{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 500, "global_step": 11719, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 2.840909090909091e-08, "loss": 2.4626, "step": 1 }, { "epoch": 0.0, "learning_rate": 5.681818181818182e-08, "loss": 2.4683, "step": 2 }, { "epoch": 0.0, "learning_rate": 8.522727272727273e-08, "loss": 2.3571, "step": 3 }, { "epoch": 0.0, "learning_rate": 1.1363636363636364e-07, "loss": 2.3632, "step": 4 }, { "epoch": 0.0, "learning_rate": 1.4204545454545455e-07, "loss": 2.3533, "step": 5 }, { "epoch": 0.0, "learning_rate": 1.7045454545454545e-07, "loss": 2.2788, "step": 6 }, { "epoch": 0.0, "learning_rate": 1.9886363636363638e-07, "loss": 2.483, "step": 7 }, { "epoch": 0.0, "learning_rate": 2.2727272727272729e-07, "loss": 2.4661, "step": 8 }, { "epoch": 0.0, "learning_rate": 2.556818181818182e-07, "loss": 2.2974, "step": 9 }, { "epoch": 0.0, "learning_rate": 2.840909090909091e-07, "loss": 2.2507, "step": 10 }, { "epoch": 0.0, "learning_rate": 3.125e-07, "loss": 2.1913, "step": 11 }, { "epoch": 0.0, "learning_rate": 3.409090909090909e-07, "loss": 2.1755, "step": 12 }, { "epoch": 0.0, "learning_rate": 3.693181818181819e-07, "loss": 2.1603, "step": 13 }, { "epoch": 0.0, "learning_rate": 3.9772727272727276e-07, "loss": 2.0207, "step": 14 }, { "epoch": 0.0, "learning_rate": 4.2613636363636364e-07, "loss": 1.9289, "step": 15 }, { "epoch": 0.0, "learning_rate": 4.5454545454545457e-07, "loss": 2.027, "step": 16 }, { "epoch": 0.0, "learning_rate": 4.829545454545455e-07, "loss": 2.0359, "step": 17 }, { "epoch": 0.0, "learning_rate": 5.113636363636364e-07, "loss": 1.9164, "step": 18 }, { "epoch": 0.0, "learning_rate": 5.397727272727273e-07, "loss": 1.7516, "step": 19 }, { "epoch": 0.0, "learning_rate": 5.681818181818182e-07, "loss": 1.8781, "step": 20 }, { "epoch": 0.0, "learning_rate": 5.965909090909092e-07, "loss": 1.7287, "step": 21 }, { "epoch": 0.0, "learning_rate": 6.25e-07, "loss": 1.7456, "step": 22 }, { "epoch": 0.0, "learning_rate": 6.534090909090909e-07, "loss": 1.7367, "step": 23 }, { "epoch": 0.0, "learning_rate": 6.818181818181818e-07, "loss": 1.6767, "step": 24 }, { "epoch": 0.0, "learning_rate": 7.102272727272729e-07, "loss": 1.6815, "step": 25 }, { "epoch": 0.0, "learning_rate": 7.386363636363638e-07, "loss": 1.2482, "step": 26 }, { "epoch": 0.0, "learning_rate": 7.670454545454547e-07, "loss": 1.523, "step": 27 }, { "epoch": 0.0, "learning_rate": 7.954545454545455e-07, "loss": 1.5421, "step": 28 }, { "epoch": 0.0, "learning_rate": 8.238636363636364e-07, "loss": 1.533, "step": 29 }, { "epoch": 0.0, "learning_rate": 8.522727272727273e-07, "loss": 1.4871, "step": 30 }, { "epoch": 0.0, "learning_rate": 8.806818181818183e-07, "loss": 1.4557, "step": 31 }, { "epoch": 0.0, "learning_rate": 9.090909090909091e-07, "loss": 1.5309, "step": 32 }, { "epoch": 0.0, "learning_rate": 9.375000000000001e-07, "loss": 1.4424, "step": 33 }, { "epoch": 0.0, "learning_rate": 9.65909090909091e-07, "loss": 1.4637, "step": 34 }, { "epoch": 0.0, "learning_rate": 9.943181818181819e-07, "loss": 1.4242, "step": 35 }, { "epoch": 0.0, "learning_rate": 1.0227272727272729e-06, "loss": 1.4108, "step": 36 }, { "epoch": 0.0, "learning_rate": 1.0511363636363639e-06, "loss": 1.3613, "step": 37 }, { "epoch": 0.0, "learning_rate": 1.0795454545454546e-06, "loss": 1.4486, "step": 38 }, { "epoch": 0.0, "learning_rate": 1.1079545454545456e-06, "loss": 1.3917, "step": 39 }, { "epoch": 0.0, "learning_rate": 1.1363636363636364e-06, "loss": 1.3509, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.1647727272727274e-06, "loss": 1.3646, "step": 41 }, { "epoch": 0.0, "learning_rate": 1.1931818181818183e-06, "loss": 1.3672, "step": 42 }, { "epoch": 0.0, "learning_rate": 1.2215909090909091e-06, "loss": 1.3639, "step": 43 }, { "epoch": 0.0, "learning_rate": 1.25e-06, "loss": 1.3419, "step": 44 }, { "epoch": 0.0, "learning_rate": 1.278409090909091e-06, "loss": 1.3217, "step": 45 }, { "epoch": 0.0, "learning_rate": 1.3068181818181819e-06, "loss": 1.3497, "step": 46 }, { "epoch": 0.0, "learning_rate": 1.3352272727272728e-06, "loss": 1.2842, "step": 47 }, { "epoch": 0.0, "learning_rate": 1.3636363636363636e-06, "loss": 1.3049, "step": 48 }, { "epoch": 0.0, "learning_rate": 1.3920454545454546e-06, "loss": 1.3072, "step": 49 }, { "epoch": 0.0, "learning_rate": 1.4204545454545458e-06, "loss": 1.3622, "step": 50 }, { "epoch": 0.0, "learning_rate": 1.4488636363636366e-06, "loss": 1.3379, "step": 51 }, { "epoch": 0.0, "learning_rate": 1.4772727272727275e-06, "loss": 1.2889, "step": 52 }, { "epoch": 0.0, "learning_rate": 1.5056818181818183e-06, "loss": 1.2406, "step": 53 }, { "epoch": 0.0, "learning_rate": 1.5340909090909093e-06, "loss": 1.2691, "step": 54 }, { "epoch": 0.0, "learning_rate": 1.5625e-06, "loss": 1.3137, "step": 55 }, { "epoch": 0.0, "learning_rate": 1.590909090909091e-06, "loss": 1.2339, "step": 56 }, { "epoch": 0.0, "learning_rate": 1.6193181818181818e-06, "loss": 1.2916, "step": 57 }, { "epoch": 0.0, "learning_rate": 1.6477272727272728e-06, "loss": 1.3007, "step": 58 }, { "epoch": 0.01, "learning_rate": 1.6761363636363636e-06, "loss": 1.1594, "step": 59 }, { "epoch": 0.01, "learning_rate": 1.7045454545454546e-06, "loss": 1.1658, "step": 60 }, { "epoch": 0.01, "learning_rate": 1.7329545454545458e-06, "loss": 1.2236, "step": 61 }, { "epoch": 0.01, "learning_rate": 1.7613636363636365e-06, "loss": 1.2254, "step": 62 }, { "epoch": 0.01, "learning_rate": 1.7897727272727275e-06, "loss": 1.2372, "step": 63 }, { "epoch": 0.01, "learning_rate": 1.8181818181818183e-06, "loss": 1.204, "step": 64 }, { "epoch": 0.01, "learning_rate": 1.8465909090909093e-06, "loss": 1.2308, "step": 65 }, { "epoch": 0.01, "learning_rate": 1.8750000000000003e-06, "loss": 1.1981, "step": 66 }, { "epoch": 0.01, "learning_rate": 1.903409090909091e-06, "loss": 1.1972, "step": 67 }, { "epoch": 0.01, "learning_rate": 1.931818181818182e-06, "loss": 1.2379, "step": 68 }, { "epoch": 0.01, "learning_rate": 1.9602272727272728e-06, "loss": 1.2171, "step": 69 }, { "epoch": 0.01, "learning_rate": 1.9886363636363638e-06, "loss": 1.1871, "step": 70 }, { "epoch": 0.01, "learning_rate": 2.0170454545454548e-06, "loss": 1.2137, "step": 71 }, { "epoch": 0.01, "learning_rate": 2.0454545454545457e-06, "loss": 1.2085, "step": 72 }, { "epoch": 0.01, "learning_rate": 2.0738636363636367e-06, "loss": 1.1259, "step": 73 }, { "epoch": 0.01, "learning_rate": 2.1022727272727277e-06, "loss": 1.2392, "step": 74 }, { "epoch": 0.01, "learning_rate": 2.1306818181818183e-06, "loss": 1.2786, "step": 75 }, { "epoch": 0.01, "learning_rate": 2.1590909090909092e-06, "loss": 1.1965, "step": 76 }, { "epoch": 0.01, "learning_rate": 2.1875000000000002e-06, "loss": 1.2248, "step": 77 }, { "epoch": 0.01, "learning_rate": 2.2159090909090912e-06, "loss": 1.2135, "step": 78 }, { "epoch": 0.01, "learning_rate": 2.2443181818181818e-06, "loss": 1.1953, "step": 79 }, { "epoch": 0.01, "learning_rate": 2.2727272727272728e-06, "loss": 1.1923, "step": 80 }, { "epoch": 0.01, "learning_rate": 2.3011363636363637e-06, "loss": 1.1635, "step": 81 }, { "epoch": 0.01, "learning_rate": 2.3295454545454547e-06, "loss": 1.1795, "step": 82 }, { "epoch": 0.01, "learning_rate": 2.3579545454545457e-06, "loss": 1.1634, "step": 83 }, { "epoch": 0.01, "learning_rate": 2.3863636363636367e-06, "loss": 0.9074, "step": 84 }, { "epoch": 0.01, "learning_rate": 2.4147727272727277e-06, "loss": 1.1687, "step": 85 }, { "epoch": 0.01, "learning_rate": 2.4431818181818182e-06, "loss": 1.1613, "step": 86 }, { "epoch": 0.01, "learning_rate": 2.4715909090909092e-06, "loss": 1.1078, "step": 87 }, { "epoch": 0.01, "learning_rate": 2.5e-06, "loss": 1.1684, "step": 88 }, { "epoch": 0.01, "learning_rate": 2.528409090909091e-06, "loss": 1.1493, "step": 89 }, { "epoch": 0.01, "learning_rate": 2.556818181818182e-06, "loss": 1.2259, "step": 90 }, { "epoch": 0.01, "learning_rate": 2.585227272727273e-06, "loss": 1.1411, "step": 91 }, { "epoch": 0.01, "learning_rate": 2.6136363636363637e-06, "loss": 1.1352, "step": 92 }, { "epoch": 0.01, "learning_rate": 2.642045454545455e-06, "loss": 1.1526, "step": 93 }, { "epoch": 0.01, "learning_rate": 2.6704545454545457e-06, "loss": 1.1651, "step": 94 }, { "epoch": 0.01, "learning_rate": 2.6988636363636367e-06, "loss": 1.129, "step": 95 }, { "epoch": 0.01, "learning_rate": 2.7272727272727272e-06, "loss": 1.2192, "step": 96 }, { "epoch": 0.01, "learning_rate": 2.7556818181818186e-06, "loss": 1.1648, "step": 97 }, { "epoch": 0.01, "learning_rate": 2.784090909090909e-06, "loss": 1.1723, "step": 98 }, { "epoch": 0.01, "learning_rate": 2.8125e-06, "loss": 1.0909, "step": 99 }, { "epoch": 0.01, "learning_rate": 2.8409090909090916e-06, "loss": 1.1307, "step": 100 }, { "epoch": 0.01, "learning_rate": 2.869318181818182e-06, "loss": 1.0981, "step": 101 }, { "epoch": 0.01, "learning_rate": 2.897727272727273e-06, "loss": 0.8085, "step": 102 }, { "epoch": 0.01, "learning_rate": 2.9261363636363637e-06, "loss": 1.0951, "step": 103 }, { "epoch": 0.01, "learning_rate": 2.954545454545455e-06, "loss": 1.1317, "step": 104 }, { "epoch": 0.01, "learning_rate": 2.9829545454545457e-06, "loss": 1.1074, "step": 105 }, { "epoch": 0.01, "learning_rate": 3.0113636363636366e-06, "loss": 1.1496, "step": 106 }, { "epoch": 0.01, "learning_rate": 3.039772727272727e-06, "loss": 1.1486, "step": 107 }, { "epoch": 0.01, "learning_rate": 3.0681818181818186e-06, "loss": 1.0684, "step": 108 }, { "epoch": 0.01, "learning_rate": 3.096590909090909e-06, "loss": 1.1273, "step": 109 }, { "epoch": 0.01, "learning_rate": 3.125e-06, "loss": 1.138, "step": 110 }, { "epoch": 0.01, "learning_rate": 3.1534090909090916e-06, "loss": 1.132, "step": 111 }, { "epoch": 0.01, "learning_rate": 3.181818181818182e-06, "loss": 1.1438, "step": 112 }, { "epoch": 0.01, "learning_rate": 3.210227272727273e-06, "loss": 1.1105, "step": 113 }, { "epoch": 0.01, "learning_rate": 3.2386363636363637e-06, "loss": 1.1079, "step": 114 }, { "epoch": 0.01, "learning_rate": 3.267045454545455e-06, "loss": 1.1211, "step": 115 }, { "epoch": 0.01, "learning_rate": 3.2954545454545456e-06, "loss": 1.0883, "step": 116 }, { "epoch": 0.01, "learning_rate": 3.3238636363636366e-06, "loss": 1.109, "step": 117 }, { "epoch": 0.01, "learning_rate": 3.352272727272727e-06, "loss": 1.0969, "step": 118 }, { "epoch": 0.01, "learning_rate": 3.3806818181818186e-06, "loss": 1.0877, "step": 119 }, { "epoch": 0.01, "learning_rate": 3.409090909090909e-06, "loss": 1.0914, "step": 120 }, { "epoch": 0.01, "learning_rate": 3.4375e-06, "loss": 0.9171, "step": 121 }, { "epoch": 0.01, "learning_rate": 3.4659090909090915e-06, "loss": 1.1463, "step": 122 }, { "epoch": 0.01, "learning_rate": 3.494318181818182e-06, "loss": 1.0308, "step": 123 }, { "epoch": 0.01, "learning_rate": 3.522727272727273e-06, "loss": 1.1102, "step": 124 }, { "epoch": 0.01, "learning_rate": 3.5511363636363636e-06, "loss": 1.1507, "step": 125 }, { "epoch": 0.01, "learning_rate": 3.579545454545455e-06, "loss": 1.1266, "step": 126 }, { "epoch": 0.01, "learning_rate": 3.6079545454545456e-06, "loss": 1.0795, "step": 127 }, { "epoch": 0.01, "learning_rate": 3.6363636363636366e-06, "loss": 1.0656, "step": 128 }, { "epoch": 0.01, "learning_rate": 3.6647727272727276e-06, "loss": 1.0954, "step": 129 }, { "epoch": 0.01, "learning_rate": 3.6931818181818186e-06, "loss": 1.1365, "step": 130 }, { "epoch": 0.01, "learning_rate": 3.721590909090909e-06, "loss": 1.0822, "step": 131 }, { "epoch": 0.01, "learning_rate": 3.7500000000000005e-06, "loss": 1.1177, "step": 132 }, { "epoch": 0.01, "learning_rate": 3.7784090909090915e-06, "loss": 1.0774, "step": 133 }, { "epoch": 0.01, "learning_rate": 3.806818181818182e-06, "loss": 1.0948, "step": 134 }, { "epoch": 0.01, "learning_rate": 3.8352272727272735e-06, "loss": 1.0655, "step": 135 }, { "epoch": 0.01, "learning_rate": 3.863636363636364e-06, "loss": 1.1023, "step": 136 }, { "epoch": 0.01, "learning_rate": 3.8920454545454554e-06, "loss": 1.0277, "step": 137 }, { "epoch": 0.01, "learning_rate": 3.9204545454545456e-06, "loss": 1.0618, "step": 138 }, { "epoch": 0.01, "learning_rate": 3.9488636363636366e-06, "loss": 1.0616, "step": 139 }, { "epoch": 0.01, "learning_rate": 3.9772727272727275e-06, "loss": 1.0872, "step": 140 }, { "epoch": 0.01, "learning_rate": 4.0056818181818185e-06, "loss": 1.0373, "step": 141 }, { "epoch": 0.01, "learning_rate": 4.0340909090909095e-06, "loss": 1.0487, "step": 142 }, { "epoch": 0.01, "learning_rate": 4.0625000000000005e-06, "loss": 1.0855, "step": 143 }, { "epoch": 0.01, "learning_rate": 4.0909090909090915e-06, "loss": 1.0457, "step": 144 }, { "epoch": 0.01, "learning_rate": 4.1193181818181825e-06, "loss": 1.0544, "step": 145 }, { "epoch": 0.01, "learning_rate": 4.1477272727272734e-06, "loss": 1.0356, "step": 146 }, { "epoch": 0.01, "learning_rate": 4.176136363636364e-06, "loss": 1.0558, "step": 147 }, { "epoch": 0.01, "learning_rate": 4.204545454545455e-06, "loss": 1.1, "step": 148 }, { "epoch": 0.01, "learning_rate": 4.2329545454545455e-06, "loss": 1.0315, "step": 149 }, { "epoch": 0.01, "learning_rate": 4.2613636363636365e-06, "loss": 1.053, "step": 150 }, { "epoch": 0.01, "learning_rate": 4.2897727272727275e-06, "loss": 1.0902, "step": 151 }, { "epoch": 0.01, "learning_rate": 4.3181818181818185e-06, "loss": 1.0643, "step": 152 }, { "epoch": 0.01, "learning_rate": 4.3465909090909095e-06, "loss": 1.0866, "step": 153 }, { "epoch": 0.01, "learning_rate": 4.3750000000000005e-06, "loss": 1.0993, "step": 154 }, { "epoch": 0.01, "learning_rate": 4.4034090909090914e-06, "loss": 1.0787, "step": 155 }, { "epoch": 0.01, "learning_rate": 4.4318181818181824e-06, "loss": 0.8752, "step": 156 }, { "epoch": 0.01, "learning_rate": 4.460227272727273e-06, "loss": 1.0271, "step": 157 }, { "epoch": 0.01, "learning_rate": 4.4886363636363636e-06, "loss": 1.1245, "step": 158 }, { "epoch": 0.01, "learning_rate": 4.517045454545455e-06, "loss": 1.0955, "step": 159 }, { "epoch": 0.01, "learning_rate": 4.5454545454545455e-06, "loss": 1.0646, "step": 160 }, { "epoch": 0.01, "learning_rate": 4.5738636363636365e-06, "loss": 0.9996, "step": 161 }, { "epoch": 0.01, "learning_rate": 4.6022727272727275e-06, "loss": 1.054, "step": 162 }, { "epoch": 0.01, "learning_rate": 4.6306818181818185e-06, "loss": 1.0399, "step": 163 }, { "epoch": 0.01, "learning_rate": 4.6590909090909095e-06, "loss": 0.9853, "step": 164 }, { "epoch": 0.01, "learning_rate": 4.6875000000000004e-06, "loss": 1.025, "step": 165 }, { "epoch": 0.01, "learning_rate": 4.715909090909091e-06, "loss": 1.0629, "step": 166 }, { "epoch": 0.01, "learning_rate": 4.744318181818182e-06, "loss": 0.7735, "step": 167 }, { "epoch": 0.01, "learning_rate": 4.772727272727273e-06, "loss": 1.0162, "step": 168 }, { "epoch": 0.01, "learning_rate": 4.8011363636363635e-06, "loss": 1.0599, "step": 169 }, { "epoch": 0.01, "learning_rate": 4.829545454545455e-06, "loss": 0.6815, "step": 170 }, { "epoch": 0.01, "learning_rate": 4.8579545454545455e-06, "loss": 1.0394, "step": 171 }, { "epoch": 0.01, "learning_rate": 4.8863636363636365e-06, "loss": 1.0195, "step": 172 }, { "epoch": 0.01, "learning_rate": 4.9147727272727275e-06, "loss": 1.0443, "step": 173 }, { "epoch": 0.01, "learning_rate": 4.9431818181818184e-06, "loss": 1.0692, "step": 174 }, { "epoch": 0.01, "learning_rate": 4.9715909090909094e-06, "loss": 1.0467, "step": 175 }, { "epoch": 0.02, "learning_rate": 5e-06, "loss": 1.0202, "step": 176 }, { "epoch": 0.02, "learning_rate": 5.028409090909091e-06, "loss": 1.0516, "step": 177 }, { "epoch": 0.02, "learning_rate": 5.056818181818182e-06, "loss": 1.0348, "step": 178 }, { "epoch": 0.02, "learning_rate": 5.085227272727273e-06, "loss": 1.0714, "step": 179 }, { "epoch": 0.02, "learning_rate": 5.113636363636364e-06, "loss": 1.0662, "step": 180 }, { "epoch": 0.02, "learning_rate": 5.1420454545454545e-06, "loss": 1.0415, "step": 181 }, { "epoch": 0.02, "learning_rate": 5.170454545454546e-06, "loss": 1.0033, "step": 182 }, { "epoch": 0.02, "learning_rate": 5.198863636363637e-06, "loss": 1.0447, "step": 183 }, { "epoch": 0.02, "learning_rate": 5.2272727272727274e-06, "loss": 1.042, "step": 184 }, { "epoch": 0.02, "learning_rate": 5.255681818181818e-06, "loss": 1.0733, "step": 185 }, { "epoch": 0.02, "learning_rate": 5.28409090909091e-06, "loss": 1.0164, "step": 186 }, { "epoch": 0.02, "learning_rate": 5.3125e-06, "loss": 0.9936, "step": 187 }, { "epoch": 0.02, "learning_rate": 5.340909090909091e-06, "loss": 1.0495, "step": 188 }, { "epoch": 0.02, "learning_rate": 5.369318181818182e-06, "loss": 0.7912, "step": 189 }, { "epoch": 0.02, "learning_rate": 5.397727272727273e-06, "loss": 1.0759, "step": 190 }, { "epoch": 0.02, "learning_rate": 5.426136363636364e-06, "loss": 1.0465, "step": 191 }, { "epoch": 0.02, "learning_rate": 5.4545454545454545e-06, "loss": 1.0453, "step": 192 }, { "epoch": 0.02, "learning_rate": 5.482954545454546e-06, "loss": 1.0428, "step": 193 }, { "epoch": 0.02, "learning_rate": 5.511363636363637e-06, "loss": 1.0607, "step": 194 }, { "epoch": 0.02, "learning_rate": 5.539772727272727e-06, "loss": 1.0004, "step": 195 }, { "epoch": 0.02, "learning_rate": 5.568181818181818e-06, "loss": 1.0871, "step": 196 }, { "epoch": 0.02, "learning_rate": 5.59659090909091e-06, "loss": 1.0001, "step": 197 }, { "epoch": 0.02, "learning_rate": 5.625e-06, "loss": 0.969, "step": 198 }, { "epoch": 0.02, "learning_rate": 5.653409090909091e-06, "loss": 0.9889, "step": 199 }, { "epoch": 0.02, "learning_rate": 5.681818181818183e-06, "loss": 0.9878, "step": 200 }, { "epoch": 0.02, "learning_rate": 5.710227272727273e-06, "loss": 0.7978, "step": 201 }, { "epoch": 0.02, "learning_rate": 5.738636363636364e-06, "loss": 1.0205, "step": 202 }, { "epoch": 0.02, "learning_rate": 5.7670454545454544e-06, "loss": 1.0365, "step": 203 }, { "epoch": 0.02, "learning_rate": 5.795454545454546e-06, "loss": 0.999, "step": 204 }, { "epoch": 0.02, "learning_rate": 5.823863636363637e-06, "loss": 1.0172, "step": 205 }, { "epoch": 0.02, "learning_rate": 5.852272727272727e-06, "loss": 1.054, "step": 206 }, { "epoch": 0.02, "learning_rate": 5.880681818181818e-06, "loss": 1.099, "step": 207 }, { "epoch": 0.02, "learning_rate": 5.90909090909091e-06, "loss": 0.9272, "step": 208 }, { "epoch": 0.02, "learning_rate": 5.9375e-06, "loss": 1.0048, "step": 209 }, { "epoch": 0.02, "learning_rate": 5.965909090909091e-06, "loss": 1.0864, "step": 210 }, { "epoch": 0.02, "learning_rate": 5.994318181818183e-06, "loss": 1.0524, "step": 211 }, { "epoch": 0.02, "learning_rate": 6.022727272727273e-06, "loss": 1.0189, "step": 212 }, { "epoch": 0.02, "learning_rate": 6.051136363636364e-06, "loss": 1.0345, "step": 213 }, { "epoch": 0.02, "learning_rate": 6.079545454545454e-06, "loss": 1.0277, "step": 214 }, { "epoch": 0.02, "learning_rate": 6.107954545454546e-06, "loss": 1.0457, "step": 215 }, { "epoch": 0.02, "learning_rate": 6.136363636363637e-06, "loss": 1.0423, "step": 216 }, { "epoch": 0.02, "learning_rate": 6.164772727272727e-06, "loss": 1.0068, "step": 217 }, { "epoch": 0.02, "learning_rate": 6.193181818181818e-06, "loss": 1.0518, "step": 218 }, { "epoch": 0.02, "learning_rate": 6.22159090909091e-06, "loss": 1.0497, "step": 219 }, { "epoch": 0.02, "learning_rate": 6.25e-06, "loss": 1.03, "step": 220 }, { "epoch": 0.02, "learning_rate": 6.278409090909091e-06, "loss": 1.0476, "step": 221 }, { "epoch": 0.02, "learning_rate": 6.306818181818183e-06, "loss": 1.0508, "step": 222 }, { "epoch": 0.02, "learning_rate": 6.335227272727273e-06, "loss": 0.9727, "step": 223 }, { "epoch": 0.02, "learning_rate": 6.363636363636364e-06, "loss": 1.0081, "step": 224 }, { "epoch": 0.02, "learning_rate": 6.392045454545454e-06, "loss": 0.9956, "step": 225 }, { "epoch": 0.02, "learning_rate": 6.420454545454546e-06, "loss": 1.0078, "step": 226 }, { "epoch": 0.02, "learning_rate": 6.448863636363637e-06, "loss": 1.0445, "step": 227 }, { "epoch": 0.02, "learning_rate": 6.477272727272727e-06, "loss": 0.9766, "step": 228 }, { "epoch": 0.02, "learning_rate": 6.505681818181818e-06, "loss": 0.9725, "step": 229 }, { "epoch": 0.02, "learning_rate": 6.53409090909091e-06, "loss": 0.9946, "step": 230 }, { "epoch": 0.02, "learning_rate": 6.5625e-06, "loss": 0.9388, "step": 231 }, { "epoch": 0.02, "learning_rate": 6.590909090909091e-06, "loss": 1.006, "step": 232 }, { "epoch": 0.02, "learning_rate": 6.619318181818183e-06, "loss": 1.0469, "step": 233 }, { "epoch": 0.02, "learning_rate": 6.647727272727273e-06, "loss": 0.7737, "step": 234 }, { "epoch": 0.02, "learning_rate": 6.676136363636364e-06, "loss": 1.0279, "step": 235 }, { "epoch": 0.02, "learning_rate": 6.704545454545454e-06, "loss": 1.0263, "step": 236 }, { "epoch": 0.02, "learning_rate": 6.732954545454546e-06, "loss": 1.0669, "step": 237 }, { "epoch": 0.02, "learning_rate": 6.761363636363637e-06, "loss": 1.0323, "step": 238 }, { "epoch": 0.02, "learning_rate": 6.789772727272727e-06, "loss": 1.0561, "step": 239 }, { "epoch": 0.02, "learning_rate": 6.818181818181818e-06, "loss": 0.998, "step": 240 }, { "epoch": 0.02, "learning_rate": 6.84659090909091e-06, "loss": 1.0571, "step": 241 }, { "epoch": 0.02, "learning_rate": 6.875e-06, "loss": 1.0066, "step": 242 }, { "epoch": 0.02, "learning_rate": 6.903409090909091e-06, "loss": 1.0263, "step": 243 }, { "epoch": 0.02, "learning_rate": 6.931818181818183e-06, "loss": 0.9689, "step": 244 }, { "epoch": 0.02, "learning_rate": 6.960227272727273e-06, "loss": 1.0469, "step": 245 }, { "epoch": 0.02, "learning_rate": 6.988636363636364e-06, "loss": 0.9805, "step": 246 }, { "epoch": 0.02, "learning_rate": 7.017045454545454e-06, "loss": 1.0316, "step": 247 }, { "epoch": 0.02, "learning_rate": 7.045454545454546e-06, "loss": 0.9688, "step": 248 }, { "epoch": 0.02, "learning_rate": 7.073863636363637e-06, "loss": 0.9987, "step": 249 }, { "epoch": 0.02, "learning_rate": 7.102272727272727e-06, "loss": 1.0369, "step": 250 }, { "epoch": 0.02, "learning_rate": 7.130681818181818e-06, "loss": 1.0501, "step": 251 }, { "epoch": 0.02, "learning_rate": 7.15909090909091e-06, "loss": 1.0113, "step": 252 }, { "epoch": 0.02, "learning_rate": 7.1875e-06, "loss": 0.9772, "step": 253 }, { "epoch": 0.02, "learning_rate": 7.215909090909091e-06, "loss": 1.0441, "step": 254 }, { "epoch": 0.02, "learning_rate": 7.244318181818183e-06, "loss": 0.9933, "step": 255 }, { "epoch": 0.02, "learning_rate": 7.272727272727273e-06, "loss": 1.0041, "step": 256 }, { "epoch": 0.02, "learning_rate": 7.301136363636364e-06, "loss": 0.7326, "step": 257 }, { "epoch": 0.02, "learning_rate": 7.329545454545455e-06, "loss": 1.0167, "step": 258 }, { "epoch": 0.02, "learning_rate": 7.357954545454546e-06, "loss": 1.0075, "step": 259 }, { "epoch": 0.02, "learning_rate": 7.386363636363637e-06, "loss": 0.9741, "step": 260 }, { "epoch": 0.02, "learning_rate": 7.414772727272728e-06, "loss": 0.967, "step": 261 }, { "epoch": 0.02, "learning_rate": 7.443181818181818e-06, "loss": 0.9895, "step": 262 }, { "epoch": 0.02, "learning_rate": 7.47159090909091e-06, "loss": 1.0138, "step": 263 }, { "epoch": 0.02, "learning_rate": 7.500000000000001e-06, "loss": 1.0047, "step": 264 }, { "epoch": 0.02, "learning_rate": 7.528409090909091e-06, "loss": 1.0068, "step": 265 }, { "epoch": 0.02, "learning_rate": 7.556818181818183e-06, "loss": 0.9889, "step": 266 }, { "epoch": 0.02, "learning_rate": 7.585227272727273e-06, "loss": 1.0038, "step": 267 }, { "epoch": 0.02, "learning_rate": 7.613636363636364e-06, "loss": 1.0135, "step": 268 }, { "epoch": 0.02, "learning_rate": 7.642045454545454e-06, "loss": 1.0355, "step": 269 }, { "epoch": 0.02, "learning_rate": 7.670454545454547e-06, "loss": 0.9533, "step": 270 }, { "epoch": 0.02, "learning_rate": 7.698863636363638e-06, "loss": 0.9987, "step": 271 }, { "epoch": 0.02, "learning_rate": 7.727272727272727e-06, "loss": 0.7268, "step": 272 }, { "epoch": 0.02, "learning_rate": 7.755681818181818e-06, "loss": 0.9892, "step": 273 }, { "epoch": 0.02, "learning_rate": 7.784090909090911e-06, "loss": 1.0228, "step": 274 }, { "epoch": 0.02, "learning_rate": 7.8125e-06, "loss": 0.9998, "step": 275 }, { "epoch": 0.02, "learning_rate": 7.840909090909091e-06, "loss": 1.004, "step": 276 }, { "epoch": 0.02, "learning_rate": 7.869318181818182e-06, "loss": 0.9843, "step": 277 }, { "epoch": 0.02, "learning_rate": 7.897727272727273e-06, "loss": 1.0417, "step": 278 }, { "epoch": 0.02, "learning_rate": 7.926136363636364e-06, "loss": 0.9753, "step": 279 }, { "epoch": 0.02, "learning_rate": 7.954545454545455e-06, "loss": 1.0547, "step": 280 }, { "epoch": 0.02, "learning_rate": 7.982954545454546e-06, "loss": 0.9877, "step": 281 }, { "epoch": 0.02, "learning_rate": 8.011363636363637e-06, "loss": 0.9536, "step": 282 }, { "epoch": 0.02, "learning_rate": 8.039772727272728e-06, "loss": 1.0098, "step": 283 }, { "epoch": 0.02, "learning_rate": 8.068181818181819e-06, "loss": 0.9383, "step": 284 }, { "epoch": 0.02, "learning_rate": 8.09659090909091e-06, "loss": 0.9996, "step": 285 }, { "epoch": 0.02, "learning_rate": 8.125000000000001e-06, "loss": 0.9865, "step": 286 }, { "epoch": 0.02, "learning_rate": 8.153409090909092e-06, "loss": 1.018, "step": 287 }, { "epoch": 0.02, "learning_rate": 8.181818181818183e-06, "loss": 1.0336, "step": 288 }, { "epoch": 0.02, "learning_rate": 8.210227272727274e-06, "loss": 1.0042, "step": 289 }, { "epoch": 0.02, "learning_rate": 8.238636363636365e-06, "loss": 0.9258, "step": 290 }, { "epoch": 0.02, "learning_rate": 8.267045454545454e-06, "loss": 1.0395, "step": 291 }, { "epoch": 0.02, "learning_rate": 8.295454545454547e-06, "loss": 0.99, "step": 292 }, { "epoch": 0.03, "learning_rate": 8.323863636363638e-06, "loss": 0.9556, "step": 293 }, { "epoch": 0.03, "learning_rate": 8.352272727272727e-06, "loss": 0.7206, "step": 294 }, { "epoch": 0.03, "learning_rate": 8.380681818181818e-06, "loss": 0.9839, "step": 295 }, { "epoch": 0.03, "learning_rate": 8.40909090909091e-06, "loss": 1.0136, "step": 296 }, { "epoch": 0.03, "learning_rate": 8.4375e-06, "loss": 0.7176, "step": 297 }, { "epoch": 0.03, "learning_rate": 8.465909090909091e-06, "loss": 0.9607, "step": 298 }, { "epoch": 0.03, "learning_rate": 8.494318181818184e-06, "loss": 0.9453, "step": 299 }, { "epoch": 0.03, "learning_rate": 8.522727272727273e-06, "loss": 0.9773, "step": 300 }, { "epoch": 0.03, "learning_rate": 8.551136363636364e-06, "loss": 1.0552, "step": 301 }, { "epoch": 0.03, "learning_rate": 8.579545454545455e-06, "loss": 0.6834, "step": 302 }, { "epoch": 0.03, "learning_rate": 8.607954545454546e-06, "loss": 1.0363, "step": 303 }, { "epoch": 0.03, "learning_rate": 8.636363636363637e-06, "loss": 1.0258, "step": 304 }, { "epoch": 0.03, "learning_rate": 8.664772727272728e-06, "loss": 0.9565, "step": 305 }, { "epoch": 0.03, "learning_rate": 8.693181818181819e-06, "loss": 1.0054, "step": 306 }, { "epoch": 0.03, "learning_rate": 8.72159090909091e-06, "loss": 0.9987, "step": 307 }, { "epoch": 0.03, "learning_rate": 8.750000000000001e-06, "loss": 0.981, "step": 308 }, { "epoch": 0.03, "learning_rate": 8.778409090909092e-06, "loss": 1.0316, "step": 309 }, { "epoch": 0.03, "learning_rate": 8.806818181818183e-06, "loss": 0.9557, "step": 310 }, { "epoch": 0.03, "learning_rate": 8.835227272727274e-06, "loss": 1.0677, "step": 311 }, { "epoch": 0.03, "learning_rate": 8.863636363636365e-06, "loss": 1.0089, "step": 312 }, { "epoch": 0.03, "learning_rate": 8.892045454545454e-06, "loss": 0.9921, "step": 313 }, { "epoch": 0.03, "learning_rate": 8.920454545454547e-06, "loss": 1.0075, "step": 314 }, { "epoch": 0.03, "learning_rate": 8.948863636363638e-06, "loss": 0.968, "step": 315 }, { "epoch": 0.03, "learning_rate": 8.977272727272727e-06, "loss": 1.0122, "step": 316 }, { "epoch": 0.03, "learning_rate": 9.005681818181818e-06, "loss": 0.9908, "step": 317 }, { "epoch": 0.03, "learning_rate": 9.03409090909091e-06, "loss": 1.0159, "step": 318 }, { "epoch": 0.03, "learning_rate": 9.0625e-06, "loss": 0.9782, "step": 319 }, { "epoch": 0.03, "learning_rate": 9.090909090909091e-06, "loss": 0.9513, "step": 320 }, { "epoch": 0.03, "learning_rate": 9.119318181818184e-06, "loss": 0.9819, "step": 321 }, { "epoch": 0.03, "learning_rate": 9.147727272727273e-06, "loss": 0.9758, "step": 322 }, { "epoch": 0.03, "learning_rate": 9.176136363636364e-06, "loss": 0.9679, "step": 323 }, { "epoch": 0.03, "learning_rate": 9.204545454545455e-06, "loss": 1.0336, "step": 324 }, { "epoch": 0.03, "learning_rate": 9.232954545454546e-06, "loss": 0.9727, "step": 325 }, { "epoch": 0.03, "learning_rate": 9.261363636363637e-06, "loss": 0.9969, "step": 326 }, { "epoch": 0.03, "learning_rate": 9.289772727272728e-06, "loss": 0.9815, "step": 327 }, { "epoch": 0.03, "learning_rate": 9.318181818181819e-06, "loss": 0.9855, "step": 328 }, { "epoch": 0.03, "learning_rate": 9.34659090909091e-06, "loss": 0.9914, "step": 329 }, { "epoch": 0.03, "learning_rate": 9.375000000000001e-06, "loss": 0.9924, "step": 330 }, { "epoch": 0.03, "learning_rate": 9.403409090909092e-06, "loss": 0.9305, "step": 331 }, { "epoch": 0.03, "learning_rate": 9.431818181818183e-06, "loss": 0.972, "step": 332 }, { "epoch": 0.03, "learning_rate": 9.460227272727274e-06, "loss": 1.0208, "step": 333 }, { "epoch": 0.03, "learning_rate": 9.488636363636365e-06, "loss": 1.0031, "step": 334 }, { "epoch": 0.03, "learning_rate": 9.517045454545454e-06, "loss": 0.9915, "step": 335 }, { "epoch": 0.03, "learning_rate": 9.545454545454547e-06, "loss": 1.0029, "step": 336 }, { "epoch": 0.03, "learning_rate": 9.573863636363638e-06, "loss": 1.0085, "step": 337 }, { "epoch": 0.03, "learning_rate": 9.602272727272727e-06, "loss": 0.9719, "step": 338 }, { "epoch": 0.03, "learning_rate": 9.630681818181818e-06, "loss": 0.9613, "step": 339 }, { "epoch": 0.03, "learning_rate": 9.65909090909091e-06, "loss": 0.7094, "step": 340 }, { "epoch": 0.03, "learning_rate": 9.6875e-06, "loss": 0.734, "step": 341 }, { "epoch": 0.03, "learning_rate": 9.715909090909091e-06, "loss": 0.9982, "step": 342 }, { "epoch": 0.03, "learning_rate": 9.744318181818184e-06, "loss": 1.0066, "step": 343 }, { "epoch": 0.03, "learning_rate": 9.772727272727273e-06, "loss": 1.0137, "step": 344 }, { "epoch": 0.03, "learning_rate": 9.801136363636364e-06, "loss": 1.0203, "step": 345 }, { "epoch": 0.03, "learning_rate": 9.829545454545455e-06, "loss": 0.9391, "step": 346 }, { "epoch": 0.03, "learning_rate": 9.857954545454546e-06, "loss": 0.9818, "step": 347 }, { "epoch": 0.03, "learning_rate": 9.886363636363637e-06, "loss": 0.9347, "step": 348 }, { "epoch": 0.03, "learning_rate": 9.914772727272728e-06, "loss": 1.0357, "step": 349 }, { "epoch": 0.03, "learning_rate": 9.943181818181819e-06, "loss": 1.0331, "step": 350 }, { "epoch": 0.03, "learning_rate": 9.97159090909091e-06, "loss": 1.0127, "step": 351 }, { "epoch": 0.03, "learning_rate": 1e-05, "loss": 1.0277, "step": 352 }, { "epoch": 0.03, "learning_rate": 9.999999809037527e-06, "loss": 1.0276, "step": 353 }, { "epoch": 0.03, "learning_rate": 9.999999236150122e-06, "loss": 1.0237, "step": 354 }, { "epoch": 0.03, "learning_rate": 9.999998281337826e-06, "loss": 0.9979, "step": 355 }, { "epoch": 0.03, "learning_rate": 9.999996944600717e-06, "loss": 0.9834, "step": 356 }, { "epoch": 0.03, "learning_rate": 9.999995225938894e-06, "loss": 1.0121, "step": 357 }, { "epoch": 0.03, "learning_rate": 9.999993125352487e-06, "loss": 0.9634, "step": 358 }, { "epoch": 0.03, "learning_rate": 9.999990642841659e-06, "loss": 1.0269, "step": 359 }, { "epoch": 0.03, "learning_rate": 9.999987778406599e-06, "loss": 1.029, "step": 360 }, { "epoch": 0.03, "learning_rate": 9.999984532047527e-06, "loss": 1.021, "step": 361 }, { "epoch": 0.03, "learning_rate": 9.999980903764688e-06, "loss": 0.9958, "step": 362 }, { "epoch": 0.03, "learning_rate": 9.999976893558361e-06, "loss": 0.9451, "step": 363 }, { "epoch": 0.03, "learning_rate": 9.999972501428852e-06, "loss": 0.6688, "step": 364 }, { "epoch": 0.03, "learning_rate": 9.999967727376496e-06, "loss": 0.9726, "step": 365 }, { "epoch": 0.03, "learning_rate": 9.999962571401659e-06, "loss": 0.9864, "step": 366 }, { "epoch": 0.03, "learning_rate": 9.999957033504734e-06, "loss": 0.9641, "step": 367 }, { "epoch": 0.03, "learning_rate": 9.999951113686143e-06, "loss": 0.9929, "step": 368 }, { "epoch": 0.03, "learning_rate": 9.99994481194634e-06, "loss": 1.0053, "step": 369 }, { "epoch": 0.03, "learning_rate": 9.999938128285806e-06, "loss": 0.9677, "step": 370 }, { "epoch": 0.03, "learning_rate": 9.99993106270505e-06, "loss": 0.9778, "step": 371 }, { "epoch": 0.03, "learning_rate": 9.999923615204614e-06, "loss": 0.9889, "step": 372 }, { "epoch": 0.03, "learning_rate": 9.999915785785065e-06, "loss": 0.985, "step": 373 }, { "epoch": 0.03, "learning_rate": 9.999907574447002e-06, "loss": 0.9768, "step": 374 }, { "epoch": 0.03, "learning_rate": 9.999898981191052e-06, "loss": 0.9943, "step": 375 }, { "epoch": 0.03, "learning_rate": 9.99989000601787e-06, "loss": 0.9287, "step": 376 }, { "epoch": 0.03, "learning_rate": 9.999880648928146e-06, "loss": 0.9815, "step": 377 }, { "epoch": 0.03, "learning_rate": 9.999870909922589e-06, "loss": 0.9578, "step": 378 }, { "epoch": 0.03, "learning_rate": 9.999860789001947e-06, "loss": 1.0212, "step": 379 }, { "epoch": 0.03, "learning_rate": 9.999850286166992e-06, "loss": 1.0386, "step": 380 }, { "epoch": 0.03, "learning_rate": 9.999839401418525e-06, "loss": 0.9946, "step": 381 }, { "epoch": 0.03, "learning_rate": 9.999828134757379e-06, "loss": 0.9977, "step": 382 }, { "epoch": 0.03, "learning_rate": 9.999816486184412e-06, "loss": 1.0256, "step": 383 }, { "epoch": 0.03, "learning_rate": 9.999804455700518e-06, "loss": 1.0189, "step": 384 }, { "epoch": 0.03, "learning_rate": 9.999792043306613e-06, "loss": 0.9537, "step": 385 }, { "epoch": 0.03, "learning_rate": 9.999779249003647e-06, "loss": 1.0094, "step": 386 }, { "epoch": 0.03, "learning_rate": 9.999766072792594e-06, "loss": 1.0187, "step": 387 }, { "epoch": 0.03, "learning_rate": 9.999752514674463e-06, "loss": 0.9531, "step": 388 }, { "epoch": 0.03, "learning_rate": 9.999738574650291e-06, "loss": 0.9604, "step": 389 }, { "epoch": 0.03, "learning_rate": 9.999724252721139e-06, "loss": 0.971, "step": 390 }, { "epoch": 0.03, "learning_rate": 9.999709548888105e-06, "loss": 0.9422, "step": 391 }, { "epoch": 0.03, "learning_rate": 9.999694463152308e-06, "loss": 0.9596, "step": 392 }, { "epoch": 0.03, "learning_rate": 9.999678995514904e-06, "loss": 0.9795, "step": 393 }, { "epoch": 0.03, "learning_rate": 9.999663145977073e-06, "loss": 0.9576, "step": 394 }, { "epoch": 0.03, "learning_rate": 9.999646914540024e-06, "loss": 1.0091, "step": 395 }, { "epoch": 0.03, "learning_rate": 9.999630301205e-06, "loss": 0.973, "step": 396 }, { "epoch": 0.03, "learning_rate": 9.999613305973268e-06, "loss": 0.9579, "step": 397 }, { "epoch": 0.03, "learning_rate": 9.999595928846126e-06, "loss": 0.9883, "step": 398 }, { "epoch": 0.03, "learning_rate": 9.999578169824903e-06, "loss": 0.9854, "step": 399 }, { "epoch": 0.03, "learning_rate": 9.999560028910954e-06, "loss": 0.974, "step": 400 }, { "epoch": 0.03, "learning_rate": 9.999541506105666e-06, "loss": 1.0147, "step": 401 }, { "epoch": 0.03, "learning_rate": 9.999522601410453e-06, "loss": 0.9853, "step": 402 }, { "epoch": 0.03, "learning_rate": 9.999503314826759e-06, "loss": 1.0133, "step": 403 }, { "epoch": 0.03, "learning_rate": 9.999483646356056e-06, "loss": 0.9513, "step": 404 }, { "epoch": 0.03, "learning_rate": 9.99946359599985e-06, "loss": 1.0021, "step": 405 }, { "epoch": 0.03, "learning_rate": 9.999443163759669e-06, "loss": 0.977, "step": 406 }, { "epoch": 0.03, "learning_rate": 9.999422349637076e-06, "loss": 0.9823, "step": 407 }, { "epoch": 0.03, "learning_rate": 9.999401153633657e-06, "loss": 0.9933, "step": 408 }, { "epoch": 0.03, "learning_rate": 9.999379575751037e-06, "loss": 0.922, "step": 409 }, { "epoch": 0.03, "learning_rate": 9.999357615990859e-06, "loss": 0.9803, "step": 410 }, { "epoch": 0.04, "learning_rate": 9.999335274354805e-06, "loss": 1.0212, "step": 411 }, { "epoch": 0.04, "learning_rate": 9.999312550844578e-06, "loss": 1.0229, "step": 412 }, { "epoch": 0.04, "learning_rate": 9.999289445461914e-06, "loss": 0.9682, "step": 413 }, { "epoch": 0.04, "learning_rate": 9.99926595820858e-06, "loss": 0.958, "step": 414 }, { "epoch": 0.04, "learning_rate": 9.999242089086367e-06, "loss": 0.9829, "step": 415 }, { "epoch": 0.04, "learning_rate": 9.999217838097101e-06, "loss": 0.9971, "step": 416 }, { "epoch": 0.04, "learning_rate": 9.999193205242634e-06, "loss": 1.0245, "step": 417 }, { "epoch": 0.04, "learning_rate": 9.999168190524847e-06, "loss": 0.9479, "step": 418 }, { "epoch": 0.04, "learning_rate": 9.99914279394565e-06, "loss": 0.9499, "step": 419 }, { "epoch": 0.04, "learning_rate": 9.999117015506985e-06, "loss": 0.9738, "step": 420 }, { "epoch": 0.04, "learning_rate": 9.99909085521082e-06, "loss": 1.0132, "step": 421 }, { "epoch": 0.04, "learning_rate": 9.999064313059151e-06, "loss": 0.9911, "step": 422 }, { "epoch": 0.04, "learning_rate": 9.99903738905401e-06, "loss": 1.0288, "step": 423 }, { "epoch": 0.04, "learning_rate": 9.999010083197449e-06, "loss": 0.9622, "step": 424 }, { "epoch": 0.04, "learning_rate": 9.998982395491556e-06, "loss": 0.9578, "step": 425 }, { "epoch": 0.04, "learning_rate": 9.998954325938445e-06, "loss": 0.9576, "step": 426 }, { "epoch": 0.04, "learning_rate": 9.998925874540264e-06, "loss": 0.9133, "step": 427 }, { "epoch": 0.04, "learning_rate": 9.99889704129918e-06, "loss": 0.9878, "step": 428 }, { "epoch": 0.04, "learning_rate": 9.9988678262174e-06, "loss": 0.9758, "step": 429 }, { "epoch": 0.04, "learning_rate": 9.998838229297154e-06, "loss": 0.9601, "step": 430 }, { "epoch": 0.04, "learning_rate": 9.998808250540704e-06, "loss": 0.9816, "step": 431 }, { "epoch": 0.04, "learning_rate": 9.998777889950337e-06, "loss": 1.0047, "step": 432 }, { "epoch": 0.04, "learning_rate": 9.998747147528375e-06, "loss": 0.9467, "step": 433 }, { "epoch": 0.04, "learning_rate": 9.998716023277163e-06, "loss": 0.9865, "step": 434 }, { "epoch": 0.04, "learning_rate": 9.998684517199084e-06, "loss": 0.9599, "step": 435 }, { "epoch": 0.04, "learning_rate": 9.99865262929654e-06, "loss": 0.9849, "step": 436 }, { "epoch": 0.04, "learning_rate": 9.998620359571968e-06, "loss": 0.9826, "step": 437 }, { "epoch": 0.04, "learning_rate": 9.998587708027831e-06, "loss": 0.9858, "step": 438 }, { "epoch": 0.04, "learning_rate": 9.998554674666627e-06, "loss": 0.9944, "step": 439 }, { "epoch": 0.04, "learning_rate": 9.998521259490876e-06, "loss": 1.0202, "step": 440 }, { "epoch": 0.04, "learning_rate": 9.998487462503132e-06, "loss": 0.9148, "step": 441 }, { "epoch": 0.04, "learning_rate": 9.998453283705978e-06, "loss": 0.972, "step": 442 }, { "epoch": 0.04, "learning_rate": 9.998418723102022e-06, "loss": 1.0131, "step": 443 }, { "epoch": 0.04, "learning_rate": 9.998383780693904e-06, "loss": 0.5846, "step": 444 }, { "epoch": 0.04, "learning_rate": 9.998348456484294e-06, "loss": 0.9834, "step": 445 }, { "epoch": 0.04, "learning_rate": 9.998312750475889e-06, "loss": 0.9896, "step": 446 }, { "epoch": 0.04, "learning_rate": 9.99827666267142e-06, "loss": 0.9991, "step": 447 }, { "epoch": 0.04, "learning_rate": 9.99824019307364e-06, "loss": 0.9767, "step": 448 }, { "epoch": 0.04, "learning_rate": 9.998203341685336e-06, "loss": 0.9909, "step": 449 }, { "epoch": 0.04, "learning_rate": 9.998166108509322e-06, "loss": 0.9672, "step": 450 }, { "epoch": 0.04, "learning_rate": 9.998128493548444e-06, "loss": 0.9974, "step": 451 }, { "epoch": 0.04, "learning_rate": 9.998090496805575e-06, "loss": 0.6093, "step": 452 }, { "epoch": 0.04, "learning_rate": 9.998052118283615e-06, "loss": 0.9816, "step": 453 }, { "epoch": 0.04, "learning_rate": 9.998013357985497e-06, "loss": 0.7312, "step": 454 }, { "epoch": 0.04, "learning_rate": 9.99797421591418e-06, "loss": 0.9692, "step": 455 }, { "epoch": 0.04, "learning_rate": 9.99793469207266e-06, "loss": 0.9442, "step": 456 }, { "epoch": 0.04, "learning_rate": 9.997894786463949e-06, "loss": 0.9705, "step": 457 }, { "epoch": 0.04, "learning_rate": 9.997854499091098e-06, "loss": 0.9933, "step": 458 }, { "epoch": 0.04, "learning_rate": 9.997813829957183e-06, "loss": 0.974, "step": 459 }, { "epoch": 0.04, "learning_rate": 9.997772779065312e-06, "loss": 0.9932, "step": 460 }, { "epoch": 0.04, "learning_rate": 9.99773134641862e-06, "loss": 0.9747, "step": 461 }, { "epoch": 0.04, "learning_rate": 9.997689532020273e-06, "loss": 1.0012, "step": 462 }, { "epoch": 0.04, "learning_rate": 9.997647335873466e-06, "loss": 0.9534, "step": 463 }, { "epoch": 0.04, "learning_rate": 9.997604757981417e-06, "loss": 0.9999, "step": 464 }, { "epoch": 0.04, "learning_rate": 9.997561798347385e-06, "loss": 0.9206, "step": 465 }, { "epoch": 0.04, "learning_rate": 9.997518456974644e-06, "loss": 0.9483, "step": 466 }, { "epoch": 0.04, "learning_rate": 9.99747473386651e-06, "loss": 0.9948, "step": 467 }, { "epoch": 0.04, "learning_rate": 9.997430629026323e-06, "loss": 0.9479, "step": 468 }, { "epoch": 0.04, "learning_rate": 9.997386142457449e-06, "loss": 0.6672, "step": 469 }, { "epoch": 0.04, "learning_rate": 9.997341274163288e-06, "loss": 0.9381, "step": 470 }, { "epoch": 0.04, "learning_rate": 9.997296024147267e-06, "loss": 0.6636, "step": 471 }, { "epoch": 0.04, "learning_rate": 9.997250392412843e-06, "loss": 1.0235, "step": 472 }, { "epoch": 0.04, "learning_rate": 9.997204378963498e-06, "loss": 0.9215, "step": 473 }, { "epoch": 0.04, "learning_rate": 9.997157983802753e-06, "loss": 0.9933, "step": 474 }, { "epoch": 0.04, "learning_rate": 9.997111206934147e-06, "loss": 1.0055, "step": 475 }, { "epoch": 0.04, "learning_rate": 9.997064048361254e-06, "loss": 0.9593, "step": 476 }, { "epoch": 0.04, "learning_rate": 9.997016508087678e-06, "loss": 0.996, "step": 477 }, { "epoch": 0.04, "learning_rate": 9.996968586117047e-06, "loss": 0.9789, "step": 478 }, { "epoch": 0.04, "learning_rate": 9.996920282453025e-06, "loss": 0.9329, "step": 479 }, { "epoch": 0.04, "learning_rate": 9.9968715970993e-06, "loss": 0.9496, "step": 480 }, { "epoch": 0.04, "learning_rate": 9.99682253005959e-06, "loss": 1.0165, "step": 481 }, { "epoch": 0.04, "learning_rate": 9.996773081337645e-06, "loss": 0.9472, "step": 482 }, { "epoch": 0.04, "learning_rate": 9.996723250937241e-06, "loss": 1.0142, "step": 483 }, { "epoch": 0.04, "learning_rate": 9.996673038862187e-06, "loss": 1.0107, "step": 484 }, { "epoch": 0.04, "learning_rate": 9.996622445116312e-06, "loss": 0.9826, "step": 485 }, { "epoch": 0.04, "learning_rate": 9.996571469703487e-06, "loss": 0.9892, "step": 486 }, { "epoch": 0.04, "learning_rate": 9.996520112627602e-06, "loss": 0.9485, "step": 487 }, { "epoch": 0.04, "learning_rate": 9.996468373892583e-06, "loss": 0.9964, "step": 488 }, { "epoch": 0.04, "learning_rate": 9.996416253502379e-06, "loss": 0.9698, "step": 489 }, { "epoch": 0.04, "learning_rate": 9.996363751460971e-06, "loss": 0.9572, "step": 490 }, { "epoch": 0.04, "learning_rate": 9.996310867772373e-06, "loss": 0.9631, "step": 491 }, { "epoch": 0.04, "learning_rate": 9.996257602440622e-06, "loss": 0.9601, "step": 492 }, { "epoch": 0.04, "learning_rate": 9.996203955469788e-06, "loss": 1.0282, "step": 493 }, { "epoch": 0.04, "learning_rate": 9.996149926863964e-06, "loss": 0.9541, "step": 494 }, { "epoch": 0.04, "learning_rate": 9.996095516627284e-06, "loss": 0.9823, "step": 495 }, { "epoch": 0.04, "learning_rate": 9.996040724763902e-06, "loss": 0.9324, "step": 496 }, { "epoch": 0.04, "learning_rate": 9.995985551278e-06, "loss": 1.0424, "step": 497 }, { "epoch": 0.04, "learning_rate": 9.995929996173797e-06, "loss": 0.9691, "step": 498 }, { "epoch": 0.04, "learning_rate": 9.995874059455532e-06, "loss": 0.9769, "step": 499 }, { "epoch": 0.04, "learning_rate": 9.995817741127481e-06, "loss": 1.0372, "step": 500 }, { "epoch": 0.04, "learning_rate": 9.995761041193943e-06, "loss": 0.9531, "step": 501 }, { "epoch": 0.04, "learning_rate": 9.995703959659254e-06, "loss": 0.9387, "step": 502 }, { "epoch": 0.04, "learning_rate": 9.995646496527768e-06, "loss": 0.9255, "step": 503 }, { "epoch": 0.04, "learning_rate": 9.995588651803881e-06, "loss": 0.9362, "step": 504 }, { "epoch": 0.04, "learning_rate": 9.995530425492003e-06, "loss": 0.9446, "step": 505 }, { "epoch": 0.04, "learning_rate": 9.99547181759659e-06, "loss": 0.9947, "step": 506 }, { "epoch": 0.04, "learning_rate": 9.995412828122115e-06, "loss": 0.944, "step": 507 }, { "epoch": 0.04, "learning_rate": 9.995353457073082e-06, "loss": 0.9528, "step": 508 }, { "epoch": 0.04, "learning_rate": 9.99529370445403e-06, "loss": 0.8968, "step": 509 }, { "epoch": 0.04, "learning_rate": 9.995233570269522e-06, "loss": 0.953, "step": 510 }, { "epoch": 0.04, "learning_rate": 9.99517305452415e-06, "loss": 0.954, "step": 511 }, { "epoch": 0.04, "learning_rate": 9.995112157222537e-06, "loss": 0.995, "step": 512 }, { "epoch": 0.04, "learning_rate": 9.995050878369334e-06, "loss": 0.9409, "step": 513 }, { "epoch": 0.04, "learning_rate": 9.994989217969224e-06, "loss": 0.9443, "step": 514 }, { "epoch": 0.04, "learning_rate": 9.994927176026915e-06, "loss": 0.9894, "step": 515 }, { "epoch": 0.04, "learning_rate": 9.994864752547147e-06, "loss": 0.9457, "step": 516 }, { "epoch": 0.04, "learning_rate": 9.994801947534686e-06, "loss": 0.9465, "step": 517 }, { "epoch": 0.04, "learning_rate": 9.994738760994334e-06, "loss": 0.9547, "step": 518 }, { "epoch": 0.04, "learning_rate": 9.994675192930915e-06, "loss": 0.9232, "step": 519 }, { "epoch": 0.04, "learning_rate": 9.994611243349282e-06, "loss": 0.9432, "step": 520 }, { "epoch": 0.04, "learning_rate": 9.994546912254323e-06, "loss": 0.9662, "step": 521 }, { "epoch": 0.04, "learning_rate": 9.994482199650953e-06, "loss": 0.9431, "step": 522 }, { "epoch": 0.04, "learning_rate": 9.99441710554411e-06, "loss": 0.9731, "step": 523 }, { "epoch": 0.04, "learning_rate": 9.994351629938772e-06, "loss": 0.9986, "step": 524 }, { "epoch": 0.04, "learning_rate": 9.994285772839936e-06, "loss": 0.9775, "step": 525 }, { "epoch": 0.04, "learning_rate": 9.994219534252633e-06, "loss": 0.921, "step": 526 }, { "epoch": 0.04, "learning_rate": 9.994152914181927e-06, "loss": 0.9635, "step": 527 }, { "epoch": 0.05, "learning_rate": 9.994085912632903e-06, "loss": 0.9771, "step": 528 }, { "epoch": 0.05, "learning_rate": 9.994018529610678e-06, "loss": 0.9644, "step": 529 }, { "epoch": 0.05, "learning_rate": 9.993950765120402e-06, "loss": 0.9495, "step": 530 }, { "epoch": 0.05, "learning_rate": 9.99388261916725e-06, "loss": 0.9562, "step": 531 }, { "epoch": 0.05, "learning_rate": 9.993814091756427e-06, "loss": 0.9406, "step": 532 }, { "epoch": 0.05, "learning_rate": 9.993745182893167e-06, "loss": 0.9595, "step": 533 }, { "epoch": 0.05, "learning_rate": 9.993675892582732e-06, "loss": 0.9124, "step": 534 }, { "epoch": 0.05, "learning_rate": 9.99360622083042e-06, "loss": 0.9347, "step": 535 }, { "epoch": 0.05, "learning_rate": 9.99353616764155e-06, "loss": 0.9211, "step": 536 }, { "epoch": 0.05, "learning_rate": 9.993465733021472e-06, "loss": 0.9707, "step": 537 }, { "epoch": 0.05, "learning_rate": 9.993394916975565e-06, "loss": 0.9231, "step": 538 }, { "epoch": 0.05, "learning_rate": 9.993323719509243e-06, "loss": 0.9614, "step": 539 }, { "epoch": 0.05, "learning_rate": 9.99325214062794e-06, "loss": 0.9895, "step": 540 }, { "epoch": 0.05, "learning_rate": 9.993180180337126e-06, "loss": 0.9984, "step": 541 }, { "epoch": 0.05, "learning_rate": 9.993107838642298e-06, "loss": 0.9842, "step": 542 }, { "epoch": 0.05, "learning_rate": 9.993035115548978e-06, "loss": 0.9711, "step": 543 }, { "epoch": 0.05, "learning_rate": 9.992962011062726e-06, "loss": 0.9828, "step": 544 }, { "epoch": 0.05, "learning_rate": 9.992888525189122e-06, "loss": 0.9528, "step": 545 }, { "epoch": 0.05, "learning_rate": 9.992814657933782e-06, "loss": 0.974, "step": 546 }, { "epoch": 0.05, "learning_rate": 9.992740409302349e-06, "loss": 0.9553, "step": 547 }, { "epoch": 0.05, "learning_rate": 9.99266577930049e-06, "loss": 0.9646, "step": 548 }, { "epoch": 0.05, "learning_rate": 9.99259076793391e-06, "loss": 0.6714, "step": 549 }, { "epoch": 0.05, "learning_rate": 9.992515375208335e-06, "loss": 0.9507, "step": 550 }, { "epoch": 0.05, "learning_rate": 9.99243960112953e-06, "loss": 0.9994, "step": 551 }, { "epoch": 0.05, "learning_rate": 9.992363445703275e-06, "loss": 0.9598, "step": 552 }, { "epoch": 0.05, "learning_rate": 9.992286908935393e-06, "loss": 0.9169, "step": 553 }, { "epoch": 0.05, "learning_rate": 9.992209990831729e-06, "loss": 0.9399, "step": 554 }, { "epoch": 0.05, "learning_rate": 9.992132691398158e-06, "loss": 1.0052, "step": 555 }, { "epoch": 0.05, "learning_rate": 9.992055010640583e-06, "loss": 0.9454, "step": 556 }, { "epoch": 0.05, "learning_rate": 9.991976948564938e-06, "loss": 0.9636, "step": 557 }, { "epoch": 0.05, "learning_rate": 9.991898505177188e-06, "loss": 0.9283, "step": 558 }, { "epoch": 0.05, "learning_rate": 9.991819680483325e-06, "loss": 0.9134, "step": 559 }, { "epoch": 0.05, "learning_rate": 9.991740474489369e-06, "loss": 0.9714, "step": 560 }, { "epoch": 0.05, "learning_rate": 9.991660887201367e-06, "loss": 0.9508, "step": 561 }, { "epoch": 0.05, "learning_rate": 9.991580918625403e-06, "loss": 0.9275, "step": 562 }, { "epoch": 0.05, "learning_rate": 9.991500568767583e-06, "loss": 1.0031, "step": 563 }, { "epoch": 0.05, "learning_rate": 9.991419837634044e-06, "loss": 0.9517, "step": 564 }, { "epoch": 0.05, "learning_rate": 9.991338725230956e-06, "loss": 1.016, "step": 565 }, { "epoch": 0.05, "learning_rate": 9.991257231564511e-06, "loss": 0.9515, "step": 566 }, { "epoch": 0.05, "learning_rate": 9.991175356640936e-06, "loss": 0.9509, "step": 567 }, { "epoch": 0.05, "learning_rate": 9.991093100466482e-06, "loss": 0.9809, "step": 568 }, { "epoch": 0.05, "learning_rate": 9.991010463047437e-06, "loss": 0.9614, "step": 569 }, { "epoch": 0.05, "learning_rate": 9.99092744439011e-06, "loss": 0.9655, "step": 570 }, { "epoch": 0.05, "learning_rate": 9.990844044500842e-06, "loss": 0.9147, "step": 571 }, { "epoch": 0.05, "learning_rate": 9.990760263386006e-06, "loss": 0.9551, "step": 572 }, { "epoch": 0.05, "learning_rate": 9.990676101052001e-06, "loss": 1.018, "step": 573 }, { "epoch": 0.05, "learning_rate": 9.990591557505254e-06, "loss": 1.007, "step": 574 }, { "epoch": 0.05, "learning_rate": 9.990506632752224e-06, "loss": 0.9351, "step": 575 }, { "epoch": 0.05, "learning_rate": 9.990421326799398e-06, "loss": 0.9629, "step": 576 }, { "epoch": 0.05, "learning_rate": 9.99033563965329e-06, "loss": 0.9696, "step": 577 }, { "epoch": 0.05, "learning_rate": 9.990249571320449e-06, "loss": 0.9675, "step": 578 }, { "epoch": 0.05, "learning_rate": 9.990163121807448e-06, "loss": 0.8714, "step": 579 }, { "epoch": 0.05, "learning_rate": 9.990076291120888e-06, "loss": 0.9434, "step": 580 }, { "epoch": 0.05, "learning_rate": 9.989989079267404e-06, "loss": 0.9815, "step": 581 }, { "epoch": 0.05, "learning_rate": 9.989901486253658e-06, "loss": 0.965, "step": 582 }, { "epoch": 0.05, "learning_rate": 9.98981351208634e-06, "loss": 0.9654, "step": 583 }, { "epoch": 0.05, "learning_rate": 9.989725156772168e-06, "loss": 0.9562, "step": 584 }, { "epoch": 0.05, "learning_rate": 9.989636420317894e-06, "loss": 0.8825, "step": 585 }, { "epoch": 0.05, "learning_rate": 9.989547302730295e-06, "loss": 0.9448, "step": 586 }, { "epoch": 0.05, "learning_rate": 9.98945780401618e-06, "loss": 1.0107, "step": 587 }, { "epoch": 0.05, "learning_rate": 9.989367924182381e-06, "loss": 0.7184, "step": 588 }, { "epoch": 0.05, "learning_rate": 9.989277663235768e-06, "loss": 0.9486, "step": 589 }, { "epoch": 0.05, "learning_rate": 9.989187021183232e-06, "loss": 0.9629, "step": 590 }, { "epoch": 0.05, "learning_rate": 9.9890959980317e-06, "loss": 0.9381, "step": 591 }, { "epoch": 0.05, "learning_rate": 9.98900459378812e-06, "loss": 0.9561, "step": 592 }, { "epoch": 0.05, "learning_rate": 9.98891280845948e-06, "loss": 0.951, "step": 593 }, { "epoch": 0.05, "learning_rate": 9.988820642052788e-06, "loss": 0.9909, "step": 594 }, { "epoch": 0.05, "learning_rate": 9.988728094575082e-06, "loss": 0.9805, "step": 595 }, { "epoch": 0.05, "learning_rate": 9.988635166033435e-06, "loss": 0.9619, "step": 596 }, { "epoch": 0.05, "learning_rate": 9.988541856434943e-06, "loss": 0.9774, "step": 597 }, { "epoch": 0.05, "learning_rate": 9.988448165786735e-06, "loss": 0.968, "step": 598 }, { "epoch": 0.05, "learning_rate": 9.988354094095967e-06, "loss": 0.968, "step": 599 }, { "epoch": 0.05, "learning_rate": 9.988259641369824e-06, "loss": 0.954, "step": 600 }, { "epoch": 0.05, "learning_rate": 9.98816480761552e-06, "loss": 0.9699, "step": 601 }, { "epoch": 0.05, "learning_rate": 9.988069592840303e-06, "loss": 0.9594, "step": 602 }, { "epoch": 0.05, "learning_rate": 9.987973997051442e-06, "loss": 0.9718, "step": 603 }, { "epoch": 0.05, "learning_rate": 9.987878020256238e-06, "loss": 0.9129, "step": 604 }, { "epoch": 0.05, "learning_rate": 9.987781662462026e-06, "loss": 0.9596, "step": 605 }, { "epoch": 0.05, "learning_rate": 9.987684923676165e-06, "loss": 0.995, "step": 606 }, { "epoch": 0.05, "learning_rate": 9.987587803906043e-06, "loss": 0.9916, "step": 607 }, { "epoch": 0.05, "learning_rate": 9.98749030315908e-06, "loss": 0.9151, "step": 608 }, { "epoch": 0.05, "learning_rate": 9.987392421442724e-06, "loss": 0.9828, "step": 609 }, { "epoch": 0.05, "learning_rate": 9.987294158764451e-06, "loss": 0.935, "step": 610 }, { "epoch": 0.05, "learning_rate": 9.987195515131765e-06, "loss": 0.6444, "step": 611 }, { "epoch": 0.05, "learning_rate": 9.987096490552203e-06, "loss": 0.9459, "step": 612 }, { "epoch": 0.05, "learning_rate": 9.986997085033327e-06, "loss": 0.9162, "step": 613 }, { "epoch": 0.05, "learning_rate": 9.986897298582733e-06, "loss": 0.9677, "step": 614 }, { "epoch": 0.05, "learning_rate": 9.986797131208044e-06, "loss": 0.8951, "step": 615 }, { "epoch": 0.05, "learning_rate": 9.986696582916907e-06, "loss": 0.914, "step": 616 }, { "epoch": 0.05, "learning_rate": 9.986595653717005e-06, "loss": 0.9727, "step": 617 }, { "epoch": 0.05, "learning_rate": 9.986494343616044e-06, "loss": 1.007, "step": 618 }, { "epoch": 0.05, "learning_rate": 9.986392652621769e-06, "loss": 0.9983, "step": 619 }, { "epoch": 0.05, "learning_rate": 9.986290580741944e-06, "loss": 0.9386, "step": 620 }, { "epoch": 0.05, "learning_rate": 9.986188127984364e-06, "loss": 0.9431, "step": 621 }, { "epoch": 0.05, "learning_rate": 9.986085294356858e-06, "loss": 1.0024, "step": 622 }, { "epoch": 0.05, "learning_rate": 9.985982079867278e-06, "loss": 0.9498, "step": 623 }, { "epoch": 0.05, "learning_rate": 9.985878484523512e-06, "loss": 0.9855, "step": 624 }, { "epoch": 0.05, "learning_rate": 9.98577450833347e-06, "loss": 0.9632, "step": 625 }, { "epoch": 0.05, "learning_rate": 9.985670151305093e-06, "loss": 0.9713, "step": 626 }, { "epoch": 0.05, "learning_rate": 9.985565413446357e-06, "loss": 0.9456, "step": 627 }, { "epoch": 0.05, "learning_rate": 9.985460294765258e-06, "loss": 0.9581, "step": 628 }, { "epoch": 0.05, "learning_rate": 9.98535479526983e-06, "loss": 0.9399, "step": 629 }, { "epoch": 0.05, "learning_rate": 9.985248914968127e-06, "loss": 0.8856, "step": 630 }, { "epoch": 0.05, "learning_rate": 9.98514265386824e-06, "loss": 0.996, "step": 631 }, { "epoch": 0.05, "learning_rate": 9.985036011978281e-06, "loss": 0.9788, "step": 632 }, { "epoch": 0.05, "learning_rate": 9.984928989306403e-06, "loss": 0.9426, "step": 633 }, { "epoch": 0.05, "learning_rate": 9.984821585860776e-06, "loss": 0.9221, "step": 634 }, { "epoch": 0.05, "learning_rate": 9.984713801649603e-06, "loss": 0.9986, "step": 635 }, { "epoch": 0.05, "learning_rate": 9.98460563668112e-06, "loss": 0.9519, "step": 636 }, { "epoch": 0.05, "learning_rate": 9.984497090963589e-06, "loss": 0.9104, "step": 637 }, { "epoch": 0.05, "learning_rate": 9.9843881645053e-06, "loss": 0.945, "step": 638 }, { "epoch": 0.05, "learning_rate": 9.984278857314575e-06, "loss": 0.9286, "step": 639 }, { "epoch": 0.05, "learning_rate": 9.984169169399763e-06, "loss": 0.9356, "step": 640 }, { "epoch": 0.05, "learning_rate": 9.984059100769239e-06, "loss": 0.9462, "step": 641 }, { "epoch": 0.05, "learning_rate": 9.983948651431416e-06, "loss": 0.9373, "step": 642 }, { "epoch": 0.05, "learning_rate": 9.983837821394726e-06, "loss": 0.9738, "step": 643 }, { "epoch": 0.05, "learning_rate": 9.98372661066764e-06, "loss": 0.9412, "step": 644 }, { "epoch": 0.06, "learning_rate": 9.983615019258647e-06, "loss": 0.9068, "step": 645 }, { "epoch": 0.06, "learning_rate": 9.983503047176274e-06, "loss": 0.9778, "step": 646 }, { "epoch": 0.06, "learning_rate": 9.983390694429073e-06, "loss": 0.96, "step": 647 }, { "epoch": 0.06, "learning_rate": 9.983277961025628e-06, "loss": 0.9645, "step": 648 }, { "epoch": 0.06, "learning_rate": 9.983164846974549e-06, "loss": 0.9066, "step": 649 }, { "epoch": 0.06, "learning_rate": 9.983051352284476e-06, "loss": 0.9364, "step": 650 }, { "epoch": 0.06, "learning_rate": 9.982937476964078e-06, "loss": 0.95, "step": 651 }, { "epoch": 0.06, "learning_rate": 9.982823221022053e-06, "loss": 1.0069, "step": 652 }, { "epoch": 0.06, "learning_rate": 9.982708584467132e-06, "loss": 0.9274, "step": 653 }, { "epoch": 0.06, "learning_rate": 9.982593567308066e-06, "loss": 0.9151, "step": 654 }, { "epoch": 0.06, "learning_rate": 9.982478169553642e-06, "loss": 0.9346, "step": 655 }, { "epoch": 0.06, "learning_rate": 9.982362391212679e-06, "loss": 0.902, "step": 656 }, { "epoch": 0.06, "learning_rate": 9.982246232294015e-06, "loss": 0.9496, "step": 657 }, { "epoch": 0.06, "learning_rate": 9.982129692806526e-06, "loss": 0.9413, "step": 658 }, { "epoch": 0.06, "learning_rate": 9.982012772759115e-06, "loss": 0.9965, "step": 659 }, { "epoch": 0.06, "learning_rate": 9.98189547216071e-06, "loss": 0.9324, "step": 660 }, { "epoch": 0.06, "learning_rate": 9.981777791020273e-06, "loss": 0.9274, "step": 661 }, { "epoch": 0.06, "learning_rate": 9.98165972934679e-06, "loss": 0.9612, "step": 662 }, { "epoch": 0.06, "learning_rate": 9.981541287149283e-06, "loss": 1.0121, "step": 663 }, { "epoch": 0.06, "learning_rate": 9.981422464436799e-06, "loss": 0.933, "step": 664 }, { "epoch": 0.06, "learning_rate": 9.981303261218412e-06, "loss": 0.9515, "step": 665 }, { "epoch": 0.06, "learning_rate": 9.981183677503228e-06, "loss": 0.9531, "step": 666 }, { "epoch": 0.06, "learning_rate": 9.98106371330038e-06, "loss": 0.9966, "step": 667 }, { "epoch": 0.06, "learning_rate": 9.980943368619036e-06, "loss": 0.9277, "step": 668 }, { "epoch": 0.06, "learning_rate": 9.980822643468383e-06, "loss": 0.9644, "step": 669 }, { "epoch": 0.06, "learning_rate": 9.980701537857646e-06, "loss": 0.9209, "step": 670 }, { "epoch": 0.06, "learning_rate": 9.980580051796076e-06, "loss": 0.9369, "step": 671 }, { "epoch": 0.06, "learning_rate": 9.980458185292952e-06, "loss": 0.923, "step": 672 }, { "epoch": 0.06, "learning_rate": 9.98033593835758e-06, "loss": 0.9357, "step": 673 }, { "epoch": 0.06, "learning_rate": 9.980213310999302e-06, "loss": 0.9955, "step": 674 }, { "epoch": 0.06, "learning_rate": 9.980090303227482e-06, "loss": 1.0023, "step": 675 }, { "epoch": 0.06, "learning_rate": 9.979966915051517e-06, "loss": 0.9289, "step": 676 }, { "epoch": 0.06, "learning_rate": 9.979843146480835e-06, "loss": 0.9177, "step": 677 }, { "epoch": 0.06, "learning_rate": 9.979718997524884e-06, "loss": 0.9796, "step": 678 }, { "epoch": 0.06, "learning_rate": 9.97959446819315e-06, "loss": 0.9638, "step": 679 }, { "epoch": 0.06, "learning_rate": 9.979469558495146e-06, "loss": 0.914, "step": 680 }, { "epoch": 0.06, "learning_rate": 9.979344268440414e-06, "loss": 0.9623, "step": 681 }, { "epoch": 0.06, "learning_rate": 9.97921859803852e-06, "loss": 0.9701, "step": 682 }, { "epoch": 0.06, "learning_rate": 9.97909254729907e-06, "loss": 0.9307, "step": 683 }, { "epoch": 0.06, "learning_rate": 9.978966116231685e-06, "loss": 0.8966, "step": 684 }, { "epoch": 0.06, "learning_rate": 9.978839304846028e-06, "loss": 0.9407, "step": 685 }, { "epoch": 0.06, "learning_rate": 9.978712113151783e-06, "loss": 0.9261, "step": 686 }, { "epoch": 0.06, "learning_rate": 9.978584541158666e-06, "loss": 0.9843, "step": 687 }, { "epoch": 0.06, "learning_rate": 9.978456588876423e-06, "loss": 0.6669, "step": 688 }, { "epoch": 0.06, "learning_rate": 9.978328256314825e-06, "loss": 0.9704, "step": 689 }, { "epoch": 0.06, "learning_rate": 9.978199543483677e-06, "loss": 0.9172, "step": 690 }, { "epoch": 0.06, "learning_rate": 9.978070450392809e-06, "loss": 1.004, "step": 691 }, { "epoch": 0.06, "learning_rate": 9.97794097705208e-06, "loss": 0.6835, "step": 692 }, { "epoch": 0.06, "learning_rate": 9.977811123471386e-06, "loss": 0.9487, "step": 693 }, { "epoch": 0.06, "learning_rate": 9.97768088966064e-06, "loss": 0.9404, "step": 694 }, { "epoch": 0.06, "learning_rate": 9.977550275629794e-06, "loss": 0.9136, "step": 695 }, { "epoch": 0.06, "learning_rate": 9.977419281388822e-06, "loss": 0.9875, "step": 696 }, { "epoch": 0.06, "learning_rate": 9.977287906947731e-06, "loss": 0.9676, "step": 697 }, { "epoch": 0.06, "learning_rate": 9.977156152316556e-06, "loss": 0.9239, "step": 698 }, { "epoch": 0.06, "learning_rate": 9.97702401750536e-06, "loss": 0.9319, "step": 699 }, { "epoch": 0.06, "learning_rate": 9.976891502524239e-06, "loss": 0.9285, "step": 700 }, { "epoch": 0.06, "learning_rate": 9.976758607383312e-06, "loss": 0.9751, "step": 701 }, { "epoch": 0.06, "learning_rate": 9.976625332092733e-06, "loss": 0.9837, "step": 702 }, { "epoch": 0.06, "learning_rate": 9.976491676662679e-06, "loss": 0.9366, "step": 703 }, { "epoch": 0.06, "learning_rate": 9.976357641103363e-06, "loss": 0.9351, "step": 704 }, { "epoch": 0.06, "learning_rate": 9.976223225425021e-06, "loss": 0.9467, "step": 705 }, { "epoch": 0.06, "learning_rate": 9.97608842963792e-06, "loss": 0.9031, "step": 706 }, { "epoch": 0.06, "learning_rate": 9.975953253752357e-06, "loss": 0.9625, "step": 707 }, { "epoch": 0.06, "learning_rate": 9.975817697778657e-06, "loss": 0.9753, "step": 708 }, { "epoch": 0.06, "learning_rate": 9.975681761727176e-06, "loss": 0.9925, "step": 709 }, { "epoch": 0.06, "learning_rate": 9.975545445608297e-06, "loss": 0.9238, "step": 710 }, { "epoch": 0.06, "learning_rate": 9.975408749432432e-06, "loss": 0.991, "step": 711 }, { "epoch": 0.06, "learning_rate": 9.975271673210021e-06, "loss": 0.9452, "step": 712 }, { "epoch": 0.06, "learning_rate": 9.975134216951537e-06, "loss": 0.9451, "step": 713 }, { "epoch": 0.06, "learning_rate": 9.974996380667477e-06, "loss": 0.9617, "step": 714 }, { "epoch": 0.06, "learning_rate": 9.974858164368373e-06, "loss": 0.9242, "step": 715 }, { "epoch": 0.06, "learning_rate": 9.97471956806478e-06, "loss": 0.9099, "step": 716 }, { "epoch": 0.06, "learning_rate": 9.974580591767286e-06, "loss": 0.9662, "step": 717 }, { "epoch": 0.06, "learning_rate": 9.974441235486506e-06, "loss": 0.9857, "step": 718 }, { "epoch": 0.06, "learning_rate": 9.974301499233086e-06, "loss": 0.9165, "step": 719 }, { "epoch": 0.06, "learning_rate": 9.974161383017697e-06, "loss": 0.6457, "step": 720 }, { "epoch": 0.06, "learning_rate": 9.974020886851046e-06, "loss": 0.9459, "step": 721 }, { "epoch": 0.06, "learning_rate": 9.97388001074386e-06, "loss": 0.9567, "step": 722 }, { "epoch": 0.06, "learning_rate": 9.973738754706903e-06, "loss": 0.9756, "step": 723 }, { "epoch": 0.06, "learning_rate": 9.973597118750964e-06, "loss": 0.9776, "step": 724 }, { "epoch": 0.06, "learning_rate": 9.973455102886862e-06, "loss": 0.9781, "step": 725 }, { "epoch": 0.06, "learning_rate": 9.973312707125445e-06, "loss": 0.9416, "step": 726 }, { "epoch": 0.06, "learning_rate": 9.973169931477588e-06, "loss": 0.9734, "step": 727 }, { "epoch": 0.06, "learning_rate": 9.973026775954202e-06, "loss": 0.8897, "step": 728 }, { "epoch": 0.06, "learning_rate": 9.972883240566216e-06, "loss": 0.9362, "step": 729 }, { "epoch": 0.06, "learning_rate": 9.972739325324596e-06, "loss": 0.942, "step": 730 }, { "epoch": 0.06, "learning_rate": 9.972595030240338e-06, "loss": 0.9008, "step": 731 }, { "epoch": 0.06, "learning_rate": 9.97245035532446e-06, "loss": 0.9863, "step": 732 }, { "epoch": 0.06, "learning_rate": 9.972305300588012e-06, "loss": 0.7748, "step": 733 }, { "epoch": 0.06, "learning_rate": 9.972159866042078e-06, "loss": 0.9155, "step": 734 }, { "epoch": 0.06, "learning_rate": 9.972014051697765e-06, "loss": 0.9649, "step": 735 }, { "epoch": 0.06, "learning_rate": 9.971867857566214e-06, "loss": 0.9417, "step": 736 }, { "epoch": 0.06, "learning_rate": 9.971721283658587e-06, "loss": 0.8669, "step": 737 }, { "epoch": 0.06, "learning_rate": 9.971574329986083e-06, "loss": 0.995, "step": 738 }, { "epoch": 0.06, "learning_rate": 9.971426996559926e-06, "loss": 0.9353, "step": 739 }, { "epoch": 0.06, "learning_rate": 9.971279283391372e-06, "loss": 0.9436, "step": 740 }, { "epoch": 0.06, "learning_rate": 9.971131190491703e-06, "loss": 0.9239, "step": 741 }, { "epoch": 0.06, "learning_rate": 9.970982717872228e-06, "loss": 0.9611, "step": 742 }, { "epoch": 0.06, "learning_rate": 9.970833865544293e-06, "loss": 0.9395, "step": 743 }, { "epoch": 0.06, "learning_rate": 9.970684633519266e-06, "loss": 0.8878, "step": 744 }, { "epoch": 0.06, "learning_rate": 9.970535021808545e-06, "loss": 0.9332, "step": 745 }, { "epoch": 0.06, "learning_rate": 9.97038503042356e-06, "loss": 0.9312, "step": 746 }, { "epoch": 0.06, "learning_rate": 9.970234659375765e-06, "loss": 0.9501, "step": 747 }, { "epoch": 0.06, "learning_rate": 9.97008390867665e-06, "loss": 0.9847, "step": 748 }, { "epoch": 0.06, "learning_rate": 9.96993277833773e-06, "loss": 0.9682, "step": 749 }, { "epoch": 0.06, "learning_rate": 9.969781268370546e-06, "loss": 0.9245, "step": 750 }, { "epoch": 0.06, "learning_rate": 9.969629378786672e-06, "loss": 0.8921, "step": 751 }, { "epoch": 0.06, "learning_rate": 9.969477109597712e-06, "loss": 0.9283, "step": 752 }, { "epoch": 0.06, "learning_rate": 9.969324460815294e-06, "loss": 0.9517, "step": 753 }, { "epoch": 0.06, "learning_rate": 9.96917143245108e-06, "loss": 0.9449, "step": 754 }, { "epoch": 0.06, "learning_rate": 9.96901802451676e-06, "loss": 0.9379, "step": 755 }, { "epoch": 0.06, "learning_rate": 9.968864237024049e-06, "loss": 0.9445, "step": 756 }, { "epoch": 0.06, "learning_rate": 9.968710069984699e-06, "loss": 0.9441, "step": 757 }, { "epoch": 0.06, "learning_rate": 9.968555523410481e-06, "loss": 0.9358, "step": 758 }, { "epoch": 0.06, "learning_rate": 9.968400597313202e-06, "loss": 1.0375, "step": 759 }, { "epoch": 0.06, "learning_rate": 9.968245291704696e-06, "loss": 0.9682, "step": 760 }, { "epoch": 0.06, "learning_rate": 9.968089606596826e-06, "loss": 0.9305, "step": 761 }, { "epoch": 0.07, "learning_rate": 9.967933542001487e-06, "loss": 0.9389, "step": 762 }, { "epoch": 0.07, "learning_rate": 9.967777097930594e-06, "loss": 0.9991, "step": 763 }, { "epoch": 0.07, "learning_rate": 9.967620274396102e-06, "loss": 0.9615, "step": 764 }, { "epoch": 0.07, "learning_rate": 9.967463071409986e-06, "loss": 0.9685, "step": 765 }, { "epoch": 0.07, "learning_rate": 9.967305488984257e-06, "loss": 0.9322, "step": 766 }, { "epoch": 0.07, "learning_rate": 9.967147527130951e-06, "loss": 0.9357, "step": 767 }, { "epoch": 0.07, "learning_rate": 9.966989185862134e-06, "loss": 0.9341, "step": 768 }, { "epoch": 0.07, "learning_rate": 9.966830465189901e-06, "loss": 0.9619, "step": 769 }, { "epoch": 0.07, "learning_rate": 9.966671365126374e-06, "loss": 0.9237, "step": 770 }, { "epoch": 0.07, "learning_rate": 9.96651188568371e-06, "loss": 0.9798, "step": 771 }, { "epoch": 0.07, "learning_rate": 9.966352026874086e-06, "loss": 0.9468, "step": 772 }, { "epoch": 0.07, "learning_rate": 9.966191788709716e-06, "loss": 0.9937, "step": 773 }, { "epoch": 0.07, "learning_rate": 9.966031171202839e-06, "loss": 0.9373, "step": 774 }, { "epoch": 0.07, "learning_rate": 9.965870174365724e-06, "loss": 0.9277, "step": 775 }, { "epoch": 0.07, "learning_rate": 9.965708798210666e-06, "loss": 0.9407, "step": 776 }, { "epoch": 0.07, "learning_rate": 9.965547042749997e-06, "loss": 0.9026, "step": 777 }, { "epoch": 0.07, "learning_rate": 9.965384907996069e-06, "loss": 0.9417, "step": 778 }, { "epoch": 0.07, "learning_rate": 9.965222393961267e-06, "loss": 0.9088, "step": 779 }, { "epoch": 0.07, "learning_rate": 9.965059500658007e-06, "loss": 0.9263, "step": 780 }, { "epoch": 0.07, "learning_rate": 9.964896228098728e-06, "loss": 0.935, "step": 781 }, { "epoch": 0.07, "learning_rate": 9.964732576295903e-06, "loss": 0.8964, "step": 782 }, { "epoch": 0.07, "learning_rate": 9.964568545262034e-06, "loss": 0.9582, "step": 783 }, { "epoch": 0.07, "learning_rate": 9.964404135009649e-06, "loss": 0.8915, "step": 784 }, { "epoch": 0.07, "learning_rate": 9.964239345551306e-06, "loss": 0.9129, "step": 785 }, { "epoch": 0.07, "learning_rate": 9.964074176899594e-06, "loss": 0.9206, "step": 786 }, { "epoch": 0.07, "learning_rate": 9.96390862906713e-06, "loss": 0.8965, "step": 787 }, { "epoch": 0.07, "learning_rate": 9.963742702066556e-06, "loss": 0.9489, "step": 788 }, { "epoch": 0.07, "learning_rate": 9.963576395910548e-06, "loss": 0.9048, "step": 789 }, { "epoch": 0.07, "learning_rate": 9.963409710611811e-06, "loss": 0.8889, "step": 790 }, { "epoch": 0.07, "learning_rate": 9.963242646183075e-06, "loss": 0.929, "step": 791 }, { "epoch": 0.07, "learning_rate": 9.963075202637104e-06, "loss": 0.9213, "step": 792 }, { "epoch": 0.07, "learning_rate": 9.962907379986685e-06, "loss": 0.9204, "step": 793 }, { "epoch": 0.07, "learning_rate": 9.962739178244638e-06, "loss": 0.9318, "step": 794 }, { "epoch": 0.07, "learning_rate": 9.962570597423813e-06, "loss": 0.911, "step": 795 }, { "epoch": 0.07, "learning_rate": 9.962401637537084e-06, "loss": 0.9422, "step": 796 }, { "epoch": 0.07, "learning_rate": 9.962232298597359e-06, "loss": 0.9148, "step": 797 }, { "epoch": 0.07, "learning_rate": 9.962062580617574e-06, "loss": 0.9633, "step": 798 }, { "epoch": 0.07, "learning_rate": 9.961892483610688e-06, "loss": 0.9279, "step": 799 }, { "epoch": 0.07, "learning_rate": 9.9617220075897e-06, "loss": 0.9211, "step": 800 }, { "epoch": 0.07, "learning_rate": 9.961551152567628e-06, "loss": 0.9441, "step": 801 }, { "epoch": 0.07, "learning_rate": 9.961379918557526e-06, "loss": 0.9152, "step": 802 }, { "epoch": 0.07, "learning_rate": 9.96120830557247e-06, "loss": 0.9532, "step": 803 }, { "epoch": 0.07, "learning_rate": 9.96103631362557e-06, "loss": 0.9603, "step": 804 }, { "epoch": 0.07, "learning_rate": 9.960863942729964e-06, "loss": 0.6995, "step": 805 }, { "epoch": 0.07, "learning_rate": 9.960691192898818e-06, "loss": 0.9402, "step": 806 }, { "epoch": 0.07, "learning_rate": 9.960518064145329e-06, "loss": 0.9308, "step": 807 }, { "epoch": 0.07, "learning_rate": 9.960344556482717e-06, "loss": 0.9435, "step": 808 }, { "epoch": 0.07, "learning_rate": 9.960170669924241e-06, "loss": 1.011, "step": 809 }, { "epoch": 0.07, "learning_rate": 9.959996404483183e-06, "loss": 0.9621, "step": 810 }, { "epoch": 0.07, "learning_rate": 9.959821760172849e-06, "loss": 0.9782, "step": 811 }, { "epoch": 0.07, "learning_rate": 9.959646737006586e-06, "loss": 0.96, "step": 812 }, { "epoch": 0.07, "learning_rate": 9.959471334997757e-06, "loss": 0.6261, "step": 813 }, { "epoch": 0.07, "learning_rate": 9.959295554159762e-06, "loss": 0.9651, "step": 814 }, { "epoch": 0.07, "learning_rate": 9.95911939450603e-06, "loss": 0.937, "step": 815 }, { "epoch": 0.07, "learning_rate": 9.958942856050015e-06, "loss": 0.9293, "step": 816 }, { "epoch": 0.07, "learning_rate": 9.958765938805204e-06, "loss": 0.9005, "step": 817 }, { "epoch": 0.07, "learning_rate": 9.958588642785108e-06, "loss": 0.88, "step": 818 }, { "epoch": 0.07, "learning_rate": 9.958410968003272e-06, "loss": 0.9686, "step": 819 }, { "epoch": 0.07, "learning_rate": 9.958232914473267e-06, "loss": 0.9231, "step": 820 }, { "epoch": 0.07, "learning_rate": 9.958054482208694e-06, "loss": 0.9272, "step": 821 }, { "epoch": 0.07, "learning_rate": 9.95787567122318e-06, "loss": 0.9449, "step": 822 }, { "epoch": 0.07, "learning_rate": 9.957696481530387e-06, "loss": 0.9724, "step": 823 }, { "epoch": 0.07, "learning_rate": 9.957516913144e-06, "loss": 0.9194, "step": 824 }, { "epoch": 0.07, "learning_rate": 9.957336966077736e-06, "loss": 0.9714, "step": 825 }, { "epoch": 0.07, "learning_rate": 9.95715664034534e-06, "loss": 0.9258, "step": 826 }, { "epoch": 0.07, "learning_rate": 9.956975935960588e-06, "loss": 0.9603, "step": 827 }, { "epoch": 0.07, "learning_rate": 9.956794852937281e-06, "loss": 0.9089, "step": 828 }, { "epoch": 0.07, "learning_rate": 9.956613391289253e-06, "loss": 0.9405, "step": 829 }, { "epoch": 0.07, "learning_rate": 9.956431551030363e-06, "loss": 0.9369, "step": 830 }, { "epoch": 0.07, "learning_rate": 9.956249332174501e-06, "loss": 0.8628, "step": 831 }, { "epoch": 0.07, "learning_rate": 9.956066734735587e-06, "loss": 0.9149, "step": 832 }, { "epoch": 0.07, "learning_rate": 9.955883758727566e-06, "loss": 0.8886, "step": 833 }, { "epoch": 0.07, "learning_rate": 9.955700404164419e-06, "loss": 1.0086, "step": 834 }, { "epoch": 0.07, "learning_rate": 9.955516671060148e-06, "loss": 0.9784, "step": 835 }, { "epoch": 0.07, "learning_rate": 9.95533255942879e-06, "loss": 0.8988, "step": 836 }, { "epoch": 0.07, "learning_rate": 9.955148069284407e-06, "loss": 0.6985, "step": 837 }, { "epoch": 0.07, "learning_rate": 9.95496320064109e-06, "loss": 0.9687, "step": 838 }, { "epoch": 0.07, "learning_rate": 9.954777953512961e-06, "loss": 0.9413, "step": 839 }, { "epoch": 0.07, "learning_rate": 9.954592327914173e-06, "loss": 0.9356, "step": 840 }, { "epoch": 0.07, "learning_rate": 9.954406323858903e-06, "loss": 0.9594, "step": 841 }, { "epoch": 0.07, "learning_rate": 9.954219941361356e-06, "loss": 0.9549, "step": 842 }, { "epoch": 0.07, "learning_rate": 9.954033180435772e-06, "loss": 0.8798, "step": 843 }, { "epoch": 0.07, "learning_rate": 9.953846041096416e-06, "loss": 0.9382, "step": 844 }, { "epoch": 0.07, "learning_rate": 9.953658523357583e-06, "loss": 0.944, "step": 845 }, { "epoch": 0.07, "learning_rate": 9.953470627233596e-06, "loss": 0.942, "step": 846 }, { "epoch": 0.07, "learning_rate": 9.953282352738808e-06, "loss": 0.9708, "step": 847 }, { "epoch": 0.07, "learning_rate": 9.953093699887599e-06, "loss": 0.9148, "step": 848 }, { "epoch": 0.07, "learning_rate": 9.952904668694382e-06, "loss": 0.9946, "step": 849 }, { "epoch": 0.07, "learning_rate": 9.952715259173593e-06, "loss": 0.9567, "step": 850 }, { "epoch": 0.07, "learning_rate": 9.952525471339703e-06, "loss": 0.9792, "step": 851 }, { "epoch": 0.07, "learning_rate": 9.952335305207206e-06, "loss": 0.948, "step": 852 }, { "epoch": 0.07, "learning_rate": 9.95214476079063e-06, "loss": 0.9492, "step": 853 }, { "epoch": 0.07, "learning_rate": 9.95195383810453e-06, "loss": 0.9317, "step": 854 }, { "epoch": 0.07, "learning_rate": 9.951762537163487e-06, "loss": 0.986, "step": 855 }, { "epoch": 0.07, "learning_rate": 9.951570857982116e-06, "loss": 0.9775, "step": 856 }, { "epoch": 0.07, "learning_rate": 9.951378800575056e-06, "loss": 0.9868, "step": 857 }, { "epoch": 0.07, "learning_rate": 9.95118636495698e-06, "loss": 0.9939, "step": 858 }, { "epoch": 0.07, "learning_rate": 9.950993551142586e-06, "loss": 0.6224, "step": 859 }, { "epoch": 0.07, "learning_rate": 9.950800359146603e-06, "loss": 0.9247, "step": 860 }, { "epoch": 0.07, "learning_rate": 9.950606788983786e-06, "loss": 0.9631, "step": 861 }, { "epoch": 0.07, "learning_rate": 9.950412840668921e-06, "loss": 0.9354, "step": 862 }, { "epoch": 0.07, "learning_rate": 9.950218514216825e-06, "loss": 0.9685, "step": 863 }, { "epoch": 0.07, "learning_rate": 9.950023809642342e-06, "loss": 0.9105, "step": 864 }, { "epoch": 0.07, "learning_rate": 9.94982872696034e-06, "loss": 0.9579, "step": 865 }, { "epoch": 0.07, "learning_rate": 9.949633266185724e-06, "loss": 0.6422, "step": 866 }, { "epoch": 0.07, "learning_rate": 9.949437427333424e-06, "loss": 0.9339, "step": 867 }, { "epoch": 0.07, "learning_rate": 9.949241210418398e-06, "loss": 0.9889, "step": 868 }, { "epoch": 0.07, "learning_rate": 9.949044615455635e-06, "loss": 0.8597, "step": 869 }, { "epoch": 0.07, "learning_rate": 9.948847642460153e-06, "loss": 0.9685, "step": 870 }, { "epoch": 0.07, "learning_rate": 9.948650291446993e-06, "loss": 0.6488, "step": 871 }, { "epoch": 0.07, "learning_rate": 9.948452562431237e-06, "loss": 0.9288, "step": 872 }, { "epoch": 0.07, "learning_rate": 9.948254455427981e-06, "loss": 0.975, "step": 873 }, { "epoch": 0.07, "learning_rate": 9.948055970452362e-06, "loss": 0.9501, "step": 874 }, { "epoch": 0.07, "learning_rate": 9.94785710751954e-06, "loss": 0.9125, "step": 875 }, { "epoch": 0.07, "learning_rate": 9.947657866644706e-06, "loss": 0.9209, "step": 876 }, { "epoch": 0.07, "learning_rate": 9.94745824784308e-06, "loss": 0.97, "step": 877 }, { "epoch": 0.07, "learning_rate": 9.947258251129906e-06, "loss": 0.9852, "step": 878 }, { "epoch": 0.08, "learning_rate": 9.947057876520464e-06, "loss": 0.9689, "step": 879 }, { "epoch": 0.08, "learning_rate": 9.946857124030059e-06, "loss": 0.9185, "step": 880 }, { "epoch": 0.08, "learning_rate": 9.946655993674023e-06, "loss": 0.946, "step": 881 }, { "epoch": 0.08, "learning_rate": 9.946454485467723e-06, "loss": 0.9204, "step": 882 }, { "epoch": 0.08, "learning_rate": 9.946252599426548e-06, "loss": 0.9305, "step": 883 }, { "epoch": 0.08, "learning_rate": 9.946050335565922e-06, "loss": 0.9428, "step": 884 }, { "epoch": 0.08, "learning_rate": 9.945847693901294e-06, "loss": 0.9444, "step": 885 }, { "epoch": 0.08, "learning_rate": 9.945644674448142e-06, "loss": 0.8825, "step": 886 }, { "epoch": 0.08, "learning_rate": 9.945441277221975e-06, "loss": 0.9526, "step": 887 }, { "epoch": 0.08, "learning_rate": 9.945237502238327e-06, "loss": 0.8756, "step": 888 }, { "epoch": 0.08, "learning_rate": 9.945033349512764e-06, "loss": 0.8648, "step": 889 }, { "epoch": 0.08, "learning_rate": 9.944828819060884e-06, "loss": 0.9378, "step": 890 }, { "epoch": 0.08, "learning_rate": 9.944623910898304e-06, "loss": 0.8754, "step": 891 }, { "epoch": 0.08, "learning_rate": 9.94441862504068e-06, "loss": 0.9472, "step": 892 }, { "epoch": 0.08, "learning_rate": 9.944212961503694e-06, "loss": 0.9706, "step": 893 }, { "epoch": 0.08, "learning_rate": 9.94400692030305e-06, "loss": 0.9301, "step": 894 }, { "epoch": 0.08, "learning_rate": 9.943800501454492e-06, "loss": 0.9832, "step": 895 }, { "epoch": 0.08, "learning_rate": 9.943593704973784e-06, "loss": 0.8927, "step": 896 }, { "epoch": 0.08, "learning_rate": 9.943386530876724e-06, "loss": 0.9193, "step": 897 }, { "epoch": 0.08, "learning_rate": 9.943178979179138e-06, "loss": 0.9287, "step": 898 }, { "epoch": 0.08, "learning_rate": 9.942971049896875e-06, "loss": 0.9593, "step": 899 }, { "epoch": 0.08, "learning_rate": 9.942762743045821e-06, "loss": 0.9437, "step": 900 }, { "epoch": 0.08, "learning_rate": 9.94255405864189e-06, "loss": 0.8887, "step": 901 }, { "epoch": 0.08, "learning_rate": 9.94234499670102e-06, "loss": 0.8868, "step": 902 }, { "epoch": 0.08, "learning_rate": 9.942135557239176e-06, "loss": 0.9069, "step": 903 }, { "epoch": 0.08, "learning_rate": 9.941925740272363e-06, "loss": 0.9423, "step": 904 }, { "epoch": 0.08, "learning_rate": 9.941715545816604e-06, "loss": 0.9371, "step": 905 }, { "epoch": 0.08, "learning_rate": 9.941504973887955e-06, "loss": 0.9687, "step": 906 }, { "epoch": 0.08, "learning_rate": 9.941294024502501e-06, "loss": 0.9419, "step": 907 }, { "epoch": 0.08, "learning_rate": 9.941082697676354e-06, "loss": 0.929, "step": 908 }, { "epoch": 0.08, "learning_rate": 9.94087099342566e-06, "loss": 0.9385, "step": 909 }, { "epoch": 0.08, "learning_rate": 9.940658911766588e-06, "loss": 0.9549, "step": 910 }, { "epoch": 0.08, "learning_rate": 9.940446452715336e-06, "loss": 0.906, "step": 911 }, { "epoch": 0.08, "learning_rate": 9.940233616288134e-06, "loss": 0.9041, "step": 912 }, { "epoch": 0.08, "learning_rate": 9.94002040250124e-06, "loss": 0.6345, "step": 913 }, { "epoch": 0.08, "learning_rate": 9.939806811370938e-06, "loss": 0.9078, "step": 914 }, { "epoch": 0.08, "learning_rate": 9.939592842913548e-06, "loss": 0.931, "step": 915 }, { "epoch": 0.08, "learning_rate": 9.93937849714541e-06, "loss": 0.9297, "step": 916 }, { "epoch": 0.08, "learning_rate": 9.939163774082899e-06, "loss": 0.9184, "step": 917 }, { "epoch": 0.08, "learning_rate": 9.938948673742414e-06, "loss": 0.9538, "step": 918 }, { "epoch": 0.08, "learning_rate": 9.938733196140386e-06, "loss": 0.8731, "step": 919 }, { "epoch": 0.08, "learning_rate": 9.938517341293277e-06, "loss": 0.938, "step": 920 }, { "epoch": 0.08, "learning_rate": 9.938301109217574e-06, "loss": 0.9461, "step": 921 }, { "epoch": 0.08, "learning_rate": 9.938084499929793e-06, "loss": 0.9115, "step": 922 }, { "epoch": 0.08, "learning_rate": 9.937867513446478e-06, "loss": 0.921, "step": 923 }, { "epoch": 0.08, "learning_rate": 9.937650149784207e-06, "loss": 0.9507, "step": 924 }, { "epoch": 0.08, "learning_rate": 9.93743240895958e-06, "loss": 0.9117, "step": 925 }, { "epoch": 0.08, "learning_rate": 9.937214290989231e-06, "loss": 0.9654, "step": 926 }, { "epoch": 0.08, "learning_rate": 9.936995795889823e-06, "loss": 0.9874, "step": 927 }, { "epoch": 0.08, "learning_rate": 9.936776923678044e-06, "loss": 0.6567, "step": 928 }, { "epoch": 0.08, "learning_rate": 9.936557674370609e-06, "loss": 0.9345, "step": 929 }, { "epoch": 0.08, "learning_rate": 9.936338047984271e-06, "loss": 0.9416, "step": 930 }, { "epoch": 0.08, "learning_rate": 9.9361180445358e-06, "loss": 0.9813, "step": 931 }, { "epoch": 0.08, "learning_rate": 9.935897664042009e-06, "loss": 0.9344, "step": 932 }, { "epoch": 0.08, "learning_rate": 9.935676906519725e-06, "loss": 0.9483, "step": 933 }, { "epoch": 0.08, "learning_rate": 9.935455771985813e-06, "loss": 0.9671, "step": 934 }, { "epoch": 0.08, "learning_rate": 9.935234260457164e-06, "loss": 0.8894, "step": 935 }, { "epoch": 0.08, "learning_rate": 9.935012371950699e-06, "loss": 0.9624, "step": 936 }, { "epoch": 0.08, "learning_rate": 9.934790106483366e-06, "loss": 0.8888, "step": 937 }, { "epoch": 0.08, "learning_rate": 9.934567464072143e-06, "loss": 0.9811, "step": 938 }, { "epoch": 0.08, "learning_rate": 9.934344444734037e-06, "loss": 0.9157, "step": 939 }, { "epoch": 0.08, "learning_rate": 9.934121048486083e-06, "loss": 0.9158, "step": 940 }, { "epoch": 0.08, "learning_rate": 9.933897275345344e-06, "loss": 0.9443, "step": 941 }, { "epoch": 0.08, "learning_rate": 9.933673125328915e-06, "loss": 0.95, "step": 942 }, { "epoch": 0.08, "learning_rate": 9.933448598453916e-06, "loss": 0.9309, "step": 943 }, { "epoch": 0.08, "learning_rate": 9.933223694737499e-06, "loss": 0.6285, "step": 944 }, { "epoch": 0.08, "learning_rate": 9.932998414196842e-06, "loss": 0.8865, "step": 945 }, { "epoch": 0.08, "learning_rate": 9.932772756849152e-06, "loss": 0.9252, "step": 946 }, { "epoch": 0.08, "learning_rate": 9.93254672271167e-06, "loss": 0.9211, "step": 947 }, { "epoch": 0.08, "learning_rate": 9.932320311801657e-06, "loss": 0.9554, "step": 948 }, { "epoch": 0.08, "learning_rate": 9.93209352413641e-06, "loss": 0.9376, "step": 949 }, { "epoch": 0.08, "learning_rate": 9.93186635973325e-06, "loss": 0.9345, "step": 950 }, { "epoch": 0.08, "learning_rate": 9.931638818609534e-06, "loss": 0.8731, "step": 951 }, { "epoch": 0.08, "learning_rate": 9.931410900782636e-06, "loss": 0.9432, "step": 952 }, { "epoch": 0.08, "learning_rate": 9.93118260626997e-06, "loss": 0.9253, "step": 953 }, { "epoch": 0.08, "learning_rate": 9.930953935088973e-06, "loss": 0.9567, "step": 954 }, { "epoch": 0.08, "learning_rate": 9.930724887257112e-06, "loss": 0.9195, "step": 955 }, { "epoch": 0.08, "learning_rate": 9.930495462791883e-06, "loss": 0.9754, "step": 956 }, { "epoch": 0.08, "learning_rate": 9.93026566171081e-06, "loss": 0.9302, "step": 957 }, { "epoch": 0.08, "learning_rate": 9.930035484031445e-06, "loss": 0.9096, "step": 958 }, { "epoch": 0.08, "learning_rate": 9.929804929771373e-06, "loss": 0.9583, "step": 959 }, { "epoch": 0.08, "learning_rate": 9.929573998948205e-06, "loss": 0.9199, "step": 960 }, { "epoch": 0.08, "learning_rate": 9.929342691579579e-06, "loss": 0.9254, "step": 961 }, { "epoch": 0.08, "learning_rate": 9.929111007683163e-06, "loss": 0.9009, "step": 962 }, { "epoch": 0.08, "learning_rate": 9.928878947276656e-06, "loss": 0.9697, "step": 963 }, { "epoch": 0.08, "learning_rate": 9.928646510377782e-06, "loss": 0.9312, "step": 964 }, { "epoch": 0.08, "learning_rate": 9.928413697004297e-06, "loss": 0.9259, "step": 965 }, { "epoch": 0.08, "learning_rate": 9.928180507173984e-06, "loss": 0.9329, "step": 966 }, { "epoch": 0.08, "learning_rate": 9.927946940904655e-06, "loss": 0.9077, "step": 967 }, { "epoch": 0.08, "learning_rate": 9.927712998214152e-06, "loss": 0.9208, "step": 968 }, { "epoch": 0.08, "learning_rate": 9.927478679120342e-06, "loss": 0.9262, "step": 969 }, { "epoch": 0.08, "learning_rate": 9.927243983641128e-06, "loss": 0.8868, "step": 970 }, { "epoch": 0.08, "learning_rate": 9.927008911794434e-06, "loss": 0.9116, "step": 971 }, { "epoch": 0.08, "learning_rate": 9.926773463598216e-06, "loss": 0.955, "step": 972 }, { "epoch": 0.08, "learning_rate": 9.926537639070457e-06, "loss": 0.9294, "step": 973 }, { "epoch": 0.08, "learning_rate": 9.926301438229176e-06, "loss": 0.9804, "step": 974 }, { "epoch": 0.08, "learning_rate": 9.92606486109241e-06, "loss": 0.9678, "step": 975 }, { "epoch": 0.08, "learning_rate": 9.925827907678234e-06, "loss": 0.9194, "step": 976 }, { "epoch": 0.08, "learning_rate": 9.925590578004742e-06, "loss": 0.9317, "step": 977 }, { "epoch": 0.08, "learning_rate": 9.92535287209007e-06, "loss": 0.9076, "step": 978 }, { "epoch": 0.08, "learning_rate": 9.925114789952369e-06, "loss": 0.9736, "step": 979 }, { "epoch": 0.08, "learning_rate": 9.924876331609828e-06, "loss": 0.8863, "step": 980 }, { "epoch": 0.08, "learning_rate": 9.924637497080658e-06, "loss": 0.9185, "step": 981 }, { "epoch": 0.08, "learning_rate": 9.924398286383107e-06, "loss": 0.9583, "step": 982 }, { "epoch": 0.08, "learning_rate": 9.924158699535445e-06, "loss": 0.9073, "step": 983 }, { "epoch": 0.08, "learning_rate": 9.923918736555972e-06, "loss": 0.8746, "step": 984 }, { "epoch": 0.08, "learning_rate": 9.923678397463022e-06, "loss": 0.9504, "step": 985 }, { "epoch": 0.08, "learning_rate": 9.923437682274948e-06, "loss": 0.8786, "step": 986 }, { "epoch": 0.08, "learning_rate": 9.923196591010138e-06, "loss": 0.9339, "step": 987 }, { "epoch": 0.08, "learning_rate": 9.922955123687009e-06, "loss": 0.933, "step": 988 }, { "epoch": 0.08, "learning_rate": 9.922713280324005e-06, "loss": 0.9166, "step": 989 }, { "epoch": 0.08, "learning_rate": 9.922471060939598e-06, "loss": 0.915, "step": 990 }, { "epoch": 0.08, "learning_rate": 9.922228465552294e-06, "loss": 0.9438, "step": 991 }, { "epoch": 0.08, "learning_rate": 9.921985494180619e-06, "loss": 0.9152, "step": 992 }, { "epoch": 0.08, "learning_rate": 9.921742146843134e-06, "loss": 0.9661, "step": 993 }, { "epoch": 0.08, "learning_rate": 9.921498423558428e-06, "loss": 0.9691, "step": 994 }, { "epoch": 0.08, "learning_rate": 9.92125432434512e-06, "loss": 0.9418, "step": 995 }, { "epoch": 0.08, "learning_rate": 9.921009849221848e-06, "loss": 0.9121, "step": 996 }, { "epoch": 0.09, "learning_rate": 9.920764998207293e-06, "loss": 0.9403, "step": 997 }, { "epoch": 0.09, "learning_rate": 9.920519771320156e-06, "loss": 0.8884, "step": 998 }, { "epoch": 0.09, "learning_rate": 9.920274168579168e-06, "loss": 0.9224, "step": 999 }, { "epoch": 0.09, "learning_rate": 9.92002819000309e-06, "loss": 0.9036, "step": 1000 }, { "epoch": 0.09, "learning_rate": 9.91978183561071e-06, "loss": 0.9463, "step": 1001 }, { "epoch": 0.09, "learning_rate": 9.91953510542085e-06, "loss": 0.8958, "step": 1002 }, { "epoch": 0.09, "learning_rate": 9.91928799945235e-06, "loss": 0.9422, "step": 1003 }, { "epoch": 0.09, "learning_rate": 9.919040517724088e-06, "loss": 0.9463, "step": 1004 }, { "epoch": 0.09, "learning_rate": 9.91879266025497e-06, "loss": 0.9334, "step": 1005 }, { "epoch": 0.09, "learning_rate": 9.918544427063926e-06, "loss": 0.9097, "step": 1006 }, { "epoch": 0.09, "learning_rate": 9.91829581816992e-06, "loss": 0.8754, "step": 1007 }, { "epoch": 0.09, "learning_rate": 9.918046833591937e-06, "loss": 0.6501, "step": 1008 }, { "epoch": 0.09, "learning_rate": 9.917797473349e-06, "loss": 0.9482, "step": 1009 }, { "epoch": 0.09, "learning_rate": 9.917547737460155e-06, "loss": 0.9424, "step": 1010 }, { "epoch": 0.09, "learning_rate": 9.917297625944479e-06, "loss": 0.9233, "step": 1011 }, { "epoch": 0.09, "learning_rate": 9.917047138821077e-06, "loss": 0.9429, "step": 1012 }, { "epoch": 0.09, "learning_rate": 9.91679627610908e-06, "loss": 0.9364, "step": 1013 }, { "epoch": 0.09, "learning_rate": 9.916545037827652e-06, "loss": 0.9175, "step": 1014 }, { "epoch": 0.09, "learning_rate": 9.916293423995982e-06, "loss": 0.8909, "step": 1015 }, { "epoch": 0.09, "learning_rate": 9.916041434633292e-06, "loss": 0.964, "step": 1016 }, { "epoch": 0.09, "learning_rate": 9.91578906975883e-06, "loss": 0.9403, "step": 1017 }, { "epoch": 0.09, "learning_rate": 9.915536329391872e-06, "loss": 0.9705, "step": 1018 }, { "epoch": 0.09, "learning_rate": 9.915283213551722e-06, "loss": 0.9509, "step": 1019 }, { "epoch": 0.09, "learning_rate": 9.915029722257718e-06, "loss": 0.9507, "step": 1020 }, { "epoch": 0.09, "learning_rate": 9.91477585552922e-06, "loss": 0.976, "step": 1021 }, { "epoch": 0.09, "learning_rate": 9.91452161338562e-06, "loss": 0.9661, "step": 1022 }, { "epoch": 0.09, "learning_rate": 9.914266995846337e-06, "loss": 0.9341, "step": 1023 }, { "epoch": 0.09, "learning_rate": 9.914012002930825e-06, "loss": 0.8608, "step": 1024 }, { "epoch": 0.09, "learning_rate": 9.913756634658556e-06, "loss": 0.8993, "step": 1025 }, { "epoch": 0.09, "learning_rate": 9.91350089104904e-06, "loss": 0.9451, "step": 1026 }, { "epoch": 0.09, "learning_rate": 9.913244772121811e-06, "loss": 0.9432, "step": 1027 }, { "epoch": 0.09, "learning_rate": 9.912988277896431e-06, "loss": 0.9213, "step": 1028 }, { "epoch": 0.09, "learning_rate": 9.912731408392493e-06, "loss": 0.9215, "step": 1029 }, { "epoch": 0.09, "learning_rate": 9.912474163629619e-06, "loss": 0.9249, "step": 1030 }, { "epoch": 0.09, "learning_rate": 9.91221654362746e-06, "loss": 0.9207, "step": 1031 }, { "epoch": 0.09, "learning_rate": 9.91195854840569e-06, "loss": 0.9175, "step": 1032 }, { "epoch": 0.09, "learning_rate": 9.91170017798402e-06, "loss": 0.9593, "step": 1033 }, { "epoch": 0.09, "learning_rate": 9.911441432382183e-06, "loss": 0.9443, "step": 1034 }, { "epoch": 0.09, "learning_rate": 9.911182311619945e-06, "loss": 0.9694, "step": 1035 }, { "epoch": 0.09, "learning_rate": 9.910922815717096e-06, "loss": 0.8907, "step": 1036 }, { "epoch": 0.09, "learning_rate": 9.910662944693463e-06, "loss": 0.878, "step": 1037 }, { "epoch": 0.09, "learning_rate": 9.910402698568891e-06, "loss": 0.928, "step": 1038 }, { "epoch": 0.09, "learning_rate": 9.910142077363263e-06, "loss": 0.8938, "step": 1039 }, { "epoch": 0.09, "learning_rate": 9.909881081096482e-06, "loss": 0.9115, "step": 1040 }, { "epoch": 0.09, "learning_rate": 9.90961970978849e-06, "loss": 0.8509, "step": 1041 }, { "epoch": 0.09, "learning_rate": 9.909357963459245e-06, "loss": 0.8912, "step": 1042 }, { "epoch": 0.09, "learning_rate": 9.909095842128746e-06, "loss": 0.9492, "step": 1043 }, { "epoch": 0.09, "learning_rate": 9.908833345817012e-06, "loss": 0.882, "step": 1044 }, { "epoch": 0.09, "learning_rate": 9.908570474544096e-06, "loss": 0.9193, "step": 1045 }, { "epoch": 0.09, "learning_rate": 9.908307228330076e-06, "loss": 0.9048, "step": 1046 }, { "epoch": 0.09, "learning_rate": 9.908043607195062e-06, "loss": 0.9609, "step": 1047 }, { "epoch": 0.09, "learning_rate": 9.907779611159187e-06, "loss": 0.9575, "step": 1048 }, { "epoch": 0.09, "learning_rate": 9.907515240242619e-06, "loss": 0.871, "step": 1049 }, { "epoch": 0.09, "learning_rate": 9.907250494465552e-06, "loss": 0.9185, "step": 1050 }, { "epoch": 0.09, "learning_rate": 9.906985373848207e-06, "loss": 0.9158, "step": 1051 }, { "epoch": 0.09, "learning_rate": 9.906719878410837e-06, "loss": 0.9836, "step": 1052 }, { "epoch": 0.09, "learning_rate": 9.906454008173722e-06, "loss": 0.9029, "step": 1053 }, { "epoch": 0.09, "learning_rate": 9.90618776315717e-06, "loss": 0.9228, "step": 1054 }, { "epoch": 0.09, "learning_rate": 9.905921143381514e-06, "loss": 0.6868, "step": 1055 }, { "epoch": 0.09, "learning_rate": 9.905654148867128e-06, "loss": 0.9038, "step": 1056 }, { "epoch": 0.09, "learning_rate": 9.905386779634399e-06, "loss": 0.9262, "step": 1057 }, { "epoch": 0.09, "learning_rate": 9.905119035703755e-06, "loss": 0.8682, "step": 1058 }, { "epoch": 0.09, "learning_rate": 9.904850917095642e-06, "loss": 0.9504, "step": 1059 }, { "epoch": 0.09, "learning_rate": 9.904582423830546e-06, "loss": 0.9742, "step": 1060 }, { "epoch": 0.09, "learning_rate": 9.904313555928975e-06, "loss": 0.9068, "step": 1061 }, { "epoch": 0.09, "learning_rate": 9.904044313411463e-06, "loss": 0.9138, "step": 1062 }, { "epoch": 0.09, "learning_rate": 9.90377469629858e-06, "loss": 0.8725, "step": 1063 }, { "epoch": 0.09, "learning_rate": 9.903504704610919e-06, "loss": 0.9418, "step": 1064 }, { "epoch": 0.09, "learning_rate": 9.9032343383691e-06, "loss": 0.9316, "step": 1065 }, { "epoch": 0.09, "learning_rate": 9.90296359759378e-06, "loss": 0.913, "step": 1066 }, { "epoch": 0.09, "learning_rate": 9.902692482305638e-06, "loss": 0.9543, "step": 1067 }, { "epoch": 0.09, "learning_rate": 9.902420992525383e-06, "loss": 0.9243, "step": 1068 }, { "epoch": 0.09, "learning_rate": 9.90214912827375e-06, "loss": 0.9157, "step": 1069 }, { "epoch": 0.09, "learning_rate": 9.901876889571511e-06, "loss": 0.9111, "step": 1070 }, { "epoch": 0.09, "learning_rate": 9.901604276439456e-06, "loss": 0.9514, "step": 1071 }, { "epoch": 0.09, "learning_rate": 9.901331288898411e-06, "loss": 0.904, "step": 1072 }, { "epoch": 0.09, "learning_rate": 9.901057926969227e-06, "loss": 0.903, "step": 1073 }, { "epoch": 0.09, "learning_rate": 9.900784190672786e-06, "loss": 0.9075, "step": 1074 }, { "epoch": 0.09, "learning_rate": 9.900510080029997e-06, "loss": 0.9085, "step": 1075 }, { "epoch": 0.09, "learning_rate": 9.900235595061795e-06, "loss": 0.909, "step": 1076 }, { "epoch": 0.09, "learning_rate": 9.899960735789153e-06, "loss": 0.9454, "step": 1077 }, { "epoch": 0.09, "learning_rate": 9.89968550223306e-06, "loss": 0.9037, "step": 1078 }, { "epoch": 0.09, "learning_rate": 9.89940989441454e-06, "loss": 0.8967, "step": 1079 }, { "epoch": 0.09, "learning_rate": 9.899133912354651e-06, "loss": 0.9702, "step": 1080 }, { "epoch": 0.09, "learning_rate": 9.898857556074469e-06, "loss": 0.9346, "step": 1081 }, { "epoch": 0.09, "learning_rate": 9.898580825595103e-06, "loss": 0.9136, "step": 1082 }, { "epoch": 0.09, "learning_rate": 9.898303720937693e-06, "loss": 0.9466, "step": 1083 }, { "epoch": 0.09, "learning_rate": 9.898026242123406e-06, "loss": 0.9792, "step": 1084 }, { "epoch": 0.09, "learning_rate": 9.897748389173436e-06, "loss": 0.8929, "step": 1085 }, { "epoch": 0.09, "learning_rate": 9.897470162109007e-06, "loss": 0.9757, "step": 1086 }, { "epoch": 0.09, "learning_rate": 9.897191560951372e-06, "loss": 0.955, "step": 1087 }, { "epoch": 0.09, "learning_rate": 9.896912585721812e-06, "loss": 0.9068, "step": 1088 }, { "epoch": 0.09, "learning_rate": 9.896633236441635e-06, "loss": 0.932, "step": 1089 }, { "epoch": 0.09, "learning_rate": 9.896353513132182e-06, "loss": 0.9281, "step": 1090 }, { "epoch": 0.09, "learning_rate": 9.896073415814816e-06, "loss": 0.8733, "step": 1091 }, { "epoch": 0.09, "learning_rate": 9.895792944510936e-06, "loss": 0.8972, "step": 1092 }, { "epoch": 0.09, "learning_rate": 9.895512099241961e-06, "loss": 0.6306, "step": 1093 }, { "epoch": 0.09, "learning_rate": 9.895230880029348e-06, "loss": 0.9372, "step": 1094 }, { "epoch": 0.09, "learning_rate": 9.894949286894577e-06, "loss": 0.6256, "step": 1095 }, { "epoch": 0.09, "learning_rate": 9.894667319859155e-06, "loss": 0.9119, "step": 1096 }, { "epoch": 0.09, "learning_rate": 9.894384978944623e-06, "loss": 0.8968, "step": 1097 }, { "epoch": 0.09, "learning_rate": 9.894102264172545e-06, "loss": 0.9157, "step": 1098 }, { "epoch": 0.09, "learning_rate": 9.89381917556452e-06, "loss": 0.9001, "step": 1099 }, { "epoch": 0.09, "learning_rate": 9.893535713142167e-06, "loss": 0.9149, "step": 1100 }, { "epoch": 0.09, "learning_rate": 9.893251876927142e-06, "loss": 0.9123, "step": 1101 }, { "epoch": 0.09, "learning_rate": 9.892967666941122e-06, "loss": 0.9208, "step": 1102 }, { "epoch": 0.09, "learning_rate": 9.892683083205821e-06, "loss": 0.9247, "step": 1103 }, { "epoch": 0.09, "learning_rate": 9.892398125742973e-06, "loss": 0.9354, "step": 1104 }, { "epoch": 0.09, "learning_rate": 9.892112794574346e-06, "loss": 0.9707, "step": 1105 }, { "epoch": 0.09, "learning_rate": 9.891827089721736e-06, "loss": 0.9274, "step": 1106 }, { "epoch": 0.09, "learning_rate": 9.891541011206965e-06, "loss": 0.6676, "step": 1107 }, { "epoch": 0.09, "learning_rate": 9.891254559051886e-06, "loss": 0.9494, "step": 1108 }, { "epoch": 0.09, "learning_rate": 9.890967733278381e-06, "loss": 0.9228, "step": 1109 }, { "epoch": 0.09, "learning_rate": 9.890680533908355e-06, "loss": 0.9143, "step": 1110 }, { "epoch": 0.09, "learning_rate": 9.89039296096375e-06, "loss": 0.9408, "step": 1111 }, { "epoch": 0.09, "learning_rate": 9.890105014466529e-06, "loss": 0.956, "step": 1112 }, { "epoch": 0.09, "learning_rate": 9.88981669443869e-06, "loss": 0.9099, "step": 1113 }, { "epoch": 0.1, "learning_rate": 9.889528000902255e-06, "loss": 0.9372, "step": 1114 }, { "epoch": 0.1, "learning_rate": 9.889238933879273e-06, "loss": 0.9407, "step": 1115 }, { "epoch": 0.1, "learning_rate": 9.88894949339183e-06, "loss": 0.9039, "step": 1116 }, { "epoch": 0.1, "learning_rate": 9.88865967946203e-06, "loss": 0.8988, "step": 1117 }, { "epoch": 0.1, "learning_rate": 9.888369492112012e-06, "loss": 0.9092, "step": 1118 }, { "epoch": 0.1, "learning_rate": 9.888078931363943e-06, "loss": 0.9538, "step": 1119 }, { "epoch": 0.1, "learning_rate": 9.887787997240017e-06, "loss": 0.9226, "step": 1120 }, { "epoch": 0.1, "learning_rate": 9.887496689762455e-06, "loss": 0.5984, "step": 1121 }, { "epoch": 0.1, "learning_rate": 9.887205008953512e-06, "loss": 0.9321, "step": 1122 }, { "epoch": 0.1, "learning_rate": 9.886912954835465e-06, "loss": 0.914, "step": 1123 }, { "epoch": 0.1, "learning_rate": 9.886620527430626e-06, "loss": 0.993, "step": 1124 }, { "epoch": 0.1, "learning_rate": 9.886327726761327e-06, "loss": 0.9507, "step": 1125 }, { "epoch": 0.1, "learning_rate": 9.886034552849939e-06, "loss": 0.9496, "step": 1126 }, { "epoch": 0.1, "learning_rate": 9.885741005718851e-06, "loss": 0.9199, "step": 1127 }, { "epoch": 0.1, "learning_rate": 9.88544708539049e-06, "loss": 0.9121, "step": 1128 }, { "epoch": 0.1, "learning_rate": 9.885152791887304e-06, "loss": 0.9134, "step": 1129 }, { "epoch": 0.1, "learning_rate": 9.884858125231776e-06, "loss": 0.9103, "step": 1130 }, { "epoch": 0.1, "learning_rate": 9.884563085446411e-06, "loss": 0.9514, "step": 1131 }, { "epoch": 0.1, "learning_rate": 9.884267672553744e-06, "loss": 0.9237, "step": 1132 }, { "epoch": 0.1, "learning_rate": 9.883971886576345e-06, "loss": 0.9418, "step": 1133 }, { "epoch": 0.1, "learning_rate": 9.883675727536805e-06, "loss": 0.9804, "step": 1134 }, { "epoch": 0.1, "learning_rate": 9.883379195457747e-06, "loss": 0.8809, "step": 1135 }, { "epoch": 0.1, "learning_rate": 9.88308229036182e-06, "loss": 0.9283, "step": 1136 }, { "epoch": 0.1, "learning_rate": 9.882785012271704e-06, "loss": 0.88, "step": 1137 }, { "epoch": 0.1, "learning_rate": 9.882487361210107e-06, "loss": 0.9601, "step": 1138 }, { "epoch": 0.1, "learning_rate": 9.882189337199764e-06, "loss": 0.9175, "step": 1139 }, { "epoch": 0.1, "learning_rate": 9.881890940263442e-06, "loss": 0.956, "step": 1140 }, { "epoch": 0.1, "learning_rate": 9.88159217042393e-06, "loss": 0.8897, "step": 1141 }, { "epoch": 0.1, "learning_rate": 9.881293027704053e-06, "loss": 0.8844, "step": 1142 }, { "epoch": 0.1, "learning_rate": 9.88099351212666e-06, "loss": 0.8785, "step": 1143 }, { "epoch": 0.1, "learning_rate": 9.88069362371463e-06, "loss": 0.9183, "step": 1144 }, { "epoch": 0.1, "learning_rate": 9.880393362490867e-06, "loss": 0.9215, "step": 1145 }, { "epoch": 0.1, "learning_rate": 9.880092728478311e-06, "loss": 0.8984, "step": 1146 }, { "epoch": 0.1, "learning_rate": 9.879791721699923e-06, "loss": 0.8516, "step": 1147 }, { "epoch": 0.1, "learning_rate": 9.879490342178697e-06, "loss": 0.9369, "step": 1148 }, { "epoch": 0.1, "learning_rate": 9.879188589937652e-06, "loss": 0.9539, "step": 1149 }, { "epoch": 0.1, "learning_rate": 9.878886464999839e-06, "loss": 0.9521, "step": 1150 }, { "epoch": 0.1, "learning_rate": 9.878583967388336e-06, "loss": 0.9129, "step": 1151 }, { "epoch": 0.1, "learning_rate": 9.878281097126245e-06, "loss": 0.9547, "step": 1152 }, { "epoch": 0.1, "learning_rate": 9.877977854236708e-06, "loss": 0.6143, "step": 1153 }, { "epoch": 0.1, "learning_rate": 9.877674238742883e-06, "loss": 0.9387, "step": 1154 }, { "epoch": 0.1, "learning_rate": 9.877370250667962e-06, "loss": 0.9692, "step": 1155 }, { "epoch": 0.1, "learning_rate": 9.877065890035167e-06, "loss": 0.9139, "step": 1156 }, { "epoch": 0.1, "learning_rate": 9.876761156867746e-06, "loss": 0.862, "step": 1157 }, { "epoch": 0.1, "learning_rate": 9.876456051188975e-06, "loss": 0.9439, "step": 1158 }, { "epoch": 0.1, "learning_rate": 9.876150573022162e-06, "loss": 0.9845, "step": 1159 }, { "epoch": 0.1, "learning_rate": 9.875844722390636e-06, "loss": 0.9695, "step": 1160 }, { "epoch": 0.1, "learning_rate": 9.875538499317765e-06, "loss": 0.9566, "step": 1161 }, { "epoch": 0.1, "learning_rate": 9.875231903826936e-06, "loss": 0.9122, "step": 1162 }, { "epoch": 0.1, "learning_rate": 9.874924935941572e-06, "loss": 0.9251, "step": 1163 }, { "epoch": 0.1, "learning_rate": 9.874617595685116e-06, "loss": 0.8963, "step": 1164 }, { "epoch": 0.1, "learning_rate": 9.874309883081048e-06, "loss": 0.9234, "step": 1165 }, { "epoch": 0.1, "learning_rate": 9.874001798152873e-06, "loss": 0.9462, "step": 1166 }, { "epoch": 0.1, "learning_rate": 9.87369334092412e-06, "loss": 0.8969, "step": 1167 }, { "epoch": 0.1, "learning_rate": 9.873384511418353e-06, "loss": 0.9251, "step": 1168 }, { "epoch": 0.1, "learning_rate": 9.873075309659162e-06, "loss": 0.9297, "step": 1169 }, { "epoch": 0.1, "learning_rate": 9.872765735670165e-06, "loss": 0.9151, "step": 1170 }, { "epoch": 0.1, "learning_rate": 9.872455789475011e-06, "loss": 0.9377, "step": 1171 }, { "epoch": 0.1, "learning_rate": 9.87214547109737e-06, "loss": 0.8868, "step": 1172 }, { "epoch": 0.1, "learning_rate": 9.871834780560951e-06, "loss": 0.9444, "step": 1173 }, { "epoch": 0.1, "learning_rate": 9.871523717889484e-06, "loss": 0.948, "step": 1174 }, { "epoch": 0.1, "learning_rate": 9.871212283106728e-06, "loss": 0.9556, "step": 1175 }, { "epoch": 0.1, "learning_rate": 9.870900476236476e-06, "loss": 0.9061, "step": 1176 }, { "epoch": 0.1, "learning_rate": 9.870588297302541e-06, "loss": 0.8554, "step": 1177 }, { "epoch": 0.1, "learning_rate": 9.87027574632877e-06, "loss": 0.8805, "step": 1178 }, { "epoch": 0.1, "learning_rate": 9.869962823339038e-06, "loss": 0.922, "step": 1179 }, { "epoch": 0.1, "learning_rate": 9.869649528357248e-06, "loss": 0.897, "step": 1180 }, { "epoch": 0.1, "learning_rate": 9.869335861407331e-06, "loss": 0.597, "step": 1181 }, { "epoch": 0.1, "learning_rate": 9.869021822513245e-06, "loss": 0.9932, "step": 1182 }, { "epoch": 0.1, "learning_rate": 9.86870741169898e-06, "loss": 0.9467, "step": 1183 }, { "epoch": 0.1, "learning_rate": 9.86839262898855e-06, "loss": 0.953, "step": 1184 }, { "epoch": 0.1, "learning_rate": 9.868077474406001e-06, "loss": 0.8945, "step": 1185 }, { "epoch": 0.1, "learning_rate": 9.867761947975406e-06, "loss": 0.9253, "step": 1186 }, { "epoch": 0.1, "learning_rate": 9.867446049720865e-06, "loss": 0.9031, "step": 1187 }, { "epoch": 0.1, "learning_rate": 9.867129779666511e-06, "loss": 0.7126, "step": 1188 }, { "epoch": 0.1, "learning_rate": 9.8668131378365e-06, "loss": 0.8995, "step": 1189 }, { "epoch": 0.1, "learning_rate": 9.866496124255017e-06, "loss": 0.5865, "step": 1190 }, { "epoch": 0.1, "learning_rate": 9.866178738946282e-06, "loss": 0.8914, "step": 1191 }, { "epoch": 0.1, "learning_rate": 9.865860981934534e-06, "loss": 0.5844, "step": 1192 }, { "epoch": 0.1, "learning_rate": 9.865542853244046e-06, "loss": 0.9418, "step": 1193 }, { "epoch": 0.1, "learning_rate": 9.86522435289912e-06, "loss": 0.8782, "step": 1194 }, { "epoch": 0.1, "learning_rate": 9.864905480924082e-06, "loss": 0.9252, "step": 1195 }, { "epoch": 0.1, "learning_rate": 9.864586237343293e-06, "loss": 0.9456, "step": 1196 }, { "epoch": 0.1, "learning_rate": 9.864266622181133e-06, "loss": 0.9733, "step": 1197 }, { "epoch": 0.1, "learning_rate": 9.863946635462021e-06, "loss": 0.94, "step": 1198 }, { "epoch": 0.1, "learning_rate": 9.863626277210396e-06, "loss": 0.8669, "step": 1199 }, { "epoch": 0.1, "learning_rate": 9.86330554745073e-06, "loss": 0.6738, "step": 1200 }, { "epoch": 0.1, "learning_rate": 9.86298444620752e-06, "loss": 0.9163, "step": 1201 }, { "epoch": 0.1, "learning_rate": 9.862662973505296e-06, "loss": 0.8507, "step": 1202 }, { "epoch": 0.1, "learning_rate": 9.862341129368612e-06, "loss": 0.9289, "step": 1203 }, { "epoch": 0.1, "learning_rate": 9.862018913822051e-06, "loss": 0.9191, "step": 1204 }, { "epoch": 0.1, "learning_rate": 9.86169632689023e-06, "loss": 0.9397, "step": 1205 }, { "epoch": 0.1, "learning_rate": 9.861373368597784e-06, "loss": 0.9342, "step": 1206 }, { "epoch": 0.1, "learning_rate": 9.861050038969385e-06, "loss": 0.948, "step": 1207 }, { "epoch": 0.1, "learning_rate": 9.86072633802973e-06, "loss": 0.9069, "step": 1208 }, { "epoch": 0.1, "learning_rate": 9.860402265803549e-06, "loss": 0.9127, "step": 1209 }, { "epoch": 0.1, "learning_rate": 9.860077822315588e-06, "loss": 0.9872, "step": 1210 }, { "epoch": 0.1, "learning_rate": 9.859753007590636e-06, "loss": 0.8759, "step": 1211 }, { "epoch": 0.1, "learning_rate": 9.859427821653502e-06, "loss": 0.908, "step": 1212 }, { "epoch": 0.1, "learning_rate": 9.859102264529025e-06, "loss": 0.8704, "step": 1213 }, { "epoch": 0.1, "learning_rate": 9.858776336242074e-06, "loss": 0.9343, "step": 1214 }, { "epoch": 0.1, "learning_rate": 9.858450036817543e-06, "loss": 0.8802, "step": 1215 }, { "epoch": 0.1, "learning_rate": 9.858123366280358e-06, "loss": 0.9085, "step": 1216 }, { "epoch": 0.1, "learning_rate": 9.85779632465547e-06, "loss": 0.9176, "step": 1217 }, { "epoch": 0.1, "learning_rate": 9.857468911967863e-06, "loss": 0.8778, "step": 1218 }, { "epoch": 0.1, "learning_rate": 9.857141128242545e-06, "loss": 0.9344, "step": 1219 }, { "epoch": 0.1, "learning_rate": 9.856812973504553e-06, "loss": 0.9609, "step": 1220 }, { "epoch": 0.1, "learning_rate": 9.856484447778952e-06, "loss": 0.8861, "step": 1221 }, { "epoch": 0.1, "learning_rate": 9.856155551090841e-06, "loss": 0.9533, "step": 1222 }, { "epoch": 0.1, "learning_rate": 9.855826283465336e-06, "loss": 0.9895, "step": 1223 }, { "epoch": 0.1, "learning_rate": 9.855496644927594e-06, "loss": 0.9372, "step": 1224 }, { "epoch": 0.1, "learning_rate": 9.855166635502793e-06, "loss": 0.9377, "step": 1225 }, { "epoch": 0.1, "learning_rate": 9.854836255216137e-06, "loss": 0.9614, "step": 1226 }, { "epoch": 0.1, "learning_rate": 9.854505504092869e-06, "loss": 0.94, "step": 1227 }, { "epoch": 0.1, "learning_rate": 9.854174382158249e-06, "loss": 0.9429, "step": 1228 }, { "epoch": 0.1, "learning_rate": 9.853842889437568e-06, "loss": 0.9523, "step": 1229 }, { "epoch": 0.1, "learning_rate": 9.85351102595615e-06, "loss": 0.8399, "step": 1230 }, { "epoch": 0.11, "learning_rate": 9.853178791739345e-06, "loss": 0.9046, "step": 1231 }, { "epoch": 0.11, "learning_rate": 9.852846186812528e-06, "loss": 0.9474, "step": 1232 }, { "epoch": 0.11, "learning_rate": 9.852513211201108e-06, "loss": 0.9441, "step": 1233 }, { "epoch": 0.11, "learning_rate": 9.852179864930517e-06, "loss": 0.9006, "step": 1234 }, { "epoch": 0.11, "learning_rate": 9.851846148026218e-06, "loss": 0.9833, "step": 1235 }, { "epoch": 0.11, "learning_rate": 9.851512060513704e-06, "loss": 0.9585, "step": 1236 }, { "epoch": 0.11, "learning_rate": 9.851177602418492e-06, "loss": 0.8708, "step": 1237 }, { "epoch": 0.11, "learning_rate": 9.850842773766131e-06, "loss": 0.9091, "step": 1238 }, { "epoch": 0.11, "learning_rate": 9.850507574582195e-06, "loss": 0.9394, "step": 1239 }, { "epoch": 0.11, "learning_rate": 9.85017200489229e-06, "loss": 0.8775, "step": 1240 }, { "epoch": 0.11, "learning_rate": 9.849836064722047e-06, "loss": 0.9172, "step": 1241 }, { "epoch": 0.11, "learning_rate": 9.84949975409713e-06, "loss": 0.9182, "step": 1242 }, { "epoch": 0.11, "learning_rate": 9.849163073043223e-06, "loss": 0.9425, "step": 1243 }, { "epoch": 0.11, "learning_rate": 9.848826021586046e-06, "loss": 0.9608, "step": 1244 }, { "epoch": 0.11, "learning_rate": 9.848488599751348e-06, "loss": 0.9102, "step": 1245 }, { "epoch": 0.11, "learning_rate": 9.848150807564896e-06, "loss": 0.8655, "step": 1246 }, { "epoch": 0.11, "learning_rate": 9.8478126450525e-06, "loss": 0.9307, "step": 1247 }, { "epoch": 0.11, "learning_rate": 9.847474112239982e-06, "loss": 0.9291, "step": 1248 }, { "epoch": 0.11, "learning_rate": 9.847135209153209e-06, "loss": 0.9131, "step": 1249 }, { "epoch": 0.11, "learning_rate": 9.846795935818063e-06, "loss": 0.9855, "step": 1250 }, { "epoch": 0.11, "learning_rate": 9.846456292260461e-06, "loss": 0.9527, "step": 1251 }, { "epoch": 0.11, "learning_rate": 9.846116278506347e-06, "loss": 0.9208, "step": 1252 }, { "epoch": 0.11, "learning_rate": 9.845775894581692e-06, "loss": 0.9242, "step": 1253 }, { "epoch": 0.11, "learning_rate": 9.845435140512496e-06, "loss": 0.9445, "step": 1254 }, { "epoch": 0.11, "learning_rate": 9.84509401632479e-06, "loss": 0.9299, "step": 1255 }, { "epoch": 0.11, "learning_rate": 9.844752522044626e-06, "loss": 0.9281, "step": 1256 }, { "epoch": 0.11, "learning_rate": 9.844410657698096e-06, "loss": 0.8853, "step": 1257 }, { "epoch": 0.11, "learning_rate": 9.844068423311306e-06, "loss": 0.9359, "step": 1258 }, { "epoch": 0.11, "learning_rate": 9.843725818910402e-06, "loss": 0.9706, "step": 1259 }, { "epoch": 0.11, "learning_rate": 9.843382844521554e-06, "loss": 0.9379, "step": 1260 }, { "epoch": 0.11, "learning_rate": 9.843039500170957e-06, "loss": 0.9023, "step": 1261 }, { "epoch": 0.11, "learning_rate": 9.842695785884839e-06, "loss": 0.8731, "step": 1262 }, { "epoch": 0.11, "learning_rate": 9.842351701689456e-06, "loss": 0.9051, "step": 1263 }, { "epoch": 0.11, "learning_rate": 9.842007247611089e-06, "loss": 0.9262, "step": 1264 }, { "epoch": 0.11, "learning_rate": 9.84166242367605e-06, "loss": 0.9427, "step": 1265 }, { "epoch": 0.11, "learning_rate": 9.841317229910678e-06, "loss": 0.9204, "step": 1266 }, { "epoch": 0.11, "learning_rate": 9.840971666341341e-06, "loss": 0.8827, "step": 1267 }, { "epoch": 0.11, "learning_rate": 9.840625732994435e-06, "loss": 0.9426, "step": 1268 }, { "epoch": 0.11, "learning_rate": 9.840279429896384e-06, "loss": 0.9042, "step": 1269 }, { "epoch": 0.11, "learning_rate": 9.83993275707364e-06, "loss": 0.9367, "step": 1270 }, { "epoch": 0.11, "learning_rate": 9.839585714552682e-06, "loss": 0.9462, "step": 1271 }, { "epoch": 0.11, "learning_rate": 9.839238302360023e-06, "loss": 0.8845, "step": 1272 }, { "epoch": 0.11, "learning_rate": 9.838890520522194e-06, "loss": 0.9334, "step": 1273 }, { "epoch": 0.11, "learning_rate": 9.838542369065768e-06, "loss": 0.9497, "step": 1274 }, { "epoch": 0.11, "learning_rate": 9.838193848017333e-06, "loss": 0.9511, "step": 1275 }, { "epoch": 0.11, "learning_rate": 9.837844957403512e-06, "loss": 0.9551, "step": 1276 }, { "epoch": 0.11, "learning_rate": 9.837495697250956e-06, "loss": 0.8969, "step": 1277 }, { "epoch": 0.11, "learning_rate": 9.837146067586341e-06, "loss": 0.9575, "step": 1278 }, { "epoch": 0.11, "learning_rate": 9.836796068436375e-06, "loss": 0.928, "step": 1279 }, { "epoch": 0.11, "learning_rate": 9.836445699827794e-06, "loss": 0.6926, "step": 1280 }, { "epoch": 0.11, "learning_rate": 9.836094961787359e-06, "loss": 0.9234, "step": 1281 }, { "epoch": 0.11, "learning_rate": 9.835743854341861e-06, "loss": 0.9551, "step": 1282 }, { "epoch": 0.11, "learning_rate": 9.83539237751812e-06, "loss": 0.5672, "step": 1283 }, { "epoch": 0.11, "learning_rate": 9.835040531342986e-06, "loss": 0.9583, "step": 1284 }, { "epoch": 0.11, "learning_rate": 9.834688315843329e-06, "loss": 0.8807, "step": 1285 }, { "epoch": 0.11, "learning_rate": 9.834335731046057e-06, "loss": 0.9236, "step": 1286 }, { "epoch": 0.11, "learning_rate": 9.833982776978103e-06, "loss": 0.9127, "step": 1287 }, { "epoch": 0.11, "learning_rate": 9.833629453666425e-06, "loss": 0.9242, "step": 1288 }, { "epoch": 0.11, "learning_rate": 9.833275761138012e-06, "loss": 0.9531, "step": 1289 }, { "epoch": 0.11, "learning_rate": 9.832921699419881e-06, "loss": 0.9219, "step": 1290 }, { "epoch": 0.11, "learning_rate": 9.83256726853908e-06, "loss": 0.9976, "step": 1291 }, { "epoch": 0.11, "learning_rate": 9.832212468522676e-06, "loss": 0.9161, "step": 1292 }, { "epoch": 0.11, "learning_rate": 9.831857299397775e-06, "loss": 0.9558, "step": 1293 }, { "epoch": 0.11, "learning_rate": 9.831501761191506e-06, "loss": 0.9351, "step": 1294 }, { "epoch": 0.11, "learning_rate": 9.831145853931027e-06, "loss": 0.9432, "step": 1295 }, { "epoch": 0.11, "learning_rate": 9.830789577643523e-06, "loss": 0.9282, "step": 1296 }, { "epoch": 0.11, "learning_rate": 9.830432932356207e-06, "loss": 0.908, "step": 1297 }, { "epoch": 0.11, "learning_rate": 9.830075918096324e-06, "loss": 0.6344, "step": 1298 }, { "epoch": 0.11, "learning_rate": 9.829718534891141e-06, "loss": 0.95, "step": 1299 }, { "epoch": 0.11, "learning_rate": 9.829360782767961e-06, "loss": 0.6167, "step": 1300 }, { "epoch": 0.11, "learning_rate": 9.82900266175411e-06, "loss": 0.9423, "step": 1301 }, { "epoch": 0.11, "learning_rate": 9.82864417187694e-06, "loss": 0.8981, "step": 1302 }, { "epoch": 0.11, "learning_rate": 9.828285313163835e-06, "loss": 0.903, "step": 1303 }, { "epoch": 0.11, "learning_rate": 9.82792608564221e-06, "loss": 0.9234, "step": 1304 }, { "epoch": 0.11, "learning_rate": 9.827566489339501e-06, "loss": 0.874, "step": 1305 }, { "epoch": 0.11, "learning_rate": 9.827206524283177e-06, "loss": 0.8913, "step": 1306 }, { "epoch": 0.11, "learning_rate": 9.826846190500733e-06, "loss": 0.9461, "step": 1307 }, { "epoch": 0.11, "learning_rate": 9.826485488019697e-06, "loss": 0.9238, "step": 1308 }, { "epoch": 0.11, "learning_rate": 9.826124416867615e-06, "loss": 0.9196, "step": 1309 }, { "epoch": 0.11, "learning_rate": 9.825762977072074e-06, "loss": 0.9471, "step": 1310 }, { "epoch": 0.11, "learning_rate": 9.825401168660676e-06, "loss": 0.943, "step": 1311 }, { "epoch": 0.11, "learning_rate": 9.825038991661062e-06, "loss": 0.8826, "step": 1312 }, { "epoch": 0.11, "learning_rate": 9.824676446100896e-06, "loss": 0.8994, "step": 1313 }, { "epoch": 0.11, "learning_rate": 9.82431353200787e-06, "loss": 0.9532, "step": 1314 }, { "epoch": 0.11, "learning_rate": 9.823950249409706e-06, "loss": 0.9256, "step": 1315 }, { "epoch": 0.11, "learning_rate": 9.823586598334155e-06, "loss": 0.9653, "step": 1316 }, { "epoch": 0.11, "learning_rate": 9.823222578808992e-06, "loss": 0.9583, "step": 1317 }, { "epoch": 0.11, "learning_rate": 9.822858190862024e-06, "loss": 0.9176, "step": 1318 }, { "epoch": 0.11, "learning_rate": 9.822493434521084e-06, "loss": 0.927, "step": 1319 }, { "epoch": 0.11, "learning_rate": 9.822128309814032e-06, "loss": 0.8732, "step": 1320 }, { "epoch": 0.11, "learning_rate": 9.821762816768762e-06, "loss": 0.9389, "step": 1321 }, { "epoch": 0.11, "learning_rate": 9.821396955413191e-06, "loss": 0.907, "step": 1322 }, { "epoch": 0.11, "learning_rate": 9.821030725775265e-06, "loss": 0.9456, "step": 1323 }, { "epoch": 0.11, "learning_rate": 9.820664127882958e-06, "loss": 0.9035, "step": 1324 }, { "epoch": 0.11, "learning_rate": 9.820297161764272e-06, "loss": 0.9258, "step": 1325 }, { "epoch": 0.11, "learning_rate": 9.819929827447238e-06, "loss": 0.9197, "step": 1326 }, { "epoch": 0.11, "learning_rate": 9.819562124959916e-06, "loss": 0.9257, "step": 1327 }, { "epoch": 0.11, "learning_rate": 9.819194054330392e-06, "loss": 0.7266, "step": 1328 }, { "epoch": 0.11, "learning_rate": 9.818825615586783e-06, "loss": 0.8939, "step": 1329 }, { "epoch": 0.11, "learning_rate": 9.818456808757228e-06, "loss": 0.921, "step": 1330 }, { "epoch": 0.11, "learning_rate": 9.818087633869902e-06, "loss": 0.9233, "step": 1331 }, { "epoch": 0.11, "learning_rate": 9.817718090953002e-06, "loss": 0.8849, "step": 1332 }, { "epoch": 0.11, "learning_rate": 9.817348180034759e-06, "loss": 0.9434, "step": 1333 }, { "epoch": 0.11, "learning_rate": 9.816977901143424e-06, "loss": 0.9183, "step": 1334 }, { "epoch": 0.11, "learning_rate": 9.816607254307285e-06, "loss": 0.8861, "step": 1335 }, { "epoch": 0.11, "learning_rate": 9.81623623955465e-06, "loss": 0.8428, "step": 1336 }, { "epoch": 0.11, "learning_rate": 9.815864856913861e-06, "loss": 0.8938, "step": 1337 }, { "epoch": 0.11, "learning_rate": 9.815493106413288e-06, "loss": 0.9097, "step": 1338 }, { "epoch": 0.11, "learning_rate": 9.815120988081324e-06, "loss": 0.9228, "step": 1339 }, { "epoch": 0.11, "learning_rate": 9.814748501946396e-06, "loss": 0.9126, "step": 1340 }, { "epoch": 0.11, "learning_rate": 9.814375648036953e-06, "loss": 0.9274, "step": 1341 }, { "epoch": 0.11, "learning_rate": 9.814002426381476e-06, "loss": 0.9006, "step": 1342 }, { "epoch": 0.11, "learning_rate": 9.813628837008477e-06, "loss": 0.8965, "step": 1343 }, { "epoch": 0.11, "learning_rate": 9.81325487994649e-06, "loss": 0.9125, "step": 1344 }, { "epoch": 0.11, "learning_rate": 9.81288055522408e-06, "loss": 0.9269, "step": 1345 }, { "epoch": 0.11, "learning_rate": 9.81250586286984e-06, "loss": 0.8978, "step": 1346 }, { "epoch": 0.11, "learning_rate": 9.81213080291239e-06, "loss": 0.8961, "step": 1347 }, { "epoch": 0.12, "learning_rate": 9.81175537538038e-06, "loss": 0.9132, "step": 1348 }, { "epoch": 0.12, "learning_rate": 9.811379580302487e-06, "loss": 0.922, "step": 1349 }, { "epoch": 0.12, "learning_rate": 9.811003417707415e-06, "loss": 0.9349, "step": 1350 }, { "epoch": 0.12, "learning_rate": 9.8106268876239e-06, "loss": 0.9559, "step": 1351 }, { "epoch": 0.12, "learning_rate": 9.810249990080698e-06, "loss": 0.9607, "step": 1352 }, { "epoch": 0.12, "learning_rate": 9.809872725106604e-06, "loss": 0.6922, "step": 1353 }, { "epoch": 0.12, "learning_rate": 9.809495092730432e-06, "loss": 0.8766, "step": 1354 }, { "epoch": 0.12, "learning_rate": 9.809117092981028e-06, "loss": 0.9405, "step": 1355 }, { "epoch": 0.12, "learning_rate": 9.808738725887268e-06, "loss": 0.962, "step": 1356 }, { "epoch": 0.12, "learning_rate": 9.808359991478049e-06, "loss": 0.9103, "step": 1357 }, { "epoch": 0.12, "learning_rate": 9.807980889782306e-06, "loss": 0.8983, "step": 1358 }, { "epoch": 0.12, "learning_rate": 9.807601420828992e-06, "loss": 0.9107, "step": 1359 }, { "epoch": 0.12, "learning_rate": 9.807221584647093e-06, "loss": 0.9132, "step": 1360 }, { "epoch": 0.12, "learning_rate": 9.806841381265626e-06, "loss": 0.9355, "step": 1361 }, { "epoch": 0.12, "learning_rate": 9.80646081071363e-06, "loss": 0.9106, "step": 1362 }, { "epoch": 0.12, "learning_rate": 9.806079873020177e-06, "loss": 0.9498, "step": 1363 }, { "epoch": 0.12, "learning_rate": 9.805698568214363e-06, "loss": 0.8771, "step": 1364 }, { "epoch": 0.12, "learning_rate": 9.805316896325316e-06, "loss": 0.9247, "step": 1365 }, { "epoch": 0.12, "learning_rate": 9.804934857382189e-06, "loss": 0.8968, "step": 1366 }, { "epoch": 0.12, "learning_rate": 9.804552451414162e-06, "loss": 0.9244, "step": 1367 }, { "epoch": 0.12, "learning_rate": 9.804169678450448e-06, "loss": 0.8819, "step": 1368 }, { "epoch": 0.12, "learning_rate": 9.803786538520284e-06, "loss": 0.9455, "step": 1369 }, { "epoch": 0.12, "learning_rate": 9.803403031652937e-06, "loss": 0.9174, "step": 1370 }, { "epoch": 0.12, "learning_rate": 9.8030191578777e-06, "loss": 0.9105, "step": 1371 }, { "epoch": 0.12, "learning_rate": 9.802634917223896e-06, "loss": 0.8992, "step": 1372 }, { "epoch": 0.12, "learning_rate": 9.802250309720873e-06, "loss": 0.8647, "step": 1373 }, { "epoch": 0.12, "learning_rate": 9.801865335398013e-06, "loss": 1.0003, "step": 1374 }, { "epoch": 0.12, "learning_rate": 9.80147999428472e-06, "loss": 0.9227, "step": 1375 }, { "epoch": 0.12, "learning_rate": 9.801094286410427e-06, "loss": 0.9563, "step": 1376 }, { "epoch": 0.12, "learning_rate": 9.8007082118046e-06, "loss": 0.7053, "step": 1377 }, { "epoch": 0.12, "learning_rate": 9.800321770496726e-06, "loss": 0.9063, "step": 1378 }, { "epoch": 0.12, "learning_rate": 9.799934962516324e-06, "loss": 0.9151, "step": 1379 }, { "epoch": 0.12, "learning_rate": 9.799547787892943e-06, "loss": 0.9234, "step": 1380 }, { "epoch": 0.12, "learning_rate": 9.799160246656153e-06, "loss": 0.9327, "step": 1381 }, { "epoch": 0.12, "learning_rate": 9.79877233883556e-06, "loss": 0.91, "step": 1382 }, { "epoch": 0.12, "learning_rate": 9.79838406446079e-06, "loss": 0.9122, "step": 1383 }, { "epoch": 0.12, "learning_rate": 9.797995423561507e-06, "loss": 0.9076, "step": 1384 }, { "epoch": 0.12, "learning_rate": 9.797606416167393e-06, "loss": 0.9389, "step": 1385 }, { "epoch": 0.12, "learning_rate": 9.797217042308164e-06, "loss": 0.9345, "step": 1386 }, { "epoch": 0.12, "learning_rate": 9.796827302013564e-06, "loss": 0.9519, "step": 1387 }, { "epoch": 0.12, "learning_rate": 9.79643719531336e-06, "loss": 0.9068, "step": 1388 }, { "epoch": 0.12, "learning_rate": 9.796046722237352e-06, "loss": 0.9021, "step": 1389 }, { "epoch": 0.12, "learning_rate": 9.795655882815365e-06, "loss": 0.9476, "step": 1390 }, { "epoch": 0.12, "learning_rate": 9.795264677077256e-06, "loss": 0.93, "step": 1391 }, { "epoch": 0.12, "learning_rate": 9.794873105052905e-06, "loss": 0.9264, "step": 1392 }, { "epoch": 0.12, "learning_rate": 9.794481166772221e-06, "loss": 0.9189, "step": 1393 }, { "epoch": 0.12, "learning_rate": 9.794088862265145e-06, "loss": 0.7006, "step": 1394 }, { "epoch": 0.12, "learning_rate": 9.793696191561644e-06, "loss": 0.9339, "step": 1395 }, { "epoch": 0.12, "learning_rate": 9.793303154691708e-06, "loss": 0.9568, "step": 1396 }, { "epoch": 0.12, "learning_rate": 9.792909751685363e-06, "loss": 0.8954, "step": 1397 }, { "epoch": 0.12, "learning_rate": 9.792515982572655e-06, "loss": 0.8852, "step": 1398 }, { "epoch": 0.12, "learning_rate": 9.792121847383666e-06, "loss": 0.9075, "step": 1399 }, { "epoch": 0.12, "learning_rate": 9.791727346148501e-06, "loss": 0.9347, "step": 1400 }, { "epoch": 0.12, "learning_rate": 9.791332478897293e-06, "loss": 0.895, "step": 1401 }, { "epoch": 0.12, "learning_rate": 9.790937245660203e-06, "loss": 0.8816, "step": 1402 }, { "epoch": 0.12, "learning_rate": 9.790541646467423e-06, "loss": 0.9362, "step": 1403 }, { "epoch": 0.12, "learning_rate": 9.790145681349172e-06, "loss": 0.5814, "step": 1404 }, { "epoch": 0.12, "learning_rate": 9.789749350335693e-06, "loss": 0.8632, "step": 1405 }, { "epoch": 0.12, "learning_rate": 9.789352653457262e-06, "loss": 0.8886, "step": 1406 }, { "epoch": 0.12, "learning_rate": 9.788955590744178e-06, "loss": 0.8712, "step": 1407 }, { "epoch": 0.12, "learning_rate": 9.78855816222677e-06, "loss": 0.8661, "step": 1408 }, { "epoch": 0.12, "learning_rate": 9.788160367935401e-06, "loss": 0.9047, "step": 1409 }, { "epoch": 0.12, "learning_rate": 9.787762207900451e-06, "loss": 0.8721, "step": 1410 }, { "epoch": 0.12, "learning_rate": 9.787363682152337e-06, "loss": 0.9252, "step": 1411 }, { "epoch": 0.12, "learning_rate": 9.786964790721498e-06, "loss": 0.8924, "step": 1412 }, { "epoch": 0.12, "learning_rate": 9.786565533638405e-06, "loss": 0.8644, "step": 1413 }, { "epoch": 0.12, "learning_rate": 9.786165910933554e-06, "loss": 0.8763, "step": 1414 }, { "epoch": 0.12, "learning_rate": 9.78576592263747e-06, "loss": 0.8569, "step": 1415 }, { "epoch": 0.12, "learning_rate": 9.785365568780709e-06, "loss": 0.9106, "step": 1416 }, { "epoch": 0.12, "learning_rate": 9.784964849393847e-06, "loss": 0.9019, "step": 1417 }, { "epoch": 0.12, "learning_rate": 9.784563764507498e-06, "loss": 0.9383, "step": 1418 }, { "epoch": 0.12, "learning_rate": 9.784162314152295e-06, "loss": 0.8898, "step": 1419 }, { "epoch": 0.12, "learning_rate": 9.783760498358906e-06, "loss": 0.9213, "step": 1420 }, { "epoch": 0.12, "learning_rate": 9.78335831715802e-06, "loss": 0.9235, "step": 1421 }, { "epoch": 0.12, "learning_rate": 9.78295577058036e-06, "loss": 0.8977, "step": 1422 }, { "epoch": 0.12, "learning_rate": 9.782552858656676e-06, "loss": 0.8905, "step": 1423 }, { "epoch": 0.12, "learning_rate": 9.782149581417741e-06, "loss": 0.9092, "step": 1424 }, { "epoch": 0.12, "learning_rate": 9.781745938894363e-06, "loss": 0.9084, "step": 1425 }, { "epoch": 0.12, "learning_rate": 9.78134193111737e-06, "loss": 0.548, "step": 1426 }, { "epoch": 0.12, "learning_rate": 9.780937558117625e-06, "loss": 0.9431, "step": 1427 }, { "epoch": 0.12, "learning_rate": 9.780532819926016e-06, "loss": 0.9129, "step": 1428 }, { "epoch": 0.12, "learning_rate": 9.780127716573458e-06, "loss": 0.9281, "step": 1429 }, { "epoch": 0.12, "learning_rate": 9.779722248090894e-06, "loss": 0.913, "step": 1430 }, { "epoch": 0.12, "learning_rate": 9.779316414509298e-06, "loss": 0.9314, "step": 1431 }, { "epoch": 0.12, "learning_rate": 9.778910215859666e-06, "loss": 0.8697, "step": 1432 }, { "epoch": 0.12, "learning_rate": 9.77850365217303e-06, "loss": 0.8955, "step": 1433 }, { "epoch": 0.12, "learning_rate": 9.778096723480445e-06, "loss": 0.9175, "step": 1434 }, { "epoch": 0.12, "learning_rate": 9.777689429812989e-06, "loss": 0.8908, "step": 1435 }, { "epoch": 0.12, "learning_rate": 9.777281771201779e-06, "loss": 0.9415, "step": 1436 }, { "epoch": 0.12, "learning_rate": 9.77687374767795e-06, "loss": 0.928, "step": 1437 }, { "epoch": 0.12, "learning_rate": 9.776465359272673e-06, "loss": 0.8871, "step": 1438 }, { "epoch": 0.12, "learning_rate": 9.776056606017138e-06, "loss": 0.9099, "step": 1439 }, { "epoch": 0.12, "learning_rate": 9.77564748794257e-06, "loss": 0.8967, "step": 1440 }, { "epoch": 0.12, "learning_rate": 9.77523800508022e-06, "loss": 0.9108, "step": 1441 }, { "epoch": 0.12, "learning_rate": 9.774828157461365e-06, "loss": 0.9003, "step": 1442 }, { "epoch": 0.12, "learning_rate": 9.774417945117312e-06, "loss": 0.9152, "step": 1443 }, { "epoch": 0.12, "learning_rate": 9.774007368079394e-06, "loss": 0.8979, "step": 1444 }, { "epoch": 0.12, "learning_rate": 9.773596426378976e-06, "loss": 0.8753, "step": 1445 }, { "epoch": 0.12, "learning_rate": 9.773185120047444e-06, "loss": 0.9688, "step": 1446 }, { "epoch": 0.12, "learning_rate": 9.772773449116218e-06, "loss": 0.9575, "step": 1447 }, { "epoch": 0.12, "learning_rate": 9.772361413616744e-06, "loss": 0.9407, "step": 1448 }, { "epoch": 0.12, "learning_rate": 9.771949013580492e-06, "loss": 0.8979, "step": 1449 }, { "epoch": 0.12, "learning_rate": 9.771536249038965e-06, "loss": 0.9518, "step": 1450 }, { "epoch": 0.12, "learning_rate": 9.771123120023693e-06, "loss": 0.8933, "step": 1451 }, { "epoch": 0.12, "learning_rate": 9.770709626566233e-06, "loss": 0.9056, "step": 1452 }, { "epoch": 0.12, "learning_rate": 9.770295768698168e-06, "loss": 0.8441, "step": 1453 }, { "epoch": 0.12, "learning_rate": 9.769881546451111e-06, "loss": 0.9029, "step": 1454 }, { "epoch": 0.12, "learning_rate": 9.769466959856703e-06, "loss": 0.871, "step": 1455 }, { "epoch": 0.12, "learning_rate": 9.76905200894661e-06, "loss": 0.9223, "step": 1456 }, { "epoch": 0.12, "learning_rate": 9.768636693752534e-06, "loss": 0.9204, "step": 1457 }, { "epoch": 0.12, "learning_rate": 9.768221014306192e-06, "loss": 0.8696, "step": 1458 }, { "epoch": 0.12, "learning_rate": 9.767804970639338e-06, "loss": 0.9344, "step": 1459 }, { "epoch": 0.12, "learning_rate": 9.767388562783753e-06, "loss": 0.8827, "step": 1460 }, { "epoch": 0.12, "learning_rate": 9.766971790771244e-06, "loss": 0.8727, "step": 1461 }, { "epoch": 0.12, "learning_rate": 9.766554654633645e-06, "loss": 0.8683, "step": 1462 }, { "epoch": 0.12, "learning_rate": 9.766137154402818e-06, "loss": 0.9021, "step": 1463 }, { "epoch": 0.12, "learning_rate": 9.765719290110655e-06, "loss": 0.8921, "step": 1464 }, { "epoch": 0.13, "learning_rate": 9.765301061789075e-06, "loss": 0.936, "step": 1465 }, { "epoch": 0.13, "learning_rate": 9.764882469470025e-06, "loss": 0.9203, "step": 1466 }, { "epoch": 0.13, "learning_rate": 9.764463513185477e-06, "loss": 0.9023, "step": 1467 }, { "epoch": 0.13, "learning_rate": 9.764044192967435e-06, "loss": 0.9309, "step": 1468 }, { "epoch": 0.13, "learning_rate": 9.763624508847926e-06, "loss": 0.9183, "step": 1469 }, { "epoch": 0.13, "learning_rate": 9.763204460859011e-06, "loss": 0.9227, "step": 1470 }, { "epoch": 0.13, "learning_rate": 9.762784049032774e-06, "loss": 0.9636, "step": 1471 }, { "epoch": 0.13, "learning_rate": 9.762363273401328e-06, "loss": 0.8869, "step": 1472 }, { "epoch": 0.13, "learning_rate": 9.761942133996813e-06, "loss": 0.9011, "step": 1473 }, { "epoch": 0.13, "learning_rate": 9.7615206308514e-06, "loss": 0.8859, "step": 1474 }, { "epoch": 0.13, "learning_rate": 9.761098763997282e-06, "loss": 0.9549, "step": 1475 }, { "epoch": 0.13, "learning_rate": 9.760676533466687e-06, "loss": 0.6067, "step": 1476 }, { "epoch": 0.13, "learning_rate": 9.760253939291864e-06, "loss": 0.9382, "step": 1477 }, { "epoch": 0.13, "learning_rate": 9.759830981505096e-06, "loss": 0.9315, "step": 1478 }, { "epoch": 0.13, "learning_rate": 9.75940766013869e-06, "loss": 0.9478, "step": 1479 }, { "epoch": 0.13, "learning_rate": 9.758983975224978e-06, "loss": 0.8609, "step": 1480 }, { "epoch": 0.13, "learning_rate": 9.758559926796327e-06, "loss": 0.919, "step": 1481 }, { "epoch": 0.13, "learning_rate": 9.758135514885125e-06, "loss": 0.9046, "step": 1482 }, { "epoch": 0.13, "learning_rate": 9.757710739523794e-06, "loss": 0.906, "step": 1483 }, { "epoch": 0.13, "learning_rate": 9.757285600744777e-06, "loss": 0.8538, "step": 1484 }, { "epoch": 0.13, "learning_rate": 9.75686009858055e-06, "loss": 0.879, "step": 1485 }, { "epoch": 0.13, "learning_rate": 9.756434233063616e-06, "loss": 0.8523, "step": 1486 }, { "epoch": 0.13, "learning_rate": 9.756008004226503e-06, "loss": 0.9209, "step": 1487 }, { "epoch": 0.13, "learning_rate": 9.75558141210177e-06, "loss": 0.9764, "step": 1488 }, { "epoch": 0.13, "learning_rate": 9.755154456721998e-06, "loss": 0.9415, "step": 1489 }, { "epoch": 0.13, "learning_rate": 9.754727138119807e-06, "loss": 0.9562, "step": 1490 }, { "epoch": 0.13, "learning_rate": 9.754299456327831e-06, "loss": 0.9135, "step": 1491 }, { "epoch": 0.13, "learning_rate": 9.753871411378743e-06, "loss": 0.8997, "step": 1492 }, { "epoch": 0.13, "learning_rate": 9.75344300330524e-06, "loss": 0.9106, "step": 1493 }, { "epoch": 0.13, "learning_rate": 9.753014232140041e-06, "loss": 0.9151, "step": 1494 }, { "epoch": 0.13, "learning_rate": 9.7525850979159e-06, "loss": 0.9386, "step": 1495 }, { "epoch": 0.13, "learning_rate": 9.752155600665598e-06, "loss": 0.9227, "step": 1496 }, { "epoch": 0.13, "learning_rate": 9.75172574042194e-06, "loss": 0.8839, "step": 1497 }, { "epoch": 0.13, "learning_rate": 9.751295517217763e-06, "loss": 0.8845, "step": 1498 }, { "epoch": 0.13, "learning_rate": 9.750864931085926e-06, "loss": 0.9175, "step": 1499 }, { "epoch": 0.13, "learning_rate": 9.750433982059324e-06, "loss": 0.8692, "step": 1500 }, { "epoch": 0.13, "learning_rate": 9.75000267017087e-06, "loss": 0.9092, "step": 1501 }, { "epoch": 0.13, "learning_rate": 9.749570995453515e-06, "loss": 0.9264, "step": 1502 }, { "epoch": 0.13, "learning_rate": 9.74913895794023e-06, "loss": 0.8769, "step": 1503 }, { "epoch": 0.13, "learning_rate": 9.748706557664014e-06, "loss": 0.9293, "step": 1504 }, { "epoch": 0.13, "learning_rate": 9.748273794657899e-06, "loss": 0.9298, "step": 1505 }, { "epoch": 0.13, "learning_rate": 9.74784066895494e-06, "loss": 0.9251, "step": 1506 }, { "epoch": 0.13, "learning_rate": 9.747407180588224e-06, "loss": 0.8992, "step": 1507 }, { "epoch": 0.13, "learning_rate": 9.746973329590858e-06, "loss": 0.8624, "step": 1508 }, { "epoch": 0.13, "learning_rate": 9.746539115995987e-06, "loss": 0.8942, "step": 1509 }, { "epoch": 0.13, "learning_rate": 9.746104539836776e-06, "loss": 0.9124, "step": 1510 }, { "epoch": 0.13, "learning_rate": 9.745669601146419e-06, "loss": 0.9447, "step": 1511 }, { "epoch": 0.13, "learning_rate": 9.74523429995814e-06, "loss": 0.8821, "step": 1512 }, { "epoch": 0.13, "learning_rate": 9.744798636305189e-06, "loss": 0.9611, "step": 1513 }, { "epoch": 0.13, "learning_rate": 9.744362610220845e-06, "loss": 0.9491, "step": 1514 }, { "epoch": 0.13, "learning_rate": 9.743926221738414e-06, "loss": 0.8777, "step": 1515 }, { "epoch": 0.13, "learning_rate": 9.743489470891227e-06, "loss": 0.928, "step": 1516 }, { "epoch": 0.13, "learning_rate": 9.743052357712648e-06, "loss": 0.903, "step": 1517 }, { "epoch": 0.13, "learning_rate": 9.742614882236066e-06, "loss": 0.9371, "step": 1518 }, { "epoch": 0.13, "learning_rate": 9.742177044494897e-06, "loss": 0.8682, "step": 1519 }, { "epoch": 0.13, "learning_rate": 9.741738844522581e-06, "loss": 0.6217, "step": 1520 }, { "epoch": 0.13, "learning_rate": 9.741300282352596e-06, "loss": 0.6543, "step": 1521 }, { "epoch": 0.13, "learning_rate": 9.740861358018442e-06, "loss": 0.9132, "step": 1522 }, { "epoch": 0.13, "learning_rate": 9.74042207155364e-06, "loss": 0.9627, "step": 1523 }, { "epoch": 0.13, "learning_rate": 9.739982422991748e-06, "loss": 0.9022, "step": 1524 }, { "epoch": 0.13, "learning_rate": 9.739542412366351e-06, "loss": 0.8715, "step": 1525 }, { "epoch": 0.13, "learning_rate": 9.739102039711056e-06, "loss": 0.9469, "step": 1526 }, { "epoch": 0.13, "learning_rate": 9.738661305059503e-06, "loss": 0.9326, "step": 1527 }, { "epoch": 0.13, "learning_rate": 9.738220208445354e-06, "loss": 0.9166, "step": 1528 }, { "epoch": 0.13, "learning_rate": 9.737778749902307e-06, "loss": 0.9488, "step": 1529 }, { "epoch": 0.13, "learning_rate": 9.73733692946408e-06, "loss": 0.9151, "step": 1530 }, { "epoch": 0.13, "learning_rate": 9.736894747164422e-06, "loss": 0.9397, "step": 1531 }, { "epoch": 0.13, "learning_rate": 9.736452203037109e-06, "loss": 0.8409, "step": 1532 }, { "epoch": 0.13, "learning_rate": 9.736009297115945e-06, "loss": 0.894, "step": 1533 }, { "epoch": 0.13, "learning_rate": 9.73556602943476e-06, "loss": 0.8745, "step": 1534 }, { "epoch": 0.13, "learning_rate": 9.735122400027414e-06, "loss": 0.9392, "step": 1535 }, { "epoch": 0.13, "learning_rate": 9.734678408927796e-06, "loss": 0.8896, "step": 1536 }, { "epoch": 0.13, "learning_rate": 9.734234056169816e-06, "loss": 0.8925, "step": 1537 }, { "epoch": 0.13, "learning_rate": 9.733789341787418e-06, "loss": 0.9406, "step": 1538 }, { "epoch": 0.13, "learning_rate": 9.73334426581457e-06, "loss": 0.9124, "step": 1539 }, { "epoch": 0.13, "learning_rate": 9.732898828285273e-06, "loss": 0.9071, "step": 1540 }, { "epoch": 0.13, "learning_rate": 9.732453029233548e-06, "loss": 0.9149, "step": 1541 }, { "epoch": 0.13, "learning_rate": 9.732006868693449e-06, "loss": 0.9238, "step": 1542 }, { "epoch": 0.13, "learning_rate": 9.731560346699056e-06, "loss": 0.8757, "step": 1543 }, { "epoch": 0.13, "learning_rate": 9.731113463284477e-06, "loss": 0.9305, "step": 1544 }, { "epoch": 0.13, "learning_rate": 9.730666218483845e-06, "loss": 0.9167, "step": 1545 }, { "epoch": 0.13, "learning_rate": 9.730218612331322e-06, "loss": 0.9342, "step": 1546 }, { "epoch": 0.13, "learning_rate": 9.729770644861103e-06, "loss": 0.947, "step": 1547 }, { "epoch": 0.13, "learning_rate": 9.729322316107404e-06, "loss": 0.9303, "step": 1548 }, { "epoch": 0.13, "learning_rate": 9.728873626104468e-06, "loss": 0.9143, "step": 1549 }, { "epoch": 0.13, "learning_rate": 9.728424574886571e-06, "loss": 0.9453, "step": 1550 }, { "epoch": 0.13, "learning_rate": 9.727975162488013e-06, "loss": 0.9104, "step": 1551 }, { "epoch": 0.13, "learning_rate": 9.727525388943123e-06, "loss": 0.9183, "step": 1552 }, { "epoch": 0.13, "learning_rate": 9.727075254286255e-06, "loss": 0.9066, "step": 1553 }, { "epoch": 0.13, "learning_rate": 9.726624758551796e-06, "loss": 0.9571, "step": 1554 }, { "epoch": 0.13, "learning_rate": 9.726173901774154e-06, "loss": 0.909, "step": 1555 }, { "epoch": 0.13, "learning_rate": 9.725722683987768e-06, "loss": 0.9037, "step": 1556 }, { "epoch": 0.13, "learning_rate": 9.725271105227105e-06, "loss": 0.9157, "step": 1557 }, { "epoch": 0.13, "learning_rate": 9.72481916552666e-06, "loss": 0.8898, "step": 1558 }, { "epoch": 0.13, "learning_rate": 9.724366864920951e-06, "loss": 0.9132, "step": 1559 }, { "epoch": 0.13, "learning_rate": 9.72391420344453e-06, "loss": 0.9126, "step": 1560 }, { "epoch": 0.13, "learning_rate": 9.723461181131974e-06, "loss": 0.8721, "step": 1561 }, { "epoch": 0.13, "learning_rate": 9.723007798017886e-06, "loss": 0.8757, "step": 1562 }, { "epoch": 0.13, "learning_rate": 9.722554054136896e-06, "loss": 0.9594, "step": 1563 }, { "epoch": 0.13, "learning_rate": 9.722099949523665e-06, "loss": 0.944, "step": 1564 }, { "epoch": 0.13, "learning_rate": 9.72164548421288e-06, "loss": 0.9523, "step": 1565 }, { "epoch": 0.13, "learning_rate": 9.721190658239255e-06, "loss": 0.9525, "step": 1566 }, { "epoch": 0.13, "learning_rate": 9.72073547163753e-06, "loss": 0.9181, "step": 1567 }, { "epoch": 0.13, "learning_rate": 9.720279924442477e-06, "loss": 0.9485, "step": 1568 }, { "epoch": 0.13, "learning_rate": 9.719824016688892e-06, "loss": 0.9603, "step": 1569 }, { "epoch": 0.13, "learning_rate": 9.7193677484116e-06, "loss": 0.9291, "step": 1570 }, { "epoch": 0.13, "learning_rate": 9.718911119645451e-06, "loss": 0.869, "step": 1571 }, { "epoch": 0.13, "learning_rate": 9.71845413042533e-06, "loss": 0.8725, "step": 1572 }, { "epoch": 0.13, "learning_rate": 9.717996780786136e-06, "loss": 0.924, "step": 1573 }, { "epoch": 0.13, "learning_rate": 9.717539070762811e-06, "loss": 0.907, "step": 1574 }, { "epoch": 0.13, "learning_rate": 9.717081000390313e-06, "loss": 0.9358, "step": 1575 }, { "epoch": 0.13, "learning_rate": 9.716622569703633e-06, "loss": 0.8934, "step": 1576 }, { "epoch": 0.13, "learning_rate": 9.716163778737788e-06, "loss": 0.9239, "step": 1577 }, { "epoch": 0.13, "learning_rate": 9.715704627527822e-06, "loss": 0.8469, "step": 1578 }, { "epoch": 0.13, "learning_rate": 9.71524511610881e-06, "loss": 0.9126, "step": 1579 }, { "epoch": 0.13, "learning_rate": 9.714785244515848e-06, "loss": 0.9188, "step": 1580 }, { "epoch": 0.13, "learning_rate": 9.714325012784065e-06, "loss": 0.9293, "step": 1581 }, { "epoch": 0.13, "learning_rate": 9.713864420948617e-06, "loss": 0.9173, "step": 1582 }, { "epoch": 0.14, "learning_rate": 9.713403469044685e-06, "loss": 0.9174, "step": 1583 }, { "epoch": 0.14, "learning_rate": 9.71294215710748e-06, "loss": 0.8791, "step": 1584 }, { "epoch": 0.14, "learning_rate": 9.712480485172235e-06, "loss": 0.8955, "step": 1585 }, { "epoch": 0.14, "learning_rate": 9.71201845327422e-06, "loss": 0.9046, "step": 1586 }, { "epoch": 0.14, "learning_rate": 9.711556061448725e-06, "loss": 0.9688, "step": 1587 }, { "epoch": 0.14, "learning_rate": 9.711093309731072e-06, "loss": 0.8868, "step": 1588 }, { "epoch": 0.14, "learning_rate": 9.710630198156606e-06, "loss": 0.9263, "step": 1589 }, { "epoch": 0.14, "learning_rate": 9.710166726760702e-06, "loss": 0.9361, "step": 1590 }, { "epoch": 0.14, "learning_rate": 9.709702895578764e-06, "loss": 0.9187, "step": 1591 }, { "epoch": 0.14, "learning_rate": 9.709238704646218e-06, "loss": 0.9439, "step": 1592 }, { "epoch": 0.14, "learning_rate": 9.708774153998523e-06, "loss": 0.8885, "step": 1593 }, { "epoch": 0.14, "learning_rate": 9.708309243671167e-06, "loss": 0.8623, "step": 1594 }, { "epoch": 0.14, "learning_rate": 9.707843973699657e-06, "loss": 0.9128, "step": 1595 }, { "epoch": 0.14, "learning_rate": 9.707378344119536e-06, "loss": 0.9346, "step": 1596 }, { "epoch": 0.14, "learning_rate": 9.70691235496637e-06, "loss": 0.9261, "step": 1597 }, { "epoch": 0.14, "learning_rate": 9.706446006275753e-06, "loss": 0.8781, "step": 1598 }, { "epoch": 0.14, "learning_rate": 9.705979298083307e-06, "loss": 0.8502, "step": 1599 }, { "epoch": 0.14, "learning_rate": 9.705512230424685e-06, "loss": 0.8927, "step": 1600 }, { "epoch": 0.14, "learning_rate": 9.705044803335559e-06, "loss": 0.8941, "step": 1601 }, { "epoch": 0.14, "learning_rate": 9.704577016851638e-06, "loss": 0.9206, "step": 1602 }, { "epoch": 0.14, "learning_rate": 9.70410887100865e-06, "loss": 0.8784, "step": 1603 }, { "epoch": 0.14, "learning_rate": 9.703640365842355e-06, "loss": 0.8901, "step": 1604 }, { "epoch": 0.14, "learning_rate": 9.703171501388542e-06, "loss": 0.9283, "step": 1605 }, { "epoch": 0.14, "learning_rate": 9.702702277683023e-06, "loss": 0.8882, "step": 1606 }, { "epoch": 0.14, "learning_rate": 9.70223269476164e-06, "loss": 0.9161, "step": 1607 }, { "epoch": 0.14, "learning_rate": 9.701762752660263e-06, "loss": 0.8893, "step": 1608 }, { "epoch": 0.14, "learning_rate": 9.701292451414787e-06, "loss": 0.9281, "step": 1609 }, { "epoch": 0.14, "learning_rate": 9.700821791061138e-06, "loss": 0.9355, "step": 1610 }, { "epoch": 0.14, "learning_rate": 9.700350771635268e-06, "loss": 0.943, "step": 1611 }, { "epoch": 0.14, "learning_rate": 9.699879393173152e-06, "loss": 0.9098, "step": 1612 }, { "epoch": 0.14, "learning_rate": 9.6994076557108e-06, "loss": 0.8787, "step": 1613 }, { "epoch": 0.14, "learning_rate": 9.698935559284243e-06, "loss": 0.9438, "step": 1614 }, { "epoch": 0.14, "learning_rate": 9.698463103929542e-06, "loss": 0.9278, "step": 1615 }, { "epoch": 0.14, "learning_rate": 9.697990289682788e-06, "loss": 0.8723, "step": 1616 }, { "epoch": 0.14, "learning_rate": 9.697517116580097e-06, "loss": 0.9613, "step": 1617 }, { "epoch": 0.14, "learning_rate": 9.697043584657609e-06, "loss": 0.9104, "step": 1618 }, { "epoch": 0.14, "learning_rate": 9.696569693951496e-06, "loss": 0.9053, "step": 1619 }, { "epoch": 0.14, "learning_rate": 9.696095444497959e-06, "loss": 0.9034, "step": 1620 }, { "epoch": 0.14, "learning_rate": 9.695620836333219e-06, "loss": 0.852, "step": 1621 }, { "epoch": 0.14, "learning_rate": 9.695145869493533e-06, "loss": 0.8994, "step": 1622 }, { "epoch": 0.14, "learning_rate": 9.69467054401518e-06, "loss": 0.9489, "step": 1623 }, { "epoch": 0.14, "learning_rate": 9.694194859934465e-06, "loss": 0.9182, "step": 1624 }, { "epoch": 0.14, "learning_rate": 9.693718817287728e-06, "loss": 0.9274, "step": 1625 }, { "epoch": 0.14, "learning_rate": 9.693242416111326e-06, "loss": 0.9136, "step": 1626 }, { "epoch": 0.14, "learning_rate": 9.692765656441655e-06, "loss": 0.9045, "step": 1627 }, { "epoch": 0.14, "learning_rate": 9.692288538315126e-06, "loss": 0.8764, "step": 1628 }, { "epoch": 0.14, "learning_rate": 9.691811061768188e-06, "loss": 0.9431, "step": 1629 }, { "epoch": 0.14, "learning_rate": 9.691333226837312e-06, "loss": 0.9283, "step": 1630 }, { "epoch": 0.14, "learning_rate": 9.690855033558997e-06, "loss": 0.9392, "step": 1631 }, { "epoch": 0.14, "learning_rate": 9.690376481969772e-06, "loss": 0.9289, "step": 1632 }, { "epoch": 0.14, "learning_rate": 9.689897572106187e-06, "loss": 0.915, "step": 1633 }, { "epoch": 0.14, "learning_rate": 9.689418304004827e-06, "loss": 0.8928, "step": 1634 }, { "epoch": 0.14, "learning_rate": 9.688938677702298e-06, "loss": 0.9769, "step": 1635 }, { "epoch": 0.14, "learning_rate": 9.68845869323524e-06, "loss": 0.9576, "step": 1636 }, { "epoch": 0.14, "learning_rate": 9.687978350640313e-06, "loss": 0.9414, "step": 1637 }, { "epoch": 0.14, "learning_rate": 9.68749764995421e-06, "loss": 0.9153, "step": 1638 }, { "epoch": 0.14, "learning_rate": 9.687016591213648e-06, "loss": 0.9312, "step": 1639 }, { "epoch": 0.14, "learning_rate": 9.686535174455374e-06, "loss": 0.9253, "step": 1640 }, { "epoch": 0.14, "learning_rate": 9.686053399716159e-06, "loss": 0.9131, "step": 1641 }, { "epoch": 0.14, "learning_rate": 9.685571267032807e-06, "loss": 0.8819, "step": 1642 }, { "epoch": 0.14, "learning_rate": 9.685088776442143e-06, "loss": 0.9524, "step": 1643 }, { "epoch": 0.14, "learning_rate": 9.684605927981022e-06, "loss": 0.9263, "step": 1644 }, { "epoch": 0.14, "learning_rate": 9.684122721686326e-06, "loss": 0.926, "step": 1645 }, { "epoch": 0.14, "learning_rate": 9.683639157594969e-06, "loss": 0.9353, "step": 1646 }, { "epoch": 0.14, "learning_rate": 9.683155235743882e-06, "loss": 0.9428, "step": 1647 }, { "epoch": 0.14, "learning_rate": 9.68267095617003e-06, "loss": 0.8607, "step": 1648 }, { "epoch": 0.14, "learning_rate": 9.68218631891041e-06, "loss": 0.9349, "step": 1649 }, { "epoch": 0.14, "learning_rate": 9.681701324002037e-06, "loss": 0.8754, "step": 1650 }, { "epoch": 0.14, "learning_rate": 9.681215971481958e-06, "loss": 0.6418, "step": 1651 }, { "epoch": 0.14, "learning_rate": 9.680730261387246e-06, "loss": 0.8795, "step": 1652 }, { "epoch": 0.14, "learning_rate": 9.680244193755003e-06, "loss": 0.8633, "step": 1653 }, { "epoch": 0.14, "learning_rate": 9.679757768622358e-06, "loss": 0.9338, "step": 1654 }, { "epoch": 0.14, "learning_rate": 9.679270986026463e-06, "loss": 0.9218, "step": 1655 }, { "epoch": 0.14, "learning_rate": 9.678783846004505e-06, "loss": 0.9017, "step": 1656 }, { "epoch": 0.14, "learning_rate": 9.678296348593691e-06, "loss": 0.9129, "step": 1657 }, { "epoch": 0.14, "learning_rate": 9.677808493831263e-06, "loss": 0.9303, "step": 1658 }, { "epoch": 0.14, "learning_rate": 9.677320281754481e-06, "loss": 0.9488, "step": 1659 }, { "epoch": 0.14, "learning_rate": 9.67683171240064e-06, "loss": 0.9052, "step": 1660 }, { "epoch": 0.14, "learning_rate": 9.676342785807059e-06, "loss": 0.9271, "step": 1661 }, { "epoch": 0.14, "learning_rate": 9.675853502011081e-06, "loss": 0.9357, "step": 1662 }, { "epoch": 0.14, "learning_rate": 9.675363861050085e-06, "loss": 0.9327, "step": 1663 }, { "epoch": 0.14, "learning_rate": 9.674873862961469e-06, "loss": 0.9087, "step": 1664 }, { "epoch": 0.14, "learning_rate": 9.674383507782662e-06, "loss": 0.8982, "step": 1665 }, { "epoch": 0.14, "learning_rate": 9.673892795551122e-06, "loss": 0.9053, "step": 1666 }, { "epoch": 0.14, "learning_rate": 9.67340172630433e-06, "loss": 0.9026, "step": 1667 }, { "epoch": 0.14, "learning_rate": 9.672910300079796e-06, "loss": 0.8592, "step": 1668 }, { "epoch": 0.14, "learning_rate": 9.672418516915058e-06, "loss": 0.9082, "step": 1669 }, { "epoch": 0.14, "learning_rate": 9.671926376847683e-06, "loss": 0.8835, "step": 1670 }, { "epoch": 0.14, "learning_rate": 9.67143387991526e-06, "loss": 0.9016, "step": 1671 }, { "epoch": 0.14, "learning_rate": 9.67094102615541e-06, "loss": 0.8778, "step": 1672 }, { "epoch": 0.14, "learning_rate": 9.670447815605779e-06, "loss": 0.9515, "step": 1673 }, { "epoch": 0.14, "learning_rate": 9.66995424830404e-06, "loss": 0.8978, "step": 1674 }, { "epoch": 0.14, "learning_rate": 9.669460324287899e-06, "loss": 0.9121, "step": 1675 }, { "epoch": 0.14, "learning_rate": 9.668966043595077e-06, "loss": 0.9045, "step": 1676 }, { "epoch": 0.14, "learning_rate": 9.668471406263334e-06, "loss": 0.895, "step": 1677 }, { "epoch": 0.14, "learning_rate": 9.667976412330454e-06, "loss": 0.9363, "step": 1678 }, { "epoch": 0.14, "learning_rate": 9.667481061834244e-06, "loss": 0.9213, "step": 1679 }, { "epoch": 0.14, "learning_rate": 9.666985354812542e-06, "loss": 0.9052, "step": 1680 }, { "epoch": 0.14, "learning_rate": 9.666489291303214e-06, "loss": 0.98, "step": 1681 }, { "epoch": 0.14, "learning_rate": 9.66599287134415e-06, "loss": 0.9151, "step": 1682 }, { "epoch": 0.14, "learning_rate": 9.665496094973271e-06, "loss": 0.8909, "step": 1683 }, { "epoch": 0.14, "learning_rate": 9.664998962228523e-06, "loss": 0.9604, "step": 1684 }, { "epoch": 0.14, "learning_rate": 9.664501473147878e-06, "loss": 0.9, "step": 1685 }, { "epoch": 0.14, "learning_rate": 9.664003627769335e-06, "loss": 0.9041, "step": 1686 }, { "epoch": 0.14, "learning_rate": 9.663505426130927e-06, "loss": 0.9108, "step": 1687 }, { "epoch": 0.14, "learning_rate": 9.663006868270708e-06, "loss": 0.9182, "step": 1688 }, { "epoch": 0.14, "learning_rate": 9.662507954226756e-06, "loss": 0.9127, "step": 1689 }, { "epoch": 0.14, "learning_rate": 9.662008684037184e-06, "loss": 0.903, "step": 1690 }, { "epoch": 0.14, "learning_rate": 9.66150905774013e-06, "loss": 0.935, "step": 1691 }, { "epoch": 0.14, "learning_rate": 9.661009075373754e-06, "loss": 0.9118, "step": 1692 }, { "epoch": 0.14, "learning_rate": 9.66050873697625e-06, "loss": 0.9064, "step": 1693 }, { "epoch": 0.14, "learning_rate": 9.660008042585836e-06, "loss": 0.9217, "step": 1694 }, { "epoch": 0.14, "learning_rate": 9.659506992240757e-06, "loss": 0.9348, "step": 1695 }, { "epoch": 0.14, "learning_rate": 9.659005585979286e-06, "loss": 0.9128, "step": 1696 }, { "epoch": 0.14, "learning_rate": 9.658503823839722e-06, "loss": 0.9015, "step": 1697 }, { "epoch": 0.14, "learning_rate": 9.658001705860393e-06, "loss": 0.8952, "step": 1698 }, { "epoch": 0.14, "learning_rate": 9.657499232079652e-06, "loss": 0.8642, "step": 1699 }, { "epoch": 0.15, "learning_rate": 9.656996402535882e-06, "loss": 0.9151, "step": 1700 }, { "epoch": 0.15, "learning_rate": 9.656493217267492e-06, "loss": 0.8658, "step": 1701 }, { "epoch": 0.15, "learning_rate": 9.655989676312918e-06, "loss": 0.9197, "step": 1702 }, { "epoch": 0.15, "learning_rate": 9.655485779710621e-06, "loss": 0.9093, "step": 1703 }, { "epoch": 0.15, "learning_rate": 9.654981527499092e-06, "loss": 0.9353, "step": 1704 }, { "epoch": 0.15, "learning_rate": 9.65447691971685e-06, "loss": 0.8914, "step": 1705 }, { "epoch": 0.15, "learning_rate": 9.653971956402434e-06, "loss": 0.9491, "step": 1706 }, { "epoch": 0.15, "learning_rate": 9.653466637594424e-06, "loss": 0.9469, "step": 1707 }, { "epoch": 0.15, "learning_rate": 9.65296096333141e-06, "loss": 0.8497, "step": 1708 }, { "epoch": 0.15, "learning_rate": 9.652454933652026e-06, "loss": 0.9194, "step": 1709 }, { "epoch": 0.15, "learning_rate": 9.65194854859492e-06, "loss": 0.911, "step": 1710 }, { "epoch": 0.15, "learning_rate": 9.651441808198773e-06, "loss": 0.8823, "step": 1711 }, { "epoch": 0.15, "learning_rate": 9.650934712502293e-06, "loss": 0.9479, "step": 1712 }, { "epoch": 0.15, "learning_rate": 9.650427261544213e-06, "loss": 0.974, "step": 1713 }, { "epoch": 0.15, "learning_rate": 9.649919455363297e-06, "loss": 0.9276, "step": 1714 }, { "epoch": 0.15, "learning_rate": 9.649411293998334e-06, "loss": 0.8919, "step": 1715 }, { "epoch": 0.15, "learning_rate": 9.648902777488138e-06, "loss": 0.9196, "step": 1716 }, { "epoch": 0.15, "learning_rate": 9.648393905871552e-06, "loss": 0.8947, "step": 1717 }, { "epoch": 0.15, "learning_rate": 9.647884679187447e-06, "loss": 0.9125, "step": 1718 }, { "epoch": 0.15, "learning_rate": 9.64737509747472e-06, "loss": 0.9088, "step": 1719 }, { "epoch": 0.15, "learning_rate": 9.646865160772295e-06, "loss": 0.865, "step": 1720 }, { "epoch": 0.15, "learning_rate": 9.646354869119125e-06, "loss": 0.9264, "step": 1721 }, { "epoch": 0.15, "learning_rate": 9.645844222554189e-06, "loss": 0.6343, "step": 1722 }, { "epoch": 0.15, "learning_rate": 9.645333221116488e-06, "loss": 0.9476, "step": 1723 }, { "epoch": 0.15, "learning_rate": 9.64482186484506e-06, "loss": 0.838, "step": 1724 }, { "epoch": 0.15, "learning_rate": 9.644310153778963e-06, "loss": 0.9165, "step": 1725 }, { "epoch": 0.15, "learning_rate": 9.643798087957284e-06, "loss": 0.8897, "step": 1726 }, { "epoch": 0.15, "learning_rate": 9.643285667419138e-06, "loss": 0.9223, "step": 1727 }, { "epoch": 0.15, "learning_rate": 9.642772892203665e-06, "loss": 0.9054, "step": 1728 }, { "epoch": 0.15, "learning_rate": 9.642259762350034e-06, "loss": 0.9327, "step": 1729 }, { "epoch": 0.15, "learning_rate": 9.64174627789744e-06, "loss": 0.861, "step": 1730 }, { "epoch": 0.15, "learning_rate": 9.641232438885106e-06, "loss": 0.8929, "step": 1731 }, { "epoch": 0.15, "learning_rate": 9.64071824535228e-06, "loss": 0.9098, "step": 1732 }, { "epoch": 0.15, "learning_rate": 9.640203697338242e-06, "loss": 0.9015, "step": 1733 }, { "epoch": 0.15, "learning_rate": 9.639688794882293e-06, "loss": 0.926, "step": 1734 }, { "epoch": 0.15, "learning_rate": 9.639173538023766e-06, "loss": 0.9236, "step": 1735 }, { "epoch": 0.15, "learning_rate": 9.638657926802016e-06, "loss": 0.9076, "step": 1736 }, { "epoch": 0.15, "learning_rate": 9.63814196125643e-06, "loss": 0.9129, "step": 1737 }, { "epoch": 0.15, "learning_rate": 9.63762564142642e-06, "loss": 0.9663, "step": 1738 }, { "epoch": 0.15, "learning_rate": 9.637108967351424e-06, "loss": 0.9071, "step": 1739 }, { "epoch": 0.15, "learning_rate": 9.63659193907091e-06, "loss": 0.9167, "step": 1740 }, { "epoch": 0.15, "learning_rate": 9.636074556624369e-06, "loss": 0.914, "step": 1741 }, { "epoch": 0.15, "learning_rate": 9.63555682005132e-06, "loss": 0.8623, "step": 1742 }, { "epoch": 0.15, "learning_rate": 9.635038729391315e-06, "loss": 0.9365, "step": 1743 }, { "epoch": 0.15, "learning_rate": 9.634520284683925e-06, "loss": 0.8544, "step": 1744 }, { "epoch": 0.15, "learning_rate": 9.634001485968754e-06, "loss": 0.9097, "step": 1745 }, { "epoch": 0.15, "learning_rate": 9.633482333285426e-06, "loss": 0.9119, "step": 1746 }, { "epoch": 0.15, "learning_rate": 9.632962826673599e-06, "loss": 0.9169, "step": 1747 }, { "epoch": 0.15, "learning_rate": 9.632442966172959e-06, "loss": 0.9329, "step": 1748 }, { "epoch": 0.15, "learning_rate": 9.63192275182321e-06, "loss": 0.9495, "step": 1749 }, { "epoch": 0.15, "learning_rate": 9.63140218366409e-06, "loss": 0.88, "step": 1750 }, { "epoch": 0.15, "learning_rate": 9.630881261735364e-06, "loss": 0.9109, "step": 1751 }, { "epoch": 0.15, "learning_rate": 9.630359986076822e-06, "loss": 0.882, "step": 1752 }, { "epoch": 0.15, "learning_rate": 9.629838356728283e-06, "loss": 0.9325, "step": 1753 }, { "epoch": 0.15, "learning_rate": 9.629316373729588e-06, "loss": 0.8925, "step": 1754 }, { "epoch": 0.15, "learning_rate": 9.628794037120613e-06, "loss": 0.9117, "step": 1755 }, { "epoch": 0.15, "learning_rate": 9.628271346941252e-06, "loss": 0.903, "step": 1756 }, { "epoch": 0.15, "learning_rate": 9.627748303231436e-06, "loss": 0.9234, "step": 1757 }, { "epoch": 0.15, "learning_rate": 9.627224906031113e-06, "loss": 0.8953, "step": 1758 }, { "epoch": 0.15, "learning_rate": 9.626701155380266e-06, "loss": 0.9092, "step": 1759 }, { "epoch": 0.15, "learning_rate": 9.626177051318899e-06, "loss": 0.8987, "step": 1760 }, { "epoch": 0.15, "learning_rate": 9.625652593887048e-06, "loss": 0.9006, "step": 1761 }, { "epoch": 0.15, "learning_rate": 9.625127783124773e-06, "loss": 0.857, "step": 1762 }, { "epoch": 0.15, "learning_rate": 9.624602619072161e-06, "loss": 0.9123, "step": 1763 }, { "epoch": 0.15, "learning_rate": 9.624077101769327e-06, "loss": 0.9131, "step": 1764 }, { "epoch": 0.15, "learning_rate": 9.623551231256412e-06, "loss": 0.9391, "step": 1765 }, { "epoch": 0.15, "learning_rate": 9.623025007573585e-06, "loss": 0.894, "step": 1766 }, { "epoch": 0.15, "learning_rate": 9.622498430761042e-06, "loss": 0.913, "step": 1767 }, { "epoch": 0.15, "learning_rate": 9.621971500859006e-06, "loss": 0.9164, "step": 1768 }, { "epoch": 0.15, "learning_rate": 9.621444217907728e-06, "loss": 0.9493, "step": 1769 }, { "epoch": 0.15, "learning_rate": 9.62091658194748e-06, "loss": 0.8739, "step": 1770 }, { "epoch": 0.15, "learning_rate": 9.620388593018569e-06, "loss": 0.9029, "step": 1771 }, { "epoch": 0.15, "learning_rate": 9.619860251161323e-06, "loss": 0.9526, "step": 1772 }, { "epoch": 0.15, "learning_rate": 9.6193315564161e-06, "loss": 0.9132, "step": 1773 }, { "epoch": 0.15, "learning_rate": 9.618802508823287e-06, "loss": 0.8934, "step": 1774 }, { "epoch": 0.15, "learning_rate": 9.618273108423292e-06, "loss": 0.9566, "step": 1775 }, { "epoch": 0.15, "learning_rate": 9.617743355256557e-06, "loss": 0.8866, "step": 1776 }, { "epoch": 0.15, "learning_rate": 9.617213249363542e-06, "loss": 0.9448, "step": 1777 }, { "epoch": 0.15, "learning_rate": 9.616682790784743e-06, "loss": 0.9156, "step": 1778 }, { "epoch": 0.15, "learning_rate": 9.61615197956068e-06, "loss": 0.9009, "step": 1779 }, { "epoch": 0.15, "learning_rate": 9.615620815731895e-06, "loss": 0.9426, "step": 1780 }, { "epoch": 0.15, "learning_rate": 9.615089299338962e-06, "loss": 0.9329, "step": 1781 }, { "epoch": 0.15, "learning_rate": 9.614557430422484e-06, "loss": 0.9691, "step": 1782 }, { "epoch": 0.15, "learning_rate": 9.614025209023084e-06, "loss": 0.9446, "step": 1783 }, { "epoch": 0.15, "learning_rate": 9.613492635181418e-06, "loss": 0.9087, "step": 1784 }, { "epoch": 0.15, "learning_rate": 9.612959708938166e-06, "loss": 0.859, "step": 1785 }, { "epoch": 0.15, "learning_rate": 9.612426430334036e-06, "loss": 0.9172, "step": 1786 }, { "epoch": 0.15, "learning_rate": 9.611892799409762e-06, "loss": 0.9475, "step": 1787 }, { "epoch": 0.15, "learning_rate": 9.611358816206104e-06, "loss": 0.901, "step": 1788 }, { "epoch": 0.15, "learning_rate": 9.610824480763855e-06, "loss": 0.9506, "step": 1789 }, { "epoch": 0.15, "learning_rate": 9.610289793123823e-06, "loss": 0.9049, "step": 1790 }, { "epoch": 0.15, "learning_rate": 9.609754753326856e-06, "loss": 0.6498, "step": 1791 }, { "epoch": 0.15, "learning_rate": 9.609219361413822e-06, "loss": 0.9375, "step": 1792 }, { "epoch": 0.15, "learning_rate": 9.608683617425616e-06, "loss": 0.9232, "step": 1793 }, { "epoch": 0.15, "learning_rate": 9.60814752140316e-06, "loss": 0.8846, "step": 1794 }, { "epoch": 0.15, "learning_rate": 9.607611073387403e-06, "loss": 0.89, "step": 1795 }, { "epoch": 0.15, "learning_rate": 9.607074273419324e-06, "loss": 0.8992, "step": 1796 }, { "epoch": 0.15, "learning_rate": 9.606537121539926e-06, "loss": 0.867, "step": 1797 }, { "epoch": 0.15, "learning_rate": 9.605999617790237e-06, "loss": 0.8945, "step": 1798 }, { "epoch": 0.15, "learning_rate": 9.60546176221132e-06, "loss": 0.8721, "step": 1799 }, { "epoch": 0.15, "learning_rate": 9.604923554844253e-06, "loss": 0.882, "step": 1800 }, { "epoch": 0.15, "learning_rate": 9.604384995730149e-06, "loss": 0.9082, "step": 1801 }, { "epoch": 0.15, "learning_rate": 9.603846084910146e-06, "loss": 0.9197, "step": 1802 }, { "epoch": 0.15, "learning_rate": 9.603306822425408e-06, "loss": 0.8968, "step": 1803 }, { "epoch": 0.15, "learning_rate": 9.602767208317127e-06, "loss": 0.9139, "step": 1804 }, { "epoch": 0.15, "learning_rate": 9.602227242626523e-06, "loss": 0.939, "step": 1805 }, { "epoch": 0.15, "learning_rate": 9.60168692539484e-06, "loss": 0.8695, "step": 1806 }, { "epoch": 0.15, "learning_rate": 9.601146256663349e-06, "loss": 0.9244, "step": 1807 }, { "epoch": 0.15, "learning_rate": 9.60060523647335e-06, "loss": 0.8894, "step": 1808 }, { "epoch": 0.15, "learning_rate": 9.60006386486617e-06, "loss": 0.8699, "step": 1809 }, { "epoch": 0.15, "learning_rate": 9.59952214188316e-06, "loss": 0.9115, "step": 1810 }, { "epoch": 0.15, "learning_rate": 9.5989800675657e-06, "loss": 0.6384, "step": 1811 }, { "epoch": 0.15, "learning_rate": 9.598437641955196e-06, "loss": 0.8988, "step": 1812 }, { "epoch": 0.15, "learning_rate": 9.59789486509308e-06, "loss": 0.9319, "step": 1813 }, { "epoch": 0.15, "learning_rate": 9.597351737020816e-06, "loss": 0.9485, "step": 1814 }, { "epoch": 0.15, "learning_rate": 9.596808257779888e-06, "loss": 0.9158, "step": 1815 }, { "epoch": 0.15, "learning_rate": 9.59626442741181e-06, "loss": 0.8625, "step": 1816 }, { "epoch": 0.16, "learning_rate": 9.595720245958122e-06, "loss": 0.8683, "step": 1817 }, { "epoch": 0.16, "learning_rate": 9.595175713460392e-06, "loss": 0.9014, "step": 1818 }, { "epoch": 0.16, "learning_rate": 9.594630829960214e-06, "loss": 0.9105, "step": 1819 }, { "epoch": 0.16, "learning_rate": 9.594085595499208e-06, "loss": 0.9043, "step": 1820 }, { "epoch": 0.16, "learning_rate": 9.593540010119023e-06, "loss": 0.9161, "step": 1821 }, { "epoch": 0.16, "learning_rate": 9.592994073861332e-06, "loss": 0.9303, "step": 1822 }, { "epoch": 0.16, "learning_rate": 9.592447786767839e-06, "loss": 0.9018, "step": 1823 }, { "epoch": 0.16, "learning_rate": 9.591901148880269e-06, "loss": 0.9131, "step": 1824 }, { "epoch": 0.16, "learning_rate": 9.59135416024038e-06, "loss": 0.8995, "step": 1825 }, { "epoch": 0.16, "learning_rate": 9.590806820889952e-06, "loss": 0.9188, "step": 1826 }, { "epoch": 0.16, "learning_rate": 9.590259130870793e-06, "loss": 0.9317, "step": 1827 }, { "epoch": 0.16, "learning_rate": 9.589711090224739e-06, "loss": 0.8451, "step": 1828 }, { "epoch": 0.16, "learning_rate": 9.589162698993652e-06, "loss": 0.9619, "step": 1829 }, { "epoch": 0.16, "learning_rate": 9.588613957219422e-06, "loss": 0.9204, "step": 1830 }, { "epoch": 0.16, "learning_rate": 9.588064864943962e-06, "loss": 0.936, "step": 1831 }, { "epoch": 0.16, "learning_rate": 9.587515422209217e-06, "loss": 0.8799, "step": 1832 }, { "epoch": 0.16, "learning_rate": 9.586965629057156e-06, "loss": 0.572, "step": 1833 }, { "epoch": 0.16, "learning_rate": 9.586415485529773e-06, "loss": 0.9859, "step": 1834 }, { "epoch": 0.16, "learning_rate": 9.585864991669092e-06, "loss": 0.8688, "step": 1835 }, { "epoch": 0.16, "learning_rate": 9.585314147517162e-06, "loss": 0.8963, "step": 1836 }, { "epoch": 0.16, "learning_rate": 9.58476295311606e-06, "loss": 0.9074, "step": 1837 }, { "epoch": 0.16, "learning_rate": 9.584211408507889e-06, "loss": 0.9793, "step": 1838 }, { "epoch": 0.16, "learning_rate": 9.583659513734777e-06, "loss": 0.9054, "step": 1839 }, { "epoch": 0.16, "learning_rate": 9.58310726883888e-06, "loss": 0.9284, "step": 1840 }, { "epoch": 0.16, "learning_rate": 9.582554673862387e-06, "loss": 0.9269, "step": 1841 }, { "epoch": 0.16, "learning_rate": 9.5820017288475e-06, "loss": 0.8311, "step": 1842 }, { "epoch": 0.16, "learning_rate": 9.581448433836463e-06, "loss": 0.8476, "step": 1843 }, { "epoch": 0.16, "learning_rate": 9.580894788871534e-06, "loss": 0.8407, "step": 1844 }, { "epoch": 0.16, "learning_rate": 9.580340793995005e-06, "loss": 0.9198, "step": 1845 }, { "epoch": 0.16, "learning_rate": 9.579786449249193e-06, "loss": 0.8933, "step": 1846 }, { "epoch": 0.16, "learning_rate": 9.579231754676443e-06, "loss": 0.9319, "step": 1847 }, { "epoch": 0.16, "learning_rate": 9.578676710319124e-06, "loss": 0.8962, "step": 1848 }, { "epoch": 0.16, "learning_rate": 9.578121316219632e-06, "loss": 0.9198, "step": 1849 }, { "epoch": 0.16, "learning_rate": 9.577565572420392e-06, "loss": 0.9136, "step": 1850 }, { "epoch": 0.16, "learning_rate": 9.577009478963855e-06, "loss": 0.9077, "step": 1851 }, { "epoch": 0.16, "learning_rate": 9.576453035892496e-06, "loss": 0.8984, "step": 1852 }, { "epoch": 0.16, "learning_rate": 9.575896243248821e-06, "loss": 0.6138, "step": 1853 }, { "epoch": 0.16, "learning_rate": 9.57533910107536e-06, "loss": 0.8963, "step": 1854 }, { "epoch": 0.16, "learning_rate": 9.574781609414672e-06, "loss": 0.9114, "step": 1855 }, { "epoch": 0.16, "learning_rate": 9.574223768309338e-06, "loss": 0.9036, "step": 1856 }, { "epoch": 0.16, "learning_rate": 9.573665577801969e-06, "loss": 0.8872, "step": 1857 }, { "epoch": 0.16, "learning_rate": 9.573107037935205e-06, "loss": 0.8394, "step": 1858 }, { "epoch": 0.16, "learning_rate": 9.572548148751708e-06, "loss": 0.8431, "step": 1859 }, { "epoch": 0.16, "learning_rate": 9.571988910294168e-06, "loss": 0.9329, "step": 1860 }, { "epoch": 0.16, "learning_rate": 9.571429322605306e-06, "loss": 0.8635, "step": 1861 }, { "epoch": 0.16, "learning_rate": 9.570869385727861e-06, "loss": 0.8941, "step": 1862 }, { "epoch": 0.16, "learning_rate": 9.57030909970461e-06, "loss": 0.9219, "step": 1863 }, { "epoch": 0.16, "learning_rate": 9.569748464578343e-06, "loss": 0.9309, "step": 1864 }, { "epoch": 0.16, "learning_rate": 9.569187480391891e-06, "loss": 0.903, "step": 1865 }, { "epoch": 0.16, "learning_rate": 9.5686261471881e-06, "loss": 0.9159, "step": 1866 }, { "epoch": 0.16, "learning_rate": 9.56806446500985e-06, "loss": 0.8881, "step": 1867 }, { "epoch": 0.16, "learning_rate": 9.567502433900044e-06, "loss": 0.9025, "step": 1868 }, { "epoch": 0.16, "learning_rate": 9.566940053901612e-06, "loss": 0.91, "step": 1869 }, { "epoch": 0.16, "learning_rate": 9.566377325057514e-06, "loss": 0.6273, "step": 1870 }, { "epoch": 0.16, "learning_rate": 9.565814247410732e-06, "loss": 0.8978, "step": 1871 }, { "epoch": 0.16, "learning_rate": 9.565250821004277e-06, "loss": 0.9424, "step": 1872 }, { "epoch": 0.16, "learning_rate": 9.564687045881187e-06, "loss": 0.8609, "step": 1873 }, { "epoch": 0.16, "learning_rate": 9.564122922084524e-06, "loss": 0.9318, "step": 1874 }, { "epoch": 0.16, "learning_rate": 9.563558449657382e-06, "loss": 0.9129, "step": 1875 }, { "epoch": 0.16, "learning_rate": 9.562993628642875e-06, "loss": 0.8883, "step": 1876 }, { "epoch": 0.16, "learning_rate": 9.56242845908415e-06, "loss": 0.8851, "step": 1877 }, { "epoch": 0.16, "learning_rate": 9.561862941024372e-06, "loss": 0.891, "step": 1878 }, { "epoch": 0.16, "learning_rate": 9.561297074506744e-06, "loss": 0.8946, "step": 1879 }, { "epoch": 0.16, "learning_rate": 9.560730859574488e-06, "loss": 0.8862, "step": 1880 }, { "epoch": 0.16, "learning_rate": 9.560164296270851e-06, "loss": 0.9243, "step": 1881 }, { "epoch": 0.16, "learning_rate": 9.559597384639116e-06, "loss": 0.9077, "step": 1882 }, { "epoch": 0.16, "learning_rate": 9.55903012472258e-06, "loss": 0.9373, "step": 1883 }, { "epoch": 0.16, "learning_rate": 9.558462516564579e-06, "loss": 0.9151, "step": 1884 }, { "epoch": 0.16, "learning_rate": 9.557894560208464e-06, "loss": 0.8905, "step": 1885 }, { "epoch": 0.16, "learning_rate": 9.557326255697624e-06, "loss": 0.9117, "step": 1886 }, { "epoch": 0.16, "learning_rate": 9.556757603075467e-06, "loss": 0.6552, "step": 1887 }, { "epoch": 0.16, "learning_rate": 9.556188602385426e-06, "loss": 0.8901, "step": 1888 }, { "epoch": 0.16, "learning_rate": 9.55561925367097e-06, "loss": 0.9095, "step": 1889 }, { "epoch": 0.16, "learning_rate": 9.555049556975583e-06, "loss": 0.9146, "step": 1890 }, { "epoch": 0.16, "learning_rate": 9.554479512342785e-06, "loss": 0.8935, "step": 1891 }, { "epoch": 0.16, "learning_rate": 9.553909119816117e-06, "loss": 0.8701, "step": 1892 }, { "epoch": 0.16, "learning_rate": 9.55333837943915e-06, "loss": 0.8733, "step": 1893 }, { "epoch": 0.16, "learning_rate": 9.552767291255478e-06, "loss": 0.8782, "step": 1894 }, { "epoch": 0.16, "learning_rate": 9.552195855308726e-06, "loss": 0.8736, "step": 1895 }, { "epoch": 0.16, "learning_rate": 9.551624071642542e-06, "loss": 0.9223, "step": 1896 }, { "epoch": 0.16, "learning_rate": 9.551051940300602e-06, "loss": 0.9234, "step": 1897 }, { "epoch": 0.16, "learning_rate": 9.550479461326607e-06, "loss": 0.9577, "step": 1898 }, { "epoch": 0.16, "learning_rate": 9.549906634764286e-06, "loss": 0.93, "step": 1899 }, { "epoch": 0.16, "learning_rate": 9.549333460657397e-06, "loss": 0.9316, "step": 1900 }, { "epoch": 0.16, "learning_rate": 9.548759939049718e-06, "loss": 0.9079, "step": 1901 }, { "epoch": 0.16, "learning_rate": 9.54818606998506e-06, "loss": 0.9039, "step": 1902 }, { "epoch": 0.16, "learning_rate": 9.547611853507257e-06, "loss": 0.8541, "step": 1903 }, { "epoch": 0.16, "learning_rate": 9.54703728966017e-06, "loss": 0.8784, "step": 1904 }, { "epoch": 0.16, "learning_rate": 9.546462378487689e-06, "loss": 0.9141, "step": 1905 }, { "epoch": 0.16, "learning_rate": 9.545887120033727e-06, "loss": 0.9232, "step": 1906 }, { "epoch": 0.16, "learning_rate": 9.545311514342228e-06, "loss": 0.8859, "step": 1907 }, { "epoch": 0.16, "learning_rate": 9.544735561457154e-06, "loss": 0.8597, "step": 1908 }, { "epoch": 0.16, "learning_rate": 9.544159261422505e-06, "loss": 0.9207, "step": 1909 }, { "epoch": 0.16, "learning_rate": 9.543582614282297e-06, "loss": 0.8786, "step": 1910 }, { "epoch": 0.16, "learning_rate": 9.54300562008058e-06, "loss": 0.8888, "step": 1911 }, { "epoch": 0.16, "learning_rate": 9.542428278861427e-06, "loss": 0.8673, "step": 1912 }, { "epoch": 0.16, "learning_rate": 9.541850590668938e-06, "loss": 0.9284, "step": 1913 }, { "epoch": 0.16, "learning_rate": 9.54127255554724e-06, "loss": 0.9049, "step": 1914 }, { "epoch": 0.16, "learning_rate": 9.540694173540485e-06, "loss": 0.9044, "step": 1915 }, { "epoch": 0.16, "learning_rate": 9.540115444692853e-06, "loss": 0.9037, "step": 1916 }, { "epoch": 0.16, "learning_rate": 9.539536369048553e-06, "loss": 0.9804, "step": 1917 }, { "epoch": 0.16, "learning_rate": 9.538956946651816e-06, "loss": 0.9217, "step": 1918 }, { "epoch": 0.16, "learning_rate": 9.538377177546898e-06, "loss": 0.8113, "step": 1919 }, { "epoch": 0.16, "learning_rate": 9.53779706177809e-06, "loss": 0.8905, "step": 1920 }, { "epoch": 0.16, "learning_rate": 9.5372165993897e-06, "loss": 0.9337, "step": 1921 }, { "epoch": 0.16, "learning_rate": 9.53663579042607e-06, "loss": 0.9117, "step": 1922 }, { "epoch": 0.16, "learning_rate": 9.536054634931561e-06, "loss": 0.8716, "step": 1923 }, { "epoch": 0.16, "learning_rate": 9.535473132950569e-06, "loss": 0.907, "step": 1924 }, { "epoch": 0.16, "learning_rate": 9.53489128452751e-06, "loss": 0.9266, "step": 1925 }, { "epoch": 0.16, "learning_rate": 9.534309089706826e-06, "loss": 0.9505, "step": 1926 }, { "epoch": 0.16, "learning_rate": 9.533726548532993e-06, "loss": 0.8284, "step": 1927 }, { "epoch": 0.16, "learning_rate": 9.533143661050504e-06, "loss": 0.8855, "step": 1928 }, { "epoch": 0.16, "learning_rate": 9.532560427303885e-06, "loss": 0.8813, "step": 1929 }, { "epoch": 0.16, "learning_rate": 9.531976847337685e-06, "loss": 0.8754, "step": 1930 }, { "epoch": 0.16, "learning_rate": 9.531392921196484e-06, "loss": 0.9264, "step": 1931 }, { "epoch": 0.16, "learning_rate": 9.530808648924879e-06, "loss": 0.9234, "step": 1932 }, { "epoch": 0.16, "learning_rate": 9.530224030567506e-06, "loss": 0.8783, "step": 1933 }, { "epoch": 0.17, "learning_rate": 9.529639066169019e-06, "loss": 0.9887, "step": 1934 }, { "epoch": 0.17, "learning_rate": 9.529053755774098e-06, "loss": 0.8993, "step": 1935 }, { "epoch": 0.17, "learning_rate": 9.528468099427454e-06, "loss": 0.9195, "step": 1936 }, { "epoch": 0.17, "learning_rate": 9.527882097173823e-06, "loss": 0.9235, "step": 1937 }, { "epoch": 0.17, "learning_rate": 9.527295749057968e-06, "loss": 0.8727, "step": 1938 }, { "epoch": 0.17, "learning_rate": 9.526709055124672e-06, "loss": 0.9554, "step": 1939 }, { "epoch": 0.17, "learning_rate": 9.526122015418755e-06, "loss": 0.8738, "step": 1940 }, { "epoch": 0.17, "learning_rate": 9.525534629985056e-06, "loss": 0.9076, "step": 1941 }, { "epoch": 0.17, "learning_rate": 9.524946898868441e-06, "loss": 0.9201, "step": 1942 }, { "epoch": 0.17, "learning_rate": 9.524358822113805e-06, "loss": 0.9286, "step": 1943 }, { "epoch": 0.17, "learning_rate": 9.52377039976607e-06, "loss": 0.8709, "step": 1944 }, { "epoch": 0.17, "learning_rate": 9.52318163187018e-06, "loss": 0.9013, "step": 1945 }, { "epoch": 0.17, "learning_rate": 9.52259251847111e-06, "loss": 0.8876, "step": 1946 }, { "epoch": 0.17, "learning_rate": 9.522003059613856e-06, "loss": 0.878, "step": 1947 }, { "epoch": 0.17, "learning_rate": 9.521413255343448e-06, "loss": 0.8917, "step": 1948 }, { "epoch": 0.17, "learning_rate": 9.520823105704938e-06, "loss": 0.9221, "step": 1949 }, { "epoch": 0.17, "learning_rate": 9.520232610743401e-06, "loss": 0.8805, "step": 1950 }, { "epoch": 0.17, "learning_rate": 9.519641770503944e-06, "loss": 0.8749, "step": 1951 }, { "epoch": 0.17, "learning_rate": 9.519050585031698e-06, "loss": 0.9505, "step": 1952 }, { "epoch": 0.17, "learning_rate": 9.518459054371822e-06, "loss": 0.8739, "step": 1953 }, { "epoch": 0.17, "learning_rate": 9.5178671785695e-06, "loss": 0.8849, "step": 1954 }, { "epoch": 0.17, "learning_rate": 9.51727495766994e-06, "loss": 0.9164, "step": 1955 }, { "epoch": 0.17, "learning_rate": 9.51668239171838e-06, "loss": 0.9031, "step": 1956 }, { "epoch": 0.17, "learning_rate": 9.516089480760085e-06, "loss": 0.8776, "step": 1957 }, { "epoch": 0.17, "learning_rate": 9.515496224840342e-06, "loss": 0.8862, "step": 1958 }, { "epoch": 0.17, "learning_rate": 9.514902624004468e-06, "loss": 0.9201, "step": 1959 }, { "epoch": 0.17, "learning_rate": 9.514308678297804e-06, "loss": 0.9315, "step": 1960 }, { "epoch": 0.17, "learning_rate": 9.51371438776572e-06, "loss": 0.9209, "step": 1961 }, { "epoch": 0.17, "learning_rate": 9.51311975245361e-06, "loss": 0.8668, "step": 1962 }, { "epoch": 0.17, "learning_rate": 9.512524772406896e-06, "loss": 0.9294, "step": 1963 }, { "epoch": 0.17, "learning_rate": 9.511929447671025e-06, "loss": 0.9106, "step": 1964 }, { "epoch": 0.17, "learning_rate": 9.511333778291472e-06, "loss": 0.9312, "step": 1965 }, { "epoch": 0.17, "learning_rate": 9.510737764313736e-06, "loss": 0.8839, "step": 1966 }, { "epoch": 0.17, "learning_rate": 9.510141405783344e-06, "loss": 0.8826, "step": 1967 }, { "epoch": 0.17, "learning_rate": 9.509544702745846e-06, "loss": 0.8655, "step": 1968 }, { "epoch": 0.17, "learning_rate": 9.508947655246826e-06, "loss": 0.9334, "step": 1969 }, { "epoch": 0.17, "learning_rate": 9.508350263331886e-06, "loss": 0.9075, "step": 1970 }, { "epoch": 0.17, "learning_rate": 9.50775252704666e-06, "loss": 0.8969, "step": 1971 }, { "epoch": 0.17, "learning_rate": 9.507154446436806e-06, "loss": 0.9168, "step": 1972 }, { "epoch": 0.17, "learning_rate": 9.506556021548006e-06, "loss": 0.9144, "step": 1973 }, { "epoch": 0.17, "learning_rate": 9.505957252425972e-06, "loss": 0.9086, "step": 1974 }, { "epoch": 0.17, "learning_rate": 9.505358139116441e-06, "loss": 0.9245, "step": 1975 }, { "epoch": 0.17, "learning_rate": 9.504758681665178e-06, "loss": 0.8722, "step": 1976 }, { "epoch": 0.17, "learning_rate": 9.504158880117969e-06, "loss": 0.9217, "step": 1977 }, { "epoch": 0.17, "learning_rate": 9.503558734520633e-06, "loss": 0.9368, "step": 1978 }, { "epoch": 0.17, "learning_rate": 9.50295824491901e-06, "loss": 0.9435, "step": 1979 }, { "epoch": 0.17, "learning_rate": 9.502357411358969e-06, "loss": 0.9367, "step": 1980 }, { "epoch": 0.17, "learning_rate": 9.501756233886406e-06, "loss": 0.6095, "step": 1981 }, { "epoch": 0.17, "learning_rate": 9.50115471254724e-06, "loss": 0.9275, "step": 1982 }, { "epoch": 0.17, "learning_rate": 9.50055284738742e-06, "loss": 0.9385, "step": 1983 }, { "epoch": 0.17, "learning_rate": 9.499950638452917e-06, "loss": 0.9187, "step": 1984 }, { "epoch": 0.17, "learning_rate": 9.499348085789733e-06, "loss": 0.5905, "step": 1985 }, { "epoch": 0.17, "learning_rate": 9.498745189443895e-06, "loss": 0.9493, "step": 1986 }, { "epoch": 0.17, "learning_rate": 9.49814194946145e-06, "loss": 0.9334, "step": 1987 }, { "epoch": 0.17, "learning_rate": 9.497538365888483e-06, "loss": 0.9179, "step": 1988 }, { "epoch": 0.17, "learning_rate": 9.496934438771093e-06, "loss": 0.8698, "step": 1989 }, { "epoch": 0.17, "learning_rate": 9.496330168155415e-06, "loss": 0.8976, "step": 1990 }, { "epoch": 0.17, "learning_rate": 9.495725554087605e-06, "loss": 0.9409, "step": 1991 }, { "epoch": 0.17, "learning_rate": 9.495120596613844e-06, "loss": 0.8406, "step": 1992 }, { "epoch": 0.17, "learning_rate": 9.494515295780347e-06, "loss": 0.9094, "step": 1993 }, { "epoch": 0.17, "learning_rate": 9.493909651633346e-06, "loss": 0.8616, "step": 1994 }, { "epoch": 0.17, "learning_rate": 9.493303664219105e-06, "loss": 0.8897, "step": 1995 }, { "epoch": 0.17, "learning_rate": 9.492697333583909e-06, "loss": 0.8735, "step": 1996 }, { "epoch": 0.17, "learning_rate": 9.492090659774076e-06, "loss": 0.8487, "step": 1997 }, { "epoch": 0.17, "learning_rate": 9.491483642835946e-06, "loss": 0.9186, "step": 1998 }, { "epoch": 0.17, "learning_rate": 9.490876282815884e-06, "loss": 0.8841, "step": 1999 }, { "epoch": 0.17, "learning_rate": 9.490268579760285e-06, "loss": 0.9814, "step": 2000 }, { "epoch": 0.17, "learning_rate": 9.48966053371557e-06, "loss": 0.9274, "step": 2001 }, { "epoch": 0.17, "learning_rate": 9.489052144728182e-06, "loss": 0.8616, "step": 2002 }, { "epoch": 0.17, "learning_rate": 9.488443412844592e-06, "loss": 0.9351, "step": 2003 }, { "epoch": 0.17, "learning_rate": 9.4878343381113e-06, "loss": 0.9406, "step": 2004 }, { "epoch": 0.17, "learning_rate": 9.487224920574831e-06, "loss": 0.9021, "step": 2005 }, { "epoch": 0.17, "learning_rate": 9.486615160281735e-06, "loss": 0.9305, "step": 2006 }, { "epoch": 0.17, "learning_rate": 9.486005057278586e-06, "loss": 0.9202, "step": 2007 }, { "epoch": 0.17, "learning_rate": 9.485394611611989e-06, "loss": 0.8822, "step": 2008 }, { "epoch": 0.17, "learning_rate": 9.48478382332857e-06, "loss": 0.8915, "step": 2009 }, { "epoch": 0.17, "learning_rate": 9.48417269247499e-06, "loss": 0.8817, "step": 2010 }, { "epoch": 0.17, "learning_rate": 9.483561219097923e-06, "loss": 0.901, "step": 2011 }, { "epoch": 0.17, "learning_rate": 9.482949403244081e-06, "loss": 0.9034, "step": 2012 }, { "epoch": 0.17, "learning_rate": 9.482337244960197e-06, "loss": 0.8782, "step": 2013 }, { "epoch": 0.17, "learning_rate": 9.48172474429303e-06, "loss": 0.8821, "step": 2014 }, { "epoch": 0.17, "learning_rate": 9.481111901289366e-06, "loss": 0.9472, "step": 2015 }, { "epoch": 0.17, "learning_rate": 9.480498715996017e-06, "loss": 0.9431, "step": 2016 }, { "epoch": 0.17, "learning_rate": 9.47988518845982e-06, "loss": 0.9033, "step": 2017 }, { "epoch": 0.17, "learning_rate": 9.47927131872764e-06, "loss": 0.863, "step": 2018 }, { "epoch": 0.17, "learning_rate": 9.47865710684637e-06, "loss": 0.8794, "step": 2019 }, { "epoch": 0.17, "learning_rate": 9.478042552862922e-06, "loss": 0.8958, "step": 2020 }, { "epoch": 0.17, "learning_rate": 9.477427656824244e-06, "loss": 0.8818, "step": 2021 }, { "epoch": 0.17, "learning_rate": 9.4768124187773e-06, "loss": 0.9069, "step": 2022 }, { "epoch": 0.17, "learning_rate": 9.476196838769087e-06, "loss": 0.8912, "step": 2023 }, { "epoch": 0.17, "learning_rate": 9.475580916846627e-06, "loss": 0.8917, "step": 2024 }, { "epoch": 0.17, "learning_rate": 9.474964653056966e-06, "loss": 0.8992, "step": 2025 }, { "epoch": 0.17, "learning_rate": 9.474348047447177e-06, "loss": 0.8538, "step": 2026 }, { "epoch": 0.17, "learning_rate": 9.47373110006436e-06, "loss": 0.9033, "step": 2027 }, { "epoch": 0.17, "learning_rate": 9.47311381095564e-06, "loss": 0.8843, "step": 2028 }, { "epoch": 0.17, "learning_rate": 9.472496180168169e-06, "loss": 0.8531, "step": 2029 }, { "epoch": 0.17, "learning_rate": 9.471878207749126e-06, "loss": 0.8705, "step": 2030 }, { "epoch": 0.17, "learning_rate": 9.471259893745711e-06, "loss": 0.9089, "step": 2031 }, { "epoch": 0.17, "learning_rate": 9.470641238205158e-06, "loss": 0.9139, "step": 2032 }, { "epoch": 0.17, "learning_rate": 9.470022241174721e-06, "loss": 0.8874, "step": 2033 }, { "epoch": 0.17, "learning_rate": 9.469402902701684e-06, "loss": 0.9462, "step": 2034 }, { "epoch": 0.17, "learning_rate": 9.468783222833354e-06, "loss": 0.8817, "step": 2035 }, { "epoch": 0.17, "learning_rate": 9.468163201617063e-06, "loss": 0.9156, "step": 2036 }, { "epoch": 0.17, "learning_rate": 9.467542839100172e-06, "loss": 0.8927, "step": 2037 }, { "epoch": 0.17, "learning_rate": 9.466922135330072e-06, "loss": 0.8701, "step": 2038 }, { "epoch": 0.17, "learning_rate": 9.46630109035417e-06, "loss": 0.8959, "step": 2039 }, { "epoch": 0.17, "learning_rate": 9.465679704219908e-06, "loss": 0.9248, "step": 2040 }, { "epoch": 0.17, "learning_rate": 9.465057976974748e-06, "loss": 0.8973, "step": 2041 }, { "epoch": 0.17, "learning_rate": 9.464435908666183e-06, "loss": 0.8788, "step": 2042 }, { "epoch": 0.17, "learning_rate": 9.463813499341726e-06, "loss": 0.9408, "step": 2043 }, { "epoch": 0.17, "learning_rate": 9.463190749048925e-06, "loss": 0.909, "step": 2044 }, { "epoch": 0.17, "learning_rate": 9.462567657835344e-06, "loss": 0.9238, "step": 2045 }, { "epoch": 0.17, "learning_rate": 9.46194422574858e-06, "loss": 0.893, "step": 2046 }, { "epoch": 0.17, "learning_rate": 9.461320452836255e-06, "loss": 0.8812, "step": 2047 }, { "epoch": 0.17, "learning_rate": 9.460696339146014e-06, "loss": 0.9041, "step": 2048 }, { "epoch": 0.17, "learning_rate": 9.46007188472553e-06, "loss": 0.9101, "step": 2049 }, { "epoch": 0.17, "learning_rate": 9.459447089622503e-06, "loss": 0.9156, "step": 2050 }, { "epoch": 0.18, "learning_rate": 9.458821953884658e-06, "loss": 0.8906, "step": 2051 }, { "epoch": 0.18, "learning_rate": 9.458196477559745e-06, "loss": 0.889, "step": 2052 }, { "epoch": 0.18, "learning_rate": 9.457570660695542e-06, "loss": 0.9016, "step": 2053 }, { "epoch": 0.18, "learning_rate": 9.456944503339848e-06, "loss": 0.9035, "step": 2054 }, { "epoch": 0.18, "learning_rate": 9.456318005540497e-06, "loss": 0.8593, "step": 2055 }, { "epoch": 0.18, "learning_rate": 9.455691167345344e-06, "loss": 0.9025, "step": 2056 }, { "epoch": 0.18, "learning_rate": 9.455063988802267e-06, "loss": 0.9224, "step": 2057 }, { "epoch": 0.18, "learning_rate": 9.454436469959176e-06, "loss": 0.8893, "step": 2058 }, { "epoch": 0.18, "learning_rate": 9.453808610864001e-06, "loss": 0.8888, "step": 2059 }, { "epoch": 0.18, "learning_rate": 9.453180411564705e-06, "loss": 0.8286, "step": 2060 }, { "epoch": 0.18, "learning_rate": 9.452551872109267e-06, "loss": 0.8743, "step": 2061 }, { "epoch": 0.18, "learning_rate": 9.451922992545703e-06, "loss": 0.6635, "step": 2062 }, { "epoch": 0.18, "learning_rate": 9.451293772922048e-06, "loss": 0.9131, "step": 2063 }, { "epoch": 0.18, "learning_rate": 9.450664213286366e-06, "loss": 0.9074, "step": 2064 }, { "epoch": 0.18, "learning_rate": 9.450034313686745e-06, "loss": 0.8511, "step": 2065 }, { "epoch": 0.18, "learning_rate": 9.449404074171301e-06, "loss": 0.8874, "step": 2066 }, { "epoch": 0.18, "learning_rate": 9.448773494788173e-06, "loss": 0.9109, "step": 2067 }, { "epoch": 0.18, "learning_rate": 9.44814257558553e-06, "loss": 0.8932, "step": 2068 }, { "epoch": 0.18, "learning_rate": 9.447511316611563e-06, "loss": 0.9, "step": 2069 }, { "epoch": 0.18, "learning_rate": 9.44687971791449e-06, "loss": 0.9606, "step": 2070 }, { "epoch": 0.18, "learning_rate": 9.44624777954256e-06, "loss": 0.904, "step": 2071 }, { "epoch": 0.18, "learning_rate": 9.445615501544037e-06, "loss": 0.968, "step": 2072 }, { "epoch": 0.18, "learning_rate": 9.444982883967222e-06, "loss": 0.9024, "step": 2073 }, { "epoch": 0.18, "learning_rate": 9.444349926860439e-06, "loss": 0.8779, "step": 2074 }, { "epoch": 0.18, "learning_rate": 9.44371663027203e-06, "loss": 0.9231, "step": 2075 }, { "epoch": 0.18, "learning_rate": 9.443082994250377e-06, "loss": 0.909, "step": 2076 }, { "epoch": 0.18, "learning_rate": 9.442449018843875e-06, "loss": 0.9173, "step": 2077 }, { "epoch": 0.18, "learning_rate": 9.441814704100952e-06, "loss": 0.5608, "step": 2078 }, { "epoch": 0.18, "learning_rate": 9.44118005007006e-06, "loss": 0.8752, "step": 2079 }, { "epoch": 0.18, "learning_rate": 9.440545056799677e-06, "loss": 0.662, "step": 2080 }, { "epoch": 0.18, "learning_rate": 9.439909724338306e-06, "loss": 0.9183, "step": 2081 }, { "epoch": 0.18, "learning_rate": 9.439274052734479e-06, "loss": 0.8953, "step": 2082 }, { "epoch": 0.18, "learning_rate": 9.43863804203675e-06, "loss": 0.9144, "step": 2083 }, { "epoch": 0.18, "learning_rate": 9.438001692293702e-06, "loss": 0.9161, "step": 2084 }, { "epoch": 0.18, "learning_rate": 9.437365003553942e-06, "loss": 0.9121, "step": 2085 }, { "epoch": 0.18, "learning_rate": 9.436727975866102e-06, "loss": 0.8957, "step": 2086 }, { "epoch": 0.18, "learning_rate": 9.436090609278843e-06, "loss": 0.901, "step": 2087 }, { "epoch": 0.18, "learning_rate": 9.43545290384085e-06, "loss": 0.8892, "step": 2088 }, { "epoch": 0.18, "learning_rate": 9.434814859600834e-06, "loss": 0.9186, "step": 2089 }, { "epoch": 0.18, "learning_rate": 9.434176476607533e-06, "loss": 0.93, "step": 2090 }, { "epoch": 0.18, "learning_rate": 9.433537754909707e-06, "loss": 0.9119, "step": 2091 }, { "epoch": 0.18, "learning_rate": 9.432898694556148e-06, "loss": 0.8663, "step": 2092 }, { "epoch": 0.18, "learning_rate": 9.43225929559567e-06, "loss": 0.9322, "step": 2093 }, { "epoch": 0.18, "learning_rate": 9.43161955807711e-06, "loss": 0.9021, "step": 2094 }, { "epoch": 0.18, "learning_rate": 9.43097948204934e-06, "loss": 0.8539, "step": 2095 }, { "epoch": 0.18, "learning_rate": 9.430339067561247e-06, "loss": 0.8471, "step": 2096 }, { "epoch": 0.18, "learning_rate": 9.429698314661752e-06, "loss": 0.8966, "step": 2097 }, { "epoch": 0.18, "learning_rate": 9.4290572233998e-06, "loss": 0.8771, "step": 2098 }, { "epoch": 0.18, "learning_rate": 9.428415793824357e-06, "loss": 0.9, "step": 2099 }, { "epoch": 0.18, "learning_rate": 9.427774025984422e-06, "loss": 0.9075, "step": 2100 }, { "epoch": 0.18, "learning_rate": 9.427131919929014e-06, "loss": 0.8771, "step": 2101 }, { "epoch": 0.18, "learning_rate": 9.426489475707181e-06, "loss": 0.9427, "step": 2102 }, { "epoch": 0.18, "learning_rate": 9.425846693367997e-06, "loss": 0.9172, "step": 2103 }, { "epoch": 0.18, "learning_rate": 9.425203572960559e-06, "loss": 0.6549, "step": 2104 }, { "epoch": 0.18, "learning_rate": 9.424560114533996e-06, "loss": 0.8705, "step": 2105 }, { "epoch": 0.18, "learning_rate": 9.423916318137454e-06, "loss": 0.9135, "step": 2106 }, { "epoch": 0.18, "learning_rate": 9.423272183820109e-06, "loss": 0.881, "step": 2107 }, { "epoch": 0.18, "learning_rate": 9.422627711631167e-06, "loss": 0.9584, "step": 2108 }, { "epoch": 0.18, "learning_rate": 9.421982901619855e-06, "loss": 0.9246, "step": 2109 }, { "epoch": 0.18, "learning_rate": 9.421337753835425e-06, "loss": 0.872, "step": 2110 }, { "epoch": 0.18, "learning_rate": 9.420692268327157e-06, "loss": 0.9538, "step": 2111 }, { "epoch": 0.18, "learning_rate": 9.420046445144358e-06, "loss": 0.8834, "step": 2112 }, { "epoch": 0.18, "learning_rate": 9.419400284336358e-06, "loss": 0.8902, "step": 2113 }, { "epoch": 0.18, "learning_rate": 9.418753785952514e-06, "loss": 0.8949, "step": 2114 }, { "epoch": 0.18, "learning_rate": 9.41810695004221e-06, "loss": 0.9065, "step": 2115 }, { "epoch": 0.18, "learning_rate": 9.417459776654852e-06, "loss": 0.9078, "step": 2116 }, { "epoch": 0.18, "learning_rate": 9.416812265839876e-06, "loss": 0.9193, "step": 2117 }, { "epoch": 0.18, "learning_rate": 9.416164417646742e-06, "loss": 0.8799, "step": 2118 }, { "epoch": 0.18, "learning_rate": 9.415516232124936e-06, "loss": 0.8865, "step": 2119 }, { "epoch": 0.18, "learning_rate": 9.414867709323969e-06, "loss": 0.6014, "step": 2120 }, { "epoch": 0.18, "learning_rate": 9.41421884929338e-06, "loss": 0.9068, "step": 2121 }, { "epoch": 0.18, "learning_rate": 9.41356965208273e-06, "loss": 0.8893, "step": 2122 }, { "epoch": 0.18, "learning_rate": 9.412920117741609e-06, "loss": 0.8636, "step": 2123 }, { "epoch": 0.18, "learning_rate": 9.412270246319633e-06, "loss": 0.8953, "step": 2124 }, { "epoch": 0.18, "learning_rate": 9.41162003786644e-06, "loss": 0.8465, "step": 2125 }, { "epoch": 0.18, "learning_rate": 9.410969492431696e-06, "loss": 0.8866, "step": 2126 }, { "epoch": 0.18, "learning_rate": 9.410318610065097e-06, "loss": 0.9119, "step": 2127 }, { "epoch": 0.18, "learning_rate": 9.409667390816355e-06, "loss": 0.8551, "step": 2128 }, { "epoch": 0.18, "learning_rate": 9.409015834735218e-06, "loss": 0.8832, "step": 2129 }, { "epoch": 0.18, "learning_rate": 9.408363941871452e-06, "loss": 0.9187, "step": 2130 }, { "epoch": 0.18, "learning_rate": 9.407711712274854e-06, "loss": 0.912, "step": 2131 }, { "epoch": 0.18, "learning_rate": 9.407059145995244e-06, "loss": 0.9428, "step": 2132 }, { "epoch": 0.18, "learning_rate": 9.406406243082467e-06, "loss": 0.8632, "step": 2133 }, { "epoch": 0.18, "learning_rate": 9.405753003586396e-06, "loss": 0.8607, "step": 2134 }, { "epoch": 0.18, "learning_rate": 9.405099427556927e-06, "loss": 0.6687, "step": 2135 }, { "epoch": 0.18, "learning_rate": 9.404445515043987e-06, "loss": 0.8993, "step": 2136 }, { "epoch": 0.18, "learning_rate": 9.403791266097523e-06, "loss": 0.8093, "step": 2137 }, { "epoch": 0.18, "learning_rate": 9.403136680767508e-06, "loss": 0.9138, "step": 2138 }, { "epoch": 0.18, "learning_rate": 9.402481759103948e-06, "loss": 0.9086, "step": 2139 }, { "epoch": 0.18, "learning_rate": 9.401826501156862e-06, "loss": 0.9434, "step": 2140 }, { "epoch": 0.18, "learning_rate": 9.401170906976307e-06, "loss": 0.9031, "step": 2141 }, { "epoch": 0.18, "learning_rate": 9.400514976612358e-06, "loss": 0.9054, "step": 2142 }, { "epoch": 0.18, "learning_rate": 9.399858710115121e-06, "loss": 0.9089, "step": 2143 }, { "epoch": 0.18, "learning_rate": 9.399202107534722e-06, "loss": 0.8911, "step": 2144 }, { "epoch": 0.18, "learning_rate": 9.398545168921315e-06, "loss": 0.9225, "step": 2145 }, { "epoch": 0.18, "learning_rate": 9.397887894325087e-06, "loss": 0.9484, "step": 2146 }, { "epoch": 0.18, "learning_rate": 9.397230283796235e-06, "loss": 0.9151, "step": 2147 }, { "epoch": 0.18, "learning_rate": 9.396572337384994e-06, "loss": 0.9418, "step": 2148 }, { "epoch": 0.18, "learning_rate": 9.395914055141622e-06, "loss": 0.8979, "step": 2149 }, { "epoch": 0.18, "learning_rate": 9.395255437116402e-06, "loss": 0.8901, "step": 2150 }, { "epoch": 0.18, "learning_rate": 9.394596483359643e-06, "loss": 0.9421, "step": 2151 }, { "epoch": 0.18, "learning_rate": 9.393937193921679e-06, "loss": 0.9206, "step": 2152 }, { "epoch": 0.18, "learning_rate": 9.393277568852866e-06, "loss": 0.8877, "step": 2153 }, { "epoch": 0.18, "learning_rate": 9.392617608203595e-06, "loss": 0.924, "step": 2154 }, { "epoch": 0.18, "learning_rate": 9.391957312024273e-06, "loss": 0.8934, "step": 2155 }, { "epoch": 0.18, "learning_rate": 9.391296680365339e-06, "loss": 0.8278, "step": 2156 }, { "epoch": 0.18, "learning_rate": 9.390635713277255e-06, "loss": 0.8988, "step": 2157 }, { "epoch": 0.18, "learning_rate": 9.389974410810509e-06, "loss": 0.9086, "step": 2158 }, { "epoch": 0.18, "learning_rate": 9.389312773015613e-06, "loss": 0.9401, "step": 2159 }, { "epoch": 0.18, "learning_rate": 9.388650799943108e-06, "loss": 0.8957, "step": 2160 }, { "epoch": 0.18, "learning_rate": 9.387988491643558e-06, "loss": 0.9037, "step": 2161 }, { "epoch": 0.18, "learning_rate": 9.387325848167554e-06, "loss": 0.913, "step": 2162 }, { "epoch": 0.18, "learning_rate": 9.386662869565711e-06, "loss": 0.898, "step": 2163 }, { "epoch": 0.18, "learning_rate": 9.385999555888672e-06, "loss": 0.9014, "step": 2164 }, { "epoch": 0.18, "learning_rate": 9.385335907187103e-06, "loss": 0.9149, "step": 2165 }, { "epoch": 0.18, "learning_rate": 9.384671923511699e-06, "loss": 0.8752, "step": 2166 }, { "epoch": 0.18, "learning_rate": 9.384007604913173e-06, "loss": 0.8989, "step": 2167 }, { "epoch": 0.18, "learning_rate": 9.383342951442276e-06, "loss": 0.907, "step": 2168 }, { "epoch": 0.19, "learning_rate": 9.382677963149773e-06, "loss": 0.9078, "step": 2169 }, { "epoch": 0.19, "learning_rate": 9.38201264008646e-06, "loss": 0.9158, "step": 2170 }, { "epoch": 0.19, "learning_rate": 9.381346982303157e-06, "loss": 0.6175, "step": 2171 }, { "epoch": 0.19, "learning_rate": 9.380680989850712e-06, "loss": 0.9351, "step": 2172 }, { "epoch": 0.19, "learning_rate": 9.380014662779996e-06, "loss": 0.9113, "step": 2173 }, { "epoch": 0.19, "learning_rate": 9.379348001141909e-06, "loss": 0.8697, "step": 2174 }, { "epoch": 0.19, "learning_rate": 9.37868100498737e-06, "loss": 0.8879, "step": 2175 }, { "epoch": 0.19, "learning_rate": 9.378013674367327e-06, "loss": 0.8711, "step": 2176 }, { "epoch": 0.19, "learning_rate": 9.377346009332758e-06, "loss": 0.8513, "step": 2177 }, { "epoch": 0.19, "learning_rate": 9.376678009934659e-06, "loss": 0.8884, "step": 2178 }, { "epoch": 0.19, "learning_rate": 9.376009676224057e-06, "loss": 0.8753, "step": 2179 }, { "epoch": 0.19, "learning_rate": 9.375341008252002e-06, "loss": 0.8734, "step": 2180 }, { "epoch": 0.19, "learning_rate": 9.374672006069572e-06, "loss": 0.9028, "step": 2181 }, { "epoch": 0.19, "learning_rate": 9.374002669727866e-06, "loss": 0.9578, "step": 2182 }, { "epoch": 0.19, "learning_rate": 9.373332999278014e-06, "loss": 0.8216, "step": 2183 }, { "epoch": 0.19, "learning_rate": 9.372662994771164e-06, "loss": 0.8591, "step": 2184 }, { "epoch": 0.19, "learning_rate": 9.3719926562585e-06, "loss": 0.8819, "step": 2185 }, { "epoch": 0.19, "learning_rate": 9.371321983791225e-06, "loss": 0.9344, "step": 2186 }, { "epoch": 0.19, "learning_rate": 9.370650977420564e-06, "loss": 0.9426, "step": 2187 }, { "epoch": 0.19, "learning_rate": 9.369979637197774e-06, "loss": 0.8878, "step": 2188 }, { "epoch": 0.19, "learning_rate": 9.369307963174137e-06, "loss": 0.874, "step": 2189 }, { "epoch": 0.19, "learning_rate": 9.368635955400958e-06, "loss": 0.8659, "step": 2190 }, { "epoch": 0.19, "learning_rate": 9.367963613929568e-06, "loss": 0.8471, "step": 2191 }, { "epoch": 0.19, "learning_rate": 9.367290938811322e-06, "loss": 0.9194, "step": 2192 }, { "epoch": 0.19, "learning_rate": 9.366617930097605e-06, "loss": 0.943, "step": 2193 }, { "epoch": 0.19, "learning_rate": 9.365944587839823e-06, "loss": 0.8605, "step": 2194 }, { "epoch": 0.19, "learning_rate": 9.365270912089412e-06, "loss": 0.9318, "step": 2195 }, { "epoch": 0.19, "learning_rate": 9.364596902897827e-06, "loss": 0.8733, "step": 2196 }, { "epoch": 0.19, "learning_rate": 9.363922560316554e-06, "loss": 0.9, "step": 2197 }, { "epoch": 0.19, "learning_rate": 9.363247884397103e-06, "loss": 0.8544, "step": 2198 }, { "epoch": 0.19, "learning_rate": 9.36257287519101e-06, "loss": 0.8283, "step": 2199 }, { "epoch": 0.19, "learning_rate": 9.361897532749832e-06, "loss": 0.8469, "step": 2200 }, { "epoch": 0.19, "learning_rate": 9.361221857125159e-06, "loss": 0.8261, "step": 2201 }, { "epoch": 0.19, "learning_rate": 9.360545848368599e-06, "loss": 0.9066, "step": 2202 }, { "epoch": 0.19, "learning_rate": 9.35986950653179e-06, "loss": 0.8427, "step": 2203 }, { "epoch": 0.19, "learning_rate": 9.359192831666397e-06, "loss": 0.8848, "step": 2204 }, { "epoch": 0.19, "learning_rate": 9.358515823824105e-06, "loss": 0.8895, "step": 2205 }, { "epoch": 0.19, "learning_rate": 9.357838483056628e-06, "loss": 0.955, "step": 2206 }, { "epoch": 0.19, "learning_rate": 9.357160809415706e-06, "loss": 0.9254, "step": 2207 }, { "epoch": 0.19, "learning_rate": 9.3564828029531e-06, "loss": 0.6006, "step": 2208 }, { "epoch": 0.19, "learning_rate": 9.355804463720601e-06, "loss": 0.9106, "step": 2209 }, { "epoch": 0.19, "learning_rate": 9.355125791770027e-06, "loss": 0.949, "step": 2210 }, { "epoch": 0.19, "learning_rate": 9.354446787153212e-06, "loss": 0.9536, "step": 2211 }, { "epoch": 0.19, "learning_rate": 9.35376744992203e-06, "loss": 0.8984, "step": 2212 }, { "epoch": 0.19, "learning_rate": 9.353087780128364e-06, "loss": 0.8932, "step": 2213 }, { "epoch": 0.19, "learning_rate": 9.352407777824134e-06, "loss": 0.9178, "step": 2214 }, { "epoch": 0.19, "learning_rate": 9.351727443061284e-06, "loss": 0.8676, "step": 2215 }, { "epoch": 0.19, "learning_rate": 9.351046775891778e-06, "loss": 0.9314, "step": 2216 }, { "epoch": 0.19, "learning_rate": 9.35036577636761e-06, "loss": 0.9087, "step": 2217 }, { "epoch": 0.19, "learning_rate": 9.3496844445408e-06, "loss": 0.9476, "step": 2218 }, { "epoch": 0.19, "learning_rate": 9.349002780463388e-06, "loss": 0.9464, "step": 2219 }, { "epoch": 0.19, "learning_rate": 9.348320784187446e-06, "loss": 0.9084, "step": 2220 }, { "epoch": 0.19, "learning_rate": 9.347638455765067e-06, "loss": 0.8961, "step": 2221 }, { "epoch": 0.19, "learning_rate": 9.346955795248369e-06, "loss": 0.8871, "step": 2222 }, { "epoch": 0.19, "learning_rate": 9.3462728026895e-06, "loss": 0.9107, "step": 2223 }, { "epoch": 0.19, "learning_rate": 9.34558947814063e-06, "loss": 0.918, "step": 2224 }, { "epoch": 0.19, "learning_rate": 9.344905821653954e-06, "loss": 0.8602, "step": 2225 }, { "epoch": 0.19, "learning_rate": 9.344221833281691e-06, "loss": 0.9036, "step": 2226 }, { "epoch": 0.19, "learning_rate": 9.34353751307609e-06, "loss": 0.932, "step": 2227 }, { "epoch": 0.19, "learning_rate": 9.342852861089422e-06, "loss": 0.9296, "step": 2228 }, { "epoch": 0.19, "learning_rate": 9.342167877373984e-06, "loss": 0.8724, "step": 2229 }, { "epoch": 0.19, "learning_rate": 9.3414825619821e-06, "loss": 0.8917, "step": 2230 }, { "epoch": 0.19, "learning_rate": 9.340796914966116e-06, "loss": 0.8677, "step": 2231 }, { "epoch": 0.19, "learning_rate": 9.340110936378406e-06, "loss": 0.8761, "step": 2232 }, { "epoch": 0.19, "learning_rate": 9.339424626271367e-06, "loss": 0.8819, "step": 2233 }, { "epoch": 0.19, "learning_rate": 9.338737984697426e-06, "loss": 0.8507, "step": 2234 }, { "epoch": 0.19, "learning_rate": 9.338051011709028e-06, "loss": 0.8807, "step": 2235 }, { "epoch": 0.19, "learning_rate": 9.337363707358651e-06, "loss": 0.921, "step": 2236 }, { "epoch": 0.19, "learning_rate": 9.336676071698792e-06, "loss": 0.9042, "step": 2237 }, { "epoch": 0.19, "learning_rate": 9.33598810478198e-06, "loss": 0.9367, "step": 2238 }, { "epoch": 0.19, "learning_rate": 9.335299806660759e-06, "loss": 0.9205, "step": 2239 }, { "epoch": 0.19, "learning_rate": 9.334611177387709e-06, "loss": 0.6169, "step": 2240 }, { "epoch": 0.19, "learning_rate": 9.333922217015432e-06, "loss": 0.8862, "step": 2241 }, { "epoch": 0.19, "learning_rate": 9.33323292559655e-06, "loss": 0.8906, "step": 2242 }, { "epoch": 0.19, "learning_rate": 9.332543303183719e-06, "loss": 0.8854, "step": 2243 }, { "epoch": 0.19, "learning_rate": 9.331853349829614e-06, "loss": 0.9223, "step": 2244 }, { "epoch": 0.19, "learning_rate": 9.331163065586935e-06, "loss": 0.8322, "step": 2245 }, { "epoch": 0.19, "learning_rate": 9.330472450508413e-06, "loss": 0.9299, "step": 2246 }, { "epoch": 0.19, "learning_rate": 9.329781504646798e-06, "loss": 0.6705, "step": 2247 }, { "epoch": 0.19, "learning_rate": 9.329090228054869e-06, "loss": 0.9057, "step": 2248 }, { "epoch": 0.19, "learning_rate": 9.328398620785429e-06, "loss": 0.9213, "step": 2249 }, { "epoch": 0.19, "learning_rate": 9.327706682891305e-06, "loss": 0.8551, "step": 2250 }, { "epoch": 0.19, "learning_rate": 9.327014414425353e-06, "loss": 0.8828, "step": 2251 }, { "epoch": 0.19, "learning_rate": 9.326321815440451e-06, "loss": 0.885, "step": 2252 }, { "epoch": 0.19, "learning_rate": 9.325628885989503e-06, "loss": 0.9011, "step": 2253 }, { "epoch": 0.19, "learning_rate": 9.324935626125437e-06, "loss": 0.8856, "step": 2254 }, { "epoch": 0.19, "learning_rate": 9.324242035901211e-06, "loss": 0.9061, "step": 2255 }, { "epoch": 0.19, "learning_rate": 9.323548115369803e-06, "loss": 0.5575, "step": 2256 }, { "epoch": 0.19, "learning_rate": 9.322853864584215e-06, "loss": 0.8853, "step": 2257 }, { "epoch": 0.19, "learning_rate": 9.322159283597484e-06, "loss": 0.6711, "step": 2258 }, { "epoch": 0.19, "learning_rate": 9.32146437246266e-06, "loss": 0.9557, "step": 2259 }, { "epoch": 0.19, "learning_rate": 9.320769131232828e-06, "loss": 0.868, "step": 2260 }, { "epoch": 0.19, "learning_rate": 9.32007355996109e-06, "loss": 0.9127, "step": 2261 }, { "epoch": 0.19, "learning_rate": 9.319377658700579e-06, "loss": 0.9193, "step": 2262 }, { "epoch": 0.19, "learning_rate": 9.318681427504451e-06, "loss": 0.8938, "step": 2263 }, { "epoch": 0.19, "learning_rate": 9.31798486642589e-06, "loss": 0.9314, "step": 2264 }, { "epoch": 0.19, "learning_rate": 9.317287975518099e-06, "loss": 0.8496, "step": 2265 }, { "epoch": 0.19, "learning_rate": 9.316590754834311e-06, "loss": 0.9209, "step": 2266 }, { "epoch": 0.19, "learning_rate": 9.315893204427786e-06, "loss": 0.5839, "step": 2267 }, { "epoch": 0.19, "learning_rate": 9.315195324351802e-06, "loss": 0.8852, "step": 2268 }, { "epoch": 0.19, "learning_rate": 9.31449711465967e-06, "loss": 0.8613, "step": 2269 }, { "epoch": 0.19, "learning_rate": 9.313798575404723e-06, "loss": 0.8481, "step": 2270 }, { "epoch": 0.19, "learning_rate": 9.313099706640317e-06, "loss": 0.8421, "step": 2271 }, { "epoch": 0.19, "learning_rate": 9.312400508419835e-06, "loss": 0.8935, "step": 2272 }, { "epoch": 0.19, "learning_rate": 9.311700980796688e-06, "loss": 0.8822, "step": 2273 }, { "epoch": 0.19, "learning_rate": 9.311001123824305e-06, "loss": 0.8971, "step": 2274 }, { "epoch": 0.19, "learning_rate": 9.310300937556147e-06, "loss": 0.9304, "step": 2275 }, { "epoch": 0.19, "learning_rate": 9.309600422045699e-06, "loss": 0.9168, "step": 2276 }, { "epoch": 0.19, "learning_rate": 9.308899577346467e-06, "loss": 0.8946, "step": 2277 }, { "epoch": 0.19, "learning_rate": 9.308198403511988e-06, "loss": 0.8939, "step": 2278 }, { "epoch": 0.19, "learning_rate": 9.307496900595819e-06, "loss": 0.9519, "step": 2279 }, { "epoch": 0.19, "learning_rate": 9.306795068651544e-06, "loss": 0.9744, "step": 2280 }, { "epoch": 0.19, "learning_rate": 9.306092907732774e-06, "loss": 0.8592, "step": 2281 }, { "epoch": 0.19, "learning_rate": 9.305390417893144e-06, "loss": 0.9065, "step": 2282 }, { "epoch": 0.19, "learning_rate": 9.304687599186311e-06, "loss": 0.9544, "step": 2283 }, { "epoch": 0.19, "learning_rate": 9.30398445166596e-06, "loss": 0.9604, "step": 2284 }, { "epoch": 0.19, "learning_rate": 9.303280975385806e-06, "loss": 0.8834, "step": 2285 }, { "epoch": 0.2, "learning_rate": 9.302577170399578e-06, "loss": 0.8953, "step": 2286 }, { "epoch": 0.2, "learning_rate": 9.30187303676104e-06, "loss": 0.9178, "step": 2287 }, { "epoch": 0.2, "learning_rate": 9.301168574523975e-06, "loss": 0.8936, "step": 2288 }, { "epoch": 0.2, "learning_rate": 9.300463783742194e-06, "loss": 0.8965, "step": 2289 }, { "epoch": 0.2, "learning_rate": 9.299758664469534e-06, "loss": 0.9091, "step": 2290 }, { "epoch": 0.2, "learning_rate": 9.299053216759853e-06, "loss": 0.9064, "step": 2291 }, { "epoch": 0.2, "learning_rate": 9.298347440667036e-06, "loss": 0.6174, "step": 2292 }, { "epoch": 0.2, "learning_rate": 9.297641336244996e-06, "loss": 0.9082, "step": 2293 }, { "epoch": 0.2, "learning_rate": 9.29693490354767e-06, "loss": 0.9053, "step": 2294 }, { "epoch": 0.2, "learning_rate": 9.296228142629016e-06, "loss": 0.9375, "step": 2295 }, { "epoch": 0.2, "learning_rate": 9.29552105354302e-06, "loss": 0.8628, "step": 2296 }, { "epoch": 0.2, "learning_rate": 9.294813636343693e-06, "loss": 0.9075, "step": 2297 }, { "epoch": 0.2, "learning_rate": 9.294105891085074e-06, "loss": 0.8968, "step": 2298 }, { "epoch": 0.2, "learning_rate": 9.293397817821222e-06, "loss": 0.9082, "step": 2299 }, { "epoch": 0.2, "learning_rate": 9.292689416606223e-06, "loss": 0.8742, "step": 2300 }, { "epoch": 0.2, "learning_rate": 9.291980687494189e-06, "loss": 0.9134, "step": 2301 }, { "epoch": 0.2, "learning_rate": 9.291271630539255e-06, "loss": 0.8533, "step": 2302 }, { "epoch": 0.2, "learning_rate": 9.290562245795583e-06, "loss": 0.8469, "step": 2303 }, { "epoch": 0.2, "learning_rate": 9.289852533317359e-06, "loss": 0.8505, "step": 2304 }, { "epoch": 0.2, "learning_rate": 9.289142493158796e-06, "loss": 0.931, "step": 2305 }, { "epoch": 0.2, "learning_rate": 9.288432125374127e-06, "loss": 0.9094, "step": 2306 }, { "epoch": 0.2, "learning_rate": 9.287721430017617e-06, "loss": 0.861, "step": 2307 }, { "epoch": 0.2, "learning_rate": 9.287010407143552e-06, "loss": 0.5831, "step": 2308 }, { "epoch": 0.2, "learning_rate": 9.286299056806241e-06, "loss": 0.8996, "step": 2309 }, { "epoch": 0.2, "learning_rate": 9.285587379060023e-06, "loss": 0.9237, "step": 2310 }, { "epoch": 0.2, "learning_rate": 9.284875373959259e-06, "loss": 0.8868, "step": 2311 }, { "epoch": 0.2, "learning_rate": 9.284163041558334e-06, "loss": 0.857, "step": 2312 }, { "epoch": 0.2, "learning_rate": 9.28345038191166e-06, "loss": 0.8812, "step": 2313 }, { "epoch": 0.2, "learning_rate": 9.282737395073676e-06, "loss": 0.9316, "step": 2314 }, { "epoch": 0.2, "learning_rate": 9.28202408109884e-06, "loss": 0.8575, "step": 2315 }, { "epoch": 0.2, "learning_rate": 9.28131044004164e-06, "loss": 0.8848, "step": 2316 }, { "epoch": 0.2, "learning_rate": 9.280596471956587e-06, "loss": 0.8699, "step": 2317 }, { "epoch": 0.2, "learning_rate": 9.279882176898218e-06, "loss": 0.9351, "step": 2318 }, { "epoch": 0.2, "learning_rate": 9.279167554921097e-06, "loss": 0.8715, "step": 2319 }, { "epoch": 0.2, "learning_rate": 9.278452606079804e-06, "loss": 0.9056, "step": 2320 }, { "epoch": 0.2, "learning_rate": 9.277737330428956e-06, "loss": 0.6015, "step": 2321 }, { "epoch": 0.2, "learning_rate": 9.277021728023187e-06, "loss": 0.9257, "step": 2322 }, { "epoch": 0.2, "learning_rate": 9.27630579891716e-06, "loss": 0.8153, "step": 2323 }, { "epoch": 0.2, "learning_rate": 9.275589543165557e-06, "loss": 0.8873, "step": 2324 }, { "epoch": 0.2, "learning_rate": 9.274872960823094e-06, "loss": 0.9314, "step": 2325 }, { "epoch": 0.2, "learning_rate": 9.274156051944507e-06, "loss": 0.9051, "step": 2326 }, { "epoch": 0.2, "learning_rate": 9.273438816584553e-06, "loss": 0.9255, "step": 2327 }, { "epoch": 0.2, "learning_rate": 9.27272125479802e-06, "loss": 0.9233, "step": 2328 }, { "epoch": 0.2, "learning_rate": 9.27200336663972e-06, "loss": 0.8862, "step": 2329 }, { "epoch": 0.2, "learning_rate": 9.271285152164488e-06, "loss": 0.8731, "step": 2330 }, { "epoch": 0.2, "learning_rate": 9.270566611427185e-06, "loss": 0.9329, "step": 2331 }, { "epoch": 0.2, "learning_rate": 9.269847744482698e-06, "loss": 0.8543, "step": 2332 }, { "epoch": 0.2, "learning_rate": 9.269128551385935e-06, "loss": 0.8809, "step": 2333 }, { "epoch": 0.2, "learning_rate": 9.268409032191833e-06, "loss": 0.8977, "step": 2334 }, { "epoch": 0.2, "learning_rate": 9.267689186955353e-06, "loss": 0.8932, "step": 2335 }, { "epoch": 0.2, "learning_rate": 9.26696901573148e-06, "loss": 0.9289, "step": 2336 }, { "epoch": 0.2, "learning_rate": 9.266248518575223e-06, "loss": 0.8761, "step": 2337 }, { "epoch": 0.2, "learning_rate": 9.265527695541618e-06, "loss": 0.8679, "step": 2338 }, { "epoch": 0.2, "learning_rate": 9.264806546685724e-06, "loss": 0.8833, "step": 2339 }, { "epoch": 0.2, "learning_rate": 9.26408507206263e-06, "loss": 0.8979, "step": 2340 }, { "epoch": 0.2, "learning_rate": 9.263363271727442e-06, "loss": 0.9028, "step": 2341 }, { "epoch": 0.2, "learning_rate": 9.262641145735295e-06, "loss": 0.8636, "step": 2342 }, { "epoch": 0.2, "learning_rate": 9.261918694141352e-06, "loss": 0.9482, "step": 2343 }, { "epoch": 0.2, "learning_rate": 9.261195917000791e-06, "loss": 0.9148, "step": 2344 }, { "epoch": 0.2, "learning_rate": 9.260472814368826e-06, "loss": 0.9257, "step": 2345 }, { "epoch": 0.2, "learning_rate": 9.259749386300693e-06, "loss": 0.9031, "step": 2346 }, { "epoch": 0.2, "learning_rate": 9.259025632851646e-06, "loss": 0.9121, "step": 2347 }, { "epoch": 0.2, "learning_rate": 9.258301554076972e-06, "loss": 0.8551, "step": 2348 }, { "epoch": 0.2, "learning_rate": 9.257577150031978e-06, "loss": 0.8406, "step": 2349 }, { "epoch": 0.2, "learning_rate": 9.256852420771999e-06, "loss": 0.8728, "step": 2350 }, { "epoch": 0.2, "learning_rate": 9.256127366352395e-06, "loss": 0.8756, "step": 2351 }, { "epoch": 0.2, "learning_rate": 9.255401986828544e-06, "loss": 0.9225, "step": 2352 }, { "epoch": 0.2, "learning_rate": 9.25467628225586e-06, "loss": 0.8631, "step": 2353 }, { "epoch": 0.2, "learning_rate": 9.253950252689772e-06, "loss": 0.933, "step": 2354 }, { "epoch": 0.2, "learning_rate": 9.253223898185737e-06, "loss": 0.8771, "step": 2355 }, { "epoch": 0.2, "learning_rate": 9.252497218799244e-06, "loss": 0.8702, "step": 2356 }, { "epoch": 0.2, "learning_rate": 9.251770214585793e-06, "loss": 0.9233, "step": 2357 }, { "epoch": 0.2, "learning_rate": 9.25104288560092e-06, "loss": 0.8779, "step": 2358 }, { "epoch": 0.2, "learning_rate": 9.250315231900182e-06, "loss": 0.8265, "step": 2359 }, { "epoch": 0.2, "learning_rate": 9.24958725353916e-06, "loss": 0.8683, "step": 2360 }, { "epoch": 0.2, "learning_rate": 9.24885895057346e-06, "loss": 0.8898, "step": 2361 }, { "epoch": 0.2, "learning_rate": 9.248130323058714e-06, "loss": 0.9077, "step": 2362 }, { "epoch": 0.2, "learning_rate": 9.247401371050579e-06, "loss": 0.878, "step": 2363 }, { "epoch": 0.2, "learning_rate": 9.246672094604736e-06, "loss": 0.8784, "step": 2364 }, { "epoch": 0.2, "learning_rate": 9.24594249377689e-06, "loss": 0.8578, "step": 2365 }, { "epoch": 0.2, "learning_rate": 9.24521256862277e-06, "loss": 0.938, "step": 2366 }, { "epoch": 0.2, "learning_rate": 9.244482319198134e-06, "loss": 0.9273, "step": 2367 }, { "epoch": 0.2, "learning_rate": 9.243751745558761e-06, "loss": 0.8679, "step": 2368 }, { "epoch": 0.2, "learning_rate": 9.243020847760456e-06, "loss": 0.9038, "step": 2369 }, { "epoch": 0.2, "learning_rate": 9.24228962585905e-06, "loss": 0.909, "step": 2370 }, { "epoch": 0.2, "learning_rate": 9.241558079910393e-06, "loss": 0.8939, "step": 2371 }, { "epoch": 0.2, "learning_rate": 9.24082620997037e-06, "loss": 0.8727, "step": 2372 }, { "epoch": 0.2, "learning_rate": 9.24009401609488e-06, "loss": 0.8975, "step": 2373 }, { "epoch": 0.2, "learning_rate": 9.239361498339854e-06, "loss": 0.8777, "step": 2374 }, { "epoch": 0.2, "learning_rate": 9.238628656761244e-06, "loss": 0.9278, "step": 2375 }, { "epoch": 0.2, "learning_rate": 9.23789549141503e-06, "loss": 0.8832, "step": 2376 }, { "epoch": 0.2, "learning_rate": 9.237162002357214e-06, "loss": 0.9423, "step": 2377 }, { "epoch": 0.2, "learning_rate": 9.236428189643825e-06, "loss": 0.9055, "step": 2378 }, { "epoch": 0.2, "learning_rate": 9.235694053330913e-06, "loss": 0.8617, "step": 2379 }, { "epoch": 0.2, "learning_rate": 9.234959593474553e-06, "loss": 0.8984, "step": 2380 }, { "epoch": 0.2, "learning_rate": 9.234224810130852e-06, "loss": 0.8539, "step": 2381 }, { "epoch": 0.2, "learning_rate": 9.233489703355933e-06, "loss": 0.8881, "step": 2382 }, { "epoch": 0.2, "learning_rate": 9.232754273205947e-06, "loss": 0.9257, "step": 2383 }, { "epoch": 0.2, "learning_rate": 9.232018519737072e-06, "loss": 0.9352, "step": 2384 }, { "epoch": 0.2, "learning_rate": 9.231282443005507e-06, "loss": 0.8801, "step": 2385 }, { "epoch": 0.2, "learning_rate": 9.230546043067477e-06, "loss": 0.8981, "step": 2386 }, { "epoch": 0.2, "learning_rate": 9.229809319979233e-06, "loss": 0.9215, "step": 2387 }, { "epoch": 0.2, "learning_rate": 9.229072273797049e-06, "loss": 0.9396, "step": 2388 }, { "epoch": 0.2, "learning_rate": 9.228334904577222e-06, "loss": 0.8852, "step": 2389 }, { "epoch": 0.2, "learning_rate": 9.227597212376081e-06, "loss": 0.9052, "step": 2390 }, { "epoch": 0.2, "learning_rate": 9.226859197249971e-06, "loss": 0.9019, "step": 2391 }, { "epoch": 0.2, "learning_rate": 9.226120859255266e-06, "loss": 0.8906, "step": 2392 }, { "epoch": 0.2, "learning_rate": 9.225382198448365e-06, "loss": 0.8934, "step": 2393 }, { "epoch": 0.2, "learning_rate": 9.22464321488569e-06, "loss": 0.8402, "step": 2394 }, { "epoch": 0.2, "learning_rate": 9.223903908623684e-06, "loss": 0.8963, "step": 2395 }, { "epoch": 0.2, "learning_rate": 9.223164279718826e-06, "loss": 0.9498, "step": 2396 }, { "epoch": 0.2, "learning_rate": 9.222424328227609e-06, "loss": 0.8876, "step": 2397 }, { "epoch": 0.2, "learning_rate": 9.221684054206554e-06, "loss": 0.9076, "step": 2398 }, { "epoch": 0.2, "learning_rate": 9.220943457712208e-06, "loss": 0.9037, "step": 2399 }, { "epoch": 0.2, "learning_rate": 9.22020253880114e-06, "loss": 0.8865, "step": 2400 }, { "epoch": 0.2, "learning_rate": 9.219461297529945e-06, "loss": 0.9177, "step": 2401 }, { "epoch": 0.2, "learning_rate": 9.218719733955245e-06, "loss": 0.873, "step": 2402 }, { "epoch": 0.21, "learning_rate": 9.217977848133683e-06, "loss": 0.8706, "step": 2403 }, { "epoch": 0.21, "learning_rate": 9.217235640121927e-06, "loss": 0.8982, "step": 2404 }, { "epoch": 0.21, "learning_rate": 9.21649310997667e-06, "loss": 0.8746, "step": 2405 }, { "epoch": 0.21, "learning_rate": 9.215750257754634e-06, "loss": 0.8781, "step": 2406 }, { "epoch": 0.21, "learning_rate": 9.21500708351256e-06, "loss": 0.898, "step": 2407 }, { "epoch": 0.21, "learning_rate": 9.214263587307212e-06, "loss": 0.8998, "step": 2408 }, { "epoch": 0.21, "learning_rate": 9.213519769195384e-06, "loss": 0.892, "step": 2409 }, { "epoch": 0.21, "learning_rate": 9.212775629233896e-06, "loss": 0.926, "step": 2410 }, { "epoch": 0.21, "learning_rate": 9.212031167479585e-06, "loss": 0.8668, "step": 2411 }, { "epoch": 0.21, "learning_rate": 9.211286383989318e-06, "loss": 0.8791, "step": 2412 }, { "epoch": 0.21, "learning_rate": 9.210541278819987e-06, "loss": 0.6064, "step": 2413 }, { "epoch": 0.21, "learning_rate": 9.209795852028503e-06, "loss": 0.8992, "step": 2414 }, { "epoch": 0.21, "learning_rate": 9.209050103671808e-06, "loss": 0.879, "step": 2415 }, { "epoch": 0.21, "learning_rate": 9.208304033806866e-06, "loss": 0.8724, "step": 2416 }, { "epoch": 0.21, "learning_rate": 9.207557642490662e-06, "loss": 0.9217, "step": 2417 }, { "epoch": 0.21, "learning_rate": 9.206810929780216e-06, "loss": 0.9285, "step": 2418 }, { "epoch": 0.21, "learning_rate": 9.20606389573256e-06, "loss": 0.9004, "step": 2419 }, { "epoch": 0.21, "learning_rate": 9.205316540404759e-06, "loss": 0.8618, "step": 2420 }, { "epoch": 0.21, "learning_rate": 9.204568863853897e-06, "loss": 0.8807, "step": 2421 }, { "epoch": 0.21, "learning_rate": 9.203820866137087e-06, "loss": 0.9209, "step": 2422 }, { "epoch": 0.21, "learning_rate": 9.203072547311467e-06, "loss": 0.9052, "step": 2423 }, { "epoch": 0.21, "learning_rate": 9.202323907434194e-06, "loss": 0.8917, "step": 2424 }, { "epoch": 0.21, "learning_rate": 9.201574946562453e-06, "loss": 0.8905, "step": 2425 }, { "epoch": 0.21, "learning_rate": 9.200825664753456e-06, "loss": 0.8456, "step": 2426 }, { "epoch": 0.21, "learning_rate": 9.200076062064433e-06, "loss": 0.9095, "step": 2427 }, { "epoch": 0.21, "learning_rate": 9.199326138552647e-06, "loss": 0.9849, "step": 2428 }, { "epoch": 0.21, "learning_rate": 9.198575894275377e-06, "loss": 0.9305, "step": 2429 }, { "epoch": 0.21, "learning_rate": 9.197825329289932e-06, "loss": 0.8649, "step": 2430 }, { "epoch": 0.21, "learning_rate": 9.197074443653643e-06, "loss": 0.9135, "step": 2431 }, { "epoch": 0.21, "learning_rate": 9.196323237423867e-06, "loss": 0.8826, "step": 2432 }, { "epoch": 0.21, "learning_rate": 9.195571710657986e-06, "loss": 0.901, "step": 2433 }, { "epoch": 0.21, "learning_rate": 9.194819863413405e-06, "loss": 0.9245, "step": 2434 }, { "epoch": 0.21, "learning_rate": 9.194067695747552e-06, "loss": 0.8881, "step": 2435 }, { "epoch": 0.21, "learning_rate": 9.193315207717883e-06, "loss": 0.927, "step": 2436 }, { "epoch": 0.21, "learning_rate": 9.192562399381878e-06, "loss": 0.8611, "step": 2437 }, { "epoch": 0.21, "learning_rate": 9.191809270797037e-06, "loss": 0.9551, "step": 2438 }, { "epoch": 0.21, "learning_rate": 9.191055822020889e-06, "loss": 0.9023, "step": 2439 }, { "epoch": 0.21, "learning_rate": 9.190302053110986e-06, "loss": 0.9305, "step": 2440 }, { "epoch": 0.21, "learning_rate": 9.189547964124907e-06, "loss": 0.9064, "step": 2441 }, { "epoch": 0.21, "learning_rate": 9.18879355512025e-06, "loss": 0.8912, "step": 2442 }, { "epoch": 0.21, "learning_rate": 9.188038826154642e-06, "loss": 0.8508, "step": 2443 }, { "epoch": 0.21, "learning_rate": 9.187283777285733e-06, "loss": 0.8942, "step": 2444 }, { "epoch": 0.21, "learning_rate": 9.186528408571196e-06, "loss": 0.8386, "step": 2445 }, { "epoch": 0.21, "learning_rate": 9.185772720068734e-06, "loss": 0.8267, "step": 2446 }, { "epoch": 0.21, "learning_rate": 9.185016711836065e-06, "loss": 0.9381, "step": 2447 }, { "epoch": 0.21, "learning_rate": 9.184260383930938e-06, "loss": 0.8992, "step": 2448 }, { "epoch": 0.21, "learning_rate": 9.183503736411126e-06, "loss": 0.948, "step": 2449 }, { "epoch": 0.21, "learning_rate": 9.182746769334427e-06, "loss": 0.8757, "step": 2450 }, { "epoch": 0.21, "learning_rate": 9.18198948275866e-06, "loss": 0.878, "step": 2451 }, { "epoch": 0.21, "learning_rate": 9.18123187674167e-06, "loss": 0.9048, "step": 2452 }, { "epoch": 0.21, "learning_rate": 9.180473951341329e-06, "loss": 0.8467, "step": 2453 }, { "epoch": 0.21, "learning_rate": 9.179715706615529e-06, "loss": 0.8559, "step": 2454 }, { "epoch": 0.21, "learning_rate": 9.178957142622187e-06, "loss": 0.9332, "step": 2455 }, { "epoch": 0.21, "learning_rate": 9.17819825941925e-06, "loss": 0.6206, "step": 2456 }, { "epoch": 0.21, "learning_rate": 9.177439057064684e-06, "loss": 0.8976, "step": 2457 }, { "epoch": 0.21, "learning_rate": 9.176679535616477e-06, "loss": 0.8663, "step": 2458 }, { "epoch": 0.21, "learning_rate": 9.175919695132652e-06, "loss": 0.8759, "step": 2459 }, { "epoch": 0.21, "learning_rate": 9.175159535671242e-06, "loss": 0.9325, "step": 2460 }, { "epoch": 0.21, "learning_rate": 9.174399057290316e-06, "loss": 0.9112, "step": 2461 }, { "epoch": 0.21, "learning_rate": 9.173638260047963e-06, "loss": 0.8762, "step": 2462 }, { "epoch": 0.21, "learning_rate": 9.172877144002296e-06, "loss": 0.8752, "step": 2463 }, { "epoch": 0.21, "learning_rate": 9.172115709211452e-06, "loss": 0.9098, "step": 2464 }, { "epoch": 0.21, "learning_rate": 9.171353955733593e-06, "loss": 0.9502, "step": 2465 }, { "epoch": 0.21, "learning_rate": 9.170591883626909e-06, "loss": 0.9456, "step": 2466 }, { "epoch": 0.21, "learning_rate": 9.169829492949605e-06, "loss": 0.8989, "step": 2467 }, { "epoch": 0.21, "learning_rate": 9.169066783759922e-06, "loss": 0.8796, "step": 2468 }, { "epoch": 0.21, "learning_rate": 9.168303756116115e-06, "loss": 0.8877, "step": 2469 }, { "epoch": 0.21, "learning_rate": 9.16754041007647e-06, "loss": 0.8715, "step": 2470 }, { "epoch": 0.21, "learning_rate": 9.166776745699297e-06, "loss": 0.8923, "step": 2471 }, { "epoch": 0.21, "learning_rate": 9.166012763042925e-06, "loss": 0.9043, "step": 2472 }, { "epoch": 0.21, "learning_rate": 9.165248462165712e-06, "loss": 0.89, "step": 2473 }, { "epoch": 0.21, "learning_rate": 9.16448384312604e-06, "loss": 0.9138, "step": 2474 }, { "epoch": 0.21, "learning_rate": 9.163718905982315e-06, "loss": 0.8677, "step": 2475 }, { "epoch": 0.21, "learning_rate": 9.162953650792962e-06, "loss": 0.9008, "step": 2476 }, { "epoch": 0.21, "learning_rate": 9.162188077616443e-06, "loss": 0.9099, "step": 2477 }, { "epoch": 0.21, "learning_rate": 9.161422186511229e-06, "loss": 0.9143, "step": 2478 }, { "epoch": 0.21, "learning_rate": 9.160655977535826e-06, "loss": 0.8422, "step": 2479 }, { "epoch": 0.21, "learning_rate": 9.15988945074876e-06, "loss": 0.8314, "step": 2480 }, { "epoch": 0.21, "learning_rate": 9.159122606208584e-06, "loss": 0.9098, "step": 2481 }, { "epoch": 0.21, "learning_rate": 9.15835544397387e-06, "loss": 0.8859, "step": 2482 }, { "epoch": 0.21, "learning_rate": 9.15758796410322e-06, "loss": 0.864, "step": 2483 }, { "epoch": 0.21, "learning_rate": 9.156820166655259e-06, "loss": 0.8509, "step": 2484 }, { "epoch": 0.21, "learning_rate": 9.156052051688633e-06, "loss": 0.8728, "step": 2485 }, { "epoch": 0.21, "learning_rate": 9.155283619262014e-06, "loss": 0.8752, "step": 2486 }, { "epoch": 0.21, "learning_rate": 9.1545148694341e-06, "loss": 0.9126, "step": 2487 }, { "epoch": 0.21, "learning_rate": 9.153745802263612e-06, "loss": 0.8689, "step": 2488 }, { "epoch": 0.21, "learning_rate": 9.152976417809296e-06, "loss": 0.8976, "step": 2489 }, { "epoch": 0.21, "learning_rate": 9.15220671612992e-06, "loss": 0.9329, "step": 2490 }, { "epoch": 0.21, "learning_rate": 9.151436697284277e-06, "loss": 0.8567, "step": 2491 }, { "epoch": 0.21, "learning_rate": 9.150666361331186e-06, "loss": 0.878, "step": 2492 }, { "epoch": 0.21, "learning_rate": 9.149895708329491e-06, "loss": 0.8984, "step": 2493 }, { "epoch": 0.21, "learning_rate": 9.149124738338053e-06, "loss": 0.9383, "step": 2494 }, { "epoch": 0.21, "learning_rate": 9.148353451415767e-06, "loss": 0.9126, "step": 2495 }, { "epoch": 0.21, "learning_rate": 9.147581847621549e-06, "loss": 0.8892, "step": 2496 }, { "epoch": 0.21, "learning_rate": 9.146809927014333e-06, "loss": 0.8722, "step": 2497 }, { "epoch": 0.21, "learning_rate": 9.146037689653087e-06, "loss": 0.9274, "step": 2498 }, { "epoch": 0.21, "learning_rate": 9.145265135596795e-06, "loss": 0.8811, "step": 2499 }, { "epoch": 0.21, "learning_rate": 9.144492264904469e-06, "loss": 0.8871, "step": 2500 }, { "epoch": 0.21, "learning_rate": 9.143719077635144e-06, "loss": 0.9295, "step": 2501 }, { "epoch": 0.21, "learning_rate": 9.142945573847884e-06, "loss": 0.8726, "step": 2502 }, { "epoch": 0.21, "learning_rate": 9.14217175360177e-06, "loss": 0.8715, "step": 2503 }, { "epoch": 0.21, "learning_rate": 9.141397616955909e-06, "loss": 0.8882, "step": 2504 }, { "epoch": 0.21, "learning_rate": 9.140623163969437e-06, "loss": 0.8982, "step": 2505 }, { "epoch": 0.21, "learning_rate": 9.139848394701507e-06, "loss": 0.8769, "step": 2506 }, { "epoch": 0.21, "learning_rate": 9.139073309211302e-06, "loss": 0.9215, "step": 2507 }, { "epoch": 0.21, "learning_rate": 9.138297907558024e-06, "loss": 0.9063, "step": 2508 }, { "epoch": 0.21, "learning_rate": 9.137522189800907e-06, "loss": 0.8879, "step": 2509 }, { "epoch": 0.21, "learning_rate": 9.1367461559992e-06, "loss": 0.882, "step": 2510 }, { "epoch": 0.21, "learning_rate": 9.135969806212182e-06, "loss": 0.9153, "step": 2511 }, { "epoch": 0.21, "learning_rate": 9.135193140499155e-06, "loss": 0.857, "step": 2512 }, { "epoch": 0.21, "learning_rate": 9.134416158919443e-06, "loss": 0.8599, "step": 2513 }, { "epoch": 0.21, "learning_rate": 9.133638861532399e-06, "loss": 0.6149, "step": 2514 }, { "epoch": 0.21, "learning_rate": 9.132861248397391e-06, "loss": 0.9317, "step": 2515 }, { "epoch": 0.21, "learning_rate": 9.132083319573822e-06, "loss": 0.9201, "step": 2516 }, { "epoch": 0.21, "learning_rate": 9.131305075121111e-06, "loss": 0.9417, "step": 2517 }, { "epoch": 0.21, "learning_rate": 9.130526515098707e-06, "loss": 0.8831, "step": 2518 }, { "epoch": 0.21, "learning_rate": 9.12974763956608e-06, "loss": 0.9038, "step": 2519 }, { "epoch": 0.22, "learning_rate": 9.128968448582722e-06, "loss": 0.8764, "step": 2520 }, { "epoch": 0.22, "learning_rate": 9.128188942208153e-06, "loss": 0.8098, "step": 2521 }, { "epoch": 0.22, "learning_rate": 9.127409120501915e-06, "loss": 0.8857, "step": 2522 }, { "epoch": 0.22, "learning_rate": 9.126628983523575e-06, "loss": 0.867, "step": 2523 }, { "epoch": 0.22, "learning_rate": 9.125848531332725e-06, "loss": 0.892, "step": 2524 }, { "epoch": 0.22, "learning_rate": 9.12506776398898e-06, "loss": 0.584, "step": 2525 }, { "epoch": 0.22, "learning_rate": 9.124286681551974e-06, "loss": 0.9175, "step": 2526 }, { "epoch": 0.22, "learning_rate": 9.123505284081376e-06, "loss": 0.8742, "step": 2527 }, { "epoch": 0.22, "learning_rate": 9.12272357163687e-06, "loss": 0.88, "step": 2528 }, { "epoch": 0.22, "learning_rate": 9.121941544278168e-06, "loss": 0.8888, "step": 2529 }, { "epoch": 0.22, "learning_rate": 9.121159202065005e-06, "loss": 0.845, "step": 2530 }, { "epoch": 0.22, "learning_rate": 9.120376545057139e-06, "loss": 0.9089, "step": 2531 }, { "epoch": 0.22, "learning_rate": 9.119593573314356e-06, "loss": 0.8733, "step": 2532 }, { "epoch": 0.22, "learning_rate": 9.11881028689646e-06, "loss": 0.8614, "step": 2533 }, { "epoch": 0.22, "learning_rate": 9.118026685863287e-06, "loss": 0.9078, "step": 2534 }, { "epoch": 0.22, "learning_rate": 9.117242770274686e-06, "loss": 0.9033, "step": 2535 }, { "epoch": 0.22, "learning_rate": 9.11645854019054e-06, "loss": 0.8926, "step": 2536 }, { "epoch": 0.22, "learning_rate": 9.115673995670752e-06, "loss": 0.8694, "step": 2537 }, { "epoch": 0.22, "learning_rate": 9.114889136775251e-06, "loss": 0.956, "step": 2538 }, { "epoch": 0.22, "learning_rate": 9.114103963563986e-06, "loss": 0.8716, "step": 2539 }, { "epoch": 0.22, "learning_rate": 9.113318476096934e-06, "loss": 0.9147, "step": 2540 }, { "epoch": 0.22, "learning_rate": 9.112532674434094e-06, "loss": 0.8826, "step": 2541 }, { "epoch": 0.22, "learning_rate": 9.111746558635488e-06, "loss": 0.8961, "step": 2542 }, { "epoch": 0.22, "learning_rate": 9.110960128761167e-06, "loss": 0.9019, "step": 2543 }, { "epoch": 0.22, "learning_rate": 9.110173384871198e-06, "loss": 0.9249, "step": 2544 }, { "epoch": 0.22, "learning_rate": 9.109386327025678e-06, "loss": 0.9098, "step": 2545 }, { "epoch": 0.22, "learning_rate": 9.10859895528473e-06, "loss": 0.8448, "step": 2546 }, { "epoch": 0.22, "learning_rate": 9.10781126970849e-06, "loss": 0.854, "step": 2547 }, { "epoch": 0.22, "learning_rate": 9.107023270357134e-06, "loss": 0.8575, "step": 2548 }, { "epoch": 0.22, "learning_rate": 9.106234957290847e-06, "loss": 0.489, "step": 2549 }, { "epoch": 0.22, "learning_rate": 9.105446330569848e-06, "loss": 0.9263, "step": 2550 }, { "epoch": 0.22, "learning_rate": 9.104657390254373e-06, "loss": 0.6048, "step": 2551 }, { "epoch": 0.22, "learning_rate": 9.103868136404688e-06, "loss": 0.9054, "step": 2552 }, { "epoch": 0.22, "learning_rate": 9.103078569081077e-06, "loss": 0.8982, "step": 2553 }, { "epoch": 0.22, "learning_rate": 9.102288688343854e-06, "loss": 0.9478, "step": 2554 }, { "epoch": 0.22, "learning_rate": 9.101498494253354e-06, "loss": 0.8947, "step": 2555 }, { "epoch": 0.22, "learning_rate": 9.100707986869935e-06, "loss": 0.9129, "step": 2556 }, { "epoch": 0.22, "learning_rate": 9.099917166253977e-06, "loss": 0.9423, "step": 2557 }, { "epoch": 0.22, "learning_rate": 9.099126032465892e-06, "loss": 0.8767, "step": 2558 }, { "epoch": 0.22, "learning_rate": 9.09833458556611e-06, "loss": 0.926, "step": 2559 }, { "epoch": 0.22, "learning_rate": 9.097542825615083e-06, "loss": 0.8694, "step": 2560 }, { "epoch": 0.22, "learning_rate": 9.09675075267329e-06, "loss": 0.9293, "step": 2561 }, { "epoch": 0.22, "learning_rate": 9.095958366801234e-06, "loss": 0.8833, "step": 2562 }, { "epoch": 0.22, "learning_rate": 9.095165668059442e-06, "loss": 0.8913, "step": 2563 }, { "epoch": 0.22, "learning_rate": 9.094372656508465e-06, "loss": 0.8637, "step": 2564 }, { "epoch": 0.22, "learning_rate": 9.093579332208875e-06, "loss": 0.892, "step": 2565 }, { "epoch": 0.22, "learning_rate": 9.092785695221271e-06, "loss": 0.5374, "step": 2566 }, { "epoch": 0.22, "learning_rate": 9.091991745606277e-06, "loss": 0.9111, "step": 2567 }, { "epoch": 0.22, "learning_rate": 9.091197483424536e-06, "loss": 0.8822, "step": 2568 }, { "epoch": 0.22, "learning_rate": 9.090402908736718e-06, "loss": 0.9212, "step": 2569 }, { "epoch": 0.22, "learning_rate": 9.089608021603516e-06, "loss": 0.888, "step": 2570 }, { "epoch": 0.22, "learning_rate": 9.088812822085651e-06, "loss": 0.9426, "step": 2571 }, { "epoch": 0.22, "learning_rate": 9.08801731024386e-06, "loss": 0.8841, "step": 2572 }, { "epoch": 0.22, "learning_rate": 9.087221486138911e-06, "loss": 0.9102, "step": 2573 }, { "epoch": 0.22, "learning_rate": 9.086425349831592e-06, "loss": 0.9077, "step": 2574 }, { "epoch": 0.22, "learning_rate": 9.085628901382718e-06, "loss": 0.8965, "step": 2575 }, { "epoch": 0.22, "learning_rate": 9.08483214085312e-06, "loss": 0.8667, "step": 2576 }, { "epoch": 0.22, "learning_rate": 9.084035068303665e-06, "loss": 0.8445, "step": 2577 }, { "epoch": 0.22, "learning_rate": 9.083237683795234e-06, "loss": 0.8708, "step": 2578 }, { "epoch": 0.22, "learning_rate": 9.082439987388736e-06, "loss": 0.8582, "step": 2579 }, { "epoch": 0.22, "learning_rate": 9.081641979145101e-06, "loss": 0.8916, "step": 2580 }, { "epoch": 0.22, "learning_rate": 9.080843659125288e-06, "loss": 0.8405, "step": 2581 }, { "epoch": 0.22, "learning_rate": 9.080045027390277e-06, "loss": 0.9035, "step": 2582 }, { "epoch": 0.22, "learning_rate": 9.079246084001069e-06, "loss": 0.9207, "step": 2583 }, { "epoch": 0.22, "learning_rate": 9.078446829018693e-06, "loss": 0.8709, "step": 2584 }, { "epoch": 0.22, "learning_rate": 9.077647262504197e-06, "loss": 0.8564, "step": 2585 }, { "epoch": 0.22, "learning_rate": 9.07684738451866e-06, "loss": 0.8964, "step": 2586 }, { "epoch": 0.22, "learning_rate": 9.076047195123177e-06, "loss": 0.6296, "step": 2587 }, { "epoch": 0.22, "learning_rate": 9.075246694378874e-06, "loss": 0.922, "step": 2588 }, { "epoch": 0.22, "learning_rate": 9.074445882346893e-06, "loss": 0.8354, "step": 2589 }, { "epoch": 0.22, "learning_rate": 9.07364475908841e-06, "loss": 0.8798, "step": 2590 }, { "epoch": 0.22, "learning_rate": 9.072843324664611e-06, "loss": 0.8933, "step": 2591 }, { "epoch": 0.22, "learning_rate": 9.07204157913672e-06, "loss": 0.8841, "step": 2592 }, { "epoch": 0.22, "learning_rate": 9.071239522565978e-06, "loss": 0.8593, "step": 2593 }, { "epoch": 0.22, "learning_rate": 9.070437155013646e-06, "loss": 0.8555, "step": 2594 }, { "epoch": 0.22, "learning_rate": 9.069634476541016e-06, "loss": 0.8994, "step": 2595 }, { "epoch": 0.22, "learning_rate": 9.068831487209399e-06, "loss": 0.6466, "step": 2596 }, { "epoch": 0.22, "learning_rate": 9.068028187080134e-06, "loss": 0.8663, "step": 2597 }, { "epoch": 0.22, "learning_rate": 9.067224576214577e-06, "loss": 0.5296, "step": 2598 }, { "epoch": 0.22, "learning_rate": 9.066420654674114e-06, "loss": 0.9002, "step": 2599 }, { "epoch": 0.22, "learning_rate": 9.065616422520153e-06, "loss": 0.8928, "step": 2600 }, { "epoch": 0.22, "learning_rate": 9.064811879814125e-06, "loss": 0.9098, "step": 2601 }, { "epoch": 0.22, "learning_rate": 9.064007026617484e-06, "loss": 0.8754, "step": 2602 }, { "epoch": 0.22, "learning_rate": 9.063201862991711e-06, "loss": 0.8792, "step": 2603 }, { "epoch": 0.22, "learning_rate": 9.062396388998306e-06, "loss": 0.8954, "step": 2604 }, { "epoch": 0.22, "learning_rate": 9.061590604698795e-06, "loss": 0.8643, "step": 2605 }, { "epoch": 0.22, "learning_rate": 9.06078451015473e-06, "loss": 0.8585, "step": 2606 }, { "epoch": 0.22, "learning_rate": 9.059978105427683e-06, "loss": 0.82, "step": 2607 }, { "epoch": 0.22, "learning_rate": 9.05917139057925e-06, "loss": 0.9264, "step": 2608 }, { "epoch": 0.22, "learning_rate": 9.058364365671055e-06, "loss": 0.8767, "step": 2609 }, { "epoch": 0.22, "learning_rate": 9.057557030764742e-06, "loss": 0.9126, "step": 2610 }, { "epoch": 0.22, "learning_rate": 9.056749385921975e-06, "loss": 0.9063, "step": 2611 }, { "epoch": 0.22, "learning_rate": 9.055941431204451e-06, "loss": 0.6084, "step": 2612 }, { "epoch": 0.22, "learning_rate": 9.055133166673884e-06, "loss": 0.6324, "step": 2613 }, { "epoch": 0.22, "learning_rate": 9.054324592392012e-06, "loss": 0.8586, "step": 2614 }, { "epoch": 0.22, "learning_rate": 9.0535157084206e-06, "loss": 0.8585, "step": 2615 }, { "epoch": 0.22, "learning_rate": 9.052706514821435e-06, "loss": 0.8408, "step": 2616 }, { "epoch": 0.22, "learning_rate": 9.051897011656323e-06, "loss": 0.6286, "step": 2617 }, { "epoch": 0.22, "learning_rate": 9.051087198987103e-06, "loss": 0.8769, "step": 2618 }, { "epoch": 0.22, "learning_rate": 9.050277076875629e-06, "loss": 0.8681, "step": 2619 }, { "epoch": 0.22, "learning_rate": 9.049466645383785e-06, "loss": 0.9032, "step": 2620 }, { "epoch": 0.22, "learning_rate": 9.048655904573473e-06, "loss": 0.9245, "step": 2621 }, { "epoch": 0.22, "learning_rate": 9.047844854506623e-06, "loss": 0.8186, "step": 2622 }, { "epoch": 0.22, "learning_rate": 9.047033495245186e-06, "loss": 0.8576, "step": 2623 }, { "epoch": 0.22, "learning_rate": 9.046221826851138e-06, "loss": 0.8979, "step": 2624 }, { "epoch": 0.22, "learning_rate": 9.04540984938648e-06, "loss": 0.89, "step": 2625 }, { "epoch": 0.22, "learning_rate": 9.044597562913233e-06, "loss": 0.8651, "step": 2626 }, { "epoch": 0.22, "learning_rate": 9.043784967493446e-06, "loss": 0.8443, "step": 2627 }, { "epoch": 0.22, "learning_rate": 9.042972063189184e-06, "loss": 0.913, "step": 2628 }, { "epoch": 0.22, "learning_rate": 9.042158850062545e-06, "loss": 0.8556, "step": 2629 }, { "epoch": 0.22, "learning_rate": 9.041345328175645e-06, "loss": 0.8867, "step": 2630 }, { "epoch": 0.22, "learning_rate": 9.040531497590627e-06, "loss": 0.8743, "step": 2631 }, { "epoch": 0.22, "learning_rate": 9.03971735836965e-06, "loss": 0.9164, "step": 2632 }, { "epoch": 0.22, "learning_rate": 9.038902910574909e-06, "loss": 0.9337, "step": 2633 }, { "epoch": 0.22, "learning_rate": 9.03808815426861e-06, "loss": 0.8569, "step": 2634 }, { "epoch": 0.22, "learning_rate": 9.03727308951299e-06, "loss": 0.8986, "step": 2635 }, { "epoch": 0.22, "learning_rate": 9.036457716370309e-06, "loss": 0.8923, "step": 2636 }, { "epoch": 0.23, "learning_rate": 9.035642034902847e-06, "loss": 0.9013, "step": 2637 }, { "epoch": 0.23, "learning_rate": 9.034826045172912e-06, "loss": 0.9319, "step": 2638 }, { "epoch": 0.23, "learning_rate": 9.034009747242832e-06, "loss": 0.922, "step": 2639 }, { "epoch": 0.23, "learning_rate": 9.03319314117496e-06, "loss": 0.8846, "step": 2640 }, { "epoch": 0.23, "learning_rate": 9.032376227031673e-06, "loss": 0.8534, "step": 2641 }, { "epoch": 0.23, "learning_rate": 9.03155900487537e-06, "loss": 0.8605, "step": 2642 }, { "epoch": 0.23, "learning_rate": 9.030741474768476e-06, "loss": 0.9073, "step": 2643 }, { "epoch": 0.23, "learning_rate": 9.029923636773438e-06, "loss": 0.8778, "step": 2644 }, { "epoch": 0.23, "learning_rate": 9.029105490952726e-06, "loss": 0.8896, "step": 2645 }, { "epoch": 0.23, "learning_rate": 9.028287037368831e-06, "loss": 0.8841, "step": 2646 }, { "epoch": 0.23, "learning_rate": 9.027468276084274e-06, "loss": 0.9384, "step": 2647 }, { "epoch": 0.23, "learning_rate": 9.026649207161597e-06, "loss": 0.8969, "step": 2648 }, { "epoch": 0.23, "learning_rate": 9.025829830663363e-06, "loss": 0.8916, "step": 2649 }, { "epoch": 0.23, "learning_rate": 9.02501014665216e-06, "loss": 0.9162, "step": 2650 }, { "epoch": 0.23, "learning_rate": 9.0241901551906e-06, "loss": 0.8751, "step": 2651 }, { "epoch": 0.23, "learning_rate": 9.023369856341315e-06, "loss": 0.9185, "step": 2652 }, { "epoch": 0.23, "learning_rate": 9.022549250166969e-06, "loss": 0.9096, "step": 2653 }, { "epoch": 0.23, "learning_rate": 9.02172833673024e-06, "loss": 0.8938, "step": 2654 }, { "epoch": 0.23, "learning_rate": 9.020907116093835e-06, "loss": 0.9062, "step": 2655 }, { "epoch": 0.23, "learning_rate": 9.020085588320483e-06, "loss": 0.8908, "step": 2656 }, { "epoch": 0.23, "learning_rate": 9.019263753472936e-06, "loss": 0.9018, "step": 2657 }, { "epoch": 0.23, "learning_rate": 9.018441611613969e-06, "loss": 0.8928, "step": 2658 }, { "epoch": 0.23, "learning_rate": 9.017619162806382e-06, "loss": 0.8343, "step": 2659 }, { "epoch": 0.23, "learning_rate": 9.016796407113e-06, "loss": 0.8842, "step": 2660 }, { "epoch": 0.23, "learning_rate": 9.015973344596665e-06, "loss": 0.8537, "step": 2661 }, { "epoch": 0.23, "learning_rate": 9.015149975320248e-06, "loss": 0.8884, "step": 2662 }, { "epoch": 0.23, "learning_rate": 9.014326299346645e-06, "loss": 0.9127, "step": 2663 }, { "epoch": 0.23, "learning_rate": 9.01350231673877e-06, "loss": 0.8561, "step": 2664 }, { "epoch": 0.23, "learning_rate": 9.012678027559561e-06, "loss": 0.9129, "step": 2665 }, { "epoch": 0.23, "learning_rate": 9.011853431871985e-06, "loss": 0.8896, "step": 2666 }, { "epoch": 0.23, "learning_rate": 9.011028529739027e-06, "loss": 0.9183, "step": 2667 }, { "epoch": 0.23, "learning_rate": 9.010203321223697e-06, "loss": 0.909, "step": 2668 }, { "epoch": 0.23, "learning_rate": 9.009377806389031e-06, "loss": 0.8758, "step": 2669 }, { "epoch": 0.23, "learning_rate": 9.00855198529808e-06, "loss": 0.9661, "step": 2670 }, { "epoch": 0.23, "learning_rate": 9.007725858013932e-06, "loss": 0.8774, "step": 2671 }, { "epoch": 0.23, "learning_rate": 9.006899424599684e-06, "loss": 0.8967, "step": 2672 }, { "epoch": 0.23, "learning_rate": 9.006072685118466e-06, "loss": 0.894, "step": 2673 }, { "epoch": 0.23, "learning_rate": 9.00524563963343e-06, "loss": 0.8887, "step": 2674 }, { "epoch": 0.23, "learning_rate": 9.004418288207747e-06, "loss": 0.8554, "step": 2675 }, { "epoch": 0.23, "learning_rate": 9.003590630904617e-06, "loss": 0.8803, "step": 2676 }, { "epoch": 0.23, "learning_rate": 9.002762667787258e-06, "loss": 0.8871, "step": 2677 }, { "epoch": 0.23, "learning_rate": 9.001934398918915e-06, "loss": 0.8988, "step": 2678 }, { "epoch": 0.23, "learning_rate": 9.001105824362857e-06, "loss": 0.8963, "step": 2679 }, { "epoch": 0.23, "learning_rate": 9.000276944182372e-06, "loss": 0.8781, "step": 2680 }, { "epoch": 0.23, "learning_rate": 8.999447758440774e-06, "loss": 0.9331, "step": 2681 }, { "epoch": 0.23, "learning_rate": 8.998618267201402e-06, "loss": 0.898, "step": 2682 }, { "epoch": 0.23, "learning_rate": 8.997788470527618e-06, "loss": 0.8646, "step": 2683 }, { "epoch": 0.23, "learning_rate": 8.996958368482802e-06, "loss": 0.8832, "step": 2684 }, { "epoch": 0.23, "learning_rate": 8.996127961130363e-06, "loss": 0.8384, "step": 2685 }, { "epoch": 0.23, "learning_rate": 8.995297248533734e-06, "loss": 0.934, "step": 2686 }, { "epoch": 0.23, "learning_rate": 8.994466230756366e-06, "loss": 0.9346, "step": 2687 }, { "epoch": 0.23, "learning_rate": 8.993634907861737e-06, "loss": 0.8757, "step": 2688 }, { "epoch": 0.23, "learning_rate": 8.99280327991335e-06, "loss": 0.9172, "step": 2689 }, { "epoch": 0.23, "learning_rate": 8.991971346974725e-06, "loss": 0.9363, "step": 2690 }, { "epoch": 0.23, "learning_rate": 8.99113910910941e-06, "loss": 0.8701, "step": 2691 }, { "epoch": 0.23, "learning_rate": 8.990306566380978e-06, "loss": 0.8419, "step": 2692 }, { "epoch": 0.23, "learning_rate": 8.98947371885302e-06, "loss": 0.9005, "step": 2693 }, { "epoch": 0.23, "learning_rate": 8.988640566589156e-06, "loss": 0.8808, "step": 2694 }, { "epoch": 0.23, "learning_rate": 8.987807109653024e-06, "loss": 0.8975, "step": 2695 }, { "epoch": 0.23, "learning_rate": 8.986973348108288e-06, "loss": 0.9383, "step": 2696 }, { "epoch": 0.23, "learning_rate": 8.986139282018636e-06, "loss": 0.8261, "step": 2697 }, { "epoch": 0.23, "learning_rate": 8.985304911447776e-06, "loss": 0.8718, "step": 2698 }, { "epoch": 0.23, "learning_rate": 8.984470236459444e-06, "loss": 0.8965, "step": 2699 }, { "epoch": 0.23, "learning_rate": 8.983635257117394e-06, "loss": 0.86, "step": 2700 }, { "epoch": 0.23, "learning_rate": 8.982799973485407e-06, "loss": 0.9348, "step": 2701 }, { "epoch": 0.23, "learning_rate": 8.981964385627287e-06, "loss": 0.8356, "step": 2702 }, { "epoch": 0.23, "learning_rate": 8.98112849360686e-06, "loss": 0.8668, "step": 2703 }, { "epoch": 0.23, "learning_rate": 8.980292297487975e-06, "loss": 0.8702, "step": 2704 }, { "epoch": 0.23, "learning_rate": 8.979455797334505e-06, "loss": 0.921, "step": 2705 }, { "epoch": 0.23, "learning_rate": 8.978618993210346e-06, "loss": 0.8517, "step": 2706 }, { "epoch": 0.23, "learning_rate": 8.977781885179417e-06, "loss": 0.8874, "step": 2707 }, { "epoch": 0.23, "learning_rate": 8.976944473305662e-06, "loss": 0.878, "step": 2708 }, { "epoch": 0.23, "learning_rate": 8.976106757653045e-06, "loss": 0.9049, "step": 2709 }, { "epoch": 0.23, "learning_rate": 8.975268738285554e-06, "loss": 0.8742, "step": 2710 }, { "epoch": 0.23, "learning_rate": 8.974430415267205e-06, "loss": 0.6179, "step": 2711 }, { "epoch": 0.23, "learning_rate": 8.97359178866203e-06, "loss": 0.9029, "step": 2712 }, { "epoch": 0.23, "learning_rate": 8.972752858534088e-06, "loss": 0.8726, "step": 2713 }, { "epoch": 0.23, "learning_rate": 8.971913624947463e-06, "loss": 0.9335, "step": 2714 }, { "epoch": 0.23, "learning_rate": 8.971074087966255e-06, "loss": 0.8833, "step": 2715 }, { "epoch": 0.23, "learning_rate": 8.970234247654596e-06, "loss": 0.865, "step": 2716 }, { "epoch": 0.23, "learning_rate": 8.969394104076637e-06, "loss": 0.8956, "step": 2717 }, { "epoch": 0.23, "learning_rate": 8.968553657296551e-06, "loss": 0.8786, "step": 2718 }, { "epoch": 0.23, "learning_rate": 8.967712907378536e-06, "loss": 0.886, "step": 2719 }, { "epoch": 0.23, "learning_rate": 8.966871854386812e-06, "loss": 0.9152, "step": 2720 }, { "epoch": 0.23, "learning_rate": 8.966030498385625e-06, "loss": 0.9471, "step": 2721 }, { "epoch": 0.23, "learning_rate": 8.96518883943924e-06, "loss": 0.8698, "step": 2722 }, { "epoch": 0.23, "learning_rate": 8.964346877611947e-06, "loss": 0.9013, "step": 2723 }, { "epoch": 0.23, "learning_rate": 8.96350461296806e-06, "loss": 0.8967, "step": 2724 }, { "epoch": 0.23, "learning_rate": 8.962662045571915e-06, "loss": 0.9136, "step": 2725 }, { "epoch": 0.23, "learning_rate": 8.96181917548787e-06, "loss": 0.8832, "step": 2726 }, { "epoch": 0.23, "learning_rate": 8.960976002780313e-06, "loss": 0.9161, "step": 2727 }, { "epoch": 0.23, "learning_rate": 8.960132527513642e-06, "loss": 0.8885, "step": 2728 }, { "epoch": 0.23, "learning_rate": 8.959288749752294e-06, "loss": 0.8609, "step": 2729 }, { "epoch": 0.23, "learning_rate": 8.958444669560714e-06, "loss": 0.8922, "step": 2730 }, { "epoch": 0.23, "learning_rate": 8.95760028700338e-06, "loss": 0.8906, "step": 2731 }, { "epoch": 0.23, "learning_rate": 8.956755602144793e-06, "loss": 0.8942, "step": 2732 }, { "epoch": 0.23, "learning_rate": 8.955910615049469e-06, "loss": 0.5936, "step": 2733 }, { "epoch": 0.23, "learning_rate": 8.955065325781956e-06, "loss": 0.9109, "step": 2734 }, { "epoch": 0.23, "learning_rate": 8.954219734406817e-06, "loss": 0.8784, "step": 2735 }, { "epoch": 0.23, "learning_rate": 8.953373840988648e-06, "loss": 0.9102, "step": 2736 }, { "epoch": 0.23, "learning_rate": 8.95252764559206e-06, "loss": 0.8786, "step": 2737 }, { "epoch": 0.23, "learning_rate": 8.951681148281688e-06, "loss": 0.825, "step": 2738 }, { "epoch": 0.23, "learning_rate": 8.950834349122195e-06, "loss": 0.8146, "step": 2739 }, { "epoch": 0.23, "learning_rate": 8.949987248178262e-06, "loss": 0.864, "step": 2740 }, { "epoch": 0.23, "learning_rate": 8.949139845514595e-06, "loss": 0.8825, "step": 2741 }, { "epoch": 0.23, "learning_rate": 8.948292141195922e-06, "loss": 0.9163, "step": 2742 }, { "epoch": 0.23, "learning_rate": 8.947444135286998e-06, "loss": 0.9224, "step": 2743 }, { "epoch": 0.23, "learning_rate": 8.946595827852594e-06, "loss": 0.8258, "step": 2744 }, { "epoch": 0.23, "learning_rate": 8.945747218957509e-06, "loss": 0.8375, "step": 2745 }, { "epoch": 0.23, "learning_rate": 8.944898308666566e-06, "loss": 0.893, "step": 2746 }, { "epoch": 0.23, "learning_rate": 8.944049097044606e-06, "loss": 0.8841, "step": 2747 }, { "epoch": 0.23, "learning_rate": 8.943199584156497e-06, "loss": 0.9173, "step": 2748 }, { "epoch": 0.23, "learning_rate": 8.94234977006713e-06, "loss": 0.8597, "step": 2749 }, { "epoch": 0.23, "learning_rate": 8.94149965484142e-06, "loss": 0.8258, "step": 2750 }, { "epoch": 0.23, "learning_rate": 8.9406492385443e-06, "loss": 0.91, "step": 2751 }, { "epoch": 0.23, "learning_rate": 8.939798521240728e-06, "loss": 0.9107, "step": 2752 }, { "epoch": 0.23, "learning_rate": 8.938947502995688e-06, "loss": 0.9361, "step": 2753 }, { "epoch": 0.24, "learning_rate": 8.938096183874186e-06, "loss": 0.9597, "step": 2754 }, { "epoch": 0.24, "learning_rate": 8.937244563941248e-06, "loss": 0.8486, "step": 2755 }, { "epoch": 0.24, "learning_rate": 8.936392643261926e-06, "loss": 0.914, "step": 2756 }, { "epoch": 0.24, "learning_rate": 8.935540421901294e-06, "loss": 0.8706, "step": 2757 }, { "epoch": 0.24, "learning_rate": 8.934687899924448e-06, "loss": 0.9089, "step": 2758 }, { "epoch": 0.24, "learning_rate": 8.933835077396509e-06, "loss": 0.84, "step": 2759 }, { "epoch": 0.24, "learning_rate": 8.932981954382619e-06, "loss": 0.8674, "step": 2760 }, { "epoch": 0.24, "learning_rate": 8.932128530947945e-06, "loss": 0.9352, "step": 2761 }, { "epoch": 0.24, "learning_rate": 8.931274807157674e-06, "loss": 0.9349, "step": 2762 }, { "epoch": 0.24, "learning_rate": 8.930420783077018e-06, "loss": 0.8629, "step": 2763 }, { "epoch": 0.24, "learning_rate": 8.929566458771213e-06, "loss": 0.9548, "step": 2764 }, { "epoch": 0.24, "learning_rate": 8.928711834305516e-06, "loss": 0.9052, "step": 2765 }, { "epoch": 0.24, "learning_rate": 8.927856909745208e-06, "loss": 0.9535, "step": 2766 }, { "epoch": 0.24, "learning_rate": 8.92700168515559e-06, "loss": 0.9089, "step": 2767 }, { "epoch": 0.24, "learning_rate": 8.92614616060199e-06, "loss": 0.5599, "step": 2768 }, { "epoch": 0.24, "learning_rate": 8.925290336149758e-06, "loss": 0.9199, "step": 2769 }, { "epoch": 0.24, "learning_rate": 8.924434211864263e-06, "loss": 0.873, "step": 2770 }, { "epoch": 0.24, "learning_rate": 8.923577787810907e-06, "loss": 0.878, "step": 2771 }, { "epoch": 0.24, "learning_rate": 8.9227210640551e-06, "loss": 0.867, "step": 2772 }, { "epoch": 0.24, "learning_rate": 8.921864040662288e-06, "loss": 0.8755, "step": 2773 }, { "epoch": 0.24, "learning_rate": 8.921006717697931e-06, "loss": 0.8782, "step": 2774 }, { "epoch": 0.24, "learning_rate": 8.920149095227517e-06, "loss": 0.8974, "step": 2775 }, { "epoch": 0.24, "learning_rate": 8.919291173316558e-06, "loss": 0.8559, "step": 2776 }, { "epoch": 0.24, "learning_rate": 8.918432952030582e-06, "loss": 0.8953, "step": 2777 }, { "epoch": 0.24, "learning_rate": 8.917574431435149e-06, "loss": 0.5848, "step": 2778 }, { "epoch": 0.24, "learning_rate": 8.916715611595833e-06, "loss": 0.9066, "step": 2779 }, { "epoch": 0.24, "learning_rate": 8.915856492578236e-06, "loss": 0.9021, "step": 2780 }, { "epoch": 0.24, "learning_rate": 8.914997074447986e-06, "loss": 0.8345, "step": 2781 }, { "epoch": 0.24, "learning_rate": 8.914137357270723e-06, "loss": 0.8875, "step": 2782 }, { "epoch": 0.24, "learning_rate": 8.913277341112118e-06, "loss": 0.8513, "step": 2783 }, { "epoch": 0.24, "learning_rate": 8.912417026037868e-06, "loss": 0.6486, "step": 2784 }, { "epoch": 0.24, "learning_rate": 8.911556412113685e-06, "loss": 0.8666, "step": 2785 }, { "epoch": 0.24, "learning_rate": 8.910695499405305e-06, "loss": 0.8972, "step": 2786 }, { "epoch": 0.24, "learning_rate": 8.909834287978492e-06, "loss": 0.8936, "step": 2787 }, { "epoch": 0.24, "learning_rate": 8.908972777899027e-06, "loss": 0.9102, "step": 2788 }, { "epoch": 0.24, "learning_rate": 8.90811096923272e-06, "loss": 0.9157, "step": 2789 }, { "epoch": 0.24, "learning_rate": 8.907248862045396e-06, "loss": 0.8683, "step": 2790 }, { "epoch": 0.24, "learning_rate": 8.90638645640291e-06, "loss": 0.8404, "step": 2791 }, { "epoch": 0.24, "learning_rate": 8.905523752371136e-06, "loss": 0.9093, "step": 2792 }, { "epoch": 0.24, "learning_rate": 8.904660750015972e-06, "loss": 0.9043, "step": 2793 }, { "epoch": 0.24, "learning_rate": 8.903797449403338e-06, "loss": 0.8767, "step": 2794 }, { "epoch": 0.24, "learning_rate": 8.902933850599176e-06, "loss": 0.8038, "step": 2795 }, { "epoch": 0.24, "learning_rate": 8.902069953669455e-06, "loss": 0.8836, "step": 2796 }, { "epoch": 0.24, "learning_rate": 8.901205758680161e-06, "loss": 0.8747, "step": 2797 }, { "epoch": 0.24, "learning_rate": 8.900341265697306e-06, "loss": 0.8902, "step": 2798 }, { "epoch": 0.24, "learning_rate": 8.899476474786925e-06, "loss": 0.8747, "step": 2799 }, { "epoch": 0.24, "learning_rate": 8.898611386015077e-06, "loss": 0.8739, "step": 2800 }, { "epoch": 0.24, "learning_rate": 8.897745999447837e-06, "loss": 0.8821, "step": 2801 }, { "epoch": 0.24, "learning_rate": 8.89688031515131e-06, "loss": 0.8557, "step": 2802 }, { "epoch": 0.24, "learning_rate": 8.896014333191624e-06, "loss": 0.882, "step": 2803 }, { "epoch": 0.24, "learning_rate": 8.895148053634923e-06, "loss": 0.8841, "step": 2804 }, { "epoch": 0.24, "learning_rate": 8.894281476547379e-06, "loss": 0.919, "step": 2805 }, { "epoch": 0.24, "learning_rate": 8.893414601995188e-06, "loss": 0.8724, "step": 2806 }, { "epoch": 0.24, "learning_rate": 8.89254743004456e-06, "loss": 0.8845, "step": 2807 }, { "epoch": 0.24, "learning_rate": 8.89167996076174e-06, "loss": 0.8314, "step": 2808 }, { "epoch": 0.24, "learning_rate": 8.890812194212987e-06, "loss": 0.905, "step": 2809 }, { "epoch": 0.24, "learning_rate": 8.889944130464587e-06, "loss": 0.8747, "step": 2810 }, { "epoch": 0.24, "learning_rate": 8.889075769582844e-06, "loss": 0.8428, "step": 2811 }, { "epoch": 0.24, "learning_rate": 8.888207111634092e-06, "loss": 0.9017, "step": 2812 }, { "epoch": 0.24, "learning_rate": 8.887338156684679e-06, "loss": 0.899, "step": 2813 }, { "epoch": 0.24, "learning_rate": 8.886468904800983e-06, "loss": 0.8864, "step": 2814 }, { "epoch": 0.24, "learning_rate": 8.8855993560494e-06, "loss": 0.7975, "step": 2815 }, { "epoch": 0.24, "learning_rate": 8.884729510496353e-06, "loss": 0.8741, "step": 2816 }, { "epoch": 0.24, "learning_rate": 8.883859368208283e-06, "loss": 0.8933, "step": 2817 }, { "epoch": 0.24, "learning_rate": 8.882988929251657e-06, "loss": 0.9015, "step": 2818 }, { "epoch": 0.24, "learning_rate": 8.88211819369296e-06, "loss": 0.8869, "step": 2819 }, { "epoch": 0.24, "learning_rate": 8.88124716159871e-06, "loss": 0.9214, "step": 2820 }, { "epoch": 0.24, "learning_rate": 8.880375833035434e-06, "loss": 0.9074, "step": 2821 }, { "epoch": 0.24, "learning_rate": 8.87950420806969e-06, "loss": 0.8812, "step": 2822 }, { "epoch": 0.24, "learning_rate": 8.878632286768062e-06, "loss": 0.9021, "step": 2823 }, { "epoch": 0.24, "learning_rate": 8.877760069197146e-06, "loss": 0.8908, "step": 2824 }, { "epoch": 0.24, "learning_rate": 8.876887555423569e-06, "loss": 0.8551, "step": 2825 }, { "epoch": 0.24, "learning_rate": 8.876014745513976e-06, "loss": 0.837, "step": 2826 }, { "epoch": 0.24, "learning_rate": 8.875141639535038e-06, "loss": 0.8848, "step": 2827 }, { "epoch": 0.24, "learning_rate": 8.874268237553448e-06, "loss": 0.849, "step": 2828 }, { "epoch": 0.24, "learning_rate": 8.873394539635918e-06, "loss": 0.6302, "step": 2829 }, { "epoch": 0.24, "learning_rate": 8.87252054584919e-06, "loss": 0.9302, "step": 2830 }, { "epoch": 0.24, "learning_rate": 8.871646256260019e-06, "loss": 0.9406, "step": 2831 }, { "epoch": 0.24, "learning_rate": 8.87077167093519e-06, "loss": 0.8944, "step": 2832 }, { "epoch": 0.24, "learning_rate": 8.869896789941507e-06, "loss": 0.8919, "step": 2833 }, { "epoch": 0.24, "learning_rate": 8.8690216133458e-06, "loss": 0.8851, "step": 2834 }, { "epoch": 0.24, "learning_rate": 8.868146141214919e-06, "loss": 0.9052, "step": 2835 }, { "epoch": 0.24, "learning_rate": 8.867270373615735e-06, "loss": 0.8863, "step": 2836 }, { "epoch": 0.24, "learning_rate": 8.866394310615144e-06, "loss": 0.9405, "step": 2837 }, { "epoch": 0.24, "learning_rate": 8.865517952280065e-06, "loss": 0.8867, "step": 2838 }, { "epoch": 0.24, "learning_rate": 8.86464129867744e-06, "loss": 0.9204, "step": 2839 }, { "epoch": 0.24, "learning_rate": 8.86376434987423e-06, "loss": 0.8144, "step": 2840 }, { "epoch": 0.24, "learning_rate": 8.862887105937419e-06, "loss": 0.8756, "step": 2841 }, { "epoch": 0.24, "learning_rate": 8.862009566934018e-06, "loss": 0.8381, "step": 2842 }, { "epoch": 0.24, "learning_rate": 8.861131732931057e-06, "loss": 0.9291, "step": 2843 }, { "epoch": 0.24, "learning_rate": 8.860253603995592e-06, "loss": 0.9051, "step": 2844 }, { "epoch": 0.24, "learning_rate": 8.859375180194694e-06, "loss": 0.8589, "step": 2845 }, { "epoch": 0.24, "learning_rate": 8.858496461595466e-06, "loss": 0.8963, "step": 2846 }, { "epoch": 0.24, "learning_rate": 8.857617448265024e-06, "loss": 0.9029, "step": 2847 }, { "epoch": 0.24, "learning_rate": 8.856738140270516e-06, "loss": 0.8838, "step": 2848 }, { "epoch": 0.24, "learning_rate": 8.855858537679104e-06, "loss": 0.8951, "step": 2849 }, { "epoch": 0.24, "learning_rate": 8.85497864055798e-06, "loss": 0.8757, "step": 2850 }, { "epoch": 0.24, "learning_rate": 8.854098448974354e-06, "loss": 0.883, "step": 2851 }, { "epoch": 0.24, "learning_rate": 8.853217962995458e-06, "loss": 0.8682, "step": 2852 }, { "epoch": 0.24, "learning_rate": 8.852337182688548e-06, "loss": 0.881, "step": 2853 }, { "epoch": 0.24, "learning_rate": 8.851456108120906e-06, "loss": 0.9245, "step": 2854 }, { "epoch": 0.24, "learning_rate": 8.850574739359827e-06, "loss": 0.8205, "step": 2855 }, { "epoch": 0.24, "learning_rate": 8.849693076472638e-06, "loss": 0.8928, "step": 2856 }, { "epoch": 0.24, "learning_rate": 8.848811119526685e-06, "loss": 0.8769, "step": 2857 }, { "epoch": 0.24, "learning_rate": 8.847928868589336e-06, "loss": 0.8988, "step": 2858 }, { "epoch": 0.24, "learning_rate": 8.84704632372798e-06, "loss": 0.8862, "step": 2859 }, { "epoch": 0.24, "learning_rate": 8.846163485010033e-06, "loss": 0.8651, "step": 2860 }, { "epoch": 0.24, "learning_rate": 8.845280352502927e-06, "loss": 0.9082, "step": 2861 }, { "epoch": 0.24, "learning_rate": 8.844396926274123e-06, "loss": 0.8892, "step": 2862 }, { "epoch": 0.24, "learning_rate": 8.8435132063911e-06, "loss": 0.937, "step": 2863 }, { "epoch": 0.24, "learning_rate": 8.842629192921364e-06, "loss": 0.8368, "step": 2864 }, { "epoch": 0.24, "learning_rate": 8.841744885932435e-06, "loss": 0.8735, "step": 2865 }, { "epoch": 0.24, "learning_rate": 8.840860285491866e-06, "loss": 0.8449, "step": 2866 }, { "epoch": 0.24, "learning_rate": 8.839975391667223e-06, "loss": 0.9309, "step": 2867 }, { "epoch": 0.24, "learning_rate": 8.839090204526101e-06, "loss": 0.9038, "step": 2868 }, { "epoch": 0.24, "learning_rate": 8.838204724136114e-06, "loss": 0.8926, "step": 2869 }, { "epoch": 0.24, "learning_rate": 8.8373189505649e-06, "loss": 0.9179, "step": 2870 }, { "epoch": 0.24, "learning_rate": 8.836432883880119e-06, "loss": 0.9433, "step": 2871 }, { "epoch": 0.25, "learning_rate": 8.835546524149452e-06, "loss": 0.8769, "step": 2872 }, { "epoch": 0.25, "learning_rate": 8.834659871440608e-06, "loss": 0.8967, "step": 2873 }, { "epoch": 0.25, "learning_rate": 8.833772925821306e-06, "loss": 0.5291, "step": 2874 }, { "epoch": 0.25, "learning_rate": 8.832885687359302e-06, "loss": 0.8311, "step": 2875 }, { "epoch": 0.25, "learning_rate": 8.831998156122364e-06, "loss": 0.8269, "step": 2876 }, { "epoch": 0.25, "learning_rate": 8.831110332178288e-06, "loss": 0.8822, "step": 2877 }, { "epoch": 0.25, "learning_rate": 8.83022221559489e-06, "loss": 0.9361, "step": 2878 }, { "epoch": 0.25, "learning_rate": 8.82933380644001e-06, "loss": 0.8985, "step": 2879 }, { "epoch": 0.25, "learning_rate": 8.828445104781508e-06, "loss": 0.8477, "step": 2880 }, { "epoch": 0.25, "learning_rate": 8.827556110687265e-06, "loss": 0.8819, "step": 2881 }, { "epoch": 0.25, "learning_rate": 8.82666682422519e-06, "loss": 0.8466, "step": 2882 }, { "epoch": 0.25, "learning_rate": 8.82577724546321e-06, "loss": 0.8333, "step": 2883 }, { "epoch": 0.25, "learning_rate": 8.824887374469277e-06, "loss": 0.9081, "step": 2884 }, { "epoch": 0.25, "learning_rate": 8.823997211311362e-06, "loss": 0.8636, "step": 2885 }, { "epoch": 0.25, "learning_rate": 8.82310675605746e-06, "loss": 0.901, "step": 2886 }, { "epoch": 0.25, "learning_rate": 8.822216008775589e-06, "loss": 0.8938, "step": 2887 }, { "epoch": 0.25, "learning_rate": 8.821324969533788e-06, "loss": 0.8823, "step": 2888 }, { "epoch": 0.25, "learning_rate": 8.820433638400121e-06, "loss": 0.8661, "step": 2889 }, { "epoch": 0.25, "learning_rate": 8.81954201544267e-06, "loss": 0.8541, "step": 2890 }, { "epoch": 0.25, "learning_rate": 8.818650100729545e-06, "loss": 0.908, "step": 2891 }, { "epoch": 0.25, "learning_rate": 8.817757894328871e-06, "loss": 0.896, "step": 2892 }, { "epoch": 0.25, "learning_rate": 8.8168653963088e-06, "loss": 0.8578, "step": 2893 }, { "epoch": 0.25, "learning_rate": 8.815972606737508e-06, "loss": 0.8821, "step": 2894 }, { "epoch": 0.25, "learning_rate": 8.815079525683188e-06, "loss": 0.8989, "step": 2895 }, { "epoch": 0.25, "learning_rate": 8.814186153214057e-06, "loss": 0.9092, "step": 2896 }, { "epoch": 0.25, "learning_rate": 8.81329248939836e-06, "loss": 0.8437, "step": 2897 }, { "epoch": 0.25, "learning_rate": 8.812398534304356e-06, "loss": 0.6181, "step": 2898 }, { "epoch": 0.25, "learning_rate": 8.81150428800033e-06, "loss": 0.871, "step": 2899 }, { "epoch": 0.25, "learning_rate": 8.81060975055459e-06, "loss": 0.889, "step": 2900 }, { "epoch": 0.25, "learning_rate": 8.809714922035463e-06, "loss": 0.9348, "step": 2901 }, { "epoch": 0.25, "learning_rate": 8.808819802511303e-06, "loss": 0.9037, "step": 2902 }, { "epoch": 0.25, "learning_rate": 8.807924392050484e-06, "loss": 0.6081, "step": 2903 }, { "epoch": 0.25, "learning_rate": 8.8070286907214e-06, "loss": 0.8683, "step": 2904 }, { "epoch": 0.25, "learning_rate": 8.806132698592469e-06, "loss": 0.8441, "step": 2905 }, { "epoch": 0.25, "learning_rate": 8.805236415732133e-06, "loss": 0.8918, "step": 2906 }, { "epoch": 0.25, "learning_rate": 8.804339842208854e-06, "loss": 0.888, "step": 2907 }, { "epoch": 0.25, "learning_rate": 8.803442978091114e-06, "loss": 0.883, "step": 2908 }, { "epoch": 0.25, "learning_rate": 8.802545823447426e-06, "loss": 0.875, "step": 2909 }, { "epoch": 0.25, "learning_rate": 8.801648378346314e-06, "loss": 0.8826, "step": 2910 }, { "epoch": 0.25, "learning_rate": 8.80075064285633e-06, "loss": 0.8776, "step": 2911 }, { "epoch": 0.25, "learning_rate": 8.79985261704605e-06, "loss": 0.8802, "step": 2912 }, { "epoch": 0.25, "learning_rate": 8.798954300984068e-06, "loss": 0.8749, "step": 2913 }, { "epoch": 0.25, "learning_rate": 8.798055694739002e-06, "loss": 0.9269, "step": 2914 }, { "epoch": 0.25, "learning_rate": 8.79715679837949e-06, "loss": 0.8572, "step": 2915 }, { "epoch": 0.25, "learning_rate": 8.7962576119742e-06, "loss": 0.8919, "step": 2916 }, { "epoch": 0.25, "learning_rate": 8.79535813559181e-06, "loss": 0.8541, "step": 2917 }, { "epoch": 0.25, "learning_rate": 8.79445836930103e-06, "loss": 0.9011, "step": 2918 }, { "epoch": 0.25, "learning_rate": 8.793558313170586e-06, "loss": 0.8955, "step": 2919 }, { "epoch": 0.25, "learning_rate": 8.792657967269234e-06, "loss": 0.8546, "step": 2920 }, { "epoch": 0.25, "learning_rate": 8.79175733166574e-06, "loss": 0.8347, "step": 2921 }, { "epoch": 0.25, "learning_rate": 8.790856406428904e-06, "loss": 0.9335, "step": 2922 }, { "epoch": 0.25, "learning_rate": 8.789955191627542e-06, "loss": 0.8993, "step": 2923 }, { "epoch": 0.25, "learning_rate": 8.789053687330492e-06, "loss": 0.9143, "step": 2924 }, { "epoch": 0.25, "learning_rate": 8.788151893606616e-06, "loss": 0.8884, "step": 2925 }, { "epoch": 0.25, "learning_rate": 8.7872498105248e-06, "loss": 0.868, "step": 2926 }, { "epoch": 0.25, "learning_rate": 8.786347438153946e-06, "loss": 0.8991, "step": 2927 }, { "epoch": 0.25, "learning_rate": 8.785444776562982e-06, "loss": 0.9069, "step": 2928 }, { "epoch": 0.25, "learning_rate": 8.78454182582086e-06, "loss": 0.8815, "step": 2929 }, { "epoch": 0.25, "learning_rate": 8.783638585996552e-06, "loss": 0.8479, "step": 2930 }, { "epoch": 0.25, "learning_rate": 8.78273505715905e-06, "loss": 0.6337, "step": 2931 }, { "epoch": 0.25, "learning_rate": 8.781831239377368e-06, "loss": 0.9162, "step": 2932 }, { "epoch": 0.25, "learning_rate": 8.780927132720549e-06, "loss": 0.9157, "step": 2933 }, { "epoch": 0.25, "learning_rate": 8.780022737257653e-06, "loss": 0.8871, "step": 2934 }, { "epoch": 0.25, "learning_rate": 8.779118053057757e-06, "loss": 0.8946, "step": 2935 }, { "epoch": 0.25, "learning_rate": 8.77821308018997e-06, "loss": 0.8226, "step": 2936 }, { "epoch": 0.25, "learning_rate": 8.777307818723418e-06, "loss": 0.615, "step": 2937 }, { "epoch": 0.25, "learning_rate": 8.776402268727247e-06, "loss": 0.8795, "step": 2938 }, { "epoch": 0.25, "learning_rate": 8.77549643027063e-06, "loss": 0.8973, "step": 2939 }, { "epoch": 0.25, "learning_rate": 8.774590303422757e-06, "loss": 0.8845, "step": 2940 }, { "epoch": 0.25, "learning_rate": 8.773683888252844e-06, "loss": 0.8622, "step": 2941 }, { "epoch": 0.25, "learning_rate": 8.772777184830128e-06, "loss": 0.8863, "step": 2942 }, { "epoch": 0.25, "learning_rate": 8.771870193223866e-06, "loss": 0.9026, "step": 2943 }, { "epoch": 0.25, "learning_rate": 8.77096291350334e-06, "loss": 0.9357, "step": 2944 }, { "epoch": 0.25, "learning_rate": 8.77005534573785e-06, "loss": 0.8764, "step": 2945 }, { "epoch": 0.25, "learning_rate": 8.769147489996724e-06, "loss": 0.8721, "step": 2946 }, { "epoch": 0.25, "learning_rate": 8.768239346349307e-06, "loss": 0.8457, "step": 2947 }, { "epoch": 0.25, "learning_rate": 8.767330914864967e-06, "loss": 0.8914, "step": 2948 }, { "epoch": 0.25, "learning_rate": 8.766422195613096e-06, "loss": 0.8799, "step": 2949 }, { "epoch": 0.25, "learning_rate": 8.765513188663106e-06, "loss": 0.8811, "step": 2950 }, { "epoch": 0.25, "learning_rate": 8.764603894084429e-06, "loss": 0.9172, "step": 2951 }, { "epoch": 0.25, "learning_rate": 8.763694311946524e-06, "loss": 0.8774, "step": 2952 }, { "epoch": 0.25, "learning_rate": 8.76278444231887e-06, "loss": 0.875, "step": 2953 }, { "epoch": 0.25, "learning_rate": 8.761874285270966e-06, "loss": 0.935, "step": 2954 }, { "epoch": 0.25, "learning_rate": 8.760963840872334e-06, "loss": 0.8753, "step": 2955 }, { "epoch": 0.25, "learning_rate": 8.760053109192519e-06, "loss": 0.8853, "step": 2956 }, { "epoch": 0.25, "learning_rate": 8.759142090301087e-06, "loss": 0.8883, "step": 2957 }, { "epoch": 0.25, "learning_rate": 8.758230784267627e-06, "loss": 0.8383, "step": 2958 }, { "epoch": 0.25, "learning_rate": 8.757319191161748e-06, "loss": 0.8197, "step": 2959 }, { "epoch": 0.25, "learning_rate": 8.756407311053082e-06, "loss": 0.884, "step": 2960 }, { "epoch": 0.25, "learning_rate": 8.755495144011284e-06, "loss": 0.8401, "step": 2961 }, { "epoch": 0.25, "learning_rate": 8.75458269010603e-06, "loss": 0.927, "step": 2962 }, { "epoch": 0.25, "learning_rate": 8.753669949407015e-06, "loss": 0.8519, "step": 2963 }, { "epoch": 0.25, "learning_rate": 8.752756921983962e-06, "loss": 0.898, "step": 2964 }, { "epoch": 0.25, "learning_rate": 8.75184360790661e-06, "loss": 0.8694, "step": 2965 }, { "epoch": 0.25, "learning_rate": 8.750930007244727e-06, "loss": 0.9309, "step": 2966 }, { "epoch": 0.25, "learning_rate": 8.750016120068094e-06, "loss": 0.8584, "step": 2967 }, { "epoch": 0.25, "learning_rate": 8.74910194644652e-06, "loss": 0.8823, "step": 2968 }, { "epoch": 0.25, "learning_rate": 8.74818748644983e-06, "loss": 0.8712, "step": 2969 }, { "epoch": 0.25, "learning_rate": 8.747272740147883e-06, "loss": 0.8862, "step": 2970 }, { "epoch": 0.25, "learning_rate": 8.746357707610544e-06, "loss": 0.8698, "step": 2971 }, { "epoch": 0.25, "learning_rate": 8.745442388907713e-06, "loss": 0.6544, "step": 2972 }, { "epoch": 0.25, "learning_rate": 8.744526784109306e-06, "loss": 0.9011, "step": 2973 }, { "epoch": 0.25, "learning_rate": 8.743610893285259e-06, "loss": 0.5945, "step": 2974 }, { "epoch": 0.25, "learning_rate": 8.742694716505533e-06, "loss": 0.8784, "step": 2975 }, { "epoch": 0.25, "learning_rate": 8.74177825384011e-06, "loss": 0.8749, "step": 2976 }, { "epoch": 0.25, "learning_rate": 8.740861505358997e-06, "loss": 0.8731, "step": 2977 }, { "epoch": 0.25, "learning_rate": 8.739944471132217e-06, "loss": 0.609, "step": 2978 }, { "epoch": 0.25, "learning_rate": 8.739027151229817e-06, "loss": 0.9041, "step": 2979 }, { "epoch": 0.25, "learning_rate": 8.738109545721869e-06, "loss": 0.8969, "step": 2980 }, { "epoch": 0.25, "learning_rate": 8.737191654678464e-06, "loss": 0.8527, "step": 2981 }, { "epoch": 0.25, "learning_rate": 8.736273478169712e-06, "loss": 0.8857, "step": 2982 }, { "epoch": 0.25, "learning_rate": 8.735355016265751e-06, "loss": 0.8744, "step": 2983 }, { "epoch": 0.25, "learning_rate": 8.734436269036737e-06, "loss": 0.8712, "step": 2984 }, { "epoch": 0.25, "learning_rate": 8.733517236552848e-06, "loss": 0.9061, "step": 2985 }, { "epoch": 0.25, "learning_rate": 8.732597918884284e-06, "loss": 0.9077, "step": 2986 }, { "epoch": 0.25, "learning_rate": 8.731678316101269e-06, "loss": 0.8461, "step": 2987 }, { "epoch": 0.25, "learning_rate": 8.730758428274044e-06, "loss": 0.9219, "step": 2988 }, { "epoch": 0.26, "learning_rate": 8.729838255472875e-06, "loss": 0.8439, "step": 2989 }, { "epoch": 0.26, "learning_rate": 8.728917797768053e-06, "loss": 0.9017, "step": 2990 }, { "epoch": 0.26, "learning_rate": 8.727997055229884e-06, "loss": 0.9214, "step": 2991 }, { "epoch": 0.26, "learning_rate": 8.727076027928698e-06, "loss": 0.912, "step": 2992 }, { "epoch": 0.26, "learning_rate": 8.72615471593485e-06, "loss": 0.9277, "step": 2993 }, { "epoch": 0.26, "learning_rate": 8.725233119318715e-06, "loss": 0.8543, "step": 2994 }, { "epoch": 0.26, "learning_rate": 8.724311238150685e-06, "loss": 0.9088, "step": 2995 }, { "epoch": 0.26, "learning_rate": 8.723389072501182e-06, "loss": 0.9075, "step": 2996 }, { "epoch": 0.26, "learning_rate": 8.722466622440644e-06, "loss": 0.9007, "step": 2997 }, { "epoch": 0.26, "learning_rate": 8.721543888039534e-06, "loss": 0.8746, "step": 2998 }, { "epoch": 0.26, "learning_rate": 8.72062086936833e-06, "loss": 0.8881, "step": 2999 }, { "epoch": 0.26, "learning_rate": 8.719697566497545e-06, "loss": 0.8822, "step": 3000 }, { "epoch": 0.26, "learning_rate": 8.718773979497697e-06, "loss": 0.9025, "step": 3001 }, { "epoch": 0.26, "learning_rate": 8.71785010843934e-06, "loss": 0.9109, "step": 3002 }, { "epoch": 0.26, "learning_rate": 8.716925953393041e-06, "loss": 0.9058, "step": 3003 }, { "epoch": 0.26, "learning_rate": 8.716001514429393e-06, "loss": 0.8566, "step": 3004 }, { "epoch": 0.26, "learning_rate": 8.71507679161901e-06, "loss": 0.8409, "step": 3005 }, { "epoch": 0.26, "learning_rate": 8.714151785032524e-06, "loss": 0.8048, "step": 3006 }, { "epoch": 0.26, "learning_rate": 8.713226494740594e-06, "loss": 0.8489, "step": 3007 }, { "epoch": 0.26, "learning_rate": 8.712300920813896e-06, "loss": 0.9124, "step": 3008 }, { "epoch": 0.26, "learning_rate": 8.711375063323134e-06, "loss": 0.9025, "step": 3009 }, { "epoch": 0.26, "learning_rate": 8.710448922339025e-06, "loss": 0.8718, "step": 3010 }, { "epoch": 0.26, "learning_rate": 8.709522497932316e-06, "loss": 0.8964, "step": 3011 }, { "epoch": 0.26, "learning_rate": 8.70859579017377e-06, "loss": 0.9502, "step": 3012 }, { "epoch": 0.26, "learning_rate": 8.707668799134173e-06, "loss": 0.9184, "step": 3013 }, { "epoch": 0.26, "learning_rate": 8.706741524884335e-06, "loss": 0.8863, "step": 3014 }, { "epoch": 0.26, "learning_rate": 8.705813967495084e-06, "loss": 0.5943, "step": 3015 }, { "epoch": 0.26, "learning_rate": 8.704886127037275e-06, "loss": 0.883, "step": 3016 }, { "epoch": 0.26, "learning_rate": 8.703958003581776e-06, "loss": 0.6494, "step": 3017 }, { "epoch": 0.26, "learning_rate": 8.703029597199486e-06, "loss": 0.8698, "step": 3018 }, { "epoch": 0.26, "learning_rate": 8.702100907961317e-06, "loss": 0.8696, "step": 3019 }, { "epoch": 0.26, "learning_rate": 8.701171935938211e-06, "loss": 0.8914, "step": 3020 }, { "epoch": 0.26, "learning_rate": 8.700242681201125e-06, "loss": 0.5308, "step": 3021 }, { "epoch": 0.26, "learning_rate": 8.699313143821042e-06, "loss": 0.9361, "step": 3022 }, { "epoch": 0.26, "learning_rate": 8.698383323868965e-06, "loss": 0.847, "step": 3023 }, { "epoch": 0.26, "learning_rate": 8.697453221415914e-06, "loss": 0.8186, "step": 3024 }, { "epoch": 0.26, "learning_rate": 8.69652283653294e-06, "loss": 0.5525, "step": 3025 }, { "epoch": 0.26, "learning_rate": 8.69559216929111e-06, "loss": 0.9137, "step": 3026 }, { "epoch": 0.26, "learning_rate": 8.694661219761507e-06, "loss": 0.8611, "step": 3027 }, { "epoch": 0.26, "learning_rate": 8.69372998801525e-06, "loss": 0.9257, "step": 3028 }, { "epoch": 0.26, "learning_rate": 8.692798474123465e-06, "loss": 0.8422, "step": 3029 }, { "epoch": 0.26, "learning_rate": 8.691866678157306e-06, "loss": 0.8665, "step": 3030 }, { "epoch": 0.26, "learning_rate": 8.690934600187954e-06, "loss": 0.8797, "step": 3031 }, { "epoch": 0.26, "learning_rate": 8.6900022402866e-06, "loss": 0.8889, "step": 3032 }, { "epoch": 0.26, "learning_rate": 8.689069598524463e-06, "loss": 0.9027, "step": 3033 }, { "epoch": 0.26, "learning_rate": 8.688136674972784e-06, "loss": 0.8935, "step": 3034 }, { "epoch": 0.26, "learning_rate": 8.687203469702826e-06, "loss": 0.8412, "step": 3035 }, { "epoch": 0.26, "learning_rate": 8.68626998278587e-06, "loss": 0.889, "step": 3036 }, { "epoch": 0.26, "learning_rate": 8.685336214293218e-06, "loss": 0.9133, "step": 3037 }, { "epoch": 0.26, "learning_rate": 8.684402164296199e-06, "loss": 0.8718, "step": 3038 }, { "epoch": 0.26, "learning_rate": 8.68346783286616e-06, "loss": 0.8795, "step": 3039 }, { "epoch": 0.26, "learning_rate": 8.682533220074471e-06, "loss": 0.8556, "step": 3040 }, { "epoch": 0.26, "learning_rate": 8.68159832599252e-06, "loss": 0.8797, "step": 3041 }, { "epoch": 0.26, "learning_rate": 8.68066315069172e-06, "loss": 0.9119, "step": 3042 }, { "epoch": 0.26, "learning_rate": 8.679727694243504e-06, "loss": 0.843, "step": 3043 }, { "epoch": 0.26, "learning_rate": 8.678791956719327e-06, "loss": 0.8667, "step": 3044 }, { "epoch": 0.26, "learning_rate": 8.677855938190668e-06, "loss": 0.835, "step": 3045 }, { "epoch": 0.26, "learning_rate": 8.67691963872902e-06, "loss": 0.9433, "step": 3046 }, { "epoch": 0.26, "learning_rate": 8.675983058405905e-06, "loss": 0.8849, "step": 3047 }, { "epoch": 0.26, "learning_rate": 8.675046197292862e-06, "loss": 0.8833, "step": 3048 }, { "epoch": 0.26, "learning_rate": 8.674109055461454e-06, "loss": 0.9188, "step": 3049 }, { "epoch": 0.26, "learning_rate": 8.673171632983265e-06, "loss": 0.5409, "step": 3050 }, { "epoch": 0.26, "learning_rate": 8.672233929929904e-06, "loss": 0.8472, "step": 3051 }, { "epoch": 0.26, "learning_rate": 8.671295946372989e-06, "loss": 0.8483, "step": 3052 }, { "epoch": 0.26, "learning_rate": 8.670357682384172e-06, "loss": 0.8649, "step": 3053 }, { "epoch": 0.26, "learning_rate": 8.669419138035124e-06, "loss": 0.8837, "step": 3054 }, { "epoch": 0.26, "learning_rate": 8.668480313397532e-06, "loss": 0.8256, "step": 3055 }, { "epoch": 0.26, "learning_rate": 8.667541208543112e-06, "loss": 0.8683, "step": 3056 }, { "epoch": 0.26, "learning_rate": 8.666601823543594e-06, "loss": 0.8992, "step": 3057 }, { "epoch": 0.26, "learning_rate": 8.665662158470737e-06, "loss": 0.8435, "step": 3058 }, { "epoch": 0.26, "learning_rate": 8.664722213396314e-06, "loss": 0.8595, "step": 3059 }, { "epoch": 0.26, "learning_rate": 8.663781988392122e-06, "loss": 0.8714, "step": 3060 }, { "epoch": 0.26, "learning_rate": 8.662841483529984e-06, "loss": 0.8056, "step": 3061 }, { "epoch": 0.26, "learning_rate": 8.661900698881737e-06, "loss": 0.8864, "step": 3062 }, { "epoch": 0.26, "learning_rate": 8.660959634519244e-06, "loss": 0.8722, "step": 3063 }, { "epoch": 0.26, "learning_rate": 8.660018290514389e-06, "loss": 0.9087, "step": 3064 }, { "epoch": 0.26, "learning_rate": 8.659076666939075e-06, "loss": 0.8649, "step": 3065 }, { "epoch": 0.26, "learning_rate": 8.658134763865227e-06, "loss": 0.8505, "step": 3066 }, { "epoch": 0.26, "learning_rate": 8.657192581364798e-06, "loss": 0.8361, "step": 3067 }, { "epoch": 0.26, "learning_rate": 8.656250119509747e-06, "loss": 0.8662, "step": 3068 }, { "epoch": 0.26, "learning_rate": 8.655307378372073e-06, "loss": 0.8654, "step": 3069 }, { "epoch": 0.26, "learning_rate": 8.654364358023784e-06, "loss": 0.8415, "step": 3070 }, { "epoch": 0.26, "learning_rate": 8.65342105853691e-06, "loss": 0.8343, "step": 3071 }, { "epoch": 0.26, "learning_rate": 8.65247747998351e-06, "loss": 0.8835, "step": 3072 }, { "epoch": 0.26, "learning_rate": 8.651533622435655e-06, "loss": 0.9152, "step": 3073 }, { "epoch": 0.26, "learning_rate": 8.650589485965443e-06, "loss": 0.5941, "step": 3074 }, { "epoch": 0.26, "learning_rate": 8.649645070644991e-06, "loss": 0.9142, "step": 3075 }, { "epoch": 0.26, "learning_rate": 8.64870037654644e-06, "loss": 0.9301, "step": 3076 }, { "epoch": 0.26, "learning_rate": 8.64775540374195e-06, "loss": 0.8637, "step": 3077 }, { "epoch": 0.26, "learning_rate": 8.646810152303702e-06, "loss": 0.9381, "step": 3078 }, { "epoch": 0.26, "learning_rate": 8.6458646223039e-06, "loss": 0.8596, "step": 3079 }, { "epoch": 0.26, "learning_rate": 8.644918813814765e-06, "loss": 0.8883, "step": 3080 }, { "epoch": 0.26, "learning_rate": 8.643972726908545e-06, "loss": 0.8687, "step": 3081 }, { "epoch": 0.26, "learning_rate": 8.64302636165751e-06, "loss": 0.8428, "step": 3082 }, { "epoch": 0.26, "learning_rate": 8.642079718133942e-06, "loss": 0.8735, "step": 3083 }, { "epoch": 0.26, "learning_rate": 8.641132796410154e-06, "loss": 0.8857, "step": 3084 }, { "epoch": 0.26, "learning_rate": 8.640185596558477e-06, "loss": 0.8099, "step": 3085 }, { "epoch": 0.26, "learning_rate": 8.639238118651261e-06, "loss": 0.8414, "step": 3086 }, { "epoch": 0.26, "learning_rate": 8.638290362760882e-06, "loss": 0.8762, "step": 3087 }, { "epoch": 0.26, "learning_rate": 8.63734232895973e-06, "loss": 0.8779, "step": 3088 }, { "epoch": 0.26, "learning_rate": 8.636394017320224e-06, "loss": 0.9215, "step": 3089 }, { "epoch": 0.26, "learning_rate": 8.635445427914797e-06, "loss": 0.9105, "step": 3090 }, { "epoch": 0.26, "learning_rate": 8.634496560815912e-06, "loss": 0.8664, "step": 3091 }, { "epoch": 0.26, "learning_rate": 8.633547416096046e-06, "loss": 0.5659, "step": 3092 }, { "epoch": 0.26, "learning_rate": 8.632597993827699e-06, "loss": 0.948, "step": 3093 }, { "epoch": 0.26, "learning_rate": 8.631648294083392e-06, "loss": 0.896, "step": 3094 }, { "epoch": 0.26, "learning_rate": 8.630698316935668e-06, "loss": 0.9084, "step": 3095 }, { "epoch": 0.26, "learning_rate": 8.629748062457093e-06, "loss": 0.8408, "step": 3096 }, { "epoch": 0.26, "learning_rate": 8.628797530720251e-06, "loss": 0.8737, "step": 3097 }, { "epoch": 0.26, "learning_rate": 8.627846721797747e-06, "loss": 0.8635, "step": 3098 }, { "epoch": 0.26, "learning_rate": 8.62689563576221e-06, "loss": 0.58, "step": 3099 }, { "epoch": 0.26, "learning_rate": 8.62594427268629e-06, "loss": 0.8443, "step": 3100 }, { "epoch": 0.26, "learning_rate": 8.624992632642653e-06, "loss": 0.8588, "step": 3101 }, { "epoch": 0.26, "learning_rate": 8.624040715703992e-06, "loss": 0.9058, "step": 3102 }, { "epoch": 0.26, "learning_rate": 8.62308852194302e-06, "loss": 0.8942, "step": 3103 }, { "epoch": 0.26, "learning_rate": 8.622136051432471e-06, "loss": 0.9025, "step": 3104 }, { "epoch": 0.26, "learning_rate": 8.621183304245095e-06, "loss": 0.8778, "step": 3105 }, { "epoch": 0.27, "learning_rate": 8.620230280453672e-06, "loss": 0.8657, "step": 3106 }, { "epoch": 0.27, "learning_rate": 8.619276980131e-06, "loss": 0.864, "step": 3107 }, { "epoch": 0.27, "learning_rate": 8.618323403349891e-06, "loss": 0.9063, "step": 3108 }, { "epoch": 0.27, "learning_rate": 8.61736955018319e-06, "loss": 0.8919, "step": 3109 }, { "epoch": 0.27, "learning_rate": 8.616415420703752e-06, "loss": 0.918, "step": 3110 }, { "epoch": 0.27, "learning_rate": 8.615461014984463e-06, "loss": 0.8796, "step": 3111 }, { "epoch": 0.27, "learning_rate": 8.61450633309822e-06, "loss": 0.8776, "step": 3112 }, { "epoch": 0.27, "learning_rate": 8.61355137511795e-06, "loss": 0.8476, "step": 3113 }, { "epoch": 0.27, "learning_rate": 8.612596141116597e-06, "loss": 0.8923, "step": 3114 }, { "epoch": 0.27, "learning_rate": 8.611640631167126e-06, "loss": 0.8209, "step": 3115 }, { "epoch": 0.27, "learning_rate": 8.610684845342524e-06, "loss": 0.8341, "step": 3116 }, { "epoch": 0.27, "learning_rate": 8.609728783715798e-06, "loss": 0.8905, "step": 3117 }, { "epoch": 0.27, "learning_rate": 8.608772446359976e-06, "loss": 0.9011, "step": 3118 }, { "epoch": 0.27, "learning_rate": 8.60781583334811e-06, "loss": 0.908, "step": 3119 }, { "epoch": 0.27, "learning_rate": 8.60685894475327e-06, "loss": 0.8318, "step": 3120 }, { "epoch": 0.27, "learning_rate": 8.605901780648549e-06, "loss": 0.9228, "step": 3121 }, { "epoch": 0.27, "learning_rate": 8.604944341107058e-06, "loss": 0.863, "step": 3122 }, { "epoch": 0.27, "learning_rate": 8.60398662620193e-06, "loss": 0.8785, "step": 3123 }, { "epoch": 0.27, "learning_rate": 8.603028636006324e-06, "loss": 0.8694, "step": 3124 }, { "epoch": 0.27, "learning_rate": 8.602070370593413e-06, "loss": 0.8609, "step": 3125 }, { "epoch": 0.27, "learning_rate": 8.601111830036396e-06, "loss": 0.8615, "step": 3126 }, { "epoch": 0.27, "learning_rate": 8.600153014408487e-06, "loss": 0.8678, "step": 3127 }, { "epoch": 0.27, "learning_rate": 8.599193923782931e-06, "loss": 0.8355, "step": 3128 }, { "epoch": 0.27, "learning_rate": 8.598234558232984e-06, "loss": 0.8174, "step": 3129 }, { "epoch": 0.27, "learning_rate": 8.597274917831929e-06, "loss": 0.8827, "step": 3130 }, { "epoch": 0.27, "learning_rate": 8.596315002653067e-06, "loss": 0.8803, "step": 3131 }, { "epoch": 0.27, "learning_rate": 8.595354812769722e-06, "loss": 0.8403, "step": 3132 }, { "epoch": 0.27, "learning_rate": 8.594394348255239e-06, "loss": 0.8609, "step": 3133 }, { "epoch": 0.27, "learning_rate": 8.59343360918298e-06, "loss": 0.8672, "step": 3134 }, { "epoch": 0.27, "learning_rate": 8.592472595626333e-06, "loss": 0.8982, "step": 3135 }, { "epoch": 0.27, "learning_rate": 8.591511307658706e-06, "loss": 0.8727, "step": 3136 }, { "epoch": 0.27, "learning_rate": 8.590549745353528e-06, "loss": 0.9081, "step": 3137 }, { "epoch": 0.27, "learning_rate": 8.589587908784242e-06, "loss": 0.8681, "step": 3138 }, { "epoch": 0.27, "learning_rate": 8.588625798024323e-06, "loss": 0.8293, "step": 3139 }, { "epoch": 0.27, "learning_rate": 8.58766341314726e-06, "loss": 0.8924, "step": 3140 }, { "epoch": 0.27, "learning_rate": 8.586700754226566e-06, "loss": 0.8853, "step": 3141 }, { "epoch": 0.27, "learning_rate": 8.585737821335773e-06, "loss": 0.8975, "step": 3142 }, { "epoch": 0.27, "learning_rate": 8.584774614548433e-06, "loss": 0.9213, "step": 3143 }, { "epoch": 0.27, "learning_rate": 8.583811133938124e-06, "loss": 0.904, "step": 3144 }, { "epoch": 0.27, "learning_rate": 8.582847379578439e-06, "loss": 0.8924, "step": 3145 }, { "epoch": 0.27, "learning_rate": 8.581883351542995e-06, "loss": 0.9511, "step": 3146 }, { "epoch": 0.27, "learning_rate": 8.580919049905428e-06, "loss": 0.8791, "step": 3147 }, { "epoch": 0.27, "learning_rate": 8.579954474739399e-06, "loss": 0.8278, "step": 3148 }, { "epoch": 0.27, "learning_rate": 8.578989626118584e-06, "loss": 0.8635, "step": 3149 }, { "epoch": 0.27, "learning_rate": 8.578024504116687e-06, "loss": 0.8901, "step": 3150 }, { "epoch": 0.27, "learning_rate": 8.577059108807422e-06, "loss": 0.869, "step": 3151 }, { "epoch": 0.27, "learning_rate": 8.57609344026454e-06, "loss": 0.8873, "step": 3152 }, { "epoch": 0.27, "learning_rate": 8.575127498561794e-06, "loss": 0.8882, "step": 3153 }, { "epoch": 0.27, "learning_rate": 8.574161283772975e-06, "loss": 0.8815, "step": 3154 }, { "epoch": 0.27, "learning_rate": 8.573194795971882e-06, "loss": 0.862, "step": 3155 }, { "epoch": 0.27, "learning_rate": 8.572228035232345e-06, "loss": 0.9341, "step": 3156 }, { "epoch": 0.27, "learning_rate": 8.571261001628204e-06, "loss": 0.8629, "step": 3157 }, { "epoch": 0.27, "learning_rate": 8.570293695233331e-06, "loss": 0.8497, "step": 3158 }, { "epoch": 0.27, "learning_rate": 8.569326116121614e-06, "loss": 0.8895, "step": 3159 }, { "epoch": 0.27, "learning_rate": 8.568358264366958e-06, "loss": 0.8897, "step": 3160 }, { "epoch": 0.27, "learning_rate": 8.567390140043292e-06, "loss": 0.8988, "step": 3161 }, { "epoch": 0.27, "learning_rate": 8.56642174322457e-06, "loss": 0.8671, "step": 3162 }, { "epoch": 0.27, "learning_rate": 8.56545307398476e-06, "loss": 0.8562, "step": 3163 }, { "epoch": 0.27, "learning_rate": 8.564484132397855e-06, "loss": 0.9035, "step": 3164 }, { "epoch": 0.27, "learning_rate": 8.563514918537868e-06, "loss": 0.8615, "step": 3165 }, { "epoch": 0.27, "learning_rate": 8.562545432478832e-06, "loss": 0.8952, "step": 3166 }, { "epoch": 0.27, "learning_rate": 8.5615756742948e-06, "loss": 0.8225, "step": 3167 }, { "epoch": 0.27, "learning_rate": 8.56060564405985e-06, "loss": 0.8556, "step": 3168 }, { "epoch": 0.27, "learning_rate": 8.559635341848072e-06, "loss": 0.8762, "step": 3169 }, { "epoch": 0.27, "learning_rate": 8.558664767733589e-06, "loss": 0.8569, "step": 3170 }, { "epoch": 0.27, "learning_rate": 8.557693921790534e-06, "loss": 0.8722, "step": 3171 }, { "epoch": 0.27, "learning_rate": 8.556722804093067e-06, "loss": 0.8896, "step": 3172 }, { "epoch": 0.27, "learning_rate": 8.555751414715367e-06, "loss": 0.8864, "step": 3173 }, { "epoch": 0.27, "learning_rate": 8.554779753731632e-06, "loss": 0.6232, "step": 3174 }, { "epoch": 0.27, "learning_rate": 8.553807821216083e-06, "loss": 0.8715, "step": 3175 }, { "epoch": 0.27, "learning_rate": 8.552835617242963e-06, "loss": 0.8746, "step": 3176 }, { "epoch": 0.27, "learning_rate": 8.55186314188653e-06, "loss": 0.8998, "step": 3177 }, { "epoch": 0.27, "learning_rate": 8.55089039522107e-06, "loss": 0.9209, "step": 3178 }, { "epoch": 0.27, "learning_rate": 8.549917377320885e-06, "loss": 0.9044, "step": 3179 }, { "epoch": 0.27, "learning_rate": 8.548944088260299e-06, "loss": 0.8258, "step": 3180 }, { "epoch": 0.27, "learning_rate": 8.547970528113653e-06, "loss": 0.8969, "step": 3181 }, { "epoch": 0.27, "learning_rate": 8.54699669695532e-06, "loss": 0.9289, "step": 3182 }, { "epoch": 0.27, "learning_rate": 8.54602259485968e-06, "loss": 0.8726, "step": 3183 }, { "epoch": 0.27, "learning_rate": 8.545048221901143e-06, "loss": 0.8864, "step": 3184 }, { "epoch": 0.27, "learning_rate": 8.544073578154133e-06, "loss": 0.8902, "step": 3185 }, { "epoch": 0.27, "learning_rate": 8.543098663693102e-06, "loss": 0.8344, "step": 3186 }, { "epoch": 0.27, "learning_rate": 8.542123478592518e-06, "loss": 0.8605, "step": 3187 }, { "epoch": 0.27, "learning_rate": 8.541148022926868e-06, "loss": 0.9197, "step": 3188 }, { "epoch": 0.27, "learning_rate": 8.540172296770663e-06, "loss": 0.8906, "step": 3189 }, { "epoch": 0.27, "learning_rate": 8.539196300198438e-06, "loss": 0.606, "step": 3190 }, { "epoch": 0.27, "learning_rate": 8.53822003328474e-06, "loss": 0.891, "step": 3191 }, { "epoch": 0.27, "learning_rate": 8.53724349610414e-06, "loss": 0.8864, "step": 3192 }, { "epoch": 0.27, "learning_rate": 8.536266688731234e-06, "loss": 0.872, "step": 3193 }, { "epoch": 0.27, "learning_rate": 8.535289611240636e-06, "loss": 0.8865, "step": 3194 }, { "epoch": 0.27, "learning_rate": 8.534312263706978e-06, "loss": 0.8713, "step": 3195 }, { "epoch": 0.27, "learning_rate": 8.533334646204915e-06, "loss": 0.9153, "step": 3196 }, { "epoch": 0.27, "learning_rate": 8.532356758809122e-06, "loss": 0.9339, "step": 3197 }, { "epoch": 0.27, "learning_rate": 8.531378601594296e-06, "loss": 0.869, "step": 3198 }, { "epoch": 0.27, "learning_rate": 8.530400174635153e-06, "loss": 0.8743, "step": 3199 }, { "epoch": 0.27, "learning_rate": 8.52942147800643e-06, "loss": 0.8964, "step": 3200 }, { "epoch": 0.27, "learning_rate": 8.528442511782885e-06, "loss": 0.8666, "step": 3201 }, { "epoch": 0.27, "learning_rate": 8.527463276039296e-06, "loss": 0.8635, "step": 3202 }, { "epoch": 0.27, "learning_rate": 8.526483770850462e-06, "loss": 0.8648, "step": 3203 }, { "epoch": 0.27, "learning_rate": 8.525503996291203e-06, "loss": 0.8861, "step": 3204 }, { "epoch": 0.27, "learning_rate": 8.524523952436357e-06, "loss": 0.8742, "step": 3205 }, { "epoch": 0.27, "learning_rate": 8.523543639360789e-06, "loss": 0.8647, "step": 3206 }, { "epoch": 0.27, "learning_rate": 8.522563057139376e-06, "loss": 0.8983, "step": 3207 }, { "epoch": 0.27, "learning_rate": 8.521582205847022e-06, "loss": 0.8789, "step": 3208 }, { "epoch": 0.27, "learning_rate": 8.520601085558648e-06, "loss": 0.885, "step": 3209 }, { "epoch": 0.27, "learning_rate": 8.519619696349197e-06, "loss": 0.8841, "step": 3210 }, { "epoch": 0.27, "learning_rate": 8.518638038293633e-06, "loss": 0.8774, "step": 3211 }, { "epoch": 0.27, "learning_rate": 8.517656111466942e-06, "loss": 0.8145, "step": 3212 }, { "epoch": 0.27, "learning_rate": 8.516673915944123e-06, "loss": 0.8892, "step": 3213 }, { "epoch": 0.27, "learning_rate": 8.515691451800206e-06, "loss": 0.8957, "step": 3214 }, { "epoch": 0.27, "learning_rate": 8.514708719110234e-06, "loss": 0.8249, "step": 3215 }, { "epoch": 0.27, "learning_rate": 8.513725717949274e-06, "loss": 0.8413, "step": 3216 }, { "epoch": 0.27, "learning_rate": 8.512742448392412e-06, "loss": 0.8741, "step": 3217 }, { "epoch": 0.27, "learning_rate": 8.511758910514755e-06, "loss": 0.8843, "step": 3218 }, { "epoch": 0.27, "learning_rate": 8.510775104391434e-06, "loss": 0.8435, "step": 3219 }, { "epoch": 0.27, "learning_rate": 8.50979103009759e-06, "loss": 0.8755, "step": 3220 }, { "epoch": 0.27, "learning_rate": 8.508806687708398e-06, "loss": 0.8733, "step": 3221 }, { "epoch": 0.27, "learning_rate": 8.507822077299043e-06, "loss": 0.9132, "step": 3222 }, { "epoch": 0.28, "learning_rate": 8.506837198944735e-06, "loss": 0.9511, "step": 3223 }, { "epoch": 0.28, "learning_rate": 8.505852052720706e-06, "loss": 0.888, "step": 3224 }, { "epoch": 0.28, "learning_rate": 8.504866638702205e-06, "loss": 0.5346, "step": 3225 }, { "epoch": 0.28, "learning_rate": 8.503880956964504e-06, "loss": 0.8844, "step": 3226 }, { "epoch": 0.28, "learning_rate": 8.502895007582892e-06, "loss": 0.8369, "step": 3227 }, { "epoch": 0.28, "learning_rate": 8.501908790632682e-06, "loss": 0.9006, "step": 3228 }, { "epoch": 0.28, "learning_rate": 8.500922306189207e-06, "loss": 0.8719, "step": 3229 }, { "epoch": 0.28, "learning_rate": 8.499935554327818e-06, "loss": 0.9442, "step": 3230 }, { "epoch": 0.28, "learning_rate": 8.49894853512389e-06, "loss": 0.8854, "step": 3231 }, { "epoch": 0.28, "learning_rate": 8.497961248652816e-06, "loss": 0.862, "step": 3232 }, { "epoch": 0.28, "learning_rate": 8.496973694990004e-06, "loss": 0.8791, "step": 3233 }, { "epoch": 0.28, "learning_rate": 8.4959858742109e-06, "loss": 0.9045, "step": 3234 }, { "epoch": 0.28, "learning_rate": 8.494997786390949e-06, "loss": 0.8424, "step": 3235 }, { "epoch": 0.28, "learning_rate": 8.494009431605628e-06, "loss": 0.9001, "step": 3236 }, { "epoch": 0.28, "learning_rate": 8.493020809930434e-06, "loss": 0.8626, "step": 3237 }, { "epoch": 0.28, "learning_rate": 8.492031921440884e-06, "loss": 0.8876, "step": 3238 }, { "epoch": 0.28, "learning_rate": 8.491042766212515e-06, "loss": 0.8576, "step": 3239 }, { "epoch": 0.28, "learning_rate": 8.490053344320877e-06, "loss": 0.524, "step": 3240 }, { "epoch": 0.28, "learning_rate": 8.489063655841552e-06, "loss": 0.8644, "step": 3241 }, { "epoch": 0.28, "learning_rate": 8.488073700850138e-06, "loss": 0.8575, "step": 3242 }, { "epoch": 0.28, "learning_rate": 8.487083479422251e-06, "loss": 0.8788, "step": 3243 }, { "epoch": 0.28, "learning_rate": 8.48609299163353e-06, "loss": 0.9376, "step": 3244 }, { "epoch": 0.28, "learning_rate": 8.485102237559631e-06, "loss": 0.8892, "step": 3245 }, { "epoch": 0.28, "learning_rate": 8.484111217276237e-06, "loss": 0.9047, "step": 3246 }, { "epoch": 0.28, "learning_rate": 8.483119930859042e-06, "loss": 0.8598, "step": 3247 }, { "epoch": 0.28, "learning_rate": 8.48212837838377e-06, "loss": 0.8688, "step": 3248 }, { "epoch": 0.28, "learning_rate": 8.481136559926157e-06, "loss": 0.9158, "step": 3249 }, { "epoch": 0.28, "learning_rate": 8.480144475561967e-06, "loss": 0.8818, "step": 3250 }, { "epoch": 0.28, "learning_rate": 8.479152125366977e-06, "loss": 0.898, "step": 3251 }, { "epoch": 0.28, "learning_rate": 8.478159509416988e-06, "loss": 0.8546, "step": 3252 }, { "epoch": 0.28, "learning_rate": 8.477166627787823e-06, "loss": 0.566, "step": 3253 }, { "epoch": 0.28, "learning_rate": 8.476173480555321e-06, "loss": 0.8591, "step": 3254 }, { "epoch": 0.28, "learning_rate": 8.475180067795345e-06, "loss": 0.8518, "step": 3255 }, { "epoch": 0.28, "learning_rate": 8.474186389583779e-06, "loss": 0.87, "step": 3256 }, { "epoch": 0.28, "learning_rate": 8.473192445996518e-06, "loss": 0.8889, "step": 3257 }, { "epoch": 0.28, "learning_rate": 8.472198237109492e-06, "loss": 0.8402, "step": 3258 }, { "epoch": 0.28, "learning_rate": 8.471203762998638e-06, "loss": 0.8574, "step": 3259 }, { "epoch": 0.28, "learning_rate": 8.470209023739925e-06, "loss": 0.9011, "step": 3260 }, { "epoch": 0.28, "learning_rate": 8.46921401940933e-06, "loss": 0.8936, "step": 3261 }, { "epoch": 0.28, "learning_rate": 8.46821875008286e-06, "loss": 0.8769, "step": 3262 }, { "epoch": 0.28, "learning_rate": 8.467223215836537e-06, "loss": 0.7649, "step": 3263 }, { "epoch": 0.28, "learning_rate": 8.466227416746404e-06, "loss": 0.873, "step": 3264 }, { "epoch": 0.28, "learning_rate": 8.465231352888528e-06, "loss": 0.8624, "step": 3265 }, { "epoch": 0.28, "learning_rate": 8.464235024338992e-06, "loss": 0.9127, "step": 3266 }, { "epoch": 0.28, "learning_rate": 8.463238431173902e-06, "loss": 0.5517, "step": 3267 }, { "epoch": 0.28, "learning_rate": 8.462241573469378e-06, "loss": 0.9038, "step": 3268 }, { "epoch": 0.28, "learning_rate": 8.461244451301572e-06, "loss": 0.8851, "step": 3269 }, { "epoch": 0.28, "learning_rate": 8.46024706474664e-06, "loss": 0.9249, "step": 3270 }, { "epoch": 0.28, "learning_rate": 8.459249413880777e-06, "loss": 0.8844, "step": 3271 }, { "epoch": 0.28, "learning_rate": 8.458251498780183e-06, "loss": 0.8751, "step": 3272 }, { "epoch": 0.28, "learning_rate": 8.457253319521085e-06, "loss": 0.864, "step": 3273 }, { "epoch": 0.28, "learning_rate": 8.456254876179729e-06, "loss": 0.902, "step": 3274 }, { "epoch": 0.28, "learning_rate": 8.45525616883238e-06, "loss": 0.8147, "step": 3275 }, { "epoch": 0.28, "learning_rate": 8.454257197555326e-06, "loss": 0.9165, "step": 3276 }, { "epoch": 0.28, "learning_rate": 8.453257962424872e-06, "loss": 0.8751, "step": 3277 }, { "epoch": 0.28, "learning_rate": 8.452258463517346e-06, "loss": 0.8577, "step": 3278 }, { "epoch": 0.28, "learning_rate": 8.451258700909093e-06, "loss": 0.9009, "step": 3279 }, { "epoch": 0.28, "learning_rate": 8.450258674676482e-06, "loss": 0.8348, "step": 3280 }, { "epoch": 0.28, "learning_rate": 8.449258384895897e-06, "loss": 0.8862, "step": 3281 }, { "epoch": 0.28, "learning_rate": 8.448257831643749e-06, "loss": 0.638, "step": 3282 }, { "epoch": 0.28, "learning_rate": 8.447257014996461e-06, "loss": 0.8596, "step": 3283 }, { "epoch": 0.28, "learning_rate": 8.446255935030485e-06, "loss": 0.8389, "step": 3284 }, { "epoch": 0.28, "learning_rate": 8.445254591822284e-06, "loss": 0.9026, "step": 3285 }, { "epoch": 0.28, "learning_rate": 8.444252985448348e-06, "loss": 0.9279, "step": 3286 }, { "epoch": 0.28, "learning_rate": 8.443251115985184e-06, "loss": 0.8451, "step": 3287 }, { "epoch": 0.28, "learning_rate": 8.44224898350932e-06, "loss": 0.8774, "step": 3288 }, { "epoch": 0.28, "learning_rate": 8.441246588097305e-06, "loss": 0.8832, "step": 3289 }, { "epoch": 0.28, "learning_rate": 8.440243929825705e-06, "loss": 0.9254, "step": 3290 }, { "epoch": 0.28, "learning_rate": 8.439241008771108e-06, "loss": 0.8916, "step": 3291 }, { "epoch": 0.28, "learning_rate": 8.438237825010126e-06, "loss": 0.8575, "step": 3292 }, { "epoch": 0.28, "learning_rate": 8.437234378619381e-06, "loss": 0.902, "step": 3293 }, { "epoch": 0.28, "learning_rate": 8.436230669675527e-06, "loss": 0.8792, "step": 3294 }, { "epoch": 0.28, "learning_rate": 8.435226698255228e-06, "loss": 0.8257, "step": 3295 }, { "epoch": 0.28, "learning_rate": 8.434222464435173e-06, "loss": 0.8754, "step": 3296 }, { "epoch": 0.28, "learning_rate": 8.433217968292074e-06, "loss": 0.9092, "step": 3297 }, { "epoch": 0.28, "learning_rate": 8.432213209902656e-06, "loss": 0.882, "step": 3298 }, { "epoch": 0.28, "learning_rate": 8.43120818934367e-06, "loss": 0.9166, "step": 3299 }, { "epoch": 0.28, "learning_rate": 8.43020290669188e-06, "loss": 0.8706, "step": 3300 }, { "epoch": 0.28, "learning_rate": 8.429197362024078e-06, "loss": 0.8817, "step": 3301 }, { "epoch": 0.28, "learning_rate": 8.428191555417073e-06, "loss": 0.896, "step": 3302 }, { "epoch": 0.28, "learning_rate": 8.427185486947691e-06, "loss": 0.8477, "step": 3303 }, { "epoch": 0.28, "learning_rate": 8.426179156692784e-06, "loss": 0.8871, "step": 3304 }, { "epoch": 0.28, "learning_rate": 8.425172564729218e-06, "loss": 0.8486, "step": 3305 }, { "epoch": 0.28, "learning_rate": 8.42416571113388e-06, "loss": 0.8537, "step": 3306 }, { "epoch": 0.28, "learning_rate": 8.423158595983682e-06, "loss": 0.9017, "step": 3307 }, { "epoch": 0.28, "learning_rate": 8.422151219355553e-06, "loss": 0.9238, "step": 3308 }, { "epoch": 0.28, "learning_rate": 8.421143581326436e-06, "loss": 0.906, "step": 3309 }, { "epoch": 0.28, "learning_rate": 8.420135681973304e-06, "loss": 0.8987, "step": 3310 }, { "epoch": 0.28, "learning_rate": 8.419127521373146e-06, "loss": 0.9079, "step": 3311 }, { "epoch": 0.28, "learning_rate": 8.418119099602965e-06, "loss": 0.8925, "step": 3312 }, { "epoch": 0.28, "learning_rate": 8.417110416739795e-06, "loss": 0.8514, "step": 3313 }, { "epoch": 0.28, "learning_rate": 8.416101472860681e-06, "loss": 0.9117, "step": 3314 }, { "epoch": 0.28, "learning_rate": 8.415092268042693e-06, "loss": 0.9153, "step": 3315 }, { "epoch": 0.28, "learning_rate": 8.41408280236292e-06, "loss": 0.8913, "step": 3316 }, { "epoch": 0.28, "learning_rate": 8.413073075898465e-06, "loss": 0.849, "step": 3317 }, { "epoch": 0.28, "learning_rate": 8.412063088726461e-06, "loss": 0.9068, "step": 3318 }, { "epoch": 0.28, "learning_rate": 8.411052840924056e-06, "loss": 0.8795, "step": 3319 }, { "epoch": 0.28, "learning_rate": 8.410042332568414e-06, "loss": 0.8417, "step": 3320 }, { "epoch": 0.28, "learning_rate": 8.409031563736724e-06, "loss": 0.835, "step": 3321 }, { "epoch": 0.28, "learning_rate": 8.408020534506195e-06, "loss": 0.8814, "step": 3322 }, { "epoch": 0.28, "learning_rate": 8.407009244954055e-06, "loss": 0.8313, "step": 3323 }, { "epoch": 0.28, "learning_rate": 8.405997695157547e-06, "loss": 0.8838, "step": 3324 }, { "epoch": 0.28, "learning_rate": 8.404985885193944e-06, "loss": 0.8499, "step": 3325 }, { "epoch": 0.28, "learning_rate": 8.403973815140528e-06, "loss": 0.8483, "step": 3326 }, { "epoch": 0.28, "learning_rate": 8.40296148507461e-06, "loss": 0.8716, "step": 3327 }, { "epoch": 0.28, "learning_rate": 8.401948895073513e-06, "loss": 0.8514, "step": 3328 }, { "epoch": 0.28, "learning_rate": 8.400936045214588e-06, "loss": 0.9411, "step": 3329 }, { "epoch": 0.28, "learning_rate": 8.399922935575198e-06, "loss": 0.86, "step": 3330 }, { "epoch": 0.28, "learning_rate": 8.398909566232731e-06, "loss": 0.8432, "step": 3331 }, { "epoch": 0.28, "learning_rate": 8.397895937264593e-06, "loss": 0.8423, "step": 3332 }, { "epoch": 0.28, "learning_rate": 8.39688204874821e-06, "loss": 0.8849, "step": 3333 }, { "epoch": 0.28, "learning_rate": 8.395867900761028e-06, "loss": 0.829, "step": 3334 }, { "epoch": 0.28, "learning_rate": 8.394853493380513e-06, "loss": 0.9064, "step": 3335 }, { "epoch": 0.28, "learning_rate": 8.393838826684148e-06, "loss": 0.8856, "step": 3336 }, { "epoch": 0.28, "learning_rate": 8.39282390074944e-06, "loss": 0.9043, "step": 3337 }, { "epoch": 0.28, "learning_rate": 8.391808715653916e-06, "loss": 0.8575, "step": 3338 }, { "epoch": 0.28, "learning_rate": 8.39079327147512e-06, "loss": 0.8693, "step": 3339 }, { "epoch": 0.29, "learning_rate": 8.389777568290614e-06, "loss": 0.8505, "step": 3340 }, { "epoch": 0.29, "learning_rate": 8.388761606177984e-06, "loss": 0.855, "step": 3341 }, { "epoch": 0.29, "learning_rate": 8.387745385214835e-06, "loss": 0.8707, "step": 3342 }, { "epoch": 0.29, "learning_rate": 8.386728905478792e-06, "loss": 0.8598, "step": 3343 }, { "epoch": 0.29, "learning_rate": 8.385712167047497e-06, "loss": 0.86, "step": 3344 }, { "epoch": 0.29, "learning_rate": 8.384695169998614e-06, "loss": 0.8633, "step": 3345 }, { "epoch": 0.29, "learning_rate": 8.383677914409827e-06, "loss": 0.6941, "step": 3346 }, { "epoch": 0.29, "learning_rate": 8.382660400358838e-06, "loss": 0.8135, "step": 3347 }, { "epoch": 0.29, "learning_rate": 8.38164262792337e-06, "loss": 0.8351, "step": 3348 }, { "epoch": 0.29, "learning_rate": 8.380624597181165e-06, "loss": 0.9114, "step": 3349 }, { "epoch": 0.29, "learning_rate": 8.379606308209987e-06, "loss": 0.9115, "step": 3350 }, { "epoch": 0.29, "learning_rate": 8.378587761087615e-06, "loss": 0.8246, "step": 3351 }, { "epoch": 0.29, "learning_rate": 8.377568955891856e-06, "loss": 0.9192, "step": 3352 }, { "epoch": 0.29, "learning_rate": 8.376549892700526e-06, "loss": 0.8714, "step": 3353 }, { "epoch": 0.29, "learning_rate": 8.375530571591468e-06, "loss": 0.9119, "step": 3354 }, { "epoch": 0.29, "learning_rate": 8.374510992642546e-06, "loss": 0.9154, "step": 3355 }, { "epoch": 0.29, "learning_rate": 8.373491155931634e-06, "loss": 0.8463, "step": 3356 }, { "epoch": 0.29, "learning_rate": 8.372471061536638e-06, "loss": 0.8514, "step": 3357 }, { "epoch": 0.29, "learning_rate": 8.371450709535476e-06, "loss": 0.8553, "step": 3358 }, { "epoch": 0.29, "learning_rate": 8.370430100006086e-06, "loss": 0.851, "step": 3359 }, { "epoch": 0.29, "learning_rate": 8.369409233026429e-06, "loss": 0.8389, "step": 3360 }, { "epoch": 0.29, "learning_rate": 8.368388108674483e-06, "loss": 0.8818, "step": 3361 }, { "epoch": 0.29, "learning_rate": 8.367366727028248e-06, "loss": 0.8706, "step": 3362 }, { "epoch": 0.29, "learning_rate": 8.366345088165741e-06, "loss": 0.8857, "step": 3363 }, { "epoch": 0.29, "learning_rate": 8.365323192165e-06, "loss": 0.8491, "step": 3364 }, { "epoch": 0.29, "learning_rate": 8.364301039104084e-06, "loss": 0.8606, "step": 3365 }, { "epoch": 0.29, "learning_rate": 8.363278629061066e-06, "loss": 0.871, "step": 3366 }, { "epoch": 0.29, "learning_rate": 8.362255962114047e-06, "loss": 0.8137, "step": 3367 }, { "epoch": 0.29, "learning_rate": 8.361233038341144e-06, "loss": 0.5348, "step": 3368 }, { "epoch": 0.29, "learning_rate": 8.360209857820487e-06, "loss": 0.8785, "step": 3369 }, { "epoch": 0.29, "learning_rate": 8.359186420630239e-06, "loss": 0.8746, "step": 3370 }, { "epoch": 0.29, "learning_rate": 8.35816272684857e-06, "loss": 0.8463, "step": 3371 }, { "epoch": 0.29, "learning_rate": 8.357138776553678e-06, "loss": 0.913, "step": 3372 }, { "epoch": 0.29, "learning_rate": 8.356114569823776e-06, "loss": 0.9292, "step": 3373 }, { "epoch": 0.29, "learning_rate": 8.355090106737098e-06, "loss": 0.8322, "step": 3374 }, { "epoch": 0.29, "learning_rate": 8.354065387371897e-06, "loss": 0.9254, "step": 3375 }, { "epoch": 0.29, "learning_rate": 8.353040411806449e-06, "loss": 0.8479, "step": 3376 }, { "epoch": 0.29, "learning_rate": 8.352015180119042e-06, "loss": 0.8792, "step": 3377 }, { "epoch": 0.29, "learning_rate": 8.350989692387992e-06, "loss": 0.856, "step": 3378 }, { "epoch": 0.29, "learning_rate": 8.349963948691631e-06, "loss": 0.879, "step": 3379 }, { "epoch": 0.29, "learning_rate": 8.348937949108309e-06, "loss": 0.8406, "step": 3380 }, { "epoch": 0.29, "learning_rate": 8.347911693716398e-06, "loss": 0.8551, "step": 3381 }, { "epoch": 0.29, "learning_rate": 8.346885182594286e-06, "loss": 0.8523, "step": 3382 }, { "epoch": 0.29, "learning_rate": 8.345858415820384e-06, "loss": 0.8976, "step": 3383 }, { "epoch": 0.29, "learning_rate": 8.344831393473125e-06, "loss": 0.8242, "step": 3384 }, { "epoch": 0.29, "learning_rate": 8.343804115630955e-06, "loss": 0.8435, "step": 3385 }, { "epoch": 0.29, "learning_rate": 8.342776582372343e-06, "loss": 0.8443, "step": 3386 }, { "epoch": 0.29, "learning_rate": 8.341748793775776e-06, "loss": 0.856, "step": 3387 }, { "epoch": 0.29, "learning_rate": 8.340720749919764e-06, "loss": 0.8694, "step": 3388 }, { "epoch": 0.29, "learning_rate": 8.339692450882833e-06, "loss": 0.9225, "step": 3389 }, { "epoch": 0.29, "learning_rate": 8.338663896743528e-06, "loss": 0.8772, "step": 3390 }, { "epoch": 0.29, "learning_rate": 8.337635087580419e-06, "loss": 0.8339, "step": 3391 }, { "epoch": 0.29, "learning_rate": 8.336606023472086e-06, "loss": 0.8377, "step": 3392 }, { "epoch": 0.29, "learning_rate": 8.335576704497141e-06, "loss": 0.9144, "step": 3393 }, { "epoch": 0.29, "learning_rate": 8.334547130734202e-06, "loss": 0.8681, "step": 3394 }, { "epoch": 0.29, "learning_rate": 8.333517302261919e-06, "loss": 0.8418, "step": 3395 }, { "epoch": 0.29, "learning_rate": 8.33248721915895e-06, "loss": 0.8695, "step": 3396 }, { "epoch": 0.29, "learning_rate": 8.33145688150398e-06, "loss": 0.8845, "step": 3397 }, { "epoch": 0.29, "learning_rate": 8.330426289375713e-06, "loss": 0.8422, "step": 3398 }, { "epoch": 0.29, "learning_rate": 8.329395442852869e-06, "loss": 0.8554, "step": 3399 }, { "epoch": 0.29, "learning_rate": 8.328364342014185e-06, "loss": 0.8857, "step": 3400 }, { "epoch": 0.29, "learning_rate": 8.327332986938433e-06, "loss": 0.8374, "step": 3401 }, { "epoch": 0.29, "learning_rate": 8.326301377704381e-06, "loss": 0.9005, "step": 3402 }, { "epoch": 0.29, "learning_rate": 8.325269514390835e-06, "loss": 0.8769, "step": 3403 }, { "epoch": 0.29, "learning_rate": 8.324237397076613e-06, "loss": 0.8376, "step": 3404 }, { "epoch": 0.29, "learning_rate": 8.323205025840553e-06, "loss": 0.9113, "step": 3405 }, { "epoch": 0.29, "learning_rate": 8.322172400761511e-06, "loss": 0.8506, "step": 3406 }, { "epoch": 0.29, "learning_rate": 8.321139521918367e-06, "loss": 0.8825, "step": 3407 }, { "epoch": 0.29, "learning_rate": 8.320106389390014e-06, "loss": 0.5942, "step": 3408 }, { "epoch": 0.29, "learning_rate": 8.319073003255372e-06, "loss": 0.9002, "step": 3409 }, { "epoch": 0.29, "learning_rate": 8.31803936359337e-06, "loss": 0.8144, "step": 3410 }, { "epoch": 0.29, "learning_rate": 8.31700547048297e-06, "loss": 0.9132, "step": 3411 }, { "epoch": 0.29, "learning_rate": 8.315971324003142e-06, "loss": 0.9084, "step": 3412 }, { "epoch": 0.29, "learning_rate": 8.314936924232878e-06, "loss": 0.867, "step": 3413 }, { "epoch": 0.29, "learning_rate": 8.313902271251192e-06, "loss": 0.8144, "step": 3414 }, { "epoch": 0.29, "learning_rate": 8.312867365137119e-06, "loss": 0.852, "step": 3415 }, { "epoch": 0.29, "learning_rate": 8.311832205969705e-06, "loss": 0.8874, "step": 3416 }, { "epoch": 0.29, "learning_rate": 8.310796793828024e-06, "loss": 0.8709, "step": 3417 }, { "epoch": 0.29, "learning_rate": 8.309761128791164e-06, "loss": 0.8635, "step": 3418 }, { "epoch": 0.29, "learning_rate": 8.308725210938238e-06, "loss": 0.8371, "step": 3419 }, { "epoch": 0.29, "learning_rate": 8.307689040348369e-06, "loss": 0.9073, "step": 3420 }, { "epoch": 0.29, "learning_rate": 8.30665261710071e-06, "loss": 0.899, "step": 3421 }, { "epoch": 0.29, "learning_rate": 8.305615941274424e-06, "loss": 0.8951, "step": 3422 }, { "epoch": 0.29, "learning_rate": 8.3045790129487e-06, "loss": 0.8877, "step": 3423 }, { "epoch": 0.29, "learning_rate": 8.303541832202744e-06, "loss": 0.9143, "step": 3424 }, { "epoch": 0.29, "learning_rate": 8.302504399115778e-06, "loss": 0.8739, "step": 3425 }, { "epoch": 0.29, "learning_rate": 8.30146671376705e-06, "loss": 0.7957, "step": 3426 }, { "epoch": 0.29, "learning_rate": 8.300428776235822e-06, "loss": 0.8997, "step": 3427 }, { "epoch": 0.29, "learning_rate": 8.29939058660138e-06, "loss": 0.8655, "step": 3428 }, { "epoch": 0.29, "learning_rate": 8.298352144943018e-06, "loss": 0.8234, "step": 3429 }, { "epoch": 0.29, "learning_rate": 8.297313451340064e-06, "loss": 0.8486, "step": 3430 }, { "epoch": 0.29, "learning_rate": 8.296274505871858e-06, "loss": 0.8524, "step": 3431 }, { "epoch": 0.29, "learning_rate": 8.295235308617758e-06, "loss": 0.9189, "step": 3432 }, { "epoch": 0.29, "learning_rate": 8.294195859657144e-06, "loss": 0.8711, "step": 3433 }, { "epoch": 0.29, "learning_rate": 8.293156159069415e-06, "loss": 0.8886, "step": 3434 }, { "epoch": 0.29, "learning_rate": 8.292116206933988e-06, "loss": 0.8412, "step": 3435 }, { "epoch": 0.29, "learning_rate": 8.291076003330298e-06, "loss": 0.8587, "step": 3436 }, { "epoch": 0.29, "learning_rate": 8.290035548337803e-06, "loss": 0.8765, "step": 3437 }, { "epoch": 0.29, "learning_rate": 8.288994842035978e-06, "loss": 0.8797, "step": 3438 }, { "epoch": 0.29, "learning_rate": 8.287953884504317e-06, "loss": 0.9208, "step": 3439 }, { "epoch": 0.29, "learning_rate": 8.286912675822331e-06, "loss": 0.8665, "step": 3440 }, { "epoch": 0.29, "learning_rate": 8.285871216069559e-06, "loss": 0.915, "step": 3441 }, { "epoch": 0.29, "learning_rate": 8.284829505325547e-06, "loss": 0.912, "step": 3442 }, { "epoch": 0.29, "learning_rate": 8.283787543669867e-06, "loss": 0.8961, "step": 3443 }, { "epoch": 0.29, "learning_rate": 8.282745331182111e-06, "loss": 0.8791, "step": 3444 }, { "epoch": 0.29, "learning_rate": 8.28170286794189e-06, "loss": 0.9142, "step": 3445 }, { "epoch": 0.29, "learning_rate": 8.280660154028827e-06, "loss": 0.8627, "step": 3446 }, { "epoch": 0.29, "learning_rate": 8.279617189522576e-06, "loss": 0.8578, "step": 3447 }, { "epoch": 0.29, "learning_rate": 8.278573974502798e-06, "loss": 0.8986, "step": 3448 }, { "epoch": 0.29, "learning_rate": 8.277530509049183e-06, "loss": 0.8526, "step": 3449 }, { "epoch": 0.29, "learning_rate": 8.276486793241433e-06, "loss": 0.8633, "step": 3450 }, { "epoch": 0.29, "learning_rate": 8.275442827159278e-06, "loss": 0.8521, "step": 3451 }, { "epoch": 0.29, "learning_rate": 8.274398610882455e-06, "loss": 0.8669, "step": 3452 }, { "epoch": 0.29, "learning_rate": 8.27335414449073e-06, "loss": 0.8736, "step": 3453 }, { "epoch": 0.29, "learning_rate": 8.272309428063882e-06, "loss": 0.8874, "step": 3454 }, { "epoch": 0.29, "learning_rate": 8.271264461681715e-06, "loss": 0.8596, "step": 3455 }, { "epoch": 0.29, "learning_rate": 8.270219245424044e-06, "loss": 0.9048, "step": 3456 }, { "epoch": 0.29, "learning_rate": 8.269173779370712e-06, "loss": 0.9238, "step": 3457 }, { "epoch": 0.3, "learning_rate": 8.268128063601574e-06, "loss": 0.841, "step": 3458 }, { "epoch": 0.3, "learning_rate": 8.267082098196511e-06, "loss": 0.5638, "step": 3459 }, { "epoch": 0.3, "learning_rate": 8.266035883235416e-06, "loss": 0.865, "step": 3460 }, { "epoch": 0.3, "learning_rate": 8.264989418798205e-06, "loss": 0.8732, "step": 3461 }, { "epoch": 0.3, "learning_rate": 8.263942704964811e-06, "loss": 0.921, "step": 3462 }, { "epoch": 0.3, "learning_rate": 8.262895741815189e-06, "loss": 0.8223, "step": 3463 }, { "epoch": 0.3, "learning_rate": 8.26184852942931e-06, "loss": 0.5749, "step": 3464 }, { "epoch": 0.3, "learning_rate": 8.260801067887165e-06, "loss": 0.8837, "step": 3465 }, { "epoch": 0.3, "learning_rate": 8.259753357268765e-06, "loss": 0.9149, "step": 3466 }, { "epoch": 0.3, "learning_rate": 8.25870539765414e-06, "loss": 0.8051, "step": 3467 }, { "epoch": 0.3, "learning_rate": 8.257657189123337e-06, "loss": 0.8965, "step": 3468 }, { "epoch": 0.3, "learning_rate": 8.256608731756425e-06, "loss": 0.8727, "step": 3469 }, { "epoch": 0.3, "learning_rate": 8.255560025633489e-06, "loss": 0.8932, "step": 3470 }, { "epoch": 0.3, "learning_rate": 8.254511070834637e-06, "loss": 0.8051, "step": 3471 }, { "epoch": 0.3, "learning_rate": 8.253461867439988e-06, "loss": 0.882, "step": 3472 }, { "epoch": 0.3, "learning_rate": 8.252412415529692e-06, "loss": 0.8977, "step": 3473 }, { "epoch": 0.3, "learning_rate": 8.251362715183907e-06, "loss": 0.8309, "step": 3474 }, { "epoch": 0.3, "learning_rate": 8.250312766482815e-06, "loss": 0.8872, "step": 3475 }, { "epoch": 0.3, "learning_rate": 8.249262569506617e-06, "loss": 0.8339, "step": 3476 }, { "epoch": 0.3, "learning_rate": 8.248212124335532e-06, "loss": 0.8516, "step": 3477 }, { "epoch": 0.3, "learning_rate": 8.2471614310498e-06, "loss": 0.9181, "step": 3478 }, { "epoch": 0.3, "learning_rate": 8.246110489729672e-06, "loss": 0.923, "step": 3479 }, { "epoch": 0.3, "learning_rate": 8.245059300455434e-06, "loss": 0.9277, "step": 3480 }, { "epoch": 0.3, "learning_rate": 8.244007863307373e-06, "loss": 0.9034, "step": 3481 }, { "epoch": 0.3, "learning_rate": 8.242956178365806e-06, "loss": 0.5977, "step": 3482 }, { "epoch": 0.3, "learning_rate": 8.241904245711065e-06, "loss": 0.8835, "step": 3483 }, { "epoch": 0.3, "learning_rate": 8.240852065423507e-06, "loss": 0.8445, "step": 3484 }, { "epoch": 0.3, "learning_rate": 8.239799637583493e-06, "loss": 0.896, "step": 3485 }, { "epoch": 0.3, "learning_rate": 8.23874696227142e-06, "loss": 0.8954, "step": 3486 }, { "epoch": 0.3, "learning_rate": 8.237694039567693e-06, "loss": 0.8687, "step": 3487 }, { "epoch": 0.3, "learning_rate": 8.236640869552744e-06, "loss": 0.821, "step": 3488 }, { "epoch": 0.3, "learning_rate": 8.235587452307016e-06, "loss": 0.9109, "step": 3489 }, { "epoch": 0.3, "learning_rate": 8.234533787910973e-06, "loss": 0.8756, "step": 3490 }, { "epoch": 0.3, "learning_rate": 8.233479876445103e-06, "loss": 0.8306, "step": 3491 }, { "epoch": 0.3, "learning_rate": 8.232425717989906e-06, "loss": 0.816, "step": 3492 }, { "epoch": 0.3, "learning_rate": 8.231371312625905e-06, "loss": 0.8561, "step": 3493 }, { "epoch": 0.3, "learning_rate": 8.23031666043364e-06, "loss": 0.8387, "step": 3494 }, { "epoch": 0.3, "learning_rate": 8.229261761493673e-06, "loss": 0.8548, "step": 3495 }, { "epoch": 0.3, "learning_rate": 8.228206615886579e-06, "loss": 0.9062, "step": 3496 }, { "epoch": 0.3, "learning_rate": 8.227151223692956e-06, "loss": 0.8574, "step": 3497 }, { "epoch": 0.3, "learning_rate": 8.226095584993424e-06, "loss": 0.8984, "step": 3498 }, { "epoch": 0.3, "learning_rate": 8.225039699868611e-06, "loss": 0.9112, "step": 3499 }, { "epoch": 0.3, "learning_rate": 8.223983568399177e-06, "loss": 0.8629, "step": 3500 }, { "epoch": 0.3, "learning_rate": 8.222927190665793e-06, "loss": 0.8452, "step": 3501 }, { "epoch": 0.3, "learning_rate": 8.221870566749148e-06, "loss": 0.5227, "step": 3502 }, { "epoch": 0.3, "learning_rate": 8.220813696729954e-06, "loss": 0.8783, "step": 3503 }, { "epoch": 0.3, "learning_rate": 8.219756580688943e-06, "loss": 0.8913, "step": 3504 }, { "epoch": 0.3, "learning_rate": 8.218699218706857e-06, "loss": 0.8766, "step": 3505 }, { "epoch": 0.3, "learning_rate": 8.217641610864466e-06, "loss": 0.8189, "step": 3506 }, { "epoch": 0.3, "learning_rate": 8.216583757242553e-06, "loss": 0.8868, "step": 3507 }, { "epoch": 0.3, "learning_rate": 8.215525657921926e-06, "loss": 0.8727, "step": 3508 }, { "epoch": 0.3, "learning_rate": 8.214467312983406e-06, "loss": 0.8677, "step": 3509 }, { "epoch": 0.3, "learning_rate": 8.213408722507834e-06, "loss": 0.8865, "step": 3510 }, { "epoch": 0.3, "learning_rate": 8.21234988657607e-06, "loss": 0.8611, "step": 3511 }, { "epoch": 0.3, "learning_rate": 8.211290805268995e-06, "loss": 0.9242, "step": 3512 }, { "epoch": 0.3, "learning_rate": 8.210231478667505e-06, "loss": 0.8588, "step": 3513 }, { "epoch": 0.3, "learning_rate": 8.209171906852518e-06, "loss": 0.5719, "step": 3514 }, { "epoch": 0.3, "learning_rate": 8.208112089904968e-06, "loss": 0.896, "step": 3515 }, { "epoch": 0.3, "learning_rate": 8.20705202790581e-06, "loss": 0.8602, "step": 3516 }, { "epoch": 0.3, "learning_rate": 8.205991720936017e-06, "loss": 0.865, "step": 3517 }, { "epoch": 0.3, "learning_rate": 8.204931169076582e-06, "loss": 0.8802, "step": 3518 }, { "epoch": 0.3, "learning_rate": 8.203870372408514e-06, "loss": 0.8546, "step": 3519 }, { "epoch": 0.3, "learning_rate": 8.202809331012839e-06, "loss": 0.897, "step": 3520 }, { "epoch": 0.3, "learning_rate": 8.20174804497061e-06, "loss": 0.8461, "step": 3521 }, { "epoch": 0.3, "learning_rate": 8.20068651436289e-06, "loss": 0.8168, "step": 3522 }, { "epoch": 0.3, "learning_rate": 8.199624739270764e-06, "loss": 0.8702, "step": 3523 }, { "epoch": 0.3, "learning_rate": 8.198562719775335e-06, "loss": 0.8722, "step": 3524 }, { "epoch": 0.3, "learning_rate": 8.197500455957728e-06, "loss": 0.8723, "step": 3525 }, { "epoch": 0.3, "learning_rate": 8.19643794789908e-06, "loss": 0.8967, "step": 3526 }, { "epoch": 0.3, "learning_rate": 8.195375195680556e-06, "loss": 0.8685, "step": 3527 }, { "epoch": 0.3, "learning_rate": 8.194312199383331e-06, "loss": 0.9027, "step": 3528 }, { "epoch": 0.3, "learning_rate": 8.193248959088604e-06, "loss": 0.8651, "step": 3529 }, { "epoch": 0.3, "learning_rate": 8.192185474877587e-06, "loss": 0.8591, "step": 3530 }, { "epoch": 0.3, "learning_rate": 8.191121746831519e-06, "loss": 0.8863, "step": 3531 }, { "epoch": 0.3, "learning_rate": 8.190057775031647e-06, "loss": 0.9013, "step": 3532 }, { "epoch": 0.3, "learning_rate": 8.188993559559247e-06, "loss": 0.8794, "step": 3533 }, { "epoch": 0.3, "learning_rate": 8.187929100495608e-06, "loss": 0.8723, "step": 3534 }, { "epoch": 0.3, "learning_rate": 8.18686439792204e-06, "loss": 0.915, "step": 3535 }, { "epoch": 0.3, "learning_rate": 8.185799451919865e-06, "loss": 0.8533, "step": 3536 }, { "epoch": 0.3, "learning_rate": 8.184734262570435e-06, "loss": 0.8489, "step": 3537 }, { "epoch": 0.3, "learning_rate": 8.183668829955111e-06, "loss": 0.871, "step": 3538 }, { "epoch": 0.3, "learning_rate": 8.182603154155279e-06, "loss": 0.8526, "step": 3539 }, { "epoch": 0.3, "learning_rate": 8.181537235252336e-06, "loss": 0.5352, "step": 3540 }, { "epoch": 0.3, "learning_rate": 8.18047107332771e-06, "loss": 0.8673, "step": 3541 }, { "epoch": 0.3, "learning_rate": 8.179404668462831e-06, "loss": 0.8756, "step": 3542 }, { "epoch": 0.3, "learning_rate": 8.178338020739161e-06, "loss": 0.8853, "step": 3543 }, { "epoch": 0.3, "learning_rate": 8.177271130238174e-06, "loss": 0.8559, "step": 3544 }, { "epoch": 0.3, "learning_rate": 8.176203997041367e-06, "loss": 0.8278, "step": 3545 }, { "epoch": 0.3, "learning_rate": 8.175136621230251e-06, "loss": 0.878, "step": 3546 }, { "epoch": 0.3, "learning_rate": 8.174069002886358e-06, "loss": 0.8602, "step": 3547 }, { "epoch": 0.3, "learning_rate": 8.173001142091238e-06, "loss": 0.6796, "step": 3548 }, { "epoch": 0.3, "learning_rate": 8.171933038926459e-06, "loss": 0.9507, "step": 3549 }, { "epoch": 0.3, "learning_rate": 8.170864693473607e-06, "loss": 0.8833, "step": 3550 }, { "epoch": 0.3, "learning_rate": 8.169796105814292e-06, "loss": 0.8757, "step": 3551 }, { "epoch": 0.3, "learning_rate": 8.168727276030132e-06, "loss": 0.908, "step": 3552 }, { "epoch": 0.3, "learning_rate": 8.167658204202774e-06, "loss": 0.5769, "step": 3553 }, { "epoch": 0.3, "learning_rate": 8.16658889041388e-06, "loss": 0.8835, "step": 3554 }, { "epoch": 0.3, "learning_rate": 8.165519334745124e-06, "loss": 0.8518, "step": 3555 }, { "epoch": 0.3, "learning_rate": 8.164449537278208e-06, "loss": 0.8936, "step": 3556 }, { "epoch": 0.3, "learning_rate": 8.163379498094848e-06, "loss": 0.5555, "step": 3557 }, { "epoch": 0.3, "learning_rate": 8.162309217276778e-06, "loss": 0.913, "step": 3558 }, { "epoch": 0.3, "learning_rate": 8.161238694905753e-06, "loss": 0.8802, "step": 3559 }, { "epoch": 0.3, "learning_rate": 8.160167931063544e-06, "loss": 0.8612, "step": 3560 }, { "epoch": 0.3, "learning_rate": 8.159096925831941e-06, "loss": 0.8379, "step": 3561 }, { "epoch": 0.3, "learning_rate": 8.158025679292752e-06, "loss": 0.8656, "step": 3562 }, { "epoch": 0.3, "learning_rate": 8.156954191527804e-06, "loss": 0.8692, "step": 3563 }, { "epoch": 0.3, "learning_rate": 8.155882462618947e-06, "loss": 0.8143, "step": 3564 }, { "epoch": 0.3, "learning_rate": 8.154810492648038e-06, "loss": 0.8467, "step": 3565 }, { "epoch": 0.3, "learning_rate": 8.153738281696966e-06, "loss": 0.837, "step": 3566 }, { "epoch": 0.3, "learning_rate": 8.152665829847626e-06, "loss": 0.8027, "step": 3567 }, { "epoch": 0.3, "learning_rate": 8.151593137181942e-06, "loss": 0.851, "step": 3568 }, { "epoch": 0.3, "learning_rate": 8.150520203781849e-06, "loss": 0.8692, "step": 3569 }, { "epoch": 0.3, "learning_rate": 8.149447029729304e-06, "loss": 0.8743, "step": 3570 }, { "epoch": 0.3, "learning_rate": 8.14837361510628e-06, "loss": 0.8545, "step": 3571 }, { "epoch": 0.3, "learning_rate": 8.147299959994772e-06, "loss": 0.8556, "step": 3572 }, { "epoch": 0.3, "learning_rate": 8.14622606447679e-06, "loss": 0.8505, "step": 3573 }, { "epoch": 0.3, "learning_rate": 8.145151928634362e-06, "loss": 0.8404, "step": 3574 }, { "epoch": 0.31, "learning_rate": 8.144077552549538e-06, "loss": 0.862, "step": 3575 }, { "epoch": 0.31, "learning_rate": 8.143002936304384e-06, "loss": 0.9031, "step": 3576 }, { "epoch": 0.31, "learning_rate": 8.141928079980983e-06, "loss": 0.8745, "step": 3577 }, { "epoch": 0.31, "learning_rate": 8.140852983661441e-06, "loss": 0.8954, "step": 3578 }, { "epoch": 0.31, "learning_rate": 8.139777647427875e-06, "loss": 0.8358, "step": 3579 }, { "epoch": 0.31, "learning_rate": 8.138702071362427e-06, "loss": 0.8702, "step": 3580 }, { "epoch": 0.31, "learning_rate": 8.137626255547258e-06, "loss": 0.8322, "step": 3581 }, { "epoch": 0.31, "learning_rate": 8.136550200064536e-06, "loss": 0.8477, "step": 3582 }, { "epoch": 0.31, "learning_rate": 8.135473904996462e-06, "loss": 0.878, "step": 3583 }, { "epoch": 0.31, "learning_rate": 8.13439737042525e-06, "loss": 0.8745, "step": 3584 }, { "epoch": 0.31, "learning_rate": 8.133320596433124e-06, "loss": 0.9364, "step": 3585 }, { "epoch": 0.31, "learning_rate": 8.132243583102339e-06, "loss": 0.8695, "step": 3586 }, { "epoch": 0.31, "learning_rate": 8.131166330515161e-06, "loss": 0.8336, "step": 3587 }, { "epoch": 0.31, "learning_rate": 8.130088838753875e-06, "loss": 0.8788, "step": 3588 }, { "epoch": 0.31, "learning_rate": 8.129011107900788e-06, "loss": 0.9027, "step": 3589 }, { "epoch": 0.31, "learning_rate": 8.12793313803822e-06, "loss": 0.8667, "step": 3590 }, { "epoch": 0.31, "learning_rate": 8.126854929248514e-06, "loss": 0.8612, "step": 3591 }, { "epoch": 0.31, "learning_rate": 8.125776481614025e-06, "loss": 0.8471, "step": 3592 }, { "epoch": 0.31, "learning_rate": 8.124697795217134e-06, "loss": 0.9015, "step": 3593 }, { "epoch": 0.31, "learning_rate": 8.123618870140235e-06, "loss": 0.9074, "step": 3594 }, { "epoch": 0.31, "learning_rate": 8.122539706465741e-06, "loss": 0.9188, "step": 3595 }, { "epoch": 0.31, "learning_rate": 8.121460304276085e-06, "loss": 0.9003, "step": 3596 }, { "epoch": 0.31, "learning_rate": 8.120380663653717e-06, "loss": 0.8651, "step": 3597 }, { "epoch": 0.31, "learning_rate": 8.119300784681106e-06, "loss": 0.8529, "step": 3598 }, { "epoch": 0.31, "learning_rate": 8.118220667440735e-06, "loss": 0.8328, "step": 3599 }, { "epoch": 0.31, "learning_rate": 8.117140312015113e-06, "loss": 0.8384, "step": 3600 }, { "epoch": 0.31, "learning_rate": 8.116059718486763e-06, "loss": 0.8052, "step": 3601 }, { "epoch": 0.31, "learning_rate": 8.11497888693822e-06, "loss": 0.8646, "step": 3602 }, { "epoch": 0.31, "learning_rate": 8.113897817452052e-06, "loss": 0.8653, "step": 3603 }, { "epoch": 0.31, "learning_rate": 8.11281651011083e-06, "loss": 0.846, "step": 3604 }, { "epoch": 0.31, "learning_rate": 8.111734964997153e-06, "loss": 0.7743, "step": 3605 }, { "epoch": 0.31, "learning_rate": 8.110653182193632e-06, "loss": 0.8196, "step": 3606 }, { "epoch": 0.31, "learning_rate": 8.109571161782905e-06, "loss": 0.9006, "step": 3607 }, { "epoch": 0.31, "learning_rate": 8.108488903847614e-06, "loss": 0.8866, "step": 3608 }, { "epoch": 0.31, "learning_rate": 8.107406408470432e-06, "loss": 0.8693, "step": 3609 }, { "epoch": 0.31, "learning_rate": 8.106323675734044e-06, "loss": 0.8654, "step": 3610 }, { "epoch": 0.31, "learning_rate": 8.105240705721156e-06, "loss": 0.8858, "step": 3611 }, { "epoch": 0.31, "learning_rate": 8.104157498514489e-06, "loss": 0.8666, "step": 3612 }, { "epoch": 0.31, "learning_rate": 8.103074054196784e-06, "loss": 0.8344, "step": 3613 }, { "epoch": 0.31, "learning_rate": 8.1019903728508e-06, "loss": 0.8615, "step": 3614 }, { "epoch": 0.31, "learning_rate": 8.100906454559317e-06, "loss": 0.9027, "step": 3615 }, { "epoch": 0.31, "learning_rate": 8.099822299405125e-06, "loss": 0.8551, "step": 3616 }, { "epoch": 0.31, "learning_rate": 8.09873790747104e-06, "loss": 0.8322, "step": 3617 }, { "epoch": 0.31, "learning_rate": 8.097653278839894e-06, "loss": 0.9002, "step": 3618 }, { "epoch": 0.31, "learning_rate": 8.096568413594533e-06, "loss": 0.9056, "step": 3619 }, { "epoch": 0.31, "learning_rate": 8.095483311817828e-06, "loss": 0.8819, "step": 3620 }, { "epoch": 0.31, "learning_rate": 8.094397973592663e-06, "loss": 0.9024, "step": 3621 }, { "epoch": 0.31, "learning_rate": 8.09331239900194e-06, "loss": 0.8069, "step": 3622 }, { "epoch": 0.31, "learning_rate": 8.092226588128585e-06, "loss": 0.5545, "step": 3623 }, { "epoch": 0.31, "learning_rate": 8.091140541055533e-06, "loss": 0.8257, "step": 3624 }, { "epoch": 0.31, "learning_rate": 8.090054257865743e-06, "loss": 0.6005, "step": 3625 }, { "epoch": 0.31, "learning_rate": 8.088967738642193e-06, "loss": 0.925, "step": 3626 }, { "epoch": 0.31, "learning_rate": 8.087880983467876e-06, "loss": 0.8671, "step": 3627 }, { "epoch": 0.31, "learning_rate": 8.086793992425798e-06, "loss": 0.5948, "step": 3628 }, { "epoch": 0.31, "learning_rate": 8.085706765598997e-06, "loss": 0.5483, "step": 3629 }, { "epoch": 0.31, "learning_rate": 8.084619303070517e-06, "loss": 0.867, "step": 3630 }, { "epoch": 0.31, "learning_rate": 8.083531604923426e-06, "loss": 0.8883, "step": 3631 }, { "epoch": 0.31, "learning_rate": 8.082443671240805e-06, "loss": 0.8445, "step": 3632 }, { "epoch": 0.31, "learning_rate": 8.081355502105756e-06, "loss": 0.8946, "step": 3633 }, { "epoch": 0.31, "learning_rate": 8.0802670976014e-06, "loss": 0.8364, "step": 3634 }, { "epoch": 0.31, "learning_rate": 8.079178457810875e-06, "loss": 0.604, "step": 3635 }, { "epoch": 0.31, "learning_rate": 8.078089582817335e-06, "loss": 0.8761, "step": 3636 }, { "epoch": 0.31, "learning_rate": 8.077000472703957e-06, "loss": 0.9013, "step": 3637 }, { "epoch": 0.31, "learning_rate": 8.075911127553928e-06, "loss": 0.8484, "step": 3638 }, { "epoch": 0.31, "learning_rate": 8.07482154745046e-06, "loss": 0.8093, "step": 3639 }, { "epoch": 0.31, "learning_rate": 8.073731732476784e-06, "loss": 0.5856, "step": 3640 }, { "epoch": 0.31, "learning_rate": 8.07264168271614e-06, "loss": 0.8624, "step": 3641 }, { "epoch": 0.31, "learning_rate": 8.071551398251793e-06, "loss": 0.9245, "step": 3642 }, { "epoch": 0.31, "learning_rate": 8.070460879167027e-06, "loss": 0.8624, "step": 3643 }, { "epoch": 0.31, "learning_rate": 8.069370125545137e-06, "loss": 0.8507, "step": 3644 }, { "epoch": 0.31, "learning_rate": 8.068279137469445e-06, "loss": 0.88, "step": 3645 }, { "epoch": 0.31, "learning_rate": 8.067187915023283e-06, "loss": 0.8704, "step": 3646 }, { "epoch": 0.31, "learning_rate": 8.066096458290003e-06, "loss": 0.8586, "step": 3647 }, { "epoch": 0.31, "learning_rate": 8.065004767352979e-06, "loss": 0.8602, "step": 3648 }, { "epoch": 0.31, "learning_rate": 8.063912842295597e-06, "loss": 0.8443, "step": 3649 }, { "epoch": 0.31, "learning_rate": 8.062820683201268e-06, "loss": 0.8595, "step": 3650 }, { "epoch": 0.31, "learning_rate": 8.06172829015341e-06, "loss": 0.801, "step": 3651 }, { "epoch": 0.31, "learning_rate": 8.060635663235472e-06, "loss": 0.8441, "step": 3652 }, { "epoch": 0.31, "learning_rate": 8.05954280253091e-06, "loss": 0.8755, "step": 3653 }, { "epoch": 0.31, "learning_rate": 8.058449708123204e-06, "loss": 0.8741, "step": 3654 }, { "epoch": 0.31, "learning_rate": 8.057356380095848e-06, "loss": 0.8898, "step": 3655 }, { "epoch": 0.31, "learning_rate": 8.05626281853236e-06, "loss": 0.8348, "step": 3656 }, { "epoch": 0.31, "learning_rate": 8.055169023516268e-06, "loss": 0.8844, "step": 3657 }, { "epoch": 0.31, "learning_rate": 8.05407499513112e-06, "loss": 0.7937, "step": 3658 }, { "epoch": 0.31, "learning_rate": 8.052980733460489e-06, "loss": 0.5819, "step": 3659 }, { "epoch": 0.31, "learning_rate": 8.051886238587956e-06, "loss": 0.8837, "step": 3660 }, { "epoch": 0.31, "learning_rate": 8.050791510597125e-06, "loss": 0.8493, "step": 3661 }, { "epoch": 0.31, "learning_rate": 8.049696549571618e-06, "loss": 0.8918, "step": 3662 }, { "epoch": 0.31, "learning_rate": 8.04860135559507e-06, "loss": 0.6627, "step": 3663 }, { "epoch": 0.31, "learning_rate": 8.047505928751141e-06, "loss": 0.8766, "step": 3664 }, { "epoch": 0.31, "learning_rate": 8.046410269123503e-06, "loss": 0.8081, "step": 3665 }, { "epoch": 0.31, "learning_rate": 8.045314376795851e-06, "loss": 0.834, "step": 3666 }, { "epoch": 0.31, "learning_rate": 8.044218251851892e-06, "loss": 0.8711, "step": 3667 }, { "epoch": 0.31, "learning_rate": 8.043121894375353e-06, "loss": 0.8473, "step": 3668 }, { "epoch": 0.31, "learning_rate": 8.042025304449983e-06, "loss": 0.9037, "step": 3669 }, { "epoch": 0.31, "learning_rate": 8.04092848215954e-06, "loss": 0.8608, "step": 3670 }, { "epoch": 0.31, "learning_rate": 8.039831427587807e-06, "loss": 0.9072, "step": 3671 }, { "epoch": 0.31, "learning_rate": 8.038734140818583e-06, "loss": 0.8824, "step": 3672 }, { "epoch": 0.31, "learning_rate": 8.037636621935686e-06, "loss": 0.8448, "step": 3673 }, { "epoch": 0.31, "learning_rate": 8.036538871022945e-06, "loss": 0.8732, "step": 3674 }, { "epoch": 0.31, "learning_rate": 8.035440888164217e-06, "loss": 0.8771, "step": 3675 }, { "epoch": 0.31, "learning_rate": 8.034342673443365e-06, "loss": 0.8732, "step": 3676 }, { "epoch": 0.31, "learning_rate": 8.033244226944284e-06, "loss": 0.8835, "step": 3677 }, { "epoch": 0.31, "learning_rate": 8.032145548750873e-06, "loss": 0.8913, "step": 3678 }, { "epoch": 0.31, "learning_rate": 8.031046638947058e-06, "loss": 0.8886, "step": 3679 }, { "epoch": 0.31, "learning_rate": 8.029947497616777e-06, "loss": 0.8457, "step": 3680 }, { "epoch": 0.31, "learning_rate": 8.02884812484399e-06, "loss": 0.9004, "step": 3681 }, { "epoch": 0.31, "learning_rate": 8.027748520712666e-06, "loss": 0.8576, "step": 3682 }, { "epoch": 0.31, "learning_rate": 8.026648685306809e-06, "loss": 0.8932, "step": 3683 }, { "epoch": 0.31, "learning_rate": 8.025548618710421e-06, "loss": 0.9186, "step": 3684 }, { "epoch": 0.31, "learning_rate": 8.024448321007535e-06, "loss": 0.8772, "step": 3685 }, { "epoch": 0.31, "learning_rate": 8.023347792282196e-06, "loss": 0.8457, "step": 3686 }, { "epoch": 0.31, "learning_rate": 8.022247032618469e-06, "loss": 0.8513, "step": 3687 }, { "epoch": 0.31, "learning_rate": 8.021146042100432e-06, "loss": 0.916, "step": 3688 }, { "epoch": 0.31, "learning_rate": 8.020044820812188e-06, "loss": 0.8938, "step": 3689 }, { "epoch": 0.31, "learning_rate": 8.018943368837854e-06, "loss": 0.8981, "step": 3690 }, { "epoch": 0.31, "learning_rate": 8.01784168626156e-06, "loss": 0.8716, "step": 3691 }, { "epoch": 0.32, "learning_rate": 8.016739773167462e-06, "loss": 0.8291, "step": 3692 }, { "epoch": 0.32, "learning_rate": 8.01563762963973e-06, "loss": 0.5673, "step": 3693 }, { "epoch": 0.32, "learning_rate": 8.014535255762547e-06, "loss": 0.8819, "step": 3694 }, { "epoch": 0.32, "learning_rate": 8.013432651620121e-06, "loss": 0.8249, "step": 3695 }, { "epoch": 0.32, "learning_rate": 8.012329817296675e-06, "loss": 0.865, "step": 3696 }, { "epoch": 0.32, "learning_rate": 8.011226752876448e-06, "loss": 0.8843, "step": 3697 }, { "epoch": 0.32, "learning_rate": 8.010123458443697e-06, "loss": 0.8581, "step": 3698 }, { "epoch": 0.32, "learning_rate": 8.009019934082699e-06, "loss": 0.8626, "step": 3699 }, { "epoch": 0.32, "learning_rate": 8.007916179877742e-06, "loss": 0.8714, "step": 3700 }, { "epoch": 0.32, "learning_rate": 8.006812195913141e-06, "loss": 0.8921, "step": 3701 }, { "epoch": 0.32, "learning_rate": 8.005707982273224e-06, "loss": 0.8479, "step": 3702 }, { "epoch": 0.32, "learning_rate": 8.004603539042333e-06, "loss": 0.9196, "step": 3703 }, { "epoch": 0.32, "learning_rate": 8.003498866304832e-06, "loss": 0.842, "step": 3704 }, { "epoch": 0.32, "learning_rate": 8.002393964145104e-06, "loss": 0.8847, "step": 3705 }, { "epoch": 0.32, "learning_rate": 8.001288832647542e-06, "loss": 0.8956, "step": 3706 }, { "epoch": 0.32, "learning_rate": 8.000183471896567e-06, "loss": 0.8405, "step": 3707 }, { "epoch": 0.32, "learning_rate": 7.999077881976609e-06, "loss": 0.8097, "step": 3708 }, { "epoch": 0.32, "learning_rate": 7.997972062972118e-06, "loss": 0.8385, "step": 3709 }, { "epoch": 0.32, "learning_rate": 7.996866014967562e-06, "loss": 0.8057, "step": 3710 }, { "epoch": 0.32, "learning_rate": 7.995759738047429e-06, "loss": 0.8686, "step": 3711 }, { "epoch": 0.32, "learning_rate": 7.994653232296217e-06, "loss": 0.8741, "step": 3712 }, { "epoch": 0.32, "learning_rate": 7.993546497798454e-06, "loss": 0.8821, "step": 3713 }, { "epoch": 0.32, "learning_rate": 7.99243953463867e-06, "loss": 0.8515, "step": 3714 }, { "epoch": 0.32, "learning_rate": 7.991332342901426e-06, "loss": 0.8782, "step": 3715 }, { "epoch": 0.32, "learning_rate": 7.990224922671293e-06, "loss": 0.8613, "step": 3716 }, { "epoch": 0.32, "learning_rate": 7.989117274032859e-06, "loss": 0.8833, "step": 3717 }, { "epoch": 0.32, "learning_rate": 7.988009397070735e-06, "loss": 0.8458, "step": 3718 }, { "epoch": 0.32, "learning_rate": 7.986901291869545e-06, "loss": 0.916, "step": 3719 }, { "epoch": 0.32, "learning_rate": 7.985792958513932e-06, "loss": 0.893, "step": 3720 }, { "epoch": 0.32, "learning_rate": 7.984684397088554e-06, "loss": 0.8279, "step": 3721 }, { "epoch": 0.32, "learning_rate": 7.983575607678092e-06, "loss": 0.8602, "step": 3722 }, { "epoch": 0.32, "learning_rate": 7.982466590367237e-06, "loss": 0.8417, "step": 3723 }, { "epoch": 0.32, "learning_rate": 7.981357345240704e-06, "loss": 0.9208, "step": 3724 }, { "epoch": 0.32, "learning_rate": 7.980247872383223e-06, "loss": 0.8971, "step": 3725 }, { "epoch": 0.32, "learning_rate": 7.979138171879536e-06, "loss": 0.8907, "step": 3726 }, { "epoch": 0.32, "learning_rate": 7.978028243814416e-06, "loss": 0.8654, "step": 3727 }, { "epoch": 0.32, "learning_rate": 7.976918088272636e-06, "loss": 0.8578, "step": 3728 }, { "epoch": 0.32, "learning_rate": 7.975807705339002e-06, "loss": 0.8808, "step": 3729 }, { "epoch": 0.32, "learning_rate": 7.974697095098329e-06, "loss": 0.838, "step": 3730 }, { "epoch": 0.32, "learning_rate": 7.973586257635448e-06, "loss": 0.8823, "step": 3731 }, { "epoch": 0.32, "learning_rate": 7.972475193035212e-06, "loss": 0.8736, "step": 3732 }, { "epoch": 0.32, "learning_rate": 7.971363901382491e-06, "loss": 0.9152, "step": 3733 }, { "epoch": 0.32, "learning_rate": 7.97025238276217e-06, "loss": 0.8763, "step": 3734 }, { "epoch": 0.32, "learning_rate": 7.969140637259152e-06, "loss": 0.8515, "step": 3735 }, { "epoch": 0.32, "learning_rate": 7.968028664958359e-06, "loss": 0.9186, "step": 3736 }, { "epoch": 0.32, "learning_rate": 7.966916465944725e-06, "loss": 0.8504, "step": 3737 }, { "epoch": 0.32, "learning_rate": 7.965804040303212e-06, "loss": 0.9174, "step": 3738 }, { "epoch": 0.32, "learning_rate": 7.964691388118784e-06, "loss": 0.8548, "step": 3739 }, { "epoch": 0.32, "learning_rate": 7.96357850947644e-06, "loss": 0.8503, "step": 3740 }, { "epoch": 0.32, "learning_rate": 7.962465404461181e-06, "loss": 0.8671, "step": 3741 }, { "epoch": 0.32, "learning_rate": 7.961352073158034e-06, "loss": 0.8608, "step": 3742 }, { "epoch": 0.32, "learning_rate": 7.960238515652041e-06, "loss": 0.8591, "step": 3743 }, { "epoch": 0.32, "learning_rate": 7.959124732028259e-06, "loss": 0.8988, "step": 3744 }, { "epoch": 0.32, "learning_rate": 7.958010722371767e-06, "loss": 0.8916, "step": 3745 }, { "epoch": 0.32, "learning_rate": 7.956896486767659e-06, "loss": 0.7881, "step": 3746 }, { "epoch": 0.32, "learning_rate": 7.95578202530104e-06, "loss": 0.9016, "step": 3747 }, { "epoch": 0.32, "learning_rate": 7.954667338057044e-06, "loss": 0.895, "step": 3748 }, { "epoch": 0.32, "learning_rate": 7.953552425120815e-06, "loss": 0.8725, "step": 3749 }, { "epoch": 0.32, "learning_rate": 7.952437286577517e-06, "loss": 0.8359, "step": 3750 }, { "epoch": 0.32, "learning_rate": 7.951321922512326e-06, "loss": 0.8471, "step": 3751 }, { "epoch": 0.32, "learning_rate": 7.950206333010441e-06, "loss": 0.8629, "step": 3752 }, { "epoch": 0.32, "learning_rate": 7.949090518157078e-06, "loss": 0.5131, "step": 3753 }, { "epoch": 0.32, "learning_rate": 7.947974478037468e-06, "loss": 0.8976, "step": 3754 }, { "epoch": 0.32, "learning_rate": 7.946858212736857e-06, "loss": 0.9025, "step": 3755 }, { "epoch": 0.32, "learning_rate": 7.945741722340514e-06, "loss": 0.8862, "step": 3756 }, { "epoch": 0.32, "learning_rate": 7.944625006933719e-06, "loss": 0.9184, "step": 3757 }, { "epoch": 0.32, "learning_rate": 7.943508066601775e-06, "loss": 0.8657, "step": 3758 }, { "epoch": 0.32, "learning_rate": 7.942390901429997e-06, "loss": 0.8487, "step": 3759 }, { "epoch": 0.32, "learning_rate": 7.941273511503724e-06, "loss": 0.8357, "step": 3760 }, { "epoch": 0.32, "learning_rate": 7.940155896908303e-06, "loss": 0.8726, "step": 3761 }, { "epoch": 0.32, "learning_rate": 7.939038057729107e-06, "loss": 0.9067, "step": 3762 }, { "epoch": 0.32, "learning_rate": 7.93791999405152e-06, "loss": 0.847, "step": 3763 }, { "epoch": 0.32, "learning_rate": 7.936801705960944e-06, "loss": 0.8175, "step": 3764 }, { "epoch": 0.32, "learning_rate": 7.9356831935428e-06, "loss": 0.535, "step": 3765 }, { "epoch": 0.32, "learning_rate": 7.934564456882528e-06, "loss": 0.8559, "step": 3766 }, { "epoch": 0.32, "learning_rate": 7.933445496065582e-06, "loss": 0.8755, "step": 3767 }, { "epoch": 0.32, "learning_rate": 7.93232631117743e-06, "loss": 0.8983, "step": 3768 }, { "epoch": 0.32, "learning_rate": 7.931206902303566e-06, "loss": 0.8851, "step": 3769 }, { "epoch": 0.32, "learning_rate": 7.930087269529494e-06, "loss": 0.8012, "step": 3770 }, { "epoch": 0.32, "learning_rate": 7.928967412940735e-06, "loss": 0.8468, "step": 3771 }, { "epoch": 0.32, "learning_rate": 7.927847332622833e-06, "loss": 0.819, "step": 3772 }, { "epoch": 0.32, "learning_rate": 7.926727028661342e-06, "loss": 0.897, "step": 3773 }, { "epoch": 0.32, "learning_rate": 7.92560650114184e-06, "loss": 0.8671, "step": 3774 }, { "epoch": 0.32, "learning_rate": 7.924485750149916e-06, "loss": 0.9083, "step": 3775 }, { "epoch": 0.32, "learning_rate": 7.923364775771179e-06, "loss": 0.8745, "step": 3776 }, { "epoch": 0.32, "learning_rate": 7.922243578091253e-06, "loss": 0.8891, "step": 3777 }, { "epoch": 0.32, "learning_rate": 7.921122157195784e-06, "loss": 0.8523, "step": 3778 }, { "epoch": 0.32, "learning_rate": 7.920000513170429e-06, "loss": 0.894, "step": 3779 }, { "epoch": 0.32, "learning_rate": 7.918878646100867e-06, "loss": 0.6275, "step": 3780 }, { "epoch": 0.32, "learning_rate": 7.917756556072792e-06, "loss": 0.8456, "step": 3781 }, { "epoch": 0.32, "learning_rate": 7.916634243171908e-06, "loss": 0.9029, "step": 3782 }, { "epoch": 0.32, "learning_rate": 7.915511707483953e-06, "loss": 0.8623, "step": 3783 }, { "epoch": 0.32, "learning_rate": 7.914388949094667e-06, "loss": 0.8745, "step": 3784 }, { "epoch": 0.32, "learning_rate": 7.913265968089809e-06, "loss": 0.8821, "step": 3785 }, { "epoch": 0.32, "learning_rate": 7.912142764555164e-06, "loss": 0.876, "step": 3786 }, { "epoch": 0.32, "learning_rate": 7.911019338576523e-06, "loss": 0.5798, "step": 3787 }, { "epoch": 0.32, "learning_rate": 7.909895690239702e-06, "loss": 0.8677, "step": 3788 }, { "epoch": 0.32, "learning_rate": 7.908771819630529e-06, "loss": 0.8484, "step": 3789 }, { "epoch": 0.32, "learning_rate": 7.907647726834851e-06, "loss": 0.8527, "step": 3790 }, { "epoch": 0.32, "learning_rate": 7.906523411938532e-06, "loss": 0.943, "step": 3791 }, { "epoch": 0.32, "learning_rate": 7.905398875027454e-06, "loss": 0.9043, "step": 3792 }, { "epoch": 0.32, "learning_rate": 7.904274116187513e-06, "loss": 0.8884, "step": 3793 }, { "epoch": 0.32, "learning_rate": 7.903149135504624e-06, "loss": 0.8025, "step": 3794 }, { "epoch": 0.32, "learning_rate": 7.902023933064721e-06, "loss": 0.8695, "step": 3795 }, { "epoch": 0.32, "learning_rate": 7.90089850895375e-06, "loss": 0.8284, "step": 3796 }, { "epoch": 0.32, "learning_rate": 7.899772863257677e-06, "loss": 0.8721, "step": 3797 }, { "epoch": 0.32, "learning_rate": 7.898646996062482e-06, "loss": 0.8117, "step": 3798 }, { "epoch": 0.32, "learning_rate": 7.89752090745417e-06, "loss": 0.9064, "step": 3799 }, { "epoch": 0.32, "learning_rate": 7.896394597518755e-06, "loss": 0.8768, "step": 3800 }, { "epoch": 0.32, "learning_rate": 7.895268066342268e-06, "loss": 0.8615, "step": 3801 }, { "epoch": 0.32, "learning_rate": 7.89414131401076e-06, "loss": 0.882, "step": 3802 }, { "epoch": 0.32, "learning_rate": 7.893014340610299e-06, "loss": 0.8784, "step": 3803 }, { "epoch": 0.32, "learning_rate": 7.891887146226968e-06, "loss": 0.834, "step": 3804 }, { "epoch": 0.32, "learning_rate": 7.890759730946867e-06, "loss": 0.8523, "step": 3805 }, { "epoch": 0.32, "learning_rate": 7.889632094856117e-06, "loss": 0.869, "step": 3806 }, { "epoch": 0.32, "learning_rate": 7.888504238040848e-06, "loss": 0.8746, "step": 3807 }, { "epoch": 0.32, "learning_rate": 7.887376160587214e-06, "loss": 0.8595, "step": 3808 }, { "epoch": 0.33, "learning_rate": 7.886247862581385e-06, "loss": 0.848, "step": 3809 }, { "epoch": 0.33, "learning_rate": 7.885119344109539e-06, "loss": 0.884, "step": 3810 }, { "epoch": 0.33, "learning_rate": 7.883990605257886e-06, "loss": 0.8832, "step": 3811 }, { "epoch": 0.33, "learning_rate": 7.88286164611264e-06, "loss": 0.9033, "step": 3812 }, { "epoch": 0.33, "learning_rate": 7.881732466760039e-06, "loss": 0.8688, "step": 3813 }, { "epoch": 0.33, "learning_rate": 7.880603067286332e-06, "loss": 0.8232, "step": 3814 }, { "epoch": 0.33, "learning_rate": 7.879473447777792e-06, "loss": 0.8137, "step": 3815 }, { "epoch": 0.33, "learning_rate": 7.878343608320702e-06, "loss": 0.8747, "step": 3816 }, { "epoch": 0.33, "learning_rate": 7.877213549001368e-06, "loss": 0.8644, "step": 3817 }, { "epoch": 0.33, "learning_rate": 7.876083269906106e-06, "loss": 0.9051, "step": 3818 }, { "epoch": 0.33, "learning_rate": 7.874952771121253e-06, "loss": 0.8546, "step": 3819 }, { "epoch": 0.33, "learning_rate": 7.873822052733164e-06, "loss": 0.8473, "step": 3820 }, { "epoch": 0.33, "learning_rate": 7.872691114828208e-06, "loss": 0.8687, "step": 3821 }, { "epoch": 0.33, "learning_rate": 7.871559957492773e-06, "loss": 0.8312, "step": 3822 }, { "epoch": 0.33, "learning_rate": 7.87042858081326e-06, "loss": 0.8751, "step": 3823 }, { "epoch": 0.33, "learning_rate": 7.869296984876089e-06, "loss": 0.8512, "step": 3824 }, { "epoch": 0.33, "learning_rate": 7.868165169767701e-06, "loss": 0.91, "step": 3825 }, { "epoch": 0.33, "learning_rate": 7.867033135574545e-06, "loss": 0.9057, "step": 3826 }, { "epoch": 0.33, "learning_rate": 7.865900882383095e-06, "loss": 0.8834, "step": 3827 }, { "epoch": 0.33, "learning_rate": 7.864768410279834e-06, "loss": 0.8552, "step": 3828 }, { "epoch": 0.33, "learning_rate": 7.86363571935127e-06, "loss": 0.8888, "step": 3829 }, { "epoch": 0.33, "learning_rate": 7.862502809683922e-06, "loss": 0.8303, "step": 3830 }, { "epoch": 0.33, "learning_rate": 7.861369681364327e-06, "loss": 0.8012, "step": 3831 }, { "epoch": 0.33, "learning_rate": 7.860236334479038e-06, "loss": 0.8488, "step": 3832 }, { "epoch": 0.33, "learning_rate": 7.85910276911463e-06, "loss": 0.901, "step": 3833 }, { "epoch": 0.33, "learning_rate": 7.857968985357686e-06, "loss": 0.8208, "step": 3834 }, { "epoch": 0.33, "learning_rate": 7.85683498329481e-06, "loss": 0.8784, "step": 3835 }, { "epoch": 0.33, "learning_rate": 7.855700763012625e-06, "loss": 0.8935, "step": 3836 }, { "epoch": 0.33, "learning_rate": 7.854566324597768e-06, "loss": 0.9203, "step": 3837 }, { "epoch": 0.33, "learning_rate": 7.853431668136892e-06, "loss": 0.826, "step": 3838 }, { "epoch": 0.33, "learning_rate": 7.85229679371667e-06, "loss": 0.8718, "step": 3839 }, { "epoch": 0.33, "learning_rate": 7.851161701423784e-06, "loss": 0.9069, "step": 3840 }, { "epoch": 0.33, "learning_rate": 7.850026391344945e-06, "loss": 0.8917, "step": 3841 }, { "epoch": 0.33, "learning_rate": 7.848890863566867e-06, "loss": 0.8837, "step": 3842 }, { "epoch": 0.33, "learning_rate": 7.847755118176295e-06, "loss": 0.923, "step": 3843 }, { "epoch": 0.33, "learning_rate": 7.846619155259976e-06, "loss": 0.6188, "step": 3844 }, { "epoch": 0.33, "learning_rate": 7.845482974904682e-06, "loss": 0.8939, "step": 3845 }, { "epoch": 0.33, "learning_rate": 7.844346577197203e-06, "loss": 0.8373, "step": 3846 }, { "epoch": 0.33, "learning_rate": 7.84320996222434e-06, "loss": 0.8603, "step": 3847 }, { "epoch": 0.33, "learning_rate": 7.842073130072916e-06, "loss": 0.8612, "step": 3848 }, { "epoch": 0.33, "learning_rate": 7.840936080829764e-06, "loss": 0.8702, "step": 3849 }, { "epoch": 0.33, "learning_rate": 7.83979881458174e-06, "loss": 0.8506, "step": 3850 }, { "epoch": 0.33, "learning_rate": 7.838661331415713e-06, "loss": 0.8334, "step": 3851 }, { "epoch": 0.33, "learning_rate": 7.837523631418573e-06, "loss": 0.876, "step": 3852 }, { "epoch": 0.33, "learning_rate": 7.836385714677218e-06, "loss": 0.8568, "step": 3853 }, { "epoch": 0.33, "learning_rate": 7.835247581278572e-06, "loss": 0.811, "step": 3854 }, { "epoch": 0.33, "learning_rate": 7.834109231309568e-06, "loss": 0.8091, "step": 3855 }, { "epoch": 0.33, "learning_rate": 7.83297066485716e-06, "loss": 0.857, "step": 3856 }, { "epoch": 0.33, "learning_rate": 7.83183188200832e-06, "loss": 0.902, "step": 3857 }, { "epoch": 0.33, "learning_rate": 7.830692882850031e-06, "loss": 0.8795, "step": 3858 }, { "epoch": 0.33, "learning_rate": 7.829553667469296e-06, "loss": 0.8829, "step": 3859 }, { "epoch": 0.33, "learning_rate": 7.828414235953133e-06, "loss": 0.9535, "step": 3860 }, { "epoch": 0.33, "learning_rate": 7.82727458838858e-06, "loss": 0.8931, "step": 3861 }, { "epoch": 0.33, "learning_rate": 7.826134724862687e-06, "loss": 0.8116, "step": 3862 }, { "epoch": 0.33, "learning_rate": 7.824994645462524e-06, "loss": 0.8654, "step": 3863 }, { "epoch": 0.33, "learning_rate": 7.823854350275173e-06, "loss": 0.8967, "step": 3864 }, { "epoch": 0.33, "learning_rate": 7.822713839387739e-06, "loss": 0.8017, "step": 3865 }, { "epoch": 0.33, "learning_rate": 7.821573112887338e-06, "loss": 0.856, "step": 3866 }, { "epoch": 0.33, "learning_rate": 7.820432170861106e-06, "loss": 0.859, "step": 3867 }, { "epoch": 0.33, "learning_rate": 7.819291013396189e-06, "loss": 0.8983, "step": 3868 }, { "epoch": 0.33, "learning_rate": 7.81814964057976e-06, "loss": 0.8512, "step": 3869 }, { "epoch": 0.33, "learning_rate": 7.817008052499002e-06, "loss": 0.8721, "step": 3870 }, { "epoch": 0.33, "learning_rate": 7.815866249241111e-06, "loss": 0.8651, "step": 3871 }, { "epoch": 0.33, "learning_rate": 7.81472423089331e-06, "loss": 0.9014, "step": 3872 }, { "epoch": 0.33, "learning_rate": 7.813581997542825e-06, "loss": 0.8749, "step": 3873 }, { "epoch": 0.33, "learning_rate": 7.81243954927691e-06, "loss": 0.865, "step": 3874 }, { "epoch": 0.33, "learning_rate": 7.81129688618283e-06, "loss": 0.9137, "step": 3875 }, { "epoch": 0.33, "learning_rate": 7.810154008347868e-06, "loss": 0.8946, "step": 3876 }, { "epoch": 0.33, "learning_rate": 7.809010915859321e-06, "loss": 0.8855, "step": 3877 }, { "epoch": 0.33, "learning_rate": 7.807867608804504e-06, "loss": 0.8687, "step": 3878 }, { "epoch": 0.33, "learning_rate": 7.806724087270752e-06, "loss": 0.8889, "step": 3879 }, { "epoch": 0.33, "learning_rate": 7.80558035134541e-06, "loss": 0.8419, "step": 3880 }, { "epoch": 0.33, "learning_rate": 7.80443640111584e-06, "loss": 0.9092, "step": 3881 }, { "epoch": 0.33, "learning_rate": 7.803292236669428e-06, "loss": 0.8803, "step": 3882 }, { "epoch": 0.33, "learning_rate": 7.802147858093567e-06, "loss": 0.9126, "step": 3883 }, { "epoch": 0.33, "learning_rate": 7.801003265475672e-06, "loss": 0.8747, "step": 3884 }, { "epoch": 0.33, "learning_rate": 7.799858458903172e-06, "loss": 0.8612, "step": 3885 }, { "epoch": 0.33, "learning_rate": 7.798713438463514e-06, "loss": 0.8149, "step": 3886 }, { "epoch": 0.33, "learning_rate": 7.797568204244159e-06, "loss": 0.8598, "step": 3887 }, { "epoch": 0.33, "learning_rate": 7.796422756332587e-06, "loss": 0.8041, "step": 3888 }, { "epoch": 0.33, "learning_rate": 7.795277094816292e-06, "loss": 0.8384, "step": 3889 }, { "epoch": 0.33, "learning_rate": 7.794131219782785e-06, "loss": 0.8664, "step": 3890 }, { "epoch": 0.33, "learning_rate": 7.792985131319594e-06, "loss": 0.8702, "step": 3891 }, { "epoch": 0.33, "learning_rate": 7.791838829514267e-06, "loss": 0.8808, "step": 3892 }, { "epoch": 0.33, "learning_rate": 7.790692314454357e-06, "loss": 0.8772, "step": 3893 }, { "epoch": 0.33, "learning_rate": 7.789545586227444e-06, "loss": 0.877, "step": 3894 }, { "epoch": 0.33, "learning_rate": 7.788398644921124e-06, "loss": 0.8999, "step": 3895 }, { "epoch": 0.33, "learning_rate": 7.787251490623003e-06, "loss": 0.9084, "step": 3896 }, { "epoch": 0.33, "learning_rate": 7.786104123420704e-06, "loss": 0.8599, "step": 3897 }, { "epoch": 0.33, "learning_rate": 7.784956543401872e-06, "loss": 0.847, "step": 3898 }, { "epoch": 0.33, "learning_rate": 7.783808750654164e-06, "loss": 0.8406, "step": 3899 }, { "epoch": 0.33, "learning_rate": 7.782660745265254e-06, "loss": 0.8746, "step": 3900 }, { "epoch": 0.33, "learning_rate": 7.781512527322832e-06, "loss": 0.8304, "step": 3901 }, { "epoch": 0.33, "learning_rate": 7.780364096914607e-06, "loss": 0.831, "step": 3902 }, { "epoch": 0.33, "learning_rate": 7.779215454128298e-06, "loss": 0.8366, "step": 3903 }, { "epoch": 0.33, "learning_rate": 7.778066599051646e-06, "loss": 0.9388, "step": 3904 }, { "epoch": 0.33, "learning_rate": 7.776917531772407e-06, "loss": 0.8807, "step": 3905 }, { "epoch": 0.33, "learning_rate": 7.775768252378352e-06, "loss": 0.8483, "step": 3906 }, { "epoch": 0.33, "learning_rate": 7.774618760957269e-06, "loss": 0.8303, "step": 3907 }, { "epoch": 0.33, "learning_rate": 7.773469057596961e-06, "loss": 0.8945, "step": 3908 }, { "epoch": 0.33, "learning_rate": 7.772319142385248e-06, "loss": 0.9013, "step": 3909 }, { "epoch": 0.33, "learning_rate": 7.771169015409967e-06, "loss": 0.8776, "step": 3910 }, { "epoch": 0.33, "learning_rate": 7.77001867675897e-06, "loss": 0.8726, "step": 3911 }, { "epoch": 0.33, "learning_rate": 7.768868126520126e-06, "loss": 0.8748, "step": 3912 }, { "epoch": 0.33, "learning_rate": 7.767717364781319e-06, "loss": 0.8977, "step": 3913 }, { "epoch": 0.33, "learning_rate": 7.76656639163045e-06, "loss": 0.8815, "step": 3914 }, { "epoch": 0.33, "learning_rate": 7.765415207155437e-06, "loss": 0.829, "step": 3915 }, { "epoch": 0.33, "learning_rate": 7.764263811444214e-06, "loss": 0.8654, "step": 3916 }, { "epoch": 0.33, "learning_rate": 7.763112204584728e-06, "loss": 0.8629, "step": 3917 }, { "epoch": 0.33, "learning_rate": 7.761960386664947e-06, "loss": 0.9136, "step": 3918 }, { "epoch": 0.33, "learning_rate": 7.760808357772848e-06, "loss": 0.8834, "step": 3919 }, { "epoch": 0.33, "learning_rate": 7.759656117996436e-06, "loss": 0.8753, "step": 3920 }, { "epoch": 0.33, "learning_rate": 7.758503667423717e-06, "loss": 0.8311, "step": 3921 }, { "epoch": 0.33, "learning_rate": 7.757351006142728e-06, "loss": 0.8547, "step": 3922 }, { "epoch": 0.33, "learning_rate": 7.756198134241508e-06, "loss": 0.9481, "step": 3923 }, { "epoch": 0.33, "learning_rate": 7.755045051808126e-06, "loss": 0.89, "step": 3924 }, { "epoch": 0.33, "learning_rate": 7.753891758930654e-06, "loss": 0.8417, "step": 3925 }, { "epoch": 0.34, "learning_rate": 7.752738255697192e-06, "loss": 0.8844, "step": 3926 }, { "epoch": 0.34, "learning_rate": 7.751584542195844e-06, "loss": 0.6171, "step": 3927 }, { "epoch": 0.34, "learning_rate": 7.750430618514744e-06, "loss": 0.8364, "step": 3928 }, { "epoch": 0.34, "learning_rate": 7.749276484742029e-06, "loss": 0.908, "step": 3929 }, { "epoch": 0.34, "learning_rate": 7.748122140965859e-06, "loss": 0.8393, "step": 3930 }, { "epoch": 0.34, "learning_rate": 7.746967587274407e-06, "loss": 0.84, "step": 3931 }, { "epoch": 0.34, "learning_rate": 7.745812823755866e-06, "loss": 0.8616, "step": 3932 }, { "epoch": 0.34, "learning_rate": 7.744657850498442e-06, "loss": 0.8564, "step": 3933 }, { "epoch": 0.34, "learning_rate": 7.743502667590356e-06, "loss": 0.8969, "step": 3934 }, { "epoch": 0.34, "learning_rate": 7.74234727511985e-06, "loss": 0.8433, "step": 3935 }, { "epoch": 0.34, "learning_rate": 7.741191673175174e-06, "loss": 0.8804, "step": 3936 }, { "epoch": 0.34, "learning_rate": 7.740035861844602e-06, "loss": 0.8759, "step": 3937 }, { "epoch": 0.34, "learning_rate": 7.73887984121642e-06, "loss": 0.8673, "step": 3938 }, { "epoch": 0.34, "learning_rate": 7.73772361137893e-06, "loss": 0.9026, "step": 3939 }, { "epoch": 0.34, "learning_rate": 7.736567172420452e-06, "loss": 0.8895, "step": 3940 }, { "epoch": 0.34, "learning_rate": 7.735410524429319e-06, "loss": 0.8554, "step": 3941 }, { "epoch": 0.34, "learning_rate": 7.734253667493882e-06, "loss": 0.8526, "step": 3942 }, { "epoch": 0.34, "learning_rate": 7.733096601702508e-06, "loss": 0.8487, "step": 3943 }, { "epoch": 0.34, "learning_rate": 7.731939327143579e-06, "loss": 0.8483, "step": 3944 }, { "epoch": 0.34, "learning_rate": 7.730781843905494e-06, "loss": 0.8339, "step": 3945 }, { "epoch": 0.34, "learning_rate": 7.729624152076665e-06, "loss": 0.8804, "step": 3946 }, { "epoch": 0.34, "learning_rate": 7.728466251745525e-06, "loss": 0.8413, "step": 3947 }, { "epoch": 0.34, "learning_rate": 7.727308143000518e-06, "loss": 0.8988, "step": 3948 }, { "epoch": 0.34, "learning_rate": 7.726149825930108e-06, "loss": 0.8669, "step": 3949 }, { "epoch": 0.34, "learning_rate": 7.724991300622776e-06, "loss": 0.8801, "step": 3950 }, { "epoch": 0.34, "learning_rate": 7.723832567167008e-06, "loss": 0.8657, "step": 3951 }, { "epoch": 0.34, "learning_rate": 7.722673625651321e-06, "loss": 0.8294, "step": 3952 }, { "epoch": 0.34, "learning_rate": 7.721514476164237e-06, "loss": 0.8253, "step": 3953 }, { "epoch": 0.34, "learning_rate": 7.720355118794299e-06, "loss": 0.8872, "step": 3954 }, { "epoch": 0.34, "learning_rate": 7.719195553630064e-06, "loss": 0.9049, "step": 3955 }, { "epoch": 0.34, "learning_rate": 7.718035780760104e-06, "loss": 0.8728, "step": 3956 }, { "epoch": 0.34, "learning_rate": 7.716875800273012e-06, "loss": 0.8229, "step": 3957 }, { "epoch": 0.34, "learning_rate": 7.715715612257389e-06, "loss": 0.892, "step": 3958 }, { "epoch": 0.34, "learning_rate": 7.71455521680186e-06, "loss": 0.6549, "step": 3959 }, { "epoch": 0.34, "learning_rate": 7.713394613995057e-06, "loss": 0.8044, "step": 3960 }, { "epoch": 0.34, "learning_rate": 7.712233803925637e-06, "loss": 0.863, "step": 3961 }, { "epoch": 0.34, "learning_rate": 7.711072786682265e-06, "loss": 0.8354, "step": 3962 }, { "epoch": 0.34, "learning_rate": 7.70991156235363e-06, "loss": 0.8679, "step": 3963 }, { "epoch": 0.34, "learning_rate": 7.708750131028426e-06, "loss": 0.8388, "step": 3964 }, { "epoch": 0.34, "learning_rate": 7.707588492795372e-06, "loss": 0.8671, "step": 3965 }, { "epoch": 0.34, "learning_rate": 7.706426647743202e-06, "loss": 0.863, "step": 3966 }, { "epoch": 0.34, "learning_rate": 7.70526459596066e-06, "loss": 0.8771, "step": 3967 }, { "epoch": 0.34, "learning_rate": 7.70410233753651e-06, "loss": 0.8483, "step": 3968 }, { "epoch": 0.34, "learning_rate": 7.702939872559532e-06, "loss": 0.8973, "step": 3969 }, { "epoch": 0.34, "learning_rate": 7.70177720111852e-06, "loss": 0.8559, "step": 3970 }, { "epoch": 0.34, "learning_rate": 7.700614323302287e-06, "loss": 0.9095, "step": 3971 }, { "epoch": 0.34, "learning_rate": 7.699451239199657e-06, "loss": 0.875, "step": 3972 }, { "epoch": 0.34, "learning_rate": 7.69828794889947e-06, "loss": 0.8745, "step": 3973 }, { "epoch": 0.34, "learning_rate": 7.69712445249059e-06, "loss": 0.8127, "step": 3974 }, { "epoch": 0.34, "learning_rate": 7.695960750061886e-06, "loss": 0.8899, "step": 3975 }, { "epoch": 0.34, "learning_rate": 7.69479684170225e-06, "loss": 0.8742, "step": 3976 }, { "epoch": 0.34, "learning_rate": 7.693632727500584e-06, "loss": 0.5806, "step": 3977 }, { "epoch": 0.34, "learning_rate": 7.692468407545812e-06, "loss": 0.8481, "step": 3978 }, { "epoch": 0.34, "learning_rate": 7.691303881926868e-06, "loss": 0.8192, "step": 3979 }, { "epoch": 0.34, "learning_rate": 7.690139150732706e-06, "loss": 0.6514, "step": 3980 }, { "epoch": 0.34, "learning_rate": 7.688974214052294e-06, "loss": 0.872, "step": 3981 }, { "epoch": 0.34, "learning_rate": 7.687809071974615e-06, "loss": 0.8413, "step": 3982 }, { "epoch": 0.34, "learning_rate": 7.686643724588668e-06, "loss": 0.8751, "step": 3983 }, { "epoch": 0.34, "learning_rate": 7.685478171983468e-06, "loss": 0.9088, "step": 3984 }, { "epoch": 0.34, "learning_rate": 7.684312414248049e-06, "loss": 0.8179, "step": 3985 }, { "epoch": 0.34, "learning_rate": 7.683146451471453e-06, "loss": 0.8383, "step": 3986 }, { "epoch": 0.34, "learning_rate": 7.681980283742743e-06, "loss": 0.8585, "step": 3987 }, { "epoch": 0.34, "learning_rate": 7.680813911150998e-06, "loss": 0.8686, "step": 3988 }, { "epoch": 0.34, "learning_rate": 7.679647333785312e-06, "loss": 0.8739, "step": 3989 }, { "epoch": 0.34, "learning_rate": 7.678480551734791e-06, "loss": 0.9037, "step": 3990 }, { "epoch": 0.34, "learning_rate": 7.677313565088562e-06, "loss": 0.8883, "step": 3991 }, { "epoch": 0.34, "learning_rate": 7.676146373935766e-06, "loss": 0.8794, "step": 3992 }, { "epoch": 0.34, "learning_rate": 7.674978978365557e-06, "loss": 0.6463, "step": 3993 }, { "epoch": 0.34, "learning_rate": 7.673811378467107e-06, "loss": 0.8771, "step": 3994 }, { "epoch": 0.34, "learning_rate": 7.672643574329603e-06, "loss": 0.8528, "step": 3995 }, { "epoch": 0.34, "learning_rate": 7.671475566042248e-06, "loss": 0.8786, "step": 3996 }, { "epoch": 0.34, "learning_rate": 7.67030735369426e-06, "loss": 0.8464, "step": 3997 }, { "epoch": 0.34, "learning_rate": 7.669138937374874e-06, "loss": 0.8629, "step": 3998 }, { "epoch": 0.34, "learning_rate": 7.667970317173338e-06, "loss": 0.8179, "step": 3999 }, { "epoch": 0.34, "learning_rate": 7.666801493178916e-06, "loss": 0.8572, "step": 4000 }, { "epoch": 0.34, "learning_rate": 7.665632465480894e-06, "loss": 0.8716, "step": 4001 }, { "epoch": 0.34, "learning_rate": 7.664463234168562e-06, "loss": 0.8992, "step": 4002 }, { "epoch": 0.34, "learning_rate": 7.663293799331234e-06, "loss": 0.9274, "step": 4003 }, { "epoch": 0.34, "learning_rate": 7.66212416105824e-06, "loss": 0.7553, "step": 4004 }, { "epoch": 0.34, "learning_rate": 7.660954319438917e-06, "loss": 0.8443, "step": 4005 }, { "epoch": 0.34, "learning_rate": 7.659784274562627e-06, "loss": 0.859, "step": 4006 }, { "epoch": 0.34, "learning_rate": 7.658614026518743e-06, "loss": 0.8691, "step": 4007 }, { "epoch": 0.34, "learning_rate": 7.657443575396657e-06, "loss": 0.8699, "step": 4008 }, { "epoch": 0.34, "learning_rate": 7.65627292128577e-06, "loss": 0.8638, "step": 4009 }, { "epoch": 0.34, "learning_rate": 7.655102064275506e-06, "loss": 0.831, "step": 4010 }, { "epoch": 0.34, "learning_rate": 7.653931004455296e-06, "loss": 0.8424, "step": 4011 }, { "epoch": 0.34, "learning_rate": 7.652759741914597e-06, "loss": 0.8166, "step": 4012 }, { "epoch": 0.34, "learning_rate": 7.651588276742872e-06, "loss": 0.89, "step": 4013 }, { "epoch": 0.34, "learning_rate": 7.650416609029606e-06, "loss": 0.894, "step": 4014 }, { "epoch": 0.34, "learning_rate": 7.649244738864294e-06, "loss": 0.8867, "step": 4015 }, { "epoch": 0.34, "learning_rate": 7.64807266633645e-06, "loss": 0.8665, "step": 4016 }, { "epoch": 0.34, "learning_rate": 7.646900391535606e-06, "loss": 0.8951, "step": 4017 }, { "epoch": 0.34, "learning_rate": 7.645727914551301e-06, "loss": 0.8875, "step": 4018 }, { "epoch": 0.34, "learning_rate": 7.6445552354731e-06, "loss": 0.8446, "step": 4019 }, { "epoch": 0.34, "learning_rate": 7.643382354390572e-06, "loss": 0.8636, "step": 4020 }, { "epoch": 0.34, "learning_rate": 7.642209271393313e-06, "loss": 0.9133, "step": 4021 }, { "epoch": 0.34, "learning_rate": 7.641035986570927e-06, "loss": 0.8649, "step": 4022 }, { "epoch": 0.34, "learning_rate": 7.639862500013036e-06, "loss": 0.8121, "step": 4023 }, { "epoch": 0.34, "learning_rate": 7.638688811809274e-06, "loss": 0.862, "step": 4024 }, { "epoch": 0.34, "learning_rate": 7.637514922049296e-06, "loss": 0.8775, "step": 4025 }, { "epoch": 0.34, "learning_rate": 7.636340830822769e-06, "loss": 0.833, "step": 4026 }, { "epoch": 0.34, "learning_rate": 7.635166538219374e-06, "loss": 0.8098, "step": 4027 }, { "epoch": 0.34, "learning_rate": 7.633992044328814e-06, "loss": 0.8564, "step": 4028 }, { "epoch": 0.34, "learning_rate": 7.632817349240796e-06, "loss": 0.897, "step": 4029 }, { "epoch": 0.34, "learning_rate": 7.631642453045053e-06, "loss": 0.8731, "step": 4030 }, { "epoch": 0.34, "learning_rate": 7.630467355831332e-06, "loss": 0.8773, "step": 4031 }, { "epoch": 0.34, "learning_rate": 7.629292057689389e-06, "loss": 0.8617, "step": 4032 }, { "epoch": 0.34, "learning_rate": 7.628116558709e-06, "loss": 0.8496, "step": 4033 }, { "epoch": 0.34, "learning_rate": 7.626940858979954e-06, "loss": 0.8756, "step": 4034 }, { "epoch": 0.34, "learning_rate": 7.62576495859206e-06, "loss": 0.8191, "step": 4035 }, { "epoch": 0.34, "learning_rate": 7.624588857635138e-06, "loss": 0.8663, "step": 4036 }, { "epoch": 0.34, "learning_rate": 7.623412556199021e-06, "loss": 0.8767, "step": 4037 }, { "epoch": 0.34, "learning_rate": 7.622236054373567e-06, "loss": 0.8224, "step": 4038 }, { "epoch": 0.34, "learning_rate": 7.621059352248639e-06, "loss": 0.8359, "step": 4039 }, { "epoch": 0.34, "learning_rate": 7.619882449914119e-06, "loss": 0.8831, "step": 4040 }, { "epoch": 0.34, "learning_rate": 7.618705347459907e-06, "loss": 0.8544, "step": 4041 }, { "epoch": 0.34, "learning_rate": 7.617528044975915e-06, "loss": 0.89, "step": 4042 }, { "epoch": 0.34, "learning_rate": 7.616350542552069e-06, "loss": 0.8252, "step": 4043 }, { "epoch": 0.35, "learning_rate": 7.615172840278316e-06, "loss": 0.7988, "step": 4044 }, { "epoch": 0.35, "learning_rate": 7.613994938244613e-06, "loss": 0.9079, "step": 4045 }, { "epoch": 0.35, "learning_rate": 7.612816836540936e-06, "loss": 0.8658, "step": 4046 }, { "epoch": 0.35, "learning_rate": 7.611638535257271e-06, "loss": 0.884, "step": 4047 }, { "epoch": 0.35, "learning_rate": 7.610460034483625e-06, "loss": 0.8997, "step": 4048 }, { "epoch": 0.35, "learning_rate": 7.609281334310017e-06, "loss": 0.8607, "step": 4049 }, { "epoch": 0.35, "learning_rate": 7.608102434826481e-06, "loss": 0.8584, "step": 4050 }, { "epoch": 0.35, "learning_rate": 7.6069233361230696e-06, "loss": 0.84, "step": 4051 }, { "epoch": 0.35, "learning_rate": 7.605744038289846e-06, "loss": 0.8312, "step": 4052 }, { "epoch": 0.35, "learning_rate": 7.60456454141689e-06, "loss": 0.8788, "step": 4053 }, { "epoch": 0.35, "learning_rate": 7.603384845594302e-06, "loss": 0.8795, "step": 4054 }, { "epoch": 0.35, "learning_rate": 7.602204950912189e-06, "loss": 0.8438, "step": 4055 }, { "epoch": 0.35, "learning_rate": 7.601024857460678e-06, "loss": 0.8481, "step": 4056 }, { "epoch": 0.35, "learning_rate": 7.599844565329911e-06, "loss": 0.831, "step": 4057 }, { "epoch": 0.35, "learning_rate": 7.598664074610045e-06, "loss": 0.8885, "step": 4058 }, { "epoch": 0.35, "learning_rate": 7.597483385391251e-06, "loss": 0.8253, "step": 4059 }, { "epoch": 0.35, "learning_rate": 7.596302497763717e-06, "loss": 0.6058, "step": 4060 }, { "epoch": 0.35, "learning_rate": 7.595121411817643e-06, "loss": 0.8299, "step": 4061 }, { "epoch": 0.35, "learning_rate": 7.593940127643248e-06, "loss": 0.8381, "step": 4062 }, { "epoch": 0.35, "learning_rate": 7.592758645330763e-06, "loss": 0.8929, "step": 4063 }, { "epoch": 0.35, "learning_rate": 7.591576964970437e-06, "loss": 0.8152, "step": 4064 }, { "epoch": 0.35, "learning_rate": 7.590395086652533e-06, "loss": 0.8564, "step": 4065 }, { "epoch": 0.35, "learning_rate": 7.589213010467326e-06, "loss": 0.8983, "step": 4066 }, { "epoch": 0.35, "learning_rate": 7.588030736505112e-06, "loss": 0.8377, "step": 4067 }, { "epoch": 0.35, "learning_rate": 7.586848264856198e-06, "loss": 0.8673, "step": 4068 }, { "epoch": 0.35, "learning_rate": 7.585665595610907e-06, "loss": 0.8625, "step": 4069 }, { "epoch": 0.35, "learning_rate": 7.584482728859575e-06, "loss": 0.9044, "step": 4070 }, { "epoch": 0.35, "learning_rate": 7.583299664692558e-06, "loss": 0.8709, "step": 4071 }, { "epoch": 0.35, "learning_rate": 7.5821164032002235e-06, "loss": 0.8546, "step": 4072 }, { "epoch": 0.35, "learning_rate": 7.580932944472955e-06, "loss": 0.8416, "step": 4073 }, { "epoch": 0.35, "learning_rate": 7.579749288601152e-06, "loss": 0.8868, "step": 4074 }, { "epoch": 0.35, "learning_rate": 7.578565435675226e-06, "loss": 0.8948, "step": 4075 }, { "epoch": 0.35, "learning_rate": 7.577381385785606e-06, "loss": 0.8723, "step": 4076 }, { "epoch": 0.35, "learning_rate": 7.576197139022736e-06, "loss": 0.8515, "step": 4077 }, { "epoch": 0.35, "learning_rate": 7.575012695477076e-06, "loss": 0.8252, "step": 4078 }, { "epoch": 0.35, "learning_rate": 7.573828055239099e-06, "loss": 0.8487, "step": 4079 }, { "epoch": 0.35, "learning_rate": 7.572643218399291e-06, "loss": 0.8619, "step": 4080 }, { "epoch": 0.35, "learning_rate": 7.571458185048162e-06, "loss": 0.887, "step": 4081 }, { "epoch": 0.35, "learning_rate": 7.570272955276223e-06, "loss": 0.8501, "step": 4082 }, { "epoch": 0.35, "learning_rate": 7.569087529174013e-06, "loss": 0.8314, "step": 4083 }, { "epoch": 0.35, "learning_rate": 7.567901906832079e-06, "loss": 0.8592, "step": 4084 }, { "epoch": 0.35, "learning_rate": 7.5667160883409865e-06, "loss": 0.8478, "step": 4085 }, { "epoch": 0.35, "learning_rate": 7.565530073791311e-06, "loss": 0.8359, "step": 4086 }, { "epoch": 0.35, "learning_rate": 7.5643438632736485e-06, "loss": 0.8586, "step": 4087 }, { "epoch": 0.35, "learning_rate": 7.563157456878609e-06, "loss": 0.8695, "step": 4088 }, { "epoch": 0.35, "learning_rate": 7.561970854696812e-06, "loss": 0.8859, "step": 4089 }, { "epoch": 0.35, "learning_rate": 7.560784056818898e-06, "loss": 0.8758, "step": 4090 }, { "epoch": 0.35, "learning_rate": 7.559597063335523e-06, "loss": 0.9135, "step": 4091 }, { "epoch": 0.35, "learning_rate": 7.558409874337352e-06, "loss": 0.8905, "step": 4092 }, { "epoch": 0.35, "learning_rate": 7.5572224899150725e-06, "loss": 0.8852, "step": 4093 }, { "epoch": 0.35, "learning_rate": 7.556034910159378e-06, "loss": 0.8969, "step": 4094 }, { "epoch": 0.35, "learning_rate": 7.554847135160984e-06, "loss": 0.8832, "step": 4095 }, { "epoch": 0.35, "learning_rate": 7.553659165010619e-06, "loss": 0.8458, "step": 4096 }, { "epoch": 0.35, "learning_rate": 7.552470999799026e-06, "loss": 0.8226, "step": 4097 }, { "epoch": 0.35, "learning_rate": 7.551282639616964e-06, "loss": 0.9146, "step": 4098 }, { "epoch": 0.35, "learning_rate": 7.550094084555201e-06, "loss": 0.8783, "step": 4099 }, { "epoch": 0.35, "learning_rate": 7.548905334704531e-06, "loss": 0.8933, "step": 4100 }, { "epoch": 0.35, "learning_rate": 7.547716390155755e-06, "loss": 0.8267, "step": 4101 }, { "epoch": 0.35, "learning_rate": 7.546527250999688e-06, "loss": 0.8942, "step": 4102 }, { "epoch": 0.35, "learning_rate": 7.5453379173271646e-06, "loss": 0.8946, "step": 4103 }, { "epoch": 0.35, "learning_rate": 7.544148389229031e-06, "loss": 0.8916, "step": 4104 }, { "epoch": 0.35, "learning_rate": 7.542958666796149e-06, "loss": 0.8716, "step": 4105 }, { "epoch": 0.35, "learning_rate": 7.5417687501193985e-06, "loss": 0.8675, "step": 4106 }, { "epoch": 0.35, "learning_rate": 7.540578639289666e-06, "loss": 0.874, "step": 4107 }, { "epoch": 0.35, "learning_rate": 7.539388334397864e-06, "loss": 0.8806, "step": 4108 }, { "epoch": 0.35, "learning_rate": 7.538197835534908e-06, "loss": 0.8287, "step": 4109 }, { "epoch": 0.35, "learning_rate": 7.537007142791739e-06, "loss": 0.8866, "step": 4110 }, { "epoch": 0.35, "learning_rate": 7.535816256259306e-06, "loss": 0.8503, "step": 4111 }, { "epoch": 0.35, "learning_rate": 7.534625176028576e-06, "loss": 0.9006, "step": 4112 }, { "epoch": 0.35, "learning_rate": 7.5334339021905275e-06, "loss": 0.8575, "step": 4113 }, { "epoch": 0.35, "learning_rate": 7.532242434836159e-06, "loss": 0.8295, "step": 4114 }, { "epoch": 0.35, "learning_rate": 7.531050774056476e-06, "loss": 0.8652, "step": 4115 }, { "epoch": 0.35, "learning_rate": 7.52985891994251e-06, "loss": 0.8441, "step": 4116 }, { "epoch": 0.35, "learning_rate": 7.528666872585295e-06, "loss": 0.8831, "step": 4117 }, { "epoch": 0.35, "learning_rate": 7.527474632075889e-06, "loss": 0.8292, "step": 4118 }, { "epoch": 0.35, "learning_rate": 7.526282198505359e-06, "loss": 0.5659, "step": 4119 }, { "epoch": 0.35, "learning_rate": 7.52508957196479e-06, "loss": 0.8647, "step": 4120 }, { "epoch": 0.35, "learning_rate": 7.523896752545282e-06, "loss": 0.8442, "step": 4121 }, { "epoch": 0.35, "learning_rate": 7.522703740337946e-06, "loss": 0.8877, "step": 4122 }, { "epoch": 0.35, "learning_rate": 7.521510535433913e-06, "loss": 0.8685, "step": 4123 }, { "epoch": 0.35, "learning_rate": 7.5203171379243235e-06, "loss": 0.8419, "step": 4124 }, { "epoch": 0.35, "learning_rate": 7.519123547900336e-06, "loss": 0.8892, "step": 4125 }, { "epoch": 0.35, "learning_rate": 7.517929765453124e-06, "loss": 0.8847, "step": 4126 }, { "epoch": 0.35, "learning_rate": 7.516735790673872e-06, "loss": 0.8704, "step": 4127 }, { "epoch": 0.35, "learning_rate": 7.515541623653786e-06, "loss": 0.8465, "step": 4128 }, { "epoch": 0.35, "learning_rate": 7.5143472644840765e-06, "loss": 0.8443, "step": 4129 }, { "epoch": 0.35, "learning_rate": 7.51315271325598e-06, "loss": 0.8872, "step": 4130 }, { "epoch": 0.35, "learning_rate": 7.511957970060741e-06, "loss": 0.8627, "step": 4131 }, { "epoch": 0.35, "learning_rate": 7.510763034989616e-06, "loss": 0.8722, "step": 4132 }, { "epoch": 0.35, "learning_rate": 7.5095679081338855e-06, "loss": 0.8939, "step": 4133 }, { "epoch": 0.35, "learning_rate": 7.508372589584836e-06, "loss": 0.807, "step": 4134 }, { "epoch": 0.35, "learning_rate": 7.507177079433774e-06, "loss": 0.88, "step": 4135 }, { "epoch": 0.35, "learning_rate": 7.505981377772016e-06, "loss": 0.8496, "step": 4136 }, { "epoch": 0.35, "learning_rate": 7.504785484690898e-06, "loss": 0.858, "step": 4137 }, { "epoch": 0.35, "learning_rate": 7.503589400281767e-06, "loss": 0.8564, "step": 4138 }, { "epoch": 0.35, "learning_rate": 7.5023931246359875e-06, "loss": 0.8318, "step": 4139 }, { "epoch": 0.35, "learning_rate": 7.5011966578449335e-06, "loss": 0.8685, "step": 4140 }, { "epoch": 0.35, "learning_rate": 7.500000000000001e-06, "loss": 0.8047, "step": 4141 }, { "epoch": 0.35, "learning_rate": 7.498803151192594e-06, "loss": 0.8723, "step": 4142 }, { "epoch": 0.35, "learning_rate": 7.497606111514135e-06, "loss": 0.876, "step": 4143 }, { "epoch": 0.35, "learning_rate": 7.496408881056061e-06, "loss": 0.9284, "step": 4144 }, { "epoch": 0.35, "learning_rate": 7.495211459909819e-06, "loss": 0.9032, "step": 4145 }, { "epoch": 0.35, "learning_rate": 7.4940138481668765e-06, "loss": 0.8433, "step": 4146 }, { "epoch": 0.35, "learning_rate": 7.492816045918714e-06, "loss": 0.8906, "step": 4147 }, { "epoch": 0.35, "learning_rate": 7.4916180532568235e-06, "loss": 0.8763, "step": 4148 }, { "epoch": 0.35, "learning_rate": 7.490419870272715e-06, "loss": 0.8226, "step": 4149 }, { "epoch": 0.35, "learning_rate": 7.4892214970579105e-06, "loss": 0.8653, "step": 4150 }, { "epoch": 0.35, "learning_rate": 7.488022933703949e-06, "loss": 0.7979, "step": 4151 }, { "epoch": 0.35, "learning_rate": 7.4868241803023815e-06, "loss": 0.8529, "step": 4152 }, { "epoch": 0.35, "learning_rate": 7.485625236944776e-06, "loss": 0.8392, "step": 4153 }, { "epoch": 0.35, "learning_rate": 7.4844261037227124e-06, "loss": 0.9006, "step": 4154 }, { "epoch": 0.35, "learning_rate": 7.483226780727787e-06, "loss": 0.6159, "step": 4155 }, { "epoch": 0.35, "learning_rate": 7.4820272680516105e-06, "loss": 0.8476, "step": 4156 }, { "epoch": 0.35, "learning_rate": 7.480827565785807e-06, "loss": 0.8906, "step": 4157 }, { "epoch": 0.35, "learning_rate": 7.479627674022018e-06, "loss": 0.8441, "step": 4158 }, { "epoch": 0.35, "learning_rate": 7.478427592851894e-06, "loss": 0.8641, "step": 4159 }, { "epoch": 0.35, "learning_rate": 7.477227322367104e-06, "loss": 0.5658, "step": 4160 }, { "epoch": 0.36, "learning_rate": 7.476026862659333e-06, "loss": 0.9048, "step": 4161 }, { "epoch": 0.36, "learning_rate": 7.474826213820275e-06, "loss": 0.8435, "step": 4162 }, { "epoch": 0.36, "learning_rate": 7.473625375941644e-06, "loss": 0.8612, "step": 4163 }, { "epoch": 0.36, "learning_rate": 7.472424349115165e-06, "loss": 0.8384, "step": 4164 }, { "epoch": 0.36, "learning_rate": 7.471223133432577e-06, "loss": 0.8365, "step": 4165 }, { "epoch": 0.36, "learning_rate": 7.470021728985637e-06, "loss": 0.8949, "step": 4166 }, { "epoch": 0.36, "learning_rate": 7.4688201358661135e-06, "loss": 0.8584, "step": 4167 }, { "epoch": 0.36, "learning_rate": 7.467618354165791e-06, "loss": 0.8736, "step": 4168 }, { "epoch": 0.36, "learning_rate": 7.466416383976465e-06, "loss": 0.547, "step": 4169 }, { "epoch": 0.36, "learning_rate": 7.46521422538995e-06, "loss": 0.8375, "step": 4170 }, { "epoch": 0.36, "learning_rate": 7.464011878498073e-06, "loss": 0.5176, "step": 4171 }, { "epoch": 0.36, "learning_rate": 7.462809343392675e-06, "loss": 0.867, "step": 4172 }, { "epoch": 0.36, "learning_rate": 7.461606620165612e-06, "loss": 0.873, "step": 4173 }, { "epoch": 0.36, "learning_rate": 7.460403708908752e-06, "loss": 0.8653, "step": 4174 }, { "epoch": 0.36, "learning_rate": 7.4592006097139815e-06, "loss": 0.8708, "step": 4175 }, { "epoch": 0.36, "learning_rate": 7.457997322673199e-06, "loss": 0.8981, "step": 4176 }, { "epoch": 0.36, "learning_rate": 7.456793847878316e-06, "loss": 0.8723, "step": 4177 }, { "epoch": 0.36, "learning_rate": 7.455590185421262e-06, "loss": 0.8526, "step": 4178 }, { "epoch": 0.36, "learning_rate": 7.4543863353939764e-06, "loss": 0.8725, "step": 4179 }, { "epoch": 0.36, "learning_rate": 7.453182297888417e-06, "loss": 0.526, "step": 4180 }, { "epoch": 0.36, "learning_rate": 7.4519780729965555e-06, "loss": 0.5502, "step": 4181 }, { "epoch": 0.36, "learning_rate": 7.450773660810373e-06, "loss": 0.8754, "step": 4182 }, { "epoch": 0.36, "learning_rate": 7.449569061421871e-06, "loss": 0.8737, "step": 4183 }, { "epoch": 0.36, "learning_rate": 7.448364274923063e-06, "loss": 0.8594, "step": 4184 }, { "epoch": 0.36, "learning_rate": 7.447159301405975e-06, "loss": 0.8594, "step": 4185 }, { "epoch": 0.36, "learning_rate": 7.44595414096265e-06, "loss": 0.8638, "step": 4186 }, { "epoch": 0.36, "learning_rate": 7.444748793685143e-06, "loss": 0.8564, "step": 4187 }, { "epoch": 0.36, "learning_rate": 7.443543259665527e-06, "loss": 0.7923, "step": 4188 }, { "epoch": 0.36, "learning_rate": 7.442337538995883e-06, "loss": 0.8459, "step": 4189 }, { "epoch": 0.36, "learning_rate": 7.441131631768315e-06, "loss": 0.809, "step": 4190 }, { "epoch": 0.36, "learning_rate": 7.439925538074932e-06, "loss": 0.8729, "step": 4191 }, { "epoch": 0.36, "learning_rate": 7.43871925800786e-06, "loss": 0.8466, "step": 4192 }, { "epoch": 0.36, "learning_rate": 7.437512791659245e-06, "loss": 0.8719, "step": 4193 }, { "epoch": 0.36, "learning_rate": 7.436306139121243e-06, "loss": 0.8179, "step": 4194 }, { "epoch": 0.36, "learning_rate": 7.435099300486021e-06, "loss": 0.83, "step": 4195 }, { "epoch": 0.36, "learning_rate": 7.433892275845764e-06, "loss": 0.8528, "step": 4196 }, { "epoch": 0.36, "learning_rate": 7.432685065292673e-06, "loss": 0.8974, "step": 4197 }, { "epoch": 0.36, "learning_rate": 7.4314776689189585e-06, "loss": 0.8618, "step": 4198 }, { "epoch": 0.36, "learning_rate": 7.430270086816846e-06, "loss": 0.8515, "step": 4199 }, { "epoch": 0.36, "learning_rate": 7.429062319078582e-06, "loss": 0.8886, "step": 4200 }, { "epoch": 0.36, "learning_rate": 7.427854365796418e-06, "loss": 0.8788, "step": 4201 }, { "epoch": 0.36, "learning_rate": 7.426646227062621e-06, "loss": 0.8574, "step": 4202 }, { "epoch": 0.36, "learning_rate": 7.4254379029694814e-06, "loss": 0.8994, "step": 4203 }, { "epoch": 0.36, "learning_rate": 7.424229393609291e-06, "loss": 0.8358, "step": 4204 }, { "epoch": 0.36, "learning_rate": 7.423020699074366e-06, "loss": 0.8802, "step": 4205 }, { "epoch": 0.36, "learning_rate": 7.421811819457029e-06, "loss": 0.8665, "step": 4206 }, { "epoch": 0.36, "learning_rate": 7.420602754849622e-06, "loss": 0.9083, "step": 4207 }, { "epoch": 0.36, "learning_rate": 7.419393505344499e-06, "loss": 0.8799, "step": 4208 }, { "epoch": 0.36, "learning_rate": 7.41818407103403e-06, "loss": 0.8413, "step": 4209 }, { "epoch": 0.36, "learning_rate": 7.416974452010596e-06, "loss": 0.8796, "step": 4210 }, { "epoch": 0.36, "learning_rate": 7.415764648366593e-06, "loss": 0.864, "step": 4211 }, { "epoch": 0.36, "learning_rate": 7.414554660194434e-06, "loss": 0.8332, "step": 4212 }, { "epoch": 0.36, "learning_rate": 7.413344487586542e-06, "loss": 0.8455, "step": 4213 }, { "epoch": 0.36, "learning_rate": 7.41213413063536e-06, "loss": 0.8301, "step": 4214 }, { "epoch": 0.36, "learning_rate": 7.410923589433334e-06, "loss": 0.902, "step": 4215 }, { "epoch": 0.36, "learning_rate": 7.409712864072938e-06, "loss": 0.8212, "step": 4216 }, { "epoch": 0.36, "learning_rate": 7.408501954646648e-06, "loss": 0.8837, "step": 4217 }, { "epoch": 0.36, "learning_rate": 7.407290861246963e-06, "loss": 0.8494, "step": 4218 }, { "epoch": 0.36, "learning_rate": 7.406079583966392e-06, "loss": 0.7918, "step": 4219 }, { "epoch": 0.36, "learning_rate": 7.404868122897456e-06, "loss": 0.8803, "step": 4220 }, { "epoch": 0.36, "learning_rate": 7.403656478132694e-06, "loss": 0.8778, "step": 4221 }, { "epoch": 0.36, "learning_rate": 7.40244464976466e-06, "loss": 0.8759, "step": 4222 }, { "epoch": 0.36, "learning_rate": 7.401232637885914e-06, "loss": 0.8369, "step": 4223 }, { "epoch": 0.36, "learning_rate": 7.40002044258904e-06, "loss": 0.8518, "step": 4224 }, { "epoch": 0.36, "learning_rate": 7.398808063966629e-06, "loss": 0.6258, "step": 4225 }, { "epoch": 0.36, "learning_rate": 7.39759550211129e-06, "loss": 0.8754, "step": 4226 }, { "epoch": 0.36, "learning_rate": 7.396382757115643e-06, "loss": 0.8371, "step": 4227 }, { "epoch": 0.36, "learning_rate": 7.395169829072326e-06, "loss": 0.8629, "step": 4228 }, { "epoch": 0.36, "learning_rate": 7.393956718073987e-06, "loss": 0.831, "step": 4229 }, { "epoch": 0.36, "learning_rate": 7.392743424213289e-06, "loss": 0.8784, "step": 4230 }, { "epoch": 0.36, "learning_rate": 7.391529947582909e-06, "loss": 0.9054, "step": 4231 }, { "epoch": 0.36, "learning_rate": 7.390316288275541e-06, "loss": 0.8155, "step": 4232 }, { "epoch": 0.36, "learning_rate": 7.389102446383887e-06, "loss": 0.8316, "step": 4233 }, { "epoch": 0.36, "learning_rate": 7.38788842200067e-06, "loss": 0.8554, "step": 4234 }, { "epoch": 0.36, "learning_rate": 7.3866742152186185e-06, "loss": 0.8596, "step": 4235 }, { "epoch": 0.36, "learning_rate": 7.385459826130484e-06, "loss": 0.8109, "step": 4236 }, { "epoch": 0.36, "learning_rate": 7.384245254829026e-06, "loss": 0.846, "step": 4237 }, { "epoch": 0.36, "learning_rate": 7.38303050140702e-06, "loss": 0.8584, "step": 4238 }, { "epoch": 0.36, "learning_rate": 7.381815565957255e-06, "loss": 0.8482, "step": 4239 }, { "epoch": 0.36, "learning_rate": 7.380600448572532e-06, "loss": 0.8969, "step": 4240 }, { "epoch": 0.36, "learning_rate": 7.3793851493456685e-06, "loss": 0.878, "step": 4241 }, { "epoch": 0.36, "learning_rate": 7.378169668369497e-06, "loss": 0.8659, "step": 4242 }, { "epoch": 0.36, "learning_rate": 7.376954005736858e-06, "loss": 0.9322, "step": 4243 }, { "epoch": 0.36, "learning_rate": 7.375738161540616e-06, "loss": 0.8705, "step": 4244 }, { "epoch": 0.36, "learning_rate": 7.374522135873637e-06, "loss": 0.8889, "step": 4245 }, { "epoch": 0.36, "learning_rate": 7.373305928828812e-06, "loss": 0.8481, "step": 4246 }, { "epoch": 0.36, "learning_rate": 7.372089540499038e-06, "loss": 0.8967, "step": 4247 }, { "epoch": 0.36, "learning_rate": 7.370872970977229e-06, "loss": 0.9077, "step": 4248 }, { "epoch": 0.36, "learning_rate": 7.369656220356314e-06, "loss": 0.8399, "step": 4249 }, { "epoch": 0.36, "learning_rate": 7.368439288729233e-06, "loss": 0.8391, "step": 4250 }, { "epoch": 0.36, "learning_rate": 7.367222176188943e-06, "loss": 0.8507, "step": 4251 }, { "epoch": 0.36, "learning_rate": 7.366004882828412e-06, "loss": 0.8655, "step": 4252 }, { "epoch": 0.36, "learning_rate": 7.364787408740621e-06, "loss": 0.8502, "step": 4253 }, { "epoch": 0.36, "learning_rate": 7.3635697540185715e-06, "loss": 0.8707, "step": 4254 }, { "epoch": 0.36, "learning_rate": 7.362351918755271e-06, "loss": 0.8582, "step": 4255 }, { "epoch": 0.36, "learning_rate": 7.361133903043744e-06, "loss": 0.8627, "step": 4256 }, { "epoch": 0.36, "learning_rate": 7.359915706977029e-06, "loss": 0.8608, "step": 4257 }, { "epoch": 0.36, "learning_rate": 7.3586973306481765e-06, "loss": 0.8493, "step": 4258 }, { "epoch": 0.36, "learning_rate": 7.357478774150254e-06, "loss": 0.571, "step": 4259 }, { "epoch": 0.36, "learning_rate": 7.3562600375763415e-06, "loss": 0.8597, "step": 4260 }, { "epoch": 0.36, "learning_rate": 7.35504112101953e-06, "loss": 0.8681, "step": 4261 }, { "epoch": 0.36, "learning_rate": 7.353822024572927e-06, "loss": 0.8744, "step": 4262 }, { "epoch": 0.36, "learning_rate": 7.352602748329655e-06, "loss": 0.8512, "step": 4263 }, { "epoch": 0.36, "learning_rate": 7.351383292382847e-06, "loss": 0.8683, "step": 4264 }, { "epoch": 0.36, "learning_rate": 7.350163656825651e-06, "loss": 0.8522, "step": 4265 }, { "epoch": 0.36, "learning_rate": 7.348943841751229e-06, "loss": 0.8739, "step": 4266 }, { "epoch": 0.36, "learning_rate": 7.347723847252756e-06, "loss": 0.9028, "step": 4267 }, { "epoch": 0.36, "learning_rate": 7.346503673423421e-06, "loss": 0.7892, "step": 4268 }, { "epoch": 0.36, "learning_rate": 7.3452833203564275e-06, "loss": 0.8503, "step": 4269 }, { "epoch": 0.36, "learning_rate": 7.344062788144996e-06, "loss": 0.8337, "step": 4270 }, { "epoch": 0.36, "learning_rate": 7.342842076882351e-06, "loss": 0.8552, "step": 4271 }, { "epoch": 0.36, "learning_rate": 7.341621186661738e-06, "loss": 0.9155, "step": 4272 }, { "epoch": 0.36, "learning_rate": 7.3404001175764165e-06, "loss": 0.8832, "step": 4273 }, { "epoch": 0.36, "learning_rate": 7.339178869719656e-06, "loss": 0.5699, "step": 4274 }, { "epoch": 0.36, "learning_rate": 7.3379574431847444e-06, "loss": 0.5691, "step": 4275 }, { "epoch": 0.36, "learning_rate": 7.336735838064975e-06, "loss": 0.8804, "step": 4276 }, { "epoch": 0.36, "learning_rate": 7.335514054453667e-06, "loss": 0.675, "step": 4277 }, { "epoch": 0.37, "learning_rate": 7.334292092444141e-06, "loss": 0.8463, "step": 4278 }, { "epoch": 0.37, "learning_rate": 7.333069952129739e-06, "loss": 0.8812, "step": 4279 }, { "epoch": 0.37, "learning_rate": 7.331847633603813e-06, "loss": 0.8371, "step": 4280 }, { "epoch": 0.37, "learning_rate": 7.330625136959729e-06, "loss": 0.8669, "step": 4281 }, { "epoch": 0.37, "learning_rate": 7.329402462290869e-06, "loss": 0.5045, "step": 4282 }, { "epoch": 0.37, "learning_rate": 7.328179609690628e-06, "loss": 0.853, "step": 4283 }, { "epoch": 0.37, "learning_rate": 7.326956579252412e-06, "loss": 0.8471, "step": 4284 }, { "epoch": 0.37, "learning_rate": 7.325733371069642e-06, "loss": 0.8334, "step": 4285 }, { "epoch": 0.37, "learning_rate": 7.324509985235753e-06, "loss": 0.8435, "step": 4286 }, { "epoch": 0.37, "learning_rate": 7.323286421844193e-06, "loss": 0.818, "step": 4287 }, { "epoch": 0.37, "learning_rate": 7.322062680988425e-06, "loss": 0.8478, "step": 4288 }, { "epoch": 0.37, "learning_rate": 7.320838762761924e-06, "loss": 0.8971, "step": 4289 }, { "epoch": 0.37, "learning_rate": 7.319614667258179e-06, "loss": 0.8196, "step": 4290 }, { "epoch": 0.37, "learning_rate": 7.318390394570692e-06, "loss": 0.8692, "step": 4291 }, { "epoch": 0.37, "learning_rate": 7.317165944792978e-06, "loss": 0.8168, "step": 4292 }, { "epoch": 0.37, "learning_rate": 7.315941318018569e-06, "loss": 0.9004, "step": 4293 }, { "epoch": 0.37, "learning_rate": 7.314716514341007e-06, "loss": 0.8318, "step": 4294 }, { "epoch": 0.37, "learning_rate": 7.313491533853848e-06, "loss": 0.8677, "step": 4295 }, { "epoch": 0.37, "learning_rate": 7.312266376650663e-06, "loss": 0.8556, "step": 4296 }, { "epoch": 0.37, "learning_rate": 7.311041042825036e-06, "loss": 0.8251, "step": 4297 }, { "epoch": 0.37, "learning_rate": 7.309815532470563e-06, "loss": 0.891, "step": 4298 }, { "epoch": 0.37, "learning_rate": 7.3085898456808545e-06, "loss": 0.855, "step": 4299 }, { "epoch": 0.37, "learning_rate": 7.307363982549537e-06, "loss": 0.857, "step": 4300 }, { "epoch": 0.37, "learning_rate": 7.306137943170244e-06, "loss": 0.8558, "step": 4301 }, { "epoch": 0.37, "learning_rate": 7.30491172763663e-06, "loss": 0.873, "step": 4302 }, { "epoch": 0.37, "learning_rate": 7.303685336042359e-06, "loss": 0.8482, "step": 4303 }, { "epoch": 0.37, "learning_rate": 7.302458768481106e-06, "loss": 0.8973, "step": 4304 }, { "epoch": 0.37, "learning_rate": 7.301232025046564e-06, "loss": 0.8873, "step": 4305 }, { "epoch": 0.37, "learning_rate": 7.300005105832438e-06, "loss": 0.7868, "step": 4306 }, { "epoch": 0.37, "learning_rate": 7.298778010932449e-06, "loss": 0.8514, "step": 4307 }, { "epoch": 0.37, "learning_rate": 7.297550740440325e-06, "loss": 0.8879, "step": 4308 }, { "epoch": 0.37, "learning_rate": 7.296323294449812e-06, "loss": 0.8363, "step": 4309 }, { "epoch": 0.37, "learning_rate": 7.295095673054669e-06, "loss": 0.8621, "step": 4310 }, { "epoch": 0.37, "learning_rate": 7.2938678763486655e-06, "loss": 0.8292, "step": 4311 }, { "epoch": 0.37, "learning_rate": 7.29263990442559e-06, "loss": 0.8878, "step": 4312 }, { "epoch": 0.37, "learning_rate": 7.291411757379241e-06, "loss": 0.9094, "step": 4313 }, { "epoch": 0.37, "learning_rate": 7.290183435303426e-06, "loss": 0.8569, "step": 4314 }, { "epoch": 0.37, "learning_rate": 7.2889549382919755e-06, "loss": 0.875, "step": 4315 }, { "epoch": 0.37, "learning_rate": 7.287726266438726e-06, "loss": 0.8657, "step": 4316 }, { "epoch": 0.37, "learning_rate": 7.2864974198375314e-06, "loss": 0.8285, "step": 4317 }, { "epoch": 0.37, "learning_rate": 7.285268398582253e-06, "loss": 0.8737, "step": 4318 }, { "epoch": 0.37, "learning_rate": 7.284039202766774e-06, "loss": 0.9271, "step": 4319 }, { "epoch": 0.37, "learning_rate": 7.282809832484985e-06, "loss": 0.8821, "step": 4320 }, { "epoch": 0.37, "learning_rate": 7.28158028783079e-06, "loss": 0.88, "step": 4321 }, { "epoch": 0.37, "learning_rate": 7.280350568898109e-06, "loss": 0.8386, "step": 4322 }, { "epoch": 0.37, "learning_rate": 7.2791206757808755e-06, "loss": 0.8301, "step": 4323 }, { "epoch": 0.37, "learning_rate": 7.277890608573031e-06, "loss": 0.8952, "step": 4324 }, { "epoch": 0.37, "learning_rate": 7.276660367368537e-06, "loss": 0.8329, "step": 4325 }, { "epoch": 0.37, "learning_rate": 7.275429952261367e-06, "loss": 0.8354, "step": 4326 }, { "epoch": 0.37, "learning_rate": 7.274199363345503e-06, "loss": 0.8103, "step": 4327 }, { "epoch": 0.37, "learning_rate": 7.272968600714945e-06, "loss": 0.8012, "step": 4328 }, { "epoch": 0.37, "learning_rate": 7.2717376644637035e-06, "loss": 0.8674, "step": 4329 }, { "epoch": 0.37, "learning_rate": 7.270506554685805e-06, "loss": 0.8964, "step": 4330 }, { "epoch": 0.37, "learning_rate": 7.269275271475289e-06, "loss": 0.8959, "step": 4331 }, { "epoch": 0.37, "learning_rate": 7.268043814926205e-06, "loss": 0.813, "step": 4332 }, { "epoch": 0.37, "learning_rate": 7.266812185132617e-06, "loss": 0.86, "step": 4333 }, { "epoch": 0.37, "learning_rate": 7.265580382188604e-06, "loss": 0.8877, "step": 4334 }, { "epoch": 0.37, "learning_rate": 7.2643484061882595e-06, "loss": 0.9019, "step": 4335 }, { "epoch": 0.37, "learning_rate": 7.263116257225684e-06, "loss": 0.854, "step": 4336 }, { "epoch": 0.37, "learning_rate": 7.261883935394999e-06, "loss": 0.8318, "step": 4337 }, { "epoch": 0.37, "learning_rate": 7.2606514407903315e-06, "loss": 0.8327, "step": 4338 }, { "epoch": 0.37, "learning_rate": 7.259418773505828e-06, "loss": 0.8702, "step": 4339 }, { "epoch": 0.37, "learning_rate": 7.258185933635649e-06, "loss": 0.8478, "step": 4340 }, { "epoch": 0.37, "learning_rate": 7.256952921273958e-06, "loss": 0.8419, "step": 4341 }, { "epoch": 0.37, "learning_rate": 7.255719736514942e-06, "loss": 0.8346, "step": 4342 }, { "epoch": 0.37, "learning_rate": 7.2544863794528e-06, "loss": 0.8465, "step": 4343 }, { "epoch": 0.37, "learning_rate": 7.253252850181738e-06, "loss": 0.8762, "step": 4344 }, { "epoch": 0.37, "learning_rate": 7.252019148795981e-06, "loss": 0.8656, "step": 4345 }, { "epoch": 0.37, "learning_rate": 7.250785275389766e-06, "loss": 0.8346, "step": 4346 }, { "epoch": 0.37, "learning_rate": 7.249551230057341e-06, "loss": 0.8572, "step": 4347 }, { "epoch": 0.37, "learning_rate": 7.248317012892969e-06, "loss": 0.7976, "step": 4348 }, { "epoch": 0.37, "learning_rate": 7.2470826239909275e-06, "loss": 0.9138, "step": 4349 }, { "epoch": 0.37, "learning_rate": 7.2458480634455034e-06, "loss": 0.8728, "step": 4350 }, { "epoch": 0.37, "learning_rate": 7.244613331350997e-06, "loss": 0.8775, "step": 4351 }, { "epoch": 0.37, "learning_rate": 7.243378427801725e-06, "loss": 0.8249, "step": 4352 }, { "epoch": 0.37, "learning_rate": 7.242143352892017e-06, "loss": 0.829, "step": 4353 }, { "epoch": 0.37, "learning_rate": 7.240908106716214e-06, "loss": 0.8497, "step": 4354 }, { "epoch": 0.37, "learning_rate": 7.239672689368667e-06, "loss": 0.8729, "step": 4355 }, { "epoch": 0.37, "learning_rate": 7.238437100943745e-06, "loss": 0.8817, "step": 4356 }, { "epoch": 0.37, "learning_rate": 7.237201341535829e-06, "loss": 0.8489, "step": 4357 }, { "epoch": 0.37, "learning_rate": 7.235965411239312e-06, "loss": 0.6183, "step": 4358 }, { "epoch": 0.37, "learning_rate": 7.234729310148603e-06, "loss": 0.8435, "step": 4359 }, { "epoch": 0.37, "learning_rate": 7.233493038358118e-06, "loss": 0.9158, "step": 4360 }, { "epoch": 0.37, "learning_rate": 7.23225659596229e-06, "loss": 0.911, "step": 4361 }, { "epoch": 0.37, "learning_rate": 7.231019983055566e-06, "loss": 0.8293, "step": 4362 }, { "epoch": 0.37, "learning_rate": 7.229783199732405e-06, "loss": 0.8593, "step": 4363 }, { "epoch": 0.37, "learning_rate": 7.228546246087279e-06, "loss": 0.8517, "step": 4364 }, { "epoch": 0.37, "learning_rate": 7.2273091222146696e-06, "loss": 0.8215, "step": 4365 }, { "epoch": 0.37, "learning_rate": 7.226071828209078e-06, "loss": 0.8863, "step": 4366 }, { "epoch": 0.37, "learning_rate": 7.224834364165012e-06, "loss": 0.8312, "step": 4367 }, { "epoch": 0.37, "learning_rate": 7.223596730176998e-06, "loss": 0.8754, "step": 4368 }, { "epoch": 0.37, "learning_rate": 7.2223589263395695e-06, "loss": 0.8477, "step": 4369 }, { "epoch": 0.37, "learning_rate": 7.22112095274728e-06, "loss": 0.8954, "step": 4370 }, { "epoch": 0.37, "learning_rate": 7.219882809494689e-06, "loss": 0.8986, "step": 4371 }, { "epoch": 0.37, "learning_rate": 7.218644496676375e-06, "loss": 0.8569, "step": 4372 }, { "epoch": 0.37, "learning_rate": 7.217406014386923e-06, "loss": 0.8323, "step": 4373 }, { "epoch": 0.37, "learning_rate": 7.2161673627209364e-06, "loss": 0.8558, "step": 4374 }, { "epoch": 0.37, "learning_rate": 7.214928541773027e-06, "loss": 0.8314, "step": 4375 }, { "epoch": 0.37, "learning_rate": 7.213689551637828e-06, "loss": 0.8273, "step": 4376 }, { "epoch": 0.37, "learning_rate": 7.2124503924099746e-06, "loss": 0.8483, "step": 4377 }, { "epoch": 0.37, "learning_rate": 7.211211064184121e-06, "loss": 0.8622, "step": 4378 }, { "epoch": 0.37, "learning_rate": 7.209971567054934e-06, "loss": 0.817, "step": 4379 }, { "epoch": 0.37, "learning_rate": 7.208731901117094e-06, "loss": 0.87, "step": 4380 }, { "epoch": 0.37, "learning_rate": 7.2074920664652894e-06, "loss": 0.8367, "step": 4381 }, { "epoch": 0.37, "learning_rate": 7.2062520631942266e-06, "loss": 0.8439, "step": 4382 }, { "epoch": 0.37, "learning_rate": 7.205011891398625e-06, "loss": 0.877, "step": 4383 }, { "epoch": 0.37, "learning_rate": 7.203771551173212e-06, "loss": 0.8659, "step": 4384 }, { "epoch": 0.37, "learning_rate": 7.2025310426127324e-06, "loss": 0.869, "step": 4385 }, { "epoch": 0.37, "learning_rate": 7.2012903658119426e-06, "loss": 0.8903, "step": 4386 }, { "epoch": 0.37, "learning_rate": 7.200049520865613e-06, "loss": 0.6214, "step": 4387 }, { "epoch": 0.37, "learning_rate": 7.198808507868522e-06, "loss": 0.8554, "step": 4388 }, { "epoch": 0.37, "learning_rate": 7.197567326915467e-06, "loss": 0.9123, "step": 4389 }, { "epoch": 0.37, "learning_rate": 7.196325978101256e-06, "loss": 0.883, "step": 4390 }, { "epoch": 0.37, "learning_rate": 7.195084461520709e-06, "loss": 0.8209, "step": 4391 }, { "epoch": 0.37, "learning_rate": 7.193842777268656e-06, "loss": 0.8203, "step": 4392 }, { "epoch": 0.37, "learning_rate": 7.1926009254399475e-06, "loss": 0.8733, "step": 4393 }, { "epoch": 0.37, "learning_rate": 7.191358906129441e-06, "loss": 0.9253, "step": 4394 }, { "epoch": 0.38, "learning_rate": 7.1901167194320065e-06, "loss": 0.8706, "step": 4395 }, { "epoch": 0.38, "learning_rate": 7.188874365442531e-06, "loss": 0.8745, "step": 4396 }, { "epoch": 0.38, "learning_rate": 7.187631844255909e-06, "loss": 0.8141, "step": 4397 }, { "epoch": 0.38, "learning_rate": 7.1863891559670525e-06, "loss": 0.8966, "step": 4398 }, { "epoch": 0.38, "learning_rate": 7.185146300670882e-06, "loss": 0.8333, "step": 4399 }, { "epoch": 0.38, "learning_rate": 7.183903278462338e-06, "loss": 0.8418, "step": 4400 }, { "epoch": 0.38, "learning_rate": 7.182660089436363e-06, "loss": 0.8572, "step": 4401 }, { "epoch": 0.38, "learning_rate": 7.1814167336879195e-06, "loss": 0.8469, "step": 4402 }, { "epoch": 0.38, "learning_rate": 7.180173211311984e-06, "loss": 0.8912, "step": 4403 }, { "epoch": 0.38, "learning_rate": 7.1789295224035385e-06, "loss": 0.9081, "step": 4404 }, { "epoch": 0.38, "learning_rate": 7.177685667057585e-06, "loss": 0.9297, "step": 4405 }, { "epoch": 0.38, "learning_rate": 7.176441645369138e-06, "loss": 0.8119, "step": 4406 }, { "epoch": 0.38, "learning_rate": 7.175197457433216e-06, "loss": 0.8112, "step": 4407 }, { "epoch": 0.38, "learning_rate": 7.173953103344859e-06, "loss": 0.8334, "step": 4408 }, { "epoch": 0.38, "learning_rate": 7.172708583199117e-06, "loss": 0.8379, "step": 4409 }, { "epoch": 0.38, "learning_rate": 7.1714638970910565e-06, "loss": 0.8106, "step": 4410 }, { "epoch": 0.38, "learning_rate": 7.170219045115746e-06, "loss": 0.859, "step": 4411 }, { "epoch": 0.38, "learning_rate": 7.1689740273682774e-06, "loss": 0.8504, "step": 4412 }, { "epoch": 0.38, "learning_rate": 7.167728843943752e-06, "loss": 0.8343, "step": 4413 }, { "epoch": 0.38, "learning_rate": 7.16648349493728e-06, "loss": 0.8438, "step": 4414 }, { "epoch": 0.38, "learning_rate": 7.165237980443991e-06, "loss": 0.8803, "step": 4415 }, { "epoch": 0.38, "learning_rate": 7.163992300559021e-06, "loss": 0.8224, "step": 4416 }, { "epoch": 0.38, "learning_rate": 7.1627464553775214e-06, "loss": 0.8704, "step": 4417 }, { "epoch": 0.38, "learning_rate": 7.161500444994657e-06, "loss": 0.8371, "step": 4418 }, { "epoch": 0.38, "learning_rate": 7.160254269505606e-06, "loss": 0.8399, "step": 4419 }, { "epoch": 0.38, "learning_rate": 7.159007929005554e-06, "loss": 0.8881, "step": 4420 }, { "epoch": 0.38, "learning_rate": 7.157761423589703e-06, "loss": 0.8284, "step": 4421 }, { "epoch": 0.38, "learning_rate": 7.15651475335327e-06, "loss": 0.8386, "step": 4422 }, { "epoch": 0.38, "learning_rate": 7.15526791839148e-06, "loss": 0.862, "step": 4423 }, { "epoch": 0.38, "learning_rate": 7.154020918799573e-06, "loss": 0.8915, "step": 4424 }, { "epoch": 0.38, "learning_rate": 7.1527737546728015e-06, "loss": 0.7983, "step": 4425 }, { "epoch": 0.38, "learning_rate": 7.15152642610643e-06, "loss": 0.8843, "step": 4426 }, { "epoch": 0.38, "learning_rate": 7.150278933195734e-06, "loss": 0.8394, "step": 4427 }, { "epoch": 0.38, "learning_rate": 7.149031276036005e-06, "loss": 0.8693, "step": 4428 }, { "epoch": 0.38, "learning_rate": 7.147783454722545e-06, "loss": 0.834, "step": 4429 }, { "epoch": 0.38, "learning_rate": 7.146535469350668e-06, "loss": 0.8922, "step": 4430 }, { "epoch": 0.38, "learning_rate": 7.145287320015702e-06, "loss": 0.8767, "step": 4431 }, { "epoch": 0.38, "learning_rate": 7.144039006812987e-06, "loss": 0.8946, "step": 4432 }, { "epoch": 0.38, "learning_rate": 7.142790529837875e-06, "loss": 0.9136, "step": 4433 }, { "epoch": 0.38, "learning_rate": 7.141541889185732e-06, "loss": 0.874, "step": 4434 }, { "epoch": 0.38, "learning_rate": 7.1402930849519345e-06, "loss": 0.8252, "step": 4435 }, { "epoch": 0.38, "learning_rate": 7.139044117231872e-06, "loss": 0.828, "step": 4436 }, { "epoch": 0.38, "learning_rate": 7.137794986120947e-06, "loss": 0.886, "step": 4437 }, { "epoch": 0.38, "learning_rate": 7.136545691714575e-06, "loss": 0.8014, "step": 4438 }, { "epoch": 0.38, "learning_rate": 7.135296234108184e-06, "loss": 0.8633, "step": 4439 }, { "epoch": 0.38, "learning_rate": 7.134046613397211e-06, "loss": 0.856, "step": 4440 }, { "epoch": 0.38, "learning_rate": 7.13279682967711e-06, "loss": 0.8554, "step": 4441 }, { "epoch": 0.38, "learning_rate": 7.131546883043348e-06, "loss": 0.851, "step": 4442 }, { "epoch": 0.38, "learning_rate": 7.130296773591399e-06, "loss": 0.8218, "step": 4443 }, { "epoch": 0.38, "learning_rate": 7.129046501416752e-06, "loss": 0.8834, "step": 4444 }, { "epoch": 0.38, "learning_rate": 7.127796066614912e-06, "loss": 0.8541, "step": 4445 }, { "epoch": 0.38, "learning_rate": 7.126545469281392e-06, "loss": 0.884, "step": 4446 }, { "epoch": 0.38, "learning_rate": 7.1252947095117176e-06, "loss": 0.8881, "step": 4447 }, { "epoch": 0.38, "learning_rate": 7.12404378740143e-06, "loss": 0.8372, "step": 4448 }, { "epoch": 0.38, "learning_rate": 7.122792703046082e-06, "loss": 0.8678, "step": 4449 }, { "epoch": 0.38, "learning_rate": 7.121541456541234e-06, "loss": 0.8942, "step": 4450 }, { "epoch": 0.38, "learning_rate": 7.120290047982464e-06, "loss": 0.8959, "step": 4451 }, { "epoch": 0.38, "learning_rate": 7.119038477465363e-06, "loss": 0.84, "step": 4452 }, { "epoch": 0.38, "learning_rate": 7.117786745085529e-06, "loss": 0.8526, "step": 4453 }, { "epoch": 0.38, "learning_rate": 7.116534850938576e-06, "loss": 0.8473, "step": 4454 }, { "epoch": 0.38, "learning_rate": 7.115282795120133e-06, "loss": 0.8604, "step": 4455 }, { "epoch": 0.38, "learning_rate": 7.1140305777258355e-06, "loss": 0.874, "step": 4456 }, { "epoch": 0.38, "learning_rate": 7.112778198851336e-06, "loss": 0.8579, "step": 4457 }, { "epoch": 0.38, "learning_rate": 7.111525658592293e-06, "loss": 0.8334, "step": 4458 }, { "epoch": 0.38, "learning_rate": 7.110272957044388e-06, "loss": 0.8938, "step": 4459 }, { "epoch": 0.38, "learning_rate": 7.109020094303304e-06, "loss": 0.8405, "step": 4460 }, { "epoch": 0.38, "learning_rate": 7.107767070464744e-06, "loss": 0.8565, "step": 4461 }, { "epoch": 0.38, "learning_rate": 7.1065138856244165e-06, "loss": 0.8816, "step": 4462 }, { "epoch": 0.38, "learning_rate": 7.105260539878049e-06, "loss": 0.9079, "step": 4463 }, { "epoch": 0.38, "learning_rate": 7.104007033321378e-06, "loss": 0.8718, "step": 4464 }, { "epoch": 0.38, "learning_rate": 7.1027533660501516e-06, "loss": 0.865, "step": 4465 }, { "epoch": 0.38, "learning_rate": 7.101499538160131e-06, "loss": 0.8858, "step": 4466 }, { "epoch": 0.38, "learning_rate": 7.1002455497470926e-06, "loss": 0.8298, "step": 4467 }, { "epoch": 0.38, "learning_rate": 7.0989914009068185e-06, "loss": 0.8272, "step": 4468 }, { "epoch": 0.38, "learning_rate": 7.097737091735109e-06, "loss": 0.8112, "step": 4469 }, { "epoch": 0.38, "learning_rate": 7.096482622327774e-06, "loss": 0.8567, "step": 4470 }, { "epoch": 0.38, "learning_rate": 7.095227992780637e-06, "loss": 0.8879, "step": 4471 }, { "epoch": 0.38, "learning_rate": 7.09397320318953e-06, "loss": 0.8756, "step": 4472 }, { "epoch": 0.38, "learning_rate": 7.0927182536503035e-06, "loss": 0.8387, "step": 4473 }, { "epoch": 0.38, "learning_rate": 7.091463144258814e-06, "loss": 0.8683, "step": 4474 }, { "epoch": 0.38, "learning_rate": 7.090207875110937e-06, "loss": 0.8402, "step": 4475 }, { "epoch": 0.38, "learning_rate": 7.0889524463025524e-06, "loss": 0.8402, "step": 4476 }, { "epoch": 0.38, "learning_rate": 7.087696857929556e-06, "loss": 0.8759, "step": 4477 }, { "epoch": 0.38, "learning_rate": 7.086441110087858e-06, "loss": 0.8316, "step": 4478 }, { "epoch": 0.38, "learning_rate": 7.085185202873379e-06, "loss": 0.6187, "step": 4479 }, { "epoch": 0.38, "learning_rate": 7.083929136382051e-06, "loss": 0.8672, "step": 4480 }, { "epoch": 0.38, "learning_rate": 7.0826729107098155e-06, "loss": 0.8637, "step": 4481 }, { "epoch": 0.38, "learning_rate": 7.081416525952634e-06, "loss": 0.8556, "step": 4482 }, { "epoch": 0.38, "learning_rate": 7.080159982206471e-06, "loss": 0.8535, "step": 4483 }, { "epoch": 0.38, "learning_rate": 7.07890327956731e-06, "loss": 0.8557, "step": 4484 }, { "epoch": 0.38, "learning_rate": 7.077646418131146e-06, "loss": 0.8608, "step": 4485 }, { "epoch": 0.38, "learning_rate": 7.076389397993982e-06, "loss": 0.8377, "step": 4486 }, { "epoch": 0.38, "learning_rate": 7.075132219251834e-06, "loss": 0.8477, "step": 4487 }, { "epoch": 0.38, "learning_rate": 7.0738748820007355e-06, "loss": 0.8572, "step": 4488 }, { "epoch": 0.38, "learning_rate": 7.072617386336726e-06, "loss": 0.8494, "step": 4489 }, { "epoch": 0.38, "learning_rate": 7.0713597323558595e-06, "loss": 0.8474, "step": 4490 }, { "epoch": 0.38, "learning_rate": 7.070101920154201e-06, "loss": 0.811, "step": 4491 }, { "epoch": 0.38, "learning_rate": 7.068843949827831e-06, "loss": 0.8421, "step": 4492 }, { "epoch": 0.38, "learning_rate": 7.067585821472836e-06, "loss": 0.8488, "step": 4493 }, { "epoch": 0.38, "learning_rate": 7.06632753518532e-06, "loss": 0.8337, "step": 4494 }, { "epoch": 0.38, "learning_rate": 7.0650690910613965e-06, "loss": 0.8458, "step": 4495 }, { "epoch": 0.38, "learning_rate": 7.063810489197194e-06, "loss": 0.8678, "step": 4496 }, { "epoch": 0.38, "learning_rate": 7.062551729688848e-06, "loss": 0.8546, "step": 4497 }, { "epoch": 0.38, "learning_rate": 7.061292812632509e-06, "loss": 0.8349, "step": 4498 }, { "epoch": 0.38, "learning_rate": 7.060033738124345e-06, "loss": 0.8089, "step": 4499 }, { "epoch": 0.38, "learning_rate": 7.058774506260522e-06, "loss": 0.8141, "step": 4500 }, { "epoch": 0.38, "learning_rate": 7.057515117137231e-06, "loss": 0.8752, "step": 4501 }, { "epoch": 0.38, "learning_rate": 7.056255570850669e-06, "loss": 0.8149, "step": 4502 }, { "epoch": 0.38, "learning_rate": 7.054995867497048e-06, "loss": 0.769, "step": 4503 }, { "epoch": 0.38, "learning_rate": 7.0537360071725895e-06, "loss": 0.8992, "step": 4504 }, { "epoch": 0.38, "learning_rate": 7.052475989973526e-06, "loss": 0.8395, "step": 4505 }, { "epoch": 0.38, "learning_rate": 7.051215815996109e-06, "loss": 0.8788, "step": 4506 }, { "epoch": 0.38, "learning_rate": 7.04995548533659e-06, "loss": 0.8509, "step": 4507 }, { "epoch": 0.38, "learning_rate": 7.048694998091246e-06, "loss": 0.8254, "step": 4508 }, { "epoch": 0.38, "learning_rate": 7.047434354356355e-06, "loss": 0.8668, "step": 4509 }, { "epoch": 0.38, "learning_rate": 7.046173554228213e-06, "loss": 0.8513, "step": 4510 }, { "epoch": 0.38, "learning_rate": 7.044912597803125e-06, "loss": 0.8527, "step": 4511 }, { "epoch": 0.39, "learning_rate": 7.04365148517741e-06, "loss": 0.605, "step": 4512 }, { "epoch": 0.39, "learning_rate": 7.0423902164474e-06, "loss": 0.9097, "step": 4513 }, { "epoch": 0.39, "learning_rate": 7.041128791709433e-06, "loss": 0.8713, "step": 4514 }, { "epoch": 0.39, "learning_rate": 7.039867211059864e-06, "loss": 0.8915, "step": 4515 }, { "epoch": 0.39, "learning_rate": 7.038605474595062e-06, "loss": 0.8365, "step": 4516 }, { "epoch": 0.39, "learning_rate": 7.037343582411401e-06, "loss": 0.7893, "step": 4517 }, { "epoch": 0.39, "learning_rate": 7.036081534605273e-06, "loss": 0.8502, "step": 4518 }, { "epoch": 0.39, "learning_rate": 7.034819331273079e-06, "loss": 0.6253, "step": 4519 }, { "epoch": 0.39, "learning_rate": 7.033556972511231e-06, "loss": 0.8606, "step": 4520 }, { "epoch": 0.39, "learning_rate": 7.0322944584161545e-06, "loss": 0.8786, "step": 4521 }, { "epoch": 0.39, "learning_rate": 7.031031789084288e-06, "loss": 0.8845, "step": 4522 }, { "epoch": 0.39, "learning_rate": 7.029768964612082e-06, "loss": 0.8495, "step": 4523 }, { "epoch": 0.39, "learning_rate": 7.028505985095992e-06, "loss": 0.8071, "step": 4524 }, { "epoch": 0.39, "learning_rate": 7.027242850632494e-06, "loss": 0.6472, "step": 4525 }, { "epoch": 0.39, "learning_rate": 7.0259795613180744e-06, "loss": 0.8614, "step": 4526 }, { "epoch": 0.39, "learning_rate": 7.024716117249227e-06, "loss": 0.8656, "step": 4527 }, { "epoch": 0.39, "learning_rate": 7.02345251852246e-06, "loss": 0.8711, "step": 4528 }, { "epoch": 0.39, "learning_rate": 7.0221887652342945e-06, "loss": 0.8244, "step": 4529 }, { "epoch": 0.39, "learning_rate": 7.02092485748126e-06, "loss": 0.8881, "step": 4530 }, { "epoch": 0.39, "learning_rate": 7.0196607953599035e-06, "loss": 0.8357, "step": 4531 }, { "epoch": 0.39, "learning_rate": 7.0183965789667785e-06, "loss": 0.8461, "step": 4532 }, { "epoch": 0.39, "learning_rate": 7.01713220839845e-06, "loss": 0.8515, "step": 4533 }, { "epoch": 0.39, "learning_rate": 7.0158676837515e-06, "loss": 0.7813, "step": 4534 }, { "epoch": 0.39, "learning_rate": 7.01460300512252e-06, "loss": 0.827, "step": 4535 }, { "epoch": 0.39, "learning_rate": 7.013338172608111e-06, "loss": 0.8379, "step": 4536 }, { "epoch": 0.39, "learning_rate": 7.012073186304885e-06, "loss": 0.8334, "step": 4537 }, { "epoch": 0.39, "learning_rate": 7.010808046309471e-06, "loss": 0.813, "step": 4538 }, { "epoch": 0.39, "learning_rate": 7.0095427527185055e-06, "loss": 0.5776, "step": 4539 }, { "epoch": 0.39, "learning_rate": 7.008277305628638e-06, "loss": 0.8704, "step": 4540 }, { "epoch": 0.39, "learning_rate": 7.00701170513653e-06, "loss": 0.8062, "step": 4541 }, { "epoch": 0.39, "learning_rate": 7.0057459513388556e-06, "loss": 0.8716, "step": 4542 }, { "epoch": 0.39, "learning_rate": 7.004480044332296e-06, "loss": 0.8805, "step": 4543 }, { "epoch": 0.39, "learning_rate": 7.003213984213551e-06, "loss": 0.849, "step": 4544 }, { "epoch": 0.39, "learning_rate": 7.001947771079326e-06, "loss": 0.8617, "step": 4545 }, { "epoch": 0.39, "learning_rate": 7.000681405026342e-06, "loss": 0.8411, "step": 4546 }, { "epoch": 0.39, "learning_rate": 6.99941488615133e-06, "loss": 0.8532, "step": 4547 }, { "epoch": 0.39, "learning_rate": 6.998148214551033e-06, "loss": 0.8556, "step": 4548 }, { "epoch": 0.39, "learning_rate": 6.996881390322207e-06, "loss": 0.8737, "step": 4549 }, { "epoch": 0.39, "learning_rate": 6.995614413561615e-06, "loss": 0.8702, "step": 4550 }, { "epoch": 0.39, "learning_rate": 6.994347284366039e-06, "loss": 0.8736, "step": 4551 }, { "epoch": 0.39, "learning_rate": 6.993080002832267e-06, "loss": 0.8674, "step": 4552 }, { "epoch": 0.39, "learning_rate": 6.991812569057099e-06, "loss": 0.5537, "step": 4553 }, { "epoch": 0.39, "learning_rate": 6.9905449831373486e-06, "loss": 0.8661, "step": 4554 }, { "epoch": 0.39, "learning_rate": 6.989277245169843e-06, "loss": 0.8746, "step": 4555 }, { "epoch": 0.39, "learning_rate": 6.988009355251415e-06, "loss": 0.8508, "step": 4556 }, { "epoch": 0.39, "learning_rate": 6.986741313478913e-06, "loss": 0.8717, "step": 4557 }, { "epoch": 0.39, "learning_rate": 6.985473119949197e-06, "loss": 0.8104, "step": 4558 }, { "epoch": 0.39, "learning_rate": 6.984204774759138e-06, "loss": 0.8719, "step": 4559 }, { "epoch": 0.39, "learning_rate": 6.9829362780056186e-06, "loss": 0.8633, "step": 4560 }, { "epoch": 0.39, "learning_rate": 6.981667629785531e-06, "loss": 0.5243, "step": 4561 }, { "epoch": 0.39, "learning_rate": 6.980398830195785e-06, "loss": 0.8853, "step": 4562 }, { "epoch": 0.39, "learning_rate": 6.979129879333294e-06, "loss": 0.8102, "step": 4563 }, { "epoch": 0.39, "learning_rate": 6.9778607772949894e-06, "loss": 0.8302, "step": 4564 }, { "epoch": 0.39, "learning_rate": 6.97659152417781e-06, "loss": 0.8661, "step": 4565 }, { "epoch": 0.39, "learning_rate": 6.975322120078706e-06, "loss": 0.8291, "step": 4566 }, { "epoch": 0.39, "learning_rate": 6.974052565094644e-06, "loss": 0.84, "step": 4567 }, { "epoch": 0.39, "learning_rate": 6.972782859322598e-06, "loss": 0.8741, "step": 4568 }, { "epoch": 0.39, "learning_rate": 6.971513002859554e-06, "loss": 0.8785, "step": 4569 }, { "epoch": 0.39, "learning_rate": 6.970242995802511e-06, "loss": 0.8401, "step": 4570 }, { "epoch": 0.39, "learning_rate": 6.968972838248476e-06, "loss": 0.8797, "step": 4571 }, { "epoch": 0.39, "learning_rate": 6.967702530294473e-06, "loss": 0.8401, "step": 4572 }, { "epoch": 0.39, "learning_rate": 6.966432072037532e-06, "loss": 0.8599, "step": 4573 }, { "epoch": 0.39, "learning_rate": 6.9651614635747e-06, "loss": 0.8394, "step": 4574 }, { "epoch": 0.39, "learning_rate": 6.963890705003029e-06, "loss": 0.8756, "step": 4575 }, { "epoch": 0.39, "learning_rate": 6.962619796419587e-06, "loss": 0.8486, "step": 4576 }, { "epoch": 0.39, "learning_rate": 6.961348737921454e-06, "loss": 0.8388, "step": 4577 }, { "epoch": 0.39, "learning_rate": 6.960077529605718e-06, "loss": 0.8664, "step": 4578 }, { "epoch": 0.39, "learning_rate": 6.958806171569481e-06, "loss": 0.8367, "step": 4579 }, { "epoch": 0.39, "learning_rate": 6.957534663909854e-06, "loss": 0.9056, "step": 4580 }, { "epoch": 0.39, "learning_rate": 6.956263006723962e-06, "loss": 0.8549, "step": 4581 }, { "epoch": 0.39, "learning_rate": 6.954991200108943e-06, "loss": 0.8498, "step": 4582 }, { "epoch": 0.39, "learning_rate": 6.953719244161942e-06, "loss": 0.84, "step": 4583 }, { "epoch": 0.39, "learning_rate": 6.9524471389801165e-06, "loss": 0.8709, "step": 4584 }, { "epoch": 0.39, "learning_rate": 6.951174884660636e-06, "loss": 0.8771, "step": 4585 }, { "epoch": 0.39, "learning_rate": 6.949902481300683e-06, "loss": 0.824, "step": 4586 }, { "epoch": 0.39, "learning_rate": 6.94862992899745e-06, "loss": 0.8324, "step": 4587 }, { "epoch": 0.39, "learning_rate": 6.947357227848141e-06, "loss": 0.5568, "step": 4588 }, { "epoch": 0.39, "learning_rate": 6.946084377949972e-06, "loss": 0.8446, "step": 4589 }, { "epoch": 0.39, "learning_rate": 6.944811379400167e-06, "loss": 0.8936, "step": 4590 }, { "epoch": 0.39, "learning_rate": 6.943538232295965e-06, "loss": 0.8413, "step": 4591 }, { "epoch": 0.39, "learning_rate": 6.942264936734616e-06, "loss": 0.841, "step": 4592 }, { "epoch": 0.39, "learning_rate": 6.940991492813381e-06, "loss": 0.8684, "step": 4593 }, { "epoch": 0.39, "learning_rate": 6.93971790062953e-06, "loss": 0.844, "step": 4594 }, { "epoch": 0.39, "learning_rate": 6.93844416028035e-06, "loss": 0.8613, "step": 4595 }, { "epoch": 0.39, "learning_rate": 6.937170271863131e-06, "loss": 0.8369, "step": 4596 }, { "epoch": 0.39, "learning_rate": 6.935896235475184e-06, "loss": 0.8807, "step": 4597 }, { "epoch": 0.39, "learning_rate": 6.934622051213822e-06, "loss": 0.8364, "step": 4598 }, { "epoch": 0.39, "learning_rate": 6.933347719176375e-06, "loss": 0.8165, "step": 4599 }, { "epoch": 0.39, "learning_rate": 6.932073239460183e-06, "loss": 0.9098, "step": 4600 }, { "epoch": 0.39, "learning_rate": 6.930798612162597e-06, "loss": 0.8794, "step": 4601 }, { "epoch": 0.39, "learning_rate": 6.9295238373809805e-06, "loss": 0.8247, "step": 4602 }, { "epoch": 0.39, "learning_rate": 6.928248915212705e-06, "loss": 0.8661, "step": 4603 }, { "epoch": 0.39, "learning_rate": 6.926973845755157e-06, "loss": 0.8677, "step": 4604 }, { "epoch": 0.39, "learning_rate": 6.925698629105733e-06, "loss": 0.8778, "step": 4605 }, { "epoch": 0.39, "learning_rate": 6.924423265361839e-06, "loss": 0.872, "step": 4606 }, { "epoch": 0.39, "learning_rate": 6.923147754620894e-06, "loss": 0.8439, "step": 4607 }, { "epoch": 0.39, "learning_rate": 6.921872096980329e-06, "loss": 0.8995, "step": 4608 }, { "epoch": 0.39, "learning_rate": 6.920596292537582e-06, "loss": 0.8279, "step": 4609 }, { "epoch": 0.39, "learning_rate": 6.91932034139011e-06, "loss": 0.8285, "step": 4610 }, { "epoch": 0.39, "learning_rate": 6.918044243635373e-06, "loss": 0.8672, "step": 4611 }, { "epoch": 0.39, "learning_rate": 6.916767999370848e-06, "loss": 0.8666, "step": 4612 }, { "epoch": 0.39, "learning_rate": 6.915491608694018e-06, "loss": 0.5722, "step": 4613 }, { "epoch": 0.39, "learning_rate": 6.9142150717023825e-06, "loss": 0.8343, "step": 4614 }, { "epoch": 0.39, "learning_rate": 6.9129383884934495e-06, "loss": 0.8251, "step": 4615 }, { "epoch": 0.39, "learning_rate": 6.911661559164739e-06, "loss": 0.8543, "step": 4616 }, { "epoch": 0.39, "learning_rate": 6.910384583813778e-06, "loss": 0.825, "step": 4617 }, { "epoch": 0.39, "learning_rate": 6.909107462538113e-06, "loss": 0.791, "step": 4618 }, { "epoch": 0.39, "learning_rate": 6.907830195435293e-06, "loss": 0.8381, "step": 4619 }, { "epoch": 0.39, "learning_rate": 6.906552782602885e-06, "loss": 0.8685, "step": 4620 }, { "epoch": 0.39, "learning_rate": 6.9052752241384625e-06, "loss": 0.9128, "step": 4621 }, { "epoch": 0.39, "learning_rate": 6.903997520139613e-06, "loss": 0.8444, "step": 4622 }, { "epoch": 0.39, "learning_rate": 6.902719670703932e-06, "loss": 0.8347, "step": 4623 }, { "epoch": 0.39, "learning_rate": 6.9014416759290285e-06, "loss": 0.8686, "step": 4624 }, { "epoch": 0.39, "learning_rate": 6.9001635359125255e-06, "loss": 0.8835, "step": 4625 }, { "epoch": 0.39, "learning_rate": 6.898885250752049e-06, "loss": 0.8707, "step": 4626 }, { "epoch": 0.39, "learning_rate": 6.8976068205452415e-06, "loss": 0.8391, "step": 4627 }, { "epoch": 0.39, "learning_rate": 6.89632824538976e-06, "loss": 0.8401, "step": 4628 }, { "epoch": 0.39, "learning_rate": 6.895049525383264e-06, "loss": 0.8305, "step": 4629 }, { "epoch": 0.4, "learning_rate": 6.893770660623431e-06, "loss": 0.8713, "step": 4630 }, { "epoch": 0.4, "learning_rate": 6.892491651207944e-06, "loss": 0.853, "step": 4631 }, { "epoch": 0.4, "learning_rate": 6.891212497234505e-06, "loss": 0.854, "step": 4632 }, { "epoch": 0.4, "learning_rate": 6.889933198800817e-06, "loss": 0.8184, "step": 4633 }, { "epoch": 0.4, "learning_rate": 6.888653756004604e-06, "loss": 0.863, "step": 4634 }, { "epoch": 0.4, "learning_rate": 6.887374168943594e-06, "loss": 0.8448, "step": 4635 }, { "epoch": 0.4, "learning_rate": 6.886094437715527e-06, "loss": 0.877, "step": 4636 }, { "epoch": 0.4, "learning_rate": 6.884814562418155e-06, "loss": 0.8922, "step": 4637 }, { "epoch": 0.4, "learning_rate": 6.883534543149245e-06, "loss": 0.9165, "step": 4638 }, { "epoch": 0.4, "learning_rate": 6.882254380006568e-06, "loss": 0.8461, "step": 4639 }, { "epoch": 0.4, "learning_rate": 6.8809740730879095e-06, "loss": 0.8067, "step": 4640 }, { "epoch": 0.4, "learning_rate": 6.879693622491067e-06, "loss": 0.8406, "step": 4641 }, { "epoch": 0.4, "learning_rate": 6.878413028313847e-06, "loss": 0.8674, "step": 4642 }, { "epoch": 0.4, "learning_rate": 6.877132290654067e-06, "loss": 0.8531, "step": 4643 }, { "epoch": 0.4, "learning_rate": 6.8758514096095596e-06, "loss": 0.8852, "step": 4644 }, { "epoch": 0.4, "learning_rate": 6.874570385278161e-06, "loss": 0.8408, "step": 4645 }, { "epoch": 0.4, "learning_rate": 6.873289217757722e-06, "loss": 0.8396, "step": 4646 }, { "epoch": 0.4, "learning_rate": 6.872007907146105e-06, "loss": 0.8748, "step": 4647 }, { "epoch": 0.4, "learning_rate": 6.8707264535411866e-06, "loss": 0.8411, "step": 4648 }, { "epoch": 0.4, "learning_rate": 6.8694448570408475e-06, "loss": 0.8603, "step": 4649 }, { "epoch": 0.4, "learning_rate": 6.868163117742982e-06, "loss": 0.9029, "step": 4650 }, { "epoch": 0.4, "learning_rate": 6.866881235745496e-06, "loss": 0.8223, "step": 4651 }, { "epoch": 0.4, "learning_rate": 6.865599211146308e-06, "loss": 0.869, "step": 4652 }, { "epoch": 0.4, "learning_rate": 6.864317044043344e-06, "loss": 0.8838, "step": 4653 }, { "epoch": 0.4, "learning_rate": 6.8630347345345415e-06, "loss": 0.8523, "step": 4654 }, { "epoch": 0.4, "learning_rate": 6.8617522827178515e-06, "loss": 0.8281, "step": 4655 }, { "epoch": 0.4, "learning_rate": 6.860469688691231e-06, "loss": 0.8428, "step": 4656 }, { "epoch": 0.4, "learning_rate": 6.859186952552656e-06, "loss": 0.8514, "step": 4657 }, { "epoch": 0.4, "learning_rate": 6.857904074400105e-06, "loss": 0.847, "step": 4658 }, { "epoch": 0.4, "learning_rate": 6.856621054331569e-06, "loss": 0.8686, "step": 4659 }, { "epoch": 0.4, "learning_rate": 6.855337892445054e-06, "loss": 0.868, "step": 4660 }, { "epoch": 0.4, "learning_rate": 6.854054588838574e-06, "loss": 0.8514, "step": 4661 }, { "epoch": 0.4, "learning_rate": 6.852771143610154e-06, "loss": 0.8794, "step": 4662 }, { "epoch": 0.4, "learning_rate": 6.851487556857831e-06, "loss": 0.8465, "step": 4663 }, { "epoch": 0.4, "learning_rate": 6.850203828679649e-06, "loss": 0.856, "step": 4664 }, { "epoch": 0.4, "learning_rate": 6.848919959173668e-06, "loss": 0.8673, "step": 4665 }, { "epoch": 0.4, "learning_rate": 6.847635948437955e-06, "loss": 0.9056, "step": 4666 }, { "epoch": 0.4, "learning_rate": 6.8463517965705905e-06, "loss": 0.8138, "step": 4667 }, { "epoch": 0.4, "learning_rate": 6.845067503669664e-06, "loss": 0.8426, "step": 4668 }, { "epoch": 0.4, "learning_rate": 6.8437830698332745e-06, "loss": 0.8263, "step": 4669 }, { "epoch": 0.4, "learning_rate": 6.842498495159534e-06, "loss": 0.8782, "step": 4670 }, { "epoch": 0.4, "learning_rate": 6.841213779746567e-06, "loss": 0.8429, "step": 4671 }, { "epoch": 0.4, "learning_rate": 6.839928923692505e-06, "loss": 0.8421, "step": 4672 }, { "epoch": 0.4, "learning_rate": 6.838643927095492e-06, "loss": 0.8345, "step": 4673 }, { "epoch": 0.4, "learning_rate": 6.8373587900536805e-06, "loss": 0.8942, "step": 4674 }, { "epoch": 0.4, "learning_rate": 6.836073512665239e-06, "loss": 0.8728, "step": 4675 }, { "epoch": 0.4, "learning_rate": 6.834788095028341e-06, "loss": 0.9379, "step": 4676 }, { "epoch": 0.4, "learning_rate": 6.833502537241173e-06, "loss": 0.8507, "step": 4677 }, { "epoch": 0.4, "learning_rate": 6.832216839401935e-06, "loss": 0.8643, "step": 4678 }, { "epoch": 0.4, "learning_rate": 6.830931001608831e-06, "loss": 0.8466, "step": 4679 }, { "epoch": 0.4, "learning_rate": 6.829645023960083e-06, "loss": 0.8848, "step": 4680 }, { "epoch": 0.4, "learning_rate": 6.82835890655392e-06, "loss": 0.8372, "step": 4681 }, { "epoch": 0.4, "learning_rate": 6.827072649488581e-06, "loss": 0.8726, "step": 4682 }, { "epoch": 0.4, "learning_rate": 6.8257862528623164e-06, "loss": 0.8192, "step": 4683 }, { "epoch": 0.4, "learning_rate": 6.8244997167733875e-06, "loss": 0.8385, "step": 4684 }, { "epoch": 0.4, "learning_rate": 6.8232130413200695e-06, "loss": 0.8444, "step": 4685 }, { "epoch": 0.4, "learning_rate": 6.821926226600641e-06, "loss": 0.6194, "step": 4686 }, { "epoch": 0.4, "learning_rate": 6.820639272713396e-06, "loss": 0.8425, "step": 4687 }, { "epoch": 0.4, "learning_rate": 6.819352179756642e-06, "loss": 0.8209, "step": 4688 }, { "epoch": 0.4, "learning_rate": 6.818064947828689e-06, "loss": 0.8539, "step": 4689 }, { "epoch": 0.4, "learning_rate": 6.8167775770278665e-06, "loss": 0.8608, "step": 4690 }, { "epoch": 0.4, "learning_rate": 6.815490067452506e-06, "loss": 0.812, "step": 4691 }, { "epoch": 0.4, "learning_rate": 6.814202419200957e-06, "loss": 0.846, "step": 4692 }, { "epoch": 0.4, "learning_rate": 6.812914632371575e-06, "loss": 0.8001, "step": 4693 }, { "epoch": 0.4, "learning_rate": 6.811626707062726e-06, "loss": 0.8552, "step": 4694 }, { "epoch": 0.4, "learning_rate": 6.810338643372795e-06, "loss": 0.8444, "step": 4695 }, { "epoch": 0.4, "learning_rate": 6.809050441400162e-06, "loss": 0.8539, "step": 4696 }, { "epoch": 0.4, "learning_rate": 6.807762101243231e-06, "loss": 0.8101, "step": 4697 }, { "epoch": 0.4, "learning_rate": 6.806473623000412e-06, "loss": 0.8588, "step": 4698 }, { "epoch": 0.4, "learning_rate": 6.805185006770125e-06, "loss": 0.8119, "step": 4699 }, { "epoch": 0.4, "learning_rate": 6.803896252650799e-06, "loss": 0.8062, "step": 4700 }, { "epoch": 0.4, "learning_rate": 6.802607360740878e-06, "loss": 0.8858, "step": 4701 }, { "epoch": 0.4, "learning_rate": 6.801318331138811e-06, "loss": 0.8198, "step": 4702 }, { "epoch": 0.4, "learning_rate": 6.800029163943063e-06, "loss": 0.8454, "step": 4703 }, { "epoch": 0.4, "learning_rate": 6.798739859252108e-06, "loss": 0.8266, "step": 4704 }, { "epoch": 0.4, "learning_rate": 6.7974504171644275e-06, "loss": 0.9388, "step": 4705 }, { "epoch": 0.4, "learning_rate": 6.796160837778515e-06, "loss": 0.8722, "step": 4706 }, { "epoch": 0.4, "learning_rate": 6.794871121192875e-06, "loss": 0.8122, "step": 4707 }, { "epoch": 0.4, "learning_rate": 6.7935812675060255e-06, "loss": 0.8645, "step": 4708 }, { "epoch": 0.4, "learning_rate": 6.792291276816489e-06, "loss": 0.8467, "step": 4709 }, { "epoch": 0.4, "learning_rate": 6.791001149222802e-06, "loss": 0.8183, "step": 4710 }, { "epoch": 0.4, "learning_rate": 6.789710884823513e-06, "loss": 0.9081, "step": 4711 }, { "epoch": 0.4, "learning_rate": 6.788420483717175e-06, "loss": 0.8835, "step": 4712 }, { "epoch": 0.4, "learning_rate": 6.787129946002358e-06, "loss": 0.8682, "step": 4713 }, { "epoch": 0.4, "learning_rate": 6.78583927177764e-06, "loss": 0.8694, "step": 4714 }, { "epoch": 0.4, "learning_rate": 6.784548461141608e-06, "loss": 0.821, "step": 4715 }, { "epoch": 0.4, "learning_rate": 6.783257514192859e-06, "loss": 0.8921, "step": 4716 }, { "epoch": 0.4, "learning_rate": 6.781966431030005e-06, "loss": 0.8689, "step": 4717 }, { "epoch": 0.4, "learning_rate": 6.780675211751664e-06, "loss": 0.8431, "step": 4718 }, { "epoch": 0.4, "learning_rate": 6.779383856456467e-06, "loss": 0.8371, "step": 4719 }, { "epoch": 0.4, "learning_rate": 6.77809236524305e-06, "loss": 0.8622, "step": 4720 }, { "epoch": 0.4, "learning_rate": 6.77680073821007e-06, "loss": 0.8606, "step": 4721 }, { "epoch": 0.4, "learning_rate": 6.775508975456182e-06, "loss": 0.8891, "step": 4722 }, { "epoch": 0.4, "learning_rate": 6.774217077080061e-06, "loss": 0.8652, "step": 4723 }, { "epoch": 0.4, "learning_rate": 6.772925043180385e-06, "loss": 0.8359, "step": 4724 }, { "epoch": 0.4, "learning_rate": 6.771632873855851e-06, "loss": 0.8567, "step": 4725 }, { "epoch": 0.4, "learning_rate": 6.7703405692051585e-06, "loss": 0.9096, "step": 4726 }, { "epoch": 0.4, "learning_rate": 6.76904812932702e-06, "loss": 0.8395, "step": 4727 }, { "epoch": 0.4, "learning_rate": 6.767755554320159e-06, "loss": 0.8365, "step": 4728 }, { "epoch": 0.4, "learning_rate": 6.766462844283308e-06, "loss": 0.8451, "step": 4729 }, { "epoch": 0.4, "learning_rate": 6.765169999315211e-06, "loss": 0.8637, "step": 4730 }, { "epoch": 0.4, "learning_rate": 6.7638770195146245e-06, "loss": 0.8104, "step": 4731 }, { "epoch": 0.4, "learning_rate": 6.762583904980309e-06, "loss": 0.8646, "step": 4732 }, { "epoch": 0.4, "learning_rate": 6.761290655811041e-06, "loss": 0.8585, "step": 4733 }, { "epoch": 0.4, "learning_rate": 6.7599972721056054e-06, "loss": 0.8838, "step": 4734 }, { "epoch": 0.4, "learning_rate": 6.758703753962797e-06, "loss": 0.8267, "step": 4735 }, { "epoch": 0.4, "learning_rate": 6.757410101481419e-06, "loss": 0.8646, "step": 4736 }, { "epoch": 0.4, "learning_rate": 6.75611631476029e-06, "loss": 0.8486, "step": 4737 }, { "epoch": 0.4, "learning_rate": 6.754822393898237e-06, "loss": 0.8604, "step": 4738 }, { "epoch": 0.4, "learning_rate": 6.7535283389940905e-06, "loss": 0.895, "step": 4739 }, { "epoch": 0.4, "learning_rate": 6.752234150146703e-06, "loss": 0.8666, "step": 4740 }, { "epoch": 0.4, "learning_rate": 6.750939827454928e-06, "loss": 0.8205, "step": 4741 }, { "epoch": 0.4, "learning_rate": 6.749645371017634e-06, "loss": 0.8387, "step": 4742 }, { "epoch": 0.4, "learning_rate": 6.748350780933695e-06, "loss": 0.8185, "step": 4743 }, { "epoch": 0.4, "learning_rate": 6.747056057302001e-06, "loss": 0.7624, "step": 4744 }, { "epoch": 0.4, "learning_rate": 6.745761200221448e-06, "loss": 0.5446, "step": 4745 }, { "epoch": 0.4, "learning_rate": 6.744466209790945e-06, "loss": 0.8433, "step": 4746 }, { "epoch": 0.41, "learning_rate": 6.743171086109408e-06, "loss": 0.8526, "step": 4747 }, { "epoch": 0.41, "learning_rate": 6.741875829275768e-06, "loss": 0.85, "step": 4748 }, { "epoch": 0.41, "learning_rate": 6.740580439388959e-06, "loss": 0.8343, "step": 4749 }, { "epoch": 0.41, "learning_rate": 6.739284916547932e-06, "loss": 0.8529, "step": 4750 }, { "epoch": 0.41, "learning_rate": 6.737989260851646e-06, "loss": 0.8635, "step": 4751 }, { "epoch": 0.41, "learning_rate": 6.73669347239907e-06, "loss": 0.8425, "step": 4752 }, { "epoch": 0.41, "learning_rate": 6.735397551289179e-06, "loss": 0.8973, "step": 4753 }, { "epoch": 0.41, "learning_rate": 6.734101497620966e-06, "loss": 0.8904, "step": 4754 }, { "epoch": 0.41, "learning_rate": 6.732805311493427e-06, "loss": 0.7864, "step": 4755 }, { "epoch": 0.41, "learning_rate": 6.731508993005574e-06, "loss": 0.8254, "step": 4756 }, { "epoch": 0.41, "learning_rate": 6.730212542256424e-06, "loss": 0.838, "step": 4757 }, { "epoch": 0.41, "learning_rate": 6.728915959345009e-06, "loss": 0.8295, "step": 4758 }, { "epoch": 0.41, "learning_rate": 6.727619244370365e-06, "loss": 0.874, "step": 4759 }, { "epoch": 0.41, "learning_rate": 6.726322397431545e-06, "loss": 0.868, "step": 4760 }, { "epoch": 0.41, "learning_rate": 6.725025418627606e-06, "loss": 0.8818, "step": 4761 }, { "epoch": 0.41, "learning_rate": 6.723728308057619e-06, "loss": 0.8244, "step": 4762 }, { "epoch": 0.41, "learning_rate": 6.722431065820663e-06, "loss": 0.8679, "step": 4763 }, { "epoch": 0.41, "learning_rate": 6.721133692015829e-06, "loss": 0.8381, "step": 4764 }, { "epoch": 0.41, "learning_rate": 6.719836186742215e-06, "loss": 0.8171, "step": 4765 }, { "epoch": 0.41, "learning_rate": 6.718538550098933e-06, "loss": 0.8879, "step": 4766 }, { "epoch": 0.41, "learning_rate": 6.717240782185102e-06, "loss": 0.8697, "step": 4767 }, { "epoch": 0.41, "learning_rate": 6.715942883099852e-06, "loss": 0.8658, "step": 4768 }, { "epoch": 0.41, "learning_rate": 6.714644852942322e-06, "loss": 0.8603, "step": 4769 }, { "epoch": 0.41, "learning_rate": 6.713346691811665e-06, "loss": 0.8758, "step": 4770 }, { "epoch": 0.41, "learning_rate": 6.71204839980704e-06, "loss": 0.8641, "step": 4771 }, { "epoch": 0.41, "learning_rate": 6.710749977027613e-06, "loss": 0.8254, "step": 4772 }, { "epoch": 0.41, "learning_rate": 6.709451423572568e-06, "loss": 0.8347, "step": 4773 }, { "epoch": 0.41, "learning_rate": 6.708152739541096e-06, "loss": 0.7973, "step": 4774 }, { "epoch": 0.41, "learning_rate": 6.706853925032394e-06, "loss": 0.8422, "step": 4775 }, { "epoch": 0.41, "learning_rate": 6.705554980145672e-06, "loss": 0.8383, "step": 4776 }, { "epoch": 0.41, "learning_rate": 6.704255904980152e-06, "loss": 0.8979, "step": 4777 }, { "epoch": 0.41, "learning_rate": 6.7029566996350636e-06, "loss": 0.8489, "step": 4778 }, { "epoch": 0.41, "learning_rate": 6.701657364209645e-06, "loss": 0.8174, "step": 4779 }, { "epoch": 0.41, "learning_rate": 6.700357898803146e-06, "loss": 0.8365, "step": 4780 }, { "epoch": 0.41, "learning_rate": 6.699058303514828e-06, "loss": 0.8592, "step": 4781 }, { "epoch": 0.41, "learning_rate": 6.697758578443958e-06, "loss": 0.9138, "step": 4782 }, { "epoch": 0.41, "learning_rate": 6.696458723689818e-06, "loss": 0.8572, "step": 4783 }, { "epoch": 0.41, "learning_rate": 6.6951587393516985e-06, "loss": 0.8307, "step": 4784 }, { "epoch": 0.41, "learning_rate": 6.693858625528894e-06, "loss": 0.8779, "step": 4785 }, { "epoch": 0.41, "learning_rate": 6.6925583823207164e-06, "loss": 0.8599, "step": 4786 }, { "epoch": 0.41, "learning_rate": 6.691258009826484e-06, "loss": 0.7966, "step": 4787 }, { "epoch": 0.41, "learning_rate": 6.689957508145528e-06, "loss": 0.8539, "step": 4788 }, { "epoch": 0.41, "learning_rate": 6.6886568773771865e-06, "loss": 0.8822, "step": 4789 }, { "epoch": 0.41, "learning_rate": 6.687356117620805e-06, "loss": 0.8705, "step": 4790 }, { "epoch": 0.41, "learning_rate": 6.686055228975747e-06, "loss": 0.8641, "step": 4791 }, { "epoch": 0.41, "learning_rate": 6.684754211541375e-06, "loss": 0.831, "step": 4792 }, { "epoch": 0.41, "learning_rate": 6.683453065417074e-06, "loss": 0.8358, "step": 4793 }, { "epoch": 0.41, "learning_rate": 6.682151790702227e-06, "loss": 0.8515, "step": 4794 }, { "epoch": 0.41, "learning_rate": 6.680850387496233e-06, "loss": 0.8662, "step": 4795 }, { "epoch": 0.41, "learning_rate": 6.679548855898499e-06, "loss": 0.8509, "step": 4796 }, { "epoch": 0.41, "learning_rate": 6.678247196008445e-06, "loss": 0.8585, "step": 4797 }, { "epoch": 0.41, "learning_rate": 6.676945407925497e-06, "loss": 0.8527, "step": 4798 }, { "epoch": 0.41, "learning_rate": 6.675643491749091e-06, "loss": 0.9104, "step": 4799 }, { "epoch": 0.41, "learning_rate": 6.674341447578674e-06, "loss": 0.8572, "step": 4800 }, { "epoch": 0.41, "learning_rate": 6.673039275513704e-06, "loss": 0.8136, "step": 4801 }, { "epoch": 0.41, "learning_rate": 6.671736975653647e-06, "loss": 0.8483, "step": 4802 }, { "epoch": 0.41, "learning_rate": 6.670434548097978e-06, "loss": 0.8651, "step": 4803 }, { "epoch": 0.41, "learning_rate": 6.669131992946185e-06, "loss": 0.7962, "step": 4804 }, { "epoch": 0.41, "learning_rate": 6.667829310297761e-06, "loss": 0.8661, "step": 4805 }, { "epoch": 0.41, "learning_rate": 6.666526500252213e-06, "loss": 0.8411, "step": 4806 }, { "epoch": 0.41, "learning_rate": 6.665223562909058e-06, "loss": 0.8829, "step": 4807 }, { "epoch": 0.41, "learning_rate": 6.663920498367818e-06, "loss": 0.8296, "step": 4808 }, { "epoch": 0.41, "learning_rate": 6.662617306728027e-06, "loss": 0.8747, "step": 4809 }, { "epoch": 0.41, "learning_rate": 6.661313988089231e-06, "loss": 0.9102, "step": 4810 }, { "epoch": 0.41, "learning_rate": 6.660010542550984e-06, "loss": 0.862, "step": 4811 }, { "epoch": 0.41, "learning_rate": 6.658706970212849e-06, "loss": 0.8543, "step": 4812 }, { "epoch": 0.41, "learning_rate": 6.6574032711744e-06, "loss": 0.8244, "step": 4813 }, { "epoch": 0.41, "learning_rate": 6.65609944553522e-06, "loss": 0.6216, "step": 4814 }, { "epoch": 0.41, "learning_rate": 6.654795493394901e-06, "loss": 0.8522, "step": 4815 }, { "epoch": 0.41, "learning_rate": 6.653491414853046e-06, "loss": 0.7987, "step": 4816 }, { "epoch": 0.41, "learning_rate": 6.652187210009268e-06, "loss": 0.8711, "step": 4817 }, { "epoch": 0.41, "learning_rate": 6.650882878963185e-06, "loss": 0.8522, "step": 4818 }, { "epoch": 0.41, "learning_rate": 6.6495784218144326e-06, "loss": 0.8419, "step": 4819 }, { "epoch": 0.41, "learning_rate": 6.648273838662649e-06, "loss": 0.811, "step": 4820 }, { "epoch": 0.41, "learning_rate": 6.646969129607487e-06, "loss": 0.8728, "step": 4821 }, { "epoch": 0.41, "learning_rate": 6.645664294748605e-06, "loss": 0.8032, "step": 4822 }, { "epoch": 0.41, "learning_rate": 6.6443593341856725e-06, "loss": 0.8347, "step": 4823 }, { "epoch": 0.41, "learning_rate": 6.6430542480183705e-06, "loss": 0.7982, "step": 4824 }, { "epoch": 0.41, "learning_rate": 6.6417490363463855e-06, "loss": 0.8752, "step": 4825 }, { "epoch": 0.41, "learning_rate": 6.640443699269421e-06, "loss": 0.8882, "step": 4826 }, { "epoch": 0.41, "learning_rate": 6.63913823688718e-06, "loss": 0.7972, "step": 4827 }, { "epoch": 0.41, "learning_rate": 6.637832649299382e-06, "loss": 0.8559, "step": 4828 }, { "epoch": 0.41, "learning_rate": 6.636526936605754e-06, "loss": 0.8546, "step": 4829 }, { "epoch": 0.41, "learning_rate": 6.635221098906035e-06, "loss": 0.6379, "step": 4830 }, { "epoch": 0.41, "learning_rate": 6.633915136299969e-06, "loss": 0.8876, "step": 4831 }, { "epoch": 0.41, "learning_rate": 6.6326090488873115e-06, "loss": 0.867, "step": 4832 }, { "epoch": 0.41, "learning_rate": 6.63130283676783e-06, "loss": 0.833, "step": 4833 }, { "epoch": 0.41, "learning_rate": 6.629996500041299e-06, "loss": 0.8345, "step": 4834 }, { "epoch": 0.41, "learning_rate": 6.628690038807503e-06, "loss": 0.8487, "step": 4835 }, { "epoch": 0.41, "learning_rate": 6.627383453166234e-06, "loss": 0.8905, "step": 4836 }, { "epoch": 0.41, "learning_rate": 6.626076743217299e-06, "loss": 0.841, "step": 4837 }, { "epoch": 0.41, "learning_rate": 6.6247699090605075e-06, "loss": 0.8426, "step": 4838 }, { "epoch": 0.41, "learning_rate": 6.623462950795685e-06, "loss": 0.8823, "step": 4839 }, { "epoch": 0.41, "learning_rate": 6.622155868522662e-06, "loss": 0.8425, "step": 4840 }, { "epoch": 0.41, "learning_rate": 6.6208486623412805e-06, "loss": 0.8313, "step": 4841 }, { "epoch": 0.41, "learning_rate": 6.61954133235139e-06, "loss": 0.8569, "step": 4842 }, { "epoch": 0.41, "learning_rate": 6.618233878652852e-06, "loss": 0.857, "step": 4843 }, { "epoch": 0.41, "learning_rate": 6.6169263013455375e-06, "loss": 0.8941, "step": 4844 }, { "epoch": 0.41, "learning_rate": 6.615618600529326e-06, "loss": 0.863, "step": 4845 }, { "epoch": 0.41, "learning_rate": 6.614310776304101e-06, "loss": 0.8562, "step": 4846 }, { "epoch": 0.41, "learning_rate": 6.613002828769768e-06, "loss": 0.885, "step": 4847 }, { "epoch": 0.41, "learning_rate": 6.611694758026228e-06, "loss": 0.8404, "step": 4848 }, { "epoch": 0.41, "learning_rate": 6.610386564173403e-06, "loss": 0.8026, "step": 4849 }, { "epoch": 0.41, "learning_rate": 6.609078247311217e-06, "loss": 0.84, "step": 4850 }, { "epoch": 0.41, "learning_rate": 6.6077698075396065e-06, "loss": 0.8617, "step": 4851 }, { "epoch": 0.41, "learning_rate": 6.606461244958515e-06, "loss": 0.8608, "step": 4852 }, { "epoch": 0.41, "learning_rate": 6.605152559667899e-06, "loss": 0.8362, "step": 4853 }, { "epoch": 0.41, "learning_rate": 6.603843751767724e-06, "loss": 0.837, "step": 4854 }, { "epoch": 0.41, "learning_rate": 6.602534821357958e-06, "loss": 0.8216, "step": 4855 }, { "epoch": 0.41, "learning_rate": 6.6012257685385875e-06, "loss": 0.8443, "step": 4856 }, { "epoch": 0.41, "learning_rate": 6.5999165934096044e-06, "loss": 0.7886, "step": 4857 }, { "epoch": 0.41, "learning_rate": 6.598607296071008e-06, "loss": 0.8046, "step": 4858 }, { "epoch": 0.41, "learning_rate": 6.597297876622811e-06, "loss": 0.8451, "step": 4859 }, { "epoch": 0.41, "learning_rate": 6.595988335165033e-06, "loss": 0.8053, "step": 4860 }, { "epoch": 0.41, "learning_rate": 6.5946786717977026e-06, "loss": 0.8601, "step": 4861 }, { "epoch": 0.41, "learning_rate": 6.593368886620859e-06, "loss": 0.8489, "step": 4862 }, { "epoch": 0.41, "learning_rate": 6.59205897973455e-06, "loss": 0.8523, "step": 4863 }, { "epoch": 0.42, "learning_rate": 6.5907489512388324e-06, "loss": 0.8618, "step": 4864 }, { "epoch": 0.42, "learning_rate": 6.589438801233773e-06, "loss": 0.9175, "step": 4865 }, { "epoch": 0.42, "learning_rate": 6.588128529819446e-06, "loss": 0.8684, "step": 4866 }, { "epoch": 0.42, "learning_rate": 6.586818137095942e-06, "loss": 0.8488, "step": 4867 }, { "epoch": 0.42, "learning_rate": 6.5855076231633496e-06, "loss": 0.9242, "step": 4868 }, { "epoch": 0.42, "learning_rate": 6.584196988121773e-06, "loss": 0.8731, "step": 4869 }, { "epoch": 0.42, "learning_rate": 6.582886232071327e-06, "loss": 0.8326, "step": 4870 }, { "epoch": 0.42, "learning_rate": 6.581575355112133e-06, "loss": 0.8818, "step": 4871 }, { "epoch": 0.42, "learning_rate": 6.580264357344321e-06, "loss": 0.8662, "step": 4872 }, { "epoch": 0.42, "learning_rate": 6.578953238868035e-06, "loss": 0.8386, "step": 4873 }, { "epoch": 0.42, "learning_rate": 6.577641999783423e-06, "loss": 0.8618, "step": 4874 }, { "epoch": 0.42, "learning_rate": 6.576330640190641e-06, "loss": 0.8534, "step": 4875 }, { "epoch": 0.42, "learning_rate": 6.575019160189861e-06, "loss": 0.8369, "step": 4876 }, { "epoch": 0.42, "learning_rate": 6.57370755988126e-06, "loss": 0.8884, "step": 4877 }, { "epoch": 0.42, "learning_rate": 6.5723958393650236e-06, "loss": 0.8599, "step": 4878 }, { "epoch": 0.42, "learning_rate": 6.571083998741346e-06, "loss": 0.8548, "step": 4879 }, { "epoch": 0.42, "learning_rate": 6.569772038110436e-06, "loss": 0.872, "step": 4880 }, { "epoch": 0.42, "learning_rate": 6.568459957572504e-06, "loss": 0.878, "step": 4881 }, { "epoch": 0.42, "learning_rate": 6.567147757227776e-06, "loss": 0.8696, "step": 4882 }, { "epoch": 0.42, "learning_rate": 6.565835437176482e-06, "loss": 0.8358, "step": 4883 }, { "epoch": 0.42, "learning_rate": 6.564522997518865e-06, "loss": 0.876, "step": 4884 }, { "epoch": 0.42, "learning_rate": 6.563210438355176e-06, "loss": 0.5739, "step": 4885 }, { "epoch": 0.42, "learning_rate": 6.561897759785674e-06, "loss": 0.8565, "step": 4886 }, { "epoch": 0.42, "learning_rate": 6.5605849619106285e-06, "loss": 0.8566, "step": 4887 }, { "epoch": 0.42, "learning_rate": 6.5592720448303174e-06, "loss": 0.8449, "step": 4888 }, { "epoch": 0.42, "learning_rate": 6.557959008645026e-06, "loss": 0.8782, "step": 4889 }, { "epoch": 0.42, "learning_rate": 6.5566458534550535e-06, "loss": 0.8263, "step": 4890 }, { "epoch": 0.42, "learning_rate": 6.555332579360703e-06, "loss": 0.8385, "step": 4891 }, { "epoch": 0.42, "learning_rate": 6.554019186462293e-06, "loss": 0.8342, "step": 4892 }, { "epoch": 0.42, "learning_rate": 6.55270567486014e-06, "loss": 0.8373, "step": 4893 }, { "epoch": 0.42, "learning_rate": 6.551392044654583e-06, "loss": 0.8866, "step": 4894 }, { "epoch": 0.42, "learning_rate": 6.5500782959459595e-06, "loss": 0.8473, "step": 4895 }, { "epoch": 0.42, "learning_rate": 6.548764428834624e-06, "loss": 0.8474, "step": 4896 }, { "epoch": 0.42, "learning_rate": 6.547450443420933e-06, "loss": 0.8246, "step": 4897 }, { "epoch": 0.42, "learning_rate": 6.546136339805256e-06, "loss": 0.8462, "step": 4898 }, { "epoch": 0.42, "learning_rate": 6.54482211808797e-06, "loss": 0.8251, "step": 4899 }, { "epoch": 0.42, "learning_rate": 6.543507778369464e-06, "loss": 0.8293, "step": 4900 }, { "epoch": 0.42, "learning_rate": 6.542193320750134e-06, "loss": 0.8415, "step": 4901 }, { "epoch": 0.42, "learning_rate": 6.540878745330382e-06, "loss": 0.8739, "step": 4902 }, { "epoch": 0.42, "learning_rate": 6.539564052210623e-06, "loss": 0.8169, "step": 4903 }, { "epoch": 0.42, "learning_rate": 6.538249241491281e-06, "loss": 0.8842, "step": 4904 }, { "epoch": 0.42, "learning_rate": 6.536934313272788e-06, "loss": 0.8737, "step": 4905 }, { "epoch": 0.42, "learning_rate": 6.535619267655582e-06, "loss": 0.8226, "step": 4906 }, { "epoch": 0.42, "learning_rate": 6.534304104740115e-06, "loss": 0.5403, "step": 4907 }, { "epoch": 0.42, "learning_rate": 6.532988824626846e-06, "loss": 0.8572, "step": 4908 }, { "epoch": 0.42, "learning_rate": 6.53167342741624e-06, "loss": 0.8485, "step": 4909 }, { "epoch": 0.42, "learning_rate": 6.530357913208779e-06, "loss": 0.8423, "step": 4910 }, { "epoch": 0.42, "learning_rate": 6.529042282104943e-06, "loss": 0.8123, "step": 4911 }, { "epoch": 0.42, "learning_rate": 6.527726534205228e-06, "loss": 0.6598, "step": 4912 }, { "epoch": 0.42, "learning_rate": 6.526410669610137e-06, "loss": 0.8818, "step": 4913 }, { "epoch": 0.42, "learning_rate": 6.525094688420185e-06, "loss": 0.8609, "step": 4914 }, { "epoch": 0.42, "learning_rate": 6.523778590735892e-06, "loss": 0.8315, "step": 4915 }, { "epoch": 0.42, "learning_rate": 6.522462376657785e-06, "loss": 0.8783, "step": 4916 }, { "epoch": 0.42, "learning_rate": 6.521146046286409e-06, "loss": 0.8576, "step": 4917 }, { "epoch": 0.42, "learning_rate": 6.519829599722305e-06, "loss": 0.8585, "step": 4918 }, { "epoch": 0.42, "learning_rate": 6.518513037066034e-06, "loss": 0.8632, "step": 4919 }, { "epoch": 0.42, "learning_rate": 6.517196358418163e-06, "loss": 0.8372, "step": 4920 }, { "epoch": 0.42, "learning_rate": 6.515879563879262e-06, "loss": 0.8037, "step": 4921 }, { "epoch": 0.42, "learning_rate": 6.514562653549916e-06, "loss": 0.8154, "step": 4922 }, { "epoch": 0.42, "learning_rate": 6.513245627530719e-06, "loss": 0.8701, "step": 4923 }, { "epoch": 0.42, "learning_rate": 6.511928485922272e-06, "loss": 0.8612, "step": 4924 }, { "epoch": 0.42, "learning_rate": 6.5106112288251825e-06, "loss": 0.8227, "step": 4925 }, { "epoch": 0.42, "learning_rate": 6.509293856340069e-06, "loss": 0.7868, "step": 4926 }, { "epoch": 0.42, "learning_rate": 6.507976368567562e-06, "loss": 0.8239, "step": 4927 }, { "epoch": 0.42, "learning_rate": 6.506658765608296e-06, "loss": 0.9144, "step": 4928 }, { "epoch": 0.42, "learning_rate": 6.505341047562915e-06, "loss": 0.8212, "step": 4929 }, { "epoch": 0.42, "learning_rate": 6.504023214532076e-06, "loss": 0.7992, "step": 4930 }, { "epoch": 0.42, "learning_rate": 6.502705266616438e-06, "loss": 0.8708, "step": 4931 }, { "epoch": 0.42, "learning_rate": 6.501387203916674e-06, "loss": 0.822, "step": 4932 }, { "epoch": 0.42, "learning_rate": 6.5000690265334655e-06, "loss": 0.8436, "step": 4933 }, { "epoch": 0.42, "learning_rate": 6.498750734567501e-06, "loss": 0.8622, "step": 4934 }, { "epoch": 0.42, "learning_rate": 6.497432328119475e-06, "loss": 0.8496, "step": 4935 }, { "epoch": 0.42, "learning_rate": 6.496113807290098e-06, "loss": 0.8446, "step": 4936 }, { "epoch": 0.42, "learning_rate": 6.494795172180084e-06, "loss": 0.865, "step": 4937 }, { "epoch": 0.42, "learning_rate": 6.493476422890157e-06, "loss": 0.8705, "step": 4938 }, { "epoch": 0.42, "learning_rate": 6.492157559521047e-06, "loss": 0.8734, "step": 4939 }, { "epoch": 0.42, "learning_rate": 6.4908385821735e-06, "loss": 0.6452, "step": 4940 }, { "epoch": 0.42, "learning_rate": 6.489519490948263e-06, "loss": 0.804, "step": 4941 }, { "epoch": 0.42, "learning_rate": 6.488200285946094e-06, "loss": 0.8775, "step": 4942 }, { "epoch": 0.42, "learning_rate": 6.4868809672677655e-06, "loss": 0.859, "step": 4943 }, { "epoch": 0.42, "learning_rate": 6.4855615350140465e-06, "loss": 0.8195, "step": 4944 }, { "epoch": 0.42, "learning_rate": 6.4842419892857265e-06, "loss": 0.8213, "step": 4945 }, { "epoch": 0.42, "learning_rate": 6.482922330183598e-06, "loss": 0.8705, "step": 4946 }, { "epoch": 0.42, "learning_rate": 6.481602557808464e-06, "loss": 0.8766, "step": 4947 }, { "epoch": 0.42, "learning_rate": 6.480282672261134e-06, "loss": 0.8458, "step": 4948 }, { "epoch": 0.42, "learning_rate": 6.478962673642427e-06, "loss": 0.8695, "step": 4949 }, { "epoch": 0.42, "learning_rate": 6.477642562053172e-06, "loss": 0.9098, "step": 4950 }, { "epoch": 0.42, "learning_rate": 6.4763223375942065e-06, "loss": 0.8391, "step": 4951 }, { "epoch": 0.42, "learning_rate": 6.475002000366375e-06, "loss": 0.8128, "step": 4952 }, { "epoch": 0.42, "learning_rate": 6.473681550470531e-06, "loss": 0.8548, "step": 4953 }, { "epoch": 0.42, "learning_rate": 6.472360988007536e-06, "loss": 0.8254, "step": 4954 }, { "epoch": 0.42, "learning_rate": 6.471040313078263e-06, "loss": 0.8755, "step": 4955 }, { "epoch": 0.42, "learning_rate": 6.469719525783592e-06, "loss": 0.9024, "step": 4956 }, { "epoch": 0.42, "learning_rate": 6.468398626224412e-06, "loss": 0.8005, "step": 4957 }, { "epoch": 0.42, "learning_rate": 6.467077614501615e-06, "loss": 0.8383, "step": 4958 }, { "epoch": 0.42, "learning_rate": 6.465756490716112e-06, "loss": 0.8177, "step": 4959 }, { "epoch": 0.42, "learning_rate": 6.4644352549688165e-06, "loss": 0.8422, "step": 4960 }, { "epoch": 0.42, "learning_rate": 6.463113907360648e-06, "loss": 0.8743, "step": 4961 }, { "epoch": 0.42, "learning_rate": 6.461792447992539e-06, "loss": 0.8295, "step": 4962 }, { "epoch": 0.42, "learning_rate": 6.46047087696543e-06, "loss": 0.7972, "step": 4963 }, { "epoch": 0.42, "learning_rate": 6.459149194380268e-06, "loss": 0.8691, "step": 4964 }, { "epoch": 0.42, "learning_rate": 6.457827400338009e-06, "loss": 0.8, "step": 4965 }, { "epoch": 0.42, "learning_rate": 6.456505494939622e-06, "loss": 0.8329, "step": 4966 }, { "epoch": 0.42, "learning_rate": 6.455183478286078e-06, "loss": 0.8745, "step": 4967 }, { "epoch": 0.42, "learning_rate": 6.4538613504783576e-06, "loss": 0.5923, "step": 4968 }, { "epoch": 0.42, "learning_rate": 6.452539111617454e-06, "loss": 0.8165, "step": 4969 }, { "epoch": 0.42, "learning_rate": 6.4512167618043665e-06, "loss": 0.8595, "step": 4970 }, { "epoch": 0.42, "learning_rate": 6.449894301140103e-06, "loss": 0.8179, "step": 4971 }, { "epoch": 0.42, "learning_rate": 6.448571729725677e-06, "loss": 0.7909, "step": 4972 }, { "epoch": 0.42, "learning_rate": 6.447249047662115e-06, "loss": 0.8571, "step": 4973 }, { "epoch": 0.42, "learning_rate": 6.44592625505045e-06, "loss": 0.8423, "step": 4974 }, { "epoch": 0.42, "learning_rate": 6.444603351991724e-06, "loss": 0.8441, "step": 4975 }, { "epoch": 0.42, "learning_rate": 6.443280338586985e-06, "loss": 0.8909, "step": 4976 }, { "epoch": 0.42, "learning_rate": 6.441957214937293e-06, "loss": 0.8457, "step": 4977 }, { "epoch": 0.42, "learning_rate": 6.440633981143715e-06, "loss": 0.9109, "step": 4978 }, { "epoch": 0.42, "learning_rate": 6.4393106373073254e-06, "loss": 0.8471, "step": 4979 }, { "epoch": 0.42, "learning_rate": 6.43798718352921e-06, "loss": 0.8168, "step": 4980 }, { "epoch": 0.43, "learning_rate": 6.436663619910456e-06, "loss": 0.5548, "step": 4981 }, { "epoch": 0.43, "learning_rate": 6.435339946552167e-06, "loss": 0.8439, "step": 4982 }, { "epoch": 0.43, "learning_rate": 6.434016163555452e-06, "loss": 0.8654, "step": 4983 }, { "epoch": 0.43, "learning_rate": 6.432692271021428e-06, "loss": 0.8452, "step": 4984 }, { "epoch": 0.43, "learning_rate": 6.431368269051219e-06, "loss": 0.8074, "step": 4985 }, { "epoch": 0.43, "learning_rate": 6.43004415774596e-06, "loss": 0.8611, "step": 4986 }, { "epoch": 0.43, "learning_rate": 6.428719937206794e-06, "loss": 0.8822, "step": 4987 }, { "epoch": 0.43, "learning_rate": 6.427395607534871e-06, "loss": 0.8622, "step": 4988 }, { "epoch": 0.43, "learning_rate": 6.426071168831349e-06, "loss": 0.8404, "step": 4989 }, { "epoch": 0.43, "learning_rate": 6.424746621197398e-06, "loss": 0.8785, "step": 4990 }, { "epoch": 0.43, "learning_rate": 6.423421964734188e-06, "loss": 0.8964, "step": 4991 }, { "epoch": 0.43, "learning_rate": 6.422097199542907e-06, "loss": 0.8453, "step": 4992 }, { "epoch": 0.43, "learning_rate": 6.420772325724749e-06, "loss": 0.8629, "step": 4993 }, { "epoch": 0.43, "learning_rate": 6.41944734338091e-06, "loss": 0.8471, "step": 4994 }, { "epoch": 0.43, "learning_rate": 6.418122252612601e-06, "loss": 0.8541, "step": 4995 }, { "epoch": 0.43, "learning_rate": 6.416797053521039e-06, "loss": 0.8373, "step": 4996 }, { "epoch": 0.43, "learning_rate": 6.415471746207449e-06, "loss": 0.7874, "step": 4997 }, { "epoch": 0.43, "learning_rate": 6.4141463307730635e-06, "loss": 0.8463, "step": 4998 }, { "epoch": 0.43, "learning_rate": 6.412820807319127e-06, "loss": 0.8386, "step": 4999 }, { "epoch": 0.43, "learning_rate": 6.411495175946889e-06, "loss": 0.8552, "step": 5000 }, { "epoch": 0.43, "learning_rate": 6.4101694367576055e-06, "loss": 0.8698, "step": 5001 }, { "epoch": 0.43, "learning_rate": 6.408843589852544e-06, "loss": 0.5876, "step": 5002 }, { "epoch": 0.43, "learning_rate": 6.40751763533298e-06, "loss": 0.8927, "step": 5003 }, { "epoch": 0.43, "learning_rate": 6.406191573300198e-06, "loss": 0.8036, "step": 5004 }, { "epoch": 0.43, "learning_rate": 6.404865403855487e-06, "loss": 0.8957, "step": 5005 }, { "epoch": 0.43, "learning_rate": 6.403539127100147e-06, "loss": 0.8313, "step": 5006 }, { "epoch": 0.43, "learning_rate": 6.402212743135484e-06, "loss": 0.8744, "step": 5007 }, { "epoch": 0.43, "learning_rate": 6.400886252062818e-06, "loss": 0.8766, "step": 5008 }, { "epoch": 0.43, "learning_rate": 6.3995596539834685e-06, "loss": 0.8465, "step": 5009 }, { "epoch": 0.43, "learning_rate": 6.39823294899877e-06, "loss": 0.8241, "step": 5010 }, { "epoch": 0.43, "learning_rate": 6.3969061372100636e-06, "loss": 0.8273, "step": 5011 }, { "epoch": 0.43, "learning_rate": 6.395579218718696e-06, "loss": 0.8786, "step": 5012 }, { "epoch": 0.43, "learning_rate": 6.394252193626026e-06, "loss": 0.8342, "step": 5013 }, { "epoch": 0.43, "learning_rate": 6.3929250620334145e-06, "loss": 0.8742, "step": 5014 }, { "epoch": 0.43, "learning_rate": 6.391597824042239e-06, "loss": 0.7933, "step": 5015 }, { "epoch": 0.43, "learning_rate": 6.390270479753877e-06, "loss": 0.8631, "step": 5016 }, { "epoch": 0.43, "learning_rate": 6.38894302926972e-06, "loss": 0.8225, "step": 5017 }, { "epoch": 0.43, "learning_rate": 6.387615472691164e-06, "loss": 0.8105, "step": 5018 }, { "epoch": 0.43, "learning_rate": 6.386287810119614e-06, "loss": 0.8438, "step": 5019 }, { "epoch": 0.43, "learning_rate": 6.384960041656486e-06, "loss": 0.8751, "step": 5020 }, { "epoch": 0.43, "learning_rate": 6.3836321674032e-06, "loss": 0.8104, "step": 5021 }, { "epoch": 0.43, "learning_rate": 6.382304187461185e-06, "loss": 0.8408, "step": 5022 }, { "epoch": 0.43, "learning_rate": 6.380976101931879e-06, "loss": 0.7976, "step": 5023 }, { "epoch": 0.43, "learning_rate": 6.379647910916729e-06, "loss": 0.8607, "step": 5024 }, { "epoch": 0.43, "learning_rate": 6.378319614517185e-06, "loss": 0.8156, "step": 5025 }, { "epoch": 0.43, "learning_rate": 6.376991212834715e-06, "loss": 0.8434, "step": 5026 }, { "epoch": 0.43, "learning_rate": 6.375662705970786e-06, "loss": 0.8619, "step": 5027 }, { "epoch": 0.43, "learning_rate": 6.374334094026874e-06, "loss": 0.853, "step": 5028 }, { "epoch": 0.43, "learning_rate": 6.373005377104466e-06, "loss": 0.8515, "step": 5029 }, { "epoch": 0.43, "learning_rate": 6.37167655530506e-06, "loss": 0.8136, "step": 5030 }, { "epoch": 0.43, "learning_rate": 6.370347628730154e-06, "loss": 0.9033, "step": 5031 }, { "epoch": 0.43, "learning_rate": 6.369018597481258e-06, "loss": 0.9217, "step": 5032 }, { "epoch": 0.43, "learning_rate": 6.36768946165989e-06, "loss": 0.5212, "step": 5033 }, { "epoch": 0.43, "learning_rate": 6.366360221367578e-06, "loss": 0.8389, "step": 5034 }, { "epoch": 0.43, "learning_rate": 6.365030876705855e-06, "loss": 0.7986, "step": 5035 }, { "epoch": 0.43, "learning_rate": 6.363701427776263e-06, "loss": 0.8844, "step": 5036 }, { "epoch": 0.43, "learning_rate": 6.3623718746803535e-06, "loss": 0.8305, "step": 5037 }, { "epoch": 0.43, "learning_rate": 6.361042217519681e-06, "loss": 0.8072, "step": 5038 }, { "epoch": 0.43, "learning_rate": 6.359712456395812e-06, "loss": 0.9307, "step": 5039 }, { "epoch": 0.43, "learning_rate": 6.358382591410324e-06, "loss": 0.8946, "step": 5040 }, { "epoch": 0.43, "learning_rate": 6.3570526226647955e-06, "loss": 0.8596, "step": 5041 }, { "epoch": 0.43, "learning_rate": 6.355722550260816e-06, "loss": 0.8084, "step": 5042 }, { "epoch": 0.43, "learning_rate": 6.3543923742999845e-06, "loss": 0.8258, "step": 5043 }, { "epoch": 0.43, "learning_rate": 6.353062094883905e-06, "loss": 0.9097, "step": 5044 }, { "epoch": 0.43, "learning_rate": 6.351731712114193e-06, "loss": 0.8277, "step": 5045 }, { "epoch": 0.43, "learning_rate": 6.350401226092468e-06, "loss": 0.9237, "step": 5046 }, { "epoch": 0.43, "learning_rate": 6.3490706369203594e-06, "loss": 0.8443, "step": 5047 }, { "epoch": 0.43, "learning_rate": 6.347739944699505e-06, "loss": 0.8499, "step": 5048 }, { "epoch": 0.43, "learning_rate": 6.346409149531547e-06, "loss": 0.8525, "step": 5049 }, { "epoch": 0.43, "learning_rate": 6.345078251518144e-06, "loss": 0.8395, "step": 5050 }, { "epoch": 0.43, "learning_rate": 6.343747250760951e-06, "loss": 0.8701, "step": 5051 }, { "epoch": 0.43, "learning_rate": 6.342416147361639e-06, "loss": 0.8256, "step": 5052 }, { "epoch": 0.43, "learning_rate": 6.341084941421883e-06, "loss": 0.838, "step": 5053 }, { "epoch": 0.43, "learning_rate": 6.339753633043369e-06, "loss": 0.8682, "step": 5054 }, { "epoch": 0.43, "learning_rate": 6.338422222327788e-06, "loss": 0.8221, "step": 5055 }, { "epoch": 0.43, "learning_rate": 6.33709070937684e-06, "loss": 0.9033, "step": 5056 }, { "epoch": 0.43, "learning_rate": 6.335759094292231e-06, "loss": 0.8307, "step": 5057 }, { "epoch": 0.43, "learning_rate": 6.334427377175678e-06, "loss": 0.8506, "step": 5058 }, { "epoch": 0.43, "learning_rate": 6.333095558128905e-06, "loss": 0.838, "step": 5059 }, { "epoch": 0.43, "learning_rate": 6.331763637253642e-06, "loss": 0.8165, "step": 5060 }, { "epoch": 0.43, "learning_rate": 6.330431614651627e-06, "loss": 0.8038, "step": 5061 }, { "epoch": 0.43, "learning_rate": 6.329099490424606e-06, "loss": 0.8494, "step": 5062 }, { "epoch": 0.43, "learning_rate": 6.327767264674336e-06, "loss": 0.8315, "step": 5063 }, { "epoch": 0.43, "learning_rate": 6.326434937502578e-06, "loss": 0.8279, "step": 5064 }, { "epoch": 0.43, "learning_rate": 6.325102509011101e-06, "loss": 0.8649, "step": 5065 }, { "epoch": 0.43, "learning_rate": 6.323769979301683e-06, "loss": 0.8671, "step": 5066 }, { "epoch": 0.43, "learning_rate": 6.322437348476107e-06, "loss": 0.8181, "step": 5067 }, { "epoch": 0.43, "learning_rate": 6.32110461663617e-06, "loss": 0.8653, "step": 5068 }, { "epoch": 0.43, "learning_rate": 6.319771783883673e-06, "loss": 0.8947, "step": 5069 }, { "epoch": 0.43, "learning_rate": 6.318438850320419e-06, "loss": 0.8394, "step": 5070 }, { "epoch": 0.43, "learning_rate": 6.317105816048228e-06, "loss": 0.8366, "step": 5071 }, { "epoch": 0.43, "learning_rate": 6.315772681168924e-06, "loss": 0.8708, "step": 5072 }, { "epoch": 0.43, "learning_rate": 6.314439445784337e-06, "loss": 0.8365, "step": 5073 }, { "epoch": 0.43, "learning_rate": 6.313106109996309e-06, "loss": 0.8666, "step": 5074 }, { "epoch": 0.43, "learning_rate": 6.3117726739066824e-06, "loss": 0.8312, "step": 5075 }, { "epoch": 0.43, "learning_rate": 6.310439137617315e-06, "loss": 0.844, "step": 5076 }, { "epoch": 0.43, "learning_rate": 6.3091055012300675e-06, "loss": 0.8422, "step": 5077 }, { "epoch": 0.43, "learning_rate": 6.307771764846811e-06, "loss": 0.9102, "step": 5078 }, { "epoch": 0.43, "learning_rate": 6.306437928569421e-06, "loss": 0.8854, "step": 5079 }, { "epoch": 0.43, "learning_rate": 6.305103992499783e-06, "loss": 0.8691, "step": 5080 }, { "epoch": 0.43, "learning_rate": 6.303769956739791e-06, "loss": 0.8527, "step": 5081 }, { "epoch": 0.43, "learning_rate": 6.302435821391347e-06, "loss": 0.8386, "step": 5082 }, { "epoch": 0.43, "learning_rate": 6.301101586556355e-06, "loss": 0.8708, "step": 5083 }, { "epoch": 0.43, "learning_rate": 6.299767252336732e-06, "loss": 0.831, "step": 5084 }, { "epoch": 0.43, "learning_rate": 6.2984328188344e-06, "loss": 0.8224, "step": 5085 }, { "epoch": 0.43, "learning_rate": 6.297098286151293e-06, "loss": 0.7909, "step": 5086 }, { "epoch": 0.43, "learning_rate": 6.295763654389347e-06, "loss": 0.586, "step": 5087 }, { "epoch": 0.43, "learning_rate": 6.294428923650506e-06, "loss": 0.8649, "step": 5088 }, { "epoch": 0.43, "learning_rate": 6.293094094036728e-06, "loss": 0.8625, "step": 5089 }, { "epoch": 0.43, "learning_rate": 6.291759165649969e-06, "loss": 0.8227, "step": 5090 }, { "epoch": 0.43, "learning_rate": 6.290424138592201e-06, "loss": 0.886, "step": 5091 }, { "epoch": 0.43, "learning_rate": 6.289089012965399e-06, "loss": 0.8822, "step": 5092 }, { "epoch": 0.43, "learning_rate": 6.287753788871547e-06, "loss": 0.7759, "step": 5093 }, { "epoch": 0.43, "learning_rate": 6.286418466412634e-06, "loss": 0.8828, "step": 5094 }, { "epoch": 0.43, "learning_rate": 6.28508304569066e-06, "loss": 0.8204, "step": 5095 }, { "epoch": 0.43, "learning_rate": 6.2837475268076335e-06, "loss": 0.8394, "step": 5096 }, { "epoch": 0.43, "learning_rate": 6.2824119098655636e-06, "loss": 0.8867, "step": 5097 }, { "epoch": 0.44, "learning_rate": 6.281076194966474e-06, "loss": 0.8637, "step": 5098 }, { "epoch": 0.44, "learning_rate": 6.279740382212394e-06, "loss": 0.8739, "step": 5099 }, { "epoch": 0.44, "learning_rate": 6.278404471705356e-06, "loss": 0.8635, "step": 5100 }, { "epoch": 0.44, "learning_rate": 6.277068463547409e-06, "loss": 0.7975, "step": 5101 }, { "epoch": 0.44, "learning_rate": 6.275732357840598e-06, "loss": 0.8876, "step": 5102 }, { "epoch": 0.44, "learning_rate": 6.274396154686987e-06, "loss": 0.7933, "step": 5103 }, { "epoch": 0.44, "learning_rate": 6.273059854188636e-06, "loss": 0.8578, "step": 5104 }, { "epoch": 0.44, "learning_rate": 6.271723456447625e-06, "loss": 0.8292, "step": 5105 }, { "epoch": 0.44, "learning_rate": 6.27038696156603e-06, "loss": 0.5422, "step": 5106 }, { "epoch": 0.44, "learning_rate": 6.269050369645941e-06, "loss": 0.8588, "step": 5107 }, { "epoch": 0.44, "learning_rate": 6.2677136807894514e-06, "loss": 0.8665, "step": 5108 }, { "epoch": 0.44, "learning_rate": 6.266376895098667e-06, "loss": 0.8527, "step": 5109 }, { "epoch": 0.44, "learning_rate": 6.265040012675696e-06, "loss": 0.8187, "step": 5110 }, { "epoch": 0.44, "learning_rate": 6.263703033622659e-06, "loss": 0.8394, "step": 5111 }, { "epoch": 0.44, "learning_rate": 6.262365958041677e-06, "loss": 0.8018, "step": 5112 }, { "epoch": 0.44, "learning_rate": 6.261028786034886e-06, "loss": 0.8282, "step": 5113 }, { "epoch": 0.44, "learning_rate": 6.259691517704424e-06, "loss": 0.9264, "step": 5114 }, { "epoch": 0.44, "learning_rate": 6.25835415315244e-06, "loss": 0.8111, "step": 5115 }, { "epoch": 0.44, "learning_rate": 6.257016692481086e-06, "loss": 0.8799, "step": 5116 }, { "epoch": 0.44, "learning_rate": 6.255679135792525e-06, "loss": 0.8615, "step": 5117 }, { "epoch": 0.44, "learning_rate": 6.254341483188926e-06, "loss": 0.8677, "step": 5118 }, { "epoch": 0.44, "learning_rate": 6.253003734772468e-06, "loss": 0.8602, "step": 5119 }, { "epoch": 0.44, "learning_rate": 6.251665890645333e-06, "loss": 0.8125, "step": 5120 }, { "epoch": 0.44, "learning_rate": 6.25032795090971e-06, "loss": 0.602, "step": 5121 }, { "epoch": 0.44, "learning_rate": 6.248989915667802e-06, "loss": 0.8654, "step": 5122 }, { "epoch": 0.44, "learning_rate": 6.247651785021811e-06, "loss": 0.8081, "step": 5123 }, { "epoch": 0.44, "learning_rate": 6.246313559073952e-06, "loss": 0.8169, "step": 5124 }, { "epoch": 0.44, "learning_rate": 6.244975237926446e-06, "loss": 0.8986, "step": 5125 }, { "epoch": 0.44, "learning_rate": 6.2436368216815205e-06, "loss": 0.5346, "step": 5126 }, { "epoch": 0.44, "learning_rate": 6.242298310441407e-06, "loss": 0.8782, "step": 5127 }, { "epoch": 0.44, "learning_rate": 6.24095970430835e-06, "loss": 0.8543, "step": 5128 }, { "epoch": 0.44, "learning_rate": 6.239621003384602e-06, "loss": 0.8698, "step": 5129 }, { "epoch": 0.44, "learning_rate": 6.238282207772415e-06, "loss": 0.843, "step": 5130 }, { "epoch": 0.44, "learning_rate": 6.236943317574054e-06, "loss": 0.8371, "step": 5131 }, { "epoch": 0.44, "learning_rate": 6.235604332891793e-06, "loss": 0.8279, "step": 5132 }, { "epoch": 0.44, "learning_rate": 6.234265253827907e-06, "loss": 0.7993, "step": 5133 }, { "epoch": 0.44, "learning_rate": 6.232926080484682e-06, "loss": 0.8579, "step": 5134 }, { "epoch": 0.44, "learning_rate": 6.231586812964412e-06, "loss": 0.8706, "step": 5135 }, { "epoch": 0.44, "learning_rate": 6.2302474513693965e-06, "loss": 0.8465, "step": 5136 }, { "epoch": 0.44, "learning_rate": 6.228907995801943e-06, "loss": 0.8797, "step": 5137 }, { "epoch": 0.44, "learning_rate": 6.2275684463643635e-06, "loss": 0.8532, "step": 5138 }, { "epoch": 0.44, "learning_rate": 6.226228803158983e-06, "loss": 0.8651, "step": 5139 }, { "epoch": 0.44, "learning_rate": 6.224889066288128e-06, "loss": 0.8447, "step": 5140 }, { "epoch": 0.44, "learning_rate": 6.223549235854134e-06, "loss": 0.8397, "step": 5141 }, { "epoch": 0.44, "learning_rate": 6.222209311959345e-06, "loss": 0.9037, "step": 5142 }, { "epoch": 0.44, "learning_rate": 6.220869294706111e-06, "loss": 0.8149, "step": 5143 }, { "epoch": 0.44, "learning_rate": 6.219529184196789e-06, "loss": 0.8333, "step": 5144 }, { "epoch": 0.44, "learning_rate": 6.218188980533743e-06, "loss": 0.8282, "step": 5145 }, { "epoch": 0.44, "learning_rate": 6.216848683819344e-06, "loss": 0.8214, "step": 5146 }, { "epoch": 0.44, "learning_rate": 6.215508294155972e-06, "loss": 0.869, "step": 5147 }, { "epoch": 0.44, "learning_rate": 6.214167811646012e-06, "loss": 0.8241, "step": 5148 }, { "epoch": 0.44, "learning_rate": 6.212827236391856e-06, "loss": 0.8479, "step": 5149 }, { "epoch": 0.44, "learning_rate": 6.211486568495903e-06, "loss": 0.8206, "step": 5150 }, { "epoch": 0.44, "learning_rate": 6.210145808060564e-06, "loss": 0.7948, "step": 5151 }, { "epoch": 0.44, "learning_rate": 6.208804955188249e-06, "loss": 0.8794, "step": 5152 }, { "epoch": 0.44, "learning_rate": 6.207464009981381e-06, "loss": 0.8399, "step": 5153 }, { "epoch": 0.44, "learning_rate": 6.206122972542386e-06, "loss": 0.8574, "step": 5154 }, { "epoch": 0.44, "learning_rate": 6.204781842973702e-06, "loss": 0.893, "step": 5155 }, { "epoch": 0.44, "learning_rate": 6.203440621377769e-06, "loss": 0.82, "step": 5156 }, { "epoch": 0.44, "learning_rate": 6.202099307857037e-06, "loss": 0.8085, "step": 5157 }, { "epoch": 0.44, "learning_rate": 6.200757902513962e-06, "loss": 0.8198, "step": 5158 }, { "epoch": 0.44, "learning_rate": 6.199416405451007e-06, "loss": 0.8455, "step": 5159 }, { "epoch": 0.44, "learning_rate": 6.198074816770644e-06, "loss": 0.8356, "step": 5160 }, { "epoch": 0.44, "learning_rate": 6.1967331365753465e-06, "loss": 0.8368, "step": 5161 }, { "epoch": 0.44, "learning_rate": 6.1953913649676035e-06, "loss": 0.8101, "step": 5162 }, { "epoch": 0.44, "learning_rate": 6.1940495020499035e-06, "loss": 0.8254, "step": 5163 }, { "epoch": 0.44, "learning_rate": 6.1927075479247435e-06, "loss": 0.8277, "step": 5164 }, { "epoch": 0.44, "learning_rate": 6.191365502694631e-06, "loss": 0.8748, "step": 5165 }, { "epoch": 0.44, "learning_rate": 6.190023366462077e-06, "loss": 0.8595, "step": 5166 }, { "epoch": 0.44, "learning_rate": 6.188681139329601e-06, "loss": 0.8884, "step": 5167 }, { "epoch": 0.44, "learning_rate": 6.1873388213997285e-06, "loss": 0.8084, "step": 5168 }, { "epoch": 0.44, "learning_rate": 6.185996412774993e-06, "loss": 0.8403, "step": 5169 }, { "epoch": 0.44, "learning_rate": 6.184653913557933e-06, "loss": 0.8718, "step": 5170 }, { "epoch": 0.44, "learning_rate": 6.183311323851098e-06, "loss": 0.7968, "step": 5171 }, { "epoch": 0.44, "learning_rate": 6.18196864375704e-06, "loss": 0.8396, "step": 5172 }, { "epoch": 0.44, "learning_rate": 6.180625873378319e-06, "loss": 0.8582, "step": 5173 }, { "epoch": 0.44, "learning_rate": 6.179283012817503e-06, "loss": 0.5255, "step": 5174 }, { "epoch": 0.44, "learning_rate": 6.177940062177166e-06, "loss": 0.8615, "step": 5175 }, { "epoch": 0.44, "learning_rate": 6.176597021559892e-06, "loss": 0.7869, "step": 5176 }, { "epoch": 0.44, "learning_rate": 6.175253891068265e-06, "loss": 0.8272, "step": 5177 }, { "epoch": 0.44, "learning_rate": 6.173910670804882e-06, "loss": 0.8402, "step": 5178 }, { "epoch": 0.44, "learning_rate": 6.172567360872347e-06, "loss": 0.526, "step": 5179 }, { "epoch": 0.44, "learning_rate": 6.171223961373264e-06, "loss": 0.8699, "step": 5180 }, { "epoch": 0.44, "learning_rate": 6.169880472410252e-06, "loss": 0.8381, "step": 5181 }, { "epoch": 0.44, "learning_rate": 6.1685368940859346e-06, "loss": 0.7856, "step": 5182 }, { "epoch": 0.44, "learning_rate": 6.167193226502937e-06, "loss": 0.8516, "step": 5183 }, { "epoch": 0.44, "learning_rate": 6.165849469763896e-06, "loss": 0.8362, "step": 5184 }, { "epoch": 0.44, "learning_rate": 6.164505623971458e-06, "loss": 0.8556, "step": 5185 }, { "epoch": 0.44, "learning_rate": 6.163161689228269e-06, "loss": 0.5352, "step": 5186 }, { "epoch": 0.44, "learning_rate": 6.161817665636987e-06, "loss": 0.8761, "step": 5187 }, { "epoch": 0.44, "learning_rate": 6.160473553300275e-06, "loss": 0.8339, "step": 5188 }, { "epoch": 0.44, "learning_rate": 6.1591293523208025e-06, "loss": 0.8697, "step": 5189 }, { "epoch": 0.44, "learning_rate": 6.157785062801247e-06, "loss": 0.7694, "step": 5190 }, { "epoch": 0.44, "learning_rate": 6.156440684844292e-06, "loss": 0.7831, "step": 5191 }, { "epoch": 0.44, "learning_rate": 6.155096218552627e-06, "loss": 0.8272, "step": 5192 }, { "epoch": 0.44, "learning_rate": 6.153751664028949e-06, "loss": 0.8402, "step": 5193 }, { "epoch": 0.44, "learning_rate": 6.152407021375964e-06, "loss": 0.8706, "step": 5194 }, { "epoch": 0.44, "learning_rate": 6.15106229069638e-06, "loss": 0.8291, "step": 5195 }, { "epoch": 0.44, "learning_rate": 6.149717472092917e-06, "loss": 0.8727, "step": 5196 }, { "epoch": 0.44, "learning_rate": 6.148372565668294e-06, "loss": 0.8625, "step": 5197 }, { "epoch": 0.44, "learning_rate": 6.147027571525246e-06, "loss": 0.8528, "step": 5198 }, { "epoch": 0.44, "learning_rate": 6.14568248976651e-06, "loss": 0.8458, "step": 5199 }, { "epoch": 0.44, "learning_rate": 6.144337320494829e-06, "loss": 0.835, "step": 5200 }, { "epoch": 0.44, "learning_rate": 6.142992063812953e-06, "loss": 0.8546, "step": 5201 }, { "epoch": 0.44, "learning_rate": 6.141646719823642e-06, "loss": 0.7851, "step": 5202 }, { "epoch": 0.44, "learning_rate": 6.140301288629656e-06, "loss": 0.8038, "step": 5203 }, { "epoch": 0.44, "learning_rate": 6.13895577033377e-06, "loss": 0.8515, "step": 5204 }, { "epoch": 0.44, "learning_rate": 6.13761016503876e-06, "loss": 0.853, "step": 5205 }, { "epoch": 0.44, "learning_rate": 6.136264472847407e-06, "loss": 0.8593, "step": 5206 }, { "epoch": 0.44, "learning_rate": 6.134918693862506e-06, "loss": 0.8163, "step": 5207 }, { "epoch": 0.44, "learning_rate": 6.133572828186851e-06, "loss": 0.8147, "step": 5208 }, { "epoch": 0.44, "learning_rate": 6.132226875923252e-06, "loss": 0.8442, "step": 5209 }, { "epoch": 0.44, "learning_rate": 6.130880837174511e-06, "loss": 0.8651, "step": 5210 }, { "epoch": 0.44, "learning_rate": 6.12953471204345e-06, "loss": 0.817, "step": 5211 }, { "epoch": 0.44, "learning_rate": 6.128188500632892e-06, "loss": 0.8382, "step": 5212 }, { "epoch": 0.44, "learning_rate": 6.126842203045667e-06, "loss": 0.8783, "step": 5213 }, { "epoch": 0.44, "learning_rate": 6.125495819384615e-06, "loss": 0.7601, "step": 5214 }, { "epoch": 0.45, "learning_rate": 6.124149349752575e-06, "loss": 0.8435, "step": 5215 }, { "epoch": 0.45, "learning_rate": 6.122802794252398e-06, "loss": 0.9162, "step": 5216 }, { "epoch": 0.45, "learning_rate": 6.121456152986941e-06, "loss": 0.8625, "step": 5217 }, { "epoch": 0.45, "learning_rate": 6.12010942605907e-06, "loss": 0.8338, "step": 5218 }, { "epoch": 0.45, "learning_rate": 6.118762613571652e-06, "loss": 0.8287, "step": 5219 }, { "epoch": 0.45, "learning_rate": 6.117415715627561e-06, "loss": 0.8453, "step": 5220 }, { "epoch": 0.45, "learning_rate": 6.116068732329684e-06, "loss": 0.8083, "step": 5221 }, { "epoch": 0.45, "learning_rate": 6.114721663780909e-06, "loss": 0.836, "step": 5222 }, { "epoch": 0.45, "learning_rate": 6.113374510084132e-06, "loss": 0.8903, "step": 5223 }, { "epoch": 0.45, "learning_rate": 6.1120272713422534e-06, "loss": 0.8463, "step": 5224 }, { "epoch": 0.45, "learning_rate": 6.110679947658186e-06, "loss": 0.7993, "step": 5225 }, { "epoch": 0.45, "learning_rate": 6.1093325391348405e-06, "loss": 0.8408, "step": 5226 }, { "epoch": 0.45, "learning_rate": 6.10798504587514e-06, "loss": 0.8275, "step": 5227 }, { "epoch": 0.45, "learning_rate": 6.106637467982015e-06, "loss": 0.8347, "step": 5228 }, { "epoch": 0.45, "learning_rate": 6.1052898055584e-06, "loss": 0.8462, "step": 5229 }, { "epoch": 0.45, "learning_rate": 6.103942058707233e-06, "loss": 0.8372, "step": 5230 }, { "epoch": 0.45, "learning_rate": 6.102594227531463e-06, "loss": 0.8842, "step": 5231 }, { "epoch": 0.45, "learning_rate": 6.101246312134046e-06, "loss": 0.7919, "step": 5232 }, { "epoch": 0.45, "learning_rate": 6.0998983126179415e-06, "loss": 0.6572, "step": 5233 }, { "epoch": 0.45, "learning_rate": 6.098550229086115e-06, "loss": 0.8611, "step": 5234 }, { "epoch": 0.45, "learning_rate": 6.0972020616415415e-06, "loss": 0.8283, "step": 5235 }, { "epoch": 0.45, "learning_rate": 6.0958538103871986e-06, "loss": 0.8582, "step": 5236 }, { "epoch": 0.45, "learning_rate": 6.0945054754260765e-06, "loss": 0.8463, "step": 5237 }, { "epoch": 0.45, "learning_rate": 6.093157056861163e-06, "loss": 0.8048, "step": 5238 }, { "epoch": 0.45, "learning_rate": 6.091808554795462e-06, "loss": 0.797, "step": 5239 }, { "epoch": 0.45, "learning_rate": 6.090459969331973e-06, "loss": 0.8206, "step": 5240 }, { "epoch": 0.45, "learning_rate": 6.089111300573714e-06, "loss": 0.8454, "step": 5241 }, { "epoch": 0.45, "learning_rate": 6.087762548623698e-06, "loss": 0.8656, "step": 5242 }, { "epoch": 0.45, "learning_rate": 6.086413713584951e-06, "loss": 0.8366, "step": 5243 }, { "epoch": 0.45, "learning_rate": 6.085064795560504e-06, "loss": 0.846, "step": 5244 }, { "epoch": 0.45, "learning_rate": 6.0837157946533955e-06, "loss": 0.8137, "step": 5245 }, { "epoch": 0.45, "learning_rate": 6.0823667109666655e-06, "loss": 0.8203, "step": 5246 }, { "epoch": 0.45, "learning_rate": 6.081017544603367e-06, "loss": 0.8606, "step": 5247 }, { "epoch": 0.45, "learning_rate": 6.079668295666555e-06, "loss": 0.8198, "step": 5248 }, { "epoch": 0.45, "learning_rate": 6.078318964259291e-06, "loss": 0.8014, "step": 5249 }, { "epoch": 0.45, "learning_rate": 6.076969550484645e-06, "loss": 0.8599, "step": 5250 }, { "epoch": 0.45, "learning_rate": 6.075620054445691e-06, "loss": 0.9026, "step": 5251 }, { "epoch": 0.45, "learning_rate": 6.074270476245512e-06, "loss": 0.832, "step": 5252 }, { "epoch": 0.45, "learning_rate": 6.072920815987191e-06, "loss": 0.8592, "step": 5253 }, { "epoch": 0.45, "learning_rate": 6.0715710737738264e-06, "loss": 0.836, "step": 5254 }, { "epoch": 0.45, "learning_rate": 6.070221249708518e-06, "loss": 0.478, "step": 5255 }, { "epoch": 0.45, "learning_rate": 6.068871343894369e-06, "loss": 0.8754, "step": 5256 }, { "epoch": 0.45, "learning_rate": 6.067521356434495e-06, "loss": 0.8418, "step": 5257 }, { "epoch": 0.45, "learning_rate": 6.066171287432013e-06, "loss": 0.8324, "step": 5258 }, { "epoch": 0.45, "learning_rate": 6.064821136990048e-06, "loss": 0.5922, "step": 5259 }, { "epoch": 0.45, "learning_rate": 6.063470905211732e-06, "loss": 0.5828, "step": 5260 }, { "epoch": 0.45, "learning_rate": 6.062120592200202e-06, "loss": 0.8493, "step": 5261 }, { "epoch": 0.45, "learning_rate": 6.060770198058603e-06, "loss": 0.8539, "step": 5262 }, { "epoch": 0.45, "learning_rate": 6.059419722890082e-06, "loss": 0.8809, "step": 5263 }, { "epoch": 0.45, "learning_rate": 6.058069166797797e-06, "loss": 0.855, "step": 5264 }, { "epoch": 0.45, "learning_rate": 6.056718529884913e-06, "loss": 0.8323, "step": 5265 }, { "epoch": 0.45, "learning_rate": 6.055367812254592e-06, "loss": 0.8251, "step": 5266 }, { "epoch": 0.45, "learning_rate": 6.054017014010012e-06, "loss": 0.8786, "step": 5267 }, { "epoch": 0.45, "learning_rate": 6.052666135254355e-06, "loss": 0.8314, "step": 5268 }, { "epoch": 0.45, "learning_rate": 6.051315176090805e-06, "loss": 0.7905, "step": 5269 }, { "epoch": 0.45, "learning_rate": 6.049964136622557e-06, "loss": 0.8112, "step": 5270 }, { "epoch": 0.45, "learning_rate": 6.048613016952809e-06, "loss": 0.8906, "step": 5271 }, { "epoch": 0.45, "learning_rate": 6.047261817184769e-06, "loss": 0.8514, "step": 5272 }, { "epoch": 0.45, "learning_rate": 6.045910537421644e-06, "loss": 0.8822, "step": 5273 }, { "epoch": 0.45, "learning_rate": 6.044559177766655e-06, "loss": 0.8411, "step": 5274 }, { "epoch": 0.45, "learning_rate": 6.0432077383230236e-06, "loss": 0.8082, "step": 5275 }, { "epoch": 0.45, "learning_rate": 6.04185621919398e-06, "loss": 0.9186, "step": 5276 }, { "epoch": 0.45, "learning_rate": 6.04050462048276e-06, "loss": 0.8342, "step": 5277 }, { "epoch": 0.45, "learning_rate": 6.039152942292606e-06, "loss": 0.6241, "step": 5278 }, { "epoch": 0.45, "learning_rate": 6.037801184726766e-06, "loss": 0.8603, "step": 5279 }, { "epoch": 0.45, "learning_rate": 6.036449347888493e-06, "loss": 0.8202, "step": 5280 }, { "epoch": 0.45, "learning_rate": 6.035097431881047e-06, "loss": 0.8611, "step": 5281 }, { "epoch": 0.45, "learning_rate": 6.033745436807696e-06, "loss": 0.867, "step": 5282 }, { "epoch": 0.45, "learning_rate": 6.032393362771709e-06, "loss": 0.8634, "step": 5283 }, { "epoch": 0.45, "learning_rate": 6.0310412098763685e-06, "loss": 0.8305, "step": 5284 }, { "epoch": 0.45, "learning_rate": 6.029688978224955e-06, "loss": 0.8944, "step": 5285 }, { "epoch": 0.45, "learning_rate": 6.0283366679207575e-06, "loss": 0.8592, "step": 5286 }, { "epoch": 0.45, "learning_rate": 6.026984279067076e-06, "loss": 0.8055, "step": 5287 }, { "epoch": 0.45, "learning_rate": 6.025631811767213e-06, "loss": 0.8253, "step": 5288 }, { "epoch": 0.45, "learning_rate": 6.024279266124474e-06, "loss": 0.8242, "step": 5289 }, { "epoch": 0.45, "learning_rate": 6.022926642242173e-06, "loss": 0.835, "step": 5290 }, { "epoch": 0.45, "learning_rate": 6.021573940223632e-06, "loss": 0.636, "step": 5291 }, { "epoch": 0.45, "learning_rate": 6.020221160172177e-06, "loss": 0.8072, "step": 5292 }, { "epoch": 0.45, "learning_rate": 6.0188683021911394e-06, "loss": 0.8218, "step": 5293 }, { "epoch": 0.45, "learning_rate": 6.017515366383856e-06, "loss": 0.6349, "step": 5294 }, { "epoch": 0.45, "learning_rate": 6.016162352853674e-06, "loss": 0.6105, "step": 5295 }, { "epoch": 0.45, "learning_rate": 6.01480926170394e-06, "loss": 0.5786, "step": 5296 }, { "epoch": 0.45, "learning_rate": 6.013456093038011e-06, "loss": 0.7975, "step": 5297 }, { "epoch": 0.45, "learning_rate": 6.012102846959251e-06, "loss": 0.8752, "step": 5298 }, { "epoch": 0.45, "learning_rate": 6.010749523571025e-06, "loss": 0.8561, "step": 5299 }, { "epoch": 0.45, "learning_rate": 6.009396122976706e-06, "loss": 0.8429, "step": 5300 }, { "epoch": 0.45, "learning_rate": 6.008042645279676e-06, "loss": 0.8679, "step": 5301 }, { "epoch": 0.45, "learning_rate": 6.00668909058332e-06, "loss": 0.8238, "step": 5302 }, { "epoch": 0.45, "learning_rate": 6.005335458991027e-06, "loss": 0.8471, "step": 5303 }, { "epoch": 0.45, "learning_rate": 6.003981750606196e-06, "loss": 0.8371, "step": 5304 }, { "epoch": 0.45, "learning_rate": 6.00262796553223e-06, "loss": 0.5539, "step": 5305 }, { "epoch": 0.45, "learning_rate": 6.001274103872537e-06, "loss": 0.8456, "step": 5306 }, { "epoch": 0.45, "learning_rate": 5.999920165730534e-06, "loss": 0.8951, "step": 5307 }, { "epoch": 0.45, "learning_rate": 5.998566151209638e-06, "loss": 0.8514, "step": 5308 }, { "epoch": 0.45, "learning_rate": 5.997212060413278e-06, "loss": 0.8446, "step": 5309 }, { "epoch": 0.45, "learning_rate": 5.995857893444884e-06, "loss": 0.8479, "step": 5310 }, { "epoch": 0.45, "learning_rate": 5.994503650407897e-06, "loss": 0.5324, "step": 5311 }, { "epoch": 0.45, "learning_rate": 5.993149331405759e-06, "loss": 0.8306, "step": 5312 }, { "epoch": 0.45, "learning_rate": 5.991794936541918e-06, "loss": 0.893, "step": 5313 }, { "epoch": 0.45, "learning_rate": 5.990440465919832e-06, "loss": 0.8521, "step": 5314 }, { "epoch": 0.45, "learning_rate": 5.989085919642963e-06, "loss": 0.8556, "step": 5315 }, { "epoch": 0.45, "learning_rate": 5.987731297814776e-06, "loss": 0.58, "step": 5316 }, { "epoch": 0.45, "learning_rate": 5.986376600538743e-06, "loss": 0.8562, "step": 5317 }, { "epoch": 0.45, "learning_rate": 5.9850218279183455e-06, "loss": 0.8572, "step": 5318 }, { "epoch": 0.45, "learning_rate": 5.9836669800570655e-06, "loss": 0.8145, "step": 5319 }, { "epoch": 0.45, "learning_rate": 5.982312057058392e-06, "loss": 0.8691, "step": 5320 }, { "epoch": 0.45, "learning_rate": 5.9809570590258245e-06, "loss": 0.802, "step": 5321 }, { "epoch": 0.45, "learning_rate": 5.979601986062863e-06, "loss": 0.8685, "step": 5322 }, { "epoch": 0.45, "learning_rate": 5.978246838273013e-06, "loss": 0.8549, "step": 5323 }, { "epoch": 0.45, "learning_rate": 5.976891615759788e-06, "loss": 0.8592, "step": 5324 }, { "epoch": 0.45, "learning_rate": 5.975536318626709e-06, "loss": 0.8427, "step": 5325 }, { "epoch": 0.45, "learning_rate": 5.974180946977299e-06, "loss": 0.8297, "step": 5326 }, { "epoch": 0.45, "learning_rate": 5.972825500915085e-06, "loss": 0.8235, "step": 5327 }, { "epoch": 0.45, "learning_rate": 5.971469980543607e-06, "loss": 0.8794, "step": 5328 }, { "epoch": 0.45, "learning_rate": 5.970114385966404e-06, "loss": 0.8611, "step": 5329 }, { "epoch": 0.45, "learning_rate": 5.968758717287025e-06, "loss": 0.588, "step": 5330 }, { "epoch": 0.45, "learning_rate": 5.9674029746090224e-06, "loss": 0.8129, "step": 5331 }, { "epoch": 0.45, "learning_rate": 5.966047158035953e-06, "loss": 0.807, "step": 5332 }, { "epoch": 0.46, "learning_rate": 5.96469126767138e-06, "loss": 0.8222, "step": 5333 }, { "epoch": 0.46, "learning_rate": 5.963335303618875e-06, "loss": 0.8492, "step": 5334 }, { "epoch": 0.46, "learning_rate": 5.9619792659820164e-06, "loss": 0.6046, "step": 5335 }, { "epoch": 0.46, "learning_rate": 5.9606231548643786e-06, "loss": 0.8298, "step": 5336 }, { "epoch": 0.46, "learning_rate": 5.959266970369553e-06, "loss": 0.8788, "step": 5337 }, { "epoch": 0.46, "learning_rate": 5.957910712601129e-06, "loss": 0.8231, "step": 5338 }, { "epoch": 0.46, "learning_rate": 5.956554381662706e-06, "loss": 0.8173, "step": 5339 }, { "epoch": 0.46, "learning_rate": 5.955197977657886e-06, "loss": 0.8202, "step": 5340 }, { "epoch": 0.46, "learning_rate": 5.953841500690281e-06, "loss": 0.8, "step": 5341 }, { "epoch": 0.46, "learning_rate": 5.952484950863499e-06, "loss": 0.8281, "step": 5342 }, { "epoch": 0.46, "learning_rate": 5.951128328281166e-06, "loss": 0.7933, "step": 5343 }, { "epoch": 0.46, "learning_rate": 5.949771633046907e-06, "loss": 0.7981, "step": 5344 }, { "epoch": 0.46, "learning_rate": 5.948414865264351e-06, "loss": 0.8037, "step": 5345 }, { "epoch": 0.46, "learning_rate": 5.9470580250371355e-06, "loss": 0.8364, "step": 5346 }, { "epoch": 0.46, "learning_rate": 5.9457011124689025e-06, "loss": 0.8444, "step": 5347 }, { "epoch": 0.46, "learning_rate": 5.944344127663301e-06, "loss": 0.8709, "step": 5348 }, { "epoch": 0.46, "learning_rate": 5.942987070723982e-06, "loss": 0.8638, "step": 5349 }, { "epoch": 0.46, "learning_rate": 5.9416299417546065e-06, "loss": 0.8657, "step": 5350 }, { "epoch": 0.46, "learning_rate": 5.9402727408588375e-06, "loss": 0.8321, "step": 5351 }, { "epoch": 0.46, "learning_rate": 5.938915468140345e-06, "loss": 0.8479, "step": 5352 }, { "epoch": 0.46, "learning_rate": 5.937558123702805e-06, "loss": 0.8651, "step": 5353 }, { "epoch": 0.46, "learning_rate": 5.936200707649897e-06, "loss": 0.8253, "step": 5354 }, { "epoch": 0.46, "learning_rate": 5.934843220085309e-06, "loss": 0.8393, "step": 5355 }, { "epoch": 0.46, "learning_rate": 5.933485661112731e-06, "loss": 0.8093, "step": 5356 }, { "epoch": 0.46, "learning_rate": 5.9321280308358595e-06, "loss": 0.7826, "step": 5357 }, { "epoch": 0.46, "learning_rate": 5.930770329358401e-06, "loss": 0.8498, "step": 5358 }, { "epoch": 0.46, "learning_rate": 5.929412556784059e-06, "loss": 0.8572, "step": 5359 }, { "epoch": 0.46, "learning_rate": 5.928054713216548e-06, "loss": 0.8867, "step": 5360 }, { "epoch": 0.46, "learning_rate": 5.9266967987595895e-06, "loss": 0.8259, "step": 5361 }, { "epoch": 0.46, "learning_rate": 5.9253388135169035e-06, "loss": 0.7901, "step": 5362 }, { "epoch": 0.46, "learning_rate": 5.923980757592223e-06, "loss": 0.8023, "step": 5363 }, { "epoch": 0.46, "learning_rate": 5.9226226310892835e-06, "loss": 0.8102, "step": 5364 }, { "epoch": 0.46, "learning_rate": 5.921264434111823e-06, "loss": 0.8501, "step": 5365 }, { "epoch": 0.46, "learning_rate": 5.919906166763588e-06, "loss": 0.8667, "step": 5366 }, { "epoch": 0.46, "learning_rate": 5.91854782914833e-06, "loss": 0.863, "step": 5367 }, { "epoch": 0.46, "learning_rate": 5.917189421369808e-06, "loss": 0.8446, "step": 5368 }, { "epoch": 0.46, "learning_rate": 5.9158309435317794e-06, "loss": 0.8259, "step": 5369 }, { "epoch": 0.46, "learning_rate": 5.9144723957380145e-06, "loss": 0.8287, "step": 5370 }, { "epoch": 0.46, "learning_rate": 5.913113778092287e-06, "loss": 0.8421, "step": 5371 }, { "epoch": 0.46, "learning_rate": 5.911755090698371e-06, "loss": 0.8615, "step": 5372 }, { "epoch": 0.46, "learning_rate": 5.910396333660053e-06, "loss": 0.8757, "step": 5373 }, { "epoch": 0.46, "learning_rate": 5.9090375070811215e-06, "loss": 0.8058, "step": 5374 }, { "epoch": 0.46, "learning_rate": 5.907678611065369e-06, "loss": 0.8872, "step": 5375 }, { "epoch": 0.46, "learning_rate": 5.906319645716594e-06, "loss": 0.8113, "step": 5376 }, { "epoch": 0.46, "learning_rate": 5.904960611138605e-06, "loss": 0.8762, "step": 5377 }, { "epoch": 0.46, "learning_rate": 5.903601507435208e-06, "loss": 0.8303, "step": 5378 }, { "epoch": 0.46, "learning_rate": 5.90224233471022e-06, "loss": 0.8081, "step": 5379 }, { "epoch": 0.46, "learning_rate": 5.900883093067459e-06, "loss": 0.8345, "step": 5380 }, { "epoch": 0.46, "learning_rate": 5.899523782610754e-06, "loss": 0.8408, "step": 5381 }, { "epoch": 0.46, "learning_rate": 5.898164403443935e-06, "loss": 0.8642, "step": 5382 }, { "epoch": 0.46, "learning_rate": 5.896804955670836e-06, "loss": 0.8357, "step": 5383 }, { "epoch": 0.46, "learning_rate": 5.895445439395302e-06, "loss": 0.8782, "step": 5384 }, { "epoch": 0.46, "learning_rate": 5.8940858547211755e-06, "loss": 0.8232, "step": 5385 }, { "epoch": 0.46, "learning_rate": 5.892726201752313e-06, "loss": 0.7907, "step": 5386 }, { "epoch": 0.46, "learning_rate": 5.891366480592566e-06, "loss": 0.8181, "step": 5387 }, { "epoch": 0.46, "learning_rate": 5.890006691345802e-06, "loss": 0.835, "step": 5388 }, { "epoch": 0.46, "learning_rate": 5.888646834115885e-06, "loss": 0.8611, "step": 5389 }, { "epoch": 0.46, "learning_rate": 5.887286909006688e-06, "loss": 0.8439, "step": 5390 }, { "epoch": 0.46, "learning_rate": 5.8859269161220935e-06, "loss": 0.8688, "step": 5391 }, { "epoch": 0.46, "learning_rate": 5.88456685556598e-06, "loss": 0.8468, "step": 5392 }, { "epoch": 0.46, "learning_rate": 5.883206727442234e-06, "loss": 0.8251, "step": 5393 }, { "epoch": 0.46, "learning_rate": 5.881846531854753e-06, "loss": 0.8252, "step": 5394 }, { "epoch": 0.46, "learning_rate": 5.880486268907435e-06, "loss": 0.8474, "step": 5395 }, { "epoch": 0.46, "learning_rate": 5.879125938704182e-06, "loss": 0.8829, "step": 5396 }, { "epoch": 0.46, "learning_rate": 5.877765541348902e-06, "loss": 0.8148, "step": 5397 }, { "epoch": 0.46, "learning_rate": 5.876405076945513e-06, "loss": 0.8284, "step": 5398 }, { "epoch": 0.46, "learning_rate": 5.875044545597929e-06, "loss": 0.8503, "step": 5399 }, { "epoch": 0.46, "learning_rate": 5.873683947410078e-06, "loss": 0.8634, "step": 5400 }, { "epoch": 0.46, "learning_rate": 5.872323282485889e-06, "loss": 0.8415, "step": 5401 }, { "epoch": 0.46, "learning_rate": 5.8709625509292935e-06, "loss": 0.8487, "step": 5402 }, { "epoch": 0.46, "learning_rate": 5.869601752844233e-06, "loss": 0.8521, "step": 5403 }, { "epoch": 0.46, "learning_rate": 5.8682408883346535e-06, "loss": 0.8077, "step": 5404 }, { "epoch": 0.46, "learning_rate": 5.866879957504501e-06, "loss": 0.8473, "step": 5405 }, { "epoch": 0.46, "learning_rate": 5.8655189604577325e-06, "loss": 0.8246, "step": 5406 }, { "epoch": 0.46, "learning_rate": 5.864157897298307e-06, "loss": 0.8187, "step": 5407 }, { "epoch": 0.46, "learning_rate": 5.862796768130191e-06, "loss": 0.8459, "step": 5408 }, { "epoch": 0.46, "learning_rate": 5.861435573057352e-06, "loss": 0.8329, "step": 5409 }, { "epoch": 0.46, "learning_rate": 5.860074312183766e-06, "loss": 0.8528, "step": 5410 }, { "epoch": 0.46, "learning_rate": 5.858712985613415e-06, "loss": 0.8877, "step": 5411 }, { "epoch": 0.46, "learning_rate": 5.8573515934502786e-06, "loss": 0.8334, "step": 5412 }, { "epoch": 0.46, "learning_rate": 5.85599013579835e-06, "loss": 0.861, "step": 5413 }, { "epoch": 0.46, "learning_rate": 5.854628612761627e-06, "loss": 0.8247, "step": 5414 }, { "epoch": 0.46, "learning_rate": 5.853267024444105e-06, "loss": 0.8705, "step": 5415 }, { "epoch": 0.46, "learning_rate": 5.851905370949789e-06, "loss": 0.8529, "step": 5416 }, { "epoch": 0.46, "learning_rate": 5.850543652382691e-06, "loss": 0.8158, "step": 5417 }, { "epoch": 0.46, "learning_rate": 5.849181868846825e-06, "loss": 0.7876, "step": 5418 }, { "epoch": 0.46, "learning_rate": 5.847820020446212e-06, "loss": 0.8379, "step": 5419 }, { "epoch": 0.46, "learning_rate": 5.8464581072848746e-06, "loss": 0.8575, "step": 5420 }, { "epoch": 0.46, "learning_rate": 5.845096129466845e-06, "loss": 0.8531, "step": 5421 }, { "epoch": 0.46, "learning_rate": 5.8437340870961545e-06, "loss": 0.8096, "step": 5422 }, { "epoch": 0.46, "learning_rate": 5.842371980276846e-06, "loss": 0.8301, "step": 5423 }, { "epoch": 0.46, "learning_rate": 5.841009809112963e-06, "loss": 0.8501, "step": 5424 }, { "epoch": 0.46, "learning_rate": 5.8396475737085545e-06, "loss": 0.8539, "step": 5425 }, { "epoch": 0.46, "learning_rate": 5.838285274167675e-06, "loss": 0.808, "step": 5426 }, { "epoch": 0.46, "learning_rate": 5.836922910594381e-06, "loss": 0.8198, "step": 5427 }, { "epoch": 0.46, "learning_rate": 5.835560483092743e-06, "loss": 0.8764, "step": 5428 }, { "epoch": 0.46, "learning_rate": 5.834197991766826e-06, "loss": 0.8837, "step": 5429 }, { "epoch": 0.46, "learning_rate": 5.832835436720703e-06, "loss": 0.799, "step": 5430 }, { "epoch": 0.46, "learning_rate": 5.831472818058455e-06, "loss": 0.8692, "step": 5431 }, { "epoch": 0.46, "learning_rate": 5.830110135884163e-06, "loss": 0.7781, "step": 5432 }, { "epoch": 0.46, "learning_rate": 5.828747390301919e-06, "loss": 0.8835, "step": 5433 }, { "epoch": 0.46, "learning_rate": 5.827384581415814e-06, "loss": 0.8677, "step": 5434 }, { "epoch": 0.46, "learning_rate": 5.826021709329945e-06, "loss": 0.8536, "step": 5435 }, { "epoch": 0.46, "learning_rate": 5.824658774148417e-06, "loss": 0.8297, "step": 5436 }, { "epoch": 0.46, "learning_rate": 5.823295775975338e-06, "loss": 0.7977, "step": 5437 }, { "epoch": 0.46, "learning_rate": 5.821932714914819e-06, "loss": 0.8655, "step": 5438 }, { "epoch": 0.46, "learning_rate": 5.820569591070979e-06, "loss": 0.8567, "step": 5439 }, { "epoch": 0.46, "learning_rate": 5.819206404547939e-06, "loss": 0.8294, "step": 5440 }, { "epoch": 0.46, "learning_rate": 5.817843155449828e-06, "loss": 0.8548, "step": 5441 }, { "epoch": 0.46, "learning_rate": 5.8164798438807745e-06, "loss": 0.9069, "step": 5442 }, { "epoch": 0.46, "learning_rate": 5.815116469944918e-06, "loss": 0.8413, "step": 5443 }, { "epoch": 0.46, "learning_rate": 5.813753033746399e-06, "loss": 0.8455, "step": 5444 }, { "epoch": 0.46, "learning_rate": 5.812389535389361e-06, "loss": 0.8441, "step": 5445 }, { "epoch": 0.46, "learning_rate": 5.8110259749779575e-06, "loss": 0.8223, "step": 5446 }, { "epoch": 0.46, "learning_rate": 5.8096623526163455e-06, "loss": 0.7797, "step": 5447 }, { "epoch": 0.46, "learning_rate": 5.808298668408682e-06, "loss": 0.8852, "step": 5448 }, { "epoch": 0.46, "learning_rate": 5.806934922459133e-06, "loss": 0.8752, "step": 5449 }, { "epoch": 0.47, "learning_rate": 5.805571114871867e-06, "loss": 0.8103, "step": 5450 }, { "epoch": 0.47, "learning_rate": 5.8042072457510634e-06, "loss": 0.8235, "step": 5451 }, { "epoch": 0.47, "learning_rate": 5.802843315200897e-06, "loss": 0.918, "step": 5452 }, { "epoch": 0.47, "learning_rate": 5.80147932332555e-06, "loss": 0.8723, "step": 5453 }, { "epoch": 0.47, "learning_rate": 5.800115270229216e-06, "loss": 0.8367, "step": 5454 }, { "epoch": 0.47, "learning_rate": 5.798751156016085e-06, "loss": 0.865, "step": 5455 }, { "epoch": 0.47, "learning_rate": 5.797386980790355e-06, "loss": 0.861, "step": 5456 }, { "epoch": 0.47, "learning_rate": 5.796022744656229e-06, "loss": 0.804, "step": 5457 }, { "epoch": 0.47, "learning_rate": 5.794658447717913e-06, "loss": 0.8103, "step": 5458 }, { "epoch": 0.47, "learning_rate": 5.793294090079621e-06, "loss": 0.8384, "step": 5459 }, { "epoch": 0.47, "learning_rate": 5.791929671845568e-06, "loss": 0.8366, "step": 5460 }, { "epoch": 0.47, "learning_rate": 5.790565193119976e-06, "loss": 0.8473, "step": 5461 }, { "epoch": 0.47, "learning_rate": 5.78920065400707e-06, "loss": 0.8556, "step": 5462 }, { "epoch": 0.47, "learning_rate": 5.7878360546110805e-06, "loss": 0.8418, "step": 5463 }, { "epoch": 0.47, "learning_rate": 5.786471395036243e-06, "loss": 0.6849, "step": 5464 }, { "epoch": 0.47, "learning_rate": 5.785106675386795e-06, "loss": 0.8775, "step": 5465 }, { "epoch": 0.47, "learning_rate": 5.783741895766982e-06, "loss": 0.8336, "step": 5466 }, { "epoch": 0.47, "learning_rate": 5.782377056281054e-06, "loss": 0.8302, "step": 5467 }, { "epoch": 0.47, "learning_rate": 5.7810121570332615e-06, "loss": 0.7655, "step": 5468 }, { "epoch": 0.47, "learning_rate": 5.779647198127862e-06, "loss": 0.8135, "step": 5469 }, { "epoch": 0.47, "learning_rate": 5.778282179669122e-06, "loss": 0.8522, "step": 5470 }, { "epoch": 0.47, "learning_rate": 5.776917101761307e-06, "loss": 0.8728, "step": 5471 }, { "epoch": 0.47, "learning_rate": 5.775551964508686e-06, "loss": 0.8505, "step": 5472 }, { "epoch": 0.47, "learning_rate": 5.774186768015535e-06, "loss": 0.8712, "step": 5473 }, { "epoch": 0.47, "learning_rate": 5.772821512386138e-06, "loss": 0.8678, "step": 5474 }, { "epoch": 0.47, "learning_rate": 5.771456197724777e-06, "loss": 0.5683, "step": 5475 }, { "epoch": 0.47, "learning_rate": 5.770090824135744e-06, "loss": 0.7768, "step": 5476 }, { "epoch": 0.47, "learning_rate": 5.7687253917233295e-06, "loss": 0.8323, "step": 5477 }, { "epoch": 0.47, "learning_rate": 5.7673599005918345e-06, "loss": 0.8518, "step": 5478 }, { "epoch": 0.47, "learning_rate": 5.765994350845562e-06, "loss": 0.8524, "step": 5479 }, { "epoch": 0.47, "learning_rate": 5.764628742588819e-06, "loss": 0.8713, "step": 5480 }, { "epoch": 0.47, "learning_rate": 5.763263075925919e-06, "loss": 0.8511, "step": 5481 }, { "epoch": 0.47, "learning_rate": 5.7618973509611755e-06, "loss": 0.8337, "step": 5482 }, { "epoch": 0.47, "learning_rate": 5.7605315677989105e-06, "loss": 0.8446, "step": 5483 }, { "epoch": 0.47, "learning_rate": 5.759165726543451e-06, "loss": 0.8285, "step": 5484 }, { "epoch": 0.47, "learning_rate": 5.757799827299125e-06, "loss": 0.792, "step": 5485 }, { "epoch": 0.47, "learning_rate": 5.756433870170267e-06, "loss": 0.8444, "step": 5486 }, { "epoch": 0.47, "learning_rate": 5.7550678552612175e-06, "loss": 0.814, "step": 5487 }, { "epoch": 0.47, "learning_rate": 5.7537017826763155e-06, "loss": 0.8155, "step": 5488 }, { "epoch": 0.47, "learning_rate": 5.752335652519912e-06, "loss": 0.8181, "step": 5489 }, { "epoch": 0.47, "learning_rate": 5.750969464896358e-06, "loss": 0.8812, "step": 5490 }, { "epoch": 0.47, "learning_rate": 5.74960321991001e-06, "loss": 0.8468, "step": 5491 }, { "epoch": 0.47, "learning_rate": 5.748236917665225e-06, "loss": 0.8231, "step": 5492 }, { "epoch": 0.47, "learning_rate": 5.746870558266373e-06, "loss": 0.8626, "step": 5493 }, { "epoch": 0.47, "learning_rate": 5.745504141817824e-06, "loss": 0.8309, "step": 5494 }, { "epoch": 0.47, "learning_rate": 5.744137668423945e-06, "loss": 0.8699, "step": 5495 }, { "epoch": 0.47, "learning_rate": 5.742771138189118e-06, "loss": 0.8712, "step": 5496 }, { "epoch": 0.47, "learning_rate": 5.741404551217729e-06, "loss": 0.8499, "step": 5497 }, { "epoch": 0.47, "learning_rate": 5.740037907614158e-06, "loss": 0.8327, "step": 5498 }, { "epoch": 0.47, "learning_rate": 5.738671207482801e-06, "loss": 0.8544, "step": 5499 }, { "epoch": 0.47, "learning_rate": 5.737304450928052e-06, "loss": 0.8404, "step": 5500 }, { "epoch": 0.47, "learning_rate": 5.735937638054308e-06, "loss": 0.8343, "step": 5501 }, { "epoch": 0.47, "learning_rate": 5.734570768965977e-06, "loss": 0.8683, "step": 5502 }, { "epoch": 0.47, "learning_rate": 5.733203843767465e-06, "loss": 0.8248, "step": 5503 }, { "epoch": 0.47, "learning_rate": 5.731836862563186e-06, "loss": 0.8314, "step": 5504 }, { "epoch": 0.47, "learning_rate": 5.730469825457555e-06, "loss": 0.8164, "step": 5505 }, { "epoch": 0.47, "learning_rate": 5.729102732554994e-06, "loss": 0.8532, "step": 5506 }, { "epoch": 0.47, "learning_rate": 5.72773558395993e-06, "loss": 0.8734, "step": 5507 }, { "epoch": 0.47, "learning_rate": 5.726368379776791e-06, "loss": 0.8692, "step": 5508 }, { "epoch": 0.47, "learning_rate": 5.72500112011001e-06, "loss": 0.8296, "step": 5509 }, { "epoch": 0.47, "learning_rate": 5.723633805064026e-06, "loss": 0.5315, "step": 5510 }, { "epoch": 0.47, "learning_rate": 5.722266434743282e-06, "loss": 0.8032, "step": 5511 }, { "epoch": 0.47, "learning_rate": 5.720899009252223e-06, "loss": 0.8202, "step": 5512 }, { "epoch": 0.47, "learning_rate": 5.719531528695303e-06, "loss": 0.8406, "step": 5513 }, { "epoch": 0.47, "learning_rate": 5.718163993176974e-06, "loss": 0.8301, "step": 5514 }, { "epoch": 0.47, "learning_rate": 5.716796402801695e-06, "loss": 0.829, "step": 5515 }, { "epoch": 0.47, "learning_rate": 5.715428757673931e-06, "loss": 0.8326, "step": 5516 }, { "epoch": 0.47, "learning_rate": 5.714061057898149e-06, "loss": 0.8642, "step": 5517 }, { "epoch": 0.47, "learning_rate": 5.712693303578821e-06, "loss": 0.8572, "step": 5518 }, { "epoch": 0.47, "learning_rate": 5.711325494820422e-06, "loss": 0.8389, "step": 5519 }, { "epoch": 0.47, "learning_rate": 5.709957631727433e-06, "loss": 0.8424, "step": 5520 }, { "epoch": 0.47, "learning_rate": 5.7085897144043376e-06, "loss": 0.8693, "step": 5521 }, { "epoch": 0.47, "learning_rate": 5.707221742955624e-06, "loss": 0.8317, "step": 5522 }, { "epoch": 0.47, "learning_rate": 5.705853717485786e-06, "loss": 0.7902, "step": 5523 }, { "epoch": 0.47, "learning_rate": 5.704485638099319e-06, "loss": 0.8171, "step": 5524 }, { "epoch": 0.47, "learning_rate": 5.7031175049007235e-06, "loss": 0.8637, "step": 5525 }, { "epoch": 0.47, "learning_rate": 5.701749317994505e-06, "loss": 0.8484, "step": 5526 }, { "epoch": 0.47, "learning_rate": 5.700381077485173e-06, "loss": 0.8156, "step": 5527 }, { "epoch": 0.47, "learning_rate": 5.699012783477239e-06, "loss": 0.8563, "step": 5528 }, { "epoch": 0.47, "learning_rate": 5.69764443607522e-06, "loss": 0.8408, "step": 5529 }, { "epoch": 0.47, "learning_rate": 5.696276035383641e-06, "loss": 0.8738, "step": 5530 }, { "epoch": 0.47, "learning_rate": 5.694907581507022e-06, "loss": 0.8099, "step": 5531 }, { "epoch": 0.47, "learning_rate": 5.693539074549896e-06, "loss": 0.8273, "step": 5532 }, { "epoch": 0.47, "learning_rate": 5.692170514616794e-06, "loss": 0.8198, "step": 5533 }, { "epoch": 0.47, "learning_rate": 5.690801901812256e-06, "loss": 0.8804, "step": 5534 }, { "epoch": 0.47, "learning_rate": 5.689433236240821e-06, "loss": 0.8477, "step": 5535 }, { "epoch": 0.47, "learning_rate": 5.688064518007036e-06, "loss": 0.8814, "step": 5536 }, { "epoch": 0.47, "learning_rate": 5.6866957472154505e-06, "loss": 0.8936, "step": 5537 }, { "epoch": 0.47, "learning_rate": 5.685326923970617e-06, "loss": 0.8537, "step": 5538 }, { "epoch": 0.47, "learning_rate": 5.683958048377093e-06, "loss": 0.8078, "step": 5539 }, { "epoch": 0.47, "learning_rate": 5.682589120539442e-06, "loss": 0.8777, "step": 5540 }, { "epoch": 0.47, "learning_rate": 5.681220140562228e-06, "loss": 0.8378, "step": 5541 }, { "epoch": 0.47, "learning_rate": 5.6798511085500206e-06, "loss": 0.805, "step": 5542 }, { "epoch": 0.47, "learning_rate": 5.678482024607392e-06, "loss": 0.8389, "step": 5543 }, { "epoch": 0.47, "learning_rate": 5.677112888838924e-06, "loss": 0.8428, "step": 5544 }, { "epoch": 0.47, "learning_rate": 5.675743701349194e-06, "loss": 0.895, "step": 5545 }, { "epoch": 0.47, "learning_rate": 5.674374462242789e-06, "loss": 0.8631, "step": 5546 }, { "epoch": 0.47, "learning_rate": 5.673005171624297e-06, "loss": 0.88, "step": 5547 }, { "epoch": 0.47, "learning_rate": 5.671635829598312e-06, "loss": 0.8641, "step": 5548 }, { "epoch": 0.47, "learning_rate": 5.6702664362694305e-06, "loss": 0.8452, "step": 5549 }, { "epoch": 0.47, "learning_rate": 5.6688969917422565e-06, "loss": 0.8432, "step": 5550 }, { "epoch": 0.47, "learning_rate": 5.667527496121392e-06, "loss": 0.8425, "step": 5551 }, { "epoch": 0.47, "learning_rate": 5.666157949511446e-06, "loss": 0.8487, "step": 5552 }, { "epoch": 0.47, "learning_rate": 5.664788352017032e-06, "loss": 0.8631, "step": 5553 }, { "epoch": 0.47, "learning_rate": 5.663418703742769e-06, "loss": 0.8644, "step": 5554 }, { "epoch": 0.47, "learning_rate": 5.662049004793274e-06, "loss": 0.8916, "step": 5555 }, { "epoch": 0.47, "learning_rate": 5.660679255273173e-06, "loss": 0.8507, "step": 5556 }, { "epoch": 0.47, "learning_rate": 5.659309455287093e-06, "loss": 0.7891, "step": 5557 }, { "epoch": 0.47, "learning_rate": 5.6579396049396684e-06, "loss": 0.8363, "step": 5558 }, { "epoch": 0.47, "learning_rate": 5.656569704335534e-06, "loss": 0.8477, "step": 5559 }, { "epoch": 0.47, "learning_rate": 5.65519975357933e-06, "loss": 0.8244, "step": 5560 }, { "epoch": 0.47, "learning_rate": 5.653829752775698e-06, "loss": 0.8633, "step": 5561 }, { "epoch": 0.47, "learning_rate": 5.652459702029288e-06, "loss": 0.8604, "step": 5562 }, { "epoch": 0.47, "learning_rate": 5.651089601444752e-06, "loss": 0.8453, "step": 5563 }, { "epoch": 0.47, "learning_rate": 5.649719451126743e-06, "loss": 0.7778, "step": 5564 }, { "epoch": 0.47, "learning_rate": 5.64834925117992e-06, "loss": 0.8169, "step": 5565 }, { "epoch": 0.47, "learning_rate": 5.646979001708947e-06, "loss": 0.8338, "step": 5566 }, { "epoch": 0.48, "learning_rate": 5.645608702818489e-06, "loss": 0.8548, "step": 5567 }, { "epoch": 0.48, "learning_rate": 5.644238354613217e-06, "loss": 0.7919, "step": 5568 }, { "epoch": 0.48, "learning_rate": 5.642867957197806e-06, "loss": 0.8053, "step": 5569 }, { "epoch": 0.48, "learning_rate": 5.641497510676933e-06, "loss": 0.8739, "step": 5570 }, { "epoch": 0.48, "learning_rate": 5.640127015155279e-06, "loss": 0.873, "step": 5571 }, { "epoch": 0.48, "learning_rate": 5.638756470737529e-06, "loss": 0.828, "step": 5572 }, { "epoch": 0.48, "learning_rate": 5.637385877528374e-06, "loss": 0.8239, "step": 5573 }, { "epoch": 0.48, "learning_rate": 5.636015235632506e-06, "loss": 0.8749, "step": 5574 }, { "epoch": 0.48, "learning_rate": 5.6346445451546194e-06, "loss": 0.8756, "step": 5575 }, { "epoch": 0.48, "learning_rate": 5.633273806199416e-06, "loss": 0.8372, "step": 5576 }, { "epoch": 0.48, "learning_rate": 5.6319030188716e-06, "loss": 0.7828, "step": 5577 }, { "epoch": 0.48, "learning_rate": 5.63053218327588e-06, "loss": 0.8706, "step": 5578 }, { "epoch": 0.48, "learning_rate": 5.629161299516964e-06, "loss": 0.8602, "step": 5579 }, { "epoch": 0.48, "learning_rate": 5.6277903676995705e-06, "loss": 0.7677, "step": 5580 }, { "epoch": 0.48, "learning_rate": 5.626419387928414e-06, "loss": 0.8462, "step": 5581 }, { "epoch": 0.48, "learning_rate": 5.6250483603082205e-06, "loss": 0.8757, "step": 5582 }, { "epoch": 0.48, "learning_rate": 5.623677284943717e-06, "loss": 0.8975, "step": 5583 }, { "epoch": 0.48, "learning_rate": 5.622306161939628e-06, "loss": 0.5592, "step": 5584 }, { "epoch": 0.48, "learning_rate": 5.620934991400689e-06, "loss": 0.7993, "step": 5585 }, { "epoch": 0.48, "learning_rate": 5.619563773431638e-06, "loss": 0.8262, "step": 5586 }, { "epoch": 0.48, "learning_rate": 5.618192508137216e-06, "loss": 0.8254, "step": 5587 }, { "epoch": 0.48, "learning_rate": 5.616821195622164e-06, "loss": 0.8015, "step": 5588 }, { "epoch": 0.48, "learning_rate": 5.615449835991233e-06, "loss": 0.8354, "step": 5589 }, { "epoch": 0.48, "learning_rate": 5.614078429349172e-06, "loss": 0.8456, "step": 5590 }, { "epoch": 0.48, "learning_rate": 5.612706975800736e-06, "loss": 0.819, "step": 5591 }, { "epoch": 0.48, "learning_rate": 5.611335475450685e-06, "loss": 0.8491, "step": 5592 }, { "epoch": 0.48, "learning_rate": 5.609963928403781e-06, "loss": 0.8436, "step": 5593 }, { "epoch": 0.48, "learning_rate": 5.608592334764787e-06, "loss": 0.8292, "step": 5594 }, { "epoch": 0.48, "learning_rate": 5.607220694638476e-06, "loss": 0.8524, "step": 5595 }, { "epoch": 0.48, "learning_rate": 5.605849008129617e-06, "loss": 0.8734, "step": 5596 }, { "epoch": 0.48, "learning_rate": 5.604477275342989e-06, "loss": 0.8071, "step": 5597 }, { "epoch": 0.48, "learning_rate": 5.60310549638337e-06, "loss": 0.7968, "step": 5598 }, { "epoch": 0.48, "learning_rate": 5.601733671355544e-06, "loss": 0.8351, "step": 5599 }, { "epoch": 0.48, "learning_rate": 5.600361800364298e-06, "loss": 0.8685, "step": 5600 }, { "epoch": 0.48, "learning_rate": 5.598989883514422e-06, "loss": 0.8434, "step": 5601 }, { "epoch": 0.48, "learning_rate": 5.597617920910711e-06, "loss": 0.8252, "step": 5602 }, { "epoch": 0.48, "learning_rate": 5.596245912657961e-06, "loss": 0.8296, "step": 5603 }, { "epoch": 0.48, "learning_rate": 5.5948738588609745e-06, "loss": 0.8518, "step": 5604 }, { "epoch": 0.48, "learning_rate": 5.593501759624552e-06, "loss": 0.8441, "step": 5605 }, { "epoch": 0.48, "learning_rate": 5.592129615053505e-06, "loss": 0.825, "step": 5606 }, { "epoch": 0.48, "learning_rate": 5.590757425252645e-06, "loss": 0.8328, "step": 5607 }, { "epoch": 0.48, "learning_rate": 5.589385190326784e-06, "loss": 0.8636, "step": 5608 }, { "epoch": 0.48, "learning_rate": 5.5880129103807415e-06, "loss": 0.8376, "step": 5609 }, { "epoch": 0.48, "learning_rate": 5.58664058551934e-06, "loss": 0.8347, "step": 5610 }, { "epoch": 0.48, "learning_rate": 5.585268215847402e-06, "loss": 0.8683, "step": 5611 }, { "epoch": 0.48, "learning_rate": 5.583895801469759e-06, "loss": 0.822, "step": 5612 }, { "epoch": 0.48, "learning_rate": 5.5825233424912415e-06, "loss": 0.807, "step": 5613 }, { "epoch": 0.48, "learning_rate": 5.581150839016683e-06, "loss": 0.8371, "step": 5614 }, { "epoch": 0.48, "learning_rate": 5.579778291150926e-06, "loss": 0.8538, "step": 5615 }, { "epoch": 0.48, "learning_rate": 5.578405698998808e-06, "loss": 0.8269, "step": 5616 }, { "epoch": 0.48, "learning_rate": 5.577033062665179e-06, "loss": 0.7825, "step": 5617 }, { "epoch": 0.48, "learning_rate": 5.5756603822548835e-06, "loss": 0.8373, "step": 5618 }, { "epoch": 0.48, "learning_rate": 5.5742876578727766e-06, "loss": 0.8189, "step": 5619 }, { "epoch": 0.48, "learning_rate": 5.572914889623715e-06, "loss": 0.808, "step": 5620 }, { "epoch": 0.48, "learning_rate": 5.571542077612554e-06, "loss": 0.8436, "step": 5621 }, { "epoch": 0.48, "learning_rate": 5.570169221944157e-06, "loss": 0.8701, "step": 5622 }, { "epoch": 0.48, "learning_rate": 5.568796322723391e-06, "loss": 0.8378, "step": 5623 }, { "epoch": 0.48, "learning_rate": 5.567423380055124e-06, "loss": 0.7972, "step": 5624 }, { "epoch": 0.48, "learning_rate": 5.566050394044228e-06, "loss": 0.8175, "step": 5625 }, { "epoch": 0.48, "learning_rate": 5.564677364795579e-06, "loss": 0.8357, "step": 5626 }, { "epoch": 0.48, "learning_rate": 5.563304292414053e-06, "loss": 0.8255, "step": 5627 }, { "epoch": 0.48, "learning_rate": 5.561931177004536e-06, "loss": 0.8231, "step": 5628 }, { "epoch": 0.48, "learning_rate": 5.560558018671913e-06, "loss": 0.8242, "step": 5629 }, { "epoch": 0.48, "learning_rate": 5.559184817521071e-06, "loss": 0.6316, "step": 5630 }, { "epoch": 0.48, "learning_rate": 5.557811573656903e-06, "loss": 0.8291, "step": 5631 }, { "epoch": 0.48, "learning_rate": 5.556438287184303e-06, "loss": 0.8549, "step": 5632 }, { "epoch": 0.48, "learning_rate": 5.555064958208172e-06, "loss": 0.7929, "step": 5633 }, { "epoch": 0.48, "learning_rate": 5.55369158683341e-06, "loss": 0.8448, "step": 5634 }, { "epoch": 0.48, "learning_rate": 5.552318173164919e-06, "loss": 0.8266, "step": 5635 }, { "epoch": 0.48, "learning_rate": 5.5509447173076135e-06, "loss": 0.8579, "step": 5636 }, { "epoch": 0.48, "learning_rate": 5.5495712193663994e-06, "loss": 0.7995, "step": 5637 }, { "epoch": 0.48, "learning_rate": 5.548197679446194e-06, "loss": 0.8898, "step": 5638 }, { "epoch": 0.48, "learning_rate": 5.546824097651916e-06, "loss": 0.8375, "step": 5639 }, { "epoch": 0.48, "learning_rate": 5.545450474088484e-06, "loss": 0.8432, "step": 5640 }, { "epoch": 0.48, "learning_rate": 5.544076808860823e-06, "loss": 0.8571, "step": 5641 }, { "epoch": 0.48, "learning_rate": 5.542703102073861e-06, "loss": 0.8575, "step": 5642 }, { "epoch": 0.48, "learning_rate": 5.54132935383253e-06, "loss": 0.8335, "step": 5643 }, { "epoch": 0.48, "learning_rate": 5.53995556424176e-06, "loss": 0.8557, "step": 5644 }, { "epoch": 0.48, "learning_rate": 5.538581733406491e-06, "loss": 0.8261, "step": 5645 }, { "epoch": 0.48, "learning_rate": 5.537207861431661e-06, "loss": 0.8421, "step": 5646 }, { "epoch": 0.48, "learning_rate": 5.535833948422216e-06, "loss": 0.8575, "step": 5647 }, { "epoch": 0.48, "learning_rate": 5.534459994483099e-06, "loss": 0.8007, "step": 5648 }, { "epoch": 0.48, "learning_rate": 5.533085999719261e-06, "loss": 0.7951, "step": 5649 }, { "epoch": 0.48, "learning_rate": 5.531711964235655e-06, "loss": 0.8512, "step": 5650 }, { "epoch": 0.48, "learning_rate": 5.530337888137236e-06, "loss": 0.8733, "step": 5651 }, { "epoch": 0.48, "learning_rate": 5.528963771528963e-06, "loss": 0.8588, "step": 5652 }, { "epoch": 0.48, "learning_rate": 5.527589614515799e-06, "loss": 0.5114, "step": 5653 }, { "epoch": 0.48, "learning_rate": 5.526215417202707e-06, "loss": 0.8115, "step": 5654 }, { "epoch": 0.48, "learning_rate": 5.5248411796946555e-06, "loss": 0.5881, "step": 5655 }, { "epoch": 0.48, "learning_rate": 5.523466902096617e-06, "loss": 0.8613, "step": 5656 }, { "epoch": 0.48, "learning_rate": 5.522092584513563e-06, "loss": 0.8654, "step": 5657 }, { "epoch": 0.48, "learning_rate": 5.520718227050474e-06, "loss": 0.8409, "step": 5658 }, { "epoch": 0.48, "learning_rate": 5.5193438298123295e-06, "loss": 0.8147, "step": 5659 }, { "epoch": 0.48, "learning_rate": 5.517969392904111e-06, "loss": 0.8258, "step": 5660 }, { "epoch": 0.48, "learning_rate": 5.516594916430806e-06, "loss": 0.6417, "step": 5661 }, { "epoch": 0.48, "learning_rate": 5.515220400497403e-06, "loss": 0.8503, "step": 5662 }, { "epoch": 0.48, "learning_rate": 5.513845845208898e-06, "loss": 0.8375, "step": 5663 }, { "epoch": 0.48, "learning_rate": 5.512471250670281e-06, "loss": 0.8339, "step": 5664 }, { "epoch": 0.48, "learning_rate": 5.511096616986552e-06, "loss": 0.8231, "step": 5665 }, { "epoch": 0.48, "learning_rate": 5.509721944262715e-06, "loss": 0.8369, "step": 5666 }, { "epoch": 0.48, "learning_rate": 5.508347232603773e-06, "loss": 0.8101, "step": 5667 }, { "epoch": 0.48, "learning_rate": 5.506972482114732e-06, "loss": 0.8234, "step": 5668 }, { "epoch": 0.48, "learning_rate": 5.5055976929006035e-06, "loss": 0.8425, "step": 5669 }, { "epoch": 0.48, "learning_rate": 5.504222865066402e-06, "loss": 0.8593, "step": 5670 }, { "epoch": 0.48, "learning_rate": 5.50284799871714e-06, "loss": 0.7947, "step": 5671 }, { "epoch": 0.48, "learning_rate": 5.501473093957841e-06, "loss": 0.8447, "step": 5672 }, { "epoch": 0.48, "learning_rate": 5.500098150893524e-06, "loss": 0.8232, "step": 5673 }, { "epoch": 0.48, "learning_rate": 5.498723169629215e-06, "loss": 0.8533, "step": 5674 }, { "epoch": 0.48, "learning_rate": 5.497348150269942e-06, "loss": 0.8348, "step": 5675 }, { "epoch": 0.48, "learning_rate": 5.4959730929207365e-06, "loss": 0.8516, "step": 5676 }, { "epoch": 0.48, "learning_rate": 5.494597997686633e-06, "loss": 0.8339, "step": 5677 }, { "epoch": 0.48, "learning_rate": 5.493222864672664e-06, "loss": 0.8541, "step": 5678 }, { "epoch": 0.48, "learning_rate": 5.491847693983872e-06, "loss": 0.7667, "step": 5679 }, { "epoch": 0.48, "learning_rate": 5.4904724857253e-06, "loss": 0.8398, "step": 5680 }, { "epoch": 0.48, "learning_rate": 5.489097240001993e-06, "loss": 0.8654, "step": 5681 }, { "epoch": 0.48, "learning_rate": 5.4877219569189965e-06, "loss": 0.8586, "step": 5682 }, { "epoch": 0.48, "learning_rate": 5.486346636581364e-06, "loss": 0.8311, "step": 5683 }, { "epoch": 0.49, "learning_rate": 5.48497127909415e-06, "loss": 0.8451, "step": 5684 }, { "epoch": 0.49, "learning_rate": 5.483595884562409e-06, "loss": 0.8812, "step": 5685 }, { "epoch": 0.49, "learning_rate": 5.482220453091203e-06, "loss": 0.8537, "step": 5686 }, { "epoch": 0.49, "learning_rate": 5.4808449847855915e-06, "loss": 0.8776, "step": 5687 }, { "epoch": 0.49, "learning_rate": 5.4794694797506395e-06, "loss": 0.8092, "step": 5688 }, { "epoch": 0.49, "learning_rate": 5.4780939380914185e-06, "loss": 0.87, "step": 5689 }, { "epoch": 0.49, "learning_rate": 5.476718359912998e-06, "loss": 0.835, "step": 5690 }, { "epoch": 0.49, "learning_rate": 5.475342745320449e-06, "loss": 0.8348, "step": 5691 }, { "epoch": 0.49, "learning_rate": 5.473967094418848e-06, "loss": 0.8478, "step": 5692 }, { "epoch": 0.49, "learning_rate": 5.472591407313278e-06, "loss": 0.8297, "step": 5693 }, { "epoch": 0.49, "learning_rate": 5.4712156841088174e-06, "loss": 0.8392, "step": 5694 }, { "epoch": 0.49, "learning_rate": 5.469839924910553e-06, "loss": 0.8395, "step": 5695 }, { "epoch": 0.49, "learning_rate": 5.468464129823569e-06, "loss": 0.8199, "step": 5696 }, { "epoch": 0.49, "learning_rate": 5.467088298952958e-06, "loss": 0.8501, "step": 5697 }, { "epoch": 0.49, "learning_rate": 5.465712432403812e-06, "loss": 0.7777, "step": 5698 }, { "epoch": 0.49, "learning_rate": 5.464336530281229e-06, "loss": 0.8425, "step": 5699 }, { "epoch": 0.49, "learning_rate": 5.462960592690304e-06, "loss": 0.8088, "step": 5700 }, { "epoch": 0.49, "learning_rate": 5.4615846197361374e-06, "loss": 0.8219, "step": 5701 }, { "epoch": 0.49, "learning_rate": 5.4602086115238355e-06, "loss": 0.8381, "step": 5702 }, { "epoch": 0.49, "learning_rate": 5.458832568158504e-06, "loss": 0.8242, "step": 5703 }, { "epoch": 0.49, "learning_rate": 5.457456489745252e-06, "loss": 0.8292, "step": 5704 }, { "epoch": 0.49, "learning_rate": 5.456080376389189e-06, "loss": 0.8441, "step": 5705 }, { "epoch": 0.49, "learning_rate": 5.4547042281954335e-06, "loss": 0.8445, "step": 5706 }, { "epoch": 0.49, "learning_rate": 5.453328045269098e-06, "loss": 0.8177, "step": 5707 }, { "epoch": 0.49, "learning_rate": 5.451951827715306e-06, "loss": 0.8682, "step": 5708 }, { "epoch": 0.49, "learning_rate": 5.4505755756391795e-06, "loss": 0.8757, "step": 5709 }, { "epoch": 0.49, "learning_rate": 5.449199289145841e-06, "loss": 0.8215, "step": 5710 }, { "epoch": 0.49, "learning_rate": 5.447822968340419e-06, "loss": 0.8253, "step": 5711 }, { "epoch": 0.49, "learning_rate": 5.446446613328046e-06, "loss": 0.862, "step": 5712 }, { "epoch": 0.49, "learning_rate": 5.445070224213852e-06, "loss": 0.839, "step": 5713 }, { "epoch": 0.49, "learning_rate": 5.4436938011029746e-06, "loss": 0.8264, "step": 5714 }, { "epoch": 0.49, "learning_rate": 5.44231734410055e-06, "loss": 0.8722, "step": 5715 }, { "epoch": 0.49, "learning_rate": 5.440940853311721e-06, "loss": 0.8444, "step": 5716 }, { "epoch": 0.49, "learning_rate": 5.439564328841628e-06, "loss": 0.8452, "step": 5717 }, { "epoch": 0.49, "learning_rate": 5.43818777079542e-06, "loss": 0.8226, "step": 5718 }, { "epoch": 0.49, "learning_rate": 5.436811179278243e-06, "loss": 0.8649, "step": 5719 }, { "epoch": 0.49, "learning_rate": 5.435434554395248e-06, "loss": 0.8606, "step": 5720 }, { "epoch": 0.49, "learning_rate": 5.43405789625159e-06, "loss": 0.872, "step": 5721 }, { "epoch": 0.49, "learning_rate": 5.432681204952424e-06, "loss": 0.8232, "step": 5722 }, { "epoch": 0.49, "learning_rate": 5.431304480602909e-06, "loss": 0.8868, "step": 5723 }, { "epoch": 0.49, "learning_rate": 5.429927723308206e-06, "loss": 0.8709, "step": 5724 }, { "epoch": 0.49, "learning_rate": 5.428550933173476e-06, "loss": 0.8953, "step": 5725 }, { "epoch": 0.49, "learning_rate": 5.4271741103038915e-06, "loss": 0.8362, "step": 5726 }, { "epoch": 0.49, "learning_rate": 5.425797254804615e-06, "loss": 0.8176, "step": 5727 }, { "epoch": 0.49, "learning_rate": 5.424420366780819e-06, "loss": 0.8697, "step": 5728 }, { "epoch": 0.49, "learning_rate": 5.423043446337679e-06, "loss": 0.8039, "step": 5729 }, { "epoch": 0.49, "learning_rate": 5.42166649358037e-06, "loss": 0.8186, "step": 5730 }, { "epoch": 0.49, "learning_rate": 5.4202895086140685e-06, "loss": 0.8563, "step": 5731 }, { "epoch": 0.49, "learning_rate": 5.41891249154396e-06, "loss": 0.8398, "step": 5732 }, { "epoch": 0.49, "learning_rate": 5.4175354424752234e-06, "loss": 0.8669, "step": 5733 }, { "epoch": 0.49, "learning_rate": 5.416158361513046e-06, "loss": 0.7866, "step": 5734 }, { "epoch": 0.49, "learning_rate": 5.4147812487626185e-06, "loss": 0.827, "step": 5735 }, { "epoch": 0.49, "learning_rate": 5.413404104329128e-06, "loss": 0.8385, "step": 5736 }, { "epoch": 0.49, "learning_rate": 5.4120269283177715e-06, "loss": 0.859, "step": 5737 }, { "epoch": 0.49, "learning_rate": 5.410649720833741e-06, "loss": 0.8158, "step": 5738 }, { "epoch": 0.49, "learning_rate": 5.409272481982236e-06, "loss": 0.8438, "step": 5739 }, { "epoch": 0.49, "learning_rate": 5.407895211868457e-06, "loss": 0.8372, "step": 5740 }, { "epoch": 0.49, "learning_rate": 5.406517910597606e-06, "loss": 0.6109, "step": 5741 }, { "epoch": 0.49, "learning_rate": 5.405140578274889e-06, "loss": 0.8463, "step": 5742 }, { "epoch": 0.49, "learning_rate": 5.403763215005514e-06, "loss": 0.8769, "step": 5743 }, { "epoch": 0.49, "learning_rate": 5.402385820894689e-06, "loss": 0.8415, "step": 5744 }, { "epoch": 0.49, "learning_rate": 5.401008396047627e-06, "loss": 0.8725, "step": 5745 }, { "epoch": 0.49, "learning_rate": 5.399630940569545e-06, "loss": 0.8822, "step": 5746 }, { "epoch": 0.49, "learning_rate": 5.398253454565656e-06, "loss": 0.8114, "step": 5747 }, { "epoch": 0.49, "learning_rate": 5.3968759381411815e-06, "loss": 0.8491, "step": 5748 }, { "epoch": 0.49, "learning_rate": 5.395498391401344e-06, "loss": 0.835, "step": 5749 }, { "epoch": 0.49, "learning_rate": 5.394120814451364e-06, "loss": 0.8208, "step": 5750 }, { "epoch": 0.49, "learning_rate": 5.392743207396471e-06, "loss": 0.824, "step": 5751 }, { "epoch": 0.49, "learning_rate": 5.391365570341893e-06, "loss": 0.7949, "step": 5752 }, { "epoch": 0.49, "learning_rate": 5.389987903392858e-06, "loss": 0.829, "step": 5753 }, { "epoch": 0.49, "learning_rate": 5.388610206654602e-06, "loss": 0.8684, "step": 5754 }, { "epoch": 0.49, "learning_rate": 5.3872324802323606e-06, "loss": 0.847, "step": 5755 }, { "epoch": 0.49, "learning_rate": 5.38585472423137e-06, "loss": 0.8253, "step": 5756 }, { "epoch": 0.49, "learning_rate": 5.384476938756869e-06, "loss": 0.8404, "step": 5757 }, { "epoch": 0.49, "learning_rate": 5.383099123914101e-06, "loss": 0.8187, "step": 5758 }, { "epoch": 0.49, "learning_rate": 5.381721279808312e-06, "loss": 0.8369, "step": 5759 }, { "epoch": 0.49, "learning_rate": 5.3803434065447465e-06, "loss": 0.5368, "step": 5760 }, { "epoch": 0.49, "learning_rate": 5.378965504228654e-06, "loss": 0.8456, "step": 5761 }, { "epoch": 0.49, "learning_rate": 5.377587572965285e-06, "loss": 0.8532, "step": 5762 }, { "epoch": 0.49, "learning_rate": 5.376209612859893e-06, "loss": 0.7964, "step": 5763 }, { "epoch": 0.49, "learning_rate": 5.3748316240177345e-06, "loss": 0.825, "step": 5764 }, { "epoch": 0.49, "learning_rate": 5.373453606544067e-06, "loss": 0.8212, "step": 5765 }, { "epoch": 0.49, "learning_rate": 5.3720755605441484e-06, "loss": 0.8345, "step": 5766 }, { "epoch": 0.49, "learning_rate": 5.370697486123242e-06, "loss": 0.8181, "step": 5767 }, { "epoch": 0.49, "learning_rate": 5.369319383386612e-06, "loss": 0.8682, "step": 5768 }, { "epoch": 0.49, "learning_rate": 5.367941252439527e-06, "loss": 0.8458, "step": 5769 }, { "epoch": 0.49, "learning_rate": 5.366563093387252e-06, "loss": 0.8427, "step": 5770 }, { "epoch": 0.49, "learning_rate": 5.365184906335058e-06, "loss": 0.8643, "step": 5771 }, { "epoch": 0.49, "learning_rate": 5.363806691388221e-06, "loss": 0.8241, "step": 5772 }, { "epoch": 0.49, "learning_rate": 5.362428448652011e-06, "loss": 0.8493, "step": 5773 }, { "epoch": 0.49, "learning_rate": 5.361050178231709e-06, "loss": 0.8264, "step": 5774 }, { "epoch": 0.49, "learning_rate": 5.359671880232593e-06, "loss": 0.8285, "step": 5775 }, { "epoch": 0.49, "learning_rate": 5.3582935547599445e-06, "loss": 0.8575, "step": 5776 }, { "epoch": 0.49, "learning_rate": 5.356915201919046e-06, "loss": 0.6279, "step": 5777 }, { "epoch": 0.49, "learning_rate": 5.355536821815184e-06, "loss": 0.8732, "step": 5778 }, { "epoch": 0.49, "learning_rate": 5.3541584145536475e-06, "loss": 0.8953, "step": 5779 }, { "epoch": 0.49, "learning_rate": 5.352779980239722e-06, "loss": 0.61, "step": 5780 }, { "epoch": 0.49, "learning_rate": 5.351401518978701e-06, "loss": 0.7714, "step": 5781 }, { "epoch": 0.49, "learning_rate": 5.350023030875879e-06, "loss": 0.8906, "step": 5782 }, { "epoch": 0.49, "learning_rate": 5.348644516036552e-06, "loss": 0.8317, "step": 5783 }, { "epoch": 0.49, "learning_rate": 5.347265974566018e-06, "loss": 0.797, "step": 5784 }, { "epoch": 0.49, "learning_rate": 5.345887406569576e-06, "loss": 0.8343, "step": 5785 }, { "epoch": 0.49, "learning_rate": 5.344508812152528e-06, "loss": 0.8565, "step": 5786 }, { "epoch": 0.49, "learning_rate": 5.343130191420177e-06, "loss": 0.8535, "step": 5787 }, { "epoch": 0.49, "learning_rate": 5.3417515444778314e-06, "loss": 0.8601, "step": 5788 }, { "epoch": 0.49, "learning_rate": 5.340372871430798e-06, "loss": 0.8449, "step": 5789 }, { "epoch": 0.49, "learning_rate": 5.338994172384384e-06, "loss": 0.8124, "step": 5790 }, { "epoch": 0.49, "learning_rate": 5.337615447443906e-06, "loss": 0.8164, "step": 5791 }, { "epoch": 0.49, "learning_rate": 5.336236696714675e-06, "loss": 0.8792, "step": 5792 }, { "epoch": 0.49, "learning_rate": 5.334857920302008e-06, "loss": 0.8626, "step": 5793 }, { "epoch": 0.49, "learning_rate": 5.333479118311221e-06, "loss": 0.8565, "step": 5794 }, { "epoch": 0.49, "learning_rate": 5.332100290847635e-06, "loss": 0.7993, "step": 5795 }, { "epoch": 0.49, "learning_rate": 5.330721438016573e-06, "loss": 0.7932, "step": 5796 }, { "epoch": 0.49, "learning_rate": 5.329342559923356e-06, "loss": 0.8945, "step": 5797 }, { "epoch": 0.49, "learning_rate": 5.327963656673313e-06, "loss": 0.7828, "step": 5798 }, { "epoch": 0.49, "learning_rate": 5.32658472837177e-06, "loss": 0.5665, "step": 5799 }, { "epoch": 0.49, "learning_rate": 5.325205775124054e-06, "loss": 0.8059, "step": 5800 }, { "epoch": 0.5, "learning_rate": 5.323826797035497e-06, "loss": 0.8411, "step": 5801 }, { "epoch": 0.5, "learning_rate": 5.322447794211437e-06, "loss": 0.8341, "step": 5802 }, { "epoch": 0.5, "learning_rate": 5.3210687667572044e-06, "loss": 0.8503, "step": 5803 }, { "epoch": 0.5, "learning_rate": 5.319689714778137e-06, "loss": 0.7964, "step": 5804 }, { "epoch": 0.5, "learning_rate": 5.3183106383795735e-06, "loss": 0.8129, "step": 5805 }, { "epoch": 0.5, "learning_rate": 5.3169315376668566e-06, "loss": 0.809, "step": 5806 }, { "epoch": 0.5, "learning_rate": 5.315552412745327e-06, "loss": 0.8316, "step": 5807 }, { "epoch": 0.5, "learning_rate": 5.31417326372033e-06, "loss": 0.8028, "step": 5808 }, { "epoch": 0.5, "learning_rate": 5.312794090697212e-06, "loss": 0.8506, "step": 5809 }, { "epoch": 0.5, "learning_rate": 5.3114148937813195e-06, "loss": 0.8377, "step": 5810 }, { "epoch": 0.5, "learning_rate": 5.310035673078004e-06, "loss": 0.8528, "step": 5811 }, { "epoch": 0.5, "learning_rate": 5.308656428692618e-06, "loss": 0.8351, "step": 5812 }, { "epoch": 0.5, "learning_rate": 5.307277160730513e-06, "loss": 0.7603, "step": 5813 }, { "epoch": 0.5, "learning_rate": 5.305897869297044e-06, "loss": 0.8382, "step": 5814 }, { "epoch": 0.5, "learning_rate": 5.304518554497572e-06, "loss": 0.8114, "step": 5815 }, { "epoch": 0.5, "learning_rate": 5.303139216437452e-06, "loss": 0.8287, "step": 5816 }, { "epoch": 0.5, "learning_rate": 5.3017598552220475e-06, "loss": 0.886, "step": 5817 }, { "epoch": 0.5, "learning_rate": 5.3003804709567176e-06, "loss": 0.7976, "step": 5818 }, { "epoch": 0.5, "learning_rate": 5.29900106374683e-06, "loss": 0.8223, "step": 5819 }, { "epoch": 0.5, "learning_rate": 5.297621633697748e-06, "loss": 0.8249, "step": 5820 }, { "epoch": 0.5, "learning_rate": 5.296242180914843e-06, "loss": 0.7984, "step": 5821 }, { "epoch": 0.5, "learning_rate": 5.294862705503482e-06, "loss": 0.8381, "step": 5822 }, { "epoch": 0.5, "learning_rate": 5.293483207569033e-06, "loss": 0.7902, "step": 5823 }, { "epoch": 0.5, "learning_rate": 5.292103687216875e-06, "loss": 0.867, "step": 5824 }, { "epoch": 0.5, "learning_rate": 5.290724144552379e-06, "loss": 0.8589, "step": 5825 }, { "epoch": 0.5, "learning_rate": 5.289344579680924e-06, "loss": 0.8812, "step": 5826 }, { "epoch": 0.5, "learning_rate": 5.287964992707885e-06, "loss": 0.8017, "step": 5827 }, { "epoch": 0.5, "learning_rate": 5.286585383738643e-06, "loss": 0.8022, "step": 5828 }, { "epoch": 0.5, "learning_rate": 5.285205752878579e-06, "loss": 0.8387, "step": 5829 }, { "epoch": 0.5, "learning_rate": 5.283826100233078e-06, "loss": 0.9004, "step": 5830 }, { "epoch": 0.5, "learning_rate": 5.282446425907523e-06, "loss": 0.8612, "step": 5831 }, { "epoch": 0.5, "learning_rate": 5.2810667300073e-06, "loss": 0.828, "step": 5832 }, { "epoch": 0.5, "learning_rate": 5.279687012637798e-06, "loss": 0.8667, "step": 5833 }, { "epoch": 0.5, "learning_rate": 5.278307273904406e-06, "loss": 0.8103, "step": 5834 }, { "epoch": 0.5, "learning_rate": 5.2769275139125175e-06, "loss": 0.8369, "step": 5835 }, { "epoch": 0.5, "learning_rate": 5.275547732767523e-06, "loss": 0.8444, "step": 5836 }, { "epoch": 0.5, "learning_rate": 5.2741679305748175e-06, "loss": 0.7744, "step": 5837 }, { "epoch": 0.5, "learning_rate": 5.272788107439797e-06, "loss": 0.869, "step": 5838 }, { "epoch": 0.5, "learning_rate": 5.271408263467861e-06, "loss": 0.8767, "step": 5839 }, { "epoch": 0.5, "learning_rate": 5.2700283987644085e-06, "loss": 0.823, "step": 5840 }, { "epoch": 0.5, "learning_rate": 5.268648513434838e-06, "loss": 0.8078, "step": 5841 }, { "epoch": 0.5, "learning_rate": 5.267268607584553e-06, "loss": 0.8706, "step": 5842 }, { "epoch": 0.5, "learning_rate": 5.265888681318959e-06, "loss": 0.8429, "step": 5843 }, { "epoch": 0.5, "learning_rate": 5.264508734743462e-06, "loss": 0.8218, "step": 5844 }, { "epoch": 0.5, "learning_rate": 5.263128767963468e-06, "loss": 0.8843, "step": 5845 }, { "epoch": 0.5, "learning_rate": 5.261748781084385e-06, "loss": 0.8521, "step": 5846 }, { "epoch": 0.5, "learning_rate": 5.260368774211623e-06, "loss": 0.8595, "step": 5847 }, { "epoch": 0.5, "learning_rate": 5.258988747450596e-06, "loss": 0.8299, "step": 5848 }, { "epoch": 0.5, "learning_rate": 5.2576087009067175e-06, "loss": 0.8778, "step": 5849 }, { "epoch": 0.5, "learning_rate": 5.2562286346854e-06, "loss": 0.8468, "step": 5850 }, { "epoch": 0.5, "learning_rate": 5.254848548892061e-06, "loss": 0.8529, "step": 5851 }, { "epoch": 0.5, "learning_rate": 5.253468443632118e-06, "loss": 0.8815, "step": 5852 }, { "epoch": 0.5, "learning_rate": 5.252088319010991e-06, "loss": 0.857, "step": 5853 }, { "epoch": 0.5, "learning_rate": 5.250708175134101e-06, "loss": 0.7816, "step": 5854 }, { "epoch": 0.5, "learning_rate": 5.249328012106869e-06, "loss": 0.8395, "step": 5855 }, { "epoch": 0.5, "learning_rate": 5.2479478300347184e-06, "loss": 0.8544, "step": 5856 }, { "epoch": 0.5, "learning_rate": 5.2465676290230775e-06, "loss": 0.8474, "step": 5857 }, { "epoch": 0.5, "learning_rate": 5.245187409177369e-06, "loss": 0.8428, "step": 5858 }, { "epoch": 0.5, "learning_rate": 5.243807170603025e-06, "loss": 0.8055, "step": 5859 }, { "epoch": 0.5, "learning_rate": 5.242426913405471e-06, "loss": 0.8445, "step": 5860 }, { "epoch": 0.5, "learning_rate": 5.24104663769014e-06, "loss": 0.861, "step": 5861 }, { "epoch": 0.5, "learning_rate": 5.239666343562465e-06, "loss": 0.8508, "step": 5862 }, { "epoch": 0.5, "learning_rate": 5.238286031127879e-06, "loss": 0.8265, "step": 5863 }, { "epoch": 0.5, "learning_rate": 5.236905700491816e-06, "loss": 0.8496, "step": 5864 }, { "epoch": 0.5, "learning_rate": 5.235525351759715e-06, "loss": 0.8348, "step": 5865 }, { "epoch": 0.5, "learning_rate": 5.234144985037011e-06, "loss": 0.8354, "step": 5866 }, { "epoch": 0.5, "learning_rate": 5.232764600429145e-06, "loss": 0.8462, "step": 5867 }, { "epoch": 0.5, "learning_rate": 5.231384198041558e-06, "loss": 0.8804, "step": 5868 }, { "epoch": 0.5, "learning_rate": 5.230003777979692e-06, "loss": 0.8551, "step": 5869 }, { "epoch": 0.5, "learning_rate": 5.228623340348988e-06, "loss": 0.7902, "step": 5870 }, { "epoch": 0.5, "learning_rate": 5.227242885254893e-06, "loss": 0.8575, "step": 5871 }, { "epoch": 0.5, "learning_rate": 5.225862412802855e-06, "loss": 0.8167, "step": 5872 }, { "epoch": 0.5, "learning_rate": 5.224481923098317e-06, "loss": 0.7984, "step": 5873 }, { "epoch": 0.5, "learning_rate": 5.22310141624673e-06, "loss": 0.8732, "step": 5874 }, { "epoch": 0.5, "learning_rate": 5.221720892353544e-06, "loss": 0.8221, "step": 5875 }, { "epoch": 0.5, "learning_rate": 5.220340351524209e-06, "loss": 0.8643, "step": 5876 }, { "epoch": 0.5, "learning_rate": 5.21895979386418e-06, "loss": 0.8292, "step": 5877 }, { "epoch": 0.5, "learning_rate": 5.21757921947891e-06, "loss": 0.7831, "step": 5878 }, { "epoch": 0.5, "learning_rate": 5.216198628473851e-06, "loss": 0.7735, "step": 5879 }, { "epoch": 0.5, "learning_rate": 5.214818020954463e-06, "loss": 0.8223, "step": 5880 }, { "epoch": 0.5, "learning_rate": 5.213437397026205e-06, "loss": 0.8353, "step": 5881 }, { "epoch": 0.5, "learning_rate": 5.212056756794531e-06, "loss": 0.8526, "step": 5882 }, { "epoch": 0.5, "learning_rate": 5.210676100364904e-06, "loss": 0.8436, "step": 5883 }, { "epoch": 0.5, "learning_rate": 5.209295427842787e-06, "loss": 0.8763, "step": 5884 }, { "epoch": 0.5, "learning_rate": 5.20791473933364e-06, "loss": 0.8266, "step": 5885 }, { "epoch": 0.5, "learning_rate": 5.206534034942927e-06, "loss": 0.8658, "step": 5886 }, { "epoch": 0.5, "learning_rate": 5.2051533147761155e-06, "loss": 0.8904, "step": 5887 }, { "epoch": 0.5, "learning_rate": 5.20377257893867e-06, "loss": 0.8262, "step": 5888 }, { "epoch": 0.5, "learning_rate": 5.202391827536057e-06, "loss": 0.7849, "step": 5889 }, { "epoch": 0.5, "learning_rate": 5.2010110606737475e-06, "loss": 0.8536, "step": 5890 }, { "epoch": 0.5, "learning_rate": 5.199630278457211e-06, "loss": 0.7851, "step": 5891 }, { "epoch": 0.5, "learning_rate": 5.1982494809919175e-06, "loss": 0.8178, "step": 5892 }, { "epoch": 0.5, "learning_rate": 5.196868668383337e-06, "loss": 0.8713, "step": 5893 }, { "epoch": 0.5, "learning_rate": 5.195487840736947e-06, "loss": 0.8108, "step": 5894 }, { "epoch": 0.5, "learning_rate": 5.194106998158221e-06, "loss": 0.8384, "step": 5895 }, { "epoch": 0.5, "learning_rate": 5.192726140752634e-06, "loss": 0.8552, "step": 5896 }, { "epoch": 0.5, "learning_rate": 5.191345268625662e-06, "loss": 0.8244, "step": 5897 }, { "epoch": 0.5, "learning_rate": 5.189964381882784e-06, "loss": 0.8218, "step": 5898 }, { "epoch": 0.5, "learning_rate": 5.188583480629478e-06, "loss": 0.7996, "step": 5899 }, { "epoch": 0.5, "learning_rate": 5.187202564971226e-06, "loss": 0.8302, "step": 5900 }, { "epoch": 0.5, "learning_rate": 5.185821635013507e-06, "loss": 0.8262, "step": 5901 }, { "epoch": 0.5, "learning_rate": 5.184440690861806e-06, "loss": 0.8094, "step": 5902 }, { "epoch": 0.5, "learning_rate": 5.183059732621602e-06, "loss": 0.8013, "step": 5903 }, { "epoch": 0.5, "learning_rate": 5.181678760398384e-06, "loss": 0.8521, "step": 5904 }, { "epoch": 0.5, "learning_rate": 5.180297774297637e-06, "loss": 0.8182, "step": 5905 }, { "epoch": 0.5, "learning_rate": 5.178916774424846e-06, "loss": 0.8213, "step": 5906 }, { "epoch": 0.5, "learning_rate": 5.177535760885497e-06, "loss": 0.8121, "step": 5907 }, { "epoch": 0.5, "learning_rate": 5.176154733785082e-06, "loss": 0.8227, "step": 5908 }, { "epoch": 0.5, "learning_rate": 5.17477369322909e-06, "loss": 0.8978, "step": 5909 }, { "epoch": 0.5, "learning_rate": 5.173392639323011e-06, "loss": 0.8458, "step": 5910 }, { "epoch": 0.5, "learning_rate": 5.172011572172336e-06, "loss": 0.7778, "step": 5911 }, { "epoch": 0.5, "learning_rate": 5.170630491882561e-06, "loss": 0.8221, "step": 5912 }, { "epoch": 0.5, "learning_rate": 5.169249398559177e-06, "loss": 0.8318, "step": 5913 }, { "epoch": 0.5, "learning_rate": 5.167868292307679e-06, "loss": 0.824, "step": 5914 }, { "epoch": 0.5, "learning_rate": 5.166487173233564e-06, "loss": 0.8549, "step": 5915 }, { "epoch": 0.5, "learning_rate": 5.1651060414423275e-06, "loss": 0.8405, "step": 5916 }, { "epoch": 0.5, "learning_rate": 5.163724897039466e-06, "loss": 0.8634, "step": 5917 }, { "epoch": 0.5, "learning_rate": 5.162343740130482e-06, "loss": 0.7829, "step": 5918 }, { "epoch": 0.51, "learning_rate": 5.160962570820873e-06, "loss": 0.8311, "step": 5919 }, { "epoch": 0.51, "learning_rate": 5.15958138921614e-06, "loss": 0.8478, "step": 5920 }, { "epoch": 0.51, "learning_rate": 5.158200195421783e-06, "loss": 0.7906, "step": 5921 }, { "epoch": 0.51, "learning_rate": 5.156818989543308e-06, "loss": 0.8098, "step": 5922 }, { "epoch": 0.51, "learning_rate": 5.1554377716862124e-06, "loss": 0.881, "step": 5923 }, { "epoch": 0.51, "learning_rate": 5.154056541956007e-06, "loss": 0.832, "step": 5924 }, { "epoch": 0.51, "learning_rate": 5.152675300458195e-06, "loss": 0.8658, "step": 5925 }, { "epoch": 0.51, "learning_rate": 5.1512940472982795e-06, "loss": 0.5864, "step": 5926 }, { "epoch": 0.51, "learning_rate": 5.149912782581771e-06, "loss": 0.829, "step": 5927 }, { "epoch": 0.51, "learning_rate": 5.148531506414176e-06, "loss": 0.8773, "step": 5928 }, { "epoch": 0.51, "learning_rate": 5.147150218901005e-06, "loss": 0.8312, "step": 5929 }, { "epoch": 0.51, "learning_rate": 5.145768920147763e-06, "loss": 0.802, "step": 5930 }, { "epoch": 0.51, "learning_rate": 5.144387610259965e-06, "loss": 0.5491, "step": 5931 }, { "epoch": 0.51, "learning_rate": 5.143006289343122e-06, "loss": 0.8585, "step": 5932 }, { "epoch": 0.51, "learning_rate": 5.141624957502745e-06, "loss": 0.8382, "step": 5933 }, { "epoch": 0.51, "learning_rate": 5.140243614844346e-06, "loss": 0.8254, "step": 5934 }, { "epoch": 0.51, "learning_rate": 5.138862261473441e-06, "loss": 0.8483, "step": 5935 }, { "epoch": 0.51, "learning_rate": 5.137480897495542e-06, "loss": 0.8257, "step": 5936 }, { "epoch": 0.51, "learning_rate": 5.136099523016167e-06, "loss": 0.8647, "step": 5937 }, { "epoch": 0.51, "learning_rate": 5.1347181381408324e-06, "loss": 0.8002, "step": 5938 }, { "epoch": 0.51, "learning_rate": 5.133336742975054e-06, "loss": 0.8856, "step": 5939 }, { "epoch": 0.51, "learning_rate": 5.131955337624348e-06, "loss": 0.891, "step": 5940 }, { "epoch": 0.51, "learning_rate": 5.130573922194236e-06, "loss": 0.8591, "step": 5941 }, { "epoch": 0.51, "learning_rate": 5.129192496790236e-06, "loss": 0.8318, "step": 5942 }, { "epoch": 0.51, "learning_rate": 5.12781106151787e-06, "loss": 0.8723, "step": 5943 }, { "epoch": 0.51, "learning_rate": 5.126429616482655e-06, "loss": 0.6102, "step": 5944 }, { "epoch": 0.51, "learning_rate": 5.125048161790118e-06, "loss": 0.7943, "step": 5945 }, { "epoch": 0.51, "learning_rate": 5.123666697545776e-06, "loss": 0.7894, "step": 5946 }, { "epoch": 0.51, "learning_rate": 5.122285223855157e-06, "loss": 0.8518, "step": 5947 }, { "epoch": 0.51, "learning_rate": 5.120903740823782e-06, "loss": 0.82, "step": 5948 }, { "epoch": 0.51, "learning_rate": 5.119522248557174e-06, "loss": 0.8778, "step": 5949 }, { "epoch": 0.51, "learning_rate": 5.11814074716086e-06, "loss": 0.8528, "step": 5950 }, { "epoch": 0.51, "learning_rate": 5.1167592367403684e-06, "loss": 0.8279, "step": 5951 }, { "epoch": 0.51, "learning_rate": 5.115377717401224e-06, "loss": 0.8422, "step": 5952 }, { "epoch": 0.51, "learning_rate": 5.1139961892489524e-06, "loss": 0.5385, "step": 5953 }, { "epoch": 0.51, "learning_rate": 5.112614652389083e-06, "loss": 0.8249, "step": 5954 }, { "epoch": 0.51, "learning_rate": 5.111233106927146e-06, "loss": 0.7737, "step": 5955 }, { "epoch": 0.51, "learning_rate": 5.109851552968667e-06, "loss": 0.855, "step": 5956 }, { "epoch": 0.51, "learning_rate": 5.108469990619181e-06, "loss": 0.8209, "step": 5957 }, { "epoch": 0.51, "learning_rate": 5.107088419984215e-06, "loss": 0.82, "step": 5958 }, { "epoch": 0.51, "learning_rate": 5.105706841169301e-06, "loss": 0.8044, "step": 5959 }, { "epoch": 0.51, "learning_rate": 5.1043252542799706e-06, "loss": 0.8426, "step": 5960 }, { "epoch": 0.51, "learning_rate": 5.102943659421757e-06, "loss": 0.811, "step": 5961 }, { "epoch": 0.51, "learning_rate": 5.101562056700195e-06, "loss": 0.8222, "step": 5962 }, { "epoch": 0.51, "learning_rate": 5.1001804462208125e-06, "loss": 0.8284, "step": 5963 }, { "epoch": 0.51, "learning_rate": 5.09879882808915e-06, "loss": 0.8233, "step": 5964 }, { "epoch": 0.51, "learning_rate": 5.097417202410741e-06, "loss": 0.8877, "step": 5965 }, { "epoch": 0.51, "learning_rate": 5.096035569291119e-06, "loss": 0.8383, "step": 5966 }, { "epoch": 0.51, "learning_rate": 5.09465392883582e-06, "loss": 0.8599, "step": 5967 }, { "epoch": 0.51, "learning_rate": 5.093272281150383e-06, "loss": 0.8399, "step": 5968 }, { "epoch": 0.51, "learning_rate": 5.091890626340342e-06, "loss": 0.8429, "step": 5969 }, { "epoch": 0.51, "learning_rate": 5.090508964511239e-06, "loss": 0.8574, "step": 5970 }, { "epoch": 0.51, "learning_rate": 5.089127295768609e-06, "loss": 0.8356, "step": 5971 }, { "epoch": 0.51, "learning_rate": 5.087745620217991e-06, "loss": 0.8296, "step": 5972 }, { "epoch": 0.51, "learning_rate": 5.086363937964923e-06, "loss": 0.8266, "step": 5973 }, { "epoch": 0.51, "learning_rate": 5.084982249114948e-06, "loss": 0.7781, "step": 5974 }, { "epoch": 0.51, "learning_rate": 5.083600553773606e-06, "loss": 0.8054, "step": 5975 }, { "epoch": 0.51, "learning_rate": 5.0822188520464365e-06, "loss": 0.8357, "step": 5976 }, { "epoch": 0.51, "learning_rate": 5.080837144038979e-06, "loss": 0.859, "step": 5977 }, { "epoch": 0.51, "learning_rate": 5.079455429856779e-06, "loss": 0.8685, "step": 5978 }, { "epoch": 0.51, "learning_rate": 5.078073709605375e-06, "loss": 0.848, "step": 5979 }, { "epoch": 0.51, "learning_rate": 5.076691983390313e-06, "loss": 0.8043, "step": 5980 }, { "epoch": 0.51, "learning_rate": 5.075310251317135e-06, "loss": 0.8086, "step": 5981 }, { "epoch": 0.51, "learning_rate": 5.073928513491383e-06, "loss": 0.7931, "step": 5982 }, { "epoch": 0.51, "learning_rate": 5.072546770018602e-06, "loss": 0.8351, "step": 5983 }, { "epoch": 0.51, "learning_rate": 5.0711650210043375e-06, "loss": 0.8464, "step": 5984 }, { "epoch": 0.51, "learning_rate": 5.069783266554133e-06, "loss": 0.8549, "step": 5985 }, { "epoch": 0.51, "learning_rate": 5.068401506773536e-06, "loss": 0.8234, "step": 5986 }, { "epoch": 0.51, "learning_rate": 5.067019741768089e-06, "loss": 0.8385, "step": 5987 }, { "epoch": 0.51, "learning_rate": 5.06563797164334e-06, "loss": 0.8398, "step": 5988 }, { "epoch": 0.51, "learning_rate": 5.064256196504837e-06, "loss": 0.86, "step": 5989 }, { "epoch": 0.51, "learning_rate": 5.062874416458123e-06, "loss": 0.8511, "step": 5990 }, { "epoch": 0.51, "learning_rate": 5.061492631608748e-06, "loss": 0.7437, "step": 5991 }, { "epoch": 0.51, "learning_rate": 5.060110842062258e-06, "loss": 0.8256, "step": 5992 }, { "epoch": 0.51, "learning_rate": 5.058729047924202e-06, "loss": 0.8346, "step": 5993 }, { "epoch": 0.51, "learning_rate": 5.0573472493001305e-06, "loss": 0.883, "step": 5994 }, { "epoch": 0.51, "learning_rate": 5.05596544629559e-06, "loss": 0.8181, "step": 5995 }, { "epoch": 0.51, "learning_rate": 5.054583639016127e-06, "loss": 0.7744, "step": 5996 }, { "epoch": 0.51, "learning_rate": 5.0532018275672945e-06, "loss": 0.8398, "step": 5997 }, { "epoch": 0.51, "learning_rate": 5.051820012054641e-06, "loss": 0.8114, "step": 5998 }, { "epoch": 0.51, "learning_rate": 5.050438192583716e-06, "loss": 0.8298, "step": 5999 }, { "epoch": 0.51, "learning_rate": 5.049056369260071e-06, "loss": 0.8079, "step": 6000 }, { "epoch": 0.51, "learning_rate": 5.047674542189256e-06, "loss": 0.8324, "step": 6001 }, { "epoch": 0.51, "learning_rate": 5.04629271147682e-06, "loss": 0.834, "step": 6002 }, { "epoch": 0.51, "learning_rate": 5.044910877228317e-06, "loss": 0.8327, "step": 6003 }, { "epoch": 0.51, "learning_rate": 5.0435290395492975e-06, "loss": 0.8199, "step": 6004 }, { "epoch": 0.51, "learning_rate": 5.042147198545312e-06, "loss": 0.8293, "step": 6005 }, { "epoch": 0.51, "learning_rate": 5.040765354321913e-06, "loss": 0.8329, "step": 6006 }, { "epoch": 0.51, "learning_rate": 5.039383506984652e-06, "loss": 0.8311, "step": 6007 }, { "epoch": 0.51, "learning_rate": 5.0380016566390855e-06, "loss": 0.7973, "step": 6008 }, { "epoch": 0.51, "learning_rate": 5.03661980339076e-06, "loss": 0.8284, "step": 6009 }, { "epoch": 0.51, "learning_rate": 5.035237947345231e-06, "loss": 0.8429, "step": 6010 }, { "epoch": 0.51, "learning_rate": 5.033856088608054e-06, "loss": 0.862, "step": 6011 }, { "epoch": 0.51, "learning_rate": 5.032474227284778e-06, "loss": 0.8119, "step": 6012 }, { "epoch": 0.51, "learning_rate": 5.031092363480959e-06, "loss": 0.846, "step": 6013 }, { "epoch": 0.51, "learning_rate": 5.029710497302152e-06, "loss": 0.8362, "step": 6014 }, { "epoch": 0.51, "learning_rate": 5.028328628853906e-06, "loss": 0.8243, "step": 6015 }, { "epoch": 0.51, "learning_rate": 5.0269467582417785e-06, "loss": 0.824, "step": 6016 }, { "epoch": 0.51, "learning_rate": 5.025564885571325e-06, "loss": 0.8215, "step": 6017 }, { "epoch": 0.51, "learning_rate": 5.024183010948097e-06, "loss": 0.8263, "step": 6018 }, { "epoch": 0.51, "learning_rate": 5.022801134477649e-06, "loss": 0.822, "step": 6019 }, { "epoch": 0.51, "learning_rate": 5.0214192562655375e-06, "loss": 0.8593, "step": 6020 }, { "epoch": 0.51, "learning_rate": 5.020037376417317e-06, "loss": 0.8517, "step": 6021 }, { "epoch": 0.51, "learning_rate": 5.018655495038542e-06, "loss": 0.8769, "step": 6022 }, { "epoch": 0.51, "learning_rate": 5.017273612234765e-06, "loss": 0.8557, "step": 6023 }, { "epoch": 0.51, "learning_rate": 5.015891728111546e-06, "loss": 0.8212, "step": 6024 }, { "epoch": 0.51, "learning_rate": 5.014509842774435e-06, "loss": 0.8122, "step": 6025 }, { "epoch": 0.51, "learning_rate": 5.013127956328991e-06, "loss": 0.7696, "step": 6026 }, { "epoch": 0.51, "learning_rate": 5.011746068880768e-06, "loss": 0.8355, "step": 6027 }, { "epoch": 0.51, "learning_rate": 5.010364180535322e-06, "loss": 0.8508, "step": 6028 }, { "epoch": 0.51, "learning_rate": 5.008982291398208e-06, "loss": 0.8503, "step": 6029 }, { "epoch": 0.51, "learning_rate": 5.007600401574981e-06, "loss": 0.8471, "step": 6030 }, { "epoch": 0.51, "learning_rate": 5.006218511171199e-06, "loss": 0.8738, "step": 6031 }, { "epoch": 0.51, "learning_rate": 5.004836620292416e-06, "loss": 0.8455, "step": 6032 }, { "epoch": 0.51, "learning_rate": 5.003454729044186e-06, "loss": 0.8199, "step": 6033 }, { "epoch": 0.51, "learning_rate": 5.002072837532068e-06, "loss": 0.8108, "step": 6034 }, { "epoch": 0.51, "learning_rate": 5.000690945861616e-06, "loss": 0.7937, "step": 6035 }, { "epoch": 0.52, "learning_rate": 4.999309054138386e-06, "loss": 0.8418, "step": 6036 }, { "epoch": 0.52, "learning_rate": 4.997927162467933e-06, "loss": 0.8653, "step": 6037 }, { "epoch": 0.52, "learning_rate": 4.9965452709558145e-06, "loss": 0.8332, "step": 6038 }, { "epoch": 0.52, "learning_rate": 4.995163379707587e-06, "loss": 0.8184, "step": 6039 }, { "epoch": 0.52, "learning_rate": 4.993781488828802e-06, "loss": 0.8582, "step": 6040 }, { "epoch": 0.52, "learning_rate": 4.99239959842502e-06, "loss": 0.8399, "step": 6041 }, { "epoch": 0.52, "learning_rate": 4.991017708601794e-06, "loss": 0.8024, "step": 6042 }, { "epoch": 0.52, "learning_rate": 4.989635819464679e-06, "loss": 0.813, "step": 6043 }, { "epoch": 0.52, "learning_rate": 4.988253931119232e-06, "loss": 0.8742, "step": 6044 }, { "epoch": 0.52, "learning_rate": 4.98687204367101e-06, "loss": 0.8194, "step": 6045 }, { "epoch": 0.52, "learning_rate": 4.985490157225568e-06, "loss": 0.7927, "step": 6046 }, { "epoch": 0.52, "learning_rate": 4.984108271888457e-06, "loss": 0.8253, "step": 6047 }, { "epoch": 0.52, "learning_rate": 4.982726387765236e-06, "loss": 0.8252, "step": 6048 }, { "epoch": 0.52, "learning_rate": 4.981344504961459e-06, "loss": 0.8374, "step": 6049 }, { "epoch": 0.52, "learning_rate": 4.979962623582684e-06, "loss": 0.8699, "step": 6050 }, { "epoch": 0.52, "learning_rate": 4.9785807437344625e-06, "loss": 0.8372, "step": 6051 }, { "epoch": 0.52, "learning_rate": 4.977198865522351e-06, "loss": 0.8318, "step": 6052 }, { "epoch": 0.52, "learning_rate": 4.975816989051906e-06, "loss": 0.8049, "step": 6053 }, { "epoch": 0.52, "learning_rate": 4.974435114428677e-06, "loss": 0.8233, "step": 6054 }, { "epoch": 0.52, "learning_rate": 4.973053241758223e-06, "loss": 0.7903, "step": 6055 }, { "epoch": 0.52, "learning_rate": 4.971671371146095e-06, "loss": 0.8353, "step": 6056 }, { "epoch": 0.52, "learning_rate": 4.97028950269785e-06, "loss": 0.8046, "step": 6057 }, { "epoch": 0.52, "learning_rate": 4.968907636519041e-06, "loss": 0.8631, "step": 6058 }, { "epoch": 0.52, "learning_rate": 4.967525772715224e-06, "loss": 0.8121, "step": 6059 }, { "epoch": 0.52, "learning_rate": 4.966143911391949e-06, "loss": 0.8504, "step": 6060 }, { "epoch": 0.52, "learning_rate": 4.96476205265477e-06, "loss": 0.8556, "step": 6061 }, { "epoch": 0.52, "learning_rate": 4.9633801966092415e-06, "loss": 0.8033, "step": 6062 }, { "epoch": 0.52, "learning_rate": 4.961998343360916e-06, "loss": 0.8086, "step": 6063 }, { "epoch": 0.52, "learning_rate": 4.960616493015347e-06, "loss": 0.8562, "step": 6064 }, { "epoch": 0.52, "learning_rate": 4.959234645678087e-06, "loss": 0.8502, "step": 6065 }, { "epoch": 0.52, "learning_rate": 4.9578528014546905e-06, "loss": 0.8657, "step": 6066 }, { "epoch": 0.52, "learning_rate": 4.956470960450704e-06, "loss": 0.844, "step": 6067 }, { "epoch": 0.52, "learning_rate": 4.955089122771684e-06, "loss": 0.7925, "step": 6068 }, { "epoch": 0.52, "learning_rate": 4.953707288523182e-06, "loss": 0.8433, "step": 6069 }, { "epoch": 0.52, "learning_rate": 4.9523254578107455e-06, "loss": 0.844, "step": 6070 }, { "epoch": 0.52, "learning_rate": 4.950943630739929e-06, "loss": 0.8577, "step": 6071 }, { "epoch": 0.52, "learning_rate": 4.949561807416285e-06, "loss": 0.8162, "step": 6072 }, { "epoch": 0.52, "learning_rate": 4.94817998794536e-06, "loss": 0.8208, "step": 6073 }, { "epoch": 0.52, "learning_rate": 4.946798172432707e-06, "loss": 0.813, "step": 6074 }, { "epoch": 0.52, "learning_rate": 4.945416360983875e-06, "loss": 0.8604, "step": 6075 }, { "epoch": 0.52, "learning_rate": 4.944034553704412e-06, "loss": 0.8168, "step": 6076 }, { "epoch": 0.52, "learning_rate": 4.9426527506998695e-06, "loss": 0.829, "step": 6077 }, { "epoch": 0.52, "learning_rate": 4.941270952075798e-06, "loss": 0.8149, "step": 6078 }, { "epoch": 0.52, "learning_rate": 4.939889157937744e-06, "loss": 0.8176, "step": 6079 }, { "epoch": 0.52, "learning_rate": 4.9385073683912545e-06, "loss": 0.8518, "step": 6080 }, { "epoch": 0.52, "learning_rate": 4.937125583541879e-06, "loss": 0.8233, "step": 6081 }, { "epoch": 0.52, "learning_rate": 4.935743803495165e-06, "loss": 0.8824, "step": 6082 }, { "epoch": 0.52, "learning_rate": 4.9343620283566615e-06, "loss": 0.7957, "step": 6083 }, { "epoch": 0.52, "learning_rate": 4.932980258231912e-06, "loss": 0.8535, "step": 6084 }, { "epoch": 0.52, "learning_rate": 4.931598493226465e-06, "loss": 0.8812, "step": 6085 }, { "epoch": 0.52, "learning_rate": 4.930216733445868e-06, "loss": 0.897, "step": 6086 }, { "epoch": 0.52, "learning_rate": 4.928834978995664e-06, "loss": 0.8564, "step": 6087 }, { "epoch": 0.52, "learning_rate": 4.9274532299814e-06, "loss": 0.8649, "step": 6088 }, { "epoch": 0.52, "learning_rate": 4.926071486508619e-06, "loss": 0.867, "step": 6089 }, { "epoch": 0.52, "learning_rate": 4.924689748682866e-06, "loss": 0.7675, "step": 6090 }, { "epoch": 0.52, "learning_rate": 4.923308016609689e-06, "loss": 0.7847, "step": 6091 }, { "epoch": 0.52, "learning_rate": 4.921926290394626e-06, "loss": 0.8459, "step": 6092 }, { "epoch": 0.52, "learning_rate": 4.920544570143223e-06, "loss": 0.8508, "step": 6093 }, { "epoch": 0.52, "learning_rate": 4.919162855961022e-06, "loss": 0.8327, "step": 6094 }, { "epoch": 0.52, "learning_rate": 4.917781147953565e-06, "loss": 0.8449, "step": 6095 }, { "epoch": 0.52, "learning_rate": 4.916399446226395e-06, "loss": 0.78, "step": 6096 }, { "epoch": 0.52, "learning_rate": 4.915017750885052e-06, "loss": 0.7952, "step": 6097 }, { "epoch": 0.52, "learning_rate": 4.913636062035076e-06, "loss": 0.854, "step": 6098 }, { "epoch": 0.52, "learning_rate": 4.912254379782012e-06, "loss": 0.867, "step": 6099 }, { "epoch": 0.52, "learning_rate": 4.910872704231393e-06, "loss": 0.8696, "step": 6100 }, { "epoch": 0.52, "learning_rate": 4.9094910354887625e-06, "loss": 0.8505, "step": 6101 }, { "epoch": 0.52, "learning_rate": 4.908109373659659e-06, "loss": 0.8319, "step": 6102 }, { "epoch": 0.52, "learning_rate": 4.906727718849619e-06, "loss": 0.85, "step": 6103 }, { "epoch": 0.52, "learning_rate": 4.905346071164181e-06, "loss": 0.8647, "step": 6104 }, { "epoch": 0.52, "learning_rate": 4.903964430708883e-06, "loss": 0.8054, "step": 6105 }, { "epoch": 0.52, "learning_rate": 4.902582797589262e-06, "loss": 0.7962, "step": 6106 }, { "epoch": 0.52, "learning_rate": 4.901201171910852e-06, "loss": 0.8155, "step": 6107 }, { "epoch": 0.52, "learning_rate": 4.899819553779188e-06, "loss": 0.8358, "step": 6108 }, { "epoch": 0.52, "learning_rate": 4.898437943299808e-06, "loss": 0.8767, "step": 6109 }, { "epoch": 0.52, "learning_rate": 4.897056340578244e-06, "loss": 0.7965, "step": 6110 }, { "epoch": 0.52, "learning_rate": 4.89567474572003e-06, "loss": 0.7781, "step": 6111 }, { "epoch": 0.52, "learning_rate": 4.894293158830702e-06, "loss": 0.8945, "step": 6112 }, { "epoch": 0.52, "learning_rate": 4.892911580015787e-06, "loss": 0.8165, "step": 6113 }, { "epoch": 0.52, "learning_rate": 4.8915300093808205e-06, "loss": 0.8345, "step": 6114 }, { "epoch": 0.52, "learning_rate": 4.8901484470313335e-06, "loss": 0.8183, "step": 6115 }, { "epoch": 0.52, "learning_rate": 4.888766893072856e-06, "loss": 0.5722, "step": 6116 }, { "epoch": 0.52, "learning_rate": 4.887385347610917e-06, "loss": 0.8387, "step": 6117 }, { "epoch": 0.52, "learning_rate": 4.8860038107510475e-06, "loss": 0.846, "step": 6118 }, { "epoch": 0.52, "learning_rate": 4.8846222825987785e-06, "loss": 0.8004, "step": 6119 }, { "epoch": 0.52, "learning_rate": 4.883240763259632e-06, "loss": 0.8916, "step": 6120 }, { "epoch": 0.52, "learning_rate": 4.881859252839141e-06, "loss": 0.8309, "step": 6121 }, { "epoch": 0.52, "learning_rate": 4.8804777514428275e-06, "loss": 0.8584, "step": 6122 }, { "epoch": 0.52, "learning_rate": 4.879096259176221e-06, "loss": 0.8261, "step": 6123 }, { "epoch": 0.52, "learning_rate": 4.877714776144845e-06, "loss": 0.8173, "step": 6124 }, { "epoch": 0.52, "learning_rate": 4.876333302454224e-06, "loss": 0.8143, "step": 6125 }, { "epoch": 0.52, "learning_rate": 4.8749518382098845e-06, "loss": 0.82, "step": 6126 }, { "epoch": 0.52, "learning_rate": 4.873570383517346e-06, "loss": 0.8286, "step": 6127 }, { "epoch": 0.52, "learning_rate": 4.872188938482131e-06, "loss": 0.8027, "step": 6128 }, { "epoch": 0.52, "learning_rate": 4.870807503209765e-06, "loss": 0.8532, "step": 6129 }, { "epoch": 0.52, "learning_rate": 4.8694260778057655e-06, "loss": 0.8422, "step": 6130 }, { "epoch": 0.52, "learning_rate": 4.868044662375653e-06, "loss": 0.5766, "step": 6131 }, { "epoch": 0.52, "learning_rate": 4.86666325702495e-06, "loss": 0.8507, "step": 6132 }, { "epoch": 0.52, "learning_rate": 4.865281861859169e-06, "loss": 0.8136, "step": 6133 }, { "epoch": 0.52, "learning_rate": 4.863900476983835e-06, "loss": 0.8108, "step": 6134 }, { "epoch": 0.52, "learning_rate": 4.8625191025044594e-06, "loss": 0.8322, "step": 6135 }, { "epoch": 0.52, "learning_rate": 4.861137738526561e-06, "loss": 0.8464, "step": 6136 }, { "epoch": 0.52, "learning_rate": 4.859756385155654e-06, "loss": 0.7933, "step": 6137 }, { "epoch": 0.52, "learning_rate": 4.8583750424972564e-06, "loss": 0.8267, "step": 6138 }, { "epoch": 0.52, "learning_rate": 4.85699371065688e-06, "loss": 0.7748, "step": 6139 }, { "epoch": 0.52, "learning_rate": 4.855612389740036e-06, "loss": 0.8232, "step": 6140 }, { "epoch": 0.52, "learning_rate": 4.854231079852238e-06, "loss": 0.8402, "step": 6141 }, { "epoch": 0.52, "learning_rate": 4.8528497810989965e-06, "loss": 0.8162, "step": 6142 }, { "epoch": 0.52, "learning_rate": 4.851468493585825e-06, "loss": 0.8617, "step": 6143 }, { "epoch": 0.52, "learning_rate": 4.850087217418229e-06, "loss": 0.8205, "step": 6144 }, { "epoch": 0.52, "learning_rate": 4.848705952701722e-06, "loss": 0.8815, "step": 6145 }, { "epoch": 0.52, "learning_rate": 4.847324699541808e-06, "loss": 0.8718, "step": 6146 }, { "epoch": 0.52, "learning_rate": 4.8459434580439936e-06, "loss": 0.8356, "step": 6147 }, { "epoch": 0.52, "learning_rate": 4.844562228313788e-06, "loss": 0.8161, "step": 6148 }, { "epoch": 0.52, "learning_rate": 4.843181010456695e-06, "loss": 0.8515, "step": 6149 }, { "epoch": 0.52, "learning_rate": 4.841799804578217e-06, "loss": 0.7932, "step": 6150 }, { "epoch": 0.52, "learning_rate": 4.840418610783861e-06, "loss": 0.8745, "step": 6151 }, { "epoch": 0.52, "learning_rate": 4.839037429179129e-06, "loss": 0.7857, "step": 6152 }, { "epoch": 0.53, "learning_rate": 4.83765625986952e-06, "loss": 0.8116, "step": 6153 }, { "epoch": 0.53, "learning_rate": 4.836275102960536e-06, "loss": 0.8385, "step": 6154 }, { "epoch": 0.53, "learning_rate": 4.834893958557675e-06, "loss": 0.8333, "step": 6155 }, { "epoch": 0.53, "learning_rate": 4.833512826766438e-06, "loss": 0.8755, "step": 6156 }, { "epoch": 0.53, "learning_rate": 4.832131707692322e-06, "loss": 0.8722, "step": 6157 }, { "epoch": 0.53, "learning_rate": 4.830750601440824e-06, "loss": 0.8264, "step": 6158 }, { "epoch": 0.53, "learning_rate": 4.829369508117441e-06, "loss": 0.8041, "step": 6159 }, { "epoch": 0.53, "learning_rate": 4.8279884278276645e-06, "loss": 0.8116, "step": 6160 }, { "epoch": 0.53, "learning_rate": 4.82660736067699e-06, "loss": 0.8291, "step": 6161 }, { "epoch": 0.53, "learning_rate": 4.825226306770911e-06, "loss": 0.8341, "step": 6162 }, { "epoch": 0.53, "learning_rate": 4.8238452662149185e-06, "loss": 0.6028, "step": 6163 }, { "epoch": 0.53, "learning_rate": 4.822464239114503e-06, "loss": 0.8487, "step": 6164 }, { "epoch": 0.53, "learning_rate": 4.821083225575157e-06, "loss": 0.8154, "step": 6165 }, { "epoch": 0.53, "learning_rate": 4.8197022257023645e-06, "loss": 0.7893, "step": 6166 }, { "epoch": 0.53, "learning_rate": 4.818321239601617e-06, "loss": 0.843, "step": 6167 }, { "epoch": 0.53, "learning_rate": 4.8169402673783985e-06, "loss": 0.8222, "step": 6168 }, { "epoch": 0.53, "learning_rate": 4.815559309138196e-06, "loss": 0.8298, "step": 6169 }, { "epoch": 0.53, "learning_rate": 4.814178364986493e-06, "loss": 0.8059, "step": 6170 }, { "epoch": 0.53, "learning_rate": 4.812797435028775e-06, "loss": 0.7969, "step": 6171 }, { "epoch": 0.53, "learning_rate": 4.811416519370524e-06, "loss": 0.8264, "step": 6172 }, { "epoch": 0.53, "learning_rate": 4.810035618117218e-06, "loss": 0.8503, "step": 6173 }, { "epoch": 0.53, "learning_rate": 4.80865473137434e-06, "loss": 0.7985, "step": 6174 }, { "epoch": 0.53, "learning_rate": 4.807273859247368e-06, "loss": 0.8135, "step": 6175 }, { "epoch": 0.53, "learning_rate": 4.80589300184178e-06, "loss": 0.8303, "step": 6176 }, { "epoch": 0.53, "learning_rate": 4.804512159263054e-06, "loss": 0.8255, "step": 6177 }, { "epoch": 0.53, "learning_rate": 4.803131331616663e-06, "loss": 0.8893, "step": 6178 }, { "epoch": 0.53, "learning_rate": 4.801750519008086e-06, "loss": 0.8168, "step": 6179 }, { "epoch": 0.53, "learning_rate": 4.800369721542791e-06, "loss": 0.7694, "step": 6180 }, { "epoch": 0.53, "learning_rate": 4.798988939326254e-06, "loss": 0.8246, "step": 6181 }, { "epoch": 0.53, "learning_rate": 4.797608172463944e-06, "loss": 0.8085, "step": 6182 }, { "epoch": 0.53, "learning_rate": 4.796227421061331e-06, "loss": 0.6577, "step": 6183 }, { "epoch": 0.53, "learning_rate": 4.7948466852238844e-06, "loss": 0.805, "step": 6184 }, { "epoch": 0.53, "learning_rate": 4.793465965057074e-06, "loss": 0.8743, "step": 6185 }, { "epoch": 0.53, "learning_rate": 4.792085260666362e-06, "loss": 0.8666, "step": 6186 }, { "epoch": 0.53, "learning_rate": 4.790704572157215e-06, "loss": 0.8364, "step": 6187 }, { "epoch": 0.53, "learning_rate": 4.7893238996350964e-06, "loss": 0.8614, "step": 6188 }, { "epoch": 0.53, "learning_rate": 4.7879432432054705e-06, "loss": 0.8257, "step": 6189 }, { "epoch": 0.53, "learning_rate": 4.786562602973797e-06, "loss": 0.8246, "step": 6190 }, { "epoch": 0.53, "learning_rate": 4.785181979045538e-06, "loss": 0.8422, "step": 6191 }, { "epoch": 0.53, "learning_rate": 4.783801371526151e-06, "loss": 0.7983, "step": 6192 }, { "epoch": 0.53, "learning_rate": 4.7824207805210935e-06, "loss": 0.8469, "step": 6193 }, { "epoch": 0.53, "learning_rate": 4.781040206135822e-06, "loss": 0.816, "step": 6194 }, { "epoch": 0.53, "learning_rate": 4.779659648475792e-06, "loss": 0.6044, "step": 6195 }, { "epoch": 0.53, "learning_rate": 4.7782791076464574e-06, "loss": 0.8132, "step": 6196 }, { "epoch": 0.53, "learning_rate": 4.77689858375327e-06, "loss": 0.8376, "step": 6197 }, { "epoch": 0.53, "learning_rate": 4.775518076901685e-06, "loss": 0.8315, "step": 6198 }, { "epoch": 0.53, "learning_rate": 4.774137587197148e-06, "loss": 0.8408, "step": 6199 }, { "epoch": 0.53, "learning_rate": 4.772757114745108e-06, "loss": 0.8176, "step": 6200 }, { "epoch": 0.53, "learning_rate": 4.771376659651013e-06, "loss": 0.8516, "step": 6201 }, { "epoch": 0.53, "learning_rate": 4.76999622202031e-06, "loss": 0.8104, "step": 6202 }, { "epoch": 0.53, "learning_rate": 4.768615801958443e-06, "loss": 0.8716, "step": 6203 }, { "epoch": 0.53, "learning_rate": 4.767235399570856e-06, "loss": 0.8226, "step": 6204 }, { "epoch": 0.53, "learning_rate": 4.765855014962993e-06, "loss": 0.8351, "step": 6205 }, { "epoch": 0.53, "learning_rate": 4.764474648240288e-06, "loss": 0.883, "step": 6206 }, { "epoch": 0.53, "learning_rate": 4.763094299508186e-06, "loss": 0.5507, "step": 6207 }, { "epoch": 0.53, "learning_rate": 4.761713968872122e-06, "loss": 0.8106, "step": 6208 }, { "epoch": 0.53, "learning_rate": 4.760333656437537e-06, "loss": 0.826, "step": 6209 }, { "epoch": 0.53, "learning_rate": 4.7589533623098604e-06, "loss": 0.7665, "step": 6210 }, { "epoch": 0.53, "learning_rate": 4.757573086594529e-06, "loss": 0.8027, "step": 6211 }, { "epoch": 0.53, "learning_rate": 4.756192829396978e-06, "loss": 0.8466, "step": 6212 }, { "epoch": 0.53, "learning_rate": 4.7548125908226315e-06, "loss": 0.8196, "step": 6213 }, { "epoch": 0.53, "learning_rate": 4.753432370976925e-06, "loss": 0.7928, "step": 6214 }, { "epoch": 0.53, "learning_rate": 4.752052169965282e-06, "loss": 0.8251, "step": 6215 }, { "epoch": 0.53, "learning_rate": 4.750671987893132e-06, "loss": 0.8017, "step": 6216 }, { "epoch": 0.53, "learning_rate": 4.749291824865901e-06, "loss": 0.8305, "step": 6217 }, { "epoch": 0.53, "learning_rate": 4.747911680989011e-06, "loss": 0.8382, "step": 6218 }, { "epoch": 0.53, "learning_rate": 4.746531556367884e-06, "loss": 0.859, "step": 6219 }, { "epoch": 0.53, "learning_rate": 4.745151451107941e-06, "loss": 0.8206, "step": 6220 }, { "epoch": 0.53, "learning_rate": 4.743771365314602e-06, "loss": 0.8471, "step": 6221 }, { "epoch": 0.53, "learning_rate": 4.742391299093283e-06, "loss": 0.8636, "step": 6222 }, { "epoch": 0.53, "learning_rate": 4.741011252549405e-06, "loss": 0.5703, "step": 6223 }, { "epoch": 0.53, "learning_rate": 4.739631225788377e-06, "loss": 0.8797, "step": 6224 }, { "epoch": 0.53, "learning_rate": 4.738251218915618e-06, "loss": 0.7539, "step": 6225 }, { "epoch": 0.53, "learning_rate": 4.736871232036534e-06, "loss": 0.8452, "step": 6226 }, { "epoch": 0.53, "learning_rate": 4.735491265256539e-06, "loss": 0.8361, "step": 6227 }, { "epoch": 0.53, "learning_rate": 4.7341113186810415e-06, "loss": 0.8483, "step": 6228 }, { "epoch": 0.53, "learning_rate": 4.732731392415448e-06, "loss": 0.8489, "step": 6229 }, { "epoch": 0.53, "learning_rate": 4.731351486565163e-06, "loss": 0.8895, "step": 6230 }, { "epoch": 0.53, "learning_rate": 4.729971601235593e-06, "loss": 0.7903, "step": 6231 }, { "epoch": 0.53, "learning_rate": 4.7285917365321396e-06, "loss": 0.8373, "step": 6232 }, { "epoch": 0.53, "learning_rate": 4.727211892560204e-06, "loss": 0.8573, "step": 6233 }, { "epoch": 0.53, "learning_rate": 4.725832069425184e-06, "loss": 0.7956, "step": 6234 }, { "epoch": 0.53, "learning_rate": 4.724452267232478e-06, "loss": 0.8234, "step": 6235 }, { "epoch": 0.53, "learning_rate": 4.723072486087484e-06, "loss": 0.8544, "step": 6236 }, { "epoch": 0.53, "learning_rate": 4.7216927260955946e-06, "loss": 0.7964, "step": 6237 }, { "epoch": 0.53, "learning_rate": 4.720312987362204e-06, "loss": 0.8383, "step": 6238 }, { "epoch": 0.53, "learning_rate": 4.718933269992702e-06, "loss": 0.8747, "step": 6239 }, { "epoch": 0.53, "learning_rate": 4.717553574092479e-06, "loss": 0.8115, "step": 6240 }, { "epoch": 0.53, "learning_rate": 4.716173899766922e-06, "loss": 0.8118, "step": 6241 }, { "epoch": 0.53, "learning_rate": 4.714794247121422e-06, "loss": 0.8872, "step": 6242 }, { "epoch": 0.53, "learning_rate": 4.713414616261358e-06, "loss": 0.8329, "step": 6243 }, { "epoch": 0.53, "learning_rate": 4.712035007292116e-06, "loss": 0.8454, "step": 6244 }, { "epoch": 0.53, "learning_rate": 4.710655420319078e-06, "loss": 0.8364, "step": 6245 }, { "epoch": 0.53, "learning_rate": 4.7092758554476215e-06, "loss": 0.8654, "step": 6246 }, { "epoch": 0.53, "learning_rate": 4.7078963127831265e-06, "loss": 0.7983, "step": 6247 }, { "epoch": 0.53, "learning_rate": 4.7065167924309675e-06, "loss": 0.824, "step": 6248 }, { "epoch": 0.53, "learning_rate": 4.70513729449652e-06, "loss": 0.8218, "step": 6249 }, { "epoch": 0.53, "learning_rate": 4.703757819085158e-06, "loss": 0.839, "step": 6250 }, { "epoch": 0.53, "learning_rate": 4.702378366302252e-06, "loss": 0.8055, "step": 6251 }, { "epoch": 0.53, "learning_rate": 4.700998936253172e-06, "loss": 0.7991, "step": 6252 }, { "epoch": 0.53, "learning_rate": 4.699619529043284e-06, "loss": 0.8448, "step": 6253 }, { "epoch": 0.53, "learning_rate": 4.698240144777955e-06, "loss": 0.8777, "step": 6254 }, { "epoch": 0.53, "learning_rate": 4.69686078356255e-06, "loss": 0.7984, "step": 6255 }, { "epoch": 0.53, "learning_rate": 4.695481445502429e-06, "loss": 0.7961, "step": 6256 }, { "epoch": 0.53, "learning_rate": 4.694102130702956e-06, "loss": 0.8164, "step": 6257 }, { "epoch": 0.53, "learning_rate": 4.69272283926949e-06, "loss": 0.8244, "step": 6258 }, { "epoch": 0.53, "learning_rate": 4.691343571307384e-06, "loss": 0.8484, "step": 6259 }, { "epoch": 0.53, "learning_rate": 4.689964326921997e-06, "loss": 0.8448, "step": 6260 }, { "epoch": 0.53, "learning_rate": 4.688585106218682e-06, "loss": 0.793, "step": 6261 }, { "epoch": 0.53, "learning_rate": 4.68720590930279e-06, "loss": 0.8262, "step": 6262 }, { "epoch": 0.53, "learning_rate": 4.6858267362796705e-06, "loss": 0.8538, "step": 6263 }, { "epoch": 0.53, "learning_rate": 4.6844475872546735e-06, "loss": 0.8093, "step": 6264 }, { "epoch": 0.53, "learning_rate": 4.683068462333144e-06, "loss": 0.8023, "step": 6265 }, { "epoch": 0.53, "learning_rate": 4.681689361620428e-06, "loss": 0.8492, "step": 6266 }, { "epoch": 0.53, "learning_rate": 4.680310285221865e-06, "loss": 0.8514, "step": 6267 }, { "epoch": 0.53, "learning_rate": 4.678931233242797e-06, "loss": 0.8691, "step": 6268 }, { "epoch": 0.53, "learning_rate": 4.677552205788565e-06, "loss": 0.8393, "step": 6269 }, { "epoch": 0.54, "learning_rate": 4.676173202964502e-06, "loss": 0.8009, "step": 6270 }, { "epoch": 0.54, "learning_rate": 4.674794224875949e-06, "loss": 0.8565, "step": 6271 }, { "epoch": 0.54, "learning_rate": 4.673415271628233e-06, "loss": 0.8648, "step": 6272 }, { "epoch": 0.54, "learning_rate": 4.6720363433266885e-06, "loss": 0.8271, "step": 6273 }, { "epoch": 0.54, "learning_rate": 4.670657440076645e-06, "loss": 0.8489, "step": 6274 }, { "epoch": 0.54, "learning_rate": 4.669278561983429e-06, "loss": 0.8252, "step": 6275 }, { "epoch": 0.54, "learning_rate": 4.6678997091523655e-06, "loss": 0.8017, "step": 6276 }, { "epoch": 0.54, "learning_rate": 4.66652088168878e-06, "loss": 0.8642, "step": 6277 }, { "epoch": 0.54, "learning_rate": 4.665142079697995e-06, "loss": 0.8704, "step": 6278 }, { "epoch": 0.54, "learning_rate": 4.663763303285327e-06, "loss": 0.8197, "step": 6279 }, { "epoch": 0.54, "learning_rate": 4.662384552556096e-06, "loss": 0.8066, "step": 6280 }, { "epoch": 0.54, "learning_rate": 4.661005827615617e-06, "loss": 0.5511, "step": 6281 }, { "epoch": 0.54, "learning_rate": 4.659627128569204e-06, "loss": 0.8013, "step": 6282 }, { "epoch": 0.54, "learning_rate": 4.65824845552217e-06, "loss": 0.8631, "step": 6283 }, { "epoch": 0.54, "learning_rate": 4.656869808579823e-06, "loss": 0.8279, "step": 6284 }, { "epoch": 0.54, "learning_rate": 4.655491187847474e-06, "loss": 0.8504, "step": 6285 }, { "epoch": 0.54, "learning_rate": 4.654112593430426e-06, "loss": 0.7851, "step": 6286 }, { "epoch": 0.54, "learning_rate": 4.652734025433983e-06, "loss": 0.8272, "step": 6287 }, { "epoch": 0.54, "learning_rate": 4.651355483963449e-06, "loss": 0.8445, "step": 6288 }, { "epoch": 0.54, "learning_rate": 4.649976969124122e-06, "loss": 0.8484, "step": 6289 }, { "epoch": 0.54, "learning_rate": 4.6485984810213e-06, "loss": 0.8019, "step": 6290 }, { "epoch": 0.54, "learning_rate": 4.6472200197602816e-06, "loss": 0.8397, "step": 6291 }, { "epoch": 0.54, "learning_rate": 4.645841585446356e-06, "loss": 0.8223, "step": 6292 }, { "epoch": 0.54, "learning_rate": 4.644463178184818e-06, "loss": 0.8307, "step": 6293 }, { "epoch": 0.54, "learning_rate": 4.643084798080955e-06, "loss": 0.826, "step": 6294 }, { "epoch": 0.54, "learning_rate": 4.641706445240057e-06, "loss": 0.7965, "step": 6295 }, { "epoch": 0.54, "learning_rate": 4.640328119767407e-06, "loss": 0.8312, "step": 6296 }, { "epoch": 0.54, "learning_rate": 4.638949821768292e-06, "loss": 0.8727, "step": 6297 }, { "epoch": 0.54, "learning_rate": 4.637571551347992e-06, "loss": 0.8325, "step": 6298 }, { "epoch": 0.54, "learning_rate": 4.636193308611783e-06, "loss": 0.8615, "step": 6299 }, { "epoch": 0.54, "learning_rate": 4.634815093664943e-06, "loss": 0.8205, "step": 6300 }, { "epoch": 0.54, "learning_rate": 4.633436906612749e-06, "loss": 0.804, "step": 6301 }, { "epoch": 0.54, "learning_rate": 4.632058747560475e-06, "loss": 0.7955, "step": 6302 }, { "epoch": 0.54, "learning_rate": 4.630680616613388e-06, "loss": 0.8648, "step": 6303 }, { "epoch": 0.54, "learning_rate": 4.629302513876758e-06, "loss": 0.7588, "step": 6304 }, { "epoch": 0.54, "learning_rate": 4.627924439455854e-06, "loss": 0.8133, "step": 6305 }, { "epoch": 0.54, "learning_rate": 4.6265463934559355e-06, "loss": 0.8385, "step": 6306 }, { "epoch": 0.54, "learning_rate": 4.625168375982267e-06, "loss": 0.8662, "step": 6307 }, { "epoch": 0.54, "learning_rate": 4.623790387140108e-06, "loss": 0.7945, "step": 6308 }, { "epoch": 0.54, "learning_rate": 4.622412427034716e-06, "loss": 0.5486, "step": 6309 }, { "epoch": 0.54, "learning_rate": 4.621034495771347e-06, "loss": 0.818, "step": 6310 }, { "epoch": 0.54, "learning_rate": 4.619656593455256e-06, "loss": 0.8151, "step": 6311 }, { "epoch": 0.54, "learning_rate": 4.6182787201916905e-06, "loss": 0.8209, "step": 6312 }, { "epoch": 0.54, "learning_rate": 4.616900876085901e-06, "loss": 0.8341, "step": 6313 }, { "epoch": 0.54, "learning_rate": 4.615523061243133e-06, "loss": 0.8271, "step": 6314 }, { "epoch": 0.54, "learning_rate": 4.614145275768632e-06, "loss": 0.8608, "step": 6315 }, { "epoch": 0.54, "learning_rate": 4.612767519767641e-06, "loss": 0.8156, "step": 6316 }, { "epoch": 0.54, "learning_rate": 4.611389793345399e-06, "loss": 0.8292, "step": 6317 }, { "epoch": 0.54, "learning_rate": 4.6100120966071445e-06, "loss": 0.8611, "step": 6318 }, { "epoch": 0.54, "learning_rate": 4.6086344296581095e-06, "loss": 0.8485, "step": 6319 }, { "epoch": 0.54, "learning_rate": 4.60725679260353e-06, "loss": 0.5735, "step": 6320 }, { "epoch": 0.54, "learning_rate": 4.605879185548637e-06, "loss": 0.876, "step": 6321 }, { "epoch": 0.54, "learning_rate": 4.604501608598658e-06, "loss": 0.8506, "step": 6322 }, { "epoch": 0.54, "learning_rate": 4.6031240618588185e-06, "loss": 0.8401, "step": 6323 }, { "epoch": 0.54, "learning_rate": 4.601746545434346e-06, "loss": 0.8207, "step": 6324 }, { "epoch": 0.54, "learning_rate": 4.600369059430457e-06, "loss": 0.8065, "step": 6325 }, { "epoch": 0.54, "learning_rate": 4.598991603952374e-06, "loss": 0.844, "step": 6326 }, { "epoch": 0.54, "learning_rate": 4.597614179105312e-06, "loss": 0.8317, "step": 6327 }, { "epoch": 0.54, "learning_rate": 4.5962367849944876e-06, "loss": 0.7994, "step": 6328 }, { "epoch": 0.54, "learning_rate": 4.594859421725111e-06, "loss": 0.787, "step": 6329 }, { "epoch": 0.54, "learning_rate": 4.593482089402395e-06, "loss": 0.8324, "step": 6330 }, { "epoch": 0.54, "learning_rate": 4.592104788131546e-06, "loss": 0.8309, "step": 6331 }, { "epoch": 0.54, "learning_rate": 4.5907275180177665e-06, "loss": 0.8411, "step": 6332 }, { "epoch": 0.54, "learning_rate": 4.589350279166261e-06, "loss": 0.845, "step": 6333 }, { "epoch": 0.54, "learning_rate": 4.58797307168223e-06, "loss": 0.7895, "step": 6334 }, { "epoch": 0.54, "learning_rate": 4.586595895670872e-06, "loss": 0.8049, "step": 6335 }, { "epoch": 0.54, "learning_rate": 4.585218751237382e-06, "loss": 0.8508, "step": 6336 }, { "epoch": 0.54, "learning_rate": 4.583841638486953e-06, "loss": 0.8228, "step": 6337 }, { "epoch": 0.54, "learning_rate": 4.582464557524778e-06, "loss": 0.5981, "step": 6338 }, { "epoch": 0.54, "learning_rate": 4.581087508456042e-06, "loss": 0.8084, "step": 6339 }, { "epoch": 0.54, "learning_rate": 4.579710491385932e-06, "loss": 0.8628, "step": 6340 }, { "epoch": 0.54, "learning_rate": 4.578333506419632e-06, "loss": 0.8951, "step": 6341 }, { "epoch": 0.54, "learning_rate": 4.576956553662322e-06, "loss": 0.5196, "step": 6342 }, { "epoch": 0.54, "learning_rate": 4.575579633219181e-06, "loss": 0.8513, "step": 6343 }, { "epoch": 0.54, "learning_rate": 4.574202745195387e-06, "loss": 0.8517, "step": 6344 }, { "epoch": 0.54, "learning_rate": 4.572825889696111e-06, "loss": 0.8017, "step": 6345 }, { "epoch": 0.54, "learning_rate": 4.5714490668265245e-06, "loss": 0.8519, "step": 6346 }, { "epoch": 0.54, "learning_rate": 4.570072276691796e-06, "loss": 0.8007, "step": 6347 }, { "epoch": 0.54, "learning_rate": 4.568695519397092e-06, "loss": 0.8612, "step": 6348 }, { "epoch": 0.54, "learning_rate": 4.567318795047577e-06, "loss": 0.836, "step": 6349 }, { "epoch": 0.54, "learning_rate": 4.565942103748411e-06, "loss": 0.8742, "step": 6350 }, { "epoch": 0.54, "learning_rate": 4.564565445604754e-06, "loss": 0.8176, "step": 6351 }, { "epoch": 0.54, "learning_rate": 4.563188820721759e-06, "loss": 0.8746, "step": 6352 }, { "epoch": 0.54, "learning_rate": 4.561812229204582e-06, "loss": 0.8082, "step": 6353 }, { "epoch": 0.54, "learning_rate": 4.560435671158373e-06, "loss": 0.8587, "step": 6354 }, { "epoch": 0.54, "learning_rate": 4.55905914668828e-06, "loss": 0.8527, "step": 6355 }, { "epoch": 0.54, "learning_rate": 4.557682655899451e-06, "loss": 0.8749, "step": 6356 }, { "epoch": 0.54, "learning_rate": 4.556306198897026e-06, "loss": 0.8364, "step": 6357 }, { "epoch": 0.54, "learning_rate": 4.5549297757861486e-06, "loss": 0.884, "step": 6358 }, { "epoch": 0.54, "learning_rate": 4.553553386671956e-06, "loss": 0.8261, "step": 6359 }, { "epoch": 0.54, "learning_rate": 4.552177031659582e-06, "loss": 0.7539, "step": 6360 }, { "epoch": 0.54, "learning_rate": 4.55080071085416e-06, "loss": 0.8703, "step": 6361 }, { "epoch": 0.54, "learning_rate": 4.549424424360821e-06, "loss": 0.8115, "step": 6362 }, { "epoch": 0.54, "learning_rate": 4.548048172284694e-06, "loss": 0.7929, "step": 6363 }, { "epoch": 0.54, "learning_rate": 4.546671954730904e-06, "loss": 0.7926, "step": 6364 }, { "epoch": 0.54, "learning_rate": 4.545295771804569e-06, "loss": 0.8578, "step": 6365 }, { "epoch": 0.54, "learning_rate": 4.543919623610812e-06, "loss": 0.808, "step": 6366 }, { "epoch": 0.54, "learning_rate": 4.54254351025475e-06, "loss": 0.8394, "step": 6367 }, { "epoch": 0.54, "learning_rate": 4.5411674318414975e-06, "loss": 0.8061, "step": 6368 }, { "epoch": 0.54, "learning_rate": 4.539791388476165e-06, "loss": 0.8309, "step": 6369 }, { "epoch": 0.54, "learning_rate": 4.538415380263863e-06, "loss": 0.8487, "step": 6370 }, { "epoch": 0.54, "learning_rate": 4.537039407309699e-06, "loss": 0.8548, "step": 6371 }, { "epoch": 0.54, "learning_rate": 4.535663469718773e-06, "loss": 0.8297, "step": 6372 }, { "epoch": 0.54, "learning_rate": 4.534287567596189e-06, "loss": 0.7485, "step": 6373 }, { "epoch": 0.54, "learning_rate": 4.532911701047043e-06, "loss": 0.826, "step": 6374 }, { "epoch": 0.54, "learning_rate": 4.5315358701764315e-06, "loss": 0.7539, "step": 6375 }, { "epoch": 0.54, "learning_rate": 4.530160075089449e-06, "loss": 0.8461, "step": 6376 }, { "epoch": 0.54, "learning_rate": 4.5287843158911825e-06, "loss": 0.8359, "step": 6377 }, { "epoch": 0.54, "learning_rate": 4.527408592686724e-06, "loss": 0.8187, "step": 6378 }, { "epoch": 0.54, "learning_rate": 4.5260329055811525e-06, "loss": 0.82, "step": 6379 }, { "epoch": 0.54, "learning_rate": 4.524657254679553e-06, "loss": 0.7756, "step": 6380 }, { "epoch": 0.54, "learning_rate": 4.523281640087004e-06, "loss": 0.8369, "step": 6381 }, { "epoch": 0.54, "learning_rate": 4.521906061908582e-06, "loss": 0.8014, "step": 6382 }, { "epoch": 0.54, "learning_rate": 4.52053052024936e-06, "loss": 0.7933, "step": 6383 }, { "epoch": 0.54, "learning_rate": 4.519155015214412e-06, "loss": 0.8136, "step": 6384 }, { "epoch": 0.54, "learning_rate": 4.517779546908799e-06, "loss": 0.8383, "step": 6385 }, { "epoch": 0.54, "learning_rate": 4.516404115437592e-06, "loss": 0.698, "step": 6386 }, { "epoch": 0.55, "learning_rate": 4.515028720905852e-06, "loss": 0.8272, "step": 6387 }, { "epoch": 0.55, "learning_rate": 4.513653363418637e-06, "loss": 0.8403, "step": 6388 }, { "epoch": 0.55, "learning_rate": 4.512278043081004e-06, "loss": 0.8227, "step": 6389 }, { "epoch": 0.55, "learning_rate": 4.51090275999801e-06, "loss": 0.8511, "step": 6390 }, { "epoch": 0.55, "learning_rate": 4.5095275142747015e-06, "loss": 0.8154, "step": 6391 }, { "epoch": 0.55, "learning_rate": 4.50815230601613e-06, "loss": 0.8278, "step": 6392 }, { "epoch": 0.55, "learning_rate": 4.506777135327338e-06, "loss": 0.8487, "step": 6393 }, { "epoch": 0.55, "learning_rate": 4.5054020023133695e-06, "loss": 0.8453, "step": 6394 }, { "epoch": 0.55, "learning_rate": 4.504026907079264e-06, "loss": 0.8276, "step": 6395 }, { "epoch": 0.55, "learning_rate": 4.502651849730058e-06, "loss": 0.812, "step": 6396 }, { "epoch": 0.55, "learning_rate": 4.501276830370787e-06, "loss": 0.8365, "step": 6397 }, { "epoch": 0.55, "learning_rate": 4.499901849106478e-06, "loss": 0.7803, "step": 6398 }, { "epoch": 0.55, "learning_rate": 4.49852690604216e-06, "loss": 0.8055, "step": 6399 }, { "epoch": 0.55, "learning_rate": 4.497152001282861e-06, "loss": 0.5184, "step": 6400 }, { "epoch": 0.55, "learning_rate": 4.4957771349336e-06, "loss": 0.8433, "step": 6401 }, { "epoch": 0.55, "learning_rate": 4.494402307099397e-06, "loss": 0.8229, "step": 6402 }, { "epoch": 0.55, "learning_rate": 4.493027517885268e-06, "loss": 0.8312, "step": 6403 }, { "epoch": 0.55, "learning_rate": 4.491652767396229e-06, "loss": 0.7973, "step": 6404 }, { "epoch": 0.55, "learning_rate": 4.490278055737286e-06, "loss": 0.8671, "step": 6405 }, { "epoch": 0.55, "learning_rate": 4.48890338301345e-06, "loss": 0.7926, "step": 6406 }, { "epoch": 0.55, "learning_rate": 4.487528749329722e-06, "loss": 0.8386, "step": 6407 }, { "epoch": 0.55, "learning_rate": 4.486154154791105e-06, "loss": 0.8423, "step": 6408 }, { "epoch": 0.55, "learning_rate": 4.484779599502598e-06, "loss": 0.8549, "step": 6409 }, { "epoch": 0.55, "learning_rate": 4.483405083569195e-06, "loss": 0.8164, "step": 6410 }, { "epoch": 0.55, "learning_rate": 4.482030607095892e-06, "loss": 0.6425, "step": 6411 }, { "epoch": 0.55, "learning_rate": 4.480656170187673e-06, "loss": 0.8034, "step": 6412 }, { "epoch": 0.55, "learning_rate": 4.479281772949527e-06, "loss": 0.8753, "step": 6413 }, { "epoch": 0.55, "learning_rate": 4.477907415486438e-06, "loss": 0.8474, "step": 6414 }, { "epoch": 0.55, "learning_rate": 4.476533097903385e-06, "loss": 0.8178, "step": 6415 }, { "epoch": 0.55, "learning_rate": 4.475158820305345e-06, "loss": 0.8401, "step": 6416 }, { "epoch": 0.55, "learning_rate": 4.473784582797296e-06, "loss": 0.8253, "step": 6417 }, { "epoch": 0.55, "learning_rate": 4.4724103854842035e-06, "loss": 0.8162, "step": 6418 }, { "epoch": 0.55, "learning_rate": 4.471036228471039e-06, "loss": 0.8378, "step": 6419 }, { "epoch": 0.55, "learning_rate": 4.469662111862766e-06, "loss": 0.8244, "step": 6420 }, { "epoch": 0.55, "learning_rate": 4.468288035764346e-06, "loss": 0.8392, "step": 6421 }, { "epoch": 0.55, "learning_rate": 4.46691400028074e-06, "loss": 0.8485, "step": 6422 }, { "epoch": 0.55, "learning_rate": 4.465540005516903e-06, "loss": 0.8398, "step": 6423 }, { "epoch": 0.55, "learning_rate": 4.464166051577786e-06, "loss": 0.8257, "step": 6424 }, { "epoch": 0.55, "learning_rate": 4.462792138568341e-06, "loss": 0.8076, "step": 6425 }, { "epoch": 0.55, "learning_rate": 4.461418266593511e-06, "loss": 0.8366, "step": 6426 }, { "epoch": 0.55, "learning_rate": 4.460044435758241e-06, "loss": 0.8174, "step": 6427 }, { "epoch": 0.55, "learning_rate": 4.458670646167473e-06, "loss": 0.8218, "step": 6428 }, { "epoch": 0.55, "learning_rate": 4.457296897926139e-06, "loss": 0.8169, "step": 6429 }, { "epoch": 0.55, "learning_rate": 4.455923191139177e-06, "loss": 0.8281, "step": 6430 }, { "epoch": 0.55, "learning_rate": 4.454549525911518e-06, "loss": 0.8379, "step": 6431 }, { "epoch": 0.55, "learning_rate": 4.453175902348086e-06, "loss": 0.8526, "step": 6432 }, { "epoch": 0.55, "learning_rate": 4.451802320553807e-06, "loss": 0.8305, "step": 6433 }, { "epoch": 0.55, "learning_rate": 4.450428780633602e-06, "loss": 0.8043, "step": 6434 }, { "epoch": 0.55, "learning_rate": 4.449055282692388e-06, "loss": 0.8346, "step": 6435 }, { "epoch": 0.55, "learning_rate": 4.447681826835081e-06, "loss": 0.8585, "step": 6436 }, { "epoch": 0.55, "learning_rate": 4.446308413166594e-06, "loss": 0.8188, "step": 6437 }, { "epoch": 0.55, "learning_rate": 4.444935041791831e-06, "loss": 0.8808, "step": 6438 }, { "epoch": 0.55, "learning_rate": 4.443561712815698e-06, "loss": 0.8521, "step": 6439 }, { "epoch": 0.55, "learning_rate": 4.442188426343098e-06, "loss": 0.7921, "step": 6440 }, { "epoch": 0.55, "learning_rate": 4.4408151824789295e-06, "loss": 0.817, "step": 6441 }, { "epoch": 0.55, "learning_rate": 4.439441981328088e-06, "loss": 0.8748, "step": 6442 }, { "epoch": 0.55, "learning_rate": 4.438068822995464e-06, "loss": 0.8435, "step": 6443 }, { "epoch": 0.55, "learning_rate": 4.4366957075859494e-06, "loss": 0.7963, "step": 6444 }, { "epoch": 0.55, "learning_rate": 4.435322635204424e-06, "loss": 0.7718, "step": 6445 }, { "epoch": 0.55, "learning_rate": 4.4339496059557745e-06, "loss": 0.7465, "step": 6446 }, { "epoch": 0.55, "learning_rate": 4.4325766199448785e-06, "loss": 0.8032, "step": 6447 }, { "epoch": 0.55, "learning_rate": 4.43120367727661e-06, "loss": 0.8792, "step": 6448 }, { "epoch": 0.55, "learning_rate": 4.429830778055843e-06, "loss": 0.8192, "step": 6449 }, { "epoch": 0.55, "learning_rate": 4.428457922387449e-06, "loss": 0.8495, "step": 6450 }, { "epoch": 0.55, "learning_rate": 4.427085110376287e-06, "loss": 0.829, "step": 6451 }, { "epoch": 0.55, "learning_rate": 4.425712342127224e-06, "loss": 0.7512, "step": 6452 }, { "epoch": 0.55, "learning_rate": 4.424339617745117e-06, "loss": 0.8269, "step": 6453 }, { "epoch": 0.55, "learning_rate": 4.4229669373348225e-06, "loss": 0.8162, "step": 6454 }, { "epoch": 0.55, "learning_rate": 4.421594301001192e-06, "loss": 0.8438, "step": 6455 }, { "epoch": 0.55, "learning_rate": 4.420221708849076e-06, "loss": 0.8373, "step": 6456 }, { "epoch": 0.55, "learning_rate": 4.418849160983319e-06, "loss": 0.8167, "step": 6457 }, { "epoch": 0.55, "learning_rate": 4.417476657508761e-06, "loss": 0.7933, "step": 6458 }, { "epoch": 0.55, "learning_rate": 4.416104198530243e-06, "loss": 0.8253, "step": 6459 }, { "epoch": 0.55, "learning_rate": 4.4147317841525984e-06, "loss": 0.8556, "step": 6460 }, { "epoch": 0.55, "learning_rate": 4.413359414480662e-06, "loss": 0.7872, "step": 6461 }, { "epoch": 0.55, "learning_rate": 4.411987089619259e-06, "loss": 0.7963, "step": 6462 }, { "epoch": 0.55, "learning_rate": 4.410614809673216e-06, "loss": 0.8633, "step": 6463 }, { "epoch": 0.55, "learning_rate": 4.409242574747357e-06, "loss": 0.8249, "step": 6464 }, { "epoch": 0.55, "learning_rate": 4.407870384946496e-06, "loss": 0.8405, "step": 6465 }, { "epoch": 0.55, "learning_rate": 4.4064982403754495e-06, "loss": 0.8589, "step": 6466 }, { "epoch": 0.55, "learning_rate": 4.405126141139028e-06, "loss": 0.789, "step": 6467 }, { "epoch": 0.55, "learning_rate": 4.40375408734204e-06, "loss": 0.8075, "step": 6468 }, { "epoch": 0.55, "learning_rate": 4.4023820790892894e-06, "loss": 0.8413, "step": 6469 }, { "epoch": 0.55, "learning_rate": 4.401010116485579e-06, "loss": 0.8447, "step": 6470 }, { "epoch": 0.55, "learning_rate": 4.399638199635704e-06, "loss": 0.8475, "step": 6471 }, { "epoch": 0.55, "learning_rate": 4.398266328644458e-06, "loss": 0.8164, "step": 6472 }, { "epoch": 0.55, "learning_rate": 4.396894503616632e-06, "loss": 0.7764, "step": 6473 }, { "epoch": 0.55, "learning_rate": 4.395522724657012e-06, "loss": 0.7933, "step": 6474 }, { "epoch": 0.55, "learning_rate": 4.394150991870385e-06, "loss": 0.8522, "step": 6475 }, { "epoch": 0.55, "learning_rate": 4.392779305361525e-06, "loss": 0.8056, "step": 6476 }, { "epoch": 0.55, "learning_rate": 4.391407665235215e-06, "loss": 0.8557, "step": 6477 }, { "epoch": 0.55, "learning_rate": 4.390036071596222e-06, "loss": 0.833, "step": 6478 }, { "epoch": 0.55, "learning_rate": 4.388664524549316e-06, "loss": 0.8096, "step": 6479 }, { "epoch": 0.55, "learning_rate": 4.387293024199265e-06, "loss": 0.8258, "step": 6480 }, { "epoch": 0.55, "learning_rate": 4.3859215706508295e-06, "loss": 0.7883, "step": 6481 }, { "epoch": 0.55, "learning_rate": 4.384550164008768e-06, "loss": 0.8517, "step": 6482 }, { "epoch": 0.55, "learning_rate": 4.3831788043778356e-06, "loss": 0.8508, "step": 6483 }, { "epoch": 0.55, "learning_rate": 4.381807491862786e-06, "loss": 0.8221, "step": 6484 }, { "epoch": 0.55, "learning_rate": 4.380436226568364e-06, "loss": 0.818, "step": 6485 }, { "epoch": 0.55, "learning_rate": 4.3790650085993125e-06, "loss": 0.7778, "step": 6486 }, { "epoch": 0.55, "learning_rate": 4.3776938380603735e-06, "loss": 0.8054, "step": 6487 }, { "epoch": 0.55, "learning_rate": 4.376322715056284e-06, "loss": 0.8332, "step": 6488 }, { "epoch": 0.55, "learning_rate": 4.374951639691779e-06, "loss": 0.8655, "step": 6489 }, { "epoch": 0.55, "learning_rate": 4.373580612071587e-06, "loss": 0.8075, "step": 6490 }, { "epoch": 0.55, "learning_rate": 4.372209632300432e-06, "loss": 0.7872, "step": 6491 }, { "epoch": 0.55, "learning_rate": 4.370838700483037e-06, "loss": 0.81, "step": 6492 }, { "epoch": 0.55, "learning_rate": 4.369467816724121e-06, "loss": 0.8546, "step": 6493 }, { "epoch": 0.55, "learning_rate": 4.3680969811284006e-06, "loss": 0.8415, "step": 6494 }, { "epoch": 0.55, "learning_rate": 4.366726193800585e-06, "loss": 0.859, "step": 6495 }, { "epoch": 0.55, "learning_rate": 4.3653554548453805e-06, "loss": 0.851, "step": 6496 }, { "epoch": 0.55, "learning_rate": 4.3639847643674965e-06, "loss": 0.8322, "step": 6497 }, { "epoch": 0.55, "learning_rate": 4.362614122471627e-06, "loss": 0.8123, "step": 6498 }, { "epoch": 0.55, "learning_rate": 4.361243529262472e-06, "loss": 0.8077, "step": 6499 }, { "epoch": 0.55, "learning_rate": 4.359872984844722e-06, "loss": 0.8358, "step": 6500 }, { "epoch": 0.55, "learning_rate": 4.3585024893230676e-06, "loss": 0.8014, "step": 6501 }, { "epoch": 0.55, "learning_rate": 4.357132042802194e-06, "loss": 0.7854, "step": 6502 }, { "epoch": 0.55, "learning_rate": 4.355761645386784e-06, "loss": 0.8125, "step": 6503 }, { "epoch": 0.55, "learning_rate": 4.3543912971815135e-06, "loss": 0.7889, "step": 6504 }, { "epoch": 0.56, "learning_rate": 4.3530209982910556e-06, "loss": 0.7941, "step": 6505 }, { "epoch": 0.56, "learning_rate": 4.351650748820082e-06, "loss": 0.771, "step": 6506 }, { "epoch": 0.56, "learning_rate": 4.350280548873258e-06, "loss": 0.8094, "step": 6507 }, { "epoch": 0.56, "learning_rate": 4.348910398555249e-06, "loss": 0.8282, "step": 6508 }, { "epoch": 0.56, "learning_rate": 4.347540297970712e-06, "loss": 0.8409, "step": 6509 }, { "epoch": 0.56, "learning_rate": 4.346170247224304e-06, "loss": 0.8001, "step": 6510 }, { "epoch": 0.56, "learning_rate": 4.344800246420673e-06, "loss": 0.8227, "step": 6511 }, { "epoch": 0.56, "learning_rate": 4.343430295664468e-06, "loss": 0.8619, "step": 6512 }, { "epoch": 0.56, "learning_rate": 4.342060395060333e-06, "loss": 0.7852, "step": 6513 }, { "epoch": 0.56, "learning_rate": 4.340690544712908e-06, "loss": 0.8795, "step": 6514 }, { "epoch": 0.56, "learning_rate": 4.339320744726828e-06, "loss": 0.8363, "step": 6515 }, { "epoch": 0.56, "learning_rate": 4.337950995206727e-06, "loss": 0.7919, "step": 6516 }, { "epoch": 0.56, "learning_rate": 4.336581296257233e-06, "loss": 0.8075, "step": 6517 }, { "epoch": 0.56, "learning_rate": 4.335211647982969e-06, "loss": 0.8512, "step": 6518 }, { "epoch": 0.56, "learning_rate": 4.333842050488555e-06, "loss": 0.8417, "step": 6519 }, { "epoch": 0.56, "learning_rate": 4.332472503878609e-06, "loss": 0.8166, "step": 6520 }, { "epoch": 0.56, "learning_rate": 4.331103008257745e-06, "loss": 0.53, "step": 6521 }, { "epoch": 0.56, "learning_rate": 4.3297335637305694e-06, "loss": 0.7942, "step": 6522 }, { "epoch": 0.56, "learning_rate": 4.328364170401691e-06, "loss": 0.8169, "step": 6523 }, { "epoch": 0.56, "learning_rate": 4.326994828375705e-06, "loss": 0.8376, "step": 6524 }, { "epoch": 0.56, "learning_rate": 4.325625537757213e-06, "loss": 0.8489, "step": 6525 }, { "epoch": 0.56, "learning_rate": 4.324256298650807e-06, "loss": 0.8283, "step": 6526 }, { "epoch": 0.56, "learning_rate": 4.322887111161077e-06, "loss": 0.8076, "step": 6527 }, { "epoch": 0.56, "learning_rate": 4.321517975392607e-06, "loss": 0.5715, "step": 6528 }, { "epoch": 0.56, "learning_rate": 4.32014889144998e-06, "loss": 0.8123, "step": 6529 }, { "epoch": 0.56, "learning_rate": 4.318779859437774e-06, "loss": 0.8511, "step": 6530 }, { "epoch": 0.56, "learning_rate": 4.3174108794605595e-06, "loss": 0.7949, "step": 6531 }, { "epoch": 0.56, "learning_rate": 4.316041951622908e-06, "loss": 0.8638, "step": 6532 }, { "epoch": 0.56, "learning_rate": 4.314673076029385e-06, "loss": 0.793, "step": 6533 }, { "epoch": 0.56, "learning_rate": 4.313304252784551e-06, "loss": 0.8193, "step": 6534 }, { "epoch": 0.56, "learning_rate": 4.311935481992965e-06, "loss": 0.7966, "step": 6535 }, { "epoch": 0.56, "learning_rate": 4.31056676375918e-06, "loss": 0.8386, "step": 6536 }, { "epoch": 0.56, "learning_rate": 4.309198098187747e-06, "loss": 0.8456, "step": 6537 }, { "epoch": 0.56, "learning_rate": 4.307829485383207e-06, "loss": 0.7997, "step": 6538 }, { "epoch": 0.56, "learning_rate": 4.3064609254501055e-06, "loss": 0.8446, "step": 6539 }, { "epoch": 0.56, "learning_rate": 4.30509241849298e-06, "loss": 0.8612, "step": 6540 }, { "epoch": 0.56, "learning_rate": 4.3037239646163606e-06, "loss": 0.8036, "step": 6541 }, { "epoch": 0.56, "learning_rate": 4.302355563924779e-06, "loss": 0.8421, "step": 6542 }, { "epoch": 0.56, "learning_rate": 4.3009872165227635e-06, "loss": 0.7787, "step": 6543 }, { "epoch": 0.56, "learning_rate": 4.2996189225148284e-06, "loss": 0.8279, "step": 6544 }, { "epoch": 0.56, "learning_rate": 4.298250682005496e-06, "loss": 0.7951, "step": 6545 }, { "epoch": 0.56, "learning_rate": 4.296882495099278e-06, "loss": 0.8533, "step": 6546 }, { "epoch": 0.56, "learning_rate": 4.295514361900682e-06, "loss": 0.8479, "step": 6547 }, { "epoch": 0.56, "learning_rate": 4.294146282514214e-06, "loss": 0.8414, "step": 6548 }, { "epoch": 0.56, "learning_rate": 4.292778257044377e-06, "loss": 0.8201, "step": 6549 }, { "epoch": 0.56, "learning_rate": 4.291410285595664e-06, "loss": 0.803, "step": 6550 }, { "epoch": 0.56, "learning_rate": 4.290042368272569e-06, "loss": 0.8442, "step": 6551 }, { "epoch": 0.56, "learning_rate": 4.28867450517958e-06, "loss": 0.7917, "step": 6552 }, { "epoch": 0.56, "learning_rate": 4.2873066964211805e-06, "loss": 0.8557, "step": 6553 }, { "epoch": 0.56, "learning_rate": 4.285938942101852e-06, "loss": 0.7806, "step": 6554 }, { "epoch": 0.56, "learning_rate": 4.28457124232607e-06, "loss": 0.8397, "step": 6555 }, { "epoch": 0.56, "learning_rate": 4.283203597198306e-06, "loss": 0.8359, "step": 6556 }, { "epoch": 0.56, "learning_rate": 4.281836006823029e-06, "loss": 0.8248, "step": 6557 }, { "epoch": 0.56, "learning_rate": 4.2804684713046985e-06, "loss": 0.8538, "step": 6558 }, { "epoch": 0.56, "learning_rate": 4.2791009907477774e-06, "loss": 0.8184, "step": 6559 }, { "epoch": 0.56, "learning_rate": 4.277733565256719e-06, "loss": 0.8225, "step": 6560 }, { "epoch": 0.56, "learning_rate": 4.276366194935975e-06, "loss": 0.8591, "step": 6561 }, { "epoch": 0.56, "learning_rate": 4.274998879889991e-06, "loss": 0.7812, "step": 6562 }, { "epoch": 0.56, "learning_rate": 4.273631620223212e-06, "loss": 0.7359, "step": 6563 }, { "epoch": 0.56, "learning_rate": 4.272264416040071e-06, "loss": 0.831, "step": 6564 }, { "epoch": 0.56, "learning_rate": 4.270897267445007e-06, "loss": 0.8266, "step": 6565 }, { "epoch": 0.56, "learning_rate": 4.269530174542446e-06, "loss": 0.8298, "step": 6566 }, { "epoch": 0.56, "learning_rate": 4.2681631374368145e-06, "loss": 0.7896, "step": 6567 }, { "epoch": 0.56, "learning_rate": 4.266796156232536e-06, "loss": 0.851, "step": 6568 }, { "epoch": 0.56, "learning_rate": 4.265429231034024e-06, "loss": 0.8091, "step": 6569 }, { "epoch": 0.56, "learning_rate": 4.264062361945694e-06, "loss": 0.8135, "step": 6570 }, { "epoch": 0.56, "learning_rate": 4.262695549071951e-06, "loss": 0.8323, "step": 6571 }, { "epoch": 0.56, "learning_rate": 4.261328792517201e-06, "loss": 0.8176, "step": 6572 }, { "epoch": 0.56, "learning_rate": 4.259962092385844e-06, "loss": 0.828, "step": 6573 }, { "epoch": 0.56, "learning_rate": 4.258595448782274e-06, "loss": 0.8087, "step": 6574 }, { "epoch": 0.56, "learning_rate": 4.257228861810882e-06, "loss": 0.86, "step": 6575 }, { "epoch": 0.56, "learning_rate": 4.255862331576058e-06, "loss": 0.7798, "step": 6576 }, { "epoch": 0.56, "learning_rate": 4.2544958581821795e-06, "loss": 0.8241, "step": 6577 }, { "epoch": 0.56, "learning_rate": 4.253129441733628e-06, "loss": 0.8315, "step": 6578 }, { "epoch": 0.56, "learning_rate": 4.251763082334776e-06, "loss": 0.7912, "step": 6579 }, { "epoch": 0.56, "learning_rate": 4.250396780089992e-06, "loss": 0.8073, "step": 6580 }, { "epoch": 0.56, "learning_rate": 4.249030535103643e-06, "loss": 0.8489, "step": 6581 }, { "epoch": 0.56, "learning_rate": 4.2476643474800885e-06, "loss": 0.8275, "step": 6582 }, { "epoch": 0.56, "learning_rate": 4.246298217323685e-06, "loss": 0.7839, "step": 6583 }, { "epoch": 0.56, "learning_rate": 4.244932144738785e-06, "loss": 0.8133, "step": 6584 }, { "epoch": 0.56, "learning_rate": 4.243566129829734e-06, "loss": 0.8024, "step": 6585 }, { "epoch": 0.56, "learning_rate": 4.242200172700876e-06, "loss": 0.8404, "step": 6586 }, { "epoch": 0.56, "learning_rate": 4.24083427345655e-06, "loss": 0.7559, "step": 6587 }, { "epoch": 0.56, "learning_rate": 4.2394684322010895e-06, "loss": 0.529, "step": 6588 }, { "epoch": 0.56, "learning_rate": 4.238102649038825e-06, "loss": 0.8455, "step": 6589 }, { "epoch": 0.56, "learning_rate": 4.236736924074083e-06, "loss": 0.8541, "step": 6590 }, { "epoch": 0.56, "learning_rate": 4.2353712574111815e-06, "loss": 0.826, "step": 6591 }, { "epoch": 0.56, "learning_rate": 4.23400564915444e-06, "loss": 0.8651, "step": 6592 }, { "epoch": 0.56, "learning_rate": 4.232640099408167e-06, "loss": 0.8555, "step": 6593 }, { "epoch": 0.56, "learning_rate": 4.231274608276671e-06, "loss": 0.7193, "step": 6594 }, { "epoch": 0.56, "learning_rate": 4.229909175864257e-06, "loss": 0.8315, "step": 6595 }, { "epoch": 0.56, "learning_rate": 4.228543802275224e-06, "loss": 0.8486, "step": 6596 }, { "epoch": 0.56, "learning_rate": 4.227178487613865e-06, "loss": 0.828, "step": 6597 }, { "epoch": 0.56, "learning_rate": 4.225813231984466e-06, "loss": 0.871, "step": 6598 }, { "epoch": 0.56, "learning_rate": 4.224448035491317e-06, "loss": 0.8016, "step": 6599 }, { "epoch": 0.56, "learning_rate": 4.223082898238695e-06, "loss": 0.8396, "step": 6600 }, { "epoch": 0.56, "learning_rate": 4.221717820330879e-06, "loss": 0.783, "step": 6601 }, { "epoch": 0.56, "learning_rate": 4.220352801872137e-06, "loss": 0.8065, "step": 6602 }, { "epoch": 0.56, "learning_rate": 4.218987842966742e-06, "loss": 0.8137, "step": 6603 }, { "epoch": 0.56, "learning_rate": 4.217622943718949e-06, "loss": 0.7877, "step": 6604 }, { "epoch": 0.56, "learning_rate": 4.216258104233019e-06, "loss": 0.8512, "step": 6605 }, { "epoch": 0.56, "learning_rate": 4.214893324613207e-06, "loss": 0.8391, "step": 6606 }, { "epoch": 0.56, "learning_rate": 4.2135286049637596e-06, "loss": 0.7889, "step": 6607 }, { "epoch": 0.56, "learning_rate": 4.21216394538892e-06, "loss": 0.8473, "step": 6608 }, { "epoch": 0.56, "learning_rate": 4.210799345992929e-06, "loss": 0.785, "step": 6609 }, { "epoch": 0.56, "learning_rate": 4.209434806880025e-06, "loss": 0.824, "step": 6610 }, { "epoch": 0.56, "learning_rate": 4.208070328154433e-06, "loss": 0.8191, "step": 6611 }, { "epoch": 0.56, "learning_rate": 4.20670590992038e-06, "loss": 0.8415, "step": 6612 }, { "epoch": 0.56, "learning_rate": 4.2053415522820875e-06, "loss": 0.8254, "step": 6613 }, { "epoch": 0.56, "learning_rate": 4.203977255343772e-06, "loss": 0.8127, "step": 6614 }, { "epoch": 0.56, "learning_rate": 4.202613019209647e-06, "loss": 0.7922, "step": 6615 }, { "epoch": 0.56, "learning_rate": 4.2012488439839185e-06, "loss": 0.8311, "step": 6616 }, { "epoch": 0.56, "learning_rate": 4.199884729770787e-06, "loss": 0.8406, "step": 6617 }, { "epoch": 0.56, "learning_rate": 4.1985206766744506e-06, "loss": 0.7866, "step": 6618 }, { "epoch": 0.56, "learning_rate": 4.1971566847991055e-06, "loss": 0.8304, "step": 6619 }, { "epoch": 0.56, "learning_rate": 4.195792754248938e-06, "loss": 0.7986, "step": 6620 }, { "epoch": 0.56, "learning_rate": 4.194428885128132e-06, "loss": 0.8178, "step": 6621 }, { "epoch": 0.57, "learning_rate": 4.193065077540868e-06, "loss": 0.8506, "step": 6622 }, { "epoch": 0.57, "learning_rate": 4.19170133159132e-06, "loss": 0.8002, "step": 6623 }, { "epoch": 0.57, "learning_rate": 4.190337647383656e-06, "loss": 0.7978, "step": 6624 }, { "epoch": 0.57, "learning_rate": 4.188974025022043e-06, "loss": 0.8212, "step": 6625 }, { "epoch": 0.57, "learning_rate": 4.187610464610641e-06, "loss": 0.7696, "step": 6626 }, { "epoch": 0.57, "learning_rate": 4.1862469662536035e-06, "loss": 0.8584, "step": 6627 }, { "epoch": 0.57, "learning_rate": 4.184883530055082e-06, "loss": 0.8055, "step": 6628 }, { "epoch": 0.57, "learning_rate": 4.1835201561192255e-06, "loss": 0.833, "step": 6629 }, { "epoch": 0.57, "learning_rate": 4.182156844550175e-06, "loss": 0.8539, "step": 6630 }, { "epoch": 0.57, "learning_rate": 4.180793595452062e-06, "loss": 0.8519, "step": 6631 }, { "epoch": 0.57, "learning_rate": 4.179430408929022e-06, "loss": 0.789, "step": 6632 }, { "epoch": 0.57, "learning_rate": 4.1780672850851815e-06, "loss": 0.5513, "step": 6633 }, { "epoch": 0.57, "learning_rate": 4.176704224024663e-06, "loss": 0.8092, "step": 6634 }, { "epoch": 0.57, "learning_rate": 4.175341225851584e-06, "loss": 0.8883, "step": 6635 }, { "epoch": 0.57, "learning_rate": 4.173978290670057e-06, "loss": 0.5505, "step": 6636 }, { "epoch": 0.57, "learning_rate": 4.17261541858419e-06, "loss": 0.8048, "step": 6637 }, { "epoch": 0.57, "learning_rate": 4.171252609698084e-06, "loss": 0.8716, "step": 6638 }, { "epoch": 0.57, "learning_rate": 4.169889864115839e-06, "loss": 0.8257, "step": 6639 }, { "epoch": 0.57, "learning_rate": 4.168527181941548e-06, "loss": 0.821, "step": 6640 }, { "epoch": 0.57, "learning_rate": 4.167164563279298e-06, "loss": 0.8543, "step": 6641 }, { "epoch": 0.57, "learning_rate": 4.165802008233176e-06, "loss": 0.7948, "step": 6642 }, { "epoch": 0.57, "learning_rate": 4.164439516907258e-06, "loss": 0.8325, "step": 6643 }, { "epoch": 0.57, "learning_rate": 4.16307708940562e-06, "loss": 0.8333, "step": 6644 }, { "epoch": 0.57, "learning_rate": 4.161714725832328e-06, "loss": 0.7953, "step": 6645 }, { "epoch": 0.57, "learning_rate": 4.160352426291447e-06, "loss": 0.8491, "step": 6646 }, { "epoch": 0.57, "learning_rate": 4.158990190887038e-06, "loss": 0.8442, "step": 6647 }, { "epoch": 0.57, "learning_rate": 4.157628019723155e-06, "loss": 0.7779, "step": 6648 }, { "epoch": 0.57, "learning_rate": 4.156265912903848e-06, "loss": 0.8377, "step": 6649 }, { "epoch": 0.57, "learning_rate": 4.154903870533158e-06, "loss": 0.8369, "step": 6650 }, { "epoch": 0.57, "learning_rate": 4.153541892715126e-06, "loss": 0.8121, "step": 6651 }, { "epoch": 0.57, "learning_rate": 4.1521799795537894e-06, "loss": 0.7736, "step": 6652 }, { "epoch": 0.57, "learning_rate": 4.150818131153176e-06, "loss": 0.8182, "step": 6653 }, { "epoch": 0.57, "learning_rate": 4.14945634761731e-06, "loss": 0.8585, "step": 6654 }, { "epoch": 0.57, "learning_rate": 4.148094629050211e-06, "loss": 0.8162, "step": 6655 }, { "epoch": 0.57, "learning_rate": 4.146732975555898e-06, "loss": 0.81, "step": 6656 }, { "epoch": 0.57, "learning_rate": 4.145371387238375e-06, "loss": 0.7936, "step": 6657 }, { "epoch": 0.57, "learning_rate": 4.1440098642016504e-06, "loss": 0.8344, "step": 6658 }, { "epoch": 0.57, "learning_rate": 4.142648406549722e-06, "loss": 0.8369, "step": 6659 }, { "epoch": 0.57, "learning_rate": 4.141287014386587e-06, "loss": 0.8724, "step": 6660 }, { "epoch": 0.57, "learning_rate": 4.1399256878162345e-06, "loss": 0.8347, "step": 6661 }, { "epoch": 0.57, "learning_rate": 4.1385644269426486e-06, "loss": 0.8488, "step": 6662 }, { "epoch": 0.57, "learning_rate": 4.137203231869812e-06, "loss": 0.7897, "step": 6663 }, { "epoch": 0.57, "learning_rate": 4.135842102701694e-06, "loss": 0.8026, "step": 6664 }, { "epoch": 0.57, "learning_rate": 4.134481039542268e-06, "loss": 0.7905, "step": 6665 }, { "epoch": 0.57, "learning_rate": 4.1331200424955e-06, "loss": 0.8597, "step": 6666 }, { "epoch": 0.57, "learning_rate": 4.131759111665349e-06, "loss": 0.8416, "step": 6667 }, { "epoch": 0.57, "learning_rate": 4.130398247155767e-06, "loss": 0.8188, "step": 6668 }, { "epoch": 0.57, "learning_rate": 4.129037449070709e-06, "loss": 0.8126, "step": 6669 }, { "epoch": 0.57, "learning_rate": 4.127676717514114e-06, "loss": 0.8005, "step": 6670 }, { "epoch": 0.57, "learning_rate": 4.126316052589923e-06, "loss": 0.8062, "step": 6671 }, { "epoch": 0.57, "learning_rate": 4.124955454402072e-06, "loss": 0.8217, "step": 6672 }, { "epoch": 0.57, "learning_rate": 4.123594923054489e-06, "loss": 0.5166, "step": 6673 }, { "epoch": 0.57, "learning_rate": 4.1222344586510985e-06, "loss": 0.8704, "step": 6674 }, { "epoch": 0.57, "learning_rate": 4.12087406129582e-06, "loss": 0.8156, "step": 6675 }, { "epoch": 0.57, "learning_rate": 4.119513731092567e-06, "loss": 0.7944, "step": 6676 }, { "epoch": 0.57, "learning_rate": 4.118153468145248e-06, "loss": 0.7996, "step": 6677 }, { "epoch": 0.57, "learning_rate": 4.116793272557767e-06, "loss": 0.7863, "step": 6678 }, { "epoch": 0.57, "learning_rate": 4.115433144434023e-06, "loss": 0.8283, "step": 6679 }, { "epoch": 0.57, "learning_rate": 4.114073083877908e-06, "loss": 0.854, "step": 6680 }, { "epoch": 0.57, "learning_rate": 4.112713090993311e-06, "loss": 0.8391, "step": 6681 }, { "epoch": 0.57, "learning_rate": 4.111353165884116e-06, "loss": 0.86, "step": 6682 }, { "epoch": 0.57, "learning_rate": 4.1099933086542e-06, "loss": 0.8258, "step": 6683 }, { "epoch": 0.57, "learning_rate": 4.108633519407435e-06, "loss": 0.8163, "step": 6684 }, { "epoch": 0.57, "learning_rate": 4.107273798247689e-06, "loss": 0.8086, "step": 6685 }, { "epoch": 0.57, "learning_rate": 4.105914145278825e-06, "loss": 0.801, "step": 6686 }, { "epoch": 0.57, "learning_rate": 4.104554560604699e-06, "loss": 0.8097, "step": 6687 }, { "epoch": 0.57, "learning_rate": 4.103195044329163e-06, "loss": 0.8364, "step": 6688 }, { "epoch": 0.57, "learning_rate": 4.101835596556067e-06, "loss": 0.8305, "step": 6689 }, { "epoch": 0.57, "learning_rate": 4.100476217389247e-06, "loss": 0.8015, "step": 6690 }, { "epoch": 0.57, "learning_rate": 4.099116906932542e-06, "loss": 0.7893, "step": 6691 }, { "epoch": 0.57, "learning_rate": 4.097757665289782e-06, "loss": 0.8052, "step": 6692 }, { "epoch": 0.57, "learning_rate": 4.096398492564793e-06, "loss": 0.8626, "step": 6693 }, { "epoch": 0.57, "learning_rate": 4.095039388861397e-06, "loss": 0.7813, "step": 6694 }, { "epoch": 0.57, "learning_rate": 4.093680354283406e-06, "loss": 0.7869, "step": 6695 }, { "epoch": 0.57, "learning_rate": 4.092321388934634e-06, "loss": 0.8343, "step": 6696 }, { "epoch": 0.57, "learning_rate": 4.090962492918881e-06, "loss": 0.8677, "step": 6697 }, { "epoch": 0.57, "learning_rate": 4.089603666339948e-06, "loss": 0.8193, "step": 6698 }, { "epoch": 0.57, "learning_rate": 4.08824490930163e-06, "loss": 0.849, "step": 6699 }, { "epoch": 0.57, "learning_rate": 4.086886221907715e-06, "loss": 0.7959, "step": 6700 }, { "epoch": 0.57, "learning_rate": 4.0855276042619855e-06, "loss": 0.7642, "step": 6701 }, { "epoch": 0.57, "learning_rate": 4.0841690564682205e-06, "loss": 0.7783, "step": 6702 }, { "epoch": 0.57, "learning_rate": 4.082810578630194e-06, "loss": 0.8056, "step": 6703 }, { "epoch": 0.57, "learning_rate": 4.081452170851671e-06, "loss": 0.7698, "step": 6704 }, { "epoch": 0.57, "learning_rate": 4.080093833236414e-06, "loss": 0.8382, "step": 6705 }, { "epoch": 0.57, "learning_rate": 4.078735565888178e-06, "loss": 0.8382, "step": 6706 }, { "epoch": 0.57, "learning_rate": 4.077377368910717e-06, "loss": 0.8174, "step": 6707 }, { "epoch": 0.57, "learning_rate": 4.0760192424077775e-06, "loss": 0.4827, "step": 6708 }, { "epoch": 0.57, "learning_rate": 4.074661186483097e-06, "loss": 0.7876, "step": 6709 }, { "epoch": 0.57, "learning_rate": 4.073303201240413e-06, "loss": 0.8239, "step": 6710 }, { "epoch": 0.57, "learning_rate": 4.071945286783453e-06, "loss": 0.8264, "step": 6711 }, { "epoch": 0.57, "learning_rate": 4.0705874432159425e-06, "loss": 0.8351, "step": 6712 }, { "epoch": 0.57, "learning_rate": 4.069229670641601e-06, "loss": 0.8111, "step": 6713 }, { "epoch": 0.57, "learning_rate": 4.0678719691641405e-06, "loss": 0.7689, "step": 6714 }, { "epoch": 0.57, "learning_rate": 4.06651433888727e-06, "loss": 0.7949, "step": 6715 }, { "epoch": 0.57, "learning_rate": 4.065156779914693e-06, "loss": 0.7861, "step": 6716 }, { "epoch": 0.57, "learning_rate": 4.063799292350104e-06, "loss": 0.8239, "step": 6717 }, { "epoch": 0.57, "learning_rate": 4.062441876297197e-06, "loss": 0.8438, "step": 6718 }, { "epoch": 0.57, "learning_rate": 4.061084531859656e-06, "loss": 0.8515, "step": 6719 }, { "epoch": 0.57, "learning_rate": 4.059727259141163e-06, "loss": 0.7786, "step": 6720 }, { "epoch": 0.57, "learning_rate": 4.0583700582453935e-06, "loss": 0.7984, "step": 6721 }, { "epoch": 0.57, "learning_rate": 4.057012929276019e-06, "loss": 0.8054, "step": 6722 }, { "epoch": 0.57, "learning_rate": 4.055655872336702e-06, "loss": 0.884, "step": 6723 }, { "epoch": 0.57, "learning_rate": 4.054298887531099e-06, "loss": 0.8148, "step": 6724 }, { "epoch": 0.57, "learning_rate": 4.052941974962866e-06, "loss": 0.9104, "step": 6725 }, { "epoch": 0.57, "learning_rate": 4.05158513473565e-06, "loss": 0.8191, "step": 6726 }, { "epoch": 0.57, "learning_rate": 4.0502283669530946e-06, "loss": 0.569, "step": 6727 }, { "epoch": 0.57, "learning_rate": 4.048871671718833e-06, "loss": 0.8063, "step": 6728 }, { "epoch": 0.57, "learning_rate": 4.047515049136503e-06, "loss": 0.8625, "step": 6729 }, { "epoch": 0.57, "learning_rate": 4.0461584993097225e-06, "loss": 0.7956, "step": 6730 }, { "epoch": 0.57, "learning_rate": 4.044802022342115e-06, "loss": 0.8905, "step": 6731 }, { "epoch": 0.57, "learning_rate": 4.043445618337296e-06, "loss": 0.8159, "step": 6732 }, { "epoch": 0.57, "learning_rate": 4.0420892873988725e-06, "loss": 0.8198, "step": 6733 }, { "epoch": 0.57, "learning_rate": 4.040733029630448e-06, "loss": 0.8291, "step": 6734 }, { "epoch": 0.57, "learning_rate": 4.039376845135621e-06, "loss": 0.8017, "step": 6735 }, { "epoch": 0.57, "learning_rate": 4.038020734017986e-06, "loss": 0.8414, "step": 6736 }, { "epoch": 0.57, "learning_rate": 4.0366646963811255e-06, "loss": 0.8298, "step": 6737 }, { "epoch": 0.57, "learning_rate": 4.035308732328621e-06, "loss": 0.8056, "step": 6738 }, { "epoch": 0.58, "learning_rate": 4.033952841964049e-06, "loss": 0.7819, "step": 6739 }, { "epoch": 0.58, "learning_rate": 4.032597025390979e-06, "loss": 0.834, "step": 6740 }, { "epoch": 0.58, "learning_rate": 4.031241282712976e-06, "loss": 0.8361, "step": 6741 }, { "epoch": 0.58, "learning_rate": 4.029885614033598e-06, "loss": 0.8407, "step": 6742 }, { "epoch": 0.58, "learning_rate": 4.028530019456395e-06, "loss": 0.7991, "step": 6743 }, { "epoch": 0.58, "learning_rate": 4.027174499084917e-06, "loss": 0.8196, "step": 6744 }, { "epoch": 0.58, "learning_rate": 4.025819053022704e-06, "loss": 0.8511, "step": 6745 }, { "epoch": 0.58, "learning_rate": 4.0244636813732925e-06, "loss": 0.8165, "step": 6746 }, { "epoch": 0.58, "learning_rate": 4.023108384240212e-06, "loss": 0.8657, "step": 6747 }, { "epoch": 0.58, "learning_rate": 4.021753161726988e-06, "loss": 0.7622, "step": 6748 }, { "epoch": 0.58, "learning_rate": 4.0203980139371394e-06, "loss": 0.7872, "step": 6749 }, { "epoch": 0.58, "learning_rate": 4.019042940974176e-06, "loss": 0.8262, "step": 6750 }, { "epoch": 0.58, "learning_rate": 4.017687942941609e-06, "loss": 0.8694, "step": 6751 }, { "epoch": 0.58, "learning_rate": 4.016333019942936e-06, "loss": 0.8532, "step": 6752 }, { "epoch": 0.58, "learning_rate": 4.014978172081655e-06, "loss": 0.7913, "step": 6753 }, { "epoch": 0.58, "learning_rate": 4.013623399461257e-06, "loss": 0.8483, "step": 6754 }, { "epoch": 0.58, "learning_rate": 4.012268702185226e-06, "loss": 0.8485, "step": 6755 }, { "epoch": 0.58, "learning_rate": 4.01091408035704e-06, "loss": 0.827, "step": 6756 }, { "epoch": 0.58, "learning_rate": 4.009559534080169e-06, "loss": 0.8239, "step": 6757 }, { "epoch": 0.58, "learning_rate": 4.0082050634580825e-06, "loss": 0.8081, "step": 6758 }, { "epoch": 0.58, "learning_rate": 4.006850668594243e-06, "loss": 0.7886, "step": 6759 }, { "epoch": 0.58, "learning_rate": 4.005496349592104e-06, "loss": 0.8285, "step": 6760 }, { "epoch": 0.58, "learning_rate": 4.004142106555117e-06, "loss": 0.8452, "step": 6761 }, { "epoch": 0.58, "learning_rate": 4.002787939586725e-06, "loss": 0.7909, "step": 6762 }, { "epoch": 0.58, "learning_rate": 4.001433848790363e-06, "loss": 0.7984, "step": 6763 }, { "epoch": 0.58, "learning_rate": 4.000079834269468e-06, "loss": 0.4757, "step": 6764 }, { "epoch": 0.58, "learning_rate": 3.9987258961274635e-06, "loss": 0.777, "step": 6765 }, { "epoch": 0.58, "learning_rate": 3.997372034467771e-06, "loss": 0.844, "step": 6766 }, { "epoch": 0.58, "learning_rate": 3.996018249393803e-06, "loss": 0.8691, "step": 6767 }, { "epoch": 0.58, "learning_rate": 3.994664541008973e-06, "loss": 0.7961, "step": 6768 }, { "epoch": 0.58, "learning_rate": 3.993310909416682e-06, "loss": 0.8383, "step": 6769 }, { "epoch": 0.58, "learning_rate": 3.9919573547203255e-06, "loss": 0.8424, "step": 6770 }, { "epoch": 0.58, "learning_rate": 3.990603877023295e-06, "loss": 0.8166, "step": 6771 }, { "epoch": 0.58, "learning_rate": 3.989250476428976e-06, "loss": 0.7803, "step": 6772 }, { "epoch": 0.58, "learning_rate": 3.98789715304075e-06, "loss": 0.8446, "step": 6773 }, { "epoch": 0.58, "learning_rate": 3.986543906961989e-06, "loss": 0.8538, "step": 6774 }, { "epoch": 0.58, "learning_rate": 3.985190738296062e-06, "loss": 0.8054, "step": 6775 }, { "epoch": 0.58, "learning_rate": 3.983837647146329e-06, "loss": 0.8364, "step": 6776 }, { "epoch": 0.58, "learning_rate": 3.9824846336161445e-06, "loss": 0.8585, "step": 6777 }, { "epoch": 0.58, "learning_rate": 3.981131697808862e-06, "loss": 0.8055, "step": 6778 }, { "epoch": 0.58, "learning_rate": 3.979778839827824e-06, "loss": 0.7986, "step": 6779 }, { "epoch": 0.58, "learning_rate": 3.978426059776369e-06, "loss": 0.7746, "step": 6780 }, { "epoch": 0.58, "learning_rate": 3.977073357757827e-06, "loss": 0.834, "step": 6781 }, { "epoch": 0.58, "learning_rate": 3.9757207338755284e-06, "loss": 0.8318, "step": 6782 }, { "epoch": 0.58, "learning_rate": 3.974368188232788e-06, "loss": 0.8219, "step": 6783 }, { "epoch": 0.58, "learning_rate": 3.973015720932925e-06, "loss": 0.8066, "step": 6784 }, { "epoch": 0.58, "learning_rate": 3.971663332079243e-06, "loss": 0.8581, "step": 6785 }, { "epoch": 0.58, "learning_rate": 3.970311021775047e-06, "loss": 0.8161, "step": 6786 }, { "epoch": 0.58, "learning_rate": 3.968958790123632e-06, "loss": 0.8053, "step": 6787 }, { "epoch": 0.58, "learning_rate": 3.967606637228291e-06, "loss": 0.8153, "step": 6788 }, { "epoch": 0.58, "learning_rate": 3.966254563192306e-06, "loss": 0.8, "step": 6789 }, { "epoch": 0.58, "learning_rate": 3.964902568118955e-06, "loss": 0.5353, "step": 6790 }, { "epoch": 0.58, "learning_rate": 3.963550652111508e-06, "loss": 0.8992, "step": 6791 }, { "epoch": 0.58, "learning_rate": 3.962198815273235e-06, "loss": 0.8162, "step": 6792 }, { "epoch": 0.58, "learning_rate": 3.960847057707395e-06, "loss": 0.8573, "step": 6793 }, { "epoch": 0.58, "learning_rate": 3.959495379517241e-06, "loss": 0.8185, "step": 6794 }, { "epoch": 0.58, "learning_rate": 3.958143780806022e-06, "loss": 0.842, "step": 6795 }, { "epoch": 0.58, "learning_rate": 3.956792261676979e-06, "loss": 0.7705, "step": 6796 }, { "epoch": 0.58, "learning_rate": 3.955440822233347e-06, "loss": 0.87, "step": 6797 }, { "epoch": 0.58, "learning_rate": 3.9540894625783576e-06, "loss": 0.7901, "step": 6798 }, { "epoch": 0.58, "learning_rate": 3.952738182815233e-06, "loss": 0.7945, "step": 6799 }, { "epoch": 0.58, "learning_rate": 3.951386983047191e-06, "loss": 0.5301, "step": 6800 }, { "epoch": 0.58, "learning_rate": 3.9500358633774434e-06, "loss": 0.775, "step": 6801 }, { "epoch": 0.58, "learning_rate": 3.948684823909197e-06, "loss": 0.8056, "step": 6802 }, { "epoch": 0.58, "learning_rate": 3.947333864745648e-06, "loss": 0.7982, "step": 6803 }, { "epoch": 0.58, "learning_rate": 3.94598298598999e-06, "loss": 0.8156, "step": 6804 }, { "epoch": 0.58, "learning_rate": 3.94463218774541e-06, "loss": 0.8095, "step": 6805 }, { "epoch": 0.58, "learning_rate": 3.943281470115089e-06, "loss": 0.8676, "step": 6806 }, { "epoch": 0.58, "learning_rate": 3.941930833202203e-06, "loss": 0.7642, "step": 6807 }, { "epoch": 0.58, "learning_rate": 3.940580277109918e-06, "loss": 0.8236, "step": 6808 }, { "epoch": 0.58, "learning_rate": 3.9392298019413995e-06, "loss": 0.8385, "step": 6809 }, { "epoch": 0.58, "learning_rate": 3.9378794077997985e-06, "loss": 0.8183, "step": 6810 }, { "epoch": 0.58, "learning_rate": 3.9365290947882686e-06, "loss": 0.8007, "step": 6811 }, { "epoch": 0.58, "learning_rate": 3.935178863009954e-06, "loss": 0.8354, "step": 6812 }, { "epoch": 0.58, "learning_rate": 3.9338287125679884e-06, "loss": 0.7957, "step": 6813 }, { "epoch": 0.58, "learning_rate": 3.932478643565506e-06, "loss": 0.8405, "step": 6814 }, { "epoch": 0.58, "learning_rate": 3.931128656105633e-06, "loss": 0.8252, "step": 6815 }, { "epoch": 0.58, "learning_rate": 3.929778750291484e-06, "loss": 0.8506, "step": 6816 }, { "epoch": 0.58, "learning_rate": 3.928428926226175e-06, "loss": 0.8316, "step": 6817 }, { "epoch": 0.58, "learning_rate": 3.92707918401281e-06, "loss": 0.785, "step": 6818 }, { "epoch": 0.58, "learning_rate": 3.92572952375449e-06, "loss": 0.7705, "step": 6819 }, { "epoch": 0.58, "learning_rate": 3.92437994555431e-06, "loss": 0.8769, "step": 6820 }, { "epoch": 0.58, "learning_rate": 3.923030449515356e-06, "loss": 0.8528, "step": 6821 }, { "epoch": 0.58, "learning_rate": 3.921681035740711e-06, "loss": 0.7735, "step": 6822 }, { "epoch": 0.58, "learning_rate": 3.920331704333447e-06, "loss": 0.8168, "step": 6823 }, { "epoch": 0.58, "learning_rate": 3.918982455396634e-06, "loss": 0.7852, "step": 6824 }, { "epoch": 0.58, "learning_rate": 3.917633289033335e-06, "loss": 0.8328, "step": 6825 }, { "epoch": 0.58, "learning_rate": 3.916284205346606e-06, "loss": 0.8529, "step": 6826 }, { "epoch": 0.58, "learning_rate": 3.914935204439496e-06, "loss": 0.5962, "step": 6827 }, { "epoch": 0.58, "learning_rate": 3.913586286415049e-06, "loss": 0.8402, "step": 6828 }, { "epoch": 0.58, "learning_rate": 3.9122374513763046e-06, "loss": 0.8889, "step": 6829 }, { "epoch": 0.58, "learning_rate": 3.910888699426289e-06, "loss": 0.8296, "step": 6830 }, { "epoch": 0.58, "learning_rate": 3.909540030668028e-06, "loss": 0.7579, "step": 6831 }, { "epoch": 0.58, "learning_rate": 3.90819144520454e-06, "loss": 0.7434, "step": 6832 }, { "epoch": 0.58, "learning_rate": 3.906842943138836e-06, "loss": 0.8217, "step": 6833 }, { "epoch": 0.58, "learning_rate": 3.905494524573925e-06, "loss": 0.8327, "step": 6834 }, { "epoch": 0.58, "learning_rate": 3.904146189612802e-06, "loss": 0.5414, "step": 6835 }, { "epoch": 0.58, "learning_rate": 3.902797938358461e-06, "loss": 0.8208, "step": 6836 }, { "epoch": 0.58, "learning_rate": 3.901449770913886e-06, "loss": 0.85, "step": 6837 }, { "epoch": 0.58, "learning_rate": 3.90010168738206e-06, "loss": 0.8462, "step": 6838 }, { "epoch": 0.58, "learning_rate": 3.898753687865956e-06, "loss": 0.8141, "step": 6839 }, { "epoch": 0.58, "learning_rate": 3.8974057724685375e-06, "loss": 0.8207, "step": 6840 }, { "epoch": 0.58, "learning_rate": 3.896057941292768e-06, "loss": 0.7994, "step": 6841 }, { "epoch": 0.58, "learning_rate": 3.894710194441603e-06, "loss": 0.7874, "step": 6842 }, { "epoch": 0.58, "learning_rate": 3.893362532017986e-06, "loss": 0.8438, "step": 6843 }, { "epoch": 0.58, "learning_rate": 3.892014954124861e-06, "loss": 0.8048, "step": 6844 }, { "epoch": 0.58, "learning_rate": 3.890667460865161e-06, "loss": 0.7577, "step": 6845 }, { "epoch": 0.58, "learning_rate": 3.8893200523418156e-06, "loss": 0.8319, "step": 6846 }, { "epoch": 0.58, "learning_rate": 3.8879727286577465e-06, "loss": 0.8317, "step": 6847 }, { "epoch": 0.58, "learning_rate": 3.88662548991587e-06, "loss": 0.8545, "step": 6848 }, { "epoch": 0.58, "learning_rate": 3.885278336219092e-06, "loss": 0.8047, "step": 6849 }, { "epoch": 0.58, "learning_rate": 3.8839312676703175e-06, "loss": 0.8209, "step": 6850 }, { "epoch": 0.58, "learning_rate": 3.8825842843724404e-06, "loss": 0.8303, "step": 6851 }, { "epoch": 0.58, "learning_rate": 3.881237386428351e-06, "loss": 0.8434, "step": 6852 }, { "epoch": 0.58, "learning_rate": 3.879890573940932e-06, "loss": 0.8541, "step": 6853 }, { "epoch": 0.58, "learning_rate": 3.878543847013059e-06, "loss": 0.7804, "step": 6854 }, { "epoch": 0.58, "learning_rate": 3.877197205747605e-06, "loss": 0.832, "step": 6855 }, { "epoch": 0.59, "learning_rate": 3.8758506502474284e-06, "loss": 0.8302, "step": 6856 }, { "epoch": 0.59, "learning_rate": 3.8745041806153875e-06, "loss": 0.7929, "step": 6857 }, { "epoch": 0.59, "learning_rate": 3.873157796954334e-06, "loss": 0.7881, "step": 6858 }, { "epoch": 0.59, "learning_rate": 3.8718114993671086e-06, "loss": 0.5862, "step": 6859 }, { "epoch": 0.59, "learning_rate": 3.8704652879565505e-06, "loss": 0.8015, "step": 6860 }, { "epoch": 0.59, "learning_rate": 3.86911916282549e-06, "loss": 0.8277, "step": 6861 }, { "epoch": 0.59, "learning_rate": 3.867773124076751e-06, "loss": 0.8214, "step": 6862 }, { "epoch": 0.59, "learning_rate": 3.8664271718131494e-06, "loss": 0.8364, "step": 6863 }, { "epoch": 0.59, "learning_rate": 3.865081306137496e-06, "loss": 0.802, "step": 6864 }, { "epoch": 0.59, "learning_rate": 3.863735527152594e-06, "loss": 0.8423, "step": 6865 }, { "epoch": 0.59, "learning_rate": 3.8623898349612425e-06, "loss": 0.8117, "step": 6866 }, { "epoch": 0.59, "learning_rate": 3.8610442296662314e-06, "loss": 0.8345, "step": 6867 }, { "epoch": 0.59, "learning_rate": 3.859698711370345e-06, "loss": 0.7934, "step": 6868 }, { "epoch": 0.59, "learning_rate": 3.858353280176361e-06, "loss": 0.8391, "step": 6869 }, { "epoch": 0.59, "learning_rate": 3.857007936187048e-06, "loss": 0.8146, "step": 6870 }, { "epoch": 0.59, "learning_rate": 3.855662679505172e-06, "loss": 0.7867, "step": 6871 }, { "epoch": 0.59, "learning_rate": 3.854317510233491e-06, "loss": 0.8313, "step": 6872 }, { "epoch": 0.59, "learning_rate": 3.852972428474754e-06, "loss": 0.793, "step": 6873 }, { "epoch": 0.59, "learning_rate": 3.8516274343317065e-06, "loss": 0.5432, "step": 6874 }, { "epoch": 0.59, "learning_rate": 3.850282527907086e-06, "loss": 0.8392, "step": 6875 }, { "epoch": 0.59, "learning_rate": 3.848937709303621e-06, "loss": 0.8155, "step": 6876 }, { "epoch": 0.59, "learning_rate": 3.847592978624037e-06, "loss": 0.8308, "step": 6877 }, { "epoch": 0.59, "learning_rate": 3.8462483359710515e-06, "loss": 0.8534, "step": 6878 }, { "epoch": 0.59, "learning_rate": 3.8449037814473735e-06, "loss": 0.758, "step": 6879 }, { "epoch": 0.59, "learning_rate": 3.843559315155709e-06, "loss": 0.8361, "step": 6880 }, { "epoch": 0.59, "learning_rate": 3.8422149371987534e-06, "loss": 0.8215, "step": 6881 }, { "epoch": 0.59, "learning_rate": 3.8408706476792e-06, "loss": 0.7247, "step": 6882 }, { "epoch": 0.59, "learning_rate": 3.839526446699727e-06, "loss": 0.8408, "step": 6883 }, { "epoch": 0.59, "learning_rate": 3.838182334363015e-06, "loss": 0.8088, "step": 6884 }, { "epoch": 0.59, "learning_rate": 3.836838310771732e-06, "loss": 0.8411, "step": 6885 }, { "epoch": 0.59, "learning_rate": 3.835494376028544e-06, "loss": 0.8057, "step": 6886 }, { "epoch": 0.59, "learning_rate": 3.834150530236104e-06, "loss": 0.814, "step": 6887 }, { "epoch": 0.59, "learning_rate": 3.832806773497067e-06, "loss": 0.8051, "step": 6888 }, { "epoch": 0.59, "learning_rate": 3.831463105914068e-06, "loss": 0.8151, "step": 6889 }, { "epoch": 0.59, "learning_rate": 3.8301195275897486e-06, "loss": 0.8492, "step": 6890 }, { "epoch": 0.59, "learning_rate": 3.828776038626737e-06, "loss": 0.8094, "step": 6891 }, { "epoch": 0.59, "learning_rate": 3.827432639127655e-06, "loss": 0.8405, "step": 6892 }, { "epoch": 0.59, "learning_rate": 3.826089329195118e-06, "loss": 0.8273, "step": 6893 }, { "epoch": 0.59, "learning_rate": 3.824746108931735e-06, "loss": 0.8238, "step": 6894 }, { "epoch": 0.59, "learning_rate": 3.823402978440111e-06, "loss": 0.7756, "step": 6895 }, { "epoch": 0.59, "learning_rate": 3.822059937822836e-06, "loss": 0.8255, "step": 6896 }, { "epoch": 0.59, "learning_rate": 3.820716987182499e-06, "loss": 0.8016, "step": 6897 }, { "epoch": 0.59, "learning_rate": 3.819374126621683e-06, "loss": 0.8245, "step": 6898 }, { "epoch": 0.59, "learning_rate": 3.818031356242961e-06, "loss": 0.7599, "step": 6899 }, { "epoch": 0.59, "learning_rate": 3.816688676148903e-06, "loss": 0.7884, "step": 6900 }, { "epoch": 0.59, "learning_rate": 3.815346086442068e-06, "loss": 0.7826, "step": 6901 }, { "epoch": 0.59, "learning_rate": 3.814003587225009e-06, "loss": 0.8312, "step": 6902 }, { "epoch": 0.59, "learning_rate": 3.812661178600273e-06, "loss": 0.7711, "step": 6903 }, { "epoch": 0.59, "learning_rate": 3.8113188606704004e-06, "loss": 0.7979, "step": 6904 }, { "epoch": 0.59, "learning_rate": 3.8099766335379246e-06, "loss": 0.8211, "step": 6905 }, { "epoch": 0.59, "learning_rate": 3.80863449730537e-06, "loss": 0.8084, "step": 6906 }, { "epoch": 0.59, "learning_rate": 3.807292452075257e-06, "loss": 0.8184, "step": 6907 }, { "epoch": 0.59, "learning_rate": 3.8059504979500995e-06, "loss": 0.8098, "step": 6908 }, { "epoch": 0.59, "learning_rate": 3.804608635032398e-06, "loss": 0.7609, "step": 6909 }, { "epoch": 0.59, "learning_rate": 3.8032668634246543e-06, "loss": 0.8459, "step": 6910 }, { "epoch": 0.59, "learning_rate": 3.801925183229358e-06, "loss": 0.8508, "step": 6911 }, { "epoch": 0.59, "learning_rate": 3.8005835945489935e-06, "loss": 0.8336, "step": 6912 }, { "epoch": 0.59, "learning_rate": 3.799242097486038e-06, "loss": 0.8115, "step": 6913 }, { "epoch": 0.59, "learning_rate": 3.797900692142964e-06, "loss": 0.9017, "step": 6914 }, { "epoch": 0.59, "learning_rate": 3.796559378622234e-06, "loss": 0.8371, "step": 6915 }, { "epoch": 0.59, "learning_rate": 3.7952181570263e-06, "loss": 0.8337, "step": 6916 }, { "epoch": 0.59, "learning_rate": 3.7938770274576155e-06, "loss": 0.8166, "step": 6917 }, { "epoch": 0.59, "learning_rate": 3.7925359900186203e-06, "loss": 0.7783, "step": 6918 }, { "epoch": 0.59, "learning_rate": 3.7911950448117524e-06, "loss": 0.8678, "step": 6919 }, { "epoch": 0.59, "learning_rate": 3.7898541919394373e-06, "loss": 0.8281, "step": 6920 }, { "epoch": 0.59, "learning_rate": 3.7885134315040985e-06, "loss": 0.8189, "step": 6921 }, { "epoch": 0.59, "learning_rate": 3.7871727636081467e-06, "loss": 0.7435, "step": 6922 }, { "epoch": 0.59, "learning_rate": 3.78583218835399e-06, "loss": 0.7883, "step": 6923 }, { "epoch": 0.59, "learning_rate": 3.78449170584403e-06, "loss": 0.848, "step": 6924 }, { "epoch": 0.59, "learning_rate": 3.783151316180657e-06, "loss": 0.7955, "step": 6925 }, { "epoch": 0.59, "learning_rate": 3.7818110194662584e-06, "loss": 0.7962, "step": 6926 }, { "epoch": 0.59, "learning_rate": 3.7804708158032123e-06, "loss": 0.861, "step": 6927 }, { "epoch": 0.59, "learning_rate": 3.7791307052938906e-06, "loss": 0.7989, "step": 6928 }, { "epoch": 0.59, "learning_rate": 3.777790688040657e-06, "loss": 0.8134, "step": 6929 }, { "epoch": 0.59, "learning_rate": 3.7764507641458674e-06, "loss": 0.8022, "step": 6930 }, { "epoch": 0.59, "learning_rate": 3.775110933711873e-06, "loss": 0.7611, "step": 6931 }, { "epoch": 0.59, "learning_rate": 3.773771196841017e-06, "loss": 0.7966, "step": 6932 }, { "epoch": 0.59, "learning_rate": 3.772431553635637e-06, "loss": 0.8202, "step": 6933 }, { "epoch": 0.59, "learning_rate": 3.771092004198058e-06, "loss": 0.8405, "step": 6934 }, { "epoch": 0.59, "learning_rate": 3.769752548630606e-06, "loss": 0.8235, "step": 6935 }, { "epoch": 0.59, "learning_rate": 3.7684131870355897e-06, "loss": 0.8124, "step": 6936 }, { "epoch": 0.59, "learning_rate": 3.767073919515319e-06, "loss": 0.5791, "step": 6937 }, { "epoch": 0.59, "learning_rate": 3.765734746172095e-06, "loss": 0.8023, "step": 6938 }, { "epoch": 0.59, "learning_rate": 3.7643956671082084e-06, "loss": 0.8401, "step": 6939 }, { "epoch": 0.59, "learning_rate": 3.7630566824259456e-06, "loss": 0.8057, "step": 6940 }, { "epoch": 0.59, "learning_rate": 3.7617177922275872e-06, "loss": 0.8378, "step": 6941 }, { "epoch": 0.59, "learning_rate": 3.7603789966154004e-06, "loss": 0.7764, "step": 6942 }, { "epoch": 0.59, "learning_rate": 3.759040295691651e-06, "loss": 0.8491, "step": 6943 }, { "epoch": 0.59, "learning_rate": 3.757701689558595e-06, "loss": 0.8577, "step": 6944 }, { "epoch": 0.59, "learning_rate": 3.7563631783184816e-06, "loss": 0.7856, "step": 6945 }, { "epoch": 0.59, "learning_rate": 3.7550247620735554e-06, "loss": 0.8285, "step": 6946 }, { "epoch": 0.59, "learning_rate": 3.753686440926048e-06, "loss": 0.8447, "step": 6947 }, { "epoch": 0.59, "learning_rate": 3.7523482149781907e-06, "loss": 0.7808, "step": 6948 }, { "epoch": 0.59, "learning_rate": 3.7510100843322e-06, "loss": 0.8181, "step": 6949 }, { "epoch": 0.59, "learning_rate": 3.7496720490902903e-06, "loss": 0.5383, "step": 6950 }, { "epoch": 0.59, "learning_rate": 3.748334109354669e-06, "loss": 0.5605, "step": 6951 }, { "epoch": 0.59, "learning_rate": 3.746996265227534e-06, "loss": 0.8299, "step": 6952 }, { "epoch": 0.59, "learning_rate": 3.7456585168110737e-06, "loss": 0.8205, "step": 6953 }, { "epoch": 0.59, "learning_rate": 3.744320864207476e-06, "loss": 0.8038, "step": 6954 }, { "epoch": 0.59, "learning_rate": 3.742983307518917e-06, "loss": 0.8466, "step": 6955 }, { "epoch": 0.59, "learning_rate": 3.7416458468475623e-06, "loss": 0.7649, "step": 6956 }, { "epoch": 0.59, "learning_rate": 3.7403084822955777e-06, "loss": 0.7855, "step": 6957 }, { "epoch": 0.59, "learning_rate": 3.738971213965115e-06, "loss": 0.7816, "step": 6958 }, { "epoch": 0.59, "learning_rate": 3.7376340419583235e-06, "loss": 0.8032, "step": 6959 }, { "epoch": 0.59, "learning_rate": 3.736296966377343e-06, "loss": 0.8119, "step": 6960 }, { "epoch": 0.59, "learning_rate": 3.734959987324305e-06, "loss": 0.8569, "step": 6961 }, { "epoch": 0.59, "learning_rate": 3.733623104901335e-06, "loss": 0.7669, "step": 6962 }, { "epoch": 0.59, "learning_rate": 3.73228631921055e-06, "loss": 0.7972, "step": 6963 }, { "epoch": 0.59, "learning_rate": 3.7309496303540606e-06, "loss": 0.8165, "step": 6964 }, { "epoch": 0.59, "learning_rate": 3.7296130384339713e-06, "loss": 0.8086, "step": 6965 }, { "epoch": 0.59, "learning_rate": 3.728276543552376e-06, "loss": 0.8227, "step": 6966 }, { "epoch": 0.59, "learning_rate": 3.726940145811363e-06, "loss": 0.7867, "step": 6967 }, { "epoch": 0.59, "learning_rate": 3.7256038453130158e-06, "loss": 0.8389, "step": 6968 }, { "epoch": 0.59, "learning_rate": 3.7242676421594025e-06, "loss": 0.8249, "step": 6969 }, { "epoch": 0.59, "learning_rate": 3.722931536452593e-06, "loss": 0.8123, "step": 6970 }, { "epoch": 0.59, "learning_rate": 3.7215955282946446e-06, "loss": 0.8487, "step": 6971 }, { "epoch": 0.59, "learning_rate": 3.7202596177876082e-06, "loss": 0.8471, "step": 6972 }, { "epoch": 0.6, "learning_rate": 3.7189238050335263e-06, "loss": 0.8181, "step": 6973 }, { "epoch": 0.6, "learning_rate": 3.7175880901344385e-06, "loss": 0.7804, "step": 6974 }, { "epoch": 0.6, "learning_rate": 3.716252473192369e-06, "loss": 0.7992, "step": 6975 }, { "epoch": 0.6, "learning_rate": 3.714916954309341e-06, "loss": 0.8305, "step": 6976 }, { "epoch": 0.6, "learning_rate": 3.7135815335873675e-06, "loss": 0.7732, "step": 6977 }, { "epoch": 0.6, "learning_rate": 3.7122462111284553e-06, "loss": 0.835, "step": 6978 }, { "epoch": 0.6, "learning_rate": 3.7109109870346026e-06, "loss": 0.827, "step": 6979 }, { "epoch": 0.6, "learning_rate": 3.7095758614078e-06, "loss": 0.8247, "step": 6980 }, { "epoch": 0.6, "learning_rate": 3.708240834350033e-06, "loss": 0.7809, "step": 6981 }, { "epoch": 0.6, "learning_rate": 3.706905905963275e-06, "loss": 0.8283, "step": 6982 }, { "epoch": 0.6, "learning_rate": 3.7055710763494945e-06, "loss": 0.8371, "step": 6983 }, { "epoch": 0.6, "learning_rate": 3.7042363456106558e-06, "loss": 0.8079, "step": 6984 }, { "epoch": 0.6, "learning_rate": 3.7029017138487083e-06, "loss": 0.8055, "step": 6985 }, { "epoch": 0.6, "learning_rate": 3.7015671811656e-06, "loss": 0.86, "step": 6986 }, { "epoch": 0.6, "learning_rate": 3.7002327476632684e-06, "loss": 0.8155, "step": 6987 }, { "epoch": 0.6, "learning_rate": 3.6988984134436474e-06, "loss": 0.7835, "step": 6988 }, { "epoch": 0.6, "learning_rate": 3.697564178608655e-06, "loss": 0.8087, "step": 6989 }, { "epoch": 0.6, "learning_rate": 3.6962300432602095e-06, "loss": 0.8249, "step": 6990 }, { "epoch": 0.6, "learning_rate": 3.694896007500217e-06, "loss": 0.83, "step": 6991 }, { "epoch": 0.6, "learning_rate": 3.6935620714305797e-06, "loss": 0.8423, "step": 6992 }, { "epoch": 0.6, "learning_rate": 3.6922282351531914e-06, "loss": 0.8361, "step": 6993 }, { "epoch": 0.6, "learning_rate": 3.6908944987699346e-06, "loss": 0.818, "step": 6994 }, { "epoch": 0.6, "learning_rate": 3.6895608623826873e-06, "loss": 0.829, "step": 6995 }, { "epoch": 0.6, "learning_rate": 3.6882273260933192e-06, "loss": 0.7793, "step": 6996 }, { "epoch": 0.6, "learning_rate": 3.6868938900036934e-06, "loss": 0.8432, "step": 6997 }, { "epoch": 0.6, "learning_rate": 3.6855605542156637e-06, "loss": 0.8096, "step": 6998 }, { "epoch": 0.6, "learning_rate": 3.6842273188310766e-06, "loss": 0.805, "step": 6999 }, { "epoch": 0.6, "learning_rate": 3.6828941839517716e-06, "loss": 0.8378, "step": 7000 }, { "epoch": 0.6, "learning_rate": 3.6815611496795835e-06, "loss": 0.863, "step": 7001 }, { "epoch": 0.6, "learning_rate": 3.68022821611633e-06, "loss": 0.8062, "step": 7002 }, { "epoch": 0.6, "learning_rate": 3.6788953833638315e-06, "loss": 0.8229, "step": 7003 }, { "epoch": 0.6, "learning_rate": 3.677562651523894e-06, "loss": 0.8584, "step": 7004 }, { "epoch": 0.6, "learning_rate": 3.676230020698319e-06, "loss": 0.7973, "step": 7005 }, { "epoch": 0.6, "learning_rate": 3.6748974909889e-06, "loss": 0.7706, "step": 7006 }, { "epoch": 0.6, "learning_rate": 3.6735650624974235e-06, "loss": 0.847, "step": 7007 }, { "epoch": 0.6, "learning_rate": 3.6722327353256664e-06, "loss": 0.8475, "step": 7008 }, { "epoch": 0.6, "learning_rate": 3.6709005095753958e-06, "loss": 0.7616, "step": 7009 }, { "epoch": 0.6, "learning_rate": 3.6695683853483753e-06, "loss": 0.8256, "step": 7010 }, { "epoch": 0.6, "learning_rate": 3.6682363627463604e-06, "loss": 0.7594, "step": 7011 }, { "epoch": 0.6, "learning_rate": 3.666904441871096e-06, "loss": 0.8422, "step": 7012 }, { "epoch": 0.6, "learning_rate": 3.665572622824323e-06, "loss": 0.8518, "step": 7013 }, { "epoch": 0.6, "learning_rate": 3.6642409057077715e-06, "loss": 0.795, "step": 7014 }, { "epoch": 0.6, "learning_rate": 3.6629092906231627e-06, "loss": 0.7626, "step": 7015 }, { "epoch": 0.6, "learning_rate": 3.661577777672214e-06, "loss": 0.8511, "step": 7016 }, { "epoch": 0.6, "learning_rate": 3.660246366956632e-06, "loss": 0.5351, "step": 7017 }, { "epoch": 0.6, "learning_rate": 3.6589150585781174e-06, "loss": 0.8367, "step": 7018 }, { "epoch": 0.6, "learning_rate": 3.6575838526383624e-06, "loss": 0.8041, "step": 7019 }, { "epoch": 0.6, "learning_rate": 3.6562527492390493e-06, "loss": 0.5611, "step": 7020 }, { "epoch": 0.6, "learning_rate": 3.6549217484818576e-06, "loss": 0.7725, "step": 7021 }, { "epoch": 0.6, "learning_rate": 3.653590850468454e-06, "loss": 0.8226, "step": 7022 }, { "epoch": 0.6, "learning_rate": 3.652260055300497e-06, "loss": 0.8466, "step": 7023 }, { "epoch": 0.6, "learning_rate": 3.650929363079642e-06, "loss": 0.8447, "step": 7024 }, { "epoch": 0.6, "learning_rate": 3.6495987739075334e-06, "loss": 0.8223, "step": 7025 }, { "epoch": 0.6, "learning_rate": 3.648268287885808e-06, "loss": 0.8119, "step": 7026 }, { "epoch": 0.6, "learning_rate": 3.6469379051160977e-06, "loss": 0.8339, "step": 7027 }, { "epoch": 0.6, "learning_rate": 3.6456076257000184e-06, "loss": 0.804, "step": 7028 }, { "epoch": 0.6, "learning_rate": 3.644277449739185e-06, "loss": 0.8548, "step": 7029 }, { "epoch": 0.6, "learning_rate": 3.6429473773352066e-06, "loss": 0.809, "step": 7030 }, { "epoch": 0.6, "learning_rate": 3.6416174085896777e-06, "loss": 0.8181, "step": 7031 }, { "epoch": 0.6, "learning_rate": 3.640287543604188e-06, "loss": 0.8155, "step": 7032 }, { "epoch": 0.6, "learning_rate": 3.63895778248032e-06, "loss": 0.8168, "step": 7033 }, { "epoch": 0.6, "learning_rate": 3.6376281253196495e-06, "loss": 0.8228, "step": 7034 }, { "epoch": 0.6, "learning_rate": 3.6362985722237375e-06, "loss": 0.4997, "step": 7035 }, { "epoch": 0.6, "learning_rate": 3.6349691232941464e-06, "loss": 0.846, "step": 7036 }, { "epoch": 0.6, "learning_rate": 3.633639778632423e-06, "loss": 0.8265, "step": 7037 }, { "epoch": 0.6, "learning_rate": 3.6323105383401102e-06, "loss": 0.8456, "step": 7038 }, { "epoch": 0.6, "learning_rate": 3.630981402518743e-06, "loss": 0.7712, "step": 7039 }, { "epoch": 0.6, "learning_rate": 3.6296523712698476e-06, "loss": 0.8429, "step": 7040 }, { "epoch": 0.6, "learning_rate": 3.6283234446949422e-06, "loss": 0.7887, "step": 7041 }, { "epoch": 0.6, "learning_rate": 3.6269946228955344e-06, "loss": 0.8409, "step": 7042 }, { "epoch": 0.6, "learning_rate": 3.6256659059731275e-06, "loss": 0.8059, "step": 7043 }, { "epoch": 0.6, "learning_rate": 3.6243372940292156e-06, "loss": 0.813, "step": 7044 }, { "epoch": 0.6, "learning_rate": 3.6230087871652863e-06, "loss": 0.8722, "step": 7045 }, { "epoch": 0.6, "learning_rate": 3.6216803854828146e-06, "loss": 0.7714, "step": 7046 }, { "epoch": 0.6, "learning_rate": 3.6203520890832746e-06, "loss": 0.8318, "step": 7047 }, { "epoch": 0.6, "learning_rate": 3.6190238980681235e-06, "loss": 0.7852, "step": 7048 }, { "epoch": 0.6, "learning_rate": 3.6176958125388173e-06, "loss": 0.7656, "step": 7049 }, { "epoch": 0.6, "learning_rate": 3.616367832596802e-06, "loss": 0.8578, "step": 7050 }, { "epoch": 0.6, "learning_rate": 3.6150399583435146e-06, "loss": 0.8013, "step": 7051 }, { "epoch": 0.6, "learning_rate": 3.613712189880386e-06, "loss": 0.8548, "step": 7052 }, { "epoch": 0.6, "learning_rate": 3.6123845273088365e-06, "loss": 0.7867, "step": 7053 }, { "epoch": 0.6, "learning_rate": 3.611056970730282e-06, "loss": 0.8494, "step": 7054 }, { "epoch": 0.6, "learning_rate": 3.6097295202461256e-06, "loss": 0.7897, "step": 7055 }, { "epoch": 0.6, "learning_rate": 3.608402175957764e-06, "loss": 0.8188, "step": 7056 }, { "epoch": 0.6, "learning_rate": 3.6070749379665863e-06, "loss": 0.8194, "step": 7057 }, { "epoch": 0.6, "learning_rate": 3.605747806373976e-06, "loss": 0.7953, "step": 7058 }, { "epoch": 0.6, "learning_rate": 3.6044207812813048e-06, "loss": 0.8431, "step": 7059 }, { "epoch": 0.6, "learning_rate": 3.603093862789937e-06, "loss": 0.8619, "step": 7060 }, { "epoch": 0.6, "learning_rate": 3.6017670510012316e-06, "loss": 0.8359, "step": 7061 }, { "epoch": 0.6, "learning_rate": 3.6004403460165327e-06, "loss": 0.8429, "step": 7062 }, { "epoch": 0.6, "learning_rate": 3.599113747937184e-06, "loss": 0.8612, "step": 7063 }, { "epoch": 0.6, "learning_rate": 3.5977872568645166e-06, "loss": 0.8119, "step": 7064 }, { "epoch": 0.6, "learning_rate": 3.596460872899855e-06, "loss": 0.837, "step": 7065 }, { "epoch": 0.6, "learning_rate": 3.595134596144514e-06, "loss": 0.8461, "step": 7066 }, { "epoch": 0.6, "learning_rate": 3.5938084266998043e-06, "loss": 0.8052, "step": 7067 }, { "epoch": 0.6, "learning_rate": 3.5924823646670205e-06, "loss": 0.8329, "step": 7068 }, { "epoch": 0.6, "learning_rate": 3.591156410147458e-06, "loss": 0.7385, "step": 7069 }, { "epoch": 0.6, "learning_rate": 3.589830563242397e-06, "loss": 0.8067, "step": 7070 }, { "epoch": 0.6, "learning_rate": 3.588504824053113e-06, "loss": 0.7812, "step": 7071 }, { "epoch": 0.6, "learning_rate": 3.587179192680873e-06, "loss": 0.8005, "step": 7072 }, { "epoch": 0.6, "learning_rate": 3.5858536692269364e-06, "loss": 0.8415, "step": 7073 }, { "epoch": 0.6, "learning_rate": 3.584528253792554e-06, "loss": 0.7781, "step": 7074 }, { "epoch": 0.6, "learning_rate": 3.583202946478963e-06, "loss": 0.8143, "step": 7075 }, { "epoch": 0.6, "learning_rate": 3.581877747387401e-06, "loss": 0.7942, "step": 7076 }, { "epoch": 0.6, "learning_rate": 3.580552656619091e-06, "loss": 0.799, "step": 7077 }, { "epoch": 0.6, "learning_rate": 3.579227674275253e-06, "loss": 0.8098, "step": 7078 }, { "epoch": 0.6, "learning_rate": 3.5779028004570925e-06, "loss": 0.4667, "step": 7079 }, { "epoch": 0.6, "learning_rate": 3.576578035265812e-06, "loss": 0.8076, "step": 7080 }, { "epoch": 0.6, "learning_rate": 3.5752533788026056e-06, "loss": 0.7771, "step": 7081 }, { "epoch": 0.6, "learning_rate": 3.573928831168652e-06, "loss": 0.8051, "step": 7082 }, { "epoch": 0.6, "learning_rate": 3.572604392465131e-06, "loss": 0.8347, "step": 7083 }, { "epoch": 0.6, "learning_rate": 3.5712800627932064e-06, "loss": 0.8182, "step": 7084 }, { "epoch": 0.6, "learning_rate": 3.56995584225404e-06, "loss": 0.8464, "step": 7085 }, { "epoch": 0.6, "learning_rate": 3.5686317309487816e-06, "loss": 0.8569, "step": 7086 }, { "epoch": 0.6, "learning_rate": 3.567307728978574e-06, "loss": 0.7871, "step": 7087 }, { "epoch": 0.6, "learning_rate": 3.5659838364445505e-06, "loss": 0.8069, "step": 7088 }, { "epoch": 0.6, "learning_rate": 3.564660053447835e-06, "loss": 0.7897, "step": 7089 }, { "epoch": 0.61, "learning_rate": 3.563336380089546e-06, "loss": 0.7572, "step": 7090 }, { "epoch": 0.61, "learning_rate": 3.562012816470792e-06, "loss": 0.8114, "step": 7091 }, { "epoch": 0.61, "learning_rate": 3.5606893626926754e-06, "loss": 0.753, "step": 7092 }, { "epoch": 0.61, "learning_rate": 3.559366018856285e-06, "loss": 0.8323, "step": 7093 }, { "epoch": 0.61, "learning_rate": 3.5580427850627085e-06, "loss": 0.8411, "step": 7094 }, { "epoch": 0.61, "learning_rate": 3.556719661413016e-06, "loss": 0.7563, "step": 7095 }, { "epoch": 0.61, "learning_rate": 3.555396648008278e-06, "loss": 0.7854, "step": 7096 }, { "epoch": 0.61, "learning_rate": 3.5540737449495515e-06, "loss": 0.7628, "step": 7097 }, { "epoch": 0.61, "learning_rate": 3.552750952337886e-06, "loss": 0.7981, "step": 7098 }, { "epoch": 0.61, "learning_rate": 3.551428270274324e-06, "loss": 0.8795, "step": 7099 }, { "epoch": 0.61, "learning_rate": 3.5501056988598996e-06, "loss": 0.8264, "step": 7100 }, { "epoch": 0.61, "learning_rate": 3.5487832381956343e-06, "loss": 0.7805, "step": 7101 }, { "epoch": 0.61, "learning_rate": 3.5474608883825475e-06, "loss": 0.8632, "step": 7102 }, { "epoch": 0.61, "learning_rate": 3.5461386495216433e-06, "loss": 0.5689, "step": 7103 }, { "epoch": 0.61, "learning_rate": 3.5448165217139243e-06, "loss": 0.8624, "step": 7104 }, { "epoch": 0.61, "learning_rate": 3.54349450506038e-06, "loss": 0.5104, "step": 7105 }, { "epoch": 0.61, "learning_rate": 3.542172599661991e-06, "loss": 0.8126, "step": 7106 }, { "epoch": 0.61, "learning_rate": 3.540850805619735e-06, "loss": 0.8526, "step": 7107 }, { "epoch": 0.61, "learning_rate": 3.539529123034573e-06, "loss": 0.7951, "step": 7108 }, { "epoch": 0.61, "learning_rate": 3.5382075520074625e-06, "loss": 0.8092, "step": 7109 }, { "epoch": 0.61, "learning_rate": 3.5368860926393535e-06, "loss": 0.8306, "step": 7110 }, { "epoch": 0.61, "learning_rate": 3.535564745031186e-06, "loss": 0.8162, "step": 7111 }, { "epoch": 0.61, "learning_rate": 3.534243509283888e-06, "loss": 0.7799, "step": 7112 }, { "epoch": 0.61, "learning_rate": 3.5329223854983847e-06, "loss": 0.8316, "step": 7113 }, { "epoch": 0.61, "learning_rate": 3.5316013737755912e-06, "loss": 0.817, "step": 7114 }, { "epoch": 0.61, "learning_rate": 3.530280474216409e-06, "loss": 0.8039, "step": 7115 }, { "epoch": 0.61, "learning_rate": 3.5289596869217384e-06, "loss": 0.8007, "step": 7116 }, { "epoch": 0.61, "learning_rate": 3.527639011992465e-06, "loss": 0.8294, "step": 7117 }, { "epoch": 0.61, "learning_rate": 3.5263184495294713e-06, "loss": 0.821, "step": 7118 }, { "epoch": 0.61, "learning_rate": 3.524997999633627e-06, "loss": 0.797, "step": 7119 }, { "epoch": 0.61, "learning_rate": 3.5236776624057956e-06, "loss": 0.8407, "step": 7120 }, { "epoch": 0.61, "learning_rate": 3.5223574379468295e-06, "loss": 0.8635, "step": 7121 }, { "epoch": 0.61, "learning_rate": 3.521037326357575e-06, "loss": 0.7933, "step": 7122 }, { "epoch": 0.61, "learning_rate": 3.5197173277388675e-06, "loss": 0.8465, "step": 7123 }, { "epoch": 0.61, "learning_rate": 3.5183974421915378e-06, "loss": 0.842, "step": 7124 }, { "epoch": 0.61, "learning_rate": 3.517077669816402e-06, "loss": 0.8191, "step": 7125 }, { "epoch": 0.61, "learning_rate": 3.515758010714273e-06, "loss": 0.8212, "step": 7126 }, { "epoch": 0.61, "learning_rate": 3.514438464985955e-06, "loss": 0.8248, "step": 7127 }, { "epoch": 0.61, "learning_rate": 3.5131190327322374e-06, "loss": 0.8084, "step": 7128 }, { "epoch": 0.61, "learning_rate": 3.5117997140539073e-06, "loss": 0.8296, "step": 7129 }, { "epoch": 0.61, "learning_rate": 3.510480509051739e-06, "loss": 0.8109, "step": 7130 }, { "epoch": 0.61, "learning_rate": 3.509161417826501e-06, "loss": 0.7887, "step": 7131 }, { "epoch": 0.61, "learning_rate": 3.507842440478953e-06, "loss": 0.8392, "step": 7132 }, { "epoch": 0.61, "learning_rate": 3.506523577109845e-06, "loss": 0.7962, "step": 7133 }, { "epoch": 0.61, "learning_rate": 3.505204827819918e-06, "loss": 0.8702, "step": 7134 }, { "epoch": 0.61, "learning_rate": 3.5038861927099037e-06, "loss": 0.8759, "step": 7135 }, { "epoch": 0.61, "learning_rate": 3.502567671880526e-06, "loss": 0.858, "step": 7136 }, { "epoch": 0.61, "learning_rate": 3.5012492654325013e-06, "loss": 0.7966, "step": 7137 }, { "epoch": 0.61, "learning_rate": 3.499930973466536e-06, "loss": 0.8293, "step": 7138 }, { "epoch": 0.61, "learning_rate": 3.4986127960833265e-06, "loss": 0.847, "step": 7139 }, { "epoch": 0.61, "learning_rate": 3.497294733383565e-06, "loss": 0.7591, "step": 7140 }, { "epoch": 0.61, "learning_rate": 3.4959767854679264e-06, "loss": 0.8061, "step": 7141 }, { "epoch": 0.61, "learning_rate": 3.494658952437086e-06, "loss": 0.8036, "step": 7142 }, { "epoch": 0.61, "learning_rate": 3.493341234391706e-06, "loss": 0.7825, "step": 7143 }, { "epoch": 0.61, "learning_rate": 3.4920236314324396e-06, "loss": 0.8644, "step": 7144 }, { "epoch": 0.61, "learning_rate": 3.490706143659931e-06, "loss": 0.8191, "step": 7145 }, { "epoch": 0.61, "learning_rate": 3.4893887711748187e-06, "loss": 0.8426, "step": 7146 }, { "epoch": 0.61, "learning_rate": 3.48807151407773e-06, "loss": 0.8732, "step": 7147 }, { "epoch": 0.61, "learning_rate": 3.486754372469283e-06, "loss": 0.5253, "step": 7148 }, { "epoch": 0.61, "learning_rate": 3.485437346450085e-06, "loss": 0.8362, "step": 7149 }, { "epoch": 0.61, "learning_rate": 3.4841204361207393e-06, "loss": 0.8332, "step": 7150 }, { "epoch": 0.61, "learning_rate": 3.482803641581839e-06, "loss": 0.8126, "step": 7151 }, { "epoch": 0.61, "learning_rate": 3.4814869629339664e-06, "loss": 0.8279, "step": 7152 }, { "epoch": 0.61, "learning_rate": 3.4801704002776963e-06, "loss": 0.7735, "step": 7153 }, { "epoch": 0.61, "learning_rate": 3.4788539537135948e-06, "loss": 0.8034, "step": 7154 }, { "epoch": 0.61, "learning_rate": 3.4775376233422152e-06, "loss": 0.8554, "step": 7155 }, { "epoch": 0.61, "learning_rate": 3.47622140926411e-06, "loss": 0.8368, "step": 7156 }, { "epoch": 0.61, "learning_rate": 3.474905311579816e-06, "loss": 0.7877, "step": 7157 }, { "epoch": 0.61, "learning_rate": 3.4735893303898627e-06, "loss": 0.7877, "step": 7158 }, { "epoch": 0.61, "learning_rate": 3.472273465794773e-06, "loss": 0.8577, "step": 7159 }, { "epoch": 0.61, "learning_rate": 3.47095771789506e-06, "loss": 0.8499, "step": 7160 }, { "epoch": 0.61, "learning_rate": 3.4696420867912234e-06, "loss": 0.5111, "step": 7161 }, { "epoch": 0.61, "learning_rate": 3.4683265725837604e-06, "loss": 0.8272, "step": 7162 }, { "epoch": 0.61, "learning_rate": 3.4670111753731557e-06, "loss": 0.5853, "step": 7163 }, { "epoch": 0.61, "learning_rate": 3.465695895259885e-06, "loss": 0.8183, "step": 7164 }, { "epoch": 0.61, "learning_rate": 3.464380732344418e-06, "loss": 0.6054, "step": 7165 }, { "epoch": 0.61, "learning_rate": 3.463065686727214e-06, "loss": 0.867, "step": 7166 }, { "epoch": 0.61, "learning_rate": 3.4617507585087207e-06, "loss": 0.7891, "step": 7167 }, { "epoch": 0.61, "learning_rate": 3.4604359477893784e-06, "loss": 0.8281, "step": 7168 }, { "epoch": 0.61, "learning_rate": 3.4591212546696196e-06, "loss": 0.7784, "step": 7169 }, { "epoch": 0.61, "learning_rate": 3.457806679249868e-06, "loss": 0.8133, "step": 7170 }, { "epoch": 0.61, "learning_rate": 3.4564922216305363e-06, "loss": 0.8185, "step": 7171 }, { "epoch": 0.61, "learning_rate": 3.45517788191203e-06, "loss": 0.8071, "step": 7172 }, { "epoch": 0.61, "learning_rate": 3.4538636601947472e-06, "loss": 0.8644, "step": 7173 }, { "epoch": 0.61, "learning_rate": 3.452549556579069e-06, "loss": 0.8142, "step": 7174 }, { "epoch": 0.61, "learning_rate": 3.4512355711653777e-06, "loss": 0.8224, "step": 7175 }, { "epoch": 0.61, "learning_rate": 3.4499217040540413e-06, "loss": 0.7712, "step": 7176 }, { "epoch": 0.61, "learning_rate": 3.4486079553454184e-06, "loss": 0.8412, "step": 7177 }, { "epoch": 0.61, "learning_rate": 3.44729432513986e-06, "loss": 0.818, "step": 7178 }, { "epoch": 0.61, "learning_rate": 3.4459808135377082e-06, "loss": 0.8022, "step": 7179 }, { "epoch": 0.61, "learning_rate": 3.444667420639298e-06, "loss": 0.7938, "step": 7180 }, { "epoch": 0.61, "learning_rate": 3.4433541465449482e-06, "loss": 0.7867, "step": 7181 }, { "epoch": 0.61, "learning_rate": 3.442040991354976e-06, "loss": 0.8012, "step": 7182 }, { "epoch": 0.61, "learning_rate": 3.4407279551696846e-06, "loss": 0.8112, "step": 7183 }, { "epoch": 0.61, "learning_rate": 3.439415038089372e-06, "loss": 0.8348, "step": 7184 }, { "epoch": 0.61, "learning_rate": 3.438102240214327e-06, "loss": 0.8076, "step": 7185 }, { "epoch": 0.61, "learning_rate": 3.436789561644824e-06, "loss": 0.8026, "step": 7186 }, { "epoch": 0.61, "learning_rate": 3.435477002481137e-06, "loss": 0.8246, "step": 7187 }, { "epoch": 0.61, "learning_rate": 3.4341645628235194e-06, "loss": 0.8265, "step": 7188 }, { "epoch": 0.61, "learning_rate": 3.4328522427722255e-06, "loss": 0.8018, "step": 7189 }, { "epoch": 0.61, "learning_rate": 3.431540042427497e-06, "loss": 0.8188, "step": 7190 }, { "epoch": 0.61, "learning_rate": 3.430227961889565e-06, "loss": 0.7737, "step": 7191 }, { "epoch": 0.61, "learning_rate": 3.4289160012586536e-06, "loss": 0.7894, "step": 7192 }, { "epoch": 0.61, "learning_rate": 3.4276041606349785e-06, "loss": 0.8444, "step": 7193 }, { "epoch": 0.61, "learning_rate": 3.4262924401187415e-06, "loss": 0.8316, "step": 7194 }, { "epoch": 0.61, "learning_rate": 3.4249808398101403e-06, "loss": 0.8166, "step": 7195 }, { "epoch": 0.61, "learning_rate": 3.4236693598093597e-06, "loss": 0.8104, "step": 7196 }, { "epoch": 0.61, "learning_rate": 3.4223580002165784e-06, "loss": 0.7905, "step": 7197 }, { "epoch": 0.61, "learning_rate": 3.421046761131965e-06, "loss": 0.821, "step": 7198 }, { "epoch": 0.61, "learning_rate": 3.4197356426556784e-06, "loss": 0.8321, "step": 7199 }, { "epoch": 0.61, "learning_rate": 3.4184246448878695e-06, "loss": 0.8146, "step": 7200 }, { "epoch": 0.61, "learning_rate": 3.4171137679286746e-06, "loss": 0.8297, "step": 7201 }, { "epoch": 0.61, "learning_rate": 3.4158030118782287e-06, "loss": 0.866, "step": 7202 }, { "epoch": 0.61, "learning_rate": 3.414492376836652e-06, "loss": 0.8199, "step": 7203 }, { "epoch": 0.61, "learning_rate": 3.41318186290406e-06, "loss": 0.4905, "step": 7204 }, { "epoch": 0.61, "learning_rate": 3.4118714701805535e-06, "loss": 0.8314, "step": 7205 }, { "epoch": 0.61, "learning_rate": 3.4105611987662278e-06, "loss": 0.8102, "step": 7206 }, { "epoch": 0.61, "learning_rate": 3.4092510487611696e-06, "loss": 0.5637, "step": 7207 }, { "epoch": 0.62, "learning_rate": 3.407941020265452e-06, "loss": 0.8174, "step": 7208 }, { "epoch": 0.62, "learning_rate": 3.406631113379143e-06, "loss": 0.8275, "step": 7209 }, { "epoch": 0.62, "learning_rate": 3.4053213282022983e-06, "loss": 0.8263, "step": 7210 }, { "epoch": 0.62, "learning_rate": 3.4040116648349673e-06, "loss": 0.815, "step": 7211 }, { "epoch": 0.62, "learning_rate": 3.4027021233771882e-06, "loss": 0.7898, "step": 7212 }, { "epoch": 0.62, "learning_rate": 3.4013927039289936e-06, "loss": 0.7469, "step": 7213 }, { "epoch": 0.62, "learning_rate": 3.400083406590398e-06, "loss": 0.807, "step": 7214 }, { "epoch": 0.62, "learning_rate": 3.398774231461414e-06, "loss": 0.8145, "step": 7215 }, { "epoch": 0.62, "learning_rate": 3.397465178642043e-06, "loss": 0.8407, "step": 7216 }, { "epoch": 0.62, "learning_rate": 3.3961562482322775e-06, "loss": 0.7999, "step": 7217 }, { "epoch": 0.62, "learning_rate": 3.3948474403321015e-06, "loss": 0.5988, "step": 7218 }, { "epoch": 0.62, "learning_rate": 3.393538755041485e-06, "loss": 0.8572, "step": 7219 }, { "epoch": 0.62, "learning_rate": 3.392230192460396e-06, "loss": 0.8107, "step": 7220 }, { "epoch": 0.62, "learning_rate": 3.3909217526887845e-06, "loss": 0.5635, "step": 7221 }, { "epoch": 0.62, "learning_rate": 3.3896134358265977e-06, "loss": 0.8256, "step": 7222 }, { "epoch": 0.62, "learning_rate": 3.388305241973773e-06, "loss": 0.7972, "step": 7223 }, { "epoch": 0.62, "learning_rate": 3.386997171230234e-06, "loss": 0.7925, "step": 7224 }, { "epoch": 0.62, "learning_rate": 3.3856892236958994e-06, "loss": 0.8211, "step": 7225 }, { "epoch": 0.62, "learning_rate": 3.3843813994706777e-06, "loss": 0.8094, "step": 7226 }, { "epoch": 0.62, "learning_rate": 3.3830736986544638e-06, "loss": 0.8116, "step": 7227 }, { "epoch": 0.62, "learning_rate": 3.381766121347149e-06, "loss": 0.8345, "step": 7228 }, { "epoch": 0.62, "learning_rate": 3.380458667648612e-06, "loss": 0.8387, "step": 7229 }, { "epoch": 0.62, "learning_rate": 3.3791513376587216e-06, "loss": 0.7765, "step": 7230 }, { "epoch": 0.62, "learning_rate": 3.3778441314773384e-06, "loss": 0.8226, "step": 7231 }, { "epoch": 0.62, "learning_rate": 3.376537049204316e-06, "loss": 0.8637, "step": 7232 }, { "epoch": 0.62, "learning_rate": 3.3752300909394946e-06, "loss": 0.7621, "step": 7233 }, { "epoch": 0.62, "learning_rate": 3.3739232567827034e-06, "loss": 0.8252, "step": 7234 }, { "epoch": 0.62, "learning_rate": 3.372616546833767e-06, "loss": 0.8155, "step": 7235 }, { "epoch": 0.62, "learning_rate": 3.3713099611924987e-06, "loss": 0.7888, "step": 7236 }, { "epoch": 0.62, "learning_rate": 3.370003499958703e-06, "loss": 0.809, "step": 7237 }, { "epoch": 0.62, "learning_rate": 3.368697163232171e-06, "loss": 0.8753, "step": 7238 }, { "epoch": 0.62, "learning_rate": 3.367390951112689e-06, "loss": 0.8069, "step": 7239 }, { "epoch": 0.62, "learning_rate": 3.366084863700034e-06, "loss": 0.8174, "step": 7240 }, { "epoch": 0.62, "learning_rate": 3.364778901093967e-06, "loss": 0.7674, "step": 7241 }, { "epoch": 0.62, "learning_rate": 3.3634730633942475e-06, "loss": 0.7825, "step": 7242 }, { "epoch": 0.62, "learning_rate": 3.3621673507006196e-06, "loss": 0.848, "step": 7243 }, { "epoch": 0.62, "learning_rate": 3.3608617631128216e-06, "loss": 0.7913, "step": 7244 }, { "epoch": 0.62, "learning_rate": 3.359556300730581e-06, "loss": 0.764, "step": 7245 }, { "epoch": 0.62, "learning_rate": 3.3582509636536153e-06, "loss": 0.8481, "step": 7246 }, { "epoch": 0.62, "learning_rate": 3.3569457519816325e-06, "loss": 0.8256, "step": 7247 }, { "epoch": 0.62, "learning_rate": 3.355640665814329e-06, "loss": 0.8086, "step": 7248 }, { "epoch": 0.62, "learning_rate": 3.354335705251397e-06, "loss": 0.8002, "step": 7249 }, { "epoch": 0.62, "learning_rate": 3.3530308703925154e-06, "loss": 0.8354, "step": 7250 }, { "epoch": 0.62, "learning_rate": 3.351726161337352e-06, "loss": 0.5716, "step": 7251 }, { "epoch": 0.62, "learning_rate": 3.3504215781855683e-06, "loss": 0.7778, "step": 7252 }, { "epoch": 0.62, "learning_rate": 3.3491171210368167e-06, "loss": 0.525, "step": 7253 }, { "epoch": 0.62, "learning_rate": 3.3478127899907343e-06, "loss": 0.8254, "step": 7254 }, { "epoch": 0.62, "learning_rate": 3.346508585146955e-06, "loss": 0.8219, "step": 7255 }, { "epoch": 0.62, "learning_rate": 3.3452045066051002e-06, "loss": 0.7697, "step": 7256 }, { "epoch": 0.62, "learning_rate": 3.3439005544647805e-06, "loss": 0.8201, "step": 7257 }, { "epoch": 0.62, "learning_rate": 3.3425967288256e-06, "loss": 0.8261, "step": 7258 }, { "epoch": 0.62, "learning_rate": 3.3412930297871516e-06, "loss": 0.7845, "step": 7259 }, { "epoch": 0.62, "learning_rate": 3.3399894574490176e-06, "loss": 0.8563, "step": 7260 }, { "epoch": 0.62, "learning_rate": 3.338686011910771e-06, "loss": 0.8208, "step": 7261 }, { "epoch": 0.62, "learning_rate": 3.337382693271975e-06, "loss": 0.836, "step": 7262 }, { "epoch": 0.62, "learning_rate": 3.3360795016321845e-06, "loss": 0.5796, "step": 7263 }, { "epoch": 0.62, "learning_rate": 3.334776437090944e-06, "loss": 0.8353, "step": 7264 }, { "epoch": 0.62, "learning_rate": 3.333473499747787e-06, "loss": 0.837, "step": 7265 }, { "epoch": 0.62, "learning_rate": 3.332170689702241e-06, "loss": 0.8393, "step": 7266 }, { "epoch": 0.62, "learning_rate": 3.3308680070538173e-06, "loss": 0.8301, "step": 7267 }, { "epoch": 0.62, "learning_rate": 3.329565451902023e-06, "loss": 0.7968, "step": 7268 }, { "epoch": 0.62, "learning_rate": 3.3282630243463552e-06, "loss": 0.8414, "step": 7269 }, { "epoch": 0.62, "learning_rate": 3.326960724486297e-06, "loss": 0.8474, "step": 7270 }, { "epoch": 0.62, "learning_rate": 3.325658552421327e-06, "loss": 0.8999, "step": 7271 }, { "epoch": 0.62, "learning_rate": 3.32435650825091e-06, "loss": 0.4897, "step": 7272 }, { "epoch": 0.62, "learning_rate": 3.3230545920745055e-06, "loss": 0.8308, "step": 7273 }, { "epoch": 0.62, "learning_rate": 3.321752803991556e-06, "loss": 0.8049, "step": 7274 }, { "epoch": 0.62, "learning_rate": 3.320451144101502e-06, "loss": 0.8296, "step": 7275 }, { "epoch": 0.62, "learning_rate": 3.3191496125037687e-06, "loss": 0.7863, "step": 7276 }, { "epoch": 0.62, "learning_rate": 3.317848209297774e-06, "loss": 0.8121, "step": 7277 }, { "epoch": 0.62, "learning_rate": 3.316546934582927e-06, "loss": 0.7717, "step": 7278 }, { "epoch": 0.62, "learning_rate": 3.3152457884586253e-06, "loss": 0.8172, "step": 7279 }, { "epoch": 0.62, "learning_rate": 3.3139447710242557e-06, "loss": 0.7891, "step": 7280 }, { "epoch": 0.62, "learning_rate": 3.3126438823791955e-06, "loss": 0.7851, "step": 7281 }, { "epoch": 0.62, "learning_rate": 3.3113431226228147e-06, "loss": 0.8299, "step": 7282 }, { "epoch": 0.62, "learning_rate": 3.3100424918544726e-06, "loss": 0.8234, "step": 7283 }, { "epoch": 0.62, "learning_rate": 3.3087419901735162e-06, "loss": 0.785, "step": 7284 }, { "epoch": 0.62, "learning_rate": 3.3074416176792844e-06, "loss": 0.7568, "step": 7285 }, { "epoch": 0.62, "learning_rate": 3.306141374471109e-06, "loss": 0.4958, "step": 7286 }, { "epoch": 0.62, "learning_rate": 3.3048412606483045e-06, "loss": 0.8123, "step": 7287 }, { "epoch": 0.62, "learning_rate": 3.303541276310184e-06, "loss": 0.738, "step": 7288 }, { "epoch": 0.62, "learning_rate": 3.302241421556043e-06, "loss": 0.8116, "step": 7289 }, { "epoch": 0.62, "learning_rate": 3.3009416964851734e-06, "loss": 0.8184, "step": 7290 }, { "epoch": 0.62, "learning_rate": 3.2996421011968546e-06, "loss": 0.8247, "step": 7291 }, { "epoch": 0.62, "learning_rate": 3.2983426357903563e-06, "loss": 0.8346, "step": 7292 }, { "epoch": 0.62, "learning_rate": 3.297043300364938e-06, "loss": 0.8645, "step": 7293 }, { "epoch": 0.62, "learning_rate": 3.2957440950198494e-06, "loss": 0.8569, "step": 7294 }, { "epoch": 0.62, "learning_rate": 3.2944450198543285e-06, "loss": 0.8519, "step": 7295 }, { "epoch": 0.62, "learning_rate": 3.293146074967607e-06, "loss": 0.8222, "step": 7296 }, { "epoch": 0.62, "learning_rate": 3.291847260458906e-06, "loss": 0.7553, "step": 7297 }, { "epoch": 0.62, "learning_rate": 3.290548576427432e-06, "loss": 0.7891, "step": 7298 }, { "epoch": 0.62, "learning_rate": 3.289250022972389e-06, "loss": 0.7871, "step": 7299 }, { "epoch": 0.62, "learning_rate": 3.2879516001929633e-06, "loss": 0.8028, "step": 7300 }, { "epoch": 0.62, "learning_rate": 3.2866533081883355e-06, "loss": 0.7924, "step": 7301 }, { "epoch": 0.62, "learning_rate": 3.2853551470576782e-06, "loss": 0.5247, "step": 7302 }, { "epoch": 0.62, "learning_rate": 3.284057116900149e-06, "loss": 0.7913, "step": 7303 }, { "epoch": 0.62, "learning_rate": 3.2827592178148985e-06, "loss": 0.5832, "step": 7304 }, { "epoch": 0.62, "learning_rate": 3.2814614499010674e-06, "loss": 0.835, "step": 7305 }, { "epoch": 0.62, "learning_rate": 3.2801638132577867e-06, "loss": 0.8198, "step": 7306 }, { "epoch": 0.62, "learning_rate": 3.2788663079841743e-06, "loss": 0.8204, "step": 7307 }, { "epoch": 0.62, "learning_rate": 3.2775689341793393e-06, "loss": 0.8154, "step": 7308 }, { "epoch": 0.62, "learning_rate": 3.276271691942383e-06, "loss": 0.5634, "step": 7309 }, { "epoch": 0.62, "learning_rate": 3.274974581372395e-06, "loss": 0.8537, "step": 7310 }, { "epoch": 0.62, "learning_rate": 3.273677602568457e-06, "loss": 0.8658, "step": 7311 }, { "epoch": 0.62, "learning_rate": 3.272380755629635e-06, "loss": 0.81, "step": 7312 }, { "epoch": 0.62, "learning_rate": 3.2710840406549937e-06, "loss": 0.8022, "step": 7313 }, { "epoch": 0.62, "learning_rate": 3.269787457743577e-06, "loss": 0.5324, "step": 7314 }, { "epoch": 0.62, "learning_rate": 3.2684910069944263e-06, "loss": 0.8184, "step": 7315 }, { "epoch": 0.62, "learning_rate": 3.267194688506574e-06, "loss": 0.8352, "step": 7316 }, { "epoch": 0.62, "learning_rate": 3.2658985023790356e-06, "loss": 0.8159, "step": 7317 }, { "epoch": 0.62, "learning_rate": 3.264602448710822e-06, "loss": 0.8046, "step": 7318 }, { "epoch": 0.62, "learning_rate": 3.2633065276009337e-06, "loss": 0.7963, "step": 7319 }, { "epoch": 0.62, "learning_rate": 3.2620107391483552e-06, "loss": 0.8195, "step": 7320 }, { "epoch": 0.62, "learning_rate": 3.2607150834520697e-06, "loss": 0.7493, "step": 7321 }, { "epoch": 0.62, "learning_rate": 3.2594195606110423e-06, "loss": 0.8324, "step": 7322 }, { "epoch": 0.62, "learning_rate": 3.258124170724234e-06, "loss": 0.7771, "step": 7323 }, { "epoch": 0.62, "learning_rate": 3.256828913890592e-06, "loss": 0.5431, "step": 7324 }, { "epoch": 0.63, "learning_rate": 3.2555337902090566e-06, "loss": 0.8511, "step": 7325 }, { "epoch": 0.63, "learning_rate": 3.2542387997785545e-06, "loss": 0.842, "step": 7326 }, { "epoch": 0.63, "learning_rate": 3.2529439426980015e-06, "loss": 0.8667, "step": 7327 }, { "epoch": 0.63, "learning_rate": 3.251649219066307e-06, "loss": 0.8527, "step": 7328 }, { "epoch": 0.63, "learning_rate": 3.250354628982368e-06, "loss": 0.8181, "step": 7329 }, { "epoch": 0.63, "learning_rate": 3.2490601725450732e-06, "loss": 0.807, "step": 7330 }, { "epoch": 0.63, "learning_rate": 3.247765849853297e-06, "loss": 0.812, "step": 7331 }, { "epoch": 0.63, "learning_rate": 3.2464716610059087e-06, "loss": 0.8737, "step": 7332 }, { "epoch": 0.63, "learning_rate": 3.245177606101766e-06, "loss": 0.8067, "step": 7333 }, { "epoch": 0.63, "learning_rate": 3.2438836852397103e-06, "loss": 0.8145, "step": 7334 }, { "epoch": 0.63, "learning_rate": 3.242589898518582e-06, "loss": 0.5215, "step": 7335 }, { "epoch": 0.63, "learning_rate": 3.2412962460372054e-06, "loss": 0.778, "step": 7336 }, { "epoch": 0.63, "learning_rate": 3.240002727894396e-06, "loss": 0.793, "step": 7337 }, { "epoch": 0.63, "learning_rate": 3.2387093441889595e-06, "loss": 0.7812, "step": 7338 }, { "epoch": 0.63, "learning_rate": 3.2374160950196933e-06, "loss": 0.8518, "step": 7339 }, { "epoch": 0.63, "learning_rate": 3.2361229804853776e-06, "loss": 0.844, "step": 7340 }, { "epoch": 0.63, "learning_rate": 3.23483000068479e-06, "loss": 0.8218, "step": 7341 }, { "epoch": 0.63, "learning_rate": 3.2335371557166926e-06, "loss": 0.823, "step": 7342 }, { "epoch": 0.63, "learning_rate": 3.2322444456798417e-06, "loss": 0.5116, "step": 7343 }, { "epoch": 0.63, "learning_rate": 3.2309518706729814e-06, "loss": 0.815, "step": 7344 }, { "epoch": 0.63, "learning_rate": 3.2296594307948428e-06, "loss": 0.7681, "step": 7345 }, { "epoch": 0.63, "learning_rate": 3.228367126144151e-06, "loss": 0.8098, "step": 7346 }, { "epoch": 0.63, "learning_rate": 3.227074956819616e-06, "loss": 0.6024, "step": 7347 }, { "epoch": 0.63, "learning_rate": 3.225782922919941e-06, "loss": 0.7963, "step": 7348 }, { "epoch": 0.63, "learning_rate": 3.22449102454382e-06, "loss": 0.8465, "step": 7349 }, { "epoch": 0.63, "learning_rate": 3.223199261789932e-06, "loss": 0.8254, "step": 7350 }, { "epoch": 0.63, "learning_rate": 3.22190763475695e-06, "loss": 0.8411, "step": 7351 }, { "epoch": 0.63, "learning_rate": 3.2206161435435363e-06, "loss": 0.8356, "step": 7352 }, { "epoch": 0.63, "learning_rate": 3.2193247882483373e-06, "loss": 0.8009, "step": 7353 }, { "epoch": 0.63, "learning_rate": 3.218033568969997e-06, "loss": 0.7856, "step": 7354 }, { "epoch": 0.63, "learning_rate": 3.2167424858071423e-06, "loss": 0.5378, "step": 7355 }, { "epoch": 0.63, "learning_rate": 3.215451538858394e-06, "loss": 0.8254, "step": 7356 }, { "epoch": 0.63, "learning_rate": 3.214160728222361e-06, "loss": 0.8096, "step": 7357 }, { "epoch": 0.63, "learning_rate": 3.212870053997642e-06, "loss": 0.8407, "step": 7358 }, { "epoch": 0.63, "learning_rate": 3.2115795162828265e-06, "loss": 0.8274, "step": 7359 }, { "epoch": 0.63, "learning_rate": 3.210289115176489e-06, "loss": 0.8399, "step": 7360 }, { "epoch": 0.63, "learning_rate": 3.2089988507771984e-06, "loss": 0.5787, "step": 7361 }, { "epoch": 0.63, "learning_rate": 3.207708723183511e-06, "loss": 0.8432, "step": 7362 }, { "epoch": 0.63, "learning_rate": 3.2064187324939753e-06, "loss": 0.7925, "step": 7363 }, { "epoch": 0.63, "learning_rate": 3.205128878807125e-06, "loss": 0.7888, "step": 7364 }, { "epoch": 0.63, "learning_rate": 3.203839162221486e-06, "loss": 0.8334, "step": 7365 }, { "epoch": 0.63, "learning_rate": 3.202549582835575e-06, "loss": 0.5359, "step": 7366 }, { "epoch": 0.63, "learning_rate": 3.2012601407478937e-06, "loss": 0.8039, "step": 7367 }, { "epoch": 0.63, "learning_rate": 3.1999708360569383e-06, "loss": 0.5016, "step": 7368 }, { "epoch": 0.63, "learning_rate": 3.19868166886119e-06, "loss": 0.8012, "step": 7369 }, { "epoch": 0.63, "learning_rate": 3.1973926392591236e-06, "loss": 0.8136, "step": 7370 }, { "epoch": 0.63, "learning_rate": 3.196103747349203e-06, "loss": 0.8464, "step": 7371 }, { "epoch": 0.63, "learning_rate": 3.194814993229878e-06, "loss": 0.7911, "step": 7372 }, { "epoch": 0.63, "learning_rate": 3.19352637699959e-06, "loss": 0.803, "step": 7373 }, { "epoch": 0.63, "learning_rate": 3.1922378987567703e-06, "loss": 0.7776, "step": 7374 }, { "epoch": 0.63, "learning_rate": 3.190949558599839e-06, "loss": 0.8869, "step": 7375 }, { "epoch": 0.63, "learning_rate": 3.189661356627207e-06, "loss": 0.5676, "step": 7376 }, { "epoch": 0.63, "learning_rate": 3.1883732929372737e-06, "loss": 0.7699, "step": 7377 }, { "epoch": 0.63, "learning_rate": 3.1870853676284264e-06, "loss": 0.7719, "step": 7378 }, { "epoch": 0.63, "learning_rate": 3.185797580799046e-06, "loss": 0.8119, "step": 7379 }, { "epoch": 0.63, "learning_rate": 3.184509932547496e-06, "loss": 0.8128, "step": 7380 }, { "epoch": 0.63, "learning_rate": 3.183222422972135e-06, "loss": 0.8482, "step": 7381 }, { "epoch": 0.63, "learning_rate": 3.181935052171312e-06, "loss": 0.8036, "step": 7382 }, { "epoch": 0.63, "learning_rate": 3.1806478202433593e-06, "loss": 0.7949, "step": 7383 }, { "epoch": 0.63, "learning_rate": 3.1793607272866033e-06, "loss": 0.7979, "step": 7384 }, { "epoch": 0.63, "learning_rate": 3.1780737733993605e-06, "loss": 0.7744, "step": 7385 }, { "epoch": 0.63, "learning_rate": 3.1767869586799325e-06, "loss": 0.8134, "step": 7386 }, { "epoch": 0.63, "learning_rate": 3.1755002832266137e-06, "loss": 0.8536, "step": 7387 }, { "epoch": 0.63, "learning_rate": 3.1742137471376857e-06, "loss": 0.8509, "step": 7388 }, { "epoch": 0.63, "learning_rate": 3.1729273505114203e-06, "loss": 0.8585, "step": 7389 }, { "epoch": 0.63, "learning_rate": 3.1716410934460816e-06, "loss": 0.8172, "step": 7390 }, { "epoch": 0.63, "learning_rate": 3.1703549760399176e-06, "loss": 0.8079, "step": 7391 }, { "epoch": 0.63, "learning_rate": 3.169068998391171e-06, "loss": 0.826, "step": 7392 }, { "epoch": 0.63, "learning_rate": 3.1677831605980674e-06, "loss": 0.842, "step": 7393 }, { "epoch": 0.63, "learning_rate": 3.166497462758828e-06, "loss": 0.7858, "step": 7394 }, { "epoch": 0.63, "learning_rate": 3.16521190497166e-06, "loss": 0.792, "step": 7395 }, { "epoch": 0.63, "learning_rate": 3.163926487334763e-06, "loss": 0.8121, "step": 7396 }, { "epoch": 0.63, "learning_rate": 3.16264120994632e-06, "loss": 0.7652, "step": 7397 }, { "epoch": 0.63, "learning_rate": 3.161356072904509e-06, "loss": 0.8292, "step": 7398 }, { "epoch": 0.63, "learning_rate": 3.1600710763074972e-06, "loss": 0.7998, "step": 7399 }, { "epoch": 0.63, "learning_rate": 3.158786220253434e-06, "loss": 0.8225, "step": 7400 }, { "epoch": 0.63, "learning_rate": 3.1575015048404674e-06, "loss": 0.8716, "step": 7401 }, { "epoch": 0.63, "learning_rate": 3.156216930166728e-06, "loss": 0.8047, "step": 7402 }, { "epoch": 0.63, "learning_rate": 3.154932496330338e-06, "loss": 0.8119, "step": 7403 }, { "epoch": 0.63, "learning_rate": 3.153648203429411e-06, "loss": 0.8253, "step": 7404 }, { "epoch": 0.63, "learning_rate": 3.152364051562047e-06, "loss": 0.8141, "step": 7405 }, { "epoch": 0.63, "learning_rate": 3.1510800408263342e-06, "loss": 0.7703, "step": 7406 }, { "epoch": 0.63, "learning_rate": 3.149796171320353e-06, "loss": 0.8422, "step": 7407 }, { "epoch": 0.63, "learning_rate": 3.148512443142171e-06, "loss": 0.7941, "step": 7408 }, { "epoch": 0.63, "learning_rate": 3.1472288563898467e-06, "loss": 0.8128, "step": 7409 }, { "epoch": 0.63, "learning_rate": 3.145945411161427e-06, "loss": 0.7843, "step": 7410 }, { "epoch": 0.63, "learning_rate": 3.1446621075549468e-06, "loss": 0.8326, "step": 7411 }, { "epoch": 0.63, "learning_rate": 3.143378945668433e-06, "loss": 0.7825, "step": 7412 }, { "epoch": 0.63, "learning_rate": 3.142095925599898e-06, "loss": 0.8133, "step": 7413 }, { "epoch": 0.63, "learning_rate": 3.1408130474473453e-06, "loss": 0.8099, "step": 7414 }, { "epoch": 0.63, "learning_rate": 3.1395303113087695e-06, "loss": 0.795, "step": 7415 }, { "epoch": 0.63, "learning_rate": 3.13824771728215e-06, "loss": 0.7766, "step": 7416 }, { "epoch": 0.63, "learning_rate": 3.1369652654654593e-06, "loss": 0.8052, "step": 7417 }, { "epoch": 0.63, "learning_rate": 3.1356829559566572e-06, "loss": 0.7788, "step": 7418 }, { "epoch": 0.63, "learning_rate": 3.134400788853693e-06, "loss": 0.8382, "step": 7419 }, { "epoch": 0.63, "learning_rate": 3.133118764254505e-06, "loss": 0.8082, "step": 7420 }, { "epoch": 0.63, "learning_rate": 3.1318368822570194e-06, "loss": 0.8517, "step": 7421 }, { "epoch": 0.63, "learning_rate": 3.130555142959154e-06, "loss": 0.7938, "step": 7422 }, { "epoch": 0.63, "learning_rate": 3.1292735464588143e-06, "loss": 0.7953, "step": 7423 }, { "epoch": 0.63, "learning_rate": 3.1279920928538944e-06, "loss": 0.8124, "step": 7424 }, { "epoch": 0.63, "learning_rate": 3.126710782242281e-06, "loss": 0.8387, "step": 7425 }, { "epoch": 0.63, "learning_rate": 3.125429614721842e-06, "loss": 0.8052, "step": 7426 }, { "epoch": 0.63, "learning_rate": 3.124148590390442e-06, "loss": 0.794, "step": 7427 }, { "epoch": 0.63, "learning_rate": 3.1228677093459336e-06, "loss": 0.7869, "step": 7428 }, { "epoch": 0.63, "learning_rate": 3.1215869716861536e-06, "loss": 0.7833, "step": 7429 }, { "epoch": 0.63, "learning_rate": 3.1203063775089336e-06, "loss": 0.8353, "step": 7430 }, { "epoch": 0.63, "learning_rate": 3.1190259269120905e-06, "loss": 0.8469, "step": 7431 }, { "epoch": 0.63, "learning_rate": 3.1177456199934344e-06, "loss": 0.7873, "step": 7432 }, { "epoch": 0.63, "learning_rate": 3.116465456850758e-06, "loss": 0.7898, "step": 7433 }, { "epoch": 0.63, "learning_rate": 3.115185437581846e-06, "loss": 0.8634, "step": 7434 }, { "epoch": 0.63, "learning_rate": 3.1139055622844748e-06, "loss": 0.8178, "step": 7435 }, { "epoch": 0.63, "learning_rate": 3.112625831056407e-06, "loss": 0.8207, "step": 7436 }, { "epoch": 0.63, "learning_rate": 3.1113462439953966e-06, "loss": 0.8108, "step": 7437 }, { "epoch": 0.63, "learning_rate": 3.1100668011991826e-06, "loss": 0.8399, "step": 7438 }, { "epoch": 0.63, "learning_rate": 3.1087875027654975e-06, "loss": 0.8083, "step": 7439 }, { "epoch": 0.63, "learning_rate": 3.1075083487920564e-06, "loss": 0.8079, "step": 7440 }, { "epoch": 0.63, "learning_rate": 3.1062293393765708e-06, "loss": 0.8374, "step": 7441 }, { "epoch": 0.64, "learning_rate": 3.104950474616738e-06, "loss": 0.8214, "step": 7442 }, { "epoch": 0.64, "learning_rate": 3.1036717546102413e-06, "loss": 0.7474, "step": 7443 }, { "epoch": 0.64, "learning_rate": 3.102393179454758e-06, "loss": 0.8285, "step": 7444 }, { "epoch": 0.64, "learning_rate": 3.101114749247954e-06, "loss": 0.8636, "step": 7445 }, { "epoch": 0.64, "learning_rate": 3.0998364640874766e-06, "loss": 0.8096, "step": 7446 }, { "epoch": 0.64, "learning_rate": 3.0985583240709728e-06, "loss": 0.8369, "step": 7447 }, { "epoch": 0.64, "learning_rate": 3.0972803292960694e-06, "loss": 0.7937, "step": 7448 }, { "epoch": 0.64, "learning_rate": 3.0960024798603884e-06, "loss": 0.8433, "step": 7449 }, { "epoch": 0.64, "learning_rate": 3.094724775861538e-06, "loss": 0.8487, "step": 7450 }, { "epoch": 0.64, "learning_rate": 3.0934472173971157e-06, "loss": 0.777, "step": 7451 }, { "epoch": 0.64, "learning_rate": 3.092169804564709e-06, "loss": 0.8312, "step": 7452 }, { "epoch": 0.64, "learning_rate": 3.090892537461889e-06, "loss": 0.5348, "step": 7453 }, { "epoch": 0.64, "learning_rate": 3.0896154161862228e-06, "loss": 0.8082, "step": 7454 }, { "epoch": 0.64, "learning_rate": 3.0883384408352624e-06, "loss": 0.7775, "step": 7455 }, { "epoch": 0.64, "learning_rate": 3.087061611506551e-06, "loss": 0.7977, "step": 7456 }, { "epoch": 0.64, "learning_rate": 3.085784928297618e-06, "loss": 0.7675, "step": 7457 }, { "epoch": 0.64, "learning_rate": 3.0845083913059812e-06, "loss": 0.8109, "step": 7458 }, { "epoch": 0.64, "learning_rate": 3.0832320006291536e-06, "loss": 0.8441, "step": 7459 }, { "epoch": 0.64, "learning_rate": 3.0819557563646273e-06, "loss": 0.8341, "step": 7460 }, { "epoch": 0.64, "learning_rate": 3.0806796586098918e-06, "loss": 0.8568, "step": 7461 }, { "epoch": 0.64, "learning_rate": 3.0794037074624183e-06, "loss": 0.8144, "step": 7462 }, { "epoch": 0.64, "learning_rate": 3.0781279030196727e-06, "loss": 0.828, "step": 7463 }, { "epoch": 0.64, "learning_rate": 3.076852245379106e-06, "loss": 0.8152, "step": 7464 }, { "epoch": 0.64, "learning_rate": 3.075576734638163e-06, "loss": 0.8083, "step": 7465 }, { "epoch": 0.64, "learning_rate": 3.0743013708942695e-06, "loss": 0.8478, "step": 7466 }, { "epoch": 0.64, "learning_rate": 3.073026154244845e-06, "loss": 0.7831, "step": 7467 }, { "epoch": 0.64, "learning_rate": 3.0717510847872955e-06, "loss": 0.8057, "step": 7468 }, { "epoch": 0.64, "learning_rate": 3.0704761626190204e-06, "loss": 0.7461, "step": 7469 }, { "epoch": 0.64, "learning_rate": 3.0692013878374037e-06, "loss": 0.8379, "step": 7470 }, { "epoch": 0.64, "learning_rate": 3.067926760539818e-06, "loss": 0.7999, "step": 7471 }, { "epoch": 0.64, "learning_rate": 3.0666522808236276e-06, "loss": 0.7608, "step": 7472 }, { "epoch": 0.64, "learning_rate": 3.06537794878618e-06, "loss": 0.7666, "step": 7473 }, { "epoch": 0.64, "learning_rate": 3.064103764524817e-06, "loss": 0.7805, "step": 7474 }, { "epoch": 0.64, "learning_rate": 3.0628297281368695e-06, "loss": 0.8144, "step": 7475 }, { "epoch": 0.64, "learning_rate": 3.061555839719651e-06, "loss": 0.7478, "step": 7476 }, { "epoch": 0.64, "learning_rate": 3.0602820993704695e-06, "loss": 0.7795, "step": 7477 }, { "epoch": 0.64, "learning_rate": 3.0590085071866216e-06, "loss": 0.7926, "step": 7478 }, { "epoch": 0.64, "learning_rate": 3.0577350632653854e-06, "loss": 0.8098, "step": 7479 }, { "epoch": 0.64, "learning_rate": 3.056461767704037e-06, "loss": 0.8323, "step": 7480 }, { "epoch": 0.64, "learning_rate": 3.055188620599835e-06, "loss": 0.7934, "step": 7481 }, { "epoch": 0.64, "learning_rate": 3.0539156220500294e-06, "loss": 0.813, "step": 7482 }, { "epoch": 0.64, "learning_rate": 3.0526427721518585e-06, "loss": 0.8281, "step": 7483 }, { "epoch": 0.64, "learning_rate": 3.0513700710025497e-06, "loss": 0.5615, "step": 7484 }, { "epoch": 0.64, "learning_rate": 3.050097518699319e-06, "loss": 0.8371, "step": 7485 }, { "epoch": 0.64, "learning_rate": 3.0488251153393655e-06, "loss": 0.797, "step": 7486 }, { "epoch": 0.64, "learning_rate": 3.0475528610198856e-06, "loss": 0.7972, "step": 7487 }, { "epoch": 0.64, "learning_rate": 3.046280755838059e-06, "loss": 0.798, "step": 7488 }, { "epoch": 0.64, "learning_rate": 3.0450087998910582e-06, "loss": 0.8431, "step": 7489 }, { "epoch": 0.64, "learning_rate": 3.043736993276038e-06, "loss": 0.7909, "step": 7490 }, { "epoch": 0.64, "learning_rate": 3.042465336090147e-06, "loss": 0.7681, "step": 7491 }, { "epoch": 0.64, "learning_rate": 3.041193828430522e-06, "loss": 0.8153, "step": 7492 }, { "epoch": 0.64, "learning_rate": 3.0399224703942844e-06, "loss": 0.802, "step": 7493 }, { "epoch": 0.64, "learning_rate": 3.038651262078548e-06, "loss": 0.8251, "step": 7494 }, { "epoch": 0.64, "learning_rate": 3.037380203580414e-06, "loss": 0.8053, "step": 7495 }, { "epoch": 0.64, "learning_rate": 3.036109294996972e-06, "loss": 0.8181, "step": 7496 }, { "epoch": 0.64, "learning_rate": 3.034838536425301e-06, "loss": 0.7699, "step": 7497 }, { "epoch": 0.64, "learning_rate": 3.0335679279624686e-06, "loss": 0.8253, "step": 7498 }, { "epoch": 0.64, "learning_rate": 3.032297469705529e-06, "loss": 0.8604, "step": 7499 }, { "epoch": 0.64, "learning_rate": 3.0310271617515254e-06, "loss": 0.7979, "step": 7500 }, { "epoch": 0.64, "learning_rate": 3.029757004197491e-06, "loss": 0.8196, "step": 7501 }, { "epoch": 0.64, "learning_rate": 3.028486997140446e-06, "loss": 0.7817, "step": 7502 }, { "epoch": 0.64, "learning_rate": 3.027217140677403e-06, "loss": 0.7751, "step": 7503 }, { "epoch": 0.64, "learning_rate": 3.0259474349053563e-06, "loss": 0.8176, "step": 7504 }, { "epoch": 0.64, "learning_rate": 3.024677879921296e-06, "loss": 0.7944, "step": 7505 }, { "epoch": 0.64, "learning_rate": 3.0234084758221934e-06, "loss": 0.8336, "step": 7506 }, { "epoch": 0.64, "learning_rate": 3.0221392227050126e-06, "loss": 0.8293, "step": 7507 }, { "epoch": 0.64, "learning_rate": 3.0208701206667075e-06, "loss": 0.7502, "step": 7508 }, { "epoch": 0.64, "learning_rate": 3.019601169804216e-06, "loss": 0.8206, "step": 7509 }, { "epoch": 0.64, "learning_rate": 3.0183323702144686e-06, "loss": 0.7559, "step": 7510 }, { "epoch": 0.64, "learning_rate": 3.017063721994383e-06, "loss": 0.8131, "step": 7511 }, { "epoch": 0.64, "learning_rate": 3.015795225240864e-06, "loss": 0.7982, "step": 7512 }, { "epoch": 0.64, "learning_rate": 3.0145268800508052e-06, "loss": 0.8283, "step": 7513 }, { "epoch": 0.64, "learning_rate": 3.013258686521089e-06, "loss": 0.839, "step": 7514 }, { "epoch": 0.64, "learning_rate": 3.011990644748587e-06, "loss": 0.8294, "step": 7515 }, { "epoch": 0.64, "learning_rate": 3.0107227548301575e-06, "loss": 0.7963, "step": 7516 }, { "epoch": 0.64, "learning_rate": 3.0094550168626514e-06, "loss": 0.8597, "step": 7517 }, { "epoch": 0.64, "learning_rate": 3.0081874309429037e-06, "loss": 0.8233, "step": 7518 }, { "epoch": 0.64, "learning_rate": 3.0069199971677354e-06, "loss": 0.7852, "step": 7519 }, { "epoch": 0.64, "learning_rate": 3.005652715633962e-06, "loss": 0.8184, "step": 7520 }, { "epoch": 0.64, "learning_rate": 3.0043855864383853e-06, "loss": 0.5338, "step": 7521 }, { "epoch": 0.64, "learning_rate": 3.0031186096777955e-06, "loss": 0.8106, "step": 7522 }, { "epoch": 0.64, "learning_rate": 3.0018517854489676e-06, "loss": 0.8054, "step": 7523 }, { "epoch": 0.64, "learning_rate": 3.000585113848671e-06, "loss": 0.8081, "step": 7524 }, { "epoch": 0.64, "learning_rate": 2.9993185949736603e-06, "loss": 0.7631, "step": 7525 }, { "epoch": 0.64, "learning_rate": 2.9980522289206758e-06, "loss": 0.836, "step": 7526 }, { "epoch": 0.64, "learning_rate": 2.9967860157864515e-06, "loss": 0.8082, "step": 7527 }, { "epoch": 0.64, "learning_rate": 2.995519955667705e-06, "loss": 0.7807, "step": 7528 }, { "epoch": 0.64, "learning_rate": 2.994254048661146e-06, "loss": 0.8263, "step": 7529 }, { "epoch": 0.64, "learning_rate": 2.99298829486347e-06, "loss": 0.8248, "step": 7530 }, { "epoch": 0.64, "learning_rate": 2.991722694371363e-06, "loss": 0.8242, "step": 7531 }, { "epoch": 0.64, "learning_rate": 2.990457247281496e-06, "loss": 0.7862, "step": 7532 }, { "epoch": 0.64, "learning_rate": 2.9891919536905305e-06, "loss": 0.8296, "step": 7533 }, { "epoch": 0.64, "learning_rate": 2.9879268136951163e-06, "loss": 0.8353, "step": 7534 }, { "epoch": 0.64, "learning_rate": 2.9866618273918906e-06, "loss": 0.7973, "step": 7535 }, { "epoch": 0.64, "learning_rate": 2.985396994877481e-06, "loss": 0.8453, "step": 7536 }, { "epoch": 0.64, "learning_rate": 2.984132316248499e-06, "loss": 0.8149, "step": 7537 }, { "epoch": 0.64, "learning_rate": 2.9828677916015514e-06, "loss": 0.8143, "step": 7538 }, { "epoch": 0.64, "learning_rate": 2.9816034210332244e-06, "loss": 0.7534, "step": 7539 }, { "epoch": 0.64, "learning_rate": 2.980339204640098e-06, "loss": 0.8161, "step": 7540 }, { "epoch": 0.64, "learning_rate": 2.9790751425187414e-06, "loss": 0.7965, "step": 7541 }, { "epoch": 0.64, "learning_rate": 2.977811234765707e-06, "loss": 0.8237, "step": 7542 }, { "epoch": 0.64, "learning_rate": 2.9765474814775406e-06, "loss": 0.8312, "step": 7543 }, { "epoch": 0.64, "learning_rate": 2.975283882750774e-06, "loss": 0.8272, "step": 7544 }, { "epoch": 0.64, "learning_rate": 2.9740204386819272e-06, "loss": 0.8398, "step": 7545 }, { "epoch": 0.64, "learning_rate": 2.972757149367507e-06, "loss": 0.8094, "step": 7546 }, { "epoch": 0.64, "learning_rate": 2.9714940149040094e-06, "loss": 0.825, "step": 7547 }, { "epoch": 0.64, "learning_rate": 2.9702310353879203e-06, "loss": 0.7519, "step": 7548 }, { "epoch": 0.64, "learning_rate": 2.9689682109157125e-06, "loss": 0.7716, "step": 7549 }, { "epoch": 0.64, "learning_rate": 2.9677055415838463e-06, "loss": 0.7576, "step": 7550 }, { "epoch": 0.64, "learning_rate": 2.966443027488772e-06, "loss": 0.8084, "step": 7551 }, { "epoch": 0.64, "learning_rate": 2.9651806687269234e-06, "loss": 0.7892, "step": 7552 }, { "epoch": 0.64, "learning_rate": 2.9639184653947284e-06, "loss": 0.784, "step": 7553 }, { "epoch": 0.64, "learning_rate": 2.9626564175886e-06, "loss": 0.8046, "step": 7554 }, { "epoch": 0.64, "learning_rate": 2.961394525404939e-06, "loss": 0.8118, "step": 7555 }, { "epoch": 0.64, "learning_rate": 2.9601327889401356e-06, "loss": 0.8039, "step": 7556 }, { "epoch": 0.64, "learning_rate": 2.9588712082905678e-06, "loss": 0.7722, "step": 7557 }, { "epoch": 0.64, "learning_rate": 2.9576097835526025e-06, "loss": 0.857, "step": 7558 }, { "epoch": 0.65, "learning_rate": 2.95634851482259e-06, "loss": 0.7927, "step": 7559 }, { "epoch": 0.65, "learning_rate": 2.955087402196877e-06, "loss": 0.789, "step": 7560 }, { "epoch": 0.65, "learning_rate": 2.953826445771788e-06, "loss": 0.7792, "step": 7561 }, { "epoch": 0.65, "learning_rate": 2.9525656456436456e-06, "loss": 0.8239, "step": 7562 }, { "epoch": 0.65, "learning_rate": 2.9513050019087552e-06, "loss": 0.8464, "step": 7563 }, { "epoch": 0.65, "learning_rate": 2.9500445146634094e-06, "loss": 0.5613, "step": 7564 }, { "epoch": 0.65, "learning_rate": 2.9487841840038945e-06, "loss": 0.8157, "step": 7565 }, { "epoch": 0.65, "learning_rate": 2.9475240100264747e-06, "loss": 0.8223, "step": 7566 }, { "epoch": 0.65, "learning_rate": 2.946263992827412e-06, "loss": 0.7989, "step": 7567 }, { "epoch": 0.65, "learning_rate": 2.9450041325029537e-06, "loss": 0.788, "step": 7568 }, { "epoch": 0.65, "learning_rate": 2.9437444291493323e-06, "loss": 0.7719, "step": 7569 }, { "epoch": 0.65, "learning_rate": 2.94248488286277e-06, "loss": 0.8574, "step": 7570 }, { "epoch": 0.65, "learning_rate": 2.94122549373948e-06, "loss": 0.8186, "step": 7571 }, { "epoch": 0.65, "learning_rate": 2.9399662618756573e-06, "loss": 0.8016, "step": 7572 }, { "epoch": 0.65, "learning_rate": 2.938707187367491e-06, "loss": 0.7914, "step": 7573 }, { "epoch": 0.65, "learning_rate": 2.937448270311153e-06, "loss": 0.7758, "step": 7574 }, { "epoch": 0.65, "learning_rate": 2.936189510802807e-06, "loss": 0.8516, "step": 7575 }, { "epoch": 0.65, "learning_rate": 2.9349309089386035e-06, "loss": 0.7892, "step": 7576 }, { "epoch": 0.65, "learning_rate": 2.933672464814682e-06, "loss": 0.8297, "step": 7577 }, { "epoch": 0.65, "learning_rate": 2.9324141785271667e-06, "loss": 0.8056, "step": 7578 }, { "epoch": 0.65, "learning_rate": 2.9311560501721726e-06, "loss": 0.7925, "step": 7579 }, { "epoch": 0.65, "learning_rate": 2.9298980798458005e-06, "loss": 0.8031, "step": 7580 }, { "epoch": 0.65, "learning_rate": 2.9286402676441417e-06, "loss": 0.8663, "step": 7581 }, { "epoch": 0.65, "learning_rate": 2.9273826136632756e-06, "loss": 0.7798, "step": 7582 }, { "epoch": 0.65, "learning_rate": 2.9261251179992644e-06, "loss": 0.7999, "step": 7583 }, { "epoch": 0.65, "learning_rate": 2.9248677807481653e-06, "loss": 0.5263, "step": 7584 }, { "epoch": 0.65, "learning_rate": 2.9236106020060203e-06, "loss": 0.8099, "step": 7585 }, { "epoch": 0.65, "learning_rate": 2.9223535818688553e-06, "loss": 0.8365, "step": 7586 }, { "epoch": 0.65, "learning_rate": 2.9210967204326905e-06, "loss": 0.8111, "step": 7587 }, { "epoch": 0.65, "learning_rate": 2.9198400177935303e-06, "loss": 0.8067, "step": 7588 }, { "epoch": 0.65, "learning_rate": 2.9185834740473685e-06, "loss": 0.7955, "step": 7589 }, { "epoch": 0.65, "learning_rate": 2.917327089290186e-06, "loss": 0.8764, "step": 7590 }, { "epoch": 0.65, "learning_rate": 2.9160708636179522e-06, "loss": 0.8023, "step": 7591 }, { "epoch": 0.65, "learning_rate": 2.914814797126624e-06, "loss": 0.8695, "step": 7592 }, { "epoch": 0.65, "learning_rate": 2.9135588899121425e-06, "loss": 0.8191, "step": 7593 }, { "epoch": 0.65, "learning_rate": 2.912303142070445e-06, "loss": 0.8413, "step": 7594 }, { "epoch": 0.65, "learning_rate": 2.91104755369745e-06, "loss": 0.8417, "step": 7595 }, { "epoch": 0.65, "learning_rate": 2.909792124889064e-06, "loss": 0.8312, "step": 7596 }, { "epoch": 0.65, "learning_rate": 2.9085368557411857e-06, "loss": 0.5906, "step": 7597 }, { "epoch": 0.65, "learning_rate": 2.9072817463496977e-06, "loss": 0.7785, "step": 7598 }, { "epoch": 0.65, "learning_rate": 2.9060267968104716e-06, "loss": 0.8192, "step": 7599 }, { "epoch": 0.65, "learning_rate": 2.904772007219366e-06, "loss": 0.8358, "step": 7600 }, { "epoch": 0.65, "learning_rate": 2.903517377672227e-06, "loss": 0.762, "step": 7601 }, { "epoch": 0.65, "learning_rate": 2.9022629082648925e-06, "loss": 0.8103, "step": 7602 }, { "epoch": 0.65, "learning_rate": 2.901008599093183e-06, "loss": 0.8277, "step": 7603 }, { "epoch": 0.65, "learning_rate": 2.8997544502529095e-06, "loss": 0.8131, "step": 7604 }, { "epoch": 0.65, "learning_rate": 2.8985004618398706e-06, "loss": 0.7814, "step": 7605 }, { "epoch": 0.65, "learning_rate": 2.8972466339498497e-06, "loss": 0.8021, "step": 7606 }, { "epoch": 0.65, "learning_rate": 2.8959929666786233e-06, "loss": 0.8317, "step": 7607 }, { "epoch": 0.65, "learning_rate": 2.8947394601219524e-06, "loss": 0.8127, "step": 7608 }, { "epoch": 0.65, "learning_rate": 2.8934861143755843e-06, "loss": 0.812, "step": 7609 }, { "epoch": 0.65, "learning_rate": 2.892232929535258e-06, "loss": 0.7937, "step": 7610 }, { "epoch": 0.65, "learning_rate": 2.8909799056966975e-06, "loss": 0.7817, "step": 7611 }, { "epoch": 0.65, "learning_rate": 2.8897270429556146e-06, "loss": 0.829, "step": 7612 }, { "epoch": 0.65, "learning_rate": 2.8884743414077087e-06, "loss": 0.8452, "step": 7613 }, { "epoch": 0.65, "learning_rate": 2.887221801148666e-06, "loss": 0.8301, "step": 7614 }, { "epoch": 0.65, "learning_rate": 2.8859694222741653e-06, "loss": 0.8204, "step": 7615 }, { "epoch": 0.65, "learning_rate": 2.8847172048798667e-06, "loss": 0.8398, "step": 7616 }, { "epoch": 0.65, "learning_rate": 2.883465149061423e-06, "loss": 0.8272, "step": 7617 }, { "epoch": 0.65, "learning_rate": 2.8822132549144743e-06, "loss": 0.8015, "step": 7618 }, { "epoch": 0.65, "learning_rate": 2.8809615225346387e-06, "loss": 0.8373, "step": 7619 }, { "epoch": 0.65, "learning_rate": 2.8797099520175376e-06, "loss": 0.8103, "step": 7620 }, { "epoch": 0.65, "learning_rate": 2.8784585434587663e-06, "loss": 0.8432, "step": 7621 }, { "epoch": 0.65, "learning_rate": 2.87720729695392e-06, "loss": 0.8496, "step": 7622 }, { "epoch": 0.65, "learning_rate": 2.8759562125985707e-06, "loss": 0.8378, "step": 7623 }, { "epoch": 0.65, "learning_rate": 2.8747052904882845e-06, "loss": 0.7953, "step": 7624 }, { "epoch": 0.65, "learning_rate": 2.8734545307186114e-06, "loss": 0.8203, "step": 7625 }, { "epoch": 0.65, "learning_rate": 2.872203933385089e-06, "loss": 0.8045, "step": 7626 }, { "epoch": 0.65, "learning_rate": 2.8709534985832497e-06, "loss": 0.8429, "step": 7627 }, { "epoch": 0.65, "learning_rate": 2.8697032264086043e-06, "loss": 0.8093, "step": 7628 }, { "epoch": 0.65, "learning_rate": 2.8684531169566533e-06, "loss": 0.8069, "step": 7629 }, { "epoch": 0.65, "learning_rate": 2.8672031703228903e-06, "loss": 0.8481, "step": 7630 }, { "epoch": 0.65, "learning_rate": 2.8659533866027904e-06, "loss": 0.7913, "step": 7631 }, { "epoch": 0.65, "learning_rate": 2.864703765891819e-06, "loss": 0.8657, "step": 7632 }, { "epoch": 0.65, "learning_rate": 2.8634543082854273e-06, "loss": 0.8215, "step": 7633 }, { "epoch": 0.65, "learning_rate": 2.8622050138790535e-06, "loss": 0.8264, "step": 7634 }, { "epoch": 0.65, "learning_rate": 2.8609558827681293e-06, "loss": 0.8137, "step": 7635 }, { "epoch": 0.65, "learning_rate": 2.859706915048067e-06, "loss": 0.8226, "step": 7636 }, { "epoch": 0.65, "learning_rate": 2.8584581108142673e-06, "loss": 0.8049, "step": 7637 }, { "epoch": 0.65, "learning_rate": 2.8572094701621265e-06, "loss": 0.8307, "step": 7638 }, { "epoch": 0.65, "learning_rate": 2.8559609931870136e-06, "loss": 0.7913, "step": 7639 }, { "epoch": 0.65, "learning_rate": 2.8547126799842996e-06, "loss": 0.8579, "step": 7640 }, { "epoch": 0.65, "learning_rate": 2.8534645306493335e-06, "loss": 0.7856, "step": 7641 }, { "epoch": 0.65, "learning_rate": 2.852216545277456e-06, "loss": 0.7982, "step": 7642 }, { "epoch": 0.65, "learning_rate": 2.850968723963996e-06, "loss": 0.8344, "step": 7643 }, { "epoch": 0.65, "learning_rate": 2.8497210668042675e-06, "loss": 0.5177, "step": 7644 }, { "epoch": 0.65, "learning_rate": 2.8484735738935724e-06, "loss": 0.8123, "step": 7645 }, { "epoch": 0.65, "learning_rate": 2.847226245327201e-06, "loss": 0.8231, "step": 7646 }, { "epoch": 0.65, "learning_rate": 2.8459790812004272e-06, "loss": 0.7728, "step": 7647 }, { "epoch": 0.65, "learning_rate": 2.844732081608521e-06, "loss": 0.7894, "step": 7648 }, { "epoch": 0.65, "learning_rate": 2.84348524664673e-06, "loss": 0.7874, "step": 7649 }, { "epoch": 0.65, "learning_rate": 2.842238576410297e-06, "loss": 0.8102, "step": 7650 }, { "epoch": 0.65, "learning_rate": 2.84099207099445e-06, "loss": 0.8331, "step": 7651 }, { "epoch": 0.65, "learning_rate": 2.8397457304943964e-06, "loss": 0.8218, "step": 7652 }, { "epoch": 0.65, "learning_rate": 2.838499555005344e-06, "loss": 0.816, "step": 7653 }, { "epoch": 0.65, "learning_rate": 2.8372535446224785e-06, "loss": 0.8434, "step": 7654 }, { "epoch": 0.65, "learning_rate": 2.8360076994409803e-06, "loss": 0.8153, "step": 7655 }, { "epoch": 0.65, "learning_rate": 2.834762019556011e-06, "loss": 0.8216, "step": 7656 }, { "epoch": 0.65, "learning_rate": 2.833516505062721e-06, "loss": 0.8299, "step": 7657 }, { "epoch": 0.65, "learning_rate": 2.832271156056251e-06, "loss": 0.8334, "step": 7658 }, { "epoch": 0.65, "learning_rate": 2.8310259726317225e-06, "loss": 0.8186, "step": 7659 }, { "epoch": 0.65, "learning_rate": 2.8297809548842547e-06, "loss": 0.8073, "step": 7660 }, { "epoch": 0.65, "learning_rate": 2.828536102908946e-06, "loss": 0.8159, "step": 7661 }, { "epoch": 0.65, "learning_rate": 2.8272914168008813e-06, "loss": 0.8022, "step": 7662 }, { "epoch": 0.65, "learning_rate": 2.8260468966551414e-06, "loss": 0.8182, "step": 7663 }, { "epoch": 0.65, "learning_rate": 2.824802542566786e-06, "loss": 0.7862, "step": 7664 }, { "epoch": 0.65, "learning_rate": 2.823558354630865e-06, "loss": 0.8143, "step": 7665 }, { "epoch": 0.65, "learning_rate": 2.8223143329424164e-06, "loss": 0.7843, "step": 7666 }, { "epoch": 0.65, "learning_rate": 2.8210704775964615e-06, "loss": 0.7713, "step": 7667 }, { "epoch": 0.65, "learning_rate": 2.819826788688018e-06, "loss": 0.7562, "step": 7668 }, { "epoch": 0.65, "learning_rate": 2.8185832663120817e-06, "loss": 0.8452, "step": 7669 }, { "epoch": 0.65, "learning_rate": 2.817339910563638e-06, "loss": 0.7504, "step": 7670 }, { "epoch": 0.65, "learning_rate": 2.8160967215376655e-06, "loss": 0.8484, "step": 7671 }, { "epoch": 0.65, "learning_rate": 2.8148536993291174e-06, "loss": 0.7901, "step": 7672 }, { "epoch": 0.65, "learning_rate": 2.813610844032949e-06, "loss": 0.7999, "step": 7673 }, { "epoch": 0.65, "learning_rate": 2.8123681557440928e-06, "loss": 0.4917, "step": 7674 }, { "epoch": 0.65, "learning_rate": 2.81112563455747e-06, "loss": 0.8142, "step": 7675 }, { "epoch": 0.66, "learning_rate": 2.8098832805679944e-06, "loss": 0.808, "step": 7676 }, { "epoch": 0.66, "learning_rate": 2.808641093870561e-06, "loss": 0.77, "step": 7677 }, { "epoch": 0.66, "learning_rate": 2.807399074560054e-06, "loss": 0.7649, "step": 7678 }, { "epoch": 0.66, "learning_rate": 2.806157222731346e-06, "loss": 0.8379, "step": 7679 }, { "epoch": 0.66, "learning_rate": 2.804915538479293e-06, "loss": 0.8347, "step": 7680 }, { "epoch": 0.66, "learning_rate": 2.8036740218987456e-06, "loss": 0.7986, "step": 7681 }, { "epoch": 0.66, "learning_rate": 2.802432673084534e-06, "loss": 0.795, "step": 7682 }, { "epoch": 0.66, "learning_rate": 2.8011914921314786e-06, "loss": 0.8141, "step": 7683 }, { "epoch": 0.66, "learning_rate": 2.7999504791343906e-06, "loss": 0.7852, "step": 7684 }, { "epoch": 0.66, "learning_rate": 2.7987096341880582e-06, "loss": 0.7589, "step": 7685 }, { "epoch": 0.66, "learning_rate": 2.7974689573872692e-06, "loss": 0.7489, "step": 7686 }, { "epoch": 0.66, "learning_rate": 2.7962284488267887e-06, "loss": 0.7763, "step": 7687 }, { "epoch": 0.66, "learning_rate": 2.7949881086013764e-06, "loss": 0.7844, "step": 7688 }, { "epoch": 0.66, "learning_rate": 2.7937479368057743e-06, "loss": 0.809, "step": 7689 }, { "epoch": 0.66, "learning_rate": 2.7925079335347105e-06, "loss": 0.8453, "step": 7690 }, { "epoch": 0.66, "learning_rate": 2.791268098882909e-06, "loss": 0.8077, "step": 7691 }, { "epoch": 0.66, "learning_rate": 2.7900284329450653e-06, "loss": 0.8046, "step": 7692 }, { "epoch": 0.66, "learning_rate": 2.788788935815879e-06, "loss": 0.8159, "step": 7693 }, { "epoch": 0.66, "learning_rate": 2.787549607590027e-06, "loss": 0.7849, "step": 7694 }, { "epoch": 0.66, "learning_rate": 2.7863104483621723e-06, "loss": 0.822, "step": 7695 }, { "epoch": 0.66, "learning_rate": 2.785071458226972e-06, "loss": 0.8295, "step": 7696 }, { "epoch": 0.66, "learning_rate": 2.7838326372790656e-06, "loss": 0.8422, "step": 7697 }, { "epoch": 0.66, "learning_rate": 2.7825939856130794e-06, "loss": 0.769, "step": 7698 }, { "epoch": 0.66, "learning_rate": 2.7813555033236282e-06, "loss": 0.8081, "step": 7699 }, { "epoch": 0.66, "learning_rate": 2.7801171905053117e-06, "loss": 0.7973, "step": 7700 }, { "epoch": 0.66, "learning_rate": 2.778879047252721e-06, "loss": 0.8376, "step": 7701 }, { "epoch": 0.66, "learning_rate": 2.777641073660431e-06, "loss": 0.7652, "step": 7702 }, { "epoch": 0.66, "learning_rate": 2.7764032698230027e-06, "loss": 0.7438, "step": 7703 }, { "epoch": 0.66, "learning_rate": 2.7751656358349903e-06, "loss": 0.8195, "step": 7704 }, { "epoch": 0.66, "learning_rate": 2.773928171790924e-06, "loss": 0.7651, "step": 7705 }, { "epoch": 0.66, "learning_rate": 2.7726908777853317e-06, "loss": 0.824, "step": 7706 }, { "epoch": 0.66, "learning_rate": 2.771453753912724e-06, "loss": 0.5688, "step": 7707 }, { "epoch": 0.66, "learning_rate": 2.7702168002675948e-06, "loss": 0.8317, "step": 7708 }, { "epoch": 0.66, "learning_rate": 2.768980016944434e-06, "loss": 0.8138, "step": 7709 }, { "epoch": 0.66, "learning_rate": 2.767743404037711e-06, "loss": 0.8323, "step": 7710 }, { "epoch": 0.66, "learning_rate": 2.766506961641885e-06, "loss": 0.8389, "step": 7711 }, { "epoch": 0.66, "learning_rate": 2.7652706898514e-06, "loss": 0.7952, "step": 7712 }, { "epoch": 0.66, "learning_rate": 2.764034588760688e-06, "loss": 0.7962, "step": 7713 }, { "epoch": 0.66, "learning_rate": 2.762798658464172e-06, "loss": 0.7946, "step": 7714 }, { "epoch": 0.66, "learning_rate": 2.761562899056257e-06, "loss": 0.8254, "step": 7715 }, { "epoch": 0.66, "learning_rate": 2.7603273106313346e-06, "loss": 0.8055, "step": 7716 }, { "epoch": 0.66, "learning_rate": 2.759091893283791e-06, "loss": 0.8213, "step": 7717 }, { "epoch": 0.66, "learning_rate": 2.7578566471079843e-06, "loss": 0.7813, "step": 7718 }, { "epoch": 0.66, "learning_rate": 2.756621572198276e-06, "loss": 0.7807, "step": 7719 }, { "epoch": 0.66, "learning_rate": 2.7553866686490054e-06, "loss": 0.8125, "step": 7720 }, { "epoch": 0.66, "learning_rate": 2.754151936554499e-06, "loss": 0.809, "step": 7721 }, { "epoch": 0.66, "learning_rate": 2.7529173760090737e-06, "loss": 0.869, "step": 7722 }, { "epoch": 0.66, "learning_rate": 2.7516829871070295e-06, "loss": 0.8196, "step": 7723 }, { "epoch": 0.66, "learning_rate": 2.7504487699426606e-06, "loss": 0.8133, "step": 7724 }, { "epoch": 0.66, "learning_rate": 2.7492147246102363e-06, "loss": 0.8184, "step": 7725 }, { "epoch": 0.66, "learning_rate": 2.7479808512040196e-06, "loss": 0.834, "step": 7726 }, { "epoch": 0.66, "learning_rate": 2.746747149818264e-06, "loss": 0.8046, "step": 7727 }, { "epoch": 0.66, "learning_rate": 2.745513620547201e-06, "loss": 0.8289, "step": 7728 }, { "epoch": 0.66, "learning_rate": 2.744280263485058e-06, "loss": 0.7931, "step": 7729 }, { "epoch": 0.66, "learning_rate": 2.7430470787260433e-06, "loss": 0.8205, "step": 7730 }, { "epoch": 0.66, "learning_rate": 2.7418140663643535e-06, "loss": 0.8181, "step": 7731 }, { "epoch": 0.66, "learning_rate": 2.740581226494172e-06, "loss": 0.808, "step": 7732 }, { "epoch": 0.66, "learning_rate": 2.739348559209668e-06, "loss": 0.7868, "step": 7733 }, { "epoch": 0.66, "learning_rate": 2.7381160646050025e-06, "loss": 0.8367, "step": 7734 }, { "epoch": 0.66, "learning_rate": 2.7368837427743174e-06, "loss": 0.797, "step": 7735 }, { "epoch": 0.66, "learning_rate": 2.7356515938117413e-06, "loss": 0.8298, "step": 7736 }, { "epoch": 0.66, "learning_rate": 2.7344196178113978e-06, "loss": 0.8003, "step": 7737 }, { "epoch": 0.66, "learning_rate": 2.733187814867384e-06, "loss": 0.8143, "step": 7738 }, { "epoch": 0.66, "learning_rate": 2.7319561850737974e-06, "loss": 0.8547, "step": 7739 }, { "epoch": 0.66, "learning_rate": 2.730724728524713e-06, "loss": 0.797, "step": 7740 }, { "epoch": 0.66, "learning_rate": 2.7294934453141943e-06, "loss": 0.7768, "step": 7741 }, { "epoch": 0.66, "learning_rate": 2.728262335536297e-06, "loss": 0.7888, "step": 7742 }, { "epoch": 0.66, "learning_rate": 2.7270313992850566e-06, "loss": 0.8232, "step": 7743 }, { "epoch": 0.66, "learning_rate": 2.7258006366544985e-06, "loss": 0.8216, "step": 7744 }, { "epoch": 0.66, "learning_rate": 2.7245700477386356e-06, "loss": 0.7743, "step": 7745 }, { "epoch": 0.66, "learning_rate": 2.7233396326314627e-06, "loss": 0.8522, "step": 7746 }, { "epoch": 0.66, "learning_rate": 2.7221093914269703e-06, "loss": 0.8077, "step": 7747 }, { "epoch": 0.66, "learning_rate": 2.7208793242191275e-06, "loss": 0.8962, "step": 7748 }, { "epoch": 0.66, "learning_rate": 2.7196494311018913e-06, "loss": 0.7875, "step": 7749 }, { "epoch": 0.66, "learning_rate": 2.718419712169213e-06, "loss": 0.8055, "step": 7750 }, { "epoch": 0.66, "learning_rate": 2.717190167515017e-06, "loss": 0.8056, "step": 7751 }, { "epoch": 0.66, "learning_rate": 2.715960797233228e-06, "loss": 0.7987, "step": 7752 }, { "epoch": 0.66, "learning_rate": 2.714731601417749e-06, "loss": 0.7835, "step": 7753 }, { "epoch": 0.66, "learning_rate": 2.7135025801624702e-06, "loss": 0.8641, "step": 7754 }, { "epoch": 0.66, "learning_rate": 2.7122737335612747e-06, "loss": 0.7985, "step": 7755 }, { "epoch": 0.66, "learning_rate": 2.711045061708024e-06, "loss": 0.8163, "step": 7756 }, { "epoch": 0.66, "learning_rate": 2.7098165646965756e-06, "loss": 0.8227, "step": 7757 }, { "epoch": 0.66, "learning_rate": 2.7085882426207635e-06, "loss": 0.7973, "step": 7758 }, { "epoch": 0.66, "learning_rate": 2.707360095574411e-06, "loss": 0.8325, "step": 7759 }, { "epoch": 0.66, "learning_rate": 2.706132123651336e-06, "loss": 0.8396, "step": 7760 }, { "epoch": 0.66, "learning_rate": 2.7049043269453323e-06, "loss": 0.8364, "step": 7761 }, { "epoch": 0.66, "learning_rate": 2.7036767055501887e-06, "loss": 0.8469, "step": 7762 }, { "epoch": 0.66, "learning_rate": 2.7024492595596764e-06, "loss": 0.8078, "step": 7763 }, { "epoch": 0.66, "learning_rate": 2.7012219890675523e-06, "loss": 0.8569, "step": 7764 }, { "epoch": 0.66, "learning_rate": 2.6999948941675625e-06, "loss": 0.784, "step": 7765 }, { "epoch": 0.66, "learning_rate": 2.6987679749534368e-06, "loss": 0.8319, "step": 7766 }, { "epoch": 0.66, "learning_rate": 2.6975412315188964e-06, "loss": 0.7844, "step": 7767 }, { "epoch": 0.66, "learning_rate": 2.6963146639576443e-06, "loss": 0.776, "step": 7768 }, { "epoch": 0.66, "learning_rate": 2.695088272363371e-06, "loss": 0.4706, "step": 7769 }, { "epoch": 0.66, "learning_rate": 2.6938620568297587e-06, "loss": 0.8143, "step": 7770 }, { "epoch": 0.66, "learning_rate": 2.692636017450465e-06, "loss": 0.7622, "step": 7771 }, { "epoch": 0.66, "learning_rate": 2.6914101543191463e-06, "loss": 0.8311, "step": 7772 }, { "epoch": 0.66, "learning_rate": 2.690184467529439e-06, "loss": 0.8275, "step": 7773 }, { "epoch": 0.66, "learning_rate": 2.6889589571749645e-06, "loss": 0.7498, "step": 7774 }, { "epoch": 0.66, "learning_rate": 2.6877336233493378e-06, "loss": 0.8198, "step": 7775 }, { "epoch": 0.66, "learning_rate": 2.686508466146153e-06, "loss": 0.8046, "step": 7776 }, { "epoch": 0.66, "learning_rate": 2.685283485658995e-06, "loss": 0.7962, "step": 7777 }, { "epoch": 0.66, "learning_rate": 2.6840586819814336e-06, "loss": 0.8149, "step": 7778 }, { "epoch": 0.66, "learning_rate": 2.6828340552070233e-06, "loss": 0.7473, "step": 7779 }, { "epoch": 0.66, "learning_rate": 2.6816096054293105e-06, "loss": 0.8083, "step": 7780 }, { "epoch": 0.66, "learning_rate": 2.6803853327418233e-06, "loss": 0.7883, "step": 7781 }, { "epoch": 0.66, "learning_rate": 2.679161237238076e-06, "loss": 0.5406, "step": 7782 }, { "epoch": 0.66, "learning_rate": 2.6779373190115774e-06, "loss": 0.8427, "step": 7783 }, { "epoch": 0.66, "learning_rate": 2.6767135781558075e-06, "loss": 0.4851, "step": 7784 }, { "epoch": 0.66, "learning_rate": 2.6754900147642486e-06, "loss": 0.7664, "step": 7785 }, { "epoch": 0.66, "learning_rate": 2.6742666289303603e-06, "loss": 0.8221, "step": 7786 }, { "epoch": 0.66, "learning_rate": 2.6730434207475888e-06, "loss": 0.7842, "step": 7787 }, { "epoch": 0.66, "learning_rate": 2.6718203903093733e-06, "loss": 0.7965, "step": 7788 }, { "epoch": 0.66, "learning_rate": 2.67059753770913e-06, "loss": 0.8441, "step": 7789 }, { "epoch": 0.66, "learning_rate": 2.6693748630402732e-06, "loss": 0.822, "step": 7790 }, { "epoch": 0.66, "learning_rate": 2.6681523663961907e-06, "loss": 0.8105, "step": 7791 }, { "epoch": 0.66, "learning_rate": 2.666930047870263e-06, "loss": 0.838, "step": 7792 }, { "epoch": 0.66, "learning_rate": 2.6657079075558613e-06, "loss": 0.8019, "step": 7793 }, { "epoch": 0.67, "learning_rate": 2.6644859455463335e-06, "loss": 0.8047, "step": 7794 }, { "epoch": 0.67, "learning_rate": 2.6632641619350243e-06, "loss": 0.7964, "step": 7795 }, { "epoch": 0.67, "learning_rate": 2.6620425568152577e-06, "loss": 0.7591, "step": 7796 }, { "epoch": 0.67, "learning_rate": 2.6608211302803457e-06, "loss": 0.7838, "step": 7797 }, { "epoch": 0.67, "learning_rate": 2.6595998824235856e-06, "loss": 0.7765, "step": 7798 }, { "epoch": 0.67, "learning_rate": 2.6583788133382625e-06, "loss": 0.8172, "step": 7799 }, { "epoch": 0.67, "learning_rate": 2.6571579231176513e-06, "loss": 0.8295, "step": 7800 }, { "epoch": 0.67, "learning_rate": 2.6559372118550063e-06, "loss": 0.8127, "step": 7801 }, { "epoch": 0.67, "learning_rate": 2.654716679643571e-06, "loss": 0.8789, "step": 7802 }, { "epoch": 0.67, "learning_rate": 2.6534963265765814e-06, "loss": 0.543, "step": 7803 }, { "epoch": 0.67, "learning_rate": 2.6522761527472464e-06, "loss": 0.7733, "step": 7804 }, { "epoch": 0.67, "learning_rate": 2.651056158248774e-06, "loss": 0.7891, "step": 7805 }, { "epoch": 0.67, "learning_rate": 2.649836343174352e-06, "loss": 0.836, "step": 7806 }, { "epoch": 0.67, "learning_rate": 2.648616707617154e-06, "loss": 0.7741, "step": 7807 }, { "epoch": 0.67, "learning_rate": 2.6473972516703463e-06, "loss": 0.8312, "step": 7808 }, { "epoch": 0.67, "learning_rate": 2.646177975427073e-06, "loss": 0.8438, "step": 7809 }, { "epoch": 0.67, "learning_rate": 2.6449588789804713e-06, "loss": 0.7612, "step": 7810 }, { "epoch": 0.67, "learning_rate": 2.643739962423661e-06, "loss": 0.7886, "step": 7811 }, { "epoch": 0.67, "learning_rate": 2.6425212258497463e-06, "loss": 0.8094, "step": 7812 }, { "epoch": 0.67, "learning_rate": 2.6413026693518247e-06, "loss": 0.8033, "step": 7813 }, { "epoch": 0.67, "learning_rate": 2.6400842930229737e-06, "loss": 0.7933, "step": 7814 }, { "epoch": 0.67, "learning_rate": 2.6388660969562567e-06, "loss": 0.8475, "step": 7815 }, { "epoch": 0.67, "learning_rate": 2.6376480812447305e-06, "loss": 0.7637, "step": 7816 }, { "epoch": 0.67, "learning_rate": 2.6364302459814293e-06, "loss": 0.7773, "step": 7817 }, { "epoch": 0.67, "learning_rate": 2.6352125912593795e-06, "loss": 0.7867, "step": 7818 }, { "epoch": 0.67, "learning_rate": 2.633995117171591e-06, "loss": 0.8113, "step": 7819 }, { "epoch": 0.67, "learning_rate": 2.632777823811058e-06, "loss": 0.7929, "step": 7820 }, { "epoch": 0.67, "learning_rate": 2.6315607112707687e-06, "loss": 0.8082, "step": 7821 }, { "epoch": 0.67, "learning_rate": 2.630343779643688e-06, "loss": 0.8068, "step": 7822 }, { "epoch": 0.67, "learning_rate": 2.6291270290227734e-06, "loss": 0.7855, "step": 7823 }, { "epoch": 0.67, "learning_rate": 2.627910459500965e-06, "loss": 0.8333, "step": 7824 }, { "epoch": 0.67, "learning_rate": 2.6266940711711896e-06, "loss": 0.8205, "step": 7825 }, { "epoch": 0.67, "learning_rate": 2.625477864126364e-06, "loss": 0.792, "step": 7826 }, { "epoch": 0.67, "learning_rate": 2.6242618384593865e-06, "loss": 0.8099, "step": 7827 }, { "epoch": 0.67, "learning_rate": 2.6230459942631413e-06, "loss": 0.7676, "step": 7828 }, { "epoch": 0.67, "learning_rate": 2.621830331630505e-06, "loss": 0.8216, "step": 7829 }, { "epoch": 0.67, "learning_rate": 2.6206148506543328e-06, "loss": 0.7936, "step": 7830 }, { "epoch": 0.67, "learning_rate": 2.6193995514274705e-06, "loss": 0.8382, "step": 7831 }, { "epoch": 0.67, "learning_rate": 2.6181844340427466e-06, "loss": 0.77, "step": 7832 }, { "epoch": 0.67, "learning_rate": 2.6169694985929804e-06, "loss": 0.8293, "step": 7833 }, { "epoch": 0.67, "learning_rate": 2.6157547451709745e-06, "loss": 0.8538, "step": 7834 }, { "epoch": 0.67, "learning_rate": 2.6145401738695154e-06, "loss": 0.7542, "step": 7835 }, { "epoch": 0.67, "learning_rate": 2.613325784781381e-06, "loss": 0.7839, "step": 7836 }, { "epoch": 0.67, "learning_rate": 2.6121115779993315e-06, "loss": 0.8407, "step": 7837 }, { "epoch": 0.67, "learning_rate": 2.610897553616114e-06, "loss": 0.8338, "step": 7838 }, { "epoch": 0.67, "learning_rate": 2.609683711724461e-06, "loss": 0.7974, "step": 7839 }, { "epoch": 0.67, "learning_rate": 2.6084700524170908e-06, "loss": 0.7707, "step": 7840 }, { "epoch": 0.67, "learning_rate": 2.607256575786712e-06, "loss": 0.7996, "step": 7841 }, { "epoch": 0.67, "learning_rate": 2.6060432819260153e-06, "loss": 0.8632, "step": 7842 }, { "epoch": 0.67, "learning_rate": 2.6048301709276756e-06, "loss": 0.8064, "step": 7843 }, { "epoch": 0.67, "learning_rate": 2.6036172428843587e-06, "loss": 0.7897, "step": 7844 }, { "epoch": 0.67, "learning_rate": 2.6024044978887107e-06, "loss": 0.7724, "step": 7845 }, { "epoch": 0.67, "learning_rate": 2.601191936033372e-06, "loss": 0.7881, "step": 7846 }, { "epoch": 0.67, "learning_rate": 2.599979557410962e-06, "loss": 0.8253, "step": 7847 }, { "epoch": 0.67, "learning_rate": 2.5987673621140864e-06, "loss": 0.7804, "step": 7848 }, { "epoch": 0.67, "learning_rate": 2.5975553502353413e-06, "loss": 0.8405, "step": 7849 }, { "epoch": 0.67, "learning_rate": 2.596343521867306e-06, "loss": 0.8128, "step": 7850 }, { "epoch": 0.67, "learning_rate": 2.595131877102545e-06, "loss": 0.8312, "step": 7851 }, { "epoch": 0.67, "learning_rate": 2.5939204160336105e-06, "loss": 0.7991, "step": 7852 }, { "epoch": 0.67, "learning_rate": 2.592709138753037e-06, "loss": 0.8243, "step": 7853 }, { "epoch": 0.67, "learning_rate": 2.591498045353352e-06, "loss": 0.5538, "step": 7854 }, { "epoch": 0.67, "learning_rate": 2.5902871359270645e-06, "loss": 0.8358, "step": 7855 }, { "epoch": 0.67, "learning_rate": 2.5890764105666677e-06, "loss": 0.8195, "step": 7856 }, { "epoch": 0.67, "learning_rate": 2.5878658693646443e-06, "loss": 0.8108, "step": 7857 }, { "epoch": 0.67, "learning_rate": 2.586655512413458e-06, "loss": 0.7972, "step": 7858 }, { "epoch": 0.67, "learning_rate": 2.5854453398055674e-06, "loss": 0.8004, "step": 7859 }, { "epoch": 0.67, "learning_rate": 2.5842353516334085e-06, "loss": 0.8226, "step": 7860 }, { "epoch": 0.67, "learning_rate": 2.583025547989405e-06, "loss": 0.8147, "step": 7861 }, { "epoch": 0.67, "learning_rate": 2.581815928965971e-06, "loss": 0.7903, "step": 7862 }, { "epoch": 0.67, "learning_rate": 2.5806064946555022e-06, "loss": 0.7597, "step": 7863 }, { "epoch": 0.67, "learning_rate": 2.57939724515038e-06, "loss": 0.7704, "step": 7864 }, { "epoch": 0.67, "learning_rate": 2.5781881805429736e-06, "loss": 0.8247, "step": 7865 }, { "epoch": 0.67, "learning_rate": 2.5769793009256363e-06, "loss": 0.7688, "step": 7866 }, { "epoch": 0.67, "learning_rate": 2.57577060639071e-06, "loss": 0.7973, "step": 7867 }, { "epoch": 0.67, "learning_rate": 2.5745620970305185e-06, "loss": 0.793, "step": 7868 }, { "epoch": 0.67, "learning_rate": 2.573353772937378e-06, "loss": 0.8067, "step": 7869 }, { "epoch": 0.67, "learning_rate": 2.572145634203584e-06, "loss": 0.8342, "step": 7870 }, { "epoch": 0.67, "learning_rate": 2.570937680921419e-06, "loss": 0.8065, "step": 7871 }, { "epoch": 0.67, "learning_rate": 2.5697299131831545e-06, "loss": 0.8359, "step": 7872 }, { "epoch": 0.67, "learning_rate": 2.5685223310810427e-06, "loss": 0.5161, "step": 7873 }, { "epoch": 0.67, "learning_rate": 2.5673149347073277e-06, "loss": 0.8003, "step": 7874 }, { "epoch": 0.67, "learning_rate": 2.5661077241542366e-06, "loss": 0.7589, "step": 7875 }, { "epoch": 0.67, "learning_rate": 2.564900699513981e-06, "loss": 0.8421, "step": 7876 }, { "epoch": 0.67, "learning_rate": 2.5636938608787597e-06, "loss": 0.7491, "step": 7877 }, { "epoch": 0.67, "learning_rate": 2.562487208340755e-06, "loss": 0.8204, "step": 7878 }, { "epoch": 0.67, "learning_rate": 2.561280741992141e-06, "loss": 0.7909, "step": 7879 }, { "epoch": 0.67, "learning_rate": 2.5600744619250716e-06, "loss": 0.7641, "step": 7880 }, { "epoch": 0.67, "learning_rate": 2.558868368231686e-06, "loss": 0.7679, "step": 7881 }, { "epoch": 0.67, "learning_rate": 2.5576624610041168e-06, "loss": 0.8371, "step": 7882 }, { "epoch": 0.67, "learning_rate": 2.556456740334474e-06, "loss": 0.8176, "step": 7883 }, { "epoch": 0.67, "learning_rate": 2.5552512063148584e-06, "loss": 0.8264, "step": 7884 }, { "epoch": 0.67, "learning_rate": 2.554045859037353e-06, "loss": 0.7811, "step": 7885 }, { "epoch": 0.67, "learning_rate": 2.552840698594026e-06, "loss": 0.7815, "step": 7886 }, { "epoch": 0.67, "learning_rate": 2.5516357250769387e-06, "loss": 0.756, "step": 7887 }, { "epoch": 0.67, "learning_rate": 2.55043093857813e-06, "loss": 0.8191, "step": 7888 }, { "epoch": 0.67, "learning_rate": 2.549226339189627e-06, "loss": 0.8404, "step": 7889 }, { "epoch": 0.67, "learning_rate": 2.548021927003448e-06, "loss": 0.8316, "step": 7890 }, { "epoch": 0.67, "learning_rate": 2.546817702111584e-06, "loss": 0.8025, "step": 7891 }, { "epoch": 0.67, "learning_rate": 2.545613664606025e-06, "loss": 0.7578, "step": 7892 }, { "epoch": 0.67, "learning_rate": 2.544409814578741e-06, "loss": 0.82, "step": 7893 }, { "epoch": 0.67, "learning_rate": 2.543206152121685e-06, "loss": 0.5095, "step": 7894 }, { "epoch": 0.67, "learning_rate": 2.5420026773268025e-06, "loss": 0.7855, "step": 7895 }, { "epoch": 0.67, "learning_rate": 2.5407993902860206e-06, "loss": 0.8234, "step": 7896 }, { "epoch": 0.67, "learning_rate": 2.5395962910912504e-06, "loss": 0.8313, "step": 7897 }, { "epoch": 0.67, "learning_rate": 2.5383933798343913e-06, "loss": 0.7937, "step": 7898 }, { "epoch": 0.67, "learning_rate": 2.5371906566073256e-06, "loss": 0.83, "step": 7899 }, { "epoch": 0.67, "learning_rate": 2.535988121501928e-06, "loss": 0.7997, "step": 7900 }, { "epoch": 0.67, "learning_rate": 2.5347857746100498e-06, "loss": 0.506, "step": 7901 }, { "epoch": 0.67, "learning_rate": 2.5335836160235357e-06, "loss": 0.7749, "step": 7902 }, { "epoch": 0.67, "learning_rate": 2.5323816458342126e-06, "loss": 0.8081, "step": 7903 }, { "epoch": 0.67, "learning_rate": 2.5311798641338873e-06, "loss": 0.8064, "step": 7904 }, { "epoch": 0.67, "learning_rate": 2.5299782710143644e-06, "loss": 0.7889, "step": 7905 }, { "epoch": 0.67, "learning_rate": 2.5287768665674227e-06, "loss": 0.8559, "step": 7906 }, { "epoch": 0.67, "learning_rate": 2.5275756508848357e-06, "loss": 0.8355, "step": 7907 }, { "epoch": 0.67, "learning_rate": 2.526374624058357e-06, "loss": 0.821, "step": 7908 }, { "epoch": 0.67, "learning_rate": 2.525173786179726e-06, "loss": 0.7928, "step": 7909 }, { "epoch": 0.67, "learning_rate": 2.5239731373406696e-06, "loss": 0.7618, "step": 7910 }, { "epoch": 0.68, "learning_rate": 2.522772677632896e-06, "loss": 0.8017, "step": 7911 }, { "epoch": 0.68, "learning_rate": 2.521572407148107e-06, "loss": 0.8521, "step": 7912 }, { "epoch": 0.68, "learning_rate": 2.5203723259779846e-06, "loss": 0.5401, "step": 7913 }, { "epoch": 0.68, "learning_rate": 2.5191724342141927e-06, "loss": 0.8165, "step": 7914 }, { "epoch": 0.68, "learning_rate": 2.5179727319483903e-06, "loss": 0.8252, "step": 7915 }, { "epoch": 0.68, "learning_rate": 2.516773219272215e-06, "loss": 0.7781, "step": 7916 }, { "epoch": 0.68, "learning_rate": 2.51557389627729e-06, "loss": 0.8157, "step": 7917 }, { "epoch": 0.68, "learning_rate": 2.5143747630552274e-06, "loss": 0.8262, "step": 7918 }, { "epoch": 0.68, "learning_rate": 2.5131758196976197e-06, "loss": 0.8402, "step": 7919 }, { "epoch": 0.68, "learning_rate": 2.511977066296053e-06, "loss": 0.799, "step": 7920 }, { "epoch": 0.68, "learning_rate": 2.5107785029420915e-06, "loss": 0.7738, "step": 7921 }, { "epoch": 0.68, "learning_rate": 2.5095801297272852e-06, "loss": 0.844, "step": 7922 }, { "epoch": 0.68, "learning_rate": 2.508381946743179e-06, "loss": 0.8076, "step": 7923 }, { "epoch": 0.68, "learning_rate": 2.507183954081287e-06, "loss": 0.8181, "step": 7924 }, { "epoch": 0.68, "learning_rate": 2.505986151833124e-06, "loss": 0.7957, "step": 7925 }, { "epoch": 0.68, "learning_rate": 2.504788540090183e-06, "loss": 0.766, "step": 7926 }, { "epoch": 0.68, "learning_rate": 2.50359111894394e-06, "loss": 0.7671, "step": 7927 }, { "epoch": 0.68, "learning_rate": 2.5023938884858655e-06, "loss": 0.767, "step": 7928 }, { "epoch": 0.68, "learning_rate": 2.5011968488074073e-06, "loss": 0.7844, "step": 7929 }, { "epoch": 0.68, "learning_rate": 2.5000000000000015e-06, "loss": 0.8387, "step": 7930 }, { "epoch": 0.68, "learning_rate": 2.498803342155068e-06, "loss": 0.7998, "step": 7931 }, { "epoch": 0.68, "learning_rate": 2.497606875364014e-06, "loss": 0.7885, "step": 7932 }, { "epoch": 0.68, "learning_rate": 2.4964105997182343e-06, "loss": 0.7761, "step": 7933 }, { "epoch": 0.68, "learning_rate": 2.495214515309102e-06, "loss": 0.8325, "step": 7934 }, { "epoch": 0.68, "learning_rate": 2.4940186222279838e-06, "loss": 0.8504, "step": 7935 }, { "epoch": 0.68, "learning_rate": 2.492822920566229e-06, "loss": 0.8401, "step": 7936 }, { "epoch": 0.68, "learning_rate": 2.491627410415165e-06, "loss": 0.7912, "step": 7937 }, { "epoch": 0.68, "learning_rate": 2.490432091866116e-06, "loss": 0.8012, "step": 7938 }, { "epoch": 0.68, "learning_rate": 2.4892369650103837e-06, "loss": 0.8043, "step": 7939 }, { "epoch": 0.68, "learning_rate": 2.488042029939261e-06, "loss": 0.7992, "step": 7940 }, { "epoch": 0.68, "learning_rate": 2.4868472867440213e-06, "loss": 0.8027, "step": 7941 }, { "epoch": 0.68, "learning_rate": 2.4856527355159226e-06, "loss": 0.7856, "step": 7942 }, { "epoch": 0.68, "learning_rate": 2.484458376346217e-06, "loss": 0.7707, "step": 7943 }, { "epoch": 0.68, "learning_rate": 2.4832642093261285e-06, "loss": 0.8191, "step": 7944 }, { "epoch": 0.68, "learning_rate": 2.4820702345468777e-06, "loss": 0.8865, "step": 7945 }, { "epoch": 0.68, "learning_rate": 2.4808764520996652e-06, "loss": 0.8297, "step": 7946 }, { "epoch": 0.68, "learning_rate": 2.479682862075677e-06, "loss": 0.7798, "step": 7947 }, { "epoch": 0.68, "learning_rate": 2.4784894645660886e-06, "loss": 0.7964, "step": 7948 }, { "epoch": 0.68, "learning_rate": 2.4772962596620546e-06, "loss": 0.8449, "step": 7949 }, { "epoch": 0.68, "learning_rate": 2.47610324745472e-06, "loss": 0.8278, "step": 7950 }, { "epoch": 0.68, "learning_rate": 2.474910428035212e-06, "loss": 0.834, "step": 7951 }, { "epoch": 0.68, "learning_rate": 2.4737178014946416e-06, "loss": 0.7738, "step": 7952 }, { "epoch": 0.68, "learning_rate": 2.4725253679241128e-06, "loss": 0.8368, "step": 7953 }, { "epoch": 0.68, "learning_rate": 2.471333127414707e-06, "loss": 0.7792, "step": 7954 }, { "epoch": 0.68, "learning_rate": 2.470141080057491e-06, "loss": 0.7827, "step": 7955 }, { "epoch": 0.68, "learning_rate": 2.4689492259435256e-06, "loss": 0.8477, "step": 7956 }, { "epoch": 0.68, "learning_rate": 2.467757565163843e-06, "loss": 0.811, "step": 7957 }, { "epoch": 0.68, "learning_rate": 2.466566097809474e-06, "loss": 0.7044, "step": 7958 }, { "epoch": 0.68, "learning_rate": 2.4653748239714265e-06, "loss": 0.8002, "step": 7959 }, { "epoch": 0.68, "learning_rate": 2.4641837437406944e-06, "loss": 0.7996, "step": 7960 }, { "epoch": 0.68, "learning_rate": 2.4629928572082623e-06, "loss": 0.8342, "step": 7961 }, { "epoch": 0.68, "learning_rate": 2.4618021644650936e-06, "loss": 0.7823, "step": 7962 }, { "epoch": 0.68, "learning_rate": 2.4606116656021397e-06, "loss": 0.7807, "step": 7963 }, { "epoch": 0.68, "learning_rate": 2.4594213607103363e-06, "loss": 0.7916, "step": 7964 }, { "epoch": 0.68, "learning_rate": 2.4582312498806036e-06, "loss": 0.5604, "step": 7965 }, { "epoch": 0.68, "learning_rate": 2.4570413332038523e-06, "loss": 0.816, "step": 7966 }, { "epoch": 0.68, "learning_rate": 2.4558516107709716e-06, "loss": 0.8355, "step": 7967 }, { "epoch": 0.68, "learning_rate": 2.4546620826728367e-06, "loss": 0.5264, "step": 7968 }, { "epoch": 0.68, "learning_rate": 2.453472749000315e-06, "loss": 0.7937, "step": 7969 }, { "epoch": 0.68, "learning_rate": 2.452283609844247e-06, "loss": 0.8003, "step": 7970 }, { "epoch": 0.68, "learning_rate": 2.4510946652954702e-06, "loss": 0.8196, "step": 7971 }, { "epoch": 0.68, "learning_rate": 2.4499059154447986e-06, "loss": 0.737, "step": 7972 }, { "epoch": 0.68, "learning_rate": 2.4487173603830383e-06, "loss": 0.7989, "step": 7973 }, { "epoch": 0.68, "learning_rate": 2.4475290002009755e-06, "loss": 0.7836, "step": 7974 }, { "epoch": 0.68, "learning_rate": 2.446340834989381e-06, "loss": 0.8063, "step": 7975 }, { "epoch": 0.68, "learning_rate": 2.4451528648390183e-06, "loss": 0.8196, "step": 7976 }, { "epoch": 0.68, "learning_rate": 2.443965089840623e-06, "loss": 0.7603, "step": 7977 }, { "epoch": 0.68, "learning_rate": 2.4427775100849295e-06, "loss": 0.8857, "step": 7978 }, { "epoch": 0.68, "learning_rate": 2.441590125662649e-06, "loss": 0.8556, "step": 7979 }, { "epoch": 0.68, "learning_rate": 2.440402936664477e-06, "loss": 0.8176, "step": 7980 }, { "epoch": 0.68, "learning_rate": 2.4392159431811018e-06, "loss": 0.7941, "step": 7981 }, { "epoch": 0.68, "learning_rate": 2.4380291453031895e-06, "loss": 0.8171, "step": 7982 }, { "epoch": 0.68, "learning_rate": 2.4368425431213938e-06, "loss": 0.82, "step": 7983 }, { "epoch": 0.68, "learning_rate": 2.435656136726353e-06, "loss": 0.7789, "step": 7984 }, { "epoch": 0.68, "learning_rate": 2.4344699262086896e-06, "loss": 0.7879, "step": 7985 }, { "epoch": 0.68, "learning_rate": 2.433283911659015e-06, "loss": 0.8083, "step": 7986 }, { "epoch": 0.68, "learning_rate": 2.4320980931679223e-06, "loss": 0.8091, "step": 7987 }, { "epoch": 0.68, "learning_rate": 2.4309124708259867e-06, "loss": 0.7694, "step": 7988 }, { "epoch": 0.68, "learning_rate": 2.429727044723779e-06, "loss": 0.7683, "step": 7989 }, { "epoch": 0.68, "learning_rate": 2.42854181495184e-06, "loss": 0.8056, "step": 7990 }, { "epoch": 0.68, "learning_rate": 2.4273567816007095e-06, "loss": 0.7844, "step": 7991 }, { "epoch": 0.68, "learning_rate": 2.4261719447609033e-06, "loss": 0.7943, "step": 7992 }, { "epoch": 0.68, "learning_rate": 2.4249873045229244e-06, "loss": 0.8142, "step": 7993 }, { "epoch": 0.68, "learning_rate": 2.423802860977264e-06, "loss": 0.7656, "step": 7994 }, { "epoch": 0.68, "learning_rate": 2.4226186142143953e-06, "loss": 0.8184, "step": 7995 }, { "epoch": 0.68, "learning_rate": 2.421434564324777e-06, "loss": 0.8077, "step": 7996 }, { "epoch": 0.68, "learning_rate": 2.420250711398851e-06, "loss": 0.7996, "step": 7997 }, { "epoch": 0.68, "learning_rate": 2.419067055527046e-06, "loss": 0.7531, "step": 7998 }, { "epoch": 0.68, "learning_rate": 2.417883596799778e-06, "loss": 0.8274, "step": 7999 }, { "epoch": 0.68, "learning_rate": 2.4167003353074435e-06, "loss": 0.5071, "step": 8000 }, { "epoch": 0.68, "learning_rate": 2.415517271140426e-06, "loss": 0.812, "step": 8001 }, { "epoch": 0.68, "learning_rate": 2.414334404389097e-06, "loss": 0.4913, "step": 8002 }, { "epoch": 0.68, "learning_rate": 2.4131517351438033e-06, "loss": 0.8207, "step": 8003 }, { "epoch": 0.68, "learning_rate": 2.4119692634948895e-06, "loss": 0.8485, "step": 8004 }, { "epoch": 0.68, "learning_rate": 2.4107869895326753e-06, "loss": 0.4689, "step": 8005 }, { "epoch": 0.68, "learning_rate": 2.409604913347468e-06, "loss": 0.8216, "step": 8006 }, { "epoch": 0.68, "learning_rate": 2.4084230350295635e-06, "loss": 0.8073, "step": 8007 }, { "epoch": 0.68, "learning_rate": 2.4072413546692364e-06, "loss": 0.7642, "step": 8008 }, { "epoch": 0.68, "learning_rate": 2.4060598723567544e-06, "loss": 0.8145, "step": 8009 }, { "epoch": 0.68, "learning_rate": 2.40487858818236e-06, "loss": 0.75, "step": 8010 }, { "epoch": 0.68, "learning_rate": 2.4036975022362845e-06, "loss": 0.8289, "step": 8011 }, { "epoch": 0.68, "learning_rate": 2.4025166146087505e-06, "loss": 0.7886, "step": 8012 }, { "epoch": 0.68, "learning_rate": 2.401335925389955e-06, "loss": 0.8273, "step": 8013 }, { "epoch": 0.68, "learning_rate": 2.400155434670089e-06, "loss": 0.7796, "step": 8014 }, { "epoch": 0.68, "learning_rate": 2.3989751425393233e-06, "loss": 0.8528, "step": 8015 }, { "epoch": 0.68, "learning_rate": 2.3977950490878126e-06, "loss": 0.7281, "step": 8016 }, { "epoch": 0.68, "learning_rate": 2.3966151544057e-06, "loss": 0.7891, "step": 8017 }, { "epoch": 0.68, "learning_rate": 2.39543545858311e-06, "loss": 0.7612, "step": 8018 }, { "epoch": 0.68, "learning_rate": 2.3942559617101557e-06, "loss": 0.7952, "step": 8019 }, { "epoch": 0.68, "learning_rate": 2.3930766638769325e-06, "loss": 0.8512, "step": 8020 }, { "epoch": 0.68, "learning_rate": 2.391897565173519e-06, "loss": 0.7338, "step": 8021 }, { "epoch": 0.68, "learning_rate": 2.390718665689986e-06, "loss": 0.7628, "step": 8022 }, { "epoch": 0.68, "learning_rate": 2.389539965516376e-06, "loss": 0.7967, "step": 8023 }, { "epoch": 0.68, "learning_rate": 2.38836146474273e-06, "loss": 0.8171, "step": 8024 }, { "epoch": 0.68, "learning_rate": 2.387183163459066e-06, "loss": 0.8446, "step": 8025 }, { "epoch": 0.68, "learning_rate": 2.3860050617553867e-06, "loss": 0.8317, "step": 8026 }, { "epoch": 0.68, "learning_rate": 2.3848271597216843e-06, "loss": 0.7803, "step": 8027 }, { "epoch": 0.69, "learning_rate": 2.383649457447932e-06, "loss": 0.7617, "step": 8028 }, { "epoch": 0.69, "learning_rate": 2.3824719550240876e-06, "loss": 0.7924, "step": 8029 }, { "epoch": 0.69, "learning_rate": 2.3812946525400954e-06, "loss": 0.8029, "step": 8030 }, { "epoch": 0.69, "learning_rate": 2.380117550085882e-06, "loss": 0.7705, "step": 8031 }, { "epoch": 0.69, "learning_rate": 2.3789406477513627e-06, "loss": 0.7384, "step": 8032 }, { "epoch": 0.69, "learning_rate": 2.377763945626435e-06, "loss": 0.5316, "step": 8033 }, { "epoch": 0.69, "learning_rate": 2.376587443800978e-06, "loss": 0.8314, "step": 8034 }, { "epoch": 0.69, "learning_rate": 2.3754111423648634e-06, "loss": 0.8536, "step": 8035 }, { "epoch": 0.69, "learning_rate": 2.3742350414079413e-06, "loss": 0.838, "step": 8036 }, { "epoch": 0.69, "learning_rate": 2.3730591410200473e-06, "loss": 0.8447, "step": 8037 }, { "epoch": 0.69, "learning_rate": 2.371883441291003e-06, "loss": 0.8004, "step": 8038 }, { "epoch": 0.69, "learning_rate": 2.3707079423106123e-06, "loss": 0.814, "step": 8039 }, { "epoch": 0.69, "learning_rate": 2.369532644168669e-06, "loss": 0.8018, "step": 8040 }, { "epoch": 0.69, "learning_rate": 2.3683575469549454e-06, "loss": 0.8214, "step": 8041 }, { "epoch": 0.69, "learning_rate": 2.367182650759206e-06, "loss": 0.7567, "step": 8042 }, { "epoch": 0.69, "learning_rate": 2.3660079556711902e-06, "loss": 0.7699, "step": 8043 }, { "epoch": 0.69, "learning_rate": 2.364833461780627e-06, "loss": 0.7622, "step": 8044 }, { "epoch": 0.69, "learning_rate": 2.3636591691772335e-06, "loss": 0.8298, "step": 8045 }, { "epoch": 0.69, "learning_rate": 2.3624850779507045e-06, "loss": 0.8066, "step": 8046 }, { "epoch": 0.69, "learning_rate": 2.3613111881907273e-06, "loss": 0.7784, "step": 8047 }, { "epoch": 0.69, "learning_rate": 2.3601374999869664e-06, "loss": 0.7912, "step": 8048 }, { "epoch": 0.69, "learning_rate": 2.3589640134290747e-06, "loss": 0.8072, "step": 8049 }, { "epoch": 0.69, "learning_rate": 2.3577907286066886e-06, "loss": 0.803, "step": 8050 }, { "epoch": 0.69, "learning_rate": 2.356617645609428e-06, "loss": 0.8317, "step": 8051 }, { "epoch": 0.69, "learning_rate": 2.355444764526903e-06, "loss": 0.7823, "step": 8052 }, { "epoch": 0.69, "learning_rate": 2.3542720854487004e-06, "loss": 0.7966, "step": 8053 }, { "epoch": 0.69, "learning_rate": 2.3530996084643954e-06, "loss": 0.7942, "step": 8054 }, { "epoch": 0.69, "learning_rate": 2.3519273336635523e-06, "loss": 0.8395, "step": 8055 }, { "epoch": 0.69, "learning_rate": 2.350755261135708e-06, "loss": 0.8364, "step": 8056 }, { "epoch": 0.69, "learning_rate": 2.349583390970397e-06, "loss": 0.7901, "step": 8057 }, { "epoch": 0.69, "learning_rate": 2.34841172325713e-06, "loss": 0.8424, "step": 8058 }, { "epoch": 0.69, "learning_rate": 2.3472402580854037e-06, "loss": 0.8124, "step": 8059 }, { "epoch": 0.69, "learning_rate": 2.346068995544704e-06, "loss": 0.8148, "step": 8060 }, { "epoch": 0.69, "learning_rate": 2.344897935724496e-06, "loss": 0.7786, "step": 8061 }, { "epoch": 0.69, "learning_rate": 2.343727078714231e-06, "loss": 0.8375, "step": 8062 }, { "epoch": 0.69, "learning_rate": 2.3425564246033455e-06, "loss": 0.8, "step": 8063 }, { "epoch": 0.69, "learning_rate": 2.3413859734812566e-06, "loss": 0.7927, "step": 8064 }, { "epoch": 0.69, "learning_rate": 2.340215725437374e-06, "loss": 0.8127, "step": 8065 }, { "epoch": 0.69, "learning_rate": 2.3390456805610857e-06, "loss": 0.8103, "step": 8066 }, { "epoch": 0.69, "learning_rate": 2.337875838941762e-06, "loss": 0.7932, "step": 8067 }, { "epoch": 0.69, "learning_rate": 2.3367062006687664e-06, "loss": 0.8284, "step": 8068 }, { "epoch": 0.69, "learning_rate": 2.3355367658314393e-06, "loss": 0.8139, "step": 8069 }, { "epoch": 0.69, "learning_rate": 2.334367534519108e-06, "loss": 0.7864, "step": 8070 }, { "epoch": 0.69, "learning_rate": 2.3331985068210843e-06, "loss": 0.7894, "step": 8071 }, { "epoch": 0.69, "learning_rate": 2.3320296828266624e-06, "loss": 0.7309, "step": 8072 }, { "epoch": 0.69, "learning_rate": 2.3308610626251272e-06, "loss": 0.8448, "step": 8073 }, { "epoch": 0.69, "learning_rate": 2.3296926463057396e-06, "loss": 0.8072, "step": 8074 }, { "epoch": 0.69, "learning_rate": 2.328524433957754e-06, "loss": 0.7662, "step": 8075 }, { "epoch": 0.69, "learning_rate": 2.3273564256703996e-06, "loss": 0.881, "step": 8076 }, { "epoch": 0.69, "learning_rate": 2.3261886215328942e-06, "loss": 0.8103, "step": 8077 }, { "epoch": 0.69, "learning_rate": 2.3250210216344447e-06, "loss": 0.8165, "step": 8078 }, { "epoch": 0.69, "learning_rate": 2.3238536260642343e-06, "loss": 0.8054, "step": 8079 }, { "epoch": 0.69, "learning_rate": 2.322686434911438e-06, "loss": 0.7909, "step": 8080 }, { "epoch": 0.69, "learning_rate": 2.3215194482652098e-06, "loss": 0.8483, "step": 8081 }, { "epoch": 0.69, "learning_rate": 2.32035266621469e-06, "loss": 0.8148, "step": 8082 }, { "epoch": 0.69, "learning_rate": 2.3191860888490037e-06, "loss": 0.8181, "step": 8083 }, { "epoch": 0.69, "learning_rate": 2.3180197162572575e-06, "loss": 0.8249, "step": 8084 }, { "epoch": 0.69, "learning_rate": 2.3168535485285493e-06, "loss": 0.8007, "step": 8085 }, { "epoch": 0.69, "learning_rate": 2.3156875857519534e-06, "loss": 0.78, "step": 8086 }, { "epoch": 0.69, "learning_rate": 2.3145218280165316e-06, "loss": 0.7838, "step": 8087 }, { "epoch": 0.69, "learning_rate": 2.3133562754113332e-06, "loss": 0.7855, "step": 8088 }, { "epoch": 0.69, "learning_rate": 2.3121909280253866e-06, "loss": 0.8135, "step": 8089 }, { "epoch": 0.69, "learning_rate": 2.311025785947708e-06, "loss": 0.8397, "step": 8090 }, { "epoch": 0.69, "learning_rate": 2.3098608492672964e-06, "loss": 0.7813, "step": 8091 }, { "epoch": 0.69, "learning_rate": 2.3086961180731334e-06, "loss": 0.8523, "step": 8092 }, { "epoch": 0.69, "learning_rate": 2.3075315924541903e-06, "loss": 0.8246, "step": 8093 }, { "epoch": 0.69, "learning_rate": 2.3063672724994177e-06, "loss": 0.8252, "step": 8094 }, { "epoch": 0.69, "learning_rate": 2.305203158297753e-06, "loss": 0.7305, "step": 8095 }, { "epoch": 0.69, "learning_rate": 2.3040392499381164e-06, "loss": 0.7954, "step": 8096 }, { "epoch": 0.69, "learning_rate": 2.302875547509411e-06, "loss": 0.8056, "step": 8097 }, { "epoch": 0.69, "learning_rate": 2.3017120511005302e-06, "loss": 0.8625, "step": 8098 }, { "epoch": 0.69, "learning_rate": 2.300548760800346e-06, "loss": 0.8091, "step": 8099 }, { "epoch": 0.69, "learning_rate": 2.2993856766977134e-06, "loss": 0.8129, "step": 8100 }, { "epoch": 0.69, "learning_rate": 2.29822279888148e-06, "loss": 0.8141, "step": 8101 }, { "epoch": 0.69, "learning_rate": 2.2970601274404693e-06, "loss": 0.7724, "step": 8102 }, { "epoch": 0.69, "learning_rate": 2.2958976624634915e-06, "loss": 0.8366, "step": 8103 }, { "epoch": 0.69, "learning_rate": 2.294735404039343e-06, "loss": 0.8127, "step": 8104 }, { "epoch": 0.69, "learning_rate": 2.2935733522567995e-06, "loss": 0.7871, "step": 8105 }, { "epoch": 0.69, "learning_rate": 2.292411507204628e-06, "loss": 0.8238, "step": 8106 }, { "epoch": 0.69, "learning_rate": 2.291249868971576e-06, "loss": 0.7698, "step": 8107 }, { "epoch": 0.69, "learning_rate": 2.290088437646373e-06, "loss": 0.8321, "step": 8108 }, { "epoch": 0.69, "learning_rate": 2.288927213317737e-06, "loss": 0.7835, "step": 8109 }, { "epoch": 0.69, "learning_rate": 2.2877661960743645e-06, "loss": 0.7975, "step": 8110 }, { "epoch": 0.69, "learning_rate": 2.2866053860049442e-06, "loss": 0.847, "step": 8111 }, { "epoch": 0.69, "learning_rate": 2.285444783198143e-06, "loss": 0.8145, "step": 8112 }, { "epoch": 0.69, "learning_rate": 2.2842843877426113e-06, "loss": 0.8175, "step": 8113 }, { "epoch": 0.69, "learning_rate": 2.2831241997269893e-06, "loss": 0.8482, "step": 8114 }, { "epoch": 0.69, "learning_rate": 2.2819642192398965e-06, "loss": 0.7987, "step": 8115 }, { "epoch": 0.69, "learning_rate": 2.2808044463699388e-06, "loss": 0.8175, "step": 8116 }, { "epoch": 0.69, "learning_rate": 2.279644881205702e-06, "loss": 0.7827, "step": 8117 }, { "epoch": 0.69, "learning_rate": 2.278485523835764e-06, "loss": 0.8196, "step": 8118 }, { "epoch": 0.69, "learning_rate": 2.2773263743486806e-06, "loss": 0.7968, "step": 8119 }, { "epoch": 0.69, "learning_rate": 2.276167432832991e-06, "loss": 0.8171, "step": 8120 }, { "epoch": 0.69, "learning_rate": 2.2750086993772254e-06, "loss": 0.8223, "step": 8121 }, { "epoch": 0.69, "learning_rate": 2.273850174069891e-06, "loss": 0.8292, "step": 8122 }, { "epoch": 0.69, "learning_rate": 2.2726918569994827e-06, "loss": 0.8301, "step": 8123 }, { "epoch": 0.69, "learning_rate": 2.2715337482544773e-06, "loss": 0.785, "step": 8124 }, { "epoch": 0.69, "learning_rate": 2.2703758479233363e-06, "loss": 0.7755, "step": 8125 }, { "epoch": 0.69, "learning_rate": 2.2692181560945086e-06, "loss": 0.7802, "step": 8126 }, { "epoch": 0.69, "learning_rate": 2.2680606728564236e-06, "loss": 0.834, "step": 8127 }, { "epoch": 0.69, "learning_rate": 2.2669033982974946e-06, "loss": 0.8065, "step": 8128 }, { "epoch": 0.69, "learning_rate": 2.265746332506121e-06, "loss": 0.7814, "step": 8129 }, { "epoch": 0.69, "learning_rate": 2.264589475570682e-06, "loss": 0.839, "step": 8130 }, { "epoch": 0.69, "learning_rate": 2.2634328275795497e-06, "loss": 0.8374, "step": 8131 }, { "epoch": 0.69, "learning_rate": 2.2622763886210715e-06, "loss": 0.8355, "step": 8132 }, { "epoch": 0.69, "learning_rate": 2.2611201587835808e-06, "loss": 0.8057, "step": 8133 }, { "epoch": 0.69, "learning_rate": 2.259964138155399e-06, "loss": 0.8216, "step": 8134 }, { "epoch": 0.69, "learning_rate": 2.2588083268248273e-06, "loss": 0.7922, "step": 8135 }, { "epoch": 0.69, "learning_rate": 2.2576527248801532e-06, "loss": 0.7098, "step": 8136 }, { "epoch": 0.69, "learning_rate": 2.2564973324096463e-06, "loss": 0.8165, "step": 8137 }, { "epoch": 0.69, "learning_rate": 2.2553421495015593e-06, "loss": 0.7512, "step": 8138 }, { "epoch": 0.69, "learning_rate": 2.2541871762441352e-06, "loss": 0.7954, "step": 8139 }, { "epoch": 0.69, "learning_rate": 2.253032412725595e-06, "loss": 0.7708, "step": 8140 }, { "epoch": 0.69, "learning_rate": 2.2518778590341424e-06, "loss": 0.7944, "step": 8141 }, { "epoch": 0.69, "learning_rate": 2.2507235152579742e-06, "loss": 0.7879, "step": 8142 }, { "epoch": 0.69, "learning_rate": 2.2495693814852578e-06, "loss": 0.784, "step": 8143 }, { "epoch": 0.69, "learning_rate": 2.248415457804156e-06, "loss": 0.7555, "step": 8144 }, { "epoch": 0.7, "learning_rate": 2.247261744302811e-06, "loss": 0.8022, "step": 8145 }, { "epoch": 0.7, "learning_rate": 2.2461082410693463e-06, "loss": 0.7814, "step": 8146 }, { "epoch": 0.7, "learning_rate": 2.244954948191876e-06, "loss": 0.8082, "step": 8147 }, { "epoch": 0.7, "learning_rate": 2.243801865758493e-06, "loss": 0.8116, "step": 8148 }, { "epoch": 0.7, "learning_rate": 2.2426489938572754e-06, "loss": 0.8446, "step": 8149 }, { "epoch": 0.7, "learning_rate": 2.241496332576285e-06, "loss": 0.8506, "step": 8150 }, { "epoch": 0.7, "learning_rate": 2.240343882003566e-06, "loss": 0.8366, "step": 8151 }, { "epoch": 0.7, "learning_rate": 2.2391916422271527e-06, "loss": 0.7602, "step": 8152 }, { "epoch": 0.7, "learning_rate": 2.238039613335054e-06, "loss": 0.812, "step": 8153 }, { "epoch": 0.7, "learning_rate": 2.2368877954152723e-06, "loss": 0.8482, "step": 8154 }, { "epoch": 0.7, "learning_rate": 2.235736188555787e-06, "loss": 0.7842, "step": 8155 }, { "epoch": 0.7, "learning_rate": 2.2345847928445635e-06, "loss": 0.8093, "step": 8156 }, { "epoch": 0.7, "learning_rate": 2.233433608369551e-06, "loss": 0.8001, "step": 8157 }, { "epoch": 0.7, "learning_rate": 2.2322826352186815e-06, "loss": 0.8265, "step": 8158 }, { "epoch": 0.7, "learning_rate": 2.2311318734798755e-06, "loss": 0.7996, "step": 8159 }, { "epoch": 0.7, "learning_rate": 2.2299813232410316e-06, "loss": 0.8231, "step": 8160 }, { "epoch": 0.7, "learning_rate": 2.2288309845900337e-06, "loss": 0.8353, "step": 8161 }, { "epoch": 0.7, "learning_rate": 2.2276808576147547e-06, "loss": 0.8073, "step": 8162 }, { "epoch": 0.7, "learning_rate": 2.2265309424030407e-06, "loss": 0.8531, "step": 8163 }, { "epoch": 0.7, "learning_rate": 2.2253812390427325e-06, "loss": 0.773, "step": 8164 }, { "epoch": 0.7, "learning_rate": 2.2242317476216496e-06, "loss": 0.8097, "step": 8165 }, { "epoch": 0.7, "learning_rate": 2.223082468227593e-06, "loss": 0.794, "step": 8166 }, { "epoch": 0.7, "learning_rate": 2.2219334009483545e-06, "loss": 0.8195, "step": 8167 }, { "epoch": 0.7, "learning_rate": 2.2207845458717038e-06, "loss": 0.8216, "step": 8168 }, { "epoch": 0.7, "learning_rate": 2.219635903085396e-06, "loss": 0.791, "step": 8169 }, { "epoch": 0.7, "learning_rate": 2.2184874726771697e-06, "loss": 0.8234, "step": 8170 }, { "epoch": 0.7, "learning_rate": 2.217339254734747e-06, "loss": 0.808, "step": 8171 }, { "epoch": 0.7, "learning_rate": 2.2161912493458376e-06, "loss": 0.7802, "step": 8172 }, { "epoch": 0.7, "learning_rate": 2.21504345659813e-06, "loss": 0.7829, "step": 8173 }, { "epoch": 0.7, "learning_rate": 2.213895876579297e-06, "loss": 0.7739, "step": 8174 }, { "epoch": 0.7, "learning_rate": 2.2127485093770013e-06, "loss": 0.806, "step": 8175 }, { "epoch": 0.7, "learning_rate": 2.2116013550788775e-06, "loss": 0.8479, "step": 8176 }, { "epoch": 0.7, "learning_rate": 2.210454413772556e-06, "loss": 0.7511, "step": 8177 }, { "epoch": 0.7, "learning_rate": 2.2093076855456446e-06, "loss": 0.8185, "step": 8178 }, { "epoch": 0.7, "learning_rate": 2.2081611704857347e-06, "loss": 0.8176, "step": 8179 }, { "epoch": 0.7, "learning_rate": 2.2070148686804054e-06, "loss": 0.8351, "step": 8180 }, { "epoch": 0.7, "learning_rate": 2.2058687802172163e-06, "loss": 0.8317, "step": 8181 }, { "epoch": 0.7, "learning_rate": 2.2047229051837107e-06, "loss": 0.7941, "step": 8182 }, { "epoch": 0.7, "learning_rate": 2.203577243667416e-06, "loss": 0.7898, "step": 8183 }, { "epoch": 0.7, "learning_rate": 2.2024317957558423e-06, "loss": 0.8063, "step": 8184 }, { "epoch": 0.7, "learning_rate": 2.2012865615364876e-06, "loss": 0.808, "step": 8185 }, { "epoch": 0.7, "learning_rate": 2.200141541096828e-06, "loss": 0.7755, "step": 8186 }, { "epoch": 0.7, "learning_rate": 2.1989967345243285e-06, "loss": 0.8466, "step": 8187 }, { "epoch": 0.7, "learning_rate": 2.197852141906436e-06, "loss": 0.8159, "step": 8188 }, { "epoch": 0.7, "learning_rate": 2.196707763330573e-06, "loss": 0.7792, "step": 8189 }, { "epoch": 0.7, "learning_rate": 2.195563598884161e-06, "loss": 0.7785, "step": 8190 }, { "epoch": 0.7, "learning_rate": 2.194419648654591e-06, "loss": 0.7938, "step": 8191 }, { "epoch": 0.7, "learning_rate": 2.1932759127292485e-06, "loss": 0.8457, "step": 8192 }, { "epoch": 0.7, "learning_rate": 2.192132391195496e-06, "loss": 0.8381, "step": 8193 }, { "epoch": 0.7, "learning_rate": 2.1909890841406787e-06, "loss": 0.8011, "step": 8194 }, { "epoch": 0.7, "learning_rate": 2.1898459916521343e-06, "loss": 0.819, "step": 8195 }, { "epoch": 0.7, "learning_rate": 2.1887031138171705e-06, "loss": 0.7413, "step": 8196 }, { "epoch": 0.7, "learning_rate": 2.187560450723091e-06, "loss": 0.824, "step": 8197 }, { "epoch": 0.7, "learning_rate": 2.186418002457177e-06, "loss": 0.8019, "step": 8198 }, { "epoch": 0.7, "learning_rate": 2.1852757691066916e-06, "loss": 0.8519, "step": 8199 }, { "epoch": 0.7, "learning_rate": 2.184133750758889e-06, "loss": 0.8224, "step": 8200 }, { "epoch": 0.7, "learning_rate": 2.182991947501e-06, "loss": 0.8372, "step": 8201 }, { "epoch": 0.7, "learning_rate": 2.1818503594202413e-06, "loss": 0.804, "step": 8202 }, { "epoch": 0.7, "learning_rate": 2.1807089866038124e-06, "loss": 0.7955, "step": 8203 }, { "epoch": 0.7, "learning_rate": 2.179567829138896e-06, "loss": 0.7649, "step": 8204 }, { "epoch": 0.7, "learning_rate": 2.1784268871126634e-06, "loss": 0.8013, "step": 8205 }, { "epoch": 0.7, "learning_rate": 2.1772861606122626e-06, "loss": 0.7866, "step": 8206 }, { "epoch": 0.7, "learning_rate": 2.176145649724827e-06, "loss": 0.8272, "step": 8207 }, { "epoch": 0.7, "learning_rate": 2.1750053545374793e-06, "loss": 0.8115, "step": 8208 }, { "epoch": 0.7, "learning_rate": 2.173865275137314e-06, "loss": 0.7971, "step": 8209 }, { "epoch": 0.7, "learning_rate": 2.1727254116114218e-06, "loss": 0.7838, "step": 8210 }, { "epoch": 0.7, "learning_rate": 2.1715857640468686e-06, "loss": 0.8234, "step": 8211 }, { "epoch": 0.7, "learning_rate": 2.170446332530705e-06, "loss": 0.7839, "step": 8212 }, { "epoch": 0.7, "learning_rate": 2.1693071171499706e-06, "loss": 0.764, "step": 8213 }, { "epoch": 0.7, "learning_rate": 2.1681681179916796e-06, "loss": 0.7772, "step": 8214 }, { "epoch": 0.7, "learning_rate": 2.167029335142841e-06, "loss": 0.7731, "step": 8215 }, { "epoch": 0.7, "learning_rate": 2.1658907686904346e-06, "loss": 0.8186, "step": 8216 }, { "epoch": 0.7, "learning_rate": 2.16475241872143e-06, "loss": 0.8204, "step": 8217 }, { "epoch": 0.7, "learning_rate": 2.163614285322783e-06, "loss": 0.8268, "step": 8218 }, { "epoch": 0.7, "learning_rate": 2.1624763685814274e-06, "loss": 0.7808, "step": 8219 }, { "epoch": 0.7, "learning_rate": 2.1613386685842863e-06, "loss": 0.8046, "step": 8220 }, { "epoch": 0.7, "learning_rate": 2.1602011854182624e-06, "loss": 0.8255, "step": 8221 }, { "epoch": 0.7, "learning_rate": 2.159063919170238e-06, "loss": 0.8047, "step": 8222 }, { "epoch": 0.7, "learning_rate": 2.1579268699270865e-06, "loss": 0.5254, "step": 8223 }, { "epoch": 0.7, "learning_rate": 2.1567900377756603e-06, "loss": 0.7977, "step": 8224 }, { "epoch": 0.7, "learning_rate": 2.1556534228027977e-06, "loss": 0.8194, "step": 8225 }, { "epoch": 0.7, "learning_rate": 2.1545170250953184e-06, "loss": 0.7368, "step": 8226 }, { "epoch": 0.7, "learning_rate": 2.1533808447400247e-06, "loss": 0.7405, "step": 8227 }, { "epoch": 0.7, "learning_rate": 2.1522448818237078e-06, "loss": 0.5718, "step": 8228 }, { "epoch": 0.7, "learning_rate": 2.151109136433132e-06, "loss": 0.7999, "step": 8229 }, { "epoch": 0.7, "learning_rate": 2.1499736086550573e-06, "loss": 0.768, "step": 8230 }, { "epoch": 0.7, "learning_rate": 2.1488382985762173e-06, "loss": 0.8054, "step": 8231 }, { "epoch": 0.7, "learning_rate": 2.147703206283332e-06, "loss": 0.6241, "step": 8232 }, { "epoch": 0.7, "learning_rate": 2.1465683318631087e-06, "loss": 0.8288, "step": 8233 }, { "epoch": 0.7, "learning_rate": 2.145433675402233e-06, "loss": 0.7776, "step": 8234 }, { "epoch": 0.7, "learning_rate": 2.1442992369873767e-06, "loss": 0.8191, "step": 8235 }, { "epoch": 0.7, "learning_rate": 2.143165016705192e-06, "loss": 0.7742, "step": 8236 }, { "epoch": 0.7, "learning_rate": 2.1420310146423157e-06, "loss": 0.8235, "step": 8237 }, { "epoch": 0.7, "learning_rate": 2.1408972308853714e-06, "loss": 0.8105, "step": 8238 }, { "epoch": 0.7, "learning_rate": 2.1397636655209617e-06, "loss": 0.7971, "step": 8239 }, { "epoch": 0.7, "learning_rate": 2.1386303186356726e-06, "loss": 0.778, "step": 8240 }, { "epoch": 0.7, "learning_rate": 2.13749719031608e-06, "loss": 0.7903, "step": 8241 }, { "epoch": 0.7, "learning_rate": 2.1363642806487307e-06, "loss": 0.7908, "step": 8242 }, { "epoch": 0.7, "learning_rate": 2.135231589720167e-06, "loss": 0.8148, "step": 8243 }, { "epoch": 0.7, "learning_rate": 2.1340991176169078e-06, "loss": 0.7703, "step": 8244 }, { "epoch": 0.7, "learning_rate": 2.132966864425455e-06, "loss": 0.7982, "step": 8245 }, { "epoch": 0.7, "learning_rate": 2.1318348302323005e-06, "loss": 0.7996, "step": 8246 }, { "epoch": 0.7, "learning_rate": 2.130703015123911e-06, "loss": 0.7395, "step": 8247 }, { "epoch": 0.7, "learning_rate": 2.129571419186742e-06, "loss": 0.7964, "step": 8248 }, { "epoch": 0.7, "learning_rate": 2.12844004250723e-06, "loss": 0.8584, "step": 8249 }, { "epoch": 0.7, "learning_rate": 2.127308885171792e-06, "loss": 0.8493, "step": 8250 }, { "epoch": 0.7, "learning_rate": 2.126177947266837e-06, "loss": 0.8234, "step": 8251 }, { "epoch": 0.7, "learning_rate": 2.1250472288787486e-06, "loss": 0.7629, "step": 8252 }, { "epoch": 0.7, "learning_rate": 2.1239167300938957e-06, "loss": 0.8082, "step": 8253 }, { "epoch": 0.7, "learning_rate": 2.1227864509986358e-06, "loss": 0.8548, "step": 8254 }, { "epoch": 0.7, "learning_rate": 2.121656391679299e-06, "loss": 0.7705, "step": 8255 }, { "epoch": 0.7, "learning_rate": 2.1205265522222102e-06, "loss": 0.7954, "step": 8256 }, { "epoch": 0.7, "learning_rate": 2.1193969327136683e-06, "loss": 0.8075, "step": 8257 }, { "epoch": 0.7, "learning_rate": 2.1182675332399627e-06, "loss": 0.8541, "step": 8258 }, { "epoch": 0.7, "learning_rate": 2.117138353887361e-06, "loss": 0.7445, "step": 8259 }, { "epoch": 0.7, "learning_rate": 2.116009394742114e-06, "loss": 0.806, "step": 8260 }, { "epoch": 0.7, "learning_rate": 2.114880655890462e-06, "loss": 0.8135, "step": 8261 }, { "epoch": 0.71, "learning_rate": 2.1137521374186176e-06, "loss": 0.8111, "step": 8262 }, { "epoch": 0.71, "learning_rate": 2.1126238394127868e-06, "loss": 0.8423, "step": 8263 }, { "epoch": 0.71, "learning_rate": 2.1114957619591536e-06, "loss": 0.7897, "step": 8264 }, { "epoch": 0.71, "learning_rate": 2.1103679051438843e-06, "loss": 0.789, "step": 8265 }, { "epoch": 0.71, "learning_rate": 2.1092402690531335e-06, "loss": 0.8387, "step": 8266 }, { "epoch": 0.71, "learning_rate": 2.1081128537730337e-06, "loss": 0.8061, "step": 8267 }, { "epoch": 0.71, "learning_rate": 2.106985659389704e-06, "loss": 0.8172, "step": 8268 }, { "epoch": 0.71, "learning_rate": 2.1058586859892426e-06, "loss": 0.8241, "step": 8269 }, { "epoch": 0.71, "learning_rate": 2.104731933657734e-06, "loss": 0.7869, "step": 8270 }, { "epoch": 0.71, "learning_rate": 2.103605402481247e-06, "loss": 0.8024, "step": 8271 }, { "epoch": 0.71, "learning_rate": 2.102479092545831e-06, "loss": 0.7933, "step": 8272 }, { "epoch": 0.71, "learning_rate": 2.101353003937517e-06, "loss": 0.7828, "step": 8273 }, { "epoch": 0.71, "learning_rate": 2.1002271367423265e-06, "loss": 0.8288, "step": 8274 }, { "epoch": 0.71, "learning_rate": 2.099101491046252e-06, "loss": 0.7607, "step": 8275 }, { "epoch": 0.71, "learning_rate": 2.0979760669352805e-06, "loss": 0.8177, "step": 8276 }, { "epoch": 0.71, "learning_rate": 2.0968508644953773e-06, "loss": 0.8195, "step": 8277 }, { "epoch": 0.71, "learning_rate": 2.0957258838124872e-06, "loss": 0.7416, "step": 8278 }, { "epoch": 0.71, "learning_rate": 2.0946011249725472e-06, "loss": 0.7817, "step": 8279 }, { "epoch": 0.71, "learning_rate": 2.0934765880614693e-06, "loss": 0.7527, "step": 8280 }, { "epoch": 0.71, "learning_rate": 2.092352273165151e-06, "loss": 0.8042, "step": 8281 }, { "epoch": 0.71, "learning_rate": 2.091228180369474e-06, "loss": 0.8154, "step": 8282 }, { "epoch": 0.71, "learning_rate": 2.0901043097602997e-06, "loss": 0.8116, "step": 8283 }, { "epoch": 0.71, "learning_rate": 2.0889806614234782e-06, "loss": 0.8356, "step": 8284 }, { "epoch": 0.71, "learning_rate": 2.0878572354448383e-06, "loss": 0.8326, "step": 8285 }, { "epoch": 0.71, "learning_rate": 2.086734031910191e-06, "loss": 0.8041, "step": 8286 }, { "epoch": 0.71, "learning_rate": 2.085611050905335e-06, "loss": 0.7691, "step": 8287 }, { "epoch": 0.71, "learning_rate": 2.084488292516048e-06, "loss": 0.8185, "step": 8288 }, { "epoch": 0.71, "learning_rate": 2.0833657568280926e-06, "loss": 0.8038, "step": 8289 }, { "epoch": 0.71, "learning_rate": 2.082243443927212e-06, "loss": 0.8182, "step": 8290 }, { "epoch": 0.71, "learning_rate": 2.0811213538991337e-06, "loss": 0.8096, "step": 8291 }, { "epoch": 0.71, "learning_rate": 2.0799994868295714e-06, "loss": 0.8051, "step": 8292 }, { "epoch": 0.71, "learning_rate": 2.0788778428042154e-06, "loss": 0.78, "step": 8293 }, { "epoch": 0.71, "learning_rate": 2.0777564219087482e-06, "loss": 0.7677, "step": 8294 }, { "epoch": 0.71, "learning_rate": 2.076635224228822e-06, "loss": 0.8213, "step": 8295 }, { "epoch": 0.71, "learning_rate": 2.075514249850085e-06, "loss": 0.5043, "step": 8296 }, { "epoch": 0.71, "learning_rate": 2.074393498858161e-06, "loss": 0.7956, "step": 8297 }, { "epoch": 0.71, "learning_rate": 2.073272971338657e-06, "loss": 0.7616, "step": 8298 }, { "epoch": 0.71, "learning_rate": 2.0721526673771674e-06, "loss": 0.7815, "step": 8299 }, { "epoch": 0.71, "learning_rate": 2.0710325870592655e-06, "loss": 0.8204, "step": 8300 }, { "epoch": 0.71, "learning_rate": 2.0699127304705087e-06, "loss": 0.8143, "step": 8301 }, { "epoch": 0.71, "learning_rate": 2.068793097696436e-06, "loss": 0.7892, "step": 8302 }, { "epoch": 0.71, "learning_rate": 2.06767368882257e-06, "loss": 0.8511, "step": 8303 }, { "epoch": 0.71, "learning_rate": 2.0665545039344206e-06, "loss": 0.8043, "step": 8304 }, { "epoch": 0.71, "learning_rate": 2.0654355431174738e-06, "loss": 0.8097, "step": 8305 }, { "epoch": 0.71, "learning_rate": 2.0643168064572e-06, "loss": 0.8217, "step": 8306 }, { "epoch": 0.71, "learning_rate": 2.063198294039059e-06, "loss": 0.7746, "step": 8307 }, { "epoch": 0.71, "learning_rate": 2.0620800059484824e-06, "loss": 0.8028, "step": 8308 }, { "epoch": 0.71, "learning_rate": 2.0609619422708947e-06, "loss": 0.5121, "step": 8309 }, { "epoch": 0.71, "learning_rate": 2.0598441030916983e-06, "loss": 0.8214, "step": 8310 }, { "epoch": 0.71, "learning_rate": 2.0587264884962764e-06, "loss": 0.8039, "step": 8311 }, { "epoch": 0.71, "learning_rate": 2.0576090985700025e-06, "loss": 0.785, "step": 8312 }, { "epoch": 0.71, "learning_rate": 2.056491933398227e-06, "loss": 0.828, "step": 8313 }, { "epoch": 0.71, "learning_rate": 2.055374993066283e-06, "loss": 0.8499, "step": 8314 }, { "epoch": 0.71, "learning_rate": 2.0542582776594894e-06, "loss": 0.7646, "step": 8315 }, { "epoch": 0.71, "learning_rate": 2.0531417872631443e-06, "loss": 0.7781, "step": 8316 }, { "epoch": 0.71, "learning_rate": 2.052025521962534e-06, "loss": 0.8349, "step": 8317 }, { "epoch": 0.71, "learning_rate": 2.0509094818429233e-06, "loss": 0.839, "step": 8318 }, { "epoch": 0.71, "learning_rate": 2.0497936669895585e-06, "loss": 0.8456, "step": 8319 }, { "epoch": 0.71, "learning_rate": 2.0486780774876745e-06, "loss": 0.8141, "step": 8320 }, { "epoch": 0.71, "learning_rate": 2.047562713422485e-06, "loss": 0.812, "step": 8321 }, { "epoch": 0.71, "learning_rate": 2.0464475748791863e-06, "loss": 0.8207, "step": 8322 }, { "epoch": 0.71, "learning_rate": 2.0453326619429577e-06, "loss": 0.801, "step": 8323 }, { "epoch": 0.71, "learning_rate": 2.044217974698961e-06, "loss": 0.8117, "step": 8324 }, { "epoch": 0.71, "learning_rate": 2.0431035132323444e-06, "loss": 0.7894, "step": 8325 }, { "epoch": 0.71, "learning_rate": 2.041989277628233e-06, "loss": 0.8001, "step": 8326 }, { "epoch": 0.71, "learning_rate": 2.0408752679717424e-06, "loss": 0.7992, "step": 8327 }, { "epoch": 0.71, "learning_rate": 2.039761484347962e-06, "loss": 0.8387, "step": 8328 }, { "epoch": 0.71, "learning_rate": 2.0386479268419667e-06, "loss": 0.8095, "step": 8329 }, { "epoch": 0.71, "learning_rate": 2.0375345955388203e-06, "loss": 0.8724, "step": 8330 }, { "epoch": 0.71, "learning_rate": 2.0364214905235607e-06, "loss": 0.8215, "step": 8331 }, { "epoch": 0.71, "learning_rate": 2.0353086118812156e-06, "loss": 0.7696, "step": 8332 }, { "epoch": 0.71, "learning_rate": 2.0341959596967907e-06, "loss": 0.8378, "step": 8333 }, { "epoch": 0.71, "learning_rate": 2.0330835340552764e-06, "loss": 0.799, "step": 8334 }, { "epoch": 0.71, "learning_rate": 2.031971335041644e-06, "loss": 0.788, "step": 8335 }, { "epoch": 0.71, "learning_rate": 2.030859362740849e-06, "loss": 0.791, "step": 8336 }, { "epoch": 0.71, "learning_rate": 2.0297476172378312e-06, "loss": 0.8341, "step": 8337 }, { "epoch": 0.71, "learning_rate": 2.02863609861751e-06, "loss": 0.8108, "step": 8338 }, { "epoch": 0.71, "learning_rate": 2.0275248069647875e-06, "loss": 0.7879, "step": 8339 }, { "epoch": 0.71, "learning_rate": 2.026413742364553e-06, "loss": 0.7546, "step": 8340 }, { "epoch": 0.71, "learning_rate": 2.0253029049016728e-06, "loss": 0.8454, "step": 8341 }, { "epoch": 0.71, "learning_rate": 2.024192294660999e-06, "loss": 0.8134, "step": 8342 }, { "epoch": 0.71, "learning_rate": 2.023081911727365e-06, "loss": 0.8415, "step": 8343 }, { "epoch": 0.71, "learning_rate": 2.0219717561855857e-06, "loss": 0.7641, "step": 8344 }, { "epoch": 0.71, "learning_rate": 2.0208618281204638e-06, "loss": 0.8084, "step": 8345 }, { "epoch": 0.71, "learning_rate": 2.0197521276167795e-06, "loss": 0.7732, "step": 8346 }, { "epoch": 0.71, "learning_rate": 2.018642654759298e-06, "loss": 0.8548, "step": 8347 }, { "epoch": 0.71, "learning_rate": 2.0175334096327655e-06, "loss": 0.8284, "step": 8348 }, { "epoch": 0.71, "learning_rate": 2.0164243923219096e-06, "loss": 0.7621, "step": 8349 }, { "epoch": 0.71, "learning_rate": 2.015315602911447e-06, "loss": 0.7967, "step": 8350 }, { "epoch": 0.71, "learning_rate": 2.0142070414860704e-06, "loss": 0.8025, "step": 8351 }, { "epoch": 0.71, "learning_rate": 2.0130987081304556e-06, "loss": 0.7833, "step": 8352 }, { "epoch": 0.71, "learning_rate": 2.011990602929265e-06, "loss": 0.7871, "step": 8353 }, { "epoch": 0.71, "learning_rate": 2.0108827259671414e-06, "loss": 0.8155, "step": 8354 }, { "epoch": 0.71, "learning_rate": 2.0097750773287093e-06, "loss": 0.8023, "step": 8355 }, { "epoch": 0.71, "learning_rate": 2.008667657098576e-06, "loss": 0.8184, "step": 8356 }, { "epoch": 0.71, "learning_rate": 2.00756046536133e-06, "loss": 0.7733, "step": 8357 }, { "epoch": 0.71, "learning_rate": 2.006453502201548e-06, "loss": 0.8171, "step": 8358 }, { "epoch": 0.71, "learning_rate": 2.005346767703782e-06, "loss": 0.7638, "step": 8359 }, { "epoch": 0.71, "learning_rate": 2.004240261952574e-06, "loss": 0.8139, "step": 8360 }, { "epoch": 0.71, "learning_rate": 2.0031339850324412e-06, "loss": 0.8166, "step": 8361 }, { "epoch": 0.71, "learning_rate": 2.0020279370278844e-06, "loss": 0.8377, "step": 8362 }, { "epoch": 0.71, "learning_rate": 2.000922118023394e-06, "loss": 0.8018, "step": 8363 }, { "epoch": 0.71, "learning_rate": 1.999816528103434e-06, "loss": 0.761, "step": 8364 }, { "epoch": 0.71, "learning_rate": 1.9987111673524584e-06, "loss": 0.8112, "step": 8365 }, { "epoch": 0.71, "learning_rate": 1.997606035854898e-06, "loss": 0.7888, "step": 8366 }, { "epoch": 0.71, "learning_rate": 1.9965011336951693e-06, "loss": 0.7927, "step": 8367 }, { "epoch": 0.71, "learning_rate": 1.9953964609576697e-06, "loss": 0.8321, "step": 8368 }, { "epoch": 0.71, "learning_rate": 1.994292017726778e-06, "loss": 0.8424, "step": 8369 }, { "epoch": 0.71, "learning_rate": 1.9931878040868595e-06, "loss": 0.7566, "step": 8370 }, { "epoch": 0.71, "learning_rate": 1.992083820122259e-06, "loss": 0.8059, "step": 8371 }, { "epoch": 0.71, "learning_rate": 1.990980065917303e-06, "loss": 0.7647, "step": 8372 }, { "epoch": 0.71, "learning_rate": 1.989876541556304e-06, "loss": 0.8211, "step": 8373 }, { "epoch": 0.71, "learning_rate": 1.988773247123553e-06, "loss": 0.7876, "step": 8374 }, { "epoch": 0.71, "learning_rate": 1.987670182703326e-06, "loss": 0.5798, "step": 8375 }, { "epoch": 0.71, "learning_rate": 1.9865673483798798e-06, "loss": 0.8394, "step": 8376 }, { "epoch": 0.71, "learning_rate": 1.985464744237453e-06, "loss": 0.8012, "step": 8377 }, { "epoch": 0.71, "learning_rate": 1.9843623703602716e-06, "loss": 0.7885, "step": 8378 }, { "epoch": 0.71, "learning_rate": 1.983260226832539e-06, "loss": 0.764, "step": 8379 }, { "epoch": 0.72, "learning_rate": 1.9821583137384413e-06, "loss": 0.6965, "step": 8380 }, { "epoch": 0.72, "learning_rate": 1.981056631162149e-06, "loss": 0.7903, "step": 8381 }, { "epoch": 0.72, "learning_rate": 1.9799551791878124e-06, "loss": 0.7703, "step": 8382 }, { "epoch": 0.72, "learning_rate": 1.9788539578995684e-06, "loss": 0.8203, "step": 8383 }, { "epoch": 0.72, "learning_rate": 1.977752967381534e-06, "loss": 0.8295, "step": 8384 }, { "epoch": 0.72, "learning_rate": 1.976652207717804e-06, "loss": 0.8051, "step": 8385 }, { "epoch": 0.72, "learning_rate": 1.9755516789924656e-06, "loss": 0.8157, "step": 8386 }, { "epoch": 0.72, "learning_rate": 1.97445138128958e-06, "loss": 0.761, "step": 8387 }, { "epoch": 0.72, "learning_rate": 1.973351314693194e-06, "loss": 0.7786, "step": 8388 }, { "epoch": 0.72, "learning_rate": 1.9722514792873348e-06, "loss": 0.7582, "step": 8389 }, { "epoch": 0.72, "learning_rate": 1.9711518751560128e-06, "loss": 0.813, "step": 8390 }, { "epoch": 0.72, "learning_rate": 1.9700525023832247e-06, "loss": 0.7964, "step": 8391 }, { "epoch": 0.72, "learning_rate": 1.9689533610529436e-06, "loss": 0.8399, "step": 8392 }, { "epoch": 0.72, "learning_rate": 1.9678544512491267e-06, "loss": 0.7785, "step": 8393 }, { "epoch": 0.72, "learning_rate": 1.9667557730557186e-06, "loss": 0.8053, "step": 8394 }, { "epoch": 0.72, "learning_rate": 1.965657326556635e-06, "loss": 0.827, "step": 8395 }, { "epoch": 0.72, "learning_rate": 1.964559111835786e-06, "loss": 0.8097, "step": 8396 }, { "epoch": 0.72, "learning_rate": 1.9634611289770554e-06, "loss": 0.778, "step": 8397 }, { "epoch": 0.72, "learning_rate": 1.962363378064316e-06, "loss": 0.8069, "step": 8398 }, { "epoch": 0.72, "learning_rate": 1.9612658591814176e-06, "loss": 0.8137, "step": 8399 }, { "epoch": 0.72, "learning_rate": 1.9601685724121944e-06, "loss": 0.8535, "step": 8400 }, { "epoch": 0.72, "learning_rate": 1.9590715178404624e-06, "loss": 0.7993, "step": 8401 }, { "epoch": 0.72, "learning_rate": 1.957974695550019e-06, "loss": 0.7753, "step": 8402 }, { "epoch": 0.72, "learning_rate": 1.9568781056246474e-06, "loss": 0.8137, "step": 8403 }, { "epoch": 0.72, "learning_rate": 1.9557817481481095e-06, "loss": 0.6993, "step": 8404 }, { "epoch": 0.72, "learning_rate": 1.954685623204149e-06, "loss": 0.7924, "step": 8405 }, { "epoch": 0.72, "learning_rate": 1.953589730876496e-06, "loss": 0.8334, "step": 8406 }, { "epoch": 0.72, "learning_rate": 1.95249407124886e-06, "loss": 0.8113, "step": 8407 }, { "epoch": 0.72, "learning_rate": 1.9513986444049315e-06, "loss": 0.8132, "step": 8408 }, { "epoch": 0.72, "learning_rate": 1.950303450428385e-06, "loss": 0.7632, "step": 8409 }, { "epoch": 0.72, "learning_rate": 1.9492084894028757e-06, "loss": 0.8176, "step": 8410 }, { "epoch": 0.72, "learning_rate": 1.9481137614120454e-06, "loss": 0.7779, "step": 8411 }, { "epoch": 0.72, "learning_rate": 1.9470192665395125e-06, "loss": 0.7773, "step": 8412 }, { "epoch": 0.72, "learning_rate": 1.9459250048688796e-06, "loss": 0.8133, "step": 8413 }, { "epoch": 0.72, "learning_rate": 1.944830976483735e-06, "loss": 0.739, "step": 8414 }, { "epoch": 0.72, "learning_rate": 1.943737181467642e-06, "loss": 0.7749, "step": 8415 }, { "epoch": 0.72, "learning_rate": 1.942643619904153e-06, "loss": 0.8114, "step": 8416 }, { "epoch": 0.72, "learning_rate": 1.941550291876798e-06, "loss": 0.7808, "step": 8417 }, { "epoch": 0.72, "learning_rate": 1.9404571974690906e-06, "loss": 0.7296, "step": 8418 }, { "epoch": 0.72, "learning_rate": 1.939364336764529e-06, "loss": 0.82, "step": 8419 }, { "epoch": 0.72, "learning_rate": 1.9382717098465903e-06, "loss": 0.8105, "step": 8420 }, { "epoch": 0.72, "learning_rate": 1.9371793167987345e-06, "loss": 0.807, "step": 8421 }, { "epoch": 0.72, "learning_rate": 1.936087157704404e-06, "loss": 0.7973, "step": 8422 }, { "epoch": 0.72, "learning_rate": 1.9349952326470215e-06, "loss": 0.8391, "step": 8423 }, { "epoch": 0.72, "learning_rate": 1.9339035417099977e-06, "loss": 0.8004, "step": 8424 }, { "epoch": 0.72, "learning_rate": 1.9328120849767198e-06, "loss": 0.7857, "step": 8425 }, { "epoch": 0.72, "learning_rate": 1.9317208625305558e-06, "loss": 0.7777, "step": 8426 }, { "epoch": 0.72, "learning_rate": 1.9306298744548647e-06, "loss": 0.8383, "step": 8427 }, { "epoch": 0.72, "learning_rate": 1.9295391208329743e-06, "loss": 0.7894, "step": 8428 }, { "epoch": 0.72, "learning_rate": 1.928448601748208e-06, "loss": 0.8076, "step": 8429 }, { "epoch": 0.72, "learning_rate": 1.927358317283862e-06, "loss": 0.7949, "step": 8430 }, { "epoch": 0.72, "learning_rate": 1.9262682675232176e-06, "loss": 0.7981, "step": 8431 }, { "epoch": 0.72, "learning_rate": 1.9251784525495397e-06, "loss": 0.8131, "step": 8432 }, { "epoch": 0.72, "learning_rate": 1.924088872446073e-06, "loss": 0.8433, "step": 8433 }, { "epoch": 0.72, "learning_rate": 1.922999527296046e-06, "loss": 0.8336, "step": 8434 }, { "epoch": 0.72, "learning_rate": 1.921910417182667e-06, "loss": 0.8075, "step": 8435 }, { "epoch": 0.72, "learning_rate": 1.9208215421891263e-06, "loss": 0.7816, "step": 8436 }, { "epoch": 0.72, "learning_rate": 1.9197329023986016e-06, "loss": 0.8425, "step": 8437 }, { "epoch": 0.72, "learning_rate": 1.918644497894244e-06, "loss": 0.8148, "step": 8438 }, { "epoch": 0.72, "learning_rate": 1.9175563287591965e-06, "loss": 0.8006, "step": 8439 }, { "epoch": 0.72, "learning_rate": 1.916468395076575e-06, "loss": 0.7932, "step": 8440 }, { "epoch": 0.72, "learning_rate": 1.915380696929483e-06, "loss": 0.7839, "step": 8441 }, { "epoch": 0.72, "learning_rate": 1.914293234401004e-06, "loss": 0.5185, "step": 8442 }, { "epoch": 0.72, "learning_rate": 1.9132060075742016e-06, "loss": 0.7558, "step": 8443 }, { "epoch": 0.72, "learning_rate": 1.9121190165321267e-06, "loss": 0.8048, "step": 8444 }, { "epoch": 0.72, "learning_rate": 1.9110322613578088e-06, "loss": 0.8861, "step": 8445 }, { "epoch": 0.72, "learning_rate": 1.9099457421342567e-06, "loss": 0.8211, "step": 8446 }, { "epoch": 0.72, "learning_rate": 1.90885945894447e-06, "loss": 0.8265, "step": 8447 }, { "epoch": 0.72, "learning_rate": 1.907773411871417e-06, "loss": 0.8226, "step": 8448 }, { "epoch": 0.72, "learning_rate": 1.9066876009980606e-06, "loss": 0.7549, "step": 8449 }, { "epoch": 0.72, "learning_rate": 1.9056020264073394e-06, "loss": 0.7937, "step": 8450 }, { "epoch": 0.72, "learning_rate": 1.904516688182173e-06, "loss": 0.7237, "step": 8451 }, { "epoch": 0.72, "learning_rate": 1.9034315864054682e-06, "loss": 0.804, "step": 8452 }, { "epoch": 0.72, "learning_rate": 1.9023467211601082e-06, "loss": 0.84, "step": 8453 }, { "epoch": 0.72, "learning_rate": 1.9012620925289615e-06, "loss": 0.7961, "step": 8454 }, { "epoch": 0.72, "learning_rate": 1.9001777005948768e-06, "loss": 0.7682, "step": 8455 }, { "epoch": 0.72, "learning_rate": 1.8990935454406839e-06, "loss": 0.8204, "step": 8456 }, { "epoch": 0.72, "learning_rate": 1.8980096271492e-06, "loss": 0.825, "step": 8457 }, { "epoch": 0.72, "learning_rate": 1.8969259458032168e-06, "loss": 0.7914, "step": 8458 }, { "epoch": 0.72, "learning_rate": 1.8958425014855108e-06, "loss": 0.5392, "step": 8459 }, { "epoch": 0.72, "learning_rate": 1.8947592942788462e-06, "loss": 0.7652, "step": 8460 }, { "epoch": 0.72, "learning_rate": 1.8936763242659568e-06, "loss": 0.8556, "step": 8461 }, { "epoch": 0.72, "learning_rate": 1.8925935915295695e-06, "loss": 0.8409, "step": 8462 }, { "epoch": 0.72, "learning_rate": 1.891511096152388e-06, "loss": 0.811, "step": 8463 }, { "epoch": 0.72, "learning_rate": 1.8904288382170966e-06, "loss": 0.7692, "step": 8464 }, { "epoch": 0.72, "learning_rate": 1.8893468178063673e-06, "loss": 0.8304, "step": 8465 }, { "epoch": 0.72, "learning_rate": 1.8882650350028463e-06, "loss": 0.7477, "step": 8466 }, { "epoch": 0.72, "learning_rate": 1.8871834898891712e-06, "loss": 0.8136, "step": 8467 }, { "epoch": 0.72, "learning_rate": 1.8861021825479504e-06, "loss": 0.8253, "step": 8468 }, { "epoch": 0.72, "learning_rate": 1.8850211130617796e-06, "loss": 0.7653, "step": 8469 }, { "epoch": 0.72, "learning_rate": 1.8839402815132401e-06, "loss": 0.8345, "step": 8470 }, { "epoch": 0.72, "learning_rate": 1.8828596879848871e-06, "loss": 0.7951, "step": 8471 }, { "epoch": 0.72, "learning_rate": 1.8817793325592653e-06, "loss": 0.8057, "step": 8472 }, { "epoch": 0.72, "learning_rate": 1.880699215318898e-06, "loss": 0.7974, "step": 8473 }, { "epoch": 0.72, "learning_rate": 1.879619336346285e-06, "loss": 0.7766, "step": 8474 }, { "epoch": 0.72, "learning_rate": 1.8785396957239171e-06, "loss": 0.7712, "step": 8475 }, { "epoch": 0.72, "learning_rate": 1.87746029353426e-06, "loss": 0.8335, "step": 8476 }, { "epoch": 0.72, "learning_rate": 1.8763811298597668e-06, "loss": 0.8482, "step": 8477 }, { "epoch": 0.72, "learning_rate": 1.8753022047828679e-06, "loss": 0.7785, "step": 8478 }, { "epoch": 0.72, "learning_rate": 1.8742235183859747e-06, "loss": 0.8067, "step": 8479 }, { "epoch": 0.72, "learning_rate": 1.8731450707514892e-06, "loss": 0.8589, "step": 8480 }, { "epoch": 0.72, "learning_rate": 1.8720668619617805e-06, "loss": 0.8107, "step": 8481 }, { "epoch": 0.72, "learning_rate": 1.870988892099213e-06, "loss": 0.7307, "step": 8482 }, { "epoch": 0.72, "learning_rate": 1.8699111612461258e-06, "loss": 0.7884, "step": 8483 }, { "epoch": 0.72, "learning_rate": 1.8688336694848396e-06, "loss": 0.819, "step": 8484 }, { "epoch": 0.72, "learning_rate": 1.867756416897662e-06, "loss": 0.8189, "step": 8485 }, { "epoch": 0.72, "learning_rate": 1.8666794035668772e-06, "loss": 0.8273, "step": 8486 }, { "epoch": 0.72, "learning_rate": 1.8656026295747532e-06, "loss": 0.7546, "step": 8487 }, { "epoch": 0.72, "learning_rate": 1.8645260950035387e-06, "loss": 0.7888, "step": 8488 }, { "epoch": 0.72, "learning_rate": 1.863449799935464e-06, "loss": 0.7675, "step": 8489 }, { "epoch": 0.72, "learning_rate": 1.8623737444527446e-06, "loss": 0.8046, "step": 8490 }, { "epoch": 0.72, "learning_rate": 1.8612979286375731e-06, "loss": 0.7887, "step": 8491 }, { "epoch": 0.72, "learning_rate": 1.8602223525721241e-06, "loss": 0.8099, "step": 8492 }, { "epoch": 0.72, "learning_rate": 1.8591470163385616e-06, "loss": 0.7947, "step": 8493 }, { "epoch": 0.72, "learning_rate": 1.8580719200190168e-06, "loss": 0.7588, "step": 8494 }, { "epoch": 0.72, "learning_rate": 1.856997063695617e-06, "loss": 0.847, "step": 8495 }, { "epoch": 0.72, "learning_rate": 1.8559224474504634e-06, "loss": 0.7995, "step": 8496 }, { "epoch": 0.73, "learning_rate": 1.8548480713656385e-06, "loss": 0.7907, "step": 8497 }, { "epoch": 0.73, "learning_rate": 1.8537739355232116e-06, "loss": 0.4639, "step": 8498 }, { "epoch": 0.73, "learning_rate": 1.852700040005228e-06, "loss": 0.7991, "step": 8499 }, { "epoch": 0.73, "learning_rate": 1.8516263848937216e-06, "loss": 0.8189, "step": 8500 }, { "epoch": 0.73, "learning_rate": 1.8505529702706986e-06, "loss": 0.7857, "step": 8501 }, { "epoch": 0.73, "learning_rate": 1.849479796218152e-06, "loss": 0.788, "step": 8502 }, { "epoch": 0.73, "learning_rate": 1.8484068628180596e-06, "loss": 0.7831, "step": 8503 }, { "epoch": 0.73, "learning_rate": 1.847334170152374e-06, "loss": 0.7787, "step": 8504 }, { "epoch": 0.73, "learning_rate": 1.8462617183030356e-06, "loss": 0.8417, "step": 8505 }, { "epoch": 0.73, "learning_rate": 1.8451895073519643e-06, "loss": 0.803, "step": 8506 }, { "epoch": 0.73, "learning_rate": 1.8441175373810554e-06, "loss": 0.8298, "step": 8507 }, { "epoch": 0.73, "learning_rate": 1.8430458084721969e-06, "loss": 0.5232, "step": 8508 }, { "epoch": 0.73, "learning_rate": 1.8419743207072487e-06, "loss": 0.8029, "step": 8509 }, { "epoch": 0.73, "learning_rate": 1.8409030741680606e-06, "loss": 0.7901, "step": 8510 }, { "epoch": 0.73, "learning_rate": 1.8398320689364574e-06, "loss": 0.7746, "step": 8511 }, { "epoch": 0.73, "learning_rate": 1.8387613050942466e-06, "loss": 0.7847, "step": 8512 }, { "epoch": 0.73, "learning_rate": 1.8376907827232239e-06, "loss": 0.7768, "step": 8513 }, { "epoch": 0.73, "learning_rate": 1.836620501905153e-06, "loss": 0.7941, "step": 8514 }, { "epoch": 0.73, "learning_rate": 1.835550462721793e-06, "loss": 0.8456, "step": 8515 }, { "epoch": 0.73, "learning_rate": 1.834480665254878e-06, "loss": 0.8152, "step": 8516 }, { "epoch": 0.73, "learning_rate": 1.8334111095861219e-06, "loss": 0.7273, "step": 8517 }, { "epoch": 0.73, "learning_rate": 1.8323417957972263e-06, "loss": 0.7623, "step": 8518 }, { "epoch": 0.73, "learning_rate": 1.8312727239698686e-06, "loss": 0.7525, "step": 8519 }, { "epoch": 0.73, "learning_rate": 1.830203894185711e-06, "loss": 0.5331, "step": 8520 }, { "epoch": 0.73, "learning_rate": 1.8291353065263945e-06, "loss": 0.5094, "step": 8521 }, { "epoch": 0.73, "learning_rate": 1.828066961073543e-06, "loss": 0.8128, "step": 8522 }, { "epoch": 0.73, "learning_rate": 1.8269988579087643e-06, "loss": 0.8089, "step": 8523 }, { "epoch": 0.73, "learning_rate": 1.8259309971136441e-06, "loss": 0.7711, "step": 8524 }, { "epoch": 0.73, "learning_rate": 1.8248633787697495e-06, "loss": 0.7771, "step": 8525 }, { "epoch": 0.73, "learning_rate": 1.8237960029586354e-06, "loss": 0.8073, "step": 8526 }, { "epoch": 0.73, "learning_rate": 1.8227288697618267e-06, "loss": 0.7891, "step": 8527 }, { "epoch": 0.73, "learning_rate": 1.821661979260841e-06, "loss": 0.8007, "step": 8528 }, { "epoch": 0.73, "learning_rate": 1.820595331537171e-06, "loss": 0.8326, "step": 8529 }, { "epoch": 0.73, "learning_rate": 1.8195289266722915e-06, "loss": 0.8165, "step": 8530 }, { "epoch": 0.73, "learning_rate": 1.8184627647476633e-06, "loss": 0.8451, "step": 8531 }, { "epoch": 0.73, "learning_rate": 1.8173968458447227e-06, "loss": 0.8193, "step": 8532 }, { "epoch": 0.73, "learning_rate": 1.8163311700448899e-06, "loss": 0.815, "step": 8533 }, { "epoch": 0.73, "learning_rate": 1.8152657374295674e-06, "loss": 0.7979, "step": 8534 }, { "epoch": 0.73, "learning_rate": 1.814200548080136e-06, "loss": 0.83, "step": 8535 }, { "epoch": 0.73, "learning_rate": 1.8131356020779633e-06, "loss": 0.8103, "step": 8536 }, { "epoch": 0.73, "learning_rate": 1.812070899504394e-06, "loss": 0.7705, "step": 8537 }, { "epoch": 0.73, "learning_rate": 1.811006440440754e-06, "loss": 0.7608, "step": 8538 }, { "epoch": 0.73, "learning_rate": 1.809942224968354e-06, "loss": 0.7577, "step": 8539 }, { "epoch": 0.73, "learning_rate": 1.808878253168484e-06, "loss": 0.834, "step": 8540 }, { "epoch": 0.73, "learning_rate": 1.8078145251224144e-06, "loss": 0.7798, "step": 8541 }, { "epoch": 0.73, "learning_rate": 1.806751040911397e-06, "loss": 0.7707, "step": 8542 }, { "epoch": 0.73, "learning_rate": 1.805687800616669e-06, "loss": 0.7603, "step": 8543 }, { "epoch": 0.73, "learning_rate": 1.8046248043194441e-06, "loss": 0.7948, "step": 8544 }, { "epoch": 0.73, "learning_rate": 1.8035620521009184e-06, "loss": 0.814, "step": 8545 }, { "epoch": 0.73, "learning_rate": 1.802499544042275e-06, "loss": 0.8143, "step": 8546 }, { "epoch": 0.73, "learning_rate": 1.8014372802246661e-06, "loss": 0.7044, "step": 8547 }, { "epoch": 0.73, "learning_rate": 1.8003752607292386e-06, "loss": 0.8122, "step": 8548 }, { "epoch": 0.73, "learning_rate": 1.799313485637113e-06, "loss": 0.8111, "step": 8549 }, { "epoch": 0.73, "learning_rate": 1.7982519550293909e-06, "loss": 0.8275, "step": 8550 }, { "epoch": 0.73, "learning_rate": 1.797190668987161e-06, "loss": 0.7412, "step": 8551 }, { "epoch": 0.73, "learning_rate": 1.7961296275914874e-06, "loss": 0.8105, "step": 8552 }, { "epoch": 0.73, "learning_rate": 1.7950688309234188e-06, "loss": 0.7894, "step": 8553 }, { "epoch": 0.73, "learning_rate": 1.7940082790639833e-06, "loss": 0.7995, "step": 8554 }, { "epoch": 0.73, "learning_rate": 1.79294797209419e-06, "loss": 0.7912, "step": 8555 }, { "epoch": 0.73, "learning_rate": 1.7918879100950332e-06, "loss": 0.802, "step": 8556 }, { "epoch": 0.73, "learning_rate": 1.7908280931474848e-06, "loss": 0.8078, "step": 8557 }, { "epoch": 0.73, "learning_rate": 1.789768521332496e-06, "loss": 0.7701, "step": 8558 }, { "epoch": 0.73, "learning_rate": 1.7887091947310087e-06, "loss": 0.7827, "step": 8559 }, { "epoch": 0.73, "learning_rate": 1.7876501134239316e-06, "loss": 0.8097, "step": 8560 }, { "epoch": 0.73, "learning_rate": 1.7865912774921685e-06, "loss": 0.772, "step": 8561 }, { "epoch": 0.73, "learning_rate": 1.7855326870165961e-06, "loss": 0.7723, "step": 8562 }, { "epoch": 0.73, "learning_rate": 1.7844743420780741e-06, "loss": 0.8002, "step": 8563 }, { "epoch": 0.73, "learning_rate": 1.783416242757447e-06, "loss": 0.7812, "step": 8564 }, { "epoch": 0.73, "learning_rate": 1.782358389135536e-06, "loss": 0.8261, "step": 8565 }, { "epoch": 0.73, "learning_rate": 1.7813007812931454e-06, "loss": 0.7984, "step": 8566 }, { "epoch": 0.73, "learning_rate": 1.7802434193110608e-06, "loss": 0.7388, "step": 8567 }, { "epoch": 0.73, "learning_rate": 1.7791863032700458e-06, "loss": 0.7913, "step": 8568 }, { "epoch": 0.73, "learning_rate": 1.7781294332508532e-06, "loss": 0.8057, "step": 8569 }, { "epoch": 0.73, "learning_rate": 1.7770728093342094e-06, "loss": 0.7186, "step": 8570 }, { "epoch": 0.73, "learning_rate": 1.7760164316008234e-06, "loss": 0.8419, "step": 8571 }, { "epoch": 0.73, "learning_rate": 1.7749603001313898e-06, "loss": 0.7905, "step": 8572 }, { "epoch": 0.73, "learning_rate": 1.7739044150065792e-06, "loss": 0.7621, "step": 8573 }, { "epoch": 0.73, "learning_rate": 1.7728487763070457e-06, "loss": 0.7483, "step": 8574 }, { "epoch": 0.73, "learning_rate": 1.7717933841134243e-06, "loss": 0.8278, "step": 8575 }, { "epoch": 0.73, "learning_rate": 1.7707382385063292e-06, "loss": 0.8008, "step": 8576 }, { "epoch": 0.73, "learning_rate": 1.769683339566361e-06, "loss": 0.8435, "step": 8577 }, { "epoch": 0.73, "learning_rate": 1.7686286873740954e-06, "loss": 0.74, "step": 8578 }, { "epoch": 0.73, "learning_rate": 1.7675742820100961e-06, "loss": 0.7812, "step": 8579 }, { "epoch": 0.73, "learning_rate": 1.7665201235548978e-06, "loss": 0.7804, "step": 8580 }, { "epoch": 0.73, "learning_rate": 1.7654662120890275e-06, "loss": 0.8331, "step": 8581 }, { "epoch": 0.73, "learning_rate": 1.7644125476929863e-06, "loss": 0.7996, "step": 8582 }, { "epoch": 0.73, "learning_rate": 1.7633591304472564e-06, "loss": 0.8302, "step": 8583 }, { "epoch": 0.73, "learning_rate": 1.7623059604323062e-06, "loss": 0.8059, "step": 8584 }, { "epoch": 0.73, "learning_rate": 1.7612530377285813e-06, "loss": 0.806, "step": 8585 }, { "epoch": 0.73, "learning_rate": 1.7602003624165087e-06, "loss": 0.783, "step": 8586 }, { "epoch": 0.73, "learning_rate": 1.7591479345764972e-06, "loss": 0.7347, "step": 8587 }, { "epoch": 0.73, "learning_rate": 1.7580957542889348e-06, "loss": 0.7833, "step": 8588 }, { "epoch": 0.73, "learning_rate": 1.757043821634195e-06, "loss": 0.8329, "step": 8589 }, { "epoch": 0.73, "learning_rate": 1.7559921366926286e-06, "loss": 0.8014, "step": 8590 }, { "epoch": 0.73, "learning_rate": 1.7549406995445673e-06, "loss": 0.7682, "step": 8591 }, { "epoch": 0.73, "learning_rate": 1.7538895102703273e-06, "loss": 0.8256, "step": 8592 }, { "epoch": 0.73, "learning_rate": 1.7528385689502025e-06, "loss": 0.5446, "step": 8593 }, { "epoch": 0.73, "learning_rate": 1.7517878756644697e-06, "loss": 0.8304, "step": 8594 }, { "epoch": 0.73, "learning_rate": 1.750737430493385e-06, "loss": 0.7822, "step": 8595 }, { "epoch": 0.73, "learning_rate": 1.7496872335171862e-06, "loss": 0.7459, "step": 8596 }, { "epoch": 0.73, "learning_rate": 1.7486372848160949e-06, "loss": 0.816, "step": 8597 }, { "epoch": 0.73, "learning_rate": 1.74758758447031e-06, "loss": 0.8455, "step": 8598 }, { "epoch": 0.73, "learning_rate": 1.7465381325600128e-06, "loss": 0.7998, "step": 8599 }, { "epoch": 0.73, "learning_rate": 1.7454889291653658e-06, "loss": 0.7674, "step": 8600 }, { "epoch": 0.73, "learning_rate": 1.7444399743665113e-06, "loss": 0.7931, "step": 8601 }, { "epoch": 0.73, "learning_rate": 1.7433912682435762e-06, "loss": 0.8182, "step": 8602 }, { "epoch": 0.73, "learning_rate": 1.7423428108766644e-06, "loss": 0.5161, "step": 8603 }, { "epoch": 0.73, "learning_rate": 1.7412946023458603e-06, "loss": 0.7978, "step": 8604 }, { "epoch": 0.73, "learning_rate": 1.7402466427312358e-06, "loss": 0.8389, "step": 8605 }, { "epoch": 0.73, "learning_rate": 1.7391989321128372e-06, "loss": 0.8416, "step": 8606 }, { "epoch": 0.73, "learning_rate": 1.7381514705706925e-06, "loss": 0.8238, "step": 8607 }, { "epoch": 0.73, "learning_rate": 1.7371042581848142e-06, "loss": 0.8103, "step": 8608 }, { "epoch": 0.73, "learning_rate": 1.73605729503519e-06, "loss": 0.7881, "step": 8609 }, { "epoch": 0.73, "learning_rate": 1.735010581201797e-06, "loss": 0.7878, "step": 8610 }, { "epoch": 0.73, "learning_rate": 1.7339641167645837e-06, "loss": 0.7892, "step": 8611 }, { "epoch": 0.73, "learning_rate": 1.7329179018034908e-06, "loss": 0.7721, "step": 8612 }, { "epoch": 0.73, "learning_rate": 1.731871936398427e-06, "loss": 0.8264, "step": 8613 }, { "epoch": 0.74, "learning_rate": 1.7308262206292898e-06, "loss": 0.8178, "step": 8614 }, { "epoch": 0.74, "learning_rate": 1.7297807545759583e-06, "loss": 0.8358, "step": 8615 }, { "epoch": 0.74, "learning_rate": 1.7287355383182868e-06, "loss": 0.7824, "step": 8616 }, { "epoch": 0.74, "learning_rate": 1.7276905719361192e-06, "loss": 0.8049, "step": 8617 }, { "epoch": 0.74, "learning_rate": 1.726645855509272e-06, "loss": 0.7611, "step": 8618 }, { "epoch": 0.74, "learning_rate": 1.7256013891175467e-06, "loss": 0.7941, "step": 8619 }, { "epoch": 0.74, "learning_rate": 1.724557172840724e-06, "loss": 0.8003, "step": 8620 }, { "epoch": 0.74, "learning_rate": 1.7235132067585658e-06, "loss": 0.8159, "step": 8621 }, { "epoch": 0.74, "learning_rate": 1.7224694909508177e-06, "loss": 0.5539, "step": 8622 }, { "epoch": 0.74, "learning_rate": 1.7214260254972033e-06, "loss": 0.8501, "step": 8623 }, { "epoch": 0.74, "learning_rate": 1.720382810477425e-06, "loss": 0.7882, "step": 8624 }, { "epoch": 0.74, "learning_rate": 1.719339845971173e-06, "loss": 0.7861, "step": 8625 }, { "epoch": 0.74, "learning_rate": 1.718297132058112e-06, "loss": 0.4851, "step": 8626 }, { "epoch": 0.74, "learning_rate": 1.7172546688178892e-06, "loss": 0.7562, "step": 8627 }, { "epoch": 0.74, "learning_rate": 1.7162124563301347e-06, "loss": 0.7755, "step": 8628 }, { "epoch": 0.74, "learning_rate": 1.7151704946744547e-06, "loss": 0.8197, "step": 8629 }, { "epoch": 0.74, "learning_rate": 1.7141287839304428e-06, "loss": 0.7557, "step": 8630 }, { "epoch": 0.74, "learning_rate": 1.7130873241776691e-06, "loss": 0.834, "step": 8631 }, { "epoch": 0.74, "learning_rate": 1.7120461154956858e-06, "loss": 0.7586, "step": 8632 }, { "epoch": 0.74, "learning_rate": 1.7110051579640247e-06, "loss": 0.8043, "step": 8633 }, { "epoch": 0.74, "learning_rate": 1.7099644516621977e-06, "loss": 0.8097, "step": 8634 }, { "epoch": 0.74, "learning_rate": 1.7089239966697035e-06, "loss": 0.804, "step": 8635 }, { "epoch": 0.74, "learning_rate": 1.7078837930660147e-06, "loss": 0.7841, "step": 8636 }, { "epoch": 0.74, "learning_rate": 1.7068438409305855e-06, "loss": 0.7723, "step": 8637 }, { "epoch": 0.74, "learning_rate": 1.7058041403428565e-06, "loss": 0.8204, "step": 8638 }, { "epoch": 0.74, "learning_rate": 1.704764691382243e-06, "loss": 0.7915, "step": 8639 }, { "epoch": 0.74, "learning_rate": 1.7037254941281438e-06, "loss": 0.7976, "step": 8640 }, { "epoch": 0.74, "learning_rate": 1.7026865486599375e-06, "loss": 0.8244, "step": 8641 }, { "epoch": 0.74, "learning_rate": 1.701647855056983e-06, "loss": 0.8032, "step": 8642 }, { "epoch": 0.74, "learning_rate": 1.7006094133986235e-06, "loss": 0.8575, "step": 8643 }, { "epoch": 0.74, "learning_rate": 1.6995712237641771e-06, "loss": 0.7618, "step": 8644 }, { "epoch": 0.74, "learning_rate": 1.6985332862329495e-06, "loss": 0.828, "step": 8645 }, { "epoch": 0.74, "learning_rate": 1.6974956008842236e-06, "loss": 0.7698, "step": 8646 }, { "epoch": 0.74, "learning_rate": 1.696458167797258e-06, "loss": 0.8285, "step": 8647 }, { "epoch": 0.74, "learning_rate": 1.6954209870513017e-06, "loss": 0.8113, "step": 8648 }, { "epoch": 0.74, "learning_rate": 1.6943840587255767e-06, "loss": 0.7687, "step": 8649 }, { "epoch": 0.74, "learning_rate": 1.6933473828992914e-06, "loss": 0.8062, "step": 8650 }, { "epoch": 0.74, "learning_rate": 1.692310959651632e-06, "loss": 0.7633, "step": 8651 }, { "epoch": 0.74, "learning_rate": 1.6912747890617643e-06, "loss": 0.8765, "step": 8652 }, { "epoch": 0.74, "learning_rate": 1.6902388712088368e-06, "loss": 0.8076, "step": 8653 }, { "epoch": 0.74, "learning_rate": 1.6892032061719765e-06, "loss": 0.8704, "step": 8654 }, { "epoch": 0.74, "learning_rate": 1.6881677940302955e-06, "loss": 0.8195, "step": 8655 }, { "epoch": 0.74, "learning_rate": 1.687132634862883e-06, "loss": 0.8129, "step": 8656 }, { "epoch": 0.74, "learning_rate": 1.6860977287488068e-06, "loss": 0.8051, "step": 8657 }, { "epoch": 0.74, "learning_rate": 1.6850630757671221e-06, "loss": 0.7855, "step": 8658 }, { "epoch": 0.74, "learning_rate": 1.6840286759968593e-06, "loss": 0.783, "step": 8659 }, { "epoch": 0.74, "learning_rate": 1.6829945295170314e-06, "loss": 0.8001, "step": 8660 }, { "epoch": 0.74, "learning_rate": 1.6819606364066303e-06, "loss": 0.5462, "step": 8661 }, { "epoch": 0.74, "learning_rate": 1.68092699674463e-06, "loss": 0.8208, "step": 8662 }, { "epoch": 0.74, "learning_rate": 1.6798936106099873e-06, "loss": 0.8277, "step": 8663 }, { "epoch": 0.74, "learning_rate": 1.6788604780816353e-06, "loss": 0.7875, "step": 8664 }, { "epoch": 0.74, "learning_rate": 1.677827599238489e-06, "loss": 0.7887, "step": 8665 }, { "epoch": 0.74, "learning_rate": 1.6767949741594503e-06, "loss": 0.7746, "step": 8666 }, { "epoch": 0.74, "learning_rate": 1.675762602923388e-06, "loss": 0.8061, "step": 8667 }, { "epoch": 0.74, "learning_rate": 1.6747304856091662e-06, "loss": 0.764, "step": 8668 }, { "epoch": 0.74, "learning_rate": 1.6736986222956209e-06, "loss": 0.7938, "step": 8669 }, { "epoch": 0.74, "learning_rate": 1.672667013061569e-06, "loss": 0.7264, "step": 8670 }, { "epoch": 0.74, "learning_rate": 1.6716356579858145e-06, "loss": 0.7984, "step": 8671 }, { "epoch": 0.74, "learning_rate": 1.6706045571471341e-06, "loss": 0.8492, "step": 8672 }, { "epoch": 0.74, "learning_rate": 1.6695737106242898e-06, "loss": 0.7754, "step": 8673 }, { "epoch": 0.74, "learning_rate": 1.6685431184960222e-06, "loss": 0.8428, "step": 8674 }, { "epoch": 0.74, "learning_rate": 1.6675127808410513e-06, "loss": 0.7951, "step": 8675 }, { "epoch": 0.74, "learning_rate": 1.6664826977380828e-06, "loss": 0.4527, "step": 8676 }, { "epoch": 0.74, "learning_rate": 1.6654528692657983e-06, "loss": 0.7976, "step": 8677 }, { "epoch": 0.74, "learning_rate": 1.6644232955028593e-06, "loss": 0.5005, "step": 8678 }, { "epoch": 0.74, "learning_rate": 1.663393976527915e-06, "loss": 0.7799, "step": 8679 }, { "epoch": 0.74, "learning_rate": 1.6623649124195828e-06, "loss": 0.8102, "step": 8680 }, { "epoch": 0.74, "learning_rate": 1.661336103256473e-06, "loss": 0.8001, "step": 8681 }, { "epoch": 0.74, "learning_rate": 1.6603075491171677e-06, "loss": 0.7644, "step": 8682 }, { "epoch": 0.74, "learning_rate": 1.6592792500802367e-06, "loss": 0.8614, "step": 8683 }, { "epoch": 0.74, "learning_rate": 1.6582512062242252e-06, "loss": 0.7879, "step": 8684 }, { "epoch": 0.74, "learning_rate": 1.6572234176276592e-06, "loss": 0.8086, "step": 8685 }, { "epoch": 0.74, "learning_rate": 1.6561958843690474e-06, "loss": 0.849, "step": 8686 }, { "epoch": 0.74, "learning_rate": 1.6551686065268758e-06, "loss": 0.826, "step": 8687 }, { "epoch": 0.74, "learning_rate": 1.654141584179616e-06, "loss": 0.7752, "step": 8688 }, { "epoch": 0.74, "learning_rate": 1.653114817405716e-06, "loss": 0.8166, "step": 8689 }, { "epoch": 0.74, "learning_rate": 1.6520883062836035e-06, "loss": 0.8109, "step": 8690 }, { "epoch": 0.74, "learning_rate": 1.6510620508916918e-06, "loss": 0.5372, "step": 8691 }, { "epoch": 0.74, "learning_rate": 1.6500360513083696e-06, "loss": 0.771, "step": 8692 }, { "epoch": 0.74, "learning_rate": 1.6490103076120084e-06, "loss": 0.773, "step": 8693 }, { "epoch": 0.74, "learning_rate": 1.6479848198809594e-06, "loss": 0.8426, "step": 8694 }, { "epoch": 0.74, "learning_rate": 1.6469595881935523e-06, "loss": 0.7795, "step": 8695 }, { "epoch": 0.74, "learning_rate": 1.6459346126281035e-06, "loss": 0.7863, "step": 8696 }, { "epoch": 0.74, "learning_rate": 1.6449098932629037e-06, "loss": 0.7954, "step": 8697 }, { "epoch": 0.74, "learning_rate": 1.643885430176224e-06, "loss": 0.772, "step": 8698 }, { "epoch": 0.74, "learning_rate": 1.6428612234463243e-06, "loss": 0.8088, "step": 8699 }, { "epoch": 0.74, "learning_rate": 1.6418372731514303e-06, "loss": 0.8401, "step": 8700 }, { "epoch": 0.74, "learning_rate": 1.6408135793697628e-06, "loss": 0.541, "step": 8701 }, { "epoch": 0.74, "learning_rate": 1.6397901421795138e-06, "loss": 0.7575, "step": 8702 }, { "epoch": 0.74, "learning_rate": 1.6387669616588576e-06, "loss": 0.7951, "step": 8703 }, { "epoch": 0.74, "learning_rate": 1.6377440378859532e-06, "loss": 0.7248, "step": 8704 }, { "epoch": 0.74, "learning_rate": 1.6367213709389351e-06, "loss": 0.7767, "step": 8705 }, { "epoch": 0.74, "learning_rate": 1.635698960895919e-06, "loss": 0.8079, "step": 8706 }, { "epoch": 0.74, "learning_rate": 1.6346768078350024e-06, "loss": 0.7851, "step": 8707 }, { "epoch": 0.74, "learning_rate": 1.6336549118342598e-06, "loss": 0.8096, "step": 8708 }, { "epoch": 0.74, "learning_rate": 1.632633272971753e-06, "loss": 0.8022, "step": 8709 }, { "epoch": 0.74, "learning_rate": 1.631611891325518e-06, "loss": 0.8299, "step": 8710 }, { "epoch": 0.74, "learning_rate": 1.6305907669735715e-06, "loss": 0.8152, "step": 8711 }, { "epoch": 0.74, "learning_rate": 1.629569899993917e-06, "loss": 0.7955, "step": 8712 }, { "epoch": 0.74, "learning_rate": 1.6285492904645262e-06, "loss": 0.4878, "step": 8713 }, { "epoch": 0.74, "learning_rate": 1.6275289384633635e-06, "loss": 0.8076, "step": 8714 }, { "epoch": 0.74, "learning_rate": 1.6265088440683673e-06, "loss": 0.765, "step": 8715 }, { "epoch": 0.74, "learning_rate": 1.625489007357456e-06, "loss": 0.7777, "step": 8716 }, { "epoch": 0.74, "learning_rate": 1.6244694284085322e-06, "loss": 0.7179, "step": 8717 }, { "epoch": 0.74, "learning_rate": 1.6234501072994735e-06, "loss": 0.5399, "step": 8718 }, { "epoch": 0.74, "learning_rate": 1.6224310441081465e-06, "loss": 0.8061, "step": 8719 }, { "epoch": 0.74, "learning_rate": 1.621412238912386e-06, "loss": 0.7904, "step": 8720 }, { "epoch": 0.74, "learning_rate": 1.6203936917900144e-06, "loss": 0.7983, "step": 8721 }, { "epoch": 0.74, "learning_rate": 1.6193754028188363e-06, "loss": 0.7446, "step": 8722 }, { "epoch": 0.74, "learning_rate": 1.618357372076631e-06, "loss": 0.8131, "step": 8723 }, { "epoch": 0.74, "learning_rate": 1.6173395996411634e-06, "loss": 0.7743, "step": 8724 }, { "epoch": 0.74, "learning_rate": 1.6163220855901745e-06, "loss": 0.8457, "step": 8725 }, { "epoch": 0.74, "learning_rate": 1.6153048300013874e-06, "loss": 0.4852, "step": 8726 }, { "epoch": 0.74, "learning_rate": 1.6142878329525047e-06, "loss": 0.7836, "step": 8727 }, { "epoch": 0.74, "learning_rate": 1.6132710945212082e-06, "loss": 0.7995, "step": 8728 }, { "epoch": 0.74, "learning_rate": 1.6122546147851647e-06, "loss": 0.795, "step": 8729 }, { "epoch": 0.74, "learning_rate": 1.6112383938220173e-06, "loss": 0.7737, "step": 8730 }, { "epoch": 0.75, "learning_rate": 1.6102224317093867e-06, "loss": 0.8004, "step": 8731 }, { "epoch": 0.75, "learning_rate": 1.6092067285248835e-06, "loss": 0.8122, "step": 8732 }, { "epoch": 0.75, "learning_rate": 1.608191284346085e-06, "loss": 0.8191, "step": 8733 }, { "epoch": 0.75, "learning_rate": 1.6071760992505609e-06, "loss": 0.8029, "step": 8734 }, { "epoch": 0.75, "learning_rate": 1.6061611733158544e-06, "loss": 0.7382, "step": 8735 }, { "epoch": 0.75, "learning_rate": 1.6051465066194894e-06, "loss": 0.8044, "step": 8736 }, { "epoch": 0.75, "learning_rate": 1.6041320992389737e-06, "loss": 0.7763, "step": 8737 }, { "epoch": 0.75, "learning_rate": 1.6031179512517913e-06, "loss": 0.7632, "step": 8738 }, { "epoch": 0.75, "learning_rate": 1.6021040627354084e-06, "loss": 0.7952, "step": 8739 }, { "epoch": 0.75, "learning_rate": 1.6010904337672705e-06, "loss": 0.5526, "step": 8740 }, { "epoch": 0.75, "learning_rate": 1.6000770644248026e-06, "loss": 0.77, "step": 8741 }, { "epoch": 0.75, "learning_rate": 1.5990639547854132e-06, "loss": 0.8038, "step": 8742 }, { "epoch": 0.75, "learning_rate": 1.598051104926488e-06, "loss": 0.7673, "step": 8743 }, { "epoch": 0.75, "learning_rate": 1.5970385149253908e-06, "loss": 0.7862, "step": 8744 }, { "epoch": 0.75, "learning_rate": 1.5960261848594743e-06, "loss": 0.8193, "step": 8745 }, { "epoch": 0.75, "learning_rate": 1.5950141148060583e-06, "loss": 0.8482, "step": 8746 }, { "epoch": 0.75, "learning_rate": 1.594002304842454e-06, "loss": 0.8136, "step": 8747 }, { "epoch": 0.75, "learning_rate": 1.5929907550459484e-06, "loss": 0.787, "step": 8748 }, { "epoch": 0.75, "learning_rate": 1.591979465493806e-06, "loss": 0.8012, "step": 8749 }, { "epoch": 0.75, "learning_rate": 1.590968436263277e-06, "loss": 0.8574, "step": 8750 }, { "epoch": 0.75, "learning_rate": 1.5899576674315865e-06, "loss": 0.819, "step": 8751 }, { "epoch": 0.75, "learning_rate": 1.588947159075947e-06, "loss": 0.8229, "step": 8752 }, { "epoch": 0.75, "learning_rate": 1.5879369112735404e-06, "loss": 0.5216, "step": 8753 }, { "epoch": 0.75, "learning_rate": 1.586926924101535e-06, "loss": 0.7937, "step": 8754 }, { "epoch": 0.75, "learning_rate": 1.5859171976370824e-06, "loss": 0.7822, "step": 8755 }, { "epoch": 0.75, "learning_rate": 1.5849077319573063e-06, "loss": 0.8062, "step": 8756 }, { "epoch": 0.75, "learning_rate": 1.5838985271393188e-06, "loss": 0.8391, "step": 8757 }, { "epoch": 0.75, "learning_rate": 1.5828895832602076e-06, "loss": 0.7626, "step": 8758 }, { "epoch": 0.75, "learning_rate": 1.5818809003970365e-06, "loss": 0.8182, "step": 8759 }, { "epoch": 0.75, "learning_rate": 1.5808724786268576e-06, "loss": 0.7889, "step": 8760 }, { "epoch": 0.75, "learning_rate": 1.5798643180266964e-06, "loss": 0.7904, "step": 8761 }, { "epoch": 0.75, "learning_rate": 1.578856418673565e-06, "loss": 0.8178, "step": 8762 }, { "epoch": 0.75, "learning_rate": 1.5778487806444499e-06, "loss": 0.8081, "step": 8763 }, { "epoch": 0.75, "learning_rate": 1.5768414040163172e-06, "loss": 0.8406, "step": 8764 }, { "epoch": 0.75, "learning_rate": 1.575834288866121e-06, "loss": 0.805, "step": 8765 }, { "epoch": 0.75, "learning_rate": 1.5748274352707831e-06, "loss": 0.7996, "step": 8766 }, { "epoch": 0.75, "learning_rate": 1.5738208433072171e-06, "loss": 0.8305, "step": 8767 }, { "epoch": 0.75, "learning_rate": 1.5728145130523092e-06, "loss": 0.832, "step": 8768 }, { "epoch": 0.75, "learning_rate": 1.5718084445829273e-06, "loss": 0.8199, "step": 8769 }, { "epoch": 0.75, "learning_rate": 1.5708026379759223e-06, "loss": 0.7512, "step": 8770 }, { "epoch": 0.75, "learning_rate": 1.5697970933081213e-06, "loss": 0.7796, "step": 8771 }, { "epoch": 0.75, "learning_rate": 1.5687918106563326e-06, "loss": 0.817, "step": 8772 }, { "epoch": 0.75, "learning_rate": 1.5677867900973459e-06, "loss": 0.7728, "step": 8773 }, { "epoch": 0.75, "learning_rate": 1.566782031707927e-06, "loss": 0.8093, "step": 8774 }, { "epoch": 0.75, "learning_rate": 1.565777535564827e-06, "loss": 0.8378, "step": 8775 }, { "epoch": 0.75, "learning_rate": 1.5647733017447741e-06, "loss": 0.8187, "step": 8776 }, { "epoch": 0.75, "learning_rate": 1.5637693303244745e-06, "loss": 0.8175, "step": 8777 }, { "epoch": 0.75, "learning_rate": 1.5627656213806215e-06, "loss": 0.759, "step": 8778 }, { "epoch": 0.75, "learning_rate": 1.5617621749898764e-06, "loss": 0.8375, "step": 8779 }, { "epoch": 0.75, "learning_rate": 1.5607589912288928e-06, "loss": 0.7906, "step": 8780 }, { "epoch": 0.75, "learning_rate": 1.5597560701742975e-06, "loss": 0.8426, "step": 8781 }, { "epoch": 0.75, "learning_rate": 1.5587534119026964e-06, "loss": 0.8102, "step": 8782 }, { "epoch": 0.75, "learning_rate": 1.5577510164906812e-06, "loss": 0.8095, "step": 8783 }, { "epoch": 0.75, "learning_rate": 1.556748884014816e-06, "loss": 0.8313, "step": 8784 }, { "epoch": 0.75, "learning_rate": 1.5557470145516546e-06, "loss": 0.7789, "step": 8785 }, { "epoch": 0.75, "learning_rate": 1.5547454081777186e-06, "loss": 0.7781, "step": 8786 }, { "epoch": 0.75, "learning_rate": 1.5537440649695169e-06, "loss": 0.8288, "step": 8787 }, { "epoch": 0.75, "learning_rate": 1.5527429850035397e-06, "loss": 0.8168, "step": 8788 }, { "epoch": 0.75, "learning_rate": 1.5517421683562517e-06, "loss": 0.824, "step": 8789 }, { "epoch": 0.75, "learning_rate": 1.5507416151041032e-06, "loss": 0.8324, "step": 8790 }, { "epoch": 0.75, "learning_rate": 1.5497413253235194e-06, "loss": 0.8106, "step": 8791 }, { "epoch": 0.75, "learning_rate": 1.548741299090908e-06, "loss": 0.8186, "step": 8792 }, { "epoch": 0.75, "learning_rate": 1.5477415364826558e-06, "loss": 0.8356, "step": 8793 }, { "epoch": 0.75, "learning_rate": 1.5467420375751285e-06, "loss": 0.7863, "step": 8794 }, { "epoch": 0.75, "learning_rate": 1.5457428024446753e-06, "loss": 0.8311, "step": 8795 }, { "epoch": 0.75, "learning_rate": 1.5447438311676215e-06, "loss": 0.8376, "step": 8796 }, { "epoch": 0.75, "learning_rate": 1.5437451238202722e-06, "loss": 0.7926, "step": 8797 }, { "epoch": 0.75, "learning_rate": 1.5427466804789177e-06, "loss": 0.7864, "step": 8798 }, { "epoch": 0.75, "learning_rate": 1.5417485012198185e-06, "loss": 0.8133, "step": 8799 }, { "epoch": 0.75, "learning_rate": 1.5407505861192246e-06, "loss": 0.8402, "step": 8800 }, { "epoch": 0.75, "learning_rate": 1.5397529352533608e-06, "loss": 0.7839, "step": 8801 }, { "epoch": 0.75, "learning_rate": 1.5387555486984302e-06, "loss": 0.8008, "step": 8802 }, { "epoch": 0.75, "learning_rate": 1.5377584265306222e-06, "loss": 0.8002, "step": 8803 }, { "epoch": 0.75, "learning_rate": 1.5367615688261e-06, "loss": 0.7944, "step": 8804 }, { "epoch": 0.75, "learning_rate": 1.5357649756610087e-06, "loss": 0.8003, "step": 8805 }, { "epoch": 0.75, "learning_rate": 1.5347686471114731e-06, "loss": 0.7746, "step": 8806 }, { "epoch": 0.75, "learning_rate": 1.533772583253596e-06, "loss": 0.811, "step": 8807 }, { "epoch": 0.75, "learning_rate": 1.5327767841634649e-06, "loss": 0.8024, "step": 8808 }, { "epoch": 0.75, "learning_rate": 1.531781249917142e-06, "loss": 0.7603, "step": 8809 }, { "epoch": 0.75, "learning_rate": 1.5307859805906706e-06, "loss": 0.7962, "step": 8810 }, { "epoch": 0.75, "learning_rate": 1.5297909762600777e-06, "loss": 0.7862, "step": 8811 }, { "epoch": 0.75, "learning_rate": 1.5287962370013615e-06, "loss": 0.7564, "step": 8812 }, { "epoch": 0.75, "learning_rate": 1.5278017628905095e-06, "loss": 0.8315, "step": 8813 }, { "epoch": 0.75, "learning_rate": 1.5268075540034832e-06, "loss": 0.8091, "step": 8814 }, { "epoch": 0.75, "learning_rate": 1.525813610416223e-06, "loss": 0.8022, "step": 8815 }, { "epoch": 0.75, "learning_rate": 1.5248199322046553e-06, "loss": 0.7664, "step": 8816 }, { "epoch": 0.75, "learning_rate": 1.5238265194446799e-06, "loss": 0.8029, "step": 8817 }, { "epoch": 0.75, "learning_rate": 1.522833372212179e-06, "loss": 0.7579, "step": 8818 }, { "epoch": 0.75, "learning_rate": 1.5218404905830142e-06, "loss": 0.7925, "step": 8819 }, { "epoch": 0.75, "learning_rate": 1.5208478746330247e-06, "loss": 0.7745, "step": 8820 }, { "epoch": 0.75, "learning_rate": 1.5198555244380353e-06, "loss": 0.758, "step": 8821 }, { "epoch": 0.75, "learning_rate": 1.5188634400738428e-06, "loss": 0.8178, "step": 8822 }, { "epoch": 0.75, "learning_rate": 1.5178716216162315e-06, "loss": 0.7863, "step": 8823 }, { "epoch": 0.75, "learning_rate": 1.516880069140959e-06, "loss": 0.7686, "step": 8824 }, { "epoch": 0.75, "learning_rate": 1.5158887827237657e-06, "loss": 0.7932, "step": 8825 }, { "epoch": 0.75, "learning_rate": 1.514897762440371e-06, "loss": 0.7627, "step": 8826 }, { "epoch": 0.75, "learning_rate": 1.513907008366472e-06, "loss": 0.7847, "step": 8827 }, { "epoch": 0.75, "learning_rate": 1.5129165205777507e-06, "loss": 0.8022, "step": 8828 }, { "epoch": 0.75, "learning_rate": 1.511926299149864e-06, "loss": 0.7793, "step": 8829 }, { "epoch": 0.75, "learning_rate": 1.510936344158448e-06, "loss": 0.8042, "step": 8830 }, { "epoch": 0.75, "learning_rate": 1.509946655679126e-06, "loss": 0.766, "step": 8831 }, { "epoch": 0.75, "learning_rate": 1.508957233787488e-06, "loss": 0.8521, "step": 8832 }, { "epoch": 0.75, "learning_rate": 1.5079680785591167e-06, "loss": 0.7897, "step": 8833 }, { "epoch": 0.75, "learning_rate": 1.5069791900695662e-06, "loss": 0.8301, "step": 8834 }, { "epoch": 0.75, "learning_rate": 1.5059905683943721e-06, "loss": 0.8235, "step": 8835 }, { "epoch": 0.75, "learning_rate": 1.505002213609053e-06, "loss": 0.8232, "step": 8836 }, { "epoch": 0.75, "learning_rate": 1.5040141257891023e-06, "loss": 0.7474, "step": 8837 }, { "epoch": 0.75, "learning_rate": 1.5030263050099957e-06, "loss": 0.8009, "step": 8838 }, { "epoch": 0.75, "learning_rate": 1.5020387513471878e-06, "loss": 0.7551, "step": 8839 }, { "epoch": 0.75, "learning_rate": 1.501051464876111e-06, "loss": 0.7761, "step": 8840 }, { "epoch": 0.75, "learning_rate": 1.5000644456721824e-06, "loss": 0.8327, "step": 8841 }, { "epoch": 0.75, "learning_rate": 1.4990776938107942e-06, "loss": 0.7812, "step": 8842 }, { "epoch": 0.75, "learning_rate": 1.4980912093673178e-06, "loss": 0.7985, "step": 8843 }, { "epoch": 0.75, "learning_rate": 1.4971049924171088e-06, "loss": 0.8556, "step": 8844 }, { "epoch": 0.75, "learning_rate": 1.4961190430354971e-06, "loss": 0.7848, "step": 8845 }, { "epoch": 0.75, "learning_rate": 1.495133361297796e-06, "loss": 0.8286, "step": 8846 }, { "epoch": 0.75, "learning_rate": 1.4941479472792953e-06, "loss": 0.8038, "step": 8847 }, { "epoch": 0.76, "learning_rate": 1.493162801055265e-06, "loss": 0.8092, "step": 8848 }, { "epoch": 0.76, "learning_rate": 1.4921779227009586e-06, "loss": 0.8196, "step": 8849 }, { "epoch": 0.76, "learning_rate": 1.4911933122916045e-06, "loss": 0.7777, "step": 8850 }, { "epoch": 0.76, "learning_rate": 1.490208969902412e-06, "loss": 0.7424, "step": 8851 }, { "epoch": 0.76, "learning_rate": 1.4892248956085697e-06, "loss": 0.7723, "step": 8852 }, { "epoch": 0.76, "learning_rate": 1.4882410894852451e-06, "loss": 0.8101, "step": 8853 }, { "epoch": 0.76, "learning_rate": 1.4872575516075892e-06, "loss": 0.7333, "step": 8854 }, { "epoch": 0.76, "learning_rate": 1.4862742820507281e-06, "loss": 0.7615, "step": 8855 }, { "epoch": 0.76, "learning_rate": 1.4852912808897667e-06, "loss": 0.8562, "step": 8856 }, { "epoch": 0.76, "learning_rate": 1.484308548199796e-06, "loss": 0.8031, "step": 8857 }, { "epoch": 0.76, "learning_rate": 1.4833260840558788e-06, "loss": 0.8068, "step": 8858 }, { "epoch": 0.76, "learning_rate": 1.4823438885330615e-06, "loss": 0.8347, "step": 8859 }, { "epoch": 0.76, "learning_rate": 1.481361961706369e-06, "loss": 0.7876, "step": 8860 }, { "epoch": 0.76, "learning_rate": 1.480380303650804e-06, "loss": 0.8097, "step": 8861 }, { "epoch": 0.76, "learning_rate": 1.479398914441354e-06, "loss": 0.8229, "step": 8862 }, { "epoch": 0.76, "learning_rate": 1.4784177941529787e-06, "loss": 0.7628, "step": 8863 }, { "epoch": 0.76, "learning_rate": 1.4774369428606261e-06, "loss": 0.7684, "step": 8864 }, { "epoch": 0.76, "learning_rate": 1.4764563606392118e-06, "loss": 0.7882, "step": 8865 }, { "epoch": 0.76, "learning_rate": 1.4754760475636432e-06, "loss": 0.7851, "step": 8866 }, { "epoch": 0.76, "learning_rate": 1.4744960037087985e-06, "loss": 0.7884, "step": 8867 }, { "epoch": 0.76, "learning_rate": 1.473516229149538e-06, "loss": 0.733, "step": 8868 }, { "epoch": 0.76, "learning_rate": 1.4725367239607048e-06, "loss": 0.8245, "step": 8869 }, { "epoch": 0.76, "learning_rate": 1.4715574882171163e-06, "loss": 0.7949, "step": 8870 }, { "epoch": 0.76, "learning_rate": 1.4705785219935714e-06, "loss": 0.8033, "step": 8871 }, { "epoch": 0.76, "learning_rate": 1.469599825364849e-06, "loss": 0.7689, "step": 8872 }, { "epoch": 0.76, "learning_rate": 1.4686213984057046e-06, "loss": 0.5003, "step": 8873 }, { "epoch": 0.76, "learning_rate": 1.4676432411908793e-06, "loss": 0.8169, "step": 8874 }, { "epoch": 0.76, "learning_rate": 1.466665353795087e-06, "loss": 0.8278, "step": 8875 }, { "epoch": 0.76, "learning_rate": 1.4656877362930228e-06, "loss": 0.7806, "step": 8876 }, { "epoch": 0.76, "learning_rate": 1.4647103887593645e-06, "loss": 0.7921, "step": 8877 }, { "epoch": 0.76, "learning_rate": 1.4637333112687662e-06, "loss": 0.7625, "step": 8878 }, { "epoch": 0.76, "learning_rate": 1.4627565038958613e-06, "loss": 0.7771, "step": 8879 }, { "epoch": 0.76, "learning_rate": 1.4617799667152633e-06, "loss": 0.5013, "step": 8880 }, { "epoch": 0.76, "learning_rate": 1.460803699801563e-06, "loss": 0.814, "step": 8881 }, { "epoch": 0.76, "learning_rate": 1.4598277032293367e-06, "loss": 0.7777, "step": 8882 }, { "epoch": 0.76, "learning_rate": 1.4588519770731336e-06, "loss": 0.8195, "step": 8883 }, { "epoch": 0.76, "learning_rate": 1.4578765214074842e-06, "loss": 0.8082, "step": 8884 }, { "epoch": 0.76, "learning_rate": 1.4569013363068995e-06, "loss": 0.8115, "step": 8885 }, { "epoch": 0.76, "learning_rate": 1.455926421845867e-06, "loss": 0.7864, "step": 8886 }, { "epoch": 0.76, "learning_rate": 1.4549517780988587e-06, "loss": 0.8167, "step": 8887 }, { "epoch": 0.76, "learning_rate": 1.4539774051403217e-06, "loss": 0.8139, "step": 8888 }, { "epoch": 0.76, "learning_rate": 1.4530033030446805e-06, "loss": 0.8187, "step": 8889 }, { "epoch": 0.76, "learning_rate": 1.4520294718863464e-06, "loss": 0.8076, "step": 8890 }, { "epoch": 0.76, "learning_rate": 1.451055911739704e-06, "loss": 0.8046, "step": 8891 }, { "epoch": 0.76, "learning_rate": 1.4500826226791175e-06, "loss": 0.7942, "step": 8892 }, { "epoch": 0.76, "learning_rate": 1.449109604778932e-06, "loss": 0.8119, "step": 8893 }, { "epoch": 0.76, "learning_rate": 1.4481368581134703e-06, "loss": 0.7386, "step": 8894 }, { "epoch": 0.76, "learning_rate": 1.4471643827570387e-06, "loss": 0.7911, "step": 8895 }, { "epoch": 0.76, "learning_rate": 1.4461921787839167e-06, "loss": 0.8144, "step": 8896 }, { "epoch": 0.76, "learning_rate": 1.4452202462683685e-06, "loss": 0.7971, "step": 8897 }, { "epoch": 0.76, "learning_rate": 1.4442485852846361e-06, "loss": 0.8207, "step": 8898 }, { "epoch": 0.76, "learning_rate": 1.4432771959069343e-06, "loss": 0.7733, "step": 8899 }, { "epoch": 0.76, "learning_rate": 1.4423060782094678e-06, "loss": 0.8022, "step": 8900 }, { "epoch": 0.76, "learning_rate": 1.441335232266412e-06, "loss": 0.8058, "step": 8901 }, { "epoch": 0.76, "learning_rate": 1.4403646581519286e-06, "loss": 0.8266, "step": 8902 }, { "epoch": 0.76, "learning_rate": 1.4393943559401529e-06, "loss": 0.7814, "step": 8903 }, { "epoch": 0.76, "learning_rate": 1.4384243257052016e-06, "loss": 0.7601, "step": 8904 }, { "epoch": 0.76, "learning_rate": 1.4374545675211703e-06, "loss": 0.7889, "step": 8905 }, { "epoch": 0.76, "learning_rate": 1.4364850814621322e-06, "loss": 0.7994, "step": 8906 }, { "epoch": 0.76, "learning_rate": 1.4355158676021457e-06, "loss": 0.8205, "step": 8907 }, { "epoch": 0.76, "learning_rate": 1.434546926015241e-06, "loss": 0.551, "step": 8908 }, { "epoch": 0.76, "learning_rate": 1.4335782567754303e-06, "loss": 0.8484, "step": 8909 }, { "epoch": 0.76, "learning_rate": 1.4326098599567084e-06, "loss": 0.7667, "step": 8910 }, { "epoch": 0.76, "learning_rate": 1.4316417356330441e-06, "loss": 0.7754, "step": 8911 }, { "epoch": 0.76, "learning_rate": 1.4306738838783886e-06, "loss": 0.771, "step": 8912 }, { "epoch": 0.76, "learning_rate": 1.42970630476667e-06, "loss": 0.8051, "step": 8913 }, { "epoch": 0.76, "learning_rate": 1.4287389983717958e-06, "loss": 0.8369, "step": 8914 }, { "epoch": 0.76, "learning_rate": 1.4277719647676576e-06, "loss": 0.7622, "step": 8915 }, { "epoch": 0.76, "learning_rate": 1.4268052040281195e-06, "loss": 0.8158, "step": 8916 }, { "epoch": 0.76, "learning_rate": 1.425838716227026e-06, "loss": 0.821, "step": 8917 }, { "epoch": 0.76, "learning_rate": 1.4248725014382076e-06, "loss": 0.7818, "step": 8918 }, { "epoch": 0.76, "learning_rate": 1.4239065597354624e-06, "loss": 0.7621, "step": 8919 }, { "epoch": 0.76, "learning_rate": 1.422940891192578e-06, "loss": 0.7972, "step": 8920 }, { "epoch": 0.76, "learning_rate": 1.4219754958833159e-06, "loss": 0.7694, "step": 8921 }, { "epoch": 0.76, "learning_rate": 1.4210103738814162e-06, "loss": 0.5189, "step": 8922 }, { "epoch": 0.76, "learning_rate": 1.4200455252606022e-06, "loss": 0.7414, "step": 8923 }, { "epoch": 0.76, "learning_rate": 1.419080950094573e-06, "loss": 0.7718, "step": 8924 }, { "epoch": 0.76, "learning_rate": 1.4181166484570074e-06, "loss": 0.8092, "step": 8925 }, { "epoch": 0.76, "learning_rate": 1.4171526204215635e-06, "loss": 0.7996, "step": 8926 }, { "epoch": 0.76, "learning_rate": 1.4161888660618772e-06, "loss": 0.8652, "step": 8927 }, { "epoch": 0.76, "learning_rate": 1.415225385451568e-06, "loss": 0.815, "step": 8928 }, { "epoch": 0.76, "learning_rate": 1.414262178664228e-06, "loss": 0.8356, "step": 8929 }, { "epoch": 0.76, "learning_rate": 1.413299245773435e-06, "loss": 0.8204, "step": 8930 }, { "epoch": 0.76, "learning_rate": 1.4123365868527428e-06, "loss": 0.7618, "step": 8931 }, { "epoch": 0.76, "learning_rate": 1.4113742019756787e-06, "loss": 0.8051, "step": 8932 }, { "epoch": 0.76, "learning_rate": 1.4104120912157605e-06, "loss": 0.8062, "step": 8933 }, { "epoch": 0.76, "learning_rate": 1.4094502546464744e-06, "loss": 0.8011, "step": 8934 }, { "epoch": 0.76, "learning_rate": 1.4084886923412944e-06, "loss": 0.8331, "step": 8935 }, { "epoch": 0.76, "learning_rate": 1.407527404373667e-06, "loss": 0.8018, "step": 8936 }, { "epoch": 0.76, "learning_rate": 1.4065663908170214e-06, "loss": 0.7856, "step": 8937 }, { "epoch": 0.76, "learning_rate": 1.4056056517447637e-06, "loss": 0.7986, "step": 8938 }, { "epoch": 0.76, "learning_rate": 1.404645187230278e-06, "loss": 0.7627, "step": 8939 }, { "epoch": 0.76, "learning_rate": 1.403684997346934e-06, "loss": 0.8338, "step": 8940 }, { "epoch": 0.76, "learning_rate": 1.4027250821680726e-06, "loss": 0.7923, "step": 8941 }, { "epoch": 0.76, "learning_rate": 1.4017654417670163e-06, "loss": 0.8586, "step": 8942 }, { "epoch": 0.76, "learning_rate": 1.4008060762170706e-06, "loss": 0.8379, "step": 8943 }, { "epoch": 0.76, "learning_rate": 1.3998469855915136e-06, "loss": 0.7784, "step": 8944 }, { "epoch": 0.76, "learning_rate": 1.398888169963607e-06, "loss": 0.7904, "step": 8945 }, { "epoch": 0.76, "learning_rate": 1.3979296294065898e-06, "loss": 0.8259, "step": 8946 }, { "epoch": 0.76, "learning_rate": 1.3969713639936771e-06, "loss": 0.7957, "step": 8947 }, { "epoch": 0.76, "learning_rate": 1.3960133737980708e-06, "loss": 0.7859, "step": 8948 }, { "epoch": 0.76, "learning_rate": 1.3950556588929443e-06, "loss": 0.8186, "step": 8949 }, { "epoch": 0.76, "learning_rate": 1.3940982193514519e-06, "loss": 0.8385, "step": 8950 }, { "epoch": 0.76, "learning_rate": 1.3931410552467312e-06, "loss": 0.7459, "step": 8951 }, { "epoch": 0.76, "learning_rate": 1.3921841666518905e-06, "loss": 0.7985, "step": 8952 }, { "epoch": 0.76, "learning_rate": 1.3912275536400249e-06, "loss": 0.8265, "step": 8953 }, { "epoch": 0.76, "learning_rate": 1.3902712162842042e-06, "loss": 0.8306, "step": 8954 }, { "epoch": 0.76, "learning_rate": 1.3893151546574768e-06, "loss": 0.7665, "step": 8955 }, { "epoch": 0.76, "learning_rate": 1.3883593688328751e-06, "loss": 0.538, "step": 8956 }, { "epoch": 0.76, "learning_rate": 1.3874038588834044e-06, "loss": 0.8045, "step": 8957 }, { "epoch": 0.76, "learning_rate": 1.3864486248820514e-06, "loss": 0.784, "step": 8958 }, { "epoch": 0.76, "learning_rate": 1.3854936669017822e-06, "loss": 0.7947, "step": 8959 }, { "epoch": 0.76, "learning_rate": 1.384538985015539e-06, "loss": 0.7317, "step": 8960 }, { "epoch": 0.76, "learning_rate": 1.3835845792962488e-06, "loss": 0.7197, "step": 8961 }, { "epoch": 0.76, "learning_rate": 1.382630449816812e-06, "loss": 0.7767, "step": 8962 }, { "epoch": 0.76, "learning_rate": 1.3816765966501089e-06, "loss": 0.7922, "step": 8963 }, { "epoch": 0.76, "learning_rate": 1.380723019869003e-06, "loss": 0.7635, "step": 8964 }, { "epoch": 0.76, "learning_rate": 1.3797697195463278e-06, "loss": 0.8248, "step": 8965 }, { "epoch": 0.77, "learning_rate": 1.378816695754906e-06, "loss": 0.7735, "step": 8966 }, { "epoch": 0.77, "learning_rate": 1.377863948567531e-06, "loss": 0.8464, "step": 8967 }, { "epoch": 0.77, "learning_rate": 1.376911478056981e-06, "loss": 0.8082, "step": 8968 }, { "epoch": 0.77, "learning_rate": 1.3759592842960096e-06, "loss": 0.8132, "step": 8969 }, { "epoch": 0.77, "learning_rate": 1.3750073673573483e-06, "loss": 0.7823, "step": 8970 }, { "epoch": 0.77, "learning_rate": 1.374055727313714e-06, "loss": 0.806, "step": 8971 }, { "epoch": 0.77, "learning_rate": 1.3731043642377911e-06, "loss": 0.824, "step": 8972 }, { "epoch": 0.77, "learning_rate": 1.372153278202254e-06, "loss": 0.765, "step": 8973 }, { "epoch": 0.77, "learning_rate": 1.371202469279751e-06, "loss": 0.8117, "step": 8974 }, { "epoch": 0.77, "learning_rate": 1.370251937542907e-06, "loss": 0.7934, "step": 8975 }, { "epoch": 0.77, "learning_rate": 1.369301683064332e-06, "loss": 0.763, "step": 8976 }, { "epoch": 0.77, "learning_rate": 1.368351705916609e-06, "loss": 0.796, "step": 8977 }, { "epoch": 0.77, "learning_rate": 1.3674020061723025e-06, "loss": 0.8262, "step": 8978 }, { "epoch": 0.77, "learning_rate": 1.3664525839039555e-06, "loss": 0.8011, "step": 8979 }, { "epoch": 0.77, "learning_rate": 1.3655034391840878e-06, "loss": 0.8685, "step": 8980 }, { "epoch": 0.77, "learning_rate": 1.3645545720852026e-06, "loss": 0.7937, "step": 8981 }, { "epoch": 0.77, "learning_rate": 1.363605982679778e-06, "loss": 0.7577, "step": 8982 }, { "epoch": 0.77, "learning_rate": 1.36265767104027e-06, "loss": 0.761, "step": 8983 }, { "epoch": 0.77, "learning_rate": 1.361709637239121e-06, "loss": 0.7545, "step": 8984 }, { "epoch": 0.77, "learning_rate": 1.3607618813487388e-06, "loss": 0.8082, "step": 8985 }, { "epoch": 0.77, "learning_rate": 1.359814403441524e-06, "loss": 0.8031, "step": 8986 }, { "epoch": 0.77, "learning_rate": 1.358867203589847e-06, "loss": 0.805, "step": 8987 }, { "epoch": 0.77, "learning_rate": 1.357920281866058e-06, "loss": 0.8176, "step": 8988 }, { "epoch": 0.77, "learning_rate": 1.356973638342492e-06, "loss": 0.7788, "step": 8989 }, { "epoch": 0.77, "learning_rate": 1.3560272730914554e-06, "loss": 0.8006, "step": 8990 }, { "epoch": 0.77, "learning_rate": 1.3550811861852369e-06, "loss": 0.7896, "step": 8991 }, { "epoch": 0.77, "learning_rate": 1.3541353776961035e-06, "loss": 0.7205, "step": 8992 }, { "epoch": 0.77, "learning_rate": 1.3531898476962996e-06, "loss": 0.7814, "step": 8993 }, { "epoch": 0.77, "learning_rate": 1.352244596258051e-06, "loss": 0.7909, "step": 8994 }, { "epoch": 0.77, "learning_rate": 1.3512996234535608e-06, "loss": 0.7779, "step": 8995 }, { "epoch": 0.77, "learning_rate": 1.350354929355009e-06, "loss": 0.7906, "step": 8996 }, { "epoch": 0.77, "learning_rate": 1.3494105140345593e-06, "loss": 0.7682, "step": 8997 }, { "epoch": 0.77, "learning_rate": 1.3484663775643465e-06, "loss": 0.7914, "step": 8998 }, { "epoch": 0.77, "learning_rate": 1.3475225200164915e-06, "loss": 0.7608, "step": 8999 }, { "epoch": 0.77, "learning_rate": 1.3465789414630908e-06, "loss": 0.8725, "step": 9000 }, { "epoch": 0.77, "learning_rate": 1.3456356419762168e-06, "loss": 0.7943, "step": 9001 }, { "epoch": 0.77, "learning_rate": 1.3446926216279272e-06, "loss": 0.7959, "step": 9002 }, { "epoch": 0.77, "learning_rate": 1.3437498804902515e-06, "loss": 0.746, "step": 9003 }, { "epoch": 0.77, "learning_rate": 1.3428074186352054e-06, "loss": 0.7943, "step": 9004 }, { "epoch": 0.77, "learning_rate": 1.3418652361347728e-06, "loss": 0.7935, "step": 9005 }, { "epoch": 0.77, "learning_rate": 1.3409233330609266e-06, "loss": 0.7945, "step": 9006 }, { "epoch": 0.77, "learning_rate": 1.3399817094856132e-06, "loss": 0.7707, "step": 9007 }, { "epoch": 0.77, "learning_rate": 1.3390403654807566e-06, "loss": 0.8245, "step": 9008 }, { "epoch": 0.77, "learning_rate": 1.3380993011182636e-06, "loss": 0.7737, "step": 9009 }, { "epoch": 0.77, "learning_rate": 1.3371585164700175e-06, "loss": 0.7941, "step": 9010 }, { "epoch": 0.77, "learning_rate": 1.3362180116078787e-06, "loss": 0.8195, "step": 9011 }, { "epoch": 0.77, "learning_rate": 1.335277786603688e-06, "loss": 0.7939, "step": 9012 }, { "epoch": 0.77, "learning_rate": 1.3343378415292641e-06, "loss": 0.8006, "step": 9013 }, { "epoch": 0.77, "learning_rate": 1.3333981764564059e-06, "loss": 0.7873, "step": 9014 }, { "epoch": 0.77, "learning_rate": 1.3324587914568899e-06, "loss": 0.8041, "step": 9015 }, { "epoch": 0.77, "learning_rate": 1.3315196866024676e-06, "loss": 0.8227, "step": 9016 }, { "epoch": 0.77, "learning_rate": 1.330580861964879e-06, "loss": 0.7763, "step": 9017 }, { "epoch": 0.77, "learning_rate": 1.329642317615829e-06, "loss": 0.8386, "step": 9018 }, { "epoch": 0.77, "learning_rate": 1.3287040536270135e-06, "loss": 0.7658, "step": 9019 }, { "epoch": 0.77, "learning_rate": 1.3277660700700996e-06, "loss": 0.7837, "step": 9020 }, { "epoch": 0.77, "learning_rate": 1.3268283670167336e-06, "loss": 0.7263, "step": 9021 }, { "epoch": 0.77, "learning_rate": 1.325890944538546e-06, "loss": 0.8148, "step": 9022 }, { "epoch": 0.77, "learning_rate": 1.324953802707139e-06, "loss": 0.712, "step": 9023 }, { "epoch": 0.77, "learning_rate": 1.3240169415940974e-06, "loss": 0.7464, "step": 9024 }, { "epoch": 0.77, "learning_rate": 1.323080361270982e-06, "loss": 0.8049, "step": 9025 }, { "epoch": 0.77, "learning_rate": 1.3221440618093335e-06, "loss": 0.5229, "step": 9026 }, { "epoch": 0.77, "learning_rate": 1.3212080432806728e-06, "loss": 0.7319, "step": 9027 }, { "epoch": 0.77, "learning_rate": 1.3202723057564965e-06, "loss": 0.7823, "step": 9028 }, { "epoch": 0.77, "learning_rate": 1.3193368493082797e-06, "loss": 0.7914, "step": 9029 }, { "epoch": 0.77, "learning_rate": 1.3184016740074824e-06, "loss": 0.821, "step": 9030 }, { "epoch": 0.77, "learning_rate": 1.3174667799255302e-06, "loss": 0.7452, "step": 9031 }, { "epoch": 0.77, "learning_rate": 1.316532167133841e-06, "loss": 0.7849, "step": 9032 }, { "epoch": 0.77, "learning_rate": 1.3155978357038025e-06, "loss": 0.8025, "step": 9033 }, { "epoch": 0.77, "learning_rate": 1.3146637857067835e-06, "loss": 0.7624, "step": 9034 }, { "epoch": 0.77, "learning_rate": 1.3137300172141333e-06, "loss": 0.7632, "step": 9035 }, { "epoch": 0.77, "learning_rate": 1.3127965302971746e-06, "loss": 0.7989, "step": 9036 }, { "epoch": 0.77, "learning_rate": 1.3118633250272172e-06, "loss": 0.7727, "step": 9037 }, { "epoch": 0.77, "learning_rate": 1.3109304014755392e-06, "loss": 0.8194, "step": 9038 }, { "epoch": 0.77, "learning_rate": 1.3099977597134023e-06, "loss": 0.7896, "step": 9039 }, { "epoch": 0.77, "learning_rate": 1.3090653998120484e-06, "loss": 0.7456, "step": 9040 }, { "epoch": 0.77, "learning_rate": 1.3081333218426928e-06, "loss": 0.5378, "step": 9041 }, { "epoch": 0.77, "learning_rate": 1.3072015258765365e-06, "loss": 0.8025, "step": 9042 }, { "epoch": 0.77, "learning_rate": 1.306270011984752e-06, "loss": 0.8178, "step": 9043 }, { "epoch": 0.77, "learning_rate": 1.3053387802384936e-06, "loss": 0.8265, "step": 9044 }, { "epoch": 0.77, "learning_rate": 1.304407830708893e-06, "loss": 0.7645, "step": 9045 }, { "epoch": 0.77, "learning_rate": 1.30347716346706e-06, "loss": 0.8009, "step": 9046 }, { "epoch": 0.77, "learning_rate": 1.3025467785840863e-06, "loss": 0.8285, "step": 9047 }, { "epoch": 0.77, "learning_rate": 1.3016166761310374e-06, "loss": 0.8265, "step": 9048 }, { "epoch": 0.77, "learning_rate": 1.3006868561789576e-06, "loss": 0.7959, "step": 9049 }, { "epoch": 0.77, "learning_rate": 1.299757318798876e-06, "loss": 0.5114, "step": 9050 }, { "epoch": 0.77, "learning_rate": 1.2988280640617901e-06, "loss": 0.7832, "step": 9051 }, { "epoch": 0.77, "learning_rate": 1.2978990920386842e-06, "loss": 0.7941, "step": 9052 }, { "epoch": 0.77, "learning_rate": 1.296970402800517e-06, "loss": 0.786, "step": 9053 }, { "epoch": 0.77, "learning_rate": 1.2960419964182247e-06, "loss": 0.7865, "step": 9054 }, { "epoch": 0.77, "learning_rate": 1.2951138729627267e-06, "loss": 0.7656, "step": 9055 }, { "epoch": 0.77, "learning_rate": 1.294186032504916e-06, "loss": 0.8259, "step": 9056 }, { "epoch": 0.77, "learning_rate": 1.2932584751156662e-06, "loss": 0.5774, "step": 9057 }, { "epoch": 0.77, "learning_rate": 1.2923312008658284e-06, "loss": 0.7579, "step": 9058 }, { "epoch": 0.77, "learning_rate": 1.2914042098262315e-06, "loss": 0.8308, "step": 9059 }, { "epoch": 0.77, "learning_rate": 1.2904775020676852e-06, "loss": 0.7976, "step": 9060 }, { "epoch": 0.77, "learning_rate": 1.289551077660976e-06, "loss": 0.5335, "step": 9061 }, { "epoch": 0.77, "learning_rate": 1.288624936676867e-06, "loss": 0.7811, "step": 9062 }, { "epoch": 0.77, "learning_rate": 1.287699079186106e-06, "loss": 0.5352, "step": 9063 }, { "epoch": 0.77, "learning_rate": 1.2867735052594083e-06, "loss": 0.8055, "step": 9064 }, { "epoch": 0.77, "learning_rate": 1.2858482149674779e-06, "loss": 0.7733, "step": 9065 }, { "epoch": 0.77, "learning_rate": 1.2849232083809925e-06, "loss": 0.6001, "step": 9066 }, { "epoch": 0.77, "learning_rate": 1.2839984855706072e-06, "loss": 0.8374, "step": 9067 }, { "epoch": 0.77, "learning_rate": 1.2830740466069596e-06, "loss": 0.8099, "step": 9068 }, { "epoch": 0.77, "learning_rate": 1.28214989156066e-06, "loss": 0.8225, "step": 9069 }, { "epoch": 0.77, "learning_rate": 1.2812260205023042e-06, "loss": 0.8144, "step": 9070 }, { "epoch": 0.77, "learning_rate": 1.2803024335024583e-06, "loss": 0.7619, "step": 9071 }, { "epoch": 0.77, "learning_rate": 1.27937913063167e-06, "loss": 0.7575, "step": 9072 }, { "epoch": 0.77, "learning_rate": 1.2784561119604683e-06, "loss": 0.7913, "step": 9073 }, { "epoch": 0.77, "learning_rate": 1.2775333775593557e-06, "loss": 0.761, "step": 9074 }, { "epoch": 0.77, "learning_rate": 1.2766109274988187e-06, "loss": 0.8346, "step": 9075 }, { "epoch": 0.77, "learning_rate": 1.2756887618493158e-06, "loss": 0.8342, "step": 9076 }, { "epoch": 0.77, "learning_rate": 1.2747668806812874e-06, "loss": 0.8141, "step": 9077 }, { "epoch": 0.77, "learning_rate": 1.2738452840651516e-06, "loss": 0.7457, "step": 9078 }, { "epoch": 0.77, "learning_rate": 1.2729239720713022e-06, "loss": 0.8011, "step": 9079 }, { "epoch": 0.77, "learning_rate": 1.2720029447701176e-06, "loss": 0.7494, "step": 9080 }, { "epoch": 0.77, "learning_rate": 1.2710822022319486e-06, "loss": 0.8147, "step": 9081 }, { "epoch": 0.77, "learning_rate": 1.2701617445271241e-06, "loss": 0.8368, "step": 9082 }, { "epoch": 0.78, "learning_rate": 1.2692415717259583e-06, "loss": 0.7414, "step": 9083 }, { "epoch": 0.78, "learning_rate": 1.2683216838987329e-06, "loss": 0.8236, "step": 9084 }, { "epoch": 0.78, "learning_rate": 1.2674020811157167e-06, "loss": 0.773, "step": 9085 }, { "epoch": 0.78, "learning_rate": 1.2664827634471539e-06, "loss": 0.7685, "step": 9086 }, { "epoch": 0.78, "learning_rate": 1.2655637309632634e-06, "loss": 0.7495, "step": 9087 }, { "epoch": 0.78, "learning_rate": 1.2646449837342494e-06, "loss": 0.794, "step": 9088 }, { "epoch": 0.78, "learning_rate": 1.2637265218302886e-06, "loss": 0.7622, "step": 9089 }, { "epoch": 0.78, "learning_rate": 1.2628083453215378e-06, "loss": 0.7937, "step": 9090 }, { "epoch": 0.78, "learning_rate": 1.2618904542781324e-06, "loss": 0.7989, "step": 9091 }, { "epoch": 0.78, "learning_rate": 1.2609728487701828e-06, "loss": 0.8325, "step": 9092 }, { "epoch": 0.78, "learning_rate": 1.260055528867784e-06, "loss": 0.7206, "step": 9093 }, { "epoch": 0.78, "learning_rate": 1.2591384946410045e-06, "loss": 0.8051, "step": 9094 }, { "epoch": 0.78, "learning_rate": 1.2582217461598895e-06, "loss": 0.7659, "step": 9095 }, { "epoch": 0.78, "learning_rate": 1.2573052834944682e-06, "loss": 0.7889, "step": 9096 }, { "epoch": 0.78, "learning_rate": 1.2563891067147432e-06, "loss": 0.7269, "step": 9097 }, { "epoch": 0.78, "learning_rate": 1.2554732158906969e-06, "loss": 0.8166, "step": 9098 }, { "epoch": 0.78, "learning_rate": 1.2545576110922886e-06, "loss": 0.8051, "step": 9099 }, { "epoch": 0.78, "learning_rate": 1.2536422923894565e-06, "loss": 0.7525, "step": 9100 }, { "epoch": 0.78, "learning_rate": 1.2527272598521196e-06, "loss": 0.7884, "step": 9101 }, { "epoch": 0.78, "learning_rate": 1.2518125135501708e-06, "loss": 0.7669, "step": 9102 }, { "epoch": 0.78, "learning_rate": 1.250898053553483e-06, "loss": 0.8264, "step": 9103 }, { "epoch": 0.78, "learning_rate": 1.2499838799319086e-06, "loss": 0.8029, "step": 9104 }, { "epoch": 0.78, "learning_rate": 1.249069992755274e-06, "loss": 0.7862, "step": 9105 }, { "epoch": 0.78, "learning_rate": 1.2481563920933892e-06, "loss": 0.794, "step": 9106 }, { "epoch": 0.78, "learning_rate": 1.2472430780160378e-06, "loss": 0.8248, "step": 9107 }, { "epoch": 0.78, "learning_rate": 1.2463300505929849e-06, "loss": 0.796, "step": 9108 }, { "epoch": 0.78, "learning_rate": 1.2454173098939715e-06, "loss": 0.8333, "step": 9109 }, { "epoch": 0.78, "learning_rate": 1.2445048559887175e-06, "loss": 0.8026, "step": 9110 }, { "epoch": 0.78, "learning_rate": 1.2435926889469197e-06, "loss": 0.8156, "step": 9111 }, { "epoch": 0.78, "learning_rate": 1.242680808838253e-06, "loss": 0.7955, "step": 9112 }, { "epoch": 0.78, "learning_rate": 1.241769215732374e-06, "loss": 0.793, "step": 9113 }, { "epoch": 0.78, "learning_rate": 1.2408579096989144e-06, "loss": 0.8669, "step": 9114 }, { "epoch": 0.78, "learning_rate": 1.2399468908074814e-06, "loss": 0.8382, "step": 9115 }, { "epoch": 0.78, "learning_rate": 1.2390361591276683e-06, "loss": 0.7691, "step": 9116 }, { "epoch": 0.78, "learning_rate": 1.238125714729036e-06, "loss": 0.7703, "step": 9117 }, { "epoch": 0.78, "learning_rate": 1.2372155576811318e-06, "loss": 0.7876, "step": 9118 }, { "epoch": 0.78, "learning_rate": 1.2363056880534774e-06, "loss": 0.7925, "step": 9119 }, { "epoch": 0.78, "learning_rate": 1.2353961059155717e-06, "loss": 0.768, "step": 9120 }, { "epoch": 0.78, "learning_rate": 1.2344868113368958e-06, "loss": 0.7978, "step": 9121 }, { "epoch": 0.78, "learning_rate": 1.233577804386905e-06, "loss": 0.8332, "step": 9122 }, { "epoch": 0.78, "learning_rate": 1.2326690851350336e-06, "loss": 0.8152, "step": 9123 }, { "epoch": 0.78, "learning_rate": 1.2317606536506942e-06, "loss": 0.826, "step": 9124 }, { "epoch": 0.78, "learning_rate": 1.2308525100032764e-06, "loss": 0.8139, "step": 9125 }, { "epoch": 0.78, "learning_rate": 1.2299446542621502e-06, "loss": 0.4894, "step": 9126 }, { "epoch": 0.78, "learning_rate": 1.2290370864966623e-06, "loss": 0.7739, "step": 9127 }, { "epoch": 0.78, "learning_rate": 1.2281298067761344e-06, "loss": 0.7825, "step": 9128 }, { "epoch": 0.78, "learning_rate": 1.2272228151698728e-06, "loss": 0.8018, "step": 9129 }, { "epoch": 0.78, "learning_rate": 1.2263161117471567e-06, "loss": 0.7537, "step": 9130 }, { "epoch": 0.78, "learning_rate": 1.2254096965772445e-06, "loss": 0.8311, "step": 9131 }, { "epoch": 0.78, "learning_rate": 1.2245035697293723e-06, "loss": 0.7965, "step": 9132 }, { "epoch": 0.78, "learning_rate": 1.2235977312727532e-06, "loss": 0.7696, "step": 9133 }, { "epoch": 0.78, "learning_rate": 1.222692181276583e-06, "loss": 0.8165, "step": 9134 }, { "epoch": 0.78, "learning_rate": 1.2217869198100307e-06, "loss": 0.7659, "step": 9135 }, { "epoch": 0.78, "learning_rate": 1.2208819469422439e-06, "loss": 0.7907, "step": 9136 }, { "epoch": 0.78, "learning_rate": 1.2199772627423501e-06, "loss": 0.8188, "step": 9137 }, { "epoch": 0.78, "learning_rate": 1.219072867279451e-06, "loss": 0.74, "step": 9138 }, { "epoch": 0.78, "learning_rate": 1.2181687606226328e-06, "loss": 0.7941, "step": 9139 }, { "epoch": 0.78, "learning_rate": 1.2172649428409532e-06, "loss": 0.7784, "step": 9140 }, { "epoch": 0.78, "learning_rate": 1.2163614140034497e-06, "loss": 0.8169, "step": 9141 }, { "epoch": 0.78, "learning_rate": 1.2154581741791404e-06, "loss": 0.794, "step": 9142 }, { "epoch": 0.78, "learning_rate": 1.2145552234370189e-06, "loss": 0.7798, "step": 9143 }, { "epoch": 0.78, "learning_rate": 1.213652561846056e-06, "loss": 0.7793, "step": 9144 }, { "epoch": 0.78, "learning_rate": 1.2127501894752025e-06, "loss": 0.7964, "step": 9145 }, { "epoch": 0.78, "learning_rate": 1.211848106393384e-06, "loss": 0.8277, "step": 9146 }, { "epoch": 0.78, "learning_rate": 1.2109463126695086e-06, "loss": 0.8003, "step": 9147 }, { "epoch": 0.78, "learning_rate": 1.2100448083724581e-06, "loss": 0.7991, "step": 9148 }, { "epoch": 0.78, "learning_rate": 1.209143593571096e-06, "loss": 0.7766, "step": 9149 }, { "epoch": 0.78, "learning_rate": 1.2082426683342603e-06, "loss": 0.7944, "step": 9150 }, { "epoch": 0.78, "learning_rate": 1.2073420327307678e-06, "loss": 0.7525, "step": 9151 }, { "epoch": 0.78, "learning_rate": 1.2064416868294143e-06, "loss": 0.7898, "step": 9152 }, { "epoch": 0.78, "learning_rate": 1.2055416306989703e-06, "loss": 0.7985, "step": 9153 }, { "epoch": 0.78, "learning_rate": 1.2046418644081904e-06, "loss": 0.7969, "step": 9154 }, { "epoch": 0.78, "learning_rate": 1.2037423880258015e-06, "loss": 0.8125, "step": 9155 }, { "epoch": 0.78, "learning_rate": 1.20284320162051e-06, "loss": 0.7887, "step": 9156 }, { "epoch": 0.78, "learning_rate": 1.2019443052610003e-06, "loss": 0.7682, "step": 9157 }, { "epoch": 0.78, "learning_rate": 1.201045699015933e-06, "loss": 0.7357, "step": 9158 }, { "epoch": 0.78, "learning_rate": 1.2001473829539513e-06, "loss": 0.7736, "step": 9159 }, { "epoch": 0.78, "learning_rate": 1.1992493571436709e-06, "loss": 0.7587, "step": 9160 }, { "epoch": 0.78, "learning_rate": 1.1983516216536867e-06, "loss": 0.7645, "step": 9161 }, { "epoch": 0.78, "learning_rate": 1.1974541765525755e-06, "loss": 0.7919, "step": 9162 }, { "epoch": 0.78, "learning_rate": 1.1965570219088862e-06, "loss": 0.8096, "step": 9163 }, { "epoch": 0.78, "learning_rate": 1.1956601577911487e-06, "loss": 0.8133, "step": 9164 }, { "epoch": 0.78, "learning_rate": 1.1947635842678695e-06, "loss": 0.7894, "step": 9165 }, { "epoch": 0.78, "learning_rate": 1.1938673014075319e-06, "loss": 0.8143, "step": 9166 }, { "epoch": 0.78, "learning_rate": 1.1929713092786021e-06, "loss": 0.8057, "step": 9167 }, { "epoch": 0.78, "learning_rate": 1.1920756079495183e-06, "loss": 0.7597, "step": 9168 }, { "epoch": 0.78, "learning_rate": 1.1911801974886971e-06, "loss": 0.8354, "step": 9169 }, { "epoch": 0.78, "learning_rate": 1.1902850779645391e-06, "loss": 0.7903, "step": 9170 }, { "epoch": 0.78, "learning_rate": 1.1893902494454117e-06, "loss": 0.7507, "step": 9171 }, { "epoch": 0.78, "learning_rate": 1.1884957119996716e-06, "loss": 0.7775, "step": 9172 }, { "epoch": 0.78, "learning_rate": 1.187601465695646e-06, "loss": 0.8053, "step": 9173 }, { "epoch": 0.78, "learning_rate": 1.1867075106016407e-06, "loss": 0.8346, "step": 9174 }, { "epoch": 0.78, "learning_rate": 1.1858138467859432e-06, "loss": 0.8067, "step": 9175 }, { "epoch": 0.78, "learning_rate": 1.184920474316814e-06, "loss": 0.5075, "step": 9176 }, { "epoch": 0.78, "learning_rate": 1.1840273932624946e-06, "loss": 0.7812, "step": 9177 }, { "epoch": 0.78, "learning_rate": 1.1831346036912018e-06, "loss": 0.7751, "step": 9178 }, { "epoch": 0.78, "learning_rate": 1.1822421056711303e-06, "loss": 0.8048, "step": 9179 }, { "epoch": 0.78, "learning_rate": 1.1813498992704569e-06, "loss": 0.8202, "step": 9180 }, { "epoch": 0.78, "learning_rate": 1.1804579845573288e-06, "loss": 0.7682, "step": 9181 }, { "epoch": 0.78, "learning_rate": 1.1795663615998793e-06, "loss": 0.7421, "step": 9182 }, { "epoch": 0.78, "learning_rate": 1.1786750304662136e-06, "loss": 0.7699, "step": 9183 }, { "epoch": 0.78, "learning_rate": 1.1777839912244127e-06, "loss": 0.7788, "step": 9184 }, { "epoch": 0.78, "learning_rate": 1.1768932439425423e-06, "loss": 0.7895, "step": 9185 }, { "epoch": 0.78, "learning_rate": 1.1760027886886395e-06, "loss": 0.7772, "step": 9186 }, { "epoch": 0.78, "learning_rate": 1.1751126255307243e-06, "loss": 0.762, "step": 9187 }, { "epoch": 0.78, "learning_rate": 1.1742227545367907e-06, "loss": 0.7862, "step": 9188 }, { "epoch": 0.78, "learning_rate": 1.1733331757748111e-06, "loss": 0.7889, "step": 9189 }, { "epoch": 0.78, "learning_rate": 1.1724438893127371e-06, "loss": 0.7842, "step": 9190 }, { "epoch": 0.78, "learning_rate": 1.1715548952184942e-06, "loss": 0.7943, "step": 9191 }, { "epoch": 0.78, "learning_rate": 1.1706661935599912e-06, "loss": 0.7793, "step": 9192 }, { "epoch": 0.78, "learning_rate": 1.1697777844051105e-06, "loss": 0.7914, "step": 9193 }, { "epoch": 0.78, "learning_rate": 1.1688896678217116e-06, "loss": 0.7919, "step": 9194 }, { "epoch": 0.78, "learning_rate": 1.1680018438776363e-06, "loss": 0.8064, "step": 9195 }, { "epoch": 0.78, "learning_rate": 1.1671143126406998e-06, "loss": 0.5576, "step": 9196 }, { "epoch": 0.78, "learning_rate": 1.1662270741786953e-06, "loss": 0.7739, "step": 9197 }, { "epoch": 0.78, "learning_rate": 1.1653401285593951e-06, "loss": 0.8098, "step": 9198 }, { "epoch": 0.78, "learning_rate": 1.1644534758505476e-06, "loss": 0.8251, "step": 9199 }, { "epoch": 0.79, "learning_rate": 1.163567116119882e-06, "loss": 0.7742, "step": 9200 }, { "epoch": 0.79, "learning_rate": 1.162681049435101e-06, "loss": 0.8266, "step": 9201 }, { "epoch": 0.79, "learning_rate": 1.161795275863886e-06, "loss": 0.7763, "step": 9202 }, { "epoch": 0.79, "learning_rate": 1.1609097954739012e-06, "loss": 0.7606, "step": 9203 }, { "epoch": 0.79, "learning_rate": 1.160024608332778e-06, "loss": 0.7961, "step": 9204 }, { "epoch": 0.79, "learning_rate": 1.1591397145081362e-06, "loss": 0.8304, "step": 9205 }, { "epoch": 0.79, "learning_rate": 1.1582551140675657e-06, "loss": 0.8104, "step": 9206 }, { "epoch": 0.79, "learning_rate": 1.157370807078637e-06, "loss": 0.8178, "step": 9207 }, { "epoch": 0.79, "learning_rate": 1.156486793608899e-06, "loss": 0.8159, "step": 9208 }, { "epoch": 0.79, "learning_rate": 1.1556030737258772e-06, "loss": 0.7872, "step": 9209 }, { "epoch": 0.79, "learning_rate": 1.1547196474970734e-06, "loss": 0.7771, "step": 9210 }, { "epoch": 0.79, "learning_rate": 1.1538365149899689e-06, "loss": 0.789, "step": 9211 }, { "epoch": 0.79, "learning_rate": 1.15295367627202e-06, "loss": 0.709, "step": 9212 }, { "epoch": 0.79, "learning_rate": 1.1520711314106648e-06, "loss": 0.8224, "step": 9213 }, { "epoch": 0.79, "learning_rate": 1.151188880473314e-06, "loss": 0.8186, "step": 9214 }, { "epoch": 0.79, "learning_rate": 1.1503069235273613e-06, "loss": 0.801, "step": 9215 }, { "epoch": 0.79, "learning_rate": 1.1494252606401745e-06, "loss": 0.7901, "step": 9216 }, { "epoch": 0.79, "learning_rate": 1.1485438918790959e-06, "loss": 0.7771, "step": 9217 }, { "epoch": 0.79, "learning_rate": 1.1476628173114523e-06, "loss": 0.799, "step": 9218 }, { "epoch": 0.79, "learning_rate": 1.1467820370045424e-06, "loss": 0.8113, "step": 9219 }, { "epoch": 0.79, "learning_rate": 1.1459015510256472e-06, "loss": 0.7437, "step": 9220 }, { "epoch": 0.79, "learning_rate": 1.1450213594420207e-06, "loss": 0.8236, "step": 9221 }, { "epoch": 0.79, "learning_rate": 1.1441414623208957e-06, "loss": 0.8267, "step": 9222 }, { "epoch": 0.79, "learning_rate": 1.143261859729487e-06, "loss": 0.7431, "step": 9223 }, { "epoch": 0.79, "learning_rate": 1.1423825517349768e-06, "loss": 0.8078, "step": 9224 }, { "epoch": 0.79, "learning_rate": 1.1415035384045365e-06, "loss": 0.8197, "step": 9225 }, { "epoch": 0.79, "learning_rate": 1.1406248198053078e-06, "loss": 0.8401, "step": 9226 }, { "epoch": 0.79, "learning_rate": 1.1397463960044093e-06, "loss": 0.7694, "step": 9227 }, { "epoch": 0.79, "learning_rate": 1.1388682670689428e-06, "loss": 0.8114, "step": 9228 }, { "epoch": 0.79, "learning_rate": 1.137990433065983e-06, "loss": 0.8397, "step": 9229 }, { "epoch": 0.79, "learning_rate": 1.1371128940625825e-06, "loss": 0.7859, "step": 9230 }, { "epoch": 0.79, "learning_rate": 1.1362356501257731e-06, "loss": 0.8057, "step": 9231 }, { "epoch": 0.79, "learning_rate": 1.135358701322561e-06, "loss": 0.805, "step": 9232 }, { "epoch": 0.79, "learning_rate": 1.1344820477199347e-06, "loss": 0.8361, "step": 9233 }, { "epoch": 0.79, "learning_rate": 1.1336056893848562e-06, "loss": 0.5536, "step": 9234 }, { "epoch": 0.79, "learning_rate": 1.1327296263842653e-06, "loss": 0.8233, "step": 9235 }, { "epoch": 0.79, "learning_rate": 1.1318538587850831e-06, "loss": 0.8104, "step": 9236 }, { "epoch": 0.79, "learning_rate": 1.1309783866542002e-06, "loss": 0.7861, "step": 9237 }, { "epoch": 0.79, "learning_rate": 1.1301032100584935e-06, "loss": 0.7982, "step": 9238 }, { "epoch": 0.79, "learning_rate": 1.129228329064812e-06, "loss": 0.5351, "step": 9239 }, { "epoch": 0.79, "learning_rate": 1.1283537437399823e-06, "loss": 0.7541, "step": 9240 }, { "epoch": 0.79, "learning_rate": 1.1274794541508121e-06, "loss": 0.7883, "step": 9241 }, { "epoch": 0.79, "learning_rate": 1.126605460364083e-06, "loss": 0.8279, "step": 9242 }, { "epoch": 0.79, "learning_rate": 1.1257317624465542e-06, "loss": 0.8087, "step": 9243 }, { "epoch": 0.79, "learning_rate": 1.1248583604649639e-06, "loss": 0.8264, "step": 9244 }, { "epoch": 0.79, "learning_rate": 1.1239852544860252e-06, "loss": 0.7793, "step": 9245 }, { "epoch": 0.79, "learning_rate": 1.123112444576433e-06, "loss": 0.7299, "step": 9246 }, { "epoch": 0.79, "learning_rate": 1.1222399308028558e-06, "loss": 0.7853, "step": 9247 }, { "epoch": 0.79, "learning_rate": 1.1213677132319384e-06, "loss": 0.7745, "step": 9248 }, { "epoch": 0.79, "learning_rate": 1.1204957919303105e-06, "loss": 0.7909, "step": 9249 }, { "epoch": 0.79, "learning_rate": 1.1196241669645675e-06, "loss": 0.8181, "step": 9250 }, { "epoch": 0.79, "learning_rate": 1.1187528384012925e-06, "loss": 0.7685, "step": 9251 }, { "epoch": 0.79, "learning_rate": 1.117881806307039e-06, "loss": 0.8096, "step": 9252 }, { "epoch": 0.79, "learning_rate": 1.1170110707483445e-06, "loss": 0.5073, "step": 9253 }, { "epoch": 0.79, "learning_rate": 1.1161406317917184e-06, "loss": 0.7808, "step": 9254 }, { "epoch": 0.79, "learning_rate": 1.115270489503647e-06, "loss": 0.8144, "step": 9255 }, { "epoch": 0.79, "learning_rate": 1.1144006439506011e-06, "loss": 0.8038, "step": 9256 }, { "epoch": 0.79, "learning_rate": 1.1135310951990174e-06, "loss": 0.7919, "step": 9257 }, { "epoch": 0.79, "learning_rate": 1.112661843315322e-06, "loss": 0.7891, "step": 9258 }, { "epoch": 0.79, "learning_rate": 1.1117928883659101e-06, "loss": 0.7733, "step": 9259 }, { "epoch": 0.79, "learning_rate": 1.1109242304171553e-06, "loss": 0.7508, "step": 9260 }, { "epoch": 0.79, "learning_rate": 1.110055869535414e-06, "loss": 0.7806, "step": 9261 }, { "epoch": 0.79, "learning_rate": 1.1091878057870137e-06, "loss": 0.7911, "step": 9262 }, { "epoch": 0.79, "learning_rate": 1.1083200392382615e-06, "loss": 0.7776, "step": 9263 }, { "epoch": 0.79, "learning_rate": 1.1074525699554417e-06, "loss": 0.7996, "step": 9264 }, { "epoch": 0.79, "learning_rate": 1.1065853980048146e-06, "loss": 0.7822, "step": 9265 }, { "epoch": 0.79, "learning_rate": 1.1057185234526218e-06, "loss": 0.7949, "step": 9266 }, { "epoch": 0.79, "learning_rate": 1.1048519463650786e-06, "loss": 0.7779, "step": 9267 }, { "epoch": 0.79, "learning_rate": 1.1039856668083765e-06, "loss": 0.8318, "step": 9268 }, { "epoch": 0.79, "learning_rate": 1.1031196848486907e-06, "loss": 0.7932, "step": 9269 }, { "epoch": 0.79, "learning_rate": 1.1022540005521643e-06, "loss": 0.7634, "step": 9270 }, { "epoch": 0.79, "learning_rate": 1.1013886139849255e-06, "loss": 0.7896, "step": 9271 }, { "epoch": 0.79, "learning_rate": 1.1005235252130758e-06, "loss": 0.7715, "step": 9272 }, { "epoch": 0.79, "learning_rate": 1.0996587343026943e-06, "loss": 0.769, "step": 9273 }, { "epoch": 0.79, "learning_rate": 1.0987942413198404e-06, "loss": 0.5425, "step": 9274 }, { "epoch": 0.79, "learning_rate": 1.097930046330547e-06, "loss": 0.7939, "step": 9275 }, { "epoch": 0.79, "learning_rate": 1.097066149400825e-06, "loss": 0.8024, "step": 9276 }, { "epoch": 0.79, "learning_rate": 1.0962025505966645e-06, "loss": 0.8177, "step": 9277 }, { "epoch": 0.79, "learning_rate": 1.0953392499840294e-06, "loss": 0.8149, "step": 9278 }, { "epoch": 0.79, "learning_rate": 1.094476247628865e-06, "loss": 0.7806, "step": 9279 }, { "epoch": 0.79, "learning_rate": 1.0936135435970912e-06, "loss": 0.7737, "step": 9280 }, { "epoch": 0.79, "learning_rate": 1.0927511379546047e-06, "loss": 0.7753, "step": 9281 }, { "epoch": 0.79, "learning_rate": 1.091889030767283e-06, "loss": 0.8107, "step": 9282 }, { "epoch": 0.79, "learning_rate": 1.0910272221009738e-06, "loss": 0.8101, "step": 9283 }, { "epoch": 0.79, "learning_rate": 1.0901657120215099e-06, "loss": 0.769, "step": 9284 }, { "epoch": 0.79, "learning_rate": 1.0893045005946967e-06, "loss": 0.8017, "step": 9285 }, { "epoch": 0.79, "learning_rate": 1.0884435878863165e-06, "loss": 0.8156, "step": 9286 }, { "epoch": 0.79, "learning_rate": 1.087582973962133e-06, "loss": 0.7507, "step": 9287 }, { "epoch": 0.79, "learning_rate": 1.0867226588878804e-06, "loss": 0.8021, "step": 9288 }, { "epoch": 0.79, "learning_rate": 1.0858626427292796e-06, "loss": 0.8413, "step": 9289 }, { "epoch": 0.79, "learning_rate": 1.0850029255520155e-06, "loss": 0.738, "step": 9290 }, { "epoch": 0.79, "learning_rate": 1.0841435074217638e-06, "loss": 0.8222, "step": 9291 }, { "epoch": 0.79, "learning_rate": 1.0832843884041683e-06, "loss": 0.8214, "step": 9292 }, { "epoch": 0.79, "learning_rate": 1.0824255685648515e-06, "loss": 0.7965, "step": 9293 }, { "epoch": 0.79, "learning_rate": 1.081567047969418e-06, "loss": 0.7902, "step": 9294 }, { "epoch": 0.79, "learning_rate": 1.0807088266834437e-06, "loss": 0.8286, "step": 9295 }, { "epoch": 0.79, "learning_rate": 1.0798509047724841e-06, "loss": 0.7601, "step": 9296 }, { "epoch": 0.79, "learning_rate": 1.0789932823020716e-06, "loss": 0.8044, "step": 9297 }, { "epoch": 0.79, "learning_rate": 1.0781359593377144e-06, "loss": 0.8072, "step": 9298 }, { "epoch": 0.79, "learning_rate": 1.0772789359449014e-06, "loss": 0.8322, "step": 9299 }, { "epoch": 0.79, "learning_rate": 1.0764222121890955e-06, "loss": 0.8423, "step": 9300 }, { "epoch": 0.79, "learning_rate": 1.0755657881357357e-06, "loss": 0.8523, "step": 9301 }, { "epoch": 0.79, "learning_rate": 1.0747096638502448e-06, "loss": 0.7334, "step": 9302 }, { "epoch": 0.79, "learning_rate": 1.0738538393980114e-06, "loss": 0.8283, "step": 9303 }, { "epoch": 0.79, "learning_rate": 1.0729983148444118e-06, "loss": 0.7734, "step": 9304 }, { "epoch": 0.79, "learning_rate": 1.0721430902547946e-06, "loss": 0.7583, "step": 9305 }, { "epoch": 0.79, "learning_rate": 1.0712881656944846e-06, "loss": 0.5014, "step": 9306 }, { "epoch": 0.79, "learning_rate": 1.0704335412287876e-06, "loss": 0.7784, "step": 9307 }, { "epoch": 0.79, "learning_rate": 1.0695792169229829e-06, "loss": 0.7933, "step": 9308 }, { "epoch": 0.79, "learning_rate": 1.068725192842328e-06, "loss": 0.8001, "step": 9309 }, { "epoch": 0.79, "learning_rate": 1.0678714690520575e-06, "loss": 0.7535, "step": 9310 }, { "epoch": 0.79, "learning_rate": 1.0670180456173818e-06, "loss": 0.7802, "step": 9311 }, { "epoch": 0.79, "learning_rate": 1.0661649226034926e-06, "loss": 0.8231, "step": 9312 }, { "epoch": 0.79, "learning_rate": 1.0653121000755535e-06, "loss": 0.8177, "step": 9313 }, { "epoch": 0.79, "learning_rate": 1.0644595780987065e-06, "loss": 0.7349, "step": 9314 }, { "epoch": 0.79, "learning_rate": 1.0636073567380762e-06, "loss": 0.8041, "step": 9315 }, { "epoch": 0.79, "learning_rate": 1.0627554360587533e-06, "loss": 0.7948, "step": 9316 }, { "epoch": 0.8, "learning_rate": 1.0619038161258155e-06, "loss": 0.8153, "step": 9317 }, { "epoch": 0.8, "learning_rate": 1.0610524970043134e-06, "loss": 0.7886, "step": 9318 }, { "epoch": 0.8, "learning_rate": 1.060201478759273e-06, "loss": 0.8102, "step": 9319 }, { "epoch": 0.8, "learning_rate": 1.059350761455702e-06, "loss": 0.7452, "step": 9320 }, { "epoch": 0.8, "learning_rate": 1.0585003451585802e-06, "loss": 0.7862, "step": 9321 }, { "epoch": 0.8, "learning_rate": 1.0576502299328707e-06, "loss": 0.7872, "step": 9322 }, { "epoch": 0.8, "learning_rate": 1.0568004158435047e-06, "loss": 0.7798, "step": 9323 }, { "epoch": 0.8, "learning_rate": 1.0559509029553955e-06, "loss": 0.7822, "step": 9324 }, { "epoch": 0.8, "learning_rate": 1.0551016913334366e-06, "loss": 0.7954, "step": 9325 }, { "epoch": 0.8, "learning_rate": 1.0542527810424918e-06, "loss": 0.8289, "step": 9326 }, { "epoch": 0.8, "learning_rate": 1.0534041721474075e-06, "loss": 0.8337, "step": 9327 }, { "epoch": 0.8, "learning_rate": 1.0525558647130036e-06, "loss": 0.7961, "step": 9328 }, { "epoch": 0.8, "learning_rate": 1.0517078588040786e-06, "loss": 0.4928, "step": 9329 }, { "epoch": 0.8, "learning_rate": 1.0508601544854063e-06, "loss": 0.7942, "step": 9330 }, { "epoch": 0.8, "learning_rate": 1.0500127518217385e-06, "loss": 0.7979, "step": 9331 }, { "epoch": 0.8, "learning_rate": 1.0491656508778054e-06, "loss": 0.8093, "step": 9332 }, { "epoch": 0.8, "learning_rate": 1.0483188517183124e-06, "loss": 0.8188, "step": 9333 }, { "epoch": 0.8, "learning_rate": 1.0474723544079406e-06, "loss": 0.7876, "step": 9334 }, { "epoch": 0.8, "learning_rate": 1.046626159011354e-06, "loss": 0.7276, "step": 9335 }, { "epoch": 0.8, "learning_rate": 1.0457802655931836e-06, "loss": 0.7321, "step": 9336 }, { "epoch": 0.8, "learning_rate": 1.0449346742180465e-06, "loss": 0.8071, "step": 9337 }, { "epoch": 0.8, "learning_rate": 1.0440893849505329e-06, "loss": 0.7925, "step": 9338 }, { "epoch": 0.8, "learning_rate": 1.0432443978552081e-06, "loss": 0.7981, "step": 9339 }, { "epoch": 0.8, "learning_rate": 1.0423997129966196e-06, "loss": 0.7791, "step": 9340 }, { "epoch": 0.8, "learning_rate": 1.0415553304392866e-06, "loss": 0.7921, "step": 9341 }, { "epoch": 0.8, "learning_rate": 1.040711250247708e-06, "loss": 0.748, "step": 9342 }, { "epoch": 0.8, "learning_rate": 1.0398674724863584e-06, "loss": 0.8399, "step": 9343 }, { "epoch": 0.8, "learning_rate": 1.0390239972196886e-06, "loss": 0.7583, "step": 9344 }, { "epoch": 0.8, "learning_rate": 1.03818082451213e-06, "loss": 0.8027, "step": 9345 }, { "epoch": 0.8, "learning_rate": 1.037337954428087e-06, "loss": 0.7762, "step": 9346 }, { "epoch": 0.8, "learning_rate": 1.036495387031941e-06, "loss": 0.7822, "step": 9347 }, { "epoch": 0.8, "learning_rate": 1.0356531223880545e-06, "loss": 0.7937, "step": 9348 }, { "epoch": 0.8, "learning_rate": 1.034811160560762e-06, "loss": 0.7976, "step": 9349 }, { "epoch": 0.8, "learning_rate": 1.033969501614377e-06, "loss": 0.8235, "step": 9350 }, { "epoch": 0.8, "learning_rate": 1.0331281456131892e-06, "loss": 0.8084, "step": 9351 }, { "epoch": 0.8, "learning_rate": 1.0322870926214645e-06, "loss": 0.8068, "step": 9352 }, { "epoch": 0.8, "learning_rate": 1.0314463427034498e-06, "loss": 0.7597, "step": 9353 }, { "epoch": 0.8, "learning_rate": 1.0306058959233627e-06, "loss": 0.7647, "step": 9354 }, { "epoch": 0.8, "learning_rate": 1.0297657523454047e-06, "loss": 0.7654, "step": 9355 }, { "epoch": 0.8, "learning_rate": 1.0289259120337464e-06, "loss": 0.7603, "step": 9356 }, { "epoch": 0.8, "learning_rate": 1.0280863750525389e-06, "loss": 0.8502, "step": 9357 }, { "epoch": 0.8, "learning_rate": 1.0272471414659125e-06, "loss": 0.7819, "step": 9358 }, { "epoch": 0.8, "learning_rate": 1.0264082113379698e-06, "loss": 0.7778, "step": 9359 }, { "epoch": 0.8, "learning_rate": 1.025569584732795e-06, "loss": 0.7803, "step": 9360 }, { "epoch": 0.8, "learning_rate": 1.0247312617144455e-06, "loss": 0.8359, "step": 9361 }, { "epoch": 0.8, "learning_rate": 1.0238932423469567e-06, "loss": 0.7653, "step": 9362 }, { "epoch": 0.8, "learning_rate": 1.0230555266943398e-06, "loss": 0.7746, "step": 9363 }, { "epoch": 0.8, "learning_rate": 1.0222181148205835e-06, "loss": 0.8057, "step": 9364 }, { "epoch": 0.8, "learning_rate": 1.0213810067896551e-06, "loss": 0.7636, "step": 9365 }, { "epoch": 0.8, "learning_rate": 1.0205442026654967e-06, "loss": 0.791, "step": 9366 }, { "epoch": 0.8, "learning_rate": 1.0197077025120256e-06, "loss": 0.8209, "step": 9367 }, { "epoch": 0.8, "learning_rate": 1.0188715063931403e-06, "loss": 0.8051, "step": 9368 }, { "epoch": 0.8, "learning_rate": 1.0180356143727132e-06, "loss": 0.8013, "step": 9369 }, { "epoch": 0.8, "learning_rate": 1.0172000265145938e-06, "loss": 0.7965, "step": 9370 }, { "epoch": 0.8, "learning_rate": 1.0163647428826078e-06, "loss": 0.7899, "step": 9371 }, { "epoch": 0.8, "learning_rate": 1.015529763540557e-06, "loss": 0.8362, "step": 9372 }, { "epoch": 0.8, "learning_rate": 1.0146950885522244e-06, "loss": 0.8236, "step": 9373 }, { "epoch": 0.8, "learning_rate": 1.0138607179813652e-06, "loss": 0.836, "step": 9374 }, { "epoch": 0.8, "learning_rate": 1.0130266518917132e-06, "loss": 0.8131, "step": 9375 }, { "epoch": 0.8, "learning_rate": 1.012192890346978e-06, "loss": 0.7704, "step": 9376 }, { "epoch": 0.8, "learning_rate": 1.0113594334108452e-06, "loss": 0.7559, "step": 9377 }, { "epoch": 0.8, "learning_rate": 1.0105262811469808e-06, "loss": 0.7613, "step": 9378 }, { "epoch": 0.8, "learning_rate": 1.009693433619024e-06, "loss": 0.7962, "step": 9379 }, { "epoch": 0.8, "learning_rate": 1.0088608908905906e-06, "loss": 0.7484, "step": 9380 }, { "epoch": 0.8, "learning_rate": 1.008028653025277e-06, "loss": 0.7902, "step": 9381 }, { "epoch": 0.8, "learning_rate": 1.0071967200866528e-06, "loss": 0.8133, "step": 9382 }, { "epoch": 0.8, "learning_rate": 1.0063650921382646e-06, "loss": 0.794, "step": 9383 }, { "epoch": 0.8, "learning_rate": 1.0055337692436363e-06, "loss": 0.795, "step": 9384 }, { "epoch": 0.8, "learning_rate": 1.0047027514662677e-06, "loss": 0.7618, "step": 9385 }, { "epoch": 0.8, "learning_rate": 1.0038720388696378e-06, "loss": 0.8344, "step": 9386 }, { "epoch": 0.8, "learning_rate": 1.0030416315172003e-06, "loss": 0.8005, "step": 9387 }, { "epoch": 0.8, "learning_rate": 1.0022115294723849e-06, "loss": 0.7475, "step": 9388 }, { "epoch": 0.8, "learning_rate": 1.0013817327985997e-06, "loss": 0.7581, "step": 9389 }, { "epoch": 0.8, "learning_rate": 1.0005522415592273e-06, "loss": 0.7861, "step": 9390 }, { "epoch": 0.8, "learning_rate": 9.997230558176301e-07, "loss": 0.8121, "step": 9391 }, { "epoch": 0.8, "learning_rate": 9.988941756371446e-07, "loss": 0.8076, "step": 9392 }, { "epoch": 0.8, "learning_rate": 9.980656010810857e-07, "loss": 0.7859, "step": 9393 }, { "epoch": 0.8, "learning_rate": 9.972373322127432e-07, "loss": 0.754, "step": 9394 }, { "epoch": 0.8, "learning_rate": 9.964093690953847e-07, "loss": 0.8023, "step": 9395 }, { "epoch": 0.8, "learning_rate": 9.955817117922546e-07, "loss": 0.754, "step": 9396 }, { "epoch": 0.8, "learning_rate": 9.947543603665711e-07, "loss": 0.7701, "step": 9397 }, { "epoch": 0.8, "learning_rate": 9.939273148815348e-07, "loss": 0.7985, "step": 9398 }, { "epoch": 0.8, "learning_rate": 9.93100575400318e-07, "loss": 0.8261, "step": 9399 }, { "epoch": 0.8, "learning_rate": 9.9227414198607e-07, "loss": 0.7872, "step": 9400 }, { "epoch": 0.8, "learning_rate": 9.914480147019202e-07, "loss": 0.7832, "step": 9401 }, { "epoch": 0.8, "learning_rate": 9.906221936109717e-07, "loss": 0.8006, "step": 9402 }, { "epoch": 0.8, "learning_rate": 9.897966787763036e-07, "loss": 0.8132, "step": 9403 }, { "epoch": 0.8, "learning_rate": 9.889714702609743e-07, "loss": 0.7803, "step": 9404 }, { "epoch": 0.8, "learning_rate": 9.881465681280155e-07, "loss": 0.8623, "step": 9405 }, { "epoch": 0.8, "learning_rate": 9.873219724404393e-07, "loss": 0.8143, "step": 9406 }, { "epoch": 0.8, "learning_rate": 9.864976832612321e-07, "loss": 0.7677, "step": 9407 }, { "epoch": 0.8, "learning_rate": 9.856737006533568e-07, "loss": 0.8097, "step": 9408 }, { "epoch": 0.8, "learning_rate": 9.848500246797527e-07, "loss": 0.8235, "step": 9409 }, { "epoch": 0.8, "learning_rate": 9.840266554033362e-07, "loss": 0.7694, "step": 9410 }, { "epoch": 0.8, "learning_rate": 9.83203592887002e-07, "loss": 0.8026, "step": 9411 }, { "epoch": 0.8, "learning_rate": 9.82380837193619e-07, "loss": 0.7859, "step": 9412 }, { "epoch": 0.8, "learning_rate": 9.815583883860313e-07, "loss": 0.7583, "step": 9413 }, { "epoch": 0.8, "learning_rate": 9.807362465270654e-07, "loss": 0.7868, "step": 9414 }, { "epoch": 0.8, "learning_rate": 9.799144116795184e-07, "loss": 0.7387, "step": 9415 }, { "epoch": 0.8, "learning_rate": 9.790928839061658e-07, "loss": 0.8, "step": 9416 }, { "epoch": 0.8, "learning_rate": 9.782716632697614e-07, "loss": 0.8029, "step": 9417 }, { "epoch": 0.8, "learning_rate": 9.77450749833032e-07, "loss": 0.7515, "step": 9418 }, { "epoch": 0.8, "learning_rate": 9.766301436586855e-07, "loss": 0.7913, "step": 9419 }, { "epoch": 0.8, "learning_rate": 9.758098448094028e-07, "loss": 0.7935, "step": 9420 }, { "epoch": 0.8, "learning_rate": 9.749898533478408e-07, "loss": 0.7898, "step": 9421 }, { "epoch": 0.8, "learning_rate": 9.74170169336639e-07, "loss": 0.7761, "step": 9422 }, { "epoch": 0.8, "learning_rate": 9.733507928384035e-07, "loss": 0.7716, "step": 9423 }, { "epoch": 0.8, "learning_rate": 9.72531723915726e-07, "loss": 0.7601, "step": 9424 }, { "epoch": 0.8, "learning_rate": 9.717129626311706e-07, "loss": 0.757, "step": 9425 }, { "epoch": 0.8, "learning_rate": 9.708945090472765e-07, "loss": 0.8202, "step": 9426 }, { "epoch": 0.8, "learning_rate": 9.700763632265637e-07, "loss": 0.7261, "step": 9427 }, { "epoch": 0.8, "learning_rate": 9.692585252315246e-07, "loss": 0.7738, "step": 9428 }, { "epoch": 0.8, "learning_rate": 9.68440995124631e-07, "loss": 0.767, "step": 9429 }, { "epoch": 0.8, "learning_rate": 9.676237729683275e-07, "loss": 0.8224, "step": 9430 }, { "epoch": 0.8, "learning_rate": 9.668068588250407e-07, "loss": 0.7696, "step": 9431 }, { "epoch": 0.8, "learning_rate": 9.65990252757169e-07, "loss": 0.7861, "step": 9432 }, { "epoch": 0.8, "learning_rate": 9.651739548270883e-07, "loss": 0.4915, "step": 9433 }, { "epoch": 0.81, "learning_rate": 9.64357965097153e-07, "loss": 0.7607, "step": 9434 }, { "epoch": 0.81, "learning_rate": 9.63542283629692e-07, "loss": 0.533, "step": 9435 }, { "epoch": 0.81, "learning_rate": 9.627269104870108e-07, "loss": 0.8016, "step": 9436 }, { "epoch": 0.81, "learning_rate": 9.619118457313915e-07, "loss": 0.7917, "step": 9437 }, { "epoch": 0.81, "learning_rate": 9.61097089425092e-07, "loss": 0.8086, "step": 9438 }, { "epoch": 0.81, "learning_rate": 9.602826416303496e-07, "loss": 0.5034, "step": 9439 }, { "epoch": 0.81, "learning_rate": 9.594685024093748e-07, "loss": 0.8263, "step": 9440 }, { "epoch": 0.81, "learning_rate": 9.586546718243556e-07, "loss": 0.7918, "step": 9441 }, { "epoch": 0.81, "learning_rate": 9.578411499374563e-07, "loss": 0.7709, "step": 9442 }, { "epoch": 0.81, "learning_rate": 9.570279368108166e-07, "loss": 0.8043, "step": 9443 }, { "epoch": 0.81, "learning_rate": 9.562150325065568e-07, "loss": 0.8066, "step": 9444 }, { "epoch": 0.81, "learning_rate": 9.554024370867682e-07, "loss": 0.76, "step": 9445 }, { "epoch": 0.81, "learning_rate": 9.545901506135201e-07, "loss": 0.8017, "step": 9446 }, { "epoch": 0.81, "learning_rate": 9.537781731488627e-07, "loss": 0.8065, "step": 9447 }, { "epoch": 0.81, "learning_rate": 9.529665047548159e-07, "loss": 0.807, "step": 9448 }, { "epoch": 0.81, "learning_rate": 9.521551454933797e-07, "loss": 0.8046, "step": 9449 }, { "epoch": 0.81, "learning_rate": 9.513440954265297e-07, "loss": 0.8203, "step": 9450 }, { "epoch": 0.81, "learning_rate": 9.505333546162171e-07, "loss": 0.81, "step": 9451 }, { "epoch": 0.81, "learning_rate": 9.497229231243721e-07, "loss": 0.8032, "step": 9452 }, { "epoch": 0.81, "learning_rate": 9.489128010128989e-07, "loss": 0.8349, "step": 9453 }, { "epoch": 0.81, "learning_rate": 9.481029883436771e-07, "loss": 0.7777, "step": 9454 }, { "epoch": 0.81, "learning_rate": 9.472934851785681e-07, "loss": 0.7693, "step": 9455 }, { "epoch": 0.81, "learning_rate": 9.464842915794003e-07, "loss": 0.8093, "step": 9456 }, { "epoch": 0.81, "learning_rate": 9.456754076079889e-07, "loss": 0.7941, "step": 9457 }, { "epoch": 0.81, "learning_rate": 9.44866833326118e-07, "loss": 0.8117, "step": 9458 }, { "epoch": 0.81, "learning_rate": 9.440585687955495e-07, "loss": 0.7322, "step": 9459 }, { "epoch": 0.81, "learning_rate": 9.432506140780262e-07, "loss": 0.7672, "step": 9460 }, { "epoch": 0.81, "learning_rate": 9.424429692352611e-07, "loss": 0.7584, "step": 9461 }, { "epoch": 0.81, "learning_rate": 9.416356343289468e-07, "loss": 0.7987, "step": 9462 }, { "epoch": 0.81, "learning_rate": 9.408286094207514e-07, "loss": 0.7691, "step": 9463 }, { "epoch": 0.81, "learning_rate": 9.400218945723189e-07, "loss": 0.752, "step": 9464 }, { "epoch": 0.81, "learning_rate": 9.392154898452716e-07, "loss": 0.7651, "step": 9465 }, { "epoch": 0.81, "learning_rate": 9.384093953012052e-07, "loss": 0.7361, "step": 9466 }, { "epoch": 0.81, "learning_rate": 9.37603611001695e-07, "loss": 0.8272, "step": 9467 }, { "epoch": 0.81, "learning_rate": 9.367981370082913e-07, "loss": 0.7905, "step": 9468 }, { "epoch": 0.81, "learning_rate": 9.359929733825157e-07, "loss": 0.8113, "step": 9469 }, { "epoch": 0.81, "learning_rate": 9.351881201858759e-07, "loss": 0.533, "step": 9470 }, { "epoch": 0.81, "learning_rate": 9.343835774798465e-07, "loss": 0.798, "step": 9471 }, { "epoch": 0.81, "learning_rate": 9.335793453258856e-07, "loss": 0.7863, "step": 9472 }, { "epoch": 0.81, "learning_rate": 9.327754237854237e-07, "loss": 0.8376, "step": 9473 }, { "epoch": 0.81, "learning_rate": 9.319718129198663e-07, "loss": 0.7197, "step": 9474 }, { "epoch": 0.81, "learning_rate": 9.311685127906017e-07, "loss": 0.7908, "step": 9475 }, { "epoch": 0.81, "learning_rate": 9.303655234589843e-07, "loss": 0.7991, "step": 9476 }, { "epoch": 0.81, "learning_rate": 9.295628449863547e-07, "loss": 0.813, "step": 9477 }, { "epoch": 0.81, "learning_rate": 9.287604774340236e-07, "loss": 0.8077, "step": 9478 }, { "epoch": 0.81, "learning_rate": 9.279584208632791e-07, "loss": 0.8223, "step": 9479 }, { "epoch": 0.81, "learning_rate": 9.271566753353889e-07, "loss": 0.5606, "step": 9480 }, { "epoch": 0.81, "learning_rate": 9.263552409115928e-07, "loss": 0.8158, "step": 9481 }, { "epoch": 0.81, "learning_rate": 9.25554117653108e-07, "loss": 0.8073, "step": 9482 }, { "epoch": 0.81, "learning_rate": 9.247533056211289e-07, "loss": 0.7379, "step": 9483 }, { "epoch": 0.81, "learning_rate": 9.239528048768242e-07, "loss": 0.7874, "step": 9484 }, { "epoch": 0.81, "learning_rate": 9.231526154813425e-07, "loss": 0.4994, "step": 9485 }, { "epoch": 0.81, "learning_rate": 9.223527374958046e-07, "loss": 0.7978, "step": 9486 }, { "epoch": 0.81, "learning_rate": 9.215531709813086e-07, "loss": 0.8221, "step": 9487 }, { "epoch": 0.81, "learning_rate": 9.207539159989331e-07, "loss": 0.7987, "step": 9488 }, { "epoch": 0.81, "learning_rate": 9.19954972609724e-07, "loss": 0.7739, "step": 9489 }, { "epoch": 0.81, "learning_rate": 9.191563408747117e-07, "loss": 0.796, "step": 9490 }, { "epoch": 0.81, "learning_rate": 9.183580208548998e-07, "loss": 0.7823, "step": 9491 }, { "epoch": 0.81, "learning_rate": 9.175600126112649e-07, "loss": 0.8005, "step": 9492 }, { "epoch": 0.81, "learning_rate": 9.167623162047673e-07, "loss": 0.7553, "step": 9493 }, { "epoch": 0.81, "learning_rate": 9.159649316963348e-07, "loss": 0.762, "step": 9494 }, { "epoch": 0.81, "learning_rate": 9.15167859146881e-07, "loss": 0.799, "step": 9495 }, { "epoch": 0.81, "learning_rate": 9.143710986172849e-07, "loss": 0.7643, "step": 9496 }, { "epoch": 0.81, "learning_rate": 9.135746501684084e-07, "loss": 0.8043, "step": 9497 }, { "epoch": 0.81, "learning_rate": 9.1277851386109e-07, "loss": 0.766, "step": 9498 }, { "epoch": 0.81, "learning_rate": 9.119826897561402e-07, "loss": 0.8403, "step": 9499 }, { "epoch": 0.81, "learning_rate": 9.111871779143505e-07, "loss": 0.7804, "step": 9500 }, { "epoch": 0.81, "learning_rate": 9.10391978396486e-07, "loss": 0.8229, "step": 9501 }, { "epoch": 0.81, "learning_rate": 9.09597091263284e-07, "loss": 0.841, "step": 9502 }, { "epoch": 0.81, "learning_rate": 9.088025165754666e-07, "loss": 0.7928, "step": 9503 }, { "epoch": 0.81, "learning_rate": 9.080082543937241e-07, "loss": 0.763, "step": 9504 }, { "epoch": 0.81, "learning_rate": 9.07214304778729e-07, "loss": 0.7441, "step": 9505 }, { "epoch": 0.81, "learning_rate": 9.064206677911258e-07, "loss": 0.7832, "step": 9506 }, { "epoch": 0.81, "learning_rate": 9.056273434915353e-07, "loss": 0.7952, "step": 9507 }, { "epoch": 0.81, "learning_rate": 9.04834331940559e-07, "loss": 0.4727, "step": 9508 }, { "epoch": 0.81, "learning_rate": 9.040416331987661e-07, "loss": 0.7813, "step": 9509 }, { "epoch": 0.81, "learning_rate": 9.032492473267113e-07, "loss": 0.7691, "step": 9510 }, { "epoch": 0.81, "learning_rate": 9.024571743849187e-07, "loss": 0.7851, "step": 9511 }, { "epoch": 0.81, "learning_rate": 9.016654144338904e-07, "loss": 0.8603, "step": 9512 }, { "epoch": 0.81, "learning_rate": 9.008739675341077e-07, "loss": 0.8502, "step": 9513 }, { "epoch": 0.81, "learning_rate": 9.000828337460226e-07, "loss": 0.7995, "step": 9514 }, { "epoch": 0.81, "learning_rate": 8.992920131300676e-07, "loss": 0.8302, "step": 9515 }, { "epoch": 0.81, "learning_rate": 8.985015057466484e-07, "loss": 0.8157, "step": 9516 }, { "epoch": 0.81, "learning_rate": 8.977113116561464e-07, "loss": 0.801, "step": 9517 }, { "epoch": 0.81, "learning_rate": 8.969214309189239e-07, "loss": 0.8237, "step": 9518 }, { "epoch": 0.81, "learning_rate": 8.961318635953148e-07, "loss": 0.8221, "step": 9519 }, { "epoch": 0.81, "learning_rate": 8.953426097456281e-07, "loss": 0.7428, "step": 9520 }, { "epoch": 0.81, "learning_rate": 8.945536694301548e-07, "loss": 0.8179, "step": 9521 }, { "epoch": 0.81, "learning_rate": 8.937650427091537e-07, "loss": 0.7976, "step": 9522 }, { "epoch": 0.81, "learning_rate": 8.929767296428677e-07, "loss": 0.8145, "step": 9523 }, { "epoch": 0.81, "learning_rate": 8.921887302915105e-07, "loss": 0.7993, "step": 9524 }, { "epoch": 0.81, "learning_rate": 8.914010447152721e-07, "loss": 0.8053, "step": 9525 }, { "epoch": 0.81, "learning_rate": 8.906136729743226e-07, "loss": 0.8041, "step": 9526 }, { "epoch": 0.81, "learning_rate": 8.898266151288043e-07, "loss": 0.8646, "step": 9527 }, { "epoch": 0.81, "learning_rate": 8.890398712388365e-07, "loss": 0.7721, "step": 9528 }, { "epoch": 0.81, "learning_rate": 8.882534413645139e-07, "loss": 0.7905, "step": 9529 }, { "epoch": 0.81, "learning_rate": 8.874673255659077e-07, "loss": 0.7822, "step": 9530 }, { "epoch": 0.81, "learning_rate": 8.866815239030674e-07, "loss": 0.7846, "step": 9531 }, { "epoch": 0.81, "learning_rate": 8.858960364360142e-07, "loss": 0.8372, "step": 9532 }, { "epoch": 0.81, "learning_rate": 8.851108632247496e-07, "loss": 0.8098, "step": 9533 }, { "epoch": 0.81, "learning_rate": 8.843260043292496e-07, "loss": 0.7952, "step": 9534 }, { "epoch": 0.81, "learning_rate": 8.83541459809461e-07, "loss": 0.8103, "step": 9535 }, { "epoch": 0.81, "learning_rate": 8.827572297253162e-07, "loss": 0.8115, "step": 9536 }, { "epoch": 0.81, "learning_rate": 8.819733141367154e-07, "loss": 0.8145, "step": 9537 }, { "epoch": 0.81, "learning_rate": 8.811897131035402e-07, "loss": 0.7681, "step": 9538 }, { "epoch": 0.81, "learning_rate": 8.804064266856454e-07, "loss": 0.7739, "step": 9539 }, { "epoch": 0.81, "learning_rate": 8.796234549428606e-07, "loss": 0.7369, "step": 9540 }, { "epoch": 0.81, "learning_rate": 8.788407979349966e-07, "loss": 0.777, "step": 9541 }, { "epoch": 0.81, "learning_rate": 8.780584557218324e-07, "loss": 0.8193, "step": 9542 }, { "epoch": 0.81, "learning_rate": 8.772764283631307e-07, "loss": 0.755, "step": 9543 }, { "epoch": 0.81, "learning_rate": 8.764947159186249e-07, "loss": 0.7929, "step": 9544 }, { "epoch": 0.81, "learning_rate": 8.757133184480254e-07, "loss": 0.7928, "step": 9545 }, { "epoch": 0.81, "learning_rate": 8.749322360110218e-07, "loss": 0.8003, "step": 9546 }, { "epoch": 0.81, "learning_rate": 8.741514686672753e-07, "loss": 0.7757, "step": 9547 }, { "epoch": 0.81, "learning_rate": 8.733710164764253e-07, "loss": 0.813, "step": 9548 }, { "epoch": 0.81, "learning_rate": 8.725908794980864e-07, "loss": 0.7822, "step": 9549 }, { "epoch": 0.81, "learning_rate": 8.718110577918476e-07, "loss": 0.8189, "step": 9550 }, { "epoch": 0.82, "learning_rate": 8.71031551417279e-07, "loss": 0.7703, "step": 9551 }, { "epoch": 0.82, "learning_rate": 8.702523604339219e-07, "loss": 0.7666, "step": 9552 }, { "epoch": 0.82, "learning_rate": 8.694734849012926e-07, "loss": 0.7719, "step": 9553 }, { "epoch": 0.82, "learning_rate": 8.6869492487889e-07, "loss": 0.8054, "step": 9554 }, { "epoch": 0.82, "learning_rate": 8.679166804261796e-07, "loss": 0.8089, "step": 9555 }, { "epoch": 0.82, "learning_rate": 8.671387516026103e-07, "loss": 0.8187, "step": 9556 }, { "epoch": 0.82, "learning_rate": 8.66361138467604e-07, "loss": 0.8256, "step": 9557 }, { "epoch": 0.82, "learning_rate": 8.655838410805572e-07, "loss": 0.7651, "step": 9558 }, { "epoch": 0.82, "learning_rate": 8.648068595008458e-07, "loss": 0.8135, "step": 9559 }, { "epoch": 0.82, "learning_rate": 8.640301937878187e-07, "loss": 0.7737, "step": 9560 }, { "epoch": 0.82, "learning_rate": 8.632538440008009e-07, "loss": 0.7735, "step": 9561 }, { "epoch": 0.82, "learning_rate": 8.62477810199095e-07, "loss": 0.7715, "step": 9562 }, { "epoch": 0.82, "learning_rate": 8.617020924419761e-07, "loss": 0.7795, "step": 9563 }, { "epoch": 0.82, "learning_rate": 8.609266907887004e-07, "loss": 0.807, "step": 9564 }, { "epoch": 0.82, "learning_rate": 8.601516052984949e-07, "loss": 0.7891, "step": 9565 }, { "epoch": 0.82, "learning_rate": 8.593768360305644e-07, "loss": 0.8262, "step": 9566 }, { "epoch": 0.82, "learning_rate": 8.586023830440926e-07, "loss": 0.8331, "step": 9567 }, { "epoch": 0.82, "learning_rate": 8.578282463982318e-07, "loss": 0.8233, "step": 9568 }, { "epoch": 0.82, "learning_rate": 8.570544261521175e-07, "loss": 0.8288, "step": 9569 }, { "epoch": 0.82, "learning_rate": 8.562809223648566e-07, "loss": 0.8118, "step": 9570 }, { "epoch": 0.82, "learning_rate": 8.555077350955321e-07, "loss": 0.7849, "step": 9571 }, { "epoch": 0.82, "learning_rate": 8.547348644032072e-07, "loss": 0.8009, "step": 9572 }, { "epoch": 0.82, "learning_rate": 8.539623103469135e-07, "loss": 0.791, "step": 9573 }, { "epoch": 0.82, "learning_rate": 8.531900729856679e-07, "loss": 0.7876, "step": 9574 }, { "epoch": 0.82, "learning_rate": 8.52418152378452e-07, "loss": 0.8026, "step": 9575 }, { "epoch": 0.82, "learning_rate": 8.516465485842329e-07, "loss": 0.7817, "step": 9576 }, { "epoch": 0.82, "learning_rate": 8.508752616619481e-07, "loss": 0.7524, "step": 9577 }, { "epoch": 0.82, "learning_rate": 8.50104291670511e-07, "loss": 0.5061, "step": 9578 }, { "epoch": 0.82, "learning_rate": 8.49333638668815e-07, "loss": 0.8418, "step": 9579 }, { "epoch": 0.82, "learning_rate": 8.485633027157247e-07, "loss": 0.8156, "step": 9580 }, { "epoch": 0.82, "learning_rate": 8.477932838700826e-07, "loss": 0.7916, "step": 9581 }, { "epoch": 0.82, "learning_rate": 8.470235821907064e-07, "loss": 0.7565, "step": 9582 }, { "epoch": 0.82, "learning_rate": 8.462541977363887e-07, "loss": 0.7526, "step": 9583 }, { "epoch": 0.82, "learning_rate": 8.454851305659012e-07, "loss": 0.8224, "step": 9584 }, { "epoch": 0.82, "learning_rate": 8.447163807379877e-07, "loss": 0.7697, "step": 9585 }, { "epoch": 0.82, "learning_rate": 8.439479483113683e-07, "loss": 0.7785, "step": 9586 }, { "epoch": 0.82, "learning_rate": 8.431798333447433e-07, "loss": 0.7982, "step": 9587 }, { "epoch": 0.82, "learning_rate": 8.424120358967803e-07, "loss": 0.7717, "step": 9588 }, { "epoch": 0.82, "learning_rate": 8.41644556026131e-07, "loss": 0.8179, "step": 9589 }, { "epoch": 0.82, "learning_rate": 8.408773937914183e-07, "loss": 0.7779, "step": 9590 }, { "epoch": 0.82, "learning_rate": 8.401105492512401e-07, "loss": 0.8124, "step": 9591 }, { "epoch": 0.82, "learning_rate": 8.393440224641747e-07, "loss": 0.8033, "step": 9592 }, { "epoch": 0.82, "learning_rate": 8.385778134887723e-07, "loss": 0.8243, "step": 9593 }, { "epoch": 0.82, "learning_rate": 8.378119223835591e-07, "loss": 0.7908, "step": 9594 }, { "epoch": 0.82, "learning_rate": 8.370463492070386e-07, "loss": 0.8459, "step": 9595 }, { "epoch": 0.82, "learning_rate": 8.362810940176869e-07, "loss": 0.804, "step": 9596 }, { "epoch": 0.82, "learning_rate": 8.355161568739611e-07, "loss": 0.8019, "step": 9597 }, { "epoch": 0.82, "learning_rate": 8.347515378342891e-07, "loss": 0.8222, "step": 9598 }, { "epoch": 0.82, "learning_rate": 8.339872369570756e-07, "loss": 0.7771, "step": 9599 }, { "epoch": 0.82, "learning_rate": 8.332232543007041e-07, "loss": 0.7336, "step": 9600 }, { "epoch": 0.82, "learning_rate": 8.324595899235305e-07, "loss": 0.8367, "step": 9601 }, { "epoch": 0.82, "learning_rate": 8.31696243883886e-07, "loss": 0.7751, "step": 9602 }, { "epoch": 0.82, "learning_rate": 8.309332162400808e-07, "loss": 0.8169, "step": 9603 }, { "epoch": 0.82, "learning_rate": 8.301705070503957e-07, "loss": 0.7746, "step": 9604 }, { "epoch": 0.82, "learning_rate": 8.294081163730933e-07, "loss": 0.7887, "step": 9605 }, { "epoch": 0.82, "learning_rate": 8.286460442664068e-07, "loss": 0.8256, "step": 9606 }, { "epoch": 0.82, "learning_rate": 8.278842907885503e-07, "loss": 0.7941, "step": 9607 }, { "epoch": 0.82, "learning_rate": 8.271228559977068e-07, "loss": 0.7558, "step": 9608 }, { "epoch": 0.82, "learning_rate": 8.26361739952038e-07, "loss": 0.518, "step": 9609 }, { "epoch": 0.82, "learning_rate": 8.256009427096851e-07, "loss": 0.786, "step": 9610 }, { "epoch": 0.82, "learning_rate": 8.248404643287589e-07, "loss": 0.806, "step": 9611 }, { "epoch": 0.82, "learning_rate": 8.240803048673501e-07, "loss": 0.8157, "step": 9612 }, { "epoch": 0.82, "learning_rate": 8.233204643835235e-07, "loss": 0.7605, "step": 9613 }, { "epoch": 0.82, "learning_rate": 8.225609429353187e-07, "loss": 0.7919, "step": 9614 }, { "epoch": 0.82, "learning_rate": 8.218017405807516e-07, "loss": 0.7643, "step": 9615 }, { "epoch": 0.82, "learning_rate": 8.210428573778134e-07, "loss": 0.8056, "step": 9616 }, { "epoch": 0.82, "learning_rate": 8.202842933844734e-07, "loss": 0.7968, "step": 9617 }, { "epoch": 0.82, "learning_rate": 8.195260486586731e-07, "loss": 0.5176, "step": 9618 }, { "epoch": 0.82, "learning_rate": 8.1876812325833e-07, "loss": 0.7766, "step": 9619 }, { "epoch": 0.82, "learning_rate": 8.18010517241341e-07, "loss": 0.7932, "step": 9620 }, { "epoch": 0.82, "learning_rate": 8.17253230665574e-07, "loss": 0.7443, "step": 9621 }, { "epoch": 0.82, "learning_rate": 8.164962635888746e-07, "loss": 0.7469, "step": 9622 }, { "epoch": 0.82, "learning_rate": 8.157396160690634e-07, "loss": 0.8048, "step": 9623 }, { "epoch": 0.82, "learning_rate": 8.149832881639363e-07, "loss": 0.8136, "step": 9624 }, { "epoch": 0.82, "learning_rate": 8.142272799312679e-07, "loss": 0.8318, "step": 9625 }, { "epoch": 0.82, "learning_rate": 8.134715914288039e-07, "loss": 0.7947, "step": 9626 }, { "epoch": 0.82, "learning_rate": 8.127162227142682e-07, "loss": 0.8222, "step": 9627 }, { "epoch": 0.82, "learning_rate": 8.119611738453592e-07, "loss": 0.758, "step": 9628 }, { "epoch": 0.82, "learning_rate": 8.112064448797502e-07, "loss": 0.7937, "step": 9629 }, { "epoch": 0.82, "learning_rate": 8.10452035875094e-07, "loss": 0.7931, "step": 9630 }, { "epoch": 0.82, "learning_rate": 8.096979468890142e-07, "loss": 0.8014, "step": 9631 }, { "epoch": 0.82, "learning_rate": 8.089441779791113e-07, "loss": 0.4884, "step": 9632 }, { "epoch": 0.82, "learning_rate": 8.081907292029639e-07, "loss": 0.8409, "step": 9633 }, { "epoch": 0.82, "learning_rate": 8.074376006181234e-07, "loss": 0.766, "step": 9634 }, { "epoch": 0.82, "learning_rate": 8.066847922821175e-07, "loss": 0.7732, "step": 9635 }, { "epoch": 0.82, "learning_rate": 8.059323042524486e-07, "loss": 0.8161, "step": 9636 }, { "epoch": 0.82, "learning_rate": 8.051801365865958e-07, "loss": 0.79, "step": 9637 }, { "epoch": 0.82, "learning_rate": 8.044282893420141e-07, "loss": 0.7735, "step": 9638 }, { "epoch": 0.82, "learning_rate": 8.036767625761321e-07, "loss": 0.7957, "step": 9639 }, { "epoch": 0.82, "learning_rate": 8.029255563463589e-07, "loss": 0.8059, "step": 9640 }, { "epoch": 0.82, "learning_rate": 8.021746707100708e-07, "loss": 0.7682, "step": 9641 }, { "epoch": 0.82, "learning_rate": 8.014241057246247e-07, "loss": 0.8004, "step": 9642 }, { "epoch": 0.82, "learning_rate": 8.006738614473552e-07, "loss": 0.81, "step": 9643 }, { "epoch": 0.82, "learning_rate": 7.999239379355672e-07, "loss": 0.8399, "step": 9644 }, { "epoch": 0.82, "learning_rate": 7.991743352465453e-07, "loss": 0.739, "step": 9645 }, { "epoch": 0.82, "learning_rate": 7.984250534375477e-07, "loss": 0.8063, "step": 9646 }, { "epoch": 0.82, "learning_rate": 7.976760925658078e-07, "loss": 0.7895, "step": 9647 }, { "epoch": 0.82, "learning_rate": 7.969274526885346e-07, "loss": 0.8329, "step": 9648 }, { "epoch": 0.82, "learning_rate": 7.961791338629127e-07, "loss": 0.7971, "step": 9649 }, { "epoch": 0.82, "learning_rate": 7.954311361461037e-07, "loss": 0.7743, "step": 9650 }, { "epoch": 0.82, "learning_rate": 7.94683459595243e-07, "loss": 0.8321, "step": 9651 }, { "epoch": 0.82, "learning_rate": 7.939361042674404e-07, "loss": 0.7823, "step": 9652 }, { "epoch": 0.82, "learning_rate": 7.931890702197848e-07, "loss": 0.7701, "step": 9653 }, { "epoch": 0.82, "learning_rate": 7.924423575093376e-07, "loss": 0.7918, "step": 9654 }, { "epoch": 0.82, "learning_rate": 7.916959661931367e-07, "loss": 0.8011, "step": 9655 }, { "epoch": 0.82, "learning_rate": 7.909498963281942e-07, "loss": 0.8274, "step": 9656 }, { "epoch": 0.82, "learning_rate": 7.902041479714984e-07, "loss": 0.8591, "step": 9657 }, { "epoch": 0.82, "learning_rate": 7.894587211800154e-07, "loss": 0.7289, "step": 9658 }, { "epoch": 0.82, "learning_rate": 7.887136160106829e-07, "loss": 0.7869, "step": 9659 }, { "epoch": 0.82, "learning_rate": 7.87968832520416e-07, "loss": 0.7765, "step": 9660 }, { "epoch": 0.82, "learning_rate": 7.872243707661054e-07, "loss": 0.8295, "step": 9661 }, { "epoch": 0.82, "learning_rate": 7.864802308046154e-07, "loss": 0.8223, "step": 9662 }, { "epoch": 0.82, "learning_rate": 7.857364126927896e-07, "loss": 0.8165, "step": 9663 }, { "epoch": 0.82, "learning_rate": 7.849929164874426e-07, "loss": 0.8157, "step": 9664 }, { "epoch": 0.82, "learning_rate": 7.842497422453666e-07, "loss": 0.8046, "step": 9665 }, { "epoch": 0.82, "learning_rate": 7.8350689002333e-07, "loss": 0.8024, "step": 9666 }, { "epoch": 0.82, "learning_rate": 7.827643598780748e-07, "loss": 0.7452, "step": 9667 }, { "epoch": 0.82, "learning_rate": 7.82022151866319e-07, "loss": 0.8312, "step": 9668 }, { "epoch": 0.83, "learning_rate": 7.812802660447571e-07, "loss": 0.8044, "step": 9669 }, { "epoch": 0.83, "learning_rate": 7.805387024700556e-07, "loss": 0.8615, "step": 9670 }, { "epoch": 0.83, "learning_rate": 7.797974611988618e-07, "loss": 0.7319, "step": 9671 }, { "epoch": 0.83, "learning_rate": 7.790565422877944e-07, "loss": 0.714, "step": 9672 }, { "epoch": 0.83, "learning_rate": 7.78315945793447e-07, "loss": 0.809, "step": 9673 }, { "epoch": 0.83, "learning_rate": 7.775756717723932e-07, "loss": 0.8174, "step": 9674 }, { "epoch": 0.83, "learning_rate": 7.768357202811749e-07, "loss": 0.8075, "step": 9675 }, { "epoch": 0.83, "learning_rate": 7.760960913763166e-07, "loss": 0.7995, "step": 9676 }, { "epoch": 0.83, "learning_rate": 7.753567851143124e-07, "loss": 0.7707, "step": 9677 }, { "epoch": 0.83, "learning_rate": 7.746178015516359e-07, "loss": 0.7788, "step": 9678 }, { "epoch": 0.83, "learning_rate": 7.738791407447344e-07, "loss": 0.8139, "step": 9679 }, { "epoch": 0.83, "learning_rate": 7.731408027500298e-07, "loss": 0.7944, "step": 9680 }, { "epoch": 0.83, "learning_rate": 7.724027876239199e-07, "loss": 0.825, "step": 9681 }, { "epoch": 0.83, "learning_rate": 7.716650954227772e-07, "loss": 0.5006, "step": 9682 }, { "epoch": 0.83, "learning_rate": 7.709277262029524e-07, "loss": 0.8425, "step": 9683 }, { "epoch": 0.83, "learning_rate": 7.701906800207687e-07, "loss": 0.7776, "step": 9684 }, { "epoch": 0.83, "learning_rate": 7.694539569325232e-07, "loss": 0.8329, "step": 9685 }, { "epoch": 0.83, "learning_rate": 7.687175569944938e-07, "loss": 0.7868, "step": 9686 }, { "epoch": 0.83, "learning_rate": 7.679814802629292e-07, "loss": 0.8447, "step": 9687 }, { "epoch": 0.83, "learning_rate": 7.672457267940542e-07, "loss": 0.8245, "step": 9688 }, { "epoch": 0.83, "learning_rate": 7.665102966440691e-07, "loss": 0.761, "step": 9689 }, { "epoch": 0.83, "learning_rate": 7.65775189869149e-07, "loss": 0.7754, "step": 9690 }, { "epoch": 0.83, "learning_rate": 7.650404065254474e-07, "loss": 0.8448, "step": 9691 }, { "epoch": 0.83, "learning_rate": 7.6430594666909e-07, "loss": 0.7881, "step": 9692 }, { "epoch": 0.83, "learning_rate": 7.635718103561768e-07, "loss": 0.7699, "step": 9693 }, { "epoch": 0.83, "learning_rate": 7.628379976427868e-07, "loss": 0.8032, "step": 9694 }, { "epoch": 0.83, "learning_rate": 7.621045085849699e-07, "loss": 0.7939, "step": 9695 }, { "epoch": 0.83, "learning_rate": 7.61371343238756e-07, "loss": 0.7946, "step": 9696 }, { "epoch": 0.83, "learning_rate": 7.606385016601475e-07, "loss": 0.7185, "step": 9697 }, { "epoch": 0.83, "learning_rate": 7.599059839051209e-07, "loss": 0.816, "step": 9698 }, { "epoch": 0.83, "learning_rate": 7.591737900296314e-07, "loss": 0.8019, "step": 9699 }, { "epoch": 0.83, "learning_rate": 7.584419200896075e-07, "loss": 0.7644, "step": 9700 }, { "epoch": 0.83, "learning_rate": 7.577103741409525e-07, "loss": 0.7911, "step": 9701 }, { "epoch": 0.83, "learning_rate": 7.569791522395454e-07, "loss": 0.8108, "step": 9702 }, { "epoch": 0.83, "learning_rate": 7.562482544412398e-07, "loss": 0.8154, "step": 9703 }, { "epoch": 0.83, "learning_rate": 7.555176808018666e-07, "loss": 0.5299, "step": 9704 }, { "epoch": 0.83, "learning_rate": 7.547874313772313e-07, "loss": 0.8117, "step": 9705 }, { "epoch": 0.83, "learning_rate": 7.540575062231115e-07, "loss": 0.4741, "step": 9706 }, { "epoch": 0.83, "learning_rate": 7.533279053952664e-07, "loss": 0.7831, "step": 9707 }, { "epoch": 0.83, "learning_rate": 7.525986289494213e-07, "loss": 0.7623, "step": 9708 }, { "epoch": 0.83, "learning_rate": 7.518696769412864e-07, "loss": 0.7881, "step": 9709 }, { "epoch": 0.83, "learning_rate": 7.511410494265415e-07, "loss": 0.7926, "step": 9710 }, { "epoch": 0.83, "learning_rate": 7.504127464608413e-07, "loss": 0.8483, "step": 9711 }, { "epoch": 0.83, "learning_rate": 7.496847680998187e-07, "loss": 0.7613, "step": 9712 }, { "epoch": 0.83, "learning_rate": 7.48957114399081e-07, "loss": 0.7805, "step": 9713 }, { "epoch": 0.83, "learning_rate": 7.482297854142085e-07, "loss": 0.7824, "step": 9714 }, { "epoch": 0.83, "learning_rate": 7.475027812007574e-07, "loss": 0.7825, "step": 9715 }, { "epoch": 0.83, "learning_rate": 7.467761018142627e-07, "loss": 0.812, "step": 9716 }, { "epoch": 0.83, "learning_rate": 7.460497473102301e-07, "loss": 0.8153, "step": 9717 }, { "epoch": 0.83, "learning_rate": 7.453237177441414e-07, "loss": 0.8212, "step": 9718 }, { "epoch": 0.83, "learning_rate": 7.44598013171457e-07, "loss": 0.503, "step": 9719 }, { "epoch": 0.83, "learning_rate": 7.438726336476076e-07, "loss": 0.761, "step": 9720 }, { "epoch": 0.83, "learning_rate": 7.431475792280018e-07, "loss": 0.81, "step": 9721 }, { "epoch": 0.83, "learning_rate": 7.424228499680236e-07, "loss": 0.7864, "step": 9722 }, { "epoch": 0.83, "learning_rate": 7.416984459230292e-07, "loss": 0.8062, "step": 9723 }, { "epoch": 0.83, "learning_rate": 7.409743671483549e-07, "loss": 0.8228, "step": 9724 }, { "epoch": 0.83, "learning_rate": 7.402506136993087e-07, "loss": 0.8094, "step": 9725 }, { "epoch": 0.83, "learning_rate": 7.395271856311731e-07, "loss": 0.8184, "step": 9726 }, { "epoch": 0.83, "learning_rate": 7.388040829992105e-07, "loss": 0.7895, "step": 9727 }, { "epoch": 0.83, "learning_rate": 7.3808130585865e-07, "loss": 0.7718, "step": 9728 }, { "epoch": 0.83, "learning_rate": 7.373588542647054e-07, "loss": 0.8009, "step": 9729 }, { "epoch": 0.83, "learning_rate": 7.366367282725595e-07, "loss": 0.7554, "step": 9730 }, { "epoch": 0.83, "learning_rate": 7.3591492793737e-07, "loss": 0.7594, "step": 9731 }, { "epoch": 0.83, "learning_rate": 7.351934533142757e-07, "loss": 0.8147, "step": 9732 }, { "epoch": 0.83, "learning_rate": 7.344723044583835e-07, "loss": 0.7362, "step": 9733 }, { "epoch": 0.83, "learning_rate": 7.337514814247792e-07, "loss": 0.7522, "step": 9734 }, { "epoch": 0.83, "learning_rate": 7.33030984268523e-07, "loss": 0.805, "step": 9735 }, { "epoch": 0.83, "learning_rate": 7.323108130446483e-07, "loss": 0.7766, "step": 9736 }, { "epoch": 0.83, "learning_rate": 7.315909678081679e-07, "loss": 0.7675, "step": 9737 }, { "epoch": 0.83, "learning_rate": 7.308714486140666e-07, "loss": 0.7669, "step": 9738 }, { "epoch": 0.83, "learning_rate": 7.30152255517303e-07, "loss": 0.8022, "step": 9739 }, { "epoch": 0.83, "learning_rate": 7.294333885728161e-07, "loss": 0.8276, "step": 9740 }, { "epoch": 0.83, "learning_rate": 7.287148478355127e-07, "loss": 0.7436, "step": 9741 }, { "epoch": 0.83, "learning_rate": 7.279966333602811e-07, "loss": 0.7269, "step": 9742 }, { "epoch": 0.83, "learning_rate": 7.272787452019813e-07, "loss": 0.7969, "step": 9743 }, { "epoch": 0.83, "learning_rate": 7.265611834154485e-07, "loss": 0.8277, "step": 9744 }, { "epoch": 0.83, "learning_rate": 7.258439480554952e-07, "loss": 0.8085, "step": 9745 }, { "epoch": 0.83, "learning_rate": 7.251270391769056e-07, "loss": 0.7677, "step": 9746 }, { "epoch": 0.83, "learning_rate": 7.244104568344439e-07, "loss": 0.83, "step": 9747 }, { "epoch": 0.83, "learning_rate": 7.23694201082843e-07, "loss": 0.8408, "step": 9748 }, { "epoch": 0.83, "learning_rate": 7.229782719768136e-07, "loss": 0.7642, "step": 9749 }, { "epoch": 0.83, "learning_rate": 7.222626695710455e-07, "loss": 0.8108, "step": 9750 }, { "epoch": 0.83, "learning_rate": 7.215473939201961e-07, "loss": 0.7599, "step": 9751 }, { "epoch": 0.83, "learning_rate": 7.20832445078905e-07, "loss": 0.8189, "step": 9752 }, { "epoch": 0.83, "learning_rate": 7.201178231017836e-07, "loss": 0.8242, "step": 9753 }, { "epoch": 0.83, "learning_rate": 7.194035280434142e-07, "loss": 0.7934, "step": 9754 }, { "epoch": 0.83, "learning_rate": 7.186895599583621e-07, "loss": 0.7659, "step": 9755 }, { "epoch": 0.83, "learning_rate": 7.179759189011615e-07, "loss": 0.7914, "step": 9756 }, { "epoch": 0.83, "learning_rate": 7.172626049263259e-07, "loss": 0.7929, "step": 9757 }, { "epoch": 0.83, "learning_rate": 7.165496180883413e-07, "loss": 0.8334, "step": 9758 }, { "epoch": 0.83, "learning_rate": 7.158369584416669e-07, "loss": 0.7469, "step": 9759 }, { "epoch": 0.83, "learning_rate": 7.151246260407435e-07, "loss": 0.7896, "step": 9760 }, { "epoch": 0.83, "learning_rate": 7.144126209399777e-07, "loss": 0.7815, "step": 9761 }, { "epoch": 0.83, "learning_rate": 7.137009431937597e-07, "loss": 0.7785, "step": 9762 }, { "epoch": 0.83, "learning_rate": 7.129895928564496e-07, "loss": 0.7892, "step": 9763 }, { "epoch": 0.83, "learning_rate": 7.122785699823831e-07, "loss": 0.7989, "step": 9764 }, { "epoch": 0.83, "learning_rate": 7.115678746258731e-07, "loss": 0.794, "step": 9765 }, { "epoch": 0.83, "learning_rate": 7.108575068412061e-07, "loss": 0.7804, "step": 9766 }, { "epoch": 0.83, "learning_rate": 7.101474666826425e-07, "loss": 0.7982, "step": 9767 }, { "epoch": 0.83, "learning_rate": 7.09437754204419e-07, "loss": 0.7858, "step": 9768 }, { "epoch": 0.83, "learning_rate": 7.087283694607461e-07, "loss": 0.8125, "step": 9769 }, { "epoch": 0.83, "learning_rate": 7.080193125058127e-07, "loss": 0.5685, "step": 9770 }, { "epoch": 0.83, "learning_rate": 7.073105833937783e-07, "loss": 0.7665, "step": 9771 }, { "epoch": 0.83, "learning_rate": 7.066021821787783e-07, "loss": 0.7148, "step": 9772 }, { "epoch": 0.83, "learning_rate": 7.058941089149274e-07, "loss": 0.8369, "step": 9773 }, { "epoch": 0.83, "learning_rate": 7.051863636563067e-07, "loss": 0.7951, "step": 9774 }, { "epoch": 0.83, "learning_rate": 7.044789464569817e-07, "loss": 0.7787, "step": 9775 }, { "epoch": 0.83, "learning_rate": 7.037718573709868e-07, "loss": 0.8202, "step": 9776 }, { "epoch": 0.83, "learning_rate": 7.030650964523317e-07, "loss": 0.8389, "step": 9777 }, { "epoch": 0.83, "learning_rate": 7.023586637550045e-07, "loss": 0.7814, "step": 9778 }, { "epoch": 0.83, "learning_rate": 7.016525593329643e-07, "loss": 0.8057, "step": 9779 }, { "epoch": 0.83, "learning_rate": 7.009467832401501e-07, "loss": 0.7524, "step": 9780 }, { "epoch": 0.83, "learning_rate": 7.002413355304688e-07, "loss": 0.809, "step": 9781 }, { "epoch": 0.83, "learning_rate": 6.995362162578068e-07, "loss": 0.8159, "step": 9782 }, { "epoch": 0.83, "learning_rate": 6.988314254760259e-07, "loss": 0.7032, "step": 9783 }, { "epoch": 0.83, "learning_rate": 6.981269632389603e-07, "loss": 0.777, "step": 9784 }, { "epoch": 0.83, "learning_rate": 6.974228296004221e-07, "loss": 0.4743, "step": 9785 }, { "epoch": 0.84, "learning_rate": 6.967190246141958e-07, "loss": 0.5285, "step": 9786 }, { "epoch": 0.84, "learning_rate": 6.960155483340397e-07, "loss": 0.8028, "step": 9787 }, { "epoch": 0.84, "learning_rate": 6.953124008136908e-07, "loss": 0.8002, "step": 9788 }, { "epoch": 0.84, "learning_rate": 6.946095821068577e-07, "loss": 0.8224, "step": 9789 }, { "epoch": 0.84, "learning_rate": 6.93907092267227e-07, "loss": 0.76, "step": 9790 }, { "epoch": 0.84, "learning_rate": 6.932049313484574e-07, "loss": 0.7541, "step": 9791 }, { "epoch": 0.84, "learning_rate": 6.925030994041826e-07, "loss": 0.7517, "step": 9792 }, { "epoch": 0.84, "learning_rate": 6.918015964880148e-07, "loss": 0.7914, "step": 9793 }, { "epoch": 0.84, "learning_rate": 6.91100422653534e-07, "loss": 0.7706, "step": 9794 }, { "epoch": 0.84, "learning_rate": 6.903995779543027e-07, "loss": 0.7676, "step": 9795 }, { "epoch": 0.84, "learning_rate": 6.896990624438543e-07, "loss": 0.8232, "step": 9796 }, { "epoch": 0.84, "learning_rate": 6.889988761756961e-07, "loss": 0.7608, "step": 9797 }, { "epoch": 0.84, "learning_rate": 6.88299019203314e-07, "loss": 0.7969, "step": 9798 }, { "epoch": 0.84, "learning_rate": 6.875994915801654e-07, "loss": 0.8426, "step": 9799 }, { "epoch": 0.84, "learning_rate": 6.869002933596841e-07, "loss": 0.7526, "step": 9800 }, { "epoch": 0.84, "learning_rate": 6.862014245952781e-07, "loss": 0.8032, "step": 9801 }, { "epoch": 0.84, "learning_rate": 6.855028853403295e-07, "loss": 0.778, "step": 9802 }, { "epoch": 0.84, "learning_rate": 6.84804675648198e-07, "loss": 0.7836, "step": 9803 }, { "epoch": 0.84, "learning_rate": 6.841067955722158e-07, "loss": 0.7814, "step": 9804 }, { "epoch": 0.84, "learning_rate": 6.834092451656892e-07, "loss": 0.7844, "step": 9805 }, { "epoch": 0.84, "learning_rate": 6.827120244819036e-07, "loss": 0.7765, "step": 9806 }, { "epoch": 0.84, "learning_rate": 6.820151335741121e-07, "loss": 0.7613, "step": 9807 }, { "epoch": 0.84, "learning_rate": 6.813185724955501e-07, "loss": 0.7382, "step": 9808 }, { "epoch": 0.84, "learning_rate": 6.806223412994228e-07, "loss": 0.7673, "step": 9809 }, { "epoch": 0.84, "learning_rate": 6.79926440038911e-07, "loss": 0.5911, "step": 9810 }, { "epoch": 0.84, "learning_rate": 6.792308687671739e-07, "loss": 0.7943, "step": 9811 }, { "epoch": 0.84, "learning_rate": 6.785356275373406e-07, "loss": 0.8035, "step": 9812 }, { "epoch": 0.84, "learning_rate": 6.778407164025175e-07, "loss": 0.4878, "step": 9813 }, { "epoch": 0.84, "learning_rate": 6.771461354157854e-07, "loss": 0.7643, "step": 9814 }, { "epoch": 0.84, "learning_rate": 6.764518846301987e-07, "loss": 0.7631, "step": 9815 }, { "epoch": 0.84, "learning_rate": 6.757579640987904e-07, "loss": 0.8111, "step": 9816 }, { "epoch": 0.84, "learning_rate": 6.750643738745627e-07, "loss": 0.7685, "step": 9817 }, { "epoch": 0.84, "learning_rate": 6.74371114010498e-07, "loss": 0.8018, "step": 9818 }, { "epoch": 0.84, "learning_rate": 6.736781845595514e-07, "loss": 0.8166, "step": 9819 }, { "epoch": 0.84, "learning_rate": 6.729855855746481e-07, "loss": 0.7272, "step": 9820 }, { "epoch": 0.84, "learning_rate": 6.722933171086965e-07, "loss": 0.7817, "step": 9821 }, { "epoch": 0.84, "learning_rate": 6.716013792145721e-07, "loss": 0.7609, "step": 9822 }, { "epoch": 0.84, "learning_rate": 6.709097719451318e-07, "loss": 0.7904, "step": 9823 }, { "epoch": 0.84, "learning_rate": 6.702184953532031e-07, "loss": 0.7653, "step": 9824 }, { "epoch": 0.84, "learning_rate": 6.69527549491587e-07, "loss": 0.7758, "step": 9825 }, { "epoch": 0.84, "learning_rate": 6.688369344130658e-07, "loss": 0.7988, "step": 9826 }, { "epoch": 0.84, "learning_rate": 6.681466501703871e-07, "loss": 0.7703, "step": 9827 }, { "epoch": 0.84, "learning_rate": 6.674566968162815e-07, "loss": 0.8281, "step": 9828 }, { "epoch": 0.84, "learning_rate": 6.667670744034498e-07, "loss": 0.7747, "step": 9829 }, { "epoch": 0.84, "learning_rate": 6.660777829845688e-07, "loss": 0.8024, "step": 9830 }, { "epoch": 0.84, "learning_rate": 6.653888226122912e-07, "loss": 0.8139, "step": 9831 }, { "epoch": 0.84, "learning_rate": 6.647001933392422e-07, "loss": 0.8174, "step": 9832 }, { "epoch": 0.84, "learning_rate": 6.64011895218023e-07, "loss": 0.8217, "step": 9833 }, { "epoch": 0.84, "learning_rate": 6.633239283012089e-07, "loss": 0.4904, "step": 9834 }, { "epoch": 0.84, "learning_rate": 6.626362926413499e-07, "loss": 0.7565, "step": 9835 }, { "epoch": 0.84, "learning_rate": 6.619489882909724e-07, "loss": 0.763, "step": 9836 }, { "epoch": 0.84, "learning_rate": 6.612620153025756e-07, "loss": 0.7995, "step": 9837 }, { "epoch": 0.84, "learning_rate": 6.605753737286325e-07, "loss": 0.8179, "step": 9838 }, { "epoch": 0.84, "learning_rate": 6.59889063621596e-07, "loss": 0.7978, "step": 9839 }, { "epoch": 0.84, "learning_rate": 6.592030850338849e-07, "loss": 0.8126, "step": 9840 }, { "epoch": 0.84, "learning_rate": 6.585174380179011e-07, "loss": 0.7586, "step": 9841 }, { "epoch": 0.84, "learning_rate": 6.578321226260165e-07, "loss": 0.8084, "step": 9842 }, { "epoch": 0.84, "learning_rate": 6.571471389105782e-07, "loss": 0.8518, "step": 9843 }, { "epoch": 0.84, "learning_rate": 6.564624869239111e-07, "loss": 0.7649, "step": 9844 }, { "epoch": 0.84, "learning_rate": 6.557781667183105e-07, "loss": 0.7913, "step": 9845 }, { "epoch": 0.84, "learning_rate": 6.550941783460485e-07, "loss": 0.7332, "step": 9846 }, { "epoch": 0.84, "learning_rate": 6.544105218593715e-07, "loss": 0.807, "step": 9847 }, { "epoch": 0.84, "learning_rate": 6.537271973105003e-07, "loss": 0.7917, "step": 9848 }, { "epoch": 0.84, "learning_rate": 6.530442047516311e-07, "loss": 0.7548, "step": 9849 }, { "epoch": 0.84, "learning_rate": 6.523615442349351e-07, "loss": 0.7993, "step": 9850 }, { "epoch": 0.84, "learning_rate": 6.516792158125545e-07, "loss": 0.811, "step": 9851 }, { "epoch": 0.84, "learning_rate": 6.509972195366127e-07, "loss": 0.7991, "step": 9852 }, { "epoch": 0.84, "learning_rate": 6.503155554592017e-07, "loss": 0.7787, "step": 9853 }, { "epoch": 0.84, "learning_rate": 6.49634223632391e-07, "loss": 0.843, "step": 9854 }, { "epoch": 0.84, "learning_rate": 6.489532241082224e-07, "loss": 0.8274, "step": 9855 }, { "epoch": 0.84, "learning_rate": 6.482725569387171e-07, "loss": 0.8093, "step": 9856 }, { "epoch": 0.84, "learning_rate": 6.475922221758668e-07, "loss": 0.8334, "step": 9857 }, { "epoch": 0.84, "learning_rate": 6.469122198716366e-07, "loss": 0.7752, "step": 9858 }, { "epoch": 0.84, "learning_rate": 6.462325500779732e-07, "loss": 0.7516, "step": 9859 }, { "epoch": 0.84, "learning_rate": 6.455532128467873e-07, "loss": 0.7565, "step": 9860 }, { "epoch": 0.84, "learning_rate": 6.448742082299747e-07, "loss": 0.5207, "step": 9861 }, { "epoch": 0.84, "learning_rate": 6.441955362793994e-07, "loss": 0.7652, "step": 9862 }, { "epoch": 0.84, "learning_rate": 6.435171970469006e-07, "loss": 0.8067, "step": 9863 }, { "epoch": 0.84, "learning_rate": 6.428391905842957e-07, "loss": 0.7588, "step": 9864 }, { "epoch": 0.84, "learning_rate": 6.421615169433726e-07, "loss": 0.8265, "step": 9865 }, { "epoch": 0.84, "learning_rate": 6.414841761758961e-07, "loss": 0.8032, "step": 9866 }, { "epoch": 0.84, "learning_rate": 6.408071683336048e-07, "loss": 0.8027, "step": 9867 }, { "epoch": 0.84, "learning_rate": 6.4013049346821e-07, "loss": 0.7957, "step": 9868 }, { "epoch": 0.84, "learning_rate": 6.394541516314029e-07, "loss": 0.8228, "step": 9869 }, { "epoch": 0.84, "learning_rate": 6.387781428748435e-07, "loss": 0.8591, "step": 9870 }, { "epoch": 0.84, "learning_rate": 6.38102467250169e-07, "loss": 0.7969, "step": 9871 }, { "epoch": 0.84, "learning_rate": 6.374271248089919e-07, "loss": 0.5372, "step": 9872 }, { "epoch": 0.84, "learning_rate": 6.367521156028978e-07, "loss": 0.8076, "step": 9873 }, { "epoch": 0.84, "learning_rate": 6.360774396834468e-07, "loss": 0.7946, "step": 9874 }, { "epoch": 0.84, "learning_rate": 6.354030971021746e-07, "loss": 0.7738, "step": 9875 }, { "epoch": 0.84, "learning_rate": 6.347290879105894e-07, "loss": 0.8226, "step": 9876 }, { "epoch": 0.84, "learning_rate": 6.340554121601771e-07, "loss": 0.7938, "step": 9877 }, { "epoch": 0.84, "learning_rate": 6.333820699023963e-07, "loss": 0.854, "step": 9878 }, { "epoch": 0.84, "learning_rate": 6.327090611886799e-07, "loss": 0.7857, "step": 9879 }, { "epoch": 0.84, "learning_rate": 6.320363860704348e-07, "loss": 0.84, "step": 9880 }, { "epoch": 0.84, "learning_rate": 6.313640445990437e-07, "loss": 0.779, "step": 9881 }, { "epoch": 0.84, "learning_rate": 6.306920368258645e-07, "loss": 0.8116, "step": 9882 }, { "epoch": 0.84, "learning_rate": 6.300203628022272e-07, "loss": 0.7641, "step": 9883 }, { "epoch": 0.84, "learning_rate": 6.29349022579438e-07, "loss": 0.8173, "step": 9884 }, { "epoch": 0.84, "learning_rate": 6.286780162087774e-07, "loss": 0.5445, "step": 9885 }, { "epoch": 0.84, "learning_rate": 6.280073437415007e-07, "loss": 0.7819, "step": 9886 }, { "epoch": 0.84, "learning_rate": 6.273370052288368e-07, "loss": 0.7647, "step": 9887 }, { "epoch": 0.84, "learning_rate": 6.266670007219894e-07, "loss": 0.768, "step": 9888 }, { "epoch": 0.84, "learning_rate": 6.25997330272135e-07, "loss": 0.8044, "step": 9889 }, { "epoch": 0.84, "learning_rate": 6.253279939304297e-07, "loss": 0.8185, "step": 9890 }, { "epoch": 0.84, "learning_rate": 6.246589917479978e-07, "loss": 0.8314, "step": 9891 }, { "epoch": 0.84, "learning_rate": 6.239903237759448e-07, "loss": 0.8071, "step": 9892 }, { "epoch": 0.84, "learning_rate": 6.233219900653437e-07, "loss": 0.7797, "step": 9893 }, { "epoch": 0.84, "learning_rate": 6.226539906672446e-07, "loss": 0.8181, "step": 9894 }, { "epoch": 0.84, "learning_rate": 6.219863256326747e-07, "loss": 0.5188, "step": 9895 }, { "epoch": 0.84, "learning_rate": 6.21318995012632e-07, "loss": 0.7654, "step": 9896 }, { "epoch": 0.84, "learning_rate": 6.206519988580927e-07, "loss": 0.7712, "step": 9897 }, { "epoch": 0.84, "learning_rate": 6.199853372200037e-07, "loss": 0.7796, "step": 9898 }, { "epoch": 0.84, "learning_rate": 6.193190101492886e-07, "loss": 0.7926, "step": 9899 }, { "epoch": 0.84, "learning_rate": 6.186530176968442e-07, "loss": 0.7926, "step": 9900 }, { "epoch": 0.84, "learning_rate": 6.179873599135416e-07, "loss": 0.7858, "step": 9901 }, { "epoch": 0.84, "learning_rate": 6.173220368502292e-07, "loss": 0.8008, "step": 9902 }, { "epoch": 0.85, "learning_rate": 6.166570485577261e-07, "loss": 0.8133, "step": 9903 }, { "epoch": 0.85, "learning_rate": 6.159923950868269e-07, "loss": 0.7921, "step": 9904 }, { "epoch": 0.85, "learning_rate": 6.153280764883035e-07, "loss": 0.7987, "step": 9905 }, { "epoch": 0.85, "learning_rate": 6.146640928128977e-07, "loss": 0.7966, "step": 9906 }, { "epoch": 0.85, "learning_rate": 6.140004441113296e-07, "loss": 0.7849, "step": 9907 }, { "epoch": 0.85, "learning_rate": 6.1333713043429e-07, "loss": 0.8105, "step": 9908 }, { "epoch": 0.85, "learning_rate": 6.12674151832447e-07, "loss": 0.7409, "step": 9909 }, { "epoch": 0.85, "learning_rate": 6.120115083564432e-07, "loss": 0.7682, "step": 9910 }, { "epoch": 0.85, "learning_rate": 6.113492000568932e-07, "loss": 0.8003, "step": 9911 }, { "epoch": 0.85, "learning_rate": 6.106872269843888e-07, "loss": 0.7102, "step": 9912 }, { "epoch": 0.85, "learning_rate": 6.100255891894935e-07, "loss": 0.8044, "step": 9913 }, { "epoch": 0.85, "learning_rate": 6.093642867227462e-07, "loss": 0.7662, "step": 9914 }, { "epoch": 0.85, "learning_rate": 6.087033196346621e-07, "loss": 0.7878, "step": 9915 }, { "epoch": 0.85, "learning_rate": 6.080426879757284e-07, "loss": 0.5407, "step": 9916 }, { "epoch": 0.85, "learning_rate": 6.073823917964061e-07, "loss": 0.7654, "step": 9917 }, { "epoch": 0.85, "learning_rate": 6.067224311471348e-07, "loss": 0.7714, "step": 9918 }, { "epoch": 0.85, "learning_rate": 6.060628060783235e-07, "loss": 0.7828, "step": 9919 }, { "epoch": 0.85, "learning_rate": 6.054035166403582e-07, "loss": 0.822, "step": 9920 }, { "epoch": 0.85, "learning_rate": 6.047445628835985e-07, "loss": 0.7743, "step": 9921 }, { "epoch": 0.85, "learning_rate": 6.04085944858378e-07, "loss": 0.8196, "step": 9922 }, { "epoch": 0.85, "learning_rate": 6.034276626150071e-07, "loss": 0.7257, "step": 9923 }, { "epoch": 0.85, "learning_rate": 6.02769716203766e-07, "loss": 0.7456, "step": 9924 }, { "epoch": 0.85, "learning_rate": 6.021121056749152e-07, "loss": 0.7597, "step": 9925 }, { "epoch": 0.85, "learning_rate": 6.014548310786849e-07, "loss": 0.7561, "step": 9926 }, { "epoch": 0.85, "learning_rate": 6.007978924652791e-07, "loss": 0.8176, "step": 9927 }, { "epoch": 0.85, "learning_rate": 6.001412898848802e-07, "loss": 0.7956, "step": 9928 }, { "epoch": 0.85, "learning_rate": 5.994850233876415e-07, "loss": 0.7917, "step": 9929 }, { "epoch": 0.85, "learning_rate": 5.988290930236934e-07, "loss": 0.7519, "step": 9930 }, { "epoch": 0.85, "learning_rate": 5.981734988431387e-07, "loss": 0.7604, "step": 9931 }, { "epoch": 0.85, "learning_rate": 5.975182408960544e-07, "loss": 0.8057, "step": 9932 }, { "epoch": 0.85, "learning_rate": 5.968633192324924e-07, "loss": 0.5478, "step": 9933 }, { "epoch": 0.85, "learning_rate": 5.962087339024781e-07, "loss": 0.818, "step": 9934 }, { "epoch": 0.85, "learning_rate": 5.955544849560141e-07, "loss": 0.8219, "step": 9935 }, { "epoch": 0.85, "learning_rate": 5.94900572443074e-07, "loss": 0.7263, "step": 9936 }, { "epoch": 0.85, "learning_rate": 5.942469964136055e-07, "loss": 0.7613, "step": 9937 }, { "epoch": 0.85, "learning_rate": 5.935937569175348e-07, "loss": 0.7918, "step": 9938 }, { "epoch": 0.85, "learning_rate": 5.929408540047582e-07, "loss": 0.7572, "step": 9939 }, { "epoch": 0.85, "learning_rate": 5.922882877251479e-07, "loss": 0.7979, "step": 9940 }, { "epoch": 0.85, "learning_rate": 5.916360581285497e-07, "loss": 0.8088, "step": 9941 }, { "epoch": 0.85, "learning_rate": 5.909841652647835e-07, "loss": 0.7755, "step": 9942 }, { "epoch": 0.85, "learning_rate": 5.903326091836459e-07, "loss": 0.8037, "step": 9943 }, { "epoch": 0.85, "learning_rate": 5.896813899349052e-07, "loss": 0.856, "step": 9944 }, { "epoch": 0.85, "learning_rate": 5.89030507568305e-07, "loss": 0.8069, "step": 9945 }, { "epoch": 0.85, "learning_rate": 5.883799621335623e-07, "loss": 0.7601, "step": 9946 }, { "epoch": 0.85, "learning_rate": 5.877297536803683e-07, "loss": 0.7713, "step": 9947 }, { "epoch": 0.85, "learning_rate": 5.870798822583918e-07, "loss": 0.7938, "step": 9948 }, { "epoch": 0.85, "learning_rate": 5.864303479172711e-07, "loss": 0.7929, "step": 9949 }, { "epoch": 0.85, "learning_rate": 5.857811507066208e-07, "loss": 0.7899, "step": 9950 }, { "epoch": 0.85, "learning_rate": 5.851322906760315e-07, "loss": 0.7645, "step": 9951 }, { "epoch": 0.85, "learning_rate": 5.844837678750654e-07, "loss": 0.7573, "step": 9952 }, { "epoch": 0.85, "learning_rate": 5.838355823532599e-07, "loss": 0.8036, "step": 9953 }, { "epoch": 0.85, "learning_rate": 5.831877341601261e-07, "loss": 0.7606, "step": 9954 }, { "epoch": 0.85, "learning_rate": 5.825402233451493e-07, "loss": 0.791, "step": 9955 }, { "epoch": 0.85, "learning_rate": 5.818930499577924e-07, "loss": 0.7723, "step": 9956 }, { "epoch": 0.85, "learning_rate": 5.812462140474868e-07, "loss": 0.769, "step": 9957 }, { "epoch": 0.85, "learning_rate": 5.805997156636428e-07, "loss": 0.8118, "step": 9958 }, { "epoch": 0.85, "learning_rate": 5.79953554855644e-07, "loss": 0.8015, "step": 9959 }, { "epoch": 0.85, "learning_rate": 5.793077316728435e-07, "loss": 0.8068, "step": 9960 }, { "epoch": 0.85, "learning_rate": 5.786622461645763e-07, "loss": 0.7932, "step": 9961 }, { "epoch": 0.85, "learning_rate": 5.780170983801458e-07, "loss": 0.7485, "step": 9962 }, { "epoch": 0.85, "learning_rate": 5.773722883688332e-07, "loss": 0.7953, "step": 9963 }, { "epoch": 0.85, "learning_rate": 5.767278161798912e-07, "loss": 0.7535, "step": 9964 }, { "epoch": 0.85, "learning_rate": 5.760836818625481e-07, "loss": 0.7905, "step": 9965 }, { "epoch": 0.85, "learning_rate": 5.754398854660065e-07, "loss": 0.7367, "step": 9966 }, { "epoch": 0.85, "learning_rate": 5.747964270394408e-07, "loss": 0.7842, "step": 9967 }, { "epoch": 0.85, "learning_rate": 5.741533066320038e-07, "loss": 0.7966, "step": 9968 }, { "epoch": 0.85, "learning_rate": 5.735105242928202e-07, "loss": 0.7813, "step": 9969 }, { "epoch": 0.85, "learning_rate": 5.728680800709869e-07, "loss": 0.7938, "step": 9970 }, { "epoch": 0.85, "learning_rate": 5.722259740155795e-07, "loss": 0.4455, "step": 9971 }, { "epoch": 0.85, "learning_rate": 5.715842061756438e-07, "loss": 0.7899, "step": 9972 }, { "epoch": 0.85, "learning_rate": 5.709427766002013e-07, "loss": 0.7771, "step": 9973 }, { "epoch": 0.85, "learning_rate": 5.703016853382481e-07, "loss": 0.8069, "step": 9974 }, { "epoch": 0.85, "learning_rate": 5.696609324387531e-07, "loss": 0.7693, "step": 9975 }, { "epoch": 0.85, "learning_rate": 5.690205179506614e-07, "loss": 0.8112, "step": 9976 }, { "epoch": 0.85, "learning_rate": 5.6838044192289e-07, "loss": 0.8092, "step": 9977 }, { "epoch": 0.85, "learning_rate": 5.677407044043309e-07, "loss": 0.7786, "step": 9978 }, { "epoch": 0.85, "learning_rate": 5.671013054438534e-07, "loss": 0.7494, "step": 9979 }, { "epoch": 0.85, "learning_rate": 5.664622450902934e-07, "loss": 0.5416, "step": 9980 }, { "epoch": 0.85, "learning_rate": 5.658235233924686e-07, "loss": 0.809, "step": 9981 }, { "epoch": 0.85, "learning_rate": 5.65185140399167e-07, "loss": 0.5295, "step": 9982 }, { "epoch": 0.85, "learning_rate": 5.645470961591504e-07, "loss": 0.7731, "step": 9983 }, { "epoch": 0.85, "learning_rate": 5.639093907211579e-07, "loss": 0.8209, "step": 9984 }, { "epoch": 0.85, "learning_rate": 5.632720241338996e-07, "loss": 0.7904, "step": 9985 }, { "epoch": 0.85, "learning_rate": 5.626349964460604e-07, "loss": 0.7605, "step": 9986 }, { "epoch": 0.85, "learning_rate": 5.619983077062996e-07, "loss": 0.811, "step": 9987 }, { "epoch": 0.85, "learning_rate": 5.613619579632506e-07, "loss": 0.7558, "step": 9988 }, { "epoch": 0.85, "learning_rate": 5.607259472655225e-07, "loss": 0.8239, "step": 9989 }, { "epoch": 0.85, "learning_rate": 5.60090275661695e-07, "loss": 0.7781, "step": 9990 }, { "epoch": 0.85, "learning_rate": 5.594549432003244e-07, "loss": 0.8223, "step": 9991 }, { "epoch": 0.85, "learning_rate": 5.588199499299423e-07, "loss": 0.8208, "step": 9992 }, { "epoch": 0.85, "learning_rate": 5.581852958990497e-07, "loss": 0.7636, "step": 9993 }, { "epoch": 0.85, "learning_rate": 5.575509811561269e-07, "loss": 0.8005, "step": 9994 }, { "epoch": 0.85, "learning_rate": 5.569170057496254e-07, "loss": 0.7402, "step": 9995 }, { "epoch": 0.85, "learning_rate": 5.562833697279702e-07, "loss": 0.7776, "step": 9996 }, { "epoch": 0.85, "learning_rate": 5.556500731395631e-07, "loss": 0.7958, "step": 9997 }, { "epoch": 0.85, "learning_rate": 5.550171160327778e-07, "loss": 0.8059, "step": 9998 }, { "epoch": 0.85, "learning_rate": 5.543844984559643e-07, "loss": 0.7832, "step": 9999 }, { "epoch": 0.85, "learning_rate": 5.537522204574425e-07, "loss": 0.8025, "step": 10000 }, { "epoch": 0.85, "learning_rate": 5.531202820855102e-07, "loss": 0.7979, "step": 10001 }, { "epoch": 0.85, "learning_rate": 5.524886833884385e-07, "loss": 0.8166, "step": 10002 }, { "epoch": 0.85, "learning_rate": 5.518574244144709e-07, "loss": 0.7727, "step": 10003 }, { "epoch": 0.85, "learning_rate": 5.512265052118271e-07, "loss": 0.7801, "step": 10004 }, { "epoch": 0.85, "learning_rate": 5.505959258286997e-07, "loss": 0.4752, "step": 10005 }, { "epoch": 0.85, "learning_rate": 5.499656863132552e-07, "loss": 0.8119, "step": 10006 }, { "epoch": 0.85, "learning_rate": 5.493357867136351e-07, "loss": 0.78, "step": 10007 }, { "epoch": 0.85, "learning_rate": 5.48706227077952e-07, "loss": 0.748, "step": 10008 }, { "epoch": 0.85, "learning_rate": 5.480770074542979e-07, "loss": 0.8267, "step": 10009 }, { "epoch": 0.85, "learning_rate": 5.47448127890734e-07, "loss": 0.8011, "step": 10010 }, { "epoch": 0.85, "learning_rate": 5.468195884352972e-07, "loss": 0.7435, "step": 10011 }, { "epoch": 0.85, "learning_rate": 5.461913891360004e-07, "loss": 0.7914, "step": 10012 }, { "epoch": 0.85, "learning_rate": 5.455635300408252e-07, "loss": 0.8163, "step": 10013 }, { "epoch": 0.85, "learning_rate": 5.449360111977336e-07, "loss": 0.8112, "step": 10014 }, { "epoch": 0.85, "learning_rate": 5.443088326546575e-07, "loss": 0.7638, "step": 10015 }, { "epoch": 0.85, "learning_rate": 5.436819944595029e-07, "loss": 0.7919, "step": 10016 }, { "epoch": 0.85, "learning_rate": 5.430554966601526e-07, "loss": 0.8037, "step": 10017 }, { "epoch": 0.85, "learning_rate": 5.42429339304461e-07, "loss": 0.8311, "step": 10018 }, { "epoch": 0.85, "learning_rate": 5.418035224402574e-07, "loss": 0.8368, "step": 10019 }, { "epoch": 0.86, "learning_rate": 5.411780461153437e-07, "loss": 0.7921, "step": 10020 }, { "epoch": 0.86, "learning_rate": 5.405529103774976e-07, "loss": 0.7756, "step": 10021 }, { "epoch": 0.86, "learning_rate": 5.399281152744707e-07, "loss": 0.7528, "step": 10022 }, { "epoch": 0.86, "learning_rate": 5.39303660853987e-07, "loss": 0.8414, "step": 10023 }, { "epoch": 0.86, "learning_rate": 5.386795471637452e-07, "loss": 0.7487, "step": 10024 }, { "epoch": 0.86, "learning_rate": 5.380557742514208e-07, "loss": 0.8297, "step": 10025 }, { "epoch": 0.86, "learning_rate": 5.374323421646566e-07, "loss": 0.5249, "step": 10026 }, { "epoch": 0.86, "learning_rate": 5.36809250951077e-07, "loss": 0.7713, "step": 10027 }, { "epoch": 0.86, "learning_rate": 5.36186500658275e-07, "loss": 0.7861, "step": 10028 }, { "epoch": 0.86, "learning_rate": 5.355640913338189e-07, "loss": 0.7859, "step": 10029 }, { "epoch": 0.86, "learning_rate": 5.349420230252534e-07, "loss": 0.7863, "step": 10030 }, { "epoch": 0.86, "learning_rate": 5.34320295780093e-07, "loss": 0.7806, "step": 10031 }, { "epoch": 0.86, "learning_rate": 5.336989096458317e-07, "loss": 0.7781, "step": 10032 }, { "epoch": 0.86, "learning_rate": 5.330778646699303e-07, "loss": 0.7794, "step": 10033 }, { "epoch": 0.86, "learning_rate": 5.32457160899828e-07, "loss": 0.7468, "step": 10034 }, { "epoch": 0.86, "learning_rate": 5.318367983829393e-07, "loss": 0.7494, "step": 10035 }, { "epoch": 0.86, "learning_rate": 5.312167771666482e-07, "loss": 0.7725, "step": 10036 }, { "epoch": 0.86, "learning_rate": 5.305970972983171e-07, "loss": 0.7804, "step": 10037 }, { "epoch": 0.86, "learning_rate": 5.299777588252791e-07, "loss": 0.7914, "step": 10038 }, { "epoch": 0.86, "learning_rate": 5.293587617948431e-07, "loss": 0.7915, "step": 10039 }, { "epoch": 0.86, "learning_rate": 5.287401062542902e-07, "loss": 0.7609, "step": 10040 }, { "epoch": 0.86, "learning_rate": 5.281217922508757e-07, "loss": 0.8084, "step": 10041 }, { "epoch": 0.86, "learning_rate": 5.275038198318322e-07, "loss": 0.7985, "step": 10042 }, { "epoch": 0.86, "learning_rate": 5.268861890443616e-07, "loss": 0.8098, "step": 10043 }, { "epoch": 0.86, "learning_rate": 5.262688999356414e-07, "loss": 0.7498, "step": 10044 }, { "epoch": 0.86, "learning_rate": 5.256519525528254e-07, "loss": 0.8139, "step": 10045 }, { "epoch": 0.86, "learning_rate": 5.250353469430353e-07, "loss": 0.838, "step": 10046 }, { "epoch": 0.86, "learning_rate": 5.24419083153374e-07, "loss": 0.8052, "step": 10047 }, { "epoch": 0.86, "learning_rate": 5.238031612309136e-07, "loss": 0.4805, "step": 10048 }, { "epoch": 0.86, "learning_rate": 5.231875812227005e-07, "loss": 0.7934, "step": 10049 }, { "epoch": 0.86, "learning_rate": 5.225723431757574e-07, "loss": 0.7886, "step": 10050 }, { "epoch": 0.86, "learning_rate": 5.21957447137078e-07, "loss": 0.7886, "step": 10051 }, { "epoch": 0.86, "learning_rate": 5.213428931536313e-07, "loss": 0.7859, "step": 10052 }, { "epoch": 0.86, "learning_rate": 5.207286812723605e-07, "loss": 0.7785, "step": 10053 }, { "epoch": 0.86, "learning_rate": 5.20114811540181e-07, "loss": 0.7959, "step": 10054 }, { "epoch": 0.86, "learning_rate": 5.195012840039854e-07, "loss": 0.8377, "step": 10055 }, { "epoch": 0.86, "learning_rate": 5.188880987106359e-07, "loss": 0.7868, "step": 10056 }, { "epoch": 0.86, "learning_rate": 5.182752557069709e-07, "loss": 0.7647, "step": 10057 }, { "epoch": 0.86, "learning_rate": 5.17662755039805e-07, "loss": 0.7667, "step": 10058 }, { "epoch": 0.86, "learning_rate": 5.170505967559198e-07, "loss": 0.7989, "step": 10059 }, { "epoch": 0.86, "learning_rate": 5.164387809020782e-07, "loss": 0.7722, "step": 10060 }, { "epoch": 0.86, "learning_rate": 5.158273075250131e-07, "loss": 0.7832, "step": 10061 }, { "epoch": 0.86, "learning_rate": 5.152161766714298e-07, "loss": 0.7422, "step": 10062 }, { "epoch": 0.86, "learning_rate": 5.146053883880131e-07, "loss": 0.7912, "step": 10063 }, { "epoch": 0.86, "learning_rate": 5.139949427214147e-07, "loss": 0.8067, "step": 10064 }, { "epoch": 0.86, "learning_rate": 5.133848397182667e-07, "loss": 0.7764, "step": 10065 }, { "epoch": 0.86, "learning_rate": 5.127750794251696e-07, "loss": 0.7839, "step": 10066 }, { "epoch": 0.86, "learning_rate": 5.121656618886994e-07, "loss": 0.7566, "step": 10067 }, { "epoch": 0.86, "learning_rate": 5.115565871554085e-07, "loss": 0.8081, "step": 10068 }, { "epoch": 0.86, "learning_rate": 5.109478552718189e-07, "loss": 0.7688, "step": 10069 }, { "epoch": 0.86, "learning_rate": 5.103394662844314e-07, "loss": 0.7853, "step": 10070 }, { "epoch": 0.86, "learning_rate": 5.097314202397164e-07, "loss": 0.8013, "step": 10071 }, { "epoch": 0.86, "learning_rate": 5.091237171841173e-07, "loss": 0.8083, "step": 10072 }, { "epoch": 0.86, "learning_rate": 5.085163571640566e-07, "loss": 0.8362, "step": 10073 }, { "epoch": 0.86, "learning_rate": 5.07909340225925e-07, "loss": 0.8164, "step": 10074 }, { "epoch": 0.86, "learning_rate": 5.073026664160924e-07, "loss": 0.5272, "step": 10075 }, { "epoch": 0.86, "learning_rate": 5.066963357808974e-07, "loss": 0.7687, "step": 10076 }, { "epoch": 0.86, "learning_rate": 5.06090348366654e-07, "loss": 0.7879, "step": 10077 }, { "epoch": 0.86, "learning_rate": 5.054847042196542e-07, "loss": 0.7972, "step": 10078 }, { "epoch": 0.86, "learning_rate": 5.048794033861554e-07, "loss": 0.8218, "step": 10079 }, { "epoch": 0.86, "learning_rate": 5.042744459123966e-07, "loss": 0.785, "step": 10080 }, { "epoch": 0.86, "learning_rate": 5.036698318445865e-07, "loss": 0.7944, "step": 10081 }, { "epoch": 0.86, "learning_rate": 5.030655612289076e-07, "loss": 0.7746, "step": 10082 }, { "epoch": 0.86, "learning_rate": 5.024616341115191e-07, "loss": 0.8042, "step": 10083 }, { "epoch": 0.86, "learning_rate": 5.018580505385506e-07, "loss": 0.7561, "step": 10084 }, { "epoch": 0.86, "learning_rate": 5.012548105561077e-07, "loss": 0.8219, "step": 10085 }, { "epoch": 0.86, "learning_rate": 5.006519142102684e-07, "loss": 0.7547, "step": 10086 }, { "epoch": 0.86, "learning_rate": 5.000493615470836e-07, "loss": 0.7686, "step": 10087 }, { "epoch": 0.86, "learning_rate": 4.994471526125816e-07, "loss": 0.8178, "step": 10088 }, { "epoch": 0.86, "learning_rate": 4.988452874527611e-07, "loss": 0.5724, "step": 10089 }, { "epoch": 0.86, "learning_rate": 4.982437661135947e-07, "loss": 0.8023, "step": 10090 }, { "epoch": 0.86, "learning_rate": 4.976425886410324e-07, "loss": 0.7427, "step": 10091 }, { "epoch": 0.86, "learning_rate": 4.970417550809909e-07, "loss": 0.8129, "step": 10092 }, { "epoch": 0.86, "learning_rate": 4.964412654793688e-07, "loss": 0.8225, "step": 10093 }, { "epoch": 0.86, "learning_rate": 4.958411198820323e-07, "loss": 0.7826, "step": 10094 }, { "epoch": 0.86, "learning_rate": 4.952413183348231e-07, "loss": 0.7552, "step": 10095 }, { "epoch": 0.86, "learning_rate": 4.94641860883559e-07, "loss": 0.7795, "step": 10096 }, { "epoch": 0.86, "learning_rate": 4.940427475740289e-07, "loss": 0.8083, "step": 10097 }, { "epoch": 0.86, "learning_rate": 4.934439784519956e-07, "loss": 0.7862, "step": 10098 }, { "epoch": 0.86, "learning_rate": 4.92845553563196e-07, "loss": 0.7772, "step": 10099 }, { "epoch": 0.86, "learning_rate": 4.9224747295334e-07, "loss": 0.7444, "step": 10100 }, { "epoch": 0.86, "learning_rate": 4.916497366681138e-07, "loss": 0.8029, "step": 10101 }, { "epoch": 0.86, "learning_rate": 4.910523447531739e-07, "loss": 0.7668, "step": 10102 }, { "epoch": 0.86, "learning_rate": 4.904552972541543e-07, "loss": 0.8261, "step": 10103 }, { "epoch": 0.86, "learning_rate": 4.898585942166584e-07, "loss": 0.7244, "step": 10104 }, { "epoch": 0.86, "learning_rate": 4.892622356862658e-07, "loss": 0.7895, "step": 10105 }, { "epoch": 0.86, "learning_rate": 4.886662217085291e-07, "loss": 0.8266, "step": 10106 }, { "epoch": 0.86, "learning_rate": 4.880705523289753e-07, "loss": 0.7969, "step": 10107 }, { "epoch": 0.86, "learning_rate": 4.874752275931043e-07, "loss": 0.7262, "step": 10108 }, { "epoch": 0.86, "learning_rate": 4.868802475463913e-07, "loss": 0.7541, "step": 10109 }, { "epoch": 0.86, "learning_rate": 4.862856122342807e-07, "loss": 0.8195, "step": 10110 }, { "epoch": 0.86, "learning_rate": 4.856913217021986e-07, "loss": 0.8071, "step": 10111 }, { "epoch": 0.86, "learning_rate": 4.85097375995534e-07, "loss": 0.7804, "step": 10112 }, { "epoch": 0.86, "learning_rate": 4.845037751596604e-07, "loss": 0.7354, "step": 10113 }, { "epoch": 0.86, "learning_rate": 4.839105192399174e-07, "loss": 0.7781, "step": 10114 }, { "epoch": 0.86, "learning_rate": 4.833176082816205e-07, "loss": 0.7444, "step": 10115 }, { "epoch": 0.86, "learning_rate": 4.827250423300617e-07, "loss": 0.8076, "step": 10116 }, { "epoch": 0.86, "learning_rate": 4.821328214305021e-07, "loss": 0.7783, "step": 10117 }, { "epoch": 0.86, "learning_rate": 4.815409456281788e-07, "loss": 0.7643, "step": 10118 }, { "epoch": 0.86, "learning_rate": 4.809494149683031e-07, "loss": 0.7597, "step": 10119 }, { "epoch": 0.86, "learning_rate": 4.803582294960568e-07, "loss": 0.7705, "step": 10120 }, { "epoch": 0.86, "learning_rate": 4.797673892566007e-07, "loss": 0.7703, "step": 10121 }, { "epoch": 0.86, "learning_rate": 4.791768942950642e-07, "loss": 0.4653, "step": 10122 }, { "epoch": 0.86, "learning_rate": 4.785867446565518e-07, "loss": 0.7795, "step": 10123 }, { "epoch": 0.86, "learning_rate": 4.779969403861439e-07, "loss": 0.7821, "step": 10124 }, { "epoch": 0.86, "learning_rate": 4.77407481528892e-07, "loss": 0.7735, "step": 10125 }, { "epoch": 0.86, "learning_rate": 4.768183681298211e-07, "loss": 0.7935, "step": 10126 }, { "epoch": 0.86, "learning_rate": 4.762296002339317e-07, "loss": 0.8032, "step": 10127 }, { "epoch": 0.86, "learning_rate": 4.756411778861947e-07, "loss": 0.7818, "step": 10128 }, { "epoch": 0.86, "learning_rate": 4.7505310113156013e-07, "loss": 0.7469, "step": 10129 }, { "epoch": 0.86, "learning_rate": 4.744653700149454e-07, "loss": 0.8051, "step": 10130 }, { "epoch": 0.86, "learning_rate": 4.7387798458124634e-07, "loss": 0.7929, "step": 10131 }, { "epoch": 0.86, "learning_rate": 4.7329094487532867e-07, "loss": 0.7516, "step": 10132 }, { "epoch": 0.86, "learning_rate": 4.727042509420338e-07, "loss": 0.8051, "step": 10133 }, { "epoch": 0.86, "learning_rate": 4.7211790282617686e-07, "loss": 0.812, "step": 10134 }, { "epoch": 0.86, "learning_rate": 4.715319005725466e-07, "loss": 0.7774, "step": 10135 }, { "epoch": 0.86, "learning_rate": 4.7094624422590273e-07, "loss": 0.7942, "step": 10136 }, { "epoch": 0.87, "learning_rate": 4.703609338309828e-07, "loss": 0.8401, "step": 10137 }, { "epoch": 0.87, "learning_rate": 4.697759694324955e-07, "loss": 0.7905, "step": 10138 }, { "epoch": 0.87, "learning_rate": 4.6919135107512225e-07, "loss": 0.7718, "step": 10139 }, { "epoch": 0.87, "learning_rate": 4.6860707880351854e-07, "loss": 0.7654, "step": 10140 }, { "epoch": 0.87, "learning_rate": 4.680231526623158e-07, "loss": 0.785, "step": 10141 }, { "epoch": 0.87, "learning_rate": 4.674395726961167e-07, "loss": 0.7623, "step": 10142 }, { "epoch": 0.87, "learning_rate": 4.6685633894949735e-07, "loss": 0.8256, "step": 10143 }, { "epoch": 0.87, "learning_rate": 4.6627345146700974e-07, "loss": 0.4772, "step": 10144 }, { "epoch": 0.87, "learning_rate": 4.6569091029317446e-07, "loss": 0.7842, "step": 10145 }, { "epoch": 0.87, "learning_rate": 4.6510871547249204e-07, "loss": 0.7848, "step": 10146 }, { "epoch": 0.87, "learning_rate": 4.645268670494324e-07, "loss": 0.7941, "step": 10147 }, { "epoch": 0.87, "learning_rate": 4.639453650684389e-07, "loss": 0.8098, "step": 10148 }, { "epoch": 0.87, "learning_rate": 4.63364209573931e-07, "loss": 0.7879, "step": 10149 }, { "epoch": 0.87, "learning_rate": 4.627834006103005e-07, "loss": 0.812, "step": 10150 }, { "epoch": 0.87, "learning_rate": 4.622029382219112e-07, "loss": 0.8192, "step": 10151 }, { "epoch": 0.87, "learning_rate": 4.6162282245310275e-07, "loss": 0.7996, "step": 10152 }, { "epoch": 0.87, "learning_rate": 4.6104305334818577e-07, "loss": 0.8661, "step": 10153 }, { "epoch": 0.87, "learning_rate": 4.604636309514482e-07, "loss": 0.8422, "step": 10154 }, { "epoch": 0.87, "learning_rate": 4.5988455530714735e-07, "loss": 0.7799, "step": 10155 }, { "epoch": 0.87, "learning_rate": 4.593058264595163e-07, "loss": 0.815, "step": 10156 }, { "epoch": 0.87, "learning_rate": 4.5872744445276227e-07, "loss": 0.7567, "step": 10157 }, { "epoch": 0.87, "learning_rate": 4.58149409331064e-07, "loss": 0.8082, "step": 10158 }, { "epoch": 0.87, "learning_rate": 4.5757172113857494e-07, "loss": 0.75, "step": 10159 }, { "epoch": 0.87, "learning_rate": 4.569943799194221e-07, "loss": 0.7914, "step": 10160 }, { "epoch": 0.87, "learning_rate": 4.56417385717704e-07, "loss": 0.7829, "step": 10161 }, { "epoch": 0.87, "learning_rate": 4.558407385774971e-07, "loss": 0.8059, "step": 10162 }, { "epoch": 0.87, "learning_rate": 4.5526443854284676e-07, "loss": 0.4708, "step": 10163 }, { "epoch": 0.87, "learning_rate": 4.5468848565777377e-07, "loss": 0.7964, "step": 10164 }, { "epoch": 0.87, "learning_rate": 4.5411287996627305e-07, "loss": 0.7401, "step": 10165 }, { "epoch": 0.87, "learning_rate": 4.5353762151231107e-07, "loss": 0.8117, "step": 10166 }, { "epoch": 0.87, "learning_rate": 4.5296271033982987e-07, "loss": 0.7907, "step": 10167 }, { "epoch": 0.87, "learning_rate": 4.523881464927443e-07, "loss": 0.756, "step": 10168 }, { "epoch": 0.87, "learning_rate": 4.51813930014941e-07, "loss": 0.7888, "step": 10169 }, { "epoch": 0.87, "learning_rate": 4.5124006095028316e-07, "loss": 0.776, "step": 10170 }, { "epoch": 0.87, "learning_rate": 4.5066653934260465e-07, "loss": 0.7965, "step": 10171 }, { "epoch": 0.87, "learning_rate": 4.5009336523571477e-07, "loss": 0.7676, "step": 10172 }, { "epoch": 0.87, "learning_rate": 4.4952053867339475e-07, "loss": 0.8074, "step": 10173 }, { "epoch": 0.87, "learning_rate": 4.4894805969939945e-07, "loss": 0.8158, "step": 10174 }, { "epoch": 0.87, "learning_rate": 4.48375928357459e-07, "loss": 0.5226, "step": 10175 }, { "epoch": 0.87, "learning_rate": 4.47804144691274e-07, "loss": 0.8219, "step": 10176 }, { "epoch": 0.87, "learning_rate": 4.472327087445216e-07, "loss": 0.7748, "step": 10177 }, { "epoch": 0.87, "learning_rate": 4.4666162056085207e-07, "loss": 0.7671, "step": 10178 }, { "epoch": 0.87, "learning_rate": 4.4609088018388424e-07, "loss": 0.7598, "step": 10179 }, { "epoch": 0.87, "learning_rate": 4.455204876572172e-07, "loss": 0.8148, "step": 10180 }, { "epoch": 0.87, "learning_rate": 4.4495044302441836e-07, "loss": 0.8016, "step": 10181 }, { "epoch": 0.87, "learning_rate": 4.443807463290323e-07, "loss": 0.7968, "step": 10182 }, { "epoch": 0.87, "learning_rate": 4.438113976145747e-07, "loss": 0.7579, "step": 10183 }, { "epoch": 0.87, "learning_rate": 4.432423969245353e-07, "loss": 0.785, "step": 10184 }, { "epoch": 0.87, "learning_rate": 4.4267374430237706e-07, "loss": 0.8022, "step": 10185 }, { "epoch": 0.87, "learning_rate": 4.421054397915359e-07, "loss": 0.7589, "step": 10186 }, { "epoch": 0.87, "learning_rate": 4.4153748343542247e-07, "loss": 0.778, "step": 10187 }, { "epoch": 0.87, "learning_rate": 4.4096987527742053e-07, "loss": 0.7849, "step": 10188 }, { "epoch": 0.87, "learning_rate": 4.4040261536088533e-07, "loss": 0.8218, "step": 10189 }, { "epoch": 0.87, "learning_rate": 4.3983570372914897e-07, "loss": 0.798, "step": 10190 }, { "epoch": 0.87, "learning_rate": 4.3926914042551393e-07, "loss": 0.7493, "step": 10191 }, { "epoch": 0.87, "learning_rate": 4.387029254932573e-07, "loss": 0.4284, "step": 10192 }, { "epoch": 0.87, "learning_rate": 4.3813705897562896e-07, "loss": 0.7618, "step": 10193 }, { "epoch": 0.87, "learning_rate": 4.375715409158521e-07, "loss": 0.7552, "step": 10194 }, { "epoch": 0.87, "learning_rate": 4.3700637135712596e-07, "loss": 0.771, "step": 10195 }, { "epoch": 0.87, "learning_rate": 4.364415503426195e-07, "loss": 0.7495, "step": 10196 }, { "epoch": 0.87, "learning_rate": 4.3587707791547696e-07, "loss": 0.8192, "step": 10197 }, { "epoch": 0.87, "learning_rate": 4.35312954118815e-07, "loss": 0.7972, "step": 10198 }, { "epoch": 0.87, "learning_rate": 4.347491789957242e-07, "loss": 0.8012, "step": 10199 }, { "epoch": 0.87, "learning_rate": 4.3418575258926954e-07, "loss": 0.8405, "step": 10200 }, { "epoch": 0.87, "learning_rate": 4.336226749424877e-07, "loss": 0.742, "step": 10201 }, { "epoch": 0.87, "learning_rate": 4.3305994609838865e-07, "loss": 0.8177, "step": 10202 }, { "epoch": 0.87, "learning_rate": 4.324975660999575e-07, "loss": 0.8195, "step": 10203 }, { "epoch": 0.87, "learning_rate": 4.319355349901522e-07, "loss": 0.5329, "step": 10204 }, { "epoch": 0.87, "learning_rate": 4.313738528119021e-07, "loss": 0.8027, "step": 10205 }, { "epoch": 0.87, "learning_rate": 4.3081251960811143e-07, "loss": 0.7449, "step": 10206 }, { "epoch": 0.87, "learning_rate": 4.3025153542165744e-07, "loss": 0.7872, "step": 10207 }, { "epoch": 0.87, "learning_rate": 4.296909002953925e-07, "loss": 0.7635, "step": 10208 }, { "epoch": 0.87, "learning_rate": 4.2913061427213854e-07, "loss": 0.7867, "step": 10209 }, { "epoch": 0.87, "learning_rate": 4.2857067739469517e-07, "loss": 0.7934, "step": 10210 }, { "epoch": 0.87, "learning_rate": 4.280110897058326e-07, "loss": 0.7529, "step": 10211 }, { "epoch": 0.87, "learning_rate": 4.274518512482934e-07, "loss": 0.7775, "step": 10212 }, { "epoch": 0.87, "learning_rate": 4.2689296206479604e-07, "loss": 0.7556, "step": 10213 }, { "epoch": 0.87, "learning_rate": 4.263344221980309e-07, "loss": 0.7636, "step": 10214 }, { "epoch": 0.87, "learning_rate": 4.257762316906633e-07, "loss": 0.7752, "step": 10215 }, { "epoch": 0.87, "learning_rate": 4.252183905853291e-07, "loss": 0.7756, "step": 10216 }, { "epoch": 0.87, "learning_rate": 4.2466089892464035e-07, "loss": 0.8515, "step": 10217 }, { "epoch": 0.87, "learning_rate": 4.2410375675117964e-07, "loss": 0.8306, "step": 10218 }, { "epoch": 0.87, "learning_rate": 4.2354696410750463e-07, "loss": 0.7724, "step": 10219 }, { "epoch": 0.87, "learning_rate": 4.229905210361468e-07, "loss": 0.8094, "step": 10220 }, { "epoch": 0.87, "learning_rate": 4.224344275796094e-07, "loss": 0.768, "step": 10221 }, { "epoch": 0.87, "learning_rate": 4.21878683780369e-07, "loss": 0.7676, "step": 10222 }, { "epoch": 0.87, "learning_rate": 4.213232896808772e-07, "loss": 0.7731, "step": 10223 }, { "epoch": 0.87, "learning_rate": 4.2076824532355787e-07, "loss": 0.8063, "step": 10224 }, { "epoch": 0.87, "learning_rate": 4.202135507508076e-07, "loss": 0.8306, "step": 10225 }, { "epoch": 0.87, "learning_rate": 4.1965920600499645e-07, "loss": 0.7996, "step": 10226 }, { "epoch": 0.87, "learning_rate": 4.1910521112846715e-07, "loss": 0.7712, "step": 10227 }, { "epoch": 0.87, "learning_rate": 4.185515661635386e-07, "loss": 0.7555, "step": 10228 }, { "epoch": 0.87, "learning_rate": 4.179982711524999e-07, "loss": 0.8018, "step": 10229 }, { "epoch": 0.87, "learning_rate": 4.1744532613761433e-07, "loss": 0.7962, "step": 10230 }, { "epoch": 0.87, "learning_rate": 4.168927311611204e-07, "loss": 0.7384, "step": 10231 }, { "epoch": 0.87, "learning_rate": 4.163404862652248e-07, "loss": 0.8042, "step": 10232 }, { "epoch": 0.87, "learning_rate": 4.1578859149211335e-07, "loss": 0.7691, "step": 10233 }, { "epoch": 0.87, "learning_rate": 4.1523704688394176e-07, "loss": 0.7714, "step": 10234 }, { "epoch": 0.87, "learning_rate": 4.1468585248283843e-07, "loss": 0.4961, "step": 10235 }, { "epoch": 0.87, "learning_rate": 4.141350083309093e-07, "loss": 0.8004, "step": 10236 }, { "epoch": 0.87, "learning_rate": 4.1358451447022827e-07, "loss": 0.8299, "step": 10237 }, { "epoch": 0.87, "learning_rate": 4.1303437094284525e-07, "loss": 0.8287, "step": 10238 }, { "epoch": 0.87, "learning_rate": 4.1248457779078374e-07, "loss": 0.7657, "step": 10239 }, { "epoch": 0.87, "learning_rate": 4.11935135056038e-07, "loss": 0.7637, "step": 10240 }, { "epoch": 0.87, "learning_rate": 4.113860427805788e-07, "loss": 0.7765, "step": 10241 }, { "epoch": 0.87, "learning_rate": 4.1083730100634766e-07, "loss": 0.7608, "step": 10242 }, { "epoch": 0.87, "learning_rate": 4.1028890977526106e-07, "loss": 0.7726, "step": 10243 }, { "epoch": 0.87, "learning_rate": 4.097408691292087e-07, "loss": 0.7941, "step": 10244 }, { "epoch": 0.87, "learning_rate": 4.091931791100495e-07, "loss": 0.7656, "step": 10245 }, { "epoch": 0.87, "learning_rate": 4.0864583975962104e-07, "loss": 0.8113, "step": 10246 }, { "epoch": 0.87, "learning_rate": 4.0809885111973094e-07, "loss": 0.792, "step": 10247 }, { "epoch": 0.87, "learning_rate": 4.075522132321619e-07, "loss": 0.8629, "step": 10248 }, { "epoch": 0.87, "learning_rate": 4.0700592613866886e-07, "loss": 0.7552, "step": 10249 }, { "epoch": 0.87, "learning_rate": 4.064599898809779e-07, "loss": 0.7936, "step": 10250 }, { "epoch": 0.87, "learning_rate": 4.05914404500794e-07, "loss": 0.7923, "step": 10251 }, { "epoch": 0.87, "learning_rate": 4.053691700397877e-07, "loss": 0.8075, "step": 10252 }, { "epoch": 0.87, "learning_rate": 4.0482428653960957e-07, "loss": 0.7498, "step": 10253 }, { "epoch": 0.87, "learning_rate": 4.042797540418797e-07, "loss": 0.7581, "step": 10254 }, { "epoch": 0.88, "learning_rate": 4.0373557258819087e-07, "loss": 0.8054, "step": 10255 }, { "epoch": 0.88, "learning_rate": 4.031917422201126e-07, "loss": 0.7565, "step": 10256 }, { "epoch": 0.88, "learning_rate": 4.0264826297918445e-07, "loss": 0.8088, "step": 10257 }, { "epoch": 0.88, "learning_rate": 4.0210513490691995e-07, "loss": 0.7683, "step": 10258 }, { "epoch": 0.88, "learning_rate": 4.015623580448058e-07, "loss": 0.8164, "step": 10259 }, { "epoch": 0.88, "learning_rate": 4.0101993243430115e-07, "loss": 0.7964, "step": 10260 }, { "epoch": 0.88, "learning_rate": 4.0047785811684116e-07, "loss": 0.8154, "step": 10261 }, { "epoch": 0.88, "learning_rate": 3.99936135133831e-07, "loss": 0.7607, "step": 10262 }, { "epoch": 0.88, "learning_rate": 3.993947635266493e-07, "loss": 0.8359, "step": 10263 }, { "epoch": 0.88, "learning_rate": 3.9885374333665184e-07, "loss": 0.8211, "step": 10264 }, { "epoch": 0.88, "learning_rate": 3.983130746051606e-07, "loss": 0.7831, "step": 10265 }, { "epoch": 0.88, "learning_rate": 3.977727573734774e-07, "loss": 0.8423, "step": 10266 }, { "epoch": 0.88, "learning_rate": 3.9723279168287333e-07, "loss": 0.8285, "step": 10267 }, { "epoch": 0.88, "learning_rate": 3.9669317757459245e-07, "loss": 0.7569, "step": 10268 }, { "epoch": 0.88, "learning_rate": 3.9615391508985524e-07, "loss": 0.789, "step": 10269 }, { "epoch": 0.88, "learning_rate": 3.9561500426985255e-07, "loss": 0.8252, "step": 10270 }, { "epoch": 0.88, "learning_rate": 3.9507644515574875e-07, "loss": 0.7614, "step": 10271 }, { "epoch": 0.88, "learning_rate": 3.9453823778868195e-07, "loss": 0.8247, "step": 10272 }, { "epoch": 0.88, "learning_rate": 3.940003822097621e-07, "loss": 0.7689, "step": 10273 }, { "epoch": 0.88, "learning_rate": 3.9346287846007467e-07, "loss": 0.8155, "step": 10274 }, { "epoch": 0.88, "learning_rate": 3.9292572658067684e-07, "loss": 0.7901, "step": 10275 }, { "epoch": 0.88, "learning_rate": 3.923889266125974e-07, "loss": 0.7683, "step": 10276 }, { "epoch": 0.88, "learning_rate": 3.9185247859684307e-07, "loss": 0.786, "step": 10277 }, { "epoch": 0.88, "learning_rate": 3.91316382574386e-07, "loss": 0.7696, "step": 10278 }, { "epoch": 0.88, "learning_rate": 3.9078063858617956e-07, "loss": 0.7694, "step": 10279 }, { "epoch": 0.88, "learning_rate": 3.9024524667314434e-07, "loss": 0.7952, "step": 10280 }, { "epoch": 0.88, "learning_rate": 3.8971020687617656e-07, "loss": 0.7912, "step": 10281 }, { "epoch": 0.88, "learning_rate": 3.8917551923614685e-07, "loss": 0.7346, "step": 10282 }, { "epoch": 0.88, "learning_rate": 3.8864118379389526e-07, "loss": 0.4915, "step": 10283 }, { "epoch": 0.88, "learning_rate": 3.881072005902392e-07, "loss": 0.7835, "step": 10284 }, { "epoch": 0.88, "learning_rate": 3.8757356966596435e-07, "loss": 0.8482, "step": 10285 }, { "epoch": 0.88, "learning_rate": 3.8704029106183425e-07, "loss": 0.8139, "step": 10286 }, { "epoch": 0.88, "learning_rate": 3.8650736481858244e-07, "loss": 0.8179, "step": 10287 }, { "epoch": 0.88, "learning_rate": 3.8597479097691626e-07, "loss": 0.7736, "step": 10288 }, { "epoch": 0.88, "learning_rate": 3.854425695775171e-07, "loss": 0.7672, "step": 10289 }, { "epoch": 0.88, "learning_rate": 3.849107006610381e-07, "loss": 0.8036, "step": 10290 }, { "epoch": 0.88, "learning_rate": 3.843791842681066e-07, "loss": 0.7954, "step": 10291 }, { "epoch": 0.88, "learning_rate": 3.838480204393219e-07, "loss": 0.817, "step": 10292 }, { "epoch": 0.88, "learning_rate": 3.833172092152565e-07, "loss": 0.813, "step": 10293 }, { "epoch": 0.88, "learning_rate": 3.82786750636458e-07, "loss": 0.7896, "step": 10294 }, { "epoch": 0.88, "learning_rate": 3.822566447434445e-07, "loss": 0.7755, "step": 10295 }, { "epoch": 0.88, "learning_rate": 3.817268915767075e-07, "loss": 0.796, "step": 10296 }, { "epoch": 0.88, "learning_rate": 3.811974911767141e-07, "loss": 0.7776, "step": 10297 }, { "epoch": 0.88, "learning_rate": 3.806684435839003e-07, "loss": 0.8257, "step": 10298 }, { "epoch": 0.88, "learning_rate": 3.8013974883867877e-07, "loss": 0.8204, "step": 10299 }, { "epoch": 0.88, "learning_rate": 3.7961140698143385e-07, "loss": 0.8151, "step": 10300 }, { "epoch": 0.88, "learning_rate": 3.790834180525216e-07, "loss": 0.795, "step": 10301 }, { "epoch": 0.88, "learning_rate": 3.7855578209227415e-07, "loss": 0.7875, "step": 10302 }, { "epoch": 0.88, "learning_rate": 3.7802849914099383e-07, "loss": 0.7415, "step": 10303 }, { "epoch": 0.88, "learning_rate": 3.7750156923895833e-07, "loss": 0.777, "step": 10304 }, { "epoch": 0.88, "learning_rate": 3.76974992426416e-07, "loss": 0.7236, "step": 10305 }, { "epoch": 0.88, "learning_rate": 3.7644876874358914e-07, "loss": 0.7589, "step": 10306 }, { "epoch": 0.88, "learning_rate": 3.7592289823067505e-07, "loss": 0.8096, "step": 10307 }, { "epoch": 0.88, "learning_rate": 3.7539738092784107e-07, "loss": 0.7735, "step": 10308 }, { "epoch": 0.88, "learning_rate": 3.748722168752278e-07, "loss": 0.7291, "step": 10309 }, { "epoch": 0.88, "learning_rate": 3.743474061129537e-07, "loss": 0.7542, "step": 10310 }, { "epoch": 0.88, "learning_rate": 3.738229486811018e-07, "loss": 0.8297, "step": 10311 }, { "epoch": 0.88, "learning_rate": 3.7329884461973555e-07, "loss": 0.7963, "step": 10312 }, { "epoch": 0.88, "learning_rate": 3.7277509396888845e-07, "loss": 0.8375, "step": 10313 }, { "epoch": 0.88, "learning_rate": 3.722516967685652e-07, "loss": 0.8236, "step": 10314 }, { "epoch": 0.88, "learning_rate": 3.717286530587483e-07, "loss": 0.7416, "step": 10315 }, { "epoch": 0.88, "learning_rate": 3.7120596287938793e-07, "loss": 0.7236, "step": 10316 }, { "epoch": 0.88, "learning_rate": 3.706836262704133e-07, "loss": 0.7855, "step": 10317 }, { "epoch": 0.88, "learning_rate": 3.7016164327171923e-07, "loss": 0.782, "step": 10318 }, { "epoch": 0.88, "learning_rate": 3.6964001392317874e-07, "loss": 0.833, "step": 10319 }, { "epoch": 0.88, "learning_rate": 3.6911873826463664e-07, "loss": 0.7718, "step": 10320 }, { "epoch": 0.88, "learning_rate": 3.6859781633591053e-07, "loss": 0.73, "step": 10321 }, { "epoch": 0.88, "learning_rate": 3.6807724817679136e-07, "loss": 0.7666, "step": 10322 }, { "epoch": 0.88, "learning_rate": 3.675570338270429e-07, "loss": 0.7701, "step": 10323 }, { "epoch": 0.88, "learning_rate": 3.670371733264011e-07, "loss": 0.7809, "step": 10324 }, { "epoch": 0.88, "learning_rate": 3.6651766671457534e-07, "loss": 0.7748, "step": 10325 }, { "epoch": 0.88, "learning_rate": 3.6599851403124776e-07, "loss": 0.801, "step": 10326 }, { "epoch": 0.88, "learning_rate": 3.654797153160755e-07, "loss": 0.7821, "step": 10327 }, { "epoch": 0.88, "learning_rate": 3.6496127060868635e-07, "loss": 0.7706, "step": 10328 }, { "epoch": 0.88, "learning_rate": 3.6444317994867975e-07, "loss": 0.8002, "step": 10329 }, { "epoch": 0.88, "learning_rate": 3.6392544337563396e-07, "loss": 0.7924, "step": 10330 }, { "epoch": 0.88, "learning_rate": 3.6340806092909186e-07, "loss": 0.8468, "step": 10331 }, { "epoch": 0.88, "learning_rate": 3.6289103264857686e-07, "loss": 0.8302, "step": 10332 }, { "epoch": 0.88, "learning_rate": 3.623743585735812e-07, "loss": 0.7858, "step": 10333 }, { "epoch": 0.88, "learning_rate": 3.6185803874356995e-07, "loss": 0.8063, "step": 10334 }, { "epoch": 0.88, "learning_rate": 3.613420731979844e-07, "loss": 0.7881, "step": 10335 }, { "epoch": 0.88, "learning_rate": 3.608264619762353e-07, "loss": 0.8199, "step": 10336 }, { "epoch": 0.88, "learning_rate": 3.6031120511770714e-07, "loss": 0.8419, "step": 10337 }, { "epoch": 0.88, "learning_rate": 3.59796302661759e-07, "loss": 0.81, "step": 10338 }, { "epoch": 0.88, "learning_rate": 3.592817546477195e-07, "loss": 0.5145, "step": 10339 }, { "epoch": 0.88, "learning_rate": 3.5876756111489554e-07, "loss": 0.7304, "step": 10340 }, { "epoch": 0.88, "learning_rate": 3.582537221025617e-07, "loss": 0.7762, "step": 10341 }, { "epoch": 0.88, "learning_rate": 3.577402376499672e-07, "loss": 0.8252, "step": 10342 }, { "epoch": 0.88, "learning_rate": 3.572271077963374e-07, "loss": 0.7983, "step": 10343 }, { "epoch": 0.88, "learning_rate": 3.5671433258086366e-07, "loss": 0.7536, "step": 10344 }, { "epoch": 0.88, "learning_rate": 3.562019120427174e-07, "loss": 0.8042, "step": 10345 }, { "epoch": 0.88, "learning_rate": 3.5568984622103844e-07, "loss": 0.8272, "step": 10346 }, { "epoch": 0.88, "learning_rate": 3.5517813515494106e-07, "loss": 0.7578, "step": 10347 }, { "epoch": 0.88, "learning_rate": 3.5466677888351285e-07, "loss": 0.7634, "step": 10348 }, { "epoch": 0.88, "learning_rate": 3.541557774458132e-07, "loss": 0.7903, "step": 10349 }, { "epoch": 0.88, "learning_rate": 3.5364513088087637e-07, "loss": 0.7399, "step": 10350 }, { "epoch": 0.88, "learning_rate": 3.531348392277062e-07, "loss": 0.796, "step": 10351 }, { "epoch": 0.88, "learning_rate": 3.526249025252815e-07, "loss": 0.7793, "step": 10352 }, { "epoch": 0.88, "learning_rate": 3.521153208125544e-07, "loss": 0.8105, "step": 10353 }, { "epoch": 0.88, "learning_rate": 3.516060941284488e-07, "loss": 0.8172, "step": 10354 }, { "epoch": 0.88, "learning_rate": 3.5109722251186365e-07, "loss": 0.7605, "step": 10355 }, { "epoch": 0.88, "learning_rate": 3.5058870600166716e-07, "loss": 0.806, "step": 10356 }, { "epoch": 0.88, "learning_rate": 3.500805446367034e-07, "loss": 0.7887, "step": 10357 }, { "epoch": 0.88, "learning_rate": 3.4957273845578785e-07, "loss": 0.5085, "step": 10358 }, { "epoch": 0.88, "learning_rate": 3.490652874977085e-07, "loss": 0.7783, "step": 10359 }, { "epoch": 0.88, "learning_rate": 3.4855819180122874e-07, "loss": 0.8328, "step": 10360 }, { "epoch": 0.88, "learning_rate": 3.48051451405082e-07, "loss": 0.5298, "step": 10361 }, { "epoch": 0.88, "learning_rate": 3.475450663479746e-07, "loss": 0.8268, "step": 10362 }, { "epoch": 0.88, "learning_rate": 3.4703903666858996e-07, "loss": 0.7723, "step": 10363 }, { "epoch": 0.88, "learning_rate": 3.465333624055778e-07, "loss": 0.8489, "step": 10364 }, { "epoch": 0.88, "learning_rate": 3.4602804359756547e-07, "loss": 0.7838, "step": 10365 }, { "epoch": 0.88, "learning_rate": 3.455230802831522e-07, "loss": 0.8001, "step": 10366 }, { "epoch": 0.88, "learning_rate": 3.4501847250090814e-07, "loss": 0.7622, "step": 10367 }, { "epoch": 0.88, "learning_rate": 3.445142202893792e-07, "loss": 0.7866, "step": 10368 }, { "epoch": 0.88, "learning_rate": 3.440103236870823e-07, "loss": 0.8063, "step": 10369 }, { "epoch": 0.88, "learning_rate": 3.435067827325078e-07, "loss": 0.8311, "step": 10370 }, { "epoch": 0.88, "learning_rate": 3.430035974641177e-07, "loss": 0.8029, "step": 10371 }, { "epoch": 0.89, "learning_rate": 3.425007679203479e-07, "loss": 0.7929, "step": 10372 }, { "epoch": 0.89, "learning_rate": 3.4199829413960826e-07, "loss": 0.8474, "step": 10373 }, { "epoch": 0.89, "learning_rate": 3.4149617616027974e-07, "loss": 0.7757, "step": 10374 }, { "epoch": 0.89, "learning_rate": 3.409944140207155e-07, "loss": 0.8057, "step": 10375 }, { "epoch": 0.89, "learning_rate": 3.4049300775924386e-07, "loss": 0.7977, "step": 10376 }, { "epoch": 0.89, "learning_rate": 3.3999195741416467e-07, "loss": 0.7875, "step": 10377 }, { "epoch": 0.89, "learning_rate": 3.394912630237507e-07, "loss": 0.7747, "step": 10378 }, { "epoch": 0.89, "learning_rate": 3.3899092462624683e-07, "loss": 0.7744, "step": 10379 }, { "epoch": 0.89, "learning_rate": 3.384909422598709e-07, "loss": 0.7494, "step": 10380 }, { "epoch": 0.89, "learning_rate": 3.3799131596281566e-07, "loss": 0.7222, "step": 10381 }, { "epoch": 0.89, "learning_rate": 3.3749204577324335e-07, "loss": 0.7105, "step": 10382 }, { "epoch": 0.89, "learning_rate": 3.369931317292935e-07, "loss": 0.7881, "step": 10383 }, { "epoch": 0.89, "learning_rate": 3.3649457386907336e-07, "loss": 0.7822, "step": 10384 }, { "epoch": 0.89, "learning_rate": 3.359963722306642e-07, "loss": 0.807, "step": 10385 }, { "epoch": 0.89, "learning_rate": 3.354985268521238e-07, "loss": 0.7957, "step": 10386 }, { "epoch": 0.89, "learning_rate": 3.3500103777147795e-07, "loss": 0.7762, "step": 10387 }, { "epoch": 0.89, "learning_rate": 3.3450390502672957e-07, "loss": 0.8155, "step": 10388 }, { "epoch": 0.89, "learning_rate": 3.340071286558505e-07, "loss": 0.8099, "step": 10389 }, { "epoch": 0.89, "learning_rate": 3.3351070869678757e-07, "loss": 0.7968, "step": 10390 }, { "epoch": 0.89, "learning_rate": 3.330146451874594e-07, "loss": 0.8126, "step": 10391 }, { "epoch": 0.89, "learning_rate": 3.3251893816575733e-07, "loss": 0.8329, "step": 10392 }, { "epoch": 0.89, "learning_rate": 3.320235876695477e-07, "loss": 0.7532, "step": 10393 }, { "epoch": 0.89, "learning_rate": 3.315285937366669e-07, "loss": 0.8181, "step": 10394 }, { "epoch": 0.89, "learning_rate": 3.3103395640492365e-07, "loss": 0.469, "step": 10395 }, { "epoch": 0.89, "learning_rate": 3.3053967571210375e-07, "loss": 0.7851, "step": 10396 }, { "epoch": 0.89, "learning_rate": 3.3004575169595987e-07, "loss": 0.8458, "step": 10397 }, { "epoch": 0.89, "learning_rate": 3.2955218439422233e-07, "loss": 0.7714, "step": 10398 }, { "epoch": 0.89, "learning_rate": 3.290589738445915e-07, "loss": 0.7922, "step": 10399 }, { "epoch": 0.89, "learning_rate": 3.2856612008474066e-07, "loss": 0.8136, "step": 10400 }, { "epoch": 0.89, "learning_rate": 3.280736231523185e-07, "loss": 0.8341, "step": 10401 }, { "epoch": 0.89, "learning_rate": 3.275814830849422e-07, "loss": 0.8089, "step": 10402 }, { "epoch": 0.89, "learning_rate": 3.2708969992020555e-07, "loss": 0.7401, "step": 10403 }, { "epoch": 0.89, "learning_rate": 3.265982736956719e-07, "loss": 0.7792, "step": 10404 }, { "epoch": 0.89, "learning_rate": 3.261072044488789e-07, "loss": 0.7669, "step": 10405 }, { "epoch": 0.89, "learning_rate": 3.256164922173388e-07, "loss": 0.8235, "step": 10406 }, { "epoch": 0.89, "learning_rate": 3.251261370385328e-07, "loss": 0.8059, "step": 10407 }, { "epoch": 0.89, "learning_rate": 3.246361389499164e-07, "loss": 0.7739, "step": 10408 }, { "epoch": 0.89, "learning_rate": 3.2414649798891975e-07, "loss": 0.8099, "step": 10409 }, { "epoch": 0.89, "learning_rate": 3.236572141929434e-07, "loss": 0.7787, "step": 10410 }, { "epoch": 0.89, "learning_rate": 3.2316828759936134e-07, "loss": 0.852, "step": 10411 }, { "epoch": 0.89, "learning_rate": 3.226797182455199e-07, "loss": 0.7034, "step": 10412 }, { "epoch": 0.89, "learning_rate": 3.221915061687375e-07, "loss": 0.782, "step": 10413 }, { "epoch": 0.89, "learning_rate": 3.217036514063082e-07, "loss": 0.7964, "step": 10414 }, { "epoch": 0.89, "learning_rate": 3.2121615399549557e-07, "loss": 0.7709, "step": 10415 }, { "epoch": 0.89, "learning_rate": 3.2072901397353807e-07, "loss": 0.8093, "step": 10416 }, { "epoch": 0.89, "learning_rate": 3.202422313776443e-07, "loss": 0.7613, "step": 10417 }, { "epoch": 0.89, "learning_rate": 3.197558062449979e-07, "loss": 0.7739, "step": 10418 }, { "epoch": 0.89, "learning_rate": 3.1926973861275526e-07, "loss": 0.8015, "step": 10419 }, { "epoch": 0.89, "learning_rate": 3.187840285180438e-07, "loss": 0.807, "step": 10420 }, { "epoch": 0.89, "learning_rate": 3.1829867599796394e-07, "loss": 0.7919, "step": 10421 }, { "epoch": 0.89, "learning_rate": 3.1781368108959096e-07, "loss": 0.8261, "step": 10422 }, { "epoch": 0.89, "learning_rate": 3.1732904382996975e-07, "loss": 0.8093, "step": 10423 }, { "epoch": 0.89, "learning_rate": 3.1684476425612063e-07, "loss": 0.7927, "step": 10424 }, { "epoch": 0.89, "learning_rate": 3.1636084240503294e-07, "loss": 0.5112, "step": 10425 }, { "epoch": 0.89, "learning_rate": 3.1587727831367374e-07, "loss": 0.7968, "step": 10426 }, { "epoch": 0.89, "learning_rate": 3.153940720189791e-07, "loss": 0.7897, "step": 10427 }, { "epoch": 0.89, "learning_rate": 3.1491122355785775e-07, "loss": 0.7998, "step": 10428 }, { "epoch": 0.89, "learning_rate": 3.1442873296719357e-07, "loss": 0.7867, "step": 10429 }, { "epoch": 0.89, "learning_rate": 3.1394660028384093e-07, "loss": 0.4859, "step": 10430 }, { "epoch": 0.89, "learning_rate": 3.134648255446271e-07, "loss": 0.8267, "step": 10431 }, { "epoch": 0.89, "learning_rate": 3.1298340878635316e-07, "loss": 0.7692, "step": 10432 }, { "epoch": 0.89, "learning_rate": 3.125023500457913e-07, "loss": 0.7507, "step": 10433 }, { "epoch": 0.89, "learning_rate": 3.1202164935968784e-07, "loss": 0.7463, "step": 10434 }, { "epoch": 0.89, "learning_rate": 3.1154130676476156e-07, "loss": 0.7982, "step": 10435 }, { "epoch": 0.89, "learning_rate": 3.110613222977027e-07, "loss": 0.7918, "step": 10436 }, { "epoch": 0.89, "learning_rate": 3.1058169599517464e-07, "loss": 0.8204, "step": 10437 }, { "epoch": 0.89, "learning_rate": 3.1010242789381364e-07, "loss": 0.7318, "step": 10438 }, { "epoch": 0.89, "learning_rate": 3.096235180302293e-07, "loss": 0.494, "step": 10439 }, { "epoch": 0.89, "learning_rate": 3.091449664410029e-07, "loss": 0.7747, "step": 10440 }, { "epoch": 0.89, "learning_rate": 3.08666773162688e-07, "loss": 0.7954, "step": 10441 }, { "epoch": 0.89, "learning_rate": 3.081889382318121e-07, "loss": 0.7864, "step": 10442 }, { "epoch": 0.89, "learning_rate": 3.0771146168487474e-07, "loss": 0.7877, "step": 10443 }, { "epoch": 0.89, "learning_rate": 3.0723434355834745e-07, "loss": 0.7944, "step": 10444 }, { "epoch": 0.89, "learning_rate": 3.0675758388867494e-07, "loss": 0.8182, "step": 10445 }, { "epoch": 0.89, "learning_rate": 3.0628118271227415e-07, "loss": 0.7406, "step": 10446 }, { "epoch": 0.89, "learning_rate": 3.0580514006553595e-07, "loss": 0.8129, "step": 10447 }, { "epoch": 0.89, "learning_rate": 3.053294559848224e-07, "loss": 0.8123, "step": 10448 }, { "epoch": 0.89, "learning_rate": 3.048541305064684e-07, "loss": 0.8187, "step": 10449 }, { "epoch": 0.89, "learning_rate": 3.04379163666782e-07, "loss": 0.7934, "step": 10450 }, { "epoch": 0.89, "learning_rate": 3.039045555020426e-07, "loss": 0.7679, "step": 10451 }, { "epoch": 0.89, "learning_rate": 3.0343030604850445e-07, "loss": 0.7372, "step": 10452 }, { "epoch": 0.89, "learning_rate": 3.029564153423925e-07, "loss": 0.7945, "step": 10453 }, { "epoch": 0.89, "learning_rate": 3.024828834199045e-07, "loss": 0.7748, "step": 10454 }, { "epoch": 0.89, "learning_rate": 3.0200971031721203e-07, "loss": 0.8199, "step": 10455 }, { "epoch": 0.89, "learning_rate": 3.015368960704584e-07, "loss": 0.7703, "step": 10456 }, { "epoch": 0.89, "learning_rate": 3.010644407157587e-07, "loss": 0.7868, "step": 10457 }, { "epoch": 0.89, "learning_rate": 3.0059234428920225e-07, "loss": 0.7935, "step": 10458 }, { "epoch": 0.89, "learning_rate": 3.001206068268486e-07, "loss": 0.8024, "step": 10459 }, { "epoch": 0.89, "learning_rate": 2.996492283647334e-07, "loss": 0.7745, "step": 10460 }, { "epoch": 0.89, "learning_rate": 2.991782089388612e-07, "loss": 0.8097, "step": 10461 }, { "epoch": 0.89, "learning_rate": 2.987075485852126e-07, "loss": 0.7562, "step": 10462 }, { "epoch": 0.89, "learning_rate": 2.982372473397377e-07, "loss": 0.8071, "step": 10463 }, { "epoch": 0.89, "learning_rate": 2.977673052383612e-07, "loss": 0.738, "step": 10464 }, { "epoch": 0.89, "learning_rate": 2.972977223169787e-07, "loss": 0.7627, "step": 10465 }, { "epoch": 0.89, "learning_rate": 2.968284986114595e-07, "loss": 0.7856, "step": 10466 }, { "epoch": 0.89, "learning_rate": 2.963596341576458e-07, "loss": 0.7508, "step": 10467 }, { "epoch": 0.89, "learning_rate": 2.9589112899135185e-07, "loss": 0.7689, "step": 10468 }, { "epoch": 0.89, "learning_rate": 2.95422983148364e-07, "loss": 0.8158, "step": 10469 }, { "epoch": 0.89, "learning_rate": 2.949551966644415e-07, "loss": 0.7655, "step": 10470 }, { "epoch": 0.89, "learning_rate": 2.944877695753157e-07, "loss": 0.7844, "step": 10471 }, { "epoch": 0.89, "learning_rate": 2.9402070191669297e-07, "loss": 0.8326, "step": 10472 }, { "epoch": 0.89, "learning_rate": 2.9355399372424817e-07, "loss": 0.7796, "step": 10473 }, { "epoch": 0.89, "learning_rate": 2.930876450336312e-07, "loss": 0.8092, "step": 10474 }, { "epoch": 0.89, "learning_rate": 2.9262165588046566e-07, "loss": 0.8053, "step": 10475 }, { "epoch": 0.89, "learning_rate": 2.921560263003448e-07, "loss": 0.7702, "step": 10476 }, { "epoch": 0.89, "learning_rate": 2.916907563288357e-07, "loss": 0.7972, "step": 10477 }, { "epoch": 0.89, "learning_rate": 2.912258460014783e-07, "loss": 0.8094, "step": 10478 }, { "epoch": 0.89, "learning_rate": 2.907612953537836e-07, "loss": 0.8133, "step": 10479 }, { "epoch": 0.89, "learning_rate": 2.9029710442123883e-07, "loss": 0.7973, "step": 10480 }, { "epoch": 0.89, "learning_rate": 2.898332732392994e-07, "loss": 0.7421, "step": 10481 }, { "epoch": 0.89, "learning_rate": 2.893698018433944e-07, "loss": 0.5053, "step": 10482 }, { "epoch": 0.89, "learning_rate": 2.8890669026892917e-07, "loss": 0.744, "step": 10483 }, { "epoch": 0.89, "learning_rate": 2.884439385512744e-07, "loss": 0.7632, "step": 10484 }, { "epoch": 0.89, "learning_rate": 2.879815467257807e-07, "loss": 0.7465, "step": 10485 }, { "epoch": 0.89, "learning_rate": 2.8751951482776597e-07, "loss": 0.7991, "step": 10486 }, { "epoch": 0.89, "learning_rate": 2.8705784289252247e-07, "loss": 0.8004, "step": 10487 }, { "epoch": 0.89, "learning_rate": 2.8659653095531647e-07, "loss": 0.7921, "step": 10488 }, { "epoch": 0.9, "learning_rate": 2.861355790513848e-07, "loss": 0.7798, "step": 10489 }, { "epoch": 0.9, "learning_rate": 2.8567498721593647e-07, "loss": 0.8208, "step": 10490 }, { "epoch": 0.9, "learning_rate": 2.8521475548415444e-07, "loss": 0.8411, "step": 10491 }, { "epoch": 0.9, "learning_rate": 2.847548838911923e-07, "loss": 0.7248, "step": 10492 }, { "epoch": 0.9, "learning_rate": 2.842953724721792e-07, "loss": 0.7732, "step": 10493 }, { "epoch": 0.9, "learning_rate": 2.838362212622131e-07, "loss": 0.7679, "step": 10494 }, { "epoch": 0.9, "learning_rate": 2.8337743029636823e-07, "loss": 0.7825, "step": 10495 }, { "epoch": 0.9, "learning_rate": 2.8291899960968925e-07, "loss": 0.7945, "step": 10496 }, { "epoch": 0.9, "learning_rate": 2.824609292371905e-07, "loss": 0.7793, "step": 10497 }, { "epoch": 0.9, "learning_rate": 2.820032192138644e-07, "loss": 0.824, "step": 10498 }, { "epoch": 0.9, "learning_rate": 2.81545869574672e-07, "loss": 0.8034, "step": 10499 }, { "epoch": 0.9, "learning_rate": 2.810888803545486e-07, "loss": 0.7577, "step": 10500 }, { "epoch": 0.9, "learning_rate": 2.8063225158840134e-07, "loss": 0.7625, "step": 10501 }, { "epoch": 0.9, "learning_rate": 2.801759833111084e-07, "loss": 0.821, "step": 10502 }, { "epoch": 0.9, "learning_rate": 2.797200755575241e-07, "loss": 0.7796, "step": 10503 }, { "epoch": 0.9, "learning_rate": 2.792645283624712e-07, "loss": 0.7808, "step": 10504 }, { "epoch": 0.9, "learning_rate": 2.788093417607474e-07, "loss": 0.7092, "step": 10505 }, { "epoch": 0.9, "learning_rate": 2.783545157871215e-07, "loss": 0.7865, "step": 10506 }, { "epoch": 0.9, "learning_rate": 2.7790005047633574e-07, "loss": 0.7994, "step": 10507 }, { "epoch": 0.9, "learning_rate": 2.7744594586310513e-07, "loss": 0.7686, "step": 10508 }, { "epoch": 0.9, "learning_rate": 2.7699220198211584e-07, "loss": 0.7613, "step": 10509 }, { "epoch": 0.9, "learning_rate": 2.765388188680268e-07, "loss": 0.8015, "step": 10510 }, { "epoch": 0.9, "learning_rate": 2.760857965554703e-07, "loss": 0.8303, "step": 10511 }, { "epoch": 0.9, "learning_rate": 2.756331350790492e-07, "loss": 0.7463, "step": 10512 }, { "epoch": 0.9, "learning_rate": 2.7518083447334145e-07, "loss": 0.807, "step": 10513 }, { "epoch": 0.9, "learning_rate": 2.74728894772896e-07, "loss": 0.8456, "step": 10514 }, { "epoch": 0.9, "learning_rate": 2.7427731601223253e-07, "loss": 0.7985, "step": 10515 }, { "epoch": 0.9, "learning_rate": 2.7382609822584783e-07, "loss": 0.7733, "step": 10516 }, { "epoch": 0.9, "learning_rate": 2.733752414482055e-07, "loss": 0.8487, "step": 10517 }, { "epoch": 0.9, "learning_rate": 2.7292474571374516e-07, "loss": 0.8138, "step": 10518 }, { "epoch": 0.9, "learning_rate": 2.724746110568782e-07, "loss": 0.7466, "step": 10519 }, { "epoch": 0.9, "learning_rate": 2.7202483751198773e-07, "loss": 0.7377, "step": 10520 }, { "epoch": 0.9, "learning_rate": 2.7157542511343e-07, "loss": 0.797, "step": 10521 }, { "epoch": 0.9, "learning_rate": 2.711263738955333e-07, "loss": 0.7774, "step": 10522 }, { "epoch": 0.9, "learning_rate": 2.706776838925984e-07, "loss": 0.7781, "step": 10523 }, { "epoch": 0.9, "learning_rate": 2.7022935513889847e-07, "loss": 0.8105, "step": 10524 }, { "epoch": 0.9, "learning_rate": 2.6978138766867833e-07, "loss": 0.7846, "step": 10525 }, { "epoch": 0.9, "learning_rate": 2.693337815161573e-07, "loss": 0.8042, "step": 10526 }, { "epoch": 0.9, "learning_rate": 2.688865367155247e-07, "loss": 0.7793, "step": 10527 }, { "epoch": 0.9, "learning_rate": 2.6843965330094477e-07, "loss": 0.7699, "step": 10528 }, { "epoch": 0.9, "learning_rate": 2.67993131306552e-07, "loss": 0.7947, "step": 10529 }, { "epoch": 0.9, "learning_rate": 2.6754697076645243e-07, "loss": 0.7553, "step": 10530 }, { "epoch": 0.9, "learning_rate": 2.671011717147276e-07, "loss": 0.8188, "step": 10531 }, { "epoch": 0.9, "learning_rate": 2.6665573418542924e-07, "loss": 0.8255, "step": 10532 }, { "epoch": 0.9, "learning_rate": 2.6621065821258283e-07, "loss": 0.7776, "step": 10533 }, { "epoch": 0.9, "learning_rate": 2.657659438301857e-07, "loss": 0.79, "step": 10534 }, { "epoch": 0.9, "learning_rate": 2.653215910722051e-07, "loss": 0.7354, "step": 10535 }, { "epoch": 0.9, "learning_rate": 2.648775999725867e-07, "loss": 0.7655, "step": 10536 }, { "epoch": 0.9, "learning_rate": 2.64433970565241e-07, "loss": 0.7918, "step": 10537 }, { "epoch": 0.9, "learning_rate": 2.639907028840566e-07, "loss": 0.7866, "step": 10538 }, { "epoch": 0.9, "learning_rate": 2.6354779696289244e-07, "loss": 0.7979, "step": 10539 }, { "epoch": 0.9, "learning_rate": 2.6310525283557875e-07, "loss": 0.79, "step": 10540 }, { "epoch": 0.9, "learning_rate": 2.626630705359207e-07, "loss": 0.7578, "step": 10541 }, { "epoch": 0.9, "learning_rate": 2.622212500976934e-07, "loss": 0.7857, "step": 10542 }, { "epoch": 0.9, "learning_rate": 2.6177979155464606e-07, "loss": 0.7552, "step": 10543 }, { "epoch": 0.9, "learning_rate": 2.613386949404989e-07, "loss": 0.8058, "step": 10544 }, { "epoch": 0.9, "learning_rate": 2.60897960288945e-07, "loss": 0.7811, "step": 10545 }, { "epoch": 0.9, "learning_rate": 2.6045758763365013e-07, "loss": 0.7848, "step": 10546 }, { "epoch": 0.9, "learning_rate": 2.6001757700825245e-07, "loss": 0.7777, "step": 10547 }, { "epoch": 0.9, "learning_rate": 2.595779284463612e-07, "loss": 0.8158, "step": 10548 }, { "epoch": 0.9, "learning_rate": 2.591386419815611e-07, "loss": 0.8, "step": 10549 }, { "epoch": 0.9, "learning_rate": 2.586997176474043e-07, "loss": 0.7899, "step": 10550 }, { "epoch": 0.9, "learning_rate": 2.5826115547741947e-07, "loss": 0.8051, "step": 10551 }, { "epoch": 0.9, "learning_rate": 2.5782295550510596e-07, "loss": 0.8183, "step": 10552 }, { "epoch": 0.9, "learning_rate": 2.573851177639353e-07, "loss": 0.7984, "step": 10553 }, { "epoch": 0.9, "learning_rate": 2.569476422873529e-07, "loss": 0.792, "step": 10554 }, { "epoch": 0.9, "learning_rate": 2.565105291087738e-07, "loss": 0.8256, "step": 10555 }, { "epoch": 0.9, "learning_rate": 2.5607377826158853e-07, "loss": 0.7405, "step": 10556 }, { "epoch": 0.9, "learning_rate": 2.556373897791564e-07, "loss": 0.7884, "step": 10557 }, { "epoch": 0.9, "learning_rate": 2.5520136369481194e-07, "loss": 0.7784, "step": 10558 }, { "epoch": 0.9, "learning_rate": 2.547657000418613e-07, "loss": 0.8302, "step": 10559 }, { "epoch": 0.9, "learning_rate": 2.543303988535828e-07, "loss": 0.8209, "step": 10560 }, { "epoch": 0.9, "learning_rate": 2.5389546016322543e-07, "loss": 0.7777, "step": 10561 }, { "epoch": 0.9, "learning_rate": 2.5346088400401427e-07, "loss": 0.808, "step": 10562 }, { "epoch": 0.9, "learning_rate": 2.5302667040914166e-07, "loss": 0.7855, "step": 10563 }, { "epoch": 0.9, "learning_rate": 2.5259281941177715e-07, "loss": 0.8292, "step": 10564 }, { "epoch": 0.9, "learning_rate": 2.521593310450593e-07, "loss": 0.8191, "step": 10565 }, { "epoch": 0.9, "learning_rate": 2.5172620534210156e-07, "loss": 0.7449, "step": 10566 }, { "epoch": 0.9, "learning_rate": 2.512934423359864e-07, "loss": 0.7551, "step": 10567 }, { "epoch": 0.9, "learning_rate": 2.5086104205977123e-07, "loss": 0.7887, "step": 10568 }, { "epoch": 0.9, "learning_rate": 2.504290045464863e-07, "loss": 0.7824, "step": 10569 }, { "epoch": 0.9, "learning_rate": 2.4999732982912973e-07, "loss": 0.813, "step": 10570 }, { "epoch": 0.9, "learning_rate": 2.495660179406778e-07, "loss": 0.821, "step": 10571 }, { "epoch": 0.9, "learning_rate": 2.491350689140748e-07, "loss": 0.8146, "step": 10572 }, { "epoch": 0.9, "learning_rate": 2.487044827822388e-07, "loss": 0.7674, "step": 10573 }, { "epoch": 0.9, "learning_rate": 2.4827425957806074e-07, "loss": 0.7577, "step": 10574 }, { "epoch": 0.9, "learning_rate": 2.478443993344032e-07, "loss": 0.7527, "step": 10575 }, { "epoch": 0.9, "learning_rate": 2.47414902084101e-07, "loss": 0.7398, "step": 10576 }, { "epoch": 0.9, "learning_rate": 2.4698576785996075e-07, "loss": 0.5391, "step": 10577 }, { "epoch": 0.9, "learning_rate": 2.4655699669476174e-07, "loss": 0.5332, "step": 10578 }, { "epoch": 0.9, "learning_rate": 2.4612858862125664e-07, "loss": 0.8334, "step": 10579 }, { "epoch": 0.9, "learning_rate": 2.4570054367216876e-07, "loss": 0.8429, "step": 10580 }, { "epoch": 0.9, "learning_rate": 2.4527286188019416e-07, "loss": 0.8212, "step": 10581 }, { "epoch": 0.9, "learning_rate": 2.4484554327800225e-07, "loss": 0.7548, "step": 10582 }, { "epoch": 0.9, "learning_rate": 2.4441858789823246e-07, "loss": 0.7664, "step": 10583 }, { "epoch": 0.9, "learning_rate": 2.439919957734987e-07, "loss": 0.765, "step": 10584 }, { "epoch": 0.9, "learning_rate": 2.4356576693638555e-07, "loss": 0.7795, "step": 10585 }, { "epoch": 0.9, "learning_rate": 2.4313990141945077e-07, "loss": 0.6949, "step": 10586 }, { "epoch": 0.9, "learning_rate": 2.427143992552244e-07, "loss": 0.8, "step": 10587 }, { "epoch": 0.9, "learning_rate": 2.422892604762078e-07, "loss": 0.7692, "step": 10588 }, { "epoch": 0.9, "learning_rate": 2.41864485114876e-07, "loss": 0.8323, "step": 10589 }, { "epoch": 0.9, "learning_rate": 2.4144007320367534e-07, "loss": 0.7635, "step": 10590 }, { "epoch": 0.9, "learning_rate": 2.410160247750232e-07, "loss": 0.804, "step": 10591 }, { "epoch": 0.9, "learning_rate": 2.4059233986131194e-07, "loss": 0.7332, "step": 10592 }, { "epoch": 0.9, "learning_rate": 2.401690184949046e-07, "loss": 0.7926, "step": 10593 }, { "epoch": 0.9, "learning_rate": 2.3974606070813586e-07, "loss": 0.8104, "step": 10594 }, { "epoch": 0.9, "learning_rate": 2.393234665333144e-07, "loss": 0.76, "step": 10595 }, { "epoch": 0.9, "learning_rate": 2.3890123600271875e-07, "loss": 0.7924, "step": 10596 }, { "epoch": 0.9, "learning_rate": 2.38479369148602e-07, "loss": 0.7783, "step": 10597 }, { "epoch": 0.9, "learning_rate": 2.380578660031885e-07, "loss": 0.7798, "step": 10598 }, { "epoch": 0.9, "learning_rate": 2.3763672659867287e-07, "loss": 0.7829, "step": 10599 }, { "epoch": 0.9, "learning_rate": 2.372159509672267e-07, "loss": 0.785, "step": 10600 }, { "epoch": 0.9, "learning_rate": 2.3679553914098873e-07, "loss": 0.8039, "step": 10601 }, { "epoch": 0.9, "learning_rate": 2.3637549115207437e-07, "loss": 0.7962, "step": 10602 }, { "epoch": 0.9, "learning_rate": 2.3595580703256682e-07, "loss": 0.7446, "step": 10603 }, { "epoch": 0.9, "learning_rate": 2.3553648681452378e-07, "loss": 0.7931, "step": 10604 }, { "epoch": 0.9, "learning_rate": 2.3511753052997633e-07, "loss": 0.7764, "step": 10605 }, { "epoch": 0.91, "learning_rate": 2.3469893821092493e-07, "loss": 0.7291, "step": 10606 }, { "epoch": 0.91, "learning_rate": 2.342807098893457e-07, "loss": 0.7626, "step": 10607 }, { "epoch": 0.91, "learning_rate": 2.3386284559718364e-07, "loss": 0.5356, "step": 10608 }, { "epoch": 0.91, "learning_rate": 2.334453453663571e-07, "loss": 0.7699, "step": 10609 }, { "epoch": 0.91, "learning_rate": 2.330282092287578e-07, "loss": 0.8339, "step": 10610 }, { "epoch": 0.91, "learning_rate": 2.3261143721624747e-07, "loss": 0.7204, "step": 10611 }, { "epoch": 0.91, "learning_rate": 2.3219502936066228e-07, "loss": 0.7893, "step": 10612 }, { "epoch": 0.91, "learning_rate": 2.3177898569380908e-07, "loss": 0.7977, "step": 10613 }, { "epoch": 0.91, "learning_rate": 2.313633062474674e-07, "loss": 0.7369, "step": 10614 }, { "epoch": 0.91, "learning_rate": 2.309479910533896e-07, "loss": 0.8334, "step": 10615 }, { "epoch": 0.91, "learning_rate": 2.3053304014329813e-07, "loss": 0.7661, "step": 10616 }, { "epoch": 0.91, "learning_rate": 2.301184535488904e-07, "loss": 0.738, "step": 10617 }, { "epoch": 0.91, "learning_rate": 2.2970423130183383e-07, "loss": 0.832, "step": 10618 }, { "epoch": 0.91, "learning_rate": 2.2929037343376815e-07, "loss": 0.5297, "step": 10619 }, { "epoch": 0.91, "learning_rate": 2.2887687997630746e-07, "loss": 0.8148, "step": 10620 }, { "epoch": 0.91, "learning_rate": 2.2846375096103545e-07, "loss": 0.7777, "step": 10621 }, { "epoch": 0.91, "learning_rate": 2.2805098641950962e-07, "loss": 0.7419, "step": 10622 }, { "epoch": 0.91, "learning_rate": 2.2763858638325808e-07, "loss": 0.7945, "step": 10623 }, { "epoch": 0.91, "learning_rate": 2.2722655088378233e-07, "loss": 0.7812, "step": 10624 }, { "epoch": 0.91, "learning_rate": 2.2681487995255601e-07, "loss": 0.7772, "step": 10625 }, { "epoch": 0.91, "learning_rate": 2.2640357362102516e-07, "loss": 0.7794, "step": 10626 }, { "epoch": 0.91, "learning_rate": 2.2599263192060516e-07, "loss": 0.7384, "step": 10627 }, { "epoch": 0.91, "learning_rate": 2.255820548826887e-07, "loss": 0.7687, "step": 10628 }, { "epoch": 0.91, "learning_rate": 2.2517184253863567e-07, "loss": 0.7779, "step": 10629 }, { "epoch": 0.91, "learning_rate": 2.2476199491978102e-07, "loss": 0.7931, "step": 10630 }, { "epoch": 0.91, "learning_rate": 2.243525120574308e-07, "loss": 0.8128, "step": 10631 }, { "epoch": 0.91, "learning_rate": 2.239433939828628e-07, "loss": 0.7624, "step": 10632 }, { "epoch": 0.91, "learning_rate": 2.2353464072732868e-07, "loss": 0.7474, "step": 10633 }, { "epoch": 0.91, "learning_rate": 2.2312625232204955e-07, "loss": 0.8365, "step": 10634 }, { "epoch": 0.91, "learning_rate": 2.2271822879822215e-07, "loss": 0.8114, "step": 10635 }, { "epoch": 0.91, "learning_rate": 2.2231057018701152e-07, "loss": 0.7889, "step": 10636 }, { "epoch": 0.91, "learning_rate": 2.2190327651955613e-07, "loss": 0.7954, "step": 10637 }, { "epoch": 0.91, "learning_rate": 2.2149634782696994e-07, "loss": 0.8051, "step": 10638 }, { "epoch": 0.91, "learning_rate": 2.210897841403331e-07, "loss": 0.7762, "step": 10639 }, { "epoch": 0.91, "learning_rate": 2.2068358549070357e-07, "loss": 0.8023, "step": 10640 }, { "epoch": 0.91, "learning_rate": 2.2027775190910705e-07, "loss": 0.7965, "step": 10641 }, { "epoch": 0.91, "learning_rate": 2.1987228342654432e-07, "loss": 0.7843, "step": 10642 }, { "epoch": 0.91, "learning_rate": 2.1946718007398616e-07, "loss": 0.7499, "step": 10643 }, { "epoch": 0.91, "learning_rate": 2.1906244188237613e-07, "loss": 0.7823, "step": 10644 }, { "epoch": 0.91, "learning_rate": 2.1865806888263118e-07, "loss": 0.7633, "step": 10645 }, { "epoch": 0.91, "learning_rate": 2.182540611056394e-07, "loss": 0.7374, "step": 10646 }, { "epoch": 0.91, "learning_rate": 2.178504185822594e-07, "loss": 0.8143, "step": 10647 }, { "epoch": 0.91, "learning_rate": 2.17447141343326e-07, "loss": 0.8129, "step": 10648 }, { "epoch": 0.91, "learning_rate": 2.1704422941964008e-07, "loss": 0.8123, "step": 10649 }, { "epoch": 0.91, "learning_rate": 2.166416828419815e-07, "loss": 0.781, "step": 10650 }, { "epoch": 0.91, "learning_rate": 2.1623950164109674e-07, "loss": 0.8195, "step": 10651 }, { "epoch": 0.91, "learning_rate": 2.1583768584770626e-07, "loss": 0.8099, "step": 10652 }, { "epoch": 0.91, "learning_rate": 2.1543623549250382e-07, "loss": 0.7847, "step": 10653 }, { "epoch": 0.91, "learning_rate": 2.1503515060615375e-07, "loss": 0.7659, "step": 10654 }, { "epoch": 0.91, "learning_rate": 2.1463443121929327e-07, "loss": 0.7469, "step": 10655 }, { "epoch": 0.91, "learning_rate": 2.142340773625312e-07, "loss": 0.7991, "step": 10656 }, { "epoch": 0.91, "learning_rate": 2.13834089066447e-07, "loss": 0.8076, "step": 10657 }, { "epoch": 0.91, "learning_rate": 2.1343446636159614e-07, "loss": 0.797, "step": 10658 }, { "epoch": 0.91, "learning_rate": 2.1303520927850263e-07, "loss": 0.775, "step": 10659 }, { "epoch": 0.91, "learning_rate": 2.126363178476637e-07, "loss": 0.7851, "step": 10660 }, { "epoch": 0.91, "learning_rate": 2.1223779209954885e-07, "loss": 0.7986, "step": 10661 }, { "epoch": 0.91, "learning_rate": 2.118396320645999e-07, "loss": 0.7978, "step": 10662 }, { "epoch": 0.91, "learning_rate": 2.1144183777322968e-07, "loss": 0.7583, "step": 10663 }, { "epoch": 0.91, "learning_rate": 2.1104440925582393e-07, "loss": 0.7718, "step": 10664 }, { "epoch": 0.91, "learning_rate": 2.1064734654273944e-07, "loss": 0.7832, "step": 10665 }, { "epoch": 0.91, "learning_rate": 2.1025064966430697e-07, "loss": 0.82, "step": 10666 }, { "epoch": 0.91, "learning_rate": 2.0985431865082783e-07, "loss": 0.7582, "step": 10667 }, { "epoch": 0.91, "learning_rate": 2.094583535325767e-07, "loss": 0.775, "step": 10668 }, { "epoch": 0.91, "learning_rate": 2.0906275433979773e-07, "loss": 0.7875, "step": 10669 }, { "epoch": 0.91, "learning_rate": 2.086675211027084e-07, "loss": 0.781, "step": 10670 }, { "epoch": 0.91, "learning_rate": 2.0827265385150064e-07, "loss": 0.8115, "step": 10671 }, { "epoch": 0.91, "learning_rate": 2.0787815261633481e-07, "loss": 0.7856, "step": 10672 }, { "epoch": 0.91, "learning_rate": 2.074840174273457e-07, "loss": 0.8379, "step": 10673 }, { "epoch": 0.91, "learning_rate": 2.0709024831463974e-07, "loss": 0.819, "step": 10674 }, { "epoch": 0.91, "learning_rate": 2.0669684530829348e-07, "loss": 0.7042, "step": 10675 }, { "epoch": 0.91, "learning_rate": 2.063038084383584e-07, "loss": 0.7838, "step": 10676 }, { "epoch": 0.91, "learning_rate": 2.059111377348555e-07, "loss": 0.7698, "step": 10677 }, { "epoch": 0.91, "learning_rate": 2.0551883322777966e-07, "loss": 0.7859, "step": 10678 }, { "epoch": 0.91, "learning_rate": 2.0512689494709691e-07, "loss": 0.8121, "step": 10679 }, { "epoch": 0.91, "learning_rate": 2.04735322922745e-07, "loss": 0.763, "step": 10680 }, { "epoch": 0.91, "learning_rate": 2.0434411718463555e-07, "loss": 0.7498, "step": 10681 }, { "epoch": 0.91, "learning_rate": 2.0395327776264906e-07, "loss": 0.6724, "step": 10682 }, { "epoch": 0.91, "learning_rate": 2.0356280468664113e-07, "loss": 0.7695, "step": 10683 }, { "epoch": 0.91, "learning_rate": 2.0317269798643734e-07, "loss": 0.7929, "step": 10684 }, { "epoch": 0.91, "learning_rate": 2.0278295769183555e-07, "loss": 0.5311, "step": 10685 }, { "epoch": 0.91, "learning_rate": 2.0239358383260687e-07, "loss": 0.7556, "step": 10686 }, { "epoch": 0.91, "learning_rate": 2.020045764384937e-07, "loss": 0.8159, "step": 10687 }, { "epoch": 0.91, "learning_rate": 2.0161593553921e-07, "loss": 0.829, "step": 10688 }, { "epoch": 0.91, "learning_rate": 2.0122766116444204e-07, "loss": 0.7384, "step": 10689 }, { "epoch": 0.91, "learning_rate": 2.0083975334384775e-07, "loss": 0.7257, "step": 10690 }, { "epoch": 0.91, "learning_rate": 2.00452212107059e-07, "loss": 0.7805, "step": 10691 }, { "epoch": 0.91, "learning_rate": 2.0006503748367655e-07, "loss": 0.7576, "step": 10692 }, { "epoch": 0.91, "learning_rate": 1.9967822950327453e-07, "loss": 0.8341, "step": 10693 }, { "epoch": 0.91, "learning_rate": 1.9929178819540095e-07, "loss": 0.8055, "step": 10694 }, { "epoch": 0.91, "learning_rate": 1.9890571358957332e-07, "loss": 0.8131, "step": 10695 }, { "epoch": 0.91, "learning_rate": 1.9852000571528142e-07, "loss": 0.7589, "step": 10696 }, { "epoch": 0.91, "learning_rate": 1.9813466460198827e-07, "loss": 0.7841, "step": 10697 }, { "epoch": 0.91, "learning_rate": 1.9774969027912705e-07, "loss": 0.7902, "step": 10698 }, { "epoch": 0.91, "learning_rate": 1.9736508277610532e-07, "loss": 0.8026, "step": 10699 }, { "epoch": 0.91, "learning_rate": 1.9698084212230128e-07, "loss": 0.7792, "step": 10700 }, { "epoch": 0.91, "learning_rate": 1.9659696834706366e-07, "loss": 0.813, "step": 10701 }, { "epoch": 0.91, "learning_rate": 1.9621346147971676e-07, "loss": 0.8205, "step": 10702 }, { "epoch": 0.91, "learning_rate": 1.9583032154955273e-07, "loss": 0.7774, "step": 10703 }, { "epoch": 0.91, "learning_rate": 1.954475485858387e-07, "loss": 0.7859, "step": 10704 }, { "epoch": 0.91, "learning_rate": 1.9506514261781296e-07, "loss": 0.8266, "step": 10705 }, { "epoch": 0.91, "learning_rate": 1.946831036746849e-07, "loss": 0.8357, "step": 10706 }, { "epoch": 0.91, "learning_rate": 1.9430143178563731e-07, "loss": 0.8033, "step": 10707 }, { "epoch": 0.91, "learning_rate": 1.939201269798241e-07, "loss": 0.7964, "step": 10708 }, { "epoch": 0.91, "learning_rate": 1.9353918928637082e-07, "loss": 0.7875, "step": 10709 }, { "epoch": 0.91, "learning_rate": 1.931586187343748e-07, "loss": 0.8083, "step": 10710 }, { "epoch": 0.91, "learning_rate": 1.9277841535290776e-07, "loss": 0.8192, "step": 10711 }, { "epoch": 0.91, "learning_rate": 1.9239857917101035e-07, "loss": 0.7799, "step": 10712 }, { "epoch": 0.91, "learning_rate": 1.9201911021769548e-07, "loss": 0.7677, "step": 10713 }, { "epoch": 0.91, "learning_rate": 1.9164000852195053e-07, "loss": 0.7883, "step": 10714 }, { "epoch": 0.91, "learning_rate": 1.9126127411273288e-07, "loss": 0.8009, "step": 10715 }, { "epoch": 0.91, "learning_rate": 1.9088290701897162e-07, "loss": 0.7571, "step": 10716 }, { "epoch": 0.91, "learning_rate": 1.905049072695686e-07, "loss": 0.8084, "step": 10717 }, { "epoch": 0.91, "learning_rate": 1.9012727489339633e-07, "loss": 0.7727, "step": 10718 }, { "epoch": 0.91, "learning_rate": 1.8975000991930171e-07, "loss": 0.8135, "step": 10719 }, { "epoch": 0.91, "learning_rate": 1.8937311237610168e-07, "loss": 0.7265, "step": 10720 }, { "epoch": 0.91, "learning_rate": 1.8899658229258543e-07, "loss": 0.7932, "step": 10721 }, { "epoch": 0.91, "learning_rate": 1.8862041969751444e-07, "loss": 0.4877, "step": 10722 }, { "epoch": 0.92, "learning_rate": 1.8824462461962068e-07, "loss": 0.7527, "step": 10723 }, { "epoch": 0.92, "learning_rate": 1.8786919708761066e-07, "loss": 0.7629, "step": 10724 }, { "epoch": 0.92, "learning_rate": 1.874941371301614e-07, "loss": 0.7985, "step": 10725 }, { "epoch": 0.92, "learning_rate": 1.8711944477592058e-07, "loss": 0.7533, "step": 10726 }, { "epoch": 0.92, "learning_rate": 1.867451200535103e-07, "loss": 0.7769, "step": 10727 }, { "epoch": 0.92, "learning_rate": 1.8637116299152324e-07, "loss": 0.7916, "step": 10728 }, { "epoch": 0.92, "learning_rate": 1.8599757361852377e-07, "loss": 0.8189, "step": 10729 }, { "epoch": 0.92, "learning_rate": 1.856243519630485e-07, "loss": 0.7552, "step": 10730 }, { "epoch": 0.92, "learning_rate": 1.8525149805360575e-07, "loss": 0.7983, "step": 10731 }, { "epoch": 0.92, "learning_rate": 1.8487901191867662e-07, "loss": 0.79, "step": 10732 }, { "epoch": 0.92, "learning_rate": 1.845068935867128e-07, "loss": 0.7703, "step": 10733 }, { "epoch": 0.92, "learning_rate": 1.8413514308613877e-07, "loss": 0.7807, "step": 10734 }, { "epoch": 0.92, "learning_rate": 1.8376376044535127e-07, "loss": 0.4935, "step": 10735 }, { "epoch": 0.92, "learning_rate": 1.8339274569271703e-07, "loss": 0.7934, "step": 10736 }, { "epoch": 0.92, "learning_rate": 1.8302209885657728e-07, "loss": 0.7617, "step": 10737 }, { "epoch": 0.92, "learning_rate": 1.8265181996524327e-07, "loss": 0.7803, "step": 10738 }, { "epoch": 0.92, "learning_rate": 1.822819090469985e-07, "loss": 0.7939, "step": 10739 }, { "epoch": 0.92, "learning_rate": 1.8191236613009976e-07, "loss": 0.5695, "step": 10740 }, { "epoch": 0.92, "learning_rate": 1.815431912427734e-07, "loss": 0.7747, "step": 10741 }, { "epoch": 0.92, "learning_rate": 1.8117438441321967e-07, "loss": 0.7811, "step": 10742 }, { "epoch": 0.92, "learning_rate": 1.808059456696093e-07, "loss": 0.7787, "step": 10743 }, { "epoch": 0.92, "learning_rate": 1.8043787504008482e-07, "loss": 0.771, "step": 10744 }, { "epoch": 0.92, "learning_rate": 1.800701725527626e-07, "loss": 0.806, "step": 10745 }, { "epoch": 0.92, "learning_rate": 1.7970283823572908e-07, "loss": 0.8023, "step": 10746 }, { "epoch": 0.92, "learning_rate": 1.793358721170435e-07, "loss": 0.7857, "step": 10747 }, { "epoch": 0.92, "learning_rate": 1.7896927422473564e-07, "loss": 0.8014, "step": 10748 }, { "epoch": 0.92, "learning_rate": 1.786030445868092e-07, "loss": 0.7833, "step": 10749 }, { "epoch": 0.92, "learning_rate": 1.7823718323123794e-07, "loss": 0.7629, "step": 10750 }, { "epoch": 0.92, "learning_rate": 1.7787169018596784e-07, "loss": 0.7613, "step": 10751 }, { "epoch": 0.92, "learning_rate": 1.7750656547891766e-07, "loss": 0.7821, "step": 10752 }, { "epoch": 0.92, "learning_rate": 1.771418091379773e-07, "loss": 0.8015, "step": 10753 }, { "epoch": 0.92, "learning_rate": 1.7677742119100838e-07, "loss": 0.788, "step": 10754 }, { "epoch": 0.92, "learning_rate": 1.764134016658464e-07, "loss": 0.7771, "step": 10755 }, { "epoch": 0.92, "learning_rate": 1.7604975059029418e-07, "loss": 0.7691, "step": 10756 }, { "epoch": 0.92, "learning_rate": 1.7568646799213107e-07, "loss": 0.8046, "step": 10757 }, { "epoch": 0.92, "learning_rate": 1.753235538991055e-07, "loss": 0.759, "step": 10758 }, { "epoch": 0.92, "learning_rate": 1.7496100833893915e-07, "loss": 0.7803, "step": 10759 }, { "epoch": 0.92, "learning_rate": 1.7459883133932543e-07, "loss": 0.8165, "step": 10760 }, { "epoch": 0.92, "learning_rate": 1.7423702292792833e-07, "loss": 0.7871, "step": 10761 }, { "epoch": 0.92, "learning_rate": 1.7387558313238517e-07, "loss": 0.7933, "step": 10762 }, { "epoch": 0.92, "learning_rate": 1.73514511980305e-07, "loss": 0.7715, "step": 10763 }, { "epoch": 0.92, "learning_rate": 1.731538094992663e-07, "loss": 0.7336, "step": 10764 }, { "epoch": 0.92, "learning_rate": 1.727934757168237e-07, "loss": 0.4919, "step": 10765 }, { "epoch": 0.92, "learning_rate": 1.724335106605002e-07, "loss": 0.8132, "step": 10766 }, { "epoch": 0.92, "learning_rate": 1.7207391435779097e-07, "loss": 0.7325, "step": 10767 }, { "epoch": 0.92, "learning_rate": 1.7171468683616578e-07, "loss": 0.8022, "step": 10768 }, { "epoch": 0.92, "learning_rate": 1.713558281230615e-07, "loss": 0.7651, "step": 10769 }, { "epoch": 0.92, "learning_rate": 1.7099733824589238e-07, "loss": 0.8203, "step": 10770 }, { "epoch": 0.92, "learning_rate": 1.706392172320398e-07, "loss": 0.8072, "step": 10771 }, { "epoch": 0.92, "learning_rate": 1.7028146510885912e-07, "loss": 0.7797, "step": 10772 }, { "epoch": 0.92, "learning_rate": 1.6992408190367739e-07, "loss": 0.7765, "step": 10773 }, { "epoch": 0.92, "learning_rate": 1.6956706764379438e-07, "loss": 0.7234, "step": 10774 }, { "epoch": 0.92, "learning_rate": 1.6921042235647888e-07, "loss": 0.8154, "step": 10775 }, { "epoch": 0.92, "learning_rate": 1.688541460689741e-07, "loss": 0.8092, "step": 10776 }, { "epoch": 0.92, "learning_rate": 1.6849823880849382e-07, "loss": 0.4985, "step": 10777 }, { "epoch": 0.92, "learning_rate": 1.6814270060222515e-07, "loss": 0.5852, "step": 10778 }, { "epoch": 0.92, "learning_rate": 1.6778753147732362e-07, "loss": 0.7875, "step": 10779 }, { "epoch": 0.92, "learning_rate": 1.6743273146092144e-07, "loss": 0.7295, "step": 10780 }, { "epoch": 0.92, "learning_rate": 1.6707830058011909e-07, "loss": 0.8143, "step": 10781 }, { "epoch": 0.92, "learning_rate": 1.6672423886198884e-07, "loss": 0.7529, "step": 10782 }, { "epoch": 0.92, "learning_rate": 1.663705463335763e-07, "loss": 0.7706, "step": 10783 }, { "epoch": 0.92, "learning_rate": 1.6601722302189814e-07, "loss": 0.8029, "step": 10784 }, { "epoch": 0.92, "learning_rate": 1.656642689539434e-07, "loss": 0.7991, "step": 10785 }, { "epoch": 0.92, "learning_rate": 1.653116841566721e-07, "loss": 0.7754, "step": 10786 }, { "epoch": 0.92, "learning_rate": 1.6495946865701605e-07, "loss": 0.8125, "step": 10787 }, { "epoch": 0.92, "learning_rate": 1.64607622481881e-07, "loss": 0.7645, "step": 10788 }, { "epoch": 0.92, "learning_rate": 1.6425614565813984e-07, "loss": 0.7838, "step": 10789 }, { "epoch": 0.92, "learning_rate": 1.6390503821264226e-07, "loss": 0.7458, "step": 10790 }, { "epoch": 0.92, "learning_rate": 1.635543001722073e-07, "loss": 0.7778, "step": 10791 }, { "epoch": 0.92, "learning_rate": 1.6320393156362525e-07, "loss": 0.8139, "step": 10792 }, { "epoch": 0.92, "learning_rate": 1.6285393241366021e-07, "loss": 0.7553, "step": 10793 }, { "epoch": 0.92, "learning_rate": 1.6250430274904637e-07, "loss": 0.7488, "step": 10794 }, { "epoch": 0.92, "learning_rate": 1.6215504259648963e-07, "loss": 0.825, "step": 10795 }, { "epoch": 0.92, "learning_rate": 1.6180615198266858e-07, "loss": 0.8105, "step": 10796 }, { "epoch": 0.92, "learning_rate": 1.6145763093423305e-07, "loss": 0.8209, "step": 10797 }, { "epoch": 0.92, "learning_rate": 1.611094794778051e-07, "loss": 0.7234, "step": 10798 }, { "epoch": 0.92, "learning_rate": 1.6076169763997895e-07, "loss": 0.7702, "step": 10799 }, { "epoch": 0.92, "learning_rate": 1.6041428544731786e-07, "loss": 0.7904, "step": 10800 }, { "epoch": 0.92, "learning_rate": 1.6006724292636166e-07, "loss": 0.7942, "step": 10801 }, { "epoch": 0.92, "learning_rate": 1.5972057010361696e-07, "loss": 0.7927, "step": 10802 }, { "epoch": 0.92, "learning_rate": 1.5937426700556535e-07, "loss": 0.756, "step": 10803 }, { "epoch": 0.92, "learning_rate": 1.59028333658659e-07, "loss": 0.7793, "step": 10804 }, { "epoch": 0.92, "learning_rate": 1.5868277008932175e-07, "loss": 0.8121, "step": 10805 }, { "epoch": 0.92, "learning_rate": 1.5833757632394976e-07, "loss": 0.8093, "step": 10806 }, { "epoch": 0.92, "learning_rate": 1.5799275238891132e-07, "loss": 0.7714, "step": 10807 }, { "epoch": 0.92, "learning_rate": 1.5764829831054484e-07, "loss": 0.8031, "step": 10808 }, { "epoch": 0.92, "learning_rate": 1.573042141151615e-07, "loss": 0.7971, "step": 10809 }, { "epoch": 0.92, "learning_rate": 1.569604998290436e-07, "loss": 0.7546, "step": 10810 }, { "epoch": 0.92, "learning_rate": 1.5661715547844737e-07, "loss": 0.7939, "step": 10811 }, { "epoch": 0.92, "learning_rate": 1.562741810895979e-07, "loss": 0.7575, "step": 10812 }, { "epoch": 0.92, "learning_rate": 1.559315766886943e-07, "loss": 0.7749, "step": 10813 }, { "epoch": 0.92, "learning_rate": 1.5558934230190615e-07, "loss": 0.7838, "step": 10814 }, { "epoch": 0.92, "learning_rate": 1.5524747795537365e-07, "loss": 0.8056, "step": 10815 }, { "epoch": 0.92, "learning_rate": 1.5490598367521204e-07, "loss": 0.8177, "step": 10816 }, { "epoch": 0.92, "learning_rate": 1.5456485948750433e-07, "loss": 0.7459, "step": 10817 }, { "epoch": 0.92, "learning_rate": 1.5422410541830913e-07, "loss": 0.8032, "step": 10818 }, { "epoch": 0.92, "learning_rate": 1.538837214936545e-07, "loss": 0.803, "step": 10819 }, { "epoch": 0.92, "learning_rate": 1.5354370773953963e-07, "loss": 0.745, "step": 10820 }, { "epoch": 0.92, "learning_rate": 1.5320406418193822e-07, "loss": 0.7978, "step": 10821 }, { "epoch": 0.92, "learning_rate": 1.5286479084679228e-07, "loss": 0.766, "step": 10822 }, { "epoch": 0.92, "learning_rate": 1.5252588776001774e-07, "loss": 0.7706, "step": 10823 }, { "epoch": 0.92, "learning_rate": 1.5218735494750225e-07, "loss": 0.7762, "step": 10824 }, { "epoch": 0.92, "learning_rate": 1.51849192435104e-07, "loss": 0.8125, "step": 10825 }, { "epoch": 0.92, "learning_rate": 1.5151140024865397e-07, "loss": 0.8353, "step": 10826 }, { "epoch": 0.92, "learning_rate": 1.5117397841395376e-07, "loss": 0.7799, "step": 10827 }, { "epoch": 0.92, "learning_rate": 1.508369269567783e-07, "loss": 0.8207, "step": 10828 }, { "epoch": 0.92, "learning_rate": 1.505002459028726e-07, "loss": 0.7709, "step": 10829 }, { "epoch": 0.92, "learning_rate": 1.5016393527795325e-07, "loss": 0.795, "step": 10830 }, { "epoch": 0.92, "learning_rate": 1.498279951077114e-07, "loss": 0.7914, "step": 10831 }, { "epoch": 0.92, "learning_rate": 1.4949242541780594e-07, "loss": 0.8182, "step": 10832 }, { "epoch": 0.92, "learning_rate": 1.491572262338703e-07, "loss": 0.7986, "step": 10833 }, { "epoch": 0.92, "learning_rate": 1.4882239758150897e-07, "loss": 0.7814, "step": 10834 }, { "epoch": 0.92, "learning_rate": 1.4848793948629648e-07, "loss": 0.7842, "step": 10835 }, { "epoch": 0.92, "learning_rate": 1.4815385197378185e-07, "loss": 0.7792, "step": 10836 }, { "epoch": 0.92, "learning_rate": 1.4782013506948356e-07, "loss": 0.7716, "step": 10837 }, { "epoch": 0.92, "learning_rate": 1.4748678879889233e-07, "loss": 0.8102, "step": 10838 }, { "epoch": 0.92, "learning_rate": 1.4715381318747225e-07, "loss": 0.7408, "step": 10839 }, { "epoch": 0.92, "learning_rate": 1.4682120826065626e-07, "loss": 0.7862, "step": 10840 }, { "epoch": 0.93, "learning_rate": 1.4648897404385076e-07, "loss": 0.7672, "step": 10841 }, { "epoch": 0.93, "learning_rate": 1.461571105624332e-07, "loss": 0.7846, "step": 10842 }, { "epoch": 0.93, "learning_rate": 1.458256178417533e-07, "loss": 0.7809, "step": 10843 }, { "epoch": 0.93, "learning_rate": 1.4549449590713193e-07, "loss": 0.8636, "step": 10844 }, { "epoch": 0.93, "learning_rate": 1.4516374478386275e-07, "loss": 0.8285, "step": 10845 }, { "epoch": 0.93, "learning_rate": 1.448333644972083e-07, "loss": 0.8141, "step": 10846 }, { "epoch": 0.93, "learning_rate": 1.4450335507240676e-07, "loss": 0.7613, "step": 10847 }, { "epoch": 0.93, "learning_rate": 1.4417371653466462e-07, "loss": 0.7809, "step": 10848 }, { "epoch": 0.93, "learning_rate": 1.4384444890916171e-07, "loss": 0.7581, "step": 10849 }, { "epoch": 0.93, "learning_rate": 1.4351555222104795e-07, "loss": 0.7879, "step": 10850 }, { "epoch": 0.93, "learning_rate": 1.4318702649544824e-07, "loss": 0.8099, "step": 10851 }, { "epoch": 0.93, "learning_rate": 1.4285887175745638e-07, "loss": 0.7633, "step": 10852 }, { "epoch": 0.93, "learning_rate": 1.4253108803213677e-07, "loss": 0.8065, "step": 10853 }, { "epoch": 0.93, "learning_rate": 1.4220367534452994e-07, "loss": 0.7619, "step": 10854 }, { "epoch": 0.93, "learning_rate": 1.418766337196431e-07, "loss": 0.8239, "step": 10855 }, { "epoch": 0.93, "learning_rate": 1.4154996318245796e-07, "loss": 0.7872, "step": 10856 }, { "epoch": 0.93, "learning_rate": 1.4122366375792785e-07, "loss": 0.8071, "step": 10857 }, { "epoch": 0.93, "learning_rate": 1.4089773547097564e-07, "loss": 0.8277, "step": 10858 }, { "epoch": 0.93, "learning_rate": 1.4057217834649916e-07, "loss": 0.7566, "step": 10859 }, { "epoch": 0.93, "learning_rate": 1.4024699240936467e-07, "loss": 0.7529, "step": 10860 }, { "epoch": 0.93, "learning_rate": 1.3992217768441285e-07, "loss": 0.7886, "step": 10861 }, { "epoch": 0.93, "learning_rate": 1.3959773419645383e-07, "loss": 0.7382, "step": 10862 }, { "epoch": 0.93, "learning_rate": 1.3927366197026949e-07, "loss": 0.7942, "step": 10863 }, { "epoch": 0.93, "learning_rate": 1.3894996103061553e-07, "loss": 0.8206, "step": 10864 }, { "epoch": 0.93, "learning_rate": 1.3862663140221722e-07, "loss": 0.7672, "step": 10865 }, { "epoch": 0.93, "learning_rate": 1.3830367310977144e-07, "loss": 0.7919, "step": 10866 }, { "epoch": 0.93, "learning_rate": 1.379810861779496e-07, "loss": 0.8245, "step": 10867 }, { "epoch": 0.93, "learning_rate": 1.3765887063138917e-07, "loss": 0.8139, "step": 10868 }, { "epoch": 0.93, "learning_rate": 1.3733702649470493e-07, "loss": 0.7735, "step": 10869 }, { "epoch": 0.93, "learning_rate": 1.3701555379248054e-07, "loss": 0.7791, "step": 10870 }, { "epoch": 0.93, "learning_rate": 1.3669445254927084e-07, "loss": 0.7617, "step": 10871 }, { "epoch": 0.93, "learning_rate": 1.3637372278960448e-07, "loss": 0.8062, "step": 10872 }, { "epoch": 0.93, "learning_rate": 1.360533645379797e-07, "loss": 0.4722, "step": 10873 }, { "epoch": 0.93, "learning_rate": 1.3573337781886686e-07, "loss": 0.7754, "step": 10874 }, { "epoch": 0.93, "learning_rate": 1.354137626567087e-07, "loss": 0.7522, "step": 10875 }, { "epoch": 0.93, "learning_rate": 1.3509451907591785e-07, "loss": 0.843, "step": 10876 }, { "epoch": 0.93, "learning_rate": 1.3477564710088097e-07, "loss": 0.7727, "step": 10877 }, { "epoch": 0.93, "learning_rate": 1.3445714675595467e-07, "loss": 0.7246, "step": 10878 }, { "epoch": 0.93, "learning_rate": 1.3413901806546725e-07, "loss": 0.8385, "step": 10879 }, { "epoch": 0.93, "learning_rate": 1.3382126105371985e-07, "loss": 0.788, "step": 10880 }, { "epoch": 0.93, "learning_rate": 1.3350387574498358e-07, "loss": 0.7882, "step": 10881 }, { "epoch": 0.93, "learning_rate": 1.3318686216350241e-07, "loss": 0.7619, "step": 10882 }, { "epoch": 0.93, "learning_rate": 1.3287022033349085e-07, "loss": 0.7895, "step": 10883 }, { "epoch": 0.93, "learning_rate": 1.3255395027913565e-07, "loss": 0.8176, "step": 10884 }, { "epoch": 0.93, "learning_rate": 1.3223805202459527e-07, "loss": 0.7968, "step": 10885 }, { "epoch": 0.93, "learning_rate": 1.3192252559399986e-07, "loss": 0.7867, "step": 10886 }, { "epoch": 0.93, "learning_rate": 1.3160737101145127e-07, "loss": 0.8085, "step": 10887 }, { "epoch": 0.93, "learning_rate": 1.3129258830102132e-07, "loss": 0.7883, "step": 10888 }, { "epoch": 0.93, "learning_rate": 1.3097817748675524e-07, "loss": 0.7594, "step": 10889 }, { "epoch": 0.93, "learning_rate": 1.3066413859266991e-07, "loss": 0.7734, "step": 10890 }, { "epoch": 0.93, "learning_rate": 1.3035047164275172e-07, "loss": 0.8206, "step": 10891 }, { "epoch": 0.93, "learning_rate": 1.3003717666096206e-07, "loss": 0.566, "step": 10892 }, { "epoch": 0.93, "learning_rate": 1.2972425367123065e-07, "loss": 0.7849, "step": 10893 }, { "epoch": 0.93, "learning_rate": 1.294117026974606e-07, "loss": 0.6275, "step": 10894 }, { "epoch": 0.93, "learning_rate": 1.290995237635262e-07, "loss": 0.8156, "step": 10895 }, { "epoch": 0.93, "learning_rate": 1.287877168932722e-07, "loss": 0.7722, "step": 10896 }, { "epoch": 0.93, "learning_rate": 1.2847628211051732e-07, "loss": 0.8106, "step": 10897 }, { "epoch": 0.93, "learning_rate": 1.281652194390498e-07, "loss": 0.8395, "step": 10898 }, { "epoch": 0.93, "learning_rate": 1.2785452890263006e-07, "loss": 0.8259, "step": 10899 }, { "epoch": 0.93, "learning_rate": 1.2754421052499133e-07, "loss": 0.7819, "step": 10900 }, { "epoch": 0.93, "learning_rate": 1.272342643298352e-07, "loss": 0.7643, "step": 10901 }, { "epoch": 0.93, "learning_rate": 1.2692469034083888e-07, "loss": 0.8002, "step": 10902 }, { "epoch": 0.93, "learning_rate": 1.2661548858164842e-07, "loss": 0.8032, "step": 10903 }, { "epoch": 0.93, "learning_rate": 1.2630665907588102e-07, "loss": 0.7322, "step": 10904 }, { "epoch": 0.93, "learning_rate": 1.2599820184712896e-07, "loss": 0.8251, "step": 10905 }, { "epoch": 0.93, "learning_rate": 1.2569011691895227e-07, "loss": 0.7795, "step": 10906 }, { "epoch": 0.93, "learning_rate": 1.2538240431488435e-07, "loss": 0.8371, "step": 10907 }, { "epoch": 0.93, "learning_rate": 1.2507506405842973e-07, "loss": 0.5271, "step": 10908 }, { "epoch": 0.93, "learning_rate": 1.2476809617306408e-07, "loss": 0.7575, "step": 10909 }, { "epoch": 0.93, "learning_rate": 1.2446150068223584e-07, "loss": 0.7967, "step": 10910 }, { "epoch": 0.93, "learning_rate": 1.2415527760936462e-07, "loss": 0.7683, "step": 10911 }, { "epoch": 0.93, "learning_rate": 1.2384942697784008e-07, "loss": 0.7671, "step": 10912 }, { "epoch": 0.93, "learning_rate": 1.2354394881102572e-07, "loss": 0.8393, "step": 10913 }, { "epoch": 0.93, "learning_rate": 1.2323884313225564e-07, "loss": 0.7729, "step": 10914 }, { "epoch": 0.93, "learning_rate": 1.22934109964834e-07, "loss": 0.7766, "step": 10915 }, { "epoch": 0.93, "learning_rate": 1.2262974933203943e-07, "loss": 0.7501, "step": 10916 }, { "epoch": 0.93, "learning_rate": 1.2232576125711882e-07, "loss": 0.7619, "step": 10917 }, { "epoch": 0.93, "learning_rate": 1.2202214576329363e-07, "loss": 0.7879, "step": 10918 }, { "epoch": 0.93, "learning_rate": 1.217189028737542e-07, "loss": 0.7538, "step": 10919 }, { "epoch": 0.93, "learning_rate": 1.214160326116659e-07, "loss": 0.798, "step": 10920 }, { "epoch": 0.93, "learning_rate": 1.2111353500016188e-07, "loss": 0.8084, "step": 10921 }, { "epoch": 0.93, "learning_rate": 1.208114100623481e-07, "loss": 0.8505, "step": 10922 }, { "epoch": 0.93, "learning_rate": 1.2050965782130386e-07, "loss": 0.7694, "step": 10923 }, { "epoch": 0.93, "learning_rate": 1.2020827830007687e-07, "loss": 0.8466, "step": 10924 }, { "epoch": 0.93, "learning_rate": 1.1990727152168923e-07, "loss": 0.8097, "step": 10925 }, { "epoch": 0.93, "learning_rate": 1.196066375091326e-07, "loss": 0.7869, "step": 10926 }, { "epoch": 0.93, "learning_rate": 1.1930637628537134e-07, "loss": 0.7993, "step": 10927 }, { "epoch": 0.93, "learning_rate": 1.1900648787334102e-07, "loss": 0.7533, "step": 10928 }, { "epoch": 0.93, "learning_rate": 1.1870697229594718e-07, "loss": 0.7755, "step": 10929 }, { "epoch": 0.93, "learning_rate": 1.1840782957607044e-07, "loss": 0.8123, "step": 10930 }, { "epoch": 0.93, "learning_rate": 1.1810905973656028e-07, "loss": 0.7429, "step": 10931 }, { "epoch": 0.93, "learning_rate": 1.1781066280023679e-07, "loss": 0.7875, "step": 10932 }, { "epoch": 0.93, "learning_rate": 1.175126387898945e-07, "loss": 0.5716, "step": 10933 }, { "epoch": 0.93, "learning_rate": 1.1721498772829743e-07, "loss": 0.8299, "step": 10934 }, { "epoch": 0.93, "learning_rate": 1.1691770963818183e-07, "loss": 0.801, "step": 10935 }, { "epoch": 0.93, "learning_rate": 1.166208045422551e-07, "loss": 0.8193, "step": 10936 }, { "epoch": 0.93, "learning_rate": 1.1632427246319577e-07, "loss": 0.7884, "step": 10937 }, { "epoch": 0.93, "learning_rate": 1.1602811342365572e-07, "loss": 0.7743, "step": 10938 }, { "epoch": 0.93, "learning_rate": 1.1573232744625685e-07, "loss": 0.7747, "step": 10939 }, { "epoch": 0.93, "learning_rate": 1.1543691455359163e-07, "loss": 0.7771, "step": 10940 }, { "epoch": 0.93, "learning_rate": 1.1514187476822591e-07, "loss": 0.7948, "step": 10941 }, { "epoch": 0.93, "learning_rate": 1.1484720811269612e-07, "loss": 0.8266, "step": 10942 }, { "epoch": 0.93, "learning_rate": 1.145529146095109e-07, "loss": 0.8098, "step": 10943 }, { "epoch": 0.93, "learning_rate": 1.142589942811495e-07, "loss": 0.8009, "step": 10944 }, { "epoch": 0.93, "learning_rate": 1.139654471500623e-07, "loss": 0.7478, "step": 10945 }, { "epoch": 0.93, "learning_rate": 1.1367227323867302e-07, "loss": 0.7661, "step": 10946 }, { "epoch": 0.93, "learning_rate": 1.1337947256937599e-07, "loss": 0.8019, "step": 10947 }, { "epoch": 0.93, "learning_rate": 1.130870451645355e-07, "loss": 0.7913, "step": 10948 }, { "epoch": 0.93, "learning_rate": 1.1279499104648928e-07, "loss": 0.7986, "step": 10949 }, { "epoch": 0.93, "learning_rate": 1.1250331023754501e-07, "loss": 0.81, "step": 10950 }, { "epoch": 0.93, "learning_rate": 1.1221200275998434e-07, "loss": 0.7802, "step": 10951 }, { "epoch": 0.93, "learning_rate": 1.1192106863605779e-07, "loss": 0.802, "step": 10952 }, { "epoch": 0.93, "learning_rate": 1.1163050788798868e-07, "loss": 0.8102, "step": 10953 }, { "epoch": 0.93, "learning_rate": 1.1134032053797151e-07, "loss": 0.7885, "step": 10954 }, { "epoch": 0.93, "learning_rate": 1.1105050660817185e-07, "loss": 0.7426, "step": 10955 }, { "epoch": 0.93, "learning_rate": 1.1076106612072756e-07, "loss": 0.7721, "step": 10956 }, { "epoch": 0.93, "learning_rate": 1.1047199909774652e-07, "loss": 0.8149, "step": 10957 }, { "epoch": 0.94, "learning_rate": 1.1018330556131052e-07, "loss": 0.8025, "step": 10958 }, { "epoch": 0.94, "learning_rate": 1.0989498553347133e-07, "loss": 0.7973, "step": 10959 }, { "epoch": 0.94, "learning_rate": 1.0960703903625136e-07, "loss": 0.7599, "step": 10960 }, { "epoch": 0.94, "learning_rate": 1.0931946609164579e-07, "loss": 0.8061, "step": 10961 }, { "epoch": 0.94, "learning_rate": 1.0903226672162092e-07, "loss": 0.7452, "step": 10962 }, { "epoch": 0.94, "learning_rate": 1.0874544094811424e-07, "loss": 0.7872, "step": 10963 }, { "epoch": 0.94, "learning_rate": 1.0845898879303596e-07, "loss": 0.8261, "step": 10964 }, { "epoch": 0.94, "learning_rate": 1.0817291027826471e-07, "loss": 0.8126, "step": 10965 }, { "epoch": 0.94, "learning_rate": 1.0788720542565467e-07, "loss": 0.8046, "step": 10966 }, { "epoch": 0.94, "learning_rate": 1.076018742570284e-07, "loss": 0.8053, "step": 10967 }, { "epoch": 0.94, "learning_rate": 1.0731691679418121e-07, "loss": 0.7779, "step": 10968 }, { "epoch": 0.94, "learning_rate": 1.0703233305887905e-07, "loss": 0.7704, "step": 10969 }, { "epoch": 0.94, "learning_rate": 1.0674812307286009e-07, "loss": 0.7931, "step": 10970 }, { "epoch": 0.94, "learning_rate": 1.0646428685783416e-07, "loss": 0.8259, "step": 10971 }, { "epoch": 0.94, "learning_rate": 1.0618082443548173e-07, "loss": 0.7524, "step": 10972 }, { "epoch": 0.94, "learning_rate": 1.0589773582745544e-07, "loss": 0.7555, "step": 10973 }, { "epoch": 0.94, "learning_rate": 1.0561502105537801e-07, "loss": 0.7579, "step": 10974 }, { "epoch": 0.94, "learning_rate": 1.0533268014084552e-07, "loss": 0.7913, "step": 10975 }, { "epoch": 0.94, "learning_rate": 1.0505071310542403e-07, "loss": 0.7663, "step": 10976 }, { "epoch": 0.94, "learning_rate": 1.0476911997065243e-07, "loss": 0.7955, "step": 10977 }, { "epoch": 0.94, "learning_rate": 1.0448790075803906e-07, "loss": 0.7927, "step": 10978 }, { "epoch": 0.94, "learning_rate": 1.0420705548906563e-07, "loss": 0.79, "step": 10979 }, { "epoch": 0.94, "learning_rate": 1.0392658418518497e-07, "loss": 0.8039, "step": 10980 }, { "epoch": 0.94, "learning_rate": 1.0364648686781997e-07, "loss": 0.5273, "step": 10981 }, { "epoch": 0.94, "learning_rate": 1.0336676355836572e-07, "loss": 0.7831, "step": 10982 }, { "epoch": 0.94, "learning_rate": 1.0308741427818903e-07, "loss": 0.7899, "step": 10983 }, { "epoch": 0.94, "learning_rate": 1.0280843904862891e-07, "loss": 0.7628, "step": 10984 }, { "epoch": 0.94, "learning_rate": 1.0252983789099391e-07, "loss": 0.7738, "step": 10985 }, { "epoch": 0.94, "learning_rate": 1.0225161082656476e-07, "loss": 0.7839, "step": 10986 }, { "epoch": 0.94, "learning_rate": 1.0197375787659558e-07, "loss": 0.8359, "step": 10987 }, { "epoch": 0.94, "learning_rate": 1.0169627906230828e-07, "loss": 0.8341, "step": 10988 }, { "epoch": 0.94, "learning_rate": 1.0141917440489868e-07, "loss": 0.7846, "step": 10989 }, { "epoch": 0.94, "learning_rate": 1.0114244392553318e-07, "loss": 0.782, "step": 10990 }, { "epoch": 0.94, "learning_rate": 1.0086608764535044e-07, "loss": 0.7943, "step": 10991 }, { "epoch": 0.94, "learning_rate": 1.0059010558545967e-07, "loss": 0.786, "step": 10992 }, { "epoch": 0.94, "learning_rate": 1.003144977669418e-07, "loss": 0.7476, "step": 10993 }, { "epoch": 0.94, "learning_rate": 1.0003926421084886e-07, "loss": 0.7985, "step": 10994 }, { "epoch": 0.94, "learning_rate": 9.976440493820461e-08, "loss": 0.7957, "step": 10995 }, { "epoch": 0.94, "learning_rate": 9.948991997000446e-08, "loss": 0.808, "step": 10996 }, { "epoch": 0.94, "learning_rate": 9.921580932721497e-08, "loss": 0.7973, "step": 10997 }, { "epoch": 0.94, "learning_rate": 9.89420730307733e-08, "loss": 0.767, "step": 10998 }, { "epoch": 0.94, "learning_rate": 9.866871110158938e-08, "loss": 0.8128, "step": 10999 }, { "epoch": 0.94, "learning_rate": 9.839572356054484e-08, "loss": 0.8326, "step": 11000 }, { "epoch": 0.94, "learning_rate": 9.812311042849021e-08, "loss": 0.7958, "step": 11001 }, { "epoch": 0.94, "learning_rate": 9.785087172624997e-08, "loss": 0.5317, "step": 11002 }, { "epoch": 0.94, "learning_rate": 9.757900747461857e-08, "loss": 0.7915, "step": 11003 }, { "epoch": 0.94, "learning_rate": 9.73075176943633e-08, "loss": 0.7875, "step": 11004 }, { "epoch": 0.94, "learning_rate": 9.703640240622092e-08, "loss": 0.7777, "step": 11005 }, { "epoch": 0.94, "learning_rate": 9.67656616309004e-08, "loss": 0.5267, "step": 11006 }, { "epoch": 0.94, "learning_rate": 9.649529538908408e-08, "loss": 0.8088, "step": 11007 }, { "epoch": 0.94, "learning_rate": 9.622530370142158e-08, "loss": 0.766, "step": 11008 }, { "epoch": 0.94, "learning_rate": 9.595568658853749e-08, "loss": 0.7815, "step": 11009 }, { "epoch": 0.94, "learning_rate": 9.568644407102645e-08, "loss": 0.7774, "step": 11010 }, { "epoch": 0.94, "learning_rate": 9.541757616945369e-08, "loss": 0.7636, "step": 11011 }, { "epoch": 0.94, "learning_rate": 9.514908290435776e-08, "loss": 0.8131, "step": 11012 }, { "epoch": 0.94, "learning_rate": 9.488096429624727e-08, "loss": 0.8012, "step": 11013 }, { "epoch": 0.94, "learning_rate": 9.46132203656025e-08, "loss": 0.7766, "step": 11014 }, { "epoch": 0.94, "learning_rate": 9.434585113287432e-08, "loss": 0.7728, "step": 11015 }, { "epoch": 0.94, "learning_rate": 9.407885661848581e-08, "loss": 0.8335, "step": 11016 }, { "epoch": 0.94, "learning_rate": 9.381223684283291e-08, "loss": 0.7752, "step": 11017 }, { "epoch": 0.94, "learning_rate": 9.354599182627932e-08, "loss": 0.7659, "step": 11018 }, { "epoch": 0.94, "learning_rate": 9.328012158916322e-08, "loss": 0.8093, "step": 11019 }, { "epoch": 0.94, "learning_rate": 9.301462615179391e-08, "loss": 0.7535, "step": 11020 }, { "epoch": 0.94, "learning_rate": 9.274950553444905e-08, "loss": 0.8053, "step": 11021 }, { "epoch": 0.94, "learning_rate": 9.248475975738192e-08, "loss": 0.8556, "step": 11022 }, { "epoch": 0.94, "learning_rate": 9.222038884081408e-08, "loss": 0.768, "step": 11023 }, { "epoch": 0.94, "learning_rate": 9.195639280493939e-08, "loss": 0.7895, "step": 11024 }, { "epoch": 0.94, "learning_rate": 9.169277166992451e-08, "loss": 0.8381, "step": 11025 }, { "epoch": 0.94, "learning_rate": 9.142952545590445e-08, "loss": 0.8054, "step": 11026 }, { "epoch": 0.94, "learning_rate": 9.116665418298864e-08, "loss": 0.7673, "step": 11027 }, { "epoch": 0.94, "learning_rate": 9.09041578712555e-08, "loss": 0.7507, "step": 11028 }, { "epoch": 0.94, "learning_rate": 9.064203654075565e-08, "loss": 0.4723, "step": 11029 }, { "epoch": 0.94, "learning_rate": 9.038029021151196e-08, "loss": 0.8096, "step": 11030 }, { "epoch": 0.94, "learning_rate": 9.011891890351788e-08, "loss": 0.753, "step": 11031 }, { "epoch": 0.94, "learning_rate": 8.985792263673798e-08, "loss": 0.8203, "step": 11032 }, { "epoch": 0.94, "learning_rate": 8.959730143110912e-08, "loss": 0.7262, "step": 11033 }, { "epoch": 0.94, "learning_rate": 8.933705530653813e-08, "loss": 0.7883, "step": 11034 }, { "epoch": 0.94, "learning_rate": 8.907718428290357e-08, "loss": 0.7756, "step": 11035 }, { "epoch": 0.94, "learning_rate": 8.881768838005622e-08, "loss": 0.8181, "step": 11036 }, { "epoch": 0.94, "learning_rate": 8.855856761781801e-08, "loss": 0.7804, "step": 11037 }, { "epoch": 0.94, "learning_rate": 8.829982201598141e-08, "loss": 0.8033, "step": 11038 }, { "epoch": 0.94, "learning_rate": 8.80414515943101e-08, "loss": 0.7906, "step": 11039 }, { "epoch": 0.94, "learning_rate": 8.778345637254159e-08, "loss": 0.8142, "step": 11040 }, { "epoch": 0.94, "learning_rate": 8.752583637038071e-08, "loss": 0.7958, "step": 11041 }, { "epoch": 0.94, "learning_rate": 8.726859160750722e-08, "loss": 0.7916, "step": 11042 }, { "epoch": 0.94, "learning_rate": 8.701172210357045e-08, "loss": 0.7924, "step": 11043 }, { "epoch": 0.94, "learning_rate": 8.675522787819023e-08, "loss": 0.7736, "step": 11044 }, { "epoch": 0.94, "learning_rate": 8.649910895096036e-08, "loss": 0.7834, "step": 11045 }, { "epoch": 0.94, "learning_rate": 8.624336534144406e-08, "loss": 0.8033, "step": 11046 }, { "epoch": 0.94, "learning_rate": 8.598799706917572e-08, "loss": 0.8166, "step": 11047 }, { "epoch": 0.94, "learning_rate": 8.573300415366248e-08, "loss": 0.823, "step": 11048 }, { "epoch": 0.94, "learning_rate": 8.547838661438157e-08, "loss": 0.4944, "step": 11049 }, { "epoch": 0.94, "learning_rate": 8.522414447078187e-08, "loss": 0.8273, "step": 11050 }, { "epoch": 0.94, "learning_rate": 8.497027774228339e-08, "loss": 0.7328, "step": 11051 }, { "epoch": 0.94, "learning_rate": 8.47167864482784e-08, "loss": 0.8013, "step": 11052 }, { "epoch": 0.94, "learning_rate": 8.446367060812977e-08, "loss": 0.7801, "step": 11053 }, { "epoch": 0.94, "learning_rate": 8.421093024117088e-08, "loss": 0.502, "step": 11054 }, { "epoch": 0.94, "learning_rate": 8.395856536670854e-08, "loss": 0.8262, "step": 11055 }, { "epoch": 0.94, "learning_rate": 8.370657600401899e-08, "loss": 0.7764, "step": 11056 }, { "epoch": 0.94, "learning_rate": 8.34549621723496e-08, "loss": 0.7941, "step": 11057 }, { "epoch": 0.94, "learning_rate": 8.320372389092168e-08, "loss": 0.7887, "step": 11058 }, { "epoch": 0.94, "learning_rate": 8.295286117892432e-08, "loss": 0.749, "step": 11059 }, { "epoch": 0.94, "learning_rate": 8.270237405552162e-08, "loss": 0.8136, "step": 11060 }, { "epoch": 0.94, "learning_rate": 8.24522625398455e-08, "loss": 0.7957, "step": 11061 }, { "epoch": 0.94, "learning_rate": 8.22025266510007e-08, "loss": 0.7625, "step": 11062 }, { "epoch": 0.94, "learning_rate": 8.195316640806417e-08, "loss": 0.8013, "step": 11063 }, { "epoch": 0.94, "learning_rate": 8.170418183008233e-08, "loss": 0.807, "step": 11064 }, { "epoch": 0.94, "learning_rate": 8.145557293607442e-08, "loss": 0.8159, "step": 11065 }, { "epoch": 0.94, "learning_rate": 8.120733974503137e-08, "loss": 0.7832, "step": 11066 }, { "epoch": 0.94, "learning_rate": 8.095948227591244e-08, "loss": 0.8049, "step": 11067 }, { "epoch": 0.94, "learning_rate": 8.071200054765138e-08, "loss": 0.7461, "step": 11068 }, { "epoch": 0.94, "learning_rate": 8.046489457915196e-08, "loss": 0.7755, "step": 11069 }, { "epoch": 0.94, "learning_rate": 8.021816438928964e-08, "loss": 0.793, "step": 11070 }, { "epoch": 0.94, "learning_rate": 7.997180999691101e-08, "loss": 0.7794, "step": 11071 }, { "epoch": 0.94, "learning_rate": 7.972583142083268e-08, "loss": 0.7436, "step": 11072 }, { "epoch": 0.94, "learning_rate": 7.948022867984517e-08, "loss": 0.7674, "step": 11073 }, { "epoch": 0.94, "learning_rate": 7.923500179270738e-08, "loss": 0.8183, "step": 11074 }, { "epoch": 0.95, "learning_rate": 7.899015077815265e-08, "loss": 0.8126, "step": 11075 }, { "epoch": 0.95, "learning_rate": 7.874567565488267e-08, "loss": 0.7709, "step": 11076 }, { "epoch": 0.95, "learning_rate": 7.850157644157141e-08, "loss": 0.8084, "step": 11077 }, { "epoch": 0.95, "learning_rate": 7.825785315686563e-08, "loss": 0.4924, "step": 11078 }, { "epoch": 0.95, "learning_rate": 7.801450581938152e-08, "loss": 0.8039, "step": 11079 }, { "epoch": 0.95, "learning_rate": 7.777153444770757e-08, "loss": 0.795, "step": 11080 }, { "epoch": 0.95, "learning_rate": 7.752893906040226e-08, "loss": 0.7555, "step": 11081 }, { "epoch": 0.95, "learning_rate": 7.728671967599632e-08, "loss": 0.7652, "step": 11082 }, { "epoch": 0.95, "learning_rate": 7.704487631299273e-08, "loss": 0.8043, "step": 11083 }, { "epoch": 0.95, "learning_rate": 7.680340898986338e-08, "loss": 0.7497, "step": 11084 }, { "epoch": 0.95, "learning_rate": 7.656231772505351e-08, "loss": 0.7531, "step": 11085 }, { "epoch": 0.95, "learning_rate": 7.632160253697951e-08, "loss": 0.7203, "step": 11086 }, { "epoch": 0.95, "learning_rate": 7.608126344402722e-08, "loss": 0.7785, "step": 11087 }, { "epoch": 0.95, "learning_rate": 7.584130046455529e-08, "loss": 0.7645, "step": 11088 }, { "epoch": 0.95, "learning_rate": 7.56017136168935e-08, "loss": 0.8193, "step": 11089 }, { "epoch": 0.95, "learning_rate": 7.536250291934222e-08, "loss": 0.7395, "step": 11090 }, { "epoch": 0.95, "learning_rate": 7.512366839017405e-08, "loss": 0.4801, "step": 11091 }, { "epoch": 0.95, "learning_rate": 7.488521004763216e-08, "loss": 0.7812, "step": 11092 }, { "epoch": 0.95, "learning_rate": 7.4647127909932e-08, "loss": 0.7686, "step": 11093 }, { "epoch": 0.95, "learning_rate": 7.440942199525791e-08, "loss": 0.7698, "step": 11094 }, { "epoch": 0.95, "learning_rate": 7.417209232176814e-08, "loss": 0.7037, "step": 11095 }, { "epoch": 0.95, "learning_rate": 7.393513890759041e-08, "loss": 0.8034, "step": 11096 }, { "epoch": 0.95, "learning_rate": 7.369856177082523e-08, "loss": 0.8234, "step": 11097 }, { "epoch": 0.95, "learning_rate": 7.346236092954318e-08, "loss": 0.8085, "step": 11098 }, { "epoch": 0.95, "learning_rate": 7.322653640178645e-08, "loss": 0.7401, "step": 11099 }, { "epoch": 0.95, "learning_rate": 7.299108820556843e-08, "loss": 0.7921, "step": 11100 }, { "epoch": 0.95, "learning_rate": 7.275601635887364e-08, "loss": 0.7925, "step": 11101 }, { "epoch": 0.95, "learning_rate": 7.252132087965824e-08, "loss": 0.7867, "step": 11102 }, { "epoch": 0.95, "learning_rate": 7.228700178584958e-08, "loss": 0.813, "step": 11103 }, { "epoch": 0.95, "learning_rate": 7.20530590953461e-08, "loss": 0.7544, "step": 11104 }, { "epoch": 0.95, "learning_rate": 7.181949282601686e-08, "loss": 0.8001, "step": 11105 }, { "epoch": 0.95, "learning_rate": 7.158630299570424e-08, "loss": 0.5196, "step": 11106 }, { "epoch": 0.95, "learning_rate": 7.1353489622219e-08, "loss": 0.819, "step": 11107 }, { "epoch": 0.95, "learning_rate": 7.112105272334523e-08, "loss": 0.7886, "step": 11108 }, { "epoch": 0.95, "learning_rate": 7.088899231683765e-08, "loss": 0.7412, "step": 11109 }, { "epoch": 0.95, "learning_rate": 7.065730842042206e-08, "loss": 0.8035, "step": 11110 }, { "epoch": 0.95, "learning_rate": 7.042600105179597e-08, "loss": 0.7367, "step": 11111 }, { "epoch": 0.95, "learning_rate": 7.019507022862693e-08, "loss": 0.7759, "step": 11112 }, { "epoch": 0.95, "learning_rate": 6.99645159685558e-08, "loss": 0.8156, "step": 11113 }, { "epoch": 0.95, "learning_rate": 6.97343382891924e-08, "loss": 0.7536, "step": 11114 }, { "epoch": 0.95, "learning_rate": 6.950453720811879e-08, "loss": 0.8309, "step": 11115 }, { "epoch": 0.95, "learning_rate": 6.927511274288923e-08, "loss": 0.7308, "step": 11116 }, { "epoch": 0.95, "learning_rate": 6.904606491102805e-08, "loss": 0.731, "step": 11117 }, { "epoch": 0.95, "learning_rate": 6.881739373003071e-08, "loss": 0.8055, "step": 11118 }, { "epoch": 0.95, "learning_rate": 6.85890992173649e-08, "loss": 0.8023, "step": 11119 }, { "epoch": 0.95, "learning_rate": 6.836118139046777e-08, "loss": 0.7635, "step": 11120 }, { "epoch": 0.95, "learning_rate": 6.81336402667493e-08, "loss": 0.8029, "step": 11121 }, { "epoch": 0.95, "learning_rate": 6.790647586359112e-08, "loss": 0.8066, "step": 11122 }, { "epoch": 0.95, "learning_rate": 6.76796881983438e-08, "loss": 0.7877, "step": 11123 }, { "epoch": 0.95, "learning_rate": 6.745327728833129e-08, "loss": 0.8226, "step": 11124 }, { "epoch": 0.95, "learning_rate": 6.722724315084805e-08, "loss": 0.8062, "step": 11125 }, { "epoch": 0.95, "learning_rate": 6.700158580315975e-08, "loss": 0.764, "step": 11126 }, { "epoch": 0.95, "learning_rate": 6.677630526250256e-08, "loss": 0.7708, "step": 11127 }, { "epoch": 0.95, "learning_rate": 6.655140154608497e-08, "loss": 0.7993, "step": 11128 }, { "epoch": 0.95, "learning_rate": 6.632687467108656e-08, "loss": 0.769, "step": 11129 }, { "epoch": 0.95, "learning_rate": 6.61027246546575e-08, "loss": 0.7629, "step": 11130 }, { "epoch": 0.95, "learning_rate": 6.587895151391854e-08, "loss": 0.793, "step": 11131 }, { "epoch": 0.95, "learning_rate": 6.565555526596434e-08, "loss": 0.8103, "step": 11132 }, { "epoch": 0.95, "learning_rate": 6.543253592785848e-08, "loss": 0.7206, "step": 11133 }, { "epoch": 0.95, "learning_rate": 6.520989351663564e-08, "loss": 0.8327, "step": 11134 }, { "epoch": 0.95, "learning_rate": 6.498762804930225e-08, "loss": 0.8138, "step": 11135 }, { "epoch": 0.95, "learning_rate": 6.476573954283694e-08, "loss": 0.7911, "step": 11136 }, { "epoch": 0.95, "learning_rate": 6.454422801418836e-08, "loss": 0.7638, "step": 11137 }, { "epoch": 0.95, "learning_rate": 6.432309348027632e-08, "loss": 0.7719, "step": 11138 }, { "epoch": 0.95, "learning_rate": 6.410233595799287e-08, "loss": 0.8255, "step": 11139 }, { "epoch": 0.95, "learning_rate": 6.38819554641995e-08, "loss": 0.7467, "step": 11140 }, { "epoch": 0.95, "learning_rate": 6.366195201573056e-08, "loss": 0.7856, "step": 11141 }, { "epoch": 0.95, "learning_rate": 6.344232562939145e-08, "loss": 0.4997, "step": 11142 }, { "epoch": 0.95, "learning_rate": 6.322307632195768e-08, "loss": 0.7521, "step": 11143 }, { "epoch": 0.95, "learning_rate": 6.300420411017694e-08, "loss": 0.7637, "step": 11144 }, { "epoch": 0.95, "learning_rate": 6.278570901076809e-08, "loss": 0.7786, "step": 11145 }, { "epoch": 0.95, "learning_rate": 6.256759104041998e-08, "loss": 0.8041, "step": 11146 }, { "epoch": 0.95, "learning_rate": 6.234985021579432e-08, "loss": 0.7564, "step": 11147 }, { "epoch": 0.95, "learning_rate": 6.213248655352222e-08, "loss": 0.8279, "step": 11148 }, { "epoch": 0.95, "learning_rate": 6.191550007020874e-08, "loss": 0.8112, "step": 11149 }, { "epoch": 0.95, "learning_rate": 6.169889078242674e-08, "loss": 0.8128, "step": 11150 }, { "epoch": 0.95, "learning_rate": 6.148265870672243e-08, "loss": 0.7998, "step": 11151 }, { "epoch": 0.95, "learning_rate": 6.12668038596137e-08, "loss": 0.8007, "step": 11152 }, { "epoch": 0.95, "learning_rate": 6.105132625758736e-08, "loss": 0.7884, "step": 11153 }, { "epoch": 0.95, "learning_rate": 6.083622591710303e-08, "loss": 0.7609, "step": 11154 }, { "epoch": 0.95, "learning_rate": 6.06215028545909e-08, "loss": 0.82, "step": 11155 }, { "epoch": 0.95, "learning_rate": 6.040715708645283e-08, "loss": 0.7493, "step": 11156 }, { "epoch": 0.95, "learning_rate": 6.019318862906188e-08, "loss": 0.8034, "step": 11157 }, { "epoch": 0.95, "learning_rate": 5.997959749876215e-08, "loss": 0.8135, "step": 11158 }, { "epoch": 0.95, "learning_rate": 5.976638371186783e-08, "loss": 0.7944, "step": 11159 }, { "epoch": 0.95, "learning_rate": 5.955354728466645e-08, "loss": 0.7702, "step": 11160 }, { "epoch": 0.95, "learning_rate": 5.9341088233414444e-08, "loss": 0.7699, "step": 11161 }, { "epoch": 0.95, "learning_rate": 5.912900657434106e-08, "loss": 0.7774, "step": 11162 }, { "epoch": 0.95, "learning_rate": 5.891730232364612e-08, "loss": 0.8392, "step": 11163 }, { "epoch": 0.95, "learning_rate": 5.8705975497500565e-08, "loss": 0.5155, "step": 11164 }, { "epoch": 0.95, "learning_rate": 5.8495026112046495e-08, "loss": 0.8114, "step": 11165 }, { "epoch": 0.95, "learning_rate": 5.828445418339823e-08, "loss": 0.7629, "step": 11166 }, { "epoch": 0.95, "learning_rate": 5.807425972763903e-08, "loss": 0.7669, "step": 11167 }, { "epoch": 0.95, "learning_rate": 5.786444276082492e-08, "loss": 0.7756, "step": 11168 }, { "epoch": 0.95, "learning_rate": 5.7655003298982547e-08, "loss": 0.7677, "step": 11169 }, { "epoch": 0.95, "learning_rate": 5.744594135811077e-08, "loss": 0.8026, "step": 11170 }, { "epoch": 0.95, "learning_rate": 5.7237256954177924e-08, "loss": 0.8044, "step": 11171 }, { "epoch": 0.95, "learning_rate": 5.7028950103125703e-08, "loss": 0.7918, "step": 11172 }, { "epoch": 0.95, "learning_rate": 5.68210208208636e-08, "loss": 0.7844, "step": 11173 }, { "epoch": 0.95, "learning_rate": 5.661346912327614e-08, "loss": 0.79, "step": 11174 }, { "epoch": 0.95, "learning_rate": 5.640629502621619e-08, "loss": 0.7368, "step": 11175 }, { "epoch": 0.95, "learning_rate": 5.619949854550888e-08, "loss": 0.8224, "step": 11176 }, { "epoch": 0.95, "learning_rate": 5.599307969695045e-08, "loss": 0.7594, "step": 11177 }, { "epoch": 0.95, "learning_rate": 5.578703849630773e-08, "loss": 0.7944, "step": 11178 }, { "epoch": 0.95, "learning_rate": 5.5581374959320366e-08, "loss": 0.8042, "step": 11179 }, { "epoch": 0.95, "learning_rate": 5.53760891016969e-08, "loss": 0.8288, "step": 11180 }, { "epoch": 0.95, "learning_rate": 5.517118093911811e-08, "loss": 0.7822, "step": 11181 }, { "epoch": 0.95, "learning_rate": 5.4966650487236485e-08, "loss": 0.7701, "step": 11182 }, { "epoch": 0.95, "learning_rate": 5.4762497761674526e-08, "loss": 0.7856, "step": 11183 }, { "epoch": 0.95, "learning_rate": 5.455872277802698e-08, "loss": 0.7881, "step": 11184 }, { "epoch": 0.95, "learning_rate": 5.435532555185863e-08, "loss": 0.8164, "step": 11185 }, { "epoch": 0.95, "learning_rate": 5.415230609870703e-08, "loss": 0.7832, "step": 11186 }, { "epoch": 0.95, "learning_rate": 5.3949664434078116e-08, "loss": 0.7946, "step": 11187 }, { "epoch": 0.95, "learning_rate": 5.374740057345229e-08, "loss": 0.7985, "step": 11188 }, { "epoch": 0.95, "learning_rate": 5.3545514532278296e-08, "loss": 0.7938, "step": 11189 }, { "epoch": 0.95, "learning_rate": 5.334400632597769e-08, "loss": 0.8338, "step": 11190 }, { "epoch": 0.95, "learning_rate": 5.314287596994316e-08, "loss": 0.7889, "step": 11191 }, { "epoch": 0.96, "learning_rate": 5.294212347953742e-08, "loss": 0.7916, "step": 11192 }, { "epoch": 0.96, "learning_rate": 5.2741748870094864e-08, "loss": 0.8236, "step": 11193 }, { "epoch": 0.96, "learning_rate": 5.2541752156921035e-08, "loss": 0.7433, "step": 11194 }, { "epoch": 0.96, "learning_rate": 5.23421333552937e-08, "loss": 0.7572, "step": 11195 }, { "epoch": 0.96, "learning_rate": 5.214289248045956e-08, "loss": 0.7949, "step": 11196 }, { "epoch": 0.96, "learning_rate": 5.1944029547638105e-08, "loss": 0.756, "step": 11197 }, { "epoch": 0.96, "learning_rate": 5.1745544572019414e-08, "loss": 0.7739, "step": 11198 }, { "epoch": 0.96, "learning_rate": 5.154743756876468e-08, "loss": 0.8075, "step": 11199 }, { "epoch": 0.96, "learning_rate": 5.134970855300681e-08, "loss": 0.7827, "step": 11200 }, { "epoch": 0.96, "learning_rate": 5.1152357539848705e-08, "loss": 0.7893, "step": 11201 }, { "epoch": 0.96, "learning_rate": 5.095538454436555e-08, "loss": 0.7516, "step": 11202 }, { "epoch": 0.96, "learning_rate": 5.075878958160252e-08, "loss": 0.7717, "step": 11203 }, { "epoch": 0.96, "learning_rate": 5.056257266657649e-08, "loss": 0.8079, "step": 11204 }, { "epoch": 0.96, "learning_rate": 5.036673381427659e-08, "loss": 0.7992, "step": 11205 }, { "epoch": 0.96, "learning_rate": 5.017127303966085e-08, "loss": 0.8327, "step": 11206 }, { "epoch": 0.96, "learning_rate": 4.9976190357660103e-08, "loss": 0.8033, "step": 11207 }, { "epoch": 0.96, "learning_rate": 4.978148578317521e-08, "loss": 0.7922, "step": 11208 }, { "epoch": 0.96, "learning_rate": 4.9587159331078716e-08, "loss": 0.832, "step": 11209 }, { "epoch": 0.96, "learning_rate": 4.9393211016215416e-08, "loss": 0.7849, "step": 11210 }, { "epoch": 0.96, "learning_rate": 4.9199640853398454e-08, "loss": 0.7823, "step": 11211 }, { "epoch": 0.96, "learning_rate": 4.9006448857414904e-08, "loss": 0.7935, "step": 11212 }, { "epoch": 0.96, "learning_rate": 4.8813635043020726e-08, "loss": 0.8107, "step": 11213 }, { "epoch": 0.96, "learning_rate": 4.86211994249447e-08, "loss": 0.8042, "step": 11214 }, { "epoch": 0.96, "learning_rate": 4.842914201788562e-08, "loss": 0.7252, "step": 11215 }, { "epoch": 0.96, "learning_rate": 4.823746283651454e-08, "loss": 0.8354, "step": 11216 }, { "epoch": 0.96, "learning_rate": 4.8046161895471956e-08, "loss": 0.7505, "step": 11217 }, { "epoch": 0.96, "learning_rate": 4.785523920937063e-08, "loss": 0.7944, "step": 11218 }, { "epoch": 0.96, "learning_rate": 4.7664694792794455e-08, "loss": 0.7815, "step": 11219 }, { "epoch": 0.96, "learning_rate": 4.747452866029789e-08, "loss": 0.7791, "step": 11220 }, { "epoch": 0.96, "learning_rate": 4.7284740826407105e-08, "loss": 0.8151, "step": 11221 }, { "epoch": 0.96, "learning_rate": 4.709533130561883e-08, "loss": 0.7596, "step": 11222 }, { "epoch": 0.96, "learning_rate": 4.690630011240094e-08, "loss": 0.7694, "step": 11223 }, { "epoch": 0.96, "learning_rate": 4.671764726119299e-08, "loss": 0.8098, "step": 11224 }, { "epoch": 0.96, "learning_rate": 4.6529372766405124e-08, "loss": 0.8179, "step": 11225 }, { "epoch": 0.96, "learning_rate": 4.634147664241862e-08, "loss": 0.759, "step": 11226 }, { "epoch": 0.96, "learning_rate": 4.615395890358532e-08, "loss": 0.7835, "step": 11227 }, { "epoch": 0.96, "learning_rate": 4.596681956422933e-08, "loss": 0.7587, "step": 11228 }, { "epoch": 0.96, "learning_rate": 4.578005863864588e-08, "loss": 0.8009, "step": 11229 }, { "epoch": 0.96, "learning_rate": 4.5593676141099106e-08, "loss": 0.7909, "step": 11230 }, { "epoch": 0.96, "learning_rate": 4.540767208582764e-08, "loss": 0.465, "step": 11231 }, { "epoch": 0.96, "learning_rate": 4.522204648703843e-08, "loss": 0.781, "step": 11232 }, { "epoch": 0.96, "learning_rate": 4.50367993589107e-08, "loss": 0.7737, "step": 11233 }, { "epoch": 0.96, "learning_rate": 4.485193071559424e-08, "loss": 0.8209, "step": 11234 }, { "epoch": 0.96, "learning_rate": 4.4667440571209976e-08, "loss": 0.7432, "step": 11235 }, { "epoch": 0.96, "learning_rate": 4.448332893985163e-08, "loss": 0.7448, "step": 11236 }, { "epoch": 0.96, "learning_rate": 4.4299595835580744e-08, "loss": 0.7933, "step": 11237 }, { "epoch": 0.96, "learning_rate": 4.41162412724333e-08, "loss": 0.8071, "step": 11238 }, { "epoch": 0.96, "learning_rate": 4.3933265264414214e-08, "loss": 0.7824, "step": 11239 }, { "epoch": 0.96, "learning_rate": 4.375066782550008e-08, "loss": 0.8151, "step": 11240 }, { "epoch": 0.96, "learning_rate": 4.3568448969638075e-08, "loss": 0.7438, "step": 11241 }, { "epoch": 0.96, "learning_rate": 4.3386608710747627e-08, "loss": 0.7377, "step": 11242 }, { "epoch": 0.96, "learning_rate": 4.320514706271872e-08, "loss": 0.771, "step": 11243 }, { "epoch": 0.96, "learning_rate": 4.302406403941195e-08, "loss": 0.8268, "step": 11244 }, { "epoch": 0.96, "learning_rate": 4.2843359654659576e-08, "loss": 0.7653, "step": 11245 }, { "epoch": 0.96, "learning_rate": 4.266303392226501e-08, "loss": 0.7643, "step": 11246 }, { "epoch": 0.96, "learning_rate": 4.248308685600111e-08, "loss": 0.5567, "step": 11247 }, { "epoch": 0.96, "learning_rate": 4.230351846961467e-08, "loss": 0.7686, "step": 11248 }, { "epoch": 0.96, "learning_rate": 4.2124328776820824e-08, "loss": 0.786, "step": 11249 }, { "epoch": 0.96, "learning_rate": 4.194551779130751e-08, "loss": 0.7667, "step": 11250 }, { "epoch": 0.96, "learning_rate": 4.176708552673381e-08, "loss": 0.7735, "step": 11251 }, { "epoch": 0.96, "learning_rate": 4.1589031996728255e-08, "loss": 0.8253, "step": 11252 }, { "epoch": 0.96, "learning_rate": 4.14113572148922e-08, "loss": 0.7729, "step": 11253 }, { "epoch": 0.96, "learning_rate": 4.1234061194797004e-08, "loss": 0.7926, "step": 11254 }, { "epoch": 0.96, "learning_rate": 4.105714394998461e-08, "loss": 0.79, "step": 11255 }, { "epoch": 0.96, "learning_rate": 4.088060549397033e-08, "loss": 0.8176, "step": 11256 }, { "epoch": 0.96, "learning_rate": 4.070444584023836e-08, "loss": 0.7876, "step": 11257 }, { "epoch": 0.96, "learning_rate": 4.052866500224462e-08, "loss": 0.7381, "step": 11258 }, { "epoch": 0.96, "learning_rate": 4.0353262993416686e-08, "loss": 0.5313, "step": 11259 }, { "epoch": 0.96, "learning_rate": 4.0178239827151077e-08, "loss": 0.7003, "step": 11260 }, { "epoch": 0.96, "learning_rate": 4.000359551681876e-08, "loss": 0.713, "step": 11261 }, { "epoch": 0.96, "learning_rate": 3.982933007575907e-08, "loss": 0.7584, "step": 11262 }, { "epoch": 0.96, "learning_rate": 3.9655443517283007e-08, "loss": 0.7876, "step": 11263 }, { "epoch": 0.96, "learning_rate": 3.94819358546733e-08, "loss": 0.8127, "step": 11264 }, { "epoch": 0.96, "learning_rate": 3.930880710118379e-08, "loss": 0.7897, "step": 11265 }, { "epoch": 0.96, "learning_rate": 3.913605727003833e-08, "loss": 0.7856, "step": 11266 }, { "epoch": 0.96, "learning_rate": 3.896368637443193e-08, "loss": 0.7929, "step": 11267 }, { "epoch": 0.96, "learning_rate": 3.8791694427531836e-08, "loss": 0.7961, "step": 11268 }, { "epoch": 0.96, "learning_rate": 3.8620081442475864e-08, "loss": 0.7796, "step": 11269 }, { "epoch": 0.96, "learning_rate": 3.8448847432372425e-08, "loss": 0.7823, "step": 11270 }, { "epoch": 0.96, "learning_rate": 3.82779924103005e-08, "loss": 0.5514, "step": 11271 }, { "epoch": 0.96, "learning_rate": 3.810751638931243e-08, "loss": 0.7984, "step": 11272 }, { "epoch": 0.96, "learning_rate": 3.7937419382428343e-08, "loss": 0.809, "step": 11273 }, { "epoch": 0.96, "learning_rate": 3.77677014026423e-08, "loss": 0.7635, "step": 11274 }, { "epoch": 0.96, "learning_rate": 3.759836246291671e-08, "loss": 0.7688, "step": 11275 }, { "epoch": 0.96, "learning_rate": 3.742940257618843e-08, "loss": 0.8299, "step": 11276 }, { "epoch": 0.96, "learning_rate": 3.726082175536272e-08, "loss": 0.813, "step": 11277 }, { "epoch": 0.96, "learning_rate": 3.709262001331648e-08, "loss": 0.8031, "step": 11278 }, { "epoch": 0.96, "learning_rate": 3.6924797362897224e-08, "loss": 0.8368, "step": 11279 }, { "epoch": 0.96, "learning_rate": 3.675735381692469e-08, "loss": 0.7775, "step": 11280 }, { "epoch": 0.96, "learning_rate": 3.6590289388189206e-08, "loss": 0.7997, "step": 11281 }, { "epoch": 0.96, "learning_rate": 3.6423604089452227e-08, "loss": 0.8466, "step": 11282 }, { "epoch": 0.96, "learning_rate": 3.6257297933445234e-08, "loss": 0.8062, "step": 11283 }, { "epoch": 0.96, "learning_rate": 3.6091370932871406e-08, "loss": 0.7788, "step": 11284 }, { "epoch": 0.96, "learning_rate": 3.5925823100406154e-08, "loss": 0.7829, "step": 11285 }, { "epoch": 0.96, "learning_rate": 3.576065444869436e-08, "loss": 0.7781, "step": 11286 }, { "epoch": 0.96, "learning_rate": 3.559586499035206e-08, "loss": 0.7878, "step": 11287 }, { "epoch": 0.96, "learning_rate": 3.5431454737966385e-08, "loss": 0.7721, "step": 11288 }, { "epoch": 0.96, "learning_rate": 3.52674237040973e-08, "loss": 0.8001, "step": 11289 }, { "epoch": 0.96, "learning_rate": 3.510377190127312e-08, "loss": 0.8067, "step": 11290 }, { "epoch": 0.96, "learning_rate": 3.494049934199384e-08, "loss": 0.7607, "step": 11291 }, { "epoch": 0.96, "learning_rate": 3.4777606038732814e-08, "loss": 0.7945, "step": 11292 }, { "epoch": 0.96, "learning_rate": 3.4615092003931204e-08, "loss": 0.7843, "step": 11293 }, { "epoch": 0.96, "learning_rate": 3.445295725000353e-08, "loss": 0.7548, "step": 11294 }, { "epoch": 0.96, "learning_rate": 3.429120178933376e-08, "loss": 0.7978, "step": 11295 }, { "epoch": 0.96, "learning_rate": 3.412982563427758e-08, "loss": 0.4913, "step": 11296 }, { "epoch": 0.96, "learning_rate": 3.3968828797162343e-08, "loss": 0.7793, "step": 11297 }, { "epoch": 0.96, "learning_rate": 3.3808211290284886e-08, "loss": 0.835, "step": 11298 }, { "epoch": 0.96, "learning_rate": 3.364797312591539e-08, "loss": 0.768, "step": 11299 }, { "epoch": 0.96, "learning_rate": 3.34881143162924e-08, "loss": 0.7475, "step": 11300 }, { "epoch": 0.96, "learning_rate": 3.3328634873626694e-08, "loss": 0.7928, "step": 11301 }, { "epoch": 0.96, "learning_rate": 3.3169534810100764e-08, "loss": 0.7479, "step": 11302 }, { "epoch": 0.96, "learning_rate": 3.301081413786711e-08, "loss": 0.8421, "step": 11303 }, { "epoch": 0.96, "learning_rate": 3.2852472869049915e-08, "loss": 0.7726, "step": 11304 }, { "epoch": 0.96, "learning_rate": 3.269451101574395e-08, "loss": 0.4755, "step": 11305 }, { "epoch": 0.96, "learning_rate": 3.253692859001512e-08, "loss": 0.7776, "step": 11306 }, { "epoch": 0.96, "learning_rate": 3.23797256038999e-08, "loss": 0.7588, "step": 11307 }, { "epoch": 0.96, "learning_rate": 3.2222902069407036e-08, "loss": 0.8103, "step": 11308 }, { "epoch": 0.97, "learning_rate": 3.206645799851471e-08, "loss": 0.7766, "step": 11309 }, { "epoch": 0.97, "learning_rate": 3.1910393403173365e-08, "loss": 0.7847, "step": 11310 }, { "epoch": 0.97, "learning_rate": 3.175470829530403e-08, "loss": 0.8058, "step": 11311 }, { "epoch": 0.97, "learning_rate": 3.1599402686798844e-08, "loss": 0.763, "step": 11312 }, { "epoch": 0.97, "learning_rate": 3.144447658951999e-08, "loss": 0.7829, "step": 11313 }, { "epoch": 0.97, "learning_rate": 3.128993001530245e-08, "loss": 0.7976, "step": 11314 }, { "epoch": 0.97, "learning_rate": 3.1135762975950666e-08, "loss": 0.7775, "step": 11315 }, { "epoch": 0.97, "learning_rate": 3.098197548324078e-08, "loss": 0.8099, "step": 11316 }, { "epoch": 0.97, "learning_rate": 3.0828567548920054e-08, "loss": 0.7907, "step": 11317 }, { "epoch": 0.97, "learning_rate": 3.0675539184706894e-08, "loss": 0.8234, "step": 11318 }, { "epoch": 0.97, "learning_rate": 3.052289040228973e-08, "loss": 0.8103, "step": 11319 }, { "epoch": 0.97, "learning_rate": 3.037062121332868e-08, "loss": 0.798, "step": 11320 }, { "epoch": 0.97, "learning_rate": 3.021873162945499e-08, "loss": 0.8393, "step": 11321 }, { "epoch": 0.97, "learning_rate": 3.006722166227105e-08, "loss": 0.8078, "step": 11322 }, { "epoch": 0.97, "learning_rate": 2.9916091323349825e-08, "loss": 0.8046, "step": 11323 }, { "epoch": 0.97, "learning_rate": 2.9765340624234306e-08, "loss": 0.7976, "step": 11324 }, { "epoch": 0.97, "learning_rate": 2.9614969576441388e-08, "loss": 0.7936, "step": 11325 }, { "epoch": 0.97, "learning_rate": 2.9464978191455773e-08, "loss": 0.7678, "step": 11326 }, { "epoch": 0.97, "learning_rate": 2.931536648073552e-08, "loss": 0.7634, "step": 11327 }, { "epoch": 0.97, "learning_rate": 2.916613445570815e-08, "loss": 0.8245, "step": 11328 }, { "epoch": 0.97, "learning_rate": 2.9017282127772328e-08, "loss": 0.7721, "step": 11329 }, { "epoch": 0.97, "learning_rate": 2.8868809508298957e-08, "loss": 0.8144, "step": 11330 }, { "epoch": 0.97, "learning_rate": 2.872071660862896e-08, "loss": 0.8325, "step": 11331 }, { "epoch": 0.97, "learning_rate": 2.857300344007441e-08, "loss": 0.8263, "step": 11332 }, { "epoch": 0.97, "learning_rate": 2.842567001391794e-08, "loss": 0.7752, "step": 11333 }, { "epoch": 0.97, "learning_rate": 2.827871634141388e-08, "loss": 0.8047, "step": 11334 }, { "epoch": 0.97, "learning_rate": 2.81321424337877e-08, "loss": 0.8094, "step": 11335 }, { "epoch": 0.97, "learning_rate": 2.7985948302234888e-08, "loss": 0.7764, "step": 11336 }, { "epoch": 0.97, "learning_rate": 2.7840133957922066e-08, "loss": 0.851, "step": 11337 }, { "epoch": 0.97, "learning_rate": 2.7694699411989213e-08, "loss": 0.7969, "step": 11338 }, { "epoch": 0.97, "learning_rate": 2.7549644675542443e-08, "loss": 0.7817, "step": 11339 }, { "epoch": 0.97, "learning_rate": 2.7404969759664556e-08, "loss": 0.8003, "step": 11340 }, { "epoch": 0.97, "learning_rate": 2.7260674675404498e-08, "loss": 0.7795, "step": 11341 }, { "epoch": 0.97, "learning_rate": 2.711675943378511e-08, "loss": 0.7908, "step": 11342 }, { "epoch": 0.97, "learning_rate": 2.6973224045799827e-08, "loss": 0.4889, "step": 11343 }, { "epoch": 0.97, "learning_rate": 2.6830068522411545e-08, "loss": 0.8129, "step": 11344 }, { "epoch": 0.97, "learning_rate": 2.668729287455596e-08, "loss": 0.7584, "step": 11345 }, { "epoch": 0.97, "learning_rate": 2.654489711313879e-08, "loss": 0.7838, "step": 11346 }, { "epoch": 0.97, "learning_rate": 2.640288124903634e-08, "loss": 0.8004, "step": 11347 }, { "epoch": 0.97, "learning_rate": 2.6261245293097704e-08, "loss": 0.7622, "step": 11348 }, { "epoch": 0.97, "learning_rate": 2.6119989256140343e-08, "loss": 0.7629, "step": 11349 }, { "epoch": 0.97, "learning_rate": 2.5979113148955624e-08, "loss": 0.6207, "step": 11350 }, { "epoch": 0.97, "learning_rate": 2.5838616982303276e-08, "loss": 0.774, "step": 11351 }, { "epoch": 0.97, "learning_rate": 2.5698500766915267e-08, "loss": 0.8058, "step": 11352 }, { "epoch": 0.97, "learning_rate": 2.5558764513494704e-08, "loss": 0.7916, "step": 11353 }, { "epoch": 0.97, "learning_rate": 2.5419408232714714e-08, "loss": 0.7862, "step": 11354 }, { "epoch": 0.97, "learning_rate": 2.528043193522067e-08, "loss": 0.7753, "step": 11355 }, { "epoch": 0.97, "learning_rate": 2.5141835631627977e-08, "loss": 0.7535, "step": 11356 }, { "epoch": 0.97, "learning_rate": 2.500361933252371e-08, "loss": 0.7809, "step": 11357 }, { "epoch": 0.97, "learning_rate": 2.4865783048464988e-08, "loss": 0.737, "step": 11358 }, { "epoch": 0.97, "learning_rate": 2.4728326789980606e-08, "loss": 0.8221, "step": 11359 }, { "epoch": 0.97, "learning_rate": 2.4591250567570502e-08, "loss": 0.7635, "step": 11360 }, { "epoch": 0.97, "learning_rate": 2.4454554391704632e-08, "loss": 0.7966, "step": 11361 }, { "epoch": 0.97, "learning_rate": 2.4318238272824647e-08, "loss": 0.8259, "step": 11362 }, { "epoch": 0.97, "learning_rate": 2.4182302221343323e-08, "loss": 0.7816, "step": 11363 }, { "epoch": 0.97, "learning_rate": 2.4046746247644027e-08, "loss": 0.7941, "step": 11364 }, { "epoch": 0.97, "learning_rate": 2.3911570362081804e-08, "loss": 0.4981, "step": 11365 }, { "epoch": 0.97, "learning_rate": 2.3776774574981177e-08, "loss": 0.7608, "step": 11366 }, { "epoch": 0.97, "learning_rate": 2.3642358896638352e-08, "loss": 0.783, "step": 11367 }, { "epoch": 0.97, "learning_rate": 2.3508323337321225e-08, "loss": 0.8089, "step": 11368 }, { "epoch": 0.97, "learning_rate": 2.3374667907268278e-08, "loss": 0.8264, "step": 11369 }, { "epoch": 0.97, "learning_rate": 2.3241392616688563e-08, "loss": 0.7785, "step": 11370 }, { "epoch": 0.97, "learning_rate": 2.310849747576227e-08, "loss": 0.7771, "step": 11371 }, { "epoch": 0.97, "learning_rate": 2.2975982494640726e-08, "loss": 0.7877, "step": 11372 }, { "epoch": 0.97, "learning_rate": 2.284384768344583e-08, "loss": 0.7899, "step": 11373 }, { "epoch": 0.97, "learning_rate": 2.2712093052270623e-08, "loss": 0.791, "step": 11374 }, { "epoch": 0.97, "learning_rate": 2.2580718611179275e-08, "loss": 0.7762, "step": 11375 }, { "epoch": 0.97, "learning_rate": 2.2449724370207092e-08, "loss": 0.7604, "step": 11376 }, { "epoch": 0.97, "learning_rate": 2.2319110339359963e-08, "loss": 0.7316, "step": 11377 }, { "epoch": 0.97, "learning_rate": 2.21888765286149e-08, "loss": 0.8146, "step": 11378 }, { "epoch": 0.97, "learning_rate": 2.2059022947919507e-08, "loss": 0.7556, "step": 11379 }, { "epoch": 0.97, "learning_rate": 2.192954960719251e-08, "loss": 0.5448, "step": 11380 }, { "epoch": 0.97, "learning_rate": 2.180045651632434e-08, "loss": 0.8199, "step": 11381 }, { "epoch": 0.97, "learning_rate": 2.1671743685175438e-08, "loss": 0.7747, "step": 11382 }, { "epoch": 0.97, "learning_rate": 2.154341112357794e-08, "loss": 0.8037, "step": 11383 }, { "epoch": 0.97, "learning_rate": 2.1415458841334e-08, "loss": 0.7479, "step": 11384 }, { "epoch": 0.97, "learning_rate": 2.1287886848216922e-08, "loss": 0.7745, "step": 11385 }, { "epoch": 0.97, "learning_rate": 2.1160695153972232e-08, "loss": 0.7453, "step": 11386 }, { "epoch": 0.97, "learning_rate": 2.1033883768314944e-08, "loss": 0.799, "step": 11387 }, { "epoch": 0.97, "learning_rate": 2.0907452700931753e-08, "loss": 0.7479, "step": 11388 }, { "epoch": 0.97, "learning_rate": 2.078140196147993e-08, "loss": 0.7986, "step": 11389 }, { "epoch": 0.97, "learning_rate": 2.0655731559587334e-08, "loss": 0.7751, "step": 11390 }, { "epoch": 0.97, "learning_rate": 2.0530441504854613e-08, "loss": 0.7779, "step": 11391 }, { "epoch": 0.97, "learning_rate": 2.0405531806850786e-08, "loss": 0.7872, "step": 11392 }, { "epoch": 0.97, "learning_rate": 2.028100247511766e-08, "loss": 0.828, "step": 11393 }, { "epoch": 0.97, "learning_rate": 2.015685351916763e-08, "loss": 0.7376, "step": 11394 }, { "epoch": 0.97, "learning_rate": 2.0033084948483104e-08, "loss": 0.7872, "step": 11395 }, { "epoch": 0.97, "learning_rate": 1.990969677251875e-08, "loss": 0.778, "step": 11396 }, { "epoch": 0.97, "learning_rate": 1.9786689000699243e-08, "loss": 0.7771, "step": 11397 }, { "epoch": 0.97, "learning_rate": 1.9664061642420406e-08, "loss": 0.7716, "step": 11398 }, { "epoch": 0.97, "learning_rate": 1.9541814707049745e-08, "loss": 0.7774, "step": 11399 }, { "epoch": 0.97, "learning_rate": 1.9419948203924788e-08, "loss": 0.7659, "step": 11400 }, { "epoch": 0.97, "learning_rate": 1.92984621423542e-08, "loss": 0.7873, "step": 11401 }, { "epoch": 0.97, "learning_rate": 1.9177356531617784e-08, "loss": 0.7953, "step": 11402 }, { "epoch": 0.97, "learning_rate": 1.9056631380965917e-08, "loss": 0.7596, "step": 11403 }, { "epoch": 0.97, "learning_rate": 1.8936286699620664e-08, "loss": 0.8103, "step": 11404 }, { "epoch": 0.97, "learning_rate": 1.8816322496773565e-08, "loss": 0.7658, "step": 11405 }, { "epoch": 0.97, "learning_rate": 1.869673878158951e-08, "loss": 0.8041, "step": 11406 }, { "epoch": 0.97, "learning_rate": 1.8577535563202298e-08, "loss": 0.8167, "step": 11407 }, { "epoch": 0.97, "learning_rate": 1.8458712850716876e-08, "loss": 0.8107, "step": 11408 }, { "epoch": 0.97, "learning_rate": 1.834027065320987e-08, "loss": 0.8153, "step": 11409 }, { "epoch": 0.97, "learning_rate": 1.8222208979728483e-08, "loss": 0.7431, "step": 11410 }, { "epoch": 0.97, "learning_rate": 1.8104527839291063e-08, "loss": 0.8048, "step": 11411 }, { "epoch": 0.97, "learning_rate": 1.7987227240885973e-08, "loss": 0.7697, "step": 11412 }, { "epoch": 0.97, "learning_rate": 1.7870307193473822e-08, "loss": 0.772, "step": 11413 }, { "epoch": 0.97, "learning_rate": 1.775376770598525e-08, "loss": 0.8327, "step": 11414 }, { "epoch": 0.97, "learning_rate": 1.763760878732257e-08, "loss": 0.8305, "step": 11415 }, { "epoch": 0.97, "learning_rate": 1.7521830446358135e-08, "loss": 0.805, "step": 11416 }, { "epoch": 0.97, "learning_rate": 1.7406432691935427e-08, "loss": 0.7529, "step": 11417 }, { "epoch": 0.97, "learning_rate": 1.729141553287017e-08, "loss": 0.8096, "step": 11418 }, { "epoch": 0.97, "learning_rate": 1.7176778977947005e-08, "loss": 0.8247, "step": 11419 }, { "epoch": 0.97, "learning_rate": 1.7062523035922262e-08, "loss": 0.7725, "step": 11420 }, { "epoch": 0.97, "learning_rate": 1.6948647715524513e-08, "loss": 0.7929, "step": 11421 }, { "epoch": 0.97, "learning_rate": 1.6835153025451246e-08, "loss": 0.8164, "step": 11422 }, { "epoch": 0.97, "learning_rate": 1.6722038974371633e-08, "loss": 0.7594, "step": 11423 }, { "epoch": 0.97, "learning_rate": 1.660930557092655e-08, "loss": 0.7618, "step": 11424 }, { "epoch": 0.97, "learning_rate": 1.649695282372632e-08, "loss": 0.8267, "step": 11425 }, { "epoch": 0.97, "learning_rate": 1.6384980741354084e-08, "loss": 0.7831, "step": 11426 }, { "epoch": 0.98, "learning_rate": 1.627338933236189e-08, "loss": 0.7808, "step": 11427 }, { "epoch": 0.98, "learning_rate": 1.6162178605274028e-08, "loss": 0.7719, "step": 11428 }, { "epoch": 0.98, "learning_rate": 1.605134856858537e-08, "loss": 0.8097, "step": 11429 }, { "epoch": 0.98, "learning_rate": 1.5940899230761363e-08, "loss": 0.7586, "step": 11430 }, { "epoch": 0.98, "learning_rate": 1.5830830600239155e-08, "loss": 0.7573, "step": 11431 }, { "epoch": 0.98, "learning_rate": 1.5721142685425905e-08, "loss": 0.7806, "step": 11432 }, { "epoch": 0.98, "learning_rate": 1.5611835494699913e-08, "loss": 0.733, "step": 11433 }, { "epoch": 0.98, "learning_rate": 1.550290903641116e-08, "loss": 0.7645, "step": 11434 }, { "epoch": 0.98, "learning_rate": 1.5394363318880223e-08, "loss": 0.7705, "step": 11435 }, { "epoch": 0.98, "learning_rate": 1.5286198350397687e-08, "loss": 0.5404, "step": 11436 }, { "epoch": 0.98, "learning_rate": 1.5178414139225828e-08, "loss": 0.8274, "step": 11437 }, { "epoch": 0.98, "learning_rate": 1.5071010693598064e-08, "loss": 0.7121, "step": 11438 }, { "epoch": 0.98, "learning_rate": 1.4963988021718388e-08, "loss": 0.8067, "step": 11439 }, { "epoch": 0.98, "learning_rate": 1.4857346131761375e-08, "loss": 0.763, "step": 11440 }, { "epoch": 0.98, "learning_rate": 1.475108503187328e-08, "loss": 0.8134, "step": 11441 }, { "epoch": 0.98, "learning_rate": 1.4645204730170393e-08, "loss": 0.7844, "step": 11442 }, { "epoch": 0.98, "learning_rate": 1.4539705234741242e-08, "loss": 0.849, "step": 11443 }, { "epoch": 0.98, "learning_rate": 1.443458655364327e-08, "loss": 0.795, "step": 11444 }, { "epoch": 0.98, "learning_rate": 1.4329848694906723e-08, "loss": 0.8039, "step": 11445 }, { "epoch": 0.98, "learning_rate": 1.422549166653131e-08, "loss": 0.8033, "step": 11446 }, { "epoch": 0.98, "learning_rate": 1.4121515476489545e-08, "loss": 0.7782, "step": 11447 }, { "epoch": 0.98, "learning_rate": 1.4017920132722295e-08, "loss": 0.8165, "step": 11448 }, { "epoch": 0.98, "learning_rate": 1.3914705643143788e-08, "loss": 0.7783, "step": 11449 }, { "epoch": 0.98, "learning_rate": 1.3811872015637162e-08, "loss": 0.7566, "step": 11450 }, { "epoch": 0.98, "learning_rate": 1.3709419258057799e-08, "loss": 0.7977, "step": 11451 }, { "epoch": 0.98, "learning_rate": 1.3607347378232217e-08, "loss": 0.7973, "step": 11452 }, { "epoch": 0.98, "learning_rate": 1.3505656383955846e-08, "loss": 0.8015, "step": 11453 }, { "epoch": 0.98, "learning_rate": 1.3404346282996915e-08, "loss": 0.4915, "step": 11454 }, { "epoch": 0.98, "learning_rate": 1.3303417083094239e-08, "loss": 0.8064, "step": 11455 }, { "epoch": 0.98, "learning_rate": 1.3202868791957203e-08, "loss": 0.7706, "step": 11456 }, { "epoch": 0.98, "learning_rate": 1.3102701417266327e-08, "loss": 0.8175, "step": 11457 }, { "epoch": 0.98, "learning_rate": 1.300291496667272e-08, "loss": 0.7494, "step": 11458 }, { "epoch": 0.98, "learning_rate": 1.2903509447798058e-08, "loss": 0.759, "step": 11459 }, { "epoch": 0.98, "learning_rate": 1.2804484868236267e-08, "loss": 0.8039, "step": 11460 }, { "epoch": 0.98, "learning_rate": 1.2705841235550742e-08, "loss": 0.7899, "step": 11461 }, { "epoch": 0.98, "learning_rate": 1.2607578557277123e-08, "loss": 0.7936, "step": 11462 }, { "epoch": 0.98, "learning_rate": 1.250969684091996e-08, "loss": 0.8, "step": 11463 }, { "epoch": 0.98, "learning_rate": 1.2412196093957163e-08, "loss": 0.7693, "step": 11464 }, { "epoch": 0.98, "learning_rate": 1.2315076323836105e-08, "loss": 0.8058, "step": 11465 }, { "epoch": 0.98, "learning_rate": 1.221833753797419e-08, "loss": 0.7294, "step": 11466 }, { "epoch": 0.98, "learning_rate": 1.2121979743762168e-08, "loss": 0.7751, "step": 11467 }, { "epoch": 0.98, "learning_rate": 1.2026002948559711e-08, "loss": 0.802, "step": 11468 }, { "epoch": 0.98, "learning_rate": 1.1930407159698176e-08, "loss": 0.7841, "step": 11469 }, { "epoch": 0.98, "learning_rate": 1.18351923844795e-08, "loss": 0.8052, "step": 11470 }, { "epoch": 0.98, "learning_rate": 1.1740358630176751e-08, "loss": 0.7425, "step": 11471 }, { "epoch": 0.98, "learning_rate": 1.1645905904034139e-08, "loss": 0.8381, "step": 11472 }, { "epoch": 0.98, "learning_rate": 1.1551834213265333e-08, "loss": 0.7926, "step": 11473 }, { "epoch": 0.98, "learning_rate": 1.1458143565057367e-08, "loss": 0.7348, "step": 11474 }, { "epoch": 0.98, "learning_rate": 1.1364833966565626e-08, "loss": 0.7871, "step": 11475 }, { "epoch": 0.98, "learning_rate": 1.1271905424918294e-08, "loss": 0.7613, "step": 11476 }, { "epoch": 0.98, "learning_rate": 1.1179357947213587e-08, "loss": 0.8008, "step": 11477 }, { "epoch": 0.98, "learning_rate": 1.1087191540520847e-08, "loss": 0.8437, "step": 11478 }, { "epoch": 0.98, "learning_rate": 1.099540621188e-08, "loss": 0.7927, "step": 11479 }, { "epoch": 0.98, "learning_rate": 1.0904001968302103e-08, "loss": 0.8502, "step": 11480 }, { "epoch": 0.98, "learning_rate": 1.0812978816768793e-08, "loss": 0.7822, "step": 11481 }, { "epoch": 0.98, "learning_rate": 1.0722336764233398e-08, "loss": 0.8448, "step": 11482 }, { "epoch": 0.98, "learning_rate": 1.063207581761927e-08, "loss": 0.826, "step": 11483 }, { "epoch": 0.98, "learning_rate": 1.0542195983821446e-08, "loss": 0.7813, "step": 11484 }, { "epoch": 0.98, "learning_rate": 1.0452697269704992e-08, "loss": 0.7954, "step": 11485 }, { "epoch": 0.98, "learning_rate": 1.0363579682106106e-08, "loss": 0.7816, "step": 11486 }, { "epoch": 0.98, "learning_rate": 1.027484322783212e-08, "loss": 0.8045, "step": 11487 }, { "epoch": 0.98, "learning_rate": 1.0186487913660947e-08, "loss": 0.7724, "step": 11488 }, { "epoch": 0.98, "learning_rate": 1.0098513746342741e-08, "loss": 0.8174, "step": 11489 }, { "epoch": 0.98, "learning_rate": 1.0010920732596019e-08, "loss": 0.8068, "step": 11490 }, { "epoch": 0.98, "learning_rate": 9.923708879112647e-09, "loss": 0.7519, "step": 11491 }, { "epoch": 0.98, "learning_rate": 9.836878192553412e-09, "loss": 0.7855, "step": 11492 }, { "epoch": 0.98, "learning_rate": 9.750428679551338e-09, "loss": 0.7844, "step": 11493 }, { "epoch": 0.98, "learning_rate": 9.664360346710033e-09, "loss": 0.7969, "step": 11494 }, { "epoch": 0.98, "learning_rate": 9.578673200603682e-09, "loss": 0.7446, "step": 11495 }, { "epoch": 0.98, "learning_rate": 9.493367247777053e-09, "loss": 0.8062, "step": 11496 }, { "epoch": 0.98, "learning_rate": 9.40844249474715e-09, "loss": 0.76, "step": 11497 }, { "epoch": 0.98, "learning_rate": 9.323898947999898e-09, "loss": 0.5294, "step": 11498 }, { "epoch": 0.98, "learning_rate": 9.23973661399402e-09, "loss": 0.787, "step": 11499 }, { "epoch": 0.98, "learning_rate": 9.155955499157709e-09, "loss": 0.7857, "step": 11500 }, { "epoch": 0.98, "learning_rate": 9.072555609890844e-09, "loss": 0.7491, "step": 11501 }, { "epoch": 0.98, "learning_rate": 8.989536952563881e-09, "loss": 0.7755, "step": 11502 }, { "epoch": 0.98, "learning_rate": 8.906899533517866e-09, "loss": 0.764, "step": 11503 }, { "epoch": 0.98, "learning_rate": 8.824643359065522e-09, "loss": 0.7639, "step": 11504 }, { "epoch": 0.98, "learning_rate": 8.742768435490156e-09, "loss": 0.795, "step": 11505 }, { "epoch": 0.98, "learning_rate": 8.661274769045658e-09, "loss": 0.7979, "step": 11506 }, { "epoch": 0.98, "learning_rate": 8.580162365955935e-09, "loss": 0.81, "step": 11507 }, { "epoch": 0.98, "learning_rate": 8.499431232418254e-09, "loss": 0.7386, "step": 11508 }, { "epoch": 0.98, "learning_rate": 8.419081374598237e-09, "loss": 0.746, "step": 11509 }, { "epoch": 0.98, "learning_rate": 8.339112798633197e-09, "loss": 0.7726, "step": 11510 }, { "epoch": 0.98, "learning_rate": 8.259525510633248e-09, "loss": 0.7724, "step": 11511 }, { "epoch": 0.98, "learning_rate": 8.180319516675196e-09, "loss": 0.7414, "step": 11512 }, { "epoch": 0.98, "learning_rate": 8.101494822811418e-09, "loss": 0.8051, "step": 11513 }, { "epoch": 0.98, "learning_rate": 8.023051435061546e-09, "loss": 0.8466, "step": 11514 }, { "epoch": 0.98, "learning_rate": 7.944989359418009e-09, "loss": 0.8085, "step": 11515 }, { "epoch": 0.98, "learning_rate": 7.867308601843815e-09, "loss": 0.7865, "step": 11516 }, { "epoch": 0.98, "learning_rate": 7.790009168271994e-09, "loss": 0.8099, "step": 11517 }, { "epoch": 0.98, "learning_rate": 7.71309106460727e-09, "loss": 0.7658, "step": 11518 }, { "epoch": 0.98, "learning_rate": 7.63655429672494e-09, "loss": 0.7953, "step": 11519 }, { "epoch": 0.98, "learning_rate": 7.560398870471997e-09, "loss": 0.7544, "step": 11520 }, { "epoch": 0.98, "learning_rate": 7.48462479166434e-09, "loss": 0.8151, "step": 11521 }, { "epoch": 0.98, "learning_rate": 7.409232066091232e-09, "loss": 0.7795, "step": 11522 }, { "epoch": 0.98, "learning_rate": 7.3342206995108415e-09, "loss": 0.7685, "step": 11523 }, { "epoch": 0.98, "learning_rate": 7.259590697653029e-09, "loss": 0.5291, "step": 11524 }, { "epoch": 0.98, "learning_rate": 7.185342066218237e-09, "loss": 0.8384, "step": 11525 }, { "epoch": 0.98, "learning_rate": 7.111474810878039e-09, "loss": 0.7758, "step": 11526 }, { "epoch": 0.98, "learning_rate": 7.037988937275142e-09, "loss": 0.7616, "step": 11527 }, { "epoch": 0.98, "learning_rate": 6.9648844510222805e-09, "loss": 0.7697, "step": 11528 }, { "epoch": 0.98, "learning_rate": 6.892161357703875e-09, "loss": 0.8168, "step": 11529 }, { "epoch": 0.98, "learning_rate": 6.819819662874372e-09, "loss": 0.8076, "step": 11530 }, { "epoch": 0.98, "learning_rate": 6.747859372060461e-09, "loss": 0.7671, "step": 11531 }, { "epoch": 0.98, "learning_rate": 6.676280490758302e-09, "loss": 0.7201, "step": 11532 }, { "epoch": 0.98, "learning_rate": 6.6050830244351885e-09, "loss": 0.7764, "step": 11533 }, { "epoch": 0.98, "learning_rate": 6.534266978530102e-09, "loss": 0.7899, "step": 11534 }, { "epoch": 0.98, "learning_rate": 6.463832358451494e-09, "loss": 0.8021, "step": 11535 }, { "epoch": 0.98, "learning_rate": 6.3937791695806166e-09, "loss": 0.7951, "step": 11536 }, { "epoch": 0.98, "learning_rate": 6.324107417267633e-09, "loss": 0.775, "step": 11537 }, { "epoch": 0.98, "learning_rate": 6.254817106834954e-09, "loss": 0.8321, "step": 11538 }, { "epoch": 0.98, "learning_rate": 6.1859082435750115e-09, "loss": 0.7295, "step": 11539 }, { "epoch": 0.98, "learning_rate": 6.117380832751374e-09, "loss": 0.7884, "step": 11540 }, { "epoch": 0.98, "learning_rate": 6.049234879598742e-09, "loss": 0.8097, "step": 11541 }, { "epoch": 0.98, "learning_rate": 5.981470389322397e-09, "loss": 0.8162, "step": 11542 }, { "epoch": 0.98, "learning_rate": 5.914087367098198e-09, "loss": 0.7981, "step": 11543 }, { "epoch": 0.99, "learning_rate": 5.847085818073695e-09, "loss": 0.7508, "step": 11544 }, { "epoch": 0.99, "learning_rate": 5.7804657473664595e-09, "loss": 0.8552, "step": 11545 }, { "epoch": 0.99, "learning_rate": 5.7142271600652e-09, "loss": 0.802, "step": 11546 }, { "epoch": 0.99, "learning_rate": 5.648370061229758e-09, "loss": 0.7791, "step": 11547 }, { "epoch": 0.99, "learning_rate": 5.582894455890553e-09, "loss": 0.7817, "step": 11548 }, { "epoch": 0.99, "learning_rate": 5.51780034904914e-09, "loss": 0.7818, "step": 11549 }, { "epoch": 0.99, "learning_rate": 5.4530877456770995e-09, "loss": 0.7853, "step": 11550 }, { "epoch": 0.99, "learning_rate": 5.388756650718807e-09, "loss": 0.8009, "step": 11551 }, { "epoch": 0.99, "learning_rate": 5.324807069086446e-09, "loss": 0.4705, "step": 11552 }, { "epoch": 0.99, "learning_rate": 5.261239005666108e-09, "loss": 0.8392, "step": 11553 }, { "epoch": 0.99, "learning_rate": 5.198052465313352e-09, "loss": 0.8201, "step": 11554 }, { "epoch": 0.99, "learning_rate": 5.1352474528537645e-09, "loss": 0.7656, "step": 11555 }, { "epoch": 0.99, "learning_rate": 5.072823973085728e-09, "loss": 0.8029, "step": 11556 }, { "epoch": 0.99, "learning_rate": 5.0107820307770945e-09, "loss": 0.8023, "step": 11557 }, { "epoch": 0.99, "learning_rate": 4.9491216306662985e-09, "loss": 0.8141, "step": 11558 }, { "epoch": 0.99, "learning_rate": 4.8878427774640135e-09, "loss": 0.8057, "step": 11559 }, { "epoch": 0.99, "learning_rate": 4.826945475850942e-09, "loss": 0.7911, "step": 11560 }, { "epoch": 0.99, "learning_rate": 4.766429730478916e-09, "loss": 0.7651, "step": 11561 }, { "epoch": 0.99, "learning_rate": 4.7062955459697966e-09, "loss": 0.7596, "step": 11562 }, { "epoch": 0.99, "learning_rate": 4.6465429269176856e-09, "loss": 0.7814, "step": 11563 }, { "epoch": 0.99, "learning_rate": 4.587171877886154e-09, "loss": 0.8119, "step": 11564 }, { "epoch": 0.99, "learning_rate": 4.5281824034104635e-09, "loss": 0.7806, "step": 11565 }, { "epoch": 0.99, "learning_rate": 4.469574507996455e-09, "loss": 0.8017, "step": 11566 }, { "epoch": 0.99, "learning_rate": 4.411348196121656e-09, "loss": 0.7263, "step": 11567 }, { "epoch": 0.99, "learning_rate": 4.353503472231957e-09, "loss": 0.8006, "step": 11568 }, { "epoch": 0.99, "learning_rate": 4.296040340747709e-09, "loss": 0.8017, "step": 11569 }, { "epoch": 0.99, "learning_rate": 4.23895880605707e-09, "loss": 0.8321, "step": 11570 }, { "epoch": 0.99, "learning_rate": 4.182258872520439e-09, "loss": 0.8262, "step": 11571 }, { "epoch": 0.99, "learning_rate": 4.125940544469353e-09, "loss": 0.7782, "step": 11572 }, { "epoch": 0.99, "learning_rate": 4.070003826204816e-09, "loss": 0.5281, "step": 11573 }, { "epoch": 0.99, "learning_rate": 4.01444872200063e-09, "loss": 0.8315, "step": 11574 }, { "epoch": 0.99, "learning_rate": 3.959275236099514e-09, "loss": 0.7273, "step": 11575 }, { "epoch": 0.99, "learning_rate": 3.904483372715873e-09, "loss": 0.7863, "step": 11576 }, { "epoch": 0.99, "learning_rate": 3.850073136035248e-09, "loss": 0.7855, "step": 11577 }, { "epoch": 0.99, "learning_rate": 3.79604453021376e-09, "loss": 0.7673, "step": 11578 }, { "epoch": 0.99, "learning_rate": 3.742397559378663e-09, "loss": 0.7939, "step": 11579 }, { "epoch": 0.99, "learning_rate": 3.68913222762779e-09, "loss": 0.7923, "step": 11580 }, { "epoch": 0.99, "learning_rate": 3.636248539028997e-09, "loss": 0.7698, "step": 11581 }, { "epoch": 0.99, "learning_rate": 3.5837464976223867e-09, "loss": 0.788, "step": 11582 }, { "epoch": 0.99, "learning_rate": 3.5316261074186398e-09, "loss": 0.7885, "step": 11583 }, { "epoch": 0.99, "learning_rate": 3.4798873723984604e-09, "loss": 0.7536, "step": 11584 }, { "epoch": 0.99, "learning_rate": 3.428530296514243e-09, "loss": 0.8179, "step": 11585 }, { "epoch": 0.99, "learning_rate": 3.3775548836884054e-09, "loss": 0.7738, "step": 11586 }, { "epoch": 0.99, "learning_rate": 3.3269611378150545e-09, "loss": 0.7918, "step": 11587 }, { "epoch": 0.99, "learning_rate": 3.276749062758877e-09, "loss": 0.8396, "step": 11588 }, { "epoch": 0.99, "learning_rate": 3.2269186623551384e-09, "loss": 0.8231, "step": 11589 }, { "epoch": 0.99, "learning_rate": 3.177469940410238e-09, "loss": 0.7497, "step": 11590 }, { "epoch": 0.99, "learning_rate": 3.1284029007011553e-09, "loss": 0.8225, "step": 11591 }, { "epoch": 0.99, "learning_rate": 3.079717546976002e-09, "loss": 0.7458, "step": 11592 }, { "epoch": 0.99, "learning_rate": 3.031413882953471e-09, "loss": 0.8132, "step": 11593 }, { "epoch": 0.99, "learning_rate": 2.9834919123233887e-09, "loss": 0.7673, "step": 11594 }, { "epoch": 0.99, "learning_rate": 2.9359516387467145e-09, "loss": 0.8317, "step": 11595 }, { "epoch": 0.99, "learning_rate": 2.888793065853879e-09, "loss": 0.7974, "step": 11596 }, { "epoch": 0.99, "learning_rate": 2.8420161972481096e-09, "loss": 0.7378, "step": 11597 }, { "epoch": 0.99, "learning_rate": 2.79562103650155e-09, "loss": 0.7788, "step": 11598 }, { "epoch": 0.99, "learning_rate": 2.749607587158587e-09, "loss": 0.8159, "step": 11599 }, { "epoch": 0.99, "learning_rate": 2.7039758527336313e-09, "loss": 0.7213, "step": 11600 }, { "epoch": 0.99, "learning_rate": 2.658725836712228e-09, "loss": 0.7728, "step": 11601 }, { "epoch": 0.99, "learning_rate": 2.613857542551612e-09, "loss": 0.7868, "step": 11602 }, { "epoch": 0.99, "learning_rate": 2.5693709736779314e-09, "loss": 0.7728, "step": 11603 }, { "epoch": 0.99, "learning_rate": 2.525266133490134e-09, "loss": 0.7798, "step": 11604 }, { "epoch": 0.99, "learning_rate": 2.4815430253566363e-09, "loss": 0.7906, "step": 11605 }, { "epoch": 0.99, "learning_rate": 2.4382016526175446e-09, "loss": 0.7782, "step": 11606 }, { "epoch": 0.99, "learning_rate": 2.3952420185835434e-09, "loss": 0.7804, "step": 11607 }, { "epoch": 0.99, "learning_rate": 2.3526641265358975e-09, "loss": 0.8219, "step": 11608 }, { "epoch": 0.99, "learning_rate": 2.3104679797270046e-09, "loss": 0.7912, "step": 11609 }, { "epoch": 0.99, "learning_rate": 2.2686535813798426e-09, "loss": 0.7899, "step": 11610 }, { "epoch": 0.99, "learning_rate": 2.2272209346885233e-09, "loss": 0.8003, "step": 11611 }, { "epoch": 0.99, "learning_rate": 2.186170042818292e-09, "loss": 0.8153, "step": 11612 }, { "epoch": 0.99, "learning_rate": 2.1455009089038635e-09, "loss": 0.7954, "step": 11613 }, { "epoch": 0.99, "learning_rate": 2.1052135360527525e-09, "loss": 0.8162, "step": 11614 }, { "epoch": 0.99, "learning_rate": 2.065307927341942e-09, "loss": 0.7664, "step": 11615 }, { "epoch": 0.99, "learning_rate": 2.0257840858189935e-09, "loss": 0.7688, "step": 11616 }, { "epoch": 0.99, "learning_rate": 1.9866420145037144e-09, "loss": 0.7699, "step": 11617 }, { "epoch": 0.99, "learning_rate": 1.9478817163859353e-09, "loss": 0.7751, "step": 11618 }, { "epoch": 0.99, "learning_rate": 1.9095031944260656e-09, "loss": 0.7325, "step": 11619 }, { "epoch": 0.99, "learning_rate": 1.8715064515562042e-09, "loss": 0.7966, "step": 11620 }, { "epoch": 0.99, "learning_rate": 1.8338914906779194e-09, "loss": 0.8046, "step": 11621 }, { "epoch": 0.99, "learning_rate": 1.7966583146644677e-09, "loss": 0.7979, "step": 11622 }, { "epoch": 0.99, "learning_rate": 1.7598069263607965e-09, "loss": 0.7384, "step": 11623 }, { "epoch": 0.99, "learning_rate": 1.7233373285807653e-09, "loss": 0.7901, "step": 11624 }, { "epoch": 0.99, "learning_rate": 1.6872495241110344e-09, "loss": 0.8122, "step": 11625 }, { "epoch": 0.99, "learning_rate": 1.6515435157071769e-09, "loss": 0.781, "step": 11626 }, { "epoch": 0.99, "learning_rate": 1.6162193060975662e-09, "loss": 0.7888, "step": 11627 }, { "epoch": 0.99, "learning_rate": 1.5812768979794891e-09, "loss": 0.8099, "step": 11628 }, { "epoch": 0.99, "learning_rate": 1.5467162940230318e-09, "loss": 0.8082, "step": 11629 }, { "epoch": 0.99, "learning_rate": 1.5125374968677498e-09, "loss": 0.7999, "step": 11630 }, { "epoch": 0.99, "learning_rate": 1.4787405091237773e-09, "loss": 0.7755, "step": 11631 }, { "epoch": 0.99, "learning_rate": 1.4453253333734929e-09, "loss": 0.7823, "step": 11632 }, { "epoch": 0.99, "learning_rate": 1.4122919721687445e-09, "loss": 0.7371, "step": 11633 }, { "epoch": 0.99, "learning_rate": 1.3796404280330688e-09, "loss": 0.7577, "step": 11634 }, { "epoch": 0.99, "learning_rate": 1.3473707034611372e-09, "loss": 0.8399, "step": 11635 }, { "epoch": 0.99, "learning_rate": 1.3154828009165343e-09, "loss": 0.7593, "step": 11636 }, { "epoch": 0.99, "learning_rate": 1.2839767228361999e-09, "loss": 0.7867, "step": 11637 }, { "epoch": 0.99, "learning_rate": 1.2528524716259872e-09, "loss": 0.7852, "step": 11638 }, { "epoch": 0.99, "learning_rate": 1.2221100496634387e-09, "loss": 0.7639, "step": 11639 }, { "epoch": 0.99, "learning_rate": 1.191749459297231e-09, "loss": 0.7934, "step": 11640 }, { "epoch": 0.99, "learning_rate": 1.1617707028460657e-09, "loss": 0.7798, "step": 11641 }, { "epoch": 0.99, "learning_rate": 1.1321737825997769e-09, "loss": 0.7691, "step": 11642 }, { "epoch": 0.99, "learning_rate": 1.1029587008198895e-09, "loss": 0.81, "step": 11643 }, { "epoch": 0.99, "learning_rate": 1.0741254597373961e-09, "loss": 0.7638, "step": 11644 }, { "epoch": 0.99, "learning_rate": 1.0456740615544248e-09, "loss": 0.812, "step": 11645 }, { "epoch": 0.99, "learning_rate": 1.0176045084447916e-09, "loss": 0.798, "step": 11646 }, { "epoch": 0.99, "learning_rate": 9.899168025523376e-10, "loss": 0.8128, "step": 11647 }, { "epoch": 0.99, "learning_rate": 9.626109459920373e-10, "loss": 0.758, "step": 11648 }, { "epoch": 0.99, "learning_rate": 9.356869408500002e-10, "loss": 0.7896, "step": 11649 }, { "epoch": 0.99, "learning_rate": 9.09144789181804e-10, "loss": 0.7318, "step": 11650 }, { "epoch": 0.99, "learning_rate": 8.829844930163811e-10, "loss": 0.7697, "step": 11651 }, { "epoch": 0.99, "learning_rate": 8.572060543504679e-10, "loss": 0.796, "step": 11652 }, { "epoch": 0.99, "learning_rate": 8.31809475154155e-10, "loss": 0.8204, "step": 11653 }, { "epoch": 0.99, "learning_rate": 8.067947573670021e-10, "loss": 0.7571, "step": 11654 }, { "epoch": 0.99, "learning_rate": 7.821619028997029e-10, "loss": 0.8082, "step": 11655 }, { "epoch": 0.99, "learning_rate": 7.579109136335306e-10, "loss": 0.7385, "step": 11656 }, { "epoch": 0.99, "learning_rate": 7.340417914220022e-10, "loss": 0.7888, "step": 11657 }, { "epoch": 0.99, "learning_rate": 7.105545380869938e-10, "loss": 0.8059, "step": 11658 }, { "epoch": 0.99, "learning_rate": 6.874491554237361e-10, "loss": 0.8097, "step": 11659 }, { "epoch": 0.99, "learning_rate": 6.647256451963735e-10, "loss": 0.7631, "step": 11660 }, { "epoch": 1.0, "learning_rate": 6.423840091412947e-10, "loss": 0.8406, "step": 11661 }, { "epoch": 1.0, "learning_rate": 6.204242489638023e-10, "loss": 0.7892, "step": 11662 }, { "epoch": 1.0, "learning_rate": 5.988463663431088e-10, "loss": 0.8198, "step": 11663 }, { "epoch": 1.0, "learning_rate": 5.776503629262298e-10, "loss": 0.8035, "step": 11664 }, { "epoch": 1.0, "learning_rate": 5.568362403318706e-10, "loss": 0.7881, "step": 11665 }, { "epoch": 1.0, "learning_rate": 5.364040001509807e-10, "loss": 0.8363, "step": 11666 }, { "epoch": 1.0, "learning_rate": 5.163536439439787e-10, "loss": 0.783, "step": 11667 }, { "epoch": 1.0, "learning_rate": 4.96685173241862e-10, "loss": 0.7609, "step": 11668 }, { "epoch": 1.0, "learning_rate": 4.773985895478727e-10, "loss": 0.7571, "step": 11669 }, { "epoch": 1.0, "learning_rate": 4.584938943347217e-10, "loss": 0.8256, "step": 11670 }, { "epoch": 1.0, "learning_rate": 4.3997108904625385e-10, "loss": 0.7546, "step": 11671 }, { "epoch": 1.0, "learning_rate": 4.218301750974485e-10, "loss": 0.7999, "step": 11672 }, { "epoch": 1.0, "learning_rate": 4.0407115387386397e-10, "loss": 0.7204, "step": 11673 }, { "epoch": 1.0, "learning_rate": 3.8669402673274794e-10, "loss": 0.7654, "step": 11674 }, { "epoch": 1.0, "learning_rate": 3.696987950008168e-10, "loss": 0.8019, "step": 11675 }, { "epoch": 1.0, "learning_rate": 3.530854599764766e-10, "loss": 0.789, "step": 11676 }, { "epoch": 1.0, "learning_rate": 3.3685402292871206e-10, "loss": 0.7493, "step": 11677 }, { "epoch": 1.0, "learning_rate": 3.210044850970873e-10, "loss": 0.7514, "step": 11678 }, { "epoch": 1.0, "learning_rate": 3.055368476923004e-10, "loss": 0.7748, "step": 11679 }, { "epoch": 1.0, "learning_rate": 2.904511118961839e-10, "loss": 0.8189, "step": 11680 }, { "epoch": 1.0, "learning_rate": 2.7574727886114925e-10, "loss": 0.7477, "step": 11681 }, { "epoch": 1.0, "learning_rate": 2.6142534971018705e-10, "loss": 0.7568, "step": 11682 }, { "epoch": 1.0, "learning_rate": 2.47485325536867e-10, "loss": 0.7285, "step": 11683 }, { "epoch": 1.0, "learning_rate": 2.3392720740644805e-10, "loss": 0.4808, "step": 11684 }, { "epoch": 1.0, "learning_rate": 2.2075099635421316e-10, "loss": 0.7683, "step": 11685 }, { "epoch": 1.0, "learning_rate": 2.0795669338713464e-10, "loss": 0.7275, "step": 11686 }, { "epoch": 1.0, "learning_rate": 1.9554429948220877e-10, "loss": 0.8022, "step": 11687 }, { "epoch": 1.0, "learning_rate": 1.83513815587566e-10, "loss": 0.7946, "step": 11688 }, { "epoch": 1.0, "learning_rate": 1.7186524262191585e-10, "loss": 0.782, "step": 11689 }, { "epoch": 1.0, "learning_rate": 1.605985814756572e-10, "loss": 0.7671, "step": 11690 }, { "epoch": 1.0, "learning_rate": 1.4971383300865782e-10, "loss": 0.818, "step": 11691 }, { "epoch": 1.0, "learning_rate": 1.3921099805302985e-10, "loss": 0.7982, "step": 11692 }, { "epoch": 1.0, "learning_rate": 1.290900774109094e-10, "loss": 0.7313, "step": 11693 }, { "epoch": 1.0, "learning_rate": 1.193510718550117e-10, "loss": 0.7474, "step": 11694 }, { "epoch": 1.0, "learning_rate": 1.0999398212918622e-10, "loss": 0.7939, "step": 11695 }, { "epoch": 1.0, "learning_rate": 1.0101880894841653e-10, "loss": 0.7602, "step": 11696 }, { "epoch": 1.0, "learning_rate": 9.242555299826539e-11, "loss": 0.8163, "step": 11697 }, { "epoch": 1.0, "learning_rate": 8.42142149354297e-11, "loss": 0.806, "step": 11698 }, { "epoch": 1.0, "learning_rate": 7.638479538663035e-11, "loss": 0.7853, "step": 11699 }, { "epoch": 1.0, "learning_rate": 6.893729495027757e-11, "loss": 0.5165, "step": 11700 }, { "epoch": 1.0, "learning_rate": 6.187171419536064e-11, "loss": 0.7929, "step": 11701 }, { "epoch": 1.0, "learning_rate": 5.518805366089286e-11, "loss": 0.793, "step": 11702 }, { "epoch": 1.0, "learning_rate": 4.888631385757681e-11, "loss": 0.7616, "step": 11703 }, { "epoch": 1.0, "learning_rate": 4.2966495267249274e-11, "loss": 0.7601, "step": 11704 }, { "epoch": 1.0, "learning_rate": 3.742859834232615e-11, "loss": 0.7887, "step": 11705 }, { "epoch": 1.0, "learning_rate": 3.2272623504692177e-11, "loss": 0.7644, "step": 11706 }, { "epoch": 1.0, "learning_rate": 2.749857114958676e-11, "loss": 0.8081, "step": 11707 }, { "epoch": 1.0, "learning_rate": 2.3106441640607934e-11, "loss": 0.7895, "step": 11708 }, { "epoch": 1.0, "learning_rate": 1.9096235313043054e-11, "loss": 0.7568, "step": 11709 }, { "epoch": 1.0, "learning_rate": 1.54679524744239e-11, "loss": 0.7349, "step": 11710 }, { "epoch": 1.0, "learning_rate": 1.2221593401195996e-11, "loss": 0.8098, "step": 11711 }, { "epoch": 1.0, "learning_rate": 9.357158340939088e-12, "loss": 0.7807, "step": 11712 }, { "epoch": 1.0, "learning_rate": 6.874647513477329e-12, "loss": 0.7754, "step": 11713 }, { "epoch": 1.0, "learning_rate": 4.774061107548633e-12, "loss": 0.8099, "step": 11714 }, { "epoch": 1.0, "learning_rate": 3.0553992835802295e-12, "loss": 0.7822, "step": 11715 }, { "epoch": 1.0, "learning_rate": 1.7186621736886567e-12, "loss": 0.7801, "step": 11716 }, { "epoch": 1.0, "learning_rate": 7.638498794593218e-13, "loss": 0.802, "step": 11717 }, { "epoch": 1.0, "learning_rate": 1.9096247361183318e-13, "loss": 0.7943, "step": 11718 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 0.6921, "step": 11719 }, { "epoch": 1.0, "step": 11719, "total_flos": 1.1834950671859712e+16, "train_loss": 0.8405589830276248, "train_runtime": 185618.2665, "train_samples_per_second": 8.081, "train_steps_per_second": 0.063 } ], "logging_steps": 1.0, "max_steps": 11719, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 1000, "total_flos": 1.1834950671859712e+16, "train_batch_size": 4, "trial_name": null, "trial_params": null }