0.1343, + "step": 8195 + }, + { + "epoch": 0.43, + "learning_rate": 2.5573425430212935e-07, + "loss": 0.133, + "step": 8196 + }, + { + "epoch": 0.43, + "learning_rate": 2.557017814257016e-07, + "loss": 0.1234, + "step": 8197 + }, + { + "epoch": 0.43, + "learning_rate": 2.5566930695729544e-07, + "loss": 0.1495, + "step": 8198 + }, + { + "epoch": 0.43, + "learning_rate": 2.55636830897839e-07, + "loss": 0.1386, + "step": 8199 + }, + { + "epoch": 0.43, + "learning_rate": 2.5560435324826043e-07, + "loss": 0.1289, + "step": 8200 + }, + { + "epoch": 0.43, + "learning_rate": 2.5557187400948796e-07, + "loss": 0.1768, + "step": 8201 + }, + { + "epoch": 0.43, + "learning_rate": 2.5553939318245e-07, + "loss": 0.1542, + "step": 8202 + }, + { + "epoch": 0.43, + "learning_rate": 2.555069107680746e-07, + "loss": 0.1262, + "step": 8203 + }, + { + "epoch": 0.43, + "learning_rate": 2.5547442676729037e-07, + "loss": 0.1116, + "step": 8204 + }, + { + "epoch": 0.43, + "learning_rate": 2.5544194118102546e-07, + "loss": 0.1574, + "step": 8205 + }, + { + "epoch": 0.43, + "learning_rate": 2.554094540102086e-07, + "loss": 0.1595, + "step": 8206 + }, + { + "epoch": 0.43, + "learning_rate": 2.5537696525576817e-07, + "loss": 0.1145, + "step": 8207 + }, + { + "epoch": 0.43, + "learning_rate": 2.5534447491863263e-07, + "loss": 0.1534, + "step": 8208 + }, + { + "epoch": 0.43, + "learning_rate": 2.5531198299973066e-07, + "loss": 0.1402, + "step": 8209 + }, + { + "epoch": 0.43, + "learning_rate": 2.5527948949999084e-07, + "loss": 0.1227, + "step": 8210 + }, + { + "epoch": 0.43, + "learning_rate": 2.5524699442034186e-07, + "loss": 0.1456, + "step": 8211 + }, + { + "epoch": 0.43, + "learning_rate": 2.5521449776171246e-07, + "loss": 0.1572, + "step": 8212 + }, + { + "epoch": 0.43, + "learning_rate": 2.551819995250314e-07, + "loss": 0.1725, + "step": 8213 + }, + { + "epoch": 0.43, + "learning_rate": 2.551494997112275e-07, + "loss": 0.1148, + "step": 8214 + }, + { + "epoch": 0.43, + "learning_rate": 2.551169983212296e-07, + "loss": 0.1516, + "step": 8215 + }, + { + "epoch": 0.43, + "learning_rate": 2.550844953559667e-07, + "loss": 0.136, + "step": 8216 + }, + { + "epoch": 0.43, + "learning_rate": 2.5505199081636757e-07, + "loss": 0.1617, + "step": 8217 + }, + { + "epoch": 0.43, + "learning_rate": 2.550194847033613e-07, + "loss": 0.1385, + "step": 8218 + }, + { + "epoch": 0.43, + "learning_rate": 2.5498697701787687e-07, + "loss": 0.1201, + "step": 8219 + }, + { + "epoch": 0.43, + "learning_rate": 2.5495446776084353e-07, + "loss": 0.1589, + "step": 8220 + }, + { + "epoch": 0.43, + "learning_rate": 2.5492195693319013e-07, + "loss": 0.1239, + "step": 8221 + }, + { + "epoch": 0.43, + "learning_rate": 2.548894445358461e-07, + "loss": 0.1297, + "step": 8222 + }, + { + "epoch": 0.43, + "learning_rate": 2.548569305697405e-07, + "loss": 0.1355, + "step": 8223 + }, + { + "epoch": 0.43, + "learning_rate": 2.548244150358028e-07, + "loss": 0.1372, + "step": 8224 + }, + { + "epoch": 0.43, + "learning_rate": 2.54791897934962e-07, + "loss": 0.1066, + "step": 8225 + }, + { + "epoch": 0.43, + "learning_rate": 2.547593792681476e-07, + "loss": 0.1536, + "step": 8226 + }, + { + "epoch": 0.43, + "learning_rate": 2.5472685903628906e-07, + "loss": 0.1498, + "step": 8227 + }, + { + "epoch": 0.43, + "learning_rate": 2.5469433724031576e-07, + "loss": 0.1417, + "step": 8228 + }, + { + "epoch": 0.43, + "learning_rate": 2.546618138811572e-07, + "loss": 0.1348, + "step": 8229 + }, + { + "epoch": 0.43, + "learning_rate": 2.546292889597428e-07, + "loss": 0.1888, + "step": 8230 + }, + { + "epoch": 0.43, + "learning_rate": 2.5459676247700234e-07, + "loss": 0.146, + "step": 8231 + }, + { + "epoch": 0.43, + "learning_rate": 2.5456423443386524e-07, + "loss": 0.1426, + "step": 8232 + }, + { + "epoch": 0.43, + "learning_rate": 2.545317048312614e-07, + "loss": 0.1676, + "step": 8233 + }, + { + "epoch": 0.43, + "learning_rate": 2.5449917367012024e-07, + "loss": 0.127, + "step": 8234 + }, + { + "epoch": 0.43, + "learning_rate": 2.5446664095137175e-07, + "loss": 0.1304, + "step": 8235 + }, + { + "epoch": 0.43, + "learning_rate": 2.544341066759456e-07, + "loss": 0.1157, + "step": 8236 + }, + { + "epoch": 0.43, + "learning_rate": 2.5440157084477167e-07, + "loss": 0.1551, + "step": 8237 + }, + { + "epoch": 0.43, + "learning_rate": 2.5436903345877984e-07, + "loss": 0.1512, + "step": 8238 + }, + { + "epoch": 0.43, + "learning_rate": 2.5433649451890003e-07, + "loss": 0.1078, + "step": 8239 + }, + { + "epoch": 0.43, + "learning_rate": 2.5430395402606224e-07, + "loss": 0.1116, + "step": 8240 + }, + { + "epoch": 0.43, + "learning_rate": 2.542714119811965e-07, + "loss": 0.1238, + "step": 8241 + }, + { + "epoch": 0.43, + "learning_rate": 2.5423886838523286e-07, + "loss": 0.1407, + "step": 8242 + }, + { + "epoch": 0.43, + "learning_rate": 2.542063232391014e-07, + "loss": 0.1484, + "step": 8243 + }, + { + "epoch": 0.43, + "learning_rate": 2.541737765437323e-07, + "loss": 0.1398, + "step": 8244 + }, + { + "epoch": 0.43, + "learning_rate": 2.5414122830005574e-07, + "loss": 0.1345, + "step": 8245 + }, + { + "epoch": 0.43, + "learning_rate": 2.54108678509002e-07, + "loss": 0.1576, + "step": 8246 + }, + { + "epoch": 0.43, + "learning_rate": 2.540761271715014e-07, + "loss": 0.1547, + "step": 8247 + }, + { + "epoch": 0.43, + "learning_rate": 2.540435742884841e-07, + "loss": 0.1579, + "step": 8248 + }, + { + "epoch": 0.43, + "learning_rate": 2.540110198608806e-07, + "loss": 0.1466, + "step": 8249 + }, + { + "epoch": 0.43, + "learning_rate": 2.539784638896213e-07, + "loss": 0.1555, + "step": 8250 + }, + { + "epoch": 0.43, + "learning_rate": 2.5394590637563676e-07, + "loss": 0.1206, + "step": 8251 + }, + { + "epoch": 0.43, + "learning_rate": 2.5391334731985726e-07, + "loss": 0.1606, + "step": 8252 + }, + { + "epoch": 0.43, + "learning_rate": 2.538807867232136e-07, + "loss": 0.1423, + "step": 8253 + }, + { + "epoch": 0.43, + "learning_rate": 2.538482245866362e-07, + "loss": 0.111, + "step": 8254 + }, + { + "epoch": 0.43, + "learning_rate": 2.5381566091105573e-07, + "loss": 0.148, + "step": 8255 + }, + { + "epoch": 0.43, + "learning_rate": 2.5378309569740293e-07, + "loss": 0.1543, + "step": 8256 + }, + { + "epoch": 0.43, + "learning_rate": 2.537505289466084e-07, + "loss": 0.1158, + "step": 8257 + }, + { + "epoch": 0.43, + "learning_rate": 2.5371796065960323e-07, + "loss": 0.1497, + "step": 8258 + }, + { + "epoch": 0.43, + "learning_rate": 2.536853908373178e-07, + "loss": 0.1783, + "step": 8259 + }, + { + "epoch": 0.43, + "learning_rate": 2.5365281948068326e-07, + "loss": 0.1462, + "step": 8260 + }, + { + "epoch": 0.43, + "learning_rate": 2.5362024659063037e-07, + "loss": 0.1329, + "step": 8261 + }, + { + "epoch": 0.43, + "learning_rate": 2.5358767216809015e-07, + "loss": 0.1571, + "step": 8262 + }, + { + "epoch": 0.43, + "learning_rate": 2.535550962139936e-07, + "loss": 0.2116, + "step": 8263 + }, + { + "epoch": 0.43, + "learning_rate": 2.5352251872927174e-07, + "loss": 0.1274, + "step": 8264 + }, + { + "epoch": 0.43, + "learning_rate": 2.534899397148556e-07, + "loss": 0.17, + "step": 8265 + }, + { + "epoch": 0.43, + "learning_rate": 2.5345735917167637e-07, + "loss": 0.1359, + "step": 8266 + }, + { + "epoch": 0.43, + "learning_rate": 2.534247771006651e-07, + "loss": 0.1256, + "step": 8267 + }, + { + "epoch": 0.43, + "learning_rate": 2.533921935027531e-07, + "loss": 0.1579, + "step": 8268 + }, + { + "epoch": 0.43, + "learning_rate": 2.5335960837887164e-07, + "loss": 0.1433, + "step": 8269 + }, + { + "epoch": 0.43, + "learning_rate": 2.53327021729952e-07, + "loss": 0.1288, + "step": 8270 + }, + { + "epoch": 0.43, + "learning_rate": 2.5329443355692543e-07, + "loss": 0.1537, + "step": 8271 + }, + { + "epoch": 0.43, + "learning_rate": 2.532618438607234e-07, + "loss": 0.1339, + "step": 8272 + }, + { + "epoch": 0.43, + "learning_rate": 2.532292526422773e-07, + "loss": 0.1507, + "step": 8273 + }, + { + "epoch": 0.43, + "learning_rate": 2.531966599025186e-07, + "loss": 0.1447, + "step": 8274 + }, + { + "epoch": 0.43, + "learning_rate": 2.5316406564237883e-07, + "loss": 0.1685, + "step": 8275 + }, + { + "epoch": 0.43, + "learning_rate": 2.531314698627896e-07, + "loss": 0.1414, + "step": 8276 + }, + { + "epoch": 0.43, + "learning_rate": 2.5309887256468234e-07, + "loss": 0.1547, + "step": 8277 + }, + { + "epoch": 0.43, + "learning_rate": 2.530662737489889e-07, + "loss": 0.1546, + "step": 8278 + }, + { + "epoch": 0.43, + "learning_rate": 2.530336734166408e-07, + "loss": 0.1209, + "step": 8279 + }, + { + "epoch": 0.43, + "learning_rate": 2.5300107156856994e-07, + "loss": 0.1229, + "step": 8280 + }, + { + "epoch": 0.43, + "learning_rate": 2.529684682057079e-07, + "loss": 0.0969, + "step": 8281 + }, + { + "epoch": 0.43, + "learning_rate": 2.5293586332898666e-07, + "loss": 0.1098, + "step": 8282 + }, + { + "epoch": 0.43, + "learning_rate": 2.5290325693933794e-07, + "loss": 0.168, + "step": 8283 + }, + { + "epoch": 0.43, + "learning_rate": 2.528706490376938e-07, + "loss": 0.1199, + "step": 8284 + }, + { + "epoch": 0.43, + "learning_rate": 2.52838039624986e-07, + "loss": 0.1381, + "step": 8285 + }, + { + "epoch": 0.43, + "learning_rate": 2.528054287021467e-07, + "loss": 0.1448, + "step": 8286 + }, + { + "epoch": 0.43, + "learning_rate": 2.527728162701078e-07, + "loss": 0.1396, + "step": 8287 + }, + { + "epoch": 0.43, + "learning_rate": 2.5274020232980156e-07, + "loss": 0.1305, + "step": 8288 + }, + { + "epoch": 0.43, + "learning_rate": 2.5270758688215985e-07, + "loss": 0.142, + "step": 8289 + }, + { + "epoch": 0.43, + "learning_rate": 2.52674969928115e-07, + "loss": 0.1242, + "step": 8290 + }, + { + "epoch": 0.43, + "learning_rate": 2.526423514685992e-07, + "loss": 0.1379, + "step": 8291 + }, + { + "epoch": 0.43, + "learning_rate": 2.526097315045446e-07, + "loss": 0.1624, + "step": 8292 + }, + { + "epoch": 0.43, + "learning_rate": 2.5257711003688363e-07, + "loss": 0.1155, + "step": 8293 + }, + { + "epoch": 0.43, + "learning_rate": 2.5254448706654855e-07, + "loss": 0.1673, + "step": 8294 + }, + { + "epoch": 0.43, + "learning_rate": 2.525118625944717e-07, + "loss": 0.0823, + "step": 8295 + }, + { + "epoch": 0.43, + "learning_rate": 2.5247923662158563e-07, + "loss": 0.1438, + "step": 8296 + }, + { + "epoch": 0.43, + "learning_rate": 2.524466091488227e-07, + "loss": 0.155, + "step": 8297 + }, + { + "epoch": 0.43, + "learning_rate": 2.524139801771154e-07, + "loss": 0.1347, + "step": 8298 + }, + { + "epoch": 0.43, + "learning_rate": 2.5238134970739635e-07, + "loss": 0.1696, + "step": 8299 + }, + { + "epoch": 0.43, + "learning_rate": 2.523487177405981e-07, + "loss": 0.169, + "step": 8300 + }, + { + "epoch": 0.43, + "learning_rate": 2.523160842776532e-07, + "loss": 0.1181, + "step": 8301 + }, + { + "epoch": 0.43, + "learning_rate": 2.5228344931949447e-07, + "loss": 0.117, + "step": 8302 + }, + { + "epoch": 0.43, + "learning_rate": 2.522508128670546e-07, + "loss": 0.1319, + "step": 8303 + }, + { + "epoch": 0.43, + "learning_rate": 2.522181749212663e-07, + "loss": 0.1349, + "step": 8304 + }, + { + "epoch": 0.43, + "learning_rate": 2.5218553548306245e-07, + "loss": 0.1674, + "step": 8305 + }, + { + "epoch": 0.43, + "learning_rate": 2.5215289455337584e-07, + "loss": 0.1737, + "step": 8306 + }, + { + "epoch": 0.43, + "learning_rate": 2.5212025213313936e-07, + "loss": 0.1417, + "step": 8307 + }, + { + "epoch": 0.43, + "learning_rate": 2.520876082232859e-07, + "loss": 0.1458, + "step": 8308 + }, + { + "epoch": 0.43, + "learning_rate": 2.5205496282474855e-07, + "loss": 0.1371, + "step": 8309 + }, + { + "epoch": 0.43, + "learning_rate": 2.5202231593846024e-07, + "loss": 0.1506, + "step": 8310 + }, + { + "epoch": 0.43, + "learning_rate": 2.51989667565354e-07, + "loss": 0.1328, + "step": 8311 + }, + { + "epoch": 0.43, + "learning_rate": 2.519570177063631e-07, + "loss": 0.1301, + "step": 8312 + }, + { + "epoch": 0.43, + "learning_rate": 2.519243663624205e-07, + "loss": 0.141, + "step": 8313 + }, + { + "epoch": 0.43, + "learning_rate": 2.518917135344596e-07, + "loss": 0.1477, + "step": 8314 + }, + { + "epoch": 0.43, + "learning_rate": 2.518590592234133e-07, + "loss": 0.1845, + "step": 8315 + }, + { + "epoch": 0.43, + "learning_rate": 2.518264034302152e-07, + "loss": 0.1371, + "step": 8316 + }, + { + "epoch": 0.43, + "learning_rate": 2.517937461557984e-07, + "loss": 0.158, + "step": 8317 + }, + { + "epoch": 0.43, + "learning_rate": 2.5176108740109646e-07, + "loss": 0.1189, + "step": 8318 + }, + { + "epoch": 0.43, + "learning_rate": 2.517284271670426e-07, + "loss": 0.143, + "step": 8319 + }, + { + "epoch": 0.43, + "learning_rate": 2.5169576545457024e-07, + "loss": 0.1528, + "step": 8320 + }, + { + "epoch": 0.43, + "learning_rate": 2.51663102264613e-07, + "loss": 0.1484, + "step": 8321 + }, + { + "epoch": 0.43, + "learning_rate": 2.516304375981044e-07, + "loss": 0.1702, + "step": 8322 + }, + { + "epoch": 0.43, + "learning_rate": 2.5159777145597797e-07, + "loss": 0.1587, + "step": 8323 + }, + { + "epoch": 0.43, + "learning_rate": 2.515651038391672e-07, + "loss": 0.1247, + "step": 8324 + }, + { + "epoch": 0.43, + "learning_rate": 2.5153243474860594e-07, + "loss": 0.1586, + "step": 8325 + }, + { + "epoch": 0.43, + "learning_rate": 2.514997641852278e-07, + "loss": 0.1135, + "step": 8326 + }, + { + "epoch": 0.43, + "learning_rate": 2.5146709214996655e-07, + "loss": 0.1429, + "step": 8327 + }, + { + "epoch": 0.43, + "learning_rate": 2.5143441864375585e-07, + "loss": 0.1273, + "step": 8328 + }, + { + "epoch": 0.43, + "learning_rate": 2.514017436675297e-07, + "loss": 0.131, + "step": 8329 + }, + { + "epoch": 0.43, + "learning_rate": 2.5136906722222175e-07, + "loss": 0.1205, + "step": 8330 + }, + { + "epoch": 0.43, + "learning_rate": 2.5133638930876617e-07, + "loss": 0.1898, + "step": 8331 + }, + { + "epoch": 0.43, + "learning_rate": 2.513037099280967e-07, + "loss": 0.1275, + "step": 8332 + }, + { + "epoch": 0.43, + "learning_rate": 2.5127102908114737e-07, + "loss": 0.1176, + "step": 8333 + }, + { + "epoch": 0.43, + "learning_rate": 2.5123834676885223e-07, + "loss": 0.1436, + "step": 8334 + }, + { + "epoch": 0.44, + "learning_rate": 2.512056629921454e-07, + "loss": 0.1262, + "step": 8335 + }, + { + "epoch": 0.44, + "learning_rate": 2.5117297775196095e-07, + "loss": 0.1407, + "step": 8336 + }, + { + "epoch": 0.44, + "learning_rate": 2.5114029104923306e-07, + "loss": 0.1255, + "step": 8337 + }, + { + "epoch": 0.44, + "learning_rate": 2.5110760288489586e-07, + "loss": 0.1115, + "step": 8338 + }, + { + "epoch": 0.44, + "learning_rate": 2.510749132598837e-07, + "loss": 0.1363, + "step": 8339 + }, + { + "epoch": 0.44, + "learning_rate": 2.5104222217513075e-07, + "loss": 0.1197, + "step": 8340 + }, + { + "epoch": 0.44, + "learning_rate": 2.5100952963157144e-07, + "loss": 0.2039, + "step": 8341 + }, + { + "epoch": 0.44, + "learning_rate": 2.5097683563014e-07, + "loss": 0.1396, + "step": 8342 + }, + { + "epoch": 0.44, + "learning_rate": 2.50944140171771e-07, + "loss": 0.1169, + "step": 8343 + }, + { + "epoch": 0.44, + "learning_rate": 2.5091144325739876e-07, + "loss": 0.1464, + "step": 8344 + }, + { + "epoch": 0.44, + "learning_rate": 2.5087874488795785e-07, + "loss": 0.1402, + "step": 8345 + }, + { + "epoch": 0.44, + "learning_rate": 2.5084604506438276e-07, + "loss": 0.1346, + "step": 8346 + }, + { + "epoch": 0.44, + "learning_rate": 2.50813343787608e-07, + "loss": 0.1416, + "step": 8347 + }, + { + "epoch": 0.44, + "learning_rate": 2.507806410585684e-07, + "loss": 0.1423, + "step": 8348 + }, + { + "epoch": 0.44, + "learning_rate": 2.5074793687819833e-07, + "loss": 0.1239, + "step": 8349 + }, + { + "epoch": 0.44, + "learning_rate": 2.5071523124743275e-07, + "loss": 0.1694, + "step": 8350 + }, + { + "epoch": 0.44, + "learning_rate": 2.5068252416720617e-07, + "loss": 0.1243, + "step": 8351 + }, + { + "epoch": 0.44, + "learning_rate": 2.506498156384536e-07, + "loss": 0.1316, + "step": 8352 + }, + { + "epoch": 0.44, + "learning_rate": 2.5061710566210965e-07, + "loss": 0.1055, + "step": 8353 + }, + { + "epoch": 0.44, + "learning_rate": 2.5058439423910937e-07, + "loss": 0.1279, + "step": 8354 + }, + { + "epoch": 0.44, + "learning_rate": 2.505516813703875e-07, + "loss": 0.1163, + "step": 8355 + }, + { + "epoch": 0.44, + "learning_rate": 2.505189670568791e-07, + "loss": 0.136, + "step": 8356 + }, + { + "epoch": 0.44, + "learning_rate": 2.5048625129951913e-07, + "loss": 0.1164, + "step": 8357 + }, + { + "epoch": 0.44, + "learning_rate": 2.504535340992426e-07, + "loss": 0.162, + "step": 8358 + }, + { + "epoch": 0.44, + "learning_rate": 2.504208154569845e-07, + "loss": 0.1283, + "step": 8359 + }, + { + "epoch": 0.44, + "learning_rate": 2.503880953736801e-07, + "loss": 0.1185, + "step": 8360 + }, + { + "epoch": 0.44, + "learning_rate": 2.5035537385026456e-07, + "loss": 0.1344, + "step": 8361 + }, + { + "epoch": 0.44, + "learning_rate": 2.503226508876729e-07, + "loss": 0.1138, + "step": 8362 + }, + { + "epoch": 0.44, + "learning_rate": 2.502899264868405e-07, + "loss": 0.153, + "step": 8363 + }, + { + "epoch": 0.44, + "learning_rate": 2.5025720064870254e-07, + "loss": 0.1291, + "step": 8364 + }, + { + "epoch": 0.44, + "learning_rate": 2.5022447337419443e-07, + "loss": 0.1276, + "step": 8365 + }, + { + "epoch": 0.44, + "learning_rate": 2.5019174466425146e-07, + "loss": 0.1533, + "step": 8366 + }, + { + "epoch": 0.44, + "learning_rate": 2.50159014519809e-07, + "loss": 0.1466, + "step": 8367 + }, + { + "epoch": 0.44, + "learning_rate": 2.5012628294180264e-07, + "loss": 0.1681, + "step": 8368 + }, + { + "epoch": 0.44, + "learning_rate": 2.5009354993116767e-07, + "loss": 0.1113, + "step": 8369 + }, + { + "epoch": 0.44, + "learning_rate": 2.500608154888398e-07, + "loss": 0.1207, + "step": 8370 + }, + { + "epoch": 0.44, + "learning_rate": 2.500280796157544e-07, + "loss": 0.1292, + "step": 8371 + }, + { + "epoch": 0.44, + "learning_rate": 2.4999534231284726e-07, + "loss": 0.1536, + "step": 8372 + }, + { + "epoch": 0.44, + "learning_rate": 2.499626035810539e-07, + "loss": 0.1506, + "step": 8373 + }, + { + "epoch": 0.44, + "learning_rate": 2.499298634213101e-07, + "loss": 0.0996, + "step": 8374 + }, + { + "epoch": 0.44, + "learning_rate": 2.498971218345514e-07, + "loss": 0.1651, + "step": 8375 + }, + { + "epoch": 0.44, + "learning_rate": 2.498643788217138e-07, + "loss": 0.1768, + "step": 8376 + }, + { + "epoch": 0.44, + "learning_rate": 2.49831634383733e-07, + "loss": 0.114, + "step": 8377 + }, + { + "epoch": 0.44, + "learning_rate": 2.497988885215448e-07, + "loss": 0.1401, + "step": 8378 + }, + { + "epoch": 0.44, + "learning_rate": 2.4976614123608515e-07, + "loss": 0.1662, + "step": 8379 + }, + { + "epoch": 0.44, + "learning_rate": 2.497333925282899e-07, + "loss": 0.136, + "step": 8380 + }, + { + "epoch": 0.44, + "learning_rate": 2.497006423990952e-07, + "loss": 0.1175, + "step": 8381 + }, + { + "epoch": 0.44, + "learning_rate": 2.4966789084943694e-07, + "loss": 0.1687, + "step": 8382 + }, + { + "epoch": 0.44, + "learning_rate": 2.4963513788025124e-07, + "loss": 0.1768, + "step": 8383 + }, + { + "epoch": 0.44, + "learning_rate": 2.4960238349247403e-07, + "loss": 0.1458, + "step": 8384 + }, + { + "epoch": 0.44, + "learning_rate": 2.4956962768704166e-07, + "loss": 0.1156, + "step": 8385 + }, + { + "epoch": 0.44, + "learning_rate": 2.4953687046489003e-07, + "loss": 0.1418, + "step": 8386 + }, + { + "epoch": 0.44, + "learning_rate": 2.4950411182695566e-07, + "loss": 0.1154, + "step": 8387 + }, + { + "epoch": 0.44, + "learning_rate": 2.4947135177417467e-07, + "loss": 0.1157, + "step": 8388 + }, + { + "epoch": 0.44, + "learning_rate": 2.4943859030748324e-07, + "loss": 0.1637, + "step": 8389 + }, + { + "epoch": 0.44, + "learning_rate": 2.4940582742781793e-07, + "loss": 0.1316, + "step": 8390 + }, + { + "epoch": 0.44, + "learning_rate": 2.493730631361149e-07, + "loss": 0.1455, + "step": 8391 + }, + { + "epoch": 0.44, + "learning_rate": 2.4934029743331077e-07, + "loss": 0.1201, + "step": 8392 + }, + { + "epoch": 0.44, + "learning_rate": 2.493075303203418e-07, + "loss": 0.1905, + "step": 8393 + }, + { + "epoch": 0.44, + "learning_rate": 2.492747617981446e-07, + "loss": 0.1418, + "step": 8394 + }, + { + "epoch": 0.44, + "learning_rate": 2.492419918676557e-07, + "loss": 0.1376, + "step": 8395 + }, + { + "epoch": 0.44, + "learning_rate": 2.4920922052981167e-07, + "loss": 0.15, + "step": 8396 + }, + { + "epoch": 0.44, + "learning_rate": 2.491764477855491e-07, + "loss": 0.152, + "step": 8397 + }, + { + "epoch": 0.44, + "learning_rate": 2.4914367363580464e-07, + "loss": 0.1269, + "step": 8398 + }, + { + "epoch": 0.44, + "learning_rate": 2.491108980815151e-07, + "loss": 0.1758, + "step": 8399 + }, + { + "epoch": 0.44, + "learning_rate": 2.4907812112361706e-07, + "loss": 0.1511, + "step": 8400 + }, + { + "epoch": 0.44, + "learning_rate": 2.490453427630475e-07, + "loss": 0.1155, + "step": 8401 + }, + { + "epoch": 0.44, + "learning_rate": 2.490125630007429e-07, + "loss": 0.1709, + "step": 8402 + }, + { + "epoch": 0.44, + "learning_rate": 2.489797818376405e-07, + "loss": 0.12, + "step": 8403 + }, + { + "epoch": 0.44, + "learning_rate": 2.48946999274677e-07, + "loss": 0.1607, + "step": 8404 + }, + { + "epoch": 0.44, + "learning_rate": 2.4891421531278927e-07, + "loss": 0.1623, + "step": 8405 + }, + { + "epoch": 0.44, + "learning_rate": 2.4888142995291443e-07, + "loss": 0.1353, + "step": 8406 + }, + { + "epoch": 0.44, + "learning_rate": 2.4884864319598946e-07, + "loss": 0.1574, + "step": 8407 + }, + { + "epoch": 0.44, + "learning_rate": 2.488158550429515e-07, + "loss": 0.1219, + "step": 8408 + }, + { + "epoch": 0.44, + "learning_rate": 2.487830654947374e-07, + "loss": 0.1493, + "step": 8409 + }, + { + "epoch": 0.44, + "learning_rate": 2.487502745522846e-07, + "loss": 0.1634, + "step": 8410 + }, + { + "epoch": 0.44, + "learning_rate": 2.4871748221653004e-07, + "loss": 0.1709, + "step": 8411 + }, + { + "epoch": 0.44, + "learning_rate": 2.4868468848841104e-07, + "loss": 0.1359, + "step": 8412 + }, + { + "epoch": 0.44, + "learning_rate": 2.486518933688649e-07, + "loss": 0.1068, + "step": 8413 + }, + { + "epoch": 0.44, + "learning_rate": 2.486190968588288e-07, + "loss": 0.1063, + "step": 8414 + }, + { + "epoch": 0.44, + "learning_rate": 2.485862989592402e-07, + "loss": 0.1393, + "step": 8415 + }, + { + "epoch": 0.44, + "learning_rate": 2.4855349967103636e-07, + "loss": 0.1462, + "step": 8416 + }, + { + "epoch": 0.44, + "learning_rate": 2.4852069899515486e-07, + "loss": 0.1549, + "step": 8417 + }, + { + "epoch": 0.44, + "learning_rate": 2.4848789693253294e-07, + "loss": 0.1121, + "step": 8418 + }, + { + "epoch": 0.44, + "learning_rate": 2.484550934841082e-07, + "loss": 0.1362, + "step": 8419 + }, + { + "epoch": 0.44, + "learning_rate": 2.484222886508183e-07, + "loss": 0.1581, + "step": 8420 + }, + { + "epoch": 0.44, + "learning_rate": 2.483894824336006e-07, + "loss": 0.1827, + "step": 8421 + }, + { + "epoch": 0.44, + "learning_rate": 2.4835667483339285e-07, + "loss": 0.1673, + "step": 8422 + }, + { + "epoch": 0.44, + "learning_rate": 2.483238658511326e-07, + "loss": 0.1227, + "step": 8423 + }, + { + "epoch": 0.44, + "learning_rate": 2.482910554877577e-07, + "loss": 0.1133, + "step": 8424 + }, + { + "epoch": 0.44, + "learning_rate": 2.4825824374420575e-07, + "loss": 0.1593, + "step": 8425 + }, + { + "epoch": 0.44, + "learning_rate": 2.482254306214146e-07, + "loss": 0.1617, + "step": 8426 + }, + { + "epoch": 0.44, + "learning_rate": 2.48192616120322e-07, + "loss": 0.1413, + "step": 8427 + }, + { + "epoch": 0.44, + "learning_rate": 2.4815980024186585e-07, + "loss": 0.1539, + "step": 8428 + }, + { + "epoch": 0.44, + "learning_rate": 2.4812698298698396e-07, + "loss": 0.1504, + "step": 8429 + }, + { + "epoch": 0.44, + "learning_rate": 2.480941643566144e-07, + "loss": 0.1803, + "step": 8430 + }, + { + "epoch": 0.44, + "learning_rate": 2.48061344351695e-07, + "loss": 0.1434, + "step": 8431 + }, + { + "epoch": 0.44, + "learning_rate": 2.4802852297316385e-07, + "loss": 0.132, + "step": 8432 + }, + { + "epoch": 0.44, + "learning_rate": 2.479957002219589e-07, + "loss": 0.1279, + "step": 8433 + }, + { + "epoch": 0.44, + "learning_rate": 2.4796287609901837e-07, + "loss": 0.1583, + "step": 8434 + }, + { + "epoch": 0.44, + "learning_rate": 2.479300506052803e-07, + "loss": 0.1383, + "step": 8435 + }, + { + "epoch": 0.44, + "learning_rate": 2.478972237416829e-07, + "loss": 0.1435, + "step": 8436 + }, + { + "epoch": 0.44, + "learning_rate": 2.478643955091644e-07, + "loss": 0.1461, + "step": 8437 + }, + { + "epoch": 0.44, + "learning_rate": 2.4783156590866293e-07, + "loss": 0.1166, + "step": 8438 + }, + { + "epoch": 0.44, + "learning_rate": 2.477987349411169e-07, + "loss": 0.158, + "step": 8439 + }, + { + "epoch": 0.44, + "learning_rate": 2.477659026074645e-07, + "loss": 0.1414, + "step": 8440 + }, + { + "epoch": 0.44, + "learning_rate": 2.4773306890864424e-07, + "loss": 0.1734, + "step": 8441 + }, + { + "epoch": 0.44, + "learning_rate": 2.477002338455944e-07, + "loss": 0.1144, + "step": 8442 + }, + { + "epoch": 0.44, + "learning_rate": 2.4766739741925344e-07, + "loss": 0.1309, + "step": 8443 + }, + { + "epoch": 0.44, + "learning_rate": 2.4763455963055994e-07, + "loss": 0.1236, + "step": 8444 + }, + { + "epoch": 0.44, + "learning_rate": 2.4760172048045224e-07, + "loss": 0.1411, + "step": 8445 + }, + { + "epoch": 0.44, + "learning_rate": 2.475688799698691e-07, + "loss": 0.1474, + "step": 8446 + }, + { + "epoch": 0.44, + "learning_rate": 2.475360380997489e-07, + "loss": 0.1459, + "step": 8447 + }, + { + "epoch": 0.44, + "learning_rate": 2.4750319487103045e-07, + "loss": 0.1331, + "step": 8448 + }, + { + "epoch": 0.44, + "learning_rate": 2.474703502846523e-07, + "loss": 0.096, + "step": 8449 + }, + { + "epoch": 0.44, + "learning_rate": 2.4743750434155327e-07, + "loss": 0.179, + "step": 8450 + }, + { + "epoch": 0.44, + "learning_rate": 2.4740465704267205e-07, + "loss": 0.1503, + "step": 8451 + }, + { + "epoch": 0.44, + "learning_rate": 2.4737180838894733e-07, + "loss": 0.1617, + "step": 8452 + }, + { + "epoch": 0.44, + "learning_rate": 2.4733895838131816e-07, + "loss": 0.1496, + "step": 8453 + }, + { + "epoch": 0.44, + "learning_rate": 2.473061070207232e-07, + "loss": 0.1027, + "step": 8454 + }, + { + "epoch": 0.44, + "learning_rate": 2.472732543081015e-07, + "loss": 0.1443, + "step": 8455 + }, + { + "epoch": 0.44, + "learning_rate": 2.4724040024439193e-07, + "loss": 0.1441, + "step": 8456 + }, + { + "epoch": 0.44, + "learning_rate": 2.472075448305335e-07, + "loss": 0.1853, + "step": 8457 + }, + { + "epoch": 0.44, + "learning_rate": 2.471746880674652e-07, + "loss": 0.1526, + "step": 8458 + }, + { + "epoch": 0.44, + "learning_rate": 2.4714182995612607e-07, + "loss": 0.1102, + "step": 8459 + }, + { + "epoch": 0.44, + "learning_rate": 2.471089704974553e-07, + "loss": 0.1344, + "step": 8460 + }, + { + "epoch": 0.44, + "learning_rate": 2.470761096923919e-07, + "loss": 0.1321, + "step": 8461 + }, + { + "epoch": 0.44, + "learning_rate": 2.4704324754187513e-07, + "loss": 0.1391, + "step": 8462 + }, + { + "epoch": 0.44, + "learning_rate": 2.4701038404684415e-07, + "loss": 0.1259, + "step": 8463 + }, + { + "epoch": 0.44, + "learning_rate": 2.469775192082384e-07, + "loss": 0.1279, + "step": 8464 + }, + { + "epoch": 0.44, + "learning_rate": 2.4694465302699685e-07, + "loss": 0.0987, + "step": 8465 + }, + { + "epoch": 0.44, + "learning_rate": 2.469117855040591e-07, + "loss": 0.1146, + "step": 8466 + }, + { + "epoch": 0.44, + "learning_rate": 2.4687891664036436e-07, + "loss": 0.1582, + "step": 8467 + }, + { + "epoch": 0.44, + "learning_rate": 2.468460464368521e-07, + "loss": 0.1146, + "step": 8468 + }, + { + "epoch": 0.44, + "learning_rate": 2.4681317489446176e-07, + "loss": 0.1153, + "step": 8469 + }, + { + "epoch": 0.44, + "learning_rate": 2.4678030201413277e-07, + "loss": 0.1313, + "step": 8470 + }, + { + "epoch": 0.44, + "learning_rate": 2.467474277968047e-07, + "loss": 0.1402, + "step": 8471 + }, + { + "epoch": 0.44, + "learning_rate": 2.4671455224341706e-07, + "loss": 0.1471, + "step": 8472 + }, + { + "epoch": 0.44, + "learning_rate": 2.466816753549096e-07, + "loss": 0.1272, + "step": 8473 + }, + { + "epoch": 0.44, + "learning_rate": 2.4664879713222177e-07, + "loss": 0.1133, + "step": 8474 + }, + { + "epoch": 0.44, + "learning_rate": 2.466159175762933e-07, + "loss": 0.149, + "step": 8475 + }, + { + "epoch": 0.44, + "learning_rate": 2.465830366880639e-07, + "loss": 0.1721, + "step": 8476 + }, + { + "epoch": 0.44, + "learning_rate": 2.4655015446847335e-07, + "loss": 0.1452, + "step": 8477 + }, + { + "epoch": 0.44, + "learning_rate": 2.4651727091846145e-07, + "loss": 0.1565, + "step": 8478 + }, + { + "epoch": 0.44, + "learning_rate": 2.464843860389679e-07, + "loss": 0.0874, + "step": 8479 + }, + { + "epoch": 0.44, + "learning_rate": 2.4645149983093273e-07, + "loss": 0.0997, + "step": 8480 + }, + { + "epoch": 0.44, + "learning_rate": 2.4641861229529575e-07, + "loss": 0.1086, + "step": 8481 + }, + { + "epoch": 0.44, + "learning_rate": 2.463857234329969e-07, + "loss": 0.1503, + "step": 8482 + }, + { + "epoch": 0.44, + "learning_rate": 2.4635283324497615e-07, + "loss": 0.1438, + "step": 8483 + }, + { + "epoch": 0.44, + "learning_rate": 2.463199417321736e-07, + "loss": 0.1262, + "step": 8484 + }, + { + "epoch": 0.44, + "learning_rate": 2.4628704889552913e-07, + "loss": 0.144, + "step": 8485 + }, + { + "epoch": 0.44, + "learning_rate": 2.46254154735983e-07, + "loss": 0.135, + "step": 8486 + }, + { + "epoch": 0.44, + "learning_rate": 2.462212592544752e-07, + "loss": 0.1561, + "step": 8487 + }, + { + "epoch": 0.44, + "learning_rate": 2.4618836245194607e-07, + "loss": 0.1472, + "step": 8488 + }, + { + "epoch": 0.44, + "learning_rate": 2.4615546432933566e-07, + "loss": 0.1638, + "step": 8489 + }, + { + "epoch": 0.44, + "learning_rate": 2.461225648875843e-07, + "loss": 0.1176, + "step": 8490 + }, + { + "epoch": 0.44, + "learning_rate": 2.4608966412763214e-07, + "loss": 0.1146, + "step": 8491 + }, + { + "epoch": 0.44, + "learning_rate": 2.4605676205041966e-07, + "loss": 0.1286, + "step": 8492 + }, + { + "epoch": 0.44, + "learning_rate": 2.460238586568872e-07, + "loss": 0.1588, + "step": 8493 + }, + { + "epoch": 0.44, + "learning_rate": 2.45990953947975e-07, + "loss": 0.1211, + "step": 8494 + }, + { + "epoch": 0.44, + "learning_rate": 2.4595804792462357e-07, + "loss": 0.1426, + "step": 8495 + }, + { + "epoch": 0.44, + "learning_rate": 2.459251405877734e-07, + "loss": 0.1348, + "step": 8496 + }, + { + "epoch": 0.44, + "learning_rate": 2.4589223193836505e-07, + "loss": 0.1658, + "step": 8497 + }, + { + "epoch": 0.44, + "learning_rate": 2.4585932197733894e-07, + "loss": 0.1544, + "step": 8498 + }, + { + "epoch": 0.44, + "learning_rate": 2.458264107056357e-07, + "loss": 0.1241, + "step": 8499 + }, + { + "epoch": 0.44, + "learning_rate": 2.4579349812419596e-07, + "loss": 0.1055, + "step": 8500 + }, + { + "epoch": 0.44, + "learning_rate": 2.4576058423396035e-07, + "loss": 0.1308, + "step": 8501 + }, + { + "epoch": 0.44, + "learning_rate": 2.4572766903586955e-07, + "loss": 0.1028, + "step": 8502 + }, + { + "epoch": 0.44, + "learning_rate": 2.456947525308644e-07, + "loss": 0.1289, + "step": 8503 + }, + { + "epoch": 0.44, + "learning_rate": 2.456618347198856e-07, + "loss": 0.1479, + "step": 8504 + }, + { + "epoch": 0.44, + "learning_rate": 2.456289156038738e-07, + "loss": 0.1546, + "step": 8505 + }, + { + "epoch": 0.44, + "learning_rate": 2.455959951837701e-07, + "loss": 0.1866, + "step": 8506 + }, + { + "epoch": 0.44, + "learning_rate": 2.4556307346051526e-07, + "loss": 0.1414, + "step": 8507 + }, + { + "epoch": 0.44, + "learning_rate": 2.4553015043505006e-07, + "loss": 0.1696, + "step": 8508 + }, + { + "epoch": 0.44, + "learning_rate": 2.454972261083157e-07, + "loss": 0.1491, + "step": 8509 + }, + { + "epoch": 0.44, + "learning_rate": 2.45464300481253e-07, + "loss": 0.1263, + "step": 8510 + }, + { + "epoch": 0.44, + "learning_rate": 2.4543137355480314e-07, + "loss": 0.1315, + "step": 8511 + }, + { + "epoch": 0.44, + "learning_rate": 2.4539844532990693e-07, + "loss": 0.1254, + "step": 8512 + }, + { + "epoch": 0.44, + "learning_rate": 2.453655158075058e-07, + "loss": 0.1726, + "step": 8513 + }, + { + "epoch": 0.44, + "learning_rate": 2.4533258498854054e-07, + "loss": 0.1866, + "step": 8514 + }, + { + "epoch": 0.44, + "learning_rate": 2.452996528739526e-07, + "loss": 0.1697, + "step": 8515 + }, + { + "epoch": 0.44, + "learning_rate": 2.452667194646831e-07, + "loss": 0.1492, + "step": 8516 + }, + { + "epoch": 0.44, + "learning_rate": 2.4523378476167325e-07, + "loss": 0.1275, + "step": 8517 + }, + { + "epoch": 0.44, + "learning_rate": 2.452008487658644e-07, + "loss": 0.1214, + "step": 8518 + }, + { + "epoch": 0.44, + "learning_rate": 2.4516791147819775e-07, + "loss": 0.091, + "step": 8519 + }, + { + "epoch": 0.44, + "learning_rate": 2.451349728996149e-07, + "loss": 0.1588, + "step": 8520 + }, + { + "epoch": 0.44, + "learning_rate": 2.45102033031057e-07, + "loss": 0.1084, + "step": 8521 + }, + { + "epoch": 0.44, + "learning_rate": 2.450690918734656e-07, + "loss": 0.1403, + "step": 8522 + }, + { + "epoch": 0.44, + "learning_rate": 2.4503614942778217e-07, + "loss": 0.1084, + "step": 8523 + }, + { + "epoch": 0.44, + "learning_rate": 2.450032056949482e-07, + "loss": 0.1492, + "step": 8524 + }, + { + "epoch": 0.44, + "learning_rate": 2.4497026067590523e-07, + "loss": 0.1432, + "step": 8525 + }, + { + "epoch": 0.45, + "learning_rate": 2.4493731437159486e-07, + "loss": 0.1617, + "step": 8526 + }, + { + "epoch": 0.45, + "learning_rate": 2.449043667829587e-07, + "loss": 0.1352, + "step": 8527 + }, + { + "epoch": 0.45, + "learning_rate": 2.4487141791093846e-07, + "loss": 0.1806, + "step": 8528 + }, + { + "epoch": 0.45, + "learning_rate": 2.448384677564757e-07, + "loss": 0.1361, + "step": 8529 + }, + { + "epoch": 0.45, + "learning_rate": 2.448055163205122e-07, + "loss": 0.1242, + "step": 8530 + }, + { + "epoch": 0.45, + "learning_rate": 2.4477256360398983e-07, + "loss": 0.1174, + "step": 8531 + }, + { + "epoch": 0.45, + "learning_rate": 2.447396096078502e-07, + "loss": 0.1344, + "step": 8532 + }, + { + "epoch": 0.45, + "learning_rate": 2.4470665433303537e-07, + "loss": 0.1844, + "step": 8533 + }, + { + "epoch": 0.45, + "learning_rate": 2.44673697780487e-07, + "loss": 0.1067, + "step": 8534 + }, + { + "epoch": 0.45, + "learning_rate": 2.446407399511471e-07, + "loss": 0.1339, + "step": 8535 + }, + { + "epoch": 0.45, + "learning_rate": 2.446077808459577e-07, + "loss": 0.1518, + "step": 8536 + }, + { + "epoch": 0.45, + "learning_rate": 2.445748204658607e-07, + "loss": 0.1352, + "step": 8537 + }, + { + "epoch": 0.45, + "learning_rate": 2.445418588117981e-07, + "loss": 0.1086, + "step": 8538 + }, + { + "epoch": 0.45, + "learning_rate": 2.4450889588471196e-07, + "loss": 0.1359, + "step": 8539 + }, + { + "epoch": 0.45, + "learning_rate": 2.444759316855444e-07, + "loss": 0.1555, + "step": 8540 + }, + { + "epoch": 0.45, + "learning_rate": 2.444429662152376e-07, + "loss": 0.1334, + "step": 8541 + }, + { + "epoch": 0.45, + "learning_rate": 2.444099994747336e-07, + "loss": 0.1442, + "step": 8542 + }, + { + "epoch": 0.45, + "learning_rate": 2.4437703146497466e-07, + "loss": 0.1252, + "step": 8543 + }, + { + "epoch": 0.45, + "learning_rate": 2.443440621869031e-07, + "loss": 0.1241, + "step": 8544 + }, + { + "epoch": 0.45, + "learning_rate": 2.443110916414611e-07, + "loss": 0.1061, + "step": 8545 + }, + { + "epoch": 0.45, + "learning_rate": 2.4427811982959097e-07, + "loss": 0.1696, + "step": 8546 + }, + { + "epoch": 0.45, + "learning_rate": 2.442451467522351e-07, + "loss": 0.1501, + "step": 8547 + }, + { + "epoch": 0.45, + "learning_rate": 2.4421217241033577e-07, + "loss": 0.1586, + "step": 8548 + }, + { + "epoch": 0.45, + "learning_rate": 2.4417919680483557e-07, + "loss": 0.0997, + "step": 8549 + }, + { + "epoch": 0.45, + "learning_rate": 2.441462199366769e-07, + "loss": 0.1566, + "step": 8550 + }, + { + "epoch": 0.45, + "learning_rate": 2.441132418068022e-07, + "loss": 0.1271, + "step": 8551 + }, + { + "epoch": 0.45, + "learning_rate": 2.44080262416154e-07, + "loss": 0.1424, + "step": 8552 + }, + { + "epoch": 0.45, + "learning_rate": 2.440472817656749e-07, + "loss": 0.1209, + "step": 8553 + }, + { + "epoch": 0.45, + "learning_rate": 2.440142998563075e-07, + "loss": 0.1304, + "step": 8554 + }, + { + "epoch": 0.45, + "learning_rate": 2.4398131668899433e-07, + "loss": 0.1604, + "step": 8555 + }, + { + "epoch": 0.45, + "learning_rate": 2.439483322646782e-07, + "loss": 0.1801, + "step": 8556 + }, + { + "epoch": 0.45, + "learning_rate": 2.4391534658430173e-07, + "loss": 0.1218, + "step": 8557 + }, + { + "epoch": 0.45, + "learning_rate": 2.438823596488079e-07, + "loss": 0.1179, + "step": 8558 + }, + { + "epoch": 0.45, + "learning_rate": 2.438493714591391e-07, + "loss": 0.175, + "step": 8559 + }, + { + "epoch": 0.45, + "learning_rate": 2.4381638201623834e-07, + "loss": 0.1145, + "step": 8560 + }, + { + "epoch": 0.45, + "learning_rate": 2.4378339132104846e-07, + "loss": 0.1248, + "step": 8561 + }, + { + "epoch": 0.45, + "learning_rate": 2.437503993745124e-07, + "loss": 0.127, + "step": 8562 + }, + { + "epoch": 0.45, + "learning_rate": 2.437174061775731e-07, + "loss": 0.1316, + "step": 8563 + }, + { + "epoch": 0.45, + "learning_rate": 2.4368441173117333e-07, + "loss": 0.1238, + "step": 8564 + }, + { + "epoch": 0.45, + "learning_rate": 2.436514160362563e-07, + "loss": 0.1452, + "step": 8565 + }, + { + "epoch": 0.45, + "learning_rate": 2.436184190937649e-07, + "loss": 0.1235, + "step": 8566 + }, + { + "epoch": 0.45, + "learning_rate": 2.4358542090464234e-07, + "loss": 0.1135, + "step": 8567 + }, + { + "epoch": 0.45, + "learning_rate": 2.435524214698315e-07, + "loss": 0.1231, + "step": 8568 + }, + { + "epoch": 0.45, + "learning_rate": 2.4351942079027566e-07, + "loss": 0.1381, + "step": 8569 + }, + { + "epoch": 0.45, + "learning_rate": 2.43486418866918e-07, + "loss": 0.1278, + "step": 8570 + }, + { + "epoch": 0.45, + "learning_rate": 2.4345341570070173e-07, + "loss": 0.1278, + "step": 8571 + }, + { + "epoch": 0.45, + "learning_rate": 2.4342041129257005e-07, + "loss": 0.1342, + "step": 8572 + }, + { + "epoch": 0.45, + "learning_rate": 2.4338740564346623e-07, + "loss": 0.1436, + "step": 8573 + }, + { + "epoch": 0.45, + "learning_rate": 2.433543987543336e-07, + "loss": 0.1541, + "step": 8574 + }, + { + "epoch": 0.45, + "learning_rate": 2.433213906261156e-07, + "loss": 0.1071, + "step": 8575 + }, + { + "epoch": 0.45, + "learning_rate": 2.432883812597554e-07, + "loss": 0.1563, + "step": 8576 + }, + { + "epoch": 0.45, + "learning_rate": 2.432553706561966e-07, + "loss": 0.1423, + "step": 8577 + }, + { + "epoch": 0.45, + "learning_rate": 2.4322235881638264e-07, + "loss": 0.1095, + "step": 8578 + }, + { + "epoch": 0.45, + "learning_rate": 2.4318934574125696e-07, + "loss": 0.1359, + "step": 8579 + }, + { + "epoch": 0.45, + "learning_rate": 2.4315633143176313e-07, + "loss": 0.1712, + "step": 8580 + }, + { + "epoch": 0.45, + "learning_rate": 2.431233158888447e-07, + "loss": 0.1393, + "step": 8581 + }, + { + "epoch": 0.45, + "learning_rate": 2.430902991134452e-07, + "loss": 0.1623, + "step": 8582 + }, + { + "epoch": 0.45, + "learning_rate": 2.4305728110650835e-07, + "loss": 0.1378, + "step": 8583 + }, + { + "epoch": 0.45, + "learning_rate": 2.430242618689778e-07, + "loss": 0.1656, + "step": 8584 + }, + { + "epoch": 0.45, + "learning_rate": 2.429912414017972e-07, + "loss": 0.1075, + "step": 8585 + }, + { + "epoch": 0.45, + "learning_rate": 2.429582197059103e-07, + "loss": 0.1642, + "step": 8586 + }, + { + "epoch": 0.45, + "learning_rate": 2.42925196782261e-07, + "loss": 0.177, + "step": 8587 + }, + { + "epoch": 0.45, + "learning_rate": 2.428921726317929e-07, + "loss": 0.1187, + "step": 8588 + }, + { + "epoch": 0.45, + "learning_rate": 2.4285914725544996e-07, + "loss": 0.1582, + "step": 8589 + }, + { + "epoch": 0.45, + "learning_rate": 2.428261206541761e-07, + "loss": 0.1277, + "step": 8590 + }, + { + "epoch": 0.45, + "learning_rate": 2.4279309282891514e-07, + "loss": 0.145, + "step": 8591 + }, + { + "epoch": 0.45, + "learning_rate": 2.427600637806111e-07, + "loss": 0.1192, + "step": 8592 + }, + { + "epoch": 0.45, + "learning_rate": 2.4272703351020795e-07, + "loss": 0.1577, + "step": 8593 + }, + { + "epoch": 0.45, + "learning_rate": 2.426940020186496e-07, + "loss": 0.1459, + "step": 8594 + }, + { + "epoch": 0.45, + "learning_rate": 2.4266096930688025e-07, + "loss": 0.1155, + "step": 8595 + }, + { + "epoch": 0.45, + "learning_rate": 2.4262793537584395e-07, + "loss": 0.1576, + "step": 8596 + }, + { + "epoch": 0.45, + "learning_rate": 2.4259490022648474e-07, + "loss": 0.1687, + "step": 8597 + }, + { + "epoch": 0.45, + "learning_rate": 2.425618638597469e-07, + "loss": 0.1345, + "step": 8598 + }, + { + "epoch": 0.45, + "learning_rate": 2.4252882627657454e-07, + "loss": 0.1293, + "step": 8599 + }, + { + "epoch": 0.45, + "learning_rate": 2.424957874779119e-07, + "loss": 0.1452, + "step": 8600 + }, + { + "epoch": 0.45, + "learning_rate": 2.424627474647032e-07, + "loss": 0.1359, + "step": 8601 + }, + { + "epoch": 0.45, + "learning_rate": 2.4242970623789283e-07, + "loss": 0.1138, + "step": 8602 + }, + { + "epoch": 0.45, + "learning_rate": 2.423966637984251e-07, + "loss": 0.163, + "step": 8603 + }, + { + "epoch": 0.45, + "learning_rate": 2.423636201472443e-07, + "loss": 0.1348, + "step": 8604 + }, + { + "epoch": 0.45, + "learning_rate": 2.42330575285295e-07, + "loss": 0.1125, + "step": 8605 + }, + { + "epoch": 0.45, + "learning_rate": 2.422975292135213e-07, + "loss": 0.1412, + "step": 8606 + }, + { + "epoch": 0.45, + "learning_rate": 2.422644819328681e-07, + "loss": 0.146, + "step": 8607 + }, + { + "epoch": 0.45, + "learning_rate": 2.4223143344427956e-07, + "loss": 0.13, + "step": 8608 + }, + { + "epoch": 0.45, + "learning_rate": 2.421983837487004e-07, + "loss": 0.1351, + "step": 8609 + }, + { + "epoch": 0.45, + "learning_rate": 2.4216533284707515e-07, + "loss": 0.1334, + "step": 8610 + }, + { + "epoch": 0.45, + "learning_rate": 2.421322807403483e-07, + "loss": 0.1295, + "step": 8611 + }, + { + "epoch": 0.45, + "learning_rate": 2.420992274294648e-07, + "loss": 0.1321, + "step": 8612 + }, + { + "epoch": 0.45, + "learning_rate": 2.4206617291536895e-07, + "loss": 0.1747, + "step": 8613 + }, + { + "epoch": 0.45, + "learning_rate": 2.420331171990058e-07, + "loss": 0.1627, + "step": 8614 + }, + { + "epoch": 0.45, + "learning_rate": 2.4200006028131984e-07, + "loss": 0.1225, + "step": 8615 + }, + { + "epoch": 0.45, + "learning_rate": 2.4196700216325593e-07, + "loss": 0.1573, + "step": 8616 + }, + { + "epoch": 0.45, + "learning_rate": 2.4193394284575885e-07, + "loss": 0.1659, + "step": 8617 + }, + { + "epoch": 0.45, + "learning_rate": 2.4190088232977363e-07, + "loss": 0.1249, + "step": 8618 + }, + { + "epoch": 0.45, + "learning_rate": 2.4186782061624497e-07, + "loss": 0.1248, + "step": 8619 + }, + { + "epoch": 0.45, + "learning_rate": 2.418347577061178e-07, + "loss": 0.1568, + "step": 8620 + }, + { + "epoch": 0.45, + "learning_rate": 2.4180169360033715e-07, + "loss": 0.1322, + "step": 8621 + }, + { + "epoch": 0.45, + "learning_rate": 2.4176862829984784e-07, + "loss": 0.1235, + "step": 8622 + }, + { + "epoch": 0.45, + "learning_rate": 2.417355618055952e-07, + "loss": 0.1357, + "step": 8623 + }, + { + "epoch": 0.45, + "learning_rate": 2.4170249411852394e-07, + "loss": 0.1135, + "step": 8624 + }, + { + "epoch": 0.45, + "learning_rate": 2.4166942523957935e-07, + "loss": 0.1342, + "step": 8625 + }, + { + "epoch": 0.45, + "learning_rate": 2.416363551697065e-07, + "loss": 0.1511, + "step": 8626 + }, + { + "epoch": 0.45, + "learning_rate": 2.4160328390985054e-07, + "loss": 0.1348, + "step": 8627 + }, + { + "epoch": 0.45, + "learning_rate": 2.415702114609567e-07, + "loss": 0.1453, + "step": 8628 + }, + { + "epoch": 0.45, + "learning_rate": 2.415371378239701e-07, + "loss": 0.1268, + "step": 8629 + }, + { + "epoch": 0.45, + "learning_rate": 2.415040629998361e-07, + "loss": 0.1436, + "step": 8630 + }, + { + "epoch": 0.45, + "learning_rate": 2.414709869895e-07, + "loss": 0.1649, + "step": 8631 + }, + { + "epoch": 0.45, + "learning_rate": 2.414379097939071e-07, + "loss": 0.1586, + "step": 8632 + }, + { + "epoch": 0.45, + "learning_rate": 2.4140483141400263e-07, + "loss": 0.1164, + "step": 8633 + }, + { + "epoch": 0.45, + "learning_rate": 2.4137175185073215e-07, + "loss": 0.1289, + "step": 8634 + }, + { + "epoch": 0.45, + "learning_rate": 2.4133867110504104e-07, + "loss": 0.1226, + "step": 8635 + }, + { + "epoch": 0.45, + "learning_rate": 2.4130558917787477e-07, + "loss": 0.1136, + "step": 8636 + }, + { + "epoch": 0.45, + "learning_rate": 2.4127250607017877e-07, + "loss": 0.1896, + "step": 8637 + }, + { + "epoch": 0.45, + "learning_rate": 2.412394217828986e-07, + "loss": 0.1506, + "step": 8638 + }, + { + "epoch": 0.45, + "learning_rate": 2.4120633631697987e-07, + "loss": 0.1546, + "step": 8639 + }, + { + "epoch": 0.45, + "learning_rate": 2.411732496733682e-07, + "loss": 0.1199, + "step": 8640 + }, + { + "epoch": 0.45, + "learning_rate": 2.4114016185300914e-07, + "loss": 0.1066, + "step": 8641 + }, + { + "epoch": 0.45, + "learning_rate": 2.411070728568483e-07, + "loss": 0.1693, + "step": 8642 + }, + { + "epoch": 0.45, + "learning_rate": 2.410739826858315e-07, + "loss": 0.1403, + "step": 8643 + }, + { + "epoch": 0.45, + "learning_rate": 2.410408913409044e-07, + "loss": 0.1429, + "step": 8644 + }, + { + "epoch": 0.45, + "learning_rate": 2.410077988230127e-07, + "loss": 0.1694, + "step": 8645 + }, + { + "epoch": 0.45, + "learning_rate": 2.4097470513310243e-07, + "loss": 0.1554, + "step": 8646 + }, + { + "epoch": 0.45, + "learning_rate": 2.409416102721192e-07, + "loss": 0.1608, + "step": 8647 + }, + { + "epoch": 0.45, + "learning_rate": 2.4090851424100895e-07, + "loss": 0.1472, + "step": 8648 + }, + { + "epoch": 0.45, + "learning_rate": 2.4087541704071753e-07, + "loss": 0.2004, + "step": 8649 + }, + { + "epoch": 0.45, + "learning_rate": 2.4084231867219097e-07, + "loss": 0.1146, + "step": 8650 + }, + { + "epoch": 0.45, + "learning_rate": 2.4080921913637513e-07, + "loss": 0.1248, + "step": 8651 + }, + { + "epoch": 0.45, + "learning_rate": 2.407761184342161e-07, + "loss": 0.1501, + "step": 8652 + }, + { + "epoch": 0.45, + "learning_rate": 2.407430165666598e-07, + "loss": 0.1281, + "step": 8653 + }, + { + "epoch": 0.45, + "learning_rate": 2.407099135346525e-07, + "loss": 0.1284, + "step": 8654 + }, + { + "epoch": 0.45, + "learning_rate": 2.4067680933914e-07, + "loss": 0.1736, + "step": 8655 + }, + { + "epoch": 0.45, + "learning_rate": 2.406437039810686e-07, + "loss": 0.1498, + "step": 8656 + }, + { + "epoch": 0.45, + "learning_rate": 2.4061059746138454e-07, + "loss": 0.1295, + "step": 8657 + }, + { + "epoch": 0.45, + "learning_rate": 2.4057748978103383e-07, + "loss": 0.1462, + "step": 8658 + }, + { + "epoch": 0.45, + "learning_rate": 2.4054438094096283e-07, + "loss": 0.116, + "step": 8659 + }, + { + "epoch": 0.45, + "learning_rate": 2.405112709421178e-07, + "loss": 0.1218, + "step": 8660 + }, + { + "epoch": 0.45, + "learning_rate": 2.4047815978544506e-07, + "loss": 0.1907, + "step": 8661 + }, + { + "epoch": 0.45, + "learning_rate": 2.404450474718908e-07, + "loss": 0.1183, + "step": 8662 + }, + { + "epoch": 0.45, + "learning_rate": 2.404119340024015e-07, + "loss": 0.1408, + "step": 8663 + }, + { + "epoch": 0.45, + "learning_rate": 2.403788193779235e-07, + "loss": 0.1938, + "step": 8664 + }, + { + "epoch": 0.45, + "learning_rate": 2.4034570359940327e-07, + "loss": 0.1619, + "step": 8665 + }, + { + "epoch": 0.45, + "learning_rate": 2.4031258666778725e-07, + "loss": 0.1712, + "step": 8666 + }, + { + "epoch": 0.45, + "learning_rate": 2.402794685840219e-07, + "loss": 0.1167, + "step": 8667 + }, + { + "epoch": 0.45, + "learning_rate": 2.402463493490539e-07, + "loss": 0.1163, + "step": 8668 + }, + { + "epoch": 0.45, + "learning_rate": 2.402132289638296e-07, + "loss": 0.1101, + "step": 8669 + }, + { + "epoch": 0.45, + "learning_rate": 2.4018010742929577e-07, + "loss": 0.1451, + "step": 8670 + }, + { + "epoch": 0.45, + "learning_rate": 2.401469847463989e-07, + "loss": 0.1474, + "step": 8671 + }, + { + "epoch": 0.45, + "learning_rate": 2.4011386091608574e-07, + "loss": 0.1314, + "step": 8672 + }, + { + "epoch": 0.45, + "learning_rate": 2.400807359393029e-07, + "loss": 0.1378, + "step": 8673 + }, + { + "epoch": 0.45, + "learning_rate": 2.400476098169972e-07, + "loss": 0.1547, + "step": 8674 + }, + { + "epoch": 0.45, + "learning_rate": 2.4001448255011534e-07, + "loss": 0.1373, + "step": 8675 + }, + { + "epoch": 0.45, + "learning_rate": 2.3998135413960405e-07, + "loss": 0.1601, + "step": 8676 + }, + { + "epoch": 0.45, + "learning_rate": 2.399482245864104e-07, + "loss": 0.1489, + "step": 8677 + }, + { + "epoch": 0.45, + "learning_rate": 2.3991509389148093e-07, + "loss": 0.1781, + "step": 8678 + }, + { + "epoch": 0.45, + "learning_rate": 2.398819620557627e-07, + "loss": 0.147, + "step": 8679 + }, + { + "epoch": 0.45, + "learning_rate": 2.398488290802026e-07, + "loss": 0.1388, + "step": 8680 + }, + { + "epoch": 0.45, + "learning_rate": 2.3981569496574756e-07, + "loss": 0.1041, + "step": 8681 + }, + { + "epoch": 0.45, + "learning_rate": 2.3978255971334464e-07, + "loss": 0.1305, + "step": 8682 + }, + { + "epoch": 0.45, + "learning_rate": 2.3974942332394075e-07, + "loss": 0.1283, + "step": 8683 + }, + { + "epoch": 0.45, + "learning_rate": 2.3971628579848303e-07, + "loss": 0.123, + "step": 8684 + }, + { + "epoch": 0.45, + "learning_rate": 2.3968314713791853e-07, + "loss": 0.1273, + "step": 8685 + }, + { + "epoch": 0.45, + "learning_rate": 2.396500073431944e-07, + "loss": 0.1906, + "step": 8686 + }, + { + "epoch": 0.45, + "learning_rate": 2.396168664152577e-07, + "loss": 0.1633, + "step": 8687 + }, + { + "epoch": 0.45, + "learning_rate": 2.3958372435505566e-07, + "loss": 0.1343, + "step": 8688 + }, + { + "epoch": 0.45, + "learning_rate": 2.395505811635355e-07, + "loss": 0.1307, + "step": 8689 + }, + { + "epoch": 0.45, + "learning_rate": 2.395174368416445e-07, + "loss": 0.136, + "step": 8690 + }, + { + "epoch": 0.45, + "learning_rate": 2.3948429139032985e-07, + "loss": 0.1484, + "step": 8691 + }, + { + "epoch": 0.45, + "learning_rate": 2.394511448105389e-07, + "loss": 0.1432, + "step": 8692 + }, + { + "epoch": 0.45, + "learning_rate": 2.3941799710321906e-07, + "loss": 0.1628, + "step": 8693 + }, + { + "epoch": 0.45, + "learning_rate": 2.393848482693176e-07, + "loss": 0.106, + "step": 8694 + }, + { + "epoch": 0.45, + "learning_rate": 2.39351698309782e-07, + "loss": 0.1602, + "step": 8695 + }, + { + "epoch": 0.45, + "learning_rate": 2.393185472255597e-07, + "loss": 0.1291, + "step": 8696 + }, + { + "epoch": 0.45, + "learning_rate": 2.392853950175981e-07, + "loss": 0.1045, + "step": 8697 + }, + { + "epoch": 0.45, + "learning_rate": 2.3925224168684473e-07, + "loss": 0.115, + "step": 8698 + }, + { + "epoch": 0.45, + "learning_rate": 2.392190872342472e-07, + "loss": 0.1228, + "step": 8699 + }, + { + "epoch": 0.45, + "learning_rate": 2.3918593166075295e-07, + "loss": 0.1395, + "step": 8700 + }, + { + "epoch": 0.45, + "learning_rate": 2.391527749673097e-07, + "loss": 0.1573, + "step": 8701 + }, + { + "epoch": 0.45, + "learning_rate": 2.3911961715486497e-07, + "loss": 0.1233, + "step": 8702 + }, + { + "epoch": 0.45, + "learning_rate": 2.390864582243665e-07, + "loss": 0.1461, + "step": 8703 + }, + { + "epoch": 0.45, + "learning_rate": 2.39053298176762e-07, + "loss": 0.17, + "step": 8704 + }, + { + "epoch": 0.45, + "learning_rate": 2.3902013701299915e-07, + "loss": 0.1417, + "step": 8705 + }, + { + "epoch": 0.45, + "learning_rate": 2.389869747340257e-07, + "loss": 0.1207, + "step": 8706 + }, + { + "epoch": 0.45, + "learning_rate": 2.389538113407894e-07, + "loss": 0.1399, + "step": 8707 + }, + { + "epoch": 0.45, + "learning_rate": 2.3892064683423834e-07, + "loss": 0.1228, + "step": 8708 + }, + { + "epoch": 0.45, + "learning_rate": 2.3888748121531997e-07, + "loss": 0.1899, + "step": 8709 + }, + { + "epoch": 0.45, + "learning_rate": 2.3885431448498246e-07, + "loss": 0.1296, + "step": 8710 + }, + { + "epoch": 0.45, + "learning_rate": 2.388211466441736e-07, + "loss": 0.1431, + "step": 8711 + }, + { + "epoch": 0.45, + "learning_rate": 2.3878797769384143e-07, + "loss": 0.1433, + "step": 8712 + }, + { + "epoch": 0.45, + "learning_rate": 2.3875480763493384e-07, + "loss": 0.1831, + "step": 8713 + }, + { + "epoch": 0.45, + "learning_rate": 2.3872163646839886e-07, + "loss": 0.1468, + "step": 8714 + }, + { + "epoch": 0.45, + "learning_rate": 2.386884641951846e-07, + "loss": 0.1615, + "step": 8715 + }, + { + "epoch": 0.45, + "learning_rate": 2.386552908162391e-07, + "loss": 0.1273, + "step": 8716 + }, + { + "epoch": 0.45, + "learning_rate": 2.386221163325105e-07, + "loss": 0.1358, + "step": 8717 + }, + { + "epoch": 0.46, + "learning_rate": 2.385889407449469e-07, + "loss": 0.137, + "step": 8718 + }, + { + "epoch": 0.46, + "learning_rate": 2.3855576405449643e-07, + "loss": 0.1228, + "step": 8719 + }, + { + "epoch": 0.46, + "learning_rate": 2.3852258626210736e-07, + "loss": 0.1421, + "step": 8720 + }, + { + "epoch": 0.46, + "learning_rate": 2.3848940736872793e-07, + "loss": 0.1269, + "step": 8721 + }, + { + "epoch": 0.46, + "learning_rate": 2.3845622737530633e-07, + "loss": 0.1208, + "step": 8722 + }, + { + "epoch": 0.46, + "learning_rate": 2.384230462827909e-07, + "loss": 0.1467, + "step": 8723 + }, + { + "epoch": 0.46, + "learning_rate": 2.3838986409213007e-07, + "loss": 0.1267, + "step": 8724 + }, + { + "epoch": 0.46, + "learning_rate": 2.3835668080427202e-07, + "loss": 0.1127, + "step": 8725 + }, + { + "epoch": 0.46, + "learning_rate": 2.3832349642016525e-07, + "loss": 0.1277, + "step": 8726 + }, + { + "epoch": 0.46, + "learning_rate": 2.3829031094075813e-07, + "loss": 0.129, + "step": 8727 + }, + { + "epoch": 0.46, + "learning_rate": 2.3825712436699922e-07, + "loss": 0.1219, + "step": 8728 + }, + { + "epoch": 0.46, + "learning_rate": 2.3822393669983681e-07, + "loss": 0.1379, + "step": 8729 + }, + { + "epoch": 0.46, + "learning_rate": 2.381907479402196e-07, + "loss": 0.1405, + "step": 8730 + }, + { + "epoch": 0.46, + "learning_rate": 2.3815755808909608e-07, + "loss": 0.1438, + "step": 8731 + }, + { + "epoch": 0.46, + "learning_rate": 2.3812436714741479e-07, + "loss": 0.1494, + "step": 8732 + }, + { + "epoch": 0.46, + "learning_rate": 2.380911751161244e-07, + "loss": 0.1587, + "step": 8733 + }, + { + "epoch": 0.46, + "learning_rate": 2.3805798199617352e-07, + "loss": 0.1286, + "step": 8734 + }, + { + "epoch": 0.46, + "learning_rate": 2.3802478778851082e-07, + "loss": 0.1585, + "step": 8735 + }, + { + "epoch": 0.46, + "learning_rate": 2.3799159249408497e-07, + "loss": 0.1261, + "step": 8736 + }, + { + "epoch": 0.46, + "learning_rate": 2.3795839611384477e-07, + "loss": 0.1478, + "step": 8737 + }, + { + "epoch": 0.46, + "learning_rate": 2.3792519864873895e-07, + "loss": 0.1345, + "step": 8738 + }, + { + "epoch": 0.46, + "learning_rate": 2.3789200009971632e-07, + "loss": 0.1399, + "step": 8739 + }, + { + "epoch": 0.46, + "learning_rate": 2.378588004677257e-07, + "loss": 0.1252, + "step": 8740 + }, + { + "epoch": 0.46, + "learning_rate": 2.3782559975371589e-07, + "loss": 0.1542, + "step": 8741 + }, + { + "epoch": 0.46, + "learning_rate": 2.377923979586359e-07, + "loss": 0.134, + "step": 8742 + }, + { + "epoch": 0.46, + "learning_rate": 2.3775919508343457e-07, + "loss": 0.153, + "step": 8743 + }, + { + "epoch": 0.46, + "learning_rate": 2.3772599112906092e-07, + "loss": 0.1052, + "step": 8744 + }, + { + "epoch": 0.46, + "learning_rate": 2.3769278609646374e-07, + "loss": 0.1216, + "step": 8745 + }, + { + "epoch": 0.46, + "learning_rate": 2.3765957998659233e-07, + "loss": 0.1358, + "step": 8746 + }, + { + "epoch": 0.46, + "learning_rate": 2.3762637280039554e-07, + "loss": 0.1038, + "step": 8747 + }, + { + "epoch": 0.46, + "learning_rate": 2.3759316453882245e-07, + "loss": 0.1207, + "step": 8748 + }, + { + "epoch": 0.46, + "learning_rate": 2.3755995520282224e-07, + "loss": 0.1169, + "step": 8749 + }, + { + "epoch": 0.46, + "learning_rate": 2.3752674479334395e-07, + "loss": 0.1513, + "step": 8750 + }, + { + "epoch": 0.46, + "learning_rate": 2.3749353331133693e-07, + "loss": 0.1531, + "step": 8751 + }, + { + "epoch": 0.46, + "learning_rate": 2.3746032075775018e-07, + "loss": 0.1698, + "step": 8752 + }, + { + "epoch": 0.46, + "learning_rate": 2.3742710713353301e-07, + "loss": 0.1198, + "step": 8753 + }, + { + "epoch": 0.46, + "learning_rate": 2.3739389243963463e-07, + "loss": 0.1455, + "step": 8754 + }, + { + "epoch": 0.46, + "learning_rate": 2.3736067667700444e-07, + "loss": 0.1263, + "step": 8755 + }, + { + "epoch": 0.46, + "learning_rate": 2.3732745984659166e-07, + "loss": 0.131, + "step": 8756 + }, + { + "epoch": 0.46, + "learning_rate": 2.372942419493456e-07, + "loss": 0.111, + "step": 8757 + }, + { + "epoch": 0.46, + "learning_rate": 2.372610229862158e-07, + "loss": 0.1446, + "step": 8758 + }, + { + "epoch": 0.46, + "learning_rate": 2.372278029581515e-07, + "loss": 0.1319, + "step": 8759 + }, + { + "epoch": 0.46, + "learning_rate": 2.3719458186610235e-07, + "loss": 0.1164, + "step": 8760 + }, + { + "epoch": 0.46, + "learning_rate": 2.3716135971101758e-07, + "loss": 0.1579, + "step": 8761 + }, + { + "epoch": 0.46, + "learning_rate": 2.3712813649384683e-07, + "loss": 0.1616, + "step": 8762 + }, + { + "epoch": 0.46, + "learning_rate": 2.370949122155396e-07, + "loss": 0.1261, + "step": 8763 + }, + { + "epoch": 0.46, + "learning_rate": 2.370616868770455e-07, + "loss": 0.1538, + "step": 8764 + }, + { + "epoch": 0.46, + "learning_rate": 2.3702846047931405e-07, + "loss": 0.1766, + "step": 8765 + }, + { + "epoch": 0.46, + "learning_rate": 2.369952330232949e-07, + "loss": 0.1196, + "step": 8766 + }, + { + "epoch": 0.46, + "learning_rate": 2.3696200450993767e-07, + "loss": 0.1457, + "step": 8767 + }, + { + "epoch": 0.46, + "learning_rate": 2.3692877494019216e-07, + "loss": 0.1608, + "step": 8768 + }, + { + "epoch": 0.46, + "learning_rate": 2.3689554431500797e-07, + "loss": 0.1644, + "step": 8769 + }, + { + "epoch": 0.46, + "learning_rate": 2.3686231263533484e-07, + "loss": 0.1166, + "step": 8770 + }, + { + "epoch": 0.46, + "learning_rate": 2.3682907990212268e-07, + "loss": 0.1573, + "step": 8771 + }, + { + "epoch": 0.46, + "learning_rate": 2.3679584611632112e-07, + "loss": 0.1576, + "step": 8772 + }, + { + "epoch": 0.46, + "learning_rate": 2.3676261127888012e-07, + "loss": 0.1393, + "step": 8773 + }, + { + "epoch": 0.46, + "learning_rate": 2.3672937539074947e-07, + "loss": 0.1603, + "step": 8774 + }, + { + "epoch": 0.46, + "learning_rate": 2.3669613845287912e-07, + "loss": 0.1332, + "step": 8775 + }, + { + "epoch": 0.46, + "learning_rate": 2.3666290046621894e-07, + "loss": 0.1581, + "step": 8776 + }, + { + "epoch": 0.46, + "learning_rate": 2.366296614317189e-07, + "loss": 0.1164, + "step": 8777 + }, + { + "epoch": 0.46, + "learning_rate": 2.3659642135032904e-07, + "loss": 0.1276, + "step": 8778 + }, + { + "epoch": 0.46, + "learning_rate": 2.3656318022299925e-07, + "loss": 0.1783, + "step": 8779 + }, + { + "epoch": 0.46, + "learning_rate": 2.3652993805067973e-07, + "loss": 0.1139, + "step": 8780 + }, + { + "epoch": 0.46, + "learning_rate": 2.3649669483432047e-07, + "loss": 0.1559, + "step": 8781 + }, + { + "epoch": 0.46, + "learning_rate": 2.3646345057487158e-07, + "loss": 0.1509, + "step": 8782 + }, + { + "epoch": 0.46, + "learning_rate": 2.3643020527328317e-07, + "loss": 0.159, + "step": 8783 + }, + { + "epoch": 0.46, + "learning_rate": 2.3639695893050545e-07, + "loss": 0.1271, + "step": 8784 + }, + { + "epoch": 0.46, + "learning_rate": 2.363637115474886e-07, + "loss": 0.1591, + "step": 8785 + }, + { + "epoch": 0.46, + "learning_rate": 2.363304631251828e-07, + "loss": 0.1121, + "step": 8786 + }, + { + "epoch": 0.46, + "learning_rate": 2.362972136645384e-07, + "loss": 0.1335, + "step": 8787 + }, + { + "epoch": 0.46, + "learning_rate": 2.3626396316650553e-07, + "loss": 0.1265, + "step": 8788 + }, + { + "epoch": 0.46, + "learning_rate": 2.3623071163203465e-07, + "loss": 0.1501, + "step": 8789 + }, + { + "epoch": 0.46, + "learning_rate": 2.3619745906207606e-07, + "loss": 0.161, + "step": 8790 + }, + { + "epoch": 0.46, + "learning_rate": 2.3616420545758014e-07, + "loss": 0.1478, + "step": 8791 + }, + { + "epoch": 0.46, + "learning_rate": 2.3613095081949722e-07, + "loss": 0.1581, + "step": 8792 + }, + { + "epoch": 0.46, + "learning_rate": 2.3609769514877777e-07, + "loss": 0.1177, + "step": 8793 + }, + { + "epoch": 0.46, + "learning_rate": 2.3606443844637227e-07, + "loss": 0.1187, + "step": 8794 + }, + { + "epoch": 0.46, + "learning_rate": 2.3603118071323117e-07, + "loss": 0.1234, + "step": 8795 + }, + { + "epoch": 0.46, + "learning_rate": 2.3599792195030508e-07, + "loss": 0.1314, + "step": 8796 + }, + { + "epoch": 0.46, + "learning_rate": 2.359646621585444e-07, + "loss": 0.197, + "step": 8797 + }, + { + "epoch": 0.46, + "learning_rate": 2.3593140133889992e-07, + "loss": 0.1584, + "step": 8798 + }, + { + "epoch": 0.46, + "learning_rate": 2.3589813949232203e-07, + "loss": 0.1227, + "step": 8799 + }, + { + "epoch": 0.46, + "learning_rate": 2.3586487661976146e-07, + "loss": 0.1309, + "step": 8800 + }, + { + "epoch": 0.46, + "learning_rate": 2.3583161272216884e-07, + "loss": 0.1122, + "step": 8801 + }, + { + "epoch": 0.46, + "learning_rate": 2.357983478004949e-07, + "loss": 0.1481, + "step": 8802 + }, + { + "epoch": 0.46, + "learning_rate": 2.3576508185569044e-07, + "loss": 0.1516, + "step": 8803 + }, + { + "epoch": 0.46, + "learning_rate": 2.3573181488870606e-07, + "loss": 0.1528, + "step": 8804 + }, + { + "epoch": 0.46, + "learning_rate": 2.3569854690049266e-07, + "loss": 0.1156, + "step": 8805 + }, + { + "epoch": 0.46, + "learning_rate": 2.3566527789200096e-07, + "loss": 0.1415, + "step": 8806 + }, + { + "epoch": 0.46, + "learning_rate": 2.3563200786418192e-07, + "loss": 0.1327, + "step": 8807 + }, + { + "epoch": 0.46, + "learning_rate": 2.3559873681798632e-07, + "loss": 0.1224, + "step": 8808 + }, + { + "epoch": 0.46, + "learning_rate": 2.3556546475436505e-07, + "loss": 0.1503, + "step": 8809 + }, + { + "epoch": 0.46, + "learning_rate": 2.3553219167426906e-07, + "loss": 0.1442, + "step": 8810 + }, + { + "epoch": 0.46, + "learning_rate": 2.3549891757864934e-07, + "loss": 0.129, + "step": 8811 + }, + { + "epoch": 0.46, + "learning_rate": 2.3546564246845686e-07, + "loss": 0.1505, + "step": 8812 + }, + { + "epoch": 0.46, + "learning_rate": 2.3543236634464256e-07, + "loss": 0.1001, + "step": 8813 + }, + { + "epoch": 0.46, + "learning_rate": 2.3539908920815761e-07, + "loss": 0.1396, + "step": 8814 + }, + { + "epoch": 0.46, + "learning_rate": 2.3536581105995307e-07, + "loss": 0.1236, + "step": 8815 + }, + { + "epoch": 0.46, + "learning_rate": 2.3533253190097997e-07, + "loss": 0.1421, + "step": 8816 + }, + { + "epoch": 0.46, + "learning_rate": 2.3529925173218942e-07, + "loss": 0.1544, + "step": 8817 + }, + { + "epoch": 0.46, + "learning_rate": 2.3526597055453265e-07, + "loss": 0.1848, + "step": 8818 + }, + { + "epoch": 0.46, + "learning_rate": 2.3523268836896078e-07, + "loss": 0.1411, + "step": 8819 + }, + { + "epoch": 0.46, + "learning_rate": 2.3519940517642518e-07, + "loss": 0.1406, + "step": 8820 + }, + { + "epoch": 0.46, + "learning_rate": 2.3516612097787695e-07, + "loss": 0.1447, + "step": 8821 + }, + { + "epoch": 0.46, + "learning_rate": 2.3513283577426737e-07, + "loss": 0.1456, + "step": 8822 + }, + { + "epoch": 0.46, + "learning_rate": 2.3509954956654777e-07, + "loss": 0.1283, + "step": 8823 + }, + { + "epoch": 0.46, + "learning_rate": 2.3506626235566958e-07, + "loss": 0.174, + "step": 8824 + }, + { + "epoch": 0.46, + "learning_rate": 2.3503297414258402e-07, + "loss": 0.116, + "step": 8825 + }, + { + "epoch": 0.46, + "learning_rate": 2.3499968492824251e-07, + "loss": 0.1451, + "step": 8826 + }, + { + "epoch": 0.46, + "learning_rate": 2.3496639471359653e-07, + "loss": 0.1336, + "step": 8827 + }, + { + "epoch": 0.46, + "learning_rate": 2.349331034995975e-07, + "loss": 0.1186, + "step": 8828 + }, + { + "epoch": 0.46, + "learning_rate": 2.348998112871969e-07, + "loss": 0.1343, + "step": 8829 + }, + { + "epoch": 0.46, + "learning_rate": 2.3486651807734613e-07, + "loss": 0.1302, + "step": 8830 + }, + { + "epoch": 0.46, + "learning_rate": 2.3483322387099692e-07, + "loss": 0.1367, + "step": 8831 + }, + { + "epoch": 0.46, + "learning_rate": 2.3479992866910067e-07, + "loss": 0.1176, + "step": 8832 + }, + { + "epoch": 0.46, + "learning_rate": 2.34766632472609e-07, + "loss": 0.1294, + "step": 8833 + }, + { + "epoch": 0.46, + "learning_rate": 2.347333352824736e-07, + "loss": 0.1172, + "step": 8834 + }, + { + "epoch": 0.46, + "learning_rate": 2.3470003709964602e-07, + "loss": 0.1625, + "step": 8835 + }, + { + "epoch": 0.46, + "learning_rate": 2.3466673792507805e-07, + "loss": 0.1465, + "step": 8836 + }, + { + "epoch": 0.46, + "learning_rate": 2.346334377597213e-07, + "loss": 0.1401, + "step": 8837 + }, + { + "epoch": 0.46, + "learning_rate": 2.3460013660452756e-07, + "loss": 0.1937, + "step": 8838 + }, + { + "epoch": 0.46, + "learning_rate": 2.3456683446044848e-07, + "loss": 0.1586, + "step": 8839 + }, + { + "epoch": 0.46, + "learning_rate": 2.3453353132843596e-07, + "loss": 0.1391, + "step": 8840 + }, + { + "epoch": 0.46, + "learning_rate": 2.345002272094418e-07, + "loss": 0.1551, + "step": 8841 + }, + { + "epoch": 0.46, + "learning_rate": 2.3446692210441781e-07, + "loss": 0.1327, + "step": 8842 + }, + { + "epoch": 0.46, + "learning_rate": 2.3443361601431592e-07, + "loss": 0.1074, + "step": 8843 + }, + { + "epoch": 0.46, + "learning_rate": 2.3440030894008792e-07, + "loss": 0.1481, + "step": 8844 + }, + { + "epoch": 0.46, + "learning_rate": 2.343670008826859e-07, + "loss": 0.1499, + "step": 8845 + }, + { + "epoch": 0.46, + "learning_rate": 2.3433369184306166e-07, + "loss": 0.1341, + "step": 8846 + }, + { + "epoch": 0.46, + "learning_rate": 2.343003818221673e-07, + "loss": 0.1679, + "step": 8847 + }, + { + "epoch": 0.46, + "learning_rate": 2.3426707082095474e-07, + "loss": 0.1543, + "step": 8848 + }, + { + "epoch": 0.46, + "learning_rate": 2.342337588403761e-07, + "loss": 0.144, + "step": 8849 + }, + { + "epoch": 0.46, + "learning_rate": 2.342004458813834e-07, + "loss": 0.1328, + "step": 8850 + }, + { + "epoch": 0.46, + "learning_rate": 2.3416713194492876e-07, + "loss": 0.1187, + "step": 8851 + }, + { + "epoch": 0.46, + "learning_rate": 2.3413381703196432e-07, + "loss": 0.1203, + "step": 8852 + }, + { + "epoch": 0.46, + "learning_rate": 2.3410050114344212e-07, + "loss": 0.107, + "step": 8853 + }, + { + "epoch": 0.46, + "learning_rate": 2.3406718428031458e-07, + "loss": 0.1254, + "step": 8854 + }, + { + "epoch": 0.46, + "learning_rate": 2.340338664435337e-07, + "loss": 0.1009, + "step": 8855 + }, + { + "epoch": 0.46, + "learning_rate": 2.3400054763405178e-07, + "loss": 0.1371, + "step": 8856 + }, + { + "epoch": 0.46, + "learning_rate": 2.3396722785282098e-07, + "loss": 0.1071, + "step": 8857 + }, + { + "epoch": 0.46, + "learning_rate": 2.3393390710079384e-07, + "loss": 0.1538, + "step": 8858 + }, + { + "epoch": 0.46, + "learning_rate": 2.3390058537892247e-07, + "loss": 0.1562, + "step": 8859 + }, + { + "epoch": 0.46, + "learning_rate": 2.3386726268815925e-07, + "loss": 0.1229, + "step": 8860 + }, + { + "epoch": 0.46, + "learning_rate": 2.3383393902945667e-07, + "loss": 0.1296, + "step": 8861 + }, + { + "epoch": 0.46, + "learning_rate": 2.3380061440376702e-07, + "loss": 0.1336, + "step": 8862 + }, + { + "epoch": 0.46, + "learning_rate": 2.3376728881204277e-07, + "loss": 0.1356, + "step": 8863 + }, + { + "epoch": 0.46, + "learning_rate": 2.3373396225523635e-07, + "loss": 0.1294, + "step": 8864 + }, + { + "epoch": 0.46, + "learning_rate": 2.3370063473430028e-07, + "loss": 0.1266, + "step": 8865 + }, + { + "epoch": 0.46, + "learning_rate": 2.33667306250187e-07, + "loss": 0.1332, + "step": 8866 + }, + { + "epoch": 0.46, + "learning_rate": 2.336339768038492e-07, + "loss": 0.1089, + "step": 8867 + }, + { + "epoch": 0.46, + "learning_rate": 2.336006463962393e-07, + "loss": 0.1597, + "step": 8868 + }, + { + "epoch": 0.46, + "learning_rate": 2.3356731502830995e-07, + "loss": 0.14, + "step": 8869 + }, + { + "epoch": 0.46, + "learning_rate": 2.3353398270101383e-07, + "loss": 0.1213, + "step": 8870 + }, + { + "epoch": 0.46, + "learning_rate": 2.3350064941530353e-07, + "loss": 0.1079, + "step": 8871 + }, + { + "epoch": 0.46, + "learning_rate": 2.3346731517213172e-07, + "loss": 0.1321, + "step": 8872 + }, + { + "epoch": 0.46, + "learning_rate": 2.3343397997245108e-07, + "loss": 0.1503, + "step": 8873 + }, + { + "epoch": 0.46, + "learning_rate": 2.3340064381721444e-07, + "loss": 0.142, + "step": 8874 + }, + { + "epoch": 0.46, + "learning_rate": 2.3336730670737452e-07, + "loss": 0.1392, + "step": 8875 + }, + { + "epoch": 0.46, + "learning_rate": 2.333339686438841e-07, + "loss": 0.1472, + "step": 8876 + }, + { + "epoch": 0.46, + "learning_rate": 2.333006296276959e-07, + "loss": 0.1078, + "step": 8877 + }, + { + "epoch": 0.46, + "learning_rate": 2.332672896597629e-07, + "loss": 0.138, + "step": 8878 + }, + { + "epoch": 0.46, + "learning_rate": 2.3323394874103792e-07, + "loss": 0.1461, + "step": 8879 + }, + { + "epoch": 0.46, + "learning_rate": 2.3320060687247384e-07, + "loss": 0.1379, + "step": 8880 + }, + { + "epoch": 0.46, + "learning_rate": 2.3316726405502365e-07, + "loss": 0.1731, + "step": 8881 + }, + { + "epoch": 0.46, + "learning_rate": 2.3313392028964022e-07, + "loss": 0.1152, + "step": 8882 + }, + { + "epoch": 0.46, + "learning_rate": 2.3310057557727655e-07, + "loss": 0.1258, + "step": 8883 + }, + { + "epoch": 0.46, + "learning_rate": 2.3306722991888566e-07, + "loss": 0.1293, + "step": 8884 + }, + { + "epoch": 0.46, + "learning_rate": 2.3303388331542058e-07, + "loss": 0.132, + "step": 8885 + }, + { + "epoch": 0.46, + "learning_rate": 2.3300053576783432e-07, + "loss": 0.1258, + "step": 8886 + }, + { + "epoch": 0.46, + "learning_rate": 2.3296718727708006e-07, + "loss": 0.1173, + "step": 8887 + }, + { + "epoch": 0.46, + "learning_rate": 2.3293383784411085e-07, + "loss": 0.1389, + "step": 8888 + }, + { + "epoch": 0.46, + "learning_rate": 2.329004874698798e-07, + "loss": 0.136, + "step": 8889 + }, + { + "epoch": 0.46, + "learning_rate": 2.3286713615534016e-07, + "loss": 0.1657, + "step": 8890 + }, + { + "epoch": 0.46, + "learning_rate": 2.3283378390144506e-07, + "loss": 0.1739, + "step": 8891 + }, + { + "epoch": 0.46, + "learning_rate": 2.3280043070914778e-07, + "loss": 0.1369, + "step": 8892 + }, + { + "epoch": 0.46, + "learning_rate": 2.3276707657940147e-07, + "loss": 0.1381, + "step": 8893 + }, + { + "epoch": 0.46, + "learning_rate": 2.327337215131595e-07, + "loss": 0.109, + "step": 8894 + }, + { + "epoch": 0.46, + "learning_rate": 2.327003655113751e-07, + "loss": 0.1394, + "step": 8895 + }, + { + "epoch": 0.46, + "learning_rate": 2.3266700857500168e-07, + "loss": 0.137, + "step": 8896 + }, + { + "epoch": 0.46, + "learning_rate": 2.3263365070499255e-07, + "loss": 0.1444, + "step": 8897 + }, + { + "epoch": 0.46, + "learning_rate": 2.32600291902301e-07, + "loss": 0.1359, + "step": 8898 + }, + { + "epoch": 0.46, + "learning_rate": 2.3256693216788057e-07, + "loss": 0.1258, + "step": 8899 + }, + { + "epoch": 0.46, + "learning_rate": 2.3253357150268463e-07, + "loss": 0.1118, + "step": 8900 + }, + { + "epoch": 0.46, + "learning_rate": 2.3250020990766676e-07, + "loss": 0.1425, + "step": 8901 + }, + { + "epoch": 0.46, + "learning_rate": 2.324668473837802e-07, + "loss": 0.1419, + "step": 8902 + }, + { + "epoch": 0.46, + "learning_rate": 2.3243348393197874e-07, + "loss": 0.12, + "step": 8903 + }, + { + "epoch": 0.46, + "learning_rate": 2.3240011955321567e-07, + "loss": 0.1346, + "step": 8904 + }, + { + "epoch": 0.46, + "learning_rate": 2.3236675424844473e-07, + "loss": 0.1467, + "step": 8905 + }, + { + "epoch": 0.46, + "learning_rate": 2.3233338801861948e-07, + "loss": 0.1553, + "step": 8906 + }, + { + "epoch": 0.46, + "learning_rate": 2.3230002086469348e-07, + "loss": 0.1591, + "step": 8907 + }, + { + "epoch": 0.46, + "learning_rate": 2.3226665278762042e-07, + "loss": 0.1117, + "step": 8908 + }, + { + "epoch": 0.47, + "learning_rate": 2.3223328378835395e-07, + "loss": 0.1549, + "step": 8909 + }, + { + "epoch": 0.47, + "learning_rate": 2.3219991386784789e-07, + "loss": 0.1777, + "step": 8910 + }, + { + "epoch": 0.47, + "learning_rate": 2.3216654302705571e-07, + "loss": 0.1456, + "step": 8911 + }, + { + "epoch": 0.47, + "learning_rate": 2.3213317126693139e-07, + "loss": 0.1395, + "step": 8912 + }, + { + "epoch": 0.47, + "learning_rate": 2.3209979858842863e-07, + "loss": 0.1269, + "step": 8913 + }, + { + "epoch": 0.47, + "learning_rate": 2.3206642499250123e-07, + "loss": 0.1346, + "step": 8914 + }, + { + "epoch": 0.47, + "learning_rate": 2.3203305048010303e-07, + "loss": 0.1383, + "step": 8915 + }, + { + "epoch": 0.47, + "learning_rate": 2.3199967505218784e-07, + "loss": 0.1408, + "step": 8916 + }, + { + "epoch": 0.47, + "learning_rate": 2.3196629870970966e-07, + "loss": 0.1386, + "step": 8917 + }, + { + "epoch": 0.47, + "learning_rate": 2.3193292145362227e-07, + "loss": 0.119, + "step": 8918 + }, + { + "epoch": 0.47, + "learning_rate": 2.318995432848797e-07, + "loss": 0.1419, + "step": 8919 + }, + { + "epoch": 0.47, + "learning_rate": 2.318661642044358e-07, + "loss": 0.1294, + "step": 8920 + }, + { + "epoch": 0.47, + "learning_rate": 2.318327842132447e-07, + "loss": 0.1147, + "step": 8921 + }, + { + "epoch": 0.47, + "learning_rate": 2.317994033122603e-07, + "loss": 0.1102, + "step": 8922 + }, + { + "epoch": 0.47, + "learning_rate": 2.3176602150243663e-07, + "loss": 0.131, + "step": 8923 + }, + { + "epoch": 0.47, + "learning_rate": 2.3173263878472792e-07, + "loss": 0.1562, + "step": 8924 + }, + { + "epoch": 0.47, + "learning_rate": 2.3169925516008802e-07, + "loss": 0.1447, + "step": 8925 + }, + { + "epoch": 0.47, + "learning_rate": 2.3166587062947122e-07, + "loss": 0.1636, + "step": 8926 + }, + { + "epoch": 0.47, + "learning_rate": 2.3163248519383166e-07, + "loss": 0.1245, + "step": 8927 + }, + { + "epoch": 0.47, + "learning_rate": 2.3159909885412343e-07, + "loss": 0.1487, + "step": 8928 + }, + { + "epoch": 0.47, + "learning_rate": 2.315657116113007e-07, + "loss": 0.171, + "step": 8929 + }, + { + "epoch": 0.47, + "learning_rate": 2.3153232346631783e-07, + "loss": 0.1568, + "step": 8930 + }, + { + "epoch": 0.47, + "learning_rate": 2.3149893442012892e-07, + "loss": 0.1656, + "step": 8931 + }, + { + "epoch": 0.47, + "learning_rate": 2.3146554447368838e-07, + "loss": 0.135, + "step": 8932 + }, + { + "epoch": 0.47, + "learning_rate": 2.3143215362795034e-07, + "loss": 0.1038, + "step": 8933 + }, + { + "epoch": 0.47, + "learning_rate": 2.3139876188386927e-07, + "loss": 0.1182, + "step": 8934 + }, + { + "epoch": 0.47, + "learning_rate": 2.313653692423995e-07, + "loss": 0.1154, + "step": 8935 + }, + { + "epoch": 0.47, + "learning_rate": 2.313319757044953e-07, + "loss": 0.1534, + "step": 8936 + }, + { + "epoch": 0.47, + "learning_rate": 2.3129858127111116e-07, + "loss": 0.1076, + "step": 8937 + }, + { + "epoch": 0.47, + "learning_rate": 2.3126518594320145e-07, + "loss": 0.1438, + "step": 8938 + }, + { + "epoch": 0.47, + "learning_rate": 2.312317897217207e-07, + "loss": 0.1304, + "step": 8939 + }, + { + "epoch": 0.47, + "learning_rate": 2.3119839260762334e-07, + "loss": 0.1511, + "step": 8940 + }, + { + "epoch": 0.47, + "learning_rate": 2.311649946018639e-07, + "loss": 0.1137, + "step": 8941 + }, + { + "epoch": 0.47, + "learning_rate": 2.311315957053968e-07, + "loss": 0.1195, + "step": 8942 + }, + { + "epoch": 0.47, + "learning_rate": 2.3109819591917677e-07, + "loss": 0.1405, + "step": 8943 + }, + { + "epoch": 0.47, + "learning_rate": 2.3106479524415826e-07, + "loss": 0.1152, + "step": 8944 + }, + { + "epoch": 0.47, + "learning_rate": 2.3103139368129586e-07, + "loss": 0.19, + "step": 8945 + }, + { + "epoch": 0.47, + "learning_rate": 2.3099799123154433e-07, + "loss": 0.1481, + "step": 8946 + }, + { + "epoch": 0.47, + "learning_rate": 2.3096458789585816e-07, + "loss": 0.1514, + "step": 8947 + }, + { + "epoch": 0.47, + "learning_rate": 2.3093118367519224e-07, + "loss": 0.1558, + "step": 8948 + }, + { + "epoch": 0.47, + "learning_rate": 2.3089777857050104e-07, + "loss": 0.1982, + "step": 8949 + }, + { + "epoch": 0.47, + "learning_rate": 2.3086437258273945e-07, + "loss": 0.1384, + "step": 8950 + }, + { + "epoch": 0.47, + "learning_rate": 2.3083096571286212e-07, + "loss": 0.1131, + "step": 8951 + }, + { + "epoch": 0.47, + "learning_rate": 2.3079755796182394e-07, + "loss": 0.1252, + "step": 8952 + }, + { + "epoch": 0.47, + "learning_rate": 2.3076414933057967e-07, + "loss": 0.1134, + "step": 8953 + }, + { + "epoch": 0.47, + "learning_rate": 2.3073073982008407e-07, + "loss": 0.1304, + "step": 8954 + }, + { + "epoch": 0.47, + "learning_rate": 2.3069732943129218e-07, + "loss": 0.1662, + "step": 8955 + }, + { + "epoch": 0.47, + "learning_rate": 2.3066391816515861e-07, + "loss": 0.1424, + "step": 8956 + }, + { + "epoch": 0.47, + "learning_rate": 2.306305060226386e-07, + "loss": 0.1141, + "step": 8957 + }, + { + "epoch": 0.47, + "learning_rate": 2.3059709300468677e-07, + "loss": 0.1708, + "step": 8958 + }, + { + "epoch": 0.47, + "learning_rate": 2.3056367911225828e-07, + "loss": 0.1516, + "step": 8959 + }, + { + "epoch": 0.47, + "learning_rate": 2.3053026434630798e-07, + "loss": 0.1395, + "step": 8960 + }, + { + "epoch": 0.47, + "learning_rate": 2.3049684870779096e-07, + "loss": 0.1708, + "step": 8961 + }, + { + "epoch": 0.47, + "learning_rate": 2.3046343219766225e-07, + "loss": 0.128, + "step": 8962 + }, + { + "epoch": 0.47, + "learning_rate": 2.3043001481687688e-07, + "loss": 0.1421, + "step": 8963 + }, + { + "epoch": 0.47, + "learning_rate": 2.3039659656638998e-07, + "loss": 0.1705, + "step": 8964 + }, + { + "epoch": 0.47, + "learning_rate": 2.3036317744715657e-07, + "loss": 0.1326, + "step": 8965 + }, + { + "epoch": 0.47, + "learning_rate": 2.3032975746013185e-07, + "loss": 0.2007, + "step": 8966 + }, + { + "epoch": 0.47, + "learning_rate": 2.3029633660627091e-07, + "loss": 0.1523, + "step": 8967 + }, + { + "epoch": 0.47, + "learning_rate": 2.3026291488652902e-07, + "loss": 0.1431, + "step": 8968 + }, + { + "epoch": 0.47, + "learning_rate": 2.302294923018613e-07, + "loss": 0.1039, + "step": 8969 + }, + { + "epoch": 0.47, + "learning_rate": 2.3019606885322305e-07, + "loss": 0.1411, + "step": 8970 + }, + { + "epoch": 0.47, + "learning_rate": 2.3016264454156947e-07, + "loss": 0.128, + "step": 8971 + }, + { + "epoch": 0.47, + "learning_rate": 2.3012921936785586e-07, + "loss": 0.1686, + "step": 8972 + }, + { + "epoch": 0.47, + "learning_rate": 2.3009579333303754e-07, + "loss": 0.1591, + "step": 8973 + }, + { + "epoch": 0.47, + "learning_rate": 2.3006236643806987e-07, + "loss": 0.12, + "step": 8974 + }, + { + "epoch": 0.47, + "learning_rate": 2.3002893868390813e-07, + "loss": 0.1479, + "step": 8975 + }, + { + "epoch": 0.47, + "learning_rate": 2.2999551007150773e-07, + "loss": 0.1524, + "step": 8976 + }, + { + "epoch": 0.47, + "learning_rate": 2.2996208060182409e-07, + "loss": 0.1674, + "step": 8977 + }, + { + "epoch": 0.47, + "learning_rate": 2.299286502758126e-07, + "loss": 0.1139, + "step": 8978 + }, + { + "epoch": 0.47, + "learning_rate": 2.2989521909442872e-07, + "loss": 0.1257, + "step": 8979 + }, + { + "epoch": 0.47, + "learning_rate": 2.2986178705862796e-07, + "loss": 0.0952, + "step": 8980 + }, + { + "epoch": 0.47, + "learning_rate": 2.2982835416936575e-07, + "loss": 0.1403, + "step": 8981 + }, + { + "epoch": 0.47, + "learning_rate": 2.2979492042759777e-07, + "loss": 0.1322, + "step": 8982 + }, + { + "epoch": 0.47, + "learning_rate": 2.2976148583427938e-07, + "loss": 0.1233, + "step": 8983 + }, + { + "epoch": 0.47, + "learning_rate": 2.2972805039036625e-07, + "loss": 0.1148, + "step": 8984 + }, + { + "epoch": 0.47, + "learning_rate": 2.2969461409681396e-07, + "loss": 0.101, + "step": 8985 + }, + { + "epoch": 0.47, + "learning_rate": 2.2966117695457816e-07, + "loss": 0.1568, + "step": 8986 + }, + { + "epoch": 0.47, + "learning_rate": 2.2962773896461446e-07, + "loss": 0.1295, + "step": 8987 + }, + { + "epoch": 0.47, + "learning_rate": 2.2959430012787854e-07, + "loss": 0.1054, + "step": 8988 + }, + { + "epoch": 0.47, + "learning_rate": 2.2956086044532608e-07, + "loss": 0.148, + "step": 8989 + }, + { + "epoch": 0.47, + "learning_rate": 2.2952741991791282e-07, + "loss": 0.136, + "step": 8990 + }, + { + "epoch": 0.47, + "learning_rate": 2.2949397854659455e-07, + "loss": 0.128, + "step": 8991 + }, + { + "epoch": 0.47, + "learning_rate": 2.294605363323269e-07, + "loss": 0.1194, + "step": 8992 + }, + { + "epoch": 0.47, + "learning_rate": 2.294270932760658e-07, + "loss": 0.1423, + "step": 8993 + }, + { + "epoch": 0.47, + "learning_rate": 2.2939364937876703e-07, + "loss": 0.1477, + "step": 8994 + }, + { + "epoch": 0.47, + "learning_rate": 2.293602046413864e-07, + "loss": 0.1547, + "step": 8995 + }, + { + "epoch": 0.47, + "learning_rate": 2.2932675906487978e-07, + "loss": 0.1674, + "step": 8996 + }, + { + "epoch": 0.47, + "learning_rate": 2.2929331265020306e-07, + "loss": 0.1622, + "step": 8997 + }, + { + "epoch": 0.47, + "learning_rate": 2.2925986539831218e-07, + "loss": 0.1525, + "step": 8998 + }, + { + "epoch": 0.47, + "learning_rate": 2.29226417310163e-07, + "loss": 0.1448, + "step": 8999 + }, + { + "epoch": 0.47, + "learning_rate": 2.291929683867116e-07, + "loss": 0.1146, + "step": 9000 + }, + { + "epoch": 0.47, + "learning_rate": 2.2915951862891383e-07, + "loss": 0.1092, + "step": 9001 + }, + { + "epoch": 0.47, + "learning_rate": 2.2912606803772578e-07, + "loss": 0.1621, + "step": 9002 + }, + { + "epoch": 0.47, + "learning_rate": 2.2909261661410347e-07, + "loss": 0.1541, + "step": 9003 + }, + { + "epoch": 0.47, + "learning_rate": 2.29059164359003e-07, + "loss": 0.1441, + "step": 9004 + }, + { + "epoch": 0.47, + "learning_rate": 2.290257112733803e-07, + "loss": 0.1406, + "step": 9005 + }, + { + "epoch": 0.47, + "learning_rate": 2.2899225735819162e-07, + "loss": 0.1552, + "step": 9006 + }, + { + "epoch": 0.47, + "learning_rate": 2.2895880261439302e-07, + "loss": 0.1173, + "step": 9007 + }, + { + "epoch": 0.47, + "learning_rate": 2.289253470429407e-07, + "loss": 0.1437, + "step": 9008 + }, + { + "epoch": 0.47, + "learning_rate": 2.2889189064479076e-07, + "loss": 0.1011, + "step": 9009 + }, + { + "epoch": 0.47, + "learning_rate": 2.2885843342089939e-07, + "loss": 0.1383, + "step": 9010 + }, + { + "epoch": 0.47, + "learning_rate": 2.2882497537222289e-07, + "loss": 0.1383, + "step": 9011 + }, + { + "epoch": 0.47, + "learning_rate": 2.2879151649971748e-07, + "loss": 0.1516, + "step": 9012 + }, + { + "epoch": 0.47, + "learning_rate": 2.2875805680433942e-07, + "loss": 0.1269, + "step": 9013 + }, + { + "epoch": 0.47, + "learning_rate": 2.2872459628704496e-07, + "loss": 0.1609, + "step": 9014 + }, + { + "epoch": 0.47, + "learning_rate": 2.286911349487905e-07, + "loss": 0.12, + "step": 9015 + }, + { + "epoch": 0.47, + "learning_rate": 2.2865767279053229e-07, + "loss": 0.1446, + "step": 9016 + }, + { + "epoch": 0.47, + "learning_rate": 2.286242098132267e-07, + "loss": 0.162, + "step": 9017 + }, + { + "epoch": 0.47, + "learning_rate": 2.285907460178302e-07, + "loss": 0.1394, + "step": 9018 + }, + { + "epoch": 0.47, + "learning_rate": 2.2855728140529908e-07, + "loss": 0.1531, + "step": 9019 + }, + { + "epoch": 0.47, + "learning_rate": 2.2852381597658987e-07, + "loss": 0.14, + "step": 9020 + }, + { + "epoch": 0.47, + "learning_rate": 2.2849034973265904e-07, + "loss": 0.1503, + "step": 9021 + }, + { + "epoch": 0.47, + "learning_rate": 2.2845688267446297e-07, + "loss": 0.1621, + "step": 9022 + }, + { + "epoch": 0.47, + "learning_rate": 2.284234148029582e-07, + "loss": 0.1177, + "step": 9023 + }, + { + "epoch": 0.47, + "learning_rate": 2.2838994611910126e-07, + "loss": 0.1378, + "step": 9024 + }, + { + "epoch": 0.47, + "learning_rate": 2.2835647662384873e-07, + "loss": 0.1363, + "step": 9025 + }, + { + "epoch": 0.47, + "learning_rate": 2.2832300631815706e-07, + "loss": 0.1436, + "step": 9026 + }, + { + "epoch": 0.47, + "learning_rate": 2.2828953520298305e-07, + "loss": 0.1724, + "step": 9027 + }, + { + "epoch": 0.47, + "learning_rate": 2.2825606327928311e-07, + "loss": 0.1679, + "step": 9028 + }, + { + "epoch": 0.47, + "learning_rate": 2.2822259054801405e-07, + "loss": 0.118, + "step": 9029 + }, + { + "epoch": 0.47, + "learning_rate": 2.2818911701013243e-07, + "loss": 0.1268, + "step": 9030 + }, + { + "epoch": 0.47, + "learning_rate": 2.2815564266659496e-07, + "loss": 0.1646, + "step": 9031 + }, + { + "epoch": 0.47, + "learning_rate": 2.2812216751835836e-07, + "loss": 0.1398, + "step": 9032 + }, + { + "epoch": 0.47, + "learning_rate": 2.2808869156637934e-07, + "loss": 0.1585, + "step": 9033 + }, + { + "epoch": 0.47, + "learning_rate": 2.2805521481161468e-07, + "loss": 0.1141, + "step": 9034 + }, + { + "epoch": 0.47, + "learning_rate": 2.2802173725502113e-07, + "loss": 0.1312, + "step": 9035 + }, + { + "epoch": 0.47, + "learning_rate": 2.279882588975556e-07, + "loss": 0.1353, + "step": 9036 + }, + { + "epoch": 0.47, + "learning_rate": 2.279547797401747e-07, + "loss": 0.14, + "step": 9037 + }, + { + "epoch": 0.47, + "learning_rate": 2.2792129978383552e-07, + "loss": 0.1861, + "step": 9038 + }, + { + "epoch": 0.47, + "learning_rate": 2.2788781902949473e-07, + "loss": 0.1207, + "step": 9039 + }, + { + "epoch": 0.47, + "learning_rate": 2.2785433747810936e-07, + "loss": 0.0973, + "step": 9040 + }, + { + "epoch": 0.47, + "learning_rate": 2.278208551306362e-07, + "loss": 0.1427, + "step": 9041 + }, + { + "epoch": 0.47, + "learning_rate": 2.2778737198803238e-07, + "loss": 0.1313, + "step": 9042 + }, + { + "epoch": 0.47, + "learning_rate": 2.277538880512546e-07, + "loss": 0.1279, + "step": 9043 + }, + { + "epoch": 0.47, + "learning_rate": 2.2772040332126007e-07, + "loss": 0.1224, + "step": 9044 + }, + { + "epoch": 0.47, + "learning_rate": 2.2768691779900569e-07, + "loss": 0.1288, + "step": 9045 + }, + { + "epoch": 0.47, + "learning_rate": 2.2765343148544854e-07, + "loss": 0.1504, + "step": 9046 + }, + { + "epoch": 0.47, + "learning_rate": 2.2761994438154563e-07, + "loss": 0.1104, + "step": 9047 + }, + { + "epoch": 0.47, + "learning_rate": 2.27586456488254e-07, + "loss": 0.1429, + "step": 9048 + }, + { + "epoch": 0.47, + "learning_rate": 2.2755296780653083e-07, + "loss": 0.1375, + "step": 9049 + }, + { + "epoch": 0.47, + "learning_rate": 2.2751947833733317e-07, + "loss": 0.1577, + "step": 9050 + }, + { + "epoch": 0.47, + "learning_rate": 2.274859880816183e-07, + "loss": 0.1261, + "step": 9051 + }, + { + "epoch": 0.47, + "learning_rate": 2.2745249704034317e-07, + "loss": 0.1245, + "step": 9052 + }, + { + "epoch": 0.47, + "learning_rate": 2.2741900521446517e-07, + "loss": 0.1155, + "step": 9053 + }, + { + "epoch": 0.47, + "learning_rate": 2.2738551260494132e-07, + "loss": 0.1465, + "step": 9054 + }, + { + "epoch": 0.47, + "learning_rate": 2.27352019212729e-07, + "loss": 0.1648, + "step": 9055 + }, + { + "epoch": 0.47, + "learning_rate": 2.2731852503878542e-07, + "loss": 0.1209, + "step": 9056 + }, + { + "epoch": 0.47, + "learning_rate": 2.272850300840678e-07, + "loss": 0.1806, + "step": 9057 + }, + { + "epoch": 0.47, + "learning_rate": 2.2725153434953355e-07, + "loss": 0.159, + "step": 9058 + }, + { + "epoch": 0.47, + "learning_rate": 2.2721803783613986e-07, + "loss": 0.145, + "step": 9059 + }, + { + "epoch": 0.47, + "learning_rate": 2.271845405448442e-07, + "loss": 0.1087, + "step": 9060 + }, + { + "epoch": 0.47, + "learning_rate": 2.2715104247660383e-07, + "loss": 0.1397, + "step": 9061 + }, + { + "epoch": 0.47, + "learning_rate": 2.2711754363237624e-07, + "loss": 0.0993, + "step": 9062 + }, + { + "epoch": 0.47, + "learning_rate": 2.270840440131187e-07, + "loss": 0.1752, + "step": 9063 + }, + { + "epoch": 0.47, + "learning_rate": 2.2705054361978883e-07, + "loss": 0.1199, + "step": 9064 + }, + { + "epoch": 0.47, + "learning_rate": 2.2701704245334397e-07, + "loss": 0.1279, + "step": 9065 + }, + { + "epoch": 0.47, + "learning_rate": 2.2698354051474151e-07, + "loss": 0.1456, + "step": 9066 + }, + { + "epoch": 0.47, + "learning_rate": 2.2695003780493912e-07, + "loss": 0.1549, + "step": 9067 + }, + { + "epoch": 0.47, + "learning_rate": 2.2691653432489425e-07, + "loss": 0.1321, + "step": 9068 + }, + { + "epoch": 0.47, + "learning_rate": 2.2688303007556443e-07, + "loss": 0.1404, + "step": 9069 + }, + { + "epoch": 0.47, + "learning_rate": 2.2684952505790716e-07, + "loss": 0.1416, + "step": 9070 + }, + { + "epoch": 0.47, + "learning_rate": 2.2681601927288021e-07, + "loss": 0.1448, + "step": 9071 + }, + { + "epoch": 0.47, + "learning_rate": 2.2678251272144102e-07, + "loss": 0.1507, + "step": 9072 + }, + { + "epoch": 0.47, + "learning_rate": 2.2674900540454726e-07, + "loss": 0.1195, + "step": 9073 + }, + { + "epoch": 0.47, + "learning_rate": 2.2671549732315665e-07, + "loss": 0.1502, + "step": 9074 + }, + { + "epoch": 0.47, + "learning_rate": 2.2668198847822678e-07, + "loss": 0.1438, + "step": 9075 + }, + { + "epoch": 0.47, + "learning_rate": 2.266484788707154e-07, + "loss": 0.1539, + "step": 9076 + }, + { + "epoch": 0.47, + "learning_rate": 2.2661496850158018e-07, + "loss": 0.1243, + "step": 9077 + }, + { + "epoch": 0.47, + "learning_rate": 2.265814573717789e-07, + "loss": 0.1425, + "step": 9078 + }, + { + "epoch": 0.47, + "learning_rate": 2.2654794548226928e-07, + "loss": 0.1262, + "step": 9079 + }, + { + "epoch": 0.47, + "learning_rate": 2.2651443283400917e-07, + "loss": 0.1231, + "step": 9080 + }, + { + "epoch": 0.47, + "learning_rate": 2.2648091942795635e-07, + "loss": 0.1767, + "step": 9081 + }, + { + "epoch": 0.47, + "learning_rate": 2.2644740526506853e-07, + "loss": 0.1176, + "step": 9082 + }, + { + "epoch": 0.47, + "learning_rate": 2.2641389034630376e-07, + "loss": 0.1152, + "step": 9083 + }, + { + "epoch": 0.47, + "learning_rate": 2.2638037467261976e-07, + "loss": 0.1309, + "step": 9084 + }, + { + "epoch": 0.47, + "learning_rate": 2.2634685824497452e-07, + "loss": 0.1383, + "step": 9085 + }, + { + "epoch": 0.47, + "learning_rate": 2.2631334106432582e-07, + "loss": 0.1254, + "step": 9086 + }, + { + "epoch": 0.47, + "learning_rate": 2.2627982313163173e-07, + "loss": 0.1364, + "step": 9087 + }, + { + "epoch": 0.47, + "learning_rate": 2.262463044478501e-07, + "loss": 0.1107, + "step": 9088 + }, + { + "epoch": 0.47, + "learning_rate": 2.2621278501393907e-07, + "loss": 0.1218, + "step": 9089 + }, + { + "epoch": 0.47, + "learning_rate": 2.2617926483085643e-07, + "loss": 0.1429, + "step": 9090 + }, + { + "epoch": 0.47, + "learning_rate": 2.2614574389956025e-07, + "loss": 0.139, + "step": 9091 + }, + { + "epoch": 0.47, + "learning_rate": 2.261122222210087e-07, + "loss": 0.1483, + "step": 9092 + }, + { + "epoch": 0.47, + "learning_rate": 2.2607869979615971e-07, + "loss": 0.1127, + "step": 9093 + }, + { + "epoch": 0.47, + "learning_rate": 2.260451766259715e-07, + "loss": 0.1401, + "step": 9094 + }, + { + "epoch": 0.47, + "learning_rate": 2.26011652711402e-07, + "loss": 0.1483, + "step": 9095 + }, + { + "epoch": 0.47, + "learning_rate": 2.2597812805340948e-07, + "loss": 0.1309, + "step": 9096 + }, + { + "epoch": 0.47, + "learning_rate": 2.2594460265295193e-07, + "loss": 0.1308, + "step": 9097 + }, + { + "epoch": 0.47, + "learning_rate": 2.2591107651098777e-07, + "loss": 0.1336, + "step": 9098 + }, + { + "epoch": 0.47, + "learning_rate": 2.2587754962847495e-07, + "loss": 0.1354, + "step": 9099 + }, + { + "epoch": 0.47, + "learning_rate": 2.258440220063718e-07, + "loss": 0.1551, + "step": 9100 + }, + { + "epoch": 0.48, + "learning_rate": 2.2581049364563646e-07, + "loss": 0.1619, + "step": 9101 + }, + { + "epoch": 0.48, + "learning_rate": 2.2577696454722736e-07, + "loss": 0.1344, + "step": 9102 + }, + { + "epoch": 0.48, + "learning_rate": 2.257434347121026e-07, + "loss": 0.1453, + "step": 9103 + }, + { + "epoch": 0.48, + "learning_rate": 2.2570990414122048e-07, + "loss": 0.2043, + "step": 9104 + }, + { + "epoch": 0.48, + "learning_rate": 2.256763728355395e-07, + "loss": 0.1465, + "step": 9105 + }, + { + "epoch": 0.48, + "learning_rate": 2.2564284079601772e-07, + "loss": 0.1552, + "step": 9106 + }, + { + "epoch": 0.48, + "learning_rate": 2.256093080236138e-07, + "loss": 0.1554, + "step": 9107 + }, + { + "epoch": 0.48, + "learning_rate": 2.255757745192859e-07, + "loss": 0.1269, + "step": 9108 + }, + { + "epoch": 0.48, + "learning_rate": 2.255422402839925e-07, + "loss": 0.1387, + "step": 9109 + }, + { + "epoch": 0.48, + "learning_rate": 2.2550870531869196e-07, + "loss": 0.1465, + "step": 9110 + }, + { + "epoch": 0.48, + "learning_rate": 2.2547516962434288e-07, + "loss": 0.1115, + "step": 9111 + }, + { + "epoch": 0.48, + "learning_rate": 2.2544163320190357e-07, + "loss": 0.1364, + "step": 9112 + }, + { + "epoch": 0.48, + "learning_rate": 2.2540809605233253e-07, + "loss": 0.1824, + "step": 9113 + }, + { + "epoch": 0.48, + "learning_rate": 2.2537455817658832e-07, + "loss": 0.1251, + "step": 9114 + }, + { + "epoch": 0.48, + "learning_rate": 2.2534101957562948e-07, + "loss": 0.1218, + "step": 9115 + }, + { + "epoch": 0.48, + "learning_rate": 2.2530748025041445e-07, + "loss": 0.1363, + "step": 9116 + }, + { + "epoch": 0.48, + "learning_rate": 2.252739402019019e-07, + "loss": 0.1298, + "step": 9117 + }, + { + "epoch": 0.48, + "learning_rate": 2.2524039943105037e-07, + "loss": 0.1147, + "step": 9118 + }, + { + "epoch": 0.48, + "learning_rate": 2.2520685793881852e-07, + "loss": 0.126, + "step": 9119 + }, + { + "epoch": 0.48, + "learning_rate": 2.2517331572616487e-07, + "loss": 0.1438, + "step": 9120 + }, + { + "epoch": 0.48, + "learning_rate": 2.251397727940482e-07, + "loss": 0.1202, + "step": 9121 + }, + { + "epoch": 0.48, + "learning_rate": 2.2510622914342704e-07, + "loss": 0.1367, + "step": 9122 + }, + { + "epoch": 0.48, + "learning_rate": 2.2507268477526026e-07, + "loss": 0.14, + "step": 9123 + }, + { + "epoch": 0.48, + "learning_rate": 2.2503913969050643e-07, + "loss": 0.1034, + "step": 9124 + }, + { + "epoch": 0.48, + "learning_rate": 2.250055938901243e-07, + "loss": 0.1294, + "step": 9125 + }, + { + "epoch": 0.48, + "learning_rate": 2.2497204737507264e-07, + "loss": 0.1553, + "step": 9126 + }, + { + "epoch": 0.48, + "learning_rate": 2.2493850014631026e-07, + "loss": 0.1584, + "step": 9127 + }, + { + "epoch": 0.48, + "learning_rate": 2.2490495220479592e-07, + "loss": 0.147, + "step": 9128 + }, + { + "epoch": 0.48, + "learning_rate": 2.248714035514884e-07, + "loss": 0.1732, + "step": 9129 + }, + { + "epoch": 0.48, + "learning_rate": 2.2483785418734658e-07, + "loss": 0.1351, + "step": 9130 + }, + { + "epoch": 0.48, + "learning_rate": 2.2480430411332927e-07, + "loss": 0.1543, + "step": 9131 + }, + { + "epoch": 0.48, + "learning_rate": 2.2477075333039547e-07, + "loss": 0.1244, + "step": 9132 + }, + { + "epoch": 0.48, + "learning_rate": 2.2473720183950385e-07, + "loss": 0.1761, + "step": 9133 + }, + { + "epoch": 0.48, + "learning_rate": 2.2470364964161353e-07, + "loss": 0.1473, + "step": 9134 + }, + { + "epoch": 0.48, + "learning_rate": 2.2467009673768334e-07, + "loss": 0.1418, + "step": 9135 + }, + { + "epoch": 0.48, + "learning_rate": 2.2463654312867226e-07, + "loss": 0.1364, + "step": 9136 + }, + { + "epoch": 0.48, + "learning_rate": 2.246029888155393e-07, + "loss": 0.1159, + "step": 9137 + }, + { + "epoch": 0.48, + "learning_rate": 2.245694337992434e-07, + "loss": 0.1265, + "step": 9138 + }, + { + "epoch": 0.48, + "learning_rate": 2.245358780807436e-07, + "loss": 0.1366, + "step": 9139 + }, + { + "epoch": 0.48, + "learning_rate": 2.2450232166099887e-07, + "loss": 0.1445, + "step": 9140 + }, + { + "epoch": 0.48, + "learning_rate": 2.2446876454096846e-07, + "loss": 0.1168, + "step": 9141 + }, + { + "epoch": 0.48, + "learning_rate": 2.2443520672161123e-07, + "loss": 0.1667, + "step": 9142 + }, + { + "epoch": 0.48, + "learning_rate": 2.2440164820388636e-07, + "loss": 0.1226, + "step": 9143 + }, + { + "epoch": 0.48, + "learning_rate": 2.2436808898875293e-07, + "loss": 0.1239, + "step": 9144 + }, + { + "epoch": 0.48, + "learning_rate": 2.2433452907717022e-07, + "loss": 0.1469, + "step": 9145 + }, + { + "epoch": 0.48, + "learning_rate": 2.2430096847009715e-07, + "loss": 0.1183, + "step": 9146 + }, + { + "epoch": 0.48, + "learning_rate": 2.242674071684931e-07, + "loss": 0.1173, + "step": 9147 + }, + { + "epoch": 0.48, + "learning_rate": 2.2423384517331714e-07, + "loss": 0.1398, + "step": 9148 + }, + { + "epoch": 0.48, + "learning_rate": 2.2420028248552855e-07, + "loss": 0.1239, + "step": 9149 + }, + { + "epoch": 0.48, + "learning_rate": 2.2416671910608657e-07, + "loss": 0.1504, + "step": 9150 + }, + { + "epoch": 0.48, + "learning_rate": 2.2413315503595037e-07, + "loss": 0.1911, + "step": 9151 + }, + { + "epoch": 0.48, + "learning_rate": 2.2409959027607935e-07, + "loss": 0.146, + "step": 9152 + }, + { + "epoch": 0.48, + "learning_rate": 2.2406602482743268e-07, + "loss": 0.1346, + "step": 9153 + }, + { + "epoch": 0.48, + "learning_rate": 2.2403245869096982e-07, + "loss": 0.134, + "step": 9154 + }, + { + "epoch": 0.48, + "learning_rate": 2.239988918676499e-07, + "loss": 0.1571, + "step": 9155 + }, + { + "epoch": 0.48, + "learning_rate": 2.2396532435843247e-07, + "loss": 0.1707, + "step": 9156 + }, + { + "epoch": 0.48, + "learning_rate": 2.2393175616427676e-07, + "loss": 0.1198, + "step": 9157 + }, + { + "epoch": 0.48, + "learning_rate": 2.2389818728614224e-07, + "loss": 0.1368, + "step": 9158 + }, + { + "epoch": 0.48, + "learning_rate": 2.2386461772498832e-07, + "loss": 0.1386, + "step": 9159 + }, + { + "epoch": 0.48, + "learning_rate": 2.2383104748177434e-07, + "loss": 0.135, + "step": 9160 + }, + { + "epoch": 0.48, + "learning_rate": 2.2379747655745992e-07, + "loss": 0.1211, + "step": 9161 + }, + { + "epoch": 0.48, + "learning_rate": 2.237639049530044e-07, + "loss": 0.1709, + "step": 9162 + }, + { + "epoch": 0.48, + "learning_rate": 2.2373033266936733e-07, + "loss": 0.1787, + "step": 9163 + }, + { + "epoch": 0.48, + "learning_rate": 2.236967597075081e-07, + "loss": 0.1967, + "step": 9164 + }, + { + "epoch": 0.48, + "learning_rate": 2.2366318606838642e-07, + "loss": 0.1461, + "step": 9165 + }, + { + "epoch": 0.48, + "learning_rate": 2.236296117529617e-07, + "loss": 0.1712, + "step": 9166 + }, + { + "epoch": 0.48, + "learning_rate": 2.2359603676219353e-07, + "loss": 0.1293, + "step": 9167 + }, + { + "epoch": 0.48, + "learning_rate": 2.235624610970416e-07, + "loss": 0.1566, + "step": 9168 + }, + { + "epoch": 0.48, + "learning_rate": 2.2352888475846537e-07, + "loss": 0.1274, + "step": 9169 + }, + { + "epoch": 0.48, + "learning_rate": 2.2349530774742455e-07, + "loss": 0.1437, + "step": 9170 + }, + { + "epoch": 0.48, + "learning_rate": 2.234617300648788e-07, + "loss": 0.1442, + "step": 9171 + }, + { + "epoch": 0.48, + "learning_rate": 2.2342815171178775e-07, + "loss": 0.1092, + "step": 9172 + }, + { + "epoch": 0.48, + "learning_rate": 2.2339457268911102e-07, + "loss": 0.1447, + "step": 9173 + }, + { + "epoch": 0.48, + "learning_rate": 2.2336099299780837e-07, + "loss": 0.1299, + "step": 9174 + }, + { + "epoch": 0.48, + "learning_rate": 2.233274126388396e-07, + "loss": 0.1309, + "step": 9175 + }, + { + "epoch": 0.48, + "learning_rate": 2.2329383161316428e-07, + "loss": 0.1446, + "step": 9176 + }, + { + "epoch": 0.48, + "learning_rate": 2.2326024992174232e-07, + "loss": 0.1294, + "step": 9177 + }, + { + "epoch": 0.48, + "learning_rate": 2.2322666756553338e-07, + "loss": 0.1182, + "step": 9178 + }, + { + "epoch": 0.48, + "learning_rate": 2.2319308454549742e-07, + "loss": 0.1173, + "step": 9179 + }, + { + "epoch": 0.48, + "learning_rate": 2.2315950086259403e-07, + "loss": 0.1082, + "step": 9180 + }, + { + "epoch": 0.48, + "learning_rate": 2.2312591651778327e-07, + "loss": 0.1636, + "step": 9181 + }, + { + "epoch": 0.48, + "learning_rate": 2.2309233151202478e-07, + "loss": 0.1385, + "step": 9182 + }, + { + "epoch": 0.48, + "learning_rate": 2.2305874584627863e-07, + "loss": 0.1608, + "step": 9183 + }, + { + "epoch": 0.48, + "learning_rate": 2.230251595215046e-07, + "loss": 0.1162, + "step": 9184 + }, + { + "epoch": 0.48, + "learning_rate": 2.2299157253866258e-07, + "loss": 0.1356, + "step": 9185 + }, + { + "epoch": 0.48, + "learning_rate": 2.229579848987126e-07, + "loss": 0.1777, + "step": 9186 + }, + { + "epoch": 0.48, + "learning_rate": 2.229243966026145e-07, + "loss": 0.158, + "step": 9187 + }, + { + "epoch": 0.48, + "learning_rate": 2.2289080765132836e-07, + "loss": 0.1725, + "step": 9188 + }, + { + "epoch": 0.48, + "learning_rate": 2.2285721804581405e-07, + "loss": 0.1587, + "step": 9189 + }, + { + "epoch": 0.48, + "learning_rate": 2.2282362778703163e-07, + "loss": 0.1342, + "step": 9190 + }, + { + "epoch": 0.48, + "learning_rate": 2.227900368759411e-07, + "loss": 0.1584, + "step": 9191 + }, + { + "epoch": 0.48, + "learning_rate": 2.2275644531350255e-07, + "loss": 0.1308, + "step": 9192 + }, + { + "epoch": 0.48, + "learning_rate": 2.22722853100676e-07, + "loss": 0.1623, + "step": 9193 + }, + { + "epoch": 0.48, + "learning_rate": 2.2268926023842152e-07, + "loss": 0.1325, + "step": 9194 + }, + { + "epoch": 0.48, + "learning_rate": 2.2265566672769922e-07, + "loss": 0.1403, + "step": 9195 + }, + { + "epoch": 0.48, + "learning_rate": 2.2262207256946924e-07, + "loss": 0.1172, + "step": 9196 + }, + { + "epoch": 0.48, + "learning_rate": 2.2258847776469173e-07, + "loss": 0.1362, + "step": 9197 + }, + { + "epoch": 0.48, + "learning_rate": 2.2255488231432675e-07, + "loss": 0.1549, + "step": 9198 + }, + { + "epoch": 0.48, + "learning_rate": 2.2252128621933454e-07, + "loss": 0.1259, + "step": 9199 + }, + { + "epoch": 0.48, + "learning_rate": 2.2248768948067527e-07, + "loss": 0.1193, + "step": 9200 + }, + { + "epoch": 0.48, + "learning_rate": 2.2245409209930922e-07, + "loss": 0.1051, + "step": 9201 + }, + { + "epoch": 0.48, + "learning_rate": 2.2242049407619646e-07, + "loss": 0.1174, + "step": 9202 + }, + { + "epoch": 0.48, + "learning_rate": 2.223868954122973e-07, + "loss": 0.1553, + "step": 9203 + }, + { + "epoch": 0.48, + "learning_rate": 2.2235329610857207e-07, + "loss": 0.1251, + "step": 9204 + }, + { + "epoch": 0.48, + "learning_rate": 2.2231969616598106e-07, + "loss": 0.1684, + "step": 9205 + }, + { + "epoch": 0.48, + "learning_rate": 2.2228609558548448e-07, + "loss": 0.1115, + "step": 9206 + }, + { + "epoch": 0.48, + "learning_rate": 2.2225249436804266e-07, + "loss": 0.1072, + "step": 9207 + }, + { + "epoch": 0.48, + "learning_rate": 2.22218892514616e-07, + "loss": 0.1148, + "step": 9208 + }, + { + "epoch": 0.48, + "learning_rate": 2.221852900261648e-07, + "loss": 0.1259, + "step": 9209 + }, + { + "epoch": 0.48, + "learning_rate": 2.2215168690364944e-07, + "loss": 0.1344, + "step": 9210 + }, + { + "epoch": 0.48, + "learning_rate": 2.221180831480303e-07, + "loss": 0.1269, + "step": 9211 + }, + { + "epoch": 0.48, + "learning_rate": 2.2208447876026785e-07, + "loss": 0.144, + "step": 9212 + }, + { + "epoch": 0.48, + "learning_rate": 2.2205087374132244e-07, + "loss": 0.1654, + "step": 9213 + }, + { + "epoch": 0.48, + "learning_rate": 2.2201726809215455e-07, + "loss": 0.1338, + "step": 9214 + }, + { + "epoch": 0.48, + "learning_rate": 2.2198366181372465e-07, + "loss": 0.1427, + "step": 9215 + }, + { + "epoch": 0.48, + "learning_rate": 2.2195005490699318e-07, + "loss": 0.1317, + "step": 9216 + }, + { + "epoch": 0.48, + "learning_rate": 2.219164473729207e-07, + "loss": 0.1612, + "step": 9217 + }, + { + "epoch": 0.48, + "learning_rate": 2.2188283921246768e-07, + "loss": 0.1374, + "step": 9218 + }, + { + "epoch": 0.48, + "learning_rate": 2.218492304265947e-07, + "loss": 0.1313, + "step": 9219 + }, + { + "epoch": 0.48, + "learning_rate": 2.218156210162622e-07, + "loss": 0.1312, + "step": 9220 + }, + { + "epoch": 0.48, + "learning_rate": 2.2178201098243088e-07, + "loss": 0.1249, + "step": 9221 + }, + { + "epoch": 0.48, + "learning_rate": 2.217484003260613e-07, + "loss": 0.1409, + "step": 9222 + }, + { + "epoch": 0.48, + "learning_rate": 2.21714789048114e-07, + "loss": 0.1284, + "step": 9223 + }, + { + "epoch": 0.48, + "learning_rate": 2.2168117714954967e-07, + "loss": 0.1514, + "step": 9224 + }, + { + "epoch": 0.48, + "learning_rate": 2.216475646313289e-07, + "loss": 0.1476, + "step": 9225 + }, + { + "epoch": 0.48, + "learning_rate": 2.216139514944124e-07, + "loss": 0.1219, + "step": 9226 + }, + { + "epoch": 0.48, + "learning_rate": 2.2158033773976084e-07, + "loss": 0.135, + "step": 9227 + }, + { + "epoch": 0.48, + "learning_rate": 2.215467233683349e-07, + "loss": 0.1507, + "step": 9228 + }, + { + "epoch": 0.48, + "learning_rate": 2.2151310838109524e-07, + "loss": 0.1291, + "step": 9229 + }, + { + "epoch": 0.48, + "learning_rate": 2.2147949277900263e-07, + "loss": 0.1348, + "step": 9230 + }, + { + "epoch": 0.48, + "learning_rate": 2.214458765630179e-07, + "loss": 0.1668, + "step": 9231 + }, + { + "epoch": 0.48, + "learning_rate": 2.2141225973410168e-07, + "loss": 0.1393, + "step": 9232 + }, + { + "epoch": 0.48, + "learning_rate": 2.213786422932148e-07, + "loss": 0.1496, + "step": 9233 + }, + { + "epoch": 0.48, + "learning_rate": 2.2134502424131808e-07, + "loss": 0.1032, + "step": 9234 + }, + { + "epoch": 0.48, + "learning_rate": 2.2131140557937242e-07, + "loss": 0.1564, + "step": 9235 + }, + { + "epoch": 0.48, + "learning_rate": 2.2127778630833846e-07, + "loss": 0.1347, + "step": 9236 + }, + { + "epoch": 0.48, + "learning_rate": 2.2124416642917718e-07, + "loss": 0.1495, + "step": 9237 + }, + { + "epoch": 0.48, + "learning_rate": 2.212105459428494e-07, + "loss": 0.123, + "step": 9238 + }, + { + "epoch": 0.48, + "learning_rate": 2.2117692485031606e-07, + "loss": 0.1279, + "step": 9239 + }, + { + "epoch": 0.48, + "learning_rate": 2.2114330315253803e-07, + "loss": 0.14, + "step": 9240 + }, + { + "epoch": 0.48, + "learning_rate": 2.2110968085047622e-07, + "loss": 0.114, + "step": 9241 + }, + { + "epoch": 0.48, + "learning_rate": 2.210760579450916e-07, + "loss": 0.097, + "step": 9242 + }, + { + "epoch": 0.48, + "learning_rate": 2.2104243443734507e-07, + "loss": 0.1535, + "step": 9243 + }, + { + "epoch": 0.48, + "learning_rate": 2.210088103281977e-07, + "loss": 0.1599, + "step": 9244 + }, + { + "epoch": 0.48, + "learning_rate": 2.2097518561861036e-07, + "loss": 0.1603, + "step": 9245 + }, + { + "epoch": 0.48, + "learning_rate": 2.2094156030954418e-07, + "loss": 0.1381, + "step": 9246 + }, + { + "epoch": 0.48, + "learning_rate": 2.2090793440196002e-07, + "loss": 0.1531, + "step": 9247 + }, + { + "epoch": 0.48, + "learning_rate": 2.208743078968191e-07, + "loss": 0.1632, + "step": 9248 + }, + { + "epoch": 0.48, + "learning_rate": 2.2084068079508245e-07, + "loss": 0.1318, + "step": 9249 + }, + { + "epoch": 0.48, + "learning_rate": 2.20807053097711e-07, + "loss": 0.1541, + "step": 9250 + }, + { + "epoch": 0.48, + "learning_rate": 2.20773424805666e-07, + "loss": 0.0908, + "step": 9251 + }, + { + "epoch": 0.48, + "learning_rate": 2.2073979591990849e-07, + "loss": 0.1405, + "step": 9252 + }, + { + "epoch": 0.48, + "learning_rate": 2.207061664413996e-07, + "loss": 0.1293, + "step": 9253 + }, + { + "epoch": 0.48, + "learning_rate": 2.2067253637110042e-07, + "loss": 0.1403, + "step": 9254 + }, + { + "epoch": 0.48, + "learning_rate": 2.206389057099723e-07, + "loss": 0.115, + "step": 9255 + }, + { + "epoch": 0.48, + "learning_rate": 2.2060527445897616e-07, + "loss": 0.1455, + "step": 9256 + }, + { + "epoch": 0.48, + "learning_rate": 2.2057164261907343e-07, + "loss": 0.1533, + "step": 9257 + }, + { + "epoch": 0.48, + "learning_rate": 2.2053801019122512e-07, + "loss": 0.1792, + "step": 9258 + }, + { + "epoch": 0.48, + "learning_rate": 2.2050437717639254e-07, + "loss": 0.1471, + "step": 9259 + }, + { + "epoch": 0.48, + "learning_rate": 2.2047074357553703e-07, + "loss": 0.1543, + "step": 9260 + }, + { + "epoch": 0.48, + "learning_rate": 2.2043710938961975e-07, + "loss": 0.1555, + "step": 9261 + }, + { + "epoch": 0.48, + "learning_rate": 2.20403474619602e-07, + "loss": 0.1296, + "step": 9262 + }, + { + "epoch": 0.48, + "learning_rate": 2.20369839266445e-07, + "loss": 0.1295, + "step": 9263 + }, + { + "epoch": 0.48, + "learning_rate": 2.2033620333111017e-07, + "loss": 0.1651, + "step": 9264 + }, + { + "epoch": 0.48, + "learning_rate": 2.2030256681455884e-07, + "loss": 0.1472, + "step": 9265 + }, + { + "epoch": 0.48, + "learning_rate": 2.2026892971775227e-07, + "loss": 0.1267, + "step": 9266 + }, + { + "epoch": 0.48, + "learning_rate": 2.2023529204165183e-07, + "loss": 0.1464, + "step": 9267 + }, + { + "epoch": 0.48, + "learning_rate": 2.20201653787219e-07, + "loss": 0.1374, + "step": 9268 + }, + { + "epoch": 0.48, + "learning_rate": 2.201680149554151e-07, + "loss": 0.1423, + "step": 9269 + }, + { + "epoch": 0.48, + "learning_rate": 2.2013437554720146e-07, + "loss": 0.1508, + "step": 9270 + }, + { + "epoch": 0.48, + "learning_rate": 2.201007355635397e-07, + "loss": 0.1267, + "step": 9271 + }, + { + "epoch": 0.48, + "learning_rate": 2.2006709500539106e-07, + "loss": 0.1331, + "step": 9272 + }, + { + "epoch": 0.48, + "learning_rate": 2.2003345387371718e-07, + "loss": 0.1237, + "step": 9273 + }, + { + "epoch": 0.48, + "learning_rate": 2.1999981216947945e-07, + "loss": 0.1218, + "step": 9274 + }, + { + "epoch": 0.48, + "learning_rate": 2.1996616989363935e-07, + "loss": 0.112, + "step": 9275 + }, + { + "epoch": 0.48, + "learning_rate": 2.1993252704715837e-07, + "loss": 0.1553, + "step": 9276 + }, + { + "epoch": 0.48, + "learning_rate": 2.1989888363099812e-07, + "loss": 0.1487, + "step": 9277 + }, + { + "epoch": 0.48, + "learning_rate": 2.1986523964612008e-07, + "loss": 0.1609, + "step": 9278 + }, + { + "epoch": 0.48, + "learning_rate": 2.198315950934858e-07, + "loss": 0.124, + "step": 9279 + }, + { + "epoch": 0.48, + "learning_rate": 2.197979499740569e-07, + "loss": 0.1413, + "step": 9280 + }, + { + "epoch": 0.48, + "learning_rate": 2.1976430428879493e-07, + "loss": 0.1384, + "step": 9281 + }, + { + "epoch": 0.48, + "learning_rate": 2.1973065803866158e-07, + "loss": 0.118, + "step": 9282 + }, + { + "epoch": 0.48, + "learning_rate": 2.1969701122461831e-07, + "loss": 0.1536, + "step": 9283 + }, + { + "epoch": 0.48, + "learning_rate": 2.1966336384762695e-07, + "loss": 0.1365, + "step": 9284 + }, + { + "epoch": 0.48, + "learning_rate": 2.1962971590864897e-07, + "loss": 0.1367, + "step": 9285 + }, + { + "epoch": 0.48, + "learning_rate": 2.1959606740864616e-07, + "loss": 0.1375, + "step": 9286 + }, + { + "epoch": 0.48, + "learning_rate": 2.195624183485802e-07, + "loss": 0.1467, + "step": 9287 + }, + { + "epoch": 0.48, + "learning_rate": 2.1952876872941272e-07, + "loss": 0.1316, + "step": 9288 + }, + { + "epoch": 0.48, + "learning_rate": 2.1949511855210558e-07, + "loss": 0.1314, + "step": 9289 + }, + { + "epoch": 0.48, + "learning_rate": 2.1946146781762032e-07, + "loss": 0.1522, + "step": 9290 + }, + { + "epoch": 0.48, + "learning_rate": 2.194278165269189e-07, + "loss": 0.1353, + "step": 9291 + }, + { + "epoch": 0.49, + "learning_rate": 2.1939416468096295e-07, + "loss": 0.1112, + "step": 9292 + }, + { + "epoch": 0.49, + "learning_rate": 2.193605122807143e-07, + "loss": 0.1392, + "step": 9293 + }, + { + "epoch": 0.49, + "learning_rate": 2.1932685932713464e-07, + "loss": 0.1206, + "step": 9294 + }, + { + "epoch": 0.49, + "learning_rate": 2.19293205821186e-07, + "loss": 0.1361, + "step": 9295 + }, + { + "epoch": 0.49, + "learning_rate": 2.1925955176383004e-07, + "loss": 0.152, + "step": 9296 + }, + { + "epoch": 0.49, + "learning_rate": 2.192258971560286e-07, + "loss": 0.1279, + "step": 9297 + }, + { + "epoch": 0.49, + "learning_rate": 2.1919224199874366e-07, + "loss": 0.1187, + "step": 9298 + }, + { + "epoch": 0.49, + "learning_rate": 2.1915858629293704e-07, + "loss": 0.1453, + "step": 9299 + }, + { + "epoch": 0.49, + "learning_rate": 2.1912493003957064e-07, + "loss": 0.1604, + "step": 9300 + }, + { + "epoch": 0.49, + "learning_rate": 2.1909127323960632e-07, + "loss": 0.144, + "step": 9301 + }, + { + "epoch": 0.49, + "learning_rate": 2.1905761589400607e-07, + "loss": 0.0998, + "step": 9302 + }, + { + "epoch": 0.49, + "learning_rate": 2.1902395800373173e-07, + "loss": 0.1365, + "step": 9303 + }, + { + "epoch": 0.49, + "learning_rate": 2.189902995697454e-07, + "loss": 0.1551, + "step": 9304 + }, + { + "epoch": 0.49, + "learning_rate": 2.1895664059300897e-07, + "loss": 0.1334, + "step": 9305 + }, + { + "epoch": 0.49, + "learning_rate": 2.1892298107448442e-07, + "loss": 0.1081, + "step": 9306 + }, + { + "epoch": 0.49, + "learning_rate": 2.1888932101513377e-07, + "loss": 0.1226, + "step": 9307 + }, + { + "epoch": 0.49, + "learning_rate": 2.1885566041591903e-07, + "loss": 0.1163, + "step": 9308 + }, + { + "epoch": 0.49, + "learning_rate": 2.1882199927780224e-07, + "loss": 0.1432, + "step": 9309 + }, + { + "epoch": 0.49, + "learning_rate": 2.1878833760174543e-07, + "loss": 0.1368, + "step": 9310 + }, + { + "epoch": 0.49, + "learning_rate": 2.1875467538871071e-07, + "loss": 0.1327, + "step": 9311 + }, + { + "epoch": 0.49, + "learning_rate": 2.187210126396601e-07, + "loss": 0.1418, + "step": 9312 + }, + { + "epoch": 0.49, + "learning_rate": 2.1868734935555574e-07, + "loss": 0.1942, + "step": 9313 + }, + { + "epoch": 0.49, + "learning_rate": 2.186536855373597e-07, + "loss": 0.1139, + "step": 9314 + }, + { + "epoch": 0.49, + "learning_rate": 2.1862002118603412e-07, + "loss": 0.1363, + "step": 9315 + }, + { + "epoch": 0.49, + "learning_rate": 2.1858635630254115e-07, + "loss": 0.1441, + "step": 9316 + }, + { + "epoch": 0.49, + "learning_rate": 2.1855269088784297e-07, + "loss": 0.1571, + "step": 9317 + }, + { + "epoch": 0.49, + "learning_rate": 2.1851902494290173e-07, + "loss": 0.1157, + "step": 9318 + }, + { + "epoch": 0.49, + "learning_rate": 2.1848535846867952e-07, + "loss": 0.1645, + "step": 9319 + }, + { + "epoch": 0.49, + "learning_rate": 2.1845169146613874e-07, + "loss": 0.1756, + "step": 9320 + }, + { + "epoch": 0.49, + "learning_rate": 2.1841802393624147e-07, + "loss": 0.1135, + "step": 9321 + }, + { + "epoch": 0.49, + "learning_rate": 2.1838435587994994e-07, + "loss": 0.1648, + "step": 9322 + }, + { + "epoch": 0.49, + "learning_rate": 2.183506872982264e-07, + "loss": 0.1325, + "step": 9323 + }, + { + "epoch": 0.49, + "learning_rate": 2.1831701819203316e-07, + "loss": 0.1047, + "step": 9324 + }, + { + "epoch": 0.49, + "learning_rate": 2.182833485623325e-07, + "loss": 0.1133, + "step": 9325 + }, + { + "epoch": 0.49, + "learning_rate": 2.182496784100866e-07, + "loss": 0.1492, + "step": 9326 + }, + { + "epoch": 0.49, + "learning_rate": 2.182160077362579e-07, + "loss": 0.1562, + "step": 9327 + }, + { + "epoch": 0.49, + "learning_rate": 2.1818233654180864e-07, + "loss": 0.1565, + "step": 9328 + }, + { + "epoch": 0.49, + "learning_rate": 2.1814866482770125e-07, + "loss": 0.1377, + "step": 9329 + }, + { + "epoch": 0.49, + "learning_rate": 2.1811499259489796e-07, + "loss": 0.1778, + "step": 9330 + }, + { + "epoch": 0.49, + "learning_rate": 2.1808131984436117e-07, + "loss": 0.1431, + "step": 9331 + }, + { + "epoch": 0.49, + "learning_rate": 2.1804764657705327e-07, + "loss": 0.1872, + "step": 9332 + }, + { + "epoch": 0.49, + "learning_rate": 2.1801397279393673e-07, + "loss": 0.1275, + "step": 9333 + }, + { + "epoch": 0.49, + "learning_rate": 2.1798029849597391e-07, + "loss": 0.1448, + "step": 9334 + }, + { + "epoch": 0.49, + "learning_rate": 2.1794662368412712e-07, + "loss": 0.1415, + "step": 9335 + }, + { + "epoch": 0.49, + "learning_rate": 2.1791294835935897e-07, + "loss": 0.1391, + "step": 9336 + }, + { + "epoch": 0.49, + "learning_rate": 2.1787927252263177e-07, + "loss": 0.142, + "step": 9337 + }, + { + "epoch": 0.49, + "learning_rate": 2.1784559617490817e-07, + "loss": 0.1645, + "step": 9338 + }, + { + "epoch": 0.49, + "learning_rate": 2.1781191931715046e-07, + "loss": 0.1265, + "step": 9339 + }, + { + "epoch": 0.49, + "learning_rate": 2.177782419503212e-07, + "loss": 0.1605, + "step": 9340 + }, + { + "epoch": 0.49, + "learning_rate": 2.1774456407538294e-07, + "loss": 0.1448, + "step": 9341 + }, + { + "epoch": 0.49, + "learning_rate": 2.1771088569329823e-07, + "loss": 0.152, + "step": 9342 + }, + { + "epoch": 0.49, + "learning_rate": 2.1767720680502957e-07, + "loss": 0.1251, + "step": 9343 + }, + { + "epoch": 0.49, + "learning_rate": 2.1764352741153947e-07, + "loss": 0.1162, + "step": 9344 + }, + { + "epoch": 0.49, + "learning_rate": 2.176098475137906e-07, + "loss": 0.1535, + "step": 9345 + }, + { + "epoch": 0.49, + "learning_rate": 2.175761671127454e-07, + "loss": 0.1225, + "step": 9346 + }, + { + "epoch": 0.49, + "learning_rate": 2.175424862093667e-07, + "loss": 0.1242, + "step": 9347 + }, + { + "epoch": 0.49, + "learning_rate": 2.1750880480461684e-07, + "loss": 0.1446, + "step": 9348 + }, + { + "epoch": 0.49, + "learning_rate": 2.1747512289945865e-07, + "loss": 0.1599, + "step": 9349 + }, + { + "epoch": 0.49, + "learning_rate": 2.1744144049485468e-07, + "loss": 0.1155, + "step": 9350 + }, + { + "epoch": 0.49, + "learning_rate": 2.174077575917676e-07, + "loss": 0.1485, + "step": 9351 + }, + { + "epoch": 0.49, + "learning_rate": 2.173740741911601e-07, + "loss": 0.1183, + "step": 9352 + }, + { + "epoch": 0.49, + "learning_rate": 2.173403902939948e-07, + "loss": 0.2085, + "step": 9353 + }, + { + "epoch": 0.49, + "learning_rate": 2.1730670590123452e-07, + "loss": 0.1529, + "step": 9354 + }, + { + "epoch": 0.49, + "learning_rate": 2.172730210138419e-07, + "loss": 0.2001, + "step": 9355 + }, + { + "epoch": 0.49, + "learning_rate": 2.172393356327797e-07, + "loss": 0.1691, + "step": 9356 + }, + { + "epoch": 0.49, + "learning_rate": 2.1720564975901054e-07, + "loss": 0.121, + "step": 9357 + }, + { + "epoch": 0.49, + "learning_rate": 2.1717196339349731e-07, + "loss": 0.1261, + "step": 9358 + }, + { + "epoch": 0.49, + "learning_rate": 2.171382765372028e-07, + "loss": 0.0805, + "step": 9359 + }, + { + "epoch": 0.49, + "learning_rate": 2.1710458919108966e-07, + "loss": 0.1165, + "step": 9360 + }, + { + "epoch": 0.49, + "learning_rate": 2.1707090135612077e-07, + "loss": 0.1196, + "step": 9361 + }, + { + "epoch": 0.49, + "learning_rate": 2.170372130332589e-07, + "loss": 0.1501, + "step": 9362 + }, + { + "epoch": 0.49, + "learning_rate": 2.1700352422346695e-07, + "loss": 0.1483, + "step": 9363 + }, + { + "epoch": 0.49, + "learning_rate": 2.169698349277077e-07, + "loss": 0.1793, + "step": 9364 + }, + { + "epoch": 0.49, + "learning_rate": 2.1693614514694402e-07, + "loss": 0.1288, + "step": 9365 + }, + { + "epoch": 0.49, + "learning_rate": 2.1690245488213876e-07, + "loss": 0.1213, + "step": 9366 + }, + { + "epoch": 0.49, + "learning_rate": 2.168687641342548e-07, + "loss": 0.1137, + "step": 9367 + }, + { + "epoch": 0.49, + "learning_rate": 2.1683507290425512e-07, + "loss": 0.1378, + "step": 9368 + }, + { + "epoch": 0.49, + "learning_rate": 2.168013811931025e-07, + "loss": 0.1229, + "step": 9369 + }, + { + "epoch": 0.49, + "learning_rate": 2.1676768900175993e-07, + "loss": 0.1803, + "step": 9370 + }, + { + "epoch": 0.49, + "learning_rate": 2.167339963311903e-07, + "loss": 0.1545, + "step": 9371 + }, + { + "epoch": 0.49, + "learning_rate": 2.167003031823567e-07, + "loss": 0.1304, + "step": 9372 + }, + { + "epoch": 0.49, + "learning_rate": 2.1666660955622187e-07, + "loss": 0.133, + "step": 9373 + }, + { + "epoch": 0.49, + "learning_rate": 2.1663291545374895e-07, + "loss": 0.1437, + "step": 9374 + }, + { + "epoch": 0.49, + "learning_rate": 2.1659922087590089e-07, + "loss": 0.1023, + "step": 9375 + }, + { + "epoch": 0.49, + "learning_rate": 2.1656552582364068e-07, + "loss": 0.1598, + "step": 9376 + }, + { + "epoch": 0.49, + "learning_rate": 2.1653183029793136e-07, + "loss": 0.1551, + "step": 9377 + }, + { + "epoch": 0.49, + "learning_rate": 2.1649813429973597e-07, + "loss": 0.1236, + "step": 9378 + }, + { + "epoch": 0.49, + "learning_rate": 2.1646443783001745e-07, + "loss": 0.1642, + "step": 9379 + }, + { + "epoch": 0.49, + "learning_rate": 2.1643074088973905e-07, + "loss": 0.1171, + "step": 9380 + }, + { + "epoch": 0.49, + "learning_rate": 2.163970434798637e-07, + "loss": 0.1539, + "step": 9381 + }, + { + "epoch": 0.49, + "learning_rate": 2.1636334560135445e-07, + "loss": 0.1504, + "step": 9382 + }, + { + "epoch": 0.49, + "learning_rate": 2.1632964725517452e-07, + "loss": 0.142, + "step": 9383 + }, + { + "epoch": 0.49, + "learning_rate": 2.1629594844228695e-07, + "loss": 0.1614, + "step": 9384 + }, + { + "epoch": 0.49, + "learning_rate": 2.1626224916365493e-07, + "loss": 0.1428, + "step": 9385 + }, + { + "epoch": 0.49, + "learning_rate": 2.162285494202415e-07, + "loss": 0.1396, + "step": 9386 + }, + { + "epoch": 0.49, + "learning_rate": 2.161948492130099e-07, + "loss": 0.1561, + "step": 9387 + }, + { + "epoch": 0.49, + "learning_rate": 2.161611485429232e-07, + "loss": 0.1738, + "step": 9388 + }, + { + "epoch": 0.49, + "learning_rate": 2.161274474109447e-07, + "loss": 0.1206, + "step": 9389 + }, + { + "epoch": 0.49, + "learning_rate": 2.160937458180375e-07, + "loss": 0.1598, + "step": 9390 + }, + { + "epoch": 0.49, + "learning_rate": 2.160600437651648e-07, + "loss": 0.1612, + "step": 9391 + }, + { + "epoch": 0.49, + "learning_rate": 2.160263412532899e-07, + "loss": 0.1466, + "step": 9392 + }, + { + "epoch": 0.49, + "learning_rate": 2.1599263828337594e-07, + "loss": 0.1753, + "step": 9393 + }, + { + "epoch": 0.49, + "learning_rate": 2.1595893485638626e-07, + "loss": 0.1323, + "step": 9394 + }, + { + "epoch": 0.49, + "learning_rate": 2.1592523097328396e-07, + "loss": 0.1213, + "step": 9395 + }, + { + "epoch": 0.49, + "learning_rate": 2.1589152663503252e-07, + "loss": 0.1686, + "step": 9396 + }, + { + "epoch": 0.49, + "learning_rate": 2.1585782184259504e-07, + "loss": 0.1349, + "step": 9397 + }, + { + "epoch": 0.49, + "learning_rate": 2.1582411659693491e-07, + "loss": 0.1154, + "step": 9398 + }, + { + "epoch": 0.49, + "learning_rate": 2.1579041089901545e-07, + "loss": 0.1589, + "step": 9399 + }, + { + "epoch": 0.49, + "learning_rate": 2.157567047497999e-07, + "loss": 0.1352, + "step": 9400 + }, + { + "epoch": 0.49, + "learning_rate": 2.157229981502517e-07, + "loss": 0.1201, + "step": 9401 + }, + { + "epoch": 0.49, + "learning_rate": 2.1568929110133412e-07, + "loss": 0.1462, + "step": 9402 + }, + { + "epoch": 0.49, + "learning_rate": 2.1565558360401054e-07, + "loss": 0.115, + "step": 9403 + }, + { + "epoch": 0.49, + "learning_rate": 2.1562187565924433e-07, + "loss": 0.1376, + "step": 9404 + }, + { + "epoch": 0.49, + "learning_rate": 2.1558816726799888e-07, + "loss": 0.1837, + "step": 9405 + }, + { + "epoch": 0.49, + "learning_rate": 2.1555445843123765e-07, + "loss": 0.1457, + "step": 9406 + }, + { + "epoch": 0.49, + "learning_rate": 2.1552074914992388e-07, + "loss": 0.1477, + "step": 9407 + }, + { + "epoch": 0.49, + "learning_rate": 2.154870394250212e-07, + "loss": 0.1278, + "step": 9408 + }, + { + "epoch": 0.49, + "learning_rate": 2.1545332925749293e-07, + "loss": 0.1501, + "step": 9409 + }, + { + "epoch": 0.49, + "learning_rate": 2.1541961864830252e-07, + "loss": 0.1167, + "step": 9410 + }, + { + "epoch": 0.49, + "learning_rate": 2.153859075984135e-07, + "loss": 0.1261, + "step": 9411 + }, + { + "epoch": 0.49, + "learning_rate": 2.1535219610878932e-07, + "loss": 0.1349, + "step": 9412 + }, + { + "epoch": 0.49, + "learning_rate": 2.1531848418039335e-07, + "loss": 0.1298, + "step": 9413 + }, + { + "epoch": 0.49, + "learning_rate": 2.152847718141893e-07, + "loss": 0.1324, + "step": 9414 + }, + { + "epoch": 0.49, + "learning_rate": 2.152510590111405e-07, + "loss": 0.1813, + "step": 9415 + }, + { + "epoch": 0.49, + "learning_rate": 2.1521734577221053e-07, + "loss": 0.0995, + "step": 9416 + }, + { + "epoch": 0.49, + "learning_rate": 2.1518363209836302e-07, + "loss": 0.1266, + "step": 9417 + }, + { + "epoch": 0.49, + "learning_rate": 2.1514991799056136e-07, + "loss": 0.1455, + "step": 9418 + }, + { + "epoch": 0.49, + "learning_rate": 2.1511620344976928e-07, + "loss": 0.1453, + "step": 9419 + }, + { + "epoch": 0.49, + "learning_rate": 2.1508248847695021e-07, + "loss": 0.1219, + "step": 9420 + }, + { + "epoch": 0.49, + "learning_rate": 2.1504877307306782e-07, + "loss": 0.1193, + "step": 9421 + }, + { + "epoch": 0.49, + "learning_rate": 2.1501505723908563e-07, + "loss": 0.133, + "step": 9422 + }, + { + "epoch": 0.49, + "learning_rate": 2.149813409759674e-07, + "loss": 0.1367, + "step": 9423 + }, + { + "epoch": 0.49, + "learning_rate": 2.1494762428467662e-07, + "loss": 0.1297, + "step": 9424 + }, + { + "epoch": 0.49, + "learning_rate": 2.1491390716617695e-07, + "loss": 0.1273, + "step": 9425 + }, + { + "epoch": 0.49, + "learning_rate": 2.1488018962143207e-07, + "loss": 0.1443, + "step": 9426 + }, + { + "epoch": 0.49, + "learning_rate": 2.148464716514056e-07, + "loss": 0.1394, + "step": 9427 + }, + { + "epoch": 0.49, + "learning_rate": 2.1481275325706135e-07, + "loss": 0.1203, + "step": 9428 + }, + { + "epoch": 0.49, + "learning_rate": 2.1477903443936278e-07, + "loss": 0.1616, + "step": 9429 + }, + { + "epoch": 0.49, + "learning_rate": 2.1474531519927375e-07, + "loss": 0.1786, + "step": 9430 + }, + { + "epoch": 0.49, + "learning_rate": 2.1471159553775786e-07, + "loss": 0.1199, + "step": 9431 + }, + { + "epoch": 0.49, + "learning_rate": 2.14677875455779e-07, + "loss": 0.1344, + "step": 9432 + }, + { + "epoch": 0.49, + "learning_rate": 2.1464415495430067e-07, + "loss": 0.1197, + "step": 9433 + }, + { + "epoch": 0.49, + "learning_rate": 2.1461043403428683e-07, + "loss": 0.1437, + "step": 9434 + }, + { + "epoch": 0.49, + "learning_rate": 2.1457671269670108e-07, + "loss": 0.1174, + "step": 9435 + }, + { + "epoch": 0.49, + "learning_rate": 2.145429909425073e-07, + "loss": 0.1646, + "step": 9436 + }, + { + "epoch": 0.49, + "learning_rate": 2.145092687726692e-07, + "loss": 0.1181, + "step": 9437 + }, + { + "epoch": 0.49, + "learning_rate": 2.1447554618815057e-07, + "loss": 0.1203, + "step": 9438 + }, + { + "epoch": 0.49, + "learning_rate": 2.144418231899153e-07, + "loss": 0.1248, + "step": 9439 + }, + { + "epoch": 0.49, + "learning_rate": 2.144080997789271e-07, + "loss": 0.1474, + "step": 9440 + }, + { + "epoch": 0.49, + "learning_rate": 2.1437437595614992e-07, + "loss": 0.1359, + "step": 9441 + }, + { + "epoch": 0.49, + "learning_rate": 2.1434065172254746e-07, + "loss": 0.1676, + "step": 9442 + }, + { + "epoch": 0.49, + "learning_rate": 2.1430692707908364e-07, + "loss": 0.1684, + "step": 9443 + }, + { + "epoch": 0.49, + "learning_rate": 2.142732020267223e-07, + "loss": 0.1511, + "step": 9444 + }, + { + "epoch": 0.49, + "learning_rate": 2.142394765664274e-07, + "loss": 0.1477, + "step": 9445 + }, + { + "epoch": 0.49, + "learning_rate": 2.1420575069916275e-07, + "loss": 0.1445, + "step": 9446 + }, + { + "epoch": 0.49, + "learning_rate": 2.1417202442589222e-07, + "loss": 0.168, + "step": 9447 + }, + { + "epoch": 0.49, + "learning_rate": 2.1413829774757982e-07, + "loss": 0.1441, + "step": 9448 + }, + { + "epoch": 0.49, + "learning_rate": 2.1410457066518935e-07, + "loss": 0.1244, + "step": 9449 + }, + { + "epoch": 0.49, + "learning_rate": 2.1407084317968486e-07, + "loss": 0.1204, + "step": 9450 + }, + { + "epoch": 0.49, + "learning_rate": 2.1403711529203017e-07, + "loss": 0.159, + "step": 9451 + }, + { + "epoch": 0.49, + "learning_rate": 2.1400338700318938e-07, + "loss": 0.1525, + "step": 9452 + }, + { + "epoch": 0.49, + "learning_rate": 2.139696583141263e-07, + "loss": 0.1298, + "step": 9453 + }, + { + "epoch": 0.49, + "learning_rate": 2.1393592922580509e-07, + "loss": 0.1111, + "step": 9454 + }, + { + "epoch": 0.49, + "learning_rate": 2.1390219973918962e-07, + "loss": 0.1548, + "step": 9455 + }, + { + "epoch": 0.49, + "learning_rate": 2.1386846985524385e-07, + "loss": 0.1252, + "step": 9456 + }, + { + "epoch": 0.49, + "learning_rate": 2.1383473957493189e-07, + "loss": 0.1476, + "step": 9457 + }, + { + "epoch": 0.49, + "learning_rate": 2.138010088992178e-07, + "loss": 0.1433, + "step": 9458 + }, + { + "epoch": 0.49, + "learning_rate": 2.1376727782906548e-07, + "loss": 0.1395, + "step": 9459 + }, + { + "epoch": 0.49, + "learning_rate": 2.1373354636543902e-07, + "loss": 0.1473, + "step": 9460 + }, + { + "epoch": 0.49, + "learning_rate": 2.1369981450930257e-07, + "loss": 0.1444, + "step": 9461 + }, + { + "epoch": 0.49, + "learning_rate": 2.1366608226162014e-07, + "loss": 0.1431, + "step": 9462 + }, + { + "epoch": 0.49, + "learning_rate": 2.1363234962335572e-07, + "loss": 0.1244, + "step": 9463 + }, + { + "epoch": 0.49, + "learning_rate": 2.1359861659547352e-07, + "loss": 0.1374, + "step": 9464 + }, + { + "epoch": 0.49, + "learning_rate": 2.135648831789376e-07, + "loss": 0.1342, + "step": 9465 + }, + { + "epoch": 0.49, + "learning_rate": 2.1353114937471215e-07, + "loss": 0.1156, + "step": 9466 + }, + { + "epoch": 0.49, + "learning_rate": 2.1349741518376114e-07, + "loss": 0.1293, + "step": 9467 + }, + { + "epoch": 0.49, + "learning_rate": 2.1346368060704884e-07, + "loss": 0.139, + "step": 9468 + }, + { + "epoch": 0.49, + "learning_rate": 2.1342994564553933e-07, + "loss": 0.1193, + "step": 9469 + }, + { + "epoch": 0.49, + "learning_rate": 2.1339621030019682e-07, + "loss": 0.1446, + "step": 9470 + }, + { + "epoch": 0.49, + "learning_rate": 2.1336247457198544e-07, + "loss": 0.1307, + "step": 9471 + }, + { + "epoch": 0.49, + "learning_rate": 2.1332873846186935e-07, + "loss": 0.1421, + "step": 9472 + }, + { + "epoch": 0.49, + "learning_rate": 2.132950019708128e-07, + "loss": 0.1169, + "step": 9473 + }, + { + "epoch": 0.49, + "learning_rate": 2.132612650997799e-07, + "loss": 0.1392, + "step": 9474 + }, + { + "epoch": 0.49, + "learning_rate": 2.1322752784973502e-07, + "loss": 0.1728, + "step": 9475 + }, + { + "epoch": 0.49, + "learning_rate": 2.1319379022164221e-07, + "loss": 0.1376, + "step": 9476 + }, + { + "epoch": 0.49, + "learning_rate": 2.1316005221646587e-07, + "loss": 0.1256, + "step": 9477 + }, + { + "epoch": 0.49, + "learning_rate": 2.1312631383517004e-07, + "loss": 0.152, + "step": 9478 + }, + { + "epoch": 0.49, + "learning_rate": 2.1309257507871923e-07, + "loss": 0.1399, + "step": 9479 + }, + { + "epoch": 0.49, + "learning_rate": 2.1305883594807743e-07, + "loss": 0.1212, + "step": 9480 + }, + { + "epoch": 0.49, + "learning_rate": 2.1302509644420917e-07, + "loss": 0.1615, + "step": 9481 + }, + { + "epoch": 0.49, + "learning_rate": 2.1299135656807853e-07, + "loss": 0.1562, + "step": 9482 + }, + { + "epoch": 0.49, + "learning_rate": 2.129576163206499e-07, + "loss": 0.1383, + "step": 9483 + }, + { + "epoch": 0.5, + "learning_rate": 2.1292387570288772e-07, + "loss": 0.1648, + "step": 9484 + }, + { + "epoch": 0.5, + "learning_rate": 2.1289013471575608e-07, + "loss": 0.127, + "step": 9485 + }, + { + "epoch": 0.5, + "learning_rate": 2.1285639336021943e-07, + "loss": 0.1262, + "step": 9486 + }, + { + "epoch": 0.5, + "learning_rate": 2.1282265163724208e-07, + "loss": 0.1404, + "step": 9487 + }, + { + "epoch": 0.5, + "learning_rate": 2.1278890954778846e-07, + "loss": 0.1097, + "step": 9488 + }, + { + "epoch": 0.5, + "learning_rate": 2.1275516709282285e-07, + "loss": 0.114, + "step": 9489 + }, + { + "epoch": 0.5, + "learning_rate": 2.1272142427330963e-07, + "loss": 0.136, + "step": 9490 + }, + { + "epoch": 0.5, + "learning_rate": 2.1268768109021312e-07, + "loss": 0.1294, + "step": 9491 + }, + { + "epoch": 0.5, + "learning_rate": 2.1265393754449787e-07, + "loss": 0.1169, + "step": 9492 + }, + { + "epoch": 0.5, + "learning_rate": 2.1262019363712822e-07, + "loss": 0.1656, + "step": 9493 + }, + { + "epoch": 0.5, + "learning_rate": 2.1258644936906846e-07, + "loss": 0.1281, + "step": 9494 + }, + { + "epoch": 0.5, + "learning_rate": 2.125527047412832e-07, + "loss": 0.113, + "step": 9495 + }, + { + "epoch": 0.5, + "learning_rate": 2.125189597547368e-07, + "loss": 0.1369, + "step": 9496 + }, + { + "epoch": 0.5, + "learning_rate": 2.1248521441039368e-07, + "loss": 0.132, + "step": 9497 + }, + { + "epoch": 0.5, + "learning_rate": 2.124514687092183e-07, + "loss": 0.1235, + "step": 9498 + }, + { + "epoch": 0.5, + "learning_rate": 2.1241772265217515e-07, + "loss": 0.0941, + "step": 9499 + }, + { + "epoch": 0.5, + "learning_rate": 2.1238397624022863e-07, + "loss": 0.098, + "step": 9500 + }, + { + "epoch": 0.5, + "learning_rate": 2.1235022947434338e-07, + "loss": 0.1519, + "step": 9501 + }, + { + "epoch": 0.5, + "learning_rate": 2.1231648235548375e-07, + "loss": 0.1209, + "step": 9502 + }, + { + "epoch": 0.5, + "learning_rate": 2.122827348846143e-07, + "loss": 0.1536, + "step": 9503 + }, + { + "epoch": 0.5, + "learning_rate": 2.122489870626996e-07, + "loss": 0.1852, + "step": 9504 + }, + { + "epoch": 0.5, + "learning_rate": 2.122152388907041e-07, + "loss": 0.1407, + "step": 9505 + }, + { + "epoch": 0.5, + "learning_rate": 2.1218149036959233e-07, + "loss": 0.1083, + "step": 9506 + }, + { + "epoch": 0.5, + "learning_rate": 2.1214774150032885e-07, + "loss": 0.1652, + "step": 9507 + }, + { + "epoch": 0.5, + "learning_rate": 2.121139922838783e-07, + "loss": 0.1184, + "step": 9508 + }, + { + "epoch": 0.5, + "learning_rate": 2.1208024272120513e-07, + "loss": 0.1499, + "step": 9509 + }, + { + "epoch": 0.5, + "learning_rate": 2.1204649281327392e-07, + "loss": 0.1425, + "step": 9510 + }, + { + "epoch": 0.5, + "learning_rate": 2.1201274256104934e-07, + "loss": 0.0893, + "step": 9511 + }, + { + "epoch": 0.5, + "learning_rate": 2.1197899196549592e-07, + "loss": 0.1565, + "step": 9512 + }, + { + "epoch": 0.5, + "learning_rate": 2.119452410275783e-07, + "loss": 0.1224, + "step": 9513 + }, + { + "epoch": 0.5, + "learning_rate": 2.1191148974826104e-07, + "loss": 0.1443, + "step": 9514 + }, + { + "epoch": 0.5, + "learning_rate": 2.118777381285089e-07, + "loss": 0.1382, + "step": 9515 + }, + { + "epoch": 0.5, + "learning_rate": 2.1184398616928631e-07, + "loss": 0.1329, + "step": 9516 + }, + { + "epoch": 0.5, + "learning_rate": 2.1181023387155809e-07, + "loss": 0.1613, + "step": 9517 + }, + { + "epoch": 0.5, + "learning_rate": 2.1177648123628881e-07, + "loss": 0.1279, + "step": 9518 + }, + { + "epoch": 0.5, + "learning_rate": 2.1174272826444314e-07, + "loss": 0.14, + "step": 9519 + }, + { + "epoch": 0.5, + "learning_rate": 2.1170897495698576e-07, + "loss": 0.115, + "step": 9520 + }, + { + "epoch": 0.5, + "learning_rate": 2.1167522131488135e-07, + "loss": 0.1361, + "step": 9521 + }, + { + "epoch": 0.5, + "learning_rate": 2.1164146733909472e-07, + "loss": 0.1141, + "step": 9522 + }, + { + "epoch": 0.5, + "learning_rate": 2.1160771303059032e-07, + "loss": 0.1336, + "step": 9523 + }, + { + "epoch": 0.5, + "learning_rate": 2.1157395839033305e-07, + "loss": 0.1236, + "step": 9524 + }, + { + "epoch": 0.5, + "learning_rate": 2.1154020341928755e-07, + "loss": 0.1095, + "step": 9525 + }, + { + "epoch": 0.5, + "learning_rate": 2.115064481184186e-07, + "loss": 0.1572, + "step": 9526 + }, + { + "epoch": 0.5, + "learning_rate": 2.1147269248869097e-07, + "loss": 0.1453, + "step": 9527 + }, + { + "epoch": 0.5, + "learning_rate": 2.1143893653106927e-07, + "loss": 0.1384, + "step": 9528 + }, + { + "epoch": 0.5, + "learning_rate": 2.1140518024651843e-07, + "loss": 0.145, + "step": 9529 + }, + { + "epoch": 0.5, + "learning_rate": 2.1137142363600305e-07, + "loss": 0.1118, + "step": 9530 + }, + { + "epoch": 0.5, + "learning_rate": 2.113376667004881e-07, + "loss": 0.1734, + "step": 9531 + }, + { + "epoch": 0.5, + "learning_rate": 2.113039094409382e-07, + "loss": 0.1321, + "step": 9532 + }, + { + "epoch": 0.5, + "learning_rate": 2.112701518583182e-07, + "loss": 0.1426, + "step": 9533 + }, + { + "epoch": 0.5, + "learning_rate": 2.1123639395359292e-07, + "loss": 0.1297, + "step": 9534 + }, + { + "epoch": 0.5, + "learning_rate": 2.1120263572772718e-07, + "loss": 0.1187, + "step": 9535 + }, + { + "epoch": 0.5, + "learning_rate": 2.1116887718168574e-07, + "loss": 0.1335, + "step": 9536 + }, + { + "epoch": 0.5, + "learning_rate": 2.111351183164335e-07, + "loss": 0.1349, + "step": 9537 + }, + { + "epoch": 0.5, + "learning_rate": 2.1110135913293527e-07, + "loss": 0.1602, + "step": 9538 + }, + { + "epoch": 0.5, + "learning_rate": 2.1106759963215596e-07, + "loss": 0.1502, + "step": 9539 + }, + { + "epoch": 0.5, + "learning_rate": 2.1103383981506035e-07, + "loss": 0.1175, + "step": 9540 + }, + { + "epoch": 0.5, + "learning_rate": 2.1100007968261334e-07, + "loss": 0.1229, + "step": 9541 + }, + { + "epoch": 0.5, + "learning_rate": 2.1096631923577978e-07, + "loss": 0.1208, + "step": 9542 + }, + { + "epoch": 0.5, + "learning_rate": 2.109325584755246e-07, + "loss": 0.1605, + "step": 9543 + }, + { + "epoch": 0.5, + "learning_rate": 2.1089879740281273e-07, + "loss": 0.1218, + "step": 9544 + }, + { + "epoch": 0.5, + "learning_rate": 2.1086503601860892e-07, + "loss": 0.1482, + "step": 9545 + }, + { + "epoch": 0.5, + "learning_rate": 2.1083127432387827e-07, + "loss": 0.1254, + "step": 9546 + }, + { + "epoch": 0.5, + "learning_rate": 2.1079751231958557e-07, + "loss": 0.1231, + "step": 9547 + }, + { + "epoch": 0.5, + "learning_rate": 2.1076375000669584e-07, + "loss": 0.134, + "step": 9548 + }, + { + "epoch": 0.5, + "learning_rate": 2.1072998738617395e-07, + "loss": 0.1473, + "step": 9549 + }, + { + "epoch": 0.5, + "learning_rate": 2.1069622445898485e-07, + "loss": 0.1517, + "step": 9550 + }, + { + "epoch": 0.5, + "learning_rate": 2.1066246122609362e-07, + "loss": 0.1093, + "step": 9551 + }, + { + "epoch": 0.5, + "learning_rate": 2.1062869768846508e-07, + "loss": 0.1212, + "step": 9552 + }, + { + "epoch": 0.5, + "learning_rate": 2.1059493384706424e-07, + "loss": 0.1567, + "step": 9553 + }, + { + "epoch": 0.5, + "learning_rate": 2.1056116970285611e-07, + "loss": 0.167, + "step": 9554 + }, + { + "epoch": 0.5, + "learning_rate": 2.1052740525680565e-07, + "loss": 0.1607, + "step": 9555 + }, + { + "epoch": 0.5, + "learning_rate": 2.104936405098779e-07, + "loss": 0.1586, + "step": 9556 + }, + { + "epoch": 0.5, + "learning_rate": 2.1045987546303788e-07, + "loss": 0.1111, + "step": 9557 + }, + { + "epoch": 0.5, + "learning_rate": 2.1042611011725055e-07, + "loss": 0.1186, + "step": 9558 + }, + { + "epoch": 0.5, + "learning_rate": 2.10392344473481e-07, + "loss": 0.1422, + "step": 9559 + }, + { + "epoch": 0.5, + "learning_rate": 2.103585785326942e-07, + "loss": 0.1408, + "step": 9560 + }, + { + "epoch": 0.5, + "learning_rate": 2.1032481229585527e-07, + "loss": 0.1244, + "step": 9561 + }, + { + "epoch": 0.5, + "learning_rate": 2.1029104576392919e-07, + "loss": 0.1, + "step": 9562 + }, + { + "epoch": 0.5, + "learning_rate": 2.1025727893788102e-07, + "loss": 0.1379, + "step": 9563 + }, + { + "epoch": 0.5, + "learning_rate": 2.1022351181867592e-07, + "loss": 0.1349, + "step": 9564 + }, + { + "epoch": 0.5, + "learning_rate": 2.1018974440727888e-07, + "loss": 0.1584, + "step": 9565 + }, + { + "epoch": 0.5, + "learning_rate": 2.1015597670465498e-07, + "loss": 0.1676, + "step": 9566 + }, + { + "epoch": 0.5, + "learning_rate": 2.1012220871176938e-07, + "loss": 0.1161, + "step": 9567 + }, + { + "epoch": 0.5, + "learning_rate": 2.100884404295871e-07, + "loss": 0.1172, + "step": 9568 + }, + { + "epoch": 0.5, + "learning_rate": 2.100546718590734e-07, + "loss": 0.1264, + "step": 9569 + }, + { + "epoch": 0.5, + "learning_rate": 2.100209030011932e-07, + "loss": 0.1489, + "step": 9570 + }, + { + "epoch": 0.5, + "learning_rate": 2.0998713385691178e-07, + "loss": 0.1604, + "step": 9571 + }, + { + "epoch": 0.5, + "learning_rate": 2.0995336442719416e-07, + "loss": 0.1536, + "step": 9572 + }, + { + "epoch": 0.5, + "learning_rate": 2.099195947130056e-07, + "loss": 0.1386, + "step": 9573 + }, + { + "epoch": 0.5, + "learning_rate": 2.0988582471531115e-07, + "loss": 0.1519, + "step": 9574 + }, + { + "epoch": 0.5, + "learning_rate": 2.0985205443507601e-07, + "loss": 0.1193, + "step": 9575 + }, + { + "epoch": 0.5, + "learning_rate": 2.098182838732654e-07, + "loss": 0.1423, + "step": 9576 + }, + { + "epoch": 0.5, + "learning_rate": 2.097845130308444e-07, + "loss": 0.1304, + "step": 9577 + }, + { + "epoch": 0.5, + "learning_rate": 2.0975074190877834e-07, + "loss": 0.153, + "step": 9578 + }, + { + "epoch": 0.5, + "learning_rate": 2.0971697050803217e-07, + "loss": 0.1363, + "step": 9579 + }, + { + "epoch": 0.5, + "learning_rate": 2.096831988295713e-07, + "loss": 0.1343, + "step": 9580 + }, + { + "epoch": 0.5, + "learning_rate": 2.0964942687436083e-07, + "loss": 0.1084, + "step": 9581 + }, + { + "epoch": 0.5, + "learning_rate": 2.096156546433661e-07, + "loss": 0.1351, + "step": 9582 + }, + { + "epoch": 0.5, + "learning_rate": 2.0958188213755222e-07, + "loss": 0.1465, + "step": 9583 + }, + { + "epoch": 0.5, + "learning_rate": 2.0954810935788435e-07, + "loss": 0.1557, + "step": 9584 + }, + { + "epoch": 0.5, + "learning_rate": 2.0951433630532796e-07, + "loss": 0.1131, + "step": 9585 + }, + { + "epoch": 0.5, + "learning_rate": 2.0948056298084813e-07, + "loss": 0.1201, + "step": 9586 + }, + { + "epoch": 0.5, + "learning_rate": 2.0944678938541011e-07, + "loss": 0.1673, + "step": 9587 + }, + { + "epoch": 0.5, + "learning_rate": 2.0941301551997923e-07, + "loss": 0.1385, + "step": 9588 + }, + { + "epoch": 0.5, + "learning_rate": 2.0937924138552073e-07, + "loss": 0.1178, + "step": 9589 + }, + { + "epoch": 0.5, + "learning_rate": 2.0934546698299984e-07, + "loss": 0.095, + "step": 9590 + }, + { + "epoch": 0.5, + "learning_rate": 2.09311692313382e-07, + "loss": 0.1088, + "step": 9591 + }, + { + "epoch": 0.5, + "learning_rate": 2.0927791737763232e-07, + "loss": 0.1535, + "step": 9592 + }, + { + "epoch": 0.5, + "learning_rate": 2.0924414217671625e-07, + "loss": 0.151, + "step": 9593 + }, + { + "epoch": 0.5, + "learning_rate": 2.0921036671159893e-07, + "loss": 0.1568, + "step": 9594 + }, + { + "epoch": 0.5, + "learning_rate": 2.0917659098324581e-07, + "loss": 0.1475, + "step": 9595 + }, + { + "epoch": 0.5, + "learning_rate": 2.0914281499262226e-07, + "loss": 0.1293, + "step": 9596 + }, + { + "epoch": 0.5, + "learning_rate": 2.091090387406934e-07, + "loss": 0.1058, + "step": 9597 + }, + { + "epoch": 0.5, + "learning_rate": 2.0907526222842478e-07, + "loss": 0.1506, + "step": 9598 + }, + { + "epoch": 0.5, + "learning_rate": 2.0904148545678164e-07, + "loss": 0.1186, + "step": 9599 + }, + { + "epoch": 0.5, + "learning_rate": 2.090077084267294e-07, + "loss": 0.131, + "step": 9600 + }, + { + "epoch": 0.5, + "learning_rate": 2.0897393113923325e-07, + "loss": 0.1638, + "step": 9601 + }, + { + "epoch": 0.5, + "learning_rate": 2.089401535952588e-07, + "loss": 0.1541, + "step": 9602 + }, + { + "epoch": 0.5, + "learning_rate": 2.0890637579577124e-07, + "loss": 0.16, + "step": 9603 + }, + { + "epoch": 0.5, + "learning_rate": 2.0887259774173605e-07, + "loss": 0.1227, + "step": 9604 + }, + { + "epoch": 0.5, + "learning_rate": 2.0883881943411864e-07, + "loss": 0.1404, + "step": 9605 + }, + { + "epoch": 0.5, + "learning_rate": 2.0880504087388427e-07, + "loss": 0.1476, + "step": 9606 + }, + { + "epoch": 0.5, + "learning_rate": 2.0877126206199848e-07, + "loss": 0.1374, + "step": 9607 + }, + { + "epoch": 0.5, + "learning_rate": 2.0873748299942667e-07, + "loss": 0.1428, + "step": 9608 + }, + { + "epoch": 0.5, + "learning_rate": 2.087037036871342e-07, + "loss": 0.1511, + "step": 9609 + }, + { + "epoch": 0.5, + "learning_rate": 2.0866992412608643e-07, + "loss": 0.1443, + "step": 9610 + }, + { + "epoch": 0.5, + "learning_rate": 2.0863614431724897e-07, + "loss": 0.119, + "step": 9611 + }, + { + "epoch": 0.5, + "learning_rate": 2.086023642615872e-07, + "loss": 0.1543, + "step": 9612 + }, + { + "epoch": 0.5, + "learning_rate": 2.0856858396006647e-07, + "loss": 0.1424, + "step": 9613 + }, + { + "epoch": 0.5, + "learning_rate": 2.085348034136523e-07, + "loss": 0.1772, + "step": 9614 + }, + { + "epoch": 0.5, + "learning_rate": 2.0850102262331015e-07, + "loss": 0.1273, + "step": 9615 + }, + { + "epoch": 0.5, + "learning_rate": 2.084672415900056e-07, + "loss": 0.1239, + "step": 9616 + }, + { + "epoch": 0.5, + "learning_rate": 2.084334603147039e-07, + "loss": 0.1572, + "step": 9617 + }, + { + "epoch": 0.5, + "learning_rate": 2.083996787983707e-07, + "loss": 0.1466, + "step": 9618 + }, + { + "epoch": 0.5, + "learning_rate": 2.083658970419714e-07, + "loss": 0.148, + "step": 9619 + }, + { + "epoch": 0.5, + "learning_rate": 2.083321150464716e-07, + "loss": 0.1322, + "step": 9620 + }, + { + "epoch": 0.5, + "learning_rate": 2.0829833281283666e-07, + "loss": 0.1194, + "step": 9621 + }, + { + "epoch": 0.5, + "learning_rate": 2.082645503420322e-07, + "loss": 0.125, + "step": 9622 + }, + { + "epoch": 0.5, + "learning_rate": 2.082307676350237e-07, + "loss": 0.1219, + "step": 9623 + }, + { + "epoch": 0.5, + "learning_rate": 2.0819698469277664e-07, + "loss": 0.1189, + "step": 9624 + }, + { + "epoch": 0.5, + "learning_rate": 2.081632015162567e-07, + "loss": 0.1512, + "step": 9625 + }, + { + "epoch": 0.5, + "learning_rate": 2.0812941810642922e-07, + "loss": 0.1398, + "step": 9626 + }, + { + "epoch": 0.5, + "learning_rate": 2.0809563446425988e-07, + "loss": 0.1521, + "step": 9627 + }, + { + "epoch": 0.5, + "learning_rate": 2.0806185059071412e-07, + "loss": 0.1643, + "step": 9628 + }, + { + "epoch": 0.5, + "learning_rate": 2.080280664867576e-07, + "loss": 0.1332, + "step": 9629 + }, + { + "epoch": 0.5, + "learning_rate": 2.0799428215335584e-07, + "loss": 0.169, + "step": 9630 + }, + { + "epoch": 0.5, + "learning_rate": 2.079604975914744e-07, + "loss": 0.1568, + "step": 9631 + }, + { + "epoch": 0.5, + "learning_rate": 2.0792671280207886e-07, + "loss": 0.1595, + "step": 9632 + }, + { + "epoch": 0.5, + "learning_rate": 2.0789292778613482e-07, + "loss": 0.1294, + "step": 9633 + }, + { + "epoch": 0.5, + "learning_rate": 2.078591425446079e-07, + "loss": 0.1286, + "step": 9634 + }, + { + "epoch": 0.5, + "learning_rate": 2.078253570784636e-07, + "loss": 0.1309, + "step": 9635 + }, + { + "epoch": 0.5, + "learning_rate": 2.077915713886676e-07, + "loss": 0.1163, + "step": 9636 + }, + { + "epoch": 0.5, + "learning_rate": 2.0775778547618542e-07, + "loss": 0.1265, + "step": 9637 + }, + { + "epoch": 0.5, + "learning_rate": 2.0772399934198281e-07, + "loss": 0.1191, + "step": 9638 + }, + { + "epoch": 0.5, + "learning_rate": 2.0769021298702526e-07, + "loss": 0.1241, + "step": 9639 + }, + { + "epoch": 0.5, + "learning_rate": 2.0765642641227846e-07, + "loss": 0.1515, + "step": 9640 + }, + { + "epoch": 0.5, + "learning_rate": 2.076226396187081e-07, + "loss": 0.1693, + "step": 9641 + }, + { + "epoch": 0.5, + "learning_rate": 2.0758885260727973e-07, + "loss": 0.1454, + "step": 9642 + }, + { + "epoch": 0.5, + "learning_rate": 2.0755506537895903e-07, + "loss": 0.1037, + "step": 9643 + }, + { + "epoch": 0.5, + "learning_rate": 2.0752127793471157e-07, + "loss": 0.1172, + "step": 9644 + }, + { + "epoch": 0.5, + "learning_rate": 2.074874902755032e-07, + "loss": 0.1178, + "step": 9645 + }, + { + "epoch": 0.5, + "learning_rate": 2.0745370240229941e-07, + "loss": 0.1298, + "step": 9646 + }, + { + "epoch": 0.5, + "learning_rate": 2.0741991431606594e-07, + "loss": 0.1125, + "step": 9647 + }, + { + "epoch": 0.5, + "learning_rate": 2.0738612601776838e-07, + "loss": 0.16, + "step": 9648 + }, + { + "epoch": 0.5, + "learning_rate": 2.0735233750837255e-07, + "loss": 0.1231, + "step": 9649 + }, + { + "epoch": 0.5, + "learning_rate": 2.0731854878884407e-07, + "loss": 0.1018, + "step": 9650 + }, + { + "epoch": 0.5, + "learning_rate": 2.0728475986014864e-07, + "loss": 0.124, + "step": 9651 + }, + { + "epoch": 0.5, + "learning_rate": 2.0725097072325198e-07, + "loss": 0.0941, + "step": 9652 + }, + { + "epoch": 0.5, + "learning_rate": 2.0721718137911973e-07, + "loss": 0.1164, + "step": 9653 + }, + { + "epoch": 0.5, + "learning_rate": 2.0718339182871766e-07, + "loss": 0.1302, + "step": 9654 + }, + { + "epoch": 0.5, + "learning_rate": 2.0714960207301155e-07, + "loss": 0.1418, + "step": 9655 + }, + { + "epoch": 0.5, + "learning_rate": 2.0711581211296703e-07, + "loss": 0.1474, + "step": 9656 + }, + { + "epoch": 0.5, + "learning_rate": 2.0708202194954977e-07, + "loss": 0.1098, + "step": 9657 + }, + { + "epoch": 0.5, + "learning_rate": 2.0704823158372564e-07, + "loss": 0.1106, + "step": 9658 + }, + { + "epoch": 0.5, + "learning_rate": 2.0701444101646035e-07, + "loss": 0.1337, + "step": 9659 + }, + { + "epoch": 0.5, + "learning_rate": 2.069806502487196e-07, + "loss": 0.157, + "step": 9660 + }, + { + "epoch": 0.5, + "learning_rate": 2.069468592814692e-07, + "loss": 0.1376, + "step": 9661 + }, + { + "epoch": 0.5, + "learning_rate": 2.0691306811567482e-07, + "loss": 0.1385, + "step": 9662 + }, + { + "epoch": 0.5, + "learning_rate": 2.0687927675230235e-07, + "loss": 0.1119, + "step": 9663 + }, + { + "epoch": 0.5, + "learning_rate": 2.0684548519231746e-07, + "loss": 0.1621, + "step": 9664 + }, + { + "epoch": 0.5, + "learning_rate": 2.0681169343668602e-07, + "loss": 0.1813, + "step": 9665 + }, + { + "epoch": 0.5, + "learning_rate": 2.0677790148637366e-07, + "loss": 0.0997, + "step": 9666 + }, + { + "epoch": 0.5, + "learning_rate": 2.0674410934234633e-07, + "loss": 0.1183, + "step": 9667 + }, + { + "epoch": 0.5, + "learning_rate": 2.0671031700556973e-07, + "loss": 0.1307, + "step": 9668 + }, + { + "epoch": 0.5, + "learning_rate": 2.0667652447700963e-07, + "loss": 0.1594, + "step": 9669 + }, + { + "epoch": 0.5, + "learning_rate": 2.0664273175763194e-07, + "loss": 0.153, + "step": 9670 + }, + { + "epoch": 0.5, + "learning_rate": 2.0660893884840237e-07, + "loss": 0.1564, + "step": 9671 + }, + { + "epoch": 0.5, + "learning_rate": 2.065751457502869e-07, + "loss": 0.1755, + "step": 9672 + }, + { + "epoch": 0.5, + "learning_rate": 2.0654135246425106e-07, + "loss": 0.1344, + "step": 9673 + }, + { + "epoch": 0.5, + "learning_rate": 2.0650755899126095e-07, + "loss": 0.1168, + "step": 9674 + }, + { + "epoch": 0.5, + "learning_rate": 2.064737653322822e-07, + "loss": 0.1406, + "step": 9675 + }, + { + "epoch": 0.51, + "learning_rate": 2.0643997148828084e-07, + "loss": 0.1325, + "step": 9676 + }, + { + "epoch": 0.51, + "learning_rate": 2.0640617746022255e-07, + "loss": 0.1255, + "step": 9677 + }, + { + "epoch": 0.51, + "learning_rate": 2.0637238324907322e-07, + "loss": 0.1289, + "step": 9678 + }, + { + "epoch": 0.51, + "learning_rate": 2.0633858885579878e-07, + "loss": 0.1492, + "step": 9679 + }, + { + "epoch": 0.51, + "learning_rate": 2.0630479428136493e-07, + "loss": 0.1174, + "step": 9680 + }, + { + "epoch": 0.51, + "learning_rate": 2.0627099952673777e-07, + "loss": 0.1302, + "step": 9681 + }, + { + "epoch": 0.51, + "learning_rate": 2.062372045928829e-07, + "loss": 0.1188, + "step": 9682 + }, + { + "epoch": 0.51, + "learning_rate": 2.0620340948076636e-07, + "loss": 0.1318, + "step": 9683 + }, + { + "epoch": 0.51, + "learning_rate": 2.0616961419135396e-07, + "loss": 0.1245, + "step": 9684 + }, + { + "epoch": 0.51, + "learning_rate": 2.061358187256116e-07, + "loss": 0.1247, + "step": 9685 + }, + { + "epoch": 0.51, + "learning_rate": 2.061020230845052e-07, + "loss": 0.1433, + "step": 9686 + }, + { + "epoch": 0.51, + "learning_rate": 2.0606822726900058e-07, + "loss": 0.1411, + "step": 9687 + }, + { + "epoch": 0.51, + "learning_rate": 2.0603443128006375e-07, + "loss": 0.1516, + "step": 9688 + }, + { + "epoch": 0.51, + "learning_rate": 2.060006351186605e-07, + "loss": 0.1195, + "step": 9689 + }, + { + "epoch": 0.51, + "learning_rate": 2.059668387857568e-07, + "loss": 0.1401, + "step": 9690 + }, + { + "epoch": 0.51, + "learning_rate": 2.0593304228231853e-07, + "loss": 0.1543, + "step": 9691 + }, + { + "epoch": 0.51, + "learning_rate": 2.0589924560931164e-07, + "loss": 0.1292, + "step": 9692 + }, + { + "epoch": 0.51, + "learning_rate": 2.0586544876770197e-07, + "loss": 0.1455, + "step": 9693 + }, + { + "epoch": 0.51, + "learning_rate": 2.058316517584556e-07, + "loss": 0.1356, + "step": 9694 + }, + { + "epoch": 0.51, + "learning_rate": 2.0579785458253833e-07, + "loss": 0.1607, + "step": 9695 + }, + { + "epoch": 0.51, + "learning_rate": 2.057640572409161e-07, + "loss": 0.1417, + "step": 9696 + }, + { + "epoch": 0.51, + "learning_rate": 2.0573025973455498e-07, + "loss": 0.1253, + "step": 9697 + }, + { + "epoch": 0.51, + "learning_rate": 2.0569646206442077e-07, + "loss": 0.1448, + "step": 9698 + }, + { + "epoch": 0.51, + "learning_rate": 2.0566266423147947e-07, + "loss": 0.1101, + "step": 9699 + }, + { + "epoch": 0.51, + "learning_rate": 2.0562886623669704e-07, + "loss": 0.1561, + "step": 9700 + }, + { + "epoch": 0.51, + "learning_rate": 2.0559506808103947e-07, + "loss": 0.1314, + "step": 9701 + }, + { + "epoch": 0.51, + "learning_rate": 2.0556126976547267e-07, + "loss": 0.1391, + "step": 9702 + }, + { + "epoch": 0.51, + "learning_rate": 2.0552747129096267e-07, + "loss": 0.1159, + "step": 9703 + }, + { + "epoch": 0.51, + "learning_rate": 2.0549367265847532e-07, + "loss": 0.1677, + "step": 9704 + }, + { + "epoch": 0.51, + "learning_rate": 2.0545987386897674e-07, + "loss": 0.1461, + "step": 9705 + }, + { + "epoch": 0.51, + "learning_rate": 2.054260749234329e-07, + "loss": 0.1532, + "step": 9706 + }, + { + "epoch": 0.51, + "learning_rate": 2.0539227582280967e-07, + "loss": 0.1468, + "step": 9707 + }, + { + "epoch": 0.51, + "learning_rate": 2.0535847656807314e-07, + "loss": 0.1353, + "step": 9708 + }, + { + "epoch": 0.51, + "learning_rate": 2.0532467716018927e-07, + "loss": 0.0889, + "step": 9709 + }, + { + "epoch": 0.51, + "learning_rate": 2.052908776001241e-07, + "loss": 0.1377, + "step": 9710 + }, + { + "epoch": 0.51, + "learning_rate": 2.0525707788884362e-07, + "loss": 0.1245, + "step": 9711 + }, + { + "epoch": 0.51, + "learning_rate": 2.052232780273138e-07, + "loss": 0.1584, + "step": 9712 + }, + { + "epoch": 0.51, + "learning_rate": 2.051894780165006e-07, + "loss": 0.1622, + "step": 9713 + }, + { + "epoch": 0.51, + "learning_rate": 2.0515567785737025e-07, + "loss": 0.1389, + "step": 9714 + }, + { + "epoch": 0.51, + "learning_rate": 2.0512187755088858e-07, + "loss": 0.1187, + "step": 9715 + }, + { + "epoch": 0.51, + "learning_rate": 2.0508807709802162e-07, + "loss": 0.1269, + "step": 9716 + }, + { + "epoch": 0.51, + "learning_rate": 2.0505427649973552e-07, + "loss": 0.1407, + "step": 9717 + }, + { + "epoch": 0.51, + "learning_rate": 2.0502047575699622e-07, + "loss": 0.1355, + "step": 9718 + }, + { + "epoch": 0.51, + "learning_rate": 2.0498667487076982e-07, + "loss": 0.1455, + "step": 9719 + }, + { + "epoch": 0.51, + "learning_rate": 2.0495287384202226e-07, + "loss": 0.1348, + "step": 9720 + }, + { + "epoch": 0.51, + "learning_rate": 2.0491907267171974e-07, + "loss": 0.1589, + "step": 9721 + }, + { + "epoch": 0.51, + "learning_rate": 2.0488527136082815e-07, + "loss": 0.1348, + "step": 9722 + }, + { + "epoch": 0.51, + "learning_rate": 2.0485146991031366e-07, + "loss": 0.1167, + "step": 9723 + }, + { + "epoch": 0.51, + "learning_rate": 2.0481766832114226e-07, + "loss": 0.1509, + "step": 9724 + }, + { + "epoch": 0.51, + "learning_rate": 2.0478386659428004e-07, + "loss": 0.1186, + "step": 9725 + }, + { + "epoch": 0.51, + "learning_rate": 2.047500647306931e-07, + "loss": 0.1337, + "step": 9726 + }, + { + "epoch": 0.51, + "learning_rate": 2.047162627313474e-07, + "loss": 0.1118, + "step": 9727 + }, + { + "epoch": 0.51, + "learning_rate": 2.046824605972092e-07, + "loss": 0.1293, + "step": 9728 + }, + { + "epoch": 0.51, + "learning_rate": 2.0464865832924438e-07, + "loss": 0.1776, + "step": 9729 + }, + { + "epoch": 0.51, + "learning_rate": 2.046148559284192e-07, + "loss": 0.1282, + "step": 9730 + }, + { + "epoch": 0.51, + "learning_rate": 2.0458105339569952e-07, + "loss": 0.1574, + "step": 9731 + }, + { + "epoch": 0.51, + "learning_rate": 2.0454725073205168e-07, + "loss": 0.1554, + "step": 9732 + }, + { + "epoch": 0.51, + "learning_rate": 2.0451344793844165e-07, + "loss": 0.167, + "step": 9733 + }, + { + "epoch": 0.51, + "learning_rate": 2.044796450158355e-07, + "loss": 0.1287, + "step": 9734 + }, + { + "epoch": 0.51, + "learning_rate": 2.0444584196519937e-07, + "loss": 0.1229, + "step": 9735 + }, + { + "epoch": 0.51, + "learning_rate": 2.0441203878749937e-07, + "loss": 0.1275, + "step": 9736 + }, + { + "epoch": 0.51, + "learning_rate": 2.043782354837016e-07, + "loss": 0.1294, + "step": 9737 + }, + { + "epoch": 0.51, + "learning_rate": 2.0434443205477211e-07, + "loss": 0.1436, + "step": 9738 + }, + { + "epoch": 0.51, + "learning_rate": 2.0431062850167716e-07, + "loss": 0.1512, + "step": 9739 + }, + { + "epoch": 0.51, + "learning_rate": 2.0427682482538273e-07, + "loss": 0.1216, + "step": 9740 + }, + { + "epoch": 0.51, + "learning_rate": 2.0424302102685506e-07, + "loss": 0.1365, + "step": 9741 + }, + { + "epoch": 0.51, + "learning_rate": 2.0420921710706014e-07, + "loss": 0.1249, + "step": 9742 + }, + { + "epoch": 0.51, + "learning_rate": 2.0417541306696418e-07, + "loss": 0.1327, + "step": 9743 + }, + { + "epoch": 0.51, + "learning_rate": 2.0414160890753337e-07, + "loss": 0.134, + "step": 9744 + }, + { + "epoch": 0.51, + "learning_rate": 2.041078046297337e-07, + "loss": 0.127, + "step": 9745 + }, + { + "epoch": 0.51, + "learning_rate": 2.0407400023453146e-07, + "loss": 0.1301, + "step": 9746 + }, + { + "epoch": 0.51, + "learning_rate": 2.0404019572289263e-07, + "loss": 0.1579, + "step": 9747 + }, + { + "epoch": 0.51, + "learning_rate": 2.0400639109578351e-07, + "loss": 0.1527, + "step": 9748 + }, + { + "epoch": 0.51, + "learning_rate": 2.0397258635417018e-07, + "loss": 0.1298, + "step": 9749 + }, + { + "epoch": 0.51, + "learning_rate": 2.0393878149901878e-07, + "loss": 0.1584, + "step": 9750 + }, + { + "epoch": 0.51, + "learning_rate": 2.039049765312955e-07, + "loss": 0.1165, + "step": 9751 + }, + { + "epoch": 0.51, + "learning_rate": 2.0387117145196643e-07, + "loss": 0.1903, + "step": 9752 + }, + { + "epoch": 0.51, + "learning_rate": 2.0383736626199788e-07, + "loss": 0.2064, + "step": 9753 + }, + { + "epoch": 0.51, + "learning_rate": 2.0380356096235586e-07, + "loss": 0.1713, + "step": 9754 + }, + { + "epoch": 0.51, + "learning_rate": 2.0376975555400663e-07, + "loss": 0.1329, + "step": 9755 + }, + { + "epoch": 0.51, + "learning_rate": 2.037359500379163e-07, + "loss": 0.1613, + "step": 9756 + }, + { + "epoch": 0.51, + "learning_rate": 2.0370214441505107e-07, + "loss": 0.1615, + "step": 9757 + }, + { + "epoch": 0.51, + "learning_rate": 2.0366833868637716e-07, + "loss": 0.1266, + "step": 9758 + }, + { + "epoch": 0.51, + "learning_rate": 2.0363453285286073e-07, + "loss": 0.1415, + "step": 9759 + }, + { + "epoch": 0.51, + "learning_rate": 2.0360072691546785e-07, + "loss": 0.1193, + "step": 9760 + }, + { + "epoch": 0.51, + "learning_rate": 2.035669208751649e-07, + "loss": 0.1376, + "step": 9761 + }, + { + "epoch": 0.51, + "learning_rate": 2.0353311473291796e-07, + "loss": 0.1105, + "step": 9762 + }, + { + "epoch": 0.51, + "learning_rate": 2.0349930848969322e-07, + "loss": 0.1538, + "step": 9763 + }, + { + "epoch": 0.51, + "learning_rate": 2.034655021464569e-07, + "loss": 0.1454, + "step": 9764 + }, + { + "epoch": 0.51, + "learning_rate": 2.0343169570417515e-07, + "loss": 0.1399, + "step": 9765 + }, + { + "epoch": 0.51, + "learning_rate": 2.0339788916381432e-07, + "loss": 0.1564, + "step": 9766 + }, + { + "epoch": 0.51, + "learning_rate": 2.033640825263404e-07, + "loss": 0.143, + "step": 9767 + }, + { + "epoch": 0.51, + "learning_rate": 2.0333027579271974e-07, + "loss": 0.1376, + "step": 9768 + }, + { + "epoch": 0.51, + "learning_rate": 2.0329646896391846e-07, + "loss": 0.1355, + "step": 9769 + }, + { + "epoch": 0.51, + "learning_rate": 2.032626620409029e-07, + "loss": 0.1359, + "step": 9770 + }, + { + "epoch": 0.51, + "learning_rate": 2.0322885502463919e-07, + "loss": 0.17, + "step": 9771 + }, + { + "epoch": 0.51, + "learning_rate": 2.031950479160935e-07, + "loss": 0.1922, + "step": 9772 + }, + { + "epoch": 0.51, + "learning_rate": 2.0316124071623216e-07, + "loss": 0.1611, + "step": 9773 + }, + { + "epoch": 0.51, + "learning_rate": 2.031274334260213e-07, + "loss": 0.1561, + "step": 9774 + }, + { + "epoch": 0.51, + "learning_rate": 2.0309362604642722e-07, + "loss": 0.1673, + "step": 9775 + }, + { + "epoch": 0.51, + "learning_rate": 2.030598185784161e-07, + "loss": 0.1425, + "step": 9776 + }, + { + "epoch": 0.51, + "learning_rate": 2.030260110229542e-07, + "loss": 0.1279, + "step": 9777 + }, + { + "epoch": 0.51, + "learning_rate": 2.029922033810077e-07, + "loss": 0.1367, + "step": 9778 + }, + { + "epoch": 0.51, + "learning_rate": 2.0295839565354292e-07, + "loss": 0.1301, + "step": 9779 + }, + { + "epoch": 0.51, + "learning_rate": 2.0292458784152605e-07, + "loss": 0.1949, + "step": 9780 + }, + { + "epoch": 0.51, + "learning_rate": 2.0289077994592328e-07, + "loss": 0.1215, + "step": 9781 + }, + { + "epoch": 0.51, + "learning_rate": 2.0285697196770092e-07, + "loss": 0.113, + "step": 9782 + }, + { + "epoch": 0.51, + "learning_rate": 2.0282316390782524e-07, + "loss": 0.1228, + "step": 9783 + }, + { + "epoch": 0.51, + "learning_rate": 2.0278935576726242e-07, + "loss": 0.1446, + "step": 9784 + }, + { + "epoch": 0.51, + "learning_rate": 2.027555475469787e-07, + "loss": 0.1144, + "step": 9785 + }, + { + "epoch": 0.51, + "learning_rate": 2.0272173924794043e-07, + "loss": 0.1327, + "step": 9786 + }, + { + "epoch": 0.51, + "learning_rate": 2.026879308711137e-07, + "loss": 0.1368, + "step": 9787 + }, + { + "epoch": 0.51, + "learning_rate": 2.0265412241746498e-07, + "loss": 0.1379, + "step": 9788 + }, + { + "epoch": 0.51, + "learning_rate": 2.0262031388796038e-07, + "loss": 0.1507, + "step": 9789 + }, + { + "epoch": 0.51, + "learning_rate": 2.0258650528356616e-07, + "loss": 0.1126, + "step": 9790 + }, + { + "epoch": 0.51, + "learning_rate": 2.0255269660524865e-07, + "loss": 0.1244, + "step": 9791 + }, + { + "epoch": 0.51, + "learning_rate": 2.0251888785397412e-07, + "loss": 0.1431, + "step": 9792 + }, + { + "epoch": 0.51, + "learning_rate": 2.024850790307088e-07, + "loss": 0.1478, + "step": 9793 + }, + { + "epoch": 0.51, + "learning_rate": 2.0245127013641888e-07, + "loss": 0.1455, + "step": 9794 + }, + { + "epoch": 0.51, + "learning_rate": 2.0241746117207076e-07, + "loss": 0.1302, + "step": 9795 + }, + { + "epoch": 0.51, + "learning_rate": 2.023836521386307e-07, + "loss": 0.122, + "step": 9796 + }, + { + "epoch": 0.51, + "learning_rate": 2.0234984303706489e-07, + "loss": 0.149, + "step": 9797 + }, + { + "epoch": 0.51, + "learning_rate": 2.023160338683397e-07, + "loss": 0.118, + "step": 9798 + }, + { + "epoch": 0.51, + "learning_rate": 2.022822246334213e-07, + "loss": 0.1773, + "step": 9799 + }, + { + "epoch": 0.51, + "learning_rate": 2.022484153332761e-07, + "loss": 0.1883, + "step": 9800 + }, + { + "epoch": 0.51, + "learning_rate": 2.0221460596887034e-07, + "loss": 0.1234, + "step": 9801 + }, + { + "epoch": 0.51, + "learning_rate": 2.0218079654117028e-07, + "loss": 0.1221, + "step": 9802 + }, + { + "epoch": 0.51, + "learning_rate": 2.0214698705114212e-07, + "loss": 0.1096, + "step": 9803 + }, + { + "epoch": 0.51, + "learning_rate": 2.0211317749975234e-07, + "loss": 0.142, + "step": 9804 + }, + { + "epoch": 0.51, + "learning_rate": 2.0207936788796714e-07, + "loss": 0.1366, + "step": 9805 + }, + { + "epoch": 0.51, + "learning_rate": 2.020455582167527e-07, + "loss": 0.1286, + "step": 9806 + }, + { + "epoch": 0.51, + "learning_rate": 2.0201174848707553e-07, + "loss": 0.1687, + "step": 9807 + }, + { + "epoch": 0.51, + "learning_rate": 2.0197793869990173e-07, + "loss": 0.1699, + "step": 9808 + }, + { + "epoch": 0.51, + "learning_rate": 2.0194412885619778e-07, + "loss": 0.1339, + "step": 9809 + }, + { + "epoch": 0.51, + "learning_rate": 2.0191031895692978e-07, + "loss": 0.1511, + "step": 9810 + }, + { + "epoch": 0.51, + "learning_rate": 2.0187650900306416e-07, + "loss": 0.1769, + "step": 9811 + }, + { + "epoch": 0.51, + "learning_rate": 2.0184269899556714e-07, + "loss": 0.1736, + "step": 9812 + }, + { + "epoch": 0.51, + "learning_rate": 2.0180888893540519e-07, + "loss": 0.1532, + "step": 9813 + }, + { + "epoch": 0.51, + "learning_rate": 2.017750788235444e-07, + "loss": 0.1334, + "step": 9814 + }, + { + "epoch": 0.51, + "learning_rate": 2.017412686609512e-07, + "loss": 0.141, + "step": 9815 + }, + { + "epoch": 0.51, + "learning_rate": 2.0170745844859184e-07, + "loss": 0.1534, + "step": 9816 + }, + { + "epoch": 0.51, + "learning_rate": 2.016736481874327e-07, + "loss": 0.1186, + "step": 9817 + }, + { + "epoch": 0.51, + "learning_rate": 2.0163983787844004e-07, + "loss": 0.1068, + "step": 9818 + }, + { + "epoch": 0.51, + "learning_rate": 2.016060275225801e-07, + "loss": 0.178, + "step": 9819 + }, + { + "epoch": 0.51, + "learning_rate": 2.015722171208194e-07, + "loss": 0.1384, + "step": 9820 + }, + { + "epoch": 0.51, + "learning_rate": 2.0153840667412401e-07, + "loss": 0.1467, + "step": 9821 + }, + { + "epoch": 0.51, + "learning_rate": 2.015045961834605e-07, + "loss": 0.1901, + "step": 9822 + }, + { + "epoch": 0.51, + "learning_rate": 2.0147078564979495e-07, + "loss": 0.1465, + "step": 9823 + }, + { + "epoch": 0.51, + "learning_rate": 2.014369750740938e-07, + "loss": 0.1413, + "step": 9824 + }, + { + "epoch": 0.51, + "learning_rate": 2.0140316445732334e-07, + "loss": 0.1485, + "step": 9825 + }, + { + "epoch": 0.51, + "learning_rate": 2.013693538004499e-07, + "loss": 0.1941, + "step": 9826 + }, + { + "epoch": 0.51, + "learning_rate": 2.0133554310443986e-07, + "loss": 0.147, + "step": 9827 + }, + { + "epoch": 0.51, + "learning_rate": 2.0130173237025935e-07, + "loss": 0.0929, + "step": 9828 + }, + { + "epoch": 0.51, + "learning_rate": 2.0126792159887493e-07, + "loss": 0.118, + "step": 9829 + }, + { + "epoch": 0.51, + "learning_rate": 2.012341107912528e-07, + "loss": 0.1366, + "step": 9830 + }, + { + "epoch": 0.51, + "learning_rate": 2.0120029994835934e-07, + "loss": 0.1565, + "step": 9831 + }, + { + "epoch": 0.51, + "learning_rate": 2.0116648907116082e-07, + "loss": 0.1211, + "step": 9832 + }, + { + "epoch": 0.51, + "learning_rate": 2.0113267816062355e-07, + "loss": 0.1171, + "step": 9833 + }, + { + "epoch": 0.51, + "learning_rate": 2.0109886721771395e-07, + "loss": 0.1104, + "step": 9834 + }, + { + "epoch": 0.51, + "learning_rate": 2.0106505624339828e-07, + "loss": 0.1149, + "step": 9835 + }, + { + "epoch": 0.51, + "learning_rate": 2.010312452386429e-07, + "loss": 0.1292, + "step": 9836 + }, + { + "epoch": 0.51, + "learning_rate": 2.009974342044141e-07, + "loss": 0.1392, + "step": 9837 + }, + { + "epoch": 0.51, + "learning_rate": 2.0096362314167829e-07, + "loss": 0.1175, + "step": 9838 + }, + { + "epoch": 0.51, + "learning_rate": 2.0092981205140176e-07, + "loss": 0.1352, + "step": 9839 + }, + { + "epoch": 0.51, + "learning_rate": 2.0089600093455083e-07, + "loss": 0.1287, + "step": 9840 + }, + { + "epoch": 0.51, + "learning_rate": 2.008621897920918e-07, + "loss": 0.1866, + "step": 9841 + }, + { + "epoch": 0.51, + "learning_rate": 2.0082837862499115e-07, + "loss": 0.1254, + "step": 9842 + }, + { + "epoch": 0.51, + "learning_rate": 2.0079456743421504e-07, + "loss": 0.1399, + "step": 9843 + }, + { + "epoch": 0.51, + "learning_rate": 2.007607562207299e-07, + "loss": 0.1328, + "step": 9844 + }, + { + "epoch": 0.51, + "learning_rate": 2.0072694498550206e-07, + "loss": 0.1426, + "step": 9845 + }, + { + "epoch": 0.51, + "learning_rate": 2.0069313372949782e-07, + "loss": 0.1368, + "step": 9846 + }, + { + "epoch": 0.51, + "learning_rate": 2.0065932245368363e-07, + "loss": 0.1376, + "step": 9847 + }, + { + "epoch": 0.51, + "learning_rate": 2.0062551115902568e-07, + "loss": 0.1779, + "step": 9848 + }, + { + "epoch": 0.51, + "learning_rate": 2.0059169984649044e-07, + "loss": 0.1316, + "step": 9849 + }, + { + "epoch": 0.51, + "learning_rate": 2.005578885170441e-07, + "loss": 0.1563, + "step": 9850 + }, + { + "epoch": 0.51, + "learning_rate": 2.005240771716532e-07, + "loss": 0.1221, + "step": 9851 + }, + { + "epoch": 0.51, + "learning_rate": 2.0049026581128392e-07, + "loss": 0.1276, + "step": 9852 + }, + { + "epoch": 0.51, + "learning_rate": 2.0045645443690265e-07, + "loss": 0.132, + "step": 9853 + }, + { + "epoch": 0.51, + "learning_rate": 2.0042264304947572e-07, + "loss": 0.1074, + "step": 9854 + }, + { + "epoch": 0.51, + "learning_rate": 2.0038883164996947e-07, + "loss": 0.1366, + "step": 9855 + }, + { + "epoch": 0.51, + "learning_rate": 2.0035502023935038e-07, + "loss": 0.1448, + "step": 9856 + }, + { + "epoch": 0.51, + "learning_rate": 2.0032120881858458e-07, + "loss": 0.1315, + "step": 9857 + }, + { + "epoch": 0.51, + "learning_rate": 2.002873973886385e-07, + "loss": 0.139, + "step": 9858 + }, + { + "epoch": 0.51, + "learning_rate": 2.002535859504785e-07, + "loss": 0.1236, + "step": 9859 + }, + { + "epoch": 0.51, + "learning_rate": 2.0021977450507093e-07, + "loss": 0.1406, + "step": 9860 + }, + { + "epoch": 0.51, + "learning_rate": 2.0018596305338212e-07, + "loss": 0.1215, + "step": 9861 + }, + { + "epoch": 0.51, + "learning_rate": 2.001521515963784e-07, + "loss": 0.1278, + "step": 9862 + }, + { + "epoch": 0.51, + "learning_rate": 2.0011834013502617e-07, + "loss": 0.1393, + "step": 9863 + }, + { + "epoch": 0.51, + "learning_rate": 2.0008452867029165e-07, + "loss": 0.1434, + "step": 9864 + }, + { + "epoch": 0.51, + "learning_rate": 2.0005071720314137e-07, + "loss": 0.1396, + "step": 9865 + }, + { + "epoch": 0.51, + "learning_rate": 2.000169057345415e-07, + "loss": 0.1203, + "step": 9866 + }, + { + "epoch": 0.52, + "learning_rate": 1.9998309426545853e-07, + "loss": 0.1388, + "step": 9867 + }, + { + "epoch": 0.52, + "learning_rate": 1.999492827968587e-07, + "loss": 0.1209, + "step": 9868 + }, + { + "epoch": 0.52, + "learning_rate": 1.9991547132970836e-07, + "loss": 0.1172, + "step": 9869 + }, + { + "epoch": 0.52, + "learning_rate": 1.9988165986497384e-07, + "loss": 0.1026, + "step": 9870 + }, + { + "epoch": 0.52, + "learning_rate": 1.998478484036216e-07, + "loss": 0.1919, + "step": 9871 + }, + { + "epoch": 0.52, + "learning_rate": 1.9981403694661786e-07, + "loss": 0.1844, + "step": 9872 + }, + { + "epoch": 0.52, + "learning_rate": 1.9978022549492908e-07, + "loss": 0.1649, + "step": 9873 + }, + { + "epoch": 0.52, + "learning_rate": 1.9974641404952148e-07, + "loss": 0.1472, + "step": 9874 + }, + { + "epoch": 0.52, + "learning_rate": 1.9971260261136148e-07, + "loss": 0.1112, + "step": 9875 + }, + { + "epoch": 0.52, + "learning_rate": 1.996787911814155e-07, + "loss": 0.131, + "step": 9876 + }, + { + "epoch": 0.52, + "learning_rate": 1.9964497976064968e-07, + "loss": 0.1348, + "step": 9877 + }, + { + "epoch": 0.52, + "learning_rate": 1.9961116835003054e-07, + "loss": 0.1758, + "step": 9878 + }, + { + "epoch": 0.52, + "learning_rate": 1.995773569505243e-07, + "loss": 0.1345, + "step": 9879 + }, + { + "epoch": 0.52, + "learning_rate": 1.9954354556309739e-07, + "loss": 0.1121, + "step": 9880 + }, + { + "epoch": 0.52, + "learning_rate": 1.995097341887161e-07, + "loss": 0.1145, + "step": 9881 + }, + { + "epoch": 0.52, + "learning_rate": 1.9947592282834683e-07, + "loss": 0.0918, + "step": 9882 + }, + { + "epoch": 0.52, + "learning_rate": 1.9944211148295585e-07, + "loss": 0.1314, + "step": 9883 + }, + { + "epoch": 0.52, + "learning_rate": 1.994083001535096e-07, + "loss": 0.1416, + "step": 9884 + }, + { + "epoch": 0.52, + "learning_rate": 1.9937448884097435e-07, + "loss": 0.166, + "step": 9885 + }, + { + "epoch": 0.52, + "learning_rate": 1.9934067754631638e-07, + "loss": 0.1301, + "step": 9886 + }, + { + "epoch": 0.52, + "learning_rate": 1.9930686627050218e-07, + "loss": 0.1458, + "step": 9887 + }, + { + "epoch": 0.52, + "learning_rate": 1.9927305501449795e-07, + "loss": 0.1225, + "step": 9888 + }, + { + "epoch": 0.52, + "learning_rate": 1.992392437792701e-07, + "loss": 0.1409, + "step": 9889 + }, + { + "epoch": 0.52, + "learning_rate": 1.9920543256578494e-07, + "loss": 0.161, + "step": 9890 + }, + { + "epoch": 0.52, + "learning_rate": 1.991716213750089e-07, + "loss": 0.1259, + "step": 9891 + }, + { + "epoch": 0.52, + "learning_rate": 1.9913781020790815e-07, + "loss": 0.1131, + "step": 9892 + }, + { + "epoch": 0.52, + "learning_rate": 1.991039990654492e-07, + "loss": 0.1276, + "step": 9893 + }, + { + "epoch": 0.52, + "learning_rate": 1.990701879485983e-07, + "loss": 0.156, + "step": 9894 + }, + { + "epoch": 0.52, + "learning_rate": 1.9903637685832172e-07, + "loss": 0.1517, + "step": 9895 + }, + { + "epoch": 0.52, + "learning_rate": 1.9900256579558592e-07, + "loss": 0.1704, + "step": 9896 + }, + { + "epoch": 0.52, + "learning_rate": 1.989687547613571e-07, + "loss": 0.1134, + "step": 9897 + }, + { + "epoch": 0.52, + "learning_rate": 1.9893494375660173e-07, + "loss": 0.1282, + "step": 9898 + }, + { + "epoch": 0.52, + "learning_rate": 1.9890113278228603e-07, + "loss": 0.1661, + "step": 9899 + }, + { + "epoch": 0.52, + "learning_rate": 1.9886732183937643e-07, + "loss": 0.12, + "step": 9900 + }, + { + "epoch": 0.52, + "learning_rate": 1.9883351092883927e-07, + "loss": 0.1237, + "step": 9901 + }, + { + "epoch": 0.52, + "learning_rate": 1.9879970005164072e-07, + "loss": 0.1063, + "step": 9902 + }, + { + "epoch": 0.52, + "learning_rate": 1.987658892087472e-07, + "loss": 0.1676, + "step": 9903 + }, + { + "epoch": 0.52, + "learning_rate": 1.9873207840112508e-07, + "loss": 0.1501, + "step": 9904 + }, + { + "epoch": 0.52, + "learning_rate": 1.9869826762974063e-07, + "loss": 0.126, + "step": 9905 + }, + { + "epoch": 0.52, + "learning_rate": 1.9866445689556018e-07, + "loss": 0.1828, + "step": 9906 + }, + { + "epoch": 0.52, + "learning_rate": 1.986306461995501e-07, + "loss": 0.1018, + "step": 9907 + }, + { + "epoch": 0.52, + "learning_rate": 1.9859683554267664e-07, + "loss": 0.087, + "step": 9908 + }, + { + "epoch": 0.52, + "learning_rate": 1.985630249259062e-07, + "loss": 0.2077, + "step": 9909 + }, + { + "epoch": 0.52, + "learning_rate": 1.985292143502051e-07, + "loss": 0.1073, + "step": 9910 + }, + { + "epoch": 0.52, + "learning_rate": 1.9849540381653956e-07, + "loss": 0.1368, + "step": 9911 + }, + { + "epoch": 0.52, + "learning_rate": 1.9846159332587597e-07, + "loss": 0.13, + "step": 9912 + }, + { + "epoch": 0.52, + "learning_rate": 1.9842778287918064e-07, + "loss": 0.1316, + "step": 9913 + }, + { + "epoch": 0.52, + "learning_rate": 1.983939724774199e-07, + "loss": 0.1333, + "step": 9914 + }, + { + "epoch": 0.52, + "learning_rate": 1.9836016212156e-07, + "loss": 0.1547, + "step": 9915 + }, + { + "epoch": 0.52, + "learning_rate": 1.9832635181256734e-07, + "loss": 0.1262, + "step": 9916 + }, + { + "epoch": 0.52, + "learning_rate": 1.9829254155140814e-07, + "loss": 0.136, + "step": 9917 + }, + { + "epoch": 0.52, + "learning_rate": 1.9825873133904887e-07, + "loss": 0.1537, + "step": 9918 + }, + { + "epoch": 0.52, + "learning_rate": 1.9822492117645567e-07, + "loss": 0.1109, + "step": 9919 + }, + { + "epoch": 0.52, + "learning_rate": 1.9819111106459487e-07, + "loss": 0.1567, + "step": 9920 + }, + { + "epoch": 0.52, + "learning_rate": 1.9815730100443287e-07, + "loss": 0.1222, + "step": 9921 + }, + { + "epoch": 0.52, + "learning_rate": 1.9812349099693587e-07, + "loss": 0.1387, + "step": 9922 + }, + { + "epoch": 0.52, + "learning_rate": 1.9808968104307025e-07, + "loss": 0.1045, + "step": 9923 + }, + { + "epoch": 0.52, + "learning_rate": 1.9805587114380223e-07, + "loss": 0.1443, + "step": 9924 + }, + { + "epoch": 0.52, + "learning_rate": 1.9802206130009825e-07, + "loss": 0.1189, + "step": 9925 + }, + { + "epoch": 0.52, + "learning_rate": 1.9798825151292445e-07, + "loss": 0.1266, + "step": 9926 + }, + { + "epoch": 0.52, + "learning_rate": 1.9795444178324728e-07, + "loss": 0.1267, + "step": 9927 + }, + { + "epoch": 0.52, + "learning_rate": 1.9792063211203292e-07, + "loss": 0.1513, + "step": 9928 + }, + { + "epoch": 0.52, + "learning_rate": 1.9788682250024767e-07, + "loss": 0.1652, + "step": 9929 + }, + { + "epoch": 0.52, + "learning_rate": 1.9785301294885786e-07, + "loss": 0.1246, + "step": 9930 + }, + { + "epoch": 0.52, + "learning_rate": 1.9781920345882975e-07, + "loss": 0.1408, + "step": 9931 + }, + { + "epoch": 0.52, + "learning_rate": 1.9778539403112967e-07, + "loss": 0.1497, + "step": 9932 + }, + { + "epoch": 0.52, + "learning_rate": 1.9775158466672386e-07, + "loss": 0.1292, + "step": 9933 + }, + { + "epoch": 0.52, + "learning_rate": 1.9771777536657866e-07, + "loss": 0.1438, + "step": 9934 + }, + { + "epoch": 0.52, + "learning_rate": 1.9768396613166028e-07, + "loss": 0.1109, + "step": 9935 + }, + { + "epoch": 0.52, + "learning_rate": 1.9765015696293512e-07, + "loss": 0.1393, + "step": 9936 + }, + { + "epoch": 0.52, + "learning_rate": 1.9761634786136935e-07, + "loss": 0.145, + "step": 9937 + }, + { + "epoch": 0.52, + "learning_rate": 1.9758253882792923e-07, + "loss": 0.14, + "step": 9938 + }, + { + "epoch": 0.52, + "learning_rate": 1.9754872986358113e-07, + "loss": 0.1084, + "step": 9939 + }, + { + "epoch": 0.52, + "learning_rate": 1.9751492096929123e-07, + "loss": 0.1258, + "step": 9940 + }, + { + "epoch": 0.52, + "learning_rate": 1.974811121460259e-07, + "loss": 0.1308, + "step": 9941 + }, + { + "epoch": 0.52, + "learning_rate": 1.974473033947513e-07, + "loss": 0.1341, + "step": 9942 + }, + { + "epoch": 0.52, + "learning_rate": 1.974134947164338e-07, + "loss": 0.1161, + "step": 9943 + }, + { + "epoch": 0.52, + "learning_rate": 1.9737968611203968e-07, + "loss": 0.1305, + "step": 9944 + }, + { + "epoch": 0.52, + "learning_rate": 1.9734587758253505e-07, + "loss": 0.1231, + "step": 9945 + }, + { + "epoch": 0.52, + "learning_rate": 1.973120691288863e-07, + "loss": 0.1397, + "step": 9946 + }, + { + "epoch": 0.52, + "learning_rate": 1.972782607520596e-07, + "loss": 0.1234, + "step": 9947 + }, + { + "epoch": 0.52, + "learning_rate": 1.972444524530213e-07, + "loss": 0.1382, + "step": 9948 + }, + { + "epoch": 0.52, + "learning_rate": 1.9721064423273759e-07, + "loss": 0.1718, + "step": 9949 + }, + { + "epoch": 0.52, + "learning_rate": 1.9717683609217477e-07, + "loss": 0.1343, + "step": 9950 + }, + { + "epoch": 0.52, + "learning_rate": 1.9714302803229904e-07, + "loss": 0.1204, + "step": 9951 + }, + { + "epoch": 0.52, + "learning_rate": 1.971092200540767e-07, + "loss": 0.1285, + "step": 9952 + }, + { + "epoch": 0.52, + "learning_rate": 1.97075412158474e-07, + "loss": 0.133, + "step": 9953 + }, + { + "epoch": 0.52, + "learning_rate": 1.9704160434645709e-07, + "loss": 0.1247, + "step": 9954 + }, + { + "epoch": 0.52, + "learning_rate": 1.9700779661899232e-07, + "loss": 0.1472, + "step": 9955 + }, + { + "epoch": 0.52, + "learning_rate": 1.969739889770458e-07, + "loss": 0.1535, + "step": 9956 + }, + { + "epoch": 0.52, + "learning_rate": 1.969401814215839e-07, + "loss": 0.131, + "step": 9957 + }, + { + "epoch": 0.52, + "learning_rate": 1.9690637395357276e-07, + "loss": 0.1263, + "step": 9958 + }, + { + "epoch": 0.52, + "learning_rate": 1.9687256657397867e-07, + "loss": 0.1324, + "step": 9959 + }, + { + "epoch": 0.52, + "learning_rate": 1.968387592837678e-07, + "loss": 0.1048, + "step": 9960 + }, + { + "epoch": 0.52, + "learning_rate": 1.9680495208390653e-07, + "loss": 0.1088, + "step": 9961 + }, + { + "epoch": 0.52, + "learning_rate": 1.9677114497536088e-07, + "loss": 0.1123, + "step": 9962 + }, + { + "epoch": 0.52, + "learning_rate": 1.9673733795909713e-07, + "loss": 0.1119, + "step": 9963 + }, + { + "epoch": 0.52, + "learning_rate": 1.9670353103608154e-07, + "loss": 0.135, + "step": 9964 + }, + { + "epoch": 0.52, + "learning_rate": 1.9666972420728027e-07, + "loss": 0.1303, + "step": 9965 + }, + { + "epoch": 0.52, + "learning_rate": 1.9663591747365964e-07, + "loss": 0.1081, + "step": 9966 + }, + { + "epoch": 0.52, + "learning_rate": 1.9660211083618572e-07, + "loss": 0.1424, + "step": 9967 + }, + { + "epoch": 0.52, + "learning_rate": 1.9656830429582484e-07, + "loss": 0.1277, + "step": 9968 + }, + { + "epoch": 0.52, + "learning_rate": 1.965344978535431e-07, + "loss": 0.1133, + "step": 9969 + }, + { + "epoch": 0.52, + "learning_rate": 1.9650069151030682e-07, + "loss": 0.1208, + "step": 9970 + }, + { + "epoch": 0.52, + "learning_rate": 1.9646688526708208e-07, + "loss": 0.1052, + "step": 9971 + }, + { + "epoch": 0.52, + "learning_rate": 1.964330791248351e-07, + "loss": 0.127, + "step": 9972 + }, + { + "epoch": 0.52, + "learning_rate": 1.9639927308453214e-07, + "loss": 0.1272, + "step": 9973 + }, + { + "epoch": 0.52, + "learning_rate": 1.9636546714713928e-07, + "loss": 0.1282, + "step": 9974 + }, + { + "epoch": 0.52, + "learning_rate": 1.9633166131362284e-07, + "loss": 0.1051, + "step": 9975 + }, + { + "epoch": 0.52, + "learning_rate": 1.9629785558494886e-07, + "loss": 0.1332, + "step": 9976 + }, + { + "epoch": 0.52, + "learning_rate": 1.9626404996208367e-07, + "loss": 0.1175, + "step": 9977 + }, + { + "epoch": 0.52, + "learning_rate": 1.9623024444599344e-07, + "loss": 0.1646, + "step": 9978 + }, + { + "epoch": 0.52, + "learning_rate": 1.9619643903764418e-07, + "loss": 0.1261, + "step": 9979 + }, + { + "epoch": 0.52, + "learning_rate": 1.9616263373800215e-07, + "loss": 0.1328, + "step": 9980 + }, + { + "epoch": 0.52, + "learning_rate": 1.9612882854803355e-07, + "loss": 0.1352, + "step": 9981 + }, + { + "epoch": 0.52, + "learning_rate": 1.9609502346870453e-07, + "loss": 0.1522, + "step": 9982 + }, + { + "epoch": 0.52, + "learning_rate": 1.9606121850098126e-07, + "loss": 0.1356, + "step": 9983 + }, + { + "epoch": 0.52, + "learning_rate": 1.9602741364582983e-07, + "loss": 0.129, + "step": 9984 + }, + { + "epoch": 0.52, + "learning_rate": 1.959936089042165e-07, + "loss": 0.1045, + "step": 9985 + }, + { + "epoch": 0.52, + "learning_rate": 1.9595980427710732e-07, + "loss": 0.1696, + "step": 9986 + }, + { + "epoch": 0.52, + "learning_rate": 1.959259997654686e-07, + "loss": 0.1137, + "step": 9987 + }, + { + "epoch": 0.52, + "learning_rate": 1.9589219537026633e-07, + "loss": 0.1431, + "step": 9988 + }, + { + "epoch": 0.52, + "learning_rate": 1.9585839109246666e-07, + "loss": 0.1319, + "step": 9989 + }, + { + "epoch": 0.52, + "learning_rate": 1.9582458693303583e-07, + "loss": 0.1833, + "step": 9990 + }, + { + "epoch": 0.52, + "learning_rate": 1.9579078289293984e-07, + "loss": 0.1453, + "step": 9991 + }, + { + "epoch": 0.52, + "learning_rate": 1.9575697897314498e-07, + "loss": 0.135, + "step": 9992 + }, + { + "epoch": 0.52, + "learning_rate": 1.9572317517461723e-07, + "loss": 0.124, + "step": 9993 + }, + { + "epoch": 0.52, + "learning_rate": 1.9568937149832282e-07, + "loss": 0.1344, + "step": 9994 + }, + { + "epoch": 0.52, + "learning_rate": 1.9565556794522782e-07, + "loss": 0.1507, + "step": 9995 + }, + { + "epoch": 0.52, + "learning_rate": 1.9562176451629844e-07, + "loss": 0.1638, + "step": 9996 + }, + { + "epoch": 0.52, + "learning_rate": 1.955879612125007e-07, + "loss": 0.1501, + "step": 9997 + }, + { + "epoch": 0.52, + "learning_rate": 1.9555415803480064e-07, + "loss": 0.1139, + "step": 9998 + }, + { + "epoch": 0.52, + "learning_rate": 1.9552035498416454e-07, + "loss": 0.1305, + "step": 9999 + }, + { + "epoch": 0.52, + "learning_rate": 1.9548655206155838e-07, + "loss": 0.1633, + "step": 10000 + }, + { + "epoch": 0.52, + "learning_rate": 1.9545274926794832e-07, + "loss": 0.1202, + "step": 10001 + }, + { + "epoch": 0.52, + "learning_rate": 1.9541894660430044e-07, + "loss": 0.1437, + "step": 10002 + }, + { + "epoch": 0.52, + "learning_rate": 1.9538514407158084e-07, + "loss": 0.1285, + "step": 10003 + }, + { + "epoch": 0.52, + "learning_rate": 1.9535134167075566e-07, + "loss": 0.1101, + "step": 10004 + }, + { + "epoch": 0.52, + "learning_rate": 1.9531753940279082e-07, + "loss": 0.1555, + "step": 10005 + }, + { + "epoch": 0.52, + "learning_rate": 1.9528373726865258e-07, + "loss": 0.1341, + "step": 10006 + }, + { + "epoch": 0.52, + "learning_rate": 1.9524993526930693e-07, + "loss": 0.1346, + "step": 10007 + }, + { + "epoch": 0.52, + "learning_rate": 1.9521613340572e-07, + "loss": 0.141, + "step": 10008 + }, + { + "epoch": 0.52, + "learning_rate": 1.9518233167885775e-07, + "loss": 0.1405, + "step": 10009 + }, + { + "epoch": 0.52, + "learning_rate": 1.9514853008968637e-07, + "loss": 0.1398, + "step": 10010 + }, + { + "epoch": 0.52, + "learning_rate": 1.9511472863917183e-07, + "loss": 0.151, + "step": 10011 + }, + { + "epoch": 0.52, + "learning_rate": 1.9508092732828027e-07, + "loss": 0.1648, + "step": 10012 + }, + { + "epoch": 0.52, + "learning_rate": 1.9504712615797775e-07, + "loss": 0.1586, + "step": 10013 + }, + { + "epoch": 0.52, + "learning_rate": 1.9501332512923022e-07, + "loss": 0.1444, + "step": 10014 + }, + { + "epoch": 0.52, + "learning_rate": 1.949795242430038e-07, + "loss": 0.1383, + "step": 10015 + }, + { + "epoch": 0.52, + "learning_rate": 1.949457235002645e-07, + "loss": 0.142, + "step": 10016 + }, + { + "epoch": 0.52, + "learning_rate": 1.9491192290197839e-07, + "loss": 0.1226, + "step": 10017 + }, + { + "epoch": 0.52, + "learning_rate": 1.9487812244911143e-07, + "loss": 0.156, + "step": 10018 + }, + { + "epoch": 0.52, + "learning_rate": 1.9484432214262978e-07, + "loss": 0.1337, + "step": 10019 + }, + { + "epoch": 0.52, + "learning_rate": 1.9481052198349932e-07, + "loss": 0.1497, + "step": 10020 + }, + { + "epoch": 0.52, + "learning_rate": 1.9477672197268625e-07, + "loss": 0.126, + "step": 10021 + }, + { + "epoch": 0.52, + "learning_rate": 1.9474292211115647e-07, + "loss": 0.1623, + "step": 10022 + }, + { + "epoch": 0.52, + "learning_rate": 1.947091223998759e-07, + "loss": 0.1563, + "step": 10023 + }, + { + "epoch": 0.52, + "learning_rate": 1.9467532283981074e-07, + "loss": 0.1316, + "step": 10024 + }, + { + "epoch": 0.52, + "learning_rate": 1.9464152343192687e-07, + "loss": 0.1586, + "step": 10025 + }, + { + "epoch": 0.52, + "learning_rate": 1.9460772417719037e-07, + "loss": 0.1114, + "step": 10026 + }, + { + "epoch": 0.52, + "learning_rate": 1.9457392507656712e-07, + "loss": 0.1164, + "step": 10027 + }, + { + "epoch": 0.52, + "learning_rate": 1.9454012613102324e-07, + "loss": 0.1402, + "step": 10028 + }, + { + "epoch": 0.52, + "learning_rate": 1.9450632734152464e-07, + "loss": 0.1065, + "step": 10029 + }, + { + "epoch": 0.52, + "learning_rate": 1.944725287090374e-07, + "loss": 0.1365, + "step": 10030 + }, + { + "epoch": 0.52, + "learning_rate": 1.9443873023452737e-07, + "loss": 0.1383, + "step": 10031 + }, + { + "epoch": 0.52, + "learning_rate": 1.9440493191896054e-07, + "loss": 0.122, + "step": 10032 + }, + { + "epoch": 0.52, + "learning_rate": 1.9437113376330297e-07, + "loss": 0.1552, + "step": 10033 + }, + { + "epoch": 0.52, + "learning_rate": 1.943373357685205e-07, + "loss": 0.1275, + "step": 10034 + }, + { + "epoch": 0.52, + "learning_rate": 1.9430353793557926e-07, + "loss": 0.1478, + "step": 10035 + }, + { + "epoch": 0.52, + "learning_rate": 1.94269740265445e-07, + "loss": 0.1005, + "step": 10036 + }, + { + "epoch": 0.52, + "learning_rate": 1.9423594275908384e-07, + "loss": 0.14, + "step": 10037 + }, + { + "epoch": 0.52, + "learning_rate": 1.9420214541746173e-07, + "loss": 0.143, + "step": 10038 + }, + { + "epoch": 0.52, + "learning_rate": 1.9416834824154443e-07, + "loss": 0.1278, + "step": 10039 + }, + { + "epoch": 0.52, + "learning_rate": 1.9413455123229801e-07, + "loss": 0.1151, + "step": 10040 + }, + { + "epoch": 0.52, + "learning_rate": 1.941007543906884e-07, + "loss": 0.1353, + "step": 10041 + }, + { + "epoch": 0.52, + "learning_rate": 1.9406695771768148e-07, + "loss": 0.1392, + "step": 10042 + }, + { + "epoch": 0.52, + "learning_rate": 1.9403316121424317e-07, + "loss": 0.1315, + "step": 10043 + }, + { + "epoch": 0.52, + "learning_rate": 1.939993648813395e-07, + "loss": 0.1981, + "step": 10044 + }, + { + "epoch": 0.52, + "learning_rate": 1.9396556871993624e-07, + "loss": 0.1265, + "step": 10045 + }, + { + "epoch": 0.52, + "learning_rate": 1.939317727309994e-07, + "loss": 0.1179, + "step": 10046 + }, + { + "epoch": 0.52, + "learning_rate": 1.9389797691549486e-07, + "loss": 0.1383, + "step": 10047 + }, + { + "epoch": 0.52, + "learning_rate": 1.9386418127438842e-07, + "loss": 0.152, + "step": 10048 + }, + { + "epoch": 0.52, + "learning_rate": 1.9383038580864608e-07, + "loss": 0.126, + "step": 10049 + }, + { + "epoch": 0.52, + "learning_rate": 1.9379659051923368e-07, + "loss": 0.1274, + "step": 10050 + }, + { + "epoch": 0.52, + "learning_rate": 1.9376279540711713e-07, + "loss": 0.1603, + "step": 10051 + }, + { + "epoch": 0.52, + "learning_rate": 1.9372900047326227e-07, + "loss": 0.1617, + "step": 10052 + }, + { + "epoch": 0.52, + "learning_rate": 1.9369520571863503e-07, + "loss": 0.1167, + "step": 10053 + }, + { + "epoch": 0.52, + "learning_rate": 1.936614111442012e-07, + "loss": 0.1705, + "step": 10054 + }, + { + "epoch": 0.52, + "learning_rate": 1.936276167509268e-07, + "loss": 0.1229, + "step": 10055 + }, + { + "epoch": 0.52, + "learning_rate": 1.9359382253977749e-07, + "loss": 0.095, + "step": 10056 + }, + { + "epoch": 0.52, + "learning_rate": 1.9356002851171917e-07, + "loss": 0.1652, + "step": 10057 + }, + { + "epoch": 0.52, + "learning_rate": 1.9352623466771778e-07, + "loss": 0.1513, + "step": 10058 + }, + { + "epoch": 0.53, + "learning_rate": 1.9349244100873906e-07, + "loss": 0.14, + "step": 10059 + }, + { + "epoch": 0.53, + "learning_rate": 1.9345864753574895e-07, + "loss": 0.1283, + "step": 10060 + }, + { + "epoch": 0.53, + "learning_rate": 1.9342485424971312e-07, + "loss": 0.1391, + "step": 10061 + }, + { + "epoch": 0.53, + "learning_rate": 1.9339106115159759e-07, + "loss": 0.1381, + "step": 10062 + }, + { + "epoch": 0.53, + "learning_rate": 1.9335726824236802e-07, + "loss": 0.137, + "step": 10063 + }, + { + "epoch": 0.53, + "learning_rate": 1.9332347552299037e-07, + "loss": 0.1281, + "step": 10064 + }, + { + "epoch": 0.53, + "learning_rate": 1.9328968299443033e-07, + "loss": 0.1408, + "step": 10065 + }, + { + "epoch": 0.53, + "learning_rate": 1.9325589065765368e-07, + "loss": 0.1414, + "step": 10066 + }, + { + "epoch": 0.53, + "learning_rate": 1.9322209851362635e-07, + "loss": 0.1666, + "step": 10067 + }, + { + "epoch": 0.53, + "learning_rate": 1.93188306563314e-07, + "loss": 0.1462, + "step": 10068 + }, + { + "epoch": 0.53, + "learning_rate": 1.9315451480768252e-07, + "loss": 0.1378, + "step": 10069 + }, + { + "epoch": 0.53, + "learning_rate": 1.931207232476976e-07, + "loss": 0.1265, + "step": 10070 + }, + { + "epoch": 0.53, + "learning_rate": 1.9308693188432514e-07, + "loss": 0.1071, + "step": 10071 + }, + { + "epoch": 0.53, + "learning_rate": 1.9305314071853078e-07, + "loss": 0.1319, + "step": 10072 + }, + { + "epoch": 0.53, + "learning_rate": 1.930193497512804e-07, + "loss": 0.0918, + "step": 10073 + }, + { + "epoch": 0.53, + "learning_rate": 1.9298555898353968e-07, + "loss": 0.1553, + "step": 10074 + }, + { + "epoch": 0.53, + "learning_rate": 1.9295176841627434e-07, + "loss": 0.1474, + "step": 10075 + }, + { + "epoch": 0.53, + "learning_rate": 1.9291797805045022e-07, + "loss": 0.1314, + "step": 10076 + }, + { + "epoch": 0.53, + "learning_rate": 1.92884187887033e-07, + "loss": 0.1322, + "step": 10077 + }, + { + "epoch": 0.53, + "learning_rate": 1.9285039792698848e-07, + "loss": 0.1255, + "step": 10078 + }, + { + "epoch": 0.53, + "learning_rate": 1.9281660817128227e-07, + "loss": 0.1313, + "step": 10079 + }, + { + "epoch": 0.53, + "learning_rate": 1.9278281862088025e-07, + "loss": 0.1603, + "step": 10080 + }, + { + "epoch": 0.53, + "learning_rate": 1.9274902927674808e-07, + "loss": 0.1112, + "step": 10081 + }, + { + "epoch": 0.53, + "learning_rate": 1.9271524013985137e-07, + "loss": 0.1512, + "step": 10082 + }, + { + "epoch": 0.53, + "learning_rate": 1.9268145121115594e-07, + "loss": 0.143, + "step": 10083 + }, + { + "epoch": 0.53, + "learning_rate": 1.9264766249162746e-07, + "loss": 0.1789, + "step": 10084 + }, + { + "epoch": 0.53, + "learning_rate": 1.926138739822316e-07, + "loss": 0.0932, + "step": 10085 + }, + { + "epoch": 0.53, + "learning_rate": 1.925800856839341e-07, + "loss": 0.1598, + "step": 10086 + }, + { + "epoch": 0.53, + "learning_rate": 1.9254629759770062e-07, + "loss": 0.1499, + "step": 10087 + }, + { + "epoch": 0.53, + "learning_rate": 1.9251250972449678e-07, + "loss": 0.1581, + "step": 10088 + }, + { + "epoch": 0.53, + "learning_rate": 1.9247872206528836e-07, + "loss": 0.1606, + "step": 10089 + }, + { + "epoch": 0.53, + "learning_rate": 1.9244493462104104e-07, + "loss": 0.1099, + "step": 10090 + }, + { + "epoch": 0.53, + "learning_rate": 1.9241114739272033e-07, + "loss": 0.1673, + "step": 10091 + }, + { + "epoch": 0.53, + "learning_rate": 1.9237736038129192e-07, + "loss": 0.1781, + "step": 10092 + }, + { + "epoch": 0.53, + "learning_rate": 1.9234357358772152e-07, + "loss": 0.1363, + "step": 10093 + }, + { + "epoch": 0.53, + "learning_rate": 1.9230978701297472e-07, + "loss": 0.1534, + "step": 10094 + }, + { + "epoch": 0.53, + "learning_rate": 1.9227600065801722e-07, + "loss": 0.1405, + "step": 10095 + }, + { + "epoch": 0.53, + "learning_rate": 1.9224221452381456e-07, + "loss": 0.1205, + "step": 10096 + }, + { + "epoch": 0.53, + "learning_rate": 1.922084286113324e-07, + "loss": 0.1335, + "step": 10097 + }, + { + "epoch": 0.53, + "learning_rate": 1.9217464292153647e-07, + "loss": 0.1406, + "step": 10098 + }, + { + "epoch": 0.53, + "learning_rate": 1.9214085745539213e-07, + "loss": 0.1781, + "step": 10099 + }, + { + "epoch": 0.53, + "learning_rate": 1.921070722138652e-07, + "loss": 0.1729, + "step": 10100 + }, + { + "epoch": 0.53, + "learning_rate": 1.9207328719792112e-07, + "loss": 0.1283, + "step": 10101 + }, + { + "epoch": 0.53, + "learning_rate": 1.920395024085256e-07, + "loss": 0.1579, + "step": 10102 + }, + { + "epoch": 0.53, + "learning_rate": 1.9200571784664414e-07, + "loss": 0.1545, + "step": 10103 + }, + { + "epoch": 0.53, + "learning_rate": 1.919719335132424e-07, + "loss": 0.1298, + "step": 10104 + }, + { + "epoch": 0.53, + "learning_rate": 1.9193814940928586e-07, + "loss": 0.1474, + "step": 10105 + }, + { + "epoch": 0.53, + "learning_rate": 1.9190436553574013e-07, + "loss": 0.1326, + "step": 10106 + }, + { + "epoch": 0.53, + "learning_rate": 1.9187058189357084e-07, + "loss": 0.1419, + "step": 10107 + }, + { + "epoch": 0.53, + "learning_rate": 1.9183679848374336e-07, + "loss": 0.1767, + "step": 10108 + }, + { + "epoch": 0.53, + "learning_rate": 1.9180301530722336e-07, + "loss": 0.1183, + "step": 10109 + }, + { + "epoch": 0.53, + "learning_rate": 1.917692323649763e-07, + "loss": 0.1235, + "step": 10110 + }, + { + "epoch": 0.53, + "learning_rate": 1.917354496579678e-07, + "loss": 0.1221, + "step": 10111 + }, + { + "epoch": 0.53, + "learning_rate": 1.9170166718716333e-07, + "loss": 0.1279, + "step": 10112 + }, + { + "epoch": 0.53, + "learning_rate": 1.9166788495352844e-07, + "loss": 0.1343, + "step": 10113 + }, + { + "epoch": 0.53, + "learning_rate": 1.9163410295802855e-07, + "loss": 0.1562, + "step": 10114 + }, + { + "epoch": 0.53, + "learning_rate": 1.9160032120162933e-07, + "loss": 0.147, + "step": 10115 + }, + { + "epoch": 0.53, + "learning_rate": 1.9156653968529614e-07, + "loss": 0.147, + "step": 10116 + }, + { + "epoch": 0.53, + "learning_rate": 1.9153275840999446e-07, + "loss": 0.162, + "step": 10117 + }, + { + "epoch": 0.53, + "learning_rate": 1.9149897737668983e-07, + "loss": 0.1211, + "step": 10118 + }, + { + "epoch": 0.53, + "learning_rate": 1.9146519658634768e-07, + "loss": 0.1364, + "step": 10119 + }, + { + "epoch": 0.53, + "learning_rate": 1.9143141603993357e-07, + "loss": 0.163, + "step": 10120 + }, + { + "epoch": 0.53, + "learning_rate": 1.913976357384128e-07, + "loss": 0.1382, + "step": 10121 + }, + { + "epoch": 0.53, + "learning_rate": 1.91363855682751e-07, + "loss": 0.1244, + "step": 10122 + }, + { + "epoch": 0.53, + "learning_rate": 1.913300758739135e-07, + "loss": 0.1313, + "step": 10123 + }, + { + "epoch": 0.53, + "learning_rate": 1.9129629631286587e-07, + "loss": 0.1782, + "step": 10124 + }, + { + "epoch": 0.53, + "learning_rate": 1.912625170005734e-07, + "loss": 0.1293, + "step": 10125 + }, + { + "epoch": 0.53, + "learning_rate": 1.912287379380015e-07, + "loss": 0.1346, + "step": 10126 + }, + { + "epoch": 0.53, + "learning_rate": 1.9119495912611574e-07, + "loss": 0.1515, + "step": 10127 + }, + { + "epoch": 0.53, + "learning_rate": 1.9116118056588137e-07, + "loss": 0.1467, + "step": 10128 + }, + { + "epoch": 0.53, + "learning_rate": 1.9112740225826394e-07, + "loss": 0.1436, + "step": 10129 + }, + { + "epoch": 0.53, + "learning_rate": 1.910936242042287e-07, + "loss": 0.121, + "step": 10130 + }, + { + "epoch": 0.53, + "learning_rate": 1.910598464047412e-07, + "loss": 0.1156, + "step": 10131 + }, + { + "epoch": 0.53, + "learning_rate": 1.9102606886076667e-07, + "loss": 0.1186, + "step": 10132 + }, + { + "epoch": 0.53, + "learning_rate": 1.9099229157327067e-07, + "loss": 0.1224, + "step": 10133 + }, + { + "epoch": 0.53, + "learning_rate": 1.909585145432184e-07, + "loss": 0.103, + "step": 10134 + }, + { + "epoch": 0.53, + "learning_rate": 1.9092473777157523e-07, + "loss": 0.1509, + "step": 10135 + }, + { + "epoch": 0.53, + "learning_rate": 1.908909612593066e-07, + "loss": 0.1491, + "step": 10136 + }, + { + "epoch": 0.53, + "learning_rate": 1.9085718500737777e-07, + "loss": 0.1505, + "step": 10137 + }, + { + "epoch": 0.53, + "learning_rate": 1.9082340901675417e-07, + "loss": 0.1552, + "step": 10138 + }, + { + "epoch": 0.53, + "learning_rate": 1.9078963328840103e-07, + "loss": 0.1463, + "step": 10139 + }, + { + "epoch": 0.53, + "learning_rate": 1.9075585782328379e-07, + "loss": 0.1761, + "step": 10140 + }, + { + "epoch": 0.53, + "learning_rate": 1.9072208262236771e-07, + "loss": 0.1676, + "step": 10141 + }, + { + "epoch": 0.53, + "learning_rate": 1.9068830768661803e-07, + "loss": 0.1294, + "step": 10142 + }, + { + "epoch": 0.53, + "learning_rate": 1.9065453301700015e-07, + "loss": 0.1789, + "step": 10143 + }, + { + "epoch": 0.53, + "learning_rate": 1.9062075861447928e-07, + "loss": 0.1557, + "step": 10144 + }, + { + "epoch": 0.53, + "learning_rate": 1.905869844800208e-07, + "loss": 0.0975, + "step": 10145 + }, + { + "epoch": 0.53, + "learning_rate": 1.9055321061458987e-07, + "loss": 0.1179, + "step": 10146 + }, + { + "epoch": 0.53, + "learning_rate": 1.905194370191519e-07, + "loss": 0.1301, + "step": 10147 + }, + { + "epoch": 0.53, + "learning_rate": 1.90485663694672e-07, + "loss": 0.1198, + "step": 10148 + }, + { + "epoch": 0.53, + "learning_rate": 1.9045189064211558e-07, + "loss": 0.119, + "step": 10149 + }, + { + "epoch": 0.53, + "learning_rate": 1.9041811786244787e-07, + "loss": 0.1413, + "step": 10150 + }, + { + "epoch": 0.53, + "learning_rate": 1.9038434535663393e-07, + "loss": 0.1215, + "step": 10151 + }, + { + "epoch": 0.53, + "learning_rate": 1.9035057312563915e-07, + "loss": 0.1323, + "step": 10152 + }, + { + "epoch": 0.53, + "learning_rate": 1.903168011704287e-07, + "loss": 0.1655, + "step": 10153 + }, + { + "epoch": 0.53, + "learning_rate": 1.9028302949196784e-07, + "loss": 0.1415, + "step": 10154 + }, + { + "epoch": 0.53, + "learning_rate": 1.902492580912217e-07, + "loss": 0.1323, + "step": 10155 + }, + { + "epoch": 0.53, + "learning_rate": 1.9021548696915557e-07, + "loss": 0.1269, + "step": 10156 + }, + { + "epoch": 0.53, + "learning_rate": 1.9018171612673458e-07, + "loss": 0.1081, + "step": 10157 + }, + { + "epoch": 0.53, + "learning_rate": 1.90147945564924e-07, + "loss": 0.1368, + "step": 10158 + }, + { + "epoch": 0.53, + "learning_rate": 1.901141752846889e-07, + "loss": 0.1664, + "step": 10159 + }, + { + "epoch": 0.53, + "learning_rate": 1.9008040528699443e-07, + "loss": 0.118, + "step": 10160 + }, + { + "epoch": 0.53, + "learning_rate": 1.9004663557280585e-07, + "loss": 0.1469, + "step": 10161 + }, + { + "epoch": 0.53, + "learning_rate": 1.9001286614308826e-07, + "loss": 0.1804, + "step": 10162 + }, + { + "epoch": 0.53, + "learning_rate": 1.899790969988068e-07, + "loss": 0.1286, + "step": 10163 + }, + { + "epoch": 0.53, + "learning_rate": 1.899453281409266e-07, + "loss": 0.1384, + "step": 10164 + }, + { + "epoch": 0.53, + "learning_rate": 1.8991155957041287e-07, + "loss": 0.1213, + "step": 10165 + }, + { + "epoch": 0.53, + "learning_rate": 1.8987779128823058e-07, + "loss": 0.1468, + "step": 10166 + }, + { + "epoch": 0.53, + "learning_rate": 1.8984402329534503e-07, + "loss": 0.1738, + "step": 10167 + }, + { + "epoch": 0.53, + "learning_rate": 1.8981025559272118e-07, + "loss": 0.1622, + "step": 10168 + }, + { + "epoch": 0.53, + "learning_rate": 1.897764881813241e-07, + "loss": 0.1237, + "step": 10169 + }, + { + "epoch": 0.53, + "learning_rate": 1.8974272106211896e-07, + "loss": 0.1606, + "step": 10170 + }, + { + "epoch": 0.53, + "learning_rate": 1.897089542360708e-07, + "loss": 0.1111, + "step": 10171 + }, + { + "epoch": 0.53, + "learning_rate": 1.8967518770414474e-07, + "loss": 0.1387, + "step": 10172 + }, + { + "epoch": 0.53, + "learning_rate": 1.8964142146730576e-07, + "loss": 0.1344, + "step": 10173 + }, + { + "epoch": 0.53, + "learning_rate": 1.89607655526519e-07, + "loss": 0.1269, + "step": 10174 + }, + { + "epoch": 0.53, + "learning_rate": 1.8957388988274946e-07, + "loss": 0.1145, + "step": 10175 + }, + { + "epoch": 0.53, + "learning_rate": 1.8954012453696213e-07, + "loss": 0.128, + "step": 10176 + }, + { + "epoch": 0.53, + "learning_rate": 1.895063594901221e-07, + "loss": 0.1311, + "step": 10177 + }, + { + "epoch": 0.53, + "learning_rate": 1.8947259474319433e-07, + "loss": 0.117, + "step": 10178 + }, + { + "epoch": 0.53, + "learning_rate": 1.8943883029714392e-07, + "loss": 0.1304, + "step": 10179 + }, + { + "epoch": 0.53, + "learning_rate": 1.8940506615293576e-07, + "loss": 0.1475, + "step": 10180 + }, + { + "epoch": 0.53, + "learning_rate": 1.8937130231153495e-07, + "loss": 0.1416, + "step": 10181 + }, + { + "epoch": 0.53, + "learning_rate": 1.8933753877390636e-07, + "loss": 0.1388, + "step": 10182 + }, + { + "epoch": 0.53, + "learning_rate": 1.8930377554101507e-07, + "loss": 0.1373, + "step": 10183 + }, + { + "epoch": 0.53, + "learning_rate": 1.8927001261382608e-07, + "loss": 0.1799, + "step": 10184 + }, + { + "epoch": 0.53, + "learning_rate": 1.892362499933042e-07, + "loss": 0.1224, + "step": 10185 + }, + { + "epoch": 0.53, + "learning_rate": 1.8920248768041444e-07, + "loss": 0.1254, + "step": 10186 + }, + { + "epoch": 0.53, + "learning_rate": 1.891687256761217e-07, + "loss": 0.1263, + "step": 10187 + }, + { + "epoch": 0.53, + "learning_rate": 1.8913496398139106e-07, + "loss": 0.1519, + "step": 10188 + }, + { + "epoch": 0.53, + "learning_rate": 1.891012025971873e-07, + "loss": 0.1208, + "step": 10189 + }, + { + "epoch": 0.53, + "learning_rate": 1.8906744152447538e-07, + "loss": 0.1437, + "step": 10190 + }, + { + "epoch": 0.53, + "learning_rate": 1.8903368076422018e-07, + "loss": 0.1352, + "step": 10191 + }, + { + "epoch": 0.53, + "learning_rate": 1.8899992031738665e-07, + "loss": 0.1238, + "step": 10192 + }, + { + "epoch": 0.53, + "learning_rate": 1.889661601849397e-07, + "loss": 0.1562, + "step": 10193 + }, + { + "epoch": 0.53, + "learning_rate": 1.8893240036784402e-07, + "loss": 0.1658, + "step": 10194 + }, + { + "epoch": 0.53, + "learning_rate": 1.8889864086706471e-07, + "loss": 0.1625, + "step": 10195 + }, + { + "epoch": 0.53, + "learning_rate": 1.8886488168356648e-07, + "loss": 0.1371, + "step": 10196 + }, + { + "epoch": 0.53, + "learning_rate": 1.8883112281831427e-07, + "loss": 0.175, + "step": 10197 + }, + { + "epoch": 0.53, + "learning_rate": 1.887973642722728e-07, + "loss": 0.1335, + "step": 10198 + }, + { + "epoch": 0.53, + "learning_rate": 1.8876360604640706e-07, + "loss": 0.1506, + "step": 10199 + }, + { + "epoch": 0.53, + "learning_rate": 1.8872984814168175e-07, + "loss": 0.1025, + "step": 10200 + }, + { + "epoch": 0.53, + "learning_rate": 1.8869609055906186e-07, + "loss": 0.1389, + "step": 10201 + }, + { + "epoch": 0.53, + "learning_rate": 1.8866233329951192e-07, + "loss": 0.1448, + "step": 10202 + }, + { + "epoch": 0.53, + "learning_rate": 1.8862857636399694e-07, + "loss": 0.1433, + "step": 10203 + }, + { + "epoch": 0.53, + "learning_rate": 1.8859481975348158e-07, + "loss": 0.1085, + "step": 10204 + }, + { + "epoch": 0.53, + "learning_rate": 1.8856106346893073e-07, + "loss": 0.1246, + "step": 10205 + }, + { + "epoch": 0.53, + "learning_rate": 1.8852730751130906e-07, + "loss": 0.114, + "step": 10206 + }, + { + "epoch": 0.53, + "learning_rate": 1.884935518815814e-07, + "loss": 0.1144, + "step": 10207 + }, + { + "epoch": 0.53, + "learning_rate": 1.8845979658071243e-07, + "loss": 0.1341, + "step": 10208 + }, + { + "epoch": 0.53, + "learning_rate": 1.8842604160966696e-07, + "loss": 0.1101, + "step": 10209 + }, + { + "epoch": 0.53, + "learning_rate": 1.8839228696940974e-07, + "loss": 0.1474, + "step": 10210 + }, + { + "epoch": 0.53, + "learning_rate": 1.8835853266090537e-07, + "loss": 0.1149, + "step": 10211 + }, + { + "epoch": 0.53, + "learning_rate": 1.8832477868511866e-07, + "loss": 0.1138, + "step": 10212 + }, + { + "epoch": 0.53, + "learning_rate": 1.8829102504301423e-07, + "loss": 0.107, + "step": 10213 + }, + { + "epoch": 0.53, + "learning_rate": 1.8825727173555687e-07, + "loss": 0.1081, + "step": 10214 + }, + { + "epoch": 0.53, + "learning_rate": 1.882235187637112e-07, + "loss": 0.1106, + "step": 10215 + }, + { + "epoch": 0.53, + "learning_rate": 1.881897661284419e-07, + "loss": 0.1348, + "step": 10216 + }, + { + "epoch": 0.53, + "learning_rate": 1.8815601383071364e-07, + "loss": 0.1466, + "step": 10217 + }, + { + "epoch": 0.53, + "learning_rate": 1.8812226187149117e-07, + "loss": 0.1191, + "step": 10218 + }, + { + "epoch": 0.53, + "learning_rate": 1.8808851025173897e-07, + "loss": 0.1206, + "step": 10219 + }, + { + "epoch": 0.53, + "learning_rate": 1.8805475897242174e-07, + "loss": 0.1626, + "step": 10220 + }, + { + "epoch": 0.53, + "learning_rate": 1.8802100803450411e-07, + "loss": 0.1402, + "step": 10221 + }, + { + "epoch": 0.53, + "learning_rate": 1.8798725743895066e-07, + "loss": 0.099, + "step": 10222 + }, + { + "epoch": 0.53, + "learning_rate": 1.8795350718672609e-07, + "loss": 0.1507, + "step": 10223 + }, + { + "epoch": 0.53, + "learning_rate": 1.879197572787949e-07, + "loss": 0.1333, + "step": 10224 + }, + { + "epoch": 0.53, + "learning_rate": 1.8788600771612172e-07, + "loss": 0.1164, + "step": 10225 + }, + { + "epoch": 0.53, + "learning_rate": 1.878522584996711e-07, + "loss": 0.143, + "step": 10226 + }, + { + "epoch": 0.53, + "learning_rate": 1.878185096304077e-07, + "loss": 0.1559, + "step": 10227 + }, + { + "epoch": 0.53, + "learning_rate": 1.8778476110929596e-07, + "loss": 0.1214, + "step": 10228 + }, + { + "epoch": 0.53, + "learning_rate": 1.8775101293730042e-07, + "loss": 0.1518, + "step": 10229 + }, + { + "epoch": 0.53, + "learning_rate": 1.8771726511538568e-07, + "loss": 0.1371, + "step": 10230 + }, + { + "epoch": 0.53, + "learning_rate": 1.8768351764451623e-07, + "loss": 0.1195, + "step": 10231 + }, + { + "epoch": 0.53, + "learning_rate": 1.8764977052565663e-07, + "loss": 0.1019, + "step": 10232 + }, + { + "epoch": 0.53, + "learning_rate": 1.876160237597713e-07, + "loss": 0.1651, + "step": 10233 + }, + { + "epoch": 0.53, + "learning_rate": 1.8758227734782486e-07, + "loss": 0.1236, + "step": 10234 + }, + { + "epoch": 0.53, + "learning_rate": 1.8754853129078175e-07, + "loss": 0.1194, + "step": 10235 + }, + { + "epoch": 0.53, + "learning_rate": 1.8751478558960636e-07, + "loss": 0.1296, + "step": 10236 + }, + { + "epoch": 0.53, + "learning_rate": 1.8748104024526323e-07, + "loss": 0.1439, + "step": 10237 + }, + { + "epoch": 0.53, + "learning_rate": 1.874472952587168e-07, + "loss": 0.1246, + "step": 10238 + }, + { + "epoch": 0.53, + "learning_rate": 1.8741355063093152e-07, + "loss": 0.1259, + "step": 10239 + }, + { + "epoch": 0.53, + "learning_rate": 1.8737980636287179e-07, + "loss": 0.1388, + "step": 10240 + }, + { + "epoch": 0.53, + "learning_rate": 1.873460624555021e-07, + "loss": 0.1335, + "step": 10241 + }, + { + "epoch": 0.53, + "learning_rate": 1.8731231890978683e-07, + "loss": 0.1624, + "step": 10242 + }, + { + "epoch": 0.53, + "learning_rate": 1.8727857572669038e-07, + "loss": 0.1597, + "step": 10243 + }, + { + "epoch": 0.53, + "learning_rate": 1.8724483290717722e-07, + "loss": 0.1633, + "step": 10244 + }, + { + "epoch": 0.53, + "learning_rate": 1.8721109045221154e-07, + "loss": 0.1313, + "step": 10245 + }, + { + "epoch": 0.53, + "learning_rate": 1.871773483627579e-07, + "loss": 0.1227, + "step": 10246 + }, + { + "epoch": 0.53, + "learning_rate": 1.8714360663978055e-07, + "loss": 0.1541, + "step": 10247 + }, + { + "epoch": 0.53, + "learning_rate": 1.8710986528424393e-07, + "loss": 0.1301, + "step": 10248 + }, + { + "epoch": 0.53, + "learning_rate": 1.870761242971123e-07, + "loss": 0.1439, + "step": 10249 + }, + { + "epoch": 0.54, + "learning_rate": 1.8704238367935004e-07, + "loss": 0.1692, + "step": 10250 + }, + { + "epoch": 0.54, + "learning_rate": 1.8700864343192146e-07, + "loss": 0.1213, + "step": 10251 + }, + { + "epoch": 0.54, + "learning_rate": 1.8697490355579087e-07, + "loss": 0.1414, + "step": 10252 + }, + { + "epoch": 0.54, + "learning_rate": 1.8694116405192258e-07, + "loss": 0.126, + "step": 10253 + }, + { + "epoch": 0.54, + "learning_rate": 1.8690742492128083e-07, + "loss": 0.125, + "step": 10254 + }, + { + "epoch": 0.54, + "learning_rate": 1.8687368616482997e-07, + "loss": 0.0995, + "step": 10255 + }, + { + "epoch": 0.54, + "learning_rate": 1.8683994778353416e-07, + "loss": 0.1481, + "step": 10256 + }, + { + "epoch": 0.54, + "learning_rate": 1.8680620977835777e-07, + "loss": 0.2071, + "step": 10257 + }, + { + "epoch": 0.54, + "learning_rate": 1.8677247215026496e-07, + "loss": 0.2165, + "step": 10258 + }, + { + "epoch": 0.54, + "learning_rate": 1.8673873490022006e-07, + "loss": 0.1244, + "step": 10259 + }, + { + "epoch": 0.54, + "learning_rate": 1.867049980291872e-07, + "loss": 0.1145, + "step": 10260 + }, + { + "epoch": 0.54, + "learning_rate": 1.8667126153813066e-07, + "loss": 0.1266, + "step": 10261 + }, + { + "epoch": 0.54, + "learning_rate": 1.8663752542801462e-07, + "loss": 0.1353, + "step": 10262 + }, + { + "epoch": 0.54, + "learning_rate": 1.8660378969980319e-07, + "loss": 0.141, + "step": 10263 + }, + { + "epoch": 0.54, + "learning_rate": 1.8657005435446068e-07, + "loss": 0.1105, + "step": 10264 + }, + { + "epoch": 0.54, + "learning_rate": 1.8653631939295115e-07, + "loss": 0.1368, + "step": 10265 + }, + { + "epoch": 0.54, + "learning_rate": 1.8650258481623886e-07, + "loss": 0.1747, + "step": 10266 + }, + { + "epoch": 0.54, + "learning_rate": 1.8646885062528786e-07, + "loss": 0.1269, + "step": 10267 + }, + { + "epoch": 0.54, + "learning_rate": 1.8643511682106237e-07, + "loss": 0.1486, + "step": 10268 + }, + { + "epoch": 0.54, + "learning_rate": 1.8640138340452643e-07, + "loss": 0.1497, + "step": 10269 + }, + { + "epoch": 0.54, + "learning_rate": 1.863676503766443e-07, + "loss": 0.1366, + "step": 10270 + }, + { + "epoch": 0.54, + "learning_rate": 1.8633391773837995e-07, + "loss": 0.1297, + "step": 10271 + }, + { + "epoch": 0.54, + "learning_rate": 1.8630018549069744e-07, + "loss": 0.1544, + "step": 10272 + }, + { + "epoch": 0.54, + "learning_rate": 1.8626645363456096e-07, + "loss": 0.1416, + "step": 10273 + }, + { + "epoch": 0.54, + "learning_rate": 1.8623272217093453e-07, + "loss": 0.1188, + "step": 10274 + }, + { + "epoch": 0.54, + "learning_rate": 1.8619899110078224e-07, + "loss": 0.1333, + "step": 10275 + }, + { + "epoch": 0.54, + "learning_rate": 1.8616526042506807e-07, + "loss": 0.1605, + "step": 10276 + }, + { + "epoch": 0.54, + "learning_rate": 1.861315301447561e-07, + "loss": 0.1433, + "step": 10277 + }, + { + "epoch": 0.54, + "learning_rate": 1.8609780026081046e-07, + "loss": 0.1404, + "step": 10278 + }, + { + "epoch": 0.54, + "learning_rate": 1.8606407077419492e-07, + "loss": 0.1361, + "step": 10279 + }, + { + "epoch": 0.54, + "learning_rate": 1.8603034168587368e-07, + "loss": 0.1431, + "step": 10280 + }, + { + "epoch": 0.54, + "learning_rate": 1.8599661299681065e-07, + "loss": 0.1472, + "step": 10281 + }, + { + "epoch": 0.54, + "learning_rate": 1.8596288470796982e-07, + "loss": 0.1346, + "step": 10282 + }, + { + "epoch": 0.54, + "learning_rate": 1.8592915682031513e-07, + "loss": 0.1446, + "step": 10283 + }, + { + "epoch": 0.54, + "learning_rate": 1.8589542933481064e-07, + "loss": 0.1489, + "step": 10284 + }, + { + "epoch": 0.54, + "learning_rate": 1.8586170225242016e-07, + "loss": 0.1404, + "step": 10285 + }, + { + "epoch": 0.54, + "learning_rate": 1.8582797557410776e-07, + "loss": 0.125, + "step": 10286 + }, + { + "epoch": 0.54, + "learning_rate": 1.857942493008373e-07, + "loss": 0.1534, + "step": 10287 + }, + { + "epoch": 0.54, + "learning_rate": 1.857605234335726e-07, + "loss": 0.1508, + "step": 10288 + }, + { + "epoch": 0.54, + "learning_rate": 1.8572679797327769e-07, + "loss": 0.1296, + "step": 10289 + }, + { + "epoch": 0.54, + "learning_rate": 1.8569307292091637e-07, + "loss": 0.1691, + "step": 10290 + }, + { + "epoch": 0.54, + "learning_rate": 1.8565934827745258e-07, + "loss": 0.1351, + "step": 10291 + }, + { + "epoch": 0.54, + "learning_rate": 1.8562562404385008e-07, + "loss": 0.1339, + "step": 10292 + }, + { + "epoch": 0.54, + "learning_rate": 1.8559190022107287e-07, + "loss": 0.1154, + "step": 10293 + }, + { + "epoch": 0.54, + "learning_rate": 1.8555817681008468e-07, + "loss": 0.1178, + "step": 10294 + }, + { + "epoch": 0.54, + "learning_rate": 1.8552445381184944e-07, + "loss": 0.1359, + "step": 10295 + }, + { + "epoch": 0.54, + "learning_rate": 1.8549073122733086e-07, + "loss": 0.1015, + "step": 10296 + }, + { + "epoch": 0.54, + "learning_rate": 1.8545700905749273e-07, + "loss": 0.1218, + "step": 10297 + }, + { + "epoch": 0.54, + "learning_rate": 1.8542328730329895e-07, + "loss": 0.1422, + "step": 10298 + }, + { + "epoch": 0.54, + "learning_rate": 1.853895659657132e-07, + "loss": 0.1484, + "step": 10299 + }, + { + "epoch": 0.54, + "learning_rate": 1.853558450456993e-07, + "loss": 0.1297, + "step": 10300 + }, + { + "epoch": 0.54, + "learning_rate": 1.8532212454422102e-07, + "loss": 0.1095, + "step": 10301 + }, + { + "epoch": 0.54, + "learning_rate": 1.8528840446224213e-07, + "loss": 0.1538, + "step": 10302 + }, + { + "epoch": 0.54, + "learning_rate": 1.8525468480072623e-07, + "loss": 0.137, + "step": 10303 + }, + { + "epoch": 0.54, + "learning_rate": 1.8522096556063728e-07, + "loss": 0.1389, + "step": 10304 + }, + { + "epoch": 0.54, + "learning_rate": 1.851872467429387e-07, + "loss": 0.1262, + "step": 10305 + }, + { + "epoch": 0.54, + "learning_rate": 1.851535283485944e-07, + "loss": 0.1352, + "step": 10306 + }, + { + "epoch": 0.54, + "learning_rate": 1.851198103785679e-07, + "loss": 0.1347, + "step": 10307 + }, + { + "epoch": 0.54, + "learning_rate": 1.8508609283382303e-07, + "loss": 0.1254, + "step": 10308 + }, + { + "epoch": 0.54, + "learning_rate": 1.8505237571532337e-07, + "loss": 0.1147, + "step": 10309 + }, + { + "epoch": 0.54, + "learning_rate": 1.8501865902403258e-07, + "loss": 0.1068, + "step": 10310 + }, + { + "epoch": 0.54, + "learning_rate": 1.849849427609143e-07, + "loss": 0.099, + "step": 10311 + }, + { + "epoch": 0.54, + "learning_rate": 1.8495122692693213e-07, + "loss": 0.1615, + "step": 10312 + }, + { + "epoch": 0.54, + "learning_rate": 1.8491751152304985e-07, + "loss": 0.1309, + "step": 10313 + }, + { + "epoch": 0.54, + "learning_rate": 1.8488379655023075e-07, + "loss": 0.1406, + "step": 10314 + }, + { + "epoch": 0.54, + "learning_rate": 1.8485008200943864e-07, + "loss": 0.1202, + "step": 10315 + }, + { + "epoch": 0.54, + "learning_rate": 1.84816367901637e-07, + "loss": 0.1381, + "step": 10316 + }, + { + "epoch": 0.54, + "learning_rate": 1.8478265422778946e-07, + "loss": 0.1423, + "step": 10317 + }, + { + "epoch": 0.54, + "learning_rate": 1.8474894098885949e-07, + "loss": 0.1277, + "step": 10318 + }, + { + "epoch": 0.54, + "learning_rate": 1.8471522818581074e-07, + "loss": 0.1198, + "step": 10319 + }, + { + "epoch": 0.54, + "learning_rate": 1.846815158196066e-07, + "loss": 0.1328, + "step": 10320 + }, + { + "epoch": 0.54, + "learning_rate": 1.8464780389121074e-07, + "loss": 0.0952, + "step": 10321 + }, + { + "epoch": 0.54, + "learning_rate": 1.8461409240158653e-07, + "loss": 0.1758, + "step": 10322 + }, + { + "epoch": 0.54, + "learning_rate": 1.8458038135169746e-07, + "loss": 0.123, + "step": 10323 + }, + { + "epoch": 0.54, + "learning_rate": 1.8454667074250708e-07, + "loss": 0.1393, + "step": 10324 + }, + { + "epoch": 0.54, + "learning_rate": 1.845129605749788e-07, + "loss": 0.1388, + "step": 10325 + }, + { + "epoch": 0.54, + "learning_rate": 1.844792508500761e-07, + "loss": 0.1069, + "step": 10326 + }, + { + "epoch": 0.54, + "learning_rate": 1.8444554156876239e-07, + "loss": 0.1112, + "step": 10327 + }, + { + "epoch": 0.54, + "learning_rate": 1.844118327320011e-07, + "loss": 0.1362, + "step": 10328 + }, + { + "epoch": 0.54, + "learning_rate": 1.8437812434075565e-07, + "loss": 0.1475, + "step": 10329 + }, + { + "epoch": 0.54, + "learning_rate": 1.843444163959895e-07, + "loss": 0.1326, + "step": 10330 + }, + { + "epoch": 0.54, + "learning_rate": 1.8431070889866595e-07, + "loss": 0.1257, + "step": 10331 + }, + { + "epoch": 0.54, + "learning_rate": 1.8427700184974833e-07, + "loss": 0.142, + "step": 10332 + }, + { + "epoch": 0.54, + "learning_rate": 1.8424329525020011e-07, + "loss": 0.1822, + "step": 10333 + }, + { + "epoch": 0.54, + "learning_rate": 1.8420958910098454e-07, + "loss": 0.1459, + "step": 10334 + }, + { + "epoch": 0.54, + "learning_rate": 1.841758834030651e-07, + "loss": 0.1389, + "step": 10335 + }, + { + "epoch": 0.54, + "learning_rate": 1.8414217815740495e-07, + "loss": 0.1399, + "step": 10336 + }, + { + "epoch": 0.54, + "learning_rate": 1.8410847336496748e-07, + "loss": 0.1558, + "step": 10337 + }, + { + "epoch": 0.54, + "learning_rate": 1.8407476902671605e-07, + "loss": 0.1725, + "step": 10338 + }, + { + "epoch": 0.54, + "learning_rate": 1.840410651436138e-07, + "loss": 0.1447, + "step": 10339 + }, + { + "epoch": 0.54, + "learning_rate": 1.840073617166241e-07, + "loss": 0.1153, + "step": 10340 + }, + { + "epoch": 0.54, + "learning_rate": 1.8397365874671013e-07, + "loss": 0.1408, + "step": 10341 + }, + { + "epoch": 0.54, + "learning_rate": 1.8393995623483522e-07, + "loss": 0.1382, + "step": 10342 + }, + { + "epoch": 0.54, + "learning_rate": 1.8390625418196253e-07, + "loss": 0.1312, + "step": 10343 + }, + { + "epoch": 0.54, + "learning_rate": 1.8387255258905534e-07, + "loss": 0.1246, + "step": 10344 + }, + { + "epoch": 0.54, + "learning_rate": 1.838388514570768e-07, + "loss": 0.1264, + "step": 10345 + }, + { + "epoch": 0.54, + "learning_rate": 1.8380515078699013e-07, + "loss": 0.1712, + "step": 10346 + }, + { + "epoch": 0.54, + "learning_rate": 1.8377145057975856e-07, + "loss": 0.1205, + "step": 10347 + }, + { + "epoch": 0.54, + "learning_rate": 1.837377508363451e-07, + "loss": 0.1589, + "step": 10348 + }, + { + "epoch": 0.54, + "learning_rate": 1.8370405155771306e-07, + "loss": 0.1705, + "step": 10349 + }, + { + "epoch": 0.54, + "learning_rate": 1.8367035274482546e-07, + "loss": 0.1324, + "step": 10350 + }, + { + "epoch": 0.54, + "learning_rate": 1.8363665439864556e-07, + "loss": 0.1155, + "step": 10351 + }, + { + "epoch": 0.54, + "learning_rate": 1.8360295652013634e-07, + "loss": 0.1568, + "step": 10352 + }, + { + "epoch": 0.54, + "learning_rate": 1.8356925911026096e-07, + "loss": 0.141, + "step": 10353 + }, + { + "epoch": 0.54, + "learning_rate": 1.8353556216998248e-07, + "loss": 0.138, + "step": 10354 + }, + { + "epoch": 0.54, + "learning_rate": 1.8350186570026407e-07, + "loss": 0.1549, + "step": 10355 + }, + { + "epoch": 0.54, + "learning_rate": 1.8346816970206867e-07, + "loss": 0.1535, + "step": 10356 + }, + { + "epoch": 0.54, + "learning_rate": 1.834344741763593e-07, + "loss": 0.1379, + "step": 10357 + }, + { + "epoch": 0.54, + "learning_rate": 1.8340077912409912e-07, + "loss": 0.1266, + "step": 10358 + }, + { + "epoch": 0.54, + "learning_rate": 1.8336708454625104e-07, + "loss": 0.1312, + "step": 10359 + }, + { + "epoch": 0.54, + "learning_rate": 1.8333339044377814e-07, + "loss": 0.1267, + "step": 10360 + }, + { + "epoch": 0.54, + "learning_rate": 1.8329969681764333e-07, + "loss": 0.1123, + "step": 10361 + }, + { + "epoch": 0.54, + "learning_rate": 1.8326600366880968e-07, + "loss": 0.1692, + "step": 10362 + }, + { + "epoch": 0.54, + "learning_rate": 1.8323231099824002e-07, + "loss": 0.1428, + "step": 10363 + }, + { + "epoch": 0.54, + "learning_rate": 1.8319861880689752e-07, + "loss": 0.1383, + "step": 10364 + }, + { + "epoch": 0.54, + "learning_rate": 1.8316492709574494e-07, + "loss": 0.134, + "step": 10365 + }, + { + "epoch": 0.54, + "learning_rate": 1.8313123586574518e-07, + "loss": 0.1532, + "step": 10366 + }, + { + "epoch": 0.54, + "learning_rate": 1.8309754511786125e-07, + "loss": 0.1461, + "step": 10367 + }, + { + "epoch": 0.54, + "learning_rate": 1.8306385485305596e-07, + "loss": 0.1336, + "step": 10368 + }, + { + "epoch": 0.54, + "learning_rate": 1.8303016507229228e-07, + "loss": 0.1778, + "step": 10369 + }, + { + "epoch": 0.54, + "learning_rate": 1.8299647577653303e-07, + "loss": 0.1399, + "step": 10370 + }, + { + "epoch": 0.54, + "learning_rate": 1.8296278696674108e-07, + "loss": 0.1379, + "step": 10371 + }, + { + "epoch": 0.54, + "learning_rate": 1.829290986438793e-07, + "loss": 0.1384, + "step": 10372 + }, + { + "epoch": 0.54, + "learning_rate": 1.8289541080891038e-07, + "loss": 0.1737, + "step": 10373 + }, + { + "epoch": 0.54, + "learning_rate": 1.8286172346279727e-07, + "loss": 0.1573, + "step": 10374 + }, + { + "epoch": 0.54, + "learning_rate": 1.8282803660650267e-07, + "loss": 0.1369, + "step": 10375 + }, + { + "epoch": 0.54, + "learning_rate": 1.8279435024098946e-07, + "loss": 0.1129, + "step": 10376 + }, + { + "epoch": 0.54, + "learning_rate": 1.8276066436722034e-07, + "loss": 0.1554, + "step": 10377 + }, + { + "epoch": 0.54, + "learning_rate": 1.827269789861581e-07, + "loss": 0.1234, + "step": 10378 + }, + { + "epoch": 0.54, + "learning_rate": 1.826932940987654e-07, + "loss": 0.1342, + "step": 10379 + }, + { + "epoch": 0.54, + "learning_rate": 1.8265960970600514e-07, + "loss": 0.1625, + "step": 10380 + }, + { + "epoch": 0.54, + "learning_rate": 1.8262592580883994e-07, + "loss": 0.1379, + "step": 10381 + }, + { + "epoch": 0.54, + "learning_rate": 1.825922424082324e-07, + "loss": 0.1516, + "step": 10382 + }, + { + "epoch": 0.54, + "learning_rate": 1.8255855950514536e-07, + "loss": 0.1256, + "step": 10383 + }, + { + "epoch": 0.54, + "learning_rate": 1.8252487710054136e-07, + "loss": 0.1236, + "step": 10384 + }, + { + "epoch": 0.54, + "learning_rate": 1.8249119519538317e-07, + "loss": 0.1677, + "step": 10385 + }, + { + "epoch": 0.54, + "learning_rate": 1.8245751379063332e-07, + "loss": 0.1475, + "step": 10386 + }, + { + "epoch": 0.54, + "learning_rate": 1.8242383288725454e-07, + "loss": 0.1875, + "step": 10387 + }, + { + "epoch": 0.54, + "learning_rate": 1.8239015248620938e-07, + "loss": 0.156, + "step": 10388 + }, + { + "epoch": 0.54, + "learning_rate": 1.823564725884605e-07, + "loss": 0.1371, + "step": 10389 + }, + { + "epoch": 0.54, + "learning_rate": 1.8232279319497047e-07, + "loss": 0.1586, + "step": 10390 + }, + { + "epoch": 0.54, + "learning_rate": 1.8228911430670175e-07, + "loss": 0.1423, + "step": 10391 + }, + { + "epoch": 0.54, + "learning_rate": 1.8225543592461704e-07, + "loss": 0.1415, + "step": 10392 + }, + { + "epoch": 0.54, + "learning_rate": 1.8222175804967877e-07, + "loss": 0.1293, + "step": 10393 + }, + { + "epoch": 0.54, + "learning_rate": 1.8218808068284958e-07, + "loss": 0.1335, + "step": 10394 + }, + { + "epoch": 0.54, + "learning_rate": 1.8215440382509184e-07, + "loss": 0.1297, + "step": 10395 + }, + { + "epoch": 0.54, + "learning_rate": 1.821207274773682e-07, + "loss": 0.147, + "step": 10396 + }, + { + "epoch": 0.54, + "learning_rate": 1.82087051640641e-07, + "loss": 0.1417, + "step": 10397 + }, + { + "epoch": 0.54, + "learning_rate": 1.820533763158729e-07, + "loss": 0.1127, + "step": 10398 + }, + { + "epoch": 0.54, + "learning_rate": 1.8201970150402617e-07, + "loss": 0.1495, + "step": 10399 + }, + { + "epoch": 0.54, + "learning_rate": 1.8198602720606325e-07, + "loss": 0.1469, + "step": 10400 + }, + { + "epoch": 0.54, + "learning_rate": 1.819523534229467e-07, + "loss": 0.1228, + "step": 10401 + }, + { + "epoch": 0.54, + "learning_rate": 1.8191868015563882e-07, + "loss": 0.1459, + "step": 10402 + }, + { + "epoch": 0.54, + "learning_rate": 1.8188500740510207e-07, + "loss": 0.1465, + "step": 10403 + }, + { + "epoch": 0.54, + "learning_rate": 1.8185133517229873e-07, + "loss": 0.1533, + "step": 10404 + }, + { + "epoch": 0.54, + "learning_rate": 1.8181766345819132e-07, + "loss": 0.1446, + "step": 10405 + }, + { + "epoch": 0.54, + "learning_rate": 1.8178399226374206e-07, + "loss": 0.1619, + "step": 10406 + }, + { + "epoch": 0.54, + "learning_rate": 1.817503215899134e-07, + "loss": 0.1671, + "step": 10407 + }, + { + "epoch": 0.54, + "learning_rate": 1.8171665143766758e-07, + "loss": 0.1135, + "step": 10408 + }, + { + "epoch": 0.54, + "learning_rate": 1.8168298180796685e-07, + "loss": 0.1306, + "step": 10409 + }, + { + "epoch": 0.54, + "learning_rate": 1.816493127017736e-07, + "loss": 0.1473, + "step": 10410 + }, + { + "epoch": 0.54, + "learning_rate": 1.816156441200501e-07, + "loss": 0.1384, + "step": 10411 + }, + { + "epoch": 0.54, + "learning_rate": 1.8158197606375856e-07, + "loss": 0.126, + "step": 10412 + }, + { + "epoch": 0.54, + "learning_rate": 1.8154830853386124e-07, + "loss": 0.1266, + "step": 10413 + }, + { + "epoch": 0.54, + "learning_rate": 1.815146415313204e-07, + "loss": 0.1554, + "step": 10414 + }, + { + "epoch": 0.54, + "learning_rate": 1.8148097505709833e-07, + "loss": 0.1507, + "step": 10415 + }, + { + "epoch": 0.54, + "learning_rate": 1.814473091121571e-07, + "loss": 0.1493, + "step": 10416 + }, + { + "epoch": 0.54, + "learning_rate": 1.8141364369745885e-07, + "loss": 0.1378, + "step": 10417 + }, + { + "epoch": 0.54, + "learning_rate": 1.8137997881396592e-07, + "loss": 0.102, + "step": 10418 + }, + { + "epoch": 0.54, + "learning_rate": 1.8134631446264033e-07, + "loss": 0.1223, + "step": 10419 + }, + { + "epoch": 0.54, + "learning_rate": 1.8131265064444424e-07, + "loss": 0.1882, + "step": 10420 + }, + { + "epoch": 0.54, + "learning_rate": 1.812789873603399e-07, + "loss": 0.1289, + "step": 10421 + }, + { + "epoch": 0.54, + "learning_rate": 1.8124532461128927e-07, + "loss": 0.1224, + "step": 10422 + }, + { + "epoch": 0.54, + "learning_rate": 1.8121166239825455e-07, + "loss": 0.115, + "step": 10423 + }, + { + "epoch": 0.54, + "learning_rate": 1.811780007221978e-07, + "loss": 0.1737, + "step": 10424 + }, + { + "epoch": 0.54, + "learning_rate": 1.8114433958408103e-07, + "loss": 0.1356, + "step": 10425 + }, + { + "epoch": 0.54, + "learning_rate": 1.8111067898486626e-07, + "loss": 0.1483, + "step": 10426 + }, + { + "epoch": 0.54, + "learning_rate": 1.8107701892551562e-07, + "loss": 0.1178, + "step": 10427 + }, + { + "epoch": 0.54, + "learning_rate": 1.8104335940699101e-07, + "loss": 0.1874, + "step": 10428 + }, + { + "epoch": 0.54, + "learning_rate": 1.810097004302546e-07, + "loss": 0.1805, + "step": 10429 + }, + { + "epoch": 0.54, + "learning_rate": 1.8097604199626822e-07, + "loss": 0.1189, + "step": 10430 + }, + { + "epoch": 0.54, + "learning_rate": 1.8094238410599394e-07, + "loss": 0.1272, + "step": 10431 + }, + { + "epoch": 0.54, + "learning_rate": 1.8090872676039374e-07, + "loss": 0.1087, + "step": 10432 + }, + { + "epoch": 0.54, + "learning_rate": 1.808750699604294e-07, + "loss": 0.1401, + "step": 10433 + }, + { + "epoch": 0.54, + "learning_rate": 1.80841413707063e-07, + "loss": 0.1503, + "step": 10434 + }, + { + "epoch": 0.54, + "learning_rate": 1.8080775800125632e-07, + "loss": 0.1095, + "step": 10435 + }, + { + "epoch": 0.54, + "learning_rate": 1.807741028439714e-07, + "loss": 0.144, + "step": 10436 + }, + { + "epoch": 0.54, + "learning_rate": 1.8074044823617e-07, + "loss": 0.1354, + "step": 10437 + }, + { + "epoch": 0.54, + "learning_rate": 1.8070679417881403e-07, + "loss": 0.1325, + "step": 10438 + }, + { + "epoch": 0.54, + "learning_rate": 1.806731406728653e-07, + "loss": 0.1316, + "step": 10439 + }, + { + "epoch": 0.54, + "learning_rate": 1.8063948771928574e-07, + "loss": 0.1586, + "step": 10440 + }, + { + "epoch": 0.54, + "learning_rate": 1.8060583531903714e-07, + "loss": 0.1498, + "step": 10441 + }, + { + "epoch": 0.55, + "learning_rate": 1.8057218347308112e-07, + "loss": 0.1232, + "step": 10442 + }, + { + "epoch": 0.55, + "learning_rate": 1.8053853218237966e-07, + "loss": 0.1293, + "step": 10443 + }, + { + "epoch": 0.55, + "learning_rate": 1.8050488144789443e-07, + "loss": 0.1077, + "step": 10444 + }, + { + "epoch": 0.55, + "learning_rate": 1.8047123127058726e-07, + "loss": 0.1346, + "step": 10445 + }, + { + "epoch": 0.55, + "learning_rate": 1.804375816514198e-07, + "loss": 0.1311, + "step": 10446 + }, + { + "epoch": 0.55, + "learning_rate": 1.8040393259135385e-07, + "loss": 0.1243, + "step": 10447 + }, + { + "epoch": 0.55, + "learning_rate": 1.80370284091351e-07, + "loss": 0.1315, + "step": 10448 + }, + { + "epoch": 0.55, + "learning_rate": 1.8033663615237309e-07, + "loss": 0.1411, + "step": 10449 + }, + { + "epoch": 0.55, + "learning_rate": 1.8030298877538172e-07, + "loss": 0.1322, + "step": 10450 + }, + { + "epoch": 0.55, + "learning_rate": 1.8026934196133849e-07, + "loss": 0.1251, + "step": 10451 + }, + { + "epoch": 0.55, + "learning_rate": 1.802356957112051e-07, + "loss": 0.1747, + "step": 10452 + }, + { + "epoch": 0.55, + "learning_rate": 1.8020205002594312e-07, + "loss": 0.0834, + "step": 10453 + }, + { + "epoch": 0.55, + "learning_rate": 1.801684049065142e-07, + "loss": 0.1697, + "step": 10454 + }, + { + "epoch": 0.55, + "learning_rate": 1.801347603538799e-07, + "loss": 0.1296, + "step": 10455 + }, + { + "epoch": 0.55, + "learning_rate": 1.8010111636900189e-07, + "loss": 0.1936, + "step": 10456 + }, + { + "epoch": 0.55, + "learning_rate": 1.800674729528416e-07, + "loss": 0.134, + "step": 10457 + }, + { + "epoch": 0.55, + "learning_rate": 1.800338301063607e-07, + "loss": 0.1421, + "step": 10458 + }, + { + "epoch": 0.55, + "learning_rate": 1.800001878305206e-07, + "loss": 0.1165, + "step": 10459 + }, + { + "epoch": 0.55, + "learning_rate": 1.7996654612628283e-07, + "loss": 0.1242, + "step": 10460 + }, + { + "epoch": 0.55, + "learning_rate": 1.7993290499460893e-07, + "loss": 0.1284, + "step": 10461 + }, + { + "epoch": 0.55, + "learning_rate": 1.7989926443646032e-07, + "loss": 0.1581, + "step": 10462 + }, + { + "epoch": 0.55, + "learning_rate": 1.7986562445279853e-07, + "loss": 0.131, + "step": 10463 + }, + { + "epoch": 0.55, + "learning_rate": 1.7983198504458492e-07, + "loss": 0.131, + "step": 10464 + }, + { + "epoch": 0.55, + "learning_rate": 1.79798346212781e-07, + "loss": 0.1349, + "step": 10465 + }, + { + "epoch": 0.55, + "learning_rate": 1.7976470795834813e-07, + "loss": 0.1202, + "step": 10466 + }, + { + "epoch": 0.55, + "learning_rate": 1.797310702822478e-07, + "loss": 0.1631, + "step": 10467 + }, + { + "epoch": 0.55, + "learning_rate": 1.7969743318544122e-07, + "loss": 0.164, + "step": 10468 + }, + { + "epoch": 0.55, + "learning_rate": 1.796637966688898e-07, + "loss": 0.1424, + "step": 10469 + }, + { + "epoch": 0.55, + "learning_rate": 1.7963016073355502e-07, + "loss": 0.129, + "step": 10470 + }, + { + "epoch": 0.55, + "learning_rate": 1.79596525380398e-07, + "loss": 0.1208, + "step": 10471 + }, + { + "epoch": 0.55, + "learning_rate": 1.7956289061038026e-07, + "loss": 0.1095, + "step": 10472 + }, + { + "epoch": 0.55, + "learning_rate": 1.7952925642446292e-07, + "loss": 0.1459, + "step": 10473 + }, + { + "epoch": 0.55, + "learning_rate": 1.794956228236074e-07, + "loss": 0.1472, + "step": 10474 + }, + { + "epoch": 0.55, + "learning_rate": 1.794619898087749e-07, + "loss": 0.1196, + "step": 10475 + }, + { + "epoch": 0.55, + "learning_rate": 1.7942835738092663e-07, + "loss": 0.1532, + "step": 10476 + }, + { + "epoch": 0.55, + "learning_rate": 1.7939472554102383e-07, + "loss": 0.134, + "step": 10477 + }, + { + "epoch": 0.55, + "learning_rate": 1.7936109429002772e-07, + "loss": 0.1349, + "step": 10478 + }, + { + "epoch": 0.55, + "learning_rate": 1.7932746362889956e-07, + "loss": 0.1092, + "step": 10479 + }, + { + "epoch": 0.55, + "learning_rate": 1.7929383355860038e-07, + "loss": 0.1321, + "step": 10480 + }, + { + "epoch": 0.55, + "learning_rate": 1.7926020408009152e-07, + "loss": 0.1787, + "step": 10481 + }, + { + "epoch": 0.55, + "learning_rate": 1.7922657519433396e-07, + "loss": 0.1353, + "step": 10482 + }, + { + "epoch": 0.55, + "learning_rate": 1.7919294690228895e-07, + "loss": 0.1167, + "step": 10483 + }, + { + "epoch": 0.55, + "learning_rate": 1.7915931920491761e-07, + "loss": 0.179, + "step": 10484 + }, + { + "epoch": 0.55, + "learning_rate": 1.7912569210318087e-07, + "loss": 0.1167, + "step": 10485 + }, + { + "epoch": 0.55, + "learning_rate": 1.7909206559803996e-07, + "loss": 0.1475, + "step": 10486 + }, + { + "epoch": 0.55, + "learning_rate": 1.7905843969045585e-07, + "loss": 0.1214, + "step": 10487 + }, + { + "epoch": 0.55, + "learning_rate": 1.7902481438138962e-07, + "loss": 0.1652, + "step": 10488 + }, + { + "epoch": 0.55, + "learning_rate": 1.7899118967180228e-07, + "loss": 0.1307, + "step": 10489 + }, + { + "epoch": 0.55, + "learning_rate": 1.7895756556265491e-07, + "loss": 0.1231, + "step": 10490 + }, + { + "epoch": 0.55, + "learning_rate": 1.7892394205490838e-07, + "loss": 0.1414, + "step": 10491 + }, + { + "epoch": 0.55, + "learning_rate": 1.788903191495238e-07, + "loss": 0.1086, + "step": 10492 + }, + { + "epoch": 0.55, + "learning_rate": 1.78856696847462e-07, + "loss": 0.1479, + "step": 10493 + }, + { + "epoch": 0.55, + "learning_rate": 1.7882307514968392e-07, + "loss": 0.1506, + "step": 10494 + }, + { + "epoch": 0.55, + "learning_rate": 1.787894540571506e-07, + "loss": 0.0938, + "step": 10495 + }, + { + "epoch": 0.55, + "learning_rate": 1.7875583357082283e-07, + "loss": 0.1526, + "step": 10496 + }, + { + "epoch": 0.55, + "learning_rate": 1.7872221369166157e-07, + "loss": 0.1336, + "step": 10497 + }, + { + "epoch": 0.55, + "learning_rate": 1.786885944206276e-07, + "loss": 0.1794, + "step": 10498 + }, + { + "epoch": 0.55, + "learning_rate": 1.7865497575868187e-07, + "loss": 0.1542, + "step": 10499 + }, + { + "epoch": 0.55, + "learning_rate": 1.7862135770678512e-07, + "loss": 0.1728, + "step": 10500 + }, + { + "epoch": 0.55, + "learning_rate": 1.7858774026589833e-07, + "loss": 0.1188, + "step": 10501 + }, + { + "epoch": 0.55, + "learning_rate": 1.7855412343698215e-07, + "loss": 0.1144, + "step": 10502 + }, + { + "epoch": 0.55, + "learning_rate": 1.7852050722099735e-07, + "loss": 0.1551, + "step": 10503 + }, + { + "epoch": 0.55, + "learning_rate": 1.7848689161890477e-07, + "loss": 0.193, + "step": 10504 + }, + { + "epoch": 0.55, + "learning_rate": 1.7845327663166512e-07, + "loss": 0.1264, + "step": 10505 + }, + { + "epoch": 0.55, + "learning_rate": 1.7841966226023917e-07, + "loss": 0.1302, + "step": 10506 + }, + { + "epoch": 0.55, + "learning_rate": 1.7838604850558757e-07, + "loss": 0.1242, + "step": 10507 + }, + { + "epoch": 0.55, + "learning_rate": 1.7835243536867106e-07, + "loss": 0.1414, + "step": 10508 + }, + { + "epoch": 0.55, + "learning_rate": 1.783188228504503e-07, + "loss": 0.1106, + "step": 10509 + }, + { + "epoch": 0.55, + "learning_rate": 1.7828521095188602e-07, + "loss": 0.1577, + "step": 10510 + }, + { + "epoch": 0.55, + "learning_rate": 1.7825159967393875e-07, + "loss": 0.1334, + "step": 10511 + }, + { + "epoch": 0.55, + "learning_rate": 1.782179890175691e-07, + "loss": 0.1505, + "step": 10512 + }, + { + "epoch": 0.55, + "learning_rate": 1.781843789837378e-07, + "loss": 0.1457, + "step": 10513 + }, + { + "epoch": 0.55, + "learning_rate": 1.781507695734053e-07, + "loss": 0.1441, + "step": 10514 + }, + { + "epoch": 0.55, + "learning_rate": 1.7811716078753233e-07, + "loss": 0.1413, + "step": 10515 + }, + { + "epoch": 0.55, + "learning_rate": 1.7808355262707928e-07, + "loss": 0.1211, + "step": 10516 + }, + { + "epoch": 0.55, + "learning_rate": 1.7804994509300678e-07, + "loss": 0.1547, + "step": 10517 + }, + { + "epoch": 0.55, + "learning_rate": 1.780163381862754e-07, + "loss": 0.1399, + "step": 10518 + }, + { + "epoch": 0.55, + "learning_rate": 1.7798273190784546e-07, + "loss": 0.1346, + "step": 10519 + }, + { + "epoch": 0.55, + "learning_rate": 1.779491262586776e-07, + "loss": 0.1353, + "step": 10520 + }, + { + "epoch": 0.55, + "learning_rate": 1.7791552123973213e-07, + "loss": 0.1066, + "step": 10521 + }, + { + "epoch": 0.55, + "learning_rate": 1.7788191685196968e-07, + "loss": 0.1081, + "step": 10522 + }, + { + "epoch": 0.55, + "learning_rate": 1.7784831309635054e-07, + "loss": 0.1352, + "step": 10523 + }, + { + "epoch": 0.55, + "learning_rate": 1.7781470997383522e-07, + "loss": 0.157, + "step": 10524 + }, + { + "epoch": 0.55, + "learning_rate": 1.7778110748538399e-07, + "loss": 0.1305, + "step": 10525 + }, + { + "epoch": 0.55, + "learning_rate": 1.777475056319573e-07, + "loss": 0.125, + "step": 10526 + }, + { + "epoch": 0.55, + "learning_rate": 1.7771390441451558e-07, + "loss": 0.1491, + "step": 10527 + }, + { + "epoch": 0.55, + "learning_rate": 1.77680303834019e-07, + "loss": 0.1379, + "step": 10528 + }, + { + "epoch": 0.55, + "learning_rate": 1.776467038914279e-07, + "loss": 0.1229, + "step": 10529 + }, + { + "epoch": 0.55, + "learning_rate": 1.7761310458770272e-07, + "loss": 0.1194, + "step": 10530 + }, + { + "epoch": 0.55, + "learning_rate": 1.7757950592380358e-07, + "loss": 0.129, + "step": 10531 + }, + { + "epoch": 0.55, + "learning_rate": 1.7754590790069082e-07, + "loss": 0.1425, + "step": 10532 + }, + { + "epoch": 0.55, + "learning_rate": 1.775123105193247e-07, + "loss": 0.1286, + "step": 10533 + }, + { + "epoch": 0.55, + "learning_rate": 1.7747871378066542e-07, + "loss": 0.1495, + "step": 10534 + }, + { + "epoch": 0.55, + "learning_rate": 1.7744511768567332e-07, + "loss": 0.1549, + "step": 10535 + }, + { + "epoch": 0.55, + "learning_rate": 1.774115222353083e-07, + "loss": 0.1448, + "step": 10536 + }, + { + "epoch": 0.55, + "learning_rate": 1.7737792743053077e-07, + "loss": 0.1395, + "step": 10537 + }, + { + "epoch": 0.55, + "learning_rate": 1.7734433327230076e-07, + "loss": 0.1155, + "step": 10538 + }, + { + "epoch": 0.55, + "learning_rate": 1.7731073976157852e-07, + "loss": 0.1469, + "step": 10539 + }, + { + "epoch": 0.55, + "learning_rate": 1.77277146899324e-07, + "loss": 0.1746, + "step": 10540 + }, + { + "epoch": 0.55, + "learning_rate": 1.7724355468649746e-07, + "loss": 0.1458, + "step": 10541 + }, + { + "epoch": 0.55, + "learning_rate": 1.7720996312405886e-07, + "loss": 0.1227, + "step": 10542 + }, + { + "epoch": 0.55, + "learning_rate": 1.7717637221296838e-07, + "loss": 0.1442, + "step": 10543 + }, + { + "epoch": 0.55, + "learning_rate": 1.7714278195418602e-07, + "loss": 0.1448, + "step": 10544 + }, + { + "epoch": 0.55, + "learning_rate": 1.771091923486717e-07, + "loss": 0.1438, + "step": 10545 + }, + { + "epoch": 0.55, + "learning_rate": 1.770756033973855e-07, + "loss": 0.1916, + "step": 10546 + }, + { + "epoch": 0.55, + "learning_rate": 1.770420151012874e-07, + "loss": 0.1377, + "step": 10547 + }, + { + "epoch": 0.55, + "learning_rate": 1.7700842746133743e-07, + "loss": 0.1599, + "step": 10548 + }, + { + "epoch": 0.55, + "learning_rate": 1.769748404784954e-07, + "loss": 0.1321, + "step": 10549 + }, + { + "epoch": 0.55, + "learning_rate": 1.7694125415372138e-07, + "loss": 0.1139, + "step": 10550 + }, + { + "epoch": 0.55, + "learning_rate": 1.7690766848797515e-07, + "loss": 0.1467, + "step": 10551 + }, + { + "epoch": 0.55, + "learning_rate": 1.768740834822168e-07, + "loss": 0.1319, + "step": 10552 + }, + { + "epoch": 0.55, + "learning_rate": 1.76840499137406e-07, + "loss": 0.133, + "step": 10553 + }, + { + "epoch": 0.55, + "learning_rate": 1.7680691545450265e-07, + "loss": 0.1555, + "step": 10554 + }, + { + "epoch": 0.55, + "learning_rate": 1.7677333243446663e-07, + "loss": 0.1064, + "step": 10555 + }, + { + "epoch": 0.55, + "learning_rate": 1.767397500782577e-07, + "loss": 0.12, + "step": 10556 + }, + { + "epoch": 0.55, + "learning_rate": 1.767061683868357e-07, + "loss": 0.1427, + "step": 10557 + }, + { + "epoch": 0.55, + "learning_rate": 1.7667258736116042e-07, + "loss": 0.13, + "step": 10558 + }, + { + "epoch": 0.55, + "learning_rate": 1.7663900700219161e-07, + "loss": 0.1207, + "step": 10559 + }, + { + "epoch": 0.55, + "learning_rate": 1.7660542731088896e-07, + "loss": 0.1773, + "step": 10560 + }, + { + "epoch": 0.55, + "learning_rate": 1.765718482882123e-07, + "loss": 0.1091, + "step": 10561 + }, + { + "epoch": 0.55, + "learning_rate": 1.7653826993512124e-07, + "loss": 0.1462, + "step": 10562 + }, + { + "epoch": 0.55, + "learning_rate": 1.7650469225257543e-07, + "loss": 0.1432, + "step": 10563 + }, + { + "epoch": 0.55, + "learning_rate": 1.7647111524153463e-07, + "loss": 0.1182, + "step": 10564 + }, + { + "epoch": 0.55, + "learning_rate": 1.764375389029584e-07, + "loss": 0.126, + "step": 10565 + }, + { + "epoch": 0.55, + "learning_rate": 1.7640396323780645e-07, + "loss": 0.1282, + "step": 10566 + }, + { + "epoch": 0.55, + "learning_rate": 1.7637038824703829e-07, + "loss": 0.1462, + "step": 10567 + }, + { + "epoch": 0.55, + "learning_rate": 1.763368139316136e-07, + "loss": 0.1529, + "step": 10568 + }, + { + "epoch": 0.55, + "learning_rate": 1.7630324029249187e-07, + "loss": 0.173, + "step": 10569 + }, + { + "epoch": 0.55, + "learning_rate": 1.7626966733063273e-07, + "loss": 0.1435, + "step": 10570 + }, + { + "epoch": 0.55, + "learning_rate": 1.7623609504699563e-07, + "loss": 0.1676, + "step": 10571 + }, + { + "epoch": 0.55, + "learning_rate": 1.762025234425401e-07, + "loss": 0.1183, + "step": 10572 + }, + { + "epoch": 0.55, + "learning_rate": 1.7616895251822564e-07, + "loss": 0.1285, + "step": 10573 + }, + { + "epoch": 0.55, + "learning_rate": 1.7613538227501168e-07, + "loss": 0.1162, + "step": 10574 + }, + { + "epoch": 0.55, + "learning_rate": 1.7610181271385777e-07, + "loss": 0.0902, + "step": 10575 + }, + { + "epoch": 0.55, + "learning_rate": 1.7606824383572322e-07, + "loss": 0.1698, + "step": 10576 + }, + { + "epoch": 0.55, + "learning_rate": 1.7603467564156754e-07, + "loss": 0.1228, + "step": 10577 + }, + { + "epoch": 0.55, + "learning_rate": 1.7600110813235015e-07, + "loss": 0.172, + "step": 10578 + }, + { + "epoch": 0.55, + "learning_rate": 1.7596754130903024e-07, + "loss": 0.1171, + "step": 10579 + }, + { + "epoch": 0.55, + "learning_rate": 1.7593397517256733e-07, + "loss": 0.1327, + "step": 10580 + }, + { + "epoch": 0.55, + "learning_rate": 1.7590040972392063e-07, + "loss": 0.1375, + "step": 10581 + }, + { + "epoch": 0.55, + "learning_rate": 1.758668449640496e-07, + "loss": 0.103, + "step": 10582 + }, + { + "epoch": 0.55, + "learning_rate": 1.758332808939134e-07, + "loss": 0.1436, + "step": 10583 + }, + { + "epoch": 0.55, + "learning_rate": 1.7579971751447143e-07, + "loss": 0.1326, + "step": 10584 + }, + { + "epoch": 0.55, + "learning_rate": 1.757661548266828e-07, + "loss": 0.1338, + "step": 10585 + }, + { + "epoch": 0.55, + "learning_rate": 1.7573259283150686e-07, + "loss": 0.1487, + "step": 10586 + }, + { + "epoch": 0.55, + "learning_rate": 1.7569903152990286e-07, + "loss": 0.1327, + "step": 10587 + }, + { + "epoch": 0.55, + "learning_rate": 1.7566547092282984e-07, + "loss": 0.1411, + "step": 10588 + }, + { + "epoch": 0.55, + "learning_rate": 1.7563191101124705e-07, + "loss": 0.1643, + "step": 10589 + }, + { + "epoch": 0.55, + "learning_rate": 1.7559835179611365e-07, + "loss": 0.1456, + "step": 10590 + }, + { + "epoch": 0.55, + "learning_rate": 1.755647932783888e-07, + "loss": 0.1635, + "step": 10591 + }, + { + "epoch": 0.55, + "learning_rate": 1.7553123545903155e-07, + "loss": 0.1229, + "step": 10592 + }, + { + "epoch": 0.55, + "learning_rate": 1.7549767833900109e-07, + "loss": 0.1276, + "step": 10593 + }, + { + "epoch": 0.55, + "learning_rate": 1.7546412191925638e-07, + "loss": 0.123, + "step": 10594 + }, + { + "epoch": 0.55, + "learning_rate": 1.7543056620075662e-07, + "loss": 0.1308, + "step": 10595 + }, + { + "epoch": 0.55, + "learning_rate": 1.7539701118446075e-07, + "loss": 0.1155, + "step": 10596 + }, + { + "epoch": 0.55, + "learning_rate": 1.7536345687132772e-07, + "loss": 0.1426, + "step": 10597 + }, + { + "epoch": 0.55, + "learning_rate": 1.7532990326231667e-07, + "loss": 0.1248, + "step": 10598 + }, + { + "epoch": 0.55, + "learning_rate": 1.7529635035838645e-07, + "loss": 0.1159, + "step": 10599 + }, + { + "epoch": 0.55, + "learning_rate": 1.7526279816049613e-07, + "loss": 0.1315, + "step": 10600 + }, + { + "epoch": 0.55, + "learning_rate": 1.7522924666960454e-07, + "loss": 0.1233, + "step": 10601 + }, + { + "epoch": 0.55, + "learning_rate": 1.7519569588667068e-07, + "loss": 0.1167, + "step": 10602 + }, + { + "epoch": 0.55, + "learning_rate": 1.7516214581265338e-07, + "loss": 0.1514, + "step": 10603 + }, + { + "epoch": 0.55, + "learning_rate": 1.7512859644851162e-07, + "loss": 0.145, + "step": 10604 + }, + { + "epoch": 0.55, + "learning_rate": 1.7509504779520412e-07, + "loss": 0.1292, + "step": 10605 + }, + { + "epoch": 0.55, + "learning_rate": 1.7506149985368974e-07, + "loss": 0.1343, + "step": 10606 + }, + { + "epoch": 0.55, + "learning_rate": 1.7502795262492736e-07, + "loss": 0.1246, + "step": 10607 + }, + { + "epoch": 0.55, + "learning_rate": 1.7499440610987569e-07, + "loss": 0.1462, + "step": 10608 + }, + { + "epoch": 0.55, + "learning_rate": 1.749608603094936e-07, + "loss": 0.1253, + "step": 10609 + }, + { + "epoch": 0.55, + "learning_rate": 1.7492731522473972e-07, + "loss": 0.1573, + "step": 10610 + }, + { + "epoch": 0.55, + "learning_rate": 1.748937708565729e-07, + "loss": 0.1445, + "step": 10611 + }, + { + "epoch": 0.55, + "learning_rate": 1.7486022720595185e-07, + "loss": 0.1214, + "step": 10612 + }, + { + "epoch": 0.55, + "learning_rate": 1.7482668427383514e-07, + "loss": 0.1535, + "step": 10613 + }, + { + "epoch": 0.55, + "learning_rate": 1.7479314206118154e-07, + "loss": 0.1374, + "step": 10614 + }, + { + "epoch": 0.55, + "learning_rate": 1.7475960056894964e-07, + "loss": 0.1392, + "step": 10615 + }, + { + "epoch": 0.55, + "learning_rate": 1.747260597980981e-07, + "loss": 0.1627, + "step": 10616 + }, + { + "epoch": 0.55, + "learning_rate": 1.7469251974958553e-07, + "loss": 0.1422, + "step": 10617 + }, + { + "epoch": 0.55, + "learning_rate": 1.7465898042437056e-07, + "loss": 0.12, + "step": 10618 + }, + { + "epoch": 0.55, + "learning_rate": 1.7462544182341164e-07, + "loss": 0.1348, + "step": 10619 + }, + { + "epoch": 0.55, + "learning_rate": 1.7459190394766745e-07, + "loss": 0.1273, + "step": 10620 + }, + { + "epoch": 0.55, + "learning_rate": 1.745583667980965e-07, + "loss": 0.1858, + "step": 10621 + }, + { + "epoch": 0.55, + "learning_rate": 1.7452483037565715e-07, + "loss": 0.1362, + "step": 10622 + }, + { + "epoch": 0.55, + "learning_rate": 1.7449129468130802e-07, + "loss": 0.1168, + "step": 10623 + }, + { + "epoch": 0.55, + "learning_rate": 1.7445775971600752e-07, + "loss": 0.1293, + "step": 10624 + }, + { + "epoch": 0.55, + "learning_rate": 1.7442422548071412e-07, + "loss": 0.1305, + "step": 10625 + }, + { + "epoch": 0.55, + "learning_rate": 1.743906919763862e-07, + "loss": 0.1421, + "step": 10626 + }, + { + "epoch": 0.55, + "learning_rate": 1.743571592039822e-07, + "loss": 0.1231, + "step": 10627 + }, + { + "epoch": 0.55, + "learning_rate": 1.743236271644605e-07, + "loss": 0.1224, + "step": 10628 + }, + { + "epoch": 0.55, + "learning_rate": 1.7429009585877945e-07, + "loss": 0.1263, + "step": 10629 + }, + { + "epoch": 0.55, + "learning_rate": 1.7425656528789748e-07, + "loss": 0.1446, + "step": 10630 + }, + { + "epoch": 0.55, + "learning_rate": 1.7422303545277268e-07, + "loss": 0.132, + "step": 10631 + }, + { + "epoch": 0.55, + "learning_rate": 1.7418950635436352e-07, + "loss": 0.1645, + "step": 10632 + }, + { + "epoch": 0.55, + "learning_rate": 1.741559779936282e-07, + "loss": 0.1262, + "step": 10633 + }, + { + "epoch": 0.56, + "learning_rate": 1.7412245037152506e-07, + "loss": 0.1564, + "step": 10634 + }, + { + "epoch": 0.56, + "learning_rate": 1.7408892348901221e-07, + "loss": 0.1129, + "step": 10635 + }, + { + "epoch": 0.56, + "learning_rate": 1.74055397347048e-07, + "loss": 0.1556, + "step": 10636 + }, + { + "epoch": 0.56, + "learning_rate": 1.7402187194659048e-07, + "loss": 0.1418, + "step": 10637 + }, + { + "epoch": 0.56, + "learning_rate": 1.7398834728859805e-07, + "loss": 0.1292, + "step": 10638 + }, + { + "epoch": 0.56, + "learning_rate": 1.7395482337402854e-07, + "loss": 0.1484, + "step": 10639 + }, + { + "epoch": 0.56, + "learning_rate": 1.739213002038403e-07, + "loss": 0.122, + "step": 10640 + }, + { + "epoch": 0.56, + "learning_rate": 1.7388777777899128e-07, + "loss": 0.1069, + "step": 10641 + }, + { + "epoch": 0.56, + "learning_rate": 1.7385425610043973e-07, + "loss": 0.147, + "step": 10642 + }, + { + "epoch": 0.56, + "learning_rate": 1.7382073516914358e-07, + "loss": 0.2195, + "step": 10643 + }, + { + "epoch": 0.56, + "learning_rate": 1.7378721498606094e-07, + "loss": 0.1514, + "step": 10644 + }, + { + "epoch": 0.56, + "learning_rate": 1.7375369555214985e-07, + "loss": 0.1231, + "step": 10645 + }, + { + "epoch": 0.56, + "learning_rate": 1.7372017686836822e-07, + "loss": 0.137, + "step": 10646 + }, + { + "epoch": 0.56, + "learning_rate": 1.7368665893567422e-07, + "loss": 0.1544, + "step": 10647 + }, + { + "epoch": 0.56, + "learning_rate": 1.736531417550255e-07, + "loss": 0.1534, + "step": 10648 + }, + { + "epoch": 0.56, + "learning_rate": 1.7361962532738025e-07, + "loss": 0.1379, + "step": 10649 + }, + { + "epoch": 0.56, + "learning_rate": 1.7358610965369625e-07, + "loss": 0.1035, + "step": 10650 + }, + { + "epoch": 0.56, + "learning_rate": 1.7355259473493145e-07, + "loss": 0.1524, + "step": 10651 + }, + { + "epoch": 0.56, + "learning_rate": 1.7351908057204368e-07, + "loss": 0.1092, + "step": 10652 + }, + { + "epoch": 0.56, + "learning_rate": 1.7348556716599084e-07, + "loss": 0.1513, + "step": 10653 + }, + { + "epoch": 0.56, + "learning_rate": 1.7345205451773067e-07, + "loss": 0.1118, + "step": 10654 + }, + { + "epoch": 0.56, + "learning_rate": 1.7341854262822115e-07, + "loss": 0.1549, + "step": 10655 + }, + { + "epoch": 0.56, + "learning_rate": 1.7338503149841988e-07, + "loss": 0.1409, + "step": 10656 + }, + { + "epoch": 0.56, + "learning_rate": 1.7335152112928464e-07, + "loss": 0.1298, + "step": 10657 + }, + { + "epoch": 0.56, + "learning_rate": 1.7331801152177325e-07, + "loss": 0.1279, + "step": 10658 + }, + { + "epoch": 0.56, + "learning_rate": 1.7328450267684335e-07, + "loss": 0.1478, + "step": 10659 + }, + { + "epoch": 0.56, + "learning_rate": 1.7325099459545273e-07, + "loss": 0.1353, + "step": 10660 + }, + { + "epoch": 0.56, + "learning_rate": 1.73217487278559e-07, + "loss": 0.1391, + "step": 10661 + }, + { + "epoch": 0.56, + "learning_rate": 1.7318398072711982e-07, + "loss": 0.1862, + "step": 10662 + }, + { + "epoch": 0.56, + "learning_rate": 1.731504749420928e-07, + "loss": 0.1355, + "step": 10663 + }, + { + "epoch": 0.56, + "learning_rate": 1.7311696992443563e-07, + "loss": 0.1283, + "step": 10664 + }, + { + "epoch": 0.56, + "learning_rate": 1.730834656751058e-07, + "loss": 0.152, + "step": 10665 + }, + { + "epoch": 0.56, + "learning_rate": 1.730499621950609e-07, + "loss": 0.1147, + "step": 10666 + }, + { + "epoch": 0.56, + "learning_rate": 1.730164594852585e-07, + "loss": 0.1762, + "step": 10667 + }, + { + "epoch": 0.56, + "learning_rate": 1.7298295754665607e-07, + "loss": 0.1439, + "step": 10668 + }, + { + "epoch": 0.56, + "learning_rate": 1.7294945638021118e-07, + "loss": 0.1367, + "step": 10669 + }, + { + "epoch": 0.56, + "learning_rate": 1.7291595598688122e-07, + "loss": 0.1838, + "step": 10670 + }, + { + "epoch": 0.56, + "learning_rate": 1.7288245636762375e-07, + "loss": 0.0954, + "step": 10671 + }, + { + "epoch": 0.56, + "learning_rate": 1.7284895752339618e-07, + "loss": 0.1445, + "step": 10672 + }, + { + "epoch": 0.56, + "learning_rate": 1.728154594551558e-07, + "loss": 0.1434, + "step": 10673 + }, + { + "epoch": 0.56, + "learning_rate": 1.7278196216386015e-07, + "loss": 0.1481, + "step": 10674 + }, + { + "epoch": 0.56, + "learning_rate": 1.7274846565046648e-07, + "loss": 0.1513, + "step": 10675 + }, + { + "epoch": 0.56, + "learning_rate": 1.7271496991593218e-07, + "loss": 0.1319, + "step": 10676 + }, + { + "epoch": 0.56, + "learning_rate": 1.7268147496121459e-07, + "loss": 0.1314, + "step": 10677 + }, + { + "epoch": 0.56, + "learning_rate": 1.7264798078727102e-07, + "loss": 0.1241, + "step": 10678 + }, + { + "epoch": 0.56, + "learning_rate": 1.7261448739505863e-07, + "loss": 0.1244, + "step": 10679 + }, + { + "epoch": 0.56, + "learning_rate": 1.7258099478553484e-07, + "loss": 0.1825, + "step": 10680 + }, + { + "epoch": 0.56, + "learning_rate": 1.7254750295965684e-07, + "loss": 0.1144, + "step": 10681 + }, + { + "epoch": 0.56, + "learning_rate": 1.7251401191838173e-07, + "loss": 0.1574, + "step": 10682 + }, + { + "epoch": 0.56, + "learning_rate": 1.724805216626668e-07, + "loss": 0.135, + "step": 10683 + }, + { + "epoch": 0.56, + "learning_rate": 1.7244703219346915e-07, + "loss": 0.1665, + "step": 10684 + }, + { + "epoch": 0.56, + "learning_rate": 1.72413543511746e-07, + "loss": 0.1219, + "step": 10685 + }, + { + "epoch": 0.56, + "learning_rate": 1.7238005561845436e-07, + "loss": 0.1578, + "step": 10686 + }, + { + "epoch": 0.56, + "learning_rate": 1.7234656851455147e-07, + "loss": 0.1376, + "step": 10687 + }, + { + "epoch": 0.56, + "learning_rate": 1.7231308220099424e-07, + "loss": 0.1438, + "step": 10688 + }, + { + "epoch": 0.56, + "learning_rate": 1.7227959667873994e-07, + "loss": 0.1646, + "step": 10689 + }, + { + "epoch": 0.56, + "learning_rate": 1.722461119487454e-07, + "loss": 0.1085, + "step": 10690 + }, + { + "epoch": 0.56, + "learning_rate": 1.7221262801196768e-07, + "loss": 0.1414, + "step": 10691 + }, + { + "epoch": 0.56, + "learning_rate": 1.721791448693638e-07, + "loss": 0.1591, + "step": 10692 + }, + { + "epoch": 0.56, + "learning_rate": 1.7214566252189065e-07, + "loss": 0.1256, + "step": 10693 + }, + { + "epoch": 0.56, + "learning_rate": 1.7211218097050527e-07, + "loss": 0.148, + "step": 10694 + }, + { + "epoch": 0.56, + "learning_rate": 1.720787002161645e-07, + "loss": 0.1236, + "step": 10695 + }, + { + "epoch": 0.56, + "learning_rate": 1.7204522025982527e-07, + "loss": 0.1572, + "step": 10696 + }, + { + "epoch": 0.56, + "learning_rate": 1.7201174110244438e-07, + "loss": 0.1225, + "step": 10697 + }, + { + "epoch": 0.56, + "learning_rate": 1.7197826274497885e-07, + "loss": 0.1459, + "step": 10698 + }, + { + "epoch": 0.56, + "learning_rate": 1.7194478518838533e-07, + "loss": 0.1152, + "step": 10699 + }, + { + "epoch": 0.56, + "learning_rate": 1.7191130843362064e-07, + "loss": 0.137, + "step": 10700 + }, + { + "epoch": 0.56, + "learning_rate": 1.7187783248164165e-07, + "loss": 0.1181, + "step": 10701 + }, + { + "epoch": 0.56, + "learning_rate": 1.7184435733340502e-07, + "loss": 0.1047, + "step": 10702 + }, + { + "epoch": 0.56, + "learning_rate": 1.7181088298986756e-07, + "loss": 0.1204, + "step": 10703 + }, + { + "epoch": 0.56, + "learning_rate": 1.717774094519859e-07, + "loss": 0.1342, + "step": 10704 + }, + { + "epoch": 0.56, + "learning_rate": 1.7174393672071684e-07, + "loss": 0.1426, + "step": 10705 + }, + { + "epoch": 0.56, + "learning_rate": 1.7171046479701693e-07, + "loss": 0.1384, + "step": 10706 + }, + { + "epoch": 0.56, + "learning_rate": 1.7167699368184292e-07, + "loss": 0.099, + "step": 10707 + }, + { + "epoch": 0.56, + "learning_rate": 1.7164352337615133e-07, + "loss": 0.1039, + "step": 10708 + }, + { + "epoch": 0.56, + "learning_rate": 1.7161005388089875e-07, + "loss": 0.1492, + "step": 10709 + }, + { + "epoch": 0.56, + "learning_rate": 1.7157658519704184e-07, + "loss": 0.1359, + "step": 10710 + }, + { + "epoch": 0.56, + "learning_rate": 1.7154311732553704e-07, + "loss": 0.1502, + "step": 10711 + }, + { + "epoch": 0.56, + "learning_rate": 1.71509650267341e-07, + "loss": 0.1722, + "step": 10712 + }, + { + "epoch": 0.56, + "learning_rate": 1.7147618402341006e-07, + "loss": 0.1188, + "step": 10713 + }, + { + "epoch": 0.56, + "learning_rate": 1.7144271859470087e-07, + "loss": 0.1171, + "step": 10714 + }, + { + "epoch": 0.56, + "learning_rate": 1.7140925398216984e-07, + "loss": 0.1394, + "step": 10715 + }, + { + "epoch": 0.56, + "learning_rate": 1.713757901867733e-07, + "loss": 0.1243, + "step": 10716 + }, + { + "epoch": 0.56, + "learning_rate": 1.7134232720946775e-07, + "loss": 0.1421, + "step": 10717 + }, + { + "epoch": 0.56, + "learning_rate": 1.7130886505120952e-07, + "loss": 0.1524, + "step": 10718 + }, + { + "epoch": 0.56, + "learning_rate": 1.7127540371295503e-07, + "loss": 0.1379, + "step": 10719 + }, + { + "epoch": 0.56, + "learning_rate": 1.7124194319566056e-07, + "loss": 0.141, + "step": 10720 + }, + { + "epoch": 0.56, + "learning_rate": 1.712084835002825e-07, + "loss": 0.1603, + "step": 10721 + }, + { + "epoch": 0.56, + "learning_rate": 1.7117502462777707e-07, + "loss": 0.1517, + "step": 10722 + }, + { + "epoch": 0.56, + "learning_rate": 1.7114156657910057e-07, + "loss": 0.1325, + "step": 10723 + }, + { + "epoch": 0.56, + "learning_rate": 1.7110810935520933e-07, + "loss": 0.1343, + "step": 10724 + }, + { + "epoch": 0.56, + "learning_rate": 1.7107465295705935e-07, + "loss": 0.1131, + "step": 10725 + }, + { + "epoch": 0.56, + "learning_rate": 1.7104119738560701e-07, + "loss": 0.1088, + "step": 10726 + }, + { + "epoch": 0.56, + "learning_rate": 1.7100774264180836e-07, + "loss": 0.1545, + "step": 10727 + }, + { + "epoch": 0.56, + "learning_rate": 1.709742887266197e-07, + "loss": 0.1147, + "step": 10728 + }, + { + "epoch": 0.56, + "learning_rate": 1.70940835640997e-07, + "loss": 0.1188, + "step": 10729 + }, + { + "epoch": 0.56, + "learning_rate": 1.709073833858965e-07, + "loss": 0.1192, + "step": 10730 + }, + { + "epoch": 0.56, + "learning_rate": 1.7087393196227415e-07, + "loss": 0.1357, + "step": 10731 + }, + { + "epoch": 0.56, + "learning_rate": 1.7084048137108618e-07, + "loss": 0.154, + "step": 10732 + }, + { + "epoch": 0.56, + "learning_rate": 1.7080703161328844e-07, + "loss": 0.1522, + "step": 10733 + }, + { + "epoch": 0.56, + "learning_rate": 1.7077358268983698e-07, + "loss": 0.1325, + "step": 10734 + }, + { + "epoch": 0.56, + "learning_rate": 1.7074013460168786e-07, + "loss": 0.1411, + "step": 10735 + }, + { + "epoch": 0.56, + "learning_rate": 1.7070668734979692e-07, + "loss": 0.1653, + "step": 10736 + }, + { + "epoch": 0.56, + "learning_rate": 1.7067324093512022e-07, + "loss": 0.1371, + "step": 10737 + }, + { + "epoch": 0.56, + "learning_rate": 1.7063979535861358e-07, + "loss": 0.1554, + "step": 10738 + }, + { + "epoch": 0.56, + "learning_rate": 1.7060635062123295e-07, + "loss": 0.1191, + "step": 10739 + }, + { + "epoch": 0.56, + "learning_rate": 1.7057290672393413e-07, + "loss": 0.096, + "step": 10740 + }, + { + "epoch": 0.56, + "learning_rate": 1.7053946366767309e-07, + "loss": 0.1558, + "step": 10741 + }, + { + "epoch": 0.56, + "learning_rate": 1.7050602145340552e-07, + "loss": 0.175, + "step": 10742 + }, + { + "epoch": 0.56, + "learning_rate": 1.7047258008208719e-07, + "loss": 0.1386, + "step": 10743 + }, + { + "epoch": 0.56, + "learning_rate": 1.7043913955467395e-07, + "loss": 0.1249, + "step": 10744 + }, + { + "epoch": 0.56, + "learning_rate": 1.7040569987212145e-07, + "loss": 0.156, + "step": 10745 + }, + { + "epoch": 0.56, + "learning_rate": 1.7037226103538555e-07, + "loss": 0.1386, + "step": 10746 + }, + { + "epoch": 0.56, + "learning_rate": 1.7033882304542182e-07, + "loss": 0.1326, + "step": 10747 + }, + { + "epoch": 0.56, + "learning_rate": 1.7030538590318603e-07, + "loss": 0.1542, + "step": 10748 + }, + { + "epoch": 0.56, + "learning_rate": 1.7027194960963378e-07, + "loss": 0.1523, + "step": 10749 + }, + { + "epoch": 0.56, + "learning_rate": 1.7023851416572068e-07, + "loss": 0.1213, + "step": 10750 + }, + { + "epoch": 0.56, + "learning_rate": 1.7020507957240226e-07, + "loss": 0.1286, + "step": 10751 + }, + { + "epoch": 0.56, + "learning_rate": 1.7017164583063424e-07, + "loss": 0.18, + "step": 10752 + }, + { + "epoch": 0.56, + "learning_rate": 1.7013821294137202e-07, + "loss": 0.1246, + "step": 10753 + }, + { + "epoch": 0.56, + "learning_rate": 1.701047809055713e-07, + "loss": 0.1387, + "step": 10754 + }, + { + "epoch": 0.56, + "learning_rate": 1.700713497241874e-07, + "loss": 0.1454, + "step": 10755 + }, + { + "epoch": 0.56, + "learning_rate": 1.700379193981759e-07, + "loss": 0.1465, + "step": 10756 + }, + { + "epoch": 0.56, + "learning_rate": 1.7000448992849223e-07, + "loss": 0.1413, + "step": 10757 + }, + { + "epoch": 0.56, + "learning_rate": 1.6997106131609188e-07, + "loss": 0.1332, + "step": 10758 + }, + { + "epoch": 0.56, + "learning_rate": 1.6993763356193017e-07, + "loss": 0.1179, + "step": 10759 + }, + { + "epoch": 0.56, + "learning_rate": 1.6990420666696244e-07, + "loss": 0.1245, + "step": 10760 + }, + { + "epoch": 0.56, + "learning_rate": 1.6987078063214415e-07, + "loss": 0.1413, + "step": 10761 + }, + { + "epoch": 0.56, + "learning_rate": 1.6983735545843051e-07, + "loss": 0.1014, + "step": 10762 + }, + { + "epoch": 0.56, + "learning_rate": 1.6980393114677698e-07, + "loss": 0.1371, + "step": 10763 + }, + { + "epoch": 0.56, + "learning_rate": 1.697705076981387e-07, + "loss": 0.1596, + "step": 10764 + }, + { + "epoch": 0.56, + "learning_rate": 1.69737085113471e-07, + "loss": 0.1398, + "step": 10765 + }, + { + "epoch": 0.56, + "learning_rate": 1.6970366339372907e-07, + "loss": 0.149, + "step": 10766 + }, + { + "epoch": 0.56, + "learning_rate": 1.6967024253986822e-07, + "loss": 0.1516, + "step": 10767 + }, + { + "epoch": 0.56, + "learning_rate": 1.696368225528435e-07, + "loss": 0.1014, + "step": 10768 + }, + { + "epoch": 0.56, + "learning_rate": 1.6960340343361006e-07, + "loss": 0.1402, + "step": 10769 + }, + { + "epoch": 0.56, + "learning_rate": 1.6956998518312312e-07, + "loss": 0.1449, + "step": 10770 + }, + { + "epoch": 0.56, + "learning_rate": 1.6953656780233773e-07, + "loss": 0.1756, + "step": 10771 + }, + { + "epoch": 0.56, + "learning_rate": 1.6950315129220903e-07, + "loss": 0.1199, + "step": 10772 + }, + { + "epoch": 0.56, + "learning_rate": 1.69469735653692e-07, + "loss": 0.1292, + "step": 10773 + }, + { + "epoch": 0.56, + "learning_rate": 1.6943632088774173e-07, + "loss": 0.1231, + "step": 10774 + }, + { + "epoch": 0.56, + "learning_rate": 1.6940290699531327e-07, + "loss": 0.1512, + "step": 10775 + }, + { + "epoch": 0.56, + "learning_rate": 1.6936949397736146e-07, + "loss": 0.1429, + "step": 10776 + }, + { + "epoch": 0.56, + "learning_rate": 1.6933608183484137e-07, + "loss": 0.1105, + "step": 10777 + }, + { + "epoch": 0.56, + "learning_rate": 1.6930267056870786e-07, + "loss": 0.1375, + "step": 10778 + }, + { + "epoch": 0.56, + "learning_rate": 1.692692601799159e-07, + "loss": 0.1518, + "step": 10779 + }, + { + "epoch": 0.56, + "learning_rate": 1.6923585066942034e-07, + "loss": 0.1257, + "step": 10780 + }, + { + "epoch": 0.56, + "learning_rate": 1.6920244203817606e-07, + "loss": 0.1329, + "step": 10781 + }, + { + "epoch": 0.56, + "learning_rate": 1.6916903428713786e-07, + "loss": 0.1397, + "step": 10782 + }, + { + "epoch": 0.56, + "learning_rate": 1.6913562741726056e-07, + "loss": 0.1127, + "step": 10783 + }, + { + "epoch": 0.56, + "learning_rate": 1.6910222142949902e-07, + "loss": 0.1377, + "step": 10784 + }, + { + "epoch": 0.56, + "learning_rate": 1.6906881632480782e-07, + "loss": 0.1626, + "step": 10785 + }, + { + "epoch": 0.56, + "learning_rate": 1.6903541210414182e-07, + "loss": 0.1414, + "step": 10786 + }, + { + "epoch": 0.56, + "learning_rate": 1.6900200876845568e-07, + "loss": 0.1192, + "step": 10787 + }, + { + "epoch": 0.56, + "learning_rate": 1.6896860631870412e-07, + "loss": 0.1354, + "step": 10788 + }, + { + "epoch": 0.56, + "learning_rate": 1.6893520475584175e-07, + "loss": 0.1535, + "step": 10789 + }, + { + "epoch": 0.56, + "learning_rate": 1.6890180408082326e-07, + "loss": 0.115, + "step": 10790 + }, + { + "epoch": 0.56, + "learning_rate": 1.6886840429460315e-07, + "loss": 0.1332, + "step": 10791 + }, + { + "epoch": 0.56, + "learning_rate": 1.6883500539813614e-07, + "loss": 0.1362, + "step": 10792 + }, + { + "epoch": 0.56, + "learning_rate": 1.688016073923767e-07, + "loss": 0.1517, + "step": 10793 + }, + { + "epoch": 0.56, + "learning_rate": 1.687682102782793e-07, + "loss": 0.1537, + "step": 10794 + }, + { + "epoch": 0.56, + "learning_rate": 1.6873481405679856e-07, + "loss": 0.148, + "step": 10795 + }, + { + "epoch": 0.56, + "learning_rate": 1.6870141872888885e-07, + "loss": 0.1337, + "step": 10796 + }, + { + "epoch": 0.56, + "learning_rate": 1.6866802429550472e-07, + "loss": 0.1662, + "step": 10797 + }, + { + "epoch": 0.56, + "learning_rate": 1.6863463075760054e-07, + "loss": 0.1164, + "step": 10798 + }, + { + "epoch": 0.56, + "learning_rate": 1.686012381161307e-07, + "loss": 0.1156, + "step": 10799 + }, + { + "epoch": 0.56, + "learning_rate": 1.6856784637204962e-07, + "loss": 0.2037, + "step": 10800 + }, + { + "epoch": 0.56, + "learning_rate": 1.6853445552631168e-07, + "loss": 0.1177, + "step": 10801 + }, + { + "epoch": 0.56, + "learning_rate": 1.6850106557987108e-07, + "loss": 0.1588, + "step": 10802 + }, + { + "epoch": 0.56, + "learning_rate": 1.684676765336822e-07, + "loss": 0.1454, + "step": 10803 + }, + { + "epoch": 0.56, + "learning_rate": 1.684342883886993e-07, + "loss": 0.1667, + "step": 10804 + }, + { + "epoch": 0.56, + "learning_rate": 1.6840090114587658e-07, + "loss": 0.1573, + "step": 10805 + }, + { + "epoch": 0.56, + "learning_rate": 1.6836751480616835e-07, + "loss": 0.1343, + "step": 10806 + }, + { + "epoch": 0.56, + "learning_rate": 1.6833412937052873e-07, + "loss": 0.1228, + "step": 10807 + }, + { + "epoch": 0.56, + "learning_rate": 1.6830074483991196e-07, + "loss": 0.1476, + "step": 10808 + }, + { + "epoch": 0.56, + "learning_rate": 1.6826736121527217e-07, + "loss": 0.1479, + "step": 10809 + }, + { + "epoch": 0.56, + "learning_rate": 1.6823397849756335e-07, + "loss": 0.1315, + "step": 10810 + }, + { + "epoch": 0.56, + "learning_rate": 1.6820059668773975e-07, + "loss": 0.1612, + "step": 10811 + }, + { + "epoch": 0.56, + "learning_rate": 1.6816721578675532e-07, + "loss": 0.1294, + "step": 10812 + }, + { + "epoch": 0.56, + "learning_rate": 1.681338357955642e-07, + "loss": 0.1535, + "step": 10813 + }, + { + "epoch": 0.56, + "learning_rate": 1.681004567151203e-07, + "loss": 0.1401, + "step": 10814 + }, + { + "epoch": 0.56, + "learning_rate": 1.680670785463777e-07, + "loss": 0.1749, + "step": 10815 + }, + { + "epoch": 0.56, + "learning_rate": 1.680337012902903e-07, + "loss": 0.1314, + "step": 10816 + }, + { + "epoch": 0.56, + "learning_rate": 1.680003249478121e-07, + "loss": 0.094, + "step": 10817 + }, + { + "epoch": 0.56, + "learning_rate": 1.67966949519897e-07, + "loss": 0.1407, + "step": 10818 + }, + { + "epoch": 0.56, + "learning_rate": 1.6793357500749878e-07, + "loss": 0.1108, + "step": 10819 + }, + { + "epoch": 0.56, + "learning_rate": 1.679002014115714e-07, + "loss": 0.1311, + "step": 10820 + }, + { + "epoch": 0.56, + "learning_rate": 1.678668287330686e-07, + "loss": 0.1267, + "step": 10821 + }, + { + "epoch": 0.56, + "learning_rate": 1.678334569729443e-07, + "loss": 0.1116, + "step": 10822 + }, + { + "epoch": 0.56, + "learning_rate": 1.6780008613215215e-07, + "loss": 0.1595, + "step": 10823 + }, + { + "epoch": 0.56, + "learning_rate": 1.6776671621164603e-07, + "loss": 0.1519, + "step": 10824 + }, + { + "epoch": 0.57, + "learning_rate": 1.6773334721237956e-07, + "loss": 0.1477, + "step": 10825 + }, + { + "epoch": 0.57, + "learning_rate": 1.676999791353065e-07, + "loss": 0.1422, + "step": 10826 + }, + { + "epoch": 0.57, + "learning_rate": 1.6766661198138059e-07, + "loss": 0.1558, + "step": 10827 + }, + { + "epoch": 0.57, + "learning_rate": 1.6763324575155528e-07, + "loss": 0.1296, + "step": 10828 + }, + { + "epoch": 0.57, + "learning_rate": 1.6759988044678434e-07, + "loss": 0.1325, + "step": 10829 + }, + { + "epoch": 0.57, + "learning_rate": 1.675665160680213e-07, + "loss": 0.1784, + "step": 10830 + }, + { + "epoch": 0.57, + "learning_rate": 1.675331526162198e-07, + "loss": 0.1274, + "step": 10831 + }, + { + "epoch": 0.57, + "learning_rate": 1.6749979009233325e-07, + "loss": 0.1198, + "step": 10832 + }, + { + "epoch": 0.57, + "learning_rate": 1.6746642849731532e-07, + "loss": 0.1413, + "step": 10833 + }, + { + "epoch": 0.57, + "learning_rate": 1.6743306783211936e-07, + "loss": 0.1212, + "step": 10834 + }, + { + "epoch": 0.57, + "learning_rate": 1.6739970809769898e-07, + "loss": 0.1411, + "step": 10835 + }, + { + "epoch": 0.57, + "learning_rate": 1.6736634929500753e-07, + "loss": 0.0999, + "step": 10836 + }, + { + "epoch": 0.57, + "learning_rate": 1.6733299142499833e-07, + "loss": 0.1281, + "step": 10837 + }, + { + "epoch": 0.57, + "learning_rate": 1.672996344886249e-07, + "loss": 0.1381, + "step": 10838 + }, + { + "epoch": 0.57, + "learning_rate": 1.6726627848684048e-07, + "loss": 0.1165, + "step": 10839 + }, + { + "epoch": 0.57, + "learning_rate": 1.672329234205985e-07, + "loss": 0.1433, + "step": 10840 + }, + { + "epoch": 0.57, + "learning_rate": 1.6719956929085223e-07, + "loss": 0.1316, + "step": 10841 + }, + { + "epoch": 0.57, + "learning_rate": 1.6716621609855492e-07, + "loss": 0.1231, + "step": 10842 + }, + { + "epoch": 0.57, + "learning_rate": 1.671328638446598e-07, + "loss": 0.1718, + "step": 10843 + }, + { + "epoch": 0.57, + "learning_rate": 1.670995125301202e-07, + "loss": 0.1426, + "step": 10844 + }, + { + "epoch": 0.57, + "learning_rate": 1.670661621558892e-07, + "loss": 0.1285, + "step": 10845 + }, + { + "epoch": 0.57, + "learning_rate": 1.6703281272291995e-07, + "loss": 0.1466, + "step": 10846 + }, + { + "epoch": 0.57, + "learning_rate": 1.6699946423216568e-07, + "loss": 0.163, + "step": 10847 + }, + { + "epoch": 0.57, + "learning_rate": 1.6696611668457942e-07, + "loss": 0.1851, + "step": 10848 + }, + { + "epoch": 0.57, + "learning_rate": 1.6693277008111435e-07, + "loss": 0.1312, + "step": 10849 + }, + { + "epoch": 0.57, + "learning_rate": 1.6689942442272343e-07, + "loss": 0.1381, + "step": 10850 + }, + { + "epoch": 0.57, + "learning_rate": 1.6686607971035976e-07, + "loss": 0.137, + "step": 10851 + }, + { + "epoch": 0.57, + "learning_rate": 1.6683273594497639e-07, + "loss": 0.1337, + "step": 10852 + }, + { + "epoch": 0.57, + "learning_rate": 1.6679939312752614e-07, + "loss": 0.0943, + "step": 10853 + }, + { + "epoch": 0.57, + "learning_rate": 1.667660512589621e-07, + "loss": 0.1364, + "step": 10854 + }, + { + "epoch": 0.57, + "learning_rate": 1.667327103402371e-07, + "loss": 0.1188, + "step": 10855 + }, + { + "epoch": 0.57, + "learning_rate": 1.6669937037230412e-07, + "loss": 0.1754, + "step": 10856 + }, + { + "epoch": 0.57, + "learning_rate": 1.6666603135611592e-07, + "loss": 0.1197, + "step": 10857 + }, + { + "epoch": 0.57, + "learning_rate": 1.6663269329262548e-07, + "loss": 0.1612, + "step": 10858 + }, + { + "epoch": 0.57, + "learning_rate": 1.6659935618278552e-07, + "loss": 0.1286, + "step": 10859 + }, + { + "epoch": 0.57, + "learning_rate": 1.6656602002754888e-07, + "loss": 0.1345, + "step": 10860 + }, + { + "epoch": 0.57, + "learning_rate": 1.6653268482786834e-07, + "loss": 0.1049, + "step": 10861 + }, + { + "epoch": 0.57, + "learning_rate": 1.6649935058469653e-07, + "loss": 0.1477, + "step": 10862 + }, + { + "epoch": 0.57, + "learning_rate": 1.664660172989862e-07, + "loss": 0.1712, + "step": 10863 + }, + { + "epoch": 0.57, + "learning_rate": 1.6643268497169006e-07, + "loss": 0.1517, + "step": 10864 + }, + { + "epoch": 0.57, + "learning_rate": 1.663993536037607e-07, + "loss": 0.0973, + "step": 10865 + }, + { + "epoch": 0.57, + "learning_rate": 1.6636602319615083e-07, + "loss": 0.1188, + "step": 10866 + }, + { + "epoch": 0.57, + "learning_rate": 1.6633269374981295e-07, + "loss": 0.1238, + "step": 10867 + }, + { + "epoch": 0.57, + "learning_rate": 1.662993652656997e-07, + "loss": 0.142, + "step": 10868 + }, + { + "epoch": 0.57, + "learning_rate": 1.6626603774476374e-07, + "loss": 0.1333, + "step": 10869 + }, + { + "epoch": 0.57, + "learning_rate": 1.662327111879573e-07, + "loss": 0.1145, + "step": 10870 + }, + { + "epoch": 0.57, + "learning_rate": 1.6619938559623304e-07, + "loss": 0.1564, + "step": 10871 + }, + { + "epoch": 0.57, + "learning_rate": 1.6616606097054333e-07, + "loss": 0.1225, + "step": 10872 + }, + { + "epoch": 0.57, + "learning_rate": 1.6613273731184073e-07, + "loss": 0.1723, + "step": 10873 + }, + { + "epoch": 0.57, + "learning_rate": 1.6609941462107752e-07, + "loss": 0.1315, + "step": 10874 + }, + { + "epoch": 0.57, + "learning_rate": 1.660660928992062e-07, + "loss": 0.1356, + "step": 10875 + }, + { + "epoch": 0.57, + "learning_rate": 1.6603277214717894e-07, + "loss": 0.1579, + "step": 10876 + }, + { + "epoch": 0.57, + "learning_rate": 1.6599945236594825e-07, + "loss": 0.146, + "step": 10877 + }, + { + "epoch": 0.57, + "learning_rate": 1.659661335564664e-07, + "loss": 0.143, + "step": 10878 + }, + { + "epoch": 0.57, + "learning_rate": 1.6593281571968546e-07, + "loss": 0.1438, + "step": 10879 + }, + { + "epoch": 0.57, + "learning_rate": 1.6589949885655786e-07, + "loss": 0.1544, + "step": 10880 + }, + { + "epoch": 0.57, + "learning_rate": 1.658661829680357e-07, + "loss": 0.14, + "step": 10881 + }, + { + "epoch": 0.57, + "learning_rate": 1.6583286805507127e-07, + "loss": 0.1222, + "step": 10882 + }, + { + "epoch": 0.57, + "learning_rate": 1.657995541186166e-07, + "loss": 0.1304, + "step": 10883 + }, + { + "epoch": 0.57, + "learning_rate": 1.657662411596239e-07, + "loss": 0.1071, + "step": 10884 + }, + { + "epoch": 0.57, + "learning_rate": 1.6573292917904522e-07, + "loss": 0.1202, + "step": 10885 + }, + { + "epoch": 0.57, + "learning_rate": 1.6569961817783269e-07, + "loss": 0.1395, + "step": 10886 + }, + { + "epoch": 0.57, + "learning_rate": 1.6566630815693837e-07, + "loss": 0.1111, + "step": 10887 + }, + { + "epoch": 0.57, + "learning_rate": 1.6563299911731412e-07, + "loss": 0.1386, + "step": 10888 + }, + { + "epoch": 0.57, + "learning_rate": 1.6559969105991206e-07, + "loss": 0.1326, + "step": 10889 + }, + { + "epoch": 0.57, + "learning_rate": 1.655663839856841e-07, + "loss": 0.1396, + "step": 10890 + }, + { + "epoch": 0.57, + "learning_rate": 1.655330778955822e-07, + "loss": 0.1384, + "step": 10891 + }, + { + "epoch": 0.57, + "learning_rate": 1.6549977279055818e-07, + "loss": 0.134, + "step": 10892 + }, + { + "epoch": 0.57, + "learning_rate": 1.6546646867156402e-07, + "loss": 0.1297, + "step": 10893 + }, + { + "epoch": 0.57, + "learning_rate": 1.6543316553955148e-07, + "loss": 0.1452, + "step": 10894 + }, + { + "epoch": 0.57, + "learning_rate": 1.6539986339547253e-07, + "loss": 0.1618, + "step": 10895 + }, + { + "epoch": 0.57, + "learning_rate": 1.6536656224027876e-07, + "loss": 0.1525, + "step": 10896 + }, + { + "epoch": 0.57, + "learning_rate": 1.6533326207492198e-07, + "loss": 0.166, + "step": 10897 + }, + { + "epoch": 0.57, + "learning_rate": 1.6529996290035396e-07, + "loss": 0.1391, + "step": 10898 + }, + { + "epoch": 0.57, + "learning_rate": 1.652666647175264e-07, + "loss": 0.1412, + "step": 10899 + }, + { + "epoch": 0.57, + "learning_rate": 1.65233367527391e-07, + "loss": 0.1242, + "step": 10900 + }, + { + "epoch": 0.57, + "learning_rate": 1.6520007133089931e-07, + "loss": 0.1687, + "step": 10901 + }, + { + "epoch": 0.57, + "learning_rate": 1.651667761290031e-07, + "loss": 0.1293, + "step": 10902 + }, + { + "epoch": 0.57, + "learning_rate": 1.6513348192265385e-07, + "loss": 0.1211, + "step": 10903 + }, + { + "epoch": 0.57, + "learning_rate": 1.6510018871280317e-07, + "loss": 0.1142, + "step": 10904 + }, + { + "epoch": 0.57, + "learning_rate": 1.6506689650040255e-07, + "loss": 0.1531, + "step": 10905 + }, + { + "epoch": 0.57, + "learning_rate": 1.6503360528640346e-07, + "loss": 0.1387, + "step": 10906 + }, + { + "epoch": 0.57, + "learning_rate": 1.6500031507175752e-07, + "loss": 0.1728, + "step": 10907 + }, + { + "epoch": 0.57, + "learning_rate": 1.64967025857416e-07, + "loss": 0.1149, + "step": 10908 + }, + { + "epoch": 0.57, + "learning_rate": 1.6493373764433046e-07, + "loss": 0.1221, + "step": 10909 + }, + { + "epoch": 0.57, + "learning_rate": 1.649004504334522e-07, + "loss": 0.1185, + "step": 10910 + }, + { + "epoch": 0.57, + "learning_rate": 1.6486716422573264e-07, + "loss": 0.1579, + "step": 10911 + }, + { + "epoch": 0.57, + "learning_rate": 1.6483387902212314e-07, + "loss": 0.1223, + "step": 10912 + }, + { + "epoch": 0.57, + "learning_rate": 1.6480059482357488e-07, + "loss": 0.127, + "step": 10913 + }, + { + "epoch": 0.57, + "learning_rate": 1.647673116310392e-07, + "loss": 0.1167, + "step": 10914 + }, + { + "epoch": 0.57, + "learning_rate": 1.6473402944546738e-07, + "loss": 0.1534, + "step": 10915 + }, + { + "epoch": 0.57, + "learning_rate": 1.6470074826781061e-07, + "loss": 0.1119, + "step": 10916 + }, + { + "epoch": 0.57, + "learning_rate": 1.6466746809902006e-07, + "loss": 0.1377, + "step": 10917 + }, + { + "epoch": 0.57, + "learning_rate": 1.6463418894004694e-07, + "loss": 0.099, + "step": 10918 + }, + { + "epoch": 0.57, + "learning_rate": 1.6460091079184234e-07, + "loss": 0.1276, + "step": 10919 + }, + { + "epoch": 0.57, + "learning_rate": 1.6456763365535737e-07, + "loss": 0.1079, + "step": 10920 + }, + { + "epoch": 0.57, + "learning_rate": 1.645343575315432e-07, + "loss": 0.1557, + "step": 10921 + }, + { + "epoch": 0.57, + "learning_rate": 1.6450108242135067e-07, + "loss": 0.1364, + "step": 10922 + }, + { + "epoch": 0.57, + "learning_rate": 1.6446780832573095e-07, + "loss": 0.1455, + "step": 10923 + }, + { + "epoch": 0.57, + "learning_rate": 1.6443453524563496e-07, + "loss": 0.1304, + "step": 10924 + }, + { + "epoch": 0.57, + "learning_rate": 1.6440126318201371e-07, + "loss": 0.1389, + "step": 10925 + }, + { + "epoch": 0.57, + "learning_rate": 1.6436799213581806e-07, + "loss": 0.0982, + "step": 10926 + }, + { + "epoch": 0.57, + "learning_rate": 1.64334722107999e-07, + "loss": 0.1641, + "step": 10927 + }, + { + "epoch": 0.57, + "learning_rate": 1.643014530995073e-07, + "loss": 0.1416, + "step": 10928 + }, + { + "epoch": 0.57, + "learning_rate": 1.6426818511129394e-07, + "loss": 0.1646, + "step": 10929 + }, + { + "epoch": 0.57, + "learning_rate": 1.642349181443096e-07, + "loss": 0.1656, + "step": 10930 + }, + { + "epoch": 0.57, + "learning_rate": 1.6420165219950507e-07, + "loss": 0.1787, + "step": 10931 + }, + { + "epoch": 0.57, + "learning_rate": 1.6416838727783117e-07, + "loss": 0.1308, + "step": 10932 + }, + { + "epoch": 0.57, + "learning_rate": 1.6413512338023857e-07, + "loss": 0.1518, + "step": 10933 + }, + { + "epoch": 0.57, + "learning_rate": 1.64101860507678e-07, + "loss": 0.1834, + "step": 10934 + }, + { + "epoch": 0.57, + "learning_rate": 1.640685986611001e-07, + "loss": 0.1716, + "step": 10935 + }, + { + "epoch": 0.57, + "learning_rate": 1.6403533784145553e-07, + "loss": 0.1218, + "step": 10936 + }, + { + "epoch": 0.57, + "learning_rate": 1.6400207804969487e-07, + "loss": 0.1533, + "step": 10937 + }, + { + "epoch": 0.57, + "learning_rate": 1.639688192867688e-07, + "loss": 0.1386, + "step": 10938 + }, + { + "epoch": 0.57, + "learning_rate": 1.6393556155362776e-07, + "loss": 0.1417, + "step": 10939 + }, + { + "epoch": 0.57, + "learning_rate": 1.6390230485122224e-07, + "loss": 0.1356, + "step": 10940 + }, + { + "epoch": 0.57, + "learning_rate": 1.6386904918050282e-07, + "loss": 0.1269, + "step": 10941 + }, + { + "epoch": 0.57, + "learning_rate": 1.6383579454241987e-07, + "loss": 0.124, + "step": 10942 + }, + { + "epoch": 0.57, + "learning_rate": 1.6380254093792392e-07, + "loss": 0.161, + "step": 10943 + }, + { + "epoch": 0.57, + "learning_rate": 1.6376928836796528e-07, + "loss": 0.1174, + "step": 10944 + }, + { + "epoch": 0.57, + "learning_rate": 1.6373603683349443e-07, + "loss": 0.1252, + "step": 10945 + }, + { + "epoch": 0.57, + "learning_rate": 1.6370278633546156e-07, + "loss": 0.1309, + "step": 10946 + }, + { + "epoch": 0.57, + "learning_rate": 1.6366953687481718e-07, + "loss": 0.1382, + "step": 10947 + }, + { + "epoch": 0.57, + "learning_rate": 1.6363628845251143e-07, + "loss": 0.1223, + "step": 10948 + }, + { + "epoch": 0.57, + "learning_rate": 1.6360304106949453e-07, + "loss": 0.1294, + "step": 10949 + }, + { + "epoch": 0.57, + "learning_rate": 1.6356979472671684e-07, + "loss": 0.1142, + "step": 10950 + }, + { + "epoch": 0.57, + "learning_rate": 1.635365494251284e-07, + "loss": 0.152, + "step": 10951 + }, + { + "epoch": 0.57, + "learning_rate": 1.635033051656795e-07, + "loss": 0.1654, + "step": 10952 + }, + { + "epoch": 0.57, + "learning_rate": 1.6347006194932023e-07, + "loss": 0.1669, + "step": 10953 + }, + { + "epoch": 0.57, + "learning_rate": 1.634368197770007e-07, + "loss": 0.1473, + "step": 10954 + }, + { + "epoch": 0.57, + "learning_rate": 1.6340357864967102e-07, + "loss": 0.149, + "step": 10955 + }, + { + "epoch": 0.57, + "learning_rate": 1.633703385682811e-07, + "loss": 0.1223, + "step": 10956 + }, + { + "epoch": 0.57, + "learning_rate": 1.633370995337811e-07, + "loss": 0.1496, + "step": 10957 + }, + { + "epoch": 0.57, + "learning_rate": 1.633038615471209e-07, + "loss": 0.1661, + "step": 10958 + }, + { + "epoch": 0.57, + "learning_rate": 1.6327062460925054e-07, + "loss": 0.154, + "step": 10959 + }, + { + "epoch": 0.57, + "learning_rate": 1.6323738872111986e-07, + "loss": 0.115, + "step": 10960 + }, + { + "epoch": 0.57, + "learning_rate": 1.6320415388367886e-07, + "loss": 0.1205, + "step": 10961 + }, + { + "epoch": 0.57, + "learning_rate": 1.631709200978773e-07, + "loss": 0.1079, + "step": 10962 + }, + { + "epoch": 0.57, + "learning_rate": 1.631376873646651e-07, + "loss": 0.1469, + "step": 10963 + }, + { + "epoch": 0.57, + "learning_rate": 1.6310445568499204e-07, + "loss": 0.1609, + "step": 10964 + }, + { + "epoch": 0.57, + "learning_rate": 1.6307122505980785e-07, + "loss": 0.1228, + "step": 10965 + }, + { + "epoch": 0.57, + "learning_rate": 1.630379954900623e-07, + "loss": 0.1325, + "step": 10966 + }, + { + "epoch": 0.57, + "learning_rate": 1.6300476697670508e-07, + "loss": 0.1882, + "step": 10967 + }, + { + "epoch": 0.57, + "learning_rate": 1.6297153952068596e-07, + "loss": 0.1258, + "step": 10968 + }, + { + "epoch": 0.57, + "learning_rate": 1.6293831312295448e-07, + "loss": 0.1854, + "step": 10969 + }, + { + "epoch": 0.57, + "learning_rate": 1.6290508778446036e-07, + "loss": 0.1594, + "step": 10970 + }, + { + "epoch": 0.57, + "learning_rate": 1.6287186350615312e-07, + "loss": 0.1701, + "step": 10971 + }, + { + "epoch": 0.57, + "learning_rate": 1.6283864028898248e-07, + "loss": 0.1458, + "step": 10972 + }, + { + "epoch": 0.57, + "learning_rate": 1.6280541813389769e-07, + "loss": 0.1244, + "step": 10973 + }, + { + "epoch": 0.57, + "learning_rate": 1.6277219704184848e-07, + "loss": 0.1148, + "step": 10974 + }, + { + "epoch": 0.57, + "learning_rate": 1.627389770137842e-07, + "loss": 0.1604, + "step": 10975 + }, + { + "epoch": 0.57, + "learning_rate": 1.627057580506544e-07, + "loss": 0.1211, + "step": 10976 + }, + { + "epoch": 0.57, + "learning_rate": 1.6267254015340838e-07, + "loss": 0.1398, + "step": 10977 + }, + { + "epoch": 0.57, + "learning_rate": 1.626393233229956e-07, + "loss": 0.1401, + "step": 10978 + }, + { + "epoch": 0.57, + "learning_rate": 1.6260610756036532e-07, + "loss": 0.1247, + "step": 10979 + }, + { + "epoch": 0.57, + "learning_rate": 1.62572892866467e-07, + "loss": 0.1748, + "step": 10980 + }, + { + "epoch": 0.57, + "learning_rate": 1.6253967924224989e-07, + "loss": 0.1317, + "step": 10981 + }, + { + "epoch": 0.57, + "learning_rate": 1.6250646668866313e-07, + "loss": 0.1221, + "step": 10982 + }, + { + "epoch": 0.57, + "learning_rate": 1.6247325520665603e-07, + "loss": 0.1462, + "step": 10983 + }, + { + "epoch": 0.57, + "learning_rate": 1.6244004479717777e-07, + "loss": 0.119, + "step": 10984 + }, + { + "epoch": 0.57, + "learning_rate": 1.6240683546117756e-07, + "loss": 0.1625, + "step": 10985 + }, + { + "epoch": 0.57, + "learning_rate": 1.6237362719960447e-07, + "loss": 0.1066, + "step": 10986 + }, + { + "epoch": 0.57, + "learning_rate": 1.6234042001340768e-07, + "loss": 0.1655, + "step": 10987 + }, + { + "epoch": 0.57, + "learning_rate": 1.623072139035362e-07, + "loss": 0.1166, + "step": 10988 + }, + { + "epoch": 0.57, + "learning_rate": 1.6227400887093917e-07, + "loss": 0.1586, + "step": 10989 + }, + { + "epoch": 0.57, + "learning_rate": 1.6224080491656547e-07, + "loss": 0.1422, + "step": 10990 + }, + { + "epoch": 0.57, + "learning_rate": 1.6220760204136412e-07, + "loss": 0.1153, + "step": 10991 + }, + { + "epoch": 0.57, + "learning_rate": 1.6217440024628412e-07, + "loss": 0.1406, + "step": 10992 + }, + { + "epoch": 0.57, + "learning_rate": 1.6214119953227434e-07, + "loss": 0.1015, + "step": 10993 + }, + { + "epoch": 0.57, + "learning_rate": 1.6210799990028372e-07, + "loss": 0.142, + "step": 10994 + }, + { + "epoch": 0.57, + "learning_rate": 1.6207480135126104e-07, + "loss": 0.1374, + "step": 10995 + }, + { + "epoch": 0.57, + "learning_rate": 1.6204160388615523e-07, + "loss": 0.168, + "step": 10996 + }, + { + "epoch": 0.57, + "learning_rate": 1.62008407505915e-07, + "loss": 0.1287, + "step": 10997 + }, + { + "epoch": 0.57, + "learning_rate": 1.6197521221148925e-07, + "loss": 0.1256, + "step": 10998 + }, + { + "epoch": 0.57, + "learning_rate": 1.6194201800382655e-07, + "loss": 0.1113, + "step": 10999 + }, + { + "epoch": 0.57, + "learning_rate": 1.619088248838756e-07, + "loss": 0.1502, + "step": 11000 + }, + { + "epoch": 0.57, + "learning_rate": 1.618756328525852e-07, + "loss": 0.1532, + "step": 11001 + }, + { + "epoch": 0.57, + "learning_rate": 1.618424419109039e-07, + "loss": 0.1125, + "step": 11002 + }, + { + "epoch": 0.57, + "learning_rate": 1.6180925205978039e-07, + "loss": 0.1647, + "step": 11003 + }, + { + "epoch": 0.57, + "learning_rate": 1.6177606330016317e-07, + "loss": 0.1314, + "step": 11004 + }, + { + "epoch": 0.57, + "learning_rate": 1.6174287563300082e-07, + "loss": 0.1221, + "step": 11005 + }, + { + "epoch": 0.57, + "learning_rate": 1.617096890592419e-07, + "loss": 0.1389, + "step": 11006 + }, + { + "epoch": 0.57, + "learning_rate": 1.6167650357983479e-07, + "loss": 0.128, + "step": 11007 + }, + { + "epoch": 0.57, + "learning_rate": 1.61643319195728e-07, + "loss": 0.1537, + "step": 11008 + }, + { + "epoch": 0.57, + "learning_rate": 1.6161013590786997e-07, + "loss": 0.176, + "step": 11009 + }, + { + "epoch": 0.57, + "learning_rate": 1.615769537172091e-07, + "loss": 0.1201, + "step": 11010 + }, + { + "epoch": 0.57, + "learning_rate": 1.6154377262469365e-07, + "loss": 0.128, + "step": 11011 + }, + { + "epoch": 0.57, + "learning_rate": 1.6151059263127208e-07, + "loss": 0.1564, + "step": 11012 + }, + { + "epoch": 0.57, + "learning_rate": 1.6147741373789262e-07, + "loss": 0.1149, + "step": 11013 + }, + { + "epoch": 0.57, + "learning_rate": 1.6144423594550355e-07, + "loss": 0.1281, + "step": 11014 + }, + { + "epoch": 0.57, + "learning_rate": 1.6141105925505315e-07, + "loss": 0.1391, + "step": 11015 + }, + { + "epoch": 0.57, + "learning_rate": 1.613778836674895e-07, + "loss": 0.155, + "step": 11016 + }, + { + "epoch": 0.58, + "learning_rate": 1.6134470918376088e-07, + "loss": 0.1336, + "step": 11017 + }, + { + "epoch": 0.58, + "learning_rate": 1.6131153580481536e-07, + "loss": 0.1089, + "step": 11018 + }, + { + "epoch": 0.58, + "learning_rate": 1.6127836353160112e-07, + "loss": 0.1325, + "step": 11019 + }, + { + "epoch": 0.58, + "learning_rate": 1.6124519236506614e-07, + "loss": 0.1429, + "step": 11020 + }, + { + "epoch": 0.58, + "learning_rate": 1.612120223061586e-07, + "loss": 0.1432, + "step": 11021 + }, + { + "epoch": 0.58, + "learning_rate": 1.6117885335582638e-07, + "loss": 0.1191, + "step": 11022 + }, + { + "epoch": 0.58, + "learning_rate": 1.6114568551501755e-07, + "loss": 0.1484, + "step": 11023 + }, + { + "epoch": 0.58, + "learning_rate": 1.6111251878468007e-07, + "loss": 0.1397, + "step": 11024 + }, + { + "epoch": 0.58, + "learning_rate": 1.6107935316576175e-07, + "loss": 0.1167, + "step": 11025 + }, + { + "epoch": 0.58, + "learning_rate": 1.6104618865921056e-07, + "loss": 0.127, + "step": 11026 + }, + { + "epoch": 0.58, + "learning_rate": 1.610130252659743e-07, + "loss": 0.149, + "step": 11027 + }, + { + "epoch": 0.58, + "learning_rate": 1.6097986298700086e-07, + "loss": 0.1394, + "step": 11028 + }, + { + "epoch": 0.58, + "learning_rate": 1.6094670182323798e-07, + "loss": 0.1328, + "step": 11029 + }, + { + "epoch": 0.58, + "learning_rate": 1.6091354177563348e-07, + "loss": 0.1324, + "step": 11030 + }, + { + "epoch": 0.58, + "learning_rate": 1.60880382845135e-07, + "loss": 0.1258, + "step": 11031 + }, + { + "epoch": 0.58, + "learning_rate": 1.6084722503269034e-07, + "loss": 0.1431, + "step": 11032 + }, + { + "epoch": 0.58, + "learning_rate": 1.6081406833924705e-07, + "loss": 0.1476, + "step": 11033 + }, + { + "epoch": 0.58, + "learning_rate": 1.6078091276575282e-07, + "loss": 0.165, + "step": 11034 + }, + { + "epoch": 0.58, + "learning_rate": 1.6074775831315528e-07, + "loss": 0.1435, + "step": 11035 + }, + { + "epoch": 0.58, + "learning_rate": 1.607146049824019e-07, + "loss": 0.1476, + "step": 11036 + }, + { + "epoch": 0.58, + "learning_rate": 1.6068145277444032e-07, + "loss": 0.1354, + "step": 11037 + }, + { + "epoch": 0.58, + "learning_rate": 1.6064830169021796e-07, + "loss": 0.133, + "step": 11038 + }, + { + "epoch": 0.58, + "learning_rate": 1.6061515173068236e-07, + "loss": 0.1254, + "step": 11039 + }, + { + "epoch": 0.58, + "learning_rate": 1.605820028967809e-07, + "loss": 0.1387, + "step": 11040 + }, + { + "epoch": 0.58, + "learning_rate": 1.605488551894611e-07, + "loss": 0.1299, + "step": 11041 + }, + { + "epoch": 0.58, + "learning_rate": 1.6051570860967018e-07, + "loss": 0.1252, + "step": 11042 + }, + { + "epoch": 0.58, + "learning_rate": 1.6048256315835553e-07, + "loss": 0.1226, + "step": 11043 + }, + { + "epoch": 0.58, + "learning_rate": 1.6044941883646453e-07, + "loss": 0.1335, + "step": 11044 + }, + { + "epoch": 0.58, + "learning_rate": 1.6041627564494435e-07, + "loss": 0.1358, + "step": 11045 + }, + { + "epoch": 0.58, + "learning_rate": 1.6038313358474233e-07, + "loss": 0.1121, + "step": 11046 + }, + { + "epoch": 0.58, + "learning_rate": 1.6034999265680563e-07, + "loss": 0.1187, + "step": 11047 + }, + { + "epoch": 0.58, + "learning_rate": 1.6031685286208145e-07, + "loss": 0.158, + "step": 11048 + }, + { + "epoch": 0.58, + "learning_rate": 1.60283714201517e-07, + "loss": 0.1355, + "step": 11049 + }, + { + "epoch": 0.58, + "learning_rate": 1.6025057667605926e-07, + "loss": 0.146, + "step": 11050 + }, + { + "epoch": 0.58, + "learning_rate": 1.602174402866554e-07, + "loss": 0.1549, + "step": 11051 + }, + { + "epoch": 0.58, + "learning_rate": 1.6018430503425242e-07, + "loss": 0.137, + "step": 11052 + }, + { + "epoch": 0.58, + "learning_rate": 1.6015117091979742e-07, + "loss": 0.1033, + "step": 11053 + }, + { + "epoch": 0.58, + "learning_rate": 1.601180379442373e-07, + "loss": 0.151, + "step": 11054 + }, + { + "epoch": 0.58, + "learning_rate": 1.6008490610851908e-07, + "loss": 0.1621, + "step": 11055 + }, + { + "epoch": 0.58, + "learning_rate": 1.600517754135896e-07, + "loss": 0.1216, + "step": 11056 + }, + { + "epoch": 0.58, + "learning_rate": 1.6001864586039585e-07, + "loss": 0.1161, + "step": 11057 + }, + { + "epoch": 0.58, + "learning_rate": 1.599855174498847e-07, + "loss": 0.1411, + "step": 11058 + }, + { + "epoch": 0.58, + "learning_rate": 1.599523901830028e-07, + "loss": 0.108, + "step": 11059 + }, + { + "epoch": 0.58, + "learning_rate": 1.599192640606971e-07, + "loss": 0.1204, + "step": 11060 + }, + { + "epoch": 0.58, + "learning_rate": 1.598861390839143e-07, + "loss": 0.1732, + "step": 11061 + }, + { + "epoch": 0.58, + "learning_rate": 1.598530152536011e-07, + "loss": 0.131, + "step": 11062 + }, + { + "epoch": 0.58, + "learning_rate": 1.5981989257070424e-07, + "loss": 0.2315, + "step": 11063 + }, + { + "epoch": 0.58, + "learning_rate": 1.597867710361704e-07, + "loss": 0.1547, + "step": 11064 + }, + { + "epoch": 0.58, + "learning_rate": 1.5975365065094608e-07, + "loss": 0.12, + "step": 11065 + }, + { + "epoch": 0.58, + "learning_rate": 1.597205314159781e-07, + "loss": 0.1423, + "step": 11066 + }, + { + "epoch": 0.58, + "learning_rate": 1.596874133322128e-07, + "loss": 0.1422, + "step": 11067 + }, + { + "epoch": 0.58, + "learning_rate": 1.5965429640059676e-07, + "loss": 0.1702, + "step": 11068 + }, + { + "epoch": 0.58, + "learning_rate": 1.5962118062207655e-07, + "loss": 0.129, + "step": 11069 + }, + { + "epoch": 0.58, + "learning_rate": 1.5958806599759852e-07, + "loss": 0.1616, + "step": 11070 + }, + { + "epoch": 0.58, + "learning_rate": 1.5955495252810923e-07, + "loss": 0.1274, + "step": 11071 + }, + { + "epoch": 0.58, + "learning_rate": 1.5952184021455497e-07, + "loss": 0.1183, + "step": 11072 + }, + { + "epoch": 0.58, + "learning_rate": 1.5948872905788218e-07, + "loss": 0.183, + "step": 11073 + }, + { + "epoch": 0.58, + "learning_rate": 1.5945561905903707e-07, + "loss": 0.1825, + "step": 11074 + }, + { + "epoch": 0.58, + "learning_rate": 1.5942251021896618e-07, + "loss": 0.1594, + "step": 11075 + }, + { + "epoch": 0.58, + "learning_rate": 1.593894025386155e-07, + "loss": 0.1481, + "step": 11076 + }, + { + "epoch": 0.58, + "learning_rate": 1.5935629601893137e-07, + "loss": 0.1959, + "step": 11077 + }, + { + "epoch": 0.58, + "learning_rate": 1.5932319066086003e-07, + "loss": 0.1033, + "step": 11078 + }, + { + "epoch": 0.58, + "learning_rate": 1.5929008646534755e-07, + "loss": 0.1656, + "step": 11079 + }, + { + "epoch": 0.58, + "learning_rate": 1.5925698343334018e-07, + "loss": 0.0934, + "step": 11080 + }, + { + "epoch": 0.58, + "learning_rate": 1.5922388156578388e-07, + "loss": 0.1066, + "step": 11081 + }, + { + "epoch": 0.58, + "learning_rate": 1.5919078086362485e-07, + "loss": 0.1419, + "step": 11082 + }, + { + "epoch": 0.58, + "learning_rate": 1.59157681327809e-07, + "loss": 0.138, + "step": 11083 + }, + { + "epoch": 0.58, + "learning_rate": 1.591245829592825e-07, + "loss": 0.1405, + "step": 11084 + }, + { + "epoch": 0.58, + "learning_rate": 1.590914857589911e-07, + "loss": 0.1361, + "step": 11085 + }, + { + "epoch": 0.58, + "learning_rate": 1.5905838972788085e-07, + "loss": 0.1552, + "step": 11086 + }, + { + "epoch": 0.58, + "learning_rate": 1.590252948668976e-07, + "loss": 0.1174, + "step": 11087 + }, + { + "epoch": 0.58, + "learning_rate": 1.5899220117698728e-07, + "loss": 0.1208, + "step": 11088 + }, + { + "epoch": 0.58, + "learning_rate": 1.5895910865909564e-07, + "loss": 0.1326, + "step": 11089 + }, + { + "epoch": 0.58, + "learning_rate": 1.5892601731416856e-07, + "loss": 0.1598, + "step": 11090 + }, + { + "epoch": 0.58, + "learning_rate": 1.588929271431517e-07, + "loss": 0.1569, + "step": 11091 + }, + { + "epoch": 0.58, + "learning_rate": 1.5885983814699095e-07, + "loss": 0.1522, + "step": 11092 + }, + { + "epoch": 0.58, + "learning_rate": 1.5882675032663186e-07, + "loss": 0.141, + "step": 11093 + }, + { + "epoch": 0.58, + "learning_rate": 1.5879366368302009e-07, + "loss": 0.1634, + "step": 11094 + }, + { + "epoch": 0.58, + "learning_rate": 1.587605782171014e-07, + "loss": 0.1324, + "step": 11095 + }, + { + "epoch": 0.58, + "learning_rate": 1.5872749392982124e-07, + "loss": 0.1439, + "step": 11096 + }, + { + "epoch": 0.58, + "learning_rate": 1.5869441082212526e-07, + "loss": 0.1298, + "step": 11097 + }, + { + "epoch": 0.58, + "learning_rate": 1.5866132889495894e-07, + "loss": 0.1362, + "step": 11098 + }, + { + "epoch": 0.58, + "learning_rate": 1.5862824814926783e-07, + "loss": 0.2142, + "step": 11099 + }, + { + "epoch": 0.58, + "learning_rate": 1.5859516858599732e-07, + "loss": 0.1018, + "step": 11100 + }, + { + "epoch": 0.58, + "learning_rate": 1.5856209020609299e-07, + "loss": 0.1246, + "step": 11101 + }, + { + "epoch": 0.58, + "learning_rate": 1.5852901301050003e-07, + "loss": 0.1523, + "step": 11102 + }, + { + "epoch": 0.58, + "learning_rate": 1.5849593700016388e-07, + "loss": 0.1167, + "step": 11103 + }, + { + "epoch": 0.58, + "learning_rate": 1.5846286217602992e-07, + "loss": 0.1103, + "step": 11104 + }, + { + "epoch": 0.58, + "learning_rate": 1.5842978853904332e-07, + "loss": 0.136, + "step": 11105 + }, + { + "epoch": 0.58, + "learning_rate": 1.5839671609014947e-07, + "loss": 0.1463, + "step": 11106 + }, + { + "epoch": 0.58, + "learning_rate": 1.5836364483029347e-07, + "loss": 0.1274, + "step": 11107 + }, + { + "epoch": 0.58, + "learning_rate": 1.5833057476042063e-07, + "loss": 0.1086, + "step": 11108 + }, + { + "epoch": 0.58, + "learning_rate": 1.582975058814761e-07, + "loss": 0.1528, + "step": 11109 + }, + { + "epoch": 0.58, + "learning_rate": 1.5826443819440488e-07, + "loss": 0.1181, + "step": 11110 + }, + { + "epoch": 0.58, + "learning_rate": 1.5823137170015215e-07, + "loss": 0.1089, + "step": 11111 + }, + { + "epoch": 0.58, + "learning_rate": 1.5819830639966288e-07, + "loss": 0.1348, + "step": 11112 + }, + { + "epoch": 0.58, + "learning_rate": 1.5816524229388223e-07, + "loss": 0.1265, + "step": 11113 + }, + { + "epoch": 0.58, + "learning_rate": 1.5813217938375504e-07, + "loss": 0.1455, + "step": 11114 + }, + { + "epoch": 0.58, + "learning_rate": 1.5809911767022638e-07, + "loss": 0.1286, + "step": 11115 + }, + { + "epoch": 0.58, + "learning_rate": 1.5806605715424105e-07, + "loss": 0.1047, + "step": 11116 + }, + { + "epoch": 0.58, + "learning_rate": 1.5803299783674408e-07, + "loss": 0.1517, + "step": 11117 + }, + { + "epoch": 0.58, + "learning_rate": 1.5799993971868025e-07, + "loss": 0.128, + "step": 11118 + }, + { + "epoch": 0.58, + "learning_rate": 1.5796688280099427e-07, + "loss": 0.1281, + "step": 11119 + }, + { + "epoch": 0.58, + "learning_rate": 1.5793382708463103e-07, + "loss": 0.1288, + "step": 11120 + }, + { + "epoch": 0.58, + "learning_rate": 1.5790077257053524e-07, + "loss": 0.0959, + "step": 11121 + }, + { + "epoch": 0.58, + "learning_rate": 1.5786771925965163e-07, + "loss": 0.1161, + "step": 11122 + }, + { + "epoch": 0.58, + "learning_rate": 1.5783466715292486e-07, + "loss": 0.1408, + "step": 11123 + }, + { + "epoch": 0.58, + "learning_rate": 1.5780161625129962e-07, + "loss": 0.1725, + "step": 11124 + }, + { + "epoch": 0.58, + "learning_rate": 1.5776856655572043e-07, + "loss": 0.1178, + "step": 11125 + }, + { + "epoch": 0.58, + "learning_rate": 1.5773551806713197e-07, + "loss": 0.1225, + "step": 11126 + }, + { + "epoch": 0.58, + "learning_rate": 1.577024707864787e-07, + "loss": 0.16, + "step": 11127 + }, + { + "epoch": 0.58, + "learning_rate": 1.576694247147051e-07, + "loss": 0.1832, + "step": 11128 + }, + { + "epoch": 0.58, + "learning_rate": 1.576363798527557e-07, + "loss": 0.1579, + "step": 11129 + }, + { + "epoch": 0.58, + "learning_rate": 1.576033362015749e-07, + "loss": 0.1377, + "step": 11130 + }, + { + "epoch": 0.58, + "learning_rate": 1.5757029376210718e-07, + "loss": 0.1169, + "step": 11131 + }, + { + "epoch": 0.58, + "learning_rate": 1.5753725253529677e-07, + "loss": 0.1639, + "step": 11132 + }, + { + "epoch": 0.58, + "learning_rate": 1.5750421252208814e-07, + "loss": 0.1211, + "step": 11133 + }, + { + "epoch": 0.58, + "learning_rate": 1.5747117372342547e-07, + "loss": 0.137, + "step": 11134 + }, + { + "epoch": 0.58, + "learning_rate": 1.5743813614025315e-07, + "loss": 0.1378, + "step": 11135 + }, + { + "epoch": 0.58, + "learning_rate": 1.574050997735153e-07, + "loss": 0.1172, + "step": 11136 + }, + { + "epoch": 0.58, + "learning_rate": 1.573720646241561e-07, + "loss": 0.1233, + "step": 11137 + }, + { + "epoch": 0.58, + "learning_rate": 1.5733903069311976e-07, + "loss": 0.1246, + "step": 11138 + }, + { + "epoch": 0.58, + "learning_rate": 1.5730599798135038e-07, + "loss": 0.1488, + "step": 11139 + }, + { + "epoch": 0.58, + "learning_rate": 1.5727296648979208e-07, + "loss": 0.1475, + "step": 11140 + }, + { + "epoch": 0.58, + "learning_rate": 1.5723993621938887e-07, + "loss": 0.1232, + "step": 11141 + }, + { + "epoch": 0.58, + "learning_rate": 1.5720690717108482e-07, + "loss": 0.1257, + "step": 11142 + }, + { + "epoch": 0.58, + "learning_rate": 1.5717387934582385e-07, + "loss": 0.1179, + "step": 11143 + }, + { + "epoch": 0.58, + "learning_rate": 1.5714085274455002e-07, + "loss": 0.1376, + "step": 11144 + }, + { + "epoch": 0.58, + "learning_rate": 1.5710782736820712e-07, + "loss": 0.1286, + "step": 11145 + }, + { + "epoch": 0.58, + "learning_rate": 1.5707480321773903e-07, + "loss": 0.1302, + "step": 11146 + }, + { + "epoch": 0.58, + "learning_rate": 1.570417802940897e-07, + "loss": 0.1313, + "step": 11147 + }, + { + "epoch": 0.58, + "learning_rate": 1.570087585982028e-07, + "loss": 0.096, + "step": 11148 + }, + { + "epoch": 0.58, + "learning_rate": 1.5697573813102222e-07, + "loss": 0.1268, + "step": 11149 + }, + { + "epoch": 0.58, + "learning_rate": 1.5694271889349163e-07, + "loss": 0.1164, + "step": 11150 + }, + { + "epoch": 0.58, + "learning_rate": 1.5690970088655478e-07, + "loss": 0.1156, + "step": 11151 + }, + { + "epoch": 0.58, + "learning_rate": 1.5687668411115537e-07, + "loss": 0.165, + "step": 11152 + }, + { + "epoch": 0.58, + "learning_rate": 1.5684366856823688e-07, + "loss": 0.1385, + "step": 11153 + }, + { + "epoch": 0.58, + "learning_rate": 1.5681065425874305e-07, + "loss": 0.1313, + "step": 11154 + }, + { + "epoch": 0.58, + "learning_rate": 1.5677764118361734e-07, + "loss": 0.1118, + "step": 11155 + }, + { + "epoch": 0.58, + "learning_rate": 1.5674462934380336e-07, + "loss": 0.1457, + "step": 11156 + }, + { + "epoch": 0.58, + "learning_rate": 1.5671161874024456e-07, + "loss": 0.1238, + "step": 11157 + }, + { + "epoch": 0.58, + "learning_rate": 1.5667860937388445e-07, + "loss": 0.1666, + "step": 11158 + }, + { + "epoch": 0.58, + "learning_rate": 1.5664560124566636e-07, + "loss": 0.1278, + "step": 11159 + }, + { + "epoch": 0.58, + "learning_rate": 1.5661259435653375e-07, + "loss": 0.1666, + "step": 11160 + }, + { + "epoch": 0.58, + "learning_rate": 1.5657958870743e-07, + "loss": 0.1472, + "step": 11161 + }, + { + "epoch": 0.58, + "learning_rate": 1.5654658429929828e-07, + "loss": 0.1177, + "step": 11162 + }, + { + "epoch": 0.58, + "learning_rate": 1.56513581133082e-07, + "loss": 0.1191, + "step": 11163 + }, + { + "epoch": 0.58, + "learning_rate": 1.5648057920972432e-07, + "loss": 0.1365, + "step": 11164 + }, + { + "epoch": 0.58, + "learning_rate": 1.564475785301685e-07, + "loss": 0.1362, + "step": 11165 + }, + { + "epoch": 0.58, + "learning_rate": 1.564145790953577e-07, + "loss": 0.19, + "step": 11166 + }, + { + "epoch": 0.58, + "learning_rate": 1.5638158090623509e-07, + "loss": 0.1771, + "step": 11167 + }, + { + "epoch": 0.58, + "learning_rate": 1.5634858396374367e-07, + "loss": 0.1133, + "step": 11168 + }, + { + "epoch": 0.58, + "learning_rate": 1.5631558826882665e-07, + "loss": 0.1354, + "step": 11169 + }, + { + "epoch": 0.58, + "learning_rate": 1.5628259382242693e-07, + "loss": 0.1088, + "step": 11170 + }, + { + "epoch": 0.58, + "learning_rate": 1.5624960062548755e-07, + "loss": 0.1507, + "step": 11171 + }, + { + "epoch": 0.58, + "learning_rate": 1.562166086789515e-07, + "loss": 0.1284, + "step": 11172 + }, + { + "epoch": 0.58, + "learning_rate": 1.5618361798376165e-07, + "loss": 0.1228, + "step": 11173 + }, + { + "epoch": 0.58, + "learning_rate": 1.5615062854086093e-07, + "loss": 0.136, + "step": 11174 + }, + { + "epoch": 0.58, + "learning_rate": 1.5611764035119215e-07, + "loss": 0.1452, + "step": 11175 + }, + { + "epoch": 0.58, + "learning_rate": 1.560846534156982e-07, + "loss": 0.1467, + "step": 11176 + }, + { + "epoch": 0.58, + "learning_rate": 1.5605166773532173e-07, + "loss": 0.1227, + "step": 11177 + }, + { + "epoch": 0.58, + "learning_rate": 1.5601868331100567e-07, + "loss": 0.1387, + "step": 11178 + }, + { + "epoch": 0.58, + "learning_rate": 1.5598570014369258e-07, + "loss": 0.1198, + "step": 11179 + }, + { + "epoch": 0.58, + "learning_rate": 1.5595271823432513e-07, + "loss": 0.1133, + "step": 11180 + }, + { + "epoch": 0.58, + "learning_rate": 1.5591973758384603e-07, + "loss": 0.1244, + "step": 11181 + }, + { + "epoch": 0.58, + "learning_rate": 1.5588675819319778e-07, + "loss": 0.1996, + "step": 11182 + }, + { + "epoch": 0.58, + "learning_rate": 1.558537800633231e-07, + "loss": 0.1273, + "step": 11183 + }, + { + "epoch": 0.58, + "learning_rate": 1.5582080319516436e-07, + "loss": 0.1431, + "step": 11184 + }, + { + "epoch": 0.58, + "learning_rate": 1.5578782758966416e-07, + "loss": 0.1536, + "step": 11185 + }, + { + "epoch": 0.58, + "learning_rate": 1.5575485324776495e-07, + "loss": 0.151, + "step": 11186 + }, + { + "epoch": 0.58, + "learning_rate": 1.5572188017040906e-07, + "loss": 0.1322, + "step": 11187 + }, + { + "epoch": 0.58, + "learning_rate": 1.5568890835853896e-07, + "loss": 0.1349, + "step": 11188 + }, + { + "epoch": 0.58, + "learning_rate": 1.5565593781309693e-07, + "loss": 0.1291, + "step": 11189 + }, + { + "epoch": 0.58, + "learning_rate": 1.5562296853502532e-07, + "loss": 0.128, + "step": 11190 + }, + { + "epoch": 0.58, + "learning_rate": 1.555900005252664e-07, + "loss": 0.1036, + "step": 11191 + }, + { + "epoch": 0.58, + "learning_rate": 1.5555703378476243e-07, + "loss": 0.1326, + "step": 11192 + }, + { + "epoch": 0.58, + "learning_rate": 1.5552406831445555e-07, + "loss": 0.154, + "step": 11193 + }, + { + "epoch": 0.58, + "learning_rate": 1.5549110411528802e-07, + "loss": 0.1134, + "step": 11194 + }, + { + "epoch": 0.58, + "learning_rate": 1.5545814118820193e-07, + "loss": 0.1255, + "step": 11195 + }, + { + "epoch": 0.58, + "learning_rate": 1.5542517953413934e-07, + "loss": 0.1229, + "step": 11196 + }, + { + "epoch": 0.58, + "learning_rate": 1.553922191540423e-07, + "loss": 0.1529, + "step": 11197 + }, + { + "epoch": 0.58, + "learning_rate": 1.5535926004885286e-07, + "loss": 0.1107, + "step": 11198 + }, + { + "epoch": 0.58, + "learning_rate": 1.55326302219513e-07, + "loss": 0.1214, + "step": 11199 + }, + { + "epoch": 0.58, + "learning_rate": 1.5529334566696467e-07, + "loss": 0.1272, + "step": 11200 + }, + { + "epoch": 0.58, + "learning_rate": 1.5526039039214977e-07, + "loss": 0.1573, + "step": 11201 + }, + { + "epoch": 0.58, + "learning_rate": 1.552274363960102e-07, + "loss": 0.1442, + "step": 11202 + }, + { + "epoch": 0.58, + "learning_rate": 1.5519448367948778e-07, + "loss": 0.1376, + "step": 11203 + }, + { + "epoch": 0.58, + "learning_rate": 1.5516153224352437e-07, + "loss": 0.1289, + "step": 11204 + }, + { + "epoch": 0.58, + "learning_rate": 1.5512858208906163e-07, + "loss": 0.1455, + "step": 11205 + }, + { + "epoch": 0.58, + "learning_rate": 1.550956332170413e-07, + "loss": 0.1755, + "step": 11206 + }, + { + "epoch": 0.58, + "learning_rate": 1.5506268562840515e-07, + "loss": 0.1541, + "step": 11207 + }, + { + "epoch": 0.59, + "learning_rate": 1.5502973932409475e-07, + "loss": 0.1458, + "step": 11208 + }, + { + "epoch": 0.59, + "learning_rate": 1.549967943050518e-07, + "loss": 0.1334, + "step": 11209 + }, + { + "epoch": 0.59, + "learning_rate": 1.5496385057221781e-07, + "loss": 0.1063, + "step": 11210 + }, + { + "epoch": 0.59, + "learning_rate": 1.5493090812653437e-07, + "loss": 0.1118, + "step": 11211 + }, + { + "epoch": 0.59, + "learning_rate": 1.5489796696894306e-07, + "loss": 0.1436, + "step": 11212 + }, + { + "epoch": 0.59, + "learning_rate": 1.5486502710038515e-07, + "loss": 0.1315, + "step": 11213 + }, + { + "epoch": 0.59, + "learning_rate": 1.5483208852180223e-07, + "loss": 0.1316, + "step": 11214 + }, + { + "epoch": 0.59, + "learning_rate": 1.5479915123413563e-07, + "loss": 0.1246, + "step": 11215 + }, + { + "epoch": 0.59, + "learning_rate": 1.5476621523832679e-07, + "loss": 0.1386, + "step": 11216 + }, + { + "epoch": 0.59, + "learning_rate": 1.547332805353169e-07, + "loss": 0.1281, + "step": 11217 + }, + { + "epoch": 0.59, + "learning_rate": 1.5470034712604741e-07, + "loss": 0.146, + "step": 11218 + }, + { + "epoch": 0.59, + "learning_rate": 1.5466741501145942e-07, + "loss": 0.1306, + "step": 11219 + }, + { + "epoch": 0.59, + "learning_rate": 1.5463448419249426e-07, + "loss": 0.1398, + "step": 11220 + }, + { + "epoch": 0.59, + "learning_rate": 1.5460155467009308e-07, + "loss": 0.1366, + "step": 11221 + }, + { + "epoch": 0.59, + "learning_rate": 1.5456862644519692e-07, + "loss": 0.1867, + "step": 11222 + }, + { + "epoch": 0.59, + "learning_rate": 1.54535699518747e-07, + "loss": 0.1279, + "step": 11223 + }, + { + "epoch": 0.59, + "learning_rate": 1.545027738916843e-07, + "loss": 0.1388, + "step": 11224 + }, + { + "epoch": 0.59, + "learning_rate": 1.5446984956494992e-07, + "loss": 0.1265, + "step": 11225 + }, + { + "epoch": 0.59, + "learning_rate": 1.5443692653948478e-07, + "loss": 0.1014, + "step": 11226 + }, + { + "epoch": 0.59, + "learning_rate": 1.544040048162299e-07, + "loss": 0.1387, + "step": 11227 + }, + { + "epoch": 0.59, + "learning_rate": 1.5437108439612614e-07, + "loss": 0.1519, + "step": 11228 + }, + { + "epoch": 0.59, + "learning_rate": 1.5433816528011446e-07, + "loss": 0.1456, + "step": 11229 + }, + { + "epoch": 0.59, + "learning_rate": 1.5430524746913563e-07, + "loss": 0.145, + "step": 11230 + }, + { + "epoch": 0.59, + "learning_rate": 1.542723309641304e-07, + "loss": 0.1247, + "step": 11231 + }, + { + "epoch": 0.59, + "learning_rate": 1.5423941576603966e-07, + "loss": 0.1386, + "step": 11232 + }, + { + "epoch": 0.59, + "learning_rate": 1.5420650187580405e-07, + "loss": 0.1211, + "step": 11233 + }, + { + "epoch": 0.59, + "learning_rate": 1.5417358929436434e-07, + "loss": 0.1372, + "step": 11234 + }, + { + "epoch": 0.59, + "learning_rate": 1.5414067802266107e-07, + "loss": 0.1637, + "step": 11235 + }, + { + "epoch": 0.59, + "learning_rate": 1.5410776806163499e-07, + "loss": 0.1746, + "step": 11236 + }, + { + "epoch": 0.59, + "learning_rate": 1.5407485941222656e-07, + "loss": 0.1538, + "step": 11237 + }, + { + "epoch": 0.59, + "learning_rate": 1.5404195207537644e-07, + "loss": 0.1759, + "step": 11238 + }, + { + "epoch": 0.59, + "learning_rate": 1.5400904605202508e-07, + "loss": 0.1741, + "step": 11239 + }, + { + "epoch": 0.59, + "learning_rate": 1.539761413431129e-07, + "loss": 0.1326, + "step": 11240 + }, + { + "epoch": 0.59, + "learning_rate": 1.5394323794958032e-07, + "loss": 0.1464, + "step": 11241 + }, + { + "epoch": 0.59, + "learning_rate": 1.539103358723678e-07, + "loss": 0.1417, + "step": 11242 + }, + { + "epoch": 0.59, + "learning_rate": 1.5387743511241574e-07, + "loss": 0.1413, + "step": 11243 + }, + { + "epoch": 0.59, + "learning_rate": 1.5384453567066432e-07, + "loss": 0.1344, + "step": 11244 + }, + { + "epoch": 0.59, + "learning_rate": 1.538116375480539e-07, + "loss": 0.1004, + "step": 11245 + }, + { + "epoch": 0.59, + "learning_rate": 1.5377874074552478e-07, + "loss": 0.1548, + "step": 11246 + }, + { + "epoch": 0.59, + "learning_rate": 1.5374584526401702e-07, + "loss": 0.1387, + "step": 11247 + }, + { + "epoch": 0.59, + "learning_rate": 1.537129511044709e-07, + "loss": 0.1297, + "step": 11248 + }, + { + "epoch": 0.59, + "learning_rate": 1.5368005826782645e-07, + "loss": 0.1409, + "step": 11249 + }, + { + "epoch": 0.59, + "learning_rate": 1.5364716675502388e-07, + "loss": 0.1098, + "step": 11250 + }, + { + "epoch": 0.59, + "learning_rate": 1.536142765670031e-07, + "loss": 0.1223, + "step": 11251 + }, + { + "epoch": 0.59, + "learning_rate": 1.5358138770470426e-07, + "loss": 0.1172, + "step": 11252 + }, + { + "epoch": 0.59, + "learning_rate": 1.5354850016906725e-07, + "loss": 0.1421, + "step": 11253 + }, + { + "epoch": 0.59, + "learning_rate": 1.5351561396103205e-07, + "loss": 0.1517, + "step": 11254 + }, + { + "epoch": 0.59, + "learning_rate": 1.5348272908153862e-07, + "loss": 0.1871, + "step": 11255 + }, + { + "epoch": 0.59, + "learning_rate": 1.5344984553152666e-07, + "loss": 0.0949, + "step": 11256 + }, + { + "epoch": 0.59, + "learning_rate": 1.534169633119361e-07, + "loss": 0.1173, + "step": 11257 + }, + { + "epoch": 0.59, + "learning_rate": 1.533840824237067e-07, + "loss": 0.126, + "step": 11258 + }, + { + "epoch": 0.59, + "learning_rate": 1.5335120286777824e-07, + "loss": 0.1295, + "step": 11259 + }, + { + "epoch": 0.59, + "learning_rate": 1.5331832464509037e-07, + "loss": 0.132, + "step": 11260 + }, + { + "epoch": 0.59, + "learning_rate": 1.5328544775658287e-07, + "loss": 0.1248, + "step": 11261 + }, + { + "epoch": 0.59, + "learning_rate": 1.5325257220319525e-07, + "loss": 0.1321, + "step": 11262 + }, + { + "epoch": 0.59, + "learning_rate": 1.5321969798586724e-07, + "loss": 0.0969, + "step": 11263 + }, + { + "epoch": 0.59, + "learning_rate": 1.531868251055383e-07, + "loss": 0.1576, + "step": 11264 + }, + { + "epoch": 0.59, + "learning_rate": 1.531539535631479e-07, + "loss": 0.1122, + "step": 11265 + }, + { + "epoch": 0.59, + "learning_rate": 1.5312108335963565e-07, + "loss": 0.1295, + "step": 11266 + }, + { + "epoch": 0.59, + "learning_rate": 1.530882144959409e-07, + "loss": 0.1292, + "step": 11267 + }, + { + "epoch": 0.59, + "learning_rate": 1.5305534697300313e-07, + "loss": 0.1375, + "step": 11268 + }, + { + "epoch": 0.59, + "learning_rate": 1.530224807917616e-07, + "loss": 0.1256, + "step": 11269 + }, + { + "epoch": 0.59, + "learning_rate": 1.5298961595315575e-07, + "loss": 0.1602, + "step": 11270 + }, + { + "epoch": 0.59, + "learning_rate": 1.529567524581248e-07, + "loss": 0.1323, + "step": 11271 + }, + { + "epoch": 0.59, + "learning_rate": 1.5292389030760814e-07, + "loss": 0.1344, + "step": 11272 + }, + { + "epoch": 0.59, + "learning_rate": 1.5289102950254477e-07, + "loss": 0.1245, + "step": 11273 + }, + { + "epoch": 0.59, + "learning_rate": 1.5285817004387394e-07, + "loss": 0.1266, + "step": 11274 + }, + { + "epoch": 0.59, + "learning_rate": 1.5282531193253485e-07, + "loss": 0.1231, + "step": 11275 + }, + { + "epoch": 0.59, + "learning_rate": 1.527924551694665e-07, + "loss": 0.1601, + "step": 11276 + }, + { + "epoch": 0.59, + "learning_rate": 1.5275959975560805e-07, + "loss": 0.1501, + "step": 11277 + }, + { + "epoch": 0.59, + "learning_rate": 1.5272674569189846e-07, + "loss": 0.1123, + "step": 11278 + }, + { + "epoch": 0.59, + "learning_rate": 1.5269389297927674e-07, + "loss": 0.1534, + "step": 11279 + }, + { + "epoch": 0.59, + "learning_rate": 1.526610416186818e-07, + "loss": 0.128, + "step": 11280 + }, + { + "epoch": 0.59, + "learning_rate": 1.5262819161105265e-07, + "loss": 0.151, + "step": 11281 + }, + { + "epoch": 0.59, + "learning_rate": 1.5259534295732804e-07, + "loss": 0.1433, + "step": 11282 + }, + { + "epoch": 0.59, + "learning_rate": 1.5256249565844676e-07, + "loss": 0.1558, + "step": 11283 + }, + { + "epoch": 0.59, + "learning_rate": 1.525296497153477e-07, + "loss": 0.1438, + "step": 11284 + }, + { + "epoch": 0.59, + "learning_rate": 1.5249680512896955e-07, + "loss": 0.1237, + "step": 11285 + }, + { + "epoch": 0.59, + "learning_rate": 1.524639619002511e-07, + "loss": 0.118, + "step": 11286 + }, + { + "epoch": 0.59, + "learning_rate": 1.5243112003013093e-07, + "loss": 0.1423, + "step": 11287 + }, + { + "epoch": 0.59, + "learning_rate": 1.5239827951954774e-07, + "loss": 0.1404, + "step": 11288 + }, + { + "epoch": 0.59, + "learning_rate": 1.5236544036944015e-07, + "loss": 0.132, + "step": 11289 + }, + { + "epoch": 0.59, + "learning_rate": 1.5233260258074657e-07, + "loss": 0.1341, + "step": 11290 + }, + { + "epoch": 0.59, + "learning_rate": 1.5229976615440564e-07, + "loss": 0.1359, + "step": 11291 + }, + { + "epoch": 0.59, + "learning_rate": 1.5226693109135577e-07, + "loss": 0.1361, + "step": 11292 + }, + { + "epoch": 0.59, + "learning_rate": 1.522340973925355e-07, + "loss": 0.1461, + "step": 11293 + }, + { + "epoch": 0.59, + "learning_rate": 1.5220126505888312e-07, + "loss": 0.1351, + "step": 11294 + }, + { + "epoch": 0.59, + "learning_rate": 1.5216843409133707e-07, + "loss": 0.1596, + "step": 11295 + }, + { + "epoch": 0.59, + "learning_rate": 1.5213560449083558e-07, + "loss": 0.1214, + "step": 11296 + }, + { + "epoch": 0.59, + "learning_rate": 1.5210277625831705e-07, + "loss": 0.1279, + "step": 11297 + }, + { + "epoch": 0.59, + "learning_rate": 1.5206994939471972e-07, + "loss": 0.1329, + "step": 11298 + }, + { + "epoch": 0.59, + "learning_rate": 1.520371239009816e-07, + "loss": 0.1366, + "step": 11299 + }, + { + "epoch": 0.59, + "learning_rate": 1.5200429977804112e-07, + "loss": 0.1595, + "step": 11300 + }, + { + "epoch": 0.59, + "learning_rate": 1.519714770268362e-07, + "loss": 0.1583, + "step": 11301 + }, + { + "epoch": 0.59, + "learning_rate": 1.5193865564830503e-07, + "loss": 0.1502, + "step": 11302 + }, + { + "epoch": 0.59, + "learning_rate": 1.5190583564338563e-07, + "loss": 0.1181, + "step": 11303 + }, + { + "epoch": 0.59, + "learning_rate": 1.5187301701301602e-07, + "loss": 0.15, + "step": 11304 + }, + { + "epoch": 0.59, + "learning_rate": 1.5184019975813414e-07, + "loss": 0.127, + "step": 11305 + }, + { + "epoch": 0.59, + "learning_rate": 1.5180738387967806e-07, + "loss": 0.1484, + "step": 11306 + }, + { + "epoch": 0.59, + "learning_rate": 1.5177456937858542e-07, + "loss": 0.129, + "step": 11307 + }, + { + "epoch": 0.59, + "learning_rate": 1.5174175625579426e-07, + "loss": 0.1381, + "step": 11308 + }, + { + "epoch": 0.59, + "learning_rate": 1.517089445122423e-07, + "loss": 0.1576, + "step": 11309 + }, + { + "epoch": 0.59, + "learning_rate": 1.5167613414886736e-07, + "loss": 0.1565, + "step": 11310 + }, + { + "epoch": 0.59, + "learning_rate": 1.5164332516660716e-07, + "loss": 0.1505, + "step": 11311 + }, + { + "epoch": 0.59, + "learning_rate": 1.516105175663994e-07, + "loss": 0.1566, + "step": 11312 + }, + { + "epoch": 0.59, + "learning_rate": 1.515777113491817e-07, + "loss": 0.1296, + "step": 11313 + }, + { + "epoch": 0.59, + "learning_rate": 1.5154490651589174e-07, + "loss": 0.142, + "step": 11314 + }, + { + "epoch": 0.59, + "learning_rate": 1.5151210306746714e-07, + "loss": 0.1451, + "step": 11315 + }, + { + "epoch": 0.59, + "learning_rate": 1.5147930100484522e-07, + "loss": 0.1465, + "step": 11316 + }, + { + "epoch": 0.59, + "learning_rate": 1.5144650032896367e-07, + "loss": 0.1217, + "step": 11317 + }, + { + "epoch": 0.59, + "learning_rate": 1.5141370104075984e-07, + "loss": 0.1254, + "step": 11318 + }, + { + "epoch": 0.59, + "learning_rate": 1.513809031411712e-07, + "loss": 0.1147, + "step": 11319 + }, + { + "epoch": 0.59, + "learning_rate": 1.513481066311351e-07, + "loss": 0.1341, + "step": 11320 + }, + { + "epoch": 0.59, + "learning_rate": 1.5131531151158897e-07, + "loss": 0.1975, + "step": 11321 + }, + { + "epoch": 0.59, + "learning_rate": 1.5128251778346994e-07, + "loss": 0.1373, + "step": 11322 + }, + { + "epoch": 0.59, + "learning_rate": 1.5124972544771545e-07, + "loss": 0.1419, + "step": 11323 + }, + { + "epoch": 0.59, + "learning_rate": 1.512169345052626e-07, + "loss": 0.1571, + "step": 11324 + }, + { + "epoch": 0.59, + "learning_rate": 1.5118414495704856e-07, + "loss": 0.1334, + "step": 11325 + }, + { + "epoch": 0.59, + "learning_rate": 1.5115135680401055e-07, + "loss": 0.1204, + "step": 11326 + }, + { + "epoch": 0.59, + "learning_rate": 1.5111857004708555e-07, + "loss": 0.1423, + "step": 11327 + }, + { + "epoch": 0.59, + "learning_rate": 1.5108578468721074e-07, + "loss": 0.1495, + "step": 11328 + }, + { + "epoch": 0.59, + "learning_rate": 1.5105300072532304e-07, + "loss": 0.1265, + "step": 11329 + }, + { + "epoch": 0.59, + "learning_rate": 1.5102021816235952e-07, + "loss": 0.1713, + "step": 11330 + }, + { + "epoch": 0.59, + "learning_rate": 1.5098743699925702e-07, + "loss": 0.1137, + "step": 11331 + }, + { + "epoch": 0.59, + "learning_rate": 1.509546572369526e-07, + "loss": 0.1464, + "step": 11332 + }, + { + "epoch": 0.59, + "learning_rate": 1.5092187887638294e-07, + "loss": 0.113, + "step": 11333 + }, + { + "epoch": 0.59, + "learning_rate": 1.508891019184849e-07, + "loss": 0.1273, + "step": 11334 + }, + { + "epoch": 0.59, + "learning_rate": 1.5085632636419534e-07, + "loss": 0.1043, + "step": 11335 + }, + { + "epoch": 0.59, + "learning_rate": 1.508235522144509e-07, + "loss": 0.1336, + "step": 11336 + }, + { + "epoch": 0.59, + "learning_rate": 1.5079077947018834e-07, + "loss": 0.1555, + "step": 11337 + }, + { + "epoch": 0.59, + "learning_rate": 1.507580081323443e-07, + "loss": 0.1276, + "step": 11338 + }, + { + "epoch": 0.59, + "learning_rate": 1.5072523820185537e-07, + "loss": 0.1493, + "step": 11339 + }, + { + "epoch": 0.59, + "learning_rate": 1.506924696796582e-07, + "loss": 0.1366, + "step": 11340 + }, + { + "epoch": 0.59, + "learning_rate": 1.506597025666893e-07, + "loss": 0.1179, + "step": 11341 + }, + { + "epoch": 0.59, + "learning_rate": 1.5062693686388511e-07, + "loss": 0.1782, + "step": 11342 + }, + { + "epoch": 0.59, + "learning_rate": 1.5059417257218213e-07, + "loss": 0.1081, + "step": 11343 + }, + { + "epoch": 0.59, + "learning_rate": 1.5056140969251677e-07, + "loss": 0.1319, + "step": 11344 + }, + { + "epoch": 0.59, + "learning_rate": 1.5052864822582537e-07, + "loss": 0.1487, + "step": 11345 + }, + { + "epoch": 0.59, + "learning_rate": 1.5049588817304432e-07, + "loss": 0.1136, + "step": 11346 + }, + { + "epoch": 0.59, + "learning_rate": 1.504631295351099e-07, + "loss": 0.0934, + "step": 11347 + }, + { + "epoch": 0.59, + "learning_rate": 1.5043037231295838e-07, + "loss": 0.1671, + "step": 11348 + }, + { + "epoch": 0.59, + "learning_rate": 1.50397616507526e-07, + "loss": 0.1964, + "step": 11349 + }, + { + "epoch": 0.59, + "learning_rate": 1.5036486211974882e-07, + "loss": 0.1043, + "step": 11350 + }, + { + "epoch": 0.59, + "learning_rate": 1.5033210915056307e-07, + "loss": 0.1677, + "step": 11351 + }, + { + "epoch": 0.59, + "learning_rate": 1.5029935760090475e-07, + "loss": 0.1378, + "step": 11352 + }, + { + "epoch": 0.59, + "learning_rate": 1.5026660747171004e-07, + "loss": 0.1253, + "step": 11353 + }, + { + "epoch": 0.59, + "learning_rate": 1.5023385876391486e-07, + "loss": 0.1617, + "step": 11354 + }, + { + "epoch": 0.59, + "learning_rate": 1.5020111147845523e-07, + "loss": 0.1418, + "step": 11355 + }, + { + "epoch": 0.59, + "learning_rate": 1.5016836561626703e-07, + "loss": 0.0995, + "step": 11356 + }, + { + "epoch": 0.59, + "learning_rate": 1.501356211782862e-07, + "loss": 0.1356, + "step": 11357 + }, + { + "epoch": 0.59, + "learning_rate": 1.5010287816544863e-07, + "loss": 0.129, + "step": 11358 + }, + { + "epoch": 0.59, + "learning_rate": 1.5007013657868996e-07, + "loss": 0.1324, + "step": 11359 + }, + { + "epoch": 0.59, + "learning_rate": 1.5003739641894614e-07, + "loss": 0.1418, + "step": 11360 + }, + { + "epoch": 0.59, + "learning_rate": 1.5000465768715272e-07, + "loss": 0.129, + "step": 11361 + }, + { + "epoch": 0.59, + "learning_rate": 1.4997192038424555e-07, + "loss": 0.1637, + "step": 11362 + }, + { + "epoch": 0.59, + "learning_rate": 1.4993918451116017e-07, + "loss": 0.1222, + "step": 11363 + }, + { + "epoch": 0.59, + "learning_rate": 1.499064500688323e-07, + "loss": 0.1989, + "step": 11364 + }, + { + "epoch": 0.59, + "learning_rate": 1.4987371705819732e-07, + "loss": 0.1463, + "step": 11365 + }, + { + "epoch": 0.59, + "learning_rate": 1.4984098548019097e-07, + "loss": 0.1485, + "step": 11366 + }, + { + "epoch": 0.59, + "learning_rate": 1.4980825533574857e-07, + "loss": 0.1456, + "step": 11367 + }, + { + "epoch": 0.59, + "learning_rate": 1.4977552662580558e-07, + "loss": 0.1224, + "step": 11368 + }, + { + "epoch": 0.59, + "learning_rate": 1.4974279935129744e-07, + "loss": 0.1576, + "step": 11369 + }, + { + "epoch": 0.59, + "learning_rate": 1.497100735131595e-07, + "loss": 0.1721, + "step": 11370 + }, + { + "epoch": 0.59, + "learning_rate": 1.496773491123271e-07, + "loss": 0.1598, + "step": 11371 + }, + { + "epoch": 0.59, + "learning_rate": 1.4964462614973542e-07, + "loss": 0.1233, + "step": 11372 + }, + { + "epoch": 0.59, + "learning_rate": 1.4961190462631982e-07, + "loss": 0.1194, + "step": 11373 + }, + { + "epoch": 0.59, + "learning_rate": 1.495791845430154e-07, + "loss": 0.1695, + "step": 11374 + }, + { + "epoch": 0.59, + "learning_rate": 1.4954646590075744e-07, + "loss": 0.1317, + "step": 11375 + }, + { + "epoch": 0.59, + "learning_rate": 1.495137487004809e-07, + "loss": 0.1274, + "step": 11376 + }, + { + "epoch": 0.59, + "learning_rate": 1.4948103294312092e-07, + "loss": 0.1175, + "step": 11377 + }, + { + "epoch": 0.59, + "learning_rate": 1.494483186296125e-07, + "loss": 0.1571, + "step": 11378 + }, + { + "epoch": 0.59, + "learning_rate": 1.4941560576089064e-07, + "loss": 0.123, + "step": 11379 + }, + { + "epoch": 0.59, + "learning_rate": 1.493828943378903e-07, + "loss": 0.1419, + "step": 11380 + }, + { + "epoch": 0.59, + "learning_rate": 1.4935018436154638e-07, + "loss": 0.1744, + "step": 11381 + }, + { + "epoch": 0.59, + "learning_rate": 1.4931747583279378e-07, + "loss": 0.1499, + "step": 11382 + }, + { + "epoch": 0.59, + "learning_rate": 1.4928476875256732e-07, + "loss": 0.1261, + "step": 11383 + }, + { + "epoch": 0.59, + "learning_rate": 1.4925206312180168e-07, + "loss": 0.1231, + "step": 11384 + }, + { + "epoch": 0.59, + "learning_rate": 1.4921935894143168e-07, + "loss": 0.1492, + "step": 11385 + }, + { + "epoch": 0.59, + "learning_rate": 1.4918665621239197e-07, + "loss": 0.1452, + "step": 11386 + }, + { + "epoch": 0.59, + "learning_rate": 1.491539549356173e-07, + "loss": 0.1115, + "step": 11387 + }, + { + "epoch": 0.59, + "learning_rate": 1.4912125511204216e-07, + "loss": 0.1482, + "step": 11388 + }, + { + "epoch": 0.59, + "learning_rate": 1.4908855674260125e-07, + "loss": 0.1442, + "step": 11389 + }, + { + "epoch": 0.59, + "learning_rate": 1.49055859828229e-07, + "loss": 0.1653, + "step": 11390 + }, + { + "epoch": 0.59, + "learning_rate": 1.4902316436986e-07, + "loss": 0.105, + "step": 11391 + }, + { + "epoch": 0.59, + "learning_rate": 1.4899047036842865e-07, + "loss": 0.1254, + "step": 11392 + }, + { + "epoch": 0.59, + "learning_rate": 1.4895777782486926e-07, + "loss": 0.184, + "step": 11393 + }, + { + "epoch": 0.59, + "learning_rate": 1.4892508674011636e-07, + "loss": 0.149, + "step": 11394 + }, + { + "epoch": 0.59, + "learning_rate": 1.4889239711510412e-07, + "loss": 0.1215, + "step": 11395 + }, + { + "epoch": 0.59, + "learning_rate": 1.4885970895076695e-07, + "loss": 0.1256, + "step": 11396 + }, + { + "epoch": 0.59, + "learning_rate": 1.48827022248039e-07, + "loss": 0.1322, + "step": 11397 + }, + { + "epoch": 0.59, + "learning_rate": 1.4879433700785457e-07, + "loss": 0.1083, + "step": 11398 + }, + { + "epoch": 0.59, + "learning_rate": 1.487616532311477e-07, + "loss": 0.1091, + "step": 11399 + }, + { + "epoch": 0.6, + "learning_rate": 1.487289709188526e-07, + "loss": 0.1153, + "step": 11400 + }, + { + "epoch": 0.6, + "learning_rate": 1.4869629007190337e-07, + "loss": 0.1478, + "step": 11401 + }, + { + "epoch": 0.6, + "learning_rate": 1.4866361069123386e-07, + "loss": 0.1529, + "step": 11402 + }, + { + "epoch": 0.6, + "learning_rate": 1.4863093277777826e-07, + "loss": 0.1287, + "step": 11403 + }, + { + "epoch": 0.6, + "learning_rate": 1.4859825633247035e-07, + "loss": 0.1245, + "step": 11404 + }, + { + "epoch": 0.6, + "learning_rate": 1.4856558135624416e-07, + "loss": 0.1207, + "step": 11405 + }, + { + "epoch": 0.6, + "learning_rate": 1.4853290785003349e-07, + "loss": 0.1455, + "step": 11406 + }, + { + "epoch": 0.6, + "learning_rate": 1.485002358147722e-07, + "loss": 0.141, + "step": 11407 + }, + { + "epoch": 0.6, + "learning_rate": 1.4846756525139402e-07, + "loss": 0.1617, + "step": 11408 + }, + { + "epoch": 0.6, + "learning_rate": 1.4843489616083285e-07, + "loss": 0.1367, + "step": 11409 + }, + { + "epoch": 0.6, + "learning_rate": 1.4840222854402212e-07, + "loss": 0.143, + "step": 11410 + }, + { + "epoch": 0.6, + "learning_rate": 1.483695624018956e-07, + "loss": 0.1598, + "step": 11411 + }, + { + "epoch": 0.6, + "learning_rate": 1.4833689773538697e-07, + "loss": 0.1169, + "step": 11412 + }, + { + "epoch": 0.6, + "learning_rate": 1.4830423454542972e-07, + "loss": 0.102, + "step": 11413 + }, + { + "epoch": 0.6, + "learning_rate": 1.4827157283295746e-07, + "loss": 0.1221, + "step": 11414 + }, + { + "epoch": 0.6, + "learning_rate": 1.4823891259890355e-07, + "loss": 0.1535, + "step": 11415 + }, + { + "epoch": 0.6, + "learning_rate": 1.4820625384420154e-07, + "loss": 0.1714, + "step": 11416 + }, + { + "epoch": 0.6, + "learning_rate": 1.4817359656978477e-07, + "loss": 0.1323, + "step": 11417 + }, + { + "epoch": 0.6, + "learning_rate": 1.4814094077658672e-07, + "loss": 0.1343, + "step": 11418 + }, + { + "epoch": 0.6, + "learning_rate": 1.4810828646554048e-07, + "loss": 0.1979, + "step": 11419 + }, + { + "epoch": 0.6, + "learning_rate": 1.480756336375795e-07, + "loss": 0.1224, + "step": 11420 + }, + { + "epoch": 0.6, + "learning_rate": 1.480429822936369e-07, + "loss": 0.1397, + "step": 11421 + }, + { + "epoch": 0.6, + "learning_rate": 1.4801033243464596e-07, + "loss": 0.0997, + "step": 11422 + }, + { + "epoch": 0.6, + "learning_rate": 1.4797768406153977e-07, + "loss": 0.1369, + "step": 11423 + }, + { + "epoch": 0.6, + "learning_rate": 1.4794503717525148e-07, + "loss": 0.1082, + "step": 11424 + }, + { + "epoch": 0.6, + "learning_rate": 1.479123917767141e-07, + "loss": 0.1634, + "step": 11425 + }, + { + "epoch": 0.6, + "learning_rate": 1.478797478668607e-07, + "loss": 0.0973, + "step": 11426 + }, + { + "epoch": 0.6, + "learning_rate": 1.4784710544662424e-07, + "loss": 0.1247, + "step": 11427 + }, + { + "epoch": 0.6, + "learning_rate": 1.4781446451693756e-07, + "loss": 0.1361, + "step": 11428 + }, + { + "epoch": 0.6, + "learning_rate": 1.4778182507873372e-07, + "loss": 0.1246, + "step": 11429 + }, + { + "epoch": 0.6, + "learning_rate": 1.477491871329454e-07, + "loss": 0.1255, + "step": 11430 + }, + { + "epoch": 0.6, + "learning_rate": 1.477165506805055e-07, + "loss": 0.1183, + "step": 11431 + }, + { + "epoch": 0.6, + "learning_rate": 1.4768391572234676e-07, + "loss": 0.1581, + "step": 11432 + }, + { + "epoch": 0.6, + "learning_rate": 1.4765128225940195e-07, + "loss": 0.101, + "step": 11433 + }, + { + "epoch": 0.6, + "learning_rate": 1.4761865029260366e-07, + "loss": 0.1312, + "step": 11434 + }, + { + "epoch": 0.6, + "learning_rate": 1.4758601982288462e-07, + "loss": 0.1522, + "step": 11435 + }, + { + "epoch": 0.6, + "learning_rate": 1.4755339085117736e-07, + "loss": 0.1169, + "step": 11436 + }, + { + "epoch": 0.6, + "learning_rate": 1.4752076337841435e-07, + "loss": 0.1062, + "step": 11437 + }, + { + "epoch": 0.6, + "learning_rate": 1.4748813740552824e-07, + "loss": 0.1579, + "step": 11438 + }, + { + "epoch": 0.6, + "learning_rate": 1.4745551293345143e-07, + "loss": 0.1474, + "step": 11439 + }, + { + "epoch": 0.6, + "learning_rate": 1.4742288996311635e-07, + "loss": 0.1086, + "step": 11440 + }, + { + "epoch": 0.6, + "learning_rate": 1.4739026849545534e-07, + "loss": 0.1236, + "step": 11441 + }, + { + "epoch": 0.6, + "learning_rate": 1.4735764853140082e-07, + "loss": 0.1365, + "step": 11442 + }, + { + "epoch": 0.6, + "learning_rate": 1.4732503007188505e-07, + "loss": 0.1792, + "step": 11443 + }, + { + "epoch": 0.6, + "learning_rate": 1.4729241311784019e-07, + "loss": 0.1466, + "step": 11444 + }, + { + "epoch": 0.6, + "learning_rate": 1.4725979767019853e-07, + "loss": 0.1092, + "step": 11445 + }, + { + "epoch": 0.6, + "learning_rate": 1.4722718372989218e-07, + "loss": 0.1425, + "step": 11446 + }, + { + "epoch": 0.6, + "learning_rate": 1.4719457129785333e-07, + "loss": 0.1224, + "step": 11447 + }, + { + "epoch": 0.6, + "learning_rate": 1.47161960375014e-07, + "loss": 0.123, + "step": 11448 + }, + { + "epoch": 0.6, + "learning_rate": 1.4712935096230626e-07, + "loss": 0.1312, + "step": 11449 + }, + { + "epoch": 0.6, + "learning_rate": 1.4709674306066204e-07, + "loss": 0.1507, + "step": 11450 + }, + { + "epoch": 0.6, + "learning_rate": 1.4706413667101335e-07, + "loss": 0.1452, + "step": 11451 + }, + { + "epoch": 0.6, + "learning_rate": 1.4703153179429216e-07, + "loss": 0.1436, + "step": 11452 + }, + { + "epoch": 0.6, + "learning_rate": 1.469989284314301e-07, + "loss": 0.1068, + "step": 11453 + }, + { + "epoch": 0.6, + "learning_rate": 1.469663265833592e-07, + "loss": 0.1083, + "step": 11454 + }, + { + "epoch": 0.6, + "learning_rate": 1.4693372625101112e-07, + "loss": 0.1319, + "step": 11455 + }, + { + "epoch": 0.6, + "learning_rate": 1.4690112743531764e-07, + "loss": 0.159, + "step": 11456 + }, + { + "epoch": 0.6, + "learning_rate": 1.4686853013721042e-07, + "loss": 0.1324, + "step": 11457 + }, + { + "epoch": 0.6, + "learning_rate": 1.4683593435762113e-07, + "loss": 0.1569, + "step": 11458 + }, + { + "epoch": 0.6, + "learning_rate": 1.4680334009748135e-07, + "loss": 0.1342, + "step": 11459 + }, + { + "epoch": 0.6, + "learning_rate": 1.467707473577227e-07, + "loss": 0.1243, + "step": 11460 + }, + { + "epoch": 0.6, + "learning_rate": 1.4673815613927666e-07, + "loss": 0.1374, + "step": 11461 + }, + { + "epoch": 0.6, + "learning_rate": 1.467055664430746e-07, + "loss": 0.1886, + "step": 11462 + }, + { + "epoch": 0.6, + "learning_rate": 1.4667297827004809e-07, + "loss": 0.1376, + "step": 11463 + }, + { + "epoch": 0.6, + "learning_rate": 1.4664039162112834e-07, + "loss": 0.1402, + "step": 11464 + }, + { + "epoch": 0.6, + "learning_rate": 1.4660780649724688e-07, + "loss": 0.1229, + "step": 11465 + }, + { + "epoch": 0.6, + "learning_rate": 1.465752228993349e-07, + "loss": 0.1314, + "step": 11466 + }, + { + "epoch": 0.6, + "learning_rate": 1.465426408283237e-07, + "loss": 0.1456, + "step": 11467 + }, + { + "epoch": 0.6, + "learning_rate": 1.465100602851444e-07, + "loss": 0.1152, + "step": 11468 + }, + { + "epoch": 0.6, + "learning_rate": 1.4647748127072832e-07, + "loss": 0.1444, + "step": 11469 + }, + { + "epoch": 0.6, + "learning_rate": 1.4644490378600645e-07, + "loss": 0.1289, + "step": 11470 + }, + { + "epoch": 0.6, + "learning_rate": 1.4641232783190986e-07, + "loss": 0.1835, + "step": 11471 + }, + { + "epoch": 0.6, + "learning_rate": 1.4637975340936967e-07, + "loss": 0.1483, + "step": 11472 + }, + { + "epoch": 0.6, + "learning_rate": 1.4634718051931677e-07, + "loss": 0.1083, + "step": 11473 + }, + { + "epoch": 0.6, + "learning_rate": 1.4631460916268221e-07, + "loss": 0.1031, + "step": 11474 + }, + { + "epoch": 0.6, + "learning_rate": 1.4628203934039683e-07, + "loss": 0.1334, + "step": 11475 + }, + { + "epoch": 0.6, + "learning_rate": 1.4624947105339152e-07, + "loss": 0.1677, + "step": 11476 + }, + { + "epoch": 0.6, + "learning_rate": 1.4621690430259706e-07, + "loss": 0.1267, + "step": 11477 + }, + { + "epoch": 0.6, + "learning_rate": 1.461843390889443e-07, + "loss": 0.1603, + "step": 11478 + }, + { + "epoch": 0.6, + "learning_rate": 1.4615177541336386e-07, + "loss": 0.1236, + "step": 11479 + }, + { + "epoch": 0.6, + "learning_rate": 1.4611921327678643e-07, + "loss": 0.171, + "step": 11480 + }, + { + "epoch": 0.6, + "learning_rate": 1.4608665268014272e-07, + "loss": 0.1027, + "step": 11481 + }, + { + "epoch": 0.6, + "learning_rate": 1.4605409362436327e-07, + "loss": 0.13, + "step": 11482 + }, + { + "epoch": 0.6, + "learning_rate": 1.4602153611037866e-07, + "loss": 0.1046, + "step": 11483 + }, + { + "epoch": 0.6, + "learning_rate": 1.4598898013911934e-07, + "loss": 0.1231, + "step": 11484 + }, + { + "epoch": 0.6, + "learning_rate": 1.459564257115159e-07, + "loss": 0.1192, + "step": 11485 + }, + { + "epoch": 0.6, + "learning_rate": 1.459238728284987e-07, + "loss": 0.13, + "step": 11486 + }, + { + "epoch": 0.6, + "learning_rate": 1.45891321490998e-07, + "loss": 0.1201, + "step": 11487 + }, + { + "epoch": 0.6, + "learning_rate": 1.4585877169994425e-07, + "loss": 0.1496, + "step": 11488 + }, + { + "epoch": 0.6, + "learning_rate": 1.458262234562677e-07, + "loss": 0.126, + "step": 11489 + }, + { + "epoch": 0.6, + "learning_rate": 1.4579367676089862e-07, + "loss": 0.1645, + "step": 11490 + }, + { + "epoch": 0.6, + "learning_rate": 1.4576113161476712e-07, + "loss": 0.1375, + "step": 11491 + }, + { + "epoch": 0.6, + "learning_rate": 1.4572858801880348e-07, + "loss": 0.136, + "step": 11492 + }, + { + "epoch": 0.6, + "learning_rate": 1.456960459739377e-07, + "loss": 0.1746, + "step": 11493 + }, + { + "epoch": 0.6, + "learning_rate": 1.4566350548109995e-07, + "loss": 0.1261, + "step": 11494 + }, + { + "epoch": 0.6, + "learning_rate": 1.456309665412202e-07, + "loss": 0.1256, + "step": 11495 + }, + { + "epoch": 0.6, + "learning_rate": 1.4559842915522834e-07, + "loss": 0.1285, + "step": 11496 + }, + { + "epoch": 0.6, + "learning_rate": 1.4556589332405443e-07, + "loss": 0.1342, + "step": 11497 + }, + { + "epoch": 0.6, + "learning_rate": 1.4553335904862826e-07, + "loss": 0.1502, + "step": 11498 + }, + { + "epoch": 0.6, + "learning_rate": 1.455008263298797e-07, + "loss": 0.1614, + "step": 11499 + }, + { + "epoch": 0.6, + "learning_rate": 1.4546829516873863e-07, + "loss": 0.0951, + "step": 11500 + }, + { + "epoch": 0.6, + "learning_rate": 1.454357655661347e-07, + "loss": 0.1297, + "step": 11501 + }, + { + "epoch": 0.6, + "learning_rate": 1.4540323752299762e-07, + "loss": 0.1335, + "step": 11502 + }, + { + "epoch": 0.6, + "learning_rate": 1.453707110402572e-07, + "loss": 0.1178, + "step": 11503 + }, + { + "epoch": 0.6, + "learning_rate": 1.4533818611884287e-07, + "loss": 0.1467, + "step": 11504 + }, + { + "epoch": 0.6, + "learning_rate": 1.4530566275968425e-07, + "loss": 0.1394, + "step": 11505 + }, + { + "epoch": 0.6, + "learning_rate": 1.4527314096371095e-07, + "loss": 0.1467, + "step": 11506 + }, + { + "epoch": 0.6, + "learning_rate": 1.4524062073185237e-07, + "loss": 0.111, + "step": 11507 + }, + { + "epoch": 0.6, + "learning_rate": 1.45208102065038e-07, + "loss": 0.1207, + "step": 11508 + }, + { + "epoch": 0.6, + "learning_rate": 1.4517558496419725e-07, + "loss": 0.1413, + "step": 11509 + }, + { + "epoch": 0.6, + "learning_rate": 1.4514306943025944e-07, + "loss": 0.1697, + "step": 11510 + }, + { + "epoch": 0.6, + "learning_rate": 1.4511055546415384e-07, + "loss": 0.1437, + "step": 11511 + }, + { + "epoch": 0.6, + "learning_rate": 1.4507804306680985e-07, + "loss": 0.1362, + "step": 11512 + }, + { + "epoch": 0.6, + "learning_rate": 1.4504553223915656e-07, + "loss": 0.1594, + "step": 11513 + }, + { + "epoch": 0.6, + "learning_rate": 1.450130229821231e-07, + "loss": 0.1543, + "step": 11514 + }, + { + "epoch": 0.6, + "learning_rate": 1.4498051529663875e-07, + "loss": 0.1281, + "step": 11515 + }, + { + "epoch": 0.6, + "learning_rate": 1.4494800918363247e-07, + "loss": 0.1402, + "step": 11516 + }, + { + "epoch": 0.6, + "learning_rate": 1.4491550464403335e-07, + "loss": 0.1454, + "step": 11517 + }, + { + "epoch": 0.6, + "learning_rate": 1.4488300167877032e-07, + "loss": 0.1038, + "step": 11518 + }, + { + "epoch": 0.6, + "learning_rate": 1.4485050028877245e-07, + "loss": 0.1494, + "step": 11519 + }, + { + "epoch": 0.6, + "learning_rate": 1.4481800047496853e-07, + "loss": 0.1192, + "step": 11520 + }, + { + "epoch": 0.6, + "learning_rate": 1.4478550223828755e-07, + "loss": 0.1129, + "step": 11521 + }, + { + "epoch": 0.6, + "learning_rate": 1.4475300557965815e-07, + "loss": 0.165, + "step": 11522 + }, + { + "epoch": 0.6, + "learning_rate": 1.4472051050000917e-07, + "loss": 0.1179, + "step": 11523 + }, + { + "epoch": 0.6, + "learning_rate": 1.4468801700026937e-07, + "loss": 0.1489, + "step": 11524 + }, + { + "epoch": 0.6, + "learning_rate": 1.4465552508136735e-07, + "loss": 0.1262, + "step": 11525 + }, + { + "epoch": 0.6, + "learning_rate": 1.4462303474423184e-07, + "loss": 0.1487, + "step": 11526 + }, + { + "epoch": 0.6, + "learning_rate": 1.4459054598979132e-07, + "loss": 0.1981, + "step": 11527 + }, + { + "epoch": 0.6, + "learning_rate": 1.4455805881897444e-07, + "loss": 0.1112, + "step": 11528 + }, + { + "epoch": 0.6, + "learning_rate": 1.445255732327097e-07, + "loss": 0.1745, + "step": 11529 + }, + { + "epoch": 0.6, + "learning_rate": 1.4449308923192543e-07, + "loss": 0.1488, + "step": 11530 + }, + { + "epoch": 0.6, + "learning_rate": 1.4446060681755005e-07, + "loss": 0.1535, + "step": 11531 + }, + { + "epoch": 0.6, + "learning_rate": 1.4442812599051202e-07, + "loss": 0.1096, + "step": 11532 + }, + { + "epoch": 0.6, + "learning_rate": 1.4439564675173957e-07, + "loss": 0.1288, + "step": 11533 + }, + { + "epoch": 0.6, + "learning_rate": 1.4436316910216105e-07, + "loss": 0.1714, + "step": 11534 + }, + { + "epoch": 0.6, + "learning_rate": 1.4433069304270456e-07, + "loss": 0.14, + "step": 11535 + }, + { + "epoch": 0.6, + "learning_rate": 1.442982185742984e-07, + "loss": 0.1358, + "step": 11536 + }, + { + "epoch": 0.6, + "learning_rate": 1.4426574569787063e-07, + "loss": 0.0884, + "step": 11537 + }, + { + "epoch": 0.6, + "learning_rate": 1.442332744143494e-07, + "loss": 0.128, + "step": 11538 + }, + { + "epoch": 0.6, + "learning_rate": 1.4420080472466274e-07, + "loss": 0.113, + "step": 11539 + }, + { + "epoch": 0.6, + "learning_rate": 1.4416833662973855e-07, + "loss": 0.1247, + "step": 11540 + }, + { + "epoch": 0.6, + "learning_rate": 1.4413587013050493e-07, + "loss": 0.1431, + "step": 11541 + }, + { + "epoch": 0.6, + "learning_rate": 1.4410340522788963e-07, + "loss": 0.1431, + "step": 11542 + }, + { + "epoch": 0.6, + "learning_rate": 1.4407094192282063e-07, + "loss": 0.1356, + "step": 11543 + }, + { + "epoch": 0.6, + "learning_rate": 1.4403848021622565e-07, + "loss": 0.1213, + "step": 11544 + }, + { + "epoch": 0.6, + "learning_rate": 1.4400602010903256e-07, + "loss": 0.1531, + "step": 11545 + }, + { + "epoch": 0.6, + "learning_rate": 1.4397356160216907e-07, + "loss": 0.1029, + "step": 11546 + }, + { + "epoch": 0.6, + "learning_rate": 1.4394110469656274e-07, + "loss": 0.1568, + "step": 11547 + }, + { + "epoch": 0.6, + "learning_rate": 1.439086493931413e-07, + "loss": 0.1314, + "step": 11548 + }, + { + "epoch": 0.6, + "learning_rate": 1.4387619569283232e-07, + "loss": 0.1371, + "step": 11549 + }, + { + "epoch": 0.6, + "learning_rate": 1.4384374359656334e-07, + "loss": 0.1523, + "step": 11550 + }, + { + "epoch": 0.6, + "learning_rate": 1.4381129310526184e-07, + "loss": 0.1154, + "step": 11551 + }, + { + "epoch": 0.6, + "learning_rate": 1.4377884421985526e-07, + "loss": 0.1981, + "step": 11552 + }, + { + "epoch": 0.6, + "learning_rate": 1.4374639694127103e-07, + "loss": 0.1415, + "step": 11553 + }, + { + "epoch": 0.6, + "learning_rate": 1.437139512704365e-07, + "loss": 0.1296, + "step": 11554 + }, + { + "epoch": 0.6, + "learning_rate": 1.43681507208279e-07, + "loss": 0.1158, + "step": 11555 + }, + { + "epoch": 0.6, + "learning_rate": 1.4364906475572572e-07, + "loss": 0.1291, + "step": 11556 + }, + { + "epoch": 0.6, + "learning_rate": 1.436166239137039e-07, + "loss": 0.108, + "step": 11557 + }, + { + "epoch": 0.6, + "learning_rate": 1.4358418468314076e-07, + "loss": 0.1143, + "step": 11558 + }, + { + "epoch": 0.6, + "learning_rate": 1.4355174706496343e-07, + "loss": 0.1542, + "step": 11559 + }, + { + "epoch": 0.6, + "learning_rate": 1.4351931106009889e-07, + "loss": 0.1129, + "step": 11560 + }, + { + "epoch": 0.6, + "learning_rate": 1.434868766694743e-07, + "loss": 0.1667, + "step": 11561 + }, + { + "epoch": 0.6, + "learning_rate": 1.4345444389401654e-07, + "loss": 0.1676, + "step": 11562 + }, + { + "epoch": 0.6, + "learning_rate": 1.4342201273465273e-07, + "loss": 0.1396, + "step": 11563 + }, + { + "epoch": 0.6, + "learning_rate": 1.4338958319230952e-07, + "loss": 0.149, + "step": 11564 + }, + { + "epoch": 0.6, + "learning_rate": 1.433571552679139e-07, + "loss": 0.1631, + "step": 11565 + }, + { + "epoch": 0.6, + "learning_rate": 1.4332472896239268e-07, + "loss": 0.129, + "step": 11566 + }, + { + "epoch": 0.6, + "learning_rate": 1.4329230427667254e-07, + "loss": 0.148, + "step": 11567 + }, + { + "epoch": 0.6, + "learning_rate": 1.4325988121168028e-07, + "loss": 0.1671, + "step": 11568 + }, + { + "epoch": 0.6, + "learning_rate": 1.4322745976834246e-07, + "loss": 0.1322, + "step": 11569 + }, + { + "epoch": 0.6, + "learning_rate": 1.4319503994758578e-07, + "loss": 0.1393, + "step": 11570 + }, + { + "epoch": 0.6, + "learning_rate": 1.4316262175033678e-07, + "loss": 0.1263, + "step": 11571 + }, + { + "epoch": 0.6, + "learning_rate": 1.4313020517752205e-07, + "loss": 0.126, + "step": 11572 + }, + { + "epoch": 0.6, + "learning_rate": 1.4309779023006797e-07, + "loss": 0.1843, + "step": 11573 + }, + { + "epoch": 0.6, + "learning_rate": 1.43065376908901e-07, + "loss": 0.1367, + "step": 11574 + }, + { + "epoch": 0.6, + "learning_rate": 1.4303296521494757e-07, + "loss": 0.1995, + "step": 11575 + }, + { + "epoch": 0.6, + "learning_rate": 1.430005551491339e-07, + "loss": 0.1211, + "step": 11576 + }, + { + "epoch": 0.6, + "learning_rate": 1.4296814671238649e-07, + "loss": 0.1492, + "step": 11577 + }, + { + "epoch": 0.6, + "learning_rate": 1.4293573990563136e-07, + "loss": 0.1332, + "step": 11578 + }, + { + "epoch": 0.6, + "learning_rate": 1.4290333472979487e-07, + "loss": 0.1392, + "step": 11579 + }, + { + "epoch": 0.6, + "learning_rate": 1.4287093118580308e-07, + "loss": 0.1294, + "step": 11580 + }, + { + "epoch": 0.6, + "learning_rate": 1.428385292745822e-07, + "loss": 0.1249, + "step": 11581 + }, + { + "epoch": 0.6, + "learning_rate": 1.4280612899705818e-07, + "loss": 0.1924, + "step": 11582 + }, + { + "epoch": 0.6, + "learning_rate": 1.4277373035415705e-07, + "loss": 0.1529, + "step": 11583 + }, + { + "epoch": 0.6, + "learning_rate": 1.4274133334680481e-07, + "loss": 0.1305, + "step": 11584 + }, + { + "epoch": 0.6, + "learning_rate": 1.4270893797592735e-07, + "loss": 0.1831, + "step": 11585 + }, + { + "epoch": 0.6, + "learning_rate": 1.426765442424506e-07, + "loss": 0.1232, + "step": 11586 + }, + { + "epoch": 0.6, + "learning_rate": 1.426441521473003e-07, + "loss": 0.1253, + "step": 11587 + }, + { + "epoch": 0.6, + "learning_rate": 1.426117616914023e-07, + "loss": 0.1277, + "step": 11588 + }, + { + "epoch": 0.6, + "learning_rate": 1.4257937287568235e-07, + "loss": 0.0964, + "step": 11589 + }, + { + "epoch": 0.6, + "learning_rate": 1.4254698570106602e-07, + "loss": 0.1425, + "step": 11590 + }, + { + "epoch": 0.6, + "learning_rate": 1.4251460016847904e-07, + "loss": 0.138, + "step": 11591 + }, + { + "epoch": 0.61, + "learning_rate": 1.4248221627884696e-07, + "loss": 0.126, + "step": 11592 + }, + { + "epoch": 0.61, + "learning_rate": 1.424498340330954e-07, + "loss": 0.1199, + "step": 11593 + }, + { + "epoch": 0.61, + "learning_rate": 1.424174534321497e-07, + "loss": 0.1867, + "step": 11594 + }, + { + "epoch": 0.61, + "learning_rate": 1.423850744769355e-07, + "loss": 0.1203, + "step": 11595 + }, + { + "epoch": 0.61, + "learning_rate": 1.4235269716837804e-07, + "loss": 0.1298, + "step": 11596 + }, + { + "epoch": 0.61, + "learning_rate": 1.4232032150740282e-07, + "loss": 0.1259, + "step": 11597 + }, + { + "epoch": 0.61, + "learning_rate": 1.422879474949351e-07, + "loss": 0.1101, + "step": 11598 + }, + { + "epoch": 0.61, + "learning_rate": 1.4225557513190004e-07, + "loss": 0.1214, + "step": 11599 + }, + { + "epoch": 0.61, + "learning_rate": 1.42223204419223e-07, + "loss": 0.129, + "step": 11600 + }, + { + "epoch": 0.61, + "learning_rate": 1.42190835357829e-07, + "loss": 0.1092, + "step": 11601 + }, + { + "epoch": 0.61, + "learning_rate": 1.421584679486433e-07, + "loss": 0.0998, + "step": 11602 + }, + { + "epoch": 0.61, + "learning_rate": 1.421261021925909e-07, + "loss": 0.1389, + "step": 11603 + }, + { + "epoch": 0.61, + "learning_rate": 1.4209373809059685e-07, + "loss": 0.136, + "step": 11604 + }, + { + "epoch": 0.61, + "learning_rate": 1.420613756435861e-07, + "loss": 0.1081, + "step": 11605 + }, + { + "epoch": 0.61, + "learning_rate": 1.4202901485248364e-07, + "loss": 0.194, + "step": 11606 + }, + { + "epoch": 0.61, + "learning_rate": 1.4199665571821431e-07, + "loss": 0.1375, + "step": 11607 + }, + { + "epoch": 0.61, + "learning_rate": 1.419642982417029e-07, + "loss": 0.1488, + "step": 11608 + }, + { + "epoch": 0.61, + "learning_rate": 1.419319424238743e-07, + "loss": 0.1273, + "step": 11609 + }, + { + "epoch": 0.61, + "learning_rate": 1.4189958826565315e-07, + "loss": 0.1597, + "step": 11610 + }, + { + "epoch": 0.61, + "learning_rate": 1.4186723576796424e-07, + "loss": 0.1175, + "step": 11611 + }, + { + "epoch": 0.61, + "learning_rate": 1.4183488493173211e-07, + "loss": 0.1346, + "step": 11612 + }, + { + "epoch": 0.61, + "learning_rate": 1.418025357578815e-07, + "loss": 0.0944, + "step": 11613 + }, + { + "epoch": 0.61, + "learning_rate": 1.4177018824733688e-07, + "loss": 0.1701, + "step": 11614 + }, + { + "epoch": 0.61, + "learning_rate": 1.4173784240102275e-07, + "loss": 0.126, + "step": 11615 + }, + { + "epoch": 0.61, + "learning_rate": 1.417054982198636e-07, + "loss": 0.148, + "step": 11616 + }, + { + "epoch": 0.61, + "learning_rate": 1.4167315570478375e-07, + "loss": 0.1492, + "step": 11617 + }, + { + "epoch": 0.61, + "learning_rate": 1.4164081485670766e-07, + "loss": 0.1423, + "step": 11618 + }, + { + "epoch": 0.61, + "learning_rate": 1.4160847567655958e-07, + "loss": 0.1537, + "step": 11619 + }, + { + "epoch": 0.61, + "learning_rate": 1.4157613816526385e-07, + "loss": 0.1554, + "step": 11620 + }, + { + "epoch": 0.61, + "learning_rate": 1.415438023237446e-07, + "loss": 0.1439, + "step": 11621 + }, + { + "epoch": 0.61, + "learning_rate": 1.415114681529261e-07, + "loss": 0.1649, + "step": 11622 + }, + { + "epoch": 0.61, + "learning_rate": 1.414791356537324e-07, + "loss": 0.1257, + "step": 11623 + }, + { + "epoch": 0.61, + "learning_rate": 1.4144680482708758e-07, + "loss": 0.1437, + "step": 11624 + }, + { + "epoch": 0.61, + "learning_rate": 1.4141447567391572e-07, + "loss": 0.1292, + "step": 11625 + }, + { + "epoch": 0.61, + "learning_rate": 1.413821481951407e-07, + "loss": 0.1311, + "step": 11626 + }, + { + "epoch": 0.61, + "learning_rate": 1.4134982239168656e-07, + "loss": 0.108, + "step": 11627 + }, + { + "epoch": 0.61, + "learning_rate": 1.413174982644771e-07, + "loss": 0.1121, + "step": 11628 + }, + { + "epoch": 0.61, + "learning_rate": 1.4128517581443622e-07, + "loss": 0.1713, + "step": 11629 + }, + { + "epoch": 0.61, + "learning_rate": 1.4125285504248763e-07, + "loss": 0.1407, + "step": 11630 + }, + { + "epoch": 0.61, + "learning_rate": 1.412205359495552e-07, + "loss": 0.1386, + "step": 11631 + }, + { + "epoch": 0.61, + "learning_rate": 1.4118821853656255e-07, + "loss": 0.098, + "step": 11632 + }, + { + "epoch": 0.61, + "learning_rate": 1.411559028044333e-07, + "loss": 0.1234, + "step": 11633 + }, + { + "epoch": 0.61, + "learning_rate": 1.4112358875409101e-07, + "loss": 0.094, + "step": 11634 + }, + { + "epoch": 0.61, + "learning_rate": 1.410912763864593e-07, + "loss": 0.1311, + "step": 11635 + }, + { + "epoch": 0.61, + "learning_rate": 1.4105896570246168e-07, + "loss": 0.1603, + "step": 11636 + }, + { + "epoch": 0.61, + "learning_rate": 1.4102665670302152e-07, + "loss": 0.1323, + "step": 11637 + }, + { + "epoch": 0.61, + "learning_rate": 1.4099434938906232e-07, + "loss": 0.1209, + "step": 11638 + }, + { + "epoch": 0.61, + "learning_rate": 1.4096204376150738e-07, + "loss": 0.1553, + "step": 11639 + }, + { + "epoch": 0.61, + "learning_rate": 1.4092973982128014e-07, + "loss": 0.1386, + "step": 11640 + }, + { + "epoch": 0.61, + "learning_rate": 1.408974375693036e-07, + "loss": 0.1884, + "step": 11641 + }, + { + "epoch": 0.61, + "learning_rate": 1.4086513700650114e-07, + "loss": 0.1372, + "step": 11642 + }, + { + "epoch": 0.61, + "learning_rate": 1.4083283813379584e-07, + "loss": 0.1323, + "step": 11643 + }, + { + "epoch": 0.61, + "learning_rate": 1.4080054095211095e-07, + "loss": 0.1731, + "step": 11644 + }, + { + "epoch": 0.61, + "learning_rate": 1.407682454623694e-07, + "loss": 0.1218, + "step": 11645 + }, + { + "epoch": 0.61, + "learning_rate": 1.407359516654943e-07, + "loss": 0.1314, + "step": 11646 + }, + { + "epoch": 0.61, + "learning_rate": 1.4070365956240854e-07, + "loss": 0.1368, + "step": 11647 + }, + { + "epoch": 0.61, + "learning_rate": 1.406713691540351e-07, + "loss": 0.1509, + "step": 11648 + }, + { + "epoch": 0.61, + "learning_rate": 1.406390804412969e-07, + "loss": 0.0868, + "step": 11649 + }, + { + "epoch": 0.61, + "learning_rate": 1.406067934251166e-07, + "loss": 0.1484, + "step": 11650 + }, + { + "epoch": 0.61, + "learning_rate": 1.4057450810641713e-07, + "loss": 0.1287, + "step": 11651 + }, + { + "epoch": 0.61, + "learning_rate": 1.405422244861211e-07, + "loss": 0.1534, + "step": 11652 + }, + { + "epoch": 0.61, + "learning_rate": 1.405099425651513e-07, + "loss": 0.1477, + "step": 11653 + }, + { + "epoch": 0.61, + "learning_rate": 1.4047766234443024e-07, + "loss": 0.134, + "step": 11654 + }, + { + "epoch": 0.61, + "learning_rate": 1.4044538382488065e-07, + "loss": 0.1319, + "step": 11655 + }, + { + "epoch": 0.61, + "learning_rate": 1.404131070074249e-07, + "loss": 0.1282, + "step": 11656 + }, + { + "epoch": 0.61, + "learning_rate": 1.403808318929856e-07, + "loss": 0.1323, + "step": 11657 + }, + { + "epoch": 0.61, + "learning_rate": 1.4034855848248523e-07, + "loss": 0.1114, + "step": 11658 + }, + { + "epoch": 0.61, + "learning_rate": 1.4031628677684597e-07, + "loss": 0.1685, + "step": 11659 + }, + { + "epoch": 0.61, + "learning_rate": 1.402840167769903e-07, + "loss": 0.1615, + "step": 11660 + }, + { + "epoch": 0.61, + "learning_rate": 1.4025174848384046e-07, + "loss": 0.1251, + "step": 11661 + }, + { + "epoch": 0.61, + "learning_rate": 1.4021948189831877e-07, + "loss": 0.1107, + "step": 11662 + }, + { + "epoch": 0.61, + "learning_rate": 1.4018721702134727e-07, + "loss": 0.1745, + "step": 11663 + }, + { + "epoch": 0.61, + "learning_rate": 1.4015495385384827e-07, + "loss": 0.1674, + "step": 11664 + }, + { + "epoch": 0.61, + "learning_rate": 1.4012269239674376e-07, + "loss": 0.127, + "step": 11665 + }, + { + "epoch": 0.61, + "learning_rate": 1.400904326509559e-07, + "loss": 0.1019, + "step": 11666 + }, + { + "epoch": 0.61, + "learning_rate": 1.4005817461740657e-07, + "loss": 0.1481, + "step": 11667 + }, + { + "epoch": 0.61, + "learning_rate": 1.400259182970177e-07, + "loss": 0.1109, + "step": 11668 + }, + { + "epoch": 0.61, + "learning_rate": 1.3999366369071127e-07, + "loss": 0.1247, + "step": 11669 + }, + { + "epoch": 0.61, + "learning_rate": 1.3996141079940905e-07, + "loss": 0.1441, + "step": 11670 + }, + { + "epoch": 0.61, + "learning_rate": 1.3992915962403294e-07, + "loss": 0.1349, + "step": 11671 + }, + { + "epoch": 0.61, + "learning_rate": 1.398969101655046e-07, + "loss": 0.1305, + "step": 11672 + }, + { + "epoch": 0.61, + "learning_rate": 1.3986466242474576e-07, + "loss": 0.1192, + "step": 11673 + }, + { + "epoch": 0.61, + "learning_rate": 1.3983241640267812e-07, + "loss": 0.1426, + "step": 11674 + }, + { + "epoch": 0.61, + "learning_rate": 1.3980017210022328e-07, + "loss": 0.1219, + "step": 11675 + }, + { + "epoch": 0.61, + "learning_rate": 1.397679295183027e-07, + "loss": 0.1193, + "step": 11676 + }, + { + "epoch": 0.61, + "learning_rate": 1.3973568865783796e-07, + "loss": 0.1285, + "step": 11677 + }, + { + "epoch": 0.61, + "learning_rate": 1.3970344951975054e-07, + "loss": 0.0978, + "step": 11678 + }, + { + "epoch": 0.61, + "learning_rate": 1.3967121210496174e-07, + "loss": 0.1496, + "step": 11679 + }, + { + "epoch": 0.61, + "learning_rate": 1.3963897641439302e-07, + "loss": 0.0932, + "step": 11680 + }, + { + "epoch": 0.61, + "learning_rate": 1.396067424489656e-07, + "loss": 0.1146, + "step": 11681 + }, + { + "epoch": 0.61, + "learning_rate": 1.395745102096009e-07, + "loss": 0.1415, + "step": 11682 + }, + { + "epoch": 0.61, + "learning_rate": 1.3954227969722e-07, + "loss": 0.1045, + "step": 11683 + }, + { + "epoch": 0.61, + "learning_rate": 1.3951005091274405e-07, + "loss": 0.1403, + "step": 11684 + }, + { + "epoch": 0.61, + "learning_rate": 1.3947782385709424e-07, + "loss": 0.1686, + "step": 11685 + }, + { + "epoch": 0.61, + "learning_rate": 1.3944559853119153e-07, + "loss": 0.1311, + "step": 11686 + }, + { + "epoch": 0.61, + "learning_rate": 1.3941337493595705e-07, + "loss": 0.1221, + "step": 11687 + }, + { + "epoch": 0.61, + "learning_rate": 1.3938115307231162e-07, + "loss": 0.1331, + "step": 11688 + }, + { + "epoch": 0.61, + "learning_rate": 1.393489329411763e-07, + "loss": 0.1725, + "step": 11689 + }, + { + "epoch": 0.61, + "learning_rate": 1.3931671454347185e-07, + "loss": 0.1376, + "step": 11690 + }, + { + "epoch": 0.61, + "learning_rate": 1.392844978801192e-07, + "loss": 0.1439, + "step": 11691 + }, + { + "epoch": 0.61, + "learning_rate": 1.39252282952039e-07, + "loss": 0.1086, + "step": 11692 + }, + { + "epoch": 0.61, + "learning_rate": 1.39220069760152e-07, + "loss": 0.1385, + "step": 11693 + }, + { + "epoch": 0.61, + "learning_rate": 1.3918785830537886e-07, + "loss": 0.0987, + "step": 11694 + }, + { + "epoch": 0.61, + "learning_rate": 1.3915564858864017e-07, + "loss": 0.1413, + "step": 11695 + }, + { + "epoch": 0.61, + "learning_rate": 1.391234406108566e-07, + "loss": 0.1179, + "step": 11696 + }, + { + "epoch": 0.61, + "learning_rate": 1.3909123437294854e-07, + "loss": 0.1582, + "step": 11697 + }, + { + "epoch": 0.61, + "learning_rate": 1.3905902987583654e-07, + "loss": 0.1109, + "step": 11698 + }, + { + "epoch": 0.61, + "learning_rate": 1.39026827120441e-07, + "loss": 0.1451, + "step": 11699 + }, + { + "epoch": 0.61, + "learning_rate": 1.3899462610768232e-07, + "loss": 0.1273, + "step": 11700 + }, + { + "epoch": 0.61, + "learning_rate": 1.3896242683848075e-07, + "loss": 0.1442, + "step": 11701 + }, + { + "epoch": 0.61, + "learning_rate": 1.3893022931375654e-07, + "loss": 0.1157, + "step": 11702 + }, + { + "epoch": 0.61, + "learning_rate": 1.3889803353443e-07, + "loss": 0.1238, + "step": 11703 + }, + { + "epoch": 0.61, + "learning_rate": 1.3886583950142118e-07, + "loss": 0.1231, + "step": 11704 + }, + { + "epoch": 0.61, + "learning_rate": 1.3883364721565033e-07, + "loss": 0.0829, + "step": 11705 + }, + { + "epoch": 0.61, + "learning_rate": 1.3880145667803742e-07, + "loss": 0.1619, + "step": 11706 + }, + { + "epoch": 0.61, + "learning_rate": 1.3876926788950255e-07, + "loss": 0.1069, + "step": 11707 + }, + { + "epoch": 0.61, + "learning_rate": 1.3873708085096557e-07, + "loss": 0.1251, + "step": 11708 + }, + { + "epoch": 0.61, + "learning_rate": 1.3870489556334657e-07, + "loss": 0.1229, + "step": 11709 + }, + { + "epoch": 0.61, + "learning_rate": 1.3867271202756526e-07, + "loss": 0.1572, + "step": 11710 + }, + { + "epoch": 0.61, + "learning_rate": 1.3864053024454148e-07, + "loss": 0.1458, + "step": 11711 + }, + { + "epoch": 0.61, + "learning_rate": 1.386083502151951e-07, + "loss": 0.1093, + "step": 11712 + }, + { + "epoch": 0.61, + "learning_rate": 1.3857617194044572e-07, + "loss": 0.1281, + "step": 11713 + }, + { + "epoch": 0.61, + "learning_rate": 1.3854399542121305e-07, + "loss": 0.1448, + "step": 11714 + }, + { + "epoch": 0.61, + "learning_rate": 1.3851182065841673e-07, + "loss": 0.146, + "step": 11715 + }, + { + "epoch": 0.61, + "learning_rate": 1.3847964765297632e-07, + "loss": 0.1462, + "step": 11716 + }, + { + "epoch": 0.61, + "learning_rate": 1.3844747640581132e-07, + "loss": 0.1366, + "step": 11717 + }, + { + "epoch": 0.61, + "learning_rate": 1.3841530691784126e-07, + "loss": 0.1192, + "step": 11718 + }, + { + "epoch": 0.61, + "learning_rate": 1.383831391899855e-07, + "loss": 0.1514, + "step": 11719 + }, + { + "epoch": 0.61, + "learning_rate": 1.3835097322316332e-07, + "loss": 0.1204, + "step": 11720 + }, + { + "epoch": 0.61, + "learning_rate": 1.383188090182942e-07, + "loss": 0.1401, + "step": 11721 + }, + { + "epoch": 0.61, + "learning_rate": 1.382866465762973e-07, + "loss": 0.1219, + "step": 11722 + }, + { + "epoch": 0.61, + "learning_rate": 1.3825448589809186e-07, + "loss": 0.1544, + "step": 11723 + }, + { + "epoch": 0.61, + "learning_rate": 1.3822232698459702e-07, + "loss": 0.1635, + "step": 11724 + }, + { + "epoch": 0.61, + "learning_rate": 1.38190169836732e-07, + "loss": 0.1302, + "step": 11725 + }, + { + "epoch": 0.61, + "learning_rate": 1.381580144554158e-07, + "loss": 0.115, + "step": 11726 + }, + { + "epoch": 0.61, + "learning_rate": 1.3812586084156733e-07, + "loss": 0.1412, + "step": 11727 + }, + { + "epoch": 0.61, + "learning_rate": 1.380937089961057e-07, + "loss": 0.1012, + "step": 11728 + }, + { + "epoch": 0.61, + "learning_rate": 1.380615589199497e-07, + "loss": 0.1122, + "step": 11729 + }, + { + "epoch": 0.61, + "learning_rate": 1.3802941061401834e-07, + "loss": 0.1564, + "step": 11730 + }, + { + "epoch": 0.61, + "learning_rate": 1.3799726407923026e-07, + "loss": 0.1546, + "step": 11731 + }, + { + "epoch": 0.61, + "learning_rate": 1.3796511931650438e-07, + "loss": 0.1235, + "step": 11732 + }, + { + "epoch": 0.61, + "learning_rate": 1.3793297632675928e-07, + "loss": 0.0999, + "step": 11733 + }, + { + "epoch": 0.61, + "learning_rate": 1.3790083511091375e-07, + "loss": 0.1134, + "step": 11734 + }, + { + "epoch": 0.61, + "learning_rate": 1.3786869566988634e-07, + "loss": 0.1603, + "step": 11735 + }, + { + "epoch": 0.61, + "learning_rate": 1.378365580045955e-07, + "loss": 0.1892, + "step": 11736 + }, + { + "epoch": 0.61, + "learning_rate": 1.3780442211595987e-07, + "loss": 0.1762, + "step": 11737 + }, + { + "epoch": 0.61, + "learning_rate": 1.3777228800489785e-07, + "loss": 0.1429, + "step": 11738 + }, + { + "epoch": 0.61, + "learning_rate": 1.377401556723279e-07, + "loss": 0.1469, + "step": 11739 + }, + { + "epoch": 0.61, + "learning_rate": 1.377080251191683e-07, + "loss": 0.1783, + "step": 11740 + }, + { + "epoch": 0.61, + "learning_rate": 1.3767589634633741e-07, + "loss": 0.124, + "step": 11741 + }, + { + "epoch": 0.61, + "learning_rate": 1.376437693547534e-07, + "loss": 0.1998, + "step": 11742 + }, + { + "epoch": 0.61, + "learning_rate": 1.3761164414533471e-07, + "loss": 0.1177, + "step": 11743 + }, + { + "epoch": 0.61, + "learning_rate": 1.3757952071899915e-07, + "loss": 0.1554, + "step": 11744 + }, + { + "epoch": 0.61, + "learning_rate": 1.3754739907666503e-07, + "loss": 0.1337, + "step": 11745 + }, + { + "epoch": 0.61, + "learning_rate": 1.3751527921925035e-07, + "loss": 0.1382, + "step": 11746 + }, + { + "epoch": 0.61, + "learning_rate": 1.3748316114767312e-07, + "loss": 0.1283, + "step": 11747 + }, + { + "epoch": 0.61, + "learning_rate": 1.3745104486285124e-07, + "loss": 0.1379, + "step": 11748 + }, + { + "epoch": 0.61, + "learning_rate": 1.3741893036570264e-07, + "loss": 0.1109, + "step": 11749 + }, + { + "epoch": 0.61, + "learning_rate": 1.3738681765714524e-07, + "loss": 0.1415, + "step": 11750 + }, + { + "epoch": 0.61, + "learning_rate": 1.3735470673809667e-07, + "loss": 0.1439, + "step": 11751 + }, + { + "epoch": 0.61, + "learning_rate": 1.3732259760947495e-07, + "loss": 0.1193, + "step": 11752 + }, + { + "epoch": 0.61, + "learning_rate": 1.372904902721974e-07, + "loss": 0.1852, + "step": 11753 + }, + { + "epoch": 0.61, + "learning_rate": 1.3725838472718195e-07, + "loss": 0.137, + "step": 11754 + }, + { + "epoch": 0.61, + "learning_rate": 1.37226280975346e-07, + "loss": 0.1081, + "step": 11755 + }, + { + "epoch": 0.61, + "learning_rate": 1.371941790176072e-07, + "loss": 0.1565, + "step": 11756 + }, + { + "epoch": 0.61, + "learning_rate": 1.37162078854883e-07, + "loss": 0.1162, + "step": 11757 + }, + { + "epoch": 0.61, + "learning_rate": 1.3712998048809087e-07, + "loss": 0.1351, + "step": 11758 + }, + { + "epoch": 0.61, + "learning_rate": 1.3709788391814817e-07, + "loss": 0.1207, + "step": 11759 + }, + { + "epoch": 0.61, + "learning_rate": 1.3706578914597225e-07, + "loss": 0.1173, + "step": 11760 + }, + { + "epoch": 0.61, + "learning_rate": 1.3703369617248038e-07, + "loss": 0.1275, + "step": 11761 + }, + { + "epoch": 0.61, + "learning_rate": 1.370016049985897e-07, + "loss": 0.1163, + "step": 11762 + }, + { + "epoch": 0.61, + "learning_rate": 1.3696951562521752e-07, + "loss": 0.1251, + "step": 11763 + }, + { + "epoch": 0.61, + "learning_rate": 1.369374280532809e-07, + "loss": 0.1521, + "step": 11764 + }, + { + "epoch": 0.61, + "learning_rate": 1.3690534228369695e-07, + "loss": 0.1233, + "step": 11765 + }, + { + "epoch": 0.61, + "learning_rate": 1.3687325831738263e-07, + "loss": 0.1425, + "step": 11766 + }, + { + "epoch": 0.61, + "learning_rate": 1.36841176155255e-07, + "loss": 0.1136, + "step": 11767 + }, + { + "epoch": 0.61, + "learning_rate": 1.368090957982309e-07, + "loss": 0.1069, + "step": 11768 + }, + { + "epoch": 0.61, + "learning_rate": 1.3677701724722733e-07, + "loss": 0.1275, + "step": 11769 + }, + { + "epoch": 0.61, + "learning_rate": 1.3674494050316096e-07, + "loss": 0.1741, + "step": 11770 + }, + { + "epoch": 0.61, + "learning_rate": 1.367128655669486e-07, + "loss": 0.1131, + "step": 11771 + }, + { + "epoch": 0.61, + "learning_rate": 1.3668079243950696e-07, + "loss": 0.1299, + "step": 11772 + }, + { + "epoch": 0.61, + "learning_rate": 1.366487211217527e-07, + "loss": 0.1358, + "step": 11773 + }, + { + "epoch": 0.61, + "learning_rate": 1.3661665161460246e-07, + "loss": 0.1257, + "step": 11774 + }, + { + "epoch": 0.61, + "learning_rate": 1.365845839189728e-07, + "loss": 0.1228, + "step": 11775 + }, + { + "epoch": 0.61, + "learning_rate": 1.3655251803578022e-07, + "loss": 0.1591, + "step": 11776 + }, + { + "epoch": 0.61, + "learning_rate": 1.3652045396594114e-07, + "loss": 0.1594, + "step": 11777 + }, + { + "epoch": 0.61, + "learning_rate": 1.3648839171037208e-07, + "loss": 0.1382, + "step": 11778 + }, + { + "epoch": 0.61, + "learning_rate": 1.3645633126998923e-07, + "loss": 0.1617, + "step": 11779 + }, + { + "epoch": 0.61, + "learning_rate": 1.3642427264570895e-07, + "loss": 0.1602, + "step": 11780 + }, + { + "epoch": 0.61, + "learning_rate": 1.3639221583844753e-07, + "loss": 0.1099, + "step": 11781 + }, + { + "epoch": 0.61, + "learning_rate": 1.3636016084912108e-07, + "loss": 0.1492, + "step": 11782 + }, + { + "epoch": 0.62, + "learning_rate": 1.3632810767864586e-07, + "loss": 0.1333, + "step": 11783 + }, + { + "epoch": 0.62, + "learning_rate": 1.3629605632793787e-07, + "loss": 0.1259, + "step": 11784 + }, + { + "epoch": 0.62, + "learning_rate": 1.362640067979132e-07, + "loss": 0.126, + "step": 11785 + }, + { + "epoch": 0.62, + "learning_rate": 1.3623195908948785e-07, + "loss": 0.1433, + "step": 11786 + }, + { + "epoch": 0.62, + "learning_rate": 1.361999132035777e-07, + "loss": 0.1319, + "step": 11787 + }, + { + "epoch": 0.62, + "learning_rate": 1.3616786914109865e-07, + "loss": 0.1443, + "step": 11788 + }, + { + "epoch": 0.62, + "learning_rate": 1.3613582690296653e-07, + "loss": 0.1322, + "step": 11789 + }, + { + "epoch": 0.62, + "learning_rate": 1.3610378649009718e-07, + "loss": 0.1401, + "step": 11790 + }, + { + "epoch": 0.62, + "learning_rate": 1.3607174790340618e-07, + "loss": 0.146, + "step": 11791 + }, + { + "epoch": 0.62, + "learning_rate": 1.3603971114380942e-07, + "loss": 0.1409, + "step": 11792 + }, + { + "epoch": 0.62, + "learning_rate": 1.360076762122223e-07, + "loss": 0.1046, + "step": 11793 + }, + { + "epoch": 0.62, + "learning_rate": 1.359756431095606e-07, + "loss": 0.1659, + "step": 11794 + }, + { + "epoch": 0.62, + "learning_rate": 1.3594361183673973e-07, + "loss": 0.1769, + "step": 11795 + }, + { + "epoch": 0.62, + "learning_rate": 1.3591158239467508e-07, + "loss": 0.1272, + "step": 11796 + }, + { + "epoch": 0.62, + "learning_rate": 1.358795547842822e-07, + "loss": 0.149, + "step": 11797 + }, + { + "epoch": 0.62, + "learning_rate": 1.3584752900647636e-07, + "loss": 0.1476, + "step": 11798 + }, + { + "epoch": 0.62, + "learning_rate": 1.3581550506217294e-07, + "loss": 0.1849, + "step": 11799 + }, + { + "epoch": 0.62, + "learning_rate": 1.357834829522871e-07, + "loss": 0.1478, + "step": 11800 + }, + { + "epoch": 0.62, + "learning_rate": 1.357514626777342e-07, + "loss": 0.1298, + "step": 11801 + }, + { + "epoch": 0.62, + "learning_rate": 1.3571944423942924e-07, + "loss": 0.1447, + "step": 11802 + }, + { + "epoch": 0.62, + "learning_rate": 1.3568742763828748e-07, + "loss": 0.1285, + "step": 11803 + }, + { + "epoch": 0.62, + "learning_rate": 1.3565541287522378e-07, + "loss": 0.1235, + "step": 11804 + }, + { + "epoch": 0.62, + "learning_rate": 1.3562339995115316e-07, + "loss": 0.1342, + "step": 11805 + }, + { + "epoch": 0.62, + "learning_rate": 1.355913888669907e-07, + "loss": 0.1266, + "step": 11806 + }, + { + "epoch": 0.62, + "learning_rate": 1.3555937962365117e-07, + "loss": 0.133, + "step": 11807 + }, + { + "epoch": 0.62, + "learning_rate": 1.355273722220495e-07, + "loss": 0.1544, + "step": 11808 + }, + { + "epoch": 0.62, + "learning_rate": 1.3549536666310037e-07, + "loss": 0.13, + "step": 11809 + }, + { + "epoch": 0.62, + "learning_rate": 1.354633629477186e-07, + "loss": 0.1244, + "step": 11810 + }, + { + "epoch": 0.62, + "learning_rate": 1.354313610768188e-07, + "loss": 0.1454, + "step": 11811 + }, + { + "epoch": 0.62, + "learning_rate": 1.3539936105131568e-07, + "loss": 0.1133, + "step": 11812 + }, + { + "epoch": 0.62, + "learning_rate": 1.3536736287212374e-07, + "loss": 0.1372, + "step": 11813 + }, + { + "epoch": 0.62, + "learning_rate": 1.3533536654015748e-07, + "loss": 0.1232, + "step": 11814 + }, + { + "epoch": 0.62, + "learning_rate": 1.3530337205633143e-07, + "loss": 0.1379, + "step": 11815 + }, + { + "epoch": 0.62, + "learning_rate": 1.3527137942155995e-07, + "loss": 0.1241, + "step": 11816 + }, + { + "epoch": 0.62, + "learning_rate": 1.3523938863675748e-07, + "loss": 0.158, + "step": 11817 + }, + { + "epoch": 0.62, + "learning_rate": 1.3520739970283823e-07, + "loss": 0.1164, + "step": 11818 + }, + { + "epoch": 0.62, + "learning_rate": 1.3517541262071657e-07, + "loss": 0.167, + "step": 11819 + }, + { + "epoch": 0.62, + "learning_rate": 1.3514342739130666e-07, + "loss": 0.1197, + "step": 11820 + }, + { + "epoch": 0.62, + "learning_rate": 1.351114440155226e-07, + "loss": 0.1299, + "step": 11821 + }, + { + "epoch": 0.62, + "learning_rate": 1.350794624942785e-07, + "loss": 0.1209, + "step": 11822 + }, + { + "epoch": 0.62, + "learning_rate": 1.3504748282848838e-07, + "loss": 0.1224, + "step": 11823 + }, + { + "epoch": 0.62, + "learning_rate": 1.3501550501906637e-07, + "loss": 0.1578, + "step": 11824 + }, + { + "epoch": 0.62, + "learning_rate": 1.3498352906692624e-07, + "loss": 0.1355, + "step": 11825 + }, + { + "epoch": 0.62, + "learning_rate": 1.34951554972982e-07, + "loss": 0.1309, + "step": 11826 + }, + { + "epoch": 0.62, + "learning_rate": 1.3491958273814738e-07, + "loss": 0.1088, + "step": 11827 + }, + { + "epoch": 0.62, + "learning_rate": 1.3488761236333624e-07, + "loss": 0.1465, + "step": 11828 + }, + { + "epoch": 0.62, + "learning_rate": 1.3485564384946235e-07, + "loss": 0.1488, + "step": 11829 + }, + { + "epoch": 0.62, + "learning_rate": 1.3482367719743918e-07, + "loss": 0.1554, + "step": 11830 + }, + { + "epoch": 0.62, + "learning_rate": 1.3479171240818054e-07, + "loss": 0.143, + "step": 11831 + }, + { + "epoch": 0.62, + "learning_rate": 1.3475974948259992e-07, + "loss": 0.1295, + "step": 11832 + }, + { + "epoch": 0.62, + "learning_rate": 1.3472778842161082e-07, + "loss": 0.1218, + "step": 11833 + }, + { + "epoch": 0.62, + "learning_rate": 1.3469582922612674e-07, + "loss": 0.1409, + "step": 11834 + }, + { + "epoch": 0.62, + "learning_rate": 1.346638718970611e-07, + "loss": 0.1938, + "step": 11835 + }, + { + "epoch": 0.62, + "learning_rate": 1.3463191643532715e-07, + "loss": 0.1611, + "step": 11836 + }, + { + "epoch": 0.62, + "learning_rate": 1.3459996284183835e-07, + "loss": 0.1433, + "step": 11837 + }, + { + "epoch": 0.62, + "learning_rate": 1.3456801111750788e-07, + "loss": 0.159, + "step": 11838 + }, + { + "epoch": 0.62, + "learning_rate": 1.3453606126324883e-07, + "loss": 0.144, + "step": 11839 + }, + { + "epoch": 0.62, + "learning_rate": 1.3450411327997447e-07, + "loss": 0.1216, + "step": 11840 + }, + { + "epoch": 0.62, + "learning_rate": 1.3447216716859776e-07, + "loss": 0.1498, + "step": 11841 + }, + { + "epoch": 0.62, + "learning_rate": 1.3444022293003187e-07, + "loss": 0.1231, + "step": 11842 + }, + { + "epoch": 0.62, + "learning_rate": 1.344082805651897e-07, + "loss": 0.1346, + "step": 11843 + }, + { + "epoch": 0.62, + "learning_rate": 1.3437634007498423e-07, + "loss": 0.1637, + "step": 11844 + }, + { + "epoch": 0.62, + "learning_rate": 1.3434440146032823e-07, + "loss": 0.1319, + "step": 11845 + }, + { + "epoch": 0.62, + "learning_rate": 1.3431246472213466e-07, + "loss": 0.149, + "step": 11846 + }, + { + "epoch": 0.62, + "learning_rate": 1.342805298613162e-07, + "loss": 0.1292, + "step": 11847 + }, + { + "epoch": 0.62, + "learning_rate": 1.342485968787855e-07, + "loss": 0.1525, + "step": 11848 + }, + { + "epoch": 0.62, + "learning_rate": 1.3421666577545532e-07, + "loss": 0.1103, + "step": 11849 + }, + { + "epoch": 0.62, + "learning_rate": 1.341847365522382e-07, + "loss": 0.1284, + "step": 11850 + }, + { + "epoch": 0.62, + "learning_rate": 1.3415280921004675e-07, + "loss": 0.1039, + "step": 11851 + }, + { + "epoch": 0.62, + "learning_rate": 1.341208837497934e-07, + "loss": 0.1404, + "step": 11852 + }, + { + "epoch": 0.62, + "learning_rate": 1.3408896017239065e-07, + "loss": 0.1467, + "step": 11853 + }, + { + "epoch": 0.62, + "learning_rate": 1.3405703847875083e-07, + "loss": 0.1551, + "step": 11854 + }, + { + "epoch": 0.62, + "learning_rate": 1.3402511866978642e-07, + "loss": 0.2048, + "step": 11855 + }, + { + "epoch": 0.62, + "learning_rate": 1.3399320074640947e-07, + "loss": 0.1371, + "step": 11856 + }, + { + "epoch": 0.62, + "learning_rate": 1.3396128470953238e-07, + "loss": 0.1755, + "step": 11857 + }, + { + "epoch": 0.62, + "learning_rate": 1.339293705600672e-07, + "loss": 0.1727, + "step": 11858 + }, + { + "epoch": 0.62, + "learning_rate": 1.3389745829892618e-07, + "loss": 0.1646, + "step": 11859 + }, + { + "epoch": 0.62, + "learning_rate": 1.3386554792702127e-07, + "loss": 0.1286, + "step": 11860 + }, + { + "epoch": 0.62, + "learning_rate": 1.3383363944526454e-07, + "loss": 0.1368, + "step": 11861 + }, + { + "epoch": 0.62, + "learning_rate": 1.3380173285456797e-07, + "loss": 0.097, + "step": 11862 + }, + { + "epoch": 0.62, + "learning_rate": 1.3376982815584345e-07, + "loss": 0.1185, + "step": 11863 + }, + { + "epoch": 0.62, + "learning_rate": 1.337379253500028e-07, + "loss": 0.1924, + "step": 11864 + }, + { + "epoch": 0.62, + "learning_rate": 1.3370602443795775e-07, + "loss": 0.1591, + "step": 11865 + }, + { + "epoch": 0.62, + "learning_rate": 1.3367412542062016e-07, + "loss": 0.1147, + "step": 11866 + }, + { + "epoch": 0.62, + "learning_rate": 1.3364222829890168e-07, + "loss": 0.1418, + "step": 11867 + }, + { + "epoch": 0.62, + "learning_rate": 1.3361033307371395e-07, + "loss": 0.1555, + "step": 11868 + }, + { + "epoch": 0.62, + "learning_rate": 1.3357843974596845e-07, + "loss": 0.1503, + "step": 11869 + }, + { + "epoch": 0.62, + "learning_rate": 1.3354654831657687e-07, + "loss": 0.1249, + "step": 11870 + }, + { + "epoch": 0.62, + "learning_rate": 1.335146587864505e-07, + "loss": 0.118, + "step": 11871 + }, + { + "epoch": 0.62, + "learning_rate": 1.3348277115650098e-07, + "loss": 0.129, + "step": 11872 + }, + { + "epoch": 0.62, + "learning_rate": 1.3345088542763946e-07, + "loss": 0.1276, + "step": 11873 + }, + { + "epoch": 0.62, + "learning_rate": 1.3341900160077732e-07, + "loss": 0.1766, + "step": 11874 + }, + { + "epoch": 0.62, + "learning_rate": 1.3338711967682584e-07, + "loss": 0.122, + "step": 11875 + }, + { + "epoch": 0.62, + "learning_rate": 1.3335523965669616e-07, + "loss": 0.106, + "step": 11876 + }, + { + "epoch": 0.62, + "learning_rate": 1.333233615412995e-07, + "loss": 0.1509, + "step": 11877 + }, + { + "epoch": 0.62, + "learning_rate": 1.3329148533154684e-07, + "loss": 0.1103, + "step": 11878 + }, + { + "epoch": 0.62, + "learning_rate": 1.3325961102834935e-07, + "loss": 0.1254, + "step": 11879 + }, + { + "epoch": 0.62, + "learning_rate": 1.3322773863261797e-07, + "loss": 0.1184, + "step": 11880 + }, + { + "epoch": 0.62, + "learning_rate": 1.3319586814526352e-07, + "loss": 0.1185, + "step": 11881 + }, + { + "epoch": 0.62, + "learning_rate": 1.3316399956719696e-07, + "loss": 0.1264, + "step": 11882 + }, + { + "epoch": 0.62, + "learning_rate": 1.3313213289932909e-07, + "loss": 0.1568, + "step": 11883 + }, + { + "epoch": 0.62, + "learning_rate": 1.3310026814257072e-07, + "loss": 0.1482, + "step": 11884 + }, + { + "epoch": 0.62, + "learning_rate": 1.3306840529783245e-07, + "loss": 0.1178, + "step": 11885 + }, + { + "epoch": 0.62, + "learning_rate": 1.3303654436602503e-07, + "loss": 0.1478, + "step": 11886 + }, + { + "epoch": 0.62, + "learning_rate": 1.33004685348059e-07, + "loss": 0.1528, + "step": 11887 + }, + { + "epoch": 0.62, + "learning_rate": 1.3297282824484498e-07, + "loss": 0.1403, + "step": 11888 + }, + { + "epoch": 0.62, + "learning_rate": 1.3294097305729342e-07, + "loss": 0.144, + "step": 11889 + }, + { + "epoch": 0.62, + "learning_rate": 1.3290911978631466e-07, + "loss": 0.1557, + "step": 11890 + }, + { + "epoch": 0.62, + "learning_rate": 1.328772684328192e-07, + "loss": 0.0986, + "step": 11891 + }, + { + "epoch": 0.62, + "learning_rate": 1.3284541899771728e-07, + "loss": 0.1389, + "step": 11892 + }, + { + "epoch": 0.62, + "learning_rate": 1.3281357148191923e-07, + "loss": 0.1175, + "step": 11893 + }, + { + "epoch": 0.62, + "learning_rate": 1.3278172588633524e-07, + "loss": 0.1402, + "step": 11894 + }, + { + "epoch": 0.62, + "learning_rate": 1.327498822118755e-07, + "loss": 0.1227, + "step": 11895 + }, + { + "epoch": 0.62, + "learning_rate": 1.3271804045945004e-07, + "loss": 0.1167, + "step": 11896 + }, + { + "epoch": 0.62, + "learning_rate": 1.32686200629969e-07, + "loss": 0.1531, + "step": 11897 + }, + { + "epoch": 0.62, + "learning_rate": 1.3265436272434235e-07, + "loss": 0.1549, + "step": 11898 + }, + { + "epoch": 0.62, + "learning_rate": 1.3262252674347995e-07, + "loss": 0.1173, + "step": 11899 + }, + { + "epoch": 0.62, + "learning_rate": 1.325906926882918e-07, + "loss": 0.1183, + "step": 11900 + }, + { + "epoch": 0.62, + "learning_rate": 1.3255886055968762e-07, + "loss": 0.1184, + "step": 11901 + }, + { + "epoch": 0.62, + "learning_rate": 1.3252703035857727e-07, + "loss": 0.1575, + "step": 11902 + }, + { + "epoch": 0.62, + "learning_rate": 1.3249520208587041e-07, + "loss": 0.1131, + "step": 11903 + }, + { + "epoch": 0.62, + "learning_rate": 1.3246337574247679e-07, + "loss": 0.1346, + "step": 11904 + }, + { + "epoch": 0.62, + "learning_rate": 1.324315513293059e-07, + "loss": 0.0924, + "step": 11905 + }, + { + "epoch": 0.62, + "learning_rate": 1.3239972884726746e-07, + "loss": 0.133, + "step": 11906 + }, + { + "epoch": 0.62, + "learning_rate": 1.3236790829727081e-07, + "loss": 0.1175, + "step": 11907 + }, + { + "epoch": 0.62, + "learning_rate": 1.323360896802254e-07, + "loss": 0.1352, + "step": 11908 + }, + { + "epoch": 0.62, + "learning_rate": 1.3230427299704073e-07, + "loss": 0.1294, + "step": 11909 + }, + { + "epoch": 0.62, + "learning_rate": 1.32272458248626e-07, + "loss": 0.1535, + "step": 11910 + }, + { + "epoch": 0.62, + "learning_rate": 1.3224064543589063e-07, + "loss": 0.138, + "step": 11911 + }, + { + "epoch": 0.62, + "learning_rate": 1.3220883455974372e-07, + "loss": 0.1493, + "step": 11912 + }, + { + "epoch": 0.62, + "learning_rate": 1.3217702562109455e-07, + "loss": 0.1387, + "step": 11913 + }, + { + "epoch": 0.62, + "learning_rate": 1.321452186208521e-07, + "loss": 0.1213, + "step": 11914 + }, + { + "epoch": 0.62, + "learning_rate": 1.3211341355992557e-07, + "loss": 0.1451, + "step": 11915 + }, + { + "epoch": 0.62, + "learning_rate": 1.3208161043922387e-07, + "loss": 0.1377, + "step": 11916 + }, + { + "epoch": 0.62, + "learning_rate": 1.3204980925965593e-07, + "loss": 0.1367, + "step": 11917 + }, + { + "epoch": 0.62, + "learning_rate": 1.320180100221307e-07, + "loss": 0.1225, + "step": 11918 + }, + { + "epoch": 0.62, + "learning_rate": 1.3198621272755691e-07, + "loss": 0.1609, + "step": 11919 + }, + { + "epoch": 0.62, + "learning_rate": 1.3195441737684352e-07, + "loss": 0.1393, + "step": 11920 + }, + { + "epoch": 0.62, + "learning_rate": 1.319226239708991e-07, + "loss": 0.1304, + "step": 11921 + }, + { + "epoch": 0.62, + "learning_rate": 1.3189083251063236e-07, + "loss": 0.143, + "step": 11922 + }, + { + "epoch": 0.62, + "learning_rate": 1.31859042996952e-07, + "loss": 0.1357, + "step": 11923 + }, + { + "epoch": 0.62, + "learning_rate": 1.318272554307664e-07, + "loss": 0.1444, + "step": 11924 + }, + { + "epoch": 0.62, + "learning_rate": 1.317954698129842e-07, + "loss": 0.1457, + "step": 11925 + }, + { + "epoch": 0.62, + "learning_rate": 1.3176368614451376e-07, + "loss": 0.1196, + "step": 11926 + }, + { + "epoch": 0.62, + "learning_rate": 1.3173190442626356e-07, + "loss": 0.1111, + "step": 11927 + }, + { + "epoch": 0.62, + "learning_rate": 1.3170012465914184e-07, + "loss": 0.1265, + "step": 11928 + }, + { + "epoch": 0.62, + "learning_rate": 1.3166834684405697e-07, + "loss": 0.1487, + "step": 11929 + }, + { + "epoch": 0.62, + "learning_rate": 1.3163657098191707e-07, + "loss": 0.1262, + "step": 11930 + }, + { + "epoch": 0.62, + "learning_rate": 1.3160479707363043e-07, + "loss": 0.125, + "step": 11931 + }, + { + "epoch": 0.62, + "learning_rate": 1.315730251201051e-07, + "loss": 0.1628, + "step": 11932 + }, + { + "epoch": 0.62, + "learning_rate": 1.3154125512224906e-07, + "loss": 0.1737, + "step": 11933 + }, + { + "epoch": 0.62, + "learning_rate": 1.3150948708097046e-07, + "loss": 0.127, + "step": 11934 + }, + { + "epoch": 0.62, + "learning_rate": 1.3147772099717705e-07, + "loss": 0.1113, + "step": 11935 + }, + { + "epoch": 0.62, + "learning_rate": 1.3144595687177692e-07, + "loss": 0.1384, + "step": 11936 + }, + { + "epoch": 0.62, + "learning_rate": 1.314141947056777e-07, + "loss": 0.1346, + "step": 11937 + }, + { + "epoch": 0.62, + "learning_rate": 1.3138243449978736e-07, + "loss": 0.1211, + "step": 11938 + }, + { + "epoch": 0.62, + "learning_rate": 1.313506762550135e-07, + "loss": 0.1281, + "step": 11939 + }, + { + "epoch": 0.62, + "learning_rate": 1.3131891997226387e-07, + "loss": 0.1178, + "step": 11940 + }, + { + "epoch": 0.62, + "learning_rate": 1.31287165652446e-07, + "loss": 0.1378, + "step": 11941 + }, + { + "epoch": 0.62, + "learning_rate": 1.3125541329646742e-07, + "loss": 0.1519, + "step": 11942 + }, + { + "epoch": 0.62, + "learning_rate": 1.3122366290523568e-07, + "loss": 0.1441, + "step": 11943 + }, + { + "epoch": 0.62, + "learning_rate": 1.311919144796582e-07, + "loss": 0.1137, + "step": 11944 + }, + { + "epoch": 0.62, + "learning_rate": 1.3116016802064237e-07, + "loss": 0.1459, + "step": 11945 + }, + { + "epoch": 0.62, + "learning_rate": 1.311284235290955e-07, + "loss": 0.1733, + "step": 11946 + }, + { + "epoch": 0.62, + "learning_rate": 1.310966810059249e-07, + "loss": 0.1092, + "step": 11947 + }, + { + "epoch": 0.62, + "learning_rate": 1.3106494045203772e-07, + "loss": 0.1151, + "step": 11948 + }, + { + "epoch": 0.62, + "learning_rate": 1.3103320186834125e-07, + "loss": 0.1628, + "step": 11949 + }, + { + "epoch": 0.62, + "learning_rate": 1.3100146525574242e-07, + "loss": 0.1338, + "step": 11950 + }, + { + "epoch": 0.62, + "learning_rate": 1.3096973061514835e-07, + "loss": 0.1338, + "step": 11951 + }, + { + "epoch": 0.62, + "learning_rate": 1.3093799794746603e-07, + "loss": 0.1314, + "step": 11952 + }, + { + "epoch": 0.62, + "learning_rate": 1.3090626725360236e-07, + "loss": 0.1514, + "step": 11953 + }, + { + "epoch": 0.62, + "learning_rate": 1.308745385344643e-07, + "loss": 0.1675, + "step": 11954 + }, + { + "epoch": 0.62, + "learning_rate": 1.3084281179095858e-07, + "loss": 0.1365, + "step": 11955 + }, + { + "epoch": 0.62, + "learning_rate": 1.3081108702399201e-07, + "loss": 0.1098, + "step": 11956 + }, + { + "epoch": 0.62, + "learning_rate": 1.307793642344713e-07, + "loss": 0.1177, + "step": 11957 + }, + { + "epoch": 0.62, + "learning_rate": 1.3074764342330308e-07, + "loss": 0.1389, + "step": 11958 + }, + { + "epoch": 0.62, + "learning_rate": 1.3071592459139394e-07, + "loss": 0.1441, + "step": 11959 + }, + { + "epoch": 0.62, + "learning_rate": 1.3068420773965036e-07, + "loss": 0.1841, + "step": 11960 + }, + { + "epoch": 0.62, + "learning_rate": 1.3065249286897894e-07, + "loss": 0.1537, + "step": 11961 + }, + { + "epoch": 0.62, + "learning_rate": 1.30620779980286e-07, + "loss": 0.1595, + "step": 11962 + }, + { + "epoch": 0.62, + "learning_rate": 1.30589069074478e-07, + "loss": 0.137, + "step": 11963 + }, + { + "epoch": 0.62, + "learning_rate": 1.3055736015246117e-07, + "loss": 0.1474, + "step": 11964 + }, + { + "epoch": 0.62, + "learning_rate": 1.3052565321514181e-07, + "loss": 0.1294, + "step": 11965 + }, + { + "epoch": 0.62, + "learning_rate": 1.3049394826342616e-07, + "loss": 0.1091, + "step": 11966 + }, + { + "epoch": 0.62, + "learning_rate": 1.3046224529822025e-07, + "loss": 0.1721, + "step": 11967 + }, + { + "epoch": 0.62, + "learning_rate": 1.3043054432043018e-07, + "loss": 0.1427, + "step": 11968 + }, + { + "epoch": 0.62, + "learning_rate": 1.3039884533096205e-07, + "loss": 0.1444, + "step": 11969 + }, + { + "epoch": 0.62, + "learning_rate": 1.3036714833072176e-07, + "loss": 0.1371, + "step": 11970 + }, + { + "epoch": 0.62, + "learning_rate": 1.3033545332061527e-07, + "loss": 0.1259, + "step": 11971 + }, + { + "epoch": 0.62, + "learning_rate": 1.303037603015484e-07, + "loss": 0.1195, + "step": 11972 + }, + { + "epoch": 0.62, + "learning_rate": 1.3027206927442692e-07, + "loss": 0.1291, + "step": 11973 + }, + { + "epoch": 0.62, + "learning_rate": 1.3024038024015672e-07, + "loss": 0.0861, + "step": 11974 + }, + { + "epoch": 0.63, + "learning_rate": 1.3020869319964339e-07, + "loss": 0.1477, + "step": 11975 + }, + { + "epoch": 0.63, + "learning_rate": 1.301770081537925e-07, + "loss": 0.1764, + "step": 11976 + }, + { + "epoch": 0.63, + "learning_rate": 1.3014532510350967e-07, + "loss": 0.1267, + "step": 11977 + }, + { + "epoch": 0.63, + "learning_rate": 1.3011364404970043e-07, + "loss": 0.1256, + "step": 11978 + }, + { + "epoch": 0.63, + "learning_rate": 1.3008196499327016e-07, + "loss": 0.1489, + "step": 11979 + }, + { + "epoch": 0.63, + "learning_rate": 1.300502879351244e-07, + "loss": 0.146, + "step": 11980 + }, + { + "epoch": 0.63, + "learning_rate": 1.3001861287616836e-07, + "loss": 0.1311, + "step": 11981 + }, + { + "epoch": 0.63, + "learning_rate": 1.299869398173074e-07, + "loss": 0.127, + "step": 11982 + }, + { + "epoch": 0.63, + "learning_rate": 1.2995526875944677e-07, + "loss": 0.1359, + "step": 11983 + }, + { + "epoch": 0.63, + "learning_rate": 1.2992359970349156e-07, + "loss": 0.124, + "step": 11984 + }, + { + "epoch": 0.63, + "learning_rate": 1.2989193265034692e-07, + "loss": 0.1193, + "step": 11985 + }, + { + "epoch": 0.63, + "learning_rate": 1.298602676009179e-07, + "loss": 0.1551, + "step": 11986 + }, + { + "epoch": 0.63, + "learning_rate": 1.2982860455610954e-07, + "loss": 0.1049, + "step": 11987 + }, + { + "epoch": 0.63, + "learning_rate": 1.297969435168267e-07, + "loss": 0.1508, + "step": 11988 + }, + { + "epoch": 0.63, + "learning_rate": 1.2976528448397437e-07, + "loss": 0.1603, + "step": 11989 + }, + { + "epoch": 0.63, + "learning_rate": 1.2973362745845726e-07, + "loss": 0.1119, + "step": 11990 + }, + { + "epoch": 0.63, + "learning_rate": 1.2970197244118027e-07, + "loss": 0.1469, + "step": 11991 + }, + { + "epoch": 0.63, + "learning_rate": 1.2967031943304804e-07, + "loss": 0.1354, + "step": 11992 + }, + { + "epoch": 0.63, + "learning_rate": 1.296386684349652e-07, + "loss": 0.1314, + "step": 11993 + }, + { + "epoch": 0.63, + "learning_rate": 1.296070194478364e-07, + "loss": 0.1332, + "step": 11994 + }, + { + "epoch": 0.63, + "learning_rate": 1.295753724725661e-07, + "loss": 0.1296, + "step": 11995 + }, + { + "epoch": 0.63, + "learning_rate": 1.2954372751005892e-07, + "loss": 0.1276, + "step": 11996 + }, + { + "epoch": 0.63, + "learning_rate": 1.2951208456121916e-07, + "loss": 0.1231, + "step": 11997 + }, + { + "epoch": 0.63, + "learning_rate": 1.2948044362695125e-07, + "loss": 0.1374, + "step": 11998 + }, + { + "epoch": 0.63, + "learning_rate": 1.2944880470815945e-07, + "loss": 0.1456, + "step": 11999 + }, + { + "epoch": 0.63, + "learning_rate": 1.2941716780574813e-07, + "loss": 0.1269, + "step": 12000 + }, + { + "epoch": 0.63, + "learning_rate": 1.2938553292062138e-07, + "loss": 0.1135, + "step": 12001 + }, + { + "epoch": 0.63, + "learning_rate": 1.293539000536833e-07, + "loss": 0.1696, + "step": 12002 + }, + { + "epoch": 0.63, + "learning_rate": 1.2932226920583805e-07, + "loss": 0.1448, + "step": 12003 + }, + { + "epoch": 0.63, + "learning_rate": 1.2929064037798965e-07, + "loss": 0.0934, + "step": 12004 + }, + { + "epoch": 0.63, + "learning_rate": 1.2925901357104203e-07, + "loss": 0.1191, + "step": 12005 + }, + { + "epoch": 0.63, + "learning_rate": 1.2922738878589909e-07, + "loss": 0.1199, + "step": 12006 + }, + { + "epoch": 0.63, + "learning_rate": 1.2919576602346473e-07, + "loss": 0.1512, + "step": 12007 + }, + { + "epoch": 0.63, + "learning_rate": 1.2916414528464267e-07, + "loss": 0.1516, + "step": 12008 + }, + { + "epoch": 0.63, + "learning_rate": 1.2913252657033677e-07, + "loss": 0.1159, + "step": 12009 + }, + { + "epoch": 0.63, + "learning_rate": 1.291009098814506e-07, + "loss": 0.126, + "step": 12010 + }, + { + "epoch": 0.63, + "learning_rate": 1.2906929521888775e-07, + "loss": 0.135, + "step": 12011 + }, + { + "epoch": 0.63, + "learning_rate": 1.2903768258355184e-07, + "loss": 0.1567, + "step": 12012 + }, + { + "epoch": 0.63, + "learning_rate": 1.290060719763463e-07, + "loss": 0.122, + "step": 12013 + }, + { + "epoch": 0.63, + "learning_rate": 1.2897446339817472e-07, + "loss": 0.1655, + "step": 12014 + }, + { + "epoch": 0.63, + "learning_rate": 1.289428568499403e-07, + "loss": 0.158, + "step": 12015 + }, + { + "epoch": 0.63, + "learning_rate": 1.2891125233254653e-07, + "loss": 0.114, + "step": 12016 + }, + { + "epoch": 0.63, + "learning_rate": 1.2887964984689665e-07, + "loss": 0.1498, + "step": 12017 + }, + { + "epoch": 0.63, + "learning_rate": 1.2884804939389378e-07, + "loss": 0.1519, + "step": 12018 + }, + { + "epoch": 0.63, + "learning_rate": 1.2881645097444113e-07, + "loss": 0.1242, + "step": 12019 + }, + { + "epoch": 0.63, + "learning_rate": 1.2878485458944176e-07, + "loss": 0.1252, + "step": 12020 + }, + { + "epoch": 0.63, + "learning_rate": 1.287532602397988e-07, + "loss": 0.1753, + "step": 12021 + }, + { + "epoch": 0.63, + "learning_rate": 1.287216679264151e-07, + "loss": 0.1206, + "step": 12022 + }, + { + "epoch": 0.63, + "learning_rate": 1.2869007765019367e-07, + "loss": 0.1653, + "step": 12023 + }, + { + "epoch": 0.63, + "learning_rate": 1.2865848941203734e-07, + "loss": 0.1336, + "step": 12024 + }, + { + "epoch": 0.63, + "learning_rate": 1.2862690321284895e-07, + "loss": 0.1504, + "step": 12025 + }, + { + "epoch": 0.63, + "learning_rate": 1.2859531905353128e-07, + "loss": 0.1636, + "step": 12026 + }, + { + "epoch": 0.63, + "learning_rate": 1.2856373693498684e-07, + "loss": 0.1086, + "step": 12027 + }, + { + "epoch": 0.63, + "learning_rate": 1.2853215685811845e-07, + "loss": 0.198, + "step": 12028 + }, + { + "epoch": 0.63, + "learning_rate": 1.2850057882382856e-07, + "loss": 0.1554, + "step": 12029 + }, + { + "epoch": 0.63, + "learning_rate": 1.2846900283301976e-07, + "loss": 0.156, + "step": 12030 + }, + { + "epoch": 0.63, + "learning_rate": 1.284374288865944e-07, + "loss": 0.1299, + "step": 12031 + }, + { + "epoch": 0.63, + "learning_rate": 1.2840585698545504e-07, + "loss": 0.1365, + "step": 12032 + }, + { + "epoch": 0.63, + "learning_rate": 1.283742871305039e-07, + "loss": 0.1186, + "step": 12033 + }, + { + "epoch": 0.63, + "learning_rate": 1.2834271932264328e-07, + "loss": 0.1342, + "step": 12034 + }, + { + "epoch": 0.63, + "learning_rate": 1.2831115356277546e-07, + "loss": 0.1499, + "step": 12035 + }, + { + "epoch": 0.63, + "learning_rate": 1.2827958985180246e-07, + "loss": 0.1614, + "step": 12036 + }, + { + "epoch": 0.63, + "learning_rate": 1.282480281906265e-07, + "loss": 0.1204, + "step": 12037 + }, + { + "epoch": 0.63, + "learning_rate": 1.2821646858014956e-07, + "loss": 0.1159, + "step": 12038 + }, + { + "epoch": 0.63, + "learning_rate": 1.2818491102127368e-07, + "loss": 0.1922, + "step": 12039 + }, + { + "epoch": 0.63, + "learning_rate": 1.2815335551490075e-07, + "loss": 0.1127, + "step": 12040 + }, + { + "epoch": 0.63, + "learning_rate": 1.281218020619327e-07, + "loss": 0.1454, + "step": 12041 + }, + { + "epoch": 0.63, + "learning_rate": 1.2809025066327128e-07, + "loss": 0.1117, + "step": 12042 + }, + { + "epoch": 0.63, + "learning_rate": 1.280587013198183e-07, + "loss": 0.1454, + "step": 12043 + }, + { + "epoch": 0.63, + "learning_rate": 1.2802715403247536e-07, + "loss": 0.1369, + "step": 12044 + }, + { + "epoch": 0.63, + "learning_rate": 1.279956088021441e-07, + "loss": 0.1221, + "step": 12045 + }, + { + "epoch": 0.63, + "learning_rate": 1.279640656297262e-07, + "loss": 0.1344, + "step": 12046 + }, + { + "epoch": 0.63, + "learning_rate": 1.279325245161231e-07, + "loss": 0.1627, + "step": 12047 + }, + { + "epoch": 0.63, + "learning_rate": 1.2790098546223626e-07, + "loss": 0.1639, + "step": 12048 + }, + { + "epoch": 0.63, + "learning_rate": 1.2786944846896705e-07, + "loss": 0.1409, + "step": 12049 + }, + { + "epoch": 0.63, + "learning_rate": 1.2783791353721692e-07, + "loss": 0.1582, + "step": 12050 + }, + { + "epoch": 0.63, + "learning_rate": 1.2780638066788702e-07, + "loss": 0.1178, + "step": 12051 + }, + { + "epoch": 0.63, + "learning_rate": 1.277748498618787e-07, + "loss": 0.1803, + "step": 12052 + }, + { + "epoch": 0.63, + "learning_rate": 1.2774332112009303e-07, + "loss": 0.1055, + "step": 12053 + }, + { + "epoch": 0.63, + "learning_rate": 1.2771179444343108e-07, + "loss": 0.1942, + "step": 12054 + }, + { + "epoch": 0.63, + "learning_rate": 1.2768026983279402e-07, + "loss": 0.1166, + "step": 12055 + }, + { + "epoch": 0.63, + "learning_rate": 1.2764874728908267e-07, + "loss": 0.1259, + "step": 12056 + }, + { + "epoch": 0.63, + "learning_rate": 1.2761722681319816e-07, + "loss": 0.1021, + "step": 12057 + }, + { + "epoch": 0.63, + "learning_rate": 1.2758570840604117e-07, + "loss": 0.1134, + "step": 12058 + }, + { + "epoch": 0.63, + "learning_rate": 1.2755419206851264e-07, + "loss": 0.1287, + "step": 12059 + }, + { + "epoch": 0.63, + "learning_rate": 1.275226778015133e-07, + "loss": 0.1547, + "step": 12060 + }, + { + "epoch": 0.63, + "learning_rate": 1.2749116560594376e-07, + "loss": 0.1154, + "step": 12061 + }, + { + "epoch": 0.63, + "learning_rate": 1.274596554827047e-07, + "loss": 0.1295, + "step": 12062 + }, + { + "epoch": 0.63, + "learning_rate": 1.2742814743269667e-07, + "loss": 0.1703, + "step": 12063 + }, + { + "epoch": 0.63, + "learning_rate": 1.273966414568202e-07, + "loss": 0.1259, + "step": 12064 + }, + { + "epoch": 0.63, + "learning_rate": 1.2736513755597575e-07, + "loss": 0.1209, + "step": 12065 + }, + { + "epoch": 0.63, + "learning_rate": 1.2733363573106375e-07, + "loss": 0.1988, + "step": 12066 + }, + { + "epoch": 0.63, + "learning_rate": 1.2730213598298446e-07, + "loss": 0.1269, + "step": 12067 + }, + { + "epoch": 0.63, + "learning_rate": 1.272706383126382e-07, + "loss": 0.1111, + "step": 12068 + }, + { + "epoch": 0.63, + "learning_rate": 1.2723914272092523e-07, + "loss": 0.1238, + "step": 12069 + }, + { + "epoch": 0.63, + "learning_rate": 1.2720764920874553e-07, + "loss": 0.1572, + "step": 12070 + }, + { + "epoch": 0.63, + "learning_rate": 1.271761577769994e-07, + "loss": 0.1054, + "step": 12071 + }, + { + "epoch": 0.63, + "learning_rate": 1.2714466842658673e-07, + "loss": 0.155, + "step": 12072 + }, + { + "epoch": 0.63, + "learning_rate": 1.2711318115840763e-07, + "loss": 0.1697, + "step": 12073 + }, + { + "epoch": 0.63, + "learning_rate": 1.2708169597336188e-07, + "loss": 0.1372, + "step": 12074 + }, + { + "epoch": 0.63, + "learning_rate": 1.2705021287234946e-07, + "loss": 0.1221, + "step": 12075 + }, + { + "epoch": 0.63, + "learning_rate": 1.270187318562701e-07, + "loss": 0.1258, + "step": 12076 + }, + { + "epoch": 0.63, + "learning_rate": 1.2698725292602363e-07, + "loss": 0.1291, + "step": 12077 + }, + { + "epoch": 0.63, + "learning_rate": 1.2695577608250956e-07, + "loss": 0.1471, + "step": 12078 + }, + { + "epoch": 0.63, + "learning_rate": 1.2692430132662767e-07, + "loss": 0.1395, + "step": 12079 + }, + { + "epoch": 0.63, + "learning_rate": 1.2689282865927738e-07, + "loss": 0.1295, + "step": 12080 + }, + { + "epoch": 0.63, + "learning_rate": 1.2686135808135832e-07, + "loss": 0.1264, + "step": 12081 + }, + { + "epoch": 0.63, + "learning_rate": 1.2682988959376986e-07, + "loss": 0.143, + "step": 12082 + }, + { + "epoch": 0.63, + "learning_rate": 1.2679842319741143e-07, + "loss": 0.1226, + "step": 12083 + }, + { + "epoch": 0.63, + "learning_rate": 1.2676695889318229e-07, + "loss": 0.1158, + "step": 12084 + }, + { + "epoch": 0.63, + "learning_rate": 1.2673549668198172e-07, + "loss": 0.1724, + "step": 12085 + }, + { + "epoch": 0.63, + "learning_rate": 1.2670403656470905e-07, + "loss": 0.1348, + "step": 12086 + }, + { + "epoch": 0.63, + "learning_rate": 1.266725785422632e-07, + "loss": 0.1366, + "step": 12087 + }, + { + "epoch": 0.63, + "learning_rate": 1.2664112261554342e-07, + "loss": 0.1995, + "step": 12088 + }, + { + "epoch": 0.63, + "learning_rate": 1.2660966878544863e-07, + "loss": 0.1289, + "step": 12089 + }, + { + "epoch": 0.63, + "learning_rate": 1.265782170528779e-07, + "loss": 0.1207, + "step": 12090 + }, + { + "epoch": 0.63, + "learning_rate": 1.2654676741872997e-07, + "loss": 0.1539, + "step": 12091 + }, + { + "epoch": 0.63, + "learning_rate": 1.2651531988390385e-07, + "loss": 0.1402, + "step": 12092 + }, + { + "epoch": 0.63, + "learning_rate": 1.2648387444929823e-07, + "loss": 0.1118, + "step": 12093 + }, + { + "epoch": 0.63, + "learning_rate": 1.2645243111581188e-07, + "loss": 0.1447, + "step": 12094 + }, + { + "epoch": 0.63, + "learning_rate": 1.2642098988434348e-07, + "loss": 0.1589, + "step": 12095 + }, + { + "epoch": 0.63, + "learning_rate": 1.2638955075579153e-07, + "loss": 0.1251, + "step": 12096 + }, + { + "epoch": 0.63, + "learning_rate": 1.2635811373105467e-07, + "loss": 0.1734, + "step": 12097 + }, + { + "epoch": 0.63, + "learning_rate": 1.263266788110313e-07, + "loss": 0.1707, + "step": 12098 + }, + { + "epoch": 0.63, + "learning_rate": 1.2629524599661997e-07, + "loss": 0.1873, + "step": 12099 + }, + { + "epoch": 0.63, + "learning_rate": 1.2626381528871885e-07, + "loss": 0.1046, + "step": 12100 + }, + { + "epoch": 0.63, + "learning_rate": 1.2623238668822643e-07, + "loss": 0.1121, + "step": 12101 + }, + { + "epoch": 0.63, + "learning_rate": 1.2620096019604085e-07, + "loss": 0.1269, + "step": 12102 + }, + { + "epoch": 0.63, + "learning_rate": 1.2616953581306037e-07, + "loss": 0.1529, + "step": 12103 + }, + { + "epoch": 0.63, + "learning_rate": 1.2613811354018306e-07, + "loss": 0.1135, + "step": 12104 + }, + { + "epoch": 0.63, + "learning_rate": 1.2610669337830691e-07, + "loss": 0.1415, + "step": 12105 + }, + { + "epoch": 0.63, + "learning_rate": 1.2607527532833002e-07, + "loss": 0.1567, + "step": 12106 + }, + { + "epoch": 0.63, + "learning_rate": 1.2604385939115026e-07, + "loss": 0.1398, + "step": 12107 + }, + { + "epoch": 0.63, + "learning_rate": 1.260124455676656e-07, + "loss": 0.1083, + "step": 12108 + }, + { + "epoch": 0.63, + "learning_rate": 1.2598103385877379e-07, + "loss": 0.1588, + "step": 12109 + }, + { + "epoch": 0.63, + "learning_rate": 1.2594962426537262e-07, + "loss": 0.1721, + "step": 12110 + }, + { + "epoch": 0.63, + "learning_rate": 1.259182167883597e-07, + "loss": 0.1356, + "step": 12111 + }, + { + "epoch": 0.63, + "learning_rate": 1.258868114286329e-07, + "loss": 0.1708, + "step": 12112 + }, + { + "epoch": 0.63, + "learning_rate": 1.2585540818708955e-07, + "loss": 0.1122, + "step": 12113 + }, + { + "epoch": 0.63, + "learning_rate": 1.2582400706462724e-07, + "loss": 0.0943, + "step": 12114 + }, + { + "epoch": 0.63, + "learning_rate": 1.2579260806214347e-07, + "loss": 0.1305, + "step": 12115 + }, + { + "epoch": 0.63, + "learning_rate": 1.2576121118053555e-07, + "loss": 0.1389, + "step": 12116 + }, + { + "epoch": 0.63, + "learning_rate": 1.2572981642070093e-07, + "loss": 0.1389, + "step": 12117 + }, + { + "epoch": 0.63, + "learning_rate": 1.2569842378353678e-07, + "loss": 0.1377, + "step": 12118 + }, + { + "epoch": 0.63, + "learning_rate": 1.2566703326994042e-07, + "loss": 0.1288, + "step": 12119 + }, + { + "epoch": 0.63, + "learning_rate": 1.25635644880809e-07, + "loss": 0.1689, + "step": 12120 + }, + { + "epoch": 0.63, + "learning_rate": 1.2560425861703942e-07, + "loss": 0.1253, + "step": 12121 + }, + { + "epoch": 0.63, + "learning_rate": 1.2557287447952894e-07, + "loss": 0.1663, + "step": 12122 + }, + { + "epoch": 0.63, + "learning_rate": 1.255414924691744e-07, + "loss": 0.1248, + "step": 12123 + }, + { + "epoch": 0.63, + "learning_rate": 1.2551011258687276e-07, + "loss": 0.1434, + "step": 12124 + }, + { + "epoch": 0.63, + "learning_rate": 1.254787348335208e-07, + "loss": 0.1336, + "step": 12125 + }, + { + "epoch": 0.63, + "learning_rate": 1.2544735921001547e-07, + "loss": 0.1784, + "step": 12126 + }, + { + "epoch": 0.63, + "learning_rate": 1.254159857172533e-07, + "loss": 0.1355, + "step": 12127 + }, + { + "epoch": 0.63, + "learning_rate": 1.253846143561311e-07, + "loss": 0.1471, + "step": 12128 + }, + { + "epoch": 0.63, + "learning_rate": 1.253532451275455e-07, + "loss": 0.1354, + "step": 12129 + }, + { + "epoch": 0.63, + "learning_rate": 1.2532187803239284e-07, + "loss": 0.1296, + "step": 12130 + }, + { + "epoch": 0.63, + "learning_rate": 1.2529051307156982e-07, + "loss": 0.1102, + "step": 12131 + }, + { + "epoch": 0.63, + "learning_rate": 1.2525915024597271e-07, + "loss": 0.148, + "step": 12132 + }, + { + "epoch": 0.63, + "learning_rate": 1.2522778955649796e-07, + "loss": 0.1195, + "step": 12133 + }, + { + "epoch": 0.63, + "learning_rate": 1.2519643100404186e-07, + "loss": 0.1461, + "step": 12134 + }, + { + "epoch": 0.63, + "learning_rate": 1.2516507458950065e-07, + "loss": 0.1752, + "step": 12135 + }, + { + "epoch": 0.63, + "learning_rate": 1.2513372031377047e-07, + "loss": 0.1365, + "step": 12136 + }, + { + "epoch": 0.63, + "learning_rate": 1.2510236817774752e-07, + "loss": 0.1244, + "step": 12137 + }, + { + "epoch": 0.63, + "learning_rate": 1.250710181823278e-07, + "loss": 0.1411, + "step": 12138 + }, + { + "epoch": 0.63, + "learning_rate": 1.2503967032840725e-07, + "loss": 0.1363, + "step": 12139 + }, + { + "epoch": 0.63, + "learning_rate": 1.2500832461688185e-07, + "loss": 0.1394, + "step": 12140 + }, + { + "epoch": 0.63, + "learning_rate": 1.2497698104864752e-07, + "loss": 0.1526, + "step": 12141 + }, + { + "epoch": 0.63, + "learning_rate": 1.2494563962460003e-07, + "loss": 0.1377, + "step": 12142 + }, + { + "epoch": 0.63, + "learning_rate": 1.249143003456351e-07, + "loss": 0.1624, + "step": 12143 + }, + { + "epoch": 0.63, + "learning_rate": 1.2488296321264851e-07, + "loss": 0.1284, + "step": 12144 + }, + { + "epoch": 0.63, + "learning_rate": 1.2485162822653577e-07, + "loss": 0.1415, + "step": 12145 + }, + { + "epoch": 0.63, + "learning_rate": 1.248202953881926e-07, + "loss": 0.148, + "step": 12146 + }, + { + "epoch": 0.63, + "learning_rate": 1.2478896469851434e-07, + "loss": 0.1509, + "step": 12147 + }, + { + "epoch": 0.63, + "learning_rate": 1.247576361583965e-07, + "loss": 0.1707, + "step": 12148 + }, + { + "epoch": 0.63, + "learning_rate": 1.2472630976873445e-07, + "loss": 0.1644, + "step": 12149 + }, + { + "epoch": 0.63, + "learning_rate": 1.2469498553042354e-07, + "loss": 0.1134, + "step": 12150 + }, + { + "epoch": 0.63, + "learning_rate": 1.2466366344435905e-07, + "loss": 0.1173, + "step": 12151 + }, + { + "epoch": 0.63, + "learning_rate": 1.2463234351143609e-07, + "loss": 0.1342, + "step": 12152 + }, + { + "epoch": 0.63, + "learning_rate": 1.2460102573254986e-07, + "loss": 0.1567, + "step": 12153 + }, + { + "epoch": 0.63, + "learning_rate": 1.2456971010859544e-07, + "loss": 0.1523, + "step": 12154 + }, + { + "epoch": 0.63, + "learning_rate": 1.2453839664046785e-07, + "loss": 0.1277, + "step": 12155 + }, + { + "epoch": 0.63, + "learning_rate": 1.24507085329062e-07, + "loss": 0.1258, + "step": 12156 + }, + { + "epoch": 0.63, + "learning_rate": 1.2447577617527273e-07, + "loss": 0.1565, + "step": 12157 + }, + { + "epoch": 0.63, + "learning_rate": 1.2444446917999504e-07, + "loss": 0.1741, + "step": 12158 + }, + { + "epoch": 0.63, + "learning_rate": 1.2441316434412347e-07, + "loss": 0.1396, + "step": 12159 + }, + { + "epoch": 0.63, + "learning_rate": 1.2438186166855292e-07, + "loss": 0.1753, + "step": 12160 + }, + { + "epoch": 0.63, + "learning_rate": 1.243505611541779e-07, + "loss": 0.1665, + "step": 12161 + }, + { + "epoch": 0.63, + "learning_rate": 1.2431926280189312e-07, + "loss": 0.1291, + "step": 12162 + }, + { + "epoch": 0.63, + "learning_rate": 1.24287966612593e-07, + "loss": 0.1374, + "step": 12163 + }, + { + "epoch": 0.63, + "learning_rate": 1.24256672587172e-07, + "loss": 0.1342, + "step": 12164 + }, + { + "epoch": 0.63, + "learning_rate": 1.2422538072652456e-07, + "loss": 0.1362, + "step": 12165 + }, + { + "epoch": 0.64, + "learning_rate": 1.2419409103154496e-07, + "loss": 0.1374, + "step": 12166 + }, + { + "epoch": 0.64, + "learning_rate": 1.2416280350312754e-07, + "loss": 0.1075, + "step": 12167 + }, + { + "epoch": 0.64, + "learning_rate": 1.241315181421664e-07, + "loss": 0.1431, + "step": 12168 + }, + { + "epoch": 0.64, + "learning_rate": 1.2410023494955584e-07, + "loss": 0.1277, + "step": 12169 + }, + { + "epoch": 0.64, + "learning_rate": 1.240689539261898e-07, + "loss": 0.1674, + "step": 12170 + }, + { + "epoch": 0.64, + "learning_rate": 1.240376750729624e-07, + "loss": 0.1628, + "step": 12171 + }, + { + "epoch": 0.64, + "learning_rate": 1.2400639839076766e-07, + "loss": 0.1587, + "step": 12172 + }, + { + "epoch": 0.64, + "learning_rate": 1.2397512388049926e-07, + "loss": 0.1138, + "step": 12173 + }, + { + "epoch": 0.64, + "learning_rate": 1.2394385154305125e-07, + "loss": 0.1379, + "step": 12174 + }, + { + "epoch": 0.64, + "learning_rate": 1.2391258137931723e-07, + "loss": 0.1152, + "step": 12175 + }, + { + "epoch": 0.64, + "learning_rate": 1.238813133901911e-07, + "loss": 0.1269, + "step": 12176 + }, + { + "epoch": 0.64, + "learning_rate": 1.2385004757656635e-07, + "loss": 0.1301, + "step": 12177 + }, + { + "epoch": 0.64, + "learning_rate": 1.2381878393933668e-07, + "loss": 0.1154, + "step": 12178 + }, + { + "epoch": 0.64, + "learning_rate": 1.2378752247939554e-07, + "loss": 0.1746, + "step": 12179 + }, + { + "epoch": 0.64, + "learning_rate": 1.237562631976365e-07, + "loss": 0.1439, + "step": 12180 + }, + { + "epoch": 0.64, + "learning_rate": 1.2372500609495287e-07, + "loss": 0.1397, + "step": 12181 + }, + { + "epoch": 0.64, + "learning_rate": 1.2369375117223797e-07, + "loss": 0.1214, + "step": 12182 + }, + { + "epoch": 0.64, + "learning_rate": 1.236624984303851e-07, + "loss": 0.123, + "step": 12183 + }, + { + "epoch": 0.64, + "learning_rate": 1.2363124787028755e-07, + "loss": 0.1538, + "step": 12184 + }, + { + "epoch": 0.64, + "learning_rate": 1.2359999949283838e-07, + "loss": 0.1458, + "step": 12185 + }, + { + "epoch": 0.64, + "learning_rate": 1.2356875329893073e-07, + "loss": 0.1364, + "step": 12186 + }, + { + "epoch": 0.64, + "learning_rate": 1.2353750928945762e-07, + "loss": 0.1175, + "step": 12187 + }, + { + "epoch": 0.64, + "learning_rate": 1.2350626746531197e-07, + "loss": 0.1228, + "step": 12188 + }, + { + "epoch": 0.64, + "learning_rate": 1.2347502782738687e-07, + "loss": 0.1322, + "step": 12189 + }, + { + "epoch": 0.64, + "learning_rate": 1.234437903765749e-07, + "loss": 0.1303, + "step": 12190 + }, + { + "epoch": 0.64, + "learning_rate": 1.23412555113769e-07, + "loss": 0.1065, + "step": 12191 + }, + { + "epoch": 0.64, + "learning_rate": 1.2338132203986176e-07, + "loss": 0.1315, + "step": 12192 + }, + { + "epoch": 0.64, + "learning_rate": 1.2335009115574595e-07, + "loss": 0.123, + "step": 12193 + }, + { + "epoch": 0.64, + "learning_rate": 1.2331886246231413e-07, + "loss": 0.1222, + "step": 12194 + }, + { + "epoch": 0.64, + "learning_rate": 1.2328763596045882e-07, + "loss": 0.1411, + "step": 12195 + }, + { + "epoch": 0.64, + "learning_rate": 1.2325641165107247e-07, + "loss": 0.1439, + "step": 12196 + }, + { + "epoch": 0.64, + "learning_rate": 1.232251895350476e-07, + "loss": 0.1172, + "step": 12197 + }, + { + "epoch": 0.64, + "learning_rate": 1.231939696132764e-07, + "loss": 0.1216, + "step": 12198 + }, + { + "epoch": 0.64, + "learning_rate": 1.2316275188665114e-07, + "loss": 0.1275, + "step": 12199 + }, + { + "epoch": 0.64, + "learning_rate": 1.2313153635606415e-07, + "loss": 0.1259, + "step": 12200 + }, + { + "epoch": 0.64, + "learning_rate": 1.2310032302240748e-07, + "loss": 0.1568, + "step": 12201 + }, + { + "epoch": 0.64, + "learning_rate": 1.230691118865733e-07, + "loss": 0.129, + "step": 12202 + }, + { + "epoch": 0.64, + "learning_rate": 1.2303790294945356e-07, + "loss": 0.11, + "step": 12203 + }, + { + "epoch": 0.64, + "learning_rate": 1.2300669621194032e-07, + "loss": 0.1491, + "step": 12204 + }, + { + "epoch": 0.64, + "learning_rate": 1.2297549167492536e-07, + "loss": 0.1036, + "step": 12205 + }, + { + "epoch": 0.64, + "learning_rate": 1.2294428933930068e-07, + "loss": 0.1358, + "step": 12206 + }, + { + "epoch": 0.64, + "learning_rate": 1.229130892059579e-07, + "loss": 0.1028, + "step": 12207 + }, + { + "epoch": 0.64, + "learning_rate": 1.2288189127578877e-07, + "loss": 0.1304, + "step": 12208 + }, + { + "epoch": 0.64, + "learning_rate": 1.22850695549685e-07, + "loss": 0.1338, + "step": 12209 + }, + { + "epoch": 0.64, + "learning_rate": 1.2281950202853805e-07, + "loss": 0.1421, + "step": 12210 + }, + { + "epoch": 0.64, + "learning_rate": 1.227883107132396e-07, + "loss": 0.1337, + "step": 12211 + }, + { + "epoch": 0.64, + "learning_rate": 1.2275712160468103e-07, + "loss": 0.1298, + "step": 12212 + }, + { + "epoch": 0.64, + "learning_rate": 1.2272593470375373e-07, + "loss": 0.1092, + "step": 12213 + }, + { + "epoch": 0.64, + "learning_rate": 1.2269475001134902e-07, + "loss": 0.1126, + "step": 12214 + }, + { + "epoch": 0.64, + "learning_rate": 1.226635675283583e-07, + "loss": 0.1107, + "step": 12215 + }, + { + "epoch": 0.64, + "learning_rate": 1.2263238725567262e-07, + "loss": 0.1367, + "step": 12216 + }, + { + "epoch": 0.64, + "learning_rate": 1.2260120919418314e-07, + "loss": 0.178, + "step": 12217 + }, + { + "epoch": 0.64, + "learning_rate": 1.22570033344781e-07, + "loss": 0.1335, + "step": 12218 + }, + { + "epoch": 0.64, + "learning_rate": 1.225388597083572e-07, + "loss": 0.1351, + "step": 12219 + }, + { + "epoch": 0.64, + "learning_rate": 1.225076882858027e-07, + "loss": 0.1196, + "step": 12220 + }, + { + "epoch": 0.64, + "learning_rate": 1.2247651907800834e-07, + "loss": 0.1687, + "step": 12221 + }, + { + "epoch": 0.64, + "learning_rate": 1.2244535208586504e-07, + "loss": 0.1205, + "step": 12222 + }, + { + "epoch": 0.64, + "learning_rate": 1.2241418731026355e-07, + "loss": 0.139, + "step": 12223 + }, + { + "epoch": 0.64, + "learning_rate": 1.223830247520945e-07, + "loss": 0.1418, + "step": 12224 + }, + { + "epoch": 0.64, + "learning_rate": 1.2235186441224857e-07, + "loss": 0.1321, + "step": 12225 + }, + { + "epoch": 0.64, + "learning_rate": 1.2232070629161628e-07, + "loss": 0.1025, + "step": 12226 + }, + { + "epoch": 0.64, + "learning_rate": 1.2228955039108827e-07, + "loss": 0.1018, + "step": 12227 + }, + { + "epoch": 0.64, + "learning_rate": 1.2225839671155483e-07, + "loss": 0.151, + "step": 12228 + }, + { + "epoch": 0.64, + "learning_rate": 1.2222724525390645e-07, + "loss": 0.1188, + "step": 12229 + }, + { + "epoch": 0.64, + "learning_rate": 1.2219609601903343e-07, + "loss": 0.1674, + "step": 12230 + }, + { + "epoch": 0.64, + "learning_rate": 1.2216494900782604e-07, + "loss": 0.1242, + "step": 12231 + }, + { + "epoch": 0.64, + "learning_rate": 1.221338042211745e-07, + "loss": 0.1353, + "step": 12232 + }, + { + "epoch": 0.64, + "learning_rate": 1.2210266165996882e-07, + "loss": 0.126, + "step": 12233 + }, + { + "epoch": 0.64, + "learning_rate": 1.2207152132509918e-07, + "loss": 0.1656, + "step": 12234 + }, + { + "epoch": 0.64, + "learning_rate": 1.220403832174555e-07, + "loss": 0.1485, + "step": 12235 + }, + { + "epoch": 0.64, + "learning_rate": 1.2200924733792778e-07, + "loss": 0.1327, + "step": 12236 + }, + { + "epoch": 0.64, + "learning_rate": 1.2197811368740588e-07, + "loss": 0.1174, + "step": 12237 + }, + { + "epoch": 0.64, + "learning_rate": 1.2194698226677963e-07, + "loss": 0.1467, + "step": 12238 + }, + { + "epoch": 0.64, + "learning_rate": 1.219158530769387e-07, + "loss": 0.1283, + "step": 12239 + }, + { + "epoch": 0.64, + "learning_rate": 1.2188472611877294e-07, + "loss": 0.109, + "step": 12240 + }, + { + "epoch": 0.64, + "learning_rate": 1.2185360139317182e-07, + "loss": 0.1544, + "step": 12241 + }, + { + "epoch": 0.64, + "learning_rate": 1.218224789010249e-07, + "loss": 0.153, + "step": 12242 + }, + { + "epoch": 0.64, + "learning_rate": 1.2179135864322172e-07, + "loss": 0.1202, + "step": 12243 + }, + { + "epoch": 0.64, + "learning_rate": 1.2176024062065172e-07, + "loss": 0.1331, + "step": 12244 + }, + { + "epoch": 0.64, + "learning_rate": 1.2172912483420423e-07, + "loss": 0.1157, + "step": 12245 + }, + { + "epoch": 0.64, + "learning_rate": 1.2169801128476856e-07, + "loss": 0.1329, + "step": 12246 + }, + { + "epoch": 0.64, + "learning_rate": 1.2166689997323398e-07, + "loss": 0.1306, + "step": 12247 + }, + { + "epoch": 0.64, + "learning_rate": 1.2163579090048962e-07, + "loss": 0.1518, + "step": 12248 + }, + { + "epoch": 0.64, + "learning_rate": 1.2160468406742464e-07, + "loss": 0.1356, + "step": 12249 + }, + { + "epoch": 0.64, + "learning_rate": 1.2157357947492805e-07, + "loss": 0.1361, + "step": 12250 + }, + { + "epoch": 0.64, + "learning_rate": 1.215424771238888e-07, + "loss": 0.1324, + "step": 12251 + }, + { + "epoch": 0.64, + "learning_rate": 1.2151137701519588e-07, + "loss": 0.1346, + "step": 12252 + }, + { + "epoch": 0.64, + "learning_rate": 1.2148027914973807e-07, + "loss": 0.1225, + "step": 12253 + }, + { + "epoch": 0.64, + "learning_rate": 1.214491835284042e-07, + "loss": 0.1529, + "step": 12254 + }, + { + "epoch": 0.64, + "learning_rate": 1.2141809015208298e-07, + "loss": 0.1425, + "step": 12255 + }, + { + "epoch": 0.64, + "learning_rate": 1.213869990216631e-07, + "loss": 0.1482, + "step": 12256 + }, + { + "epoch": 0.64, + "learning_rate": 1.2135591013803316e-07, + "loss": 0.1393, + "step": 12257 + }, + { + "epoch": 0.64, + "learning_rate": 1.2132482350208161e-07, + "loss": 0.1173, + "step": 12258 + }, + { + "epoch": 0.64, + "learning_rate": 1.2129373911469706e-07, + "loss": 0.1064, + "step": 12259 + }, + { + "epoch": 0.64, + "learning_rate": 1.2126265697676773e-07, + "loss": 0.137, + "step": 12260 + }, + { + "epoch": 0.64, + "learning_rate": 1.2123157708918212e-07, + "loss": 0.1942, + "step": 12261 + }, + { + "epoch": 0.64, + "learning_rate": 1.212004994528284e-07, + "loss": 0.1085, + "step": 12262 + }, + { + "epoch": 0.64, + "learning_rate": 1.2116942406859484e-07, + "loss": 0.1617, + "step": 12263 + }, + { + "epoch": 0.64, + "learning_rate": 1.2113835093736958e-07, + "loss": 0.1369, + "step": 12264 + }, + { + "epoch": 0.64, + "learning_rate": 1.211072800600407e-07, + "loss": 0.149, + "step": 12265 + }, + { + "epoch": 0.64, + "learning_rate": 1.2107621143749627e-07, + "loss": 0.1322, + "step": 12266 + }, + { + "epoch": 0.64, + "learning_rate": 1.210451450706241e-07, + "loss": 0.1557, + "step": 12267 + }, + { + "epoch": 0.64, + "learning_rate": 1.210140809603122e-07, + "loss": 0.1175, + "step": 12268 + }, + { + "epoch": 0.64, + "learning_rate": 1.209830191074483e-07, + "loss": 0.1201, + "step": 12269 + }, + { + "epoch": 0.64, + "learning_rate": 1.209519595129203e-07, + "loss": 0.1434, + "step": 12270 + }, + { + "epoch": 0.64, + "learning_rate": 1.2092090217761576e-07, + "loss": 0.1214, + "step": 12271 + }, + { + "epoch": 0.64, + "learning_rate": 1.2088984710242237e-07, + "loss": 0.1633, + "step": 12272 + }, + { + "epoch": 0.64, + "learning_rate": 1.2085879428822766e-07, + "loss": 0.1245, + "step": 12273 + }, + { + "epoch": 0.64, + "learning_rate": 1.2082774373591927e-07, + "loss": 0.1309, + "step": 12274 + }, + { + "epoch": 0.64, + "learning_rate": 1.2079669544638446e-07, + "loss": 0.1196, + "step": 12275 + }, + { + "epoch": 0.64, + "learning_rate": 1.2076564942051065e-07, + "loss": 0.1062, + "step": 12276 + }, + { + "epoch": 0.64, + "learning_rate": 1.2073460565918518e-07, + "loss": 0.1383, + "step": 12277 + }, + { + "epoch": 0.64, + "learning_rate": 1.2070356416329527e-07, + "loss": 0.1438, + "step": 12278 + }, + { + "epoch": 0.64, + "learning_rate": 1.2067252493372815e-07, + "loss": 0.1174, + "step": 12279 + }, + { + "epoch": 0.64, + "learning_rate": 1.2064148797137081e-07, + "loss": 0.1399, + "step": 12280 + }, + { + "epoch": 0.64, + "learning_rate": 1.2061045327711045e-07, + "loss": 0.1256, + "step": 12281 + }, + { + "epoch": 0.64, + "learning_rate": 1.2057942085183396e-07, + "loss": 0.1589, + "step": 12282 + }, + { + "epoch": 0.64, + "learning_rate": 1.2054839069642833e-07, + "loss": 0.1519, + "step": 12283 + }, + { + "epoch": 0.64, + "learning_rate": 1.2051736281178035e-07, + "loss": 0.1235, + "step": 12284 + }, + { + "epoch": 0.64, + "learning_rate": 1.2048633719877676e-07, + "loss": 0.1495, + "step": 12285 + }, + { + "epoch": 0.64, + "learning_rate": 1.204553138583044e-07, + "loss": 0.1184, + "step": 12286 + }, + { + "epoch": 0.64, + "learning_rate": 1.204242927912498e-07, + "loss": 0.1573, + "step": 12287 + }, + { + "epoch": 0.64, + "learning_rate": 1.2039327399849972e-07, + "loss": 0.1369, + "step": 12288 + }, + { + "epoch": 0.64, + "learning_rate": 1.2036225748094055e-07, + "loss": 0.1624, + "step": 12289 + }, + { + "epoch": 0.64, + "learning_rate": 1.2033124323945884e-07, + "loss": 0.1398, + "step": 12290 + }, + { + "epoch": 0.64, + "learning_rate": 1.2030023127494092e-07, + "loss": 0.1276, + "step": 12291 + }, + { + "epoch": 0.64, + "learning_rate": 1.2026922158827323e-07, + "loss": 0.1885, + "step": 12292 + }, + { + "epoch": 0.64, + "learning_rate": 1.2023821418034192e-07, + "loss": 0.1522, + "step": 12293 + }, + { + "epoch": 0.64, + "learning_rate": 1.202072090520332e-07, + "loss": 0.1158, + "step": 12294 + }, + { + "epoch": 0.64, + "learning_rate": 1.201762062042333e-07, + "loss": 0.1573, + "step": 12295 + }, + { + "epoch": 0.64, + "learning_rate": 1.2014520563782822e-07, + "loss": 0.1607, + "step": 12296 + }, + { + "epoch": 0.64, + "learning_rate": 1.20114207353704e-07, + "loss": 0.1404, + "step": 12297 + }, + { + "epoch": 0.64, + "learning_rate": 1.2008321135274653e-07, + "loss": 0.1644, + "step": 12298 + }, + { + "epoch": 0.64, + "learning_rate": 1.200522176358418e-07, + "loss": 0.1495, + "step": 12299 + }, + { + "epoch": 0.64, + "learning_rate": 1.2002122620387555e-07, + "loss": 0.1687, + "step": 12300 + }, + { + "epoch": 0.64, + "learning_rate": 1.1999023705773352e-07, + "loss": 0.1401, + "step": 12301 + }, + { + "epoch": 0.64, + "learning_rate": 1.1995925019830135e-07, + "loss": 0.1082, + "step": 12302 + }, + { + "epoch": 0.64, + "learning_rate": 1.1992826562646477e-07, + "loss": 0.1163, + "step": 12303 + }, + { + "epoch": 0.64, + "learning_rate": 1.198972833431092e-07, + "loss": 0.1337, + "step": 12304 + }, + { + "epoch": 0.64, + "learning_rate": 1.1986630334912026e-07, + "loss": 0.1253, + "step": 12305 + }, + { + "epoch": 0.64, + "learning_rate": 1.1983532564538323e-07, + "loss": 0.1522, + "step": 12306 + }, + { + "epoch": 0.64, + "learning_rate": 1.198043502327836e-07, + "loss": 0.1143, + "step": 12307 + }, + { + "epoch": 0.64, + "learning_rate": 1.1977337711220657e-07, + "loss": 0.1595, + "step": 12308 + }, + { + "epoch": 0.64, + "learning_rate": 1.1974240628453746e-07, + "loss": 0.1676, + "step": 12309 + }, + { + "epoch": 0.64, + "learning_rate": 1.1971143775066135e-07, + "loss": 0.1275, + "step": 12310 + }, + { + "epoch": 0.64, + "learning_rate": 1.1968047151146325e-07, + "loss": 0.1592, + "step": 12311 + }, + { + "epoch": 0.64, + "learning_rate": 1.1964950756782836e-07, + "loss": 0.1444, + "step": 12312 + }, + { + "epoch": 0.64, + "learning_rate": 1.196185459206415e-07, + "loss": 0.1436, + "step": 12313 + }, + { + "epoch": 0.64, + "learning_rate": 1.195875865707877e-07, + "loss": 0.135, + "step": 12314 + }, + { + "epoch": 0.64, + "learning_rate": 1.1955662951915165e-07, + "loss": 0.131, + "step": 12315 + }, + { + "epoch": 0.64, + "learning_rate": 1.1952567476661822e-07, + "loss": 0.1751, + "step": 12316 + }, + { + "epoch": 0.64, + "learning_rate": 1.1949472231407214e-07, + "loss": 0.1248, + "step": 12317 + }, + { + "epoch": 0.64, + "learning_rate": 1.1946377216239788e-07, + "loss": 0.1747, + "step": 12318 + }, + { + "epoch": 0.64, + "learning_rate": 1.1943282431248012e-07, + "loss": 0.1179, + "step": 12319 + }, + { + "epoch": 0.64, + "learning_rate": 1.1940187876520332e-07, + "loss": 0.1431, + "step": 12320 + }, + { + "epoch": 0.64, + "learning_rate": 1.1937093552145198e-07, + "loss": 0.1421, + "step": 12321 + }, + { + "epoch": 0.64, + "learning_rate": 1.1933999458211038e-07, + "loss": 0.1387, + "step": 12322 + }, + { + "epoch": 0.64, + "learning_rate": 1.1930905594806292e-07, + "loss": 0.1624, + "step": 12323 + }, + { + "epoch": 0.64, + "learning_rate": 1.1927811962019376e-07, + "loss": 0.114, + "step": 12324 + }, + { + "epoch": 0.64, + "learning_rate": 1.1924718559938716e-07, + "loss": 0.1615, + "step": 12325 + }, + { + "epoch": 0.64, + "learning_rate": 1.1921625388652717e-07, + "loss": 0.0921, + "step": 12326 + }, + { + "epoch": 0.64, + "learning_rate": 1.1918532448249776e-07, + "loss": 0.1727, + "step": 12327 + }, + { + "epoch": 0.64, + "learning_rate": 1.19154397388183e-07, + "loss": 0.1206, + "step": 12328 + }, + { + "epoch": 0.64, + "learning_rate": 1.1912347260446675e-07, + "loss": 0.1298, + "step": 12329 + }, + { + "epoch": 0.64, + "learning_rate": 1.1909255013223294e-07, + "loss": 0.1347, + "step": 12330 + }, + { + "epoch": 0.64, + "learning_rate": 1.190616299723652e-07, + "loss": 0.136, + "step": 12331 + }, + { + "epoch": 0.64, + "learning_rate": 1.1903071212574738e-07, + "loss": 0.141, + "step": 12332 + }, + { + "epoch": 0.64, + "learning_rate": 1.1899979659326302e-07, + "loss": 0.1235, + "step": 12333 + }, + { + "epoch": 0.64, + "learning_rate": 1.1896888337579583e-07, + "loss": 0.172, + "step": 12334 + }, + { + "epoch": 0.64, + "learning_rate": 1.1893797247422919e-07, + "loss": 0.1687, + "step": 12335 + }, + { + "epoch": 0.64, + "learning_rate": 1.1890706388944653e-07, + "loss": 0.1231, + "step": 12336 + }, + { + "epoch": 0.64, + "learning_rate": 1.1887615762233134e-07, + "loss": 0.1203, + "step": 12337 + }, + { + "epoch": 0.64, + "learning_rate": 1.1884525367376687e-07, + "loss": 0.2026, + "step": 12338 + }, + { + "epoch": 0.64, + "learning_rate": 1.1881435204463639e-07, + "loss": 0.1198, + "step": 12339 + }, + { + "epoch": 0.64, + "learning_rate": 1.1878345273582304e-07, + "loss": 0.1316, + "step": 12340 + }, + { + "epoch": 0.64, + "learning_rate": 1.1875255574820999e-07, + "loss": 0.1368, + "step": 12341 + }, + { + "epoch": 0.64, + "learning_rate": 1.1872166108268022e-07, + "loss": 0.1421, + "step": 12342 + }, + { + "epoch": 0.64, + "learning_rate": 1.1869076874011685e-07, + "loss": 0.131, + "step": 12343 + }, + { + "epoch": 0.64, + "learning_rate": 1.1865987872140266e-07, + "loss": 0.1395, + "step": 12344 + }, + { + "epoch": 0.64, + "learning_rate": 1.1862899102742047e-07, + "loss": 0.1747, + "step": 12345 + }, + { + "epoch": 0.64, + "learning_rate": 1.1859810565905321e-07, + "loss": 0.1164, + "step": 12346 + }, + { + "epoch": 0.64, + "learning_rate": 1.1856722261718344e-07, + "loss": 0.1116, + "step": 12347 + }, + { + "epoch": 0.64, + "learning_rate": 1.1853634190269395e-07, + "loss": 0.1449, + "step": 12348 + }, + { + "epoch": 0.64, + "learning_rate": 1.185054635164672e-07, + "loss": 0.1317, + "step": 12349 + }, + { + "epoch": 0.64, + "learning_rate": 1.1847458745938581e-07, + "loss": 0.1753, + "step": 12350 + }, + { + "epoch": 0.64, + "learning_rate": 1.1844371373233213e-07, + "loss": 0.1137, + "step": 12351 + }, + { + "epoch": 0.64, + "learning_rate": 1.1841284233618869e-07, + "loss": 0.1685, + "step": 12352 + }, + { + "epoch": 0.64, + "learning_rate": 1.1838197327183768e-07, + "loss": 0.1494, + "step": 12353 + }, + { + "epoch": 0.64, + "learning_rate": 1.1835110654016131e-07, + "loss": 0.1194, + "step": 12354 + }, + { + "epoch": 0.64, + "learning_rate": 1.183202421420419e-07, + "loss": 0.118, + "step": 12355 + }, + { + "epoch": 0.64, + "learning_rate": 1.1828938007836145e-07, + "loss": 0.1672, + "step": 12356 + }, + { + "epoch": 0.64, + "learning_rate": 1.1825852035000208e-07, + "loss": 0.1199, + "step": 12357 + }, + { + "epoch": 0.65, + "learning_rate": 1.1822766295784576e-07, + "loss": 0.1537, + "step": 12358 + }, + { + "epoch": 0.65, + "learning_rate": 1.1819680790277442e-07, + "loss": 0.1293, + "step": 12359 + }, + { + "epoch": 0.65, + "learning_rate": 1.1816595518566992e-07, + "loss": 0.1174, + "step": 12360 + }, + { + "epoch": 0.65, + "learning_rate": 1.1813510480741393e-07, + "loss": 0.1245, + "step": 12361 + }, + { + "epoch": 0.65, + "learning_rate": 1.1810425676888831e-07, + "loss": 0.1394, + "step": 12362 + }, + { + "epoch": 0.65, + "learning_rate": 1.180734110709746e-07, + "loss": 0.1643, + "step": 12363 + }, + { + "epoch": 0.65, + "learning_rate": 1.1804256771455445e-07, + "loss": 0.1469, + "step": 12364 + }, + { + "epoch": 0.65, + "learning_rate": 1.1801172670050932e-07, + "loss": 0.1376, + "step": 12365 + }, + { + "epoch": 0.65, + "learning_rate": 1.1798088802972073e-07, + "loss": 0.1026, + "step": 12366 + }, + { + "epoch": 0.65, + "learning_rate": 1.1795005170306999e-07, + "loss": 0.1447, + "step": 12367 + }, + { + "epoch": 0.65, + "learning_rate": 1.179192177214385e-07, + "loss": 0.1148, + "step": 12368 + }, + { + "epoch": 0.65, + "learning_rate": 1.1788838608570745e-07, + "loss": 0.1155, + "step": 12369 + }, + { + "epoch": 0.65, + "learning_rate": 1.17857556796758e-07, + "loss": 0.1125, + "step": 12370 + }, + { + "epoch": 0.65, + "learning_rate": 1.1782672985547128e-07, + "loss": 0.1676, + "step": 12371 + }, + { + "epoch": 0.65, + "learning_rate": 1.1779590526272833e-07, + "loss": 0.1538, + "step": 12372 + }, + { + "epoch": 0.65, + "learning_rate": 1.1776508301941018e-07, + "loss": 0.1155, + "step": 12373 + }, + { + "epoch": 0.65, + "learning_rate": 1.1773426312639765e-07, + "loss": 0.1251, + "step": 12374 + }, + { + "epoch": 0.65, + "learning_rate": 1.1770344558457168e-07, + "loss": 0.1422, + "step": 12375 + }, + { + "epoch": 0.65, + "learning_rate": 1.1767263039481297e-07, + "loss": 0.1088, + "step": 12376 + }, + { + "epoch": 0.65, + "learning_rate": 1.1764181755800236e-07, + "loss": 0.106, + "step": 12377 + }, + { + "epoch": 0.65, + "learning_rate": 1.1761100707502034e-07, + "loss": 0.1383, + "step": 12378 + }, + { + "epoch": 0.65, + "learning_rate": 1.175801989467475e-07, + "loss": 0.172, + "step": 12379 + }, + { + "epoch": 0.65, + "learning_rate": 1.1754939317406445e-07, + "loss": 0.1504, + "step": 12380 + }, + { + "epoch": 0.65, + "learning_rate": 1.175185897578515e-07, + "loss": 0.1476, + "step": 12381 + }, + { + "epoch": 0.65, + "learning_rate": 1.1748778869898916e-07, + "loss": 0.1235, + "step": 12382 + }, + { + "epoch": 0.65, + "learning_rate": 1.1745698999835762e-07, + "loss": 0.1241, + "step": 12383 + }, + { + "epoch": 0.65, + "learning_rate": 1.1742619365683721e-07, + "loss": 0.1288, + "step": 12384 + }, + { + "epoch": 0.65, + "learning_rate": 1.1739539967530805e-07, + "loss": 0.1438, + "step": 12385 + }, + { + "epoch": 0.65, + "learning_rate": 1.1736460805465029e-07, + "loss": 0.1445, + "step": 12386 + }, + { + "epoch": 0.65, + "learning_rate": 1.1733381879574394e-07, + "loss": 0.1281, + "step": 12387 + }, + { + "epoch": 0.65, + "learning_rate": 1.173030318994689e-07, + "loss": 0.1606, + "step": 12388 + }, + { + "epoch": 0.65, + "learning_rate": 1.1727224736670516e-07, + "loss": 0.1235, + "step": 12389 + }, + { + "epoch": 0.65, + "learning_rate": 1.1724146519833249e-07, + "loss": 0.1663, + "step": 12390 + }, + { + "epoch": 0.65, + "learning_rate": 1.1721068539523073e-07, + "loss": 0.1314, + "step": 12391 + }, + { + "epoch": 0.65, + "learning_rate": 1.1717990795827954e-07, + "loss": 0.1514, + "step": 12392 + }, + { + "epoch": 0.65, + "learning_rate": 1.1714913288835857e-07, + "loss": 0.1216, + "step": 12393 + }, + { + "epoch": 0.65, + "learning_rate": 1.1711836018634738e-07, + "loss": 0.1355, + "step": 12394 + }, + { + "epoch": 0.65, + "learning_rate": 1.1708758985312539e-07, + "loss": 0.1261, + "step": 12395 + }, + { + "epoch": 0.65, + "learning_rate": 1.1705682188957213e-07, + "loss": 0.1512, + "step": 12396 + }, + { + "epoch": 0.65, + "learning_rate": 1.170260562965669e-07, + "loss": 0.1435, + "step": 12397 + }, + { + "epoch": 0.65, + "learning_rate": 1.1699529307498903e-07, + "loss": 0.1243, + "step": 12398 + }, + { + "epoch": 0.65, + "learning_rate": 1.169645322257177e-07, + "loss": 0.1197, + "step": 12399 + }, + { + "epoch": 0.65, + "learning_rate": 1.1693377374963209e-07, + "loss": 0.1195, + "step": 12400 + }, + { + "epoch": 0.65, + "learning_rate": 1.169030176476113e-07, + "loss": 0.1401, + "step": 12401 + }, + { + "epoch": 0.65, + "learning_rate": 1.1687226392053435e-07, + "loss": 0.1078, + "step": 12402 + }, + { + "epoch": 0.65, + "learning_rate": 1.1684151256928028e-07, + "loss": 0.1763, + "step": 12403 + }, + { + "epoch": 0.65, + "learning_rate": 1.1681076359472775e-07, + "loss": 0.1049, + "step": 12404 + }, + { + "epoch": 0.65, + "learning_rate": 1.167800169977558e-07, + "loss": 0.1816, + "step": 12405 + }, + { + "epoch": 0.65, + "learning_rate": 1.1674927277924305e-07, + "loss": 0.1328, + "step": 12406 + }, + { + "epoch": 0.65, + "learning_rate": 1.1671853094006814e-07, + "loss": 0.1255, + "step": 12407 + }, + { + "epoch": 0.65, + "learning_rate": 1.1668779148110987e-07, + "loss": 0.1228, + "step": 12408 + }, + { + "epoch": 0.65, + "learning_rate": 1.1665705440324667e-07, + "loss": 0.1272, + "step": 12409 + }, + { + "epoch": 0.65, + "learning_rate": 1.1662631970735701e-07, + "loss": 0.1307, + "step": 12410 + }, + { + "epoch": 0.65, + "learning_rate": 1.1659558739431928e-07, + "loss": 0.1396, + "step": 12411 + }, + { + "epoch": 0.65, + "learning_rate": 1.1656485746501199e-07, + "loss": 0.1377, + "step": 12412 + }, + { + "epoch": 0.65, + "learning_rate": 1.1653412992031322e-07, + "loss": 0.143, + "step": 12413 + }, + { + "epoch": 0.65, + "learning_rate": 1.1650340476110123e-07, + "loss": 0.1348, + "step": 12414 + }, + { + "epoch": 0.65, + "learning_rate": 1.1647268198825409e-07, + "loss": 0.111, + "step": 12415 + }, + { + "epoch": 0.65, + "learning_rate": 1.1644196160265e-07, + "loss": 0.1492, + "step": 12416 + }, + { + "epoch": 0.65, + "learning_rate": 1.1641124360516692e-07, + "loss": 0.1693, + "step": 12417 + }, + { + "epoch": 0.65, + "learning_rate": 1.1638052799668279e-07, + "loss": 0.1757, + "step": 12418 + }, + { + "epoch": 0.65, + "learning_rate": 1.1634981477807533e-07, + "loss": 0.1266, + "step": 12419 + }, + { + "epoch": 0.65, + "learning_rate": 1.1631910395022262e-07, + "loss": 0.1256, + "step": 12420 + }, + { + "epoch": 0.65, + "learning_rate": 1.1628839551400216e-07, + "loss": 0.1661, + "step": 12421 + }, + { + "epoch": 0.65, + "learning_rate": 1.162576894702916e-07, + "loss": 0.1249, + "step": 12422 + }, + { + "epoch": 0.65, + "learning_rate": 1.1622698581996865e-07, + "loss": 0.1813, + "step": 12423 + }, + { + "epoch": 0.65, + "learning_rate": 1.1619628456391084e-07, + "loss": 0.1546, + "step": 12424 + }, + { + "epoch": 0.65, + "learning_rate": 1.1616558570299551e-07, + "loss": 0.1395, + "step": 12425 + }, + { + "epoch": 0.65, + "learning_rate": 1.1613488923810007e-07, + "loss": 0.1209, + "step": 12426 + }, + { + "epoch": 0.65, + "learning_rate": 1.1610419517010192e-07, + "loss": 0.1202, + "step": 12427 + }, + { + "epoch": 0.65, + "learning_rate": 1.1607350349987828e-07, + "loss": 0.1506, + "step": 12428 + }, + { + "epoch": 0.65, + "learning_rate": 1.1604281422830629e-07, + "loss": 0.1382, + "step": 12429 + }, + { + "epoch": 0.65, + "learning_rate": 1.160121273562631e-07, + "loss": 0.117, + "step": 12430 + }, + { + "epoch": 0.65, + "learning_rate": 1.159814428846257e-07, + "loss": 0.1267, + "step": 12431 + }, + { + "epoch": 0.65, + "learning_rate": 1.1595076081427114e-07, + "loss": 0.1556, + "step": 12432 + }, + { + "epoch": 0.65, + "learning_rate": 1.1592008114607616e-07, + "loss": 0.151, + "step": 12433 + }, + { + "epoch": 0.65, + "learning_rate": 1.1588940388091782e-07, + "loss": 0.1318, + "step": 12434 + }, + { + "epoch": 0.65, + "learning_rate": 1.1585872901967278e-07, + "loss": 0.1329, + "step": 12435 + }, + { + "epoch": 0.65, + "learning_rate": 1.1582805656321775e-07, + "loss": 0.1208, + "step": 12436 + }, + { + "epoch": 0.65, + "learning_rate": 1.1579738651242936e-07, + "loss": 0.0979, + "step": 12437 + }, + { + "epoch": 0.65, + "learning_rate": 1.1576671886818417e-07, + "loss": 0.1616, + "step": 12438 + }, + { + "epoch": 0.65, + "learning_rate": 1.1573605363135867e-07, + "loss": 0.112, + "step": 12439 + }, + { + "epoch": 0.65, + "learning_rate": 1.157053908028292e-07, + "loss": 0.1581, + "step": 12440 + }, + { + "epoch": 0.65, + "learning_rate": 1.1567473038347228e-07, + "loss": 0.1443, + "step": 12441 + }, + { + "epoch": 0.65, + "learning_rate": 1.1564407237416416e-07, + "loss": 0.1548, + "step": 12442 + }, + { + "epoch": 0.65, + "learning_rate": 1.1561341677578096e-07, + "loss": 0.1433, + "step": 12443 + }, + { + "epoch": 0.65, + "learning_rate": 1.1558276358919883e-07, + "loss": 0.1606, + "step": 12444 + }, + { + "epoch": 0.65, + "learning_rate": 1.15552112815294e-07, + "loss": 0.1801, + "step": 12445 + }, + { + "epoch": 0.65, + "learning_rate": 1.1552146445494244e-07, + "loss": 0.1635, + "step": 12446 + }, + { + "epoch": 0.65, + "learning_rate": 1.154908185090199e-07, + "loss": 0.1363, + "step": 12447 + }, + { + "epoch": 0.65, + "learning_rate": 1.1546017497840244e-07, + "loss": 0.1234, + "step": 12448 + }, + { + "epoch": 0.65, + "learning_rate": 1.1542953386396583e-07, + "loss": 0.1147, + "step": 12449 + }, + { + "epoch": 0.65, + "learning_rate": 1.1539889516658581e-07, + "loss": 0.1108, + "step": 12450 + }, + { + "epoch": 0.65, + "learning_rate": 1.153682588871379e-07, + "loss": 0.1326, + "step": 12451 + }, + { + "epoch": 0.65, + "learning_rate": 1.1533762502649796e-07, + "loss": 0.1986, + "step": 12452 + }, + { + "epoch": 0.65, + "learning_rate": 1.1530699358554132e-07, + "loss": 0.1324, + "step": 12453 + }, + { + "epoch": 0.65, + "learning_rate": 1.1527636456514354e-07, + "loss": 0.1159, + "step": 12454 + }, + { + "epoch": 0.65, + "learning_rate": 1.1524573796617993e-07, + "loss": 0.1274, + "step": 12455 + }, + { + "epoch": 0.65, + "learning_rate": 1.1521511378952588e-07, + "loss": 0.1294, + "step": 12456 + }, + { + "epoch": 0.65, + "learning_rate": 1.151844920360566e-07, + "loss": 0.1495, + "step": 12457 + }, + { + "epoch": 0.65, + "learning_rate": 1.1515387270664723e-07, + "loss": 0.1212, + "step": 12458 + }, + { + "epoch": 0.65, + "learning_rate": 1.1512325580217295e-07, + "loss": 0.1233, + "step": 12459 + }, + { + "epoch": 0.65, + "learning_rate": 1.1509264132350885e-07, + "loss": 0.1498, + "step": 12460 + }, + { + "epoch": 0.65, + "learning_rate": 1.1506202927152977e-07, + "loss": 0.1584, + "step": 12461 + }, + { + "epoch": 0.65, + "learning_rate": 1.1503141964711065e-07, + "loss": 0.14, + "step": 12462 + }, + { + "epoch": 0.65, + "learning_rate": 1.150008124511265e-07, + "loss": 0.1634, + "step": 12463 + }, + { + "epoch": 0.65, + "learning_rate": 1.1497020768445187e-07, + "loss": 0.1162, + "step": 12464 + }, + { + "epoch": 0.65, + "learning_rate": 1.1493960534796146e-07, + "loss": 0.1351, + "step": 12465 + }, + { + "epoch": 0.65, + "learning_rate": 1.1490900544253004e-07, + "loss": 0.153, + "step": 12466 + }, + { + "epoch": 0.65, + "learning_rate": 1.1487840796903208e-07, + "loss": 0.1325, + "step": 12467 + }, + { + "epoch": 0.65, + "learning_rate": 1.1484781292834207e-07, + "loss": 0.1369, + "step": 12468 + }, + { + "epoch": 0.65, + "learning_rate": 1.1481722032133438e-07, + "loss": 0.1392, + "step": 12469 + }, + { + "epoch": 0.65, + "learning_rate": 1.147866301488835e-07, + "loss": 0.1334, + "step": 12470 + }, + { + "epoch": 0.65, + "learning_rate": 1.1475604241186361e-07, + "loss": 0.1356, + "step": 12471 + }, + { + "epoch": 0.65, + "learning_rate": 1.1472545711114894e-07, + "loss": 0.1322, + "step": 12472 + }, + { + "epoch": 0.65, + "learning_rate": 1.1469487424761363e-07, + "loss": 0.1325, + "step": 12473 + }, + { + "epoch": 0.65, + "learning_rate": 1.1466429382213174e-07, + "loss": 0.0909, + "step": 12474 + }, + { + "epoch": 0.65, + "learning_rate": 1.1463371583557728e-07, + "loss": 0.1592, + "step": 12475 + }, + { + "epoch": 0.65, + "learning_rate": 1.1460314028882412e-07, + "loss": 0.1588, + "step": 12476 + }, + { + "epoch": 0.65, + "learning_rate": 1.1457256718274621e-07, + "loss": 0.129, + "step": 12477 + }, + { + "epoch": 0.65, + "learning_rate": 1.1454199651821734e-07, + "loss": 0.1483, + "step": 12478 + }, + { + "epoch": 0.65, + "learning_rate": 1.1451142829611118e-07, + "loss": 0.1266, + "step": 12479 + }, + { + "epoch": 0.65, + "learning_rate": 1.1448086251730142e-07, + "loss": 0.1454, + "step": 12480 + }, + { + "epoch": 0.65, + "learning_rate": 1.1445029918266163e-07, + "loss": 0.1335, + "step": 12481 + }, + { + "epoch": 0.65, + "learning_rate": 1.144197382930653e-07, + "loss": 0.1603, + "step": 12482 + }, + { + "epoch": 0.65, + "learning_rate": 1.1438917984938581e-07, + "loss": 0.0945, + "step": 12483 + }, + { + "epoch": 0.65, + "learning_rate": 1.1435862385249669e-07, + "loss": 0.1153, + "step": 12484 + }, + { + "epoch": 0.65, + "learning_rate": 1.1432807030327114e-07, + "loss": 0.1412, + "step": 12485 + }, + { + "epoch": 0.65, + "learning_rate": 1.1429751920258243e-07, + "loss": 0.1493, + "step": 12486 + }, + { + "epoch": 0.65, + "learning_rate": 1.1426697055130361e-07, + "loss": 0.151, + "step": 12487 + }, + { + "epoch": 0.65, + "learning_rate": 1.1423642435030798e-07, + "loss": 0.1181, + "step": 12488 + }, + { + "epoch": 0.65, + "learning_rate": 1.1420588060046852e-07, + "loss": 0.1285, + "step": 12489 + }, + { + "epoch": 0.65, + "learning_rate": 1.1417533930265792e-07, + "loss": 0.1322, + "step": 12490 + }, + { + "epoch": 0.65, + "learning_rate": 1.1414480045774939e-07, + "loss": 0.1403, + "step": 12491 + }, + { + "epoch": 0.65, + "learning_rate": 1.1411426406661554e-07, + "loss": 0.1773, + "step": 12492 + }, + { + "epoch": 0.65, + "learning_rate": 1.1408373013012924e-07, + "loss": 0.1638, + "step": 12493 + }, + { + "epoch": 0.65, + "learning_rate": 1.1405319864916298e-07, + "loss": 0.1492, + "step": 12494 + }, + { + "epoch": 0.65, + "learning_rate": 1.1402266962458958e-07, + "loss": 0.1474, + "step": 12495 + }, + { + "epoch": 0.65, + "learning_rate": 1.1399214305728149e-07, + "loss": 0.1034, + "step": 12496 + }, + { + "epoch": 0.65, + "learning_rate": 1.1396161894811114e-07, + "loss": 0.1325, + "step": 12497 + }, + { + "epoch": 0.65, + "learning_rate": 1.1393109729795089e-07, + "loss": 0.1344, + "step": 12498 + }, + { + "epoch": 0.65, + "learning_rate": 1.1390057810767317e-07, + "loss": 0.1144, + "step": 12499 + }, + { + "epoch": 0.65, + "learning_rate": 1.1387006137815011e-07, + "loss": 0.1092, + "step": 12500 + }, + { + "epoch": 0.65, + "learning_rate": 1.1383954711025391e-07, + "loss": 0.1393, + "step": 12501 + }, + { + "epoch": 0.65, + "learning_rate": 1.1380903530485677e-07, + "loss": 0.1211, + "step": 12502 + }, + { + "epoch": 0.65, + "learning_rate": 1.1377852596283069e-07, + "loss": 0.1517, + "step": 12503 + }, + { + "epoch": 0.65, + "learning_rate": 1.1374801908504764e-07, + "loss": 0.1177, + "step": 12504 + }, + { + "epoch": 0.65, + "learning_rate": 1.1371751467237938e-07, + "loss": 0.1578, + "step": 12505 + }, + { + "epoch": 0.65, + "learning_rate": 1.1368701272569803e-07, + "loss": 0.1047, + "step": 12506 + }, + { + "epoch": 0.65, + "learning_rate": 1.1365651324587511e-07, + "loss": 0.1093, + "step": 12507 + }, + { + "epoch": 0.65, + "learning_rate": 1.1362601623378228e-07, + "loss": 0.115, + "step": 12508 + }, + { + "epoch": 0.65, + "learning_rate": 1.1359552169029134e-07, + "loss": 0.1213, + "step": 12509 + }, + { + "epoch": 0.65, + "learning_rate": 1.1356502961627375e-07, + "loss": 0.1785, + "step": 12510 + }, + { + "epoch": 0.65, + "learning_rate": 1.1353454001260097e-07, + "loss": 0.111, + "step": 12511 + }, + { + "epoch": 0.65, + "learning_rate": 1.1350405288014431e-07, + "loss": 0.138, + "step": 12512 + }, + { + "epoch": 0.65, + "learning_rate": 1.1347356821977535e-07, + "loss": 0.1107, + "step": 12513 + }, + { + "epoch": 0.65, + "learning_rate": 1.1344308603236523e-07, + "loss": 0.1271, + "step": 12514 + }, + { + "epoch": 0.65, + "learning_rate": 1.13412606318785e-07, + "loss": 0.1531, + "step": 12515 + }, + { + "epoch": 0.65, + "learning_rate": 1.1338212907990596e-07, + "loss": 0.1391, + "step": 12516 + }, + { + "epoch": 0.65, + "learning_rate": 1.1335165431659914e-07, + "loss": 0.1371, + "step": 12517 + }, + { + "epoch": 0.65, + "learning_rate": 1.1332118202973551e-07, + "loss": 0.1911, + "step": 12518 + }, + { + "epoch": 0.65, + "learning_rate": 1.1329071222018582e-07, + "loss": 0.1025, + "step": 12519 + }, + { + "epoch": 0.65, + "learning_rate": 1.1326024488882118e-07, + "loss": 0.1217, + "step": 12520 + }, + { + "epoch": 0.65, + "learning_rate": 1.132297800365122e-07, + "loss": 0.186, + "step": 12521 + }, + { + "epoch": 0.65, + "learning_rate": 1.1319931766412961e-07, + "loss": 0.1462, + "step": 12522 + }, + { + "epoch": 0.65, + "learning_rate": 1.1316885777254402e-07, + "loss": 0.1377, + "step": 12523 + }, + { + "epoch": 0.65, + "learning_rate": 1.1313840036262599e-07, + "loss": 0.1084, + "step": 12524 + }, + { + "epoch": 0.65, + "learning_rate": 1.1310794543524602e-07, + "loss": 0.1561, + "step": 12525 + }, + { + "epoch": 0.65, + "learning_rate": 1.1307749299127452e-07, + "loss": 0.1408, + "step": 12526 + }, + { + "epoch": 0.65, + "learning_rate": 1.1304704303158175e-07, + "loss": 0.1226, + "step": 12527 + }, + { + "epoch": 0.65, + "learning_rate": 1.1301659555703811e-07, + "loss": 0.1193, + "step": 12528 + }, + { + "epoch": 0.65, + "learning_rate": 1.1298615056851378e-07, + "loss": 0.137, + "step": 12529 + }, + { + "epoch": 0.65, + "learning_rate": 1.1295570806687882e-07, + "loss": 0.1282, + "step": 12530 + }, + { + "epoch": 0.65, + "learning_rate": 1.1292526805300329e-07, + "loss": 0.1607, + "step": 12531 + }, + { + "epoch": 0.65, + "learning_rate": 1.1289483052775732e-07, + "loss": 0.1237, + "step": 12532 + }, + { + "epoch": 0.65, + "learning_rate": 1.1286439549201068e-07, + "loss": 0.1409, + "step": 12533 + }, + { + "epoch": 0.65, + "learning_rate": 1.1283396294663319e-07, + "loss": 0.1435, + "step": 12534 + }, + { + "epoch": 0.65, + "learning_rate": 1.1280353289249473e-07, + "loss": 0.1417, + "step": 12535 + }, + { + "epoch": 0.65, + "learning_rate": 1.1277310533046498e-07, + "loss": 0.1338, + "step": 12536 + }, + { + "epoch": 0.65, + "learning_rate": 1.1274268026141354e-07, + "loss": 0.1343, + "step": 12537 + }, + { + "epoch": 0.65, + "learning_rate": 1.1271225768620991e-07, + "loss": 0.1504, + "step": 12538 + }, + { + "epoch": 0.65, + "learning_rate": 1.1268183760572374e-07, + "loss": 0.1744, + "step": 12539 + }, + { + "epoch": 0.65, + "learning_rate": 1.1265142002082443e-07, + "loss": 0.1186, + "step": 12540 + }, + { + "epoch": 0.65, + "learning_rate": 1.1262100493238109e-07, + "loss": 0.121, + "step": 12541 + }, + { + "epoch": 0.65, + "learning_rate": 1.1259059234126324e-07, + "loss": 0.1136, + "step": 12542 + }, + { + "epoch": 0.65, + "learning_rate": 1.1256018224834001e-07, + "loss": 0.1151, + "step": 12543 + }, + { + "epoch": 0.65, + "learning_rate": 1.1252977465448052e-07, + "loss": 0.1974, + "step": 12544 + }, + { + "epoch": 0.65, + "learning_rate": 1.1249936956055379e-07, + "loss": 0.126, + "step": 12545 + }, + { + "epoch": 0.65, + "learning_rate": 1.1246896696742891e-07, + "loss": 0.1301, + "step": 12546 + }, + { + "epoch": 0.65, + "learning_rate": 1.1243856687597475e-07, + "loss": 0.1292, + "step": 12547 + }, + { + "epoch": 0.65, + "learning_rate": 1.1240816928706014e-07, + "loss": 0.1299, + "step": 12548 + }, + { + "epoch": 0.66, + "learning_rate": 1.1237777420155387e-07, + "loss": 0.149, + "step": 12549 + }, + { + "epoch": 0.66, + "learning_rate": 1.1234738162032464e-07, + "loss": 0.1159, + "step": 12550 + }, + { + "epoch": 0.66, + "learning_rate": 1.1231699154424111e-07, + "loss": 0.1132, + "step": 12551 + }, + { + "epoch": 0.66, + "learning_rate": 1.1228660397417169e-07, + "loss": 0.1654, + "step": 12552 + }, + { + "epoch": 0.66, + "learning_rate": 1.1225621891098507e-07, + "loss": 0.1202, + "step": 12553 + }, + { + "epoch": 0.66, + "learning_rate": 1.1222583635554961e-07, + "loss": 0.1284, + "step": 12554 + }, + { + "epoch": 0.66, + "learning_rate": 1.1219545630873362e-07, + "loss": 0.1392, + "step": 12555 + }, + { + "epoch": 0.66, + "learning_rate": 1.1216507877140531e-07, + "loss": 0.1314, + "step": 12556 + }, + { + "epoch": 0.66, + "learning_rate": 1.121347037444331e-07, + "loss": 0.133, + "step": 12557 + }, + { + "epoch": 0.66, + "learning_rate": 1.121043312286849e-07, + "loss": 0.1164, + "step": 12558 + }, + { + "epoch": 0.66, + "learning_rate": 1.1207396122502877e-07, + "loss": 0.1364, + "step": 12559 + }, + { + "epoch": 0.66, + "learning_rate": 1.1204359373433287e-07, + "loss": 0.1288, + "step": 12560 + }, + { + "epoch": 0.66, + "learning_rate": 1.1201322875746499e-07, + "loss": 0.1686, + "step": 12561 + }, + { + "epoch": 0.66, + "learning_rate": 1.11982866295293e-07, + "loss": 0.143, + "step": 12562 + }, + { + "epoch": 0.66, + "learning_rate": 1.1195250634868461e-07, + "loss": 0.1209, + "step": 12563 + }, + { + "epoch": 0.66, + "learning_rate": 1.1192214891850765e-07, + "loss": 0.1357, + "step": 12564 + }, + { + "epoch": 0.66, + "learning_rate": 1.1189179400562967e-07, + "loss": 0.1239, + "step": 12565 + }, + { + "epoch": 0.66, + "learning_rate": 1.1186144161091822e-07, + "loss": 0.1497, + "step": 12566 + }, + { + "epoch": 0.66, + "learning_rate": 1.1183109173524081e-07, + "loss": 0.1565, + "step": 12567 + }, + { + "epoch": 0.66, + "learning_rate": 1.1180074437946483e-07, + "loss": 0.1163, + "step": 12568 + }, + { + "epoch": 0.66, + "learning_rate": 1.1177039954445762e-07, + "loss": 0.1207, + "step": 12569 + }, + { + "epoch": 0.66, + "learning_rate": 1.1174005723108639e-07, + "loss": 0.1464, + "step": 12570 + }, + { + "epoch": 0.66, + "learning_rate": 1.1170971744021845e-07, + "loss": 0.1733, + "step": 12571 + }, + { + "epoch": 0.66, + "learning_rate": 1.1167938017272089e-07, + "loss": 0.1588, + "step": 12572 + }, + { + "epoch": 0.66, + "learning_rate": 1.1164904542946073e-07, + "loss": 0.2174, + "step": 12573 + }, + { + "epoch": 0.66, + "learning_rate": 1.1161871321130493e-07, + "loss": 0.1404, + "step": 12574 + }, + { + "epoch": 0.66, + "learning_rate": 1.1158838351912048e-07, + "loss": 0.1273, + "step": 12575 + }, + { + "epoch": 0.66, + "learning_rate": 1.1155805635377413e-07, + "loss": 0.1111, + "step": 12576 + }, + { + "epoch": 0.66, + "learning_rate": 1.1152773171613259e-07, + "loss": 0.1253, + "step": 12577 + }, + { + "epoch": 0.66, + "learning_rate": 1.1149740960706271e-07, + "loss": 0.1176, + "step": 12578 + }, + { + "epoch": 0.66, + "learning_rate": 1.1146709002743104e-07, + "loss": 0.1462, + "step": 12579 + }, + { + "epoch": 0.66, + "learning_rate": 1.114367729781041e-07, + "loss": 0.1454, + "step": 12580 + }, + { + "epoch": 0.66, + "learning_rate": 1.1140645845994829e-07, + "loss": 0.1135, + "step": 12581 + }, + { + "epoch": 0.66, + "learning_rate": 1.1137614647383022e-07, + "loss": 0.1454, + "step": 12582 + }, + { + "epoch": 0.66, + "learning_rate": 1.1134583702061613e-07, + "loss": 0.1737, + "step": 12583 + }, + { + "epoch": 0.66, + "learning_rate": 1.1131553010117212e-07, + "loss": 0.1398, + "step": 12584 + }, + { + "epoch": 0.66, + "learning_rate": 1.1128522571636457e-07, + "loss": 0.145, + "step": 12585 + }, + { + "epoch": 0.66, + "learning_rate": 1.1125492386705948e-07, + "loss": 0.1396, + "step": 12586 + }, + { + "epoch": 0.66, + "learning_rate": 1.1122462455412294e-07, + "loss": 0.1369, + "step": 12587 + }, + { + "epoch": 0.66, + "learning_rate": 1.1119432777842083e-07, + "loss": 0.0973, + "step": 12588 + }, + { + "epoch": 0.66, + "learning_rate": 1.111640335408192e-07, + "loss": 0.1362, + "step": 12589 + }, + { + "epoch": 0.66, + "learning_rate": 1.1113374184218378e-07, + "loss": 0.1524, + "step": 12590 + }, + { + "epoch": 0.66, + "learning_rate": 1.1110345268338033e-07, + "loss": 0.1198, + "step": 12591 + }, + { + "epoch": 0.66, + "learning_rate": 1.110731660652745e-07, + "loss": 0.1344, + "step": 12592 + }, + { + "epoch": 0.66, + "learning_rate": 1.1104288198873193e-07, + "loss": 0.1183, + "step": 12593 + }, + { + "epoch": 0.66, + "learning_rate": 1.1101260045461813e-07, + "loss": 0.1565, + "step": 12594 + }, + { + "epoch": 0.66, + "learning_rate": 1.1098232146379847e-07, + "loss": 0.134, + "step": 12595 + }, + { + "epoch": 0.66, + "learning_rate": 1.1095204501713852e-07, + "loss": 0.1694, + "step": 12596 + }, + { + "epoch": 0.66, + "learning_rate": 1.109217711155035e-07, + "loss": 0.1696, + "step": 12597 + }, + { + "epoch": 0.66, + "learning_rate": 1.1089149975975865e-07, + "loss": 0.1188, + "step": 12598 + }, + { + "epoch": 0.66, + "learning_rate": 1.1086123095076905e-07, + "loss": 0.1455, + "step": 12599 + }, + { + "epoch": 0.66, + "learning_rate": 1.1083096468940003e-07, + "loss": 0.1318, + "step": 12600 + }, + { + "epoch": 0.66, + "learning_rate": 1.108007009765164e-07, + "loss": 0.1277, + "step": 12601 + }, + { + "epoch": 0.66, + "learning_rate": 1.1077043981298308e-07, + "loss": 0.1398, + "step": 12602 + }, + { + "epoch": 0.66, + "learning_rate": 1.1074018119966513e-07, + "loss": 0.1323, + "step": 12603 + }, + { + "epoch": 0.66, + "learning_rate": 1.1070992513742725e-07, + "loss": 0.1348, + "step": 12604 + }, + { + "epoch": 0.66, + "learning_rate": 1.1067967162713419e-07, + "loss": 0.1252, + "step": 12605 + }, + { + "epoch": 0.66, + "learning_rate": 1.1064942066965049e-07, + "loss": 0.0974, + "step": 12606 + }, + { + "epoch": 0.66, + "learning_rate": 1.1061917226584094e-07, + "loss": 0.1118, + "step": 12607 + }, + { + "epoch": 0.66, + "learning_rate": 1.1058892641656994e-07, + "loss": 0.1552, + "step": 12608 + }, + { + "epoch": 0.66, + "learning_rate": 1.1055868312270195e-07, + "loss": 0.1585, + "step": 12609 + }, + { + "epoch": 0.66, + "learning_rate": 1.1052844238510133e-07, + "loss": 0.1212, + "step": 12610 + }, + { + "epoch": 0.66, + "learning_rate": 1.1049820420463233e-07, + "loss": 0.1335, + "step": 12611 + }, + { + "epoch": 0.66, + "learning_rate": 1.1046796858215926e-07, + "loss": 0.1639, + "step": 12612 + }, + { + "epoch": 0.66, + "learning_rate": 1.104377355185461e-07, + "loss": 0.122, + "step": 12613 + }, + { + "epoch": 0.66, + "learning_rate": 1.1040750501465712e-07, + "loss": 0.1254, + "step": 12614 + }, + { + "epoch": 0.66, + "learning_rate": 1.103772770713562e-07, + "loss": 0.1556, + "step": 12615 + }, + { + "epoch": 0.66, + "learning_rate": 1.1034705168950736e-07, + "loss": 0.1203, + "step": 12616 + }, + { + "epoch": 0.66, + "learning_rate": 1.1031682886997434e-07, + "loss": 0.1172, + "step": 12617 + }, + { + "epoch": 0.66, + "learning_rate": 1.1028660861362098e-07, + "loss": 0.1219, + "step": 12618 + }, + { + "epoch": 0.66, + "learning_rate": 1.10256390921311e-07, + "loss": 0.1201, + "step": 12619 + }, + { + "epoch": 0.66, + "learning_rate": 1.1022617579390794e-07, + "loss": 0.1025, + "step": 12620 + }, + { + "epoch": 0.66, + "learning_rate": 1.1019596323227549e-07, + "loss": 0.1608, + "step": 12621 + }, + { + "epoch": 0.66, + "learning_rate": 1.1016575323727706e-07, + "loss": 0.1348, + "step": 12622 + }, + { + "epoch": 0.66, + "learning_rate": 1.1013554580977613e-07, + "loss": 0.1322, + "step": 12623 + }, + { + "epoch": 0.66, + "learning_rate": 1.1010534095063588e-07, + "loss": 0.1336, + "step": 12624 + }, + { + "epoch": 0.66, + "learning_rate": 1.1007513866071976e-07, + "loss": 0.1185, + "step": 12625 + }, + { + "epoch": 0.66, + "learning_rate": 1.1004493894089099e-07, + "loss": 0.1137, + "step": 12626 + }, + { + "epoch": 0.66, + "learning_rate": 1.1001474179201245e-07, + "loss": 0.1198, + "step": 12627 + }, + { + "epoch": 0.66, + "learning_rate": 1.0998454721494737e-07, + "loss": 0.1502, + "step": 12628 + }, + { + "epoch": 0.66, + "learning_rate": 1.0995435521055869e-07, + "loss": 0.1364, + "step": 12629 + }, + { + "epoch": 0.66, + "learning_rate": 1.0992416577970931e-07, + "loss": 0.1355, + "step": 12630 + }, + { + "epoch": 0.66, + "learning_rate": 1.0989397892326196e-07, + "loss": 0.1399, + "step": 12631 + }, + { + "epoch": 0.66, + "learning_rate": 1.0986379464207955e-07, + "loss": 0.11, + "step": 12632 + }, + { + "epoch": 0.66, + "learning_rate": 1.0983361293702469e-07, + "loss": 0.1561, + "step": 12633 + }, + { + "epoch": 0.66, + "learning_rate": 1.0980343380896006e-07, + "loss": 0.1134, + "step": 12634 + }, + { + "epoch": 0.66, + "learning_rate": 1.0977325725874795e-07, + "loss": 0.1291, + "step": 12635 + }, + { + "epoch": 0.66, + "learning_rate": 1.0974308328725106e-07, + "loss": 0.1362, + "step": 12636 + }, + { + "epoch": 0.66, + "learning_rate": 1.0971291189533171e-07, + "loss": 0.1591, + "step": 12637 + }, + { + "epoch": 0.66, + "learning_rate": 1.0968274308385217e-07, + "loss": 0.1467, + "step": 12638 + }, + { + "epoch": 0.66, + "learning_rate": 1.0965257685367462e-07, + "loss": 0.1345, + "step": 12639 + }, + { + "epoch": 0.66, + "learning_rate": 1.0962241320566139e-07, + "loss": 0.1343, + "step": 12640 + }, + { + "epoch": 0.66, + "learning_rate": 1.0959225214067447e-07, + "loss": 0.1386, + "step": 12641 + }, + { + "epoch": 0.66, + "learning_rate": 1.0956209365957589e-07, + "loss": 0.146, + "step": 12642 + }, + { + "epoch": 0.66, + "learning_rate": 1.095319377632276e-07, + "loss": 0.1302, + "step": 12643 + }, + { + "epoch": 0.66, + "learning_rate": 1.0950178445249143e-07, + "loss": 0.1498, + "step": 12644 + }, + { + "epoch": 0.66, + "learning_rate": 1.0947163372822919e-07, + "loss": 0.1538, + "step": 12645 + }, + { + "epoch": 0.66, + "learning_rate": 1.0944148559130255e-07, + "loss": 0.1051, + "step": 12646 + }, + { + "epoch": 0.66, + "learning_rate": 1.0941134004257328e-07, + "loss": 0.1742, + "step": 12647 + }, + { + "epoch": 0.66, + "learning_rate": 1.0938119708290288e-07, + "loss": 0.15, + "step": 12648 + }, + { + "epoch": 0.66, + "learning_rate": 1.0935105671315285e-07, + "loss": 0.1246, + "step": 12649 + }, + { + "epoch": 0.66, + "learning_rate": 1.0932091893418456e-07, + "loss": 0.1517, + "step": 12650 + }, + { + "epoch": 0.66, + "learning_rate": 1.0929078374685954e-07, + "loss": 0.1341, + "step": 12651 + }, + { + "epoch": 0.66, + "learning_rate": 1.0926065115203886e-07, + "loss": 0.1379, + "step": 12652 + }, + { + "epoch": 0.66, + "learning_rate": 1.0923052115058373e-07, + "loss": 0.1354, + "step": 12653 + }, + { + "epoch": 0.66, + "learning_rate": 1.0920039374335541e-07, + "loss": 0.1037, + "step": 12654 + }, + { + "epoch": 0.66, + "learning_rate": 1.091702689312149e-07, + "loss": 0.1403, + "step": 12655 + }, + { + "epoch": 0.66, + "learning_rate": 1.0914014671502316e-07, + "loss": 0.1366, + "step": 12656 + }, + { + "epoch": 0.66, + "learning_rate": 1.0911002709564105e-07, + "loss": 0.107, + "step": 12657 + }, + { + "epoch": 0.66, + "learning_rate": 1.0907991007392951e-07, + "loss": 0.1175, + "step": 12658 + }, + { + "epoch": 0.66, + "learning_rate": 1.0904979565074923e-07, + "loss": 0.1557, + "step": 12659 + }, + { + "epoch": 0.66, + "learning_rate": 1.0901968382696093e-07, + "loss": 0.1108, + "step": 12660 + }, + { + "epoch": 0.66, + "learning_rate": 1.0898957460342517e-07, + "loss": 0.1409, + "step": 12661 + }, + { + "epoch": 0.66, + "learning_rate": 1.089594679810025e-07, + "loss": 0.0893, + "step": 12662 + }, + { + "epoch": 0.66, + "learning_rate": 1.0892936396055339e-07, + "loss": 0.1505, + "step": 12663 + }, + { + "epoch": 0.66, + "learning_rate": 1.0889926254293813e-07, + "loss": 0.132, + "step": 12664 + }, + { + "epoch": 0.66, + "learning_rate": 1.0886916372901717e-07, + "loss": 0.1411, + "step": 12665 + }, + { + "epoch": 0.66, + "learning_rate": 1.0883906751965073e-07, + "loss": 0.1683, + "step": 12666 + }, + { + "epoch": 0.66, + "learning_rate": 1.088089739156989e-07, + "loss": 0.1391, + "step": 12667 + }, + { + "epoch": 0.66, + "learning_rate": 1.0877888291802175e-07, + "loss": 0.1385, + "step": 12668 + }, + { + "epoch": 0.66, + "learning_rate": 1.0874879452747947e-07, + "loss": 0.1266, + "step": 12669 + }, + { + "epoch": 0.66, + "learning_rate": 1.087187087449318e-07, + "loss": 0.1165, + "step": 12670 + }, + { + "epoch": 0.66, + "learning_rate": 1.0868862557123861e-07, + "loss": 0.1654, + "step": 12671 + }, + { + "epoch": 0.66, + "learning_rate": 1.0865854500725981e-07, + "loss": 0.1356, + "step": 12672 + }, + { + "epoch": 0.66, + "learning_rate": 1.0862846705385507e-07, + "loss": 0.1123, + "step": 12673 + }, + { + "epoch": 0.66, + "learning_rate": 1.08598391711884e-07, + "loss": 0.1059, + "step": 12674 + }, + { + "epoch": 0.66, + "learning_rate": 1.085683189822061e-07, + "loss": 0.1075, + "step": 12675 + }, + { + "epoch": 0.66, + "learning_rate": 1.0853824886568104e-07, + "loss": 0.1018, + "step": 12676 + }, + { + "epoch": 0.66, + "learning_rate": 1.085081813631682e-07, + "loss": 0.1344, + "step": 12677 + }, + { + "epoch": 0.66, + "learning_rate": 1.0847811647552669e-07, + "loss": 0.1263, + "step": 12678 + }, + { + "epoch": 0.66, + "learning_rate": 1.0844805420361604e-07, + "loss": 0.1347, + "step": 12679 + }, + { + "epoch": 0.66, + "learning_rate": 1.0841799454829532e-07, + "loss": 0.146, + "step": 12680 + }, + { + "epoch": 0.66, + "learning_rate": 1.083879375104237e-07, + "loss": 0.1413, + "step": 12681 + }, + { + "epoch": 0.66, + "learning_rate": 1.0835788309086007e-07, + "loss": 0.1421, + "step": 12682 + }, + { + "epoch": 0.66, + "learning_rate": 1.0832783129046364e-07, + "loss": 0.1388, + "step": 12683 + }, + { + "epoch": 0.66, + "learning_rate": 1.0829778211009317e-07, + "loss": 0.1381, + "step": 12684 + }, + { + "epoch": 0.66, + "learning_rate": 1.0826773555060746e-07, + "loss": 0.1619, + "step": 12685 + }, + { + "epoch": 0.66, + "learning_rate": 1.0823769161286531e-07, + "loss": 0.1284, + "step": 12686 + }, + { + "epoch": 0.66, + "learning_rate": 1.0820765029772534e-07, + "loss": 0.1651, + "step": 12687 + }, + { + "epoch": 0.66, + "learning_rate": 1.0817761160604617e-07, + "loss": 0.1102, + "step": 12688 + }, + { + "epoch": 0.66, + "learning_rate": 1.0814757553868624e-07, + "loss": 0.1904, + "step": 12689 + }, + { + "epoch": 0.66, + "learning_rate": 1.0811754209650413e-07, + "loss": 0.0986, + "step": 12690 + }, + { + "epoch": 0.66, + "learning_rate": 1.0808751128035813e-07, + "loss": 0.1464, + "step": 12691 + }, + { + "epoch": 0.66, + "learning_rate": 1.0805748309110654e-07, + "loss": 0.0864, + "step": 12692 + }, + { + "epoch": 0.66, + "learning_rate": 1.0802745752960749e-07, + "loss": 0.1571, + "step": 12693 + }, + { + "epoch": 0.66, + "learning_rate": 1.079974345967194e-07, + "loss": 0.107, + "step": 12694 + }, + { + "epoch": 0.66, + "learning_rate": 1.0796741429330003e-07, + "loss": 0.1047, + "step": 12695 + }, + { + "epoch": 0.66, + "learning_rate": 1.0793739662020743e-07, + "loss": 0.1217, + "step": 12696 + }, + { + "epoch": 0.66, + "learning_rate": 1.0790738157829964e-07, + "loss": 0.1241, + "step": 12697 + }, + { + "epoch": 0.66, + "learning_rate": 1.0787736916843444e-07, + "loss": 0.1239, + "step": 12698 + }, + { + "epoch": 0.66, + "learning_rate": 1.0784735939146958e-07, + "loss": 0.1422, + "step": 12699 + }, + { + "epoch": 0.66, + "learning_rate": 1.078173522482627e-07, + "loss": 0.1744, + "step": 12700 + }, + { + "epoch": 0.66, + "learning_rate": 1.0778734773967157e-07, + "loss": 0.1519, + "step": 12701 + }, + { + "epoch": 0.66, + "learning_rate": 1.0775734586655363e-07, + "loss": 0.1207, + "step": 12702 + }, + { + "epoch": 0.66, + "learning_rate": 1.0772734662976634e-07, + "loss": 0.1654, + "step": 12703 + }, + { + "epoch": 0.66, + "learning_rate": 1.076973500301671e-07, + "loss": 0.118, + "step": 12704 + }, + { + "epoch": 0.66, + "learning_rate": 1.0766735606861326e-07, + "loss": 0.1461, + "step": 12705 + }, + { + "epoch": 0.66, + "learning_rate": 1.07637364745962e-07, + "loss": 0.1155, + "step": 12706 + }, + { + "epoch": 0.66, + "learning_rate": 1.0760737606307044e-07, + "loss": 0.1407, + "step": 12707 + }, + { + "epoch": 0.66, + "learning_rate": 1.0757739002079582e-07, + "loss": 0.1468, + "step": 12708 + }, + { + "epoch": 0.66, + "learning_rate": 1.0754740661999506e-07, + "loss": 0.1459, + "step": 12709 + }, + { + "epoch": 0.66, + "learning_rate": 1.0751742586152514e-07, + "loss": 0.2086, + "step": 12710 + }, + { + "epoch": 0.66, + "learning_rate": 1.0748744774624285e-07, + "loss": 0.1059, + "step": 12711 + }, + { + "epoch": 0.66, + "learning_rate": 1.0745747227500506e-07, + "loss": 0.1302, + "step": 12712 + }, + { + "epoch": 0.66, + "learning_rate": 1.074274994486684e-07, + "loss": 0.1143, + "step": 12713 + }, + { + "epoch": 0.66, + "learning_rate": 1.0739752926808949e-07, + "loss": 0.1103, + "step": 12714 + }, + { + "epoch": 0.66, + "learning_rate": 1.07367561734125e-07, + "loss": 0.1316, + "step": 12715 + }, + { + "epoch": 0.66, + "learning_rate": 1.0733759684763137e-07, + "loss": 0.1533, + "step": 12716 + }, + { + "epoch": 0.66, + "learning_rate": 1.0730763460946499e-07, + "loss": 0.1663, + "step": 12717 + }, + { + "epoch": 0.66, + "learning_rate": 1.072776750204821e-07, + "loss": 0.1329, + "step": 12718 + }, + { + "epoch": 0.66, + "learning_rate": 1.0724771808153917e-07, + "loss": 0.1338, + "step": 12719 + }, + { + "epoch": 0.66, + "learning_rate": 1.072177637934923e-07, + "loss": 0.1307, + "step": 12720 + }, + { + "epoch": 0.66, + "learning_rate": 1.0718781215719742e-07, + "loss": 0.1199, + "step": 12721 + }, + { + "epoch": 0.66, + "learning_rate": 1.0715786317351079e-07, + "loss": 0.1055, + "step": 12722 + }, + { + "epoch": 0.66, + "learning_rate": 1.0712791684328823e-07, + "loss": 0.1237, + "step": 12723 + }, + { + "epoch": 0.66, + "learning_rate": 1.0709797316738569e-07, + "loss": 0.1419, + "step": 12724 + }, + { + "epoch": 0.66, + "learning_rate": 1.0706803214665887e-07, + "loss": 0.1295, + "step": 12725 + }, + { + "epoch": 0.66, + "learning_rate": 1.0703809378196363e-07, + "loss": 0.1264, + "step": 12726 + }, + { + "epoch": 0.66, + "learning_rate": 1.0700815807415557e-07, + "loss": 0.1287, + "step": 12727 + }, + { + "epoch": 0.66, + "learning_rate": 1.0697822502409021e-07, + "loss": 0.1593, + "step": 12728 + }, + { + "epoch": 0.66, + "learning_rate": 1.0694829463262313e-07, + "loss": 0.124, + "step": 12729 + }, + { + "epoch": 0.66, + "learning_rate": 1.0691836690060971e-07, + "loss": 0.1467, + "step": 12730 + }, + { + "epoch": 0.66, + "learning_rate": 1.0688844182890529e-07, + "loss": 0.1136, + "step": 12731 + }, + { + "epoch": 0.66, + "learning_rate": 1.0685851941836505e-07, + "loss": 0.1352, + "step": 12732 + }, + { + "epoch": 0.66, + "learning_rate": 1.0682859966984439e-07, + "loss": 0.1408, + "step": 12733 + }, + { + "epoch": 0.66, + "learning_rate": 1.067986825841983e-07, + "loss": 0.1334, + "step": 12734 + }, + { + "epoch": 0.66, + "learning_rate": 1.0676876816228184e-07, + "loss": 0.1021, + "step": 12735 + }, + { + "epoch": 0.66, + "learning_rate": 1.0673885640494993e-07, + "loss": 0.1642, + "step": 12736 + }, + { + "epoch": 0.66, + "learning_rate": 1.0670894731305762e-07, + "loss": 0.1085, + "step": 12737 + }, + { + "epoch": 0.66, + "learning_rate": 1.0667904088745956e-07, + "loss": 0.1498, + "step": 12738 + }, + { + "epoch": 0.66, + "learning_rate": 1.0664913712901051e-07, + "loss": 0.1526, + "step": 12739 + }, + { + "epoch": 0.66, + "learning_rate": 1.0661923603856511e-07, + "loss": 0.1259, + "step": 12740 + }, + { + "epoch": 0.67, + "learning_rate": 1.0658933761697806e-07, + "loss": 0.1341, + "step": 12741 + }, + { + "epoch": 0.67, + "learning_rate": 1.0655944186510377e-07, + "loss": 0.1428, + "step": 12742 + }, + { + "epoch": 0.67, + "learning_rate": 1.0652954878379667e-07, + "loss": 0.1392, + "step": 12743 + }, + { + "epoch": 0.67, + "learning_rate": 1.0649965837391119e-07, + "loss": 0.1622, + "step": 12744 + }, + { + "epoch": 0.67, + "learning_rate": 1.064697706363016e-07, + "loss": 0.1496, + "step": 12745 + }, + { + "epoch": 0.67, + "learning_rate": 1.0643988557182213e-07, + "loss": 0.11, + "step": 12746 + }, + { + "epoch": 0.67, + "learning_rate": 1.0641000318132671e-07, + "loss": 0.1411, + "step": 12747 + }, + { + "epoch": 0.67, + "learning_rate": 1.063801234656696e-07, + "loss": 0.1177, + "step": 12748 + }, + { + "epoch": 0.67, + "learning_rate": 1.0635024642570469e-07, + "loss": 0.1653, + "step": 12749 + }, + { + "epoch": 0.67, + "learning_rate": 1.0632037206228591e-07, + "loss": 0.1623, + "step": 12750 + }, + { + "epoch": 0.67, + "learning_rate": 1.0629050037626695e-07, + "loss": 0.1082, + "step": 12751 + }, + { + "epoch": 0.67, + "learning_rate": 1.0626063136850176e-07, + "loss": 0.1404, + "step": 12752 + }, + { + "epoch": 0.67, + "learning_rate": 1.0623076503984391e-07, + "loss": 0.124, + "step": 12753 + }, + { + "epoch": 0.67, + "learning_rate": 1.0620090139114701e-07, + "loss": 0.1518, + "step": 12754 + }, + { + "epoch": 0.67, + "learning_rate": 1.0617104042326457e-07, + "loss": 0.1333, + "step": 12755 + }, + { + "epoch": 0.67, + "learning_rate": 1.0614118213704997e-07, + "loss": 0.1229, + "step": 12756 + }, + { + "epoch": 0.67, + "learning_rate": 1.0611132653335664e-07, + "loss": 0.1641, + "step": 12757 + }, + { + "epoch": 0.67, + "learning_rate": 1.0608147361303777e-07, + "loss": 0.1465, + "step": 12758 + }, + { + "epoch": 0.67, + "learning_rate": 1.060516233769467e-07, + "loss": 0.117, + "step": 12759 + }, + { + "epoch": 0.67, + "learning_rate": 1.0602177582593652e-07, + "loss": 0.1481, + "step": 12760 + }, + { + "epoch": 0.67, + "learning_rate": 1.0599193096086028e-07, + "loss": 0.135, + "step": 12761 + }, + { + "epoch": 0.67, + "learning_rate": 1.0596208878257085e-07, + "loss": 0.1404, + "step": 12762 + }, + { + "epoch": 0.67, + "learning_rate": 1.0593224929192137e-07, + "loss": 0.1262, + "step": 12763 + }, + { + "epoch": 0.67, + "learning_rate": 1.0590241248976446e-07, + "loss": 0.127, + "step": 12764 + }, + { + "epoch": 0.67, + "learning_rate": 1.0587257837695282e-07, + "loss": 0.1371, + "step": 12765 + }, + { + "epoch": 0.67, + "learning_rate": 1.0584274695433935e-07, + "loss": 0.1349, + "step": 12766 + }, + { + "epoch": 0.67, + "learning_rate": 1.058129182227765e-07, + "loss": 0.1195, + "step": 12767 + }, + { + "epoch": 0.67, + "learning_rate": 1.0578309218311679e-07, + "loss": 0.1248, + "step": 12768 + }, + { + "epoch": 0.67, + "learning_rate": 1.057532688362126e-07, + "loss": 0.1351, + "step": 12769 + }, + { + "epoch": 0.67, + "learning_rate": 1.0572344818291646e-07, + "loss": 0.1324, + "step": 12770 + }, + { + "epoch": 0.67, + "learning_rate": 1.0569363022408065e-07, + "loss": 0.1237, + "step": 12771 + }, + { + "epoch": 0.67, + "learning_rate": 1.0566381496055709e-07, + "loss": 0.0982, + "step": 12772 + }, + { + "epoch": 0.67, + "learning_rate": 1.0563400239319826e-07, + "loss": 0.0977, + "step": 12773 + }, + { + "epoch": 0.67, + "learning_rate": 1.0560419252285604e-07, + "loss": 0.1556, + "step": 12774 + }, + { + "epoch": 0.67, + "learning_rate": 1.0557438535038243e-07, + "loss": 0.1231, + "step": 12775 + }, + { + "epoch": 0.67, + "learning_rate": 1.0554458087662928e-07, + "loss": 0.138, + "step": 12776 + }, + { + "epoch": 0.67, + "learning_rate": 1.0551477910244852e-07, + "loss": 0.1318, + "step": 12777 + }, + { + "epoch": 0.67, + "learning_rate": 1.0548498002869188e-07, + "loss": 0.1543, + "step": 12778 + }, + { + "epoch": 0.67, + "learning_rate": 1.0545518365621097e-07, + "loss": 0.1326, + "step": 12779 + }, + { + "epoch": 0.67, + "learning_rate": 1.0542538998585744e-07, + "loss": 0.0872, + "step": 12780 + }, + { + "epoch": 0.67, + "learning_rate": 1.0539559901848276e-07, + "loss": 0.1292, + "step": 12781 + }, + { + "epoch": 0.67, + "learning_rate": 1.053658107549384e-07, + "loss": 0.15, + "step": 12782 + }, + { + "epoch": 0.67, + "learning_rate": 1.0533602519607559e-07, + "loss": 0.1175, + "step": 12783 + }, + { + "epoch": 0.67, + "learning_rate": 1.0530624234274582e-07, + "loss": 0.1473, + "step": 12784 + }, + { + "epoch": 0.67, + "learning_rate": 1.0527646219580019e-07, + "loss": 0.1202, + "step": 12785 + }, + { + "epoch": 0.67, + "learning_rate": 1.0524668475608985e-07, + "loss": 0.1288, + "step": 12786 + }, + { + "epoch": 0.67, + "learning_rate": 1.0521691002446575e-07, + "loss": 0.154, + "step": 12787 + }, + { + "epoch": 0.67, + "learning_rate": 1.0518713800177903e-07, + "loss": 0.1581, + "step": 12788 + }, + { + "epoch": 0.67, + "learning_rate": 1.0515736868888058e-07, + "loss": 0.1329, + "step": 12789 + }, + { + "epoch": 0.67, + "learning_rate": 1.0512760208662102e-07, + "loss": 0.1504, + "step": 12790 + }, + { + "epoch": 0.67, + "learning_rate": 1.0509783819585125e-07, + "loss": 0.1325, + "step": 12791 + }, + { + "epoch": 0.67, + "learning_rate": 1.0506807701742193e-07, + "loss": 0.1289, + "step": 12792 + }, + { + "epoch": 0.67, + "learning_rate": 1.0503831855218362e-07, + "loss": 0.1572, + "step": 12793 + }, + { + "epoch": 0.67, + "learning_rate": 1.0500856280098674e-07, + "loss": 0.1742, + "step": 12794 + }, + { + "epoch": 0.67, + "learning_rate": 1.0497880976468185e-07, + "loss": 0.1408, + "step": 12795 + }, + { + "epoch": 0.67, + "learning_rate": 1.0494905944411929e-07, + "loss": 0.1421, + "step": 12796 + }, + { + "epoch": 0.67, + "learning_rate": 1.049193118401493e-07, + "loss": 0.2002, + "step": 12797 + }, + { + "epoch": 0.67, + "learning_rate": 1.048895669536221e-07, + "loss": 0.1104, + "step": 12798 + }, + { + "epoch": 0.67, + "learning_rate": 1.0485982478538775e-07, + "loss": 0.1032, + "step": 12799 + }, + { + "epoch": 0.67, + "learning_rate": 1.0483008533629638e-07, + "loss": 0.1665, + "step": 12800 + }, + { + "epoch": 0.67, + "learning_rate": 1.0480034860719782e-07, + "loss": 0.1386, + "step": 12801 + }, + { + "epoch": 0.67, + "learning_rate": 1.0477061459894211e-07, + "loss": 0.1221, + "step": 12802 + }, + { + "epoch": 0.67, + "learning_rate": 1.0474088331237902e-07, + "loss": 0.1196, + "step": 12803 + }, + { + "epoch": 0.67, + "learning_rate": 1.0471115474835823e-07, + "loss": 0.1117, + "step": 12804 + }, + { + "epoch": 0.67, + "learning_rate": 1.0468142890772936e-07, + "loss": 0.1678, + "step": 12805 + }, + { + "epoch": 0.67, + "learning_rate": 1.0465170579134222e-07, + "loss": 0.1533, + "step": 12806 + }, + { + "epoch": 0.67, + "learning_rate": 1.0462198540004603e-07, + "loss": 0.0889, + "step": 12807 + }, + { + "epoch": 0.67, + "learning_rate": 1.0459226773469026e-07, + "loss": 0.1051, + "step": 12808 + }, + { + "epoch": 0.67, + "learning_rate": 1.0456255279612439e-07, + "loss": 0.1322, + "step": 12809 + }, + { + "epoch": 0.67, + "learning_rate": 1.0453284058519758e-07, + "loss": 0.1144, + "step": 12810 + }, + { + "epoch": 0.67, + "learning_rate": 1.0450313110275905e-07, + "loss": 0.1033, + "step": 12811 + }, + { + "epoch": 0.67, + "learning_rate": 1.0447342434965782e-07, + "loss": 0.1268, + "step": 12812 + }, + { + "epoch": 0.67, + "learning_rate": 1.0444372032674309e-07, + "loss": 0.0842, + "step": 12813 + }, + { + "epoch": 0.67, + "learning_rate": 1.0441401903486377e-07, + "loss": 0.1127, + "step": 12814 + }, + { + "epoch": 0.67, + "learning_rate": 1.0438432047486859e-07, + "loss": 0.146, + "step": 12815 + }, + { + "epoch": 0.67, + "learning_rate": 1.0435462464760645e-07, + "loss": 0.1727, + "step": 12816 + }, + { + "epoch": 0.67, + "learning_rate": 1.043249315539261e-07, + "loss": 0.1502, + "step": 12817 + }, + { + "epoch": 0.67, + "learning_rate": 1.0429524119467612e-07, + "loss": 0.0999, + "step": 12818 + }, + { + "epoch": 0.67, + "learning_rate": 1.0426555357070504e-07, + "loss": 0.1346, + "step": 12819 + }, + { + "epoch": 0.67, + "learning_rate": 1.0423586868286145e-07, + "loss": 0.1301, + "step": 12820 + }, + { + "epoch": 0.67, + "learning_rate": 1.0420618653199373e-07, + "loss": 0.1179, + "step": 12821 + }, + { + "epoch": 0.67, + "learning_rate": 1.0417650711895017e-07, + "loss": 0.0972, + "step": 12822 + }, + { + "epoch": 0.67, + "learning_rate": 1.04146830444579e-07, + "loss": 0.1659, + "step": 12823 + }, + { + "epoch": 0.67, + "learning_rate": 1.0411715650972843e-07, + "loss": 0.1276, + "step": 12824 + }, + { + "epoch": 0.67, + "learning_rate": 1.0408748531524656e-07, + "loss": 0.1285, + "step": 12825 + }, + { + "epoch": 0.67, + "learning_rate": 1.0405781686198129e-07, + "loss": 0.132, + "step": 12826 + }, + { + "epoch": 0.67, + "learning_rate": 1.0402815115078075e-07, + "loss": 0.1461, + "step": 12827 + }, + { + "epoch": 0.67, + "learning_rate": 1.0399848818249268e-07, + "loss": 0.1509, + "step": 12828 + }, + { + "epoch": 0.67, + "learning_rate": 1.0396882795796487e-07, + "loss": 0.1419, + "step": 12829 + }, + { + "epoch": 0.67, + "learning_rate": 1.0393917047804497e-07, + "loss": 0.1579, + "step": 12830 + }, + { + "epoch": 0.67, + "learning_rate": 1.039095157435808e-07, + "loss": 0.126, + "step": 12831 + }, + { + "epoch": 0.67, + "learning_rate": 1.0387986375541971e-07, + "loss": 0.1668, + "step": 12832 + }, + { + "epoch": 0.67, + "learning_rate": 1.0385021451440912e-07, + "loss": 0.1495, + "step": 12833 + }, + { + "epoch": 0.67, + "learning_rate": 1.038205680213966e-07, + "loss": 0.1121, + "step": 12834 + }, + { + "epoch": 0.67, + "learning_rate": 1.0379092427722938e-07, + "loss": 0.1167, + "step": 12835 + }, + { + "epoch": 0.67, + "learning_rate": 1.037612832827547e-07, + "loss": 0.1154, + "step": 12836 + }, + { + "epoch": 0.67, + "learning_rate": 1.0373164503881959e-07, + "loss": 0.1224, + "step": 12837 + }, + { + "epoch": 0.67, + "learning_rate": 1.0370200954627132e-07, + "loss": 0.1249, + "step": 12838 + }, + { + "epoch": 0.67, + "learning_rate": 1.0367237680595679e-07, + "loss": 0.1186, + "step": 12839 + }, + { + "epoch": 0.67, + "learning_rate": 1.0364274681872291e-07, + "loss": 0.1717, + "step": 12840 + }, + { + "epoch": 0.67, + "learning_rate": 1.0361311958541654e-07, + "loss": 0.1588, + "step": 12841 + }, + { + "epoch": 0.67, + "learning_rate": 1.035834951068844e-07, + "loss": 0.1262, + "step": 12842 + }, + { + "epoch": 0.67, + "learning_rate": 1.0355387338397321e-07, + "loss": 0.1608, + "step": 12843 + }, + { + "epoch": 0.67, + "learning_rate": 1.0352425441752948e-07, + "loss": 0.1399, + "step": 12844 + }, + { + "epoch": 0.67, + "learning_rate": 1.0349463820839986e-07, + "loss": 0.1592, + "step": 12845 + }, + { + "epoch": 0.67, + "learning_rate": 1.0346502475743073e-07, + "loss": 0.0957, + "step": 12846 + }, + { + "epoch": 0.67, + "learning_rate": 1.0343541406546848e-07, + "loss": 0.143, + "step": 12847 + }, + { + "epoch": 0.67, + "learning_rate": 1.0340580613335928e-07, + "loss": 0.1454, + "step": 12848 + }, + { + "epoch": 0.67, + "learning_rate": 1.0337620096194959e-07, + "loss": 0.1242, + "step": 12849 + }, + { + "epoch": 0.67, + "learning_rate": 1.0334659855208529e-07, + "loss": 0.1316, + "step": 12850 + }, + { + "epoch": 0.67, + "learning_rate": 1.0331699890461254e-07, + "loss": 0.146, + "step": 12851 + }, + { + "epoch": 0.67, + "learning_rate": 1.0328740202037718e-07, + "loss": 0.1226, + "step": 12852 + }, + { + "epoch": 0.67, + "learning_rate": 1.0325780790022529e-07, + "loss": 0.1407, + "step": 12853 + }, + { + "epoch": 0.67, + "learning_rate": 1.032282165450026e-07, + "loss": 0.1023, + "step": 12854 + }, + { + "epoch": 0.67, + "learning_rate": 1.0319862795555476e-07, + "loss": 0.1215, + "step": 12855 + }, + { + "epoch": 0.67, + "learning_rate": 1.031690421327276e-07, + "loss": 0.1297, + "step": 12856 + }, + { + "epoch": 0.67, + "learning_rate": 1.0313945907736666e-07, + "loss": 0.1245, + "step": 12857 + }, + { + "epoch": 0.67, + "learning_rate": 1.0310987879031727e-07, + "loss": 0.1397, + "step": 12858 + }, + { + "epoch": 0.67, + "learning_rate": 1.0308030127242488e-07, + "loss": 0.1175, + "step": 12859 + }, + { + "epoch": 0.67, + "learning_rate": 1.0305072652453499e-07, + "loss": 0.1636, + "step": 12860 + }, + { + "epoch": 0.67, + "learning_rate": 1.0302115454749274e-07, + "loss": 0.1277, + "step": 12861 + }, + { + "epoch": 0.67, + "learning_rate": 1.0299158534214336e-07, + "loss": 0.1454, + "step": 12862 + }, + { + "epoch": 0.67, + "learning_rate": 1.0296201890933183e-07, + "loss": 0.1789, + "step": 12863 + }, + { + "epoch": 0.67, + "learning_rate": 1.0293245524990333e-07, + "loss": 0.1456, + "step": 12864 + }, + { + "epoch": 0.67, + "learning_rate": 1.0290289436470275e-07, + "loss": 0.1258, + "step": 12865 + }, + { + "epoch": 0.67, + "learning_rate": 1.0287333625457488e-07, + "loss": 0.1352, + "step": 12866 + }, + { + "epoch": 0.67, + "learning_rate": 1.0284378092036461e-07, + "loss": 0.1052, + "step": 12867 + }, + { + "epoch": 0.67, + "learning_rate": 1.0281422836291658e-07, + "loss": 0.1163, + "step": 12868 + }, + { + "epoch": 0.67, + "learning_rate": 1.027846785830754e-07, + "loss": 0.1407, + "step": 12869 + }, + { + "epoch": 0.67, + "learning_rate": 1.0275513158168556e-07, + "loss": 0.1689, + "step": 12870 + }, + { + "epoch": 0.67, + "learning_rate": 1.0272558735959166e-07, + "loss": 0.1126, + "step": 12871 + }, + { + "epoch": 0.67, + "learning_rate": 1.0269604591763808e-07, + "loss": 0.1366, + "step": 12872 + }, + { + "epoch": 0.67, + "learning_rate": 1.0266650725666904e-07, + "loss": 0.1576, + "step": 12873 + }, + { + "epoch": 0.67, + "learning_rate": 1.026369713775288e-07, + "loss": 0.1242, + "step": 12874 + }, + { + "epoch": 0.67, + "learning_rate": 1.0260743828106151e-07, + "loss": 0.1202, + "step": 12875 + }, + { + "epoch": 0.67, + "learning_rate": 1.0257790796811121e-07, + "loss": 0.1037, + "step": 12876 + }, + { + "epoch": 0.67, + "learning_rate": 1.0254838043952182e-07, + "loss": 0.1535, + "step": 12877 + }, + { + "epoch": 0.67, + "learning_rate": 1.0251885569613745e-07, + "loss": 0.1344, + "step": 12878 + }, + { + "epoch": 0.67, + "learning_rate": 1.0248933373880179e-07, + "loss": 0.1481, + "step": 12879 + }, + { + "epoch": 0.67, + "learning_rate": 1.0245981456835866e-07, + "loss": 0.1325, + "step": 12880 + }, + { + "epoch": 0.67, + "learning_rate": 1.0243029818565156e-07, + "loss": 0.114, + "step": 12881 + }, + { + "epoch": 0.67, + "learning_rate": 1.0240078459152431e-07, + "loss": 0.1408, + "step": 12882 + }, + { + "epoch": 0.67, + "learning_rate": 1.0237127378682036e-07, + "loss": 0.1294, + "step": 12883 + }, + { + "epoch": 0.67, + "learning_rate": 1.0234176577238297e-07, + "loss": 0.1512, + "step": 12884 + }, + { + "epoch": 0.67, + "learning_rate": 1.0231226054905566e-07, + "loss": 0.1596, + "step": 12885 + }, + { + "epoch": 0.67, + "learning_rate": 1.0228275811768167e-07, + "loss": 0.1691, + "step": 12886 + }, + { + "epoch": 0.67, + "learning_rate": 1.0225325847910418e-07, + "loss": 0.1266, + "step": 12887 + }, + { + "epoch": 0.67, + "learning_rate": 1.0222376163416622e-07, + "loss": 0.1357, + "step": 12888 + }, + { + "epoch": 0.67, + "learning_rate": 1.02194267583711e-07, + "loss": 0.1041, + "step": 12889 + }, + { + "epoch": 0.67, + "learning_rate": 1.0216477632858134e-07, + "loss": 0.1182, + "step": 12890 + }, + { + "epoch": 0.67, + "learning_rate": 1.0213528786962013e-07, + "loss": 0.1343, + "step": 12891 + }, + { + "epoch": 0.67, + "learning_rate": 1.0210580220767018e-07, + "loss": 0.129, + "step": 12892 + }, + { + "epoch": 0.67, + "learning_rate": 1.0207631934357419e-07, + "loss": 0.1611, + "step": 12893 + }, + { + "epoch": 0.67, + "learning_rate": 1.0204683927817482e-07, + "loss": 0.1369, + "step": 12894 + }, + { + "epoch": 0.67, + "learning_rate": 1.020173620123145e-07, + "loss": 0.134, + "step": 12895 + }, + { + "epoch": 0.67, + "learning_rate": 1.0198788754683587e-07, + "loss": 0.1559, + "step": 12896 + }, + { + "epoch": 0.67, + "learning_rate": 1.0195841588258125e-07, + "loss": 0.131, + "step": 12897 + }, + { + "epoch": 0.67, + "learning_rate": 1.0192894702039296e-07, + "loss": 0.1676, + "step": 12898 + }, + { + "epoch": 0.67, + "learning_rate": 1.0189948096111316e-07, + "loss": 0.1603, + "step": 12899 + }, + { + "epoch": 0.67, + "learning_rate": 1.0187001770558421e-07, + "loss": 0.1555, + "step": 12900 + }, + { + "epoch": 0.67, + "learning_rate": 1.0184055725464795e-07, + "loss": 0.1408, + "step": 12901 + }, + { + "epoch": 0.67, + "learning_rate": 1.018110996091464e-07, + "loss": 0.1287, + "step": 12902 + }, + { + "epoch": 0.67, + "learning_rate": 1.0178164476992157e-07, + "loss": 0.1111, + "step": 12903 + }, + { + "epoch": 0.67, + "learning_rate": 1.0175219273781528e-07, + "loss": 0.1206, + "step": 12904 + }, + { + "epoch": 0.67, + "learning_rate": 1.0172274351366925e-07, + "loss": 0.124, + "step": 12905 + }, + { + "epoch": 0.67, + "learning_rate": 1.0169329709832504e-07, + "loss": 0.1291, + "step": 12906 + }, + { + "epoch": 0.67, + "learning_rate": 1.0166385349262445e-07, + "loss": 0.1091, + "step": 12907 + }, + { + "epoch": 0.67, + "learning_rate": 1.0163441269740898e-07, + "loss": 0.1764, + "step": 12908 + }, + { + "epoch": 0.67, + "learning_rate": 1.0160497471351979e-07, + "loss": 0.1203, + "step": 12909 + }, + { + "epoch": 0.67, + "learning_rate": 1.0157553954179849e-07, + "loss": 0.1756, + "step": 12910 + }, + { + "epoch": 0.67, + "learning_rate": 1.0154610718308627e-07, + "loss": 0.1514, + "step": 12911 + }, + { + "epoch": 0.67, + "learning_rate": 1.0151667763822432e-07, + "loss": 0.1215, + "step": 12912 + }, + { + "epoch": 0.67, + "learning_rate": 1.0148725090805363e-07, + "loss": 0.1434, + "step": 12913 + }, + { + "epoch": 0.67, + "learning_rate": 1.0145782699341545e-07, + "loss": 0.1576, + "step": 12914 + }, + { + "epoch": 0.67, + "learning_rate": 1.014284058951506e-07, + "loss": 0.1121, + "step": 12915 + }, + { + "epoch": 0.67, + "learning_rate": 1.0139898761409997e-07, + "loss": 0.1229, + "step": 12916 + }, + { + "epoch": 0.67, + "learning_rate": 1.0136957215110435e-07, + "loss": 0.1276, + "step": 12917 + }, + { + "epoch": 0.67, + "learning_rate": 1.0134015950700441e-07, + "loss": 0.1508, + "step": 12918 + }, + { + "epoch": 0.67, + "learning_rate": 1.0131074968264082e-07, + "loss": 0.1556, + "step": 12919 + }, + { + "epoch": 0.67, + "learning_rate": 1.01281342678854e-07, + "loss": 0.1194, + "step": 12920 + }, + { + "epoch": 0.67, + "learning_rate": 1.0125193849648459e-07, + "loss": 0.1465, + "step": 12921 + }, + { + "epoch": 0.67, + "learning_rate": 1.0122253713637292e-07, + "loss": 0.1396, + "step": 12922 + }, + { + "epoch": 0.67, + "learning_rate": 1.0119313859935928e-07, + "loss": 0.133, + "step": 12923 + }, + { + "epoch": 0.67, + "learning_rate": 1.0116374288628377e-07, + "loss": 0.1247, + "step": 12924 + }, + { + "epoch": 0.67, + "learning_rate": 1.0113434999798673e-07, + "loss": 0.1263, + "step": 12925 + }, + { + "epoch": 0.67, + "learning_rate": 1.0110495993530823e-07, + "loss": 0.1318, + "step": 12926 + }, + { + "epoch": 0.67, + "learning_rate": 1.01075572699088e-07, + "loss": 0.1649, + "step": 12927 + }, + { + "epoch": 0.67, + "learning_rate": 1.0104618829016613e-07, + "loss": 0.1451, + "step": 12928 + }, + { + "epoch": 0.67, + "learning_rate": 1.0101680670938244e-07, + "loss": 0.1536, + "step": 12929 + }, + { + "epoch": 0.67, + "learning_rate": 1.009874279575766e-07, + "loss": 0.1078, + "step": 12930 + }, + { + "epoch": 0.67, + "learning_rate": 1.0095805203558822e-07, + "loss": 0.1232, + "step": 12931 + }, + { + "epoch": 0.67, + "learning_rate": 1.0092867894425703e-07, + "loss": 0.1362, + "step": 12932 + }, + { + "epoch": 0.68, + "learning_rate": 1.0089930868442245e-07, + "loss": 0.1137, + "step": 12933 + }, + { + "epoch": 0.68, + "learning_rate": 1.0086994125692387e-07, + "loss": 0.1079, + "step": 12934 + }, + { + "epoch": 0.68, + "learning_rate": 1.0084057666260064e-07, + "loss": 0.1272, + "step": 12935 + }, + { + "epoch": 0.68, + "learning_rate": 1.0081121490229201e-07, + "loss": 0.149, + "step": 12936 + }, + { + "epoch": 0.68, + "learning_rate": 1.0078185597683715e-07, + "loss": 0.138, + "step": 12937 + }, + { + "epoch": 0.68, + "learning_rate": 1.0075249988707506e-07, + "loss": 0.1489, + "step": 12938 + }, + { + "epoch": 0.68, + "learning_rate": 1.0072314663384491e-07, + "loss": 0.133, + "step": 12939 + }, + { + "epoch": 0.68, + "learning_rate": 1.0069379621798557e-07, + "loss": 0.1534, + "step": 12940 + }, + { + "epoch": 0.68, + "learning_rate": 1.0066444864033587e-07, + "loss": 0.1498, + "step": 12941 + }, + { + "epoch": 0.68, + "learning_rate": 1.0063510390173448e-07, + "loss": 0.1883, + "step": 12942 + }, + { + "epoch": 0.68, + "learning_rate": 1.0060576200302038e-07, + "loss": 0.1527, + "step": 12943 + }, + { + "epoch": 0.68, + "learning_rate": 1.0057642294503184e-07, + "loss": 0.1745, + "step": 12944 + }, + { + "epoch": 0.68, + "learning_rate": 1.0054708672860747e-07, + "loss": 0.1149, + "step": 12945 + }, + { + "epoch": 0.68, + "learning_rate": 1.0051775335458581e-07, + "loss": 0.1066, + "step": 12946 + }, + { + "epoch": 0.68, + "learning_rate": 1.0048842282380516e-07, + "loss": 0.1061, + "step": 12947 + }, + { + "epoch": 0.68, + "learning_rate": 1.0045909513710382e-07, + "loss": 0.0977, + "step": 12948 + }, + { + "epoch": 0.68, + "learning_rate": 1.004297702953199e-07, + "loss": 0.1739, + "step": 12949 + }, + { + "epoch": 0.68, + "learning_rate": 1.0040044829929164e-07, + "loss": 0.1539, + "step": 12950 + }, + { + "epoch": 0.68, + "learning_rate": 1.003711291498571e-07, + "loss": 0.1142, + "step": 12951 + }, + { + "epoch": 0.68, + "learning_rate": 1.0034181284785399e-07, + "loss": 0.1224, + "step": 12952 + }, + { + "epoch": 0.68, + "learning_rate": 1.0031249939412041e-07, + "loss": 0.1184, + "step": 12953 + }, + { + "epoch": 0.68, + "learning_rate": 1.0028318878949411e-07, + "loss": 0.0978, + "step": 12954 + }, + { + "epoch": 0.68, + "learning_rate": 1.0025388103481276e-07, + "loss": 0.1082, + "step": 12955 + }, + { + "epoch": 0.68, + "learning_rate": 1.002245761309139e-07, + "loss": 0.112, + "step": 12956 + }, + { + "epoch": 0.68, + "learning_rate": 1.0019527407863526e-07, + "loss": 0.1554, + "step": 12957 + }, + { + "epoch": 0.68, + "learning_rate": 1.0016597487881422e-07, + "loss": 0.14, + "step": 12958 + }, + { + "epoch": 0.68, + "learning_rate": 1.0013667853228813e-07, + "loss": 0.1385, + "step": 12959 + }, + { + "epoch": 0.68, + "learning_rate": 1.0010738503989436e-07, + "loss": 0.1549, + "step": 12960 + }, + { + "epoch": 0.68, + "learning_rate": 1.0007809440247006e-07, + "loss": 0.1413, + "step": 12961 + }, + { + "epoch": 0.68, + "learning_rate": 1.000488066208524e-07, + "loss": 0.1253, + "step": 12962 + }, + { + "epoch": 0.68, + "learning_rate": 1.0001952169587844e-07, + "loss": 0.1498, + "step": 12963 + }, + { + "epoch": 0.68, + "learning_rate": 9.999023962838507e-08, + "loss": 0.168, + "step": 12964 + }, + { + "epoch": 0.68, + "learning_rate": 9.996096041920932e-08, + "loss": 0.1513, + "step": 12965 + }, + { + "epoch": 0.68, + "learning_rate": 9.993168406918794e-08, + "loss": 0.1279, + "step": 12966 + }, + { + "epoch": 0.68, + "learning_rate": 9.990241057915761e-08, + "loss": 0.1699, + "step": 12967 + }, + { + "epoch": 0.68, + "learning_rate": 9.987313994995516e-08, + "loss": 0.1687, + "step": 12968 + }, + { + "epoch": 0.68, + "learning_rate": 9.984387218241697e-08, + "loss": 0.1189, + "step": 12969 + }, + { + "epoch": 0.68, + "learning_rate": 9.981460727737953e-08, + "loss": 0.1405, + "step": 12970 + }, + { + "epoch": 0.68, + "learning_rate": 9.978534523567925e-08, + "loss": 0.1118, + "step": 12971 + }, + { + "epoch": 0.68, + "learning_rate": 9.975608605815256e-08, + "loss": 0.154, + "step": 12972 + }, + { + "epoch": 0.68, + "learning_rate": 9.972682974563562e-08, + "loss": 0.1277, + "step": 12973 + }, + { + "epoch": 0.68, + "learning_rate": 9.96975762989646e-08, + "loss": 0.1145, + "step": 12974 + }, + { + "epoch": 0.68, + "learning_rate": 9.966832571897549e-08, + "loss": 0.1305, + "step": 12975 + }, + { + "epoch": 0.68, + "learning_rate": 9.963907800650446e-08, + "loss": 0.1742, + "step": 12976 + }, + { + "epoch": 0.68, + "learning_rate": 9.960983316238738e-08, + "loss": 0.1318, + "step": 12977 + }, + { + "epoch": 0.68, + "learning_rate": 9.958059118745987e-08, + "loss": 0.1445, + "step": 12978 + }, + { + "epoch": 0.68, + "learning_rate": 9.955135208255791e-08, + "loss": 0.1241, + "step": 12979 + }, + { + "epoch": 0.68, + "learning_rate": 9.952211584851708e-08, + "loss": 0.1589, + "step": 12980 + }, + { + "epoch": 0.68, + "learning_rate": 9.949288248617298e-08, + "loss": 0.1315, + "step": 12981 + }, + { + "epoch": 0.68, + "learning_rate": 9.946365199636102e-08, + "loss": 0.1458, + "step": 12982 + }, + { + "epoch": 0.68, + "learning_rate": 9.943442437991675e-08, + "loss": 0.1266, + "step": 12983 + }, + { + "epoch": 0.68, + "learning_rate": 9.940519963767548e-08, + "loss": 0.1677, + "step": 12984 + }, + { + "epoch": 0.68, + "learning_rate": 9.937597777047244e-08, + "loss": 0.155, + "step": 12985 + }, + { + "epoch": 0.68, + "learning_rate": 9.934675877914279e-08, + "loss": 0.118, + "step": 12986 + }, + { + "epoch": 0.68, + "learning_rate": 9.931754266452162e-08, + "loss": 0.1199, + "step": 12987 + }, + { + "epoch": 0.68, + "learning_rate": 9.928832942744396e-08, + "loss": 0.1289, + "step": 12988 + }, + { + "epoch": 0.68, + "learning_rate": 9.925911906874465e-08, + "loss": 0.1565, + "step": 12989 + }, + { + "epoch": 0.68, + "learning_rate": 9.922991158925867e-08, + "loss": 0.1117, + "step": 12990 + }, + { + "epoch": 0.68, + "learning_rate": 9.920070698982073e-08, + "loss": 0.1505, + "step": 12991 + }, + { + "epoch": 0.68, + "learning_rate": 9.91715052712655e-08, + "loss": 0.1303, + "step": 12992 + }, + { + "epoch": 0.68, + "learning_rate": 9.914230643442748e-08, + "loss": 0.1273, + "step": 12993 + }, + { + "epoch": 0.68, + "learning_rate": 9.911311048014143e-08, + "loss": 0.1389, + "step": 12994 + }, + { + "epoch": 0.68, + "learning_rate": 9.908391740924155e-08, + "loss": 0.1593, + "step": 12995 + }, + { + "epoch": 0.68, + "learning_rate": 9.905472722256223e-08, + "loss": 0.128, + "step": 12996 + }, + { + "epoch": 0.68, + "learning_rate": 9.90255399209378e-08, + "loss": 0.1215, + "step": 12997 + }, + { + "epoch": 0.68, + "learning_rate": 9.899635550520245e-08, + "loss": 0.157, + "step": 12998 + }, + { + "epoch": 0.68, + "learning_rate": 9.896717397619026e-08, + "loss": 0.1013, + "step": 12999 + }, + { + "epoch": 0.68, + "learning_rate": 9.893799533473514e-08, + "loss": 0.1424, + "step": 13000 + }, + { + "epoch": 0.68, + "learning_rate": 9.89088195816712e-08, + "loss": 0.1289, + "step": 13001 + }, + { + "epoch": 0.68, + "learning_rate": 9.887964671783224e-08, + "loss": 0.1403, + "step": 13002 + }, + { + "epoch": 0.68, + "learning_rate": 9.8850476744052e-08, + "loss": 0.148, + "step": 13003 + }, + { + "epoch": 0.68, + "learning_rate": 9.88213096611642e-08, + "loss": 0.154, + "step": 13004 + }, + { + "epoch": 0.68, + "learning_rate": 9.879214547000241e-08, + "loss": 0.1444, + "step": 13005 + }, + { + "epoch": 0.68, + "learning_rate": 9.876298417140019e-08, + "loss": 0.1284, + "step": 13006 + }, + { + "epoch": 0.68, + "learning_rate": 9.873382576619087e-08, + "loss": 0.1473, + "step": 13007 + }, + { + "epoch": 0.68, + "learning_rate": 9.870467025520799e-08, + "loss": 0.1363, + "step": 13008 + }, + { + "epoch": 0.68, + "learning_rate": 9.867551763928472e-08, + "loss": 0.1182, + "step": 13009 + }, + { + "epoch": 0.68, + "learning_rate": 9.864636791925429e-08, + "loss": 0.1321, + "step": 13010 + }, + { + "epoch": 0.68, + "learning_rate": 9.861722109594978e-08, + "loss": 0.1233, + "step": 13011 + }, + { + "epoch": 0.68, + "learning_rate": 9.858807717020423e-08, + "loss": 0.1444, + "step": 13012 + }, + { + "epoch": 0.68, + "learning_rate": 9.85589361428506e-08, + "loss": 0.1475, + "step": 13013 + }, + { + "epoch": 0.68, + "learning_rate": 9.852979801472165e-08, + "loss": 0.1556, + "step": 13014 + }, + { + "epoch": 0.68, + "learning_rate": 9.850066278665032e-08, + "loss": 0.1237, + "step": 13015 + }, + { + "epoch": 0.68, + "learning_rate": 9.847153045946924e-08, + "loss": 0.1321, + "step": 13016 + }, + { + "epoch": 0.68, + "learning_rate": 9.844240103401098e-08, + "loss": 0.0977, + "step": 13017 + }, + { + "epoch": 0.68, + "learning_rate": 9.841327451110806e-08, + "loss": 0.1702, + "step": 13018 + }, + { + "epoch": 0.68, + "learning_rate": 9.838415089159303e-08, + "loss": 0.1151, + "step": 13019 + }, + { + "epoch": 0.68, + "learning_rate": 9.835503017629827e-08, + "loss": 0.1866, + "step": 13020 + }, + { + "epoch": 0.68, + "learning_rate": 9.832591236605585e-08, + "loss": 0.1227, + "step": 13021 + }, + { + "epoch": 0.68, + "learning_rate": 9.829679746169817e-08, + "loss": 0.1134, + "step": 13022 + }, + { + "epoch": 0.68, + "learning_rate": 9.826768546405728e-08, + "loss": 0.105, + "step": 13023 + }, + { + "epoch": 0.68, + "learning_rate": 9.823857637396523e-08, + "loss": 0.151, + "step": 13024 + }, + { + "epoch": 0.68, + "learning_rate": 9.820947019225385e-08, + "loss": 0.1586, + "step": 13025 + }, + { + "epoch": 0.68, + "learning_rate": 9.818036691975517e-08, + "loss": 0.1449, + "step": 13026 + }, + { + "epoch": 0.68, + "learning_rate": 9.815126655730095e-08, + "loss": 0.1862, + "step": 13027 + }, + { + "epoch": 0.68, + "learning_rate": 9.812216910572282e-08, + "loss": 0.1027, + "step": 13028 + }, + { + "epoch": 0.68, + "learning_rate": 9.809307456585245e-08, + "loss": 0.1352, + "step": 13029 + }, + { + "epoch": 0.68, + "learning_rate": 9.806398293852134e-08, + "loss": 0.1496, + "step": 13030 + }, + { + "epoch": 0.68, + "learning_rate": 9.803489422456093e-08, + "loss": 0.134, + "step": 13031 + }, + { + "epoch": 0.68, + "learning_rate": 9.800580842480255e-08, + "loss": 0.1599, + "step": 13032 + }, + { + "epoch": 0.68, + "learning_rate": 9.79767255400776e-08, + "loss": 0.1389, + "step": 13033 + }, + { + "epoch": 0.68, + "learning_rate": 9.794764557121724e-08, + "loss": 0.1478, + "step": 13034 + }, + { + "epoch": 0.68, + "learning_rate": 9.791856851905255e-08, + "loss": 0.1391, + "step": 13035 + }, + { + "epoch": 0.68, + "learning_rate": 9.788949438441452e-08, + "loss": 0.1683, + "step": 13036 + }, + { + "epoch": 0.68, + "learning_rate": 9.786042316813427e-08, + "loss": 0.1278, + "step": 13037 + }, + { + "epoch": 0.68, + "learning_rate": 9.783135487104253e-08, + "loss": 0.1519, + "step": 13038 + }, + { + "epoch": 0.68, + "learning_rate": 9.780228949397001e-08, + "loss": 0.104, + "step": 13039 + }, + { + "epoch": 0.68, + "learning_rate": 9.777322703774757e-08, + "loss": 0.1451, + "step": 13040 + }, + { + "epoch": 0.68, + "learning_rate": 9.77441675032058e-08, + "loss": 0.1272, + "step": 13041 + }, + { + "epoch": 0.68, + "learning_rate": 9.771511089117517e-08, + "loss": 0.1254, + "step": 13042 + }, + { + "epoch": 0.68, + "learning_rate": 9.768605720248609e-08, + "loss": 0.143, + "step": 13043 + }, + { + "epoch": 0.68, + "learning_rate": 9.765700643796906e-08, + "loss": 0.1519, + "step": 13044 + }, + { + "epoch": 0.68, + "learning_rate": 9.762795859845429e-08, + "loss": 0.1394, + "step": 13045 + }, + { + "epoch": 0.68, + "learning_rate": 9.7598913684772e-08, + "loss": 0.1524, + "step": 13046 + }, + { + "epoch": 0.68, + "learning_rate": 9.756987169775229e-08, + "loss": 0.1128, + "step": 13047 + }, + { + "epoch": 0.68, + "learning_rate": 9.754083263822519e-08, + "loss": 0.1169, + "step": 13048 + }, + { + "epoch": 0.68, + "learning_rate": 9.751179650702062e-08, + "loss": 0.1277, + "step": 13049 + }, + { + "epoch": 0.68, + "learning_rate": 9.74827633049684e-08, + "loss": 0.1377, + "step": 13050 + }, + { + "epoch": 0.68, + "learning_rate": 9.745373303289847e-08, + "loss": 0.1415, + "step": 13051 + }, + { + "epoch": 0.68, + "learning_rate": 9.742470569164045e-08, + "loss": 0.1703, + "step": 13052 + }, + { + "epoch": 0.68, + "learning_rate": 9.739568128202391e-08, + "loss": 0.1373, + "step": 13053 + }, + { + "epoch": 0.68, + "learning_rate": 9.736665980487844e-08, + "loss": 0.1068, + "step": 13054 + }, + { + "epoch": 0.68, + "learning_rate": 9.733764126103343e-08, + "loss": 0.1155, + "step": 13055 + }, + { + "epoch": 0.68, + "learning_rate": 9.730862565131827e-08, + "loss": 0.1671, + "step": 13056 + }, + { + "epoch": 0.68, + "learning_rate": 9.727961297656216e-08, + "loss": 0.181, + "step": 13057 + }, + { + "epoch": 0.68, + "learning_rate": 9.725060323759441e-08, + "loss": 0.1113, + "step": 13058 + }, + { + "epoch": 0.68, + "learning_rate": 9.722159643524412e-08, + "loss": 0.1744, + "step": 13059 + }, + { + "epoch": 0.68, + "learning_rate": 9.719259257034027e-08, + "loss": 0.1431, + "step": 13060 + }, + { + "epoch": 0.68, + "learning_rate": 9.716359164371175e-08, + "loss": 0.1341, + "step": 13061 + }, + { + "epoch": 0.68, + "learning_rate": 9.713459365618755e-08, + "loss": 0.1386, + "step": 13062 + }, + { + "epoch": 0.68, + "learning_rate": 9.710559860859644e-08, + "loss": 0.1372, + "step": 13063 + }, + { + "epoch": 0.68, + "learning_rate": 9.707660650176688e-08, + "loss": 0.1572, + "step": 13064 + }, + { + "epoch": 0.68, + "learning_rate": 9.704761733652774e-08, + "loss": 0.1337, + "step": 13065 + }, + { + "epoch": 0.68, + "learning_rate": 9.701863111370742e-08, + "loss": 0.1177, + "step": 13066 + }, + { + "epoch": 0.68, + "learning_rate": 9.69896478341344e-08, + "loss": 0.1375, + "step": 13067 + }, + { + "epoch": 0.68, + "learning_rate": 9.696066749863691e-08, + "loss": 0.1282, + "step": 13068 + }, + { + "epoch": 0.68, + "learning_rate": 9.69316901080434e-08, + "loss": 0.1494, + "step": 13069 + }, + { + "epoch": 0.68, + "learning_rate": 9.690271566318197e-08, + "loss": 0.1602, + "step": 13070 + }, + { + "epoch": 0.68, + "learning_rate": 9.687374416488084e-08, + "loss": 0.1502, + "step": 13071 + }, + { + "epoch": 0.68, + "learning_rate": 9.684477561396774e-08, + "loss": 0.1609, + "step": 13072 + }, + { + "epoch": 0.68, + "learning_rate": 9.681581001127084e-08, + "loss": 0.1375, + "step": 13073 + }, + { + "epoch": 0.68, + "learning_rate": 9.678684735761791e-08, + "loss": 0.1539, + "step": 13074 + }, + { + "epoch": 0.68, + "learning_rate": 9.675788765383675e-08, + "loss": 0.1471, + "step": 13075 + }, + { + "epoch": 0.68, + "learning_rate": 9.672893090075493e-08, + "loss": 0.1298, + "step": 13076 + }, + { + "epoch": 0.68, + "learning_rate": 9.669997709920022e-08, + "loss": 0.1347, + "step": 13077 + }, + { + "epoch": 0.68, + "learning_rate": 9.667102625000003e-08, + "loss": 0.1342, + "step": 13078 + }, + { + "epoch": 0.68, + "learning_rate": 9.664207835398173e-08, + "loss": 0.1445, + "step": 13079 + }, + { + "epoch": 0.68, + "learning_rate": 9.661313341197288e-08, + "loss": 0.1495, + "step": 13080 + }, + { + "epoch": 0.68, + "learning_rate": 9.658419142480053e-08, + "loss": 0.1334, + "step": 13081 + }, + { + "epoch": 0.68, + "learning_rate": 9.655525239329187e-08, + "loss": 0.1379, + "step": 13082 + }, + { + "epoch": 0.68, + "learning_rate": 9.652631631827399e-08, + "loss": 0.1241, + "step": 13083 + }, + { + "epoch": 0.68, + "learning_rate": 9.6497383200574e-08, + "loss": 0.1021, + "step": 13084 + }, + { + "epoch": 0.68, + "learning_rate": 9.646845304101876e-08, + "loss": 0.1209, + "step": 13085 + }, + { + "epoch": 0.68, + "learning_rate": 9.643952584043508e-08, + "loss": 0.1359, + "step": 13086 + }, + { + "epoch": 0.68, + "learning_rate": 9.641060159964967e-08, + "loss": 0.1326, + "step": 13087 + }, + { + "epoch": 0.68, + "learning_rate": 9.63816803194894e-08, + "loss": 0.1383, + "step": 13088 + }, + { + "epoch": 0.68, + "learning_rate": 9.635276200078065e-08, + "loss": 0.1068, + "step": 13089 + }, + { + "epoch": 0.68, + "learning_rate": 9.632384664434987e-08, + "loss": 0.0984, + "step": 13090 + }, + { + "epoch": 0.68, + "learning_rate": 9.629493425102367e-08, + "loss": 0.1217, + "step": 13091 + }, + { + "epoch": 0.68, + "learning_rate": 9.626602482162826e-08, + "loss": 0.1094, + "step": 13092 + }, + { + "epoch": 0.68, + "learning_rate": 9.623711835698993e-08, + "loss": 0.0966, + "step": 13093 + }, + { + "epoch": 0.68, + "learning_rate": 9.620821485793473e-08, + "loss": 0.1524, + "step": 13094 + }, + { + "epoch": 0.68, + "learning_rate": 9.617931432528891e-08, + "loss": 0.1237, + "step": 13095 + }, + { + "epoch": 0.68, + "learning_rate": 9.615041675987834e-08, + "loss": 0.1463, + "step": 13096 + }, + { + "epoch": 0.68, + "learning_rate": 9.612152216252898e-08, + "loss": 0.1325, + "step": 13097 + }, + { + "epoch": 0.68, + "learning_rate": 9.609263053406658e-08, + "loss": 0.152, + "step": 13098 + }, + { + "epoch": 0.68, + "learning_rate": 9.606374187531694e-08, + "loss": 0.1718, + "step": 13099 + }, + { + "epoch": 0.68, + "learning_rate": 9.603485618710567e-08, + "loss": 0.1341, + "step": 13100 + }, + { + "epoch": 0.68, + "learning_rate": 9.600597347025827e-08, + "loss": 0.1203, + "step": 13101 + }, + { + "epoch": 0.68, + "learning_rate": 9.597709372560036e-08, + "loss": 0.118, + "step": 13102 + }, + { + "epoch": 0.68, + "learning_rate": 9.594821695395727e-08, + "loss": 0.1599, + "step": 13103 + }, + { + "epoch": 0.68, + "learning_rate": 9.591934315615434e-08, + "loss": 0.1267, + "step": 13104 + }, + { + "epoch": 0.68, + "learning_rate": 9.589047233301664e-08, + "loss": 0.113, + "step": 13105 + }, + { + "epoch": 0.68, + "learning_rate": 9.586160448536959e-08, + "loss": 0.1375, + "step": 13106 + }, + { + "epoch": 0.68, + "learning_rate": 9.583273961403803e-08, + "loss": 0.157, + "step": 13107 + }, + { + "epoch": 0.68, + "learning_rate": 9.58038777198469e-08, + "loss": 0.119, + "step": 13108 + }, + { + "epoch": 0.68, + "learning_rate": 9.577501880362125e-08, + "loss": 0.1256, + "step": 13109 + }, + { + "epoch": 0.68, + "learning_rate": 9.57461628661858e-08, + "loss": 0.117, + "step": 13110 + }, + { + "epoch": 0.68, + "learning_rate": 9.571730990836525e-08, + "loss": 0.1217, + "step": 13111 + }, + { + "epoch": 0.68, + "learning_rate": 9.568845993098416e-08, + "loss": 0.109, + "step": 13112 + }, + { + "epoch": 0.68, + "learning_rate": 9.565961293486724e-08, + "loss": 0.1186, + "step": 13113 + }, + { + "epoch": 0.68, + "learning_rate": 9.563076892083895e-08, + "loss": 0.1642, + "step": 13114 + }, + { + "epoch": 0.68, + "learning_rate": 9.560192788972339e-08, + "loss": 0.1317, + "step": 13115 + }, + { + "epoch": 0.68, + "learning_rate": 9.557308984234513e-08, + "loss": 0.1382, + "step": 13116 + }, + { + "epoch": 0.68, + "learning_rate": 9.554425477952828e-08, + "loss": 0.1181, + "step": 13117 + }, + { + "epoch": 0.68, + "learning_rate": 9.551542270209694e-08, + "loss": 0.1451, + "step": 13118 + }, + { + "epoch": 0.68, + "learning_rate": 9.548659361087507e-08, + "loss": 0.1725, + "step": 13119 + }, + { + "epoch": 0.68, + "learning_rate": 9.545776750668679e-08, + "loss": 0.1542, + "step": 13120 + }, + { + "epoch": 0.68, + "learning_rate": 9.542894439035587e-08, + "loss": 0.1432, + "step": 13121 + }, + { + "epoch": 0.68, + "learning_rate": 9.54001242627061e-08, + "loss": 0.1404, + "step": 13122 + }, + { + "epoch": 0.68, + "learning_rate": 9.537130712456113e-08, + "loss": 0.1316, + "step": 13123 + }, + { + "epoch": 0.69, + "learning_rate": 9.534249297674464e-08, + "loss": 0.103, + "step": 13124 + }, + { + "epoch": 0.69, + "learning_rate": 9.531368182008006e-08, + "loss": 0.1463, + "step": 13125 + }, + { + "epoch": 0.69, + "learning_rate": 9.52848736553908e-08, + "loss": 0.1192, + "step": 13126 + }, + { + "epoch": 0.69, + "learning_rate": 9.525606848350037e-08, + "loss": 0.1294, + "step": 13127 + }, + { + "epoch": 0.69, + "learning_rate": 9.52272663052319e-08, + "loss": 0.1231, + "step": 13128 + }, + { + "epoch": 0.69, + "learning_rate": 9.519846712140864e-08, + "loss": 0.1904, + "step": 13129 + }, + { + "epoch": 0.69, + "learning_rate": 9.516967093285355e-08, + "loss": 0.1142, + "step": 13130 + }, + { + "epoch": 0.69, + "learning_rate": 9.51408777403899e-08, + "loss": 0.1692, + "step": 13131 + }, + { + "epoch": 0.69, + "learning_rate": 9.511208754484037e-08, + "loss": 0.1279, + "step": 13132 + }, + { + "epoch": 0.69, + "learning_rate": 9.508330034702778e-08, + "loss": 0.1534, + "step": 13133 + }, + { + "epoch": 0.69, + "learning_rate": 9.505451614777504e-08, + "loss": 0.1446, + "step": 13134 + }, + { + "epoch": 0.69, + "learning_rate": 9.502573494790476e-08, + "loss": 0.1292, + "step": 13135 + }, + { + "epoch": 0.69, + "learning_rate": 9.499695674823948e-08, + "loss": 0.1437, + "step": 13136 + }, + { + "epoch": 0.69, + "learning_rate": 9.496818154960164e-08, + "loss": 0.1265, + "step": 13137 + }, + { + "epoch": 0.69, + "learning_rate": 9.493940935281381e-08, + "loss": 0.1558, + "step": 13138 + }, + { + "epoch": 0.69, + "learning_rate": 9.491064015869821e-08, + "loss": 0.1046, + "step": 13139 + }, + { + "epoch": 0.69, + "learning_rate": 9.488187396807707e-08, + "loss": 0.1088, + "step": 13140 + }, + { + "epoch": 0.69, + "learning_rate": 9.485311078177256e-08, + "loss": 0.118, + "step": 13141 + }, + { + "epoch": 0.69, + "learning_rate": 9.482435060060674e-08, + "loss": 0.135, + "step": 13142 + }, + { + "epoch": 0.69, + "learning_rate": 9.479559342540158e-08, + "loss": 0.2168, + "step": 13143 + }, + { + "epoch": 0.69, + "learning_rate": 9.47668392569789e-08, + "loss": 0.1135, + "step": 13144 + }, + { + "epoch": 0.69, + "learning_rate": 9.473808809616065e-08, + "loss": 0.1423, + "step": 13145 + }, + { + "epoch": 0.69, + "learning_rate": 9.470933994376847e-08, + "loss": 0.1157, + "step": 13146 + }, + { + "epoch": 0.69, + "learning_rate": 9.468059480062401e-08, + "loss": 0.142, + "step": 13147 + }, + { + "epoch": 0.69, + "learning_rate": 9.465185266754881e-08, + "loss": 0.1465, + "step": 13148 + }, + { + "epoch": 0.69, + "learning_rate": 9.462311354536432e-08, + "loss": 0.1385, + "step": 13149 + }, + { + "epoch": 0.69, + "learning_rate": 9.459437743489191e-08, + "loss": 0.156, + "step": 13150 + }, + { + "epoch": 0.69, + "learning_rate": 9.456564433695285e-08, + "loss": 0.1463, + "step": 13151 + }, + { + "epoch": 0.69, + "learning_rate": 9.453691425236842e-08, + "loss": 0.1314, + "step": 13152 + }, + { + "epoch": 0.69, + "learning_rate": 9.45081871819597e-08, + "loss": 0.1138, + "step": 13153 + }, + { + "epoch": 0.69, + "learning_rate": 9.447946312654771e-08, + "loss": 0.1529, + "step": 13154 + }, + { + "epoch": 0.69, + "learning_rate": 9.445074208695334e-08, + "loss": 0.1406, + "step": 13155 + }, + { + "epoch": 0.69, + "learning_rate": 9.442202406399758e-08, + "loss": 0.1668, + "step": 13156 + }, + { + "epoch": 0.69, + "learning_rate": 9.439330905850118e-08, + "loss": 0.1455, + "step": 13157 + }, + { + "epoch": 0.69, + "learning_rate": 9.436459707128466e-08, + "loss": 0.1833, + "step": 13158 + }, + { + "epoch": 0.69, + "learning_rate": 9.433588810316877e-08, + "loss": 0.1506, + "step": 13159 + }, + { + "epoch": 0.69, + "learning_rate": 9.430718215497402e-08, + "loss": 0.1144, + "step": 13160 + }, + { + "epoch": 0.69, + "learning_rate": 9.427847922752079e-08, + "loss": 0.1439, + "step": 13161 + }, + { + "epoch": 0.69, + "learning_rate": 9.424977932162939e-08, + "loss": 0.1423, + "step": 13162 + }, + { + "epoch": 0.69, + "learning_rate": 9.422108243812014e-08, + "loss": 0.1122, + "step": 13163 + }, + { + "epoch": 0.69, + "learning_rate": 9.419238857781324e-08, + "loss": 0.1139, + "step": 13164 + }, + { + "epoch": 0.69, + "learning_rate": 9.41636977415287e-08, + "loss": 0.1252, + "step": 13165 + }, + { + "epoch": 0.69, + "learning_rate": 9.413500993008653e-08, + "loss": 0.1696, + "step": 13166 + }, + { + "epoch": 0.69, + "learning_rate": 9.410632514430666e-08, + "loss": 0.1183, + "step": 13167 + }, + { + "epoch": 0.69, + "learning_rate": 9.407764338500889e-08, + "loss": 0.1546, + "step": 13168 + }, + { + "epoch": 0.69, + "learning_rate": 9.404896465301289e-08, + "loss": 0.1527, + "step": 13169 + }, + { + "epoch": 0.69, + "learning_rate": 9.402028894913844e-08, + "loss": 0.158, + "step": 13170 + }, + { + "epoch": 0.69, + "learning_rate": 9.399161627420506e-08, + "loss": 0.1603, + "step": 13171 + }, + { + "epoch": 0.69, + "learning_rate": 9.396294662903222e-08, + "loss": 0.1097, + "step": 13172 + }, + { + "epoch": 0.69, + "learning_rate": 9.39342800144392e-08, + "loss": 0.1782, + "step": 13173 + }, + { + "epoch": 0.69, + "learning_rate": 9.390561643124557e-08, + "loss": 0.1107, + "step": 13174 + }, + { + "epoch": 0.69, + "learning_rate": 9.387695588027029e-08, + "loss": 0.1597, + "step": 13175 + }, + { + "epoch": 0.69, + "learning_rate": 9.384829836233252e-08, + "loss": 0.1279, + "step": 13176 + }, + { + "epoch": 0.69, + "learning_rate": 9.381964387825144e-08, + "loss": 0.1115, + "step": 13177 + }, + { + "epoch": 0.69, + "learning_rate": 9.379099242884589e-08, + "loss": 0.1575, + "step": 13178 + }, + { + "epoch": 0.69, + "learning_rate": 9.376234401493481e-08, + "loss": 0.1839, + "step": 13179 + }, + { + "epoch": 0.69, + "learning_rate": 9.373369863733687e-08, + "loss": 0.1546, + "step": 13180 + }, + { + "epoch": 0.69, + "learning_rate": 9.370505629687091e-08, + "loss": 0.1401, + "step": 13181 + }, + { + "epoch": 0.69, + "learning_rate": 9.367641699435549e-08, + "loss": 0.1565, + "step": 13182 + }, + { + "epoch": 0.69, + "learning_rate": 9.36477807306092e-08, + "loss": 0.1189, + "step": 13183 + }, + { + "epoch": 0.69, + "learning_rate": 9.361914750645021e-08, + "loss": 0.1358, + "step": 13184 + }, + { + "epoch": 0.69, + "learning_rate": 9.359051732269714e-08, + "loss": 0.1382, + "step": 13185 + }, + { + "epoch": 0.69, + "learning_rate": 9.356189018016818e-08, + "loss": 0.1284, + "step": 13186 + }, + { + "epoch": 0.69, + "learning_rate": 9.353326607968145e-08, + "loss": 0.1221, + "step": 13187 + }, + { + "epoch": 0.69, + "learning_rate": 9.350464502205504e-08, + "loss": 0.1407, + "step": 13188 + }, + { + "epoch": 0.69, + "learning_rate": 9.347602700810706e-08, + "loss": 0.1372, + "step": 13189 + }, + { + "epoch": 0.69, + "learning_rate": 9.344741203865534e-08, + "loss": 0.1484, + "step": 13190 + }, + { + "epoch": 0.69, + "learning_rate": 9.34188001145177e-08, + "loss": 0.1328, + "step": 13191 + }, + { + "epoch": 0.69, + "learning_rate": 9.339019123651192e-08, + "loss": 0.142, + "step": 13192 + }, + { + "epoch": 0.69, + "learning_rate": 9.336158540545563e-08, + "loss": 0.1296, + "step": 13193 + }, + { + "epoch": 0.69, + "learning_rate": 9.333298262216638e-08, + "loss": 0.1449, + "step": 13194 + }, + { + "epoch": 0.69, + "learning_rate": 9.330438288746161e-08, + "loss": 0.1428, + "step": 13195 + }, + { + "epoch": 0.69, + "learning_rate": 9.327578620215881e-08, + "loss": 0.1547, + "step": 13196 + }, + { + "epoch": 0.69, + "learning_rate": 9.324719256707527e-08, + "loss": 0.1166, + "step": 13197 + }, + { + "epoch": 0.69, + "learning_rate": 9.321860198302818e-08, + "loss": 0.1253, + "step": 13198 + }, + { + "epoch": 0.69, + "learning_rate": 9.319001445083457e-08, + "loss": 0.1407, + "step": 13199 + }, + { + "epoch": 0.69, + "learning_rate": 9.316142997131173e-08, + "loss": 0.1586, + "step": 13200 + }, + { + "epoch": 0.69, + "learning_rate": 9.313284854527641e-08, + "loss": 0.1366, + "step": 13201 + }, + { + "epoch": 0.69, + "learning_rate": 9.310427017354543e-08, + "loss": 0.1216, + "step": 13202 + }, + { + "epoch": 0.69, + "learning_rate": 9.307569485693579e-08, + "loss": 0.114, + "step": 13203 + }, + { + "epoch": 0.69, + "learning_rate": 9.304712259626404e-08, + "loss": 0.138, + "step": 13204 + }, + { + "epoch": 0.69, + "learning_rate": 9.301855339234681e-08, + "loss": 0.13, + "step": 13205 + }, + { + "epoch": 0.69, + "learning_rate": 9.298998724600056e-08, + "loss": 0.1236, + "step": 13206 + }, + { + "epoch": 0.69, + "learning_rate": 9.296142415804184e-08, + "loss": 0.127, + "step": 13207 + }, + { + "epoch": 0.69, + "learning_rate": 9.293286412928704e-08, + "loss": 0.1423, + "step": 13208 + }, + { + "epoch": 0.69, + "learning_rate": 9.290430716055215e-08, + "loss": 0.1471, + "step": 13209 + }, + { + "epoch": 0.69, + "learning_rate": 9.287575325265358e-08, + "loss": 0.1606, + "step": 13210 + }, + { + "epoch": 0.69, + "learning_rate": 9.284720240640735e-08, + "loss": 0.0907, + "step": 13211 + }, + { + "epoch": 0.69, + "learning_rate": 9.281865462262943e-08, + "loss": 0.1412, + "step": 13212 + }, + { + "epoch": 0.69, + "learning_rate": 9.279010990213565e-08, + "loss": 0.1325, + "step": 13213 + }, + { + "epoch": 0.69, + "learning_rate": 9.276156824574202e-08, + "loss": 0.1297, + "step": 13214 + }, + { + "epoch": 0.69, + "learning_rate": 9.273302965426414e-08, + "loss": 0.1322, + "step": 13215 + }, + { + "epoch": 0.69, + "learning_rate": 9.27044941285177e-08, + "loss": 0.1262, + "step": 13216 + }, + { + "epoch": 0.69, + "learning_rate": 9.267596166931824e-08, + "loss": 0.1357, + "step": 13217 + }, + { + "epoch": 0.69, + "learning_rate": 9.264743227748119e-08, + "loss": 0.1473, + "step": 13218 + }, + { + "epoch": 0.69, + "learning_rate": 9.261890595382201e-08, + "loss": 0.1286, + "step": 13219 + }, + { + "epoch": 0.69, + "learning_rate": 9.259038269915586e-08, + "loss": 0.1602, + "step": 13220 + }, + { + "epoch": 0.69, + "learning_rate": 9.256186251429811e-08, + "loss": 0.1484, + "step": 13221 + }, + { + "epoch": 0.69, + "learning_rate": 9.25333454000638e-08, + "loss": 0.1165, + "step": 13222 + }, + { + "epoch": 0.69, + "learning_rate": 9.250483135726799e-08, + "loss": 0.1106, + "step": 13223 + }, + { + "epoch": 0.69, + "learning_rate": 9.24763203867255e-08, + "loss": 0.0992, + "step": 13224 + }, + { + "epoch": 0.69, + "learning_rate": 9.244781248925145e-08, + "loss": 0.1022, + "step": 13225 + }, + { + "epoch": 0.69, + "learning_rate": 9.241930766566035e-08, + "loss": 0.1604, + "step": 13226 + }, + { + "epoch": 0.69, + "learning_rate": 9.239080591676688e-08, + "loss": 0.1367, + "step": 13227 + }, + { + "epoch": 0.69, + "learning_rate": 9.23623072433858e-08, + "loss": 0.1413, + "step": 13228 + }, + { + "epoch": 0.69, + "learning_rate": 9.233381164633154e-08, + "loss": 0.1106, + "step": 13229 + }, + { + "epoch": 0.69, + "learning_rate": 9.23053191264185e-08, + "loss": 0.1567, + "step": 13230 + }, + { + "epoch": 0.69, + "learning_rate": 9.227682968446095e-08, + "loss": 0.1018, + "step": 13231 + }, + { + "epoch": 0.69, + "learning_rate": 9.224834332127326e-08, + "loss": 0.1216, + "step": 13232 + }, + { + "epoch": 0.69, + "learning_rate": 9.221986003766951e-08, + "loss": 0.1356, + "step": 13233 + }, + { + "epoch": 0.69, + "learning_rate": 9.219137983446377e-08, + "loss": 0.1579, + "step": 13234 + }, + { + "epoch": 0.69, + "learning_rate": 9.216290271247003e-08, + "loss": 0.1395, + "step": 13235 + }, + { + "epoch": 0.69, + "learning_rate": 9.213442867250214e-08, + "loss": 0.137, + "step": 13236 + }, + { + "epoch": 0.69, + "learning_rate": 9.210595771537391e-08, + "loss": 0.1533, + "step": 13237 + }, + { + "epoch": 0.69, + "learning_rate": 9.2077489841899e-08, + "loss": 0.142, + "step": 13238 + }, + { + "epoch": 0.69, + "learning_rate": 9.204902505289116e-08, + "loss": 0.1459, + "step": 13239 + }, + { + "epoch": 0.69, + "learning_rate": 9.202056334916389e-08, + "loss": 0.1343, + "step": 13240 + }, + { + "epoch": 0.69, + "learning_rate": 9.199210473153056e-08, + "loss": 0.1436, + "step": 13241 + }, + { + "epoch": 0.69, + "learning_rate": 9.19636492008045e-08, + "loss": 0.1247, + "step": 13242 + }, + { + "epoch": 0.69, + "learning_rate": 9.193519675779921e-08, + "loss": 0.155, + "step": 13243 + }, + { + "epoch": 0.69, + "learning_rate": 9.190674740332767e-08, + "loss": 0.1382, + "step": 13244 + }, + { + "epoch": 0.69, + "learning_rate": 9.187830113820289e-08, + "loss": 0.144, + "step": 13245 + }, + { + "epoch": 0.69, + "learning_rate": 9.184985796323811e-08, + "loss": 0.1416, + "step": 13246 + }, + { + "epoch": 0.69, + "learning_rate": 9.182141787924616e-08, + "loss": 0.1693, + "step": 13247 + }, + { + "epoch": 0.69, + "learning_rate": 9.179298088703982e-08, + "loss": 0.1242, + "step": 13248 + }, + { + "epoch": 0.69, + "learning_rate": 9.176454698743176e-08, + "loss": 0.1417, + "step": 13249 + }, + { + "epoch": 0.69, + "learning_rate": 9.173611618123484e-08, + "loss": 0.1603, + "step": 13250 + }, + { + "epoch": 0.69, + "learning_rate": 9.170768846926158e-08, + "loss": 0.1175, + "step": 13251 + }, + { + "epoch": 0.69, + "learning_rate": 9.167926385232423e-08, + "loss": 0.1392, + "step": 13252 + }, + { + "epoch": 0.69, + "learning_rate": 9.165084233123544e-08, + "loss": 0.1418, + "step": 13253 + }, + { + "epoch": 0.69, + "learning_rate": 9.162242390680737e-08, + "loss": 0.1216, + "step": 13254 + }, + { + "epoch": 0.69, + "learning_rate": 9.159400857985227e-08, + "loss": 0.1175, + "step": 13255 + }, + { + "epoch": 0.69, + "learning_rate": 9.156559635118217e-08, + "loss": 0.1451, + "step": 13256 + }, + { + "epoch": 0.69, + "learning_rate": 9.153718722160928e-08, + "loss": 0.1208, + "step": 13257 + }, + { + "epoch": 0.69, + "learning_rate": 9.150878119194545e-08, + "loss": 0.1377, + "step": 13258 + }, + { + "epoch": 0.69, + "learning_rate": 9.148037826300252e-08, + "loss": 0.1471, + "step": 13259 + }, + { + "epoch": 0.69, + "learning_rate": 9.14519784355923e-08, + "loss": 0.1236, + "step": 13260 + }, + { + "epoch": 0.69, + "learning_rate": 9.142358171052642e-08, + "loss": 0.1297, + "step": 13261 + }, + { + "epoch": 0.69, + "learning_rate": 9.139518808861648e-08, + "loss": 0.1235, + "step": 13262 + }, + { + "epoch": 0.69, + "learning_rate": 9.136679757067393e-08, + "loss": 0.1631, + "step": 13263 + }, + { + "epoch": 0.69, + "learning_rate": 9.133841015751031e-08, + "loss": 0.122, + "step": 13264 + }, + { + "epoch": 0.69, + "learning_rate": 9.131002584993688e-08, + "loss": 0.1677, + "step": 13265 + }, + { + "epoch": 0.69, + "learning_rate": 9.128164464876486e-08, + "loss": 0.1625, + "step": 13266 + }, + { + "epoch": 0.69, + "learning_rate": 9.125326655480534e-08, + "loss": 0.136, + "step": 13267 + }, + { + "epoch": 0.69, + "learning_rate": 9.122489156886959e-08, + "loss": 0.1342, + "step": 13268 + }, + { + "epoch": 0.69, + "learning_rate": 9.119651969176836e-08, + "loss": 0.1382, + "step": 13269 + }, + { + "epoch": 0.69, + "learning_rate": 9.116815092431252e-08, + "loss": 0.1488, + "step": 13270 + }, + { + "epoch": 0.69, + "learning_rate": 9.1139785267313e-08, + "loss": 0.1416, + "step": 13271 + }, + { + "epoch": 0.69, + "learning_rate": 9.111142272158048e-08, + "loss": 0.1569, + "step": 13272 + }, + { + "epoch": 0.69, + "learning_rate": 9.10830632879255e-08, + "loss": 0.1435, + "step": 13273 + }, + { + "epoch": 0.69, + "learning_rate": 9.105470696715856e-08, + "loss": 0.1136, + "step": 13274 + }, + { + "epoch": 0.69, + "learning_rate": 9.10263537600902e-08, + "loss": 0.143, + "step": 13275 + }, + { + "epoch": 0.69, + "learning_rate": 9.099800366753075e-08, + "loss": 0.1057, + "step": 13276 + }, + { + "epoch": 0.69, + "learning_rate": 9.096965669029042e-08, + "loss": 0.1064, + "step": 13277 + }, + { + "epoch": 0.69, + "learning_rate": 9.094131282917939e-08, + "loss": 0.1155, + "step": 13278 + }, + { + "epoch": 0.69, + "learning_rate": 9.091297208500774e-08, + "loss": 0.1522, + "step": 13279 + }, + { + "epoch": 0.69, + "learning_rate": 9.088463445858547e-08, + "loss": 0.187, + "step": 13280 + }, + { + "epoch": 0.69, + "learning_rate": 9.08562999507224e-08, + "loss": 0.1527, + "step": 13281 + }, + { + "epoch": 0.69, + "learning_rate": 9.082796856222846e-08, + "loss": 0.1039, + "step": 13282 + }, + { + "epoch": 0.69, + "learning_rate": 9.079964029391334e-08, + "loss": 0.1003, + "step": 13283 + }, + { + "epoch": 0.69, + "learning_rate": 9.077131514658665e-08, + "loss": 0.1224, + "step": 13284 + }, + { + "epoch": 0.69, + "learning_rate": 9.074299312105796e-08, + "loss": 0.1031, + "step": 13285 + }, + { + "epoch": 0.69, + "learning_rate": 9.071467421813668e-08, + "loss": 0.1187, + "step": 13286 + }, + { + "epoch": 0.69, + "learning_rate": 9.06863584386322e-08, + "loss": 0.1318, + "step": 13287 + }, + { + "epoch": 0.69, + "learning_rate": 9.065804578335372e-08, + "loss": 0.1068, + "step": 13288 + }, + { + "epoch": 0.69, + "learning_rate": 9.062973625311059e-08, + "loss": 0.1368, + "step": 13289 + }, + { + "epoch": 0.69, + "learning_rate": 9.060142984871182e-08, + "loss": 0.1253, + "step": 13290 + }, + { + "epoch": 0.69, + "learning_rate": 9.05731265709664e-08, + "loss": 0.1557, + "step": 13291 + }, + { + "epoch": 0.69, + "learning_rate": 9.05448264206832e-08, + "loss": 0.1755, + "step": 13292 + }, + { + "epoch": 0.69, + "learning_rate": 9.051652939867119e-08, + "loss": 0.1608, + "step": 13293 + }, + { + "epoch": 0.69, + "learning_rate": 9.04882355057391e-08, + "loss": 0.1228, + "step": 13294 + }, + { + "epoch": 0.69, + "learning_rate": 9.045994474269545e-08, + "loss": 0.1829, + "step": 13295 + }, + { + "epoch": 0.69, + "learning_rate": 9.043165711034879e-08, + "loss": 0.1119, + "step": 13296 + }, + { + "epoch": 0.69, + "learning_rate": 9.040337260950775e-08, + "loss": 0.1385, + "step": 13297 + }, + { + "epoch": 0.69, + "learning_rate": 9.037509124098062e-08, + "loss": 0.1494, + "step": 13298 + }, + { + "epoch": 0.69, + "learning_rate": 9.034681300557569e-08, + "loss": 0.1467, + "step": 13299 + }, + { + "epoch": 0.69, + "learning_rate": 9.031853790410113e-08, + "loss": 0.1301, + "step": 13300 + }, + { + "epoch": 0.69, + "learning_rate": 9.029026593736515e-08, + "loss": 0.1321, + "step": 13301 + }, + { + "epoch": 0.69, + "learning_rate": 9.026199710617572e-08, + "loss": 0.1313, + "step": 13302 + }, + { + "epoch": 0.69, + "learning_rate": 9.02337314113408e-08, + "loss": 0.1249, + "step": 13303 + }, + { + "epoch": 0.69, + "learning_rate": 9.02054688536682e-08, + "loss": 0.1149, + "step": 13304 + }, + { + "epoch": 0.69, + "learning_rate": 9.017720943396568e-08, + "loss": 0.1083, + "step": 13305 + }, + { + "epoch": 0.69, + "learning_rate": 9.014895315304089e-08, + "loss": 0.2155, + "step": 13306 + }, + { + "epoch": 0.69, + "learning_rate": 9.012070001170138e-08, + "loss": 0.1206, + "step": 13307 + }, + { + "epoch": 0.69, + "learning_rate": 9.009245001075474e-08, + "loss": 0.1291, + "step": 13308 + }, + { + "epoch": 0.69, + "learning_rate": 9.00642031510083e-08, + "loss": 0.1787, + "step": 13309 + }, + { + "epoch": 0.69, + "learning_rate": 9.00359594332694e-08, + "loss": 0.1024, + "step": 13310 + }, + { + "epoch": 0.69, + "learning_rate": 9.000771885834522e-08, + "loss": 0.1201, + "step": 13311 + }, + { + "epoch": 0.69, + "learning_rate": 8.997948142704291e-08, + "loss": 0.1281, + "step": 13312 + }, + { + "epoch": 0.69, + "learning_rate": 8.995124714016947e-08, + "loss": 0.1345, + "step": 13313 + }, + { + "epoch": 0.69, + "learning_rate": 8.992301599853181e-08, + "loss": 0.1292, + "step": 13314 + }, + { + "epoch": 0.69, + "learning_rate": 8.989478800293691e-08, + "loss": 0.132, + "step": 13315 + }, + { + "epoch": 0.7, + "learning_rate": 8.986656315419145e-08, + "loss": 0.1431, + "step": 13316 + }, + { + "epoch": 0.7, + "learning_rate": 8.983834145310219e-08, + "loss": 0.1208, + "step": 13317 + }, + { + "epoch": 0.7, + "learning_rate": 8.981012290047554e-08, + "loss": 0.1137, + "step": 13318 + }, + { + "epoch": 0.7, + "learning_rate": 8.978190749711821e-08, + "loss": 0.1295, + "step": 13319 + }, + { + "epoch": 0.7, + "learning_rate": 8.975369524383656e-08, + "loss": 0.1431, + "step": 13320 + }, + { + "epoch": 0.7, + "learning_rate": 8.972548614143673e-08, + "loss": 0.1402, + "step": 13321 + }, + { + "epoch": 0.7, + "learning_rate": 8.969728019072516e-08, + "loss": 0.1529, + "step": 13322 + }, + { + "epoch": 0.7, + "learning_rate": 8.966907739250789e-08, + "loss": 0.1182, + "step": 13323 + }, + { + "epoch": 0.7, + "learning_rate": 8.964087774759097e-08, + "loss": 0.1851, + "step": 13324 + }, + { + "epoch": 0.7, + "learning_rate": 8.961268125678029e-08, + "loss": 0.1085, + "step": 13325 + }, + { + "epoch": 0.7, + "learning_rate": 8.95844879208819e-08, + "loss": 0.1417, + "step": 13326 + }, + { + "epoch": 0.7, + "learning_rate": 8.955629774070146e-08, + "loss": 0.1449, + "step": 13327 + }, + { + "epoch": 0.7, + "learning_rate": 8.952811071704465e-08, + "loss": 0.1271, + "step": 13328 + }, + { + "epoch": 0.7, + "learning_rate": 8.949992685071707e-08, + "loss": 0.1358, + "step": 13329 + }, + { + "epoch": 0.7, + "learning_rate": 8.947174614252428e-08, + "loss": 0.1423, + "step": 13330 + }, + { + "epoch": 0.7, + "learning_rate": 8.944356859327163e-08, + "loss": 0.1612, + "step": 13331 + }, + { + "epoch": 0.7, + "learning_rate": 8.941539420376439e-08, + "loss": 0.1256, + "step": 13332 + }, + { + "epoch": 0.7, + "learning_rate": 8.938722297480796e-08, + "loss": 0.1413, + "step": 13333 + }, + { + "epoch": 0.7, + "learning_rate": 8.935905490720738e-08, + "loss": 0.1272, + "step": 13334 + }, + { + "epoch": 0.7, + "learning_rate": 8.933089000176775e-08, + "loss": 0.1358, + "step": 13335 + }, + { + "epoch": 0.7, + "learning_rate": 8.93027282592939e-08, + "loss": 0.1561, + "step": 13336 + }, + { + "epoch": 0.7, + "learning_rate": 8.9274569680591e-08, + "loss": 0.1323, + "step": 13337 + }, + { + "epoch": 0.7, + "learning_rate": 8.924641426646353e-08, + "loss": 0.1229, + "step": 13338 + }, + { + "epoch": 0.7, + "learning_rate": 8.921826201771623e-08, + "loss": 0.1228, + "step": 13339 + }, + { + "epoch": 0.7, + "learning_rate": 8.919011293515386e-08, + "loss": 0.1194, + "step": 13340 + }, + { + "epoch": 0.7, + "learning_rate": 8.916196701958081e-08, + "loss": 0.1494, + "step": 13341 + }, + { + "epoch": 0.7, + "learning_rate": 8.913382427180156e-08, + "loss": 0.1212, + "step": 13342 + }, + { + "epoch": 0.7, + "learning_rate": 8.910568469262032e-08, + "loss": 0.1061, + "step": 13343 + }, + { + "epoch": 0.7, + "learning_rate": 8.907754828284148e-08, + "loss": 0.1245, + "step": 13344 + }, + { + "epoch": 0.7, + "learning_rate": 8.904941504326924e-08, + "loss": 0.1307, + "step": 13345 + }, + { + "epoch": 0.7, + "learning_rate": 8.902128497470737e-08, + "loss": 0.1014, + "step": 13346 + }, + { + "epoch": 0.7, + "learning_rate": 8.89931580779601e-08, + "loss": 0.1121, + "step": 13347 + }, + { + "epoch": 0.7, + "learning_rate": 8.896503435383125e-08, + "loss": 0.1311, + "step": 13348 + }, + { + "epoch": 0.7, + "learning_rate": 8.893691380312456e-08, + "loss": 0.1332, + "step": 13349 + }, + { + "epoch": 0.7, + "learning_rate": 8.890879642664366e-08, + "loss": 0.1437, + "step": 13350 + }, + { + "epoch": 0.7, + "learning_rate": 8.888068222519234e-08, + "loss": 0.1336, + "step": 13351 + }, + { + "epoch": 0.7, + "learning_rate": 8.885257119957401e-08, + "loss": 0.1372, + "step": 13352 + }, + { + "epoch": 0.7, + "learning_rate": 8.88244633505921e-08, + "loss": 0.1654, + "step": 13353 + }, + { + "epoch": 0.7, + "learning_rate": 8.879635867904995e-08, + "loss": 0.1334, + "step": 13354 + }, + { + "epoch": 0.7, + "learning_rate": 8.87682571857508e-08, + "loss": 0.1464, + "step": 13355 + }, + { + "epoch": 0.7, + "learning_rate": 8.874015887149782e-08, + "loss": 0.1217, + "step": 13356 + }, + { + "epoch": 0.7, + "learning_rate": 8.871206373709394e-08, + "loss": 0.1499, + "step": 13357 + }, + { + "epoch": 0.7, + "learning_rate": 8.868397178334233e-08, + "loss": 0.1441, + "step": 13358 + }, + { + "epoch": 0.7, + "learning_rate": 8.865588301104578e-08, + "loss": 0.1648, + "step": 13359 + }, + { + "epoch": 0.7, + "learning_rate": 8.862779742100708e-08, + "loss": 0.1156, + "step": 13360 + }, + { + "epoch": 0.7, + "learning_rate": 8.859971501402884e-08, + "loss": 0.1012, + "step": 13361 + }, + { + "epoch": 0.7, + "learning_rate": 8.857163579091382e-08, + "loss": 0.1403, + "step": 13362 + }, + { + "epoch": 0.7, + "learning_rate": 8.854355975246454e-08, + "loss": 0.1456, + "step": 13363 + }, + { + "epoch": 0.7, + "learning_rate": 8.851548689948321e-08, + "loss": 0.1459, + "step": 13364 + }, + { + "epoch": 0.7, + "learning_rate": 8.848741723277238e-08, + "loss": 0.1645, + "step": 13365 + }, + { + "epoch": 0.7, + "learning_rate": 8.84593507531342e-08, + "loss": 0.1374, + "step": 13366 + }, + { + "epoch": 0.7, + "learning_rate": 8.843128746137083e-08, + "loss": 0.1406, + "step": 13367 + }, + { + "epoch": 0.7, + "learning_rate": 8.840322735828427e-08, + "loss": 0.1592, + "step": 13368 + }, + { + "epoch": 0.7, + "learning_rate": 8.837517044467664e-08, + "loss": 0.166, + "step": 13369 + }, + { + "epoch": 0.7, + "learning_rate": 8.834711672134971e-08, + "loss": 0.1652, + "step": 13370 + }, + { + "epoch": 0.7, + "learning_rate": 8.831906618910529e-08, + "loss": 0.1007, + "step": 13371 + }, + { + "epoch": 0.7, + "learning_rate": 8.829101884874508e-08, + "loss": 0.1499, + "step": 13372 + }, + { + "epoch": 0.7, + "learning_rate": 8.826297470107067e-08, + "loss": 0.1542, + "step": 13373 + }, + { + "epoch": 0.7, + "learning_rate": 8.823493374688357e-08, + "loss": 0.1273, + "step": 13374 + }, + { + "epoch": 0.7, + "learning_rate": 8.820689598698514e-08, + "loss": 0.1351, + "step": 13375 + }, + { + "epoch": 0.7, + "learning_rate": 8.817886142217687e-08, + "loss": 0.1142, + "step": 13376 + }, + { + "epoch": 0.7, + "learning_rate": 8.81508300532599e-08, + "loss": 0.1634, + "step": 13377 + }, + { + "epoch": 0.7, + "learning_rate": 8.812280188103538e-08, + "loss": 0.093, + "step": 13378 + }, + { + "epoch": 0.7, + "learning_rate": 8.809477690630429e-08, + "loss": 0.0943, + "step": 13379 + }, + { + "epoch": 0.7, + "learning_rate": 8.806675512986783e-08, + "loss": 0.1727, + "step": 13380 + }, + { + "epoch": 0.7, + "learning_rate": 8.803873655252665e-08, + "loss": 0.1372, + "step": 13381 + }, + { + "epoch": 0.7, + "learning_rate": 8.801072117508153e-08, + "loss": 0.1449, + "step": 13382 + }, + { + "epoch": 0.7, + "learning_rate": 8.798270899833328e-08, + "loss": 0.1369, + "step": 13383 + }, + { + "epoch": 0.7, + "learning_rate": 8.795470002308247e-08, + "loss": 0.184, + "step": 13384 + }, + { + "epoch": 0.7, + "learning_rate": 8.792669425012958e-08, + "loss": 0.1103, + "step": 13385 + }, + { + "epoch": 0.7, + "learning_rate": 8.789869168027497e-08, + "loss": 0.1509, + "step": 13386 + }, + { + "epoch": 0.7, + "learning_rate": 8.787069231431909e-08, + "loss": 0.1484, + "step": 13387 + }, + { + "epoch": 0.7, + "learning_rate": 8.784269615306219e-08, + "loss": 0.1454, + "step": 13388 + }, + { + "epoch": 0.7, + "learning_rate": 8.781470319730418e-08, + "loss": 0.1471, + "step": 13389 + }, + { + "epoch": 0.7, + "learning_rate": 8.778671344784535e-08, + "loss": 0.1354, + "step": 13390 + }, + { + "epoch": 0.7, + "learning_rate": 8.775872690548554e-08, + "loss": 0.1346, + "step": 13391 + }, + { + "epoch": 0.7, + "learning_rate": 8.773074357102465e-08, + "loss": 0.1785, + "step": 13392 + }, + { + "epoch": 0.7, + "learning_rate": 8.770276344526239e-08, + "loss": 0.1331, + "step": 13393 + }, + { + "epoch": 0.7, + "learning_rate": 8.767478652899857e-08, + "loss": 0.1535, + "step": 13394 + }, + { + "epoch": 0.7, + "learning_rate": 8.764681282303273e-08, + "loss": 0.1146, + "step": 13395 + }, + { + "epoch": 0.7, + "learning_rate": 8.761884232816433e-08, + "loss": 0.1518, + "step": 13396 + }, + { + "epoch": 0.7, + "learning_rate": 8.75908750451928e-08, + "loss": 0.1224, + "step": 13397 + }, + { + "epoch": 0.7, + "learning_rate": 8.756291097491746e-08, + "loss": 0.142, + "step": 13398 + }, + { + "epoch": 0.7, + "learning_rate": 8.753495011813752e-08, + "loss": 0.1043, + "step": 13399 + }, + { + "epoch": 0.7, + "learning_rate": 8.750699247565205e-08, + "loss": 0.1019, + "step": 13400 + }, + { + "epoch": 0.7, + "learning_rate": 8.747903804826025e-08, + "loss": 0.1362, + "step": 13401 + }, + { + "epoch": 0.7, + "learning_rate": 8.745108683676097e-08, + "loss": 0.1672, + "step": 13402 + }, + { + "epoch": 0.7, + "learning_rate": 8.742313884195309e-08, + "loss": 0.0942, + "step": 13403 + }, + { + "epoch": 0.7, + "learning_rate": 8.739519406463527e-08, + "loss": 0.1555, + "step": 13404 + }, + { + "epoch": 0.7, + "learning_rate": 8.736725250560644e-08, + "loss": 0.1301, + "step": 13405 + }, + { + "epoch": 0.7, + "learning_rate": 8.733931416566491e-08, + "loss": 0.1155, + "step": 13406 + }, + { + "epoch": 0.7, + "learning_rate": 8.73113790456093e-08, + "loss": 0.1205, + "step": 13407 + }, + { + "epoch": 0.7, + "learning_rate": 8.72834471462379e-08, + "loss": 0.1564, + "step": 13408 + }, + { + "epoch": 0.7, + "learning_rate": 8.725551846834916e-08, + "loss": 0.1435, + "step": 13409 + }, + { + "epoch": 0.7, + "learning_rate": 8.722759301274123e-08, + "loss": 0.1182, + "step": 13410 + }, + { + "epoch": 0.7, + "learning_rate": 8.719967078021222e-08, + "loss": 0.1662, + "step": 13411 + }, + { + "epoch": 0.7, + "learning_rate": 8.717175177156009e-08, + "loss": 0.1313, + "step": 13412 + }, + { + "epoch": 0.7, + "learning_rate": 8.714383598758292e-08, + "loss": 0.1264, + "step": 13413 + }, + { + "epoch": 0.7, + "learning_rate": 8.711592342907859e-08, + "loss": 0.1114, + "step": 13414 + }, + { + "epoch": 0.7, + "learning_rate": 8.708801409684456e-08, + "loss": 0.1178, + "step": 13415 + }, + { + "epoch": 0.7, + "learning_rate": 8.706010799167878e-08, + "loss": 0.1197, + "step": 13416 + }, + { + "epoch": 0.7, + "learning_rate": 8.70322051143787e-08, + "loss": 0.1318, + "step": 13417 + }, + { + "epoch": 0.7, + "learning_rate": 8.70043054657418e-08, + "loss": 0.1735, + "step": 13418 + }, + { + "epoch": 0.7, + "learning_rate": 8.69764090465654e-08, + "loss": 0.1002, + "step": 13419 + }, + { + "epoch": 0.7, + "learning_rate": 8.694851585764693e-08, + "loss": 0.1288, + "step": 13420 + }, + { + "epoch": 0.7, + "learning_rate": 8.692062589978353e-08, + "loss": 0.1263, + "step": 13421 + }, + { + "epoch": 0.7, + "learning_rate": 8.689273917377229e-08, + "loss": 0.1293, + "step": 13422 + }, + { + "epoch": 0.7, + "learning_rate": 8.686485568041022e-08, + "loss": 0.1575, + "step": 13423 + }, + { + "epoch": 0.7, + "learning_rate": 8.683697542049426e-08, + "loss": 0.1218, + "step": 13424 + }, + { + "epoch": 0.7, + "learning_rate": 8.680909839482124e-08, + "loss": 0.1586, + "step": 13425 + }, + { + "epoch": 0.7, + "learning_rate": 8.67812246041878e-08, + "loss": 0.1203, + "step": 13426 + }, + { + "epoch": 0.7, + "learning_rate": 8.675335404939073e-08, + "loss": 0.1442, + "step": 13427 + }, + { + "epoch": 0.7, + "learning_rate": 8.672548673122653e-08, + "loss": 0.1139, + "step": 13428 + }, + { + "epoch": 0.7, + "learning_rate": 8.669762265049165e-08, + "loss": 0.1349, + "step": 13429 + }, + { + "epoch": 0.7, + "learning_rate": 8.666976180798237e-08, + "loss": 0.1366, + "step": 13430 + }, + { + "epoch": 0.7, + "learning_rate": 8.664190420449519e-08, + "loss": 0.1299, + "step": 13431 + }, + { + "epoch": 0.7, + "learning_rate": 8.661404984082609e-08, + "loss": 0.134, + "step": 13432 + }, + { + "epoch": 0.7, + "learning_rate": 8.658619871777114e-08, + "loss": 0.1344, + "step": 13433 + }, + { + "epoch": 0.7, + "learning_rate": 8.655835083612648e-08, + "loss": 0.1696, + "step": 13434 + }, + { + "epoch": 0.7, + "learning_rate": 8.653050619668796e-08, + "loss": 0.1146, + "step": 13435 + }, + { + "epoch": 0.7, + "learning_rate": 8.650266480025137e-08, + "loss": 0.1363, + "step": 13436 + }, + { + "epoch": 0.7, + "learning_rate": 8.647482664761234e-08, + "loss": 0.1477, + "step": 13437 + }, + { + "epoch": 0.7, + "learning_rate": 8.644699173956669e-08, + "loss": 0.1246, + "step": 13438 + }, + { + "epoch": 0.7, + "learning_rate": 8.641916007690984e-08, + "loss": 0.147, + "step": 13439 + }, + { + "epoch": 0.7, + "learning_rate": 8.639133166043723e-08, + "loss": 0.116, + "step": 13440 + }, + { + "epoch": 0.7, + "learning_rate": 8.636350649094424e-08, + "loss": 0.1601, + "step": 13441 + }, + { + "epoch": 0.7, + "learning_rate": 8.63356845692261e-08, + "loss": 0.1822, + "step": 13442 + }, + { + "epoch": 0.7, + "learning_rate": 8.630786589607798e-08, + "loss": 0.1689, + "step": 13443 + }, + { + "epoch": 0.7, + "learning_rate": 8.628005047229485e-08, + "loss": 0.1077, + "step": 13444 + }, + { + "epoch": 0.7, + "learning_rate": 8.625223829867185e-08, + "loss": 0.1209, + "step": 13445 + }, + { + "epoch": 0.7, + "learning_rate": 8.62244293760038e-08, + "loss": 0.1137, + "step": 13446 + }, + { + "epoch": 0.7, + "learning_rate": 8.61966237050855e-08, + "loss": 0.1472, + "step": 13447 + }, + { + "epoch": 0.7, + "learning_rate": 8.61688212867116e-08, + "loss": 0.1293, + "step": 13448 + }, + { + "epoch": 0.7, + "learning_rate": 8.614102212167674e-08, + "loss": 0.1429, + "step": 13449 + }, + { + "epoch": 0.7, + "learning_rate": 8.611322621077541e-08, + "loss": 0.1599, + "step": 13450 + }, + { + "epoch": 0.7, + "learning_rate": 8.608543355480198e-08, + "loss": 0.1464, + "step": 13451 + }, + { + "epoch": 0.7, + "learning_rate": 8.60576441545509e-08, + "loss": 0.1538, + "step": 13452 + }, + { + "epoch": 0.7, + "learning_rate": 8.602985801081633e-08, + "loss": 0.1447, + "step": 13453 + }, + { + "epoch": 0.7, + "learning_rate": 8.600207512439244e-08, + "loss": 0.1301, + "step": 13454 + }, + { + "epoch": 0.7, + "learning_rate": 8.597429549607315e-08, + "loss": 0.1253, + "step": 13455 + }, + { + "epoch": 0.7, + "learning_rate": 8.594651912665261e-08, + "loss": 0.1551, + "step": 13456 + }, + { + "epoch": 0.7, + "learning_rate": 8.591874601692464e-08, + "loss": 0.1755, + "step": 13457 + }, + { + "epoch": 0.7, + "learning_rate": 8.589097616768279e-08, + "loss": 0.1377, + "step": 13458 + }, + { + "epoch": 0.7, + "learning_rate": 8.586320957972099e-08, + "loss": 0.1098, + "step": 13459 + }, + { + "epoch": 0.7, + "learning_rate": 8.58354462538327e-08, + "loss": 0.1278, + "step": 13460 + }, + { + "epoch": 0.7, + "learning_rate": 8.580768619081145e-08, + "loss": 0.1375, + "step": 13461 + }, + { + "epoch": 0.7, + "learning_rate": 8.577992939145053e-08, + "loss": 0.135, + "step": 13462 + }, + { + "epoch": 0.7, + "learning_rate": 8.575217585654339e-08, + "loss": 0.1304, + "step": 13463 + }, + { + "epoch": 0.7, + "learning_rate": 8.572442558688318e-08, + "loss": 0.1384, + "step": 13464 + }, + { + "epoch": 0.7, + "learning_rate": 8.5696678583263e-08, + "loss": 0.1468, + "step": 13465 + }, + { + "epoch": 0.7, + "learning_rate": 8.566893484647583e-08, + "loss": 0.1239, + "step": 13466 + }, + { + "epoch": 0.7, + "learning_rate": 8.56411943773147e-08, + "loss": 0.1549, + "step": 13467 + }, + { + "epoch": 0.7, + "learning_rate": 8.561345717657233e-08, + "loss": 0.1451, + "step": 13468 + }, + { + "epoch": 0.7, + "learning_rate": 8.558572324504148e-08, + "loss": 0.1209, + "step": 13469 + }, + { + "epoch": 0.7, + "learning_rate": 8.555799258351488e-08, + "loss": 0.1272, + "step": 13470 + }, + { + "epoch": 0.7, + "learning_rate": 8.553026519278504e-08, + "loss": 0.1294, + "step": 13471 + }, + { + "epoch": 0.7, + "learning_rate": 8.55025410736444e-08, + "loss": 0.113, + "step": 13472 + }, + { + "epoch": 0.7, + "learning_rate": 8.547482022688526e-08, + "loss": 0.1049, + "step": 13473 + }, + { + "epoch": 0.7, + "learning_rate": 8.544710265330012e-08, + "loss": 0.1325, + "step": 13474 + }, + { + "epoch": 0.7, + "learning_rate": 8.541938835368095e-08, + "loss": 0.1311, + "step": 13475 + }, + { + "epoch": 0.7, + "learning_rate": 8.53916773288198e-08, + "loss": 0.1244, + "step": 13476 + }, + { + "epoch": 0.7, + "learning_rate": 8.536396957950884e-08, + "loss": 0.1416, + "step": 13477 + }, + { + "epoch": 0.7, + "learning_rate": 8.533626510653988e-08, + "loss": 0.1467, + "step": 13478 + }, + { + "epoch": 0.7, + "learning_rate": 8.530856391070473e-08, + "loss": 0.1273, + "step": 13479 + }, + { + "epoch": 0.7, + "learning_rate": 8.528086599279503e-08, + "loss": 0.1361, + "step": 13480 + }, + { + "epoch": 0.7, + "learning_rate": 8.525317135360252e-08, + "loss": 0.1731, + "step": 13481 + }, + { + "epoch": 0.7, + "learning_rate": 8.522547999391869e-08, + "loss": 0.1338, + "step": 13482 + }, + { + "epoch": 0.7, + "learning_rate": 8.519779191453495e-08, + "loss": 0.1181, + "step": 13483 + }, + { + "epoch": 0.7, + "learning_rate": 8.517010711624261e-08, + "loss": 0.1258, + "step": 13484 + }, + { + "epoch": 0.7, + "learning_rate": 8.514242559983297e-08, + "loss": 0.1295, + "step": 13485 + }, + { + "epoch": 0.7, + "learning_rate": 8.511474736609714e-08, + "loss": 0.1352, + "step": 13486 + }, + { + "epoch": 0.7, + "learning_rate": 8.508707241582609e-08, + "loss": 0.1448, + "step": 13487 + }, + { + "epoch": 0.7, + "learning_rate": 8.505940074981094e-08, + "loss": 0.1178, + "step": 13488 + }, + { + "epoch": 0.7, + "learning_rate": 8.503173236884252e-08, + "loss": 0.1513, + "step": 13489 + }, + { + "epoch": 0.7, + "learning_rate": 8.500406727371154e-08, + "loss": 0.1574, + "step": 13490 + }, + { + "epoch": 0.7, + "learning_rate": 8.497640546520873e-08, + "loss": 0.1238, + "step": 13491 + }, + { + "epoch": 0.7, + "learning_rate": 8.494874694412464e-08, + "loss": 0.112, + "step": 13492 + }, + { + "epoch": 0.7, + "learning_rate": 8.492109171124979e-08, + "loss": 0.1067, + "step": 13493 + }, + { + "epoch": 0.7, + "learning_rate": 8.489343976737448e-08, + "loss": 0.1621, + "step": 13494 + }, + { + "epoch": 0.7, + "learning_rate": 8.486579111328917e-08, + "loss": 0.1468, + "step": 13495 + }, + { + "epoch": 0.7, + "learning_rate": 8.4838145749784e-08, + "loss": 0.1293, + "step": 13496 + }, + { + "epoch": 0.7, + "learning_rate": 8.481050367764908e-08, + "loss": 0.1364, + "step": 13497 + }, + { + "epoch": 0.7, + "learning_rate": 8.478286489767437e-08, + "loss": 0.1395, + "step": 13498 + }, + { + "epoch": 0.7, + "learning_rate": 8.475522941064991e-08, + "loss": 0.1394, + "step": 13499 + }, + { + "epoch": 0.7, + "learning_rate": 8.472759721736558e-08, + "loss": 0.1191, + "step": 13500 + }, + { + "epoch": 0.7, + "learning_rate": 8.469996831861086e-08, + "loss": 0.1296, + "step": 13501 + }, + { + "epoch": 0.7, + "learning_rate": 8.467234271517565e-08, + "loss": 0.1333, + "step": 13502 + }, + { + "epoch": 0.7, + "learning_rate": 8.464472040784937e-08, + "loss": 0.1342, + "step": 13503 + }, + { + "epoch": 0.7, + "learning_rate": 8.461710139742152e-08, + "loss": 0.1294, + "step": 13504 + }, + { + "epoch": 0.7, + "learning_rate": 8.45894856846814e-08, + "loss": 0.0978, + "step": 13505 + }, + { + "epoch": 0.7, + "learning_rate": 8.456187327041842e-08, + "loss": 0.1221, + "step": 13506 + }, + { + "epoch": 0.71, + "learning_rate": 8.453426415542164e-08, + "loss": 0.1546, + "step": 13507 + }, + { + "epoch": 0.71, + "learning_rate": 8.450665834048016e-08, + "loss": 0.1329, + "step": 13508 + }, + { + "epoch": 0.71, + "learning_rate": 8.447905582638297e-08, + "loss": 0.1472, + "step": 13509 + }, + { + "epoch": 0.71, + "learning_rate": 8.445145661391896e-08, + "loss": 0.1477, + "step": 13510 + }, + { + "epoch": 0.71, + "learning_rate": 8.442386070387695e-08, + "loss": 0.1265, + "step": 13511 + }, + { + "epoch": 0.71, + "learning_rate": 8.439626809704555e-08, + "loss": 0.127, + "step": 13512 + }, + { + "epoch": 0.71, + "learning_rate": 8.436867879421347e-08, + "loss": 0.1271, + "step": 13513 + }, + { + "epoch": 0.71, + "learning_rate": 8.43410927961692e-08, + "loss": 0.1614, + "step": 13514 + }, + { + "epoch": 0.71, + "learning_rate": 8.431351010370115e-08, + "loss": 0.1345, + "step": 13515 + }, + { + "epoch": 0.71, + "learning_rate": 8.428593071759758e-08, + "loss": 0.1636, + "step": 13516 + }, + { + "epoch": 0.71, + "learning_rate": 8.425835463864692e-08, + "loss": 0.1728, + "step": 13517 + }, + { + "epoch": 0.71, + "learning_rate": 8.423078186763708e-08, + "loss": 0.1546, + "step": 13518 + }, + { + "epoch": 0.71, + "learning_rate": 8.42032124053562e-08, + "loss": 0.1586, + "step": 13519 + }, + { + "epoch": 0.71, + "learning_rate": 8.417564625259212e-08, + "loss": 0.138, + "step": 13520 + }, + { + "epoch": 0.71, + "learning_rate": 8.414808341013288e-08, + "loss": 0.139, + "step": 13521 + }, + { + "epoch": 0.71, + "learning_rate": 8.412052387876614e-08, + "loss": 0.1497, + "step": 13522 + }, + { + "epoch": 0.71, + "learning_rate": 8.409296765927953e-08, + "loss": 0.1252, + "step": 13523 + }, + { + "epoch": 0.71, + "learning_rate": 8.40654147524606e-08, + "loss": 0.1181, + "step": 13524 + }, + { + "epoch": 0.71, + "learning_rate": 8.403786515909701e-08, + "loss": 0.0991, + "step": 13525 + }, + { + "epoch": 0.71, + "learning_rate": 8.401031887997594e-08, + "loss": 0.0884, + "step": 13526 + }, + { + "epoch": 0.71, + "learning_rate": 8.398277591588465e-08, + "loss": 0.1472, + "step": 13527 + }, + { + "epoch": 0.71, + "learning_rate": 8.395523626761047e-08, + "loss": 0.1466, + "step": 13528 + }, + { + "epoch": 0.71, + "learning_rate": 8.392769993594047e-08, + "loss": 0.1346, + "step": 13529 + }, + { + "epoch": 0.71, + "learning_rate": 8.390016692166158e-08, + "loss": 0.1332, + "step": 13530 + }, + { + "epoch": 0.71, + "learning_rate": 8.387263722556068e-08, + "loss": 0.1541, + "step": 13531 + }, + { + "epoch": 0.71, + "learning_rate": 8.384511084842471e-08, + "loss": 0.1762, + "step": 13532 + }, + { + "epoch": 0.71, + "learning_rate": 8.381758779104031e-08, + "loss": 0.1579, + "step": 13533 + }, + { + "epoch": 0.71, + "learning_rate": 8.379006805419411e-08, + "loss": 0.1927, + "step": 13534 + }, + { + "epoch": 0.71, + "learning_rate": 8.37625516386726e-08, + "loss": 0.1548, + "step": 13535 + }, + { + "epoch": 0.71, + "learning_rate": 8.373503854526227e-08, + "loss": 0.1172, + "step": 13536 + }, + { + "epoch": 0.71, + "learning_rate": 8.37075287747494e-08, + "loss": 0.1332, + "step": 13537 + }, + { + "epoch": 0.71, + "learning_rate": 8.36800223279202e-08, + "loss": 0.1184, + "step": 13538 + }, + { + "epoch": 0.71, + "learning_rate": 8.365251920556091e-08, + "loss": 0.1514, + "step": 13539 + }, + { + "epoch": 0.71, + "learning_rate": 8.362501940845756e-08, + "loss": 0.1161, + "step": 13540 + }, + { + "epoch": 0.71, + "learning_rate": 8.359752293739608e-08, + "loss": 0.1326, + "step": 13541 + }, + { + "epoch": 0.71, + "learning_rate": 8.357002979316233e-08, + "loss": 0.162, + "step": 13542 + }, + { + "epoch": 0.71, + "learning_rate": 8.354253997654205e-08, + "loss": 0.1626, + "step": 13543 + }, + { + "epoch": 0.71, + "learning_rate": 8.351505348832095e-08, + "loss": 0.1094, + "step": 13544 + }, + { + "epoch": 0.71, + "learning_rate": 8.348757032928452e-08, + "loss": 0.1335, + "step": 13545 + }, + { + "epoch": 0.71, + "learning_rate": 8.346009050021838e-08, + "loss": 0.1201, + "step": 13546 + }, + { + "epoch": 0.71, + "learning_rate": 8.343261400190784e-08, + "loss": 0.1481, + "step": 13547 + }, + { + "epoch": 0.71, + "learning_rate": 8.340514083513819e-08, + "loss": 0.1354, + "step": 13548 + }, + { + "epoch": 0.71, + "learning_rate": 8.337767100069455e-08, + "loss": 0.1276, + "step": 13549 + }, + { + "epoch": 0.71, + "learning_rate": 8.335020449936216e-08, + "loss": 0.1485, + "step": 13550 + }, + { + "epoch": 0.71, + "learning_rate": 8.332274133192604e-08, + "loss": 0.1796, + "step": 13551 + }, + { + "epoch": 0.71, + "learning_rate": 8.32952814991709e-08, + "loss": 0.1438, + "step": 13552 + }, + { + "epoch": 0.71, + "learning_rate": 8.32678250018817e-08, + "loss": 0.142, + "step": 13553 + }, + { + "epoch": 0.71, + "learning_rate": 8.324037184084314e-08, + "loss": 0.1155, + "step": 13554 + }, + { + "epoch": 0.71, + "learning_rate": 8.321292201683981e-08, + "loss": 0.1358, + "step": 13555 + }, + { + "epoch": 0.71, + "learning_rate": 8.318547553065618e-08, + "loss": 0.1264, + "step": 13556 + }, + { + "epoch": 0.71, + "learning_rate": 8.315803238307684e-08, + "loss": 0.1344, + "step": 13557 + }, + { + "epoch": 0.71, + "learning_rate": 8.313059257488603e-08, + "loss": 0.1385, + "step": 13558 + }, + { + "epoch": 0.71, + "learning_rate": 8.310315610686803e-08, + "loss": 0.1357, + "step": 13559 + }, + { + "epoch": 0.71, + "learning_rate": 8.307572297980692e-08, + "loss": 0.1131, + "step": 13560 + }, + { + "epoch": 0.71, + "learning_rate": 8.30482931944868e-08, + "loss": 0.124, + "step": 13561 + }, + { + "epoch": 0.71, + "learning_rate": 8.302086675169162e-08, + "loss": 0.1461, + "step": 13562 + }, + { + "epoch": 0.71, + "learning_rate": 8.299344365220513e-08, + "loss": 0.1226, + "step": 13563 + }, + { + "epoch": 0.71, + "learning_rate": 8.296602389681127e-08, + "loss": 0.155, + "step": 13564 + }, + { + "epoch": 0.71, + "learning_rate": 8.293860748629362e-08, + "loss": 0.1283, + "step": 13565 + }, + { + "epoch": 0.71, + "learning_rate": 8.291119442143577e-08, + "loss": 0.1467, + "step": 13566 + }, + { + "epoch": 0.71, + "learning_rate": 8.288378470302111e-08, + "loss": 0.1053, + "step": 13567 + }, + { + "epoch": 0.71, + "learning_rate": 8.285637833183325e-08, + "loss": 0.1308, + "step": 13568 + }, + { + "epoch": 0.71, + "learning_rate": 8.282897530865523e-08, + "loss": 0.1294, + "step": 13569 + }, + { + "epoch": 0.71, + "learning_rate": 8.280157563427026e-08, + "loss": 0.1426, + "step": 13570 + }, + { + "epoch": 0.71, + "learning_rate": 8.277417930946158e-08, + "loss": 0.1301, + "step": 13571 + }, + { + "epoch": 0.71, + "learning_rate": 8.27467863350121e-08, + "loss": 0.1224, + "step": 13572 + }, + { + "epoch": 0.71, + "learning_rate": 8.271939671170474e-08, + "loss": 0.1318, + "step": 13573 + }, + { + "epoch": 0.71, + "learning_rate": 8.269201044032221e-08, + "loss": 0.1647, + "step": 13574 + }, + { + "epoch": 0.71, + "learning_rate": 8.26646275216474e-08, + "loss": 0.1294, + "step": 13575 + }, + { + "epoch": 0.71, + "learning_rate": 8.263724795646282e-08, + "loss": 0.1147, + "step": 13576 + }, + { + "epoch": 0.71, + "learning_rate": 8.260987174555101e-08, + "loss": 0.1475, + "step": 13577 + }, + { + "epoch": 0.71, + "learning_rate": 8.258249888969437e-08, + "loss": 0.1236, + "step": 13578 + }, + { + "epoch": 0.71, + "learning_rate": 8.255512938967524e-08, + "loss": 0.1165, + "step": 13579 + }, + { + "epoch": 0.71, + "learning_rate": 8.252776324627585e-08, + "loss": 0.1777, + "step": 13580 + }, + { + "epoch": 0.71, + "learning_rate": 8.250040046027826e-08, + "loss": 0.1405, + "step": 13581 + }, + { + "epoch": 0.71, + "learning_rate": 8.247304103246466e-08, + "loss": 0.1621, + "step": 13582 + }, + { + "epoch": 0.71, + "learning_rate": 8.244568496361692e-08, + "loss": 0.1722, + "step": 13583 + }, + { + "epoch": 0.71, + "learning_rate": 8.241833225451689e-08, + "loss": 0.1887, + "step": 13584 + }, + { + "epoch": 0.71, + "learning_rate": 8.23909829059463e-08, + "loss": 0.1083, + "step": 13585 + }, + { + "epoch": 0.71, + "learning_rate": 8.236363691868684e-08, + "loss": 0.1089, + "step": 13586 + }, + { + "epoch": 0.71, + "learning_rate": 8.233629429352005e-08, + "loss": 0.1084, + "step": 13587 + }, + { + "epoch": 0.71, + "learning_rate": 8.230895503122731e-08, + "loss": 0.1412, + "step": 13588 + }, + { + "epoch": 0.71, + "learning_rate": 8.228161913259013e-08, + "loss": 0.1626, + "step": 13589 + }, + { + "epoch": 0.71, + "learning_rate": 8.225428659838973e-08, + "loss": 0.1895, + "step": 13590 + }, + { + "epoch": 0.71, + "learning_rate": 8.222695742940726e-08, + "loss": 0.1239, + "step": 13591 + }, + { + "epoch": 0.71, + "learning_rate": 8.219963162642377e-08, + "loss": 0.1402, + "step": 13592 + }, + { + "epoch": 0.71, + "learning_rate": 8.217230919022032e-08, + "loss": 0.1494, + "step": 13593 + }, + { + "epoch": 0.71, + "learning_rate": 8.214499012157786e-08, + "loss": 0.1296, + "step": 13594 + }, + { + "epoch": 0.71, + "learning_rate": 8.211767442127691e-08, + "loss": 0.1345, + "step": 13595 + }, + { + "epoch": 0.71, + "learning_rate": 8.209036209009844e-08, + "loss": 0.0874, + "step": 13596 + }, + { + "epoch": 0.71, + "learning_rate": 8.206305312882289e-08, + "loss": 0.1416, + "step": 13597 + }, + { + "epoch": 0.71, + "learning_rate": 8.203574753823084e-08, + "loss": 0.153, + "step": 13598 + }, + { + "epoch": 0.71, + "learning_rate": 8.200844531910258e-08, + "loss": 0.1397, + "step": 13599 + }, + { + "epoch": 0.71, + "learning_rate": 8.198114647221859e-08, + "loss": 0.1888, + "step": 13600 + }, + { + "epoch": 0.71, + "learning_rate": 8.195385099835899e-08, + "loss": 0.1272, + "step": 13601 + }, + { + "epoch": 0.71, + "learning_rate": 8.192655889830389e-08, + "loss": 0.1182, + "step": 13602 + }, + { + "epoch": 0.71, + "learning_rate": 8.189927017283334e-08, + "loss": 0.1137, + "step": 13603 + }, + { + "epoch": 0.71, + "learning_rate": 8.187198482272722e-08, + "loss": 0.1077, + "step": 13604 + }, + { + "epoch": 0.71, + "learning_rate": 8.184470284876536e-08, + "loss": 0.1227, + "step": 13605 + }, + { + "epoch": 0.71, + "learning_rate": 8.181742425172746e-08, + "loss": 0.1194, + "step": 13606 + }, + { + "epoch": 0.71, + "learning_rate": 8.179014903239326e-08, + "loss": 0.1491, + "step": 13607 + }, + { + "epoch": 0.71, + "learning_rate": 8.176287719154224e-08, + "loss": 0.1576, + "step": 13608 + }, + { + "epoch": 0.71, + "learning_rate": 8.173560872995382e-08, + "loss": 0.1254, + "step": 13609 + }, + { + "epoch": 0.71, + "learning_rate": 8.17083436484073e-08, + "loss": 0.1615, + "step": 13610 + }, + { + "epoch": 0.71, + "learning_rate": 8.168108194768213e-08, + "loss": 0.1794, + "step": 13611 + }, + { + "epoch": 0.71, + "learning_rate": 8.165382362855721e-08, + "loss": 0.126, + "step": 13612 + }, + { + "epoch": 0.71, + "learning_rate": 8.162656869181166e-08, + "loss": 0.1146, + "step": 13613 + }, + { + "epoch": 0.71, + "learning_rate": 8.159931713822454e-08, + "loss": 0.1079, + "step": 13614 + }, + { + "epoch": 0.71, + "learning_rate": 8.157206896857462e-08, + "loss": 0.1412, + "step": 13615 + }, + { + "epoch": 0.71, + "learning_rate": 8.154482418364073e-08, + "loss": 0.1407, + "step": 13616 + }, + { + "epoch": 0.71, + "learning_rate": 8.151758278420137e-08, + "loss": 0.1111, + "step": 13617 + }, + { + "epoch": 0.71, + "learning_rate": 8.149034477103534e-08, + "loss": 0.1615, + "step": 13618 + }, + { + "epoch": 0.71, + "learning_rate": 8.146311014492099e-08, + "loss": 0.1151, + "step": 13619 + }, + { + "epoch": 0.71, + "learning_rate": 8.143587890663672e-08, + "loss": 0.1306, + "step": 13620 + }, + { + "epoch": 0.71, + "learning_rate": 8.140865105696082e-08, + "loss": 0.15, + "step": 13621 + }, + { + "epoch": 0.71, + "learning_rate": 8.138142659667142e-08, + "loss": 0.1151, + "step": 13622 + }, + { + "epoch": 0.71, + "learning_rate": 8.135420552654665e-08, + "loss": 0.1061, + "step": 13623 + }, + { + "epoch": 0.71, + "learning_rate": 8.132698784736443e-08, + "loss": 0.1267, + "step": 13624 + }, + { + "epoch": 0.71, + "learning_rate": 8.129977355990276e-08, + "loss": 0.1405, + "step": 13625 + }, + { + "epoch": 0.71, + "learning_rate": 8.127256266493941e-08, + "loss": 0.183, + "step": 13626 + }, + { + "epoch": 0.71, + "learning_rate": 8.124535516325201e-08, + "loss": 0.1323, + "step": 13627 + }, + { + "epoch": 0.71, + "learning_rate": 8.121815105561825e-08, + "loss": 0.1283, + "step": 13628 + }, + { + "epoch": 0.71, + "learning_rate": 8.119095034281556e-08, + "loss": 0.1333, + "step": 13629 + }, + { + "epoch": 0.71, + "learning_rate": 8.116375302562137e-08, + "loss": 0.1537, + "step": 13630 + }, + { + "epoch": 0.71, + "learning_rate": 8.1136559104813e-08, + "loss": 0.114, + "step": 13631 + }, + { + "epoch": 0.71, + "learning_rate": 8.110936858116761e-08, + "loss": 0.1307, + "step": 13632 + }, + { + "epoch": 0.71, + "learning_rate": 8.108218145546242e-08, + "loss": 0.1265, + "step": 13633 + }, + { + "epoch": 0.71, + "learning_rate": 8.105499772847438e-08, + "loss": 0.1491, + "step": 13634 + }, + { + "epoch": 0.71, + "learning_rate": 8.102781740098041e-08, + "loss": 0.1346, + "step": 13635 + }, + { + "epoch": 0.71, + "learning_rate": 8.10006404737573e-08, + "loss": 0.1172, + "step": 13636 + }, + { + "epoch": 0.71, + "learning_rate": 8.097346694758195e-08, + "loss": 0.123, + "step": 13637 + }, + { + "epoch": 0.71, + "learning_rate": 8.094629682323082e-08, + "loss": 0.121, + "step": 13638 + }, + { + "epoch": 0.71, + "learning_rate": 8.091913010148039e-08, + "loss": 0.1825, + "step": 13639 + }, + { + "epoch": 0.71, + "learning_rate": 8.089196678310728e-08, + "loss": 0.1431, + "step": 13640 + }, + { + "epoch": 0.71, + "learning_rate": 8.086480686888772e-08, + "loss": 0.1833, + "step": 13641 + }, + { + "epoch": 0.71, + "learning_rate": 8.083765035959798e-08, + "loss": 0.1283, + "step": 13642 + }, + { + "epoch": 0.71, + "learning_rate": 8.081049725601412e-08, + "loss": 0.124, + "step": 13643 + }, + { + "epoch": 0.71, + "learning_rate": 8.078334755891234e-08, + "loss": 0.1283, + "step": 13644 + }, + { + "epoch": 0.71, + "learning_rate": 8.075620126906857e-08, + "loss": 0.141, + "step": 13645 + }, + { + "epoch": 0.71, + "learning_rate": 8.072905838725845e-08, + "loss": 0.1248, + "step": 13646 + }, + { + "epoch": 0.71, + "learning_rate": 8.070191891425797e-08, + "loss": 0.1201, + "step": 13647 + }, + { + "epoch": 0.71, + "learning_rate": 8.06747828508427e-08, + "loss": 0.1762, + "step": 13648 + }, + { + "epoch": 0.71, + "learning_rate": 8.064765019778821e-08, + "loss": 0.1453, + "step": 13649 + }, + { + "epoch": 0.71, + "learning_rate": 8.062052095586986e-08, + "loss": 0.1551, + "step": 13650 + }, + { + "epoch": 0.71, + "learning_rate": 8.059339512586318e-08, + "loss": 0.1209, + "step": 13651 + }, + { + "epoch": 0.71, + "learning_rate": 8.056627270854337e-08, + "loss": 0.1317, + "step": 13652 + }, + { + "epoch": 0.71, + "learning_rate": 8.053915370468561e-08, + "loss": 0.149, + "step": 13653 + }, + { + "epoch": 0.71, + "learning_rate": 8.051203811506493e-08, + "loss": 0.1554, + "step": 13654 + }, + { + "epoch": 0.71, + "learning_rate": 8.048492594045633e-08, + "loss": 0.1244, + "step": 13655 + }, + { + "epoch": 0.71, + "learning_rate": 8.045781718163469e-08, + "loss": 0.1683, + "step": 13656 + }, + { + "epoch": 0.71, + "learning_rate": 8.043071183937471e-08, + "loss": 0.1382, + "step": 13657 + }, + { + "epoch": 0.71, + "learning_rate": 8.04036099144512e-08, + "loss": 0.0932, + "step": 13658 + }, + { + "epoch": 0.71, + "learning_rate": 8.03765114076387e-08, + "loss": 0.1347, + "step": 13659 + }, + { + "epoch": 0.71, + "learning_rate": 8.03494163197117e-08, + "loss": 0.138, + "step": 13660 + }, + { + "epoch": 0.71, + "learning_rate": 8.032232465144448e-08, + "loss": 0.1379, + "step": 13661 + }, + { + "epoch": 0.71, + "learning_rate": 8.029523640361157e-08, + "loss": 0.1355, + "step": 13662 + }, + { + "epoch": 0.71, + "learning_rate": 8.026815157698694e-08, + "loss": 0.1633, + "step": 13663 + }, + { + "epoch": 0.71, + "learning_rate": 8.02410701723447e-08, + "loss": 0.1202, + "step": 13664 + }, + { + "epoch": 0.71, + "learning_rate": 8.021399219045896e-08, + "loss": 0.1705, + "step": 13665 + }, + { + "epoch": 0.71, + "learning_rate": 8.018691763210357e-08, + "loss": 0.1309, + "step": 13666 + }, + { + "epoch": 0.71, + "learning_rate": 8.015984649805232e-08, + "loss": 0.1199, + "step": 13667 + }, + { + "epoch": 0.71, + "learning_rate": 8.013277878907887e-08, + "loss": 0.1174, + "step": 13668 + }, + { + "epoch": 0.71, + "learning_rate": 8.010571450595693e-08, + "loss": 0.1422, + "step": 13669 + }, + { + "epoch": 0.71, + "learning_rate": 8.007865364945996e-08, + "loss": 0.1266, + "step": 13670 + }, + { + "epoch": 0.71, + "learning_rate": 8.005159622036135e-08, + "loss": 0.1131, + "step": 13671 + }, + { + "epoch": 0.71, + "learning_rate": 8.002454221943444e-08, + "loss": 0.1155, + "step": 13672 + }, + { + "epoch": 0.71, + "learning_rate": 7.999749164745244e-08, + "loss": 0.1488, + "step": 13673 + }, + { + "epoch": 0.71, + "learning_rate": 7.997044450518844e-08, + "loss": 0.1573, + "step": 13674 + }, + { + "epoch": 0.71, + "learning_rate": 7.994340079341539e-08, + "loss": 0.1182, + "step": 13675 + }, + { + "epoch": 0.71, + "learning_rate": 7.991636051290638e-08, + "loss": 0.1457, + "step": 13676 + }, + { + "epoch": 0.71, + "learning_rate": 7.988932366443413e-08, + "loss": 0.1192, + "step": 13677 + }, + { + "epoch": 0.71, + "learning_rate": 7.986229024877138e-08, + "loss": 0.1034, + "step": 13678 + }, + { + "epoch": 0.71, + "learning_rate": 7.98352602666907e-08, + "loss": 0.1275, + "step": 13679 + }, + { + "epoch": 0.71, + "learning_rate": 7.980823371896479e-08, + "loss": 0.131, + "step": 13680 + }, + { + "epoch": 0.71, + "learning_rate": 7.978121060636589e-08, + "loss": 0.1226, + "step": 13681 + }, + { + "epoch": 0.71, + "learning_rate": 7.975419092966633e-08, + "loss": 0.1243, + "step": 13682 + }, + { + "epoch": 0.71, + "learning_rate": 7.972717468963848e-08, + "loss": 0.1468, + "step": 13683 + }, + { + "epoch": 0.71, + "learning_rate": 7.970016188705441e-08, + "loss": 0.1451, + "step": 13684 + }, + { + "epoch": 0.71, + "learning_rate": 7.967315252268616e-08, + "loss": 0.1044, + "step": 13685 + }, + { + "epoch": 0.71, + "learning_rate": 7.96461465973056e-08, + "loss": 0.1514, + "step": 13686 + }, + { + "epoch": 0.71, + "learning_rate": 7.96191441116847e-08, + "loss": 0.1014, + "step": 13687 + }, + { + "epoch": 0.71, + "learning_rate": 7.959214506659517e-08, + "loss": 0.1305, + "step": 13688 + }, + { + "epoch": 0.71, + "learning_rate": 7.956514946280853e-08, + "loss": 0.1261, + "step": 13689 + }, + { + "epoch": 0.71, + "learning_rate": 7.953815730109644e-08, + "loss": 0.1135, + "step": 13690 + }, + { + "epoch": 0.71, + "learning_rate": 7.951116858223032e-08, + "loss": 0.1285, + "step": 13691 + }, + { + "epoch": 0.71, + "learning_rate": 7.948418330698152e-08, + "loss": 0.108, + "step": 13692 + }, + { + "epoch": 0.71, + "learning_rate": 7.945720147612123e-08, + "loss": 0.1233, + "step": 13693 + }, + { + "epoch": 0.71, + "learning_rate": 7.94302230904207e-08, + "loss": 0.1231, + "step": 13694 + }, + { + "epoch": 0.71, + "learning_rate": 7.940324815065096e-08, + "loss": 0.1737, + "step": 13695 + }, + { + "epoch": 0.71, + "learning_rate": 7.937627665758293e-08, + "loss": 0.1049, + "step": 13696 + }, + { + "epoch": 0.71, + "learning_rate": 7.93493086119875e-08, + "loss": 0.1351, + "step": 13697 + }, + { + "epoch": 0.71, + "learning_rate": 7.932234401463538e-08, + "loss": 0.1214, + "step": 13698 + }, + { + "epoch": 0.72, + "learning_rate": 7.929538286629726e-08, + "loss": 0.139, + "step": 13699 + }, + { + "epoch": 0.72, + "learning_rate": 7.926842516774364e-08, + "loss": 0.1759, + "step": 13700 + }, + { + "epoch": 0.72, + "learning_rate": 7.924147091974509e-08, + "loss": 0.1424, + "step": 13701 + }, + { + "epoch": 0.72, + "learning_rate": 7.921452012307193e-08, + "loss": 0.1341, + "step": 13702 + }, + { + "epoch": 0.72, + "learning_rate": 7.918757277849439e-08, + "loss": 0.1057, + "step": 13703 + }, + { + "epoch": 0.72, + "learning_rate": 7.91606288867826e-08, + "loss": 0.1348, + "step": 13704 + }, + { + "epoch": 0.72, + "learning_rate": 7.913368844870684e-08, + "loss": 0.1559, + "step": 13705 + }, + { + "epoch": 0.72, + "learning_rate": 7.910675146503684e-08, + "loss": 0.1578, + "step": 13706 + }, + { + "epoch": 0.72, + "learning_rate": 7.907981793654251e-08, + "loss": 0.1259, + "step": 13707 + }, + { + "epoch": 0.72, + "learning_rate": 7.905288786399368e-08, + "loss": 0.1343, + "step": 13708 + }, + { + "epoch": 0.72, + "learning_rate": 7.902596124816006e-08, + "loss": 0.1457, + "step": 13709 + }, + { + "epoch": 0.72, + "learning_rate": 7.899903808981113e-08, + "loss": 0.157, + "step": 13710 + }, + { + "epoch": 0.72, + "learning_rate": 7.897211838971633e-08, + "loss": 0.1343, + "step": 13711 + }, + { + "epoch": 0.72, + "learning_rate": 7.894520214864519e-08, + "loss": 0.1389, + "step": 13712 + }, + { + "epoch": 0.72, + "learning_rate": 7.891828936736693e-08, + "loss": 0.1421, + "step": 13713 + }, + { + "epoch": 0.72, + "learning_rate": 7.889138004665067e-08, + "loss": 0.1606, + "step": 13714 + }, + { + "epoch": 0.72, + "learning_rate": 7.886447418726553e-08, + "loss": 0.1501, + "step": 13715 + }, + { + "epoch": 0.72, + "learning_rate": 7.883757178998047e-08, + "loss": 0.1444, + "step": 13716 + }, + { + "epoch": 0.72, + "learning_rate": 7.88106728555644e-08, + "loss": 0.161, + "step": 13717 + }, + { + "epoch": 0.72, + "learning_rate": 7.878377738478601e-08, + "loss": 0.1246, + "step": 13718 + }, + { + "epoch": 0.72, + "learning_rate": 7.875688537841411e-08, + "loss": 0.1475, + "step": 13719 + }, + { + "epoch": 0.72, + "learning_rate": 7.872999683721725e-08, + "loss": 0.1605, + "step": 13720 + }, + { + "epoch": 0.72, + "learning_rate": 7.870311176196389e-08, + "loss": 0.1449, + "step": 13721 + }, + { + "epoch": 0.72, + "learning_rate": 7.867623015342241e-08, + "loss": 0.1433, + "step": 13722 + }, + { + "epoch": 0.72, + "learning_rate": 7.864935201236113e-08, + "loss": 0.2009, + "step": 13723 + }, + { + "epoch": 0.72, + "learning_rate": 7.862247733954822e-08, + "loss": 0.1373, + "step": 13724 + }, + { + "epoch": 0.72, + "learning_rate": 7.859560613575167e-08, + "loss": 0.1521, + "step": 13725 + }, + { + "epoch": 0.72, + "learning_rate": 7.856873840173964e-08, + "loss": 0.1491, + "step": 13726 + }, + { + "epoch": 0.72, + "learning_rate": 7.854187413827992e-08, + "loss": 0.1209, + "step": 13727 + }, + { + "epoch": 0.72, + "learning_rate": 7.851501334614036e-08, + "loss": 0.1642, + "step": 13728 + }, + { + "epoch": 0.72, + "learning_rate": 7.848815602608854e-08, + "loss": 0.1618, + "step": 13729 + }, + { + "epoch": 0.72, + "learning_rate": 7.84613021788922e-08, + "loss": 0.1285, + "step": 13730 + }, + { + "epoch": 0.72, + "learning_rate": 7.843445180531882e-08, + "loss": 0.1667, + "step": 13731 + }, + { + "epoch": 0.72, + "learning_rate": 7.84076049061356e-08, + "loss": 0.0976, + "step": 13732 + }, + { + "epoch": 0.72, + "learning_rate": 7.838076148211003e-08, + "loss": 0.1405, + "step": 13733 + }, + { + "epoch": 0.72, + "learning_rate": 7.835392153400928e-08, + "loss": 0.1651, + "step": 13734 + }, + { + "epoch": 0.72, + "learning_rate": 7.83270850626004e-08, + "loss": 0.1271, + "step": 13735 + }, + { + "epoch": 0.72, + "learning_rate": 7.830025206865032e-08, + "loss": 0.1308, + "step": 13736 + }, + { + "epoch": 0.72, + "learning_rate": 7.827342255292609e-08, + "loss": 0.1634, + "step": 13737 + }, + { + "epoch": 0.72, + "learning_rate": 7.824659651619443e-08, + "loss": 0.1408, + "step": 13738 + }, + { + "epoch": 0.72, + "learning_rate": 7.821977395922203e-08, + "loss": 0.122, + "step": 13739 + }, + { + "epoch": 0.72, + "learning_rate": 7.819295488277554e-08, + "loss": 0.1491, + "step": 13740 + }, + { + "epoch": 0.72, + "learning_rate": 7.81661392876214e-08, + "loss": 0.1432, + "step": 13741 + }, + { + "epoch": 0.72, + "learning_rate": 7.813932717452602e-08, + "loss": 0.1293, + "step": 13742 + }, + { + "epoch": 0.72, + "learning_rate": 7.811251854425573e-08, + "loss": 0.1438, + "step": 13743 + }, + { + "epoch": 0.72, + "learning_rate": 7.808571339757664e-08, + "loss": 0.1252, + "step": 13744 + }, + { + "epoch": 0.72, + "learning_rate": 7.805891173525498e-08, + "loss": 0.117, + "step": 13745 + }, + { + "epoch": 0.72, + "learning_rate": 7.803211355805672e-08, + "loss": 0.1337, + "step": 13746 + }, + { + "epoch": 0.72, + "learning_rate": 7.800531886674772e-08, + "loss": 0.1635, + "step": 13747 + }, + { + "epoch": 0.72, + "learning_rate": 7.79785276620938e-08, + "loss": 0.1391, + "step": 13748 + }, + { + "epoch": 0.72, + "learning_rate": 7.795173994486069e-08, + "loss": 0.1487, + "step": 13749 + }, + { + "epoch": 0.72, + "learning_rate": 7.792495571581394e-08, + "loss": 0.1226, + "step": 13750 + }, + { + "epoch": 0.72, + "learning_rate": 7.789817497571905e-08, + "loss": 0.1608, + "step": 13751 + }, + { + "epoch": 0.72, + "learning_rate": 7.787139772534148e-08, + "loss": 0.1865, + "step": 13752 + }, + { + "epoch": 0.72, + "learning_rate": 7.784462396544653e-08, + "loss": 0.1251, + "step": 13753 + }, + { + "epoch": 0.72, + "learning_rate": 7.781785369679938e-08, + "loss": 0.1321, + "step": 13754 + }, + { + "epoch": 0.72, + "learning_rate": 7.779108692016508e-08, + "loss": 0.1266, + "step": 13755 + }, + { + "epoch": 0.72, + "learning_rate": 7.776432363630874e-08, + "loss": 0.1407, + "step": 13756 + }, + { + "epoch": 0.72, + "learning_rate": 7.77375638459953e-08, + "loss": 0.1245, + "step": 13757 + }, + { + "epoch": 0.72, + "learning_rate": 7.771080754998936e-08, + "loss": 0.1433, + "step": 13758 + }, + { + "epoch": 0.72, + "learning_rate": 7.768405474905579e-08, + "loss": 0.1376, + "step": 13759 + }, + { + "epoch": 0.72, + "learning_rate": 7.765730544395916e-08, + "loss": 0.1362, + "step": 13760 + }, + { + "epoch": 0.72, + "learning_rate": 7.763055963546399e-08, + "loss": 0.1325, + "step": 13761 + }, + { + "epoch": 0.72, + "learning_rate": 7.760381732433459e-08, + "loss": 0.1241, + "step": 13762 + }, + { + "epoch": 0.72, + "learning_rate": 7.757707851133539e-08, + "loss": 0.1327, + "step": 13763 + }, + { + "epoch": 0.72, + "learning_rate": 7.755034319723057e-08, + "loss": 0.146, + "step": 13764 + }, + { + "epoch": 0.72, + "learning_rate": 7.75236113827842e-08, + "loss": 0.1874, + "step": 13765 + }, + { + "epoch": 0.72, + "learning_rate": 7.74968830687603e-08, + "loss": 0.1317, + "step": 13766 + }, + { + "epoch": 0.72, + "learning_rate": 7.747015825592276e-08, + "loss": 0.1365, + "step": 13767 + }, + { + "epoch": 0.72, + "learning_rate": 7.744343694503541e-08, + "loss": 0.1365, + "step": 13768 + }, + { + "epoch": 0.72, + "learning_rate": 7.741671913686187e-08, + "loss": 0.1246, + "step": 13769 + }, + { + "epoch": 0.72, + "learning_rate": 7.739000483216589e-08, + "loss": 0.1586, + "step": 13770 + }, + { + "epoch": 0.72, + "learning_rate": 7.736329403171091e-08, + "loss": 0.1267, + "step": 13771 + }, + { + "epoch": 0.72, + "learning_rate": 7.733658673626033e-08, + "loss": 0.098, + "step": 13772 + }, + { + "epoch": 0.72, + "learning_rate": 7.730988294657736e-08, + "loss": 0.1423, + "step": 13773 + }, + { + "epoch": 0.72, + "learning_rate": 7.728318266342546e-08, + "loss": 0.1351, + "step": 13774 + }, + { + "epoch": 0.72, + "learning_rate": 7.72564858875675e-08, + "loss": 0.1546, + "step": 13775 + }, + { + "epoch": 0.72, + "learning_rate": 7.72297926197665e-08, + "loss": 0.1306, + "step": 13776 + }, + { + "epoch": 0.72, + "learning_rate": 7.720310286078549e-08, + "loss": 0.1146, + "step": 13777 + }, + { + "epoch": 0.72, + "learning_rate": 7.717641661138718e-08, + "loss": 0.138, + "step": 13778 + }, + { + "epoch": 0.72, + "learning_rate": 7.71497338723343e-08, + "loss": 0.1319, + "step": 13779 + }, + { + "epoch": 0.72, + "learning_rate": 7.71230546443894e-08, + "loss": 0.1431, + "step": 13780 + }, + { + "epoch": 0.72, + "learning_rate": 7.709637892831509e-08, + "loss": 0.1282, + "step": 13781 + }, + { + "epoch": 0.72, + "learning_rate": 7.706970672487381e-08, + "loss": 0.1407, + "step": 13782 + }, + { + "epoch": 0.72, + "learning_rate": 7.704303803482759e-08, + "loss": 0.1565, + "step": 13783 + }, + { + "epoch": 0.72, + "learning_rate": 7.701637285893889e-08, + "loss": 0.1191, + "step": 13784 + }, + { + "epoch": 0.72, + "learning_rate": 7.698971119796976e-08, + "loss": 0.1529, + "step": 13785 + }, + { + "epoch": 0.72, + "learning_rate": 7.696305305268213e-08, + "loss": 0.1613, + "step": 13786 + }, + { + "epoch": 0.72, + "learning_rate": 7.69363984238379e-08, + "loss": 0.1297, + "step": 13787 + }, + { + "epoch": 0.72, + "learning_rate": 7.690974731219897e-08, + "loss": 0.113, + "step": 13788 + }, + { + "epoch": 0.72, + "learning_rate": 7.688309971852697e-08, + "loss": 0.1584, + "step": 13789 + }, + { + "epoch": 0.72, + "learning_rate": 7.685645564358351e-08, + "loss": 0.1303, + "step": 13790 + }, + { + "epoch": 0.72, + "learning_rate": 7.682981508813011e-08, + "loss": 0.1288, + "step": 13791 + }, + { + "epoch": 0.72, + "learning_rate": 7.680317805292812e-08, + "loss": 0.1422, + "step": 13792 + }, + { + "epoch": 0.72, + "learning_rate": 7.677654453873887e-08, + "loss": 0.1481, + "step": 13793 + }, + { + "epoch": 0.72, + "learning_rate": 7.674991454632347e-08, + "loss": 0.1325, + "step": 13794 + }, + { + "epoch": 0.72, + "learning_rate": 7.672328807644317e-08, + "loss": 0.1625, + "step": 13795 + }, + { + "epoch": 0.72, + "learning_rate": 7.669666512985889e-08, + "loss": 0.1597, + "step": 13796 + }, + { + "epoch": 0.72, + "learning_rate": 7.667004570733149e-08, + "loss": 0.1205, + "step": 13797 + }, + { + "epoch": 0.72, + "learning_rate": 7.664342980962175e-08, + "loss": 0.1385, + "step": 13798 + }, + { + "epoch": 0.72, + "learning_rate": 7.661681743749047e-08, + "loss": 0.171, + "step": 13799 + }, + { + "epoch": 0.72, + "learning_rate": 7.659020859169825e-08, + "loss": 0.1217, + "step": 13800 + }, + { + "epoch": 0.72, + "learning_rate": 7.656360327300538e-08, + "loss": 0.1299, + "step": 13801 + }, + { + "epoch": 0.72, + "learning_rate": 7.653700148217244e-08, + "loss": 0.1789, + "step": 13802 + }, + { + "epoch": 0.72, + "learning_rate": 7.651040321995965e-08, + "loss": 0.1458, + "step": 13803 + }, + { + "epoch": 0.72, + "learning_rate": 7.648380848712721e-08, + "loss": 0.1097, + "step": 13804 + }, + { + "epoch": 0.72, + "learning_rate": 7.645721728443515e-08, + "loss": 0.1222, + "step": 13805 + }, + { + "epoch": 0.72, + "learning_rate": 7.643062961264357e-08, + "loss": 0.1091, + "step": 13806 + }, + { + "epoch": 0.72, + "learning_rate": 7.640404547251229e-08, + "loss": 0.1384, + "step": 13807 + }, + { + "epoch": 0.72, + "learning_rate": 7.637746486480112e-08, + "loss": 0.1367, + "step": 13808 + }, + { + "epoch": 0.72, + "learning_rate": 7.635088779026973e-08, + "loss": 0.1205, + "step": 13809 + }, + { + "epoch": 0.72, + "learning_rate": 7.63243142496777e-08, + "loss": 0.1293, + "step": 13810 + }, + { + "epoch": 0.72, + "learning_rate": 7.62977442437845e-08, + "loss": 0.1298, + "step": 13811 + }, + { + "epoch": 0.72, + "learning_rate": 7.627117777334943e-08, + "loss": 0.1249, + "step": 13812 + }, + { + "epoch": 0.72, + "learning_rate": 7.624461483913198e-08, + "loss": 0.1403, + "step": 13813 + }, + { + "epoch": 0.72, + "learning_rate": 7.621805544189119e-08, + "loss": 0.1295, + "step": 13814 + }, + { + "epoch": 0.72, + "learning_rate": 7.619149958238618e-08, + "loss": 0.1275, + "step": 13815 + }, + { + "epoch": 0.72, + "learning_rate": 7.616494726137581e-08, + "loss": 0.1491, + "step": 13816 + }, + { + "epoch": 0.72, + "learning_rate": 7.61383984796192e-08, + "loss": 0.1494, + "step": 13817 + }, + { + "epoch": 0.72, + "learning_rate": 7.61118532378749e-08, + "loss": 0.128, + "step": 13818 + }, + { + "epoch": 0.72, + "learning_rate": 7.608531153690161e-08, + "loss": 0.1245, + "step": 13819 + }, + { + "epoch": 0.72, + "learning_rate": 7.6058773377458e-08, + "loss": 0.156, + "step": 13820 + }, + { + "epoch": 0.72, + "learning_rate": 7.603223876030251e-08, + "loss": 0.138, + "step": 13821 + }, + { + "epoch": 0.72, + "learning_rate": 7.60057076861935e-08, + "loss": 0.114, + "step": 13822 + }, + { + "epoch": 0.72, + "learning_rate": 7.597918015588913e-08, + "loss": 0.1472, + "step": 13823 + }, + { + "epoch": 0.72, + "learning_rate": 7.595265617014777e-08, + "loss": 0.1331, + "step": 13824 + }, + { + "epoch": 0.72, + "learning_rate": 7.592613572972743e-08, + "loss": 0.1339, + "step": 13825 + }, + { + "epoch": 0.72, + "learning_rate": 7.589961883538592e-08, + "loss": 0.1209, + "step": 13826 + }, + { + "epoch": 0.72, + "learning_rate": 7.587310548788124e-08, + "loss": 0.1006, + "step": 13827 + }, + { + "epoch": 0.72, + "learning_rate": 7.584659568797116e-08, + "loss": 0.1581, + "step": 13828 + }, + { + "epoch": 0.72, + "learning_rate": 7.582008943641327e-08, + "loss": 0.1265, + "step": 13829 + }, + { + "epoch": 0.72, + "learning_rate": 7.579358673396512e-08, + "loss": 0.1478, + "step": 13830 + }, + { + "epoch": 0.72, + "learning_rate": 7.576708758138425e-08, + "loss": 0.1492, + "step": 13831 + }, + { + "epoch": 0.72, + "learning_rate": 7.574059197942801e-08, + "loss": 0.149, + "step": 13832 + }, + { + "epoch": 0.72, + "learning_rate": 7.57140999288536e-08, + "loss": 0.1311, + "step": 13833 + }, + { + "epoch": 0.72, + "learning_rate": 7.568761143041819e-08, + "loss": 0.1176, + "step": 13834 + }, + { + "epoch": 0.72, + "learning_rate": 7.566112648487886e-08, + "loss": 0.1407, + "step": 13835 + }, + { + "epoch": 0.72, + "learning_rate": 7.563464509299252e-08, + "loss": 0.1047, + "step": 13836 + }, + { + "epoch": 0.72, + "learning_rate": 7.560816725551593e-08, + "loss": 0.1801, + "step": 13837 + }, + { + "epoch": 0.72, + "learning_rate": 7.558169297320605e-08, + "loss": 0.1439, + "step": 13838 + }, + { + "epoch": 0.72, + "learning_rate": 7.555522224681941e-08, + "loss": 0.1304, + "step": 13839 + }, + { + "epoch": 0.72, + "learning_rate": 7.552875507711254e-08, + "loss": 0.1122, + "step": 13840 + }, + { + "epoch": 0.72, + "learning_rate": 7.550229146484185e-08, + "loss": 0.1557, + "step": 13841 + }, + { + "epoch": 0.72, + "learning_rate": 7.547583141076385e-08, + "loss": 0.1218, + "step": 13842 + }, + { + "epoch": 0.72, + "learning_rate": 7.54493749156346e-08, + "loss": 0.1121, + "step": 13843 + }, + { + "epoch": 0.72, + "learning_rate": 7.542292198021025e-08, + "loss": 0.1171, + "step": 13844 + }, + { + "epoch": 0.72, + "learning_rate": 7.539647260524693e-08, + "loss": 0.1558, + "step": 13845 + }, + { + "epoch": 0.72, + "learning_rate": 7.537002679150054e-08, + "loss": 0.115, + "step": 13846 + }, + { + "epoch": 0.72, + "learning_rate": 7.534358453972687e-08, + "loss": 0.1556, + "step": 13847 + }, + { + "epoch": 0.72, + "learning_rate": 7.531714585068163e-08, + "loss": 0.1191, + "step": 13848 + }, + { + "epoch": 0.72, + "learning_rate": 7.529071072512054e-08, + "loss": 0.1278, + "step": 13849 + }, + { + "epoch": 0.72, + "learning_rate": 7.52642791637991e-08, + "loss": 0.1404, + "step": 13850 + }, + { + "epoch": 0.72, + "learning_rate": 7.523785116747276e-08, + "loss": 0.1076, + "step": 13851 + }, + { + "epoch": 0.72, + "learning_rate": 7.521142673689669e-08, + "loss": 0.1408, + "step": 13852 + }, + { + "epoch": 0.72, + "learning_rate": 7.518500587282625e-08, + "loss": 0.144, + "step": 13853 + }, + { + "epoch": 0.72, + "learning_rate": 7.515858857601656e-08, + "loss": 0.1524, + "step": 13854 + }, + { + "epoch": 0.72, + "learning_rate": 7.513217484722259e-08, + "loss": 0.1342, + "step": 13855 + }, + { + "epoch": 0.72, + "learning_rate": 7.510576468719921e-08, + "loss": 0.1247, + "step": 13856 + }, + { + "epoch": 0.72, + "learning_rate": 7.507935809670134e-08, + "loss": 0.1172, + "step": 13857 + }, + { + "epoch": 0.72, + "learning_rate": 7.505295507648363e-08, + "loss": 0.1369, + "step": 13858 + }, + { + "epoch": 0.72, + "learning_rate": 7.502655562730072e-08, + "loss": 0.14, + "step": 13859 + }, + { + "epoch": 0.72, + "learning_rate": 7.500015974990707e-08, + "loss": 0.1025, + "step": 13860 + }, + { + "epoch": 0.72, + "learning_rate": 7.497376744505715e-08, + "loss": 0.1407, + "step": 13861 + }, + { + "epoch": 0.72, + "learning_rate": 7.494737871350519e-08, + "loss": 0.1517, + "step": 13862 + }, + { + "epoch": 0.72, + "learning_rate": 7.492099355600536e-08, + "loss": 0.1345, + "step": 13863 + }, + { + "epoch": 0.72, + "learning_rate": 7.489461197331188e-08, + "loss": 0.1147, + "step": 13864 + }, + { + "epoch": 0.72, + "learning_rate": 7.48682339661787e-08, + "loss": 0.145, + "step": 13865 + }, + { + "epoch": 0.72, + "learning_rate": 7.484185953535969e-08, + "loss": 0.1588, + "step": 13866 + }, + { + "epoch": 0.72, + "learning_rate": 7.481548868160859e-08, + "loss": 0.1502, + "step": 13867 + }, + { + "epoch": 0.72, + "learning_rate": 7.478912140567928e-08, + "loss": 0.1606, + "step": 13868 + }, + { + "epoch": 0.72, + "learning_rate": 7.476275770832516e-08, + "loss": 0.1352, + "step": 13869 + }, + { + "epoch": 0.72, + "learning_rate": 7.473639759029968e-08, + "loss": 0.1529, + "step": 13870 + }, + { + "epoch": 0.72, + "learning_rate": 7.47100410523564e-08, + "loss": 0.1053, + "step": 13871 + }, + { + "epoch": 0.72, + "learning_rate": 7.46836880952485e-08, + "loss": 0.1657, + "step": 13872 + }, + { + "epoch": 0.72, + "learning_rate": 7.465733871972919e-08, + "loss": 0.1696, + "step": 13873 + }, + { + "epoch": 0.72, + "learning_rate": 7.463099292655143e-08, + "loss": 0.1419, + "step": 13874 + }, + { + "epoch": 0.72, + "learning_rate": 7.460465071646838e-08, + "loss": 0.1316, + "step": 13875 + }, + { + "epoch": 0.72, + "learning_rate": 7.457831209023282e-08, + "loss": 0.1754, + "step": 13876 + }, + { + "epoch": 0.72, + "learning_rate": 7.455197704859752e-08, + "loss": 0.1463, + "step": 13877 + }, + { + "epoch": 0.72, + "learning_rate": 7.452564559231513e-08, + "loss": 0.126, + "step": 13878 + }, + { + "epoch": 0.72, + "learning_rate": 7.449931772213824e-08, + "loss": 0.1475, + "step": 13879 + }, + { + "epoch": 0.72, + "learning_rate": 7.447299343881927e-08, + "loss": 0.1177, + "step": 13880 + }, + { + "epoch": 0.72, + "learning_rate": 7.444667274311059e-08, + "loss": 0.1212, + "step": 13881 + }, + { + "epoch": 0.72, + "learning_rate": 7.44203556357645e-08, + "loss": 0.1297, + "step": 13882 + }, + { + "epoch": 0.72, + "learning_rate": 7.439404211753313e-08, + "loss": 0.1172, + "step": 13883 + }, + { + "epoch": 0.72, + "learning_rate": 7.436773218916852e-08, + "loss": 0.1128, + "step": 13884 + }, + { + "epoch": 0.72, + "learning_rate": 7.434142585142264e-08, + "loss": 0.1024, + "step": 13885 + }, + { + "epoch": 0.72, + "learning_rate": 7.43151231050473e-08, + "loss": 0.1276, + "step": 13886 + }, + { + "epoch": 0.72, + "learning_rate": 7.428882395079428e-08, + "loss": 0.1209, + "step": 13887 + }, + { + "epoch": 0.72, + "learning_rate": 7.426252838941512e-08, + "loss": 0.1232, + "step": 13888 + }, + { + "epoch": 0.72, + "learning_rate": 7.42362364216615e-08, + "loss": 0.1293, + "step": 13889 + }, + { + "epoch": 0.72, + "learning_rate": 7.420994804828479e-08, + "loss": 0.1442, + "step": 13890 + }, + { + "epoch": 0.73, + "learning_rate": 7.418366327003636e-08, + "loss": 0.1201, + "step": 13891 + }, + { + "epoch": 0.73, + "learning_rate": 7.41573820876673e-08, + "loss": 0.1211, + "step": 13892 + }, + { + "epoch": 0.73, + "learning_rate": 7.413110450192892e-08, + "loss": 0.1472, + "step": 13893 + }, + { + "epoch": 0.73, + "learning_rate": 7.410483051357224e-08, + "loss": 0.094, + "step": 13894 + }, + { + "epoch": 0.73, + "learning_rate": 7.407856012334797e-08, + "loss": 0.1315, + "step": 13895 + }, + { + "epoch": 0.73, + "learning_rate": 7.405229333200711e-08, + "loss": 0.139, + "step": 13896 + }, + { + "epoch": 0.73, + "learning_rate": 7.402603014030033e-08, + "loss": 0.1527, + "step": 13897 + }, + { + "epoch": 0.73, + "learning_rate": 7.399977054897825e-08, + "loss": 0.1317, + "step": 13898 + }, + { + "epoch": 0.73, + "learning_rate": 7.397351455879131e-08, + "loss": 0.1585, + "step": 13899 + }, + { + "epoch": 0.73, + "learning_rate": 7.394726217049003e-08, + "loss": 0.1129, + "step": 13900 + }, + { + "epoch": 0.73, + "learning_rate": 7.392101338482466e-08, + "loss": 0.1094, + "step": 13901 + }, + { + "epoch": 0.73, + "learning_rate": 7.389476820254538e-08, + "loss": 0.1357, + "step": 13902 + }, + { + "epoch": 0.73, + "learning_rate": 7.386852662440233e-08, + "loss": 0.1582, + "step": 13903 + }, + { + "epoch": 0.73, + "learning_rate": 7.384228865114548e-08, + "loss": 0.139, + "step": 13904 + }, + { + "epoch": 0.73, + "learning_rate": 7.38160542835247e-08, + "loss": 0.1532, + "step": 13905 + }, + { + "epoch": 0.73, + "learning_rate": 7.378982352228977e-08, + "loss": 0.1436, + "step": 13906 + }, + { + "epoch": 0.73, + "learning_rate": 7.376359636819045e-08, + "loss": 0.1168, + "step": 13907 + }, + { + "epoch": 0.73, + "learning_rate": 7.373737282197628e-08, + "loss": 0.1204, + "step": 13908 + }, + { + "epoch": 0.73, + "learning_rate": 7.371115288439673e-08, + "loss": 0.1216, + "step": 13909 + }, + { + "epoch": 0.73, + "learning_rate": 7.368493655620113e-08, + "loss": 0.1718, + "step": 13910 + }, + { + "epoch": 0.73, + "learning_rate": 7.365872383813894e-08, + "loss": 0.1418, + "step": 13911 + }, + { + "epoch": 0.73, + "learning_rate": 7.363251473095913e-08, + "loss": 0.1454, + "step": 13912 + }, + { + "epoch": 0.73, + "learning_rate": 7.360630923541076e-08, + "loss": 0.14, + "step": 13913 + }, + { + "epoch": 0.73, + "learning_rate": 7.358010735224292e-08, + "loss": 0.1232, + "step": 13914 + }, + { + "epoch": 0.73, + "learning_rate": 7.355390908220443e-08, + "loss": 0.1564, + "step": 13915 + }, + { + "epoch": 0.73, + "learning_rate": 7.352771442604402e-08, + "loss": 0.1756, + "step": 13916 + }, + { + "epoch": 0.73, + "learning_rate": 7.35015233845103e-08, + "loss": 0.1597, + "step": 13917 + }, + { + "epoch": 0.73, + "learning_rate": 7.347533595835193e-08, + "loss": 0.2191, + "step": 13918 + }, + { + "epoch": 0.73, + "learning_rate": 7.344915214831737e-08, + "loss": 0.1492, + "step": 13919 + }, + { + "epoch": 0.73, + "learning_rate": 7.342297195515477e-08, + "loss": 0.1452, + "step": 13920 + }, + { + "epoch": 0.73, + "learning_rate": 7.339679537961256e-08, + "loss": 0.1263, + "step": 13921 + }, + { + "epoch": 0.73, + "learning_rate": 7.33706224224388e-08, + "loss": 0.1269, + "step": 13922 + }, + { + "epoch": 0.73, + "learning_rate": 7.334445308438155e-08, + "loss": 0.1233, + "step": 13923 + }, + { + "epoch": 0.73, + "learning_rate": 7.331828736618865e-08, + "loss": 0.1198, + "step": 13924 + }, + { + "epoch": 0.73, + "learning_rate": 7.329212526860807e-08, + "loss": 0.141, + "step": 13925 + }, + { + "epoch": 0.73, + "learning_rate": 7.326596679238746e-08, + "loss": 0.1226, + "step": 13926 + }, + { + "epoch": 0.73, + "learning_rate": 7.323981193827444e-08, + "loss": 0.1319, + "step": 13927 + }, + { + "epoch": 0.73, + "learning_rate": 7.321366070701656e-08, + "loss": 0.0984, + "step": 13928 + }, + { + "epoch": 0.73, + "learning_rate": 7.318751309936117e-08, + "loss": 0.1004, + "step": 13929 + }, + { + "epoch": 0.73, + "learning_rate": 7.316136911605561e-08, + "loss": 0.1249, + "step": 13930 + }, + { + "epoch": 0.73, + "learning_rate": 7.313522875784703e-08, + "loss": 0.1427, + "step": 13931 + }, + { + "epoch": 0.73, + "learning_rate": 7.310909202548265e-08, + "loss": 0.1292, + "step": 13932 + }, + { + "epoch": 0.73, + "learning_rate": 7.308295891970942e-08, + "loss": 0.1185, + "step": 13933 + }, + { + "epoch": 0.73, + "learning_rate": 7.305682944127418e-08, + "loss": 0.1296, + "step": 13934 + }, + { + "epoch": 0.73, + "learning_rate": 7.303070359092372e-08, + "loss": 0.1559, + "step": 13935 + }, + { + "epoch": 0.73, + "learning_rate": 7.300458136940483e-08, + "loss": 0.1216, + "step": 13936 + }, + { + "epoch": 0.73, + "learning_rate": 7.297846277746411e-08, + "loss": 0.1585, + "step": 13937 + }, + { + "epoch": 0.73, + "learning_rate": 7.29523478158478e-08, + "loss": 0.1383, + "step": 13938 + }, + { + "epoch": 0.73, + "learning_rate": 7.292623648530252e-08, + "loss": 0.1155, + "step": 13939 + }, + { + "epoch": 0.73, + "learning_rate": 7.290012878657446e-08, + "loss": 0.1099, + "step": 13940 + }, + { + "epoch": 0.73, + "learning_rate": 7.287402472040976e-08, + "loss": 0.1531, + "step": 13941 + }, + { + "epoch": 0.73, + "learning_rate": 7.284792428755446e-08, + "loss": 0.1321, + "step": 13942 + }, + { + "epoch": 0.73, + "learning_rate": 7.282182748875463e-08, + "loss": 0.1057, + "step": 13943 + }, + { + "epoch": 0.73, + "learning_rate": 7.279573432475607e-08, + "loss": 0.1147, + "step": 13944 + }, + { + "epoch": 0.73, + "learning_rate": 7.276964479630454e-08, + "loss": 0.1293, + "step": 13945 + }, + { + "epoch": 0.73, + "learning_rate": 7.274355890414566e-08, + "loss": 0.0968, + "step": 13946 + }, + { + "epoch": 0.73, + "learning_rate": 7.271747664902499e-08, + "loss": 0.1699, + "step": 13947 + }, + { + "epoch": 0.73, + "learning_rate": 7.269139803168798e-08, + "loss": 0.1153, + "step": 13948 + }, + { + "epoch": 0.73, + "learning_rate": 7.266532305287991e-08, + "loss": 0.1569, + "step": 13949 + }, + { + "epoch": 0.73, + "learning_rate": 7.263925171334611e-08, + "loss": 0.1358, + "step": 13950 + }, + { + "epoch": 0.73, + "learning_rate": 7.261318401383167e-08, + "loss": 0.1245, + "step": 13951 + }, + { + "epoch": 0.73, + "learning_rate": 7.25871199550816e-08, + "loss": 0.1379, + "step": 13952 + }, + { + "epoch": 0.73, + "learning_rate": 7.256105953784077e-08, + "loss": 0.1281, + "step": 13953 + }, + { + "epoch": 0.73, + "learning_rate": 7.253500276285418e-08, + "loss": 0.1253, + "step": 13954 + }, + { + "epoch": 0.73, + "learning_rate": 7.250894963086636e-08, + "loss": 0.1279, + "step": 13955 + }, + { + "epoch": 0.73, + "learning_rate": 7.248290014262191e-08, + "loss": 0.1351, + "step": 13956 + }, + { + "epoch": 0.73, + "learning_rate": 7.245685429886545e-08, + "loss": 0.1557, + "step": 13957 + }, + { + "epoch": 0.73, + "learning_rate": 7.243081210034134e-08, + "loss": 0.1522, + "step": 13958 + }, + { + "epoch": 0.73, + "learning_rate": 7.240477354779386e-08, + "loss": 0.1207, + "step": 13959 + }, + { + "epoch": 0.73, + "learning_rate": 7.237873864196715e-08, + "loss": 0.1051, + "step": 13960 + }, + { + "epoch": 0.73, + "learning_rate": 7.235270738360544e-08, + "loss": 0.1165, + "step": 13961 + }, + { + "epoch": 0.73, + "learning_rate": 7.232667977345269e-08, + "loss": 0.1299, + "step": 13962 + }, + { + "epoch": 0.73, + "learning_rate": 7.230065581225262e-08, + "loss": 0.1141, + "step": 13963 + }, + { + "epoch": 0.73, + "learning_rate": 7.227463550074906e-08, + "loss": 0.1311, + "step": 13964 + }, + { + "epoch": 0.73, + "learning_rate": 7.22486188396858e-08, + "loss": 0.1069, + "step": 13965 + }, + { + "epoch": 0.73, + "learning_rate": 7.222260582980633e-08, + "loss": 0.1217, + "step": 13966 + }, + { + "epoch": 0.73, + "learning_rate": 7.21965964718541e-08, + "loss": 0.1262, + "step": 13967 + }, + { + "epoch": 0.73, + "learning_rate": 7.217059076657241e-08, + "loss": 0.1691, + "step": 13968 + }, + { + "epoch": 0.73, + "learning_rate": 7.214458871470466e-08, + "loss": 0.1034, + "step": 13969 + }, + { + "epoch": 0.73, + "learning_rate": 7.211859031699392e-08, + "loss": 0.1126, + "step": 13970 + }, + { + "epoch": 0.73, + "learning_rate": 7.209259557418324e-08, + "loss": 0.1431, + "step": 13971 + }, + { + "epoch": 0.73, + "learning_rate": 7.206660448701554e-08, + "loss": 0.1129, + "step": 13972 + }, + { + "epoch": 0.73, + "learning_rate": 7.204061705623367e-08, + "loss": 0.1528, + "step": 13973 + }, + { + "epoch": 0.73, + "learning_rate": 7.20146332825804e-08, + "loss": 0.1469, + "step": 13974 + }, + { + "epoch": 0.73, + "learning_rate": 7.198865316679821e-08, + "loss": 0.1377, + "step": 13975 + }, + { + "epoch": 0.73, + "learning_rate": 7.196267670962981e-08, + "loss": 0.1301, + "step": 13976 + }, + { + "epoch": 0.73, + "learning_rate": 7.193670391181755e-08, + "loss": 0.1537, + "step": 13977 + }, + { + "epoch": 0.73, + "learning_rate": 7.19107347741037e-08, + "loss": 0.1363, + "step": 13978 + }, + { + "epoch": 0.73, + "learning_rate": 7.188476929723053e-08, + "loss": 0.1442, + "step": 13979 + }, + { + "epoch": 0.73, + "learning_rate": 7.185880748194009e-08, + "loss": 0.1243, + "step": 13980 + }, + { + "epoch": 0.73, + "learning_rate": 7.18328493289744e-08, + "loss": 0.1395, + "step": 13981 + }, + { + "epoch": 0.73, + "learning_rate": 7.18068948390753e-08, + "loss": 0.1542, + "step": 13982 + }, + { + "epoch": 0.73, + "learning_rate": 7.178094401298469e-08, + "loss": 0.1588, + "step": 13983 + }, + { + "epoch": 0.73, + "learning_rate": 7.17549968514442e-08, + "loss": 0.1318, + "step": 13984 + }, + { + "epoch": 0.73, + "learning_rate": 7.172905335519542e-08, + "loss": 0.1439, + "step": 13985 + }, + { + "epoch": 0.73, + "learning_rate": 7.170311352497971e-08, + "loss": 0.1426, + "step": 13986 + }, + { + "epoch": 0.73, + "learning_rate": 7.167717736153864e-08, + "loss": 0.1231, + "step": 13987 + }, + { + "epoch": 0.73, + "learning_rate": 7.165124486561344e-08, + "loss": 0.1412, + "step": 13988 + }, + { + "epoch": 0.73, + "learning_rate": 7.162531603794508e-08, + "loss": 0.1222, + "step": 13989 + }, + { + "epoch": 0.73, + "learning_rate": 7.15993908792748e-08, + "loss": 0.1295, + "step": 13990 + }, + { + "epoch": 0.73, + "learning_rate": 7.15734693903435e-08, + "loss": 0.1655, + "step": 13991 + }, + { + "epoch": 0.73, + "learning_rate": 7.154755157189203e-08, + "loss": 0.1682, + "step": 13992 + }, + { + "epoch": 0.73, + "learning_rate": 7.152163742466104e-08, + "loss": 0.1366, + "step": 13993 + }, + { + "epoch": 0.73, + "learning_rate": 7.149572694939134e-08, + "loss": 0.1367, + "step": 13994 + }, + { + "epoch": 0.73, + "learning_rate": 7.146982014682337e-08, + "loss": 0.1337, + "step": 13995 + }, + { + "epoch": 0.73, + "learning_rate": 7.144391701769754e-08, + "loss": 0.1214, + "step": 13996 + }, + { + "epoch": 0.73, + "learning_rate": 7.141801756275418e-08, + "loss": 0.1325, + "step": 13997 + }, + { + "epoch": 0.73, + "learning_rate": 7.139212178273355e-08, + "loss": 0.1292, + "step": 13998 + }, + { + "epoch": 0.73, + "learning_rate": 7.13662296783757e-08, + "loss": 0.1187, + "step": 13999 + }, + { + "epoch": 0.73, + "learning_rate": 7.134034125042061e-08, + "loss": 0.1224, + "step": 14000 + }, + { + "epoch": 0.73, + "learning_rate": 7.13144564996083e-08, + "loss": 0.1195, + "step": 14001 + }, + { + "epoch": 0.73, + "learning_rate": 7.12885754266785e-08, + "loss": 0.1121, + "step": 14002 + }, + { + "epoch": 0.73, + "learning_rate": 7.126269803237091e-08, + "loss": 0.0976, + "step": 14003 + }, + { + "epoch": 0.73, + "learning_rate": 7.123682431742502e-08, + "loss": 0.1414, + "step": 14004 + }, + { + "epoch": 0.73, + "learning_rate": 7.121095428258055e-08, + "loss": 0.1403, + "step": 14005 + }, + { + "epoch": 0.73, + "learning_rate": 7.118508792857665e-08, + "loss": 0.16, + "step": 14006 + }, + { + "epoch": 0.73, + "learning_rate": 7.115922525615259e-08, + "loss": 0.1322, + "step": 14007 + }, + { + "epoch": 0.73, + "learning_rate": 7.113336626604769e-08, + "loss": 0.1684, + "step": 14008 + }, + { + "epoch": 0.73, + "learning_rate": 7.110751095900094e-08, + "loss": 0.1294, + "step": 14009 + }, + { + "epoch": 0.73, + "learning_rate": 7.108165933575126e-08, + "loss": 0.12, + "step": 14010 + }, + { + "epoch": 0.73, + "learning_rate": 7.105581139703747e-08, + "loss": 0.1253, + "step": 14011 + }, + { + "epoch": 0.73, + "learning_rate": 7.102996714359843e-08, + "loss": 0.1392, + "step": 14012 + }, + { + "epoch": 0.73, + "learning_rate": 7.100412657617272e-08, + "loss": 0.1433, + "step": 14013 + }, + { + "epoch": 0.73, + "learning_rate": 7.097828969549886e-08, + "loss": 0.1272, + "step": 14014 + }, + { + "epoch": 0.73, + "learning_rate": 7.09524565023153e-08, + "loss": 0.136, + "step": 14015 + }, + { + "epoch": 0.73, + "learning_rate": 7.092662699736033e-08, + "loss": 0.1701, + "step": 14016 + }, + { + "epoch": 0.73, + "learning_rate": 7.090080118137223e-08, + "loss": 0.1434, + "step": 14017 + }, + { + "epoch": 0.73, + "learning_rate": 7.087497905508897e-08, + "loss": 0.136, + "step": 14018 + }, + { + "epoch": 0.73, + "learning_rate": 7.084916061924873e-08, + "loss": 0.1025, + "step": 14019 + }, + { + "epoch": 0.73, + "learning_rate": 7.082334587458933e-08, + "loss": 0.1116, + "step": 14020 + }, + { + "epoch": 0.73, + "learning_rate": 7.079753482184858e-08, + "loss": 0.1784, + "step": 14021 + }, + { + "epoch": 0.73, + "learning_rate": 7.077172746176416e-08, + "loss": 0.1539, + "step": 14022 + }, + { + "epoch": 0.73, + "learning_rate": 7.074592379507365e-08, + "loss": 0.1885, + "step": 14023 + }, + { + "epoch": 0.73, + "learning_rate": 7.072012382251453e-08, + "loss": 0.0966, + "step": 14024 + }, + { + "epoch": 0.73, + "learning_rate": 7.069432754482414e-08, + "loss": 0.1496, + "step": 14025 + }, + { + "epoch": 0.73, + "learning_rate": 7.066853496273983e-08, + "loss": 0.1375, + "step": 14026 + }, + { + "epoch": 0.73, + "learning_rate": 7.064274607699872e-08, + "loss": 0.1281, + "step": 14027 + }, + { + "epoch": 0.73, + "learning_rate": 7.061696088833786e-08, + "loss": 0.1169, + "step": 14028 + }, + { + "epoch": 0.73, + "learning_rate": 7.059117939749415e-08, + "loss": 0.1319, + "step": 14029 + }, + { + "epoch": 0.73, + "learning_rate": 7.056540160520455e-08, + "loss": 0.1258, + "step": 14030 + }, + { + "epoch": 0.73, + "learning_rate": 7.053962751220582e-08, + "loss": 0.1335, + "step": 14031 + }, + { + "epoch": 0.73, + "learning_rate": 7.051385711923437e-08, + "loss": 0.111, + "step": 14032 + }, + { + "epoch": 0.73, + "learning_rate": 7.048809042702692e-08, + "loss": 0.1258, + "step": 14033 + }, + { + "epoch": 0.73, + "learning_rate": 7.046232743631986e-08, + "loss": 0.1398, + "step": 14034 + }, + { + "epoch": 0.73, + "learning_rate": 7.043656814784949e-08, + "loss": 0.1228, + "step": 14035 + }, + { + "epoch": 0.73, + "learning_rate": 7.041081256235192e-08, + "loss": 0.1131, + "step": 14036 + }, + { + "epoch": 0.73, + "learning_rate": 7.038506068056347e-08, + "loss": 0.1285, + "step": 14037 + }, + { + "epoch": 0.73, + "learning_rate": 7.035931250321998e-08, + "loss": 0.1079, + "step": 14038 + }, + { + "epoch": 0.73, + "learning_rate": 7.033356803105741e-08, + "loss": 0.1251, + "step": 14039 + }, + { + "epoch": 0.73, + "learning_rate": 7.030782726481153e-08, + "loss": 0.1354, + "step": 14040 + }, + { + "epoch": 0.73, + "learning_rate": 7.028209020521798e-08, + "loss": 0.1135, + "step": 14041 + }, + { + "epoch": 0.73, + "learning_rate": 7.025635685301239e-08, + "loss": 0.1442, + "step": 14042 + }, + { + "epoch": 0.73, + "learning_rate": 7.023062720893014e-08, + "loss": 0.1303, + "step": 14043 + }, + { + "epoch": 0.73, + "learning_rate": 7.020490127370673e-08, + "loss": 0.1308, + "step": 14044 + }, + { + "epoch": 0.73, + "learning_rate": 7.017917904807734e-08, + "loss": 0.1278, + "step": 14045 + }, + { + "epoch": 0.73, + "learning_rate": 7.015346053277711e-08, + "loss": 0.1277, + "step": 14046 + }, + { + "epoch": 0.73, + "learning_rate": 7.012774572854105e-08, + "loss": 0.1382, + "step": 14047 + }, + { + "epoch": 0.73, + "learning_rate": 7.010203463610432e-08, + "loss": 0.1379, + "step": 14048 + }, + { + "epoch": 0.73, + "learning_rate": 7.007632725620147e-08, + "loss": 0.1123, + "step": 14049 + }, + { + "epoch": 0.73, + "learning_rate": 7.005062358956731e-08, + "loss": 0.1069, + "step": 14050 + }, + { + "epoch": 0.73, + "learning_rate": 7.002492363693655e-08, + "loss": 0.1296, + "step": 14051 + }, + { + "epoch": 0.73, + "learning_rate": 6.999922739904365e-08, + "loss": 0.1614, + "step": 14052 + }, + { + "epoch": 0.73, + "learning_rate": 6.9973534876623e-08, + "loss": 0.1309, + "step": 14053 + }, + { + "epoch": 0.73, + "learning_rate": 6.994784607040887e-08, + "loss": 0.0923, + "step": 14054 + }, + { + "epoch": 0.73, + "learning_rate": 6.992216098113557e-08, + "loss": 0.1118, + "step": 14055 + }, + { + "epoch": 0.73, + "learning_rate": 6.989647960953711e-08, + "loss": 0.1243, + "step": 14056 + }, + { + "epoch": 0.73, + "learning_rate": 6.98708019563475e-08, + "loss": 0.1343, + "step": 14057 + }, + { + "epoch": 0.73, + "learning_rate": 6.98451280223006e-08, + "loss": 0.1435, + "step": 14058 + }, + { + "epoch": 0.73, + "learning_rate": 6.98194578081302e-08, + "loss": 0.1119, + "step": 14059 + }, + { + "epoch": 0.73, + "learning_rate": 6.979379131456992e-08, + "loss": 0.1259, + "step": 14060 + }, + { + "epoch": 0.73, + "learning_rate": 6.97681285423533e-08, + "loss": 0.1253, + "step": 14061 + }, + { + "epoch": 0.73, + "learning_rate": 6.974246949221391e-08, + "loss": 0.1284, + "step": 14062 + }, + { + "epoch": 0.73, + "learning_rate": 6.971681416488504e-08, + "loss": 0.1245, + "step": 14063 + }, + { + "epoch": 0.73, + "learning_rate": 6.969116256109988e-08, + "loss": 0.1297, + "step": 14064 + }, + { + "epoch": 0.73, + "learning_rate": 6.966551468159163e-08, + "loss": 0.157, + "step": 14065 + }, + { + "epoch": 0.73, + "learning_rate": 6.963987052709328e-08, + "loss": 0.168, + "step": 14066 + }, + { + "epoch": 0.73, + "learning_rate": 6.961423009833774e-08, + "loss": 0.1466, + "step": 14067 + }, + { + "epoch": 0.73, + "learning_rate": 6.958859339605778e-08, + "loss": 0.1303, + "step": 14068 + }, + { + "epoch": 0.73, + "learning_rate": 6.956296042098624e-08, + "loss": 0.1228, + "step": 14069 + }, + { + "epoch": 0.73, + "learning_rate": 6.953733117385564e-08, + "loss": 0.1307, + "step": 14070 + }, + { + "epoch": 0.73, + "learning_rate": 6.951170565539846e-08, + "loss": 0.1552, + "step": 14071 + }, + { + "epoch": 0.73, + "learning_rate": 6.948608386634705e-08, + "loss": 0.1308, + "step": 14072 + }, + { + "epoch": 0.73, + "learning_rate": 6.946046580743381e-08, + "loss": 0.1294, + "step": 14073 + }, + { + "epoch": 0.73, + "learning_rate": 6.943485147939092e-08, + "loss": 0.1192, + "step": 14074 + }, + { + "epoch": 0.73, + "learning_rate": 6.940924088295033e-08, + "loss": 0.1384, + "step": 14075 + }, + { + "epoch": 0.73, + "learning_rate": 6.938363401884395e-08, + "loss": 0.1784, + "step": 14076 + }, + { + "epoch": 0.73, + "learning_rate": 6.935803088780383e-08, + "loss": 0.1166, + "step": 14077 + }, + { + "epoch": 0.73, + "learning_rate": 6.933243149056161e-08, + "loss": 0.123, + "step": 14078 + }, + { + "epoch": 0.73, + "learning_rate": 6.930683582784893e-08, + "loss": 0.1283, + "step": 14079 + }, + { + "epoch": 0.73, + "learning_rate": 6.928124390039727e-08, + "loss": 0.1271, + "step": 14080 + }, + { + "epoch": 0.73, + "learning_rate": 6.925565570893819e-08, + "loss": 0.1364, + "step": 14081 + }, + { + "epoch": 0.74, + "learning_rate": 6.923007125420302e-08, + "loss": 0.1673, + "step": 14082 + }, + { + "epoch": 0.74, + "learning_rate": 6.920449053692274e-08, + "loss": 0.1038, + "step": 14083 + }, + { + "epoch": 0.74, + "learning_rate": 6.91789135578287e-08, + "loss": 0.1673, + "step": 14084 + }, + { + "epoch": 0.74, + "learning_rate": 6.915334031765181e-08, + "loss": 0.1128, + "step": 14085 + }, + { + "epoch": 0.74, + "learning_rate": 6.912777081712295e-08, + "loss": 0.1579, + "step": 14086 + }, + { + "epoch": 0.74, + "learning_rate": 6.910220505697289e-08, + "loss": 0.1343, + "step": 14087 + }, + { + "epoch": 0.74, + "learning_rate": 6.907664303793236e-08, + "loss": 0.1159, + "step": 14088 + }, + { + "epoch": 0.74, + "learning_rate": 6.905108476073196e-08, + "loss": 0.141, + "step": 14089 + }, + { + "epoch": 0.74, + "learning_rate": 6.902553022610207e-08, + "loss": 0.1393, + "step": 14090 + }, + { + "epoch": 0.74, + "learning_rate": 6.899997943477312e-08, + "loss": 0.1044, + "step": 14091 + }, + { + "epoch": 0.74, + "learning_rate": 6.897443238747531e-08, + "loss": 0.1267, + "step": 14092 + }, + { + "epoch": 0.74, + "learning_rate": 6.894888908493881e-08, + "loss": 0.1287, + "step": 14093 + }, + { + "epoch": 0.74, + "learning_rate": 6.892334952789359e-08, + "loss": 0.1126, + "step": 14094 + }, + { + "epoch": 0.74, + "learning_rate": 6.889781371706969e-08, + "loss": 0.1155, + "step": 14095 + }, + { + "epoch": 0.74, + "learning_rate": 6.887228165319689e-08, + "loss": 0.1129, + "step": 14096 + }, + { + "epoch": 0.74, + "learning_rate": 6.88467533370049e-08, + "loss": 0.0989, + "step": 14097 + }, + { + "epoch": 0.74, + "learning_rate": 6.882122876922327e-08, + "loss": 0.1577, + "step": 14098 + }, + { + "epoch": 0.74, + "learning_rate": 6.879570795058169e-08, + "loss": 0.1928, + "step": 14099 + }, + { + "epoch": 0.74, + "learning_rate": 6.877019088180935e-08, + "loss": 0.2002, + "step": 14100 + }, + { + "epoch": 0.74, + "learning_rate": 6.874467756363558e-08, + "loss": 0.1599, + "step": 14101 + }, + { + "epoch": 0.74, + "learning_rate": 6.871916799678963e-08, + "loss": 0.1377, + "step": 14102 + }, + { + "epoch": 0.74, + "learning_rate": 6.869366218200055e-08, + "loss": 0.1364, + "step": 14103 + }, + { + "epoch": 0.74, + "learning_rate": 6.866816011999729e-08, + "loss": 0.1644, + "step": 14104 + }, + { + "epoch": 0.74, + "learning_rate": 6.864266181150867e-08, + "loss": 0.1198, + "step": 14105 + }, + { + "epoch": 0.74, + "learning_rate": 6.861716725726352e-08, + "loss": 0.0933, + "step": 14106 + }, + { + "epoch": 0.74, + "learning_rate": 6.859167645799046e-08, + "loss": 0.1396, + "step": 14107 + }, + { + "epoch": 0.74, + "learning_rate": 6.856618941441801e-08, + "loss": 0.1225, + "step": 14108 + }, + { + "epoch": 0.74, + "learning_rate": 6.854070612727461e-08, + "loss": 0.1111, + "step": 14109 + }, + { + "epoch": 0.74, + "learning_rate": 6.851522659728859e-08, + "loss": 0.1345, + "step": 14110 + }, + { + "epoch": 0.74, + "learning_rate": 6.848975082518813e-08, + "loss": 0.1363, + "step": 14111 + }, + { + "epoch": 0.74, + "learning_rate": 6.84642788117013e-08, + "loss": 0.1339, + "step": 14112 + }, + { + "epoch": 0.74, + "learning_rate": 6.843881055755623e-08, + "loss": 0.1362, + "step": 14113 + }, + { + "epoch": 0.74, + "learning_rate": 6.841334606348074e-08, + "loss": 0.1446, + "step": 14114 + }, + { + "epoch": 0.74, + "learning_rate": 6.838788533020261e-08, + "loss": 0.1083, + "step": 14115 + }, + { + "epoch": 0.74, + "learning_rate": 6.836242835844947e-08, + "loss": 0.1377, + "step": 14116 + }, + { + "epoch": 0.74, + "learning_rate": 6.833697514894908e-08, + "loss": 0.1076, + "step": 14117 + }, + { + "epoch": 0.74, + "learning_rate": 6.831152570242868e-08, + "loss": 0.1304, + "step": 14118 + }, + { + "epoch": 0.74, + "learning_rate": 6.828608001961566e-08, + "loss": 0.1447, + "step": 14119 + }, + { + "epoch": 0.74, + "learning_rate": 6.82606381012374e-08, + "loss": 0.1082, + "step": 14120 + }, + { + "epoch": 0.74, + "learning_rate": 6.823519994802094e-08, + "loss": 0.1208, + "step": 14121 + }, + { + "epoch": 0.74, + "learning_rate": 6.820976556069336e-08, + "loss": 0.1184, + "step": 14122 + }, + { + "epoch": 0.74, + "learning_rate": 6.818433493998146e-08, + "loss": 0.1469, + "step": 14123 + }, + { + "epoch": 0.74, + "learning_rate": 6.815890808661222e-08, + "loss": 0.1243, + "step": 14124 + }, + { + "epoch": 0.74, + "learning_rate": 6.813348500131235e-08, + "loss": 0.1037, + "step": 14125 + }, + { + "epoch": 0.74, + "learning_rate": 6.810806568480827e-08, + "loss": 0.1476, + "step": 14126 + }, + { + "epoch": 0.74, + "learning_rate": 6.808265013782664e-08, + "loss": 0.1046, + "step": 14127 + }, + { + "epoch": 0.74, + "learning_rate": 6.80572383610938e-08, + "loss": 0.1621, + "step": 14128 + }, + { + "epoch": 0.74, + "learning_rate": 6.803183035533601e-08, + "loss": 0.1426, + "step": 14129 + }, + { + "epoch": 0.74, + "learning_rate": 6.80064261212794e-08, + "loss": 0.1139, + "step": 14130 + }, + { + "epoch": 0.74, + "learning_rate": 6.798102565965016e-08, + "loss": 0.1682, + "step": 14131 + }, + { + "epoch": 0.74, + "learning_rate": 6.795562897117415e-08, + "loss": 0.1441, + "step": 14132 + }, + { + "epoch": 0.74, + "learning_rate": 6.793023605657724e-08, + "loss": 0.1433, + "step": 14133 + }, + { + "epoch": 0.74, + "learning_rate": 6.790484691658517e-08, + "loss": 0.1119, + "step": 14134 + }, + { + "epoch": 0.74, + "learning_rate": 6.787946155192358e-08, + "loss": 0.1346, + "step": 14135 + }, + { + "epoch": 0.74, + "learning_rate": 6.785407996331797e-08, + "loss": 0.156, + "step": 14136 + }, + { + "epoch": 0.74, + "learning_rate": 6.782870215149372e-08, + "loss": 0.1501, + "step": 14137 + }, + { + "epoch": 0.74, + "learning_rate": 6.780332811717624e-08, + "loss": 0.1616, + "step": 14138 + }, + { + "epoch": 0.74, + "learning_rate": 6.777795786109068e-08, + "loss": 0.1078, + "step": 14139 + }, + { + "epoch": 0.74, + "learning_rate": 6.775259138396212e-08, + "loss": 0.1816, + "step": 14140 + }, + { + "epoch": 0.74, + "learning_rate": 6.772722868651551e-08, + "loss": 0.1157, + "step": 14141 + }, + { + "epoch": 0.74, + "learning_rate": 6.770186976947588e-08, + "loss": 0.1503, + "step": 14142 + }, + { + "epoch": 0.74, + "learning_rate": 6.767651463356781e-08, + "loss": 0.1269, + "step": 14143 + }, + { + "epoch": 0.74, + "learning_rate": 6.765116327951596e-08, + "loss": 0.1396, + "step": 14144 + }, + { + "epoch": 0.74, + "learning_rate": 6.762581570804505e-08, + "loss": 0.1577, + "step": 14145 + }, + { + "epoch": 0.74, + "learning_rate": 6.76004719198794e-08, + "loss": 0.1023, + "step": 14146 + }, + { + "epoch": 0.74, + "learning_rate": 6.757513191574338e-08, + "loss": 0.1425, + "step": 14147 + }, + { + "epoch": 0.74, + "learning_rate": 6.754979569636117e-08, + "loss": 0.1308, + "step": 14148 + }, + { + "epoch": 0.74, + "learning_rate": 6.752446326245697e-08, + "loss": 0.1349, + "step": 14149 + }, + { + "epoch": 0.74, + "learning_rate": 6.749913461475477e-08, + "loss": 0.1653, + "step": 14150 + }, + { + "epoch": 0.74, + "learning_rate": 6.747380975397843e-08, + "loss": 0.1234, + "step": 14151 + }, + { + "epoch": 0.74, + "learning_rate": 6.744848868085177e-08, + "loss": 0.1194, + "step": 14152 + }, + { + "epoch": 0.74, + "learning_rate": 6.74231713960985e-08, + "loss": 0.1222, + "step": 14153 + }, + { + "epoch": 0.74, + "learning_rate": 6.739785790044215e-08, + "loss": 0.1283, + "step": 14154 + }, + { + "epoch": 0.74, + "learning_rate": 6.737254819460614e-08, + "loss": 0.104, + "step": 14155 + }, + { + "epoch": 0.74, + "learning_rate": 6.734724227931399e-08, + "loss": 0.1147, + "step": 14156 + }, + { + "epoch": 0.74, + "learning_rate": 6.732194015528885e-08, + "loss": 0.1951, + "step": 14157 + }, + { + "epoch": 0.74, + "learning_rate": 6.729664182325388e-08, + "loss": 0.128, + "step": 14158 + }, + { + "epoch": 0.74, + "learning_rate": 6.727134728393211e-08, + "loss": 0.1391, + "step": 14159 + }, + { + "epoch": 0.74, + "learning_rate": 6.724605653804647e-08, + "loss": 0.1046, + "step": 14160 + }, + { + "epoch": 0.74, + "learning_rate": 6.722076958631979e-08, + "loss": 0.1033, + "step": 14161 + }, + { + "epoch": 0.74, + "learning_rate": 6.719548642947472e-08, + "loss": 0.1107, + "step": 14162 + }, + { + "epoch": 0.74, + "learning_rate": 6.717020706823398e-08, + "loss": 0.1256, + "step": 14163 + }, + { + "epoch": 0.74, + "learning_rate": 6.714493150331997e-08, + "loss": 0.1169, + "step": 14164 + }, + { + "epoch": 0.74, + "learning_rate": 6.711965973545513e-08, + "loss": 0.1489, + "step": 14165 + }, + { + "epoch": 0.74, + "learning_rate": 6.709439176536164e-08, + "loss": 0.1632, + "step": 14166 + }, + { + "epoch": 0.74, + "learning_rate": 6.70691275937618e-08, + "loss": 0.1182, + "step": 14167 + }, + { + "epoch": 0.74, + "learning_rate": 6.70438672213777e-08, + "loss": 0.1374, + "step": 14168 + }, + { + "epoch": 0.74, + "learning_rate": 6.701861064893104e-08, + "loss": 0.1291, + "step": 14169 + }, + { + "epoch": 0.74, + "learning_rate": 6.69933578771439e-08, + "loss": 0.1621, + "step": 14170 + }, + { + "epoch": 0.74, + "learning_rate": 6.696810890673795e-08, + "loss": 0.1537, + "step": 14171 + }, + { + "epoch": 0.74, + "learning_rate": 6.694286373843479e-08, + "loss": 0.1283, + "step": 14172 + }, + { + "epoch": 0.74, + "learning_rate": 6.691762237295588e-08, + "loss": 0.1812, + "step": 14173 + }, + { + "epoch": 0.74, + "learning_rate": 6.689238481102278e-08, + "loss": 0.1321, + "step": 14174 + }, + { + "epoch": 0.74, + "learning_rate": 6.686715105335667e-08, + "loss": 0.1662, + "step": 14175 + }, + { + "epoch": 0.74, + "learning_rate": 6.684192110067888e-08, + "loss": 0.162, + "step": 14176 + }, + { + "epoch": 0.74, + "learning_rate": 6.681669495371025e-08, + "loss": 0.1301, + "step": 14177 + }, + { + "epoch": 0.74, + "learning_rate": 6.679147261317195e-08, + "loss": 0.1292, + "step": 14178 + }, + { + "epoch": 0.74, + "learning_rate": 6.676625407978478e-08, + "loss": 0.1089, + "step": 14179 + }, + { + "epoch": 0.74, + "learning_rate": 6.674103935426952e-08, + "loss": 0.1363, + "step": 14180 + }, + { + "epoch": 0.74, + "learning_rate": 6.671582843734673e-08, + "loss": 0.1846, + "step": 14181 + }, + { + "epoch": 0.74, + "learning_rate": 6.669062132973706e-08, + "loss": 0.1575, + "step": 14182 + }, + { + "epoch": 0.74, + "learning_rate": 6.666541803216093e-08, + "loss": 0.1337, + "step": 14183 + }, + { + "epoch": 0.74, + "learning_rate": 6.664021854533853e-08, + "loss": 0.1129, + "step": 14184 + }, + { + "epoch": 0.74, + "learning_rate": 6.66150228699903e-08, + "loss": 0.1398, + "step": 14185 + }, + { + "epoch": 0.74, + "learning_rate": 6.658983100683618e-08, + "loss": 0.1378, + "step": 14186 + }, + { + "epoch": 0.74, + "learning_rate": 6.656464295659614e-08, + "loss": 0.1855, + "step": 14187 + }, + { + "epoch": 0.74, + "learning_rate": 6.653945871999007e-08, + "loss": 0.1345, + "step": 14188 + }, + { + "epoch": 0.74, + "learning_rate": 6.651427829773786e-08, + "loss": 0.1289, + "step": 14189 + }, + { + "epoch": 0.74, + "learning_rate": 6.64891016905591e-08, + "loss": 0.1558, + "step": 14190 + }, + { + "epoch": 0.74, + "learning_rate": 6.646392889917337e-08, + "loss": 0.1588, + "step": 14191 + }, + { + "epoch": 0.74, + "learning_rate": 6.643875992430001e-08, + "loss": 0.1575, + "step": 14192 + }, + { + "epoch": 0.74, + "learning_rate": 6.641359476665856e-08, + "loss": 0.132, + "step": 14193 + }, + { + "epoch": 0.74, + "learning_rate": 6.638843342696815e-08, + "loss": 0.1226, + "step": 14194 + }, + { + "epoch": 0.74, + "learning_rate": 6.636327590594781e-08, + "loss": 0.117, + "step": 14195 + }, + { + "epoch": 0.74, + "learning_rate": 6.633812220431667e-08, + "loss": 0.1283, + "step": 14196 + }, + { + "epoch": 0.74, + "learning_rate": 6.63129723227936e-08, + "loss": 0.1473, + "step": 14197 + }, + { + "epoch": 0.74, + "learning_rate": 6.628782626209741e-08, + "loss": 0.1518, + "step": 14198 + }, + { + "epoch": 0.74, + "learning_rate": 6.626268402294669e-08, + "loss": 0.1366, + "step": 14199 + }, + { + "epoch": 0.74, + "learning_rate": 6.623754560606014e-08, + "loss": 0.167, + "step": 14200 + }, + { + "epoch": 0.74, + "learning_rate": 6.621241101215617e-08, + "loss": 0.1397, + "step": 14201 + }, + { + "epoch": 0.74, + "learning_rate": 6.618728024195315e-08, + "loss": 0.1357, + "step": 14202 + }, + { + "epoch": 0.74, + "learning_rate": 6.616215329616932e-08, + "loss": 0.1549, + "step": 14203 + }, + { + "epoch": 0.74, + "learning_rate": 6.613703017552281e-08, + "loss": 0.1017, + "step": 14204 + }, + { + "epoch": 0.74, + "learning_rate": 6.611191088073167e-08, + "loss": 0.1361, + "step": 14205 + }, + { + "epoch": 0.74, + "learning_rate": 6.608679541251371e-08, + "loss": 0.1223, + "step": 14206 + }, + { + "epoch": 0.74, + "learning_rate": 6.606168377158692e-08, + "loss": 0.2054, + "step": 14207 + }, + { + "epoch": 0.74, + "learning_rate": 6.603657595866892e-08, + "loss": 0.1198, + "step": 14208 + }, + { + "epoch": 0.74, + "learning_rate": 6.601147197447727e-08, + "loss": 0.1186, + "step": 14209 + }, + { + "epoch": 0.74, + "learning_rate": 6.598637181972941e-08, + "loss": 0.1057, + "step": 14210 + }, + { + "epoch": 0.74, + "learning_rate": 6.596127549514294e-08, + "loss": 0.1403, + "step": 14211 + }, + { + "epoch": 0.74, + "learning_rate": 6.593618300143486e-08, + "loss": 0.1529, + "step": 14212 + }, + { + "epoch": 0.74, + "learning_rate": 6.591109433932238e-08, + "loss": 0.199, + "step": 14213 + }, + { + "epoch": 0.74, + "learning_rate": 6.588600950952266e-08, + "loss": 0.1567, + "step": 14214 + }, + { + "epoch": 0.74, + "learning_rate": 6.586092851275253e-08, + "loss": 0.1342, + "step": 14215 + }, + { + "epoch": 0.74, + "learning_rate": 6.583585134972886e-08, + "loss": 0.1137, + "step": 14216 + }, + { + "epoch": 0.74, + "learning_rate": 6.58107780211683e-08, + "loss": 0.1192, + "step": 14217 + }, + { + "epoch": 0.74, + "learning_rate": 6.578570852778756e-08, + "loss": 0.1399, + "step": 14218 + }, + { + "epoch": 0.74, + "learning_rate": 6.576064287030314e-08, + "loss": 0.1301, + "step": 14219 + }, + { + "epoch": 0.74, + "learning_rate": 6.573558104943124e-08, + "loss": 0.1488, + "step": 14220 + }, + { + "epoch": 0.74, + "learning_rate": 6.571052306588835e-08, + "loss": 0.1403, + "step": 14221 + }, + { + "epoch": 0.74, + "learning_rate": 6.568546892039053e-08, + "loss": 0.1273, + "step": 14222 + }, + { + "epoch": 0.74, + "learning_rate": 6.566041861365388e-08, + "loss": 0.1476, + "step": 14223 + }, + { + "epoch": 0.74, + "learning_rate": 6.563537214639426e-08, + "loss": 0.1528, + "step": 14224 + }, + { + "epoch": 0.74, + "learning_rate": 6.561032951932762e-08, + "loss": 0.1108, + "step": 14225 + }, + { + "epoch": 0.74, + "learning_rate": 6.558529073316962e-08, + "loss": 0.1225, + "step": 14226 + }, + { + "epoch": 0.74, + "learning_rate": 6.556025578863595e-08, + "loss": 0.1406, + "step": 14227 + }, + { + "epoch": 0.74, + "learning_rate": 6.553522468644204e-08, + "loss": 0.1233, + "step": 14228 + }, + { + "epoch": 0.74, + "learning_rate": 6.551019742730333e-08, + "loss": 0.1238, + "step": 14229 + }, + { + "epoch": 0.74, + "learning_rate": 6.54851740119351e-08, + "loss": 0.138, + "step": 14230 + }, + { + "epoch": 0.74, + "learning_rate": 6.546015444105245e-08, + "loss": 0.1253, + "step": 14231 + }, + { + "epoch": 0.74, + "learning_rate": 6.543513871537058e-08, + "loss": 0.1399, + "step": 14232 + }, + { + "epoch": 0.74, + "learning_rate": 6.54101268356044e-08, + "loss": 0.1132, + "step": 14233 + }, + { + "epoch": 0.74, + "learning_rate": 6.538511880246875e-08, + "loss": 0.1253, + "step": 14234 + }, + { + "epoch": 0.74, + "learning_rate": 6.536011461667831e-08, + "loss": 0.1568, + "step": 14235 + }, + { + "epoch": 0.74, + "learning_rate": 6.533511427894791e-08, + "loss": 0.1313, + "step": 14236 + }, + { + "epoch": 0.74, + "learning_rate": 6.531011778999185e-08, + "loss": 0.1537, + "step": 14237 + }, + { + "epoch": 0.74, + "learning_rate": 6.528512515052455e-08, + "loss": 0.1181, + "step": 14238 + }, + { + "epoch": 0.74, + "learning_rate": 6.526013636126046e-08, + "loss": 0.1299, + "step": 14239 + }, + { + "epoch": 0.74, + "learning_rate": 6.52351514229137e-08, + "loss": 0.1255, + "step": 14240 + }, + { + "epoch": 0.74, + "learning_rate": 6.521017033619832e-08, + "loss": 0.142, + "step": 14241 + }, + { + "epoch": 0.74, + "learning_rate": 6.518519310182822e-08, + "loss": 0.151, + "step": 14242 + }, + { + "epoch": 0.74, + "learning_rate": 6.516021972051742e-08, + "loss": 0.1416, + "step": 14243 + }, + { + "epoch": 0.74, + "learning_rate": 6.513525019297961e-08, + "loss": 0.1572, + "step": 14244 + }, + { + "epoch": 0.74, + "learning_rate": 6.511028451992839e-08, + "loss": 0.1297, + "step": 14245 + }, + { + "epoch": 0.74, + "learning_rate": 6.508532270207733e-08, + "loss": 0.1286, + "step": 14246 + }, + { + "epoch": 0.74, + "learning_rate": 6.506036474013982e-08, + "loss": 0.1394, + "step": 14247 + }, + { + "epoch": 0.74, + "learning_rate": 6.50354106348292e-08, + "loss": 0.1272, + "step": 14248 + }, + { + "epoch": 0.74, + "learning_rate": 6.501046038685855e-08, + "loss": 0.1353, + "step": 14249 + }, + { + "epoch": 0.74, + "learning_rate": 6.498551399694115e-08, + "loss": 0.1388, + "step": 14250 + }, + { + "epoch": 0.74, + "learning_rate": 6.496057146578986e-08, + "loss": 0.1462, + "step": 14251 + }, + { + "epoch": 0.74, + "learning_rate": 6.493563279411755e-08, + "loss": 0.1004, + "step": 14252 + }, + { + "epoch": 0.74, + "learning_rate": 6.491069798263696e-08, + "loss": 0.155, + "step": 14253 + }, + { + "epoch": 0.74, + "learning_rate": 6.48857670320609e-08, + "loss": 0.1415, + "step": 14254 + }, + { + "epoch": 0.74, + "learning_rate": 6.48608399431017e-08, + "loss": 0.131, + "step": 14255 + }, + { + "epoch": 0.74, + "learning_rate": 6.483591671647179e-08, + "loss": 0.1002, + "step": 14256 + }, + { + "epoch": 0.74, + "learning_rate": 6.481099735288364e-08, + "loss": 0.1175, + "step": 14257 + }, + { + "epoch": 0.74, + "learning_rate": 6.478608185304937e-08, + "loss": 0.1179, + "step": 14258 + }, + { + "epoch": 0.74, + "learning_rate": 6.476117021768108e-08, + "loss": 0.1362, + "step": 14259 + }, + { + "epoch": 0.74, + "learning_rate": 6.473626244749068e-08, + "loss": 0.1747, + "step": 14260 + }, + { + "epoch": 0.74, + "learning_rate": 6.471135854319019e-08, + "loss": 0.1005, + "step": 14261 + }, + { + "epoch": 0.74, + "learning_rate": 6.468645850549136e-08, + "loss": 0.1089, + "step": 14262 + }, + { + "epoch": 0.74, + "learning_rate": 6.466156233510566e-08, + "loss": 0.1169, + "step": 14263 + }, + { + "epoch": 0.74, + "learning_rate": 6.463667003274482e-08, + "loss": 0.1415, + "step": 14264 + }, + { + "epoch": 0.74, + "learning_rate": 6.461178159912019e-08, + "loss": 0.1211, + "step": 14265 + }, + { + "epoch": 0.74, + "learning_rate": 6.45868970349431e-08, + "loss": 0.0846, + "step": 14266 + }, + { + "epoch": 0.74, + "learning_rate": 6.456201634092472e-08, + "loss": 0.119, + "step": 14267 + }, + { + "epoch": 0.74, + "learning_rate": 6.453713951777629e-08, + "loss": 0.1195, + "step": 14268 + }, + { + "epoch": 0.74, + "learning_rate": 6.451226656620868e-08, + "loss": 0.1286, + "step": 14269 + }, + { + "epoch": 0.74, + "learning_rate": 6.448739748693277e-08, + "loss": 0.1361, + "step": 14270 + }, + { + "epoch": 0.74, + "learning_rate": 6.446253228065939e-08, + "loss": 0.1246, + "step": 14271 + }, + { + "epoch": 0.74, + "learning_rate": 6.443767094809914e-08, + "loss": 0.1184, + "step": 14272 + }, + { + "epoch": 0.74, + "learning_rate": 6.441281348996259e-08, + "loss": 0.1175, + "step": 14273 + }, + { + "epoch": 0.75, + "learning_rate": 6.438795990696011e-08, + "loss": 0.1529, + "step": 14274 + }, + { + "epoch": 0.75, + "learning_rate": 6.436311019980216e-08, + "loss": 0.1371, + "step": 14275 + }, + { + "epoch": 0.75, + "learning_rate": 6.433826436919887e-08, + "loss": 0.1544, + "step": 14276 + }, + { + "epoch": 0.75, + "learning_rate": 6.431342241586035e-08, + "loss": 0.1382, + "step": 14277 + }, + { + "epoch": 0.75, + "learning_rate": 6.428858434049655e-08, + "loss": 0.1369, + "step": 14278 + }, + { + "epoch": 0.75, + "learning_rate": 6.426375014381753e-08, + "loss": 0.1462, + "step": 14279 + }, + { + "epoch": 0.75, + "learning_rate": 6.423891982653284e-08, + "loss": 0.1191, + "step": 14280 + }, + { + "epoch": 0.75, + "learning_rate": 6.42140933893522e-08, + "loss": 0.1251, + "step": 14281 + }, + { + "epoch": 0.75, + "learning_rate": 6.418927083298522e-08, + "loss": 0.1416, + "step": 14282 + }, + { + "epoch": 0.75, + "learning_rate": 6.416445215814132e-08, + "loss": 0.1448, + "step": 14283 + }, + { + "epoch": 0.75, + "learning_rate": 6.413963736552983e-08, + "loss": 0.1703, + "step": 14284 + }, + { + "epoch": 0.75, + "learning_rate": 6.411482645585988e-08, + "loss": 0.1649, + "step": 14285 + }, + { + "epoch": 0.75, + "learning_rate": 6.409001942984069e-08, + "loss": 0.1874, + "step": 14286 + }, + { + "epoch": 0.75, + "learning_rate": 6.406521628818124e-08, + "loss": 0.1348, + "step": 14287 + }, + { + "epoch": 0.75, + "learning_rate": 6.404041703159044e-08, + "loss": 0.1128, + "step": 14288 + }, + { + "epoch": 0.75, + "learning_rate": 6.401562166077687e-08, + "loss": 0.1473, + "step": 14289 + }, + { + "epoch": 0.75, + "learning_rate": 6.399083017644939e-08, + "loss": 0.1243, + "step": 14290 + }, + { + "epoch": 0.75, + "learning_rate": 6.396604257931651e-08, + "loss": 0.1525, + "step": 14291 + }, + { + "epoch": 0.75, + "learning_rate": 6.394125887008663e-08, + "loss": 0.1414, + "step": 14292 + }, + { + "epoch": 0.75, + "learning_rate": 6.391647904946804e-08, + "loss": 0.1032, + "step": 14293 + }, + { + "epoch": 0.75, + "learning_rate": 6.389170311816907e-08, + "loss": 0.1567, + "step": 14294 + }, + { + "epoch": 0.75, + "learning_rate": 6.386693107689775e-08, + "loss": 0.1307, + "step": 14295 + }, + { + "epoch": 0.75, + "learning_rate": 6.384216292636212e-08, + "loss": 0.1252, + "step": 14296 + }, + { + "epoch": 0.75, + "learning_rate": 6.381739866727003e-08, + "loss": 0.1518, + "step": 14297 + }, + { + "epoch": 0.75, + "learning_rate": 6.379263830032926e-08, + "loss": 0.1092, + "step": 14298 + }, + { + "epoch": 0.75, + "learning_rate": 6.376788182624746e-08, + "loss": 0.1295, + "step": 14299 + }, + { + "epoch": 0.75, + "learning_rate": 6.374312924573213e-08, + "loss": 0.134, + "step": 14300 + }, + { + "epoch": 0.75, + "learning_rate": 6.371838055949084e-08, + "loss": 0.1411, + "step": 14301 + }, + { + "epoch": 0.75, + "learning_rate": 6.369363576823084e-08, + "loss": 0.1295, + "step": 14302 + }, + { + "epoch": 0.75, + "learning_rate": 6.366889487265932e-08, + "loss": 0.1377, + "step": 14303 + }, + { + "epoch": 0.75, + "learning_rate": 6.364415787348339e-08, + "loss": 0.1729, + "step": 14304 + }, + { + "epoch": 0.75, + "learning_rate": 6.361942477141016e-08, + "loss": 0.1133, + "step": 14305 + }, + { + "epoch": 0.75, + "learning_rate": 6.359469556714638e-08, + "loss": 0.149, + "step": 14306 + }, + { + "epoch": 0.75, + "learning_rate": 6.356997026139879e-08, + "loss": 0.1365, + "step": 14307 + }, + { + "epoch": 0.75, + "learning_rate": 6.354524885487417e-08, + "loss": 0.1505, + "step": 14308 + }, + { + "epoch": 0.75, + "learning_rate": 6.352053134827902e-08, + "loss": 0.1359, + "step": 14309 + }, + { + "epoch": 0.75, + "learning_rate": 6.349581774231978e-08, + "loss": 0.1531, + "step": 14310 + }, + { + "epoch": 0.75, + "learning_rate": 6.347110803770267e-08, + "loss": 0.1463, + "step": 14311 + }, + { + "epoch": 0.75, + "learning_rate": 6.344640223513408e-08, + "loss": 0.153, + "step": 14312 + }, + { + "epoch": 0.75, + "learning_rate": 6.342170033532004e-08, + "loss": 0.1421, + "step": 14313 + }, + { + "epoch": 0.75, + "learning_rate": 6.339700233896652e-08, + "loss": 0.1111, + "step": 14314 + }, + { + "epoch": 0.75, + "learning_rate": 6.337230824677942e-08, + "loss": 0.1382, + "step": 14315 + }, + { + "epoch": 0.75, + "learning_rate": 6.334761805946449e-08, + "loss": 0.1391, + "step": 14316 + }, + { + "epoch": 0.75, + "learning_rate": 6.332293177772738e-08, + "loss": 0.1405, + "step": 14317 + }, + { + "epoch": 0.75, + "learning_rate": 6.329824940227357e-08, + "loss": 0.1263, + "step": 14318 + }, + { + "epoch": 0.75, + "learning_rate": 6.327357093380866e-08, + "loss": 0.128, + "step": 14319 + }, + { + "epoch": 0.75, + "learning_rate": 6.324889637303783e-08, + "loss": 0.1474, + "step": 14320 + }, + { + "epoch": 0.75, + "learning_rate": 6.322422572066637e-08, + "loss": 0.1392, + "step": 14321 + }, + { + "epoch": 0.75, + "learning_rate": 6.319955897739932e-08, + "loss": 0.1431, + "step": 14322 + }, + { + "epoch": 0.75, + "learning_rate": 6.31748961439417e-08, + "loss": 0.1189, + "step": 14323 + }, + { + "epoch": 0.75, + "learning_rate": 6.315023722099835e-08, + "loss": 0.163, + "step": 14324 + }, + { + "epoch": 0.75, + "learning_rate": 6.312558220927399e-08, + "loss": 0.0971, + "step": 14325 + }, + { + "epoch": 0.75, + "learning_rate": 6.310093110947339e-08, + "loss": 0.1454, + "step": 14326 + }, + { + "epoch": 0.75, + "learning_rate": 6.307628392230105e-08, + "loss": 0.1336, + "step": 14327 + }, + { + "epoch": 0.75, + "learning_rate": 6.305164064846135e-08, + "loss": 0.1168, + "step": 14328 + }, + { + "epoch": 0.75, + "learning_rate": 6.302700128865856e-08, + "loss": 0.1018, + "step": 14329 + }, + { + "epoch": 0.75, + "learning_rate": 6.300236584359704e-08, + "loss": 0.1133, + "step": 14330 + }, + { + "epoch": 0.75, + "learning_rate": 6.297773431398083e-08, + "loss": 0.1348, + "step": 14331 + }, + { + "epoch": 0.75, + "learning_rate": 6.295310670051375e-08, + "loss": 0.1611, + "step": 14332 + }, + { + "epoch": 0.75, + "learning_rate": 6.292848300389986e-08, + "loss": 0.1377, + "step": 14333 + }, + { + "epoch": 0.75, + "learning_rate": 6.290386322484283e-08, + "loss": 0.1788, + "step": 14334 + }, + { + "epoch": 0.75, + "learning_rate": 6.28792473640463e-08, + "loss": 0.1275, + "step": 14335 + }, + { + "epoch": 0.75, + "learning_rate": 6.285463542221379e-08, + "loss": 0.1358, + "step": 14336 + }, + { + "epoch": 0.75, + "learning_rate": 6.283002740004877e-08, + "loss": 0.1077, + "step": 14337 + }, + { + "epoch": 0.75, + "learning_rate": 6.280542329825454e-08, + "loss": 0.1278, + "step": 14338 + }, + { + "epoch": 0.75, + "learning_rate": 6.278082311753426e-08, + "loss": 0.1174, + "step": 14339 + }, + { + "epoch": 0.75, + "learning_rate": 6.275622685859104e-08, + "loss": 0.146, + "step": 14340 + }, + { + "epoch": 0.75, + "learning_rate": 6.273163452212784e-08, + "loss": 0.1572, + "step": 14341 + }, + { + "epoch": 0.75, + "learning_rate": 6.27070461088475e-08, + "loss": 0.1387, + "step": 14342 + }, + { + "epoch": 0.75, + "learning_rate": 6.268246161945271e-08, + "loss": 0.1305, + "step": 14343 + }, + { + "epoch": 0.75, + "learning_rate": 6.265788105464626e-08, + "loss": 0.1671, + "step": 14344 + }, + { + "epoch": 0.75, + "learning_rate": 6.263330441513059e-08, + "loss": 0.093, + "step": 14345 + }, + { + "epoch": 0.75, + "learning_rate": 6.260873170160812e-08, + "loss": 0.1181, + "step": 14346 + }, + { + "epoch": 0.75, + "learning_rate": 6.258416291478106e-08, + "loss": 0.155, + "step": 14347 + }, + { + "epoch": 0.75, + "learning_rate": 6.255959805535179e-08, + "loss": 0.1357, + "step": 14348 + }, + { + "epoch": 0.75, + "learning_rate": 6.253503712402222e-08, + "loss": 0.1144, + "step": 14349 + }, + { + "epoch": 0.75, + "learning_rate": 6.251048012149427e-08, + "loss": 0.1475, + "step": 14350 + }, + { + "epoch": 0.75, + "learning_rate": 6.248592704846996e-08, + "loss": 0.1388, + "step": 14351 + }, + { + "epoch": 0.75, + "learning_rate": 6.246137790565092e-08, + "loss": 0.1455, + "step": 14352 + }, + { + "epoch": 0.75, + "learning_rate": 6.243683269373881e-08, + "loss": 0.1147, + "step": 14353 + }, + { + "epoch": 0.75, + "learning_rate": 6.241229141343505e-08, + "loss": 0.162, + "step": 14354 + }, + { + "epoch": 0.75, + "learning_rate": 6.238775406544119e-08, + "loss": 0.1259, + "step": 14355 + }, + { + "epoch": 0.75, + "learning_rate": 6.23632206504585e-08, + "loss": 0.1384, + "step": 14356 + }, + { + "epoch": 0.75, + "learning_rate": 6.233869116918798e-08, + "loss": 0.1146, + "step": 14357 + }, + { + "epoch": 0.75, + "learning_rate": 6.231416562233084e-08, + "loss": 0.155, + "step": 14358 + }, + { + "epoch": 0.75, + "learning_rate": 6.228964401058802e-08, + "loss": 0.1355, + "step": 14359 + }, + { + "epoch": 0.75, + "learning_rate": 6.226512633466035e-08, + "loss": 0.1136, + "step": 14360 + }, + { + "epoch": 0.75, + "learning_rate": 6.224061259524846e-08, + "loss": 0.1433, + "step": 14361 + }, + { + "epoch": 0.75, + "learning_rate": 6.221610279305312e-08, + "loss": 0.1289, + "step": 14362 + }, + { + "epoch": 0.75, + "learning_rate": 6.219159692877473e-08, + "loss": 0.1333, + "step": 14363 + }, + { + "epoch": 0.75, + "learning_rate": 6.216709500311372e-08, + "loss": 0.1291, + "step": 14364 + }, + { + "epoch": 0.75, + "learning_rate": 6.214259701677034e-08, + "loss": 0.1261, + "step": 14365 + }, + { + "epoch": 0.75, + "learning_rate": 6.211810297044475e-08, + "loss": 0.136, + "step": 14366 + }, + { + "epoch": 0.75, + "learning_rate": 6.2093612864837e-08, + "loss": 0.1387, + "step": 14367 + }, + { + "epoch": 0.75, + "learning_rate": 6.206912670064698e-08, + "loss": 0.1304, + "step": 14368 + }, + { + "epoch": 0.75, + "learning_rate": 6.204464447857462e-08, + "loss": 0.1905, + "step": 14369 + }, + { + "epoch": 0.75, + "learning_rate": 6.202016619931957e-08, + "loss": 0.1026, + "step": 14370 + }, + { + "epoch": 0.75, + "learning_rate": 6.199569186358147e-08, + "loss": 0.1521, + "step": 14371 + }, + { + "epoch": 0.75, + "learning_rate": 6.197122147205968e-08, + "loss": 0.1558, + "step": 14372 + }, + { + "epoch": 0.75, + "learning_rate": 6.194675502545373e-08, + "loss": 0.1331, + "step": 14373 + }, + { + "epoch": 0.75, + "learning_rate": 6.192229252446289e-08, + "loss": 0.1335, + "step": 14374 + }, + { + "epoch": 0.75, + "learning_rate": 6.189783396978609e-08, + "loss": 0.1193, + "step": 14375 + }, + { + "epoch": 0.75, + "learning_rate": 6.187337936212259e-08, + "loss": 0.1354, + "step": 14376 + }, + { + "epoch": 0.75, + "learning_rate": 6.18489287021712e-08, + "loss": 0.144, + "step": 14377 + }, + { + "epoch": 0.75, + "learning_rate": 6.18244819906308e-08, + "loss": 0.1597, + "step": 14378 + }, + { + "epoch": 0.75, + "learning_rate": 6.180003922819995e-08, + "loss": 0.1489, + "step": 14379 + }, + { + "epoch": 0.75, + "learning_rate": 6.177560041557741e-08, + "loss": 0.1487, + "step": 14380 + }, + { + "epoch": 0.75, + "learning_rate": 6.175116555346156e-08, + "loss": 0.1322, + "step": 14381 + }, + { + "epoch": 0.75, + "learning_rate": 6.172673464255076e-08, + "loss": 0.1354, + "step": 14382 + }, + { + "epoch": 0.75, + "learning_rate": 6.170230768354327e-08, + "loss": 0.1201, + "step": 14383 + }, + { + "epoch": 0.75, + "learning_rate": 6.167788467713724e-08, + "loss": 0.1459, + "step": 14384 + }, + { + "epoch": 0.75, + "learning_rate": 6.165346562403065e-08, + "loss": 0.0987, + "step": 14385 + }, + { + "epoch": 0.75, + "learning_rate": 6.162905052492134e-08, + "loss": 0.1203, + "step": 14386 + }, + { + "epoch": 0.75, + "learning_rate": 6.160463938050724e-08, + "loss": 0.1176, + "step": 14387 + }, + { + "epoch": 0.75, + "learning_rate": 6.1580232191486e-08, + "loss": 0.1578, + "step": 14388 + }, + { + "epoch": 0.75, + "learning_rate": 6.155582895855513e-08, + "loss": 0.1168, + "step": 14389 + }, + { + "epoch": 0.75, + "learning_rate": 6.153142968241207e-08, + "loss": 0.1042, + "step": 14390 + }, + { + "epoch": 0.75, + "learning_rate": 6.150703436375433e-08, + "loss": 0.1034, + "step": 14391 + }, + { + "epoch": 0.75, + "learning_rate": 6.148264300327894e-08, + "loss": 0.1257, + "step": 14392 + }, + { + "epoch": 0.75, + "learning_rate": 6.145825560168303e-08, + "loss": 0.1165, + "step": 14393 + }, + { + "epoch": 0.75, + "learning_rate": 6.143387215966373e-08, + "loss": 0.1121, + "step": 14394 + }, + { + "epoch": 0.75, + "learning_rate": 6.140949267791785e-08, + "loss": 0.1499, + "step": 14395 + }, + { + "epoch": 0.75, + "learning_rate": 6.138511715714217e-08, + "loss": 0.1594, + "step": 14396 + }, + { + "epoch": 0.75, + "learning_rate": 6.136074559803326e-08, + "loss": 0.143, + "step": 14397 + }, + { + "epoch": 0.75, + "learning_rate": 6.133637800128786e-08, + "loss": 0.1194, + "step": 14398 + }, + { + "epoch": 0.75, + "learning_rate": 6.131201436760236e-08, + "loss": 0.1212, + "step": 14399 + }, + { + "epoch": 0.75, + "learning_rate": 6.128765469767297e-08, + "loss": 0.1503, + "step": 14400 + }, + { + "epoch": 0.75, + "learning_rate": 6.126329899219587e-08, + "loss": 0.1346, + "step": 14401 + }, + { + "epoch": 0.75, + "learning_rate": 6.123894725186734e-08, + "loss": 0.1507, + "step": 14402 + }, + { + "epoch": 0.75, + "learning_rate": 6.121459947738323e-08, + "loss": 0.1331, + "step": 14403 + }, + { + "epoch": 0.75, + "learning_rate": 6.119025566943947e-08, + "loss": 0.1187, + "step": 14404 + }, + { + "epoch": 0.75, + "learning_rate": 6.116591582873172e-08, + "loss": 0.1133, + "step": 14405 + }, + { + "epoch": 0.75, + "learning_rate": 6.114157995595575e-08, + "loss": 0.0949, + "step": 14406 + }, + { + "epoch": 0.75, + "learning_rate": 6.111724805180703e-08, + "loss": 0.1365, + "step": 14407 + }, + { + "epoch": 0.75, + "learning_rate": 6.109292011698097e-08, + "loss": 0.1087, + "step": 14408 + }, + { + "epoch": 0.75, + "learning_rate": 6.106859615217288e-08, + "loss": 0.1385, + "step": 14409 + }, + { + "epoch": 0.75, + "learning_rate": 6.104427615807794e-08, + "loss": 0.1318, + "step": 14410 + }, + { + "epoch": 0.75, + "learning_rate": 6.101996013539123e-08, + "loss": 0.1462, + "step": 14411 + }, + { + "epoch": 0.75, + "learning_rate": 6.099564808480766e-08, + "loss": 0.1328, + "step": 14412 + }, + { + "epoch": 0.75, + "learning_rate": 6.09713400070222e-08, + "loss": 0.1397, + "step": 14413 + }, + { + "epoch": 0.75, + "learning_rate": 6.09470359027295e-08, + "loss": 0.1443, + "step": 14414 + }, + { + "epoch": 0.75, + "learning_rate": 6.092273577262419e-08, + "loss": 0.1512, + "step": 14415 + }, + { + "epoch": 0.75, + "learning_rate": 6.089843961740078e-08, + "loss": 0.1069, + "step": 14416 + }, + { + "epoch": 0.75, + "learning_rate": 6.087414743775365e-08, + "loss": 0.1762, + "step": 14417 + }, + { + "epoch": 0.75, + "learning_rate": 6.084985923437714e-08, + "loss": 0.12, + "step": 14418 + }, + { + "epoch": 0.75, + "learning_rate": 6.082557500796526e-08, + "loss": 0.1761, + "step": 14419 + }, + { + "epoch": 0.75, + "learning_rate": 6.080129475921225e-08, + "loss": 0.1322, + "step": 14420 + }, + { + "epoch": 0.75, + "learning_rate": 6.077701848881198e-08, + "loss": 0.1141, + "step": 14421 + }, + { + "epoch": 0.75, + "learning_rate": 6.075274619745828e-08, + "loss": 0.1498, + "step": 14422 + }, + { + "epoch": 0.75, + "learning_rate": 6.072847788584476e-08, + "loss": 0.1316, + "step": 14423 + }, + { + "epoch": 0.75, + "learning_rate": 6.070421355466517e-08, + "loss": 0.1094, + "step": 14424 + }, + { + "epoch": 0.75, + "learning_rate": 6.067995320461301e-08, + "loss": 0.172, + "step": 14425 + }, + { + "epoch": 0.75, + "learning_rate": 6.065569683638143e-08, + "loss": 0.1431, + "step": 14426 + }, + { + "epoch": 0.75, + "learning_rate": 6.063144445066391e-08, + "loss": 0.1201, + "step": 14427 + }, + { + "epoch": 0.75, + "learning_rate": 6.06071960481535e-08, + "loss": 0.1387, + "step": 14428 + }, + { + "epoch": 0.75, + "learning_rate": 6.058295162954321e-08, + "loss": 0.1108, + "step": 14429 + }, + { + "epoch": 0.75, + "learning_rate": 6.055871119552595e-08, + "loss": 0.0975, + "step": 14430 + }, + { + "epoch": 0.75, + "learning_rate": 6.053447474679461e-08, + "loss": 0.1174, + "step": 14431 + }, + { + "epoch": 0.75, + "learning_rate": 6.051024228404183e-08, + "loss": 0.1427, + "step": 14432 + }, + { + "epoch": 0.75, + "learning_rate": 6.048601380796016e-08, + "loss": 0.171, + "step": 14433 + }, + { + "epoch": 0.75, + "learning_rate": 6.046178931924209e-08, + "loss": 0.1337, + "step": 14434 + }, + { + "epoch": 0.75, + "learning_rate": 6.043756881857993e-08, + "loss": 0.1124, + "step": 14435 + }, + { + "epoch": 0.75, + "learning_rate": 6.041335230666593e-08, + "loss": 0.1417, + "step": 14436 + }, + { + "epoch": 0.75, + "learning_rate": 6.038913978419216e-08, + "loss": 0.1438, + "step": 14437 + }, + { + "epoch": 0.75, + "learning_rate": 6.036493125185071e-08, + "loss": 0.1243, + "step": 14438 + }, + { + "epoch": 0.75, + "learning_rate": 6.034072671033346e-08, + "loss": 0.1555, + "step": 14439 + }, + { + "epoch": 0.75, + "learning_rate": 6.031652616033211e-08, + "loss": 0.1437, + "step": 14440 + }, + { + "epoch": 0.75, + "learning_rate": 6.029232960253834e-08, + "loss": 0.1335, + "step": 14441 + }, + { + "epoch": 0.75, + "learning_rate": 6.026813703764384e-08, + "loss": 0.1316, + "step": 14442 + }, + { + "epoch": 0.75, + "learning_rate": 6.024394846633987e-08, + "loss": 0.1246, + "step": 14443 + }, + { + "epoch": 0.75, + "learning_rate": 6.021976388931774e-08, + "loss": 0.1308, + "step": 14444 + }, + { + "epoch": 0.75, + "learning_rate": 6.019558330726875e-08, + "loss": 0.1459, + "step": 14445 + }, + { + "epoch": 0.75, + "learning_rate": 6.017140672088399e-08, + "loss": 0.1104, + "step": 14446 + }, + { + "epoch": 0.75, + "learning_rate": 6.01472341308544e-08, + "loss": 0.1462, + "step": 14447 + }, + { + "epoch": 0.75, + "learning_rate": 6.012306553787079e-08, + "loss": 0.1243, + "step": 14448 + }, + { + "epoch": 0.75, + "learning_rate": 6.009890094262403e-08, + "loss": 0.1002, + "step": 14449 + }, + { + "epoch": 0.75, + "learning_rate": 6.007474034580467e-08, + "loss": 0.1314, + "step": 14450 + }, + { + "epoch": 0.75, + "learning_rate": 6.005058374810324e-08, + "loss": 0.1336, + "step": 14451 + }, + { + "epoch": 0.75, + "learning_rate": 6.002643115021018e-08, + "loss": 0.1404, + "step": 14452 + }, + { + "epoch": 0.75, + "learning_rate": 6.000228255281574e-08, + "loss": 0.1422, + "step": 14453 + }, + { + "epoch": 0.75, + "learning_rate": 5.997813795661011e-08, + "loss": 0.1183, + "step": 14454 + }, + { + "epoch": 0.75, + "learning_rate": 5.995399736228328e-08, + "loss": 0.1255, + "step": 14455 + }, + { + "epoch": 0.75, + "learning_rate": 5.992986077052533e-08, + "loss": 0.1299, + "step": 14456 + }, + { + "epoch": 0.75, + "learning_rate": 5.990572818202606e-08, + "loss": 0.1504, + "step": 14457 + }, + { + "epoch": 0.75, + "learning_rate": 5.988159959747514e-08, + "loss": 0.1484, + "step": 14458 + }, + { + "epoch": 0.75, + "learning_rate": 5.98574750175622e-08, + "loss": 0.1351, + "step": 14459 + }, + { + "epoch": 0.75, + "learning_rate": 5.98333544429767e-08, + "loss": 0.1096, + "step": 14460 + }, + { + "epoch": 0.75, + "learning_rate": 5.980923787440808e-08, + "loss": 0.1604, + "step": 14461 + }, + { + "epoch": 0.75, + "learning_rate": 5.978512531254548e-08, + "loss": 0.1191, + "step": 14462 + }, + { + "epoch": 0.75, + "learning_rate": 5.976101675807819e-08, + "loss": 0.1212, + "step": 14463 + }, + { + "epoch": 0.75, + "learning_rate": 5.973691221169517e-08, + "loss": 0.1264, + "step": 14464 + }, + { + "epoch": 0.76, + "learning_rate": 5.971281167408535e-08, + "loss": 0.1889, + "step": 14465 + }, + { + "epoch": 0.76, + "learning_rate": 5.968871514593747e-08, + "loss": 0.13, + "step": 14466 + }, + { + "epoch": 0.76, + "learning_rate": 5.966462262794034e-08, + "loss": 0.1454, + "step": 14467 + }, + { + "epoch": 0.76, + "learning_rate": 5.964053412078253e-08, + "loss": 0.1491, + "step": 14468 + }, + { + "epoch": 0.76, + "learning_rate": 5.96164496251523e-08, + "loss": 0.1045, + "step": 14469 + }, + { + "epoch": 0.76, + "learning_rate": 5.959236914173822e-08, + "loss": 0.1256, + "step": 14470 + }, + { + "epoch": 0.76, + "learning_rate": 5.956829267122843e-08, + "loss": 0.1088, + "step": 14471 + }, + { + "epoch": 0.76, + "learning_rate": 5.9544220214311046e-08, + "loss": 0.1437, + "step": 14472 + }, + { + "epoch": 0.76, + "learning_rate": 5.9520151771674e-08, + "loss": 0.1312, + "step": 14473 + }, + { + "epoch": 0.76, + "learning_rate": 5.949608734400531e-08, + "loss": 0.1325, + "step": 14474 + }, + { + "epoch": 0.76, + "learning_rate": 5.947202693199267e-08, + "loss": 0.1086, + "step": 14475 + }, + { + "epoch": 0.76, + "learning_rate": 5.944797053632378e-08, + "loss": 0.1165, + "step": 14476 + }, + { + "epoch": 0.76, + "learning_rate": 5.9423918157686125e-08, + "loss": 0.11, + "step": 14477 + }, + { + "epoch": 0.76, + "learning_rate": 5.9399869796767164e-08, + "loss": 0.1515, + "step": 14478 + }, + { + "epoch": 0.76, + "learning_rate": 5.9375825454254216e-08, + "loss": 0.129, + "step": 14479 + }, + { + "epoch": 0.76, + "learning_rate": 5.9351785130834385e-08, + "loss": 0.1177, + "step": 14480 + }, + { + "epoch": 0.76, + "learning_rate": 5.932774882719489e-08, + "loss": 0.1526, + "step": 14481 + }, + { + "epoch": 0.76, + "learning_rate": 5.930371654402266e-08, + "loss": 0.1343, + "step": 14482 + }, + { + "epoch": 0.76, + "learning_rate": 5.927968828200452e-08, + "loss": 0.1161, + "step": 14483 + }, + { + "epoch": 0.76, + "learning_rate": 5.925566404182711e-08, + "loss": 0.1304, + "step": 14484 + }, + { + "epoch": 0.76, + "learning_rate": 5.923164382417734e-08, + "loss": 0.116, + "step": 14485 + }, + { + "epoch": 0.76, + "learning_rate": 5.920762762974143e-08, + "loss": 0.1282, + "step": 14486 + }, + { + "epoch": 0.76, + "learning_rate": 5.9183615459205805e-08, + "loss": 0.1011, + "step": 14487 + }, + { + "epoch": 0.76, + "learning_rate": 5.915960731325691e-08, + "loss": 0.1384, + "step": 14488 + }, + { + "epoch": 0.76, + "learning_rate": 5.913560319258077e-08, + "loss": 0.1681, + "step": 14489 + }, + { + "epoch": 0.76, + "learning_rate": 5.9111603097863496e-08, + "loss": 0.1347, + "step": 14490 + }, + { + "epoch": 0.76, + "learning_rate": 5.908760702979092e-08, + "loss": 0.1469, + "step": 14491 + }, + { + "epoch": 0.76, + "learning_rate": 5.9063614989048994e-08, + "loss": 0.1206, + "step": 14492 + }, + { + "epoch": 0.76, + "learning_rate": 5.903962697632343e-08, + "loss": 0.1797, + "step": 14493 + }, + { + "epoch": 0.76, + "learning_rate": 5.9015642992299596e-08, + "loss": 0.1139, + "step": 14494 + }, + { + "epoch": 0.76, + "learning_rate": 5.8991663037663206e-08, + "loss": 0.1567, + "step": 14495 + }, + { + "epoch": 0.76, + "learning_rate": 5.89676871130995e-08, + "loss": 0.154, + "step": 14496 + }, + { + "epoch": 0.76, + "learning_rate": 5.894371521929375e-08, + "loss": 0.1153, + "step": 14497 + }, + { + "epoch": 0.76, + "learning_rate": 5.891974735693099e-08, + "loss": 0.1537, + "step": 14498 + }, + { + "epoch": 0.76, + "learning_rate": 5.8895783526696396e-08, + "loss": 0.1221, + "step": 14499 + }, + { + "epoch": 0.76, + "learning_rate": 5.887182372927477e-08, + "loss": 0.1343, + "step": 14500 + }, + { + "epoch": 0.76, + "learning_rate": 5.88478679653509e-08, + "loss": 0.1698, + "step": 14501 + }, + { + "epoch": 0.76, + "learning_rate": 5.882391623560947e-08, + "loss": 0.1255, + "step": 14502 + }, + { + "epoch": 0.76, + "learning_rate": 5.8799968540735014e-08, + "loss": 0.1199, + "step": 14503 + }, + { + "epoch": 0.76, + "learning_rate": 5.8776024881411954e-08, + "loss": 0.1558, + "step": 14504 + }, + { + "epoch": 0.76, + "learning_rate": 5.8752085258324554e-08, + "loss": 0.12, + "step": 14505 + }, + { + "epoch": 0.76, + "learning_rate": 5.872814967215716e-08, + "loss": 0.1352, + "step": 14506 + }, + { + "epoch": 0.76, + "learning_rate": 5.8704218123593784e-08, + "loss": 0.1043, + "step": 14507 + }, + { + "epoch": 0.76, + "learning_rate": 5.8680290613318385e-08, + "loss": 0.151, + "step": 14508 + }, + { + "epoch": 0.76, + "learning_rate": 5.865636714201478e-08, + "loss": 0.1191, + "step": 14509 + }, + { + "epoch": 0.76, + "learning_rate": 5.863244771036684e-08, + "loss": 0.1668, + "step": 14510 + }, + { + "epoch": 0.76, + "learning_rate": 5.8608532319058156e-08, + "loss": 0.1474, + "step": 14511 + }, + { + "epoch": 0.76, + "learning_rate": 5.858462096877217e-08, + "loss": 0.1368, + "step": 14512 + }, + { + "epoch": 0.76, + "learning_rate": 5.856071366019222e-08, + "loss": 0.1377, + "step": 14513 + }, + { + "epoch": 0.76, + "learning_rate": 5.853681039400176e-08, + "loss": 0.133, + "step": 14514 + }, + { + "epoch": 0.76, + "learning_rate": 5.851291117088384e-08, + "loss": 0.1254, + "step": 14515 + }, + { + "epoch": 0.76, + "learning_rate": 5.8489015991521564e-08, + "loss": 0.1106, + "step": 14516 + }, + { + "epoch": 0.76, + "learning_rate": 5.8465124856597757e-08, + "loss": 0.1839, + "step": 14517 + }, + { + "epoch": 0.76, + "learning_rate": 5.8441237766795395e-08, + "loss": 0.2096, + "step": 14518 + }, + { + "epoch": 0.76, + "learning_rate": 5.841735472279716e-08, + "loss": 0.1276, + "step": 14519 + }, + { + "epoch": 0.76, + "learning_rate": 5.839347572528548e-08, + "loss": 0.1546, + "step": 14520 + }, + { + "epoch": 0.76, + "learning_rate": 5.836960077494297e-08, + "loss": 0.1333, + "step": 14521 + }, + { + "epoch": 0.76, + "learning_rate": 5.8345729872451944e-08, + "loss": 0.1217, + "step": 14522 + }, + { + "epoch": 0.76, + "learning_rate": 5.8321863018494644e-08, + "loss": 0.1123, + "step": 14523 + }, + { + "epoch": 0.76, + "learning_rate": 5.829800021375313e-08, + "loss": 0.1367, + "step": 14524 + }, + { + "epoch": 0.76, + "learning_rate": 5.8274141458909545e-08, + "loss": 0.1284, + "step": 14525 + }, + { + "epoch": 0.76, + "learning_rate": 5.8250286754645695e-08, + "loss": 0.1494, + "step": 14526 + }, + { + "epoch": 0.76, + "learning_rate": 5.822643610164339e-08, + "loss": 0.1459, + "step": 14527 + }, + { + "epoch": 0.76, + "learning_rate": 5.8202589500584256e-08, + "loss": 0.1455, + "step": 14528 + }, + { + "epoch": 0.76, + "learning_rate": 5.817874695214986e-08, + "loss": 0.1738, + "step": 14529 + }, + { + "epoch": 0.76, + "learning_rate": 5.815490845702162e-08, + "loss": 0.1288, + "step": 14530 + }, + { + "epoch": 0.76, + "learning_rate": 5.813107401588078e-08, + "loss": 0.1324, + "step": 14531 + }, + { + "epoch": 0.76, + "learning_rate": 5.810724362940869e-08, + "loss": 0.1183, + "step": 14532 + }, + { + "epoch": 0.76, + "learning_rate": 5.808341729828634e-08, + "loss": 0.1096, + "step": 14533 + }, + { + "epoch": 0.76, + "learning_rate": 5.805959502319473e-08, + "loss": 0.1406, + "step": 14534 + }, + { + "epoch": 0.76, + "learning_rate": 5.803577680481462e-08, + "loss": 0.1323, + "step": 14535 + }, + { + "epoch": 0.76, + "learning_rate": 5.801196264382693e-08, + "loss": 0.1206, + "step": 14536 + }, + { + "epoch": 0.76, + "learning_rate": 5.7988152540912095e-08, + "loss": 0.1448, + "step": 14537 + }, + { + "epoch": 0.76, + "learning_rate": 5.796434649675064e-08, + "loss": 0.1252, + "step": 14538 + }, + { + "epoch": 0.76, + "learning_rate": 5.7940544512023035e-08, + "loss": 0.144, + "step": 14539 + }, + { + "epoch": 0.76, + "learning_rate": 5.79167465874095e-08, + "loss": 0.1543, + "step": 14540 + }, + { + "epoch": 0.76, + "learning_rate": 5.789295272359023e-08, + "loss": 0.1174, + "step": 14541 + }, + { + "epoch": 0.76, + "learning_rate": 5.786916292124513e-08, + "loss": 0.1407, + "step": 14542 + }, + { + "epoch": 0.76, + "learning_rate": 5.784537718105431e-08, + "loss": 0.1238, + "step": 14543 + }, + { + "epoch": 0.76, + "learning_rate": 5.782159550369747e-08, + "loss": 0.119, + "step": 14544 + }, + { + "epoch": 0.76, + "learning_rate": 5.779781788985434e-08, + "loss": 0.1141, + "step": 14545 + }, + { + "epoch": 0.76, + "learning_rate": 5.7774044340204456e-08, + "loss": 0.1339, + "step": 14546 + }, + { + "epoch": 0.76, + "learning_rate": 5.7750274855427294e-08, + "loss": 0.12, + "step": 14547 + }, + { + "epoch": 0.76, + "learning_rate": 5.7726509436202185e-08, + "loss": 0.15, + "step": 14548 + }, + { + "epoch": 0.76, + "learning_rate": 5.77027480832083e-08, + "loss": 0.1396, + "step": 14549 + }, + { + "epoch": 0.76, + "learning_rate": 5.767899079712488e-08, + "loss": 0.1562, + "step": 14550 + }, + { + "epoch": 0.76, + "learning_rate": 5.765523757863082e-08, + "loss": 0.1386, + "step": 14551 + }, + { + "epoch": 0.76, + "learning_rate": 5.7631488428405037e-08, + "loss": 0.1317, + "step": 14552 + }, + { + "epoch": 0.76, + "learning_rate": 5.760774334712628e-08, + "loss": 0.1136, + "step": 14553 + }, + { + "epoch": 0.76, + "learning_rate": 5.758400233547318e-08, + "loss": 0.113, + "step": 14554 + }, + { + "epoch": 0.76, + "learning_rate": 5.756026539412426e-08, + "loss": 0.1344, + "step": 14555 + }, + { + "epoch": 0.76, + "learning_rate": 5.75365325237579e-08, + "loss": 0.1454, + "step": 14556 + }, + { + "epoch": 0.76, + "learning_rate": 5.7512803725052494e-08, + "loss": 0.1216, + "step": 14557 + }, + { + "epoch": 0.76, + "learning_rate": 5.748907899868616e-08, + "loss": 0.1324, + "step": 14558 + }, + { + "epoch": 0.76, + "learning_rate": 5.746535834533695e-08, + "loss": 0.112, + "step": 14559 + }, + { + "epoch": 0.76, + "learning_rate": 5.744164176568278e-08, + "loss": 0.1136, + "step": 14560 + }, + { + "epoch": 0.76, + "learning_rate": 5.7417929260401564e-08, + "loss": 0.123, + "step": 14561 + }, + { + "epoch": 0.76, + "learning_rate": 5.7394220830171044e-08, + "loss": 0.1451, + "step": 14562 + }, + { + "epoch": 0.76, + "learning_rate": 5.737051647566862e-08, + "loss": 0.1131, + "step": 14563 + }, + { + "epoch": 0.76, + "learning_rate": 5.7346816197571955e-08, + "loss": 0.1476, + "step": 14564 + }, + { + "epoch": 0.76, + "learning_rate": 5.732311999655837e-08, + "loss": 0.1292, + "step": 14565 + }, + { + "epoch": 0.76, + "learning_rate": 5.7299427873305087e-08, + "loss": 0.1508, + "step": 14566 + }, + { + "epoch": 0.76, + "learning_rate": 5.727573982848917e-08, + "loss": 0.1245, + "step": 14567 + }, + { + "epoch": 0.76, + "learning_rate": 5.725205586278776e-08, + "loss": 0.1333, + "step": 14568 + }, + { + "epoch": 0.76, + "learning_rate": 5.722837597687773e-08, + "loss": 0.1183, + "step": 14569 + }, + { + "epoch": 0.76, + "learning_rate": 5.7204700171435795e-08, + "loss": 0.125, + "step": 14570 + }, + { + "epoch": 0.76, + "learning_rate": 5.718102844713868e-08, + "loss": 0.1435, + "step": 14571 + }, + { + "epoch": 0.76, + "learning_rate": 5.715736080466289e-08, + "loss": 0.1286, + "step": 14572 + }, + { + "epoch": 0.76, + "learning_rate": 5.713369724468489e-08, + "loss": 0.1219, + "step": 14573 + }, + { + "epoch": 0.76, + "learning_rate": 5.71100377678809e-08, + "loss": 0.1383, + "step": 14574 + }, + { + "epoch": 0.76, + "learning_rate": 5.7086382374927244e-08, + "loss": 0.1918, + "step": 14575 + }, + { + "epoch": 0.76, + "learning_rate": 5.706273106649997e-08, + "loss": 0.1285, + "step": 14576 + }, + { + "epoch": 0.76, + "learning_rate": 5.7039083843275e-08, + "loss": 0.1397, + "step": 14577 + }, + { + "epoch": 0.76, + "learning_rate": 5.701544070592816e-08, + "loss": 0.1268, + "step": 14578 + }, + { + "epoch": 0.76, + "learning_rate": 5.699180165513533e-08, + "loss": 0.1048, + "step": 14579 + }, + { + "epoch": 0.76, + "learning_rate": 5.696816669157196e-08, + "loss": 0.1548, + "step": 14580 + }, + { + "epoch": 0.76, + "learning_rate": 5.6944535815913544e-08, + "loss": 0.1118, + "step": 14581 + }, + { + "epoch": 0.76, + "learning_rate": 5.6920909028835594e-08, + "loss": 0.1345, + "step": 14582 + }, + { + "epoch": 0.76, + "learning_rate": 5.689728633101327e-08, + "loss": 0.1282, + "step": 14583 + }, + { + "epoch": 0.76, + "learning_rate": 5.687366772312177e-08, + "loss": 0.1273, + "step": 14584 + }, + { + "epoch": 0.76, + "learning_rate": 5.6850053205836024e-08, + "loss": 0.1737, + "step": 14585 + }, + { + "epoch": 0.76, + "learning_rate": 5.6826442779831084e-08, + "loss": 0.1266, + "step": 14586 + }, + { + "epoch": 0.76, + "learning_rate": 5.680283644578168e-08, + "loss": 0.1328, + "step": 14587 + }, + { + "epoch": 0.76, + "learning_rate": 5.677923420436248e-08, + "loss": 0.1556, + "step": 14588 + }, + { + "epoch": 0.76, + "learning_rate": 5.675563605624807e-08, + "loss": 0.1458, + "step": 14589 + }, + { + "epoch": 0.76, + "learning_rate": 5.673204200211286e-08, + "loss": 0.1262, + "step": 14590 + }, + { + "epoch": 0.76, + "learning_rate": 5.6708452042631214e-08, + "loss": 0.1297, + "step": 14591 + }, + { + "epoch": 0.76, + "learning_rate": 5.668486617847725e-08, + "loss": 0.1109, + "step": 14592 + }, + { + "epoch": 0.76, + "learning_rate": 5.6661284410325204e-08, + "loss": 0.1796, + "step": 14593 + }, + { + "epoch": 0.76, + "learning_rate": 5.6637706738849e-08, + "loss": 0.1051, + "step": 14594 + }, + { + "epoch": 0.76, + "learning_rate": 5.661413316472246e-08, + "loss": 0.1269, + "step": 14595 + }, + { + "epoch": 0.76, + "learning_rate": 5.6590563688619366e-08, + "loss": 0.1668, + "step": 14596 + }, + { + "epoch": 0.76, + "learning_rate": 5.65669983112133e-08, + "loss": 0.167, + "step": 14597 + }, + { + "epoch": 0.76, + "learning_rate": 5.654343703317781e-08, + "loss": 0.1116, + "step": 14598 + }, + { + "epoch": 0.76, + "learning_rate": 5.6519879855186206e-08, + "loss": 0.1921, + "step": 14599 + }, + { + "epoch": 0.76, + "learning_rate": 5.649632677791187e-08, + "loss": 0.1222, + "step": 14600 + }, + { + "epoch": 0.76, + "learning_rate": 5.647277780202793e-08, + "loss": 0.165, + "step": 14601 + }, + { + "epoch": 0.76, + "learning_rate": 5.64492329282074e-08, + "loss": 0.1437, + "step": 14602 + }, + { + "epoch": 0.76, + "learning_rate": 5.642569215712314e-08, + "loss": 0.1507, + "step": 14603 + }, + { + "epoch": 0.76, + "learning_rate": 5.640215548944811e-08, + "loss": 0.1342, + "step": 14604 + }, + { + "epoch": 0.76, + "learning_rate": 5.637862292585496e-08, + "loss": 0.1417, + "step": 14605 + }, + { + "epoch": 0.76, + "learning_rate": 5.6355094467016077e-08, + "loss": 0.1182, + "step": 14606 + }, + { + "epoch": 0.76, + "learning_rate": 5.63315701136041e-08, + "loss": 0.1102, + "step": 14607 + }, + { + "epoch": 0.76, + "learning_rate": 5.630804986629134e-08, + "loss": 0.1256, + "step": 14608 + }, + { + "epoch": 0.76, + "learning_rate": 5.628453372574997e-08, + "loss": 0.1289, + "step": 14609 + }, + { + "epoch": 0.76, + "learning_rate": 5.6261021692652034e-08, + "loss": 0.1275, + "step": 14610 + }, + { + "epoch": 0.76, + "learning_rate": 5.623751376766965e-08, + "loss": 0.1793, + "step": 14611 + }, + { + "epoch": 0.76, + "learning_rate": 5.621400995147465e-08, + "loss": 0.1248, + "step": 14612 + }, + { + "epoch": 0.76, + "learning_rate": 5.6190510244738735e-08, + "loss": 0.118, + "step": 14613 + }, + { + "epoch": 0.76, + "learning_rate": 5.616701464813354e-08, + "loss": 0.1288, + "step": 14614 + }, + { + "epoch": 0.76, + "learning_rate": 5.6143523162330595e-08, + "loss": 0.1503, + "step": 14615 + }, + { + "epoch": 0.76, + "learning_rate": 5.6120035788001306e-08, + "loss": 0.1447, + "step": 14616 + }, + { + "epoch": 0.76, + "learning_rate": 5.6096552525816863e-08, + "loss": 0.1035, + "step": 14617 + }, + { + "epoch": 0.76, + "learning_rate": 5.6073073376448575e-08, + "loss": 0.1127, + "step": 14618 + }, + { + "epoch": 0.76, + "learning_rate": 5.604959834056742e-08, + "loss": 0.1517, + "step": 14619 + }, + { + "epoch": 0.76, + "learning_rate": 5.602612741884429e-08, + "loss": 0.1262, + "step": 14620 + }, + { + "epoch": 0.76, + "learning_rate": 5.600266061194996e-08, + "loss": 0.1621, + "step": 14621 + }, + { + "epoch": 0.76, + "learning_rate": 5.597919792055532e-08, + "loss": 0.1432, + "step": 14622 + }, + { + "epoch": 0.76, + "learning_rate": 5.595573934533071e-08, + "loss": 0.1636, + "step": 14623 + }, + { + "epoch": 0.76, + "learning_rate": 5.59322848869467e-08, + "loss": 0.132, + "step": 14624 + }, + { + "epoch": 0.76, + "learning_rate": 5.590883454607353e-08, + "loss": 0.1608, + "step": 14625 + }, + { + "epoch": 0.76, + "learning_rate": 5.588538832338157e-08, + "loss": 0.1496, + "step": 14626 + }, + { + "epoch": 0.76, + "learning_rate": 5.586194621954083e-08, + "loss": 0.1607, + "step": 14627 + }, + { + "epoch": 0.76, + "learning_rate": 5.583850823522132e-08, + "loss": 0.1794, + "step": 14628 + }, + { + "epoch": 0.76, + "learning_rate": 5.581507437109283e-08, + "loss": 0.1555, + "step": 14629 + }, + { + "epoch": 0.76, + "learning_rate": 5.579164462782524e-08, + "loss": 0.1524, + "step": 14630 + }, + { + "epoch": 0.76, + "learning_rate": 5.576821900608817e-08, + "loss": 0.1146, + "step": 14631 + }, + { + "epoch": 0.76, + "learning_rate": 5.5744797506550967e-08, + "loss": 0.1494, + "step": 14632 + }, + { + "epoch": 0.76, + "learning_rate": 5.572138012988319e-08, + "loss": 0.1248, + "step": 14633 + }, + { + "epoch": 0.76, + "learning_rate": 5.569796687675408e-08, + "loss": 0.1404, + "step": 14634 + }, + { + "epoch": 0.76, + "learning_rate": 5.5674557747832785e-08, + "loss": 0.1651, + "step": 14635 + }, + { + "epoch": 0.76, + "learning_rate": 5.565115274378827e-08, + "loss": 0.1501, + "step": 14636 + }, + { + "epoch": 0.76, + "learning_rate": 5.562775186528959e-08, + "loss": 0.1645, + "step": 14637 + }, + { + "epoch": 0.76, + "learning_rate": 5.560435511300552e-08, + "loss": 0.0867, + "step": 14638 + }, + { + "epoch": 0.76, + "learning_rate": 5.558096248760469e-08, + "loss": 0.1451, + "step": 14639 + }, + { + "epoch": 0.76, + "learning_rate": 5.5557573989755736e-08, + "loss": 0.1725, + "step": 14640 + }, + { + "epoch": 0.76, + "learning_rate": 5.553418962012706e-08, + "loss": 0.1315, + "step": 14641 + }, + { + "epoch": 0.76, + "learning_rate": 5.551080937938701e-08, + "loss": 0.1481, + "step": 14642 + }, + { + "epoch": 0.76, + "learning_rate": 5.548743326820375e-08, + "loss": 0.1431, + "step": 14643 + }, + { + "epoch": 0.76, + "learning_rate": 5.54640612872455e-08, + "loss": 0.1315, + "step": 14644 + }, + { + "epoch": 0.76, + "learning_rate": 5.5440693437180165e-08, + "loss": 0.1343, + "step": 14645 + }, + { + "epoch": 0.76, + "learning_rate": 5.5417329718675634e-08, + "loss": 0.163, + "step": 14646 + }, + { + "epoch": 0.76, + "learning_rate": 5.539397013239957e-08, + "loss": 0.1389, + "step": 14647 + }, + { + "epoch": 0.76, + "learning_rate": 5.5370614679019776e-08, + "loss": 0.1266, + "step": 14648 + }, + { + "epoch": 0.76, + "learning_rate": 5.534726335920359e-08, + "loss": 0.1184, + "step": 14649 + }, + { + "epoch": 0.76, + "learning_rate": 5.532391617361838e-08, + "loss": 0.1235, + "step": 14650 + }, + { + "epoch": 0.76, + "learning_rate": 5.530057312293159e-08, + "loss": 0.1637, + "step": 14651 + }, + { + "epoch": 0.76, + "learning_rate": 5.527723420781025e-08, + "loss": 0.1521, + "step": 14652 + }, + { + "epoch": 0.76, + "learning_rate": 5.525389942892145e-08, + "loss": 0.1284, + "step": 14653 + }, + { + "epoch": 0.76, + "learning_rate": 5.5230568786932e-08, + "loss": 0.1258, + "step": 14654 + }, + { + "epoch": 0.76, + "learning_rate": 5.520724228250886e-08, + "loss": 0.1297, + "step": 14655 + }, + { + "epoch": 0.76, + "learning_rate": 5.518391991631868e-08, + "loss": 0.0862, + "step": 14656 + }, + { + "epoch": 0.77, + "learning_rate": 5.5160601689027853e-08, + "loss": 0.1222, + "step": 14657 + }, + { + "epoch": 0.77, + "learning_rate": 5.5137287601303006e-08, + "loss": 0.1294, + "step": 14658 + }, + { + "epoch": 0.77, + "learning_rate": 5.511397765381041e-08, + "loss": 0.1036, + "step": 14659 + }, + { + "epoch": 0.77, + "learning_rate": 5.5090671847216255e-08, + "loss": 0.1285, + "step": 14660 + }, + { + "epoch": 0.77, + "learning_rate": 5.506737018218657e-08, + "loss": 0.1451, + "step": 14661 + }, + { + "epoch": 0.77, + "learning_rate": 5.5044072659387465e-08, + "loss": 0.1456, + "step": 14662 + }, + { + "epoch": 0.77, + "learning_rate": 5.502077927948472e-08, + "loss": 0.1467, + "step": 14663 + }, + { + "epoch": 0.77, + "learning_rate": 5.499749004314409e-08, + "loss": 0.1208, + "step": 14664 + }, + { + "epoch": 0.77, + "learning_rate": 5.4974204951031155e-08, + "loss": 0.119, + "step": 14665 + }, + { + "epoch": 0.77, + "learning_rate": 5.4950924003811426e-08, + "loss": 0.1254, + "step": 14666 + }, + { + "epoch": 0.77, + "learning_rate": 5.4927647202150286e-08, + "loss": 0.151, + "step": 14667 + }, + { + "epoch": 0.77, + "learning_rate": 5.490437454671293e-08, + "loss": 0.1287, + "step": 14668 + }, + { + "epoch": 0.77, + "learning_rate": 5.4881106038164625e-08, + "loss": 0.1686, + "step": 14669 + }, + { + "epoch": 0.77, + "learning_rate": 5.4857841677170316e-08, + "loss": 0.1549, + "step": 14670 + }, + { + "epoch": 0.77, + "learning_rate": 5.483458146439495e-08, + "loss": 0.1288, + "step": 14671 + }, + { + "epoch": 0.77, + "learning_rate": 5.481132540050319e-08, + "loss": 0.157, + "step": 14672 + }, + { + "epoch": 0.77, + "learning_rate": 5.478807348615995e-08, + "loss": 0.1401, + "step": 14673 + }, + { + "epoch": 0.77, + "learning_rate": 5.476482572202956e-08, + "loss": 0.1686, + "step": 14674 + }, + { + "epoch": 0.77, + "learning_rate": 5.474158210877646e-08, + "loss": 0.137, + "step": 14675 + }, + { + "epoch": 0.77, + "learning_rate": 5.4718342647065075e-08, + "loss": 0.1318, + "step": 14676 + }, + { + "epoch": 0.77, + "learning_rate": 5.469510733755953e-08, + "loss": 0.1593, + "step": 14677 + }, + { + "epoch": 0.77, + "learning_rate": 5.467187618092392e-08, + "loss": 0.1518, + "step": 14678 + }, + { + "epoch": 0.77, + "learning_rate": 5.4648649177822126e-08, + "loss": 0.1353, + "step": 14679 + }, + { + "epoch": 0.77, + "learning_rate": 5.462542632891811e-08, + "loss": 0.2129, + "step": 14680 + }, + { + "epoch": 0.77, + "learning_rate": 5.460220763487555e-08, + "loss": 0.1294, + "step": 14681 + }, + { + "epoch": 0.77, + "learning_rate": 5.457899309635801e-08, + "loss": 0.135, + "step": 14682 + }, + { + "epoch": 0.77, + "learning_rate": 5.455578271402899e-08, + "loss": 0.2003, + "step": 14683 + }, + { + "epoch": 0.77, + "learning_rate": 5.453257648855183e-08, + "loss": 0.1546, + "step": 14684 + }, + { + "epoch": 0.77, + "learning_rate": 5.4509374420589805e-08, + "loss": 0.1062, + "step": 14685 + }, + { + "epoch": 0.77, + "learning_rate": 5.448617651080596e-08, + "loss": 0.169, + "step": 14686 + }, + { + "epoch": 0.77, + "learning_rate": 5.446298275986343e-08, + "loss": 0.15, + "step": 14687 + }, + { + "epoch": 0.77, + "learning_rate": 5.443979316842502e-08, + "loss": 0.1121, + "step": 14688 + }, + { + "epoch": 0.77, + "learning_rate": 5.441660773715353e-08, + "loss": 0.1471, + "step": 14689 + }, + { + "epoch": 0.77, + "learning_rate": 5.439342646671152e-08, + "loss": 0.1409, + "step": 14690 + }, + { + "epoch": 0.77, + "learning_rate": 5.4370249357761736e-08, + "loss": 0.1309, + "step": 14691 + }, + { + "epoch": 0.77, + "learning_rate": 5.4347076410966365e-08, + "loss": 0.1298, + "step": 14692 + }, + { + "epoch": 0.77, + "learning_rate": 5.432390762698771e-08, + "loss": 0.1206, + "step": 14693 + }, + { + "epoch": 0.77, + "learning_rate": 5.430074300648808e-08, + "loss": 0.1295, + "step": 14694 + }, + { + "epoch": 0.77, + "learning_rate": 5.427758255012947e-08, + "loss": 0.1186, + "step": 14695 + }, + { + "epoch": 0.77, + "learning_rate": 5.4254426258573795e-08, + "loss": 0.098, + "step": 14696 + }, + { + "epoch": 0.77, + "learning_rate": 5.423127413248281e-08, + "loss": 0.1235, + "step": 14697 + }, + { + "epoch": 0.77, + "learning_rate": 5.4208126172518355e-08, + "loss": 0.143, + "step": 14698 + }, + { + "epoch": 0.77, + "learning_rate": 5.418498237934199e-08, + "loss": 0.1388, + "step": 14699 + }, + { + "epoch": 0.77, + "learning_rate": 5.4161842753615015e-08, + "loss": 0.1484, + "step": 14700 + }, + { + "epoch": 0.77, + "learning_rate": 5.413870729599892e-08, + "loss": 0.1427, + "step": 14701 + }, + { + "epoch": 0.77, + "learning_rate": 5.4115576007154884e-08, + "loss": 0.1397, + "step": 14702 + }, + { + "epoch": 0.77, + "learning_rate": 5.409244888774398e-08, + "loss": 0.1441, + "step": 14703 + }, + { + "epoch": 0.77, + "learning_rate": 5.406932593842719e-08, + "loss": 0.1471, + "step": 14704 + }, + { + "epoch": 0.77, + "learning_rate": 5.4046207159865434e-08, + "loss": 0.1509, + "step": 14705 + }, + { + "epoch": 0.77, + "learning_rate": 5.402309255271942e-08, + "loss": 0.1291, + "step": 14706 + }, + { + "epoch": 0.77, + "learning_rate": 5.399998211764978e-08, + "loss": 0.146, + "step": 14707 + }, + { + "epoch": 0.77, + "learning_rate": 5.3976875855317026e-08, + "loss": 0.1182, + "step": 14708 + }, + { + "epoch": 0.77, + "learning_rate": 5.3953773766381526e-08, + "loss": 0.1269, + "step": 14709 + }, + { + "epoch": 0.77, + "learning_rate": 5.393067585150353e-08, + "loss": 0.1111, + "step": 14710 + }, + { + "epoch": 0.77, + "learning_rate": 5.3907582111343164e-08, + "loss": 0.1616, + "step": 14711 + }, + { + "epoch": 0.77, + "learning_rate": 5.3884492546560555e-08, + "loss": 0.1446, + "step": 14712 + }, + { + "epoch": 0.77, + "learning_rate": 5.3861407157815555e-08, + "loss": 0.1207, + "step": 14713 + }, + { + "epoch": 0.77, + "learning_rate": 5.383832594576796e-08, + "loss": 0.1248, + "step": 14714 + }, + { + "epoch": 0.77, + "learning_rate": 5.381524891107736e-08, + "loss": 0.155, + "step": 14715 + }, + { + "epoch": 0.77, + "learning_rate": 5.3792176054403516e-08, + "loss": 0.1238, + "step": 14716 + }, + { + "epoch": 0.77, + "learning_rate": 5.3769107376405653e-08, + "loss": 0.1206, + "step": 14717 + }, + { + "epoch": 0.77, + "learning_rate": 5.374604287774309e-08, + "loss": 0.14, + "step": 14718 + }, + { + "epoch": 0.77, + "learning_rate": 5.372298255907515e-08, + "loss": 0.1514, + "step": 14719 + }, + { + "epoch": 0.77, + "learning_rate": 5.3699926421060827e-08, + "loss": 0.113, + "step": 14720 + }, + { + "epoch": 0.77, + "learning_rate": 5.367687446435909e-08, + "loss": 0.1283, + "step": 14721 + }, + { + "epoch": 0.77, + "learning_rate": 5.36538266896287e-08, + "loss": 0.1599, + "step": 14722 + }, + { + "epoch": 0.77, + "learning_rate": 5.363078309752851e-08, + "loss": 0.1301, + "step": 14723 + }, + { + "epoch": 0.77, + "learning_rate": 5.360774368871703e-08, + "loss": 0.1622, + "step": 14724 + }, + { + "epoch": 0.77, + "learning_rate": 5.358470846385277e-08, + "loss": 0.1291, + "step": 14725 + }, + { + "epoch": 0.77, + "learning_rate": 5.3561677423594056e-08, + "loss": 0.1259, + "step": 14726 + }, + { + "epoch": 0.77, + "learning_rate": 5.353865056859912e-08, + "loss": 0.1678, + "step": 14727 + }, + { + "epoch": 0.77, + "learning_rate": 5.3515627899526085e-08, + "loss": 0.1296, + "step": 14728 + }, + { + "epoch": 0.77, + "learning_rate": 5.3492609417032905e-08, + "loss": 0.2029, + "step": 14729 + }, + { + "epoch": 0.77, + "learning_rate": 5.346959512177758e-08, + "loss": 0.1099, + "step": 14730 + }, + { + "epoch": 0.77, + "learning_rate": 5.3446585014417764e-08, + "loss": 0.1328, + "step": 14731 + }, + { + "epoch": 0.77, + "learning_rate": 5.342357909561115e-08, + "loss": 0.1293, + "step": 14732 + }, + { + "epoch": 0.77, + "learning_rate": 5.340057736601522e-08, + "loss": 0.1375, + "step": 14733 + }, + { + "epoch": 0.77, + "learning_rate": 5.337757982628737e-08, + "loss": 0.1386, + "step": 14734 + }, + { + "epoch": 0.77, + "learning_rate": 5.3354586477084906e-08, + "loss": 0.1394, + "step": 14735 + }, + { + "epoch": 0.77, + "learning_rate": 5.333159731906496e-08, + "loss": 0.1491, + "step": 14736 + }, + { + "epoch": 0.77, + "learning_rate": 5.330861235288453e-08, + "loss": 0.1515, + "step": 14737 + }, + { + "epoch": 0.77, + "learning_rate": 5.328563157920065e-08, + "loss": 0.1334, + "step": 14738 + }, + { + "epoch": 0.77, + "learning_rate": 5.3262654998670044e-08, + "loss": 0.1049, + "step": 14739 + }, + { + "epoch": 0.77, + "learning_rate": 5.323968261194942e-08, + "loss": 0.1372, + "step": 14740 + }, + { + "epoch": 0.77, + "learning_rate": 5.3216714419695244e-08, + "loss": 0.1271, + "step": 14741 + }, + { + "epoch": 0.77, + "learning_rate": 5.3193750422564176e-08, + "loss": 0.1532, + "step": 14742 + }, + { + "epoch": 0.77, + "learning_rate": 5.317079062121232e-08, + "loss": 0.122, + "step": 14743 + }, + { + "epoch": 0.77, + "learning_rate": 5.314783501629589e-08, + "loss": 0.1478, + "step": 14744 + }, + { + "epoch": 0.77, + "learning_rate": 5.3124883608471094e-08, + "loss": 0.1323, + "step": 14745 + }, + { + "epoch": 0.77, + "learning_rate": 5.31019363983938e-08, + "loss": 0.0879, + "step": 14746 + }, + { + "epoch": 0.77, + "learning_rate": 5.30789933867199e-08, + "loss": 0.1364, + "step": 14747 + }, + { + "epoch": 0.77, + "learning_rate": 5.305605457410503e-08, + "loss": 0.1276, + "step": 14748 + }, + { + "epoch": 0.77, + "learning_rate": 5.303311996120487e-08, + "loss": 0.1365, + "step": 14749 + }, + { + "epoch": 0.77, + "learning_rate": 5.3010189548674975e-08, + "loss": 0.1311, + "step": 14750 + }, + { + "epoch": 0.77, + "learning_rate": 5.298726333717045e-08, + "loss": 0.1389, + "step": 14751 + }, + { + "epoch": 0.77, + "learning_rate": 5.296434132734677e-08, + "loss": 0.1396, + "step": 14752 + }, + { + "epoch": 0.77, + "learning_rate": 5.2941423519859e-08, + "loss": 0.1343, + "step": 14753 + }, + { + "epoch": 0.77, + "learning_rate": 5.291850991536209e-08, + "loss": 0.1185, + "step": 14754 + }, + { + "epoch": 0.77, + "learning_rate": 5.28956005145109e-08, + "loss": 0.125, + "step": 14755 + }, + { + "epoch": 0.77, + "learning_rate": 5.287269531796028e-08, + "loss": 0.1856, + "step": 14756 + }, + { + "epoch": 0.77, + "learning_rate": 5.2849794326364826e-08, + "loss": 0.1327, + "step": 14757 + }, + { + "epoch": 0.77, + "learning_rate": 5.282689754037906e-08, + "loss": 0.1201, + "step": 14758 + }, + { + "epoch": 0.77, + "learning_rate": 5.280400496065738e-08, + "loss": 0.1288, + "step": 14759 + }, + { + "epoch": 0.77, + "learning_rate": 5.2781116587854046e-08, + "loss": 0.1605, + "step": 14760 + }, + { + "epoch": 0.77, + "learning_rate": 5.275823242262325e-08, + "loss": 0.1539, + "step": 14761 + }, + { + "epoch": 0.77, + "learning_rate": 5.2735352465618956e-08, + "loss": 0.1183, + "step": 14762 + }, + { + "epoch": 0.77, + "learning_rate": 5.271247671749517e-08, + "loss": 0.1075, + "step": 14763 + }, + { + "epoch": 0.77, + "learning_rate": 5.268960517890568e-08, + "loss": 0.0896, + "step": 14764 + }, + { + "epoch": 0.77, + "learning_rate": 5.266673785050415e-08, + "loss": 0.1073, + "step": 14765 + }, + { + "epoch": 0.77, + "learning_rate": 5.264387473294405e-08, + "loss": 0.1175, + "step": 14766 + }, + { + "epoch": 0.77, + "learning_rate": 5.262101582687897e-08, + "loss": 0.1236, + "step": 14767 + }, + { + "epoch": 0.77, + "learning_rate": 5.2598161132962203e-08, + "loss": 0.15, + "step": 14768 + }, + { + "epoch": 0.77, + "learning_rate": 5.257531065184677e-08, + "loss": 0.1409, + "step": 14769 + }, + { + "epoch": 0.77, + "learning_rate": 5.2552464384185943e-08, + "loss": 0.1459, + "step": 14770 + }, + { + "epoch": 0.77, + "learning_rate": 5.252962233063258e-08, + "loss": 0.1398, + "step": 14771 + }, + { + "epoch": 0.77, + "learning_rate": 5.250678449183956e-08, + "loss": 0.1207, + "step": 14772 + }, + { + "epoch": 0.77, + "learning_rate": 5.248395086845951e-08, + "loss": 0.1076, + "step": 14773 + }, + { + "epoch": 0.77, + "learning_rate": 5.246112146114516e-08, + "loss": 0.1372, + "step": 14774 + }, + { + "epoch": 0.77, + "learning_rate": 5.243829627054892e-08, + "loss": 0.1947, + "step": 14775 + }, + { + "epoch": 0.77, + "learning_rate": 5.241547529732313e-08, + "loss": 0.1635, + "step": 14776 + }, + { + "epoch": 0.77, + "learning_rate": 5.239265854212003e-08, + "loss": 0.1477, + "step": 14777 + }, + { + "epoch": 0.77, + "learning_rate": 5.236984600559171e-08, + "loss": 0.1325, + "step": 14778 + }, + { + "epoch": 0.77, + "learning_rate": 5.2347037688390215e-08, + "loss": 0.1378, + "step": 14779 + }, + { + "epoch": 0.77, + "learning_rate": 5.232423359116729e-08, + "loss": 0.1695, + "step": 14780 + }, + { + "epoch": 0.77, + "learning_rate": 5.230143371457487e-08, + "loss": 0.1208, + "step": 14781 + }, + { + "epoch": 0.77, + "learning_rate": 5.2278638059264465e-08, + "loss": 0.128, + "step": 14782 + }, + { + "epoch": 0.77, + "learning_rate": 5.2255846625887625e-08, + "loss": 0.1579, + "step": 14783 + }, + { + "epoch": 0.77, + "learning_rate": 5.223305941509566e-08, + "loss": 0.1246, + "step": 14784 + }, + { + "epoch": 0.77, + "learning_rate": 5.221027642754001e-08, + "loss": 0.1097, + "step": 14785 + }, + { + "epoch": 0.77, + "learning_rate": 5.2187497663871673e-08, + "loss": 0.1139, + "step": 14786 + }, + { + "epoch": 0.77, + "learning_rate": 5.216472312474165e-08, + "loss": 0.1833, + "step": 14787 + }, + { + "epoch": 0.77, + "learning_rate": 5.2141952810800957e-08, + "loss": 0.1008, + "step": 14788 + }, + { + "epoch": 0.77, + "learning_rate": 5.211918672270035e-08, + "loss": 0.1458, + "step": 14789 + }, + { + "epoch": 0.77, + "learning_rate": 5.2096424861090447e-08, + "loss": 0.1633, + "step": 14790 + }, + { + "epoch": 0.77, + "learning_rate": 5.207366722662179e-08, + "loss": 0.1172, + "step": 14791 + }, + { + "epoch": 0.77, + "learning_rate": 5.205091381994486e-08, + "loss": 0.1396, + "step": 14792 + }, + { + "epoch": 0.77, + "learning_rate": 5.2028164641710004e-08, + "loss": 0.1603, + "step": 14793 + }, + { + "epoch": 0.77, + "learning_rate": 5.20054196925672e-08, + "loss": 0.121, + "step": 14794 + }, + { + "epoch": 0.77, + "learning_rate": 5.198267897316671e-08, + "loss": 0.1274, + "step": 14795 + }, + { + "epoch": 0.77, + "learning_rate": 5.195994248415836e-08, + "loss": 0.1515, + "step": 14796 + }, + { + "epoch": 0.77, + "learning_rate": 5.193721022619204e-08, + "loss": 0.1385, + "step": 14797 + }, + { + "epoch": 0.77, + "learning_rate": 5.191448219991732e-08, + "loss": 0.1398, + "step": 14798 + }, + { + "epoch": 0.77, + "learning_rate": 5.1891758405983945e-08, + "loss": 0.1073, + "step": 14799 + }, + { + "epoch": 0.77, + "learning_rate": 5.186903884504128e-08, + "loss": 0.1298, + "step": 14800 + }, + { + "epoch": 0.77, + "learning_rate": 5.184632351773868e-08, + "loss": 0.1227, + "step": 14801 + }, + { + "epoch": 0.77, + "learning_rate": 5.1823612424725325e-08, + "loss": 0.1333, + "step": 14802 + }, + { + "epoch": 0.77, + "learning_rate": 5.180090556665035e-08, + "loss": 0.1244, + "step": 14803 + }, + { + "epoch": 0.77, + "learning_rate": 5.1778202944162705e-08, + "loss": 0.1414, + "step": 14804 + }, + { + "epoch": 0.77, + "learning_rate": 5.175550455791116e-08, + "loss": 0.1292, + "step": 14805 + }, + { + "epoch": 0.77, + "learning_rate": 5.17328104085446e-08, + "loss": 0.1235, + "step": 14806 + }, + { + "epoch": 0.77, + "learning_rate": 5.171012049671155e-08, + "loss": 0.1906, + "step": 14807 + }, + { + "epoch": 0.77, + "learning_rate": 5.16874348230605e-08, + "loss": 0.1153, + "step": 14808 + }, + { + "epoch": 0.77, + "learning_rate": 5.1664753388239765e-08, + "loss": 0.1115, + "step": 14809 + }, + { + "epoch": 0.77, + "learning_rate": 5.1642076192897756e-08, + "loss": 0.1322, + "step": 14810 + }, + { + "epoch": 0.77, + "learning_rate": 5.1619403237682415e-08, + "loss": 0.1033, + "step": 14811 + }, + { + "epoch": 0.77, + "learning_rate": 5.1596734523241755e-08, + "loss": 0.1293, + "step": 14812 + }, + { + "epoch": 0.77, + "learning_rate": 5.1574070050223785e-08, + "loss": 0.1205, + "step": 14813 + }, + { + "epoch": 0.77, + "learning_rate": 5.155140981927617e-08, + "loss": 0.1231, + "step": 14814 + }, + { + "epoch": 0.77, + "learning_rate": 5.152875383104658e-08, + "loss": 0.1452, + "step": 14815 + }, + { + "epoch": 0.77, + "learning_rate": 5.1506102086182443e-08, + "loss": 0.1028, + "step": 14816 + }, + { + "epoch": 0.77, + "learning_rate": 5.148345458533132e-08, + "loss": 0.1365, + "step": 14817 + }, + { + "epoch": 0.77, + "learning_rate": 5.1460811329140396e-08, + "loss": 0.1323, + "step": 14818 + }, + { + "epoch": 0.77, + "learning_rate": 5.1438172318256803e-08, + "loss": 0.1422, + "step": 14819 + }, + { + "epoch": 0.77, + "learning_rate": 5.141553755332762e-08, + "loss": 0.1419, + "step": 14820 + }, + { + "epoch": 0.77, + "learning_rate": 5.139290703499972e-08, + "loss": 0.1208, + "step": 14821 + }, + { + "epoch": 0.77, + "learning_rate": 5.1370280763919917e-08, + "loss": 0.1481, + "step": 14822 + }, + { + "epoch": 0.77, + "learning_rate": 5.134765874073479e-08, + "loss": 0.122, + "step": 14823 + }, + { + "epoch": 0.77, + "learning_rate": 5.132504096609103e-08, + "loss": 0.1417, + "step": 14824 + }, + { + "epoch": 0.77, + "learning_rate": 5.1302427440635e-08, + "loss": 0.1367, + "step": 14825 + }, + { + "epoch": 0.77, + "learning_rate": 5.127981816501299e-08, + "loss": 0.1314, + "step": 14826 + }, + { + "epoch": 0.77, + "learning_rate": 5.1257213139871125e-08, + "loss": 0.1221, + "step": 14827 + }, + { + "epoch": 0.77, + "learning_rate": 5.123461236585564e-08, + "loss": 0.1352, + "step": 14828 + }, + { + "epoch": 0.77, + "learning_rate": 5.121201584361232e-08, + "loss": 0.1234, + "step": 14829 + }, + { + "epoch": 0.77, + "learning_rate": 5.118942357378697e-08, + "loss": 0.1713, + "step": 14830 + }, + { + "epoch": 0.77, + "learning_rate": 5.116683555702539e-08, + "loss": 0.1694, + "step": 14831 + }, + { + "epoch": 0.77, + "learning_rate": 5.1144251793973105e-08, + "loss": 0.1223, + "step": 14832 + }, + { + "epoch": 0.77, + "learning_rate": 5.1121672285275576e-08, + "loss": 0.1312, + "step": 14833 + }, + { + "epoch": 0.77, + "learning_rate": 5.1099097031578044e-08, + "loss": 0.0888, + "step": 14834 + }, + { + "epoch": 0.77, + "learning_rate": 5.107652603352588e-08, + "loss": 0.1362, + "step": 14835 + }, + { + "epoch": 0.77, + "learning_rate": 5.105395929176415e-08, + "loss": 0.1565, + "step": 14836 + }, + { + "epoch": 0.77, + "learning_rate": 5.1031396806937646e-08, + "loss": 0.1298, + "step": 14837 + }, + { + "epoch": 0.77, + "learning_rate": 5.100883857969138e-08, + "loss": 0.1408, + "step": 14838 + }, + { + "epoch": 0.77, + "learning_rate": 5.0986284610670027e-08, + "loss": 0.1366, + "step": 14839 + }, + { + "epoch": 0.77, + "learning_rate": 5.0963734900518175e-08, + "loss": 0.1611, + "step": 14840 + }, + { + "epoch": 0.77, + "learning_rate": 5.0941189449880265e-08, + "loss": 0.1249, + "step": 14841 + }, + { + "epoch": 0.77, + "learning_rate": 5.091864825940075e-08, + "loss": 0.1452, + "step": 14842 + }, + { + "epoch": 0.77, + "learning_rate": 5.089611132972385e-08, + "loss": 0.1125, + "step": 14843 + }, + { + "epoch": 0.77, + "learning_rate": 5.0873578661493623e-08, + "loss": 0.118, + "step": 14844 + }, + { + "epoch": 0.77, + "learning_rate": 5.0851050255354076e-08, + "loss": 0.108, + "step": 14845 + }, + { + "epoch": 0.77, + "learning_rate": 5.082852611194912e-08, + "loss": 0.1571, + "step": 14846 + }, + { + "epoch": 0.77, + "learning_rate": 5.080600623192246e-08, + "loss": 0.1119, + "step": 14847 + }, + { + "epoch": 0.77, + "learning_rate": 5.0783490615917734e-08, + "loss": 0.1606, + "step": 14848 + }, + { + "epoch": 0.78, + "learning_rate": 5.076097926457839e-08, + "loss": 0.123, + "step": 14849 + }, + { + "epoch": 0.78, + "learning_rate": 5.073847217854792e-08, + "loss": 0.1604, + "step": 14850 + }, + { + "epoch": 0.78, + "learning_rate": 5.071596935846956e-08, + "loss": 0.1396, + "step": 14851 + }, + { + "epoch": 0.78, + "learning_rate": 5.0693470804986404e-08, + "loss": 0.2012, + "step": 14852 + }, + { + "epoch": 0.78, + "learning_rate": 5.06709765187415e-08, + "loss": 0.1327, + "step": 14853 + }, + { + "epoch": 0.78, + "learning_rate": 5.064848650037774e-08, + "loss": 0.1079, + "step": 14854 + }, + { + "epoch": 0.78, + "learning_rate": 5.062600075053789e-08, + "loss": 0.1205, + "step": 14855 + }, + { + "epoch": 0.78, + "learning_rate": 5.060351926986453e-08, + "loss": 0.1499, + "step": 14856 + }, + { + "epoch": 0.78, + "learning_rate": 5.058104205900033e-08, + "loss": 0.1254, + "step": 14857 + }, + { + "epoch": 0.78, + "learning_rate": 5.0558569118587625e-08, + "loss": 0.1517, + "step": 14858 + }, + { + "epoch": 0.78, + "learning_rate": 5.0536100449268706e-08, + "loss": 0.1286, + "step": 14859 + }, + { + "epoch": 0.78, + "learning_rate": 5.051363605168568e-08, + "loss": 0.1258, + "step": 14860 + }, + { + "epoch": 0.78, + "learning_rate": 5.049117592648071e-08, + "loss": 0.1095, + "step": 14861 + }, + { + "epoch": 0.78, + "learning_rate": 5.04687200742957e-08, + "loss": 0.1211, + "step": 14862 + }, + { + "epoch": 0.78, + "learning_rate": 5.044626849577227e-08, + "loss": 0.1216, + "step": 14863 + }, + { + "epoch": 0.78, + "learning_rate": 5.042382119155229e-08, + "loss": 0.1261, + "step": 14864 + }, + { + "epoch": 0.78, + "learning_rate": 5.040137816227723e-08, + "loss": 0.1051, + "step": 14865 + }, + { + "epoch": 0.78, + "learning_rate": 5.037893940858855e-08, + "loss": 0.1105, + "step": 14866 + }, + { + "epoch": 0.78, + "learning_rate": 5.035650493112746e-08, + "loss": 0.1303, + "step": 14867 + }, + { + "epoch": 0.78, + "learning_rate": 5.03340747305353e-08, + "loss": 0.1329, + "step": 14868 + }, + { + "epoch": 0.78, + "learning_rate": 5.0311648807453064e-08, + "loss": 0.1179, + "step": 14869 + }, + { + "epoch": 0.78, + "learning_rate": 5.0289227162521685e-08, + "loss": 0.1247, + "step": 14870 + }, + { + "epoch": 0.78, + "learning_rate": 5.026680979638198e-08, + "loss": 0.1247, + "step": 14871 + }, + { + "epoch": 0.78, + "learning_rate": 5.024439670967466e-08, + "loss": 0.1362, + "step": 14872 + }, + { + "epoch": 0.78, + "learning_rate": 5.022198790304031e-08, + "loss": 0.1101, + "step": 14873 + }, + { + "epoch": 0.78, + "learning_rate": 5.019958337711929e-08, + "loss": 0.133, + "step": 14874 + }, + { + "epoch": 0.78, + "learning_rate": 5.017718313255206e-08, + "loss": 0.1363, + "step": 14875 + }, + { + "epoch": 0.78, + "learning_rate": 5.015478716997878e-08, + "loss": 0.1349, + "step": 14876 + }, + { + "epoch": 0.78, + "learning_rate": 5.0132395490039535e-08, + "loss": 0.1421, + "step": 14877 + }, + { + "epoch": 0.78, + "learning_rate": 5.0110008093374225e-08, + "loss": 0.1333, + "step": 14878 + }, + { + "epoch": 0.78, + "learning_rate": 5.0087624980622845e-08, + "loss": 0.1483, + "step": 14879 + }, + { + "epoch": 0.78, + "learning_rate": 5.006524615242498e-08, + "loss": 0.1757, + "step": 14880 + }, + { + "epoch": 0.78, + "learning_rate": 5.0042871609420177e-08, + "loss": 0.1546, + "step": 14881 + }, + { + "epoch": 0.78, + "learning_rate": 5.0020501352248066e-08, + "loss": 0.1149, + "step": 14882 + }, + { + "epoch": 0.78, + "learning_rate": 4.999813538154793e-08, + "loss": 0.1622, + "step": 14883 + }, + { + "epoch": 0.78, + "learning_rate": 4.997577369795898e-08, + "loss": 0.1281, + "step": 14884 + }, + { + "epoch": 0.78, + "learning_rate": 4.9953416302120285e-08, + "loss": 0.1164, + "step": 14885 + }, + { + "epoch": 0.78, + "learning_rate": 4.993106319467093e-08, + "loss": 0.1139, + "step": 14886 + }, + { + "epoch": 0.78, + "learning_rate": 4.990871437624971e-08, + "loss": 0.1435, + "step": 14887 + }, + { + "epoch": 0.78, + "learning_rate": 4.988636984749539e-08, + "loss": 0.1315, + "step": 14888 + }, + { + "epoch": 0.78, + "learning_rate": 4.986402960904656e-08, + "loss": 0.1233, + "step": 14889 + }, + { + "epoch": 0.78, + "learning_rate": 4.984169366154172e-08, + "loss": 0.111, + "step": 14890 + }, + { + "epoch": 0.78, + "learning_rate": 4.981936200561925e-08, + "loss": 0.1508, + "step": 14891 + }, + { + "epoch": 0.78, + "learning_rate": 4.979703464191734e-08, + "loss": 0.1203, + "step": 14892 + }, + { + "epoch": 0.78, + "learning_rate": 4.9774711571074203e-08, + "loss": 0.1506, + "step": 14893 + }, + { + "epoch": 0.78, + "learning_rate": 4.97523927937278e-08, + "loss": 0.1296, + "step": 14894 + }, + { + "epoch": 0.78, + "learning_rate": 4.973007831051603e-08, + "loss": 0.136, + "step": 14895 + }, + { + "epoch": 0.78, + "learning_rate": 4.970776812207662e-08, + "loss": 0.1261, + "step": 14896 + }, + { + "epoch": 0.78, + "learning_rate": 4.968546222904722e-08, + "loss": 0.1327, + "step": 14897 + }, + { + "epoch": 0.78, + "learning_rate": 4.9663160632065346e-08, + "loss": 0.1582, + "step": 14898 + }, + { + "epoch": 0.78, + "learning_rate": 4.9640863331768314e-08, + "loss": 0.1078, + "step": 14899 + }, + { + "epoch": 0.78, + "learning_rate": 4.9618570328793485e-08, + "loss": 0.1129, + "step": 14900 + }, + { + "epoch": 0.78, + "learning_rate": 4.9596281623778e-08, + "loss": 0.1373, + "step": 14901 + }, + { + "epoch": 0.78, + "learning_rate": 4.9573997217358845e-08, + "loss": 0.1748, + "step": 14902 + }, + { + "epoch": 0.78, + "learning_rate": 4.955171711017285e-08, + "loss": 0.1533, + "step": 14903 + }, + { + "epoch": 0.78, + "learning_rate": 4.952944130285692e-08, + "loss": 0.1119, + "step": 14904 + }, + { + "epoch": 0.78, + "learning_rate": 4.95071697960477e-08, + "loss": 0.1699, + "step": 14905 + }, + { + "epoch": 0.78, + "learning_rate": 4.948490259038156e-08, + "loss": 0.2008, + "step": 14906 + }, + { + "epoch": 0.78, + "learning_rate": 4.946263968649504e-08, + "loss": 0.1225, + "step": 14907 + }, + { + "epoch": 0.78, + "learning_rate": 4.9440381085024417e-08, + "loss": 0.1209, + "step": 14908 + }, + { + "epoch": 0.78, + "learning_rate": 4.9418126786605815e-08, + "loss": 0.1194, + "step": 14909 + }, + { + "epoch": 0.78, + "learning_rate": 4.939587679187522e-08, + "loss": 0.1723, + "step": 14910 + }, + { + "epoch": 0.78, + "learning_rate": 4.9373631101468685e-08, + "loss": 0.12, + "step": 14911 + }, + { + "epoch": 0.78, + "learning_rate": 4.935138971602191e-08, + "loss": 0.111, + "step": 14912 + }, + { + "epoch": 0.78, + "learning_rate": 4.932915263617055e-08, + "loss": 0.1263, + "step": 14913 + }, + { + "epoch": 0.78, + "learning_rate": 4.93069198625502e-08, + "loss": 0.1687, + "step": 14914 + }, + { + "epoch": 0.78, + "learning_rate": 4.928469139579623e-08, + "loss": 0.1085, + "step": 14915 + }, + { + "epoch": 0.78, + "learning_rate": 4.926246723654397e-08, + "loss": 0.1282, + "step": 14916 + }, + { + "epoch": 0.78, + "learning_rate": 4.924024738542854e-08, + "loss": 0.1244, + "step": 14917 + }, + { + "epoch": 0.78, + "learning_rate": 4.921803184308508e-08, + "loss": 0.1498, + "step": 14918 + }, + { + "epoch": 0.78, + "learning_rate": 4.9195820610148505e-08, + "loss": 0.1481, + "step": 14919 + }, + { + "epoch": 0.78, + "learning_rate": 4.917361368725357e-08, + "loss": 0.1355, + "step": 14920 + }, + { + "epoch": 0.78, + "learning_rate": 4.915141107503493e-08, + "loss": 0.1491, + "step": 14921 + }, + { + "epoch": 0.78, + "learning_rate": 4.912921277412732e-08, + "loss": 0.1192, + "step": 14922 + }, + { + "epoch": 0.78, + "learning_rate": 4.910701878516497e-08, + "loss": 0.1231, + "step": 14923 + }, + { + "epoch": 0.78, + "learning_rate": 4.908482910878222e-08, + "loss": 0.1025, + "step": 14924 + }, + { + "epoch": 0.78, + "learning_rate": 4.9062643745613374e-08, + "loss": 0.1731, + "step": 14925 + }, + { + "epoch": 0.78, + "learning_rate": 4.904046269629245e-08, + "loss": 0.14, + "step": 14926 + }, + { + "epoch": 0.78, + "learning_rate": 4.9018285961453365e-08, + "loss": 0.1576, + "step": 14927 + }, + { + "epoch": 0.78, + "learning_rate": 4.89961135417299e-08, + "loss": 0.1374, + "step": 14928 + }, + { + "epoch": 0.78, + "learning_rate": 4.8973945437755856e-08, + "loss": 0.1432, + "step": 14929 + }, + { + "epoch": 0.78, + "learning_rate": 4.895178165016482e-08, + "loss": 0.2091, + "step": 14930 + }, + { + "epoch": 0.78, + "learning_rate": 4.892962217959007e-08, + "loss": 0.1618, + "step": 14931 + }, + { + "epoch": 0.78, + "learning_rate": 4.890746702666508e-08, + "loss": 0.1526, + "step": 14932 + }, + { + "epoch": 0.78, + "learning_rate": 4.888531619202301e-08, + "loss": 0.1567, + "step": 14933 + }, + { + "epoch": 0.78, + "learning_rate": 4.8863169676296936e-08, + "loss": 0.1248, + "step": 14934 + }, + { + "epoch": 0.78, + "learning_rate": 4.8841027480119754e-08, + "loss": 0.1212, + "step": 14935 + }, + { + "epoch": 0.78, + "learning_rate": 4.8818889604124435e-08, + "loss": 0.121, + "step": 14936 + }, + { + "epoch": 0.78, + "learning_rate": 4.87967560489436e-08, + "loss": 0.1225, + "step": 14937 + }, + { + "epoch": 0.78, + "learning_rate": 4.877462681520985e-08, + "loss": 0.1076, + "step": 14938 + }, + { + "epoch": 0.78, + "learning_rate": 4.8752501903555664e-08, + "loss": 0.1327, + "step": 14939 + }, + { + "epoch": 0.78, + "learning_rate": 4.8730381314613334e-08, + "loss": 0.1301, + "step": 14940 + }, + { + "epoch": 0.78, + "learning_rate": 4.870826504901511e-08, + "loss": 0.1255, + "step": 14941 + }, + { + "epoch": 0.78, + "learning_rate": 4.8686153107393014e-08, + "loss": 0.1445, + "step": 14942 + }, + { + "epoch": 0.78, + "learning_rate": 4.8664045490379146e-08, + "loss": 0.1585, + "step": 14943 + }, + { + "epoch": 0.78, + "learning_rate": 4.8641942198605265e-08, + "loss": 0.137, + "step": 14944 + }, + { + "epoch": 0.78, + "learning_rate": 4.861984323270312e-08, + "loss": 0.1454, + "step": 14945 + }, + { + "epoch": 0.78, + "learning_rate": 4.859774859330423e-08, + "loss": 0.1292, + "step": 14946 + }, + { + "epoch": 0.78, + "learning_rate": 4.857565828104018e-08, + "loss": 0.1262, + "step": 14947 + }, + { + "epoch": 0.78, + "learning_rate": 4.855357229654236e-08, + "loss": 0.2001, + "step": 14948 + }, + { + "epoch": 0.78, + "learning_rate": 4.8531490640441754e-08, + "loss": 0.1408, + "step": 14949 + }, + { + "epoch": 0.78, + "learning_rate": 4.850941331336971e-08, + "loss": 0.0967, + "step": 14950 + }, + { + "epoch": 0.78, + "learning_rate": 4.848734031595709e-08, + "loss": 0.1485, + "step": 14951 + }, + { + "epoch": 0.78, + "learning_rate": 4.846527164883478e-08, + "loss": 0.1041, + "step": 14952 + }, + { + "epoch": 0.78, + "learning_rate": 4.844320731263345e-08, + "loss": 0.1133, + "step": 14953 + }, + { + "epoch": 0.78, + "learning_rate": 4.842114730798381e-08, + "loss": 0.1383, + "step": 14954 + }, + { + "epoch": 0.78, + "learning_rate": 4.8399091635516295e-08, + "loss": 0.1269, + "step": 14955 + }, + { + "epoch": 0.78, + "learning_rate": 4.837704029586134e-08, + "loss": 0.1653, + "step": 14956 + }, + { + "epoch": 0.78, + "learning_rate": 4.8354993289648984e-08, + "loss": 0.1244, + "step": 14957 + }, + { + "epoch": 0.78, + "learning_rate": 4.83329506175095e-08, + "loss": 0.1175, + "step": 14958 + }, + { + "epoch": 0.78, + "learning_rate": 4.831091228007287e-08, + "loss": 0.1556, + "step": 14959 + }, + { + "epoch": 0.78, + "learning_rate": 4.82888782779689e-08, + "loss": 0.1652, + "step": 14960 + }, + { + "epoch": 0.78, + "learning_rate": 4.8266848611827306e-08, + "loss": 0.128, + "step": 14961 + }, + { + "epoch": 0.78, + "learning_rate": 4.824482328227781e-08, + "loss": 0.1442, + "step": 14962 + }, + { + "epoch": 0.78, + "learning_rate": 4.822280228994984e-08, + "loss": 0.1426, + "step": 14963 + }, + { + "epoch": 0.78, + "learning_rate": 4.820078563547279e-08, + "loss": 0.1436, + "step": 14964 + }, + { + "epoch": 0.78, + "learning_rate": 4.817877331947589e-08, + "loss": 0.1355, + "step": 14965 + }, + { + "epoch": 0.78, + "learning_rate": 4.815676534258823e-08, + "loss": 0.1258, + "step": 14966 + }, + { + "epoch": 0.78, + "learning_rate": 4.8134761705438844e-08, + "loss": 0.139, + "step": 14967 + }, + { + "epoch": 0.78, + "learning_rate": 4.811276240865654e-08, + "loss": 0.1312, + "step": 14968 + }, + { + "epoch": 0.78, + "learning_rate": 4.8090767452870194e-08, + "loss": 0.1046, + "step": 14969 + }, + { + "epoch": 0.78, + "learning_rate": 4.8068776838708334e-08, + "loss": 0.1121, + "step": 14970 + }, + { + "epoch": 0.78, + "learning_rate": 4.80467905667995e-08, + "loss": 0.1094, + "step": 14971 + }, + { + "epoch": 0.78, + "learning_rate": 4.802480863777199e-08, + "loss": 0.1092, + "step": 14972 + }, + { + "epoch": 0.78, + "learning_rate": 4.800283105225422e-08, + "loss": 0.1191, + "step": 14973 + }, + { + "epoch": 0.78, + "learning_rate": 4.798085781087415e-08, + "loss": 0.1102, + "step": 14974 + }, + { + "epoch": 0.78, + "learning_rate": 4.795888891425981e-08, + "loss": 0.1434, + "step": 14975 + }, + { + "epoch": 0.78, + "learning_rate": 4.793692436303918e-08, + "loss": 0.1435, + "step": 14976 + }, + { + "epoch": 0.78, + "learning_rate": 4.7914964157839955e-08, + "loss": 0.1371, + "step": 14977 + }, + { + "epoch": 0.78, + "learning_rate": 4.789300829928975e-08, + "loss": 0.134, + "step": 14978 + }, + { + "epoch": 0.78, + "learning_rate": 4.787105678801604e-08, + "loss": 0.1018, + "step": 14979 + }, + { + "epoch": 0.78, + "learning_rate": 4.78491096246463e-08, + "loss": 0.1532, + "step": 14980 + }, + { + "epoch": 0.78, + "learning_rate": 4.7827166809807764e-08, + "loss": 0.1263, + "step": 14981 + }, + { + "epoch": 0.78, + "learning_rate": 4.7805228344127527e-08, + "loss": 0.1522, + "step": 14982 + }, + { + "epoch": 0.78, + "learning_rate": 4.778329422823262e-08, + "loss": 0.1487, + "step": 14983 + }, + { + "epoch": 0.78, + "learning_rate": 4.7761364462749945e-08, + "loss": 0.1261, + "step": 14984 + }, + { + "epoch": 0.78, + "learning_rate": 4.7739439048306215e-08, + "loss": 0.1381, + "step": 14985 + }, + { + "epoch": 0.78, + "learning_rate": 4.7717517985528054e-08, + "loss": 0.1088, + "step": 14986 + }, + { + "epoch": 0.78, + "learning_rate": 4.769560127504206e-08, + "loss": 0.1257, + "step": 14987 + }, + { + "epoch": 0.78, + "learning_rate": 4.767368891747461e-08, + "loss": 0.1526, + "step": 14988 + }, + { + "epoch": 0.78, + "learning_rate": 4.76517809134519e-08, + "loss": 0.1027, + "step": 14989 + }, + { + "epoch": 0.78, + "learning_rate": 4.762987726360013e-08, + "loss": 0.1706, + "step": 14990 + }, + { + "epoch": 0.78, + "learning_rate": 4.760797796854525e-08, + "loss": 0.1464, + "step": 14991 + }, + { + "epoch": 0.78, + "learning_rate": 4.758608302891323e-08, + "loss": 0.1213, + "step": 14992 + }, + { + "epoch": 0.78, + "learning_rate": 4.756419244532972e-08, + "loss": 0.1427, + "step": 14993 + }, + { + "epoch": 0.78, + "learning_rate": 4.754230621842048e-08, + "loss": 0.1499, + "step": 14994 + }, + { + "epoch": 0.78, + "learning_rate": 4.752042434881101e-08, + "loss": 0.1009, + "step": 14995 + }, + { + "epoch": 0.78, + "learning_rate": 4.749854683712666e-08, + "loss": 0.1315, + "step": 14996 + }, + { + "epoch": 0.78, + "learning_rate": 4.7476673683992665e-08, + "loss": 0.1165, + "step": 14997 + }, + { + "epoch": 0.78, + "learning_rate": 4.745480489003426e-08, + "loss": 0.1732, + "step": 14998 + }, + { + "epoch": 0.78, + "learning_rate": 4.743294045587647e-08, + "loss": 0.1224, + "step": 14999 + }, + { + "epoch": 0.78, + "learning_rate": 4.7411080382144055e-08, + "loss": 0.1585, + "step": 15000 + }, + { + "epoch": 0.78, + "learning_rate": 4.7389224669461894e-08, + "loss": 0.1261, + "step": 15001 + }, + { + "epoch": 0.78, + "learning_rate": 4.7367373318454616e-08, + "loss": 0.119, + "step": 15002 + }, + { + "epoch": 0.78, + "learning_rate": 4.734552632974673e-08, + "loss": 0.1012, + "step": 15003 + }, + { + "epoch": 0.78, + "learning_rate": 4.732368370396258e-08, + "loss": 0.1188, + "step": 15004 + }, + { + "epoch": 0.78, + "learning_rate": 4.730184544172655e-08, + "loss": 0.1277, + "step": 15005 + }, + { + "epoch": 0.78, + "learning_rate": 4.728001154366272e-08, + "loss": 0.1158, + "step": 15006 + }, + { + "epoch": 0.78, + "learning_rate": 4.725818201039511e-08, + "loss": 0.1163, + "step": 15007 + }, + { + "epoch": 0.78, + "learning_rate": 4.723635684254763e-08, + "loss": 0.137, + "step": 15008 + }, + { + "epoch": 0.78, + "learning_rate": 4.721453604074403e-08, + "loss": 0.1317, + "step": 15009 + }, + { + "epoch": 0.78, + "learning_rate": 4.7192719605608e-08, + "loss": 0.1585, + "step": 15010 + }, + { + "epoch": 0.78, + "learning_rate": 4.717090753776296e-08, + "loss": 0.1261, + "step": 15011 + }, + { + "epoch": 0.78, + "learning_rate": 4.714909983783242e-08, + "loss": 0.1159, + "step": 15012 + }, + { + "epoch": 0.78, + "learning_rate": 4.712729650643963e-08, + "loss": 0.137, + "step": 15013 + }, + { + "epoch": 0.78, + "learning_rate": 4.710549754420772e-08, + "loss": 0.1322, + "step": 15014 + }, + { + "epoch": 0.78, + "learning_rate": 4.7083702951759655e-08, + "loss": 0.1745, + "step": 15015 + }, + { + "epoch": 0.78, + "learning_rate": 4.7061912729718513e-08, + "loss": 0.1293, + "step": 15016 + }, + { + "epoch": 0.78, + "learning_rate": 4.704012687870689e-08, + "loss": 0.1265, + "step": 15017 + }, + { + "epoch": 0.78, + "learning_rate": 4.701834539934741e-08, + "loss": 0.1115, + "step": 15018 + }, + { + "epoch": 0.78, + "learning_rate": 4.699656829226275e-08, + "loss": 0.1095, + "step": 15019 + }, + { + "epoch": 0.78, + "learning_rate": 4.697479555807522e-08, + "loss": 0.1188, + "step": 15020 + }, + { + "epoch": 0.78, + "learning_rate": 4.6953027197407125e-08, + "loss": 0.1611, + "step": 15021 + }, + { + "epoch": 0.78, + "learning_rate": 4.6931263210880525e-08, + "loss": 0.1296, + "step": 15022 + }, + { + "epoch": 0.78, + "learning_rate": 4.6909503599117585e-08, + "loss": 0.1411, + "step": 15023 + }, + { + "epoch": 0.78, + "learning_rate": 4.688774836274012e-08, + "loss": 0.1307, + "step": 15024 + }, + { + "epoch": 0.78, + "learning_rate": 4.68659975023699e-08, + "loss": 0.1034, + "step": 15025 + }, + { + "epoch": 0.78, + "learning_rate": 4.6844251018628613e-08, + "loss": 0.1513, + "step": 15026 + }, + { + "epoch": 0.78, + "learning_rate": 4.682250891213775e-08, + "loss": 0.141, + "step": 15027 + }, + { + "epoch": 0.78, + "learning_rate": 4.680077118351871e-08, + "loss": 0.1522, + "step": 15028 + }, + { + "epoch": 0.78, + "learning_rate": 4.67790378333927e-08, + "loss": 0.1228, + "step": 15029 + }, + { + "epoch": 0.78, + "learning_rate": 4.6757308862381004e-08, + "loss": 0.1244, + "step": 15030 + }, + { + "epoch": 0.78, + "learning_rate": 4.6735584271104577e-08, + "loss": 0.1279, + "step": 15031 + }, + { + "epoch": 0.78, + "learning_rate": 4.671386406018434e-08, + "loss": 0.1321, + "step": 15032 + }, + { + "epoch": 0.78, + "learning_rate": 4.669214823024101e-08, + "loss": 0.1435, + "step": 15033 + }, + { + "epoch": 0.78, + "learning_rate": 4.667043678189529e-08, + "loss": 0.1261, + "step": 15034 + }, + { + "epoch": 0.78, + "learning_rate": 4.6648729715767655e-08, + "loss": 0.1525, + "step": 15035 + }, + { + "epoch": 0.78, + "learning_rate": 4.6627027032478496e-08, + "loss": 0.1237, + "step": 15036 + }, + { + "epoch": 0.78, + "learning_rate": 4.6605328732648154e-08, + "loss": 0.1533, + "step": 15037 + }, + { + "epoch": 0.78, + "learning_rate": 4.658363481689673e-08, + "loss": 0.1045, + "step": 15038 + }, + { + "epoch": 0.78, + "learning_rate": 4.656194528584425e-08, + "loss": 0.1209, + "step": 15039 + }, + { + "epoch": 0.79, + "learning_rate": 4.654026014011054e-08, + "loss": 0.1261, + "step": 15040 + }, + { + "epoch": 0.79, + "learning_rate": 4.651857938031551e-08, + "loss": 0.156, + "step": 15041 + }, + { + "epoch": 0.79, + "learning_rate": 4.649690300707878e-08, + "loss": 0.1287, + "step": 15042 + }, + { + "epoch": 0.79, + "learning_rate": 4.647523102101971e-08, + "loss": 0.1458, + "step": 15043 + }, + { + "epoch": 0.79, + "learning_rate": 4.645356342275786e-08, + "loss": 0.0958, + "step": 15044 + }, + { + "epoch": 0.79, + "learning_rate": 4.6431900212912435e-08, + "loss": 0.1551, + "step": 15045 + }, + { + "epoch": 0.79, + "learning_rate": 4.641024139210261e-08, + "loss": 0.1522, + "step": 15046 + }, + { + "epoch": 0.79, + "learning_rate": 4.63885869609473e-08, + "loss": 0.1665, + "step": 15047 + }, + { + "epoch": 0.79, + "learning_rate": 4.6366936920065525e-08, + "loss": 0.0916, + "step": 15048 + }, + { + "epoch": 0.79, + "learning_rate": 4.6345291270076004e-08, + "loss": 0.1317, + "step": 15049 + }, + { + "epoch": 0.79, + "learning_rate": 4.6323650011597396e-08, + "loss": 0.131, + "step": 15050 + }, + { + "epoch": 0.79, + "learning_rate": 4.630201314524818e-08, + "loss": 0.0839, + "step": 15051 + }, + { + "epoch": 0.79, + "learning_rate": 4.6280380671646765e-08, + "loss": 0.1608, + "step": 15052 + }, + { + "epoch": 0.79, + "learning_rate": 4.625875259141141e-08, + "loss": 0.1374, + "step": 15053 + }, + { + "epoch": 0.79, + "learning_rate": 4.6237128905160186e-08, + "loss": 0.1642, + "step": 15054 + }, + { + "epoch": 0.79, + "learning_rate": 4.621550961351126e-08, + "loss": 0.1598, + "step": 15055 + }, + { + "epoch": 0.79, + "learning_rate": 4.619389471708241e-08, + "loss": 0.1321, + "step": 15056 + }, + { + "epoch": 0.79, + "learning_rate": 4.617228421649144e-08, + "loss": 0.162, + "step": 15057 + }, + { + "epoch": 0.79, + "learning_rate": 4.6150678112355914e-08, + "loss": 0.1493, + "step": 15058 + }, + { + "epoch": 0.79, + "learning_rate": 4.612907640529351e-08, + "loss": 0.122, + "step": 15059 + }, + { + "epoch": 0.79, + "learning_rate": 4.610747909592145e-08, + "loss": 0.1245, + "step": 15060 + }, + { + "epoch": 0.79, + "learning_rate": 4.608588618485698e-08, + "loss": 0.1547, + "step": 15061 + }, + { + "epoch": 0.79, + "learning_rate": 4.606429767271736e-08, + "loss": 0.1162, + "step": 15062 + }, + { + "epoch": 0.79, + "learning_rate": 4.604271356011953e-08, + "loss": 0.1116, + "step": 15063 + }, + { + "epoch": 0.79, + "learning_rate": 4.60211338476804e-08, + "loss": 0.1445, + "step": 15064 + }, + { + "epoch": 0.79, + "learning_rate": 4.599955853601662e-08, + "loss": 0.1331, + "step": 15065 + }, + { + "epoch": 0.79, + "learning_rate": 4.597798762574498e-08, + "loss": 0.1797, + "step": 15066 + }, + { + "epoch": 0.79, + "learning_rate": 4.595642111748199e-08, + "loss": 0.1216, + "step": 15067 + }, + { + "epoch": 0.79, + "learning_rate": 4.593485901184386e-08, + "loss": 0.1601, + "step": 15068 + }, + { + "epoch": 0.79, + "learning_rate": 4.5913301309446906e-08, + "loss": 0.1576, + "step": 15069 + }, + { + "epoch": 0.79, + "learning_rate": 4.5891748010907313e-08, + "loss": 0.1088, + "step": 15070 + }, + { + "epoch": 0.79, + "learning_rate": 4.587019911684109e-08, + "loss": 0.1196, + "step": 15071 + }, + { + "epoch": 0.79, + "learning_rate": 4.584865462786405e-08, + "loss": 0.1449, + "step": 15072 + }, + { + "epoch": 0.79, + "learning_rate": 4.582711454459191e-08, + "loss": 0.1665, + "step": 15073 + }, + { + "epoch": 0.79, + "learning_rate": 4.580557886764043e-08, + "loss": 0.1314, + "step": 15074 + }, + { + "epoch": 0.79, + "learning_rate": 4.5784047597624996e-08, + "loss": 0.1171, + "step": 15075 + }, + { + "epoch": 0.79, + "learning_rate": 4.576252073516105e-08, + "loss": 0.1274, + "step": 15076 + }, + { + "epoch": 0.79, + "learning_rate": 4.5740998280863795e-08, + "loss": 0.1561, + "step": 15077 + }, + { + "epoch": 0.79, + "learning_rate": 4.571948023534835e-08, + "loss": 0.1194, + "step": 15078 + }, + { + "epoch": 0.79, + "learning_rate": 4.569796659922971e-08, + "loss": 0.1328, + "step": 15079 + }, + { + "epoch": 0.79, + "learning_rate": 4.567645737312271e-08, + "loss": 0.1017, + "step": 15080 + }, + { + "epoch": 0.79, + "learning_rate": 4.56549525576422e-08, + "loss": 0.1577, + "step": 15081 + }, + { + "epoch": 0.79, + "learning_rate": 4.563345215340271e-08, + "loss": 0.1386, + "step": 15082 + }, + { + "epoch": 0.79, + "learning_rate": 4.5611956161018764e-08, + "loss": 0.1514, + "step": 15083 + }, + { + "epoch": 0.79, + "learning_rate": 4.5590464581104626e-08, + "loss": 0.1281, + "step": 15084 + }, + { + "epoch": 0.79, + "learning_rate": 4.556897741427475e-08, + "loss": 0.1726, + "step": 15085 + }, + { + "epoch": 0.79, + "learning_rate": 4.554749466114307e-08, + "loss": 0.1413, + "step": 15086 + }, + { + "epoch": 0.79, + "learning_rate": 4.552601632232354e-08, + "loss": 0.1552, + "step": 15087 + }, + { + "epoch": 0.79, + "learning_rate": 4.550454239843018e-08, + "loss": 0.1252, + "step": 15088 + }, + { + "epoch": 0.79, + "learning_rate": 4.54830728900766e-08, + "loss": 0.1399, + "step": 15089 + }, + { + "epoch": 0.79, + "learning_rate": 4.5461607797876486e-08, + "loss": 0.1227, + "step": 15090 + }, + { + "epoch": 0.79, + "learning_rate": 4.54401471224432e-08, + "loss": 0.127, + "step": 15091 + }, + { + "epoch": 0.79, + "learning_rate": 4.541869086439023e-08, + "loss": 0.0879, + "step": 15092 + }, + { + "epoch": 0.79, + "learning_rate": 4.5397239024330816e-08, + "loss": 0.1474, + "step": 15093 + }, + { + "epoch": 0.79, + "learning_rate": 4.5375791602877876e-08, + "loss": 0.1186, + "step": 15094 + }, + { + "epoch": 0.79, + "learning_rate": 4.535434860064458e-08, + "loss": 0.158, + "step": 15095 + }, + { + "epoch": 0.79, + "learning_rate": 4.5332910018243664e-08, + "loss": 0.1286, + "step": 15096 + }, + { + "epoch": 0.79, + "learning_rate": 4.531147585628792e-08, + "loss": 0.1626, + "step": 15097 + }, + { + "epoch": 0.79, + "learning_rate": 4.529004611538982e-08, + "loss": 0.1309, + "step": 15098 + }, + { + "epoch": 0.79, + "learning_rate": 4.5268620796162006e-08, + "loss": 0.1423, + "step": 15099 + }, + { + "epoch": 0.79, + "learning_rate": 4.524719989921675e-08, + "loss": 0.1278, + "step": 15100 + }, + { + "epoch": 0.79, + "learning_rate": 4.5225783425166234e-08, + "loss": 0.1841, + "step": 15101 + }, + { + "epoch": 0.79, + "learning_rate": 4.520437137462261e-08, + "loss": 0.143, + "step": 15102 + }, + { + "epoch": 0.79, + "learning_rate": 4.518296374819779e-08, + "loss": 0.121, + "step": 15103 + }, + { + "epoch": 0.79, + "learning_rate": 4.5161560546503616e-08, + "loss": 0.1545, + "step": 15104 + }, + { + "epoch": 0.79, + "learning_rate": 4.5140161770151784e-08, + "loss": 0.1228, + "step": 15105 + }, + { + "epoch": 0.79, + "learning_rate": 4.511876741975396e-08, + "loss": 0.1345, + "step": 15106 + }, + { + "epoch": 0.79, + "learning_rate": 4.509737749592155e-08, + "loss": 0.1278, + "step": 15107 + }, + { + "epoch": 0.79, + "learning_rate": 4.5075991999265884e-08, + "loss": 0.1338, + "step": 15108 + }, + { + "epoch": 0.79, + "learning_rate": 4.505461093039813e-08, + "loss": 0.1103, + "step": 15109 + }, + { + "epoch": 0.79, + "learning_rate": 4.503323428992951e-08, + "loss": 0.1482, + "step": 15110 + }, + { + "epoch": 0.79, + "learning_rate": 4.501186207847083e-08, + "loss": 0.1388, + "step": 15111 + }, + { + "epoch": 0.79, + "learning_rate": 4.499049429663291e-08, + "loss": 0.153, + "step": 15112 + }, + { + "epoch": 0.79, + "learning_rate": 4.4969130945026545e-08, + "loss": 0.1475, + "step": 15113 + }, + { + "epoch": 0.79, + "learning_rate": 4.494777202426227e-08, + "loss": 0.118, + "step": 15114 + }, + { + "epoch": 0.79, + "learning_rate": 4.4926417534950545e-08, + "loss": 0.1102, + "step": 15115 + }, + { + "epoch": 0.79, + "learning_rate": 4.49050674777016e-08, + "loss": 0.1262, + "step": 15116 + }, + { + "epoch": 0.79, + "learning_rate": 4.488372185312577e-08, + "loss": 0.1836, + "step": 15117 + }, + { + "epoch": 0.79, + "learning_rate": 4.4862380661833035e-08, + "loss": 0.1774, + "step": 15118 + }, + { + "epoch": 0.79, + "learning_rate": 4.4841043904433354e-08, + "loss": 0.1119, + "step": 15119 + }, + { + "epoch": 0.79, + "learning_rate": 4.481971158153652e-08, + "loss": 0.1484, + "step": 15120 + }, + { + "epoch": 0.79, + "learning_rate": 4.479838369375226e-08, + "loss": 0.1239, + "step": 15121 + }, + { + "epoch": 0.79, + "learning_rate": 4.47770602416901e-08, + "loss": 0.1717, + "step": 15122 + }, + { + "epoch": 0.79, + "learning_rate": 4.475574122595942e-08, + "loss": 0.1485, + "step": 15123 + }, + { + "epoch": 0.79, + "learning_rate": 4.4734426647169646e-08, + "loss": 0.1363, + "step": 15124 + }, + { + "epoch": 0.79, + "learning_rate": 4.471311650592988e-08, + "loss": 0.1723, + "step": 15125 + }, + { + "epoch": 0.79, + "learning_rate": 4.469181080284921e-08, + "loss": 0.1683, + "step": 15126 + }, + { + "epoch": 0.79, + "learning_rate": 4.467050953853653e-08, + "loss": 0.1537, + "step": 15127 + }, + { + "epoch": 0.79, + "learning_rate": 4.464921271360067e-08, + "loss": 0.1082, + "step": 15128 + }, + { + "epoch": 0.79, + "learning_rate": 4.462792032865026e-08, + "loss": 0.1269, + "step": 15129 + }, + { + "epoch": 0.79, + "learning_rate": 4.460663238429383e-08, + "loss": 0.1485, + "step": 15130 + }, + { + "epoch": 0.79, + "learning_rate": 4.458534888113987e-08, + "loss": 0.1296, + "step": 15131 + }, + { + "epoch": 0.79, + "learning_rate": 4.4564069819796656e-08, + "loss": 0.1306, + "step": 15132 + }, + { + "epoch": 0.79, + "learning_rate": 4.4542795200872316e-08, + "loss": 0.1404, + "step": 15133 + }, + { + "epoch": 0.79, + "learning_rate": 4.4521525024974837e-08, + "loss": 0.1249, + "step": 15134 + }, + { + "epoch": 0.79, + "learning_rate": 4.450025929271226e-08, + "loss": 0.1357, + "step": 15135 + }, + { + "epoch": 0.79, + "learning_rate": 4.447899800469237e-08, + "loss": 0.1222, + "step": 15136 + }, + { + "epoch": 0.79, + "learning_rate": 4.445774116152263e-08, + "loss": 0.1408, + "step": 15137 + }, + { + "epoch": 0.79, + "learning_rate": 4.443648876381077e-08, + "loss": 0.1345, + "step": 15138 + }, + { + "epoch": 0.79, + "learning_rate": 4.4415240812164076e-08, + "loss": 0.1506, + "step": 15139 + }, + { + "epoch": 0.79, + "learning_rate": 4.439399730718989e-08, + "loss": 0.1465, + "step": 15140 + }, + { + "epoch": 0.79, + "learning_rate": 4.437275824949527e-08, + "loss": 0.1401, + "step": 15141 + }, + { + "epoch": 0.79, + "learning_rate": 4.435152363968735e-08, + "loss": 0.1112, + "step": 15142 + }, + { + "epoch": 0.79, + "learning_rate": 4.433029347837296e-08, + "loss": 0.1749, + "step": 15143 + }, + { + "epoch": 0.79, + "learning_rate": 4.430906776615888e-08, + "loss": 0.1308, + "step": 15144 + }, + { + "epoch": 0.79, + "learning_rate": 4.428784650365176e-08, + "loss": 0.1555, + "step": 15145 + }, + { + "epoch": 0.79, + "learning_rate": 4.426662969145809e-08, + "loss": 0.1076, + "step": 15146 + }, + { + "epoch": 0.79, + "learning_rate": 4.4245417330184256e-08, + "loss": 0.1371, + "step": 15147 + }, + { + "epoch": 0.79, + "learning_rate": 4.422420942043647e-08, + "loss": 0.1279, + "step": 15148 + }, + { + "epoch": 0.79, + "learning_rate": 4.420300596282098e-08, + "loss": 0.1218, + "step": 15149 + }, + { + "epoch": 0.79, + "learning_rate": 4.4181806957943715e-08, + "loss": 0.1336, + "step": 15150 + }, + { + "epoch": 0.79, + "learning_rate": 4.416061240641058e-08, + "loss": 0.1207, + "step": 15151 + }, + { + "epoch": 0.79, + "learning_rate": 4.413942230882721e-08, + "loss": 0.1291, + "step": 15152 + }, + { + "epoch": 0.79, + "learning_rate": 4.4118236665799456e-08, + "loss": 0.1563, + "step": 15153 + }, + { + "epoch": 0.79, + "learning_rate": 4.409705547793261e-08, + "loss": 0.104, + "step": 15154 + }, + { + "epoch": 0.79, + "learning_rate": 4.407587874583205e-08, + "loss": 0.1298, + "step": 15155 + }, + { + "epoch": 0.79, + "learning_rate": 4.405470647010312e-08, + "loss": 0.1716, + "step": 15156 + }, + { + "epoch": 0.79, + "learning_rate": 4.403353865135089e-08, + "loss": 0.0947, + "step": 15157 + }, + { + "epoch": 0.79, + "learning_rate": 4.4012375290180313e-08, + "loss": 0.1118, + "step": 15158 + }, + { + "epoch": 0.79, + "learning_rate": 4.3991216387196226e-08, + "loss": 0.1389, + "step": 15159 + }, + { + "epoch": 0.79, + "learning_rate": 4.3970061943003453e-08, + "loss": 0.1458, + "step": 15160 + }, + { + "epoch": 0.79, + "learning_rate": 4.394891195820656e-08, + "loss": 0.1401, + "step": 15161 + }, + { + "epoch": 0.79, + "learning_rate": 4.3927766433409983e-08, + "loss": 0.1374, + "step": 15162 + }, + { + "epoch": 0.79, + "learning_rate": 4.3906625369218095e-08, + "loss": 0.1327, + "step": 15163 + }, + { + "epoch": 0.79, + "learning_rate": 4.3885488766235125e-08, + "loss": 0.1178, + "step": 15164 + }, + { + "epoch": 0.79, + "learning_rate": 4.3864356625065134e-08, + "loss": 0.1321, + "step": 15165 + }, + { + "epoch": 0.79, + "learning_rate": 4.384322894631205e-08, + "loss": 0.147, + "step": 15166 + }, + { + "epoch": 0.79, + "learning_rate": 4.382210573057983e-08, + "loss": 0.1286, + "step": 15167 + }, + { + "epoch": 0.79, + "learning_rate": 4.3800986978472124e-08, + "loss": 0.1082, + "step": 15168 + }, + { + "epoch": 0.79, + "learning_rate": 4.377987269059251e-08, + "loss": 0.1651, + "step": 15169 + }, + { + "epoch": 0.79, + "learning_rate": 4.375876286754443e-08, + "loss": 0.1979, + "step": 15170 + }, + { + "epoch": 0.79, + "learning_rate": 4.3737657509931236e-08, + "loss": 0.1228, + "step": 15171 + }, + { + "epoch": 0.79, + "learning_rate": 4.371655661835609e-08, + "loss": 0.1536, + "step": 15172 + }, + { + "epoch": 0.79, + "learning_rate": 4.3695460193422054e-08, + "loss": 0.1409, + "step": 15173 + }, + { + "epoch": 0.79, + "learning_rate": 4.3674368235732164e-08, + "loss": 0.1587, + "step": 15174 + }, + { + "epoch": 0.79, + "learning_rate": 4.3653280745889164e-08, + "loss": 0.1784, + "step": 15175 + }, + { + "epoch": 0.79, + "learning_rate": 4.363219772449576e-08, + "loss": 0.1308, + "step": 15176 + }, + { + "epoch": 0.79, + "learning_rate": 4.361111917215443e-08, + "loss": 0.1184, + "step": 15177 + }, + { + "epoch": 0.79, + "learning_rate": 4.3590045089467775e-08, + "loss": 0.1319, + "step": 15178 + }, + { + "epoch": 0.79, + "learning_rate": 4.3568975477038063e-08, + "loss": 0.1777, + "step": 15179 + }, + { + "epoch": 0.79, + "learning_rate": 4.3547910335467364e-08, + "loss": 0.1178, + "step": 15180 + }, + { + "epoch": 0.79, + "learning_rate": 4.352684966535771e-08, + "loss": 0.1374, + "step": 15181 + }, + { + "epoch": 0.79, + "learning_rate": 4.350579346731116e-08, + "loss": 0.1226, + "step": 15182 + }, + { + "epoch": 0.79, + "learning_rate": 4.348474174192946e-08, + "loss": 0.1296, + "step": 15183 + }, + { + "epoch": 0.79, + "learning_rate": 4.3463694489814265e-08, + "loss": 0.1225, + "step": 15184 + }, + { + "epoch": 0.79, + "learning_rate": 4.3442651711567026e-08, + "loss": 0.1322, + "step": 15185 + }, + { + "epoch": 0.79, + "learning_rate": 4.342161340778932e-08, + "loss": 0.122, + "step": 15186 + }, + { + "epoch": 0.79, + "learning_rate": 4.3400579579082387e-08, + "loss": 0.1389, + "step": 15187 + }, + { + "epoch": 0.79, + "learning_rate": 4.3379550226047243e-08, + "loss": 0.119, + "step": 15188 + }, + { + "epoch": 0.79, + "learning_rate": 4.335852534928508e-08, + "loss": 0.1496, + "step": 15189 + }, + { + "epoch": 0.79, + "learning_rate": 4.333750494939672e-08, + "loss": 0.1172, + "step": 15190 + }, + { + "epoch": 0.79, + "learning_rate": 4.3316489026982925e-08, + "loss": 0.1379, + "step": 15191 + }, + { + "epoch": 0.79, + "learning_rate": 4.329547758264433e-08, + "loss": 0.1276, + "step": 15192 + }, + { + "epoch": 0.79, + "learning_rate": 4.327447061698153e-08, + "loss": 0.1261, + "step": 15193 + }, + { + "epoch": 0.79, + "learning_rate": 4.325346813059487e-08, + "loss": 0.1324, + "step": 15194 + }, + { + "epoch": 0.79, + "learning_rate": 4.32324701240846e-08, + "loss": 0.1282, + "step": 15195 + }, + { + "epoch": 0.79, + "learning_rate": 4.3211476598050844e-08, + "loss": 0.098, + "step": 15196 + }, + { + "epoch": 0.79, + "learning_rate": 4.31904875530936e-08, + "loss": 0.1628, + "step": 15197 + }, + { + "epoch": 0.79, + "learning_rate": 4.316950298981279e-08, + "loss": 0.141, + "step": 15198 + }, + { + "epoch": 0.79, + "learning_rate": 4.314852290880804e-08, + "loss": 0.1394, + "step": 15199 + }, + { + "epoch": 0.79, + "learning_rate": 4.312754731067914e-08, + "loss": 0.1306, + "step": 15200 + }, + { + "epoch": 0.79, + "learning_rate": 4.310657619602547e-08, + "loss": 0.1511, + "step": 15201 + }, + { + "epoch": 0.79, + "learning_rate": 4.308560956544642e-08, + "loss": 0.1536, + "step": 15202 + }, + { + "epoch": 0.79, + "learning_rate": 4.306464741954119e-08, + "loss": 0.1291, + "step": 15203 + }, + { + "epoch": 0.79, + "learning_rate": 4.304368975890895e-08, + "loss": 0.102, + "step": 15204 + }, + { + "epoch": 0.79, + "learning_rate": 4.3022736584148724e-08, + "loss": 0.1418, + "step": 15205 + }, + { + "epoch": 0.79, + "learning_rate": 4.3001787895859174e-08, + "loss": 0.1342, + "step": 15206 + }, + { + "epoch": 0.79, + "learning_rate": 4.298084369463919e-08, + "loss": 0.1309, + "step": 15207 + }, + { + "epoch": 0.79, + "learning_rate": 4.295990398108729e-08, + "loss": 0.1296, + "step": 15208 + }, + { + "epoch": 0.79, + "learning_rate": 4.293896875580196e-08, + "loss": 0.121, + "step": 15209 + }, + { + "epoch": 0.79, + "learning_rate": 4.2918038019381497e-08, + "loss": 0.1503, + "step": 15210 + }, + { + "epoch": 0.79, + "learning_rate": 4.289711177242419e-08, + "loss": 0.1304, + "step": 15211 + }, + { + "epoch": 0.79, + "learning_rate": 4.287619001552809e-08, + "loss": 0.13, + "step": 15212 + }, + { + "epoch": 0.79, + "learning_rate": 4.2855272749291125e-08, + "loss": 0.1437, + "step": 15213 + }, + { + "epoch": 0.79, + "learning_rate": 4.283435997431113e-08, + "loss": 0.1426, + "step": 15214 + }, + { + "epoch": 0.79, + "learning_rate": 4.2813451691185796e-08, + "loss": 0.1415, + "step": 15215 + }, + { + "epoch": 0.79, + "learning_rate": 4.27925479005127e-08, + "loss": 0.1267, + "step": 15216 + }, + { + "epoch": 0.79, + "learning_rate": 4.27716486028892e-08, + "loss": 0.1399, + "step": 15217 + }, + { + "epoch": 0.79, + "learning_rate": 4.2750753798912755e-08, + "loss": 0.1077, + "step": 15218 + }, + { + "epoch": 0.79, + "learning_rate": 4.2729863489180465e-08, + "loss": 0.0955, + "step": 15219 + }, + { + "epoch": 0.79, + "learning_rate": 4.2708977674289384e-08, + "loss": 0.125, + "step": 15220 + }, + { + "epoch": 0.79, + "learning_rate": 4.268809635483639e-08, + "loss": 0.1151, + "step": 15221 + }, + { + "epoch": 0.79, + "learning_rate": 4.2667219531418453e-08, + "loss": 0.1376, + "step": 15222 + }, + { + "epoch": 0.79, + "learning_rate": 4.264634720463205e-08, + "loss": 0.2098, + "step": 15223 + }, + { + "epoch": 0.79, + "learning_rate": 4.262547937507372e-08, + "loss": 0.1104, + "step": 15224 + }, + { + "epoch": 0.79, + "learning_rate": 4.260461604334002e-08, + "loss": 0.1106, + "step": 15225 + }, + { + "epoch": 0.79, + "learning_rate": 4.258375721002714e-08, + "loss": 0.1076, + "step": 15226 + }, + { + "epoch": 0.79, + "learning_rate": 4.2562902875731255e-08, + "loss": 0.1161, + "step": 15227 + }, + { + "epoch": 0.79, + "learning_rate": 4.254205304104832e-08, + "loss": 0.1216, + "step": 15228 + }, + { + "epoch": 0.79, + "learning_rate": 4.252120770657437e-08, + "loss": 0.1229, + "step": 15229 + }, + { + "epoch": 0.79, + "learning_rate": 4.2500366872905126e-08, + "loss": 0.1499, + "step": 15230 + }, + { + "epoch": 0.79, + "learning_rate": 4.2479530540636086e-08, + "loss": 0.1283, + "step": 15231 + }, + { + "epoch": 0.8, + "learning_rate": 4.245869871036296e-08, + "loss": 0.1305, + "step": 15232 + }, + { + "epoch": 0.8, + "learning_rate": 4.2437871382681e-08, + "loss": 0.1405, + "step": 15233 + }, + { + "epoch": 0.8, + "learning_rate": 4.241704855818553e-08, + "loss": 0.1373, + "step": 15234 + }, + { + "epoch": 0.8, + "learning_rate": 4.239623023747159e-08, + "loss": 0.1703, + "step": 15235 + }, + { + "epoch": 0.8, + "learning_rate": 4.237541642113427e-08, + "loss": 0.1528, + "step": 15236 + }, + { + "epoch": 0.8, + "learning_rate": 4.235460710976842e-08, + "loss": 0.1276, + "step": 15237 + }, + { + "epoch": 0.8, + "learning_rate": 4.233380230396872e-08, + "loss": 0.1442, + "step": 15238 + }, + { + "epoch": 0.8, + "learning_rate": 4.231300200432986e-08, + "loss": 0.1121, + "step": 15239 + }, + { + "epoch": 0.8, + "learning_rate": 4.229220621144625e-08, + "loss": 0.1063, + "step": 15240 + }, + { + "epoch": 0.8, + "learning_rate": 4.227141492591226e-08, + "loss": 0.1719, + "step": 15241 + }, + { + "epoch": 0.8, + "learning_rate": 4.225062814832208e-08, + "loss": 0.1603, + "step": 15242 + }, + { + "epoch": 0.8, + "learning_rate": 4.22298458792699e-08, + "loss": 0.11, + "step": 15243 + }, + { + "epoch": 0.8, + "learning_rate": 4.2209068119349635e-08, + "loss": 0.1414, + "step": 15244 + }, + { + "epoch": 0.8, + "learning_rate": 4.2188294869155093e-08, + "loss": 0.1644, + "step": 15245 + }, + { + "epoch": 0.8, + "learning_rate": 4.2167526129279984e-08, + "loss": 0.124, + "step": 15246 + }, + { + "epoch": 0.8, + "learning_rate": 4.2146761900317983e-08, + "loss": 0.1528, + "step": 15247 + }, + { + "epoch": 0.8, + "learning_rate": 4.212600218286244e-08, + "loss": 0.1082, + "step": 15248 + }, + { + "epoch": 0.8, + "learning_rate": 4.210524697750664e-08, + "loss": 0.1542, + "step": 15249 + }, + { + "epoch": 0.8, + "learning_rate": 4.208449628484387e-08, + "loss": 0.1654, + "step": 15250 + }, + { + "epoch": 0.8, + "learning_rate": 4.206375010546717e-08, + "loss": 0.1006, + "step": 15251 + }, + { + "epoch": 0.8, + "learning_rate": 4.2043008439969464e-08, + "loss": 0.1464, + "step": 15252 + }, + { + "epoch": 0.8, + "learning_rate": 4.2022271288943487e-08, + "loss": 0.1782, + "step": 15253 + }, + { + "epoch": 0.8, + "learning_rate": 4.200153865298206e-08, + "loss": 0.113, + "step": 15254 + }, + { + "epoch": 0.8, + "learning_rate": 4.1980810532677636e-08, + "loss": 0.1491, + "step": 15255 + }, + { + "epoch": 0.8, + "learning_rate": 4.196008692862265e-08, + "loss": 0.1581, + "step": 15256 + }, + { + "epoch": 0.8, + "learning_rate": 4.193936784140939e-08, + "loss": 0.1369, + "step": 15257 + }, + { + "epoch": 0.8, + "learning_rate": 4.191865327163e-08, + "loss": 0.1118, + "step": 15258 + }, + { + "epoch": 0.8, + "learning_rate": 4.189794321987654e-08, + "loss": 0.1676, + "step": 15259 + }, + { + "epoch": 0.8, + "learning_rate": 4.1877237686740854e-08, + "loss": 0.1721, + "step": 15260 + }, + { + "epoch": 0.8, + "learning_rate": 4.1856536672814794e-08, + "loss": 0.1096, + "step": 15261 + }, + { + "epoch": 0.8, + "learning_rate": 4.183584017868995e-08, + "loss": 0.1628, + "step": 15262 + }, + { + "epoch": 0.8, + "learning_rate": 4.181514820495788e-08, + "loss": 0.1274, + "step": 15263 + }, + { + "epoch": 0.8, + "learning_rate": 4.179446075220987e-08, + "loss": 0.1422, + "step": 15264 + }, + { + "epoch": 0.8, + "learning_rate": 4.177377782103737e-08, + "loss": 0.1619, + "step": 15265 + }, + { + "epoch": 0.8, + "learning_rate": 4.1753099412031314e-08, + "loss": 0.1335, + "step": 15266 + }, + { + "epoch": 0.8, + "learning_rate": 4.173242552578271e-08, + "loss": 0.1299, + "step": 15267 + }, + { + "epoch": 0.8, + "learning_rate": 4.171175616288254e-08, + "loss": 0.1813, + "step": 15268 + }, + { + "epoch": 0.8, + "learning_rate": 4.1691091323921504e-08, + "loss": 0.1592, + "step": 15269 + }, + { + "epoch": 0.8, + "learning_rate": 4.167043100949015e-08, + "loss": 0.1393, + "step": 15270 + }, + { + "epoch": 0.8, + "learning_rate": 4.1649775220178985e-08, + "loss": 0.1262, + "step": 15271 + }, + { + "epoch": 0.8, + "learning_rate": 4.1629123956578406e-08, + "loss": 0.1186, + "step": 15272 + }, + { + "epoch": 0.8, + "learning_rate": 4.1608477219278646e-08, + "loss": 0.1253, + "step": 15273 + }, + { + "epoch": 0.8, + "learning_rate": 4.1587835008869664e-08, + "loss": 0.1171, + "step": 15274 + }, + { + "epoch": 0.8, + "learning_rate": 4.156719732594156e-08, + "loss": 0.1146, + "step": 15275 + }, + { + "epoch": 0.8, + "learning_rate": 4.1546564171084106e-08, + "loss": 0.1324, + "step": 15276 + }, + { + "epoch": 0.8, + "learning_rate": 4.152593554488704e-08, + "loss": 0.1351, + "step": 15277 + }, + { + "epoch": 0.8, + "learning_rate": 4.150531144793985e-08, + "loss": 0.1695, + "step": 15278 + }, + { + "epoch": 0.8, + "learning_rate": 4.148469188083208e-08, + "loss": 0.1304, + "step": 15279 + }, + { + "epoch": 0.8, + "learning_rate": 4.1464076844153026e-08, + "loss": 0.1025, + "step": 15280 + }, + { + "epoch": 0.8, + "learning_rate": 4.144346633849185e-08, + "loss": 0.1396, + "step": 15281 + }, + { + "epoch": 0.8, + "learning_rate": 4.142286036443763e-08, + "loss": 0.189, + "step": 15282 + }, + { + "epoch": 0.8, + "learning_rate": 4.140225892257927e-08, + "loss": 0.1355, + "step": 15283 + }, + { + "epoch": 0.8, + "learning_rate": 4.138166201350557e-08, + "loss": 0.1725, + "step": 15284 + }, + { + "epoch": 0.8, + "learning_rate": 4.136106963780514e-08, + "loss": 0.1492, + "step": 15285 + }, + { + "epoch": 0.8, + "learning_rate": 4.134048179606666e-08, + "loss": 0.11, + "step": 15286 + }, + { + "epoch": 0.8, + "learning_rate": 4.131989848887845e-08, + "loss": 0.1667, + "step": 15287 + }, + { + "epoch": 0.8, + "learning_rate": 4.129931971682879e-08, + "loss": 0.1201, + "step": 15288 + }, + { + "epoch": 0.8, + "learning_rate": 4.127874548050578e-08, + "loss": 0.1473, + "step": 15289 + }, + { + "epoch": 0.8, + "learning_rate": 4.1258175780497594e-08, + "loss": 0.136, + "step": 15290 + }, + { + "epoch": 0.8, + "learning_rate": 4.1237610617391995e-08, + "loss": 0.1475, + "step": 15291 + }, + { + "epoch": 0.8, + "learning_rate": 4.121704999177677e-08, + "loss": 0.1608, + "step": 15292 + }, + { + "epoch": 0.8, + "learning_rate": 4.119649390423949e-08, + "loss": 0.1361, + "step": 15293 + }, + { + "epoch": 0.8, + "learning_rate": 4.117594235536779e-08, + "loss": 0.0996, + "step": 15294 + }, + { + "epoch": 0.8, + "learning_rate": 4.1155395345748947e-08, + "loss": 0.1367, + "step": 15295 + }, + { + "epoch": 0.8, + "learning_rate": 4.113485287597023e-08, + "loss": 0.1355, + "step": 15296 + }, + { + "epoch": 0.8, + "learning_rate": 4.11143149466187e-08, + "loss": 0.1229, + "step": 15297 + }, + { + "epoch": 0.8, + "learning_rate": 4.1093781558281424e-08, + "loss": 0.1208, + "step": 15298 + }, + { + "epoch": 0.8, + "learning_rate": 4.107325271154529e-08, + "loss": 0.1316, + "step": 15299 + }, + { + "epoch": 0.8, + "learning_rate": 4.1052728406996826e-08, + "loss": 0.1595, + "step": 15300 + }, + { + "epoch": 0.8, + "learning_rate": 4.103220864522279e-08, + "loss": 0.1104, + "step": 15301 + }, + { + "epoch": 0.8, + "learning_rate": 4.1011693426809614e-08, + "loss": 0.1347, + "step": 15302 + }, + { + "epoch": 0.8, + "learning_rate": 4.099118275234361e-08, + "loss": 0.1524, + "step": 15303 + }, + { + "epoch": 0.8, + "learning_rate": 4.097067662241091e-08, + "loss": 0.1226, + "step": 15304 + }, + { + "epoch": 0.8, + "learning_rate": 4.095017503759775e-08, + "loss": 0.1705, + "step": 15305 + }, + { + "epoch": 0.8, + "learning_rate": 4.0929677998489985e-08, + "loss": 0.1524, + "step": 15306 + }, + { + "epoch": 0.8, + "learning_rate": 4.090918550567342e-08, + "loss": 0.1606, + "step": 15307 + }, + { + "epoch": 0.8, + "learning_rate": 4.088869755973374e-08, + "loss": 0.123, + "step": 15308 + }, + { + "epoch": 0.8, + "learning_rate": 4.086821416125652e-08, + "loss": 0.1501, + "step": 15309 + }, + { + "epoch": 0.8, + "learning_rate": 4.084773531082717e-08, + "loss": 0.1192, + "step": 15310 + }, + { + "epoch": 0.8, + "learning_rate": 4.082726100903093e-08, + "loss": 0.1202, + "step": 15311 + }, + { + "epoch": 0.8, + "learning_rate": 4.0806791256453054e-08, + "loss": 0.1128, + "step": 15312 + }, + { + "epoch": 0.8, + "learning_rate": 4.078632605367855e-08, + "loss": 0.1394, + "step": 15313 + }, + { + "epoch": 0.8, + "learning_rate": 4.0765865401292323e-08, + "loss": 0.1273, + "step": 15314 + }, + { + "epoch": 0.8, + "learning_rate": 4.0745409299879065e-08, + "loss": 0.156, + "step": 15315 + }, + { + "epoch": 0.8, + "learning_rate": 4.0724957750023606e-08, + "loss": 0.1374, + "step": 15316 + }, + { + "epoch": 0.8, + "learning_rate": 4.0704510752310274e-08, + "loss": 0.1252, + "step": 15317 + }, + { + "epoch": 0.8, + "learning_rate": 4.068406830732347e-08, + "loss": 0.1196, + "step": 15318 + }, + { + "epoch": 0.8, + "learning_rate": 4.066363041564755e-08, + "loss": 0.1594, + "step": 15319 + }, + { + "epoch": 0.8, + "learning_rate": 4.064319707786659e-08, + "loss": 0.1256, + "step": 15320 + }, + { + "epoch": 0.8, + "learning_rate": 4.0622768294564566e-08, + "loss": 0.1453, + "step": 15321 + }, + { + "epoch": 0.8, + "learning_rate": 4.06023440663253e-08, + "loss": 0.1398, + "step": 15322 + }, + { + "epoch": 0.8, + "learning_rate": 4.0581924393732626e-08, + "loss": 0.1294, + "step": 15323 + }, + { + "epoch": 0.8, + "learning_rate": 4.05615092773701e-08, + "loss": 0.165, + "step": 15324 + }, + { + "epoch": 0.8, + "learning_rate": 4.0541098717821185e-08, + "loss": 0.1759, + "step": 15325 + }, + { + "epoch": 0.8, + "learning_rate": 4.0520692715669225e-08, + "loss": 0.1276, + "step": 15326 + }, + { + "epoch": 0.8, + "learning_rate": 4.0500291271497435e-08, + "loss": 0.1345, + "step": 15327 + }, + { + "epoch": 0.8, + "learning_rate": 4.047989438588888e-08, + "loss": 0.1607, + "step": 15328 + }, + { + "epoch": 0.8, + "learning_rate": 4.04595020594265e-08, + "loss": 0.1447, + "step": 15329 + }, + { + "epoch": 0.8, + "learning_rate": 4.0439114292693154e-08, + "loss": 0.1171, + "step": 15330 + }, + { + "epoch": 0.8, + "learning_rate": 4.041873108627154e-08, + "loss": 0.1123, + "step": 15331 + }, + { + "epoch": 0.8, + "learning_rate": 4.039835244074421e-08, + "loss": 0.1863, + "step": 15332 + }, + { + "epoch": 0.8, + "learning_rate": 4.037797835669357e-08, + "loss": 0.0992, + "step": 15333 + }, + { + "epoch": 0.8, + "learning_rate": 4.035760883470192e-08, + "loss": 0.1461, + "step": 15334 + }, + { + "epoch": 0.8, + "learning_rate": 4.033724387535142e-08, + "loss": 0.1207, + "step": 15335 + }, + { + "epoch": 0.8, + "learning_rate": 4.0316883479224105e-08, + "loss": 0.1288, + "step": 15336 + }, + { + "epoch": 0.8, + "learning_rate": 4.029652764690194e-08, + "loss": 0.1085, + "step": 15337 + }, + { + "epoch": 0.8, + "learning_rate": 4.0276176378966676e-08, + "loss": 0.1264, + "step": 15338 + }, + { + "epoch": 0.8, + "learning_rate": 4.0255829675999964e-08, + "loss": 0.1124, + "step": 15339 + }, + { + "epoch": 0.8, + "learning_rate": 4.0235487538583235e-08, + "loss": 0.1573, + "step": 15340 + }, + { + "epoch": 0.8, + "learning_rate": 4.0215149967298e-08, + "loss": 0.1226, + "step": 15341 + }, + { + "epoch": 0.8, + "learning_rate": 4.019481696272551e-08, + "loss": 0.1382, + "step": 15342 + }, + { + "epoch": 0.8, + "learning_rate": 4.017448852544676e-08, + "loss": 0.0935, + "step": 15343 + }, + { + "epoch": 0.8, + "learning_rate": 4.015416465604287e-08, + "loss": 0.1353, + "step": 15344 + }, + { + "epoch": 0.8, + "learning_rate": 4.0133845355094655e-08, + "loss": 0.1274, + "step": 15345 + }, + { + "epoch": 0.8, + "learning_rate": 4.011353062318283e-08, + "loss": 0.131, + "step": 15346 + }, + { + "epoch": 0.8, + "learning_rate": 4.009322046088801e-08, + "loss": 0.1273, + "step": 15347 + }, + { + "epoch": 0.8, + "learning_rate": 4.007291486879072e-08, + "loss": 0.1434, + "step": 15348 + }, + { + "epoch": 0.8, + "learning_rate": 4.005261384747125e-08, + "loss": 0.1166, + "step": 15349 + }, + { + "epoch": 0.8, + "learning_rate": 4.003231739750983e-08, + "loss": 0.1771, + "step": 15350 + }, + { + "epoch": 0.8, + "learning_rate": 4.0012025519486523e-08, + "loss": 0.1538, + "step": 15351 + }, + { + "epoch": 0.8, + "learning_rate": 3.9991738213981315e-08, + "loss": 0.1153, + "step": 15352 + }, + { + "epoch": 0.8, + "learning_rate": 3.997145548157397e-08, + "loss": 0.1224, + "step": 15353 + }, + { + "epoch": 0.8, + "learning_rate": 3.9951177322844164e-08, + "loss": 0.1559, + "step": 15354 + }, + { + "epoch": 0.8, + "learning_rate": 3.993090373837156e-08, + "loss": 0.1697, + "step": 15355 + }, + { + "epoch": 0.8, + "learning_rate": 3.991063472873551e-08, + "loss": 0.1259, + "step": 15356 + }, + { + "epoch": 0.8, + "learning_rate": 3.9890370294515295e-08, + "loss": 0.1314, + "step": 15357 + }, + { + "epoch": 0.8, + "learning_rate": 3.987011043629007e-08, + "loss": 0.1394, + "step": 15358 + }, + { + "epoch": 0.8, + "learning_rate": 3.984985515463903e-08, + "loss": 0.1221, + "step": 15359 + }, + { + "epoch": 0.8, + "learning_rate": 3.9829604450140876e-08, + "loss": 0.1625, + "step": 15360 + }, + { + "epoch": 0.8, + "learning_rate": 3.980935832337442e-08, + "loss": 0.157, + "step": 15361 + }, + { + "epoch": 0.8, + "learning_rate": 3.9789116774918386e-08, + "loss": 0.1302, + "step": 15362 + }, + { + "epoch": 0.8, + "learning_rate": 3.9768879805351265e-08, + "loss": 0.1376, + "step": 15363 + }, + { + "epoch": 0.8, + "learning_rate": 3.974864741525139e-08, + "loss": 0.156, + "step": 15364 + }, + { + "epoch": 0.8, + "learning_rate": 3.9728419605196995e-08, + "loss": 0.1088, + "step": 15365 + }, + { + "epoch": 0.8, + "learning_rate": 3.970819637576628e-08, + "loss": 0.1738, + "step": 15366 + }, + { + "epoch": 0.8, + "learning_rate": 3.9687977727537245e-08, + "loss": 0.2029, + "step": 15367 + }, + { + "epoch": 0.8, + "learning_rate": 3.966776366108759e-08, + "loss": 0.1514, + "step": 15368 + }, + { + "epoch": 0.8, + "learning_rate": 3.964755417699521e-08, + "loss": 0.1514, + "step": 15369 + }, + { + "epoch": 0.8, + "learning_rate": 3.962734927583762e-08, + "loss": 0.1104, + "step": 15370 + }, + { + "epoch": 0.8, + "learning_rate": 3.96071489581923e-08, + "loss": 0.1609, + "step": 15371 + }, + { + "epoch": 0.8, + "learning_rate": 3.958695322463654e-08, + "loss": 0.1379, + "step": 15372 + }, + { + "epoch": 0.8, + "learning_rate": 3.9566762075747625e-08, + "loss": 0.15, + "step": 15373 + }, + { + "epoch": 0.8, + "learning_rate": 3.9546575512102584e-08, + "loss": 0.1502, + "step": 15374 + }, + { + "epoch": 0.8, + "learning_rate": 3.9526393534278355e-08, + "loss": 0.1687, + "step": 15375 + }, + { + "epoch": 0.8, + "learning_rate": 3.950621614285174e-08, + "loss": 0.1469, + "step": 15376 + }, + { + "epoch": 0.8, + "learning_rate": 3.948604333839945e-08, + "loss": 0.2044, + "step": 15377 + }, + { + "epoch": 0.8, + "learning_rate": 3.946587512149797e-08, + "loss": 0.1367, + "step": 15378 + }, + { + "epoch": 0.8, + "learning_rate": 3.944571149272373e-08, + "loss": 0.1508, + "step": 15379 + }, + { + "epoch": 0.8, + "learning_rate": 3.9425552452653065e-08, + "loss": 0.1702, + "step": 15380 + }, + { + "epoch": 0.8, + "learning_rate": 3.940539800186211e-08, + "loss": 0.1201, + "step": 15381 + }, + { + "epoch": 0.8, + "learning_rate": 3.938524814092687e-08, + "loss": 0.1315, + "step": 15382 + }, + { + "epoch": 0.8, + "learning_rate": 3.936510287042321e-08, + "loss": 0.1165, + "step": 15383 + }, + { + "epoch": 0.8, + "learning_rate": 3.9344962190927e-08, + "loss": 0.1468, + "step": 15384 + }, + { + "epoch": 0.8, + "learning_rate": 3.932482610301375e-08, + "loss": 0.1248, + "step": 15385 + }, + { + "epoch": 0.8, + "learning_rate": 3.930469460725894e-08, + "loss": 0.119, + "step": 15386 + }, + { + "epoch": 0.8, + "learning_rate": 3.928456770423805e-08, + "loss": 0.1072, + "step": 15387 + }, + { + "epoch": 0.8, + "learning_rate": 3.926444539452625e-08, + "loss": 0.1224, + "step": 15388 + }, + { + "epoch": 0.8, + "learning_rate": 3.9244327678698654e-08, + "loss": 0.1392, + "step": 15389 + }, + { + "epoch": 0.8, + "learning_rate": 3.922421455733018e-08, + "loss": 0.1251, + "step": 15390 + }, + { + "epoch": 0.8, + "learning_rate": 3.9204106030995773e-08, + "loss": 0.1377, + "step": 15391 + }, + { + "epoch": 0.8, + "learning_rate": 3.918400210027009e-08, + "loss": 0.1553, + "step": 15392 + }, + { + "epoch": 0.8, + "learning_rate": 3.9163902765727764e-08, + "loss": 0.1124, + "step": 15393 + }, + { + "epoch": 0.8, + "learning_rate": 3.914380802794308e-08, + "loss": 0.1449, + "step": 15394 + }, + { + "epoch": 0.8, + "learning_rate": 3.9123717887490516e-08, + "loss": 0.122, + "step": 15395 + }, + { + "epoch": 0.8, + "learning_rate": 3.910363234494418e-08, + "loss": 0.1111, + "step": 15396 + }, + { + "epoch": 0.8, + "learning_rate": 3.908355140087811e-08, + "loss": 0.121, + "step": 15397 + }, + { + "epoch": 0.8, + "learning_rate": 3.906347505586629e-08, + "loss": 0.141, + "step": 15398 + }, + { + "epoch": 0.8, + "learning_rate": 3.90434033104825e-08, + "loss": 0.1356, + "step": 15399 + }, + { + "epoch": 0.8, + "learning_rate": 3.902333616530038e-08, + "loss": 0.1368, + "step": 15400 + }, + { + "epoch": 0.8, + "learning_rate": 3.9003273620893394e-08, + "loss": 0.1318, + "step": 15401 + }, + { + "epoch": 0.8, + "learning_rate": 3.898321567783511e-08, + "loss": 0.1237, + "step": 15402 + }, + { + "epoch": 0.8, + "learning_rate": 3.896316233669863e-08, + "loss": 0.1227, + "step": 15403 + }, + { + "epoch": 0.8, + "learning_rate": 3.894311359805713e-08, + "loss": 0.1304, + "step": 15404 + }, + { + "epoch": 0.8, + "learning_rate": 3.8923069462483585e-08, + "loss": 0.1072, + "step": 15405 + }, + { + "epoch": 0.8, + "learning_rate": 3.8903029930550926e-08, + "loss": 0.1491, + "step": 15406 + }, + { + "epoch": 0.8, + "learning_rate": 3.8882995002831856e-08, + "loss": 0.2009, + "step": 15407 + }, + { + "epoch": 0.8, + "learning_rate": 3.8862964679899004e-08, + "loss": 0.1686, + "step": 15408 + }, + { + "epoch": 0.8, + "learning_rate": 3.884293896232478e-08, + "loss": 0.1163, + "step": 15409 + }, + { + "epoch": 0.8, + "learning_rate": 3.882291785068166e-08, + "loss": 0.1109, + "step": 15410 + }, + { + "epoch": 0.8, + "learning_rate": 3.8802901345541716e-08, + "loss": 0.1326, + "step": 15411 + }, + { + "epoch": 0.8, + "learning_rate": 3.878288944747703e-08, + "loss": 0.1236, + "step": 15412 + }, + { + "epoch": 0.8, + "learning_rate": 3.876288215705967e-08, + "loss": 0.1223, + "step": 15413 + }, + { + "epoch": 0.8, + "learning_rate": 3.874287947486137e-08, + "loss": 0.0995, + "step": 15414 + }, + { + "epoch": 0.8, + "learning_rate": 3.8722881401453814e-08, + "loss": 0.1261, + "step": 15415 + }, + { + "epoch": 0.8, + "learning_rate": 3.8702887937408523e-08, + "loss": 0.1156, + "step": 15416 + }, + { + "epoch": 0.8, + "learning_rate": 3.868289908329703e-08, + "loss": 0.1544, + "step": 15417 + }, + { + "epoch": 0.8, + "learning_rate": 3.866291483969053e-08, + "loss": 0.1835, + "step": 15418 + }, + { + "epoch": 0.8, + "learning_rate": 3.8642935207160224e-08, + "loss": 0.1491, + "step": 15419 + }, + { + "epoch": 0.8, + "learning_rate": 3.862296018627711e-08, + "loss": 0.1238, + "step": 15420 + }, + { + "epoch": 0.8, + "learning_rate": 3.86029897776121e-08, + "loss": 0.1253, + "step": 15421 + }, + { + "epoch": 0.8, + "learning_rate": 3.858302398173594e-08, + "loss": 0.1208, + "step": 15422 + }, + { + "epoch": 0.81, + "learning_rate": 3.8563062799219214e-08, + "loss": 0.1517, + "step": 15423 + }, + { + "epoch": 0.81, + "learning_rate": 3.854310623063253e-08, + "loss": 0.1213, + "step": 15424 + }, + { + "epoch": 0.81, + "learning_rate": 3.85231542765462e-08, + "loss": 0.1279, + "step": 15425 + }, + { + "epoch": 0.81, + "learning_rate": 3.850320693753046e-08, + "loss": 0.1193, + "step": 15426 + }, + { + "epoch": 0.81, + "learning_rate": 3.84832642141554e-08, + "loss": 0.1367, + "step": 15427 + }, + { + "epoch": 0.81, + "learning_rate": 3.846332610699101e-08, + "loss": 0.1545, + "step": 15428 + }, + { + "epoch": 0.81, + "learning_rate": 3.844339261660712e-08, + "loss": 0.1523, + "step": 15429 + }, + { + "epoch": 0.81, + "learning_rate": 3.842346374357337e-08, + "loss": 0.1672, + "step": 15430 + }, + { + "epoch": 0.81, + "learning_rate": 3.840353948845947e-08, + "loss": 0.1445, + "step": 15431 + }, + { + "epoch": 0.81, + "learning_rate": 3.838361985183478e-08, + "loss": 0.1058, + "step": 15432 + }, + { + "epoch": 0.81, + "learning_rate": 3.836370483426863e-08, + "loss": 0.1387, + "step": 15433 + }, + { + "epoch": 0.81, + "learning_rate": 3.834379443633016e-08, + "loss": 0.122, + "step": 15434 + }, + { + "epoch": 0.81, + "learning_rate": 3.832388865858847e-08, + "loss": 0.1545, + "step": 15435 + }, + { + "epoch": 0.81, + "learning_rate": 3.8303987501612523e-08, + "loss": 0.1368, + "step": 15436 + }, + { + "epoch": 0.81, + "learning_rate": 3.828409096597094e-08, + "loss": 0.1465, + "step": 15437 + }, + { + "epoch": 0.81, + "learning_rate": 3.8264199052232503e-08, + "loss": 0.1234, + "step": 15438 + }, + { + "epoch": 0.81, + "learning_rate": 3.8244311760965696e-08, + "loss": 0.1751, + "step": 15439 + }, + { + "epoch": 0.81, + "learning_rate": 3.822442909273891e-08, + "loss": 0.1515, + "step": 15440 + }, + { + "epoch": 0.81, + "learning_rate": 3.820455104812033e-08, + "loss": 0.1294, + "step": 15441 + }, + { + "epoch": 0.81, + "learning_rate": 3.818467762767821e-08, + "loss": 0.1332, + "step": 15442 + }, + { + "epoch": 0.81, + "learning_rate": 3.8164808831980453e-08, + "loss": 0.1257, + "step": 15443 + }, + { + "epoch": 0.81, + "learning_rate": 3.8144944661594946e-08, + "loss": 0.1552, + "step": 15444 + }, + { + "epoch": 0.81, + "learning_rate": 3.8125085117089384e-08, + "loss": 0.0979, + "step": 15445 + }, + { + "epoch": 0.81, + "learning_rate": 3.8105230199031405e-08, + "loss": 0.1246, + "step": 15446 + }, + { + "epoch": 0.81, + "learning_rate": 3.8085379907988414e-08, + "loss": 0.1794, + "step": 15447 + }, + { + "epoch": 0.81, + "learning_rate": 3.806553424452774e-08, + "loss": 0.1296, + "step": 15448 + }, + { + "epoch": 0.81, + "learning_rate": 3.804569320921667e-08, + "loss": 0.1018, + "step": 15449 + }, + { + "epoch": 0.81, + "learning_rate": 3.802585680262218e-08, + "loss": 0.1069, + "step": 15450 + }, + { + "epoch": 0.81, + "learning_rate": 3.800602502531125e-08, + "loss": 0.1474, + "step": 15451 + }, + { + "epoch": 0.81, + "learning_rate": 3.798619787785062e-08, + "loss": 0.1366, + "step": 15452 + }, + { + "epoch": 0.81, + "learning_rate": 3.796637536080709e-08, + "loss": 0.1292, + "step": 15453 + }, + { + "epoch": 0.81, + "learning_rate": 3.794655747474704e-08, + "loss": 0.1626, + "step": 15454 + }, + { + "epoch": 0.81, + "learning_rate": 3.7926744220236895e-08, + "loss": 0.0994, + "step": 15455 + }, + { + "epoch": 0.81, + "learning_rate": 3.790693559784301e-08, + "loss": 0.1063, + "step": 15456 + }, + { + "epoch": 0.81, + "learning_rate": 3.7887131608131506e-08, + "loss": 0.1324, + "step": 15457 + }, + { + "epoch": 0.81, + "learning_rate": 3.7867332251668336e-08, + "loss": 0.1871, + "step": 15458 + }, + { + "epoch": 0.81, + "learning_rate": 3.784753752901937e-08, + "loss": 0.1077, + "step": 15459 + }, + { + "epoch": 0.81, + "learning_rate": 3.782774744075041e-08, + "loss": 0.127, + "step": 15460 + }, + { + "epoch": 0.81, + "learning_rate": 3.780796198742704e-08, + "loss": 0.1413, + "step": 15461 + }, + { + "epoch": 0.81, + "learning_rate": 3.778818116961473e-08, + "loss": 0.0862, + "step": 15462 + }, + { + "epoch": 0.81, + "learning_rate": 3.776840498787881e-08, + "loss": 0.1051, + "step": 15463 + }, + { + "epoch": 0.81, + "learning_rate": 3.7748633442784516e-08, + "loss": 0.1468, + "step": 15464 + }, + { + "epoch": 0.81, + "learning_rate": 3.772886653489691e-08, + "loss": 0.0988, + "step": 15465 + }, + { + "epoch": 0.81, + "learning_rate": 3.770910426478089e-08, + "loss": 0.121, + "step": 15466 + }, + { + "epoch": 0.81, + "learning_rate": 3.7689346633001385e-08, + "loss": 0.1742, + "step": 15467 + }, + { + "epoch": 0.81, + "learning_rate": 3.7669593640123006e-08, + "loss": 0.1298, + "step": 15468 + }, + { + "epoch": 0.81, + "learning_rate": 3.764984528671029e-08, + "loss": 0.1548, + "step": 15469 + }, + { + "epoch": 0.81, + "learning_rate": 3.7630101573327686e-08, + "loss": 0.1021, + "step": 15470 + }, + { + "epoch": 0.81, + "learning_rate": 3.761036250053946e-08, + "loss": 0.1267, + "step": 15471 + }, + { + "epoch": 0.81, + "learning_rate": 3.759062806890976e-08, + "loss": 0.1226, + "step": 15472 + }, + { + "epoch": 0.81, + "learning_rate": 3.757089827900255e-08, + "loss": 0.1412, + "step": 15473 + }, + { + "epoch": 0.81, + "learning_rate": 3.755117313138181e-08, + "loss": 0.156, + "step": 15474 + }, + { + "epoch": 0.81, + "learning_rate": 3.753145262661126e-08, + "loss": 0.1568, + "step": 15475 + }, + { + "epoch": 0.81, + "learning_rate": 3.751173676525452e-08, + "loss": 0.1387, + "step": 15476 + }, + { + "epoch": 0.81, + "learning_rate": 3.7492025547875026e-08, + "loss": 0.143, + "step": 15477 + }, + { + "epoch": 0.81, + "learning_rate": 3.7472318975036224e-08, + "loss": 0.1604, + "step": 15478 + }, + { + "epoch": 0.81, + "learning_rate": 3.7452617047301336e-08, + "loss": 0.1294, + "step": 15479 + }, + { + "epoch": 0.81, + "learning_rate": 3.743291976523331e-08, + "loss": 0.126, + "step": 15480 + }, + { + "epoch": 0.81, + "learning_rate": 3.741322712939523e-08, + "loss": 0.1291, + "step": 15481 + }, + { + "epoch": 0.81, + "learning_rate": 3.7393539140349884e-08, + "loss": 0.132, + "step": 15482 + }, + { + "epoch": 0.81, + "learning_rate": 3.737385579865997e-08, + "loss": 0.1816, + "step": 15483 + }, + { + "epoch": 0.81, + "learning_rate": 3.7354177104887976e-08, + "loss": 0.1144, + "step": 15484 + }, + { + "epoch": 0.81, + "learning_rate": 3.7334503059596444e-08, + "loss": 0.1306, + "step": 15485 + }, + { + "epoch": 0.81, + "learning_rate": 3.731483366334762e-08, + "loss": 0.0991, + "step": 15486 + }, + { + "epoch": 0.81, + "learning_rate": 3.7295168916703635e-08, + "loss": 0.1385, + "step": 15487 + }, + { + "epoch": 0.81, + "learning_rate": 3.727550882022654e-08, + "loss": 0.135, + "step": 15488 + }, + { + "epoch": 0.81, + "learning_rate": 3.725585337447823e-08, + "loss": 0.1016, + "step": 15489 + }, + { + "epoch": 0.81, + "learning_rate": 3.723620258002045e-08, + "loss": 0.1267, + "step": 15490 + }, + { + "epoch": 0.81, + "learning_rate": 3.721655643741477e-08, + "loss": 0.1508, + "step": 15491 + }, + { + "epoch": 0.81, + "learning_rate": 3.7196914947222815e-08, + "loss": 0.1732, + "step": 15492 + }, + { + "epoch": 0.81, + "learning_rate": 3.717727811000588e-08, + "loss": 0.1421, + "step": 15493 + }, + { + "epoch": 0.81, + "learning_rate": 3.7157645926325174e-08, + "loss": 0.1431, + "step": 15494 + }, + { + "epoch": 0.81, + "learning_rate": 3.713801839674178e-08, + "loss": 0.1313, + "step": 15495 + }, + { + "epoch": 0.81, + "learning_rate": 3.711839552181679e-08, + "loss": 0.1541, + "step": 15496 + }, + { + "epoch": 0.81, + "learning_rate": 3.709877730211086e-08, + "loss": 0.1239, + "step": 15497 + }, + { + "epoch": 0.81, + "learning_rate": 3.707916373818474e-08, + "loss": 0.1726, + "step": 15498 + }, + { + "epoch": 0.81, + "learning_rate": 3.705955483059906e-08, + "loss": 0.1312, + "step": 15499 + }, + { + "epoch": 0.81, + "learning_rate": 3.70399505799142e-08, + "loss": 0.1374, + "step": 15500 + }, + { + "epoch": 0.81, + "learning_rate": 3.702035098669045e-08, + "loss": 0.1481, + "step": 15501 + }, + { + "epoch": 0.81, + "learning_rate": 3.700075605148796e-08, + "loss": 0.1272, + "step": 15502 + }, + { + "epoch": 0.81, + "learning_rate": 3.698116577486681e-08, + "loss": 0.133, + "step": 15503 + }, + { + "epoch": 0.81, + "learning_rate": 3.696158015738693e-08, + "loss": 0.1455, + "step": 15504 + }, + { + "epoch": 0.81, + "learning_rate": 3.694199919960799e-08, + "loss": 0.1151, + "step": 15505 + }, + { + "epoch": 0.81, + "learning_rate": 3.69224229020896e-08, + "loss": 0.1299, + "step": 15506 + }, + { + "epoch": 0.81, + "learning_rate": 3.690285126539137e-08, + "loss": 0.1248, + "step": 15507 + }, + { + "epoch": 0.81, + "learning_rate": 3.688328429007261e-08, + "loss": 0.1334, + "step": 15508 + }, + { + "epoch": 0.81, + "learning_rate": 3.686372197669252e-08, + "loss": 0.1715, + "step": 15509 + }, + { + "epoch": 0.81, + "learning_rate": 3.684416432581026e-08, + "loss": 0.1283, + "step": 15510 + }, + { + "epoch": 0.81, + "learning_rate": 3.682461133798478e-08, + "loss": 0.1082, + "step": 15511 + }, + { + "epoch": 0.81, + "learning_rate": 3.680506301377488e-08, + "loss": 0.1439, + "step": 15512 + }, + { + "epoch": 0.81, + "learning_rate": 3.67855193537393e-08, + "loss": 0.1461, + "step": 15513 + }, + { + "epoch": 0.81, + "learning_rate": 3.676598035843659e-08, + "loss": 0.1085, + "step": 15514 + }, + { + "epoch": 0.81, + "learning_rate": 3.674644602842516e-08, + "loss": 0.1234, + "step": 15515 + }, + { + "epoch": 0.81, + "learning_rate": 3.672691636426335e-08, + "loss": 0.1336, + "step": 15516 + }, + { + "epoch": 0.81, + "learning_rate": 3.670739136650922e-08, + "loss": 0.1215, + "step": 15517 + }, + { + "epoch": 0.81, + "learning_rate": 3.6687871035720954e-08, + "loss": 0.1141, + "step": 15518 + }, + { + "epoch": 0.81, + "learning_rate": 3.666835537245636e-08, + "loss": 0.1448, + "step": 15519 + }, + { + "epoch": 0.81, + "learning_rate": 3.664884437727322e-08, + "loss": 0.1377, + "step": 15520 + }, + { + "epoch": 0.81, + "learning_rate": 3.662933805072914e-08, + "loss": 0.1634, + "step": 15521 + }, + { + "epoch": 0.81, + "learning_rate": 3.660983639338173e-08, + "loss": 0.1701, + "step": 15522 + }, + { + "epoch": 0.81, + "learning_rate": 3.659033940578822e-08, + "loss": 0.1263, + "step": 15523 + }, + { + "epoch": 0.81, + "learning_rate": 3.657084708850584e-08, + "loss": 0.1299, + "step": 15524 + }, + { + "epoch": 0.81, + "learning_rate": 3.65513594420918e-08, + "loss": 0.1364, + "step": 15525 + }, + { + "epoch": 0.81, + "learning_rate": 3.653187646710301e-08, + "loss": 0.1274, + "step": 15526 + }, + { + "epoch": 0.81, + "learning_rate": 3.651239816409628e-08, + "loss": 0.1522, + "step": 15527 + }, + { + "epoch": 0.81, + "learning_rate": 3.649292453362829e-08, + "loss": 0.1484, + "step": 15528 + }, + { + "epoch": 0.81, + "learning_rate": 3.647345557625567e-08, + "loss": 0.1158, + "step": 15529 + }, + { + "epoch": 0.81, + "learning_rate": 3.645399129253488e-08, + "loss": 0.1675, + "step": 15530 + }, + { + "epoch": 0.81, + "learning_rate": 3.643453168302204e-08, + "loss": 0.1336, + "step": 15531 + }, + { + "epoch": 0.81, + "learning_rate": 3.641507674827349e-08, + "loss": 0.1189, + "step": 15532 + }, + { + "epoch": 0.81, + "learning_rate": 3.63956264888452e-08, + "loss": 0.1388, + "step": 15533 + }, + { + "epoch": 0.81, + "learning_rate": 3.6376180905293043e-08, + "loss": 0.126, + "step": 15534 + }, + { + "epoch": 0.81, + "learning_rate": 3.635673999817277e-08, + "loss": 0.1639, + "step": 15535 + }, + { + "epoch": 0.81, + "learning_rate": 3.6337303768040074e-08, + "loss": 0.1493, + "step": 15536 + }, + { + "epoch": 0.81, + "learning_rate": 3.631787221545044e-08, + "loss": 0.1232, + "step": 15537 + }, + { + "epoch": 0.81, + "learning_rate": 3.629844534095918e-08, + "loss": 0.1229, + "step": 15538 + }, + { + "epoch": 0.81, + "learning_rate": 3.627902314512157e-08, + "loss": 0.1177, + "step": 15539 + }, + { + "epoch": 0.81, + "learning_rate": 3.625960562849266e-08, + "loss": 0.1526, + "step": 15540 + }, + { + "epoch": 0.81, + "learning_rate": 3.6240192791627444e-08, + "loss": 0.1292, + "step": 15541 + }, + { + "epoch": 0.81, + "learning_rate": 3.622078463508067e-08, + "loss": 0.1503, + "step": 15542 + }, + { + "epoch": 0.81, + "learning_rate": 3.620138115940716e-08, + "loss": 0.1247, + "step": 15543 + }, + { + "epoch": 0.81, + "learning_rate": 3.618198236516141e-08, + "loss": 0.1561, + "step": 15544 + }, + { + "epoch": 0.81, + "learning_rate": 3.616258825289784e-08, + "loss": 0.1466, + "step": 15545 + }, + { + "epoch": 0.81, + "learning_rate": 3.61431988231707e-08, + "loss": 0.1544, + "step": 15546 + }, + { + "epoch": 0.81, + "learning_rate": 3.6123814076534284e-08, + "loss": 0.1113, + "step": 15547 + }, + { + "epoch": 0.81, + "learning_rate": 3.6104434013542485e-08, + "loss": 0.1112, + "step": 15548 + }, + { + "epoch": 0.81, + "learning_rate": 3.608505863474918e-08, + "loss": 0.1156, + "step": 15549 + }, + { + "epoch": 0.81, + "learning_rate": 3.6065687940708236e-08, + "loss": 0.1029, + "step": 15550 + }, + { + "epoch": 0.81, + "learning_rate": 3.604632193197319e-08, + "loss": 0.1465, + "step": 15551 + }, + { + "epoch": 0.81, + "learning_rate": 3.602696060909758e-08, + "loss": 0.1397, + "step": 15552 + }, + { + "epoch": 0.81, + "learning_rate": 3.600760397263467e-08, + "loss": 0.1354, + "step": 15553 + }, + { + "epoch": 0.81, + "learning_rate": 3.59882520231378e-08, + "loss": 0.1762, + "step": 15554 + }, + { + "epoch": 0.81, + "learning_rate": 3.596890476116001e-08, + "loss": 0.14, + "step": 15555 + }, + { + "epoch": 0.81, + "learning_rate": 3.594956218725422e-08, + "loss": 0.1434, + "step": 15556 + }, + { + "epoch": 0.81, + "learning_rate": 3.59302243019733e-08, + "loss": 0.1173, + "step": 15557 + }, + { + "epoch": 0.81, + "learning_rate": 3.591089110586989e-08, + "loss": 0.149, + "step": 15558 + }, + { + "epoch": 0.81, + "learning_rate": 3.589156259949657e-08, + "loss": 0.1128, + "step": 15559 + }, + { + "epoch": 0.81, + "learning_rate": 3.587223878340569e-08, + "loss": 0.1255, + "step": 15560 + }, + { + "epoch": 0.81, + "learning_rate": 3.585291965814963e-08, + "loss": 0.13, + "step": 15561 + }, + { + "epoch": 0.81, + "learning_rate": 3.583360522428049e-08, + "loss": 0.1281, + "step": 15562 + }, + { + "epoch": 0.81, + "learning_rate": 3.581429548235029e-08, + "loss": 0.1345, + "step": 15563 + }, + { + "epoch": 0.81, + "learning_rate": 3.5794990432910923e-08, + "loss": 0.121, + "step": 15564 + }, + { + "epoch": 0.81, + "learning_rate": 3.5775690076514084e-08, + "loss": 0.1234, + "step": 15565 + }, + { + "epoch": 0.81, + "learning_rate": 3.575639441371146e-08, + "loss": 0.1592, + "step": 15566 + }, + { + "epoch": 0.81, + "learning_rate": 3.5737103445054405e-08, + "loss": 0.1219, + "step": 15567 + }, + { + "epoch": 0.81, + "learning_rate": 3.571781717109441e-08, + "loss": 0.1263, + "step": 15568 + }, + { + "epoch": 0.81, + "learning_rate": 3.569853559238263e-08, + "loss": 0.1395, + "step": 15569 + }, + { + "epoch": 0.81, + "learning_rate": 3.5679258709470106e-08, + "loss": 0.1602, + "step": 15570 + }, + { + "epoch": 0.81, + "learning_rate": 3.565998652290778e-08, + "loss": 0.1322, + "step": 15571 + }, + { + "epoch": 0.81, + "learning_rate": 3.56407190332465e-08, + "loss": 0.1561, + "step": 15572 + }, + { + "epoch": 0.81, + "learning_rate": 3.562145624103699e-08, + "loss": 0.1241, + "step": 15573 + }, + { + "epoch": 0.81, + "learning_rate": 3.560219814682961e-08, + "loss": 0.2122, + "step": 15574 + }, + { + "epoch": 0.81, + "learning_rate": 3.5582944751174934e-08, + "loss": 0.168, + "step": 15575 + }, + { + "epoch": 0.81, + "learning_rate": 3.556369605462315e-08, + "loss": 0.1343, + "step": 15576 + }, + { + "epoch": 0.81, + "learning_rate": 3.554445205772443e-08, + "loss": 0.1463, + "step": 15577 + }, + { + "epoch": 0.81, + "learning_rate": 3.5525212761028686e-08, + "loss": 0.1114, + "step": 15578 + }, + { + "epoch": 0.81, + "learning_rate": 3.5505978165085914e-08, + "loss": 0.1282, + "step": 15579 + }, + { + "epoch": 0.81, + "learning_rate": 3.548674827044578e-08, + "loss": 0.1327, + "step": 15580 + }, + { + "epoch": 0.81, + "learning_rate": 3.54675230776579e-08, + "loss": 0.1285, + "step": 15581 + }, + { + "epoch": 0.81, + "learning_rate": 3.5448302587271717e-08, + "loss": 0.1119, + "step": 15582 + }, + { + "epoch": 0.81, + "learning_rate": 3.542908679983658e-08, + "loss": 0.1351, + "step": 15583 + }, + { + "epoch": 0.81, + "learning_rate": 3.540987571590168e-08, + "loss": 0.1507, + "step": 15584 + }, + { + "epoch": 0.81, + "learning_rate": 3.5390669336016e-08, + "loss": 0.1481, + "step": 15585 + }, + { + "epoch": 0.81, + "learning_rate": 3.5371467660728606e-08, + "loss": 0.1845, + "step": 15586 + }, + { + "epoch": 0.81, + "learning_rate": 3.535227069058819e-08, + "loss": 0.1364, + "step": 15587 + }, + { + "epoch": 0.81, + "learning_rate": 3.533307842614346e-08, + "loss": 0.1895, + "step": 15588 + }, + { + "epoch": 0.81, + "learning_rate": 3.5313890867942855e-08, + "loss": 0.1252, + "step": 15589 + }, + { + "epoch": 0.81, + "learning_rate": 3.5294708016534936e-08, + "loss": 0.1522, + "step": 15590 + }, + { + "epoch": 0.81, + "learning_rate": 3.52755298724678e-08, + "loss": 0.1149, + "step": 15591 + }, + { + "epoch": 0.81, + "learning_rate": 3.525635643628955e-08, + "loss": 0.1697, + "step": 15592 + }, + { + "epoch": 0.81, + "learning_rate": 3.5237187708548265e-08, + "loss": 0.1286, + "step": 15593 + }, + { + "epoch": 0.81, + "learning_rate": 3.5218023689791785e-08, + "loss": 0.1451, + "step": 15594 + }, + { + "epoch": 0.81, + "learning_rate": 3.519886438056778e-08, + "loss": 0.1256, + "step": 15595 + }, + { + "epoch": 0.81, + "learning_rate": 3.517970978142382e-08, + "loss": 0.1971, + "step": 15596 + }, + { + "epoch": 0.81, + "learning_rate": 3.5160559892907426e-08, + "loss": 0.1152, + "step": 15597 + }, + { + "epoch": 0.81, + "learning_rate": 3.5141414715565867e-08, + "loss": 0.148, + "step": 15598 + }, + { + "epoch": 0.81, + "learning_rate": 3.512227424994629e-08, + "loss": 0.129, + "step": 15599 + }, + { + "epoch": 0.81, + "learning_rate": 3.510313849659581e-08, + "loss": 0.119, + "step": 15600 + }, + { + "epoch": 0.81, + "learning_rate": 3.508400745606126e-08, + "loss": 0.142, + "step": 15601 + }, + { + "epoch": 0.81, + "learning_rate": 3.506488112888943e-08, + "loss": 0.1473, + "step": 15602 + }, + { + "epoch": 0.81, + "learning_rate": 3.504575951562694e-08, + "loss": 0.1793, + "step": 15603 + }, + { + "epoch": 0.81, + "learning_rate": 3.5026642616820356e-08, + "loss": 0.1065, + "step": 15604 + }, + { + "epoch": 0.81, + "learning_rate": 3.500753043301603e-08, + "loss": 0.1337, + "step": 15605 + }, + { + "epoch": 0.81, + "learning_rate": 3.498842296476017e-08, + "loss": 0.135, + "step": 15606 + }, + { + "epoch": 0.81, + "learning_rate": 3.496932021259889e-08, + "loss": 0.1586, + "step": 15607 + }, + { + "epoch": 0.81, + "learning_rate": 3.4950222177078124e-08, + "loss": 0.1318, + "step": 15608 + }, + { + "epoch": 0.81, + "learning_rate": 3.493112885874374e-08, + "loss": 0.137, + "step": 15609 + }, + { + "epoch": 0.81, + "learning_rate": 3.491204025814136e-08, + "loss": 0.1137, + "step": 15610 + }, + { + "epoch": 0.81, + "learning_rate": 3.489295637581666e-08, + "loss": 0.0961, + "step": 15611 + }, + { + "epoch": 0.81, + "learning_rate": 3.4873877212315004e-08, + "loss": 0.1518, + "step": 15612 + }, + { + "epoch": 0.81, + "learning_rate": 3.485480276818167e-08, + "loss": 0.1588, + "step": 15613 + }, + { + "epoch": 0.81, + "learning_rate": 3.483573304396179e-08, + "loss": 0.1347, + "step": 15614 + }, + { + "epoch": 0.82, + "learning_rate": 3.481666804020047e-08, + "loss": 0.1306, + "step": 15615 + }, + { + "epoch": 0.82, + "learning_rate": 3.479760775744258e-08, + "loss": 0.1313, + "step": 15616 + }, + { + "epoch": 0.82, + "learning_rate": 3.477855219623278e-08, + "loss": 0.1494, + "step": 15617 + }, + { + "epoch": 0.82, + "learning_rate": 3.475950135711572e-08, + "loss": 0.1377, + "step": 15618 + }, + { + "epoch": 0.82, + "learning_rate": 3.4740455240635914e-08, + "loss": 0.1419, + "step": 15619 + }, + { + "epoch": 0.82, + "learning_rate": 3.4721413847337713e-08, + "loss": 0.1435, + "step": 15620 + }, + { + "epoch": 0.82, + "learning_rate": 3.4702377177765315e-08, + "loss": 0.1738, + "step": 15621 + }, + { + "epoch": 0.82, + "learning_rate": 3.468334523246272e-08, + "loss": 0.1178, + "step": 15622 + }, + { + "epoch": 0.82, + "learning_rate": 3.4664318011974e-08, + "loss": 0.126, + "step": 15623 + }, + { + "epoch": 0.82, + "learning_rate": 3.464529551684294e-08, + "loss": 0.1633, + "step": 15624 + }, + { + "epoch": 0.82, + "learning_rate": 3.4626277747613085e-08, + "loss": 0.1323, + "step": 15625 + }, + { + "epoch": 0.82, + "learning_rate": 3.4607264704828085e-08, + "loss": 0.1112, + "step": 15626 + }, + { + "epoch": 0.82, + "learning_rate": 3.4588256389031315e-08, + "loss": 0.1287, + "step": 15627 + }, + { + "epoch": 0.82, + "learning_rate": 3.4569252800766033e-08, + "loss": 0.1374, + "step": 15628 + }, + { + "epoch": 0.82, + "learning_rate": 3.4550253940575315e-08, + "loss": 0.1799, + "step": 15629 + }, + { + "epoch": 0.82, + "learning_rate": 3.4531259809002266e-08, + "loss": 0.1149, + "step": 15630 + }, + { + "epoch": 0.82, + "learning_rate": 3.451227040658968e-08, + "loss": 0.1118, + "step": 15631 + }, + { + "epoch": 0.82, + "learning_rate": 3.449328573388028e-08, + "loss": 0.1201, + "step": 15632 + }, + { + "epoch": 0.82, + "learning_rate": 3.447430579141668e-08, + "loss": 0.1274, + "step": 15633 + }, + { + "epoch": 0.82, + "learning_rate": 3.4455330579741324e-08, + "loss": 0.1388, + "step": 15634 + }, + { + "epoch": 0.82, + "learning_rate": 3.4436360099396524e-08, + "loss": 0.1487, + "step": 15635 + }, + { + "epoch": 0.82, + "learning_rate": 3.441739435092441e-08, + "loss": 0.1416, + "step": 15636 + }, + { + "epoch": 0.82, + "learning_rate": 3.439843333486714e-08, + "loss": 0.1457, + "step": 15637 + }, + { + "epoch": 0.82, + "learning_rate": 3.437947705176658e-08, + "loss": 0.1326, + "step": 15638 + }, + { + "epoch": 0.82, + "learning_rate": 3.4360525502164505e-08, + "loss": 0.1612, + "step": 15639 + }, + { + "epoch": 0.82, + "learning_rate": 3.434157868660248e-08, + "loss": 0.1989, + "step": 15640 + }, + { + "epoch": 0.82, + "learning_rate": 3.432263660562216e-08, + "loss": 0.1151, + "step": 15641 + }, + { + "epoch": 0.82, + "learning_rate": 3.430369925976488e-08, + "loss": 0.1512, + "step": 15642 + }, + { + "epoch": 0.82, + "learning_rate": 3.4284766649571763e-08, + "loss": 0.1417, + "step": 15643 + }, + { + "epoch": 0.82, + "learning_rate": 3.426583877558402e-08, + "loss": 0.1334, + "step": 15644 + }, + { + "epoch": 0.82, + "learning_rate": 3.424691563834259e-08, + "loss": 0.1633, + "step": 15645 + }, + { + "epoch": 0.82, + "learning_rate": 3.4227997238388274e-08, + "loss": 0.1366, + "step": 15646 + }, + { + "epoch": 0.82, + "learning_rate": 3.4209083576261775e-08, + "loss": 0.1559, + "step": 15647 + }, + { + "epoch": 0.82, + "learning_rate": 3.419017465250369e-08, + "loss": 0.1287, + "step": 15648 + }, + { + "epoch": 0.82, + "learning_rate": 3.417127046765444e-08, + "loss": 0.1297, + "step": 15649 + }, + { + "epoch": 0.82, + "learning_rate": 3.41523710222543e-08, + "loss": 0.1707, + "step": 15650 + }, + { + "epoch": 0.82, + "learning_rate": 3.4133476316843425e-08, + "loss": 0.1389, + "step": 15651 + }, + { + "epoch": 0.82, + "learning_rate": 3.41145863519618e-08, + "loss": 0.0991, + "step": 15652 + }, + { + "epoch": 0.82, + "learning_rate": 3.4095701128149365e-08, + "loss": 0.1206, + "step": 15653 + }, + { + "epoch": 0.82, + "learning_rate": 3.407682064594577e-08, + "loss": 0.1279, + "step": 15654 + }, + { + "epoch": 0.82, + "learning_rate": 3.4057944905890754e-08, + "loss": 0.1241, + "step": 15655 + }, + { + "epoch": 0.82, + "learning_rate": 3.403907390852372e-08, + "loss": 0.1321, + "step": 15656 + }, + { + "epoch": 0.82, + "learning_rate": 3.4020207654384025e-08, + "loss": 0.1553, + "step": 15657 + }, + { + "epoch": 0.82, + "learning_rate": 3.400134614401082e-08, + "loss": 0.1241, + "step": 15658 + }, + { + "epoch": 0.82, + "learning_rate": 3.398248937794335e-08, + "loss": 0.127, + "step": 15659 + }, + { + "epoch": 0.82, + "learning_rate": 3.396363735672034e-08, + "loss": 0.1112, + "step": 15660 + }, + { + "epoch": 0.82, + "learning_rate": 3.3944790080880625e-08, + "loss": 0.1395, + "step": 15661 + }, + { + "epoch": 0.82, + "learning_rate": 3.3925947550962965e-08, + "loss": 0.1397, + "step": 15662 + }, + { + "epoch": 0.82, + "learning_rate": 3.3907109767505837e-08, + "loss": 0.1161, + "step": 15663 + }, + { + "epoch": 0.82, + "learning_rate": 3.388827673104764e-08, + "loss": 0.1344, + "step": 15664 + }, + { + "epoch": 0.82, + "learning_rate": 3.3869448442126556e-08, + "loss": 0.1088, + "step": 15665 + }, + { + "epoch": 0.82, + "learning_rate": 3.385062490128083e-08, + "loss": 0.1105, + "step": 15666 + }, + { + "epoch": 0.82, + "learning_rate": 3.383180610904843e-08, + "loss": 0.0993, + "step": 15667 + }, + { + "epoch": 0.82, + "learning_rate": 3.381299206596706e-08, + "loss": 0.1575, + "step": 15668 + }, + { + "epoch": 0.82, + "learning_rate": 3.379418277257457e-08, + "loss": 0.1187, + "step": 15669 + }, + { + "epoch": 0.82, + "learning_rate": 3.37753782294085e-08, + "loss": 0.1574, + "step": 15670 + }, + { + "epoch": 0.82, + "learning_rate": 3.37565784370063e-08, + "loss": 0.1176, + "step": 15671 + }, + { + "epoch": 0.82, + "learning_rate": 3.3737783395905226e-08, + "loss": 0.1108, + "step": 15672 + }, + { + "epoch": 0.82, + "learning_rate": 3.371899310664253e-08, + "loss": 0.1164, + "step": 15673 + }, + { + "epoch": 0.82, + "learning_rate": 3.370020756975518e-08, + "loss": 0.1164, + "step": 15674 + }, + { + "epoch": 0.82, + "learning_rate": 3.3681426785780096e-08, + "loss": 0.129, + "step": 15675 + }, + { + "epoch": 0.82, + "learning_rate": 3.366265075525407e-08, + "loss": 0.1157, + "step": 15676 + }, + { + "epoch": 0.82, + "learning_rate": 3.364387947871368e-08, + "loss": 0.1117, + "step": 15677 + }, + { + "epoch": 0.82, + "learning_rate": 3.362511295669543e-08, + "loss": 0.1245, + "step": 15678 + }, + { + "epoch": 0.82, + "learning_rate": 3.3606351189735624e-08, + "loss": 0.1584, + "step": 15679 + }, + { + "epoch": 0.82, + "learning_rate": 3.3587594178370606e-08, + "loss": 0.1258, + "step": 15680 + }, + { + "epoch": 0.82, + "learning_rate": 3.356884192313638e-08, + "loss": 0.1241, + "step": 15681 + }, + { + "epoch": 0.82, + "learning_rate": 3.355009442456889e-08, + "loss": 0.1402, + "step": 15682 + }, + { + "epoch": 0.82, + "learning_rate": 3.353135168320391e-08, + "loss": 0.1192, + "step": 15683 + }, + { + "epoch": 0.82, + "learning_rate": 3.351261369957727e-08, + "loss": 0.1216, + "step": 15684 + }, + { + "epoch": 0.82, + "learning_rate": 3.349388047422437e-08, + "loss": 0.1293, + "step": 15685 + }, + { + "epoch": 0.82, + "learning_rate": 3.347515200768056e-08, + "loss": 0.1187, + "step": 15686 + }, + { + "epoch": 0.82, + "learning_rate": 3.3456428300481255e-08, + "loss": 0.1401, + "step": 15687 + }, + { + "epoch": 0.82, + "learning_rate": 3.343770935316153e-08, + "loss": 0.1409, + "step": 15688 + }, + { + "epoch": 0.82, + "learning_rate": 3.3418995166256346e-08, + "loss": 0.1265, + "step": 15689 + }, + { + "epoch": 0.82, + "learning_rate": 3.340028574030056e-08, + "loss": 0.1299, + "step": 15690 + }, + { + "epoch": 0.82, + "learning_rate": 3.338158107582896e-08, + "loss": 0.1804, + "step": 15691 + }, + { + "epoch": 0.82, + "learning_rate": 3.336288117337609e-08, + "loss": 0.1506, + "step": 15692 + }, + { + "epoch": 0.82, + "learning_rate": 3.334418603347642e-08, + "loss": 0.112, + "step": 15693 + }, + { + "epoch": 0.82, + "learning_rate": 3.332549565666425e-08, + "loss": 0.1343, + "step": 15694 + }, + { + "epoch": 0.82, + "learning_rate": 3.3306810043473735e-08, + "loss": 0.1453, + "step": 15695 + }, + { + "epoch": 0.82, + "learning_rate": 3.328812919443895e-08, + "loss": 0.1292, + "step": 15696 + }, + { + "epoch": 0.82, + "learning_rate": 3.326945311009375e-08, + "loss": 0.1376, + "step": 15697 + }, + { + "epoch": 0.82, + "learning_rate": 3.325078179097198e-08, + "loss": 0.1152, + "step": 15698 + }, + { + "epoch": 0.82, + "learning_rate": 3.323211523760725e-08, + "loss": 0.1468, + "step": 15699 + }, + { + "epoch": 0.82, + "learning_rate": 3.3213453450533036e-08, + "loss": 0.149, + "step": 15700 + }, + { + "epoch": 0.82, + "learning_rate": 3.3194796430282713e-08, + "loss": 0.1537, + "step": 15701 + }, + { + "epoch": 0.82, + "learning_rate": 3.3176144177389496e-08, + "loss": 0.1369, + "step": 15702 + }, + { + "epoch": 0.82, + "learning_rate": 3.31574966923865e-08, + "loss": 0.1447, + "step": 15703 + }, + { + "epoch": 0.82, + "learning_rate": 3.313885397580658e-08, + "loss": 0.12, + "step": 15704 + }, + { + "epoch": 0.82, + "learning_rate": 3.312021602818269e-08, + "loss": 0.1415, + "step": 15705 + }, + { + "epoch": 0.82, + "learning_rate": 3.310158285004745e-08, + "loss": 0.1352, + "step": 15706 + }, + { + "epoch": 0.82, + "learning_rate": 3.308295444193338e-08, + "loss": 0.1284, + "step": 15707 + }, + { + "epoch": 0.82, + "learning_rate": 3.30643308043729e-08, + "loss": 0.1377, + "step": 15708 + }, + { + "epoch": 0.82, + "learning_rate": 3.3045711937898315e-08, + "loss": 0.1406, + "step": 15709 + }, + { + "epoch": 0.82, + "learning_rate": 3.3027097843041764e-08, + "loss": 0.1154, + "step": 15710 + }, + { + "epoch": 0.82, + "learning_rate": 3.300848852033513e-08, + "loss": 0.1327, + "step": 15711 + }, + { + "epoch": 0.82, + "learning_rate": 3.298988397031042e-08, + "loss": 0.1293, + "step": 15712 + }, + { + "epoch": 0.82, + "learning_rate": 3.2971284193499306e-08, + "loss": 0.1379, + "step": 15713 + }, + { + "epoch": 0.82, + "learning_rate": 3.295268919043335e-08, + "loss": 0.1798, + "step": 15714 + }, + { + "epoch": 0.82, + "learning_rate": 3.293409896164398e-08, + "loss": 0.1656, + "step": 15715 + }, + { + "epoch": 0.82, + "learning_rate": 3.2915513507662595e-08, + "loss": 0.1331, + "step": 15716 + }, + { + "epoch": 0.82, + "learning_rate": 3.289693282902035e-08, + "loss": 0.1197, + "step": 15717 + }, + { + "epoch": 0.82, + "learning_rate": 3.2878356926248275e-08, + "loss": 0.1195, + "step": 15718 + }, + { + "epoch": 0.82, + "learning_rate": 3.285978579987727e-08, + "loss": 0.1375, + "step": 15719 + }, + { + "epoch": 0.82, + "learning_rate": 3.2841219450438095e-08, + "loss": 0.1791, + "step": 15720 + }, + { + "epoch": 0.82, + "learning_rate": 3.282265787846141e-08, + "loss": 0.1369, + "step": 15721 + }, + { + "epoch": 0.82, + "learning_rate": 3.280410108447764e-08, + "loss": 0.1309, + "step": 15722 + }, + { + "epoch": 0.82, + "learning_rate": 3.278554906901727e-08, + "loss": 0.1689, + "step": 15723 + }, + { + "epoch": 0.82, + "learning_rate": 3.276700183261045e-08, + "loss": 0.1222, + "step": 15724 + }, + { + "epoch": 0.82, + "learning_rate": 3.27484593757873e-08, + "loss": 0.1167, + "step": 15725 + }, + { + "epoch": 0.82, + "learning_rate": 3.2729921699077666e-08, + "loss": 0.1425, + "step": 15726 + }, + { + "epoch": 0.82, + "learning_rate": 3.271138880301154e-08, + "loss": 0.1603, + "step": 15727 + }, + { + "epoch": 0.82, + "learning_rate": 3.269286068811847e-08, + "loss": 0.1346, + "step": 15728 + }, + { + "epoch": 0.82, + "learning_rate": 3.267433735492804e-08, + "loss": 0.1243, + "step": 15729 + }, + { + "epoch": 0.82, + "learning_rate": 3.2655818803969573e-08, + "loss": 0.1358, + "step": 15730 + }, + { + "epoch": 0.82, + "learning_rate": 3.263730503577244e-08, + "loss": 0.1274, + "step": 15731 + }, + { + "epoch": 0.82, + "learning_rate": 3.261879605086577e-08, + "loss": 0.1318, + "step": 15732 + }, + { + "epoch": 0.82, + "learning_rate": 3.26002918497785e-08, + "loss": 0.1274, + "step": 15733 + }, + { + "epoch": 0.82, + "learning_rate": 3.258179243303949e-08, + "loss": 0.1351, + "step": 15734 + }, + { + "epoch": 0.82, + "learning_rate": 3.2563297801177506e-08, + "loss": 0.1263, + "step": 15735 + }, + { + "epoch": 0.82, + "learning_rate": 3.254480795472116e-08, + "loss": 0.1543, + "step": 15736 + }, + { + "epoch": 0.82, + "learning_rate": 3.2526322894198766e-08, + "loss": 0.1544, + "step": 15737 + }, + { + "epoch": 0.82, + "learning_rate": 3.2507842620138746e-08, + "loss": 0.117, + "step": 15738 + }, + { + "epoch": 0.82, + "learning_rate": 3.248936713306925e-08, + "loss": 0.1257, + "step": 15739 + }, + { + "epoch": 0.82, + "learning_rate": 3.2470896433518305e-08, + "loss": 0.1198, + "step": 15740 + }, + { + "epoch": 0.82, + "learning_rate": 3.245243052201376e-08, + "loss": 0.1156, + "step": 15741 + }, + { + "epoch": 0.82, + "learning_rate": 3.2433969399083495e-08, + "loss": 0.0999, + "step": 15742 + }, + { + "epoch": 0.82, + "learning_rate": 3.241551306525505e-08, + "loss": 0.1142, + "step": 15743 + }, + { + "epoch": 0.82, + "learning_rate": 3.239706152105595e-08, + "loss": 0.1583, + "step": 15744 + }, + { + "epoch": 0.82, + "learning_rate": 3.237861476701354e-08, + "loss": 0.167, + "step": 15745 + }, + { + "epoch": 0.82, + "learning_rate": 3.2360172803654996e-08, + "loss": 0.1422, + "step": 15746 + }, + { + "epoch": 0.82, + "learning_rate": 3.234173563150746e-08, + "loss": 0.1416, + "step": 15747 + }, + { + "epoch": 0.82, + "learning_rate": 3.232330325109778e-08, + "loss": 0.1366, + "step": 15748 + }, + { + "epoch": 0.82, + "learning_rate": 3.230487566295288e-08, + "loss": 0.1022, + "step": 15749 + }, + { + "epoch": 0.82, + "learning_rate": 3.2286452867599364e-08, + "loss": 0.1556, + "step": 15750 + }, + { + "epoch": 0.82, + "learning_rate": 3.226803486556378e-08, + "loss": 0.1203, + "step": 15751 + }, + { + "epoch": 0.82, + "learning_rate": 3.224962165737244e-08, + "loss": 0.1266, + "step": 15752 + }, + { + "epoch": 0.82, + "learning_rate": 3.223121324355178e-08, + "loss": 0.1257, + "step": 15753 + }, + { + "epoch": 0.82, + "learning_rate": 3.221280962462776e-08, + "loss": 0.1417, + "step": 15754 + }, + { + "epoch": 0.82, + "learning_rate": 3.219441080112637e-08, + "loss": 0.1084, + "step": 15755 + }, + { + "epoch": 0.82, + "learning_rate": 3.217601677357355e-08, + "loss": 0.1268, + "step": 15756 + }, + { + "epoch": 0.82, + "learning_rate": 3.2157627542494956e-08, + "loss": 0.1177, + "step": 15757 + }, + { + "epoch": 0.82, + "learning_rate": 3.2139243108416156e-08, + "loss": 0.1498, + "step": 15758 + }, + { + "epoch": 0.82, + "learning_rate": 3.212086347186253e-08, + "loss": 0.1575, + "step": 15759 + }, + { + "epoch": 0.82, + "learning_rate": 3.210248863335949e-08, + "loss": 0.1081, + "step": 15760 + }, + { + "epoch": 0.82, + "learning_rate": 3.2084118593432205e-08, + "loss": 0.1368, + "step": 15761 + }, + { + "epoch": 0.82, + "learning_rate": 3.206575335260553e-08, + "loss": 0.1271, + "step": 15762 + }, + { + "epoch": 0.82, + "learning_rate": 3.204739291140448e-08, + "loss": 0.1696, + "step": 15763 + }, + { + "epoch": 0.82, + "learning_rate": 3.2029037270353796e-08, + "loss": 0.1592, + "step": 15764 + }, + { + "epoch": 0.82, + "learning_rate": 3.2010686429978065e-08, + "loss": 0.13, + "step": 15765 + }, + { + "epoch": 0.82, + "learning_rate": 3.199234039080172e-08, + "loss": 0.1287, + "step": 15766 + }, + { + "epoch": 0.82, + "learning_rate": 3.1973999153349196e-08, + "loss": 0.144, + "step": 15767 + }, + { + "epoch": 0.82, + "learning_rate": 3.1955662718144625e-08, + "loss": 0.1226, + "step": 15768 + }, + { + "epoch": 0.82, + "learning_rate": 3.1937331085712104e-08, + "loss": 0.1018, + "step": 15769 + }, + { + "epoch": 0.82, + "learning_rate": 3.191900425657552e-08, + "loss": 0.105, + "step": 15770 + }, + { + "epoch": 0.82, + "learning_rate": 3.19006822312587e-08, + "loss": 0.1229, + "step": 15771 + }, + { + "epoch": 0.82, + "learning_rate": 3.188236501028525e-08, + "loss": 0.1246, + "step": 15772 + }, + { + "epoch": 0.82, + "learning_rate": 3.186405259417868e-08, + "loss": 0.1398, + "step": 15773 + }, + { + "epoch": 0.82, + "learning_rate": 3.1845744983462444e-08, + "loss": 0.12, + "step": 15774 + }, + { + "epoch": 0.82, + "learning_rate": 3.1827442178659716e-08, + "loss": 0.1114, + "step": 15775 + }, + { + "epoch": 0.82, + "learning_rate": 3.180914418029361e-08, + "loss": 0.1222, + "step": 15776 + }, + { + "epoch": 0.82, + "learning_rate": 3.179085098888707e-08, + "loss": 0.132, + "step": 15777 + }, + { + "epoch": 0.82, + "learning_rate": 3.177256260496297e-08, + "loss": 0.1157, + "step": 15778 + }, + { + "epoch": 0.82, + "learning_rate": 3.175427902904402e-08, + "loss": 0.1463, + "step": 15779 + }, + { + "epoch": 0.82, + "learning_rate": 3.173600026165264e-08, + "loss": 0.2358, + "step": 15780 + }, + { + "epoch": 0.82, + "learning_rate": 3.171772630331137e-08, + "loss": 0.1303, + "step": 15781 + }, + { + "epoch": 0.82, + "learning_rate": 3.169945715454245e-08, + "loss": 0.1266, + "step": 15782 + }, + { + "epoch": 0.82, + "learning_rate": 3.168119281586801e-08, + "loss": 0.1612, + "step": 15783 + }, + { + "epoch": 0.82, + "learning_rate": 3.166293328781e-08, + "loss": 0.0989, + "step": 15784 + }, + { + "epoch": 0.82, + "learning_rate": 3.164467857089041e-08, + "loss": 0.1357, + "step": 15785 + }, + { + "epoch": 0.82, + "learning_rate": 3.16264286656309e-08, + "loss": 0.1364, + "step": 15786 + }, + { + "epoch": 0.82, + "learning_rate": 3.1608183572553035e-08, + "loss": 0.171, + "step": 15787 + }, + { + "epoch": 0.82, + "learning_rate": 3.15899432921783e-08, + "loss": 0.1929, + "step": 15788 + }, + { + "epoch": 0.82, + "learning_rate": 3.1571707825028006e-08, + "loss": 0.1408, + "step": 15789 + }, + { + "epoch": 0.82, + "learning_rate": 3.155347717162331e-08, + "loss": 0.1145, + "step": 15790 + }, + { + "epoch": 0.82, + "learning_rate": 3.153525133248522e-08, + "loss": 0.1206, + "step": 15791 + }, + { + "epoch": 0.82, + "learning_rate": 3.151703030813473e-08, + "loss": 0.1363, + "step": 15792 + }, + { + "epoch": 0.82, + "learning_rate": 3.149881409909254e-08, + "loss": 0.1521, + "step": 15793 + }, + { + "epoch": 0.82, + "learning_rate": 3.148060270587931e-08, + "loss": 0.1834, + "step": 15794 + }, + { + "epoch": 0.82, + "learning_rate": 3.146239612901547e-08, + "loss": 0.1206, + "step": 15795 + }, + { + "epoch": 0.82, + "learning_rate": 3.144419436902148e-08, + "loss": 0.1469, + "step": 15796 + }, + { + "epoch": 0.82, + "learning_rate": 3.142599742641745e-08, + "loss": 0.1288, + "step": 15797 + }, + { + "epoch": 0.82, + "learning_rate": 3.1407805301723446e-08, + "loss": 0.1276, + "step": 15798 + }, + { + "epoch": 0.82, + "learning_rate": 3.1389617995459496e-08, + "loss": 0.1182, + "step": 15799 + }, + { + "epoch": 0.82, + "learning_rate": 3.137143550814534e-08, + "loss": 0.1653, + "step": 15800 + }, + { + "epoch": 0.82, + "learning_rate": 3.135325784030067e-08, + "loss": 0.1266, + "step": 15801 + }, + { + "epoch": 0.82, + "learning_rate": 3.133508499244495e-08, + "loss": 0.1684, + "step": 15802 + }, + { + "epoch": 0.82, + "learning_rate": 3.131691696509764e-08, + "loss": 0.1802, + "step": 15803 + }, + { + "epoch": 0.82, + "learning_rate": 3.1298753758778015e-08, + "loss": 0.1161, + "step": 15804 + }, + { + "epoch": 0.82, + "learning_rate": 3.128059537400505e-08, + "loss": 0.1072, + "step": 15805 + }, + { + "epoch": 0.83, + "learning_rate": 3.126244181129785e-08, + "loss": 0.1425, + "step": 15806 + }, + { + "epoch": 0.83, + "learning_rate": 3.12442930711752e-08, + "loss": 0.1221, + "step": 15807 + }, + { + "epoch": 0.83, + "learning_rate": 3.122614915415578e-08, + "loss": 0.1261, + "step": 15808 + }, + { + "epoch": 0.83, + "learning_rate": 3.120801006075815e-08, + "loss": 0.137, + "step": 15809 + }, + { + "epoch": 0.83, + "learning_rate": 3.1189875791500786e-08, + "loss": 0.1142, + "step": 15810 + }, + { + "epoch": 0.83, + "learning_rate": 3.117174634690194e-08, + "loss": 0.1083, + "step": 15811 + }, + { + "epoch": 0.83, + "learning_rate": 3.115362172747975e-08, + "loss": 0.1093, + "step": 15812 + }, + { + "epoch": 0.83, + "learning_rate": 3.113550193375225e-08, + "loss": 0.1154, + "step": 15813 + }, + { + "epoch": 0.83, + "learning_rate": 3.1117386966237294e-08, + "loss": 0.1361, + "step": 15814 + }, + { + "epoch": 0.83, + "learning_rate": 3.1099276825452594e-08, + "loss": 0.1465, + "step": 15815 + }, + { + "epoch": 0.83, + "learning_rate": 3.108117151191574e-08, + "loss": 0.1227, + "step": 15816 + }, + { + "epoch": 0.83, + "learning_rate": 3.1063071026144226e-08, + "loss": 0.1357, + "step": 15817 + }, + { + "epoch": 0.83, + "learning_rate": 3.10449753686554e-08, + "loss": 0.1389, + "step": 15818 + }, + { + "epoch": 0.83, + "learning_rate": 3.102688453996638e-08, + "loss": 0.1313, + "step": 15819 + }, + { + "epoch": 0.83, + "learning_rate": 3.1008798540594173e-08, + "loss": 0.1446, + "step": 15820 + }, + { + "epoch": 0.83, + "learning_rate": 3.0990717371055854e-08, + "loss": 0.1496, + "step": 15821 + }, + { + "epoch": 0.83, + "learning_rate": 3.097264103186801e-08, + "loss": 0.1561, + "step": 15822 + }, + { + "epoch": 0.83, + "learning_rate": 3.095456952354731e-08, + "loss": 0.1568, + "step": 15823 + }, + { + "epoch": 0.83, + "learning_rate": 3.093650284661031e-08, + "loss": 0.1563, + "step": 15824 + }, + { + "epoch": 0.83, + "learning_rate": 3.091844100157333e-08, + "loss": 0.1542, + "step": 15825 + }, + { + "epoch": 0.83, + "learning_rate": 3.090038398895256e-08, + "loss": 0.1669, + "step": 15826 + }, + { + "epoch": 0.83, + "learning_rate": 3.0882331809264074e-08, + "loss": 0.1278, + "step": 15827 + }, + { + "epoch": 0.83, + "learning_rate": 3.0864284463023846e-08, + "loss": 0.1405, + "step": 15828 + }, + { + "epoch": 0.83, + "learning_rate": 3.0846241950747676e-08, + "loss": 0.1323, + "step": 15829 + }, + { + "epoch": 0.83, + "learning_rate": 3.082820427295123e-08, + "loss": 0.1195, + "step": 15830 + }, + { + "epoch": 0.83, + "learning_rate": 3.0810171430149987e-08, + "loss": 0.132, + "step": 15831 + }, + { + "epoch": 0.83, + "learning_rate": 3.079214342285936e-08, + "loss": 0.1241, + "step": 15832 + }, + { + "epoch": 0.83, + "learning_rate": 3.0774120251594607e-08, + "loss": 0.1375, + "step": 15833 + }, + { + "epoch": 0.83, + "learning_rate": 3.075610191687077e-08, + "loss": 0.1564, + "step": 15834 + }, + { + "epoch": 0.83, + "learning_rate": 3.0738088419202934e-08, + "loss": 0.1131, + "step": 15835 + }, + { + "epoch": 0.83, + "learning_rate": 3.072007975910587e-08, + "loss": 0.1315, + "step": 15836 + }, + { + "epoch": 0.83, + "learning_rate": 3.070207593709426e-08, + "loss": 0.1255, + "step": 15837 + }, + { + "epoch": 0.83, + "learning_rate": 3.0684076953682624e-08, + "loss": 0.1185, + "step": 15838 + }, + { + "epoch": 0.83, + "learning_rate": 3.066608280938554e-08, + "loss": 0.1321, + "step": 15839 + }, + { + "epoch": 0.83, + "learning_rate": 3.064809350471713e-08, + "loss": 0.1647, + "step": 15840 + }, + { + "epoch": 0.83, + "learning_rate": 3.063010904019159e-08, + "loss": 0.1168, + "step": 15841 + }, + { + "epoch": 0.83, + "learning_rate": 3.0612129416322875e-08, + "loss": 0.1611, + "step": 15842 + }, + { + "epoch": 0.83, + "learning_rate": 3.0594154633624914e-08, + "loss": 0.1447, + "step": 15843 + }, + { + "epoch": 0.83, + "learning_rate": 3.057618469261143e-08, + "loss": 0.1338, + "step": 15844 + }, + { + "epoch": 0.83, + "learning_rate": 3.0558219593795985e-08, + "loss": 0.1331, + "step": 15845 + }, + { + "epoch": 0.83, + "learning_rate": 3.054025933769197e-08, + "loss": 0.1318, + "step": 15846 + }, + { + "epoch": 0.83, + "learning_rate": 3.05223039248129e-08, + "loss": 0.1397, + "step": 15847 + }, + { + "epoch": 0.83, + "learning_rate": 3.050435335567172e-08, + "loss": 0.1386, + "step": 15848 + }, + { + "epoch": 0.83, + "learning_rate": 3.048640763078152e-08, + "loss": 0.1344, + "step": 15849 + }, + { + "epoch": 0.83, + "learning_rate": 3.0468466750655284e-08, + "loss": 0.1339, + "step": 15850 + }, + { + "epoch": 0.83, + "learning_rate": 3.04505307158057e-08, + "loss": 0.1294, + "step": 15851 + }, + { + "epoch": 0.83, + "learning_rate": 3.043259952674542e-08, + "loss": 0.1354, + "step": 15852 + }, + { + "epoch": 0.83, + "learning_rate": 3.041467318398685e-08, + "loss": 0.1502, + "step": 15853 + }, + { + "epoch": 0.83, + "learning_rate": 3.0396751688042435e-08, + "loss": 0.1477, + "step": 15854 + }, + { + "epoch": 0.83, + "learning_rate": 3.037883503942431e-08, + "loss": 0.1352, + "step": 15855 + }, + { + "epoch": 0.83, + "learning_rate": 3.0360923238644586e-08, + "loss": 0.1317, + "step": 15856 + }, + { + "epoch": 0.83, + "learning_rate": 3.034301628621514e-08, + "loss": 0.1452, + "step": 15857 + }, + { + "epoch": 0.83, + "learning_rate": 3.0325114182647804e-08, + "loss": 0.1307, + "step": 15858 + }, + { + "epoch": 0.83, + "learning_rate": 3.030721692845419e-08, + "loss": 0.1266, + "step": 15859 + }, + { + "epoch": 0.83, + "learning_rate": 3.0289324524145786e-08, + "loss": 0.1259, + "step": 15860 + }, + { + "epoch": 0.83, + "learning_rate": 3.027143697023404e-08, + "loss": 0.1219, + "step": 15861 + }, + { + "epoch": 0.83, + "learning_rate": 3.025355426723015e-08, + "loss": 0.1285, + "step": 15862 + }, + { + "epoch": 0.83, + "learning_rate": 3.023567641564522e-08, + "loss": 0.1553, + "step": 15863 + }, + { + "epoch": 0.83, + "learning_rate": 3.021780341599018e-08, + "loss": 0.1611, + "step": 15864 + }, + { + "epoch": 0.83, + "learning_rate": 3.019993526877589e-08, + "loss": 0.1337, + "step": 15865 + }, + { + "epoch": 0.83, + "learning_rate": 3.0182071974512966e-08, + "loss": 0.1103, + "step": 15866 + }, + { + "epoch": 0.83, + "learning_rate": 3.016421353371195e-08, + "loss": 0.1455, + "step": 15867 + }, + { + "epoch": 0.83, + "learning_rate": 3.0146359946883305e-08, + "loss": 0.1131, + "step": 15868 + }, + { + "epoch": 0.83, + "learning_rate": 3.012851121453728e-08, + "loss": 0.1321, + "step": 15869 + }, + { + "epoch": 0.83, + "learning_rate": 3.011066733718397e-08, + "loss": 0.1772, + "step": 15870 + }, + { + "epoch": 0.83, + "learning_rate": 3.009282831533333e-08, + "loss": 0.1522, + "step": 15871 + }, + { + "epoch": 0.83, + "learning_rate": 3.00749941494953e-08, + "loss": 0.1608, + "step": 15872 + }, + { + "epoch": 0.83, + "learning_rate": 3.005716484017957e-08, + "loss": 0.1179, + "step": 15873 + }, + { + "epoch": 0.83, + "learning_rate": 3.003934038789559e-08, + "loss": 0.1093, + "step": 15874 + }, + { + "epoch": 0.83, + "learning_rate": 3.002152079315292e-08, + "loss": 0.1571, + "step": 15875 + }, + { + "epoch": 0.83, + "learning_rate": 3.000370605646079e-08, + "loss": 0.1312, + "step": 15876 + }, + { + "epoch": 0.83, + "learning_rate": 2.998589617832838e-08, + "loss": 0.1354, + "step": 15877 + }, + { + "epoch": 0.83, + "learning_rate": 2.996809115926464e-08, + "loss": 0.1487, + "step": 15878 + }, + { + "epoch": 0.83, + "learning_rate": 2.9950290999778524e-08, + "loss": 0.1102, + "step": 15879 + }, + { + "epoch": 0.83, + "learning_rate": 2.993249570037877e-08, + "loss": 0.1598, + "step": 15880 + }, + { + "epoch": 0.83, + "learning_rate": 2.9914705261573914e-08, + "loss": 0.162, + "step": 15881 + }, + { + "epoch": 0.83, + "learning_rate": 2.989691968387245e-08, + "loss": 0.151, + "step": 15882 + }, + { + "epoch": 0.83, + "learning_rate": 2.987913896778269e-08, + "loss": 0.13, + "step": 15883 + }, + { + "epoch": 0.83, + "learning_rate": 2.98613631138128e-08, + "loss": 0.1097, + "step": 15884 + }, + { + "epoch": 0.83, + "learning_rate": 2.984359212247081e-08, + "loss": 0.1411, + "step": 15885 + }, + { + "epoch": 0.83, + "learning_rate": 2.982582599426466e-08, + "loss": 0.1495, + "step": 15886 + }, + { + "epoch": 0.83, + "learning_rate": 2.980806472970212e-08, + "loss": 0.1166, + "step": 15887 + }, + { + "epoch": 0.83, + "learning_rate": 2.979030832929079e-08, + "loss": 0.198, + "step": 15888 + }, + { + "epoch": 0.83, + "learning_rate": 2.9772556793538094e-08, + "loss": 0.1168, + "step": 15889 + }, + { + "epoch": 0.83, + "learning_rate": 2.9754810122951534e-08, + "loss": 0.1428, + "step": 15890 + }, + { + "epoch": 0.83, + "learning_rate": 2.9737068318038196e-08, + "loss": 0.1486, + "step": 15891 + }, + { + "epoch": 0.83, + "learning_rate": 2.9719331379305112e-08, + "loss": 0.1793, + "step": 15892 + }, + { + "epoch": 0.83, + "learning_rate": 2.9701599307259307e-08, + "loss": 0.1272, + "step": 15893 + }, + { + "epoch": 0.83, + "learning_rate": 2.968387210240755e-08, + "loss": 0.1166, + "step": 15894 + }, + { + "epoch": 0.83, + "learning_rate": 2.9666149765256476e-08, + "loss": 0.1318, + "step": 15895 + }, + { + "epoch": 0.83, + "learning_rate": 2.9648432296312554e-08, + "loss": 0.1428, + "step": 15896 + }, + { + "epoch": 0.83, + "learning_rate": 2.9630719696082242e-08, + "loss": 0.1279, + "step": 15897 + }, + { + "epoch": 0.83, + "learning_rate": 2.9613011965071732e-08, + "loss": 0.1729, + "step": 15898 + }, + { + "epoch": 0.83, + "learning_rate": 2.9595309103787113e-08, + "loss": 0.1192, + "step": 15899 + }, + { + "epoch": 0.83, + "learning_rate": 2.9577611112734335e-08, + "loss": 0.1251, + "step": 15900 + }, + { + "epoch": 0.83, + "learning_rate": 2.9559917992419235e-08, + "loss": 0.1078, + "step": 15901 + }, + { + "epoch": 0.83, + "learning_rate": 2.954222974334748e-08, + "loss": 0.1513, + "step": 15902 + }, + { + "epoch": 0.83, + "learning_rate": 2.952454636602455e-08, + "loss": 0.0939, + "step": 15903 + }, + { + "epoch": 0.83, + "learning_rate": 2.950686786095593e-08, + "loss": 0.12, + "step": 15904 + }, + { + "epoch": 0.83, + "learning_rate": 2.9489194228646842e-08, + "loss": 0.1611, + "step": 15905 + }, + { + "epoch": 0.83, + "learning_rate": 2.947152546960241e-08, + "loss": 0.1307, + "step": 15906 + }, + { + "epoch": 0.83, + "learning_rate": 2.9453861584327633e-08, + "loss": 0.1244, + "step": 15907 + }, + { + "epoch": 0.83, + "learning_rate": 2.9436202573327308e-08, + "loss": 0.1396, + "step": 15908 + }, + { + "epoch": 0.83, + "learning_rate": 2.941854843710616e-08, + "loss": 0.1646, + "step": 15909 + }, + { + "epoch": 0.83, + "learning_rate": 2.9400899176168702e-08, + "loss": 0.1307, + "step": 15910 + }, + { + "epoch": 0.83, + "learning_rate": 2.9383254791019463e-08, + "loss": 0.1356, + "step": 15911 + }, + { + "epoch": 0.83, + "learning_rate": 2.9365615282162635e-08, + "loss": 0.1392, + "step": 15912 + }, + { + "epoch": 0.83, + "learning_rate": 2.9347980650102422e-08, + "loss": 0.1479, + "step": 15913 + }, + { + "epoch": 0.83, + "learning_rate": 2.9330350895342748e-08, + "loss": 0.1493, + "step": 15914 + }, + { + "epoch": 0.83, + "learning_rate": 2.931272601838759e-08, + "loss": 0.1529, + "step": 15915 + }, + { + "epoch": 0.83, + "learning_rate": 2.9295106019740635e-08, + "loss": 0.1277, + "step": 15916 + }, + { + "epoch": 0.83, + "learning_rate": 2.9277490899905365e-08, + "loss": 0.1291, + "step": 15917 + }, + { + "epoch": 0.83, + "learning_rate": 2.925988065938536e-08, + "loss": 0.1189, + "step": 15918 + }, + { + "epoch": 0.83, + "learning_rate": 2.9242275298683884e-08, + "loss": 0.0907, + "step": 15919 + }, + { + "epoch": 0.83, + "learning_rate": 2.9224674818304085e-08, + "loss": 0.1084, + "step": 15920 + }, + { + "epoch": 0.83, + "learning_rate": 2.9207079218748965e-08, + "loss": 0.1392, + "step": 15921 + }, + { + "epoch": 0.83, + "learning_rate": 2.9189488500521497e-08, + "loss": 0.1604, + "step": 15922 + }, + { + "epoch": 0.83, + "learning_rate": 2.917190266412439e-08, + "loss": 0.1192, + "step": 15923 + }, + { + "epoch": 0.83, + "learning_rate": 2.915432171006027e-08, + "loss": 0.1039, + "step": 15924 + }, + { + "epoch": 0.83, + "learning_rate": 2.9136745638831573e-08, + "loss": 0.0935, + "step": 15925 + }, + { + "epoch": 0.83, + "learning_rate": 2.911917445094063e-08, + "loss": 0.133, + "step": 15926 + }, + { + "epoch": 0.83, + "learning_rate": 2.9101608146889667e-08, + "loss": 0.1225, + "step": 15927 + }, + { + "epoch": 0.83, + "learning_rate": 2.9084046727180656e-08, + "loss": 0.1542, + "step": 15928 + }, + { + "epoch": 0.83, + "learning_rate": 2.906649019231564e-08, + "loss": 0.1376, + "step": 15929 + }, + { + "epoch": 0.83, + "learning_rate": 2.904893854279631e-08, + "loss": 0.122, + "step": 15930 + }, + { + "epoch": 0.83, + "learning_rate": 2.9031391779124303e-08, + "loss": 0.1548, + "step": 15931 + }, + { + "epoch": 0.83, + "learning_rate": 2.9013849901801068e-08, + "loss": 0.1857, + "step": 15932 + }, + { + "epoch": 0.83, + "learning_rate": 2.8996312911328112e-08, + "loss": 0.1825, + "step": 15933 + }, + { + "epoch": 0.83, + "learning_rate": 2.8978780808206506e-08, + "loss": 0.1489, + "step": 15934 + }, + { + "epoch": 0.83, + "learning_rate": 2.8961253592937305e-08, + "loss": 0.1078, + "step": 15935 + }, + { + "epoch": 0.83, + "learning_rate": 2.8943731266021564e-08, + "loss": 0.1457, + "step": 15936 + }, + { + "epoch": 0.83, + "learning_rate": 2.892621382796001e-08, + "loss": 0.1045, + "step": 15937 + }, + { + "epoch": 0.83, + "learning_rate": 2.890870127925331e-08, + "loss": 0.113, + "step": 15938 + }, + { + "epoch": 0.83, + "learning_rate": 2.8891193620401933e-08, + "loss": 0.1399, + "step": 15939 + }, + { + "epoch": 0.83, + "learning_rate": 2.887369085190634e-08, + "loss": 0.1532, + "step": 15940 + }, + { + "epoch": 0.83, + "learning_rate": 2.8856192974266758e-08, + "loss": 0.1147, + "step": 15941 + }, + { + "epoch": 0.83, + "learning_rate": 2.883869998798316e-08, + "loss": 0.128, + "step": 15942 + }, + { + "epoch": 0.83, + "learning_rate": 2.882121189355564e-08, + "loss": 0.1522, + "step": 15943 + }, + { + "epoch": 0.83, + "learning_rate": 2.8803728691483976e-08, + "loss": 0.1303, + "step": 15944 + }, + { + "epoch": 0.83, + "learning_rate": 2.878625038226783e-08, + "loss": 0.1192, + "step": 15945 + }, + { + "epoch": 0.83, + "learning_rate": 2.8768776966406693e-08, + "loss": 0.1595, + "step": 15946 + }, + { + "epoch": 0.83, + "learning_rate": 2.8751308444400057e-08, + "loss": 0.1209, + "step": 15947 + }, + { + "epoch": 0.83, + "learning_rate": 2.8733844816747143e-08, + "loss": 0.1454, + "step": 15948 + }, + { + "epoch": 0.83, + "learning_rate": 2.871638608394704e-08, + "loss": 0.1305, + "step": 15949 + }, + { + "epoch": 0.83, + "learning_rate": 2.8698932246498776e-08, + "loss": 0.1462, + "step": 15950 + }, + { + "epoch": 0.83, + "learning_rate": 2.8681483304901122e-08, + "loss": 0.1382, + "step": 15951 + }, + { + "epoch": 0.83, + "learning_rate": 2.8664039259652838e-08, + "loss": 0.1328, + "step": 15952 + }, + { + "epoch": 0.83, + "learning_rate": 2.864660011125244e-08, + "loss": 0.1368, + "step": 15953 + }, + { + "epoch": 0.83, + "learning_rate": 2.8629165860198325e-08, + "loss": 0.1226, + "step": 15954 + }, + { + "epoch": 0.83, + "learning_rate": 2.8611736506988848e-08, + "loss": 0.1437, + "step": 15955 + }, + { + "epoch": 0.83, + "learning_rate": 2.8594312052122104e-08, + "loss": 0.1175, + "step": 15956 + }, + { + "epoch": 0.83, + "learning_rate": 2.8576892496096074e-08, + "loss": 0.1034, + "step": 15957 + }, + { + "epoch": 0.83, + "learning_rate": 2.855947783940862e-08, + "loss": 0.1248, + "step": 15958 + }, + { + "epoch": 0.83, + "learning_rate": 2.8542068082557526e-08, + "loss": 0.1157, + "step": 15959 + }, + { + "epoch": 0.83, + "learning_rate": 2.85246632260403e-08, + "loss": 0.1531, + "step": 15960 + }, + { + "epoch": 0.83, + "learning_rate": 2.850726327035433e-08, + "loss": 0.1544, + "step": 15961 + }, + { + "epoch": 0.83, + "learning_rate": 2.8489868215997038e-08, + "loss": 0.103, + "step": 15962 + }, + { + "epoch": 0.83, + "learning_rate": 2.847247806346551e-08, + "loss": 0.1411, + "step": 15963 + }, + { + "epoch": 0.83, + "learning_rate": 2.8455092813256776e-08, + "loss": 0.1272, + "step": 15964 + }, + { + "epoch": 0.83, + "learning_rate": 2.843771246586768e-08, + "loss": 0.1283, + "step": 15965 + }, + { + "epoch": 0.83, + "learning_rate": 2.8420337021795025e-08, + "loss": 0.1475, + "step": 15966 + }, + { + "epoch": 0.83, + "learning_rate": 2.8402966481535417e-08, + "loss": 0.1396, + "step": 15967 + }, + { + "epoch": 0.83, + "learning_rate": 2.8385600845585188e-08, + "loss": 0.1198, + "step": 15968 + }, + { + "epoch": 0.83, + "learning_rate": 2.836824011444079e-08, + "loss": 0.1591, + "step": 15969 + }, + { + "epoch": 0.83, + "learning_rate": 2.8350884288598332e-08, + "loss": 0.1698, + "step": 15970 + }, + { + "epoch": 0.83, + "learning_rate": 2.8333533368553863e-08, + "loss": 0.1371, + "step": 15971 + }, + { + "epoch": 0.83, + "learning_rate": 2.8316187354803255e-08, + "loss": 0.1211, + "step": 15972 + }, + { + "epoch": 0.83, + "learning_rate": 2.829884624784231e-08, + "loss": 0.1206, + "step": 15973 + }, + { + "epoch": 0.83, + "learning_rate": 2.8281510048166635e-08, + "loss": 0.1367, + "step": 15974 + }, + { + "epoch": 0.83, + "learning_rate": 2.8264178756271694e-08, + "loss": 0.1594, + "step": 15975 + }, + { + "epoch": 0.83, + "learning_rate": 2.8246852372652807e-08, + "loss": 0.1481, + "step": 15976 + }, + { + "epoch": 0.83, + "learning_rate": 2.82295308978052e-08, + "loss": 0.1327, + "step": 15977 + }, + { + "epoch": 0.83, + "learning_rate": 2.8212214332223895e-08, + "loss": 0.1555, + "step": 15978 + }, + { + "epoch": 0.83, + "learning_rate": 2.8194902676403785e-08, + "loss": 0.115, + "step": 15979 + }, + { + "epoch": 0.83, + "learning_rate": 2.8177595930839726e-08, + "loss": 0.1675, + "step": 15980 + }, + { + "epoch": 0.83, + "learning_rate": 2.8160294096026292e-08, + "loss": 0.1155, + "step": 15981 + }, + { + "epoch": 0.83, + "learning_rate": 2.8142997172458004e-08, + "loss": 0.1211, + "step": 15982 + }, + { + "epoch": 0.83, + "learning_rate": 2.8125705160629153e-08, + "loss": 0.1463, + "step": 15983 + }, + { + "epoch": 0.83, + "learning_rate": 2.8108418061034078e-08, + "loss": 0.1364, + "step": 15984 + }, + { + "epoch": 0.83, + "learning_rate": 2.8091135874166738e-08, + "loss": 0.1346, + "step": 15985 + }, + { + "epoch": 0.83, + "learning_rate": 2.807385860052105e-08, + "loss": 0.1374, + "step": 15986 + }, + { + "epoch": 0.83, + "learning_rate": 2.8056586240590906e-08, + "loss": 0.1193, + "step": 15987 + }, + { + "epoch": 0.83, + "learning_rate": 2.803931879486989e-08, + "loss": 0.1316, + "step": 15988 + }, + { + "epoch": 0.83, + "learning_rate": 2.8022056263851546e-08, + "loss": 0.135, + "step": 15989 + }, + { + "epoch": 0.83, + "learning_rate": 2.800479864802916e-08, + "loss": 0.1598, + "step": 15990 + }, + { + "epoch": 0.83, + "learning_rate": 2.7987545947896075e-08, + "loss": 0.1485, + "step": 15991 + }, + { + "epoch": 0.83, + "learning_rate": 2.797029816394534e-08, + "loss": 0.1501, + "step": 15992 + }, + { + "epoch": 0.83, + "learning_rate": 2.795305529666989e-08, + "loss": 0.1728, + "step": 15993 + }, + { + "epoch": 0.83, + "learning_rate": 2.793581734656252e-08, + "loss": 0.1282, + "step": 15994 + }, + { + "epoch": 0.83, + "learning_rate": 2.7918584314115935e-08, + "loss": 0.0994, + "step": 15995 + }, + { + "epoch": 0.83, + "learning_rate": 2.7901356199822633e-08, + "loss": 0.1168, + "step": 15996 + }, + { + "epoch": 0.83, + "learning_rate": 2.788413300417496e-08, + "loss": 0.1029, + "step": 15997 + }, + { + "epoch": 0.84, + "learning_rate": 2.7866914727665246e-08, + "loss": 0.154, + "step": 15998 + }, + { + "epoch": 0.84, + "learning_rate": 2.784970137078557e-08, + "loss": 0.1301, + "step": 15999 + }, + { + "epoch": 0.84, + "learning_rate": 2.7832492934027895e-08, + "loss": 0.1161, + "step": 16000 + }, + { + "epoch": 0.84, + "learning_rate": 2.7815289417884026e-08, + "loss": 0.1501, + "step": 16001 + }, + { + "epoch": 0.84, + "learning_rate": 2.7798090822845654e-08, + "loss": 0.1436, + "step": 16002 + }, + { + "epoch": 0.84, + "learning_rate": 2.7780897149404326e-08, + "loss": 0.1313, + "step": 16003 + }, + { + "epoch": 0.84, + "learning_rate": 2.7763708398051377e-08, + "loss": 0.1259, + "step": 16004 + }, + { + "epoch": 0.84, + "learning_rate": 2.7746524569278195e-08, + "loss": 0.1259, + "step": 16005 + }, + { + "epoch": 0.84, + "learning_rate": 2.7729345663575832e-08, + "loss": 0.1478, + "step": 16006 + }, + { + "epoch": 0.84, + "learning_rate": 2.771217168143527e-08, + "loss": 0.1345, + "step": 16007 + }, + { + "epoch": 0.84, + "learning_rate": 2.769500262334732e-08, + "loss": 0.1279, + "step": 16008 + }, + { + "epoch": 0.84, + "learning_rate": 2.767783848980274e-08, + "loss": 0.1383, + "step": 16009 + }, + { + "epoch": 0.84, + "learning_rate": 2.766067928129212e-08, + "loss": 0.129, + "step": 16010 + }, + { + "epoch": 0.84, + "learning_rate": 2.7643524998305712e-08, + "loss": 0.1027, + "step": 16011 + }, + { + "epoch": 0.84, + "learning_rate": 2.7626375641333966e-08, + "loss": 0.1682, + "step": 16012 + }, + { + "epoch": 0.84, + "learning_rate": 2.7609231210866913e-08, + "loss": 0.137, + "step": 16013 + }, + { + "epoch": 0.84, + "learning_rate": 2.7592091707394605e-08, + "loss": 0.1493, + "step": 16014 + }, + { + "epoch": 0.84, + "learning_rate": 2.7574957131406805e-08, + "loss": 0.1353, + "step": 16015 + }, + { + "epoch": 0.84, + "learning_rate": 2.755782748339337e-08, + "loss": 0.1397, + "step": 16016 + }, + { + "epoch": 0.84, + "learning_rate": 2.754070276384377e-08, + "loss": 0.1631, + "step": 16017 + }, + { + "epoch": 0.84, + "learning_rate": 2.7523582973247462e-08, + "loss": 0.0988, + "step": 16018 + }, + { + "epoch": 0.84, + "learning_rate": 2.7506468112093738e-08, + "loss": 0.1324, + "step": 16019 + }, + { + "epoch": 0.84, + "learning_rate": 2.7489358180871747e-08, + "loss": 0.1103, + "step": 16020 + }, + { + "epoch": 0.84, + "learning_rate": 2.7472253180070493e-08, + "loss": 0.1459, + "step": 16021 + }, + { + "epoch": 0.84, + "learning_rate": 2.745515311017881e-08, + "loss": 0.1405, + "step": 16022 + }, + { + "epoch": 0.84, + "learning_rate": 2.74380579716855e-08, + "loss": 0.1337, + "step": 16023 + }, + { + "epoch": 0.84, + "learning_rate": 2.7420967765079096e-08, + "loss": 0.1406, + "step": 16024 + }, + { + "epoch": 0.84, + "learning_rate": 2.7403882490848086e-08, + "loss": 0.1396, + "step": 16025 + }, + { + "epoch": 0.84, + "learning_rate": 2.7386802149480658e-08, + "loss": 0.156, + "step": 16026 + }, + { + "epoch": 0.84, + "learning_rate": 2.736972674146518e-08, + "loss": 0.1258, + "step": 16027 + }, + { + "epoch": 0.84, + "learning_rate": 2.7352656267289488e-08, + "loss": 0.1507, + "step": 16028 + }, + { + "epoch": 0.84, + "learning_rate": 2.7335590727441493e-08, + "loss": 0.1563, + "step": 16029 + }, + { + "epoch": 0.84, + "learning_rate": 2.7318530122409033e-08, + "loss": 0.109, + "step": 16030 + }, + { + "epoch": 0.84, + "learning_rate": 2.730147445267963e-08, + "loss": 0.148, + "step": 16031 + }, + { + "epoch": 0.84, + "learning_rate": 2.7284423718740735e-08, + "loss": 0.1398, + "step": 16032 + }, + { + "epoch": 0.84, + "learning_rate": 2.726737792107967e-08, + "loss": 0.1173, + "step": 16033 + }, + { + "epoch": 0.84, + "learning_rate": 2.7250337060183647e-08, + "loss": 0.1243, + "step": 16034 + }, + { + "epoch": 0.84, + "learning_rate": 2.72333011365397e-08, + "loss": 0.1322, + "step": 16035 + }, + { + "epoch": 0.84, + "learning_rate": 2.721627015063468e-08, + "loss": 0.1438, + "step": 16036 + }, + { + "epoch": 0.84, + "learning_rate": 2.7199244102955375e-08, + "loss": 0.1421, + "step": 16037 + }, + { + "epoch": 0.84, + "learning_rate": 2.718222299398838e-08, + "loss": 0.1274, + "step": 16038 + }, + { + "epoch": 0.84, + "learning_rate": 2.7165206824220164e-08, + "loss": 0.1408, + "step": 16039 + }, + { + "epoch": 0.84, + "learning_rate": 2.7148195594137012e-08, + "loss": 0.1273, + "step": 16040 + }, + { + "epoch": 0.84, + "learning_rate": 2.7131189304225198e-08, + "loss": 0.1271, + "step": 16041 + }, + { + "epoch": 0.84, + "learning_rate": 2.7114187954970737e-08, + "loss": 0.1434, + "step": 16042 + }, + { + "epoch": 0.84, + "learning_rate": 2.709719154685952e-08, + "loss": 0.1138, + "step": 16043 + }, + { + "epoch": 0.84, + "learning_rate": 2.7080200080377325e-08, + "loss": 0.1286, + "step": 16044 + }, + { + "epoch": 0.84, + "learning_rate": 2.7063213556009755e-08, + "loss": 0.1131, + "step": 16045 + }, + { + "epoch": 0.84, + "learning_rate": 2.7046231974242318e-08, + "loss": 0.166, + "step": 16046 + }, + { + "epoch": 0.84, + "learning_rate": 2.7029255335560263e-08, + "loss": 0.1735, + "step": 16047 + }, + { + "epoch": 0.84, + "learning_rate": 2.7012283640448942e-08, + "loss": 0.1472, + "step": 16048 + }, + { + "epoch": 0.84, + "learning_rate": 2.6995316889393316e-08, + "loss": 0.1315, + "step": 16049 + }, + { + "epoch": 0.84, + "learning_rate": 2.6978355082878312e-08, + "loss": 0.1062, + "step": 16050 + }, + { + "epoch": 0.84, + "learning_rate": 2.6961398221388677e-08, + "loss": 0.1192, + "step": 16051 + }, + { + "epoch": 0.84, + "learning_rate": 2.694444630540913e-08, + "loss": 0.1014, + "step": 16052 + }, + { + "epoch": 0.84, + "learning_rate": 2.6927499335424154e-08, + "loss": 0.1374, + "step": 16053 + }, + { + "epoch": 0.84, + "learning_rate": 2.691055731191798e-08, + "loss": 0.1254, + "step": 16054 + }, + { + "epoch": 0.84, + "learning_rate": 2.6893620235374915e-08, + "loss": 0.1445, + "step": 16055 + }, + { + "epoch": 0.84, + "learning_rate": 2.6876688106279033e-08, + "loss": 0.1463, + "step": 16056 + }, + { + "epoch": 0.84, + "learning_rate": 2.6859760925114238e-08, + "loss": 0.1541, + "step": 16057 + }, + { + "epoch": 0.84, + "learning_rate": 2.6842838692364257e-08, + "loss": 0.1341, + "step": 16058 + }, + { + "epoch": 0.84, + "learning_rate": 2.6825921408512853e-08, + "loss": 0.1548, + "step": 16059 + }, + { + "epoch": 0.84, + "learning_rate": 2.6809009074043443e-08, + "loss": 0.2178, + "step": 16060 + }, + { + "epoch": 0.84, + "learning_rate": 2.6792101689439485e-08, + "loss": 0.1694, + "step": 16061 + }, + { + "epoch": 0.84, + "learning_rate": 2.6775199255184012e-08, + "loss": 0.1284, + "step": 16062 + }, + { + "epoch": 0.84, + "learning_rate": 2.675830177176026e-08, + "loss": 0.1367, + "step": 16063 + }, + { + "epoch": 0.84, + "learning_rate": 2.6741409239651137e-08, + "loss": 0.1182, + "step": 16064 + }, + { + "epoch": 0.84, + "learning_rate": 2.6724521659339428e-08, + "loss": 0.1409, + "step": 16065 + }, + { + "epoch": 0.84, + "learning_rate": 2.6707639031307728e-08, + "loss": 0.138, + "step": 16066 + }, + { + "epoch": 0.84, + "learning_rate": 2.6690761356038627e-08, + "loss": 0.1354, + "step": 16067 + }, + { + "epoch": 0.84, + "learning_rate": 2.6673888634014496e-08, + "loss": 0.1143, + "step": 16068 + }, + { + "epoch": 0.84, + "learning_rate": 2.6657020865717526e-08, + "loss": 0.1385, + "step": 16069 + }, + { + "epoch": 0.84, + "learning_rate": 2.6640158051629824e-08, + "loss": 0.1096, + "step": 16070 + }, + { + "epoch": 0.84, + "learning_rate": 2.6623300192233333e-08, + "loss": 0.1444, + "step": 16071 + }, + { + "epoch": 0.84, + "learning_rate": 2.6606447288009827e-08, + "loss": 0.1815, + "step": 16072 + }, + { + "epoch": 0.84, + "learning_rate": 2.6589599339440982e-08, + "loss": 0.1025, + "step": 16073 + }, + { + "epoch": 0.84, + "learning_rate": 2.6572756347008352e-08, + "loss": 0.1483, + "step": 16074 + }, + { + "epoch": 0.84, + "learning_rate": 2.6555918311193304e-08, + "loss": 0.1316, + "step": 16075 + }, + { + "epoch": 0.84, + "learning_rate": 2.6539085232477077e-08, + "loss": 0.0966, + "step": 16076 + }, + { + "epoch": 0.84, + "learning_rate": 2.6522257111340707e-08, + "loss": 0.1804, + "step": 16077 + }, + { + "epoch": 0.84, + "learning_rate": 2.6505433948265276e-08, + "loss": 0.1307, + "step": 16078 + }, + { + "epoch": 0.84, + "learning_rate": 2.648861574373147e-08, + "loss": 0.1142, + "step": 16079 + }, + { + "epoch": 0.84, + "learning_rate": 2.6471802498219985e-08, + "loss": 0.1129, + "step": 16080 + }, + { + "epoch": 0.84, + "learning_rate": 2.6454994212211402e-08, + "loss": 0.1791, + "step": 16081 + }, + { + "epoch": 0.84, + "learning_rate": 2.6438190886186085e-08, + "loss": 0.1247, + "step": 16082 + }, + { + "epoch": 0.84, + "learning_rate": 2.6421392520624276e-08, + "loss": 0.119, + "step": 16083 + }, + { + "epoch": 0.84, + "learning_rate": 2.640459911600603e-08, + "loss": 0.1246, + "step": 16084 + }, + { + "epoch": 0.84, + "learning_rate": 2.6387810672811396e-08, + "loss": 0.0977, + "step": 16085 + }, + { + "epoch": 0.84, + "learning_rate": 2.6371027191520155e-08, + "loss": 0.1278, + "step": 16086 + }, + { + "epoch": 0.84, + "learning_rate": 2.6354248672611978e-08, + "loss": 0.1429, + "step": 16087 + }, + { + "epoch": 0.84, + "learning_rate": 2.6337475116566433e-08, + "loss": 0.1142, + "step": 16088 + }, + { + "epoch": 0.84, + "learning_rate": 2.632070652386287e-08, + "loss": 0.1477, + "step": 16089 + }, + { + "epoch": 0.84, + "learning_rate": 2.6303942894980567e-08, + "loss": 0.1059, + "step": 16090 + }, + { + "epoch": 0.84, + "learning_rate": 2.6287184230398575e-08, + "loss": 0.1051, + "step": 16091 + }, + { + "epoch": 0.84, + "learning_rate": 2.6270430530595965e-08, + "loss": 0.1453, + "step": 16092 + }, + { + "epoch": 0.84, + "learning_rate": 2.6253681796051542e-08, + "loss": 0.1617, + "step": 16093 + }, + { + "epoch": 0.84, + "learning_rate": 2.6236938027243937e-08, + "loss": 0.1391, + "step": 16094 + }, + { + "epoch": 0.84, + "learning_rate": 2.6220199224651686e-08, + "loss": 0.156, + "step": 16095 + }, + { + "epoch": 0.84, + "learning_rate": 2.620346538875331e-08, + "loss": 0.1432, + "step": 16096 + }, + { + "epoch": 0.84, + "learning_rate": 2.6186736520026964e-08, + "loss": 0.1546, + "step": 16097 + }, + { + "epoch": 0.84, + "learning_rate": 2.617001261895071e-08, + "loss": 0.1475, + "step": 16098 + }, + { + "epoch": 0.84, + "learning_rate": 2.6153293686002677e-08, + "loss": 0.1339, + "step": 16099 + }, + { + "epoch": 0.84, + "learning_rate": 2.613657972166059e-08, + "loss": 0.1151, + "step": 16100 + }, + { + "epoch": 0.84, + "learning_rate": 2.6119870726402204e-08, + "loss": 0.1271, + "step": 16101 + }, + { + "epoch": 0.84, + "learning_rate": 2.610316670070496e-08, + "loss": 0.1461, + "step": 16102 + }, + { + "epoch": 0.84, + "learning_rate": 2.6086467645046406e-08, + "loss": 0.161, + "step": 16103 + }, + { + "epoch": 0.84, + "learning_rate": 2.6069773559903785e-08, + "loss": 0.17, + "step": 16104 + }, + { + "epoch": 0.84, + "learning_rate": 2.6053084445754115e-08, + "loss": 0.1208, + "step": 16105 + }, + { + "epoch": 0.84, + "learning_rate": 2.6036400303074458e-08, + "loss": 0.1178, + "step": 16106 + }, + { + "epoch": 0.84, + "learning_rate": 2.6019721132341633e-08, + "loss": 0.16, + "step": 16107 + }, + { + "epoch": 0.84, + "learning_rate": 2.6003046934032346e-08, + "loss": 0.1373, + "step": 16108 + }, + { + "epoch": 0.84, + "learning_rate": 2.5986377708623108e-08, + "loss": 0.1255, + "step": 16109 + }, + { + "epoch": 0.84, + "learning_rate": 2.5969713456590425e-08, + "loss": 0.1619, + "step": 16110 + }, + { + "epoch": 0.84, + "learning_rate": 2.5953054178410495e-08, + "loss": 0.135, + "step": 16111 + }, + { + "epoch": 0.84, + "learning_rate": 2.593639987455949e-08, + "loss": 0.1121, + "step": 16112 + }, + { + "epoch": 0.84, + "learning_rate": 2.591975054551334e-08, + "loss": 0.1429, + "step": 16113 + }, + { + "epoch": 0.84, + "learning_rate": 2.5903106191747958e-08, + "loss": 0.1535, + "step": 16114 + }, + { + "epoch": 0.84, + "learning_rate": 2.5886466813738984e-08, + "loss": 0.1414, + "step": 16115 + }, + { + "epoch": 0.84, + "learning_rate": 2.5869832411961967e-08, + "loss": 0.1092, + "step": 16116 + }, + { + "epoch": 0.84, + "learning_rate": 2.58532029868924e-08, + "loss": 0.114, + "step": 16117 + }, + { + "epoch": 0.84, + "learning_rate": 2.5836578539005514e-08, + "loss": 0.1679, + "step": 16118 + }, + { + "epoch": 0.84, + "learning_rate": 2.5819959068776454e-08, + "loss": 0.106, + "step": 16119 + }, + { + "epoch": 0.84, + "learning_rate": 2.5803344576680164e-08, + "loss": 0.1518, + "step": 16120 + }, + { + "epoch": 0.84, + "learning_rate": 2.5786735063191622e-08, + "loss": 0.1291, + "step": 16121 + }, + { + "epoch": 0.84, + "learning_rate": 2.5770130528785384e-08, + "loss": 0.1346, + "step": 16122 + }, + { + "epoch": 0.84, + "learning_rate": 2.5753530973936045e-08, + "loss": 0.1384, + "step": 16123 + }, + { + "epoch": 0.84, + "learning_rate": 2.5736936399118093e-08, + "loss": 0.1157, + "step": 16124 + }, + { + "epoch": 0.84, + "learning_rate": 2.5720346804805793e-08, + "loss": 0.1077, + "step": 16125 + }, + { + "epoch": 0.84, + "learning_rate": 2.5703762191473234e-08, + "loss": 0.1249, + "step": 16126 + }, + { + "epoch": 0.84, + "learning_rate": 2.568718255959439e-08, + "loss": 0.1456, + "step": 16127 + }, + { + "epoch": 0.84, + "learning_rate": 2.5670607909643216e-08, + "loss": 0.1349, + "step": 16128 + }, + { + "epoch": 0.84, + "learning_rate": 2.565403824209338e-08, + "loss": 0.14, + "step": 16129 + }, + { + "epoch": 0.84, + "learning_rate": 2.5637473557418408e-08, + "loss": 0.1113, + "step": 16130 + }, + { + "epoch": 0.84, + "learning_rate": 2.5620913856091776e-08, + "loss": 0.1681, + "step": 16131 + }, + { + "epoch": 0.84, + "learning_rate": 2.560435913858674e-08, + "loss": 0.1471, + "step": 16132 + }, + { + "epoch": 0.84, + "learning_rate": 2.558780940537644e-08, + "loss": 0.1387, + "step": 16133 + }, + { + "epoch": 0.84, + "learning_rate": 2.5571264656933822e-08, + "loss": 0.1276, + "step": 16134 + }, + { + "epoch": 0.84, + "learning_rate": 2.555472489373185e-08, + "loss": 0.1021, + "step": 16135 + }, + { + "epoch": 0.84, + "learning_rate": 2.5538190116243163e-08, + "loss": 0.1078, + "step": 16136 + }, + { + "epoch": 0.84, + "learning_rate": 2.5521660324940387e-08, + "loss": 0.1441, + "step": 16137 + }, + { + "epoch": 0.84, + "learning_rate": 2.5505135520295896e-08, + "loss": 0.1172, + "step": 16138 + }, + { + "epoch": 0.84, + "learning_rate": 2.548861570278198e-08, + "loss": 0.1322, + "step": 16139 + }, + { + "epoch": 0.84, + "learning_rate": 2.5472100872870817e-08, + "loss": 0.1235, + "step": 16140 + }, + { + "epoch": 0.84, + "learning_rate": 2.545559103103434e-08, + "loss": 0.1434, + "step": 16141 + }, + { + "epoch": 0.84, + "learning_rate": 2.5439086177744485e-08, + "loss": 0.1296, + "step": 16142 + }, + { + "epoch": 0.84, + "learning_rate": 2.5422586313472937e-08, + "loss": 0.1558, + "step": 16143 + }, + { + "epoch": 0.84, + "learning_rate": 2.5406091438691256e-08, + "loss": 0.1305, + "step": 16144 + }, + { + "epoch": 0.84, + "learning_rate": 2.5389601553870867e-08, + "loss": 0.14, + "step": 16145 + }, + { + "epoch": 0.84, + "learning_rate": 2.5373116659483096e-08, + "loss": 0.1131, + "step": 16146 + }, + { + "epoch": 0.84, + "learning_rate": 2.5356636755999107e-08, + "loss": 0.1358, + "step": 16147 + }, + { + "epoch": 0.84, + "learning_rate": 2.5340161843889786e-08, + "loss": 0.0994, + "step": 16148 + }, + { + "epoch": 0.84, + "learning_rate": 2.532369192362609e-08, + "loss": 0.1117, + "step": 16149 + }, + { + "epoch": 0.84, + "learning_rate": 2.530722699567871e-08, + "loss": 0.1098, + "step": 16150 + }, + { + "epoch": 0.84, + "learning_rate": 2.5290767060518224e-08, + "loss": 0.1647, + "step": 16151 + }, + { + "epoch": 0.84, + "learning_rate": 2.527431211861504e-08, + "loss": 0.1426, + "step": 16152 + }, + { + "epoch": 0.84, + "learning_rate": 2.5257862170439483e-08, + "loss": 0.1546, + "step": 16153 + }, + { + "epoch": 0.84, + "learning_rate": 2.524141721646169e-08, + "loss": 0.1191, + "step": 16154 + }, + { + "epoch": 0.84, + "learning_rate": 2.5224977257151647e-08, + "loss": 0.1438, + "step": 16155 + }, + { + "epoch": 0.84, + "learning_rate": 2.520854229297924e-08, + "loss": 0.1525, + "step": 16156 + }, + { + "epoch": 0.84, + "learning_rate": 2.5192112324414162e-08, + "loss": 0.1482, + "step": 16157 + }, + { + "epoch": 0.84, + "learning_rate": 2.5175687351926013e-08, + "loss": 0.1318, + "step": 16158 + }, + { + "epoch": 0.84, + "learning_rate": 2.515926737598415e-08, + "loss": 0.1087, + "step": 16159 + }, + { + "epoch": 0.84, + "learning_rate": 2.5142852397057955e-08, + "loss": 0.1175, + "step": 16160 + }, + { + "epoch": 0.84, + "learning_rate": 2.5126442415616543e-08, + "loss": 0.1194, + "step": 16161 + }, + { + "epoch": 0.84, + "learning_rate": 2.511003743212894e-08, + "loss": 0.1242, + "step": 16162 + }, + { + "epoch": 0.84, + "learning_rate": 2.509363744706392e-08, + "loss": 0.1513, + "step": 16163 + }, + { + "epoch": 0.84, + "learning_rate": 2.5077242460890335e-08, + "loss": 0.1265, + "step": 16164 + }, + { + "epoch": 0.84, + "learning_rate": 2.5060852474076655e-08, + "loss": 0.1275, + "step": 16165 + }, + { + "epoch": 0.84, + "learning_rate": 2.50444674870913e-08, + "loss": 0.127, + "step": 16166 + }, + { + "epoch": 0.84, + "learning_rate": 2.5028087500402662e-08, + "loss": 0.1321, + "step": 16167 + }, + { + "epoch": 0.84, + "learning_rate": 2.5011712514478822e-08, + "loss": 0.1224, + "step": 16168 + }, + { + "epoch": 0.84, + "learning_rate": 2.4995342529787788e-08, + "loss": 0.1393, + "step": 16169 + }, + { + "epoch": 0.84, + "learning_rate": 2.4978977546797385e-08, + "loss": 0.1122, + "step": 16170 + }, + { + "epoch": 0.84, + "learning_rate": 2.4962617565975398e-08, + "loss": 0.1362, + "step": 16171 + }, + { + "epoch": 0.84, + "learning_rate": 2.494626258778938e-08, + "loss": 0.1475, + "step": 16172 + }, + { + "epoch": 0.84, + "learning_rate": 2.4929912612706805e-08, + "loss": 0.1313, + "step": 16173 + }, + { + "epoch": 0.84, + "learning_rate": 2.491356764119481e-08, + "loss": 0.124, + "step": 16174 + }, + { + "epoch": 0.84, + "learning_rate": 2.489722767372069e-08, + "loss": 0.136, + "step": 16175 + }, + { + "epoch": 0.84, + "learning_rate": 2.48808927107514e-08, + "loss": 0.1241, + "step": 16176 + }, + { + "epoch": 0.84, + "learning_rate": 2.4864562752753813e-08, + "loss": 0.1057, + "step": 16177 + }, + { + "epoch": 0.84, + "learning_rate": 2.484823780019458e-08, + "loss": 0.132, + "step": 16178 + }, + { + "epoch": 0.84, + "learning_rate": 2.483191785354035e-08, + "loss": 0.1425, + "step": 16179 + }, + { + "epoch": 0.84, + "learning_rate": 2.481560291325755e-08, + "loss": 0.1344, + "step": 16180 + }, + { + "epoch": 0.84, + "learning_rate": 2.479929297981245e-08, + "loss": 0.148, + "step": 16181 + }, + { + "epoch": 0.84, + "learning_rate": 2.4782988053671184e-08, + "loss": 0.1443, + "step": 16182 + }, + { + "epoch": 0.84, + "learning_rate": 2.476668813529976e-08, + "loss": 0.1719, + "step": 16183 + }, + { + "epoch": 0.84, + "learning_rate": 2.4750393225164035e-08, + "loss": 0.1312, + "step": 16184 + }, + { + "epoch": 0.84, + "learning_rate": 2.473410332372967e-08, + "loss": 0.1227, + "step": 16185 + }, + { + "epoch": 0.84, + "learning_rate": 2.471781843146237e-08, + "loss": 0.1502, + "step": 16186 + }, + { + "epoch": 0.84, + "learning_rate": 2.470153854882746e-08, + "loss": 0.0932, + "step": 16187 + }, + { + "epoch": 0.84, + "learning_rate": 2.4685263676290246e-08, + "loss": 0.1968, + "step": 16188 + }, + { + "epoch": 0.84, + "learning_rate": 2.466899381431584e-08, + "loss": 0.1586, + "step": 16189 + }, + { + "epoch": 0.85, + "learning_rate": 2.465272896336936e-08, + "loss": 0.1224, + "step": 16190 + }, + { + "epoch": 0.85, + "learning_rate": 2.4636469123915527e-08, + "loss": 0.1788, + "step": 16191 + }, + { + "epoch": 0.85, + "learning_rate": 2.4620214296419074e-08, + "loss": 0.1514, + "step": 16192 + }, + { + "epoch": 0.85, + "learning_rate": 2.460396448134463e-08, + "loss": 0.1082, + "step": 16193 + }, + { + "epoch": 0.85, + "learning_rate": 2.4587719679156584e-08, + "loss": 0.1289, + "step": 16194 + }, + { + "epoch": 0.85, + "learning_rate": 2.4571479890319245e-08, + "loss": 0.1095, + "step": 16195 + }, + { + "epoch": 0.85, + "learning_rate": 2.4555245115296675e-08, + "loss": 0.1343, + "step": 16196 + }, + { + "epoch": 0.85, + "learning_rate": 2.4539015354552984e-08, + "loss": 0.129, + "step": 16197 + }, + { + "epoch": 0.85, + "learning_rate": 2.4522790608552002e-08, + "loss": 0.1229, + "step": 16198 + }, + { + "epoch": 0.85, + "learning_rate": 2.4506570877757338e-08, + "loss": 0.1358, + "step": 16199 + }, + { + "epoch": 0.85, + "learning_rate": 2.4490356162632643e-08, + "loss": 0.1255, + "step": 16200 + }, + { + "epoch": 0.85, + "learning_rate": 2.4474146463641342e-08, + "loss": 0.1216, + "step": 16201 + }, + { + "epoch": 0.85, + "learning_rate": 2.4457941781246695e-08, + "loss": 0.1358, + "step": 16202 + }, + { + "epoch": 0.85, + "learning_rate": 2.444174211591177e-08, + "loss": 0.161, + "step": 16203 + }, + { + "epoch": 0.85, + "learning_rate": 2.4425547468099683e-08, + "loss": 0.1388, + "step": 16204 + }, + { + "epoch": 0.85, + "learning_rate": 2.440935783827325e-08, + "loss": 0.1481, + "step": 16205 + }, + { + "epoch": 0.85, + "learning_rate": 2.4393173226895137e-08, + "loss": 0.1149, + "step": 16206 + }, + { + "epoch": 0.85, + "learning_rate": 2.437699363442791e-08, + "loss": 0.1741, + "step": 16207 + }, + { + "epoch": 0.85, + "learning_rate": 2.4360819061334027e-08, + "loss": 0.1105, + "step": 16208 + }, + { + "epoch": 0.85, + "learning_rate": 2.434464950807573e-08, + "loss": 0.1264, + "step": 16209 + }, + { + "epoch": 0.85, + "learning_rate": 2.4328484975115105e-08, + "loss": 0.1503, + "step": 16210 + }, + { + "epoch": 0.85, + "learning_rate": 2.4312325462914263e-08, + "loss": 0.1245, + "step": 16211 + }, + { + "epoch": 0.85, + "learning_rate": 2.429617097193495e-08, + "loss": 0.1415, + "step": 16212 + }, + { + "epoch": 0.85, + "learning_rate": 2.4280021502638926e-08, + "loss": 0.1419, + "step": 16213 + }, + { + "epoch": 0.85, + "learning_rate": 2.4263877055487668e-08, + "loss": 0.1261, + "step": 16214 + }, + { + "epoch": 0.85, + "learning_rate": 2.4247737630942698e-08, + "loss": 0.1126, + "step": 16215 + }, + { + "epoch": 0.85, + "learning_rate": 2.423160322946527e-08, + "loss": 0.1114, + "step": 16216 + }, + { + "epoch": 0.85, + "learning_rate": 2.4215473851516366e-08, + "loss": 0.1371, + "step": 16217 + }, + { + "epoch": 0.85, + "learning_rate": 2.4199349497557155e-08, + "loss": 0.1834, + "step": 16218 + }, + { + "epoch": 0.85, + "learning_rate": 2.4183230168048395e-08, + "loss": 0.1614, + "step": 16219 + }, + { + "epoch": 0.85, + "learning_rate": 2.4167115863450773e-08, + "loss": 0.1264, + "step": 16220 + }, + { + "epoch": 0.85, + "learning_rate": 2.415100658422482e-08, + "loss": 0.1221, + "step": 16221 + }, + { + "epoch": 0.85, + "learning_rate": 2.413490233083104e-08, + "loss": 0.1065, + "step": 16222 + }, + { + "epoch": 0.85, + "learning_rate": 2.4118803103729623e-08, + "loss": 0.156, + "step": 16223 + }, + { + "epoch": 0.85, + "learning_rate": 2.410270890338073e-08, + "loss": 0.099, + "step": 16224 + }, + { + "epoch": 0.85, + "learning_rate": 2.4086619730244306e-08, + "loss": 0.1363, + "step": 16225 + }, + { + "epoch": 0.85, + "learning_rate": 2.4070535584780206e-08, + "loss": 0.118, + "step": 16226 + }, + { + "epoch": 0.85, + "learning_rate": 2.4054456467448103e-08, + "loss": 0.1217, + "step": 16227 + }, + { + "epoch": 0.85, + "learning_rate": 2.4038382378707543e-08, + "loss": 0.1327, + "step": 16228 + }, + { + "epoch": 0.85, + "learning_rate": 2.402231331901796e-08, + "loss": 0.1299, + "step": 16229 + }, + { + "epoch": 0.85, + "learning_rate": 2.4006249288838586e-08, + "loss": 0.1032, + "step": 16230 + }, + { + "epoch": 0.85, + "learning_rate": 2.3990190288628564e-08, + "loss": 0.1113, + "step": 16231 + }, + { + "epoch": 0.85, + "learning_rate": 2.3974136318846816e-08, + "loss": 0.1093, + "step": 16232 + }, + { + "epoch": 0.85, + "learning_rate": 2.3958087379952285e-08, + "loss": 0.1414, + "step": 16233 + }, + { + "epoch": 0.85, + "learning_rate": 2.3942043472403517e-08, + "loss": 0.1056, + "step": 16234 + }, + { + "epoch": 0.85, + "learning_rate": 2.39260045966591e-08, + "loss": 0.1355, + "step": 16235 + }, + { + "epoch": 0.85, + "learning_rate": 2.3909970753177466e-08, + "loss": 0.1298, + "step": 16236 + }, + { + "epoch": 0.85, + "learning_rate": 2.3893941942416828e-08, + "loss": 0.103, + "step": 16237 + }, + { + "epoch": 0.85, + "learning_rate": 2.3877918164835332e-08, + "loss": 0.1754, + "step": 16238 + }, + { + "epoch": 0.85, + "learning_rate": 2.3861899420890896e-08, + "loss": 0.1523, + "step": 16239 + }, + { + "epoch": 0.85, + "learning_rate": 2.38458857110414e-08, + "loss": 0.1294, + "step": 16240 + }, + { + "epoch": 0.85, + "learning_rate": 2.382987703574453e-08, + "loss": 0.1438, + "step": 16241 + }, + { + "epoch": 0.85, + "learning_rate": 2.3813873395457706e-08, + "loss": 0.1092, + "step": 16242 + }, + { + "epoch": 0.85, + "learning_rate": 2.3797874790638415e-08, + "loss": 0.1758, + "step": 16243 + }, + { + "epoch": 0.85, + "learning_rate": 2.3781881221743914e-08, + "loss": 0.1194, + "step": 16244 + }, + { + "epoch": 0.85, + "learning_rate": 2.376589268923126e-08, + "loss": 0.1409, + "step": 16245 + }, + { + "epoch": 0.85, + "learning_rate": 2.3749909193557393e-08, + "loss": 0.1388, + "step": 16246 + }, + { + "epoch": 0.85, + "learning_rate": 2.3733930735179196e-08, + "loss": 0.1228, + "step": 16247 + }, + { + "epoch": 0.85, + "learning_rate": 2.3717957314553305e-08, + "loss": 0.1551, + "step": 16248 + }, + { + "epoch": 0.85, + "learning_rate": 2.3701988932136242e-08, + "loss": 0.1353, + "step": 16249 + }, + { + "epoch": 0.85, + "learning_rate": 2.3686025588384395e-08, + "loss": 0.1095, + "step": 16250 + }, + { + "epoch": 0.85, + "learning_rate": 2.3670067283754004e-08, + "loss": 0.1781, + "step": 16251 + }, + { + "epoch": 0.85, + "learning_rate": 2.3654114018701142e-08, + "loss": 0.1174, + "step": 16252 + }, + { + "epoch": 0.85, + "learning_rate": 2.3638165793681762e-08, + "loss": 0.1324, + "step": 16253 + }, + { + "epoch": 0.85, + "learning_rate": 2.362222260915172e-08, + "loss": 0.1831, + "step": 16254 + }, + { + "epoch": 0.85, + "learning_rate": 2.360628446556665e-08, + "loss": 0.1483, + "step": 16255 + }, + { + "epoch": 0.85, + "learning_rate": 2.3590351363382055e-08, + "loss": 0.15, + "step": 16256 + }, + { + "epoch": 0.85, + "learning_rate": 2.357442330305328e-08, + "loss": 0.1552, + "step": 16257 + }, + { + "epoch": 0.85, + "learning_rate": 2.3558500285035675e-08, + "loss": 0.1374, + "step": 16258 + }, + { + "epoch": 0.85, + "learning_rate": 2.354258230978421e-08, + "loss": 0.1306, + "step": 16259 + }, + { + "epoch": 0.85, + "learning_rate": 2.3526669377753805e-08, + "loss": 0.1208, + "step": 16260 + }, + { + "epoch": 0.85, + "learning_rate": 2.3510761489399366e-08, + "loss": 0.1447, + "step": 16261 + }, + { + "epoch": 0.85, + "learning_rate": 2.3494858645175507e-08, + "loss": 0.1382, + "step": 16262 + }, + { + "epoch": 0.85, + "learning_rate": 2.3478960845536712e-08, + "loss": 0.1347, + "step": 16263 + }, + { + "epoch": 0.85, + "learning_rate": 2.346306809093732e-08, + "loss": 0.1236, + "step": 16264 + }, + { + "epoch": 0.85, + "learning_rate": 2.3447180381831622e-08, + "loss": 0.1303, + "step": 16265 + }, + { + "epoch": 0.85, + "learning_rate": 2.3431297718673692e-08, + "loss": 0.1135, + "step": 16266 + }, + { + "epoch": 0.85, + "learning_rate": 2.3415420101917416e-08, + "loss": 0.0927, + "step": 16267 + }, + { + "epoch": 0.85, + "learning_rate": 2.3399547532016605e-08, + "loss": 0.1518, + "step": 16268 + }, + { + "epoch": 0.85, + "learning_rate": 2.33836800094249e-08, + "loss": 0.1069, + "step": 16269 + }, + { + "epoch": 0.85, + "learning_rate": 2.33678175345958e-08, + "loss": 0.1414, + "step": 16270 + }, + { + "epoch": 0.85, + "learning_rate": 2.335196010798264e-08, + "loss": 0.1232, + "step": 16271 + }, + { + "epoch": 0.85, + "learning_rate": 2.3336107730038668e-08, + "loss": 0.1046, + "step": 16272 + }, + { + "epoch": 0.85, + "learning_rate": 2.332026040121695e-08, + "loss": 0.1293, + "step": 16273 + }, + { + "epoch": 0.85, + "learning_rate": 2.3304418121970394e-08, + "loss": 0.137, + "step": 16274 + }, + { + "epoch": 0.85, + "learning_rate": 2.3288580892751742e-08, + "loss": 0.1477, + "step": 16275 + }, + { + "epoch": 0.85, + "learning_rate": 2.3272748714013747e-08, + "loss": 0.1077, + "step": 16276 + }, + { + "epoch": 0.85, + "learning_rate": 2.3256921586208754e-08, + "loss": 0.1524, + "step": 16277 + }, + { + "epoch": 0.85, + "learning_rate": 2.3241099509789163e-08, + "loss": 0.1329, + "step": 16278 + }, + { + "epoch": 0.85, + "learning_rate": 2.322528248520721e-08, + "loss": 0.1978, + "step": 16279 + }, + { + "epoch": 0.85, + "learning_rate": 2.320947051291491e-08, + "loss": 0.1406, + "step": 16280 + }, + { + "epoch": 0.85, + "learning_rate": 2.319366359336421e-08, + "loss": 0.168, + "step": 16281 + }, + { + "epoch": 0.85, + "learning_rate": 2.317786172700682e-08, + "loss": 0.1172, + "step": 16282 + }, + { + "epoch": 0.85, + "learning_rate": 2.3162064914294444e-08, + "loss": 0.1346, + "step": 16283 + }, + { + "epoch": 0.85, + "learning_rate": 2.3146273155678543e-08, + "loss": 0.1167, + "step": 16284 + }, + { + "epoch": 0.85, + "learning_rate": 2.3130486451610398e-08, + "loss": 0.1278, + "step": 16285 + }, + { + "epoch": 0.85, + "learning_rate": 2.3114704802541186e-08, + "loss": 0.1319, + "step": 16286 + }, + { + "epoch": 0.85, + "learning_rate": 2.309892820892203e-08, + "loss": 0.149, + "step": 16287 + }, + { + "epoch": 0.85, + "learning_rate": 2.3083156671203796e-08, + "loss": 0.1318, + "step": 16288 + }, + { + "epoch": 0.85, + "learning_rate": 2.3067390189837255e-08, + "loss": 0.1177, + "step": 16289 + }, + { + "epoch": 0.85, + "learning_rate": 2.3051628765272957e-08, + "loss": 0.1168, + "step": 16290 + }, + { + "epoch": 0.85, + "learning_rate": 2.3035872397961454e-08, + "loss": 0.1448, + "step": 16291 + }, + { + "epoch": 0.85, + "learning_rate": 2.302012108835305e-08, + "loss": 0.1434, + "step": 16292 + }, + { + "epoch": 0.85, + "learning_rate": 2.30043748368979e-08, + "loss": 0.136, + "step": 16293 + }, + { + "epoch": 0.85, + "learning_rate": 2.2988633644046018e-08, + "loss": 0.1346, + "step": 16294 + }, + { + "epoch": 0.85, + "learning_rate": 2.2972897510247357e-08, + "loss": 0.1417, + "step": 16295 + }, + { + "epoch": 0.85, + "learning_rate": 2.2957166435951604e-08, + "loss": 0.1339, + "step": 16296 + }, + { + "epoch": 0.85, + "learning_rate": 2.2941440421608348e-08, + "loss": 0.1384, + "step": 16297 + }, + { + "epoch": 0.85, + "learning_rate": 2.2925719467667103e-08, + "loss": 0.1369, + "step": 16298 + }, + { + "epoch": 0.85, + "learning_rate": 2.2910003574577152e-08, + "loss": 0.1212, + "step": 16299 + }, + { + "epoch": 0.85, + "learning_rate": 2.289429274278769e-08, + "loss": 0.147, + "step": 16300 + }, + { + "epoch": 0.85, + "learning_rate": 2.2878586972747694e-08, + "loss": 0.1343, + "step": 16301 + }, + { + "epoch": 0.85, + "learning_rate": 2.2862886264906045e-08, + "loss": 0.1103, + "step": 16302 + }, + { + "epoch": 0.85, + "learning_rate": 2.28471906197115e-08, + "loss": 0.1228, + "step": 16303 + }, + { + "epoch": 0.85, + "learning_rate": 2.2831500037612606e-08, + "loss": 0.1259, + "step": 16304 + }, + { + "epoch": 0.85, + "learning_rate": 2.2815814519057854e-08, + "loss": 0.1517, + "step": 16305 + }, + { + "epoch": 0.85, + "learning_rate": 2.2800134064495546e-08, + "loss": 0.133, + "step": 16306 + }, + { + "epoch": 0.85, + "learning_rate": 2.2784458674373795e-08, + "loss": 0.1158, + "step": 16307 + }, + { + "epoch": 0.85, + "learning_rate": 2.2768788349140577e-08, + "loss": 0.1317, + "step": 16308 + }, + { + "epoch": 0.85, + "learning_rate": 2.2753123089243863e-08, + "loss": 0.1379, + "step": 16309 + }, + { + "epoch": 0.85, + "learning_rate": 2.2737462895131365e-08, + "loss": 0.1552, + "step": 16310 + }, + { + "epoch": 0.85, + "learning_rate": 2.272180776725052e-08, + "loss": 0.1206, + "step": 16311 + }, + { + "epoch": 0.85, + "learning_rate": 2.2706157706048868e-08, + "loss": 0.094, + "step": 16312 + }, + { + "epoch": 0.85, + "learning_rate": 2.2690512711973687e-08, + "loss": 0.1272, + "step": 16313 + }, + { + "epoch": 0.85, + "learning_rate": 2.267487278547211e-08, + "loss": 0.1554, + "step": 16314 + }, + { + "epoch": 0.85, + "learning_rate": 2.265923792699107e-08, + "loss": 0.1374, + "step": 16315 + }, + { + "epoch": 0.85, + "learning_rate": 2.2643608136977498e-08, + "loss": 0.1328, + "step": 16316 + }, + { + "epoch": 0.85, + "learning_rate": 2.2627983415878083e-08, + "loss": 0.0903, + "step": 16317 + }, + { + "epoch": 0.85, + "learning_rate": 2.26123637641394e-08, + "loss": 0.1168, + "step": 16318 + }, + { + "epoch": 0.85, + "learning_rate": 2.2596749182207798e-08, + "loss": 0.1269, + "step": 16319 + }, + { + "epoch": 0.85, + "learning_rate": 2.2581139670529615e-08, + "loss": 0.1184, + "step": 16320 + }, + { + "epoch": 0.85, + "learning_rate": 2.2565535229550937e-08, + "loss": 0.1199, + "step": 16321 + }, + { + "epoch": 0.85, + "learning_rate": 2.2549935859717738e-08, + "loss": 0.1383, + "step": 16322 + }, + { + "epoch": 0.85, + "learning_rate": 2.2534341561475888e-08, + "loss": 0.1511, + "step": 16323 + }, + { + "epoch": 0.85, + "learning_rate": 2.2518752335271074e-08, + "loss": 0.1127, + "step": 16324 + }, + { + "epoch": 0.85, + "learning_rate": 2.2503168181548847e-08, + "loss": 0.1414, + "step": 16325 + }, + { + "epoch": 0.85, + "learning_rate": 2.2487589100754543e-08, + "loss": 0.1856, + "step": 16326 + }, + { + "epoch": 0.85, + "learning_rate": 2.247201509333354e-08, + "loss": 0.1106, + "step": 16327 + }, + { + "epoch": 0.85, + "learning_rate": 2.245644615973086e-08, + "loss": 0.1355, + "step": 16328 + }, + { + "epoch": 0.85, + "learning_rate": 2.2440882300391428e-08, + "loss": 0.1624, + "step": 16329 + }, + { + "epoch": 0.85, + "learning_rate": 2.2425323515760186e-08, + "loss": 0.1472, + "step": 16330 + }, + { + "epoch": 0.85, + "learning_rate": 2.240976980628173e-08, + "loss": 0.1498, + "step": 16331 + }, + { + "epoch": 0.85, + "learning_rate": 2.239422117240062e-08, + "loss": 0.1382, + "step": 16332 + }, + { + "epoch": 0.85, + "learning_rate": 2.2378677614561182e-08, + "loss": 0.1403, + "step": 16333 + }, + { + "epoch": 0.85, + "learning_rate": 2.2363139133207754e-08, + "loss": 0.1257, + "step": 16334 + }, + { + "epoch": 0.85, + "learning_rate": 2.2347605728784424e-08, + "loss": 0.1248, + "step": 16335 + }, + { + "epoch": 0.85, + "learning_rate": 2.2332077401735017e-08, + "loss": 0.1182, + "step": 16336 + }, + { + "epoch": 0.85, + "learning_rate": 2.2316554152503465e-08, + "loss": 0.1468, + "step": 16337 + }, + { + "epoch": 0.85, + "learning_rate": 2.230103598153339e-08, + "loss": 0.1658, + "step": 16338 + }, + { + "epoch": 0.85, + "learning_rate": 2.2285522889268305e-08, + "loss": 0.1761, + "step": 16339 + }, + { + "epoch": 0.85, + "learning_rate": 2.2270014876151543e-08, + "loss": 0.1346, + "step": 16340 + }, + { + "epoch": 0.85, + "learning_rate": 2.2254511942626418e-08, + "loss": 0.1367, + "step": 16341 + }, + { + "epoch": 0.85, + "learning_rate": 2.2239014089135933e-08, + "loss": 0.1427, + "step": 16342 + }, + { + "epoch": 0.85, + "learning_rate": 2.2223521316123063e-08, + "loss": 0.1468, + "step": 16343 + }, + { + "epoch": 0.85, + "learning_rate": 2.2208033624030586e-08, + "loss": 0.1067, + "step": 16344 + }, + { + "epoch": 0.85, + "learning_rate": 2.2192551013301132e-08, + "loss": 0.1645, + "step": 16345 + }, + { + "epoch": 0.85, + "learning_rate": 2.2177073484377208e-08, + "loss": 0.1121, + "step": 16346 + }, + { + "epoch": 0.85, + "learning_rate": 2.216160103770115e-08, + "loss": 0.1324, + "step": 16347 + }, + { + "epoch": 0.85, + "learning_rate": 2.2146133673715208e-08, + "loss": 0.1166, + "step": 16348 + }, + { + "epoch": 0.85, + "learning_rate": 2.2130671392861444e-08, + "loss": 0.11, + "step": 16349 + }, + { + "epoch": 0.85, + "learning_rate": 2.2115214195581733e-08, + "loss": 0.1353, + "step": 16350 + }, + { + "epoch": 0.85, + "learning_rate": 2.209976208231783e-08, + "loss": 0.1322, + "step": 16351 + }, + { + "epoch": 0.85, + "learning_rate": 2.208431505351147e-08, + "loss": 0.157, + "step": 16352 + }, + { + "epoch": 0.85, + "learning_rate": 2.206887310960408e-08, + "loss": 0.0864, + "step": 16353 + }, + { + "epoch": 0.85, + "learning_rate": 2.2053436251036927e-08, + "loss": 0.1388, + "step": 16354 + }, + { + "epoch": 0.85, + "learning_rate": 2.2038004478251283e-08, + "loss": 0.1466, + "step": 16355 + }, + { + "epoch": 0.85, + "learning_rate": 2.202257779168819e-08, + "loss": 0.1405, + "step": 16356 + }, + { + "epoch": 0.85, + "learning_rate": 2.2007156191788502e-08, + "loss": 0.1087, + "step": 16357 + }, + { + "epoch": 0.85, + "learning_rate": 2.1991739678992995e-08, + "loss": 0.1295, + "step": 16358 + }, + { + "epoch": 0.85, + "learning_rate": 2.1976328253742295e-08, + "loss": 0.1274, + "step": 16359 + }, + { + "epoch": 0.85, + "learning_rate": 2.1960921916476873e-08, + "loss": 0.1664, + "step": 16360 + }, + { + "epoch": 0.85, + "learning_rate": 2.194552066763704e-08, + "loss": 0.1182, + "step": 16361 + }, + { + "epoch": 0.85, + "learning_rate": 2.193012450766294e-08, + "loss": 0.1471, + "step": 16362 + }, + { + "epoch": 0.85, + "learning_rate": 2.1914733436994636e-08, + "loss": 0.149, + "step": 16363 + }, + { + "epoch": 0.85, + "learning_rate": 2.1899347456072004e-08, + "loss": 0.1379, + "step": 16364 + }, + { + "epoch": 0.85, + "learning_rate": 2.1883966565334732e-08, + "loss": 0.1401, + "step": 16365 + }, + { + "epoch": 0.85, + "learning_rate": 2.1868590765222496e-08, + "loss": 0.1588, + "step": 16366 + }, + { + "epoch": 0.85, + "learning_rate": 2.1853220056174694e-08, + "loss": 0.1641, + "step": 16367 + }, + { + "epoch": 0.85, + "learning_rate": 2.183785443863064e-08, + "loss": 0.1305, + "step": 16368 + }, + { + "epoch": 0.85, + "learning_rate": 2.182249391302946e-08, + "loss": 0.1188, + "step": 16369 + }, + { + "epoch": 0.85, + "learning_rate": 2.180713847981024e-08, + "loss": 0.1195, + "step": 16370 + }, + { + "epoch": 0.85, + "learning_rate": 2.179178813941176e-08, + "loss": 0.142, + "step": 16371 + }, + { + "epoch": 0.85, + "learning_rate": 2.177644289227274e-08, + "loss": 0.1221, + "step": 16372 + }, + { + "epoch": 0.85, + "learning_rate": 2.1761102738831805e-08, + "loss": 0.1236, + "step": 16373 + }, + { + "epoch": 0.85, + "learning_rate": 2.174576767952738e-08, + "loss": 0.1651, + "step": 16374 + }, + { + "epoch": 0.85, + "learning_rate": 2.1730437714797724e-08, + "loss": 0.1878, + "step": 16375 + }, + { + "epoch": 0.85, + "learning_rate": 2.171511284508092e-08, + "loss": 0.1082, + "step": 16376 + }, + { + "epoch": 0.85, + "learning_rate": 2.1699793070815087e-08, + "loss": 0.1227, + "step": 16377 + }, + { + "epoch": 0.85, + "learning_rate": 2.1684478392438012e-08, + "loss": 0.1235, + "step": 16378 + }, + { + "epoch": 0.85, + "learning_rate": 2.166916881038732e-08, + "loss": 0.1263, + "step": 16379 + }, + { + "epoch": 0.85, + "learning_rate": 2.1653864325100635e-08, + "loss": 0.1421, + "step": 16380 + }, + { + "epoch": 0.86, + "learning_rate": 2.163856493701539e-08, + "loss": 0.116, + "step": 16381 + }, + { + "epoch": 0.86, + "learning_rate": 2.162327064656879e-08, + "loss": 0.1683, + "step": 16382 + }, + { + "epoch": 0.86, + "learning_rate": 2.1607981454197954e-08, + "loss": 0.1199, + "step": 16383 + }, + { + "epoch": 0.86, + "learning_rate": 2.1592697360339905e-08, + "loss": 0.1342, + "step": 16384 + }, + { + "epoch": 0.86, + "learning_rate": 2.1577418365431434e-08, + "loss": 0.1398, + "step": 16385 + }, + { + "epoch": 0.86, + "learning_rate": 2.1562144469909227e-08, + "loss": 0.1698, + "step": 16386 + }, + { + "epoch": 0.86, + "learning_rate": 2.154687567420983e-08, + "loss": 0.1302, + "step": 16387 + }, + { + "epoch": 0.86, + "learning_rate": 2.1531611978769605e-08, + "loss": 0.1183, + "step": 16388 + }, + { + "epoch": 0.86, + "learning_rate": 2.15163533840248e-08, + "loss": 0.1483, + "step": 16389 + }, + { + "epoch": 0.86, + "learning_rate": 2.1501099890411467e-08, + "loss": 0.1325, + "step": 16390 + }, + { + "epoch": 0.86, + "learning_rate": 2.148585149836566e-08, + "loss": 0.1593, + "step": 16391 + }, + { + "epoch": 0.86, + "learning_rate": 2.147060820832314e-08, + "loss": 0.1191, + "step": 16392 + }, + { + "epoch": 0.86, + "learning_rate": 2.1455370020719532e-08, + "loss": 0.207, + "step": 16393 + }, + { + "epoch": 0.86, + "learning_rate": 2.144013693599036e-08, + "loss": 0.1297, + "step": 16394 + }, + { + "epoch": 0.86, + "learning_rate": 2.1424908954571074e-08, + "loss": 0.1381, + "step": 16395 + }, + { + "epoch": 0.86, + "learning_rate": 2.1409686076896795e-08, + "loss": 0.143, + "step": 16396 + }, + { + "epoch": 0.86, + "learning_rate": 2.1394468303402612e-08, + "loss": 0.1, + "step": 16397 + }, + { + "epoch": 0.86, + "learning_rate": 2.1379255634523453e-08, + "loss": 0.1036, + "step": 16398 + }, + { + "epoch": 0.86, + "learning_rate": 2.136404807069414e-08, + "loss": 0.1188, + "step": 16399 + }, + { + "epoch": 0.86, + "learning_rate": 2.134884561234931e-08, + "loss": 0.1508, + "step": 16400 + }, + { + "epoch": 0.86, + "learning_rate": 2.1333648259923453e-08, + "loss": 0.172, + "step": 16401 + }, + { + "epoch": 0.86, + "learning_rate": 2.1318456013850827e-08, + "loss": 0.1303, + "step": 16402 + }, + { + "epoch": 0.86, + "learning_rate": 2.130326887456577e-08, + "loss": 0.1843, + "step": 16403 + }, + { + "epoch": 0.86, + "learning_rate": 2.1288086842502296e-08, + "loss": 0.1435, + "step": 16404 + }, + { + "epoch": 0.86, + "learning_rate": 2.127290991809423e-08, + "loss": 0.1097, + "step": 16405 + }, + { + "epoch": 0.86, + "learning_rate": 2.1257738101775426e-08, + "loss": 0.1078, + "step": 16406 + }, + { + "epoch": 0.86, + "learning_rate": 2.1242571393979468e-08, + "loss": 0.0888, + "step": 16407 + }, + { + "epoch": 0.86, + "learning_rate": 2.1227409795139818e-08, + "loss": 0.1278, + "step": 16408 + }, + { + "epoch": 0.86, + "learning_rate": 2.1212253305689786e-08, + "loss": 0.0973, + "step": 16409 + }, + { + "epoch": 0.86, + "learning_rate": 2.1197101926062587e-08, + "loss": 0.1121, + "step": 16410 + }, + { + "epoch": 0.86, + "learning_rate": 2.1181955656691276e-08, + "loss": 0.1449, + "step": 16411 + }, + { + "epoch": 0.86, + "learning_rate": 2.1166814498008677e-08, + "loss": 0.1814, + "step": 16412 + }, + { + "epoch": 0.86, + "learning_rate": 2.115167845044756e-08, + "loss": 0.1125, + "step": 16413 + }, + { + "epoch": 0.86, + "learning_rate": 2.1136547514440518e-08, + "loss": 0.1678, + "step": 16414 + }, + { + "epoch": 0.86, + "learning_rate": 2.1121421690419993e-08, + "loss": 0.1062, + "step": 16415 + }, + { + "epoch": 0.86, + "learning_rate": 2.1106300978818248e-08, + "loss": 0.149, + "step": 16416 + }, + { + "epoch": 0.86, + "learning_rate": 2.1091185380067534e-08, + "loss": 0.1487, + "step": 16417 + }, + { + "epoch": 0.86, + "learning_rate": 2.1076074894599793e-08, + "loss": 0.1093, + "step": 16418 + }, + { + "epoch": 0.86, + "learning_rate": 2.1060969522846928e-08, + "loss": 0.1631, + "step": 16419 + }, + { + "epoch": 0.86, + "learning_rate": 2.104586926524057e-08, + "loss": 0.1169, + "step": 16420 + }, + { + "epoch": 0.86, + "learning_rate": 2.1030774122212437e-08, + "loss": 0.1569, + "step": 16421 + }, + { + "epoch": 0.86, + "learning_rate": 2.101568409419381e-08, + "loss": 0.1006, + "step": 16422 + }, + { + "epoch": 0.86, + "learning_rate": 2.100059918161601e-08, + "loss": 0.0969, + "step": 16423 + }, + { + "epoch": 0.86, + "learning_rate": 2.0985519384910223e-08, + "loss": 0.1154, + "step": 16424 + }, + { + "epoch": 0.86, + "learning_rate": 2.0970444704507396e-08, + "loss": 0.1578, + "step": 16425 + }, + { + "epoch": 0.86, + "learning_rate": 2.095537514083836e-08, + "loss": 0.1473, + "step": 16426 + }, + { + "epoch": 0.86, + "learning_rate": 2.0940310694333795e-08, + "loss": 0.1476, + "step": 16427 + }, + { + "epoch": 0.86, + "learning_rate": 2.092525136542431e-08, + "loss": 0.1189, + "step": 16428 + }, + { + "epoch": 0.86, + "learning_rate": 2.0910197154540278e-08, + "loss": 0.1337, + "step": 16429 + }, + { + "epoch": 0.86, + "learning_rate": 2.0895148062111944e-08, + "loss": 0.1257, + "step": 16430 + }, + { + "epoch": 0.86, + "learning_rate": 2.0880104088569417e-08, + "loss": 0.1485, + "step": 16431 + }, + { + "epoch": 0.86, + "learning_rate": 2.086506523434266e-08, + "loss": 0.1103, + "step": 16432 + }, + { + "epoch": 0.86, + "learning_rate": 2.0850031499861487e-08, + "loss": 0.1311, + "step": 16433 + }, + { + "epoch": 0.86, + "learning_rate": 2.083500288555553e-08, + "loss": 0.1029, + "step": 16434 + }, + { + "epoch": 0.86, + "learning_rate": 2.0819979391854402e-08, + "loss": 0.1437, + "step": 16435 + }, + { + "epoch": 0.86, + "learning_rate": 2.080496101918745e-08, + "loss": 0.1153, + "step": 16436 + }, + { + "epoch": 0.86, + "learning_rate": 2.078994776798386e-08, + "loss": 0.0955, + "step": 16437 + }, + { + "epoch": 0.86, + "learning_rate": 2.077493963867276e-08, + "loss": 0.1603, + "step": 16438 + }, + { + "epoch": 0.86, + "learning_rate": 2.0759936631683074e-08, + "loss": 0.1487, + "step": 16439 + }, + { + "epoch": 0.86, + "learning_rate": 2.0744938747443586e-08, + "loss": 0.1361, + "step": 16440 + }, + { + "epoch": 0.86, + "learning_rate": 2.0729945986382913e-08, + "loss": 0.1686, + "step": 16441 + }, + { + "epoch": 0.86, + "learning_rate": 2.0714958348929644e-08, + "loss": 0.1188, + "step": 16442 + }, + { + "epoch": 0.86, + "learning_rate": 2.0699975835512064e-08, + "loss": 0.1614, + "step": 16443 + }, + { + "epoch": 0.86, + "learning_rate": 2.0684998446558376e-08, + "loss": 0.1057, + "step": 16444 + }, + { + "epoch": 0.86, + "learning_rate": 2.0670026182496624e-08, + "loss": 0.1192, + "step": 16445 + }, + { + "epoch": 0.86, + "learning_rate": 2.0655059043754798e-08, + "loss": 0.1418, + "step": 16446 + }, + { + "epoch": 0.86, + "learning_rate": 2.0640097030760663e-08, + "loss": 0.1422, + "step": 16447 + }, + { + "epoch": 0.86, + "learning_rate": 2.0625140143941698e-08, + "loss": 0.1287, + "step": 16448 + }, + { + "epoch": 0.86, + "learning_rate": 2.0610188383725524e-08, + "loss": 0.1264, + "step": 16449 + }, + { + "epoch": 0.86, + "learning_rate": 2.0595241750539415e-08, + "loss": 0.1707, + "step": 16450 + }, + { + "epoch": 0.86, + "learning_rate": 2.0580300244810544e-08, + "loss": 0.1363, + "step": 16451 + }, + { + "epoch": 0.86, + "learning_rate": 2.0565363866965944e-08, + "loss": 0.1388, + "step": 16452 + }, + { + "epoch": 0.86, + "learning_rate": 2.055043261743252e-08, + "loss": 0.1227, + "step": 16453 + }, + { + "epoch": 0.86, + "learning_rate": 2.0535506496637024e-08, + "loss": 0.1556, + "step": 16454 + }, + { + "epoch": 0.86, + "learning_rate": 2.0520585505006038e-08, + "loss": 0.1264, + "step": 16455 + }, + { + "epoch": 0.86, + "learning_rate": 2.0505669642965983e-08, + "loss": 0.1396, + "step": 16456 + }, + { + "epoch": 0.86, + "learning_rate": 2.0490758910943208e-08, + "loss": 0.1221, + "step": 16457 + }, + { + "epoch": 0.86, + "learning_rate": 2.0475853309363832e-08, + "loss": 0.1207, + "step": 16458 + }, + { + "epoch": 0.86, + "learning_rate": 2.046095283865381e-08, + "loss": 0.1999, + "step": 16459 + }, + { + "epoch": 0.86, + "learning_rate": 2.0446057499239133e-08, + "loss": 0.1272, + "step": 16460 + }, + { + "epoch": 0.86, + "learning_rate": 2.0431167291545438e-08, + "loss": 0.1283, + "step": 16461 + }, + { + "epoch": 0.86, + "learning_rate": 2.041628221599829e-08, + "loss": 0.1198, + "step": 16462 + }, + { + "epoch": 0.86, + "learning_rate": 2.0401402273023095e-08, + "loss": 0.1314, + "step": 16463 + }, + { + "epoch": 0.86, + "learning_rate": 2.0386527463045232e-08, + "loss": 0.1299, + "step": 16464 + }, + { + "epoch": 0.86, + "learning_rate": 2.0371657786489727e-08, + "loss": 0.1163, + "step": 16465 + }, + { + "epoch": 0.86, + "learning_rate": 2.0356793243781524e-08, + "loss": 0.1242, + "step": 16466 + }, + { + "epoch": 0.86, + "learning_rate": 2.034193383534557e-08, + "loss": 0.1356, + "step": 16467 + }, + { + "epoch": 0.86, + "learning_rate": 2.0327079561606507e-08, + "loss": 0.1388, + "step": 16468 + }, + { + "epoch": 0.86, + "learning_rate": 2.0312230422988862e-08, + "loss": 0.1232, + "step": 16469 + }, + { + "epoch": 0.86, + "learning_rate": 2.0297386419917007e-08, + "loss": 0.1397, + "step": 16470 + }, + { + "epoch": 0.86, + "learning_rate": 2.0282547552815244e-08, + "loss": 0.1211, + "step": 16471 + }, + { + "epoch": 0.86, + "learning_rate": 2.0267713822107657e-08, + "loss": 0.2132, + "step": 16472 + }, + { + "epoch": 0.86, + "learning_rate": 2.0252885228218197e-08, + "loss": 0.1306, + "step": 16473 + }, + { + "epoch": 0.86, + "learning_rate": 2.0238061771570658e-08, + "loss": 0.1311, + "step": 16474 + }, + { + "epoch": 0.86, + "learning_rate": 2.0223243452588722e-08, + "loss": 0.1297, + "step": 16475 + }, + { + "epoch": 0.86, + "learning_rate": 2.0208430271695876e-08, + "loss": 0.1467, + "step": 16476 + }, + { + "epoch": 0.86, + "learning_rate": 2.0193622229315467e-08, + "loss": 0.1308, + "step": 16477 + }, + { + "epoch": 0.86, + "learning_rate": 2.0178819325870777e-08, + "loss": 0.1126, + "step": 16478 + }, + { + "epoch": 0.86, + "learning_rate": 2.0164021561784872e-08, + "loss": 0.1269, + "step": 16479 + }, + { + "epoch": 0.86, + "learning_rate": 2.0149228937480635e-08, + "loss": 0.1723, + "step": 16480 + }, + { + "epoch": 0.86, + "learning_rate": 2.013444145338086e-08, + "loss": 0.1208, + "step": 16481 + }, + { + "epoch": 0.86, + "learning_rate": 2.0119659109908184e-08, + "loss": 0.1307, + "step": 16482 + }, + { + "epoch": 0.86, + "learning_rate": 2.0104881907485115e-08, + "loss": 0.1702, + "step": 16483 + }, + { + "epoch": 0.86, + "learning_rate": 2.0090109846533898e-08, + "loss": 0.1505, + "step": 16484 + }, + { + "epoch": 0.86, + "learning_rate": 2.0075342927476857e-08, + "loss": 0.1561, + "step": 16485 + }, + { + "epoch": 0.86, + "learning_rate": 2.0060581150735967e-08, + "loss": 0.1499, + "step": 16486 + }, + { + "epoch": 0.86, + "learning_rate": 2.004582451673311e-08, + "loss": 0.171, + "step": 16487 + }, + { + "epoch": 0.86, + "learning_rate": 2.0031073025890044e-08, + "loss": 0.1482, + "step": 16488 + }, + { + "epoch": 0.86, + "learning_rate": 2.0016326678628424e-08, + "loss": 0.1317, + "step": 16489 + }, + { + "epoch": 0.86, + "learning_rate": 2.00015854753697e-08, + "loss": 0.14, + "step": 16490 + }, + { + "epoch": 0.86, + "learning_rate": 1.9986849416535056e-08, + "loss": 0.1501, + "step": 16491 + }, + { + "epoch": 0.86, + "learning_rate": 1.997211850254581e-08, + "loss": 0.1255, + "step": 16492 + }, + { + "epoch": 0.86, + "learning_rate": 1.9957392733822908e-08, + "loss": 0.1051, + "step": 16493 + }, + { + "epoch": 0.86, + "learning_rate": 1.9942672110787218e-08, + "loss": 0.1314, + "step": 16494 + }, + { + "epoch": 0.86, + "learning_rate": 1.9927956633859445e-08, + "loss": 0.1128, + "step": 16495 + }, + { + "epoch": 0.86, + "learning_rate": 1.9913246303460207e-08, + "loss": 0.1238, + "step": 16496 + }, + { + "epoch": 0.86, + "learning_rate": 1.9898541120009925e-08, + "loss": 0.1016, + "step": 16497 + }, + { + "epoch": 0.86, + "learning_rate": 1.9883841083928908e-08, + "loss": 0.1418, + "step": 16498 + }, + { + "epoch": 0.86, + "learning_rate": 1.9869146195637153e-08, + "loss": 0.118, + "step": 16499 + }, + { + "epoch": 0.86, + "learning_rate": 1.985445645555477e-08, + "loss": 0.1356, + "step": 16500 + }, + { + "epoch": 0.86, + "learning_rate": 1.9839771864101572e-08, + "loss": 0.1339, + "step": 16501 + }, + { + "epoch": 0.86, + "learning_rate": 1.982509242169721e-08, + "loss": 0.1098, + "step": 16502 + }, + { + "epoch": 0.86, + "learning_rate": 1.9810418128761298e-08, + "loss": 0.1406, + "step": 16503 + }, + { + "epoch": 0.86, + "learning_rate": 1.979574898571319e-08, + "loss": 0.1227, + "step": 16504 + }, + { + "epoch": 0.86, + "learning_rate": 1.9781084992972152e-08, + "loss": 0.1962, + "step": 16505 + }, + { + "epoch": 0.86, + "learning_rate": 1.9766426150957226e-08, + "loss": 0.121, + "step": 16506 + }, + { + "epoch": 0.86, + "learning_rate": 1.9751772460087502e-08, + "loss": 0.12, + "step": 16507 + }, + { + "epoch": 0.86, + "learning_rate": 1.9737123920781664e-08, + "loss": 0.1247, + "step": 16508 + }, + { + "epoch": 0.86, + "learning_rate": 1.972248053345842e-08, + "loss": 0.167, + "step": 16509 + }, + { + "epoch": 0.86, + "learning_rate": 1.9707842298536216e-08, + "loss": 0.1069, + "step": 16510 + }, + { + "epoch": 0.86, + "learning_rate": 1.9693209216433538e-08, + "loss": 0.1379, + "step": 16511 + }, + { + "epoch": 0.86, + "learning_rate": 1.9678581287568542e-08, + "loss": 0.1385, + "step": 16512 + }, + { + "epoch": 0.86, + "learning_rate": 1.966395851235929e-08, + "loss": 0.1773, + "step": 16513 + }, + { + "epoch": 0.86, + "learning_rate": 1.9649340891223697e-08, + "loss": 0.1623, + "step": 16514 + }, + { + "epoch": 0.86, + "learning_rate": 1.963472842457963e-08, + "loss": 0.1896, + "step": 16515 + }, + { + "epoch": 0.86, + "learning_rate": 1.9620121112844633e-08, + "loss": 0.1542, + "step": 16516 + }, + { + "epoch": 0.86, + "learning_rate": 1.9605518956436162e-08, + "loss": 0.1273, + "step": 16517 + }, + { + "epoch": 0.86, + "learning_rate": 1.9590921955771634e-08, + "loss": 0.1774, + "step": 16518 + }, + { + "epoch": 0.86, + "learning_rate": 1.957633011126818e-08, + "loss": 0.1128, + "step": 16519 + }, + { + "epoch": 0.86, + "learning_rate": 1.9561743423342892e-08, + "loss": 0.1645, + "step": 16520 + }, + { + "epoch": 0.86, + "learning_rate": 1.954716189241259e-08, + "loss": 0.1449, + "step": 16521 + }, + { + "epoch": 0.86, + "learning_rate": 1.9532585518894094e-08, + "loss": 0.1216, + "step": 16522 + }, + { + "epoch": 0.86, + "learning_rate": 1.9518014303203968e-08, + "loss": 0.1135, + "step": 16523 + }, + { + "epoch": 0.86, + "learning_rate": 1.950344824575867e-08, + "loss": 0.1397, + "step": 16524 + }, + { + "epoch": 0.86, + "learning_rate": 1.9488887346974514e-08, + "loss": 0.1336, + "step": 16525 + }, + { + "epoch": 0.86, + "learning_rate": 1.9474331607267615e-08, + "loss": 0.1217, + "step": 16526 + }, + { + "epoch": 0.86, + "learning_rate": 1.9459781027054034e-08, + "loss": 0.106, + "step": 16527 + }, + { + "epoch": 0.86, + "learning_rate": 1.9445235606749554e-08, + "loss": 0.1387, + "step": 16528 + }, + { + "epoch": 0.86, + "learning_rate": 1.9430695346769977e-08, + "loss": 0.1422, + "step": 16529 + }, + { + "epoch": 0.86, + "learning_rate": 1.9416160247530855e-08, + "loss": 0.164, + "step": 16530 + }, + { + "epoch": 0.86, + "learning_rate": 1.940163030944757e-08, + "loss": 0.1442, + "step": 16531 + }, + { + "epoch": 0.86, + "learning_rate": 1.9387105532935366e-08, + "loss": 0.1478, + "step": 16532 + }, + { + "epoch": 0.86, + "learning_rate": 1.937258591840949e-08, + "loss": 0.1814, + "step": 16533 + }, + { + "epoch": 0.86, + "learning_rate": 1.9358071466284788e-08, + "loss": 0.1664, + "step": 16534 + }, + { + "epoch": 0.86, + "learning_rate": 1.9343562176976126e-08, + "loss": 0.1344, + "step": 16535 + }, + { + "epoch": 0.86, + "learning_rate": 1.9329058050898216e-08, + "loss": 0.1495, + "step": 16536 + }, + { + "epoch": 0.86, + "learning_rate": 1.9314559088465576e-08, + "loss": 0.1279, + "step": 16537 + }, + { + "epoch": 0.86, + "learning_rate": 1.9300065290092603e-08, + "loss": 0.1229, + "step": 16538 + }, + { + "epoch": 0.86, + "learning_rate": 1.9285576656193458e-08, + "loss": 0.1633, + "step": 16539 + }, + { + "epoch": 0.86, + "learning_rate": 1.9271093187182364e-08, + "loss": 0.1147, + "step": 16540 + }, + { + "epoch": 0.86, + "learning_rate": 1.925661488347321e-08, + "loss": 0.1358, + "step": 16541 + }, + { + "epoch": 0.86, + "learning_rate": 1.9242141745479712e-08, + "loss": 0.1372, + "step": 16542 + }, + { + "epoch": 0.86, + "learning_rate": 1.9227673773615606e-08, + "loss": 0.1308, + "step": 16543 + }, + { + "epoch": 0.86, + "learning_rate": 1.921321096829438e-08, + "loss": 0.121, + "step": 16544 + }, + { + "epoch": 0.86, + "learning_rate": 1.9198753329929374e-08, + "loss": 0.1228, + "step": 16545 + }, + { + "epoch": 0.86, + "learning_rate": 1.9184300858933745e-08, + "loss": 0.1026, + "step": 16546 + }, + { + "epoch": 0.86, + "learning_rate": 1.916985355572065e-08, + "loss": 0.1364, + "step": 16547 + }, + { + "epoch": 0.86, + "learning_rate": 1.9155411420702938e-08, + "loss": 0.1354, + "step": 16548 + }, + { + "epoch": 0.86, + "learning_rate": 1.914097445429339e-08, + "loss": 0.1383, + "step": 16549 + }, + { + "epoch": 0.86, + "learning_rate": 1.912654265690461e-08, + "loss": 0.1761, + "step": 16550 + }, + { + "epoch": 0.86, + "learning_rate": 1.9112116028949065e-08, + "loss": 0.1205, + "step": 16551 + }, + { + "epoch": 0.86, + "learning_rate": 1.9097694570839073e-08, + "loss": 0.1131, + "step": 16552 + }, + { + "epoch": 0.86, + "learning_rate": 1.908327828298677e-08, + "loss": 0.137, + "step": 16553 + }, + { + "epoch": 0.86, + "learning_rate": 1.906886716580425e-08, + "loss": 0.1703, + "step": 16554 + }, + { + "epoch": 0.86, + "learning_rate": 1.9054461219703354e-08, + "loss": 0.1385, + "step": 16555 + }, + { + "epoch": 0.86, + "learning_rate": 1.9040060445095806e-08, + "loss": 0.1182, + "step": 16556 + }, + { + "epoch": 0.86, + "learning_rate": 1.9025664842393142e-08, + "loss": 0.1132, + "step": 16557 + }, + { + "epoch": 0.86, + "learning_rate": 1.901127441200694e-08, + "loss": 0.1452, + "step": 16558 + }, + { + "epoch": 0.86, + "learning_rate": 1.8996889154348318e-08, + "loss": 0.1535, + "step": 16559 + }, + { + "epoch": 0.86, + "learning_rate": 1.8982509069828457e-08, + "loss": 0.1354, + "step": 16560 + }, + { + "epoch": 0.86, + "learning_rate": 1.8968134158858407e-08, + "loss": 0.1208, + "step": 16561 + }, + { + "epoch": 0.86, + "learning_rate": 1.895376442184895e-08, + "loss": 0.1271, + "step": 16562 + }, + { + "epoch": 0.86, + "learning_rate": 1.8939399859210803e-08, + "loss": 0.1251, + "step": 16563 + }, + { + "epoch": 0.86, + "learning_rate": 1.892504047135448e-08, + "loss": 0.123, + "step": 16564 + }, + { + "epoch": 0.86, + "learning_rate": 1.8910686258690432e-08, + "loss": 0.1169, + "step": 16565 + }, + { + "epoch": 0.86, + "learning_rate": 1.8896337221628865e-08, + "loss": 0.1474, + "step": 16566 + }, + { + "epoch": 0.86, + "learning_rate": 1.8881993360579918e-08, + "loss": 0.1421, + "step": 16567 + }, + { + "epoch": 0.86, + "learning_rate": 1.8867654675953504e-08, + "loss": 0.1296, + "step": 16568 + }, + { + "epoch": 0.86, + "learning_rate": 1.8853321168159452e-08, + "loss": 0.1185, + "step": 16569 + }, + { + "epoch": 0.86, + "learning_rate": 1.883899283760739e-08, + "loss": 0.1084, + "step": 16570 + }, + { + "epoch": 0.86, + "learning_rate": 1.882466968470684e-08, + "loss": 0.1124, + "step": 16571 + }, + { + "epoch": 0.86, + "learning_rate": 1.88103517098672e-08, + "loss": 0.1294, + "step": 16572 + }, + { + "epoch": 0.87, + "learning_rate": 1.8796038913497635e-08, + "loss": 0.1322, + "step": 16573 + }, + { + "epoch": 0.87, + "learning_rate": 1.878173129600724e-08, + "loss": 0.1713, + "step": 16574 + }, + { + "epoch": 0.87, + "learning_rate": 1.8767428857804936e-08, + "loss": 0.1322, + "step": 16575 + }, + { + "epoch": 0.87, + "learning_rate": 1.8753131599299475e-08, + "loss": 0.1135, + "step": 16576 + }, + { + "epoch": 0.87, + "learning_rate": 1.8738839520899476e-08, + "loss": 0.122, + "step": 16577 + }, + { + "epoch": 0.87, + "learning_rate": 1.8724552623013356e-08, + "loss": 0.1371, + "step": 16578 + }, + { + "epoch": 0.87, + "learning_rate": 1.8710270906049574e-08, + "loss": 0.1204, + "step": 16579 + }, + { + "epoch": 0.87, + "learning_rate": 1.8695994370416225e-08, + "loss": 0.1177, + "step": 16580 + }, + { + "epoch": 0.87, + "learning_rate": 1.8681723016521334e-08, + "loss": 0.1075, + "step": 16581 + }, + { + "epoch": 0.87, + "learning_rate": 1.8667456844772776e-08, + "loss": 0.1378, + "step": 16582 + }, + { + "epoch": 0.87, + "learning_rate": 1.865319585557832e-08, + "loss": 0.1671, + "step": 16583 + }, + { + "epoch": 0.87, + "learning_rate": 1.8638940049345585e-08, + "loss": 0.1243, + "step": 16584 + }, + { + "epoch": 0.87, + "learning_rate": 1.8624689426481876e-08, + "loss": 0.1335, + "step": 16585 + }, + { + "epoch": 0.87, + "learning_rate": 1.8610443987394596e-08, + "loss": 0.1083, + "step": 16586 + }, + { + "epoch": 0.87, + "learning_rate": 1.8596203732490867e-08, + "loss": 0.1467, + "step": 16587 + }, + { + "epoch": 0.87, + "learning_rate": 1.858196866217765e-08, + "loss": 0.1424, + "step": 16588 + }, + { + "epoch": 0.87, + "learning_rate": 1.856773877686175e-08, + "loss": 0.1268, + "step": 16589 + }, + { + "epoch": 0.87, + "learning_rate": 1.8553514076949984e-08, + "loss": 0.1809, + "step": 16590 + }, + { + "epoch": 0.87, + "learning_rate": 1.8539294562848817e-08, + "loss": 0.1426, + "step": 16591 + }, + { + "epoch": 0.87, + "learning_rate": 1.8525080234964664e-08, + "loss": 0.1177, + "step": 16592 + }, + { + "epoch": 0.87, + "learning_rate": 1.8510871093703772e-08, + "loss": 0.0998, + "step": 16593 + }, + { + "epoch": 0.87, + "learning_rate": 1.8496667139472243e-08, + "loss": 0.1139, + "step": 16594 + }, + { + "epoch": 0.87, + "learning_rate": 1.848246837267604e-08, + "loss": 0.1188, + "step": 16595 + }, + { + "epoch": 0.87, + "learning_rate": 1.8468274793720927e-08, + "loss": 0.119, + "step": 16596 + }, + { + "epoch": 0.87, + "learning_rate": 1.8454086403012627e-08, + "loss": 0.1267, + "step": 16597 + }, + { + "epoch": 0.87, + "learning_rate": 1.8439903200956632e-08, + "loss": 0.1017, + "step": 16598 + }, + { + "epoch": 0.87, + "learning_rate": 1.8425725187958286e-08, + "loss": 0.1136, + "step": 16599 + }, + { + "epoch": 0.87, + "learning_rate": 1.841155236442278e-08, + "loss": 0.1316, + "step": 16600 + }, + { + "epoch": 0.87, + "learning_rate": 1.839738473075527e-08, + "loss": 0.1358, + "step": 16601 + }, + { + "epoch": 0.87, + "learning_rate": 1.8383222287360577e-08, + "loss": 0.1623, + "step": 16602 + }, + { + "epoch": 0.87, + "learning_rate": 1.836906503464346e-08, + "loss": 0.1127, + "step": 16603 + }, + { + "epoch": 0.87, + "learning_rate": 1.835491297300864e-08, + "loss": 0.107, + "step": 16604 + }, + { + "epoch": 0.87, + "learning_rate": 1.8340766102860505e-08, + "loss": 0.1554, + "step": 16605 + }, + { + "epoch": 0.87, + "learning_rate": 1.832662442460342e-08, + "loss": 0.155, + "step": 16606 + }, + { + "epoch": 0.87, + "learning_rate": 1.8312487938641485e-08, + "loss": 0.1266, + "step": 16607 + }, + { + "epoch": 0.87, + "learning_rate": 1.8298356645378845e-08, + "loss": 0.1562, + "step": 16608 + }, + { + "epoch": 0.87, + "learning_rate": 1.8284230545219303e-08, + "loss": 0.1635, + "step": 16609 + }, + { + "epoch": 0.87, + "learning_rate": 1.8270109638566634e-08, + "loss": 0.1566, + "step": 16610 + }, + { + "epoch": 0.87, + "learning_rate": 1.825599392582431e-08, + "loss": 0.1853, + "step": 16611 + }, + { + "epoch": 0.87, + "learning_rate": 1.8241883407395896e-08, + "loss": 0.1489, + "step": 16612 + }, + { + "epoch": 0.87, + "learning_rate": 1.822777808368461e-08, + "loss": 0.1317, + "step": 16613 + }, + { + "epoch": 0.87, + "learning_rate": 1.8213677955093543e-08, + "loss": 0.162, + "step": 16614 + }, + { + "epoch": 0.87, + "learning_rate": 1.81995830220258e-08, + "loss": 0.1322, + "step": 16615 + }, + { + "epoch": 0.87, + "learning_rate": 1.8185493284884145e-08, + "loss": 0.1269, + "step": 16616 + }, + { + "epoch": 0.87, + "learning_rate": 1.817140874407126e-08, + "loss": 0.1647, + "step": 16617 + }, + { + "epoch": 0.87, + "learning_rate": 1.815732939998973e-08, + "loss": 0.1481, + "step": 16618 + }, + { + "epoch": 0.87, + "learning_rate": 1.8143255253041922e-08, + "loss": 0.1283, + "step": 16619 + }, + { + "epoch": 0.87, + "learning_rate": 1.8129186303630073e-08, + "loss": 0.1262, + "step": 16620 + }, + { + "epoch": 0.87, + "learning_rate": 1.8115122552156303e-08, + "loss": 0.0927, + "step": 16621 + }, + { + "epoch": 0.87, + "learning_rate": 1.8101063999022493e-08, + "loss": 0.1787, + "step": 16622 + }, + { + "epoch": 0.87, + "learning_rate": 1.8087010644630542e-08, + "loss": 0.1884, + "step": 16623 + }, + { + "epoch": 0.87, + "learning_rate": 1.807296248938206e-08, + "loss": 0.1435, + "step": 16624 + }, + { + "epoch": 0.87, + "learning_rate": 1.805891953367853e-08, + "loss": 0.1384, + "step": 16625 + }, + { + "epoch": 0.87, + "learning_rate": 1.804488177792127e-08, + "loss": 0.1183, + "step": 16626 + }, + { + "epoch": 0.87, + "learning_rate": 1.8030849222511614e-08, + "loss": 0.1393, + "step": 16627 + }, + { + "epoch": 0.87, + "learning_rate": 1.8016821867850497e-08, + "loss": 0.1432, + "step": 16628 + }, + { + "epoch": 0.87, + "learning_rate": 1.8002799714338822e-08, + "loss": 0.1275, + "step": 16629 + }, + { + "epoch": 0.87, + "learning_rate": 1.798878276237743e-08, + "loss": 0.1517, + "step": 16630 + }, + { + "epoch": 0.87, + "learning_rate": 1.7974771012366886e-08, + "loss": 0.1378, + "step": 16631 + }, + { + "epoch": 0.87, + "learning_rate": 1.7960764464707668e-08, + "loss": 0.1102, + "step": 16632 + }, + { + "epoch": 0.87, + "learning_rate": 1.7946763119800034e-08, + "loss": 0.1412, + "step": 16633 + }, + { + "epoch": 0.87, + "learning_rate": 1.7932766978044222e-08, + "loss": 0.1116, + "step": 16634 + }, + { + "epoch": 0.87, + "learning_rate": 1.7918776039840246e-08, + "loss": 0.1337, + "step": 16635 + }, + { + "epoch": 0.87, + "learning_rate": 1.7904790305587868e-08, + "loss": 0.184, + "step": 16636 + }, + { + "epoch": 0.87, + "learning_rate": 1.7890809775686932e-08, + "loss": 0.1277, + "step": 16637 + }, + { + "epoch": 0.87, + "learning_rate": 1.7876834450536936e-08, + "loss": 0.1546, + "step": 16638 + }, + { + "epoch": 0.87, + "learning_rate": 1.786286433053732e-08, + "loss": 0.1555, + "step": 16639 + }, + { + "epoch": 0.87, + "learning_rate": 1.7848899416087337e-08, + "loss": 0.1175, + "step": 16640 + }, + { + "epoch": 0.87, + "learning_rate": 1.783493970758614e-08, + "loss": 0.1491, + "step": 16641 + }, + { + "epoch": 0.87, + "learning_rate": 1.7820985205432694e-08, + "loss": 0.1324, + "step": 16642 + }, + { + "epoch": 0.87, + "learning_rate": 1.7807035910025834e-08, + "loss": 0.137, + "step": 16643 + }, + { + "epoch": 0.87, + "learning_rate": 1.7793091821764206e-08, + "loss": 0.1131, + "step": 16644 + }, + { + "epoch": 0.87, + "learning_rate": 1.777915294104635e-08, + "loss": 0.0996, + "step": 16645 + }, + { + "epoch": 0.87, + "learning_rate": 1.7765219268270658e-08, + "loss": 0.1356, + "step": 16646 + }, + { + "epoch": 0.87, + "learning_rate": 1.7751290803835307e-08, + "loss": 0.1305, + "step": 16647 + }, + { + "epoch": 0.87, + "learning_rate": 1.773736754813846e-08, + "loss": 0.1684, + "step": 16648 + }, + { + "epoch": 0.87, + "learning_rate": 1.7723449501578e-08, + "loss": 0.2015, + "step": 16649 + }, + { + "epoch": 0.87, + "learning_rate": 1.7709536664551728e-08, + "loss": 0.1436, + "step": 16650 + }, + { + "epoch": 0.87, + "learning_rate": 1.7695629037457226e-08, + "loss": 0.1626, + "step": 16651 + }, + { + "epoch": 0.87, + "learning_rate": 1.7681726620692093e-08, + "loss": 0.1194, + "step": 16652 + }, + { + "epoch": 0.87, + "learning_rate": 1.766782941465357e-08, + "loss": 0.1149, + "step": 16653 + }, + { + "epoch": 0.87, + "learning_rate": 1.765393741973884e-08, + "loss": 0.1494, + "step": 16654 + }, + { + "epoch": 0.87, + "learning_rate": 1.7640050636344993e-08, + "loss": 0.1151, + "step": 16655 + }, + { + "epoch": 0.87, + "learning_rate": 1.762616906486889e-08, + "loss": 0.1241, + "step": 16656 + }, + { + "epoch": 0.87, + "learning_rate": 1.76122927057073e-08, + "loss": 0.1418, + "step": 16657 + }, + { + "epoch": 0.87, + "learning_rate": 1.7598421559256728e-08, + "loss": 0.1373, + "step": 16658 + }, + { + "epoch": 0.87, + "learning_rate": 1.7584555625913744e-08, + "loss": 0.1375, + "step": 16659 + }, + { + "epoch": 0.87, + "learning_rate": 1.757069490607457e-08, + "loss": 0.1162, + "step": 16660 + }, + { + "epoch": 0.87, + "learning_rate": 1.755683940013537e-08, + "loss": 0.095, + "step": 16661 + }, + { + "epoch": 0.87, + "learning_rate": 1.7542989108492123e-08, + "loss": 0.1632, + "step": 16662 + }, + { + "epoch": 0.87, + "learning_rate": 1.7529144031540665e-08, + "loss": 0.13, + "step": 16663 + }, + { + "epoch": 0.87, + "learning_rate": 1.7515304169676726e-08, + "loss": 0.138, + "step": 16664 + }, + { + "epoch": 0.87, + "learning_rate": 1.750146952329581e-08, + "loss": 0.1757, + "step": 16665 + }, + { + "epoch": 0.87, + "learning_rate": 1.7487640092793378e-08, + "loss": 0.1573, + "step": 16666 + }, + { + "epoch": 0.87, + "learning_rate": 1.7473815878564647e-08, + "loss": 0.1343, + "step": 16667 + }, + { + "epoch": 0.87, + "learning_rate": 1.7459996881004723e-08, + "loss": 0.1251, + "step": 16668 + }, + { + "epoch": 0.87, + "learning_rate": 1.7446183100508514e-08, + "loss": 0.1238, + "step": 16669 + }, + { + "epoch": 0.87, + "learning_rate": 1.743237453747095e-08, + "loss": 0.1164, + "step": 16670 + }, + { + "epoch": 0.87, + "learning_rate": 1.7418571192286558e-08, + "loss": 0.1418, + "step": 16671 + }, + { + "epoch": 0.87, + "learning_rate": 1.7404773065349842e-08, + "loss": 0.1367, + "step": 16672 + }, + { + "epoch": 0.87, + "learning_rate": 1.7390980157055245e-08, + "loss": 0.1304, + "step": 16673 + }, + { + "epoch": 0.87, + "learning_rate": 1.737719246779692e-08, + "loss": 0.102, + "step": 16674 + }, + { + "epoch": 0.87, + "learning_rate": 1.7363409997968947e-08, + "loss": 0.1015, + "step": 16675 + }, + { + "epoch": 0.87, + "learning_rate": 1.734963274796517e-08, + "loss": 0.1414, + "step": 16676 + }, + { + "epoch": 0.87, + "learning_rate": 1.7335860718179452e-08, + "loss": 0.1123, + "step": 16677 + }, + { + "epoch": 0.87, + "learning_rate": 1.7322093909005363e-08, + "loss": 0.0915, + "step": 16678 + }, + { + "epoch": 0.87, + "learning_rate": 1.7308332320836306e-08, + "loss": 0.1212, + "step": 16679 + }, + { + "epoch": 0.87, + "learning_rate": 1.729457595406565e-08, + "loss": 0.1244, + "step": 16680 + }, + { + "epoch": 0.87, + "learning_rate": 1.728082480908657e-08, + "loss": 0.1095, + "step": 16681 + }, + { + "epoch": 0.87, + "learning_rate": 1.726707888629204e-08, + "loss": 0.1194, + "step": 16682 + }, + { + "epoch": 0.87, + "learning_rate": 1.7253338186074885e-08, + "loss": 0.1099, + "step": 16683 + }, + { + "epoch": 0.87, + "learning_rate": 1.7239602708827938e-08, + "loss": 0.1414, + "step": 16684 + }, + { + "epoch": 0.87, + "learning_rate": 1.722587245494367e-08, + "loss": 0.199, + "step": 16685 + }, + { + "epoch": 0.87, + "learning_rate": 1.7212147424814538e-08, + "loss": 0.1361, + "step": 16686 + }, + { + "epoch": 0.87, + "learning_rate": 1.719842761883279e-08, + "loss": 0.1383, + "step": 16687 + }, + { + "epoch": 0.87, + "learning_rate": 1.718471303739055e-08, + "loss": 0.1556, + "step": 16688 + }, + { + "epoch": 0.87, + "learning_rate": 1.7171003680879804e-08, + "loss": 0.1674, + "step": 16689 + }, + { + "epoch": 0.87, + "learning_rate": 1.7157299549692294e-08, + "loss": 0.1302, + "step": 16690 + }, + { + "epoch": 0.87, + "learning_rate": 1.71436006442198e-08, + "loss": 0.143, + "step": 16691 + }, + { + "epoch": 0.87, + "learning_rate": 1.712990696485377e-08, + "loss": 0.126, + "step": 16692 + }, + { + "epoch": 0.87, + "learning_rate": 1.7116218511985613e-08, + "loss": 0.1307, + "step": 16693 + }, + { + "epoch": 0.87, + "learning_rate": 1.7102535286006493e-08, + "loss": 0.1003, + "step": 16694 + }, + { + "epoch": 0.87, + "learning_rate": 1.708885728730758e-08, + "loss": 0.1629, + "step": 16695 + }, + { + "epoch": 0.87, + "learning_rate": 1.7075184516279695e-08, + "loss": 0.1105, + "step": 16696 + }, + { + "epoch": 0.87, + "learning_rate": 1.706151697331364e-08, + "loss": 0.1439, + "step": 16697 + }, + { + "epoch": 0.87, + "learning_rate": 1.7047854658800076e-08, + "loss": 0.1391, + "step": 16698 + }, + { + "epoch": 0.87, + "learning_rate": 1.703419757312945e-08, + "loss": 0.1487, + "step": 16699 + }, + { + "epoch": 0.87, + "learning_rate": 1.702054571669209e-08, + "loss": 0.2121, + "step": 16700 + }, + { + "epoch": 0.87, + "learning_rate": 1.7006899089878136e-08, + "loss": 0.1747, + "step": 16701 + }, + { + "epoch": 0.87, + "learning_rate": 1.699325769307769e-08, + "loss": 0.1416, + "step": 16702 + }, + { + "epoch": 0.87, + "learning_rate": 1.6979621526680576e-08, + "loss": 0.1311, + "step": 16703 + }, + { + "epoch": 0.87, + "learning_rate": 1.696599059107653e-08, + "loss": 0.1626, + "step": 16704 + }, + { + "epoch": 0.87, + "learning_rate": 1.6952364886655146e-08, + "loss": 0.1327, + "step": 16705 + }, + { + "epoch": 0.87, + "learning_rate": 1.6938744413805828e-08, + "loss": 0.1607, + "step": 16706 + }, + { + "epoch": 0.87, + "learning_rate": 1.692512917291786e-08, + "loss": 0.1331, + "step": 16707 + }, + { + "epoch": 0.87, + "learning_rate": 1.691151916438034e-08, + "loss": 0.1504, + "step": 16708 + }, + { + "epoch": 0.87, + "learning_rate": 1.6897914388582323e-08, + "loss": 0.1215, + "step": 16709 + }, + { + "epoch": 0.87, + "learning_rate": 1.6884314845912596e-08, + "loss": 0.1173, + "step": 16710 + }, + { + "epoch": 0.87, + "learning_rate": 1.687072053675984e-08, + "loss": 0.1566, + "step": 16711 + }, + { + "epoch": 0.87, + "learning_rate": 1.6857131461512574e-08, + "loss": 0.1515, + "step": 16712 + }, + { + "epoch": 0.87, + "learning_rate": 1.6843547620559197e-08, + "loss": 0.1297, + "step": 16713 + }, + { + "epoch": 0.87, + "learning_rate": 1.682996901428795e-08, + "loss": 0.1261, + "step": 16714 + }, + { + "epoch": 0.87, + "learning_rate": 1.6816395643086856e-08, + "loss": 0.1498, + "step": 16715 + }, + { + "epoch": 0.87, + "learning_rate": 1.680282750734392e-08, + "loss": 0.1568, + "step": 16716 + }, + { + "epoch": 0.87, + "learning_rate": 1.6789264607446896e-08, + "loss": 0.1304, + "step": 16717 + }, + { + "epoch": 0.87, + "learning_rate": 1.6775706943783406e-08, + "loss": 0.0827, + "step": 16718 + }, + { + "epoch": 0.87, + "learning_rate": 1.6762154516740924e-08, + "loss": 0.117, + "step": 16719 + }, + { + "epoch": 0.87, + "learning_rate": 1.6748607326706842e-08, + "loss": 0.1185, + "step": 16720 + }, + { + "epoch": 0.87, + "learning_rate": 1.6735065374068326e-08, + "loss": 0.1319, + "step": 16721 + }, + { + "epoch": 0.87, + "learning_rate": 1.672152865921237e-08, + "loss": 0.1266, + "step": 16722 + }, + { + "epoch": 0.87, + "learning_rate": 1.6707997182525825e-08, + "loss": 0.1242, + "step": 16723 + }, + { + "epoch": 0.87, + "learning_rate": 1.669447094439551e-08, + "loss": 0.1541, + "step": 16724 + }, + { + "epoch": 0.87, + "learning_rate": 1.668094994520799e-08, + "loss": 0.1201, + "step": 16725 + }, + { + "epoch": 0.87, + "learning_rate": 1.666743418534966e-08, + "loss": 0.1372, + "step": 16726 + }, + { + "epoch": 0.87, + "learning_rate": 1.665392366520686e-08, + "loss": 0.1369, + "step": 16727 + }, + { + "epoch": 0.87, + "learning_rate": 1.66404183851657e-08, + "loss": 0.1366, + "step": 16728 + }, + { + "epoch": 0.87, + "learning_rate": 1.6626918345612184e-08, + "loss": 0.1497, + "step": 16729 + }, + { + "epoch": 0.87, + "learning_rate": 1.6613423546932116e-08, + "loss": 0.1324, + "step": 16730 + }, + { + "epoch": 0.87, + "learning_rate": 1.6599933989511183e-08, + "loss": 0.1246, + "step": 16731 + }, + { + "epoch": 0.87, + "learning_rate": 1.6586449673734947e-08, + "loss": 0.1164, + "step": 16732 + }, + { + "epoch": 0.87, + "learning_rate": 1.657297059998879e-08, + "loss": 0.1349, + "step": 16733 + }, + { + "epoch": 0.87, + "learning_rate": 1.655949676865791e-08, + "loss": 0.1232, + "step": 16734 + }, + { + "epoch": 0.87, + "learning_rate": 1.654602818012747e-08, + "loss": 0.124, + "step": 16735 + }, + { + "epoch": 0.87, + "learning_rate": 1.6532564834782336e-08, + "loss": 0.1256, + "step": 16736 + }, + { + "epoch": 0.87, + "learning_rate": 1.651910673300734e-08, + "loss": 0.1222, + "step": 16737 + }, + { + "epoch": 0.87, + "learning_rate": 1.650565387518712e-08, + "loss": 0.1145, + "step": 16738 + }, + { + "epoch": 0.87, + "learning_rate": 1.649220626170613e-08, + "loss": 0.1149, + "step": 16739 + }, + { + "epoch": 0.87, + "learning_rate": 1.6478763892948733e-08, + "loss": 0.1363, + "step": 16740 + }, + { + "epoch": 0.87, + "learning_rate": 1.646532676929906e-08, + "loss": 0.1165, + "step": 16741 + }, + { + "epoch": 0.87, + "learning_rate": 1.6451894891141248e-08, + "loss": 0.1097, + "step": 16742 + }, + { + "epoch": 0.87, + "learning_rate": 1.643846825885915e-08, + "loss": 0.1399, + "step": 16743 + }, + { + "epoch": 0.87, + "learning_rate": 1.6425046872836457e-08, + "loss": 0.1177, + "step": 16744 + }, + { + "epoch": 0.87, + "learning_rate": 1.6411630733456794e-08, + "loss": 0.16, + "step": 16745 + }, + { + "epoch": 0.87, + "learning_rate": 1.639821984110361e-08, + "loss": 0.1484, + "step": 16746 + }, + { + "epoch": 0.87, + "learning_rate": 1.6384814196160222e-08, + "loss": 0.1476, + "step": 16747 + }, + { + "epoch": 0.87, + "learning_rate": 1.6371413799009638e-08, + "loss": 0.1287, + "step": 16748 + }, + { + "epoch": 0.87, + "learning_rate": 1.6358018650034987e-08, + "loss": 0.1248, + "step": 16749 + }, + { + "epoch": 0.87, + "learning_rate": 1.6344628749619062e-08, + "loss": 0.12, + "step": 16750 + }, + { + "epoch": 0.87, + "learning_rate": 1.6331244098144547e-08, + "loss": 0.144, + "step": 16751 + }, + { + "epoch": 0.87, + "learning_rate": 1.631786469599392e-08, + "loss": 0.1591, + "step": 16752 + }, + { + "epoch": 0.87, + "learning_rate": 1.6304490543549697e-08, + "loss": 0.1404, + "step": 16753 + }, + { + "epoch": 0.87, + "learning_rate": 1.6291121641194017e-08, + "loss": 0.1656, + "step": 16754 + }, + { + "epoch": 0.87, + "learning_rate": 1.6277757989309016e-08, + "loss": 0.1621, + "step": 16755 + }, + { + "epoch": 0.87, + "learning_rate": 1.626439958827661e-08, + "loss": 0.1416, + "step": 16756 + }, + { + "epoch": 0.87, + "learning_rate": 1.625104643847859e-08, + "loss": 0.1273, + "step": 16757 + }, + { + "epoch": 0.87, + "learning_rate": 1.6237698540296617e-08, + "loss": 0.1244, + "step": 16758 + }, + { + "epoch": 0.87, + "learning_rate": 1.6224355894112108e-08, + "loss": 0.1288, + "step": 16759 + }, + { + "epoch": 0.87, + "learning_rate": 1.6211018500306483e-08, + "loss": 0.1395, + "step": 16760 + }, + { + "epoch": 0.87, + "learning_rate": 1.6197686359260908e-08, + "loss": 0.112, + "step": 16761 + }, + { + "epoch": 0.87, + "learning_rate": 1.618435947135641e-08, + "loss": 0.1257, + "step": 16762 + }, + { + "epoch": 0.87, + "learning_rate": 1.617103783697382e-08, + "loss": 0.1172, + "step": 16763 + }, + { + "epoch": 0.88, + "learning_rate": 1.6157721456494032e-08, + "loss": 0.1324, + "step": 16764 + }, + { + "epoch": 0.88, + "learning_rate": 1.6144410330297497e-08, + "loss": 0.158, + "step": 16765 + }, + { + "epoch": 0.88, + "learning_rate": 1.6131104458764645e-08, + "loss": 0.1457, + "step": 16766 + }, + { + "epoch": 0.88, + "learning_rate": 1.6117803842275857e-08, + "loss": 0.1428, + "step": 16767 + }, + { + "epoch": 0.88, + "learning_rate": 1.610450848121121e-08, + "loss": 0.13, + "step": 16768 + }, + { + "epoch": 0.88, + "learning_rate": 1.609121837595071e-08, + "loss": 0.1493, + "step": 16769 + }, + { + "epoch": 0.88, + "learning_rate": 1.6077933526874142e-08, + "loss": 0.2179, + "step": 16770 + }, + { + "epoch": 0.88, + "learning_rate": 1.606465393436127e-08, + "loss": 0.1127, + "step": 16771 + }, + { + "epoch": 0.88, + "learning_rate": 1.6051379598791636e-08, + "loss": 0.1287, + "step": 16772 + }, + { + "epoch": 0.88, + "learning_rate": 1.603811052054451e-08, + "loss": 0.1462, + "step": 16773 + }, + { + "epoch": 0.88, + "learning_rate": 1.602484669999924e-08, + "loss": 0.1462, + "step": 16774 + }, + { + "epoch": 0.88, + "learning_rate": 1.601158813753485e-08, + "loss": 0.1627, + "step": 16775 + }, + { + "epoch": 0.88, + "learning_rate": 1.5998334833530325e-08, + "loss": 0.1048, + "step": 16776 + }, + { + "epoch": 0.88, + "learning_rate": 1.598508678836439e-08, + "loss": 0.1683, + "step": 16777 + }, + { + "epoch": 0.88, + "learning_rate": 1.5971844002415737e-08, + "loss": 0.1016, + "step": 16778 + }, + { + "epoch": 0.88, + "learning_rate": 1.595860647606282e-08, + "loss": 0.1499, + "step": 16779 + }, + { + "epoch": 0.88, + "learning_rate": 1.594537420968398e-08, + "loss": 0.1358, + "step": 16780 + }, + { + "epoch": 0.88, + "learning_rate": 1.5932147203657386e-08, + "loss": 0.1793, + "step": 16781 + }, + { + "epoch": 0.88, + "learning_rate": 1.5918925458361087e-08, + "loss": 0.1261, + "step": 16782 + }, + { + "epoch": 0.88, + "learning_rate": 1.5905708974172983e-08, + "loss": 0.1128, + "step": 16783 + }, + { + "epoch": 0.88, + "learning_rate": 1.5892497751470724e-08, + "loss": 0.1185, + "step": 16784 + }, + { + "epoch": 0.88, + "learning_rate": 1.587929179063201e-08, + "loss": 0.1278, + "step": 16785 + }, + { + "epoch": 0.88, + "learning_rate": 1.5866091092034183e-08, + "loss": 0.1246, + "step": 16786 + }, + { + "epoch": 0.88, + "learning_rate": 1.585289565605459e-08, + "loss": 0.1233, + "step": 16787 + }, + { + "epoch": 0.88, + "learning_rate": 1.5839705483070275e-08, + "loss": 0.1328, + "step": 16788 + }, + { + "epoch": 0.88, + "learning_rate": 1.58265205734583e-08, + "loss": 0.1328, + "step": 16789 + }, + { + "epoch": 0.88, + "learning_rate": 1.5813340927595498e-08, + "loss": 0.1195, + "step": 16790 + }, + { + "epoch": 0.88, + "learning_rate": 1.580016654585845e-08, + "loss": 0.1462, + "step": 16791 + }, + { + "epoch": 0.88, + "learning_rate": 1.5786997428623794e-08, + "loss": 0.1477, + "step": 16792 + }, + { + "epoch": 0.88, + "learning_rate": 1.577383357626787e-08, + "loss": 0.154, + "step": 16793 + }, + { + "epoch": 0.88, + "learning_rate": 1.576067498916689e-08, + "loss": 0.1624, + "step": 16794 + }, + { + "epoch": 0.88, + "learning_rate": 1.5747521667696904e-08, + "loss": 0.1658, + "step": 16795 + }, + { + "epoch": 0.88, + "learning_rate": 1.573437361223391e-08, + "loss": 0.1298, + "step": 16796 + }, + { + "epoch": 0.88, + "learning_rate": 1.5721230823153665e-08, + "loss": 0.127, + "step": 16797 + }, + { + "epoch": 0.88, + "learning_rate": 1.5708093300831782e-08, + "loss": 0.1383, + "step": 16798 + }, + { + "epoch": 0.88, + "learning_rate": 1.5694961045643717e-08, + "loss": 0.1819, + "step": 16799 + }, + { + "epoch": 0.88, + "learning_rate": 1.5681834057964838e-08, + "loss": 0.1437, + "step": 16800 + }, + { + "epoch": 0.88, + "learning_rate": 1.5668712338170265e-08, + "loss": 0.1212, + "step": 16801 + }, + { + "epoch": 0.88, + "learning_rate": 1.5655595886635054e-08, + "loss": 0.1161, + "step": 16802 + }, + { + "epoch": 0.88, + "learning_rate": 1.564248470373408e-08, + "loss": 0.1526, + "step": 16803 + }, + { + "epoch": 0.88, + "learning_rate": 1.5629378789842075e-08, + "loss": 0.1224, + "step": 16804 + }, + { + "epoch": 0.88, + "learning_rate": 1.5616278145333617e-08, + "loss": 0.1255, + "step": 16805 + }, + { + "epoch": 0.88, + "learning_rate": 1.5603182770583058e-08, + "loss": 0.1219, + "step": 16806 + }, + { + "epoch": 0.88, + "learning_rate": 1.559009266596478e-08, + "loss": 0.1534, + "step": 16807 + }, + { + "epoch": 0.88, + "learning_rate": 1.5577007831852807e-08, + "loss": 0.1228, + "step": 16808 + }, + { + "epoch": 0.88, + "learning_rate": 1.5563928268621117e-08, + "loss": 0.1126, + "step": 16809 + }, + { + "epoch": 0.88, + "learning_rate": 1.5550853976643597e-08, + "loss": 0.1307, + "step": 16810 + }, + { + "epoch": 0.88, + "learning_rate": 1.5537784956293875e-08, + "loss": 0.1394, + "step": 16811 + }, + { + "epoch": 0.88, + "learning_rate": 1.5524721207945456e-08, + "loss": 0.1289, + "step": 16812 + }, + { + "epoch": 0.88, + "learning_rate": 1.551166273197171e-08, + "loss": 0.1563, + "step": 16813 + }, + { + "epoch": 0.88, + "learning_rate": 1.5498609528745865e-08, + "loss": 0.1074, + "step": 16814 + }, + { + "epoch": 0.88, + "learning_rate": 1.548556159864105e-08, + "loss": 0.128, + "step": 16815 + }, + { + "epoch": 0.88, + "learning_rate": 1.5472518942030032e-08, + "loss": 0.1819, + "step": 16816 + }, + { + "epoch": 0.88, + "learning_rate": 1.545948155928569e-08, + "loss": 0.1234, + "step": 16817 + }, + { + "epoch": 0.88, + "learning_rate": 1.5446449450780597e-08, + "loss": 0.1549, + "step": 16818 + }, + { + "epoch": 0.88, + "learning_rate": 1.5433422616887225e-08, + "loss": 0.1552, + "step": 16819 + }, + { + "epoch": 0.88, + "learning_rate": 1.542040105797786e-08, + "loss": 0.1374, + "step": 16820 + }, + { + "epoch": 0.88, + "learning_rate": 1.5407384774424714e-08, + "loss": 0.1406, + "step": 16821 + }, + { + "epoch": 0.88, + "learning_rate": 1.539437376659978e-08, + "loss": 0.1407, + "step": 16822 + }, + { + "epoch": 0.88, + "learning_rate": 1.538136803487491e-08, + "loss": 0.1465, + "step": 16823 + }, + { + "epoch": 0.88, + "learning_rate": 1.536836757962181e-08, + "loss": 0.1487, + "step": 16824 + }, + { + "epoch": 0.88, + "learning_rate": 1.535537240121203e-08, + "loss": 0.1191, + "step": 16825 + }, + { + "epoch": 0.88, + "learning_rate": 1.534238250001698e-08, + "loss": 0.1219, + "step": 16826 + }, + { + "epoch": 0.88, + "learning_rate": 1.53293978764079e-08, + "loss": 0.1709, + "step": 16827 + }, + { + "epoch": 0.88, + "learning_rate": 1.531641853075596e-08, + "loss": 0.1688, + "step": 16828 + }, + { + "epoch": 0.88, + "learning_rate": 1.530344446343208e-08, + "loss": 0.1446, + "step": 16829 + }, + { + "epoch": 0.88, + "learning_rate": 1.5290475674807034e-08, + "loss": 0.1317, + "step": 16830 + }, + { + "epoch": 0.88, + "learning_rate": 1.5277512165251482e-08, + "loss": 0.153, + "step": 16831 + }, + { + "epoch": 0.88, + "learning_rate": 1.5264553935135992e-08, + "loss": 0.1405, + "step": 16832 + }, + { + "epoch": 0.88, + "learning_rate": 1.525160098483085e-08, + "loss": 0.1248, + "step": 16833 + }, + { + "epoch": 0.88, + "learning_rate": 1.523865331470626e-08, + "loss": 0.138, + "step": 16834 + }, + { + "epoch": 0.88, + "learning_rate": 1.522571092513225e-08, + "loss": 0.1217, + "step": 16835 + }, + { + "epoch": 0.88, + "learning_rate": 1.5212773816478807e-08, + "loss": 0.1147, + "step": 16836 + }, + { + "epoch": 0.88, + "learning_rate": 1.519984198911559e-08, + "loss": 0.1483, + "step": 16837 + }, + { + "epoch": 0.88, + "learning_rate": 1.5186915443412217e-08, + "loss": 0.1289, + "step": 16838 + }, + { + "epoch": 0.88, + "learning_rate": 1.517399417973817e-08, + "loss": 0.1478, + "step": 16839 + }, + { + "epoch": 0.88, + "learning_rate": 1.516107819846273e-08, + "loss": 0.1136, + "step": 16840 + }, + { + "epoch": 0.88, + "learning_rate": 1.5148167499955067e-08, + "loss": 0.1721, + "step": 16841 + }, + { + "epoch": 0.88, + "learning_rate": 1.5135262084584045e-08, + "loss": 0.1217, + "step": 16842 + }, + { + "epoch": 0.88, + "learning_rate": 1.5122361952718655e-08, + "loss": 0.1045, + "step": 16843 + }, + { + "epoch": 0.88, + "learning_rate": 1.5109467104727514e-08, + "loss": 0.1297, + "step": 16844 + }, + { + "epoch": 0.88, + "learning_rate": 1.5096577540979195e-08, + "loss": 0.1776, + "step": 16845 + }, + { + "epoch": 0.88, + "learning_rate": 1.5083693261842024e-08, + "loss": 0.1427, + "step": 16846 + }, + { + "epoch": 0.88, + "learning_rate": 1.5070814267684305e-08, + "loss": 0.1132, + "step": 16847 + }, + { + "epoch": 0.88, + "learning_rate": 1.5057940558874128e-08, + "loss": 0.0935, + "step": 16848 + }, + { + "epoch": 0.88, + "learning_rate": 1.504507213577939e-08, + "loss": 0.1249, + "step": 16849 + }, + { + "epoch": 0.88, + "learning_rate": 1.5032208998767894e-08, + "loss": 0.1277, + "step": 16850 + }, + { + "epoch": 0.88, + "learning_rate": 1.501935114820727e-08, + "loss": 0.1049, + "step": 16851 + }, + { + "epoch": 0.88, + "learning_rate": 1.500649858446501e-08, + "loss": 0.1194, + "step": 16852 + }, + { + "epoch": 0.88, + "learning_rate": 1.4993651307908394e-08, + "loss": 0.1091, + "step": 16853 + }, + { + "epoch": 0.88, + "learning_rate": 1.4980809318904684e-08, + "loss": 0.1238, + "step": 16854 + }, + { + "epoch": 0.88, + "learning_rate": 1.4967972617820855e-08, + "loss": 0.101, + "step": 16855 + }, + { + "epoch": 0.88, + "learning_rate": 1.495514120502379e-08, + "loss": 0.1476, + "step": 16856 + }, + { + "epoch": 0.88, + "learning_rate": 1.494231508088022e-08, + "loss": 0.1652, + "step": 16857 + }, + { + "epoch": 0.88, + "learning_rate": 1.4929494245756756e-08, + "loss": 0.1602, + "step": 16858 + }, + { + "epoch": 0.88, + "learning_rate": 1.4916678700019782e-08, + "loss": 0.1298, + "step": 16859 + }, + { + "epoch": 0.88, + "learning_rate": 1.490386844403555e-08, + "loss": 0.1141, + "step": 16860 + }, + { + "epoch": 0.88, + "learning_rate": 1.489106347817024e-08, + "loss": 0.1329, + "step": 16861 + }, + { + "epoch": 0.88, + "learning_rate": 1.4878263802789781e-08, + "loss": 0.1282, + "step": 16862 + }, + { + "epoch": 0.88, + "learning_rate": 1.4865469418260012e-08, + "loss": 0.1385, + "step": 16863 + }, + { + "epoch": 0.88, + "learning_rate": 1.4852680324946576e-08, + "loss": 0.1521, + "step": 16864 + }, + { + "epoch": 0.88, + "learning_rate": 1.4839896523215045e-08, + "loss": 0.1502, + "step": 16865 + }, + { + "epoch": 0.88, + "learning_rate": 1.4827118013430729e-08, + "loss": 0.1402, + "step": 16866 + }, + { + "epoch": 0.88, + "learning_rate": 1.4814344795958889e-08, + "loss": 0.1228, + "step": 16867 + }, + { + "epoch": 0.88, + "learning_rate": 1.4801576871164545e-08, + "loss": 0.137, + "step": 16868 + }, + { + "epoch": 0.88, + "learning_rate": 1.4788814239412628e-08, + "loss": 0.1579, + "step": 16869 + }, + { + "epoch": 0.88, + "learning_rate": 1.4776056901067913e-08, + "loss": 0.1416, + "step": 16870 + }, + { + "epoch": 0.88, + "learning_rate": 1.4763304856494951e-08, + "loss": 0.149, + "step": 16871 + }, + { + "epoch": 0.88, + "learning_rate": 1.4750558106058275e-08, + "loss": 0.1212, + "step": 16872 + }, + { + "epoch": 0.88, + "learning_rate": 1.4737816650122171e-08, + "loss": 0.1346, + "step": 16873 + }, + { + "epoch": 0.88, + "learning_rate": 1.472508048905079e-08, + "loss": 0.1167, + "step": 16874 + }, + { + "epoch": 0.88, + "learning_rate": 1.4712349623208109e-08, + "loss": 0.1479, + "step": 16875 + }, + { + "epoch": 0.88, + "learning_rate": 1.4699624052958015e-08, + "loss": 0.1261, + "step": 16876 + }, + { + "epoch": 0.88, + "learning_rate": 1.4686903778664216e-08, + "loss": 0.1521, + "step": 16877 + }, + { + "epoch": 0.88, + "learning_rate": 1.4674188800690201e-08, + "loss": 0.1186, + "step": 16878 + }, + { + "epoch": 0.88, + "learning_rate": 1.4661479119399434e-08, + "loss": 0.1296, + "step": 16879 + }, + { + "epoch": 0.88, + "learning_rate": 1.4648774735155157e-08, + "loss": 0.1533, + "step": 16880 + }, + { + "epoch": 0.88, + "learning_rate": 1.4636075648320434e-08, + "loss": 0.1383, + "step": 16881 + }, + { + "epoch": 0.88, + "learning_rate": 1.4623381859258222e-08, + "loss": 0.1923, + "step": 16882 + }, + { + "epoch": 0.88, + "learning_rate": 1.4610693368331317e-08, + "loss": 0.1304, + "step": 16883 + }, + { + "epoch": 0.88, + "learning_rate": 1.4598010175902431e-08, + "loss": 0.1453, + "step": 16884 + }, + { + "epoch": 0.88, + "learning_rate": 1.4585332282333895e-08, + "loss": 0.1212, + "step": 16885 + }, + { + "epoch": 0.88, + "learning_rate": 1.4572659687988176e-08, + "loss": 0.1432, + "step": 16886 + }, + { + "epoch": 0.88, + "learning_rate": 1.4559992393227428e-08, + "loss": 0.1125, + "step": 16887 + }, + { + "epoch": 0.88, + "learning_rate": 1.4547330398413693e-08, + "loss": 0.1306, + "step": 16888 + }, + { + "epoch": 0.88, + "learning_rate": 1.4534673703908795e-08, + "loss": 0.1498, + "step": 16889 + }, + { + "epoch": 0.88, + "learning_rate": 1.4522022310074578e-08, + "loss": 0.116, + "step": 16890 + }, + { + "epoch": 0.88, + "learning_rate": 1.450937621727255e-08, + "loss": 0.1278, + "step": 16891 + }, + { + "epoch": 0.88, + "learning_rate": 1.449673542586416e-08, + "loss": 0.1419, + "step": 16892 + }, + { + "epoch": 0.88, + "learning_rate": 1.448409993621067e-08, + "loss": 0.1343, + "step": 16893 + }, + { + "epoch": 0.88, + "learning_rate": 1.4471469748673238e-08, + "loss": 0.1177, + "step": 16894 + }, + { + "epoch": 0.88, + "learning_rate": 1.4458844863612819e-08, + "loss": 0.1051, + "step": 16895 + }, + { + "epoch": 0.88, + "learning_rate": 1.4446225281390211e-08, + "loss": 0.1143, + "step": 16896 + }, + { + "epoch": 0.88, + "learning_rate": 1.443361100236613e-08, + "loss": 0.1517, + "step": 16897 + }, + { + "epoch": 0.88, + "learning_rate": 1.4421002026901086e-08, + "loss": 0.1395, + "step": 16898 + }, + { + "epoch": 0.88, + "learning_rate": 1.4408398355355456e-08, + "loss": 0.134, + "step": 16899 + }, + { + "epoch": 0.88, + "learning_rate": 1.439579998808942e-08, + "loss": 0.0981, + "step": 16900 + }, + { + "epoch": 0.88, + "learning_rate": 1.438320692546311e-08, + "loss": 0.1306, + "step": 16901 + }, + { + "epoch": 0.88, + "learning_rate": 1.4370619167836395e-08, + "loss": 0.113, + "step": 16902 + }, + { + "epoch": 0.88, + "learning_rate": 1.435803671556901e-08, + "loss": 0.1468, + "step": 16903 + }, + { + "epoch": 0.88, + "learning_rate": 1.4345459569020623e-08, + "loss": 0.1556, + "step": 16904 + }, + { + "epoch": 0.88, + "learning_rate": 1.4332887728550657e-08, + "loss": 0.1139, + "step": 16905 + }, + { + "epoch": 0.88, + "learning_rate": 1.4320321194518448e-08, + "loss": 0.1709, + "step": 16906 + }, + { + "epoch": 0.88, + "learning_rate": 1.4307759967283128e-08, + "loss": 0.1322, + "step": 16907 + }, + { + "epoch": 0.88, + "learning_rate": 1.4295204047203724e-08, + "loss": 0.1156, + "step": 16908 + }, + { + "epoch": 0.88, + "learning_rate": 1.428265343463908e-08, + "loss": 0.1334, + "step": 16909 + }, + { + "epoch": 0.88, + "learning_rate": 1.4270108129947888e-08, + "loss": 0.1586, + "step": 16910 + }, + { + "epoch": 0.88, + "learning_rate": 1.4257568133488707e-08, + "loss": 0.1567, + "step": 16911 + }, + { + "epoch": 0.88, + "learning_rate": 1.4245033445619936e-08, + "loss": 0.1444, + "step": 16912 + }, + { + "epoch": 0.88, + "learning_rate": 1.4232504066699824e-08, + "loss": 0.1115, + "step": 16913 + }, + { + "epoch": 0.88, + "learning_rate": 1.4219979997086418e-08, + "loss": 0.1444, + "step": 16914 + }, + { + "epoch": 0.88, + "learning_rate": 1.4207461237137742e-08, + "loss": 0.106, + "step": 16915 + }, + { + "epoch": 0.88, + "learning_rate": 1.419494778721153e-08, + "loss": 0.1294, + "step": 16916 + }, + { + "epoch": 0.88, + "learning_rate": 1.4182439647665456e-08, + "loss": 0.1285, + "step": 16917 + }, + { + "epoch": 0.88, + "learning_rate": 1.4169936818856987e-08, + "loss": 0.1302, + "step": 16918 + }, + { + "epoch": 0.88, + "learning_rate": 1.415743930114346e-08, + "loss": 0.1865, + "step": 16919 + }, + { + "epoch": 0.88, + "learning_rate": 1.4144947094882053e-08, + "loss": 0.1304, + "step": 16920 + }, + { + "epoch": 0.88, + "learning_rate": 1.4132460200429775e-08, + "loss": 0.1318, + "step": 16921 + }, + { + "epoch": 0.88, + "learning_rate": 1.411997861814358e-08, + "loss": 0.1453, + "step": 16922 + }, + { + "epoch": 0.88, + "learning_rate": 1.410750234838014e-08, + "loss": 0.2236, + "step": 16923 + }, + { + "epoch": 0.88, + "learning_rate": 1.4095031391496059e-08, + "loss": 0.1234, + "step": 16924 + }, + { + "epoch": 0.88, + "learning_rate": 1.4082565747847697e-08, + "loss": 0.1471, + "step": 16925 + }, + { + "epoch": 0.88, + "learning_rate": 1.4070105417791411e-08, + "loss": 0.1451, + "step": 16926 + }, + { + "epoch": 0.88, + "learning_rate": 1.4057650401683341e-08, + "loss": 0.146, + "step": 16927 + }, + { + "epoch": 0.88, + "learning_rate": 1.4045200699879312e-08, + "loss": 0.1589, + "step": 16928 + }, + { + "epoch": 0.88, + "learning_rate": 1.4032756312735283e-08, + "loss": 0.1191, + "step": 16929 + }, + { + "epoch": 0.88, + "learning_rate": 1.402031724060686e-08, + "loss": 0.1194, + "step": 16930 + }, + { + "epoch": 0.88, + "learning_rate": 1.4007883483849581e-08, + "loss": 0.131, + "step": 16931 + }, + { + "epoch": 0.88, + "learning_rate": 1.3995455042818738e-08, + "loss": 0.1415, + "step": 16932 + }, + { + "epoch": 0.88, + "learning_rate": 1.3983031917869648e-08, + "loss": 0.1219, + "step": 16933 + }, + { + "epoch": 0.88, + "learning_rate": 1.3970614109357314e-08, + "loss": 0.1355, + "step": 16934 + }, + { + "epoch": 0.88, + "learning_rate": 1.3958201617636634e-08, + "loss": 0.1319, + "step": 16935 + }, + { + "epoch": 0.88, + "learning_rate": 1.3945794443062364e-08, + "loss": 0.1411, + "step": 16936 + }, + { + "epoch": 0.88, + "learning_rate": 1.3933392585989135e-08, + "loss": 0.1924, + "step": 16937 + }, + { + "epoch": 0.88, + "learning_rate": 1.392099604677135e-08, + "loss": 0.1679, + "step": 16938 + }, + { + "epoch": 0.88, + "learning_rate": 1.3908604825763303e-08, + "loss": 0.1331, + "step": 16939 + }, + { + "epoch": 0.88, + "learning_rate": 1.3896218923319225e-08, + "loss": 0.1417, + "step": 16940 + }, + { + "epoch": 0.88, + "learning_rate": 1.388383833979303e-08, + "loss": 0.1225, + "step": 16941 + }, + { + "epoch": 0.88, + "learning_rate": 1.3871463075538593e-08, + "loss": 0.1485, + "step": 16942 + }, + { + "epoch": 0.88, + "learning_rate": 1.385909313090956e-08, + "loss": 0.1186, + "step": 16943 + }, + { + "epoch": 0.88, + "learning_rate": 1.3846728506259565e-08, + "loss": 0.1276, + "step": 16944 + }, + { + "epoch": 0.88, + "learning_rate": 1.3834369201941898e-08, + "loss": 0.1448, + "step": 16945 + }, + { + "epoch": 0.88, + "learning_rate": 1.3822015218309834e-08, + "loss": 0.1293, + "step": 16946 + }, + { + "epoch": 0.88, + "learning_rate": 1.3809666555716403e-08, + "loss": 0.0867, + "step": 16947 + }, + { + "epoch": 0.88, + "learning_rate": 1.3797323214514634e-08, + "loss": 0.1212, + "step": 16948 + }, + { + "epoch": 0.88, + "learning_rate": 1.378498519505722e-08, + "loss": 0.118, + "step": 16949 + }, + { + "epoch": 0.88, + "learning_rate": 1.377265249769679e-08, + "loss": 0.1382, + "step": 16950 + }, + { + "epoch": 0.88, + "learning_rate": 1.3760325122785888e-08, + "loss": 0.1235, + "step": 16951 + }, + { + "epoch": 0.88, + "learning_rate": 1.3748003070676827e-08, + "loss": 0.1225, + "step": 16952 + }, + { + "epoch": 0.88, + "learning_rate": 1.3735686341721685e-08, + "loss": 0.1521, + "step": 16953 + }, + { + "epoch": 0.88, + "learning_rate": 1.3723374936272514e-08, + "loss": 0.1532, + "step": 16954 + }, + { + "epoch": 0.88, + "learning_rate": 1.371106885468123e-08, + "loss": 0.1379, + "step": 16955 + }, + { + "epoch": 0.89, + "learning_rate": 1.3698768097299506e-08, + "loss": 0.154, + "step": 16956 + }, + { + "epoch": 0.89, + "learning_rate": 1.368647266447891e-08, + "loss": 0.1221, + "step": 16957 + }, + { + "epoch": 0.89, + "learning_rate": 1.3674182556570823e-08, + "loss": 0.1529, + "step": 16958 + }, + { + "epoch": 0.89, + "learning_rate": 1.3661897773926545e-08, + "loss": 0.1566, + "step": 16959 + }, + { + "epoch": 0.89, + "learning_rate": 1.3649618316897171e-08, + "loss": 0.139, + "step": 16960 + }, + { + "epoch": 0.89, + "learning_rate": 1.3637344185833644e-08, + "loss": 0.1395, + "step": 16961 + }, + { + "epoch": 0.89, + "learning_rate": 1.362507538108677e-08, + "loss": 0.1373, + "step": 16962 + }, + { + "epoch": 0.89, + "learning_rate": 1.3612811903007182e-08, + "loss": 0.1225, + "step": 16963 + }, + { + "epoch": 0.89, + "learning_rate": 1.3600553751945398e-08, + "loss": 0.1609, + "step": 16964 + }, + { + "epoch": 0.89, + "learning_rate": 1.3588300928251694e-08, + "loss": 0.1112, + "step": 16965 + }, + { + "epoch": 0.89, + "learning_rate": 1.3576053432276347e-08, + "loss": 0.1655, + "step": 16966 + }, + { + "epoch": 0.89, + "learning_rate": 1.3563811264369363e-08, + "loss": 0.1201, + "step": 16967 + }, + { + "epoch": 0.89, + "learning_rate": 1.3551574424880641e-08, + "loss": 0.1267, + "step": 16968 + }, + { + "epoch": 0.89, + "learning_rate": 1.353934291415988e-08, + "loss": 0.1642, + "step": 16969 + }, + { + "epoch": 0.89, + "learning_rate": 1.3527116732556664e-08, + "loss": 0.1486, + "step": 16970 + }, + { + "epoch": 0.89, + "learning_rate": 1.3514895880420473e-08, + "loss": 0.1142, + "step": 16971 + }, + { + "epoch": 0.89, + "learning_rate": 1.3502680358100494e-08, + "loss": 0.1375, + "step": 16972 + }, + { + "epoch": 0.89, + "learning_rate": 1.3490470165945933e-08, + "loss": 0.1052, + "step": 16973 + }, + { + "epoch": 0.89, + "learning_rate": 1.3478265304305758e-08, + "loss": 0.148, + "step": 16974 + }, + { + "epoch": 0.89, + "learning_rate": 1.3466065773528734e-08, + "loss": 0.1549, + "step": 16975 + }, + { + "epoch": 0.89, + "learning_rate": 1.345387157396356e-08, + "loss": 0.178, + "step": 16976 + }, + { + "epoch": 0.89, + "learning_rate": 1.3441682705958756e-08, + "loss": 0.1601, + "step": 16977 + }, + { + "epoch": 0.89, + "learning_rate": 1.3429499169862734e-08, + "loss": 0.1622, + "step": 16978 + }, + { + "epoch": 0.89, + "learning_rate": 1.3417320966023571e-08, + "loss": 0.098, + "step": 16979 + }, + { + "epoch": 0.89, + "learning_rate": 1.3405148094789431e-08, + "loss": 0.1109, + "step": 16980 + }, + { + "epoch": 0.89, + "learning_rate": 1.3392980556508193e-08, + "loss": 0.1126, + "step": 16981 + }, + { + "epoch": 0.89, + "learning_rate": 1.3380818351527622e-08, + "loss": 0.1545, + "step": 16982 + }, + { + "epoch": 0.89, + "learning_rate": 1.3368661480195265e-08, + "loss": 0.1285, + "step": 16983 + }, + { + "epoch": 0.89, + "learning_rate": 1.335650994285864e-08, + "loss": 0.1843, + "step": 16984 + }, + { + "epoch": 0.89, + "learning_rate": 1.3344363739865006e-08, + "loss": 0.1867, + "step": 16985 + }, + { + "epoch": 0.89, + "learning_rate": 1.3332222871561527e-08, + "loss": 0.1632, + "step": 16986 + }, + { + "epoch": 0.89, + "learning_rate": 1.332008733829515e-08, + "loss": 0.1322, + "step": 16987 + }, + { + "epoch": 0.89, + "learning_rate": 1.3307957140412773e-08, + "loss": 0.1674, + "step": 16988 + }, + { + "epoch": 0.89, + "learning_rate": 1.3295832278261055e-08, + "loss": 0.1266, + "step": 16989 + }, + { + "epoch": 0.89, + "learning_rate": 1.3283712752186472e-08, + "loss": 0.1471, + "step": 16990 + }, + { + "epoch": 0.89, + "learning_rate": 1.3271598562535502e-08, + "loss": 0.1565, + "step": 16991 + }, + { + "epoch": 0.89, + "learning_rate": 1.3259489709654315e-08, + "loss": 0.127, + "step": 16992 + }, + { + "epoch": 0.89, + "learning_rate": 1.324738619388901e-08, + "loss": 0.1257, + "step": 16993 + }, + { + "epoch": 0.89, + "learning_rate": 1.3235288015585489e-08, + "loss": 0.1516, + "step": 16994 + }, + { + "epoch": 0.89, + "learning_rate": 1.3223195175089585e-08, + "loss": 0.1376, + "step": 16995 + }, + { + "epoch": 0.89, + "learning_rate": 1.3211107672746824e-08, + "loss": 0.1259, + "step": 16996 + }, + { + "epoch": 0.89, + "learning_rate": 1.3199025508902683e-08, + "loss": 0.1391, + "step": 16997 + }, + { + "epoch": 0.89, + "learning_rate": 1.3186948683902555e-08, + "loss": 0.1203, + "step": 16998 + }, + { + "epoch": 0.89, + "learning_rate": 1.3174877198091538e-08, + "loss": 0.1296, + "step": 16999 + }, + { + "epoch": 0.89, + "learning_rate": 1.316281105181465e-08, + "loss": 0.1342, + "step": 17000 + }, + { + "epoch": 0.89, + "learning_rate": 1.315075024541672e-08, + "loss": 0.1545, + "step": 17001 + }, + { + "epoch": 0.89, + "learning_rate": 1.3138694779242498e-08, + "loss": 0.1518, + "step": 17002 + }, + { + "epoch": 0.89, + "learning_rate": 1.3126644653636532e-08, + "loss": 0.1289, + "step": 17003 + }, + { + "epoch": 0.89, + "learning_rate": 1.3114599868943188e-08, + "loss": 0.1375, + "step": 17004 + }, + { + "epoch": 0.89, + "learning_rate": 1.3102560425506748e-08, + "loss": 0.1178, + "step": 17005 + }, + { + "epoch": 0.89, + "learning_rate": 1.309052632367127e-08, + "loss": 0.1346, + "step": 17006 + }, + { + "epoch": 0.89, + "learning_rate": 1.3078497563780699e-08, + "loss": 0.1455, + "step": 17007 + }, + { + "epoch": 0.89, + "learning_rate": 1.3066474146178785e-08, + "loss": 0.1669, + "step": 17008 + }, + { + "epoch": 0.89, + "learning_rate": 1.3054456071209253e-08, + "loss": 0.1212, + "step": 17009 + }, + { + "epoch": 0.89, + "learning_rate": 1.3042443339215537e-08, + "loss": 0.1159, + "step": 17010 + }, + { + "epoch": 0.89, + "learning_rate": 1.303043595054094e-08, + "loss": 0.1525, + "step": 17011 + }, + { + "epoch": 0.89, + "learning_rate": 1.301843390552868e-08, + "loss": 0.1447, + "step": 17012 + }, + { + "epoch": 0.89, + "learning_rate": 1.3006437204521769e-08, + "loss": 0.113, + "step": 17013 + }, + { + "epoch": 0.89, + "learning_rate": 1.2994445847863045e-08, + "loss": 0.12, + "step": 17014 + }, + { + "epoch": 0.89, + "learning_rate": 1.2982459835895232e-08, + "loss": 0.144, + "step": 17015 + }, + { + "epoch": 0.89, + "learning_rate": 1.2970479168960946e-08, + "loss": 0.1349, + "step": 17016 + }, + { + "epoch": 0.89, + "learning_rate": 1.2958503847402557e-08, + "loss": 0.0981, + "step": 17017 + }, + { + "epoch": 0.89, + "learning_rate": 1.2946533871562348e-08, + "loss": 0.1199, + "step": 17018 + }, + { + "epoch": 0.89, + "learning_rate": 1.2934569241782378e-08, + "loss": 0.1265, + "step": 17019 + }, + { + "epoch": 0.89, + "learning_rate": 1.2922609958404663e-08, + "loss": 0.1379, + "step": 17020 + }, + { + "epoch": 0.89, + "learning_rate": 1.2910656021770993e-08, + "loss": 0.0979, + "step": 17021 + }, + { + "epoch": 0.89, + "learning_rate": 1.2898707432222944e-08, + "loss": 0.1295, + "step": 17022 + }, + { + "epoch": 0.89, + "learning_rate": 1.2886764190102083e-08, + "loss": 0.159, + "step": 17023 + }, + { + "epoch": 0.89, + "learning_rate": 1.2874826295749763e-08, + "loss": 0.1601, + "step": 17024 + }, + { + "epoch": 0.89, + "learning_rate": 1.286289374950713e-08, + "loss": 0.1308, + "step": 17025 + }, + { + "epoch": 0.89, + "learning_rate": 1.2850966551715202e-08, + "loss": 0.1193, + "step": 17026 + }, + { + "epoch": 0.89, + "learning_rate": 1.283904470271493e-08, + "loss": 0.1369, + "step": 17027 + }, + { + "epoch": 0.89, + "learning_rate": 1.2827128202847016e-08, + "loss": 0.106, + "step": 17028 + }, + { + "epoch": 0.89, + "learning_rate": 1.281521705245201e-08, + "loss": 0.1304, + "step": 17029 + }, + { + "epoch": 0.89, + "learning_rate": 1.2803311251870397e-08, + "loss": 0.1271, + "step": 17030 + }, + { + "epoch": 0.89, + "learning_rate": 1.2791410801442393e-08, + "loss": 0.1326, + "step": 17031 + }, + { + "epoch": 0.89, + "learning_rate": 1.2779515701508148e-08, + "loss": 0.1629, + "step": 17032 + }, + { + "epoch": 0.89, + "learning_rate": 1.2767625952407568e-08, + "loss": 0.1242, + "step": 17033 + }, + { + "epoch": 0.89, + "learning_rate": 1.2755741554480559e-08, + "loss": 0.148, + "step": 17034 + }, + { + "epoch": 0.89, + "learning_rate": 1.274386250806676e-08, + "loss": 0.1599, + "step": 17035 + }, + { + "epoch": 0.89, + "learning_rate": 1.2731988813505633e-08, + "loss": 0.1285, + "step": 17036 + }, + { + "epoch": 0.89, + "learning_rate": 1.2720120471136531e-08, + "loss": 0.1289, + "step": 17037 + }, + { + "epoch": 0.89, + "learning_rate": 1.2708257481298756e-08, + "loss": 0.1191, + "step": 17038 + }, + { + "epoch": 0.89, + "learning_rate": 1.269639984433124e-08, + "loss": 0.1405, + "step": 17039 + }, + { + "epoch": 0.89, + "learning_rate": 1.2684547560572911e-08, + "loss": 0.1531, + "step": 17040 + }, + { + "epoch": 0.89, + "learning_rate": 1.2672700630362542e-08, + "loss": 0.0982, + "step": 17041 + }, + { + "epoch": 0.89, + "learning_rate": 1.2660859054038708e-08, + "loss": 0.1407, + "step": 17042 + }, + { + "epoch": 0.89, + "learning_rate": 1.2649022831939826e-08, + "loss": 0.1202, + "step": 17043 + }, + { + "epoch": 0.89, + "learning_rate": 1.263719196440418e-08, + "loss": 0.1467, + "step": 17044 + }, + { + "epoch": 0.89, + "learning_rate": 1.2625366451769947e-08, + "loss": 0.1549, + "step": 17045 + }, + { + "epoch": 0.89, + "learning_rate": 1.2613546294375077e-08, + "loss": 0.1774, + "step": 17046 + }, + { + "epoch": 0.89, + "learning_rate": 1.2601731492557388e-08, + "loss": 0.1347, + "step": 17047 + }, + { + "epoch": 0.89, + "learning_rate": 1.2589922046654567e-08, + "loss": 0.1546, + "step": 17048 + }, + { + "epoch": 0.89, + "learning_rate": 1.257811795700412e-08, + "loss": 0.1282, + "step": 17049 + }, + { + "epoch": 0.89, + "learning_rate": 1.25663192239434e-08, + "loss": 0.1553, + "step": 17050 + }, + { + "epoch": 0.89, + "learning_rate": 1.2554525847809605e-08, + "loss": 0.1398, + "step": 17051 + }, + { + "epoch": 0.89, + "learning_rate": 1.2542737828939864e-08, + "loss": 0.1298, + "step": 17052 + }, + { + "epoch": 0.89, + "learning_rate": 1.253095516767102e-08, + "loss": 0.1686, + "step": 17053 + }, + { + "epoch": 0.89, + "learning_rate": 1.251917786433987e-08, + "loss": 0.1058, + "step": 17054 + }, + { + "epoch": 0.89, + "learning_rate": 1.2507405919282987e-08, + "loss": 0.1421, + "step": 17055 + }, + { + "epoch": 0.89, + "learning_rate": 1.2495639332836816e-08, + "loss": 0.1795, + "step": 17056 + }, + { + "epoch": 0.89, + "learning_rate": 1.2483878105337664e-08, + "loss": 0.1093, + "step": 17057 + }, + { + "epoch": 0.89, + "learning_rate": 1.2472122237121662e-08, + "loss": 0.1703, + "step": 17058 + }, + { + "epoch": 0.89, + "learning_rate": 1.2460371728524766e-08, + "loss": 0.1192, + "step": 17059 + }, + { + "epoch": 0.89, + "learning_rate": 1.2448626579882881e-08, + "loss": 0.1268, + "step": 17060 + }, + { + "epoch": 0.89, + "learning_rate": 1.243688679153163e-08, + "loss": 0.1227, + "step": 17061 + }, + { + "epoch": 0.89, + "learning_rate": 1.2425152363806591e-08, + "loss": 0.1497, + "step": 17062 + }, + { + "epoch": 0.89, + "learning_rate": 1.2413423297043047e-08, + "loss": 0.1231, + "step": 17063 + }, + { + "epoch": 0.89, + "learning_rate": 1.2401699591576376e-08, + "loss": 0.1371, + "step": 17064 + }, + { + "epoch": 0.89, + "learning_rate": 1.2389981247741488e-08, + "loss": 0.1408, + "step": 17065 + }, + { + "epoch": 0.89, + "learning_rate": 1.2378268265873337e-08, + "loss": 0.1419, + "step": 17066 + }, + { + "epoch": 0.89, + "learning_rate": 1.2366560646306767e-08, + "loss": 0.1398, + "step": 17067 + }, + { + "epoch": 0.89, + "learning_rate": 1.2354858389376288e-08, + "loss": 0.1187, + "step": 17068 + }, + { + "epoch": 0.89, + "learning_rate": 1.2343161495416432e-08, + "loss": 0.1707, + "step": 17069 + }, + { + "epoch": 0.89, + "learning_rate": 1.2331469964761398e-08, + "loss": 0.1412, + "step": 17070 + }, + { + "epoch": 0.89, + "learning_rate": 1.2319783797745454e-08, + "loss": 0.1486, + "step": 17071 + }, + { + "epoch": 0.89, + "learning_rate": 1.2308102994702552e-08, + "loss": 0.1204, + "step": 17072 + }, + { + "epoch": 0.89, + "learning_rate": 1.2296427555966493e-08, + "loss": 0.1361, + "step": 17073 + }, + { + "epoch": 0.89, + "learning_rate": 1.228475748187099e-08, + "loss": 0.1477, + "step": 17074 + }, + { + "epoch": 0.89, + "learning_rate": 1.2273092772749617e-08, + "loss": 0.1507, + "step": 17075 + }, + { + "epoch": 0.89, + "learning_rate": 1.2261433428935685e-08, + "loss": 0.1175, + "step": 17076 + }, + { + "epoch": 0.89, + "learning_rate": 1.2249779450762466e-08, + "loss": 0.1317, + "step": 17077 + }, + { + "epoch": 0.89, + "learning_rate": 1.2238130838563043e-08, + "loss": 0.1172, + "step": 17078 + }, + { + "epoch": 0.89, + "learning_rate": 1.222648759267031e-08, + "loss": 0.1529, + "step": 17079 + }, + { + "epoch": 0.89, + "learning_rate": 1.2214849713417064e-08, + "loss": 0.1547, + "step": 17080 + }, + { + "epoch": 0.89, + "learning_rate": 1.2203217201135907e-08, + "loss": 0.1541, + "step": 17081 + }, + { + "epoch": 0.89, + "learning_rate": 1.2191590056159306e-08, + "loss": 0.1056, + "step": 17082 + }, + { + "epoch": 0.89, + "learning_rate": 1.2179968278819553e-08, + "loss": 0.1118, + "step": 17083 + }, + { + "epoch": 0.89, + "learning_rate": 1.2168351869448779e-08, + "loss": 0.1254, + "step": 17084 + }, + { + "epoch": 0.89, + "learning_rate": 1.2156740828379052e-08, + "loss": 0.1757, + "step": 17085 + }, + { + "epoch": 0.89, + "learning_rate": 1.2145135155942199e-08, + "loss": 0.1283, + "step": 17086 + }, + { + "epoch": 0.89, + "learning_rate": 1.2133534852469885e-08, + "loss": 0.1438, + "step": 17087 + }, + { + "epoch": 0.89, + "learning_rate": 1.2121939918293645e-08, + "loss": 0.1398, + "step": 17088 + }, + { + "epoch": 0.89, + "learning_rate": 1.2110350353744947e-08, + "loss": 0.1537, + "step": 17089 + }, + { + "epoch": 0.89, + "learning_rate": 1.209876615915495e-08, + "loss": 0.1091, + "step": 17090 + }, + { + "epoch": 0.89, + "learning_rate": 1.208718733485472e-08, + "loss": 0.1429, + "step": 17091 + }, + { + "epoch": 0.89, + "learning_rate": 1.207561388117524e-08, + "loss": 0.1387, + "step": 17092 + }, + { + "epoch": 0.89, + "learning_rate": 1.2064045798447264e-08, + "loss": 0.1102, + "step": 17093 + }, + { + "epoch": 0.89, + "learning_rate": 1.2052483087001397e-08, + "loss": 0.1345, + "step": 17094 + }, + { + "epoch": 0.89, + "learning_rate": 1.2040925747168106e-08, + "loss": 0.0968, + "step": 17095 + }, + { + "epoch": 0.89, + "learning_rate": 1.2029373779277752e-08, + "loss": 0.1543, + "step": 17096 + }, + { + "epoch": 0.89, + "learning_rate": 1.2017827183660445e-08, + "loss": 0.0975, + "step": 17097 + }, + { + "epoch": 0.89, + "learning_rate": 1.2006285960646213e-08, + "loss": 0.1652, + "step": 17098 + }, + { + "epoch": 0.89, + "learning_rate": 1.19947501105649e-08, + "loss": 0.1189, + "step": 17099 + }, + { + "epoch": 0.89, + "learning_rate": 1.19832196337462e-08, + "loss": 0.1291, + "step": 17100 + }, + { + "epoch": 0.89, + "learning_rate": 1.1971694530519672e-08, + "loss": 0.1571, + "step": 17101 + }, + { + "epoch": 0.89, + "learning_rate": 1.1960174801214674e-08, + "loss": 0.1284, + "step": 17102 + }, + { + "epoch": 0.89, + "learning_rate": 1.1948660446160497e-08, + "loss": 0.1099, + "step": 17103 + }, + { + "epoch": 0.89, + "learning_rate": 1.193715146568619e-08, + "loss": 0.143, + "step": 17104 + }, + { + "epoch": 0.89, + "learning_rate": 1.1925647860120692e-08, + "loss": 0.1322, + "step": 17105 + }, + { + "epoch": 0.89, + "learning_rate": 1.1914149629792757e-08, + "loss": 0.1109, + "step": 17106 + }, + { + "epoch": 0.89, + "learning_rate": 1.1902656775031105e-08, + "loss": 0.1441, + "step": 17107 + }, + { + "epoch": 0.89, + "learning_rate": 1.1891169296164094e-08, + "loss": 0.1409, + "step": 17108 + }, + { + "epoch": 0.89, + "learning_rate": 1.187968719352006e-08, + "loss": 0.1704, + "step": 17109 + }, + { + "epoch": 0.89, + "learning_rate": 1.1868210467427209e-08, + "loss": 0.1098, + "step": 17110 + }, + { + "epoch": 0.89, + "learning_rate": 1.1856739118213544e-08, + "loss": 0.1284, + "step": 17111 + }, + { + "epoch": 0.89, + "learning_rate": 1.1845273146206892e-08, + "loss": 0.117, + "step": 17112 + }, + { + "epoch": 0.89, + "learning_rate": 1.1833812551734944e-08, + "loss": 0.1244, + "step": 17113 + }, + { + "epoch": 0.89, + "learning_rate": 1.1822357335125288e-08, + "loss": 0.1504, + "step": 17114 + }, + { + "epoch": 0.89, + "learning_rate": 1.1810907496705347e-08, + "loss": 0.1634, + "step": 17115 + }, + { + "epoch": 0.89, + "learning_rate": 1.179946303680226e-08, + "loss": 0.1279, + "step": 17116 + }, + { + "epoch": 0.89, + "learning_rate": 1.1788023955743209e-08, + "loss": 0.14, + "step": 17117 + }, + { + "epoch": 0.89, + "learning_rate": 1.1776590253855068e-08, + "loss": 0.1208, + "step": 17118 + }, + { + "epoch": 0.89, + "learning_rate": 1.1765161931464662e-08, + "loss": 0.1105, + "step": 17119 + }, + { + "epoch": 0.89, + "learning_rate": 1.1753738988898555e-08, + "loss": 0.1381, + "step": 17120 + }, + { + "epoch": 0.89, + "learning_rate": 1.1742321426483281e-08, + "loss": 0.1324, + "step": 17121 + }, + { + "epoch": 0.89, + "learning_rate": 1.173090924454514e-08, + "loss": 0.1078, + "step": 17122 + }, + { + "epoch": 0.89, + "learning_rate": 1.171950244341029e-08, + "loss": 0.1283, + "step": 17123 + }, + { + "epoch": 0.89, + "learning_rate": 1.170810102340476e-08, + "loss": 0.1085, + "step": 17124 + }, + { + "epoch": 0.89, + "learning_rate": 1.1696704984854378e-08, + "loss": 0.1162, + "step": 17125 + }, + { + "epoch": 0.89, + "learning_rate": 1.168531432808486e-08, + "loss": 0.1215, + "step": 17126 + }, + { + "epoch": 0.89, + "learning_rate": 1.1673929053421704e-08, + "loss": 0.1043, + "step": 17127 + }, + { + "epoch": 0.89, + "learning_rate": 1.1662549161190426e-08, + "loss": 0.1165, + "step": 17128 + }, + { + "epoch": 0.89, + "learning_rate": 1.1651174651716166e-08, + "loss": 0.1338, + "step": 17129 + }, + { + "epoch": 0.89, + "learning_rate": 1.1639805525324065e-08, + "loss": 0.1043, + "step": 17130 + }, + { + "epoch": 0.89, + "learning_rate": 1.1628441782338994e-08, + "loss": 0.1235, + "step": 17131 + }, + { + "epoch": 0.89, + "learning_rate": 1.1617083423085849e-08, + "loss": 0.1609, + "step": 17132 + }, + { + "epoch": 0.89, + "learning_rate": 1.160573044788915e-08, + "loss": 0.1088, + "step": 17133 + }, + { + "epoch": 0.89, + "learning_rate": 1.159438285707337e-08, + "loss": 0.0946, + "step": 17134 + }, + { + "epoch": 0.89, + "learning_rate": 1.1583040650962872e-08, + "loss": 0.1148, + "step": 17135 + }, + { + "epoch": 0.89, + "learning_rate": 1.1571703829881841e-08, + "loss": 0.1362, + "step": 17136 + }, + { + "epoch": 0.89, + "learning_rate": 1.1560372394154217e-08, + "loss": 0.143, + "step": 17137 + }, + { + "epoch": 0.89, + "learning_rate": 1.1549046344103897e-08, + "loss": 0.1236, + "step": 17138 + }, + { + "epoch": 0.89, + "learning_rate": 1.15377256800546e-08, + "loss": 0.1446, + "step": 17139 + }, + { + "epoch": 0.89, + "learning_rate": 1.1526410402329867e-08, + "loss": 0.1309, + "step": 17140 + }, + { + "epoch": 0.89, + "learning_rate": 1.1515100511253062e-08, + "loss": 0.168, + "step": 17141 + }, + { + "epoch": 0.89, + "learning_rate": 1.1503796007147459e-08, + "loss": 0.1573, + "step": 17142 + }, + { + "epoch": 0.89, + "learning_rate": 1.1492496890336134e-08, + "loss": 0.131, + "step": 17143 + }, + { + "epoch": 0.89, + "learning_rate": 1.1481203161142005e-08, + "loss": 0.1292, + "step": 17144 + }, + { + "epoch": 0.89, + "learning_rate": 1.1469914819887861e-08, + "loss": 0.1047, + "step": 17145 + }, + { + "epoch": 0.89, + "learning_rate": 1.1458631866896351e-08, + "loss": 0.1535, + "step": 17146 + }, + { + "epoch": 0.89, + "learning_rate": 1.1447354302489908e-08, + "loss": 0.1308, + "step": 17147 + }, + { + "epoch": 0.9, + "learning_rate": 1.1436082126990899e-08, + "loss": 0.1107, + "step": 17148 + }, + { + "epoch": 0.9, + "learning_rate": 1.1424815340721439e-08, + "loss": 0.1432, + "step": 17149 + }, + { + "epoch": 0.9, + "learning_rate": 1.141355394400354e-08, + "loss": 0.1052, + "step": 17150 + }, + { + "epoch": 0.9, + "learning_rate": 1.14022979371591e-08, + "loss": 0.12, + "step": 17151 + }, + { + "epoch": 0.9, + "learning_rate": 1.139104732050975e-08, + "loss": 0.1327, + "step": 17152 + }, + { + "epoch": 0.9, + "learning_rate": 1.13798020943771e-08, + "loss": 0.1574, + "step": 17153 + }, + { + "epoch": 0.9, + "learning_rate": 1.1368562259082515e-08, + "loss": 0.1341, + "step": 17154 + }, + { + "epoch": 0.9, + "learning_rate": 1.1357327814947249e-08, + "loss": 0.1407, + "step": 17155 + }, + { + "epoch": 0.9, + "learning_rate": 1.1346098762292356e-08, + "loss": 0.1258, + "step": 17156 + }, + { + "epoch": 0.9, + "learning_rate": 1.1334875101438823e-08, + "loss": 0.1306, + "step": 17157 + }, + { + "epoch": 0.9, + "learning_rate": 1.1323656832707396e-08, + "loss": 0.141, + "step": 17158 + }, + { + "epoch": 0.9, + "learning_rate": 1.1312443956418661e-08, + "loss": 0.1259, + "step": 17159 + }, + { + "epoch": 0.9, + "learning_rate": 1.1301236472893117e-08, + "loss": 0.1099, + "step": 17160 + }, + { + "epoch": 0.9, + "learning_rate": 1.1290034382451108e-08, + "loss": 0.1166, + "step": 17161 + }, + { + "epoch": 0.9, + "learning_rate": 1.1278837685412757e-08, + "loss": 0.1092, + "step": 17162 + }, + { + "epoch": 0.9, + "learning_rate": 1.1267646382098028e-08, + "loss": 0.1294, + "step": 17163 + }, + { + "epoch": 0.9, + "learning_rate": 1.125646047282689e-08, + "loss": 0.1326, + "step": 17164 + }, + { + "epoch": 0.9, + "learning_rate": 1.1245279957918952e-08, + "loss": 0.1544, + "step": 17165 + }, + { + "epoch": 0.9, + "learning_rate": 1.123410483769378e-08, + "loss": 0.1337, + "step": 17166 + }, + { + "epoch": 0.9, + "learning_rate": 1.1222935112470767e-08, + "loss": 0.1191, + "step": 17167 + }, + { + "epoch": 0.9, + "learning_rate": 1.1211770782569142e-08, + "loss": 0.1294, + "step": 17168 + }, + { + "epoch": 0.9, + "learning_rate": 1.1200611848307984e-08, + "loss": 0.115, + "step": 17169 + }, + { + "epoch": 0.9, + "learning_rate": 1.1189458310006239e-08, + "loss": 0.1387, + "step": 17170 + }, + { + "epoch": 0.9, + "learning_rate": 1.1178310167982607e-08, + "loss": 0.177, + "step": 17171 + }, + { + "epoch": 0.9, + "learning_rate": 1.1167167422555813e-08, + "loss": 0.1423, + "step": 17172 + }, + { + "epoch": 0.9, + "learning_rate": 1.1156030074044264e-08, + "loss": 0.1705, + "step": 17173 + }, + { + "epoch": 0.9, + "learning_rate": 1.1144898122766289e-08, + "loss": 0.1376, + "step": 17174 + }, + { + "epoch": 0.9, + "learning_rate": 1.1133771569040007e-08, + "loss": 0.1713, + "step": 17175 + }, + { + "epoch": 0.9, + "learning_rate": 1.1122650413183455e-08, + "loss": 0.137, + "step": 17176 + }, + { + "epoch": 0.9, + "learning_rate": 1.1111534655514464e-08, + "loss": 0.1304, + "step": 17177 + }, + { + "epoch": 0.9, + "learning_rate": 1.1100424296350719e-08, + "loss": 0.1193, + "step": 17178 + }, + { + "epoch": 0.9, + "learning_rate": 1.1089319336009784e-08, + "loss": 0.1587, + "step": 17179 + }, + { + "epoch": 0.9, + "learning_rate": 1.1078219774809027e-08, + "loss": 0.1375, + "step": 17180 + }, + { + "epoch": 0.9, + "learning_rate": 1.1067125613065686e-08, + "loss": 0.144, + "step": 17181 + }, + { + "epoch": 0.9, + "learning_rate": 1.1056036851096818e-08, + "loss": 0.168, + "step": 17182 + }, + { + "epoch": 0.9, + "learning_rate": 1.1044953489219366e-08, + "loss": 0.1096, + "step": 17183 + }, + { + "epoch": 0.9, + "learning_rate": 1.1033875527750103e-08, + "loss": 0.1402, + "step": 17184 + }, + { + "epoch": 0.9, + "learning_rate": 1.1022802967005596e-08, + "loss": 0.1105, + "step": 17185 + }, + { + "epoch": 0.9, + "learning_rate": 1.1011735807302347e-08, + "loss": 0.1491, + "step": 17186 + }, + { + "epoch": 0.9, + "learning_rate": 1.1000674048956637e-08, + "loss": 0.0884, + "step": 17187 + }, + { + "epoch": 0.9, + "learning_rate": 1.0989617692284637e-08, + "loss": 0.1477, + "step": 17188 + }, + { + "epoch": 0.9, + "learning_rate": 1.0978566737602313e-08, + "loss": 0.1737, + "step": 17189 + }, + { + "epoch": 0.9, + "learning_rate": 1.0967521185225526e-08, + "loss": 0.111, + "step": 17190 + }, + { + "epoch": 0.9, + "learning_rate": 1.0956481035469978e-08, + "loss": 0.1574, + "step": 17191 + }, + { + "epoch": 0.9, + "learning_rate": 1.094544628865115e-08, + "loss": 0.1354, + "step": 17192 + }, + { + "epoch": 0.9, + "learning_rate": 1.093441694508448e-08, + "loss": 0.1398, + "step": 17193 + }, + { + "epoch": 0.9, + "learning_rate": 1.0923393005085135e-08, + "loss": 0.1256, + "step": 17194 + }, + { + "epoch": 0.9, + "learning_rate": 1.0912374468968222e-08, + "loss": 0.1569, + "step": 17195 + }, + { + "epoch": 0.9, + "learning_rate": 1.090136133704862e-08, + "loss": 0.1467, + "step": 17196 + }, + { + "epoch": 0.9, + "learning_rate": 1.0890353609641123e-08, + "loss": 0.1544, + "step": 17197 + }, + { + "epoch": 0.9, + "learning_rate": 1.0879351287060345e-08, + "loss": 0.1025, + "step": 17198 + }, + { + "epoch": 0.9, + "learning_rate": 1.0868354369620702e-08, + "loss": 0.0886, + "step": 17199 + }, + { + "epoch": 0.9, + "learning_rate": 1.0857362857636477e-08, + "loss": 0.1415, + "step": 17200 + }, + { + "epoch": 0.9, + "learning_rate": 1.0846376751421882e-08, + "loss": 0.1258, + "step": 17201 + }, + { + "epoch": 0.9, + "learning_rate": 1.0835396051290846e-08, + "loss": 0.1481, + "step": 17202 + }, + { + "epoch": 0.9, + "learning_rate": 1.0824420757557184e-08, + "loss": 0.1364, + "step": 17203 + }, + { + "epoch": 0.9, + "learning_rate": 1.0813450870534646e-08, + "loss": 0.1294, + "step": 17204 + }, + { + "epoch": 0.9, + "learning_rate": 1.080248639053669e-08, + "loss": 0.1307, + "step": 17205 + }, + { + "epoch": 0.9, + "learning_rate": 1.0791527317876736e-08, + "loss": 0.1272, + "step": 17206 + }, + { + "epoch": 0.9, + "learning_rate": 1.078057365286793e-08, + "loss": 0.1261, + "step": 17207 + }, + { + "epoch": 0.9, + "learning_rate": 1.0769625395823401e-08, + "loss": 0.1456, + "step": 17208 + }, + { + "epoch": 0.9, + "learning_rate": 1.0758682547056075e-08, + "loss": 0.0964, + "step": 17209 + }, + { + "epoch": 0.9, + "learning_rate": 1.074774510687857e-08, + "loss": 0.136, + "step": 17210 + }, + { + "epoch": 0.9, + "learning_rate": 1.0736813075603612e-08, + "loss": 0.1552, + "step": 17211 + }, + { + "epoch": 0.9, + "learning_rate": 1.0725886453543598e-08, + "loss": 0.1278, + "step": 17212 + }, + { + "epoch": 0.9, + "learning_rate": 1.0714965241010831e-08, + "loss": 0.1731, + "step": 17213 + }, + { + "epoch": 0.9, + "learning_rate": 1.0704049438317397e-08, + "loss": 0.1123, + "step": 17214 + }, + { + "epoch": 0.9, + "learning_rate": 1.0693139045775313e-08, + "loss": 0.1178, + "step": 17215 + }, + { + "epoch": 0.9, + "learning_rate": 1.0682234063696416e-08, + "loss": 0.1332, + "step": 17216 + }, + { + "epoch": 0.9, + "learning_rate": 1.0671334492392348e-08, + "loss": 0.1719, + "step": 17217 + }, + { + "epoch": 0.9, + "learning_rate": 1.0660440332174613e-08, + "loss": 0.1488, + "step": 17218 + }, + { + "epoch": 0.9, + "learning_rate": 1.064955158335461e-08, + "loss": 0.1685, + "step": 17219 + }, + { + "epoch": 0.9, + "learning_rate": 1.063866824624351e-08, + "loss": 0.1448, + "step": 17220 + }, + { + "epoch": 0.9, + "learning_rate": 1.062779032115233e-08, + "loss": 0.1038, + "step": 17221 + }, + { + "epoch": 0.9, + "learning_rate": 1.0616917808392045e-08, + "loss": 0.1299, + "step": 17222 + }, + { + "epoch": 0.9, + "learning_rate": 1.0606050708273362e-08, + "loss": 0.1864, + "step": 17223 + }, + { + "epoch": 0.9, + "learning_rate": 1.0595189021106854e-08, + "loss": 0.1912, + "step": 17224 + }, + { + "epoch": 0.9, + "learning_rate": 1.058433274720294e-08, + "loss": 0.1075, + "step": 17225 + }, + { + "epoch": 0.9, + "learning_rate": 1.0573481886871948e-08, + "loss": 0.1176, + "step": 17226 + }, + { + "epoch": 0.9, + "learning_rate": 1.056263644042401e-08, + "loss": 0.1548, + "step": 17227 + }, + { + "epoch": 0.9, + "learning_rate": 1.055179640816899e-08, + "loss": 0.128, + "step": 17228 + }, + { + "epoch": 0.9, + "learning_rate": 1.054096179041677e-08, + "loss": 0.1192, + "step": 17229 + }, + { + "epoch": 0.9, + "learning_rate": 1.0530132587477015e-08, + "loss": 0.1359, + "step": 17230 + }, + { + "epoch": 0.9, + "learning_rate": 1.051930879965921e-08, + "loss": 0.1552, + "step": 17231 + }, + { + "epoch": 0.9, + "learning_rate": 1.0508490427272687e-08, + "loss": 0.1423, + "step": 17232 + }, + { + "epoch": 0.9, + "learning_rate": 1.0497677470626686e-08, + "loss": 0.127, + "step": 17233 + }, + { + "epoch": 0.9, + "learning_rate": 1.0486869930030206e-08, + "loss": 0.1269, + "step": 17234 + }, + { + "epoch": 0.9, + "learning_rate": 1.0476067805792154e-08, + "loss": 0.15, + "step": 17235 + }, + { + "epoch": 0.9, + "learning_rate": 1.046527109822124e-08, + "loss": 0.1204, + "step": 17236 + }, + { + "epoch": 0.9, + "learning_rate": 1.045447980762606e-08, + "loss": 0.1173, + "step": 17237 + }, + { + "epoch": 0.9, + "learning_rate": 1.0443693934315013e-08, + "loss": 0.1497, + "step": 17238 + }, + { + "epoch": 0.9, + "learning_rate": 1.043291347859634e-08, + "loss": 0.1313, + "step": 17239 + }, + { + "epoch": 0.9, + "learning_rate": 1.0422138440778194e-08, + "loss": 0.1755, + "step": 17240 + }, + { + "epoch": 0.9, + "learning_rate": 1.0411368821168531e-08, + "loss": 0.1358, + "step": 17241 + }, + { + "epoch": 0.9, + "learning_rate": 1.0400604620075126e-08, + "loss": 0.1305, + "step": 17242 + }, + { + "epoch": 0.9, + "learning_rate": 1.0389845837805623e-08, + "loss": 0.1288, + "step": 17243 + }, + { + "epoch": 0.9, + "learning_rate": 1.037909247466755e-08, + "loss": 0.1409, + "step": 17244 + }, + { + "epoch": 0.9, + "learning_rate": 1.036834453096822e-08, + "loss": 0.1611, + "step": 17245 + }, + { + "epoch": 0.9, + "learning_rate": 1.0357602007014766e-08, + "loss": 0.1326, + "step": 17246 + }, + { + "epoch": 0.9, + "learning_rate": 1.0346864903114272e-08, + "loss": 0.1106, + "step": 17247 + }, + { + "epoch": 0.9, + "learning_rate": 1.0336133219573606e-08, + "loss": 0.1208, + "step": 17248 + }, + { + "epoch": 0.9, + "learning_rate": 1.0325406956699478e-08, + "loss": 0.1358, + "step": 17249 + }, + { + "epoch": 0.9, + "learning_rate": 1.03146861147984e-08, + "loss": 0.1031, + "step": 17250 + }, + { + "epoch": 0.9, + "learning_rate": 1.0303970694176834e-08, + "loss": 0.1313, + "step": 17251 + }, + { + "epoch": 0.9, + "learning_rate": 1.029326069514107e-08, + "loss": 0.1485, + "step": 17252 + }, + { + "epoch": 0.9, + "learning_rate": 1.0282556117997087e-08, + "loss": 0.1312, + "step": 17253 + }, + { + "epoch": 0.9, + "learning_rate": 1.0271856963050907e-08, + "loss": 0.146, + "step": 17254 + }, + { + "epoch": 0.9, + "learning_rate": 1.0261163230608306e-08, + "loss": 0.1226, + "step": 17255 + }, + { + "epoch": 0.9, + "learning_rate": 1.0250474920974905e-08, + "loss": 0.127, + "step": 17256 + }, + { + "epoch": 0.9, + "learning_rate": 1.0239792034456151e-08, + "loss": 0.1234, + "step": 17257 + }, + { + "epoch": 0.9, + "learning_rate": 1.0229114571357422e-08, + "loss": 0.1582, + "step": 17258 + }, + { + "epoch": 0.9, + "learning_rate": 1.0218442531983851e-08, + "loss": 0.1259, + "step": 17259 + }, + { + "epoch": 0.9, + "learning_rate": 1.020777591664046e-08, + "loss": 0.1522, + "step": 17260 + }, + { + "epoch": 0.9, + "learning_rate": 1.0197114725632117e-08, + "loss": 0.1162, + "step": 17261 + }, + { + "epoch": 0.9, + "learning_rate": 1.0186458959263489e-08, + "loss": 0.0998, + "step": 17262 + }, + { + "epoch": 0.9, + "learning_rate": 1.0175808617839132e-08, + "loss": 0.1174, + "step": 17263 + }, + { + "epoch": 0.9, + "learning_rate": 1.0165163701663427e-08, + "loss": 0.0998, + "step": 17264 + }, + { + "epoch": 0.9, + "learning_rate": 1.015452421104066e-08, + "loss": 0.1544, + "step": 17265 + }, + { + "epoch": 0.9, + "learning_rate": 1.014389014627488e-08, + "loss": 0.1546, + "step": 17266 + }, + { + "epoch": 0.9, + "learning_rate": 1.0133261507669999e-08, + "loss": 0.1111, + "step": 17267 + }, + { + "epoch": 0.9, + "learning_rate": 1.0122638295529795e-08, + "loss": 0.1438, + "step": 17268 + }, + { + "epoch": 0.9, + "learning_rate": 1.0112020510157915e-08, + "loss": 0.1411, + "step": 17269 + }, + { + "epoch": 0.9, + "learning_rate": 1.0101408151857782e-08, + "loss": 0.13, + "step": 17270 + }, + { + "epoch": 0.9, + "learning_rate": 1.0090801220932687e-08, + "loss": 0.175, + "step": 17271 + }, + { + "epoch": 0.9, + "learning_rate": 1.0080199717685834e-08, + "loss": 0.1547, + "step": 17272 + }, + { + "epoch": 0.9, + "learning_rate": 1.0069603642420177e-08, + "loss": 0.1328, + "step": 17273 + }, + { + "epoch": 0.9, + "learning_rate": 1.0059012995438566e-08, + "loss": 0.1387, + "step": 17274 + }, + { + "epoch": 0.9, + "learning_rate": 1.0048427777043666e-08, + "loss": 0.1122, + "step": 17275 + }, + { + "epoch": 0.9, + "learning_rate": 1.0037847987538062e-08, + "loss": 0.1579, + "step": 17276 + }, + { + "epoch": 0.9, + "learning_rate": 1.0027273627224086e-08, + "loss": 0.1124, + "step": 17277 + }, + { + "epoch": 0.9, + "learning_rate": 1.001670469640401e-08, + "loss": 0.1538, + "step": 17278 + }, + { + "epoch": 0.9, + "learning_rate": 1.0006141195379792e-08, + "loss": 0.1218, + "step": 17279 + }, + { + "epoch": 0.9, + "learning_rate": 9.995583124453433e-09, + "loss": 0.137, + "step": 17280 + }, + { + "epoch": 0.9, + "learning_rate": 9.98503048392665e-09, + "loss": 0.1542, + "step": 17281 + }, + { + "epoch": 0.9, + "learning_rate": 9.974483274101042e-09, + "loss": 0.1463, + "step": 17282 + }, + { + "epoch": 0.9, + "learning_rate": 9.963941495278061e-09, + "loss": 0.1227, + "step": 17283 + }, + { + "epoch": 0.9, + "learning_rate": 9.953405147758998e-09, + "loss": 0.0935, + "step": 17284 + }, + { + "epoch": 0.9, + "learning_rate": 9.942874231844989e-09, + "loss": 0.1226, + "step": 17285 + }, + { + "epoch": 0.9, + "learning_rate": 9.932348747836994e-09, + "loss": 0.118, + "step": 17286 + }, + { + "epoch": 0.9, + "learning_rate": 9.921828696035839e-09, + "loss": 0.1551, + "step": 17287 + }, + { + "epoch": 0.9, + "learning_rate": 9.911314076742217e-09, + "loss": 0.1149, + "step": 17288 + }, + { + "epoch": 0.9, + "learning_rate": 9.90080489025662e-09, + "loss": 0.1153, + "step": 17289 + }, + { + "epoch": 0.9, + "learning_rate": 9.890301136879364e-09, + "loss": 0.1325, + "step": 17290 + }, + { + "epoch": 0.9, + "learning_rate": 9.87980281691072e-09, + "loss": 0.1367, + "step": 17291 + }, + { + "epoch": 0.9, + "learning_rate": 9.869309930650716e-09, + "loss": 0.1276, + "step": 17292 + }, + { + "epoch": 0.9, + "learning_rate": 9.858822478399221e-09, + "loss": 0.123, + "step": 17293 + }, + { + "epoch": 0.9, + "learning_rate": 9.848340460455973e-09, + "loss": 0.1291, + "step": 17294 + }, + { + "epoch": 0.9, + "learning_rate": 9.837863877120601e-09, + "loss": 0.118, + "step": 17295 + }, + { + "epoch": 0.9, + "learning_rate": 9.827392728692462e-09, + "loss": 0.1522, + "step": 17296 + }, + { + "epoch": 0.9, + "learning_rate": 9.816927015470833e-09, + "loss": 0.1276, + "step": 17297 + }, + { + "epoch": 0.9, + "learning_rate": 9.806466737754871e-09, + "loss": 0.1591, + "step": 17298 + }, + { + "epoch": 0.9, + "learning_rate": 9.796011895843493e-09, + "loss": 0.168, + "step": 17299 + }, + { + "epoch": 0.9, + "learning_rate": 9.785562490035548e-09, + "loss": 0.1388, + "step": 17300 + }, + { + "epoch": 0.9, + "learning_rate": 9.775118520629621e-09, + "loss": 0.1181, + "step": 17301 + }, + { + "epoch": 0.9, + "learning_rate": 9.76467998792425e-09, + "loss": 0.1405, + "step": 17302 + }, + { + "epoch": 0.9, + "learning_rate": 9.754246892217799e-09, + "loss": 0.126, + "step": 17303 + }, + { + "epoch": 0.9, + "learning_rate": 9.743819233808382e-09, + "loss": 0.1246, + "step": 17304 + }, + { + "epoch": 0.9, + "learning_rate": 9.73339701299407e-09, + "loss": 0.1333, + "step": 17305 + }, + { + "epoch": 0.9, + "learning_rate": 9.72298023007272e-09, + "loss": 0.1199, + "step": 17306 + }, + { + "epoch": 0.9, + "learning_rate": 9.712568885342066e-09, + "loss": 0.1048, + "step": 17307 + }, + { + "epoch": 0.9, + "learning_rate": 9.702162979099605e-09, + "loss": 0.1523, + "step": 17308 + }, + { + "epoch": 0.9, + "learning_rate": 9.691762511642809e-09, + "loss": 0.1478, + "step": 17309 + }, + { + "epoch": 0.9, + "learning_rate": 9.681367483268933e-09, + "loss": 0.1121, + "step": 17310 + }, + { + "epoch": 0.9, + "learning_rate": 9.670977894275022e-09, + "loss": 0.1208, + "step": 17311 + }, + { + "epoch": 0.9, + "learning_rate": 9.660593744958045e-09, + "loss": 0.1376, + "step": 17312 + }, + { + "epoch": 0.9, + "learning_rate": 9.650215035614783e-09, + "loss": 0.1301, + "step": 17313 + }, + { + "epoch": 0.9, + "learning_rate": 9.639841766541846e-09, + "loss": 0.1103, + "step": 17314 + }, + { + "epoch": 0.9, + "learning_rate": 9.629473938035704e-09, + "loss": 0.1354, + "step": 17315 + }, + { + "epoch": 0.9, + "learning_rate": 9.6191115503927e-09, + "loss": 0.1465, + "step": 17316 + }, + { + "epoch": 0.9, + "learning_rate": 9.608754603908997e-09, + "loss": 0.1766, + "step": 17317 + }, + { + "epoch": 0.9, + "learning_rate": 9.59840309888058e-09, + "loss": 0.1134, + "step": 17318 + }, + { + "epoch": 0.9, + "learning_rate": 9.58805703560326e-09, + "loss": 0.1651, + "step": 17319 + }, + { + "epoch": 0.9, + "learning_rate": 9.577716414372838e-09, + "loss": 0.1317, + "step": 17320 + }, + { + "epoch": 0.9, + "learning_rate": 9.567381235484796e-09, + "loss": 0.1591, + "step": 17321 + }, + { + "epoch": 0.9, + "learning_rate": 9.557051499234448e-09, + "loss": 0.1456, + "step": 17322 + }, + { + "epoch": 0.9, + "learning_rate": 9.546727205917137e-09, + "loss": 0.1172, + "step": 17323 + }, + { + "epoch": 0.9, + "learning_rate": 9.536408355827873e-09, + "loss": 0.121, + "step": 17324 + }, + { + "epoch": 0.9, + "learning_rate": 9.526094949261599e-09, + "loss": 0.1479, + "step": 17325 + }, + { + "epoch": 0.9, + "learning_rate": 9.515786986513008e-09, + "loss": 0.116, + "step": 17326 + }, + { + "epoch": 0.9, + "learning_rate": 9.5054844678768e-09, + "loss": 0.1074, + "step": 17327 + }, + { + "epoch": 0.9, + "learning_rate": 9.495187393647385e-09, + "loss": 0.1454, + "step": 17328 + }, + { + "epoch": 0.9, + "learning_rate": 9.48489576411906e-09, + "loss": 0.1169, + "step": 17329 + }, + { + "epoch": 0.9, + "learning_rate": 9.474609579585968e-09, + "loss": 0.146, + "step": 17330 + }, + { + "epoch": 0.9, + "learning_rate": 9.464328840342073e-09, + "loss": 0.1605, + "step": 17331 + }, + { + "epoch": 0.9, + "learning_rate": 9.454053546681207e-09, + "loss": 0.1387, + "step": 17332 + }, + { + "epoch": 0.9, + "learning_rate": 9.443783698897023e-09, + "loss": 0.1344, + "step": 17333 + }, + { + "epoch": 0.9, + "learning_rate": 9.43351929728311e-09, + "loss": 0.1387, + "step": 17334 + }, + { + "epoch": 0.9, + "learning_rate": 9.423260342132767e-09, + "loss": 0.1465, + "step": 17335 + }, + { + "epoch": 0.9, + "learning_rate": 9.413006833739202e-09, + "loss": 0.1421, + "step": 17336 + }, + { + "epoch": 0.9, + "learning_rate": 9.40275877239547e-09, + "loss": 0.1641, + "step": 17337 + }, + { + "epoch": 0.9, + "learning_rate": 9.392516158394514e-09, + "loss": 0.1313, + "step": 17338 + }, + { + "epoch": 0.91, + "learning_rate": 9.382278992028992e-09, + "loss": 0.1828, + "step": 17339 + }, + { + "epoch": 0.91, + "learning_rate": 9.372047273591511e-09, + "loss": 0.1256, + "step": 17340 + }, + { + "epoch": 0.91, + "learning_rate": 9.361821003374525e-09, + "loss": 0.1151, + "step": 17341 + }, + { + "epoch": 0.91, + "learning_rate": 9.351600181670295e-09, + "loss": 0.1461, + "step": 17342 + }, + { + "epoch": 0.91, + "learning_rate": 9.341384808770936e-09, + "loss": 0.1116, + "step": 17343 + }, + { + "epoch": 0.91, + "learning_rate": 9.331174884968352e-09, + "loss": 0.1338, + "step": 17344 + }, + { + "epoch": 0.91, + "learning_rate": 9.320970410554441e-09, + "loss": 0.1533, + "step": 17345 + }, + { + "epoch": 0.91, + "learning_rate": 9.310771385820837e-09, + "loss": 0.1441, + "step": 17346 + }, + { + "epoch": 0.91, + "learning_rate": 9.30057781105893e-09, + "loss": 0.1373, + "step": 17347 + }, + { + "epoch": 0.91, + "learning_rate": 9.290389686560152e-09, + "loss": 0.1326, + "step": 17348 + }, + { + "epoch": 0.91, + "learning_rate": 9.28020701261567e-09, + "loss": 0.1138, + "step": 17349 + }, + { + "epoch": 0.91, + "learning_rate": 9.270029789516497e-09, + "loss": 0.1271, + "step": 17350 + }, + { + "epoch": 0.91, + "learning_rate": 9.259858017553468e-09, + "loss": 0.1443, + "step": 17351 + }, + { + "epoch": 0.91, + "learning_rate": 9.24969169701737e-09, + "loss": 0.1124, + "step": 17352 + }, + { + "epoch": 0.91, + "learning_rate": 9.239530828198704e-09, + "loss": 0.1308, + "step": 17353 + }, + { + "epoch": 0.91, + "learning_rate": 9.229375411387885e-09, + "loss": 0.135, + "step": 17354 + }, + { + "epoch": 0.91, + "learning_rate": 9.219225446875168e-09, + "loss": 0.16, + "step": 17355 + }, + { + "epoch": 0.91, + "learning_rate": 9.209080934950653e-09, + "loss": 0.1416, + "step": 17356 + }, + { + "epoch": 0.91, + "learning_rate": 9.198941875904242e-09, + "loss": 0.111, + "step": 17357 + }, + { + "epoch": 0.91, + "learning_rate": 9.188808270025705e-09, + "loss": 0.1864, + "step": 17358 + }, + { + "epoch": 0.91, + "learning_rate": 9.178680117604698e-09, + "loss": 0.1348, + "step": 17359 + }, + { + "epoch": 0.91, + "learning_rate": 9.1685574189307e-09, + "loss": 0.1185, + "step": 17360 + }, + { + "epoch": 0.91, + "learning_rate": 9.158440174292991e-09, + "loss": 0.1194, + "step": 17361 + }, + { + "epoch": 0.91, + "learning_rate": 9.148328383980718e-09, + "loss": 0.1491, + "step": 17362 + }, + { + "epoch": 0.91, + "learning_rate": 9.138222048282895e-09, + "loss": 0.1329, + "step": 17363 + }, + { + "epoch": 0.91, + "learning_rate": 9.128121167488423e-09, + "loss": 0.1355, + "step": 17364 + }, + { + "epoch": 0.91, + "learning_rate": 9.118025741885871e-09, + "loss": 0.1646, + "step": 17365 + }, + { + "epoch": 0.91, + "learning_rate": 9.107935771763853e-09, + "loss": 0.169, + "step": 17366 + }, + { + "epoch": 0.91, + "learning_rate": 9.097851257410716e-09, + "loss": 0.1087, + "step": 17367 + }, + { + "epoch": 0.91, + "learning_rate": 9.087772199114695e-09, + "loss": 0.1371, + "step": 17368 + }, + { + "epoch": 0.91, + "learning_rate": 9.077698597163808e-09, + "loss": 0.1563, + "step": 17369 + }, + { + "epoch": 0.91, + "learning_rate": 9.067630451846041e-09, + "loss": 0.1522, + "step": 17370 + }, + { + "epoch": 0.91, + "learning_rate": 9.057567763449081e-09, + "loss": 0.1345, + "step": 17371 + }, + { + "epoch": 0.91, + "learning_rate": 9.04751053226056e-09, + "loss": 0.1275, + "step": 17372 + }, + { + "epoch": 0.91, + "learning_rate": 9.037458758567917e-09, + "loss": 0.128, + "step": 17373 + }, + { + "epoch": 0.91, + "learning_rate": 9.027412442658389e-09, + "loss": 0.1218, + "step": 17374 + }, + { + "epoch": 0.91, + "learning_rate": 9.017371584819166e-09, + "loss": 0.1306, + "step": 17375 + }, + { + "epoch": 0.91, + "learning_rate": 9.007336185337155e-09, + "loss": 0.1096, + "step": 17376 + }, + { + "epoch": 0.91, + "learning_rate": 8.997306244499214e-09, + "loss": 0.1147, + "step": 17377 + }, + { + "epoch": 0.91, + "learning_rate": 8.987281762592024e-09, + "loss": 0.1072, + "step": 17378 + }, + { + "epoch": 0.91, + "learning_rate": 8.977262739902047e-09, + "loss": 0.175, + "step": 17379 + }, + { + "epoch": 0.91, + "learning_rate": 8.967249176715607e-09, + "loss": 0.1313, + "step": 17380 + }, + { + "epoch": 0.91, + "learning_rate": 8.957241073318988e-09, + "loss": 0.1308, + "step": 17381 + }, + { + "epoch": 0.91, + "learning_rate": 8.947238429998139e-09, + "loss": 0.1467, + "step": 17382 + }, + { + "epoch": 0.91, + "learning_rate": 8.937241247038962e-09, + "loss": 0.1422, + "step": 17383 + }, + { + "epoch": 0.91, + "learning_rate": 8.927249524727209e-09, + "loss": 0.1099, + "step": 17384 + }, + { + "epoch": 0.91, + "learning_rate": 8.917263263348407e-09, + "loss": 0.1392, + "step": 17385 + }, + { + "epoch": 0.91, + "learning_rate": 8.907282463188015e-09, + "loss": 0.1379, + "step": 17386 + }, + { + "epoch": 0.91, + "learning_rate": 8.897307124531206e-09, + "loss": 0.12, + "step": 17387 + }, + { + "epoch": 0.91, + "learning_rate": 8.887337247663173e-09, + "loss": 0.1307, + "step": 17388 + }, + { + "epoch": 0.91, + "learning_rate": 8.877372832868846e-09, + "loss": 0.1174, + "step": 17389 + }, + { + "epoch": 0.91, + "learning_rate": 8.86741388043295e-09, + "loss": 0.1294, + "step": 17390 + }, + { + "epoch": 0.91, + "learning_rate": 8.857460390640126e-09, + "loss": 0.1016, + "step": 17391 + }, + { + "epoch": 0.91, + "learning_rate": 8.8475123637749e-09, + "loss": 0.1191, + "step": 17392 + }, + { + "epoch": 0.91, + "learning_rate": 8.837569800121559e-09, + "loss": 0.1181, + "step": 17393 + }, + { + "epoch": 0.91, + "learning_rate": 8.827632699964271e-09, + "loss": 0.1349, + "step": 17394 + }, + { + "epoch": 0.91, + "learning_rate": 8.817701063587014e-09, + "loss": 0.1598, + "step": 17395 + }, + { + "epoch": 0.91, + "learning_rate": 8.807774891273712e-09, + "loss": 0.1242, + "step": 17396 + }, + { + "epoch": 0.91, + "learning_rate": 8.797854183307985e-09, + "loss": 0.1295, + "step": 17397 + }, + { + "epoch": 0.91, + "learning_rate": 8.787938939973405e-09, + "loss": 0.1271, + "step": 17398 + }, + { + "epoch": 0.91, + "learning_rate": 8.778029161553347e-09, + "loss": 0.1904, + "step": 17399 + }, + { + "epoch": 0.91, + "learning_rate": 8.768124848331026e-09, + "loss": 0.1251, + "step": 17400 + }, + { + "epoch": 0.91, + "learning_rate": 8.75822600058953e-09, + "loss": 0.1372, + "step": 17401 + }, + { + "epoch": 0.91, + "learning_rate": 8.74833261861172e-09, + "loss": 0.1323, + "step": 17402 + }, + { + "epoch": 0.91, + "learning_rate": 8.738444702680437e-09, + "loss": 0.1256, + "step": 17403 + }, + { + "epoch": 0.91, + "learning_rate": 8.72856225307823e-09, + "loss": 0.1029, + "step": 17404 + }, + { + "epoch": 0.91, + "learning_rate": 8.718685270087567e-09, + "loss": 0.1434, + "step": 17405 + }, + { + "epoch": 0.91, + "learning_rate": 8.708813753990706e-09, + "loss": 0.1525, + "step": 17406 + }, + { + "epoch": 0.91, + "learning_rate": 8.698947705069804e-09, + "loss": 0.1337, + "step": 17407 + }, + { + "epoch": 0.91, + "learning_rate": 8.689087123606809e-09, + "loss": 0.1302, + "step": 17408 + }, + { + "epoch": 0.91, + "learning_rate": 8.679232009883542e-09, + "loss": 0.1206, + "step": 17409 + }, + { + "epoch": 0.91, + "learning_rate": 8.66938236418171e-09, + "loss": 0.1423, + "step": 17410 + }, + { + "epoch": 0.91, + "learning_rate": 8.659538186782778e-09, + "loss": 0.1288, + "step": 17411 + }, + { + "epoch": 0.91, + "learning_rate": 8.64969947796812e-09, + "loss": 0.1405, + "step": 17412 + }, + { + "epoch": 0.91, + "learning_rate": 8.639866238018912e-09, + "loss": 0.1299, + "step": 17413 + }, + { + "epoch": 0.91, + "learning_rate": 8.630038467216194e-09, + "loss": 0.1291, + "step": 17414 + }, + { + "epoch": 0.91, + "learning_rate": 8.620216165840898e-09, + "loss": 0.2186, + "step": 17415 + }, + { + "epoch": 0.91, + "learning_rate": 8.610399334173646e-09, + "loss": 0.1304, + "step": 17416 + }, + { + "epoch": 0.91, + "learning_rate": 8.600587972495076e-09, + "loss": 0.1473, + "step": 17417 + }, + { + "epoch": 0.91, + "learning_rate": 8.590782081085613e-09, + "loss": 0.1352, + "step": 17418 + }, + { + "epoch": 0.91, + "learning_rate": 8.580981660225472e-09, + "loss": 0.1521, + "step": 17419 + }, + { + "epoch": 0.91, + "learning_rate": 8.571186710194767e-09, + "loss": 0.1424, + "step": 17420 + }, + { + "epoch": 0.91, + "learning_rate": 8.561397231273448e-09, + "loss": 0.1418, + "step": 17421 + }, + { + "epoch": 0.91, + "learning_rate": 8.551613223741316e-09, + "loss": 0.1507, + "step": 17422 + }, + { + "epoch": 0.91, + "learning_rate": 8.541834687877969e-09, + "loss": 0.1384, + "step": 17423 + }, + { + "epoch": 0.91, + "learning_rate": 8.532061623962916e-09, + "loss": 0.1085, + "step": 17424 + }, + { + "epoch": 0.91, + "learning_rate": 8.522294032275424e-09, + "loss": 0.1312, + "step": 17425 + }, + { + "epoch": 0.91, + "learning_rate": 8.512531913094712e-09, + "loss": 0.1301, + "step": 17426 + }, + { + "epoch": 0.91, + "learning_rate": 8.502775266699735e-09, + "loss": 0.1095, + "step": 17427 + }, + { + "epoch": 0.91, + "learning_rate": 8.493024093369383e-09, + "loss": 0.1373, + "step": 17428 + }, + { + "epoch": 0.91, + "learning_rate": 8.483278393382343e-09, + "loss": 0.1209, + "step": 17429 + }, + { + "epoch": 0.91, + "learning_rate": 8.473538167017125e-09, + "loss": 0.1326, + "step": 17430 + }, + { + "epoch": 0.91, + "learning_rate": 8.463803414552106e-09, + "loss": 0.1117, + "step": 17431 + }, + { + "epoch": 0.91, + "learning_rate": 8.454074136265555e-09, + "loss": 0.118, + "step": 17432 + }, + { + "epoch": 0.91, + "learning_rate": 8.444350332435512e-09, + "loss": 0.1105, + "step": 17433 + }, + { + "epoch": 0.91, + "learning_rate": 8.434632003339849e-09, + "loss": 0.098, + "step": 17434 + }, + { + "epoch": 0.91, + "learning_rate": 8.424919149256382e-09, + "loss": 0.137, + "step": 17435 + }, + { + "epoch": 0.91, + "learning_rate": 8.415211770462672e-09, + "loss": 0.1377, + "step": 17436 + }, + { + "epoch": 0.91, + "learning_rate": 8.405509867236182e-09, + "loss": 0.1336, + "step": 17437 + }, + { + "epoch": 0.91, + "learning_rate": 8.395813439854137e-09, + "loss": 0.1339, + "step": 17438 + }, + { + "epoch": 0.91, + "learning_rate": 8.38612248859376e-09, + "loss": 0.1486, + "step": 17439 + }, + { + "epoch": 0.91, + "learning_rate": 8.376437013731963e-09, + "loss": 0.1425, + "step": 17440 + }, + { + "epoch": 0.91, + "learning_rate": 8.366757015545568e-09, + "loss": 0.1181, + "step": 17441 + }, + { + "epoch": 0.91, + "learning_rate": 8.35708249431124e-09, + "loss": 0.114, + "step": 17442 + }, + { + "epoch": 0.91, + "learning_rate": 8.347413450305473e-09, + "loss": 0.131, + "step": 17443 + }, + { + "epoch": 0.91, + "learning_rate": 8.337749883804624e-09, + "loss": 0.1063, + "step": 17444 + }, + { + "epoch": 0.91, + "learning_rate": 8.328091795084825e-09, + "loss": 0.1351, + "step": 17445 + }, + { + "epoch": 0.91, + "learning_rate": 8.31843918442221e-09, + "loss": 0.1394, + "step": 17446 + }, + { + "epoch": 0.91, + "learning_rate": 8.308792052092584e-09, + "loss": 0.1326, + "step": 17447 + }, + { + "epoch": 0.91, + "learning_rate": 8.299150398371679e-09, + "loss": 0.1859, + "step": 17448 + }, + { + "epoch": 0.91, + "learning_rate": 8.289514223535054e-09, + "loss": 0.1353, + "step": 17449 + }, + { + "epoch": 0.91, + "learning_rate": 8.279883527858134e-09, + "loss": 0.1624, + "step": 17450 + }, + { + "epoch": 0.91, + "learning_rate": 8.27025831161614e-09, + "loss": 0.1557, + "step": 17451 + }, + { + "epoch": 0.91, + "learning_rate": 8.260638575084166e-09, + "loss": 0.1203, + "step": 17452 + }, + { + "epoch": 0.91, + "learning_rate": 8.25102431853717e-09, + "loss": 0.1272, + "step": 17453 + }, + { + "epoch": 0.91, + "learning_rate": 8.241415542249951e-09, + "loss": 0.1406, + "step": 17454 + }, + { + "epoch": 0.91, + "learning_rate": 8.231812246497071e-09, + "loss": 0.1244, + "step": 17455 + }, + { + "epoch": 0.91, + "learning_rate": 8.22221443155302e-09, + "loss": 0.1178, + "step": 17456 + }, + { + "epoch": 0.91, + "learning_rate": 8.212622097692134e-09, + "loss": 0.1222, + "step": 17457 + }, + { + "epoch": 0.91, + "learning_rate": 8.203035245188573e-09, + "loss": 0.1202, + "step": 17458 + }, + { + "epoch": 0.91, + "learning_rate": 8.193453874316247e-09, + "loss": 0.1399, + "step": 17459 + }, + { + "epoch": 0.91, + "learning_rate": 8.183877985349098e-09, + "loss": 0.1261, + "step": 17460 + }, + { + "epoch": 0.91, + "learning_rate": 8.174307578560745e-09, + "loss": 0.1205, + "step": 17461 + }, + { + "epoch": 0.91, + "learning_rate": 8.164742654224733e-09, + "loss": 0.1214, + "step": 17462 + }, + { + "epoch": 0.91, + "learning_rate": 8.155183212614413e-09, + "loss": 0.1176, + "step": 17463 + }, + { + "epoch": 0.91, + "learning_rate": 8.145629254003039e-09, + "loss": 0.1087, + "step": 17464 + }, + { + "epoch": 0.91, + "learning_rate": 8.136080778663657e-09, + "loss": 0.1525, + "step": 17465 + }, + { + "epoch": 0.91, + "learning_rate": 8.126537786869137e-09, + "loss": 0.1193, + "step": 17466 + }, + { + "epoch": 0.91, + "learning_rate": 8.11700027889226e-09, + "loss": 0.114, + "step": 17467 + }, + { + "epoch": 0.91, + "learning_rate": 8.107468255005567e-09, + "loss": 0.1596, + "step": 17468 + }, + { + "epoch": 0.91, + "learning_rate": 8.097941715481504e-09, + "loss": 0.1494, + "step": 17469 + }, + { + "epoch": 0.91, + "learning_rate": 8.088420660592343e-09, + "loss": 0.1249, + "step": 17470 + }, + { + "epoch": 0.91, + "learning_rate": 8.078905090610221e-09, + "loss": 0.1404, + "step": 17471 + }, + { + "epoch": 0.91, + "learning_rate": 8.069395005807079e-09, + "loss": 0.1229, + "step": 17472 + }, + { + "epoch": 0.91, + "learning_rate": 8.059890406454739e-09, + "loss": 0.1178, + "step": 17473 + }, + { + "epoch": 0.91, + "learning_rate": 8.050391292824787e-09, + "loss": 0.1368, + "step": 17474 + }, + { + "epoch": 0.91, + "learning_rate": 8.040897665188783e-09, + "loss": 0.1147, + "step": 17475 + }, + { + "epoch": 0.91, + "learning_rate": 8.031409523818023e-09, + "loss": 0.1635, + "step": 17476 + }, + { + "epoch": 0.91, + "learning_rate": 8.021926868983664e-09, + "loss": 0.143, + "step": 17477 + }, + { + "epoch": 0.91, + "learning_rate": 8.01244970095678e-09, + "loss": 0.1175, + "step": 17478 + }, + { + "epoch": 0.91, + "learning_rate": 8.002978020008177e-09, + "loss": 0.1404, + "step": 17479 + }, + { + "epoch": 0.91, + "learning_rate": 7.993511826408595e-09, + "loss": 0.1278, + "step": 17480 + }, + { + "epoch": 0.91, + "learning_rate": 7.984051120428526e-09, + "loss": 0.1226, + "step": 17481 + }, + { + "epoch": 0.91, + "learning_rate": 7.974595902338443e-09, + "loss": 0.116, + "step": 17482 + }, + { + "epoch": 0.91, + "learning_rate": 7.965146172408533e-09, + "loss": 0.126, + "step": 17483 + }, + { + "epoch": 0.91, + "learning_rate": 7.955701930908865e-09, + "loss": 0.1279, + "step": 17484 + }, + { + "epoch": 0.91, + "learning_rate": 7.94626317810938e-09, + "loss": 0.1641, + "step": 17485 + }, + { + "epoch": 0.91, + "learning_rate": 7.936829914279841e-09, + "loss": 0.1284, + "step": 17486 + }, + { + "epoch": 0.91, + "learning_rate": 7.927402139689854e-09, + "loss": 0.1545, + "step": 17487 + }, + { + "epoch": 0.91, + "learning_rate": 7.917979854608825e-09, + "loss": 0.1203, + "step": 17488 + }, + { + "epoch": 0.91, + "learning_rate": 7.908563059306095e-09, + "loss": 0.1637, + "step": 17489 + }, + { + "epoch": 0.91, + "learning_rate": 7.899151754050804e-09, + "loss": 0.1271, + "step": 17490 + }, + { + "epoch": 0.91, + "learning_rate": 7.889745939111913e-09, + "loss": 0.1252, + "step": 17491 + }, + { + "epoch": 0.91, + "learning_rate": 7.880345614758233e-09, + "loss": 0.1074, + "step": 17492 + }, + { + "epoch": 0.91, + "learning_rate": 7.870950781258457e-09, + "loss": 0.1409, + "step": 17493 + }, + { + "epoch": 0.91, + "learning_rate": 7.86156143888108e-09, + "loss": 0.1119, + "step": 17494 + }, + { + "epoch": 0.91, + "learning_rate": 7.852177587894404e-09, + "loss": 0.1342, + "step": 17495 + }, + { + "epoch": 0.91, + "learning_rate": 7.84279922856672e-09, + "loss": 0.1202, + "step": 17496 + }, + { + "epoch": 0.91, + "learning_rate": 7.833426361165996e-09, + "loss": 0.1203, + "step": 17497 + }, + { + "epoch": 0.91, + "learning_rate": 7.824058985960125e-09, + "loss": 0.1257, + "step": 17498 + }, + { + "epoch": 0.91, + "learning_rate": 7.814697103216828e-09, + "loss": 0.1171, + "step": 17499 + }, + { + "epoch": 0.91, + "learning_rate": 7.805340713203689e-09, + "loss": 0.1376, + "step": 17500 + }, + { + "epoch": 0.91, + "learning_rate": 7.795989816188142e-09, + "loss": 0.1272, + "step": 17501 + }, + { + "epoch": 0.91, + "learning_rate": 7.786644412437372e-09, + "loss": 0.1346, + "step": 17502 + }, + { + "epoch": 0.91, + "learning_rate": 7.777304502218495e-09, + "loss": 0.1719, + "step": 17503 + }, + { + "epoch": 0.91, + "learning_rate": 7.767970085798481e-09, + "loss": 0.1496, + "step": 17504 + }, + { + "epoch": 0.91, + "learning_rate": 7.758641163444068e-09, + "loss": 0.1388, + "step": 17505 + }, + { + "epoch": 0.91, + "learning_rate": 7.749317735421934e-09, + "loss": 0.1175, + "step": 17506 + }, + { + "epoch": 0.91, + "learning_rate": 7.739999801998487e-09, + "loss": 0.142, + "step": 17507 + }, + { + "epoch": 0.91, + "learning_rate": 7.730687363440069e-09, + "loss": 0.1595, + "step": 17508 + }, + { + "epoch": 0.91, + "learning_rate": 7.721380420012868e-09, + "loss": 0.1714, + "step": 17509 + }, + { + "epoch": 0.91, + "learning_rate": 7.712078971982782e-09, + "loss": 0.1307, + "step": 17510 + }, + { + "epoch": 0.91, + "learning_rate": 7.702783019615754e-09, + "loss": 0.1167, + "step": 17511 + }, + { + "epoch": 0.91, + "learning_rate": 7.693492563177418e-09, + "loss": 0.1519, + "step": 17512 + }, + { + "epoch": 0.91, + "learning_rate": 7.68420760293329e-09, + "loss": 0.1679, + "step": 17513 + }, + { + "epoch": 0.91, + "learning_rate": 7.674928139148717e-09, + "loss": 0.1471, + "step": 17514 + }, + { + "epoch": 0.91, + "learning_rate": 7.665654172088998e-09, + "loss": 0.1171, + "step": 17515 + }, + { + "epoch": 0.91, + "learning_rate": 7.656385702019096e-09, + "loss": 0.1047, + "step": 17516 + }, + { + "epoch": 0.91, + "learning_rate": 7.64712272920398e-09, + "loss": 0.1193, + "step": 17517 + }, + { + "epoch": 0.91, + "learning_rate": 7.637865253908327e-09, + "loss": 0.1063, + "step": 17518 + }, + { + "epoch": 0.91, + "learning_rate": 7.628613276396745e-09, + "loss": 0.1509, + "step": 17519 + }, + { + "epoch": 0.91, + "learning_rate": 7.619366796933668e-09, + "loss": 0.1608, + "step": 17520 + }, + { + "epoch": 0.91, + "learning_rate": 7.610125815783331e-09, + "loss": 0.1544, + "step": 17521 + }, + { + "epoch": 0.91, + "learning_rate": 7.600890333209875e-09, + "loss": 0.1238, + "step": 17522 + }, + { + "epoch": 0.91, + "learning_rate": 7.59166034947727e-09, + "loss": 0.1254, + "step": 17523 + }, + { + "epoch": 0.91, + "learning_rate": 7.582435864849279e-09, + "loss": 0.1255, + "step": 17524 + }, + { + "epoch": 0.91, + "learning_rate": 7.573216879589518e-09, + "loss": 0.1516, + "step": 17525 + }, + { + "epoch": 0.91, + "learning_rate": 7.564003393961549e-09, + "loss": 0.1102, + "step": 17526 + }, + { + "epoch": 0.91, + "learning_rate": 7.554795408228654e-09, + "loss": 0.1467, + "step": 17527 + }, + { + "epoch": 0.91, + "learning_rate": 7.545592922653954e-09, + "loss": 0.1113, + "step": 17528 + }, + { + "epoch": 0.91, + "learning_rate": 7.536395937500528e-09, + "loss": 0.162, + "step": 17529 + }, + { + "epoch": 0.91, + "learning_rate": 7.52720445303121e-09, + "loss": 0.1591, + "step": 17530 + }, + { + "epoch": 0.92, + "learning_rate": 7.518018469508681e-09, + "loss": 0.1378, + "step": 17531 + }, + { + "epoch": 0.92, + "learning_rate": 7.508837987195482e-09, + "loss": 0.1222, + "step": 17532 + }, + { + "epoch": 0.92, + "learning_rate": 7.499663006354007e-09, + "loss": 0.119, + "step": 17533 + }, + { + "epoch": 0.92, + "learning_rate": 7.490493527246489e-09, + "loss": 0.0996, + "step": 17534 + }, + { + "epoch": 0.92, + "learning_rate": 7.481329550134962e-09, + "loss": 0.1001, + "step": 17535 + }, + { + "epoch": 0.92, + "learning_rate": 7.472171075281376e-09, + "loss": 0.1296, + "step": 17536 + }, + { + "epoch": 0.92, + "learning_rate": 7.46301810294747e-09, + "loss": 0.1372, + "step": 17537 + }, + { + "epoch": 0.92, + "learning_rate": 7.45387063339482e-09, + "loss": 0.1165, + "step": 17538 + }, + { + "epoch": 0.92, + "learning_rate": 7.444728666884858e-09, + "loss": 0.1313, + "step": 17539 + }, + { + "epoch": 0.92, + "learning_rate": 7.435592203678908e-09, + "loss": 0.1554, + "step": 17540 + }, + { + "epoch": 0.92, + "learning_rate": 7.426461244038074e-09, + "loss": 0.1347, + "step": 17541 + }, + { + "epoch": 0.92, + "learning_rate": 7.4173357882233004e-09, + "loss": 0.1743, + "step": 17542 + }, + { + "epoch": 0.92, + "learning_rate": 7.4082158364954016e-09, + "loss": 0.1384, + "step": 17543 + }, + { + "epoch": 0.92, + "learning_rate": 7.399101389115104e-09, + "loss": 0.151, + "step": 17544 + }, + { + "epoch": 0.92, + "learning_rate": 7.389992446342797e-09, + "loss": 0.1567, + "step": 17545 + }, + { + "epoch": 0.92, + "learning_rate": 7.3808890084388505e-09, + "loss": 0.1072, + "step": 17546 + }, + { + "epoch": 0.92, + "learning_rate": 7.3717910756634805e-09, + "loss": 0.1563, + "step": 17547 + }, + { + "epoch": 0.92, + "learning_rate": 7.362698648276677e-09, + "loss": 0.1613, + "step": 17548 + }, + { + "epoch": 0.92, + "learning_rate": 7.3536117265383e-09, + "loss": 0.1192, + "step": 17549 + }, + { + "epoch": 0.92, + "learning_rate": 7.344530310708052e-09, + "loss": 0.1556, + "step": 17550 + }, + { + "epoch": 0.92, + "learning_rate": 7.335454401045527e-09, + "loss": 0.1278, + "step": 17551 + }, + { + "epoch": 0.92, + "learning_rate": 7.326383997810115e-09, + "loss": 0.1488, + "step": 17552 + }, + { + "epoch": 0.92, + "learning_rate": 7.317319101260988e-09, + "loss": 0.1242, + "step": 17553 + }, + { + "epoch": 0.92, + "learning_rate": 7.308259711657272e-09, + "loss": 0.1419, + "step": 17554 + }, + { + "epoch": 0.92, + "learning_rate": 7.299205829257893e-09, + "loss": 0.114, + "step": 17555 + }, + { + "epoch": 0.92, + "learning_rate": 7.290157454321599e-09, + "loss": 0.1322, + "step": 17556 + }, + { + "epoch": 0.92, + "learning_rate": 7.2811145871069626e-09, + "loss": 0.1095, + "step": 17557 + }, + { + "epoch": 0.92, + "learning_rate": 7.272077227872509e-09, + "loss": 0.1351, + "step": 17558 + }, + { + "epoch": 0.92, + "learning_rate": 7.263045376876475e-09, + "loss": 0.1733, + "step": 17559 + }, + { + "epoch": 0.92, + "learning_rate": 7.254019034377013e-09, + "loss": 0.1132, + "step": 17560 + }, + { + "epoch": 0.92, + "learning_rate": 7.244998200632113e-09, + "loss": 0.1311, + "step": 17561 + }, + { + "epoch": 0.92, + "learning_rate": 7.235982875899549e-09, + "loss": 0.1532, + "step": 17562 + }, + { + "epoch": 0.92, + "learning_rate": 7.226973060437025e-09, + "loss": 0.132, + "step": 17563 + }, + { + "epoch": 0.92, + "learning_rate": 7.217968754502002e-09, + "loss": 0.12, + "step": 17564 + }, + { + "epoch": 0.92, + "learning_rate": 7.208969958351874e-09, + "loss": 0.1247, + "step": 17565 + }, + { + "epoch": 0.92, + "learning_rate": 7.199976672243791e-09, + "loss": 0.1214, + "step": 17566 + }, + { + "epoch": 0.92, + "learning_rate": 7.190988896434835e-09, + "loss": 0.1261, + "step": 17567 + }, + { + "epoch": 0.92, + "learning_rate": 7.182006631181803e-09, + "loss": 0.1201, + "step": 17568 + }, + { + "epoch": 0.92, + "learning_rate": 7.17302987674151e-09, + "loss": 0.1266, + "step": 17569 + }, + { + "epoch": 0.92, + "learning_rate": 7.16405863337044e-09, + "loss": 0.143, + "step": 17570 + }, + { + "epoch": 0.92, + "learning_rate": 7.155092901324989e-09, + "loss": 0.1269, + "step": 17571 + }, + { + "epoch": 0.92, + "learning_rate": 7.146132680861461e-09, + "loss": 0.1497, + "step": 17572 + }, + { + "epoch": 0.92, + "learning_rate": 7.137177972235897e-09, + "loss": 0.139, + "step": 17573 + }, + { + "epoch": 0.92, + "learning_rate": 7.128228775704248e-09, + "loss": 0.1481, + "step": 17574 + }, + { + "epoch": 0.92, + "learning_rate": 7.119285091522265e-09, + "loss": 0.1312, + "step": 17575 + }, + { + "epoch": 0.92, + "learning_rate": 7.110346919945565e-09, + "loss": 0.1869, + "step": 17576 + }, + { + "epoch": 0.92, + "learning_rate": 7.101414261229633e-09, + "loss": 0.1474, + "step": 17577 + }, + { + "epoch": 0.92, + "learning_rate": 7.092487115629753e-09, + "loss": 0.1401, + "step": 17578 + }, + { + "epoch": 0.92, + "learning_rate": 7.083565483401055e-09, + "loss": 0.1723, + "step": 17579 + }, + { + "epoch": 0.92, + "learning_rate": 7.074649364798535e-09, + "loss": 0.1313, + "step": 17580 + }, + { + "epoch": 0.92, + "learning_rate": 7.065738760076989e-09, + "loss": 0.1203, + "step": 17581 + }, + { + "epoch": 0.92, + "learning_rate": 7.056833669491102e-09, + "loss": 0.1596, + "step": 17582 + }, + { + "epoch": 0.92, + "learning_rate": 7.047934093295427e-09, + "loss": 0.1111, + "step": 17583 + }, + { + "epoch": 0.92, + "learning_rate": 7.03904003174427e-09, + "loss": 0.1361, + "step": 17584 + }, + { + "epoch": 0.92, + "learning_rate": 7.030151485091828e-09, + "loss": 0.1543, + "step": 17585 + }, + { + "epoch": 0.92, + "learning_rate": 7.021268453592166e-09, + "loss": 0.1075, + "step": 17586 + }, + { + "epoch": 0.92, + "learning_rate": 7.012390937499146e-09, + "loss": 0.1207, + "step": 17587 + }, + { + "epoch": 0.92, + "learning_rate": 7.003518937066499e-09, + "loss": 0.1618, + "step": 17588 + }, + { + "epoch": 0.92, + "learning_rate": 6.994652452547756e-09, + "loss": 0.1388, + "step": 17589 + }, + { + "epoch": 0.92, + "learning_rate": 6.9857914841963796e-09, + "loss": 0.1407, + "step": 17590 + }, + { + "epoch": 0.92, + "learning_rate": 6.97693603226559e-09, + "loss": 0.1286, + "step": 17591 + }, + { + "epoch": 0.92, + "learning_rate": 6.9680860970084964e-09, + "loss": 0.1158, + "step": 17592 + }, + { + "epoch": 0.92, + "learning_rate": 6.959241678677985e-09, + "loss": 0.1434, + "step": 17593 + }, + { + "epoch": 0.92, + "learning_rate": 6.950402777526876e-09, + "loss": 0.1195, + "step": 17594 + }, + { + "epoch": 0.92, + "learning_rate": 6.941569393807834e-09, + "loss": 0.1504, + "step": 17595 + }, + { + "epoch": 0.92, + "learning_rate": 6.932741527773189e-09, + "loss": 0.1378, + "step": 17596 + }, + { + "epoch": 0.92, + "learning_rate": 6.923919179675364e-09, + "loss": 0.1132, + "step": 17597 + }, + { + "epoch": 0.92, + "learning_rate": 6.915102349766466e-09, + "loss": 0.1678, + "step": 17598 + }, + { + "epoch": 0.92, + "learning_rate": 6.906291038298473e-09, + "loss": 0.1308, + "step": 17599 + }, + { + "epoch": 0.92, + "learning_rate": 6.897485245523205e-09, + "loss": 0.1422, + "step": 17600 + }, + { + "epoch": 0.92, + "learning_rate": 6.888684971692371e-09, + "loss": 0.128, + "step": 17601 + }, + { + "epoch": 0.92, + "learning_rate": 6.8798902170574825e-09, + "loss": 0.1673, + "step": 17602 + }, + { + "epoch": 0.92, + "learning_rate": 6.8711009818698924e-09, + "loss": 0.1441, + "step": 17603 + }, + { + "epoch": 0.92, + "learning_rate": 6.862317266380757e-09, + "loss": 0.1233, + "step": 17604 + }, + { + "epoch": 0.92, + "learning_rate": 6.853539070841163e-09, + "loss": 0.1295, + "step": 17605 + }, + { + "epoch": 0.92, + "learning_rate": 6.844766395502e-09, + "loss": 0.1734, + "step": 17606 + }, + { + "epoch": 0.92, + "learning_rate": 6.8359992406139325e-09, + "loss": 0.1422, + "step": 17607 + }, + { + "epoch": 0.92, + "learning_rate": 6.827237606427627e-09, + "loss": 0.155, + "step": 17608 + }, + { + "epoch": 0.92, + "learning_rate": 6.818481493193418e-09, + "loss": 0.1663, + "step": 17609 + }, + { + "epoch": 0.92, + "learning_rate": 6.809730901161592e-09, + "loss": 0.1901, + "step": 17610 + }, + { + "epoch": 0.92, + "learning_rate": 6.800985830582218e-09, + "loss": 0.1617, + "step": 17611 + }, + { + "epoch": 0.92, + "learning_rate": 6.792246281705294e-09, + "loss": 0.1588, + "step": 17612 + }, + { + "epoch": 0.92, + "learning_rate": 6.783512254780532e-09, + "loss": 0.1255, + "step": 17613 + }, + { + "epoch": 0.92, + "learning_rate": 6.7747837500575776e-09, + "loss": 0.1184, + "step": 17614 + }, + { + "epoch": 0.92, + "learning_rate": 6.766060767785875e-09, + "loss": 0.1187, + "step": 17615 + }, + { + "epoch": 0.92, + "learning_rate": 6.75734330821478e-09, + "loss": 0.1577, + "step": 17616 + }, + { + "epoch": 0.92, + "learning_rate": 6.748631371593405e-09, + "loss": 0.1505, + "step": 17617 + }, + { + "epoch": 0.92, + "learning_rate": 6.739924958170751e-09, + "loss": 0.1064, + "step": 17618 + }, + { + "epoch": 0.92, + "learning_rate": 6.731224068195618e-09, + "loss": 0.13, + "step": 17619 + }, + { + "epoch": 0.92, + "learning_rate": 6.722528701916741e-09, + "loss": 0.1377, + "step": 17620 + }, + { + "epoch": 0.92, + "learning_rate": 6.713838859582632e-09, + "loss": 0.1369, + "step": 17621 + }, + { + "epoch": 0.92, + "learning_rate": 6.705154541441582e-09, + "loss": 0.1034, + "step": 17622 + }, + { + "epoch": 0.92, + "learning_rate": 6.696475747741837e-09, + "loss": 0.1548, + "step": 17623 + }, + { + "epoch": 0.92, + "learning_rate": 6.687802478731463e-09, + "loss": 0.1254, + "step": 17624 + }, + { + "epoch": 0.92, + "learning_rate": 6.679134734658309e-09, + "loss": 0.1617, + "step": 17625 + }, + { + "epoch": 0.92, + "learning_rate": 6.670472515770087e-09, + "loss": 0.1149, + "step": 17626 + }, + { + "epoch": 0.92, + "learning_rate": 6.661815822314421e-09, + "loss": 0.1475, + "step": 17627 + }, + { + "epoch": 0.92, + "learning_rate": 6.65316465453869e-09, + "loss": 0.0872, + "step": 17628 + }, + { + "epoch": 0.92, + "learning_rate": 6.644519012690164e-09, + "loss": 0.1134, + "step": 17629 + }, + { + "epoch": 0.92, + "learning_rate": 6.635878897015912e-09, + "loss": 0.1672, + "step": 17630 + }, + { + "epoch": 0.92, + "learning_rate": 6.627244307762891e-09, + "loss": 0.1343, + "step": 17631 + }, + { + "epoch": 0.92, + "learning_rate": 6.618615245177883e-09, + "loss": 0.103, + "step": 17632 + }, + { + "epoch": 0.92, + "learning_rate": 6.6099917095074675e-09, + "loss": 0.1205, + "step": 17633 + }, + { + "epoch": 0.92, + "learning_rate": 6.601373700998181e-09, + "loss": 0.1477, + "step": 17634 + }, + { + "epoch": 0.92, + "learning_rate": 6.592761219896293e-09, + "loss": 0.1294, + "step": 17635 + }, + { + "epoch": 0.92, + "learning_rate": 6.584154266447961e-09, + "loss": 0.1376, + "step": 17636 + }, + { + "epoch": 0.92, + "learning_rate": 6.575552840899123e-09, + "loss": 0.1444, + "step": 17637 + }, + { + "epoch": 0.92, + "learning_rate": 6.566956943495716e-09, + "loss": 0.1548, + "step": 17638 + }, + { + "epoch": 0.92, + "learning_rate": 6.55836657448332e-09, + "loss": 0.1268, + "step": 17639 + }, + { + "epoch": 0.92, + "learning_rate": 6.5497817341074515e-09, + "loss": 0.1251, + "step": 17640 + }, + { + "epoch": 0.92, + "learning_rate": 6.541202422613534e-09, + "loss": 0.1288, + "step": 17641 + }, + { + "epoch": 0.92, + "learning_rate": 6.532628640246751e-09, + "loss": 0.1301, + "step": 17642 + }, + { + "epoch": 0.92, + "learning_rate": 6.524060387252106e-09, + "loss": 0.1515, + "step": 17643 + }, + { + "epoch": 0.92, + "learning_rate": 6.515497663874492e-09, + "loss": 0.1527, + "step": 17644 + }, + { + "epoch": 0.92, + "learning_rate": 6.506940470358646e-09, + "loss": 0.1242, + "step": 17645 + }, + { + "epoch": 0.92, + "learning_rate": 6.498388806949195e-09, + "loss": 0.1278, + "step": 17646 + }, + { + "epoch": 0.92, + "learning_rate": 6.489842673890411e-09, + "loss": 0.1608, + "step": 17647 + }, + { + "epoch": 0.92, + "learning_rate": 6.481302071426675e-09, + "loss": 0.1464, + "step": 17648 + }, + { + "epoch": 0.92, + "learning_rate": 6.472766999802015e-09, + "loss": 0.172, + "step": 17649 + }, + { + "epoch": 0.92, + "learning_rate": 6.464237459260391e-09, + "loss": 0.1658, + "step": 17650 + }, + { + "epoch": 0.92, + "learning_rate": 6.455713450045563e-09, + "loss": 0.143, + "step": 17651 + }, + { + "epoch": 0.92, + "learning_rate": 6.44719497240116e-09, + "loss": 0.1242, + "step": 17652 + }, + { + "epoch": 0.92, + "learning_rate": 6.438682026570652e-09, + "loss": 0.1056, + "step": 17653 + }, + { + "epoch": 0.92, + "learning_rate": 6.430174612797334e-09, + "loss": 0.1694, + "step": 17654 + }, + { + "epoch": 0.92, + "learning_rate": 6.421672731324368e-09, + "loss": 0.1308, + "step": 17655 + }, + { + "epoch": 0.92, + "learning_rate": 6.4131763823946916e-09, + "loss": 0.1104, + "step": 17656 + }, + { + "epoch": 0.92, + "learning_rate": 6.4046855662512e-09, + "loss": 0.13, + "step": 17657 + }, + { + "epoch": 0.92, + "learning_rate": 6.396200283136477e-09, + "loss": 0.1515, + "step": 17658 + }, + { + "epoch": 0.92, + "learning_rate": 6.387720533293128e-09, + "loss": 0.1344, + "step": 17659 + }, + { + "epoch": 0.92, + "learning_rate": 6.379246316963471e-09, + "loss": 0.1254, + "step": 17660 + }, + { + "epoch": 0.92, + "learning_rate": 6.3707776343896905e-09, + "loss": 0.0996, + "step": 17661 + }, + { + "epoch": 0.92, + "learning_rate": 6.362314485813813e-09, + "loss": 0.1184, + "step": 17662 + }, + { + "epoch": 0.92, + "learning_rate": 6.353856871477781e-09, + "loss": 0.1291, + "step": 17663 + }, + { + "epoch": 0.92, + "learning_rate": 6.345404791623266e-09, + "loss": 0.1384, + "step": 17664 + }, + { + "epoch": 0.92, + "learning_rate": 6.336958246491808e-09, + "loss": 0.1121, + "step": 17665 + }, + { + "epoch": 0.92, + "learning_rate": 6.32851723632486e-09, + "loss": 0.1201, + "step": 17666 + }, + { + "epoch": 0.92, + "learning_rate": 6.320081761363649e-09, + "loss": 0.1353, + "step": 17667 + }, + { + "epoch": 0.92, + "learning_rate": 6.3116518218492955e-09, + "loss": 0.1489, + "step": 17668 + }, + { + "epoch": 0.92, + "learning_rate": 6.3032274180226715e-09, + "loss": 0.0928, + "step": 17669 + }, + { + "epoch": 0.92, + "learning_rate": 6.294808550124586e-09, + "loss": 0.1289, + "step": 17670 + }, + { + "epoch": 0.92, + "learning_rate": 6.286395218395668e-09, + "loss": 0.1297, + "step": 17671 + }, + { + "epoch": 0.92, + "learning_rate": 6.277987423076347e-09, + "loss": 0.1539, + "step": 17672 + }, + { + "epoch": 0.92, + "learning_rate": 6.269585164406943e-09, + "loss": 0.1269, + "step": 17673 + }, + { + "epoch": 0.92, + "learning_rate": 6.2611884426275744e-09, + "loss": 0.1566, + "step": 17674 + }, + { + "epoch": 0.92, + "learning_rate": 6.252797257978226e-09, + "loss": 0.1386, + "step": 17675 + }, + { + "epoch": 0.92, + "learning_rate": 6.244411610698708e-09, + "loss": 0.1435, + "step": 17676 + }, + { + "epoch": 0.92, + "learning_rate": 6.236031501028738e-09, + "loss": 0.1129, + "step": 17677 + }, + { + "epoch": 0.92, + "learning_rate": 6.22765692920777e-09, + "loss": 0.1327, + "step": 17678 + }, + { + "epoch": 0.92, + "learning_rate": 6.2192878954751675e-09, + "loss": 0.1433, + "step": 17679 + }, + { + "epoch": 0.92, + "learning_rate": 6.210924400070117e-09, + "loss": 0.1378, + "step": 17680 + }, + { + "epoch": 0.92, + "learning_rate": 6.202566443231694e-09, + "loss": 0.1477, + "step": 17681 + }, + { + "epoch": 0.92, + "learning_rate": 6.194214025198708e-09, + "loss": 0.1599, + "step": 17682 + }, + { + "epoch": 0.92, + "learning_rate": 6.1858671462098996e-09, + "loss": 0.1399, + "step": 17683 + }, + { + "epoch": 0.92, + "learning_rate": 6.177525806503836e-09, + "loss": 0.1433, + "step": 17684 + }, + { + "epoch": 0.92, + "learning_rate": 6.169190006318903e-09, + "loss": 0.1212, + "step": 17685 + }, + { + "epoch": 0.92, + "learning_rate": 6.160859745893354e-09, + "loss": 0.1405, + "step": 17686 + }, + { + "epoch": 0.92, + "learning_rate": 6.1525350254652444e-09, + "loss": 0.1284, + "step": 17687 + }, + { + "epoch": 0.92, + "learning_rate": 6.144215845272516e-09, + "loss": 0.1296, + "step": 17688 + }, + { + "epoch": 0.92, + "learning_rate": 6.135902205552978e-09, + "loss": 0.1507, + "step": 17689 + }, + { + "epoch": 0.92, + "learning_rate": 6.127594106544154e-09, + "loss": 0.1337, + "step": 17690 + }, + { + "epoch": 0.92, + "learning_rate": 6.119291548483541e-09, + "loss": 0.1292, + "step": 17691 + }, + { + "epoch": 0.92, + "learning_rate": 6.110994531608416e-09, + "loss": 0.1311, + "step": 17692 + }, + { + "epoch": 0.92, + "learning_rate": 6.102703056155945e-09, + "loss": 0.1178, + "step": 17693 + }, + { + "epoch": 0.92, + "learning_rate": 6.094417122363027e-09, + "loss": 0.1269, + "step": 17694 + }, + { + "epoch": 0.92, + "learning_rate": 6.08613673046654e-09, + "loss": 0.134, + "step": 17695 + }, + { + "epoch": 0.92, + "learning_rate": 6.077861880703139e-09, + "loss": 0.1374, + "step": 17696 + }, + { + "epoch": 0.92, + "learning_rate": 6.069592573309301e-09, + "loss": 0.0998, + "step": 17697 + }, + { + "epoch": 0.92, + "learning_rate": 6.061328808521371e-09, + "loss": 0.1215, + "step": 17698 + }, + { + "epoch": 0.92, + "learning_rate": 6.053070586575537e-09, + "loss": 0.1413, + "step": 17699 + }, + { + "epoch": 0.92, + "learning_rate": 6.044817907707811e-09, + "loss": 0.1617, + "step": 17700 + }, + { + "epoch": 0.92, + "learning_rate": 6.036570772154026e-09, + "loss": 0.1162, + "step": 17701 + }, + { + "epoch": 0.92, + "learning_rate": 6.028329180149972e-09, + "loss": 0.1134, + "step": 17702 + }, + { + "epoch": 0.92, + "learning_rate": 6.020093131931148e-09, + "loss": 0.1115, + "step": 17703 + }, + { + "epoch": 0.92, + "learning_rate": 6.011862627732922e-09, + "loss": 0.1194, + "step": 17704 + }, + { + "epoch": 0.92, + "learning_rate": 6.003637667790551e-09, + "loss": 0.1166, + "step": 17705 + }, + { + "epoch": 0.92, + "learning_rate": 5.995418252339157e-09, + "loss": 0.137, + "step": 17706 + }, + { + "epoch": 0.92, + "learning_rate": 5.987204381613553e-09, + "loss": 0.1476, + "step": 17707 + }, + { + "epoch": 0.92, + "learning_rate": 5.9789960558485504e-09, + "loss": 0.1545, + "step": 17708 + }, + { + "epoch": 0.92, + "learning_rate": 5.97079327527874e-09, + "loss": 0.1623, + "step": 17709 + }, + { + "epoch": 0.92, + "learning_rate": 5.962596040138579e-09, + "loss": 0.1621, + "step": 17710 + }, + { + "epoch": 0.92, + "learning_rate": 5.954404350662301e-09, + "loss": 0.1159, + "step": 17711 + }, + { + "epoch": 0.92, + "learning_rate": 5.9462182070840525e-09, + "loss": 0.1207, + "step": 17712 + }, + { + "epoch": 0.92, + "learning_rate": 5.938037609637825e-09, + "loss": 0.116, + "step": 17713 + }, + { + "epoch": 0.92, + "learning_rate": 5.929862558557408e-09, + "loss": 0.1131, + "step": 17714 + }, + { + "epoch": 0.92, + "learning_rate": 5.921693054076438e-09, + "loss": 0.1361, + "step": 17715 + }, + { + "epoch": 0.92, + "learning_rate": 5.913529096428371e-09, + "loss": 0.143, + "step": 17716 + }, + { + "epoch": 0.92, + "learning_rate": 5.905370685846578e-09, + "loss": 0.145, + "step": 17717 + }, + { + "epoch": 0.92, + "learning_rate": 5.897217822564226e-09, + "loss": 0.1863, + "step": 17718 + }, + { + "epoch": 0.92, + "learning_rate": 5.8890705068142866e-09, + "loss": 0.1378, + "step": 17719 + }, + { + "epoch": 0.92, + "learning_rate": 5.880928738829682e-09, + "loss": 0.1555, + "step": 17720 + }, + { + "epoch": 0.92, + "learning_rate": 5.8727925188430725e-09, + "loss": 0.1097, + "step": 17721 + }, + { + "epoch": 0.93, + "learning_rate": 5.864661847086982e-09, + "loss": 0.1896, + "step": 17722 + }, + { + "epoch": 0.93, + "learning_rate": 5.856536723793826e-09, + "loss": 0.1245, + "step": 17723 + }, + { + "epoch": 0.93, + "learning_rate": 5.848417149195772e-09, + "loss": 0.1406, + "step": 17724 + }, + { + "epoch": 0.93, + "learning_rate": 5.840303123524904e-09, + "loss": 0.1627, + "step": 17725 + }, + { + "epoch": 0.93, + "learning_rate": 5.832194647013145e-09, + "loss": 0.1354, + "step": 17726 + }, + { + "epoch": 0.93, + "learning_rate": 5.824091719892177e-09, + "loss": 0.1373, + "step": 17727 + }, + { + "epoch": 0.93, + "learning_rate": 5.8159943423936595e-09, + "loss": 0.1396, + "step": 17728 + }, + { + "epoch": 0.93, + "learning_rate": 5.807902514748985e-09, + "loss": 0.1449, + "step": 17729 + }, + { + "epoch": 0.93, + "learning_rate": 5.7998162371894365e-09, + "loss": 0.1118, + "step": 17730 + }, + { + "epoch": 0.93, + "learning_rate": 5.791735509946072e-09, + "loss": 0.1438, + "step": 17731 + }, + { + "epoch": 0.93, + "learning_rate": 5.783660333249929e-09, + "loss": 0.1483, + "step": 17732 + }, + { + "epoch": 0.93, + "learning_rate": 5.775590707331712e-09, + "loss": 0.136, + "step": 17733 + }, + { + "epoch": 0.93, + "learning_rate": 5.7675266324220815e-09, + "loss": 0.1469, + "step": 17734 + }, + { + "epoch": 0.93, + "learning_rate": 5.759468108751542e-09, + "loss": 0.1386, + "step": 17735 + }, + { + "epoch": 0.93, + "learning_rate": 5.751415136550375e-09, + "loss": 0.103, + "step": 17736 + }, + { + "epoch": 0.93, + "learning_rate": 5.743367716048775e-09, + "loss": 0.1204, + "step": 17737 + }, + { + "epoch": 0.93, + "learning_rate": 5.735325847476668e-09, + "loss": 0.1443, + "step": 17738 + }, + { + "epoch": 0.93, + "learning_rate": 5.727289531063983e-09, + "loss": 0.1206, + "step": 17739 + }, + { + "epoch": 0.93, + "learning_rate": 5.719258767040358e-09, + "loss": 0.1026, + "step": 17740 + }, + { + "epoch": 0.93, + "learning_rate": 5.71123355563532e-09, + "loss": 0.1284, + "step": 17741 + }, + { + "epoch": 0.93, + "learning_rate": 5.703213897078219e-09, + "loss": 0.1627, + "step": 17742 + }, + { + "epoch": 0.93, + "learning_rate": 5.695199791598271e-09, + "loss": 0.1299, + "step": 17743 + }, + { + "epoch": 0.93, + "learning_rate": 5.687191239424516e-09, + "loss": 0.1074, + "step": 17744 + }, + { + "epoch": 0.93, + "learning_rate": 5.679188240785837e-09, + "loss": 0.1399, + "step": 17745 + }, + { + "epoch": 0.93, + "learning_rate": 5.6711907959110076e-09, + "loss": 0.1739, + "step": 17746 + }, + { + "epoch": 0.93, + "learning_rate": 5.663198905028555e-09, + "loss": 0.1471, + "step": 17747 + }, + { + "epoch": 0.93, + "learning_rate": 5.655212568366896e-09, + "loss": 0.1332, + "step": 17748 + }, + { + "epoch": 0.93, + "learning_rate": 5.6472317861542715e-09, + "loss": 0.1651, + "step": 17749 + }, + { + "epoch": 0.93, + "learning_rate": 5.63925655861881e-09, + "loss": 0.1477, + "step": 17750 + }, + { + "epoch": 0.93, + "learning_rate": 5.631286885988418e-09, + "loss": 0.1574, + "step": 17751 + }, + { + "epoch": 0.93, + "learning_rate": 5.623322768490868e-09, + "loss": 0.1172, + "step": 17752 + }, + { + "epoch": 0.93, + "learning_rate": 5.6153642063538005e-09, + "loss": 0.1343, + "step": 17753 + }, + { + "epoch": 0.93, + "learning_rate": 5.607411199804679e-09, + "loss": 0.1292, + "step": 17754 + }, + { + "epoch": 0.93, + "learning_rate": 5.599463749070765e-09, + "loss": 0.1158, + "step": 17755 + }, + { + "epoch": 0.93, + "learning_rate": 5.5915218543792335e-09, + "loss": 0.1448, + "step": 17756 + }, + { + "epoch": 0.93, + "learning_rate": 5.583585515957057e-09, + "loss": 0.1546, + "step": 17757 + }, + { + "epoch": 0.93, + "learning_rate": 5.575654734031077e-09, + "loss": 0.1394, + "step": 17758 + }, + { + "epoch": 0.93, + "learning_rate": 5.567729508827912e-09, + "loss": 0.1171, + "step": 17759 + }, + { + "epoch": 0.93, + "learning_rate": 5.559809840574092e-09, + "loss": 0.1128, + "step": 17760 + }, + { + "epoch": 0.93, + "learning_rate": 5.551895729495992e-09, + "loss": 0.1147, + "step": 17761 + }, + { + "epoch": 0.93, + "learning_rate": 5.5439871758197645e-09, + "loss": 0.1129, + "step": 17762 + }, + { + "epoch": 0.93, + "learning_rate": 5.53608417977145e-09, + "loss": 0.1294, + "step": 17763 + }, + { + "epoch": 0.93, + "learning_rate": 5.528186741576912e-09, + "loss": 0.0997, + "step": 17764 + }, + { + "epoch": 0.93, + "learning_rate": 5.520294861461882e-09, + "loss": 0.1111, + "step": 17765 + }, + { + "epoch": 0.93, + "learning_rate": 5.512408539651914e-09, + "loss": 0.1162, + "step": 17766 + }, + { + "epoch": 0.93, + "learning_rate": 5.504527776372381e-09, + "loss": 0.1435, + "step": 17767 + }, + { + "epoch": 0.93, + "learning_rate": 5.4966525718485254e-09, + "loss": 0.1384, + "step": 17768 + }, + { + "epoch": 0.93, + "learning_rate": 5.488782926305435e-09, + "loss": 0.1235, + "step": 17769 + }, + { + "epoch": 0.93, + "learning_rate": 5.480918839967974e-09, + "loss": 0.1159, + "step": 17770 + }, + { + "epoch": 0.93, + "learning_rate": 5.473060313060984e-09, + "loss": 0.1126, + "step": 17771 + }, + { + "epoch": 0.93, + "learning_rate": 5.4652073458090195e-09, + "loss": 0.1296, + "step": 17772 + }, + { + "epoch": 0.93, + "learning_rate": 5.457359938436523e-09, + "loss": 0.1495, + "step": 17773 + }, + { + "epoch": 0.93, + "learning_rate": 5.449518091167782e-09, + "loss": 0.1308, + "step": 17774 + }, + { + "epoch": 0.93, + "learning_rate": 5.441681804226927e-09, + "loss": 0.1431, + "step": 17775 + }, + { + "epoch": 0.93, + "learning_rate": 5.4338510778379145e-09, + "loss": 0.1054, + "step": 17776 + }, + { + "epoch": 0.93, + "learning_rate": 5.4260259122245185e-09, + "loss": 0.1242, + "step": 17777 + }, + { + "epoch": 0.93, + "learning_rate": 5.418206307610429e-09, + "loss": 0.1229, + "step": 17778 + }, + { + "epoch": 0.93, + "learning_rate": 5.4103922642191324e-09, + "loss": 0.1776, + "step": 17779 + }, + { + "epoch": 0.93, + "learning_rate": 5.40258378227394e-09, + "loss": 0.1142, + "step": 17780 + }, + { + "epoch": 0.93, + "learning_rate": 5.394780861998005e-09, + "loss": 0.1614, + "step": 17781 + }, + { + "epoch": 0.93, + "learning_rate": 5.3869835036143725e-09, + "loss": 0.1252, + "step": 17782 + }, + { + "epoch": 0.93, + "learning_rate": 5.379191707345909e-09, + "loss": 0.1204, + "step": 17783 + }, + { + "epoch": 0.93, + "learning_rate": 5.371405473415236e-09, + "loss": 0.1423, + "step": 17784 + }, + { + "epoch": 0.93, + "learning_rate": 5.363624802044952e-09, + "loss": 0.1425, + "step": 17785 + }, + { + "epoch": 0.93, + "learning_rate": 5.355849693457393e-09, + "loss": 0.1425, + "step": 17786 + }, + { + "epoch": 0.93, + "learning_rate": 5.3480801478748e-09, + "loss": 0.1413, + "step": 17787 + }, + { + "epoch": 0.93, + "learning_rate": 5.340316165519199e-09, + "loss": 0.1363, + "step": 17788 + }, + { + "epoch": 0.93, + "learning_rate": 5.3325577466125425e-09, + "loss": 0.1323, + "step": 17789 + }, + { + "epoch": 0.93, + "learning_rate": 5.324804891376522e-09, + "loss": 0.1224, + "step": 17790 + }, + { + "epoch": 0.93, + "learning_rate": 5.317057600032715e-09, + "loss": 0.1236, + "step": 17791 + }, + { + "epoch": 0.93, + "learning_rate": 5.309315872802589e-09, + "loss": 0.1318, + "step": 17792 + }, + { + "epoch": 0.93, + "learning_rate": 5.301579709907344e-09, + "loss": 0.1415, + "step": 17793 + }, + { + "epoch": 0.93, + "learning_rate": 5.293849111568116e-09, + "loss": 0.1334, + "step": 17794 + }, + { + "epoch": 0.93, + "learning_rate": 5.286124078005838e-09, + "loss": 0.1219, + "step": 17795 + }, + { + "epoch": 0.93, + "learning_rate": 5.278404609441311e-09, + "loss": 0.147, + "step": 17796 + }, + { + "epoch": 0.93, + "learning_rate": 5.27069070609516e-09, + "loss": 0.1532, + "step": 17797 + }, + { + "epoch": 0.93, + "learning_rate": 5.26298236818783e-09, + "loss": 0.1279, + "step": 17798 + }, + { + "epoch": 0.93, + "learning_rate": 5.255279595939632e-09, + "loss": 0.1188, + "step": 17799 + }, + { + "epoch": 0.93, + "learning_rate": 5.247582389570726e-09, + "loss": 0.1314, + "step": 17800 + }, + { + "epoch": 0.93, + "learning_rate": 5.239890749301135e-09, + "loss": 0.1395, + "step": 17801 + }, + { + "epoch": 0.93, + "learning_rate": 5.2322046753505935e-09, + "loss": 0.1382, + "step": 17802 + }, + { + "epoch": 0.93, + "learning_rate": 5.224524167938882e-09, + "loss": 0.1485, + "step": 17803 + }, + { + "epoch": 0.93, + "learning_rate": 5.216849227285425e-09, + "loss": 0.1435, + "step": 17804 + }, + { + "epoch": 0.93, + "learning_rate": 5.209179853609624e-09, + "loss": 0.161, + "step": 17805 + }, + { + "epoch": 0.93, + "learning_rate": 5.201516047130638e-09, + "loss": 0.1337, + "step": 17806 + }, + { + "epoch": 0.93, + "learning_rate": 5.193857808067559e-09, + "loss": 0.1444, + "step": 17807 + }, + { + "epoch": 0.93, + "learning_rate": 5.18620513663921e-09, + "loss": 0.1079, + "step": 17808 + }, + { + "epoch": 0.93, + "learning_rate": 5.178558033064328e-09, + "loss": 0.1135, + "step": 17809 + }, + { + "epoch": 0.93, + "learning_rate": 5.17091649756145e-09, + "loss": 0.1128, + "step": 17810 + }, + { + "epoch": 0.93, + "learning_rate": 5.163280530349023e-09, + "loss": 0.1374, + "step": 17811 + }, + { + "epoch": 0.93, + "learning_rate": 5.155650131645228e-09, + "loss": 0.1195, + "step": 17812 + }, + { + "epoch": 0.93, + "learning_rate": 5.1480253016681576e-09, + "loss": 0.1355, + "step": 17813 + }, + { + "epoch": 0.93, + "learning_rate": 5.14040604063577e-09, + "loss": 0.1448, + "step": 17814 + }, + { + "epoch": 0.93, + "learning_rate": 5.13279234876578e-09, + "loss": 0.1432, + "step": 17815 + }, + { + "epoch": 0.93, + "learning_rate": 5.125184226275836e-09, + "loss": 0.1202, + "step": 17816 + }, + { + "epoch": 0.93, + "learning_rate": 5.11758167338332e-09, + "loss": 0.1523, + "step": 17817 + }, + { + "epoch": 0.93, + "learning_rate": 5.109984690305613e-09, + "loss": 0.1103, + "step": 17818 + }, + { + "epoch": 0.93, + "learning_rate": 5.1023932772597425e-09, + "loss": 0.1225, + "step": 17819 + }, + { + "epoch": 0.93, + "learning_rate": 5.094807434462689e-09, + "loss": 0.1207, + "step": 17820 + }, + { + "epoch": 0.93, + "learning_rate": 5.087227162131303e-09, + "loss": 0.1079, + "step": 17821 + }, + { + "epoch": 0.93, + "learning_rate": 5.079652460482209e-09, + "loss": 0.1412, + "step": 17822 + }, + { + "epoch": 0.93, + "learning_rate": 5.07208332973188e-09, + "loss": 0.1306, + "step": 17823 + }, + { + "epoch": 0.93, + "learning_rate": 5.064519770096653e-09, + "loss": 0.1137, + "step": 17824 + }, + { + "epoch": 0.93, + "learning_rate": 5.0569617817927344e-09, + "loss": 0.1255, + "step": 17825 + }, + { + "epoch": 0.93, + "learning_rate": 5.049409365036106e-09, + "loss": 0.1487, + "step": 17826 + }, + { + "epoch": 0.93, + "learning_rate": 5.041862520042594e-09, + "loss": 0.1544, + "step": 17827 + }, + { + "epoch": 0.93, + "learning_rate": 5.034321247027917e-09, + "loss": 0.1432, + "step": 17828 + }, + { + "epoch": 0.93, + "learning_rate": 5.026785546207591e-09, + "loss": 0.1728, + "step": 17829 + }, + { + "epoch": 0.93, + "learning_rate": 5.019255417797019e-09, + "loss": 0.148, + "step": 17830 + }, + { + "epoch": 0.93, + "learning_rate": 5.011730862011387e-09, + "loss": 0.1315, + "step": 17831 + }, + { + "epoch": 0.93, + "learning_rate": 5.004211879065767e-09, + "loss": 0.194, + "step": 17832 + }, + { + "epoch": 0.93, + "learning_rate": 4.996698469175054e-09, + "loss": 0.1241, + "step": 17833 + }, + { + "epoch": 0.93, + "learning_rate": 4.989190632553963e-09, + "loss": 0.1453, + "step": 17834 + }, + { + "epoch": 0.93, + "learning_rate": 4.981688369417103e-09, + "loss": 0.1341, + "step": 17835 + }, + { + "epoch": 0.93, + "learning_rate": 4.974191679978856e-09, + "loss": 0.1237, + "step": 17836 + }, + { + "epoch": 0.93, + "learning_rate": 4.966700564453518e-09, + "loss": 0.1033, + "step": 17837 + }, + { + "epoch": 0.93, + "learning_rate": 4.959215023055163e-09, + "loss": 0.1837, + "step": 17838 + }, + { + "epoch": 0.93, + "learning_rate": 4.951735055997708e-09, + "loss": 0.1359, + "step": 17839 + }, + { + "epoch": 0.93, + "learning_rate": 4.944260663494981e-09, + "loss": 0.1242, + "step": 17840 + }, + { + "epoch": 0.93, + "learning_rate": 4.936791845760591e-09, + "loss": 0.1242, + "step": 17841 + }, + { + "epoch": 0.93, + "learning_rate": 4.929328603007987e-09, + "loss": 0.1078, + "step": 17842 + }, + { + "epoch": 0.93, + "learning_rate": 4.92187093545049e-09, + "loss": 0.1383, + "step": 17843 + }, + { + "epoch": 0.93, + "learning_rate": 4.914418843301216e-09, + "loss": 0.1268, + "step": 17844 + }, + { + "epoch": 0.93, + "learning_rate": 4.906972326773151e-09, + "loss": 0.1277, + "step": 17845 + }, + { + "epoch": 0.93, + "learning_rate": 4.8995313860791256e-09, + "loss": 0.1118, + "step": 17846 + }, + { + "epoch": 0.93, + "learning_rate": 4.892096021431835e-09, + "loss": 0.1704, + "step": 17847 + }, + { + "epoch": 0.93, + "learning_rate": 4.884666233043755e-09, + "loss": 0.1555, + "step": 17848 + }, + { + "epoch": 0.93, + "learning_rate": 4.877242021127226e-09, + "loss": 0.1068, + "step": 17849 + }, + { + "epoch": 0.93, + "learning_rate": 4.869823385894434e-09, + "loss": 0.1193, + "step": 17850 + }, + { + "epoch": 0.93, + "learning_rate": 4.862410327557431e-09, + "loss": 0.1256, + "step": 17851 + }, + { + "epoch": 0.93, + "learning_rate": 4.8550028463280935e-09, + "loss": 0.1436, + "step": 17852 + }, + { + "epoch": 0.93, + "learning_rate": 4.847600942418073e-09, + "loss": 0.1223, + "step": 17853 + }, + { + "epoch": 0.93, + "learning_rate": 4.840204616038979e-09, + "loss": 0.1507, + "step": 17854 + }, + { + "epoch": 0.93, + "learning_rate": 4.8328138674021525e-09, + "loss": 0.1549, + "step": 17855 + }, + { + "epoch": 0.93, + "learning_rate": 4.825428696718869e-09, + "loss": 0.1512, + "step": 17856 + }, + { + "epoch": 0.93, + "learning_rate": 4.818049104200161e-09, + "loss": 0.1479, + "step": 17857 + }, + { + "epoch": 0.93, + "learning_rate": 4.8106750900569704e-09, + "loss": 0.1212, + "step": 17858 + }, + { + "epoch": 0.93, + "learning_rate": 4.803306654500017e-09, + "loss": 0.1204, + "step": 17859 + }, + { + "epoch": 0.93, + "learning_rate": 4.795943797739932e-09, + "loss": 0.1608, + "step": 17860 + }, + { + "epoch": 0.93, + "learning_rate": 4.788586519987148e-09, + "loss": 0.1752, + "step": 17861 + }, + { + "epoch": 0.93, + "learning_rate": 4.781234821451896e-09, + "loss": 0.1117, + "step": 17862 + }, + { + "epoch": 0.93, + "learning_rate": 4.7738887023443195e-09, + "loss": 0.1612, + "step": 17863 + }, + { + "epoch": 0.93, + "learning_rate": 4.76654816287434e-09, + "loss": 0.1337, + "step": 17864 + }, + { + "epoch": 0.93, + "learning_rate": 4.759213203251811e-09, + "loss": 0.1279, + "step": 17865 + }, + { + "epoch": 0.93, + "learning_rate": 4.751883823686342e-09, + "loss": 0.1322, + "step": 17866 + }, + { + "epoch": 0.93, + "learning_rate": 4.744560024387412e-09, + "loss": 0.1411, + "step": 17867 + }, + { + "epoch": 0.93, + "learning_rate": 4.737241805564296e-09, + "loss": 0.1246, + "step": 17868 + }, + { + "epoch": 0.93, + "learning_rate": 4.72992916742625e-09, + "loss": 0.1201, + "step": 17869 + }, + { + "epoch": 0.93, + "learning_rate": 4.722622110182173e-09, + "loss": 0.1265, + "step": 17870 + }, + { + "epoch": 0.93, + "learning_rate": 4.71532063404092e-09, + "loss": 0.1227, + "step": 17871 + }, + { + "epoch": 0.93, + "learning_rate": 4.7080247392112135e-09, + "loss": 0.1478, + "step": 17872 + }, + { + "epoch": 0.93, + "learning_rate": 4.7007344259015536e-09, + "loss": 0.1246, + "step": 17873 + }, + { + "epoch": 0.93, + "learning_rate": 4.693449694320284e-09, + "loss": 0.1339, + "step": 17874 + }, + { + "epoch": 0.93, + "learning_rate": 4.686170544675616e-09, + "loss": 0.1105, + "step": 17875 + }, + { + "epoch": 0.93, + "learning_rate": 4.678896977175606e-09, + "loss": 0.1493, + "step": 17876 + }, + { + "epoch": 0.93, + "learning_rate": 4.671628992028132e-09, + "loss": 0.1353, + "step": 17877 + }, + { + "epoch": 0.93, + "learning_rate": 4.664366589440894e-09, + "loss": 0.1076, + "step": 17878 + }, + { + "epoch": 0.93, + "learning_rate": 4.657109769621459e-09, + "loss": 0.1437, + "step": 17879 + }, + { + "epoch": 0.93, + "learning_rate": 4.64985853277724e-09, + "loss": 0.1338, + "step": 17880 + }, + { + "epoch": 0.93, + "learning_rate": 4.6426128791154924e-09, + "loss": 0.1823, + "step": 17881 + }, + { + "epoch": 0.93, + "learning_rate": 4.635372808843252e-09, + "loss": 0.1471, + "step": 17882 + }, + { + "epoch": 0.93, + "learning_rate": 4.628138322167508e-09, + "loss": 0.1206, + "step": 17883 + }, + { + "epoch": 0.93, + "learning_rate": 4.620909419294983e-09, + "loss": 0.126, + "step": 17884 + }, + { + "epoch": 0.93, + "learning_rate": 4.61368610043229e-09, + "loss": 0.12, + "step": 17885 + }, + { + "epoch": 0.93, + "learning_rate": 4.606468365785865e-09, + "loss": 0.13, + "step": 17886 + }, + { + "epoch": 0.93, + "learning_rate": 4.599256215562009e-09, + "loss": 0.1395, + "step": 17887 + }, + { + "epoch": 0.93, + "learning_rate": 4.592049649966845e-09, + "loss": 0.1308, + "step": 17888 + }, + { + "epoch": 0.93, + "learning_rate": 4.584848669206321e-09, + "loss": 0.1789, + "step": 17889 + }, + { + "epoch": 0.93, + "learning_rate": 4.577653273486271e-09, + "loss": 0.1288, + "step": 17890 + }, + { + "epoch": 0.93, + "learning_rate": 4.570463463012353e-09, + "loss": 0.1717, + "step": 17891 + }, + { + "epoch": 0.93, + "learning_rate": 4.563279237990003e-09, + "loss": 0.115, + "step": 17892 + }, + { + "epoch": 0.93, + "learning_rate": 4.5561005986245684e-09, + "loss": 0.1158, + "step": 17893 + }, + { + "epoch": 0.93, + "learning_rate": 4.5489275451212614e-09, + "loss": 0.1358, + "step": 17894 + }, + { + "epoch": 0.93, + "learning_rate": 4.5417600776850525e-09, + "loss": 0.1438, + "step": 17895 + }, + { + "epoch": 0.93, + "learning_rate": 4.534598196520778e-09, + "loss": 0.1256, + "step": 17896 + }, + { + "epoch": 0.93, + "learning_rate": 4.527441901833162e-09, + "loss": 0.1347, + "step": 17897 + }, + { + "epoch": 0.93, + "learning_rate": 4.520291193826708e-09, + "loss": 0.1407, + "step": 17898 + }, + { + "epoch": 0.93, + "learning_rate": 4.513146072705787e-09, + "loss": 0.1364, + "step": 17899 + }, + { + "epoch": 0.93, + "learning_rate": 4.5060065386746115e-09, + "loss": 0.1486, + "step": 17900 + }, + { + "epoch": 0.93, + "learning_rate": 4.498872591937264e-09, + "loss": 0.1302, + "step": 17901 + }, + { + "epoch": 0.93, + "learning_rate": 4.49174423269758e-09, + "loss": 0.165, + "step": 17902 + }, + { + "epoch": 0.93, + "learning_rate": 4.484621461159355e-09, + "loss": 0.1456, + "step": 17903 + }, + { + "epoch": 0.93, + "learning_rate": 4.47750427752609e-09, + "loss": 0.1591, + "step": 17904 + }, + { + "epoch": 0.93, + "learning_rate": 4.470392682001245e-09, + "loss": 0.1044, + "step": 17905 + }, + { + "epoch": 0.93, + "learning_rate": 4.4632866747880804e-09, + "loss": 0.1148, + "step": 17906 + }, + { + "epoch": 0.93, + "learning_rate": 4.456186256089611e-09, + "loss": 0.1068, + "step": 17907 + }, + { + "epoch": 0.93, + "learning_rate": 4.449091426108875e-09, + "loss": 0.0949, + "step": 17908 + }, + { + "epoch": 0.93, + "learning_rate": 4.4420021850485765e-09, + "loss": 0.1059, + "step": 17909 + }, + { + "epoch": 0.93, + "learning_rate": 4.434918533111376e-09, + "loss": 0.1272, + "step": 17910 + }, + { + "epoch": 0.93, + "learning_rate": 4.4278404704996445e-09, + "loss": 0.1308, + "step": 17911 + }, + { + "epoch": 0.93, + "learning_rate": 4.420767997415798e-09, + "loss": 0.1258, + "step": 17912 + }, + { + "epoch": 0.93, + "learning_rate": 4.413701114061874e-09, + "loss": 0.1627, + "step": 17913 + }, + { + "epoch": 0.94, + "learning_rate": 4.406639820639846e-09, + "loss": 0.1361, + "step": 17914 + }, + { + "epoch": 0.94, + "learning_rate": 4.399584117351596e-09, + "loss": 0.1334, + "step": 17915 + }, + { + "epoch": 0.94, + "learning_rate": 4.39253400439874e-09, + "loss": 0.1292, + "step": 17916 + }, + { + "epoch": 0.94, + "learning_rate": 4.385489481982763e-09, + "loss": 0.1466, + "step": 17917 + }, + { + "epoch": 0.94, + "learning_rate": 4.378450550304991e-09, + "loss": 0.1141, + "step": 17918 + }, + { + "epoch": 0.94, + "learning_rate": 4.371417209566641e-09, + "loss": 0.1288, + "step": 17919 + }, + { + "epoch": 0.94, + "learning_rate": 4.364389459968731e-09, + "loss": 0.1641, + "step": 17920 + }, + { + "epoch": 0.94, + "learning_rate": 4.357367301712056e-09, + "loss": 0.1531, + "step": 17921 + }, + { + "epoch": 0.94, + "learning_rate": 4.350350734997366e-09, + "loss": 0.1411, + "step": 17922 + }, + { + "epoch": 0.94, + "learning_rate": 4.343339760025189e-09, + "loss": 0.1271, + "step": 17923 + }, + { + "epoch": 0.94, + "learning_rate": 4.336334376995876e-09, + "loss": 0.1736, + "step": 17924 + }, + { + "epoch": 0.94, + "learning_rate": 4.329334586109645e-09, + "loss": 0.1269, + "step": 17925 + }, + { + "epoch": 0.94, + "learning_rate": 4.3223403875666035e-09, + "loss": 0.1781, + "step": 17926 + }, + { + "epoch": 0.94, + "learning_rate": 4.31535178156659e-09, + "loss": 0.1109, + "step": 17927 + }, + { + "epoch": 0.94, + "learning_rate": 4.308368768309379e-09, + "loss": 0.1422, + "step": 17928 + }, + { + "epoch": 0.94, + "learning_rate": 4.301391347994521e-09, + "loss": 0.1283, + "step": 17929 + }, + { + "epoch": 0.94, + "learning_rate": 4.294419520821457e-09, + "loss": 0.1409, + "step": 17930 + }, + { + "epoch": 0.94, + "learning_rate": 4.287453286989429e-09, + "loss": 0.1174, + "step": 17931 + }, + { + "epoch": 0.94, + "learning_rate": 4.28049264669752e-09, + "loss": 0.1328, + "step": 17932 + }, + { + "epoch": 0.94, + "learning_rate": 4.273537600144705e-09, + "loss": 0.1479, + "step": 17933 + }, + { + "epoch": 0.94, + "learning_rate": 4.26658814752976e-09, + "loss": 0.1485, + "step": 17934 + }, + { + "epoch": 0.94, + "learning_rate": 4.259644289051256e-09, + "loss": 0.1742, + "step": 17935 + }, + { + "epoch": 0.94, + "learning_rate": 4.252706024907682e-09, + "loss": 0.1436, + "step": 17936 + }, + { + "epoch": 0.94, + "learning_rate": 4.245773355297344e-09, + "loss": 0.1622, + "step": 17937 + }, + { + "epoch": 0.94, + "learning_rate": 4.238846280418418e-09, + "loss": 0.1507, + "step": 17938 + }, + { + "epoch": 0.94, + "learning_rate": 4.231924800468789e-09, + "loss": 0.1086, + "step": 17939 + }, + { + "epoch": 0.94, + "learning_rate": 4.225008915646322e-09, + "loss": 0.1169, + "step": 17940 + }, + { + "epoch": 0.94, + "learning_rate": 4.218098626148681e-09, + "loss": 0.1257, + "step": 17941 + }, + { + "epoch": 0.94, + "learning_rate": 4.2111939321733514e-09, + "loss": 0.1254, + "step": 17942 + }, + { + "epoch": 0.94, + "learning_rate": 4.204294833917688e-09, + "loss": 0.1524, + "step": 17943 + }, + { + "epoch": 0.94, + "learning_rate": 4.197401331578865e-09, + "loss": 0.1277, + "step": 17944 + }, + { + "epoch": 0.94, + "learning_rate": 4.1905134253539035e-09, + "loss": 0.1045, + "step": 17945 + }, + { + "epoch": 0.94, + "learning_rate": 4.1836311154396896e-09, + "loss": 0.1123, + "step": 17946 + }, + { + "epoch": 0.94, + "learning_rate": 4.1767544020328446e-09, + "loss": 0.1551, + "step": 17947 + }, + { + "epoch": 0.94, + "learning_rate": 4.169883285329967e-09, + "loss": 0.1751, + "step": 17948 + }, + { + "epoch": 0.94, + "learning_rate": 4.163017765527432e-09, + "loss": 0.1794, + "step": 17949 + }, + { + "epoch": 0.94, + "learning_rate": 4.15615784282144e-09, + "loss": 0.1353, + "step": 17950 + }, + { + "epoch": 0.94, + "learning_rate": 4.149303517408054e-09, + "loss": 0.1671, + "step": 17951 + }, + { + "epoch": 0.94, + "learning_rate": 4.142454789483185e-09, + "loss": 0.1127, + "step": 17952 + }, + { + "epoch": 0.94, + "learning_rate": 4.135611659242566e-09, + "loss": 0.1569, + "step": 17953 + }, + { + "epoch": 0.94, + "learning_rate": 4.128774126881796e-09, + "loss": 0.154, + "step": 17954 + }, + { + "epoch": 0.94, + "learning_rate": 4.121942192596251e-09, + "loss": 0.1394, + "step": 17955 + }, + { + "epoch": 0.94, + "learning_rate": 4.115115856581241e-09, + "loss": 0.1584, + "step": 17956 + }, + { + "epoch": 0.94, + "learning_rate": 4.108295119031813e-09, + "loss": 0.1381, + "step": 17957 + }, + { + "epoch": 0.94, + "learning_rate": 4.101479980142919e-09, + "loss": 0.095, + "step": 17958 + }, + { + "epoch": 0.94, + "learning_rate": 4.094670440109382e-09, + "loss": 0.1452, + "step": 17959 + }, + { + "epoch": 0.94, + "learning_rate": 4.0878664991257805e-09, + "loss": 0.1177, + "step": 17960 + }, + { + "epoch": 0.94, + "learning_rate": 4.081068157386603e-09, + "loss": 0.1114, + "step": 17961 + }, + { + "epoch": 0.94, + "learning_rate": 4.074275415086093e-09, + "loss": 0.1375, + "step": 17962 + }, + { + "epoch": 0.94, + "learning_rate": 4.067488272418473e-09, + "loss": 0.1127, + "step": 17963 + }, + { + "epoch": 0.94, + "learning_rate": 4.060706729577657e-09, + "loss": 0.1112, + "step": 17964 + }, + { + "epoch": 0.94, + "learning_rate": 4.053930786757443e-09, + "loss": 0.1158, + "step": 17965 + }, + { + "epoch": 0.94, + "learning_rate": 4.047160444151565e-09, + "loss": 0.1335, + "step": 17966 + }, + { + "epoch": 0.94, + "learning_rate": 4.040395701953492e-09, + "loss": 0.1069, + "step": 17967 + }, + { + "epoch": 0.94, + "learning_rate": 4.033636560356557e-09, + "loss": 0.1232, + "step": 17968 + }, + { + "epoch": 0.94, + "learning_rate": 4.026883019553917e-09, + "loss": 0.1495, + "step": 17969 + }, + { + "epoch": 0.94, + "learning_rate": 4.02013507973864e-09, + "loss": 0.1424, + "step": 17970 + }, + { + "epoch": 0.94, + "learning_rate": 4.01339274110355e-09, + "loss": 0.122, + "step": 17971 + }, + { + "epoch": 0.94, + "learning_rate": 4.006656003841335e-09, + "loss": 0.1436, + "step": 17972 + }, + { + "epoch": 0.94, + "learning_rate": 3.999924868144578e-09, + "loss": 0.1509, + "step": 17973 + }, + { + "epoch": 0.94, + "learning_rate": 3.993199334205633e-09, + "loss": 0.1302, + "step": 17974 + }, + { + "epoch": 0.94, + "learning_rate": 3.9864794022167025e-09, + "loss": 0.1121, + "step": 17975 + }, + { + "epoch": 0.94, + "learning_rate": 3.979765072369856e-09, + "loss": 0.1629, + "step": 17976 + }, + { + "epoch": 0.94, + "learning_rate": 3.973056344857006e-09, + "loss": 0.1567, + "step": 17977 + }, + { + "epoch": 0.94, + "learning_rate": 3.966353219869867e-09, + "loss": 0.1242, + "step": 17978 + }, + { + "epoch": 0.94, + "learning_rate": 3.959655697600062e-09, + "loss": 0.1195, + "step": 17979 + }, + { + "epoch": 0.94, + "learning_rate": 3.952963778238949e-09, + "loss": 0.1305, + "step": 17980 + }, + { + "epoch": 0.94, + "learning_rate": 3.94627746197782e-09, + "loss": 0.0981, + "step": 17981 + }, + { + "epoch": 0.94, + "learning_rate": 3.939596749007768e-09, + "loss": 0.1542, + "step": 17982 + }, + { + "epoch": 0.94, + "learning_rate": 3.932921639519704e-09, + "loss": 0.1744, + "step": 17983 + }, + { + "epoch": 0.94, + "learning_rate": 3.926252133704455e-09, + "loss": 0.1198, + "step": 17984 + }, + { + "epoch": 0.94, + "learning_rate": 3.919588231752602e-09, + "loss": 0.1563, + "step": 17985 + }, + { + "epoch": 0.94, + "learning_rate": 3.912929933854614e-09, + "loss": 0.1826, + "step": 17986 + }, + { + "epoch": 0.94, + "learning_rate": 3.906277240200784e-09, + "loss": 0.1171, + "step": 17987 + }, + { + "epoch": 0.94, + "learning_rate": 3.899630150981248e-09, + "loss": 0.1694, + "step": 17988 + }, + { + "epoch": 0.94, + "learning_rate": 3.89298866638601e-09, + "loss": 0.158, + "step": 17989 + }, + { + "epoch": 0.94, + "learning_rate": 3.88635278660483e-09, + "loss": 0.133, + "step": 17990 + }, + { + "epoch": 0.94, + "learning_rate": 3.879722511827399e-09, + "loss": 0.1702, + "step": 17991 + }, + { + "epoch": 0.94, + "learning_rate": 3.873097842243234e-09, + "loss": 0.1576, + "step": 17992 + }, + { + "epoch": 0.94, + "learning_rate": 3.866478778041626e-09, + "loss": 0.127, + "step": 17993 + }, + { + "epoch": 0.94, + "learning_rate": 3.859865319411737e-09, + "loss": 0.1403, + "step": 17994 + }, + { + "epoch": 0.94, + "learning_rate": 3.8532574665426585e-09, + "loss": 0.1182, + "step": 17995 + }, + { + "epoch": 0.94, + "learning_rate": 3.846655219623196e-09, + "loss": 0.1365, + "step": 17996 + }, + { + "epoch": 0.94, + "learning_rate": 3.840058578842042e-09, + "loss": 0.1421, + "step": 17997 + }, + { + "epoch": 0.94, + "learning_rate": 3.833467544387759e-09, + "loss": 0.1064, + "step": 17998 + }, + { + "epoch": 0.94, + "learning_rate": 3.826882116448682e-09, + "loss": 0.092, + "step": 17999 + }, + { + "epoch": 0.94, + "learning_rate": 3.8203022952130405e-09, + "loss": 0.0952, + "step": 18000 + }, + { + "epoch": 0.94, + "learning_rate": 3.813728080868883e-09, + "loss": 0.1397, + "step": 18001 + }, + { + "epoch": 0.94, + "learning_rate": 3.807159473604149e-09, + "loss": 0.1457, + "step": 18002 + }, + { + "epoch": 0.94, + "learning_rate": 3.80059647360651e-09, + "loss": 0.1438, + "step": 18003 + }, + { + "epoch": 0.94, + "learning_rate": 3.794039081063572e-09, + "loss": 0.1242, + "step": 18004 + }, + { + "epoch": 0.94, + "learning_rate": 3.787487296162717e-09, + "loss": 0.133, + "step": 18005 + }, + { + "epoch": 0.94, + "learning_rate": 3.7809411190912635e-09, + "loss": 0.17, + "step": 18006 + }, + { + "epoch": 0.94, + "learning_rate": 3.774400550036239e-09, + "loss": 0.1633, + "step": 18007 + }, + { + "epoch": 0.94, + "learning_rate": 3.7678655891845825e-09, + "loss": 0.1325, + "step": 18008 + }, + { + "epoch": 0.94, + "learning_rate": 3.7613362367231e-09, + "loss": 0.138, + "step": 18009 + }, + { + "epoch": 0.94, + "learning_rate": 3.754812492838377e-09, + "loss": 0.1602, + "step": 18010 + }, + { + "epoch": 0.94, + "learning_rate": 3.748294357716886e-09, + "loss": 0.1127, + "step": 18011 + }, + { + "epoch": 0.94, + "learning_rate": 3.741781831544877e-09, + "loss": 0.1158, + "step": 18012 + }, + { + "epoch": 0.94, + "learning_rate": 3.735274914508513e-09, + "loss": 0.1535, + "step": 18013 + }, + { + "epoch": 0.94, + "learning_rate": 3.728773606793778e-09, + "loss": 0.1509, + "step": 18014 + }, + { + "epoch": 0.94, + "learning_rate": 3.7222779085864576e-09, + "loss": 0.1475, + "step": 18015 + }, + { + "epoch": 0.94, + "learning_rate": 3.7157878200721805e-09, + "loss": 0.1187, + "step": 18016 + }, + { + "epoch": 0.94, + "learning_rate": 3.7093033414364872e-09, + "loss": 0.1352, + "step": 18017 + }, + { + "epoch": 0.94, + "learning_rate": 3.7028244728646742e-09, + "loss": 0.1186, + "step": 18018 + }, + { + "epoch": 0.94, + "learning_rate": 3.696351214541904e-09, + "loss": 0.1591, + "step": 18019 + }, + { + "epoch": 0.94, + "learning_rate": 3.689883566653207e-09, + "loss": 0.1675, + "step": 18020 + }, + { + "epoch": 0.94, + "learning_rate": 3.6834215293834126e-09, + "loss": 0.0984, + "step": 18021 + }, + { + "epoch": 0.94, + "learning_rate": 3.676965102917218e-09, + "loss": 0.1192, + "step": 18022 + }, + { + "epoch": 0.94, + "learning_rate": 3.6705142874391637e-09, + "loss": 0.1287, + "step": 18023 + }, + { + "epoch": 0.94, + "learning_rate": 3.664069083133592e-09, + "loss": 0.1077, + "step": 18024 + }, + { + "epoch": 0.94, + "learning_rate": 3.6576294901847325e-09, + "loss": 0.1216, + "step": 18025 + }, + { + "epoch": 0.94, + "learning_rate": 3.651195508776572e-09, + "loss": 0.1574, + "step": 18026 + }, + { + "epoch": 0.94, + "learning_rate": 3.6447671390930747e-09, + "loss": 0.1405, + "step": 18027 + }, + { + "epoch": 0.94, + "learning_rate": 3.638344381317937e-09, + "loss": 0.1229, + "step": 18028 + }, + { + "epoch": 0.94, + "learning_rate": 3.6319272356347243e-09, + "loss": 0.1471, + "step": 18029 + }, + { + "epoch": 0.94, + "learning_rate": 3.6255157022268223e-09, + "loss": 0.1067, + "step": 18030 + }, + { + "epoch": 0.94, + "learning_rate": 3.619109781277485e-09, + "loss": 0.1677, + "step": 18031 + }, + { + "epoch": 0.94, + "learning_rate": 3.6127094729698327e-09, + "loss": 0.132, + "step": 18032 + }, + { + "epoch": 0.94, + "learning_rate": 3.6063147774867187e-09, + "loss": 0.1414, + "step": 18033 + }, + { + "epoch": 0.94, + "learning_rate": 3.5999256950109746e-09, + "loss": 0.1134, + "step": 18034 + }, + { + "epoch": 0.94, + "learning_rate": 3.593542225725166e-09, + "loss": 0.1118, + "step": 18035 + }, + { + "epoch": 0.94, + "learning_rate": 3.587164369811724e-09, + "loss": 0.1361, + "step": 18036 + }, + { + "epoch": 0.94, + "learning_rate": 3.5807921274529473e-09, + "loss": 0.1292, + "step": 18037 + }, + { + "epoch": 0.94, + "learning_rate": 3.57442549883098e-09, + "loss": 0.1286, + "step": 18038 + }, + { + "epoch": 0.94, + "learning_rate": 3.568064484127742e-09, + "loss": 0.1389, + "step": 18039 + }, + { + "epoch": 0.94, + "learning_rate": 3.5617090835250663e-09, + "loss": 0.1179, + "step": 18040 + }, + { + "epoch": 0.94, + "learning_rate": 3.555359297204563e-09, + "loss": 0.1028, + "step": 18041 + }, + { + "epoch": 0.94, + "learning_rate": 3.5490151253477317e-09, + "loss": 0.121, + "step": 18042 + }, + { + "epoch": 0.94, + "learning_rate": 3.5426765681358937e-09, + "loss": 0.1126, + "step": 18043 + }, + { + "epoch": 0.94, + "learning_rate": 3.536343625750171e-09, + "loss": 0.1281, + "step": 18044 + }, + { + "epoch": 0.94, + "learning_rate": 3.5300162983716187e-09, + "loss": 0.1069, + "step": 18045 + }, + { + "epoch": 0.94, + "learning_rate": 3.5236945861810252e-09, + "loss": 0.1066, + "step": 18046 + }, + { + "epoch": 0.94, + "learning_rate": 3.517378489359113e-09, + "loss": 0.1217, + "step": 18047 + }, + { + "epoch": 0.94, + "learning_rate": 3.5110680080863375e-09, + "loss": 0.109, + "step": 18048 + }, + { + "epoch": 0.94, + "learning_rate": 3.504763142543132e-09, + "loss": 0.1308, + "step": 18049 + }, + { + "epoch": 0.94, + "learning_rate": 3.4984638929096422e-09, + "loss": 0.1466, + "step": 18050 + }, + { + "epoch": 0.94, + "learning_rate": 3.492170259365901e-09, + "loss": 0.1208, + "step": 18051 + }, + { + "epoch": 0.94, + "learning_rate": 3.4858822420917868e-09, + "loss": 0.1858, + "step": 18052 + }, + { + "epoch": 0.94, + "learning_rate": 3.4795998412670223e-09, + "loss": 0.1445, + "step": 18053 + }, + { + "epoch": 0.94, + "learning_rate": 3.4733230570711757e-09, + "loss": 0.1614, + "step": 18054 + }, + { + "epoch": 0.94, + "learning_rate": 3.467051889683592e-09, + "loss": 0.1378, + "step": 18055 + }, + { + "epoch": 0.94, + "learning_rate": 3.46078633928355e-09, + "loss": 0.1009, + "step": 18056 + }, + { + "epoch": 0.94, + "learning_rate": 3.4545264060501067e-09, + "loss": 0.1121, + "step": 18057 + }, + { + "epoch": 0.94, + "learning_rate": 3.4482720901622074e-09, + "loss": 0.1594, + "step": 18058 + }, + { + "epoch": 0.94, + "learning_rate": 3.44202339179851e-09, + "loss": 0.1535, + "step": 18059 + }, + { + "epoch": 0.94, + "learning_rate": 3.4357803111376925e-09, + "loss": 0.1239, + "step": 18060 + }, + { + "epoch": 0.94, + "learning_rate": 3.429542848358169e-09, + "loss": 0.1158, + "step": 18061 + }, + { + "epoch": 0.94, + "learning_rate": 3.423311003638174e-09, + "loss": 0.1562, + "step": 18062 + }, + { + "epoch": 0.94, + "learning_rate": 3.417084777155832e-09, + "loss": 0.1736, + "step": 18063 + }, + { + "epoch": 0.94, + "learning_rate": 3.4108641690891117e-09, + "loss": 0.1431, + "step": 18064 + }, + { + "epoch": 0.94, + "learning_rate": 3.4046491796157596e-09, + "loss": 0.1136, + "step": 18065 + }, + { + "epoch": 0.94, + "learning_rate": 3.398439808913456e-09, + "loss": 0.1702, + "step": 18066 + }, + { + "epoch": 0.94, + "learning_rate": 3.3922360571596364e-09, + "loss": 0.1079, + "step": 18067 + }, + { + "epoch": 0.94, + "learning_rate": 3.3860379245316038e-09, + "loss": 0.1461, + "step": 18068 + }, + { + "epoch": 0.94, + "learning_rate": 3.3798454112065055e-09, + "loss": 0.1611, + "step": 18069 + }, + { + "epoch": 0.94, + "learning_rate": 3.373658517361311e-09, + "loss": 0.1246, + "step": 18070 + }, + { + "epoch": 0.94, + "learning_rate": 3.367477243172878e-09, + "loss": 0.1279, + "step": 18071 + }, + { + "epoch": 0.94, + "learning_rate": 3.361301588817844e-09, + "loss": 0.1256, + "step": 18072 + }, + { + "epoch": 0.94, + "learning_rate": 3.355131554472734e-09, + "loss": 0.141, + "step": 18073 + }, + { + "epoch": 0.94, + "learning_rate": 3.348967140313874e-09, + "loss": 0.1275, + "step": 18074 + }, + { + "epoch": 0.94, + "learning_rate": 3.342808346517456e-09, + "loss": 0.1425, + "step": 18075 + }, + { + "epoch": 0.94, + "learning_rate": 3.336655173259495e-09, + "loss": 0.1418, + "step": 18076 + }, + { + "epoch": 0.94, + "learning_rate": 3.3305076207158058e-09, + "loss": 0.1304, + "step": 18077 + }, + { + "epoch": 0.94, + "learning_rate": 3.324365689062181e-09, + "loss": 0.1436, + "step": 18078 + }, + { + "epoch": 0.94, + "learning_rate": 3.31822937847408e-09, + "loss": 0.1093, + "step": 18079 + }, + { + "epoch": 0.94, + "learning_rate": 3.3120986891269186e-09, + "loss": 0.1917, + "step": 18080 + }, + { + "epoch": 0.94, + "learning_rate": 3.305973621195912e-09, + "loss": 0.1187, + "step": 18081 + }, + { + "epoch": 0.94, + "learning_rate": 3.2998541748560983e-09, + "loss": 0.1496, + "step": 18082 + }, + { + "epoch": 0.94, + "learning_rate": 3.2937403502824035e-09, + "loss": 0.1475, + "step": 18083 + }, + { + "epoch": 0.94, + "learning_rate": 3.287632147649533e-09, + "loss": 0.1392, + "step": 18084 + }, + { + "epoch": 0.94, + "learning_rate": 3.28152956713208e-09, + "loss": 0.1406, + "step": 18085 + }, + { + "epoch": 0.94, + "learning_rate": 3.2754326089044605e-09, + "loss": 0.1085, + "step": 18086 + }, + { + "epoch": 0.94, + "learning_rate": 3.269341273140913e-09, + "loss": 0.1507, + "step": 18087 + }, + { + "epoch": 0.94, + "learning_rate": 3.263255560015521e-09, + "loss": 0.1149, + "step": 18088 + }, + { + "epoch": 0.94, + "learning_rate": 3.2571754697022336e-09, + "loss": 0.1299, + "step": 18089 + }, + { + "epoch": 0.94, + "learning_rate": 3.2511010023748454e-09, + "loss": 0.1141, + "step": 18090 + }, + { + "epoch": 0.94, + "learning_rate": 3.2450321582069285e-09, + "loss": 0.1311, + "step": 18091 + }, + { + "epoch": 0.94, + "learning_rate": 3.2389689373719664e-09, + "loss": 0.135, + "step": 18092 + }, + { + "epoch": 0.94, + "learning_rate": 3.2329113400432205e-09, + "loss": 0.1344, + "step": 18093 + }, + { + "epoch": 0.94, + "learning_rate": 3.2268593663938193e-09, + "loss": 0.1728, + "step": 18094 + }, + { + "epoch": 0.94, + "learning_rate": 3.220813016596735e-09, + "loss": 0.1332, + "step": 18095 + }, + { + "epoch": 0.94, + "learning_rate": 3.2147722908247852e-09, + "loss": 0.1152, + "step": 18096 + }, + { + "epoch": 0.94, + "learning_rate": 3.20873718925061e-09, + "loss": 0.1302, + "step": 18097 + }, + { + "epoch": 0.94, + "learning_rate": 3.202707712046693e-09, + "loss": 0.1186, + "step": 18098 + }, + { + "epoch": 0.94, + "learning_rate": 3.196683859385363e-09, + "loss": 0.1449, + "step": 18099 + }, + { + "epoch": 0.94, + "learning_rate": 3.1906656314387937e-09, + "loss": 0.1542, + "step": 18100 + }, + { + "epoch": 0.94, + "learning_rate": 3.1846530283789806e-09, + "loss": 0.139, + "step": 18101 + }, + { + "epoch": 0.94, + "learning_rate": 3.17864605037772e-09, + "loss": 0.1217, + "step": 18102 + }, + { + "epoch": 0.94, + "learning_rate": 3.172644697606763e-09, + "loss": 0.0945, + "step": 18103 + }, + { + "epoch": 0.94, + "learning_rate": 3.1666489702376177e-09, + "loss": 0.1342, + "step": 18104 + }, + { + "epoch": 0.94, + "learning_rate": 3.1606588684416123e-09, + "loss": 0.1101, + "step": 18105 + }, + { + "epoch": 0.95, + "learning_rate": 3.1546743923899665e-09, + "loss": 0.1536, + "step": 18106 + }, + { + "epoch": 0.95, + "learning_rate": 3.148695542253721e-09, + "loss": 0.1608, + "step": 18107 + }, + { + "epoch": 0.95, + "learning_rate": 3.142722318203739e-09, + "loss": 0.1556, + "step": 18108 + }, + { + "epoch": 0.95, + "learning_rate": 3.1367547204107724e-09, + "loss": 0.1192, + "step": 18109 + }, + { + "epoch": 0.95, + "learning_rate": 3.13079274904533e-09, + "loss": 0.157, + "step": 18110 + }, + { + "epoch": 0.95, + "learning_rate": 3.124836404277853e-09, + "loss": 0.164, + "step": 18111 + }, + { + "epoch": 0.95, + "learning_rate": 3.1188856862785386e-09, + "loss": 0.1274, + "step": 18112 + }, + { + "epoch": 0.95, + "learning_rate": 3.11294059521745e-09, + "loss": 0.1295, + "step": 18113 + }, + { + "epoch": 0.95, + "learning_rate": 3.1070011312645416e-09, + "loss": 0.1387, + "step": 18114 + }, + { + "epoch": 0.95, + "learning_rate": 3.1010672945895654e-09, + "loss": 0.1119, + "step": 18115 + }, + { + "epoch": 0.95, + "learning_rate": 3.095139085362097e-09, + "loss": 0.1337, + "step": 18116 + }, + { + "epoch": 0.95, + "learning_rate": 3.089216503751535e-09, + "loss": 0.1344, + "step": 18117 + }, + { + "epoch": 0.95, + "learning_rate": 3.083299549927232e-09, + "loss": 0.1641, + "step": 18118 + }, + { + "epoch": 0.95, + "learning_rate": 3.0773882240582083e-09, + "loss": 0.1135, + "step": 18119 + }, + { + "epoch": 0.95, + "learning_rate": 3.07148252631344e-09, + "loss": 0.1049, + "step": 18120 + }, + { + "epoch": 0.95, + "learning_rate": 3.065582456861748e-09, + "loss": 0.1364, + "step": 18121 + }, + { + "epoch": 0.95, + "learning_rate": 3.059688015871731e-09, + "loss": 0.1466, + "step": 18122 + }, + { + "epoch": 0.95, + "learning_rate": 3.0537992035118533e-09, + "loss": 0.1379, + "step": 18123 + }, + { + "epoch": 0.95, + "learning_rate": 3.0479160199504027e-09, + "loss": 0.134, + "step": 18124 + }, + { + "epoch": 0.95, + "learning_rate": 3.0420384653555566e-09, + "loss": 0.1518, + "step": 18125 + }, + { + "epoch": 0.95, + "learning_rate": 3.036166539895313e-09, + "loss": 0.1413, + "step": 18126 + }, + { + "epoch": 0.95, + "learning_rate": 3.030300243737427e-09, + "loss": 0.12, + "step": 18127 + }, + { + "epoch": 0.95, + "learning_rate": 3.024439577049609e-09, + "loss": 0.1049, + "step": 18128 + }, + { + "epoch": 0.95, + "learning_rate": 3.0185845399993694e-09, + "loss": 0.1003, + "step": 18129 + }, + { + "epoch": 0.95, + "learning_rate": 3.0127351327539962e-09, + "loss": 0.1441, + "step": 18130 + }, + { + "epoch": 0.95, + "learning_rate": 3.006891355480712e-09, + "loss": 0.1452, + "step": 18131 + }, + { + "epoch": 0.95, + "learning_rate": 3.0010532083465157e-09, + "loss": 0.1254, + "step": 18132 + }, + { + "epoch": 0.95, + "learning_rate": 2.995220691518274e-09, + "loss": 0.1171, + "step": 18133 + }, + { + "epoch": 0.95, + "learning_rate": 2.9893938051626767e-09, + "loss": 0.1426, + "step": 18134 + }, + { + "epoch": 0.95, + "learning_rate": 2.983572549446256e-09, + "loss": 0.1156, + "step": 18135 + }, + { + "epoch": 0.95, + "learning_rate": 2.9777569245353906e-09, + "loss": 0.0965, + "step": 18136 + }, + { + "epoch": 0.95, + "learning_rate": 2.9719469305962806e-09, + "loss": 0.1218, + "step": 18137 + }, + { + "epoch": 0.95, + "learning_rate": 2.9661425677949713e-09, + "loss": 0.1063, + "step": 18138 + }, + { + "epoch": 0.95, + "learning_rate": 2.9603438362973964e-09, + "loss": 0.1254, + "step": 18139 + }, + { + "epoch": 0.95, + "learning_rate": 2.9545507362692677e-09, + "loss": 0.1258, + "step": 18140 + }, + { + "epoch": 0.95, + "learning_rate": 2.94876326787612e-09, + "loss": 0.1091, + "step": 18141 + }, + { + "epoch": 0.95, + "learning_rate": 2.9429814312833977e-09, + "loss": 0.1766, + "step": 18142 + }, + { + "epoch": 0.95, + "learning_rate": 2.9372052266563475e-09, + "loss": 0.1585, + "step": 18143 + }, + { + "epoch": 0.95, + "learning_rate": 2.9314346541600587e-09, + "loss": 0.1665, + "step": 18144 + }, + { + "epoch": 0.95, + "learning_rate": 2.9256697139594e-09, + "loss": 0.1374, + "step": 18145 + }, + { + "epoch": 0.95, + "learning_rate": 2.9199104062192172e-09, + "loss": 0.136, + "step": 18146 + }, + { + "epoch": 0.95, + "learning_rate": 2.9141567311040672e-09, + "loss": 0.111, + "step": 18147 + }, + { + "epoch": 0.95, + "learning_rate": 2.9084086887783966e-09, + "loss": 0.1226, + "step": 18148 + }, + { + "epoch": 0.95, + "learning_rate": 2.9026662794064738e-09, + "loss": 0.1684, + "step": 18149 + }, + { + "epoch": 0.95, + "learning_rate": 2.8969295031524565e-09, + "loss": 0.1307, + "step": 18150 + }, + { + "epoch": 0.95, + "learning_rate": 2.89119836018028e-09, + "loss": 0.1227, + "step": 18151 + }, + { + "epoch": 0.95, + "learning_rate": 2.8854728506537474e-09, + "loss": 0.1104, + "step": 18152 + }, + { + "epoch": 0.95, + "learning_rate": 2.8797529747364824e-09, + "loss": 0.1613, + "step": 18153 + }, + { + "epoch": 0.95, + "learning_rate": 2.8740387325919767e-09, + "loss": 0.1522, + "step": 18154 + }, + { + "epoch": 0.95, + "learning_rate": 2.8683301243835666e-09, + "loss": 0.1528, + "step": 18155 + }, + { + "epoch": 0.95, + "learning_rate": 2.862627150274344e-09, + "loss": 0.113, + "step": 18156 + }, + { + "epoch": 0.95, + "learning_rate": 2.856929810427355e-09, + "loss": 0.1136, + "step": 18157 + }, + { + "epoch": 0.95, + "learning_rate": 2.851238105005427e-09, + "loss": 0.1096, + "step": 18158 + }, + { + "epoch": 0.95, + "learning_rate": 2.845552034171228e-09, + "loss": 0.1422, + "step": 18159 + }, + { + "epoch": 0.95, + "learning_rate": 2.8398715980872513e-09, + "loss": 0.1326, + "step": 18160 + }, + { + "epoch": 0.95, + "learning_rate": 2.8341967969158552e-09, + "loss": 0.1239, + "step": 18161 + }, + { + "epoch": 0.95, + "learning_rate": 2.828527630819222e-09, + "loss": 0.1467, + "step": 18162 + }, + { + "epoch": 0.95, + "learning_rate": 2.822864099959399e-09, + "loss": 0.168, + "step": 18163 + }, + { + "epoch": 0.95, + "learning_rate": 2.8172062044982124e-09, + "loss": 0.1132, + "step": 18164 + }, + { + "epoch": 0.95, + "learning_rate": 2.8115539445974e-09, + "loss": 0.131, + "step": 18165 + }, + { + "epoch": 0.95, + "learning_rate": 2.8059073204185215e-09, + "loss": 0.1481, + "step": 18166 + }, + { + "epoch": 0.95, + "learning_rate": 2.800266332122914e-09, + "loss": 0.1366, + "step": 18167 + }, + { + "epoch": 0.95, + "learning_rate": 2.794630979871826e-09, + "loss": 0.1376, + "step": 18168 + }, + { + "epoch": 0.95, + "learning_rate": 2.7890012638263295e-09, + "loss": 0.1426, + "step": 18169 + }, + { + "epoch": 0.95, + "learning_rate": 2.7833771841472952e-09, + "loss": 0.1528, + "step": 18170 + }, + { + "epoch": 0.95, + "learning_rate": 2.7777587409954617e-09, + "loss": 0.1745, + "step": 18171 + }, + { + "epoch": 0.95, + "learning_rate": 2.7721459345314336e-09, + "loss": 0.1258, + "step": 18172 + }, + { + "epoch": 0.95, + "learning_rate": 2.7665387649156156e-09, + "loss": 0.1362, + "step": 18173 + }, + { + "epoch": 0.95, + "learning_rate": 2.760937232308258e-09, + "loss": 0.1297, + "step": 18174 + }, + { + "epoch": 0.95, + "learning_rate": 2.7553413368694323e-09, + "loss": 0.1299, + "step": 18175 + }, + { + "epoch": 0.95, + "learning_rate": 2.7497510787591217e-09, + "loss": 0.1098, + "step": 18176 + }, + { + "epoch": 0.95, + "learning_rate": 2.7441664581370648e-09, + "loss": 0.1259, + "step": 18177 + }, + { + "epoch": 0.95, + "learning_rate": 2.7385874751628902e-09, + "loss": 0.1587, + "step": 18178 + }, + { + "epoch": 0.95, + "learning_rate": 2.7330141299960252e-09, + "loss": 0.1829, + "step": 18179 + }, + { + "epoch": 0.95, + "learning_rate": 2.727446422795765e-09, + "loss": 0.1151, + "step": 18180 + }, + { + "epoch": 0.95, + "learning_rate": 2.721884353721249e-09, + "loss": 0.1265, + "step": 18181 + }, + { + "epoch": 0.95, + "learning_rate": 2.716327922931416e-09, + "loss": 0.2135, + "step": 18182 + }, + { + "epoch": 0.95, + "learning_rate": 2.710777130585118e-09, + "loss": 0.146, + "step": 18183 + }, + { + "epoch": 0.95, + "learning_rate": 2.7052319768409603e-09, + "loss": 0.1426, + "step": 18184 + }, + { + "epoch": 0.95, + "learning_rate": 2.699692461857439e-09, + "loss": 0.1799, + "step": 18185 + }, + { + "epoch": 0.95, + "learning_rate": 2.6941585857928497e-09, + "loss": 0.1165, + "step": 18186 + }, + { + "epoch": 0.95, + "learning_rate": 2.6886303488053983e-09, + "loss": 0.1334, + "step": 18187 + }, + { + "epoch": 0.95, + "learning_rate": 2.6831077510530708e-09, + "loss": 0.1293, + "step": 18188 + }, + { + "epoch": 0.95, + "learning_rate": 2.6775907926936736e-09, + "loss": 0.123, + "step": 18189 + }, + { + "epoch": 0.95, + "learning_rate": 2.672079473884903e-09, + "loss": 0.1605, + "step": 18190 + }, + { + "epoch": 0.95, + "learning_rate": 2.6665737947842993e-09, + "loss": 0.1296, + "step": 18191 + }, + { + "epoch": 0.95, + "learning_rate": 2.6610737555491814e-09, + "loss": 0.1432, + "step": 18192 + }, + { + "epoch": 0.95, + "learning_rate": 2.6555793563367344e-09, + "loss": 0.1206, + "step": 18193 + }, + { + "epoch": 0.95, + "learning_rate": 2.6500905973040555e-09, + "loss": 0.1423, + "step": 18194 + }, + { + "epoch": 0.95, + "learning_rate": 2.6446074786079518e-09, + "loss": 0.1212, + "step": 18195 + }, + { + "epoch": 0.95, + "learning_rate": 2.6391300004051433e-09, + "loss": 0.127, + "step": 18196 + }, + { + "epoch": 0.95, + "learning_rate": 2.6336581628522146e-09, + "loss": 0.0989, + "step": 18197 + }, + { + "epoch": 0.95, + "learning_rate": 2.6281919661055084e-09, + "loss": 0.1184, + "step": 18198 + }, + { + "epoch": 0.95, + "learning_rate": 2.6227314103212773e-09, + "loss": 0.1241, + "step": 18199 + }, + { + "epoch": 0.95, + "learning_rate": 2.6172764956555515e-09, + "loss": 0.1307, + "step": 18200 + }, + { + "epoch": 0.95, + "learning_rate": 2.6118272222642958e-09, + "loss": 0.1159, + "step": 18201 + }, + { + "epoch": 0.95, + "learning_rate": 2.606383590303185e-09, + "loss": 0.1385, + "step": 18202 + }, + { + "epoch": 0.95, + "learning_rate": 2.6009455999278728e-09, + "loss": 0.1595, + "step": 18203 + }, + { + "epoch": 0.95, + "learning_rate": 2.595513251293702e-09, + "loss": 0.1164, + "step": 18204 + }, + { + "epoch": 0.95, + "learning_rate": 2.5900865445559917e-09, + "loss": 0.1179, + "step": 18205 + }, + { + "epoch": 0.95, + "learning_rate": 2.5846654798698188e-09, + "loss": 0.1154, + "step": 18206 + }, + { + "epoch": 0.95, + "learning_rate": 2.579250057390081e-09, + "loss": 0.1338, + "step": 18207 + }, + { + "epoch": 0.95, + "learning_rate": 2.5738402772716104e-09, + "loss": 0.1142, + "step": 18208 + }, + { + "epoch": 0.95, + "learning_rate": 2.568436139669017e-09, + "loss": 0.1364, + "step": 18209 + }, + { + "epoch": 0.95, + "learning_rate": 2.5630376447367097e-09, + "loss": 0.126, + "step": 18210 + }, + { + "epoch": 0.95, + "learning_rate": 2.5576447926290103e-09, + "loss": 0.1242, + "step": 18211 + }, + { + "epoch": 0.95, + "learning_rate": 2.5522575835000613e-09, + "loss": 0.1226, + "step": 18212 + }, + { + "epoch": 0.95, + "learning_rate": 2.5468760175038074e-09, + "loss": 0.1293, + "step": 18213 + }, + { + "epoch": 0.95, + "learning_rate": 2.541500094794036e-09, + "loss": 0.1053, + "step": 18214 + }, + { + "epoch": 0.95, + "learning_rate": 2.536129815524446e-09, + "loss": 0.1442, + "step": 18215 + }, + { + "epoch": 0.95, + "learning_rate": 2.530765179848471e-09, + "loss": 0.1358, + "step": 18216 + }, + { + "epoch": 0.95, + "learning_rate": 2.5254061879194766e-09, + "loss": 0.1402, + "step": 18217 + }, + { + "epoch": 0.95, + "learning_rate": 2.520052839890585e-09, + "loss": 0.1541, + "step": 18218 + }, + { + "epoch": 0.95, + "learning_rate": 2.514705135914852e-09, + "loss": 0.1691, + "step": 18219 + }, + { + "epoch": 0.95, + "learning_rate": 2.5093630761450656e-09, + "loss": 0.1329, + "step": 18220 + }, + { + "epoch": 0.95, + "learning_rate": 2.504026660733927e-09, + "loss": 0.1404, + "step": 18221 + }, + { + "epoch": 0.95, + "learning_rate": 2.498695889833935e-09, + "loss": 0.1201, + "step": 18222 + }, + { + "epoch": 0.95, + "learning_rate": 2.493370763597458e-09, + "loss": 0.1001, + "step": 18223 + }, + { + "epoch": 0.95, + "learning_rate": 2.488051282176706e-09, + "loss": 0.1492, + "step": 18224 + }, + { + "epoch": 0.95, + "learning_rate": 2.4827374457236707e-09, + "loss": 0.1075, + "step": 18225 + }, + { + "epoch": 0.95, + "learning_rate": 2.477429254390273e-09, + "loss": 0.1668, + "step": 18226 + }, + { + "epoch": 0.95, + "learning_rate": 2.4721267083281926e-09, + "loss": 0.1568, + "step": 18227 + }, + { + "epoch": 0.95, + "learning_rate": 2.466829807688975e-09, + "loss": 0.1259, + "step": 18228 + }, + { + "epoch": 0.95, + "learning_rate": 2.4615385526240317e-09, + "loss": 0.12, + "step": 18229 + }, + { + "epoch": 0.95, + "learning_rate": 2.4562529432845534e-09, + "loss": 0.1082, + "step": 18230 + }, + { + "epoch": 0.95, + "learning_rate": 2.4509729798216416e-09, + "loss": 0.1295, + "step": 18231 + }, + { + "epoch": 0.95, + "learning_rate": 2.445698662386175e-09, + "loss": 0.1083, + "step": 18232 + }, + { + "epoch": 0.95, + "learning_rate": 2.4404299911289007e-09, + "loss": 0.1332, + "step": 18233 + }, + { + "epoch": 0.95, + "learning_rate": 2.4351669662004082e-09, + "loss": 0.1601, + "step": 18234 + }, + { + "epoch": 0.95, + "learning_rate": 2.4299095877511333e-09, + "loss": 0.1354, + "step": 18235 + }, + { + "epoch": 0.95, + "learning_rate": 2.424657855931267e-09, + "loss": 0.1388, + "step": 18236 + }, + { + "epoch": 0.95, + "learning_rate": 2.419411770891e-09, + "loss": 0.1421, + "step": 18237 + }, + { + "epoch": 0.95, + "learning_rate": 2.4141713327801905e-09, + "loss": 0.1072, + "step": 18238 + }, + { + "epoch": 0.95, + "learning_rate": 2.4089365417486297e-09, + "loss": 0.1179, + "step": 18239 + }, + { + "epoch": 0.95, + "learning_rate": 2.4037073979459755e-09, + "loss": 0.1399, + "step": 18240 + }, + { + "epoch": 0.95, + "learning_rate": 2.3984839015216196e-09, + "loss": 0.155, + "step": 18241 + }, + { + "epoch": 0.95, + "learning_rate": 2.393266052624887e-09, + "loss": 0.155, + "step": 18242 + }, + { + "epoch": 0.95, + "learning_rate": 2.3880538514048586e-09, + "loss": 0.1385, + "step": 18243 + }, + { + "epoch": 0.95, + "learning_rate": 2.3828472980105705e-09, + "loss": 0.1422, + "step": 18244 + }, + { + "epoch": 0.95, + "learning_rate": 2.377646392590793e-09, + "loss": 0.1523, + "step": 18245 + }, + { + "epoch": 0.95, + "learning_rate": 2.372451135294162e-09, + "loss": 0.1271, + "step": 18246 + }, + { + "epoch": 0.95, + "learning_rate": 2.367261526269182e-09, + "loss": 0.1505, + "step": 18247 + }, + { + "epoch": 0.95, + "learning_rate": 2.3620775656641557e-09, + "loss": 0.1093, + "step": 18248 + }, + { + "epoch": 0.95, + "learning_rate": 2.3568992536272537e-09, + "loss": 0.1276, + "step": 18249 + }, + { + "epoch": 0.95, + "learning_rate": 2.3517265903064464e-09, + "loss": 0.1587, + "step": 18250 + }, + { + "epoch": 0.95, + "learning_rate": 2.346559575849616e-09, + "loss": 0.14, + "step": 18251 + }, + { + "epoch": 0.95, + "learning_rate": 2.3413982104044217e-09, + "loss": 0.1265, + "step": 18252 + }, + { + "epoch": 0.95, + "learning_rate": 2.3362424941183457e-09, + "loss": 0.123, + "step": 18253 + }, + { + "epoch": 0.95, + "learning_rate": 2.331092427138781e-09, + "loss": 0.1452, + "step": 18254 + }, + { + "epoch": 0.95, + "learning_rate": 2.325948009612899e-09, + "loss": 0.1347, + "step": 18255 + }, + { + "epoch": 0.95, + "learning_rate": 2.320809241687738e-09, + "loss": 0.1298, + "step": 18256 + }, + { + "epoch": 0.95, + "learning_rate": 2.3156761235101352e-09, + "loss": 0.1333, + "step": 18257 + }, + { + "epoch": 0.95, + "learning_rate": 2.3105486552268625e-09, + "loss": 0.1302, + "step": 18258 + }, + { + "epoch": 0.95, + "learning_rate": 2.305426836984403e-09, + "loss": 0.1593, + "step": 18259 + }, + { + "epoch": 0.95, + "learning_rate": 2.3003106689291508e-09, + "loss": 0.1221, + "step": 18260 + }, + { + "epoch": 0.95, + "learning_rate": 2.295200151207344e-09, + "loss": 0.1341, + "step": 18261 + }, + { + "epoch": 0.95, + "learning_rate": 2.2900952839650434e-09, + "loss": 0.0955, + "step": 18262 + }, + { + "epoch": 0.95, + "learning_rate": 2.284996067348177e-09, + "loss": 0.1186, + "step": 18263 + }, + { + "epoch": 0.95, + "learning_rate": 2.2799025015024064e-09, + "loss": 0.1271, + "step": 18264 + }, + { + "epoch": 0.95, + "learning_rate": 2.2748145865733703e-09, + "loss": 0.1403, + "step": 18265 + }, + { + "epoch": 0.95, + "learning_rate": 2.2697323227064413e-09, + "loss": 0.1529, + "step": 18266 + }, + { + "epoch": 0.95, + "learning_rate": 2.2646557100469256e-09, + "loss": 0.1307, + "step": 18267 + }, + { + "epoch": 0.95, + "learning_rate": 2.25958474873984e-09, + "loss": 0.0986, + "step": 18268 + }, + { + "epoch": 0.95, + "learning_rate": 2.25451943893018e-09, + "loss": 0.1273, + "step": 18269 + }, + { + "epoch": 0.95, + "learning_rate": 2.249459780762697e-09, + "loss": 0.1455, + "step": 18270 + }, + { + "epoch": 0.95, + "learning_rate": 2.2444057743820076e-09, + "loss": 0.1267, + "step": 18271 + }, + { + "epoch": 0.95, + "learning_rate": 2.2393574199325303e-09, + "loss": 0.1511, + "step": 18272 + }, + { + "epoch": 0.95, + "learning_rate": 2.2343147175585498e-09, + "loss": 0.1246, + "step": 18273 + }, + { + "epoch": 0.95, + "learning_rate": 2.229277667404217e-09, + "loss": 0.1227, + "step": 18274 + }, + { + "epoch": 0.95, + "learning_rate": 2.224246269613461e-09, + "loss": 0.152, + "step": 18275 + }, + { + "epoch": 0.95, + "learning_rate": 2.2192205243300787e-09, + "loss": 0.1401, + "step": 18276 + }, + { + "epoch": 0.95, + "learning_rate": 2.214200431697755e-09, + "loss": 0.1376, + "step": 18277 + }, + { + "epoch": 0.95, + "learning_rate": 2.2091859918599297e-09, + "loss": 0.1287, + "step": 18278 + }, + { + "epoch": 0.95, + "learning_rate": 2.2041772049599115e-09, + "loss": 0.1578, + "step": 18279 + }, + { + "epoch": 0.95, + "learning_rate": 2.199174071140897e-09, + "loss": 0.142, + "step": 18280 + }, + { + "epoch": 0.95, + "learning_rate": 2.1941765905458154e-09, + "loss": 0.1251, + "step": 18281 + }, + { + "epoch": 0.95, + "learning_rate": 2.189184763317531e-09, + "loss": 0.1173, + "step": 18282 + }, + { + "epoch": 0.95, + "learning_rate": 2.184198589598707e-09, + "loss": 0.139, + "step": 18283 + }, + { + "epoch": 0.95, + "learning_rate": 2.1792180695318742e-09, + "loss": 0.1133, + "step": 18284 + }, + { + "epoch": 0.95, + "learning_rate": 2.1742432032593405e-09, + "loss": 0.1344, + "step": 18285 + }, + { + "epoch": 0.95, + "learning_rate": 2.1692739909233036e-09, + "loss": 0.1274, + "step": 18286 + }, + { + "epoch": 0.95, + "learning_rate": 2.164310432665761e-09, + "loss": 0.1009, + "step": 18287 + }, + { + "epoch": 0.95, + "learning_rate": 2.159352528628644e-09, + "loss": 0.1359, + "step": 18288 + }, + { + "epoch": 0.95, + "learning_rate": 2.154400278953594e-09, + "loss": 0.1477, + "step": 18289 + }, + { + "epoch": 0.95, + "learning_rate": 2.1494536837821433e-09, + "loss": 0.1214, + "step": 18290 + }, + { + "epoch": 0.95, + "learning_rate": 2.1445127432556885e-09, + "loss": 0.1889, + "step": 18291 + }, + { + "epoch": 0.95, + "learning_rate": 2.1395774575154514e-09, + "loss": 0.0936, + "step": 18292 + }, + { + "epoch": 0.95, + "learning_rate": 2.1346478267024738e-09, + "loss": 0.1548, + "step": 18293 + }, + { + "epoch": 0.95, + "learning_rate": 2.129723850957621e-09, + "loss": 0.1532, + "step": 18294 + }, + { + "epoch": 0.95, + "learning_rate": 2.1248055304216693e-09, + "loss": 0.1245, + "step": 18295 + }, + { + "epoch": 0.95, + "learning_rate": 2.119892865235151e-09, + "loss": 0.1291, + "step": 18296 + }, + { + "epoch": 0.96, + "learning_rate": 2.114985855538487e-09, + "loss": 0.1557, + "step": 18297 + }, + { + "epoch": 0.96, + "learning_rate": 2.1100845014719203e-09, + "loss": 0.1534, + "step": 18298 + }, + { + "epoch": 0.96, + "learning_rate": 2.1051888031755392e-09, + "loss": 0.1425, + "step": 18299 + }, + { + "epoch": 0.96, + "learning_rate": 2.100298760789254e-09, + "loss": 0.1377, + "step": 18300 + }, + { + "epoch": 0.96, + "learning_rate": 2.0954143744527973e-09, + "loss": 0.1633, + "step": 18301 + }, + { + "epoch": 0.96, + "learning_rate": 2.0905356443058354e-09, + "loss": 0.1035, + "step": 18302 + }, + { + "epoch": 0.96, + "learning_rate": 2.085662570487745e-09, + "loss": 0.121, + "step": 18303 + }, + { + "epoch": 0.96, + "learning_rate": 2.0807951531378375e-09, + "loss": 0.1329, + "step": 18304 + }, + { + "epoch": 0.96, + "learning_rate": 2.0759333923951795e-09, + "loss": 0.1178, + "step": 18305 + }, + { + "epoch": 0.96, + "learning_rate": 2.0710772883987925e-09, + "loss": 0.137, + "step": 18306 + }, + { + "epoch": 0.96, + "learning_rate": 2.0662268412874105e-09, + "loss": 0.1521, + "step": 18307 + }, + { + "epoch": 0.96, + "learning_rate": 2.061382051199656e-09, + "loss": 0.1528, + "step": 18308 + }, + { + "epoch": 0.96, + "learning_rate": 2.0565429182740403e-09, + "loss": 0.1601, + "step": 18309 + }, + { + "epoch": 0.96, + "learning_rate": 2.05170944264883e-09, + "loss": 0.1665, + "step": 18310 + }, + { + "epoch": 0.96, + "learning_rate": 2.0468816244621824e-09, + "loss": 0.1467, + "step": 18311 + }, + { + "epoch": 0.96, + "learning_rate": 2.0420594638520527e-09, + "loss": 0.1737, + "step": 18312 + }, + { + "epoch": 0.96, + "learning_rate": 2.037242960956309e-09, + "loss": 0.1667, + "step": 18313 + }, + { + "epoch": 0.96, + "learning_rate": 2.0324321159125744e-09, + "loss": 0.1592, + "step": 18314 + }, + { + "epoch": 0.96, + "learning_rate": 2.0276269288583614e-09, + "loss": 0.1139, + "step": 18315 + }, + { + "epoch": 0.96, + "learning_rate": 2.022827399930982e-09, + "loss": 0.1345, + "step": 18316 + }, + { + "epoch": 0.96, + "learning_rate": 2.0180335292676376e-09, + "loss": 0.1945, + "step": 18317 + }, + { + "epoch": 0.96, + "learning_rate": 2.0132453170053077e-09, + "loss": 0.1318, + "step": 18318 + }, + { + "epoch": 0.96, + "learning_rate": 2.0084627632808603e-09, + "loss": 0.1307, + "step": 18319 + }, + { + "epoch": 0.96, + "learning_rate": 2.0036858682309863e-09, + "loss": 0.1345, + "step": 18320 + }, + { + "epoch": 0.96, + "learning_rate": 1.998914631992199e-09, + "loss": 0.1322, + "step": 18321 + }, + { + "epoch": 0.96, + "learning_rate": 1.9941490547008776e-09, + "loss": 0.1329, + "step": 18322 + }, + { + "epoch": 0.96, + "learning_rate": 1.9893891364932023e-09, + "loss": 0.1356, + "step": 18323 + }, + { + "epoch": 0.96, + "learning_rate": 1.9846348775052425e-09, + "loss": 0.1735, + "step": 18324 + }, + { + "epoch": 0.96, + "learning_rate": 1.979886277872844e-09, + "loss": 0.1189, + "step": 18325 + }, + { + "epoch": 0.96, + "learning_rate": 1.9751433377317217e-09, + "loss": 0.1024, + "step": 18326 + }, + { + "epoch": 0.96, + "learning_rate": 1.9704060572174776e-09, + "loss": 0.166, + "step": 18327 + }, + { + "epoch": 0.96, + "learning_rate": 1.96567443646547e-09, + "loss": 0.1415, + "step": 18328 + }, + { + "epoch": 0.96, + "learning_rate": 1.960948475610924e-09, + "loss": 0.1532, + "step": 18329 + }, + { + "epoch": 0.96, + "learning_rate": 1.956228174788932e-09, + "loss": 0.1218, + "step": 18330 + }, + { + "epoch": 0.96, + "learning_rate": 1.951513534134386e-09, + "loss": 0.1275, + "step": 18331 + }, + { + "epoch": 0.96, + "learning_rate": 1.9468045537820664e-09, + "loss": 0.1177, + "step": 18332 + }, + { + "epoch": 0.96, + "learning_rate": 1.9421012338664887e-09, + "loss": 0.1536, + "step": 18333 + }, + { + "epoch": 0.96, + "learning_rate": 1.9374035745221227e-09, + "loss": 0.1092, + "step": 18334 + }, + { + "epoch": 0.96, + "learning_rate": 1.932711575883239e-09, + "loss": 0.1442, + "step": 18335 + }, + { + "epoch": 0.96, + "learning_rate": 1.928025238083908e-09, + "loss": 0.1209, + "step": 18336 + }, + { + "epoch": 0.96, + "learning_rate": 1.9233445612580666e-09, + "loss": 0.1475, + "step": 18337 + }, + { + "epoch": 0.96, + "learning_rate": 1.9186695455395198e-09, + "loss": 0.1148, + "step": 18338 + }, + { + "epoch": 0.96, + "learning_rate": 1.914000191061871e-09, + "loss": 0.1338, + "step": 18339 + }, + { + "epoch": 0.96, + "learning_rate": 1.909336497958569e-09, + "loss": 0.1575, + "step": 18340 + }, + { + "epoch": 0.96, + "learning_rate": 1.9046784663628857e-09, + "loss": 0.1421, + "step": 18341 + }, + { + "epoch": 0.96, + "learning_rate": 1.9000260964079807e-09, + "loss": 0.1279, + "step": 18342 + }, + { + "epoch": 0.96, + "learning_rate": 1.895379388226792e-09, + "loss": 0.1256, + "step": 18343 + }, + { + "epoch": 0.96, + "learning_rate": 1.890738341952125e-09, + "loss": 0.1114, + "step": 18344 + }, + { + "epoch": 0.96, + "learning_rate": 1.8861029577166286e-09, + "loss": 0.1436, + "step": 18345 + }, + { + "epoch": 0.96, + "learning_rate": 1.8814732356528194e-09, + "loss": 0.1186, + "step": 18346 + }, + { + "epoch": 0.96, + "learning_rate": 1.8768491758929695e-09, + "loss": 0.1263, + "step": 18347 + }, + { + "epoch": 0.96, + "learning_rate": 1.8722307785692394e-09, + "loss": 0.0997, + "step": 18348 + }, + { + "epoch": 0.96, + "learning_rate": 1.867618043813679e-09, + "loss": 0.1578, + "step": 18349 + }, + { + "epoch": 0.96, + "learning_rate": 1.863010971758072e-09, + "loss": 0.1226, + "step": 18350 + }, + { + "epoch": 0.96, + "learning_rate": 1.858409562534069e-09, + "loss": 0.1812, + "step": 18351 + }, + { + "epoch": 0.96, + "learning_rate": 1.853813816273231e-09, + "loss": 0.1266, + "step": 18352 + }, + { + "epoch": 0.96, + "learning_rate": 1.8492237331068749e-09, + "loss": 0.1593, + "step": 18353 + }, + { + "epoch": 0.96, + "learning_rate": 1.8446393131662075e-09, + "loss": 0.1312, + "step": 18354 + }, + { + "epoch": 0.96, + "learning_rate": 1.8400605565822346e-09, + "loss": 0.1394, + "step": 18355 + }, + { + "epoch": 0.96, + "learning_rate": 1.835487463485852e-09, + "loss": 0.1608, + "step": 18356 + }, + { + "epoch": 0.96, + "learning_rate": 1.8309200340077325e-09, + "loss": 0.1581, + "step": 18357 + }, + { + "epoch": 0.96, + "learning_rate": 1.826358268278394e-09, + "loss": 0.1221, + "step": 18358 + }, + { + "epoch": 0.96, + "learning_rate": 1.8218021664282658e-09, + "loss": 0.1186, + "step": 18359 + }, + { + "epoch": 0.96, + "learning_rate": 1.8172517285875322e-09, + "loss": 0.1483, + "step": 18360 + }, + { + "epoch": 0.96, + "learning_rate": 1.8127069548862451e-09, + "loss": 0.1186, + "step": 18361 + }, + { + "epoch": 0.96, + "learning_rate": 1.808167845454278e-09, + "loss": 0.1769, + "step": 18362 + }, + { + "epoch": 0.96, + "learning_rate": 1.8036344004214165e-09, + "loss": 0.1287, + "step": 18363 + }, + { + "epoch": 0.96, + "learning_rate": 1.799106619917179e-09, + "loss": 0.1464, + "step": 18364 + }, + { + "epoch": 0.96, + "learning_rate": 1.794584504070995e-09, + "loss": 0.1118, + "step": 18365 + }, + { + "epoch": 0.96, + "learning_rate": 1.7900680530120949e-09, + "loss": 0.1135, + "step": 18366 + }, + { + "epoch": 0.96, + "learning_rate": 1.7855572668695751e-09, + "loss": 0.1579, + "step": 18367 + }, + { + "epoch": 0.96, + "learning_rate": 1.7810521457723548e-09, + "loss": 0.1486, + "step": 18368 + }, + { + "epoch": 0.96, + "learning_rate": 1.7765526898491535e-09, + "loss": 0.1245, + "step": 18369 + }, + { + "epoch": 0.96, + "learning_rate": 1.7720588992286233e-09, + "loss": 0.1587, + "step": 18370 + }, + { + "epoch": 0.96, + "learning_rate": 1.7675707740391732e-09, + "loss": 0.1285, + "step": 18371 + }, + { + "epoch": 0.96, + "learning_rate": 1.7630883144090558e-09, + "loss": 0.1458, + "step": 18372 + }, + { + "epoch": 0.96, + "learning_rate": 1.7586115204664132e-09, + "loss": 0.2151, + "step": 18373 + }, + { + "epoch": 0.96, + "learning_rate": 1.7541403923391873e-09, + "loss": 0.1357, + "step": 18374 + }, + { + "epoch": 0.96, + "learning_rate": 1.749674930155165e-09, + "loss": 0.1501, + "step": 18375 + }, + { + "epoch": 0.96, + "learning_rate": 1.7452151340419551e-09, + "loss": 0.1453, + "step": 18376 + }, + { + "epoch": 0.96, + "learning_rate": 1.7407610041270338e-09, + "loss": 0.1321, + "step": 18377 + }, + { + "epoch": 0.96, + "learning_rate": 1.7363125405376987e-09, + "loss": 0.1454, + "step": 18378 + }, + { + "epoch": 0.96, + "learning_rate": 1.731869743401071e-09, + "loss": 0.1303, + "step": 18379 + }, + { + "epoch": 0.96, + "learning_rate": 1.7274326128441596e-09, + "loss": 0.1587, + "step": 18380 + }, + { + "epoch": 0.96, + "learning_rate": 1.7230011489937746e-09, + "loss": 0.1225, + "step": 18381 + }, + { + "epoch": 0.96, + "learning_rate": 1.7185753519765477e-09, + "loss": 0.1226, + "step": 18382 + }, + { + "epoch": 0.96, + "learning_rate": 1.7141552219190003e-09, + "loss": 0.1404, + "step": 18383 + }, + { + "epoch": 0.96, + "learning_rate": 1.709740758947409e-09, + "loss": 0.1066, + "step": 18384 + }, + { + "epoch": 0.96, + "learning_rate": 1.705331963188006e-09, + "loss": 0.1693, + "step": 18385 + }, + { + "epoch": 0.96, + "learning_rate": 1.7009288347667571e-09, + "loss": 0.1197, + "step": 18386 + }, + { + "epoch": 0.96, + "learning_rate": 1.6965313738094956e-09, + "loss": 0.1305, + "step": 18387 + }, + { + "epoch": 0.96, + "learning_rate": 1.6921395804419203e-09, + "loss": 0.1415, + "step": 18388 + }, + { + "epoch": 0.96, + "learning_rate": 1.6877534547895755e-09, + "loss": 0.1151, + "step": 18389 + }, + { + "epoch": 0.96, + "learning_rate": 1.6833729969777832e-09, + "loss": 0.1118, + "step": 18390 + }, + { + "epoch": 0.96, + "learning_rate": 1.6789982071317322e-09, + "loss": 0.1275, + "step": 18391 + }, + { + "epoch": 0.96, + "learning_rate": 1.6746290853765e-09, + "loss": 0.1102, + "step": 18392 + }, + { + "epoch": 0.96, + "learning_rate": 1.6702656318369202e-09, + "loss": 0.1338, + "step": 18393 + }, + { + "epoch": 0.96, + "learning_rate": 1.6659078466377152e-09, + "loss": 0.0977, + "step": 18394 + }, + { + "epoch": 0.96, + "learning_rate": 1.6615557299034078e-09, + "loss": 0.1186, + "step": 18395 + }, + { + "epoch": 0.96, + "learning_rate": 1.6572092817584094e-09, + "loss": 0.1329, + "step": 18396 + }, + { + "epoch": 0.96, + "learning_rate": 1.652868502326954e-09, + "loss": 0.1228, + "step": 18397 + }, + { + "epoch": 0.96, + "learning_rate": 1.6485333917330756e-09, + "loss": 0.1467, + "step": 18398 + }, + { + "epoch": 0.96, + "learning_rate": 1.6442039501006755e-09, + "loss": 0.1217, + "step": 18399 + }, + { + "epoch": 0.96, + "learning_rate": 1.639880177553521e-09, + "loss": 0.1345, + "step": 18400 + }, + { + "epoch": 0.96, + "learning_rate": 1.6355620742151577e-09, + "loss": 0.1672, + "step": 18401 + }, + { + "epoch": 0.96, + "learning_rate": 1.6312496402089981e-09, + "loss": 0.154, + "step": 18402 + }, + { + "epoch": 0.96, + "learning_rate": 1.6269428756582993e-09, + "loss": 0.1179, + "step": 18403 + }, + { + "epoch": 0.96, + "learning_rate": 1.6226417806861626e-09, + "loss": 0.143, + "step": 18404 + }, + { + "epoch": 0.96, + "learning_rate": 1.618346355415512e-09, + "loss": 0.1269, + "step": 18405 + }, + { + "epoch": 0.96, + "learning_rate": 1.6140565999690937e-09, + "loss": 0.1744, + "step": 18406 + }, + { + "epoch": 0.96, + "learning_rate": 1.6097725144695206e-09, + "loss": 0.1295, + "step": 18407 + }, + { + "epoch": 0.96, + "learning_rate": 1.6054940990392507e-09, + "loss": 0.1764, + "step": 18408 + }, + { + "epoch": 0.96, + "learning_rate": 1.6012213538005415e-09, + "loss": 0.1305, + "step": 18409 + }, + { + "epoch": 0.96, + "learning_rate": 1.5969542788755174e-09, + "loss": 0.1396, + "step": 18410 + }, + { + "epoch": 0.96, + "learning_rate": 1.5926928743861256e-09, + "loss": 0.1361, + "step": 18411 + }, + { + "epoch": 0.96, + "learning_rate": 1.5884371404541575e-09, + "loss": 0.1376, + "step": 18412 + }, + { + "epoch": 0.96, + "learning_rate": 1.584187077201249e-09, + "loss": 0.1497, + "step": 18413 + }, + { + "epoch": 0.96, + "learning_rate": 1.5799426847488807e-09, + "loss": 0.1253, + "step": 18414 + }, + { + "epoch": 0.96, + "learning_rate": 1.5757039632183556e-09, + "loss": 0.1366, + "step": 18415 + }, + { + "epoch": 0.96, + "learning_rate": 1.571470912730799e-09, + "loss": 0.1442, + "step": 18416 + }, + { + "epoch": 0.96, + "learning_rate": 1.5672435334072031e-09, + "loss": 0.1634, + "step": 18417 + }, + { + "epoch": 0.96, + "learning_rate": 1.5630218253684047e-09, + "loss": 0.1125, + "step": 18418 + }, + { + "epoch": 0.96, + "learning_rate": 1.558805788735018e-09, + "loss": 0.1395, + "step": 18419 + }, + { + "epoch": 0.96, + "learning_rate": 1.5545954236275915e-09, + "loss": 0.1909, + "step": 18420 + }, + { + "epoch": 0.96, + "learning_rate": 1.5503907301664287e-09, + "loss": 0.1316, + "step": 18421 + }, + { + "epoch": 0.96, + "learning_rate": 1.5461917084717003e-09, + "loss": 0.1201, + "step": 18422 + }, + { + "epoch": 0.96, + "learning_rate": 1.5419983586634432e-09, + "loss": 0.1837, + "step": 18423 + }, + { + "epoch": 0.96, + "learning_rate": 1.5378106808614733e-09, + "loss": 0.1189, + "step": 18424 + }, + { + "epoch": 0.96, + "learning_rate": 1.533628675185472e-09, + "loss": 0.1451, + "step": 18425 + }, + { + "epoch": 0.96, + "learning_rate": 1.529452341755011e-09, + "loss": 0.114, + "step": 18426 + }, + { + "epoch": 0.96, + "learning_rate": 1.5252816806893942e-09, + "loss": 0.1448, + "step": 18427 + }, + { + "epoch": 0.96, + "learning_rate": 1.52111669210786e-09, + "loss": 0.1333, + "step": 18428 + }, + { + "epoch": 0.96, + "learning_rate": 1.5169573761294464e-09, + "loss": 0.1511, + "step": 18429 + }, + { + "epoch": 0.96, + "learning_rate": 1.5128037328729915e-09, + "loss": 0.1381, + "step": 18430 + }, + { + "epoch": 0.96, + "learning_rate": 1.5086557624572227e-09, + "loss": 0.1385, + "step": 18431 + }, + { + "epoch": 0.96, + "learning_rate": 1.5045134650007118e-09, + "loss": 0.1387, + "step": 18432 + }, + { + "epoch": 0.96, + "learning_rate": 1.5003768406218309e-09, + "loss": 0.1169, + "step": 18433 + }, + { + "epoch": 0.96, + "learning_rate": 1.4962458894387964e-09, + "loss": 0.1442, + "step": 18434 + }, + { + "epoch": 0.96, + "learning_rate": 1.4921206115696917e-09, + "loss": 0.1753, + "step": 18435 + }, + { + "epoch": 0.96, + "learning_rate": 1.4880010071324223e-09, + "loss": 0.1528, + "step": 18436 + }, + { + "epoch": 0.96, + "learning_rate": 1.4838870762447164e-09, + "loss": 0.0975, + "step": 18437 + }, + { + "epoch": 0.96, + "learning_rate": 1.4797788190241245e-09, + "loss": 0.1268, + "step": 18438 + }, + { + "epoch": 0.96, + "learning_rate": 1.4756762355881303e-09, + "loss": 0.1479, + "step": 18439 + }, + { + "epoch": 0.96, + "learning_rate": 1.4715793260539288e-09, + "loss": 0.1221, + "step": 18440 + }, + { + "epoch": 0.96, + "learning_rate": 1.4674880905386266e-09, + "loss": 0.1073, + "step": 18441 + }, + { + "epoch": 0.96, + "learning_rate": 1.4634025291591523e-09, + "loss": 0.1018, + "step": 18442 + }, + { + "epoch": 0.96, + "learning_rate": 1.4593226420323013e-09, + "loss": 0.1151, + "step": 18443 + }, + { + "epoch": 0.96, + "learning_rate": 1.4552484292746248e-09, + "loss": 0.1114, + "step": 18444 + }, + { + "epoch": 0.96, + "learning_rate": 1.4511798910026075e-09, + "loss": 0.1913, + "step": 18445 + }, + { + "epoch": 0.96, + "learning_rate": 1.4471170273325118e-09, + "loss": 0.0993, + "step": 18446 + }, + { + "epoch": 0.96, + "learning_rate": 1.443059838380467e-09, + "loss": 0.1541, + "step": 18447 + }, + { + "epoch": 0.96, + "learning_rate": 1.439008324262425e-09, + "loss": 0.1855, + "step": 18448 + }, + { + "epoch": 0.96, + "learning_rate": 1.43496248509416e-09, + "loss": 0.1382, + "step": 18449 + }, + { + "epoch": 0.96, + "learning_rate": 1.4309223209913347e-09, + "loss": 0.1394, + "step": 18450 + }, + { + "epoch": 0.96, + "learning_rate": 1.42688783206939e-09, + "loss": 0.1647, + "step": 18451 + }, + { + "epoch": 0.96, + "learning_rate": 1.4228590184436785e-09, + "loss": 0.1454, + "step": 18452 + }, + { + "epoch": 0.96, + "learning_rate": 1.4188358802292855e-09, + "loss": 0.1304, + "step": 18453 + }, + { + "epoch": 0.96, + "learning_rate": 1.4148184175412303e-09, + "loss": 0.139, + "step": 18454 + }, + { + "epoch": 0.96, + "learning_rate": 1.4108066304943323e-09, + "loss": 0.1202, + "step": 18455 + }, + { + "epoch": 0.96, + "learning_rate": 1.406800519203233e-09, + "loss": 0.1247, + "step": 18456 + }, + { + "epoch": 0.96, + "learning_rate": 1.4028000837824627e-09, + "loss": 0.1448, + "step": 18457 + }, + { + "epoch": 0.96, + "learning_rate": 1.3988053243463083e-09, + "loss": 0.1125, + "step": 18458 + }, + { + "epoch": 0.96, + "learning_rate": 1.3948162410089893e-09, + "loss": 0.1093, + "step": 18459 + }, + { + "epoch": 0.96, + "learning_rate": 1.390832833884481e-09, + "loss": 0.1512, + "step": 18460 + }, + { + "epoch": 0.96, + "learning_rate": 1.3868551030866261e-09, + "loss": 0.1544, + "step": 18461 + }, + { + "epoch": 0.96, + "learning_rate": 1.3828830487291332e-09, + "loss": 0.1517, + "step": 18462 + }, + { + "epoch": 0.96, + "learning_rate": 1.3789166709255118e-09, + "loss": 0.1099, + "step": 18463 + }, + { + "epoch": 0.96, + "learning_rate": 1.3749559697891378e-09, + "loss": 0.117, + "step": 18464 + }, + { + "epoch": 0.96, + "learning_rate": 1.3710009454332095e-09, + "loss": 0.1492, + "step": 18465 + }, + { + "epoch": 0.96, + "learning_rate": 1.3670515979707254e-09, + "loss": 0.1607, + "step": 18466 + }, + { + "epoch": 0.96, + "learning_rate": 1.3631079275145952e-09, + "loss": 0.1431, + "step": 18467 + }, + { + "epoch": 0.96, + "learning_rate": 1.3591699341775287e-09, + "loss": 0.1547, + "step": 18468 + }, + { + "epoch": 0.96, + "learning_rate": 1.3552376180720803e-09, + "loss": 0.1634, + "step": 18469 + }, + { + "epoch": 0.96, + "learning_rate": 1.3513109793106047e-09, + "loss": 0.1209, + "step": 18470 + }, + { + "epoch": 0.96, + "learning_rate": 1.3473900180053676e-09, + "loss": 0.1515, + "step": 18471 + }, + { + "epoch": 0.96, + "learning_rate": 1.3434747342683906e-09, + "loss": 0.151, + "step": 18472 + }, + { + "epoch": 0.96, + "learning_rate": 1.3395651282116061e-09, + "loss": 0.1647, + "step": 18473 + }, + { + "epoch": 0.96, + "learning_rate": 1.335661199946725e-09, + "loss": 0.1228, + "step": 18474 + }, + { + "epoch": 0.96, + "learning_rate": 1.3317629495853688e-09, + "loss": 0.138, + "step": 18475 + }, + { + "epoch": 0.96, + "learning_rate": 1.3278703772388934e-09, + "loss": 0.1289, + "step": 18476 + }, + { + "epoch": 0.96, + "learning_rate": 1.323983483018587e-09, + "loss": 0.1257, + "step": 18477 + }, + { + "epoch": 0.96, + "learning_rate": 1.3201022670355387e-09, + "loss": 0.1383, + "step": 18478 + }, + { + "epoch": 0.96, + "learning_rate": 1.3162267294006601e-09, + "loss": 0.1721, + "step": 18479 + }, + { + "epoch": 0.96, + "learning_rate": 1.312356870224729e-09, + "loss": 0.1577, + "step": 18480 + }, + { + "epoch": 0.96, + "learning_rate": 1.3084926896183234e-09, + "loss": 0.108, + "step": 18481 + }, + { + "epoch": 0.96, + "learning_rate": 1.3046341876918887e-09, + "loss": 0.1814, + "step": 18482 + }, + { + "epoch": 0.96, + "learning_rate": 1.3007813645557363e-09, + "loss": 0.1258, + "step": 18483 + }, + { + "epoch": 0.96, + "learning_rate": 1.2969342203199563e-09, + "loss": 0.1368, + "step": 18484 + }, + { + "epoch": 0.96, + "learning_rate": 1.2930927550944825e-09, + "loss": 0.1488, + "step": 18485 + }, + { + "epoch": 0.96, + "learning_rate": 1.2892569689891608e-09, + "loss": 0.1788, + "step": 18486 + }, + { + "epoch": 0.96, + "learning_rate": 1.28542686211357e-09, + "loss": 0.0978, + "step": 18487 + }, + { + "epoch": 0.96, + "learning_rate": 1.2816024345771781e-09, + "loss": 0.1598, + "step": 18488 + }, + { + "epoch": 0.97, + "learning_rate": 1.27778368648932e-09, + "loss": 0.1146, + "step": 18489 + }, + { + "epoch": 0.97, + "learning_rate": 1.273970617959108e-09, + "loss": 0.1139, + "step": 18490 + }, + { + "epoch": 0.97, + "learning_rate": 1.2701632290955444e-09, + "loss": 0.143, + "step": 18491 + }, + { + "epoch": 0.97, + "learning_rate": 1.2663615200074085e-09, + "loss": 0.1425, + "step": 18492 + }, + { + "epoch": 0.97, + "learning_rate": 1.2625654908033912e-09, + "loss": 0.145, + "step": 18493 + }, + { + "epoch": 0.97, + "learning_rate": 1.2587751415919835e-09, + "loss": 0.1323, + "step": 18494 + }, + { + "epoch": 0.97, + "learning_rate": 1.2549904724814987e-09, + "loss": 0.1534, + "step": 18495 + }, + { + "epoch": 0.97, + "learning_rate": 1.2512114835800946e-09, + "loss": 0.125, + "step": 18496 + }, + { + "epoch": 0.97, + "learning_rate": 1.2474381749958184e-09, + "loss": 0.1324, + "step": 18497 + }, + { + "epoch": 0.97, + "learning_rate": 1.2436705468364728e-09, + "loss": 0.1617, + "step": 18498 + }, + { + "epoch": 0.97, + "learning_rate": 1.2399085992097492e-09, + "loss": 0.1465, + "step": 18499 + }, + { + "epoch": 0.97, + "learning_rate": 1.2361523322231393e-09, + "loss": 0.1154, + "step": 18500 + }, + { + "epoch": 0.97, + "learning_rate": 1.2324017459840685e-09, + "loss": 0.1378, + "step": 18501 + }, + { + "epoch": 0.97, + "learning_rate": 1.2286568405996734e-09, + "loss": 0.1299, + "step": 18502 + }, + { + "epoch": 0.97, + "learning_rate": 1.2249176161769791e-09, + "loss": 0.1369, + "step": 18503 + }, + { + "epoch": 0.97, + "learning_rate": 1.2211840728229005e-09, + "loss": 0.137, + "step": 18504 + }, + { + "epoch": 0.97, + "learning_rate": 1.2174562106441077e-09, + "loss": 0.1578, + "step": 18505 + }, + { + "epoch": 0.97, + "learning_rate": 1.21373402974716e-09, + "loss": 0.1194, + "step": 18506 + }, + { + "epoch": 0.97, + "learning_rate": 1.2100175302384163e-09, + "loss": 0.1367, + "step": 18507 + }, + { + "epoch": 0.97, + "learning_rate": 1.2063067122241255e-09, + "loss": 0.1535, + "step": 18508 + }, + { + "epoch": 0.97, + "learning_rate": 1.2026015758103358e-09, + "loss": 0.1427, + "step": 18509 + }, + { + "epoch": 0.97, + "learning_rate": 1.1989021211029405e-09, + "loss": 0.1465, + "step": 18510 + }, + { + "epoch": 0.97, + "learning_rate": 1.1952083482076546e-09, + "loss": 0.1103, + "step": 18511 + }, + { + "epoch": 0.97, + "learning_rate": 1.191520257230061e-09, + "loss": 0.0925, + "step": 18512 + }, + { + "epoch": 0.97, + "learning_rate": 1.1878378482755857e-09, + "loss": 0.1282, + "step": 18513 + }, + { + "epoch": 0.97, + "learning_rate": 1.184161121449412e-09, + "loss": 0.1176, + "step": 18514 + }, + { + "epoch": 0.97, + "learning_rate": 1.1804900768566995e-09, + "loss": 0.1319, + "step": 18515 + }, + { + "epoch": 0.97, + "learning_rate": 1.1768247146023202e-09, + "loss": 0.1398, + "step": 18516 + }, + { + "epoch": 0.97, + "learning_rate": 1.1731650347910572e-09, + "loss": 0.1253, + "step": 18517 + }, + { + "epoch": 0.97, + "learning_rate": 1.1695110375274708e-09, + "loss": 0.1793, + "step": 18518 + }, + { + "epoch": 0.97, + "learning_rate": 1.165862722916011e-09, + "loss": 0.142, + "step": 18519 + }, + { + "epoch": 0.97, + "learning_rate": 1.1622200910609947e-09, + "loss": 0.1745, + "step": 18520 + }, + { + "epoch": 0.97, + "learning_rate": 1.1585831420664494e-09, + "loss": 0.1341, + "step": 18521 + }, + { + "epoch": 0.97, + "learning_rate": 1.1549518760363587e-09, + "loss": 0.1167, + "step": 18522 + }, + { + "epoch": 0.97, + "learning_rate": 1.1513262930745282e-09, + "loss": 0.1166, + "step": 18523 + }, + { + "epoch": 0.97, + "learning_rate": 1.1477063932845421e-09, + "loss": 0.1358, + "step": 18524 + }, + { + "epoch": 0.97, + "learning_rate": 1.144092176769873e-09, + "loss": 0.1422, + "step": 18525 + }, + { + "epoch": 0.97, + "learning_rate": 1.140483643633816e-09, + "loss": 0.1291, + "step": 18526 + }, + { + "epoch": 0.97, + "learning_rate": 1.136880793979489e-09, + "loss": 0.1336, + "step": 18527 + }, + { + "epoch": 0.97, + "learning_rate": 1.1332836279098979e-09, + "loss": 0.1237, + "step": 18528 + }, + { + "epoch": 0.97, + "learning_rate": 1.1296921455278275e-09, + "loss": 0.1334, + "step": 18529 + }, + { + "epoch": 0.97, + "learning_rate": 1.1261063469359289e-09, + "loss": 0.1052, + "step": 18530 + }, + { + "epoch": 0.97, + "learning_rate": 1.1225262322366757e-09, + "loss": 0.1152, + "step": 18531 + }, + { + "epoch": 0.97, + "learning_rate": 1.1189518015323863e-09, + "loss": 0.1377, + "step": 18532 + }, + { + "epoch": 0.97, + "learning_rate": 1.1153830549252452e-09, + "loss": 0.1292, + "step": 18533 + }, + { + "epoch": 0.97, + "learning_rate": 1.1118199925172377e-09, + "loss": 0.1275, + "step": 18534 + }, + { + "epoch": 0.97, + "learning_rate": 1.1082626144101936e-09, + "loss": 0.1303, + "step": 18535 + }, + { + "epoch": 0.97, + "learning_rate": 1.104710920705787e-09, + "loss": 0.1919, + "step": 18536 + }, + { + "epoch": 0.97, + "learning_rate": 1.1011649115055143e-09, + "loss": 0.1752, + "step": 18537 + }, + { + "epoch": 0.97, + "learning_rate": 1.0976245869107393e-09, + "loss": 0.1584, + "step": 18538 + }, + { + "epoch": 0.97, + "learning_rate": 1.094089947022625e-09, + "loss": 0.1435, + "step": 18539 + }, + { + "epoch": 0.97, + "learning_rate": 1.0905609919422022e-09, + "loss": 0.1171, + "step": 18540 + }, + { + "epoch": 0.97, + "learning_rate": 1.0870377217703452e-09, + "loss": 0.1565, + "step": 18541 + }, + { + "epoch": 0.97, + "learning_rate": 1.0835201366077517e-09, + "loss": 0.1399, + "step": 18542 + }, + { + "epoch": 0.97, + "learning_rate": 1.0800082365548969e-09, + "loss": 0.1265, + "step": 18543 + }, + { + "epoch": 0.97, + "learning_rate": 1.0765020217122333e-09, + "loss": 0.1726, + "step": 18544 + }, + { + "epoch": 0.97, + "learning_rate": 1.0730014921799257e-09, + "loss": 0.1209, + "step": 18545 + }, + { + "epoch": 0.97, + "learning_rate": 1.0695066480580495e-09, + "loss": 0.1237, + "step": 18546 + }, + { + "epoch": 0.97, + "learning_rate": 1.0660174894464358e-09, + "loss": 0.1242, + "step": 18547 + }, + { + "epoch": 0.97, + "learning_rate": 1.0625340164448492e-09, + "loss": 0.0991, + "step": 18548 + }, + { + "epoch": 0.97, + "learning_rate": 1.0590562291528548e-09, + "loss": 0.1479, + "step": 18549 + }, + { + "epoch": 0.97, + "learning_rate": 1.0555841276698175e-09, + "loss": 0.1198, + "step": 18550 + }, + { + "epoch": 0.97, + "learning_rate": 1.0521177120949687e-09, + "loss": 0.1498, + "step": 18551 + }, + { + "epoch": 0.97, + "learning_rate": 1.0486569825274294e-09, + "loss": 0.1364, + "step": 18552 + }, + { + "epoch": 0.97, + "learning_rate": 1.0452019390660538e-09, + "loss": 0.1399, + "step": 18553 + }, + { + "epoch": 0.97, + "learning_rate": 1.0417525818096296e-09, + "loss": 0.1816, + "step": 18554 + }, + { + "epoch": 0.97, + "learning_rate": 1.0383089108567221e-09, + "loss": 0.1329, + "step": 18555 + }, + { + "epoch": 0.97, + "learning_rate": 1.034870926305742e-09, + "loss": 0.1115, + "step": 18556 + }, + { + "epoch": 0.97, + "learning_rate": 1.0314386282549436e-09, + "loss": 0.1312, + "step": 18557 + }, + { + "epoch": 0.97, + "learning_rate": 1.0280120168024487e-09, + "loss": 0.1511, + "step": 18558 + }, + { + "epoch": 0.97, + "learning_rate": 1.0245910920462008e-09, + "loss": 0.153, + "step": 18559 + }, + { + "epoch": 0.97, + "learning_rate": 1.0211758540839444e-09, + "loss": 0.1183, + "step": 18560 + }, + { + "epoch": 0.97, + "learning_rate": 1.0177663030132678e-09, + "loss": 0.1238, + "step": 18561 + }, + { + "epoch": 0.97, + "learning_rate": 1.0143624389316707e-09, + "loss": 0.1583, + "step": 18562 + }, + { + "epoch": 0.97, + "learning_rate": 1.0109642619364312e-09, + "loss": 0.1341, + "step": 18563 + }, + { + "epoch": 0.97, + "learning_rate": 1.0075717721246046e-09, + "loss": 0.1597, + "step": 18564 + }, + { + "epoch": 0.97, + "learning_rate": 1.0041849695932248e-09, + "loss": 0.1082, + "step": 18565 + }, + { + "epoch": 0.97, + "learning_rate": 1.0008038544390584e-09, + "loss": 0.1061, + "step": 18566 + }, + { + "epoch": 0.97, + "learning_rate": 9.974284267587397e-10, + "loss": 0.1431, + "step": 18567 + }, + { + "epoch": 0.97, + "learning_rate": 9.940586866487244e-10, + "loss": 0.0857, + "step": 18568 + }, + { + "epoch": 0.97, + "learning_rate": 9.906946342053357e-10, + "loss": 0.1368, + "step": 18569 + }, + { + "epoch": 0.97, + "learning_rate": 9.87336269524741e-10, + "loss": 0.1626, + "step": 18570 + }, + { + "epoch": 0.97, + "learning_rate": 9.83983592702886e-10, + "loss": 0.1122, + "step": 18571 + }, + { + "epoch": 0.97, + "learning_rate": 9.80636603835605e-10, + "loss": 0.111, + "step": 18572 + }, + { + "epoch": 0.97, + "learning_rate": 9.772953030185771e-10, + "loss": 0.1477, + "step": 18573 + }, + { + "epoch": 0.97, + "learning_rate": 9.739596903472592e-10, + "loss": 0.1319, + "step": 18574 + }, + { + "epoch": 0.97, + "learning_rate": 9.706297659170193e-10, + "loss": 0.1214, + "step": 18575 + }, + { + "epoch": 0.97, + "learning_rate": 9.673055298230038e-10, + "loss": 0.1734, + "step": 18576 + }, + { + "epoch": 0.97, + "learning_rate": 9.639869821602253e-10, + "loss": 0.1788, + "step": 18577 + }, + { + "epoch": 0.97, + "learning_rate": 9.606741230235638e-10, + "loss": 0.1257, + "step": 18578 + }, + { + "epoch": 0.97, + "learning_rate": 9.573669525076322e-10, + "loss": 0.1344, + "step": 18579 + }, + { + "epoch": 0.97, + "learning_rate": 9.540654707070217e-10, + "loss": 0.1275, + "step": 18580 + }, + { + "epoch": 0.97, + "learning_rate": 9.507696777160568e-10, + "loss": 0.1397, + "step": 18581 + }, + { + "epoch": 0.97, + "learning_rate": 9.474795736289287e-10, + "loss": 0.1294, + "step": 18582 + }, + { + "epoch": 0.97, + "learning_rate": 9.441951585396735e-10, + "loss": 0.1699, + "step": 18583 + }, + { + "epoch": 0.97, + "learning_rate": 9.409164325421714e-10, + "loss": 0.1187, + "step": 18584 + }, + { + "epoch": 0.97, + "learning_rate": 9.376433957301255e-10, + "loss": 0.1352, + "step": 18585 + }, + { + "epoch": 0.97, + "learning_rate": 9.343760481970608e-10, + "loss": 0.1206, + "step": 18586 + }, + { + "epoch": 0.97, + "learning_rate": 9.311143900363694e-10, + "loss": 0.1209, + "step": 18587 + }, + { + "epoch": 0.97, + "learning_rate": 9.278584213413099e-10, + "loss": 0.1518, + "step": 18588 + }, + { + "epoch": 0.97, + "learning_rate": 9.246081422048746e-10, + "loss": 0.1158, + "step": 18589 + }, + { + "epoch": 0.97, + "learning_rate": 9.213635527200114e-10, + "loss": 0.1387, + "step": 18590 + }, + { + "epoch": 0.97, + "learning_rate": 9.18124652979424e-10, + "loss": 0.1384, + "step": 18591 + }, + { + "epoch": 0.97, + "learning_rate": 9.148914430756827e-10, + "loss": 0.1744, + "step": 18592 + }, + { + "epoch": 0.97, + "learning_rate": 9.116639231012024e-10, + "loss": 0.1349, + "step": 18593 + }, + { + "epoch": 0.97, + "learning_rate": 9.084420931482206e-10, + "loss": 0.1223, + "step": 18594 + }, + { + "epoch": 0.97, + "learning_rate": 9.052259533088192e-10, + "loss": 0.1331, + "step": 18595 + }, + { + "epoch": 0.97, + "learning_rate": 9.020155036749244e-10, + "loss": 0.1458, + "step": 18596 + }, + { + "epoch": 0.97, + "learning_rate": 8.988107443382853e-10, + "loss": 0.1055, + "step": 18597 + }, + { + "epoch": 0.97, + "learning_rate": 8.956116753904952e-10, + "loss": 0.1156, + "step": 18598 + }, + { + "epoch": 0.97, + "learning_rate": 8.924182969229699e-10, + "loss": 0.1438, + "step": 18599 + }, + { + "epoch": 0.97, + "learning_rate": 8.892306090270141e-10, + "loss": 0.0992, + "step": 18600 + }, + { + "epoch": 0.97, + "learning_rate": 8.860486117936883e-10, + "loss": 0.1273, + "step": 18601 + }, + { + "epoch": 0.97, + "learning_rate": 8.828723053139864e-10, + "loss": 0.1089, + "step": 18602 + }, + { + "epoch": 0.97, + "learning_rate": 8.79701689678658e-10, + "loss": 0.1212, + "step": 18603 + }, + { + "epoch": 0.97, + "learning_rate": 8.765367649782973e-10, + "loss": 0.1389, + "step": 18604 + }, + { + "epoch": 0.97, + "learning_rate": 8.733775313034098e-10, + "loss": 0.1425, + "step": 18605 + }, + { + "epoch": 0.97, + "learning_rate": 8.702239887442564e-10, + "loss": 0.1332, + "step": 18606 + }, + { + "epoch": 0.97, + "learning_rate": 8.670761373909874e-10, + "loss": 0.1382, + "step": 18607 + }, + { + "epoch": 0.97, + "learning_rate": 8.639339773335307e-10, + "loss": 0.1317, + "step": 18608 + }, + { + "epoch": 0.97, + "learning_rate": 8.607975086617258e-10, + "loss": 0.1212, + "step": 18609 + }, + { + "epoch": 0.97, + "learning_rate": 8.576667314651898e-10, + "loss": 0.1312, + "step": 18610 + }, + { + "epoch": 0.97, + "learning_rate": 8.545416458334509e-10, + "loss": 0.1485, + "step": 18611 + }, + { + "epoch": 0.97, + "learning_rate": 8.51422251855749e-10, + "loss": 0.139, + "step": 18612 + }, + { + "epoch": 0.97, + "learning_rate": 8.483085496213016e-10, + "loss": 0.1273, + "step": 18613 + }, + { + "epoch": 0.97, + "learning_rate": 8.45200539219082e-10, + "loss": 0.1081, + "step": 18614 + }, + { + "epoch": 0.97, + "learning_rate": 8.420982207378857e-10, + "loss": 0.1319, + "step": 18615 + }, + { + "epoch": 0.97, + "learning_rate": 8.390015942664197e-10, + "loss": 0.1139, + "step": 18616 + }, + { + "epoch": 0.97, + "learning_rate": 8.359106598931908e-10, + "loss": 0.1238, + "step": 18617 + }, + { + "epoch": 0.97, + "learning_rate": 8.328254177064841e-10, + "loss": 0.1283, + "step": 18618 + }, + { + "epoch": 0.97, + "learning_rate": 8.297458677945401e-10, + "loss": 0.1406, + "step": 18619 + }, + { + "epoch": 0.97, + "learning_rate": 8.266720102453328e-10, + "loss": 0.1417, + "step": 18620 + }, + { + "epoch": 0.97, + "learning_rate": 8.236038451467253e-10, + "loss": 0.1297, + "step": 18621 + }, + { + "epoch": 0.97, + "learning_rate": 8.205413725864252e-10, + "loss": 0.1703, + "step": 18622 + }, + { + "epoch": 0.97, + "learning_rate": 8.174845926519402e-10, + "loss": 0.1403, + "step": 18623 + }, + { + "epoch": 0.97, + "learning_rate": 8.144335054306451e-10, + "loss": 0.1055, + "step": 18624 + }, + { + "epoch": 0.97, + "learning_rate": 8.113881110097142e-10, + "loss": 0.1963, + "step": 18625 + }, + { + "epoch": 0.97, + "learning_rate": 8.083484094762116e-10, + "loss": 0.1255, + "step": 18626 + }, + { + "epoch": 0.97, + "learning_rate": 8.05314400917001e-10, + "loss": 0.1242, + "step": 18627 + }, + { + "epoch": 0.97, + "learning_rate": 8.02286085418813e-10, + "loss": 0.1548, + "step": 18628 + }, + { + "epoch": 0.97, + "learning_rate": 7.992634630682005e-10, + "loss": 0.1233, + "step": 18629 + }, + { + "epoch": 0.97, + "learning_rate": 7.962465339514946e-10, + "loss": 0.1269, + "step": 18630 + }, + { + "epoch": 0.97, + "learning_rate": 7.932352981550039e-10, + "loss": 0.1085, + "step": 18631 + }, + { + "epoch": 0.97, + "learning_rate": 7.902297557647486e-10, + "loss": 0.1136, + "step": 18632 + }, + { + "epoch": 0.97, + "learning_rate": 7.872299068665933e-10, + "loss": 0.1506, + "step": 18633 + }, + { + "epoch": 0.97, + "learning_rate": 7.84235751546336e-10, + "loss": 0.1419, + "step": 18634 + }, + { + "epoch": 0.97, + "learning_rate": 7.812472898895306e-10, + "loss": 0.1147, + "step": 18635 + }, + { + "epoch": 0.97, + "learning_rate": 7.782645219815753e-10, + "loss": 0.1143, + "step": 18636 + }, + { + "epoch": 0.97, + "learning_rate": 7.752874479077354e-10, + "loss": 0.1638, + "step": 18637 + }, + { + "epoch": 0.97, + "learning_rate": 7.72316067753076e-10, + "loss": 0.1384, + "step": 18638 + }, + { + "epoch": 0.97, + "learning_rate": 7.693503816025515e-10, + "loss": 0.1063, + "step": 18639 + }, + { + "epoch": 0.97, + "learning_rate": 7.66390389540894e-10, + "loss": 0.1482, + "step": 18640 + }, + { + "epoch": 0.97, + "learning_rate": 7.634360916527027e-10, + "loss": 0.1524, + "step": 18641 + }, + { + "epoch": 0.97, + "learning_rate": 7.604874880224211e-10, + "loss": 0.1259, + "step": 18642 + }, + { + "epoch": 0.97, + "learning_rate": 7.575445787343371e-10, + "loss": 0.2096, + "step": 18643 + }, + { + "epoch": 0.97, + "learning_rate": 7.54607363872517e-10, + "loss": 0.1331, + "step": 18644 + }, + { + "epoch": 0.97, + "learning_rate": 7.516758435209603e-10, + "loss": 0.1337, + "step": 18645 + }, + { + "epoch": 0.97, + "learning_rate": 7.48750017763422e-10, + "loss": 0.1079, + "step": 18646 + }, + { + "epoch": 0.97, + "learning_rate": 7.458298866835244e-10, + "loss": 0.1289, + "step": 18647 + }, + { + "epoch": 0.97, + "learning_rate": 7.429154503647117e-10, + "loss": 0.1281, + "step": 18648 + }, + { + "epoch": 0.97, + "learning_rate": 7.400067088903173e-10, + "loss": 0.1437, + "step": 18649 + }, + { + "epoch": 0.97, + "learning_rate": 7.371036623434523e-10, + "loss": 0.1448, + "step": 18650 + }, + { + "epoch": 0.97, + "learning_rate": 7.342063108070729e-10, + "loss": 0.1241, + "step": 18651 + }, + { + "epoch": 0.97, + "learning_rate": 7.313146543640236e-10, + "loss": 0.1195, + "step": 18652 + }, + { + "epoch": 0.97, + "learning_rate": 7.284286930969053e-10, + "loss": 0.1268, + "step": 18653 + }, + { + "epoch": 0.97, + "learning_rate": 7.255484270882517e-10, + "loss": 0.1755, + "step": 18654 + }, + { + "epoch": 0.97, + "learning_rate": 7.226738564203084e-10, + "loss": 0.1456, + "step": 18655 + }, + { + "epoch": 0.97, + "learning_rate": 7.198049811752982e-10, + "loss": 0.1317, + "step": 18656 + }, + { + "epoch": 0.97, + "learning_rate": 7.169418014352002e-10, + "loss": 0.1064, + "step": 18657 + }, + { + "epoch": 0.97, + "learning_rate": 7.140843172818379e-10, + "loss": 0.1094, + "step": 18658 + }, + { + "epoch": 0.97, + "learning_rate": 7.112325287968568e-10, + "loss": 0.1337, + "step": 18659 + }, + { + "epoch": 0.97, + "learning_rate": 7.083864360617919e-10, + "loss": 0.1622, + "step": 18660 + }, + { + "epoch": 0.97, + "learning_rate": 7.055460391579782e-10, + "loss": 0.1539, + "step": 18661 + }, + { + "epoch": 0.97, + "learning_rate": 7.02711338166595e-10, + "loss": 0.1676, + "step": 18662 + }, + { + "epoch": 0.97, + "learning_rate": 6.998823331686665e-10, + "loss": 0.1307, + "step": 18663 + }, + { + "epoch": 0.97, + "learning_rate": 6.970590242450169e-10, + "loss": 0.1481, + "step": 18664 + }, + { + "epoch": 0.97, + "learning_rate": 6.942414114763817e-10, + "loss": 0.1779, + "step": 18665 + }, + { + "epoch": 0.97, + "learning_rate": 6.91429494943252e-10, + "loss": 0.1581, + "step": 18666 + }, + { + "epoch": 0.97, + "learning_rate": 6.886232747260301e-10, + "loss": 0.1516, + "step": 18667 + }, + { + "epoch": 0.97, + "learning_rate": 6.858227509048741e-10, + "loss": 0.1504, + "step": 18668 + }, + { + "epoch": 0.97, + "learning_rate": 6.830279235598535e-10, + "loss": 0.1277, + "step": 18669 + }, + { + "epoch": 0.97, + "learning_rate": 6.802387927708375e-10, + "loss": 0.1445, + "step": 18670 + }, + { + "epoch": 0.97, + "learning_rate": 6.774553586175402e-10, + "loss": 0.1116, + "step": 18671 + }, + { + "epoch": 0.97, + "learning_rate": 6.746776211795202e-10, + "loss": 0.1412, + "step": 18672 + }, + { + "epoch": 0.97, + "learning_rate": 6.719055805361584e-10, + "loss": 0.1684, + "step": 18673 + }, + { + "epoch": 0.97, + "learning_rate": 6.691392367666804e-10, + "loss": 0.1355, + "step": 18674 + }, + { + "epoch": 0.97, + "learning_rate": 6.663785899501561e-10, + "loss": 0.1413, + "step": 18675 + }, + { + "epoch": 0.97, + "learning_rate": 6.636236401654782e-10, + "loss": 0.1387, + "step": 18676 + }, + { + "epoch": 0.97, + "learning_rate": 6.608743874913835e-10, + "loss": 0.1484, + "step": 18677 + }, + { + "epoch": 0.97, + "learning_rate": 6.58130832006476e-10, + "loss": 0.111, + "step": 18678 + }, + { + "epoch": 0.97, + "learning_rate": 6.553929737891151e-10, + "loss": 0.0976, + "step": 18679 + }, + { + "epoch": 0.98, + "learning_rate": 6.526608129175937e-10, + "loss": 0.1291, + "step": 18680 + }, + { + "epoch": 0.98, + "learning_rate": 6.499343494699827e-10, + "loss": 0.164, + "step": 18681 + }, + { + "epoch": 0.98, + "learning_rate": 6.472135835241976e-10, + "loss": 0.1285, + "step": 18682 + }, + { + "epoch": 0.98, + "learning_rate": 6.444985151579984e-10, + "loss": 0.1146, + "step": 18683 + }, + { + "epoch": 0.98, + "learning_rate": 6.41789144449012e-10, + "loss": 0.1334, + "step": 18684 + }, + { + "epoch": 0.98, + "learning_rate": 6.39085471474643e-10, + "loss": 0.1098, + "step": 18685 + }, + { + "epoch": 0.98, + "learning_rate": 6.363874963121629e-10, + "loss": 0.1433, + "step": 18686 + }, + { + "epoch": 0.98, + "learning_rate": 6.336952190386879e-10, + "loss": 0.1253, + "step": 18687 + }, + { + "epoch": 0.98, + "learning_rate": 6.310086397311787e-10, + "loss": 0.124, + "step": 18688 + }, + { + "epoch": 0.98, + "learning_rate": 6.283277584664182e-10, + "loss": 0.1045, + "step": 18689 + }, + { + "epoch": 0.98, + "learning_rate": 6.256525753209896e-10, + "loss": 0.1041, + "step": 18690 + }, + { + "epoch": 0.98, + "learning_rate": 6.229830903713873e-10, + "loss": 0.1362, + "step": 18691 + }, + { + "epoch": 0.98, + "learning_rate": 6.203193036939058e-10, + "loss": 0.1116, + "step": 18692 + }, + { + "epoch": 0.98, + "learning_rate": 6.176612153646621e-10, + "loss": 0.1276, + "step": 18693 + }, + { + "epoch": 0.98, + "learning_rate": 6.150088254596175e-10, + "loss": 0.1142, + "step": 18694 + }, + { + "epoch": 0.98, + "learning_rate": 6.123621340546003e-10, + "loss": 0.1145, + "step": 18695 + }, + { + "epoch": 0.98, + "learning_rate": 6.09721141225239e-10, + "loss": 0.1332, + "step": 18696 + }, + { + "epoch": 0.98, + "learning_rate": 6.070858470470286e-10, + "loss": 0.1047, + "step": 18697 + }, + { + "epoch": 0.98, + "learning_rate": 6.044562515952645e-10, + "loss": 0.1962, + "step": 18698 + }, + { + "epoch": 0.98, + "learning_rate": 6.01832354945131e-10, + "loss": 0.1262, + "step": 18699 + }, + { + "epoch": 0.98, + "learning_rate": 5.992141571716125e-10, + "loss": 0.1133, + "step": 18700 + }, + { + "epoch": 0.98, + "learning_rate": 5.966016583495159e-10, + "loss": 0.1547, + "step": 18701 + }, + { + "epoch": 0.98, + "learning_rate": 5.939948585535148e-10, + "loss": 0.1723, + "step": 18702 + }, + { + "epoch": 0.98, + "learning_rate": 5.913937578581274e-10, + "loss": 0.1288, + "step": 18703 + }, + { + "epoch": 0.98, + "learning_rate": 5.887983563376719e-10, + "loss": 0.0955, + "step": 18704 + }, + { + "epoch": 0.98, + "learning_rate": 5.862086540663558e-10, + "loss": 0.1239, + "step": 18705 + }, + { + "epoch": 0.98, + "learning_rate": 5.836246511181864e-10, + "loss": 0.1078, + "step": 18706 + }, + { + "epoch": 0.98, + "learning_rate": 5.810463475669936e-10, + "loss": 0.1618, + "step": 18707 + }, + { + "epoch": 0.98, + "learning_rate": 5.784737434864739e-10, + "loss": 0.1565, + "step": 18708 + }, + { + "epoch": 0.98, + "learning_rate": 5.759068389501686e-10, + "loss": 0.1211, + "step": 18709 + }, + { + "epoch": 0.98, + "learning_rate": 5.733456340314192e-10, + "loss": 0.1441, + "step": 18710 + }, + { + "epoch": 0.98, + "learning_rate": 5.707901288034555e-10, + "loss": 0.1261, + "step": 18711 + }, + { + "epoch": 0.98, + "learning_rate": 5.682403233392863e-10, + "loss": 0.1617, + "step": 18712 + }, + { + "epoch": 0.98, + "learning_rate": 5.656962177117863e-10, + "loss": 0.1187, + "step": 18713 + }, + { + "epoch": 0.98, + "learning_rate": 5.631578119936975e-10, + "loss": 0.1269, + "step": 18714 + }, + { + "epoch": 0.98, + "learning_rate": 5.606251062575396e-10, + "loss": 0.1374, + "step": 18715 + }, + { + "epoch": 0.98, + "learning_rate": 5.580981005757213e-10, + "loss": 0.1084, + "step": 18716 + }, + { + "epoch": 0.98, + "learning_rate": 5.555767950204293e-10, + "loss": 0.1333, + "step": 18717 + }, + { + "epoch": 0.98, + "learning_rate": 5.530611896637616e-10, + "loss": 0.1341, + "step": 18718 + }, + { + "epoch": 0.98, + "learning_rate": 5.505512845775939e-10, + "loss": 0.114, + "step": 18719 + }, + { + "epoch": 0.98, + "learning_rate": 5.48047079833669e-10, + "loss": 0.1055, + "step": 18720 + }, + { + "epoch": 0.98, + "learning_rate": 5.455485755035516e-10, + "loss": 0.1058, + "step": 18721 + }, + { + "epoch": 0.98, + "learning_rate": 5.430557716586515e-10, + "loss": 0.1231, + "step": 18722 + }, + { + "epoch": 0.98, + "learning_rate": 5.405686683702227e-10, + "loss": 0.1588, + "step": 18723 + }, + { + "epoch": 0.98, + "learning_rate": 5.380872657093194e-10, + "loss": 0.1321, + "step": 18724 + }, + { + "epoch": 0.98, + "learning_rate": 5.356115637469072e-10, + "loss": 0.1321, + "step": 18725 + }, + { + "epoch": 0.98, + "learning_rate": 5.331415625537072e-10, + "loss": 0.1467, + "step": 18726 + }, + { + "epoch": 0.98, + "learning_rate": 5.306772622003298e-10, + "loss": 0.0991, + "step": 18727 + }, + { + "epoch": 0.98, + "learning_rate": 5.28218662757185e-10, + "loss": 0.1403, + "step": 18728 + }, + { + "epoch": 0.98, + "learning_rate": 5.257657642945723e-10, + "loss": 0.1579, + "step": 18729 + }, + { + "epoch": 0.98, + "learning_rate": 5.23318566882569e-10, + "loss": 0.1103, + "step": 18730 + }, + { + "epoch": 0.98, + "learning_rate": 5.208770705911192e-10, + "loss": 0.1629, + "step": 18731 + }, + { + "epoch": 0.98, + "learning_rate": 5.184412754900336e-10, + "loss": 0.1852, + "step": 18732 + }, + { + "epoch": 0.98, + "learning_rate": 5.160111816488788e-10, + "loss": 0.1264, + "step": 18733 + }, + { + "epoch": 0.98, + "learning_rate": 5.135867891371548e-10, + "loss": 0.1256, + "step": 18734 + }, + { + "epoch": 0.98, + "learning_rate": 5.111680980241173e-10, + "loss": 0.1447, + "step": 18735 + }, + { + "epoch": 0.98, + "learning_rate": 5.08755108378911e-10, + "loss": 0.1681, + "step": 18736 + }, + { + "epoch": 0.98, + "learning_rate": 5.063478202704807e-10, + "loss": 0.1111, + "step": 18737 + }, + { + "epoch": 0.98, + "learning_rate": 5.039462337676381e-10, + "loss": 0.0891, + "step": 18738 + }, + { + "epoch": 0.98, + "learning_rate": 5.015503489390393e-10, + "loss": 0.0968, + "step": 18739 + }, + { + "epoch": 0.98, + "learning_rate": 4.99160165853163e-10, + "loss": 0.1881, + "step": 18740 + }, + { + "epoch": 0.98, + "learning_rate": 4.967756845782877e-10, + "loss": 0.1467, + "step": 18741 + }, + { + "epoch": 0.98, + "learning_rate": 4.943969051825592e-10, + "loss": 0.1437, + "step": 18742 + }, + { + "epoch": 0.98, + "learning_rate": 4.920238277340116e-10, + "loss": 0.1113, + "step": 18743 + }, + { + "epoch": 0.98, + "learning_rate": 4.896564523004353e-10, + "loss": 0.1573, + "step": 18744 + }, + { + "epoch": 0.98, + "learning_rate": 4.872947789494874e-10, + "loss": 0.1433, + "step": 18745 + }, + { + "epoch": 0.98, + "learning_rate": 4.849388077486915e-10, + "loss": 0.1453, + "step": 18746 + }, + { + "epoch": 0.98, + "learning_rate": 4.825885387653495e-10, + "loss": 0.1324, + "step": 18747 + }, + { + "epoch": 0.98, + "learning_rate": 4.802439720666518e-10, + "loss": 0.1634, + "step": 18748 + }, + { + "epoch": 0.98, + "learning_rate": 4.779051077195895e-10, + "loss": 0.123, + "step": 18749 + }, + { + "epoch": 0.98, + "learning_rate": 4.755719457910424e-10, + "loss": 0.1046, + "step": 18750 + }, + { + "epoch": 0.98, + "learning_rate": 4.732444863476681e-10, + "loss": 0.1288, + "step": 18751 + }, + { + "epoch": 0.98, + "learning_rate": 4.709227294559915e-10, + "loss": 0.2238, + "step": 18752 + }, + { + "epoch": 0.98, + "learning_rate": 4.686066751823814e-10, + "loss": 0.1094, + "step": 18753 + }, + { + "epoch": 0.98, + "learning_rate": 4.662963235930073e-10, + "loss": 0.1477, + "step": 18754 + }, + { + "epoch": 0.98, + "learning_rate": 4.639916747539052e-10, + "loss": 0.1112, + "step": 18755 + }, + { + "epoch": 0.98, + "learning_rate": 4.616927287309558e-10, + "loss": 0.1303, + "step": 18756 + }, + { + "epoch": 0.98, + "learning_rate": 4.5939948558986194e-10, + "loss": 0.1435, + "step": 18757 + }, + { + "epoch": 0.98, + "learning_rate": 4.5711194539617135e-10, + "loss": 0.1641, + "step": 18758 + }, + { + "epoch": 0.98, + "learning_rate": 4.5483010821523173e-10, + "loss": 0.1805, + "step": 18759 + }, + { + "epoch": 0.98, + "learning_rate": 4.5255397411227967e-10, + "loss": 0.1586, + "step": 18760 + }, + { + "epoch": 0.98, + "learning_rate": 4.502835431523744e-10, + "loss": 0.1591, + "step": 18761 + }, + { + "epoch": 0.98, + "learning_rate": 4.4801881540041943e-10, + "loss": 0.1243, + "step": 18762 + }, + { + "epoch": 0.98, + "learning_rate": 4.457597909210964e-10, + "loss": 0.1539, + "step": 18763 + }, + { + "epoch": 0.98, + "learning_rate": 4.4350646977899806e-10, + "loss": 0.1348, + "step": 18764 + }, + { + "epoch": 0.98, + "learning_rate": 4.412588520385174e-10, + "loss": 0.1162, + "step": 18765 + }, + { + "epoch": 0.98, + "learning_rate": 4.3901693776391415e-10, + "loss": 0.12, + "step": 18766 + }, + { + "epoch": 0.98, + "learning_rate": 4.367807270192259e-10, + "loss": 0.1232, + "step": 18767 + }, + { + "epoch": 0.98, + "learning_rate": 4.3455021986837924e-10, + "loss": 0.1352, + "step": 18768 + }, + { + "epoch": 0.98, + "learning_rate": 4.323254163751455e-10, + "loss": 0.1371, + "step": 18769 + }, + { + "epoch": 0.98, + "learning_rate": 4.3010631660309604e-10, + "loss": 0.1437, + "step": 18770 + }, + { + "epoch": 0.98, + "learning_rate": 4.2789292061564674e-10, + "loss": 0.1161, + "step": 18771 + }, + { + "epoch": 0.98, + "learning_rate": 4.2568522847603595e-10, + "loss": 0.1138, + "step": 18772 + }, + { + "epoch": 0.98, + "learning_rate": 4.234832402474131e-10, + "loss": 0.1631, + "step": 18773 + }, + { + "epoch": 0.98, + "learning_rate": 4.212869559926613e-10, + "loss": 0.154, + "step": 18774 + }, + { + "epoch": 0.98, + "learning_rate": 4.190963757745969e-10, + "loss": 0.1211, + "step": 18775 + }, + { + "epoch": 0.98, + "learning_rate": 4.169114996557921e-10, + "loss": 0.1282, + "step": 18776 + }, + { + "epoch": 0.98, + "learning_rate": 4.14732327698708e-10, + "loss": 0.159, + "step": 18777 + }, + { + "epoch": 0.98, + "learning_rate": 4.1255885996560605e-10, + "loss": 0.1047, + "step": 18778 + }, + { + "epoch": 0.98, + "learning_rate": 4.1039109651863633e-10, + "loss": 0.1272, + "step": 18779 + }, + { + "epoch": 0.98, + "learning_rate": 4.0822903741972724e-10, + "loss": 0.1393, + "step": 18780 + }, + { + "epoch": 0.98, + "learning_rate": 4.0607268273069595e-10, + "loss": 0.148, + "step": 18781 + }, + { + "epoch": 0.98, + "learning_rate": 4.039220325131376e-10, + "loss": 0.1113, + "step": 18782 + }, + { + "epoch": 0.98, + "learning_rate": 4.017770868285586e-10, + "loss": 0.1236, + "step": 18783 + }, + { + "epoch": 0.98, + "learning_rate": 3.9963784573824323e-10, + "loss": 0.0907, + "step": 18784 + }, + { + "epoch": 0.98, + "learning_rate": 3.9750430930332035e-10, + "loss": 0.1414, + "step": 18785 + }, + { + "epoch": 0.98, + "learning_rate": 3.9537647758478566e-10, + "loss": 0.1468, + "step": 18786 + }, + { + "epoch": 0.98, + "learning_rate": 3.9325435064345713e-10, + "loss": 0.1312, + "step": 18787 + }, + { + "epoch": 0.98, + "learning_rate": 3.91137928539953e-10, + "loss": 0.1335, + "step": 18788 + }, + { + "epoch": 0.98, + "learning_rate": 3.890272113348025e-10, + "loss": 0.1338, + "step": 18789 + }, + { + "epoch": 0.98, + "learning_rate": 3.869221990882909e-10, + "loss": 0.1267, + "step": 18790 + }, + { + "epoch": 0.98, + "learning_rate": 3.8482289186061444e-10, + "loss": 0.1037, + "step": 18791 + }, + { + "epoch": 0.98, + "learning_rate": 3.8272928971174733e-10, + "loss": 0.0926, + "step": 18792 + }, + { + "epoch": 0.98, + "learning_rate": 3.8064139270155283e-10, + "loss": 0.1288, + "step": 18793 + }, + { + "epoch": 0.98, + "learning_rate": 3.785592008896943e-10, + "loss": 0.1811, + "step": 18794 + }, + { + "epoch": 0.98, + "learning_rate": 3.7648271433565746e-10, + "loss": 0.13, + "step": 18795 + }, + { + "epoch": 0.98, + "learning_rate": 3.744119330988171e-10, + "loss": 0.1448, + "step": 18796 + }, + { + "epoch": 0.98, + "learning_rate": 3.723468572383481e-10, + "loss": 0.1358, + "step": 18797 + }, + { + "epoch": 0.98, + "learning_rate": 3.7028748681326995e-10, + "loss": 0.1215, + "step": 18798 + }, + { + "epoch": 0.98, + "learning_rate": 3.6823382188242437e-10, + "loss": 0.1296, + "step": 18799 + }, + { + "epoch": 0.98, + "learning_rate": 3.661858625045422e-10, + "loss": 0.1636, + "step": 18800 + }, + { + "epoch": 0.98, + "learning_rate": 3.6414360873811e-10, + "loss": 0.1412, + "step": 18801 + }, + { + "epoch": 0.98, + "learning_rate": 3.621070606415477e-10, + "loss": 0.1184, + "step": 18802 + }, + { + "epoch": 0.98, + "learning_rate": 3.6007621827300883e-10, + "loss": 0.1251, + "step": 18803 + }, + { + "epoch": 0.98, + "learning_rate": 3.580510816905802e-10, + "loss": 0.1379, + "step": 18804 + }, + { + "epoch": 0.98, + "learning_rate": 3.560316509521266e-10, + "loss": 0.1305, + "step": 18805 + }, + { + "epoch": 0.98, + "learning_rate": 3.5401792611533533e-10, + "loss": 0.1411, + "step": 18806 + }, + { + "epoch": 0.98, + "learning_rate": 3.5200990723780464e-10, + "loss": 0.1692, + "step": 18807 + }, + { + "epoch": 0.98, + "learning_rate": 3.500075943768888e-10, + "loss": 0.1356, + "step": 18808 + }, + { + "epoch": 0.98, + "learning_rate": 3.480109875898307e-10, + "loss": 0.1232, + "step": 18809 + }, + { + "epoch": 0.98, + "learning_rate": 3.46020086933696e-10, + "loss": 0.1366, + "step": 18810 + }, + { + "epoch": 0.98, + "learning_rate": 3.4403489246537244e-10, + "loss": 0.1867, + "step": 18811 + }, + { + "epoch": 0.98, + "learning_rate": 3.420554042416146e-10, + "loss": 0.1156, + "step": 18812 + }, + { + "epoch": 0.98, + "learning_rate": 3.400816223189773e-10, + "loss": 0.1349, + "step": 18813 + }, + { + "epoch": 0.98, + "learning_rate": 3.3811354675390425e-10, + "loss": 0.1558, + "step": 18814 + }, + { + "epoch": 0.98, + "learning_rate": 3.3615117760261713e-10, + "loss": 0.1593, + "step": 18815 + }, + { + "epoch": 0.98, + "learning_rate": 3.3419451492120444e-10, + "loss": 0.1314, + "step": 18816 + }, + { + "epoch": 0.98, + "learning_rate": 3.32243558765577e-10, + "loss": 0.1523, + "step": 18817 + }, + { + "epoch": 0.98, + "learning_rate": 3.302983091915124e-10, + "loss": 0.1578, + "step": 18818 + }, + { + "epoch": 0.98, + "learning_rate": 3.2835876625461055e-10, + "loss": 0.1148, + "step": 18819 + }, + { + "epoch": 0.98, + "learning_rate": 3.264249300103161e-10, + "loss": 0.1667, + "step": 18820 + }, + { + "epoch": 0.98, + "learning_rate": 3.2449680051385155e-10, + "loss": 0.1351, + "step": 18821 + }, + { + "epoch": 0.98, + "learning_rate": 3.2257437782035047e-10, + "loss": 0.1445, + "step": 18822 + }, + { + "epoch": 0.98, + "learning_rate": 3.2065766198476893e-10, + "loss": 0.1411, + "step": 18823 + }, + { + "epoch": 0.98, + "learning_rate": 3.187466530618632e-10, + "loss": 0.1333, + "step": 18824 + }, + { + "epoch": 0.98, + "learning_rate": 3.168413511062784e-10, + "loss": 0.1178, + "step": 18825 + }, + { + "epoch": 0.98, + "learning_rate": 3.149417561724377e-10, + "loss": 0.1504, + "step": 18826 + }, + { + "epoch": 0.98, + "learning_rate": 3.1304786831467533e-10, + "loss": 0.1256, + "step": 18827 + }, + { + "epoch": 0.98, + "learning_rate": 3.1115968758705925e-10, + "loss": 0.0958, + "step": 18828 + }, + { + "epoch": 0.98, + "learning_rate": 3.0927721404361284e-10, + "loss": 0.1482, + "step": 18829 + }, + { + "epoch": 0.98, + "learning_rate": 3.0740044773809316e-10, + "loss": 0.1262, + "step": 18830 + }, + { + "epoch": 0.98, + "learning_rate": 3.0552938872416836e-10, + "loss": 0.102, + "step": 18831 + }, + { + "epoch": 0.98, + "learning_rate": 3.0366403705528455e-10, + "loss": 0.1253, + "step": 18832 + }, + { + "epoch": 0.98, + "learning_rate": 3.0180439278479907e-10, + "loss": 0.1103, + "step": 18833 + }, + { + "epoch": 0.98, + "learning_rate": 2.99950455965825e-10, + "loss": 0.1539, + "step": 18834 + }, + { + "epoch": 0.98, + "learning_rate": 2.981022266513644e-10, + "loss": 0.1355, + "step": 18835 + }, + { + "epoch": 0.98, + "learning_rate": 2.9625970489421947e-10, + "loss": 0.1371, + "step": 18836 + }, + { + "epoch": 0.98, + "learning_rate": 2.9442289074708136e-10, + "loss": 0.1336, + "step": 18837 + }, + { + "epoch": 0.98, + "learning_rate": 2.9259178426241924e-10, + "loss": 0.123, + "step": 18838 + }, + { + "epoch": 0.98, + "learning_rate": 2.9076638549256904e-10, + "loss": 0.1646, + "step": 18839 + }, + { + "epoch": 0.98, + "learning_rate": 2.889466944897334e-10, + "loss": 0.1296, + "step": 18840 + }, + { + "epoch": 0.98, + "learning_rate": 2.87132711305893e-10, + "loss": 0.1411, + "step": 18841 + }, + { + "epoch": 0.98, + "learning_rate": 2.8532443599289525e-10, + "loss": 0.1295, + "step": 18842 + }, + { + "epoch": 0.98, + "learning_rate": 2.835218686024099e-10, + "loss": 0.1068, + "step": 18843 + }, + { + "epoch": 0.98, + "learning_rate": 2.8172500918597354e-10, + "loss": 0.1452, + "step": 18844 + }, + { + "epoch": 0.98, + "learning_rate": 2.7993385779494505e-10, + "loss": 0.1144, + "step": 18845 + }, + { + "epoch": 0.98, + "learning_rate": 2.7814841448050575e-10, + "loss": 0.1371, + "step": 18846 + }, + { + "epoch": 0.98, + "learning_rate": 2.7636867929368145e-10, + "loss": 0.1253, + "step": 18847 + }, + { + "epoch": 0.98, + "learning_rate": 2.745946522853426e-10, + "loss": 0.1368, + "step": 18848 + }, + { + "epoch": 0.98, + "learning_rate": 2.72826333506182e-10, + "loss": 0.1236, + "step": 18849 + }, + { + "epoch": 0.98, + "learning_rate": 2.7106372300675915e-10, + "loss": 0.1292, + "step": 18850 + }, + { + "epoch": 0.98, + "learning_rate": 2.6930682083743385e-10, + "loss": 0.0828, + "step": 18851 + }, + { + "epoch": 0.98, + "learning_rate": 2.675556270484325e-10, + "loss": 0.1502, + "step": 18852 + }, + { + "epoch": 0.98, + "learning_rate": 2.6581014168978176e-10, + "loss": 0.1297, + "step": 18853 + }, + { + "epoch": 0.98, + "learning_rate": 2.6407036481137513e-10, + "loss": 0.1135, + "step": 18854 + }, + { + "epoch": 0.98, + "learning_rate": 2.623362964629727e-10, + "loss": 0.1405, + "step": 18855 + }, + { + "epoch": 0.98, + "learning_rate": 2.6060793669409055e-10, + "loss": 0.1265, + "step": 18856 + }, + { + "epoch": 0.98, + "learning_rate": 2.5888528555413347e-10, + "loss": 0.1326, + "step": 18857 + }, + { + "epoch": 0.98, + "learning_rate": 2.571683430923288e-10, + "loss": 0.1998, + "step": 18858 + }, + { + "epoch": 0.98, + "learning_rate": 2.5545710935779286e-10, + "loss": 0.128, + "step": 18859 + }, + { + "epoch": 0.98, + "learning_rate": 2.537515843993754e-10, + "loss": 0.1439, + "step": 18860 + }, + { + "epoch": 0.98, + "learning_rate": 2.5205176826585963e-10, + "loss": 0.1426, + "step": 18861 + }, + { + "epoch": 0.98, + "learning_rate": 2.503576610058289e-10, + "loss": 0.1285, + "step": 18862 + }, + { + "epoch": 0.98, + "learning_rate": 2.486692626676667e-10, + "loss": 0.1392, + "step": 18863 + }, + { + "epoch": 0.98, + "learning_rate": 2.469865732996679e-10, + "loss": 0.1245, + "step": 18864 + }, + { + "epoch": 0.98, + "learning_rate": 2.4530959294988275e-10, + "loss": 0.0982, + "step": 18865 + }, + { + "epoch": 0.98, + "learning_rate": 2.4363832166627295e-10, + "loss": 0.1734, + "step": 18866 + }, + { + "epoch": 0.98, + "learning_rate": 2.4197275949660035e-10, + "loss": 0.1501, + "step": 18867 + }, + { + "epoch": 0.98, + "learning_rate": 2.4031290648844905e-10, + "loss": 0.157, + "step": 18868 + }, + { + "epoch": 0.98, + "learning_rate": 2.3865876268927e-10, + "loss": 0.1062, + "step": 18869 + }, + { + "epoch": 0.98, + "learning_rate": 2.370103281463587e-10, + "loss": 0.1434, + "step": 18870 + }, + { + "epoch": 0.98, + "learning_rate": 2.353676029067886e-10, + "loss": 0.152, + "step": 18871 + }, + { + "epoch": 0.99, + "learning_rate": 2.3373058701754433e-10, + "loss": 0.1092, + "step": 18872 + }, + { + "epoch": 0.99, + "learning_rate": 2.3209928052538853e-10, + "loss": 0.1711, + "step": 18873 + }, + { + "epoch": 0.99, + "learning_rate": 2.304736834769505e-10, + "loss": 0.1368, + "step": 18874 + }, + { + "epoch": 0.99, + "learning_rate": 2.2885379591868203e-10, + "loss": 0.0979, + "step": 18875 + }, + { + "epoch": 0.99, + "learning_rate": 2.272396178969016e-10, + "loss": 0.1603, + "step": 18876 + }, + { + "epoch": 0.99, + "learning_rate": 2.2563114945775008e-10, + "loss": 0.1312, + "step": 18877 + }, + { + "epoch": 0.99, + "learning_rate": 2.240283906471685e-10, + "loss": 0.1586, + "step": 18878 + }, + { + "epoch": 0.99, + "learning_rate": 2.2243134151096466e-10, + "loss": 0.1269, + "step": 18879 + }, + { + "epoch": 0.99, + "learning_rate": 2.2084000209481312e-10, + "loss": 0.1922, + "step": 18880 + }, + { + "epoch": 0.99, + "learning_rate": 2.1925437244416645e-10, + "loss": 0.1366, + "step": 18881 + }, + { + "epoch": 0.99, + "learning_rate": 2.1767445260436612e-10, + "loss": 0.11, + "step": 18882 + }, + { + "epoch": 0.99, + "learning_rate": 2.1610024262053162e-10, + "loss": 0.1311, + "step": 18883 + }, + { + "epoch": 0.99, + "learning_rate": 2.1453174253769358e-10, + "loss": 0.1145, + "step": 18884 + }, + { + "epoch": 0.99, + "learning_rate": 2.129689524006606e-10, + "loss": 0.1473, + "step": 18885 + }, + { + "epoch": 0.99, + "learning_rate": 2.1141187225408586e-10, + "loss": 0.138, + "step": 18886 + }, + { + "epoch": 0.99, + "learning_rate": 2.098605021425115e-10, + "loss": 0.1382, + "step": 18887 + }, + { + "epoch": 0.99, + "learning_rate": 2.0831484211023543e-10, + "loss": 0.1449, + "step": 18888 + }, + { + "epoch": 0.99, + "learning_rate": 2.0677489220144452e-10, + "loss": 0.1258, + "step": 18889 + }, + { + "epoch": 0.99, + "learning_rate": 2.0524065246017019e-10, + "loss": 0.1104, + "step": 18890 + }, + { + "epoch": 0.99, + "learning_rate": 2.037121229302219e-10, + "loss": 0.17, + "step": 18891 + }, + { + "epoch": 0.99, + "learning_rate": 2.0218930365534236e-10, + "loss": 0.1317, + "step": 18892 + }, + { + "epoch": 0.99, + "learning_rate": 2.0067219467898578e-10, + "loss": 0.1339, + "step": 18893 + }, + { + "epoch": 0.99, + "learning_rate": 1.9916079604458403e-10, + "loss": 0.1248, + "step": 18894 + }, + { + "epoch": 0.99, + "learning_rate": 1.9765510779528038e-10, + "loss": 0.1221, + "step": 18895 + }, + { + "epoch": 0.99, + "learning_rate": 1.961551299741293e-10, + "loss": 0.1298, + "step": 18896 + }, + { + "epoch": 0.99, + "learning_rate": 1.946608626239854e-10, + "loss": 0.1434, + "step": 18897 + }, + { + "epoch": 0.99, + "learning_rate": 1.9317230578757006e-10, + "loss": 0.1554, + "step": 18898 + }, + { + "epoch": 0.99, + "learning_rate": 1.916894595074492e-10, + "loss": 0.1234, + "step": 18899 + }, + { + "epoch": 0.99, + "learning_rate": 1.9021232382594453e-10, + "loss": 0.1224, + "step": 18900 + }, + { + "epoch": 0.99, + "learning_rate": 1.8874089878531119e-10, + "loss": 0.1507, + "step": 18901 + }, + { + "epoch": 0.99, + "learning_rate": 1.8727518442760436e-10, + "loss": 0.1408, + "step": 18902 + }, + { + "epoch": 0.99, + "learning_rate": 1.858151807947239e-10, + "loss": 0.1708, + "step": 18903 + }, + { + "epoch": 0.99, + "learning_rate": 1.843608879283476e-10, + "loss": 0.1159, + "step": 18904 + }, + { + "epoch": 0.99, + "learning_rate": 1.8291230587010875e-10, + "loss": 0.1187, + "step": 18905 + }, + { + "epoch": 0.99, + "learning_rate": 1.8146943466135211e-10, + "loss": 0.114, + "step": 18906 + }, + { + "epoch": 0.99, + "learning_rate": 1.8003227434333357e-10, + "loss": 0.1187, + "step": 18907 + }, + { + "epoch": 0.99, + "learning_rate": 1.7860082495715356e-10, + "loss": 0.1256, + "step": 18908 + }, + { + "epoch": 0.99, + "learning_rate": 1.771750865436683e-10, + "loss": 0.1369, + "step": 18909 + }, + { + "epoch": 0.99, + "learning_rate": 1.757550591436896e-10, + "loss": 0.1421, + "step": 18910 + }, + { + "epoch": 0.99, + "learning_rate": 1.7434074279776278e-10, + "loss": 0.1733, + "step": 18911 + }, + { + "epoch": 0.99, + "learning_rate": 1.729321375463e-10, + "loss": 0.1107, + "step": 18912 + }, + { + "epoch": 0.99, + "learning_rate": 1.7152924342960228e-10, + "loss": 0.0907, + "step": 18913 + }, + { + "epoch": 0.99, + "learning_rate": 1.7013206048772654e-10, + "loss": 0.1028, + "step": 18914 + }, + { + "epoch": 0.99, + "learning_rate": 1.6874058876061858e-10, + "loss": 0.1794, + "step": 18915 + }, + { + "epoch": 0.99, + "learning_rate": 1.673548282880466e-10, + "loss": 0.145, + "step": 18916 + }, + { + "epoch": 0.99, + "learning_rate": 1.6597477910962332e-10, + "loss": 0.1205, + "step": 18917 + }, + { + "epoch": 0.99, + "learning_rate": 1.646004412647839e-10, + "loss": 0.1174, + "step": 18918 + }, + { + "epoch": 0.99, + "learning_rate": 1.6323181479280802e-10, + "loss": 0.1234, + "step": 18919 + }, + { + "epoch": 0.99, + "learning_rate": 1.6186889973279772e-10, + "loss": 0.1373, + "step": 18920 + }, + { + "epoch": 0.99, + "learning_rate": 1.6051169612374404e-10, + "loss": 0.1336, + "step": 18921 + }, + { + "epoch": 0.99, + "learning_rate": 1.591602040043938e-10, + "loss": 0.1354, + "step": 18922 + }, + { + "epoch": 0.99, + "learning_rate": 1.5781442341338268e-10, + "loss": 0.1174, + "step": 18923 + }, + { + "epoch": 0.99, + "learning_rate": 1.5647435438921332e-10, + "loss": 0.1469, + "step": 18924 + }, + { + "epoch": 0.99, + "learning_rate": 1.551399969701217e-10, + "loss": 0.1238, + "step": 18925 + }, + { + "epoch": 0.99, + "learning_rate": 1.5381135119429955e-10, + "loss": 0.1186, + "step": 18926 + }, + { + "epoch": 0.99, + "learning_rate": 1.5248841709969428e-10, + "loss": 0.1304, + "step": 18927 + }, + { + "epoch": 0.99, + "learning_rate": 1.511711947240979e-10, + "loss": 0.1246, + "step": 18928 + }, + { + "epoch": 0.99, + "learning_rate": 1.4985968410519133e-10, + "loss": 0.1431, + "step": 18929 + }, + { + "epoch": 0.99, + "learning_rate": 1.4855388528045575e-10, + "loss": 0.1265, + "step": 18930 + }, + { + "epoch": 0.99, + "learning_rate": 1.4725379828719464e-10, + "loss": 0.1278, + "step": 18931 + }, + { + "epoch": 0.99, + "learning_rate": 1.4595942316257825e-10, + "loss": 0.1319, + "step": 18932 + }, + { + "epoch": 0.99, + "learning_rate": 1.4467075994357703e-10, + "loss": 0.1275, + "step": 18933 + }, + { + "epoch": 0.99, + "learning_rate": 1.4338780866705035e-10, + "loss": 0.1127, + "step": 18934 + }, + { + "epoch": 0.99, + "learning_rate": 1.421105693696356e-10, + "loss": 0.1498, + "step": 18935 + }, + { + "epoch": 0.99, + "learning_rate": 1.4083904208785912e-10, + "loss": 0.1142, + "step": 18936 + }, + { + "epoch": 0.99, + "learning_rate": 1.395732268580696e-10, + "loss": 0.1133, + "step": 18937 + }, + { + "epoch": 0.99, + "learning_rate": 1.383131237164159e-10, + "loss": 0.1222, + "step": 18938 + }, + { + "epoch": 0.99, + "learning_rate": 1.3705873269893586e-10, + "loss": 0.148, + "step": 18939 + }, + { + "epoch": 0.99, + "learning_rate": 1.358100538414675e-10, + "loss": 0.1281, + "step": 18940 + }, + { + "epoch": 0.99, + "learning_rate": 1.3456708717971555e-10, + "loss": 0.109, + "step": 18941 + }, + { + "epoch": 0.99, + "learning_rate": 1.3332983274918498e-10, + "loss": 0.1508, + "step": 18942 + }, + { + "epoch": 0.99, + "learning_rate": 1.3209829058524747e-10, + "loss": 0.0964, + "step": 18943 + }, + { + "epoch": 0.99, + "learning_rate": 1.308724607230749e-10, + "loss": 0.1042, + "step": 18944 + }, + { + "epoch": 0.99, + "learning_rate": 1.2965234319775032e-10, + "loss": 0.1212, + "step": 18945 + }, + { + "epoch": 0.99, + "learning_rate": 1.2843793804411252e-10, + "loss": 0.1128, + "step": 18946 + }, + { + "epoch": 0.99, + "learning_rate": 1.2722924529686706e-10, + "loss": 0.1219, + "step": 18947 + }, + { + "epoch": 0.99, + "learning_rate": 1.2602626499058633e-10, + "loss": 0.1402, + "step": 18948 + }, + { + "epoch": 0.99, + "learning_rate": 1.2482899715962058e-10, + "loss": 0.1504, + "step": 18949 + }, + { + "epoch": 0.99, + "learning_rate": 1.2363744183820912e-10, + "loss": 0.1086, + "step": 18950 + }, + { + "epoch": 0.99, + "learning_rate": 1.2245159906039137e-10, + "loss": 0.1231, + "step": 18951 + }, + { + "epoch": 0.99, + "learning_rate": 1.2127146886007356e-10, + "loss": 0.1448, + "step": 18952 + }, + { + "epoch": 0.99, + "learning_rate": 1.2009705127098423e-10, + "loss": 0.1265, + "step": 18953 + }, + { + "epoch": 0.99, + "learning_rate": 1.1892834632667437e-10, + "loss": 0.1276, + "step": 18954 + }, + { + "epoch": 0.99, + "learning_rate": 1.1776535406053945e-10, + "loss": 0.1339, + "step": 18955 + }, + { + "epoch": 0.99, + "learning_rate": 1.1660807450584176e-10, + "loss": 0.1166, + "step": 18956 + }, + { + "epoch": 0.99, + "learning_rate": 1.1545650769566595e-10, + "loss": 0.1611, + "step": 18957 + }, + { + "epoch": 0.99, + "learning_rate": 1.1431065366289683e-10, + "loss": 0.1222, + "step": 18958 + }, + { + "epoch": 0.99, + "learning_rate": 1.1317051244028596e-10, + "loss": 0.1003, + "step": 18959 + }, + { + "epoch": 0.99, + "learning_rate": 1.1203608406042952e-10, + "loss": 0.1289, + "step": 18960 + }, + { + "epoch": 0.99, + "learning_rate": 1.10907368555746e-10, + "loss": 0.1009, + "step": 18961 + }, + { + "epoch": 0.99, + "learning_rate": 1.097843659584985e-10, + "loss": 0.127, + "step": 18962 + }, + { + "epoch": 0.99, + "learning_rate": 1.0866707630077243e-10, + "loss": 0.1332, + "step": 18963 + }, + { + "epoch": 0.99, + "learning_rate": 1.0755549961452005e-10, + "loss": 0.1739, + "step": 18964 + }, + { + "epoch": 0.99, + "learning_rate": 1.0644963593149369e-10, + "loss": 0.1122, + "step": 18965 + }, + { + "epoch": 0.99, + "learning_rate": 1.0534948528329036e-10, + "loss": 0.1244, + "step": 18966 + }, + { + "epoch": 0.99, + "learning_rate": 1.0425504770137372e-10, + "loss": 0.1335, + "step": 18967 + }, + { + "epoch": 0.99, + "learning_rate": 1.0316632321702989e-10, + "loss": 0.1489, + "step": 18968 + }, + { + "epoch": 0.99, + "learning_rate": 1.0208331186134511e-10, + "loss": 0.1388, + "step": 18969 + }, + { + "epoch": 0.99, + "learning_rate": 1.0100601366527239e-10, + "loss": 0.1478, + "step": 18970 + }, + { + "epoch": 0.99, + "learning_rate": 9.993442865963153e-11, + "loss": 0.1627, + "step": 18971 + }, + { + "epoch": 0.99, + "learning_rate": 9.886855687504247e-11, + "loss": 0.1003, + "step": 18972 + }, + { + "epoch": 0.99, + "learning_rate": 9.780839834194754e-11, + "loss": 0.1305, + "step": 18973 + }, + { + "epoch": 0.99, + "learning_rate": 9.675395309067802e-11, + "loss": 0.1058, + "step": 18974 + }, + { + "epoch": 0.99, + "learning_rate": 9.570522115132096e-11, + "loss": 0.1476, + "step": 18975 + }, + { + "epoch": 0.99, + "learning_rate": 9.466220255389679e-11, + "loss": 0.1189, + "step": 18976 + }, + { + "epoch": 0.99, + "learning_rate": 9.362489732820389e-11, + "loss": 0.163, + "step": 18977 + }, + { + "epoch": 0.99, + "learning_rate": 9.259330550388523e-11, + "loss": 0.1116, + "step": 18978 + }, + { + "epoch": 0.99, + "learning_rate": 9.15674271104061e-11, + "loss": 0.124, + "step": 18979 + }, + { + "epoch": 0.99, + "learning_rate": 9.054726217709863e-11, + "loss": 0.1293, + "step": 18980 + }, + { + "epoch": 0.99, + "learning_rate": 8.953281073313945e-11, + "loss": 0.1408, + "step": 18981 + }, + { + "epoch": 0.99, + "learning_rate": 8.85240728075054e-11, + "loss": 0.126, + "step": 18982 + }, + { + "epoch": 0.99, + "learning_rate": 8.752104842901787e-11, + "loss": 0.1724, + "step": 18983 + }, + { + "epoch": 0.99, + "learning_rate": 8.6523737626365e-11, + "loss": 0.1298, + "step": 18984 + }, + { + "epoch": 0.99, + "learning_rate": 8.553214042801293e-11, + "loss": 0.1502, + "step": 18985 + }, + { + "epoch": 0.99, + "learning_rate": 8.454625686236116e-11, + "loss": 0.1689, + "step": 18986 + }, + { + "epoch": 0.99, + "learning_rate": 8.356608695752054e-11, + "loss": 0.151, + "step": 18987 + }, + { + "epoch": 0.99, + "learning_rate": 8.25916307415575e-11, + "loss": 0.1366, + "step": 18988 + }, + { + "epoch": 0.99, + "learning_rate": 8.162288824229424e-11, + "loss": 0.1414, + "step": 18989 + }, + { + "epoch": 0.99, + "learning_rate": 8.065985948741972e-11, + "loss": 0.144, + "step": 18990 + }, + { + "epoch": 0.99, + "learning_rate": 7.970254450446745e-11, + "loss": 0.1895, + "step": 18991 + }, + { + "epoch": 0.99, + "learning_rate": 7.875094332079335e-11, + "loss": 0.142, + "step": 18992 + }, + { + "epoch": 0.99, + "learning_rate": 7.780505596359788e-11, + "loss": 0.1018, + "step": 18993 + }, + { + "epoch": 0.99, + "learning_rate": 7.686488245990386e-11, + "loss": 0.1324, + "step": 18994 + }, + { + "epoch": 0.99, + "learning_rate": 7.59304228366009e-11, + "loss": 0.1317, + "step": 18995 + }, + { + "epoch": 0.99, + "learning_rate": 7.500167712037875e-11, + "loss": 0.1299, + "step": 18996 + }, + { + "epoch": 0.99, + "learning_rate": 7.407864533779395e-11, + "loss": 0.139, + "step": 18997 + }, + { + "epoch": 0.99, + "learning_rate": 7.316132751520321e-11, + "loss": 0.1244, + "step": 18998 + }, + { + "epoch": 0.99, + "learning_rate": 7.224972367885219e-11, + "loss": 0.1113, + "step": 18999 + }, + { + "epoch": 0.99, + "learning_rate": 7.134383385478671e-11, + "loss": 0.1657, + "step": 19000 + }, + { + "epoch": 0.99, + "learning_rate": 7.044365806889719e-11, + "loss": 0.1448, + "step": 19001 + }, + { + "epoch": 0.99, + "learning_rate": 6.954919634691858e-11, + "loss": 0.1337, + "step": 19002 + }, + { + "epoch": 0.99, + "learning_rate": 6.866044871438603e-11, + "loss": 0.1582, + "step": 19003 + }, + { + "epoch": 0.99, + "learning_rate": 6.777741519672365e-11, + "loss": 0.1048, + "step": 19004 + }, + { + "epoch": 0.99, + "learning_rate": 6.690009581915567e-11, + "loss": 0.1517, + "step": 19005 + }, + { + "epoch": 0.99, + "learning_rate": 6.602849060677318e-11, + "loss": 0.0989, + "step": 19006 + }, + { + "epoch": 0.99, + "learning_rate": 6.516259958448955e-11, + "loss": 0.1334, + "step": 19007 + }, + { + "epoch": 0.99, + "learning_rate": 6.430242277701837e-11, + "loss": 0.1357, + "step": 19008 + }, + { + "epoch": 0.99, + "learning_rate": 6.344796020898435e-11, + "loss": 0.117, + "step": 19009 + }, + { + "epoch": 0.99, + "learning_rate": 6.259921190479023e-11, + "loss": 0.1308, + "step": 19010 + }, + { + "epoch": 0.99, + "learning_rate": 6.175617788868326e-11, + "loss": 0.1186, + "step": 19011 + }, + { + "epoch": 0.99, + "learning_rate": 6.091885818477749e-11, + "loss": 0.1719, + "step": 19012 + }, + { + "epoch": 0.99, + "learning_rate": 6.008725281698712e-11, + "loss": 0.0931, + "step": 19013 + }, + { + "epoch": 0.99, + "learning_rate": 5.926136180909313e-11, + "loss": 0.1308, + "step": 19014 + }, + { + "epoch": 0.99, + "learning_rate": 5.844118518469887e-11, + "loss": 0.1141, + "step": 19015 + }, + { + "epoch": 0.99, + "learning_rate": 5.762672296723003e-11, + "loss": 0.1148, + "step": 19016 + }, + { + "epoch": 0.99, + "learning_rate": 5.6817975179979106e-11, + "loss": 0.1404, + "step": 19017 + }, + { + "epoch": 0.99, + "learning_rate": 5.6014941846060924e-11, + "loss": 0.1667, + "step": 19018 + }, + { + "epoch": 0.99, + "learning_rate": 5.521762298841271e-11, + "loss": 0.1567, + "step": 19019 + }, + { + "epoch": 0.99, + "learning_rate": 5.442601862983842e-11, + "loss": 0.141, + "step": 19020 + }, + { + "epoch": 0.99, + "learning_rate": 5.364012879296442e-11, + "loss": 0.1649, + "step": 19021 + }, + { + "epoch": 0.99, + "learning_rate": 5.285995350023942e-11, + "loss": 0.1618, + "step": 19022 + }, + { + "epoch": 0.99, + "learning_rate": 5.2085492773956684e-11, + "loss": 0.1223, + "step": 19023 + }, + { + "epoch": 0.99, + "learning_rate": 5.131674663625407e-11, + "loss": 0.1254, + "step": 19024 + }, + { + "epoch": 0.99, + "learning_rate": 5.055371510911399e-11, + "loss": 0.1145, + "step": 19025 + }, + { + "epoch": 0.99, + "learning_rate": 4.979639821434123e-11, + "loss": 0.1556, + "step": 19026 + }, + { + "epoch": 0.99, + "learning_rate": 4.904479597356292e-11, + "loss": 0.1202, + "step": 19027 + }, + { + "epoch": 0.99, + "learning_rate": 4.8298908408272996e-11, + "loss": 0.1701, + "step": 19028 + }, + { + "epoch": 0.99, + "learning_rate": 4.7558735539809933e-11, + "loss": 0.1503, + "step": 19029 + }, + { + "epoch": 0.99, + "learning_rate": 4.6824277389290176e-11, + "loss": 0.1287, + "step": 19030 + }, + { + "epoch": 0.99, + "learning_rate": 4.609553397771915e-11, + "loss": 0.1517, + "step": 19031 + }, + { + "epoch": 0.99, + "learning_rate": 4.537250532594683e-11, + "loss": 0.1429, + "step": 19032 + }, + { + "epoch": 0.99, + "learning_rate": 4.465519145460117e-11, + "loss": 0.1504, + "step": 19033 + }, + { + "epoch": 0.99, + "learning_rate": 4.394359238419909e-11, + "loss": 0.1407, + "step": 19034 + }, + { + "epoch": 0.99, + "learning_rate": 4.323770813507987e-11, + "loss": 0.1173, + "step": 19035 + }, + { + "epoch": 0.99, + "learning_rate": 4.253753872742738e-11, + "loss": 0.1822, + "step": 19036 + }, + { + "epoch": 0.99, + "learning_rate": 4.184308418124782e-11, + "loss": 0.1225, + "step": 19037 + }, + { + "epoch": 0.99, + "learning_rate": 4.1154344516369787e-11, + "loss": 0.1271, + "step": 19038 + }, + { + "epoch": 0.99, + "learning_rate": 4.0471319752510834e-11, + "loss": 0.1277, + "step": 19039 + }, + { + "epoch": 0.99, + "learning_rate": 3.979400990916648e-11, + "loss": 0.1294, + "step": 19040 + }, + { + "epoch": 0.99, + "learning_rate": 3.912241500569902e-11, + "loss": 0.1715, + "step": 19041 + }, + { + "epoch": 0.99, + "learning_rate": 3.8456535061293093e-11, + "loss": 0.1259, + "step": 19042 + }, + { + "epoch": 0.99, + "learning_rate": 3.779637009500014e-11, + "loss": 0.1318, + "step": 19043 + }, + { + "epoch": 0.99, + "learning_rate": 3.714192012569395e-11, + "loss": 0.1228, + "step": 19044 + }, + { + "epoch": 0.99, + "learning_rate": 3.649318517204847e-11, + "loss": 0.1434, + "step": 19045 + }, + { + "epoch": 0.99, + "learning_rate": 3.585016525262663e-11, + "loss": 0.1348, + "step": 19046 + }, + { + "epoch": 0.99, + "learning_rate": 3.521286038581373e-11, + "loss": 0.1266, + "step": 19047 + }, + { + "epoch": 0.99, + "learning_rate": 3.458127058979521e-11, + "loss": 0.1556, + "step": 19048 + }, + { + "epoch": 0.99, + "learning_rate": 3.395539588264551e-11, + "loss": 0.1515, + "step": 19049 + }, + { + "epoch": 0.99, + "learning_rate": 3.333523628223922e-11, + "loss": 0.1378, + "step": 19050 + }, + { + "epoch": 0.99, + "learning_rate": 3.2720791806317704e-11, + "loss": 0.1637, + "step": 19051 + }, + { + "epoch": 0.99, + "learning_rate": 3.211206247242249e-11, + "loss": 0.1238, + "step": 19052 + }, + { + "epoch": 0.99, + "learning_rate": 3.150904829796186e-11, + "loss": 0.1256, + "step": 19053 + }, + { + "epoch": 0.99, + "learning_rate": 3.091174930016649e-11, + "loss": 0.1259, + "step": 19054 + }, + { + "epoch": 0.99, + "learning_rate": 3.0320165496111605e-11, + "loss": 0.1128, + "step": 19055 + }, + { + "epoch": 0.99, + "learning_rate": 2.97342969026948e-11, + "loss": 0.0935, + "step": 19056 + }, + { + "epoch": 0.99, + "learning_rate": 2.915414353668044e-11, + "loss": 0.1225, + "step": 19057 + }, + { + "epoch": 0.99, + "learning_rate": 2.857970541463306e-11, + "loss": 0.1456, + "step": 19058 + }, + { + "epoch": 0.99, + "learning_rate": 2.8010982552961748e-11, + "loss": 0.1064, + "step": 19059 + }, + { + "epoch": 0.99, + "learning_rate": 2.7447974967942377e-11, + "loss": 0.1265, + "step": 19060 + }, + { + "epoch": 0.99, + "learning_rate": 2.689068267567318e-11, + "loss": 0.1259, + "step": 19061 + }, + { + "epoch": 0.99, + "learning_rate": 2.6339105692052556e-11, + "loss": 0.1252, + "step": 19062 + }, + { + "epoch": 1.0, + "learning_rate": 2.5793244032845663e-11, + "loss": 0.1333, + "step": 19063 + }, + { + "epoch": 1.0, + "learning_rate": 2.5253097713684444e-11, + "loss": 0.1417, + "step": 19064 + }, + { + "epoch": 1.0, + "learning_rate": 2.47186667499788e-11, + "loss": 0.1668, + "step": 19065 + }, + { + "epoch": 1.0, + "learning_rate": 2.4189951157005394e-11, + "loss": 0.1467, + "step": 19066 + }, + { + "epoch": 1.0, + "learning_rate": 2.3666950949885468e-11, + "loss": 0.1357, + "step": 19067 + }, + { + "epoch": 1.0, + "learning_rate": 2.3149666143584822e-11, + "loss": 0.1418, + "step": 19068 + }, + { + "epoch": 1.0, + "learning_rate": 2.2638096752847224e-11, + "loss": 0.1228, + "step": 19069 + }, + { + "epoch": 1.0, + "learning_rate": 2.2132242792305412e-11, + "loss": 0.1638, + "step": 19070 + }, + { + "epoch": 1.0, + "learning_rate": 2.163210427643669e-11, + "loss": 0.1431, + "step": 19071 + }, + { + "epoch": 1.0, + "learning_rate": 2.1137681219496327e-11, + "loss": 0.1052, + "step": 19072 + }, + { + "epoch": 1.0, + "learning_rate": 2.064897363567297e-11, + "loss": 0.1219, + "step": 19073 + }, + { + "epoch": 1.0, + "learning_rate": 2.016598153888882e-11, + "loss": 0.1237, + "step": 19074 + }, + { + "epoch": 1.0, + "learning_rate": 1.9688704942955047e-11, + "loss": 0.1073, + "step": 19075 + }, + { + "epoch": 1.0, + "learning_rate": 1.9217143861527395e-11, + "loss": 0.1264, + "step": 19076 + }, + { + "epoch": 1.0, + "learning_rate": 1.875129830806177e-11, + "loss": 0.1168, + "step": 19077 + }, + { + "epoch": 1.0, + "learning_rate": 1.8291168295903047e-11, + "loss": 0.1074, + "step": 19078 + }, + { + "epoch": 1.0, + "learning_rate": 1.783675383817407e-11, + "loss": 0.1176, + "step": 19079 + }, + { + "epoch": 1.0, + "learning_rate": 1.738805494786444e-11, + "loss": 0.1113, + "step": 19080 + }, + { + "epoch": 1.0, + "learning_rate": 1.694507163780834e-11, + "loss": 0.1254, + "step": 19081 + }, + { + "epoch": 1.0, + "learning_rate": 1.650780392066231e-11, + "loss": 0.1553, + "step": 19082 + }, + { + "epoch": 1.0, + "learning_rate": 1.6076251808927465e-11, + "loss": 0.1095, + "step": 19083 + }, + { + "epoch": 1.0, + "learning_rate": 1.5650415314949483e-11, + "loss": 0.1405, + "step": 19084 + }, + { + "epoch": 1.0, + "learning_rate": 1.5230294450874203e-11, + "loss": 0.1475, + "step": 19085 + }, + { + "epoch": 1.0, + "learning_rate": 1.481588922871424e-11, + "loss": 0.1338, + "step": 19086 + }, + { + "epoch": 1.0, + "learning_rate": 1.440719966030457e-11, + "loss": 0.1375, + "step": 19087 + }, + { + "epoch": 1.0, + "learning_rate": 1.4004225757369148e-11, + "loss": 0.1188, + "step": 19088 + }, + { + "epoch": 1.0, + "learning_rate": 1.3606967531365477e-11, + "loss": 0.1715, + "step": 19089 + }, + { + "epoch": 1.0, + "learning_rate": 1.3215424993706648e-11, + "loss": 0.1448, + "step": 19090 + }, + { + "epoch": 1.0, + "learning_rate": 1.28295981555393e-11, + "loss": 0.1381, + "step": 19091 + }, + { + "epoch": 1.0, + "learning_rate": 1.2449487027899053e-11, + "loss": 0.1528, + "step": 19092 + }, + { + "epoch": 1.0, + "learning_rate": 1.2075091621666089e-11, + "loss": 0.1206, + "step": 19093 + }, + { + "epoch": 1.0, + "learning_rate": 1.1706411947520756e-11, + "loss": 0.1527, + "step": 19094 + }, + { + "epoch": 1.0, + "learning_rate": 1.1343448016010171e-11, + "loss": 0.1526, + "step": 19095 + }, + { + "epoch": 1.0, + "learning_rate": 1.0986199837526022e-11, + "loss": 0.1613, + "step": 19096 + }, + { + "epoch": 1.0, + "learning_rate": 1.0634667422237954e-11, + "loss": 0.1208, + "step": 19097 + }, + { + "epoch": 1.0, + "learning_rate": 1.0288850780226787e-11, + "loss": 0.1486, + "step": 19098 + }, + { + "epoch": 1.0, + "learning_rate": 9.948749921351308e-12, + "loss": 0.1552, + "step": 19099 + }, + { + "epoch": 1.0, + "learning_rate": 9.61436485535927e-12, + "loss": 0.1315, + "step": 19100 + }, + { + "epoch": 1.0, + "learning_rate": 9.285695591776388e-12, + "loss": 0.11, + "step": 19101 + }, + { + "epoch": 1.0, + "learning_rate": 8.962742140017354e-12, + "loss": 0.1235, + "step": 19102 + }, + { + "epoch": 1.0, + "learning_rate": 8.645504509319224e-12, + "loss": 0.1749, + "step": 19103 + }, + { + "epoch": 1.0, + "learning_rate": 8.333982708719211e-12, + "loss": 0.12, + "step": 19104 + }, + { + "epoch": 1.0, + "learning_rate": 8.028176747165715e-12, + "loss": 0.1217, + "step": 19105 + }, + { + "epoch": 1.0, + "learning_rate": 7.72808663334068e-12, + "loss": 0.1508, + "step": 19106 + }, + { + "epoch": 1.0, + "learning_rate": 7.43371237588164e-12, + "loss": 0.1055, + "step": 19107 + }, + { + "epoch": 1.0, + "learning_rate": 7.145053983137472e-12, + "loss": 0.1396, + "step": 19108 + }, + { + "epoch": 1.0, + "learning_rate": 6.862111463412646e-12, + "loss": 0.1293, + "step": 19109 + }, + { + "epoch": 1.0, + "learning_rate": 6.584884824789582e-12, + "loss": 0.1167, + "step": 19110 + }, + { + "epoch": 1.0, + "learning_rate": 6.313374075150868e-12, + "loss": 0.147, + "step": 19111 + }, + { + "epoch": 1.0, + "learning_rate": 6.0475792222902665e-12, + "loss": 0.1066, + "step": 19112 + }, + { + "epoch": 1.0, + "learning_rate": 5.787500273801704e-12, + "loss": 0.1599, + "step": 19113 + }, + { + "epoch": 1.0, + "learning_rate": 5.533137237101471e-12, + "loss": 0.1467, + "step": 19114 + }, + { + "epoch": 1.0, + "learning_rate": 5.284490119472629e-12, + "loss": 0.1275, + "step": 19115 + }, + { + "epoch": 1.0, + "learning_rate": 5.041558928020606e-12, + "loss": 0.1422, + "step": 19116 + }, + { + "epoch": 1.0, + "learning_rate": 4.804343669673194e-12, + "loss": 0.1354, + "step": 19117 + }, + { + "epoch": 1.0, + "learning_rate": 4.572844351247163e-12, + "loss": 0.1381, + "step": 19118 + }, + { + "epoch": 1.0, + "learning_rate": 4.347060979315032e-12, + "loss": 0.1287, + "step": 19119 + }, + { + "epoch": 1.0, + "learning_rate": 4.1269935603605034e-12, + "loss": 0.1358, + "step": 19120 + }, + { + "epoch": 1.0, + "learning_rate": 3.9126421006674404e-12, + "loss": 0.1085, + "step": 19121 + }, + { + "epoch": 1.0, + "learning_rate": 3.704006606342069e-12, + "loss": 0.1659, + "step": 19122 + }, + { + "epoch": 1.0, + "learning_rate": 3.50108708335739e-12, + "loss": 0.1665, + "step": 19123 + }, + { + "epoch": 1.0, + "learning_rate": 3.3038835375309715e-12, + "loss": 0.2083, + "step": 19124 + }, + { + "epoch": 1.0, + "learning_rate": 3.112395974502746e-12, + "loss": 0.156, + "step": 19125 + }, + { + "epoch": 1.0, + "learning_rate": 2.926624399712807e-12, + "loss": 0.1, + "step": 19126 + }, + { + "epoch": 1.0, + "learning_rate": 2.746568818490225e-12, + "loss": 0.1081, + "step": 19127 + }, + { + "epoch": 1.0, + "learning_rate": 2.5722292359642295e-12, + "loss": 0.1577, + "step": 19128 + }, + { + "epoch": 1.0, + "learning_rate": 2.4036056571308248e-12, + "loss": 0.1373, + "step": 19129 + }, + { + "epoch": 1.0, + "learning_rate": 2.2406980868305835e-12, + "loss": 0.1663, + "step": 19130 + }, + { + "epoch": 1.0, + "learning_rate": 2.083506529682033e-12, + "loss": 0.1199, + "step": 19131 + }, + { + "epoch": 1.0, + "learning_rate": 1.932030990192679e-12, + "loss": 0.1314, + "step": 19132 + }, + { + "epoch": 1.0, + "learning_rate": 1.7862714726923911e-12, + "loss": 0.1236, + "step": 19133 + }, + { + "epoch": 1.0, + "learning_rate": 1.6462279813556079e-12, + "loss": 0.1545, + "step": 19134 + }, + { + "epoch": 1.0, + "learning_rate": 1.5119005201791324e-12, + "loss": 0.1226, + "step": 19135 + }, + { + "epoch": 1.0, + "learning_rate": 1.3832890930043362e-12, + "loss": 0.1243, + "step": 19136 + }, + { + "epoch": 1.0, + "learning_rate": 1.2603937034949552e-12, + "loss": 0.137, + "step": 19137 + }, + { + "epoch": 1.0, + "learning_rate": 1.1432143551592943e-12, + "loss": 0.1425, + "step": 19138 + }, + { + "epoch": 1.0, + "learning_rate": 1.0317510513724315e-12, + "loss": 0.1043, + "step": 19139 + }, + { + "epoch": 1.0, + "learning_rate": 9.260037952874e-13, + "loss": 0.1334, + "step": 19140 + }, + { + "epoch": 1.0, + "learning_rate": 8.259725899684155e-13, + "loss": 0.13, + "step": 19141 + }, + { + "epoch": 1.0, + "learning_rate": 7.316574382354446e-13, + "loss": 0.1158, + "step": 19142 + }, + { + "epoch": 1.0, + "learning_rate": 6.430583427974312e-13, + "loss": 0.1211, + "step": 19143 + }, + { + "epoch": 1.0, + "learning_rate": 5.601753062078885e-13, + "loss": 0.161, + "step": 19144 + }, + { + "epoch": 1.0, + "learning_rate": 4.830083307982846e-13, + "loss": 0.1176, + "step": 19145 + }, + { + "epoch": 1.0, + "learning_rate": 4.115574187890658e-13, + "loss": 0.1432, + "step": 19146 + }, + { + "epoch": 1.0, + "learning_rate": 3.4582257222304236e-13, + "loss": 0.1089, + "step": 19147 + }, + { + "epoch": 1.0, + "learning_rate": 2.8580379300979783e-13, + "loss": 0.1189, + "step": 19148 + }, + { + "epoch": 1.0, + "learning_rate": 2.3150108281466685e-13, + "loss": 0.1329, + "step": 19149 + }, + { + "epoch": 1.0, + "learning_rate": 1.8291444319196158e-13, + "loss": 0.1554, + "step": 19150 + }, + { + "epoch": 1.0, + "learning_rate": 1.4004387558497198e-13, + "loss": 0.1403, + "step": 19151 + }, + { + "epoch": 1.0, + "learning_rate": 1.0288938117053447e-13, + "loss": 0.1283, + "step": 19152 + }, + { + "epoch": 1.0, + "learning_rate": 7.145096101446313e-14, + "loss": 0.1404, + "step": 19153 + }, + { + "epoch": 1.0, + "learning_rate": 4.572861602714084e-14, + "loss": 0.131, + "step": 19154 + }, + { + "epoch": 1.0, + "learning_rate": 2.572234694131481e-14, + "loss": 0.1165, + "step": 19155 + }, + { + "epoch": 1.0, + "learning_rate": 1.1432154334301002e-14, + "loss": 0.1428, + "step": 19156 + }, + { + "epoch": 1.0, + "learning_rate": 2.8580386057797113e-15, + "loss": 0.1288, + "step": 19157 + }, + { + "epoch": 1.0, + "learning_rate": 0.0, + "loss": 0.1177, + "step": 19158 + }, + { + "epoch": 1.0, + "step": 19158, + "total_flos": 9.2413134033624e+21, + "train_loss": 0.1417416125416961, + "train_runtime": 226046.9031, + "train_samples_per_second": 10.849, + "train_steps_per_second": 0.085 + } + ], + "logging_steps": 1.0, + "max_steps": 19158, + "num_input_tokens_seen": 0, + "num_train_epochs": 1, + "save_steps": 500.0, + "total_flos": 9.2413134033624e+21, + "train_batch_size": 2, + "trial_name": null, + "trial_params": null +}