{ "best_metric": 0.06538490206003189, "best_model_checkpoint": "email-spam-detection-roberta/checkpoint-8345", "epoch": 1.0, "eval_steps": 1, "global_step": 8345, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 6.000000000000001e-08, "loss": 0.6937, "step": 1 }, { "epoch": 0.0, "learning_rate": 1.2000000000000002e-07, "loss": 0.6792, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.8e-07, "loss": 0.6904, "step": 3 }, { "epoch": 0.0, "learning_rate": 2.4000000000000003e-07, "loss": 0.7266, "step": 4 }, { "epoch": 0.0, "learning_rate": 3.0000000000000004e-07, "loss": 0.714, "step": 5 }, { "epoch": 0.0, "learning_rate": 3.6e-07, "loss": 0.7124, "step": 6 }, { "epoch": 0.0, "learning_rate": 4.2000000000000006e-07, "loss": 0.6731, "step": 7 }, { "epoch": 0.0, "learning_rate": 4.800000000000001e-07, "loss": 0.6605, "step": 8 }, { "epoch": 0.0, "learning_rate": 5.4e-07, "loss": 0.6648, "step": 9 }, { "epoch": 0.0, "learning_rate": 6.000000000000001e-07, "loss": 0.7258, "step": 10 }, { "epoch": 0.0, "learning_rate": 6.6e-07, "loss": 0.6749, "step": 11 }, { "epoch": 0.0, "learning_rate": 7.2e-07, "loss": 0.7003, "step": 12 }, { "epoch": 0.0, "learning_rate": 7.8e-07, "loss": 0.6821, "step": 13 }, { "epoch": 0.0, "learning_rate": 8.400000000000001e-07, "loss": 0.676, "step": 14 }, { "epoch": 0.0, "learning_rate": 9e-07, "loss": 0.6913, "step": 15 }, { "epoch": 0.0, "learning_rate": 9.600000000000001e-07, "loss": 0.6948, "step": 16 }, { "epoch": 0.0, "learning_rate": 1.0200000000000002e-06, "loss": 0.6956, "step": 17 }, { "epoch": 0.0, "learning_rate": 1.08e-06, "loss": 0.7193, "step": 18 }, { "epoch": 0.0, "learning_rate": 1.14e-06, "loss": 0.696, "step": 19 }, { "epoch": 0.0, "learning_rate": 1.2000000000000002e-06, "loss": 0.6884, "step": 20 }, { "epoch": 0.0, "learning_rate": 1.26e-06, "loss": 0.7021, "step": 21 }, { "epoch": 0.0, "learning_rate": 1.32e-06, "loss": 0.6803, "step": 22 }, { "epoch": 0.0, "learning_rate": 1.3800000000000001e-06, "loss": 0.6844, "step": 23 }, { "epoch": 0.0, "learning_rate": 1.44e-06, "loss": 0.6849, "step": 24 }, { "epoch": 0.0, "learning_rate": 1.5e-06, "loss": 0.7154, "step": 25 }, { "epoch": 0.0, "learning_rate": 1.56e-06, "loss": 0.7066, "step": 26 }, { "epoch": 0.0, "learning_rate": 1.6200000000000002e-06, "loss": 0.6905, "step": 27 }, { "epoch": 0.0, "learning_rate": 1.6800000000000002e-06, "loss": 0.6848, "step": 28 }, { "epoch": 0.0, "learning_rate": 1.7399999999999999e-06, "loss": 0.6752, "step": 29 }, { "epoch": 0.0, "learning_rate": 1.8e-06, "loss": 0.699, "step": 30 }, { "epoch": 0.0, "learning_rate": 1.86e-06, "loss": 0.7, "step": 31 }, { "epoch": 0.0, "learning_rate": 1.9200000000000003e-06, "loss": 0.7487, "step": 32 }, { "epoch": 0.0, "learning_rate": 1.98e-06, "loss": 0.6824, "step": 33 }, { "epoch": 0.0, "learning_rate": 2.0400000000000004e-06, "loss": 0.6962, "step": 34 }, { "epoch": 0.0, "learning_rate": 2.1e-06, "loss": 0.6992, "step": 35 }, { "epoch": 0.0, "learning_rate": 2.16e-06, "loss": 0.677, "step": 36 }, { "epoch": 0.0, "learning_rate": 2.22e-06, "loss": 0.6891, "step": 37 }, { "epoch": 0.0, "learning_rate": 2.28e-06, "loss": 0.6881, "step": 38 }, { "epoch": 0.0, "learning_rate": 2.34e-06, "loss": 0.6826, "step": 39 }, { "epoch": 0.0, "learning_rate": 2.4000000000000003e-06, "loss": 0.7039, "step": 40 }, { "epoch": 0.0, "learning_rate": 2.4599999999999997e-06, "loss": 0.676, "step": 41 }, { "epoch": 0.01, "learning_rate": 2.52e-06, "loss": 0.6853, "step": 42 }, { "epoch": 0.01, "learning_rate": 2.58e-06, "loss": 0.6925, "step": 43 }, { "epoch": 0.01, "learning_rate": 2.64e-06, "loss": 0.6677, "step": 44 }, { "epoch": 0.01, "learning_rate": 2.7e-06, "loss": 0.6864, "step": 45 }, { "epoch": 0.01, "learning_rate": 2.7600000000000003e-06, "loss": 0.6949, "step": 46 }, { "epoch": 0.01, "learning_rate": 2.82e-06, "loss": 0.6862, "step": 47 }, { "epoch": 0.01, "learning_rate": 2.88e-06, "loss": 0.6707, "step": 48 }, { "epoch": 0.01, "learning_rate": 2.9400000000000002e-06, "loss": 0.683, "step": 49 }, { "epoch": 0.01, "learning_rate": 3e-06, "loss": 0.6959, "step": 50 }, { "epoch": 0.01, "learning_rate": 2.9996383363471973e-06, "loss": 0.666, "step": 51 }, { "epoch": 0.01, "learning_rate": 2.999276672694394e-06, "loss": 0.6933, "step": 52 }, { "epoch": 0.01, "learning_rate": 2.9989150090415913e-06, "loss": 0.6784, "step": 53 }, { "epoch": 0.01, "learning_rate": 2.9985533453887886e-06, "loss": 0.6764, "step": 54 }, { "epoch": 0.01, "learning_rate": 2.998191681735986e-06, "loss": 0.6711, "step": 55 }, { "epoch": 0.01, "learning_rate": 2.9978300180831826e-06, "loss": 0.723, "step": 56 }, { "epoch": 0.01, "learning_rate": 2.99746835443038e-06, "loss": 0.659, "step": 57 }, { "epoch": 0.01, "learning_rate": 2.997106690777577e-06, "loss": 0.6803, "step": 58 }, { "epoch": 0.01, "learning_rate": 2.9967450271247743e-06, "loss": 0.6849, "step": 59 }, { "epoch": 0.01, "learning_rate": 2.996383363471971e-06, "loss": 0.6671, "step": 60 }, { "epoch": 0.01, "learning_rate": 2.9960216998191683e-06, "loss": 0.6928, "step": 61 }, { "epoch": 0.01, "learning_rate": 2.995660036166365e-06, "loss": 0.6772, "step": 62 }, { "epoch": 0.01, "learning_rate": 2.9952983725135624e-06, "loss": 0.6826, "step": 63 }, { "epoch": 0.01, "learning_rate": 2.9949367088607596e-06, "loss": 0.6627, "step": 64 }, { "epoch": 0.01, "learning_rate": 2.9945750452079564e-06, "loss": 0.6815, "step": 65 }, { "epoch": 0.01, "learning_rate": 2.9942133815551537e-06, "loss": 0.6382, "step": 66 }, { "epoch": 0.01, "learning_rate": 2.993851717902351e-06, "loss": 0.6241, "step": 67 }, { "epoch": 0.01, "learning_rate": 2.993490054249548e-06, "loss": 0.6584, "step": 68 }, { "epoch": 0.01, "learning_rate": 2.993128390596745e-06, "loss": 0.6966, "step": 69 }, { "epoch": 0.01, "learning_rate": 2.992766726943942e-06, "loss": 0.6823, "step": 70 }, { "epoch": 0.01, "learning_rate": 2.9924050632911394e-06, "loss": 0.6278, "step": 71 }, { "epoch": 0.01, "learning_rate": 2.992043399638336e-06, "loss": 0.6406, "step": 72 }, { "epoch": 0.01, "learning_rate": 2.9916817359855334e-06, "loss": 0.6376, "step": 73 }, { "epoch": 0.01, "learning_rate": 2.9913200723327307e-06, "loss": 0.7178, "step": 74 }, { "epoch": 0.01, "learning_rate": 2.990958408679928e-06, "loss": 0.669, "step": 75 }, { "epoch": 0.01, "learning_rate": 2.9905967450271247e-06, "loss": 0.6778, "step": 76 }, { "epoch": 0.01, "learning_rate": 2.990235081374322e-06, "loss": 0.6515, "step": 77 }, { "epoch": 0.01, "learning_rate": 2.989873417721519e-06, "loss": 0.6865, "step": 78 }, { "epoch": 0.01, "learning_rate": 2.9895117540687164e-06, "loss": 0.7508, "step": 79 }, { "epoch": 0.01, "learning_rate": 2.989150090415913e-06, "loss": 0.6932, "step": 80 }, { "epoch": 0.01, "learning_rate": 2.9887884267631104e-06, "loss": 0.6252, "step": 81 }, { "epoch": 0.01, "learning_rate": 2.9884267631103077e-06, "loss": 0.6753, "step": 82 }, { "epoch": 0.01, "learning_rate": 2.988065099457505e-06, "loss": 0.6028, "step": 83 }, { "epoch": 0.01, "learning_rate": 2.9877034358047017e-06, "loss": 0.7477, "step": 84 }, { "epoch": 0.01, "learning_rate": 2.987341772151899e-06, "loss": 0.6045, "step": 85 }, { "epoch": 0.01, "learning_rate": 2.9869801084990957e-06, "loss": 0.6451, "step": 86 }, { "epoch": 0.01, "learning_rate": 2.986618444846293e-06, "loss": 0.6667, "step": 87 }, { "epoch": 0.01, "learning_rate": 2.98625678119349e-06, "loss": 0.6673, "step": 88 }, { "epoch": 0.01, "learning_rate": 2.985895117540687e-06, "loss": 0.6404, "step": 89 }, { "epoch": 0.01, "learning_rate": 2.9855334538878842e-06, "loss": 0.6362, "step": 90 }, { "epoch": 0.01, "learning_rate": 2.9851717902350815e-06, "loss": 0.6501, "step": 91 }, { "epoch": 0.01, "learning_rate": 2.9848101265822783e-06, "loss": 0.5689, "step": 92 }, { "epoch": 0.01, "learning_rate": 2.9844484629294755e-06, "loss": 0.6588, "step": 93 }, { "epoch": 0.01, "learning_rate": 2.9840867992766727e-06, "loss": 0.5602, "step": 94 }, { "epoch": 0.01, "learning_rate": 2.98372513562387e-06, "loss": 0.5826, "step": 95 }, { "epoch": 0.01, "learning_rate": 2.9833634719710668e-06, "loss": 0.5189, "step": 96 }, { "epoch": 0.01, "learning_rate": 2.983001808318264e-06, "loss": 0.5703, "step": 97 }, { "epoch": 0.01, "learning_rate": 2.9826401446654612e-06, "loss": 0.5546, "step": 98 }, { "epoch": 0.01, "learning_rate": 2.9822784810126585e-06, "loss": 0.5888, "step": 99 }, { "epoch": 0.01, "learning_rate": 2.9819168173598553e-06, "loss": 0.4787, "step": 100 }, { "epoch": 0.01, "learning_rate": 2.9815551537070525e-06, "loss": 0.5301, "step": 101 }, { "epoch": 0.01, "learning_rate": 2.9811934900542497e-06, "loss": 0.5279, "step": 102 }, { "epoch": 0.01, "learning_rate": 2.980831826401447e-06, "loss": 0.5466, "step": 103 }, { "epoch": 0.01, "learning_rate": 2.9804701627486438e-06, "loss": 0.4475, "step": 104 }, { "epoch": 0.01, "learning_rate": 2.980108499095841e-06, "loss": 0.5345, "step": 105 }, { "epoch": 0.01, "learning_rate": 2.9797468354430382e-06, "loss": 0.4166, "step": 106 }, { "epoch": 0.01, "learning_rate": 2.9793851717902355e-06, "loss": 0.5127, "step": 107 }, { "epoch": 0.01, "learning_rate": 2.9790235081374323e-06, "loss": 0.6167, "step": 108 }, { "epoch": 0.01, "learning_rate": 2.9786618444846295e-06, "loss": 0.4234, "step": 109 }, { "epoch": 0.01, "learning_rate": 2.9783001808318263e-06, "loss": 0.5026, "step": 110 }, { "epoch": 0.01, "learning_rate": 2.9779385171790236e-06, "loss": 0.3723, "step": 111 }, { "epoch": 0.01, "learning_rate": 2.9775768535262204e-06, "loss": 0.5485, "step": 112 }, { "epoch": 0.01, "learning_rate": 2.9772151898734176e-06, "loss": 0.5197, "step": 113 }, { "epoch": 0.01, "learning_rate": 2.976853526220615e-06, "loss": 0.4378, "step": 114 }, { "epoch": 0.01, "learning_rate": 2.976491862567812e-06, "loss": 0.5589, "step": 115 }, { "epoch": 0.01, "learning_rate": 2.976130198915009e-06, "loss": 0.4933, "step": 116 }, { "epoch": 0.01, "learning_rate": 2.975768535262206e-06, "loss": 0.3765, "step": 117 }, { "epoch": 0.01, "learning_rate": 2.9754068716094033e-06, "loss": 0.3891, "step": 118 }, { "epoch": 0.01, "learning_rate": 2.9750452079566006e-06, "loss": 0.4171, "step": 119 }, { "epoch": 0.01, "learning_rate": 2.9746835443037974e-06, "loss": 0.2326, "step": 120 }, { "epoch": 0.01, "learning_rate": 2.9743218806509946e-06, "loss": 0.4467, "step": 121 }, { "epoch": 0.01, "learning_rate": 2.973960216998192e-06, "loss": 0.3009, "step": 122 }, { "epoch": 0.01, "learning_rate": 2.973598553345389e-06, "loss": 0.3877, "step": 123 }, { "epoch": 0.01, "learning_rate": 2.973236889692586e-06, "loss": 0.5998, "step": 124 }, { "epoch": 0.01, "learning_rate": 2.972875226039783e-06, "loss": 0.4372, "step": 125 }, { "epoch": 0.02, "learning_rate": 2.9725135623869803e-06, "loss": 0.3109, "step": 126 }, { "epoch": 0.02, "learning_rate": 2.9721518987341776e-06, "loss": 0.3411, "step": 127 }, { "epoch": 0.02, "learning_rate": 2.9717902350813744e-06, "loss": 0.213, "step": 128 }, { "epoch": 0.02, "learning_rate": 2.9714285714285716e-06, "loss": 0.414, "step": 129 }, { "epoch": 0.02, "learning_rate": 2.971066907775769e-06, "loss": 0.4402, "step": 130 }, { "epoch": 0.02, "learning_rate": 2.9707052441229656e-06, "loss": 0.4468, "step": 131 }, { "epoch": 0.02, "learning_rate": 2.970343580470163e-06, "loss": 0.4433, "step": 132 }, { "epoch": 0.02, "learning_rate": 2.96998191681736e-06, "loss": 0.1103, "step": 133 }, { "epoch": 0.02, "learning_rate": 2.969620253164557e-06, "loss": 0.2949, "step": 134 }, { "epoch": 0.02, "learning_rate": 2.969258589511754e-06, "loss": 0.272, "step": 135 }, { "epoch": 0.02, "learning_rate": 2.968896925858951e-06, "loss": 0.3632, "step": 136 }, { "epoch": 0.02, "learning_rate": 2.968535262206148e-06, "loss": 0.4219, "step": 137 }, { "epoch": 0.02, "learning_rate": 2.9681735985533454e-06, "loss": 0.3187, "step": 138 }, { "epoch": 0.02, "learning_rate": 2.9678119349005426e-06, "loss": 0.1422, "step": 139 }, { "epoch": 0.02, "learning_rate": 2.9674502712477394e-06, "loss": 0.4607, "step": 140 }, { "epoch": 0.02, "learning_rate": 2.9670886075949367e-06, "loss": 0.1419, "step": 141 }, { "epoch": 0.02, "learning_rate": 2.966726943942134e-06, "loss": 0.1187, "step": 142 }, { "epoch": 0.02, "learning_rate": 2.966365280289331e-06, "loss": 0.2841, "step": 143 }, { "epoch": 0.02, "learning_rate": 2.966003616636528e-06, "loss": 0.4261, "step": 144 }, { "epoch": 0.02, "learning_rate": 2.965641952983725e-06, "loss": 0.1203, "step": 145 }, { "epoch": 0.02, "learning_rate": 2.9652802893309224e-06, "loss": 0.522, "step": 146 }, { "epoch": 0.02, "learning_rate": 2.9649186256781196e-06, "loss": 0.3277, "step": 147 }, { "epoch": 0.02, "learning_rate": 2.9645569620253164e-06, "loss": 0.1366, "step": 148 }, { "epoch": 0.02, "learning_rate": 2.9641952983725137e-06, "loss": 0.1753, "step": 149 }, { "epoch": 0.02, "learning_rate": 2.963833634719711e-06, "loss": 0.2735, "step": 150 }, { "epoch": 0.02, "learning_rate": 2.9634719710669077e-06, "loss": 0.3221, "step": 151 }, { "epoch": 0.02, "learning_rate": 2.963110307414105e-06, "loss": 0.2371, "step": 152 }, { "epoch": 0.02, "learning_rate": 2.962748643761302e-06, "loss": 0.1003, "step": 153 }, { "epoch": 0.02, "learning_rate": 2.9623869801084994e-06, "loss": 0.1256, "step": 154 }, { "epoch": 0.02, "learning_rate": 2.9620253164556962e-06, "loss": 0.1615, "step": 155 }, { "epoch": 0.02, "learning_rate": 2.9616636528028934e-06, "loss": 0.4206, "step": 156 }, { "epoch": 0.02, "learning_rate": 2.9613019891500907e-06, "loss": 0.1749, "step": 157 }, { "epoch": 0.02, "learning_rate": 2.9609403254972875e-06, "loss": 0.1897, "step": 158 }, { "epoch": 0.02, "learning_rate": 2.9605786618444847e-06, "loss": 0.0864, "step": 159 }, { "epoch": 0.02, "learning_rate": 2.9602169981916815e-06, "loss": 0.1223, "step": 160 }, { "epoch": 0.02, "learning_rate": 2.9598553345388788e-06, "loss": 0.308, "step": 161 }, { "epoch": 0.02, "learning_rate": 2.959493670886076e-06, "loss": 0.3341, "step": 162 }, { "epoch": 0.02, "learning_rate": 2.9591320072332732e-06, "loss": 0.2211, "step": 163 }, { "epoch": 0.02, "learning_rate": 2.95877034358047e-06, "loss": 0.2708, "step": 164 }, { "epoch": 0.02, "learning_rate": 2.9584086799276673e-06, "loss": 0.4518, "step": 165 }, { "epoch": 0.02, "learning_rate": 2.9580470162748645e-06, "loss": 0.1286, "step": 166 }, { "epoch": 0.02, "learning_rate": 2.9576853526220617e-06, "loss": 0.0926, "step": 167 }, { "epoch": 0.02, "learning_rate": 2.9573236889692585e-06, "loss": 0.0483, "step": 168 }, { "epoch": 0.02, "learning_rate": 2.9569620253164558e-06, "loss": 0.1073, "step": 169 }, { "epoch": 0.02, "learning_rate": 2.956600361663653e-06, "loss": 0.298, "step": 170 }, { "epoch": 0.02, "learning_rate": 2.95623869801085e-06, "loss": 0.0656, "step": 171 }, { "epoch": 0.02, "learning_rate": 2.955877034358047e-06, "loss": 0.1313, "step": 172 }, { "epoch": 0.02, "learning_rate": 2.9555153707052443e-06, "loss": 0.0638, "step": 173 }, { "epoch": 0.02, "learning_rate": 2.9551537070524415e-06, "loss": 0.141, "step": 174 }, { "epoch": 0.02, "learning_rate": 2.9547920433996383e-06, "loss": 0.0583, "step": 175 }, { "epoch": 0.02, "learning_rate": 2.9544303797468355e-06, "loss": 0.0988, "step": 176 }, { "epoch": 0.02, "learning_rate": 2.9540687160940328e-06, "loss": 0.1143, "step": 177 }, { "epoch": 0.02, "learning_rate": 2.95370705244123e-06, "loss": 0.1068, "step": 178 }, { "epoch": 0.02, "learning_rate": 2.953345388788427e-06, "loss": 0.0463, "step": 179 }, { "epoch": 0.02, "learning_rate": 2.952983725135624e-06, "loss": 0.075, "step": 180 }, { "epoch": 0.02, "learning_rate": 2.9526220614828213e-06, "loss": 0.3276, "step": 181 }, { "epoch": 0.02, "learning_rate": 2.952260397830018e-06, "loss": 0.3698, "step": 182 }, { "epoch": 0.02, "learning_rate": 2.9518987341772153e-06, "loss": 0.0427, "step": 183 }, { "epoch": 0.02, "learning_rate": 2.951537070524412e-06, "loss": 0.0531, "step": 184 }, { "epoch": 0.02, "learning_rate": 2.9511754068716093e-06, "loss": 0.3523, "step": 185 }, { "epoch": 0.02, "learning_rate": 2.9508137432188066e-06, "loss": 0.1887, "step": 186 }, { "epoch": 0.02, "learning_rate": 2.950452079566004e-06, "loss": 0.2075, "step": 187 }, { "epoch": 0.02, "learning_rate": 2.9500904159132006e-06, "loss": 0.0643, "step": 188 }, { "epoch": 0.02, "learning_rate": 2.949728752260398e-06, "loss": 0.0486, "step": 189 }, { "epoch": 0.02, "learning_rate": 2.949367088607595e-06, "loss": 0.627, "step": 190 }, { "epoch": 0.02, "learning_rate": 2.949005424954792e-06, "loss": 0.041, "step": 191 }, { "epoch": 0.02, "learning_rate": 2.948643761301989e-06, "loss": 0.0861, "step": 192 }, { "epoch": 0.02, "learning_rate": 2.9482820976491863e-06, "loss": 0.1554, "step": 193 }, { "epoch": 0.02, "learning_rate": 2.9479204339963836e-06, "loss": 0.051, "step": 194 }, { "epoch": 0.02, "learning_rate": 2.9475587703435804e-06, "loss": 0.0564, "step": 195 }, { "epoch": 0.02, "learning_rate": 2.9471971066907776e-06, "loss": 0.15, "step": 196 }, { "epoch": 0.02, "learning_rate": 2.946835443037975e-06, "loss": 0.2048, "step": 197 }, { "epoch": 0.02, "learning_rate": 2.946473779385172e-06, "loss": 0.0779, "step": 198 }, { "epoch": 0.02, "learning_rate": 2.946112115732369e-06, "loss": 0.0345, "step": 199 }, { "epoch": 0.02, "learning_rate": 2.945750452079566e-06, "loss": 0.0267, "step": 200 }, { "epoch": 0.02, "learning_rate": 2.9453887884267633e-06, "loss": 0.2386, "step": 201 }, { "epoch": 0.02, "learning_rate": 2.9450271247739606e-06, "loss": 0.0967, "step": 202 }, { "epoch": 0.02, "learning_rate": 2.9446654611211574e-06, "loss": 0.1996, "step": 203 }, { "epoch": 0.02, "learning_rate": 2.9443037974683546e-06, "loss": 0.3576, "step": 204 }, { "epoch": 0.02, "learning_rate": 2.943942133815552e-06, "loss": 0.3909, "step": 205 }, { "epoch": 0.02, "learning_rate": 2.9435804701627487e-06, "loss": 0.0239, "step": 206 }, { "epoch": 0.02, "learning_rate": 2.943218806509946e-06, "loss": 0.02, "step": 207 }, { "epoch": 0.02, "learning_rate": 2.9428571428571427e-06, "loss": 0.0623, "step": 208 }, { "epoch": 0.03, "learning_rate": 2.94249547920434e-06, "loss": 0.0592, "step": 209 }, { "epoch": 0.03, "learning_rate": 2.942133815551537e-06, "loss": 0.3401, "step": 210 }, { "epoch": 0.03, "learning_rate": 2.941772151898734e-06, "loss": 0.0243, "step": 211 }, { "epoch": 0.03, "learning_rate": 2.941410488245931e-06, "loss": 0.4765, "step": 212 }, { "epoch": 0.03, "learning_rate": 2.9410488245931284e-06, "loss": 0.0235, "step": 213 }, { "epoch": 0.03, "learning_rate": 2.9406871609403257e-06, "loss": 0.4395, "step": 214 }, { "epoch": 0.03, "learning_rate": 2.9403254972875225e-06, "loss": 0.0388, "step": 215 }, { "epoch": 0.03, "learning_rate": 2.9399638336347197e-06, "loss": 0.1226, "step": 216 }, { "epoch": 0.03, "learning_rate": 2.939602169981917e-06, "loss": 0.0741, "step": 217 }, { "epoch": 0.03, "learning_rate": 2.939240506329114e-06, "loss": 0.02, "step": 218 }, { "epoch": 0.03, "learning_rate": 2.938878842676311e-06, "loss": 0.4179, "step": 219 }, { "epoch": 0.03, "learning_rate": 2.938517179023508e-06, "loss": 0.4706, "step": 220 }, { "epoch": 0.03, "learning_rate": 2.9381555153707054e-06, "loss": 0.1489, "step": 221 }, { "epoch": 0.03, "learning_rate": 2.9377938517179027e-06, "loss": 0.0204, "step": 222 }, { "epoch": 0.03, "learning_rate": 2.9374321880650995e-06, "loss": 0.182, "step": 223 }, { "epoch": 0.03, "learning_rate": 2.9370705244122967e-06, "loss": 0.0154, "step": 224 }, { "epoch": 0.03, "learning_rate": 2.936708860759494e-06, "loss": 0.0237, "step": 225 }, { "epoch": 0.03, "learning_rate": 2.936347197106691e-06, "loss": 0.0635, "step": 226 }, { "epoch": 0.03, "learning_rate": 2.935985533453888e-06, "loss": 0.0998, "step": 227 }, { "epoch": 0.03, "learning_rate": 2.935623869801085e-06, "loss": 0.3425, "step": 228 }, { "epoch": 0.03, "learning_rate": 2.9352622061482824e-06, "loss": 0.0332, "step": 229 }, { "epoch": 0.03, "learning_rate": 2.9349005424954792e-06, "loss": 0.3403, "step": 230 }, { "epoch": 0.03, "learning_rate": 2.934538878842676e-06, "loss": 0.4081, "step": 231 }, { "epoch": 0.03, "learning_rate": 2.9341772151898733e-06, "loss": 0.4624, "step": 232 }, { "epoch": 0.03, "learning_rate": 2.9338155515370705e-06, "loss": 0.4829, "step": 233 }, { "epoch": 0.03, "learning_rate": 2.9334538878842677e-06, "loss": 0.0972, "step": 234 }, { "epoch": 0.03, "learning_rate": 2.9330922242314645e-06, "loss": 0.0257, "step": 235 }, { "epoch": 0.03, "learning_rate": 2.9327305605786618e-06, "loss": 0.3762, "step": 236 }, { "epoch": 0.03, "learning_rate": 2.932368896925859e-06, "loss": 0.0841, "step": 237 }, { "epoch": 0.03, "learning_rate": 2.9320072332730562e-06, "loss": 0.0576, "step": 238 }, { "epoch": 0.03, "learning_rate": 2.931645569620253e-06, "loss": 0.0208, "step": 239 }, { "epoch": 0.03, "learning_rate": 2.9312839059674503e-06, "loss": 0.0812, "step": 240 }, { "epoch": 0.03, "learning_rate": 2.9309222423146475e-06, "loss": 0.0519, "step": 241 }, { "epoch": 0.03, "learning_rate": 2.9305605786618447e-06, "loss": 0.0316, "step": 242 }, { "epoch": 0.03, "learning_rate": 2.9301989150090415e-06, "loss": 0.0147, "step": 243 }, { "epoch": 0.03, "learning_rate": 2.9298372513562388e-06, "loss": 0.0147, "step": 244 }, { "epoch": 0.03, "learning_rate": 2.929475587703436e-06, "loss": 0.0116, "step": 245 }, { "epoch": 0.03, "learning_rate": 2.9291139240506332e-06, "loss": 0.1418, "step": 246 }, { "epoch": 0.03, "learning_rate": 2.92875226039783e-06, "loss": 0.0655, "step": 247 }, { "epoch": 0.03, "learning_rate": 2.9283905967450273e-06, "loss": 0.0579, "step": 248 }, { "epoch": 0.03, "learning_rate": 2.9280289330922245e-06, "loss": 0.3237, "step": 249 }, { "epoch": 0.03, "learning_rate": 2.9276672694394217e-06, "loss": 0.0139, "step": 250 }, { "epoch": 0.03, "learning_rate": 2.9273056057866186e-06, "loss": 0.0331, "step": 251 }, { "epoch": 0.03, "learning_rate": 2.9269439421338158e-06, "loss": 0.023, "step": 252 }, { "epoch": 0.03, "learning_rate": 2.926582278481013e-06, "loss": 0.0679, "step": 253 }, { "epoch": 0.03, "learning_rate": 2.92622061482821e-06, "loss": 0.0114, "step": 254 }, { "epoch": 0.03, "learning_rate": 2.9258589511754066e-06, "loss": 0.0191, "step": 255 }, { "epoch": 0.03, "learning_rate": 2.925497287522604e-06, "loss": 0.0813, "step": 256 }, { "epoch": 0.03, "learning_rate": 2.925135623869801e-06, "loss": 0.4516, "step": 257 }, { "epoch": 0.03, "learning_rate": 2.9247739602169983e-06, "loss": 0.0174, "step": 258 }, { "epoch": 0.03, "learning_rate": 2.924412296564195e-06, "loss": 0.0118, "step": 259 }, { "epoch": 0.03, "learning_rate": 2.9240506329113924e-06, "loss": 0.0106, "step": 260 }, { "epoch": 0.03, "learning_rate": 2.9236889692585896e-06, "loss": 0.0171, "step": 261 }, { "epoch": 0.03, "learning_rate": 2.923327305605787e-06, "loss": 0.0066, "step": 262 }, { "epoch": 0.03, "learning_rate": 2.9229656419529836e-06, "loss": 0.6778, "step": 263 }, { "epoch": 0.03, "learning_rate": 2.922603978300181e-06, "loss": 0.0147, "step": 264 }, { "epoch": 0.03, "learning_rate": 2.922242314647378e-06, "loss": 0.0153, "step": 265 }, { "epoch": 0.03, "learning_rate": 2.9218806509945753e-06, "loss": 0.7883, "step": 266 }, { "epoch": 0.03, "learning_rate": 2.921518987341772e-06, "loss": 0.1478, "step": 267 }, { "epoch": 0.03, "learning_rate": 2.9211573236889694e-06, "loss": 0.63, "step": 268 }, { "epoch": 0.03, "learning_rate": 2.9207956600361666e-06, "loss": 0.0324, "step": 269 }, { "epoch": 0.03, "learning_rate": 2.9204339963833634e-06, "loss": 0.0087, "step": 270 }, { "epoch": 0.03, "learning_rate": 2.9200723327305606e-06, "loss": 0.357, "step": 271 }, { "epoch": 0.03, "learning_rate": 2.919710669077758e-06, "loss": 0.1481, "step": 272 }, { "epoch": 0.03, "learning_rate": 2.919349005424955e-06, "loss": 0.0063, "step": 273 }, { "epoch": 0.03, "learning_rate": 2.918987341772152e-06, "loss": 0.0152, "step": 274 }, { "epoch": 0.03, "learning_rate": 2.918625678119349e-06, "loss": 0.4222, "step": 275 }, { "epoch": 0.03, "learning_rate": 2.9182640144665464e-06, "loss": 0.0725, "step": 276 }, { "epoch": 0.03, "learning_rate": 2.9179023508137436e-06, "loss": 0.007, "step": 277 }, { "epoch": 0.03, "learning_rate": 2.9175406871609404e-06, "loss": 0.2973, "step": 278 }, { "epoch": 0.03, "learning_rate": 2.9171790235081372e-06, "loss": 0.0077, "step": 279 }, { "epoch": 0.03, "learning_rate": 2.9168173598553344e-06, "loss": 0.0075, "step": 280 }, { "epoch": 0.03, "learning_rate": 2.9164556962025317e-06, "loss": 0.0605, "step": 281 }, { "epoch": 0.03, "learning_rate": 2.916094032549729e-06, "loss": 0.1778, "step": 282 }, { "epoch": 0.03, "learning_rate": 2.9157323688969257e-06, "loss": 0.0294, "step": 283 }, { "epoch": 0.03, "learning_rate": 2.915370705244123e-06, "loss": 0.0451, "step": 284 }, { "epoch": 0.03, "learning_rate": 2.91500904159132e-06, "loss": 0.0649, "step": 285 }, { "epoch": 0.03, "learning_rate": 2.9146473779385174e-06, "loss": 0.011, "step": 286 }, { "epoch": 0.03, "learning_rate": 2.9142857142857142e-06, "loss": 0.3507, "step": 287 }, { "epoch": 0.03, "learning_rate": 2.9139240506329114e-06, "loss": 0.4663, "step": 288 }, { "epoch": 0.03, "learning_rate": 2.9135623869801087e-06, "loss": 0.0056, "step": 289 }, { "epoch": 0.03, "learning_rate": 2.9132007233273055e-06, "loss": 0.0454, "step": 290 }, { "epoch": 0.03, "learning_rate": 2.9128390596745027e-06, "loss": 0.0071, "step": 291 }, { "epoch": 0.03, "learning_rate": 2.9124773960217e-06, "loss": 0.0113, "step": 292 }, { "epoch": 0.04, "learning_rate": 2.912115732368897e-06, "loss": 0.0058, "step": 293 }, { "epoch": 0.04, "learning_rate": 2.911754068716094e-06, "loss": 0.0265, "step": 294 }, { "epoch": 0.04, "learning_rate": 2.9113924050632912e-06, "loss": 0.6139, "step": 295 }, { "epoch": 0.04, "learning_rate": 2.9110307414104884e-06, "loss": 0.0071, "step": 296 }, { "epoch": 0.04, "learning_rate": 2.9106690777576857e-06, "loss": 0.5708, "step": 297 }, { "epoch": 0.04, "learning_rate": 2.9103074141048825e-06, "loss": 0.3155, "step": 298 }, { "epoch": 0.04, "learning_rate": 2.9099457504520797e-06, "loss": 0.0633, "step": 299 }, { "epoch": 0.04, "learning_rate": 2.909584086799277e-06, "loss": 0.2406, "step": 300 }, { "epoch": 0.04, "learning_rate": 2.909222423146474e-06, "loss": 0.0057, "step": 301 }, { "epoch": 0.04, "learning_rate": 2.908860759493671e-06, "loss": 0.1354, "step": 302 }, { "epoch": 0.04, "learning_rate": 2.908499095840868e-06, "loss": 0.0387, "step": 303 }, { "epoch": 0.04, "learning_rate": 2.908137432188065e-06, "loss": 0.0141, "step": 304 }, { "epoch": 0.04, "learning_rate": 2.9077757685352623e-06, "loss": 0.035, "step": 305 }, { "epoch": 0.04, "learning_rate": 2.9074141048824595e-06, "loss": 0.5731, "step": 306 }, { "epoch": 0.04, "learning_rate": 2.9070524412296563e-06, "loss": 1.204, "step": 307 }, { "epoch": 0.04, "learning_rate": 2.9066907775768535e-06, "loss": 0.1692, "step": 308 }, { "epoch": 0.04, "learning_rate": 2.9063291139240508e-06, "loss": 0.5332, "step": 309 }, { "epoch": 0.04, "learning_rate": 2.9059674502712476e-06, "loss": 0.26, "step": 310 }, { "epoch": 0.04, "learning_rate": 2.905605786618445e-06, "loss": 0.015, "step": 311 }, { "epoch": 0.04, "learning_rate": 2.905244122965642e-06, "loss": 0.3684, "step": 312 }, { "epoch": 0.04, "learning_rate": 2.9048824593128393e-06, "loss": 0.0255, "step": 313 }, { "epoch": 0.04, "learning_rate": 2.904520795660036e-06, "loss": 0.0085, "step": 314 }, { "epoch": 0.04, "learning_rate": 2.9041591320072333e-06, "loss": 0.0121, "step": 315 }, { "epoch": 0.04, "learning_rate": 2.9037974683544305e-06, "loss": 0.0061, "step": 316 }, { "epoch": 0.04, "learning_rate": 2.9034358047016278e-06, "loss": 0.0158, "step": 317 }, { "epoch": 0.04, "learning_rate": 2.9030741410488246e-06, "loss": 0.2376, "step": 318 }, { "epoch": 0.04, "learning_rate": 2.902712477396022e-06, "loss": 0.2219, "step": 319 }, { "epoch": 0.04, "learning_rate": 2.902350813743219e-06, "loss": 0.1393, "step": 320 }, { "epoch": 0.04, "learning_rate": 2.9019891500904163e-06, "loss": 0.0065, "step": 321 }, { "epoch": 0.04, "learning_rate": 2.901627486437613e-06, "loss": 0.287, "step": 322 }, { "epoch": 0.04, "learning_rate": 2.9012658227848103e-06, "loss": 0.0054, "step": 323 }, { "epoch": 0.04, "learning_rate": 2.9009041591320075e-06, "loss": 0.0087, "step": 324 }, { "epoch": 0.04, "learning_rate": 2.9005424954792043e-06, "loss": 0.3064, "step": 325 }, { "epoch": 0.04, "learning_rate": 2.9001808318264016e-06, "loss": 0.4763, "step": 326 }, { "epoch": 0.04, "learning_rate": 2.8998191681735984e-06, "loss": 0.1457, "step": 327 }, { "epoch": 0.04, "learning_rate": 2.8994575045207956e-06, "loss": 0.7692, "step": 328 }, { "epoch": 0.04, "learning_rate": 2.899095840867993e-06, "loss": 0.004, "step": 329 }, { "epoch": 0.04, "learning_rate": 2.8987341772151896e-06, "loss": 0.0074, "step": 330 }, { "epoch": 0.04, "learning_rate": 2.898372513562387e-06, "loss": 0.007, "step": 331 }, { "epoch": 0.04, "learning_rate": 2.898010849909584e-06, "loss": 0.5415, "step": 332 }, { "epoch": 0.04, "learning_rate": 2.8976491862567813e-06, "loss": 0.0052, "step": 333 }, { "epoch": 0.04, "learning_rate": 2.897287522603978e-06, "loss": 0.5926, "step": 334 }, { "epoch": 0.04, "learning_rate": 2.8969258589511754e-06, "loss": 0.1236, "step": 335 }, { "epoch": 0.04, "learning_rate": 2.8965641952983726e-06, "loss": 0.167, "step": 336 }, { "epoch": 0.04, "learning_rate": 2.89620253164557e-06, "loss": 0.0508, "step": 337 }, { "epoch": 0.04, "learning_rate": 2.8958408679927667e-06, "loss": 0.0474, "step": 338 }, { "epoch": 0.04, "learning_rate": 2.895479204339964e-06, "loss": 0.3562, "step": 339 }, { "epoch": 0.04, "learning_rate": 2.895117540687161e-06, "loss": 0.0306, "step": 340 }, { "epoch": 0.04, "learning_rate": 2.8947558770343583e-06, "loss": 0.0127, "step": 341 }, { "epoch": 0.04, "learning_rate": 2.894394213381555e-06, "loss": 0.3177, "step": 342 }, { "epoch": 0.04, "learning_rate": 2.8940325497287524e-06, "loss": 0.5094, "step": 343 }, { "epoch": 0.04, "learning_rate": 2.8936708860759496e-06, "loss": 0.4139, "step": 344 }, { "epoch": 0.04, "learning_rate": 2.893309222423147e-06, "loss": 0.006, "step": 345 }, { "epoch": 0.04, "learning_rate": 2.8929475587703437e-06, "loss": 0.0061, "step": 346 }, { "epoch": 0.04, "learning_rate": 2.892585895117541e-06, "loss": 0.0779, "step": 347 }, { "epoch": 0.04, "learning_rate": 2.892224231464738e-06, "loss": 0.0056, "step": 348 }, { "epoch": 0.04, "learning_rate": 2.891862567811935e-06, "loss": 0.7973, "step": 349 }, { "epoch": 0.04, "learning_rate": 2.8915009041591317e-06, "loss": 0.0894, "step": 350 }, { "epoch": 0.04, "learning_rate": 2.891139240506329e-06, "loss": 0.0108, "step": 351 }, { "epoch": 0.04, "learning_rate": 2.890777576853526e-06, "loss": 0.0087, "step": 352 }, { "epoch": 0.04, "learning_rate": 2.8904159132007234e-06, "loss": 0.497, "step": 353 }, { "epoch": 0.04, "learning_rate": 2.8900542495479202e-06, "loss": 0.0133, "step": 354 }, { "epoch": 0.04, "learning_rate": 2.8896925858951175e-06, "loss": 0.0069, "step": 355 }, { "epoch": 0.04, "learning_rate": 2.8893309222423147e-06, "loss": 0.03, "step": 356 }, { "epoch": 0.04, "learning_rate": 2.888969258589512e-06, "loss": 0.0044, "step": 357 }, { "epoch": 0.04, "learning_rate": 2.8886075949367087e-06, "loss": 0.6211, "step": 358 }, { "epoch": 0.04, "learning_rate": 2.888245931283906e-06, "loss": 0.6506, "step": 359 }, { "epoch": 0.04, "learning_rate": 2.887884267631103e-06, "loss": 0.0194, "step": 360 }, { "epoch": 0.04, "learning_rate": 2.8875226039783004e-06, "loss": 0.5466, "step": 361 }, { "epoch": 0.04, "learning_rate": 2.8871609403254972e-06, "loss": 0.1724, "step": 362 }, { "epoch": 0.04, "learning_rate": 2.8867992766726945e-06, "loss": 0.0047, "step": 363 }, { "epoch": 0.04, "learning_rate": 2.8864376130198917e-06, "loss": 0.0057, "step": 364 }, { "epoch": 0.04, "learning_rate": 2.886075949367089e-06, "loss": 0.005, "step": 365 }, { "epoch": 0.04, "learning_rate": 2.8857142857142857e-06, "loss": 0.2229, "step": 366 }, { "epoch": 0.04, "learning_rate": 2.885352622061483e-06, "loss": 0.1335, "step": 367 }, { "epoch": 0.04, "learning_rate": 2.88499095840868e-06, "loss": 0.0072, "step": 368 }, { "epoch": 0.04, "learning_rate": 2.8846292947558774e-06, "loss": 0.007, "step": 369 }, { "epoch": 0.04, "learning_rate": 2.8842676311030742e-06, "loss": 0.0066, "step": 370 }, { "epoch": 0.04, "learning_rate": 2.8839059674502715e-06, "loss": 0.0179, "step": 371 }, { "epoch": 0.04, "learning_rate": 2.8835443037974687e-06, "loss": 0.1895, "step": 372 }, { "epoch": 0.04, "learning_rate": 2.8831826401446655e-06, "loss": 0.5474, "step": 373 }, { "epoch": 0.04, "learning_rate": 2.8828209764918623e-06, "loss": 0.0555, "step": 374 }, { "epoch": 0.04, "learning_rate": 2.8824593128390595e-06, "loss": 0.0054, "step": 375 }, { "epoch": 0.05, "learning_rate": 2.8820976491862568e-06, "loss": 0.0205, "step": 376 }, { "epoch": 0.05, "learning_rate": 2.881735985533454e-06, "loss": 0.0094, "step": 377 }, { "epoch": 0.05, "learning_rate": 2.881374321880651e-06, "loss": 0.0049, "step": 378 }, { "epoch": 0.05, "learning_rate": 2.881012658227848e-06, "loss": 0.0037, "step": 379 }, { "epoch": 0.05, "learning_rate": 2.8806509945750453e-06, "loss": 0.0052, "step": 380 }, { "epoch": 0.05, "learning_rate": 2.8802893309222425e-06, "loss": 0.0047, "step": 381 }, { "epoch": 0.05, "learning_rate": 2.8799276672694393e-06, "loss": 0.0057, "step": 382 }, { "epoch": 0.05, "learning_rate": 2.8795660036166365e-06, "loss": 0.0037, "step": 383 }, { "epoch": 0.05, "learning_rate": 2.8792043399638338e-06, "loss": 0.0094, "step": 384 }, { "epoch": 0.05, "learning_rate": 2.878842676311031e-06, "loss": 0.0039, "step": 385 }, { "epoch": 0.05, "learning_rate": 2.878481012658228e-06, "loss": 0.5598, "step": 386 }, { "epoch": 0.05, "learning_rate": 2.878119349005425e-06, "loss": 0.0034, "step": 387 }, { "epoch": 0.05, "learning_rate": 2.8777576853526223e-06, "loss": 0.0543, "step": 388 }, { "epoch": 0.05, "learning_rate": 2.8773960216998195e-06, "loss": 0.3103, "step": 389 }, { "epoch": 0.05, "learning_rate": 2.8770343580470163e-06, "loss": 0.3494, "step": 390 }, { "epoch": 0.05, "learning_rate": 2.8766726943942135e-06, "loss": 0.1869, "step": 391 }, { "epoch": 0.05, "learning_rate": 2.8763110307414108e-06, "loss": 0.0029, "step": 392 }, { "epoch": 0.05, "learning_rate": 2.8759493670886076e-06, "loss": 0.1012, "step": 393 }, { "epoch": 0.05, "learning_rate": 2.875587703435805e-06, "loss": 0.0042, "step": 394 }, { "epoch": 0.05, "learning_rate": 2.875226039783002e-06, "loss": 0.4597, "step": 395 }, { "epoch": 0.05, "learning_rate": 2.8748643761301993e-06, "loss": 0.0043, "step": 396 }, { "epoch": 0.05, "learning_rate": 2.874502712477396e-06, "loss": 0.0035, "step": 397 }, { "epoch": 0.05, "learning_rate": 2.874141048824593e-06, "loss": 0.0031, "step": 398 }, { "epoch": 0.05, "learning_rate": 2.87377938517179e-06, "loss": 1.2048, "step": 399 }, { "epoch": 0.05, "learning_rate": 2.8734177215189874e-06, "loss": 0.0042, "step": 400 }, { "epoch": 0.05, "learning_rate": 2.8730560578661846e-06, "loss": 0.6547, "step": 401 }, { "epoch": 0.05, "learning_rate": 2.8726943942133814e-06, "loss": 0.7885, "step": 402 }, { "epoch": 0.05, "learning_rate": 2.8723327305605786e-06, "loss": 0.0037, "step": 403 }, { "epoch": 0.05, "learning_rate": 2.871971066907776e-06, "loss": 0.0047, "step": 404 }, { "epoch": 0.05, "learning_rate": 2.871609403254973e-06, "loss": 0.1216, "step": 405 }, { "epoch": 0.05, "learning_rate": 2.87124773960217e-06, "loss": 0.0034, "step": 406 }, { "epoch": 0.05, "learning_rate": 2.870886075949367e-06, "loss": 0.0056, "step": 407 }, { "epoch": 0.05, "learning_rate": 2.8705244122965644e-06, "loss": 0.1177, "step": 408 }, { "epoch": 0.05, "learning_rate": 2.870162748643761e-06, "loss": 0.2242, "step": 409 }, { "epoch": 0.05, "learning_rate": 2.8698010849909584e-06, "loss": 0.0034, "step": 410 }, { "epoch": 0.05, "learning_rate": 2.8694394213381556e-06, "loss": 0.003, "step": 411 }, { "epoch": 0.05, "learning_rate": 2.869077757685353e-06, "loss": 0.8431, "step": 412 }, { "epoch": 0.05, "learning_rate": 2.8687160940325497e-06, "loss": 0.5649, "step": 413 }, { "epoch": 0.05, "learning_rate": 2.868354430379747e-06, "loss": 0.032, "step": 414 }, { "epoch": 0.05, "learning_rate": 2.867992766726944e-06, "loss": 0.7653, "step": 415 }, { "epoch": 0.05, "learning_rate": 2.8676311030741414e-06, "loss": 0.0039, "step": 416 }, { "epoch": 0.05, "learning_rate": 2.867269439421338e-06, "loss": 0.0646, "step": 417 }, { "epoch": 0.05, "learning_rate": 2.8669077757685354e-06, "loss": 0.8512, "step": 418 }, { "epoch": 0.05, "learning_rate": 2.8665461121157326e-06, "loss": 0.0328, "step": 419 }, { "epoch": 0.05, "learning_rate": 2.86618444846293e-06, "loss": 0.0034, "step": 420 }, { "epoch": 0.05, "learning_rate": 2.8658227848101267e-06, "loss": 0.0035, "step": 421 }, { "epoch": 0.05, "learning_rate": 2.8654611211573235e-06, "loss": 0.0029, "step": 422 }, { "epoch": 0.05, "learning_rate": 2.8650994575045207e-06, "loss": 0.8072, "step": 423 }, { "epoch": 0.05, "learning_rate": 2.864737793851718e-06, "loss": 0.4187, "step": 424 }, { "epoch": 0.05, "learning_rate": 2.864376130198915e-06, "loss": 1.1613, "step": 425 }, { "epoch": 0.05, "learning_rate": 2.864014466546112e-06, "loss": 0.0057, "step": 426 }, { "epoch": 0.05, "learning_rate": 2.8636528028933092e-06, "loss": 0.0051, "step": 427 }, { "epoch": 0.05, "learning_rate": 2.8632911392405064e-06, "loss": 0.0079, "step": 428 }, { "epoch": 0.05, "learning_rate": 2.8629294755877033e-06, "loss": 0.0028, "step": 429 }, { "epoch": 0.05, "learning_rate": 2.8625678119349005e-06, "loss": 0.3632, "step": 430 }, { "epoch": 0.05, "learning_rate": 2.8622061482820977e-06, "loss": 0.0043, "step": 431 }, { "epoch": 0.05, "learning_rate": 2.861844484629295e-06, "loss": 0.003, "step": 432 }, { "epoch": 0.05, "learning_rate": 2.8614828209764918e-06, "loss": 0.0085, "step": 433 }, { "epoch": 0.05, "learning_rate": 2.861121157323689e-06, "loss": 0.0033, "step": 434 }, { "epoch": 0.05, "learning_rate": 2.8607594936708862e-06, "loss": 0.4541, "step": 435 }, { "epoch": 0.05, "learning_rate": 2.8603978300180834e-06, "loss": 0.3978, "step": 436 }, { "epoch": 0.05, "learning_rate": 2.8600361663652803e-06, "loss": 0.005, "step": 437 }, { "epoch": 0.05, "learning_rate": 2.8596745027124775e-06, "loss": 0.0028, "step": 438 }, { "epoch": 0.05, "learning_rate": 2.8593128390596747e-06, "loss": 0.0047, "step": 439 }, { "epoch": 0.05, "learning_rate": 2.858951175406872e-06, "loss": 0.5185, "step": 440 }, { "epoch": 0.05, "learning_rate": 2.8585895117540688e-06, "loss": 0.0028, "step": 441 }, { "epoch": 0.05, "learning_rate": 2.858227848101266e-06, "loss": 0.0836, "step": 442 }, { "epoch": 0.05, "learning_rate": 2.8578661844484632e-06, "loss": 0.3483, "step": 443 }, { "epoch": 0.05, "learning_rate": 2.8575045207956604e-06, "loss": 0.0634, "step": 444 }, { "epoch": 0.05, "learning_rate": 2.8571428571428573e-06, "loss": 0.0033, "step": 445 }, { "epoch": 0.05, "learning_rate": 2.856781193490054e-06, "loss": 0.0037, "step": 446 }, { "epoch": 0.05, "learning_rate": 2.8564195298372513e-06, "loss": 0.0032, "step": 447 }, { "epoch": 0.05, "learning_rate": 2.8560578661844485e-06, "loss": 0.0028, "step": 448 }, { "epoch": 0.05, "learning_rate": 2.8556962025316453e-06, "loss": 0.0031, "step": 449 }, { "epoch": 0.05, "learning_rate": 2.8553345388788426e-06, "loss": 0.0085, "step": 450 }, { "epoch": 0.05, "learning_rate": 2.85497287522604e-06, "loss": 0.0054, "step": 451 }, { "epoch": 0.05, "learning_rate": 2.854611211573237e-06, "loss": 0.6337, "step": 452 }, { "epoch": 0.05, "learning_rate": 2.854249547920434e-06, "loss": 0.247, "step": 453 }, { "epoch": 0.05, "learning_rate": 2.853887884267631e-06, "loss": 0.0024, "step": 454 }, { "epoch": 0.05, "learning_rate": 2.8535262206148283e-06, "loss": 0.0028, "step": 455 }, { "epoch": 0.05, "learning_rate": 2.8531645569620255e-06, "loss": 0.0097, "step": 456 }, { "epoch": 0.05, "learning_rate": 2.8528028933092223e-06, "loss": 0.6322, "step": 457 }, { "epoch": 0.05, "learning_rate": 2.8524412296564196e-06, "loss": 0.0608, "step": 458 }, { "epoch": 0.06, "learning_rate": 2.852079566003617e-06, "loss": 0.0027, "step": 459 }, { "epoch": 0.06, "learning_rate": 2.851717902350814e-06, "loss": 0.0027, "step": 460 }, { "epoch": 0.06, "learning_rate": 2.851356238698011e-06, "loss": 0.0034, "step": 461 }, { "epoch": 0.06, "learning_rate": 2.850994575045208e-06, "loss": 0.0133, "step": 462 }, { "epoch": 0.06, "learning_rate": 2.8506329113924053e-06, "loss": 0.0028, "step": 463 }, { "epoch": 0.06, "learning_rate": 2.8502712477396025e-06, "loss": 0.0037, "step": 464 }, { "epoch": 0.06, "learning_rate": 2.8499095840867993e-06, "loss": 0.0053, "step": 465 }, { "epoch": 0.06, "learning_rate": 2.8495479204339966e-06, "loss": 0.0117, "step": 466 }, { "epoch": 0.06, "learning_rate": 2.849186256781194e-06, "loss": 0.003, "step": 467 }, { "epoch": 0.06, "learning_rate": 2.848824593128391e-06, "loss": 0.0116, "step": 468 }, { "epoch": 0.06, "learning_rate": 2.8484629294755874e-06, "loss": 0.0036, "step": 469 }, { "epoch": 0.06, "learning_rate": 2.8481012658227846e-06, "loss": 0.0024, "step": 470 }, { "epoch": 0.06, "learning_rate": 2.847739602169982e-06, "loss": 0.0024, "step": 471 }, { "epoch": 0.06, "learning_rate": 2.847377938517179e-06, "loss": 0.0055, "step": 472 }, { "epoch": 0.06, "learning_rate": 2.847016274864376e-06, "loss": 0.3293, "step": 473 }, { "epoch": 0.06, "learning_rate": 2.846654611211573e-06, "loss": 0.0023, "step": 474 }, { "epoch": 0.06, "learning_rate": 2.8462929475587704e-06, "loss": 0.0044, "step": 475 }, { "epoch": 0.06, "learning_rate": 2.8459312839059676e-06, "loss": 0.0022, "step": 476 }, { "epoch": 0.06, "learning_rate": 2.8455696202531644e-06, "loss": 0.0021, "step": 477 }, { "epoch": 0.06, "learning_rate": 2.8452079566003616e-06, "loss": 0.0029, "step": 478 }, { "epoch": 0.06, "learning_rate": 2.844846292947559e-06, "loss": 0.6968, "step": 479 }, { "epoch": 0.06, "learning_rate": 2.844484629294756e-06, "loss": 0.0143, "step": 480 }, { "epoch": 0.06, "learning_rate": 2.844122965641953e-06, "loss": 0.0023, "step": 481 }, { "epoch": 0.06, "learning_rate": 2.84376130198915e-06, "loss": 0.0021, "step": 482 }, { "epoch": 0.06, "learning_rate": 2.8433996383363474e-06, "loss": 0.0036, "step": 483 }, { "epoch": 0.06, "learning_rate": 2.8430379746835446e-06, "loss": 0.002, "step": 484 }, { "epoch": 0.06, "learning_rate": 2.8426763110307414e-06, "loss": 0.0022, "step": 485 }, { "epoch": 0.06, "learning_rate": 2.8423146473779387e-06, "loss": 0.0032, "step": 486 }, { "epoch": 0.06, "learning_rate": 2.841952983725136e-06, "loss": 0.0039, "step": 487 }, { "epoch": 0.06, "learning_rate": 2.841591320072333e-06, "loss": 0.004, "step": 488 }, { "epoch": 0.06, "learning_rate": 2.84122965641953e-06, "loss": 0.0567, "step": 489 }, { "epoch": 0.06, "learning_rate": 2.840867992766727e-06, "loss": 0.2362, "step": 490 }, { "epoch": 0.06, "learning_rate": 2.8405063291139244e-06, "loss": 0.0026, "step": 491 }, { "epoch": 0.06, "learning_rate": 2.840144665461121e-06, "loss": 0.0022, "step": 492 }, { "epoch": 0.06, "learning_rate": 2.839783001808318e-06, "loss": 0.0022, "step": 493 }, { "epoch": 0.06, "learning_rate": 2.8394213381555152e-06, "loss": 0.003, "step": 494 }, { "epoch": 0.06, "learning_rate": 2.8390596745027125e-06, "loss": 0.0029, "step": 495 }, { "epoch": 0.06, "learning_rate": 2.8386980108499097e-06, "loss": 0.0021, "step": 496 }, { "epoch": 0.06, "learning_rate": 2.8383363471971065e-06, "loss": 0.101, "step": 497 }, { "epoch": 0.06, "learning_rate": 2.8379746835443037e-06, "loss": 0.7092, "step": 498 }, { "epoch": 0.06, "learning_rate": 2.837613019891501e-06, "loss": 0.0028, "step": 499 }, { "epoch": 0.06, "learning_rate": 2.837251356238698e-06, "loss": 0.0017, "step": 500 }, { "epoch": 0.06, "learning_rate": 2.836889692585895e-06, "loss": 0.109, "step": 501 }, { "epoch": 0.06, "learning_rate": 2.8365280289330922e-06, "loss": 0.0099, "step": 502 }, { "epoch": 0.06, "learning_rate": 2.8361663652802895e-06, "loss": 0.0031, "step": 503 }, { "epoch": 0.06, "learning_rate": 2.8358047016274867e-06, "loss": 0.3253, "step": 504 }, { "epoch": 0.06, "learning_rate": 2.8354430379746835e-06, "loss": 0.0019, "step": 505 }, { "epoch": 0.06, "learning_rate": 2.8350813743218807e-06, "loss": 0.0018, "step": 506 }, { "epoch": 0.06, "learning_rate": 2.834719710669078e-06, "loss": 1.1213, "step": 507 }, { "epoch": 0.06, "learning_rate": 2.834358047016275e-06, "loss": 0.0016, "step": 508 }, { "epoch": 0.06, "learning_rate": 2.833996383363472e-06, "loss": 0.2754, "step": 509 }, { "epoch": 0.06, "learning_rate": 2.8336347197106692e-06, "loss": 0.3531, "step": 510 }, { "epoch": 0.06, "learning_rate": 2.8332730560578665e-06, "loss": 0.003, "step": 511 }, { "epoch": 0.06, "learning_rate": 2.8329113924050633e-06, "loss": 0.0017, "step": 512 }, { "epoch": 0.06, "learning_rate": 2.8325497287522605e-06, "loss": 0.0378, "step": 513 }, { "epoch": 0.06, "learning_rate": 2.8321880650994577e-06, "loss": 0.0017, "step": 514 }, { "epoch": 0.06, "learning_rate": 2.831826401446655e-06, "loss": 0.2998, "step": 515 }, { "epoch": 0.06, "learning_rate": 2.8314647377938518e-06, "loss": 0.0486, "step": 516 }, { "epoch": 0.06, "learning_rate": 2.8311030741410486e-06, "loss": 0.0018, "step": 517 }, { "epoch": 0.06, "learning_rate": 2.830741410488246e-06, "loss": 0.0051, "step": 518 }, { "epoch": 0.06, "learning_rate": 2.830379746835443e-06, "loss": 0.0017, "step": 519 }, { "epoch": 0.06, "learning_rate": 2.8300180831826403e-06, "loss": 2.0192, "step": 520 }, { "epoch": 0.06, "learning_rate": 2.829656419529837e-06, "loss": 0.2164, "step": 521 }, { "epoch": 0.06, "learning_rate": 2.8292947558770343e-06, "loss": 0.7646, "step": 522 }, { "epoch": 0.06, "learning_rate": 2.8289330922242315e-06, "loss": 0.0016, "step": 523 }, { "epoch": 0.06, "learning_rate": 2.8285714285714288e-06, "loss": 0.004, "step": 524 }, { "epoch": 0.06, "learning_rate": 2.8282097649186256e-06, "loss": 0.0014, "step": 525 }, { "epoch": 0.06, "learning_rate": 2.827848101265823e-06, "loss": 0.0024, "step": 526 }, { "epoch": 0.06, "learning_rate": 2.82748643761302e-06, "loss": 0.0012, "step": 527 }, { "epoch": 0.06, "learning_rate": 2.8271247739602173e-06, "loss": 0.4217, "step": 528 }, { "epoch": 0.06, "learning_rate": 2.826763110307414e-06, "loss": 0.0045, "step": 529 }, { "epoch": 0.06, "learning_rate": 2.8264014466546113e-06, "loss": 0.0354, "step": 530 }, { "epoch": 0.06, "learning_rate": 2.8260397830018085e-06, "loss": 0.0038, "step": 531 }, { "epoch": 0.06, "learning_rate": 2.8256781193490054e-06, "loss": 0.0017, "step": 532 }, { "epoch": 0.06, "learning_rate": 2.8253164556962026e-06, "loss": 0.0017, "step": 533 }, { "epoch": 0.06, "learning_rate": 2.8249547920434e-06, "loss": 0.0015, "step": 534 }, { "epoch": 0.06, "learning_rate": 2.824593128390597e-06, "loss": 0.9161, "step": 535 }, { "epoch": 0.06, "learning_rate": 2.824231464737794e-06, "loss": 0.0014, "step": 536 }, { "epoch": 0.06, "learning_rate": 2.823869801084991e-06, "loss": 0.0024, "step": 537 }, { "epoch": 0.06, "learning_rate": 2.8235081374321883e-06, "loss": 0.0018, "step": 538 }, { "epoch": 0.06, "learning_rate": 2.8231464737793855e-06, "loss": 0.0018, "step": 539 }, { "epoch": 0.06, "learning_rate": 2.8227848101265824e-06, "loss": 0.0112, "step": 540 }, { "epoch": 0.06, "learning_rate": 2.822423146473779e-06, "loss": 0.0014, "step": 541 }, { "epoch": 0.06, "learning_rate": 2.8220614828209764e-06, "loss": 0.6375, "step": 542 }, { "epoch": 0.07, "learning_rate": 2.8216998191681736e-06, "loss": 0.7348, "step": 543 }, { "epoch": 0.07, "learning_rate": 2.821338155515371e-06, "loss": 0.0017, "step": 544 }, { "epoch": 0.07, "learning_rate": 2.8209764918625677e-06, "loss": 0.0029, "step": 545 }, { "epoch": 0.07, "learning_rate": 2.820614828209765e-06, "loss": 0.0017, "step": 546 }, { "epoch": 0.07, "learning_rate": 2.820253164556962e-06, "loss": 0.0045, "step": 547 }, { "epoch": 0.07, "learning_rate": 2.8198915009041594e-06, "loss": 0.0024, "step": 548 }, { "epoch": 0.07, "learning_rate": 2.819529837251356e-06, "loss": 0.0065, "step": 549 }, { "epoch": 0.07, "learning_rate": 2.8191681735985534e-06, "loss": 0.0024, "step": 550 }, { "epoch": 0.07, "learning_rate": 2.8188065099457506e-06, "loss": 0.7615, "step": 551 }, { "epoch": 0.07, "learning_rate": 2.8184448462929474e-06, "loss": 0.7474, "step": 552 }, { "epoch": 0.07, "learning_rate": 2.8180831826401447e-06, "loss": 0.0025, "step": 553 }, { "epoch": 0.07, "learning_rate": 2.817721518987342e-06, "loss": 0.002, "step": 554 }, { "epoch": 0.07, "learning_rate": 2.817359855334539e-06, "loss": 0.0017, "step": 555 }, { "epoch": 0.07, "learning_rate": 2.816998191681736e-06, "loss": 0.0022, "step": 556 }, { "epoch": 0.07, "learning_rate": 2.816636528028933e-06, "loss": 0.0024, "step": 557 }, { "epoch": 0.07, "learning_rate": 2.8162748643761304e-06, "loss": 0.0023, "step": 558 }, { "epoch": 0.07, "learning_rate": 2.8159132007233276e-06, "loss": 0.8247, "step": 559 }, { "epoch": 0.07, "learning_rate": 2.8155515370705244e-06, "loss": 0.1295, "step": 560 }, { "epoch": 0.07, "learning_rate": 2.8151898734177217e-06, "loss": 0.004, "step": 561 }, { "epoch": 0.07, "learning_rate": 2.814828209764919e-06, "loss": 0.0018, "step": 562 }, { "epoch": 0.07, "learning_rate": 2.814466546112116e-06, "loss": 0.7272, "step": 563 }, { "epoch": 0.07, "learning_rate": 2.814104882459313e-06, "loss": 0.1324, "step": 564 }, { "epoch": 0.07, "learning_rate": 2.8137432188065097e-06, "loss": 0.0061, "step": 565 }, { "epoch": 0.07, "learning_rate": 2.813381555153707e-06, "loss": 0.0013, "step": 566 }, { "epoch": 0.07, "learning_rate": 2.813019891500904e-06, "loss": 0.4842, "step": 567 }, { "epoch": 0.07, "learning_rate": 2.812658227848101e-06, "loss": 0.002, "step": 568 }, { "epoch": 0.07, "learning_rate": 2.8122965641952982e-06, "loss": 0.002, "step": 569 }, { "epoch": 0.07, "learning_rate": 2.8119349005424955e-06, "loss": 0.0017, "step": 570 }, { "epoch": 0.07, "learning_rate": 2.8115732368896927e-06, "loss": 0.0017, "step": 571 }, { "epoch": 0.07, "learning_rate": 2.8112115732368895e-06, "loss": 0.0014, "step": 572 }, { "epoch": 0.07, "learning_rate": 2.8108499095840868e-06, "loss": 0.0019, "step": 573 }, { "epoch": 0.07, "learning_rate": 2.810488245931284e-06, "loss": 0.0018, "step": 574 }, { "epoch": 0.07, "learning_rate": 2.8101265822784812e-06, "loss": 0.0031, "step": 575 }, { "epoch": 0.07, "learning_rate": 2.809764918625678e-06, "loss": 0.0024, "step": 576 }, { "epoch": 0.07, "learning_rate": 2.8094032549728753e-06, "loss": 0.1583, "step": 577 }, { "epoch": 0.07, "learning_rate": 2.8090415913200725e-06, "loss": 0.2637, "step": 578 }, { "epoch": 0.07, "learning_rate": 2.8086799276672697e-06, "loss": 0.0024, "step": 579 }, { "epoch": 0.07, "learning_rate": 2.8083182640144665e-06, "loss": 0.0017, "step": 580 }, { "epoch": 0.07, "learning_rate": 2.8079566003616638e-06, "loss": 0.0016, "step": 581 }, { "epoch": 0.07, "learning_rate": 2.807594936708861e-06, "loss": 0.0014, "step": 582 }, { "epoch": 0.07, "learning_rate": 2.8072332730560582e-06, "loss": 0.0019, "step": 583 }, { "epoch": 0.07, "learning_rate": 2.806871609403255e-06, "loss": 0.0017, "step": 584 }, { "epoch": 0.07, "learning_rate": 2.8065099457504523e-06, "loss": 0.0045, "step": 585 }, { "epoch": 0.07, "learning_rate": 2.8061482820976495e-06, "loss": 0.0036, "step": 586 }, { "epoch": 0.07, "learning_rate": 2.8057866184448467e-06, "loss": 0.7185, "step": 587 }, { "epoch": 0.07, "learning_rate": 2.805424954792043e-06, "loss": 0.0017, "step": 588 }, { "epoch": 0.07, "learning_rate": 2.8050632911392403e-06, "loss": 0.4729, "step": 589 }, { "epoch": 0.07, "learning_rate": 2.8047016274864376e-06, "loss": 0.002, "step": 590 }, { "epoch": 0.07, "learning_rate": 2.804339963833635e-06, "loss": 0.0016, "step": 591 }, { "epoch": 0.07, "learning_rate": 2.8039783001808316e-06, "loss": 0.0018, "step": 592 }, { "epoch": 0.07, "learning_rate": 2.803616636528029e-06, "loss": 0.0012, "step": 593 }, { "epoch": 0.07, "learning_rate": 2.803254972875226e-06, "loss": 0.0014, "step": 594 }, { "epoch": 0.07, "learning_rate": 2.8028933092224233e-06, "loss": 0.0015, "step": 595 }, { "epoch": 0.07, "learning_rate": 2.80253164556962e-06, "loss": 0.3357, "step": 596 }, { "epoch": 0.07, "learning_rate": 2.8021699819168173e-06, "loss": 0.7223, "step": 597 }, { "epoch": 0.07, "learning_rate": 2.8018083182640146e-06, "loss": 0.0029, "step": 598 }, { "epoch": 0.07, "learning_rate": 2.801446654611212e-06, "loss": 0.016, "step": 599 }, { "epoch": 0.07, "learning_rate": 2.8010849909584086e-06, "loss": 0.0369, "step": 600 }, { "epoch": 0.07, "learning_rate": 2.800723327305606e-06, "loss": 0.002, "step": 601 }, { "epoch": 0.07, "learning_rate": 2.800361663652803e-06, "loss": 0.415, "step": 602 }, { "epoch": 0.07, "learning_rate": 2.8000000000000003e-06, "loss": 0.0015, "step": 603 }, { "epoch": 0.07, "learning_rate": 2.799638336347197e-06, "loss": 0.0014, "step": 604 }, { "epoch": 0.07, "learning_rate": 2.7992766726943943e-06, "loss": 0.0014, "step": 605 }, { "epoch": 0.07, "learning_rate": 2.7989150090415916e-06, "loss": 0.2792, "step": 606 }, { "epoch": 0.07, "learning_rate": 2.798553345388789e-06, "loss": 0.023, "step": 607 }, { "epoch": 0.07, "learning_rate": 2.7981916817359856e-06, "loss": 0.0019, "step": 608 }, { "epoch": 0.07, "learning_rate": 2.797830018083183e-06, "loss": 0.0889, "step": 609 }, { "epoch": 0.07, "learning_rate": 2.79746835443038e-06, "loss": 0.0012, "step": 610 }, { "epoch": 0.07, "learning_rate": 2.797106690777577e-06, "loss": 0.0017, "step": 611 }, { "epoch": 0.07, "learning_rate": 2.7967450271247737e-06, "loss": 0.0017, "step": 612 }, { "epoch": 0.07, "learning_rate": 2.796383363471971e-06, "loss": 0.0018, "step": 613 }, { "epoch": 0.07, "learning_rate": 2.796021699819168e-06, "loss": 0.0014, "step": 614 }, { "epoch": 0.07, "learning_rate": 2.7956600361663654e-06, "loss": 0.0012, "step": 615 }, { "epoch": 0.07, "learning_rate": 2.795298372513562e-06, "loss": 0.0013, "step": 616 }, { "epoch": 0.07, "learning_rate": 2.7949367088607594e-06, "loss": 0.8109, "step": 617 }, { "epoch": 0.07, "learning_rate": 2.7945750452079566e-06, "loss": 0.0043, "step": 618 }, { "epoch": 0.07, "learning_rate": 2.794213381555154e-06, "loss": 0.0013, "step": 619 }, { "epoch": 0.07, "learning_rate": 2.7938517179023507e-06, "loss": 0.5301, "step": 620 }, { "epoch": 0.07, "learning_rate": 2.793490054249548e-06, "loss": 0.0012, "step": 621 }, { "epoch": 0.07, "learning_rate": 2.793128390596745e-06, "loss": 0.0052, "step": 622 }, { "epoch": 0.07, "learning_rate": 2.7927667269439424e-06, "loss": 0.0331, "step": 623 }, { "epoch": 0.07, "learning_rate": 2.792405063291139e-06, "loss": 0.4297, "step": 624 }, { "epoch": 0.07, "learning_rate": 2.7920433996383364e-06, "loss": 0.0015, "step": 625 }, { "epoch": 0.08, "learning_rate": 2.7916817359855336e-06, "loss": 0.0012, "step": 626 }, { "epoch": 0.08, "learning_rate": 2.791320072332731e-06, "loss": 0.0014, "step": 627 }, { "epoch": 0.08, "learning_rate": 2.7909584086799277e-06, "loss": 0.0014, "step": 628 }, { "epoch": 0.08, "learning_rate": 2.790596745027125e-06, "loss": 0.0034, "step": 629 }, { "epoch": 0.08, "learning_rate": 2.790235081374322e-06, "loss": 0.8178, "step": 630 }, { "epoch": 0.08, "learning_rate": 2.789873417721519e-06, "loss": 0.0844, "step": 631 }, { "epoch": 0.08, "learning_rate": 2.789511754068716e-06, "loss": 0.0016, "step": 632 }, { "epoch": 0.08, "learning_rate": 2.7891500904159134e-06, "loss": 0.0013, "step": 633 }, { "epoch": 0.08, "learning_rate": 2.7887884267631107e-06, "loss": 0.0012, "step": 634 }, { "epoch": 0.08, "learning_rate": 2.7884267631103075e-06, "loss": 0.0015, "step": 635 }, { "epoch": 0.08, "learning_rate": 2.7880650994575043e-06, "loss": 0.0018, "step": 636 }, { "epoch": 0.08, "learning_rate": 2.7877034358047015e-06, "loss": 0.0013, "step": 637 }, { "epoch": 0.08, "learning_rate": 2.7873417721518987e-06, "loss": 0.0277, "step": 638 }, { "epoch": 0.08, "learning_rate": 2.786980108499096e-06, "loss": 0.0013, "step": 639 }, { "epoch": 0.08, "learning_rate": 2.7866184448462928e-06, "loss": 0.0016, "step": 640 }, { "epoch": 0.08, "learning_rate": 2.78625678119349e-06, "loss": 0.0015, "step": 641 }, { "epoch": 0.08, "learning_rate": 2.7858951175406872e-06, "loss": 0.1892, "step": 642 }, { "epoch": 0.08, "learning_rate": 2.7855334538878845e-06, "loss": 0.7164, "step": 643 }, { "epoch": 0.08, "learning_rate": 2.7851717902350813e-06, "loss": 0.0017, "step": 644 }, { "epoch": 0.08, "learning_rate": 2.7848101265822785e-06, "loss": 0.0015, "step": 645 }, { "epoch": 0.08, "learning_rate": 2.7844484629294757e-06, "loss": 0.0029, "step": 646 }, { "epoch": 0.08, "learning_rate": 2.784086799276673e-06, "loss": 0.0055, "step": 647 }, { "epoch": 0.08, "learning_rate": 2.7837251356238698e-06, "loss": 0.0016, "step": 648 }, { "epoch": 0.08, "learning_rate": 2.783363471971067e-06, "loss": 0.0013, "step": 649 }, { "epoch": 0.08, "learning_rate": 2.7830018083182642e-06, "loss": 0.3424, "step": 650 }, { "epoch": 0.08, "learning_rate": 2.782640144665461e-06, "loss": 0.002, "step": 651 }, { "epoch": 0.08, "learning_rate": 2.7822784810126583e-06, "loss": 0.0014, "step": 652 }, { "epoch": 0.08, "learning_rate": 2.7819168173598555e-06, "loss": 0.3674, "step": 653 }, { "epoch": 0.08, "learning_rate": 2.7815551537070527e-06, "loss": 0.1662, "step": 654 }, { "epoch": 0.08, "learning_rate": 2.7811934900542495e-06, "loss": 0.7966, "step": 655 }, { "epoch": 0.08, "learning_rate": 2.7808318264014468e-06, "loss": 0.0082, "step": 656 }, { "epoch": 0.08, "learning_rate": 2.780470162748644e-06, "loss": 0.0077, "step": 657 }, { "epoch": 0.08, "learning_rate": 2.7801084990958412e-06, "loss": 0.8501, "step": 658 }, { "epoch": 0.08, "learning_rate": 2.779746835443038e-06, "loss": 0.0013, "step": 659 }, { "epoch": 0.08, "learning_rate": 2.779385171790235e-06, "loss": 0.0015, "step": 660 }, { "epoch": 0.08, "learning_rate": 2.779023508137432e-06, "loss": 0.0015, "step": 661 }, { "epoch": 0.08, "learning_rate": 2.7786618444846293e-06, "loss": 0.0016, "step": 662 }, { "epoch": 0.08, "learning_rate": 2.7783001808318265e-06, "loss": 0.0019, "step": 663 }, { "epoch": 0.08, "learning_rate": 2.7779385171790234e-06, "loss": 0.0017, "step": 664 }, { "epoch": 0.08, "learning_rate": 2.7775768535262206e-06, "loss": 0.0016, "step": 665 }, { "epoch": 0.08, "learning_rate": 2.777215189873418e-06, "loss": 0.0017, "step": 666 }, { "epoch": 0.08, "learning_rate": 2.776853526220615e-06, "loss": 0.0012, "step": 667 }, { "epoch": 0.08, "learning_rate": 2.776491862567812e-06, "loss": 0.4939, "step": 668 }, { "epoch": 0.08, "learning_rate": 2.776130198915009e-06, "loss": 0.7207, "step": 669 }, { "epoch": 0.08, "learning_rate": 2.7757685352622063e-06, "loss": 0.009, "step": 670 }, { "epoch": 0.08, "learning_rate": 2.775406871609403e-06, "loss": 0.5574, "step": 671 }, { "epoch": 0.08, "learning_rate": 2.7750452079566004e-06, "loss": 0.0021, "step": 672 }, { "epoch": 0.08, "learning_rate": 2.7746835443037976e-06, "loss": 0.0171, "step": 673 }, { "epoch": 0.08, "learning_rate": 2.774321880650995e-06, "loss": 0.6123, "step": 674 }, { "epoch": 0.08, "learning_rate": 2.7739602169981916e-06, "loss": 0.0292, "step": 675 }, { "epoch": 0.08, "learning_rate": 2.773598553345389e-06, "loss": 0.1652, "step": 676 }, { "epoch": 0.08, "learning_rate": 2.773236889692586e-06, "loss": 0.0039, "step": 677 }, { "epoch": 0.08, "learning_rate": 2.7728752260397833e-06, "loss": 0.0013, "step": 678 }, { "epoch": 0.08, "learning_rate": 2.77251356238698e-06, "loss": 0.0054, "step": 679 }, { "epoch": 0.08, "learning_rate": 2.7721518987341774e-06, "loss": 0.0017, "step": 680 }, { "epoch": 0.08, "learning_rate": 2.7717902350813746e-06, "loss": 0.0014, "step": 681 }, { "epoch": 0.08, "learning_rate": 2.771428571428572e-06, "loss": 0.0019, "step": 682 }, { "epoch": 0.08, "learning_rate": 2.7710669077757686e-06, "loss": 0.0011, "step": 683 }, { "epoch": 0.08, "learning_rate": 2.7707052441229654e-06, "loss": 0.0012, "step": 684 }, { "epoch": 0.08, "learning_rate": 2.7703435804701627e-06, "loss": 0.0016, "step": 685 }, { "epoch": 0.08, "learning_rate": 2.76998191681736e-06, "loss": 0.0017, "step": 686 }, { "epoch": 0.08, "learning_rate": 2.769620253164557e-06, "loss": 0.0018, "step": 687 }, { "epoch": 0.08, "learning_rate": 2.769258589511754e-06, "loss": 0.0012, "step": 688 }, { "epoch": 0.08, "learning_rate": 2.768896925858951e-06, "loss": 0.0012, "step": 689 }, { "epoch": 0.08, "learning_rate": 2.7685352622061484e-06, "loss": 0.0012, "step": 690 }, { "epoch": 0.08, "learning_rate": 2.768173598553345e-06, "loss": 0.8812, "step": 691 }, { "epoch": 0.08, "learning_rate": 2.7678119349005424e-06, "loss": 0.7634, "step": 692 }, { "epoch": 0.08, "learning_rate": 2.7674502712477397e-06, "loss": 0.0019, "step": 693 }, { "epoch": 0.08, "learning_rate": 2.767088607594937e-06, "loss": 0.723, "step": 694 }, { "epoch": 0.08, "learning_rate": 2.7667269439421337e-06, "loss": 0.0033, "step": 695 }, { "epoch": 0.08, "learning_rate": 2.766365280289331e-06, "loss": 0.0015, "step": 696 }, { "epoch": 0.08, "learning_rate": 2.766003616636528e-06, "loss": 0.0015, "step": 697 }, { "epoch": 0.08, "learning_rate": 2.7656419529837254e-06, "loss": 0.0073, "step": 698 }, { "epoch": 0.08, "learning_rate": 2.765280289330922e-06, "loss": 0.6803, "step": 699 }, { "epoch": 0.08, "learning_rate": 2.7649186256781194e-06, "loss": 0.0016, "step": 700 }, { "epoch": 0.08, "learning_rate": 2.7645569620253167e-06, "loss": 0.0016, "step": 701 }, { "epoch": 0.08, "learning_rate": 2.764195298372514e-06, "loss": 0.9066, "step": 702 }, { "epoch": 0.08, "learning_rate": 2.7638336347197107e-06, "loss": 0.3678, "step": 703 }, { "epoch": 0.08, "learning_rate": 2.763471971066908e-06, "loss": 0.7641, "step": 704 }, { "epoch": 0.08, "learning_rate": 2.763110307414105e-06, "loss": 1.4345, "step": 705 }, { "epoch": 0.08, "learning_rate": 2.7627486437613024e-06, "loss": 0.0012, "step": 706 }, { "epoch": 0.08, "learning_rate": 2.762386980108499e-06, "loss": 0.002, "step": 707 }, { "epoch": 0.08, "learning_rate": 2.762025316455696e-06, "loss": 0.0015, "step": 708 }, { "epoch": 0.08, "learning_rate": 2.7616636528028932e-06, "loss": 0.0014, "step": 709 }, { "epoch": 0.09, "learning_rate": 2.7613019891500905e-06, "loss": 0.0027, "step": 710 }, { "epoch": 0.09, "learning_rate": 2.7609403254972873e-06, "loss": 0.0017, "step": 711 }, { "epoch": 0.09, "learning_rate": 2.7605786618444845e-06, "loss": 0.002, "step": 712 }, { "epoch": 0.09, "learning_rate": 2.7602169981916817e-06, "loss": 0.0052, "step": 713 }, { "epoch": 0.09, "learning_rate": 2.759855334538879e-06, "loss": 0.0013, "step": 714 }, { "epoch": 0.09, "learning_rate": 2.7594936708860758e-06, "loss": 0.0016, "step": 715 }, { "epoch": 0.09, "learning_rate": 2.759132007233273e-06, "loss": 0.0774, "step": 716 }, { "epoch": 0.09, "learning_rate": 2.7587703435804702e-06, "loss": 0.0018, "step": 717 }, { "epoch": 0.09, "learning_rate": 2.7584086799276675e-06, "loss": 0.6517, "step": 718 }, { "epoch": 0.09, "learning_rate": 2.7580470162748643e-06, "loss": 0.0021, "step": 719 }, { "epoch": 0.09, "learning_rate": 2.7576853526220615e-06, "loss": 0.0193, "step": 720 }, { "epoch": 0.09, "learning_rate": 2.7573236889692588e-06, "loss": 0.022, "step": 721 }, { "epoch": 0.09, "learning_rate": 2.756962025316456e-06, "loss": 0.025, "step": 722 }, { "epoch": 0.09, "learning_rate": 2.756600361663653e-06, "loss": 0.0016, "step": 723 }, { "epoch": 0.09, "learning_rate": 2.75623869801085e-06, "loss": 0.0063, "step": 724 }, { "epoch": 0.09, "learning_rate": 2.7558770343580473e-06, "loss": 0.723, "step": 725 }, { "epoch": 0.09, "learning_rate": 2.7555153707052445e-06, "loss": 0.0272, "step": 726 }, { "epoch": 0.09, "learning_rate": 2.7551537070524413e-06, "loss": 0.1946, "step": 727 }, { "epoch": 0.09, "learning_rate": 2.7547920433996385e-06, "loss": 0.0019, "step": 728 }, { "epoch": 0.09, "learning_rate": 2.7544303797468358e-06, "loss": 0.771, "step": 729 }, { "epoch": 0.09, "learning_rate": 2.7540687160940326e-06, "loss": 0.0019, "step": 730 }, { "epoch": 0.09, "learning_rate": 2.75370705244123e-06, "loss": 0.0041, "step": 731 }, { "epoch": 0.09, "learning_rate": 2.7533453887884266e-06, "loss": 0.006, "step": 732 }, { "epoch": 0.09, "learning_rate": 2.752983725135624e-06, "loss": 0.6061, "step": 733 }, { "epoch": 0.09, "learning_rate": 2.752622061482821e-06, "loss": 0.008, "step": 734 }, { "epoch": 0.09, "learning_rate": 2.752260397830018e-06, "loss": 0.0016, "step": 735 }, { "epoch": 0.09, "learning_rate": 2.751898734177215e-06, "loss": 0.5488, "step": 736 }, { "epoch": 0.09, "learning_rate": 2.7515370705244123e-06, "loss": 0.0259, "step": 737 }, { "epoch": 0.09, "learning_rate": 2.7511754068716096e-06, "loss": 0.8052, "step": 738 }, { "epoch": 0.09, "learning_rate": 2.7508137432188064e-06, "loss": 0.0084, "step": 739 }, { "epoch": 0.09, "learning_rate": 2.7504520795660036e-06, "loss": 0.0017, "step": 740 }, { "epoch": 0.09, "learning_rate": 2.750090415913201e-06, "loss": 0.8191, "step": 741 }, { "epoch": 0.09, "learning_rate": 2.749728752260398e-06, "loss": 0.0072, "step": 742 }, { "epoch": 0.09, "learning_rate": 2.749367088607595e-06, "loss": 0.0044, "step": 743 }, { "epoch": 0.09, "learning_rate": 2.749005424954792e-06, "loss": 0.0073, "step": 744 }, { "epoch": 0.09, "learning_rate": 2.7486437613019893e-06, "loss": 0.0014, "step": 745 }, { "epoch": 0.09, "learning_rate": 2.7482820976491866e-06, "loss": 0.0019, "step": 746 }, { "epoch": 0.09, "learning_rate": 2.7479204339963834e-06, "loss": 0.0013, "step": 747 }, { "epoch": 0.09, "learning_rate": 2.7475587703435806e-06, "loss": 0.0014, "step": 748 }, { "epoch": 0.09, "learning_rate": 2.747197106690778e-06, "loss": 0.0352, "step": 749 }, { "epoch": 0.09, "learning_rate": 2.7468354430379746e-06, "loss": 0.7246, "step": 750 }, { "epoch": 0.09, "learning_rate": 2.746473779385172e-06, "loss": 0.0043, "step": 751 }, { "epoch": 0.09, "learning_rate": 2.746112115732369e-06, "loss": 0.0014, "step": 752 }, { "epoch": 0.09, "learning_rate": 2.7457504520795663e-06, "loss": 0.0016, "step": 753 }, { "epoch": 0.09, "learning_rate": 2.745388788426763e-06, "loss": 0.0017, "step": 754 }, { "epoch": 0.09, "learning_rate": 2.7450271247739604e-06, "loss": 0.0016, "step": 755 }, { "epoch": 0.09, "learning_rate": 2.744665461121157e-06, "loss": 0.0014, "step": 756 }, { "epoch": 0.09, "learning_rate": 2.7443037974683544e-06, "loss": 0.0054, "step": 757 }, { "epoch": 0.09, "learning_rate": 2.7439421338155516e-06, "loss": 0.7638, "step": 758 }, { "epoch": 0.09, "learning_rate": 2.7435804701627485e-06, "loss": 0.0016, "step": 759 }, { "epoch": 0.09, "learning_rate": 2.7432188065099457e-06, "loss": 0.513, "step": 760 }, { "epoch": 0.09, "learning_rate": 2.742857142857143e-06, "loss": 0.0021, "step": 761 }, { "epoch": 0.09, "learning_rate": 2.74249547920434e-06, "loss": 0.6245, "step": 762 }, { "epoch": 0.09, "learning_rate": 2.742133815551537e-06, "loss": 0.0017, "step": 763 }, { "epoch": 0.09, "learning_rate": 2.741772151898734e-06, "loss": 0.0013, "step": 764 }, { "epoch": 0.09, "learning_rate": 2.7414104882459314e-06, "loss": 0.0256, "step": 765 }, { "epoch": 0.09, "learning_rate": 2.7410488245931286e-06, "loss": 0.515, "step": 766 }, { "epoch": 0.09, "learning_rate": 2.7406871609403255e-06, "loss": 0.7336, "step": 767 }, { "epoch": 0.09, "learning_rate": 2.7403254972875227e-06, "loss": 0.0022, "step": 768 }, { "epoch": 0.09, "learning_rate": 2.73996383363472e-06, "loss": 0.0152, "step": 769 }, { "epoch": 0.09, "learning_rate": 2.7396021699819167e-06, "loss": 0.0012, "step": 770 }, { "epoch": 0.09, "learning_rate": 2.739240506329114e-06, "loss": 0.0016, "step": 771 }, { "epoch": 0.09, "learning_rate": 2.738878842676311e-06, "loss": 0.1283, "step": 772 }, { "epoch": 0.09, "learning_rate": 2.7385171790235084e-06, "loss": 0.002, "step": 773 }, { "epoch": 0.09, "learning_rate": 2.7381555153707052e-06, "loss": 1.0149, "step": 774 }, { "epoch": 0.09, "learning_rate": 2.7377938517179025e-06, "loss": 0.4035, "step": 775 }, { "epoch": 0.09, "learning_rate": 2.7374321880650997e-06, "loss": 0.0029, "step": 776 }, { "epoch": 0.09, "learning_rate": 2.737070524412297e-06, "loss": 0.0088, "step": 777 }, { "epoch": 0.09, "learning_rate": 2.7367088607594937e-06, "loss": 0.0012, "step": 778 }, { "epoch": 0.09, "learning_rate": 2.736347197106691e-06, "loss": 0.0708, "step": 779 }, { "epoch": 0.09, "learning_rate": 2.7359855334538878e-06, "loss": 0.0425, "step": 780 }, { "epoch": 0.09, "learning_rate": 2.735623869801085e-06, "loss": 0.0019, "step": 781 }, { "epoch": 0.09, "learning_rate": 2.7352622061482822e-06, "loss": 0.4199, "step": 782 }, { "epoch": 0.09, "learning_rate": 2.734900542495479e-06, "loss": 0.0013, "step": 783 }, { "epoch": 0.09, "learning_rate": 2.7345388788426763e-06, "loss": 0.0015, "step": 784 }, { "epoch": 0.09, "learning_rate": 2.7341772151898735e-06, "loss": 0.0015, "step": 785 }, { "epoch": 0.09, "learning_rate": 2.7338155515370707e-06, "loss": 0.4673, "step": 786 }, { "epoch": 0.09, "learning_rate": 2.7334538878842675e-06, "loss": 0.0013, "step": 787 }, { "epoch": 0.09, "learning_rate": 2.7330922242314648e-06, "loss": 0.0015, "step": 788 }, { "epoch": 0.09, "learning_rate": 2.732730560578662e-06, "loss": 0.0258, "step": 789 }, { "epoch": 0.09, "learning_rate": 2.732368896925859e-06, "loss": 0.0166, "step": 790 }, { "epoch": 0.09, "learning_rate": 2.732007233273056e-06, "loss": 0.7838, "step": 791 }, { "epoch": 0.09, "learning_rate": 2.7316455696202533e-06, "loss": 0.0013, "step": 792 }, { "epoch": 0.1, "learning_rate": 2.7312839059674505e-06, "loss": 0.1622, "step": 793 }, { "epoch": 0.1, "learning_rate": 2.7309222423146473e-06, "loss": 0.0014, "step": 794 }, { "epoch": 0.1, "learning_rate": 2.7305605786618445e-06, "loss": 0.0012, "step": 795 }, { "epoch": 0.1, "learning_rate": 2.7301989150090418e-06, "loss": 0.001, "step": 796 }, { "epoch": 0.1, "learning_rate": 2.729837251356239e-06, "loss": 0.0011, "step": 797 }, { "epoch": 0.1, "learning_rate": 2.729475587703436e-06, "loss": 0.0012, "step": 798 }, { "epoch": 0.1, "learning_rate": 2.729113924050633e-06, "loss": 0.0015, "step": 799 }, { "epoch": 0.1, "learning_rate": 2.7287522603978303e-06, "loss": 0.0011, "step": 800 }, { "epoch": 0.1, "learning_rate": 2.7283905967450275e-06, "loss": 0.007, "step": 801 }, { "epoch": 0.1, "learning_rate": 2.7280289330922243e-06, "loss": 0.0036, "step": 802 }, { "epoch": 0.1, "learning_rate": 2.7276672694394215e-06, "loss": 0.0059, "step": 803 }, { "epoch": 0.1, "learning_rate": 2.7273056057866183e-06, "loss": 0.4898, "step": 804 }, { "epoch": 0.1, "learning_rate": 2.7269439421338156e-06, "loss": 0.5824, "step": 805 }, { "epoch": 0.1, "learning_rate": 2.726582278481013e-06, "loss": 0.3402, "step": 806 }, { "epoch": 0.1, "learning_rate": 2.7262206148282096e-06, "loss": 0.721, "step": 807 }, { "epoch": 0.1, "learning_rate": 2.725858951175407e-06, "loss": 0.146, "step": 808 }, { "epoch": 0.1, "learning_rate": 2.725497287522604e-06, "loss": 0.0014, "step": 809 }, { "epoch": 0.1, "learning_rate": 2.725135623869801e-06, "loss": 0.0013, "step": 810 }, { "epoch": 0.1, "learning_rate": 2.724773960216998e-06, "loss": 0.0029, "step": 811 }, { "epoch": 0.1, "learning_rate": 2.7244122965641954e-06, "loss": 0.0025, "step": 812 }, { "epoch": 0.1, "learning_rate": 2.7240506329113926e-06, "loss": 0.7846, "step": 813 }, { "epoch": 0.1, "learning_rate": 2.7236889692585894e-06, "loss": 0.0011, "step": 814 }, { "epoch": 0.1, "learning_rate": 2.7233273056057866e-06, "loss": 0.0012, "step": 815 }, { "epoch": 0.1, "learning_rate": 2.722965641952984e-06, "loss": 0.0015, "step": 816 }, { "epoch": 0.1, "learning_rate": 2.722603978300181e-06, "loss": 0.3214, "step": 817 }, { "epoch": 0.1, "learning_rate": 2.722242314647378e-06, "loss": 0.0012, "step": 818 }, { "epoch": 0.1, "learning_rate": 2.721880650994575e-06, "loss": 0.7269, "step": 819 }, { "epoch": 0.1, "learning_rate": 2.7215189873417724e-06, "loss": 0.0014, "step": 820 }, { "epoch": 0.1, "learning_rate": 2.7211573236889696e-06, "loss": 0.0013, "step": 821 }, { "epoch": 0.1, "learning_rate": 2.7207956600361664e-06, "loss": 0.0022, "step": 822 }, { "epoch": 0.1, "learning_rate": 2.7204339963833636e-06, "loss": 0.8211, "step": 823 }, { "epoch": 0.1, "learning_rate": 2.720072332730561e-06, "loss": 0.0009, "step": 824 }, { "epoch": 0.1, "learning_rate": 2.719710669077758e-06, "loss": 0.7548, "step": 825 }, { "epoch": 0.1, "learning_rate": 2.719349005424955e-06, "loss": 0.0011, "step": 826 }, { "epoch": 0.1, "learning_rate": 2.7189873417721517e-06, "loss": 0.0427, "step": 827 }, { "epoch": 0.1, "learning_rate": 2.718625678119349e-06, "loss": 0.0011, "step": 828 }, { "epoch": 0.1, "learning_rate": 2.718264014466546e-06, "loss": 0.001, "step": 829 }, { "epoch": 0.1, "learning_rate": 2.717902350813743e-06, "loss": 0.001, "step": 830 }, { "epoch": 0.1, "learning_rate": 2.71754068716094e-06, "loss": 0.0012, "step": 831 }, { "epoch": 0.1, "learning_rate": 2.7171790235081374e-06, "loss": 1.5676, "step": 832 }, { "epoch": 0.1, "learning_rate": 2.7168173598553347e-06, "loss": 0.0152, "step": 833 }, { "epoch": 0.1, "learning_rate": 2.7164556962025315e-06, "loss": 0.0026, "step": 834 }, { "epoch": 0.1, "learning_rate": 2.7160940325497287e-06, "loss": 0.224, "step": 835 }, { "epoch": 0.1, "learning_rate": 2.715732368896926e-06, "loss": 0.0016, "step": 836 }, { "epoch": 0.1, "learning_rate": 2.715370705244123e-06, "loss": 0.0016, "step": 837 }, { "epoch": 0.1, "learning_rate": 2.71500904159132e-06, "loss": 0.0014, "step": 838 }, { "epoch": 0.1, "learning_rate": 2.714647377938517e-06, "loss": 0.0305, "step": 839 }, { "epoch": 0.1, "learning_rate": 2.7142857142857144e-06, "loss": 0.0015, "step": 840 }, { "epoch": 0.1, "learning_rate": 2.7139240506329117e-06, "loss": 0.002, "step": 841 }, { "epoch": 0.1, "learning_rate": 2.7135623869801085e-06, "loss": 0.0223, "step": 842 }, { "epoch": 0.1, "learning_rate": 2.7132007233273057e-06, "loss": 0.0015, "step": 843 }, { "epoch": 0.1, "learning_rate": 2.712839059674503e-06, "loss": 1.0373, "step": 844 }, { "epoch": 0.1, "learning_rate": 2.7124773960217e-06, "loss": 0.0025, "step": 845 }, { "epoch": 0.1, "learning_rate": 2.712115732368897e-06, "loss": 0.0013, "step": 846 }, { "epoch": 0.1, "learning_rate": 2.711754068716094e-06, "loss": 0.3817, "step": 847 }, { "epoch": 0.1, "learning_rate": 2.7113924050632914e-06, "loss": 0.0012, "step": 848 }, { "epoch": 0.1, "learning_rate": 2.7110307414104882e-06, "loss": 0.0013, "step": 849 }, { "epoch": 0.1, "learning_rate": 2.7106690777576855e-06, "loss": 0.0863, "step": 850 }, { "epoch": 0.1, "learning_rate": 2.7103074141048823e-06, "loss": 0.0099, "step": 851 }, { "epoch": 0.1, "learning_rate": 2.7099457504520795e-06, "loss": 0.0028, "step": 852 }, { "epoch": 0.1, "learning_rate": 2.7095840867992767e-06, "loss": 0.0013, "step": 853 }, { "epoch": 0.1, "learning_rate": 2.7092224231464736e-06, "loss": 0.0016, "step": 854 }, { "epoch": 0.1, "learning_rate": 2.7088607594936708e-06, "loss": 0.0016, "step": 855 }, { "epoch": 0.1, "learning_rate": 2.708499095840868e-06, "loss": 0.0021, "step": 856 }, { "epoch": 0.1, "learning_rate": 2.7081374321880652e-06, "loss": 0.0012, "step": 857 }, { "epoch": 0.1, "learning_rate": 2.707775768535262e-06, "loss": 0.0044, "step": 858 }, { "epoch": 0.1, "learning_rate": 2.7074141048824593e-06, "loss": 0.3029, "step": 859 }, { "epoch": 0.1, "learning_rate": 2.7070524412296565e-06, "loss": 0.0018, "step": 860 }, { "epoch": 0.1, "learning_rate": 2.7066907775768537e-06, "loss": 0.004, "step": 861 }, { "epoch": 0.1, "learning_rate": 2.7063291139240506e-06, "loss": 0.098, "step": 862 }, { "epoch": 0.1, "learning_rate": 2.7059674502712478e-06, "loss": 0.0012, "step": 863 }, { "epoch": 0.1, "learning_rate": 2.705605786618445e-06, "loss": 0.001, "step": 864 }, { "epoch": 0.1, "learning_rate": 2.7052441229656422e-06, "loss": 0.0011, "step": 865 }, { "epoch": 0.1, "learning_rate": 2.704882459312839e-06, "loss": 1.6932, "step": 866 }, { "epoch": 0.1, "learning_rate": 2.7045207956600363e-06, "loss": 0.0014, "step": 867 }, { "epoch": 0.1, "learning_rate": 2.7041591320072335e-06, "loss": 0.7574, "step": 868 }, { "epoch": 0.1, "learning_rate": 2.7037974683544303e-06, "loss": 0.001, "step": 869 }, { "epoch": 0.1, "learning_rate": 2.7034358047016276e-06, "loss": 0.0014, "step": 870 }, { "epoch": 0.1, "learning_rate": 2.703074141048825e-06, "loss": 0.0009, "step": 871 }, { "epoch": 0.1, "learning_rate": 2.702712477396022e-06, "loss": 0.0016, "step": 872 }, { "epoch": 0.1, "learning_rate": 2.702350813743219e-06, "loss": 0.0024, "step": 873 }, { "epoch": 0.1, "learning_rate": 2.701989150090416e-06, "loss": 0.6935, "step": 874 }, { "epoch": 0.1, "learning_rate": 2.701627486437613e-06, "loss": 0.0029, "step": 875 }, { "epoch": 0.1, "learning_rate": 2.70126582278481e-06, "loss": 0.0011, "step": 876 }, { "epoch": 0.11, "learning_rate": 2.7009041591320073e-06, "loss": 0.0011, "step": 877 }, { "epoch": 0.11, "learning_rate": 2.700542495479204e-06, "loss": 0.0012, "step": 878 }, { "epoch": 0.11, "learning_rate": 2.7001808318264014e-06, "loss": 0.8404, "step": 879 }, { "epoch": 0.11, "learning_rate": 2.6998191681735986e-06, "loss": 0.015, "step": 880 }, { "epoch": 0.11, "learning_rate": 2.699457504520796e-06, "loss": 0.6826, "step": 881 }, { "epoch": 0.11, "learning_rate": 2.6990958408679926e-06, "loss": 0.0008, "step": 882 }, { "epoch": 0.11, "learning_rate": 2.69873417721519e-06, "loss": 0.0134, "step": 883 }, { "epoch": 0.11, "learning_rate": 2.698372513562387e-06, "loss": 0.2818, "step": 884 }, { "epoch": 0.11, "learning_rate": 2.6980108499095843e-06, "loss": 0.0014, "step": 885 }, { "epoch": 0.11, "learning_rate": 2.697649186256781e-06, "loss": 0.0012, "step": 886 }, { "epoch": 0.11, "learning_rate": 2.6972875226039784e-06, "loss": 0.0266, "step": 887 }, { "epoch": 0.11, "learning_rate": 2.6969258589511756e-06, "loss": 0.0596, "step": 888 }, { "epoch": 0.11, "learning_rate": 2.6965641952983724e-06, "loss": 0.001, "step": 889 }, { "epoch": 0.11, "learning_rate": 2.6962025316455696e-06, "loss": 0.7597, "step": 890 }, { "epoch": 0.11, "learning_rate": 2.695840867992767e-06, "loss": 0.0016, "step": 891 }, { "epoch": 0.11, "learning_rate": 2.695479204339964e-06, "loss": 0.0009, "step": 892 }, { "epoch": 0.11, "learning_rate": 2.695117540687161e-06, "loss": 0.0014, "step": 893 }, { "epoch": 0.11, "learning_rate": 2.694755877034358e-06, "loss": 0.6644, "step": 894 }, { "epoch": 0.11, "learning_rate": 2.6943942133815554e-06, "loss": 0.6738, "step": 895 }, { "epoch": 0.11, "learning_rate": 2.6940325497287526e-06, "loss": 0.0024, "step": 896 }, { "epoch": 0.11, "learning_rate": 2.6936708860759494e-06, "loss": 0.0014, "step": 897 }, { "epoch": 0.11, "learning_rate": 2.6933092224231466e-06, "loss": 0.4127, "step": 898 }, { "epoch": 0.11, "learning_rate": 2.6929475587703435e-06, "loss": 0.0011, "step": 899 }, { "epoch": 0.11, "learning_rate": 2.6925858951175407e-06, "loss": 0.0015, "step": 900 }, { "epoch": 0.11, "learning_rate": 2.692224231464738e-06, "loss": 0.001, "step": 901 }, { "epoch": 0.11, "learning_rate": 2.6918625678119347e-06, "loss": 0.0011, "step": 902 }, { "epoch": 0.11, "learning_rate": 2.691500904159132e-06, "loss": 0.001, "step": 903 }, { "epoch": 0.11, "learning_rate": 2.691139240506329e-06, "loss": 0.0015, "step": 904 }, { "epoch": 0.11, "learning_rate": 2.6907775768535264e-06, "loss": 0.0519, "step": 905 }, { "epoch": 0.11, "learning_rate": 2.6904159132007232e-06, "loss": 0.001, "step": 906 }, { "epoch": 0.11, "learning_rate": 2.6900542495479205e-06, "loss": 0.0021, "step": 907 }, { "epoch": 0.11, "learning_rate": 2.6896925858951177e-06, "loss": 0.0011, "step": 908 }, { "epoch": 0.11, "learning_rate": 2.6893309222423145e-06, "loss": 0.001, "step": 909 }, { "epoch": 0.11, "learning_rate": 2.6889692585895117e-06, "loss": 0.5719, "step": 910 }, { "epoch": 0.11, "learning_rate": 2.688607594936709e-06, "loss": 0.001, "step": 911 }, { "epoch": 0.11, "learning_rate": 2.688245931283906e-06, "loss": 0.0009, "step": 912 }, { "epoch": 0.11, "learning_rate": 2.687884267631103e-06, "loss": 0.0126, "step": 913 }, { "epoch": 0.11, "learning_rate": 2.6875226039783002e-06, "loss": 0.002, "step": 914 }, { "epoch": 0.11, "learning_rate": 2.6871609403254975e-06, "loss": 0.0021, "step": 915 }, { "epoch": 0.11, "learning_rate": 2.6867992766726947e-06, "loss": 0.001, "step": 916 }, { "epoch": 0.11, "learning_rate": 2.6864376130198915e-06, "loss": 0.2224, "step": 917 }, { "epoch": 0.11, "learning_rate": 2.6860759493670887e-06, "loss": 0.0009, "step": 918 }, { "epoch": 0.11, "learning_rate": 2.685714285714286e-06, "loss": 0.0011, "step": 919 }, { "epoch": 0.11, "learning_rate": 2.685352622061483e-06, "loss": 0.0013, "step": 920 }, { "epoch": 0.11, "learning_rate": 2.68499095840868e-06, "loss": 0.0091, "step": 921 }, { "epoch": 0.11, "learning_rate": 2.6846292947558772e-06, "loss": 0.0012, "step": 922 }, { "epoch": 0.11, "learning_rate": 2.684267631103074e-06, "loss": 0.0009, "step": 923 }, { "epoch": 0.11, "learning_rate": 2.6839059674502713e-06, "loss": 0.5024, "step": 924 }, { "epoch": 0.11, "learning_rate": 2.6835443037974685e-06, "loss": 0.0016, "step": 925 }, { "epoch": 0.11, "learning_rate": 2.6831826401446653e-06, "loss": 0.001, "step": 926 }, { "epoch": 0.11, "learning_rate": 2.6828209764918625e-06, "loss": 0.0009, "step": 927 }, { "epoch": 0.11, "learning_rate": 2.6824593128390598e-06, "loss": 0.0008, "step": 928 }, { "epoch": 0.11, "learning_rate": 2.6820976491862566e-06, "loss": 0.9051, "step": 929 }, { "epoch": 0.11, "learning_rate": 2.681735985533454e-06, "loss": 0.2501, "step": 930 }, { "epoch": 0.11, "learning_rate": 2.681374321880651e-06, "loss": 0.0013, "step": 931 }, { "epoch": 0.11, "learning_rate": 2.6810126582278483e-06, "loss": 0.7314, "step": 932 }, { "epoch": 0.11, "learning_rate": 2.680650994575045e-06, "loss": 0.0007, "step": 933 }, { "epoch": 0.11, "learning_rate": 2.6802893309222423e-06, "loss": 0.0011, "step": 934 }, { "epoch": 0.11, "learning_rate": 2.6799276672694395e-06, "loss": 0.003, "step": 935 }, { "epoch": 0.11, "learning_rate": 2.6795660036166368e-06, "loss": 0.0012, "step": 936 }, { "epoch": 0.11, "learning_rate": 2.6792043399638336e-06, "loss": 0.001, "step": 937 }, { "epoch": 0.11, "learning_rate": 2.678842676311031e-06, "loss": 0.0006, "step": 938 }, { "epoch": 0.11, "learning_rate": 2.678481012658228e-06, "loss": 0.0008, "step": 939 }, { "epoch": 0.11, "learning_rate": 2.6781193490054253e-06, "loss": 0.0009, "step": 940 }, { "epoch": 0.11, "learning_rate": 2.677757685352622e-06, "loss": 0.0014, "step": 941 }, { "epoch": 0.11, "learning_rate": 2.6773960216998193e-06, "loss": 0.0012, "step": 942 }, { "epoch": 0.11, "learning_rate": 2.6770343580470165e-06, "loss": 0.0009, "step": 943 }, { "epoch": 0.11, "learning_rate": 2.6766726943942138e-06, "loss": 0.7488, "step": 944 }, { "epoch": 0.11, "learning_rate": 2.6763110307414106e-06, "loss": 0.7822, "step": 945 }, { "epoch": 0.11, "learning_rate": 2.675949367088608e-06, "loss": 0.0027, "step": 946 }, { "epoch": 0.11, "learning_rate": 2.6755877034358046e-06, "loss": 0.0033, "step": 947 }, { "epoch": 0.11, "learning_rate": 2.675226039783002e-06, "loss": 0.001, "step": 948 }, { "epoch": 0.11, "learning_rate": 2.6748643761301987e-06, "loss": 0.0028, "step": 949 }, { "epoch": 0.11, "learning_rate": 2.674502712477396e-06, "loss": 0.0455, "step": 950 }, { "epoch": 0.11, "learning_rate": 2.674141048824593e-06, "loss": 0.2509, "step": 951 }, { "epoch": 0.11, "learning_rate": 2.6737793851717903e-06, "loss": 0.001, "step": 952 }, { "epoch": 0.11, "learning_rate": 2.673417721518987e-06, "loss": 0.001, "step": 953 }, { "epoch": 0.11, "learning_rate": 2.6730560578661844e-06, "loss": 0.086, "step": 954 }, { "epoch": 0.11, "learning_rate": 2.6726943942133816e-06, "loss": 0.001, "step": 955 }, { "epoch": 0.11, "learning_rate": 2.672332730560579e-06, "loss": 0.0008, "step": 956 }, { "epoch": 0.11, "learning_rate": 2.6719710669077757e-06, "loss": 0.001, "step": 957 }, { "epoch": 0.11, "learning_rate": 2.671609403254973e-06, "loss": 0.7348, "step": 958 }, { "epoch": 0.11, "learning_rate": 2.67124773960217e-06, "loss": 0.0014, "step": 959 }, { "epoch": 0.12, "learning_rate": 2.6708860759493674e-06, "loss": 0.025, "step": 960 }, { "epoch": 0.12, "learning_rate": 2.670524412296564e-06, "loss": 0.7521, "step": 961 }, { "epoch": 0.12, "learning_rate": 2.6701627486437614e-06, "loss": 0.0018, "step": 962 }, { "epoch": 0.12, "learning_rate": 2.6698010849909586e-06, "loss": 0.001, "step": 963 }, { "epoch": 0.12, "learning_rate": 2.669439421338156e-06, "loss": 0.0012, "step": 964 }, { "epoch": 0.12, "learning_rate": 2.6690777576853527e-06, "loss": 0.2813, "step": 965 }, { "epoch": 0.12, "learning_rate": 2.66871609403255e-06, "loss": 0.0006, "step": 966 }, { "epoch": 0.12, "learning_rate": 2.668354430379747e-06, "loss": 0.001, "step": 967 }, { "epoch": 0.12, "learning_rate": 2.6679927667269444e-06, "loss": 0.0008, "step": 968 }, { "epoch": 0.12, "learning_rate": 2.667631103074141e-06, "loss": 0.0011, "step": 969 }, { "epoch": 0.12, "learning_rate": 2.6672694394213384e-06, "loss": 0.0014, "step": 970 }, { "epoch": 0.12, "learning_rate": 2.666907775768535e-06, "loss": 0.5517, "step": 971 }, { "epoch": 0.12, "learning_rate": 2.6665461121157324e-06, "loss": 0.0009, "step": 972 }, { "epoch": 0.12, "learning_rate": 2.6661844484629292e-06, "loss": 0.0011, "step": 973 }, { "epoch": 0.12, "learning_rate": 2.6658227848101265e-06, "loss": 0.0009, "step": 974 }, { "epoch": 0.12, "learning_rate": 2.6654611211573237e-06, "loss": 0.001, "step": 975 }, { "epoch": 0.12, "learning_rate": 2.665099457504521e-06, "loss": 0.8137, "step": 976 }, { "epoch": 0.12, "learning_rate": 2.6647377938517177e-06, "loss": 0.0265, "step": 977 }, { "epoch": 0.12, "learning_rate": 2.664376130198915e-06, "loss": 0.0009, "step": 978 }, { "epoch": 0.12, "learning_rate": 2.664014466546112e-06, "loss": 0.001, "step": 979 }, { "epoch": 0.12, "learning_rate": 2.6636528028933094e-06, "loss": 0.4869, "step": 980 }, { "epoch": 0.12, "learning_rate": 2.6632911392405062e-06, "loss": 0.0019, "step": 981 }, { "epoch": 0.12, "learning_rate": 2.6629294755877035e-06, "loss": 0.7462, "step": 982 }, { "epoch": 0.12, "learning_rate": 2.6625678119349007e-06, "loss": 0.001, "step": 983 }, { "epoch": 0.12, "learning_rate": 2.662206148282098e-06, "loss": 0.9377, "step": 984 }, { "epoch": 0.12, "learning_rate": 2.6618444846292947e-06, "loss": 0.0014, "step": 985 }, { "epoch": 0.12, "learning_rate": 2.661482820976492e-06, "loss": 0.3775, "step": 986 }, { "epoch": 0.12, "learning_rate": 2.661121157323689e-06, "loss": 0.001, "step": 987 }, { "epoch": 0.12, "learning_rate": 2.660759493670886e-06, "loss": 0.0025, "step": 988 }, { "epoch": 0.12, "learning_rate": 2.6603978300180832e-06, "loss": 0.0008, "step": 989 }, { "epoch": 0.12, "learning_rate": 2.6600361663652805e-06, "loss": 0.0025, "step": 990 }, { "epoch": 0.12, "learning_rate": 2.6596745027124777e-06, "loss": 0.0015, "step": 991 }, { "epoch": 0.12, "learning_rate": 2.6593128390596745e-06, "loss": 0.8463, "step": 992 }, { "epoch": 0.12, "learning_rate": 2.6589511754068717e-06, "loss": 0.0033, "step": 993 }, { "epoch": 0.12, "learning_rate": 2.658589511754069e-06, "loss": 0.0009, "step": 994 }, { "epoch": 0.12, "learning_rate": 2.6582278481012658e-06, "loss": 0.0009, "step": 995 }, { "epoch": 0.12, "learning_rate": 2.657866184448463e-06, "loss": 0.0008, "step": 996 }, { "epoch": 0.12, "learning_rate": 2.65750452079566e-06, "loss": 0.0017, "step": 997 }, { "epoch": 0.12, "learning_rate": 2.657142857142857e-06, "loss": 0.0017, "step": 998 }, { "epoch": 0.12, "learning_rate": 2.6567811934900543e-06, "loss": 0.0051, "step": 999 }, { "epoch": 0.12, "learning_rate": 2.6564195298372515e-06, "loss": 0.0011, "step": 1000 }, { "epoch": 0.12, "learning_rate": 2.6560578661844483e-06, "loss": 0.904, "step": 1001 }, { "epoch": 0.12, "learning_rate": 2.6556962025316456e-06, "loss": 0.0015, "step": 1002 }, { "epoch": 0.12, "learning_rate": 2.6553345388788428e-06, "loss": 0.0009, "step": 1003 }, { "epoch": 0.12, "learning_rate": 2.65497287522604e-06, "loss": 0.4277, "step": 1004 }, { "epoch": 0.12, "learning_rate": 2.654611211573237e-06, "loss": 0.0557, "step": 1005 }, { "epoch": 0.12, "learning_rate": 2.654249547920434e-06, "loss": 0.001, "step": 1006 }, { "epoch": 0.12, "learning_rate": 2.6538878842676313e-06, "loss": 0.0012, "step": 1007 }, { "epoch": 0.12, "learning_rate": 2.653526220614828e-06, "loss": 0.001, "step": 1008 }, { "epoch": 0.12, "learning_rate": 2.6531645569620253e-06, "loss": 0.5312, "step": 1009 }, { "epoch": 0.12, "learning_rate": 2.6528028933092226e-06, "loss": 0.0011, "step": 1010 }, { "epoch": 0.12, "learning_rate": 2.6524412296564198e-06, "loss": 0.0009, "step": 1011 }, { "epoch": 0.12, "learning_rate": 2.6520795660036166e-06, "loss": 0.0011, "step": 1012 }, { "epoch": 0.12, "learning_rate": 2.651717902350814e-06, "loss": 0.0013, "step": 1013 }, { "epoch": 0.12, "learning_rate": 2.651356238698011e-06, "loss": 0.0008, "step": 1014 }, { "epoch": 0.12, "learning_rate": 2.6509945750452083e-06, "loss": 0.7669, "step": 1015 }, { "epoch": 0.12, "learning_rate": 2.650632911392405e-06, "loss": 0.0011, "step": 1016 }, { "epoch": 0.12, "learning_rate": 2.6502712477396023e-06, "loss": 0.0012, "step": 1017 }, { "epoch": 0.12, "learning_rate": 2.6499095840867996e-06, "loss": 0.001, "step": 1018 }, { "epoch": 0.12, "learning_rate": 2.6495479204339964e-06, "loss": 0.0014, "step": 1019 }, { "epoch": 0.12, "learning_rate": 2.6491862567811936e-06, "loss": 0.0044, "step": 1020 }, { "epoch": 0.12, "learning_rate": 2.6488245931283904e-06, "loss": 0.0017, "step": 1021 }, { "epoch": 0.12, "learning_rate": 2.6484629294755876e-06, "loss": 0.001, "step": 1022 }, { "epoch": 0.12, "learning_rate": 2.648101265822785e-06, "loss": 0.0009, "step": 1023 }, { "epoch": 0.12, "learning_rate": 2.647739602169982e-06, "loss": 0.0011, "step": 1024 }, { "epoch": 0.12, "learning_rate": 2.647377938517179e-06, "loss": 0.0011, "step": 1025 }, { "epoch": 0.12, "learning_rate": 2.647016274864376e-06, "loss": 0.0008, "step": 1026 }, { "epoch": 0.12, "learning_rate": 2.6466546112115734e-06, "loss": 0.0008, "step": 1027 }, { "epoch": 0.12, "learning_rate": 2.64629294755877e-06, "loss": 0.649, "step": 1028 }, { "epoch": 0.12, "learning_rate": 2.6459312839059674e-06, "loss": 0.0011, "step": 1029 }, { "epoch": 0.12, "learning_rate": 2.6455696202531646e-06, "loss": 0.0011, "step": 1030 }, { "epoch": 0.12, "learning_rate": 2.645207956600362e-06, "loss": 0.3929, "step": 1031 }, { "epoch": 0.12, "learning_rate": 2.6448462929475587e-06, "loss": 0.0008, "step": 1032 }, { "epoch": 0.12, "learning_rate": 2.644484629294756e-06, "loss": 0.0013, "step": 1033 }, { "epoch": 0.12, "learning_rate": 2.644122965641953e-06, "loss": 0.0011, "step": 1034 }, { "epoch": 0.12, "learning_rate": 2.6437613019891504e-06, "loss": 0.2455, "step": 1035 }, { "epoch": 0.12, "learning_rate": 2.643399638336347e-06, "loss": 0.0009, "step": 1036 }, { "epoch": 0.12, "learning_rate": 2.6430379746835444e-06, "loss": 0.0026, "step": 1037 }, { "epoch": 0.12, "learning_rate": 2.6426763110307416e-06, "loss": 0.1014, "step": 1038 }, { "epoch": 0.12, "learning_rate": 2.642314647377939e-06, "loss": 0.2952, "step": 1039 }, { "epoch": 0.12, "learning_rate": 2.6419529837251357e-06, "loss": 0.0026, "step": 1040 }, { "epoch": 0.12, "learning_rate": 2.641591320072333e-06, "loss": 0.0024, "step": 1041 }, { "epoch": 0.12, "learning_rate": 2.64122965641953e-06, "loss": 0.0012, "step": 1042 }, { "epoch": 0.12, "learning_rate": 2.640867992766727e-06, "loss": 0.0018, "step": 1043 }, { "epoch": 0.13, "learning_rate": 2.640506329113924e-06, "loss": 0.8675, "step": 1044 }, { "epoch": 0.13, "learning_rate": 2.640144665461121e-06, "loss": 0.0012, "step": 1045 }, { "epoch": 0.13, "learning_rate": 2.6397830018083182e-06, "loss": 0.0008, "step": 1046 }, { "epoch": 0.13, "learning_rate": 2.6394213381555155e-06, "loss": 0.0012, "step": 1047 }, { "epoch": 0.13, "learning_rate": 2.6390596745027123e-06, "loss": 0.291, "step": 1048 }, { "epoch": 0.13, "learning_rate": 2.6386980108499095e-06, "loss": 0.6893, "step": 1049 }, { "epoch": 0.13, "learning_rate": 2.6383363471971067e-06, "loss": 0.001, "step": 1050 }, { "epoch": 0.13, "learning_rate": 2.637974683544304e-06, "loss": 0.001, "step": 1051 }, { "epoch": 0.13, "learning_rate": 2.6376130198915008e-06, "loss": 0.0008, "step": 1052 }, { "epoch": 0.13, "learning_rate": 2.637251356238698e-06, "loss": 0.0019, "step": 1053 }, { "epoch": 0.13, "learning_rate": 2.6368896925858952e-06, "loss": 0.0803, "step": 1054 }, { "epoch": 0.13, "learning_rate": 2.6365280289330925e-06, "loss": 0.0083, "step": 1055 }, { "epoch": 0.13, "learning_rate": 2.6361663652802893e-06, "loss": 0.0012, "step": 1056 }, { "epoch": 0.13, "learning_rate": 2.6358047016274865e-06, "loss": 0.0009, "step": 1057 }, { "epoch": 0.13, "learning_rate": 2.6354430379746837e-06, "loss": 0.0009, "step": 1058 }, { "epoch": 0.13, "learning_rate": 2.635081374321881e-06, "loss": 0.8084, "step": 1059 }, { "epoch": 0.13, "learning_rate": 2.6347197106690778e-06, "loss": 0.1182, "step": 1060 }, { "epoch": 0.13, "learning_rate": 2.634358047016275e-06, "loss": 0.2185, "step": 1061 }, { "epoch": 0.13, "learning_rate": 2.6339963833634722e-06, "loss": 0.0008, "step": 1062 }, { "epoch": 0.13, "learning_rate": 2.6336347197106695e-06, "loss": 0.7721, "step": 1063 }, { "epoch": 0.13, "learning_rate": 2.6332730560578663e-06, "loss": 0.1146, "step": 1064 }, { "epoch": 0.13, "learning_rate": 2.6329113924050635e-06, "loss": 0.0034, "step": 1065 }, { "epoch": 0.13, "learning_rate": 2.6325497287522603e-06, "loss": 0.001, "step": 1066 }, { "epoch": 0.13, "learning_rate": 2.6321880650994575e-06, "loss": 0.6872, "step": 1067 }, { "epoch": 0.13, "learning_rate": 2.6318264014466543e-06, "loss": 0.0011, "step": 1068 }, { "epoch": 0.13, "learning_rate": 2.6314647377938516e-06, "loss": 0.0008, "step": 1069 }, { "epoch": 0.13, "learning_rate": 2.631103074141049e-06, "loss": 0.001, "step": 1070 }, { "epoch": 0.13, "learning_rate": 2.630741410488246e-06, "loss": 0.0505, "step": 1071 }, { "epoch": 0.13, "learning_rate": 2.630379746835443e-06, "loss": 0.001, "step": 1072 }, { "epoch": 0.13, "learning_rate": 2.63001808318264e-06, "loss": 1.7661, "step": 1073 }, { "epoch": 0.13, "learning_rate": 2.6296564195298373e-06, "loss": 0.0101, "step": 1074 }, { "epoch": 0.13, "learning_rate": 2.6292947558770345e-06, "loss": 0.0009, "step": 1075 }, { "epoch": 0.13, "learning_rate": 2.6289330922242313e-06, "loss": 0.001, "step": 1076 }, { "epoch": 0.13, "learning_rate": 2.6285714285714286e-06, "loss": 0.0112, "step": 1077 }, { "epoch": 0.13, "learning_rate": 2.628209764918626e-06, "loss": 0.001, "step": 1078 }, { "epoch": 0.13, "learning_rate": 2.627848101265823e-06, "loss": 0.0029, "step": 1079 }, { "epoch": 0.13, "learning_rate": 2.62748643761302e-06, "loss": 0.1713, "step": 1080 }, { "epoch": 0.13, "learning_rate": 2.627124773960217e-06, "loss": 0.0014, "step": 1081 }, { "epoch": 0.13, "learning_rate": 2.6267631103074143e-06, "loss": 0.0017, "step": 1082 }, { "epoch": 0.13, "learning_rate": 2.6264014466546115e-06, "loss": 0.0008, "step": 1083 }, { "epoch": 0.13, "learning_rate": 2.6260397830018083e-06, "loss": 0.1261, "step": 1084 }, { "epoch": 0.13, "learning_rate": 2.6256781193490056e-06, "loss": 0.6892, "step": 1085 }, { "epoch": 0.13, "learning_rate": 2.625316455696203e-06, "loss": 0.0007, "step": 1086 }, { "epoch": 0.13, "learning_rate": 2.6249547920434e-06, "loss": 0.1815, "step": 1087 }, { "epoch": 0.13, "learning_rate": 2.624593128390597e-06, "loss": 0.0008, "step": 1088 }, { "epoch": 0.13, "learning_rate": 2.624231464737794e-06, "loss": 0.0008, "step": 1089 }, { "epoch": 0.13, "learning_rate": 2.623869801084991e-06, "loss": 0.001, "step": 1090 }, { "epoch": 0.13, "learning_rate": 2.623508137432188e-06, "loss": 0.0009, "step": 1091 }, { "epoch": 0.13, "learning_rate": 2.623146473779385e-06, "loss": 0.0008, "step": 1092 }, { "epoch": 0.13, "learning_rate": 2.622784810126582e-06, "loss": 0.0007, "step": 1093 }, { "epoch": 0.13, "learning_rate": 2.6224231464737794e-06, "loss": 0.0008, "step": 1094 }, { "epoch": 0.13, "learning_rate": 2.6220614828209766e-06, "loss": 0.001, "step": 1095 }, { "epoch": 0.13, "learning_rate": 2.6216998191681734e-06, "loss": 0.0011, "step": 1096 }, { "epoch": 0.13, "learning_rate": 2.6213381555153707e-06, "loss": 0.0007, "step": 1097 }, { "epoch": 0.13, "learning_rate": 2.620976491862568e-06, "loss": 0.0014, "step": 1098 }, { "epoch": 0.13, "learning_rate": 2.620614828209765e-06, "loss": 0.0356, "step": 1099 }, { "epoch": 0.13, "learning_rate": 2.620253164556962e-06, "loss": 0.001, "step": 1100 }, { "epoch": 0.13, "learning_rate": 2.619891500904159e-06, "loss": 0.5485, "step": 1101 }, { "epoch": 0.13, "learning_rate": 2.6195298372513564e-06, "loss": 0.0006, "step": 1102 }, { "epoch": 0.13, "learning_rate": 2.6191681735985536e-06, "loss": 0.7367, "step": 1103 }, { "epoch": 0.13, "learning_rate": 2.6188065099457504e-06, "loss": 0.0337, "step": 1104 }, { "epoch": 0.13, "learning_rate": 2.6184448462929477e-06, "loss": 0.0008, "step": 1105 }, { "epoch": 0.13, "learning_rate": 2.618083182640145e-06, "loss": 0.967, "step": 1106 }, { "epoch": 0.13, "learning_rate": 2.617721518987342e-06, "loss": 0.0007, "step": 1107 }, { "epoch": 0.13, "learning_rate": 2.617359855334539e-06, "loss": 0.8566, "step": 1108 }, { "epoch": 0.13, "learning_rate": 2.616998191681736e-06, "loss": 0.0009, "step": 1109 }, { "epoch": 0.13, "learning_rate": 2.6166365280289334e-06, "loss": 0.0007, "step": 1110 }, { "epoch": 0.13, "learning_rate": 2.61627486437613e-06, "loss": 0.7036, "step": 1111 }, { "epoch": 0.13, "learning_rate": 2.6159132007233274e-06, "loss": 0.0239, "step": 1112 }, { "epoch": 0.13, "learning_rate": 2.6155515370705247e-06, "loss": 0.0007, "step": 1113 }, { "epoch": 0.13, "learning_rate": 2.6151898734177215e-06, "loss": 0.001, "step": 1114 }, { "epoch": 0.13, "learning_rate": 2.6148282097649187e-06, "loss": 0.8387, "step": 1115 }, { "epoch": 0.13, "learning_rate": 2.6144665461121155e-06, "loss": 0.0009, "step": 1116 }, { "epoch": 0.13, "learning_rate": 2.6141048824593127e-06, "loss": 0.0027, "step": 1117 }, { "epoch": 0.13, "learning_rate": 2.61374321880651e-06, "loss": 0.3823, "step": 1118 }, { "epoch": 0.13, "learning_rate": 2.613381555153707e-06, "loss": 0.0022, "step": 1119 }, { "epoch": 0.13, "learning_rate": 2.613019891500904e-06, "loss": 0.0011, "step": 1120 }, { "epoch": 0.13, "learning_rate": 2.6126582278481012e-06, "loss": 0.001, "step": 1121 }, { "epoch": 0.13, "learning_rate": 2.6122965641952985e-06, "loss": 0.0054, "step": 1122 }, { "epoch": 0.13, "learning_rate": 2.6119349005424957e-06, "loss": 0.0017, "step": 1123 }, { "epoch": 0.13, "learning_rate": 2.6115732368896925e-06, "loss": 0.0011, "step": 1124 }, { "epoch": 0.13, "learning_rate": 2.6112115732368897e-06, "loss": 0.001, "step": 1125 }, { "epoch": 0.13, "learning_rate": 2.610849909584087e-06, "loss": 0.001, "step": 1126 }, { "epoch": 0.14, "learning_rate": 2.6104882459312838e-06, "loss": 0.0032, "step": 1127 }, { "epoch": 0.14, "learning_rate": 2.610126582278481e-06, "loss": 0.0017, "step": 1128 }, { "epoch": 0.14, "learning_rate": 2.6097649186256782e-06, "loss": 0.0079, "step": 1129 }, { "epoch": 0.14, "learning_rate": 2.6094032549728755e-06, "loss": 0.0008, "step": 1130 }, { "epoch": 0.14, "learning_rate": 2.6090415913200723e-06, "loss": 0.0794, "step": 1131 }, { "epoch": 0.14, "learning_rate": 2.6086799276672695e-06, "loss": 0.0019, "step": 1132 }, { "epoch": 0.14, "learning_rate": 2.6083182640144667e-06, "loss": 0.0012, "step": 1133 }, { "epoch": 0.14, "learning_rate": 2.607956600361664e-06, "loss": 0.001, "step": 1134 }, { "epoch": 0.14, "learning_rate": 2.6075949367088608e-06, "loss": 0.0014, "step": 1135 }, { "epoch": 0.14, "learning_rate": 2.607233273056058e-06, "loss": 0.1373, "step": 1136 }, { "epoch": 0.14, "learning_rate": 2.6068716094032552e-06, "loss": 0.2415, "step": 1137 }, { "epoch": 0.14, "learning_rate": 2.606509945750452e-06, "loss": 0.0009, "step": 1138 }, { "epoch": 0.14, "learning_rate": 2.6061482820976493e-06, "loss": 0.0009, "step": 1139 }, { "epoch": 0.14, "learning_rate": 2.605786618444846e-06, "loss": 0.0036, "step": 1140 }, { "epoch": 0.14, "learning_rate": 2.6054249547920433e-06, "loss": 0.0026, "step": 1141 }, { "epoch": 0.14, "learning_rate": 2.6050632911392406e-06, "loss": 0.0008, "step": 1142 }, { "epoch": 0.14, "learning_rate": 2.6047016274864378e-06, "loss": 0.5724, "step": 1143 }, { "epoch": 0.14, "learning_rate": 2.6043399638336346e-06, "loss": 0.0437, "step": 1144 }, { "epoch": 0.14, "learning_rate": 2.603978300180832e-06, "loss": 0.001, "step": 1145 }, { "epoch": 0.14, "learning_rate": 2.603616636528029e-06, "loss": 0.0008, "step": 1146 }, { "epoch": 0.14, "learning_rate": 2.603254972875226e-06, "loss": 0.3496, "step": 1147 }, { "epoch": 0.14, "learning_rate": 2.602893309222423e-06, "loss": 0.0009, "step": 1148 }, { "epoch": 0.14, "learning_rate": 2.6025316455696203e-06, "loss": 0.0648, "step": 1149 }, { "epoch": 0.14, "learning_rate": 2.6021699819168176e-06, "loss": 1.4393, "step": 1150 }, { "epoch": 0.14, "learning_rate": 2.6018083182640144e-06, "loss": 0.0011, "step": 1151 }, { "epoch": 0.14, "learning_rate": 2.6014466546112116e-06, "loss": 0.0009, "step": 1152 }, { "epoch": 0.14, "learning_rate": 2.601084990958409e-06, "loss": 0.0017, "step": 1153 }, { "epoch": 0.14, "learning_rate": 2.600723327305606e-06, "loss": 0.002, "step": 1154 }, { "epoch": 0.14, "learning_rate": 2.600361663652803e-06, "loss": 0.0015, "step": 1155 }, { "epoch": 0.14, "learning_rate": 2.6e-06, "loss": 0.0012, "step": 1156 }, { "epoch": 0.14, "learning_rate": 2.5996383363471973e-06, "loss": 0.001, "step": 1157 }, { "epoch": 0.14, "learning_rate": 2.5992766726943946e-06, "loss": 0.0007, "step": 1158 }, { "epoch": 0.14, "learning_rate": 2.5989150090415914e-06, "loss": 0.8518, "step": 1159 }, { "epoch": 0.14, "learning_rate": 2.5985533453887886e-06, "loss": 0.6396, "step": 1160 }, { "epoch": 0.14, "learning_rate": 2.598191681735986e-06, "loss": 0.0009, "step": 1161 }, { "epoch": 0.14, "learning_rate": 2.5978300180831826e-06, "loss": 0.0008, "step": 1162 }, { "epoch": 0.14, "learning_rate": 2.59746835443038e-06, "loss": 0.3063, "step": 1163 }, { "epoch": 0.14, "learning_rate": 2.5971066907775767e-06, "loss": 0.0013, "step": 1164 }, { "epoch": 0.14, "learning_rate": 2.596745027124774e-06, "loss": 0.0022, "step": 1165 }, { "epoch": 0.14, "learning_rate": 2.596383363471971e-06, "loss": 0.0017, "step": 1166 }, { "epoch": 0.14, "learning_rate": 2.596021699819168e-06, "loss": 0.0007, "step": 1167 }, { "epoch": 0.14, "learning_rate": 2.595660036166365e-06, "loss": 0.001, "step": 1168 }, { "epoch": 0.14, "learning_rate": 2.5952983725135624e-06, "loss": 0.0561, "step": 1169 }, { "epoch": 0.14, "learning_rate": 2.5949367088607596e-06, "loss": 0.0009, "step": 1170 }, { "epoch": 0.14, "learning_rate": 2.5945750452079564e-06, "loss": 0.0008, "step": 1171 }, { "epoch": 0.14, "learning_rate": 2.5942133815551537e-06, "loss": 0.9405, "step": 1172 }, { "epoch": 0.14, "learning_rate": 2.593851717902351e-06, "loss": 0.0011, "step": 1173 }, { "epoch": 0.14, "learning_rate": 2.593490054249548e-06, "loss": 0.0009, "step": 1174 }, { "epoch": 0.14, "learning_rate": 2.593128390596745e-06, "loss": 0.0009, "step": 1175 }, { "epoch": 0.14, "learning_rate": 2.592766726943942e-06, "loss": 0.001, "step": 1176 }, { "epoch": 0.14, "learning_rate": 2.5924050632911394e-06, "loss": 0.001, "step": 1177 }, { "epoch": 0.14, "learning_rate": 2.5920433996383366e-06, "loss": 0.3647, "step": 1178 }, { "epoch": 0.14, "learning_rate": 2.5916817359855334e-06, "loss": 0.0008, "step": 1179 }, { "epoch": 0.14, "learning_rate": 2.5913200723327307e-06, "loss": 0.001, "step": 1180 }, { "epoch": 0.14, "learning_rate": 2.590958408679928e-06, "loss": 0.0007, "step": 1181 }, { "epoch": 0.14, "learning_rate": 2.590596745027125e-06, "loss": 0.0036, "step": 1182 }, { "epoch": 0.14, "learning_rate": 2.590235081374322e-06, "loss": 0.0129, "step": 1183 }, { "epoch": 0.14, "learning_rate": 2.589873417721519e-06, "loss": 0.0015, "step": 1184 }, { "epoch": 0.14, "learning_rate": 2.5895117540687164e-06, "loss": 0.0008, "step": 1185 }, { "epoch": 0.14, "learning_rate": 2.5891500904159132e-06, "loss": 0.7697, "step": 1186 }, { "epoch": 0.14, "learning_rate": 2.58878842676311e-06, "loss": 0.0094, "step": 1187 }, { "epoch": 0.14, "learning_rate": 2.5884267631103073e-06, "loss": 0.0009, "step": 1188 }, { "epoch": 0.14, "learning_rate": 2.5880650994575045e-06, "loss": 0.0201, "step": 1189 }, { "epoch": 0.14, "learning_rate": 2.5877034358047017e-06, "loss": 0.0009, "step": 1190 }, { "epoch": 0.14, "learning_rate": 2.5873417721518985e-06, "loss": 0.9054, "step": 1191 }, { "epoch": 0.14, "learning_rate": 2.5869801084990958e-06, "loss": 0.0011, "step": 1192 }, { "epoch": 0.14, "learning_rate": 2.586618444846293e-06, "loss": 0.0007, "step": 1193 }, { "epoch": 0.14, "learning_rate": 2.5862567811934902e-06, "loss": 0.0007, "step": 1194 }, { "epoch": 0.14, "learning_rate": 2.585895117540687e-06, "loss": 0.001, "step": 1195 }, { "epoch": 0.14, "learning_rate": 2.5855334538878843e-06, "loss": 0.0007, "step": 1196 }, { "epoch": 0.14, "learning_rate": 2.5851717902350815e-06, "loss": 0.0007, "step": 1197 }, { "epoch": 0.14, "learning_rate": 2.5848101265822787e-06, "loss": 0.0008, "step": 1198 }, { "epoch": 0.14, "learning_rate": 2.5844484629294755e-06, "loss": 0.0009, "step": 1199 }, { "epoch": 0.14, "learning_rate": 2.5840867992766728e-06, "loss": 0.0009, "step": 1200 }, { "epoch": 0.14, "learning_rate": 2.58372513562387e-06, "loss": 0.7633, "step": 1201 }, { "epoch": 0.14, "learning_rate": 2.5833634719710672e-06, "loss": 0.0016, "step": 1202 }, { "epoch": 0.14, "learning_rate": 2.583001808318264e-06, "loss": 0.0898, "step": 1203 }, { "epoch": 0.14, "learning_rate": 2.5826401446654613e-06, "loss": 0.0021, "step": 1204 }, { "epoch": 0.14, "learning_rate": 2.5822784810126585e-06, "loss": 0.0007, "step": 1205 }, { "epoch": 0.14, "learning_rate": 2.5819168173598557e-06, "loss": 0.001, "step": 1206 }, { "epoch": 0.14, "learning_rate": 2.5815551537070525e-06, "loss": 0.7408, "step": 1207 }, { "epoch": 0.14, "learning_rate": 2.5811934900542498e-06, "loss": 0.0007, "step": 1208 }, { "epoch": 0.14, "learning_rate": 2.580831826401447e-06, "loss": 0.0009, "step": 1209 }, { "epoch": 0.14, "learning_rate": 2.580470162748644e-06, "loss": 0.0007, "step": 1210 }, { "epoch": 0.15, "learning_rate": 2.5801084990958406e-06, "loss": 0.0016, "step": 1211 }, { "epoch": 0.15, "learning_rate": 2.579746835443038e-06, "loss": 0.1773, "step": 1212 }, { "epoch": 0.15, "learning_rate": 2.579385171790235e-06, "loss": 0.0008, "step": 1213 }, { "epoch": 0.15, "learning_rate": 2.5790235081374323e-06, "loss": 0.0013, "step": 1214 }, { "epoch": 0.15, "learning_rate": 2.578661844484629e-06, "loss": 0.0008, "step": 1215 }, { "epoch": 0.15, "learning_rate": 2.5783001808318263e-06, "loss": 0.0008, "step": 1216 }, { "epoch": 0.15, "learning_rate": 2.5779385171790236e-06, "loss": 0.4379, "step": 1217 }, { "epoch": 0.15, "learning_rate": 2.577576853526221e-06, "loss": 0.0007, "step": 1218 }, { "epoch": 0.15, "learning_rate": 2.5772151898734176e-06, "loss": 0.0069, "step": 1219 }, { "epoch": 0.15, "learning_rate": 2.576853526220615e-06, "loss": 0.0009, "step": 1220 }, { "epoch": 0.15, "learning_rate": 2.576491862567812e-06, "loss": 0.0007, "step": 1221 }, { "epoch": 0.15, "learning_rate": 2.5761301989150093e-06, "loss": 0.0007, "step": 1222 }, { "epoch": 0.15, "learning_rate": 2.575768535262206e-06, "loss": 0.0016, "step": 1223 }, { "epoch": 0.15, "learning_rate": 2.5754068716094033e-06, "loss": 0.8036, "step": 1224 }, { "epoch": 0.15, "learning_rate": 2.5750452079566006e-06, "loss": 0.0009, "step": 1225 }, { "epoch": 0.15, "learning_rate": 2.574683544303798e-06, "loss": 0.0013, "step": 1226 }, { "epoch": 0.15, "learning_rate": 2.5743218806509946e-06, "loss": 0.0008, "step": 1227 }, { "epoch": 0.15, "learning_rate": 2.573960216998192e-06, "loss": 0.0006, "step": 1228 }, { "epoch": 0.15, "learning_rate": 2.573598553345389e-06, "loss": 0.0333, "step": 1229 }, { "epoch": 0.15, "learning_rate": 2.573236889692586e-06, "loss": 0.007, "step": 1230 }, { "epoch": 0.15, "learning_rate": 2.572875226039783e-06, "loss": 0.9043, "step": 1231 }, { "epoch": 0.15, "learning_rate": 2.5725135623869803e-06, "loss": 0.0007, "step": 1232 }, { "epoch": 0.15, "learning_rate": 2.5721518987341776e-06, "loss": 0.0008, "step": 1233 }, { "epoch": 0.15, "learning_rate": 2.5717902350813744e-06, "loss": 0.0008, "step": 1234 }, { "epoch": 0.15, "learning_rate": 2.571428571428571e-06, "loss": 0.8769, "step": 1235 }, { "epoch": 0.15, "learning_rate": 2.5710669077757684e-06, "loss": 0.0012, "step": 1236 }, { "epoch": 0.15, "learning_rate": 2.5707052441229657e-06, "loss": 0.0009, "step": 1237 }, { "epoch": 0.15, "learning_rate": 2.570343580470163e-06, "loss": 0.5142, "step": 1238 }, { "epoch": 0.15, "learning_rate": 2.5699819168173597e-06, "loss": 0.0006, "step": 1239 }, { "epoch": 0.15, "learning_rate": 2.569620253164557e-06, "loss": 0.0011, "step": 1240 }, { "epoch": 0.15, "learning_rate": 2.569258589511754e-06, "loss": 0.0007, "step": 1241 }, { "epoch": 0.15, "learning_rate": 2.5688969258589514e-06, "loss": 0.8691, "step": 1242 }, { "epoch": 0.15, "learning_rate": 2.568535262206148e-06, "loss": 0.0009, "step": 1243 }, { "epoch": 0.15, "learning_rate": 2.5681735985533454e-06, "loss": 0.0008, "step": 1244 }, { "epoch": 0.15, "learning_rate": 2.5678119349005427e-06, "loss": 0.0018, "step": 1245 }, { "epoch": 0.15, "learning_rate": 2.56745027124774e-06, "loss": 0.016, "step": 1246 }, { "epoch": 0.15, "learning_rate": 2.5670886075949367e-06, "loss": 0.0413, "step": 1247 }, { "epoch": 0.15, "learning_rate": 2.566726943942134e-06, "loss": 0.0012, "step": 1248 }, { "epoch": 0.15, "learning_rate": 2.566365280289331e-06, "loss": 0.0007, "step": 1249 }, { "epoch": 0.15, "learning_rate": 2.566003616636528e-06, "loss": 0.2044, "step": 1250 }, { "epoch": 0.15, "learning_rate": 2.565641952983725e-06, "loss": 0.0014, "step": 1251 }, { "epoch": 0.15, "learning_rate": 2.5652802893309224e-06, "loss": 0.0008, "step": 1252 }, { "epoch": 0.15, "learning_rate": 2.5649186256781197e-06, "loss": 1.2803, "step": 1253 }, { "epoch": 0.15, "learning_rate": 2.5645569620253165e-06, "loss": 0.519, "step": 1254 }, { "epoch": 0.15, "learning_rate": 2.5641952983725137e-06, "loss": 0.8631, "step": 1255 }, { "epoch": 0.15, "learning_rate": 2.563833634719711e-06, "loss": 0.0398, "step": 1256 }, { "epoch": 0.15, "learning_rate": 2.563471971066908e-06, "loss": 0.0059, "step": 1257 }, { "epoch": 0.15, "learning_rate": 2.563110307414105e-06, "loss": 0.5624, "step": 1258 }, { "epoch": 0.15, "learning_rate": 2.5627486437613018e-06, "loss": 0.6324, "step": 1259 }, { "epoch": 0.15, "learning_rate": 2.562386980108499e-06, "loss": 0.0016, "step": 1260 }, { "epoch": 0.15, "learning_rate": 2.5620253164556962e-06, "loss": 0.001, "step": 1261 }, { "epoch": 0.15, "learning_rate": 2.5616636528028935e-06, "loss": 0.0054, "step": 1262 }, { "epoch": 0.15, "learning_rate": 2.5613019891500903e-06, "loss": 0.0022, "step": 1263 }, { "epoch": 0.15, "learning_rate": 2.5609403254972875e-06, "loss": 0.0022, "step": 1264 }, { "epoch": 0.15, "learning_rate": 2.5605786618444847e-06, "loss": 0.0012, "step": 1265 }, { "epoch": 0.15, "learning_rate": 2.5602169981916815e-06, "loss": 0.0011, "step": 1266 }, { "epoch": 0.15, "learning_rate": 2.5598553345388788e-06, "loss": 0.0699, "step": 1267 }, { "epoch": 0.15, "learning_rate": 2.559493670886076e-06, "loss": 1.4501, "step": 1268 }, { "epoch": 0.15, "learning_rate": 2.5591320072332732e-06, "loss": 0.0017, "step": 1269 }, { "epoch": 0.15, "learning_rate": 2.55877034358047e-06, "loss": 0.002, "step": 1270 }, { "epoch": 0.15, "learning_rate": 2.5584086799276673e-06, "loss": 0.0142, "step": 1271 }, { "epoch": 0.15, "learning_rate": 2.5580470162748645e-06, "loss": 0.001, "step": 1272 }, { "epoch": 0.15, "learning_rate": 2.5576853526220617e-06, "loss": 0.0012, "step": 1273 }, { "epoch": 0.15, "learning_rate": 2.5573236889692585e-06, "loss": 0.0016, "step": 1274 }, { "epoch": 0.15, "learning_rate": 2.5569620253164558e-06, "loss": 0.0561, "step": 1275 }, { "epoch": 0.15, "learning_rate": 2.556600361663653e-06, "loss": 0.0015, "step": 1276 }, { "epoch": 0.15, "learning_rate": 2.5562386980108502e-06, "loss": 0.0007, "step": 1277 }, { "epoch": 0.15, "learning_rate": 2.555877034358047e-06, "loss": 0.0013, "step": 1278 }, { "epoch": 0.15, "learning_rate": 2.5555153707052443e-06, "loss": 0.4543, "step": 1279 }, { "epoch": 0.15, "learning_rate": 2.5551537070524415e-06, "loss": 0.0103, "step": 1280 }, { "epoch": 0.15, "learning_rate": 2.5547920433996387e-06, "loss": 0.001, "step": 1281 }, { "epoch": 0.15, "learning_rate": 2.5544303797468355e-06, "loss": 0.0012, "step": 1282 }, { "epoch": 0.15, "learning_rate": 2.5540687160940324e-06, "loss": 0.0012, "step": 1283 }, { "epoch": 0.15, "learning_rate": 2.5537070524412296e-06, "loss": 0.4454, "step": 1284 }, { "epoch": 0.15, "learning_rate": 2.553345388788427e-06, "loss": 0.0011, "step": 1285 }, { "epoch": 0.15, "learning_rate": 2.5529837251356236e-06, "loss": 0.0008, "step": 1286 }, { "epoch": 0.15, "learning_rate": 2.552622061482821e-06, "loss": 0.0006, "step": 1287 }, { "epoch": 0.15, "learning_rate": 2.552260397830018e-06, "loss": 0.0012, "step": 1288 }, { "epoch": 0.15, "learning_rate": 2.5518987341772153e-06, "loss": 0.0367, "step": 1289 }, { "epoch": 0.15, "learning_rate": 2.551537070524412e-06, "loss": 0.002, "step": 1290 }, { "epoch": 0.15, "learning_rate": 2.5511754068716094e-06, "loss": 0.0015, "step": 1291 }, { "epoch": 0.15, "learning_rate": 2.5508137432188066e-06, "loss": 0.0011, "step": 1292 }, { "epoch": 0.15, "learning_rate": 2.550452079566004e-06, "loss": 0.0172, "step": 1293 }, { "epoch": 0.16, "learning_rate": 2.5500904159132006e-06, "loss": 0.0012, "step": 1294 }, { "epoch": 0.16, "learning_rate": 2.549728752260398e-06, "loss": 0.0008, "step": 1295 }, { "epoch": 0.16, "learning_rate": 2.549367088607595e-06, "loss": 0.5449, "step": 1296 }, { "epoch": 0.16, "learning_rate": 2.5490054249547923e-06, "loss": 0.0007, "step": 1297 }, { "epoch": 0.16, "learning_rate": 2.548643761301989e-06, "loss": 0.0658, "step": 1298 }, { "epoch": 0.16, "learning_rate": 2.5482820976491864e-06, "loss": 0.0472, "step": 1299 }, { "epoch": 0.16, "learning_rate": 2.5479204339963836e-06, "loss": 0.0009, "step": 1300 }, { "epoch": 0.16, "learning_rate": 2.547558770343581e-06, "loss": 0.0014, "step": 1301 }, { "epoch": 0.16, "learning_rate": 2.5471971066907776e-06, "loss": 0.0009, "step": 1302 }, { "epoch": 0.16, "learning_rate": 2.546835443037975e-06, "loss": 0.0021, "step": 1303 }, { "epoch": 0.16, "learning_rate": 2.546473779385172e-06, "loss": 0.001, "step": 1304 }, { "epoch": 0.16, "learning_rate": 2.5461121157323693e-06, "loss": 0.6583, "step": 1305 }, { "epoch": 0.16, "learning_rate": 2.5457504520795657e-06, "loss": 0.0009, "step": 1306 }, { "epoch": 0.16, "learning_rate": 2.545388788426763e-06, "loss": 0.0011, "step": 1307 }, { "epoch": 0.16, "learning_rate": 2.54502712477396e-06, "loss": 0.0009, "step": 1308 }, { "epoch": 0.16, "learning_rate": 2.5446654611211574e-06, "loss": 0.3945, "step": 1309 }, { "epoch": 0.16, "learning_rate": 2.544303797468354e-06, "loss": 0.4089, "step": 1310 }, { "epoch": 0.16, "learning_rate": 2.5439421338155514e-06, "loss": 0.001, "step": 1311 }, { "epoch": 0.16, "learning_rate": 2.5435804701627487e-06, "loss": 0.0013, "step": 1312 }, { "epoch": 0.16, "learning_rate": 2.543218806509946e-06, "loss": 0.028, "step": 1313 }, { "epoch": 0.16, "learning_rate": 2.5428571428571427e-06, "loss": 0.0009, "step": 1314 }, { "epoch": 0.16, "learning_rate": 2.54249547920434e-06, "loss": 0.0215, "step": 1315 }, { "epoch": 0.16, "learning_rate": 2.542133815551537e-06, "loss": 0.0007, "step": 1316 }, { "epoch": 0.16, "learning_rate": 2.5417721518987344e-06, "loss": 0.0077, "step": 1317 }, { "epoch": 0.16, "learning_rate": 2.5414104882459312e-06, "loss": 0.0659, "step": 1318 }, { "epoch": 0.16, "learning_rate": 2.5410488245931284e-06, "loss": 0.0167, "step": 1319 }, { "epoch": 0.16, "learning_rate": 2.5406871609403257e-06, "loss": 0.0019, "step": 1320 }, { "epoch": 0.16, "learning_rate": 2.540325497287523e-06, "loss": 0.0008, "step": 1321 }, { "epoch": 0.16, "learning_rate": 2.5399638336347197e-06, "loss": 0.0006, "step": 1322 }, { "epoch": 0.16, "learning_rate": 2.539602169981917e-06, "loss": 0.0008, "step": 1323 }, { "epoch": 0.16, "learning_rate": 2.539240506329114e-06, "loss": 0.0007, "step": 1324 }, { "epoch": 0.16, "learning_rate": 2.5388788426763114e-06, "loss": 0.0007, "step": 1325 }, { "epoch": 0.16, "learning_rate": 2.5385171790235082e-06, "loss": 0.0007, "step": 1326 }, { "epoch": 0.16, "learning_rate": 2.5381555153707054e-06, "loss": 0.002, "step": 1327 }, { "epoch": 0.16, "learning_rate": 2.5377938517179027e-06, "loss": 0.0005, "step": 1328 }, { "epoch": 0.16, "learning_rate": 2.5374321880650995e-06, "loss": 0.0006, "step": 1329 }, { "epoch": 0.16, "learning_rate": 2.5370705244122963e-06, "loss": 0.6811, "step": 1330 }, { "epoch": 0.16, "learning_rate": 2.5367088607594935e-06, "loss": 0.0011, "step": 1331 }, { "epoch": 0.16, "learning_rate": 2.5363471971066908e-06, "loss": 0.8269, "step": 1332 }, { "epoch": 0.16, "learning_rate": 2.535985533453888e-06, "loss": 0.0005, "step": 1333 }, { "epoch": 0.16, "learning_rate": 2.535623869801085e-06, "loss": 0.8256, "step": 1334 }, { "epoch": 0.16, "learning_rate": 2.535262206148282e-06, "loss": 0.0022, "step": 1335 }, { "epoch": 0.16, "learning_rate": 2.5349005424954793e-06, "loss": 0.169, "step": 1336 }, { "epoch": 0.16, "learning_rate": 2.5345388788426765e-06, "loss": 0.8593, "step": 1337 }, { "epoch": 0.16, "learning_rate": 2.5341772151898733e-06, "loss": 0.001, "step": 1338 }, { "epoch": 0.16, "learning_rate": 2.5338155515370705e-06, "loss": 0.0007, "step": 1339 }, { "epoch": 0.16, "learning_rate": 2.5334538878842678e-06, "loss": 0.0008, "step": 1340 }, { "epoch": 0.16, "learning_rate": 2.533092224231465e-06, "loss": 0.0007, "step": 1341 }, { "epoch": 0.16, "learning_rate": 2.532730560578662e-06, "loss": 0.0007, "step": 1342 }, { "epoch": 0.16, "learning_rate": 2.532368896925859e-06, "loss": 0.0006, "step": 1343 }, { "epoch": 0.16, "learning_rate": 2.5320072332730563e-06, "loss": 0.611, "step": 1344 }, { "epoch": 0.16, "learning_rate": 2.5316455696202535e-06, "loss": 0.0012, "step": 1345 }, { "epoch": 0.16, "learning_rate": 2.5312839059674503e-06, "loss": 0.1574, "step": 1346 }, { "epoch": 0.16, "learning_rate": 2.5309222423146475e-06, "loss": 0.0081, "step": 1347 }, { "epoch": 0.16, "learning_rate": 2.5305605786618448e-06, "loss": 0.0016, "step": 1348 }, { "epoch": 0.16, "learning_rate": 2.5301989150090416e-06, "loss": 0.0013, "step": 1349 }, { "epoch": 0.16, "learning_rate": 2.529837251356239e-06, "loss": 0.0046, "step": 1350 }, { "epoch": 0.16, "learning_rate": 2.529475587703436e-06, "loss": 0.0015, "step": 1351 }, { "epoch": 0.16, "learning_rate": 2.5291139240506333e-06, "loss": 0.0009, "step": 1352 }, { "epoch": 0.16, "learning_rate": 2.52875226039783e-06, "loss": 0.5389, "step": 1353 }, { "epoch": 0.16, "learning_rate": 2.528390596745027e-06, "loss": 0.0007, "step": 1354 }, { "epoch": 0.16, "learning_rate": 2.528028933092224e-06, "loss": 0.1, "step": 1355 }, { "epoch": 0.16, "learning_rate": 2.5276672694394213e-06, "loss": 0.0009, "step": 1356 }, { "epoch": 0.16, "learning_rate": 2.5273056057866186e-06, "loss": 0.0014, "step": 1357 }, { "epoch": 0.16, "learning_rate": 2.5269439421338154e-06, "loss": 0.0234, "step": 1358 }, { "epoch": 0.16, "learning_rate": 2.5265822784810126e-06, "loss": 0.0009, "step": 1359 }, { "epoch": 0.16, "learning_rate": 2.52622061482821e-06, "loss": 0.0136, "step": 1360 }, { "epoch": 0.16, "learning_rate": 2.525858951175407e-06, "loss": 0.003, "step": 1361 }, { "epoch": 0.16, "learning_rate": 2.525497287522604e-06, "loss": 0.0011, "step": 1362 }, { "epoch": 0.16, "learning_rate": 2.525135623869801e-06, "loss": 0.8355, "step": 1363 }, { "epoch": 0.16, "learning_rate": 2.5247739602169983e-06, "loss": 0.0006, "step": 1364 }, { "epoch": 0.16, "learning_rate": 2.5244122965641956e-06, "loss": 0.0005, "step": 1365 }, { "epoch": 0.16, "learning_rate": 2.5240506329113924e-06, "loss": 0.0173, "step": 1366 }, { "epoch": 0.16, "learning_rate": 2.5236889692585896e-06, "loss": 0.0007, "step": 1367 }, { "epoch": 0.16, "learning_rate": 2.523327305605787e-06, "loss": 0.0007, "step": 1368 }, { "epoch": 0.16, "learning_rate": 2.5229656419529836e-06, "loss": 0.0009, "step": 1369 }, { "epoch": 0.16, "learning_rate": 2.522603978300181e-06, "loss": 0.0008, "step": 1370 }, { "epoch": 0.16, "learning_rate": 2.522242314647378e-06, "loss": 0.0008, "step": 1371 }, { "epoch": 0.16, "learning_rate": 2.5218806509945753e-06, "loss": 0.3076, "step": 1372 }, { "epoch": 0.16, "learning_rate": 2.521518987341772e-06, "loss": 0.001, "step": 1373 }, { "epoch": 0.16, "learning_rate": 2.5211573236889694e-06, "loss": 0.0023, "step": 1374 }, { "epoch": 0.16, "learning_rate": 2.5207956600361666e-06, "loss": 0.0009, "step": 1375 }, { "epoch": 0.16, "learning_rate": 2.520433996383364e-06, "loss": 0.0006, "step": 1376 }, { "epoch": 0.17, "learning_rate": 2.5200723327305607e-06, "loss": 0.3299, "step": 1377 }, { "epoch": 0.17, "learning_rate": 2.5197106690777575e-06, "loss": 0.0043, "step": 1378 }, { "epoch": 0.17, "learning_rate": 2.5193490054249547e-06, "loss": 0.0008, "step": 1379 }, { "epoch": 0.17, "learning_rate": 2.518987341772152e-06, "loss": 0.3947, "step": 1380 }, { "epoch": 0.17, "learning_rate": 2.518625678119349e-06, "loss": 0.0012, "step": 1381 }, { "epoch": 0.17, "learning_rate": 2.518264014466546e-06, "loss": 0.0007, "step": 1382 }, { "epoch": 0.17, "learning_rate": 2.517902350813743e-06, "loss": 0.0006, "step": 1383 }, { "epoch": 0.17, "learning_rate": 2.5175406871609404e-06, "loss": 0.0009, "step": 1384 }, { "epoch": 0.17, "learning_rate": 2.5171790235081377e-06, "loss": 0.0008, "step": 1385 }, { "epoch": 0.17, "learning_rate": 2.5168173598553345e-06, "loss": 0.0008, "step": 1386 }, { "epoch": 0.17, "learning_rate": 2.5164556962025317e-06, "loss": 0.0007, "step": 1387 }, { "epoch": 0.17, "learning_rate": 2.516094032549729e-06, "loss": 0.3692, "step": 1388 }, { "epoch": 0.17, "learning_rate": 2.5157323688969257e-06, "loss": 0.0009, "step": 1389 }, { "epoch": 0.17, "learning_rate": 2.515370705244123e-06, "loss": 0.0009, "step": 1390 }, { "epoch": 0.17, "learning_rate": 2.51500904159132e-06, "loss": 0.0007, "step": 1391 }, { "epoch": 0.17, "learning_rate": 2.5146473779385174e-06, "loss": 0.008, "step": 1392 }, { "epoch": 0.17, "learning_rate": 2.5142857142857142e-06, "loss": 0.0009, "step": 1393 }, { "epoch": 0.17, "learning_rate": 2.5139240506329115e-06, "loss": 0.0012, "step": 1394 }, { "epoch": 0.17, "learning_rate": 2.5135623869801087e-06, "loss": 0.0005, "step": 1395 }, { "epoch": 0.17, "learning_rate": 2.513200723327306e-06, "loss": 0.0032, "step": 1396 }, { "epoch": 0.17, "learning_rate": 2.5128390596745027e-06, "loss": 0.006, "step": 1397 }, { "epoch": 0.17, "learning_rate": 2.5124773960217e-06, "loss": 0.0007, "step": 1398 }, { "epoch": 0.17, "learning_rate": 2.512115732368897e-06, "loss": 0.0007, "step": 1399 }, { "epoch": 0.17, "learning_rate": 2.5117540687160944e-06, "loss": 0.0011, "step": 1400 }, { "epoch": 0.17, "learning_rate": 2.5113924050632912e-06, "loss": 0.0008, "step": 1401 }, { "epoch": 0.17, "learning_rate": 2.511030741410488e-06, "loss": 0.0013, "step": 1402 }, { "epoch": 0.17, "learning_rate": 2.5106690777576853e-06, "loss": 0.0007, "step": 1403 }, { "epoch": 0.17, "learning_rate": 2.5103074141048825e-06, "loss": 0.7766, "step": 1404 }, { "epoch": 0.17, "learning_rate": 2.5099457504520793e-06, "loss": 0.0007, "step": 1405 }, { "epoch": 0.17, "learning_rate": 2.5095840867992765e-06, "loss": 0.0011, "step": 1406 }, { "epoch": 0.17, "learning_rate": 2.5092224231464738e-06, "loss": 0.1515, "step": 1407 }, { "epoch": 0.17, "learning_rate": 2.508860759493671e-06, "loss": 0.0007, "step": 1408 }, { "epoch": 0.17, "learning_rate": 2.508499095840868e-06, "loss": 0.0007, "step": 1409 }, { "epoch": 0.17, "learning_rate": 2.508137432188065e-06, "loss": 0.0012, "step": 1410 }, { "epoch": 0.17, "learning_rate": 2.5077757685352623e-06, "loss": 0.8441, "step": 1411 }, { "epoch": 0.17, "learning_rate": 2.5074141048824595e-06, "loss": 0.0006, "step": 1412 }, { "epoch": 0.17, "learning_rate": 2.5070524412296563e-06, "loss": 0.0007, "step": 1413 }, { "epoch": 0.17, "learning_rate": 2.5066907775768535e-06, "loss": 0.0009, "step": 1414 }, { "epoch": 0.17, "learning_rate": 2.5063291139240508e-06, "loss": 0.0008, "step": 1415 }, { "epoch": 0.17, "learning_rate": 2.505967450271248e-06, "loss": 0.0006, "step": 1416 }, { "epoch": 0.17, "learning_rate": 2.505605786618445e-06, "loss": 0.0522, "step": 1417 }, { "epoch": 0.17, "learning_rate": 2.505244122965642e-06, "loss": 0.0008, "step": 1418 }, { "epoch": 0.17, "learning_rate": 2.5048824593128393e-06, "loss": 0.0007, "step": 1419 }, { "epoch": 0.17, "learning_rate": 2.5045207956600365e-06, "loss": 0.063, "step": 1420 }, { "epoch": 0.17, "learning_rate": 2.5041591320072333e-06, "loss": 0.3445, "step": 1421 }, { "epoch": 0.17, "learning_rate": 2.5037974683544305e-06, "loss": 0.2513, "step": 1422 }, { "epoch": 0.17, "learning_rate": 2.5034358047016278e-06, "loss": 0.0008, "step": 1423 }, { "epoch": 0.17, "learning_rate": 2.503074141048825e-06, "loss": 0.0009, "step": 1424 }, { "epoch": 0.17, "learning_rate": 2.5027124773960214e-06, "loss": 0.0005, "step": 1425 }, { "epoch": 0.17, "learning_rate": 2.5023508137432186e-06, "loss": 0.0006, "step": 1426 }, { "epoch": 0.17, "learning_rate": 2.501989150090416e-06, "loss": 0.001, "step": 1427 }, { "epoch": 0.17, "learning_rate": 2.501627486437613e-06, "loss": 0.0009, "step": 1428 }, { "epoch": 0.17, "learning_rate": 2.50126582278481e-06, "loss": 0.0728, "step": 1429 }, { "epoch": 0.17, "learning_rate": 2.500904159132007e-06, "loss": 0.0375, "step": 1430 }, { "epoch": 0.17, "learning_rate": 2.5005424954792044e-06, "loss": 0.0007, "step": 1431 }, { "epoch": 0.17, "learning_rate": 2.5001808318264016e-06, "loss": 0.0006, "step": 1432 }, { "epoch": 0.17, "learning_rate": 2.4998191681735984e-06, "loss": 0.0007, "step": 1433 }, { "epoch": 0.17, "learning_rate": 2.4994575045207956e-06, "loss": 0.0009, "step": 1434 }, { "epoch": 0.17, "learning_rate": 2.499095840867993e-06, "loss": 0.075, "step": 1435 }, { "epoch": 0.17, "learning_rate": 2.49873417721519e-06, "loss": 0.0093, "step": 1436 }, { "epoch": 0.17, "learning_rate": 2.498372513562387e-06, "loss": 0.0007, "step": 1437 }, { "epoch": 0.17, "learning_rate": 2.498010849909584e-06, "loss": 0.0006, "step": 1438 }, { "epoch": 0.17, "learning_rate": 2.4976491862567814e-06, "loss": 0.0007, "step": 1439 }, { "epoch": 0.17, "learning_rate": 2.4972875226039786e-06, "loss": 0.7793, "step": 1440 }, { "epoch": 0.17, "learning_rate": 2.4969258589511754e-06, "loss": 0.0006, "step": 1441 }, { "epoch": 0.17, "learning_rate": 2.4965641952983726e-06, "loss": 0.8945, "step": 1442 }, { "epoch": 0.17, "learning_rate": 2.49620253164557e-06, "loss": 0.0005, "step": 1443 }, { "epoch": 0.17, "learning_rate": 2.495840867992767e-06, "loss": 0.0015, "step": 1444 }, { "epoch": 0.17, "learning_rate": 2.495479204339964e-06, "loss": 0.0537, "step": 1445 }, { "epoch": 0.17, "learning_rate": 2.495117540687161e-06, "loss": 0.0007, "step": 1446 }, { "epoch": 0.17, "learning_rate": 2.4947558770343584e-06, "loss": 0.8036, "step": 1447 }, { "epoch": 0.17, "learning_rate": 2.494394213381555e-06, "loss": 0.1306, "step": 1448 }, { "epoch": 0.17, "learning_rate": 2.494032549728752e-06, "loss": 0.0006, "step": 1449 }, { "epoch": 0.17, "learning_rate": 2.493670886075949e-06, "loss": 0.0007, "step": 1450 }, { "epoch": 0.17, "learning_rate": 2.4933092224231464e-06, "loss": 0.0007, "step": 1451 }, { "epoch": 0.17, "learning_rate": 2.4929475587703437e-06, "loss": 0.4274, "step": 1452 }, { "epoch": 0.17, "learning_rate": 2.4925858951175405e-06, "loss": 0.0006, "step": 1453 }, { "epoch": 0.17, "learning_rate": 2.4922242314647377e-06, "loss": 0.0006, "step": 1454 }, { "epoch": 0.17, "learning_rate": 2.491862567811935e-06, "loss": 0.0023, "step": 1455 }, { "epoch": 0.17, "learning_rate": 2.491500904159132e-06, "loss": 0.0007, "step": 1456 }, { "epoch": 0.17, "learning_rate": 2.491139240506329e-06, "loss": 0.0007, "step": 1457 }, { "epoch": 0.17, "learning_rate": 2.490777576853526e-06, "loss": 0.0007, "step": 1458 }, { "epoch": 0.17, "learning_rate": 2.4904159132007234e-06, "loss": 0.3419, "step": 1459 }, { "epoch": 0.17, "learning_rate": 2.4900542495479207e-06, "loss": 0.0029, "step": 1460 }, { "epoch": 0.18, "learning_rate": 2.4896925858951175e-06, "loss": 0.0009, "step": 1461 }, { "epoch": 0.18, "learning_rate": 2.4893309222423147e-06, "loss": 0.001, "step": 1462 }, { "epoch": 0.18, "learning_rate": 2.488969258589512e-06, "loss": 0.8532, "step": 1463 }, { "epoch": 0.18, "learning_rate": 2.488607594936709e-06, "loss": 0.27, "step": 1464 }, { "epoch": 0.18, "learning_rate": 2.488245931283906e-06, "loss": 0.0006, "step": 1465 }, { "epoch": 0.18, "learning_rate": 2.4878842676311032e-06, "loss": 0.0007, "step": 1466 }, { "epoch": 0.18, "learning_rate": 2.4875226039783004e-06, "loss": 0.0006, "step": 1467 }, { "epoch": 0.18, "learning_rate": 2.4871609403254973e-06, "loss": 0.0548, "step": 1468 }, { "epoch": 0.18, "learning_rate": 2.4867992766726945e-06, "loss": 0.873, "step": 1469 }, { "epoch": 0.18, "learning_rate": 2.4864376130198917e-06, "loss": 0.6588, "step": 1470 }, { "epoch": 0.18, "learning_rate": 2.486075949367089e-06, "loss": 0.0012, "step": 1471 }, { "epoch": 0.18, "learning_rate": 2.4857142857142858e-06, "loss": 0.0005, "step": 1472 }, { "epoch": 0.18, "learning_rate": 2.4853526220614826e-06, "loss": 0.0009, "step": 1473 }, { "epoch": 0.18, "learning_rate": 2.48499095840868e-06, "loss": 0.8183, "step": 1474 }, { "epoch": 0.18, "learning_rate": 2.484629294755877e-06, "loss": 0.0022, "step": 1475 }, { "epoch": 0.18, "learning_rate": 2.4842676311030743e-06, "loss": 0.0019, "step": 1476 }, { "epoch": 0.18, "learning_rate": 2.483905967450271e-06, "loss": 0.0017, "step": 1477 }, { "epoch": 0.18, "learning_rate": 2.4835443037974683e-06, "loss": 0.0053, "step": 1478 }, { "epoch": 0.18, "learning_rate": 2.4831826401446655e-06, "loss": 0.125, "step": 1479 }, { "epoch": 0.18, "learning_rate": 2.4828209764918628e-06, "loss": 0.0008, "step": 1480 }, { "epoch": 0.18, "learning_rate": 2.4824593128390596e-06, "loss": 0.117, "step": 1481 }, { "epoch": 0.18, "learning_rate": 2.482097649186257e-06, "loss": 0.4171, "step": 1482 }, { "epoch": 0.18, "learning_rate": 2.481735985533454e-06, "loss": 0.0009, "step": 1483 }, { "epoch": 0.18, "learning_rate": 2.4813743218806513e-06, "loss": 0.0014, "step": 1484 }, { "epoch": 0.18, "learning_rate": 2.481012658227848e-06, "loss": 0.0009, "step": 1485 }, { "epoch": 0.18, "learning_rate": 2.4806509945750453e-06, "loss": 0.0012, "step": 1486 }, { "epoch": 0.18, "learning_rate": 2.4802893309222425e-06, "loss": 0.001, "step": 1487 }, { "epoch": 0.18, "learning_rate": 2.4799276672694393e-06, "loss": 0.0009, "step": 1488 }, { "epoch": 0.18, "learning_rate": 2.4795660036166366e-06, "loss": 0.3706, "step": 1489 }, { "epoch": 0.18, "learning_rate": 2.479204339963834e-06, "loss": 0.0006, "step": 1490 }, { "epoch": 0.18, "learning_rate": 2.478842676311031e-06, "loss": 0.0007, "step": 1491 }, { "epoch": 0.18, "learning_rate": 2.478481012658228e-06, "loss": 0.0007, "step": 1492 }, { "epoch": 0.18, "learning_rate": 2.478119349005425e-06, "loss": 0.7861, "step": 1493 }, { "epoch": 0.18, "learning_rate": 2.4777576853526223e-06, "loss": 0.0039, "step": 1494 }, { "epoch": 0.18, "learning_rate": 2.4773960216998195e-06, "loss": 0.0034, "step": 1495 }, { "epoch": 0.18, "learning_rate": 2.4770343580470163e-06, "loss": 0.0008, "step": 1496 }, { "epoch": 0.18, "learning_rate": 2.476672694394213e-06, "loss": 0.7965, "step": 1497 }, { "epoch": 0.18, "learning_rate": 2.4763110307414104e-06, "loss": 0.0009, "step": 1498 }, { "epoch": 0.18, "learning_rate": 2.4759493670886076e-06, "loss": 0.0011, "step": 1499 }, { "epoch": 0.18, "learning_rate": 2.475587703435805e-06, "loss": 0.0205, "step": 1500 }, { "epoch": 0.18, "learning_rate": 2.4752260397830016e-06, "loss": 0.4313, "step": 1501 }, { "epoch": 0.18, "learning_rate": 2.474864376130199e-06, "loss": 0.0006, "step": 1502 }, { "epoch": 0.18, "learning_rate": 2.474502712477396e-06, "loss": 0.0034, "step": 1503 }, { "epoch": 0.18, "learning_rate": 2.4741410488245933e-06, "loss": 0.0009, "step": 1504 }, { "epoch": 0.18, "learning_rate": 2.47377938517179e-06, "loss": 0.0011, "step": 1505 }, { "epoch": 0.18, "learning_rate": 2.4734177215189874e-06, "loss": 0.6256, "step": 1506 }, { "epoch": 0.18, "learning_rate": 2.4730560578661846e-06, "loss": 0.279, "step": 1507 }, { "epoch": 0.18, "learning_rate": 2.4726943942133814e-06, "loss": 0.0013, "step": 1508 }, { "epoch": 0.18, "learning_rate": 2.4723327305605786e-06, "loss": 0.0009, "step": 1509 }, { "epoch": 0.18, "learning_rate": 2.471971066907776e-06, "loss": 0.0009, "step": 1510 }, { "epoch": 0.18, "learning_rate": 2.471609403254973e-06, "loss": 0.1074, "step": 1511 }, { "epoch": 0.18, "learning_rate": 2.47124773960217e-06, "loss": 0.0025, "step": 1512 }, { "epoch": 0.18, "learning_rate": 2.470886075949367e-06, "loss": 0.0564, "step": 1513 }, { "epoch": 0.18, "learning_rate": 2.4705244122965644e-06, "loss": 0.0013, "step": 1514 }, { "epoch": 0.18, "learning_rate": 2.4701627486437616e-06, "loss": 0.624, "step": 1515 }, { "epoch": 0.18, "learning_rate": 2.4698010849909584e-06, "loss": 0.664, "step": 1516 }, { "epoch": 0.18, "learning_rate": 2.4694394213381556e-06, "loss": 0.001, "step": 1517 }, { "epoch": 0.18, "learning_rate": 2.469077757685353e-06, "loss": 0.0011, "step": 1518 }, { "epoch": 0.18, "learning_rate": 2.46871609403255e-06, "loss": 0.0013, "step": 1519 }, { "epoch": 0.18, "learning_rate": 2.468354430379747e-06, "loss": 0.0011, "step": 1520 }, { "epoch": 0.18, "learning_rate": 2.4679927667269437e-06, "loss": 0.0069, "step": 1521 }, { "epoch": 0.18, "learning_rate": 2.467631103074141e-06, "loss": 0.0007, "step": 1522 }, { "epoch": 0.18, "learning_rate": 2.467269439421338e-06, "loss": 0.0008, "step": 1523 }, { "epoch": 0.18, "learning_rate": 2.4669077757685354e-06, "loss": 0.0034, "step": 1524 }, { "epoch": 0.18, "learning_rate": 2.4665461121157322e-06, "loss": 0.0008, "step": 1525 }, { "epoch": 0.18, "learning_rate": 2.4661844484629295e-06, "loss": 0.3367, "step": 1526 }, { "epoch": 0.18, "learning_rate": 2.4658227848101267e-06, "loss": 0.0039, "step": 1527 }, { "epoch": 0.18, "learning_rate": 2.4654611211573235e-06, "loss": 0.002, "step": 1528 }, { "epoch": 0.18, "learning_rate": 2.4650994575045207e-06, "loss": 0.001, "step": 1529 }, { "epoch": 0.18, "learning_rate": 2.464737793851718e-06, "loss": 0.0005, "step": 1530 }, { "epoch": 0.18, "learning_rate": 2.464376130198915e-06, "loss": 0.0008, "step": 1531 }, { "epoch": 0.18, "learning_rate": 2.464014466546112e-06, "loss": 0.0012, "step": 1532 }, { "epoch": 0.18, "learning_rate": 2.4636528028933092e-06, "loss": 0.0005, "step": 1533 }, { "epoch": 0.18, "learning_rate": 2.4632911392405065e-06, "loss": 0.0008, "step": 1534 }, { "epoch": 0.18, "learning_rate": 2.4629294755877037e-06, "loss": 0.0901, "step": 1535 }, { "epoch": 0.18, "learning_rate": 2.4625678119349005e-06, "loss": 0.001, "step": 1536 }, { "epoch": 0.18, "learning_rate": 2.4622061482820977e-06, "loss": 0.0007, "step": 1537 }, { "epoch": 0.18, "learning_rate": 2.461844484629295e-06, "loss": 0.0005, "step": 1538 }, { "epoch": 0.18, "learning_rate": 2.461482820976492e-06, "loss": 0.001, "step": 1539 }, { "epoch": 0.18, "learning_rate": 2.461121157323689e-06, "loss": 0.0005, "step": 1540 }, { "epoch": 0.18, "learning_rate": 2.4607594936708862e-06, "loss": 0.0005, "step": 1541 }, { "epoch": 0.18, "learning_rate": 2.4603978300180835e-06, "loss": 0.0006, "step": 1542 }, { "epoch": 0.18, "learning_rate": 2.4600361663652807e-06, "loss": 0.0004, "step": 1543 }, { "epoch": 0.19, "learning_rate": 2.4596745027124775e-06, "loss": 0.0232, "step": 1544 }, { "epoch": 0.19, "learning_rate": 2.4593128390596743e-06, "loss": 0.7462, "step": 1545 }, { "epoch": 0.19, "learning_rate": 2.4589511754068715e-06, "loss": 0.001, "step": 1546 }, { "epoch": 0.19, "learning_rate": 2.4585895117540688e-06, "loss": 0.0009, "step": 1547 }, { "epoch": 0.19, "learning_rate": 2.4582278481012656e-06, "loss": 0.0006, "step": 1548 }, { "epoch": 0.19, "learning_rate": 2.457866184448463e-06, "loss": 0.0006, "step": 1549 }, { "epoch": 0.19, "learning_rate": 2.45750452079566e-06, "loss": 0.0005, "step": 1550 }, { "epoch": 0.19, "learning_rate": 2.4571428571428573e-06, "loss": 0.0009, "step": 1551 }, { "epoch": 0.19, "learning_rate": 2.456781193490054e-06, "loss": 0.0229, "step": 1552 }, { "epoch": 0.19, "learning_rate": 2.4564195298372513e-06, "loss": 0.0005, "step": 1553 }, { "epoch": 0.19, "learning_rate": 2.4560578661844485e-06, "loss": 0.0006, "step": 1554 }, { "epoch": 0.19, "learning_rate": 2.4556962025316458e-06, "loss": 0.904, "step": 1555 }, { "epoch": 0.19, "learning_rate": 2.4553345388788426e-06, "loss": 0.0007, "step": 1556 }, { "epoch": 0.19, "learning_rate": 2.45497287522604e-06, "loss": 0.7207, "step": 1557 }, { "epoch": 0.19, "learning_rate": 2.454611211573237e-06, "loss": 0.0006, "step": 1558 }, { "epoch": 0.19, "learning_rate": 2.4542495479204343e-06, "loss": 0.0006, "step": 1559 }, { "epoch": 0.19, "learning_rate": 2.453887884267631e-06, "loss": 0.0007, "step": 1560 }, { "epoch": 0.19, "learning_rate": 2.4535262206148283e-06, "loss": 0.0051, "step": 1561 }, { "epoch": 0.19, "learning_rate": 2.4531645569620255e-06, "loss": 0.0008, "step": 1562 }, { "epoch": 0.19, "learning_rate": 2.4528028933092228e-06, "loss": 0.0008, "step": 1563 }, { "epoch": 0.19, "learning_rate": 2.4524412296564196e-06, "loss": 0.0005, "step": 1564 }, { "epoch": 0.19, "learning_rate": 2.452079566003617e-06, "loss": 0.0005, "step": 1565 }, { "epoch": 0.19, "learning_rate": 2.451717902350814e-06, "loss": 0.0007, "step": 1566 }, { "epoch": 0.19, "learning_rate": 2.451356238698011e-06, "loss": 0.0008, "step": 1567 }, { "epoch": 0.19, "learning_rate": 2.4509945750452077e-06, "loss": 0.0006, "step": 1568 }, { "epoch": 0.19, "learning_rate": 2.450632911392405e-06, "loss": 0.0005, "step": 1569 }, { "epoch": 0.19, "learning_rate": 2.450271247739602e-06, "loss": 0.0005, "step": 1570 }, { "epoch": 0.19, "learning_rate": 2.4499095840867994e-06, "loss": 0.0004, "step": 1571 }, { "epoch": 0.19, "learning_rate": 2.449547920433996e-06, "loss": 0.0011, "step": 1572 }, { "epoch": 0.19, "learning_rate": 2.4491862567811934e-06, "loss": 0.0006, "step": 1573 }, { "epoch": 0.19, "learning_rate": 2.4488245931283906e-06, "loss": 0.0007, "step": 1574 }, { "epoch": 0.19, "learning_rate": 2.448462929475588e-06, "loss": 0.094, "step": 1575 }, { "epoch": 0.19, "learning_rate": 2.4481012658227847e-06, "loss": 0.0005, "step": 1576 }, { "epoch": 0.19, "learning_rate": 2.447739602169982e-06, "loss": 0.0007, "step": 1577 }, { "epoch": 0.19, "learning_rate": 2.447377938517179e-06, "loss": 0.0005, "step": 1578 }, { "epoch": 0.19, "learning_rate": 2.4470162748643764e-06, "loss": 0.0005, "step": 1579 }, { "epoch": 0.19, "learning_rate": 2.446654611211573e-06, "loss": 0.914, "step": 1580 }, { "epoch": 0.19, "learning_rate": 2.4462929475587704e-06, "loss": 0.0005, "step": 1581 }, { "epoch": 0.19, "learning_rate": 2.4459312839059676e-06, "loss": 0.0009, "step": 1582 }, { "epoch": 0.19, "learning_rate": 2.445569620253165e-06, "loss": 0.1151, "step": 1583 }, { "epoch": 0.19, "learning_rate": 2.4452079566003617e-06, "loss": 0.0006, "step": 1584 }, { "epoch": 0.19, "learning_rate": 2.444846292947559e-06, "loss": 0.0004, "step": 1585 }, { "epoch": 0.19, "learning_rate": 2.444484629294756e-06, "loss": 0.0016, "step": 1586 }, { "epoch": 0.19, "learning_rate": 2.444122965641953e-06, "loss": 0.0004, "step": 1587 }, { "epoch": 0.19, "learning_rate": 2.44376130198915e-06, "loss": 0.0004, "step": 1588 }, { "epoch": 0.19, "learning_rate": 2.4433996383363474e-06, "loss": 0.0009, "step": 1589 }, { "epoch": 0.19, "learning_rate": 2.4430379746835446e-06, "loss": 0.0005, "step": 1590 }, { "epoch": 0.19, "learning_rate": 2.4426763110307414e-06, "loss": 0.0083, "step": 1591 }, { "epoch": 0.19, "learning_rate": 2.4423146473779382e-06, "loss": 0.0004, "step": 1592 }, { "epoch": 0.19, "learning_rate": 2.4419529837251355e-06, "loss": 0.0737, "step": 1593 }, { "epoch": 0.19, "learning_rate": 2.4415913200723327e-06, "loss": 0.0008, "step": 1594 }, { "epoch": 0.19, "learning_rate": 2.44122965641953e-06, "loss": 0.0006, "step": 1595 }, { "epoch": 0.19, "learning_rate": 2.4408679927667267e-06, "loss": 0.0005, "step": 1596 }, { "epoch": 0.19, "learning_rate": 2.440506329113924e-06, "loss": 0.0006, "step": 1597 }, { "epoch": 0.19, "learning_rate": 2.440144665461121e-06, "loss": 0.3038, "step": 1598 }, { "epoch": 0.19, "learning_rate": 2.4397830018083184e-06, "loss": 0.001, "step": 1599 }, { "epoch": 0.19, "learning_rate": 2.4394213381555152e-06, "loss": 0.8537, "step": 1600 }, { "epoch": 0.19, "learning_rate": 2.4390596745027125e-06, "loss": 0.0005, "step": 1601 }, { "epoch": 0.19, "learning_rate": 2.4386980108499097e-06, "loss": 0.0005, "step": 1602 }, { "epoch": 0.19, "learning_rate": 2.438336347197107e-06, "loss": 0.0007, "step": 1603 }, { "epoch": 0.19, "learning_rate": 2.4379746835443037e-06, "loss": 0.871, "step": 1604 }, { "epoch": 0.19, "learning_rate": 2.437613019891501e-06, "loss": 0.8406, "step": 1605 }, { "epoch": 0.19, "learning_rate": 2.437251356238698e-06, "loss": 0.7218, "step": 1606 }, { "epoch": 0.19, "learning_rate": 2.436889692585895e-06, "loss": 0.0006, "step": 1607 }, { "epoch": 0.19, "learning_rate": 2.4365280289330922e-06, "loss": 0.0006, "step": 1608 }, { "epoch": 0.19, "learning_rate": 2.4361663652802895e-06, "loss": 0.0008, "step": 1609 }, { "epoch": 0.19, "learning_rate": 2.4358047016274867e-06, "loss": 0.0005, "step": 1610 }, { "epoch": 0.19, "learning_rate": 2.4354430379746835e-06, "loss": 0.0006, "step": 1611 }, { "epoch": 0.19, "learning_rate": 2.4350813743218808e-06, "loss": 0.0005, "step": 1612 }, { "epoch": 0.19, "learning_rate": 2.434719710669078e-06, "loss": 0.0018, "step": 1613 }, { "epoch": 0.19, "learning_rate": 2.4343580470162752e-06, "loss": 0.0283, "step": 1614 }, { "epoch": 0.19, "learning_rate": 2.433996383363472e-06, "loss": 1.2287, "step": 1615 }, { "epoch": 0.19, "learning_rate": 2.433634719710669e-06, "loss": 0.3696, "step": 1616 }, { "epoch": 0.19, "learning_rate": 2.433273056057866e-06, "loss": 0.0006, "step": 1617 }, { "epoch": 0.19, "learning_rate": 2.4329113924050633e-06, "loss": 0.0006, "step": 1618 }, { "epoch": 0.19, "learning_rate": 2.4325497287522605e-06, "loss": 0.0005, "step": 1619 }, { "epoch": 0.19, "learning_rate": 2.4321880650994573e-06, "loss": 0.0007, "step": 1620 }, { "epoch": 0.19, "learning_rate": 2.4318264014466546e-06, "loss": 0.0005, "step": 1621 }, { "epoch": 0.19, "learning_rate": 2.431464737793852e-06, "loss": 0.0005, "step": 1622 }, { "epoch": 0.19, "learning_rate": 2.431103074141049e-06, "loss": 0.0022, "step": 1623 }, { "epoch": 0.19, "learning_rate": 2.430741410488246e-06, "loss": 0.0011, "step": 1624 }, { "epoch": 0.19, "learning_rate": 2.430379746835443e-06, "loss": 0.0006, "step": 1625 }, { "epoch": 0.19, "learning_rate": 2.4300180831826403e-06, "loss": 0.0019, "step": 1626 }, { "epoch": 0.19, "learning_rate": 2.429656419529837e-06, "loss": 0.0006, "step": 1627 }, { "epoch": 0.2, "learning_rate": 2.4292947558770343e-06, "loss": 0.7804, "step": 1628 }, { "epoch": 0.2, "learning_rate": 2.4289330922242316e-06, "loss": 0.0005, "step": 1629 }, { "epoch": 0.2, "learning_rate": 2.428571428571429e-06, "loss": 0.0015, "step": 1630 }, { "epoch": 0.2, "learning_rate": 2.4282097649186256e-06, "loss": 0.8847, "step": 1631 }, { "epoch": 0.2, "learning_rate": 2.427848101265823e-06, "loss": 0.0005, "step": 1632 }, { "epoch": 0.2, "learning_rate": 2.42748643761302e-06, "loss": 0.3242, "step": 1633 }, { "epoch": 0.2, "learning_rate": 2.4271247739602173e-06, "loss": 0.0008, "step": 1634 }, { "epoch": 0.2, "learning_rate": 2.426763110307414e-06, "loss": 0.0005, "step": 1635 }, { "epoch": 0.2, "learning_rate": 2.4264014466546113e-06, "loss": 0.5854, "step": 1636 }, { "epoch": 0.2, "learning_rate": 2.4260397830018086e-06, "loss": 0.0005, "step": 1637 }, { "epoch": 0.2, "learning_rate": 2.425678119349006e-06, "loss": 0.0007, "step": 1638 }, { "epoch": 0.2, "learning_rate": 2.4253164556962026e-06, "loss": 0.0005, "step": 1639 }, { "epoch": 0.2, "learning_rate": 2.4249547920433994e-06, "loss": 0.0005, "step": 1640 }, { "epoch": 0.2, "learning_rate": 2.4245931283905966e-06, "loss": 0.0005, "step": 1641 }, { "epoch": 0.2, "learning_rate": 2.424231464737794e-06, "loss": 0.0019, "step": 1642 }, { "epoch": 0.2, "learning_rate": 2.423869801084991e-06, "loss": 0.1418, "step": 1643 }, { "epoch": 0.2, "learning_rate": 2.423508137432188e-06, "loss": 0.9828, "step": 1644 }, { "epoch": 0.2, "learning_rate": 2.423146473779385e-06, "loss": 0.0014, "step": 1645 }, { "epoch": 0.2, "learning_rate": 2.4227848101265824e-06, "loss": 0.0008, "step": 1646 }, { "epoch": 0.2, "learning_rate": 2.422423146473779e-06, "loss": 0.0005, "step": 1647 }, { "epoch": 0.2, "learning_rate": 2.4220614828209764e-06, "loss": 0.0005, "step": 1648 }, { "epoch": 0.2, "learning_rate": 2.4216998191681736e-06, "loss": 0.098, "step": 1649 }, { "epoch": 0.2, "learning_rate": 2.421338155515371e-06, "loss": 0.0005, "step": 1650 }, { "epoch": 0.2, "learning_rate": 2.4209764918625677e-06, "loss": 0.0011, "step": 1651 }, { "epoch": 0.2, "learning_rate": 2.420614828209765e-06, "loss": 0.0007, "step": 1652 }, { "epoch": 0.2, "learning_rate": 2.420253164556962e-06, "loss": 0.0776, "step": 1653 }, { "epoch": 0.2, "learning_rate": 2.4198915009041594e-06, "loss": 0.0006, "step": 1654 }, { "epoch": 0.2, "learning_rate": 2.419529837251356e-06, "loss": 0.0007, "step": 1655 }, { "epoch": 0.2, "learning_rate": 2.4191681735985534e-06, "loss": 0.0009, "step": 1656 }, { "epoch": 0.2, "learning_rate": 2.4188065099457506e-06, "loss": 0.9879, "step": 1657 }, { "epoch": 0.2, "learning_rate": 2.418444846292948e-06, "loss": 0.0025, "step": 1658 }, { "epoch": 0.2, "learning_rate": 2.4180831826401447e-06, "loss": 0.0029, "step": 1659 }, { "epoch": 0.2, "learning_rate": 2.417721518987342e-06, "loss": 0.0007, "step": 1660 }, { "epoch": 0.2, "learning_rate": 2.417359855334539e-06, "loss": 0.0054, "step": 1661 }, { "epoch": 0.2, "learning_rate": 2.4169981916817364e-06, "loss": 0.0024, "step": 1662 }, { "epoch": 0.2, "learning_rate": 2.416636528028933e-06, "loss": 0.0005, "step": 1663 }, { "epoch": 0.2, "learning_rate": 2.41627486437613e-06, "loss": 0.0017, "step": 1664 }, { "epoch": 0.2, "learning_rate": 2.4159132007233272e-06, "loss": 0.0005, "step": 1665 }, { "epoch": 0.2, "learning_rate": 2.4155515370705245e-06, "loss": 0.0009, "step": 1666 }, { "epoch": 0.2, "learning_rate": 2.4151898734177213e-06, "loss": 1.5253, "step": 1667 }, { "epoch": 0.2, "learning_rate": 2.4148282097649185e-06, "loss": 0.0005, "step": 1668 }, { "epoch": 0.2, "learning_rate": 2.4144665461121157e-06, "loss": 0.0006, "step": 1669 }, { "epoch": 0.2, "learning_rate": 2.414104882459313e-06, "loss": 0.6573, "step": 1670 }, { "epoch": 0.2, "learning_rate": 2.4137432188065098e-06, "loss": 0.0018, "step": 1671 }, { "epoch": 0.2, "learning_rate": 2.413381555153707e-06, "loss": 0.0005, "step": 1672 }, { "epoch": 0.2, "learning_rate": 2.4130198915009042e-06, "loss": 0.0014, "step": 1673 }, { "epoch": 0.2, "learning_rate": 2.4126582278481015e-06, "loss": 0.0006, "step": 1674 }, { "epoch": 0.2, "learning_rate": 2.4122965641952983e-06, "loss": 0.0232, "step": 1675 }, { "epoch": 0.2, "learning_rate": 2.4119349005424955e-06, "loss": 0.0008, "step": 1676 }, { "epoch": 0.2, "learning_rate": 2.4115732368896927e-06, "loss": 0.0013, "step": 1677 }, { "epoch": 0.2, "learning_rate": 2.41121157323689e-06, "loss": 0.0007, "step": 1678 }, { "epoch": 0.2, "learning_rate": 2.4108499095840868e-06, "loss": 0.001, "step": 1679 }, { "epoch": 0.2, "learning_rate": 2.410488245931284e-06, "loss": 0.0004, "step": 1680 }, { "epoch": 0.2, "learning_rate": 2.4101265822784812e-06, "loss": 0.8901, "step": 1681 }, { "epoch": 0.2, "learning_rate": 2.4097649186256785e-06, "loss": 0.0005, "step": 1682 }, { "epoch": 0.2, "learning_rate": 2.4094032549728753e-06, "loss": 0.0005, "step": 1683 }, { "epoch": 0.2, "learning_rate": 2.4090415913200725e-06, "loss": 0.0008, "step": 1684 }, { "epoch": 0.2, "learning_rate": 2.4086799276672697e-06, "loss": 0.6902, "step": 1685 }, { "epoch": 0.2, "learning_rate": 2.408318264014467e-06, "loss": 0.0016, "step": 1686 }, { "epoch": 0.2, "learning_rate": 2.4079566003616638e-06, "loss": 0.0008, "step": 1687 }, { "epoch": 0.2, "learning_rate": 2.4075949367088606e-06, "loss": 0.0005, "step": 1688 }, { "epoch": 0.2, "learning_rate": 2.407233273056058e-06, "loss": 0.0005, "step": 1689 }, { "epoch": 0.2, "learning_rate": 2.406871609403255e-06, "loss": 0.0006, "step": 1690 }, { "epoch": 0.2, "learning_rate": 2.406509945750452e-06, "loss": 0.0008, "step": 1691 }, { "epoch": 0.2, "learning_rate": 2.406148282097649e-06, "loss": 0.0006, "step": 1692 }, { "epoch": 0.2, "learning_rate": 2.4057866184448463e-06, "loss": 0.0008, "step": 1693 }, { "epoch": 0.2, "learning_rate": 2.4054249547920435e-06, "loss": 0.0007, "step": 1694 }, { "epoch": 0.2, "learning_rate": 2.4050632911392403e-06, "loss": 0.0006, "step": 1695 }, { "epoch": 0.2, "learning_rate": 2.4047016274864376e-06, "loss": 0.3372, "step": 1696 }, { "epoch": 0.2, "learning_rate": 2.404339963833635e-06, "loss": 0.0007, "step": 1697 }, { "epoch": 0.2, "learning_rate": 2.403978300180832e-06, "loss": 0.0006, "step": 1698 }, { "epoch": 0.2, "learning_rate": 2.403616636528029e-06, "loss": 0.0006, "step": 1699 }, { "epoch": 0.2, "learning_rate": 2.403254972875226e-06, "loss": 0.0006, "step": 1700 }, { "epoch": 0.2, "learning_rate": 2.4028933092224233e-06, "loss": 0.0007, "step": 1701 }, { "epoch": 0.2, "learning_rate": 2.4025316455696205e-06, "loss": 0.0006, "step": 1702 }, { "epoch": 0.2, "learning_rate": 2.4021699819168174e-06, "loss": 0.0006, "step": 1703 }, { "epoch": 0.2, "learning_rate": 2.4018083182640146e-06, "loss": 0.0017, "step": 1704 }, { "epoch": 0.2, "learning_rate": 2.401446654611212e-06, "loss": 0.032, "step": 1705 }, { "epoch": 0.2, "learning_rate": 2.4010849909584086e-06, "loss": 0.0017, "step": 1706 }, { "epoch": 0.2, "learning_rate": 2.400723327305606e-06, "loss": 0.0008, "step": 1707 }, { "epoch": 0.2, "learning_rate": 2.400361663652803e-06, "loss": 0.0006, "step": 1708 }, { "epoch": 0.2, "learning_rate": 2.4000000000000003e-06, "loss": 0.2097, "step": 1709 }, { "epoch": 0.2, "learning_rate": 2.399638336347197e-06, "loss": 0.9601, "step": 1710 }, { "epoch": 0.21, "learning_rate": 2.3992766726943944e-06, "loss": 0.0005, "step": 1711 }, { "epoch": 0.21, "learning_rate": 2.398915009041591e-06, "loss": 0.0007, "step": 1712 }, { "epoch": 0.21, "learning_rate": 2.3985533453887884e-06, "loss": 0.0006, "step": 1713 }, { "epoch": 0.21, "learning_rate": 2.3981916817359856e-06, "loss": 0.0043, "step": 1714 }, { "epoch": 0.21, "learning_rate": 2.3978300180831824e-06, "loss": 0.0007, "step": 1715 }, { "epoch": 0.21, "learning_rate": 2.3974683544303797e-06, "loss": 0.0084, "step": 1716 }, { "epoch": 0.21, "learning_rate": 2.397106690777577e-06, "loss": 0.0063, "step": 1717 }, { "epoch": 0.21, "learning_rate": 2.396745027124774e-06, "loss": 0.0007, "step": 1718 }, { "epoch": 0.21, "learning_rate": 2.396383363471971e-06, "loss": 0.0006, "step": 1719 }, { "epoch": 0.21, "learning_rate": 2.396021699819168e-06, "loss": 0.5174, "step": 1720 }, { "epoch": 0.21, "learning_rate": 2.3956600361663654e-06, "loss": 0.0006, "step": 1721 }, { "epoch": 0.21, "learning_rate": 2.3952983725135626e-06, "loss": 0.0025, "step": 1722 }, { "epoch": 0.21, "learning_rate": 2.3949367088607594e-06, "loss": 0.0008, "step": 1723 }, { "epoch": 0.21, "learning_rate": 2.3945750452079567e-06, "loss": 0.0007, "step": 1724 }, { "epoch": 0.21, "learning_rate": 2.394213381555154e-06, "loss": 0.0006, "step": 1725 }, { "epoch": 0.21, "learning_rate": 2.3938517179023507e-06, "loss": 0.0008, "step": 1726 }, { "epoch": 0.21, "learning_rate": 2.393490054249548e-06, "loss": 0.1579, "step": 1727 }, { "epoch": 0.21, "learning_rate": 2.393128390596745e-06, "loss": 0.0006, "step": 1728 }, { "epoch": 0.21, "learning_rate": 2.3927667269439424e-06, "loss": 0.0006, "step": 1729 }, { "epoch": 0.21, "learning_rate": 2.392405063291139e-06, "loss": 0.0005, "step": 1730 }, { "epoch": 0.21, "learning_rate": 2.3920433996383364e-06, "loss": 0.0012, "step": 1731 }, { "epoch": 0.21, "learning_rate": 2.3916817359855337e-06, "loss": 0.0006, "step": 1732 }, { "epoch": 0.21, "learning_rate": 2.391320072332731e-06, "loss": 0.0005, "step": 1733 }, { "epoch": 0.21, "learning_rate": 2.3909584086799277e-06, "loss": 0.0048, "step": 1734 }, { "epoch": 0.21, "learning_rate": 2.390596745027125e-06, "loss": 0.0008, "step": 1735 }, { "epoch": 0.21, "learning_rate": 2.3902350813743217e-06, "loss": 0.0007, "step": 1736 }, { "epoch": 0.21, "learning_rate": 2.389873417721519e-06, "loss": 0.0006, "step": 1737 }, { "epoch": 0.21, "learning_rate": 2.389511754068716e-06, "loss": 0.0007, "step": 1738 }, { "epoch": 0.21, "learning_rate": 2.389150090415913e-06, "loss": 0.0004, "step": 1739 }, { "epoch": 0.21, "learning_rate": 2.3887884267631102e-06, "loss": 0.0005, "step": 1740 }, { "epoch": 0.21, "learning_rate": 2.3884267631103075e-06, "loss": 0.0793, "step": 1741 }, { "epoch": 0.21, "learning_rate": 2.3880650994575047e-06, "loss": 0.0007, "step": 1742 }, { "epoch": 0.21, "learning_rate": 2.3877034358047015e-06, "loss": 0.0006, "step": 1743 }, { "epoch": 0.21, "learning_rate": 2.3873417721518987e-06, "loss": 0.0006, "step": 1744 }, { "epoch": 0.21, "learning_rate": 2.386980108499096e-06, "loss": 0.0007, "step": 1745 }, { "epoch": 0.21, "learning_rate": 2.3866184448462928e-06, "loss": 0.7995, "step": 1746 }, { "epoch": 0.21, "learning_rate": 2.38625678119349e-06, "loss": 0.0436, "step": 1747 }, { "epoch": 0.21, "learning_rate": 2.3858951175406872e-06, "loss": 0.0006, "step": 1748 }, { "epoch": 0.21, "learning_rate": 2.3855334538878845e-06, "loss": 0.0006, "step": 1749 }, { "epoch": 0.21, "learning_rate": 2.3851717902350813e-06, "loss": 0.0005, "step": 1750 }, { "epoch": 0.21, "learning_rate": 2.3848101265822785e-06, "loss": 0.0005, "step": 1751 }, { "epoch": 0.21, "learning_rate": 2.3844484629294757e-06, "loss": 0.0005, "step": 1752 }, { "epoch": 0.21, "learning_rate": 2.384086799276673e-06, "loss": 0.0007, "step": 1753 }, { "epoch": 0.21, "learning_rate": 2.3837251356238698e-06, "loss": 0.001, "step": 1754 }, { "epoch": 0.21, "learning_rate": 2.383363471971067e-06, "loss": 0.0023, "step": 1755 }, { "epoch": 0.21, "learning_rate": 2.3830018083182643e-06, "loss": 0.0008, "step": 1756 }, { "epoch": 0.21, "learning_rate": 2.3826401446654615e-06, "loss": 0.0032, "step": 1757 }, { "epoch": 0.21, "learning_rate": 2.3822784810126583e-06, "loss": 0.0006, "step": 1758 }, { "epoch": 0.21, "learning_rate": 2.3819168173598555e-06, "loss": 0.1279, "step": 1759 }, { "epoch": 0.21, "learning_rate": 2.3815551537070523e-06, "loss": 0.0005, "step": 1760 }, { "epoch": 0.21, "learning_rate": 2.3811934900542496e-06, "loss": 0.0006, "step": 1761 }, { "epoch": 0.21, "learning_rate": 2.380831826401447e-06, "loss": 0.0006, "step": 1762 }, { "epoch": 0.21, "learning_rate": 2.3804701627486436e-06, "loss": 0.0018, "step": 1763 }, { "epoch": 0.21, "learning_rate": 2.380108499095841e-06, "loss": 0.0006, "step": 1764 }, { "epoch": 0.21, "learning_rate": 2.379746835443038e-06, "loss": 0.0004, "step": 1765 }, { "epoch": 0.21, "learning_rate": 2.379385171790235e-06, "loss": 0.0004, "step": 1766 }, { "epoch": 0.21, "learning_rate": 2.379023508137432e-06, "loss": 0.0007, "step": 1767 }, { "epoch": 0.21, "learning_rate": 2.3786618444846293e-06, "loss": 0.0006, "step": 1768 }, { "epoch": 0.21, "learning_rate": 2.3783001808318266e-06, "loss": 0.0008, "step": 1769 }, { "epoch": 0.21, "learning_rate": 2.3779385171790234e-06, "loss": 0.0007, "step": 1770 }, { "epoch": 0.21, "learning_rate": 2.3775768535262206e-06, "loss": 0.0006, "step": 1771 }, { "epoch": 0.21, "learning_rate": 2.377215189873418e-06, "loss": 0.0005, "step": 1772 }, { "epoch": 0.21, "learning_rate": 2.376853526220615e-06, "loss": 0.0005, "step": 1773 }, { "epoch": 0.21, "learning_rate": 2.376491862567812e-06, "loss": 0.6038, "step": 1774 }, { "epoch": 0.21, "learning_rate": 2.376130198915009e-06, "loss": 0.0007, "step": 1775 }, { "epoch": 0.21, "learning_rate": 2.3757685352622063e-06, "loss": 0.0004, "step": 1776 }, { "epoch": 0.21, "learning_rate": 2.3754068716094036e-06, "loss": 0.0006, "step": 1777 }, { "epoch": 0.21, "learning_rate": 2.3750452079566004e-06, "loss": 0.8566, "step": 1778 }, { "epoch": 0.21, "learning_rate": 2.3746835443037976e-06, "loss": 0.0204, "step": 1779 }, { "epoch": 0.21, "learning_rate": 2.374321880650995e-06, "loss": 0.0005, "step": 1780 }, { "epoch": 0.21, "learning_rate": 2.373960216998192e-06, "loss": 0.0046, "step": 1781 }, { "epoch": 0.21, "learning_rate": 2.373598553345389e-06, "loss": 0.0007, "step": 1782 }, { "epoch": 0.21, "learning_rate": 2.373236889692586e-06, "loss": 0.0004, "step": 1783 }, { "epoch": 0.21, "learning_rate": 2.372875226039783e-06, "loss": 0.0017, "step": 1784 }, { "epoch": 0.21, "learning_rate": 2.37251356238698e-06, "loss": 0.0005, "step": 1785 }, { "epoch": 0.21, "learning_rate": 2.372151898734177e-06, "loss": 0.0005, "step": 1786 }, { "epoch": 0.21, "learning_rate": 2.371790235081374e-06, "loss": 0.0005, "step": 1787 }, { "epoch": 0.21, "learning_rate": 2.3714285714285714e-06, "loss": 0.0005, "step": 1788 }, { "epoch": 0.21, "learning_rate": 2.3710669077757686e-06, "loss": 0.0004, "step": 1789 }, { "epoch": 0.21, "learning_rate": 2.3707052441229655e-06, "loss": 0.0004, "step": 1790 }, { "epoch": 0.21, "learning_rate": 2.3703435804701627e-06, "loss": 0.0007, "step": 1791 }, { "epoch": 0.21, "learning_rate": 2.36998191681736e-06, "loss": 0.0005, "step": 1792 }, { "epoch": 0.21, "learning_rate": 2.369620253164557e-06, "loss": 0.0005, "step": 1793 }, { "epoch": 0.21, "learning_rate": 2.369258589511754e-06, "loss": 0.0004, "step": 1794 }, { "epoch": 0.22, "learning_rate": 2.368896925858951e-06, "loss": 0.0005, "step": 1795 }, { "epoch": 0.22, "learning_rate": 2.3685352622061484e-06, "loss": 0.0047, "step": 1796 }, { "epoch": 0.22, "learning_rate": 2.3681735985533456e-06, "loss": 0.0014, "step": 1797 }, { "epoch": 0.22, "learning_rate": 2.3678119349005425e-06, "loss": 0.0005, "step": 1798 }, { "epoch": 0.22, "learning_rate": 2.3674502712477397e-06, "loss": 0.0006, "step": 1799 }, { "epoch": 0.22, "learning_rate": 2.367088607594937e-06, "loss": 0.0004, "step": 1800 }, { "epoch": 0.22, "learning_rate": 2.366726943942134e-06, "loss": 0.9257, "step": 1801 }, { "epoch": 0.22, "learning_rate": 2.366365280289331e-06, "loss": 0.0004, "step": 1802 }, { "epoch": 0.22, "learning_rate": 2.366003616636528e-06, "loss": 0.0007, "step": 1803 }, { "epoch": 0.22, "learning_rate": 2.3656419529837254e-06, "loss": 0.0004, "step": 1804 }, { "epoch": 0.22, "learning_rate": 2.3652802893309226e-06, "loss": 0.0005, "step": 1805 }, { "epoch": 0.22, "learning_rate": 2.3649186256781195e-06, "loss": 0.0005, "step": 1806 }, { "epoch": 0.22, "learning_rate": 2.3645569620253167e-06, "loss": 0.0004, "step": 1807 }, { "epoch": 0.22, "learning_rate": 2.3641952983725135e-06, "loss": 0.0008, "step": 1808 }, { "epoch": 0.22, "learning_rate": 2.3638336347197107e-06, "loss": 0.0005, "step": 1809 }, { "epoch": 0.22, "learning_rate": 2.3634719710669075e-06, "loss": 0.0005, "step": 1810 }, { "epoch": 0.22, "learning_rate": 2.3631103074141048e-06, "loss": 0.0005, "step": 1811 }, { "epoch": 0.22, "learning_rate": 2.362748643761302e-06, "loss": 0.7247, "step": 1812 }, { "epoch": 0.22, "learning_rate": 2.3623869801084992e-06, "loss": 0.0005, "step": 1813 }, { "epoch": 0.22, "learning_rate": 2.362025316455696e-06, "loss": 0.088, "step": 1814 }, { "epoch": 0.22, "learning_rate": 2.3616636528028933e-06, "loss": 0.0015, "step": 1815 }, { "epoch": 0.22, "learning_rate": 2.3613019891500905e-06, "loss": 0.0113, "step": 1816 }, { "epoch": 0.22, "learning_rate": 2.3609403254972877e-06, "loss": 0.0523, "step": 1817 }, { "epoch": 0.22, "learning_rate": 2.3605786618444845e-06, "loss": 0.0017, "step": 1818 }, { "epoch": 0.22, "learning_rate": 2.3602169981916818e-06, "loss": 0.0005, "step": 1819 }, { "epoch": 0.22, "learning_rate": 2.359855334538879e-06, "loss": 0.0008, "step": 1820 }, { "epoch": 0.22, "learning_rate": 2.3594936708860762e-06, "loss": 0.0005, "step": 1821 }, { "epoch": 0.22, "learning_rate": 2.359132007233273e-06, "loss": 0.0005, "step": 1822 }, { "epoch": 0.22, "learning_rate": 2.3587703435804703e-06, "loss": 0.0005, "step": 1823 }, { "epoch": 0.22, "learning_rate": 2.3584086799276675e-06, "loss": 0.0005, "step": 1824 }, { "epoch": 0.22, "learning_rate": 2.3580470162748647e-06, "loss": 0.1467, "step": 1825 }, { "epoch": 0.22, "learning_rate": 2.3576853526220615e-06, "loss": 0.7237, "step": 1826 }, { "epoch": 0.22, "learning_rate": 2.3573236889692588e-06, "loss": 0.0008, "step": 1827 }, { "epoch": 0.22, "learning_rate": 2.356962025316456e-06, "loss": 0.1902, "step": 1828 }, { "epoch": 0.22, "learning_rate": 2.356600361663653e-06, "loss": 0.985, "step": 1829 }, { "epoch": 0.22, "learning_rate": 2.35623869801085e-06, "loss": 0.0004, "step": 1830 }, { "epoch": 0.22, "learning_rate": 2.355877034358047e-06, "loss": 0.0013, "step": 1831 }, { "epoch": 0.22, "learning_rate": 2.355515370705244e-06, "loss": 0.0006, "step": 1832 }, { "epoch": 0.22, "learning_rate": 2.3551537070524413e-06, "loss": 0.5635, "step": 1833 }, { "epoch": 0.22, "learning_rate": 2.354792043399638e-06, "loss": 0.0005, "step": 1834 }, { "epoch": 0.22, "learning_rate": 2.3544303797468353e-06, "loss": 0.0005, "step": 1835 }, { "epoch": 0.22, "learning_rate": 2.3540687160940326e-06, "loss": 0.0008, "step": 1836 }, { "epoch": 0.22, "learning_rate": 2.35370705244123e-06, "loss": 0.0005, "step": 1837 }, { "epoch": 0.22, "learning_rate": 2.3533453887884266e-06, "loss": 0.0004, "step": 1838 }, { "epoch": 0.22, "learning_rate": 2.352983725135624e-06, "loss": 0.0339, "step": 1839 }, { "epoch": 0.22, "learning_rate": 2.352622061482821e-06, "loss": 0.0006, "step": 1840 }, { "epoch": 0.22, "learning_rate": 2.3522603978300183e-06, "loss": 0.0006, "step": 1841 }, { "epoch": 0.22, "learning_rate": 2.351898734177215e-06, "loss": 0.0161, "step": 1842 }, { "epoch": 0.22, "learning_rate": 2.3515370705244123e-06, "loss": 0.0004, "step": 1843 }, { "epoch": 0.22, "learning_rate": 2.3511754068716096e-06, "loss": 0.0005, "step": 1844 }, { "epoch": 0.22, "learning_rate": 2.3508137432188064e-06, "loss": 0.0006, "step": 1845 }, { "epoch": 0.22, "learning_rate": 2.3504520795660036e-06, "loss": 0.0007, "step": 1846 }, { "epoch": 0.22, "learning_rate": 2.350090415913201e-06, "loss": 0.0006, "step": 1847 }, { "epoch": 0.22, "learning_rate": 2.349728752260398e-06, "loss": 0.0006, "step": 1848 }, { "epoch": 0.22, "learning_rate": 2.349367088607595e-06, "loss": 0.0004, "step": 1849 }, { "epoch": 0.22, "learning_rate": 2.349005424954792e-06, "loss": 0.0012, "step": 1850 }, { "epoch": 0.22, "learning_rate": 2.3486437613019894e-06, "loss": 0.0004, "step": 1851 }, { "epoch": 0.22, "learning_rate": 2.3482820976491866e-06, "loss": 0.0006, "step": 1852 }, { "epoch": 0.22, "learning_rate": 2.3479204339963834e-06, "loss": 0.0006, "step": 1853 }, { "epoch": 0.22, "learning_rate": 2.3475587703435806e-06, "loss": 0.8808, "step": 1854 }, { "epoch": 0.22, "learning_rate": 2.3471971066907774e-06, "loss": 0.49, "step": 1855 }, { "epoch": 0.22, "learning_rate": 2.3468354430379747e-06, "loss": 0.0005, "step": 1856 }, { "epoch": 0.22, "learning_rate": 2.346473779385172e-06, "loss": 0.0005, "step": 1857 }, { "epoch": 0.22, "learning_rate": 2.3461121157323687e-06, "loss": 0.0006, "step": 1858 }, { "epoch": 0.22, "learning_rate": 2.345750452079566e-06, "loss": 0.0006, "step": 1859 }, { "epoch": 0.22, "learning_rate": 2.345388788426763e-06, "loss": 0.0006, "step": 1860 }, { "epoch": 0.22, "learning_rate": 2.3450271247739604e-06, "loss": 0.0004, "step": 1861 }, { "epoch": 0.22, "learning_rate": 2.344665461121157e-06, "loss": 0.0007, "step": 1862 }, { "epoch": 0.22, "learning_rate": 2.3443037974683544e-06, "loss": 0.0005, "step": 1863 }, { "epoch": 0.22, "learning_rate": 2.3439421338155517e-06, "loss": 0.0006, "step": 1864 }, { "epoch": 0.22, "learning_rate": 2.3435804701627485e-06, "loss": 0.0007, "step": 1865 }, { "epoch": 0.22, "learning_rate": 2.3432188065099457e-06, "loss": 0.0006, "step": 1866 }, { "epoch": 0.22, "learning_rate": 2.342857142857143e-06, "loss": 0.0004, "step": 1867 }, { "epoch": 0.22, "learning_rate": 2.34249547920434e-06, "loss": 0.0006, "step": 1868 }, { "epoch": 0.22, "learning_rate": 2.342133815551537e-06, "loss": 0.0004, "step": 1869 }, { "epoch": 0.22, "learning_rate": 2.341772151898734e-06, "loss": 0.0005, "step": 1870 }, { "epoch": 0.22, "learning_rate": 2.3414104882459314e-06, "loss": 1.3054, "step": 1871 }, { "epoch": 0.22, "learning_rate": 2.3410488245931287e-06, "loss": 0.9436, "step": 1872 }, { "epoch": 0.22, "learning_rate": 2.3406871609403255e-06, "loss": 0.0006, "step": 1873 }, { "epoch": 0.22, "learning_rate": 2.3403254972875227e-06, "loss": 0.47, "step": 1874 }, { "epoch": 0.22, "learning_rate": 2.33996383363472e-06, "loss": 0.0006, "step": 1875 }, { "epoch": 0.22, "learning_rate": 2.339602169981917e-06, "loss": 0.0015, "step": 1876 }, { "epoch": 0.22, "learning_rate": 2.339240506329114e-06, "loss": 0.0047, "step": 1877 }, { "epoch": 0.23, "learning_rate": 2.338878842676311e-06, "loss": 0.3001, "step": 1878 }, { "epoch": 0.23, "learning_rate": 2.338517179023508e-06, "loss": 0.0008, "step": 1879 }, { "epoch": 0.23, "learning_rate": 2.3381555153707052e-06, "loss": 0.001, "step": 1880 }, { "epoch": 0.23, "learning_rate": 2.3377938517179025e-06, "loss": 0.0008, "step": 1881 }, { "epoch": 0.23, "learning_rate": 2.3374321880650993e-06, "loss": 0.0007, "step": 1882 }, { "epoch": 0.23, "learning_rate": 2.3370705244122965e-06, "loss": 0.0008, "step": 1883 }, { "epoch": 0.23, "learning_rate": 2.3367088607594937e-06, "loss": 0.0005, "step": 1884 }, { "epoch": 0.23, "learning_rate": 2.3363471971066906e-06, "loss": 0.771, "step": 1885 }, { "epoch": 0.23, "learning_rate": 2.3359855334538878e-06, "loss": 0.0006, "step": 1886 }, { "epoch": 0.23, "learning_rate": 2.335623869801085e-06, "loss": 0.5016, "step": 1887 }, { "epoch": 0.23, "learning_rate": 2.3352622061482822e-06, "loss": 0.0005, "step": 1888 }, { "epoch": 0.23, "learning_rate": 2.334900542495479e-06, "loss": 0.0006, "step": 1889 }, { "epoch": 0.23, "learning_rate": 2.3345388788426763e-06, "loss": 0.0006, "step": 1890 }, { "epoch": 0.23, "learning_rate": 2.3341772151898735e-06, "loss": 0.0011, "step": 1891 }, { "epoch": 0.23, "learning_rate": 2.3338155515370707e-06, "loss": 0.0006, "step": 1892 }, { "epoch": 0.23, "learning_rate": 2.3334538878842676e-06, "loss": 0.0008, "step": 1893 }, { "epoch": 0.23, "learning_rate": 2.3330922242314648e-06, "loss": 0.717, "step": 1894 }, { "epoch": 0.23, "learning_rate": 2.332730560578662e-06, "loss": 0.031, "step": 1895 }, { "epoch": 0.23, "learning_rate": 2.3323688969258592e-06, "loss": 0.0004, "step": 1896 }, { "epoch": 0.23, "learning_rate": 2.332007233273056e-06, "loss": 0.0037, "step": 1897 }, { "epoch": 0.23, "learning_rate": 2.3316455696202533e-06, "loss": 0.0004, "step": 1898 }, { "epoch": 0.23, "learning_rate": 2.3312839059674505e-06, "loss": 0.0008, "step": 1899 }, { "epoch": 0.23, "learning_rate": 2.3309222423146477e-06, "loss": 0.0651, "step": 1900 }, { "epoch": 0.23, "learning_rate": 2.3305605786618446e-06, "loss": 0.0009, "step": 1901 }, { "epoch": 0.23, "learning_rate": 2.3301989150090418e-06, "loss": 0.0207, "step": 1902 }, { "epoch": 0.23, "learning_rate": 2.3298372513562386e-06, "loss": 0.0009, "step": 1903 }, { "epoch": 0.23, "learning_rate": 2.329475587703436e-06, "loss": 0.0009, "step": 1904 }, { "epoch": 0.23, "learning_rate": 2.3291139240506326e-06, "loss": 0.0005, "step": 1905 }, { "epoch": 0.23, "learning_rate": 2.32875226039783e-06, "loss": 0.8886, "step": 1906 }, { "epoch": 0.23, "learning_rate": 2.328390596745027e-06, "loss": 0.0238, "step": 1907 }, { "epoch": 0.23, "learning_rate": 2.3280289330922243e-06, "loss": 0.0106, "step": 1908 }, { "epoch": 0.23, "learning_rate": 2.327667269439421e-06, "loss": 0.0013, "step": 1909 }, { "epoch": 0.23, "learning_rate": 2.3273056057866184e-06, "loss": 0.0005, "step": 1910 }, { "epoch": 0.23, "learning_rate": 2.3269439421338156e-06, "loss": 0.0008, "step": 1911 }, { "epoch": 0.23, "learning_rate": 2.326582278481013e-06, "loss": 0.0005, "step": 1912 }, { "epoch": 0.23, "learning_rate": 2.3262206148282096e-06, "loss": 0.0005, "step": 1913 }, { "epoch": 0.23, "learning_rate": 2.325858951175407e-06, "loss": 0.5053, "step": 1914 }, { "epoch": 0.23, "learning_rate": 2.325497287522604e-06, "loss": 0.0005, "step": 1915 }, { "epoch": 0.23, "learning_rate": 2.3251356238698013e-06, "loss": 0.0008, "step": 1916 }, { "epoch": 0.23, "learning_rate": 2.324773960216998e-06, "loss": 0.8556, "step": 1917 }, { "epoch": 0.23, "learning_rate": 2.3244122965641954e-06, "loss": 0.0008, "step": 1918 }, { "epoch": 0.23, "learning_rate": 2.3240506329113926e-06, "loss": 0.0008, "step": 1919 }, { "epoch": 0.23, "learning_rate": 2.32368896925859e-06, "loss": 0.0005, "step": 1920 }, { "epoch": 0.23, "learning_rate": 2.3233273056057866e-06, "loss": 0.0009, "step": 1921 }, { "epoch": 0.23, "learning_rate": 2.322965641952984e-06, "loss": 0.0197, "step": 1922 }, { "epoch": 0.23, "learning_rate": 2.322603978300181e-06, "loss": 0.277, "step": 1923 }, { "epoch": 0.23, "learning_rate": 2.3222423146473783e-06, "loss": 0.0004, "step": 1924 }, { "epoch": 0.23, "learning_rate": 2.321880650994575e-06, "loss": 0.001, "step": 1925 }, { "epoch": 0.23, "learning_rate": 2.3215189873417724e-06, "loss": 0.0006, "step": 1926 }, { "epoch": 0.23, "learning_rate": 2.321157323688969e-06, "loss": 0.0006, "step": 1927 }, { "epoch": 0.23, "learning_rate": 2.3207956600361664e-06, "loss": 0.0006, "step": 1928 }, { "epoch": 0.23, "learning_rate": 2.3204339963833632e-06, "loss": 0.0013, "step": 1929 }, { "epoch": 0.23, "learning_rate": 2.3200723327305604e-06, "loss": 0.0009, "step": 1930 }, { "epoch": 0.23, "learning_rate": 2.3197106690777577e-06, "loss": 0.0006, "step": 1931 }, { "epoch": 0.23, "learning_rate": 2.319349005424955e-06, "loss": 0.0014, "step": 1932 }, { "epoch": 0.23, "learning_rate": 2.3189873417721517e-06, "loss": 0.0004, "step": 1933 }, { "epoch": 0.23, "learning_rate": 2.318625678119349e-06, "loss": 0.0005, "step": 1934 }, { "epoch": 0.23, "learning_rate": 2.318264014466546e-06, "loss": 0.0072, "step": 1935 }, { "epoch": 0.23, "learning_rate": 2.3179023508137434e-06, "loss": 0.0008, "step": 1936 }, { "epoch": 0.23, "learning_rate": 2.3175406871609402e-06, "loss": 0.0006, "step": 1937 }, { "epoch": 0.23, "learning_rate": 2.3171790235081375e-06, "loss": 0.0006, "step": 1938 }, { "epoch": 0.23, "learning_rate": 2.3168173598553347e-06, "loss": 0.0016, "step": 1939 }, { "epoch": 0.23, "learning_rate": 2.316455696202532e-06, "loss": 0.0007, "step": 1940 }, { "epoch": 0.23, "learning_rate": 2.3160940325497287e-06, "loss": 0.5611, "step": 1941 }, { "epoch": 0.23, "learning_rate": 2.315732368896926e-06, "loss": 0.0012, "step": 1942 }, { "epoch": 0.23, "learning_rate": 2.315370705244123e-06, "loss": 0.0004, "step": 1943 }, { "epoch": 0.23, "learning_rate": 2.3150090415913204e-06, "loss": 0.0009, "step": 1944 }, { "epoch": 0.23, "learning_rate": 2.3146473779385172e-06, "loss": 0.0003, "step": 1945 }, { "epoch": 0.23, "learning_rate": 2.3142857142857145e-06, "loss": 0.2246, "step": 1946 }, { "epoch": 0.23, "learning_rate": 2.3139240506329117e-06, "loss": 0.0005, "step": 1947 }, { "epoch": 0.23, "learning_rate": 2.3135623869801085e-06, "loss": 0.0005, "step": 1948 }, { "epoch": 0.23, "learning_rate": 2.3132007233273057e-06, "loss": 0.0017, "step": 1949 }, { "epoch": 0.23, "learning_rate": 2.312839059674503e-06, "loss": 0.0007, "step": 1950 }, { "epoch": 0.23, "learning_rate": 2.3124773960216998e-06, "loss": 0.0033, "step": 1951 }, { "epoch": 0.23, "learning_rate": 2.312115732368897e-06, "loss": 0.002, "step": 1952 }, { "epoch": 0.23, "learning_rate": 2.311754068716094e-06, "loss": 0.0004, "step": 1953 }, { "epoch": 0.23, "learning_rate": 2.311392405063291e-06, "loss": 0.8801, "step": 1954 }, { "epoch": 0.23, "learning_rate": 2.3110307414104883e-06, "loss": 0.0005, "step": 1955 }, { "epoch": 0.23, "learning_rate": 2.3106690777576855e-06, "loss": 0.0004, "step": 1956 }, { "epoch": 0.23, "learning_rate": 2.3103074141048823e-06, "loss": 0.0005, "step": 1957 }, { "epoch": 0.23, "learning_rate": 2.3099457504520795e-06, "loss": 0.0009, "step": 1958 }, { "epoch": 0.23, "learning_rate": 2.3095840867992768e-06, "loss": 0.0003, "step": 1959 }, { "epoch": 0.23, "learning_rate": 2.309222423146474e-06, "loss": 0.0009, "step": 1960 }, { "epoch": 0.23, "learning_rate": 2.308860759493671e-06, "loss": 0.0005, "step": 1961 }, { "epoch": 0.24, "learning_rate": 2.308499095840868e-06, "loss": 0.0009, "step": 1962 }, { "epoch": 0.24, "learning_rate": 2.3081374321880653e-06, "loss": 0.0005, "step": 1963 }, { "epoch": 0.24, "learning_rate": 2.3077757685352625e-06, "loss": 0.0004, "step": 1964 }, { "epoch": 0.24, "learning_rate": 2.3074141048824593e-06, "loss": 0.0007, "step": 1965 }, { "epoch": 0.24, "learning_rate": 2.3070524412296565e-06, "loss": 0.0005, "step": 1966 }, { "epoch": 0.24, "learning_rate": 2.3066907775768538e-06, "loss": 0.0008, "step": 1967 }, { "epoch": 0.24, "learning_rate": 2.3063291139240506e-06, "loss": 0.1082, "step": 1968 }, { "epoch": 0.24, "learning_rate": 2.305967450271248e-06, "loss": 0.0009, "step": 1969 }, { "epoch": 0.24, "learning_rate": 2.305605786618445e-06, "loss": 0.8238, "step": 1970 }, { "epoch": 0.24, "learning_rate": 2.3052441229656423e-06, "loss": 0.0005, "step": 1971 }, { "epoch": 0.24, "learning_rate": 2.304882459312839e-06, "loss": 0.0005, "step": 1972 }, { "epoch": 0.24, "learning_rate": 2.3045207956600363e-06, "loss": 0.0004, "step": 1973 }, { "epoch": 0.24, "learning_rate": 2.3041591320072335e-06, "loss": 0.0006, "step": 1974 }, { "epoch": 0.24, "learning_rate": 2.3037974683544303e-06, "loss": 0.0008, "step": 1975 }, { "epoch": 0.24, "learning_rate": 2.3034358047016276e-06, "loss": 0.0041, "step": 1976 }, { "epoch": 0.24, "learning_rate": 2.3030741410488244e-06, "loss": 0.006, "step": 1977 }, { "epoch": 0.24, "learning_rate": 2.3027124773960216e-06, "loss": 0.0006, "step": 1978 }, { "epoch": 0.24, "learning_rate": 2.302350813743219e-06, "loss": 0.0006, "step": 1979 }, { "epoch": 0.24, "learning_rate": 2.301989150090416e-06, "loss": 0.0125, "step": 1980 }, { "epoch": 0.24, "learning_rate": 2.301627486437613e-06, "loss": 0.0005, "step": 1981 }, { "epoch": 0.24, "learning_rate": 2.30126582278481e-06, "loss": 0.0007, "step": 1982 }, { "epoch": 0.24, "learning_rate": 2.3009041591320073e-06, "loss": 0.8284, "step": 1983 }, { "epoch": 0.24, "learning_rate": 2.300542495479204e-06, "loss": 0.0013, "step": 1984 }, { "epoch": 0.24, "learning_rate": 2.3001808318264014e-06, "loss": 0.0004, "step": 1985 }, { "epoch": 0.24, "learning_rate": 2.2998191681735986e-06, "loss": 0.0006, "step": 1986 }, { "epoch": 0.24, "learning_rate": 2.299457504520796e-06, "loss": 0.0006, "step": 1987 }, { "epoch": 0.24, "learning_rate": 2.2990958408679927e-06, "loss": 0.0008, "step": 1988 }, { "epoch": 0.24, "learning_rate": 2.29873417721519e-06, "loss": 0.963, "step": 1989 }, { "epoch": 0.24, "learning_rate": 2.298372513562387e-06, "loss": 0.0016, "step": 1990 }, { "epoch": 0.24, "learning_rate": 2.2980108499095843e-06, "loss": 0.0004, "step": 1991 }, { "epoch": 0.24, "learning_rate": 2.297649186256781e-06, "loss": 0.0006, "step": 1992 }, { "epoch": 0.24, "learning_rate": 2.2972875226039784e-06, "loss": 0.0003, "step": 1993 }, { "epoch": 0.24, "learning_rate": 2.2969258589511756e-06, "loss": 0.8549, "step": 1994 }, { "epoch": 0.24, "learning_rate": 2.296564195298373e-06, "loss": 0.6268, "step": 1995 }, { "epoch": 0.24, "learning_rate": 2.2962025316455697e-06, "loss": 0.0004, "step": 1996 }, { "epoch": 0.24, "learning_rate": 2.295840867992767e-06, "loss": 0.0004, "step": 1997 }, { "epoch": 0.24, "learning_rate": 2.295479204339964e-06, "loss": 0.0006, "step": 1998 }, { "epoch": 0.24, "learning_rate": 2.295117540687161e-06, "loss": 0.0008, "step": 1999 }, { "epoch": 0.24, "learning_rate": 2.294755877034358e-06, "loss": 0.0005, "step": 2000 }, { "epoch": 0.24, "learning_rate": 2.294394213381555e-06, "loss": 0.0011, "step": 2001 }, { "epoch": 0.24, "learning_rate": 2.294032549728752e-06, "loss": 0.0005, "step": 2002 }, { "epoch": 0.24, "learning_rate": 2.2936708860759494e-06, "loss": 0.0014, "step": 2003 }, { "epoch": 0.24, "learning_rate": 2.2933092224231462e-06, "loss": 0.0004, "step": 2004 }, { "epoch": 0.24, "learning_rate": 2.2929475587703435e-06, "loss": 0.0005, "step": 2005 }, { "epoch": 0.24, "learning_rate": 2.2925858951175407e-06, "loss": 0.0007, "step": 2006 }, { "epoch": 0.24, "learning_rate": 2.292224231464738e-06, "loss": 0.0011, "step": 2007 }, { "epoch": 0.24, "learning_rate": 2.2918625678119347e-06, "loss": 0.0006, "step": 2008 }, { "epoch": 0.24, "learning_rate": 2.291500904159132e-06, "loss": 0.0006, "step": 2009 }, { "epoch": 0.24, "learning_rate": 2.291139240506329e-06, "loss": 0.0007, "step": 2010 }, { "epoch": 0.24, "learning_rate": 2.2907775768535264e-06, "loss": 0.8388, "step": 2011 }, { "epoch": 0.24, "learning_rate": 2.2904159132007232e-06, "loss": 0.3367, "step": 2012 }, { "epoch": 0.24, "learning_rate": 2.2900542495479205e-06, "loss": 0.0006, "step": 2013 }, { "epoch": 0.24, "learning_rate": 2.2896925858951177e-06, "loss": 0.1191, "step": 2014 }, { "epoch": 0.24, "learning_rate": 2.289330922242315e-06, "loss": 0.0007, "step": 2015 }, { "epoch": 0.24, "learning_rate": 2.2889692585895117e-06, "loss": 0.0006, "step": 2016 }, { "epoch": 0.24, "learning_rate": 2.288607594936709e-06, "loss": 0.001, "step": 2017 }, { "epoch": 0.24, "learning_rate": 2.288245931283906e-06, "loss": 0.0005, "step": 2018 }, { "epoch": 0.24, "learning_rate": 2.2878842676311034e-06, "loss": 0.0006, "step": 2019 }, { "epoch": 0.24, "learning_rate": 2.2875226039783002e-06, "loss": 0.7186, "step": 2020 }, { "epoch": 0.24, "learning_rate": 2.2871609403254975e-06, "loss": 0.2696, "step": 2021 }, { "epoch": 0.24, "learning_rate": 2.2867992766726947e-06, "loss": 0.0006, "step": 2022 }, { "epoch": 0.24, "learning_rate": 2.2864376130198915e-06, "loss": 0.0007, "step": 2023 }, { "epoch": 0.24, "learning_rate": 2.2860759493670883e-06, "loss": 0.0006, "step": 2024 }, { "epoch": 0.24, "learning_rate": 2.2857142857142856e-06, "loss": 0.0004, "step": 2025 }, { "epoch": 0.24, "learning_rate": 2.2853526220614828e-06, "loss": 0.5443, "step": 2026 }, { "epoch": 0.24, "learning_rate": 2.28499095840868e-06, "loss": 0.0006, "step": 2027 }, { "epoch": 0.24, "learning_rate": 2.284629294755877e-06, "loss": 0.0003, "step": 2028 }, { "epoch": 0.24, "learning_rate": 2.284267631103074e-06, "loss": 0.0011, "step": 2029 }, { "epoch": 0.24, "learning_rate": 2.2839059674502713e-06, "loss": 0.0015, "step": 2030 }, { "epoch": 0.24, "learning_rate": 2.2835443037974685e-06, "loss": 0.0253, "step": 2031 }, { "epoch": 0.24, "learning_rate": 2.2831826401446653e-06, "loss": 0.0008, "step": 2032 }, { "epoch": 0.24, "learning_rate": 2.2828209764918626e-06, "loss": 0.0007, "step": 2033 }, { "epoch": 0.24, "learning_rate": 2.2824593128390598e-06, "loss": 0.0006, "step": 2034 }, { "epoch": 0.24, "learning_rate": 2.282097649186257e-06, "loss": 0.0009, "step": 2035 }, { "epoch": 0.24, "learning_rate": 2.281735985533454e-06, "loss": 0.0007, "step": 2036 }, { "epoch": 0.24, "learning_rate": 2.281374321880651e-06, "loss": 0.0006, "step": 2037 }, { "epoch": 0.24, "learning_rate": 2.2810126582278483e-06, "loss": 0.0008, "step": 2038 }, { "epoch": 0.24, "learning_rate": 2.2806509945750455e-06, "loss": 0.0011, "step": 2039 }, { "epoch": 0.24, "learning_rate": 2.2802893309222423e-06, "loss": 0.001, "step": 2040 }, { "epoch": 0.24, "learning_rate": 2.2799276672694396e-06, "loss": 0.0004, "step": 2041 }, { "epoch": 0.24, "learning_rate": 2.2795660036166368e-06, "loss": 0.0696, "step": 2042 }, { "epoch": 0.24, "learning_rate": 2.279204339963834e-06, "loss": 0.0005, "step": 2043 }, { "epoch": 0.24, "learning_rate": 2.278842676311031e-06, "loss": 0.0009, "step": 2044 }, { "epoch": 0.25, "learning_rate": 2.278481012658228e-06, "loss": 0.0056, "step": 2045 }, { "epoch": 0.25, "learning_rate": 2.2781193490054253e-06, "loss": 0.0007, "step": 2046 }, { "epoch": 0.25, "learning_rate": 2.277757685352622e-06, "loss": 0.0007, "step": 2047 }, { "epoch": 0.25, "learning_rate": 2.277396021699819e-06, "loss": 0.0006, "step": 2048 }, { "epoch": 0.25, "learning_rate": 2.277034358047016e-06, "loss": 0.0381, "step": 2049 }, { "epoch": 0.25, "learning_rate": 2.2766726943942134e-06, "loss": 0.0004, "step": 2050 }, { "epoch": 0.25, "learning_rate": 2.2763110307414106e-06, "loss": 0.0004, "step": 2051 }, { "epoch": 0.25, "learning_rate": 2.2759493670886074e-06, "loss": 0.0005, "step": 2052 }, { "epoch": 0.25, "learning_rate": 2.2755877034358046e-06, "loss": 0.0013, "step": 2053 }, { "epoch": 0.25, "learning_rate": 2.275226039783002e-06, "loss": 0.0006, "step": 2054 }, { "epoch": 0.25, "learning_rate": 2.274864376130199e-06, "loss": 0.0005, "step": 2055 }, { "epoch": 0.25, "learning_rate": 2.274502712477396e-06, "loss": 0.0009, "step": 2056 }, { "epoch": 0.25, "learning_rate": 2.274141048824593e-06, "loss": 0.0035, "step": 2057 }, { "epoch": 0.25, "learning_rate": 2.2737793851717904e-06, "loss": 0.0004, "step": 2058 }, { "epoch": 0.25, "learning_rate": 2.2734177215189876e-06, "loss": 0.376, "step": 2059 }, { "epoch": 0.25, "learning_rate": 2.2730560578661844e-06, "loss": 0.0006, "step": 2060 }, { "epoch": 0.25, "learning_rate": 2.2726943942133816e-06, "loss": 0.0004, "step": 2061 }, { "epoch": 0.25, "learning_rate": 2.272332730560579e-06, "loss": 0.0038, "step": 2062 }, { "epoch": 0.25, "learning_rate": 2.271971066907776e-06, "loss": 0.0006, "step": 2063 }, { "epoch": 0.25, "learning_rate": 2.271609403254973e-06, "loss": 0.0005, "step": 2064 }, { "epoch": 0.25, "learning_rate": 2.27124773960217e-06, "loss": 0.0004, "step": 2065 }, { "epoch": 0.25, "learning_rate": 2.2708860759493674e-06, "loss": 0.0003, "step": 2066 }, { "epoch": 0.25, "learning_rate": 2.270524412296564e-06, "loss": 0.0004, "step": 2067 }, { "epoch": 0.25, "learning_rate": 2.2701627486437614e-06, "loss": 0.0007, "step": 2068 }, { "epoch": 0.25, "learning_rate": 2.2698010849909586e-06, "loss": 0.0006, "step": 2069 }, { "epoch": 0.25, "learning_rate": 2.2694394213381554e-06, "loss": 0.0011, "step": 2070 }, { "epoch": 0.25, "learning_rate": 2.2690777576853527e-06, "loss": 0.0008, "step": 2071 }, { "epoch": 0.25, "learning_rate": 2.2687160940325495e-06, "loss": 0.0005, "step": 2072 }, { "epoch": 0.25, "learning_rate": 2.2683544303797467e-06, "loss": 1.8116, "step": 2073 }, { "epoch": 0.25, "learning_rate": 2.267992766726944e-06, "loss": 0.0004, "step": 2074 }, { "epoch": 0.25, "learning_rate": 2.267631103074141e-06, "loss": 0.0004, "step": 2075 }, { "epoch": 0.25, "learning_rate": 2.267269439421338e-06, "loss": 0.0004, "step": 2076 }, { "epoch": 0.25, "learning_rate": 2.2669077757685352e-06, "loss": 0.0004, "step": 2077 }, { "epoch": 0.25, "learning_rate": 2.2665461121157324e-06, "loss": 0.0006, "step": 2078 }, { "epoch": 0.25, "learning_rate": 2.2661844484629297e-06, "loss": 0.0003, "step": 2079 }, { "epoch": 0.25, "learning_rate": 2.2658227848101265e-06, "loss": 0.0004, "step": 2080 }, { "epoch": 0.25, "learning_rate": 2.2654611211573237e-06, "loss": 0.5398, "step": 2081 }, { "epoch": 0.25, "learning_rate": 2.265099457504521e-06, "loss": 0.0012, "step": 2082 }, { "epoch": 0.25, "learning_rate": 2.264737793851718e-06, "loss": 0.0011, "step": 2083 }, { "epoch": 0.25, "learning_rate": 2.264376130198915e-06, "loss": 0.0003, "step": 2084 }, { "epoch": 0.25, "learning_rate": 2.2640144665461122e-06, "loss": 0.0003, "step": 2085 }, { "epoch": 0.25, "learning_rate": 2.2636528028933095e-06, "loss": 0.0004, "step": 2086 }, { "epoch": 0.25, "learning_rate": 2.2632911392405063e-06, "loss": 0.0004, "step": 2087 }, { "epoch": 0.25, "learning_rate": 2.2629294755877035e-06, "loss": 0.0005, "step": 2088 }, { "epoch": 0.25, "learning_rate": 2.2625678119349007e-06, "loss": 0.0009, "step": 2089 }, { "epoch": 0.25, "learning_rate": 2.262206148282098e-06, "loss": 0.0004, "step": 2090 }, { "epoch": 0.25, "learning_rate": 2.2618444846292948e-06, "loss": 0.0005, "step": 2091 }, { "epoch": 0.25, "learning_rate": 2.261482820976492e-06, "loss": 0.0004, "step": 2092 }, { "epoch": 0.25, "learning_rate": 2.2611211573236892e-06, "loss": 0.0004, "step": 2093 }, { "epoch": 0.25, "learning_rate": 2.260759493670886e-06, "loss": 0.0062, "step": 2094 }, { "epoch": 0.25, "learning_rate": 2.2603978300180833e-06, "loss": 0.0115, "step": 2095 }, { "epoch": 0.25, "learning_rate": 2.26003616636528e-06, "loss": 0.0004, "step": 2096 }, { "epoch": 0.25, "learning_rate": 2.2596745027124773e-06, "loss": 0.0003, "step": 2097 }, { "epoch": 0.25, "learning_rate": 2.2593128390596745e-06, "loss": 0.0003, "step": 2098 }, { "epoch": 0.25, "learning_rate": 2.2589511754068718e-06, "loss": 0.0004, "step": 2099 }, { "epoch": 0.25, "learning_rate": 2.2585895117540686e-06, "loss": 0.4176, "step": 2100 }, { "epoch": 0.25, "learning_rate": 2.258227848101266e-06, "loss": 0.0004, "step": 2101 }, { "epoch": 0.25, "learning_rate": 2.257866184448463e-06, "loss": 0.14, "step": 2102 }, { "epoch": 0.25, "learning_rate": 2.2575045207956603e-06, "loss": 0.0015, "step": 2103 }, { "epoch": 0.25, "learning_rate": 2.257142857142857e-06, "loss": 0.0004, "step": 2104 }, { "epoch": 0.25, "learning_rate": 2.2567811934900543e-06, "loss": 0.0011, "step": 2105 }, { "epoch": 0.25, "learning_rate": 2.2564195298372515e-06, "loss": 0.0003, "step": 2106 }, { "epoch": 0.25, "learning_rate": 2.2560578661844483e-06, "loss": 0.0005, "step": 2107 }, { "epoch": 0.25, "learning_rate": 2.2556962025316456e-06, "loss": 0.8456, "step": 2108 }, { "epoch": 0.25, "learning_rate": 2.255334538878843e-06, "loss": 0.0003, "step": 2109 }, { "epoch": 0.25, "learning_rate": 2.25497287522604e-06, "loss": 0.0003, "step": 2110 }, { "epoch": 0.25, "learning_rate": 2.254611211573237e-06, "loss": 0.0011, "step": 2111 }, { "epoch": 0.25, "learning_rate": 2.254249547920434e-06, "loss": 0.0008, "step": 2112 }, { "epoch": 0.25, "learning_rate": 2.2538878842676313e-06, "loss": 0.0006, "step": 2113 }, { "epoch": 0.25, "learning_rate": 2.2535262206148285e-06, "loss": 0.0004, "step": 2114 }, { "epoch": 0.25, "learning_rate": 2.2531645569620253e-06, "loss": 0.0029, "step": 2115 }, { "epoch": 0.25, "learning_rate": 2.2528028933092226e-06, "loss": 0.0006, "step": 2116 }, { "epoch": 0.25, "learning_rate": 2.25244122965642e-06, "loss": 0.0003, "step": 2117 }, { "epoch": 0.25, "learning_rate": 2.2520795660036166e-06, "loss": 0.0003, "step": 2118 }, { "epoch": 0.25, "learning_rate": 2.251717902350814e-06, "loss": 1.012, "step": 2119 }, { "epoch": 0.25, "learning_rate": 2.2513562386980107e-06, "loss": 0.0003, "step": 2120 }, { "epoch": 0.25, "learning_rate": 2.250994575045208e-06, "loss": 0.0062, "step": 2121 }, { "epoch": 0.25, "learning_rate": 2.250632911392405e-06, "loss": 0.8546, "step": 2122 }, { "epoch": 0.25, "learning_rate": 2.250271247739602e-06, "loss": 0.8422, "step": 2123 }, { "epoch": 0.25, "learning_rate": 2.249909584086799e-06, "loss": 0.0004, "step": 2124 }, { "epoch": 0.25, "learning_rate": 2.2495479204339964e-06, "loss": 0.0004, "step": 2125 }, { "epoch": 0.25, "learning_rate": 2.2491862567811936e-06, "loss": 0.0004, "step": 2126 }, { "epoch": 0.25, "learning_rate": 2.2488245931283904e-06, "loss": 0.0219, "step": 2127 }, { "epoch": 0.26, "learning_rate": 2.2484629294755877e-06, "loss": 0.2049, "step": 2128 }, { "epoch": 0.26, "learning_rate": 2.248101265822785e-06, "loss": 0.0005, "step": 2129 }, { "epoch": 0.26, "learning_rate": 2.247739602169982e-06, "loss": 0.0005, "step": 2130 }, { "epoch": 0.26, "learning_rate": 2.247377938517179e-06, "loss": 0.0004, "step": 2131 }, { "epoch": 0.26, "learning_rate": 2.247016274864376e-06, "loss": 0.0006, "step": 2132 }, { "epoch": 0.26, "learning_rate": 2.2466546112115734e-06, "loss": 0.0004, "step": 2133 }, { "epoch": 0.26, "learning_rate": 2.2462929475587706e-06, "loss": 0.0005, "step": 2134 }, { "epoch": 0.26, "learning_rate": 2.2459312839059674e-06, "loss": 0.0004, "step": 2135 }, { "epoch": 0.26, "learning_rate": 2.2455696202531647e-06, "loss": 0.7772, "step": 2136 }, { "epoch": 0.26, "learning_rate": 2.245207956600362e-06, "loss": 0.0006, "step": 2137 }, { "epoch": 0.26, "learning_rate": 2.244846292947559e-06, "loss": 0.0005, "step": 2138 }, { "epoch": 0.26, "learning_rate": 2.244484629294756e-06, "loss": 0.0004, "step": 2139 }, { "epoch": 0.26, "learning_rate": 2.244122965641953e-06, "loss": 0.3752, "step": 2140 }, { "epoch": 0.26, "learning_rate": 2.2437613019891504e-06, "loss": 0.0012, "step": 2141 }, { "epoch": 0.26, "learning_rate": 2.243399638336347e-06, "loss": 0.0004, "step": 2142 }, { "epoch": 0.26, "learning_rate": 2.243037974683544e-06, "loss": 0.0121, "step": 2143 }, { "epoch": 0.26, "learning_rate": 2.2426763110307412e-06, "loss": 0.6834, "step": 2144 }, { "epoch": 0.26, "learning_rate": 2.2423146473779385e-06, "loss": 0.7304, "step": 2145 }, { "epoch": 0.26, "learning_rate": 2.2419529837251357e-06, "loss": 0.8672, "step": 2146 }, { "epoch": 0.26, "learning_rate": 2.2415913200723325e-06, "loss": 0.0006, "step": 2147 }, { "epoch": 0.26, "learning_rate": 2.2412296564195297e-06, "loss": 0.0005, "step": 2148 }, { "epoch": 0.26, "learning_rate": 2.240867992766727e-06, "loss": 0.0005, "step": 2149 }, { "epoch": 0.26, "learning_rate": 2.240506329113924e-06, "loss": 0.0005, "step": 2150 }, { "epoch": 0.26, "learning_rate": 2.240144665461121e-06, "loss": 0.0006, "step": 2151 }, { "epoch": 0.26, "learning_rate": 2.2397830018083182e-06, "loss": 0.0005, "step": 2152 }, { "epoch": 0.26, "learning_rate": 2.2394213381555155e-06, "loss": 0.0018, "step": 2153 }, { "epoch": 0.26, "learning_rate": 2.2390596745027127e-06, "loss": 0.0004, "step": 2154 }, { "epoch": 0.26, "learning_rate": 2.2386980108499095e-06, "loss": 0.0005, "step": 2155 }, { "epoch": 0.26, "learning_rate": 2.2383363471971067e-06, "loss": 0.0005, "step": 2156 }, { "epoch": 0.26, "learning_rate": 2.237974683544304e-06, "loss": 0.0005, "step": 2157 }, { "epoch": 0.26, "learning_rate": 2.237613019891501e-06, "loss": 0.531, "step": 2158 }, { "epoch": 0.26, "learning_rate": 2.237251356238698e-06, "loss": 0.0007, "step": 2159 }, { "epoch": 0.26, "learning_rate": 2.2368896925858952e-06, "loss": 0.0004, "step": 2160 }, { "epoch": 0.26, "learning_rate": 2.2365280289330925e-06, "loss": 0.0009, "step": 2161 }, { "epoch": 0.26, "learning_rate": 2.2361663652802897e-06, "loss": 0.9445, "step": 2162 }, { "epoch": 0.26, "learning_rate": 2.2358047016274865e-06, "loss": 0.0006, "step": 2163 }, { "epoch": 0.26, "learning_rate": 2.2354430379746837e-06, "loss": 0.0005, "step": 2164 }, { "epoch": 0.26, "learning_rate": 2.235081374321881e-06, "loss": 0.0005, "step": 2165 }, { "epoch": 0.26, "learning_rate": 2.2347197106690778e-06, "loss": 0.5568, "step": 2166 }, { "epoch": 0.26, "learning_rate": 2.2343580470162746e-06, "loss": 0.0005, "step": 2167 }, { "epoch": 0.26, "learning_rate": 2.233996383363472e-06, "loss": 0.0007, "step": 2168 }, { "epoch": 0.26, "learning_rate": 2.233634719710669e-06, "loss": 0.002, "step": 2169 }, { "epoch": 0.26, "learning_rate": 2.2332730560578663e-06, "loss": 0.8863, "step": 2170 }, { "epoch": 0.26, "learning_rate": 2.232911392405063e-06, "loss": 0.0004, "step": 2171 }, { "epoch": 0.26, "learning_rate": 2.2325497287522603e-06, "loss": 0.0005, "step": 2172 }, { "epoch": 0.26, "learning_rate": 2.2321880650994576e-06, "loss": 0.0092, "step": 2173 }, { "epoch": 0.26, "learning_rate": 2.2318264014466548e-06, "loss": 0.001, "step": 2174 }, { "epoch": 0.26, "learning_rate": 2.2314647377938516e-06, "loss": 0.001, "step": 2175 }, { "epoch": 0.26, "learning_rate": 2.231103074141049e-06, "loss": 0.0184, "step": 2176 }, { "epoch": 0.26, "learning_rate": 2.230741410488246e-06, "loss": 0.0012, "step": 2177 }, { "epoch": 0.26, "learning_rate": 2.2303797468354433e-06, "loss": 0.0015, "step": 2178 }, { "epoch": 0.26, "learning_rate": 2.23001808318264e-06, "loss": 0.7497, "step": 2179 }, { "epoch": 0.26, "learning_rate": 2.2296564195298373e-06, "loss": 0.0007, "step": 2180 }, { "epoch": 0.26, "learning_rate": 2.2292947558770346e-06, "loss": 0.0005, "step": 2181 }, { "epoch": 0.26, "learning_rate": 2.2289330922242318e-06, "loss": 0.1341, "step": 2182 }, { "epoch": 0.26, "learning_rate": 2.2285714285714286e-06, "loss": 0.0006, "step": 2183 }, { "epoch": 0.26, "learning_rate": 2.228209764918626e-06, "loss": 0.001, "step": 2184 }, { "epoch": 0.26, "learning_rate": 2.227848101265823e-06, "loss": 0.7435, "step": 2185 }, { "epoch": 0.26, "learning_rate": 2.22748643761302e-06, "loss": 0.0008, "step": 2186 }, { "epoch": 0.26, "learning_rate": 2.227124773960217e-06, "loss": 0.0015, "step": 2187 }, { "epoch": 0.26, "learning_rate": 2.2267631103074143e-06, "loss": 0.0008, "step": 2188 }, { "epoch": 0.26, "learning_rate": 2.2264014466546116e-06, "loss": 0.0007, "step": 2189 }, { "epoch": 0.26, "learning_rate": 2.2260397830018084e-06, "loss": 0.5226, "step": 2190 }, { "epoch": 0.26, "learning_rate": 2.225678119349005e-06, "loss": 0.0008, "step": 2191 }, { "epoch": 0.26, "learning_rate": 2.2253164556962024e-06, "loss": 0.0025, "step": 2192 }, { "epoch": 0.26, "learning_rate": 2.2249547920433996e-06, "loss": 0.0005, "step": 2193 }, { "epoch": 0.26, "learning_rate": 2.224593128390597e-06, "loss": 0.0006, "step": 2194 }, { "epoch": 0.26, "learning_rate": 2.2242314647377937e-06, "loss": 0.0009, "step": 2195 }, { "epoch": 0.26, "learning_rate": 2.223869801084991e-06, "loss": 0.0008, "step": 2196 }, { "epoch": 0.26, "learning_rate": 2.223508137432188e-06, "loss": 0.008, "step": 2197 }, { "epoch": 0.26, "learning_rate": 2.2231464737793854e-06, "loss": 0.0005, "step": 2198 }, { "epoch": 0.26, "learning_rate": 2.222784810126582e-06, "loss": 0.0004, "step": 2199 }, { "epoch": 0.26, "learning_rate": 2.2224231464737794e-06, "loss": 0.7628, "step": 2200 }, { "epoch": 0.26, "learning_rate": 2.2220614828209766e-06, "loss": 0.4538, "step": 2201 }, { "epoch": 0.26, "learning_rate": 2.221699819168174e-06, "loss": 0.002, "step": 2202 }, { "epoch": 0.26, "learning_rate": 2.2213381555153707e-06, "loss": 0.0007, "step": 2203 }, { "epoch": 0.26, "learning_rate": 2.220976491862568e-06, "loss": 1.233, "step": 2204 }, { "epoch": 0.26, "learning_rate": 2.220614828209765e-06, "loss": 0.0009, "step": 2205 }, { "epoch": 0.26, "learning_rate": 2.220253164556962e-06, "loss": 0.0006, "step": 2206 }, { "epoch": 0.26, "learning_rate": 2.219891500904159e-06, "loss": 0.0004, "step": 2207 }, { "epoch": 0.26, "learning_rate": 2.2195298372513564e-06, "loss": 0.0006, "step": 2208 }, { "epoch": 0.26, "learning_rate": 2.2191681735985536e-06, "loss": 0.0005, "step": 2209 }, { "epoch": 0.26, "learning_rate": 2.2188065099457504e-06, "loss": 0.6154, "step": 2210 }, { "epoch": 0.26, "learning_rate": 2.2184448462929477e-06, "loss": 0.0005, "step": 2211 }, { "epoch": 0.27, "learning_rate": 2.218083182640145e-06, "loss": 0.0005, "step": 2212 }, { "epoch": 0.27, "learning_rate": 2.217721518987342e-06, "loss": 0.0004, "step": 2213 }, { "epoch": 0.27, "learning_rate": 2.217359855334539e-06, "loss": 0.0005, "step": 2214 }, { "epoch": 0.27, "learning_rate": 2.2169981916817358e-06, "loss": 0.5639, "step": 2215 }, { "epoch": 0.27, "learning_rate": 2.216636528028933e-06, "loss": 0.0004, "step": 2216 }, { "epoch": 0.27, "learning_rate": 2.2162748643761302e-06, "loss": 0.0004, "step": 2217 }, { "epoch": 0.27, "learning_rate": 2.2159132007233274e-06, "loss": 0.0006, "step": 2218 }, { "epoch": 0.27, "learning_rate": 2.2155515370705243e-06, "loss": 0.0332, "step": 2219 }, { "epoch": 0.27, "learning_rate": 2.2151898734177215e-06, "loss": 0.0006, "step": 2220 }, { "epoch": 0.27, "learning_rate": 2.2148282097649187e-06, "loss": 0.0005, "step": 2221 }, { "epoch": 0.27, "learning_rate": 2.214466546112116e-06, "loss": 0.6834, "step": 2222 }, { "epoch": 0.27, "learning_rate": 2.2141048824593128e-06, "loss": 0.0004, "step": 2223 }, { "epoch": 0.27, "learning_rate": 2.21374321880651e-06, "loss": 0.0005, "step": 2224 }, { "epoch": 0.27, "learning_rate": 2.2133815551537072e-06, "loss": 0.7968, "step": 2225 }, { "epoch": 0.27, "learning_rate": 2.213019891500904e-06, "loss": 0.0004, "step": 2226 }, { "epoch": 0.27, "learning_rate": 2.2126582278481013e-06, "loss": 0.0007, "step": 2227 }, { "epoch": 0.27, "learning_rate": 2.2122965641952985e-06, "loss": 0.0005, "step": 2228 }, { "epoch": 0.27, "learning_rate": 2.2119349005424957e-06, "loss": 0.4568, "step": 2229 }, { "epoch": 0.27, "learning_rate": 2.2115732368896925e-06, "loss": 0.0006, "step": 2230 }, { "epoch": 0.27, "learning_rate": 2.2112115732368898e-06, "loss": 0.0005, "step": 2231 }, { "epoch": 0.27, "learning_rate": 2.210849909584087e-06, "loss": 0.0004, "step": 2232 }, { "epoch": 0.27, "learning_rate": 2.2104882459312842e-06, "loss": 0.001, "step": 2233 }, { "epoch": 0.27, "learning_rate": 2.210126582278481e-06, "loss": 0.0037, "step": 2234 }, { "epoch": 0.27, "learning_rate": 2.2097649186256783e-06, "loss": 1.6128, "step": 2235 }, { "epoch": 0.27, "learning_rate": 2.2094032549728755e-06, "loss": 0.001, "step": 2236 }, { "epoch": 0.27, "learning_rate": 2.2090415913200727e-06, "loss": 0.0005, "step": 2237 }, { "epoch": 0.27, "learning_rate": 2.2086799276672695e-06, "loss": 0.0007, "step": 2238 }, { "epoch": 0.27, "learning_rate": 2.2083182640144663e-06, "loss": 0.6958, "step": 2239 }, { "epoch": 0.27, "learning_rate": 2.2079566003616636e-06, "loss": 0.0311, "step": 2240 }, { "epoch": 0.27, "learning_rate": 2.207594936708861e-06, "loss": 0.0004, "step": 2241 }, { "epoch": 0.27, "learning_rate": 2.207233273056058e-06, "loss": 0.0004, "step": 2242 }, { "epoch": 0.27, "learning_rate": 2.206871609403255e-06, "loss": 0.0005, "step": 2243 }, { "epoch": 0.27, "learning_rate": 2.206509945750452e-06, "loss": 0.0004, "step": 2244 }, { "epoch": 0.27, "learning_rate": 2.2061482820976493e-06, "loss": 0.0006, "step": 2245 }, { "epoch": 0.27, "learning_rate": 2.205786618444846e-06, "loss": 0.4308, "step": 2246 }, { "epoch": 0.27, "learning_rate": 2.2054249547920433e-06, "loss": 0.0006, "step": 2247 }, { "epoch": 0.27, "learning_rate": 2.2050632911392406e-06, "loss": 0.0046, "step": 2248 }, { "epoch": 0.27, "learning_rate": 2.204701627486438e-06, "loss": 0.001, "step": 2249 }, { "epoch": 0.27, "learning_rate": 2.2043399638336346e-06, "loss": 0.0003, "step": 2250 }, { "epoch": 0.27, "learning_rate": 2.203978300180832e-06, "loss": 0.0007, "step": 2251 }, { "epoch": 0.27, "learning_rate": 2.203616636528029e-06, "loss": 0.0006, "step": 2252 }, { "epoch": 0.27, "learning_rate": 2.2032549728752263e-06, "loss": 0.0008, "step": 2253 }, { "epoch": 0.27, "learning_rate": 2.202893309222423e-06, "loss": 0.9014, "step": 2254 }, { "epoch": 0.27, "learning_rate": 2.2025316455696203e-06, "loss": 0.0008, "step": 2255 }, { "epoch": 0.27, "learning_rate": 2.2021699819168176e-06, "loss": 0.0004, "step": 2256 }, { "epoch": 0.27, "learning_rate": 2.201808318264015e-06, "loss": 0.9228, "step": 2257 }, { "epoch": 0.27, "learning_rate": 2.2014466546112116e-06, "loss": 0.0004, "step": 2258 }, { "epoch": 0.27, "learning_rate": 2.201084990958409e-06, "loss": 0.0005, "step": 2259 }, { "epoch": 0.27, "learning_rate": 2.200723327305606e-06, "loss": 0.0006, "step": 2260 }, { "epoch": 0.27, "learning_rate": 2.2003616636528033e-06, "loss": 0.0294, "step": 2261 }, { "epoch": 0.27, "learning_rate": 2.1999999999999997e-06, "loss": 0.0015, "step": 2262 }, { "epoch": 0.27, "learning_rate": 2.199638336347197e-06, "loss": 0.0004, "step": 2263 }, { "epoch": 0.27, "learning_rate": 2.199276672694394e-06, "loss": 0.0005, "step": 2264 }, { "epoch": 0.27, "learning_rate": 2.1989150090415914e-06, "loss": 0.0009, "step": 2265 }, { "epoch": 0.27, "learning_rate": 2.198553345388788e-06, "loss": 0.5859, "step": 2266 }, { "epoch": 0.27, "learning_rate": 2.1981916817359854e-06, "loss": 0.0005, "step": 2267 }, { "epoch": 0.27, "learning_rate": 2.1978300180831827e-06, "loss": 0.0004, "step": 2268 }, { "epoch": 0.27, "learning_rate": 2.19746835443038e-06, "loss": 0.0014, "step": 2269 }, { "epoch": 0.27, "learning_rate": 2.1971066907775767e-06, "loss": 0.8342, "step": 2270 }, { "epoch": 0.27, "learning_rate": 2.196745027124774e-06, "loss": 0.0008, "step": 2271 }, { "epoch": 0.27, "learning_rate": 2.196383363471971e-06, "loss": 0.0004, "step": 2272 }, { "epoch": 0.27, "learning_rate": 2.1960216998191684e-06, "loss": 0.0011, "step": 2273 }, { "epoch": 0.27, "learning_rate": 2.195660036166365e-06, "loss": 0.0023, "step": 2274 }, { "epoch": 0.27, "learning_rate": 2.1952983725135624e-06, "loss": 0.9025, "step": 2275 }, { "epoch": 0.27, "learning_rate": 2.1949367088607597e-06, "loss": 0.0004, "step": 2276 }, { "epoch": 0.27, "learning_rate": 2.194575045207957e-06, "loss": 0.0004, "step": 2277 }, { "epoch": 0.27, "learning_rate": 2.1942133815551537e-06, "loss": 0.0337, "step": 2278 }, { "epoch": 0.27, "learning_rate": 2.193851717902351e-06, "loss": 0.0004, "step": 2279 }, { "epoch": 0.27, "learning_rate": 2.193490054249548e-06, "loss": 0.0006, "step": 2280 }, { "epoch": 0.27, "learning_rate": 2.1931283905967454e-06, "loss": 0.8594, "step": 2281 }, { "epoch": 0.27, "learning_rate": 2.192766726943942e-06, "loss": 0.0005, "step": 2282 }, { "epoch": 0.27, "learning_rate": 2.1924050632911394e-06, "loss": 0.0005, "step": 2283 }, { "epoch": 0.27, "learning_rate": 2.1920433996383367e-06, "loss": 0.0017, "step": 2284 }, { "epoch": 0.27, "learning_rate": 2.1916817359855335e-06, "loss": 0.0012, "step": 2285 }, { "epoch": 0.27, "learning_rate": 2.1913200723327303e-06, "loss": 0.5834, "step": 2286 }, { "epoch": 0.27, "learning_rate": 2.1909584086799275e-06, "loss": 0.0005, "step": 2287 }, { "epoch": 0.27, "learning_rate": 2.1905967450271247e-06, "loss": 0.0004, "step": 2288 }, { "epoch": 0.27, "learning_rate": 2.190235081374322e-06, "loss": 0.107, "step": 2289 }, { "epoch": 0.27, "learning_rate": 2.1898734177215188e-06, "loss": 0.0005, "step": 2290 }, { "epoch": 0.27, "learning_rate": 2.189511754068716e-06, "loss": 0.0003, "step": 2291 }, { "epoch": 0.27, "learning_rate": 2.1891500904159132e-06, "loss": 0.0005, "step": 2292 }, { "epoch": 0.27, "learning_rate": 2.1887884267631105e-06, "loss": 0.0007, "step": 2293 }, { "epoch": 0.27, "learning_rate": 2.1884267631103073e-06, "loss": 0.0006, "step": 2294 }, { "epoch": 0.28, "learning_rate": 2.1880650994575045e-06, "loss": 0.0004, "step": 2295 }, { "epoch": 0.28, "learning_rate": 2.1877034358047017e-06, "loss": 0.0004, "step": 2296 }, { "epoch": 0.28, "learning_rate": 2.187341772151899e-06, "loss": 0.8001, "step": 2297 }, { "epoch": 0.28, "learning_rate": 2.1869801084990958e-06, "loss": 0.0004, "step": 2298 }, { "epoch": 0.28, "learning_rate": 2.186618444846293e-06, "loss": 0.0004, "step": 2299 }, { "epoch": 0.28, "learning_rate": 2.1862567811934902e-06, "loss": 0.7731, "step": 2300 }, { "epoch": 0.28, "learning_rate": 2.1858951175406875e-06, "loss": 0.0007, "step": 2301 }, { "epoch": 0.28, "learning_rate": 2.1855334538878843e-06, "loss": 0.0006, "step": 2302 }, { "epoch": 0.28, "learning_rate": 2.1851717902350815e-06, "loss": 0.0004, "step": 2303 }, { "epoch": 0.28, "learning_rate": 2.1848101265822787e-06, "loss": 0.9427, "step": 2304 }, { "epoch": 0.28, "learning_rate": 2.1844484629294755e-06, "loss": 0.0009, "step": 2305 }, { "epoch": 0.28, "learning_rate": 2.1840867992766728e-06, "loss": 0.0006, "step": 2306 }, { "epoch": 0.28, "learning_rate": 2.18372513562387e-06, "loss": 0.795, "step": 2307 }, { "epoch": 0.28, "learning_rate": 2.1833634719710672e-06, "loss": 0.0006, "step": 2308 }, { "epoch": 0.28, "learning_rate": 2.183001808318264e-06, "loss": 0.0064, "step": 2309 }, { "epoch": 0.28, "learning_rate": 2.182640144665461e-06, "loss": 0.0007, "step": 2310 }, { "epoch": 0.28, "learning_rate": 2.182278481012658e-06, "loss": 0.0019, "step": 2311 }, { "epoch": 0.28, "learning_rate": 2.1819168173598553e-06, "loss": 0.1639, "step": 2312 }, { "epoch": 0.28, "learning_rate": 2.1815551537070525e-06, "loss": 0.0006, "step": 2313 }, { "epoch": 0.28, "learning_rate": 2.1811934900542494e-06, "loss": 0.0005, "step": 2314 }, { "epoch": 0.28, "learning_rate": 2.1808318264014466e-06, "loss": 0.0006, "step": 2315 }, { "epoch": 0.28, "learning_rate": 2.180470162748644e-06, "loss": 0.0006, "step": 2316 }, { "epoch": 0.28, "learning_rate": 2.180108499095841e-06, "loss": 0.0004, "step": 2317 }, { "epoch": 0.28, "learning_rate": 2.179746835443038e-06, "loss": 0.0005, "step": 2318 }, { "epoch": 0.28, "learning_rate": 2.179385171790235e-06, "loss": 0.0026, "step": 2319 }, { "epoch": 0.28, "learning_rate": 2.1790235081374323e-06, "loss": 0.0752, "step": 2320 }, { "epoch": 0.28, "learning_rate": 2.1786618444846296e-06, "loss": 0.0012, "step": 2321 }, { "epoch": 0.28, "learning_rate": 2.1783001808318264e-06, "loss": 0.0006, "step": 2322 }, { "epoch": 0.28, "learning_rate": 2.1779385171790236e-06, "loss": 0.0006, "step": 2323 }, { "epoch": 0.28, "learning_rate": 2.177576853526221e-06, "loss": 0.0017, "step": 2324 }, { "epoch": 0.28, "learning_rate": 2.1772151898734176e-06, "loss": 0.0004, "step": 2325 }, { "epoch": 0.28, "learning_rate": 2.176853526220615e-06, "loss": 0.0006, "step": 2326 }, { "epoch": 0.28, "learning_rate": 2.176491862567812e-06, "loss": 0.0005, "step": 2327 }, { "epoch": 0.28, "learning_rate": 2.1761301989150093e-06, "loss": 0.0007, "step": 2328 }, { "epoch": 0.28, "learning_rate": 2.175768535262206e-06, "loss": 0.0156, "step": 2329 }, { "epoch": 0.28, "learning_rate": 2.1754068716094034e-06, "loss": 0.0004, "step": 2330 }, { "epoch": 0.28, "learning_rate": 2.1750452079566006e-06, "loss": 0.001, "step": 2331 }, { "epoch": 0.28, "learning_rate": 2.174683544303798e-06, "loss": 1.7027, "step": 2332 }, { "epoch": 0.28, "learning_rate": 2.1743218806509946e-06, "loss": 0.0073, "step": 2333 }, { "epoch": 0.28, "learning_rate": 2.1739602169981914e-06, "loss": 0.0004, "step": 2334 }, { "epoch": 0.28, "learning_rate": 2.1735985533453887e-06, "loss": 0.0258, "step": 2335 }, { "epoch": 0.28, "learning_rate": 2.173236889692586e-06, "loss": 0.0004, "step": 2336 }, { "epoch": 0.28, "learning_rate": 2.172875226039783e-06, "loss": 1.0237, "step": 2337 }, { "epoch": 0.28, "learning_rate": 2.17251356238698e-06, "loss": 0.2054, "step": 2338 }, { "epoch": 0.28, "learning_rate": 2.172151898734177e-06, "loss": 0.0004, "step": 2339 }, { "epoch": 0.28, "learning_rate": 2.1717902350813744e-06, "loss": 0.0005, "step": 2340 }, { "epoch": 0.28, "learning_rate": 2.1714285714285716e-06, "loss": 0.0015, "step": 2341 }, { "epoch": 0.28, "learning_rate": 2.1710669077757684e-06, "loss": 0.0005, "step": 2342 }, { "epoch": 0.28, "learning_rate": 2.1707052441229657e-06, "loss": 0.0006, "step": 2343 }, { "epoch": 0.28, "learning_rate": 2.170343580470163e-06, "loss": 0.0014, "step": 2344 }, { "epoch": 0.28, "learning_rate": 2.1699819168173597e-06, "loss": 0.0005, "step": 2345 }, { "epoch": 0.28, "learning_rate": 2.169620253164557e-06, "loss": 0.0007, "step": 2346 }, { "epoch": 0.28, "learning_rate": 2.169258589511754e-06, "loss": 0.0005, "step": 2347 }, { "epoch": 0.28, "learning_rate": 2.1688969258589514e-06, "loss": 0.0006, "step": 2348 }, { "epoch": 0.28, "learning_rate": 2.168535262206148e-06, "loss": 0.0004, "step": 2349 }, { "epoch": 0.28, "learning_rate": 2.1681735985533454e-06, "loss": 0.0008, "step": 2350 }, { "epoch": 0.28, "learning_rate": 2.1678119349005427e-06, "loss": 0.9017, "step": 2351 }, { "epoch": 0.28, "learning_rate": 2.16745027124774e-06, "loss": 0.0005, "step": 2352 }, { "epoch": 0.28, "learning_rate": 2.1670886075949367e-06, "loss": 0.0004, "step": 2353 }, { "epoch": 0.28, "learning_rate": 2.166726943942134e-06, "loss": 0.0018, "step": 2354 }, { "epoch": 0.28, "learning_rate": 2.166365280289331e-06, "loss": 0.0002, "step": 2355 }, { "epoch": 0.28, "learning_rate": 2.1660036166365284e-06, "loss": 0.0004, "step": 2356 }, { "epoch": 0.28, "learning_rate": 2.1656419529837252e-06, "loss": 0.0009, "step": 2357 }, { "epoch": 0.28, "learning_rate": 2.165280289330922e-06, "loss": 0.0005, "step": 2358 }, { "epoch": 0.28, "learning_rate": 2.1649186256781193e-06, "loss": 0.0007, "step": 2359 }, { "epoch": 0.28, "learning_rate": 2.1645569620253165e-06, "loss": 0.0009, "step": 2360 }, { "epoch": 0.28, "learning_rate": 2.1641952983725137e-06, "loss": 0.0004, "step": 2361 }, { "epoch": 0.28, "learning_rate": 2.1638336347197105e-06, "loss": 0.001, "step": 2362 }, { "epoch": 0.28, "learning_rate": 2.1634719710669078e-06, "loss": 0.0005, "step": 2363 }, { "epoch": 0.28, "learning_rate": 2.163110307414105e-06, "loss": 0.0008, "step": 2364 }, { "epoch": 0.28, "learning_rate": 2.162748643761302e-06, "loss": 0.8066, "step": 2365 }, { "epoch": 0.28, "learning_rate": 2.162386980108499e-06, "loss": 0.0004, "step": 2366 }, { "epoch": 0.28, "learning_rate": 2.1620253164556963e-06, "loss": 0.1615, "step": 2367 }, { "epoch": 0.28, "learning_rate": 2.1616636528028935e-06, "loss": 0.0004, "step": 2368 }, { "epoch": 0.28, "learning_rate": 2.1613019891500903e-06, "loss": 0.0005, "step": 2369 }, { "epoch": 0.28, "learning_rate": 2.1609403254972875e-06, "loss": 0.0005, "step": 2370 }, { "epoch": 0.28, "learning_rate": 2.1605786618444848e-06, "loss": 0.0796, "step": 2371 }, { "epoch": 0.28, "learning_rate": 2.160216998191682e-06, "loss": 0.8457, "step": 2372 }, { "epoch": 0.28, "learning_rate": 2.159855334538879e-06, "loss": 0.0006, "step": 2373 }, { "epoch": 0.28, "learning_rate": 2.159493670886076e-06, "loss": 0.0034, "step": 2374 }, { "epoch": 0.28, "learning_rate": 2.1591320072332733e-06, "loss": 0.0007, "step": 2375 }, { "epoch": 0.28, "learning_rate": 2.1587703435804705e-06, "loss": 0.0004, "step": 2376 }, { "epoch": 0.28, "learning_rate": 2.1584086799276673e-06, "loss": 0.2872, "step": 2377 }, { "epoch": 0.28, "learning_rate": 2.1580470162748645e-06, "loss": 0.3717, "step": 2378 }, { "epoch": 0.29, "learning_rate": 2.1576853526220618e-06, "loss": 0.0007, "step": 2379 }, { "epoch": 0.29, "learning_rate": 2.157323688969259e-06, "loss": 0.0006, "step": 2380 }, { "epoch": 0.29, "learning_rate": 2.156962025316456e-06, "loss": 0.0004, "step": 2381 }, { "epoch": 0.29, "learning_rate": 2.1566003616636526e-06, "loss": 0.0008, "step": 2382 }, { "epoch": 0.29, "learning_rate": 2.15623869801085e-06, "loss": 0.4826, "step": 2383 }, { "epoch": 0.29, "learning_rate": 2.155877034358047e-06, "loss": 0.0005, "step": 2384 }, { "epoch": 0.29, "learning_rate": 2.155515370705244e-06, "loss": 0.4738, "step": 2385 }, { "epoch": 0.29, "learning_rate": 2.155153707052441e-06, "loss": 0.0006, "step": 2386 }, { "epoch": 0.29, "learning_rate": 2.1547920433996383e-06, "loss": 0.0018, "step": 2387 }, { "epoch": 0.29, "learning_rate": 2.1544303797468356e-06, "loss": 0.0008, "step": 2388 }, { "epoch": 0.29, "learning_rate": 2.1540687160940324e-06, "loss": 0.0004, "step": 2389 }, { "epoch": 0.29, "learning_rate": 2.1537070524412296e-06, "loss": 0.0005, "step": 2390 }, { "epoch": 0.29, "learning_rate": 2.153345388788427e-06, "loss": 0.001, "step": 2391 }, { "epoch": 0.29, "learning_rate": 2.152983725135624e-06, "loss": 0.3446, "step": 2392 }, { "epoch": 0.29, "learning_rate": 2.152622061482821e-06, "loss": 0.0006, "step": 2393 }, { "epoch": 0.29, "learning_rate": 2.152260397830018e-06, "loss": 0.0005, "step": 2394 }, { "epoch": 0.29, "learning_rate": 2.1518987341772153e-06, "loss": 0.001, "step": 2395 }, { "epoch": 0.29, "learning_rate": 2.1515370705244126e-06, "loss": 0.0004, "step": 2396 }, { "epoch": 0.29, "learning_rate": 2.1511754068716094e-06, "loss": 0.0005, "step": 2397 }, { "epoch": 0.29, "learning_rate": 2.1508137432188066e-06, "loss": 0.0007, "step": 2398 }, { "epoch": 0.29, "learning_rate": 2.150452079566004e-06, "loss": 0.0005, "step": 2399 }, { "epoch": 0.29, "learning_rate": 2.150090415913201e-06, "loss": 0.0005, "step": 2400 }, { "epoch": 0.29, "learning_rate": 2.149728752260398e-06, "loss": 0.0005, "step": 2401 }, { "epoch": 0.29, "learning_rate": 2.149367088607595e-06, "loss": 0.0004, "step": 2402 }, { "epoch": 0.29, "learning_rate": 2.1490054249547923e-06, "loss": 0.0005, "step": 2403 }, { "epoch": 0.29, "learning_rate": 2.1486437613019896e-06, "loss": 0.3438, "step": 2404 }, { "epoch": 0.29, "learning_rate": 2.148282097649186e-06, "loss": 0.0006, "step": 2405 }, { "epoch": 0.29, "learning_rate": 2.147920433996383e-06, "loss": 0.0006, "step": 2406 }, { "epoch": 0.29, "learning_rate": 2.1475587703435804e-06, "loss": 0.0007, "step": 2407 }, { "epoch": 0.29, "learning_rate": 2.1471971066907777e-06, "loss": 0.0488, "step": 2408 }, { "epoch": 0.29, "learning_rate": 2.1468354430379745e-06, "loss": 0.0003, "step": 2409 }, { "epoch": 0.29, "learning_rate": 2.1464737793851717e-06, "loss": 0.8128, "step": 2410 }, { "epoch": 0.29, "learning_rate": 2.146112115732369e-06, "loss": 1.1209, "step": 2411 }, { "epoch": 0.29, "learning_rate": 2.145750452079566e-06, "loss": 0.0004, "step": 2412 }, { "epoch": 0.29, "learning_rate": 2.145388788426763e-06, "loss": 0.0005, "step": 2413 }, { "epoch": 0.29, "learning_rate": 2.14502712477396e-06, "loss": 0.0004, "step": 2414 }, { "epoch": 0.29, "learning_rate": 2.1446654611211574e-06, "loss": 0.0004, "step": 2415 }, { "epoch": 0.29, "learning_rate": 2.1443037974683547e-06, "loss": 0.0005, "step": 2416 }, { "epoch": 0.29, "learning_rate": 2.1439421338155515e-06, "loss": 0.0006, "step": 2417 }, { "epoch": 0.29, "learning_rate": 2.1435804701627487e-06, "loss": 0.0004, "step": 2418 }, { "epoch": 0.29, "learning_rate": 2.143218806509946e-06, "loss": 0.2783, "step": 2419 }, { "epoch": 0.29, "learning_rate": 2.142857142857143e-06, "loss": 0.0054, "step": 2420 }, { "epoch": 0.29, "learning_rate": 2.14249547920434e-06, "loss": 0.0107, "step": 2421 }, { "epoch": 0.29, "learning_rate": 2.142133815551537e-06, "loss": 0.0004, "step": 2422 }, { "epoch": 0.29, "learning_rate": 2.1417721518987344e-06, "loss": 0.0006, "step": 2423 }, { "epoch": 0.29, "learning_rate": 2.1414104882459312e-06, "loss": 0.0004, "step": 2424 }, { "epoch": 0.29, "learning_rate": 2.1410488245931285e-06, "loss": 0.7244, "step": 2425 }, { "epoch": 0.29, "learning_rate": 2.1406871609403257e-06, "loss": 0.0005, "step": 2426 }, { "epoch": 0.29, "learning_rate": 2.140325497287523e-06, "loss": 0.0006, "step": 2427 }, { "epoch": 0.29, "learning_rate": 2.1399638336347197e-06, "loss": 0.0008, "step": 2428 }, { "epoch": 0.29, "learning_rate": 2.1396021699819165e-06, "loss": 0.0045, "step": 2429 }, { "epoch": 0.29, "learning_rate": 2.1392405063291138e-06, "loss": 0.0041, "step": 2430 }, { "epoch": 0.29, "learning_rate": 2.138878842676311e-06, "loss": 0.0007, "step": 2431 }, { "epoch": 0.29, "learning_rate": 2.1385171790235082e-06, "loss": 0.8846, "step": 2432 }, { "epoch": 0.29, "learning_rate": 2.138155515370705e-06, "loss": 0.0008, "step": 2433 }, { "epoch": 0.29, "learning_rate": 2.1377938517179023e-06, "loss": 0.0005, "step": 2434 }, { "epoch": 0.29, "learning_rate": 2.1374321880650995e-06, "loss": 0.0671, "step": 2435 }, { "epoch": 0.29, "learning_rate": 2.1370705244122967e-06, "loss": 0.0006, "step": 2436 }, { "epoch": 0.29, "learning_rate": 2.1367088607594935e-06, "loss": 0.0006, "step": 2437 }, { "epoch": 0.29, "learning_rate": 2.1363471971066908e-06, "loss": 0.0005, "step": 2438 }, { "epoch": 0.29, "learning_rate": 2.135985533453888e-06, "loss": 0.7182, "step": 2439 }, { "epoch": 0.29, "learning_rate": 2.1356238698010852e-06, "loss": 0.0135, "step": 2440 }, { "epoch": 0.29, "learning_rate": 2.135262206148282e-06, "loss": 0.0011, "step": 2441 }, { "epoch": 0.29, "learning_rate": 2.1349005424954793e-06, "loss": 0.0007, "step": 2442 }, { "epoch": 0.29, "learning_rate": 2.1345388788426765e-06, "loss": 0.0006, "step": 2443 }, { "epoch": 0.29, "learning_rate": 2.1341772151898733e-06, "loss": 0.0006, "step": 2444 }, { "epoch": 0.29, "learning_rate": 2.1338155515370705e-06, "loss": 0.0005, "step": 2445 }, { "epoch": 0.29, "learning_rate": 2.1334538878842678e-06, "loss": 0.0006, "step": 2446 }, { "epoch": 0.29, "learning_rate": 2.133092224231465e-06, "loss": 0.0012, "step": 2447 }, { "epoch": 0.29, "learning_rate": 2.132730560578662e-06, "loss": 0.0022, "step": 2448 }, { "epoch": 0.29, "learning_rate": 2.132368896925859e-06, "loss": 0.0008, "step": 2449 }, { "epoch": 0.29, "learning_rate": 2.1320072332730563e-06, "loss": 0.0007, "step": 2450 }, { "epoch": 0.29, "learning_rate": 2.1316455696202535e-06, "loss": 1.7435, "step": 2451 }, { "epoch": 0.29, "learning_rate": 2.1312839059674503e-06, "loss": 0.9831, "step": 2452 }, { "epoch": 0.29, "learning_rate": 2.130922242314647e-06, "loss": 0.0018, "step": 2453 }, { "epoch": 0.29, "learning_rate": 2.1305605786618444e-06, "loss": 0.0006, "step": 2454 }, { "epoch": 0.29, "learning_rate": 2.1301989150090416e-06, "loss": 0.0021, "step": 2455 }, { "epoch": 0.29, "learning_rate": 2.129837251356239e-06, "loss": 0.0005, "step": 2456 }, { "epoch": 0.29, "learning_rate": 2.1294755877034356e-06, "loss": 0.0005, "step": 2457 }, { "epoch": 0.29, "learning_rate": 2.129113924050633e-06, "loss": 0.0005, "step": 2458 }, { "epoch": 0.29, "learning_rate": 2.12875226039783e-06, "loss": 0.0534, "step": 2459 }, { "epoch": 0.29, "learning_rate": 2.1283905967450273e-06, "loss": 0.1314, "step": 2460 }, { "epoch": 0.29, "learning_rate": 2.128028933092224e-06, "loss": 0.0026, "step": 2461 }, { "epoch": 0.3, "learning_rate": 2.1276672694394214e-06, "loss": 0.0011, "step": 2462 }, { "epoch": 0.3, "learning_rate": 2.1273056057866186e-06, "loss": 0.0008, "step": 2463 }, { "epoch": 0.3, "learning_rate": 2.1269439421338154e-06, "loss": 0.0008, "step": 2464 }, { "epoch": 0.3, "learning_rate": 2.1265822784810126e-06, "loss": 0.0012, "step": 2465 }, { "epoch": 0.3, "learning_rate": 2.12622061482821e-06, "loss": 0.0006, "step": 2466 }, { "epoch": 0.3, "learning_rate": 2.125858951175407e-06, "loss": 0.7668, "step": 2467 }, { "epoch": 0.3, "learning_rate": 2.125497287522604e-06, "loss": 0.0009, "step": 2468 }, { "epoch": 0.3, "learning_rate": 2.125135623869801e-06, "loss": 0.0005, "step": 2469 }, { "epoch": 0.3, "learning_rate": 2.1247739602169984e-06, "loss": 0.001, "step": 2470 }, { "epoch": 0.3, "learning_rate": 2.1244122965641956e-06, "loss": 0.0006, "step": 2471 }, { "epoch": 0.3, "learning_rate": 2.1240506329113924e-06, "loss": 0.0005, "step": 2472 }, { "epoch": 0.3, "learning_rate": 2.1236889692585896e-06, "loss": 0.7426, "step": 2473 }, { "epoch": 0.3, "learning_rate": 2.123327305605787e-06, "loss": 0.0003, "step": 2474 }, { "epoch": 0.3, "learning_rate": 2.122965641952984e-06, "loss": 0.3706, "step": 2475 }, { "epoch": 0.3, "learning_rate": 2.122603978300181e-06, "loss": 0.0006, "step": 2476 }, { "epoch": 0.3, "learning_rate": 2.1222423146473777e-06, "loss": 0.5702, "step": 2477 }, { "epoch": 0.3, "learning_rate": 2.121880650994575e-06, "loss": 0.0005, "step": 2478 }, { "epoch": 0.3, "learning_rate": 2.121518987341772e-06, "loss": 0.0005, "step": 2479 }, { "epoch": 0.3, "learning_rate": 2.1211573236889694e-06, "loss": 0.0005, "step": 2480 }, { "epoch": 0.3, "learning_rate": 2.120795660036166e-06, "loss": 0.0929, "step": 2481 }, { "epoch": 0.3, "learning_rate": 2.1204339963833634e-06, "loss": 0.0006, "step": 2482 }, { "epoch": 0.3, "learning_rate": 2.1200723327305607e-06, "loss": 0.9287, "step": 2483 }, { "epoch": 0.3, "learning_rate": 2.1197106690777575e-06, "loss": 0.0006, "step": 2484 }, { "epoch": 0.3, "learning_rate": 2.1193490054249547e-06, "loss": 0.0007, "step": 2485 }, { "epoch": 0.3, "learning_rate": 2.118987341772152e-06, "loss": 0.0018, "step": 2486 }, { "epoch": 0.3, "learning_rate": 2.118625678119349e-06, "loss": 0.0005, "step": 2487 }, { "epoch": 0.3, "learning_rate": 2.118264014466546e-06, "loss": 0.1911, "step": 2488 }, { "epoch": 0.3, "learning_rate": 2.117902350813743e-06, "loss": 0.0007, "step": 2489 }, { "epoch": 0.3, "learning_rate": 2.1175406871609404e-06, "loss": 0.0006, "step": 2490 }, { "epoch": 0.3, "learning_rate": 2.1171790235081377e-06, "loss": 0.0008, "step": 2491 }, { "epoch": 0.3, "learning_rate": 2.1168173598553345e-06, "loss": 0.0004, "step": 2492 }, { "epoch": 0.3, "learning_rate": 2.1164556962025317e-06, "loss": 0.0003, "step": 2493 }, { "epoch": 0.3, "learning_rate": 2.116094032549729e-06, "loss": 0.7054, "step": 2494 }, { "epoch": 0.3, "learning_rate": 2.115732368896926e-06, "loss": 0.0006, "step": 2495 }, { "epoch": 0.3, "learning_rate": 2.115370705244123e-06, "loss": 0.0011, "step": 2496 }, { "epoch": 0.3, "learning_rate": 2.11500904159132e-06, "loss": 0.0005, "step": 2497 }, { "epoch": 0.3, "learning_rate": 2.1146473779385174e-06, "loss": 0.0007, "step": 2498 }, { "epoch": 0.3, "learning_rate": 2.1142857142857147e-06, "loss": 0.0005, "step": 2499 }, { "epoch": 0.3, "learning_rate": 2.1139240506329115e-06, "loss": 0.0005, "step": 2500 }, { "epoch": 0.3, "learning_rate": 2.1135623869801083e-06, "loss": 0.0005, "step": 2501 }, { "epoch": 0.3, "learning_rate": 2.1132007233273055e-06, "loss": 0.8977, "step": 2502 }, { "epoch": 0.3, "learning_rate": 2.1128390596745028e-06, "loss": 0.0006, "step": 2503 }, { "epoch": 0.3, "learning_rate": 2.1124773960216996e-06, "loss": 0.0004, "step": 2504 }, { "epoch": 0.3, "learning_rate": 2.112115732368897e-06, "loss": 0.0008, "step": 2505 }, { "epoch": 0.3, "learning_rate": 2.111754068716094e-06, "loss": 0.0071, "step": 2506 }, { "epoch": 0.3, "learning_rate": 2.1113924050632913e-06, "loss": 0.0275, "step": 2507 }, { "epoch": 0.3, "learning_rate": 2.111030741410488e-06, "loss": 0.0005, "step": 2508 }, { "epoch": 0.3, "learning_rate": 2.1106690777576853e-06, "loss": 0.0005, "step": 2509 }, { "epoch": 0.3, "learning_rate": 2.1103074141048825e-06, "loss": 0.0481, "step": 2510 }, { "epoch": 0.3, "learning_rate": 2.1099457504520798e-06, "loss": 0.8434, "step": 2511 }, { "epoch": 0.3, "learning_rate": 2.1095840867992766e-06, "loss": 0.0009, "step": 2512 }, { "epoch": 0.3, "learning_rate": 2.109222423146474e-06, "loss": 0.0005, "step": 2513 }, { "epoch": 0.3, "learning_rate": 2.108860759493671e-06, "loss": 0.0006, "step": 2514 }, { "epoch": 0.3, "learning_rate": 2.1084990958408683e-06, "loss": 0.0005, "step": 2515 }, { "epoch": 0.3, "learning_rate": 2.108137432188065e-06, "loss": 0.0005, "step": 2516 }, { "epoch": 0.3, "learning_rate": 2.1077757685352623e-06, "loss": 0.0011, "step": 2517 }, { "epoch": 0.3, "learning_rate": 2.1074141048824595e-06, "loss": 0.0003, "step": 2518 }, { "epoch": 0.3, "learning_rate": 2.1070524412296568e-06, "loss": 0.0005, "step": 2519 }, { "epoch": 0.3, "learning_rate": 2.1066907775768536e-06, "loss": 0.0007, "step": 2520 }, { "epoch": 0.3, "learning_rate": 2.106329113924051e-06, "loss": 0.0009, "step": 2521 }, { "epoch": 0.3, "learning_rate": 2.105967450271248e-06, "loss": 0.0005, "step": 2522 }, { "epoch": 0.3, "learning_rate": 2.1056057866184453e-06, "loss": 0.0005, "step": 2523 }, { "epoch": 0.3, "learning_rate": 2.105244122965642e-06, "loss": 0.8654, "step": 2524 }, { "epoch": 0.3, "learning_rate": 2.104882459312839e-06, "loss": 0.0006, "step": 2525 }, { "epoch": 0.3, "learning_rate": 2.104520795660036e-06, "loss": 0.0006, "step": 2526 }, { "epoch": 0.3, "learning_rate": 2.1041591320072333e-06, "loss": 0.0004, "step": 2527 }, { "epoch": 0.3, "learning_rate": 2.10379746835443e-06, "loss": 0.0006, "step": 2528 }, { "epoch": 0.3, "learning_rate": 2.1034358047016274e-06, "loss": 0.0005, "step": 2529 }, { "epoch": 0.3, "learning_rate": 2.1030741410488246e-06, "loss": 0.0298, "step": 2530 }, { "epoch": 0.3, "learning_rate": 2.102712477396022e-06, "loss": 0.0012, "step": 2531 }, { "epoch": 0.3, "learning_rate": 2.1023508137432186e-06, "loss": 0.0019, "step": 2532 }, { "epoch": 0.3, "learning_rate": 2.101989150090416e-06, "loss": 0.0004, "step": 2533 }, { "epoch": 0.3, "learning_rate": 2.101627486437613e-06, "loss": 0.0005, "step": 2534 }, { "epoch": 0.3, "learning_rate": 2.1012658227848103e-06, "loss": 0.0005, "step": 2535 }, { "epoch": 0.3, "learning_rate": 2.100904159132007e-06, "loss": 0.0005, "step": 2536 }, { "epoch": 0.3, "learning_rate": 2.1005424954792044e-06, "loss": 0.0005, "step": 2537 }, { "epoch": 0.3, "learning_rate": 2.1001808318264016e-06, "loss": 0.9553, "step": 2538 }, { "epoch": 0.3, "learning_rate": 2.099819168173599e-06, "loss": 0.0005, "step": 2539 }, { "epoch": 0.3, "learning_rate": 2.0994575045207956e-06, "loss": 0.0005, "step": 2540 }, { "epoch": 0.3, "learning_rate": 2.099095840867993e-06, "loss": 0.0036, "step": 2541 }, { "epoch": 0.3, "learning_rate": 2.09873417721519e-06, "loss": 0.0005, "step": 2542 }, { "epoch": 0.3, "learning_rate": 2.0983725135623873e-06, "loss": 0.0005, "step": 2543 }, { "epoch": 0.3, "learning_rate": 2.098010849909584e-06, "loss": 0.0004, "step": 2544 }, { "epoch": 0.3, "learning_rate": 2.0976491862567814e-06, "loss": 0.0005, "step": 2545 }, { "epoch": 0.31, "learning_rate": 2.0972875226039786e-06, "loss": 0.0005, "step": 2546 }, { "epoch": 0.31, "learning_rate": 2.0969258589511754e-06, "loss": 0.0005, "step": 2547 }, { "epoch": 0.31, "learning_rate": 2.0965641952983726e-06, "loss": 0.0006, "step": 2548 }, { "epoch": 0.31, "learning_rate": 2.0962025316455695e-06, "loss": 0.0005, "step": 2549 }, { "epoch": 0.31, "learning_rate": 2.0958408679927667e-06, "loss": 0.0005, "step": 2550 }, { "epoch": 0.31, "learning_rate": 2.095479204339964e-06, "loss": 0.0004, "step": 2551 }, { "epoch": 0.31, "learning_rate": 2.0951175406871607e-06, "loss": 0.0004, "step": 2552 }, { "epoch": 0.31, "learning_rate": 2.094755877034358e-06, "loss": 0.0006, "step": 2553 }, { "epoch": 0.31, "learning_rate": 2.094394213381555e-06, "loss": 0.0004, "step": 2554 }, { "epoch": 0.31, "learning_rate": 2.0940325497287524e-06, "loss": 0.0006, "step": 2555 }, { "epoch": 0.31, "learning_rate": 2.0936708860759492e-06, "loss": 0.0008, "step": 2556 }, { "epoch": 0.31, "learning_rate": 2.0933092224231465e-06, "loss": 0.0009, "step": 2557 }, { "epoch": 0.31, "learning_rate": 2.0929475587703437e-06, "loss": 0.0007, "step": 2558 }, { "epoch": 0.31, "learning_rate": 2.092585895117541e-06, "loss": 0.0006, "step": 2559 }, { "epoch": 0.31, "learning_rate": 2.0922242314647377e-06, "loss": 0.0005, "step": 2560 }, { "epoch": 0.31, "learning_rate": 2.091862567811935e-06, "loss": 0.0005, "step": 2561 }, { "epoch": 0.31, "learning_rate": 2.091500904159132e-06, "loss": 0.4695, "step": 2562 }, { "epoch": 0.31, "learning_rate": 2.091139240506329e-06, "loss": 0.0005, "step": 2563 }, { "epoch": 0.31, "learning_rate": 2.0907775768535262e-06, "loss": 0.1667, "step": 2564 }, { "epoch": 0.31, "learning_rate": 2.0904159132007235e-06, "loss": 0.9597, "step": 2565 }, { "epoch": 0.31, "learning_rate": 2.0900542495479207e-06, "loss": 0.0007, "step": 2566 }, { "epoch": 0.31, "learning_rate": 2.0896925858951175e-06, "loss": 0.0007, "step": 2567 }, { "epoch": 0.31, "learning_rate": 2.0893309222423147e-06, "loss": 0.0005, "step": 2568 }, { "epoch": 0.31, "learning_rate": 2.088969258589512e-06, "loss": 0.8962, "step": 2569 }, { "epoch": 0.31, "learning_rate": 2.088607594936709e-06, "loss": 0.0006, "step": 2570 }, { "epoch": 0.31, "learning_rate": 2.088245931283906e-06, "loss": 0.0005, "step": 2571 }, { "epoch": 0.31, "learning_rate": 2.087884267631103e-06, "loss": 0.0008, "step": 2572 }, { "epoch": 0.31, "learning_rate": 2.0875226039783e-06, "loss": 0.0004, "step": 2573 }, { "epoch": 0.31, "learning_rate": 2.0871609403254973e-06, "loss": 0.001, "step": 2574 }, { "epoch": 0.31, "learning_rate": 2.0867992766726945e-06, "loss": 0.2251, "step": 2575 }, { "epoch": 0.31, "learning_rate": 2.0864376130198913e-06, "loss": 0.0012, "step": 2576 }, { "epoch": 0.31, "learning_rate": 2.0860759493670885e-06, "loss": 0.0005, "step": 2577 }, { "epoch": 0.31, "learning_rate": 2.0857142857142858e-06, "loss": 0.0005, "step": 2578 }, { "epoch": 0.31, "learning_rate": 2.085352622061483e-06, "loss": 0.0005, "step": 2579 }, { "epoch": 0.31, "learning_rate": 2.08499095840868e-06, "loss": 0.0006, "step": 2580 }, { "epoch": 0.31, "learning_rate": 2.084629294755877e-06, "loss": 0.0005, "step": 2581 }, { "epoch": 0.31, "learning_rate": 2.0842676311030743e-06, "loss": 0.0005, "step": 2582 }, { "epoch": 0.31, "learning_rate": 2.083905967450271e-06, "loss": 0.88, "step": 2583 }, { "epoch": 0.31, "learning_rate": 2.0835443037974683e-06, "loss": 0.0006, "step": 2584 }, { "epoch": 0.31, "learning_rate": 2.0831826401446655e-06, "loss": 0.0006, "step": 2585 }, { "epoch": 0.31, "learning_rate": 2.0828209764918628e-06, "loss": 0.0005, "step": 2586 }, { "epoch": 0.31, "learning_rate": 2.0824593128390596e-06, "loss": 0.0044, "step": 2587 }, { "epoch": 0.31, "learning_rate": 2.082097649186257e-06, "loss": 0.0006, "step": 2588 }, { "epoch": 0.31, "learning_rate": 2.081735985533454e-06, "loss": 0.0004, "step": 2589 }, { "epoch": 0.31, "learning_rate": 2.0813743218806513e-06, "loss": 0.5952, "step": 2590 }, { "epoch": 0.31, "learning_rate": 2.081012658227848e-06, "loss": 0.0005, "step": 2591 }, { "epoch": 0.31, "learning_rate": 2.0806509945750453e-06, "loss": 0.0006, "step": 2592 }, { "epoch": 0.31, "learning_rate": 2.0802893309222425e-06, "loss": 0.0004, "step": 2593 }, { "epoch": 0.31, "learning_rate": 2.0799276672694398e-06, "loss": 0.0007, "step": 2594 }, { "epoch": 0.31, "learning_rate": 2.0795660036166366e-06, "loss": 0.1063, "step": 2595 }, { "epoch": 0.31, "learning_rate": 2.0792043399638334e-06, "loss": 0.0006, "step": 2596 }, { "epoch": 0.31, "learning_rate": 2.0788426763110306e-06, "loss": 0.0005, "step": 2597 }, { "epoch": 0.31, "learning_rate": 2.078481012658228e-06, "loss": 0.0013, "step": 2598 }, { "epoch": 0.31, "learning_rate": 2.078119349005425e-06, "loss": 0.7051, "step": 2599 }, { "epoch": 0.31, "learning_rate": 2.077757685352622e-06, "loss": 0.0008, "step": 2600 }, { "epoch": 0.31, "learning_rate": 2.077396021699819e-06, "loss": 0.0007, "step": 2601 }, { "epoch": 0.31, "learning_rate": 2.0770343580470164e-06, "loss": 0.0014, "step": 2602 }, { "epoch": 0.31, "learning_rate": 2.076672694394213e-06, "loss": 0.0006, "step": 2603 }, { "epoch": 0.31, "learning_rate": 2.0763110307414104e-06, "loss": 0.0008, "step": 2604 }, { "epoch": 0.31, "learning_rate": 2.0759493670886076e-06, "loss": 0.0005, "step": 2605 }, { "epoch": 0.31, "learning_rate": 2.075587703435805e-06, "loss": 0.9334, "step": 2606 }, { "epoch": 0.31, "learning_rate": 2.0752260397830017e-06, "loss": 0.0019, "step": 2607 }, { "epoch": 0.31, "learning_rate": 2.074864376130199e-06, "loss": 0.0007, "step": 2608 }, { "epoch": 0.31, "learning_rate": 2.074502712477396e-06, "loss": 0.4305, "step": 2609 }, { "epoch": 0.31, "learning_rate": 2.0741410488245934e-06, "loss": 0.0013, "step": 2610 }, { "epoch": 0.31, "learning_rate": 2.07377938517179e-06, "loss": 0.0019, "step": 2611 }, { "epoch": 0.31, "learning_rate": 2.0734177215189874e-06, "loss": 0.2552, "step": 2612 }, { "epoch": 0.31, "learning_rate": 2.0730560578661846e-06, "loss": 0.0007, "step": 2613 }, { "epoch": 0.31, "learning_rate": 2.072694394213382e-06, "loss": 0.0034, "step": 2614 }, { "epoch": 0.31, "learning_rate": 2.0723327305605787e-06, "loss": 0.0011, "step": 2615 }, { "epoch": 0.31, "learning_rate": 2.071971066907776e-06, "loss": 0.0829, "step": 2616 }, { "epoch": 0.31, "learning_rate": 2.071609403254973e-06, "loss": 0.0007, "step": 2617 }, { "epoch": 0.31, "learning_rate": 2.0712477396021704e-06, "loss": 0.0008, "step": 2618 }, { "epoch": 0.31, "learning_rate": 2.070886075949367e-06, "loss": 0.2195, "step": 2619 }, { "epoch": 0.31, "learning_rate": 2.070524412296564e-06, "loss": 0.0003, "step": 2620 }, { "epoch": 0.31, "learning_rate": 2.070162748643761e-06, "loss": 0.513, "step": 2621 }, { "epoch": 0.31, "learning_rate": 2.0698010849909584e-06, "loss": 0.0006, "step": 2622 }, { "epoch": 0.31, "learning_rate": 2.0694394213381552e-06, "loss": 0.0481, "step": 2623 }, { "epoch": 0.31, "learning_rate": 2.0690777576853525e-06, "loss": 0.0006, "step": 2624 }, { "epoch": 0.31, "learning_rate": 2.0687160940325497e-06, "loss": 0.0018, "step": 2625 }, { "epoch": 0.31, "learning_rate": 2.068354430379747e-06, "loss": 0.0003, "step": 2626 }, { "epoch": 0.31, "learning_rate": 2.0679927667269437e-06, "loss": 0.1253, "step": 2627 }, { "epoch": 0.31, "learning_rate": 2.067631103074141e-06, "loss": 0.691, "step": 2628 }, { "epoch": 0.32, "learning_rate": 2.067269439421338e-06, "loss": 0.0197, "step": 2629 }, { "epoch": 0.32, "learning_rate": 2.0669077757685354e-06, "loss": 0.0006, "step": 2630 }, { "epoch": 0.32, "learning_rate": 2.0665461121157322e-06, "loss": 0.0006, "step": 2631 }, { "epoch": 0.32, "learning_rate": 2.0661844484629295e-06, "loss": 0.0007, "step": 2632 }, { "epoch": 0.32, "learning_rate": 2.0658227848101267e-06, "loss": 0.0004, "step": 2633 }, { "epoch": 0.32, "learning_rate": 2.065461121157324e-06, "loss": 0.6268, "step": 2634 }, { "epoch": 0.32, "learning_rate": 2.0650994575045207e-06, "loss": 0.0071, "step": 2635 }, { "epoch": 0.32, "learning_rate": 2.064737793851718e-06, "loss": 0.0004, "step": 2636 }, { "epoch": 0.32, "learning_rate": 2.064376130198915e-06, "loss": 0.3339, "step": 2637 }, { "epoch": 0.32, "learning_rate": 2.0640144665461124e-06, "loss": 0.0006, "step": 2638 }, { "epoch": 0.32, "learning_rate": 2.0636528028933092e-06, "loss": 0.0004, "step": 2639 }, { "epoch": 0.32, "learning_rate": 2.0632911392405065e-06, "loss": 0.0005, "step": 2640 }, { "epoch": 0.32, "learning_rate": 2.0629294755877037e-06, "loss": 0.0004, "step": 2641 }, { "epoch": 0.32, "learning_rate": 2.062567811934901e-06, "loss": 0.0004, "step": 2642 }, { "epoch": 0.32, "learning_rate": 2.0622061482820977e-06, "loss": 0.0005, "step": 2643 }, { "epoch": 0.32, "learning_rate": 2.0618444846292946e-06, "loss": 0.0015, "step": 2644 }, { "epoch": 0.32, "learning_rate": 2.061482820976492e-06, "loss": 0.0333, "step": 2645 }, { "epoch": 0.32, "learning_rate": 2.061121157323689e-06, "loss": 0.0006, "step": 2646 }, { "epoch": 0.32, "learning_rate": 2.060759493670886e-06, "loss": 0.896, "step": 2647 }, { "epoch": 0.32, "learning_rate": 2.060397830018083e-06, "loss": 0.0259, "step": 2648 }, { "epoch": 0.32, "learning_rate": 2.0600361663652803e-06, "loss": 0.0005, "step": 2649 }, { "epoch": 0.32, "learning_rate": 2.0596745027124775e-06, "loss": 0.0004, "step": 2650 }, { "epoch": 0.32, "learning_rate": 2.0593128390596743e-06, "loss": 0.0003, "step": 2651 }, { "epoch": 0.32, "learning_rate": 2.0589511754068716e-06, "loss": 0.0004, "step": 2652 }, { "epoch": 0.32, "learning_rate": 2.058589511754069e-06, "loss": 0.0005, "step": 2653 }, { "epoch": 0.32, "learning_rate": 2.058227848101266e-06, "loss": 0.0007, "step": 2654 }, { "epoch": 0.32, "learning_rate": 2.057866184448463e-06, "loss": 0.0005, "step": 2655 }, { "epoch": 0.32, "learning_rate": 2.05750452079566e-06, "loss": 0.0004, "step": 2656 }, { "epoch": 0.32, "learning_rate": 2.0571428571428573e-06, "loss": 0.003, "step": 2657 }, { "epoch": 0.32, "learning_rate": 2.0567811934900545e-06, "loss": 0.001, "step": 2658 }, { "epoch": 0.32, "learning_rate": 2.0564195298372513e-06, "loss": 0.0005, "step": 2659 }, { "epoch": 0.32, "learning_rate": 2.0560578661844486e-06, "loss": 0.5533, "step": 2660 }, { "epoch": 0.32, "learning_rate": 2.055696202531646e-06, "loss": 0.001, "step": 2661 }, { "epoch": 0.32, "learning_rate": 2.055334538878843e-06, "loss": 0.0007, "step": 2662 }, { "epoch": 0.32, "learning_rate": 2.05497287522604e-06, "loss": 0.0004, "step": 2663 }, { "epoch": 0.32, "learning_rate": 2.054611211573237e-06, "loss": 0.0008, "step": 2664 }, { "epoch": 0.32, "learning_rate": 2.0542495479204343e-06, "loss": 0.0003, "step": 2665 }, { "epoch": 0.32, "learning_rate": 2.053887884267631e-06, "loss": 0.0004, "step": 2666 }, { "epoch": 0.32, "learning_rate": 2.0535262206148283e-06, "loss": 0.0004, "step": 2667 }, { "epoch": 0.32, "learning_rate": 2.053164556962025e-06, "loss": 0.0006, "step": 2668 }, { "epoch": 0.32, "learning_rate": 2.0528028933092224e-06, "loss": 0.0004, "step": 2669 }, { "epoch": 0.32, "learning_rate": 2.0524412296564196e-06, "loss": 0.0004, "step": 2670 }, { "epoch": 0.32, "learning_rate": 2.0520795660036164e-06, "loss": 0.6395, "step": 2671 }, { "epoch": 0.32, "learning_rate": 2.0517179023508136e-06, "loss": 0.0179, "step": 2672 }, { "epoch": 0.32, "learning_rate": 2.051356238698011e-06, "loss": 0.0004, "step": 2673 }, { "epoch": 0.32, "learning_rate": 2.050994575045208e-06, "loss": 0.0004, "step": 2674 }, { "epoch": 0.32, "learning_rate": 2.050632911392405e-06, "loss": 0.0004, "step": 2675 }, { "epoch": 0.32, "learning_rate": 2.050271247739602e-06, "loss": 0.0007, "step": 2676 }, { "epoch": 0.32, "learning_rate": 2.0499095840867994e-06, "loss": 0.034, "step": 2677 }, { "epoch": 0.32, "learning_rate": 2.0495479204339966e-06, "loss": 0.0004, "step": 2678 }, { "epoch": 0.32, "learning_rate": 2.0491862567811934e-06, "loss": 0.0004, "step": 2679 }, { "epoch": 0.32, "learning_rate": 2.0488245931283906e-06, "loss": 0.0008, "step": 2680 }, { "epoch": 0.32, "learning_rate": 2.048462929475588e-06, "loss": 0.0004, "step": 2681 }, { "epoch": 0.32, "learning_rate": 2.048101265822785e-06, "loss": 0.0004, "step": 2682 }, { "epoch": 0.32, "learning_rate": 2.047739602169982e-06, "loss": 0.0004, "step": 2683 }, { "epoch": 0.32, "learning_rate": 2.047377938517179e-06, "loss": 0.0004, "step": 2684 }, { "epoch": 0.32, "learning_rate": 2.0470162748643764e-06, "loss": 0.0004, "step": 2685 }, { "epoch": 0.32, "learning_rate": 2.046654611211573e-06, "loss": 0.0005, "step": 2686 }, { "epoch": 0.32, "learning_rate": 2.0462929475587704e-06, "loss": 0.0003, "step": 2687 }, { "epoch": 0.32, "learning_rate": 2.0459312839059676e-06, "loss": 0.0008, "step": 2688 }, { "epoch": 0.32, "learning_rate": 2.045569620253165e-06, "loss": 0.0287, "step": 2689 }, { "epoch": 0.32, "learning_rate": 2.0452079566003617e-06, "loss": 0.0003, "step": 2690 }, { "epoch": 0.32, "learning_rate": 2.044846292947559e-06, "loss": 0.9211, "step": 2691 }, { "epoch": 0.32, "learning_rate": 2.0444846292947557e-06, "loss": 0.0006, "step": 2692 }, { "epoch": 0.32, "learning_rate": 2.044122965641953e-06, "loss": 0.7727, "step": 2693 }, { "epoch": 0.32, "learning_rate": 2.04376130198915e-06, "loss": 0.0004, "step": 2694 }, { "epoch": 0.32, "learning_rate": 2.043399638336347e-06, "loss": 0.0003, "step": 2695 }, { "epoch": 0.32, "learning_rate": 2.0430379746835442e-06, "loss": 0.0004, "step": 2696 }, { "epoch": 0.32, "learning_rate": 2.0426763110307415e-06, "loss": 0.0004, "step": 2697 }, { "epoch": 0.32, "learning_rate": 2.0423146473779387e-06, "loss": 0.0003, "step": 2698 }, { "epoch": 0.32, "learning_rate": 2.0419529837251355e-06, "loss": 0.0004, "step": 2699 }, { "epoch": 0.32, "learning_rate": 2.0415913200723327e-06, "loss": 0.063, "step": 2700 }, { "epoch": 0.32, "learning_rate": 2.04122965641953e-06, "loss": 0.0004, "step": 2701 }, { "epoch": 0.32, "learning_rate": 2.0408679927667268e-06, "loss": 0.1952, "step": 2702 }, { "epoch": 0.32, "learning_rate": 2.040506329113924e-06, "loss": 1.0291, "step": 2703 }, { "epoch": 0.32, "learning_rate": 2.0401446654611212e-06, "loss": 0.9123, "step": 2704 }, { "epoch": 0.32, "learning_rate": 2.0397830018083185e-06, "loss": 0.0004, "step": 2705 }, { "epoch": 0.32, "learning_rate": 2.0394213381555153e-06, "loss": 0.0005, "step": 2706 }, { "epoch": 0.32, "learning_rate": 2.0390596745027125e-06, "loss": 0.0005, "step": 2707 }, { "epoch": 0.32, "learning_rate": 2.0386980108499097e-06, "loss": 0.0005, "step": 2708 }, { "epoch": 0.32, "learning_rate": 2.038336347197107e-06, "loss": 0.0012, "step": 2709 }, { "epoch": 0.32, "learning_rate": 2.0379746835443038e-06, "loss": 0.0007, "step": 2710 }, { "epoch": 0.32, "learning_rate": 2.037613019891501e-06, "loss": 0.0008, "step": 2711 }, { "epoch": 0.32, "learning_rate": 2.0372513562386982e-06, "loss": 0.0004, "step": 2712 }, { "epoch": 0.33, "learning_rate": 2.0368896925858955e-06, "loss": 0.0004, "step": 2713 }, { "epoch": 0.33, "learning_rate": 2.0365280289330923e-06, "loss": 0.0005, "step": 2714 }, { "epoch": 0.33, "learning_rate": 2.0361663652802895e-06, "loss": 0.512, "step": 2715 }, { "epoch": 0.33, "learning_rate": 2.0358047016274863e-06, "loss": 0.0007, "step": 2716 }, { "epoch": 0.33, "learning_rate": 2.0354430379746835e-06, "loss": 0.0004, "step": 2717 }, { "epoch": 0.33, "learning_rate": 2.0350813743218808e-06, "loss": 0.8897, "step": 2718 }, { "epoch": 0.33, "learning_rate": 2.0347197106690776e-06, "loss": 0.0016, "step": 2719 }, { "epoch": 0.33, "learning_rate": 2.034358047016275e-06, "loss": 0.0009, "step": 2720 }, { "epoch": 0.33, "learning_rate": 2.033996383363472e-06, "loss": 0.0005, "step": 2721 }, { "epoch": 0.33, "learning_rate": 2.033634719710669e-06, "loss": 0.0006, "step": 2722 }, { "epoch": 0.33, "learning_rate": 2.033273056057866e-06, "loss": 0.001, "step": 2723 }, { "epoch": 0.33, "learning_rate": 2.0329113924050633e-06, "loss": 0.001, "step": 2724 }, { "epoch": 0.33, "learning_rate": 2.0325497287522605e-06, "loss": 0.0006, "step": 2725 }, { "epoch": 0.33, "learning_rate": 2.0321880650994573e-06, "loss": 0.0004, "step": 2726 }, { "epoch": 0.33, "learning_rate": 2.0318264014466546e-06, "loss": 0.0008, "step": 2727 }, { "epoch": 0.33, "learning_rate": 2.031464737793852e-06, "loss": 0.0004, "step": 2728 }, { "epoch": 0.33, "learning_rate": 2.031103074141049e-06, "loss": 0.0333, "step": 2729 }, { "epoch": 0.33, "learning_rate": 2.030741410488246e-06, "loss": 0.223, "step": 2730 }, { "epoch": 0.33, "learning_rate": 2.030379746835443e-06, "loss": 0.0005, "step": 2731 }, { "epoch": 0.33, "learning_rate": 2.0300180831826403e-06, "loss": 0.0101, "step": 2732 }, { "epoch": 0.33, "learning_rate": 2.0296564195298375e-06, "loss": 0.0027, "step": 2733 }, { "epoch": 0.33, "learning_rate": 2.0292947558770344e-06, "loss": 1.0027, "step": 2734 }, { "epoch": 0.33, "learning_rate": 2.0289330922242316e-06, "loss": 0.0006, "step": 2735 }, { "epoch": 0.33, "learning_rate": 2.028571428571429e-06, "loss": 0.0014, "step": 2736 }, { "epoch": 0.33, "learning_rate": 2.028209764918626e-06, "loss": 0.001, "step": 2737 }, { "epoch": 0.33, "learning_rate": 2.027848101265823e-06, "loss": 0.0005, "step": 2738 }, { "epoch": 0.33, "learning_rate": 2.02748643761302e-06, "loss": 0.0007, "step": 2739 }, { "epoch": 0.33, "learning_rate": 2.027124773960217e-06, "loss": 0.0006, "step": 2740 }, { "epoch": 0.33, "learning_rate": 2.026763110307414e-06, "loss": 0.0006, "step": 2741 }, { "epoch": 0.33, "learning_rate": 2.026401446654611e-06, "loss": 0.0008, "step": 2742 }, { "epoch": 0.33, "learning_rate": 2.026039783001808e-06, "loss": 0.0011, "step": 2743 }, { "epoch": 0.33, "learning_rate": 2.0256781193490054e-06, "loss": 0.0008, "step": 2744 }, { "epoch": 0.33, "learning_rate": 2.0253164556962026e-06, "loss": 0.0004, "step": 2745 }, { "epoch": 0.33, "learning_rate": 2.0249547920433994e-06, "loss": 0.4037, "step": 2746 }, { "epoch": 0.33, "learning_rate": 2.0245931283905967e-06, "loss": 0.0003, "step": 2747 }, { "epoch": 0.33, "learning_rate": 2.024231464737794e-06, "loss": 0.0122, "step": 2748 }, { "epoch": 0.33, "learning_rate": 2.023869801084991e-06, "loss": 0.0006, "step": 2749 }, { "epoch": 0.33, "learning_rate": 2.023508137432188e-06, "loss": 0.0006, "step": 2750 }, { "epoch": 0.33, "learning_rate": 2.023146473779385e-06, "loss": 0.0005, "step": 2751 }, { "epoch": 0.33, "learning_rate": 2.0227848101265824e-06, "loss": 0.0086, "step": 2752 }, { "epoch": 0.33, "learning_rate": 2.0224231464737796e-06, "loss": 0.0015, "step": 2753 }, { "epoch": 0.33, "learning_rate": 2.0220614828209764e-06, "loss": 0.0004, "step": 2754 }, { "epoch": 0.33, "learning_rate": 2.0216998191681737e-06, "loss": 0.001, "step": 2755 }, { "epoch": 0.33, "learning_rate": 2.021338155515371e-06, "loss": 0.0007, "step": 2756 }, { "epoch": 0.33, "learning_rate": 2.020976491862568e-06, "loss": 0.0005, "step": 2757 }, { "epoch": 0.33, "learning_rate": 2.020614828209765e-06, "loss": 0.002, "step": 2758 }, { "epoch": 0.33, "learning_rate": 2.020253164556962e-06, "loss": 0.0005, "step": 2759 }, { "epoch": 0.33, "learning_rate": 2.0198915009041594e-06, "loss": 0.0004, "step": 2760 }, { "epoch": 0.33, "learning_rate": 2.0195298372513566e-06, "loss": 0.0007, "step": 2761 }, { "epoch": 0.33, "learning_rate": 2.0191681735985534e-06, "loss": 0.0005, "step": 2762 }, { "epoch": 0.33, "learning_rate": 2.0188065099457507e-06, "loss": 0.0007, "step": 2763 }, { "epoch": 0.33, "learning_rate": 2.0184448462929475e-06, "loss": 0.0038, "step": 2764 }, { "epoch": 0.33, "learning_rate": 2.0180831826401447e-06, "loss": 0.0003, "step": 2765 }, { "epoch": 0.33, "learning_rate": 2.0177215189873415e-06, "loss": 0.0005, "step": 2766 }, { "epoch": 0.33, "learning_rate": 2.0173598553345387e-06, "loss": 0.0004, "step": 2767 }, { "epoch": 0.33, "learning_rate": 2.016998191681736e-06, "loss": 0.0491, "step": 2768 }, { "epoch": 0.33, "learning_rate": 2.016636528028933e-06, "loss": 0.0012, "step": 2769 }, { "epoch": 0.33, "learning_rate": 2.01627486437613e-06, "loss": 0.0006, "step": 2770 }, { "epoch": 0.33, "learning_rate": 2.0159132007233272e-06, "loss": 0.0004, "step": 2771 }, { "epoch": 0.33, "learning_rate": 2.0155515370705245e-06, "loss": 0.0006, "step": 2772 }, { "epoch": 0.33, "learning_rate": 2.0151898734177217e-06, "loss": 0.0006, "step": 2773 }, { "epoch": 0.33, "learning_rate": 2.0148282097649185e-06, "loss": 0.0006, "step": 2774 }, { "epoch": 0.33, "learning_rate": 2.0144665461121157e-06, "loss": 0.0005, "step": 2775 }, { "epoch": 0.33, "learning_rate": 2.014104882459313e-06, "loss": 1.558, "step": 2776 }, { "epoch": 0.33, "learning_rate": 2.01374321880651e-06, "loss": 0.0006, "step": 2777 }, { "epoch": 0.33, "learning_rate": 2.013381555153707e-06, "loss": 0.0003, "step": 2778 }, { "epoch": 0.33, "learning_rate": 2.0130198915009042e-06, "loss": 0.0015, "step": 2779 }, { "epoch": 0.33, "learning_rate": 2.0126582278481015e-06, "loss": 0.0009, "step": 2780 }, { "epoch": 0.33, "learning_rate": 2.0122965641952987e-06, "loss": 0.0004, "step": 2781 }, { "epoch": 0.33, "learning_rate": 2.0119349005424955e-06, "loss": 0.0004, "step": 2782 }, { "epoch": 0.33, "learning_rate": 2.0115732368896927e-06, "loss": 0.0073, "step": 2783 }, { "epoch": 0.33, "learning_rate": 2.01121157323689e-06, "loss": 0.0005, "step": 2784 }, { "epoch": 0.33, "learning_rate": 2.0108499095840868e-06, "loss": 0.0007, "step": 2785 }, { "epoch": 0.33, "learning_rate": 2.010488245931284e-06, "loss": 0.013, "step": 2786 }, { "epoch": 0.33, "learning_rate": 2.0101265822784812e-06, "loss": 0.8377, "step": 2787 }, { "epoch": 0.33, "learning_rate": 2.009764918625678e-06, "loss": 0.0136, "step": 2788 }, { "epoch": 0.33, "learning_rate": 2.0094032549728753e-06, "loss": 0.0004, "step": 2789 }, { "epoch": 0.33, "learning_rate": 2.009041591320072e-06, "loss": 0.0003, "step": 2790 }, { "epoch": 0.33, "learning_rate": 2.0086799276672693e-06, "loss": 0.0003, "step": 2791 }, { "epoch": 0.33, "learning_rate": 2.0083182640144666e-06, "loss": 0.0004, "step": 2792 }, { "epoch": 0.33, "learning_rate": 2.007956600361664e-06, "loss": 0.0317, "step": 2793 }, { "epoch": 0.33, "learning_rate": 2.0075949367088606e-06, "loss": 0.0005, "step": 2794 }, { "epoch": 0.33, "learning_rate": 2.007233273056058e-06, "loss": 0.0003, "step": 2795 }, { "epoch": 0.34, "learning_rate": 2.006871609403255e-06, "loss": 0.0005, "step": 2796 }, { "epoch": 0.34, "learning_rate": 2.0065099457504523e-06, "loss": 0.4022, "step": 2797 }, { "epoch": 0.34, "learning_rate": 2.006148282097649e-06, "loss": 0.0003, "step": 2798 }, { "epoch": 0.34, "learning_rate": 2.0057866184448463e-06, "loss": 0.0003, "step": 2799 }, { "epoch": 0.34, "learning_rate": 2.0054249547920436e-06, "loss": 0.0008, "step": 2800 }, { "epoch": 0.34, "learning_rate": 2.005063291139241e-06, "loss": 0.0005, "step": 2801 }, { "epoch": 0.34, "learning_rate": 2.0047016274864376e-06, "loss": 0.0003, "step": 2802 }, { "epoch": 0.34, "learning_rate": 2.004339963833635e-06, "loss": 0.0005, "step": 2803 }, { "epoch": 0.34, "learning_rate": 2.003978300180832e-06, "loss": 0.0003, "step": 2804 }, { "epoch": 0.34, "learning_rate": 2.003616636528029e-06, "loss": 0.0004, "step": 2805 }, { "epoch": 0.34, "learning_rate": 2.003254972875226e-06, "loss": 0.0005, "step": 2806 }, { "epoch": 0.34, "learning_rate": 2.0028933092224233e-06, "loss": 0.0003, "step": 2807 }, { "epoch": 0.34, "learning_rate": 2.0025316455696206e-06, "loss": 0.0003, "step": 2808 }, { "epoch": 0.34, "learning_rate": 2.0021699819168174e-06, "loss": 0.0004, "step": 2809 }, { "epoch": 0.34, "learning_rate": 2.0018083182640146e-06, "loss": 0.0004, "step": 2810 }, { "epoch": 0.34, "learning_rate": 2.001446654611212e-06, "loss": 0.0003, "step": 2811 }, { "epoch": 0.34, "learning_rate": 2.0010849909584086e-06, "loss": 0.6124, "step": 2812 }, { "epoch": 0.34, "learning_rate": 2.000723327305606e-06, "loss": 0.0004, "step": 2813 }, { "epoch": 0.34, "learning_rate": 2.0003616636528027e-06, "loss": 0.0151, "step": 2814 }, { "epoch": 0.34, "learning_rate": 2e-06, "loss": 0.0004, "step": 2815 }, { "epoch": 0.34, "learning_rate": 1.999638336347197e-06, "loss": 0.0007, "step": 2816 }, { "epoch": 0.34, "learning_rate": 1.9992766726943944e-06, "loss": 0.0005, "step": 2817 }, { "epoch": 0.34, "learning_rate": 1.998915009041591e-06, "loss": 0.1773, "step": 2818 }, { "epoch": 0.34, "learning_rate": 1.9985533453887884e-06, "loss": 0.0004, "step": 2819 }, { "epoch": 0.34, "learning_rate": 1.9981916817359856e-06, "loss": 0.0004, "step": 2820 }, { "epoch": 0.34, "learning_rate": 1.997830018083183e-06, "loss": 0.0038, "step": 2821 }, { "epoch": 0.34, "learning_rate": 1.9974683544303797e-06, "loss": 0.0004, "step": 2822 }, { "epoch": 0.34, "learning_rate": 1.997106690777577e-06, "loss": 0.0003, "step": 2823 }, { "epoch": 0.34, "learning_rate": 1.996745027124774e-06, "loss": 0.0004, "step": 2824 }, { "epoch": 0.34, "learning_rate": 1.996383363471971e-06, "loss": 0.0003, "step": 2825 }, { "epoch": 0.34, "learning_rate": 1.996021699819168e-06, "loss": 0.0004, "step": 2826 }, { "epoch": 0.34, "learning_rate": 1.9956600361663654e-06, "loss": 0.0005, "step": 2827 }, { "epoch": 0.34, "learning_rate": 1.9952983725135626e-06, "loss": 0.0004, "step": 2828 }, { "epoch": 0.34, "learning_rate": 1.9949367088607595e-06, "loss": 0.0003, "step": 2829 }, { "epoch": 0.34, "learning_rate": 1.9945750452079567e-06, "loss": 0.0005, "step": 2830 }, { "epoch": 0.34, "learning_rate": 1.994213381555154e-06, "loss": 0.0003, "step": 2831 }, { "epoch": 0.34, "learning_rate": 1.993851717902351e-06, "loss": 0.6588, "step": 2832 }, { "epoch": 0.34, "learning_rate": 1.993490054249548e-06, "loss": 0.0004, "step": 2833 }, { "epoch": 0.34, "learning_rate": 1.993128390596745e-06, "loss": 0.0003, "step": 2834 }, { "epoch": 0.34, "learning_rate": 1.992766726943942e-06, "loss": 0.0003, "step": 2835 }, { "epoch": 0.34, "learning_rate": 1.9924050632911392e-06, "loss": 0.0003, "step": 2836 }, { "epoch": 0.34, "learning_rate": 1.9920433996383365e-06, "loss": 0.0003, "step": 2837 }, { "epoch": 0.34, "learning_rate": 1.9916817359855333e-06, "loss": 0.001, "step": 2838 }, { "epoch": 0.34, "learning_rate": 1.9913200723327305e-06, "loss": 0.8943, "step": 2839 }, { "epoch": 0.34, "learning_rate": 1.9909584086799277e-06, "loss": 0.9332, "step": 2840 }, { "epoch": 0.34, "learning_rate": 1.9905967450271245e-06, "loss": 0.0006, "step": 2841 }, { "epoch": 0.34, "learning_rate": 1.9902350813743218e-06, "loss": 0.6075, "step": 2842 }, { "epoch": 0.34, "learning_rate": 1.989873417721519e-06, "loss": 0.0003, "step": 2843 }, { "epoch": 0.34, "learning_rate": 1.9895117540687162e-06, "loss": 0.0005, "step": 2844 }, { "epoch": 0.34, "learning_rate": 1.989150090415913e-06, "loss": 0.0004, "step": 2845 }, { "epoch": 0.34, "learning_rate": 1.9887884267631103e-06, "loss": 0.0005, "step": 2846 }, { "epoch": 0.34, "learning_rate": 1.9884267631103075e-06, "loss": 0.0004, "step": 2847 }, { "epoch": 0.34, "learning_rate": 1.9880650994575047e-06, "loss": 0.0004, "step": 2848 }, { "epoch": 0.34, "learning_rate": 1.9877034358047015e-06, "loss": 0.1424, "step": 2849 }, { "epoch": 0.34, "learning_rate": 1.9873417721518988e-06, "loss": 0.0008, "step": 2850 }, { "epoch": 0.34, "learning_rate": 1.986980108499096e-06, "loss": 0.0078, "step": 2851 }, { "epoch": 0.34, "learning_rate": 1.9866184448462932e-06, "loss": 0.0007, "step": 2852 }, { "epoch": 0.34, "learning_rate": 1.98625678119349e-06, "loss": 0.0004, "step": 2853 }, { "epoch": 0.34, "learning_rate": 1.9858951175406873e-06, "loss": 0.0237, "step": 2854 }, { "epoch": 0.34, "learning_rate": 1.9855334538878845e-06, "loss": 0.0005, "step": 2855 }, { "epoch": 0.34, "learning_rate": 1.9851717902350817e-06, "loss": 0.0836, "step": 2856 }, { "epoch": 0.34, "learning_rate": 1.9848101265822785e-06, "loss": 0.002, "step": 2857 }, { "epoch": 0.34, "learning_rate": 1.9844484629294758e-06, "loss": 0.0006, "step": 2858 }, { "epoch": 0.34, "learning_rate": 1.9840867992766726e-06, "loss": 0.0003, "step": 2859 }, { "epoch": 0.34, "learning_rate": 1.98372513562387e-06, "loss": 0.0058, "step": 2860 }, { "epoch": 0.34, "learning_rate": 1.9833634719710666e-06, "loss": 0.0008, "step": 2861 }, { "epoch": 0.34, "learning_rate": 1.983001808318264e-06, "loss": 0.0005, "step": 2862 }, { "epoch": 0.34, "learning_rate": 1.982640144665461e-06, "loss": 0.0017, "step": 2863 }, { "epoch": 0.34, "learning_rate": 1.9822784810126583e-06, "loss": 0.0004, "step": 2864 }, { "epoch": 0.34, "learning_rate": 1.981916817359855e-06, "loss": 0.9297, "step": 2865 }, { "epoch": 0.34, "learning_rate": 1.9815551537070523e-06, "loss": 0.0004, "step": 2866 }, { "epoch": 0.34, "learning_rate": 1.9811934900542496e-06, "loss": 0.0006, "step": 2867 }, { "epoch": 0.34, "learning_rate": 1.980831826401447e-06, "loss": 0.0003, "step": 2868 }, { "epoch": 0.34, "learning_rate": 1.9804701627486436e-06, "loss": 0.0097, "step": 2869 }, { "epoch": 0.34, "learning_rate": 1.980108499095841e-06, "loss": 0.0005, "step": 2870 }, { "epoch": 0.34, "learning_rate": 1.979746835443038e-06, "loss": 1.3488, "step": 2871 }, { "epoch": 0.34, "learning_rate": 1.9793851717902353e-06, "loss": 0.0004, "step": 2872 }, { "epoch": 0.34, "learning_rate": 1.979023508137432e-06, "loss": 0.0007, "step": 2873 }, { "epoch": 0.34, "learning_rate": 1.9786618444846293e-06, "loss": 0.0004, "step": 2874 }, { "epoch": 0.34, "learning_rate": 1.9783001808318266e-06, "loss": 0.8789, "step": 2875 }, { "epoch": 0.34, "learning_rate": 1.977938517179024e-06, "loss": 0.0007, "step": 2876 }, { "epoch": 0.34, "learning_rate": 1.9775768535262206e-06, "loss": 0.0004, "step": 2877 }, { "epoch": 0.34, "learning_rate": 1.977215189873418e-06, "loss": 0.0004, "step": 2878 }, { "epoch": 0.34, "learning_rate": 1.976853526220615e-06, "loss": 0.0014, "step": 2879 }, { "epoch": 0.35, "learning_rate": 1.9764918625678123e-06, "loss": 0.0006, "step": 2880 }, { "epoch": 0.35, "learning_rate": 1.976130198915009e-06, "loss": 0.0004, "step": 2881 }, { "epoch": 0.35, "learning_rate": 1.9757685352622064e-06, "loss": 0.0004, "step": 2882 }, { "epoch": 0.35, "learning_rate": 1.975406871609403e-06, "loss": 0.0061, "step": 2883 }, { "epoch": 0.35, "learning_rate": 1.9750452079566004e-06, "loss": 0.0005, "step": 2884 }, { "epoch": 0.35, "learning_rate": 1.974683544303797e-06, "loss": 0.0005, "step": 2885 }, { "epoch": 0.35, "learning_rate": 1.9743218806509944e-06, "loss": 0.9235, "step": 2886 }, { "epoch": 0.35, "learning_rate": 1.9739602169981917e-06, "loss": 0.0005, "step": 2887 }, { "epoch": 0.35, "learning_rate": 1.973598553345389e-06, "loss": 0.0003, "step": 2888 }, { "epoch": 0.35, "learning_rate": 1.9732368896925857e-06, "loss": 0.0014, "step": 2889 }, { "epoch": 0.35, "learning_rate": 1.972875226039783e-06, "loss": 0.0004, "step": 2890 }, { "epoch": 0.35, "learning_rate": 1.97251356238698e-06, "loss": 0.0007, "step": 2891 }, { "epoch": 0.35, "learning_rate": 1.9721518987341774e-06, "loss": 0.0006, "step": 2892 }, { "epoch": 0.35, "learning_rate": 1.971790235081374e-06, "loss": 0.0006, "step": 2893 }, { "epoch": 0.35, "learning_rate": 1.9714285714285714e-06, "loss": 0.597, "step": 2894 }, { "epoch": 0.35, "learning_rate": 1.9710669077757687e-06, "loss": 0.0006, "step": 2895 }, { "epoch": 0.35, "learning_rate": 1.970705244122966e-06, "loss": 0.4378, "step": 2896 }, { "epoch": 0.35, "learning_rate": 1.9703435804701627e-06, "loss": 1.0447, "step": 2897 }, { "epoch": 0.35, "learning_rate": 1.96998191681736e-06, "loss": 0.0008, "step": 2898 }, { "epoch": 0.35, "learning_rate": 1.969620253164557e-06, "loss": 0.0013, "step": 2899 }, { "epoch": 0.35, "learning_rate": 1.9692585895117544e-06, "loss": 0.0009, "step": 2900 }, { "epoch": 0.35, "learning_rate": 1.968896925858951e-06, "loss": 0.0005, "step": 2901 }, { "epoch": 0.35, "learning_rate": 1.9685352622061484e-06, "loss": 0.0006, "step": 2902 }, { "epoch": 0.35, "learning_rate": 1.9681735985533457e-06, "loss": 0.5016, "step": 2903 }, { "epoch": 0.35, "learning_rate": 1.9678119349005425e-06, "loss": 0.0006, "step": 2904 }, { "epoch": 0.35, "learning_rate": 1.9674502712477397e-06, "loss": 1.4274, "step": 2905 }, { "epoch": 0.35, "learning_rate": 1.967088607594937e-06, "loss": 0.1104, "step": 2906 }, { "epoch": 0.35, "learning_rate": 1.9667269439421337e-06, "loss": 0.0004, "step": 2907 }, { "epoch": 0.35, "learning_rate": 1.966365280289331e-06, "loss": 1.8272, "step": 2908 }, { "epoch": 0.35, "learning_rate": 1.9660036166365278e-06, "loss": 0.0006, "step": 2909 }, { "epoch": 0.35, "learning_rate": 1.965641952983725e-06, "loss": 0.0006, "step": 2910 }, { "epoch": 0.35, "learning_rate": 1.9652802893309222e-06, "loss": 0.0015, "step": 2911 }, { "epoch": 0.35, "learning_rate": 1.9649186256781195e-06, "loss": 0.0005, "step": 2912 }, { "epoch": 0.35, "learning_rate": 1.9645569620253163e-06, "loss": 0.001, "step": 2913 }, { "epoch": 0.35, "learning_rate": 1.9641952983725135e-06, "loss": 0.0009, "step": 2914 }, { "epoch": 0.35, "learning_rate": 1.9638336347197107e-06, "loss": 0.0004, "step": 2915 }, { "epoch": 0.35, "learning_rate": 1.963471971066908e-06, "loss": 0.0009, "step": 2916 }, { "epoch": 0.35, "learning_rate": 1.9631103074141048e-06, "loss": 0.0009, "step": 2917 }, { "epoch": 0.35, "learning_rate": 1.962748643761302e-06, "loss": 0.0006, "step": 2918 }, { "epoch": 0.35, "learning_rate": 1.9623869801084992e-06, "loss": 0.1883, "step": 2919 }, { "epoch": 0.35, "learning_rate": 1.9620253164556965e-06, "loss": 0.0006, "step": 2920 }, { "epoch": 0.35, "learning_rate": 1.9616636528028933e-06, "loss": 0.0007, "step": 2921 }, { "epoch": 0.35, "learning_rate": 1.9613019891500905e-06, "loss": 0.0009, "step": 2922 }, { "epoch": 0.35, "learning_rate": 1.9609403254972877e-06, "loss": 0.774, "step": 2923 }, { "epoch": 0.35, "learning_rate": 1.9605786618444846e-06, "loss": 0.0006, "step": 2924 }, { "epoch": 0.35, "learning_rate": 1.9602169981916818e-06, "loss": 0.0014, "step": 2925 }, { "epoch": 0.35, "learning_rate": 1.959855334538879e-06, "loss": 0.001, "step": 2926 }, { "epoch": 0.35, "learning_rate": 1.9594936708860762e-06, "loss": 0.0018, "step": 2927 }, { "epoch": 0.35, "learning_rate": 1.959132007233273e-06, "loss": 0.0042, "step": 2928 }, { "epoch": 0.35, "learning_rate": 1.9587703435804703e-06, "loss": 0.0246, "step": 2929 }, { "epoch": 0.35, "learning_rate": 1.9584086799276675e-06, "loss": 0.0007, "step": 2930 }, { "epoch": 0.35, "learning_rate": 1.9580470162748643e-06, "loss": 0.2712, "step": 2931 }, { "epoch": 0.35, "learning_rate": 1.9576853526220616e-06, "loss": 1.3844, "step": 2932 }, { "epoch": 0.35, "learning_rate": 1.9573236889692584e-06, "loss": 0.0009, "step": 2933 }, { "epoch": 0.35, "learning_rate": 1.9569620253164556e-06, "loss": 0.0008, "step": 2934 }, { "epoch": 0.35, "learning_rate": 1.956600361663653e-06, "loss": 0.8735, "step": 2935 }, { "epoch": 0.35, "learning_rate": 1.95623869801085e-06, "loss": 0.0008, "step": 2936 }, { "epoch": 0.35, "learning_rate": 1.955877034358047e-06, "loss": 0.0005, "step": 2937 }, { "epoch": 0.35, "learning_rate": 1.955515370705244e-06, "loss": 0.0006, "step": 2938 }, { "epoch": 0.35, "learning_rate": 1.9551537070524413e-06, "loss": 0.0006, "step": 2939 }, { "epoch": 0.35, "learning_rate": 1.9547920433996386e-06, "loss": 0.0682, "step": 2940 }, { "epoch": 0.35, "learning_rate": 1.9544303797468354e-06, "loss": 0.001, "step": 2941 }, { "epoch": 0.35, "learning_rate": 1.9540687160940326e-06, "loss": 0.0006, "step": 2942 }, { "epoch": 0.35, "learning_rate": 1.95370705244123e-06, "loss": 0.0006, "step": 2943 }, { "epoch": 0.35, "learning_rate": 1.9533453887884266e-06, "loss": 0.0005, "step": 2944 }, { "epoch": 0.35, "learning_rate": 1.952983725135624e-06, "loss": 0.6496, "step": 2945 }, { "epoch": 0.35, "learning_rate": 1.952622061482821e-06, "loss": 0.3408, "step": 2946 }, { "epoch": 0.35, "learning_rate": 1.9522603978300183e-06, "loss": 0.0024, "step": 2947 }, { "epoch": 0.35, "learning_rate": 1.951898734177215e-06, "loss": 0.001, "step": 2948 }, { "epoch": 0.35, "learning_rate": 1.9515370705244124e-06, "loss": 0.0048, "step": 2949 }, { "epoch": 0.35, "learning_rate": 1.9511754068716096e-06, "loss": 0.4811, "step": 2950 }, { "epoch": 0.35, "learning_rate": 1.950813743218807e-06, "loss": 0.0015, "step": 2951 }, { "epoch": 0.35, "learning_rate": 1.9504520795660036e-06, "loss": 0.0007, "step": 2952 }, { "epoch": 0.35, "learning_rate": 1.950090415913201e-06, "loss": 0.0009, "step": 2953 }, { "epoch": 0.35, "learning_rate": 1.949728752260398e-06, "loss": 0.0006, "step": 2954 }, { "epoch": 0.35, "learning_rate": 1.949367088607595e-06, "loss": 0.0006, "step": 2955 }, { "epoch": 0.35, "learning_rate": 1.949005424954792e-06, "loss": 0.9021, "step": 2956 }, { "epoch": 0.35, "learning_rate": 1.948643761301989e-06, "loss": 0.3892, "step": 2957 }, { "epoch": 0.35, "learning_rate": 1.948282097649186e-06, "loss": 0.0005, "step": 2958 }, { "epoch": 0.35, "learning_rate": 1.9479204339963834e-06, "loss": 0.0005, "step": 2959 }, { "epoch": 0.35, "learning_rate": 1.9475587703435806e-06, "loss": 0.9273, "step": 2960 }, { "epoch": 0.35, "learning_rate": 1.9471971066907774e-06, "loss": 0.0005, "step": 2961 }, { "epoch": 0.35, "learning_rate": 1.9468354430379747e-06, "loss": 0.0007, "step": 2962 }, { "epoch": 0.36, "learning_rate": 1.946473779385172e-06, "loss": 0.0014, "step": 2963 }, { "epoch": 0.36, "learning_rate": 1.9461121157323687e-06, "loss": 0.0011, "step": 2964 }, { "epoch": 0.36, "learning_rate": 1.945750452079566e-06, "loss": 0.0008, "step": 2965 }, { "epoch": 0.36, "learning_rate": 1.945388788426763e-06, "loss": 0.0004, "step": 2966 }, { "epoch": 0.36, "learning_rate": 1.9450271247739604e-06, "loss": 0.0006, "step": 2967 }, { "epoch": 0.36, "learning_rate": 1.9446654611211572e-06, "loss": 0.0005, "step": 2968 }, { "epoch": 0.36, "learning_rate": 1.9443037974683544e-06, "loss": 0.0003, "step": 2969 }, { "epoch": 0.36, "learning_rate": 1.9439421338155517e-06, "loss": 0.0008, "step": 2970 }, { "epoch": 0.36, "learning_rate": 1.943580470162749e-06, "loss": 0.0012, "step": 2971 }, { "epoch": 0.36, "learning_rate": 1.9432188065099457e-06, "loss": 0.0005, "step": 2972 }, { "epoch": 0.36, "learning_rate": 1.942857142857143e-06, "loss": 0.0006, "step": 2973 }, { "epoch": 0.36, "learning_rate": 1.94249547920434e-06, "loss": 0.1963, "step": 2974 }, { "epoch": 0.36, "learning_rate": 1.9421338155515374e-06, "loss": 0.0012, "step": 2975 }, { "epoch": 0.36, "learning_rate": 1.9417721518987342e-06, "loss": 0.0005, "step": 2976 }, { "epoch": 0.36, "learning_rate": 1.9414104882459315e-06, "loss": 0.0009, "step": 2977 }, { "epoch": 0.36, "learning_rate": 1.9410488245931287e-06, "loss": 0.0012, "step": 2978 }, { "epoch": 0.36, "learning_rate": 1.9406871609403255e-06, "loss": 0.0006, "step": 2979 }, { "epoch": 0.36, "learning_rate": 1.9403254972875223e-06, "loss": 0.0007, "step": 2980 }, { "epoch": 0.36, "learning_rate": 1.9399638336347195e-06, "loss": 0.0006, "step": 2981 }, { "epoch": 0.36, "learning_rate": 1.9396021699819168e-06, "loss": 0.0009, "step": 2982 }, { "epoch": 0.36, "learning_rate": 1.939240506329114e-06, "loss": 0.0004, "step": 2983 }, { "epoch": 0.36, "learning_rate": 1.938878842676311e-06, "loss": 0.0008, "step": 2984 }, { "epoch": 0.36, "learning_rate": 1.938517179023508e-06, "loss": 0.0005, "step": 2985 }, { "epoch": 0.36, "learning_rate": 1.9381555153707053e-06, "loss": 0.8107, "step": 2986 }, { "epoch": 0.36, "learning_rate": 1.9377938517179025e-06, "loss": 0.0006, "step": 2987 }, { "epoch": 0.36, "learning_rate": 1.9374321880650993e-06, "loss": 0.0008, "step": 2988 }, { "epoch": 0.36, "learning_rate": 1.9370705244122965e-06, "loss": 0.0007, "step": 2989 }, { "epoch": 0.36, "learning_rate": 1.9367088607594938e-06, "loss": 0.0006, "step": 2990 }, { "epoch": 0.36, "learning_rate": 1.936347197106691e-06, "loss": 0.0014, "step": 2991 }, { "epoch": 0.36, "learning_rate": 1.935985533453888e-06, "loss": 0.0023, "step": 2992 }, { "epoch": 0.36, "learning_rate": 1.935623869801085e-06, "loss": 0.0007, "step": 2993 }, { "epoch": 0.36, "learning_rate": 1.9352622061482823e-06, "loss": 0.0009, "step": 2994 }, { "epoch": 0.36, "learning_rate": 1.9349005424954795e-06, "loss": 0.797, "step": 2995 }, { "epoch": 0.36, "learning_rate": 1.9345388788426763e-06, "loss": 0.0007, "step": 2996 }, { "epoch": 0.36, "learning_rate": 1.9341772151898735e-06, "loss": 0.8052, "step": 2997 }, { "epoch": 0.36, "learning_rate": 1.9338155515370708e-06, "loss": 0.0128, "step": 2998 }, { "epoch": 0.36, "learning_rate": 1.933453887884268e-06, "loss": 0.0005, "step": 2999 }, { "epoch": 0.36, "learning_rate": 1.933092224231465e-06, "loss": 0.0005, "step": 3000 }, { "epoch": 0.36, "learning_rate": 1.932730560578662e-06, "loss": 0.001, "step": 3001 }, { "epoch": 0.36, "learning_rate": 1.9323688969258593e-06, "loss": 0.0012, "step": 3002 }, { "epoch": 0.36, "learning_rate": 1.932007233273056e-06, "loss": 0.0006, "step": 3003 }, { "epoch": 0.36, "learning_rate": 1.931645569620253e-06, "loss": 0.0382, "step": 3004 }, { "epoch": 0.36, "learning_rate": 1.93128390596745e-06, "loss": 0.6709, "step": 3005 }, { "epoch": 0.36, "learning_rate": 1.9309222423146473e-06, "loss": 0.0013, "step": 3006 }, { "epoch": 0.36, "learning_rate": 1.9305605786618446e-06, "loss": 0.0006, "step": 3007 }, { "epoch": 0.36, "learning_rate": 1.9301989150090414e-06, "loss": 0.0004, "step": 3008 }, { "epoch": 0.36, "learning_rate": 1.9298372513562386e-06, "loss": 0.7363, "step": 3009 }, { "epoch": 0.36, "learning_rate": 1.929475587703436e-06, "loss": 0.0007, "step": 3010 }, { "epoch": 0.36, "learning_rate": 1.929113924050633e-06, "loss": 0.0005, "step": 3011 }, { "epoch": 0.36, "learning_rate": 1.92875226039783e-06, "loss": 0.0005, "step": 3012 }, { "epoch": 0.36, "learning_rate": 1.928390596745027e-06, "loss": 0.0004, "step": 3013 }, { "epoch": 0.36, "learning_rate": 1.9280289330922243e-06, "loss": 0.0005, "step": 3014 }, { "epoch": 0.36, "learning_rate": 1.9276672694394216e-06, "loss": 0.001, "step": 3015 }, { "epoch": 0.36, "learning_rate": 1.9273056057866184e-06, "loss": 0.0004, "step": 3016 }, { "epoch": 0.36, "learning_rate": 1.9269439421338156e-06, "loss": 0.0005, "step": 3017 }, { "epoch": 0.36, "learning_rate": 1.926582278481013e-06, "loss": 0.0005, "step": 3018 }, { "epoch": 0.36, "learning_rate": 1.92622061482821e-06, "loss": 0.0006, "step": 3019 }, { "epoch": 0.36, "learning_rate": 1.925858951175407e-06, "loss": 0.0005, "step": 3020 }, { "epoch": 0.36, "learning_rate": 1.925497287522604e-06, "loss": 0.0005, "step": 3021 }, { "epoch": 0.36, "learning_rate": 1.9251356238698013e-06, "loss": 0.0005, "step": 3022 }, { "epoch": 0.36, "learning_rate": 1.924773960216998e-06, "loss": 0.0007, "step": 3023 }, { "epoch": 0.36, "learning_rate": 1.9244122965641954e-06, "loss": 0.9697, "step": 3024 }, { "epoch": 0.36, "learning_rate": 1.9240506329113926e-06, "loss": 0.0191, "step": 3025 }, { "epoch": 0.36, "learning_rate": 1.92368896925859e-06, "loss": 0.1499, "step": 3026 }, { "epoch": 0.36, "learning_rate": 1.9233273056057867e-06, "loss": 0.0005, "step": 3027 }, { "epoch": 0.36, "learning_rate": 1.9229656419529835e-06, "loss": 0.0006, "step": 3028 }, { "epoch": 0.36, "learning_rate": 1.9226039783001807e-06, "loss": 0.0042, "step": 3029 }, { "epoch": 0.36, "learning_rate": 1.922242314647378e-06, "loss": 0.0441, "step": 3030 }, { "epoch": 0.36, "learning_rate": 1.921880650994575e-06, "loss": 0.0003, "step": 3031 }, { "epoch": 0.36, "learning_rate": 1.921518987341772e-06, "loss": 0.0006, "step": 3032 }, { "epoch": 0.36, "learning_rate": 1.921157323688969e-06, "loss": 0.0006, "step": 3033 }, { "epoch": 0.36, "learning_rate": 1.9207956600361664e-06, "loss": 0.0945, "step": 3034 }, { "epoch": 0.36, "learning_rate": 1.9204339963833637e-06, "loss": 0.0004, "step": 3035 }, { "epoch": 0.36, "learning_rate": 1.9200723327305605e-06, "loss": 0.0004, "step": 3036 }, { "epoch": 0.36, "learning_rate": 1.9197106690777577e-06, "loss": 0.0006, "step": 3037 }, { "epoch": 0.36, "learning_rate": 1.919349005424955e-06, "loss": 0.001, "step": 3038 }, { "epoch": 0.36, "learning_rate": 1.918987341772152e-06, "loss": 0.0007, "step": 3039 }, { "epoch": 0.36, "learning_rate": 1.918625678119349e-06, "loss": 0.0017, "step": 3040 }, { "epoch": 0.36, "learning_rate": 1.918264014466546e-06, "loss": 0.2017, "step": 3041 }, { "epoch": 0.36, "learning_rate": 1.9179023508137434e-06, "loss": 0.0006, "step": 3042 }, { "epoch": 0.36, "learning_rate": 1.9175406871609402e-06, "loss": 0.0005, "step": 3043 }, { "epoch": 0.36, "learning_rate": 1.9171790235081375e-06, "loss": 0.5947, "step": 3044 }, { "epoch": 0.36, "learning_rate": 1.9168173598553347e-06, "loss": 0.0005, "step": 3045 }, { "epoch": 0.37, "learning_rate": 1.916455696202532e-06, "loss": 0.0031, "step": 3046 }, { "epoch": 0.37, "learning_rate": 1.9160940325497287e-06, "loss": 0.0006, "step": 3047 }, { "epoch": 0.37, "learning_rate": 1.915732368896926e-06, "loss": 0.4689, "step": 3048 }, { "epoch": 0.37, "learning_rate": 1.915370705244123e-06, "loss": 0.0007, "step": 3049 }, { "epoch": 0.37, "learning_rate": 1.9150090415913204e-06, "loss": 0.0004, "step": 3050 }, { "epoch": 0.37, "learning_rate": 1.9146473779385172e-06, "loss": 0.0006, "step": 3051 }, { "epoch": 0.37, "learning_rate": 1.914285714285714e-06, "loss": 0.2435, "step": 3052 }, { "epoch": 0.37, "learning_rate": 1.9139240506329113e-06, "loss": 0.0004, "step": 3053 }, { "epoch": 0.37, "learning_rate": 1.9135623869801085e-06, "loss": 0.885, "step": 3054 }, { "epoch": 0.37, "learning_rate": 1.9132007233273057e-06, "loss": 0.0005, "step": 3055 }, { "epoch": 0.37, "learning_rate": 1.9128390596745025e-06, "loss": 0.0004, "step": 3056 }, { "epoch": 0.37, "learning_rate": 1.9124773960216998e-06, "loss": 0.0006, "step": 3057 }, { "epoch": 0.37, "learning_rate": 1.912115732368897e-06, "loss": 0.0005, "step": 3058 }, { "epoch": 0.37, "learning_rate": 1.9117540687160942e-06, "loss": 0.0007, "step": 3059 }, { "epoch": 0.37, "learning_rate": 1.911392405063291e-06, "loss": 0.0004, "step": 3060 }, { "epoch": 0.37, "learning_rate": 1.9110307414104883e-06, "loss": 0.0004, "step": 3061 }, { "epoch": 0.37, "learning_rate": 1.9106690777576855e-06, "loss": 0.0008, "step": 3062 }, { "epoch": 0.37, "learning_rate": 1.9103074141048823e-06, "loss": 0.666, "step": 3063 }, { "epoch": 0.37, "learning_rate": 1.9099457504520796e-06, "loss": 0.0004, "step": 3064 }, { "epoch": 0.37, "learning_rate": 1.9095840867992768e-06, "loss": 0.0007, "step": 3065 }, { "epoch": 0.37, "learning_rate": 1.909222423146474e-06, "loss": 0.001, "step": 3066 }, { "epoch": 0.37, "learning_rate": 1.908860759493671e-06, "loss": 0.0003, "step": 3067 }, { "epoch": 0.37, "learning_rate": 1.908499095840868e-06, "loss": 0.0005, "step": 3068 }, { "epoch": 0.37, "learning_rate": 1.9081374321880653e-06, "loss": 0.0008, "step": 3069 }, { "epoch": 0.37, "learning_rate": 1.9077757685352625e-06, "loss": 0.0007, "step": 3070 }, { "epoch": 0.37, "learning_rate": 1.9074141048824593e-06, "loss": 0.0016, "step": 3071 }, { "epoch": 0.37, "learning_rate": 1.9070524412296566e-06, "loss": 0.7406, "step": 3072 }, { "epoch": 0.37, "learning_rate": 1.9066907775768538e-06, "loss": 0.0008, "step": 3073 }, { "epoch": 0.37, "learning_rate": 1.9063291139240506e-06, "loss": 0.0004, "step": 3074 }, { "epoch": 0.37, "learning_rate": 1.9059674502712476e-06, "loss": 0.0006, "step": 3075 }, { "epoch": 0.37, "learning_rate": 1.9056057866184448e-06, "loss": 0.0003, "step": 3076 }, { "epoch": 0.37, "learning_rate": 1.9052441229656419e-06, "loss": 0.0025, "step": 3077 }, { "epoch": 0.37, "learning_rate": 1.904882459312839e-06, "loss": 0.0005, "step": 3078 }, { "epoch": 0.37, "learning_rate": 1.9045207956600361e-06, "loss": 0.0004, "step": 3079 }, { "epoch": 0.37, "learning_rate": 1.9041591320072333e-06, "loss": 0.0006, "step": 3080 }, { "epoch": 0.37, "learning_rate": 1.9037974683544304e-06, "loss": 0.0004, "step": 3081 }, { "epoch": 0.37, "learning_rate": 1.9034358047016274e-06, "loss": 0.0261, "step": 3082 }, { "epoch": 0.37, "learning_rate": 1.9030741410488246e-06, "loss": 0.0011, "step": 3083 }, { "epoch": 0.37, "learning_rate": 1.9027124773960216e-06, "loss": 0.3903, "step": 3084 }, { "epoch": 0.37, "learning_rate": 1.9023508137432189e-06, "loss": 0.1738, "step": 3085 }, { "epoch": 0.37, "learning_rate": 1.9019891500904159e-06, "loss": 0.006, "step": 3086 }, { "epoch": 0.37, "learning_rate": 1.9016274864376131e-06, "loss": 1.7165, "step": 3087 }, { "epoch": 0.37, "learning_rate": 1.9012658227848101e-06, "loss": 0.0005, "step": 3088 }, { "epoch": 0.37, "learning_rate": 1.9009041591320074e-06, "loss": 0.0007, "step": 3089 }, { "epoch": 0.37, "learning_rate": 1.9005424954792044e-06, "loss": 0.0006, "step": 3090 }, { "epoch": 0.37, "learning_rate": 1.9001808318264016e-06, "loss": 0.0004, "step": 3091 }, { "epoch": 0.37, "learning_rate": 1.8998191681735986e-06, "loss": 0.0004, "step": 3092 }, { "epoch": 0.37, "learning_rate": 1.8994575045207959e-06, "loss": 0.0005, "step": 3093 }, { "epoch": 0.37, "learning_rate": 1.8990958408679929e-06, "loss": 0.0004, "step": 3094 }, { "epoch": 0.37, "learning_rate": 1.8987341772151901e-06, "loss": 0.0005, "step": 3095 }, { "epoch": 0.37, "learning_rate": 1.8983725135623871e-06, "loss": 0.0005, "step": 3096 }, { "epoch": 0.37, "learning_rate": 1.8980108499095844e-06, "loss": 0.0004, "step": 3097 }, { "epoch": 0.37, "learning_rate": 1.8976491862567812e-06, "loss": 0.0005, "step": 3098 }, { "epoch": 0.37, "learning_rate": 1.8972875226039782e-06, "loss": 0.8599, "step": 3099 }, { "epoch": 0.37, "learning_rate": 1.8969258589511752e-06, "loss": 0.0004, "step": 3100 }, { "epoch": 0.37, "learning_rate": 1.8965641952983724e-06, "loss": 0.0047, "step": 3101 }, { "epoch": 0.37, "learning_rate": 1.8962025316455695e-06, "loss": 0.0008, "step": 3102 }, { "epoch": 0.37, "learning_rate": 1.8958408679927667e-06, "loss": 0.2625, "step": 3103 }, { "epoch": 0.37, "learning_rate": 1.8954792043399637e-06, "loss": 0.0006, "step": 3104 }, { "epoch": 0.37, "learning_rate": 1.895117540687161e-06, "loss": 0.0004, "step": 3105 }, { "epoch": 0.37, "learning_rate": 1.894755877034358e-06, "loss": 0.0004, "step": 3106 }, { "epoch": 0.37, "learning_rate": 1.8943942133815552e-06, "loss": 0.8792, "step": 3107 }, { "epoch": 0.37, "learning_rate": 1.8940325497287522e-06, "loss": 0.0004, "step": 3108 }, { "epoch": 0.37, "learning_rate": 1.8936708860759494e-06, "loss": 0.0005, "step": 3109 }, { "epoch": 0.37, "learning_rate": 1.8933092224231465e-06, "loss": 0.0072, "step": 3110 }, { "epoch": 0.37, "learning_rate": 1.8929475587703437e-06, "loss": 0.0005, "step": 3111 }, { "epoch": 0.37, "learning_rate": 1.8925858951175407e-06, "loss": 0.0005, "step": 3112 }, { "epoch": 0.37, "learning_rate": 1.892224231464738e-06, "loss": 0.0006, "step": 3113 }, { "epoch": 0.37, "learning_rate": 1.891862567811935e-06, "loss": 0.001, "step": 3114 }, { "epoch": 0.37, "learning_rate": 1.8915009041591322e-06, "loss": 0.0029, "step": 3115 }, { "epoch": 0.37, "learning_rate": 1.8911392405063292e-06, "loss": 0.0004, "step": 3116 }, { "epoch": 0.37, "learning_rate": 1.8907775768535264e-06, "loss": 0.0509, "step": 3117 }, { "epoch": 0.37, "learning_rate": 1.8904159132007235e-06, "loss": 0.2387, "step": 3118 }, { "epoch": 0.37, "learning_rate": 1.8900542495479207e-06, "loss": 0.0004, "step": 3119 }, { "epoch": 0.37, "learning_rate": 1.8896925858951177e-06, "loss": 0.0008, "step": 3120 }, { "epoch": 0.37, "learning_rate": 1.889330922242315e-06, "loss": 0.6576, "step": 3121 }, { "epoch": 0.37, "learning_rate": 1.8889692585895115e-06, "loss": 0.0006, "step": 3122 }, { "epoch": 0.37, "learning_rate": 1.8886075949367088e-06, "loss": 0.0007, "step": 3123 }, { "epoch": 0.37, "learning_rate": 1.8882459312839058e-06, "loss": 0.0003, "step": 3124 }, { "epoch": 0.37, "learning_rate": 1.887884267631103e-06, "loss": 0.0004, "step": 3125 }, { "epoch": 0.37, "learning_rate": 1.8875226039783e-06, "loss": 0.0004, "step": 3126 }, { "epoch": 0.37, "learning_rate": 1.8871609403254973e-06, "loss": 0.0006, "step": 3127 }, { "epoch": 0.37, "learning_rate": 1.8867992766726943e-06, "loss": 0.0004, "step": 3128 }, { "epoch": 0.37, "learning_rate": 1.8864376130198915e-06, "loss": 0.0005, "step": 3129 }, { "epoch": 0.38, "learning_rate": 1.8860759493670885e-06, "loss": 0.0004, "step": 3130 }, { "epoch": 0.38, "learning_rate": 1.8857142857142858e-06, "loss": 0.0007, "step": 3131 }, { "epoch": 0.38, "learning_rate": 1.8853526220614828e-06, "loss": 0.0005, "step": 3132 }, { "epoch": 0.38, "learning_rate": 1.88499095840868e-06, "loss": 0.0005, "step": 3133 }, { "epoch": 0.38, "learning_rate": 1.884629294755877e-06, "loss": 0.0004, "step": 3134 }, { "epoch": 0.38, "learning_rate": 1.8842676311030743e-06, "loss": 0.0004, "step": 3135 }, { "epoch": 0.38, "learning_rate": 1.8839059674502713e-06, "loss": 0.0004, "step": 3136 }, { "epoch": 0.38, "learning_rate": 1.8835443037974685e-06, "loss": 0.0006, "step": 3137 }, { "epoch": 0.38, "learning_rate": 1.8831826401446656e-06, "loss": 0.0003, "step": 3138 }, { "epoch": 0.38, "learning_rate": 1.8828209764918628e-06, "loss": 0.0005, "step": 3139 }, { "epoch": 0.38, "learning_rate": 1.8824593128390598e-06, "loss": 0.0003, "step": 3140 }, { "epoch": 0.38, "learning_rate": 1.882097649186257e-06, "loss": 0.0003, "step": 3141 }, { "epoch": 0.38, "learning_rate": 1.881735985533454e-06, "loss": 0.0005, "step": 3142 }, { "epoch": 0.38, "learning_rate": 1.881374321880651e-06, "loss": 0.0005, "step": 3143 }, { "epoch": 0.38, "learning_rate": 1.8810126582278483e-06, "loss": 0.0012, "step": 3144 }, { "epoch": 0.38, "learning_rate": 1.8806509945750453e-06, "loss": 0.0004, "step": 3145 }, { "epoch": 0.38, "learning_rate": 1.8802893309222421e-06, "loss": 0.8212, "step": 3146 }, { "epoch": 0.38, "learning_rate": 1.8799276672694394e-06, "loss": 0.0004, "step": 3147 }, { "epoch": 0.38, "learning_rate": 1.8795660036166364e-06, "loss": 0.0004, "step": 3148 }, { "epoch": 0.38, "learning_rate": 1.8792043399638336e-06, "loss": 0.0004, "step": 3149 }, { "epoch": 0.38, "learning_rate": 1.8788426763110306e-06, "loss": 0.9131, "step": 3150 }, { "epoch": 0.38, "learning_rate": 1.8784810126582279e-06, "loss": 0.0005, "step": 3151 }, { "epoch": 0.38, "learning_rate": 1.8781193490054249e-06, "loss": 0.0003, "step": 3152 }, { "epoch": 0.38, "learning_rate": 1.8777576853526221e-06, "loss": 0.0003, "step": 3153 }, { "epoch": 0.38, "learning_rate": 1.8773960216998191e-06, "loss": 0.0004, "step": 3154 }, { "epoch": 0.38, "learning_rate": 1.8770343580470164e-06, "loss": 0.0004, "step": 3155 }, { "epoch": 0.38, "learning_rate": 1.8766726943942134e-06, "loss": 0.0005, "step": 3156 }, { "epoch": 0.38, "learning_rate": 1.8763110307414106e-06, "loss": 0.0004, "step": 3157 }, { "epoch": 0.38, "learning_rate": 1.8759493670886076e-06, "loss": 0.8385, "step": 3158 }, { "epoch": 0.38, "learning_rate": 1.8755877034358049e-06, "loss": 0.0004, "step": 3159 }, { "epoch": 0.38, "learning_rate": 1.8752260397830019e-06, "loss": 0.1289, "step": 3160 }, { "epoch": 0.38, "learning_rate": 1.8748643761301991e-06, "loss": 0.0006, "step": 3161 }, { "epoch": 0.38, "learning_rate": 1.8745027124773961e-06, "loss": 0.0004, "step": 3162 }, { "epoch": 0.38, "learning_rate": 1.8741410488245932e-06, "loss": 0.0005, "step": 3163 }, { "epoch": 0.38, "learning_rate": 1.8737793851717904e-06, "loss": 0.0009, "step": 3164 }, { "epoch": 0.38, "learning_rate": 1.8734177215189874e-06, "loss": 0.0013, "step": 3165 }, { "epoch": 0.38, "learning_rate": 1.8730560578661846e-06, "loss": 0.0004, "step": 3166 }, { "epoch": 0.38, "learning_rate": 1.8726943942133817e-06, "loss": 0.1482, "step": 3167 }, { "epoch": 0.38, "learning_rate": 1.8723327305605789e-06, "loss": 0.0003, "step": 3168 }, { "epoch": 0.38, "learning_rate": 1.871971066907776e-06, "loss": 0.0003, "step": 3169 }, { "epoch": 0.38, "learning_rate": 1.8716094032549727e-06, "loss": 0.0006, "step": 3170 }, { "epoch": 0.38, "learning_rate": 1.87124773960217e-06, "loss": 0.0003, "step": 3171 }, { "epoch": 0.38, "learning_rate": 1.870886075949367e-06, "loss": 0.011, "step": 3172 }, { "epoch": 0.38, "learning_rate": 1.8705244122965642e-06, "loss": 0.0004, "step": 3173 }, { "epoch": 0.38, "learning_rate": 1.8701627486437612e-06, "loss": 0.0003, "step": 3174 }, { "epoch": 0.38, "learning_rate": 1.8698010849909584e-06, "loss": 0.0004, "step": 3175 }, { "epoch": 0.38, "learning_rate": 1.8694394213381555e-06, "loss": 0.0003, "step": 3176 }, { "epoch": 0.38, "learning_rate": 1.8690777576853527e-06, "loss": 0.0005, "step": 3177 }, { "epoch": 0.38, "learning_rate": 1.8687160940325497e-06, "loss": 0.0003, "step": 3178 }, { "epoch": 0.38, "learning_rate": 1.868354430379747e-06, "loss": 0.0171, "step": 3179 }, { "epoch": 0.38, "learning_rate": 1.867992766726944e-06, "loss": 0.0022, "step": 3180 }, { "epoch": 0.38, "learning_rate": 1.867631103074141e-06, "loss": 0.0009, "step": 3181 }, { "epoch": 0.38, "learning_rate": 1.8672694394213382e-06, "loss": 0.6379, "step": 3182 }, { "epoch": 0.38, "learning_rate": 1.8669077757685352e-06, "loss": 0.0003, "step": 3183 }, { "epoch": 0.38, "learning_rate": 1.8665461121157325e-06, "loss": 0.0005, "step": 3184 }, { "epoch": 0.38, "learning_rate": 1.8661844484629295e-06, "loss": 0.0002, "step": 3185 }, { "epoch": 0.38, "learning_rate": 1.8658227848101267e-06, "loss": 0.806, "step": 3186 }, { "epoch": 0.38, "learning_rate": 1.8654611211573237e-06, "loss": 0.0017, "step": 3187 }, { "epoch": 0.38, "learning_rate": 1.865099457504521e-06, "loss": 0.0004, "step": 3188 }, { "epoch": 0.38, "learning_rate": 1.864737793851718e-06, "loss": 0.0003, "step": 3189 }, { "epoch": 0.38, "learning_rate": 1.8643761301989152e-06, "loss": 0.0005, "step": 3190 }, { "epoch": 0.38, "learning_rate": 1.8640144665461122e-06, "loss": 0.0005, "step": 3191 }, { "epoch": 0.38, "learning_rate": 1.8636528028933095e-06, "loss": 0.0005, "step": 3192 }, { "epoch": 0.38, "learning_rate": 1.8632911392405065e-06, "loss": 0.0003, "step": 3193 }, { "epoch": 0.38, "learning_rate": 1.8629294755877033e-06, "loss": 0.0003, "step": 3194 }, { "epoch": 0.38, "learning_rate": 1.8625678119349005e-06, "loss": 0.0004, "step": 3195 }, { "epoch": 0.38, "learning_rate": 1.8622061482820975e-06, "loss": 0.0005, "step": 3196 }, { "epoch": 0.38, "learning_rate": 1.8618444846292948e-06, "loss": 0.0004, "step": 3197 }, { "epoch": 0.38, "learning_rate": 1.8614828209764918e-06, "loss": 0.0014, "step": 3198 }, { "epoch": 0.38, "learning_rate": 1.861121157323689e-06, "loss": 0.8839, "step": 3199 }, { "epoch": 0.38, "learning_rate": 1.860759493670886e-06, "loss": 0.0004, "step": 3200 }, { "epoch": 0.38, "learning_rate": 1.860397830018083e-06, "loss": 0.0021, "step": 3201 }, { "epoch": 0.38, "learning_rate": 1.8600361663652803e-06, "loss": 0.0045, "step": 3202 }, { "epoch": 0.38, "learning_rate": 1.8596745027124773e-06, "loss": 0.0003, "step": 3203 }, { "epoch": 0.38, "learning_rate": 1.8593128390596745e-06, "loss": 0.9494, "step": 3204 }, { "epoch": 0.38, "learning_rate": 1.8589511754068716e-06, "loss": 0.0016, "step": 3205 }, { "epoch": 0.38, "learning_rate": 1.8585895117540688e-06, "loss": 0.0002, "step": 3206 }, { "epoch": 0.38, "learning_rate": 1.8582278481012658e-06, "loss": 0.0003, "step": 3207 }, { "epoch": 0.38, "learning_rate": 1.857866184448463e-06, "loss": 0.0005, "step": 3208 }, { "epoch": 0.38, "learning_rate": 1.85750452079566e-06, "loss": 0.2536, "step": 3209 }, { "epoch": 0.38, "learning_rate": 1.8571428571428573e-06, "loss": 0.0003, "step": 3210 }, { "epoch": 0.38, "learning_rate": 1.8567811934900543e-06, "loss": 0.0004, "step": 3211 }, { "epoch": 0.38, "learning_rate": 1.8564195298372516e-06, "loss": 0.0007, "step": 3212 }, { "epoch": 0.39, "learning_rate": 1.8560578661844486e-06, "loss": 0.0003, "step": 3213 }, { "epoch": 0.39, "learning_rate": 1.8556962025316458e-06, "loss": 0.0007, "step": 3214 }, { "epoch": 0.39, "learning_rate": 1.8553345388788428e-06, "loss": 0.0005, "step": 3215 }, { "epoch": 0.39, "learning_rate": 1.85497287522604e-06, "loss": 0.0004, "step": 3216 }, { "epoch": 0.39, "learning_rate": 1.854611211573237e-06, "loss": 0.0807, "step": 3217 }, { "epoch": 0.39, "learning_rate": 1.8542495479204339e-06, "loss": 0.0004, "step": 3218 }, { "epoch": 0.39, "learning_rate": 1.8538878842676311e-06, "loss": 0.0004, "step": 3219 }, { "epoch": 0.39, "learning_rate": 1.8535262206148281e-06, "loss": 0.7662, "step": 3220 }, { "epoch": 0.39, "learning_rate": 1.8531645569620251e-06, "loss": 0.0004, "step": 3221 }, { "epoch": 0.39, "learning_rate": 1.8528028933092224e-06, "loss": 0.7022, "step": 3222 }, { "epoch": 0.39, "learning_rate": 1.8524412296564194e-06, "loss": 0.0005, "step": 3223 }, { "epoch": 0.39, "learning_rate": 1.8520795660036166e-06, "loss": 0.0005, "step": 3224 }, { "epoch": 0.39, "learning_rate": 1.8517179023508137e-06, "loss": 0.0004, "step": 3225 }, { "epoch": 0.39, "learning_rate": 1.8513562386980109e-06, "loss": 0.0036, "step": 3226 }, { "epoch": 0.39, "learning_rate": 1.850994575045208e-06, "loss": 0.0006, "step": 3227 }, { "epoch": 0.39, "learning_rate": 1.8506329113924051e-06, "loss": 0.0004, "step": 3228 }, { "epoch": 0.39, "learning_rate": 1.8502712477396022e-06, "loss": 0.0004, "step": 3229 }, { "epoch": 0.39, "learning_rate": 1.8499095840867994e-06, "loss": 0.0004, "step": 3230 }, { "epoch": 0.39, "learning_rate": 1.8495479204339964e-06, "loss": 0.0023, "step": 3231 }, { "epoch": 0.39, "learning_rate": 1.8491862567811936e-06, "loss": 0.0005, "step": 3232 }, { "epoch": 0.39, "learning_rate": 1.8488245931283907e-06, "loss": 0.0015, "step": 3233 }, { "epoch": 0.39, "learning_rate": 1.8484629294755879e-06, "loss": 0.0004, "step": 3234 }, { "epoch": 0.39, "learning_rate": 1.848101265822785e-06, "loss": 0.0004, "step": 3235 }, { "epoch": 0.39, "learning_rate": 1.8477396021699821e-06, "loss": 0.0005, "step": 3236 }, { "epoch": 0.39, "learning_rate": 1.8473779385171792e-06, "loss": 0.3569, "step": 3237 }, { "epoch": 0.39, "learning_rate": 1.8470162748643764e-06, "loss": 0.0004, "step": 3238 }, { "epoch": 0.39, "learning_rate": 1.8466546112115734e-06, "loss": 0.0003, "step": 3239 }, { "epoch": 0.39, "learning_rate": 1.8462929475587706e-06, "loss": 0.0003, "step": 3240 }, { "epoch": 0.39, "learning_rate": 1.8459312839059677e-06, "loss": 0.0055, "step": 3241 }, { "epoch": 0.39, "learning_rate": 1.8455696202531645e-06, "loss": 0.0003, "step": 3242 }, { "epoch": 0.39, "learning_rate": 1.8452079566003615e-06, "loss": 0.0004, "step": 3243 }, { "epoch": 0.39, "learning_rate": 1.8448462929475587e-06, "loss": 0.2258, "step": 3244 }, { "epoch": 0.39, "learning_rate": 1.8444846292947557e-06, "loss": 0.0004, "step": 3245 }, { "epoch": 0.39, "learning_rate": 1.844122965641953e-06, "loss": 0.0004, "step": 3246 }, { "epoch": 0.39, "learning_rate": 1.84376130198915e-06, "loss": 0.0107, "step": 3247 }, { "epoch": 0.39, "learning_rate": 1.8433996383363472e-06, "loss": 0.0006, "step": 3248 }, { "epoch": 0.39, "learning_rate": 1.8430379746835442e-06, "loss": 0.0042, "step": 3249 }, { "epoch": 0.39, "learning_rate": 1.8426763110307415e-06, "loss": 0.0016, "step": 3250 }, { "epoch": 0.39, "learning_rate": 1.8423146473779385e-06, "loss": 0.0009, "step": 3251 }, { "epoch": 0.39, "learning_rate": 1.8419529837251357e-06, "loss": 0.0004, "step": 3252 }, { "epoch": 0.39, "learning_rate": 1.8415913200723327e-06, "loss": 0.0012, "step": 3253 }, { "epoch": 0.39, "learning_rate": 1.84122965641953e-06, "loss": 0.0004, "step": 3254 }, { "epoch": 0.39, "learning_rate": 1.840867992766727e-06, "loss": 0.0004, "step": 3255 }, { "epoch": 0.39, "learning_rate": 1.8405063291139242e-06, "loss": 0.0294, "step": 3256 }, { "epoch": 0.39, "learning_rate": 1.8401446654611212e-06, "loss": 0.0004, "step": 3257 }, { "epoch": 0.39, "learning_rate": 1.8397830018083185e-06, "loss": 0.0003, "step": 3258 }, { "epoch": 0.39, "learning_rate": 1.8394213381555155e-06, "loss": 0.0007, "step": 3259 }, { "epoch": 0.39, "learning_rate": 1.8390596745027127e-06, "loss": 0.0021, "step": 3260 }, { "epoch": 0.39, "learning_rate": 1.8386980108499097e-06, "loss": 0.0003, "step": 3261 }, { "epoch": 0.39, "learning_rate": 1.8383363471971068e-06, "loss": 0.0006, "step": 3262 }, { "epoch": 0.39, "learning_rate": 1.837974683544304e-06, "loss": 0.0003, "step": 3263 }, { "epoch": 0.39, "learning_rate": 1.837613019891501e-06, "loss": 0.0003, "step": 3264 }, { "epoch": 0.39, "learning_rate": 1.8372513562386982e-06, "loss": 0.004, "step": 3265 }, { "epoch": 0.39, "learning_rate": 1.836889692585895e-06, "loss": 0.0004, "step": 3266 }, { "epoch": 0.39, "learning_rate": 1.836528028933092e-06, "loss": 0.0003, "step": 3267 }, { "epoch": 0.39, "learning_rate": 1.8361663652802893e-06, "loss": 0.0003, "step": 3268 }, { "epoch": 0.39, "learning_rate": 1.8358047016274863e-06, "loss": 0.2128, "step": 3269 }, { "epoch": 0.39, "learning_rate": 1.8354430379746835e-06, "loss": 0.0004, "step": 3270 }, { "epoch": 0.39, "learning_rate": 1.8350813743218806e-06, "loss": 0.0011, "step": 3271 }, { "epoch": 0.39, "learning_rate": 1.8347197106690778e-06, "loss": 0.2007, "step": 3272 }, { "epoch": 0.39, "learning_rate": 1.8343580470162748e-06, "loss": 0.0003, "step": 3273 }, { "epoch": 0.39, "learning_rate": 1.833996383363472e-06, "loss": 0.0004, "step": 3274 }, { "epoch": 0.39, "learning_rate": 1.833634719710669e-06, "loss": 0.0004, "step": 3275 }, { "epoch": 0.39, "learning_rate": 1.8332730560578663e-06, "loss": 0.333, "step": 3276 }, { "epoch": 0.39, "learning_rate": 1.8329113924050633e-06, "loss": 0.0003, "step": 3277 }, { "epoch": 0.39, "learning_rate": 1.8325497287522605e-06, "loss": 0.0003, "step": 3278 }, { "epoch": 0.39, "learning_rate": 1.8321880650994576e-06, "loss": 0.0004, "step": 3279 }, { "epoch": 0.39, "learning_rate": 1.8318264014466548e-06, "loss": 0.0003, "step": 3280 }, { "epoch": 0.39, "learning_rate": 1.8314647377938518e-06, "loss": 0.0003, "step": 3281 }, { "epoch": 0.39, "learning_rate": 1.8311030741410488e-06, "loss": 0.0004, "step": 3282 }, { "epoch": 0.39, "learning_rate": 1.830741410488246e-06, "loss": 0.0003, "step": 3283 }, { "epoch": 0.39, "learning_rate": 1.830379746835443e-06, "loss": 0.0003, "step": 3284 }, { "epoch": 0.39, "learning_rate": 1.8300180831826403e-06, "loss": 0.5824, "step": 3285 }, { "epoch": 0.39, "learning_rate": 1.8296564195298373e-06, "loss": 0.0011, "step": 3286 }, { "epoch": 0.39, "learning_rate": 1.8292947558770346e-06, "loss": 2.0181, "step": 3287 }, { "epoch": 0.39, "learning_rate": 1.8289330922242316e-06, "loss": 0.0003, "step": 3288 }, { "epoch": 0.39, "learning_rate": 1.8285714285714288e-06, "loss": 0.9789, "step": 3289 }, { "epoch": 0.39, "learning_rate": 1.8282097649186256e-06, "loss": 0.0004, "step": 3290 }, { "epoch": 0.39, "learning_rate": 1.8278481012658226e-06, "loss": 0.0006, "step": 3291 }, { "epoch": 0.39, "learning_rate": 1.8274864376130199e-06, "loss": 0.0016, "step": 3292 }, { "epoch": 0.39, "learning_rate": 1.827124773960217e-06, "loss": 0.0003, "step": 3293 }, { "epoch": 0.39, "learning_rate": 1.8267631103074141e-06, "loss": 0.0006, "step": 3294 }, { "epoch": 0.39, "learning_rate": 1.8264014466546111e-06, "loss": 0.0064, "step": 3295 }, { "epoch": 0.39, "learning_rate": 1.8260397830018084e-06, "loss": 0.0007, "step": 3296 }, { "epoch": 0.4, "learning_rate": 1.8256781193490054e-06, "loss": 0.0003, "step": 3297 }, { "epoch": 0.4, "learning_rate": 1.8253164556962026e-06, "loss": 0.0005, "step": 3298 }, { "epoch": 0.4, "learning_rate": 1.8249547920433997e-06, "loss": 0.0005, "step": 3299 }, { "epoch": 0.4, "learning_rate": 1.8245931283905969e-06, "loss": 0.0004, "step": 3300 }, { "epoch": 0.4, "learning_rate": 1.824231464737794e-06, "loss": 0.0005, "step": 3301 }, { "epoch": 0.4, "learning_rate": 1.823869801084991e-06, "loss": 0.0004, "step": 3302 }, { "epoch": 0.4, "learning_rate": 1.8235081374321882e-06, "loss": 0.0025, "step": 3303 }, { "epoch": 0.4, "learning_rate": 1.8231464737793852e-06, "loss": 0.0003, "step": 3304 }, { "epoch": 0.4, "learning_rate": 1.8227848101265824e-06, "loss": 0.0004, "step": 3305 }, { "epoch": 0.4, "learning_rate": 1.8224231464737794e-06, "loss": 0.0007, "step": 3306 }, { "epoch": 0.4, "learning_rate": 1.8220614828209767e-06, "loss": 0.0004, "step": 3307 }, { "epoch": 0.4, "learning_rate": 1.8216998191681737e-06, "loss": 0.0005, "step": 3308 }, { "epoch": 0.4, "learning_rate": 1.821338155515371e-06, "loss": 0.0004, "step": 3309 }, { "epoch": 0.4, "learning_rate": 1.820976491862568e-06, "loss": 0.0005, "step": 3310 }, { "epoch": 0.4, "learning_rate": 1.8206148282097652e-06, "loss": 0.0004, "step": 3311 }, { "epoch": 0.4, "learning_rate": 1.8202531645569622e-06, "loss": 0.0005, "step": 3312 }, { "epoch": 0.4, "learning_rate": 1.8198915009041594e-06, "loss": 0.0011, "step": 3313 }, { "epoch": 0.4, "learning_rate": 1.8195298372513562e-06, "loss": 0.0004, "step": 3314 }, { "epoch": 0.4, "learning_rate": 1.8191681735985532e-06, "loss": 0.0005, "step": 3315 }, { "epoch": 0.4, "learning_rate": 1.8188065099457505e-06, "loss": 0.0004, "step": 3316 }, { "epoch": 0.4, "learning_rate": 1.8184448462929475e-06, "loss": 0.0004, "step": 3317 }, { "epoch": 0.4, "learning_rate": 1.8180831826401447e-06, "loss": 0.0004, "step": 3318 }, { "epoch": 0.4, "learning_rate": 1.8177215189873417e-06, "loss": 0.3825, "step": 3319 }, { "epoch": 0.4, "learning_rate": 1.8173598553345388e-06, "loss": 0.0004, "step": 3320 }, { "epoch": 0.4, "learning_rate": 1.816998191681736e-06, "loss": 0.0005, "step": 3321 }, { "epoch": 0.4, "learning_rate": 1.816636528028933e-06, "loss": 0.7545, "step": 3322 }, { "epoch": 0.4, "learning_rate": 1.8162748643761302e-06, "loss": 0.0003, "step": 3323 }, { "epoch": 0.4, "learning_rate": 1.8159132007233273e-06, "loss": 0.0382, "step": 3324 }, { "epoch": 0.4, "learning_rate": 1.8155515370705245e-06, "loss": 0.0003, "step": 3325 }, { "epoch": 0.4, "learning_rate": 1.8151898734177215e-06, "loss": 0.0004, "step": 3326 }, { "epoch": 0.4, "learning_rate": 1.8148282097649187e-06, "loss": 0.0006, "step": 3327 }, { "epoch": 0.4, "learning_rate": 1.8144665461121158e-06, "loss": 0.0005, "step": 3328 }, { "epoch": 0.4, "learning_rate": 1.814104882459313e-06, "loss": 0.0003, "step": 3329 }, { "epoch": 0.4, "learning_rate": 1.81374321880651e-06, "loss": 0.0004, "step": 3330 }, { "epoch": 0.4, "learning_rate": 1.8133815551537072e-06, "loss": 0.0003, "step": 3331 }, { "epoch": 0.4, "learning_rate": 1.8130198915009043e-06, "loss": 0.0007, "step": 3332 }, { "epoch": 0.4, "learning_rate": 1.8126582278481015e-06, "loss": 0.0004, "step": 3333 }, { "epoch": 0.4, "learning_rate": 1.8122965641952985e-06, "loss": 0.0006, "step": 3334 }, { "epoch": 0.4, "learning_rate": 1.8119349005424957e-06, "loss": 0.0004, "step": 3335 }, { "epoch": 0.4, "learning_rate": 1.8115732368896928e-06, "loss": 0.0004, "step": 3336 }, { "epoch": 0.4, "learning_rate": 1.8112115732368896e-06, "loss": 0.3401, "step": 3337 }, { "epoch": 0.4, "learning_rate": 1.8108499095840868e-06, "loss": 0.0003, "step": 3338 }, { "epoch": 0.4, "learning_rate": 1.8104882459312838e-06, "loss": 0.0003, "step": 3339 }, { "epoch": 0.4, "learning_rate": 1.8101265822784808e-06, "loss": 0.0004, "step": 3340 }, { "epoch": 0.4, "learning_rate": 1.809764918625678e-06, "loss": 0.0003, "step": 3341 }, { "epoch": 0.4, "learning_rate": 1.809403254972875e-06, "loss": 0.0142, "step": 3342 }, { "epoch": 0.4, "learning_rate": 1.8090415913200723e-06, "loss": 0.0003, "step": 3343 }, { "epoch": 0.4, "learning_rate": 1.8086799276672693e-06, "loss": 0.0002, "step": 3344 }, { "epoch": 0.4, "learning_rate": 1.8083182640144666e-06, "loss": 0.0003, "step": 3345 }, { "epoch": 0.4, "learning_rate": 1.8079566003616636e-06, "loss": 0.0004, "step": 3346 }, { "epoch": 0.4, "learning_rate": 1.8075949367088608e-06, "loss": 0.0004, "step": 3347 }, { "epoch": 0.4, "learning_rate": 1.8072332730560578e-06, "loss": 0.0171, "step": 3348 }, { "epoch": 0.4, "learning_rate": 1.806871609403255e-06, "loss": 0.0005, "step": 3349 }, { "epoch": 0.4, "learning_rate": 1.806509945750452e-06, "loss": 0.0003, "step": 3350 }, { "epoch": 0.4, "learning_rate": 1.8061482820976493e-06, "loss": 0.0003, "step": 3351 }, { "epoch": 0.4, "learning_rate": 1.8057866184448463e-06, "loss": 0.2991, "step": 3352 }, { "epoch": 0.4, "learning_rate": 1.8054249547920436e-06, "loss": 0.0003, "step": 3353 }, { "epoch": 0.4, "learning_rate": 1.8050632911392406e-06, "loss": 0.9208, "step": 3354 }, { "epoch": 0.4, "learning_rate": 1.8047016274864378e-06, "loss": 0.0005, "step": 3355 }, { "epoch": 0.4, "learning_rate": 1.8043399638336348e-06, "loss": 0.0003, "step": 3356 }, { "epoch": 0.4, "learning_rate": 1.803978300180832e-06, "loss": 0.0003, "step": 3357 }, { "epoch": 0.4, "learning_rate": 1.803616636528029e-06, "loss": 0.001, "step": 3358 }, { "epoch": 0.4, "learning_rate": 1.8032549728752263e-06, "loss": 0.0003, "step": 3359 }, { "epoch": 0.4, "learning_rate": 1.8028933092224233e-06, "loss": 0.0004, "step": 3360 }, { "epoch": 0.4, "learning_rate": 1.8025316455696201e-06, "loss": 0.0005, "step": 3361 }, { "epoch": 0.4, "learning_rate": 1.8021699819168172e-06, "loss": 0.0004, "step": 3362 }, { "epoch": 0.4, "learning_rate": 1.8018083182640144e-06, "loss": 0.0005, "step": 3363 }, { "epoch": 0.4, "learning_rate": 1.8014466546112114e-06, "loss": 0.0004, "step": 3364 }, { "epoch": 0.4, "learning_rate": 1.8010849909584086e-06, "loss": 0.0003, "step": 3365 }, { "epoch": 0.4, "learning_rate": 1.8007233273056057e-06, "loss": 0.0518, "step": 3366 }, { "epoch": 0.4, "learning_rate": 1.800361663652803e-06, "loss": 0.0004, "step": 3367 }, { "epoch": 0.4, "learning_rate": 1.8e-06, "loss": 0.0003, "step": 3368 }, { "epoch": 0.4, "learning_rate": 1.7996383363471971e-06, "loss": 0.0005, "step": 3369 }, { "epoch": 0.4, "learning_rate": 1.7992766726943942e-06, "loss": 0.0007, "step": 3370 }, { "epoch": 0.4, "learning_rate": 1.7989150090415914e-06, "loss": 0.8035, "step": 3371 }, { "epoch": 0.4, "learning_rate": 1.7985533453887884e-06, "loss": 0.0004, "step": 3372 }, { "epoch": 0.4, "learning_rate": 1.7981916817359857e-06, "loss": 0.0003, "step": 3373 }, { "epoch": 0.4, "learning_rate": 1.7978300180831827e-06, "loss": 0.0711, "step": 3374 }, { "epoch": 0.4, "learning_rate": 1.79746835443038e-06, "loss": 0.0003, "step": 3375 }, { "epoch": 0.4, "learning_rate": 1.797106690777577e-06, "loss": 0.0005, "step": 3376 }, { "epoch": 0.4, "learning_rate": 1.7967450271247742e-06, "loss": 0.0003, "step": 3377 }, { "epoch": 0.4, "learning_rate": 1.7963833634719712e-06, "loss": 0.0005, "step": 3378 }, { "epoch": 0.4, "learning_rate": 1.7960216998191684e-06, "loss": 0.001, "step": 3379 }, { "epoch": 0.41, "learning_rate": 1.7956600361663654e-06, "loss": 0.0003, "step": 3380 }, { "epoch": 0.41, "learning_rate": 1.7952983725135624e-06, "loss": 0.0004, "step": 3381 }, { "epoch": 0.41, "learning_rate": 1.7949367088607597e-06, "loss": 0.0003, "step": 3382 }, { "epoch": 0.41, "learning_rate": 1.7945750452079567e-06, "loss": 0.0004, "step": 3383 }, { "epoch": 0.41, "learning_rate": 1.794213381555154e-06, "loss": 0.8786, "step": 3384 }, { "epoch": 0.41, "learning_rate": 1.7938517179023507e-06, "loss": 0.0002, "step": 3385 }, { "epoch": 0.41, "learning_rate": 1.7934900542495478e-06, "loss": 0.0004, "step": 3386 }, { "epoch": 0.41, "learning_rate": 1.793128390596745e-06, "loss": 0.0004, "step": 3387 }, { "epoch": 0.41, "learning_rate": 1.792766726943942e-06, "loss": 0.0003, "step": 3388 }, { "epoch": 0.41, "learning_rate": 1.7924050632911392e-06, "loss": 0.0075, "step": 3389 }, { "epoch": 0.41, "learning_rate": 1.7920433996383363e-06, "loss": 0.0006, "step": 3390 }, { "epoch": 0.41, "learning_rate": 1.7916817359855335e-06, "loss": 0.0002, "step": 3391 }, { "epoch": 0.41, "learning_rate": 1.7913200723327305e-06, "loss": 0.0003, "step": 3392 }, { "epoch": 0.41, "learning_rate": 1.7909584086799277e-06, "loss": 0.0005, "step": 3393 }, { "epoch": 0.41, "learning_rate": 1.7905967450271248e-06, "loss": 0.0003, "step": 3394 }, { "epoch": 0.41, "learning_rate": 1.790235081374322e-06, "loss": 0.0105, "step": 3395 }, { "epoch": 0.41, "learning_rate": 1.789873417721519e-06, "loss": 0.0004, "step": 3396 }, { "epoch": 0.41, "learning_rate": 1.7895117540687162e-06, "loss": 0.0004, "step": 3397 }, { "epoch": 0.41, "learning_rate": 1.7891500904159133e-06, "loss": 0.0003, "step": 3398 }, { "epoch": 0.41, "learning_rate": 1.7887884267631105e-06, "loss": 0.0006, "step": 3399 }, { "epoch": 0.41, "learning_rate": 1.7884267631103075e-06, "loss": 0.8481, "step": 3400 }, { "epoch": 0.41, "learning_rate": 1.7880650994575045e-06, "loss": 0.0003, "step": 3401 }, { "epoch": 0.41, "learning_rate": 1.7877034358047018e-06, "loss": 0.0003, "step": 3402 }, { "epoch": 0.41, "learning_rate": 1.7873417721518988e-06, "loss": 0.0003, "step": 3403 }, { "epoch": 0.41, "learning_rate": 1.786980108499096e-06, "loss": 0.8839, "step": 3404 }, { "epoch": 0.41, "learning_rate": 1.786618444846293e-06, "loss": 0.0003, "step": 3405 }, { "epoch": 0.41, "learning_rate": 1.7862567811934903e-06, "loss": 0.0003, "step": 3406 }, { "epoch": 0.41, "learning_rate": 1.7858951175406873e-06, "loss": 0.0021, "step": 3407 }, { "epoch": 0.41, "learning_rate": 1.7855334538878845e-06, "loss": 0.1532, "step": 3408 }, { "epoch": 0.41, "learning_rate": 1.7851717902350813e-06, "loss": 0.0004, "step": 3409 }, { "epoch": 0.41, "learning_rate": 1.7848101265822783e-06, "loss": 0.0004, "step": 3410 }, { "epoch": 0.41, "learning_rate": 1.7844484629294756e-06, "loss": 0.0004, "step": 3411 }, { "epoch": 0.41, "learning_rate": 1.7840867992766726e-06, "loss": 0.0005, "step": 3412 }, { "epoch": 0.41, "learning_rate": 1.7837251356238698e-06, "loss": 0.0004, "step": 3413 }, { "epoch": 0.41, "learning_rate": 1.7833634719710668e-06, "loss": 0.0027, "step": 3414 }, { "epoch": 0.41, "learning_rate": 1.783001808318264e-06, "loss": 0.0003, "step": 3415 }, { "epoch": 0.41, "learning_rate": 1.782640144665461e-06, "loss": 0.0005, "step": 3416 }, { "epoch": 0.41, "learning_rate": 1.7822784810126583e-06, "loss": 0.7349, "step": 3417 }, { "epoch": 0.41, "learning_rate": 1.7819168173598553e-06, "loss": 0.0004, "step": 3418 }, { "epoch": 0.41, "learning_rate": 1.7815551537070526e-06, "loss": 0.0005, "step": 3419 }, { "epoch": 0.41, "learning_rate": 1.7811934900542496e-06, "loss": 0.0004, "step": 3420 }, { "epoch": 0.41, "learning_rate": 1.7808318264014466e-06, "loss": 0.0007, "step": 3421 }, { "epoch": 0.41, "learning_rate": 1.7804701627486438e-06, "loss": 0.0005, "step": 3422 }, { "epoch": 0.41, "learning_rate": 1.7801084990958409e-06, "loss": 0.0004, "step": 3423 }, { "epoch": 0.41, "learning_rate": 1.779746835443038e-06, "loss": 0.0004, "step": 3424 }, { "epoch": 0.41, "learning_rate": 1.7793851717902351e-06, "loss": 0.0159, "step": 3425 }, { "epoch": 0.41, "learning_rate": 1.7790235081374323e-06, "loss": 0.0225, "step": 3426 }, { "epoch": 0.41, "learning_rate": 1.7786618444846294e-06, "loss": 0.0848, "step": 3427 }, { "epoch": 0.41, "learning_rate": 1.7783001808318266e-06, "loss": 0.0005, "step": 3428 }, { "epoch": 0.41, "learning_rate": 1.7779385171790236e-06, "loss": 0.0006, "step": 3429 }, { "epoch": 0.41, "learning_rate": 1.7775768535262208e-06, "loss": 0.0004, "step": 3430 }, { "epoch": 0.41, "learning_rate": 1.7772151898734179e-06, "loss": 0.0003, "step": 3431 }, { "epoch": 0.41, "learning_rate": 1.776853526220615e-06, "loss": 0.0043, "step": 3432 }, { "epoch": 0.41, "learning_rate": 1.776491862567812e-06, "loss": 0.6287, "step": 3433 }, { "epoch": 0.41, "learning_rate": 1.776130198915009e-06, "loss": 0.9403, "step": 3434 }, { "epoch": 0.41, "learning_rate": 1.7757685352622061e-06, "loss": 0.0032, "step": 3435 }, { "epoch": 0.41, "learning_rate": 1.7754068716094032e-06, "loss": 0.871, "step": 3436 }, { "epoch": 0.41, "learning_rate": 1.7750452079566004e-06, "loss": 0.0002, "step": 3437 }, { "epoch": 0.41, "learning_rate": 1.7746835443037974e-06, "loss": 0.0004, "step": 3438 }, { "epoch": 0.41, "learning_rate": 1.7743218806509946e-06, "loss": 0.0004, "step": 3439 }, { "epoch": 0.41, "learning_rate": 1.7739602169981917e-06, "loss": 0.0005, "step": 3440 }, { "epoch": 0.41, "learning_rate": 1.7735985533453887e-06, "loss": 0.0004, "step": 3441 }, { "epoch": 0.41, "learning_rate": 1.773236889692586e-06, "loss": 0.0004, "step": 3442 }, { "epoch": 0.41, "learning_rate": 1.772875226039783e-06, "loss": 0.0004, "step": 3443 }, { "epoch": 0.41, "learning_rate": 1.7725135623869802e-06, "loss": 0.0005, "step": 3444 }, { "epoch": 0.41, "learning_rate": 1.7721518987341772e-06, "loss": 0.0004, "step": 3445 }, { "epoch": 0.41, "learning_rate": 1.7717902350813744e-06, "loss": 0.0002, "step": 3446 }, { "epoch": 0.41, "learning_rate": 1.7714285714285714e-06, "loss": 0.8766, "step": 3447 }, { "epoch": 0.41, "learning_rate": 1.7710669077757687e-06, "loss": 0.0003, "step": 3448 }, { "epoch": 0.41, "learning_rate": 1.7707052441229657e-06, "loss": 0.0005, "step": 3449 }, { "epoch": 0.41, "learning_rate": 1.770343580470163e-06, "loss": 0.0004, "step": 3450 }, { "epoch": 0.41, "learning_rate": 1.76998191681736e-06, "loss": 0.0005, "step": 3451 }, { "epoch": 0.41, "learning_rate": 1.7696202531645572e-06, "loss": 0.0003, "step": 3452 }, { "epoch": 0.41, "learning_rate": 1.7692585895117542e-06, "loss": 0.0004, "step": 3453 }, { "epoch": 0.41, "learning_rate": 1.7688969258589514e-06, "loss": 0.0003, "step": 3454 }, { "epoch": 0.41, "learning_rate": 1.7685352622061484e-06, "loss": 0.0003, "step": 3455 }, { "epoch": 0.41, "learning_rate": 1.7681735985533457e-06, "loss": 0.0004, "step": 3456 }, { "epoch": 0.41, "learning_rate": 1.7678119349005425e-06, "loss": 0.0021, "step": 3457 }, { "epoch": 0.41, "learning_rate": 1.7674502712477395e-06, "loss": 0.0003, "step": 3458 }, { "epoch": 0.41, "learning_rate": 1.7670886075949365e-06, "loss": 0.0021, "step": 3459 }, { "epoch": 0.41, "learning_rate": 1.7667269439421338e-06, "loss": 0.0003, "step": 3460 }, { "epoch": 0.41, "learning_rate": 1.7663652802893308e-06, "loss": 0.0002, "step": 3461 }, { "epoch": 0.41, "learning_rate": 1.766003616636528e-06, "loss": 0.0004, "step": 3462 }, { "epoch": 0.41, "learning_rate": 1.765641952983725e-06, "loss": 0.4772, "step": 3463 }, { "epoch": 0.42, "learning_rate": 1.7652802893309223e-06, "loss": 0.0005, "step": 3464 }, { "epoch": 0.42, "learning_rate": 1.7649186256781193e-06, "loss": 0.0005, "step": 3465 }, { "epoch": 0.42, "learning_rate": 1.7645569620253165e-06, "loss": 0.0685, "step": 3466 }, { "epoch": 0.42, "learning_rate": 1.7641952983725135e-06, "loss": 0.0003, "step": 3467 }, { "epoch": 0.42, "learning_rate": 1.7638336347197108e-06, "loss": 0.0003, "step": 3468 }, { "epoch": 0.42, "learning_rate": 1.7634719710669078e-06, "loss": 0.0003, "step": 3469 }, { "epoch": 0.42, "learning_rate": 1.763110307414105e-06, "loss": 0.0003, "step": 3470 }, { "epoch": 0.42, "learning_rate": 1.762748643761302e-06, "loss": 0.0003, "step": 3471 }, { "epoch": 0.42, "learning_rate": 1.7623869801084993e-06, "loss": 0.0002, "step": 3472 }, { "epoch": 0.42, "learning_rate": 1.7620253164556963e-06, "loss": 0.3625, "step": 3473 }, { "epoch": 0.42, "learning_rate": 1.7616636528028935e-06, "loss": 0.3983, "step": 3474 }, { "epoch": 0.42, "learning_rate": 1.7613019891500905e-06, "loss": 0.0003, "step": 3475 }, { "epoch": 0.42, "learning_rate": 1.7609403254972878e-06, "loss": 0.0004, "step": 3476 }, { "epoch": 0.42, "learning_rate": 1.7605786618444848e-06, "loss": 0.1238, "step": 3477 }, { "epoch": 0.42, "learning_rate": 1.760216998191682e-06, "loss": 0.6219, "step": 3478 }, { "epoch": 0.42, "learning_rate": 1.759855334538879e-06, "loss": 0.0003, "step": 3479 }, { "epoch": 0.42, "learning_rate": 1.7594936708860763e-06, "loss": 0.8001, "step": 3480 }, { "epoch": 0.42, "learning_rate": 1.7591320072332729e-06, "loss": 0.0006, "step": 3481 }, { "epoch": 0.42, "learning_rate": 1.75877034358047e-06, "loss": 0.0003, "step": 3482 }, { "epoch": 0.42, "learning_rate": 1.758408679927667e-06, "loss": 0.222, "step": 3483 }, { "epoch": 0.42, "learning_rate": 1.7580470162748643e-06, "loss": 0.001, "step": 3484 }, { "epoch": 0.42, "learning_rate": 1.7576853526220614e-06, "loss": 0.0024, "step": 3485 }, { "epoch": 0.42, "learning_rate": 1.7573236889692586e-06, "loss": 0.0003, "step": 3486 }, { "epoch": 0.42, "learning_rate": 1.7569620253164556e-06, "loss": 0.0004, "step": 3487 }, { "epoch": 0.42, "learning_rate": 1.7566003616636528e-06, "loss": 0.0003, "step": 3488 }, { "epoch": 0.42, "learning_rate": 1.7562386980108499e-06, "loss": 0.0003, "step": 3489 }, { "epoch": 0.42, "learning_rate": 1.755877034358047e-06, "loss": 0.0004, "step": 3490 }, { "epoch": 0.42, "learning_rate": 1.755515370705244e-06, "loss": 0.0013, "step": 3491 }, { "epoch": 0.42, "learning_rate": 1.7551537070524413e-06, "loss": 0.0003, "step": 3492 }, { "epoch": 0.42, "learning_rate": 1.7547920433996384e-06, "loss": 0.0003, "step": 3493 }, { "epoch": 0.42, "learning_rate": 1.7544303797468356e-06, "loss": 0.0003, "step": 3494 }, { "epoch": 0.42, "learning_rate": 1.7540687160940326e-06, "loss": 0.0003, "step": 3495 }, { "epoch": 0.42, "learning_rate": 1.7537070524412298e-06, "loss": 0.0003, "step": 3496 }, { "epoch": 0.42, "learning_rate": 1.7533453887884269e-06, "loss": 0.0003, "step": 3497 }, { "epoch": 0.42, "learning_rate": 1.752983725135624e-06, "loss": 0.0003, "step": 3498 }, { "epoch": 0.42, "learning_rate": 1.7526220614828211e-06, "loss": 0.5663, "step": 3499 }, { "epoch": 0.42, "learning_rate": 1.7522603978300183e-06, "loss": 0.0003, "step": 3500 }, { "epoch": 0.42, "learning_rate": 1.7518987341772154e-06, "loss": 0.0003, "step": 3501 }, { "epoch": 0.42, "learning_rate": 1.7515370705244124e-06, "loss": 0.0002, "step": 3502 }, { "epoch": 0.42, "learning_rate": 1.7511754068716096e-06, "loss": 0.9949, "step": 3503 }, { "epoch": 0.42, "learning_rate": 1.7508137432188066e-06, "loss": 0.0003, "step": 3504 }, { "epoch": 0.42, "learning_rate": 1.7504520795660034e-06, "loss": 0.0003, "step": 3505 }, { "epoch": 0.42, "learning_rate": 1.7500904159132007e-06, "loss": 0.0003, "step": 3506 }, { "epoch": 0.42, "learning_rate": 1.7497287522603977e-06, "loss": 0.0005, "step": 3507 }, { "epoch": 0.42, "learning_rate": 1.749367088607595e-06, "loss": 0.0003, "step": 3508 }, { "epoch": 0.42, "learning_rate": 1.749005424954792e-06, "loss": 0.5555, "step": 3509 }, { "epoch": 0.42, "learning_rate": 1.7486437613019892e-06, "loss": 0.9679, "step": 3510 }, { "epoch": 0.42, "learning_rate": 1.7482820976491862e-06, "loss": 0.0003, "step": 3511 }, { "epoch": 0.42, "learning_rate": 1.7479204339963834e-06, "loss": 1.0538, "step": 3512 }, { "epoch": 0.42, "learning_rate": 1.7475587703435804e-06, "loss": 0.0003, "step": 3513 }, { "epoch": 0.42, "learning_rate": 1.7471971066907777e-06, "loss": 0.0005, "step": 3514 }, { "epoch": 0.42, "learning_rate": 1.7468354430379747e-06, "loss": 0.0002, "step": 3515 }, { "epoch": 0.42, "learning_rate": 1.746473779385172e-06, "loss": 0.0003, "step": 3516 }, { "epoch": 0.42, "learning_rate": 1.746112115732369e-06, "loss": 0.0003, "step": 3517 }, { "epoch": 0.42, "learning_rate": 1.7457504520795662e-06, "loss": 0.0155, "step": 3518 }, { "epoch": 0.42, "learning_rate": 1.7453887884267632e-06, "loss": 0.0003, "step": 3519 }, { "epoch": 0.42, "learning_rate": 1.7450271247739604e-06, "loss": 0.0012, "step": 3520 }, { "epoch": 0.42, "learning_rate": 1.7446654611211574e-06, "loss": 0.8334, "step": 3521 }, { "epoch": 0.42, "learning_rate": 1.7443037974683545e-06, "loss": 0.0003, "step": 3522 }, { "epoch": 0.42, "learning_rate": 1.7439421338155517e-06, "loss": 0.0004, "step": 3523 }, { "epoch": 0.42, "learning_rate": 1.7435804701627487e-06, "loss": 0.0162, "step": 3524 }, { "epoch": 0.42, "learning_rate": 1.743218806509946e-06, "loss": 0.0118, "step": 3525 }, { "epoch": 0.42, "learning_rate": 1.742857142857143e-06, "loss": 0.0003, "step": 3526 }, { "epoch": 0.42, "learning_rate": 1.7424954792043402e-06, "loss": 0.1943, "step": 3527 }, { "epoch": 0.42, "learning_rate": 1.7421338155515372e-06, "loss": 0.0002, "step": 3528 }, { "epoch": 0.42, "learning_rate": 1.741772151898734e-06, "loss": 0.0005, "step": 3529 }, { "epoch": 0.42, "learning_rate": 1.7414104882459312e-06, "loss": 0.0004, "step": 3530 }, { "epoch": 0.42, "learning_rate": 1.7410488245931283e-06, "loss": 0.8398, "step": 3531 }, { "epoch": 0.42, "learning_rate": 1.7406871609403255e-06, "loss": 0.0003, "step": 3532 }, { "epoch": 0.42, "learning_rate": 1.7403254972875225e-06, "loss": 0.0003, "step": 3533 }, { "epoch": 0.42, "learning_rate": 1.7399638336347198e-06, "loss": 0.0003, "step": 3534 }, { "epoch": 0.42, "learning_rate": 1.7396021699819168e-06, "loss": 0.0003, "step": 3535 }, { "epoch": 0.42, "learning_rate": 1.739240506329114e-06, "loss": 0.0099, "step": 3536 }, { "epoch": 0.42, "learning_rate": 1.738878842676311e-06, "loss": 0.0002, "step": 3537 }, { "epoch": 0.42, "learning_rate": 1.7385171790235083e-06, "loss": 0.0004, "step": 3538 }, { "epoch": 0.42, "learning_rate": 1.7381555153707053e-06, "loss": 0.0003, "step": 3539 }, { "epoch": 0.42, "learning_rate": 1.7377938517179023e-06, "loss": 0.0002, "step": 3540 }, { "epoch": 0.42, "learning_rate": 1.7374321880650995e-06, "loss": 0.0003, "step": 3541 }, { "epoch": 0.42, "learning_rate": 1.7370705244122965e-06, "loss": 0.0069, "step": 3542 }, { "epoch": 0.42, "learning_rate": 1.7367088607594938e-06, "loss": 0.0003, "step": 3543 }, { "epoch": 0.42, "learning_rate": 1.7363471971066908e-06, "loss": 0.0003, "step": 3544 }, { "epoch": 0.42, "learning_rate": 1.735985533453888e-06, "loss": 0.0003, "step": 3545 }, { "epoch": 0.42, "learning_rate": 1.735623869801085e-06, "loss": 0.0003, "step": 3546 }, { "epoch": 0.43, "learning_rate": 1.7352622061482823e-06, "loss": 0.0004, "step": 3547 }, { "epoch": 0.43, "learning_rate": 1.7349005424954793e-06, "loss": 0.0003, "step": 3548 }, { "epoch": 0.43, "learning_rate": 1.7345388788426765e-06, "loss": 0.0003, "step": 3549 }, { "epoch": 0.43, "learning_rate": 1.7341772151898735e-06, "loss": 0.0003, "step": 3550 }, { "epoch": 0.43, "learning_rate": 1.7338155515370708e-06, "loss": 0.0003, "step": 3551 }, { "epoch": 0.43, "learning_rate": 1.7334538878842678e-06, "loss": 0.0002, "step": 3552 }, { "epoch": 0.43, "learning_rate": 1.7330922242314646e-06, "loss": 0.0003, "step": 3553 }, { "epoch": 0.43, "learning_rate": 1.7327305605786618e-06, "loss": 0.0006, "step": 3554 }, { "epoch": 0.43, "learning_rate": 1.7323688969258589e-06, "loss": 1.0107, "step": 3555 }, { "epoch": 0.43, "learning_rate": 1.732007233273056e-06, "loss": 0.0027, "step": 3556 }, { "epoch": 0.43, "learning_rate": 1.731645569620253e-06, "loss": 0.0003, "step": 3557 }, { "epoch": 0.43, "learning_rate": 1.7312839059674503e-06, "loss": 0.0003, "step": 3558 }, { "epoch": 0.43, "learning_rate": 1.7309222423146474e-06, "loss": 0.4761, "step": 3559 }, { "epoch": 0.43, "learning_rate": 1.7305605786618444e-06, "loss": 0.0003, "step": 3560 }, { "epoch": 0.43, "learning_rate": 1.7301989150090416e-06, "loss": 0.0004, "step": 3561 }, { "epoch": 0.43, "learning_rate": 1.7298372513562386e-06, "loss": 0.0003, "step": 3562 }, { "epoch": 0.43, "learning_rate": 1.7294755877034359e-06, "loss": 0.0003, "step": 3563 }, { "epoch": 0.43, "learning_rate": 1.7291139240506329e-06, "loss": 0.0003, "step": 3564 }, { "epoch": 0.43, "learning_rate": 1.72875226039783e-06, "loss": 0.0003, "step": 3565 }, { "epoch": 0.43, "learning_rate": 1.7283905967450271e-06, "loss": 0.0003, "step": 3566 }, { "epoch": 0.43, "learning_rate": 1.7280289330922244e-06, "loss": 1.0801, "step": 3567 }, { "epoch": 0.43, "learning_rate": 1.7276672694394214e-06, "loss": 0.0003, "step": 3568 }, { "epoch": 0.43, "learning_rate": 1.7273056057866186e-06, "loss": 0.0003, "step": 3569 }, { "epoch": 0.43, "learning_rate": 1.7269439421338156e-06, "loss": 0.4064, "step": 3570 }, { "epoch": 0.43, "learning_rate": 1.7265822784810129e-06, "loss": 0.0025, "step": 3571 }, { "epoch": 0.43, "learning_rate": 1.7262206148282099e-06, "loss": 0.0003, "step": 3572 }, { "epoch": 0.43, "learning_rate": 1.7258589511754071e-06, "loss": 0.0003, "step": 3573 }, { "epoch": 0.43, "learning_rate": 1.7254972875226041e-06, "loss": 0.0003, "step": 3574 }, { "epoch": 0.43, "learning_rate": 1.7251356238698014e-06, "loss": 0.0005, "step": 3575 }, { "epoch": 0.43, "learning_rate": 1.7247739602169982e-06, "loss": 0.0003, "step": 3576 }, { "epoch": 0.43, "learning_rate": 1.7244122965641952e-06, "loss": 0.0004, "step": 3577 }, { "epoch": 0.43, "learning_rate": 1.7240506329113924e-06, "loss": 0.0005, "step": 3578 }, { "epoch": 0.43, "learning_rate": 1.7236889692585894e-06, "loss": 0.8358, "step": 3579 }, { "epoch": 0.43, "learning_rate": 1.7233273056057865e-06, "loss": 0.0002, "step": 3580 }, { "epoch": 0.43, "learning_rate": 1.7229656419529837e-06, "loss": 0.0005, "step": 3581 }, { "epoch": 0.43, "learning_rate": 1.7226039783001807e-06, "loss": 0.0004, "step": 3582 }, { "epoch": 0.43, "learning_rate": 1.722242314647378e-06, "loss": 0.7364, "step": 3583 }, { "epoch": 0.43, "learning_rate": 1.721880650994575e-06, "loss": 0.0005, "step": 3584 }, { "epoch": 0.43, "learning_rate": 1.7215189873417722e-06, "loss": 0.0006, "step": 3585 }, { "epoch": 0.43, "learning_rate": 1.7211573236889692e-06, "loss": 0.0003, "step": 3586 }, { "epoch": 0.43, "learning_rate": 1.7207956600361664e-06, "loss": 0.0004, "step": 3587 }, { "epoch": 0.43, "learning_rate": 1.7204339963833635e-06, "loss": 0.0003, "step": 3588 }, { "epoch": 0.43, "learning_rate": 1.7200723327305607e-06, "loss": 0.0003, "step": 3589 }, { "epoch": 0.43, "learning_rate": 1.7197106690777577e-06, "loss": 0.0003, "step": 3590 }, { "epoch": 0.43, "learning_rate": 1.719349005424955e-06, "loss": 0.1431, "step": 3591 }, { "epoch": 0.43, "learning_rate": 1.718987341772152e-06, "loss": 0.0002, "step": 3592 }, { "epoch": 0.43, "learning_rate": 1.7186256781193492e-06, "loss": 0.0011, "step": 3593 }, { "epoch": 0.43, "learning_rate": 1.7182640144665462e-06, "loss": 0.9389, "step": 3594 }, { "epoch": 0.43, "learning_rate": 1.7179023508137434e-06, "loss": 0.0005, "step": 3595 }, { "epoch": 0.43, "learning_rate": 1.7175406871609405e-06, "loss": 0.0291, "step": 3596 }, { "epoch": 0.43, "learning_rate": 1.7171790235081377e-06, "loss": 0.0003, "step": 3597 }, { "epoch": 0.43, "learning_rate": 1.7168173598553347e-06, "loss": 0.03, "step": 3598 }, { "epoch": 0.43, "learning_rate": 1.716455696202532e-06, "loss": 0.0003, "step": 3599 }, { "epoch": 0.43, "learning_rate": 1.7160940325497285e-06, "loss": 0.0003, "step": 3600 }, { "epoch": 0.43, "learning_rate": 1.7157323688969258e-06, "loss": 0.0055, "step": 3601 }, { "epoch": 0.43, "learning_rate": 1.7153707052441228e-06, "loss": 0.0003, "step": 3602 }, { "epoch": 0.43, "learning_rate": 1.71500904159132e-06, "loss": 0.0003, "step": 3603 }, { "epoch": 0.43, "learning_rate": 1.714647377938517e-06, "loss": 0.0812, "step": 3604 }, { "epoch": 0.43, "learning_rate": 1.7142857142857143e-06, "loss": 0.9659, "step": 3605 }, { "epoch": 0.43, "learning_rate": 1.7139240506329113e-06, "loss": 0.0003, "step": 3606 }, { "epoch": 0.43, "learning_rate": 1.7135623869801085e-06, "loss": 0.0003, "step": 3607 }, { "epoch": 0.43, "learning_rate": 1.7132007233273055e-06, "loss": 0.0003, "step": 3608 }, { "epoch": 0.43, "learning_rate": 1.7128390596745028e-06, "loss": 0.0003, "step": 3609 }, { "epoch": 0.43, "learning_rate": 1.7124773960216998e-06, "loss": 0.0003, "step": 3610 }, { "epoch": 0.43, "learning_rate": 1.712115732368897e-06, "loss": 0.0004, "step": 3611 }, { "epoch": 0.43, "learning_rate": 1.711754068716094e-06, "loss": 0.4355, "step": 3612 }, { "epoch": 0.43, "learning_rate": 1.7113924050632913e-06, "loss": 0.0003, "step": 3613 }, { "epoch": 0.43, "learning_rate": 1.7110307414104883e-06, "loss": 0.0003, "step": 3614 }, { "epoch": 0.43, "learning_rate": 1.7106690777576855e-06, "loss": 0.0005, "step": 3615 }, { "epoch": 0.43, "learning_rate": 1.7103074141048825e-06, "loss": 0.0005, "step": 3616 }, { "epoch": 0.43, "learning_rate": 1.7099457504520798e-06, "loss": 0.0004, "step": 3617 }, { "epoch": 0.43, "learning_rate": 1.7095840867992768e-06, "loss": 0.0004, "step": 3618 }, { "epoch": 0.43, "learning_rate": 1.709222423146474e-06, "loss": 0.4786, "step": 3619 }, { "epoch": 0.43, "learning_rate": 1.708860759493671e-06, "loss": 0.0004, "step": 3620 }, { "epoch": 0.43, "learning_rate": 1.708499095840868e-06, "loss": 0.0003, "step": 3621 }, { "epoch": 0.43, "learning_rate": 1.7081374321880653e-06, "loss": 0.0009, "step": 3622 }, { "epoch": 0.43, "learning_rate": 1.7077757685352623e-06, "loss": 0.0005, "step": 3623 }, { "epoch": 0.43, "learning_rate": 1.7074141048824591e-06, "loss": 0.0004, "step": 3624 }, { "epoch": 0.43, "learning_rate": 1.7070524412296564e-06, "loss": 0.0004, "step": 3625 }, { "epoch": 0.43, "learning_rate": 1.7066907775768534e-06, "loss": 0.0005, "step": 3626 }, { "epoch": 0.43, "learning_rate": 1.7063291139240506e-06, "loss": 0.0003, "step": 3627 }, { "epoch": 0.43, "learning_rate": 1.7059674502712476e-06, "loss": 0.0004, "step": 3628 }, { "epoch": 0.43, "learning_rate": 1.7056057866184449e-06, "loss": 0.7856, "step": 3629 }, { "epoch": 0.43, "learning_rate": 1.7052441229656419e-06, "loss": 0.0003, "step": 3630 }, { "epoch": 0.44, "learning_rate": 1.704882459312839e-06, "loss": 0.0005, "step": 3631 }, { "epoch": 0.44, "learning_rate": 1.7045207956600361e-06, "loss": 0.0003, "step": 3632 }, { "epoch": 0.44, "learning_rate": 1.7041591320072334e-06, "loss": 0.0004, "step": 3633 }, { "epoch": 0.44, "learning_rate": 1.7037974683544304e-06, "loss": 0.0006, "step": 3634 }, { "epoch": 0.44, "learning_rate": 1.7034358047016276e-06, "loss": 0.0003, "step": 3635 }, { "epoch": 0.44, "learning_rate": 1.7030741410488246e-06, "loss": 0.0004, "step": 3636 }, { "epoch": 0.44, "learning_rate": 1.7027124773960219e-06, "loss": 0.6656, "step": 3637 }, { "epoch": 0.44, "learning_rate": 1.7023508137432189e-06, "loss": 0.0602, "step": 3638 }, { "epoch": 0.44, "learning_rate": 1.701989150090416e-06, "loss": 0.0004, "step": 3639 }, { "epoch": 0.44, "learning_rate": 1.7016274864376131e-06, "loss": 0.4809, "step": 3640 }, { "epoch": 0.44, "learning_rate": 1.7012658227848101e-06, "loss": 0.2632, "step": 3641 }, { "epoch": 0.44, "learning_rate": 1.7009041591320074e-06, "loss": 0.0005, "step": 3642 }, { "epoch": 0.44, "learning_rate": 1.7005424954792044e-06, "loss": 0.0003, "step": 3643 }, { "epoch": 0.44, "learning_rate": 1.7001808318264016e-06, "loss": 0.0057, "step": 3644 }, { "epoch": 0.44, "learning_rate": 1.6998191681735986e-06, "loss": 0.0004, "step": 3645 }, { "epoch": 0.44, "learning_rate": 1.6994575045207959e-06, "loss": 0.0009, "step": 3646 }, { "epoch": 0.44, "learning_rate": 1.699095840867993e-06, "loss": 0.0004, "step": 3647 }, { "epoch": 0.44, "learning_rate": 1.6987341772151897e-06, "loss": 0.001, "step": 3648 }, { "epoch": 0.44, "learning_rate": 1.698372513562387e-06, "loss": 0.001, "step": 3649 }, { "epoch": 0.44, "learning_rate": 1.698010849909584e-06, "loss": 0.0019, "step": 3650 }, { "epoch": 0.44, "learning_rate": 1.6976491862567812e-06, "loss": 0.0004, "step": 3651 }, { "epoch": 0.44, "learning_rate": 1.6972875226039782e-06, "loss": 0.0004, "step": 3652 }, { "epoch": 0.44, "learning_rate": 1.6969258589511754e-06, "loss": 0.6319, "step": 3653 }, { "epoch": 0.44, "learning_rate": 1.6965641952983725e-06, "loss": 0.0003, "step": 3654 }, { "epoch": 0.44, "learning_rate": 1.6962025316455697e-06, "loss": 0.0003, "step": 3655 }, { "epoch": 0.44, "learning_rate": 1.6958408679927667e-06, "loss": 0.0003, "step": 3656 }, { "epoch": 0.44, "learning_rate": 1.695479204339964e-06, "loss": 0.0002, "step": 3657 }, { "epoch": 0.44, "learning_rate": 1.695117540687161e-06, "loss": 0.0005, "step": 3658 }, { "epoch": 0.44, "learning_rate": 1.6947558770343582e-06, "loss": 0.6594, "step": 3659 }, { "epoch": 0.44, "learning_rate": 1.6943942133815552e-06, "loss": 0.9084, "step": 3660 }, { "epoch": 0.44, "learning_rate": 1.6940325497287522e-06, "loss": 0.0003, "step": 3661 }, { "epoch": 0.44, "learning_rate": 1.6936708860759495e-06, "loss": 0.6942, "step": 3662 }, { "epoch": 0.44, "learning_rate": 1.6933092224231465e-06, "loss": 0.0003, "step": 3663 }, { "epoch": 0.44, "learning_rate": 1.6929475587703437e-06, "loss": 0.0003, "step": 3664 }, { "epoch": 0.44, "learning_rate": 1.6925858951175407e-06, "loss": 0.0003, "step": 3665 }, { "epoch": 0.44, "learning_rate": 1.692224231464738e-06, "loss": 0.1232, "step": 3666 }, { "epoch": 0.44, "learning_rate": 1.691862567811935e-06, "loss": 0.0003, "step": 3667 }, { "epoch": 0.44, "learning_rate": 1.6915009041591322e-06, "loss": 0.0007, "step": 3668 }, { "epoch": 0.44, "learning_rate": 1.6911392405063292e-06, "loss": 0.0003, "step": 3669 }, { "epoch": 0.44, "learning_rate": 1.6907775768535265e-06, "loss": 1.0084, "step": 3670 }, { "epoch": 0.44, "learning_rate": 1.6904159132007235e-06, "loss": 0.0003, "step": 3671 }, { "epoch": 0.44, "learning_rate": 1.6900542495479203e-06, "loss": 0.0006, "step": 3672 }, { "epoch": 0.44, "learning_rate": 1.6896925858951175e-06, "loss": 0.0004, "step": 3673 }, { "epoch": 0.44, "learning_rate": 1.6893309222423145e-06, "loss": 0.0012, "step": 3674 }, { "epoch": 0.44, "learning_rate": 1.6889692585895118e-06, "loss": 0.9214, "step": 3675 }, { "epoch": 0.44, "learning_rate": 1.6886075949367088e-06, "loss": 0.6976, "step": 3676 }, { "epoch": 0.44, "learning_rate": 1.688245931283906e-06, "loss": 0.0004, "step": 3677 }, { "epoch": 0.44, "learning_rate": 1.687884267631103e-06, "loss": 0.0004, "step": 3678 }, { "epoch": 0.44, "learning_rate": 1.6875226039783e-06, "loss": 0.0006, "step": 3679 }, { "epoch": 0.44, "learning_rate": 1.6871609403254973e-06, "loss": 0.0005, "step": 3680 }, { "epoch": 0.44, "learning_rate": 1.6867992766726943e-06, "loss": 0.0007, "step": 3681 }, { "epoch": 0.44, "learning_rate": 1.6864376130198915e-06, "loss": 0.0008, "step": 3682 }, { "epoch": 0.44, "learning_rate": 1.6860759493670886e-06, "loss": 0.0003, "step": 3683 }, { "epoch": 0.44, "learning_rate": 1.6857142857142858e-06, "loss": 0.0005, "step": 3684 }, { "epoch": 0.44, "learning_rate": 1.6853526220614828e-06, "loss": 0.5632, "step": 3685 }, { "epoch": 0.44, "learning_rate": 1.68499095840868e-06, "loss": 0.0011, "step": 3686 }, { "epoch": 0.44, "learning_rate": 1.684629294755877e-06, "loss": 0.0003, "step": 3687 }, { "epoch": 0.44, "learning_rate": 1.6842676311030743e-06, "loss": 0.0004, "step": 3688 }, { "epoch": 0.44, "learning_rate": 1.6839059674502713e-06, "loss": 0.0003, "step": 3689 }, { "epoch": 0.44, "learning_rate": 1.6835443037974685e-06, "loss": 0.0007, "step": 3690 }, { "epoch": 0.44, "learning_rate": 1.6831826401446656e-06, "loss": 1.0491, "step": 3691 }, { "epoch": 0.44, "learning_rate": 1.6828209764918628e-06, "loss": 0.0009, "step": 3692 }, { "epoch": 0.44, "learning_rate": 1.6824593128390598e-06, "loss": 0.0003, "step": 3693 }, { "epoch": 0.44, "learning_rate": 1.682097649186257e-06, "loss": 0.0004, "step": 3694 }, { "epoch": 0.44, "learning_rate": 1.681735985533454e-06, "loss": 0.0006, "step": 3695 }, { "epoch": 0.44, "learning_rate": 1.6813743218806509e-06, "loss": 0.0003, "step": 3696 }, { "epoch": 0.44, "learning_rate": 1.681012658227848e-06, "loss": 0.0005, "step": 3697 }, { "epoch": 0.44, "learning_rate": 1.6806509945750451e-06, "loss": 0.0004, "step": 3698 }, { "epoch": 0.44, "learning_rate": 1.6802893309222421e-06, "loss": 0.0005, "step": 3699 }, { "epoch": 0.44, "learning_rate": 1.6799276672694394e-06, "loss": 0.0553, "step": 3700 }, { "epoch": 0.44, "learning_rate": 1.6795660036166364e-06, "loss": 0.001, "step": 3701 }, { "epoch": 0.44, "learning_rate": 1.6792043399638336e-06, "loss": 0.0004, "step": 3702 }, { "epoch": 0.44, "learning_rate": 1.6788426763110306e-06, "loss": 0.0011, "step": 3703 }, { "epoch": 0.44, "learning_rate": 1.6784810126582279e-06, "loss": 0.0003, "step": 3704 }, { "epoch": 0.44, "learning_rate": 1.6781193490054249e-06, "loss": 0.0004, "step": 3705 }, { "epoch": 0.44, "learning_rate": 1.6777576853526221e-06, "loss": 0.0006, "step": 3706 }, { "epoch": 0.44, "learning_rate": 1.6773960216998191e-06, "loss": 0.0152, "step": 3707 }, { "epoch": 0.44, "learning_rate": 1.6770343580470164e-06, "loss": 0.0003, "step": 3708 }, { "epoch": 0.44, "learning_rate": 1.6766726943942134e-06, "loss": 0.0003, "step": 3709 }, { "epoch": 0.44, "learning_rate": 1.6763110307414106e-06, "loss": 0.0003, "step": 3710 }, { "epoch": 0.44, "learning_rate": 1.6759493670886076e-06, "loss": 0.0003, "step": 3711 }, { "epoch": 0.44, "learning_rate": 1.6755877034358049e-06, "loss": 0.0004, "step": 3712 }, { "epoch": 0.44, "learning_rate": 1.6752260397830019e-06, "loss": 0.3975, "step": 3713 }, { "epoch": 0.45, "learning_rate": 1.6748643761301991e-06, "loss": 0.0006, "step": 3714 }, { "epoch": 0.45, "learning_rate": 1.6745027124773961e-06, "loss": 0.0003, "step": 3715 }, { "epoch": 0.45, "learning_rate": 1.6741410488245934e-06, "loss": 0.6014, "step": 3716 }, { "epoch": 0.45, "learning_rate": 1.6737793851717904e-06, "loss": 0.6313, "step": 3717 }, { "epoch": 0.45, "learning_rate": 1.6734177215189876e-06, "loss": 0.0003, "step": 3718 }, { "epoch": 0.45, "learning_rate": 1.6730560578661846e-06, "loss": 0.0004, "step": 3719 }, { "epoch": 0.45, "learning_rate": 1.6726943942133815e-06, "loss": 0.0212, "step": 3720 }, { "epoch": 0.45, "learning_rate": 1.6723327305605785e-06, "loss": 0.0007, "step": 3721 }, { "epoch": 0.45, "learning_rate": 1.6719710669077757e-06, "loss": 0.0003, "step": 3722 }, { "epoch": 0.45, "learning_rate": 1.6716094032549727e-06, "loss": 0.0777, "step": 3723 }, { "epoch": 0.45, "learning_rate": 1.67124773960217e-06, "loss": 0.0004, "step": 3724 }, { "epoch": 0.45, "learning_rate": 1.670886075949367e-06, "loss": 0.9305, "step": 3725 }, { "epoch": 0.45, "learning_rate": 1.6705244122965642e-06, "loss": 0.0004, "step": 3726 }, { "epoch": 0.45, "learning_rate": 1.6701627486437612e-06, "loss": 0.0003, "step": 3727 }, { "epoch": 0.45, "learning_rate": 1.6698010849909585e-06, "loss": 0.0004, "step": 3728 }, { "epoch": 0.45, "learning_rate": 1.6694394213381555e-06, "loss": 0.0004, "step": 3729 }, { "epoch": 0.45, "learning_rate": 1.6690777576853527e-06, "loss": 0.0004, "step": 3730 }, { "epoch": 0.45, "learning_rate": 1.6687160940325497e-06, "loss": 0.0005, "step": 3731 }, { "epoch": 0.45, "learning_rate": 1.668354430379747e-06, "loss": 0.0003, "step": 3732 }, { "epoch": 0.45, "learning_rate": 1.667992766726944e-06, "loss": 0.0007, "step": 3733 }, { "epoch": 0.45, "learning_rate": 1.6676311030741412e-06, "loss": 0.0003, "step": 3734 }, { "epoch": 0.45, "learning_rate": 1.6672694394213382e-06, "loss": 0.0006, "step": 3735 }, { "epoch": 0.45, "learning_rate": 1.6669077757685355e-06, "loss": 0.0004, "step": 3736 }, { "epoch": 0.45, "learning_rate": 1.6665461121157325e-06, "loss": 0.0505, "step": 3737 }, { "epoch": 0.45, "learning_rate": 1.6661844484629297e-06, "loss": 0.0004, "step": 3738 }, { "epoch": 0.45, "learning_rate": 1.6658227848101267e-06, "loss": 0.0006, "step": 3739 }, { "epoch": 0.45, "learning_rate": 1.6654611211573237e-06, "loss": 0.0004, "step": 3740 }, { "epoch": 0.45, "learning_rate": 1.665099457504521e-06, "loss": 0.0004, "step": 3741 }, { "epoch": 0.45, "learning_rate": 1.664737793851718e-06, "loss": 0.2613, "step": 3742 }, { "epoch": 0.45, "learning_rate": 1.6643761301989152e-06, "loss": 0.0003, "step": 3743 }, { "epoch": 0.45, "learning_rate": 1.664014466546112e-06, "loss": 0.0003, "step": 3744 }, { "epoch": 0.45, "learning_rate": 1.663652802893309e-06, "loss": 0.0004, "step": 3745 }, { "epoch": 0.45, "learning_rate": 1.6632911392405063e-06, "loss": 0.0005, "step": 3746 }, { "epoch": 0.45, "learning_rate": 1.6629294755877033e-06, "loss": 0.0006, "step": 3747 }, { "epoch": 0.45, "learning_rate": 1.6625678119349005e-06, "loss": 0.0415, "step": 3748 }, { "epoch": 0.45, "learning_rate": 1.6622061482820976e-06, "loss": 0.0005, "step": 3749 }, { "epoch": 0.45, "learning_rate": 1.6618444846292948e-06, "loss": 0.8911, "step": 3750 }, { "epoch": 0.45, "learning_rate": 1.6614828209764918e-06, "loss": 0.0004, "step": 3751 }, { "epoch": 0.45, "learning_rate": 1.661121157323689e-06, "loss": 0.0007, "step": 3752 }, { "epoch": 0.45, "learning_rate": 1.660759493670886e-06, "loss": 0.0905, "step": 3753 }, { "epoch": 0.45, "learning_rate": 1.6603978300180833e-06, "loss": 0.0003, "step": 3754 }, { "epoch": 0.45, "learning_rate": 1.6600361663652803e-06, "loss": 0.0004, "step": 3755 }, { "epoch": 0.45, "learning_rate": 1.6596745027124775e-06, "loss": 0.0003, "step": 3756 }, { "epoch": 0.45, "learning_rate": 1.6593128390596746e-06, "loss": 0.0004, "step": 3757 }, { "epoch": 0.45, "learning_rate": 1.6589511754068718e-06, "loss": 0.0005, "step": 3758 }, { "epoch": 0.45, "learning_rate": 1.6585895117540688e-06, "loss": 0.0004, "step": 3759 }, { "epoch": 0.45, "learning_rate": 1.6582278481012658e-06, "loss": 0.8224, "step": 3760 }, { "epoch": 0.45, "learning_rate": 1.657866184448463e-06, "loss": 0.912, "step": 3761 }, { "epoch": 0.45, "learning_rate": 1.65750452079566e-06, "loss": 0.0004, "step": 3762 }, { "epoch": 0.45, "learning_rate": 1.6571428571428573e-06, "loss": 0.0006, "step": 3763 }, { "epoch": 0.45, "learning_rate": 1.6567811934900543e-06, "loss": 0.0004, "step": 3764 }, { "epoch": 0.45, "learning_rate": 1.6564195298372516e-06, "loss": 0.0005, "step": 3765 }, { "epoch": 0.45, "learning_rate": 1.6560578661844486e-06, "loss": 0.0003, "step": 3766 }, { "epoch": 0.45, "learning_rate": 1.6556962025316458e-06, "loss": 1.0545, "step": 3767 }, { "epoch": 0.45, "learning_rate": 1.6553345388788426e-06, "loss": 0.0479, "step": 3768 }, { "epoch": 0.45, "learning_rate": 1.6549728752260396e-06, "loss": 0.1062, "step": 3769 }, { "epoch": 0.45, "learning_rate": 1.6546112115732369e-06, "loss": 0.0035, "step": 3770 }, { "epoch": 0.45, "learning_rate": 1.6542495479204339e-06, "loss": 0.0091, "step": 3771 }, { "epoch": 0.45, "learning_rate": 1.6538878842676311e-06, "loss": 0.0004, "step": 3772 }, { "epoch": 0.45, "learning_rate": 1.6535262206148281e-06, "loss": 0.0005, "step": 3773 }, { "epoch": 0.45, "learning_rate": 1.6531645569620254e-06, "loss": 0.0003, "step": 3774 }, { "epoch": 0.45, "learning_rate": 1.6528028933092224e-06, "loss": 0.0009, "step": 3775 }, { "epoch": 0.45, "learning_rate": 1.6524412296564196e-06, "loss": 0.0005, "step": 3776 }, { "epoch": 0.45, "learning_rate": 1.6520795660036166e-06, "loss": 0.0006, "step": 3777 }, { "epoch": 0.45, "learning_rate": 1.6517179023508139e-06, "loss": 0.0008, "step": 3778 }, { "epoch": 0.45, "learning_rate": 1.6513562386980109e-06, "loss": 0.0007, "step": 3779 }, { "epoch": 0.45, "learning_rate": 1.650994575045208e-06, "loss": 0.0089, "step": 3780 }, { "epoch": 0.45, "learning_rate": 1.6506329113924051e-06, "loss": 0.001, "step": 3781 }, { "epoch": 0.45, "learning_rate": 1.6502712477396022e-06, "loss": 0.0006, "step": 3782 }, { "epoch": 0.45, "learning_rate": 1.6499095840867994e-06, "loss": 0.0005, "step": 3783 }, { "epoch": 0.45, "learning_rate": 1.6495479204339964e-06, "loss": 0.0007, "step": 3784 }, { "epoch": 0.45, "learning_rate": 1.6491862567811936e-06, "loss": 0.0009, "step": 3785 }, { "epoch": 0.45, "learning_rate": 1.6488245931283907e-06, "loss": 0.0679, "step": 3786 }, { "epoch": 0.45, "learning_rate": 1.6484629294755879e-06, "loss": 0.0006, "step": 3787 }, { "epoch": 0.45, "learning_rate": 1.648101265822785e-06, "loss": 0.0005, "step": 3788 }, { "epoch": 0.45, "learning_rate": 1.6477396021699821e-06, "loss": 0.0006, "step": 3789 }, { "epoch": 0.45, "learning_rate": 1.6473779385171792e-06, "loss": 0.3645, "step": 3790 }, { "epoch": 0.45, "learning_rate": 1.6470162748643764e-06, "loss": 0.0039, "step": 3791 }, { "epoch": 0.45, "learning_rate": 1.6466546112115732e-06, "loss": 0.0005, "step": 3792 }, { "epoch": 0.45, "learning_rate": 1.6462929475587702e-06, "loss": 0.001, "step": 3793 }, { "epoch": 0.45, "learning_rate": 1.6459312839059675e-06, "loss": 0.0004, "step": 3794 }, { "epoch": 0.45, "learning_rate": 1.6455696202531645e-06, "loss": 0.0004, "step": 3795 }, { "epoch": 0.45, "learning_rate": 1.6452079566003617e-06, "loss": 0.0006, "step": 3796 }, { "epoch": 0.46, "learning_rate": 1.6448462929475587e-06, "loss": 0.0005, "step": 3797 }, { "epoch": 0.46, "learning_rate": 1.644484629294756e-06, "loss": 0.0009, "step": 3798 }, { "epoch": 0.46, "learning_rate": 1.644122965641953e-06, "loss": 0.0004, "step": 3799 }, { "epoch": 0.46, "learning_rate": 1.64376130198915e-06, "loss": 0.002, "step": 3800 }, { "epoch": 0.46, "learning_rate": 1.6433996383363472e-06, "loss": 0.0006, "step": 3801 }, { "epoch": 0.46, "learning_rate": 1.6430379746835442e-06, "loss": 0.0006, "step": 3802 }, { "epoch": 0.46, "learning_rate": 1.6426763110307415e-06, "loss": 0.0006, "step": 3803 }, { "epoch": 0.46, "learning_rate": 1.6423146473779385e-06, "loss": 0.0003, "step": 3804 }, { "epoch": 0.46, "learning_rate": 1.6419529837251357e-06, "loss": 0.0004, "step": 3805 }, { "epoch": 0.46, "learning_rate": 1.6415913200723327e-06, "loss": 0.0004, "step": 3806 }, { "epoch": 0.46, "learning_rate": 1.64122965641953e-06, "loss": 0.0005, "step": 3807 }, { "epoch": 0.46, "learning_rate": 1.640867992766727e-06, "loss": 0.0004, "step": 3808 }, { "epoch": 0.46, "learning_rate": 1.6405063291139242e-06, "loss": 0.0004, "step": 3809 }, { "epoch": 0.46, "learning_rate": 1.6401446654611212e-06, "loss": 0.0008, "step": 3810 }, { "epoch": 0.46, "learning_rate": 1.6397830018083185e-06, "loss": 0.0004, "step": 3811 }, { "epoch": 0.46, "learning_rate": 1.6394213381555155e-06, "loss": 0.0005, "step": 3812 }, { "epoch": 0.46, "learning_rate": 1.6390596745027127e-06, "loss": 0.0003, "step": 3813 }, { "epoch": 0.46, "learning_rate": 1.6386980108499097e-06, "loss": 0.0003, "step": 3814 }, { "epoch": 0.46, "learning_rate": 1.638336347197107e-06, "loss": 0.0004, "step": 3815 }, { "epoch": 0.46, "learning_rate": 1.6379746835443038e-06, "loss": 0.0004, "step": 3816 }, { "epoch": 0.46, "learning_rate": 1.6376130198915008e-06, "loss": 0.0006, "step": 3817 }, { "epoch": 0.46, "learning_rate": 1.6372513562386978e-06, "loss": 0.0013, "step": 3818 }, { "epoch": 0.46, "learning_rate": 1.636889692585895e-06, "loss": 0.0004, "step": 3819 }, { "epoch": 0.46, "learning_rate": 1.636528028933092e-06, "loss": 0.0005, "step": 3820 }, { "epoch": 0.46, "learning_rate": 1.6361663652802893e-06, "loss": 0.0004, "step": 3821 }, { "epoch": 0.46, "learning_rate": 1.6358047016274863e-06, "loss": 0.0106, "step": 3822 }, { "epoch": 0.46, "learning_rate": 1.6354430379746836e-06, "loss": 0.001, "step": 3823 }, { "epoch": 0.46, "learning_rate": 1.6350813743218806e-06, "loss": 0.0004, "step": 3824 }, { "epoch": 0.46, "learning_rate": 1.6347197106690778e-06, "loss": 0.4335, "step": 3825 }, { "epoch": 0.46, "learning_rate": 1.6343580470162748e-06, "loss": 0.0007, "step": 3826 }, { "epoch": 0.46, "learning_rate": 1.633996383363472e-06, "loss": 0.0011, "step": 3827 }, { "epoch": 0.46, "learning_rate": 1.633634719710669e-06, "loss": 0.0003, "step": 3828 }, { "epoch": 0.46, "learning_rate": 1.6332730560578663e-06, "loss": 0.0006, "step": 3829 }, { "epoch": 0.46, "learning_rate": 1.6329113924050633e-06, "loss": 0.0004, "step": 3830 }, { "epoch": 0.46, "learning_rate": 1.6325497287522606e-06, "loss": 0.0004, "step": 3831 }, { "epoch": 0.46, "learning_rate": 1.6321880650994576e-06, "loss": 0.0007, "step": 3832 }, { "epoch": 0.46, "learning_rate": 1.6318264014466548e-06, "loss": 0.0004, "step": 3833 }, { "epoch": 0.46, "learning_rate": 1.6314647377938518e-06, "loss": 0.0004, "step": 3834 }, { "epoch": 0.46, "learning_rate": 1.631103074141049e-06, "loss": 0.0004, "step": 3835 }, { "epoch": 0.46, "learning_rate": 1.630741410488246e-06, "loss": 0.0004, "step": 3836 }, { "epoch": 0.46, "learning_rate": 1.6303797468354433e-06, "loss": 1.3404, "step": 3837 }, { "epoch": 0.46, "learning_rate": 1.6300180831826403e-06, "loss": 0.0017, "step": 3838 }, { "epoch": 0.46, "learning_rate": 1.6296564195298371e-06, "loss": 0.0004, "step": 3839 }, { "epoch": 0.46, "learning_rate": 1.6292947558770342e-06, "loss": 0.0006, "step": 3840 }, { "epoch": 0.46, "learning_rate": 1.6289330922242314e-06, "loss": 0.0002, "step": 3841 }, { "epoch": 0.46, "learning_rate": 1.6285714285714284e-06, "loss": 0.0003, "step": 3842 }, { "epoch": 0.46, "learning_rate": 1.6282097649186256e-06, "loss": 0.0002, "step": 3843 }, { "epoch": 0.46, "learning_rate": 1.6278481012658227e-06, "loss": 0.0004, "step": 3844 }, { "epoch": 0.46, "learning_rate": 1.6274864376130199e-06, "loss": 0.0003, "step": 3845 }, { "epoch": 0.46, "learning_rate": 1.627124773960217e-06, "loss": 0.0004, "step": 3846 }, { "epoch": 0.46, "learning_rate": 1.6267631103074141e-06, "loss": 0.0005, "step": 3847 }, { "epoch": 0.46, "learning_rate": 1.6264014466546112e-06, "loss": 0.0005, "step": 3848 }, { "epoch": 0.46, "learning_rate": 1.6260397830018084e-06, "loss": 0.0004, "step": 3849 }, { "epoch": 0.46, "learning_rate": 1.6256781193490054e-06, "loss": 0.9304, "step": 3850 }, { "epoch": 0.46, "learning_rate": 1.6253164556962026e-06, "loss": 0.0017, "step": 3851 }, { "epoch": 0.46, "learning_rate": 1.6249547920433997e-06, "loss": 0.0004, "step": 3852 }, { "epoch": 0.46, "learning_rate": 1.6245931283905969e-06, "loss": 0.0003, "step": 3853 }, { "epoch": 0.46, "learning_rate": 1.624231464737794e-06, "loss": 0.0061, "step": 3854 }, { "epoch": 0.46, "learning_rate": 1.6238698010849911e-06, "loss": 0.0006, "step": 3855 }, { "epoch": 0.46, "learning_rate": 1.6235081374321882e-06, "loss": 0.0004, "step": 3856 }, { "epoch": 0.46, "learning_rate": 1.6231464737793854e-06, "loss": 0.0005, "step": 3857 }, { "epoch": 0.46, "learning_rate": 1.6227848101265824e-06, "loss": 0.0003, "step": 3858 }, { "epoch": 0.46, "learning_rate": 1.6224231464737796e-06, "loss": 0.0003, "step": 3859 }, { "epoch": 0.46, "learning_rate": 1.6220614828209767e-06, "loss": 0.0006, "step": 3860 }, { "epoch": 0.46, "learning_rate": 1.6216998191681737e-06, "loss": 0.0003, "step": 3861 }, { "epoch": 0.46, "learning_rate": 1.621338155515371e-06, "loss": 0.0005, "step": 3862 }, { "epoch": 0.46, "learning_rate": 1.6209764918625677e-06, "loss": 0.0005, "step": 3863 }, { "epoch": 0.46, "learning_rate": 1.6206148282097647e-06, "loss": 0.7823, "step": 3864 }, { "epoch": 0.46, "learning_rate": 1.620253164556962e-06, "loss": 0.0005, "step": 3865 }, { "epoch": 0.46, "learning_rate": 1.619891500904159e-06, "loss": 0.0006, "step": 3866 }, { "epoch": 0.46, "learning_rate": 1.6195298372513562e-06, "loss": 0.0005, "step": 3867 }, { "epoch": 0.46, "learning_rate": 1.6191681735985532e-06, "loss": 0.0007, "step": 3868 }, { "epoch": 0.46, "learning_rate": 1.6188065099457505e-06, "loss": 0.0006, "step": 3869 }, { "epoch": 0.46, "learning_rate": 1.6184448462929475e-06, "loss": 0.0003, "step": 3870 }, { "epoch": 0.46, "learning_rate": 1.6180831826401447e-06, "loss": 0.0004, "step": 3871 }, { "epoch": 0.46, "learning_rate": 1.6177215189873417e-06, "loss": 0.001, "step": 3872 }, { "epoch": 0.46, "learning_rate": 1.617359855334539e-06, "loss": 0.0009, "step": 3873 }, { "epoch": 0.46, "learning_rate": 1.616998191681736e-06, "loss": 0.0004, "step": 3874 }, { "epoch": 0.46, "learning_rate": 1.6166365280289332e-06, "loss": 0.0066, "step": 3875 }, { "epoch": 0.46, "learning_rate": 1.6162748643761302e-06, "loss": 0.0004, "step": 3876 }, { "epoch": 0.46, "learning_rate": 1.6159132007233275e-06, "loss": 0.0004, "step": 3877 }, { "epoch": 0.46, "learning_rate": 1.6155515370705245e-06, "loss": 0.0004, "step": 3878 }, { "epoch": 0.46, "learning_rate": 1.6151898734177215e-06, "loss": 0.0005, "step": 3879 }, { "epoch": 0.46, "learning_rate": 1.6148282097649187e-06, "loss": 0.0003, "step": 3880 }, { "epoch": 0.47, "learning_rate": 1.6144665461121158e-06, "loss": 0.0004, "step": 3881 }, { "epoch": 0.47, "learning_rate": 1.614104882459313e-06, "loss": 0.0003, "step": 3882 }, { "epoch": 0.47, "learning_rate": 1.61374321880651e-06, "loss": 0.0003, "step": 3883 }, { "epoch": 0.47, "learning_rate": 1.6133815551537072e-06, "loss": 0.0003, "step": 3884 }, { "epoch": 0.47, "learning_rate": 1.6130198915009043e-06, "loss": 0.0006, "step": 3885 }, { "epoch": 0.47, "learning_rate": 1.6126582278481015e-06, "loss": 0.0003, "step": 3886 }, { "epoch": 0.47, "learning_rate": 1.6122965641952983e-06, "loss": 0.0002, "step": 3887 }, { "epoch": 0.47, "learning_rate": 1.6119349005424953e-06, "loss": 0.0003, "step": 3888 }, { "epoch": 0.47, "learning_rate": 1.6115732368896926e-06, "loss": 0.0007, "step": 3889 }, { "epoch": 0.47, "learning_rate": 1.6112115732368896e-06, "loss": 0.001, "step": 3890 }, { "epoch": 0.47, "learning_rate": 1.6108499095840868e-06, "loss": 0.7379, "step": 3891 }, { "epoch": 0.47, "learning_rate": 1.6104882459312838e-06, "loss": 0.0003, "step": 3892 }, { "epoch": 0.47, "learning_rate": 1.610126582278481e-06, "loss": 0.7807, "step": 3893 }, { "epoch": 0.47, "learning_rate": 1.609764918625678e-06, "loss": 0.0002, "step": 3894 }, { "epoch": 0.47, "learning_rate": 1.6094032549728753e-06, "loss": 0.0005, "step": 3895 }, { "epoch": 0.47, "learning_rate": 1.6090415913200723e-06, "loss": 0.0003, "step": 3896 }, { "epoch": 0.47, "learning_rate": 1.6086799276672696e-06, "loss": 0.0003, "step": 3897 }, { "epoch": 0.47, "learning_rate": 1.6083182640144666e-06, "loss": 0.8143, "step": 3898 }, { "epoch": 0.47, "learning_rate": 1.6079566003616636e-06, "loss": 0.0005, "step": 3899 }, { "epoch": 0.47, "learning_rate": 1.6075949367088608e-06, "loss": 0.0003, "step": 3900 }, { "epoch": 0.47, "learning_rate": 1.6072332730560578e-06, "loss": 0.0844, "step": 3901 }, { "epoch": 0.47, "learning_rate": 1.606871609403255e-06, "loss": 0.8561, "step": 3902 }, { "epoch": 0.47, "learning_rate": 1.606509945750452e-06, "loss": 0.0003, "step": 3903 }, { "epoch": 0.47, "learning_rate": 1.6061482820976493e-06, "loss": 0.6897, "step": 3904 }, { "epoch": 0.47, "learning_rate": 1.6057866184448463e-06, "loss": 0.0003, "step": 3905 }, { "epoch": 0.47, "learning_rate": 1.6054249547920436e-06, "loss": 0.004, "step": 3906 }, { "epoch": 0.47, "learning_rate": 1.6050632911392406e-06, "loss": 0.0008, "step": 3907 }, { "epoch": 0.47, "learning_rate": 1.6047016274864378e-06, "loss": 0.0003, "step": 3908 }, { "epoch": 0.47, "learning_rate": 1.6043399638336348e-06, "loss": 0.0003, "step": 3909 }, { "epoch": 0.47, "learning_rate": 1.603978300180832e-06, "loss": 0.0038, "step": 3910 }, { "epoch": 0.47, "learning_rate": 1.6036166365280289e-06, "loss": 0.6809, "step": 3911 }, { "epoch": 0.47, "learning_rate": 1.603254972875226e-06, "loss": 0.0008, "step": 3912 }, { "epoch": 0.47, "learning_rate": 1.6028933092224231e-06, "loss": 0.0004, "step": 3913 }, { "epoch": 0.47, "learning_rate": 1.6025316455696202e-06, "loss": 0.0003, "step": 3914 }, { "epoch": 0.47, "learning_rate": 1.6021699819168174e-06, "loss": 0.0003, "step": 3915 }, { "epoch": 0.47, "learning_rate": 1.6018083182640144e-06, "loss": 0.0004, "step": 3916 }, { "epoch": 0.47, "learning_rate": 1.6014466546112116e-06, "loss": 0.0003, "step": 3917 }, { "epoch": 0.47, "learning_rate": 1.6010849909584087e-06, "loss": 0.0003, "step": 3918 }, { "epoch": 0.47, "learning_rate": 1.6007233273056057e-06, "loss": 0.0265, "step": 3919 }, { "epoch": 0.47, "learning_rate": 1.600361663652803e-06, "loss": 0.0002, "step": 3920 }, { "epoch": 0.47, "learning_rate": 1.6e-06, "loss": 0.0003, "step": 3921 }, { "epoch": 0.47, "learning_rate": 1.5996383363471972e-06, "loss": 0.0043, "step": 3922 }, { "epoch": 0.47, "learning_rate": 1.5992766726943942e-06, "loss": 0.0004, "step": 3923 }, { "epoch": 0.47, "learning_rate": 1.5989150090415914e-06, "loss": 0.3927, "step": 3924 }, { "epoch": 0.47, "learning_rate": 1.5985533453887884e-06, "loss": 0.0002, "step": 3925 }, { "epoch": 0.47, "learning_rate": 1.5981916817359857e-06, "loss": 0.9368, "step": 3926 }, { "epoch": 0.47, "learning_rate": 1.5978300180831827e-06, "loss": 0.0003, "step": 3927 }, { "epoch": 0.47, "learning_rate": 1.59746835443038e-06, "loss": 0.0003, "step": 3928 }, { "epoch": 0.47, "learning_rate": 1.597106690777577e-06, "loss": 0.0004, "step": 3929 }, { "epoch": 0.47, "learning_rate": 1.5967450271247742e-06, "loss": 0.0003, "step": 3930 }, { "epoch": 0.47, "learning_rate": 1.5963833634719712e-06, "loss": 0.0003, "step": 3931 }, { "epoch": 0.47, "learning_rate": 1.5960216998191684e-06, "loss": 0.0028, "step": 3932 }, { "epoch": 0.47, "learning_rate": 1.5956600361663654e-06, "loss": 0.0004, "step": 3933 }, { "epoch": 0.47, "learning_rate": 1.5952983725135627e-06, "loss": 0.0003, "step": 3934 }, { "epoch": 0.47, "learning_rate": 1.5949367088607595e-06, "loss": 0.0008, "step": 3935 }, { "epoch": 0.47, "learning_rate": 1.5945750452079565e-06, "loss": 0.0003, "step": 3936 }, { "epoch": 0.47, "learning_rate": 1.5942133815551537e-06, "loss": 0.0003, "step": 3937 }, { "epoch": 0.47, "learning_rate": 1.5938517179023507e-06, "loss": 0.0003, "step": 3938 }, { "epoch": 0.47, "learning_rate": 1.5934900542495478e-06, "loss": 0.0005, "step": 3939 }, { "epoch": 0.47, "learning_rate": 1.593128390596745e-06, "loss": 0.0003, "step": 3940 }, { "epoch": 0.47, "learning_rate": 1.592766726943942e-06, "loss": 0.6895, "step": 3941 }, { "epoch": 0.47, "learning_rate": 1.5924050632911392e-06, "loss": 0.0004, "step": 3942 }, { "epoch": 0.47, "learning_rate": 1.5920433996383363e-06, "loss": 0.0004, "step": 3943 }, { "epoch": 0.47, "learning_rate": 1.5916817359855335e-06, "loss": 0.0002, "step": 3944 }, { "epoch": 0.47, "learning_rate": 1.5913200723327305e-06, "loss": 0.0004, "step": 3945 }, { "epoch": 0.47, "learning_rate": 1.5909584086799277e-06, "loss": 0.0003, "step": 3946 }, { "epoch": 0.47, "learning_rate": 1.5905967450271248e-06, "loss": 0.0003, "step": 3947 }, { "epoch": 0.47, "learning_rate": 1.590235081374322e-06, "loss": 0.0005, "step": 3948 }, { "epoch": 0.47, "learning_rate": 1.589873417721519e-06, "loss": 0.0004, "step": 3949 }, { "epoch": 0.47, "learning_rate": 1.5895117540687162e-06, "loss": 0.0002, "step": 3950 }, { "epoch": 0.47, "learning_rate": 1.5891500904159133e-06, "loss": 0.0003, "step": 3951 }, { "epoch": 0.47, "learning_rate": 1.5887884267631105e-06, "loss": 0.0003, "step": 3952 }, { "epoch": 0.47, "learning_rate": 1.5884267631103075e-06, "loss": 0.0005, "step": 3953 }, { "epoch": 0.47, "learning_rate": 1.5880650994575047e-06, "loss": 0.0004, "step": 3954 }, { "epoch": 0.47, "learning_rate": 1.5877034358047018e-06, "loss": 0.0013, "step": 3955 }, { "epoch": 0.47, "learning_rate": 1.587341772151899e-06, "loss": 0.0004, "step": 3956 }, { "epoch": 0.47, "learning_rate": 1.586980108499096e-06, "loss": 0.5796, "step": 3957 }, { "epoch": 0.47, "learning_rate": 1.5866184448462932e-06, "loss": 0.0003, "step": 3958 }, { "epoch": 0.47, "learning_rate": 1.5862567811934898e-06, "loss": 0.0008, "step": 3959 }, { "epoch": 0.47, "learning_rate": 1.585895117540687e-06, "loss": 0.0003, "step": 3960 }, { "epoch": 0.47, "learning_rate": 1.585533453887884e-06, "loss": 0.0004, "step": 3961 }, { "epoch": 0.47, "learning_rate": 1.5851717902350813e-06, "loss": 0.0003, "step": 3962 }, { "epoch": 0.47, "learning_rate": 1.5848101265822783e-06, "loss": 0.9003, "step": 3963 }, { "epoch": 0.48, "learning_rate": 1.5844484629294756e-06, "loss": 0.0007, "step": 3964 }, { "epoch": 0.48, "learning_rate": 1.5840867992766726e-06, "loss": 0.0057, "step": 3965 }, { "epoch": 0.48, "learning_rate": 1.5837251356238698e-06, "loss": 0.0003, "step": 3966 }, { "epoch": 0.48, "learning_rate": 1.5833634719710668e-06, "loss": 0.0002, "step": 3967 }, { "epoch": 0.48, "learning_rate": 1.583001808318264e-06, "loss": 0.0003, "step": 3968 }, { "epoch": 0.48, "learning_rate": 1.582640144665461e-06, "loss": 0.0005, "step": 3969 }, { "epoch": 0.48, "learning_rate": 1.5822784810126583e-06, "loss": 0.0004, "step": 3970 }, { "epoch": 0.48, "learning_rate": 1.5819168173598553e-06, "loss": 0.0007, "step": 3971 }, { "epoch": 0.48, "learning_rate": 1.5815551537070526e-06, "loss": 0.0004, "step": 3972 }, { "epoch": 0.48, "learning_rate": 1.5811934900542496e-06, "loss": 0.7521, "step": 3973 }, { "epoch": 0.48, "learning_rate": 1.5808318264014468e-06, "loss": 0.0002, "step": 3974 }, { "epoch": 0.48, "learning_rate": 1.5804701627486438e-06, "loss": 0.0003, "step": 3975 }, { "epoch": 0.48, "learning_rate": 1.580108499095841e-06, "loss": 0.0003, "step": 3976 }, { "epoch": 0.48, "learning_rate": 1.579746835443038e-06, "loss": 0.6951, "step": 3977 }, { "epoch": 0.48, "learning_rate": 1.5793851717902353e-06, "loss": 0.0008, "step": 3978 }, { "epoch": 0.48, "learning_rate": 1.5790235081374323e-06, "loss": 0.0003, "step": 3979 }, { "epoch": 0.48, "learning_rate": 1.5786618444846294e-06, "loss": 0.0003, "step": 3980 }, { "epoch": 0.48, "learning_rate": 1.5783001808318266e-06, "loss": 0.0038, "step": 3981 }, { "epoch": 0.48, "learning_rate": 1.5779385171790236e-06, "loss": 1.128, "step": 3982 }, { "epoch": 0.48, "learning_rate": 1.5775768535262204e-06, "loss": 0.0005, "step": 3983 }, { "epoch": 0.48, "learning_rate": 1.5772151898734177e-06, "loss": 0.0014, "step": 3984 }, { "epoch": 0.48, "learning_rate": 1.5768535262206147e-06, "loss": 0.1858, "step": 3985 }, { "epoch": 0.48, "learning_rate": 1.576491862567812e-06, "loss": 0.0003, "step": 3986 }, { "epoch": 0.48, "learning_rate": 1.576130198915009e-06, "loss": 0.0003, "step": 3987 }, { "epoch": 0.48, "learning_rate": 1.5757685352622062e-06, "loss": 0.0006, "step": 3988 }, { "epoch": 0.48, "learning_rate": 1.5754068716094032e-06, "loss": 0.0003, "step": 3989 }, { "epoch": 0.48, "learning_rate": 1.5750452079566004e-06, "loss": 0.0011, "step": 3990 }, { "epoch": 0.48, "learning_rate": 1.5746835443037974e-06, "loss": 0.0003, "step": 3991 }, { "epoch": 0.48, "learning_rate": 1.5743218806509947e-06, "loss": 0.0002, "step": 3992 }, { "epoch": 0.48, "learning_rate": 1.5739602169981917e-06, "loss": 0.0003, "step": 3993 }, { "epoch": 0.48, "learning_rate": 1.573598553345389e-06, "loss": 0.0002, "step": 3994 }, { "epoch": 0.48, "learning_rate": 1.573236889692586e-06, "loss": 0.0005, "step": 3995 }, { "epoch": 0.48, "learning_rate": 1.5728752260397832e-06, "loss": 0.0004, "step": 3996 }, { "epoch": 0.48, "learning_rate": 1.5725135623869802e-06, "loss": 0.0002, "step": 3997 }, { "epoch": 0.48, "learning_rate": 1.5721518987341774e-06, "loss": 0.0008, "step": 3998 }, { "epoch": 0.48, "learning_rate": 1.5717902350813744e-06, "loss": 0.0003, "step": 3999 }, { "epoch": 0.48, "learning_rate": 1.5714285714285714e-06, "loss": 0.0003, "step": 4000 }, { "epoch": 0.48, "learning_rate": 1.5710669077757687e-06, "loss": 0.0003, "step": 4001 }, { "epoch": 0.48, "learning_rate": 1.5707052441229657e-06, "loss": 0.0003, "step": 4002 }, { "epoch": 0.48, "learning_rate": 1.570343580470163e-06, "loss": 0.0004, "step": 4003 }, { "epoch": 0.48, "learning_rate": 1.56998191681736e-06, "loss": 0.0002, "step": 4004 }, { "epoch": 0.48, "learning_rate": 1.5696202531645572e-06, "loss": 0.0326, "step": 4005 }, { "epoch": 0.48, "learning_rate": 1.5692585895117542e-06, "loss": 0.0008, "step": 4006 }, { "epoch": 0.48, "learning_rate": 1.568896925858951e-06, "loss": 0.0002, "step": 4007 }, { "epoch": 0.48, "learning_rate": 1.5685352622061482e-06, "loss": 0.0003, "step": 4008 }, { "epoch": 0.48, "learning_rate": 1.5681735985533453e-06, "loss": 0.0003, "step": 4009 }, { "epoch": 0.48, "learning_rate": 1.5678119349005425e-06, "loss": 0.0003, "step": 4010 }, { "epoch": 0.48, "learning_rate": 1.5674502712477395e-06, "loss": 0.0003, "step": 4011 }, { "epoch": 0.48, "learning_rate": 1.5670886075949367e-06, "loss": 0.0002, "step": 4012 }, { "epoch": 0.48, "learning_rate": 1.5667269439421338e-06, "loss": 0.0003, "step": 4013 }, { "epoch": 0.48, "learning_rate": 1.566365280289331e-06, "loss": 0.0009, "step": 4014 }, { "epoch": 0.48, "learning_rate": 1.566003616636528e-06, "loss": 0.7457, "step": 4015 }, { "epoch": 0.48, "learning_rate": 1.5656419529837252e-06, "loss": 0.0003, "step": 4016 }, { "epoch": 0.48, "learning_rate": 1.5652802893309223e-06, "loss": 0.0003, "step": 4017 }, { "epoch": 0.48, "learning_rate": 1.5649186256781195e-06, "loss": 0.0003, "step": 4018 }, { "epoch": 0.48, "learning_rate": 1.5645569620253165e-06, "loss": 0.0002, "step": 4019 }, { "epoch": 0.48, "learning_rate": 1.5641952983725135e-06, "loss": 0.0027, "step": 4020 }, { "epoch": 0.48, "learning_rate": 1.5638336347197108e-06, "loss": 0.0003, "step": 4021 }, { "epoch": 0.48, "learning_rate": 1.5634719710669078e-06, "loss": 0.0003, "step": 4022 }, { "epoch": 0.48, "learning_rate": 1.563110307414105e-06, "loss": 0.0003, "step": 4023 }, { "epoch": 0.48, "learning_rate": 1.562748643761302e-06, "loss": 0.0002, "step": 4024 }, { "epoch": 0.48, "learning_rate": 1.5623869801084993e-06, "loss": 0.0003, "step": 4025 }, { "epoch": 0.48, "learning_rate": 1.5620253164556963e-06, "loss": 0.0004, "step": 4026 }, { "epoch": 0.48, "learning_rate": 1.5616636528028935e-06, "loss": 0.0003, "step": 4027 }, { "epoch": 0.48, "learning_rate": 1.5613019891500905e-06, "loss": 0.0003, "step": 4028 }, { "epoch": 0.48, "learning_rate": 1.5609403254972878e-06, "loss": 0.0003, "step": 4029 }, { "epoch": 0.48, "learning_rate": 1.5605786618444848e-06, "loss": 0.0004, "step": 4030 }, { "epoch": 0.48, "learning_rate": 1.5602169981916816e-06, "loss": 0.8112, "step": 4031 }, { "epoch": 0.48, "learning_rate": 1.5598553345388788e-06, "loss": 0.0089, "step": 4032 }, { "epoch": 0.48, "learning_rate": 1.5594936708860758e-06, "loss": 0.0004, "step": 4033 }, { "epoch": 0.48, "learning_rate": 1.559132007233273e-06, "loss": 0.7439, "step": 4034 }, { "epoch": 0.48, "learning_rate": 1.55877034358047e-06, "loss": 0.0003, "step": 4035 }, { "epoch": 0.48, "learning_rate": 1.5584086799276673e-06, "loss": 0.6874, "step": 4036 }, { "epoch": 0.48, "learning_rate": 1.5580470162748643e-06, "loss": 0.0003, "step": 4037 }, { "epoch": 0.48, "learning_rate": 1.5576853526220614e-06, "loss": 0.0003, "step": 4038 }, { "epoch": 0.48, "learning_rate": 1.5573236889692586e-06, "loss": 0.0004, "step": 4039 }, { "epoch": 0.48, "learning_rate": 1.5569620253164556e-06, "loss": 0.0006, "step": 4040 }, { "epoch": 0.48, "learning_rate": 1.5566003616636528e-06, "loss": 0.9096, "step": 4041 }, { "epoch": 0.48, "learning_rate": 1.5562386980108499e-06, "loss": 0.0003, "step": 4042 }, { "epoch": 0.48, "learning_rate": 1.555877034358047e-06, "loss": 0.0003, "step": 4043 }, { "epoch": 0.48, "learning_rate": 1.5555153707052441e-06, "loss": 0.0003, "step": 4044 }, { "epoch": 0.48, "learning_rate": 1.5551537070524413e-06, "loss": 0.0003, "step": 4045 }, { "epoch": 0.48, "learning_rate": 1.5547920433996384e-06, "loss": 0.0004, "step": 4046 }, { "epoch": 0.48, "learning_rate": 1.5544303797468356e-06, "loss": 0.0003, "step": 4047 }, { "epoch": 0.49, "learning_rate": 1.5540687160940326e-06, "loss": 0.0003, "step": 4048 }, { "epoch": 0.49, "learning_rate": 1.5537070524412298e-06, "loss": 0.0003, "step": 4049 }, { "epoch": 0.49, "learning_rate": 1.5533453887884269e-06, "loss": 0.0004, "step": 4050 }, { "epoch": 0.49, "learning_rate": 1.552983725135624e-06, "loss": 0.0009, "step": 4051 }, { "epoch": 0.49, "learning_rate": 1.5526220614828211e-06, "loss": 0.5165, "step": 4052 }, { "epoch": 0.49, "learning_rate": 1.5522603978300183e-06, "loss": 0.0003, "step": 4053 }, { "epoch": 0.49, "learning_rate": 1.5518987341772154e-06, "loss": 0.0008, "step": 4054 }, { "epoch": 0.49, "learning_rate": 1.5515370705244122e-06, "loss": 0.0004, "step": 4055 }, { "epoch": 0.49, "learning_rate": 1.5511754068716094e-06, "loss": 0.0003, "step": 4056 }, { "epoch": 0.49, "learning_rate": 1.5508137432188064e-06, "loss": 0.0003, "step": 4057 }, { "epoch": 0.49, "learning_rate": 1.5504520795660034e-06, "loss": 0.0002, "step": 4058 }, { "epoch": 0.49, "learning_rate": 1.5500904159132007e-06, "loss": 0.0003, "step": 4059 }, { "epoch": 0.49, "learning_rate": 1.5497287522603977e-06, "loss": 0.0009, "step": 4060 }, { "epoch": 0.49, "learning_rate": 1.549367088607595e-06, "loss": 0.0003, "step": 4061 }, { "epoch": 0.49, "learning_rate": 1.549005424954792e-06, "loss": 0.0003, "step": 4062 }, { "epoch": 0.49, "learning_rate": 1.5486437613019892e-06, "loss": 0.0002, "step": 4063 }, { "epoch": 0.49, "learning_rate": 1.5482820976491862e-06, "loss": 0.0002, "step": 4064 }, { "epoch": 0.49, "learning_rate": 1.5479204339963834e-06, "loss": 0.2904, "step": 4065 }, { "epoch": 0.49, "learning_rate": 1.5475587703435804e-06, "loss": 0.0003, "step": 4066 }, { "epoch": 0.49, "learning_rate": 1.5471971066907777e-06, "loss": 0.0002, "step": 4067 }, { "epoch": 0.49, "learning_rate": 1.5468354430379747e-06, "loss": 0.0007, "step": 4068 }, { "epoch": 0.49, "learning_rate": 1.546473779385172e-06, "loss": 0.0002, "step": 4069 }, { "epoch": 0.49, "learning_rate": 1.546112115732369e-06, "loss": 0.4826, "step": 4070 }, { "epoch": 0.49, "learning_rate": 1.5457504520795662e-06, "loss": 0.0003, "step": 4071 }, { "epoch": 0.49, "learning_rate": 1.5453887884267632e-06, "loss": 0.0005, "step": 4072 }, { "epoch": 0.49, "learning_rate": 1.5450271247739604e-06, "loss": 1.1352, "step": 4073 }, { "epoch": 0.49, "learning_rate": 1.5446654611211574e-06, "loss": 0.0003, "step": 4074 }, { "epoch": 0.49, "learning_rate": 1.5443037974683547e-06, "loss": 0.0002, "step": 4075 }, { "epoch": 0.49, "learning_rate": 1.5439421338155517e-06, "loss": 0.0003, "step": 4076 }, { "epoch": 0.49, "learning_rate": 1.543580470162749e-06, "loss": 0.0002, "step": 4077 }, { "epoch": 0.49, "learning_rate": 1.5432188065099455e-06, "loss": 0.0003, "step": 4078 }, { "epoch": 0.49, "learning_rate": 1.5428571428571428e-06, "loss": 0.0003, "step": 4079 }, { "epoch": 0.49, "learning_rate": 1.5424954792043398e-06, "loss": 0.1524, "step": 4080 }, { "epoch": 0.49, "learning_rate": 1.542133815551537e-06, "loss": 0.0003, "step": 4081 }, { "epoch": 0.49, "learning_rate": 1.541772151898734e-06, "loss": 0.0208, "step": 4082 }, { "epoch": 0.49, "learning_rate": 1.5414104882459313e-06, "loss": 0.0011, "step": 4083 }, { "epoch": 0.49, "learning_rate": 1.5410488245931283e-06, "loss": 0.0003, "step": 4084 }, { "epoch": 0.49, "learning_rate": 1.5406871609403255e-06, "loss": 0.001, "step": 4085 }, { "epoch": 0.49, "learning_rate": 1.5403254972875225e-06, "loss": 0.0003, "step": 4086 }, { "epoch": 0.49, "learning_rate": 1.5399638336347198e-06, "loss": 0.0003, "step": 4087 }, { "epoch": 0.49, "learning_rate": 1.5396021699819168e-06, "loss": 0.0005, "step": 4088 }, { "epoch": 0.49, "learning_rate": 1.539240506329114e-06, "loss": 0.0004, "step": 4089 }, { "epoch": 0.49, "learning_rate": 1.538878842676311e-06, "loss": 0.0008, "step": 4090 }, { "epoch": 0.49, "learning_rate": 1.5385171790235083e-06, "loss": 0.0008, "step": 4091 }, { "epoch": 0.49, "learning_rate": 1.5381555153707053e-06, "loss": 0.0003, "step": 4092 }, { "epoch": 0.49, "learning_rate": 1.5377938517179025e-06, "loss": 0.0003, "step": 4093 }, { "epoch": 0.49, "learning_rate": 1.5374321880650995e-06, "loss": 0.0004, "step": 4094 }, { "epoch": 0.49, "learning_rate": 1.5370705244122968e-06, "loss": 0.0009, "step": 4095 }, { "epoch": 0.49, "learning_rate": 1.5367088607594938e-06, "loss": 0.0003, "step": 4096 }, { "epoch": 0.49, "learning_rate": 1.536347197106691e-06, "loss": 0.0003, "step": 4097 }, { "epoch": 0.49, "learning_rate": 1.535985533453888e-06, "loss": 0.0003, "step": 4098 }, { "epoch": 0.49, "learning_rate": 1.535623869801085e-06, "loss": 0.0005, "step": 4099 }, { "epoch": 0.49, "learning_rate": 1.5352622061482823e-06, "loss": 0.0007, "step": 4100 }, { "epoch": 0.49, "learning_rate": 1.5349005424954793e-06, "loss": 0.0004, "step": 4101 }, { "epoch": 0.49, "learning_rate": 1.5345388788426761e-06, "loss": 0.0003, "step": 4102 }, { "epoch": 0.49, "learning_rate": 1.5341772151898733e-06, "loss": 0.0053, "step": 4103 }, { "epoch": 0.49, "learning_rate": 1.5338155515370704e-06, "loss": 0.4938, "step": 4104 }, { "epoch": 0.49, "learning_rate": 1.5334538878842676e-06, "loss": 0.0003, "step": 4105 }, { "epoch": 0.49, "learning_rate": 1.5330922242314646e-06, "loss": 0.0006, "step": 4106 }, { "epoch": 0.49, "learning_rate": 1.5327305605786618e-06, "loss": 0.0031, "step": 4107 }, { "epoch": 0.49, "learning_rate": 1.5323688969258589e-06, "loss": 0.0003, "step": 4108 }, { "epoch": 0.49, "learning_rate": 1.532007233273056e-06, "loss": 0.3132, "step": 4109 }, { "epoch": 0.49, "learning_rate": 1.5316455696202531e-06, "loss": 0.0004, "step": 4110 }, { "epoch": 0.49, "learning_rate": 1.5312839059674503e-06, "loss": 0.0003, "step": 4111 }, { "epoch": 0.49, "learning_rate": 1.5309222423146474e-06, "loss": 0.0003, "step": 4112 }, { "epoch": 0.49, "learning_rate": 1.5305605786618446e-06, "loss": 0.0003, "step": 4113 }, { "epoch": 0.49, "learning_rate": 1.5301989150090416e-06, "loss": 0.001, "step": 4114 }, { "epoch": 0.49, "learning_rate": 1.5298372513562388e-06, "loss": 0.0004, "step": 4115 }, { "epoch": 0.49, "learning_rate": 1.5294755877034359e-06, "loss": 0.0002, "step": 4116 }, { "epoch": 0.49, "learning_rate": 1.529113924050633e-06, "loss": 0.0003, "step": 4117 }, { "epoch": 0.49, "learning_rate": 1.5287522603978301e-06, "loss": 0.0005, "step": 4118 }, { "epoch": 0.49, "learning_rate": 1.5283905967450271e-06, "loss": 0.0004, "step": 4119 }, { "epoch": 0.49, "learning_rate": 1.5280289330922244e-06, "loss": 0.0002, "step": 4120 }, { "epoch": 0.49, "learning_rate": 1.5276672694394214e-06, "loss": 0.0003, "step": 4121 }, { "epoch": 0.49, "learning_rate": 1.5273056057866186e-06, "loss": 0.9441, "step": 4122 }, { "epoch": 0.49, "learning_rate": 1.5269439421338156e-06, "loss": 0.0003, "step": 4123 }, { "epoch": 0.49, "learning_rate": 1.5265822784810129e-06, "loss": 0.0002, "step": 4124 }, { "epoch": 0.49, "learning_rate": 1.5262206148282099e-06, "loss": 0.0011, "step": 4125 }, { "epoch": 0.49, "learning_rate": 1.5258589511754067e-06, "loss": 0.002, "step": 4126 }, { "epoch": 0.49, "learning_rate": 1.525497287522604e-06, "loss": 0.0003, "step": 4127 }, { "epoch": 0.49, "learning_rate": 1.525135623869801e-06, "loss": 0.0673, "step": 4128 }, { "epoch": 0.49, "learning_rate": 1.5247739602169982e-06, "loss": 0.001, "step": 4129 }, { "epoch": 0.49, "learning_rate": 1.5244122965641952e-06, "loss": 0.0003, "step": 4130 }, { "epoch": 0.5, "learning_rate": 1.5240506329113924e-06, "loss": 0.0066, "step": 4131 }, { "epoch": 0.5, "learning_rate": 1.5236889692585894e-06, "loss": 0.8411, "step": 4132 }, { "epoch": 0.5, "learning_rate": 1.5233273056057867e-06, "loss": 0.0003, "step": 4133 }, { "epoch": 0.5, "learning_rate": 1.5229656419529837e-06, "loss": 0.0004, "step": 4134 }, { "epoch": 0.5, "learning_rate": 1.522603978300181e-06, "loss": 0.0004, "step": 4135 }, { "epoch": 0.5, "learning_rate": 1.522242314647378e-06, "loss": 0.1411, "step": 4136 }, { "epoch": 0.5, "learning_rate": 1.5218806509945752e-06, "loss": 0.0004, "step": 4137 }, { "epoch": 0.5, "learning_rate": 1.5215189873417722e-06, "loss": 0.0003, "step": 4138 }, { "epoch": 0.5, "learning_rate": 1.5211573236889692e-06, "loss": 0.4919, "step": 4139 }, { "epoch": 0.5, "learning_rate": 1.5207956600361664e-06, "loss": 0.0005, "step": 4140 }, { "epoch": 0.5, "learning_rate": 1.5204339963833635e-06, "loss": 0.0004, "step": 4141 }, { "epoch": 0.5, "learning_rate": 1.5200723327305607e-06, "loss": 0.0004, "step": 4142 }, { "epoch": 0.5, "learning_rate": 1.5197106690777577e-06, "loss": 0.0003, "step": 4143 }, { "epoch": 0.5, "learning_rate": 1.519349005424955e-06, "loss": 0.0003, "step": 4144 }, { "epoch": 0.5, "learning_rate": 1.518987341772152e-06, "loss": 0.045, "step": 4145 }, { "epoch": 0.5, "learning_rate": 1.5186256781193492e-06, "loss": 0.0003, "step": 4146 }, { "epoch": 0.5, "learning_rate": 1.5182640144665462e-06, "loss": 0.0006, "step": 4147 }, { "epoch": 0.5, "learning_rate": 1.5179023508137434e-06, "loss": 1.0016, "step": 4148 }, { "epoch": 0.5, "learning_rate": 1.5175406871609405e-06, "loss": 0.0011, "step": 4149 }, { "epoch": 0.5, "learning_rate": 1.5171790235081373e-06, "loss": 0.0004, "step": 4150 }, { "epoch": 0.5, "learning_rate": 1.5168173598553345e-06, "loss": 0.0004, "step": 4151 }, { "epoch": 0.5, "learning_rate": 1.5164556962025315e-06, "loss": 0.0005, "step": 4152 }, { "epoch": 0.5, "learning_rate": 1.5160940325497288e-06, "loss": 0.0004, "step": 4153 }, { "epoch": 0.5, "learning_rate": 1.5157323688969258e-06, "loss": 0.0006, "step": 4154 }, { "epoch": 0.5, "learning_rate": 1.515370705244123e-06, "loss": 0.0005, "step": 4155 }, { "epoch": 0.5, "learning_rate": 1.51500904159132e-06, "loss": 0.0005, "step": 4156 }, { "epoch": 0.5, "learning_rate": 1.5146473779385173e-06, "loss": 0.0002, "step": 4157 }, { "epoch": 0.5, "learning_rate": 1.5142857142857143e-06, "loss": 0.0005, "step": 4158 }, { "epoch": 0.5, "learning_rate": 1.5139240506329113e-06, "loss": 0.0005, "step": 4159 }, { "epoch": 0.5, "learning_rate": 1.5135623869801085e-06, "loss": 0.0006, "step": 4160 }, { "epoch": 0.5, "learning_rate": 1.5132007233273055e-06, "loss": 0.0005, "step": 4161 }, { "epoch": 0.5, "learning_rate": 1.5128390596745028e-06, "loss": 0.0005, "step": 4162 }, { "epoch": 0.5, "learning_rate": 1.5124773960216998e-06, "loss": 0.0007, "step": 4163 }, { "epoch": 0.5, "learning_rate": 1.512115732368897e-06, "loss": 0.0003, "step": 4164 }, { "epoch": 0.5, "learning_rate": 1.511754068716094e-06, "loss": 0.0062, "step": 4165 }, { "epoch": 0.5, "learning_rate": 1.5113924050632913e-06, "loss": 0.0006, "step": 4166 }, { "epoch": 0.5, "learning_rate": 1.5110307414104883e-06, "loss": 0.0005, "step": 4167 }, { "epoch": 0.5, "learning_rate": 1.5106690777576855e-06, "loss": 0.0008, "step": 4168 }, { "epoch": 0.5, "learning_rate": 1.5103074141048825e-06, "loss": 0.0003, "step": 4169 }, { "epoch": 0.5, "learning_rate": 1.5099457504520798e-06, "loss": 0.0007, "step": 4170 }, { "epoch": 0.5, "learning_rate": 1.5095840867992768e-06, "loss": 0.0014, "step": 4171 }, { "epoch": 0.5, "learning_rate": 1.509222423146474e-06, "loss": 0.0145, "step": 4172 }, { "epoch": 0.5, "learning_rate": 1.508860759493671e-06, "loss": 0.0005, "step": 4173 }, { "epoch": 0.5, "learning_rate": 1.5084990958408679e-06, "loss": 0.0003, "step": 4174 }, { "epoch": 0.5, "learning_rate": 1.508137432188065e-06, "loss": 0.0003, "step": 4175 }, { "epoch": 0.5, "learning_rate": 1.5077757685352621e-06, "loss": 0.0006, "step": 4176 }, { "epoch": 0.5, "learning_rate": 1.5074141048824591e-06, "loss": 0.0008, "step": 4177 }, { "epoch": 0.5, "learning_rate": 1.5070524412296564e-06, "loss": 0.0013, "step": 4178 }, { "epoch": 0.5, "learning_rate": 1.5066907775768534e-06, "loss": 0.0005, "step": 4179 }, { "epoch": 0.5, "learning_rate": 1.5063291139240506e-06, "loss": 0.0007, "step": 4180 }, { "epoch": 0.5, "learning_rate": 1.5059674502712476e-06, "loss": 0.0014, "step": 4181 }, { "epoch": 0.5, "learning_rate": 1.5056057866184449e-06, "loss": 0.0008, "step": 4182 }, { "epoch": 0.5, "learning_rate": 1.5052441229656419e-06, "loss": 0.0004, "step": 4183 }, { "epoch": 0.5, "learning_rate": 1.5048824593128391e-06, "loss": 0.0004, "step": 4184 }, { "epoch": 0.5, "learning_rate": 1.5045207956600361e-06, "loss": 0.0005, "step": 4185 }, { "epoch": 0.5, "learning_rate": 1.5041591320072334e-06, "loss": 0.0003, "step": 4186 }, { "epoch": 0.5, "learning_rate": 1.5037974683544304e-06, "loss": 0.0045, "step": 4187 }, { "epoch": 0.5, "learning_rate": 1.5034358047016276e-06, "loss": 0.0004, "step": 4188 }, { "epoch": 0.5, "learning_rate": 1.5030741410488246e-06, "loss": 0.0005, "step": 4189 }, { "epoch": 0.5, "learning_rate": 1.5027124773960219e-06, "loss": 0.0005, "step": 4190 }, { "epoch": 0.5, "learning_rate": 1.5023508137432189e-06, "loss": 1.1073, "step": 4191 }, { "epoch": 0.5, "learning_rate": 1.5019891500904161e-06, "loss": 0.0002, "step": 4192 }, { "epoch": 0.5, "learning_rate": 1.5016274864376131e-06, "loss": 0.0003, "step": 4193 }, { "epoch": 0.5, "learning_rate": 1.5012658227848104e-06, "loss": 0.0004, "step": 4194 }, { "epoch": 0.5, "learning_rate": 1.5009041591320074e-06, "loss": 0.0003, "step": 4195 }, { "epoch": 0.5, "learning_rate": 1.5005424954792046e-06, "loss": 0.0023, "step": 4196 }, { "epoch": 0.5, "learning_rate": 1.5001808318264016e-06, "loss": 1.0719, "step": 4197 }, { "epoch": 0.5, "learning_rate": 1.4998191681735987e-06, "loss": 0.0006, "step": 4198 }, { "epoch": 0.5, "learning_rate": 1.4994575045207957e-06, "loss": 0.0003, "step": 4199 }, { "epoch": 0.5, "learning_rate": 1.499095840867993e-06, "loss": 0.0006, "step": 4200 }, { "epoch": 0.5, "learning_rate": 1.49873417721519e-06, "loss": 1.0631, "step": 4201 }, { "epoch": 0.5, "learning_rate": 1.4983725135623872e-06, "loss": 0.0004, "step": 4202 }, { "epoch": 0.5, "learning_rate": 1.4980108499095842e-06, "loss": 0.0005, "step": 4203 }, { "epoch": 0.5, "learning_rate": 1.4976491862567812e-06, "loss": 0.0002, "step": 4204 }, { "epoch": 0.5, "learning_rate": 1.4972875226039782e-06, "loss": 0.0005, "step": 4205 }, { "epoch": 0.5, "learning_rate": 1.4969258589511754e-06, "loss": 0.0004, "step": 4206 }, { "epoch": 0.5, "learning_rate": 1.4965641952983725e-06, "loss": 0.0002, "step": 4207 }, { "epoch": 0.5, "learning_rate": 1.4962025316455697e-06, "loss": 0.0003, "step": 4208 }, { "epoch": 0.5, "learning_rate": 1.4958408679927667e-06, "loss": 0.9331, "step": 4209 }, { "epoch": 0.5, "learning_rate": 1.495479204339964e-06, "loss": 0.0004, "step": 4210 }, { "epoch": 0.5, "learning_rate": 1.495117540687161e-06, "loss": 0.3075, "step": 4211 }, { "epoch": 0.5, "learning_rate": 1.4947558770343582e-06, "loss": 0.0005, "step": 4212 }, { "epoch": 0.5, "learning_rate": 1.4943942133815552e-06, "loss": 0.001, "step": 4213 }, { "epoch": 0.5, "learning_rate": 1.4940325497287524e-06, "loss": 0.9407, "step": 4214 }, { "epoch": 0.51, "learning_rate": 1.4936708860759495e-06, "loss": 0.0004, "step": 4215 }, { "epoch": 0.51, "learning_rate": 1.4933092224231465e-06, "loss": 0.0202, "step": 4216 }, { "epoch": 0.51, "learning_rate": 1.4929475587703435e-06, "loss": 0.0004, "step": 4217 }, { "epoch": 0.51, "learning_rate": 1.4925858951175407e-06, "loss": 0.0006, "step": 4218 }, { "epoch": 0.51, "learning_rate": 1.4922242314647378e-06, "loss": 0.001, "step": 4219 }, { "epoch": 0.51, "learning_rate": 1.491862567811935e-06, "loss": 1.0363, "step": 4220 }, { "epoch": 0.51, "learning_rate": 1.491500904159132e-06, "loss": 0.0003, "step": 4221 }, { "epoch": 0.51, "learning_rate": 1.4911392405063292e-06, "loss": 0.0006, "step": 4222 }, { "epoch": 0.51, "learning_rate": 1.4907775768535263e-06, "loss": 0.0007, "step": 4223 }, { "epoch": 0.51, "learning_rate": 1.4904159132007235e-06, "loss": 0.0005, "step": 4224 }, { "epoch": 0.51, "learning_rate": 1.4900542495479205e-06, "loss": 0.0008, "step": 4225 }, { "epoch": 0.51, "learning_rate": 1.4896925858951177e-06, "loss": 0.0019, "step": 4226 }, { "epoch": 0.51, "learning_rate": 1.4893309222423148e-06, "loss": 0.0007, "step": 4227 }, { "epoch": 0.51, "learning_rate": 1.4889692585895118e-06, "loss": 0.0006, "step": 4228 }, { "epoch": 0.51, "learning_rate": 1.4886075949367088e-06, "loss": 0.0005, "step": 4229 }, { "epoch": 0.51, "learning_rate": 1.488245931283906e-06, "loss": 0.9597, "step": 4230 }, { "epoch": 0.51, "learning_rate": 1.487884267631103e-06, "loss": 0.0005, "step": 4231 }, { "epoch": 0.51, "learning_rate": 1.4875226039783003e-06, "loss": 0.1462, "step": 4232 }, { "epoch": 0.51, "learning_rate": 1.4871609403254973e-06, "loss": 0.0008, "step": 4233 }, { "epoch": 0.51, "learning_rate": 1.4867992766726945e-06, "loss": 0.0007, "step": 4234 }, { "epoch": 0.51, "learning_rate": 1.4864376130198915e-06, "loss": 0.0008, "step": 4235 }, { "epoch": 0.51, "learning_rate": 1.4860759493670888e-06, "loss": 0.0016, "step": 4236 }, { "epoch": 0.51, "learning_rate": 1.4857142857142858e-06, "loss": 0.0005, "step": 4237 }, { "epoch": 0.51, "learning_rate": 1.4853526220614828e-06, "loss": 0.0005, "step": 4238 }, { "epoch": 0.51, "learning_rate": 1.48499095840868e-06, "loss": 0.3941, "step": 4239 }, { "epoch": 0.51, "learning_rate": 1.484629294755877e-06, "loss": 0.576, "step": 4240 }, { "epoch": 0.51, "learning_rate": 1.484267631103074e-06, "loss": 0.0005, "step": 4241 }, { "epoch": 0.51, "learning_rate": 1.4839059674502713e-06, "loss": 0.0005, "step": 4242 }, { "epoch": 0.51, "learning_rate": 1.4835443037974683e-06, "loss": 0.0006, "step": 4243 }, { "epoch": 0.51, "learning_rate": 1.4831826401446656e-06, "loss": 0.001, "step": 4244 }, { "epoch": 0.51, "learning_rate": 1.4828209764918626e-06, "loss": 0.0009, "step": 4245 }, { "epoch": 0.51, "learning_rate": 1.4824593128390598e-06, "loss": 0.0006, "step": 4246 }, { "epoch": 0.51, "learning_rate": 1.4820976491862568e-06, "loss": 0.0011, "step": 4247 }, { "epoch": 0.51, "learning_rate": 1.4817359855334539e-06, "loss": 0.0007, "step": 4248 }, { "epoch": 0.51, "learning_rate": 1.481374321880651e-06, "loss": 0.0005, "step": 4249 }, { "epoch": 0.51, "learning_rate": 1.4810126582278481e-06, "loss": 0.4853, "step": 4250 }, { "epoch": 0.51, "learning_rate": 1.4806509945750453e-06, "loss": 0.0005, "step": 4251 }, { "epoch": 0.51, "learning_rate": 1.4802893309222424e-06, "loss": 0.0006, "step": 4252 }, { "epoch": 0.51, "learning_rate": 1.4799276672694394e-06, "loss": 0.0005, "step": 4253 }, { "epoch": 0.51, "learning_rate": 1.4795660036166366e-06, "loss": 0.001, "step": 4254 }, { "epoch": 0.51, "learning_rate": 1.4792043399638336e-06, "loss": 0.0008, "step": 4255 }, { "epoch": 0.51, "learning_rate": 1.4788426763110309e-06, "loss": 0.0009, "step": 4256 }, { "epoch": 0.51, "learning_rate": 1.4784810126582279e-06, "loss": 0.0004, "step": 4257 }, { "epoch": 0.51, "learning_rate": 1.478119349005425e-06, "loss": 0.7707, "step": 4258 }, { "epoch": 0.51, "learning_rate": 1.4777576853526221e-06, "loss": 0.0044, "step": 4259 }, { "epoch": 0.51, "learning_rate": 1.4773960216998192e-06, "loss": 0.0004, "step": 4260 }, { "epoch": 0.51, "learning_rate": 1.4770343580470164e-06, "loss": 0.0028, "step": 4261 }, { "epoch": 0.51, "learning_rate": 1.4766726943942134e-06, "loss": 0.5407, "step": 4262 }, { "epoch": 0.51, "learning_rate": 1.4763110307414106e-06, "loss": 0.0004, "step": 4263 }, { "epoch": 0.51, "learning_rate": 1.4759493670886077e-06, "loss": 0.0005, "step": 4264 }, { "epoch": 0.51, "learning_rate": 1.4755877034358047e-06, "loss": 0.0005, "step": 4265 }, { "epoch": 0.51, "learning_rate": 1.475226039783002e-06, "loss": 0.0003, "step": 4266 }, { "epoch": 0.51, "learning_rate": 1.474864376130199e-06, "loss": 0.0006, "step": 4267 }, { "epoch": 0.51, "learning_rate": 1.474502712477396e-06, "loss": 0.0004, "step": 4268 }, { "epoch": 0.51, "learning_rate": 1.4741410488245932e-06, "loss": 0.0008, "step": 4269 }, { "epoch": 0.51, "learning_rate": 1.4737793851717902e-06, "loss": 0.0008, "step": 4270 }, { "epoch": 0.51, "learning_rate": 1.4734177215189874e-06, "loss": 0.0005, "step": 4271 }, { "epoch": 0.51, "learning_rate": 1.4730560578661844e-06, "loss": 0.0004, "step": 4272 }, { "epoch": 0.51, "learning_rate": 1.4726943942133817e-06, "loss": 0.0007, "step": 4273 }, { "epoch": 0.51, "learning_rate": 1.4723327305605787e-06, "loss": 0.0006, "step": 4274 }, { "epoch": 0.51, "learning_rate": 1.471971066907776e-06, "loss": 0.0006, "step": 4275 }, { "epoch": 0.51, "learning_rate": 1.471609403254973e-06, "loss": 0.0005, "step": 4276 }, { "epoch": 0.51, "learning_rate": 1.47124773960217e-06, "loss": 0.0004, "step": 4277 }, { "epoch": 0.51, "learning_rate": 1.470886075949367e-06, "loss": 0.0004, "step": 4278 }, { "epoch": 0.51, "learning_rate": 1.4705244122965642e-06, "loss": 0.0003, "step": 4279 }, { "epoch": 0.51, "learning_rate": 1.4701627486437612e-06, "loss": 0.0003, "step": 4280 }, { "epoch": 0.51, "learning_rate": 1.4698010849909585e-06, "loss": 0.0005, "step": 4281 }, { "epoch": 0.51, "learning_rate": 1.4694394213381555e-06, "loss": 0.0005, "step": 4282 }, { "epoch": 0.51, "learning_rate": 1.4690777576853527e-06, "loss": 0.3585, "step": 4283 }, { "epoch": 0.51, "learning_rate": 1.4687160940325497e-06, "loss": 0.0006, "step": 4284 }, { "epoch": 0.51, "learning_rate": 1.468354430379747e-06, "loss": 0.0047, "step": 4285 }, { "epoch": 0.51, "learning_rate": 1.467992766726944e-06, "loss": 0.0003, "step": 4286 }, { "epoch": 0.51, "learning_rate": 1.4676311030741412e-06, "loss": 0.0003, "step": 4287 }, { "epoch": 0.51, "learning_rate": 1.467269439421338e-06, "loss": 0.8202, "step": 4288 }, { "epoch": 0.51, "learning_rate": 1.4669077757685353e-06, "loss": 0.0007, "step": 4289 }, { "epoch": 0.51, "learning_rate": 1.4665461121157323e-06, "loss": 0.0005, "step": 4290 }, { "epoch": 0.51, "learning_rate": 1.4661844484629295e-06, "loss": 0.0007, "step": 4291 }, { "epoch": 0.51, "learning_rate": 1.4658227848101265e-06, "loss": 0.0003, "step": 4292 }, { "epoch": 0.51, "learning_rate": 1.4654611211573238e-06, "loss": 0.4267, "step": 4293 }, { "epoch": 0.51, "learning_rate": 1.4650994575045208e-06, "loss": 0.0005, "step": 4294 }, { "epoch": 0.51, "learning_rate": 1.464737793851718e-06, "loss": 0.0005, "step": 4295 }, { "epoch": 0.51, "learning_rate": 1.464376130198915e-06, "loss": 0.0004, "step": 4296 }, { "epoch": 0.51, "learning_rate": 1.4640144665461123e-06, "loss": 0.0004, "step": 4297 }, { "epoch": 0.52, "learning_rate": 1.4636528028933093e-06, "loss": 0.0003, "step": 4298 }, { "epoch": 0.52, "learning_rate": 1.4632911392405065e-06, "loss": 0.0003, "step": 4299 }, { "epoch": 0.52, "learning_rate": 1.4629294755877033e-06, "loss": 0.0006, "step": 4300 }, { "epoch": 0.52, "learning_rate": 1.4625678119349005e-06, "loss": 0.0008, "step": 4301 }, { "epoch": 0.52, "learning_rate": 1.4622061482820976e-06, "loss": 0.0005, "step": 4302 }, { "epoch": 0.52, "learning_rate": 1.4618444846292948e-06, "loss": 0.8241, "step": 4303 }, { "epoch": 0.52, "learning_rate": 1.4614828209764918e-06, "loss": 0.7617, "step": 4304 }, { "epoch": 0.52, "learning_rate": 1.461121157323689e-06, "loss": 0.0405, "step": 4305 }, { "epoch": 0.52, "learning_rate": 1.460759493670886e-06, "loss": 0.8183, "step": 4306 }, { "epoch": 0.52, "learning_rate": 1.4603978300180833e-06, "loss": 0.0006, "step": 4307 }, { "epoch": 0.52, "learning_rate": 1.4600361663652803e-06, "loss": 0.0003, "step": 4308 }, { "epoch": 0.52, "learning_rate": 1.4596745027124775e-06, "loss": 0.0004, "step": 4309 }, { "epoch": 0.52, "learning_rate": 1.4593128390596746e-06, "loss": 0.0003, "step": 4310 }, { "epoch": 0.52, "learning_rate": 1.4589511754068718e-06, "loss": 0.0006, "step": 4311 }, { "epoch": 0.52, "learning_rate": 1.4585895117540686e-06, "loss": 0.0003, "step": 4312 }, { "epoch": 0.52, "learning_rate": 1.4582278481012658e-06, "loss": 0.0003, "step": 4313 }, { "epoch": 0.52, "learning_rate": 1.4578661844484629e-06, "loss": 0.0005, "step": 4314 }, { "epoch": 0.52, "learning_rate": 1.45750452079566e-06, "loss": 0.0005, "step": 4315 }, { "epoch": 0.52, "learning_rate": 1.4571428571428571e-06, "loss": 0.0003, "step": 4316 }, { "epoch": 0.52, "learning_rate": 1.4567811934900543e-06, "loss": 0.8254, "step": 4317 }, { "epoch": 0.52, "learning_rate": 1.4564195298372514e-06, "loss": 0.0003, "step": 4318 }, { "epoch": 0.52, "learning_rate": 1.4560578661844486e-06, "loss": 0.0004, "step": 4319 }, { "epoch": 0.52, "learning_rate": 1.4556962025316456e-06, "loss": 0.0004, "step": 4320 }, { "epoch": 0.52, "learning_rate": 1.4553345388788428e-06, "loss": 0.7158, "step": 4321 }, { "epoch": 0.52, "learning_rate": 1.4549728752260399e-06, "loss": 0.0004, "step": 4322 }, { "epoch": 0.52, "learning_rate": 1.454611211573237e-06, "loss": 0.0003, "step": 4323 }, { "epoch": 0.52, "learning_rate": 1.454249547920434e-06, "loss": 0.0004, "step": 4324 }, { "epoch": 0.52, "learning_rate": 1.4538878842676311e-06, "loss": 0.0004, "step": 4325 }, { "epoch": 0.52, "learning_rate": 1.4535262206148281e-06, "loss": 0.0025, "step": 4326 }, { "epoch": 0.52, "learning_rate": 1.4531645569620254e-06, "loss": 0.0004, "step": 4327 }, { "epoch": 0.52, "learning_rate": 1.4528028933092224e-06, "loss": 0.0003, "step": 4328 }, { "epoch": 0.52, "learning_rate": 1.4524412296564196e-06, "loss": 0.0006, "step": 4329 }, { "epoch": 0.52, "learning_rate": 1.4520795660036166e-06, "loss": 0.0007, "step": 4330 }, { "epoch": 0.52, "learning_rate": 1.4517179023508139e-06, "loss": 0.0005, "step": 4331 }, { "epoch": 0.52, "learning_rate": 1.451356238698011e-06, "loss": 0.0008, "step": 4332 }, { "epoch": 0.52, "learning_rate": 1.4509945750452081e-06, "loss": 0.0108, "step": 4333 }, { "epoch": 0.52, "learning_rate": 1.4506329113924052e-06, "loss": 0.0004, "step": 4334 }, { "epoch": 0.52, "learning_rate": 1.4502712477396022e-06, "loss": 0.0004, "step": 4335 }, { "epoch": 0.52, "learning_rate": 1.4499095840867992e-06, "loss": 0.0005, "step": 4336 }, { "epoch": 0.52, "learning_rate": 1.4495479204339964e-06, "loss": 0.6884, "step": 4337 }, { "epoch": 0.52, "learning_rate": 1.4491862567811934e-06, "loss": 0.032, "step": 4338 }, { "epoch": 0.52, "learning_rate": 1.4488245931283907e-06, "loss": 0.0002, "step": 4339 }, { "epoch": 0.52, "learning_rate": 1.4484629294755877e-06, "loss": 0.0006, "step": 4340 }, { "epoch": 0.52, "learning_rate": 1.448101265822785e-06, "loss": 0.0014, "step": 4341 }, { "epoch": 0.52, "learning_rate": 1.447739602169982e-06, "loss": 0.0004, "step": 4342 }, { "epoch": 0.52, "learning_rate": 1.4473779385171792e-06, "loss": 0.0019, "step": 4343 }, { "epoch": 0.52, "learning_rate": 1.4470162748643762e-06, "loss": 0.0004, "step": 4344 }, { "epoch": 0.52, "learning_rate": 1.4466546112115734e-06, "loss": 0.0087, "step": 4345 }, { "epoch": 0.52, "learning_rate": 1.4462929475587704e-06, "loss": 0.0004, "step": 4346 }, { "epoch": 0.52, "learning_rate": 1.4459312839059675e-06, "loss": 0.9477, "step": 4347 }, { "epoch": 0.52, "learning_rate": 1.4455696202531645e-06, "loss": 0.0003, "step": 4348 }, { "epoch": 0.52, "learning_rate": 1.4452079566003617e-06, "loss": 0.0003, "step": 4349 }, { "epoch": 0.52, "learning_rate": 1.4448462929475587e-06, "loss": 0.1569, "step": 4350 }, { "epoch": 0.52, "learning_rate": 1.444484629294756e-06, "loss": 0.0004, "step": 4351 }, { "epoch": 0.52, "learning_rate": 1.444122965641953e-06, "loss": 0.0004, "step": 4352 }, { "epoch": 0.52, "learning_rate": 1.4437613019891502e-06, "loss": 0.0004, "step": 4353 }, { "epoch": 0.52, "learning_rate": 1.4433996383363472e-06, "loss": 0.0004, "step": 4354 }, { "epoch": 0.52, "learning_rate": 1.4430379746835445e-06, "loss": 0.0004, "step": 4355 }, { "epoch": 0.52, "learning_rate": 1.4426763110307415e-06, "loss": 0.0069, "step": 4356 }, { "epoch": 0.52, "learning_rate": 1.4423146473779387e-06, "loss": 0.0004, "step": 4357 }, { "epoch": 0.52, "learning_rate": 1.4419529837251357e-06, "loss": 0.0004, "step": 4358 }, { "epoch": 0.52, "learning_rate": 1.4415913200723328e-06, "loss": 0.0003, "step": 4359 }, { "epoch": 0.52, "learning_rate": 1.4412296564195298e-06, "loss": 0.0003, "step": 4360 }, { "epoch": 0.52, "learning_rate": 1.440867992766727e-06, "loss": 0.0007, "step": 4361 }, { "epoch": 0.52, "learning_rate": 1.440506329113924e-06, "loss": 0.0004, "step": 4362 }, { "epoch": 0.52, "learning_rate": 1.4401446654611213e-06, "loss": 0.0004, "step": 4363 }, { "epoch": 0.52, "learning_rate": 1.4397830018083183e-06, "loss": 0.0004, "step": 4364 }, { "epoch": 0.52, "learning_rate": 1.4394213381555155e-06, "loss": 0.0005, "step": 4365 }, { "epoch": 0.52, "learning_rate": 1.4390596745027125e-06, "loss": 0.0003, "step": 4366 }, { "epoch": 0.52, "learning_rate": 1.4386980108499098e-06, "loss": 0.8764, "step": 4367 }, { "epoch": 0.52, "learning_rate": 1.4383363471971068e-06, "loss": 0.0004, "step": 4368 }, { "epoch": 0.52, "learning_rate": 1.4379746835443038e-06, "loss": 0.0005, "step": 4369 }, { "epoch": 0.52, "learning_rate": 1.437613019891501e-06, "loss": 0.0004, "step": 4370 }, { "epoch": 0.52, "learning_rate": 1.437251356238698e-06, "loss": 0.0005, "step": 4371 }, { "epoch": 0.52, "learning_rate": 1.436889692585895e-06, "loss": 0.0009, "step": 4372 }, { "epoch": 0.52, "learning_rate": 1.4365280289330923e-06, "loss": 0.0004, "step": 4373 }, { "epoch": 0.52, "learning_rate": 1.4361663652802893e-06, "loss": 0.0004, "step": 4374 }, { "epoch": 0.52, "learning_rate": 1.4358047016274865e-06, "loss": 0.0004, "step": 4375 }, { "epoch": 0.52, "learning_rate": 1.4354430379746836e-06, "loss": 0.0006, "step": 4376 }, { "epoch": 0.52, "learning_rate": 1.4350813743218806e-06, "loss": 0.0004, "step": 4377 }, { "epoch": 0.52, "learning_rate": 1.4347197106690778e-06, "loss": 0.0005, "step": 4378 }, { "epoch": 0.52, "learning_rate": 1.4343580470162748e-06, "loss": 0.001, "step": 4379 }, { "epoch": 0.52, "learning_rate": 1.433996383363472e-06, "loss": 0.0007, "step": 4380 }, { "epoch": 0.52, "learning_rate": 1.433634719710669e-06, "loss": 0.0004, "step": 4381 }, { "epoch": 0.53, "learning_rate": 1.4332730560578663e-06, "loss": 0.5889, "step": 4382 }, { "epoch": 0.53, "learning_rate": 1.4329113924050633e-06, "loss": 0.0013, "step": 4383 }, { "epoch": 0.53, "learning_rate": 1.4325497287522604e-06, "loss": 0.0004, "step": 4384 }, { "epoch": 0.53, "learning_rate": 1.4321880650994576e-06, "loss": 0.0012, "step": 4385 }, { "epoch": 0.53, "learning_rate": 1.4318264014466546e-06, "loss": 0.0006, "step": 4386 }, { "epoch": 0.53, "learning_rate": 1.4314647377938516e-06, "loss": 0.949, "step": 4387 }, { "epoch": 0.53, "learning_rate": 1.4311030741410489e-06, "loss": 0.0005, "step": 4388 }, { "epoch": 0.53, "learning_rate": 1.4307414104882459e-06, "loss": 0.0004, "step": 4389 }, { "epoch": 0.53, "learning_rate": 1.4303797468354431e-06, "loss": 0.0003, "step": 4390 }, { "epoch": 0.53, "learning_rate": 1.4300180831826401e-06, "loss": 0.0003, "step": 4391 }, { "epoch": 0.53, "learning_rate": 1.4296564195298374e-06, "loss": 0.9995, "step": 4392 }, { "epoch": 0.53, "learning_rate": 1.4292947558770344e-06, "loss": 0.0003, "step": 4393 }, { "epoch": 0.53, "learning_rate": 1.4289330922242316e-06, "loss": 0.0003, "step": 4394 }, { "epoch": 0.53, "learning_rate": 1.4285714285714286e-06, "loss": 0.0063, "step": 4395 }, { "epoch": 0.53, "learning_rate": 1.4282097649186256e-06, "loss": 0.0004, "step": 4396 }, { "epoch": 0.53, "learning_rate": 1.4278481012658227e-06, "loss": 0.0004, "step": 4397 }, { "epoch": 0.53, "learning_rate": 1.42748643761302e-06, "loss": 0.0005, "step": 4398 }, { "epoch": 0.53, "learning_rate": 1.427124773960217e-06, "loss": 0.0005, "step": 4399 }, { "epoch": 0.53, "learning_rate": 1.4267631103074141e-06, "loss": 0.0007, "step": 4400 }, { "epoch": 0.53, "learning_rate": 1.4264014466546112e-06, "loss": 0.0005, "step": 4401 }, { "epoch": 0.53, "learning_rate": 1.4260397830018084e-06, "loss": 0.0007, "step": 4402 }, { "epoch": 0.53, "learning_rate": 1.4256781193490054e-06, "loss": 0.0003, "step": 4403 }, { "epoch": 0.53, "learning_rate": 1.4253164556962026e-06, "loss": 0.0004, "step": 4404 }, { "epoch": 0.53, "learning_rate": 1.4249547920433997e-06, "loss": 0.0006, "step": 4405 }, { "epoch": 0.53, "learning_rate": 1.424593128390597e-06, "loss": 0.0003, "step": 4406 }, { "epoch": 0.53, "learning_rate": 1.4242314647377937e-06, "loss": 0.0005, "step": 4407 }, { "epoch": 0.53, "learning_rate": 1.423869801084991e-06, "loss": 0.0003, "step": 4408 }, { "epoch": 0.53, "learning_rate": 1.423508137432188e-06, "loss": 0.0004, "step": 4409 }, { "epoch": 0.53, "learning_rate": 1.4231464737793852e-06, "loss": 0.0002, "step": 4410 }, { "epoch": 0.53, "learning_rate": 1.4227848101265822e-06, "loss": 0.5427, "step": 4411 }, { "epoch": 0.53, "learning_rate": 1.4224231464737794e-06, "loss": 0.0026, "step": 4412 }, { "epoch": 0.53, "learning_rate": 1.4220614828209765e-06, "loss": 0.0005, "step": 4413 }, { "epoch": 0.53, "learning_rate": 1.4216998191681737e-06, "loss": 0.0004, "step": 4414 }, { "epoch": 0.53, "learning_rate": 1.4213381555153707e-06, "loss": 0.0005, "step": 4415 }, { "epoch": 0.53, "learning_rate": 1.420976491862568e-06, "loss": 0.0005, "step": 4416 }, { "epoch": 0.53, "learning_rate": 1.420614828209765e-06, "loss": 0.0005, "step": 4417 }, { "epoch": 0.53, "learning_rate": 1.4202531645569622e-06, "loss": 0.0258, "step": 4418 }, { "epoch": 0.53, "learning_rate": 1.419891500904159e-06, "loss": 0.0004, "step": 4419 }, { "epoch": 0.53, "learning_rate": 1.4195298372513562e-06, "loss": 0.0004, "step": 4420 }, { "epoch": 0.53, "learning_rate": 1.4191681735985532e-06, "loss": 0.0003, "step": 4421 }, { "epoch": 0.53, "learning_rate": 1.4188065099457505e-06, "loss": 0.0004, "step": 4422 }, { "epoch": 0.53, "learning_rate": 1.4184448462929475e-06, "loss": 0.0003, "step": 4423 }, { "epoch": 0.53, "learning_rate": 1.4180831826401447e-06, "loss": 0.0004, "step": 4424 }, { "epoch": 0.53, "learning_rate": 1.4177215189873418e-06, "loss": 0.0003, "step": 4425 }, { "epoch": 0.53, "learning_rate": 1.417359855334539e-06, "loss": 0.0003, "step": 4426 }, { "epoch": 0.53, "learning_rate": 1.416998191681736e-06, "loss": 0.0004, "step": 4427 }, { "epoch": 0.53, "learning_rate": 1.4166365280289332e-06, "loss": 0.625, "step": 4428 }, { "epoch": 0.53, "learning_rate": 1.4162748643761303e-06, "loss": 0.0003, "step": 4429 }, { "epoch": 0.53, "learning_rate": 1.4159132007233275e-06, "loss": 0.0009, "step": 4430 }, { "epoch": 0.53, "learning_rate": 1.4155515370705243e-06, "loss": 0.0008, "step": 4431 }, { "epoch": 0.53, "learning_rate": 1.4151898734177215e-06, "loss": 0.0004, "step": 4432 }, { "epoch": 0.53, "learning_rate": 1.4148282097649185e-06, "loss": 0.0004, "step": 4433 }, { "epoch": 0.53, "learning_rate": 1.4144665461121158e-06, "loss": 0.0004, "step": 4434 }, { "epoch": 0.53, "learning_rate": 1.4141048824593128e-06, "loss": 0.0002, "step": 4435 }, { "epoch": 0.53, "learning_rate": 1.41374321880651e-06, "loss": 0.0003, "step": 4436 }, { "epoch": 0.53, "learning_rate": 1.413381555153707e-06, "loss": 0.0003, "step": 4437 }, { "epoch": 0.53, "learning_rate": 1.4130198915009043e-06, "loss": 0.0002, "step": 4438 }, { "epoch": 0.53, "learning_rate": 1.4126582278481013e-06, "loss": 0.0008, "step": 4439 }, { "epoch": 0.53, "learning_rate": 1.4122965641952985e-06, "loss": 0.0002, "step": 4440 }, { "epoch": 0.53, "learning_rate": 1.4119349005424955e-06, "loss": 0.0006, "step": 4441 }, { "epoch": 0.53, "learning_rate": 1.4115732368896928e-06, "loss": 0.0005, "step": 4442 }, { "epoch": 0.53, "learning_rate": 1.4112115732368896e-06, "loss": 0.0004, "step": 4443 }, { "epoch": 0.53, "learning_rate": 1.4108499095840868e-06, "loss": 0.0003, "step": 4444 }, { "epoch": 0.53, "learning_rate": 1.4104882459312838e-06, "loss": 0.0005, "step": 4445 }, { "epoch": 0.53, "learning_rate": 1.410126582278481e-06, "loss": 0.0004, "step": 4446 }, { "epoch": 0.53, "learning_rate": 1.409764918625678e-06, "loss": 0.0004, "step": 4447 }, { "epoch": 0.53, "learning_rate": 1.4094032549728753e-06, "loss": 0.0003, "step": 4448 }, { "epoch": 0.53, "learning_rate": 1.4090415913200723e-06, "loss": 0.0004, "step": 4449 }, { "epoch": 0.53, "learning_rate": 1.4086799276672696e-06, "loss": 0.0004, "step": 4450 }, { "epoch": 0.53, "learning_rate": 1.4083182640144666e-06, "loss": 0.0004, "step": 4451 }, { "epoch": 0.53, "learning_rate": 1.4079566003616638e-06, "loss": 0.0003, "step": 4452 }, { "epoch": 0.53, "learning_rate": 1.4075949367088608e-06, "loss": 0.83, "step": 4453 }, { "epoch": 0.53, "learning_rate": 1.407233273056058e-06, "loss": 0.0003, "step": 4454 }, { "epoch": 0.53, "learning_rate": 1.4068716094032549e-06, "loss": 0.0003, "step": 4455 }, { "epoch": 0.53, "learning_rate": 1.406509945750452e-06, "loss": 0.0047, "step": 4456 }, { "epoch": 0.53, "learning_rate": 1.4061482820976491e-06, "loss": 0.0002, "step": 4457 }, { "epoch": 0.53, "learning_rate": 1.4057866184448464e-06, "loss": 0.0003, "step": 4458 }, { "epoch": 0.53, "learning_rate": 1.4054249547920434e-06, "loss": 0.7607, "step": 4459 }, { "epoch": 0.53, "learning_rate": 1.4050632911392406e-06, "loss": 0.0008, "step": 4460 }, { "epoch": 0.53, "learning_rate": 1.4047016274864376e-06, "loss": 0.9253, "step": 4461 }, { "epoch": 0.53, "learning_rate": 1.4043399638336349e-06, "loss": 0.2611, "step": 4462 }, { "epoch": 0.53, "learning_rate": 1.4039783001808319e-06, "loss": 0.0003, "step": 4463 }, { "epoch": 0.53, "learning_rate": 1.4036166365280291e-06, "loss": 0.0003, "step": 4464 }, { "epoch": 0.54, "learning_rate": 1.4032549728752261e-06, "loss": 0.8892, "step": 4465 }, { "epoch": 0.54, "learning_rate": 1.4028933092224234e-06, "loss": 0.0023, "step": 4466 }, { "epoch": 0.54, "learning_rate": 1.4025316455696202e-06, "loss": 0.0003, "step": 4467 }, { "epoch": 0.54, "learning_rate": 1.4021699819168174e-06, "loss": 0.0005, "step": 4468 }, { "epoch": 0.54, "learning_rate": 1.4018083182640144e-06, "loss": 0.0002, "step": 4469 }, { "epoch": 0.54, "learning_rate": 1.4014466546112116e-06, "loss": 0.0003, "step": 4470 }, { "epoch": 0.54, "learning_rate": 1.4010849909584087e-06, "loss": 0.0003, "step": 4471 }, { "epoch": 0.54, "learning_rate": 1.400723327305606e-06, "loss": 0.0003, "step": 4472 }, { "epoch": 0.54, "learning_rate": 1.400361663652803e-06, "loss": 0.0004, "step": 4473 }, { "epoch": 0.54, "learning_rate": 1.4000000000000001e-06, "loss": 0.0002, "step": 4474 }, { "epoch": 0.54, "learning_rate": 1.3996383363471972e-06, "loss": 0.002, "step": 4475 }, { "epoch": 0.54, "learning_rate": 1.3992766726943944e-06, "loss": 0.0003, "step": 4476 }, { "epoch": 0.54, "learning_rate": 1.3989150090415914e-06, "loss": 0.0003, "step": 4477 }, { "epoch": 0.54, "learning_rate": 1.3985533453887884e-06, "loss": 0.0003, "step": 4478 }, { "epoch": 0.54, "learning_rate": 1.3981916817359855e-06, "loss": 0.0004, "step": 4479 }, { "epoch": 0.54, "learning_rate": 1.3978300180831827e-06, "loss": 0.0003, "step": 4480 }, { "epoch": 0.54, "learning_rate": 1.3974683544303797e-06, "loss": 0.0667, "step": 4481 }, { "epoch": 0.54, "learning_rate": 1.397106690777577e-06, "loss": 0.0003, "step": 4482 }, { "epoch": 0.54, "learning_rate": 1.396745027124774e-06, "loss": 0.0004, "step": 4483 }, { "epoch": 0.54, "learning_rate": 1.3963833634719712e-06, "loss": 0.8324, "step": 4484 }, { "epoch": 0.54, "learning_rate": 1.3960216998191682e-06, "loss": 0.0002, "step": 4485 }, { "epoch": 0.54, "learning_rate": 1.3956600361663654e-06, "loss": 0.0003, "step": 4486 }, { "epoch": 0.54, "learning_rate": 1.3952983725135625e-06, "loss": 0.0002, "step": 4487 }, { "epoch": 0.54, "learning_rate": 1.3949367088607595e-06, "loss": 0.0053, "step": 4488 }, { "epoch": 0.54, "learning_rate": 1.3945750452079567e-06, "loss": 0.0004, "step": 4489 }, { "epoch": 0.54, "learning_rate": 1.3942133815551537e-06, "loss": 0.0003, "step": 4490 }, { "epoch": 0.54, "learning_rate": 1.3938517179023507e-06, "loss": 0.0003, "step": 4491 }, { "epoch": 0.54, "learning_rate": 1.393490054249548e-06, "loss": 0.0004, "step": 4492 }, { "epoch": 0.54, "learning_rate": 1.393128390596745e-06, "loss": 0.0003, "step": 4493 }, { "epoch": 0.54, "learning_rate": 1.3927667269439422e-06, "loss": 0.0002, "step": 4494 }, { "epoch": 0.54, "learning_rate": 1.3924050632911392e-06, "loss": 0.0003, "step": 4495 }, { "epoch": 0.54, "learning_rate": 1.3920433996383365e-06, "loss": 0.8456, "step": 4496 }, { "epoch": 0.54, "learning_rate": 1.3916817359855335e-06, "loss": 0.0003, "step": 4497 }, { "epoch": 0.54, "learning_rate": 1.3913200723327305e-06, "loss": 0.0002, "step": 4498 }, { "epoch": 0.54, "learning_rate": 1.3909584086799278e-06, "loss": 0.0003, "step": 4499 }, { "epoch": 0.54, "learning_rate": 1.3905967450271248e-06, "loss": 0.0152, "step": 4500 }, { "epoch": 0.54, "learning_rate": 1.390235081374322e-06, "loss": 0.0005, "step": 4501 }, { "epoch": 0.54, "learning_rate": 1.389873417721519e-06, "loss": 0.843, "step": 4502 }, { "epoch": 0.54, "learning_rate": 1.389511754068716e-06, "loss": 0.0002, "step": 4503 }, { "epoch": 0.54, "learning_rate": 1.3891500904159133e-06, "loss": 0.0002, "step": 4504 }, { "epoch": 0.54, "learning_rate": 1.3887884267631103e-06, "loss": 0.0005, "step": 4505 }, { "epoch": 0.54, "learning_rate": 1.3884267631103075e-06, "loss": 0.0002, "step": 4506 }, { "epoch": 0.54, "learning_rate": 1.3880650994575045e-06, "loss": 0.0002, "step": 4507 }, { "epoch": 0.54, "learning_rate": 1.3877034358047016e-06, "loss": 0.0003, "step": 4508 }, { "epoch": 0.54, "learning_rate": 1.3873417721518988e-06, "loss": 0.0004, "step": 4509 }, { "epoch": 0.54, "learning_rate": 1.3869801084990958e-06, "loss": 0.0003, "step": 4510 }, { "epoch": 0.54, "learning_rate": 1.386618444846293e-06, "loss": 0.0003, "step": 4511 }, { "epoch": 0.54, "learning_rate": 1.38625678119349e-06, "loss": 0.0003, "step": 4512 }, { "epoch": 0.54, "learning_rate": 1.3858951175406873e-06, "loss": 0.0005, "step": 4513 }, { "epoch": 0.54, "learning_rate": 1.3855334538878843e-06, "loss": 0.0005, "step": 4514 }, { "epoch": 0.54, "learning_rate": 1.3851717902350813e-06, "loss": 0.0002, "step": 4515 }, { "epoch": 0.54, "learning_rate": 1.3848101265822786e-06, "loss": 0.1475, "step": 4516 }, { "epoch": 0.54, "learning_rate": 1.3844484629294756e-06, "loss": 0.0012, "step": 4517 }, { "epoch": 0.54, "learning_rate": 1.3840867992766726e-06, "loss": 0.0008, "step": 4518 }, { "epoch": 0.54, "learning_rate": 1.3837251356238698e-06, "loss": 0.0004, "step": 4519 }, { "epoch": 0.54, "learning_rate": 1.3833634719710669e-06, "loss": 0.0666, "step": 4520 }, { "epoch": 0.54, "learning_rate": 1.383001808318264e-06, "loss": 0.8839, "step": 4521 }, { "epoch": 0.54, "learning_rate": 1.382640144665461e-06, "loss": 0.0002, "step": 4522 }, { "epoch": 0.54, "learning_rate": 1.3822784810126583e-06, "loss": 0.0003, "step": 4523 }, { "epoch": 0.54, "learning_rate": 1.3819168173598554e-06, "loss": 0.0004, "step": 4524 }, { "epoch": 0.54, "learning_rate": 1.3815551537070526e-06, "loss": 0.0003, "step": 4525 }, { "epoch": 0.54, "learning_rate": 1.3811934900542496e-06, "loss": 0.8093, "step": 4526 }, { "epoch": 0.54, "learning_rate": 1.3808318264014466e-06, "loss": 0.0004, "step": 4527 }, { "epoch": 0.54, "learning_rate": 1.3804701627486436e-06, "loss": 0.0002, "step": 4528 }, { "epoch": 0.54, "learning_rate": 1.3801084990958409e-06, "loss": 0.0004, "step": 4529 }, { "epoch": 0.54, "learning_rate": 1.3797468354430379e-06, "loss": 0.0003, "step": 4530 }, { "epoch": 0.54, "learning_rate": 1.3793851717902351e-06, "loss": 0.0003, "step": 4531 }, { "epoch": 0.54, "learning_rate": 1.3790235081374321e-06, "loss": 0.0003, "step": 4532 }, { "epoch": 0.54, "learning_rate": 1.3786618444846294e-06, "loss": 0.0005, "step": 4533 }, { "epoch": 0.54, "learning_rate": 1.3783001808318264e-06, "loss": 0.0003, "step": 4534 }, { "epoch": 0.54, "learning_rate": 1.3779385171790236e-06, "loss": 0.0002, "step": 4535 }, { "epoch": 0.54, "learning_rate": 1.3775768535262206e-06, "loss": 0.0005, "step": 4536 }, { "epoch": 0.54, "learning_rate": 1.3772151898734179e-06, "loss": 0.0002, "step": 4537 }, { "epoch": 0.54, "learning_rate": 1.376853526220615e-06, "loss": 0.0002, "step": 4538 }, { "epoch": 0.54, "learning_rate": 1.376491862567812e-06, "loss": 0.0003, "step": 4539 }, { "epoch": 0.54, "learning_rate": 1.376130198915009e-06, "loss": 0.0003, "step": 4540 }, { "epoch": 0.54, "learning_rate": 1.3757685352622062e-06, "loss": 0.0003, "step": 4541 }, { "epoch": 0.54, "learning_rate": 1.3754068716094032e-06, "loss": 0.0003, "step": 4542 }, { "epoch": 0.54, "learning_rate": 1.3750452079566004e-06, "loss": 0.0003, "step": 4543 }, { "epoch": 0.54, "learning_rate": 1.3746835443037974e-06, "loss": 0.4045, "step": 4544 }, { "epoch": 0.54, "learning_rate": 1.3743218806509947e-06, "loss": 0.0003, "step": 4545 }, { "epoch": 0.54, "learning_rate": 1.3739602169981917e-06, "loss": 0.0003, "step": 4546 }, { "epoch": 0.54, "learning_rate": 1.373598553345389e-06, "loss": 0.0002, "step": 4547 }, { "epoch": 0.54, "learning_rate": 1.373236889692586e-06, "loss": 0.0002, "step": 4548 }, { "epoch": 0.55, "learning_rate": 1.3728752260397832e-06, "loss": 0.0002, "step": 4549 }, { "epoch": 0.55, "learning_rate": 1.3725135623869802e-06, "loss": 0.0002, "step": 4550 }, { "epoch": 0.55, "learning_rate": 1.3721518987341772e-06, "loss": 0.0003, "step": 4551 }, { "epoch": 0.55, "learning_rate": 1.3717902350813742e-06, "loss": 0.0002, "step": 4552 }, { "epoch": 0.55, "learning_rate": 1.3714285714285715e-06, "loss": 0.0003, "step": 4553 }, { "epoch": 0.55, "learning_rate": 1.3710669077757685e-06, "loss": 0.0007, "step": 4554 }, { "epoch": 0.55, "learning_rate": 1.3707052441229657e-06, "loss": 0.623, "step": 4555 }, { "epoch": 0.55, "learning_rate": 1.3703435804701627e-06, "loss": 0.0003, "step": 4556 }, { "epoch": 0.55, "learning_rate": 1.36998191681736e-06, "loss": 0.0003, "step": 4557 }, { "epoch": 0.55, "learning_rate": 1.369620253164557e-06, "loss": 0.0003, "step": 4558 }, { "epoch": 0.55, "learning_rate": 1.3692585895117542e-06, "loss": 0.0003, "step": 4559 }, { "epoch": 0.55, "learning_rate": 1.3688969258589512e-06, "loss": 0.0003, "step": 4560 }, { "epoch": 0.55, "learning_rate": 1.3685352622061485e-06, "loss": 0.0003, "step": 4561 }, { "epoch": 0.55, "learning_rate": 1.3681735985533455e-06, "loss": 0.0003, "step": 4562 }, { "epoch": 0.55, "learning_rate": 1.3678119349005425e-06, "loss": 0.0002, "step": 4563 }, { "epoch": 0.55, "learning_rate": 1.3674502712477395e-06, "loss": 0.0005, "step": 4564 }, { "epoch": 0.55, "learning_rate": 1.3670886075949367e-06, "loss": 0.0003, "step": 4565 }, { "epoch": 0.55, "learning_rate": 1.3667269439421338e-06, "loss": 0.0005, "step": 4566 }, { "epoch": 0.55, "learning_rate": 1.366365280289331e-06, "loss": 0.0003, "step": 4567 }, { "epoch": 0.55, "learning_rate": 1.366003616636528e-06, "loss": 0.0003, "step": 4568 }, { "epoch": 0.55, "learning_rate": 1.3656419529837252e-06, "loss": 0.0002, "step": 4569 }, { "epoch": 0.55, "learning_rate": 1.3652802893309223e-06, "loss": 0.0004, "step": 4570 }, { "epoch": 0.55, "learning_rate": 1.3649186256781195e-06, "loss": 0.0004, "step": 4571 }, { "epoch": 0.55, "learning_rate": 1.3645569620253165e-06, "loss": 0.0002, "step": 4572 }, { "epoch": 0.55, "learning_rate": 1.3641952983725138e-06, "loss": 0.0003, "step": 4573 }, { "epoch": 0.55, "learning_rate": 1.3638336347197108e-06, "loss": 0.0002, "step": 4574 }, { "epoch": 0.55, "learning_rate": 1.3634719710669078e-06, "loss": 0.0003, "step": 4575 }, { "epoch": 0.55, "learning_rate": 1.3631103074141048e-06, "loss": 0.0003, "step": 4576 }, { "epoch": 0.55, "learning_rate": 1.362748643761302e-06, "loss": 0.0007, "step": 4577 }, { "epoch": 0.55, "learning_rate": 1.362386980108499e-06, "loss": 0.0001, "step": 4578 }, { "epoch": 0.55, "learning_rate": 1.3620253164556963e-06, "loss": 0.0002, "step": 4579 }, { "epoch": 0.55, "learning_rate": 1.3616636528028933e-06, "loss": 0.0005, "step": 4580 }, { "epoch": 0.55, "learning_rate": 1.3613019891500905e-06, "loss": 0.0004, "step": 4581 }, { "epoch": 0.55, "learning_rate": 1.3609403254972876e-06, "loss": 0.0003, "step": 4582 }, { "epoch": 0.55, "learning_rate": 1.3605786618444848e-06, "loss": 0.0003, "step": 4583 }, { "epoch": 0.55, "learning_rate": 1.3602169981916818e-06, "loss": 0.0003, "step": 4584 }, { "epoch": 0.55, "learning_rate": 1.359855334538879e-06, "loss": 0.0003, "step": 4585 }, { "epoch": 0.55, "learning_rate": 1.3594936708860759e-06, "loss": 0.0009, "step": 4586 }, { "epoch": 0.55, "learning_rate": 1.359132007233273e-06, "loss": 0.1927, "step": 4587 }, { "epoch": 0.55, "learning_rate": 1.35877034358047e-06, "loss": 0.0002, "step": 4588 }, { "epoch": 0.55, "learning_rate": 1.3584086799276673e-06, "loss": 0.0002, "step": 4589 }, { "epoch": 0.55, "learning_rate": 1.3580470162748644e-06, "loss": 0.0003, "step": 4590 }, { "epoch": 0.55, "learning_rate": 1.3576853526220616e-06, "loss": 0.0003, "step": 4591 }, { "epoch": 0.55, "learning_rate": 1.3573236889692586e-06, "loss": 0.0002, "step": 4592 }, { "epoch": 0.55, "learning_rate": 1.3569620253164558e-06, "loss": 0.0016, "step": 4593 }, { "epoch": 0.55, "learning_rate": 1.3566003616636529e-06, "loss": 0.0004, "step": 4594 }, { "epoch": 0.55, "learning_rate": 1.35623869801085e-06, "loss": 0.0003, "step": 4595 }, { "epoch": 0.55, "learning_rate": 1.355877034358047e-06, "loss": 0.0003, "step": 4596 }, { "epoch": 0.55, "learning_rate": 1.3555153707052441e-06, "loss": 0.0002, "step": 4597 }, { "epoch": 0.55, "learning_rate": 1.3551537070524411e-06, "loss": 0.0003, "step": 4598 }, { "epoch": 0.55, "learning_rate": 1.3547920433996384e-06, "loss": 0.0002, "step": 4599 }, { "epoch": 0.55, "learning_rate": 1.3544303797468354e-06, "loss": 1.5932, "step": 4600 }, { "epoch": 0.55, "learning_rate": 1.3540687160940326e-06, "loss": 0.0002, "step": 4601 }, { "epoch": 0.55, "learning_rate": 1.3537070524412296e-06, "loss": 0.0003, "step": 4602 }, { "epoch": 0.55, "learning_rate": 1.3533453887884269e-06, "loss": 0.0002, "step": 4603 }, { "epoch": 0.55, "learning_rate": 1.3529837251356239e-06, "loss": 0.0007, "step": 4604 }, { "epoch": 0.55, "learning_rate": 1.3526220614828211e-06, "loss": 0.0002, "step": 4605 }, { "epoch": 0.55, "learning_rate": 1.3522603978300181e-06, "loss": 0.0941, "step": 4606 }, { "epoch": 0.55, "learning_rate": 1.3518987341772152e-06, "loss": 0.0003, "step": 4607 }, { "epoch": 0.55, "learning_rate": 1.3515370705244124e-06, "loss": 0.0001, "step": 4608 }, { "epoch": 0.55, "learning_rate": 1.3511754068716094e-06, "loss": 0.0003, "step": 4609 }, { "epoch": 0.55, "learning_rate": 1.3508137432188064e-06, "loss": 0.0004, "step": 4610 }, { "epoch": 0.55, "learning_rate": 1.3504520795660037e-06, "loss": 0.0002, "step": 4611 }, { "epoch": 0.55, "learning_rate": 1.3500904159132007e-06, "loss": 0.0003, "step": 4612 }, { "epoch": 0.55, "learning_rate": 1.349728752260398e-06, "loss": 0.0003, "step": 4613 }, { "epoch": 0.55, "learning_rate": 1.349367088607595e-06, "loss": 0.0003, "step": 4614 }, { "epoch": 0.55, "learning_rate": 1.3490054249547922e-06, "loss": 0.2905, "step": 4615 }, { "epoch": 0.55, "learning_rate": 1.3486437613019892e-06, "loss": 0.0002, "step": 4616 }, { "epoch": 0.55, "learning_rate": 1.3482820976491862e-06, "loss": 0.0002, "step": 4617 }, { "epoch": 0.55, "learning_rate": 1.3479204339963834e-06, "loss": 0.0004, "step": 4618 }, { "epoch": 0.55, "learning_rate": 1.3475587703435805e-06, "loss": 0.0003, "step": 4619 }, { "epoch": 0.55, "learning_rate": 1.3471971066907777e-06, "loss": 0.0002, "step": 4620 }, { "epoch": 0.55, "learning_rate": 1.3468354430379747e-06, "loss": 0.0002, "step": 4621 }, { "epoch": 0.55, "learning_rate": 1.3464737793851717e-06, "loss": 0.1257, "step": 4622 }, { "epoch": 0.55, "learning_rate": 1.346112115732369e-06, "loss": 0.0002, "step": 4623 }, { "epoch": 0.55, "learning_rate": 1.345750452079566e-06, "loss": 0.9021, "step": 4624 }, { "epoch": 0.55, "learning_rate": 1.3453887884267632e-06, "loss": 0.2049, "step": 4625 }, { "epoch": 0.55, "learning_rate": 1.3450271247739602e-06, "loss": 0.0399, "step": 4626 }, { "epoch": 0.55, "learning_rate": 1.3446654611211572e-06, "loss": 0.0195, "step": 4627 }, { "epoch": 0.55, "learning_rate": 1.3443037974683545e-06, "loss": 0.0003, "step": 4628 }, { "epoch": 0.55, "learning_rate": 1.3439421338155515e-06, "loss": 0.0505, "step": 4629 }, { "epoch": 0.55, "learning_rate": 1.3435804701627487e-06, "loss": 0.506, "step": 4630 }, { "epoch": 0.55, "learning_rate": 1.3432188065099457e-06, "loss": 0.0003, "step": 4631 }, { "epoch": 0.56, "learning_rate": 1.342857142857143e-06, "loss": 0.268, "step": 4632 }, { "epoch": 0.56, "learning_rate": 1.34249547920434e-06, "loss": 1.6228, "step": 4633 }, { "epoch": 0.56, "learning_rate": 1.342133815551537e-06, "loss": 1.0962, "step": 4634 }, { "epoch": 0.56, "learning_rate": 1.3417721518987342e-06, "loss": 0.0003, "step": 4635 }, { "epoch": 0.56, "learning_rate": 1.3414104882459313e-06, "loss": 0.0077, "step": 4636 }, { "epoch": 0.56, "learning_rate": 1.3410488245931283e-06, "loss": 0.0007, "step": 4637 }, { "epoch": 0.56, "learning_rate": 1.3406871609403255e-06, "loss": 1.1167, "step": 4638 }, { "epoch": 0.56, "learning_rate": 1.3403254972875225e-06, "loss": 0.0003, "step": 4639 }, { "epoch": 0.56, "learning_rate": 1.3399638336347198e-06, "loss": 0.0003, "step": 4640 }, { "epoch": 0.56, "learning_rate": 1.3396021699819168e-06, "loss": 0.0002, "step": 4641 }, { "epoch": 0.56, "learning_rate": 1.339240506329114e-06, "loss": 0.0002, "step": 4642 }, { "epoch": 0.56, "learning_rate": 1.338878842676311e-06, "loss": 0.0005, "step": 4643 }, { "epoch": 0.56, "learning_rate": 1.3385171790235083e-06, "loss": 0.0004, "step": 4644 }, { "epoch": 0.56, "learning_rate": 1.3381555153707053e-06, "loss": 0.0003, "step": 4645 }, { "epoch": 0.56, "learning_rate": 1.3377938517179023e-06, "loss": 0.0002, "step": 4646 }, { "epoch": 0.56, "learning_rate": 1.3374321880650993e-06, "loss": 0.0003, "step": 4647 }, { "epoch": 0.56, "learning_rate": 1.3370705244122966e-06, "loss": 0.0004, "step": 4648 }, { "epoch": 0.56, "learning_rate": 1.3367088607594936e-06, "loss": 0.0004, "step": 4649 }, { "epoch": 0.56, "learning_rate": 1.3363471971066908e-06, "loss": 0.0005, "step": 4650 }, { "epoch": 0.56, "learning_rate": 1.3359855334538878e-06, "loss": 0.0002, "step": 4651 }, { "epoch": 0.56, "learning_rate": 1.335623869801085e-06, "loss": 0.1622, "step": 4652 }, { "epoch": 0.56, "learning_rate": 1.335262206148282e-06, "loss": 0.0043, "step": 4653 }, { "epoch": 0.56, "learning_rate": 1.3349005424954793e-06, "loss": 0.0003, "step": 4654 }, { "epoch": 0.56, "learning_rate": 1.3345388788426763e-06, "loss": 0.0002, "step": 4655 }, { "epoch": 0.56, "learning_rate": 1.3341772151898736e-06, "loss": 0.0004, "step": 4656 }, { "epoch": 0.56, "learning_rate": 1.3338155515370706e-06, "loss": 0.0013, "step": 4657 }, { "epoch": 0.56, "learning_rate": 1.3334538878842676e-06, "loss": 0.0004, "step": 4658 }, { "epoch": 0.56, "learning_rate": 1.3330922242314646e-06, "loss": 0.0004, "step": 4659 }, { "epoch": 0.56, "learning_rate": 1.3327305605786619e-06, "loss": 0.0003, "step": 4660 }, { "epoch": 0.56, "learning_rate": 1.3323688969258589e-06, "loss": 0.0005, "step": 4661 }, { "epoch": 0.56, "learning_rate": 1.332007233273056e-06, "loss": 0.0002, "step": 4662 }, { "epoch": 0.56, "learning_rate": 1.3316455696202531e-06, "loss": 0.0003, "step": 4663 }, { "epoch": 0.56, "learning_rate": 1.3312839059674504e-06, "loss": 0.0003, "step": 4664 }, { "epoch": 0.56, "learning_rate": 1.3309222423146474e-06, "loss": 0.0003, "step": 4665 }, { "epoch": 0.56, "learning_rate": 1.3305605786618446e-06, "loss": 0.0004, "step": 4666 }, { "epoch": 0.56, "learning_rate": 1.3301989150090416e-06, "loss": 0.0007, "step": 4667 }, { "epoch": 0.56, "learning_rate": 1.3298372513562389e-06, "loss": 0.0004, "step": 4668 }, { "epoch": 0.56, "learning_rate": 1.3294755877034359e-06, "loss": 0.0006, "step": 4669 }, { "epoch": 0.56, "learning_rate": 1.3291139240506329e-06, "loss": 0.0004, "step": 4670 }, { "epoch": 0.56, "learning_rate": 1.32875226039783e-06, "loss": 0.0003, "step": 4671 }, { "epoch": 0.56, "learning_rate": 1.3283905967450271e-06, "loss": 0.0004, "step": 4672 }, { "epoch": 0.56, "learning_rate": 1.3280289330922242e-06, "loss": 0.0004, "step": 4673 }, { "epoch": 0.56, "learning_rate": 1.3276672694394214e-06, "loss": 0.0016, "step": 4674 }, { "epoch": 0.56, "learning_rate": 1.3273056057866184e-06, "loss": 0.0003, "step": 4675 }, { "epoch": 0.56, "learning_rate": 1.3269439421338156e-06, "loss": 0.0003, "step": 4676 }, { "epoch": 0.56, "learning_rate": 1.3265822784810127e-06, "loss": 0.7934, "step": 4677 }, { "epoch": 0.56, "learning_rate": 1.3262206148282099e-06, "loss": 0.0003, "step": 4678 }, { "epoch": 0.56, "learning_rate": 1.325858951175407e-06, "loss": 0.8793, "step": 4679 }, { "epoch": 0.56, "learning_rate": 1.3254972875226041e-06, "loss": 0.0003, "step": 4680 }, { "epoch": 0.56, "learning_rate": 1.3251356238698012e-06, "loss": 0.0003, "step": 4681 }, { "epoch": 0.56, "learning_rate": 1.3247739602169982e-06, "loss": 0.0004, "step": 4682 }, { "epoch": 0.56, "learning_rate": 1.3244122965641952e-06, "loss": 0.0004, "step": 4683 }, { "epoch": 0.56, "learning_rate": 1.3240506329113924e-06, "loss": 0.0913, "step": 4684 }, { "epoch": 0.56, "learning_rate": 1.3236889692585895e-06, "loss": 0.0015, "step": 4685 }, { "epoch": 0.56, "learning_rate": 1.3233273056057867e-06, "loss": 0.0009, "step": 4686 }, { "epoch": 0.56, "learning_rate": 1.3229656419529837e-06, "loss": 0.0002, "step": 4687 }, { "epoch": 0.56, "learning_rate": 1.322603978300181e-06, "loss": 0.0005, "step": 4688 }, { "epoch": 0.56, "learning_rate": 1.322242314647378e-06, "loss": 0.0004, "step": 4689 }, { "epoch": 0.56, "learning_rate": 1.3218806509945752e-06, "loss": 0.7002, "step": 4690 }, { "epoch": 0.56, "learning_rate": 1.3215189873417722e-06, "loss": 0.6104, "step": 4691 }, { "epoch": 0.56, "learning_rate": 1.3211573236889694e-06, "loss": 0.0004, "step": 4692 }, { "epoch": 0.56, "learning_rate": 1.3207956600361665e-06, "loss": 0.0003, "step": 4693 }, { "epoch": 0.56, "learning_rate": 1.3204339963833635e-06, "loss": 0.0002, "step": 4694 }, { "epoch": 0.56, "learning_rate": 1.3200723327305605e-06, "loss": 0.0004, "step": 4695 }, { "epoch": 0.56, "learning_rate": 1.3197106690777577e-06, "loss": 0.0002, "step": 4696 }, { "epoch": 0.56, "learning_rate": 1.3193490054249547e-06, "loss": 0.0023, "step": 4697 }, { "epoch": 0.56, "learning_rate": 1.318987341772152e-06, "loss": 0.0004, "step": 4698 }, { "epoch": 0.56, "learning_rate": 1.318625678119349e-06, "loss": 0.0003, "step": 4699 }, { "epoch": 0.56, "learning_rate": 1.3182640144665462e-06, "loss": 0.0004, "step": 4700 }, { "epoch": 0.56, "learning_rate": 1.3179023508137432e-06, "loss": 0.0082, "step": 4701 }, { "epoch": 0.56, "learning_rate": 1.3175406871609405e-06, "loss": 0.0003, "step": 4702 }, { "epoch": 0.56, "learning_rate": 1.3171790235081375e-06, "loss": 0.0003, "step": 4703 }, { "epoch": 0.56, "learning_rate": 1.3168173598553347e-06, "loss": 0.0003, "step": 4704 }, { "epoch": 0.56, "learning_rate": 1.3164556962025317e-06, "loss": 0.0004, "step": 4705 }, { "epoch": 0.56, "learning_rate": 1.3160940325497288e-06, "loss": 0.0041, "step": 4706 }, { "epoch": 0.56, "learning_rate": 1.3157323688969258e-06, "loss": 0.0002, "step": 4707 }, { "epoch": 0.56, "learning_rate": 1.315370705244123e-06, "loss": 0.0003, "step": 4708 }, { "epoch": 0.56, "learning_rate": 1.31500904159132e-06, "loss": 0.0003, "step": 4709 }, { "epoch": 0.56, "learning_rate": 1.3146473779385173e-06, "loss": 0.0318, "step": 4710 }, { "epoch": 0.56, "learning_rate": 1.3142857142857143e-06, "loss": 0.0003, "step": 4711 }, { "epoch": 0.56, "learning_rate": 1.3139240506329115e-06, "loss": 0.0002, "step": 4712 }, { "epoch": 0.56, "learning_rate": 1.3135623869801085e-06, "loss": 0.0005, "step": 4713 }, { "epoch": 0.56, "learning_rate": 1.3132007233273058e-06, "loss": 0.0002, "step": 4714 }, { "epoch": 0.57, "learning_rate": 1.3128390596745028e-06, "loss": 0.3328, "step": 4715 }, { "epoch": 0.57, "learning_rate": 1.3124773960217e-06, "loss": 0.0003, "step": 4716 }, { "epoch": 0.57, "learning_rate": 1.312115732368897e-06, "loss": 0.0005, "step": 4717 }, { "epoch": 0.57, "learning_rate": 1.311754068716094e-06, "loss": 0.0012, "step": 4718 }, { "epoch": 0.57, "learning_rate": 1.311392405063291e-06, "loss": 0.0002, "step": 4719 }, { "epoch": 0.57, "learning_rate": 1.3110307414104883e-06, "loss": 0.0007, "step": 4720 }, { "epoch": 0.57, "learning_rate": 1.3106690777576853e-06, "loss": 0.0003, "step": 4721 }, { "epoch": 0.57, "learning_rate": 1.3103074141048826e-06, "loss": 0.0003, "step": 4722 }, { "epoch": 0.57, "learning_rate": 1.3099457504520796e-06, "loss": 0.0003, "step": 4723 }, { "epoch": 0.57, "learning_rate": 1.3095840867992768e-06, "loss": 0.0019, "step": 4724 }, { "epoch": 0.57, "learning_rate": 1.3092224231464738e-06, "loss": 0.0004, "step": 4725 }, { "epoch": 0.57, "learning_rate": 1.308860759493671e-06, "loss": 0.0023, "step": 4726 }, { "epoch": 0.57, "learning_rate": 1.308499095840868e-06, "loss": 0.0037, "step": 4727 }, { "epoch": 0.57, "learning_rate": 1.308137432188065e-06, "loss": 0.7674, "step": 4728 }, { "epoch": 0.57, "learning_rate": 1.3077757685352623e-06, "loss": 0.0339, "step": 4729 }, { "epoch": 0.57, "learning_rate": 1.3074141048824593e-06, "loss": 0.0004, "step": 4730 }, { "epoch": 0.57, "learning_rate": 1.3070524412296564e-06, "loss": 0.0002, "step": 4731 }, { "epoch": 0.57, "learning_rate": 1.3066907775768536e-06, "loss": 0.0003, "step": 4732 }, { "epoch": 0.57, "learning_rate": 1.3063291139240506e-06, "loss": 0.5138, "step": 4733 }, { "epoch": 0.57, "learning_rate": 1.3059674502712479e-06, "loss": 0.0005, "step": 4734 }, { "epoch": 0.57, "learning_rate": 1.3056057866184449e-06, "loss": 0.0003, "step": 4735 }, { "epoch": 0.57, "learning_rate": 1.3052441229656419e-06, "loss": 0.4749, "step": 4736 }, { "epoch": 0.57, "learning_rate": 1.3048824593128391e-06, "loss": 0.0002, "step": 4737 }, { "epoch": 0.57, "learning_rate": 1.3045207956600361e-06, "loss": 0.0002, "step": 4738 }, { "epoch": 0.57, "learning_rate": 1.3041591320072334e-06, "loss": 0.0003, "step": 4739 }, { "epoch": 0.57, "learning_rate": 1.3037974683544304e-06, "loss": 0.0004, "step": 4740 }, { "epoch": 0.57, "learning_rate": 1.3034358047016276e-06, "loss": 0.291, "step": 4741 }, { "epoch": 0.57, "learning_rate": 1.3030741410488246e-06, "loss": 0.0005, "step": 4742 }, { "epoch": 0.57, "learning_rate": 1.3027124773960217e-06, "loss": 0.0002, "step": 4743 }, { "epoch": 0.57, "learning_rate": 1.3023508137432189e-06, "loss": 0.0003, "step": 4744 }, { "epoch": 0.57, "learning_rate": 1.301989150090416e-06, "loss": 0.0004, "step": 4745 }, { "epoch": 0.57, "learning_rate": 1.301627486437613e-06, "loss": 0.0002, "step": 4746 }, { "epoch": 0.57, "learning_rate": 1.3012658227848102e-06, "loss": 0.9931, "step": 4747 }, { "epoch": 0.57, "learning_rate": 1.3009041591320072e-06, "loss": 0.0003, "step": 4748 }, { "epoch": 0.57, "learning_rate": 1.3005424954792044e-06, "loss": 0.0013, "step": 4749 }, { "epoch": 0.57, "learning_rate": 1.3001808318264014e-06, "loss": 0.0003, "step": 4750 }, { "epoch": 0.57, "learning_rate": 1.2998191681735987e-06, "loss": 0.0004, "step": 4751 }, { "epoch": 0.57, "learning_rate": 1.2994575045207957e-06, "loss": 0.0002, "step": 4752 }, { "epoch": 0.57, "learning_rate": 1.299095840867993e-06, "loss": 0.0009, "step": 4753 }, { "epoch": 0.57, "learning_rate": 1.29873417721519e-06, "loss": 0.0005, "step": 4754 }, { "epoch": 0.57, "learning_rate": 1.298372513562387e-06, "loss": 0.0002, "step": 4755 }, { "epoch": 0.57, "learning_rate": 1.298010849909584e-06, "loss": 0.0003, "step": 4756 }, { "epoch": 0.57, "learning_rate": 1.2976491862567812e-06, "loss": 0.0002, "step": 4757 }, { "epoch": 0.57, "learning_rate": 1.2972875226039782e-06, "loss": 0.0018, "step": 4758 }, { "epoch": 0.57, "learning_rate": 1.2969258589511755e-06, "loss": 0.0003, "step": 4759 }, { "epoch": 0.57, "learning_rate": 1.2965641952983725e-06, "loss": 0.0002, "step": 4760 }, { "epoch": 0.57, "learning_rate": 1.2962025316455697e-06, "loss": 0.1475, "step": 4761 }, { "epoch": 0.57, "learning_rate": 1.2958408679927667e-06, "loss": 0.0002, "step": 4762 }, { "epoch": 0.57, "learning_rate": 1.295479204339964e-06, "loss": 0.0002, "step": 4763 }, { "epoch": 0.57, "learning_rate": 1.295117540687161e-06, "loss": 0.0002, "step": 4764 }, { "epoch": 0.57, "learning_rate": 1.2947558770343582e-06, "loss": 0.0009, "step": 4765 }, { "epoch": 0.57, "learning_rate": 1.294394213381555e-06, "loss": 0.0003, "step": 4766 }, { "epoch": 0.57, "learning_rate": 1.2940325497287522e-06, "loss": 0.0007, "step": 4767 }, { "epoch": 0.57, "learning_rate": 1.2936708860759493e-06, "loss": 0.0002, "step": 4768 }, { "epoch": 0.57, "learning_rate": 1.2933092224231465e-06, "loss": 0.002, "step": 4769 }, { "epoch": 0.57, "learning_rate": 1.2929475587703435e-06, "loss": 0.0003, "step": 4770 }, { "epoch": 0.57, "learning_rate": 1.2925858951175407e-06, "loss": 0.0003, "step": 4771 }, { "epoch": 0.57, "learning_rate": 1.2922242314647378e-06, "loss": 0.0443, "step": 4772 }, { "epoch": 0.57, "learning_rate": 1.291862567811935e-06, "loss": 0.0018, "step": 4773 }, { "epoch": 0.57, "learning_rate": 1.291500904159132e-06, "loss": 0.0005, "step": 4774 }, { "epoch": 0.57, "learning_rate": 1.2911392405063292e-06, "loss": 0.0003, "step": 4775 }, { "epoch": 0.57, "learning_rate": 1.2907775768535263e-06, "loss": 0.8023, "step": 4776 }, { "epoch": 0.57, "learning_rate": 1.2904159132007235e-06, "loss": 0.0003, "step": 4777 }, { "epoch": 0.57, "learning_rate": 1.2900542495479203e-06, "loss": 0.0003, "step": 4778 }, { "epoch": 0.57, "learning_rate": 1.2896925858951175e-06, "loss": 0.0003, "step": 4779 }, { "epoch": 0.57, "learning_rate": 1.2893309222423146e-06, "loss": 0.0002, "step": 4780 }, { "epoch": 0.57, "learning_rate": 1.2889692585895118e-06, "loss": 0.0003, "step": 4781 }, { "epoch": 0.57, "learning_rate": 1.2886075949367088e-06, "loss": 0.9047, "step": 4782 }, { "epoch": 0.57, "learning_rate": 1.288245931283906e-06, "loss": 0.0007, "step": 4783 }, { "epoch": 0.57, "learning_rate": 1.287884267631103e-06, "loss": 0.0021, "step": 4784 }, { "epoch": 0.57, "learning_rate": 1.2875226039783003e-06, "loss": 0.0002, "step": 4785 }, { "epoch": 0.57, "learning_rate": 1.2871609403254973e-06, "loss": 0.0107, "step": 4786 }, { "epoch": 0.57, "learning_rate": 1.2867992766726945e-06, "loss": 0.0003, "step": 4787 }, { "epoch": 0.57, "learning_rate": 1.2864376130198916e-06, "loss": 0.0004, "step": 4788 }, { "epoch": 0.57, "learning_rate": 1.2860759493670888e-06, "loss": 0.001, "step": 4789 }, { "epoch": 0.57, "learning_rate": 1.2857142857142856e-06, "loss": 0.0003, "step": 4790 }, { "epoch": 0.57, "learning_rate": 1.2853526220614828e-06, "loss": 0.5865, "step": 4791 }, { "epoch": 0.57, "learning_rate": 1.2849909584086798e-06, "loss": 0.0002, "step": 4792 }, { "epoch": 0.57, "learning_rate": 1.284629294755877e-06, "loss": 0.0002, "step": 4793 }, { "epoch": 0.57, "learning_rate": 1.284267631103074e-06, "loss": 0.0004, "step": 4794 }, { "epoch": 0.57, "learning_rate": 1.2839059674502713e-06, "loss": 0.0003, "step": 4795 }, { "epoch": 0.57, "learning_rate": 1.2835443037974683e-06, "loss": 0.0027, "step": 4796 }, { "epoch": 0.57, "learning_rate": 1.2831826401446656e-06, "loss": 0.0002, "step": 4797 }, { "epoch": 0.57, "learning_rate": 1.2828209764918626e-06, "loss": 0.0003, "step": 4798 }, { "epoch": 0.58, "learning_rate": 1.2824593128390598e-06, "loss": 0.597, "step": 4799 }, { "epoch": 0.58, "learning_rate": 1.2820976491862568e-06, "loss": 0.0009, "step": 4800 }, { "epoch": 0.58, "learning_rate": 1.281735985533454e-06, "loss": 0.0006, "step": 4801 }, { "epoch": 0.58, "learning_rate": 1.2813743218806509e-06, "loss": 0.0002, "step": 4802 }, { "epoch": 0.58, "learning_rate": 1.2810126582278481e-06, "loss": 0.5799, "step": 4803 }, { "epoch": 0.58, "learning_rate": 1.2806509945750451e-06, "loss": 0.0003, "step": 4804 }, { "epoch": 0.58, "learning_rate": 1.2802893309222424e-06, "loss": 0.0006, "step": 4805 }, { "epoch": 0.58, "learning_rate": 1.2799276672694394e-06, "loss": 0.0004, "step": 4806 }, { "epoch": 0.58, "learning_rate": 1.2795660036166366e-06, "loss": 0.0002, "step": 4807 }, { "epoch": 0.58, "learning_rate": 1.2792043399638336e-06, "loss": 0.0002, "step": 4808 }, { "epoch": 0.58, "learning_rate": 1.2788426763110309e-06, "loss": 0.0002, "step": 4809 }, { "epoch": 0.58, "learning_rate": 1.2784810126582279e-06, "loss": 0.0002, "step": 4810 }, { "epoch": 0.58, "learning_rate": 1.2781193490054251e-06, "loss": 0.0002, "step": 4811 }, { "epoch": 0.58, "learning_rate": 1.2777576853526221e-06, "loss": 0.0002, "step": 4812 }, { "epoch": 0.58, "learning_rate": 1.2773960216998194e-06, "loss": 0.0002, "step": 4813 }, { "epoch": 0.58, "learning_rate": 1.2770343580470162e-06, "loss": 0.0003, "step": 4814 }, { "epoch": 0.58, "learning_rate": 1.2766726943942134e-06, "loss": 0.0003, "step": 4815 }, { "epoch": 0.58, "learning_rate": 1.2763110307414104e-06, "loss": 0.0002, "step": 4816 }, { "epoch": 0.58, "learning_rate": 1.2759493670886077e-06, "loss": 0.0003, "step": 4817 }, { "epoch": 0.58, "learning_rate": 1.2755877034358047e-06, "loss": 0.0003, "step": 4818 }, { "epoch": 0.58, "learning_rate": 1.275226039783002e-06, "loss": 0.0003, "step": 4819 }, { "epoch": 0.58, "learning_rate": 1.274864376130199e-06, "loss": 0.0002, "step": 4820 }, { "epoch": 0.58, "learning_rate": 1.2745027124773962e-06, "loss": 0.0002, "step": 4821 }, { "epoch": 0.58, "learning_rate": 1.2741410488245932e-06, "loss": 0.0002, "step": 4822 }, { "epoch": 0.58, "learning_rate": 1.2737793851717904e-06, "loss": 1.0762, "step": 4823 }, { "epoch": 0.58, "learning_rate": 1.2734177215189874e-06, "loss": 0.0002, "step": 4824 }, { "epoch": 0.58, "learning_rate": 1.2730560578661847e-06, "loss": 0.0002, "step": 4825 }, { "epoch": 0.58, "learning_rate": 1.2726943942133815e-06, "loss": 0.0005, "step": 4826 }, { "epoch": 0.58, "learning_rate": 1.2723327305605787e-06, "loss": 0.0007, "step": 4827 }, { "epoch": 0.58, "learning_rate": 1.2719710669077757e-06, "loss": 0.0002, "step": 4828 }, { "epoch": 0.58, "learning_rate": 1.271609403254973e-06, "loss": 0.0002, "step": 4829 }, { "epoch": 0.58, "learning_rate": 1.27124773960217e-06, "loss": 0.0007, "step": 4830 }, { "epoch": 0.58, "learning_rate": 1.2708860759493672e-06, "loss": 0.0002, "step": 4831 }, { "epoch": 0.58, "learning_rate": 1.2705244122965642e-06, "loss": 0.0004, "step": 4832 }, { "epoch": 0.58, "learning_rate": 1.2701627486437615e-06, "loss": 0.0003, "step": 4833 }, { "epoch": 0.58, "learning_rate": 1.2698010849909585e-06, "loss": 0.0002, "step": 4834 }, { "epoch": 0.58, "learning_rate": 1.2694394213381557e-06, "loss": 0.7067, "step": 4835 }, { "epoch": 0.58, "learning_rate": 1.2690777576853527e-06, "loss": 0.0003, "step": 4836 }, { "epoch": 0.58, "learning_rate": 1.2687160940325497e-06, "loss": 0.0002, "step": 4837 }, { "epoch": 0.58, "learning_rate": 1.2683544303797468e-06, "loss": 0.0003, "step": 4838 }, { "epoch": 0.58, "learning_rate": 1.267992766726944e-06, "loss": 0.0003, "step": 4839 }, { "epoch": 0.58, "learning_rate": 1.267631103074141e-06, "loss": 0.0002, "step": 4840 }, { "epoch": 0.58, "learning_rate": 1.2672694394213382e-06, "loss": 0.0002, "step": 4841 }, { "epoch": 0.58, "learning_rate": 1.2669077757685353e-06, "loss": 0.0003, "step": 4842 }, { "epoch": 0.58, "learning_rate": 1.2665461121157325e-06, "loss": 0.0002, "step": 4843 }, { "epoch": 0.58, "learning_rate": 1.2661844484629295e-06, "loss": 0.0004, "step": 4844 }, { "epoch": 0.58, "learning_rate": 1.2658227848101267e-06, "loss": 0.0003, "step": 4845 }, { "epoch": 0.58, "learning_rate": 1.2654611211573238e-06, "loss": 0.0002, "step": 4846 }, { "epoch": 0.58, "learning_rate": 1.2650994575045208e-06, "loss": 0.0003, "step": 4847 }, { "epoch": 0.58, "learning_rate": 1.264737793851718e-06, "loss": 0.0002, "step": 4848 }, { "epoch": 0.58, "learning_rate": 1.264376130198915e-06, "loss": 1.0225, "step": 4849 }, { "epoch": 0.58, "learning_rate": 1.264014466546112e-06, "loss": 0.0002, "step": 4850 }, { "epoch": 0.58, "learning_rate": 1.2636528028933093e-06, "loss": 1.0487, "step": 4851 }, { "epoch": 0.58, "learning_rate": 1.2632911392405063e-06, "loss": 0.0004, "step": 4852 }, { "epoch": 0.58, "learning_rate": 1.2629294755877035e-06, "loss": 0.0003, "step": 4853 }, { "epoch": 0.58, "learning_rate": 1.2625678119349006e-06, "loss": 0.0009, "step": 4854 }, { "epoch": 0.58, "learning_rate": 1.2622061482820978e-06, "loss": 1.0062, "step": 4855 }, { "epoch": 0.58, "learning_rate": 1.2618444846292948e-06, "loss": 0.0002, "step": 4856 }, { "epoch": 0.58, "learning_rate": 1.2614828209764918e-06, "loss": 0.0002, "step": 4857 }, { "epoch": 0.58, "learning_rate": 1.261121157323689e-06, "loss": 0.0002, "step": 4858 }, { "epoch": 0.58, "learning_rate": 1.260759493670886e-06, "loss": 0.0004, "step": 4859 }, { "epoch": 0.58, "learning_rate": 1.2603978300180833e-06, "loss": 0.0003, "step": 4860 }, { "epoch": 0.58, "learning_rate": 1.2600361663652803e-06, "loss": 0.0003, "step": 4861 }, { "epoch": 0.58, "learning_rate": 1.2596745027124773e-06, "loss": 0.0004, "step": 4862 }, { "epoch": 0.58, "learning_rate": 1.2593128390596746e-06, "loss": 0.0005, "step": 4863 }, { "epoch": 0.58, "learning_rate": 1.2589511754068716e-06, "loss": 0.0003, "step": 4864 }, { "epoch": 0.58, "learning_rate": 1.2585895117540688e-06, "loss": 0.0002, "step": 4865 }, { "epoch": 0.58, "learning_rate": 1.2582278481012658e-06, "loss": 0.0006, "step": 4866 }, { "epoch": 0.58, "learning_rate": 1.2578661844484629e-06, "loss": 0.0003, "step": 4867 }, { "epoch": 0.58, "learning_rate": 1.25750452079566e-06, "loss": 0.0002, "step": 4868 }, { "epoch": 0.58, "learning_rate": 1.2571428571428571e-06, "loss": 0.0004, "step": 4869 }, { "epoch": 0.58, "learning_rate": 1.2567811934900543e-06, "loss": 0.0003, "step": 4870 }, { "epoch": 0.58, "learning_rate": 1.2564195298372514e-06, "loss": 0.0079, "step": 4871 }, { "epoch": 0.58, "learning_rate": 1.2560578661844486e-06, "loss": 0.0004, "step": 4872 }, { "epoch": 0.58, "learning_rate": 1.2556962025316456e-06, "loss": 0.0005, "step": 4873 }, { "epoch": 0.58, "learning_rate": 1.2553345388788426e-06, "loss": 0.0448, "step": 4874 }, { "epoch": 0.58, "learning_rate": 1.2549728752260397e-06, "loss": 0.0003, "step": 4875 }, { "epoch": 0.58, "learning_rate": 1.2546112115732369e-06, "loss": 0.0004, "step": 4876 }, { "epoch": 0.58, "learning_rate": 1.254249547920434e-06, "loss": 0.0004, "step": 4877 }, { "epoch": 0.58, "learning_rate": 1.2538878842676311e-06, "loss": 0.0002, "step": 4878 }, { "epoch": 0.58, "learning_rate": 1.2535262206148282e-06, "loss": 0.0003, "step": 4879 }, { "epoch": 0.58, "learning_rate": 1.2531645569620254e-06, "loss": 0.0004, "step": 4880 }, { "epoch": 0.58, "learning_rate": 1.2528028933092224e-06, "loss": 0.0005, "step": 4881 }, { "epoch": 0.59, "learning_rate": 1.2524412296564196e-06, "loss": 0.0005, "step": 4882 }, { "epoch": 0.59, "learning_rate": 1.2520795660036167e-06, "loss": 0.0003, "step": 4883 }, { "epoch": 0.59, "learning_rate": 1.2517179023508139e-06, "loss": 0.0004, "step": 4884 }, { "epoch": 0.59, "learning_rate": 1.2513562386980107e-06, "loss": 0.0004, "step": 4885 }, { "epoch": 0.59, "learning_rate": 1.250994575045208e-06, "loss": 0.0005, "step": 4886 }, { "epoch": 0.59, "learning_rate": 1.250632911392405e-06, "loss": 0.0004, "step": 4887 }, { "epoch": 0.59, "learning_rate": 1.2502712477396022e-06, "loss": 0.0003, "step": 4888 }, { "epoch": 0.59, "learning_rate": 1.2499095840867992e-06, "loss": 0.0003, "step": 4889 }, { "epoch": 0.59, "learning_rate": 1.2495479204339964e-06, "loss": 0.0017, "step": 4890 }, { "epoch": 0.59, "learning_rate": 1.2491862567811934e-06, "loss": 0.0003, "step": 4891 }, { "epoch": 0.59, "learning_rate": 1.2488245931283907e-06, "loss": 0.0003, "step": 4892 }, { "epoch": 0.59, "learning_rate": 1.2484629294755877e-06, "loss": 0.0004, "step": 4893 }, { "epoch": 0.59, "learning_rate": 1.248101265822785e-06, "loss": 0.0003, "step": 4894 }, { "epoch": 0.59, "learning_rate": 1.247739602169982e-06, "loss": 0.0004, "step": 4895 }, { "epoch": 0.59, "learning_rate": 1.2473779385171792e-06, "loss": 0.8794, "step": 4896 }, { "epoch": 0.59, "learning_rate": 1.247016274864376e-06, "loss": 0.0006, "step": 4897 }, { "epoch": 0.59, "learning_rate": 1.2466546112115732e-06, "loss": 0.0006, "step": 4898 }, { "epoch": 0.59, "learning_rate": 1.2462929475587702e-06, "loss": 0.0919, "step": 4899 }, { "epoch": 0.59, "learning_rate": 1.2459312839059675e-06, "loss": 0.0004, "step": 4900 }, { "epoch": 0.59, "learning_rate": 1.2455696202531645e-06, "loss": 0.0005, "step": 4901 }, { "epoch": 0.59, "learning_rate": 1.2452079566003617e-06, "loss": 1.0666, "step": 4902 }, { "epoch": 0.59, "learning_rate": 1.2448462929475587e-06, "loss": 0.0004, "step": 4903 }, { "epoch": 0.59, "learning_rate": 1.244484629294756e-06, "loss": 0.0003, "step": 4904 }, { "epoch": 0.59, "learning_rate": 1.244122965641953e-06, "loss": 0.0003, "step": 4905 }, { "epoch": 0.59, "learning_rate": 1.2437613019891502e-06, "loss": 0.0003, "step": 4906 }, { "epoch": 0.59, "learning_rate": 1.2433996383363472e-06, "loss": 0.0004, "step": 4907 }, { "epoch": 0.59, "learning_rate": 1.2430379746835445e-06, "loss": 0.0009, "step": 4908 }, { "epoch": 0.59, "learning_rate": 1.2426763110307413e-06, "loss": 0.0011, "step": 4909 }, { "epoch": 0.59, "learning_rate": 1.2423146473779385e-06, "loss": 0.0004, "step": 4910 }, { "epoch": 0.59, "learning_rate": 1.2419529837251355e-06, "loss": 0.0003, "step": 4911 }, { "epoch": 0.59, "learning_rate": 1.2415913200723328e-06, "loss": 0.0003, "step": 4912 }, { "epoch": 0.59, "learning_rate": 1.2412296564195298e-06, "loss": 0.0011, "step": 4913 }, { "epoch": 0.59, "learning_rate": 1.240867992766727e-06, "loss": 0.0005, "step": 4914 }, { "epoch": 0.59, "learning_rate": 1.240506329113924e-06, "loss": 0.0004, "step": 4915 }, { "epoch": 0.59, "learning_rate": 1.2401446654611213e-06, "loss": 0.0006, "step": 4916 }, { "epoch": 0.59, "learning_rate": 1.2397830018083183e-06, "loss": 0.0004, "step": 4917 }, { "epoch": 0.59, "learning_rate": 1.2394213381555155e-06, "loss": 0.0004, "step": 4918 }, { "epoch": 0.59, "learning_rate": 1.2390596745027125e-06, "loss": 0.0004, "step": 4919 }, { "epoch": 0.59, "learning_rate": 1.2386980108499098e-06, "loss": 0.0004, "step": 4920 }, { "epoch": 0.59, "learning_rate": 1.2383363471971066e-06, "loss": 0.0005, "step": 4921 }, { "epoch": 0.59, "learning_rate": 1.2379746835443038e-06, "loss": 0.0004, "step": 4922 }, { "epoch": 0.59, "learning_rate": 1.2376130198915008e-06, "loss": 0.2746, "step": 4923 }, { "epoch": 0.59, "learning_rate": 1.237251356238698e-06, "loss": 0.4101, "step": 4924 }, { "epoch": 0.59, "learning_rate": 1.236889692585895e-06, "loss": 0.0005, "step": 4925 }, { "epoch": 0.59, "learning_rate": 1.2365280289330923e-06, "loss": 0.0006, "step": 4926 }, { "epoch": 0.59, "learning_rate": 1.2361663652802893e-06, "loss": 0.0006, "step": 4927 }, { "epoch": 0.59, "learning_rate": 1.2358047016274866e-06, "loss": 0.0005, "step": 4928 }, { "epoch": 0.59, "learning_rate": 1.2354430379746836e-06, "loss": 0.0004, "step": 4929 }, { "epoch": 0.59, "learning_rate": 1.2350813743218808e-06, "loss": 0.0004, "step": 4930 }, { "epoch": 0.59, "learning_rate": 1.2347197106690778e-06, "loss": 0.0005, "step": 4931 }, { "epoch": 0.59, "learning_rate": 1.234358047016275e-06, "loss": 0.0003, "step": 4932 }, { "epoch": 0.59, "learning_rate": 1.2339963833634719e-06, "loss": 0.0006, "step": 4933 }, { "epoch": 0.59, "learning_rate": 1.233634719710669e-06, "loss": 0.0891, "step": 4934 }, { "epoch": 0.59, "learning_rate": 1.2332730560578661e-06, "loss": 0.0003, "step": 4935 }, { "epoch": 0.59, "learning_rate": 1.2329113924050633e-06, "loss": 0.0003, "step": 4936 }, { "epoch": 0.59, "learning_rate": 1.2325497287522604e-06, "loss": 0.9706, "step": 4937 }, { "epoch": 0.59, "learning_rate": 1.2321880650994576e-06, "loss": 0.0003, "step": 4938 }, { "epoch": 0.59, "learning_rate": 1.2318264014466546e-06, "loss": 0.0006, "step": 4939 }, { "epoch": 0.59, "learning_rate": 1.2314647377938518e-06, "loss": 0.0005, "step": 4940 }, { "epoch": 0.59, "learning_rate": 1.2311030741410489e-06, "loss": 0.0003, "step": 4941 }, { "epoch": 0.59, "learning_rate": 1.230741410488246e-06, "loss": 0.0004, "step": 4942 }, { "epoch": 0.59, "learning_rate": 1.2303797468354431e-06, "loss": 0.0004, "step": 4943 }, { "epoch": 0.59, "learning_rate": 1.2300180831826403e-06, "loss": 0.0005, "step": 4944 }, { "epoch": 0.59, "learning_rate": 1.2296564195298372e-06, "loss": 0.0005, "step": 4945 }, { "epoch": 0.59, "learning_rate": 1.2292947558770344e-06, "loss": 0.001, "step": 4946 }, { "epoch": 0.59, "learning_rate": 1.2289330922242314e-06, "loss": 0.0004, "step": 4947 }, { "epoch": 0.59, "learning_rate": 1.2285714285714286e-06, "loss": 0.0003, "step": 4948 }, { "epoch": 0.59, "learning_rate": 1.2282097649186257e-06, "loss": 0.0006, "step": 4949 }, { "epoch": 0.59, "learning_rate": 1.2278481012658229e-06, "loss": 1.4742, "step": 4950 }, { "epoch": 0.59, "learning_rate": 1.22748643761302e-06, "loss": 0.0005, "step": 4951 }, { "epoch": 0.59, "learning_rate": 1.2271247739602171e-06, "loss": 0.0009, "step": 4952 }, { "epoch": 0.59, "learning_rate": 1.2267631103074142e-06, "loss": 0.0005, "step": 4953 }, { "epoch": 0.59, "learning_rate": 1.2264014466546114e-06, "loss": 0.0003, "step": 4954 }, { "epoch": 0.59, "learning_rate": 1.2260397830018084e-06, "loss": 0.8315, "step": 4955 }, { "epoch": 0.59, "learning_rate": 1.2256781193490054e-06, "loss": 0.8404, "step": 4956 }, { "epoch": 0.59, "learning_rate": 1.2253164556962024e-06, "loss": 0.0005, "step": 4957 }, { "epoch": 0.59, "learning_rate": 1.2249547920433997e-06, "loss": 0.0046, "step": 4958 }, { "epoch": 0.59, "learning_rate": 1.2245931283905967e-06, "loss": 0.0004, "step": 4959 }, { "epoch": 0.59, "learning_rate": 1.224231464737794e-06, "loss": 0.0224, "step": 4960 }, { "epoch": 0.59, "learning_rate": 1.223869801084991e-06, "loss": 0.0004, "step": 4961 }, { "epoch": 0.59, "learning_rate": 1.2235081374321882e-06, "loss": 0.0004, "step": 4962 }, { "epoch": 0.59, "learning_rate": 1.2231464737793852e-06, "loss": 0.0021, "step": 4963 }, { "epoch": 0.59, "learning_rate": 1.2227848101265824e-06, "loss": 0.0004, "step": 4964 }, { "epoch": 0.59, "learning_rate": 1.2224231464737794e-06, "loss": 0.0004, "step": 4965 }, { "epoch": 0.6, "learning_rate": 1.2220614828209765e-06, "loss": 0.0019, "step": 4966 }, { "epoch": 0.6, "learning_rate": 1.2216998191681737e-06, "loss": 0.0004, "step": 4967 }, { "epoch": 0.6, "learning_rate": 1.2213381555153707e-06, "loss": 0.0004, "step": 4968 }, { "epoch": 0.6, "learning_rate": 1.2209764918625677e-06, "loss": 0.4654, "step": 4969 }, { "epoch": 0.6, "learning_rate": 1.220614828209765e-06, "loss": 0.9234, "step": 4970 }, { "epoch": 0.6, "learning_rate": 1.220253164556962e-06, "loss": 0.0003, "step": 4971 }, { "epoch": 0.6, "learning_rate": 1.2198915009041592e-06, "loss": 0.0005, "step": 4972 }, { "epoch": 0.6, "learning_rate": 1.2195298372513562e-06, "loss": 0.0005, "step": 4973 }, { "epoch": 0.6, "learning_rate": 1.2191681735985535e-06, "loss": 0.0004, "step": 4974 }, { "epoch": 0.6, "learning_rate": 1.2188065099457505e-06, "loss": 0.0004, "step": 4975 }, { "epoch": 0.6, "learning_rate": 1.2184448462929475e-06, "loss": 0.0005, "step": 4976 }, { "epoch": 0.6, "learning_rate": 1.2180831826401447e-06, "loss": 0.0005, "step": 4977 }, { "epoch": 0.6, "learning_rate": 1.2177215189873418e-06, "loss": 0.0005, "step": 4978 }, { "epoch": 0.6, "learning_rate": 1.217359855334539e-06, "loss": 0.9344, "step": 4979 }, { "epoch": 0.6, "learning_rate": 1.216998191681736e-06, "loss": 0.0004, "step": 4980 }, { "epoch": 0.6, "learning_rate": 1.216636528028933e-06, "loss": 0.0005, "step": 4981 }, { "epoch": 0.6, "learning_rate": 1.2162748643761303e-06, "loss": 0.0006, "step": 4982 }, { "epoch": 0.6, "learning_rate": 1.2159132007233273e-06, "loss": 0.0004, "step": 4983 }, { "epoch": 0.6, "learning_rate": 1.2155515370705245e-06, "loss": 0.0003, "step": 4984 }, { "epoch": 0.6, "learning_rate": 1.2151898734177215e-06, "loss": 0.0004, "step": 4985 }, { "epoch": 0.6, "learning_rate": 1.2148282097649186e-06, "loss": 0.0004, "step": 4986 }, { "epoch": 0.6, "learning_rate": 1.2144665461121158e-06, "loss": 0.0005, "step": 4987 }, { "epoch": 0.6, "learning_rate": 1.2141048824593128e-06, "loss": 0.0007, "step": 4988 }, { "epoch": 0.6, "learning_rate": 1.21374321880651e-06, "loss": 0.0006, "step": 4989 }, { "epoch": 0.6, "learning_rate": 1.213381555153707e-06, "loss": 0.0003, "step": 4990 }, { "epoch": 0.6, "learning_rate": 1.2130198915009043e-06, "loss": 0.0011, "step": 4991 }, { "epoch": 0.6, "learning_rate": 1.2126582278481013e-06, "loss": 0.8305, "step": 4992 }, { "epoch": 0.6, "learning_rate": 1.2122965641952983e-06, "loss": 0.0005, "step": 4993 }, { "epoch": 0.6, "learning_rate": 1.2119349005424956e-06, "loss": 0.0005, "step": 4994 }, { "epoch": 0.6, "learning_rate": 1.2115732368896926e-06, "loss": 0.0004, "step": 4995 }, { "epoch": 0.6, "learning_rate": 1.2112115732368896e-06, "loss": 0.0005, "step": 4996 }, { "epoch": 0.6, "learning_rate": 1.2108499095840868e-06, "loss": 0.0004, "step": 4997 }, { "epoch": 0.6, "learning_rate": 1.2104882459312838e-06, "loss": 0.0006, "step": 4998 }, { "epoch": 0.6, "learning_rate": 1.210126582278481e-06, "loss": 0.0005, "step": 4999 }, { "epoch": 0.6, "learning_rate": 1.209764918625678e-06, "loss": 0.0004, "step": 5000 }, { "epoch": 0.6, "learning_rate": 1.2094032549728753e-06, "loss": 0.0007, "step": 5001 }, { "epoch": 0.6, "learning_rate": 1.2090415913200723e-06, "loss": 0.0018, "step": 5002 }, { "epoch": 0.6, "learning_rate": 1.2086799276672696e-06, "loss": 0.0006, "step": 5003 }, { "epoch": 0.6, "learning_rate": 1.2083182640144666e-06, "loss": 0.8049, "step": 5004 }, { "epoch": 0.6, "learning_rate": 1.2079566003616636e-06, "loss": 0.3383, "step": 5005 }, { "epoch": 0.6, "learning_rate": 1.2075949367088606e-06, "loss": 0.0005, "step": 5006 }, { "epoch": 0.6, "learning_rate": 1.2072332730560579e-06, "loss": 0.8383, "step": 5007 }, { "epoch": 0.6, "learning_rate": 1.2068716094032549e-06, "loss": 0.0005, "step": 5008 }, { "epoch": 0.6, "learning_rate": 1.2065099457504521e-06, "loss": 0.0002, "step": 5009 }, { "epoch": 0.6, "learning_rate": 1.2061482820976491e-06, "loss": 0.143, "step": 5010 }, { "epoch": 0.6, "learning_rate": 1.2057866184448464e-06, "loss": 0.0009, "step": 5011 }, { "epoch": 0.6, "learning_rate": 1.2054249547920434e-06, "loss": 0.0004, "step": 5012 }, { "epoch": 0.6, "learning_rate": 1.2050632911392406e-06, "loss": 0.0005, "step": 5013 }, { "epoch": 0.6, "learning_rate": 1.2047016274864376e-06, "loss": 0.0003, "step": 5014 }, { "epoch": 0.6, "learning_rate": 1.2043399638336349e-06, "loss": 0.0004, "step": 5015 }, { "epoch": 0.6, "learning_rate": 1.2039783001808319e-06, "loss": 0.0004, "step": 5016 }, { "epoch": 0.6, "learning_rate": 1.203616636528029e-06, "loss": 0.0007, "step": 5017 }, { "epoch": 0.6, "learning_rate": 1.203254972875226e-06, "loss": 0.7184, "step": 5018 }, { "epoch": 0.6, "learning_rate": 1.2028933092224232e-06, "loss": 0.0026, "step": 5019 }, { "epoch": 0.6, "learning_rate": 1.2025316455696202e-06, "loss": 0.0006, "step": 5020 }, { "epoch": 0.6, "learning_rate": 1.2021699819168174e-06, "loss": 0.0006, "step": 5021 }, { "epoch": 0.6, "learning_rate": 1.2018083182640144e-06, "loss": 0.015, "step": 5022 }, { "epoch": 0.6, "learning_rate": 1.2014466546112117e-06, "loss": 0.0005, "step": 5023 }, { "epoch": 0.6, "learning_rate": 1.2010849909584087e-06, "loss": 0.0004, "step": 5024 }, { "epoch": 0.6, "learning_rate": 1.200723327305606e-06, "loss": 0.0003, "step": 5025 }, { "epoch": 0.6, "learning_rate": 1.200361663652803e-06, "loss": 0.0006, "step": 5026 }, { "epoch": 0.6, "learning_rate": 1.2000000000000002e-06, "loss": 0.0003, "step": 5027 }, { "epoch": 0.6, "learning_rate": 1.1996383363471972e-06, "loss": 0.0003, "step": 5028 }, { "epoch": 0.6, "learning_rate": 1.1992766726943942e-06, "loss": 0.0004, "step": 5029 }, { "epoch": 0.6, "learning_rate": 1.1989150090415912e-06, "loss": 0.0005, "step": 5030 }, { "epoch": 0.6, "learning_rate": 1.1985533453887884e-06, "loss": 0.0005, "step": 5031 }, { "epoch": 0.6, "learning_rate": 1.1981916817359855e-06, "loss": 0.0003, "step": 5032 }, { "epoch": 0.6, "learning_rate": 1.1978300180831827e-06, "loss": 0.0003, "step": 5033 }, { "epoch": 0.6, "learning_rate": 1.1974683544303797e-06, "loss": 0.0188, "step": 5034 }, { "epoch": 0.6, "learning_rate": 1.197106690777577e-06, "loss": 0.001, "step": 5035 }, { "epoch": 0.6, "learning_rate": 1.196745027124774e-06, "loss": 0.0004, "step": 5036 }, { "epoch": 0.6, "learning_rate": 1.1963833634719712e-06, "loss": 0.8289, "step": 5037 }, { "epoch": 0.6, "learning_rate": 1.1960216998191682e-06, "loss": 0.0016, "step": 5038 }, { "epoch": 0.6, "learning_rate": 1.1956600361663654e-06, "loss": 0.0004, "step": 5039 }, { "epoch": 0.6, "learning_rate": 1.1952983725135625e-06, "loss": 0.0008, "step": 5040 }, { "epoch": 0.6, "learning_rate": 1.1949367088607595e-06, "loss": 0.9076, "step": 5041 }, { "epoch": 0.6, "learning_rate": 1.1945750452079565e-06, "loss": 0.7211, "step": 5042 }, { "epoch": 0.6, "learning_rate": 1.1942133815551537e-06, "loss": 0.0006, "step": 5043 }, { "epoch": 0.6, "learning_rate": 1.1938517179023508e-06, "loss": 0.0004, "step": 5044 }, { "epoch": 0.6, "learning_rate": 1.193490054249548e-06, "loss": 0.0003, "step": 5045 }, { "epoch": 0.6, "learning_rate": 1.193128390596745e-06, "loss": 0.0005, "step": 5046 }, { "epoch": 0.6, "learning_rate": 1.1927667269439422e-06, "loss": 0.009, "step": 5047 }, { "epoch": 0.6, "learning_rate": 1.1924050632911393e-06, "loss": 0.0004, "step": 5048 }, { "epoch": 0.61, "learning_rate": 1.1920433996383365e-06, "loss": 0.0005, "step": 5049 }, { "epoch": 0.61, "learning_rate": 1.1916817359855335e-06, "loss": 0.6364, "step": 5050 }, { "epoch": 0.61, "learning_rate": 1.1913200723327307e-06, "loss": 0.0141, "step": 5051 }, { "epoch": 0.61, "learning_rate": 1.1909584086799278e-06, "loss": 0.0005, "step": 5052 }, { "epoch": 0.61, "learning_rate": 1.1905967450271248e-06, "loss": 0.25, "step": 5053 }, { "epoch": 0.61, "learning_rate": 1.1902350813743218e-06, "loss": 0.005, "step": 5054 }, { "epoch": 0.61, "learning_rate": 1.189873417721519e-06, "loss": 0.003, "step": 5055 }, { "epoch": 0.61, "learning_rate": 1.189511754068716e-06, "loss": 0.0003, "step": 5056 }, { "epoch": 0.61, "learning_rate": 1.1891500904159133e-06, "loss": 0.0005, "step": 5057 }, { "epoch": 0.61, "learning_rate": 1.1887884267631103e-06, "loss": 0.0091, "step": 5058 }, { "epoch": 0.61, "learning_rate": 1.1884267631103075e-06, "loss": 0.0005, "step": 5059 }, { "epoch": 0.61, "learning_rate": 1.1880650994575046e-06, "loss": 0.0008, "step": 5060 }, { "epoch": 0.61, "learning_rate": 1.1877034358047018e-06, "loss": 0.0006, "step": 5061 }, { "epoch": 0.61, "learning_rate": 1.1873417721518988e-06, "loss": 0.0006, "step": 5062 }, { "epoch": 0.61, "learning_rate": 1.186980108499096e-06, "loss": 0.0004, "step": 5063 }, { "epoch": 0.61, "learning_rate": 1.186618444846293e-06, "loss": 0.001, "step": 5064 }, { "epoch": 0.61, "learning_rate": 1.18625678119349e-06, "loss": 0.0004, "step": 5065 }, { "epoch": 0.61, "learning_rate": 1.185895117540687e-06, "loss": 0.0005, "step": 5066 }, { "epoch": 0.61, "learning_rate": 1.1855334538878843e-06, "loss": 0.0003, "step": 5067 }, { "epoch": 0.61, "learning_rate": 1.1851717902350813e-06, "loss": 0.0006, "step": 5068 }, { "epoch": 0.61, "learning_rate": 1.1848101265822786e-06, "loss": 0.0005, "step": 5069 }, { "epoch": 0.61, "learning_rate": 1.1844484629294756e-06, "loss": 0.0005, "step": 5070 }, { "epoch": 0.61, "learning_rate": 1.1840867992766728e-06, "loss": 0.0005, "step": 5071 }, { "epoch": 0.61, "learning_rate": 1.1837251356238698e-06, "loss": 0.003, "step": 5072 }, { "epoch": 0.61, "learning_rate": 1.183363471971067e-06, "loss": 0.0015, "step": 5073 }, { "epoch": 0.61, "learning_rate": 1.183001808318264e-06, "loss": 0.0006, "step": 5074 }, { "epoch": 0.61, "learning_rate": 1.1826401446654613e-06, "loss": 0.0004, "step": 5075 }, { "epoch": 0.61, "learning_rate": 1.1822784810126583e-06, "loss": 0.0019, "step": 5076 }, { "epoch": 0.61, "learning_rate": 1.1819168173598554e-06, "loss": 0.0003, "step": 5077 }, { "epoch": 0.61, "learning_rate": 1.1815551537070524e-06, "loss": 0.0007, "step": 5078 }, { "epoch": 0.61, "learning_rate": 1.1811934900542496e-06, "loss": 0.0005, "step": 5079 }, { "epoch": 0.61, "learning_rate": 1.1808318264014466e-06, "loss": 0.0008, "step": 5080 }, { "epoch": 0.61, "learning_rate": 1.1804701627486439e-06, "loss": 0.4716, "step": 5081 }, { "epoch": 0.61, "learning_rate": 1.1801084990958409e-06, "loss": 0.0004, "step": 5082 }, { "epoch": 0.61, "learning_rate": 1.1797468354430381e-06, "loss": 0.0003, "step": 5083 }, { "epoch": 0.61, "learning_rate": 1.1793851717902351e-06, "loss": 0.0012, "step": 5084 }, { "epoch": 0.61, "learning_rate": 1.1790235081374324e-06, "loss": 0.0004, "step": 5085 }, { "epoch": 0.61, "learning_rate": 1.1786618444846294e-06, "loss": 0.0003, "step": 5086 }, { "epoch": 0.61, "learning_rate": 1.1783001808318264e-06, "loss": 0.8328, "step": 5087 }, { "epoch": 0.61, "learning_rate": 1.1779385171790234e-06, "loss": 0.0003, "step": 5088 }, { "epoch": 0.61, "learning_rate": 1.1775768535262207e-06, "loss": 0.0009, "step": 5089 }, { "epoch": 0.61, "learning_rate": 1.1772151898734177e-06, "loss": 0.0003, "step": 5090 }, { "epoch": 0.61, "learning_rate": 1.176853526220615e-06, "loss": 0.0007, "step": 5091 }, { "epoch": 0.61, "learning_rate": 1.176491862567812e-06, "loss": 0.0004, "step": 5092 }, { "epoch": 0.61, "learning_rate": 1.1761301989150092e-06, "loss": 0.0005, "step": 5093 }, { "epoch": 0.61, "learning_rate": 1.1757685352622062e-06, "loss": 0.0004, "step": 5094 }, { "epoch": 0.61, "learning_rate": 1.1754068716094032e-06, "loss": 0.7363, "step": 5095 }, { "epoch": 0.61, "learning_rate": 1.1750452079566004e-06, "loss": 0.0003, "step": 5096 }, { "epoch": 0.61, "learning_rate": 1.1746835443037974e-06, "loss": 0.0005, "step": 5097 }, { "epoch": 0.61, "learning_rate": 1.1743218806509947e-06, "loss": 0.0006, "step": 5098 }, { "epoch": 0.61, "learning_rate": 1.1739602169981917e-06, "loss": 0.0008, "step": 5099 }, { "epoch": 0.61, "learning_rate": 1.1735985533453887e-06, "loss": 0.0006, "step": 5100 }, { "epoch": 0.61, "learning_rate": 1.173236889692586e-06, "loss": 0.0003, "step": 5101 }, { "epoch": 0.61, "learning_rate": 1.172875226039783e-06, "loss": 0.0004, "step": 5102 }, { "epoch": 0.61, "learning_rate": 1.1725135623869802e-06, "loss": 0.5982, "step": 5103 }, { "epoch": 0.61, "learning_rate": 1.1721518987341772e-06, "loss": 0.0005, "step": 5104 }, { "epoch": 0.61, "learning_rate": 1.1717902350813742e-06, "loss": 0.0006, "step": 5105 }, { "epoch": 0.61, "learning_rate": 1.1714285714285715e-06, "loss": 0.0014, "step": 5106 }, { "epoch": 0.61, "learning_rate": 1.1710669077757685e-06, "loss": 0.0004, "step": 5107 }, { "epoch": 0.61, "learning_rate": 1.1707052441229657e-06, "loss": 0.0003, "step": 5108 }, { "epoch": 0.61, "learning_rate": 1.1703435804701627e-06, "loss": 0.0003, "step": 5109 }, { "epoch": 0.61, "learning_rate": 1.16998191681736e-06, "loss": 0.0071, "step": 5110 }, { "epoch": 0.61, "learning_rate": 1.169620253164557e-06, "loss": 0.0004, "step": 5111 }, { "epoch": 0.61, "learning_rate": 1.169258589511754e-06, "loss": 0.0003, "step": 5112 }, { "epoch": 0.61, "learning_rate": 1.1688969258589512e-06, "loss": 0.491, "step": 5113 }, { "epoch": 0.61, "learning_rate": 1.1685352622061483e-06, "loss": 0.0004, "step": 5114 }, { "epoch": 0.61, "learning_rate": 1.1681735985533453e-06, "loss": 0.0008, "step": 5115 }, { "epoch": 0.61, "learning_rate": 1.1678119349005425e-06, "loss": 0.0018, "step": 5116 }, { "epoch": 0.61, "learning_rate": 1.1674502712477395e-06, "loss": 0.0004, "step": 5117 }, { "epoch": 0.61, "learning_rate": 1.1670886075949368e-06, "loss": 0.0003, "step": 5118 }, { "epoch": 0.61, "learning_rate": 1.1667269439421338e-06, "loss": 0.0003, "step": 5119 }, { "epoch": 0.61, "learning_rate": 1.166365280289331e-06, "loss": 0.0003, "step": 5120 }, { "epoch": 0.61, "learning_rate": 1.166003616636528e-06, "loss": 0.0005, "step": 5121 }, { "epoch": 0.61, "learning_rate": 1.1656419529837253e-06, "loss": 0.0003, "step": 5122 }, { "epoch": 0.61, "learning_rate": 1.1652802893309223e-06, "loss": 0.0008, "step": 5123 }, { "epoch": 0.61, "learning_rate": 1.1649186256781193e-06, "loss": 0.0004, "step": 5124 }, { "epoch": 0.61, "learning_rate": 1.1645569620253163e-06, "loss": 0.0002, "step": 5125 }, { "epoch": 0.61, "learning_rate": 1.1641952983725135e-06, "loss": 0.0005, "step": 5126 }, { "epoch": 0.61, "learning_rate": 1.1638336347197106e-06, "loss": 0.0005, "step": 5127 }, { "epoch": 0.61, "learning_rate": 1.1634719710669078e-06, "loss": 0.0003, "step": 5128 }, { "epoch": 0.61, "learning_rate": 1.1631103074141048e-06, "loss": 0.0003, "step": 5129 }, { "epoch": 0.61, "learning_rate": 1.162748643761302e-06, "loss": 0.0004, "step": 5130 }, { "epoch": 0.61, "learning_rate": 1.162386980108499e-06, "loss": 0.0004, "step": 5131 }, { "epoch": 0.61, "learning_rate": 1.1620253164556963e-06, "loss": 0.1226, "step": 5132 }, { "epoch": 0.62, "learning_rate": 1.1616636528028933e-06, "loss": 0.0004, "step": 5133 }, { "epoch": 0.62, "learning_rate": 1.1613019891500906e-06, "loss": 0.0004, "step": 5134 }, { "epoch": 0.62, "learning_rate": 1.1609403254972876e-06, "loss": 0.0007, "step": 5135 }, { "epoch": 0.62, "learning_rate": 1.1605786618444846e-06, "loss": 0.5672, "step": 5136 }, { "epoch": 0.62, "learning_rate": 1.1602169981916816e-06, "loss": 0.0008, "step": 5137 }, { "epoch": 0.62, "learning_rate": 1.1598553345388788e-06, "loss": 0.0004, "step": 5138 }, { "epoch": 0.62, "learning_rate": 1.1594936708860759e-06, "loss": 0.0004, "step": 5139 }, { "epoch": 0.62, "learning_rate": 1.159132007233273e-06, "loss": 0.0003, "step": 5140 }, { "epoch": 0.62, "learning_rate": 1.1587703435804701e-06, "loss": 0.0004, "step": 5141 }, { "epoch": 0.62, "learning_rate": 1.1584086799276673e-06, "loss": 0.0004, "step": 5142 }, { "epoch": 0.62, "learning_rate": 1.1580470162748644e-06, "loss": 1.5234, "step": 5143 }, { "epoch": 0.62, "learning_rate": 1.1576853526220616e-06, "loss": 0.0003, "step": 5144 }, { "epoch": 0.62, "learning_rate": 1.1573236889692586e-06, "loss": 0.7839, "step": 5145 }, { "epoch": 0.62, "learning_rate": 1.1569620253164558e-06, "loss": 0.0003, "step": 5146 }, { "epoch": 0.62, "learning_rate": 1.1566003616636529e-06, "loss": 0.0003, "step": 5147 }, { "epoch": 0.62, "learning_rate": 1.1562386980108499e-06, "loss": 0.0004, "step": 5148 }, { "epoch": 0.62, "learning_rate": 1.155877034358047e-06, "loss": 0.0003, "step": 5149 }, { "epoch": 0.62, "learning_rate": 1.1555153707052441e-06, "loss": 0.0004, "step": 5150 }, { "epoch": 0.62, "learning_rate": 1.1551537070524412e-06, "loss": 0.0005, "step": 5151 }, { "epoch": 0.62, "learning_rate": 1.1547920433996384e-06, "loss": 0.0006, "step": 5152 }, { "epoch": 0.62, "learning_rate": 1.1544303797468354e-06, "loss": 0.0004, "step": 5153 }, { "epoch": 0.62, "learning_rate": 1.1540687160940326e-06, "loss": 0.0004, "step": 5154 }, { "epoch": 0.62, "learning_rate": 1.1537070524412297e-06, "loss": 0.0003, "step": 5155 }, { "epoch": 0.62, "learning_rate": 1.1533453887884269e-06, "loss": 0.0003, "step": 5156 }, { "epoch": 0.62, "learning_rate": 1.152983725135624e-06, "loss": 0.0002, "step": 5157 }, { "epoch": 0.62, "learning_rate": 1.1526220614828211e-06, "loss": 0.0003, "step": 5158 }, { "epoch": 0.62, "learning_rate": 1.1522603978300182e-06, "loss": 0.0003, "step": 5159 }, { "epoch": 0.62, "learning_rate": 1.1518987341772152e-06, "loss": 0.0004, "step": 5160 }, { "epoch": 0.62, "learning_rate": 1.1515370705244122e-06, "loss": 0.0003, "step": 5161 }, { "epoch": 0.62, "learning_rate": 1.1511754068716094e-06, "loss": 0.0005, "step": 5162 }, { "epoch": 0.62, "learning_rate": 1.1508137432188064e-06, "loss": 0.8662, "step": 5163 }, { "epoch": 0.62, "learning_rate": 1.1504520795660037e-06, "loss": 0.0003, "step": 5164 }, { "epoch": 0.62, "learning_rate": 1.1500904159132007e-06, "loss": 0.0005, "step": 5165 }, { "epoch": 0.62, "learning_rate": 1.149728752260398e-06, "loss": 0.0004, "step": 5166 }, { "epoch": 0.62, "learning_rate": 1.149367088607595e-06, "loss": 0.0003, "step": 5167 }, { "epoch": 0.62, "learning_rate": 1.1490054249547922e-06, "loss": 0.0007, "step": 5168 }, { "epoch": 0.62, "learning_rate": 1.1486437613019892e-06, "loss": 0.0003, "step": 5169 }, { "epoch": 0.62, "learning_rate": 1.1482820976491864e-06, "loss": 0.4474, "step": 5170 }, { "epoch": 0.62, "learning_rate": 1.1479204339963834e-06, "loss": 0.0004, "step": 5171 }, { "epoch": 0.62, "learning_rate": 1.1475587703435805e-06, "loss": 0.0006, "step": 5172 }, { "epoch": 0.62, "learning_rate": 1.1471971066907775e-06, "loss": 0.0004, "step": 5173 }, { "epoch": 0.62, "learning_rate": 1.1468354430379747e-06, "loss": 0.9849, "step": 5174 }, { "epoch": 0.62, "learning_rate": 1.1464737793851717e-06, "loss": 0.9426, "step": 5175 }, { "epoch": 0.62, "learning_rate": 1.146112115732369e-06, "loss": 0.0007, "step": 5176 }, { "epoch": 0.62, "learning_rate": 1.145750452079566e-06, "loss": 0.0004, "step": 5177 }, { "epoch": 0.62, "learning_rate": 1.1453887884267632e-06, "loss": 0.0002, "step": 5178 }, { "epoch": 0.62, "learning_rate": 1.1450271247739602e-06, "loss": 0.0004, "step": 5179 }, { "epoch": 0.62, "learning_rate": 1.1446654611211575e-06, "loss": 0.0002, "step": 5180 }, { "epoch": 0.62, "learning_rate": 1.1443037974683545e-06, "loss": 0.0009, "step": 5181 }, { "epoch": 0.62, "learning_rate": 1.1439421338155517e-06, "loss": 0.0004, "step": 5182 }, { "epoch": 0.62, "learning_rate": 1.1435804701627487e-06, "loss": 0.0005, "step": 5183 }, { "epoch": 0.62, "learning_rate": 1.1432188065099458e-06, "loss": 1.5782, "step": 5184 }, { "epoch": 0.62, "learning_rate": 1.1428571428571428e-06, "loss": 0.0004, "step": 5185 }, { "epoch": 0.62, "learning_rate": 1.14249547920434e-06, "loss": 0.0005, "step": 5186 }, { "epoch": 0.62, "learning_rate": 1.142133815551537e-06, "loss": 0.0004, "step": 5187 }, { "epoch": 0.62, "learning_rate": 1.1417721518987343e-06, "loss": 0.0003, "step": 5188 }, { "epoch": 0.62, "learning_rate": 1.1414104882459313e-06, "loss": 0.0007, "step": 5189 }, { "epoch": 0.62, "learning_rate": 1.1410488245931285e-06, "loss": 0.0993, "step": 5190 }, { "epoch": 0.62, "learning_rate": 1.1406871609403255e-06, "loss": 0.0004, "step": 5191 }, { "epoch": 0.62, "learning_rate": 1.1403254972875228e-06, "loss": 0.0006, "step": 5192 }, { "epoch": 0.62, "learning_rate": 1.1399638336347198e-06, "loss": 0.1335, "step": 5193 }, { "epoch": 0.62, "learning_rate": 1.139602169981917e-06, "loss": 0.0005, "step": 5194 }, { "epoch": 0.62, "learning_rate": 1.139240506329114e-06, "loss": 0.0013, "step": 5195 }, { "epoch": 0.62, "learning_rate": 1.138878842676311e-06, "loss": 0.0004, "step": 5196 }, { "epoch": 0.62, "learning_rate": 1.138517179023508e-06, "loss": 0.0464, "step": 5197 }, { "epoch": 0.62, "learning_rate": 1.1381555153707053e-06, "loss": 0.0006, "step": 5198 }, { "epoch": 0.62, "learning_rate": 1.1377938517179023e-06, "loss": 0.0007, "step": 5199 }, { "epoch": 0.62, "learning_rate": 1.1374321880650995e-06, "loss": 0.0027, "step": 5200 }, { "epoch": 0.62, "learning_rate": 1.1370705244122966e-06, "loss": 0.0005, "step": 5201 }, { "epoch": 0.62, "learning_rate": 1.1367088607594938e-06, "loss": 0.0003, "step": 5202 }, { "epoch": 0.62, "learning_rate": 1.1363471971066908e-06, "loss": 0.0004, "step": 5203 }, { "epoch": 0.62, "learning_rate": 1.135985533453888e-06, "loss": 0.0006, "step": 5204 }, { "epoch": 0.62, "learning_rate": 1.135623869801085e-06, "loss": 0.0007, "step": 5205 }, { "epoch": 0.62, "learning_rate": 1.135262206148282e-06, "loss": 0.0004, "step": 5206 }, { "epoch": 0.62, "learning_rate": 1.1349005424954793e-06, "loss": 0.0006, "step": 5207 }, { "epoch": 0.62, "learning_rate": 1.1345388788426763e-06, "loss": 0.0003, "step": 5208 }, { "epoch": 0.62, "learning_rate": 1.1341772151898734e-06, "loss": 0.0008, "step": 5209 }, { "epoch": 0.62, "learning_rate": 1.1338155515370706e-06, "loss": 0.0005, "step": 5210 }, { "epoch": 0.62, "learning_rate": 1.1334538878842676e-06, "loss": 0.0007, "step": 5211 }, { "epoch": 0.62, "learning_rate": 1.1330922242314648e-06, "loss": 0.0008, "step": 5212 }, { "epoch": 0.62, "learning_rate": 1.1327305605786619e-06, "loss": 0.0008, "step": 5213 }, { "epoch": 0.62, "learning_rate": 1.132368896925859e-06, "loss": 0.0003, "step": 5214 }, { "epoch": 0.62, "learning_rate": 1.1320072332730561e-06, "loss": 0.0006, "step": 5215 }, { "epoch": 0.63, "learning_rate": 1.1316455696202531e-06, "loss": 0.633, "step": 5216 }, { "epoch": 0.63, "learning_rate": 1.1312839059674504e-06, "loss": 0.0005, "step": 5217 }, { "epoch": 0.63, "learning_rate": 1.1309222423146474e-06, "loss": 0.0016, "step": 5218 }, { "epoch": 0.63, "learning_rate": 1.1305605786618446e-06, "loss": 0.0004, "step": 5219 }, { "epoch": 0.63, "learning_rate": 1.1301989150090416e-06, "loss": 0.0007, "step": 5220 }, { "epoch": 0.63, "learning_rate": 1.1298372513562386e-06, "loss": 0.0005, "step": 5221 }, { "epoch": 0.63, "learning_rate": 1.1294755877034359e-06, "loss": 0.0004, "step": 5222 }, { "epoch": 0.63, "learning_rate": 1.129113924050633e-06, "loss": 0.4368, "step": 5223 }, { "epoch": 0.63, "learning_rate": 1.1287522603978301e-06, "loss": 0.0002, "step": 5224 }, { "epoch": 0.63, "learning_rate": 1.1283905967450272e-06, "loss": 0.0004, "step": 5225 }, { "epoch": 0.63, "learning_rate": 1.1280289330922242e-06, "loss": 0.0004, "step": 5226 }, { "epoch": 0.63, "learning_rate": 1.1276672694394214e-06, "loss": 0.0003, "step": 5227 }, { "epoch": 0.63, "learning_rate": 1.1273056057866184e-06, "loss": 0.5689, "step": 5228 }, { "epoch": 0.63, "learning_rate": 1.1269439421338157e-06, "loss": 0.0006, "step": 5229 }, { "epoch": 0.63, "learning_rate": 1.1265822784810127e-06, "loss": 0.0005, "step": 5230 }, { "epoch": 0.63, "learning_rate": 1.12622061482821e-06, "loss": 0.0004, "step": 5231 }, { "epoch": 0.63, "learning_rate": 1.125858951175407e-06, "loss": 0.0008, "step": 5232 }, { "epoch": 0.63, "learning_rate": 1.125497287522604e-06, "loss": 0.0004, "step": 5233 }, { "epoch": 0.63, "learning_rate": 1.125135623869801e-06, "loss": 0.0013, "step": 5234 }, { "epoch": 0.63, "learning_rate": 1.1247739602169982e-06, "loss": 0.0015, "step": 5235 }, { "epoch": 0.63, "learning_rate": 1.1244122965641952e-06, "loss": 0.8335, "step": 5236 }, { "epoch": 0.63, "learning_rate": 1.1240506329113924e-06, "loss": 0.0006, "step": 5237 }, { "epoch": 0.63, "learning_rate": 1.1236889692585895e-06, "loss": 0.8209, "step": 5238 }, { "epoch": 0.63, "learning_rate": 1.1233273056057867e-06, "loss": 0.0002, "step": 5239 }, { "epoch": 0.63, "learning_rate": 1.1229656419529837e-06, "loss": 0.0003, "step": 5240 }, { "epoch": 0.63, "learning_rate": 1.122603978300181e-06, "loss": 0.0004, "step": 5241 }, { "epoch": 0.63, "learning_rate": 1.122242314647378e-06, "loss": 0.0008, "step": 5242 }, { "epoch": 0.63, "learning_rate": 1.1218806509945752e-06, "loss": 0.0004, "step": 5243 }, { "epoch": 0.63, "learning_rate": 1.121518987341772e-06, "loss": 0.0003, "step": 5244 }, { "epoch": 0.63, "learning_rate": 1.1211573236889692e-06, "loss": 0.0003, "step": 5245 }, { "epoch": 0.63, "learning_rate": 1.1207956600361663e-06, "loss": 0.4467, "step": 5246 }, { "epoch": 0.63, "learning_rate": 1.1204339963833635e-06, "loss": 0.0003, "step": 5247 }, { "epoch": 0.63, "learning_rate": 1.1200723327305605e-06, "loss": 0.0009, "step": 5248 }, { "epoch": 0.63, "learning_rate": 1.1197106690777577e-06, "loss": 0.0103, "step": 5249 }, { "epoch": 0.63, "learning_rate": 1.1193490054249548e-06, "loss": 0.0003, "step": 5250 }, { "epoch": 0.63, "learning_rate": 1.118987341772152e-06, "loss": 0.0006, "step": 5251 }, { "epoch": 0.63, "learning_rate": 1.118625678119349e-06, "loss": 0.0003, "step": 5252 }, { "epoch": 0.63, "learning_rate": 1.1182640144665462e-06, "loss": 0.0006, "step": 5253 }, { "epoch": 0.63, "learning_rate": 1.1179023508137433e-06, "loss": 0.0003, "step": 5254 }, { "epoch": 0.63, "learning_rate": 1.1175406871609405e-06, "loss": 0.0005, "step": 5255 }, { "epoch": 0.63, "learning_rate": 1.1171790235081373e-06, "loss": 0.0005, "step": 5256 }, { "epoch": 0.63, "learning_rate": 1.1168173598553345e-06, "loss": 0.0009, "step": 5257 }, { "epoch": 0.63, "learning_rate": 1.1164556962025315e-06, "loss": 0.0004, "step": 5258 }, { "epoch": 0.63, "learning_rate": 1.1160940325497288e-06, "loss": 0.0004, "step": 5259 }, { "epoch": 0.63, "learning_rate": 1.1157323688969258e-06, "loss": 0.662, "step": 5260 }, { "epoch": 0.63, "learning_rate": 1.115370705244123e-06, "loss": 0.0003, "step": 5261 }, { "epoch": 0.63, "learning_rate": 1.11500904159132e-06, "loss": 0.0004, "step": 5262 }, { "epoch": 0.63, "learning_rate": 1.1146473779385173e-06, "loss": 0.0004, "step": 5263 }, { "epoch": 0.63, "learning_rate": 1.1142857142857143e-06, "loss": 0.0003, "step": 5264 }, { "epoch": 0.63, "learning_rate": 1.1139240506329115e-06, "loss": 0.0475, "step": 5265 }, { "epoch": 0.63, "learning_rate": 1.1135623869801085e-06, "loss": 0.0003, "step": 5266 }, { "epoch": 0.63, "learning_rate": 1.1132007233273058e-06, "loss": 0.0006, "step": 5267 }, { "epoch": 0.63, "learning_rate": 1.1128390596745026e-06, "loss": 0.0003, "step": 5268 }, { "epoch": 0.63, "learning_rate": 1.1124773960216998e-06, "loss": 0.0003, "step": 5269 }, { "epoch": 0.63, "learning_rate": 1.1121157323688968e-06, "loss": 0.3917, "step": 5270 }, { "epoch": 0.63, "learning_rate": 1.111754068716094e-06, "loss": 0.7273, "step": 5271 }, { "epoch": 0.63, "learning_rate": 1.111392405063291e-06, "loss": 0.0004, "step": 5272 }, { "epoch": 0.63, "learning_rate": 1.1110307414104883e-06, "loss": 0.0456, "step": 5273 }, { "epoch": 0.63, "learning_rate": 1.1106690777576853e-06, "loss": 0.0004, "step": 5274 }, { "epoch": 0.63, "learning_rate": 1.1103074141048826e-06, "loss": 0.0003, "step": 5275 }, { "epoch": 0.63, "learning_rate": 1.1099457504520796e-06, "loss": 0.0003, "step": 5276 }, { "epoch": 0.63, "learning_rate": 1.1095840867992768e-06, "loss": 0.0005, "step": 5277 }, { "epoch": 0.63, "learning_rate": 1.1092224231464738e-06, "loss": 0.0002, "step": 5278 }, { "epoch": 0.63, "learning_rate": 1.108860759493671e-06, "loss": 0.0003, "step": 5279 }, { "epoch": 0.63, "learning_rate": 1.1084990958408679e-06, "loss": 0.0003, "step": 5280 }, { "epoch": 0.63, "learning_rate": 1.1081374321880651e-06, "loss": 0.0004, "step": 5281 }, { "epoch": 0.63, "learning_rate": 1.1077757685352621e-06, "loss": 0.0003, "step": 5282 }, { "epoch": 0.63, "learning_rate": 1.1074141048824594e-06, "loss": 0.0004, "step": 5283 }, { "epoch": 0.63, "learning_rate": 1.1070524412296564e-06, "loss": 0.0003, "step": 5284 }, { "epoch": 0.63, "learning_rate": 1.1066907775768536e-06, "loss": 0.0007, "step": 5285 }, { "epoch": 0.63, "learning_rate": 1.1063291139240506e-06, "loss": 0.0004, "step": 5286 }, { "epoch": 0.63, "learning_rate": 1.1059674502712479e-06, "loss": 0.0005, "step": 5287 }, { "epoch": 0.63, "learning_rate": 1.1056057866184449e-06, "loss": 0.0003, "step": 5288 }, { "epoch": 0.63, "learning_rate": 1.1052441229656421e-06, "loss": 0.0003, "step": 5289 }, { "epoch": 0.63, "learning_rate": 1.1048824593128391e-06, "loss": 0.0003, "step": 5290 }, { "epoch": 0.63, "learning_rate": 1.1045207956600364e-06, "loss": 0.0003, "step": 5291 }, { "epoch": 0.63, "learning_rate": 1.1041591320072332e-06, "loss": 0.0005, "step": 5292 }, { "epoch": 0.63, "learning_rate": 1.1037974683544304e-06, "loss": 0.0052, "step": 5293 }, { "epoch": 0.63, "learning_rate": 1.1034358047016274e-06, "loss": 0.0003, "step": 5294 }, { "epoch": 0.63, "learning_rate": 1.1030741410488247e-06, "loss": 0.0004, "step": 5295 }, { "epoch": 0.63, "learning_rate": 1.1027124773960217e-06, "loss": 0.0004, "step": 5296 }, { "epoch": 0.63, "learning_rate": 1.102350813743219e-06, "loss": 0.0003, "step": 5297 }, { "epoch": 0.63, "learning_rate": 1.101989150090416e-06, "loss": 0.0004, "step": 5298 }, { "epoch": 0.63, "learning_rate": 1.1016274864376132e-06, "loss": 0.0007, "step": 5299 }, { "epoch": 0.64, "learning_rate": 1.1012658227848102e-06, "loss": 0.0003, "step": 5300 }, { "epoch": 0.64, "learning_rate": 1.1009041591320074e-06, "loss": 0.0005, "step": 5301 }, { "epoch": 0.64, "learning_rate": 1.1005424954792044e-06, "loss": 0.0131, "step": 5302 }, { "epoch": 0.64, "learning_rate": 1.1001808318264017e-06, "loss": 0.0003, "step": 5303 }, { "epoch": 0.64, "learning_rate": 1.0998191681735985e-06, "loss": 0.0007, "step": 5304 }, { "epoch": 0.64, "learning_rate": 1.0994575045207957e-06, "loss": 0.0003, "step": 5305 }, { "epoch": 0.64, "learning_rate": 1.0990958408679927e-06, "loss": 0.0003, "step": 5306 }, { "epoch": 0.64, "learning_rate": 1.09873417721519e-06, "loss": 0.0003, "step": 5307 }, { "epoch": 0.64, "learning_rate": 1.098372513562387e-06, "loss": 0.0002, "step": 5308 }, { "epoch": 0.64, "learning_rate": 1.0980108499095842e-06, "loss": 0.0066, "step": 5309 }, { "epoch": 0.64, "learning_rate": 1.0976491862567812e-06, "loss": 0.0005, "step": 5310 }, { "epoch": 0.64, "learning_rate": 1.0972875226039784e-06, "loss": 0.0004, "step": 5311 }, { "epoch": 0.64, "learning_rate": 1.0969258589511755e-06, "loss": 0.0002, "step": 5312 }, { "epoch": 0.64, "learning_rate": 1.0965641952983727e-06, "loss": 0.0299, "step": 5313 }, { "epoch": 0.64, "learning_rate": 1.0962025316455697e-06, "loss": 0.0004, "step": 5314 }, { "epoch": 0.64, "learning_rate": 1.0958408679927667e-06, "loss": 0.0008, "step": 5315 }, { "epoch": 0.64, "learning_rate": 1.0954792043399638e-06, "loss": 0.0003, "step": 5316 }, { "epoch": 0.64, "learning_rate": 1.095117540687161e-06, "loss": 0.9359, "step": 5317 }, { "epoch": 0.64, "learning_rate": 1.094755877034358e-06, "loss": 0.0006, "step": 5318 }, { "epoch": 0.64, "learning_rate": 1.0943942133815552e-06, "loss": 0.0003, "step": 5319 }, { "epoch": 0.64, "learning_rate": 1.0940325497287523e-06, "loss": 0.0004, "step": 5320 }, { "epoch": 0.64, "learning_rate": 1.0936708860759495e-06, "loss": 0.7477, "step": 5321 }, { "epoch": 0.64, "learning_rate": 1.0933092224231465e-06, "loss": 0.0004, "step": 5322 }, { "epoch": 0.64, "learning_rate": 1.0929475587703437e-06, "loss": 0.0003, "step": 5323 }, { "epoch": 0.64, "learning_rate": 1.0925858951175408e-06, "loss": 0.0003, "step": 5324 }, { "epoch": 0.64, "learning_rate": 1.0922242314647378e-06, "loss": 0.0003, "step": 5325 }, { "epoch": 0.64, "learning_rate": 1.091862567811935e-06, "loss": 0.0003, "step": 5326 }, { "epoch": 0.64, "learning_rate": 1.091500904159132e-06, "loss": 0.0004, "step": 5327 }, { "epoch": 0.64, "learning_rate": 1.091139240506329e-06, "loss": 0.2927, "step": 5328 }, { "epoch": 0.64, "learning_rate": 1.0907775768535263e-06, "loss": 0.0004, "step": 5329 }, { "epoch": 0.64, "learning_rate": 1.0904159132007233e-06, "loss": 0.6995, "step": 5330 }, { "epoch": 0.64, "learning_rate": 1.0900542495479205e-06, "loss": 0.4272, "step": 5331 }, { "epoch": 0.64, "learning_rate": 1.0896925858951175e-06, "loss": 0.0003, "step": 5332 }, { "epoch": 0.64, "learning_rate": 1.0893309222423148e-06, "loss": 0.0006, "step": 5333 }, { "epoch": 0.64, "learning_rate": 1.0889692585895118e-06, "loss": 0.0003, "step": 5334 }, { "epoch": 0.64, "learning_rate": 1.0886075949367088e-06, "loss": 0.0009, "step": 5335 }, { "epoch": 0.64, "learning_rate": 1.088245931283906e-06, "loss": 0.0004, "step": 5336 }, { "epoch": 0.64, "learning_rate": 1.087884267631103e-06, "loss": 0.526, "step": 5337 }, { "epoch": 0.64, "learning_rate": 1.0875226039783003e-06, "loss": 0.0628, "step": 5338 }, { "epoch": 0.64, "learning_rate": 1.0871609403254973e-06, "loss": 0.0002, "step": 5339 }, { "epoch": 0.64, "learning_rate": 1.0867992766726943e-06, "loss": 0.8904, "step": 5340 }, { "epoch": 0.64, "learning_rate": 1.0864376130198916e-06, "loss": 0.0002, "step": 5341 }, { "epoch": 0.64, "learning_rate": 1.0860759493670886e-06, "loss": 0.0004, "step": 5342 }, { "epoch": 0.64, "learning_rate": 1.0857142857142858e-06, "loss": 0.0005, "step": 5343 }, { "epoch": 0.64, "learning_rate": 1.0853526220614828e-06, "loss": 0.0004, "step": 5344 }, { "epoch": 0.64, "learning_rate": 1.0849909584086799e-06, "loss": 0.0003, "step": 5345 }, { "epoch": 0.64, "learning_rate": 1.084629294755877e-06, "loss": 0.0003, "step": 5346 }, { "epoch": 0.64, "learning_rate": 1.084267631103074e-06, "loss": 0.0002, "step": 5347 }, { "epoch": 0.64, "learning_rate": 1.0839059674502713e-06, "loss": 0.0002, "step": 5348 }, { "epoch": 0.64, "learning_rate": 1.0835443037974684e-06, "loss": 0.0257, "step": 5349 }, { "epoch": 0.64, "learning_rate": 1.0831826401446656e-06, "loss": 0.0003, "step": 5350 }, { "epoch": 0.64, "learning_rate": 1.0828209764918626e-06, "loss": 0.0004, "step": 5351 }, { "epoch": 0.64, "learning_rate": 1.0824593128390596e-06, "loss": 0.0006, "step": 5352 }, { "epoch": 0.64, "learning_rate": 1.0820976491862569e-06, "loss": 0.0004, "step": 5353 }, { "epoch": 0.64, "learning_rate": 1.0817359855334539e-06, "loss": 0.0003, "step": 5354 }, { "epoch": 0.64, "learning_rate": 1.081374321880651e-06, "loss": 0.0004, "step": 5355 }, { "epoch": 0.64, "learning_rate": 1.0810126582278481e-06, "loss": 0.0003, "step": 5356 }, { "epoch": 0.64, "learning_rate": 1.0806509945750451e-06, "loss": 0.305, "step": 5357 }, { "epoch": 0.64, "learning_rate": 1.0802893309222424e-06, "loss": 0.0003, "step": 5358 }, { "epoch": 0.64, "learning_rate": 1.0799276672694394e-06, "loss": 0.0002, "step": 5359 }, { "epoch": 0.64, "learning_rate": 1.0795660036166366e-06, "loss": 0.0271, "step": 5360 }, { "epoch": 0.64, "learning_rate": 1.0792043399638336e-06, "loss": 0.3117, "step": 5361 }, { "epoch": 0.64, "learning_rate": 1.0788426763110309e-06, "loss": 0.0004, "step": 5362 }, { "epoch": 0.64, "learning_rate": 1.078481012658228e-06, "loss": 0.0002, "step": 5363 }, { "epoch": 0.64, "learning_rate": 1.078119349005425e-06, "loss": 0.0003, "step": 5364 }, { "epoch": 0.64, "learning_rate": 1.077757685352622e-06, "loss": 0.0002, "step": 5365 }, { "epoch": 0.64, "learning_rate": 1.0773960216998192e-06, "loss": 0.0003, "step": 5366 }, { "epoch": 0.64, "learning_rate": 1.0770343580470162e-06, "loss": 0.0029, "step": 5367 }, { "epoch": 0.64, "learning_rate": 1.0766726943942134e-06, "loss": 0.0003, "step": 5368 }, { "epoch": 0.64, "learning_rate": 1.0763110307414104e-06, "loss": 0.0002, "step": 5369 }, { "epoch": 0.64, "learning_rate": 1.0759493670886077e-06, "loss": 0.0004, "step": 5370 }, { "epoch": 0.64, "learning_rate": 1.0755877034358047e-06, "loss": 0.0003, "step": 5371 }, { "epoch": 0.64, "learning_rate": 1.075226039783002e-06, "loss": 0.0003, "step": 5372 }, { "epoch": 0.64, "learning_rate": 1.074864376130199e-06, "loss": 0.0003, "step": 5373 }, { "epoch": 0.64, "learning_rate": 1.0745027124773962e-06, "loss": 0.0006, "step": 5374 }, { "epoch": 0.64, "learning_rate": 1.074141048824593e-06, "loss": 0.0003, "step": 5375 }, { "epoch": 0.64, "learning_rate": 1.0737793851717902e-06, "loss": 0.0005, "step": 5376 }, { "epoch": 0.64, "learning_rate": 1.0734177215189872e-06, "loss": 0.0002, "step": 5377 }, { "epoch": 0.64, "learning_rate": 1.0730560578661845e-06, "loss": 0.0003, "step": 5378 }, { "epoch": 0.64, "learning_rate": 1.0726943942133815e-06, "loss": 0.0002, "step": 5379 }, { "epoch": 0.64, "learning_rate": 1.0723327305605787e-06, "loss": 0.0002, "step": 5380 }, { "epoch": 0.64, "learning_rate": 1.0719710669077757e-06, "loss": 0.0002, "step": 5381 }, { "epoch": 0.64, "learning_rate": 1.071609403254973e-06, "loss": 0.0003, "step": 5382 }, { "epoch": 0.65, "learning_rate": 1.07124773960217e-06, "loss": 0.0481, "step": 5383 }, { "epoch": 0.65, "learning_rate": 1.0708860759493672e-06, "loss": 0.0003, "step": 5384 }, { "epoch": 0.65, "learning_rate": 1.0705244122965642e-06, "loss": 0.0009, "step": 5385 }, { "epoch": 0.65, "learning_rate": 1.0701627486437615e-06, "loss": 0.0029, "step": 5386 }, { "epoch": 0.65, "learning_rate": 1.0698010849909583e-06, "loss": 0.0003, "step": 5387 }, { "epoch": 0.65, "learning_rate": 1.0694394213381555e-06, "loss": 0.0003, "step": 5388 }, { "epoch": 0.65, "learning_rate": 1.0690777576853525e-06, "loss": 0.0004, "step": 5389 }, { "epoch": 0.65, "learning_rate": 1.0687160940325498e-06, "loss": 0.0007, "step": 5390 }, { "epoch": 0.65, "learning_rate": 1.0683544303797468e-06, "loss": 0.0003, "step": 5391 }, { "epoch": 0.65, "learning_rate": 1.067992766726944e-06, "loss": 1.3908, "step": 5392 }, { "epoch": 0.65, "learning_rate": 1.067631103074141e-06, "loss": 0.0003, "step": 5393 }, { "epoch": 0.65, "learning_rate": 1.0672694394213383e-06, "loss": 0.0002, "step": 5394 }, { "epoch": 0.65, "learning_rate": 1.0669077757685353e-06, "loss": 0.0002, "step": 5395 }, { "epoch": 0.65, "learning_rate": 1.0665461121157325e-06, "loss": 0.0002, "step": 5396 }, { "epoch": 0.65, "learning_rate": 1.0661844484629295e-06, "loss": 0.0002, "step": 5397 }, { "epoch": 0.65, "learning_rate": 1.0658227848101268e-06, "loss": 0.0004, "step": 5398 }, { "epoch": 0.65, "learning_rate": 1.0654611211573236e-06, "loss": 0.0003, "step": 5399 }, { "epoch": 0.65, "learning_rate": 1.0650994575045208e-06, "loss": 0.0002, "step": 5400 }, { "epoch": 0.65, "learning_rate": 1.0647377938517178e-06, "loss": 0.0002, "step": 5401 }, { "epoch": 0.65, "learning_rate": 1.064376130198915e-06, "loss": 0.0003, "step": 5402 }, { "epoch": 0.65, "learning_rate": 1.064014466546112e-06, "loss": 0.7133, "step": 5403 }, { "epoch": 0.65, "learning_rate": 1.0636528028933093e-06, "loss": 0.0009, "step": 5404 }, { "epoch": 0.65, "learning_rate": 1.0632911392405063e-06, "loss": 0.0002, "step": 5405 }, { "epoch": 0.65, "learning_rate": 1.0629294755877035e-06, "loss": 0.0003, "step": 5406 }, { "epoch": 0.65, "learning_rate": 1.0625678119349006e-06, "loss": 0.0003, "step": 5407 }, { "epoch": 0.65, "learning_rate": 1.0622061482820978e-06, "loss": 0.1394, "step": 5408 }, { "epoch": 0.65, "learning_rate": 1.0618444846292948e-06, "loss": 0.0003, "step": 5409 }, { "epoch": 0.65, "learning_rate": 1.061482820976492e-06, "loss": 0.0027, "step": 5410 }, { "epoch": 0.65, "learning_rate": 1.0611211573236889e-06, "loss": 0.9566, "step": 5411 }, { "epoch": 0.65, "learning_rate": 1.060759493670886e-06, "loss": 0.0002, "step": 5412 }, { "epoch": 0.65, "learning_rate": 1.060397830018083e-06, "loss": 0.0026, "step": 5413 }, { "epoch": 0.65, "learning_rate": 1.0600361663652803e-06, "loss": 0.5651, "step": 5414 }, { "epoch": 0.65, "learning_rate": 1.0596745027124774e-06, "loss": 0.0002, "step": 5415 }, { "epoch": 0.65, "learning_rate": 1.0593128390596746e-06, "loss": 1.3473, "step": 5416 }, { "epoch": 0.65, "learning_rate": 1.0589511754068716e-06, "loss": 0.0003, "step": 5417 }, { "epoch": 0.65, "learning_rate": 1.0585895117540688e-06, "loss": 0.0003, "step": 5418 }, { "epoch": 0.65, "learning_rate": 1.0582278481012659e-06, "loss": 0.0002, "step": 5419 }, { "epoch": 0.65, "learning_rate": 1.057866184448463e-06, "loss": 0.0016, "step": 5420 }, { "epoch": 0.65, "learning_rate": 1.05750452079566e-06, "loss": 0.0031, "step": 5421 }, { "epoch": 0.65, "learning_rate": 1.0571428571428573e-06, "loss": 0.0002, "step": 5422 }, { "epoch": 0.65, "learning_rate": 1.0567811934900541e-06, "loss": 0.0004, "step": 5423 }, { "epoch": 0.65, "learning_rate": 1.0564195298372514e-06, "loss": 0.0003, "step": 5424 }, { "epoch": 0.65, "learning_rate": 1.0560578661844484e-06, "loss": 0.0059, "step": 5425 }, { "epoch": 0.65, "learning_rate": 1.0556962025316456e-06, "loss": 1.6123, "step": 5426 }, { "epoch": 0.65, "learning_rate": 1.0553345388788426e-06, "loss": 0.0002, "step": 5427 }, { "epoch": 0.65, "learning_rate": 1.0549728752260399e-06, "loss": 0.0003, "step": 5428 }, { "epoch": 0.65, "learning_rate": 1.054611211573237e-06, "loss": 0.0002, "step": 5429 }, { "epoch": 0.65, "learning_rate": 1.0542495479204341e-06, "loss": 0.0003, "step": 5430 }, { "epoch": 0.65, "learning_rate": 1.0538878842676311e-06, "loss": 0.0002, "step": 5431 }, { "epoch": 0.65, "learning_rate": 1.0535262206148284e-06, "loss": 0.9622, "step": 5432 }, { "epoch": 0.65, "learning_rate": 1.0531645569620254e-06, "loss": 0.0002, "step": 5433 }, { "epoch": 0.65, "learning_rate": 1.0528028933092226e-06, "loss": 0.0003, "step": 5434 }, { "epoch": 0.65, "learning_rate": 1.0524412296564194e-06, "loss": 0.001, "step": 5435 }, { "epoch": 0.65, "learning_rate": 1.0520795660036167e-06, "loss": 0.0002, "step": 5436 }, { "epoch": 0.65, "learning_rate": 1.0517179023508137e-06, "loss": 0.0229, "step": 5437 }, { "epoch": 0.65, "learning_rate": 1.051356238698011e-06, "loss": 0.0004, "step": 5438 }, { "epoch": 0.65, "learning_rate": 1.050994575045208e-06, "loss": 0.0002, "step": 5439 }, { "epoch": 0.65, "learning_rate": 1.0506329113924052e-06, "loss": 0.0002, "step": 5440 }, { "epoch": 0.65, "learning_rate": 1.0502712477396022e-06, "loss": 0.0002, "step": 5441 }, { "epoch": 0.65, "learning_rate": 1.0499095840867994e-06, "loss": 0.0005, "step": 5442 }, { "epoch": 0.65, "learning_rate": 1.0495479204339964e-06, "loss": 0.0003, "step": 5443 }, { "epoch": 0.65, "learning_rate": 1.0491862567811937e-06, "loss": 0.0005, "step": 5444 }, { "epoch": 0.65, "learning_rate": 1.0488245931283907e-06, "loss": 0.0005, "step": 5445 }, { "epoch": 0.65, "learning_rate": 1.0484629294755877e-06, "loss": 0.0002, "step": 5446 }, { "epoch": 0.65, "learning_rate": 1.0481012658227847e-06, "loss": 0.0002, "step": 5447 }, { "epoch": 0.65, "learning_rate": 1.047739602169982e-06, "loss": 0.0003, "step": 5448 }, { "epoch": 0.65, "learning_rate": 1.047377938517179e-06, "loss": 0.0002, "step": 5449 }, { "epoch": 0.65, "learning_rate": 1.0470162748643762e-06, "loss": 0.0003, "step": 5450 }, { "epoch": 0.65, "learning_rate": 1.0466546112115732e-06, "loss": 0.0004, "step": 5451 }, { "epoch": 0.65, "learning_rate": 1.0462929475587705e-06, "loss": 0.0019, "step": 5452 }, { "epoch": 0.65, "learning_rate": 1.0459312839059675e-06, "loss": 0.0002, "step": 5453 }, { "epoch": 0.65, "learning_rate": 1.0455696202531645e-06, "loss": 0.0004, "step": 5454 }, { "epoch": 0.65, "learning_rate": 1.0452079566003617e-06, "loss": 0.0016, "step": 5455 }, { "epoch": 0.65, "learning_rate": 1.0448462929475587e-06, "loss": 0.0003, "step": 5456 }, { "epoch": 0.65, "learning_rate": 1.044484629294756e-06, "loss": 0.0002, "step": 5457 }, { "epoch": 0.65, "learning_rate": 1.044122965641953e-06, "loss": 0.6202, "step": 5458 }, { "epoch": 0.65, "learning_rate": 1.04376130198915e-06, "loss": 0.0003, "step": 5459 }, { "epoch": 0.65, "learning_rate": 1.0433996383363473e-06, "loss": 0.0004, "step": 5460 }, { "epoch": 0.65, "learning_rate": 1.0430379746835443e-06, "loss": 0.0003, "step": 5461 }, { "epoch": 0.65, "learning_rate": 1.0426763110307415e-06, "loss": 0.0003, "step": 5462 }, { "epoch": 0.65, "learning_rate": 1.0423146473779385e-06, "loss": 0.0004, "step": 5463 }, { "epoch": 0.65, "learning_rate": 1.0419529837251355e-06, "loss": 0.0005, "step": 5464 }, { "epoch": 0.65, "learning_rate": 1.0415913200723328e-06, "loss": 0.0003, "step": 5465 }, { "epoch": 0.66, "learning_rate": 1.0412296564195298e-06, "loss": 0.0003, "step": 5466 }, { "epoch": 0.66, "learning_rate": 1.040867992766727e-06, "loss": 0.0002, "step": 5467 }, { "epoch": 0.66, "learning_rate": 1.040506329113924e-06, "loss": 0.0003, "step": 5468 }, { "epoch": 0.66, "learning_rate": 1.0401446654611213e-06, "loss": 0.0004, "step": 5469 }, { "epoch": 0.66, "learning_rate": 1.0397830018083183e-06, "loss": 0.9907, "step": 5470 }, { "epoch": 0.66, "learning_rate": 1.0394213381555153e-06, "loss": 0.0002, "step": 5471 }, { "epoch": 0.66, "learning_rate": 1.0390596745027125e-06, "loss": 0.0003, "step": 5472 }, { "epoch": 0.66, "learning_rate": 1.0386980108499096e-06, "loss": 0.0003, "step": 5473 }, { "epoch": 0.66, "learning_rate": 1.0383363471971066e-06, "loss": 0.0003, "step": 5474 }, { "epoch": 0.66, "learning_rate": 1.0379746835443038e-06, "loss": 0.0003, "step": 5475 }, { "epoch": 0.66, "learning_rate": 1.0376130198915008e-06, "loss": 0.0004, "step": 5476 }, { "epoch": 0.66, "learning_rate": 1.037251356238698e-06, "loss": 0.0003, "step": 5477 }, { "epoch": 0.66, "learning_rate": 1.036889692585895e-06, "loss": 0.0002, "step": 5478 }, { "epoch": 0.66, "learning_rate": 1.0365280289330923e-06, "loss": 0.0003, "step": 5479 }, { "epoch": 0.66, "learning_rate": 1.0361663652802893e-06, "loss": 0.0002, "step": 5480 }, { "epoch": 0.66, "learning_rate": 1.0358047016274866e-06, "loss": 0.0004, "step": 5481 }, { "epoch": 0.66, "learning_rate": 1.0354430379746836e-06, "loss": 0.0007, "step": 5482 }, { "epoch": 0.66, "learning_rate": 1.0350813743218806e-06, "loss": 0.0004, "step": 5483 }, { "epoch": 0.66, "learning_rate": 1.0347197106690776e-06, "loss": 0.0003, "step": 5484 }, { "epoch": 0.66, "learning_rate": 1.0343580470162749e-06, "loss": 0.7035, "step": 5485 }, { "epoch": 0.66, "learning_rate": 1.0339963833634719e-06, "loss": 0.0011, "step": 5486 }, { "epoch": 0.66, "learning_rate": 1.033634719710669e-06, "loss": 0.0002, "step": 5487 }, { "epoch": 0.66, "learning_rate": 1.0332730560578661e-06, "loss": 0.0003, "step": 5488 }, { "epoch": 0.66, "learning_rate": 1.0329113924050634e-06, "loss": 0.0054, "step": 5489 }, { "epoch": 0.66, "learning_rate": 1.0325497287522604e-06, "loss": 0.0004, "step": 5490 }, { "epoch": 0.66, "learning_rate": 1.0321880650994576e-06, "loss": 0.0003, "step": 5491 }, { "epoch": 0.66, "learning_rate": 1.0318264014466546e-06, "loss": 0.8776, "step": 5492 }, { "epoch": 0.66, "learning_rate": 1.0314647377938519e-06, "loss": 0.0003, "step": 5493 }, { "epoch": 0.66, "learning_rate": 1.0311030741410489e-06, "loss": 0.0002, "step": 5494 }, { "epoch": 0.66, "learning_rate": 1.030741410488246e-06, "loss": 0.0003, "step": 5495 }, { "epoch": 0.66, "learning_rate": 1.030379746835443e-06, "loss": 0.0004, "step": 5496 }, { "epoch": 0.66, "learning_rate": 1.0300180831826401e-06, "loss": 0.9327, "step": 5497 }, { "epoch": 0.66, "learning_rate": 1.0296564195298372e-06, "loss": 0.0002, "step": 5498 }, { "epoch": 0.66, "learning_rate": 1.0292947558770344e-06, "loss": 0.0003, "step": 5499 }, { "epoch": 0.66, "learning_rate": 1.0289330922242314e-06, "loss": 0.0003, "step": 5500 }, { "epoch": 0.66, "learning_rate": 1.0285714285714286e-06, "loss": 0.0004, "step": 5501 }, { "epoch": 0.66, "learning_rate": 1.0282097649186257e-06, "loss": 0.0002, "step": 5502 }, { "epoch": 0.66, "learning_rate": 1.027848101265823e-06, "loss": 0.0005, "step": 5503 }, { "epoch": 0.66, "learning_rate": 1.02748643761302e-06, "loss": 0.3938, "step": 5504 }, { "epoch": 0.66, "learning_rate": 1.0271247739602171e-06, "loss": 0.0001, "step": 5505 }, { "epoch": 0.66, "learning_rate": 1.0267631103074142e-06, "loss": 0.0004, "step": 5506 }, { "epoch": 0.66, "learning_rate": 1.0264014466546112e-06, "loss": 0.0003, "step": 5507 }, { "epoch": 0.66, "learning_rate": 1.0260397830018082e-06, "loss": 0.0005, "step": 5508 }, { "epoch": 0.66, "learning_rate": 1.0256781193490054e-06, "loss": 0.0004, "step": 5509 }, { "epoch": 0.66, "learning_rate": 1.0253164556962025e-06, "loss": 0.0003, "step": 5510 }, { "epoch": 0.66, "learning_rate": 1.0249547920433997e-06, "loss": 0.9047, "step": 5511 }, { "epoch": 0.66, "learning_rate": 1.0245931283905967e-06, "loss": 0.0112, "step": 5512 }, { "epoch": 0.66, "learning_rate": 1.024231464737794e-06, "loss": 0.7263, "step": 5513 }, { "epoch": 0.66, "learning_rate": 1.023869801084991e-06, "loss": 0.1249, "step": 5514 }, { "epoch": 0.66, "learning_rate": 1.0235081374321882e-06, "loss": 0.0003, "step": 5515 }, { "epoch": 0.66, "learning_rate": 1.0231464737793852e-06, "loss": 0.0003, "step": 5516 }, { "epoch": 0.66, "learning_rate": 1.0227848101265824e-06, "loss": 0.0003, "step": 5517 }, { "epoch": 0.66, "learning_rate": 1.0224231464737795e-06, "loss": 0.0003, "step": 5518 }, { "epoch": 0.66, "learning_rate": 1.0220614828209765e-06, "loss": 0.0004, "step": 5519 }, { "epoch": 0.66, "learning_rate": 1.0216998191681735e-06, "loss": 0.0003, "step": 5520 }, { "epoch": 0.66, "learning_rate": 1.0213381555153707e-06, "loss": 0.0002, "step": 5521 }, { "epoch": 0.66, "learning_rate": 1.0209764918625677e-06, "loss": 0.0012, "step": 5522 }, { "epoch": 0.66, "learning_rate": 1.020614828209765e-06, "loss": 0.0003, "step": 5523 }, { "epoch": 0.66, "learning_rate": 1.020253164556962e-06, "loss": 0.0003, "step": 5524 }, { "epoch": 0.66, "learning_rate": 1.0198915009041592e-06, "loss": 0.0004, "step": 5525 }, { "epoch": 0.66, "learning_rate": 1.0195298372513562e-06, "loss": 0.0002, "step": 5526 }, { "epoch": 0.66, "learning_rate": 1.0191681735985535e-06, "loss": 0.0011, "step": 5527 }, { "epoch": 0.66, "learning_rate": 1.0188065099457505e-06, "loss": 0.0003, "step": 5528 }, { "epoch": 0.66, "learning_rate": 1.0184448462929477e-06, "loss": 0.8913, "step": 5529 }, { "epoch": 0.66, "learning_rate": 1.0180831826401447e-06, "loss": 0.0004, "step": 5530 }, { "epoch": 0.66, "learning_rate": 1.0177215189873418e-06, "loss": 0.0003, "step": 5531 }, { "epoch": 0.66, "learning_rate": 1.0173598553345388e-06, "loss": 0.0002, "step": 5532 }, { "epoch": 0.66, "learning_rate": 1.016998191681736e-06, "loss": 0.0003, "step": 5533 }, { "epoch": 0.66, "learning_rate": 1.016636528028933e-06, "loss": 0.0002, "step": 5534 }, { "epoch": 0.66, "learning_rate": 1.0162748643761303e-06, "loss": 0.4973, "step": 5535 }, { "epoch": 0.66, "learning_rate": 1.0159132007233273e-06, "loss": 0.0003, "step": 5536 }, { "epoch": 0.66, "learning_rate": 1.0155515370705245e-06, "loss": 0.0002, "step": 5537 }, { "epoch": 0.66, "learning_rate": 1.0151898734177215e-06, "loss": 0.0013, "step": 5538 }, { "epoch": 0.66, "learning_rate": 1.0148282097649188e-06, "loss": 0.0002, "step": 5539 }, { "epoch": 0.66, "learning_rate": 1.0144665461121158e-06, "loss": 0.0003, "step": 5540 }, { "epoch": 0.66, "learning_rate": 1.014104882459313e-06, "loss": 0.0003, "step": 5541 }, { "epoch": 0.66, "learning_rate": 1.01374321880651e-06, "loss": 0.0006, "step": 5542 }, { "epoch": 0.66, "learning_rate": 1.013381555153707e-06, "loss": 0.0004, "step": 5543 }, { "epoch": 0.66, "learning_rate": 1.013019891500904e-06, "loss": 0.0005, "step": 5544 }, { "epoch": 0.66, "learning_rate": 1.0126582278481013e-06, "loss": 0.0003, "step": 5545 }, { "epoch": 0.66, "learning_rate": 1.0122965641952983e-06, "loss": 0.0003, "step": 5546 }, { "epoch": 0.66, "learning_rate": 1.0119349005424956e-06, "loss": 0.0003, "step": 5547 }, { "epoch": 0.66, "learning_rate": 1.0115732368896926e-06, "loss": 0.0002, "step": 5548 }, { "epoch": 0.66, "learning_rate": 1.0112115732368898e-06, "loss": 0.0002, "step": 5549 }, { "epoch": 0.67, "learning_rate": 1.0108499095840868e-06, "loss": 0.0003, "step": 5550 }, { "epoch": 0.67, "learning_rate": 1.010488245931284e-06, "loss": 0.0003, "step": 5551 }, { "epoch": 0.67, "learning_rate": 1.010126582278481e-06, "loss": 0.0003, "step": 5552 }, { "epoch": 0.67, "learning_rate": 1.0097649186256783e-06, "loss": 0.0003, "step": 5553 }, { "epoch": 0.67, "learning_rate": 1.0094032549728753e-06, "loss": 0.0003, "step": 5554 }, { "epoch": 0.67, "learning_rate": 1.0090415913200724e-06, "loss": 0.0003, "step": 5555 }, { "epoch": 0.67, "learning_rate": 1.0086799276672694e-06, "loss": 0.0002, "step": 5556 }, { "epoch": 0.67, "learning_rate": 1.0083182640144666e-06, "loss": 0.1687, "step": 5557 }, { "epoch": 0.67, "learning_rate": 1.0079566003616636e-06, "loss": 0.0003, "step": 5558 }, { "epoch": 0.67, "learning_rate": 1.0075949367088609e-06, "loss": 0.0015, "step": 5559 }, { "epoch": 0.67, "learning_rate": 1.0072332730560579e-06, "loss": 0.0003, "step": 5560 }, { "epoch": 0.67, "learning_rate": 1.006871609403255e-06, "loss": 0.0002, "step": 5561 }, { "epoch": 0.67, "learning_rate": 1.0065099457504521e-06, "loss": 0.0004, "step": 5562 }, { "epoch": 0.67, "learning_rate": 1.0061482820976494e-06, "loss": 0.0064, "step": 5563 }, { "epoch": 0.67, "learning_rate": 1.0057866184448464e-06, "loss": 0.0005, "step": 5564 }, { "epoch": 0.67, "learning_rate": 1.0054249547920434e-06, "loss": 0.0002, "step": 5565 }, { "epoch": 0.67, "learning_rate": 1.0050632911392406e-06, "loss": 0.0003, "step": 5566 }, { "epoch": 0.67, "learning_rate": 1.0047016274864376e-06, "loss": 0.0003, "step": 5567 }, { "epoch": 0.67, "learning_rate": 1.0043399638336347e-06, "loss": 0.0004, "step": 5568 }, { "epoch": 0.67, "learning_rate": 1.003978300180832e-06, "loss": 0.0002, "step": 5569 }, { "epoch": 0.67, "learning_rate": 1.003616636528029e-06, "loss": 0.0001, "step": 5570 }, { "epoch": 0.67, "learning_rate": 1.0032549728752261e-06, "loss": 0.5598, "step": 5571 }, { "epoch": 0.67, "learning_rate": 1.0028933092224232e-06, "loss": 0.0003, "step": 5572 }, { "epoch": 0.67, "learning_rate": 1.0025316455696204e-06, "loss": 0.0173, "step": 5573 }, { "epoch": 0.67, "learning_rate": 1.0021699819168174e-06, "loss": 0.0002, "step": 5574 }, { "epoch": 0.67, "learning_rate": 1.0018083182640144e-06, "loss": 0.0003, "step": 5575 }, { "epoch": 0.67, "learning_rate": 1.0014466546112117e-06, "loss": 0.0003, "step": 5576 }, { "epoch": 0.67, "learning_rate": 1.0010849909584087e-06, "loss": 0.0002, "step": 5577 }, { "epoch": 0.67, "learning_rate": 1.000723327305606e-06, "loss": 0.0005, "step": 5578 }, { "epoch": 0.67, "learning_rate": 1.000361663652803e-06, "loss": 0.0003, "step": 5579 }, { "epoch": 0.67, "learning_rate": 1e-06, "loss": 0.0003, "step": 5580 }, { "epoch": 0.67, "learning_rate": 9.996383363471972e-07, "loss": 0.0007, "step": 5581 }, { "epoch": 0.67, "learning_rate": 9.992766726943942e-07, "loss": 0.5781, "step": 5582 }, { "epoch": 0.67, "learning_rate": 9.989150090415914e-07, "loss": 0.0002, "step": 5583 }, { "epoch": 0.67, "learning_rate": 9.985533453887885e-07, "loss": 0.0005, "step": 5584 }, { "epoch": 0.67, "learning_rate": 9.981916817359855e-07, "loss": 0.0003, "step": 5585 }, { "epoch": 0.67, "learning_rate": 9.978300180831827e-07, "loss": 0.0731, "step": 5586 }, { "epoch": 0.67, "learning_rate": 9.974683544303797e-07, "loss": 0.0005, "step": 5587 }, { "epoch": 0.67, "learning_rate": 9.97106690777577e-07, "loss": 0.0002, "step": 5588 }, { "epoch": 0.67, "learning_rate": 9.96745027124774e-07, "loss": 0.0003, "step": 5589 }, { "epoch": 0.67, "learning_rate": 9.96383363471971e-07, "loss": 0.0047, "step": 5590 }, { "epoch": 0.67, "learning_rate": 9.960216998191682e-07, "loss": 0.0014, "step": 5591 }, { "epoch": 0.67, "learning_rate": 9.956600361663652e-07, "loss": 0.0003, "step": 5592 }, { "epoch": 0.67, "learning_rate": 9.952983725135623e-07, "loss": 0.001, "step": 5593 }, { "epoch": 0.67, "learning_rate": 9.949367088607595e-07, "loss": 0.0002, "step": 5594 }, { "epoch": 0.67, "learning_rate": 9.945750452079565e-07, "loss": 0.0004, "step": 5595 }, { "epoch": 0.67, "learning_rate": 9.942133815551537e-07, "loss": 0.0003, "step": 5596 }, { "epoch": 0.67, "learning_rate": 9.938517179023508e-07, "loss": 0.0004, "step": 5597 }, { "epoch": 0.67, "learning_rate": 9.93490054249548e-07, "loss": 1.1013, "step": 5598 }, { "epoch": 0.67, "learning_rate": 9.93128390596745e-07, "loss": 0.0003, "step": 5599 }, { "epoch": 0.67, "learning_rate": 9.927667269439422e-07, "loss": 0.0003, "step": 5600 }, { "epoch": 0.67, "learning_rate": 9.924050632911393e-07, "loss": 0.0008, "step": 5601 }, { "epoch": 0.67, "learning_rate": 9.920433996383363e-07, "loss": 0.0003, "step": 5602 }, { "epoch": 0.67, "learning_rate": 9.916817359855333e-07, "loss": 0.0003, "step": 5603 }, { "epoch": 0.67, "learning_rate": 9.913200723327305e-07, "loss": 0.0007, "step": 5604 }, { "epoch": 0.67, "learning_rate": 9.909584086799276e-07, "loss": 0.0002, "step": 5605 }, { "epoch": 0.67, "learning_rate": 9.905967450271248e-07, "loss": 0.0006, "step": 5606 }, { "epoch": 0.67, "learning_rate": 9.902350813743218e-07, "loss": 0.0082, "step": 5607 }, { "epoch": 0.67, "learning_rate": 9.89873417721519e-07, "loss": 0.8321, "step": 5608 }, { "epoch": 0.67, "learning_rate": 9.89511754068716e-07, "loss": 0.0002, "step": 5609 }, { "epoch": 0.67, "learning_rate": 9.891500904159133e-07, "loss": 0.0003, "step": 5610 }, { "epoch": 0.67, "learning_rate": 9.887884267631103e-07, "loss": 0.0002, "step": 5611 }, { "epoch": 0.67, "learning_rate": 9.884267631103075e-07, "loss": 0.0002, "step": 5612 }, { "epoch": 0.67, "learning_rate": 9.880650994575046e-07, "loss": 0.579, "step": 5613 }, { "epoch": 0.67, "learning_rate": 9.877034358047016e-07, "loss": 0.002, "step": 5614 }, { "epoch": 0.67, "learning_rate": 9.873417721518986e-07, "loss": 0.0002, "step": 5615 }, { "epoch": 0.67, "learning_rate": 9.869801084990958e-07, "loss": 0.0004, "step": 5616 }, { "epoch": 0.67, "learning_rate": 9.866184448462928e-07, "loss": 0.0005, "step": 5617 }, { "epoch": 0.67, "learning_rate": 9.8625678119349e-07, "loss": 0.0022, "step": 5618 }, { "epoch": 0.67, "learning_rate": 9.85895117540687e-07, "loss": 0.0002, "step": 5619 }, { "epoch": 0.67, "learning_rate": 9.855334538878843e-07, "loss": 0.0002, "step": 5620 }, { "epoch": 0.67, "learning_rate": 9.851717902350814e-07, "loss": 0.0002, "step": 5621 }, { "epoch": 0.67, "learning_rate": 9.848101265822786e-07, "loss": 0.0002, "step": 5622 }, { "epoch": 0.67, "learning_rate": 9.844484629294756e-07, "loss": 0.0002, "step": 5623 }, { "epoch": 0.67, "learning_rate": 9.840867992766728e-07, "loss": 0.0024, "step": 5624 }, { "epoch": 0.67, "learning_rate": 9.837251356238699e-07, "loss": 0.0003, "step": 5625 }, { "epoch": 0.67, "learning_rate": 9.833634719710669e-07, "loss": 0.0004, "step": 5626 }, { "epoch": 0.67, "learning_rate": 9.830018083182639e-07, "loss": 0.0003, "step": 5627 }, { "epoch": 0.67, "learning_rate": 9.826401446654611e-07, "loss": 0.0002, "step": 5628 }, { "epoch": 0.67, "learning_rate": 9.822784810126581e-07, "loss": 0.0002, "step": 5629 }, { "epoch": 0.67, "learning_rate": 9.819168173598554e-07, "loss": 0.0021, "step": 5630 }, { "epoch": 0.67, "learning_rate": 9.815551537070524e-07, "loss": 0.0003, "step": 5631 }, { "epoch": 0.67, "learning_rate": 9.811934900542496e-07, "loss": 0.0002, "step": 5632 }, { "epoch": 0.68, "learning_rate": 9.808318264014466e-07, "loss": 0.0004, "step": 5633 }, { "epoch": 0.68, "learning_rate": 9.804701627486439e-07, "loss": 0.0003, "step": 5634 }, { "epoch": 0.68, "learning_rate": 9.801084990958409e-07, "loss": 0.1934, "step": 5635 }, { "epoch": 0.68, "learning_rate": 9.797468354430381e-07, "loss": 0.825, "step": 5636 }, { "epoch": 0.68, "learning_rate": 9.793851717902351e-07, "loss": 0.0004, "step": 5637 }, { "epoch": 0.68, "learning_rate": 9.790235081374322e-07, "loss": 0.0003, "step": 5638 }, { "epoch": 0.68, "learning_rate": 9.786618444846292e-07, "loss": 0.001, "step": 5639 }, { "epoch": 0.68, "learning_rate": 9.783001808318264e-07, "loss": 0.0002, "step": 5640 }, { "epoch": 0.68, "learning_rate": 9.779385171790234e-07, "loss": 0.0003, "step": 5641 }, { "epoch": 0.68, "learning_rate": 9.775768535262207e-07, "loss": 0.0003, "step": 5642 }, { "epoch": 0.68, "learning_rate": 9.772151898734177e-07, "loss": 0.0002, "step": 5643 }, { "epoch": 0.68, "learning_rate": 9.76853526220615e-07, "loss": 0.0003, "step": 5644 }, { "epoch": 0.68, "learning_rate": 9.76491862567812e-07, "loss": 0.0002, "step": 5645 }, { "epoch": 0.68, "learning_rate": 9.761301989150092e-07, "loss": 0.0002, "step": 5646 }, { "epoch": 0.68, "learning_rate": 9.757685352622062e-07, "loss": 0.0007, "step": 5647 }, { "epoch": 0.68, "learning_rate": 9.754068716094034e-07, "loss": 0.0003, "step": 5648 }, { "epoch": 0.68, "learning_rate": 9.750452079566004e-07, "loss": 0.0031, "step": 5649 }, { "epoch": 0.68, "learning_rate": 9.746835443037975e-07, "loss": 0.0002, "step": 5650 }, { "epoch": 0.68, "learning_rate": 9.743218806509945e-07, "loss": 0.0003, "step": 5651 }, { "epoch": 0.68, "learning_rate": 9.739602169981917e-07, "loss": 0.1507, "step": 5652 }, { "epoch": 0.68, "learning_rate": 9.735985533453887e-07, "loss": 0.0003, "step": 5653 }, { "epoch": 0.68, "learning_rate": 9.73236889692586e-07, "loss": 0.0003, "step": 5654 }, { "epoch": 0.68, "learning_rate": 9.72875226039783e-07, "loss": 0.0007, "step": 5655 }, { "epoch": 0.68, "learning_rate": 9.725135623869802e-07, "loss": 0.6346, "step": 5656 }, { "epoch": 0.68, "learning_rate": 9.721518987341772e-07, "loss": 0.0003, "step": 5657 }, { "epoch": 0.68, "learning_rate": 9.717902350813745e-07, "loss": 0.0003, "step": 5658 }, { "epoch": 0.68, "learning_rate": 9.714285714285715e-07, "loss": 0.0002, "step": 5659 }, { "epoch": 0.68, "learning_rate": 9.710669077757687e-07, "loss": 0.0002, "step": 5660 }, { "epoch": 0.68, "learning_rate": 9.707052441229657e-07, "loss": 0.0002, "step": 5661 }, { "epoch": 0.68, "learning_rate": 9.703435804701627e-07, "loss": 0.0002, "step": 5662 }, { "epoch": 0.68, "learning_rate": 9.699819168173598e-07, "loss": 0.0003, "step": 5663 }, { "epoch": 0.68, "learning_rate": 9.69620253164557e-07, "loss": 0.0005, "step": 5664 }, { "epoch": 0.68, "learning_rate": 9.69258589511754e-07, "loss": 0.0004, "step": 5665 }, { "epoch": 0.68, "learning_rate": 9.688969258589512e-07, "loss": 0.9518, "step": 5666 }, { "epoch": 0.68, "learning_rate": 9.685352622061483e-07, "loss": 0.0011, "step": 5667 }, { "epoch": 0.68, "learning_rate": 9.681735985533455e-07, "loss": 0.0006, "step": 5668 }, { "epoch": 0.68, "learning_rate": 9.678119349005425e-07, "loss": 0.5331, "step": 5669 }, { "epoch": 0.68, "learning_rate": 9.674502712477397e-07, "loss": 0.0003, "step": 5670 }, { "epoch": 0.68, "learning_rate": 9.670886075949368e-07, "loss": 0.0002, "step": 5671 }, { "epoch": 0.68, "learning_rate": 9.66726943942134e-07, "loss": 0.0002, "step": 5672 }, { "epoch": 0.68, "learning_rate": 9.66365280289331e-07, "loss": 0.0003, "step": 5673 }, { "epoch": 0.68, "learning_rate": 9.66003616636528e-07, "loss": 0.0003, "step": 5674 }, { "epoch": 0.68, "learning_rate": 9.65641952983725e-07, "loss": 0.0002, "step": 5675 }, { "epoch": 0.68, "learning_rate": 9.652802893309223e-07, "loss": 0.002, "step": 5676 }, { "epoch": 0.68, "learning_rate": 9.649186256781193e-07, "loss": 0.0004, "step": 5677 }, { "epoch": 0.68, "learning_rate": 9.645569620253165e-07, "loss": 0.0003, "step": 5678 }, { "epoch": 0.68, "learning_rate": 9.641952983725136e-07, "loss": 0.0003, "step": 5679 }, { "epoch": 0.68, "learning_rate": 9.638336347197108e-07, "loss": 0.0002, "step": 5680 }, { "epoch": 0.68, "learning_rate": 9.634719710669078e-07, "loss": 0.0003, "step": 5681 }, { "epoch": 0.68, "learning_rate": 9.63110307414105e-07, "loss": 0.0003, "step": 5682 }, { "epoch": 0.68, "learning_rate": 9.62748643761302e-07, "loss": 0.0002, "step": 5683 }, { "epoch": 0.68, "learning_rate": 9.62386980108499e-07, "loss": 0.092, "step": 5684 }, { "epoch": 0.68, "learning_rate": 9.620253164556963e-07, "loss": 0.0001, "step": 5685 }, { "epoch": 0.68, "learning_rate": 9.616636528028933e-07, "loss": 0.4645, "step": 5686 }, { "epoch": 0.68, "learning_rate": 9.613019891500903e-07, "loss": 0.0003, "step": 5687 }, { "epoch": 0.68, "learning_rate": 9.609403254972876e-07, "loss": 0.0002, "step": 5688 }, { "epoch": 0.68, "learning_rate": 9.605786618444846e-07, "loss": 1.0413, "step": 5689 }, { "epoch": 0.68, "learning_rate": 9.602169981916818e-07, "loss": 0.0003, "step": 5690 }, { "epoch": 0.68, "learning_rate": 9.598553345388788e-07, "loss": 0.0277, "step": 5691 }, { "epoch": 0.68, "learning_rate": 9.59493670886076e-07, "loss": 0.0003, "step": 5692 }, { "epoch": 0.68, "learning_rate": 9.59132007233273e-07, "loss": 0.0066, "step": 5693 }, { "epoch": 0.68, "learning_rate": 9.587703435804701e-07, "loss": 0.0002, "step": 5694 }, { "epoch": 0.68, "learning_rate": 9.584086799276674e-07, "loss": 0.0002, "step": 5695 }, { "epoch": 0.68, "learning_rate": 9.580470162748644e-07, "loss": 0.0006, "step": 5696 }, { "epoch": 0.68, "learning_rate": 9.576853526220616e-07, "loss": 0.0004, "step": 5697 }, { "epoch": 0.68, "learning_rate": 9.573236889692586e-07, "loss": 0.0003, "step": 5698 }, { "epoch": 0.68, "learning_rate": 9.569620253164556e-07, "loss": 0.0002, "step": 5699 }, { "epoch": 0.68, "learning_rate": 9.566003616636529e-07, "loss": 0.0003, "step": 5700 }, { "epoch": 0.68, "learning_rate": 9.562386980108499e-07, "loss": 0.0003, "step": 5701 }, { "epoch": 0.68, "learning_rate": 9.558770343580471e-07, "loss": 0.7408, "step": 5702 }, { "epoch": 0.68, "learning_rate": 9.555153707052441e-07, "loss": 0.0002, "step": 5703 }, { "epoch": 0.68, "learning_rate": 9.551537070524412e-07, "loss": 0.0003, "step": 5704 }, { "epoch": 0.68, "learning_rate": 9.547920433996384e-07, "loss": 0.0002, "step": 5705 }, { "epoch": 0.68, "learning_rate": 9.544303797468354e-07, "loss": 0.0003, "step": 5706 }, { "epoch": 0.68, "learning_rate": 9.540687160940326e-07, "loss": 0.0002, "step": 5707 }, { "epoch": 0.68, "learning_rate": 9.537070524412297e-07, "loss": 0.0003, "step": 5708 }, { "epoch": 0.68, "learning_rate": 9.533453887884269e-07, "loss": 0.0003, "step": 5709 }, { "epoch": 0.68, "learning_rate": 9.529837251356238e-07, "loss": 0.0004, "step": 5710 }, { "epoch": 0.68, "learning_rate": 9.526220614828209e-07, "loss": 0.0003, "step": 5711 }, { "epoch": 0.68, "learning_rate": 9.522603978300181e-07, "loss": 0.0002, "step": 5712 }, { "epoch": 0.68, "learning_rate": 9.518987341772152e-07, "loss": 0.0002, "step": 5713 }, { "epoch": 0.68, "learning_rate": 9.515370705244123e-07, "loss": 0.0006, "step": 5714 }, { "epoch": 0.68, "learning_rate": 9.511754068716094e-07, "loss": 0.0002, "step": 5715 }, { "epoch": 0.68, "learning_rate": 9.508137432188066e-07, "loss": 0.0002, "step": 5716 }, { "epoch": 0.69, "learning_rate": 9.504520795660037e-07, "loss": 0.5864, "step": 5717 }, { "epoch": 0.69, "learning_rate": 9.500904159132008e-07, "loss": 0.0005, "step": 5718 }, { "epoch": 0.69, "learning_rate": 9.497287522603979e-07, "loss": 0.0002, "step": 5719 }, { "epoch": 0.69, "learning_rate": 9.493670886075951e-07, "loss": 0.0002, "step": 5720 }, { "epoch": 0.69, "learning_rate": 9.490054249547922e-07, "loss": 0.0007, "step": 5721 }, { "epoch": 0.69, "learning_rate": 9.486437613019891e-07, "loss": 0.0145, "step": 5722 }, { "epoch": 0.69, "learning_rate": 9.482820976491862e-07, "loss": 0.0002, "step": 5723 }, { "epoch": 0.69, "learning_rate": 9.479204339963833e-07, "loss": 0.5746, "step": 5724 }, { "epoch": 0.69, "learning_rate": 9.475587703435805e-07, "loss": 0.0004, "step": 5725 }, { "epoch": 0.69, "learning_rate": 9.471971066907776e-07, "loss": 0.0002, "step": 5726 }, { "epoch": 0.69, "learning_rate": 9.468354430379747e-07, "loss": 0.0004, "step": 5727 }, { "epoch": 0.69, "learning_rate": 9.464737793851718e-07, "loss": 0.0002, "step": 5728 }, { "epoch": 0.69, "learning_rate": 9.46112115732369e-07, "loss": 0.0002, "step": 5729 }, { "epoch": 0.69, "learning_rate": 9.457504520795661e-07, "loss": 0.0002, "step": 5730 }, { "epoch": 0.69, "learning_rate": 9.453887884267632e-07, "loss": 0.891, "step": 5731 }, { "epoch": 0.69, "learning_rate": 9.450271247739604e-07, "loss": 0.0003, "step": 5732 }, { "epoch": 0.69, "learning_rate": 9.446654611211575e-07, "loss": 0.0074, "step": 5733 }, { "epoch": 0.69, "learning_rate": 9.443037974683544e-07, "loss": 0.0015, "step": 5734 }, { "epoch": 0.69, "learning_rate": 9.439421338155515e-07, "loss": 0.0004, "step": 5735 }, { "epoch": 0.69, "learning_rate": 9.435804701627486e-07, "loss": 0.0002, "step": 5736 }, { "epoch": 0.69, "learning_rate": 9.432188065099458e-07, "loss": 0.0003, "step": 5737 }, { "epoch": 0.69, "learning_rate": 9.428571428571429e-07, "loss": 0.0003, "step": 5738 }, { "epoch": 0.69, "learning_rate": 9.4249547920434e-07, "loss": 0.0003, "step": 5739 }, { "epoch": 0.69, "learning_rate": 9.421338155515371e-07, "loss": 0.0003, "step": 5740 }, { "epoch": 0.69, "learning_rate": 9.417721518987343e-07, "loss": 0.0002, "step": 5741 }, { "epoch": 0.69, "learning_rate": 9.414104882459314e-07, "loss": 1.0935, "step": 5742 }, { "epoch": 0.69, "learning_rate": 9.410488245931285e-07, "loss": 0.0007, "step": 5743 }, { "epoch": 0.69, "learning_rate": 9.406871609403255e-07, "loss": 0.0002, "step": 5744 }, { "epoch": 0.69, "learning_rate": 9.403254972875227e-07, "loss": 0.0002, "step": 5745 }, { "epoch": 0.69, "learning_rate": 9.399638336347197e-07, "loss": 1.0633, "step": 5746 }, { "epoch": 0.69, "learning_rate": 9.396021699819168e-07, "loss": 0.0002, "step": 5747 }, { "epoch": 0.69, "learning_rate": 9.392405063291139e-07, "loss": 0.0003, "step": 5748 }, { "epoch": 0.69, "learning_rate": 9.388788426763111e-07, "loss": 0.0003, "step": 5749 }, { "epoch": 0.69, "learning_rate": 9.385171790235082e-07, "loss": 0.7409, "step": 5750 }, { "epoch": 0.69, "learning_rate": 9.381555153707053e-07, "loss": 0.5078, "step": 5751 }, { "epoch": 0.69, "learning_rate": 9.377938517179024e-07, "loss": 0.0001, "step": 5752 }, { "epoch": 0.69, "learning_rate": 9.374321880650996e-07, "loss": 0.0003, "step": 5753 }, { "epoch": 0.69, "learning_rate": 9.370705244122966e-07, "loss": 0.0006, "step": 5754 }, { "epoch": 0.69, "learning_rate": 9.367088607594937e-07, "loss": 0.0002, "step": 5755 }, { "epoch": 0.69, "learning_rate": 9.363471971066908e-07, "loss": 0.9116, "step": 5756 }, { "epoch": 0.69, "learning_rate": 9.35985533453888e-07, "loss": 0.0003, "step": 5757 }, { "epoch": 0.69, "learning_rate": 9.35623869801085e-07, "loss": 0.0003, "step": 5758 }, { "epoch": 0.69, "learning_rate": 9.352622061482821e-07, "loss": 0.0004, "step": 5759 }, { "epoch": 0.69, "learning_rate": 9.349005424954792e-07, "loss": 0.0004, "step": 5760 }, { "epoch": 0.69, "learning_rate": 9.345388788426763e-07, "loss": 0.1478, "step": 5761 }, { "epoch": 0.69, "learning_rate": 9.341772151898735e-07, "loss": 0.0003, "step": 5762 }, { "epoch": 0.69, "learning_rate": 9.338155515370705e-07, "loss": 0.5247, "step": 5763 }, { "epoch": 0.69, "learning_rate": 9.334538878842676e-07, "loss": 0.012, "step": 5764 }, { "epoch": 0.69, "learning_rate": 9.330922242314647e-07, "loss": 0.0004, "step": 5765 }, { "epoch": 0.69, "learning_rate": 9.327305605786619e-07, "loss": 0.001, "step": 5766 }, { "epoch": 0.69, "learning_rate": 9.32368896925859e-07, "loss": 0.0004, "step": 5767 }, { "epoch": 0.69, "learning_rate": 9.320072332730561e-07, "loss": 0.0004, "step": 5768 }, { "epoch": 0.69, "learning_rate": 9.316455696202532e-07, "loss": 0.0003, "step": 5769 }, { "epoch": 0.69, "learning_rate": 9.312839059674503e-07, "loss": 0.0003, "step": 5770 }, { "epoch": 0.69, "learning_rate": 9.309222423146474e-07, "loss": 0.0004, "step": 5771 }, { "epoch": 0.69, "learning_rate": 9.305605786618445e-07, "loss": 0.0031, "step": 5772 }, { "epoch": 0.69, "learning_rate": 9.301989150090415e-07, "loss": 0.0003, "step": 5773 }, { "epoch": 0.69, "learning_rate": 9.298372513562387e-07, "loss": 0.0007, "step": 5774 }, { "epoch": 0.69, "learning_rate": 9.294755877034358e-07, "loss": 0.9264, "step": 5775 }, { "epoch": 0.69, "learning_rate": 9.291139240506329e-07, "loss": 0.7435, "step": 5776 }, { "epoch": 0.69, "learning_rate": 9.2875226039783e-07, "loss": 0.0002, "step": 5777 }, { "epoch": 0.69, "learning_rate": 9.283905967450272e-07, "loss": 0.0003, "step": 5778 }, { "epoch": 0.69, "learning_rate": 9.280289330922243e-07, "loss": 0.0003, "step": 5779 }, { "epoch": 0.69, "learning_rate": 9.276672694394214e-07, "loss": 0.0005, "step": 5780 }, { "epoch": 0.69, "learning_rate": 9.273056057866185e-07, "loss": 0.0004, "step": 5781 }, { "epoch": 0.69, "learning_rate": 9.269439421338156e-07, "loss": 0.0004, "step": 5782 }, { "epoch": 0.69, "learning_rate": 9.265822784810126e-07, "loss": 0.0004, "step": 5783 }, { "epoch": 0.69, "learning_rate": 9.262206148282097e-07, "loss": 0.0003, "step": 5784 }, { "epoch": 0.69, "learning_rate": 9.258589511754068e-07, "loss": 0.0005, "step": 5785 }, { "epoch": 0.69, "learning_rate": 9.25497287522604e-07, "loss": 0.0005, "step": 5786 }, { "epoch": 0.69, "learning_rate": 9.251356238698011e-07, "loss": 0.0003, "step": 5787 }, { "epoch": 0.69, "learning_rate": 9.247739602169982e-07, "loss": 0.0003, "step": 5788 }, { "epoch": 0.69, "learning_rate": 9.244122965641953e-07, "loss": 0.0003, "step": 5789 }, { "epoch": 0.69, "learning_rate": 9.240506329113925e-07, "loss": 0.0007, "step": 5790 }, { "epoch": 0.69, "learning_rate": 9.236889692585896e-07, "loss": 0.0002, "step": 5791 }, { "epoch": 0.69, "learning_rate": 9.233273056057867e-07, "loss": 0.0003, "step": 5792 }, { "epoch": 0.69, "learning_rate": 9.229656419529838e-07, "loss": 0.0007, "step": 5793 }, { "epoch": 0.69, "learning_rate": 9.226039783001807e-07, "loss": 0.0003, "step": 5794 }, { "epoch": 0.69, "learning_rate": 9.222423146473779e-07, "loss": 0.0003, "step": 5795 }, { "epoch": 0.69, "learning_rate": 9.21880650994575e-07, "loss": 0.0003, "step": 5796 }, { "epoch": 0.69, "learning_rate": 9.215189873417721e-07, "loss": 0.0003, "step": 5797 }, { "epoch": 0.69, "learning_rate": 9.211573236889692e-07, "loss": 0.0005, "step": 5798 }, { "epoch": 0.69, "learning_rate": 9.207956600361664e-07, "loss": 0.0004, "step": 5799 }, { "epoch": 0.7, "learning_rate": 9.204339963833635e-07, "loss": 0.0003, "step": 5800 }, { "epoch": 0.7, "learning_rate": 9.200723327305606e-07, "loss": 0.0003, "step": 5801 }, { "epoch": 0.7, "learning_rate": 9.197106690777577e-07, "loss": 0.0004, "step": 5802 }, { "epoch": 0.7, "learning_rate": 9.193490054249549e-07, "loss": 0.0003, "step": 5803 }, { "epoch": 0.7, "learning_rate": 9.18987341772152e-07, "loss": 0.0002, "step": 5804 }, { "epoch": 0.7, "learning_rate": 9.186256781193491e-07, "loss": 0.0003, "step": 5805 }, { "epoch": 0.7, "learning_rate": 9.18264014466546e-07, "loss": 0.0002, "step": 5806 }, { "epoch": 0.7, "learning_rate": 9.179023508137432e-07, "loss": 0.0002, "step": 5807 }, { "epoch": 0.7, "learning_rate": 9.175406871609403e-07, "loss": 0.0004, "step": 5808 }, { "epoch": 0.7, "learning_rate": 9.171790235081374e-07, "loss": 0.0002, "step": 5809 }, { "epoch": 0.7, "learning_rate": 9.168173598553345e-07, "loss": 0.0006, "step": 5810 }, { "epoch": 0.7, "learning_rate": 9.164556962025317e-07, "loss": 0.0002, "step": 5811 }, { "epoch": 0.7, "learning_rate": 9.160940325497288e-07, "loss": 0.0004, "step": 5812 }, { "epoch": 0.7, "learning_rate": 9.157323688969259e-07, "loss": 0.0003, "step": 5813 }, { "epoch": 0.7, "learning_rate": 9.15370705244123e-07, "loss": 0.7344, "step": 5814 }, { "epoch": 0.7, "learning_rate": 9.150090415913202e-07, "loss": 0.0002, "step": 5815 }, { "epoch": 0.7, "learning_rate": 9.146473779385173e-07, "loss": 0.0003, "step": 5816 }, { "epoch": 0.7, "learning_rate": 9.142857142857144e-07, "loss": 0.0002, "step": 5817 }, { "epoch": 0.7, "learning_rate": 9.139240506329113e-07, "loss": 0.0003, "step": 5818 }, { "epoch": 0.7, "learning_rate": 9.135623869801084e-07, "loss": 0.0003, "step": 5819 }, { "epoch": 0.7, "learning_rate": 9.132007233273056e-07, "loss": 0.0002, "step": 5820 }, { "epoch": 0.7, "learning_rate": 9.128390596745027e-07, "loss": 0.0003, "step": 5821 }, { "epoch": 0.7, "learning_rate": 9.124773960216998e-07, "loss": 0.5592, "step": 5822 }, { "epoch": 0.7, "learning_rate": 9.12115732368897e-07, "loss": 0.0003, "step": 5823 }, { "epoch": 0.7, "learning_rate": 9.117540687160941e-07, "loss": 0.0003, "step": 5824 }, { "epoch": 0.7, "learning_rate": 9.113924050632912e-07, "loss": 0.0003, "step": 5825 }, { "epoch": 0.7, "learning_rate": 9.110307414104883e-07, "loss": 0.0003, "step": 5826 }, { "epoch": 0.7, "learning_rate": 9.106690777576855e-07, "loss": 0.5382, "step": 5827 }, { "epoch": 0.7, "learning_rate": 9.103074141048826e-07, "loss": 0.0002, "step": 5828 }, { "epoch": 0.7, "learning_rate": 9.099457504520797e-07, "loss": 0.0004, "step": 5829 }, { "epoch": 0.7, "learning_rate": 9.095840867992766e-07, "loss": 0.0003, "step": 5830 }, { "epoch": 0.7, "learning_rate": 9.092224231464737e-07, "loss": 0.0003, "step": 5831 }, { "epoch": 0.7, "learning_rate": 9.088607594936709e-07, "loss": 0.1023, "step": 5832 }, { "epoch": 0.7, "learning_rate": 9.08499095840868e-07, "loss": 0.0004, "step": 5833 }, { "epoch": 0.7, "learning_rate": 9.081374321880651e-07, "loss": 0.0002, "step": 5834 }, { "epoch": 0.7, "learning_rate": 9.077757685352622e-07, "loss": 0.0003, "step": 5835 }, { "epoch": 0.7, "learning_rate": 9.074141048824594e-07, "loss": 0.0004, "step": 5836 }, { "epoch": 0.7, "learning_rate": 9.070524412296565e-07, "loss": 0.0003, "step": 5837 }, { "epoch": 0.7, "learning_rate": 9.066907775768536e-07, "loss": 0.0003, "step": 5838 }, { "epoch": 0.7, "learning_rate": 9.063291139240507e-07, "loss": 0.0002, "step": 5839 }, { "epoch": 0.7, "learning_rate": 9.059674502712479e-07, "loss": 0.0003, "step": 5840 }, { "epoch": 0.7, "learning_rate": 9.056057866184448e-07, "loss": 0.0011, "step": 5841 }, { "epoch": 0.7, "learning_rate": 9.052441229656419e-07, "loss": 0.0004, "step": 5842 }, { "epoch": 0.7, "learning_rate": 9.04882459312839e-07, "loss": 0.8954, "step": 5843 }, { "epoch": 0.7, "learning_rate": 9.045207956600362e-07, "loss": 0.7344, "step": 5844 }, { "epoch": 0.7, "learning_rate": 9.041591320072333e-07, "loss": 0.0006, "step": 5845 }, { "epoch": 0.7, "learning_rate": 9.037974683544304e-07, "loss": 0.528, "step": 5846 }, { "epoch": 0.7, "learning_rate": 9.034358047016275e-07, "loss": 0.0004, "step": 5847 }, { "epoch": 0.7, "learning_rate": 9.030741410488247e-07, "loss": 0.0003, "step": 5848 }, { "epoch": 0.7, "learning_rate": 9.027124773960218e-07, "loss": 0.0002, "step": 5849 }, { "epoch": 0.7, "learning_rate": 9.023508137432189e-07, "loss": 0.0005, "step": 5850 }, { "epoch": 0.7, "learning_rate": 9.01989150090416e-07, "loss": 0.0003, "step": 5851 }, { "epoch": 0.7, "learning_rate": 9.016274864376132e-07, "loss": 0.5795, "step": 5852 }, { "epoch": 0.7, "learning_rate": 9.012658227848101e-07, "loss": 0.0776, "step": 5853 }, { "epoch": 0.7, "learning_rate": 9.009041591320072e-07, "loss": 0.0002, "step": 5854 }, { "epoch": 0.7, "learning_rate": 9.005424954792043e-07, "loss": 0.4284, "step": 5855 }, { "epoch": 0.7, "learning_rate": 9.001808318264014e-07, "loss": 0.0003, "step": 5856 }, { "epoch": 0.7, "learning_rate": 8.998191681735986e-07, "loss": 0.0003, "step": 5857 }, { "epoch": 0.7, "learning_rate": 8.994575045207957e-07, "loss": 0.0004, "step": 5858 }, { "epoch": 0.7, "learning_rate": 8.990958408679928e-07, "loss": 0.0005, "step": 5859 }, { "epoch": 0.7, "learning_rate": 8.9873417721519e-07, "loss": 0.0002, "step": 5860 }, { "epoch": 0.7, "learning_rate": 8.983725135623871e-07, "loss": 0.0003, "step": 5861 }, { "epoch": 0.7, "learning_rate": 8.980108499095842e-07, "loss": 0.0002, "step": 5862 }, { "epoch": 0.7, "learning_rate": 8.976491862567812e-07, "loss": 0.0003, "step": 5863 }, { "epoch": 0.7, "learning_rate": 8.972875226039783e-07, "loss": 0.0003, "step": 5864 }, { "epoch": 0.7, "learning_rate": 8.969258589511754e-07, "loss": 0.0003, "step": 5865 }, { "epoch": 0.7, "learning_rate": 8.965641952983725e-07, "loss": 0.0003, "step": 5866 }, { "epoch": 0.7, "learning_rate": 8.962025316455696e-07, "loss": 0.0003, "step": 5867 }, { "epoch": 0.7, "learning_rate": 8.958408679927667e-07, "loss": 0.0003, "step": 5868 }, { "epoch": 0.7, "learning_rate": 8.954792043399639e-07, "loss": 0.0006, "step": 5869 }, { "epoch": 0.7, "learning_rate": 8.95117540687161e-07, "loss": 0.0003, "step": 5870 }, { "epoch": 0.7, "learning_rate": 8.947558770343581e-07, "loss": 0.0009, "step": 5871 }, { "epoch": 0.7, "learning_rate": 8.943942133815552e-07, "loss": 0.0004, "step": 5872 }, { "epoch": 0.7, "learning_rate": 8.940325497287523e-07, "loss": 0.0003, "step": 5873 }, { "epoch": 0.7, "learning_rate": 8.936708860759494e-07, "loss": 0.0098, "step": 5874 }, { "epoch": 0.7, "learning_rate": 8.933092224231465e-07, "loss": 0.0003, "step": 5875 }, { "epoch": 0.7, "learning_rate": 8.929475587703436e-07, "loss": 0.0002, "step": 5876 }, { "epoch": 0.7, "learning_rate": 8.925858951175407e-07, "loss": 0.0003, "step": 5877 }, { "epoch": 0.7, "learning_rate": 8.922242314647378e-07, "loss": 0.0036, "step": 5878 }, { "epoch": 0.7, "learning_rate": 8.918625678119349e-07, "loss": 0.0004, "step": 5879 }, { "epoch": 0.7, "learning_rate": 8.91500904159132e-07, "loss": 0.0003, "step": 5880 }, { "epoch": 0.7, "learning_rate": 8.911392405063292e-07, "loss": 0.0003, "step": 5881 }, { "epoch": 0.7, "learning_rate": 8.907775768535263e-07, "loss": 0.0002, "step": 5882 }, { "epoch": 0.7, "learning_rate": 8.904159132007233e-07, "loss": 0.0003, "step": 5883 }, { "epoch": 0.71, "learning_rate": 8.900542495479204e-07, "loss": 0.0003, "step": 5884 }, { "epoch": 0.71, "learning_rate": 8.896925858951176e-07, "loss": 0.0004, "step": 5885 }, { "epoch": 0.71, "learning_rate": 8.893309222423147e-07, "loss": 0.0003, "step": 5886 }, { "epoch": 0.71, "learning_rate": 8.889692585895118e-07, "loss": 0.0002, "step": 5887 }, { "epoch": 0.71, "learning_rate": 8.886075949367089e-07, "loss": 0.0003, "step": 5888 }, { "epoch": 0.71, "learning_rate": 8.88245931283906e-07, "loss": 0.0002, "step": 5889 }, { "epoch": 0.71, "learning_rate": 8.878842676311031e-07, "loss": 0.0003, "step": 5890 }, { "epoch": 0.71, "learning_rate": 8.875226039783002e-07, "loss": 0.0003, "step": 5891 }, { "epoch": 0.71, "learning_rate": 8.871609403254973e-07, "loss": 0.2866, "step": 5892 }, { "epoch": 0.71, "learning_rate": 8.867992766726943e-07, "loss": 0.0003, "step": 5893 }, { "epoch": 0.71, "learning_rate": 8.864376130198915e-07, "loss": 0.0249, "step": 5894 }, { "epoch": 0.71, "learning_rate": 8.860759493670886e-07, "loss": 0.0003, "step": 5895 }, { "epoch": 0.71, "learning_rate": 8.857142857142857e-07, "loss": 0.0004, "step": 5896 }, { "epoch": 0.71, "learning_rate": 8.853526220614828e-07, "loss": 0.0084, "step": 5897 }, { "epoch": 0.71, "learning_rate": 8.8499095840868e-07, "loss": 0.0002, "step": 5898 }, { "epoch": 0.71, "learning_rate": 8.846292947558771e-07, "loss": 0.0003, "step": 5899 }, { "epoch": 0.71, "learning_rate": 8.842676311030742e-07, "loss": 0.0004, "step": 5900 }, { "epoch": 0.71, "learning_rate": 8.839059674502712e-07, "loss": 0.0002, "step": 5901 }, { "epoch": 0.71, "learning_rate": 8.835443037974683e-07, "loss": 0.0003, "step": 5902 }, { "epoch": 0.71, "learning_rate": 8.831826401446654e-07, "loss": 0.0007, "step": 5903 }, { "epoch": 0.71, "learning_rate": 8.828209764918625e-07, "loss": 0.0002, "step": 5904 }, { "epoch": 0.71, "learning_rate": 8.824593128390596e-07, "loss": 0.0003, "step": 5905 }, { "epoch": 0.71, "learning_rate": 8.820976491862568e-07, "loss": 0.0003, "step": 5906 }, { "epoch": 0.71, "learning_rate": 8.817359855334539e-07, "loss": 0.0002, "step": 5907 }, { "epoch": 0.71, "learning_rate": 8.81374321880651e-07, "loss": 0.0003, "step": 5908 }, { "epoch": 0.71, "learning_rate": 8.810126582278481e-07, "loss": 0.0006, "step": 5909 }, { "epoch": 0.71, "learning_rate": 8.806509945750453e-07, "loss": 0.0002, "step": 5910 }, { "epoch": 0.71, "learning_rate": 8.802893309222424e-07, "loss": 0.0002, "step": 5911 }, { "epoch": 0.71, "learning_rate": 8.799276672694395e-07, "loss": 0.0002, "step": 5912 }, { "epoch": 0.71, "learning_rate": 8.795660036166364e-07, "loss": 0.0002, "step": 5913 }, { "epoch": 0.71, "learning_rate": 8.792043399638336e-07, "loss": 0.0003, "step": 5914 }, { "epoch": 0.71, "learning_rate": 8.788426763110307e-07, "loss": 0.0002, "step": 5915 }, { "epoch": 0.71, "learning_rate": 8.784810126582278e-07, "loss": 0.0003, "step": 5916 }, { "epoch": 0.71, "learning_rate": 8.781193490054249e-07, "loss": 0.001, "step": 5917 }, { "epoch": 0.71, "learning_rate": 8.77757685352622e-07, "loss": 0.0002, "step": 5918 }, { "epoch": 0.71, "learning_rate": 8.773960216998192e-07, "loss": 0.0032, "step": 5919 }, { "epoch": 0.71, "learning_rate": 8.770343580470163e-07, "loss": 1.3877, "step": 5920 }, { "epoch": 0.71, "learning_rate": 8.766726943942134e-07, "loss": 0.0003, "step": 5921 }, { "epoch": 0.71, "learning_rate": 8.763110307414106e-07, "loss": 0.0002, "step": 5922 }, { "epoch": 0.71, "learning_rate": 8.759493670886077e-07, "loss": 0.0002, "step": 5923 }, { "epoch": 0.71, "learning_rate": 8.755877034358048e-07, "loss": 0.9983, "step": 5924 }, { "epoch": 0.71, "learning_rate": 8.752260397830017e-07, "loss": 0.0004, "step": 5925 }, { "epoch": 0.71, "learning_rate": 8.748643761301988e-07, "loss": 0.0003, "step": 5926 }, { "epoch": 0.71, "learning_rate": 8.74502712477396e-07, "loss": 0.0002, "step": 5927 }, { "epoch": 0.71, "learning_rate": 8.741410488245931e-07, "loss": 0.0002, "step": 5928 }, { "epoch": 0.71, "learning_rate": 8.737793851717902e-07, "loss": 0.0002, "step": 5929 }, { "epoch": 0.71, "learning_rate": 8.734177215189873e-07, "loss": 0.227, "step": 5930 }, { "epoch": 0.71, "learning_rate": 8.730560578661845e-07, "loss": 0.0004, "step": 5931 }, { "epoch": 0.71, "learning_rate": 8.726943942133816e-07, "loss": 0.0004, "step": 5932 }, { "epoch": 0.71, "learning_rate": 8.723327305605787e-07, "loss": 0.0002, "step": 5933 }, { "epoch": 0.71, "learning_rate": 8.719710669077758e-07, "loss": 0.0003, "step": 5934 }, { "epoch": 0.71, "learning_rate": 8.71609403254973e-07, "loss": 0.0002, "step": 5935 }, { "epoch": 0.71, "learning_rate": 8.712477396021701e-07, "loss": 0.0002, "step": 5936 }, { "epoch": 0.71, "learning_rate": 8.70886075949367e-07, "loss": 0.0002, "step": 5937 }, { "epoch": 0.71, "learning_rate": 8.705244122965641e-07, "loss": 0.0004, "step": 5938 }, { "epoch": 0.71, "learning_rate": 8.701627486437613e-07, "loss": 0.0042, "step": 5939 }, { "epoch": 0.71, "learning_rate": 8.698010849909584e-07, "loss": 0.0003, "step": 5940 }, { "epoch": 0.71, "learning_rate": 8.694394213381555e-07, "loss": 0.0003, "step": 5941 }, { "epoch": 0.71, "learning_rate": 8.690777576853526e-07, "loss": 0.0003, "step": 5942 }, { "epoch": 0.71, "learning_rate": 8.687160940325498e-07, "loss": 0.4476, "step": 5943 }, { "epoch": 0.71, "learning_rate": 8.683544303797469e-07, "loss": 0.0003, "step": 5944 }, { "epoch": 0.71, "learning_rate": 8.67992766726944e-07, "loss": 0.0002, "step": 5945 }, { "epoch": 0.71, "learning_rate": 8.676311030741411e-07, "loss": 0.0002, "step": 5946 }, { "epoch": 0.71, "learning_rate": 8.672694394213383e-07, "loss": 0.0003, "step": 5947 }, { "epoch": 0.71, "learning_rate": 8.669077757685354e-07, "loss": 0.0002, "step": 5948 }, { "epoch": 0.71, "learning_rate": 8.665461121157323e-07, "loss": 0.0002, "step": 5949 }, { "epoch": 0.71, "learning_rate": 8.661844484629294e-07, "loss": 0.1351, "step": 5950 }, { "epoch": 0.71, "learning_rate": 8.658227848101266e-07, "loss": 0.1144, "step": 5951 }, { "epoch": 0.71, "learning_rate": 8.654611211573237e-07, "loss": 0.0004, "step": 5952 }, { "epoch": 0.71, "learning_rate": 8.650994575045208e-07, "loss": 0.001, "step": 5953 }, { "epoch": 0.71, "learning_rate": 8.647377938517179e-07, "loss": 0.0013, "step": 5954 }, { "epoch": 0.71, "learning_rate": 8.64376130198915e-07, "loss": 0.0002, "step": 5955 }, { "epoch": 0.71, "learning_rate": 8.640144665461122e-07, "loss": 0.0002, "step": 5956 }, { "epoch": 0.71, "learning_rate": 8.636528028933093e-07, "loss": 0.0002, "step": 5957 }, { "epoch": 0.71, "learning_rate": 8.632911392405064e-07, "loss": 0.0003, "step": 5958 }, { "epoch": 0.71, "learning_rate": 8.629294755877036e-07, "loss": 0.0003, "step": 5959 }, { "epoch": 0.71, "learning_rate": 8.625678119349007e-07, "loss": 0.0002, "step": 5960 }, { "epoch": 0.71, "learning_rate": 8.622061482820976e-07, "loss": 0.0003, "step": 5961 }, { "epoch": 0.71, "learning_rate": 8.618444846292947e-07, "loss": 0.0003, "step": 5962 }, { "epoch": 0.71, "learning_rate": 8.614828209764918e-07, "loss": 0.0002, "step": 5963 }, { "epoch": 0.71, "learning_rate": 8.61121157323689e-07, "loss": 0.0002, "step": 5964 }, { "epoch": 0.71, "learning_rate": 8.607594936708861e-07, "loss": 0.0003, "step": 5965 }, { "epoch": 0.71, "learning_rate": 8.603978300180832e-07, "loss": 0.0036, "step": 5966 }, { "epoch": 0.72, "learning_rate": 8.600361663652803e-07, "loss": 0.0002, "step": 5967 }, { "epoch": 0.72, "learning_rate": 8.596745027124775e-07, "loss": 0.0002, "step": 5968 }, { "epoch": 0.72, "learning_rate": 8.593128390596746e-07, "loss": 0.0002, "step": 5969 }, { "epoch": 0.72, "learning_rate": 8.589511754068717e-07, "loss": 0.0003, "step": 5970 }, { "epoch": 0.72, "learning_rate": 8.585895117540688e-07, "loss": 0.0003, "step": 5971 }, { "epoch": 0.72, "learning_rate": 8.58227848101266e-07, "loss": 0.0002, "step": 5972 }, { "epoch": 0.72, "learning_rate": 8.578661844484629e-07, "loss": 0.0002, "step": 5973 }, { "epoch": 0.72, "learning_rate": 8.5750452079566e-07, "loss": 0.0003, "step": 5974 }, { "epoch": 0.72, "learning_rate": 8.571428571428571e-07, "loss": 0.9661, "step": 5975 }, { "epoch": 0.72, "learning_rate": 8.567811934900543e-07, "loss": 0.0002, "step": 5976 }, { "epoch": 0.72, "learning_rate": 8.564195298372514e-07, "loss": 0.0002, "step": 5977 }, { "epoch": 0.72, "learning_rate": 8.560578661844485e-07, "loss": 0.0003, "step": 5978 }, { "epoch": 0.72, "learning_rate": 8.556962025316456e-07, "loss": 0.0003, "step": 5979 }, { "epoch": 0.72, "learning_rate": 8.553345388788428e-07, "loss": 0.0003, "step": 5980 }, { "epoch": 0.72, "learning_rate": 8.549728752260399e-07, "loss": 0.0002, "step": 5981 }, { "epoch": 0.72, "learning_rate": 8.54611211573237e-07, "loss": 0.0002, "step": 5982 }, { "epoch": 0.72, "learning_rate": 8.54249547920434e-07, "loss": 0.0003, "step": 5983 }, { "epoch": 0.72, "learning_rate": 8.538878842676312e-07, "loss": 0.0002, "step": 5984 }, { "epoch": 0.72, "learning_rate": 8.535262206148282e-07, "loss": 0.0004, "step": 5985 }, { "epoch": 0.72, "learning_rate": 8.531645569620253e-07, "loss": 0.0002, "step": 5986 }, { "epoch": 0.72, "learning_rate": 8.528028933092224e-07, "loss": 0.0002, "step": 5987 }, { "epoch": 0.72, "learning_rate": 8.524412296564196e-07, "loss": 0.0002, "step": 5988 }, { "epoch": 0.72, "learning_rate": 8.520795660036167e-07, "loss": 0.0002, "step": 5989 }, { "epoch": 0.72, "learning_rate": 8.517179023508138e-07, "loss": 0.0002, "step": 5990 }, { "epoch": 0.72, "learning_rate": 8.513562386980109e-07, "loss": 0.5143, "step": 5991 }, { "epoch": 0.72, "learning_rate": 8.50994575045208e-07, "loss": 0.0003, "step": 5992 }, { "epoch": 0.72, "learning_rate": 8.506329113924051e-07, "loss": 0.0002, "step": 5993 }, { "epoch": 0.72, "learning_rate": 8.502712477396022e-07, "loss": 0.978, "step": 5994 }, { "epoch": 0.72, "learning_rate": 8.499095840867993e-07, "loss": 0.0003, "step": 5995 }, { "epoch": 0.72, "learning_rate": 8.495479204339964e-07, "loss": 0.0002, "step": 5996 }, { "epoch": 0.72, "learning_rate": 8.491862567811935e-07, "loss": 0.7089, "step": 5997 }, { "epoch": 0.72, "learning_rate": 8.488245931283906e-07, "loss": 0.0003, "step": 5998 }, { "epoch": 0.72, "learning_rate": 8.484629294755877e-07, "loss": 0.0002, "step": 5999 }, { "epoch": 0.72, "learning_rate": 8.481012658227848e-07, "loss": 0.0006, "step": 6000 }, { "epoch": 0.72, "learning_rate": 8.47739602169982e-07, "loss": 0.0002, "step": 6001 }, { "epoch": 0.72, "learning_rate": 8.473779385171791e-07, "loss": 0.1844, "step": 6002 }, { "epoch": 0.72, "learning_rate": 8.470162748643761e-07, "loss": 0.0002, "step": 6003 }, { "epoch": 0.72, "learning_rate": 8.466546112115732e-07, "loss": 0.0002, "step": 6004 }, { "epoch": 0.72, "learning_rate": 8.462929475587704e-07, "loss": 0.0002, "step": 6005 }, { "epoch": 0.72, "learning_rate": 8.459312839059675e-07, "loss": 0.0002, "step": 6006 }, { "epoch": 0.72, "learning_rate": 8.455696202531646e-07, "loss": 0.9354, "step": 6007 }, { "epoch": 0.72, "learning_rate": 8.452079566003617e-07, "loss": 0.0004, "step": 6008 }, { "epoch": 0.72, "learning_rate": 8.448462929475588e-07, "loss": 0.0002, "step": 6009 }, { "epoch": 0.72, "learning_rate": 8.444846292947559e-07, "loss": 0.0002, "step": 6010 }, { "epoch": 0.72, "learning_rate": 8.44122965641953e-07, "loss": 0.0002, "step": 6011 }, { "epoch": 0.72, "learning_rate": 8.4376130198915e-07, "loss": 0.0004, "step": 6012 }, { "epoch": 0.72, "learning_rate": 8.433996383363472e-07, "loss": 0.0002, "step": 6013 }, { "epoch": 0.72, "learning_rate": 8.430379746835443e-07, "loss": 0.0003, "step": 6014 }, { "epoch": 0.72, "learning_rate": 8.426763110307414e-07, "loss": 0.0002, "step": 6015 }, { "epoch": 0.72, "learning_rate": 8.423146473779385e-07, "loss": 0.0002, "step": 6016 }, { "epoch": 0.72, "learning_rate": 8.419529837251357e-07, "loss": 0.0008, "step": 6017 }, { "epoch": 0.72, "learning_rate": 8.415913200723328e-07, "loss": 0.0004, "step": 6018 }, { "epoch": 0.72, "learning_rate": 8.412296564195299e-07, "loss": 0.0002, "step": 6019 }, { "epoch": 0.72, "learning_rate": 8.40867992766727e-07, "loss": 0.0006, "step": 6020 }, { "epoch": 0.72, "learning_rate": 8.40506329113924e-07, "loss": 0.0004, "step": 6021 }, { "epoch": 0.72, "learning_rate": 8.401446654611211e-07, "loss": 0.0002, "step": 6022 }, { "epoch": 0.72, "learning_rate": 8.397830018083182e-07, "loss": 0.0028, "step": 6023 }, { "epoch": 0.72, "learning_rate": 8.394213381555153e-07, "loss": 0.0003, "step": 6024 }, { "epoch": 0.72, "learning_rate": 8.390596745027124e-07, "loss": 0.5265, "step": 6025 }, { "epoch": 0.72, "learning_rate": 8.386980108499096e-07, "loss": 0.6152, "step": 6026 }, { "epoch": 0.72, "learning_rate": 8.383363471971067e-07, "loss": 0.0005, "step": 6027 }, { "epoch": 0.72, "learning_rate": 8.379746835443038e-07, "loss": 0.0008, "step": 6028 }, { "epoch": 0.72, "learning_rate": 8.376130198915009e-07, "loss": 0.0002, "step": 6029 }, { "epoch": 0.72, "learning_rate": 8.372513562386981e-07, "loss": 0.0003, "step": 6030 }, { "epoch": 0.72, "learning_rate": 8.368896925858952e-07, "loss": 0.4355, "step": 6031 }, { "epoch": 0.72, "learning_rate": 8.365280289330923e-07, "loss": 0.0004, "step": 6032 }, { "epoch": 0.72, "learning_rate": 8.361663652802892e-07, "loss": 0.0002, "step": 6033 }, { "epoch": 0.72, "learning_rate": 8.358047016274864e-07, "loss": 0.0003, "step": 6034 }, { "epoch": 0.72, "learning_rate": 8.354430379746835e-07, "loss": 0.0002, "step": 6035 }, { "epoch": 0.72, "learning_rate": 8.350813743218806e-07, "loss": 0.0002, "step": 6036 }, { "epoch": 0.72, "learning_rate": 8.347197106690777e-07, "loss": 0.0002, "step": 6037 }, { "epoch": 0.72, "learning_rate": 8.343580470162749e-07, "loss": 0.0002, "step": 6038 }, { "epoch": 0.72, "learning_rate": 8.33996383363472e-07, "loss": 0.0002, "step": 6039 }, { "epoch": 0.72, "learning_rate": 8.336347197106691e-07, "loss": 0.5901, "step": 6040 }, { "epoch": 0.72, "learning_rate": 8.332730560578662e-07, "loss": 0.0002, "step": 6041 }, { "epoch": 0.72, "learning_rate": 8.329113924050634e-07, "loss": 0.5122, "step": 6042 }, { "epoch": 0.72, "learning_rate": 8.325497287522605e-07, "loss": 0.0002, "step": 6043 }, { "epoch": 0.72, "learning_rate": 8.321880650994576e-07, "loss": 0.0002, "step": 6044 }, { "epoch": 0.72, "learning_rate": 8.318264014466545e-07, "loss": 0.0002, "step": 6045 }, { "epoch": 0.72, "learning_rate": 8.314647377938517e-07, "loss": 0.0002, "step": 6046 }, { "epoch": 0.72, "learning_rate": 8.311030741410488e-07, "loss": 0.0002, "step": 6047 }, { "epoch": 0.72, "learning_rate": 8.307414104882459e-07, "loss": 0.0002, "step": 6048 }, { "epoch": 0.72, "learning_rate": 8.30379746835443e-07, "loss": 0.0005, "step": 6049 }, { "epoch": 0.72, "learning_rate": 8.300180831826402e-07, "loss": 0.5832, "step": 6050 }, { "epoch": 0.73, "learning_rate": 8.296564195298373e-07, "loss": 0.0002, "step": 6051 }, { "epoch": 0.73, "learning_rate": 8.292947558770344e-07, "loss": 0.0002, "step": 6052 }, { "epoch": 0.73, "learning_rate": 8.289330922242315e-07, "loss": 0.0003, "step": 6053 }, { "epoch": 0.73, "learning_rate": 8.285714285714287e-07, "loss": 0.0002, "step": 6054 }, { "epoch": 0.73, "learning_rate": 8.282097649186258e-07, "loss": 0.0002, "step": 6055 }, { "epoch": 0.73, "learning_rate": 8.278481012658229e-07, "loss": 0.4766, "step": 6056 }, { "epoch": 0.73, "learning_rate": 8.274864376130198e-07, "loss": 0.0004, "step": 6057 }, { "epoch": 0.73, "learning_rate": 8.271247739602169e-07, "loss": 0.0002, "step": 6058 }, { "epoch": 0.73, "learning_rate": 8.267631103074141e-07, "loss": 0.0002, "step": 6059 }, { "epoch": 0.73, "learning_rate": 8.264014466546112e-07, "loss": 0.0003, "step": 6060 }, { "epoch": 0.73, "learning_rate": 8.260397830018083e-07, "loss": 0.0003, "step": 6061 }, { "epoch": 0.73, "learning_rate": 8.256781193490054e-07, "loss": 0.0002, "step": 6062 }, { "epoch": 0.73, "learning_rate": 8.253164556962026e-07, "loss": 0.0003, "step": 6063 }, { "epoch": 0.73, "learning_rate": 8.249547920433997e-07, "loss": 0.0002, "step": 6064 }, { "epoch": 0.73, "learning_rate": 8.245931283905968e-07, "loss": 0.0002, "step": 6065 }, { "epoch": 0.73, "learning_rate": 8.242314647377939e-07, "loss": 0.0004, "step": 6066 }, { "epoch": 0.73, "learning_rate": 8.238698010849911e-07, "loss": 0.0002, "step": 6067 }, { "epoch": 0.73, "learning_rate": 8.235081374321882e-07, "loss": 0.0004, "step": 6068 }, { "epoch": 0.73, "learning_rate": 8.231464737793851e-07, "loss": 0.9005, "step": 6069 }, { "epoch": 0.73, "learning_rate": 8.227848101265822e-07, "loss": 0.0001, "step": 6070 }, { "epoch": 0.73, "learning_rate": 8.224231464737794e-07, "loss": 0.0002, "step": 6071 }, { "epoch": 0.73, "learning_rate": 8.220614828209765e-07, "loss": 0.0019, "step": 6072 }, { "epoch": 0.73, "learning_rate": 8.216998191681736e-07, "loss": 0.0002, "step": 6073 }, { "epoch": 0.73, "learning_rate": 8.213381555153707e-07, "loss": 0.0002, "step": 6074 }, { "epoch": 0.73, "learning_rate": 8.209764918625679e-07, "loss": 0.0002, "step": 6075 }, { "epoch": 0.73, "learning_rate": 8.20614828209765e-07, "loss": 0.0002, "step": 6076 }, { "epoch": 0.73, "learning_rate": 8.202531645569621e-07, "loss": 0.0002, "step": 6077 }, { "epoch": 0.73, "learning_rate": 8.198915009041592e-07, "loss": 0.0005, "step": 6078 }, { "epoch": 0.73, "learning_rate": 8.195298372513564e-07, "loss": 0.0003, "step": 6079 }, { "epoch": 0.73, "learning_rate": 8.191681735985535e-07, "loss": 0.0003, "step": 6080 }, { "epoch": 0.73, "learning_rate": 8.188065099457504e-07, "loss": 0.0007, "step": 6081 }, { "epoch": 0.73, "learning_rate": 8.184448462929475e-07, "loss": 0.0003, "step": 6082 }, { "epoch": 0.73, "learning_rate": 8.180831826401447e-07, "loss": 0.0003, "step": 6083 }, { "epoch": 0.73, "learning_rate": 8.177215189873418e-07, "loss": 0.0003, "step": 6084 }, { "epoch": 0.73, "learning_rate": 8.173598553345389e-07, "loss": 0.0002, "step": 6085 }, { "epoch": 0.73, "learning_rate": 8.16998191681736e-07, "loss": 0.0005, "step": 6086 }, { "epoch": 0.73, "learning_rate": 8.166365280289332e-07, "loss": 0.0003, "step": 6087 }, { "epoch": 0.73, "learning_rate": 8.162748643761303e-07, "loss": 0.0003, "step": 6088 }, { "epoch": 0.73, "learning_rate": 8.159132007233274e-07, "loss": 0.0002, "step": 6089 }, { "epoch": 0.73, "learning_rate": 8.155515370705245e-07, "loss": 0.0003, "step": 6090 }, { "epoch": 0.73, "learning_rate": 8.151898734177217e-07, "loss": 0.0002, "step": 6091 }, { "epoch": 0.73, "learning_rate": 8.148282097649186e-07, "loss": 0.0075, "step": 6092 }, { "epoch": 0.73, "learning_rate": 8.144665461121157e-07, "loss": 0.531, "step": 6093 }, { "epoch": 0.73, "learning_rate": 8.141048824593128e-07, "loss": 0.001, "step": 6094 }, { "epoch": 0.73, "learning_rate": 8.137432188065099e-07, "loss": 0.0007, "step": 6095 }, { "epoch": 0.73, "learning_rate": 8.133815551537071e-07, "loss": 0.0002, "step": 6096 }, { "epoch": 0.73, "learning_rate": 8.130198915009042e-07, "loss": 0.0002, "step": 6097 }, { "epoch": 0.73, "learning_rate": 8.126582278481013e-07, "loss": 0.0005, "step": 6098 }, { "epoch": 0.73, "learning_rate": 8.122965641952984e-07, "loss": 0.0001, "step": 6099 }, { "epoch": 0.73, "learning_rate": 8.119349005424956e-07, "loss": 0.0002, "step": 6100 }, { "epoch": 0.73, "learning_rate": 8.115732368896927e-07, "loss": 0.0002, "step": 6101 }, { "epoch": 0.73, "learning_rate": 8.112115732368898e-07, "loss": 0.0077, "step": 6102 }, { "epoch": 0.73, "learning_rate": 8.108499095840868e-07, "loss": 0.0003, "step": 6103 }, { "epoch": 0.73, "learning_rate": 8.104882459312839e-07, "loss": 0.0001, "step": 6104 }, { "epoch": 0.73, "learning_rate": 8.10126582278481e-07, "loss": 0.0002, "step": 6105 }, { "epoch": 0.73, "learning_rate": 8.097649186256781e-07, "loss": 0.0007, "step": 6106 }, { "epoch": 0.73, "learning_rate": 8.094032549728752e-07, "loss": 0.0002, "step": 6107 }, { "epoch": 0.73, "learning_rate": 8.090415913200724e-07, "loss": 0.0004, "step": 6108 }, { "epoch": 0.73, "learning_rate": 8.086799276672695e-07, "loss": 0.0002, "step": 6109 }, { "epoch": 0.73, "learning_rate": 8.083182640144666e-07, "loss": 0.0007, "step": 6110 }, { "epoch": 0.73, "learning_rate": 8.079566003616637e-07, "loss": 0.0002, "step": 6111 }, { "epoch": 0.73, "learning_rate": 8.075949367088608e-07, "loss": 0.0003, "step": 6112 }, { "epoch": 0.73, "learning_rate": 8.072332730560579e-07, "loss": 0.6724, "step": 6113 }, { "epoch": 0.73, "learning_rate": 8.06871609403255e-07, "loss": 0.0002, "step": 6114 }, { "epoch": 0.73, "learning_rate": 8.065099457504521e-07, "loss": 0.0002, "step": 6115 }, { "epoch": 0.73, "learning_rate": 8.061482820976492e-07, "loss": 0.0002, "step": 6116 }, { "epoch": 0.73, "learning_rate": 8.057866184448463e-07, "loss": 0.0002, "step": 6117 }, { "epoch": 0.73, "learning_rate": 8.054249547920434e-07, "loss": 0.0003, "step": 6118 }, { "epoch": 0.73, "learning_rate": 8.050632911392405e-07, "loss": 0.0002, "step": 6119 }, { "epoch": 0.73, "learning_rate": 8.047016274864377e-07, "loss": 0.0003, "step": 6120 }, { "epoch": 0.73, "learning_rate": 8.043399638336348e-07, "loss": 0.0002, "step": 6121 }, { "epoch": 0.73, "learning_rate": 8.039783001808318e-07, "loss": 0.0002, "step": 6122 }, { "epoch": 0.73, "learning_rate": 8.036166365280289e-07, "loss": 0.0003, "step": 6123 }, { "epoch": 0.73, "learning_rate": 8.03254972875226e-07, "loss": 0.0002, "step": 6124 }, { "epoch": 0.73, "learning_rate": 8.028933092224232e-07, "loss": 0.0026, "step": 6125 }, { "epoch": 0.73, "learning_rate": 8.025316455696203e-07, "loss": 0.0003, "step": 6126 }, { "epoch": 0.73, "learning_rate": 8.021699819168174e-07, "loss": 0.0002, "step": 6127 }, { "epoch": 0.73, "learning_rate": 8.018083182640144e-07, "loss": 0.6892, "step": 6128 }, { "epoch": 0.73, "learning_rate": 8.014466546112116e-07, "loss": 0.5881, "step": 6129 }, { "epoch": 0.73, "learning_rate": 8.010849909584087e-07, "loss": 0.0004, "step": 6130 }, { "epoch": 0.73, "learning_rate": 8.007233273056058e-07, "loss": 0.0003, "step": 6131 }, { "epoch": 0.73, "learning_rate": 8.003616636528028e-07, "loss": 0.0016, "step": 6132 }, { "epoch": 0.73, "learning_rate": 8e-07, "loss": 0.0016, "step": 6133 }, { "epoch": 0.74, "learning_rate": 7.996383363471971e-07, "loss": 0.0003, "step": 6134 }, { "epoch": 0.74, "learning_rate": 7.992766726943942e-07, "loss": 0.0002, "step": 6135 }, { "epoch": 0.74, "learning_rate": 7.989150090415913e-07, "loss": 0.0002, "step": 6136 }, { "epoch": 0.74, "learning_rate": 7.985533453887885e-07, "loss": 0.0002, "step": 6137 }, { "epoch": 0.74, "learning_rate": 7.981916817359856e-07, "loss": 0.0002, "step": 6138 }, { "epoch": 0.74, "learning_rate": 7.978300180831827e-07, "loss": 0.0002, "step": 6139 }, { "epoch": 0.74, "learning_rate": 7.974683544303797e-07, "loss": 0.0002, "step": 6140 }, { "epoch": 0.74, "learning_rate": 7.971066907775769e-07, "loss": 0.0002, "step": 6141 }, { "epoch": 0.74, "learning_rate": 7.967450271247739e-07, "loss": 0.0003, "step": 6142 }, { "epoch": 0.74, "learning_rate": 7.96383363471971e-07, "loss": 0.0004, "step": 6143 }, { "epoch": 0.74, "learning_rate": 7.960216998191681e-07, "loss": 0.0002, "step": 6144 }, { "epoch": 0.74, "learning_rate": 7.956600361663653e-07, "loss": 1.083, "step": 6145 }, { "epoch": 0.74, "learning_rate": 7.952983725135624e-07, "loss": 0.0002, "step": 6146 }, { "epoch": 0.74, "learning_rate": 7.949367088607595e-07, "loss": 0.0003, "step": 6147 }, { "epoch": 0.74, "learning_rate": 7.945750452079566e-07, "loss": 0.6371, "step": 6148 }, { "epoch": 0.74, "learning_rate": 7.942133815551538e-07, "loss": 0.0016, "step": 6149 }, { "epoch": 0.74, "learning_rate": 7.938517179023509e-07, "loss": 0.0002, "step": 6150 }, { "epoch": 0.74, "learning_rate": 7.93490054249548e-07, "loss": 0.0002, "step": 6151 }, { "epoch": 0.74, "learning_rate": 7.931283905967449e-07, "loss": 0.9711, "step": 6152 }, { "epoch": 0.74, "learning_rate": 7.92766726943942e-07, "loss": 0.0012, "step": 6153 }, { "epoch": 0.74, "learning_rate": 7.924050632911392e-07, "loss": 0.0002, "step": 6154 }, { "epoch": 0.74, "learning_rate": 7.920433996383363e-07, "loss": 0.0003, "step": 6155 }, { "epoch": 0.74, "learning_rate": 7.916817359855334e-07, "loss": 0.0002, "step": 6156 }, { "epoch": 0.74, "learning_rate": 7.913200723327305e-07, "loss": 0.0003, "step": 6157 }, { "epoch": 0.74, "learning_rate": 7.909584086799277e-07, "loss": 0.0003, "step": 6158 }, { "epoch": 0.74, "learning_rate": 7.905967450271248e-07, "loss": 0.0001, "step": 6159 }, { "epoch": 0.74, "learning_rate": 7.902350813743219e-07, "loss": 0.0002, "step": 6160 }, { "epoch": 0.74, "learning_rate": 7.89873417721519e-07, "loss": 0.8434, "step": 6161 }, { "epoch": 0.74, "learning_rate": 7.895117540687162e-07, "loss": 0.8286, "step": 6162 }, { "epoch": 0.74, "learning_rate": 7.891500904159133e-07, "loss": 0.0002, "step": 6163 }, { "epoch": 0.74, "learning_rate": 7.887884267631102e-07, "loss": 0.0003, "step": 6164 }, { "epoch": 0.74, "learning_rate": 7.884267631103073e-07, "loss": 0.0002, "step": 6165 }, { "epoch": 0.74, "learning_rate": 7.880650994575045e-07, "loss": 0.0003, "step": 6166 }, { "epoch": 0.74, "learning_rate": 7.877034358047016e-07, "loss": 0.0002, "step": 6167 }, { "epoch": 0.74, "learning_rate": 7.873417721518987e-07, "loss": 1.01, "step": 6168 }, { "epoch": 0.74, "learning_rate": 7.869801084990958e-07, "loss": 0.0003, "step": 6169 }, { "epoch": 0.74, "learning_rate": 7.86618444846293e-07, "loss": 0.0002, "step": 6170 }, { "epoch": 0.74, "learning_rate": 7.862567811934901e-07, "loss": 0.0003, "step": 6171 }, { "epoch": 0.74, "learning_rate": 7.858951175406872e-07, "loss": 0.0002, "step": 6172 }, { "epoch": 0.74, "learning_rate": 7.855334538878843e-07, "loss": 0.0001, "step": 6173 }, { "epoch": 0.74, "learning_rate": 7.851717902350815e-07, "loss": 0.0002, "step": 6174 }, { "epoch": 0.74, "learning_rate": 7.848101265822786e-07, "loss": 0.891, "step": 6175 }, { "epoch": 0.74, "learning_rate": 7.844484629294755e-07, "loss": 0.0002, "step": 6176 }, { "epoch": 0.74, "learning_rate": 7.840867992766726e-07, "loss": 0.5877, "step": 6177 }, { "epoch": 0.74, "learning_rate": 7.837251356238698e-07, "loss": 0.0004, "step": 6178 }, { "epoch": 0.74, "learning_rate": 7.833634719710669e-07, "loss": 0.0004, "step": 6179 }, { "epoch": 0.74, "learning_rate": 7.83001808318264e-07, "loss": 0.0002, "step": 6180 }, { "epoch": 0.74, "learning_rate": 7.826401446654611e-07, "loss": 0.0003, "step": 6181 }, { "epoch": 0.74, "learning_rate": 7.822784810126583e-07, "loss": 0.0002, "step": 6182 }, { "epoch": 0.74, "learning_rate": 7.819168173598554e-07, "loss": 0.0002, "step": 6183 }, { "epoch": 0.74, "learning_rate": 7.815551537070525e-07, "loss": 0.0003, "step": 6184 }, { "epoch": 0.74, "learning_rate": 7.811934900542496e-07, "loss": 0.5456, "step": 6185 }, { "epoch": 0.74, "learning_rate": 7.808318264014468e-07, "loss": 0.0002, "step": 6186 }, { "epoch": 0.74, "learning_rate": 7.804701627486439e-07, "loss": 0.0002, "step": 6187 }, { "epoch": 0.74, "learning_rate": 7.801084990958408e-07, "loss": 0.0003, "step": 6188 }, { "epoch": 0.74, "learning_rate": 7.797468354430379e-07, "loss": 0.0005, "step": 6189 }, { "epoch": 0.74, "learning_rate": 7.79385171790235e-07, "loss": 0.1004, "step": 6190 }, { "epoch": 0.74, "learning_rate": 7.790235081374322e-07, "loss": 0.8455, "step": 6191 }, { "epoch": 0.74, "learning_rate": 7.786618444846293e-07, "loss": 0.0002, "step": 6192 }, { "epoch": 0.74, "learning_rate": 7.783001808318264e-07, "loss": 0.0003, "step": 6193 }, { "epoch": 0.74, "learning_rate": 7.779385171790235e-07, "loss": 0.0003, "step": 6194 }, { "epoch": 0.74, "learning_rate": 7.775768535262207e-07, "loss": 0.0002, "step": 6195 }, { "epoch": 0.74, "learning_rate": 7.772151898734178e-07, "loss": 0.0002, "step": 6196 }, { "epoch": 0.74, "learning_rate": 7.768535262206149e-07, "loss": 0.0006, "step": 6197 }, { "epoch": 0.74, "learning_rate": 7.76491862567812e-07, "loss": 0.0002, "step": 6198 }, { "epoch": 0.74, "learning_rate": 7.761301989150092e-07, "loss": 0.0002, "step": 6199 }, { "epoch": 0.74, "learning_rate": 7.757685352622061e-07, "loss": 0.0002, "step": 6200 }, { "epoch": 0.74, "learning_rate": 7.754068716094032e-07, "loss": 0.0003, "step": 6201 }, { "epoch": 0.74, "learning_rate": 7.750452079566003e-07, "loss": 0.0002, "step": 6202 }, { "epoch": 0.74, "learning_rate": 7.746835443037975e-07, "loss": 0.0001, "step": 6203 }, { "epoch": 0.74, "learning_rate": 7.743218806509946e-07, "loss": 0.355, "step": 6204 }, { "epoch": 0.74, "learning_rate": 7.739602169981917e-07, "loss": 0.0012, "step": 6205 }, { "epoch": 0.74, "learning_rate": 7.735985533453888e-07, "loss": 0.0004, "step": 6206 }, { "epoch": 0.74, "learning_rate": 7.73236889692586e-07, "loss": 0.0002, "step": 6207 }, { "epoch": 0.74, "learning_rate": 7.728752260397831e-07, "loss": 0.0003, "step": 6208 }, { "epoch": 0.74, "learning_rate": 7.725135623869802e-07, "loss": 0.0002, "step": 6209 }, { "epoch": 0.74, "learning_rate": 7.721518987341773e-07, "loss": 0.0002, "step": 6210 }, { "epoch": 0.74, "learning_rate": 7.717902350813745e-07, "loss": 0.0002, "step": 6211 }, { "epoch": 0.74, "learning_rate": 7.714285714285714e-07, "loss": 0.0007, "step": 6212 }, { "epoch": 0.74, "learning_rate": 7.710669077757685e-07, "loss": 0.103, "step": 6213 }, { "epoch": 0.74, "learning_rate": 7.707052441229656e-07, "loss": 0.0003, "step": 6214 }, { "epoch": 0.74, "learning_rate": 7.703435804701628e-07, "loss": 0.0003, "step": 6215 }, { "epoch": 0.74, "learning_rate": 7.699819168173599e-07, "loss": 0.001, "step": 6216 }, { "epoch": 0.74, "learning_rate": 7.69620253164557e-07, "loss": 0.0004, "step": 6217 }, { "epoch": 0.75, "learning_rate": 7.692585895117541e-07, "loss": 0.0005, "step": 6218 }, { "epoch": 0.75, "learning_rate": 7.688969258589513e-07, "loss": 0.0003, "step": 6219 }, { "epoch": 0.75, "learning_rate": 7.685352622061484e-07, "loss": 0.0003, "step": 6220 }, { "epoch": 0.75, "learning_rate": 7.681735985533455e-07, "loss": 0.0002, "step": 6221 }, { "epoch": 0.75, "learning_rate": 7.678119349005425e-07, "loss": 0.0003, "step": 6222 }, { "epoch": 0.75, "learning_rate": 7.674502712477397e-07, "loss": 0.0002, "step": 6223 }, { "epoch": 0.75, "learning_rate": 7.670886075949367e-07, "loss": 0.8921, "step": 6224 }, { "epoch": 0.75, "learning_rate": 7.667269439421338e-07, "loss": 0.0001, "step": 6225 }, { "epoch": 0.75, "learning_rate": 7.663652802893309e-07, "loss": 0.0002, "step": 6226 }, { "epoch": 0.75, "learning_rate": 7.66003616636528e-07, "loss": 0.3234, "step": 6227 }, { "epoch": 0.75, "learning_rate": 7.656419529837252e-07, "loss": 0.0002, "step": 6228 }, { "epoch": 0.75, "learning_rate": 7.652802893309223e-07, "loss": 0.3453, "step": 6229 }, { "epoch": 0.75, "learning_rate": 7.649186256781194e-07, "loss": 0.0003, "step": 6230 }, { "epoch": 0.75, "learning_rate": 7.645569620253165e-07, "loss": 0.0002, "step": 6231 }, { "epoch": 0.75, "learning_rate": 7.641952983725136e-07, "loss": 0.2491, "step": 6232 }, { "epoch": 0.75, "learning_rate": 7.638336347197107e-07, "loss": 0.0002, "step": 6233 }, { "epoch": 0.75, "learning_rate": 7.634719710669078e-07, "loss": 0.0002, "step": 6234 }, { "epoch": 0.75, "learning_rate": 7.631103074141049e-07, "loss": 0.0003, "step": 6235 }, { "epoch": 0.75, "learning_rate": 7.62748643761302e-07, "loss": 0.0002, "step": 6236 }, { "epoch": 0.75, "learning_rate": 7.623869801084991e-07, "loss": 0.1202, "step": 6237 }, { "epoch": 0.75, "learning_rate": 7.620253164556962e-07, "loss": 0.0002, "step": 6238 }, { "epoch": 0.75, "learning_rate": 7.616636528028933e-07, "loss": 0.0003, "step": 6239 }, { "epoch": 0.75, "learning_rate": 7.613019891500905e-07, "loss": 0.0003, "step": 6240 }, { "epoch": 0.75, "learning_rate": 7.609403254972876e-07, "loss": 0.0002, "step": 6241 }, { "epoch": 0.75, "learning_rate": 7.605786618444846e-07, "loss": 0.0004, "step": 6242 }, { "epoch": 0.75, "learning_rate": 7.602169981916817e-07, "loss": 0.0043, "step": 6243 }, { "epoch": 0.75, "learning_rate": 7.598553345388789e-07, "loss": 0.0002, "step": 6244 }, { "epoch": 0.75, "learning_rate": 7.59493670886076e-07, "loss": 0.0002, "step": 6245 }, { "epoch": 0.75, "learning_rate": 7.591320072332731e-07, "loss": 0.044, "step": 6246 }, { "epoch": 0.75, "learning_rate": 7.587703435804702e-07, "loss": 0.0002, "step": 6247 }, { "epoch": 0.75, "learning_rate": 7.584086799276673e-07, "loss": 0.0002, "step": 6248 }, { "epoch": 0.75, "learning_rate": 7.580470162748644e-07, "loss": 0.0002, "step": 6249 }, { "epoch": 0.75, "learning_rate": 7.576853526220615e-07, "loss": 0.0002, "step": 6250 }, { "epoch": 0.75, "learning_rate": 7.573236889692586e-07, "loss": 0.0002, "step": 6251 }, { "epoch": 0.75, "learning_rate": 7.569620253164556e-07, "loss": 0.0003, "step": 6252 }, { "epoch": 0.75, "learning_rate": 7.566003616636528e-07, "loss": 0.6366, "step": 6253 }, { "epoch": 0.75, "learning_rate": 7.562386980108499e-07, "loss": 0.4815, "step": 6254 }, { "epoch": 0.75, "learning_rate": 7.55877034358047e-07, "loss": 0.0002, "step": 6255 }, { "epoch": 0.75, "learning_rate": 7.555153707052441e-07, "loss": 0.0002, "step": 6256 }, { "epoch": 0.75, "learning_rate": 7.551537070524413e-07, "loss": 0.0002, "step": 6257 }, { "epoch": 0.75, "learning_rate": 7.547920433996384e-07, "loss": 0.0002, "step": 6258 }, { "epoch": 0.75, "learning_rate": 7.544303797468355e-07, "loss": 0.0002, "step": 6259 }, { "epoch": 0.75, "learning_rate": 7.540687160940325e-07, "loss": 0.0002, "step": 6260 }, { "epoch": 0.75, "learning_rate": 7.537070524412296e-07, "loss": 0.0003, "step": 6261 }, { "epoch": 0.75, "learning_rate": 7.533453887884267e-07, "loss": 0.0003, "step": 6262 }, { "epoch": 0.75, "learning_rate": 7.529837251356238e-07, "loss": 0.0002, "step": 6263 }, { "epoch": 0.75, "learning_rate": 7.526220614828209e-07, "loss": 0.0001, "step": 6264 }, { "epoch": 0.75, "learning_rate": 7.522603978300181e-07, "loss": 0.001, "step": 6265 }, { "epoch": 0.75, "learning_rate": 7.518987341772152e-07, "loss": 0.0002, "step": 6266 }, { "epoch": 0.75, "learning_rate": 7.515370705244123e-07, "loss": 0.0003, "step": 6267 }, { "epoch": 0.75, "learning_rate": 7.511754068716094e-07, "loss": 0.0002, "step": 6268 }, { "epoch": 0.75, "learning_rate": 7.508137432188066e-07, "loss": 0.0001, "step": 6269 }, { "epoch": 0.75, "learning_rate": 7.504520795660037e-07, "loss": 0.0002, "step": 6270 }, { "epoch": 0.75, "learning_rate": 7.500904159132008e-07, "loss": 0.0002, "step": 6271 }, { "epoch": 0.75, "learning_rate": 7.497287522603978e-07, "loss": 0.0002, "step": 6272 }, { "epoch": 0.75, "learning_rate": 7.49367088607595e-07, "loss": 0.0001, "step": 6273 }, { "epoch": 0.75, "learning_rate": 7.490054249547921e-07, "loss": 0.0003, "step": 6274 }, { "epoch": 0.75, "learning_rate": 7.486437613019891e-07, "loss": 0.4071, "step": 6275 }, { "epoch": 0.75, "learning_rate": 7.482820976491862e-07, "loss": 0.0003, "step": 6276 }, { "epoch": 0.75, "learning_rate": 7.479204339963834e-07, "loss": 0.0006, "step": 6277 }, { "epoch": 0.75, "learning_rate": 7.475587703435805e-07, "loss": 0.0002, "step": 6278 }, { "epoch": 0.75, "learning_rate": 7.471971066907776e-07, "loss": 0.001, "step": 6279 }, { "epoch": 0.75, "learning_rate": 7.468354430379747e-07, "loss": 0.5446, "step": 6280 }, { "epoch": 0.75, "learning_rate": 7.464737793851718e-07, "loss": 0.0002, "step": 6281 }, { "epoch": 0.75, "learning_rate": 7.461121157323689e-07, "loss": 0.0002, "step": 6282 }, { "epoch": 0.75, "learning_rate": 7.45750452079566e-07, "loss": 0.0003, "step": 6283 }, { "epoch": 0.75, "learning_rate": 7.453887884267631e-07, "loss": 0.0002, "step": 6284 }, { "epoch": 0.75, "learning_rate": 7.450271247739603e-07, "loss": 0.0001, "step": 6285 }, { "epoch": 0.75, "learning_rate": 7.446654611211574e-07, "loss": 0.0009, "step": 6286 }, { "epoch": 0.75, "learning_rate": 7.443037974683544e-07, "loss": 0.0002, "step": 6287 }, { "epoch": 0.75, "learning_rate": 7.439421338155515e-07, "loss": 1.0087, "step": 6288 }, { "epoch": 0.75, "learning_rate": 7.435804701627486e-07, "loss": 0.0002, "step": 6289 }, { "epoch": 0.75, "learning_rate": 7.432188065099458e-07, "loss": 1.0858, "step": 6290 }, { "epoch": 0.75, "learning_rate": 7.428571428571429e-07, "loss": 0.0002, "step": 6291 }, { "epoch": 0.75, "learning_rate": 7.4249547920434e-07, "loss": 0.0004, "step": 6292 }, { "epoch": 0.75, "learning_rate": 7.42133815551537e-07, "loss": 0.3945, "step": 6293 }, { "epoch": 0.75, "learning_rate": 7.417721518987342e-07, "loss": 0.0003, "step": 6294 }, { "epoch": 0.75, "learning_rate": 7.414104882459313e-07, "loss": 0.0004, "step": 6295 }, { "epoch": 0.75, "learning_rate": 7.410488245931284e-07, "loss": 0.0002, "step": 6296 }, { "epoch": 0.75, "learning_rate": 7.406871609403255e-07, "loss": 0.0003, "step": 6297 }, { "epoch": 0.75, "learning_rate": 7.403254972875227e-07, "loss": 0.0002, "step": 6298 }, { "epoch": 0.75, "learning_rate": 7.399638336347197e-07, "loss": 0.0002, "step": 6299 }, { "epoch": 0.75, "learning_rate": 7.396021699819168e-07, "loss": 0.0002, "step": 6300 }, { "epoch": 0.76, "learning_rate": 7.392405063291139e-07, "loss": 0.0002, "step": 6301 }, { "epoch": 0.76, "learning_rate": 7.388788426763111e-07, "loss": 0.9084, "step": 6302 }, { "epoch": 0.76, "learning_rate": 7.385171790235082e-07, "loss": 0.0002, "step": 6303 }, { "epoch": 0.76, "learning_rate": 7.381555153707053e-07, "loss": 0.0003, "step": 6304 }, { "epoch": 0.76, "learning_rate": 7.377938517179023e-07, "loss": 0.0002, "step": 6305 }, { "epoch": 0.76, "learning_rate": 7.374321880650995e-07, "loss": 0.0003, "step": 6306 }, { "epoch": 0.76, "learning_rate": 7.370705244122966e-07, "loss": 0.0004, "step": 6307 }, { "epoch": 0.76, "learning_rate": 7.367088607594937e-07, "loss": 0.0002, "step": 6308 }, { "epoch": 0.76, "learning_rate": 7.363471971066908e-07, "loss": 0.0005, "step": 6309 }, { "epoch": 0.76, "learning_rate": 7.35985533453888e-07, "loss": 0.0003, "step": 6310 }, { "epoch": 0.76, "learning_rate": 7.35623869801085e-07, "loss": 0.0002, "step": 6311 }, { "epoch": 0.76, "learning_rate": 7.352622061482821e-07, "loss": 0.0007, "step": 6312 }, { "epoch": 0.76, "learning_rate": 7.349005424954792e-07, "loss": 0.0003, "step": 6313 }, { "epoch": 0.76, "learning_rate": 7.345388788426764e-07, "loss": 0.0055, "step": 6314 }, { "epoch": 0.76, "learning_rate": 7.341772151898735e-07, "loss": 0.9488, "step": 6315 }, { "epoch": 0.76, "learning_rate": 7.338155515370706e-07, "loss": 0.0002, "step": 6316 }, { "epoch": 0.76, "learning_rate": 7.334538878842676e-07, "loss": 0.0003, "step": 6317 }, { "epoch": 0.76, "learning_rate": 7.330922242314648e-07, "loss": 0.4746, "step": 6318 }, { "epoch": 0.76, "learning_rate": 7.327305605786619e-07, "loss": 0.0003, "step": 6319 }, { "epoch": 0.76, "learning_rate": 7.32368896925859e-07, "loss": 0.0017, "step": 6320 }, { "epoch": 0.76, "learning_rate": 7.320072332730561e-07, "loss": 0.0002, "step": 6321 }, { "epoch": 0.76, "learning_rate": 7.316455696202533e-07, "loss": 0.0241, "step": 6322 }, { "epoch": 0.76, "learning_rate": 7.312839059674503e-07, "loss": 0.0002, "step": 6323 }, { "epoch": 0.76, "learning_rate": 7.309222423146474e-07, "loss": 0.4825, "step": 6324 }, { "epoch": 0.76, "learning_rate": 7.305605786618445e-07, "loss": 0.0002, "step": 6325 }, { "epoch": 0.76, "learning_rate": 7.301989150090416e-07, "loss": 0.0002, "step": 6326 }, { "epoch": 0.76, "learning_rate": 7.298372513562388e-07, "loss": 0.0002, "step": 6327 }, { "epoch": 0.76, "learning_rate": 7.294755877034359e-07, "loss": 0.0003, "step": 6328 }, { "epoch": 0.76, "learning_rate": 7.291139240506329e-07, "loss": 0.0002, "step": 6329 }, { "epoch": 0.76, "learning_rate": 7.2875226039783e-07, "loss": 0.0002, "step": 6330 }, { "epoch": 0.76, "learning_rate": 7.283905967450272e-07, "loss": 2.0022, "step": 6331 }, { "epoch": 0.76, "learning_rate": 7.280289330922243e-07, "loss": 0.0002, "step": 6332 }, { "epoch": 0.76, "learning_rate": 7.276672694394214e-07, "loss": 0.0002, "step": 6333 }, { "epoch": 0.76, "learning_rate": 7.273056057866185e-07, "loss": 0.0115, "step": 6334 }, { "epoch": 0.76, "learning_rate": 7.269439421338156e-07, "loss": 0.0002, "step": 6335 }, { "epoch": 0.76, "learning_rate": 7.265822784810127e-07, "loss": 0.0008, "step": 6336 }, { "epoch": 0.76, "learning_rate": 7.262206148282098e-07, "loss": 0.0002, "step": 6337 }, { "epoch": 0.76, "learning_rate": 7.258589511754069e-07, "loss": 0.0002, "step": 6338 }, { "epoch": 0.76, "learning_rate": 7.254972875226041e-07, "loss": 0.0003, "step": 6339 }, { "epoch": 0.76, "learning_rate": 7.251356238698011e-07, "loss": 0.0004, "step": 6340 }, { "epoch": 0.76, "learning_rate": 7.247739602169982e-07, "loss": 0.0003, "step": 6341 }, { "epoch": 0.76, "learning_rate": 7.244122965641953e-07, "loss": 0.0003, "step": 6342 }, { "epoch": 0.76, "learning_rate": 7.240506329113925e-07, "loss": 0.0003, "step": 6343 }, { "epoch": 0.76, "learning_rate": 7.236889692585896e-07, "loss": 0.0003, "step": 6344 }, { "epoch": 0.76, "learning_rate": 7.233273056057867e-07, "loss": 0.0003, "step": 6345 }, { "epoch": 0.76, "learning_rate": 7.229656419529837e-07, "loss": 0.0002, "step": 6346 }, { "epoch": 0.76, "learning_rate": 7.226039783001809e-07, "loss": 0.0001, "step": 6347 }, { "epoch": 0.76, "learning_rate": 7.22242314647378e-07, "loss": 0.0002, "step": 6348 }, { "epoch": 0.76, "learning_rate": 7.218806509945751e-07, "loss": 0.0003, "step": 6349 }, { "epoch": 0.76, "learning_rate": 7.215189873417722e-07, "loss": 0.3253, "step": 6350 }, { "epoch": 0.76, "learning_rate": 7.211573236889694e-07, "loss": 0.0002, "step": 6351 }, { "epoch": 0.76, "learning_rate": 7.207956600361664e-07, "loss": 0.0003, "step": 6352 }, { "epoch": 0.76, "learning_rate": 7.204339963833635e-07, "loss": 0.0008, "step": 6353 }, { "epoch": 0.76, "learning_rate": 7.200723327305606e-07, "loss": 0.0003, "step": 6354 }, { "epoch": 0.76, "learning_rate": 7.197106690777578e-07, "loss": 0.0003, "step": 6355 }, { "epoch": 0.76, "learning_rate": 7.193490054249549e-07, "loss": 0.0002, "step": 6356 }, { "epoch": 0.76, "learning_rate": 7.189873417721519e-07, "loss": 0.0002, "step": 6357 }, { "epoch": 0.76, "learning_rate": 7.18625678119349e-07, "loss": 0.0003, "step": 6358 }, { "epoch": 0.76, "learning_rate": 7.182640144665461e-07, "loss": 0.0268, "step": 6359 }, { "epoch": 0.76, "learning_rate": 7.179023508137433e-07, "loss": 0.0003, "step": 6360 }, { "epoch": 0.76, "learning_rate": 7.175406871609403e-07, "loss": 0.0003, "step": 6361 }, { "epoch": 0.76, "learning_rate": 7.171790235081374e-07, "loss": 0.0003, "step": 6362 }, { "epoch": 0.76, "learning_rate": 7.168173598553345e-07, "loss": 0.0003, "step": 6363 }, { "epoch": 0.76, "learning_rate": 7.164556962025317e-07, "loss": 0.003, "step": 6364 }, { "epoch": 0.76, "learning_rate": 7.160940325497288e-07, "loss": 0.0002, "step": 6365 }, { "epoch": 0.76, "learning_rate": 7.157323688969258e-07, "loss": 0.0004, "step": 6366 }, { "epoch": 0.76, "learning_rate": 7.153707052441229e-07, "loss": 0.0026, "step": 6367 }, { "epoch": 0.76, "learning_rate": 7.150090415913201e-07, "loss": 0.0003, "step": 6368 }, { "epoch": 0.76, "learning_rate": 7.146473779385172e-07, "loss": 0.0003, "step": 6369 }, { "epoch": 0.76, "learning_rate": 7.142857142857143e-07, "loss": 0.0003, "step": 6370 }, { "epoch": 0.76, "learning_rate": 7.139240506329113e-07, "loss": 1.0015, "step": 6371 }, { "epoch": 0.76, "learning_rate": 7.135623869801085e-07, "loss": 0.0008, "step": 6372 }, { "epoch": 0.76, "learning_rate": 7.132007233273056e-07, "loss": 0.0411, "step": 6373 }, { "epoch": 0.76, "learning_rate": 7.128390596745027e-07, "loss": 0.0003, "step": 6374 }, { "epoch": 0.76, "learning_rate": 7.124773960216998e-07, "loss": 0.0002, "step": 6375 }, { "epoch": 0.76, "learning_rate": 7.121157323688969e-07, "loss": 0.003, "step": 6376 }, { "epoch": 0.76, "learning_rate": 7.11754068716094e-07, "loss": 0.0003, "step": 6377 }, { "epoch": 0.76, "learning_rate": 7.113924050632911e-07, "loss": 0.0003, "step": 6378 }, { "epoch": 0.76, "learning_rate": 7.110307414104882e-07, "loss": 0.0003, "step": 6379 }, { "epoch": 0.76, "learning_rate": 7.106690777576854e-07, "loss": 0.0003, "step": 6380 }, { "epoch": 0.76, "learning_rate": 7.103074141048825e-07, "loss": 0.0002, "step": 6381 }, { "epoch": 0.76, "learning_rate": 7.099457504520795e-07, "loss": 0.0002, "step": 6382 }, { "epoch": 0.76, "learning_rate": 7.095840867992766e-07, "loss": 0.0003, "step": 6383 }, { "epoch": 0.77, "learning_rate": 7.092224231464738e-07, "loss": 0.0002, "step": 6384 }, { "epoch": 0.77, "learning_rate": 7.088607594936709e-07, "loss": 0.0002, "step": 6385 }, { "epoch": 0.77, "learning_rate": 7.08499095840868e-07, "loss": 0.446, "step": 6386 }, { "epoch": 0.77, "learning_rate": 7.081374321880651e-07, "loss": 0.0002, "step": 6387 }, { "epoch": 0.77, "learning_rate": 7.077757685352621e-07, "loss": 0.0003, "step": 6388 }, { "epoch": 0.77, "learning_rate": 7.074141048824593e-07, "loss": 0.0003, "step": 6389 }, { "epoch": 0.77, "learning_rate": 7.070524412296564e-07, "loss": 0.0002, "step": 6390 }, { "epoch": 0.77, "learning_rate": 7.066907775768535e-07, "loss": 0.0003, "step": 6391 }, { "epoch": 0.77, "learning_rate": 7.063291139240506e-07, "loss": 0.0003, "step": 6392 }, { "epoch": 0.77, "learning_rate": 7.059674502712478e-07, "loss": 0.8115, "step": 6393 }, { "epoch": 0.77, "learning_rate": 7.056057866184448e-07, "loss": 0.0002, "step": 6394 }, { "epoch": 0.77, "learning_rate": 7.052441229656419e-07, "loss": 1.1067, "step": 6395 }, { "epoch": 0.77, "learning_rate": 7.04882459312839e-07, "loss": 0.0002, "step": 6396 }, { "epoch": 0.77, "learning_rate": 7.045207956600362e-07, "loss": 0.0002, "step": 6397 }, { "epoch": 0.77, "learning_rate": 7.041591320072333e-07, "loss": 0.0002, "step": 6398 }, { "epoch": 0.77, "learning_rate": 7.037974683544304e-07, "loss": 0.0141, "step": 6399 }, { "epoch": 0.77, "learning_rate": 7.034358047016274e-07, "loss": 0.0003, "step": 6400 }, { "epoch": 0.77, "learning_rate": 7.030741410488246e-07, "loss": 0.0002, "step": 6401 }, { "epoch": 0.77, "learning_rate": 7.027124773960217e-07, "loss": 0.0004, "step": 6402 }, { "epoch": 0.77, "learning_rate": 7.023508137432188e-07, "loss": 0.5655, "step": 6403 }, { "epoch": 0.77, "learning_rate": 7.019891500904159e-07, "loss": 0.0005, "step": 6404 }, { "epoch": 0.77, "learning_rate": 7.016274864376131e-07, "loss": 0.0001, "step": 6405 }, { "epoch": 0.77, "learning_rate": 7.012658227848101e-07, "loss": 0.0008, "step": 6406 }, { "epoch": 0.77, "learning_rate": 7.009041591320072e-07, "loss": 0.0588, "step": 6407 }, { "epoch": 0.77, "learning_rate": 7.005424954792043e-07, "loss": 0.0002, "step": 6408 }, { "epoch": 0.77, "learning_rate": 7.001808318264015e-07, "loss": 0.0002, "step": 6409 }, { "epoch": 0.77, "learning_rate": 6.998191681735986e-07, "loss": 0.0002, "step": 6410 }, { "epoch": 0.77, "learning_rate": 6.994575045207957e-07, "loss": 0.0004, "step": 6411 }, { "epoch": 0.77, "learning_rate": 6.990958408679927e-07, "loss": 1.0307, "step": 6412 }, { "epoch": 0.77, "learning_rate": 6.987341772151899e-07, "loss": 0.0002, "step": 6413 }, { "epoch": 0.77, "learning_rate": 6.98372513562387e-07, "loss": 0.0276, "step": 6414 }, { "epoch": 0.77, "learning_rate": 6.980108499095841e-07, "loss": 0.0002, "step": 6415 }, { "epoch": 0.77, "learning_rate": 6.976491862567812e-07, "loss": 0.0003, "step": 6416 }, { "epoch": 0.77, "learning_rate": 6.972875226039784e-07, "loss": 0.0002, "step": 6417 }, { "epoch": 0.77, "learning_rate": 6.969258589511754e-07, "loss": 0.0002, "step": 6418 }, { "epoch": 0.77, "learning_rate": 6.965641952983725e-07, "loss": 0.0004, "step": 6419 }, { "epoch": 0.77, "learning_rate": 6.962025316455696e-07, "loss": 0.0002, "step": 6420 }, { "epoch": 0.77, "learning_rate": 6.958408679927668e-07, "loss": 0.0003, "step": 6421 }, { "epoch": 0.77, "learning_rate": 6.954792043399639e-07, "loss": 0.0076, "step": 6422 }, { "epoch": 0.77, "learning_rate": 6.95117540687161e-07, "loss": 0.0003, "step": 6423 }, { "epoch": 0.77, "learning_rate": 6.94755877034358e-07, "loss": 0.1193, "step": 6424 }, { "epoch": 0.77, "learning_rate": 6.943942133815551e-07, "loss": 0.0005, "step": 6425 }, { "epoch": 0.77, "learning_rate": 6.940325497287523e-07, "loss": 0.0003, "step": 6426 }, { "epoch": 0.77, "learning_rate": 6.936708860759494e-07, "loss": 0.1006, "step": 6427 }, { "epoch": 0.77, "learning_rate": 6.933092224231465e-07, "loss": 0.0003, "step": 6428 }, { "epoch": 0.77, "learning_rate": 6.929475587703436e-07, "loss": 0.0004, "step": 6429 }, { "epoch": 0.77, "learning_rate": 6.925858951175407e-07, "loss": 0.2283, "step": 6430 }, { "epoch": 0.77, "learning_rate": 6.922242314647378e-07, "loss": 0.0003, "step": 6431 }, { "epoch": 0.77, "learning_rate": 6.918625678119349e-07, "loss": 0.0003, "step": 6432 }, { "epoch": 0.77, "learning_rate": 6.91500904159132e-07, "loss": 0.0002, "step": 6433 }, { "epoch": 0.77, "learning_rate": 6.911392405063292e-07, "loss": 0.0002, "step": 6434 }, { "epoch": 0.77, "learning_rate": 6.907775768535263e-07, "loss": 0.0002, "step": 6435 }, { "epoch": 0.77, "learning_rate": 6.904159132007233e-07, "loss": 0.0003, "step": 6436 }, { "epoch": 0.77, "learning_rate": 6.900542495479204e-07, "loss": 0.0002, "step": 6437 }, { "epoch": 0.77, "learning_rate": 6.896925858951176e-07, "loss": 0.0005, "step": 6438 }, { "epoch": 0.77, "learning_rate": 6.893309222423147e-07, "loss": 0.0002, "step": 6439 }, { "epoch": 0.77, "learning_rate": 6.889692585895118e-07, "loss": 0.0003, "step": 6440 }, { "epoch": 0.77, "learning_rate": 6.886075949367089e-07, "loss": 0.0009, "step": 6441 }, { "epoch": 0.77, "learning_rate": 6.88245931283906e-07, "loss": 0.0002, "step": 6442 }, { "epoch": 0.77, "learning_rate": 6.878842676311031e-07, "loss": 0.0002, "step": 6443 }, { "epoch": 0.77, "learning_rate": 6.875226039783002e-07, "loss": 0.0002, "step": 6444 }, { "epoch": 0.77, "learning_rate": 6.871609403254973e-07, "loss": 0.1821, "step": 6445 }, { "epoch": 0.77, "learning_rate": 6.867992766726945e-07, "loss": 0.0003, "step": 6446 }, { "epoch": 0.77, "learning_rate": 6.864376130198916e-07, "loss": 0.0004, "step": 6447 }, { "epoch": 0.77, "learning_rate": 6.860759493670886e-07, "loss": 0.0004, "step": 6448 }, { "epoch": 0.77, "learning_rate": 6.857142857142857e-07, "loss": 0.0002, "step": 6449 }, { "epoch": 0.77, "learning_rate": 6.853526220614829e-07, "loss": 0.0003, "step": 6450 }, { "epoch": 0.77, "learning_rate": 6.8499095840868e-07, "loss": 0.0002, "step": 6451 }, { "epoch": 0.77, "learning_rate": 6.846292947558771e-07, "loss": 0.0003, "step": 6452 }, { "epoch": 0.77, "learning_rate": 6.842676311030742e-07, "loss": 0.0002, "step": 6453 }, { "epoch": 0.77, "learning_rate": 6.839059674502712e-07, "loss": 0.0003, "step": 6454 }, { "epoch": 0.77, "learning_rate": 6.835443037974684e-07, "loss": 0.0003, "step": 6455 }, { "epoch": 0.77, "learning_rate": 6.831826401446655e-07, "loss": 0.0004, "step": 6456 }, { "epoch": 0.77, "learning_rate": 6.828209764918626e-07, "loss": 0.0003, "step": 6457 }, { "epoch": 0.77, "learning_rate": 6.824593128390598e-07, "loss": 0.0003, "step": 6458 }, { "epoch": 0.77, "learning_rate": 6.820976491862569e-07, "loss": 0.0001, "step": 6459 }, { "epoch": 0.77, "learning_rate": 6.817359855334539e-07, "loss": 0.0003, "step": 6460 }, { "epoch": 0.77, "learning_rate": 6.81374321880651e-07, "loss": 0.001, "step": 6461 }, { "epoch": 0.77, "learning_rate": 6.810126582278481e-07, "loss": 0.0002, "step": 6462 }, { "epoch": 0.77, "learning_rate": 6.806509945750453e-07, "loss": 0.0003, "step": 6463 }, { "epoch": 0.77, "learning_rate": 6.802893309222424e-07, "loss": 0.0002, "step": 6464 }, { "epoch": 0.77, "learning_rate": 6.799276672694395e-07, "loss": 0.0003, "step": 6465 }, { "epoch": 0.77, "learning_rate": 6.795660036166365e-07, "loss": 0.0003, "step": 6466 }, { "epoch": 0.77, "learning_rate": 6.792043399638337e-07, "loss": 0.0002, "step": 6467 }, { "epoch": 0.78, "learning_rate": 6.788426763110308e-07, "loss": 0.0002, "step": 6468 }, { "epoch": 0.78, "learning_rate": 6.784810126582279e-07, "loss": 0.0003, "step": 6469 }, { "epoch": 0.78, "learning_rate": 6.78119349005425e-07, "loss": 0.0003, "step": 6470 }, { "epoch": 0.78, "learning_rate": 6.777576853526221e-07, "loss": 0.0002, "step": 6471 }, { "epoch": 0.78, "learning_rate": 6.773960216998192e-07, "loss": 0.0002, "step": 6472 }, { "epoch": 0.78, "learning_rate": 6.770343580470163e-07, "loss": 0.0003, "step": 6473 }, { "epoch": 0.78, "learning_rate": 6.766726943942134e-07, "loss": 0.0003, "step": 6474 }, { "epoch": 0.78, "learning_rate": 6.763110307414106e-07, "loss": 0.0069, "step": 6475 }, { "epoch": 0.78, "learning_rate": 6.759493670886076e-07, "loss": 0.0002, "step": 6476 }, { "epoch": 0.78, "learning_rate": 6.755877034358047e-07, "loss": 0.0003, "step": 6477 }, { "epoch": 0.78, "learning_rate": 6.752260397830018e-07, "loss": 0.0002, "step": 6478 }, { "epoch": 0.78, "learning_rate": 6.74864376130199e-07, "loss": 0.0004, "step": 6479 }, { "epoch": 0.78, "learning_rate": 6.745027124773961e-07, "loss": 0.0002, "step": 6480 }, { "epoch": 0.78, "learning_rate": 6.741410488245931e-07, "loss": 0.6835, "step": 6481 }, { "epoch": 0.78, "learning_rate": 6.737793851717902e-07, "loss": 0.0004, "step": 6482 }, { "epoch": 0.78, "learning_rate": 6.734177215189874e-07, "loss": 0.3002, "step": 6483 }, { "epoch": 0.78, "learning_rate": 6.730560578661845e-07, "loss": 0.0174, "step": 6484 }, { "epoch": 0.78, "learning_rate": 6.726943942133816e-07, "loss": 0.2462, "step": 6485 }, { "epoch": 0.78, "learning_rate": 6.723327305605786e-07, "loss": 0.0046, "step": 6486 }, { "epoch": 0.78, "learning_rate": 6.719710669077757e-07, "loss": 0.0006, "step": 6487 }, { "epoch": 0.78, "learning_rate": 6.716094032549729e-07, "loss": 0.0002, "step": 6488 }, { "epoch": 0.78, "learning_rate": 6.7124773960217e-07, "loss": 0.0002, "step": 6489 }, { "epoch": 0.78, "learning_rate": 6.708860759493671e-07, "loss": 0.0003, "step": 6490 }, { "epoch": 0.78, "learning_rate": 6.705244122965641e-07, "loss": 0.0007, "step": 6491 }, { "epoch": 0.78, "learning_rate": 6.701627486437613e-07, "loss": 0.0003, "step": 6492 }, { "epoch": 0.78, "learning_rate": 6.698010849909584e-07, "loss": 0.2187, "step": 6493 }, { "epoch": 0.78, "learning_rate": 6.694394213381555e-07, "loss": 0.0002, "step": 6494 }, { "epoch": 0.78, "learning_rate": 6.690777576853526e-07, "loss": 0.0003, "step": 6495 }, { "epoch": 0.78, "learning_rate": 6.687160940325497e-07, "loss": 0.0033, "step": 6496 }, { "epoch": 0.78, "learning_rate": 6.683544303797468e-07, "loss": 0.0002, "step": 6497 }, { "epoch": 0.78, "learning_rate": 6.679927667269439e-07, "loss": 0.0002, "step": 6498 }, { "epoch": 0.78, "learning_rate": 6.67631103074141e-07, "loss": 0.0008, "step": 6499 }, { "epoch": 0.78, "learning_rate": 6.672694394213382e-07, "loss": 0.0002, "step": 6500 }, { "epoch": 0.78, "learning_rate": 6.669077757685353e-07, "loss": 0.0004, "step": 6501 }, { "epoch": 0.78, "learning_rate": 6.665461121157323e-07, "loss": 0.0002, "step": 6502 }, { "epoch": 0.78, "learning_rate": 6.661844484629294e-07, "loss": 0.0003, "step": 6503 }, { "epoch": 0.78, "learning_rate": 6.658227848101266e-07, "loss": 0.0002, "step": 6504 }, { "epoch": 0.78, "learning_rate": 6.654611211573237e-07, "loss": 0.0023, "step": 6505 }, { "epoch": 0.78, "learning_rate": 6.650994575045208e-07, "loss": 0.0002, "step": 6506 }, { "epoch": 0.78, "learning_rate": 6.647377938517179e-07, "loss": 0.0004, "step": 6507 }, { "epoch": 0.78, "learning_rate": 6.64376130198915e-07, "loss": 0.0003, "step": 6508 }, { "epoch": 0.78, "learning_rate": 6.640144665461121e-07, "loss": 0.0002, "step": 6509 }, { "epoch": 0.78, "learning_rate": 6.636528028933092e-07, "loss": 0.0002, "step": 6510 }, { "epoch": 0.78, "learning_rate": 6.632911392405063e-07, "loss": 0.0002, "step": 6511 }, { "epoch": 0.78, "learning_rate": 6.629294755877035e-07, "loss": 0.0002, "step": 6512 }, { "epoch": 0.78, "learning_rate": 6.625678119349006e-07, "loss": 0.0002, "step": 6513 }, { "epoch": 0.78, "learning_rate": 6.622061482820976e-07, "loss": 0.0003, "step": 6514 }, { "epoch": 0.78, "learning_rate": 6.618444846292947e-07, "loss": 0.0003, "step": 6515 }, { "epoch": 0.78, "learning_rate": 6.614828209764919e-07, "loss": 0.0002, "step": 6516 }, { "epoch": 0.78, "learning_rate": 6.61121157323689e-07, "loss": 0.0002, "step": 6517 }, { "epoch": 0.78, "learning_rate": 6.607594936708861e-07, "loss": 0.0002, "step": 6518 }, { "epoch": 0.78, "learning_rate": 6.603978300180832e-07, "loss": 0.0002, "step": 6519 }, { "epoch": 0.78, "learning_rate": 6.600361663652802e-07, "loss": 0.0002, "step": 6520 }, { "epoch": 0.78, "learning_rate": 6.596745027124774e-07, "loss": 0.0003, "step": 6521 }, { "epoch": 0.78, "learning_rate": 6.593128390596745e-07, "loss": 0.0003, "step": 6522 }, { "epoch": 0.78, "learning_rate": 6.589511754068716e-07, "loss": 0.0003, "step": 6523 }, { "epoch": 0.78, "learning_rate": 6.585895117540687e-07, "loss": 0.039, "step": 6524 }, { "epoch": 0.78, "learning_rate": 6.582278481012659e-07, "loss": 0.0017, "step": 6525 }, { "epoch": 0.78, "learning_rate": 6.578661844484629e-07, "loss": 0.3734, "step": 6526 }, { "epoch": 0.78, "learning_rate": 6.5750452079566e-07, "loss": 0.0002, "step": 6527 }, { "epoch": 0.78, "learning_rate": 6.571428571428571e-07, "loss": 0.0003, "step": 6528 }, { "epoch": 0.78, "learning_rate": 6.567811934900543e-07, "loss": 0.0002, "step": 6529 }, { "epoch": 0.78, "learning_rate": 6.564195298372514e-07, "loss": 1.0551, "step": 6530 }, { "epoch": 0.78, "learning_rate": 6.560578661844485e-07, "loss": 0.0003, "step": 6531 }, { "epoch": 0.78, "learning_rate": 6.556962025316455e-07, "loss": 0.0003, "step": 6532 }, { "epoch": 0.78, "learning_rate": 6.553345388788427e-07, "loss": 0.0002, "step": 6533 }, { "epoch": 0.78, "learning_rate": 6.549728752260398e-07, "loss": 0.0003, "step": 6534 }, { "epoch": 0.78, "learning_rate": 6.546112115732369e-07, "loss": 0.0003, "step": 6535 }, { "epoch": 0.78, "learning_rate": 6.54249547920434e-07, "loss": 0.0002, "step": 6536 }, { "epoch": 0.78, "learning_rate": 6.538878842676312e-07, "loss": 0.0002, "step": 6537 }, { "epoch": 0.78, "learning_rate": 6.535262206148282e-07, "loss": 0.0004, "step": 6538 }, { "epoch": 0.78, "learning_rate": 6.531645569620253e-07, "loss": 0.0003, "step": 6539 }, { "epoch": 0.78, "learning_rate": 6.528028933092224e-07, "loss": 0.0002, "step": 6540 }, { "epoch": 0.78, "learning_rate": 6.524412296564196e-07, "loss": 0.0003, "step": 6541 }, { "epoch": 0.78, "learning_rate": 6.520795660036167e-07, "loss": 0.0002, "step": 6542 }, { "epoch": 0.78, "learning_rate": 6.517179023508138e-07, "loss": 0.0046, "step": 6543 }, { "epoch": 0.78, "learning_rate": 6.513562386980108e-07, "loss": 0.0002, "step": 6544 }, { "epoch": 0.78, "learning_rate": 6.50994575045208e-07, "loss": 0.0006, "step": 6545 }, { "epoch": 0.78, "learning_rate": 6.506329113924051e-07, "loss": 0.0002, "step": 6546 }, { "epoch": 0.78, "learning_rate": 6.502712477396022e-07, "loss": 0.0004, "step": 6547 }, { "epoch": 0.78, "learning_rate": 6.499095840867993e-07, "loss": 0.0035, "step": 6548 }, { "epoch": 0.78, "learning_rate": 6.495479204339965e-07, "loss": 0.0003, "step": 6549 }, { "epoch": 0.78, "learning_rate": 6.491862567811935e-07, "loss": 0.0014, "step": 6550 }, { "epoch": 0.79, "learning_rate": 6.488245931283906e-07, "loss": 0.0004, "step": 6551 }, { "epoch": 0.79, "learning_rate": 6.484629294755877e-07, "loss": 0.0002, "step": 6552 }, { "epoch": 0.79, "learning_rate": 6.481012658227849e-07, "loss": 0.0003, "step": 6553 }, { "epoch": 0.79, "learning_rate": 6.47739602169982e-07, "loss": 0.0004, "step": 6554 }, { "epoch": 0.79, "learning_rate": 6.473779385171791e-07, "loss": 0.0002, "step": 6555 }, { "epoch": 0.79, "learning_rate": 6.470162748643761e-07, "loss": 0.0003, "step": 6556 }, { "epoch": 0.79, "learning_rate": 6.466546112115732e-07, "loss": 0.0002, "step": 6557 }, { "epoch": 0.79, "learning_rate": 6.462929475587704e-07, "loss": 0.0002, "step": 6558 }, { "epoch": 0.79, "learning_rate": 6.459312839059675e-07, "loss": 0.0005, "step": 6559 }, { "epoch": 0.79, "learning_rate": 6.455696202531646e-07, "loss": 0.0001, "step": 6560 }, { "epoch": 0.79, "learning_rate": 6.452079566003617e-07, "loss": 0.0002, "step": 6561 }, { "epoch": 0.79, "learning_rate": 6.448462929475588e-07, "loss": 0.0002, "step": 6562 }, { "epoch": 0.79, "learning_rate": 6.444846292947559e-07, "loss": 0.0013, "step": 6563 }, { "epoch": 0.79, "learning_rate": 6.44122965641953e-07, "loss": 0.0002, "step": 6564 }, { "epoch": 0.79, "learning_rate": 6.437613019891501e-07, "loss": 0.0002, "step": 6565 }, { "epoch": 0.79, "learning_rate": 6.433996383363473e-07, "loss": 0.0002, "step": 6566 }, { "epoch": 0.79, "learning_rate": 6.430379746835444e-07, "loss": 0.0002, "step": 6567 }, { "epoch": 0.79, "learning_rate": 6.426763110307414e-07, "loss": 0.0004, "step": 6568 }, { "epoch": 0.79, "learning_rate": 6.423146473779385e-07, "loss": 0.0128, "step": 6569 }, { "epoch": 0.79, "learning_rate": 6.419529837251357e-07, "loss": 0.0002, "step": 6570 }, { "epoch": 0.79, "learning_rate": 6.415913200723328e-07, "loss": 0.9704, "step": 6571 }, { "epoch": 0.79, "learning_rate": 6.412296564195299e-07, "loss": 0.0002, "step": 6572 }, { "epoch": 0.79, "learning_rate": 6.40867992766727e-07, "loss": 0.0003, "step": 6573 }, { "epoch": 0.79, "learning_rate": 6.405063291139241e-07, "loss": 0.0573, "step": 6574 }, { "epoch": 0.79, "learning_rate": 6.401446654611212e-07, "loss": 0.0002, "step": 6575 }, { "epoch": 0.79, "learning_rate": 6.397830018083183e-07, "loss": 0.0003, "step": 6576 }, { "epoch": 0.79, "learning_rate": 6.394213381555154e-07, "loss": 0.0002, "step": 6577 }, { "epoch": 0.79, "learning_rate": 6.390596745027126e-07, "loss": 0.0002, "step": 6578 }, { "epoch": 0.79, "learning_rate": 6.386980108499097e-07, "loss": 0.0003, "step": 6579 }, { "epoch": 0.79, "learning_rate": 6.383363471971067e-07, "loss": 0.0031, "step": 6580 }, { "epoch": 0.79, "learning_rate": 6.379746835443038e-07, "loss": 0.0002, "step": 6581 }, { "epoch": 0.79, "learning_rate": 6.37613019891501e-07, "loss": 0.0005, "step": 6582 }, { "epoch": 0.79, "learning_rate": 6.372513562386981e-07, "loss": 0.0002, "step": 6583 }, { "epoch": 0.79, "learning_rate": 6.368896925858952e-07, "loss": 0.0004, "step": 6584 }, { "epoch": 0.79, "learning_rate": 6.365280289330923e-07, "loss": 0.0001, "step": 6585 }, { "epoch": 0.79, "learning_rate": 6.361663652802894e-07, "loss": 0.0003, "step": 6586 }, { "epoch": 0.79, "learning_rate": 6.358047016274865e-07, "loss": 0.0003, "step": 6587 }, { "epoch": 0.79, "learning_rate": 6.354430379746836e-07, "loss": 0.0002, "step": 6588 }, { "epoch": 0.79, "learning_rate": 6.350813743218807e-07, "loss": 0.7483, "step": 6589 }, { "epoch": 0.79, "learning_rate": 6.347197106690779e-07, "loss": 0.0003, "step": 6590 }, { "epoch": 0.79, "learning_rate": 6.343580470162749e-07, "loss": 0.0005, "step": 6591 }, { "epoch": 0.79, "learning_rate": 6.33996383363472e-07, "loss": 0.0003, "step": 6592 }, { "epoch": 0.79, "learning_rate": 6.336347197106691e-07, "loss": 0.0002, "step": 6593 }, { "epoch": 0.79, "learning_rate": 6.332730560578662e-07, "loss": 0.0004, "step": 6594 }, { "epoch": 0.79, "learning_rate": 6.329113924050634e-07, "loss": 0.1359, "step": 6595 }, { "epoch": 0.79, "learning_rate": 6.325497287522604e-07, "loss": 0.0003, "step": 6596 }, { "epoch": 0.79, "learning_rate": 6.321880650994575e-07, "loss": 0.0002, "step": 6597 }, { "epoch": 0.79, "learning_rate": 6.318264014466546e-07, "loss": 0.0013, "step": 6598 }, { "epoch": 0.79, "learning_rate": 6.314647377938518e-07, "loss": 0.0001, "step": 6599 }, { "epoch": 0.79, "learning_rate": 6.311030741410489e-07, "loss": 0.0003, "step": 6600 }, { "epoch": 0.79, "learning_rate": 6.307414104882459e-07, "loss": 0.0003, "step": 6601 }, { "epoch": 0.79, "learning_rate": 6.30379746835443e-07, "loss": 0.5011, "step": 6602 }, { "epoch": 0.79, "learning_rate": 6.300180831826402e-07, "loss": 0.0004, "step": 6603 }, { "epoch": 0.79, "learning_rate": 6.296564195298373e-07, "loss": 0.0014, "step": 6604 }, { "epoch": 0.79, "learning_rate": 6.292947558770344e-07, "loss": 0.0002, "step": 6605 }, { "epoch": 0.79, "learning_rate": 6.289330922242314e-07, "loss": 0.0004, "step": 6606 }, { "epoch": 0.79, "learning_rate": 6.285714285714286e-07, "loss": 0.0003, "step": 6607 }, { "epoch": 0.79, "learning_rate": 6.282097649186257e-07, "loss": 0.0002, "step": 6608 }, { "epoch": 0.79, "learning_rate": 6.278481012658228e-07, "loss": 0.0002, "step": 6609 }, { "epoch": 0.79, "learning_rate": 6.274864376130198e-07, "loss": 0.0002, "step": 6610 }, { "epoch": 0.79, "learning_rate": 6.27124773960217e-07, "loss": 0.0002, "step": 6611 }, { "epoch": 0.79, "learning_rate": 6.267631103074141e-07, "loss": 0.0002, "step": 6612 }, { "epoch": 0.79, "learning_rate": 6.264014466546112e-07, "loss": 0.0002, "step": 6613 }, { "epoch": 0.79, "learning_rate": 6.260397830018083e-07, "loss": 0.0009, "step": 6614 }, { "epoch": 0.79, "learning_rate": 6.256781193490053e-07, "loss": 0.0003, "step": 6615 }, { "epoch": 0.79, "learning_rate": 6.253164556962025e-07, "loss": 0.0003, "step": 6616 }, { "epoch": 0.79, "learning_rate": 6.249547920433996e-07, "loss": 0.0003, "step": 6617 }, { "epoch": 0.79, "learning_rate": 6.245931283905967e-07, "loss": 0.0002, "step": 6618 }, { "epoch": 0.79, "learning_rate": 6.242314647377938e-07, "loss": 0.0002, "step": 6619 }, { "epoch": 0.79, "learning_rate": 6.23869801084991e-07, "loss": 0.0002, "step": 6620 }, { "epoch": 0.79, "learning_rate": 6.23508137432188e-07, "loss": 0.0017, "step": 6621 }, { "epoch": 0.79, "learning_rate": 6.231464737793851e-07, "loss": 0.8656, "step": 6622 }, { "epoch": 0.79, "learning_rate": 6.227848101265822e-07, "loss": 0.0003, "step": 6623 }, { "epoch": 0.79, "learning_rate": 6.224231464737794e-07, "loss": 0.4041, "step": 6624 }, { "epoch": 0.79, "learning_rate": 6.220614828209765e-07, "loss": 0.0004, "step": 6625 }, { "epoch": 0.79, "learning_rate": 6.216998191681736e-07, "loss": 0.0002, "step": 6626 }, { "epoch": 0.79, "learning_rate": 6.213381555153706e-07, "loss": 0.0002, "step": 6627 }, { "epoch": 0.79, "learning_rate": 6.209764918625678e-07, "loss": 0.0006, "step": 6628 }, { "epoch": 0.79, "learning_rate": 6.206148282097649e-07, "loss": 0.0002, "step": 6629 }, { "epoch": 0.79, "learning_rate": 6.20253164556962e-07, "loss": 0.0003, "step": 6630 }, { "epoch": 0.79, "learning_rate": 6.198915009041591e-07, "loss": 0.0002, "step": 6631 }, { "epoch": 0.79, "learning_rate": 6.195298372513563e-07, "loss": 0.0001, "step": 6632 }, { "epoch": 0.79, "learning_rate": 6.191681735985533e-07, "loss": 0.2051, "step": 6633 }, { "epoch": 0.79, "learning_rate": 6.188065099457504e-07, "loss": 0.0002, "step": 6634 }, { "epoch": 0.8, "learning_rate": 6.184448462929475e-07, "loss": 0.0002, "step": 6635 }, { "epoch": 0.8, "learning_rate": 6.180831826401447e-07, "loss": 0.0002, "step": 6636 }, { "epoch": 0.8, "learning_rate": 6.177215189873418e-07, "loss": 0.0002, "step": 6637 }, { "epoch": 0.8, "learning_rate": 6.173598553345389e-07, "loss": 0.4301, "step": 6638 }, { "epoch": 0.8, "learning_rate": 6.169981916817359e-07, "loss": 0.0001, "step": 6639 }, { "epoch": 0.8, "learning_rate": 6.166365280289331e-07, "loss": 0.0001, "step": 6640 }, { "epoch": 0.8, "learning_rate": 6.162748643761302e-07, "loss": 0.0002, "step": 6641 }, { "epoch": 0.8, "learning_rate": 6.159132007233273e-07, "loss": 0.0002, "step": 6642 }, { "epoch": 0.8, "learning_rate": 6.155515370705244e-07, "loss": 0.0002, "step": 6643 }, { "epoch": 0.8, "learning_rate": 6.151898734177216e-07, "loss": 0.6564, "step": 6644 }, { "epoch": 0.8, "learning_rate": 6.148282097649186e-07, "loss": 0.0014, "step": 6645 }, { "epoch": 0.8, "learning_rate": 6.144665461121157e-07, "loss": 0.0002, "step": 6646 }, { "epoch": 0.8, "learning_rate": 6.141048824593128e-07, "loss": 0.0003, "step": 6647 }, { "epoch": 0.8, "learning_rate": 6.1374321880651e-07, "loss": 0.0002, "step": 6648 }, { "epoch": 0.8, "learning_rate": 6.133815551537071e-07, "loss": 0.0003, "step": 6649 }, { "epoch": 0.8, "learning_rate": 6.130198915009042e-07, "loss": 0.7378, "step": 6650 }, { "epoch": 0.8, "learning_rate": 6.126582278481012e-07, "loss": 0.0002, "step": 6651 }, { "epoch": 0.8, "learning_rate": 6.122965641952983e-07, "loss": 0.0004, "step": 6652 }, { "epoch": 0.8, "learning_rate": 6.119349005424955e-07, "loss": 0.0003, "step": 6653 }, { "epoch": 0.8, "learning_rate": 6.115732368896926e-07, "loss": 0.0002, "step": 6654 }, { "epoch": 0.8, "learning_rate": 6.112115732368897e-07, "loss": 0.0034, "step": 6655 }, { "epoch": 0.8, "learning_rate": 6.108499095840868e-07, "loss": 0.0212, "step": 6656 }, { "epoch": 0.8, "learning_rate": 6.104882459312839e-07, "loss": 0.0002, "step": 6657 }, { "epoch": 0.8, "learning_rate": 6.10126582278481e-07, "loss": 0.0002, "step": 6658 }, { "epoch": 0.8, "learning_rate": 6.097649186256781e-07, "loss": 0.0002, "step": 6659 }, { "epoch": 0.8, "learning_rate": 6.094032549728752e-07, "loss": 0.0003, "step": 6660 }, { "epoch": 0.8, "learning_rate": 6.090415913200724e-07, "loss": 1.0346, "step": 6661 }, { "epoch": 0.8, "learning_rate": 6.086799276672695e-07, "loss": 0.0003, "step": 6662 }, { "epoch": 0.8, "learning_rate": 6.083182640144665e-07, "loss": 0.0003, "step": 6663 }, { "epoch": 0.8, "learning_rate": 6.079566003616636e-07, "loss": 0.1048, "step": 6664 }, { "epoch": 0.8, "learning_rate": 6.075949367088608e-07, "loss": 0.0002, "step": 6665 }, { "epoch": 0.8, "learning_rate": 6.072332730560579e-07, "loss": 0.0006, "step": 6666 }, { "epoch": 0.8, "learning_rate": 6.06871609403255e-07, "loss": 0.9012, "step": 6667 }, { "epoch": 0.8, "learning_rate": 6.065099457504521e-07, "loss": 1.0183, "step": 6668 }, { "epoch": 0.8, "learning_rate": 6.061482820976492e-07, "loss": 0.0002, "step": 6669 }, { "epoch": 0.8, "learning_rate": 6.057866184448463e-07, "loss": 0.0004, "step": 6670 }, { "epoch": 0.8, "learning_rate": 6.054249547920434e-07, "loss": 0.0002, "step": 6671 }, { "epoch": 0.8, "learning_rate": 6.050632911392405e-07, "loss": 0.0002, "step": 6672 }, { "epoch": 0.8, "learning_rate": 6.047016274864377e-07, "loss": 0.0003, "step": 6673 }, { "epoch": 0.8, "learning_rate": 6.043399638336348e-07, "loss": 0.0003, "step": 6674 }, { "epoch": 0.8, "learning_rate": 6.039783001808318e-07, "loss": 0.0003, "step": 6675 }, { "epoch": 0.8, "learning_rate": 6.036166365280289e-07, "loss": 0.0003, "step": 6676 }, { "epoch": 0.8, "learning_rate": 6.032549728752261e-07, "loss": 0.9793, "step": 6677 }, { "epoch": 0.8, "learning_rate": 6.028933092224232e-07, "loss": 0.0003, "step": 6678 }, { "epoch": 0.8, "learning_rate": 6.025316455696203e-07, "loss": 0.0003, "step": 6679 }, { "epoch": 0.8, "learning_rate": 6.021699819168174e-07, "loss": 0.5649, "step": 6680 }, { "epoch": 0.8, "learning_rate": 6.018083182640145e-07, "loss": 0.0001, "step": 6681 }, { "epoch": 0.8, "learning_rate": 6.014466546112116e-07, "loss": 0.0006, "step": 6682 }, { "epoch": 0.8, "learning_rate": 6.010849909584087e-07, "loss": 0.0007, "step": 6683 }, { "epoch": 0.8, "learning_rate": 6.007233273056058e-07, "loss": 0.0003, "step": 6684 }, { "epoch": 0.8, "learning_rate": 6.00361663652803e-07, "loss": 0.0006, "step": 6685 }, { "epoch": 0.8, "learning_rate": 6.000000000000001e-07, "loss": 0.0004, "step": 6686 }, { "epoch": 0.8, "learning_rate": 5.996383363471971e-07, "loss": 0.0003, "step": 6687 }, { "epoch": 0.8, "learning_rate": 5.992766726943942e-07, "loss": 0.001, "step": 6688 }, { "epoch": 0.8, "learning_rate": 5.989150090415913e-07, "loss": 0.0007, "step": 6689 }, { "epoch": 0.8, "learning_rate": 5.985533453887885e-07, "loss": 0.0106, "step": 6690 }, { "epoch": 0.8, "learning_rate": 5.981916817359856e-07, "loss": 0.0003, "step": 6691 }, { "epoch": 0.8, "learning_rate": 5.978300180831827e-07, "loss": 0.1945, "step": 6692 }, { "epoch": 0.8, "learning_rate": 5.974683544303797e-07, "loss": 0.0005, "step": 6693 }, { "epoch": 0.8, "learning_rate": 5.971066907775769e-07, "loss": 0.0047, "step": 6694 }, { "epoch": 0.8, "learning_rate": 5.96745027124774e-07, "loss": 0.0006, "step": 6695 }, { "epoch": 0.8, "learning_rate": 5.963833634719711e-07, "loss": 0.1268, "step": 6696 }, { "epoch": 0.8, "learning_rate": 5.960216998191682e-07, "loss": 0.0003, "step": 6697 }, { "epoch": 0.8, "learning_rate": 5.956600361663654e-07, "loss": 0.0002, "step": 6698 }, { "epoch": 0.8, "learning_rate": 5.952983725135624e-07, "loss": 0.0002, "step": 6699 }, { "epoch": 0.8, "learning_rate": 5.949367088607595e-07, "loss": 0.0002, "step": 6700 }, { "epoch": 0.8, "learning_rate": 5.945750452079566e-07, "loss": 0.0003, "step": 6701 }, { "epoch": 0.8, "learning_rate": 5.942133815551538e-07, "loss": 1.0721, "step": 6702 }, { "epoch": 0.8, "learning_rate": 5.938517179023509e-07, "loss": 0.4715, "step": 6703 }, { "epoch": 0.8, "learning_rate": 5.93490054249548e-07, "loss": 0.0008, "step": 6704 }, { "epoch": 0.8, "learning_rate": 5.93128390596745e-07, "loss": 0.0004, "step": 6705 }, { "epoch": 0.8, "learning_rate": 5.927667269439422e-07, "loss": 0.0009, "step": 6706 }, { "epoch": 0.8, "learning_rate": 5.924050632911393e-07, "loss": 1.0649, "step": 6707 }, { "epoch": 0.8, "learning_rate": 5.920433996383364e-07, "loss": 0.0002, "step": 6708 }, { "epoch": 0.8, "learning_rate": 5.916817359855335e-07, "loss": 0.0002, "step": 6709 }, { "epoch": 0.8, "learning_rate": 5.913200723327307e-07, "loss": 0.0003, "step": 6710 }, { "epoch": 0.8, "learning_rate": 5.909584086799277e-07, "loss": 0.0002, "step": 6711 }, { "epoch": 0.8, "learning_rate": 5.905967450271248e-07, "loss": 0.0079, "step": 6712 }, { "epoch": 0.8, "learning_rate": 5.902350813743219e-07, "loss": 0.0003, "step": 6713 }, { "epoch": 0.8, "learning_rate": 5.898734177215191e-07, "loss": 0.0005, "step": 6714 }, { "epoch": 0.8, "learning_rate": 5.895117540687162e-07, "loss": 0.0018, "step": 6715 }, { "epoch": 0.8, "learning_rate": 5.891500904159132e-07, "loss": 0.0004, "step": 6716 }, { "epoch": 0.8, "learning_rate": 5.887884267631103e-07, "loss": 0.0003, "step": 6717 }, { "epoch": 0.81, "learning_rate": 5.884267631103075e-07, "loss": 0.0002, "step": 6718 }, { "epoch": 0.81, "learning_rate": 5.880650994575046e-07, "loss": 0.0017, "step": 6719 }, { "epoch": 0.81, "learning_rate": 5.877034358047016e-07, "loss": 0.0002, "step": 6720 }, { "epoch": 0.81, "learning_rate": 5.873417721518987e-07, "loss": 0.0004, "step": 6721 }, { "epoch": 0.81, "learning_rate": 5.869801084990958e-07, "loss": 0.0013, "step": 6722 }, { "epoch": 0.81, "learning_rate": 5.86618444846293e-07, "loss": 0.0004, "step": 6723 }, { "epoch": 0.81, "learning_rate": 5.862567811934901e-07, "loss": 0.0003, "step": 6724 }, { "epoch": 0.81, "learning_rate": 5.858951175406871e-07, "loss": 0.0004, "step": 6725 }, { "epoch": 0.81, "learning_rate": 5.855334538878842e-07, "loss": 0.0003, "step": 6726 }, { "epoch": 0.81, "learning_rate": 5.851717902350814e-07, "loss": 0.0007, "step": 6727 }, { "epoch": 0.81, "learning_rate": 5.848101265822785e-07, "loss": 0.0005, "step": 6728 }, { "epoch": 0.81, "learning_rate": 5.844484629294756e-07, "loss": 0.0003, "step": 6729 }, { "epoch": 0.81, "learning_rate": 5.840867992766726e-07, "loss": 0.0004, "step": 6730 }, { "epoch": 0.81, "learning_rate": 5.837251356238698e-07, "loss": 0.0003, "step": 6731 }, { "epoch": 0.81, "learning_rate": 5.833634719710669e-07, "loss": 0.0002, "step": 6732 }, { "epoch": 0.81, "learning_rate": 5.83001808318264e-07, "loss": 0.0129, "step": 6733 }, { "epoch": 0.81, "learning_rate": 5.826401446654611e-07, "loss": 0.0002, "step": 6734 }, { "epoch": 0.81, "learning_rate": 5.822784810126582e-07, "loss": 0.0003, "step": 6735 }, { "epoch": 0.81, "learning_rate": 5.819168173598553e-07, "loss": 0.8634, "step": 6736 }, { "epoch": 0.81, "learning_rate": 5.815551537070524e-07, "loss": 0.0003, "step": 6737 }, { "epoch": 0.81, "learning_rate": 5.811934900542495e-07, "loss": 0.0003, "step": 6738 }, { "epoch": 0.81, "learning_rate": 5.808318264014467e-07, "loss": 0.0002, "step": 6739 }, { "epoch": 0.81, "learning_rate": 5.804701627486438e-07, "loss": 0.0003, "step": 6740 }, { "epoch": 0.81, "learning_rate": 5.801084990958408e-07, "loss": 0.8731, "step": 6741 }, { "epoch": 0.81, "learning_rate": 5.797468354430379e-07, "loss": 0.0004, "step": 6742 }, { "epoch": 0.81, "learning_rate": 5.793851717902351e-07, "loss": 0.001, "step": 6743 }, { "epoch": 0.81, "learning_rate": 5.790235081374322e-07, "loss": 0.0003, "step": 6744 }, { "epoch": 0.81, "learning_rate": 5.786618444846293e-07, "loss": 0.1322, "step": 6745 }, { "epoch": 0.81, "learning_rate": 5.783001808318264e-07, "loss": 0.0683, "step": 6746 }, { "epoch": 0.81, "learning_rate": 5.779385171790235e-07, "loss": 0.0002, "step": 6747 }, { "epoch": 0.81, "learning_rate": 5.775768535262206e-07, "loss": 0.9317, "step": 6748 }, { "epoch": 0.81, "learning_rate": 5.772151898734177e-07, "loss": 0.0003, "step": 6749 }, { "epoch": 0.81, "learning_rate": 5.768535262206148e-07, "loss": 0.0003, "step": 6750 }, { "epoch": 0.81, "learning_rate": 5.76491862567812e-07, "loss": 0.147, "step": 6751 }, { "epoch": 0.81, "learning_rate": 5.761301989150091e-07, "loss": 0.0003, "step": 6752 }, { "epoch": 0.81, "learning_rate": 5.757685352622061e-07, "loss": 0.2253, "step": 6753 }, { "epoch": 0.81, "learning_rate": 5.754068716094032e-07, "loss": 0.0003, "step": 6754 }, { "epoch": 0.81, "learning_rate": 5.750452079566003e-07, "loss": 0.0004, "step": 6755 }, { "epoch": 0.81, "learning_rate": 5.746835443037975e-07, "loss": 0.0003, "step": 6756 }, { "epoch": 0.81, "learning_rate": 5.743218806509946e-07, "loss": 0.0005, "step": 6757 }, { "epoch": 0.81, "learning_rate": 5.739602169981917e-07, "loss": 0.0003, "step": 6758 }, { "epoch": 0.81, "learning_rate": 5.735985533453887e-07, "loss": 0.0001, "step": 6759 }, { "epoch": 0.81, "learning_rate": 5.732368896925859e-07, "loss": 0.0003, "step": 6760 }, { "epoch": 0.81, "learning_rate": 5.72875226039783e-07, "loss": 0.0004, "step": 6761 }, { "epoch": 0.81, "learning_rate": 5.725135623869801e-07, "loss": 0.0004, "step": 6762 }, { "epoch": 0.81, "learning_rate": 5.721518987341772e-07, "loss": 0.0002, "step": 6763 }, { "epoch": 0.81, "learning_rate": 5.717902350813744e-07, "loss": 0.0003, "step": 6764 }, { "epoch": 0.81, "learning_rate": 5.714285714285714e-07, "loss": 0.5956, "step": 6765 }, { "epoch": 0.81, "learning_rate": 5.710669077757685e-07, "loss": 0.0004, "step": 6766 }, { "epoch": 0.81, "learning_rate": 5.707052441229656e-07, "loss": 0.0004, "step": 6767 }, { "epoch": 0.81, "learning_rate": 5.703435804701628e-07, "loss": 0.0105, "step": 6768 }, { "epoch": 0.81, "learning_rate": 5.699819168173599e-07, "loss": 0.0004, "step": 6769 }, { "epoch": 0.81, "learning_rate": 5.69620253164557e-07, "loss": 0.8494, "step": 6770 }, { "epoch": 0.81, "learning_rate": 5.69258589511754e-07, "loss": 0.0002, "step": 6771 }, { "epoch": 0.81, "learning_rate": 5.688969258589512e-07, "loss": 0.0007, "step": 6772 }, { "epoch": 0.81, "learning_rate": 5.685352622061483e-07, "loss": 0.0002, "step": 6773 }, { "epoch": 0.81, "learning_rate": 5.681735985533454e-07, "loss": 0.0009, "step": 6774 }, { "epoch": 0.81, "learning_rate": 5.678119349005425e-07, "loss": 0.0005, "step": 6775 }, { "epoch": 0.81, "learning_rate": 5.674502712477397e-07, "loss": 0.0004, "step": 6776 }, { "epoch": 0.81, "learning_rate": 5.670886075949367e-07, "loss": 0.0005, "step": 6777 }, { "epoch": 0.81, "learning_rate": 5.667269439421338e-07, "loss": 0.0003, "step": 6778 }, { "epoch": 0.81, "learning_rate": 5.663652802893309e-07, "loss": 0.0003, "step": 6779 }, { "epoch": 0.81, "learning_rate": 5.660036166365281e-07, "loss": 0.0004, "step": 6780 }, { "epoch": 0.81, "learning_rate": 5.656419529837252e-07, "loss": 0.0005, "step": 6781 }, { "epoch": 0.81, "learning_rate": 5.652802893309223e-07, "loss": 1.0327, "step": 6782 }, { "epoch": 0.81, "learning_rate": 5.649186256781193e-07, "loss": 0.0005, "step": 6783 }, { "epoch": 0.81, "learning_rate": 5.645569620253165e-07, "loss": 0.0004, "step": 6784 }, { "epoch": 0.81, "learning_rate": 5.641952983725136e-07, "loss": 0.0003, "step": 6785 }, { "epoch": 0.81, "learning_rate": 5.638336347197107e-07, "loss": 0.0005, "step": 6786 }, { "epoch": 0.81, "learning_rate": 5.634719710669078e-07, "loss": 0.0005, "step": 6787 }, { "epoch": 0.81, "learning_rate": 5.63110307414105e-07, "loss": 0.0033, "step": 6788 }, { "epoch": 0.81, "learning_rate": 5.62748643761302e-07, "loss": 0.4977, "step": 6789 }, { "epoch": 0.81, "learning_rate": 5.623869801084991e-07, "loss": 0.0004, "step": 6790 }, { "epoch": 0.81, "learning_rate": 5.620253164556962e-07, "loss": 0.0004, "step": 6791 }, { "epoch": 0.81, "learning_rate": 5.616636528028933e-07, "loss": 0.0005, "step": 6792 }, { "epoch": 0.81, "learning_rate": 5.613019891500905e-07, "loss": 0.0005, "step": 6793 }, { "epoch": 0.81, "learning_rate": 5.609403254972876e-07, "loss": 0.5843, "step": 6794 }, { "epoch": 0.81, "learning_rate": 5.605786618444846e-07, "loss": 0.0004, "step": 6795 }, { "epoch": 0.81, "learning_rate": 5.602169981916817e-07, "loss": 0.0006, "step": 6796 }, { "epoch": 0.81, "learning_rate": 5.598553345388789e-07, "loss": 0.0013, "step": 6797 }, { "epoch": 0.81, "learning_rate": 5.59493670886076e-07, "loss": 0.0003, "step": 6798 }, { "epoch": 0.81, "learning_rate": 5.591320072332731e-07, "loss": 0.0005, "step": 6799 }, { "epoch": 0.81, "learning_rate": 5.587703435804702e-07, "loss": 0.0005, "step": 6800 }, { "epoch": 0.81, "learning_rate": 5.584086799276673e-07, "loss": 0.0003, "step": 6801 }, { "epoch": 0.82, "learning_rate": 5.580470162748644e-07, "loss": 0.0003, "step": 6802 }, { "epoch": 0.82, "learning_rate": 5.576853526220615e-07, "loss": 0.0006, "step": 6803 }, { "epoch": 0.82, "learning_rate": 5.573236889692586e-07, "loss": 0.0007, "step": 6804 }, { "epoch": 0.82, "learning_rate": 5.569620253164558e-07, "loss": 0.01, "step": 6805 }, { "epoch": 0.82, "learning_rate": 5.566003616636529e-07, "loss": 0.0109, "step": 6806 }, { "epoch": 0.82, "learning_rate": 5.562386980108499e-07, "loss": 0.0299, "step": 6807 }, { "epoch": 0.82, "learning_rate": 5.55877034358047e-07, "loss": 0.0007, "step": 6808 }, { "epoch": 0.82, "learning_rate": 5.555153707052442e-07, "loss": 0.0006, "step": 6809 }, { "epoch": 0.82, "learning_rate": 5.551537070524413e-07, "loss": 0.0003, "step": 6810 }, { "epoch": 0.82, "learning_rate": 5.547920433996384e-07, "loss": 0.0006, "step": 6811 }, { "epoch": 0.82, "learning_rate": 5.544303797468355e-07, "loss": 0.0003, "step": 6812 }, { "epoch": 0.82, "learning_rate": 5.540687160940326e-07, "loss": 0.0002, "step": 6813 }, { "epoch": 0.82, "learning_rate": 5.537070524412297e-07, "loss": 0.0007, "step": 6814 }, { "epoch": 0.82, "learning_rate": 5.533453887884268e-07, "loss": 0.0004, "step": 6815 }, { "epoch": 0.82, "learning_rate": 5.529837251356239e-07, "loss": 0.0006, "step": 6816 }, { "epoch": 0.82, "learning_rate": 5.526220614828211e-07, "loss": 0.0005, "step": 6817 }, { "epoch": 0.82, "learning_rate": 5.522603978300182e-07, "loss": 0.0035, "step": 6818 }, { "epoch": 0.82, "learning_rate": 5.518987341772152e-07, "loss": 0.0003, "step": 6819 }, { "epoch": 0.82, "learning_rate": 5.515370705244123e-07, "loss": 0.0003, "step": 6820 }, { "epoch": 0.82, "learning_rate": 5.511754068716095e-07, "loss": 0.0004, "step": 6821 }, { "epoch": 0.82, "learning_rate": 5.508137432188066e-07, "loss": 0.0969, "step": 6822 }, { "epoch": 0.82, "learning_rate": 5.504520795660037e-07, "loss": 0.0004, "step": 6823 }, { "epoch": 0.82, "learning_rate": 5.500904159132008e-07, "loss": 0.0003, "step": 6824 }, { "epoch": 0.82, "learning_rate": 5.497287522603978e-07, "loss": 0.0003, "step": 6825 }, { "epoch": 0.82, "learning_rate": 5.49367088607595e-07, "loss": 0.0003, "step": 6826 }, { "epoch": 0.82, "learning_rate": 5.490054249547921e-07, "loss": 0.0004, "step": 6827 }, { "epoch": 0.82, "learning_rate": 5.486437613019892e-07, "loss": 0.0003, "step": 6828 }, { "epoch": 0.82, "learning_rate": 5.482820976491863e-07, "loss": 0.0003, "step": 6829 }, { "epoch": 0.82, "learning_rate": 5.479204339963834e-07, "loss": 0.0004, "step": 6830 }, { "epoch": 0.82, "learning_rate": 5.475587703435805e-07, "loss": 0.0005, "step": 6831 }, { "epoch": 0.82, "learning_rate": 5.471971066907776e-07, "loss": 0.0003, "step": 6832 }, { "epoch": 0.82, "learning_rate": 5.468354430379747e-07, "loss": 0.0004, "step": 6833 }, { "epoch": 0.82, "learning_rate": 5.464737793851719e-07, "loss": 0.0004, "step": 6834 }, { "epoch": 0.82, "learning_rate": 5.461121157323689e-07, "loss": 0.5114, "step": 6835 }, { "epoch": 0.82, "learning_rate": 5.45750452079566e-07, "loss": 0.0003, "step": 6836 }, { "epoch": 0.82, "learning_rate": 5.453887884267631e-07, "loss": 0.0003, "step": 6837 }, { "epoch": 0.82, "learning_rate": 5.450271247739603e-07, "loss": 0.9517, "step": 6838 }, { "epoch": 0.82, "learning_rate": 5.446654611211574e-07, "loss": 0.0004, "step": 6839 }, { "epoch": 0.82, "learning_rate": 5.443037974683544e-07, "loss": 0.0003, "step": 6840 }, { "epoch": 0.82, "learning_rate": 5.439421338155515e-07, "loss": 0.0003, "step": 6841 }, { "epoch": 0.82, "learning_rate": 5.435804701627487e-07, "loss": 0.0002, "step": 6842 }, { "epoch": 0.82, "learning_rate": 5.432188065099458e-07, "loss": 0.0004, "step": 6843 }, { "epoch": 0.82, "learning_rate": 5.428571428571429e-07, "loss": 0.0004, "step": 6844 }, { "epoch": 0.82, "learning_rate": 5.424954792043399e-07, "loss": 0.2617, "step": 6845 }, { "epoch": 0.82, "learning_rate": 5.42133815551537e-07, "loss": 0.0004, "step": 6846 }, { "epoch": 0.82, "learning_rate": 5.417721518987342e-07, "loss": 0.0075, "step": 6847 }, { "epoch": 0.82, "learning_rate": 5.414104882459313e-07, "loss": 0.0003, "step": 6848 }, { "epoch": 0.82, "learning_rate": 5.410488245931284e-07, "loss": 0.0004, "step": 6849 }, { "epoch": 0.82, "learning_rate": 5.406871609403254e-07, "loss": 0.0003, "step": 6850 }, { "epoch": 0.82, "learning_rate": 5.403254972875226e-07, "loss": 0.0005, "step": 6851 }, { "epoch": 0.82, "learning_rate": 5.399638336347197e-07, "loss": 0.5981, "step": 6852 }, { "epoch": 0.82, "learning_rate": 5.396021699819168e-07, "loss": 0.0007, "step": 6853 }, { "epoch": 0.82, "learning_rate": 5.39240506329114e-07, "loss": 0.0003, "step": 6854 }, { "epoch": 0.82, "learning_rate": 5.38878842676311e-07, "loss": 0.0002, "step": 6855 }, { "epoch": 0.82, "learning_rate": 5.385171790235081e-07, "loss": 0.0008, "step": 6856 }, { "epoch": 0.82, "learning_rate": 5.381555153707052e-07, "loss": 0.0004, "step": 6857 }, { "epoch": 0.82, "learning_rate": 5.377938517179023e-07, "loss": 0.0002, "step": 6858 }, { "epoch": 0.82, "learning_rate": 5.374321880650995e-07, "loss": 0.0003, "step": 6859 }, { "epoch": 0.82, "learning_rate": 5.370705244122965e-07, "loss": 0.0007, "step": 6860 }, { "epoch": 0.82, "learning_rate": 5.367088607594936e-07, "loss": 0.0029, "step": 6861 }, { "epoch": 0.82, "learning_rate": 5.363471971066907e-07, "loss": 0.0003, "step": 6862 }, { "epoch": 0.82, "learning_rate": 5.359855334538879e-07, "loss": 0.0273, "step": 6863 }, { "epoch": 0.82, "learning_rate": 5.35623869801085e-07, "loss": 0.0003, "step": 6864 }, { "epoch": 0.82, "learning_rate": 5.352622061482821e-07, "loss": 0.0003, "step": 6865 }, { "epoch": 0.82, "learning_rate": 5.349005424954791e-07, "loss": 0.0003, "step": 6866 }, { "epoch": 0.82, "learning_rate": 5.345388788426763e-07, "loss": 0.0007, "step": 6867 }, { "epoch": 0.82, "learning_rate": 5.341772151898734e-07, "loss": 0.0004, "step": 6868 }, { "epoch": 0.82, "learning_rate": 5.338155515370705e-07, "loss": 0.0005, "step": 6869 }, { "epoch": 0.82, "learning_rate": 5.334538878842676e-07, "loss": 0.0002, "step": 6870 }, { "epoch": 0.82, "learning_rate": 5.330922242314648e-07, "loss": 0.001, "step": 6871 }, { "epoch": 0.82, "learning_rate": 5.327305605786618e-07, "loss": 0.8651, "step": 6872 }, { "epoch": 0.82, "learning_rate": 5.323688969258589e-07, "loss": 0.0003, "step": 6873 }, { "epoch": 0.82, "learning_rate": 5.32007233273056e-07, "loss": 0.0003, "step": 6874 }, { "epoch": 0.82, "learning_rate": 5.316455696202532e-07, "loss": 0.0002, "step": 6875 }, { "epoch": 0.82, "learning_rate": 5.312839059674503e-07, "loss": 0.0003, "step": 6876 }, { "epoch": 0.82, "learning_rate": 5.309222423146474e-07, "loss": 0.0003, "step": 6877 }, { "epoch": 0.82, "learning_rate": 5.305605786618444e-07, "loss": 0.0002, "step": 6878 }, { "epoch": 0.82, "learning_rate": 5.301989150090416e-07, "loss": 0.0004, "step": 6879 }, { "epoch": 0.82, "learning_rate": 5.298372513562387e-07, "loss": 0.0002, "step": 6880 }, { "epoch": 0.82, "learning_rate": 5.294755877034358e-07, "loss": 0.0004, "step": 6881 }, { "epoch": 0.82, "learning_rate": 5.291139240506329e-07, "loss": 0.0011, "step": 6882 }, { "epoch": 0.82, "learning_rate": 5.2875226039783e-07, "loss": 0.0003, "step": 6883 }, { "epoch": 0.82, "learning_rate": 5.283905967450271e-07, "loss": 0.0001, "step": 6884 }, { "epoch": 0.83, "learning_rate": 5.280289330922242e-07, "loss": 0.2938, "step": 6885 }, { "epoch": 0.83, "learning_rate": 5.276672694394213e-07, "loss": 0.5266, "step": 6886 }, { "epoch": 0.83, "learning_rate": 5.273056057866184e-07, "loss": 0.0004, "step": 6887 }, { "epoch": 0.83, "learning_rate": 5.269439421338156e-07, "loss": 0.0004, "step": 6888 }, { "epoch": 0.83, "learning_rate": 5.265822784810127e-07, "loss": 0.0003, "step": 6889 }, { "epoch": 0.83, "learning_rate": 5.262206148282097e-07, "loss": 0.0002, "step": 6890 }, { "epoch": 0.83, "learning_rate": 5.258589511754068e-07, "loss": 0.0004, "step": 6891 }, { "epoch": 0.83, "learning_rate": 5.25497287522604e-07, "loss": 0.6306, "step": 6892 }, { "epoch": 0.83, "learning_rate": 5.251356238698011e-07, "loss": 0.0003, "step": 6893 }, { "epoch": 0.83, "learning_rate": 5.247739602169982e-07, "loss": 0.0034, "step": 6894 }, { "epoch": 0.83, "learning_rate": 5.244122965641953e-07, "loss": 0.0003, "step": 6895 }, { "epoch": 0.83, "learning_rate": 5.240506329113924e-07, "loss": 0.0002, "step": 6896 }, { "epoch": 0.83, "learning_rate": 5.236889692585895e-07, "loss": 0.0006, "step": 6897 }, { "epoch": 0.83, "learning_rate": 5.233273056057866e-07, "loss": 0.0002, "step": 6898 }, { "epoch": 0.83, "learning_rate": 5.229656419529837e-07, "loss": 0.7266, "step": 6899 }, { "epoch": 0.83, "learning_rate": 5.226039783001809e-07, "loss": 0.0005, "step": 6900 }, { "epoch": 0.83, "learning_rate": 5.22242314647378e-07, "loss": 0.1166, "step": 6901 }, { "epoch": 0.83, "learning_rate": 5.21880650994575e-07, "loss": 0.0004, "step": 6902 }, { "epoch": 0.83, "learning_rate": 5.215189873417721e-07, "loss": 1.0247, "step": 6903 }, { "epoch": 0.83, "learning_rate": 5.211573236889693e-07, "loss": 0.0006, "step": 6904 }, { "epoch": 0.83, "learning_rate": 5.207956600361664e-07, "loss": 0.0005, "step": 6905 }, { "epoch": 0.83, "learning_rate": 5.204339963833635e-07, "loss": 0.0002, "step": 6906 }, { "epoch": 0.83, "learning_rate": 5.200723327305606e-07, "loss": 0.0009, "step": 6907 }, { "epoch": 0.83, "learning_rate": 5.197106690777577e-07, "loss": 0.0656, "step": 6908 }, { "epoch": 0.83, "learning_rate": 5.193490054249548e-07, "loss": 0.0006, "step": 6909 }, { "epoch": 0.83, "learning_rate": 5.189873417721519e-07, "loss": 0.0003, "step": 6910 }, { "epoch": 0.83, "learning_rate": 5.18625678119349e-07, "loss": 0.0006, "step": 6911 }, { "epoch": 0.83, "learning_rate": 5.182640144665462e-07, "loss": 0.8834, "step": 6912 }, { "epoch": 0.83, "learning_rate": 5.179023508137433e-07, "loss": 0.0006, "step": 6913 }, { "epoch": 0.83, "learning_rate": 5.175406871609403e-07, "loss": 0.0003, "step": 6914 }, { "epoch": 0.83, "learning_rate": 5.171790235081374e-07, "loss": 0.0009, "step": 6915 }, { "epoch": 0.83, "learning_rate": 5.168173598553346e-07, "loss": 0.0006, "step": 6916 }, { "epoch": 0.83, "learning_rate": 5.164556962025317e-07, "loss": 0.0905, "step": 6917 }, { "epoch": 0.83, "learning_rate": 5.160940325497288e-07, "loss": 0.0003, "step": 6918 }, { "epoch": 0.83, "learning_rate": 5.157323688969259e-07, "loss": 0.0337, "step": 6919 }, { "epoch": 0.83, "learning_rate": 5.15370705244123e-07, "loss": 0.0004, "step": 6920 }, { "epoch": 0.83, "learning_rate": 5.150090415913201e-07, "loss": 0.0004, "step": 6921 }, { "epoch": 0.83, "learning_rate": 5.146473779385172e-07, "loss": 0.0002, "step": 6922 }, { "epoch": 0.83, "learning_rate": 5.142857142857143e-07, "loss": 0.0006, "step": 6923 }, { "epoch": 0.83, "learning_rate": 5.139240506329114e-07, "loss": 0.0007, "step": 6924 }, { "epoch": 0.83, "learning_rate": 5.135623869801086e-07, "loss": 0.0003, "step": 6925 }, { "epoch": 0.83, "learning_rate": 5.132007233273056e-07, "loss": 0.0007, "step": 6926 }, { "epoch": 0.83, "learning_rate": 5.128390596745027e-07, "loss": 0.0004, "step": 6927 }, { "epoch": 0.83, "learning_rate": 5.124773960216998e-07, "loss": 0.0004, "step": 6928 }, { "epoch": 0.83, "learning_rate": 5.12115732368897e-07, "loss": 0.0004, "step": 6929 }, { "epoch": 0.83, "learning_rate": 5.117540687160941e-07, "loss": 0.0004, "step": 6930 }, { "epoch": 0.83, "learning_rate": 5.113924050632912e-07, "loss": 0.0003, "step": 6931 }, { "epoch": 0.83, "learning_rate": 5.110307414104882e-07, "loss": 0.0003, "step": 6932 }, { "epoch": 0.83, "learning_rate": 5.106690777576854e-07, "loss": 0.5482, "step": 6933 }, { "epoch": 0.83, "learning_rate": 5.103074141048825e-07, "loss": 0.7974, "step": 6934 }, { "epoch": 0.83, "learning_rate": 5.099457504520796e-07, "loss": 0.0014, "step": 6935 }, { "epoch": 0.83, "learning_rate": 5.095840867992767e-07, "loss": 0.0003, "step": 6936 }, { "epoch": 0.83, "learning_rate": 5.092224231464739e-07, "loss": 0.0002, "step": 6937 }, { "epoch": 0.83, "learning_rate": 5.088607594936709e-07, "loss": 0.0004, "step": 6938 }, { "epoch": 0.83, "learning_rate": 5.08499095840868e-07, "loss": 0.0008, "step": 6939 }, { "epoch": 0.83, "learning_rate": 5.081374321880651e-07, "loss": 0.0004, "step": 6940 }, { "epoch": 0.83, "learning_rate": 5.077757685352623e-07, "loss": 0.0004, "step": 6941 }, { "epoch": 0.83, "learning_rate": 5.074141048824594e-07, "loss": 0.0006, "step": 6942 }, { "epoch": 0.83, "learning_rate": 5.070524412296565e-07, "loss": 0.0005, "step": 6943 }, { "epoch": 0.83, "learning_rate": 5.066907775768535e-07, "loss": 0.0003, "step": 6944 }, { "epoch": 0.83, "learning_rate": 5.063291139240507e-07, "loss": 0.975, "step": 6945 }, { "epoch": 0.83, "learning_rate": 5.059674502712478e-07, "loss": 0.0004, "step": 6946 }, { "epoch": 0.83, "learning_rate": 5.056057866184449e-07, "loss": 0.0005, "step": 6947 }, { "epoch": 0.83, "learning_rate": 5.05244122965642e-07, "loss": 0.0004, "step": 6948 }, { "epoch": 0.83, "learning_rate": 5.048824593128392e-07, "loss": 0.0003, "step": 6949 }, { "epoch": 0.83, "learning_rate": 5.045207956600362e-07, "loss": 0.0004, "step": 6950 }, { "epoch": 0.83, "learning_rate": 5.041591320072333e-07, "loss": 0.0005, "step": 6951 }, { "epoch": 0.83, "learning_rate": 5.037974683544304e-07, "loss": 0.4147, "step": 6952 }, { "epoch": 0.83, "learning_rate": 5.034358047016276e-07, "loss": 0.0005, "step": 6953 }, { "epoch": 0.83, "learning_rate": 5.030741410488247e-07, "loss": 0.0007, "step": 6954 }, { "epoch": 0.83, "learning_rate": 5.027124773960217e-07, "loss": 0.0004, "step": 6955 }, { "epoch": 0.83, "learning_rate": 5.023508137432188e-07, "loss": 0.0004, "step": 6956 }, { "epoch": 0.83, "learning_rate": 5.01989150090416e-07, "loss": 0.0004, "step": 6957 }, { "epoch": 0.83, "learning_rate": 5.016274864376131e-07, "loss": 0.3355, "step": 6958 }, { "epoch": 0.83, "learning_rate": 5.012658227848102e-07, "loss": 0.0003, "step": 6959 }, { "epoch": 0.83, "learning_rate": 5.009041591320072e-07, "loss": 0.0005, "step": 6960 }, { "epoch": 0.83, "learning_rate": 5.005424954792043e-07, "loss": 0.0005, "step": 6961 }, { "epoch": 0.83, "learning_rate": 5.001808318264015e-07, "loss": 0.0005, "step": 6962 }, { "epoch": 0.83, "learning_rate": 4.998191681735986e-07, "loss": 0.0003, "step": 6963 }, { "epoch": 0.83, "learning_rate": 4.994575045207957e-07, "loss": 0.0005, "step": 6964 }, { "epoch": 0.83, "learning_rate": 4.990958408679927e-07, "loss": 0.0005, "step": 6965 }, { "epoch": 0.83, "learning_rate": 4.987341772151899e-07, "loss": 0.0004, "step": 6966 }, { "epoch": 0.83, "learning_rate": 4.98372513562387e-07, "loss": 0.0006, "step": 6967 }, { "epoch": 0.83, "learning_rate": 4.980108499095841e-07, "loss": 0.0003, "step": 6968 }, { "epoch": 0.84, "learning_rate": 4.976491862567811e-07, "loss": 0.0004, "step": 6969 }, { "epoch": 0.84, "learning_rate": 4.972875226039783e-07, "loss": 0.0003, "step": 6970 }, { "epoch": 0.84, "learning_rate": 4.969258589511754e-07, "loss": 0.0907, "step": 6971 }, { "epoch": 0.84, "learning_rate": 4.965641952983725e-07, "loss": 0.6326, "step": 6972 }, { "epoch": 0.84, "learning_rate": 4.962025316455696e-07, "loss": 0.0004, "step": 6973 }, { "epoch": 0.84, "learning_rate": 4.958408679927667e-07, "loss": 0.0003, "step": 6974 }, { "epoch": 0.84, "learning_rate": 4.954792043399638e-07, "loss": 0.0006, "step": 6975 }, { "epoch": 0.84, "learning_rate": 4.951175406871609e-07, "loss": 0.0007, "step": 6976 }, { "epoch": 0.84, "learning_rate": 4.94755877034358e-07, "loss": 0.0005, "step": 6977 }, { "epoch": 0.84, "learning_rate": 4.943942133815552e-07, "loss": 0.0003, "step": 6978 }, { "epoch": 0.84, "learning_rate": 4.940325497287523e-07, "loss": 0.0002, "step": 6979 }, { "epoch": 0.84, "learning_rate": 4.936708860759493e-07, "loss": 0.0005, "step": 6980 }, { "epoch": 0.84, "learning_rate": 4.933092224231464e-07, "loss": 0.0005, "step": 6981 }, { "epoch": 0.84, "learning_rate": 4.929475587703435e-07, "loss": 0.0003, "step": 6982 }, { "epoch": 0.84, "learning_rate": 4.925858951175407e-07, "loss": 0.0004, "step": 6983 }, { "epoch": 0.84, "learning_rate": 4.922242314647378e-07, "loss": 0.0002, "step": 6984 }, { "epoch": 0.84, "learning_rate": 4.918625678119349e-07, "loss": 0.0004, "step": 6985 }, { "epoch": 0.84, "learning_rate": 4.915009041591319e-07, "loss": 0.0003, "step": 6986 }, { "epoch": 0.84, "learning_rate": 4.911392405063291e-07, "loss": 0.0004, "step": 6987 }, { "epoch": 0.84, "learning_rate": 4.907775768535262e-07, "loss": 0.0007, "step": 6988 }, { "epoch": 0.84, "learning_rate": 4.904159132007233e-07, "loss": 0.0004, "step": 6989 }, { "epoch": 0.84, "learning_rate": 4.900542495479204e-07, "loss": 0.0006, "step": 6990 }, { "epoch": 0.84, "learning_rate": 4.896925858951176e-07, "loss": 0.0004, "step": 6991 }, { "epoch": 0.84, "learning_rate": 4.893309222423146e-07, "loss": 0.0004, "step": 6992 }, { "epoch": 0.84, "learning_rate": 4.889692585895117e-07, "loss": 0.0002, "step": 6993 }, { "epoch": 0.84, "learning_rate": 4.886075949367088e-07, "loss": 0.0021, "step": 6994 }, { "epoch": 0.84, "learning_rate": 4.88245931283906e-07, "loss": 0.0069, "step": 6995 }, { "epoch": 0.84, "learning_rate": 4.878842676311031e-07, "loss": 0.0004, "step": 6996 }, { "epoch": 0.84, "learning_rate": 4.875226039783002e-07, "loss": 0.0004, "step": 6997 }, { "epoch": 0.84, "learning_rate": 4.871609403254972e-07, "loss": 0.0013, "step": 6998 }, { "epoch": 0.84, "learning_rate": 4.867992766726944e-07, "loss": 0.0006, "step": 6999 }, { "epoch": 0.84, "learning_rate": 4.864376130198915e-07, "loss": 0.0003, "step": 7000 }, { "epoch": 0.84, "learning_rate": 4.860759493670886e-07, "loss": 0.0003, "step": 7001 }, { "epoch": 0.84, "learning_rate": 4.857142857142857e-07, "loss": 0.0004, "step": 7002 }, { "epoch": 0.84, "learning_rate": 4.853526220614829e-07, "loss": 0.0003, "step": 7003 }, { "epoch": 0.84, "learning_rate": 4.849909584086799e-07, "loss": 0.0006, "step": 7004 }, { "epoch": 0.84, "learning_rate": 4.84629294755877e-07, "loss": 0.0008, "step": 7005 }, { "epoch": 0.84, "learning_rate": 4.842676311030741e-07, "loss": 0.0003, "step": 7006 }, { "epoch": 0.84, "learning_rate": 4.839059674502713e-07, "loss": 0.0005, "step": 7007 }, { "epoch": 0.84, "learning_rate": 4.835443037974684e-07, "loss": 0.0005, "step": 7008 }, { "epoch": 0.84, "learning_rate": 4.831826401446655e-07, "loss": 0.0004, "step": 7009 }, { "epoch": 0.84, "learning_rate": 4.828209764918625e-07, "loss": 0.0003, "step": 7010 }, { "epoch": 0.84, "learning_rate": 4.824593128390597e-07, "loss": 0.0003, "step": 7011 }, { "epoch": 0.84, "learning_rate": 4.820976491862568e-07, "loss": 0.6548, "step": 7012 }, { "epoch": 0.84, "learning_rate": 4.817359855334539e-07, "loss": 0.0003, "step": 7013 }, { "epoch": 0.84, "learning_rate": 4.81374321880651e-07, "loss": 0.134, "step": 7014 }, { "epoch": 0.84, "learning_rate": 4.810126582278482e-07, "loss": 0.0003, "step": 7015 }, { "epoch": 0.84, "learning_rate": 4.806509945750452e-07, "loss": 0.0011, "step": 7016 }, { "epoch": 0.84, "learning_rate": 4.802893309222423e-07, "loss": 0.0002, "step": 7017 }, { "epoch": 0.84, "learning_rate": 4.799276672694394e-07, "loss": 0.0003, "step": 7018 }, { "epoch": 0.84, "learning_rate": 4.795660036166365e-07, "loss": 0.0002, "step": 7019 }, { "epoch": 0.84, "learning_rate": 4.792043399638337e-07, "loss": 0.0003, "step": 7020 }, { "epoch": 0.84, "learning_rate": 4.788426763110308e-07, "loss": 0.0007, "step": 7021 }, { "epoch": 0.84, "learning_rate": 4.784810126582278e-07, "loss": 0.0004, "step": 7022 }, { "epoch": 0.84, "learning_rate": 4.781193490054249e-07, "loss": 0.0004, "step": 7023 }, { "epoch": 0.84, "learning_rate": 4.777576853526221e-07, "loss": 0.0005, "step": 7024 }, { "epoch": 0.84, "learning_rate": 4.773960216998192e-07, "loss": 0.0003, "step": 7025 }, { "epoch": 0.84, "learning_rate": 4.770343580470163e-07, "loss": 0.0002, "step": 7026 }, { "epoch": 0.84, "learning_rate": 4.7667269439421345e-07, "loss": 0.0003, "step": 7027 }, { "epoch": 0.84, "learning_rate": 4.7631103074141047e-07, "loss": 0.0002, "step": 7028 }, { "epoch": 0.84, "learning_rate": 4.759493670886076e-07, "loss": 0.0004, "step": 7029 }, { "epoch": 0.84, "learning_rate": 4.755877034358047e-07, "loss": 0.0816, "step": 7030 }, { "epoch": 0.84, "learning_rate": 4.7522603978300184e-07, "loss": 0.0003, "step": 7031 }, { "epoch": 0.84, "learning_rate": 4.7486437613019897e-07, "loss": 0.0001, "step": 7032 }, { "epoch": 0.84, "learning_rate": 4.745027124773961e-07, "loss": 0.0002, "step": 7033 }, { "epoch": 0.84, "learning_rate": 4.741410488245931e-07, "loss": 0.0002, "step": 7034 }, { "epoch": 0.84, "learning_rate": 4.7377938517179024e-07, "loss": 0.0003, "step": 7035 }, { "epoch": 0.84, "learning_rate": 4.7341772151898736e-07, "loss": 0.0005, "step": 7036 }, { "epoch": 0.84, "learning_rate": 4.730560578661845e-07, "loss": 0.0004, "step": 7037 }, { "epoch": 0.84, "learning_rate": 4.726943942133816e-07, "loss": 0.0003, "step": 7038 }, { "epoch": 0.84, "learning_rate": 4.7233273056057874e-07, "loss": 0.0003, "step": 7039 }, { "epoch": 0.84, "learning_rate": 4.7197106690777576e-07, "loss": 0.0003, "step": 7040 }, { "epoch": 0.84, "learning_rate": 4.716094032549729e-07, "loss": 0.0002, "step": 7041 }, { "epoch": 0.84, "learning_rate": 4.7124773960217e-07, "loss": 0.0002, "step": 7042 }, { "epoch": 0.84, "learning_rate": 4.7088607594936713e-07, "loss": 0.9512, "step": 7043 }, { "epoch": 0.84, "learning_rate": 4.7052441229656426e-07, "loss": 0.0003, "step": 7044 }, { "epoch": 0.84, "learning_rate": 4.7016274864376133e-07, "loss": 0.0004, "step": 7045 }, { "epoch": 0.84, "learning_rate": 4.698010849909584e-07, "loss": 0.068, "step": 7046 }, { "epoch": 0.84, "learning_rate": 4.6943942133815553e-07, "loss": 0.0002, "step": 7047 }, { "epoch": 0.84, "learning_rate": 4.6907775768535265e-07, "loss": 0.0004, "step": 7048 }, { "epoch": 0.84, "learning_rate": 4.687160940325498e-07, "loss": 0.0004, "step": 7049 }, { "epoch": 0.84, "learning_rate": 4.6835443037974685e-07, "loss": 0.0002, "step": 7050 }, { "epoch": 0.84, "learning_rate": 4.67992766726944e-07, "loss": 0.506, "step": 7051 }, { "epoch": 0.85, "learning_rate": 4.6763110307414105e-07, "loss": 0.0003, "step": 7052 }, { "epoch": 0.85, "learning_rate": 4.672694394213382e-07, "loss": 0.0005, "step": 7053 }, { "epoch": 0.85, "learning_rate": 4.6690777576853525e-07, "loss": 0.0004, "step": 7054 }, { "epoch": 0.85, "learning_rate": 4.6654611211573237e-07, "loss": 0.0002, "step": 7055 }, { "epoch": 0.85, "learning_rate": 4.661844484629295e-07, "loss": 0.0003, "step": 7056 }, { "epoch": 0.85, "learning_rate": 4.658227848101266e-07, "loss": 0.0013, "step": 7057 }, { "epoch": 0.85, "learning_rate": 4.654611211573237e-07, "loss": 0.0004, "step": 7058 }, { "epoch": 0.85, "learning_rate": 4.6509945750452077e-07, "loss": 0.0003, "step": 7059 }, { "epoch": 0.85, "learning_rate": 4.647377938517179e-07, "loss": 0.0004, "step": 7060 }, { "epoch": 0.85, "learning_rate": 4.64376130198915e-07, "loss": 0.0004, "step": 7061 }, { "epoch": 0.85, "learning_rate": 4.6401446654611214e-07, "loss": 0.0003, "step": 7062 }, { "epoch": 0.85, "learning_rate": 4.6365280289330927e-07, "loss": 0.0004, "step": 7063 }, { "epoch": 0.85, "learning_rate": 4.632911392405063e-07, "loss": 0.0941, "step": 7064 }, { "epoch": 0.85, "learning_rate": 4.629294755877034e-07, "loss": 0.0853, "step": 7065 }, { "epoch": 0.85, "learning_rate": 4.6256781193490054e-07, "loss": 0.0005, "step": 7066 }, { "epoch": 0.85, "learning_rate": 4.6220614828209766e-07, "loss": 0.0003, "step": 7067 }, { "epoch": 0.85, "learning_rate": 4.618444846292948e-07, "loss": 0.0004, "step": 7068 }, { "epoch": 0.85, "learning_rate": 4.614828209764919e-07, "loss": 0.0099, "step": 7069 }, { "epoch": 0.85, "learning_rate": 4.6112115732368893e-07, "loss": 0.7961, "step": 7070 }, { "epoch": 0.85, "learning_rate": 4.6075949367088606e-07, "loss": 0.0004, "step": 7071 }, { "epoch": 0.85, "learning_rate": 4.603978300180832e-07, "loss": 0.0005, "step": 7072 }, { "epoch": 0.85, "learning_rate": 4.600361663652803e-07, "loss": 0.0003, "step": 7073 }, { "epoch": 0.85, "learning_rate": 4.5967450271247743e-07, "loss": 0.0009, "step": 7074 }, { "epoch": 0.85, "learning_rate": 4.5931283905967456e-07, "loss": 0.0004, "step": 7075 }, { "epoch": 0.85, "learning_rate": 4.589511754068716e-07, "loss": 0.0005, "step": 7076 }, { "epoch": 0.85, "learning_rate": 4.585895117540687e-07, "loss": 0.9811, "step": 7077 }, { "epoch": 0.85, "learning_rate": 4.5822784810126583e-07, "loss": 0.0006, "step": 7078 }, { "epoch": 0.85, "learning_rate": 4.5786618444846295e-07, "loss": 0.0003, "step": 7079 }, { "epoch": 0.85, "learning_rate": 4.575045207956601e-07, "loss": 0.0003, "step": 7080 }, { "epoch": 0.85, "learning_rate": 4.571428571428572e-07, "loss": 0.0003, "step": 7081 }, { "epoch": 0.85, "learning_rate": 4.567811934900542e-07, "loss": 0.0007, "step": 7082 }, { "epoch": 0.85, "learning_rate": 4.5641952983725135e-07, "loss": 0.0003, "step": 7083 }, { "epoch": 0.85, "learning_rate": 4.560578661844485e-07, "loss": 0.0009, "step": 7084 }, { "epoch": 0.85, "learning_rate": 4.556962025316456e-07, "loss": 0.0008, "step": 7085 }, { "epoch": 0.85, "learning_rate": 4.553345388788427e-07, "loss": 0.0007, "step": 7086 }, { "epoch": 0.85, "learning_rate": 4.5497287522603985e-07, "loss": 0.0003, "step": 7087 }, { "epoch": 0.85, "learning_rate": 4.5461121157323687e-07, "loss": 0.4469, "step": 7088 }, { "epoch": 0.85, "learning_rate": 4.54249547920434e-07, "loss": 0.0003, "step": 7089 }, { "epoch": 0.85, "learning_rate": 4.538878842676311e-07, "loss": 0.9219, "step": 7090 }, { "epoch": 0.85, "learning_rate": 4.5352622061482825e-07, "loss": 0.0003, "step": 7091 }, { "epoch": 0.85, "learning_rate": 4.5316455696202537e-07, "loss": 0.0004, "step": 7092 }, { "epoch": 0.85, "learning_rate": 4.528028933092224e-07, "loss": 0.0007, "step": 7093 }, { "epoch": 0.85, "learning_rate": 4.524412296564195e-07, "loss": 0.0023, "step": 7094 }, { "epoch": 0.85, "learning_rate": 4.5207956600361664e-07, "loss": 0.0004, "step": 7095 }, { "epoch": 0.85, "learning_rate": 4.5171790235081377e-07, "loss": 0.0003, "step": 7096 }, { "epoch": 0.85, "learning_rate": 4.513562386980109e-07, "loss": 0.0005, "step": 7097 }, { "epoch": 0.85, "learning_rate": 4.50994575045208e-07, "loss": 0.0003, "step": 7098 }, { "epoch": 0.85, "learning_rate": 4.5063291139240504e-07, "loss": 0.0003, "step": 7099 }, { "epoch": 0.85, "learning_rate": 4.5027124773960216e-07, "loss": 0.5339, "step": 7100 }, { "epoch": 0.85, "learning_rate": 4.499095840867993e-07, "loss": 0.8026, "step": 7101 }, { "epoch": 0.85, "learning_rate": 4.495479204339964e-07, "loss": 0.0004, "step": 7102 }, { "epoch": 0.85, "learning_rate": 4.4918625678119354e-07, "loss": 0.0002, "step": 7103 }, { "epoch": 0.85, "learning_rate": 4.488245931283906e-07, "loss": 0.0004, "step": 7104 }, { "epoch": 0.85, "learning_rate": 4.484629294755877e-07, "loss": 0.0004, "step": 7105 }, { "epoch": 0.85, "learning_rate": 4.481012658227848e-07, "loss": 0.0011, "step": 7106 }, { "epoch": 0.85, "learning_rate": 4.4773960216998193e-07, "loss": 0.0004, "step": 7107 }, { "epoch": 0.85, "learning_rate": 4.4737793851717906e-07, "loss": 0.0003, "step": 7108 }, { "epoch": 0.85, "learning_rate": 4.4701627486437613e-07, "loss": 0.0003, "step": 7109 }, { "epoch": 0.85, "learning_rate": 4.4665461121157326e-07, "loss": 0.0001, "step": 7110 }, { "epoch": 0.85, "learning_rate": 4.4629294755877033e-07, "loss": 0.0003, "step": 7111 }, { "epoch": 0.85, "learning_rate": 4.4593128390596745e-07, "loss": 0.0003, "step": 7112 }, { "epoch": 0.85, "learning_rate": 4.455696202531646e-07, "loss": 0.0001, "step": 7113 }, { "epoch": 0.85, "learning_rate": 4.4520795660036165e-07, "loss": 0.0006, "step": 7114 }, { "epoch": 0.85, "learning_rate": 4.448462929475588e-07, "loss": 0.0003, "step": 7115 }, { "epoch": 0.85, "learning_rate": 4.444846292947559e-07, "loss": 0.0003, "step": 7116 }, { "epoch": 0.85, "learning_rate": 4.44122965641953e-07, "loss": 0.0003, "step": 7117 }, { "epoch": 0.85, "learning_rate": 4.437613019891501e-07, "loss": 0.0003, "step": 7118 }, { "epoch": 0.85, "learning_rate": 4.4339963833634717e-07, "loss": 0.0002, "step": 7119 }, { "epoch": 0.85, "learning_rate": 4.430379746835443e-07, "loss": 0.0002, "step": 7120 }, { "epoch": 0.85, "learning_rate": 4.426763110307414e-07, "loss": 0.0002, "step": 7121 }, { "epoch": 0.85, "learning_rate": 4.4231464737793855e-07, "loss": 0.0002, "step": 7122 }, { "epoch": 0.85, "learning_rate": 4.419529837251356e-07, "loss": 0.0005, "step": 7123 }, { "epoch": 0.85, "learning_rate": 4.415913200723327e-07, "loss": 0.0003, "step": 7124 }, { "epoch": 0.85, "learning_rate": 4.412296564195298e-07, "loss": 0.0003, "step": 7125 }, { "epoch": 0.85, "learning_rate": 4.4086799276672694e-07, "loss": 0.0002, "step": 7126 }, { "epoch": 0.85, "learning_rate": 4.4050632911392407e-07, "loss": 0.0003, "step": 7127 }, { "epoch": 0.85, "learning_rate": 4.401446654611212e-07, "loss": 0.0003, "step": 7128 }, { "epoch": 0.85, "learning_rate": 4.397830018083182e-07, "loss": 0.0002, "step": 7129 }, { "epoch": 0.85, "learning_rate": 4.3942133815551534e-07, "loss": 0.0003, "step": 7130 }, { "epoch": 0.85, "learning_rate": 4.3905967450271246e-07, "loss": 0.0004, "step": 7131 }, { "epoch": 0.85, "learning_rate": 4.386980108499096e-07, "loss": 0.0002, "step": 7132 }, { "epoch": 0.85, "learning_rate": 4.383363471971067e-07, "loss": 0.0002, "step": 7133 }, { "epoch": 0.85, "learning_rate": 4.3797468354430384e-07, "loss": 0.0004, "step": 7134 }, { "epoch": 0.86, "learning_rate": 4.3761301989150086e-07, "loss": 0.0003, "step": 7135 }, { "epoch": 0.86, "learning_rate": 4.37251356238698e-07, "loss": 0.0003, "step": 7136 }, { "epoch": 0.86, "learning_rate": 4.368896925858951e-07, "loss": 0.0003, "step": 7137 }, { "epoch": 0.86, "learning_rate": 4.3652802893309223e-07, "loss": 0.001, "step": 7138 }, { "epoch": 0.86, "learning_rate": 4.3616636528028936e-07, "loss": 0.0163, "step": 7139 }, { "epoch": 0.86, "learning_rate": 4.358047016274865e-07, "loss": 0.0004, "step": 7140 }, { "epoch": 0.86, "learning_rate": 4.354430379746835e-07, "loss": 0.0003, "step": 7141 }, { "epoch": 0.86, "learning_rate": 4.3508137432188063e-07, "loss": 0.0022, "step": 7142 }, { "epoch": 0.86, "learning_rate": 4.3471971066907776e-07, "loss": 0.8695, "step": 7143 }, { "epoch": 0.86, "learning_rate": 4.343580470162749e-07, "loss": 0.4137, "step": 7144 }, { "epoch": 0.86, "learning_rate": 4.33996383363472e-07, "loss": 0.0003, "step": 7145 }, { "epoch": 0.86, "learning_rate": 4.3363471971066913e-07, "loss": 0.0002, "step": 7146 }, { "epoch": 0.86, "learning_rate": 4.3327305605786615e-07, "loss": 0.0009, "step": 7147 }, { "epoch": 0.86, "learning_rate": 4.329113924050633e-07, "loss": 0.0002, "step": 7148 }, { "epoch": 0.86, "learning_rate": 4.325497287522604e-07, "loss": 0.0003, "step": 7149 }, { "epoch": 0.86, "learning_rate": 4.321880650994575e-07, "loss": 0.0004, "step": 7150 }, { "epoch": 0.86, "learning_rate": 4.3182640144665465e-07, "loss": 0.0002, "step": 7151 }, { "epoch": 0.86, "learning_rate": 4.314647377938518e-07, "loss": 0.0002, "step": 7152 }, { "epoch": 0.86, "learning_rate": 4.311030741410488e-07, "loss": 0.0002, "step": 7153 }, { "epoch": 0.86, "learning_rate": 4.307414104882459e-07, "loss": 0.0003, "step": 7154 }, { "epoch": 0.86, "learning_rate": 4.3037974683544305e-07, "loss": 0.0003, "step": 7155 }, { "epoch": 0.86, "learning_rate": 4.3001808318264017e-07, "loss": 0.0004, "step": 7156 }, { "epoch": 0.86, "learning_rate": 4.296564195298373e-07, "loss": 0.0002, "step": 7157 }, { "epoch": 0.86, "learning_rate": 4.292947558770344e-07, "loss": 0.0002, "step": 7158 }, { "epoch": 0.86, "learning_rate": 4.2893309222423144e-07, "loss": 0.0003, "step": 7159 }, { "epoch": 0.86, "learning_rate": 4.2857142857142857e-07, "loss": 0.0003, "step": 7160 }, { "epoch": 0.86, "learning_rate": 4.282097649186257e-07, "loss": 0.0003, "step": 7161 }, { "epoch": 0.86, "learning_rate": 4.278481012658228e-07, "loss": 0.0003, "step": 7162 }, { "epoch": 0.86, "learning_rate": 4.2748643761301994e-07, "loss": 0.0002, "step": 7163 }, { "epoch": 0.86, "learning_rate": 4.27124773960217e-07, "loss": 0.3414, "step": 7164 }, { "epoch": 0.86, "learning_rate": 4.267631103074141e-07, "loss": 0.0003, "step": 7165 }, { "epoch": 0.86, "learning_rate": 4.264014466546112e-07, "loss": 0.0002, "step": 7166 }, { "epoch": 0.86, "learning_rate": 4.2603978300180834e-07, "loss": 0.5282, "step": 7167 }, { "epoch": 0.86, "learning_rate": 4.2567811934900546e-07, "loss": 0.0002, "step": 7168 }, { "epoch": 0.86, "learning_rate": 4.2531645569620254e-07, "loss": 0.001, "step": 7169 }, { "epoch": 0.86, "learning_rate": 4.2495479204339966e-07, "loss": 0.9898, "step": 7170 }, { "epoch": 0.86, "learning_rate": 4.2459312839059673e-07, "loss": 0.0002, "step": 7171 }, { "epoch": 0.86, "learning_rate": 4.2423146473779386e-07, "loss": 0.0043, "step": 7172 }, { "epoch": 0.86, "learning_rate": 4.23869801084991e-07, "loss": 0.0003, "step": 7173 }, { "epoch": 0.86, "learning_rate": 4.2350813743218806e-07, "loss": 0.0002, "step": 7174 }, { "epoch": 0.86, "learning_rate": 4.231464737793852e-07, "loss": 0.0002, "step": 7175 }, { "epoch": 0.86, "learning_rate": 4.227848101265823e-07, "loss": 0.0003, "step": 7176 }, { "epoch": 0.86, "learning_rate": 4.224231464737794e-07, "loss": 0.0002, "step": 7177 }, { "epoch": 0.86, "learning_rate": 4.220614828209765e-07, "loss": 0.0003, "step": 7178 }, { "epoch": 0.86, "learning_rate": 4.216998191681736e-07, "loss": 0.0003, "step": 7179 }, { "epoch": 0.86, "learning_rate": 4.213381555153707e-07, "loss": 0.0002, "step": 7180 }, { "epoch": 0.86, "learning_rate": 4.2097649186256783e-07, "loss": 0.0003, "step": 7181 }, { "epoch": 0.86, "learning_rate": 4.2061482820976495e-07, "loss": 0.0003, "step": 7182 }, { "epoch": 0.86, "learning_rate": 4.20253164556962e-07, "loss": 0.0003, "step": 7183 }, { "epoch": 0.86, "learning_rate": 4.198915009041591e-07, "loss": 0.0002, "step": 7184 }, { "epoch": 0.86, "learning_rate": 4.195298372513562e-07, "loss": 0.0002, "step": 7185 }, { "epoch": 0.86, "learning_rate": 4.1916817359855335e-07, "loss": 0.0002, "step": 7186 }, { "epoch": 0.86, "learning_rate": 4.1880650994575047e-07, "loss": 0.0003, "step": 7187 }, { "epoch": 0.86, "learning_rate": 4.184448462929476e-07, "loss": 0.0002, "step": 7188 }, { "epoch": 0.86, "learning_rate": 4.180831826401446e-07, "loss": 0.0002, "step": 7189 }, { "epoch": 0.86, "learning_rate": 4.1772151898734174e-07, "loss": 0.0002, "step": 7190 }, { "epoch": 0.86, "learning_rate": 4.1735985533453887e-07, "loss": 0.0005, "step": 7191 }, { "epoch": 0.86, "learning_rate": 4.16998191681736e-07, "loss": 0.0047, "step": 7192 }, { "epoch": 0.86, "learning_rate": 4.166365280289331e-07, "loss": 0.0003, "step": 7193 }, { "epoch": 0.86, "learning_rate": 4.1627486437613024e-07, "loss": 0.0003, "step": 7194 }, { "epoch": 0.86, "learning_rate": 4.1591320072332726e-07, "loss": 1.0028, "step": 7195 }, { "epoch": 0.86, "learning_rate": 4.155515370705244e-07, "loss": 0.0003, "step": 7196 }, { "epoch": 0.86, "learning_rate": 4.151898734177215e-07, "loss": 0.0003, "step": 7197 }, { "epoch": 0.86, "learning_rate": 4.1482820976491864e-07, "loss": 0.0002, "step": 7198 }, { "epoch": 0.86, "learning_rate": 4.1446654611211576e-07, "loss": 0.8207, "step": 7199 }, { "epoch": 0.86, "learning_rate": 4.141048824593129e-07, "loss": 0.0003, "step": 7200 }, { "epoch": 0.86, "learning_rate": 4.137432188065099e-07, "loss": 0.0003, "step": 7201 }, { "epoch": 0.86, "learning_rate": 4.1338155515370703e-07, "loss": 0.0003, "step": 7202 }, { "epoch": 0.86, "learning_rate": 4.1301989150090416e-07, "loss": 0.0003, "step": 7203 }, { "epoch": 0.86, "learning_rate": 4.126582278481013e-07, "loss": 0.0002, "step": 7204 }, { "epoch": 0.86, "learning_rate": 4.122965641952984e-07, "loss": 0.0049, "step": 7205 }, { "epoch": 0.86, "learning_rate": 4.1193490054249554e-07, "loss": 0.0003, "step": 7206 }, { "epoch": 0.86, "learning_rate": 4.1157323688969256e-07, "loss": 0.0003, "step": 7207 }, { "epoch": 0.86, "learning_rate": 4.112115732368897e-07, "loss": 0.0004, "step": 7208 }, { "epoch": 0.86, "learning_rate": 4.108499095840868e-07, "loss": 0.5517, "step": 7209 }, { "epoch": 0.86, "learning_rate": 4.1048824593128393e-07, "loss": 0.0002, "step": 7210 }, { "epoch": 0.86, "learning_rate": 4.1012658227848106e-07, "loss": 0.0002, "step": 7211 }, { "epoch": 0.86, "learning_rate": 4.097649186256782e-07, "loss": 0.0004, "step": 7212 }, { "epoch": 0.86, "learning_rate": 4.094032549728752e-07, "loss": 0.4276, "step": 7213 }, { "epoch": 0.86, "learning_rate": 4.090415913200723e-07, "loss": 0.0007, "step": 7214 }, { "epoch": 0.86, "learning_rate": 4.0867992766726945e-07, "loss": 0.0003, "step": 7215 }, { "epoch": 0.86, "learning_rate": 4.083182640144666e-07, "loss": 0.0002, "step": 7216 }, { "epoch": 0.86, "learning_rate": 4.079566003616637e-07, "loss": 0.0007, "step": 7217 }, { "epoch": 0.86, "learning_rate": 4.0759493670886083e-07, "loss": 0.0003, "step": 7218 }, { "epoch": 0.87, "learning_rate": 4.0723327305605785e-07, "loss": 0.0002, "step": 7219 }, { "epoch": 0.87, "learning_rate": 4.0687160940325497e-07, "loss": 0.0003, "step": 7220 }, { "epoch": 0.87, "learning_rate": 4.065099457504521e-07, "loss": 0.0002, "step": 7221 }, { "epoch": 0.87, "learning_rate": 4.061482820976492e-07, "loss": 0.0003, "step": 7222 }, { "epoch": 0.87, "learning_rate": 4.0578661844484635e-07, "loss": 0.0003, "step": 7223 }, { "epoch": 0.87, "learning_rate": 4.054249547920434e-07, "loss": 0.0002, "step": 7224 }, { "epoch": 0.87, "learning_rate": 4.050632911392405e-07, "loss": 0.0006, "step": 7225 }, { "epoch": 0.87, "learning_rate": 4.047016274864376e-07, "loss": 0.1986, "step": 7226 }, { "epoch": 0.87, "learning_rate": 4.0433996383363474e-07, "loss": 0.0004, "step": 7227 }, { "epoch": 0.87, "learning_rate": 4.0397830018083187e-07, "loss": 0.0005, "step": 7228 }, { "epoch": 0.87, "learning_rate": 4.0361663652802894e-07, "loss": 0.0002, "step": 7229 }, { "epoch": 0.87, "learning_rate": 4.0325497287522607e-07, "loss": 0.0003, "step": 7230 }, { "epoch": 0.87, "learning_rate": 4.0289330922242314e-07, "loss": 0.0003, "step": 7231 }, { "epoch": 0.87, "learning_rate": 4.0253164556962026e-07, "loss": 0.0002, "step": 7232 }, { "epoch": 0.87, "learning_rate": 4.021699819168174e-07, "loss": 0.0004, "step": 7233 }, { "epoch": 0.87, "learning_rate": 4.0180831826401446e-07, "loss": 0.0003, "step": 7234 }, { "epoch": 0.87, "learning_rate": 4.014466546112116e-07, "loss": 0.0004, "step": 7235 }, { "epoch": 0.87, "learning_rate": 4.010849909584087e-07, "loss": 0.0003, "step": 7236 }, { "epoch": 0.87, "learning_rate": 4.007233273056058e-07, "loss": 0.0002, "step": 7237 }, { "epoch": 0.87, "learning_rate": 4.003616636528029e-07, "loss": 0.0003, "step": 7238 }, { "epoch": 0.87, "learning_rate": 4e-07, "loss": 0.0002, "step": 7239 }, { "epoch": 0.87, "learning_rate": 3.996383363471971e-07, "loss": 0.0162, "step": 7240 }, { "epoch": 0.87, "learning_rate": 3.9927667269439423e-07, "loss": 0.0004, "step": 7241 }, { "epoch": 0.87, "learning_rate": 3.9891500904159136e-07, "loss": 0.0002, "step": 7242 }, { "epoch": 0.87, "learning_rate": 3.9855334538878843e-07, "loss": 0.0034, "step": 7243 }, { "epoch": 0.87, "learning_rate": 3.981916817359855e-07, "loss": 0.0003, "step": 7244 }, { "epoch": 0.87, "learning_rate": 3.9783001808318263e-07, "loss": 0.0006, "step": 7245 }, { "epoch": 0.87, "learning_rate": 3.9746835443037975e-07, "loss": 0.0002, "step": 7246 }, { "epoch": 0.87, "learning_rate": 3.971066907775769e-07, "loss": 0.0003, "step": 7247 }, { "epoch": 0.87, "learning_rate": 3.96745027124774e-07, "loss": 0.0003, "step": 7248 }, { "epoch": 0.87, "learning_rate": 3.96383363471971e-07, "loss": 0.9879, "step": 7249 }, { "epoch": 0.87, "learning_rate": 3.9602169981916815e-07, "loss": 0.0003, "step": 7250 }, { "epoch": 0.87, "learning_rate": 3.9566003616636527e-07, "loss": 0.0002, "step": 7251 }, { "epoch": 0.87, "learning_rate": 3.952983725135624e-07, "loss": 0.0003, "step": 7252 }, { "epoch": 0.87, "learning_rate": 3.949367088607595e-07, "loss": 0.0074, "step": 7253 }, { "epoch": 0.87, "learning_rate": 3.9457504520795665e-07, "loss": 0.0004, "step": 7254 }, { "epoch": 0.87, "learning_rate": 3.9421338155515367e-07, "loss": 0.8771, "step": 7255 }, { "epoch": 0.87, "learning_rate": 3.938517179023508e-07, "loss": 0.0004, "step": 7256 }, { "epoch": 0.87, "learning_rate": 3.934900542495479e-07, "loss": 0.0002, "step": 7257 }, { "epoch": 0.87, "learning_rate": 3.9312839059674504e-07, "loss": 0.0003, "step": 7258 }, { "epoch": 0.87, "learning_rate": 3.9276672694394217e-07, "loss": 0.0002, "step": 7259 }, { "epoch": 0.87, "learning_rate": 3.924050632911393e-07, "loss": 0.0004, "step": 7260 }, { "epoch": 0.87, "learning_rate": 3.920433996383363e-07, "loss": 0.0004, "step": 7261 }, { "epoch": 0.87, "learning_rate": 3.9168173598553344e-07, "loss": 0.0002, "step": 7262 }, { "epoch": 0.87, "learning_rate": 3.9132007233273057e-07, "loss": 0.0003, "step": 7263 }, { "epoch": 0.87, "learning_rate": 3.909584086799277e-07, "loss": 0.0002, "step": 7264 }, { "epoch": 0.87, "learning_rate": 3.905967450271248e-07, "loss": 0.0003, "step": 7265 }, { "epoch": 0.87, "learning_rate": 3.9023508137432194e-07, "loss": 0.0002, "step": 7266 }, { "epoch": 0.87, "learning_rate": 3.8987341772151896e-07, "loss": 0.0003, "step": 7267 }, { "epoch": 0.87, "learning_rate": 3.895117540687161e-07, "loss": 0.0002, "step": 7268 }, { "epoch": 0.87, "learning_rate": 3.891500904159132e-07, "loss": 0.0003, "step": 7269 }, { "epoch": 0.87, "learning_rate": 3.8878842676311034e-07, "loss": 0.0003, "step": 7270 }, { "epoch": 0.87, "learning_rate": 3.8842676311030746e-07, "loss": 0.0005, "step": 7271 }, { "epoch": 0.87, "learning_rate": 3.880650994575046e-07, "loss": 0.0002, "step": 7272 }, { "epoch": 0.87, "learning_rate": 3.877034358047016e-07, "loss": 0.0003, "step": 7273 }, { "epoch": 0.87, "learning_rate": 3.8734177215189873e-07, "loss": 0.0004, "step": 7274 }, { "epoch": 0.87, "learning_rate": 3.8698010849909586e-07, "loss": 0.0002, "step": 7275 }, { "epoch": 0.87, "learning_rate": 3.86618444846293e-07, "loss": 0.0003, "step": 7276 }, { "epoch": 0.87, "learning_rate": 3.862567811934901e-07, "loss": 0.0003, "step": 7277 }, { "epoch": 0.87, "learning_rate": 3.8589511754068723e-07, "loss": 0.0003, "step": 7278 }, { "epoch": 0.87, "learning_rate": 3.8553345388788425e-07, "loss": 0.0003, "step": 7279 }, { "epoch": 0.87, "learning_rate": 3.851717902350814e-07, "loss": 0.0005, "step": 7280 }, { "epoch": 0.87, "learning_rate": 3.848101265822785e-07, "loss": 0.0002, "step": 7281 }, { "epoch": 0.87, "learning_rate": 3.8444846292947563e-07, "loss": 0.8491, "step": 7282 }, { "epoch": 0.87, "learning_rate": 3.8408679927667275e-07, "loss": 0.0002, "step": 7283 }, { "epoch": 0.87, "learning_rate": 3.837251356238698e-07, "loss": 0.0003, "step": 7284 }, { "epoch": 0.87, "learning_rate": 3.833634719710669e-07, "loss": 0.8989, "step": 7285 }, { "epoch": 0.87, "learning_rate": 3.83001808318264e-07, "loss": 0.0003, "step": 7286 }, { "epoch": 0.87, "learning_rate": 3.8264014466546115e-07, "loss": 0.0003, "step": 7287 }, { "epoch": 0.87, "learning_rate": 3.822784810126583e-07, "loss": 0.0006, "step": 7288 }, { "epoch": 0.87, "learning_rate": 3.8191681735985535e-07, "loss": 0.0004, "step": 7289 }, { "epoch": 0.87, "learning_rate": 3.8155515370705247e-07, "loss": 0.0002, "step": 7290 }, { "epoch": 0.87, "learning_rate": 3.8119349005424954e-07, "loss": 0.6494, "step": 7291 }, { "epoch": 0.87, "learning_rate": 3.8083182640144667e-07, "loss": 0.0003, "step": 7292 }, { "epoch": 0.87, "learning_rate": 3.804701627486438e-07, "loss": 0.8374, "step": 7293 }, { "epoch": 0.87, "learning_rate": 3.8010849909584087e-07, "loss": 0.0004, "step": 7294 }, { "epoch": 0.87, "learning_rate": 3.79746835443038e-07, "loss": 0.0003, "step": 7295 }, { "epoch": 0.87, "learning_rate": 3.793851717902351e-07, "loss": 0.0003, "step": 7296 }, { "epoch": 0.87, "learning_rate": 3.790235081374322e-07, "loss": 0.0004, "step": 7297 }, { "epoch": 0.87, "learning_rate": 3.786618444846293e-07, "loss": 0.4508, "step": 7298 }, { "epoch": 0.87, "learning_rate": 3.783001808318264e-07, "loss": 0.155, "step": 7299 }, { "epoch": 0.87, "learning_rate": 3.779385171790235e-07, "loss": 0.0003, "step": 7300 }, { "epoch": 0.87, "learning_rate": 3.7757685352622064e-07, "loss": 0.0003, "step": 7301 }, { "epoch": 0.88, "learning_rate": 3.7721518987341776e-07, "loss": 0.0002, "step": 7302 }, { "epoch": 0.88, "learning_rate": 3.768535262206148e-07, "loss": 0.0002, "step": 7303 }, { "epoch": 0.88, "learning_rate": 3.764918625678119e-07, "loss": 0.0004, "step": 7304 }, { "epoch": 0.88, "learning_rate": 3.7613019891500903e-07, "loss": 0.0004, "step": 7305 }, { "epoch": 0.88, "learning_rate": 3.7576853526220616e-07, "loss": 0.7985, "step": 7306 }, { "epoch": 0.88, "learning_rate": 3.754068716094033e-07, "loss": 0.0006, "step": 7307 }, { "epoch": 0.88, "learning_rate": 3.750452079566004e-07, "loss": 0.0003, "step": 7308 }, { "epoch": 0.88, "learning_rate": 3.746835443037975e-07, "loss": 0.0004, "step": 7309 }, { "epoch": 0.88, "learning_rate": 3.7432188065099455e-07, "loss": 0.0003, "step": 7310 }, { "epoch": 0.88, "learning_rate": 3.739602169981917e-07, "loss": 0.0003, "step": 7311 }, { "epoch": 0.88, "learning_rate": 3.735985533453888e-07, "loss": 0.0004, "step": 7312 }, { "epoch": 0.88, "learning_rate": 3.732368896925859e-07, "loss": 0.0003, "step": 7313 }, { "epoch": 0.88, "learning_rate": 3.72875226039783e-07, "loss": 0.0008, "step": 7314 }, { "epoch": 0.88, "learning_rate": 3.7251356238698013e-07, "loss": 0.0003, "step": 7315 }, { "epoch": 0.88, "learning_rate": 3.721518987341772e-07, "loss": 0.0002, "step": 7316 }, { "epoch": 0.88, "learning_rate": 3.717902350813743e-07, "loss": 0.0004, "step": 7317 }, { "epoch": 0.88, "learning_rate": 3.7142857142857145e-07, "loss": 0.0003, "step": 7318 }, { "epoch": 0.88, "learning_rate": 3.710669077757685e-07, "loss": 0.0002, "step": 7319 }, { "epoch": 0.88, "learning_rate": 3.7070524412296565e-07, "loss": 0.0003, "step": 7320 }, { "epoch": 0.88, "learning_rate": 3.7034358047016277e-07, "loss": 0.0003, "step": 7321 }, { "epoch": 0.88, "learning_rate": 3.6998191681735984e-07, "loss": 0.0008, "step": 7322 }, { "epoch": 0.88, "learning_rate": 3.6962025316455697e-07, "loss": 0.0003, "step": 7323 }, { "epoch": 0.88, "learning_rate": 3.692585895117541e-07, "loss": 0.0002, "step": 7324 }, { "epoch": 0.88, "learning_rate": 3.6889692585895117e-07, "loss": 0.0001, "step": 7325 }, { "epoch": 0.88, "learning_rate": 3.685352622061483e-07, "loss": 0.0003, "step": 7326 }, { "epoch": 0.88, "learning_rate": 3.681735985533454e-07, "loss": 0.0003, "step": 7327 }, { "epoch": 0.88, "learning_rate": 3.678119349005425e-07, "loss": 0.0045, "step": 7328 }, { "epoch": 0.88, "learning_rate": 3.674502712477396e-07, "loss": 0.0003, "step": 7329 }, { "epoch": 0.88, "learning_rate": 3.6708860759493674e-07, "loss": 0.0004, "step": 7330 }, { "epoch": 0.88, "learning_rate": 3.667269439421338e-07, "loss": 0.0301, "step": 7331 }, { "epoch": 0.88, "learning_rate": 3.6636528028933094e-07, "loss": 0.1998, "step": 7332 }, { "epoch": 0.88, "learning_rate": 3.6600361663652806e-07, "loss": 0.0004, "step": 7333 }, { "epoch": 0.88, "learning_rate": 3.6564195298372514e-07, "loss": 0.0003, "step": 7334 }, { "epoch": 0.88, "learning_rate": 3.6528028933092226e-07, "loss": 0.0002, "step": 7335 }, { "epoch": 0.88, "learning_rate": 3.649186256781194e-07, "loss": 0.0002, "step": 7336 }, { "epoch": 0.88, "learning_rate": 3.6455696202531646e-07, "loss": 0.0004, "step": 7337 }, { "epoch": 0.88, "learning_rate": 3.641952983725136e-07, "loss": 0.0003, "step": 7338 }, { "epoch": 0.88, "learning_rate": 3.638336347197107e-07, "loss": 0.0006, "step": 7339 }, { "epoch": 0.88, "learning_rate": 3.634719710669078e-07, "loss": 0.0002, "step": 7340 }, { "epoch": 0.88, "learning_rate": 3.631103074141049e-07, "loss": 0.0002, "step": 7341 }, { "epoch": 0.88, "learning_rate": 3.6274864376130203e-07, "loss": 0.0003, "step": 7342 }, { "epoch": 0.88, "learning_rate": 3.623869801084991e-07, "loss": 0.0003, "step": 7343 }, { "epoch": 0.88, "learning_rate": 3.6202531645569623e-07, "loss": 0.0003, "step": 7344 }, { "epoch": 0.88, "learning_rate": 3.6166365280289336e-07, "loss": 0.0003, "step": 7345 }, { "epoch": 0.88, "learning_rate": 3.6130198915009043e-07, "loss": 0.0002, "step": 7346 }, { "epoch": 0.88, "learning_rate": 3.6094032549728755e-07, "loss": 0.0002, "step": 7347 }, { "epoch": 0.88, "learning_rate": 3.605786618444847e-07, "loss": 0.0002, "step": 7348 }, { "epoch": 0.88, "learning_rate": 3.6021699819168175e-07, "loss": 0.0003, "step": 7349 }, { "epoch": 0.88, "learning_rate": 3.598553345388789e-07, "loss": 0.0002, "step": 7350 }, { "epoch": 0.88, "learning_rate": 3.5949367088607595e-07, "loss": 0.0005, "step": 7351 }, { "epoch": 0.88, "learning_rate": 3.591320072332731e-07, "loss": 0.0003, "step": 7352 }, { "epoch": 0.88, "learning_rate": 3.5877034358047015e-07, "loss": 0.0003, "step": 7353 }, { "epoch": 0.88, "learning_rate": 3.5840867992766727e-07, "loss": 0.0003, "step": 7354 }, { "epoch": 0.88, "learning_rate": 3.580470162748644e-07, "loss": 0.0003, "step": 7355 }, { "epoch": 0.88, "learning_rate": 3.5768535262206147e-07, "loss": 0.0002, "step": 7356 }, { "epoch": 0.88, "learning_rate": 3.573236889692586e-07, "loss": 0.0004, "step": 7357 }, { "epoch": 0.88, "learning_rate": 3.5696202531645567e-07, "loss": 0.0002, "step": 7358 }, { "epoch": 0.88, "learning_rate": 3.566003616636528e-07, "loss": 0.0011, "step": 7359 }, { "epoch": 0.88, "learning_rate": 3.562386980108499e-07, "loss": 0.0003, "step": 7360 }, { "epoch": 0.88, "learning_rate": 3.55877034358047e-07, "loss": 0.0002, "step": 7361 }, { "epoch": 0.88, "learning_rate": 3.555153707052441e-07, "loss": 0.0001, "step": 7362 }, { "epoch": 0.88, "learning_rate": 3.5515370705244124e-07, "loss": 0.4055, "step": 7363 }, { "epoch": 0.88, "learning_rate": 3.547920433996383e-07, "loss": 0.0004, "step": 7364 }, { "epoch": 0.88, "learning_rate": 3.5443037974683544e-07, "loss": 0.0004, "step": 7365 }, { "epoch": 0.88, "learning_rate": 3.5406871609403256e-07, "loss": 0.0003, "step": 7366 }, { "epoch": 0.88, "learning_rate": 3.5370705244122964e-07, "loss": 0.0002, "step": 7367 }, { "epoch": 0.88, "learning_rate": 3.5334538878842676e-07, "loss": 0.0003, "step": 7368 }, { "epoch": 0.88, "learning_rate": 3.529837251356239e-07, "loss": 0.2835, "step": 7369 }, { "epoch": 0.88, "learning_rate": 3.5262206148282096e-07, "loss": 0.0002, "step": 7370 }, { "epoch": 0.88, "learning_rate": 3.522603978300181e-07, "loss": 0.0002, "step": 7371 }, { "epoch": 0.88, "learning_rate": 3.518987341772152e-07, "loss": 0.0006, "step": 7372 }, { "epoch": 0.88, "learning_rate": 3.515370705244123e-07, "loss": 0.0004, "step": 7373 }, { "epoch": 0.88, "learning_rate": 3.511754068716094e-07, "loss": 0.0002, "step": 7374 }, { "epoch": 0.88, "learning_rate": 3.5081374321880653e-07, "loss": 0.0004, "step": 7375 }, { "epoch": 0.88, "learning_rate": 3.504520795660036e-07, "loss": 0.0003, "step": 7376 }, { "epoch": 0.88, "learning_rate": 3.5009041591320073e-07, "loss": 0.0003, "step": 7377 }, { "epoch": 0.88, "learning_rate": 3.4972875226039785e-07, "loss": 0.0003, "step": 7378 }, { "epoch": 0.88, "learning_rate": 3.4936708860759493e-07, "loss": 0.0004, "step": 7379 }, { "epoch": 0.88, "learning_rate": 3.4900542495479205e-07, "loss": 0.0002, "step": 7380 }, { "epoch": 0.88, "learning_rate": 3.486437613019892e-07, "loss": 0.0002, "step": 7381 }, { "epoch": 0.88, "learning_rate": 3.4828209764918625e-07, "loss": 0.0002, "step": 7382 }, { "epoch": 0.88, "learning_rate": 3.479204339963834e-07, "loss": 0.0002, "step": 7383 }, { "epoch": 0.88, "learning_rate": 3.475587703435805e-07, "loss": 0.0003, "step": 7384 }, { "epoch": 0.88, "learning_rate": 3.4719710669077757e-07, "loss": 0.0003, "step": 7385 }, { "epoch": 0.89, "learning_rate": 3.468354430379747e-07, "loss": 0.0003, "step": 7386 }, { "epoch": 0.89, "learning_rate": 3.464737793851718e-07, "loss": 0.0004, "step": 7387 }, { "epoch": 0.89, "learning_rate": 3.461121157323689e-07, "loss": 0.6269, "step": 7388 }, { "epoch": 0.89, "learning_rate": 3.45750452079566e-07, "loss": 0.0003, "step": 7389 }, { "epoch": 0.89, "learning_rate": 3.4538878842676315e-07, "loss": 0.0003, "step": 7390 }, { "epoch": 0.89, "learning_rate": 3.450271247739602e-07, "loss": 0.0006, "step": 7391 }, { "epoch": 0.89, "learning_rate": 3.4466546112115734e-07, "loss": 0.0002, "step": 7392 }, { "epoch": 0.89, "learning_rate": 3.4430379746835447e-07, "loss": 0.0003, "step": 7393 }, { "epoch": 0.89, "learning_rate": 3.4394213381555154e-07, "loss": 0.0003, "step": 7394 }, { "epoch": 0.89, "learning_rate": 3.4358047016274867e-07, "loss": 0.0006, "step": 7395 }, { "epoch": 0.89, "learning_rate": 3.432188065099458e-07, "loss": 0.0002, "step": 7396 }, { "epoch": 0.89, "learning_rate": 3.4285714285714286e-07, "loss": 0.0002, "step": 7397 }, { "epoch": 0.89, "learning_rate": 3.4249547920434e-07, "loss": 0.001, "step": 7398 }, { "epoch": 0.89, "learning_rate": 3.421338155515371e-07, "loss": 0.0001, "step": 7399 }, { "epoch": 0.89, "learning_rate": 3.417721518987342e-07, "loss": 0.9631, "step": 7400 }, { "epoch": 0.89, "learning_rate": 3.414104882459313e-07, "loss": 0.0003, "step": 7401 }, { "epoch": 0.89, "learning_rate": 3.4104882459312844e-07, "loss": 0.8682, "step": 7402 }, { "epoch": 0.89, "learning_rate": 3.406871609403255e-07, "loss": 0.0017, "step": 7403 }, { "epoch": 0.89, "learning_rate": 3.4032549728752264e-07, "loss": 0.0002, "step": 7404 }, { "epoch": 0.89, "learning_rate": 3.3996383363471976e-07, "loss": 0.0004, "step": 7405 }, { "epoch": 0.89, "learning_rate": 3.3960216998191683e-07, "loss": 0.0003, "step": 7406 }, { "epoch": 0.89, "learning_rate": 3.3924050632911396e-07, "loss": 0.0239, "step": 7407 }, { "epoch": 0.89, "learning_rate": 3.3887884267631103e-07, "loss": 0.0004, "step": 7408 }, { "epoch": 0.89, "learning_rate": 3.3851717902350816e-07, "loss": 0.0002, "step": 7409 }, { "epoch": 0.89, "learning_rate": 3.381555153707053e-07, "loss": 0.0003, "step": 7410 }, { "epoch": 0.89, "learning_rate": 3.3779385171790235e-07, "loss": 0.0002, "step": 7411 }, { "epoch": 0.89, "learning_rate": 3.374321880650995e-07, "loss": 0.0003, "step": 7412 }, { "epoch": 0.89, "learning_rate": 3.3707052441229655e-07, "loss": 0.0002, "step": 7413 }, { "epoch": 0.89, "learning_rate": 3.367088607594937e-07, "loss": 0.0003, "step": 7414 }, { "epoch": 0.89, "learning_rate": 3.363471971066908e-07, "loss": 0.0002, "step": 7415 }, { "epoch": 0.89, "learning_rate": 3.359855334538879e-07, "loss": 0.0002, "step": 7416 }, { "epoch": 0.89, "learning_rate": 3.35623869801085e-07, "loss": 0.0002, "step": 7417 }, { "epoch": 0.89, "learning_rate": 3.3526220614828207e-07, "loss": 0.0002, "step": 7418 }, { "epoch": 0.89, "learning_rate": 3.349005424954792e-07, "loss": 0.0003, "step": 7419 }, { "epoch": 0.89, "learning_rate": 3.345388788426763e-07, "loss": 0.027, "step": 7420 }, { "epoch": 0.89, "learning_rate": 3.341772151898734e-07, "loss": 0.0002, "step": 7421 }, { "epoch": 0.89, "learning_rate": 3.338155515370705e-07, "loss": 0.6901, "step": 7422 }, { "epoch": 0.89, "learning_rate": 3.3345388788426765e-07, "loss": 0.0003, "step": 7423 }, { "epoch": 0.89, "learning_rate": 3.330922242314647e-07, "loss": 0.0002, "step": 7424 }, { "epoch": 0.89, "learning_rate": 3.3273056057866184e-07, "loss": 0.0004, "step": 7425 }, { "epoch": 0.89, "learning_rate": 3.3236889692585897e-07, "loss": 0.0003, "step": 7426 }, { "epoch": 0.89, "learning_rate": 3.3200723327305604e-07, "loss": 0.0002, "step": 7427 }, { "epoch": 0.89, "learning_rate": 3.3164556962025317e-07, "loss": 0.0003, "step": 7428 }, { "epoch": 0.89, "learning_rate": 3.312839059674503e-07, "loss": 0.0076, "step": 7429 }, { "epoch": 0.89, "learning_rate": 3.3092224231464736e-07, "loss": 0.0003, "step": 7430 }, { "epoch": 0.89, "learning_rate": 3.305605786618445e-07, "loss": 0.0003, "step": 7431 }, { "epoch": 0.89, "learning_rate": 3.301989150090416e-07, "loss": 0.0004, "step": 7432 }, { "epoch": 0.89, "learning_rate": 3.298372513562387e-07, "loss": 0.7009, "step": 7433 }, { "epoch": 0.89, "learning_rate": 3.294755877034358e-07, "loss": 0.0002, "step": 7434 }, { "epoch": 0.89, "learning_rate": 3.2911392405063294e-07, "loss": 0.0002, "step": 7435 }, { "epoch": 0.89, "learning_rate": 3.2875226039783e-07, "loss": 0.0001, "step": 7436 }, { "epoch": 0.89, "learning_rate": 3.2839059674502713e-07, "loss": 0.0027, "step": 7437 }, { "epoch": 0.89, "learning_rate": 3.2802893309222426e-07, "loss": 0.0002, "step": 7438 }, { "epoch": 0.89, "learning_rate": 3.2766726943942133e-07, "loss": 0.028, "step": 7439 }, { "epoch": 0.89, "learning_rate": 3.2730560578661846e-07, "loss": 0.0002, "step": 7440 }, { "epoch": 0.89, "learning_rate": 3.269439421338156e-07, "loss": 0.0007, "step": 7441 }, { "epoch": 0.89, "learning_rate": 3.2658227848101265e-07, "loss": 0.0003, "step": 7442 }, { "epoch": 0.89, "learning_rate": 3.262206148282098e-07, "loss": 0.0003, "step": 7443 }, { "epoch": 0.89, "learning_rate": 3.258589511754069e-07, "loss": 0.0004, "step": 7444 }, { "epoch": 0.89, "learning_rate": 3.25497287522604e-07, "loss": 0.0002, "step": 7445 }, { "epoch": 0.89, "learning_rate": 3.251356238698011e-07, "loss": 0.0003, "step": 7446 }, { "epoch": 0.89, "learning_rate": 3.2477396021699823e-07, "loss": 0.0283, "step": 7447 }, { "epoch": 0.89, "learning_rate": 3.244122965641953e-07, "loss": 0.0002, "step": 7448 }, { "epoch": 0.89, "learning_rate": 3.240506329113924e-07, "loss": 0.0003, "step": 7449 }, { "epoch": 0.89, "learning_rate": 3.2368896925858955e-07, "loss": 0.0003, "step": 7450 }, { "epoch": 0.89, "learning_rate": 3.233273056057866e-07, "loss": 1.077, "step": 7451 }, { "epoch": 0.89, "learning_rate": 3.2296564195298375e-07, "loss": 0.0004, "step": 7452 }, { "epoch": 0.89, "learning_rate": 3.226039783001809e-07, "loss": 0.0003, "step": 7453 }, { "epoch": 0.89, "learning_rate": 3.2224231464737795e-07, "loss": 0.0003, "step": 7454 }, { "epoch": 0.89, "learning_rate": 3.2188065099457507e-07, "loss": 0.0103, "step": 7455 }, { "epoch": 0.89, "learning_rate": 3.215189873417722e-07, "loss": 0.0004, "step": 7456 }, { "epoch": 0.89, "learning_rate": 3.2115732368896927e-07, "loss": 0.0003, "step": 7457 }, { "epoch": 0.89, "learning_rate": 3.207956600361664e-07, "loss": 0.0004, "step": 7458 }, { "epoch": 0.89, "learning_rate": 3.204339963833635e-07, "loss": 0.0002, "step": 7459 }, { "epoch": 0.89, "learning_rate": 3.200723327305606e-07, "loss": 0.0003, "step": 7460 }, { "epoch": 0.89, "learning_rate": 3.197106690777577e-07, "loss": 0.0003, "step": 7461 }, { "epoch": 0.89, "learning_rate": 3.1934900542495484e-07, "loss": 0.0002, "step": 7462 }, { "epoch": 0.89, "learning_rate": 3.189873417721519e-07, "loss": 0.0003, "step": 7463 }, { "epoch": 0.89, "learning_rate": 3.1862567811934904e-07, "loss": 0.5128, "step": 7464 }, { "epoch": 0.89, "learning_rate": 3.1826401446654617e-07, "loss": 0.0036, "step": 7465 }, { "epoch": 0.89, "learning_rate": 3.1790235081374324e-07, "loss": 0.0002, "step": 7466 }, { "epoch": 0.89, "learning_rate": 3.1754068716094036e-07, "loss": 0.0002, "step": 7467 }, { "epoch": 0.89, "learning_rate": 3.1717902350813744e-07, "loss": 0.0004, "step": 7468 }, { "epoch": 0.9, "learning_rate": 3.1681735985533456e-07, "loss": 0.0002, "step": 7469 }, { "epoch": 0.9, "learning_rate": 3.164556962025317e-07, "loss": 0.0003, "step": 7470 }, { "epoch": 0.9, "learning_rate": 3.1609403254972876e-07, "loss": 0.0003, "step": 7471 }, { "epoch": 0.9, "learning_rate": 3.157323688969259e-07, "loss": 0.6132, "step": 7472 }, { "epoch": 0.9, "learning_rate": 3.1537070524412296e-07, "loss": 0.0004, "step": 7473 }, { "epoch": 0.9, "learning_rate": 3.150090415913201e-07, "loss": 0.0003, "step": 7474 }, { "epoch": 0.9, "learning_rate": 3.146473779385172e-07, "loss": 0.001, "step": 7475 }, { "epoch": 0.9, "learning_rate": 3.142857142857143e-07, "loss": 0.0003, "step": 7476 }, { "epoch": 0.9, "learning_rate": 3.139240506329114e-07, "loss": 0.0002, "step": 7477 }, { "epoch": 0.9, "learning_rate": 3.135623869801085e-07, "loss": 0.0002, "step": 7478 }, { "epoch": 0.9, "learning_rate": 3.132007233273056e-07, "loss": 0.0003, "step": 7479 }, { "epoch": 0.9, "learning_rate": 3.128390596745027e-07, "loss": 0.0092, "step": 7480 }, { "epoch": 0.9, "learning_rate": 3.124773960216998e-07, "loss": 0.0055, "step": 7481 }, { "epoch": 0.9, "learning_rate": 3.121157323688969e-07, "loss": 0.0002, "step": 7482 }, { "epoch": 0.9, "learning_rate": 3.11754068716094e-07, "loss": 0.0003, "step": 7483 }, { "epoch": 0.9, "learning_rate": 3.113924050632911e-07, "loss": 0.0004, "step": 7484 }, { "epoch": 0.9, "learning_rate": 3.1103074141048825e-07, "loss": 0.0005, "step": 7485 }, { "epoch": 0.9, "learning_rate": 3.106690777576853e-07, "loss": 0.0111, "step": 7486 }, { "epoch": 0.9, "learning_rate": 3.1030741410488245e-07, "loss": 0.0003, "step": 7487 }, { "epoch": 0.9, "learning_rate": 3.0994575045207957e-07, "loss": 0.0008, "step": 7488 }, { "epoch": 0.9, "learning_rate": 3.0958408679927664e-07, "loss": 0.0002, "step": 7489 }, { "epoch": 0.9, "learning_rate": 3.0922242314647377e-07, "loss": 0.0002, "step": 7490 }, { "epoch": 0.9, "learning_rate": 3.088607594936709e-07, "loss": 0.0002, "step": 7491 }, { "epoch": 0.9, "learning_rate": 3.0849909584086797e-07, "loss": 0.0003, "step": 7492 }, { "epoch": 0.9, "learning_rate": 3.081374321880651e-07, "loss": 0.0536, "step": 7493 }, { "epoch": 0.9, "learning_rate": 3.077757685352622e-07, "loss": 0.2004, "step": 7494 }, { "epoch": 0.9, "learning_rate": 3.074141048824593e-07, "loss": 0.327, "step": 7495 }, { "epoch": 0.9, "learning_rate": 3.070524412296564e-07, "loss": 0.0003, "step": 7496 }, { "epoch": 0.9, "learning_rate": 3.0669077757685354e-07, "loss": 0.0002, "step": 7497 }, { "epoch": 0.9, "learning_rate": 3.063291139240506e-07, "loss": 0.0003, "step": 7498 }, { "epoch": 0.9, "learning_rate": 3.0596745027124774e-07, "loss": 0.0003, "step": 7499 }, { "epoch": 0.9, "learning_rate": 3.0560578661844486e-07, "loss": 0.0003, "step": 7500 }, { "epoch": 0.9, "learning_rate": 3.0524412296564193e-07, "loss": 0.0001, "step": 7501 }, { "epoch": 0.9, "learning_rate": 3.0488245931283906e-07, "loss": 0.0041, "step": 7502 }, { "epoch": 0.9, "learning_rate": 3.045207956600362e-07, "loss": 0.0037, "step": 7503 }, { "epoch": 0.9, "learning_rate": 3.0415913200723326e-07, "loss": 0.0002, "step": 7504 }, { "epoch": 0.9, "learning_rate": 3.037974683544304e-07, "loss": 0.0004, "step": 7505 }, { "epoch": 0.9, "learning_rate": 3.034358047016275e-07, "loss": 0.0009, "step": 7506 }, { "epoch": 0.9, "learning_rate": 3.030741410488246e-07, "loss": 0.0002, "step": 7507 }, { "epoch": 0.9, "learning_rate": 3.027124773960217e-07, "loss": 0.0004, "step": 7508 }, { "epoch": 0.9, "learning_rate": 3.0235081374321883e-07, "loss": 0.0003, "step": 7509 }, { "epoch": 0.9, "learning_rate": 3.019891500904159e-07, "loss": 0.0003, "step": 7510 }, { "epoch": 0.9, "learning_rate": 3.0162748643761303e-07, "loss": 0.0005, "step": 7511 }, { "epoch": 0.9, "learning_rate": 3.0126582278481015e-07, "loss": 0.0003, "step": 7512 }, { "epoch": 0.9, "learning_rate": 3.009041591320072e-07, "loss": 0.0003, "step": 7513 }, { "epoch": 0.9, "learning_rate": 3.0054249547920435e-07, "loss": 0.4219, "step": 7514 }, { "epoch": 0.9, "learning_rate": 3.001808318264015e-07, "loss": 0.0003, "step": 7515 }, { "epoch": 0.9, "learning_rate": 2.9981916817359855e-07, "loss": 0.0003, "step": 7516 }, { "epoch": 0.9, "learning_rate": 2.994575045207957e-07, "loss": 0.0004, "step": 7517 }, { "epoch": 0.9, "learning_rate": 2.990958408679928e-07, "loss": 0.0003, "step": 7518 }, { "epoch": 0.9, "learning_rate": 2.9873417721518987e-07, "loss": 0.9417, "step": 7519 }, { "epoch": 0.9, "learning_rate": 2.98372513562387e-07, "loss": 0.0003, "step": 7520 }, { "epoch": 0.9, "learning_rate": 2.980108499095841e-07, "loss": 0.0002, "step": 7521 }, { "epoch": 0.9, "learning_rate": 2.976491862567812e-07, "loss": 0.0003, "step": 7522 }, { "epoch": 0.9, "learning_rate": 2.972875226039783e-07, "loss": 0.0003, "step": 7523 }, { "epoch": 0.9, "learning_rate": 2.9692585895117545e-07, "loss": 0.0003, "step": 7524 }, { "epoch": 0.9, "learning_rate": 2.965641952983725e-07, "loss": 0.0003, "step": 7525 }, { "epoch": 0.9, "learning_rate": 2.9620253164556964e-07, "loss": 0.0002, "step": 7526 }, { "epoch": 0.9, "learning_rate": 2.9584086799276677e-07, "loss": 0.0003, "step": 7527 }, { "epoch": 0.9, "learning_rate": 2.9547920433996384e-07, "loss": 0.0002, "step": 7528 }, { "epoch": 0.9, "learning_rate": 2.9511754068716097e-07, "loss": 0.0002, "step": 7529 }, { "epoch": 0.9, "learning_rate": 2.947558770343581e-07, "loss": 0.0003, "step": 7530 }, { "epoch": 0.9, "learning_rate": 2.9439421338155516e-07, "loss": 0.0002, "step": 7531 }, { "epoch": 0.9, "learning_rate": 2.940325497287523e-07, "loss": 0.0003, "step": 7532 }, { "epoch": 0.9, "learning_rate": 2.9367088607594936e-07, "loss": 0.0002, "step": 7533 }, { "epoch": 0.9, "learning_rate": 2.933092224231465e-07, "loss": 0.0002, "step": 7534 }, { "epoch": 0.9, "learning_rate": 2.9294755877034356e-07, "loss": 0.0002, "step": 7535 }, { "epoch": 0.9, "learning_rate": 2.925858951175407e-07, "loss": 1.0077, "step": 7536 }, { "epoch": 0.9, "learning_rate": 2.922242314647378e-07, "loss": 0.0002, "step": 7537 }, { "epoch": 0.9, "learning_rate": 2.918625678119349e-07, "loss": 0.0003, "step": 7538 }, { "epoch": 0.9, "learning_rate": 2.91500904159132e-07, "loss": 0.0002, "step": 7539 }, { "epoch": 0.9, "learning_rate": 2.911392405063291e-07, "loss": 0.0002, "step": 7540 }, { "epoch": 0.9, "learning_rate": 2.907775768535262e-07, "loss": 0.0018, "step": 7541 }, { "epoch": 0.9, "learning_rate": 2.9041591320072333e-07, "loss": 0.0002, "step": 7542 }, { "epoch": 0.9, "learning_rate": 2.900542495479204e-07, "loss": 0.0002, "step": 7543 }, { "epoch": 0.9, "learning_rate": 2.8969258589511753e-07, "loss": 0.0002, "step": 7544 }, { "epoch": 0.9, "learning_rate": 2.8933092224231465e-07, "loss": 0.0003, "step": 7545 }, { "epoch": 0.9, "learning_rate": 2.889692585895117e-07, "loss": 0.0287, "step": 7546 }, { "epoch": 0.9, "learning_rate": 2.8860759493670885e-07, "loss": 0.0003, "step": 7547 }, { "epoch": 0.9, "learning_rate": 2.88245931283906e-07, "loss": 0.0002, "step": 7548 }, { "epoch": 0.9, "learning_rate": 2.8788426763110305e-07, "loss": 0.0003, "step": 7549 }, { "epoch": 0.9, "learning_rate": 2.8752260397830017e-07, "loss": 0.0002, "step": 7550 }, { "epoch": 0.9, "learning_rate": 2.871609403254973e-07, "loss": 0.0002, "step": 7551 }, { "epoch": 0.9, "learning_rate": 2.8679927667269437e-07, "loss": 0.0002, "step": 7552 }, { "epoch": 0.91, "learning_rate": 2.864376130198915e-07, "loss": 0.0003, "step": 7553 }, { "epoch": 0.91, "learning_rate": 2.860759493670886e-07, "loss": 0.0003, "step": 7554 }, { "epoch": 0.91, "learning_rate": 2.857142857142857e-07, "loss": 0.0004, "step": 7555 }, { "epoch": 0.91, "learning_rate": 2.853526220614828e-07, "loss": 0.0003, "step": 7556 }, { "epoch": 0.91, "learning_rate": 2.8499095840867994e-07, "loss": 0.0003, "step": 7557 }, { "epoch": 0.91, "learning_rate": 2.84629294755877e-07, "loss": 0.0011, "step": 7558 }, { "epoch": 0.91, "learning_rate": 2.8426763110307414e-07, "loss": 0.0003, "step": 7559 }, { "epoch": 0.91, "learning_rate": 2.8390596745027127e-07, "loss": 0.0002, "step": 7560 }, { "epoch": 0.91, "learning_rate": 2.8354430379746834e-07, "loss": 0.8648, "step": 7561 }, { "epoch": 0.91, "learning_rate": 2.8318264014466546e-07, "loss": 0.0002, "step": 7562 }, { "epoch": 0.91, "learning_rate": 2.828209764918626e-07, "loss": 0.0002, "step": 7563 }, { "epoch": 0.91, "learning_rate": 2.8245931283905966e-07, "loss": 0.0006, "step": 7564 }, { "epoch": 0.91, "learning_rate": 2.820976491862568e-07, "loss": 0.0003, "step": 7565 }, { "epoch": 0.91, "learning_rate": 2.817359855334539e-07, "loss": 0.0003, "step": 7566 }, { "epoch": 0.91, "learning_rate": 2.81374321880651e-07, "loss": 0.0004, "step": 7567 }, { "epoch": 0.91, "learning_rate": 2.810126582278481e-07, "loss": 0.0002, "step": 7568 }, { "epoch": 0.91, "learning_rate": 2.8065099457504524e-07, "loss": 0.0003, "step": 7569 }, { "epoch": 0.91, "learning_rate": 2.802893309222423e-07, "loss": 0.0004, "step": 7570 }, { "epoch": 0.91, "learning_rate": 2.7992766726943943e-07, "loss": 0.0197, "step": 7571 }, { "epoch": 0.91, "learning_rate": 2.7956600361663656e-07, "loss": 0.7784, "step": 7572 }, { "epoch": 0.91, "learning_rate": 2.7920433996383363e-07, "loss": 0.0005, "step": 7573 }, { "epoch": 0.91, "learning_rate": 2.7884267631103076e-07, "loss": 0.0003, "step": 7574 }, { "epoch": 0.91, "learning_rate": 2.784810126582279e-07, "loss": 0.0002, "step": 7575 }, { "epoch": 0.91, "learning_rate": 2.7811934900542495e-07, "loss": 0.0002, "step": 7576 }, { "epoch": 0.91, "learning_rate": 2.777576853526221e-07, "loss": 0.861, "step": 7577 }, { "epoch": 0.91, "learning_rate": 2.773960216998192e-07, "loss": 0.0003, "step": 7578 }, { "epoch": 0.91, "learning_rate": 2.770343580470163e-07, "loss": 0.0002, "step": 7579 }, { "epoch": 0.91, "learning_rate": 2.766726943942134e-07, "loss": 0.0004, "step": 7580 }, { "epoch": 0.91, "learning_rate": 2.7631103074141053e-07, "loss": 0.0004, "step": 7581 }, { "epoch": 0.91, "learning_rate": 2.759493670886076e-07, "loss": 0.0003, "step": 7582 }, { "epoch": 0.91, "learning_rate": 2.755877034358047e-07, "loss": 0.0003, "step": 7583 }, { "epoch": 0.91, "learning_rate": 2.7522603978300185e-07, "loss": 0.0003, "step": 7584 }, { "epoch": 0.91, "learning_rate": 2.748643761301989e-07, "loss": 0.0002, "step": 7585 }, { "epoch": 0.91, "learning_rate": 2.7450271247739605e-07, "loss": 0.0005, "step": 7586 }, { "epoch": 0.91, "learning_rate": 2.7414104882459317e-07, "loss": 0.0001, "step": 7587 }, { "epoch": 0.91, "learning_rate": 2.7377938517179025e-07, "loss": 0.0012, "step": 7588 }, { "epoch": 0.91, "learning_rate": 2.7341772151898737e-07, "loss": 0.0002, "step": 7589 }, { "epoch": 0.91, "learning_rate": 2.7305605786618444e-07, "loss": 0.0003, "step": 7590 }, { "epoch": 0.91, "learning_rate": 2.7269439421338157e-07, "loss": 0.0003, "step": 7591 }, { "epoch": 0.91, "learning_rate": 2.723327305605787e-07, "loss": 0.0022, "step": 7592 }, { "epoch": 0.91, "learning_rate": 2.7197106690777577e-07, "loss": 1.3712, "step": 7593 }, { "epoch": 0.91, "learning_rate": 2.716094032549729e-07, "loss": 0.0004, "step": 7594 }, { "epoch": 0.91, "learning_rate": 2.7124773960216996e-07, "loss": 0.2827, "step": 7595 }, { "epoch": 0.91, "learning_rate": 2.708860759493671e-07, "loss": 0.0005, "step": 7596 }, { "epoch": 0.91, "learning_rate": 2.705244122965642e-07, "loss": 0.0106, "step": 7597 }, { "epoch": 0.91, "learning_rate": 2.701627486437613e-07, "loss": 0.0082, "step": 7598 }, { "epoch": 0.91, "learning_rate": 2.698010849909584e-07, "loss": 0.0003, "step": 7599 }, { "epoch": 0.91, "learning_rate": 2.694394213381555e-07, "loss": 0.9285, "step": 7600 }, { "epoch": 0.91, "learning_rate": 2.690777576853526e-07, "loss": 0.0002, "step": 7601 }, { "epoch": 0.91, "learning_rate": 2.6871609403254973e-07, "loss": 0.0003, "step": 7602 }, { "epoch": 0.91, "learning_rate": 2.683544303797468e-07, "loss": 0.0002, "step": 7603 }, { "epoch": 0.91, "learning_rate": 2.6799276672694393e-07, "loss": 0.0002, "step": 7604 }, { "epoch": 0.91, "learning_rate": 2.6763110307414106e-07, "loss": 0.0002, "step": 7605 }, { "epoch": 0.91, "learning_rate": 2.6726943942133813e-07, "loss": 0.0001, "step": 7606 }, { "epoch": 0.91, "learning_rate": 2.6690777576853526e-07, "loss": 0.0005, "step": 7607 }, { "epoch": 0.91, "learning_rate": 2.665461121157324e-07, "loss": 0.0003, "step": 7608 }, { "epoch": 0.91, "learning_rate": 2.6618444846292945e-07, "loss": 0.0005, "step": 7609 }, { "epoch": 0.91, "learning_rate": 2.658227848101266e-07, "loss": 0.0005, "step": 7610 }, { "epoch": 0.91, "learning_rate": 2.654611211573237e-07, "loss": 0.0002, "step": 7611 }, { "epoch": 0.91, "learning_rate": 2.650994575045208e-07, "loss": 0.0002, "step": 7612 }, { "epoch": 0.91, "learning_rate": 2.647377938517179e-07, "loss": 0.0002, "step": 7613 }, { "epoch": 0.91, "learning_rate": 2.64376130198915e-07, "loss": 0.0002, "step": 7614 }, { "epoch": 0.91, "learning_rate": 2.640144665461121e-07, "loss": 0.9909, "step": 7615 }, { "epoch": 0.91, "learning_rate": 2.636528028933092e-07, "loss": 0.0003, "step": 7616 }, { "epoch": 0.91, "learning_rate": 2.6329113924050635e-07, "loss": 0.0003, "step": 7617 }, { "epoch": 0.91, "learning_rate": 2.629294755877034e-07, "loss": 0.0002, "step": 7618 }, { "epoch": 0.91, "learning_rate": 2.6256781193490055e-07, "loss": 0.0002, "step": 7619 }, { "epoch": 0.91, "learning_rate": 2.6220614828209767e-07, "loss": 0.0039, "step": 7620 }, { "epoch": 0.91, "learning_rate": 2.6184448462929474e-07, "loss": 0.0002, "step": 7621 }, { "epoch": 0.91, "learning_rate": 2.6148282097649187e-07, "loss": 0.0003, "step": 7622 }, { "epoch": 0.91, "learning_rate": 2.61121157323689e-07, "loss": 0.0002, "step": 7623 }, { "epoch": 0.91, "learning_rate": 2.6075949367088607e-07, "loss": 0.001, "step": 7624 }, { "epoch": 0.91, "learning_rate": 2.603978300180832e-07, "loss": 0.0003, "step": 7625 }, { "epoch": 0.91, "learning_rate": 2.600361663652803e-07, "loss": 0.0027, "step": 7626 }, { "epoch": 0.91, "learning_rate": 2.596745027124774e-07, "loss": 0.0002, "step": 7627 }, { "epoch": 0.91, "learning_rate": 2.593128390596745e-07, "loss": 0.0009, "step": 7628 }, { "epoch": 0.91, "learning_rate": 2.5895117540687164e-07, "loss": 0.0002, "step": 7629 }, { "epoch": 0.91, "learning_rate": 2.585895117540687e-07, "loss": 0.0004, "step": 7630 }, { "epoch": 0.91, "learning_rate": 2.5822784810126584e-07, "loss": 0.0003, "step": 7631 }, { "epoch": 0.91, "learning_rate": 2.5786618444846296e-07, "loss": 0.0003, "step": 7632 }, { "epoch": 0.91, "learning_rate": 2.5750452079566004e-07, "loss": 0.0003, "step": 7633 }, { "epoch": 0.91, "learning_rate": 2.5714285714285716e-07, "loss": 0.0002, "step": 7634 }, { "epoch": 0.91, "learning_rate": 2.567811934900543e-07, "loss": 0.0002, "step": 7635 }, { "epoch": 0.92, "learning_rate": 2.5641952983725136e-07, "loss": 0.0002, "step": 7636 }, { "epoch": 0.92, "learning_rate": 2.560578661844485e-07, "loss": 0.0002, "step": 7637 }, { "epoch": 0.92, "learning_rate": 2.556962025316456e-07, "loss": 0.0004, "step": 7638 }, { "epoch": 0.92, "learning_rate": 2.553345388788427e-07, "loss": 1.0165, "step": 7639 }, { "epoch": 0.92, "learning_rate": 2.549728752260398e-07, "loss": 0.0002, "step": 7640 }, { "epoch": 0.92, "learning_rate": 2.5461121157323693e-07, "loss": 0.0001, "step": 7641 }, { "epoch": 0.92, "learning_rate": 2.54249547920434e-07, "loss": 0.0006, "step": 7642 }, { "epoch": 0.92, "learning_rate": 2.5388788426763113e-07, "loss": 0.0011, "step": 7643 }, { "epoch": 0.92, "learning_rate": 2.5352622061482826e-07, "loss": 0.0003, "step": 7644 }, { "epoch": 0.92, "learning_rate": 2.5316455696202533e-07, "loss": 0.0003, "step": 7645 }, { "epoch": 0.92, "learning_rate": 2.5280289330922245e-07, "loss": 0.4212, "step": 7646 }, { "epoch": 0.92, "learning_rate": 2.524412296564196e-07, "loss": 0.0002, "step": 7647 }, { "epoch": 0.92, "learning_rate": 2.5207956600361665e-07, "loss": 0.0002, "step": 7648 }, { "epoch": 0.92, "learning_rate": 2.517179023508138e-07, "loss": 0.0002, "step": 7649 }, { "epoch": 0.92, "learning_rate": 2.5135623869801085e-07, "loss": 0.0002, "step": 7650 }, { "epoch": 0.92, "learning_rate": 2.50994575045208e-07, "loss": 0.0001, "step": 7651 }, { "epoch": 0.92, "learning_rate": 2.506329113924051e-07, "loss": 0.0002, "step": 7652 }, { "epoch": 0.92, "learning_rate": 2.5027124773960217e-07, "loss": 0.0003, "step": 7653 }, { "epoch": 0.92, "learning_rate": 2.499095840867993e-07, "loss": 0.0003, "step": 7654 }, { "epoch": 0.92, "learning_rate": 2.4954792043399637e-07, "loss": 0.0002, "step": 7655 }, { "epoch": 0.92, "learning_rate": 2.491862567811935e-07, "loss": 0.0003, "step": 7656 }, { "epoch": 0.92, "learning_rate": 2.4882459312839057e-07, "loss": 0.0002, "step": 7657 }, { "epoch": 0.92, "learning_rate": 2.484629294755877e-07, "loss": 0.0002, "step": 7658 }, { "epoch": 0.92, "learning_rate": 2.481012658227848e-07, "loss": 0.0003, "step": 7659 }, { "epoch": 0.92, "learning_rate": 2.477396021699819e-07, "loss": 0.0002, "step": 7660 }, { "epoch": 0.92, "learning_rate": 2.47377938517179e-07, "loss": 0.0003, "step": 7661 }, { "epoch": 0.92, "learning_rate": 2.4701627486437614e-07, "loss": 0.0003, "step": 7662 }, { "epoch": 0.92, "learning_rate": 2.466546112115732e-07, "loss": 0.0003, "step": 7663 }, { "epoch": 0.92, "learning_rate": 2.4629294755877034e-07, "loss": 0.0002, "step": 7664 }, { "epoch": 0.92, "learning_rate": 2.4593128390596746e-07, "loss": 0.0003, "step": 7665 }, { "epoch": 0.92, "learning_rate": 2.4556962025316454e-07, "loss": 0.9599, "step": 7666 }, { "epoch": 0.92, "learning_rate": 2.4520795660036166e-07, "loss": 0.0004, "step": 7667 }, { "epoch": 0.92, "learning_rate": 2.448462929475588e-07, "loss": 0.0003, "step": 7668 }, { "epoch": 0.92, "learning_rate": 2.4448462929475586e-07, "loss": 0.0002, "step": 7669 }, { "epoch": 0.92, "learning_rate": 2.44122965641953e-07, "loss": 0.0002, "step": 7670 }, { "epoch": 0.92, "learning_rate": 2.437613019891501e-07, "loss": 0.0003, "step": 7671 }, { "epoch": 0.92, "learning_rate": 2.433996383363472e-07, "loss": 0.0003, "step": 7672 }, { "epoch": 0.92, "learning_rate": 2.430379746835443e-07, "loss": 0.0004, "step": 7673 }, { "epoch": 0.92, "learning_rate": 2.4267631103074143e-07, "loss": 0.0003, "step": 7674 }, { "epoch": 0.92, "learning_rate": 2.423146473779385e-07, "loss": 0.0002, "step": 7675 }, { "epoch": 0.92, "learning_rate": 2.4195298372513563e-07, "loss": 0.0004, "step": 7676 }, { "epoch": 0.92, "learning_rate": 2.4159132007233275e-07, "loss": 0.0003, "step": 7677 }, { "epoch": 0.92, "learning_rate": 2.4122965641952983e-07, "loss": 0.0002, "step": 7678 }, { "epoch": 0.92, "learning_rate": 2.4086799276672695e-07, "loss": 0.0002, "step": 7679 }, { "epoch": 0.92, "learning_rate": 2.405063291139241e-07, "loss": 0.0001, "step": 7680 }, { "epoch": 0.92, "learning_rate": 2.4014466546112115e-07, "loss": 0.5084, "step": 7681 }, { "epoch": 0.92, "learning_rate": 2.397830018083183e-07, "loss": 0.0002, "step": 7682 }, { "epoch": 0.92, "learning_rate": 2.394213381555154e-07, "loss": 0.0002, "step": 7683 }, { "epoch": 0.92, "learning_rate": 2.3905967450271247e-07, "loss": 0.012, "step": 7684 }, { "epoch": 0.92, "learning_rate": 2.386980108499096e-07, "loss": 0.0003, "step": 7685 }, { "epoch": 0.92, "learning_rate": 2.3833634719710672e-07, "loss": 0.0002, "step": 7686 }, { "epoch": 0.92, "learning_rate": 2.379746835443038e-07, "loss": 0.0002, "step": 7687 }, { "epoch": 0.92, "learning_rate": 2.3761301989150092e-07, "loss": 0.0003, "step": 7688 }, { "epoch": 0.92, "learning_rate": 2.3725135623869805e-07, "loss": 0.0002, "step": 7689 }, { "epoch": 0.92, "learning_rate": 2.3688969258589512e-07, "loss": 0.0003, "step": 7690 }, { "epoch": 0.92, "learning_rate": 2.3652802893309224e-07, "loss": 0.0001, "step": 7691 }, { "epoch": 0.92, "learning_rate": 2.3616636528028937e-07, "loss": 0.0004, "step": 7692 }, { "epoch": 0.92, "learning_rate": 2.3580470162748644e-07, "loss": 0.2186, "step": 7693 }, { "epoch": 0.92, "learning_rate": 2.3544303797468357e-07, "loss": 0.0003, "step": 7694 }, { "epoch": 0.92, "learning_rate": 2.3508137432188067e-07, "loss": 0.0002, "step": 7695 }, { "epoch": 0.92, "learning_rate": 2.3471971066907776e-07, "loss": 0.0002, "step": 7696 }, { "epoch": 0.92, "learning_rate": 2.343580470162749e-07, "loss": 0.0002, "step": 7697 }, { "epoch": 0.92, "learning_rate": 2.33996383363472e-07, "loss": 0.0002, "step": 7698 }, { "epoch": 0.92, "learning_rate": 2.336347197106691e-07, "loss": 0.0002, "step": 7699 }, { "epoch": 0.92, "learning_rate": 2.3327305605786619e-07, "loss": 0.0003, "step": 7700 }, { "epoch": 0.92, "learning_rate": 2.329113924050633e-07, "loss": 0.0003, "step": 7701 }, { "epoch": 0.92, "learning_rate": 2.3254972875226038e-07, "loss": 0.0003, "step": 7702 }, { "epoch": 0.92, "learning_rate": 2.321880650994575e-07, "loss": 0.0017, "step": 7703 }, { "epoch": 0.92, "learning_rate": 2.3182640144665463e-07, "loss": 0.0002, "step": 7704 }, { "epoch": 0.92, "learning_rate": 2.314647377938517e-07, "loss": 0.0005, "step": 7705 }, { "epoch": 0.92, "learning_rate": 2.3110307414104883e-07, "loss": 0.0004, "step": 7706 }, { "epoch": 0.92, "learning_rate": 2.3074141048824596e-07, "loss": 0.0002, "step": 7707 }, { "epoch": 0.92, "learning_rate": 2.3037974683544303e-07, "loss": 0.0002, "step": 7708 }, { "epoch": 0.92, "learning_rate": 2.3001808318264015e-07, "loss": 0.0003, "step": 7709 }, { "epoch": 0.92, "learning_rate": 2.2965641952983728e-07, "loss": 0.0002, "step": 7710 }, { "epoch": 0.92, "learning_rate": 2.2929475587703435e-07, "loss": 0.0002, "step": 7711 }, { "epoch": 0.92, "learning_rate": 2.2893309222423148e-07, "loss": 0.0004, "step": 7712 }, { "epoch": 0.92, "learning_rate": 2.285714285714286e-07, "loss": 0.0049, "step": 7713 }, { "epoch": 0.92, "learning_rate": 2.2820976491862568e-07, "loss": 0.0002, "step": 7714 }, { "epoch": 0.92, "learning_rate": 2.278481012658228e-07, "loss": 0.0003, "step": 7715 }, { "epoch": 0.92, "learning_rate": 2.2748643761301993e-07, "loss": 0.6677, "step": 7716 }, { "epoch": 0.92, "learning_rate": 2.27124773960217e-07, "loss": 0.0003, "step": 7717 }, { "epoch": 0.92, "learning_rate": 2.2676311030741412e-07, "loss": 0.0002, "step": 7718 }, { "epoch": 0.92, "learning_rate": 2.264014466546112e-07, "loss": 0.0003, "step": 7719 }, { "epoch": 0.93, "learning_rate": 2.2603978300180832e-07, "loss": 0.1686, "step": 7720 }, { "epoch": 0.93, "learning_rate": 2.2567811934900545e-07, "loss": 0.0004, "step": 7721 }, { "epoch": 0.93, "learning_rate": 2.2531645569620252e-07, "loss": 0.0002, "step": 7722 }, { "epoch": 0.93, "learning_rate": 2.2495479204339964e-07, "loss": 0.0002, "step": 7723 }, { "epoch": 0.93, "learning_rate": 2.2459312839059677e-07, "loss": 0.0002, "step": 7724 }, { "epoch": 0.93, "learning_rate": 2.2423146473779384e-07, "loss": 0.4123, "step": 7725 }, { "epoch": 0.93, "learning_rate": 2.2386980108499097e-07, "loss": 0.0002, "step": 7726 }, { "epoch": 0.93, "learning_rate": 2.2350813743218807e-07, "loss": 0.4172, "step": 7727 }, { "epoch": 0.93, "learning_rate": 2.2314647377938516e-07, "loss": 0.0002, "step": 7728 }, { "epoch": 0.93, "learning_rate": 2.227848101265823e-07, "loss": 0.0002, "step": 7729 }, { "epoch": 0.93, "learning_rate": 2.224231464737794e-07, "loss": 0.0002, "step": 7730 }, { "epoch": 0.93, "learning_rate": 2.220614828209765e-07, "loss": 0.0002, "step": 7731 }, { "epoch": 0.93, "learning_rate": 2.2169981916817359e-07, "loss": 0.0002, "step": 7732 }, { "epoch": 0.93, "learning_rate": 2.213381555153707e-07, "loss": 0.0003, "step": 7733 }, { "epoch": 0.93, "learning_rate": 2.209764918625678e-07, "loss": 0.0003, "step": 7734 }, { "epoch": 0.93, "learning_rate": 2.206148282097649e-07, "loss": 0.0002, "step": 7735 }, { "epoch": 0.93, "learning_rate": 2.2025316455696203e-07, "loss": 0.0003, "step": 7736 }, { "epoch": 0.93, "learning_rate": 2.198915009041591e-07, "loss": 0.0003, "step": 7737 }, { "epoch": 0.93, "learning_rate": 2.1952983725135623e-07, "loss": 0.0002, "step": 7738 }, { "epoch": 0.93, "learning_rate": 2.1916817359855336e-07, "loss": 1.872, "step": 7739 }, { "epoch": 0.93, "learning_rate": 2.1880650994575043e-07, "loss": 0.0001, "step": 7740 }, { "epoch": 0.93, "learning_rate": 2.1844484629294755e-07, "loss": 0.0003, "step": 7741 }, { "epoch": 0.93, "learning_rate": 2.1808318264014468e-07, "loss": 0.0003, "step": 7742 }, { "epoch": 0.93, "learning_rate": 2.1772151898734175e-07, "loss": 0.0002, "step": 7743 }, { "epoch": 0.93, "learning_rate": 2.1735985533453888e-07, "loss": 0.0004, "step": 7744 }, { "epoch": 0.93, "learning_rate": 2.16998191681736e-07, "loss": 0.0003, "step": 7745 }, { "epoch": 0.93, "learning_rate": 2.1663652802893308e-07, "loss": 0.0003, "step": 7746 }, { "epoch": 0.93, "learning_rate": 2.162748643761302e-07, "loss": 0.0454, "step": 7747 }, { "epoch": 0.93, "learning_rate": 2.1591320072332733e-07, "loss": 0.6313, "step": 7748 }, { "epoch": 0.93, "learning_rate": 2.155515370705244e-07, "loss": 0.0002, "step": 7749 }, { "epoch": 0.93, "learning_rate": 2.1518987341772152e-07, "loss": 0.856, "step": 7750 }, { "epoch": 0.93, "learning_rate": 2.1482820976491865e-07, "loss": 0.0002, "step": 7751 }, { "epoch": 0.93, "learning_rate": 2.1446654611211572e-07, "loss": 0.0003, "step": 7752 }, { "epoch": 0.93, "learning_rate": 2.1410488245931285e-07, "loss": 0.0019, "step": 7753 }, { "epoch": 0.93, "learning_rate": 2.1374321880650997e-07, "loss": 0.0003, "step": 7754 }, { "epoch": 0.93, "learning_rate": 2.1338155515370704e-07, "loss": 0.0009, "step": 7755 }, { "epoch": 0.93, "learning_rate": 2.1301989150090417e-07, "loss": 0.0003, "step": 7756 }, { "epoch": 0.93, "learning_rate": 2.1265822784810127e-07, "loss": 0.0073, "step": 7757 }, { "epoch": 0.93, "learning_rate": 2.1229656419529837e-07, "loss": 0.0003, "step": 7758 }, { "epoch": 0.93, "learning_rate": 2.119349005424955e-07, "loss": 0.0003, "step": 7759 }, { "epoch": 0.93, "learning_rate": 2.115732368896926e-07, "loss": 0.0002, "step": 7760 }, { "epoch": 0.93, "learning_rate": 2.112115732368897e-07, "loss": 0.0002, "step": 7761 }, { "epoch": 0.93, "learning_rate": 2.108499095840868e-07, "loss": 0.0002, "step": 7762 }, { "epoch": 0.93, "learning_rate": 2.1048824593128391e-07, "loss": 0.2517, "step": 7763 }, { "epoch": 0.93, "learning_rate": 2.10126582278481e-07, "loss": 0.0004, "step": 7764 }, { "epoch": 0.93, "learning_rate": 2.097649186256781e-07, "loss": 0.0002, "step": 7765 }, { "epoch": 0.93, "learning_rate": 2.0940325497287524e-07, "loss": 0.0002, "step": 7766 }, { "epoch": 0.93, "learning_rate": 2.090415913200723e-07, "loss": 0.0001, "step": 7767 }, { "epoch": 0.93, "learning_rate": 2.0867992766726943e-07, "loss": 0.0003, "step": 7768 }, { "epoch": 0.93, "learning_rate": 2.0831826401446656e-07, "loss": 0.0002, "step": 7769 }, { "epoch": 0.93, "learning_rate": 2.0795660036166363e-07, "loss": 0.0002, "step": 7770 }, { "epoch": 0.93, "learning_rate": 2.0759493670886076e-07, "loss": 0.0006, "step": 7771 }, { "epoch": 0.93, "learning_rate": 2.0723327305605788e-07, "loss": 0.0002, "step": 7772 }, { "epoch": 0.93, "learning_rate": 2.0687160940325495e-07, "loss": 0.0002, "step": 7773 }, { "epoch": 0.93, "learning_rate": 2.0650994575045208e-07, "loss": 0.001, "step": 7774 }, { "epoch": 0.93, "learning_rate": 2.061482820976492e-07, "loss": 0.0002, "step": 7775 }, { "epoch": 0.93, "learning_rate": 2.0578661844484628e-07, "loss": 0.0003, "step": 7776 }, { "epoch": 0.93, "learning_rate": 2.054249547920434e-07, "loss": 0.0002, "step": 7777 }, { "epoch": 0.93, "learning_rate": 2.0506329113924053e-07, "loss": 0.0003, "step": 7778 }, { "epoch": 0.93, "learning_rate": 2.047016274864376e-07, "loss": 0.0002, "step": 7779 }, { "epoch": 0.93, "learning_rate": 2.0433996383363473e-07, "loss": 0.0005, "step": 7780 }, { "epoch": 0.93, "learning_rate": 2.0397830018083185e-07, "loss": 0.4722, "step": 7781 }, { "epoch": 0.93, "learning_rate": 2.0361663652802892e-07, "loss": 0.0002, "step": 7782 }, { "epoch": 0.93, "learning_rate": 2.0325497287522605e-07, "loss": 0.0005, "step": 7783 }, { "epoch": 0.93, "learning_rate": 2.0289330922242317e-07, "loss": 0.0003, "step": 7784 }, { "epoch": 0.93, "learning_rate": 2.0253164556962025e-07, "loss": 0.0008, "step": 7785 }, { "epoch": 0.93, "learning_rate": 2.0216998191681737e-07, "loss": 0.0003, "step": 7786 }, { "epoch": 0.93, "learning_rate": 2.0180831826401447e-07, "loss": 0.0001, "step": 7787 }, { "epoch": 0.93, "learning_rate": 2.0144665461121157e-07, "loss": 0.0002, "step": 7788 }, { "epoch": 0.93, "learning_rate": 2.010849909584087e-07, "loss": 0.0003, "step": 7789 }, { "epoch": 0.93, "learning_rate": 2.007233273056058e-07, "loss": 0.0002, "step": 7790 }, { "epoch": 0.93, "learning_rate": 2.003616636528029e-07, "loss": 1.1069, "step": 7791 }, { "epoch": 0.93, "learning_rate": 2e-07, "loss": 0.0003, "step": 7792 }, { "epoch": 0.93, "learning_rate": 1.9963833634719712e-07, "loss": 0.0003, "step": 7793 }, { "epoch": 0.93, "learning_rate": 1.9927667269439422e-07, "loss": 0.0002, "step": 7794 }, { "epoch": 0.93, "learning_rate": 1.9891500904159131e-07, "loss": 0.0002, "step": 7795 }, { "epoch": 0.93, "learning_rate": 1.9855334538878844e-07, "loss": 0.0002, "step": 7796 }, { "epoch": 0.93, "learning_rate": 1.981916817359855e-07, "loss": 0.0002, "step": 7797 }, { "epoch": 0.93, "learning_rate": 1.9783001808318264e-07, "loss": 0.0005, "step": 7798 }, { "epoch": 0.93, "learning_rate": 1.9746835443037976e-07, "loss": 0.0002, "step": 7799 }, { "epoch": 0.93, "learning_rate": 1.9710669077757683e-07, "loss": 0.0002, "step": 7800 }, { "epoch": 0.93, "learning_rate": 1.9674502712477396e-07, "loss": 0.9043, "step": 7801 }, { "epoch": 0.93, "learning_rate": 1.9638336347197108e-07, "loss": 0.0002, "step": 7802 }, { "epoch": 0.94, "learning_rate": 1.9602169981916816e-07, "loss": 0.0002, "step": 7803 }, { "epoch": 0.94, "learning_rate": 1.9566003616636528e-07, "loss": 0.0003, "step": 7804 }, { "epoch": 0.94, "learning_rate": 1.952983725135624e-07, "loss": 0.0003, "step": 7805 }, { "epoch": 0.94, "learning_rate": 1.9493670886075948e-07, "loss": 0.0251, "step": 7806 }, { "epoch": 0.94, "learning_rate": 1.945750452079566e-07, "loss": 0.0003, "step": 7807 }, { "epoch": 0.94, "learning_rate": 1.9421338155515373e-07, "loss": 0.0003, "step": 7808 }, { "epoch": 0.94, "learning_rate": 1.938517179023508e-07, "loss": 0.0002, "step": 7809 }, { "epoch": 0.94, "learning_rate": 1.9349005424954793e-07, "loss": 0.0002, "step": 7810 }, { "epoch": 0.94, "learning_rate": 1.9312839059674505e-07, "loss": 0.0002, "step": 7811 }, { "epoch": 0.94, "learning_rate": 1.9276672694394213e-07, "loss": 0.0002, "step": 7812 }, { "epoch": 0.94, "learning_rate": 1.9240506329113925e-07, "loss": 0.0004, "step": 7813 }, { "epoch": 0.94, "learning_rate": 1.9204339963833638e-07, "loss": 0.0003, "step": 7814 }, { "epoch": 0.94, "learning_rate": 1.9168173598553345e-07, "loss": 0.0004, "step": 7815 }, { "epoch": 0.94, "learning_rate": 1.9132007233273057e-07, "loss": 0.0002, "step": 7816 }, { "epoch": 0.94, "learning_rate": 1.9095840867992767e-07, "loss": 0.0003, "step": 7817 }, { "epoch": 0.94, "learning_rate": 1.9059674502712477e-07, "loss": 0.0015, "step": 7818 }, { "epoch": 0.94, "learning_rate": 1.902350813743219e-07, "loss": 0.0002, "step": 7819 }, { "epoch": 0.94, "learning_rate": 1.89873417721519e-07, "loss": 0.0003, "step": 7820 }, { "epoch": 0.94, "learning_rate": 1.895117540687161e-07, "loss": 0.0002, "step": 7821 }, { "epoch": 0.94, "learning_rate": 1.891500904159132e-07, "loss": 0.0003, "step": 7822 }, { "epoch": 0.94, "learning_rate": 1.8878842676311032e-07, "loss": 0.0003, "step": 7823 }, { "epoch": 0.94, "learning_rate": 1.884267631103074e-07, "loss": 0.0003, "step": 7824 }, { "epoch": 0.94, "learning_rate": 1.8806509945750452e-07, "loss": 0.0003, "step": 7825 }, { "epoch": 0.94, "learning_rate": 1.8770343580470164e-07, "loss": 0.0002, "step": 7826 }, { "epoch": 0.94, "learning_rate": 1.8734177215189874e-07, "loss": 0.7834, "step": 7827 }, { "epoch": 0.94, "learning_rate": 1.8698010849909584e-07, "loss": 0.0003, "step": 7828 }, { "epoch": 0.94, "learning_rate": 1.8661844484629294e-07, "loss": 0.0004, "step": 7829 }, { "epoch": 0.94, "learning_rate": 1.8625678119349006e-07, "loss": 0.0002, "step": 7830 }, { "epoch": 0.94, "learning_rate": 1.8589511754068716e-07, "loss": 0.0002, "step": 7831 }, { "epoch": 0.94, "learning_rate": 1.8553345388788426e-07, "loss": 0.7288, "step": 7832 }, { "epoch": 0.94, "learning_rate": 1.8517179023508139e-07, "loss": 0.0002, "step": 7833 }, { "epoch": 0.94, "learning_rate": 1.8481012658227849e-07, "loss": 0.0003, "step": 7834 }, { "epoch": 0.94, "learning_rate": 1.8444846292947558e-07, "loss": 0.0049, "step": 7835 }, { "epoch": 0.94, "learning_rate": 1.840867992766727e-07, "loss": 0.0003, "step": 7836 }, { "epoch": 0.94, "learning_rate": 1.837251356238698e-07, "loss": 0.0005, "step": 7837 }, { "epoch": 0.94, "learning_rate": 1.833634719710669e-07, "loss": 0.3987, "step": 7838 }, { "epoch": 0.94, "learning_rate": 1.8300180831826403e-07, "loss": 0.0002, "step": 7839 }, { "epoch": 0.94, "learning_rate": 1.8264014466546113e-07, "loss": 0.0002, "step": 7840 }, { "epoch": 0.94, "learning_rate": 1.8227848101265823e-07, "loss": 0.0003, "step": 7841 }, { "epoch": 0.94, "learning_rate": 1.8191681735985535e-07, "loss": 0.0003, "step": 7842 }, { "epoch": 0.94, "learning_rate": 1.8155515370705245e-07, "loss": 0.0002, "step": 7843 }, { "epoch": 0.94, "learning_rate": 1.8119349005424955e-07, "loss": 0.0003, "step": 7844 }, { "epoch": 0.94, "learning_rate": 1.8083182640144668e-07, "loss": 0.0002, "step": 7845 }, { "epoch": 0.94, "learning_rate": 1.8047016274864378e-07, "loss": 0.0002, "step": 7846 }, { "epoch": 0.94, "learning_rate": 1.8010849909584088e-07, "loss": 0.0002, "step": 7847 }, { "epoch": 0.94, "learning_rate": 1.7974683544303797e-07, "loss": 0.0003, "step": 7848 }, { "epoch": 0.94, "learning_rate": 1.7938517179023507e-07, "loss": 0.0002, "step": 7849 }, { "epoch": 0.94, "learning_rate": 1.790235081374322e-07, "loss": 0.0002, "step": 7850 }, { "epoch": 0.94, "learning_rate": 1.786618444846293e-07, "loss": 0.0004, "step": 7851 }, { "epoch": 0.94, "learning_rate": 1.783001808318264e-07, "loss": 0.0002, "step": 7852 }, { "epoch": 0.94, "learning_rate": 1.779385171790235e-07, "loss": 0.0002, "step": 7853 }, { "epoch": 0.94, "learning_rate": 1.7757685352622062e-07, "loss": 0.0002, "step": 7854 }, { "epoch": 0.94, "learning_rate": 1.7721518987341772e-07, "loss": 0.0002, "step": 7855 }, { "epoch": 0.94, "learning_rate": 1.7685352622061482e-07, "loss": 0.0007, "step": 7856 }, { "epoch": 0.94, "learning_rate": 1.7649186256781194e-07, "loss": 0.0002, "step": 7857 }, { "epoch": 0.94, "learning_rate": 1.7613019891500904e-07, "loss": 0.0002, "step": 7858 }, { "epoch": 0.94, "learning_rate": 1.7576853526220614e-07, "loss": 0.0003, "step": 7859 }, { "epoch": 0.94, "learning_rate": 1.7540687160940327e-07, "loss": 0.0003, "step": 7860 }, { "epoch": 0.94, "learning_rate": 1.7504520795660036e-07, "loss": 0.0075, "step": 7861 }, { "epoch": 0.94, "learning_rate": 1.7468354430379746e-07, "loss": 0.0004, "step": 7862 }, { "epoch": 0.94, "learning_rate": 1.743218806509946e-07, "loss": 0.0641, "step": 7863 }, { "epoch": 0.94, "learning_rate": 1.739602169981917e-07, "loss": 0.0002, "step": 7864 }, { "epoch": 0.94, "learning_rate": 1.7359855334538879e-07, "loss": 0.0002, "step": 7865 }, { "epoch": 0.94, "learning_rate": 1.732368896925859e-07, "loss": 0.0002, "step": 7866 }, { "epoch": 0.94, "learning_rate": 1.72875226039783e-07, "loss": 0.0002, "step": 7867 }, { "epoch": 0.94, "learning_rate": 1.725135623869801e-07, "loss": 0.0032, "step": 7868 }, { "epoch": 0.94, "learning_rate": 1.7215189873417723e-07, "loss": 0.0003, "step": 7869 }, { "epoch": 0.94, "learning_rate": 1.7179023508137433e-07, "loss": 0.0003, "step": 7870 }, { "epoch": 0.94, "learning_rate": 1.7142857142857143e-07, "loss": 0.0005, "step": 7871 }, { "epoch": 0.94, "learning_rate": 1.7106690777576856e-07, "loss": 0.0003, "step": 7872 }, { "epoch": 0.94, "learning_rate": 1.7070524412296566e-07, "loss": 0.0002, "step": 7873 }, { "epoch": 0.94, "learning_rate": 1.7034358047016276e-07, "loss": 0.0023, "step": 7874 }, { "epoch": 0.94, "learning_rate": 1.6998191681735988e-07, "loss": 0.0002, "step": 7875 }, { "epoch": 0.94, "learning_rate": 1.6962025316455698e-07, "loss": 0.0002, "step": 7876 }, { "epoch": 0.94, "learning_rate": 1.6925858951175408e-07, "loss": 0.0003, "step": 7877 }, { "epoch": 0.94, "learning_rate": 1.6889692585895118e-07, "loss": 0.0005, "step": 7878 }, { "epoch": 0.94, "learning_rate": 1.6853526220614828e-07, "loss": 0.0004, "step": 7879 }, { "epoch": 0.94, "learning_rate": 1.681735985533454e-07, "loss": 0.0003, "step": 7880 }, { "epoch": 0.94, "learning_rate": 1.678119349005425e-07, "loss": 0.0002, "step": 7881 }, { "epoch": 0.94, "learning_rate": 1.674502712477396e-07, "loss": 0.0002, "step": 7882 }, { "epoch": 0.94, "learning_rate": 1.670886075949367e-07, "loss": 0.0002, "step": 7883 }, { "epoch": 0.94, "learning_rate": 1.6672694394213382e-07, "loss": 1.04, "step": 7884 }, { "epoch": 0.94, "learning_rate": 1.6636528028933092e-07, "loss": 0.0003, "step": 7885 }, { "epoch": 0.94, "learning_rate": 1.6600361663652802e-07, "loss": 0.4064, "step": 7886 }, { "epoch": 0.95, "learning_rate": 1.6564195298372515e-07, "loss": 0.0003, "step": 7887 }, { "epoch": 0.95, "learning_rate": 1.6528028933092224e-07, "loss": 0.0002, "step": 7888 }, { "epoch": 0.95, "learning_rate": 1.6491862567811934e-07, "loss": 0.0015, "step": 7889 }, { "epoch": 0.95, "learning_rate": 1.6455696202531647e-07, "loss": 0.0002, "step": 7890 }, { "epoch": 0.95, "learning_rate": 1.6419529837251357e-07, "loss": 0.0003, "step": 7891 }, { "epoch": 0.95, "learning_rate": 1.6383363471971067e-07, "loss": 0.0002, "step": 7892 }, { "epoch": 0.95, "learning_rate": 1.634719710669078e-07, "loss": 0.0002, "step": 7893 }, { "epoch": 0.95, "learning_rate": 1.631103074141049e-07, "loss": 0.0002, "step": 7894 }, { "epoch": 0.95, "learning_rate": 1.62748643761302e-07, "loss": 0.0003, "step": 7895 }, { "epoch": 0.95, "learning_rate": 1.6238698010849911e-07, "loss": 0.0012, "step": 7896 }, { "epoch": 0.95, "learning_rate": 1.620253164556962e-07, "loss": 0.0002, "step": 7897 }, { "epoch": 0.95, "learning_rate": 1.616636528028933e-07, "loss": 0.0003, "step": 7898 }, { "epoch": 0.95, "learning_rate": 1.6130198915009044e-07, "loss": 0.0002, "step": 7899 }, { "epoch": 0.95, "learning_rate": 1.6094032549728754e-07, "loss": 0.0002, "step": 7900 }, { "epoch": 0.95, "learning_rate": 1.6057866184448463e-07, "loss": 0.0002, "step": 7901 }, { "epoch": 0.95, "learning_rate": 1.6021699819168176e-07, "loss": 0.0002, "step": 7902 }, { "epoch": 0.95, "learning_rate": 1.5985533453887886e-07, "loss": 0.0002, "step": 7903 }, { "epoch": 0.95, "learning_rate": 1.5949367088607596e-07, "loss": 0.0002, "step": 7904 }, { "epoch": 0.95, "learning_rate": 1.5913200723327308e-07, "loss": 0.0003, "step": 7905 }, { "epoch": 0.95, "learning_rate": 1.5877034358047018e-07, "loss": 0.0005, "step": 7906 }, { "epoch": 0.95, "learning_rate": 1.5840867992766728e-07, "loss": 0.5702, "step": 7907 }, { "epoch": 0.95, "learning_rate": 1.5804701627486438e-07, "loss": 0.0002, "step": 7908 }, { "epoch": 0.95, "learning_rate": 1.5768535262206148e-07, "loss": 0.0002, "step": 7909 }, { "epoch": 0.95, "learning_rate": 1.573236889692586e-07, "loss": 0.003, "step": 7910 }, { "epoch": 0.95, "learning_rate": 1.569620253164557e-07, "loss": 0.0004, "step": 7911 }, { "epoch": 0.95, "learning_rate": 1.566003616636528e-07, "loss": 0.0003, "step": 7912 }, { "epoch": 0.95, "learning_rate": 1.562386980108499e-07, "loss": 0.3911, "step": 7913 }, { "epoch": 0.95, "learning_rate": 1.55877034358047e-07, "loss": 0.0003, "step": 7914 }, { "epoch": 0.95, "learning_rate": 1.5551537070524412e-07, "loss": 0.3088, "step": 7915 }, { "epoch": 0.95, "learning_rate": 1.5515370705244122e-07, "loss": 0.7914, "step": 7916 }, { "epoch": 0.95, "learning_rate": 1.5479204339963832e-07, "loss": 0.0002, "step": 7917 }, { "epoch": 0.95, "learning_rate": 1.5443037974683545e-07, "loss": 0.0003, "step": 7918 }, { "epoch": 0.95, "learning_rate": 1.5406871609403255e-07, "loss": 0.0002, "step": 7919 }, { "epoch": 0.95, "learning_rate": 1.5370705244122964e-07, "loss": 0.0008, "step": 7920 }, { "epoch": 0.95, "learning_rate": 1.5334538878842677e-07, "loss": 0.0002, "step": 7921 }, { "epoch": 0.95, "learning_rate": 1.5298372513562387e-07, "loss": 0.0002, "step": 7922 }, { "epoch": 0.95, "learning_rate": 1.5262206148282097e-07, "loss": 0.2963, "step": 7923 }, { "epoch": 0.95, "learning_rate": 1.522603978300181e-07, "loss": 0.0002, "step": 7924 }, { "epoch": 0.95, "learning_rate": 1.518987341772152e-07, "loss": 0.0004, "step": 7925 }, { "epoch": 0.95, "learning_rate": 1.515370705244123e-07, "loss": 0.002, "step": 7926 }, { "epoch": 0.95, "learning_rate": 1.5117540687160942e-07, "loss": 0.0002, "step": 7927 }, { "epoch": 0.95, "learning_rate": 1.5081374321880651e-07, "loss": 0.0018, "step": 7928 }, { "epoch": 0.95, "learning_rate": 1.504520795660036e-07, "loss": 0.0003, "step": 7929 }, { "epoch": 0.95, "learning_rate": 1.5009041591320074e-07, "loss": 0.0002, "step": 7930 }, { "epoch": 0.95, "learning_rate": 1.4972875226039784e-07, "loss": 0.0002, "step": 7931 }, { "epoch": 0.95, "learning_rate": 1.4936708860759494e-07, "loss": 0.0003, "step": 7932 }, { "epoch": 0.95, "learning_rate": 1.4900542495479206e-07, "loss": 0.0002, "step": 7933 }, { "epoch": 0.95, "learning_rate": 1.4864376130198916e-07, "loss": 0.0002, "step": 7934 }, { "epoch": 0.95, "learning_rate": 1.4828209764918626e-07, "loss": 0.0002, "step": 7935 }, { "epoch": 0.95, "learning_rate": 1.4792043399638338e-07, "loss": 0.0002, "step": 7936 }, { "epoch": 0.95, "learning_rate": 1.4755877034358048e-07, "loss": 0.0003, "step": 7937 }, { "epoch": 0.95, "learning_rate": 1.4719710669077758e-07, "loss": 0.0006, "step": 7938 }, { "epoch": 0.95, "learning_rate": 1.4683544303797468e-07, "loss": 0.0004, "step": 7939 }, { "epoch": 0.95, "learning_rate": 1.4647377938517178e-07, "loss": 0.0002, "step": 7940 }, { "epoch": 0.95, "learning_rate": 1.461121157323689e-07, "loss": 0.0002, "step": 7941 }, { "epoch": 0.95, "learning_rate": 1.45750452079566e-07, "loss": 0.0002, "step": 7942 }, { "epoch": 0.95, "learning_rate": 1.453887884267631e-07, "loss": 0.0003, "step": 7943 }, { "epoch": 0.95, "learning_rate": 1.450271247739602e-07, "loss": 0.0003, "step": 7944 }, { "epoch": 0.95, "learning_rate": 1.4466546112115733e-07, "loss": 0.0004, "step": 7945 }, { "epoch": 0.95, "learning_rate": 1.4430379746835443e-07, "loss": 0.0002, "step": 7946 }, { "epoch": 0.95, "learning_rate": 1.4394213381555152e-07, "loss": 0.0004, "step": 7947 }, { "epoch": 0.95, "learning_rate": 1.4358047016274865e-07, "loss": 0.0003, "step": 7948 }, { "epoch": 0.95, "learning_rate": 1.4321880650994575e-07, "loss": 0.9194, "step": 7949 }, { "epoch": 0.95, "learning_rate": 1.4285714285714285e-07, "loss": 0.0003, "step": 7950 }, { "epoch": 0.95, "learning_rate": 1.4249547920433997e-07, "loss": 0.0001, "step": 7951 }, { "epoch": 0.95, "learning_rate": 1.4213381555153707e-07, "loss": 0.0002, "step": 7952 }, { "epoch": 0.95, "learning_rate": 1.4177215189873417e-07, "loss": 0.0002, "step": 7953 }, { "epoch": 0.95, "learning_rate": 1.414104882459313e-07, "loss": 0.075, "step": 7954 }, { "epoch": 0.95, "learning_rate": 1.410488245931284e-07, "loss": 0.6474, "step": 7955 }, { "epoch": 0.95, "learning_rate": 1.406871609403255e-07, "loss": 0.0003, "step": 7956 }, { "epoch": 0.95, "learning_rate": 1.4032549728752262e-07, "loss": 0.0003, "step": 7957 }, { "epoch": 0.95, "learning_rate": 1.3996383363471972e-07, "loss": 0.0015, "step": 7958 }, { "epoch": 0.95, "learning_rate": 1.3960216998191682e-07, "loss": 0.0003, "step": 7959 }, { "epoch": 0.95, "learning_rate": 1.3924050632911394e-07, "loss": 0.0002, "step": 7960 }, { "epoch": 0.95, "learning_rate": 1.3887884267631104e-07, "loss": 0.0002, "step": 7961 }, { "epoch": 0.95, "learning_rate": 1.3851717902350814e-07, "loss": 0.0002, "step": 7962 }, { "epoch": 0.95, "learning_rate": 1.3815551537070526e-07, "loss": 0.0002, "step": 7963 }, { "epoch": 0.95, "learning_rate": 1.3779385171790236e-07, "loss": 0.0004, "step": 7964 }, { "epoch": 0.95, "learning_rate": 1.3743218806509946e-07, "loss": 0.8559, "step": 7965 }, { "epoch": 0.95, "learning_rate": 1.3707052441229659e-07, "loss": 0.0003, "step": 7966 }, { "epoch": 0.95, "learning_rate": 1.3670886075949369e-07, "loss": 0.0003, "step": 7967 }, { "epoch": 0.95, "learning_rate": 1.3634719710669078e-07, "loss": 0.0003, "step": 7968 }, { "epoch": 0.95, "learning_rate": 1.3598553345388788e-07, "loss": 0.0003, "step": 7969 }, { "epoch": 0.96, "learning_rate": 1.3562386980108498e-07, "loss": 0.8779, "step": 7970 }, { "epoch": 0.96, "learning_rate": 1.352622061482821e-07, "loss": 0.0002, "step": 7971 }, { "epoch": 0.96, "learning_rate": 1.349005424954792e-07, "loss": 0.0003, "step": 7972 }, { "epoch": 0.96, "learning_rate": 1.345388788426763e-07, "loss": 0.0003, "step": 7973 }, { "epoch": 0.96, "learning_rate": 1.341772151898734e-07, "loss": 0.3485, "step": 7974 }, { "epoch": 0.96, "learning_rate": 1.3381555153707053e-07, "loss": 0.0002, "step": 7975 }, { "epoch": 0.96, "learning_rate": 1.3345388788426763e-07, "loss": 0.0004, "step": 7976 }, { "epoch": 0.96, "learning_rate": 1.3309222423146473e-07, "loss": 0.0004, "step": 7977 }, { "epoch": 0.96, "learning_rate": 1.3273056057866185e-07, "loss": 0.0002, "step": 7978 }, { "epoch": 0.96, "learning_rate": 1.3236889692585895e-07, "loss": 0.0002, "step": 7979 }, { "epoch": 0.96, "learning_rate": 1.3200723327305605e-07, "loss": 0.0003, "step": 7980 }, { "epoch": 0.96, "learning_rate": 1.3164556962025317e-07, "loss": 0.0003, "step": 7981 }, { "epoch": 0.96, "learning_rate": 1.3128390596745027e-07, "loss": 0.0002, "step": 7982 }, { "epoch": 0.96, "learning_rate": 1.3092224231464737e-07, "loss": 0.0002, "step": 7983 }, { "epoch": 0.96, "learning_rate": 1.305605786618445e-07, "loss": 0.0002, "step": 7984 }, { "epoch": 0.96, "learning_rate": 1.301989150090416e-07, "loss": 0.0002, "step": 7985 }, { "epoch": 0.96, "learning_rate": 1.298372513562387e-07, "loss": 0.0003, "step": 7986 }, { "epoch": 0.96, "learning_rate": 1.2947558770343582e-07, "loss": 0.0002, "step": 7987 }, { "epoch": 0.96, "learning_rate": 1.2911392405063292e-07, "loss": 0.0002, "step": 7988 }, { "epoch": 0.96, "learning_rate": 1.2875226039783002e-07, "loss": 0.0003, "step": 7989 }, { "epoch": 0.96, "learning_rate": 1.2839059674502714e-07, "loss": 0.0003, "step": 7990 }, { "epoch": 0.96, "learning_rate": 1.2802893309222424e-07, "loss": 0.0002, "step": 7991 }, { "epoch": 0.96, "learning_rate": 1.2766726943942134e-07, "loss": 0.0002, "step": 7992 }, { "epoch": 0.96, "learning_rate": 1.2730560578661847e-07, "loss": 0.0003, "step": 7993 }, { "epoch": 0.96, "learning_rate": 1.2694394213381557e-07, "loss": 0.0014, "step": 7994 }, { "epoch": 0.96, "learning_rate": 1.2658227848101266e-07, "loss": 0.0003, "step": 7995 }, { "epoch": 0.96, "learning_rate": 1.262206148282098e-07, "loss": 0.0001, "step": 7996 }, { "epoch": 0.96, "learning_rate": 1.258589511754069e-07, "loss": 0.0002, "step": 7997 }, { "epoch": 0.96, "learning_rate": 1.25497287522604e-07, "loss": 0.0004, "step": 7998 }, { "epoch": 0.96, "learning_rate": 1.2513562386980109e-07, "loss": 0.0002, "step": 7999 }, { "epoch": 0.96, "learning_rate": 1.2477396021699818e-07, "loss": 0.0002, "step": 8000 }, { "epoch": 0.96, "learning_rate": 1.2441229656419528e-07, "loss": 0.0002, "step": 8001 }, { "epoch": 0.96, "learning_rate": 1.240506329113924e-07, "loss": 0.0001, "step": 8002 }, { "epoch": 0.96, "learning_rate": 1.236889692585895e-07, "loss": 0.5797, "step": 8003 }, { "epoch": 0.96, "learning_rate": 1.233273056057866e-07, "loss": 0.0003, "step": 8004 }, { "epoch": 0.96, "learning_rate": 1.2296564195298373e-07, "loss": 0.0003, "step": 8005 }, { "epoch": 0.96, "learning_rate": 1.2260397830018083e-07, "loss": 0.0002, "step": 8006 }, { "epoch": 0.96, "learning_rate": 1.2224231464737793e-07, "loss": 0.0002, "step": 8007 }, { "epoch": 0.96, "learning_rate": 1.2188065099457505e-07, "loss": 0.0003, "step": 8008 }, { "epoch": 0.96, "learning_rate": 1.2151898734177215e-07, "loss": 0.0002, "step": 8009 }, { "epoch": 0.96, "learning_rate": 1.2115732368896925e-07, "loss": 0.0003, "step": 8010 }, { "epoch": 0.96, "learning_rate": 1.2079566003616638e-07, "loss": 0.0002, "step": 8011 }, { "epoch": 0.96, "learning_rate": 1.2043399638336348e-07, "loss": 0.8541, "step": 8012 }, { "epoch": 0.96, "learning_rate": 1.2007233273056057e-07, "loss": 0.0002, "step": 8013 }, { "epoch": 0.96, "learning_rate": 1.197106690777577e-07, "loss": 0.0005, "step": 8014 }, { "epoch": 0.96, "learning_rate": 1.193490054249548e-07, "loss": 0.0002, "step": 8015 }, { "epoch": 0.96, "learning_rate": 1.189873417721519e-07, "loss": 0.0003, "step": 8016 }, { "epoch": 0.96, "learning_rate": 1.1862567811934902e-07, "loss": 0.3809, "step": 8017 }, { "epoch": 0.96, "learning_rate": 1.1826401446654612e-07, "loss": 0.0002, "step": 8018 }, { "epoch": 0.96, "learning_rate": 1.1790235081374322e-07, "loss": 0.7938, "step": 8019 }, { "epoch": 0.96, "learning_rate": 1.1754068716094033e-07, "loss": 0.0003, "step": 8020 }, { "epoch": 0.96, "learning_rate": 1.1717902350813744e-07, "loss": 0.0002, "step": 8021 }, { "epoch": 0.96, "learning_rate": 1.1681735985533454e-07, "loss": 0.0002, "step": 8022 }, { "epoch": 0.96, "learning_rate": 1.1645569620253166e-07, "loss": 0.0004, "step": 8023 }, { "epoch": 0.96, "learning_rate": 1.1609403254972875e-07, "loss": 0.0002, "step": 8024 }, { "epoch": 0.96, "learning_rate": 1.1573236889692585e-07, "loss": 0.0026, "step": 8025 }, { "epoch": 0.96, "learning_rate": 1.1537070524412298e-07, "loss": 0.0002, "step": 8026 }, { "epoch": 0.96, "learning_rate": 1.1500904159132008e-07, "loss": 0.0003, "step": 8027 }, { "epoch": 0.96, "learning_rate": 1.1464737793851718e-07, "loss": 0.0003, "step": 8028 }, { "epoch": 0.96, "learning_rate": 1.142857142857143e-07, "loss": 0.0002, "step": 8029 }, { "epoch": 0.96, "learning_rate": 1.139240506329114e-07, "loss": 0.0002, "step": 8030 }, { "epoch": 0.96, "learning_rate": 1.135623869801085e-07, "loss": 0.0002, "step": 8031 }, { "epoch": 0.96, "learning_rate": 1.132007233273056e-07, "loss": 0.0002, "step": 8032 }, { "epoch": 0.96, "learning_rate": 1.1283905967450272e-07, "loss": 0.0003, "step": 8033 }, { "epoch": 0.96, "learning_rate": 1.1247739602169982e-07, "loss": 0.0006, "step": 8034 }, { "epoch": 0.96, "learning_rate": 1.1211573236889692e-07, "loss": 0.0003, "step": 8035 }, { "epoch": 0.96, "learning_rate": 1.1175406871609403e-07, "loss": 0.0002, "step": 8036 }, { "epoch": 0.96, "learning_rate": 1.1139240506329114e-07, "loss": 0.0005, "step": 8037 }, { "epoch": 0.96, "learning_rate": 1.1103074141048824e-07, "loss": 0.0004, "step": 8038 }, { "epoch": 0.96, "learning_rate": 1.1066907775768536e-07, "loss": 0.0002, "step": 8039 }, { "epoch": 0.96, "learning_rate": 1.1030741410488245e-07, "loss": 0.0003, "step": 8040 }, { "epoch": 0.96, "learning_rate": 1.0994575045207955e-07, "loss": 0.0004, "step": 8041 }, { "epoch": 0.96, "learning_rate": 1.0958408679927668e-07, "loss": 0.0002, "step": 8042 }, { "epoch": 0.96, "learning_rate": 1.0922242314647378e-07, "loss": 0.0004, "step": 8043 }, { "epoch": 0.96, "learning_rate": 1.0886075949367088e-07, "loss": 0.0003, "step": 8044 }, { "epoch": 0.96, "learning_rate": 1.08499095840868e-07, "loss": 0.3872, "step": 8045 }, { "epoch": 0.96, "learning_rate": 1.081374321880651e-07, "loss": 0.0003, "step": 8046 }, { "epoch": 0.96, "learning_rate": 1.077757685352622e-07, "loss": 0.0002, "step": 8047 }, { "epoch": 0.96, "learning_rate": 1.0741410488245932e-07, "loss": 0.0002, "step": 8048 }, { "epoch": 0.96, "learning_rate": 1.0705244122965642e-07, "loss": 0.0002, "step": 8049 }, { "epoch": 0.96, "learning_rate": 1.0669077757685352e-07, "loss": 0.0003, "step": 8050 }, { "epoch": 0.96, "learning_rate": 1.0632911392405063e-07, "loss": 0.0002, "step": 8051 }, { "epoch": 0.96, "learning_rate": 1.0596745027124775e-07, "loss": 0.0002, "step": 8052 }, { "epoch": 0.97, "learning_rate": 1.0560578661844484e-07, "loss": 0.0002, "step": 8053 }, { "epoch": 0.97, "learning_rate": 1.0524412296564196e-07, "loss": 0.0004, "step": 8054 }, { "epoch": 0.97, "learning_rate": 1.0488245931283906e-07, "loss": 0.0002, "step": 8055 }, { "epoch": 0.97, "learning_rate": 1.0452079566003615e-07, "loss": 0.0003, "step": 8056 }, { "epoch": 0.97, "learning_rate": 1.0415913200723328e-07, "loss": 0.0005, "step": 8057 }, { "epoch": 0.97, "learning_rate": 1.0379746835443038e-07, "loss": 0.0005, "step": 8058 }, { "epoch": 0.97, "learning_rate": 1.0343580470162748e-07, "loss": 0.0006, "step": 8059 }, { "epoch": 0.97, "learning_rate": 1.030741410488246e-07, "loss": 0.0002, "step": 8060 }, { "epoch": 0.97, "learning_rate": 1.027124773960217e-07, "loss": 0.0003, "step": 8061 }, { "epoch": 0.97, "learning_rate": 1.023508137432188e-07, "loss": 0.0003, "step": 8062 }, { "epoch": 0.97, "learning_rate": 1.0198915009041593e-07, "loss": 0.0005, "step": 8063 }, { "epoch": 0.97, "learning_rate": 1.0162748643761302e-07, "loss": 0.0002, "step": 8064 }, { "epoch": 0.97, "learning_rate": 1.0126582278481012e-07, "loss": 0.0002, "step": 8065 }, { "epoch": 0.97, "learning_rate": 1.0090415913200724e-07, "loss": 0.0004, "step": 8066 }, { "epoch": 0.97, "learning_rate": 1.0054249547920435e-07, "loss": 0.0002, "step": 8067 }, { "epoch": 0.97, "learning_rate": 1.0018083182640145e-07, "loss": 0.0003, "step": 8068 }, { "epoch": 0.97, "learning_rate": 9.981916817359856e-08, "loss": 0.6823, "step": 8069 }, { "epoch": 0.97, "learning_rate": 9.945750452079566e-08, "loss": 0.0002, "step": 8070 }, { "epoch": 0.97, "learning_rate": 9.909584086799276e-08, "loss": 0.3229, "step": 8071 }, { "epoch": 0.97, "learning_rate": 9.873417721518988e-08, "loss": 0.0009, "step": 8072 }, { "epoch": 0.97, "learning_rate": 9.837251356238698e-08, "loss": 0.0003, "step": 8073 }, { "epoch": 0.97, "learning_rate": 9.801084990958408e-08, "loss": 0.0006, "step": 8074 }, { "epoch": 0.97, "learning_rate": 9.76491862567812e-08, "loss": 0.0001, "step": 8075 }, { "epoch": 0.97, "learning_rate": 9.72875226039783e-08, "loss": 0.0003, "step": 8076 }, { "epoch": 0.97, "learning_rate": 9.69258589511754e-08, "loss": 0.0002, "step": 8077 }, { "epoch": 0.97, "learning_rate": 9.656419529837253e-08, "loss": 0.0003, "step": 8078 }, { "epoch": 0.97, "learning_rate": 9.620253164556963e-08, "loss": 0.0002, "step": 8079 }, { "epoch": 0.97, "learning_rate": 9.584086799276672e-08, "loss": 0.0003, "step": 8080 }, { "epoch": 0.97, "learning_rate": 9.547920433996384e-08, "loss": 0.0004, "step": 8081 }, { "epoch": 0.97, "learning_rate": 9.511754068716095e-08, "loss": 0.0012, "step": 8082 }, { "epoch": 0.97, "learning_rate": 9.475587703435805e-08, "loss": 0.0002, "step": 8083 }, { "epoch": 0.97, "learning_rate": 9.439421338155516e-08, "loss": 0.0005, "step": 8084 }, { "epoch": 0.97, "learning_rate": 9.403254972875226e-08, "loss": 0.603, "step": 8085 }, { "epoch": 0.97, "learning_rate": 9.367088607594937e-08, "loss": 0.1056, "step": 8086 }, { "epoch": 0.97, "learning_rate": 9.330922242314647e-08, "loss": 0.0003, "step": 8087 }, { "epoch": 0.97, "learning_rate": 9.294755877034358e-08, "loss": 0.0004, "step": 8088 }, { "epoch": 0.97, "learning_rate": 9.258589511754069e-08, "loss": 0.0159, "step": 8089 }, { "epoch": 0.97, "learning_rate": 9.222423146473779e-08, "loss": 0.0003, "step": 8090 }, { "epoch": 0.97, "learning_rate": 9.18625678119349e-08, "loss": 0.1501, "step": 8091 }, { "epoch": 0.97, "learning_rate": 9.150090415913202e-08, "loss": 0.0019, "step": 8092 }, { "epoch": 0.97, "learning_rate": 9.113924050632911e-08, "loss": 0.0012, "step": 8093 }, { "epoch": 0.97, "learning_rate": 9.077757685352623e-08, "loss": 0.0002, "step": 8094 }, { "epoch": 0.97, "learning_rate": 9.041591320072334e-08, "loss": 0.0002, "step": 8095 }, { "epoch": 0.97, "learning_rate": 9.005424954792044e-08, "loss": 0.0006, "step": 8096 }, { "epoch": 0.97, "learning_rate": 8.969258589511754e-08, "loss": 0.0002, "step": 8097 }, { "epoch": 0.97, "learning_rate": 8.933092224231465e-08, "loss": 0.0002, "step": 8098 }, { "epoch": 0.97, "learning_rate": 8.896925858951175e-08, "loss": 0.0002, "step": 8099 }, { "epoch": 0.97, "learning_rate": 8.860759493670886e-08, "loss": 0.0077, "step": 8100 }, { "epoch": 0.97, "learning_rate": 8.824593128390597e-08, "loss": 0.0002, "step": 8101 }, { "epoch": 0.97, "learning_rate": 8.788426763110307e-08, "loss": 0.0005, "step": 8102 }, { "epoch": 0.97, "learning_rate": 8.752260397830018e-08, "loss": 0.0031, "step": 8103 }, { "epoch": 0.97, "learning_rate": 8.71609403254973e-08, "loss": 0.0028, "step": 8104 }, { "epoch": 0.97, "learning_rate": 8.679927667269439e-08, "loss": 0.0002, "step": 8105 }, { "epoch": 0.97, "learning_rate": 8.64376130198915e-08, "loss": 0.8618, "step": 8106 }, { "epoch": 0.97, "learning_rate": 8.607594936708862e-08, "loss": 0.0001, "step": 8107 }, { "epoch": 0.97, "learning_rate": 8.571428571428572e-08, "loss": 0.0002, "step": 8108 }, { "epoch": 0.97, "learning_rate": 8.535262206148283e-08, "loss": 0.0001, "step": 8109 }, { "epoch": 0.97, "learning_rate": 8.499095840867994e-08, "loss": 0.0002, "step": 8110 }, { "epoch": 0.97, "learning_rate": 8.462929475587704e-08, "loss": 0.0002, "step": 8111 }, { "epoch": 0.97, "learning_rate": 8.426763110307414e-08, "loss": 0.0001, "step": 8112 }, { "epoch": 0.97, "learning_rate": 8.390596745027125e-08, "loss": 0.0002, "step": 8113 }, { "epoch": 0.97, "learning_rate": 8.354430379746835e-08, "loss": 0.3792, "step": 8114 }, { "epoch": 0.97, "learning_rate": 8.318264014466546e-08, "loss": 0.0003, "step": 8115 }, { "epoch": 0.97, "learning_rate": 8.282097649186257e-08, "loss": 0.0002, "step": 8116 }, { "epoch": 0.97, "learning_rate": 8.245931283905967e-08, "loss": 0.0002, "step": 8117 }, { "epoch": 0.97, "learning_rate": 8.209764918625678e-08, "loss": 0.0003, "step": 8118 }, { "epoch": 0.97, "learning_rate": 8.17359855334539e-08, "loss": 0.0002, "step": 8119 }, { "epoch": 0.97, "learning_rate": 8.1374321880651e-08, "loss": 0.0002, "step": 8120 }, { "epoch": 0.97, "learning_rate": 8.10126582278481e-08, "loss": 0.0002, "step": 8121 }, { "epoch": 0.97, "learning_rate": 8.065099457504522e-08, "loss": 0.0002, "step": 8122 }, { "epoch": 0.97, "learning_rate": 8.028933092224232e-08, "loss": 0.0002, "step": 8123 }, { "epoch": 0.97, "learning_rate": 7.992766726943943e-08, "loss": 0.0005, "step": 8124 }, { "epoch": 0.97, "learning_rate": 7.956600361663654e-08, "loss": 0.0002, "step": 8125 }, { "epoch": 0.97, "learning_rate": 7.920433996383364e-08, "loss": 0.0001, "step": 8126 }, { "epoch": 0.97, "learning_rate": 7.884267631103074e-08, "loss": 0.0002, "step": 8127 }, { "epoch": 0.97, "learning_rate": 7.848101265822785e-08, "loss": 0.0002, "step": 8128 }, { "epoch": 0.97, "learning_rate": 7.811934900542495e-08, "loss": 0.0002, "step": 8129 }, { "epoch": 0.97, "learning_rate": 7.775768535262206e-08, "loss": 0.0002, "step": 8130 }, { "epoch": 0.97, "learning_rate": 7.739602169981916e-08, "loss": 0.0001, "step": 8131 }, { "epoch": 0.97, "learning_rate": 7.703435804701627e-08, "loss": 0.0002, "step": 8132 }, { "epoch": 0.97, "learning_rate": 7.667269439421338e-08, "loss": 0.0001, "step": 8133 }, { "epoch": 0.97, "learning_rate": 7.631103074141048e-08, "loss": 0.0002, "step": 8134 }, { "epoch": 0.97, "learning_rate": 7.59493670886076e-08, "loss": 0.0002, "step": 8135 }, { "epoch": 0.97, "learning_rate": 7.558770343580471e-08, "loss": 0.0006, "step": 8136 }, { "epoch": 0.98, "learning_rate": 7.52260397830018e-08, "loss": 0.0003, "step": 8137 }, { "epoch": 0.98, "learning_rate": 7.486437613019892e-08, "loss": 0.0003, "step": 8138 }, { "epoch": 0.98, "learning_rate": 7.450271247739603e-08, "loss": 0.0002, "step": 8139 }, { "epoch": 0.98, "learning_rate": 7.414104882459313e-08, "loss": 0.0053, "step": 8140 }, { "epoch": 0.98, "learning_rate": 7.377938517179024e-08, "loss": 0.0002, "step": 8141 }, { "epoch": 0.98, "learning_rate": 7.341772151898734e-08, "loss": 0.0002, "step": 8142 }, { "epoch": 0.98, "learning_rate": 7.305605786618445e-08, "loss": 0.0002, "step": 8143 }, { "epoch": 0.98, "learning_rate": 7.269439421338155e-08, "loss": 0.0003, "step": 8144 }, { "epoch": 0.98, "learning_rate": 7.233273056057866e-08, "loss": 0.0005, "step": 8145 }, { "epoch": 0.98, "learning_rate": 7.197106690777576e-08, "loss": 0.0005, "step": 8146 }, { "epoch": 0.98, "learning_rate": 7.160940325497287e-08, "loss": 0.0021, "step": 8147 }, { "epoch": 0.98, "learning_rate": 7.124773960216999e-08, "loss": 0.0002, "step": 8148 }, { "epoch": 0.98, "learning_rate": 7.088607594936708e-08, "loss": 0.0002, "step": 8149 }, { "epoch": 0.98, "learning_rate": 7.05244122965642e-08, "loss": 0.0002, "step": 8150 }, { "epoch": 0.98, "learning_rate": 7.016274864376131e-08, "loss": 0.0002, "step": 8151 }, { "epoch": 0.98, "learning_rate": 6.980108499095841e-08, "loss": 0.0001, "step": 8152 }, { "epoch": 0.98, "learning_rate": 6.943942133815552e-08, "loss": 0.0003, "step": 8153 }, { "epoch": 0.98, "learning_rate": 6.907775768535263e-08, "loss": 0.0004, "step": 8154 }, { "epoch": 0.98, "learning_rate": 6.871609403254973e-08, "loss": 0.0005, "step": 8155 }, { "epoch": 0.98, "learning_rate": 6.835443037974684e-08, "loss": 0.0002, "step": 8156 }, { "epoch": 0.98, "learning_rate": 6.799276672694394e-08, "loss": 0.5644, "step": 8157 }, { "epoch": 0.98, "learning_rate": 6.763110307414105e-08, "loss": 0.0002, "step": 8158 }, { "epoch": 0.98, "learning_rate": 6.726943942133815e-08, "loss": 0.0003, "step": 8159 }, { "epoch": 0.98, "learning_rate": 6.690777576853526e-08, "loss": 0.0003, "step": 8160 }, { "epoch": 0.98, "learning_rate": 6.654611211573236e-08, "loss": 0.0003, "step": 8161 }, { "epoch": 0.98, "learning_rate": 6.618444846292948e-08, "loss": 0.0002, "step": 8162 }, { "epoch": 0.98, "learning_rate": 6.582278481012659e-08, "loss": 0.0009, "step": 8163 }, { "epoch": 0.98, "learning_rate": 6.546112115732369e-08, "loss": 0.0003, "step": 8164 }, { "epoch": 0.98, "learning_rate": 6.50994575045208e-08, "loss": 0.0002, "step": 8165 }, { "epoch": 0.98, "learning_rate": 6.473779385171791e-08, "loss": 0.0002, "step": 8166 }, { "epoch": 0.98, "learning_rate": 6.437613019891501e-08, "loss": 0.0003, "step": 8167 }, { "epoch": 0.98, "learning_rate": 6.401446654611212e-08, "loss": 0.0002, "step": 8168 }, { "epoch": 0.98, "learning_rate": 6.365280289330923e-08, "loss": 0.0006, "step": 8169 }, { "epoch": 0.98, "learning_rate": 6.329113924050633e-08, "loss": 0.0002, "step": 8170 }, { "epoch": 0.98, "learning_rate": 6.292947558770344e-08, "loss": 0.0111, "step": 8171 }, { "epoch": 0.98, "learning_rate": 6.256781193490054e-08, "loss": 0.1747, "step": 8172 }, { "epoch": 0.98, "learning_rate": 6.220614828209764e-08, "loss": 0.006, "step": 8173 }, { "epoch": 0.98, "learning_rate": 6.184448462929475e-08, "loss": 0.0003, "step": 8174 }, { "epoch": 0.98, "learning_rate": 6.148282097649187e-08, "loss": 0.2313, "step": 8175 }, { "epoch": 0.98, "learning_rate": 6.112115732368896e-08, "loss": 0.0002, "step": 8176 }, { "epoch": 0.98, "learning_rate": 6.075949367088608e-08, "loss": 0.0002, "step": 8177 }, { "epoch": 0.98, "learning_rate": 6.039783001808319e-08, "loss": 0.0004, "step": 8178 }, { "epoch": 0.98, "learning_rate": 6.003616636528029e-08, "loss": 0.0001, "step": 8179 }, { "epoch": 0.98, "learning_rate": 5.96745027124774e-08, "loss": 0.0003, "step": 8180 }, { "epoch": 0.98, "learning_rate": 5.931283905967451e-08, "loss": 0.0002, "step": 8181 }, { "epoch": 0.98, "learning_rate": 5.895117540687161e-08, "loss": 0.0001, "step": 8182 }, { "epoch": 0.98, "learning_rate": 5.858951175406872e-08, "loss": 0.0002, "step": 8183 }, { "epoch": 0.98, "learning_rate": 5.822784810126583e-08, "loss": 0.0003, "step": 8184 }, { "epoch": 0.98, "learning_rate": 5.7866184448462927e-08, "loss": 0.0003, "step": 8185 }, { "epoch": 0.98, "learning_rate": 5.750452079566004e-08, "loss": 0.0002, "step": 8186 }, { "epoch": 0.98, "learning_rate": 5.714285714285715e-08, "loss": 0.0006, "step": 8187 }, { "epoch": 0.98, "learning_rate": 5.678119349005425e-08, "loss": 0.0017, "step": 8188 }, { "epoch": 0.98, "learning_rate": 5.641952983725136e-08, "loss": 0.0002, "step": 8189 }, { "epoch": 0.98, "learning_rate": 5.605786618444846e-08, "loss": 0.0002, "step": 8190 }, { "epoch": 0.98, "learning_rate": 5.569620253164557e-08, "loss": 0.5021, "step": 8191 }, { "epoch": 0.98, "learning_rate": 5.533453887884268e-08, "loss": 0.0002, "step": 8192 }, { "epoch": 0.98, "learning_rate": 5.4972875226039777e-08, "loss": 0.0002, "step": 8193 }, { "epoch": 0.98, "learning_rate": 5.461121157323689e-08, "loss": 0.0003, "step": 8194 }, { "epoch": 0.98, "learning_rate": 5.4249547920434e-08, "loss": 0.0006, "step": 8195 }, { "epoch": 0.98, "learning_rate": 5.38878842676311e-08, "loss": 0.0002, "step": 8196 }, { "epoch": 0.98, "learning_rate": 5.352622061482821e-08, "loss": 0.0001, "step": 8197 }, { "epoch": 0.98, "learning_rate": 5.316455696202532e-08, "loss": 0.0003, "step": 8198 }, { "epoch": 0.98, "learning_rate": 5.280289330922242e-08, "loss": 0.0002, "step": 8199 }, { "epoch": 0.98, "learning_rate": 5.244122965641953e-08, "loss": 0.0003, "step": 8200 }, { "epoch": 0.98, "learning_rate": 5.207956600361664e-08, "loss": 0.0003, "step": 8201 }, { "epoch": 0.98, "learning_rate": 5.171790235081374e-08, "loss": 0.0002, "step": 8202 }, { "epoch": 0.98, "learning_rate": 5.135623869801085e-08, "loss": 0.0003, "step": 8203 }, { "epoch": 0.98, "learning_rate": 5.099457504520796e-08, "loss": 0.0002, "step": 8204 }, { "epoch": 0.98, "learning_rate": 5.063291139240506e-08, "loss": 0.0002, "step": 8205 }, { "epoch": 0.98, "learning_rate": 5.0271247739602174e-08, "loss": 0.0002, "step": 8206 }, { "epoch": 0.98, "learning_rate": 4.990958408679928e-08, "loss": 0.0002, "step": 8207 }, { "epoch": 0.98, "learning_rate": 4.954792043399638e-08, "loss": 0.0003, "step": 8208 }, { "epoch": 0.98, "learning_rate": 4.918625678119349e-08, "loss": 0.002, "step": 8209 }, { "epoch": 0.98, "learning_rate": 4.88245931283906e-08, "loss": 0.437, "step": 8210 }, { "epoch": 0.98, "learning_rate": 4.84629294755877e-08, "loss": 0.0003, "step": 8211 }, { "epoch": 0.98, "learning_rate": 4.810126582278481e-08, "loss": 0.0002, "step": 8212 }, { "epoch": 0.98, "learning_rate": 4.773960216998192e-08, "loss": 0.0001, "step": 8213 }, { "epoch": 0.98, "learning_rate": 4.7377938517179024e-08, "loss": 0.0002, "step": 8214 }, { "epoch": 0.98, "learning_rate": 4.701627486437613e-08, "loss": 0.0003, "step": 8215 }, { "epoch": 0.98, "learning_rate": 4.6654611211573235e-08, "loss": 0.0003, "step": 8216 }, { "epoch": 0.98, "learning_rate": 4.6292947558770347e-08, "loss": 0.0002, "step": 8217 }, { "epoch": 0.98, "learning_rate": 4.593128390596745e-08, "loss": 0.0003, "step": 8218 }, { "epoch": 0.98, "learning_rate": 4.556962025316456e-08, "loss": 0.0002, "step": 8219 }, { "epoch": 0.99, "learning_rate": 4.520795660036167e-08, "loss": 0.0003, "step": 8220 }, { "epoch": 0.99, "learning_rate": 4.484629294755877e-08, "loss": 0.0003, "step": 8221 }, { "epoch": 0.99, "learning_rate": 4.4484629294755874e-08, "loss": 0.0004, "step": 8222 }, { "epoch": 0.99, "learning_rate": 4.4122965641952986e-08, "loss": 0.0002, "step": 8223 }, { "epoch": 0.99, "learning_rate": 4.376130198915009e-08, "loss": 0.0002, "step": 8224 }, { "epoch": 0.99, "learning_rate": 4.3399638336347197e-08, "loss": 0.0003, "step": 8225 }, { "epoch": 0.99, "learning_rate": 4.303797468354431e-08, "loss": 0.0002, "step": 8226 }, { "epoch": 0.99, "learning_rate": 4.2676311030741414e-08, "loss": 0.203, "step": 8227 }, { "epoch": 0.99, "learning_rate": 4.231464737793852e-08, "loss": 0.0402, "step": 8228 }, { "epoch": 0.99, "learning_rate": 4.1952983725135625e-08, "loss": 0.0002, "step": 8229 }, { "epoch": 0.99, "learning_rate": 4.159132007233273e-08, "loss": 0.0002, "step": 8230 }, { "epoch": 0.99, "learning_rate": 4.1229656419529836e-08, "loss": 0.0002, "step": 8231 }, { "epoch": 0.99, "learning_rate": 4.086799276672695e-08, "loss": 0.0002, "step": 8232 }, { "epoch": 0.99, "learning_rate": 4.050632911392405e-08, "loss": 0.7401, "step": 8233 }, { "epoch": 0.99, "learning_rate": 4.014466546112116e-08, "loss": 0.0264, "step": 8234 }, { "epoch": 0.99, "learning_rate": 3.978300180831827e-08, "loss": 0.0002, "step": 8235 }, { "epoch": 0.99, "learning_rate": 3.942133815551537e-08, "loss": 0.0002, "step": 8236 }, { "epoch": 0.99, "learning_rate": 3.9059674502712475e-08, "loss": 0.0003, "step": 8237 }, { "epoch": 0.99, "learning_rate": 3.869801084990958e-08, "loss": 0.0002, "step": 8238 }, { "epoch": 0.99, "learning_rate": 3.833634719710669e-08, "loss": 0.0002, "step": 8239 }, { "epoch": 0.99, "learning_rate": 3.79746835443038e-08, "loss": 0.0002, "step": 8240 }, { "epoch": 0.99, "learning_rate": 3.76130198915009e-08, "loss": 0.6404, "step": 8241 }, { "epoch": 0.99, "learning_rate": 3.7251356238698015e-08, "loss": 0.0002, "step": 8242 }, { "epoch": 0.99, "learning_rate": 3.688969258589512e-08, "loss": 1.0389, "step": 8243 }, { "epoch": 0.99, "learning_rate": 3.6528028933092226e-08, "loss": 0.0002, "step": 8244 }, { "epoch": 0.99, "learning_rate": 3.616636528028933e-08, "loss": 0.0001, "step": 8245 }, { "epoch": 0.99, "learning_rate": 3.580470162748644e-08, "loss": 0.0196, "step": 8246 }, { "epoch": 0.99, "learning_rate": 3.544303797468354e-08, "loss": 0.0002, "step": 8247 }, { "epoch": 0.99, "learning_rate": 3.5081374321880654e-08, "loss": 0.0003, "step": 8248 }, { "epoch": 0.99, "learning_rate": 3.471971066907776e-08, "loss": 0.0008, "step": 8249 }, { "epoch": 0.99, "learning_rate": 3.4358047016274865e-08, "loss": 0.7288, "step": 8250 }, { "epoch": 0.99, "learning_rate": 3.399638336347197e-08, "loss": 0.0708, "step": 8251 }, { "epoch": 0.99, "learning_rate": 3.3634719710669076e-08, "loss": 0.0002, "step": 8252 }, { "epoch": 0.99, "learning_rate": 3.327305605786618e-08, "loss": 0.0002, "step": 8253 }, { "epoch": 0.99, "learning_rate": 3.2911392405063294e-08, "loss": 0.0002, "step": 8254 }, { "epoch": 0.99, "learning_rate": 3.25497287522604e-08, "loss": 0.0002, "step": 8255 }, { "epoch": 0.99, "learning_rate": 3.2188065099457505e-08, "loss": 0.2207, "step": 8256 }, { "epoch": 0.99, "learning_rate": 3.1826401446654617e-08, "loss": 0.8379, "step": 8257 }, { "epoch": 0.99, "learning_rate": 3.146473779385172e-08, "loss": 0.002, "step": 8258 }, { "epoch": 0.99, "learning_rate": 3.110307414104882e-08, "loss": 0.0003, "step": 8259 }, { "epoch": 0.99, "learning_rate": 3.074141048824593e-08, "loss": 0.0003, "step": 8260 }, { "epoch": 0.99, "learning_rate": 3.037974683544304e-08, "loss": 0.8002, "step": 8261 }, { "epoch": 0.99, "learning_rate": 3.0018083182640144e-08, "loss": 0.0003, "step": 8262 }, { "epoch": 0.99, "learning_rate": 2.9656419529837256e-08, "loss": 0.0005, "step": 8263 }, { "epoch": 0.99, "learning_rate": 2.929475587703436e-08, "loss": 0.0002, "step": 8264 }, { "epoch": 0.99, "learning_rate": 2.8933092224231463e-08, "loss": 0.0002, "step": 8265 }, { "epoch": 0.99, "learning_rate": 2.8571428571428575e-08, "loss": 0.0005, "step": 8266 }, { "epoch": 0.99, "learning_rate": 2.820976491862568e-08, "loss": 0.0002, "step": 8267 }, { "epoch": 0.99, "learning_rate": 2.7848101265822786e-08, "loss": 0.0003, "step": 8268 }, { "epoch": 0.99, "learning_rate": 2.7486437613019888e-08, "loss": 0.0002, "step": 8269 }, { "epoch": 0.99, "learning_rate": 2.7124773960217e-08, "loss": 0.002, "step": 8270 }, { "epoch": 0.99, "learning_rate": 2.6763110307414106e-08, "loss": 0.0002, "step": 8271 }, { "epoch": 0.99, "learning_rate": 2.640144665461121e-08, "loss": 0.0004, "step": 8272 }, { "epoch": 0.99, "learning_rate": 2.603978300180832e-08, "loss": 0.0005, "step": 8273 }, { "epoch": 0.99, "learning_rate": 2.5678119349005425e-08, "loss": 0.0002, "step": 8274 }, { "epoch": 0.99, "learning_rate": 2.531645569620253e-08, "loss": 0.0002, "step": 8275 }, { "epoch": 0.99, "learning_rate": 2.495479204339964e-08, "loss": 0.0002, "step": 8276 }, { "epoch": 0.99, "learning_rate": 2.4593128390596745e-08, "loss": 0.0001, "step": 8277 }, { "epoch": 0.99, "learning_rate": 2.423146473779385e-08, "loss": 0.0008, "step": 8278 }, { "epoch": 0.99, "learning_rate": 2.386980108499096e-08, "loss": 0.0002, "step": 8279 }, { "epoch": 0.99, "learning_rate": 2.3508137432188065e-08, "loss": 0.0003, "step": 8280 }, { "epoch": 0.99, "learning_rate": 2.3146473779385173e-08, "loss": 0.0002, "step": 8281 }, { "epoch": 0.99, "learning_rate": 2.278481012658228e-08, "loss": 0.0002, "step": 8282 }, { "epoch": 0.99, "learning_rate": 2.2423146473779384e-08, "loss": 0.0003, "step": 8283 }, { "epoch": 0.99, "learning_rate": 2.2061482820976493e-08, "loss": 0.0002, "step": 8284 }, { "epoch": 0.99, "learning_rate": 2.1699819168173598e-08, "loss": 0.0002, "step": 8285 }, { "epoch": 0.99, "learning_rate": 2.1338155515370707e-08, "loss": 0.0002, "step": 8286 }, { "epoch": 0.99, "learning_rate": 2.0976491862567812e-08, "loss": 0.0002, "step": 8287 }, { "epoch": 0.99, "learning_rate": 2.0614828209764918e-08, "loss": 0.0004, "step": 8288 }, { "epoch": 0.99, "learning_rate": 2.0253164556962027e-08, "loss": 0.0003, "step": 8289 }, { "epoch": 0.99, "learning_rate": 1.9891500904159135e-08, "loss": 0.0002, "step": 8290 }, { "epoch": 0.99, "learning_rate": 1.9529837251356237e-08, "loss": 0.0002, "step": 8291 }, { "epoch": 0.99, "learning_rate": 1.9168173598553346e-08, "loss": 0.0002, "step": 8292 }, { "epoch": 0.99, "learning_rate": 1.880650994575045e-08, "loss": 0.0004, "step": 8293 }, { "epoch": 0.99, "learning_rate": 1.844484629294756e-08, "loss": 0.0003, "step": 8294 }, { "epoch": 0.99, "learning_rate": 1.8083182640144666e-08, "loss": 0.0002, "step": 8295 }, { "epoch": 0.99, "learning_rate": 1.772151898734177e-08, "loss": 0.0003, "step": 8296 }, { "epoch": 0.99, "learning_rate": 1.735985533453888e-08, "loss": 0.0003, "step": 8297 }, { "epoch": 0.99, "learning_rate": 1.6998191681735985e-08, "loss": 0.0002, "step": 8298 }, { "epoch": 0.99, "learning_rate": 1.663652802893309e-08, "loss": 0.0004, "step": 8299 }, { "epoch": 0.99, "learning_rate": 1.62748643761302e-08, "loss": 0.0002, "step": 8300 }, { "epoch": 0.99, "learning_rate": 1.5913200723327308e-08, "loss": 0.0002, "step": 8301 }, { "epoch": 0.99, "learning_rate": 1.555153707052441e-08, "loss": 0.0002, "step": 8302 }, { "epoch": 0.99, "learning_rate": 1.518987341772152e-08, "loss": 0.0002, "step": 8303 }, { "epoch": 1.0, "learning_rate": 1.4828209764918628e-08, "loss": 0.0002, "step": 8304 }, { "epoch": 1.0, "learning_rate": 1.4466546112115732e-08, "loss": 0.0002, "step": 8305 }, { "epoch": 1.0, "learning_rate": 1.410488245931284e-08, "loss": 0.002, "step": 8306 }, { "epoch": 1.0, "learning_rate": 1.3743218806509944e-08, "loss": 0.0005, "step": 8307 }, { "epoch": 1.0, "learning_rate": 1.3381555153707053e-08, "loss": 0.0002, "step": 8308 }, { "epoch": 1.0, "learning_rate": 1.301989150090416e-08, "loss": 0.0448, "step": 8309 }, { "epoch": 1.0, "learning_rate": 1.2658227848101265e-08, "loss": 0.0002, "step": 8310 }, { "epoch": 1.0, "learning_rate": 1.2296564195298372e-08, "loss": 0.0002, "step": 8311 }, { "epoch": 1.0, "learning_rate": 1.193490054249548e-08, "loss": 0.0003, "step": 8312 }, { "epoch": 1.0, "learning_rate": 1.1573236889692587e-08, "loss": 0.0003, "step": 8313 }, { "epoch": 1.0, "learning_rate": 1.1211573236889692e-08, "loss": 0.0004, "step": 8314 }, { "epoch": 1.0, "learning_rate": 1.0849909584086799e-08, "loss": 0.0002, "step": 8315 }, { "epoch": 1.0, "learning_rate": 1.0488245931283906e-08, "loss": 0.1257, "step": 8316 }, { "epoch": 1.0, "learning_rate": 1.0126582278481013e-08, "loss": 0.0003, "step": 8317 }, { "epoch": 1.0, "learning_rate": 9.764918625678119e-09, "loss": 0.0004, "step": 8318 }, { "epoch": 1.0, "learning_rate": 9.403254972875226e-09, "loss": 0.0002, "step": 8319 }, { "epoch": 1.0, "learning_rate": 9.041591320072333e-09, "loss": 0.001, "step": 8320 }, { "epoch": 1.0, "learning_rate": 8.67992766726944e-09, "loss": 0.0002, "step": 8321 }, { "epoch": 1.0, "learning_rate": 8.318264014466545e-09, "loss": 0.0001, "step": 8322 }, { "epoch": 1.0, "learning_rate": 7.956600361663654e-09, "loss": 0.0004, "step": 8323 }, { "epoch": 1.0, "learning_rate": 7.59493670886076e-09, "loss": 0.0002, "step": 8324 }, { "epoch": 1.0, "learning_rate": 7.233273056057866e-09, "loss": 0.0002, "step": 8325 }, { "epoch": 1.0, "learning_rate": 6.871609403254972e-09, "loss": 0.0002, "step": 8326 }, { "epoch": 1.0, "learning_rate": 6.50994575045208e-09, "loss": 0.0002, "step": 8327 }, { "epoch": 1.0, "learning_rate": 6.148282097649186e-09, "loss": 0.0002, "step": 8328 }, { "epoch": 1.0, "learning_rate": 5.786618444846293e-09, "loss": 0.0002, "step": 8329 }, { "epoch": 1.0, "learning_rate": 5.4249547920433996e-09, "loss": 0.0002, "step": 8330 }, { "epoch": 1.0, "learning_rate": 5.063291139240507e-09, "loss": 0.0002, "step": 8331 }, { "epoch": 1.0, "learning_rate": 4.701627486437613e-09, "loss": 0.0002, "step": 8332 }, { "epoch": 1.0, "learning_rate": 4.33996383363472e-09, "loss": 0.0002, "step": 8333 }, { "epoch": 1.0, "learning_rate": 3.978300180831827e-09, "loss": 0.0004, "step": 8334 }, { "epoch": 1.0, "learning_rate": 3.616636528028933e-09, "loss": 0.0007, "step": 8335 }, { "epoch": 1.0, "learning_rate": 3.25497287522604e-09, "loss": 0.0001, "step": 8336 }, { "epoch": 1.0, "learning_rate": 2.8933092224231467e-09, "loss": 0.0002, "step": 8337 }, { "epoch": 1.0, "learning_rate": 2.5316455696202533e-09, "loss": 0.0002, "step": 8338 }, { "epoch": 1.0, "learning_rate": 2.16998191681736e-09, "loss": 0.5259, "step": 8339 }, { "epoch": 1.0, "learning_rate": 1.8083182640144665e-09, "loss": 0.0002, "step": 8340 }, { "epoch": 1.0, "learning_rate": 1.4466546112115733e-09, "loss": 0.0002, "step": 8341 }, { "epoch": 1.0, "learning_rate": 1.08499095840868e-09, "loss": 0.0007, "step": 8342 }, { "epoch": 1.0, "learning_rate": 7.233273056057867e-10, "loss": 0.0003, "step": 8343 }, { "epoch": 1.0, "learning_rate": 3.6166365280289333e-10, "loss": 0.0003, "step": 8344 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 1.2189, "step": 8345 }, { "epoch": 1.0, "eval_accuracy": 0.9873576992210905, "eval_loss": 0.06538490206003189, "eval_runtime": 302.9264, "eval_samples_per_second": 55.096, "eval_steps_per_second": 0.862, "step": 8345 } ], "logging_steps": 1, "max_steps": 8345, "num_train_epochs": 1, "save_steps": 500, "total_flos": 1.699967003679588e+16, "trial_name": null, "trial_params": null }