{ "best_metric": null, "best_model_checkpoint": null, "epoch": 2.0, "eval_steps": 500, "global_step": 10586, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 1.5723270440251573e-08, "loss": 2.7808, "step": 1 }, { "epoch": 0.0, "learning_rate": 3.1446540880503146e-08, "loss": 2.6527, "step": 2 }, { "epoch": 0.0, "learning_rate": 4.716981132075472e-08, "loss": 2.6386, "step": 3 }, { "epoch": 0.0, "learning_rate": 6.289308176100629e-08, "loss": 2.7258, "step": 4 }, { "epoch": 0.0, "learning_rate": 7.861635220125787e-08, "loss": 2.7652, "step": 5 }, { "epoch": 0.0, "learning_rate": 9.433962264150944e-08, "loss": 3.1064, "step": 6 }, { "epoch": 0.0, "learning_rate": 1.1006289308176101e-07, "loss": 2.5104, "step": 7 }, { "epoch": 0.0, "learning_rate": 1.2578616352201258e-07, "loss": 2.7967, "step": 8 }, { "epoch": 0.0, "learning_rate": 1.4150943396226417e-07, "loss": 2.5806, "step": 9 }, { "epoch": 0.0, "learning_rate": 1.5723270440251575e-07, "loss": 2.777, "step": 10 }, { "epoch": 0.0, "learning_rate": 1.729559748427673e-07, "loss": 2.6001, "step": 11 }, { "epoch": 0.0, "learning_rate": 1.886792452830189e-07, "loss": 2.6179, "step": 12 }, { "epoch": 0.0, "learning_rate": 2.0440251572327044e-07, "loss": 2.6668, "step": 13 }, { "epoch": 0.0, "learning_rate": 2.2012578616352203e-07, "loss": 2.6982, "step": 14 }, { "epoch": 0.0, "learning_rate": 2.358490566037736e-07, "loss": 2.6537, "step": 15 }, { "epoch": 0.0, "learning_rate": 2.5157232704402517e-07, "loss": 2.7373, "step": 16 }, { "epoch": 0.0, "learning_rate": 2.672955974842768e-07, "loss": 2.8534, "step": 17 }, { "epoch": 0.0, "learning_rate": 2.8301886792452833e-07, "loss": 2.6937, "step": 18 }, { "epoch": 0.0, "learning_rate": 2.987421383647799e-07, "loss": 2.481, "step": 19 }, { "epoch": 0.0, "learning_rate": 3.144654088050315e-07, "loss": 2.5344, "step": 20 }, { "epoch": 0.0, "learning_rate": 3.3018867924528305e-07, "loss": 2.5091, "step": 21 }, { "epoch": 0.0, "learning_rate": 3.459119496855346e-07, "loss": 2.5396, "step": 22 }, { "epoch": 0.0, "learning_rate": 3.6163522012578617e-07, "loss": 2.5352, "step": 23 }, { "epoch": 0.0, "learning_rate": 3.773584905660378e-07, "loss": 2.5305, "step": 24 }, { "epoch": 0.0, "learning_rate": 3.9308176100628933e-07, "loss": 2.1951, "step": 25 }, { "epoch": 0.0, "learning_rate": 4.088050314465409e-07, "loss": 2.1388, "step": 26 }, { "epoch": 0.01, "learning_rate": 4.2452830188679244e-07, "loss": 2.1283, "step": 27 }, { "epoch": 0.01, "learning_rate": 4.4025157232704405e-07, "loss": 2.2242, "step": 28 }, { "epoch": 0.01, "learning_rate": 4.559748427672956e-07, "loss": 2.0627, "step": 29 }, { "epoch": 0.01, "learning_rate": 4.716981132075472e-07, "loss": 2.0812, "step": 30 }, { "epoch": 0.01, "learning_rate": 4.874213836477988e-07, "loss": 2.059, "step": 31 }, { "epoch": 0.01, "learning_rate": 5.031446540880503e-07, "loss": 1.8717, "step": 32 }, { "epoch": 0.01, "learning_rate": 5.18867924528302e-07, "loss": 1.551, "step": 33 }, { "epoch": 0.01, "learning_rate": 5.345911949685535e-07, "loss": 1.4032, "step": 34 }, { "epoch": 0.01, "learning_rate": 5.503144654088051e-07, "loss": 1.3448, "step": 35 }, { "epoch": 0.01, "learning_rate": 5.660377358490567e-07, "loss": 1.2512, "step": 36 }, { "epoch": 0.01, "learning_rate": 5.817610062893082e-07, "loss": 1.2342, "step": 37 }, { "epoch": 0.01, "learning_rate": 5.974842767295598e-07, "loss": 1.3001, "step": 38 }, { "epoch": 0.01, "learning_rate": 6.132075471698113e-07, "loss": 1.2434, "step": 39 }, { "epoch": 0.01, "learning_rate": 6.28930817610063e-07, "loss": 1.1507, "step": 40 }, { "epoch": 0.01, "learning_rate": 6.446540880503145e-07, "loss": 1.2928, "step": 41 }, { "epoch": 0.01, "learning_rate": 6.603773584905661e-07, "loss": 1.1141, "step": 42 }, { "epoch": 0.01, "learning_rate": 6.761006289308177e-07, "loss": 1.2072, "step": 43 }, { "epoch": 0.01, "learning_rate": 6.918238993710692e-07, "loss": 1.0, "step": 44 }, { "epoch": 0.01, "learning_rate": 7.075471698113208e-07, "loss": 0.9924, "step": 45 }, { "epoch": 0.01, "learning_rate": 7.232704402515723e-07, "loss": 0.9186, "step": 46 }, { "epoch": 0.01, "learning_rate": 7.389937106918239e-07, "loss": 0.9249, "step": 47 }, { "epoch": 0.01, "learning_rate": 7.547169811320755e-07, "loss": 1.0979, "step": 48 }, { "epoch": 0.01, "learning_rate": 7.704402515723271e-07, "loss": 0.9668, "step": 49 }, { "epoch": 0.01, "learning_rate": 7.861635220125787e-07, "loss": 0.8887, "step": 50 }, { "epoch": 0.01, "learning_rate": 8.018867924528302e-07, "loss": 0.848, "step": 51 }, { "epoch": 0.01, "learning_rate": 8.176100628930818e-07, "loss": 0.9248, "step": 52 }, { "epoch": 0.01, "learning_rate": 8.333333333333333e-07, "loss": 0.7705, "step": 53 }, { "epoch": 0.01, "learning_rate": 8.490566037735849e-07, "loss": 0.8206, "step": 54 }, { "epoch": 0.01, "learning_rate": 8.647798742138364e-07, "loss": 0.7784, "step": 55 }, { "epoch": 0.01, "learning_rate": 8.805031446540881e-07, "loss": 0.9494, "step": 56 }, { "epoch": 0.01, "learning_rate": 8.962264150943397e-07, "loss": 0.9923, "step": 57 }, { "epoch": 0.01, "learning_rate": 9.119496855345912e-07, "loss": 0.8169, "step": 58 }, { "epoch": 0.01, "learning_rate": 9.276729559748428e-07, "loss": 0.7043, "step": 59 }, { "epoch": 0.01, "learning_rate": 9.433962264150944e-07, "loss": 0.8339, "step": 60 }, { "epoch": 0.01, "learning_rate": 9.59119496855346e-07, "loss": 1.0004, "step": 61 }, { "epoch": 0.01, "learning_rate": 9.748427672955975e-07, "loss": 0.7242, "step": 62 }, { "epoch": 0.01, "learning_rate": 9.90566037735849e-07, "loss": 0.7855, "step": 63 }, { "epoch": 0.01, "learning_rate": 1.0062893081761007e-06, "loss": 0.9545, "step": 64 }, { "epoch": 0.01, "learning_rate": 1.0220125786163524e-06, "loss": 0.6996, "step": 65 }, { "epoch": 0.01, "learning_rate": 1.037735849056604e-06, "loss": 0.8018, "step": 66 }, { "epoch": 0.01, "learning_rate": 1.0534591194968555e-06, "loss": 0.8343, "step": 67 }, { "epoch": 0.01, "learning_rate": 1.069182389937107e-06, "loss": 0.712, "step": 68 }, { "epoch": 0.01, "learning_rate": 1.0849056603773587e-06, "loss": 1.0122, "step": 69 }, { "epoch": 0.01, "learning_rate": 1.1006289308176102e-06, "loss": 0.698, "step": 70 }, { "epoch": 0.01, "learning_rate": 1.1163522012578618e-06, "loss": 0.8088, "step": 71 }, { "epoch": 0.01, "learning_rate": 1.1320754716981133e-06, "loss": 0.7575, "step": 72 }, { "epoch": 0.01, "learning_rate": 1.1477987421383649e-06, "loss": 0.7593, "step": 73 }, { "epoch": 0.01, "learning_rate": 1.1635220125786164e-06, "loss": 0.7502, "step": 74 }, { "epoch": 0.01, "learning_rate": 1.179245283018868e-06, "loss": 0.8196, "step": 75 }, { "epoch": 0.01, "learning_rate": 1.1949685534591195e-06, "loss": 0.648, "step": 76 }, { "epoch": 0.01, "learning_rate": 1.210691823899371e-06, "loss": 0.886, "step": 77 }, { "epoch": 0.01, "learning_rate": 1.2264150943396227e-06, "loss": 0.7335, "step": 78 }, { "epoch": 0.01, "learning_rate": 1.2421383647798742e-06, "loss": 0.6533, "step": 79 }, { "epoch": 0.02, "learning_rate": 1.257861635220126e-06, "loss": 0.8097, "step": 80 }, { "epoch": 0.02, "learning_rate": 1.2735849056603775e-06, "loss": 0.7698, "step": 81 }, { "epoch": 0.02, "learning_rate": 1.289308176100629e-06, "loss": 0.6607, "step": 82 }, { "epoch": 0.02, "learning_rate": 1.3050314465408807e-06, "loss": 0.8117, "step": 83 }, { "epoch": 0.02, "learning_rate": 1.3207547169811322e-06, "loss": 0.9845, "step": 84 }, { "epoch": 0.02, "learning_rate": 1.3364779874213838e-06, "loss": 0.7628, "step": 85 }, { "epoch": 0.02, "learning_rate": 1.3522012578616353e-06, "loss": 0.6588, "step": 86 }, { "epoch": 0.02, "learning_rate": 1.3679245283018869e-06, "loss": 0.7068, "step": 87 }, { "epoch": 0.02, "learning_rate": 1.3836477987421384e-06, "loss": 0.7452, "step": 88 }, { "epoch": 0.02, "learning_rate": 1.39937106918239e-06, "loss": 0.7519, "step": 89 }, { "epoch": 0.02, "learning_rate": 1.4150943396226415e-06, "loss": 0.682, "step": 90 }, { "epoch": 0.02, "learning_rate": 1.430817610062893e-06, "loss": 0.6806, "step": 91 }, { "epoch": 0.02, "learning_rate": 1.4465408805031447e-06, "loss": 0.7548, "step": 92 }, { "epoch": 0.02, "learning_rate": 1.4622641509433962e-06, "loss": 0.8695, "step": 93 }, { "epoch": 0.02, "learning_rate": 1.4779874213836478e-06, "loss": 0.6032, "step": 94 }, { "epoch": 0.02, "learning_rate": 1.4937106918238995e-06, "loss": 0.8407, "step": 95 }, { "epoch": 0.02, "learning_rate": 1.509433962264151e-06, "loss": 0.737, "step": 96 }, { "epoch": 0.02, "learning_rate": 1.5251572327044027e-06, "loss": 0.8484, "step": 97 }, { "epoch": 0.02, "learning_rate": 1.5408805031446542e-06, "loss": 0.8983, "step": 98 }, { "epoch": 0.02, "learning_rate": 1.5566037735849058e-06, "loss": 0.6957, "step": 99 }, { "epoch": 0.02, "learning_rate": 1.5723270440251573e-06, "loss": 0.7259, "step": 100 }, { "epoch": 0.02, "learning_rate": 1.5880503144654089e-06, "loss": 0.6688, "step": 101 }, { "epoch": 0.02, "learning_rate": 1.6037735849056604e-06, "loss": 0.6857, "step": 102 }, { "epoch": 0.02, "learning_rate": 1.619496855345912e-06, "loss": 0.9028, "step": 103 }, { "epoch": 0.02, "learning_rate": 1.6352201257861635e-06, "loss": 0.78, "step": 104 }, { "epoch": 0.02, "learning_rate": 1.650943396226415e-06, "loss": 0.9109, "step": 105 }, { "epoch": 0.02, "learning_rate": 1.6666666666666667e-06, "loss": 0.7513, "step": 106 }, { "epoch": 0.02, "learning_rate": 1.6823899371069182e-06, "loss": 0.7688, "step": 107 }, { "epoch": 0.02, "learning_rate": 1.6981132075471698e-06, "loss": 0.6129, "step": 108 }, { "epoch": 0.02, "learning_rate": 1.7138364779874213e-06, "loss": 0.8979, "step": 109 }, { "epoch": 0.02, "learning_rate": 1.7295597484276729e-06, "loss": 0.9362, "step": 110 }, { "epoch": 0.02, "learning_rate": 1.7452830188679247e-06, "loss": 0.9093, "step": 111 }, { "epoch": 0.02, "learning_rate": 1.7610062893081762e-06, "loss": 0.9452, "step": 112 }, { "epoch": 0.02, "learning_rate": 1.7767295597484278e-06, "loss": 0.7018, "step": 113 }, { "epoch": 0.02, "learning_rate": 1.7924528301886793e-06, "loss": 0.5615, "step": 114 }, { "epoch": 0.02, "learning_rate": 1.8081761006289309e-06, "loss": 0.729, "step": 115 }, { "epoch": 0.02, "learning_rate": 1.8238993710691824e-06, "loss": 0.8445, "step": 116 }, { "epoch": 0.02, "learning_rate": 1.839622641509434e-06, "loss": 0.8604, "step": 117 }, { "epoch": 0.02, "learning_rate": 1.8553459119496855e-06, "loss": 0.861, "step": 118 }, { "epoch": 0.02, "learning_rate": 1.871069182389937e-06, "loss": 0.9081, "step": 119 }, { "epoch": 0.02, "learning_rate": 1.8867924528301889e-06, "loss": 0.7913, "step": 120 }, { "epoch": 0.02, "learning_rate": 1.9025157232704406e-06, "loss": 0.6733, "step": 121 }, { "epoch": 0.02, "learning_rate": 1.918238993710692e-06, "loss": 0.6843, "step": 122 }, { "epoch": 0.02, "learning_rate": 1.9339622641509438e-06, "loss": 0.883, "step": 123 }, { "epoch": 0.02, "learning_rate": 1.949685534591195e-06, "loss": 0.6127, "step": 124 }, { "epoch": 0.02, "learning_rate": 1.965408805031447e-06, "loss": 0.7255, "step": 125 }, { "epoch": 0.02, "learning_rate": 1.981132075471698e-06, "loss": 0.5711, "step": 126 }, { "epoch": 0.02, "learning_rate": 1.99685534591195e-06, "loss": 0.8474, "step": 127 }, { "epoch": 0.02, "learning_rate": 2.0125786163522013e-06, "loss": 0.64, "step": 128 }, { "epoch": 0.02, "learning_rate": 2.028301886792453e-06, "loss": 0.9183, "step": 129 }, { "epoch": 0.02, "learning_rate": 2.044025157232705e-06, "loss": 0.743, "step": 130 }, { "epoch": 0.02, "learning_rate": 2.059748427672956e-06, "loss": 0.7033, "step": 131 }, { "epoch": 0.02, "learning_rate": 2.075471698113208e-06, "loss": 0.6887, "step": 132 }, { "epoch": 0.03, "learning_rate": 2.0911949685534593e-06, "loss": 0.7111, "step": 133 }, { "epoch": 0.03, "learning_rate": 2.106918238993711e-06, "loss": 0.8199, "step": 134 }, { "epoch": 0.03, "learning_rate": 2.1226415094339624e-06, "loss": 0.7741, "step": 135 }, { "epoch": 0.03, "learning_rate": 2.138364779874214e-06, "loss": 0.6956, "step": 136 }, { "epoch": 0.03, "learning_rate": 2.1540880503144655e-06, "loss": 0.8314, "step": 137 }, { "epoch": 0.03, "learning_rate": 2.1698113207547173e-06, "loss": 0.7197, "step": 138 }, { "epoch": 0.03, "learning_rate": 2.1855345911949687e-06, "loss": 0.635, "step": 139 }, { "epoch": 0.03, "learning_rate": 2.2012578616352204e-06, "loss": 0.6295, "step": 140 }, { "epoch": 0.03, "learning_rate": 2.2169811320754718e-06, "loss": 0.9109, "step": 141 }, { "epoch": 0.03, "learning_rate": 2.2327044025157235e-06, "loss": 0.6702, "step": 142 }, { "epoch": 0.03, "learning_rate": 2.248427672955975e-06, "loss": 0.9534, "step": 143 }, { "epoch": 0.03, "learning_rate": 2.2641509433962266e-06, "loss": 0.7433, "step": 144 }, { "epoch": 0.03, "learning_rate": 2.2798742138364784e-06, "loss": 0.7291, "step": 145 }, { "epoch": 0.03, "learning_rate": 2.2955974842767298e-06, "loss": 0.6996, "step": 146 }, { "epoch": 0.03, "learning_rate": 2.3113207547169815e-06, "loss": 0.6238, "step": 147 }, { "epoch": 0.03, "learning_rate": 2.327044025157233e-06, "loss": 0.7415, "step": 148 }, { "epoch": 0.03, "learning_rate": 2.3427672955974846e-06, "loss": 0.7203, "step": 149 }, { "epoch": 0.03, "learning_rate": 2.358490566037736e-06, "loss": 0.7741, "step": 150 }, { "epoch": 0.03, "learning_rate": 2.3742138364779878e-06, "loss": 0.8278, "step": 151 }, { "epoch": 0.03, "learning_rate": 2.389937106918239e-06, "loss": 0.7715, "step": 152 }, { "epoch": 0.03, "learning_rate": 2.405660377358491e-06, "loss": 0.7308, "step": 153 }, { "epoch": 0.03, "learning_rate": 2.421383647798742e-06, "loss": 0.6523, "step": 154 }, { "epoch": 0.03, "learning_rate": 2.437106918238994e-06, "loss": 0.8163, "step": 155 }, { "epoch": 0.03, "learning_rate": 2.4528301886792453e-06, "loss": 0.7879, "step": 156 }, { "epoch": 0.03, "learning_rate": 2.468553459119497e-06, "loss": 0.5718, "step": 157 }, { "epoch": 0.03, "learning_rate": 2.4842767295597484e-06, "loss": 0.7462, "step": 158 }, { "epoch": 0.03, "learning_rate": 2.5e-06, "loss": 0.6578, "step": 159 }, { "epoch": 0.03, "learning_rate": 2.515723270440252e-06, "loss": 0.6862, "step": 160 }, { "epoch": 0.03, "learning_rate": 2.5314465408805033e-06, "loss": 0.8248, "step": 161 }, { "epoch": 0.03, "learning_rate": 2.547169811320755e-06, "loss": 0.6853, "step": 162 }, { "epoch": 0.03, "learning_rate": 2.5628930817610064e-06, "loss": 0.5841, "step": 163 }, { "epoch": 0.03, "learning_rate": 2.578616352201258e-06, "loss": 0.7693, "step": 164 }, { "epoch": 0.03, "learning_rate": 2.5943396226415095e-06, "loss": 0.6727, "step": 165 }, { "epoch": 0.03, "learning_rate": 2.6100628930817613e-06, "loss": 0.7299, "step": 166 }, { "epoch": 0.03, "learning_rate": 2.6257861635220127e-06, "loss": 0.7157, "step": 167 }, { "epoch": 0.03, "learning_rate": 2.6415094339622644e-06, "loss": 0.7802, "step": 168 }, { "epoch": 0.03, "learning_rate": 2.6572327044025158e-06, "loss": 0.5971, "step": 169 }, { "epoch": 0.03, "learning_rate": 2.6729559748427675e-06, "loss": 0.6473, "step": 170 }, { "epoch": 0.03, "learning_rate": 2.688679245283019e-06, "loss": 0.7696, "step": 171 }, { "epoch": 0.03, "learning_rate": 2.7044025157232706e-06, "loss": 0.7576, "step": 172 }, { "epoch": 0.03, "learning_rate": 2.720125786163522e-06, "loss": 0.8093, "step": 173 }, { "epoch": 0.03, "learning_rate": 2.7358490566037738e-06, "loss": 0.6103, "step": 174 }, { "epoch": 0.03, "learning_rate": 2.7515723270440255e-06, "loss": 0.7083, "step": 175 }, { "epoch": 0.03, "learning_rate": 2.767295597484277e-06, "loss": 0.6427, "step": 176 }, { "epoch": 0.03, "learning_rate": 2.7830188679245286e-06, "loss": 0.6642, "step": 177 }, { "epoch": 0.03, "learning_rate": 2.79874213836478e-06, "loss": 0.6687, "step": 178 }, { "epoch": 0.03, "learning_rate": 2.8144654088050318e-06, "loss": 0.7135, "step": 179 }, { "epoch": 0.03, "learning_rate": 2.830188679245283e-06, "loss": 0.6518, "step": 180 }, { "epoch": 0.03, "learning_rate": 2.845911949685535e-06, "loss": 0.6362, "step": 181 }, { "epoch": 0.03, "learning_rate": 2.861635220125786e-06, "loss": 0.8378, "step": 182 }, { "epoch": 0.03, "learning_rate": 2.877358490566038e-06, "loss": 0.6202, "step": 183 }, { "epoch": 0.03, "learning_rate": 2.8930817610062893e-06, "loss": 0.5915, "step": 184 }, { "epoch": 0.03, "learning_rate": 2.908805031446541e-06, "loss": 0.8871, "step": 185 }, { "epoch": 0.04, "learning_rate": 2.9245283018867924e-06, "loss": 0.7785, "step": 186 }, { "epoch": 0.04, "learning_rate": 2.940251572327044e-06, "loss": 0.685, "step": 187 }, { "epoch": 0.04, "learning_rate": 2.9559748427672955e-06, "loss": 0.6042, "step": 188 }, { "epoch": 0.04, "learning_rate": 2.9716981132075473e-06, "loss": 0.6649, "step": 189 }, { "epoch": 0.04, "learning_rate": 2.987421383647799e-06, "loss": 0.8161, "step": 190 }, { "epoch": 0.04, "learning_rate": 3.0031446540880504e-06, "loss": 0.5941, "step": 191 }, { "epoch": 0.04, "learning_rate": 3.018867924528302e-06, "loss": 0.6455, "step": 192 }, { "epoch": 0.04, "learning_rate": 3.0345911949685535e-06, "loss": 0.751, "step": 193 }, { "epoch": 0.04, "learning_rate": 3.0503144654088053e-06, "loss": 0.7229, "step": 194 }, { "epoch": 0.04, "learning_rate": 3.0660377358490567e-06, "loss": 0.5967, "step": 195 }, { "epoch": 0.04, "learning_rate": 3.0817610062893084e-06, "loss": 0.6686, "step": 196 }, { "epoch": 0.04, "learning_rate": 3.0974842767295598e-06, "loss": 0.8315, "step": 197 }, { "epoch": 0.04, "learning_rate": 3.1132075471698115e-06, "loss": 0.6755, "step": 198 }, { "epoch": 0.04, "learning_rate": 3.128930817610063e-06, "loss": 0.7015, "step": 199 }, { "epoch": 0.04, "learning_rate": 3.1446540880503146e-06, "loss": 0.7074, "step": 200 }, { "epoch": 0.04, "learning_rate": 3.160377358490566e-06, "loss": 0.6937, "step": 201 }, { "epoch": 0.04, "learning_rate": 3.1761006289308178e-06, "loss": 0.7865, "step": 202 }, { "epoch": 0.04, "learning_rate": 3.191823899371069e-06, "loss": 0.7407, "step": 203 }, { "epoch": 0.04, "learning_rate": 3.207547169811321e-06, "loss": 0.6649, "step": 204 }, { "epoch": 0.04, "learning_rate": 3.2232704402515726e-06, "loss": 0.5785, "step": 205 }, { "epoch": 0.04, "learning_rate": 3.238993710691824e-06, "loss": 0.7513, "step": 206 }, { "epoch": 0.04, "learning_rate": 3.2547169811320758e-06, "loss": 0.839, "step": 207 }, { "epoch": 0.04, "learning_rate": 3.270440251572327e-06, "loss": 0.7405, "step": 208 }, { "epoch": 0.04, "learning_rate": 3.286163522012579e-06, "loss": 0.6009, "step": 209 }, { "epoch": 0.04, "learning_rate": 3.30188679245283e-06, "loss": 0.6251, "step": 210 }, { "epoch": 0.04, "learning_rate": 3.317610062893082e-06, "loss": 0.6663, "step": 211 }, { "epoch": 0.04, "learning_rate": 3.3333333333333333e-06, "loss": 0.6254, "step": 212 }, { "epoch": 0.04, "learning_rate": 3.349056603773585e-06, "loss": 0.6426, "step": 213 }, { "epoch": 0.04, "learning_rate": 3.3647798742138364e-06, "loss": 0.6757, "step": 214 }, { "epoch": 0.04, "learning_rate": 3.380503144654088e-06, "loss": 0.492, "step": 215 }, { "epoch": 0.04, "learning_rate": 3.3962264150943395e-06, "loss": 0.657, "step": 216 }, { "epoch": 0.04, "learning_rate": 3.4119496855345913e-06, "loss": 0.6219, "step": 217 }, { "epoch": 0.04, "learning_rate": 3.4276729559748427e-06, "loss": 0.7962, "step": 218 }, { "epoch": 0.04, "learning_rate": 3.4433962264150944e-06, "loss": 0.7135, "step": 219 }, { "epoch": 0.04, "learning_rate": 3.4591194968553458e-06, "loss": 0.5905, "step": 220 }, { "epoch": 0.04, "learning_rate": 3.4748427672955975e-06, "loss": 0.5959, "step": 221 }, { "epoch": 0.04, "learning_rate": 3.4905660377358493e-06, "loss": 0.743, "step": 222 }, { "epoch": 0.04, "learning_rate": 3.5062893081761007e-06, "loss": 0.7392, "step": 223 }, { "epoch": 0.04, "learning_rate": 3.5220125786163524e-06, "loss": 0.5915, "step": 224 }, { "epoch": 0.04, "learning_rate": 3.5377358490566038e-06, "loss": 0.5773, "step": 225 }, { "epoch": 0.04, "learning_rate": 3.5534591194968555e-06, "loss": 0.5546, "step": 226 }, { "epoch": 0.04, "learning_rate": 3.569182389937107e-06, "loss": 0.6948, "step": 227 }, { "epoch": 0.04, "learning_rate": 3.5849056603773586e-06, "loss": 0.7665, "step": 228 }, { "epoch": 0.04, "learning_rate": 3.60062893081761e-06, "loss": 0.5934, "step": 229 }, { "epoch": 0.04, "learning_rate": 3.6163522012578618e-06, "loss": 0.6811, "step": 230 }, { "epoch": 0.04, "learning_rate": 3.632075471698113e-06, "loss": 0.533, "step": 231 }, { "epoch": 0.04, "learning_rate": 3.647798742138365e-06, "loss": 0.6793, "step": 232 }, { "epoch": 0.04, "learning_rate": 3.6635220125786162e-06, "loss": 0.7059, "step": 233 }, { "epoch": 0.04, "learning_rate": 3.679245283018868e-06, "loss": 0.6229, "step": 234 }, { "epoch": 0.04, "learning_rate": 3.6949685534591193e-06, "loss": 0.7526, "step": 235 }, { "epoch": 0.04, "learning_rate": 3.710691823899371e-06, "loss": 0.6228, "step": 236 }, { "epoch": 0.04, "learning_rate": 3.726415094339623e-06, "loss": 0.6163, "step": 237 }, { "epoch": 0.04, "learning_rate": 3.742138364779874e-06, "loss": 0.6097, "step": 238 }, { "epoch": 0.05, "learning_rate": 3.7578616352201264e-06, "loss": 0.7193, "step": 239 }, { "epoch": 0.05, "learning_rate": 3.7735849056603777e-06, "loss": 0.5285, "step": 240 }, { "epoch": 0.05, "learning_rate": 3.7893081761006295e-06, "loss": 0.6127, "step": 241 }, { "epoch": 0.05, "learning_rate": 3.8050314465408813e-06, "loss": 0.6106, "step": 242 }, { "epoch": 0.05, "learning_rate": 3.820754716981133e-06, "loss": 0.6562, "step": 243 }, { "epoch": 0.05, "learning_rate": 3.836477987421384e-06, "loss": 0.6662, "step": 244 }, { "epoch": 0.05, "learning_rate": 3.852201257861636e-06, "loss": 0.6857, "step": 245 }, { "epoch": 0.05, "learning_rate": 3.8679245283018875e-06, "loss": 0.5149, "step": 246 }, { "epoch": 0.05, "learning_rate": 3.883647798742139e-06, "loss": 0.6026, "step": 247 }, { "epoch": 0.05, "learning_rate": 3.89937106918239e-06, "loss": 0.6271, "step": 248 }, { "epoch": 0.05, "learning_rate": 3.915094339622642e-06, "loss": 0.7472, "step": 249 }, { "epoch": 0.05, "learning_rate": 3.930817610062894e-06, "loss": 0.5357, "step": 250 }, { "epoch": 0.05, "learning_rate": 3.9465408805031455e-06, "loss": 0.7588, "step": 251 }, { "epoch": 0.05, "learning_rate": 3.962264150943396e-06, "loss": 0.6952, "step": 252 }, { "epoch": 0.05, "learning_rate": 3.977987421383648e-06, "loss": 0.6225, "step": 253 }, { "epoch": 0.05, "learning_rate": 3.9937106918239e-06, "loss": 0.6105, "step": 254 }, { "epoch": 0.05, "learning_rate": 4.009433962264152e-06, "loss": 0.7512, "step": 255 }, { "epoch": 0.05, "learning_rate": 4.025157232704403e-06, "loss": 0.5791, "step": 256 }, { "epoch": 0.05, "learning_rate": 4.040880503144654e-06, "loss": 0.7109, "step": 257 }, { "epoch": 0.05, "learning_rate": 4.056603773584906e-06, "loss": 0.7638, "step": 258 }, { "epoch": 0.05, "learning_rate": 4.072327044025158e-06, "loss": 0.6339, "step": 259 }, { "epoch": 0.05, "learning_rate": 4.08805031446541e-06, "loss": 0.6621, "step": 260 }, { "epoch": 0.05, "learning_rate": 4.103773584905661e-06, "loss": 0.8603, "step": 261 }, { "epoch": 0.05, "learning_rate": 4.119496855345912e-06, "loss": 0.6604, "step": 262 }, { "epoch": 0.05, "learning_rate": 4.135220125786164e-06, "loss": 0.7141, "step": 263 }, { "epoch": 0.05, "learning_rate": 4.150943396226416e-06, "loss": 0.591, "step": 264 }, { "epoch": 0.05, "learning_rate": 4.166666666666667e-06, "loss": 0.6969, "step": 265 }, { "epoch": 0.05, "learning_rate": 4.182389937106919e-06, "loss": 0.5364, "step": 266 }, { "epoch": 0.05, "learning_rate": 4.19811320754717e-06, "loss": 0.5523, "step": 267 }, { "epoch": 0.05, "learning_rate": 4.213836477987422e-06, "loss": 0.6758, "step": 268 }, { "epoch": 0.05, "learning_rate": 4.229559748427673e-06, "loss": 0.5448, "step": 269 }, { "epoch": 0.05, "learning_rate": 4.245283018867925e-06, "loss": 0.4931, "step": 270 }, { "epoch": 0.05, "learning_rate": 4.261006289308177e-06, "loss": 0.7019, "step": 271 }, { "epoch": 0.05, "learning_rate": 4.276729559748428e-06, "loss": 0.6192, "step": 272 }, { "epoch": 0.05, "learning_rate": 4.29245283018868e-06, "loss": 0.6548, "step": 273 }, { "epoch": 0.05, "learning_rate": 4.308176100628931e-06, "loss": 0.5934, "step": 274 }, { "epoch": 0.05, "learning_rate": 4.323899371069183e-06, "loss": 0.6678, "step": 275 }, { "epoch": 0.05, "learning_rate": 4.339622641509435e-06, "loss": 0.6678, "step": 276 }, { "epoch": 0.05, "learning_rate": 4.355345911949686e-06, "loss": 0.7895, "step": 277 }, { "epoch": 0.05, "learning_rate": 4.371069182389937e-06, "loss": 0.5792, "step": 278 }, { "epoch": 0.05, "learning_rate": 4.386792452830189e-06, "loss": 0.7584, "step": 279 }, { "epoch": 0.05, "learning_rate": 4.402515723270441e-06, "loss": 0.6075, "step": 280 }, { "epoch": 0.05, "learning_rate": 4.418238993710693e-06, "loss": 0.649, "step": 281 }, { "epoch": 0.05, "learning_rate": 4.4339622641509435e-06, "loss": 0.6591, "step": 282 }, { "epoch": 0.05, "learning_rate": 4.449685534591195e-06, "loss": 0.5151, "step": 283 }, { "epoch": 0.05, "learning_rate": 4.465408805031447e-06, "loss": 0.5724, "step": 284 }, { "epoch": 0.05, "learning_rate": 4.481132075471699e-06, "loss": 0.7841, "step": 285 }, { "epoch": 0.05, "learning_rate": 4.49685534591195e-06, "loss": 0.4508, "step": 286 }, { "epoch": 0.05, "learning_rate": 4.5125786163522015e-06, "loss": 0.6647, "step": 287 }, { "epoch": 0.05, "learning_rate": 4.528301886792453e-06, "loss": 0.5574, "step": 288 }, { "epoch": 0.05, "learning_rate": 4.544025157232705e-06, "loss": 0.643, "step": 289 }, { "epoch": 0.05, "learning_rate": 4.559748427672957e-06, "loss": 0.6141, "step": 290 }, { "epoch": 0.05, "learning_rate": 4.575471698113208e-06, "loss": 0.5661, "step": 291 }, { "epoch": 0.06, "learning_rate": 4.5911949685534595e-06, "loss": 0.6946, "step": 292 }, { "epoch": 0.06, "learning_rate": 4.606918238993711e-06, "loss": 0.6618, "step": 293 }, { "epoch": 0.06, "learning_rate": 4.622641509433963e-06, "loss": 0.6165, "step": 294 }, { "epoch": 0.06, "learning_rate": 4.638364779874214e-06, "loss": 0.7274, "step": 295 }, { "epoch": 0.06, "learning_rate": 4.654088050314466e-06, "loss": 0.5521, "step": 296 }, { "epoch": 0.06, "learning_rate": 4.6698113207547175e-06, "loss": 0.6427, "step": 297 }, { "epoch": 0.06, "learning_rate": 4.685534591194969e-06, "loss": 0.7399, "step": 298 }, { "epoch": 0.06, "learning_rate": 4.70125786163522e-06, "loss": 0.7546, "step": 299 }, { "epoch": 0.06, "learning_rate": 4.716981132075472e-06, "loss": 0.7118, "step": 300 }, { "epoch": 0.06, "learning_rate": 4.732704402515724e-06, "loss": 0.615, "step": 301 }, { "epoch": 0.06, "learning_rate": 4.7484276729559755e-06, "loss": 0.5798, "step": 302 }, { "epoch": 0.06, "learning_rate": 4.764150943396227e-06, "loss": 0.5389, "step": 303 }, { "epoch": 0.06, "learning_rate": 4.779874213836478e-06, "loss": 0.6527, "step": 304 }, { "epoch": 0.06, "learning_rate": 4.79559748427673e-06, "loss": 0.6666, "step": 305 }, { "epoch": 0.06, "learning_rate": 4.811320754716982e-06, "loss": 0.6153, "step": 306 }, { "epoch": 0.06, "learning_rate": 4.8270440251572335e-06, "loss": 0.6605, "step": 307 }, { "epoch": 0.06, "learning_rate": 4.842767295597484e-06, "loss": 0.7642, "step": 308 }, { "epoch": 0.06, "learning_rate": 4.858490566037736e-06, "loss": 0.5562, "step": 309 }, { "epoch": 0.06, "learning_rate": 4.874213836477988e-06, "loss": 0.5084, "step": 310 }, { "epoch": 0.06, "learning_rate": 4.88993710691824e-06, "loss": 0.7175, "step": 311 }, { "epoch": 0.06, "learning_rate": 4.905660377358491e-06, "loss": 0.628, "step": 312 }, { "epoch": 0.06, "learning_rate": 4.921383647798742e-06, "loss": 0.8491, "step": 313 }, { "epoch": 0.06, "learning_rate": 4.937106918238994e-06, "loss": 0.6645, "step": 314 }, { "epoch": 0.06, "learning_rate": 4.952830188679246e-06, "loss": 0.6514, "step": 315 }, { "epoch": 0.06, "learning_rate": 4.968553459119497e-06, "loss": 0.6911, "step": 316 }, { "epoch": 0.06, "learning_rate": 4.984276729559749e-06, "loss": 0.6577, "step": 317 }, { "epoch": 0.06, "learning_rate": 5e-06, "loss": 0.7123, "step": 318 }, { "epoch": 0.06, "learning_rate": 4.9999998829859444e-06, "loss": 0.4807, "step": 319 }, { "epoch": 0.06, "learning_rate": 4.999999531943786e-06, "loss": 0.6456, "step": 320 }, { "epoch": 0.06, "learning_rate": 4.9999989468735575e-06, "loss": 0.7427, "step": 321 }, { "epoch": 0.06, "learning_rate": 4.9999981277753164e-06, "loss": 0.6699, "step": 322 }, { "epoch": 0.06, "learning_rate": 4.999997074649137e-06, "loss": 0.6876, "step": 323 }, { "epoch": 0.06, "learning_rate": 4.999995787495118e-06, "loss": 0.4868, "step": 324 }, { "epoch": 0.06, "learning_rate": 4.999994266313382e-06, "loss": 0.5276, "step": 325 }, { "epoch": 0.06, "learning_rate": 4.999992511104068e-06, "loss": 0.5843, "step": 326 }, { "epoch": 0.06, "learning_rate": 4.9999905218673425e-06, "loss": 0.6036, "step": 327 }, { "epoch": 0.06, "learning_rate": 4.9999882986033925e-06, "loss": 0.578, "step": 328 }, { "epoch": 0.06, "learning_rate": 4.999985841312424e-06, "loss": 0.647, "step": 329 }, { "epoch": 0.06, "learning_rate": 4.999983149994668e-06, "loss": 0.5754, "step": 330 }, { "epoch": 0.06, "learning_rate": 4.999980224650377e-06, "loss": 0.6436, "step": 331 }, { "epoch": 0.06, "learning_rate": 4.999977065279824e-06, "loss": 0.5558, "step": 332 }, { "epoch": 0.06, "learning_rate": 4.999973671883305e-06, "loss": 0.6244, "step": 333 }, { "epoch": 0.06, "learning_rate": 4.999970044461138e-06, "loss": 0.8004, "step": 334 }, { "epoch": 0.06, "learning_rate": 4.999966183013663e-06, "loss": 0.7306, "step": 335 }, { "epoch": 0.06, "learning_rate": 4.9999620875412394e-06, "loss": 0.6686, "step": 336 }, { "epoch": 0.06, "learning_rate": 4.999957758044252e-06, "loss": 0.6553, "step": 337 }, { "epoch": 0.06, "learning_rate": 4.999953194523106e-06, "loss": 0.6828, "step": 338 }, { "epoch": 0.06, "learning_rate": 4.999948396978229e-06, "loss": 0.6025, "step": 339 }, { "epoch": 0.06, "learning_rate": 4.9999433654100705e-06, "loss": 0.6103, "step": 340 }, { "epoch": 0.06, "learning_rate": 4.999938099819099e-06, "loss": 0.6742, "step": 341 }, { "epoch": 0.06, "learning_rate": 4.999932600205811e-06, "loss": 0.5984, "step": 342 }, { "epoch": 0.06, "learning_rate": 4.999926866570718e-06, "loss": 0.6543, "step": 343 }, { "epoch": 0.06, "learning_rate": 4.999920898914359e-06, "loss": 0.4975, "step": 344 }, { "epoch": 0.07, "learning_rate": 4.999914697237292e-06, "loss": 0.6401, "step": 345 }, { "epoch": 0.07, "learning_rate": 4.9999082615400966e-06, "loss": 0.6208, "step": 346 }, { "epoch": 0.07, "learning_rate": 4.999901591823377e-06, "loss": 0.714, "step": 347 }, { "epoch": 0.07, "learning_rate": 4.9998946880877556e-06, "loss": 0.6239, "step": 348 }, { "epoch": 0.07, "learning_rate": 4.999887550333881e-06, "loss": 0.6707, "step": 349 }, { "epoch": 0.07, "learning_rate": 4.999880178562418e-06, "loss": 0.5439, "step": 350 }, { "epoch": 0.07, "learning_rate": 4.9998725727740604e-06, "loss": 0.6828, "step": 351 }, { "epoch": 0.07, "learning_rate": 4.999864732969518e-06, "loss": 0.6417, "step": 352 }, { "epoch": 0.07, "learning_rate": 4.999856659149526e-06, "loss": 0.6107, "step": 353 }, { "epoch": 0.07, "learning_rate": 4.9998483513148385e-06, "loss": 0.6431, "step": 354 }, { "epoch": 0.07, "learning_rate": 4.999839809466234e-06, "loss": 0.8053, "step": 355 }, { "epoch": 0.07, "learning_rate": 4.999831033604513e-06, "loss": 0.7545, "step": 356 }, { "epoch": 0.07, "learning_rate": 4.9998220237304955e-06, "loss": 0.7174, "step": 357 }, { "epoch": 0.07, "learning_rate": 4.999812779845026e-06, "loss": 0.6643, "step": 358 }, { "epoch": 0.07, "learning_rate": 4.99980330194897e-06, "loss": 0.5694, "step": 359 }, { "epoch": 0.07, "learning_rate": 4.999793590043214e-06, "loss": 0.5768, "step": 360 }, { "epoch": 0.07, "learning_rate": 4.999783644128667e-06, "loss": 0.6122, "step": 361 }, { "epoch": 0.07, "learning_rate": 4.99977346420626e-06, "loss": 0.5873, "step": 362 }, { "epoch": 0.07, "learning_rate": 4.999763050276948e-06, "loss": 0.6007, "step": 363 }, { "epoch": 0.07, "learning_rate": 4.999752402341702e-06, "loss": 0.6457, "step": 364 }, { "epoch": 0.07, "learning_rate": 4.999741520401523e-06, "loss": 0.7708, "step": 365 }, { "epoch": 0.07, "learning_rate": 4.999730404457427e-06, "loss": 0.6037, "step": 366 }, { "epoch": 0.07, "learning_rate": 4.9997190545104556e-06, "loss": 0.4465, "step": 367 }, { "epoch": 0.07, "learning_rate": 4.999707470561671e-06, "loss": 0.6266, "step": 368 }, { "epoch": 0.07, "learning_rate": 4.999695652612156e-06, "loss": 0.7344, "step": 369 }, { "epoch": 0.07, "learning_rate": 4.99968360066302e-06, "loss": 0.583, "step": 370 }, { "epoch": 0.07, "learning_rate": 4.999671314715389e-06, "loss": 0.6694, "step": 371 }, { "epoch": 0.07, "learning_rate": 4.9996587947704146e-06, "loss": 0.6115, "step": 372 }, { "epoch": 0.07, "learning_rate": 4.999646040829268e-06, "loss": 0.4649, "step": 373 }, { "epoch": 0.07, "learning_rate": 4.999633052893142e-06, "loss": 0.6749, "step": 374 }, { "epoch": 0.07, "learning_rate": 4.999619830963254e-06, "loss": 0.5075, "step": 375 }, { "epoch": 0.07, "learning_rate": 4.999606375040841e-06, "loss": 0.6569, "step": 376 }, { "epoch": 0.07, "learning_rate": 4.9995926851271625e-06, "loss": 0.7136, "step": 377 }, { "epoch": 0.07, "learning_rate": 4.999578761223501e-06, "loss": 0.6973, "step": 378 }, { "epoch": 0.07, "learning_rate": 4.999564603331159e-06, "loss": 0.6069, "step": 379 }, { "epoch": 0.07, "learning_rate": 4.999550211451462e-06, "loss": 0.5818, "step": 380 }, { "epoch": 0.07, "learning_rate": 4.999535585585757e-06, "loss": 0.5332, "step": 381 }, { "epoch": 0.07, "learning_rate": 4.999520725735415e-06, "loss": 0.562, "step": 382 }, { "epoch": 0.07, "learning_rate": 4.999505631901824e-06, "loss": 0.5682, "step": 383 }, { "epoch": 0.07, "learning_rate": 4.9994903040864e-06, "loss": 0.6315, "step": 384 }, { "epoch": 0.07, "learning_rate": 4.999474742290576e-06, "loss": 0.6317, "step": 385 }, { "epoch": 0.07, "learning_rate": 4.999458946515808e-06, "loss": 0.6515, "step": 386 }, { "epoch": 0.07, "learning_rate": 4.999442916763576e-06, "loss": 0.5692, "step": 387 }, { "epoch": 0.07, "learning_rate": 4.9994266530353805e-06, "loss": 0.5318, "step": 388 }, { "epoch": 0.07, "learning_rate": 4.999410155332745e-06, "loss": 0.6741, "step": 389 }, { "epoch": 0.07, "learning_rate": 4.9993934236572115e-06, "loss": 0.7101, "step": 390 }, { "epoch": 0.07, "learning_rate": 4.999376458010347e-06, "loss": 0.8358, "step": 391 }, { "epoch": 0.07, "learning_rate": 4.999359258393742e-06, "loss": 0.5429, "step": 392 }, { "epoch": 0.07, "learning_rate": 4.999341824809003e-06, "loss": 0.6504, "step": 393 }, { "epoch": 0.07, "learning_rate": 4.999324157257765e-06, "loss": 0.9476, "step": 394 }, { "epoch": 0.07, "learning_rate": 4.999306255741679e-06, "loss": 0.6352, "step": 395 }, { "epoch": 0.07, "learning_rate": 4.999288120262424e-06, "loss": 0.6281, "step": 396 }, { "epoch": 0.08, "learning_rate": 4.999269750821695e-06, "loss": 0.6593, "step": 397 }, { "epoch": 0.08, "learning_rate": 4.999251147421213e-06, "loss": 0.6777, "step": 398 }, { "epoch": 0.08, "learning_rate": 4.9992323100627194e-06, "loss": 0.7051, "step": 399 }, { "epoch": 0.08, "learning_rate": 4.999213238747976e-06, "loss": 0.632, "step": 400 }, { "epoch": 0.08, "learning_rate": 4.999193933478771e-06, "loss": 0.5528, "step": 401 }, { "epoch": 0.08, "learning_rate": 4.999174394256909e-06, "loss": 0.759, "step": 402 }, { "epoch": 0.08, "learning_rate": 4.999154621084221e-06, "loss": 0.5059, "step": 403 }, { "epoch": 0.08, "learning_rate": 4.999134613962557e-06, "loss": 0.4467, "step": 404 }, { "epoch": 0.08, "learning_rate": 4.999114372893789e-06, "loss": 0.6034, "step": 405 }, { "epoch": 0.08, "learning_rate": 4.9990938978798135e-06, "loss": 0.5509, "step": 406 }, { "epoch": 0.08, "learning_rate": 4.999073188922547e-06, "loss": 0.7033, "step": 407 }, { "epoch": 0.08, "learning_rate": 4.999052246023925e-06, "loss": 0.5676, "step": 408 }, { "epoch": 0.08, "learning_rate": 4.999031069185913e-06, "loss": 0.5682, "step": 409 }, { "epoch": 0.08, "learning_rate": 4.99900965841049e-06, "loss": 0.766, "step": 410 }, { "epoch": 0.08, "learning_rate": 4.9989880136996614e-06, "loss": 0.5491, "step": 411 }, { "epoch": 0.08, "learning_rate": 4.998966135055454e-06, "loss": 0.619, "step": 412 }, { "epoch": 0.08, "learning_rate": 4.998944022479913e-06, "loss": 0.5, "step": 413 }, { "epoch": 0.08, "learning_rate": 4.9989216759751124e-06, "loss": 0.5593, "step": 414 }, { "epoch": 0.08, "learning_rate": 4.9988990955431416e-06, "loss": 0.5751, "step": 415 }, { "epoch": 0.08, "learning_rate": 4.998876281186115e-06, "loss": 0.6146, "step": 416 }, { "epoch": 0.08, "learning_rate": 4.998853232906168e-06, "loss": 0.523, "step": 417 }, { "epoch": 0.08, "learning_rate": 4.998829950705458e-06, "loss": 0.6343, "step": 418 }, { "epoch": 0.08, "learning_rate": 4.998806434586166e-06, "loss": 0.6576, "step": 419 }, { "epoch": 0.08, "learning_rate": 4.9987826845504916e-06, "loss": 0.5941, "step": 420 }, { "epoch": 0.08, "learning_rate": 4.998758700600659e-06, "loss": 0.6927, "step": 421 }, { "epoch": 0.08, "learning_rate": 4.998734482738913e-06, "loss": 0.5648, "step": 422 }, { "epoch": 0.08, "learning_rate": 4.99871003096752e-06, "loss": 0.686, "step": 423 }, { "epoch": 0.08, "learning_rate": 4.998685345288771e-06, "loss": 0.5096, "step": 424 }, { "epoch": 0.08, "learning_rate": 4.9986604257049745e-06, "loss": 0.6102, "step": 425 }, { "epoch": 0.08, "learning_rate": 4.998635272218464e-06, "loss": 0.5855, "step": 426 }, { "epoch": 0.08, "learning_rate": 4.9986098848315955e-06, "loss": 0.5897, "step": 427 }, { "epoch": 0.08, "learning_rate": 4.998584263546744e-06, "loss": 0.5423, "step": 428 }, { "epoch": 0.08, "learning_rate": 4.9985584083663085e-06, "loss": 0.6114, "step": 429 }, { "epoch": 0.08, "learning_rate": 4.99853231929271e-06, "loss": 0.59, "step": 430 }, { "epoch": 0.08, "learning_rate": 4.9985059963283886e-06, "loss": 0.5653, "step": 431 }, { "epoch": 0.08, "learning_rate": 4.998479439475811e-06, "loss": 0.4309, "step": 432 }, { "epoch": 0.08, "learning_rate": 4.9984526487374615e-06, "loss": 0.5366, "step": 433 }, { "epoch": 0.08, "learning_rate": 4.998425624115849e-06, "loss": 0.5943, "step": 434 }, { "epoch": 0.08, "learning_rate": 4.998398365613503e-06, "loss": 0.6533, "step": 435 }, { "epoch": 0.08, "learning_rate": 4.998370873232974e-06, "loss": 0.618, "step": 436 }, { "epoch": 0.08, "learning_rate": 4.998343146976837e-06, "loss": 0.5628, "step": 437 }, { "epoch": 0.08, "learning_rate": 4.9983151868476884e-06, "loss": 0.5537, "step": 438 }, { "epoch": 0.08, "learning_rate": 4.998286992848144e-06, "loss": 0.6095, "step": 439 }, { "epoch": 0.08, "learning_rate": 4.998258564980842e-06, "loss": 0.6659, "step": 440 }, { "epoch": 0.08, "learning_rate": 4.998229903248446e-06, "loss": 0.5004, "step": 441 }, { "epoch": 0.08, "learning_rate": 4.998201007653638e-06, "loss": 0.625, "step": 442 }, { "epoch": 0.08, "learning_rate": 4.998171878199124e-06, "loss": 0.675, "step": 443 }, { "epoch": 0.08, "learning_rate": 4.998142514887628e-06, "loss": 0.5824, "step": 444 }, { "epoch": 0.08, "learning_rate": 4.998112917721902e-06, "loss": 0.6107, "step": 445 }, { "epoch": 0.08, "learning_rate": 4.998083086704715e-06, "loss": 0.7933, "step": 446 }, { "epoch": 0.08, "learning_rate": 4.998053021838859e-06, "loss": 0.7569, "step": 447 }, { "epoch": 0.08, "learning_rate": 4.99802272312715e-06, "loss": 0.7477, "step": 448 }, { "epoch": 0.08, "learning_rate": 4.997992190572422e-06, "loss": 0.715, "step": 449 }, { "epoch": 0.09, "learning_rate": 4.9979614241775365e-06, "loss": 0.5806, "step": 450 }, { "epoch": 0.09, "learning_rate": 4.997930423945371e-06, "loss": 0.6414, "step": 451 }, { "epoch": 0.09, "learning_rate": 4.997899189878828e-06, "loss": 0.7399, "step": 452 }, { "epoch": 0.09, "learning_rate": 4.997867721980832e-06, "loss": 0.5778, "step": 453 }, { "epoch": 0.09, "learning_rate": 4.997836020254328e-06, "loss": 0.6263, "step": 454 }, { "epoch": 0.09, "learning_rate": 4.997804084702284e-06, "loss": 0.7672, "step": 455 }, { "epoch": 0.09, "learning_rate": 4.99777191532769e-06, "loss": 0.4963, "step": 456 }, { "epoch": 0.09, "learning_rate": 4.997739512133556e-06, "loss": 0.5704, "step": 457 }, { "epoch": 0.09, "learning_rate": 4.997706875122916e-06, "loss": 0.4742, "step": 458 }, { "epoch": 0.09, "learning_rate": 4.997674004298826e-06, "loss": 0.7426, "step": 459 }, { "epoch": 0.09, "learning_rate": 4.997640899664362e-06, "loss": 0.6105, "step": 460 }, { "epoch": 0.09, "learning_rate": 4.997607561222624e-06, "loss": 0.6304, "step": 461 }, { "epoch": 0.09, "learning_rate": 4.997573988976732e-06, "loss": 0.6571, "step": 462 }, { "epoch": 0.09, "learning_rate": 4.997540182929829e-06, "loss": 0.8022, "step": 463 }, { "epoch": 0.09, "learning_rate": 4.997506143085079e-06, "loss": 0.57, "step": 464 }, { "epoch": 0.09, "learning_rate": 4.99747186944567e-06, "loss": 0.6342, "step": 465 }, { "epoch": 0.09, "learning_rate": 4.99743736201481e-06, "loss": 0.4605, "step": 466 }, { "epoch": 0.09, "learning_rate": 4.9974026207957286e-06, "loss": 0.7152, "step": 467 }, { "epoch": 0.09, "learning_rate": 4.997367645791678e-06, "loss": 0.7093, "step": 468 }, { "epoch": 0.09, "learning_rate": 4.997332437005932e-06, "loss": 0.5887, "step": 469 }, { "epoch": 0.09, "learning_rate": 4.997296994441787e-06, "loss": 0.5676, "step": 470 }, { "epoch": 0.09, "learning_rate": 4.9972613181025605e-06, "loss": 0.5717, "step": 471 }, { "epoch": 0.09, "learning_rate": 4.997225407991593e-06, "loss": 0.7426, "step": 472 }, { "epoch": 0.09, "learning_rate": 4.997189264112246e-06, "loss": 0.5358, "step": 473 }, { "epoch": 0.09, "learning_rate": 4.997152886467902e-06, "loss": 0.6457, "step": 474 }, { "epoch": 0.09, "learning_rate": 4.997116275061967e-06, "loss": 0.5944, "step": 475 }, { "epoch": 0.09, "learning_rate": 4.997079429897868e-06, "loss": 0.4824, "step": 476 }, { "epoch": 0.09, "learning_rate": 4.997042350979055e-06, "loss": 0.5731, "step": 477 }, { "epoch": 0.09, "learning_rate": 4.997005038308999e-06, "loss": 0.573, "step": 478 }, { "epoch": 0.09, "learning_rate": 4.996967491891191e-06, "loss": 0.6677, "step": 479 }, { "epoch": 0.09, "learning_rate": 4.996929711729147e-06, "loss": 0.5757, "step": 480 }, { "epoch": 0.09, "learning_rate": 4.996891697826404e-06, "loss": 0.5538, "step": 481 }, { "epoch": 0.09, "learning_rate": 4.99685345018652e-06, "loss": 0.4886, "step": 482 }, { "epoch": 0.09, "learning_rate": 4.996814968813075e-06, "loss": 0.5507, "step": 483 }, { "epoch": 0.09, "learning_rate": 4.9967762537096725e-06, "loss": 0.5565, "step": 484 }, { "epoch": 0.09, "learning_rate": 4.9967373048799365e-06, "loss": 0.4996, "step": 485 }, { "epoch": 0.09, "learning_rate": 4.996698122327511e-06, "loss": 0.5682, "step": 486 }, { "epoch": 0.09, "learning_rate": 4.996658706056067e-06, "loss": 0.6274, "step": 487 }, { "epoch": 0.09, "learning_rate": 4.996619056069292e-06, "loss": 0.6888, "step": 488 }, { "epoch": 0.09, "learning_rate": 4.996579172370899e-06, "loss": 0.5651, "step": 489 }, { "epoch": 0.09, "learning_rate": 4.99653905496462e-06, "loss": 0.5683, "step": 490 }, { "epoch": 0.09, "learning_rate": 4.996498703854212e-06, "loss": 0.61, "step": 491 }, { "epoch": 0.09, "learning_rate": 4.996458119043452e-06, "loss": 0.6499, "step": 492 }, { "epoch": 0.09, "learning_rate": 4.996417300536139e-06, "loss": 0.5621, "step": 493 }, { "epoch": 0.09, "learning_rate": 4.996376248336094e-06, "loss": 0.7707, "step": 494 }, { "epoch": 0.09, "learning_rate": 4.99633496244716e-06, "loss": 0.5034, "step": 495 }, { "epoch": 0.09, "learning_rate": 4.996293442873202e-06, "loss": 0.4858, "step": 496 }, { "epoch": 0.09, "learning_rate": 4.996251689618106e-06, "loss": 0.6947, "step": 497 }, { "epoch": 0.09, "learning_rate": 4.996209702685781e-06, "loss": 0.496, "step": 498 }, { "epoch": 0.09, "learning_rate": 4.996167482080159e-06, "loss": 0.7663, "step": 499 }, { "epoch": 0.09, "learning_rate": 4.996125027805189e-06, "loss": 0.7248, "step": 500 }, { "epoch": 0.09, "learning_rate": 4.9960823398648485e-06, "loss": 0.6228, "step": 501 }, { "epoch": 0.09, "learning_rate": 4.996039418263131e-06, "loss": 0.7003, "step": 502 }, { "epoch": 0.1, "learning_rate": 4.995996263004056e-06, "loss": 0.6567, "step": 503 }, { "epoch": 0.1, "learning_rate": 4.995952874091663e-06, "loss": 0.6935, "step": 504 }, { "epoch": 0.1, "learning_rate": 4.995909251530014e-06, "loss": 0.6219, "step": 505 }, { "epoch": 0.1, "learning_rate": 4.995865395323192e-06, "loss": 0.5546, "step": 506 }, { "epoch": 0.1, "learning_rate": 4.9958213054753026e-06, "loss": 0.586, "step": 507 }, { "epoch": 0.1, "learning_rate": 4.995776981990472e-06, "loss": 0.6492, "step": 508 }, { "epoch": 0.1, "learning_rate": 4.995732424872852e-06, "loss": 0.7048, "step": 509 }, { "epoch": 0.1, "learning_rate": 4.995687634126611e-06, "loss": 0.5452, "step": 510 }, { "epoch": 0.1, "learning_rate": 4.995642609755944e-06, "loss": 0.5958, "step": 511 }, { "epoch": 0.1, "learning_rate": 4.995597351765065e-06, "loss": 0.6744, "step": 512 }, { "epoch": 0.1, "learning_rate": 4.995551860158209e-06, "loss": 0.5281, "step": 513 }, { "epoch": 0.1, "learning_rate": 4.995506134939637e-06, "loss": 0.4769, "step": 514 }, { "epoch": 0.1, "learning_rate": 4.995460176113628e-06, "loss": 0.6111, "step": 515 }, { "epoch": 0.1, "learning_rate": 4.995413983684485e-06, "loss": 0.771, "step": 516 }, { "epoch": 0.1, "learning_rate": 4.995367557656532e-06, "loss": 0.6556, "step": 517 }, { "epoch": 0.1, "learning_rate": 4.995320898034115e-06, "loss": 0.7412, "step": 518 }, { "epoch": 0.1, "learning_rate": 4.995274004821601e-06, "loss": 0.6719, "step": 519 }, { "epoch": 0.1, "learning_rate": 4.99522687802338e-06, "loss": 0.6179, "step": 520 }, { "epoch": 0.1, "learning_rate": 4.995179517643865e-06, "loss": 0.795, "step": 521 }, { "epoch": 0.1, "learning_rate": 4.995131923687488e-06, "loss": 0.517, "step": 522 }, { "epoch": 0.1, "learning_rate": 4.995084096158704e-06, "loss": 0.6217, "step": 523 }, { "epoch": 0.1, "learning_rate": 4.995036035061993e-06, "loss": 0.5917, "step": 524 }, { "epoch": 0.1, "learning_rate": 4.994987740401851e-06, "loss": 0.5738, "step": 525 }, { "epoch": 0.1, "learning_rate": 4.994939212182799e-06, "loss": 0.6571, "step": 526 }, { "epoch": 0.1, "learning_rate": 4.994890450409382e-06, "loss": 0.7902, "step": 527 }, { "epoch": 0.1, "learning_rate": 4.994841455086162e-06, "loss": 0.6279, "step": 528 }, { "epoch": 0.1, "learning_rate": 4.9947922262177285e-06, "loss": 0.5872, "step": 529 }, { "epoch": 0.1, "learning_rate": 4.994742763808688e-06, "loss": 0.5671, "step": 530 }, { "epoch": 0.1, "learning_rate": 4.9946930678636705e-06, "loss": 0.7272, "step": 531 }, { "epoch": 0.1, "learning_rate": 4.994643138387329e-06, "loss": 0.5517, "step": 532 }, { "epoch": 0.1, "learning_rate": 4.994592975384337e-06, "loss": 0.6949, "step": 533 }, { "epoch": 0.1, "learning_rate": 4.99454257885939e-06, "loss": 0.5198, "step": 534 }, { "epoch": 0.1, "learning_rate": 4.994491948817206e-06, "loss": 0.6724, "step": 535 }, { "epoch": 0.1, "learning_rate": 4.994441085262525e-06, "loss": 0.6089, "step": 536 }, { "epoch": 0.1, "learning_rate": 4.994389988200109e-06, "loss": 0.5904, "step": 537 }, { "epoch": 0.1, "learning_rate": 4.994338657634738e-06, "loss": 0.7284, "step": 538 }, { "epoch": 0.1, "learning_rate": 4.9942870935712215e-06, "loss": 0.5079, "step": 539 }, { "epoch": 0.1, "learning_rate": 4.994235296014384e-06, "loss": 0.5816, "step": 540 }, { "epoch": 0.1, "learning_rate": 4.994183264969074e-06, "loss": 0.6283, "step": 541 }, { "epoch": 0.1, "learning_rate": 4.994131000440163e-06, "loss": 0.6465, "step": 542 }, { "epoch": 0.1, "learning_rate": 4.994078502432544e-06, "loss": 0.5163, "step": 543 }, { "epoch": 0.1, "learning_rate": 4.994025770951131e-06, "loss": 0.694, "step": 544 }, { "epoch": 0.1, "learning_rate": 4.9939728060008595e-06, "loss": 0.5769, "step": 545 }, { "epoch": 0.1, "learning_rate": 4.993919607586688e-06, "loss": 0.5954, "step": 546 }, { "epoch": 0.1, "learning_rate": 4.993866175713598e-06, "loss": 0.7324, "step": 547 }, { "epoch": 0.1, "learning_rate": 4.993812510386589e-06, "loss": 0.6245, "step": 548 }, { "epoch": 0.1, "learning_rate": 4.993758611610685e-06, "loss": 0.6939, "step": 549 }, { "epoch": 0.1, "learning_rate": 4.993704479390934e-06, "loss": 0.5646, "step": 550 }, { "epoch": 0.1, "learning_rate": 4.9936501137324e-06, "loss": 0.6816, "step": 551 }, { "epoch": 0.1, "learning_rate": 4.993595514640175e-06, "loss": 0.5478, "step": 552 }, { "epoch": 0.1, "learning_rate": 4.993540682119369e-06, "loss": 0.5366, "step": 553 }, { "epoch": 0.1, "learning_rate": 4.993485616175114e-06, "loss": 0.6288, "step": 554 }, { "epoch": 0.1, "learning_rate": 4.993430316812566e-06, "loss": 0.6975, "step": 555 }, { "epoch": 0.11, "learning_rate": 4.993374784036902e-06, "loss": 0.4986, "step": 556 }, { "epoch": 0.11, "learning_rate": 4.993319017853319e-06, "loss": 0.6118, "step": 557 }, { "epoch": 0.11, "learning_rate": 4.993263018267039e-06, "loss": 0.5222, "step": 558 }, { "epoch": 0.11, "learning_rate": 4.993206785283302e-06, "loss": 0.6622, "step": 559 }, { "epoch": 0.11, "learning_rate": 4.993150318907374e-06, "loss": 0.6904, "step": 560 }, { "epoch": 0.11, "learning_rate": 4.993093619144541e-06, "loss": 0.68, "step": 561 }, { "epoch": 0.11, "learning_rate": 4.993036686000109e-06, "loss": 0.6402, "step": 562 }, { "epoch": 0.11, "learning_rate": 4.99297951947941e-06, "loss": 0.6804, "step": 563 }, { "epoch": 0.11, "learning_rate": 4.992922119587792e-06, "loss": 0.4777, "step": 564 }, { "epoch": 0.11, "learning_rate": 4.992864486330632e-06, "loss": 0.6577, "step": 565 }, { "epoch": 0.11, "learning_rate": 4.992806619713322e-06, "loss": 0.5159, "step": 566 }, { "epoch": 0.11, "learning_rate": 4.992748519741282e-06, "loss": 0.6356, "step": 567 }, { "epoch": 0.11, "learning_rate": 4.992690186419948e-06, "loss": 0.694, "step": 568 }, { "epoch": 0.11, "learning_rate": 4.992631619754782e-06, "loss": 0.6718, "step": 569 }, { "epoch": 0.11, "learning_rate": 4.992572819751267e-06, "loss": 0.6694, "step": 570 }, { "epoch": 0.11, "learning_rate": 4.992513786414906e-06, "loss": 0.5242, "step": 571 }, { "epoch": 0.11, "learning_rate": 4.992454519751227e-06, "loss": 0.6744, "step": 572 }, { "epoch": 0.11, "learning_rate": 4.992395019765776e-06, "loss": 0.6638, "step": 573 }, { "epoch": 0.11, "learning_rate": 4.992335286464123e-06, "loss": 0.4489, "step": 574 }, { "epoch": 0.11, "learning_rate": 4.9922753198518614e-06, "loss": 0.6465, "step": 575 }, { "epoch": 0.11, "learning_rate": 4.992215119934604e-06, "loss": 0.3804, "step": 576 }, { "epoch": 0.11, "learning_rate": 4.992154686717985e-06, "loss": 0.6432, "step": 577 }, { "epoch": 0.11, "learning_rate": 4.992094020207664e-06, "loss": 0.5377, "step": 578 }, { "epoch": 0.11, "learning_rate": 4.992033120409318e-06, "loss": 0.5549, "step": 579 }, { "epoch": 0.11, "learning_rate": 4.991971987328649e-06, "loss": 0.6473, "step": 580 }, { "epoch": 0.11, "learning_rate": 4.991910620971379e-06, "loss": 0.6309, "step": 581 }, { "epoch": 0.11, "learning_rate": 4.9918490213432525e-06, "loss": 0.6174, "step": 582 }, { "epoch": 0.11, "learning_rate": 4.991787188450037e-06, "loss": 0.5346, "step": 583 }, { "epoch": 0.11, "learning_rate": 4.991725122297521e-06, "loss": 0.5502, "step": 584 }, { "epoch": 0.11, "learning_rate": 4.991662822891512e-06, "loss": 0.4878, "step": 585 }, { "epoch": 0.11, "learning_rate": 4.991600290237845e-06, "loss": 0.7417, "step": 586 }, { "epoch": 0.11, "learning_rate": 4.991537524342371e-06, "loss": 0.5949, "step": 587 }, { "epoch": 0.11, "learning_rate": 4.991474525210967e-06, "loss": 0.496, "step": 588 }, { "epoch": 0.11, "learning_rate": 4.991411292849532e-06, "loss": 0.7647, "step": 589 }, { "epoch": 0.11, "learning_rate": 4.991347827263983e-06, "loss": 0.5644, "step": 590 }, { "epoch": 0.11, "learning_rate": 4.991284128460261e-06, "loss": 0.5216, "step": 591 }, { "epoch": 0.11, "learning_rate": 4.99122019644433e-06, "loss": 0.6682, "step": 592 }, { "epoch": 0.11, "learning_rate": 4.991156031222175e-06, "loss": 0.4488, "step": 593 }, { "epoch": 0.11, "learning_rate": 4.991091632799801e-06, "loss": 0.6987, "step": 594 }, { "epoch": 0.11, "learning_rate": 4.99102700118324e-06, "loss": 0.6496, "step": 595 }, { "epoch": 0.11, "learning_rate": 4.990962136378537e-06, "loss": 0.7498, "step": 596 }, { "epoch": 0.11, "learning_rate": 4.9908970383917676e-06, "loss": 0.5851, "step": 597 }, { "epoch": 0.11, "learning_rate": 4.990831707229025e-06, "loss": 0.583, "step": 598 }, { "epoch": 0.11, "learning_rate": 4.990766142896425e-06, "loss": 0.5957, "step": 599 }, { "epoch": 0.11, "learning_rate": 4.990700345400105e-06, "loss": 0.4994, "step": 600 }, { "epoch": 0.11, "learning_rate": 4.990634314746223e-06, "loss": 0.5826, "step": 601 }, { "epoch": 0.11, "learning_rate": 4.990568050940963e-06, "loss": 0.5534, "step": 602 }, { "epoch": 0.11, "learning_rate": 4.990501553990526e-06, "loss": 0.5684, "step": 603 }, { "epoch": 0.11, "learning_rate": 4.990434823901138e-06, "loss": 0.5945, "step": 604 }, { "epoch": 0.11, "learning_rate": 4.990367860679045e-06, "loss": 0.6116, "step": 605 }, { "epoch": 0.11, "learning_rate": 4.990300664330515e-06, "loss": 0.7104, "step": 606 }, { "epoch": 0.11, "learning_rate": 4.99023323486184e-06, "loss": 0.6545, "step": 607 }, { "epoch": 0.11, "learning_rate": 4.99016557227933e-06, "loss": 0.4887, "step": 608 }, { "epoch": 0.12, "learning_rate": 4.99009767658932e-06, "loss": 0.5269, "step": 609 }, { "epoch": 0.12, "learning_rate": 4.990029547798168e-06, "loss": 0.5799, "step": 610 }, { "epoch": 0.12, "learning_rate": 4.9899611859122475e-06, "loss": 0.7287, "step": 611 }, { "epoch": 0.12, "learning_rate": 4.98989259093796e-06, "loss": 0.6534, "step": 612 }, { "epoch": 0.12, "learning_rate": 4.989823762881728e-06, "loss": 0.5336, "step": 613 }, { "epoch": 0.12, "learning_rate": 4.989754701749993e-06, "loss": 0.6175, "step": 614 }, { "epoch": 0.12, "learning_rate": 4.9896854075492195e-06, "loss": 0.5731, "step": 615 }, { "epoch": 0.12, "learning_rate": 4.989615880285896e-06, "loss": 0.6189, "step": 616 }, { "epoch": 0.12, "learning_rate": 4.989546119966529e-06, "loss": 0.6523, "step": 617 }, { "epoch": 0.12, "learning_rate": 4.989476126597651e-06, "loss": 0.6382, "step": 618 }, { "epoch": 0.12, "learning_rate": 4.989405900185812e-06, "loss": 0.6064, "step": 619 }, { "epoch": 0.12, "learning_rate": 4.989335440737587e-06, "loss": 0.8097, "step": 620 }, { "epoch": 0.12, "learning_rate": 4.989264748259571e-06, "loss": 0.7016, "step": 621 }, { "epoch": 0.12, "learning_rate": 4.989193822758384e-06, "loss": 0.5089, "step": 622 }, { "epoch": 0.12, "learning_rate": 4.9891226642406635e-06, "loss": 0.5301, "step": 623 }, { "epoch": 0.12, "learning_rate": 4.98905127271307e-06, "loss": 0.5359, "step": 624 }, { "epoch": 0.12, "learning_rate": 4.988979648182288e-06, "loss": 0.5489, "step": 625 }, { "epoch": 0.12, "learning_rate": 4.988907790655024e-06, "loss": 0.7118, "step": 626 }, { "epoch": 0.12, "learning_rate": 4.988835700137999e-06, "loss": 0.5416, "step": 627 }, { "epoch": 0.12, "learning_rate": 4.988763376637968e-06, "loss": 0.6279, "step": 628 }, { "epoch": 0.12, "learning_rate": 4.988690820161696e-06, "loss": 0.5507, "step": 629 }, { "epoch": 0.12, "learning_rate": 4.98861803071598e-06, "loss": 0.6155, "step": 630 }, { "epoch": 0.12, "learning_rate": 4.98854500830763e-06, "loss": 0.5185, "step": 631 }, { "epoch": 0.12, "learning_rate": 4.9884717529434835e-06, "loss": 0.6748, "step": 632 }, { "epoch": 0.12, "learning_rate": 4.988398264630398e-06, "loss": 0.7438, "step": 633 }, { "epoch": 0.12, "learning_rate": 4.988324543375252e-06, "loss": 0.5973, "step": 634 }, { "epoch": 0.12, "learning_rate": 4.9882505891849475e-06, "loss": 0.4889, "step": 635 }, { "epoch": 0.12, "learning_rate": 4.988176402066407e-06, "loss": 0.5988, "step": 636 }, { "epoch": 0.12, "learning_rate": 4.988101982026574e-06, "loss": 0.4996, "step": 637 }, { "epoch": 0.12, "learning_rate": 4.988027329072418e-06, "loss": 0.5911, "step": 638 }, { "epoch": 0.12, "learning_rate": 4.9879524432109255e-06, "loss": 0.712, "step": 639 }, { "epoch": 0.12, "learning_rate": 4.987877324449107e-06, "loss": 0.4892, "step": 640 }, { "epoch": 0.12, "learning_rate": 4.987801972793993e-06, "loss": 0.6114, "step": 641 }, { "epoch": 0.12, "learning_rate": 4.98772638825264e-06, "loss": 0.6342, "step": 642 }, { "epoch": 0.12, "learning_rate": 4.987650570832122e-06, "loss": 0.5971, "step": 643 }, { "epoch": 0.12, "learning_rate": 4.9875745205395365e-06, "loss": 0.506, "step": 644 }, { "epoch": 0.12, "learning_rate": 4.987498237382002e-06, "loss": 0.4412, "step": 645 }, { "epoch": 0.12, "learning_rate": 4.987421721366661e-06, "loss": 0.5027, "step": 646 }, { "epoch": 0.12, "learning_rate": 4.987344972500675e-06, "loss": 0.5487, "step": 647 }, { "epoch": 0.12, "learning_rate": 4.987267990791229e-06, "loss": 0.5902, "step": 648 }, { "epoch": 0.12, "learning_rate": 4.987190776245529e-06, "loss": 0.4537, "step": 649 }, { "epoch": 0.12, "learning_rate": 4.987113328870805e-06, "loss": 0.5989, "step": 650 }, { "epoch": 0.12, "learning_rate": 4.987035648674305e-06, "loss": 0.5567, "step": 651 }, { "epoch": 0.12, "learning_rate": 4.9869577356633e-06, "loss": 0.5675, "step": 652 }, { "epoch": 0.12, "learning_rate": 4.986879589845085e-06, "loss": 0.7715, "step": 653 }, { "epoch": 0.12, "learning_rate": 4.986801211226976e-06, "loss": 0.5601, "step": 654 }, { "epoch": 0.12, "learning_rate": 4.986722599816308e-06, "loss": 0.6122, "step": 655 }, { "epoch": 0.12, "learning_rate": 4.986643755620441e-06, "loss": 0.5409, "step": 656 }, { "epoch": 0.12, "learning_rate": 4.986564678646756e-06, "loss": 0.5765, "step": 657 }, { "epoch": 0.12, "learning_rate": 4.986485368902656e-06, "loss": 0.5027, "step": 658 }, { "epoch": 0.12, "learning_rate": 4.9864058263955635e-06, "loss": 0.5433, "step": 659 }, { "epoch": 0.12, "learning_rate": 4.986326051132927e-06, "loss": 0.5614, "step": 660 }, { "epoch": 0.12, "learning_rate": 4.986246043122212e-06, "loss": 0.4941, "step": 661 }, { "epoch": 0.13, "learning_rate": 4.986165802370909e-06, "loss": 0.5986, "step": 662 }, { "epoch": 0.13, "learning_rate": 4.98608532888653e-06, "loss": 0.7352, "step": 663 }, { "epoch": 0.13, "learning_rate": 4.986004622676608e-06, "loss": 0.4618, "step": 664 }, { "epoch": 0.13, "learning_rate": 4.985923683748697e-06, "loss": 0.6956, "step": 665 }, { "epoch": 0.13, "learning_rate": 4.985842512110375e-06, "loss": 0.4751, "step": 666 }, { "epoch": 0.13, "learning_rate": 4.98576110776924e-06, "loss": 0.6259, "step": 667 }, { "epoch": 0.13, "learning_rate": 4.985679470732913e-06, "loss": 0.5796, "step": 668 }, { "epoch": 0.13, "learning_rate": 4.9855976010090356e-06, "loss": 0.6487, "step": 669 }, { "epoch": 0.13, "learning_rate": 4.985515498605272e-06, "loss": 0.6116, "step": 670 }, { "epoch": 0.13, "learning_rate": 4.985433163529307e-06, "loss": 0.5679, "step": 671 }, { "epoch": 0.13, "learning_rate": 4.98535059578885e-06, "loss": 0.5487, "step": 672 }, { "epoch": 0.13, "learning_rate": 4.985267795391627e-06, "loss": 0.6978, "step": 673 }, { "epoch": 0.13, "learning_rate": 4.985184762345393e-06, "loss": 0.4644, "step": 674 }, { "epoch": 0.13, "learning_rate": 4.985101496657918e-06, "loss": 0.6359, "step": 675 }, { "epoch": 0.13, "learning_rate": 4.985017998336998e-06, "loss": 0.663, "step": 676 }, { "epoch": 0.13, "learning_rate": 4.9849342673904485e-06, "loss": 0.6591, "step": 677 }, { "epoch": 0.13, "learning_rate": 4.984850303826109e-06, "loss": 0.6061, "step": 678 }, { "epoch": 0.13, "learning_rate": 4.984766107651837e-06, "loss": 0.5456, "step": 679 }, { "epoch": 0.13, "learning_rate": 4.984681678875517e-06, "loss": 0.5457, "step": 680 }, { "epoch": 0.13, "learning_rate": 4.98459701750505e-06, "loss": 0.5851, "step": 681 }, { "epoch": 0.13, "learning_rate": 4.984512123548364e-06, "loss": 0.5066, "step": 682 }, { "epoch": 0.13, "learning_rate": 4.984426997013404e-06, "loss": 0.6103, "step": 683 }, { "epoch": 0.13, "learning_rate": 4.984341637908139e-06, "loss": 0.6734, "step": 684 }, { "epoch": 0.13, "learning_rate": 4.98425604624056e-06, "loss": 0.6392, "step": 685 }, { "epoch": 0.13, "learning_rate": 4.984170222018679e-06, "loss": 0.7044, "step": 686 }, { "epoch": 0.13, "learning_rate": 4.984084165250531e-06, "loss": 0.7324, "step": 687 }, { "epoch": 0.13, "learning_rate": 4.98399787594417e-06, "loss": 0.5385, "step": 688 }, { "epoch": 0.13, "learning_rate": 4.983911354107676e-06, "loss": 0.6363, "step": 689 }, { "epoch": 0.13, "learning_rate": 4.983824599749146e-06, "loss": 0.6429, "step": 690 }, { "epoch": 0.13, "learning_rate": 4.983737612876704e-06, "loss": 0.5885, "step": 691 }, { "epoch": 0.13, "learning_rate": 4.98365039349849e-06, "loss": 0.5466, "step": 692 }, { "epoch": 0.13, "learning_rate": 4.983562941622672e-06, "loss": 0.5293, "step": 693 }, { "epoch": 0.13, "learning_rate": 4.983475257257433e-06, "loss": 0.594, "step": 694 }, { "epoch": 0.13, "learning_rate": 4.983387340410984e-06, "loss": 0.6543, "step": 695 }, { "epoch": 0.13, "learning_rate": 4.983299191091553e-06, "loss": 0.5417, "step": 696 }, { "epoch": 0.13, "learning_rate": 4.983210809307392e-06, "loss": 0.5799, "step": 697 }, { "epoch": 0.13, "learning_rate": 4.983122195066777e-06, "loss": 0.5437, "step": 698 }, { "epoch": 0.13, "learning_rate": 4.983033348377999e-06, "loss": 0.6054, "step": 699 }, { "epoch": 0.13, "learning_rate": 4.982944269249379e-06, "loss": 0.6686, "step": 700 }, { "epoch": 0.13, "learning_rate": 4.982854957689255e-06, "loss": 0.5314, "step": 701 }, { "epoch": 0.13, "learning_rate": 4.982765413705986e-06, "loss": 0.6099, "step": 702 }, { "epoch": 0.13, "learning_rate": 4.982675637307954e-06, "loss": 0.5742, "step": 703 }, { "epoch": 0.13, "learning_rate": 4.982585628503566e-06, "loss": 0.5222, "step": 704 }, { "epoch": 0.13, "learning_rate": 4.982495387301245e-06, "loss": 0.5115, "step": 705 }, { "epoch": 0.13, "learning_rate": 4.982404913709441e-06, "loss": 0.5097, "step": 706 }, { "epoch": 0.13, "learning_rate": 4.982314207736621e-06, "loss": 0.5132, "step": 707 }, { "epoch": 0.13, "learning_rate": 4.9822232693912775e-06, "loss": 0.6161, "step": 708 }, { "epoch": 0.13, "learning_rate": 4.982132098681923e-06, "loss": 0.5563, "step": 709 }, { "epoch": 0.13, "learning_rate": 4.982040695617092e-06, "loss": 0.5952, "step": 710 }, { "epoch": 0.13, "learning_rate": 4.9819490602053405e-06, "loss": 0.7291, "step": 711 }, { "epoch": 0.13, "learning_rate": 4.9818571924552474e-06, "loss": 0.7277, "step": 712 }, { "epoch": 0.13, "learning_rate": 4.981765092375412e-06, "loss": 0.4393, "step": 713 }, { "epoch": 0.13, "learning_rate": 4.981672759974457e-06, "loss": 0.6617, "step": 714 }, { "epoch": 0.14, "learning_rate": 4.981580195261024e-06, "loss": 0.5948, "step": 715 }, { "epoch": 0.14, "learning_rate": 4.981487398243779e-06, "loss": 0.5079, "step": 716 }, { "epoch": 0.14, "learning_rate": 4.9813943689314095e-06, "loss": 0.5578, "step": 717 }, { "epoch": 0.14, "learning_rate": 4.981301107332623e-06, "loss": 0.5826, "step": 718 }, { "epoch": 0.14, "learning_rate": 4.9812076134561495e-06, "loss": 0.6198, "step": 719 }, { "epoch": 0.14, "learning_rate": 4.981113887310743e-06, "loss": 0.5769, "step": 720 }, { "epoch": 0.14, "learning_rate": 4.9810199289051755e-06, "loss": 0.5392, "step": 721 }, { "epoch": 0.14, "learning_rate": 4.980925738248243e-06, "loss": 0.572, "step": 722 }, { "epoch": 0.14, "learning_rate": 4.980831315348763e-06, "loss": 0.5468, "step": 723 }, { "epoch": 0.14, "learning_rate": 4.980736660215575e-06, "loss": 0.6885, "step": 724 }, { "epoch": 0.14, "learning_rate": 4.980641772857539e-06, "loss": 0.6501, "step": 725 }, { "epoch": 0.14, "learning_rate": 4.980546653283538e-06, "loss": 0.6385, "step": 726 }, { "epoch": 0.14, "learning_rate": 4.9804513015024755e-06, "loss": 0.4479, "step": 727 }, { "epoch": 0.14, "learning_rate": 4.98035571752328e-06, "loss": 0.578, "step": 728 }, { "epoch": 0.14, "learning_rate": 4.9802599013548955e-06, "loss": 0.481, "step": 729 }, { "epoch": 0.14, "learning_rate": 4.980163853006294e-06, "loss": 0.5737, "step": 730 }, { "epoch": 0.14, "learning_rate": 4.980067572486466e-06, "loss": 0.5789, "step": 731 }, { "epoch": 0.14, "learning_rate": 4.979971059804424e-06, "loss": 0.6247, "step": 732 }, { "epoch": 0.14, "learning_rate": 4.979874314969203e-06, "loss": 0.7557, "step": 733 }, { "epoch": 0.14, "learning_rate": 4.979777337989861e-06, "loss": 0.6473, "step": 734 }, { "epoch": 0.14, "learning_rate": 4.979680128875473e-06, "loss": 0.5863, "step": 735 }, { "epoch": 0.14, "learning_rate": 4.979582687635141e-06, "loss": 0.4832, "step": 736 }, { "epoch": 0.14, "learning_rate": 4.979485014277986e-06, "loss": 0.5136, "step": 737 }, { "epoch": 0.14, "learning_rate": 4.979387108813152e-06, "loss": 0.4169, "step": 738 }, { "epoch": 0.14, "learning_rate": 4.979288971249804e-06, "loss": 0.4883, "step": 739 }, { "epoch": 0.14, "learning_rate": 4.979190601597127e-06, "loss": 0.5772, "step": 740 }, { "epoch": 0.14, "learning_rate": 4.979091999864332e-06, "loss": 0.6558, "step": 741 }, { "epoch": 0.14, "learning_rate": 4.978993166060647e-06, "loss": 0.5553, "step": 742 }, { "epoch": 0.14, "learning_rate": 4.978894100195325e-06, "loss": 0.5659, "step": 743 }, { "epoch": 0.14, "learning_rate": 4.97879480227764e-06, "loss": 0.4901, "step": 744 }, { "epoch": 0.14, "learning_rate": 4.978695272316888e-06, "loss": 0.5705, "step": 745 }, { "epoch": 0.14, "learning_rate": 4.978595510322383e-06, "loss": 0.6571, "step": 746 }, { "epoch": 0.14, "learning_rate": 4.9784955163034674e-06, "loss": 0.5555, "step": 747 }, { "epoch": 0.14, "learning_rate": 4.978395290269501e-06, "loss": 0.5216, "step": 748 }, { "epoch": 0.14, "learning_rate": 4.9782948322298646e-06, "loss": 0.633, "step": 749 }, { "epoch": 0.14, "learning_rate": 4.978194142193963e-06, "loss": 0.5766, "step": 750 }, { "epoch": 0.14, "learning_rate": 4.978093220171223e-06, "loss": 0.6546, "step": 751 }, { "epoch": 0.14, "learning_rate": 4.97799206617109e-06, "loss": 0.5886, "step": 752 }, { "epoch": 0.14, "learning_rate": 4.977890680203035e-06, "loss": 0.7086, "step": 753 }, { "epoch": 0.14, "learning_rate": 4.9777890622765464e-06, "loss": 0.5933, "step": 754 }, { "epoch": 0.14, "learning_rate": 4.97768721240114e-06, "loss": 0.6023, "step": 755 }, { "epoch": 0.14, "learning_rate": 4.977585130586349e-06, "loss": 0.8361, "step": 756 }, { "epoch": 0.14, "learning_rate": 4.977482816841728e-06, "loss": 0.5993, "step": 757 }, { "epoch": 0.14, "learning_rate": 4.977380271176855e-06, "loss": 0.5394, "step": 758 }, { "epoch": 0.14, "learning_rate": 4.977277493601331e-06, "loss": 0.6325, "step": 759 }, { "epoch": 0.14, "learning_rate": 4.977174484124776e-06, "loss": 0.5887, "step": 760 }, { "epoch": 0.14, "learning_rate": 4.977071242756834e-06, "loss": 0.6638, "step": 761 }, { "epoch": 0.14, "learning_rate": 4.976967769507167e-06, "loss": 0.5613, "step": 762 }, { "epoch": 0.14, "learning_rate": 4.976864064385464e-06, "loss": 0.4926, "step": 763 }, { "epoch": 0.14, "learning_rate": 4.976760127401431e-06, "loss": 0.7191, "step": 764 }, { "epoch": 0.14, "learning_rate": 4.976655958564799e-06, "loss": 0.5547, "step": 765 }, { "epoch": 0.14, "learning_rate": 4.976551557885318e-06, "loss": 0.5023, "step": 766 }, { "epoch": 0.14, "learning_rate": 4.976446925372763e-06, "loss": 0.6377, "step": 767 }, { "epoch": 0.15, "learning_rate": 4.976342061036928e-06, "loss": 0.507, "step": 768 }, { "epoch": 0.15, "learning_rate": 4.976236964887628e-06, "loss": 0.6426, "step": 769 }, { "epoch": 0.15, "learning_rate": 4.976131636934703e-06, "loss": 0.6072, "step": 770 }, { "epoch": 0.15, "learning_rate": 4.976026077188013e-06, "loss": 0.5482, "step": 771 }, { "epoch": 0.15, "learning_rate": 4.975920285657438e-06, "loss": 0.7026, "step": 772 }, { "epoch": 0.15, "learning_rate": 4.9758142623528825e-06, "loss": 0.5872, "step": 773 }, { "epoch": 0.15, "learning_rate": 4.9757080072842715e-06, "loss": 0.4455, "step": 774 }, { "epoch": 0.15, "learning_rate": 4.975601520461551e-06, "loss": 0.6185, "step": 775 }, { "epoch": 0.15, "learning_rate": 4.975494801894689e-06, "loss": 0.5688, "step": 776 }, { "epoch": 0.15, "learning_rate": 4.975387851593677e-06, "loss": 0.4708, "step": 777 }, { "epoch": 0.15, "learning_rate": 4.9752806695685265e-06, "loss": 0.6324, "step": 778 }, { "epoch": 0.15, "learning_rate": 4.9751732558292684e-06, "loss": 0.6222, "step": 779 }, { "epoch": 0.15, "learning_rate": 4.975065610385962e-06, "loss": 0.4227, "step": 780 }, { "epoch": 0.15, "learning_rate": 4.974957733248681e-06, "loss": 0.5411, "step": 781 }, { "epoch": 0.15, "learning_rate": 4.9748496244275244e-06, "loss": 0.5096, "step": 782 }, { "epoch": 0.15, "learning_rate": 4.974741283932614e-06, "loss": 0.5729, "step": 783 }, { "epoch": 0.15, "learning_rate": 4.9746327117740885e-06, "loss": 0.4924, "step": 784 }, { "epoch": 0.15, "learning_rate": 4.974523907962115e-06, "loss": 0.5287, "step": 785 }, { "epoch": 0.15, "learning_rate": 4.974414872506878e-06, "loss": 0.6618, "step": 786 }, { "epoch": 0.15, "learning_rate": 4.974305605418583e-06, "loss": 0.4304, "step": 787 }, { "epoch": 0.15, "learning_rate": 4.9741961067074595e-06, "loss": 0.6281, "step": 788 }, { "epoch": 0.15, "learning_rate": 4.974086376383758e-06, "loss": 0.5365, "step": 789 }, { "epoch": 0.15, "learning_rate": 4.973976414457749e-06, "loss": 0.7778, "step": 790 }, { "epoch": 0.15, "learning_rate": 4.973866220939728e-06, "loss": 0.5566, "step": 791 }, { "epoch": 0.15, "learning_rate": 4.97375579584001e-06, "loss": 0.666, "step": 792 }, { "epoch": 0.15, "learning_rate": 4.973645139168931e-06, "loss": 0.5362, "step": 793 }, { "epoch": 0.15, "learning_rate": 4.973534250936851e-06, "loss": 0.644, "step": 794 }, { "epoch": 0.15, "learning_rate": 4.97342313115415e-06, "loss": 0.4832, "step": 795 }, { "epoch": 0.15, "learning_rate": 4.973311779831229e-06, "loss": 0.631, "step": 796 }, { "epoch": 0.15, "learning_rate": 4.973200196978514e-06, "loss": 0.6427, "step": 797 }, { "epoch": 0.15, "learning_rate": 4.973088382606448e-06, "loss": 0.5767, "step": 798 }, { "epoch": 0.15, "learning_rate": 4.972976336725499e-06, "loss": 0.581, "step": 799 }, { "epoch": 0.15, "learning_rate": 4.9728640593461566e-06, "loss": 0.6559, "step": 800 }, { "epoch": 0.15, "learning_rate": 4.97275155047893e-06, "loss": 0.5141, "step": 801 }, { "epoch": 0.15, "learning_rate": 4.9726388101343526e-06, "loss": 0.5971, "step": 802 }, { "epoch": 0.15, "learning_rate": 4.972525838322976e-06, "loss": 0.5155, "step": 803 }, { "epoch": 0.15, "learning_rate": 4.972412635055378e-06, "loss": 0.6231, "step": 804 }, { "epoch": 0.15, "learning_rate": 4.972299200342154e-06, "loss": 0.5221, "step": 805 }, { "epoch": 0.15, "learning_rate": 4.972185534193924e-06, "loss": 0.7368, "step": 806 }, { "epoch": 0.15, "learning_rate": 4.972071636621328e-06, "loss": 0.5874, "step": 807 }, { "epoch": 0.15, "learning_rate": 4.9719575076350275e-06, "loss": 0.6764, "step": 808 }, { "epoch": 0.15, "learning_rate": 4.971843147245707e-06, "loss": 0.6155, "step": 809 }, { "epoch": 0.15, "learning_rate": 4.971728555464071e-06, "loss": 0.5299, "step": 810 }, { "epoch": 0.15, "learning_rate": 4.9716137323008485e-06, "loss": 0.5607, "step": 811 }, { "epoch": 0.15, "learning_rate": 4.971498677766786e-06, "loss": 0.6525, "step": 812 }, { "epoch": 0.15, "learning_rate": 4.971383391872655e-06, "loss": 0.6013, "step": 813 }, { "epoch": 0.15, "learning_rate": 4.9712678746292475e-06, "loss": 0.6076, "step": 814 }, { "epoch": 0.15, "learning_rate": 4.971152126047377e-06, "loss": 0.526, "step": 815 }, { "epoch": 0.15, "learning_rate": 4.971036146137879e-06, "loss": 0.6355, "step": 816 }, { "epoch": 0.15, "learning_rate": 4.970919934911611e-06, "loss": 0.4774, "step": 817 }, { "epoch": 0.15, "learning_rate": 4.9708034923794516e-06, "loss": 0.529, "step": 818 }, { "epoch": 0.15, "learning_rate": 4.970686818552299e-06, "loss": 0.5809, "step": 819 }, { "epoch": 0.15, "learning_rate": 4.970569913441078e-06, "loss": 0.5911, "step": 820 }, { "epoch": 0.16, "learning_rate": 4.970452777056731e-06, "loss": 0.6341, "step": 821 }, { "epoch": 0.16, "learning_rate": 4.970335409410224e-06, "loss": 0.7212, "step": 822 }, { "epoch": 0.16, "learning_rate": 4.9702178105125425e-06, "loss": 0.5603, "step": 823 }, { "epoch": 0.16, "learning_rate": 4.9700999803746965e-06, "loss": 0.5847, "step": 824 }, { "epoch": 0.16, "learning_rate": 4.969981919007716e-06, "loss": 0.5912, "step": 825 }, { "epoch": 0.16, "learning_rate": 4.969863626422652e-06, "loss": 0.5319, "step": 826 }, { "epoch": 0.16, "learning_rate": 4.9697451026305774e-06, "loss": 0.6873, "step": 827 }, { "epoch": 0.16, "learning_rate": 4.96962634764259e-06, "loss": 0.7584, "step": 828 }, { "epoch": 0.16, "learning_rate": 4.9695073614698055e-06, "loss": 0.6413, "step": 829 }, { "epoch": 0.16, "learning_rate": 4.969388144123361e-06, "loss": 0.6329, "step": 830 }, { "epoch": 0.16, "learning_rate": 4.969268695614418e-06, "loss": 0.5873, "step": 831 }, { "epoch": 0.16, "learning_rate": 4.969149015954157e-06, "loss": 0.6676, "step": 832 }, { "epoch": 0.16, "learning_rate": 4.969029105153783e-06, "loss": 0.6462, "step": 833 }, { "epoch": 0.16, "learning_rate": 4.96890896322452e-06, "loss": 0.6295, "step": 834 }, { "epoch": 0.16, "learning_rate": 4.9687885901776144e-06, "loss": 0.4607, "step": 835 }, { "epoch": 0.16, "learning_rate": 4.9686679860243345e-06, "loss": 0.5876, "step": 836 }, { "epoch": 0.16, "learning_rate": 4.968547150775972e-06, "loss": 0.6106, "step": 837 }, { "epoch": 0.16, "learning_rate": 4.968426084443836e-06, "loss": 0.6116, "step": 838 }, { "epoch": 0.16, "learning_rate": 4.968304787039261e-06, "loss": 0.4074, "step": 839 }, { "epoch": 0.16, "learning_rate": 4.968183258573601e-06, "loss": 0.5525, "step": 840 }, { "epoch": 0.16, "learning_rate": 4.968061499058233e-06, "loss": 0.5435, "step": 841 }, { "epoch": 0.16, "learning_rate": 4.967939508504555e-06, "loss": 0.5236, "step": 842 }, { "epoch": 0.16, "learning_rate": 4.967817286923986e-06, "loss": 0.5731, "step": 843 }, { "epoch": 0.16, "learning_rate": 4.967694834327969e-06, "loss": 0.6252, "step": 844 }, { "epoch": 0.16, "learning_rate": 4.967572150727965e-06, "loss": 0.7052, "step": 845 }, { "epoch": 0.16, "learning_rate": 4.9674492361354604e-06, "loss": 0.5264, "step": 846 }, { "epoch": 0.16, "learning_rate": 4.96732609056196e-06, "loss": 0.776, "step": 847 }, { "epoch": 0.16, "learning_rate": 4.967202714018991e-06, "loss": 0.5925, "step": 848 }, { "epoch": 0.16, "learning_rate": 4.967079106518106e-06, "loss": 0.5191, "step": 849 }, { "epoch": 0.16, "learning_rate": 4.966955268070873e-06, "loss": 0.5696, "step": 850 }, { "epoch": 0.16, "learning_rate": 4.9668311986888855e-06, "loss": 0.6267, "step": 851 }, { "epoch": 0.16, "learning_rate": 4.966706898383757e-06, "loss": 0.7658, "step": 852 }, { "epoch": 0.16, "learning_rate": 4.966582367167126e-06, "loss": 0.5608, "step": 853 }, { "epoch": 0.16, "learning_rate": 4.966457605050647e-06, "loss": 0.5059, "step": 854 }, { "epoch": 0.16, "learning_rate": 4.966332612046001e-06, "loss": 0.5736, "step": 855 }, { "epoch": 0.16, "learning_rate": 4.966207388164888e-06, "loss": 0.5535, "step": 856 }, { "epoch": 0.16, "learning_rate": 4.96608193341903e-06, "loss": 0.5594, "step": 857 }, { "epoch": 0.16, "learning_rate": 4.9659562478201724e-06, "loss": 0.6303, "step": 858 }, { "epoch": 0.16, "learning_rate": 4.965830331380081e-06, "loss": 0.554, "step": 859 }, { "epoch": 0.16, "learning_rate": 4.96570418411054e-06, "loss": 0.6388, "step": 860 }, { "epoch": 0.16, "learning_rate": 4.965577806023361e-06, "loss": 0.5712, "step": 861 }, { "epoch": 0.16, "learning_rate": 4.965451197130373e-06, "loss": 0.6741, "step": 862 }, { "epoch": 0.16, "learning_rate": 4.965324357443428e-06, "loss": 0.6766, "step": 863 }, { "epoch": 0.16, "learning_rate": 4.965197286974401e-06, "loss": 0.6713, "step": 864 }, { "epoch": 0.16, "learning_rate": 4.965069985735187e-06, "loss": 0.6835, "step": 865 }, { "epoch": 0.16, "learning_rate": 4.9649424537377024e-06, "loss": 0.5357, "step": 866 }, { "epoch": 0.16, "learning_rate": 4.964814690993885e-06, "loss": 0.6439, "step": 867 }, { "epoch": 0.16, "learning_rate": 4.964686697515694e-06, "loss": 0.4858, "step": 868 }, { "epoch": 0.16, "learning_rate": 4.9645584733151135e-06, "loss": 0.6212, "step": 869 }, { "epoch": 0.16, "learning_rate": 4.964430018404145e-06, "loss": 0.4649, "step": 870 }, { "epoch": 0.16, "learning_rate": 4.9643013327948144e-06, "loss": 0.5111, "step": 871 }, { "epoch": 0.16, "learning_rate": 4.964172416499168e-06, "loss": 0.5837, "step": 872 }, { "epoch": 0.16, "learning_rate": 4.964043269529273e-06, "loss": 0.601, "step": 873 }, { "epoch": 0.17, "learning_rate": 4.963913891897219e-06, "loss": 0.4329, "step": 874 }, { "epoch": 0.17, "learning_rate": 4.9637842836151175e-06, "loss": 0.5925, "step": 875 }, { "epoch": 0.17, "learning_rate": 4.963654444695102e-06, "loss": 0.549, "step": 876 }, { "epoch": 0.17, "learning_rate": 4.963524375149327e-06, "loss": 0.6177, "step": 877 }, { "epoch": 0.17, "learning_rate": 4.9633940749899665e-06, "loss": 0.55, "step": 878 }, { "epoch": 0.17, "learning_rate": 4.963263544229219e-06, "loss": 0.642, "step": 879 }, { "epoch": 0.17, "learning_rate": 4.963132782879305e-06, "loss": 0.5282, "step": 880 }, { "epoch": 0.17, "learning_rate": 4.9630017909524645e-06, "loss": 0.5271, "step": 881 }, { "epoch": 0.17, "learning_rate": 4.962870568460958e-06, "loss": 0.5904, "step": 882 }, { "epoch": 0.17, "learning_rate": 4.962739115417073e-06, "loss": 0.5979, "step": 883 }, { "epoch": 0.17, "learning_rate": 4.96260743183311e-06, "loss": 0.6564, "step": 884 }, { "epoch": 0.17, "learning_rate": 4.962475517721401e-06, "loss": 0.5383, "step": 885 }, { "epoch": 0.17, "learning_rate": 4.962343373094293e-06, "loss": 0.6621, "step": 886 }, { "epoch": 0.17, "learning_rate": 4.962210997964154e-06, "loss": 0.5828, "step": 887 }, { "epoch": 0.17, "learning_rate": 4.962078392343379e-06, "loss": 0.6148, "step": 888 }, { "epoch": 0.17, "learning_rate": 4.96194555624438e-06, "loss": 0.5458, "step": 889 }, { "epoch": 0.17, "learning_rate": 4.961812489679592e-06, "loss": 0.543, "step": 890 }, { "epoch": 0.17, "learning_rate": 4.961679192661471e-06, "loss": 0.6564, "step": 891 }, { "epoch": 0.17, "learning_rate": 4.961545665202496e-06, "loss": 0.6033, "step": 892 }, { "epoch": 0.17, "learning_rate": 4.9614119073151654e-06, "loss": 0.6096, "step": 893 }, { "epoch": 0.17, "learning_rate": 4.961277919012003e-06, "loss": 0.5601, "step": 894 }, { "epoch": 0.17, "learning_rate": 4.961143700305549e-06, "loss": 0.5278, "step": 895 }, { "epoch": 0.17, "learning_rate": 4.961009251208368e-06, "loss": 0.6234, "step": 896 }, { "epoch": 0.17, "learning_rate": 4.960874571733047e-06, "loss": 0.6833, "step": 897 }, { "epoch": 0.17, "learning_rate": 4.960739661892193e-06, "loss": 0.6945, "step": 898 }, { "epoch": 0.17, "learning_rate": 4.9606045216984365e-06, "loss": 0.4474, "step": 899 }, { "epoch": 0.17, "learning_rate": 4.960469151164426e-06, "loss": 0.5842, "step": 900 }, { "epoch": 0.17, "learning_rate": 4.9603335503028335e-06, "loss": 0.5698, "step": 901 }, { "epoch": 0.17, "learning_rate": 4.9601977191263556e-06, "loss": 0.6572, "step": 902 }, { "epoch": 0.17, "learning_rate": 4.960061657647705e-06, "loss": 0.5918, "step": 903 }, { "epoch": 0.17, "learning_rate": 4.95992536587962e-06, "loss": 0.6452, "step": 904 }, { "epoch": 0.17, "learning_rate": 4.959788843834858e-06, "loss": 0.6204, "step": 905 }, { "epoch": 0.17, "learning_rate": 4.959652091526199e-06, "loss": 0.5946, "step": 906 }, { "epoch": 0.17, "learning_rate": 4.959515108966446e-06, "loss": 0.5904, "step": 907 }, { "epoch": 0.17, "learning_rate": 4.959377896168421e-06, "loss": 0.7734, "step": 908 }, { "epoch": 0.17, "learning_rate": 4.959240453144969e-06, "loss": 0.5135, "step": 909 }, { "epoch": 0.17, "learning_rate": 4.959102779908955e-06, "loss": 0.5411, "step": 910 }, { "epoch": 0.17, "learning_rate": 4.958964876473269e-06, "loss": 0.6505, "step": 911 }, { "epoch": 0.17, "learning_rate": 4.958826742850819e-06, "loss": 0.4978, "step": 912 }, { "epoch": 0.17, "learning_rate": 4.9586883790545345e-06, "loss": 0.5985, "step": 913 }, { "epoch": 0.17, "learning_rate": 4.958549785097371e-06, "loss": 0.6368, "step": 914 }, { "epoch": 0.17, "learning_rate": 4.9584109609923006e-06, "loss": 0.695, "step": 915 }, { "epoch": 0.17, "learning_rate": 4.9582719067523185e-06, "loss": 0.5091, "step": 916 }, { "epoch": 0.17, "learning_rate": 4.958132622390442e-06, "loss": 0.5002, "step": 917 }, { "epoch": 0.17, "learning_rate": 4.95799310791971e-06, "loss": 0.7569, "step": 918 }, { "epoch": 0.17, "learning_rate": 4.957853363353184e-06, "loss": 0.5638, "step": 919 }, { "epoch": 0.17, "learning_rate": 4.957713388703943e-06, "loss": 0.7089, "step": 920 }, { "epoch": 0.17, "learning_rate": 4.957573183985092e-06, "loss": 0.5681, "step": 921 }, { "epoch": 0.17, "learning_rate": 4.957432749209755e-06, "loss": 0.5849, "step": 922 }, { "epoch": 0.17, "learning_rate": 4.957292084391078e-06, "loss": 0.5675, "step": 923 }, { "epoch": 0.17, "learning_rate": 4.9571511895422294e-06, "loss": 0.3648, "step": 924 }, { "epoch": 0.17, "learning_rate": 4.957010064676398e-06, "loss": 0.7103, "step": 925 }, { "epoch": 0.17, "learning_rate": 4.956868709806797e-06, "loss": 0.5813, "step": 926 }, { "epoch": 0.18, "learning_rate": 4.956727124946655e-06, "loss": 0.4868, "step": 927 }, { "epoch": 0.18, "learning_rate": 4.956585310109229e-06, "loss": 0.6653, "step": 928 }, { "epoch": 0.18, "learning_rate": 4.956443265307792e-06, "loss": 0.5769, "step": 929 }, { "epoch": 0.18, "learning_rate": 4.956300990555643e-06, "loss": 0.5626, "step": 930 }, { "epoch": 0.18, "learning_rate": 4.956158485866099e-06, "loss": 0.5069, "step": 931 }, { "epoch": 0.18, "learning_rate": 4.956015751252502e-06, "loss": 0.5449, "step": 932 }, { "epoch": 0.18, "learning_rate": 4.955872786728212e-06, "loss": 0.5526, "step": 933 }, { "epoch": 0.18, "learning_rate": 4.955729592306613e-06, "loss": 0.6356, "step": 934 }, { "epoch": 0.18, "learning_rate": 4.955586168001108e-06, "loss": 0.5932, "step": 935 }, { "epoch": 0.18, "learning_rate": 4.955442513825125e-06, "loss": 0.4863, "step": 936 }, { "epoch": 0.18, "learning_rate": 4.95529862979211e-06, "loss": 0.5767, "step": 937 }, { "epoch": 0.18, "learning_rate": 4.955154515915533e-06, "loss": 0.5504, "step": 938 }, { "epoch": 0.18, "learning_rate": 4.955010172208886e-06, "loss": 0.6052, "step": 939 }, { "epoch": 0.18, "learning_rate": 4.954865598685679e-06, "loss": 0.5778, "step": 940 }, { "epoch": 0.18, "learning_rate": 4.954720795359446e-06, "loss": 0.8284, "step": 941 }, { "epoch": 0.18, "learning_rate": 4.954575762243745e-06, "loss": 0.5265, "step": 942 }, { "epoch": 0.18, "learning_rate": 4.954430499352148e-06, "loss": 0.7167, "step": 943 }, { "epoch": 0.18, "learning_rate": 4.954285006698258e-06, "loss": 0.3703, "step": 944 }, { "epoch": 0.18, "learning_rate": 4.954139284295691e-06, "loss": 0.7103, "step": 945 }, { "epoch": 0.18, "learning_rate": 4.9539933321580905e-06, "loss": 0.5411, "step": 946 }, { "epoch": 0.18, "learning_rate": 4.953847150299119e-06, "loss": 0.6294, "step": 947 }, { "epoch": 0.18, "learning_rate": 4.9537007387324595e-06, "loss": 0.6033, "step": 948 }, { "epoch": 0.18, "learning_rate": 4.953554097471819e-06, "loss": 0.5929, "step": 949 }, { "epoch": 0.18, "learning_rate": 4.953407226530924e-06, "loss": 0.5894, "step": 950 }, { "epoch": 0.18, "learning_rate": 4.953260125923523e-06, "loss": 0.5678, "step": 951 }, { "epoch": 0.18, "learning_rate": 4.9531127956633886e-06, "loss": 0.6685, "step": 952 }, { "epoch": 0.18, "learning_rate": 4.95296523576431e-06, "loss": 0.5664, "step": 953 }, { "epoch": 0.18, "learning_rate": 4.952817446240101e-06, "loss": 0.5736, "step": 954 }, { "epoch": 0.18, "learning_rate": 4.952669427104597e-06, "loss": 0.6019, "step": 955 }, { "epoch": 0.18, "learning_rate": 4.952521178371654e-06, "loss": 0.5536, "step": 956 }, { "epoch": 0.18, "learning_rate": 4.95237270005515e-06, "loss": 0.6431, "step": 957 }, { "epoch": 0.18, "learning_rate": 4.952223992168984e-06, "loss": 0.5594, "step": 958 }, { "epoch": 0.18, "learning_rate": 4.952075054727076e-06, "loss": 0.5973, "step": 959 }, { "epoch": 0.18, "learning_rate": 4.951925887743369e-06, "loss": 0.5228, "step": 960 }, { "epoch": 0.18, "learning_rate": 4.951776491231827e-06, "loss": 0.4993, "step": 961 }, { "epoch": 0.18, "learning_rate": 4.951626865206435e-06, "loss": 0.674, "step": 962 }, { "epoch": 0.18, "learning_rate": 4.951477009681199e-06, "loss": 0.5603, "step": 963 }, { "epoch": 0.18, "learning_rate": 4.951326924670148e-06, "loss": 0.6902, "step": 964 }, { "epoch": 0.18, "learning_rate": 4.951176610187331e-06, "loss": 0.5415, "step": 965 }, { "epoch": 0.18, "learning_rate": 4.9510260662468195e-06, "loss": 0.5537, "step": 966 }, { "epoch": 0.18, "learning_rate": 4.9508752928627065e-06, "loss": 0.5849, "step": 967 }, { "epoch": 0.18, "learning_rate": 4.950724290049105e-06, "loss": 0.589, "step": 968 }, { "epoch": 0.18, "learning_rate": 4.950573057820151e-06, "loss": 0.5472, "step": 969 }, { "epoch": 0.18, "learning_rate": 4.950421596190002e-06, "loss": 0.5292, "step": 970 }, { "epoch": 0.18, "learning_rate": 4.9502699051728365e-06, "loss": 0.562, "step": 971 }, { "epoch": 0.18, "learning_rate": 4.950117984782854e-06, "loss": 0.5238, "step": 972 }, { "epoch": 0.18, "learning_rate": 4.949965835034276e-06, "loss": 0.4386, "step": 973 }, { "epoch": 0.18, "learning_rate": 4.949813455941346e-06, "loss": 0.5677, "step": 974 }, { "epoch": 0.18, "learning_rate": 4.949660847518328e-06, "loss": 0.6885, "step": 975 }, { "epoch": 0.18, "learning_rate": 4.949508009779507e-06, "loss": 0.556, "step": 976 }, { "epoch": 0.18, "learning_rate": 4.949354942739192e-06, "loss": 0.7733, "step": 977 }, { "epoch": 0.18, "learning_rate": 4.94920164641171e-06, "loss": 0.6054, "step": 978 }, { "epoch": 0.18, "learning_rate": 4.949048120811413e-06, "loss": 0.5777, "step": 979 }, { "epoch": 0.19, "learning_rate": 4.9488943659526715e-06, "loss": 0.5596, "step": 980 }, { "epoch": 0.19, "learning_rate": 4.948740381849879e-06, "loss": 0.5232, "step": 981 }, { "epoch": 0.19, "learning_rate": 4.948586168517452e-06, "loss": 0.6399, "step": 982 }, { "epoch": 0.19, "learning_rate": 4.948431725969824e-06, "loss": 0.5957, "step": 983 }, { "epoch": 0.19, "learning_rate": 4.948277054221452e-06, "loss": 0.6572, "step": 984 }, { "epoch": 0.19, "learning_rate": 4.9481221532868175e-06, "loss": 0.6251, "step": 985 }, { "epoch": 0.19, "learning_rate": 4.94796702318042e-06, "loss": 0.5022, "step": 986 }, { "epoch": 0.19, "learning_rate": 4.947811663916782e-06, "loss": 0.5117, "step": 987 }, { "epoch": 0.19, "learning_rate": 4.947656075510445e-06, "loss": 0.6018, "step": 988 }, { "epoch": 0.19, "learning_rate": 4.947500257975976e-06, "loss": 0.7343, "step": 989 }, { "epoch": 0.19, "learning_rate": 4.94734421132796e-06, "loss": 0.4866, "step": 990 }, { "epoch": 0.19, "learning_rate": 4.947187935581005e-06, "loss": 0.6585, "step": 991 }, { "epoch": 0.19, "learning_rate": 4.9470314307497395e-06, "loss": 0.6245, "step": 992 }, { "epoch": 0.19, "learning_rate": 4.946874696848816e-06, "loss": 0.5593, "step": 993 }, { "epoch": 0.19, "learning_rate": 4.946717733892904e-06, "loss": 0.5504, "step": 994 }, { "epoch": 0.19, "learning_rate": 4.9465605418967e-06, "loss": 0.5393, "step": 995 }, { "epoch": 0.19, "learning_rate": 4.9464031208749165e-06, "loss": 0.6646, "step": 996 }, { "epoch": 0.19, "learning_rate": 4.9462454708422906e-06, "loss": 0.6789, "step": 997 }, { "epoch": 0.19, "learning_rate": 4.946087591813581e-06, "loss": 0.5656, "step": 998 }, { "epoch": 0.19, "learning_rate": 4.945929483803565e-06, "loss": 0.5213, "step": 999 }, { "epoch": 0.19, "learning_rate": 4.945771146827045e-06, "loss": 0.5675, "step": 1000 }, { "epoch": 0.19, "learning_rate": 4.945612580898843e-06, "loss": 0.6907, "step": 1001 }, { "epoch": 0.19, "learning_rate": 4.945453786033801e-06, "loss": 0.56, "step": 1002 }, { "epoch": 0.19, "learning_rate": 4.945294762246786e-06, "loss": 0.5886, "step": 1003 }, { "epoch": 0.19, "learning_rate": 4.945135509552683e-06, "loss": 0.6047, "step": 1004 }, { "epoch": 0.19, "learning_rate": 4.944976027966401e-06, "loss": 0.523, "step": 1005 }, { "epoch": 0.19, "learning_rate": 4.9448163175028685e-06, "loss": 0.4645, "step": 1006 }, { "epoch": 0.19, "learning_rate": 4.944656378177036e-06, "loss": 0.5346, "step": 1007 }, { "epoch": 0.19, "learning_rate": 4.944496210003876e-06, "loss": 0.4798, "step": 1008 }, { "epoch": 0.19, "learning_rate": 4.944335812998382e-06, "loss": 0.5602, "step": 1009 }, { "epoch": 0.19, "learning_rate": 4.944175187175569e-06, "loss": 0.5246, "step": 1010 }, { "epoch": 0.19, "learning_rate": 4.944014332550473e-06, "loss": 0.6319, "step": 1011 }, { "epoch": 0.19, "learning_rate": 4.943853249138153e-06, "loss": 0.6521, "step": 1012 }, { "epoch": 0.19, "learning_rate": 4.9436919369536864e-06, "loss": 0.614, "step": 1013 }, { "epoch": 0.19, "learning_rate": 4.943530396012175e-06, "loss": 0.4612, "step": 1014 }, { "epoch": 0.19, "learning_rate": 4.943368626328741e-06, "loss": 0.6588, "step": 1015 }, { "epoch": 0.19, "learning_rate": 4.9432066279185276e-06, "loss": 0.7137, "step": 1016 }, { "epoch": 0.19, "learning_rate": 4.943044400796699e-06, "loss": 0.5173, "step": 1017 }, { "epoch": 0.19, "learning_rate": 4.942881944978444e-06, "loss": 0.761, "step": 1018 }, { "epoch": 0.19, "learning_rate": 4.942719260478966e-06, "loss": 0.6483, "step": 1019 }, { "epoch": 0.19, "learning_rate": 4.942556347313497e-06, "loss": 0.4534, "step": 1020 }, { "epoch": 0.19, "learning_rate": 4.942393205497288e-06, "loss": 0.6584, "step": 1021 }, { "epoch": 0.19, "learning_rate": 4.942229835045609e-06, "loss": 0.5754, "step": 1022 }, { "epoch": 0.19, "learning_rate": 4.942066235973753e-06, "loss": 0.5946, "step": 1023 }, { "epoch": 0.19, "learning_rate": 4.941902408297038e-06, "loss": 0.6186, "step": 1024 }, { "epoch": 0.19, "learning_rate": 4.941738352030797e-06, "loss": 0.6042, "step": 1025 }, { "epoch": 0.19, "learning_rate": 4.941574067190389e-06, "loss": 0.5281, "step": 1026 }, { "epoch": 0.19, "learning_rate": 4.941409553791192e-06, "loss": 0.369, "step": 1027 }, { "epoch": 0.19, "learning_rate": 4.941244811848607e-06, "loss": 0.564, "step": 1028 }, { "epoch": 0.19, "learning_rate": 4.9410798413780546e-06, "loss": 0.6234, "step": 1029 }, { "epoch": 0.19, "learning_rate": 4.940914642394979e-06, "loss": 0.6412, "step": 1030 }, { "epoch": 0.19, "learning_rate": 4.940749214914845e-06, "loss": 0.6182, "step": 1031 }, { "epoch": 0.19, "learning_rate": 4.940583558953138e-06, "loss": 0.7293, "step": 1032 }, { "epoch": 0.2, "learning_rate": 4.940417674525364e-06, "loss": 0.5649, "step": 1033 }, { "epoch": 0.2, "learning_rate": 4.940251561647054e-06, "loss": 0.5613, "step": 1034 }, { "epoch": 0.2, "learning_rate": 4.940085220333757e-06, "loss": 0.5851, "step": 1035 }, { "epoch": 0.2, "learning_rate": 4.939918650601043e-06, "loss": 0.5149, "step": 1036 }, { "epoch": 0.2, "learning_rate": 4.939751852464507e-06, "loss": 0.5265, "step": 1037 }, { "epoch": 0.2, "learning_rate": 4.9395848259397615e-06, "loss": 0.5597, "step": 1038 }, { "epoch": 0.2, "learning_rate": 4.939417571042443e-06, "loss": 0.5823, "step": 1039 }, { "epoch": 0.2, "learning_rate": 4.939250087788208e-06, "loss": 0.6568, "step": 1040 }, { "epoch": 0.2, "learning_rate": 4.939082376192735e-06, "loss": 0.548, "step": 1041 }, { "epoch": 0.2, "learning_rate": 4.9389144362717245e-06, "loss": 0.7159, "step": 1042 }, { "epoch": 0.2, "learning_rate": 4.938746268040896e-06, "loss": 0.5247, "step": 1043 }, { "epoch": 0.2, "learning_rate": 4.938577871515993e-06, "loss": 0.6437, "step": 1044 }, { "epoch": 0.2, "learning_rate": 4.93840924671278e-06, "loss": 0.5933, "step": 1045 }, { "epoch": 0.2, "learning_rate": 4.93824039364704e-06, "loss": 0.6193, "step": 1046 }, { "epoch": 0.2, "learning_rate": 4.93807131233458e-06, "loss": 0.4508, "step": 1047 }, { "epoch": 0.2, "learning_rate": 4.9379020027912305e-06, "loss": 0.7454, "step": 1048 }, { "epoch": 0.2, "learning_rate": 4.937732465032838e-06, "loss": 0.557, "step": 1049 }, { "epoch": 0.2, "learning_rate": 4.937562699075274e-06, "loss": 0.4324, "step": 1050 }, { "epoch": 0.2, "learning_rate": 4.937392704934431e-06, "loss": 0.6337, "step": 1051 }, { "epoch": 0.2, "learning_rate": 4.9372224826262216e-06, "loss": 0.6242, "step": 1052 }, { "epoch": 0.2, "learning_rate": 4.937052032166581e-06, "loss": 0.5962, "step": 1053 }, { "epoch": 0.2, "learning_rate": 4.936881353571466e-06, "loss": 0.5471, "step": 1054 }, { "epoch": 0.2, "learning_rate": 4.936710446856851e-06, "loss": 0.6264, "step": 1055 }, { "epoch": 0.2, "learning_rate": 4.936539312038739e-06, "loss": 0.5006, "step": 1056 }, { "epoch": 0.2, "learning_rate": 4.936367949133147e-06, "loss": 0.488, "step": 1057 }, { "epoch": 0.2, "learning_rate": 4.936196358156118e-06, "loss": 0.4935, "step": 1058 }, { "epoch": 0.2, "learning_rate": 4.936024539123715e-06, "loss": 0.5917, "step": 1059 }, { "epoch": 0.2, "learning_rate": 4.935852492052021e-06, "loss": 0.6988, "step": 1060 }, { "epoch": 0.2, "learning_rate": 4.935680216957142e-06, "loss": 0.6366, "step": 1061 }, { "epoch": 0.2, "learning_rate": 4.935507713855205e-06, "loss": 0.4681, "step": 1062 }, { "epoch": 0.2, "learning_rate": 4.93533498276236e-06, "loss": 0.488, "step": 1063 }, { "epoch": 0.2, "learning_rate": 4.935162023694773e-06, "loss": 0.5025, "step": 1064 }, { "epoch": 0.2, "learning_rate": 4.9349888366686376e-06, "loss": 0.5734, "step": 1065 }, { "epoch": 0.2, "learning_rate": 4.934815421700166e-06, "loss": 0.4219, "step": 1066 }, { "epoch": 0.2, "learning_rate": 4.934641778805589e-06, "loss": 0.6705, "step": 1067 }, { "epoch": 0.2, "learning_rate": 4.934467908001165e-06, "loss": 0.5473, "step": 1068 }, { "epoch": 0.2, "learning_rate": 4.9342938093031686e-06, "loss": 0.5889, "step": 1069 }, { "epoch": 0.2, "learning_rate": 4.934119482727897e-06, "loss": 0.6812, "step": 1070 }, { "epoch": 0.2, "learning_rate": 4.9339449282916705e-06, "loss": 0.6537, "step": 1071 }, { "epoch": 0.2, "learning_rate": 4.9337701460108285e-06, "loss": 0.4684, "step": 1072 }, { "epoch": 0.2, "learning_rate": 4.933595135901733e-06, "loss": 0.6257, "step": 1073 }, { "epoch": 0.2, "learning_rate": 4.933419897980766e-06, "loss": 0.6084, "step": 1074 }, { "epoch": 0.2, "learning_rate": 4.933244432264333e-06, "loss": 0.5462, "step": 1075 }, { "epoch": 0.2, "learning_rate": 4.933068738768858e-06, "loss": 0.581, "step": 1076 }, { "epoch": 0.2, "learning_rate": 4.932892817510789e-06, "loss": 0.5719, "step": 1077 }, { "epoch": 0.2, "learning_rate": 4.9327166685065944e-06, "loss": 0.5577, "step": 1078 }, { "epoch": 0.2, "learning_rate": 4.932540291772763e-06, "loss": 0.6809, "step": 1079 }, { "epoch": 0.2, "learning_rate": 4.932363687325807e-06, "loss": 0.6563, "step": 1080 }, { "epoch": 0.2, "learning_rate": 4.932186855182256e-06, "loss": 0.6622, "step": 1081 }, { "epoch": 0.2, "learning_rate": 4.932009795358666e-06, "loss": 0.6037, "step": 1082 }, { "epoch": 0.2, "learning_rate": 4.931832507871611e-06, "loss": 0.6053, "step": 1083 }, { "epoch": 0.2, "learning_rate": 4.931654992737686e-06, "loss": 0.6004, "step": 1084 }, { "epoch": 0.2, "learning_rate": 4.9314772499735094e-06, "loss": 0.5488, "step": 1085 }, { "epoch": 0.21, "learning_rate": 4.931299279595721e-06, "loss": 0.6135, "step": 1086 }, { "epoch": 0.21, "learning_rate": 4.931121081620978e-06, "loss": 0.5006, "step": 1087 }, { "epoch": 0.21, "learning_rate": 4.9309426560659655e-06, "loss": 0.6732, "step": 1088 }, { "epoch": 0.21, "learning_rate": 4.9307640029473826e-06, "loss": 0.605, "step": 1089 }, { "epoch": 0.21, "learning_rate": 4.930585122281956e-06, "loss": 0.5596, "step": 1090 }, { "epoch": 0.21, "learning_rate": 4.930406014086428e-06, "loss": 0.5708, "step": 1091 }, { "epoch": 0.21, "learning_rate": 4.930226678377568e-06, "loss": 0.5677, "step": 1092 }, { "epoch": 0.21, "learning_rate": 4.930047115172162e-06, "loss": 0.5272, "step": 1093 }, { "epoch": 0.21, "learning_rate": 4.929867324487019e-06, "loss": 0.5646, "step": 1094 }, { "epoch": 0.21, "learning_rate": 4.9296873063389715e-06, "loss": 0.5356, "step": 1095 }, { "epoch": 0.21, "learning_rate": 4.9295070607448695e-06, "loss": 0.7101, "step": 1096 }, { "epoch": 0.21, "learning_rate": 4.929326587721586e-06, "loss": 0.8068, "step": 1097 }, { "epoch": 0.21, "learning_rate": 4.929145887286017e-06, "loss": 0.5315, "step": 1098 }, { "epoch": 0.21, "learning_rate": 4.928964959455075e-06, "loss": 0.5868, "step": 1099 }, { "epoch": 0.21, "learning_rate": 4.9287838042457e-06, "loss": 0.5719, "step": 1100 }, { "epoch": 0.21, "learning_rate": 4.928602421674849e-06, "loss": 0.5656, "step": 1101 }, { "epoch": 0.21, "learning_rate": 4.9284208117595e-06, "loss": 0.528, "step": 1102 }, { "epoch": 0.21, "learning_rate": 4.928238974516656e-06, "loss": 0.5618, "step": 1103 }, { "epoch": 0.21, "learning_rate": 4.928056909963338e-06, "loss": 0.4753, "step": 1104 }, { "epoch": 0.21, "learning_rate": 4.927874618116589e-06, "loss": 0.5092, "step": 1105 }, { "epoch": 0.21, "learning_rate": 4.927692098993474e-06, "loss": 0.4398, "step": 1106 }, { "epoch": 0.21, "learning_rate": 4.927509352611078e-06, "loss": 0.5692, "step": 1107 }, { "epoch": 0.21, "learning_rate": 4.9273263789865105e-06, "loss": 0.7236, "step": 1108 }, { "epoch": 0.21, "learning_rate": 4.927143178136897e-06, "loss": 0.5763, "step": 1109 }, { "epoch": 0.21, "learning_rate": 4.926959750079389e-06, "loss": 0.4626, "step": 1110 }, { "epoch": 0.21, "learning_rate": 4.926776094831157e-06, "loss": 0.5762, "step": 1111 }, { "epoch": 0.21, "learning_rate": 4.926592212409393e-06, "loss": 0.5756, "step": 1112 }, { "epoch": 0.21, "learning_rate": 4.92640810283131e-06, "loss": 0.5629, "step": 1113 }, { "epoch": 0.21, "learning_rate": 4.926223766114143e-06, "loss": 0.5607, "step": 1114 }, { "epoch": 0.21, "learning_rate": 4.926039202275149e-06, "loss": 0.6372, "step": 1115 }, { "epoch": 0.21, "learning_rate": 4.9258544113316045e-06, "loss": 0.5954, "step": 1116 }, { "epoch": 0.21, "learning_rate": 4.925669393300808e-06, "loss": 0.4867, "step": 1117 }, { "epoch": 0.21, "learning_rate": 4.925484148200078e-06, "loss": 0.6455, "step": 1118 }, { "epoch": 0.21, "learning_rate": 4.925298676046758e-06, "loss": 0.535, "step": 1119 }, { "epoch": 0.21, "learning_rate": 4.925112976858209e-06, "loss": 0.5415, "step": 1120 }, { "epoch": 0.21, "learning_rate": 4.9249270506518144e-06, "loss": 0.6022, "step": 1121 }, { "epoch": 0.21, "learning_rate": 4.924740897444979e-06, "loss": 0.7029, "step": 1122 }, { "epoch": 0.21, "learning_rate": 4.924554517255129e-06, "loss": 0.5441, "step": 1123 }, { "epoch": 0.21, "learning_rate": 4.924367910099712e-06, "loss": 0.4847, "step": 1124 }, { "epoch": 0.21, "learning_rate": 4.924181075996196e-06, "loss": 0.5534, "step": 1125 }, { "epoch": 0.21, "learning_rate": 4.9239940149620705e-06, "loss": 0.6948, "step": 1126 }, { "epoch": 0.21, "learning_rate": 4.923806727014847e-06, "loss": 0.614, "step": 1127 }, { "epoch": 0.21, "learning_rate": 4.923619212172059e-06, "loss": 0.6101, "step": 1128 }, { "epoch": 0.21, "learning_rate": 4.923431470451258e-06, "loss": 0.6022, "step": 1129 }, { "epoch": 0.21, "learning_rate": 4.92324350187002e-06, "loss": 0.5337, "step": 1130 }, { "epoch": 0.21, "learning_rate": 4.923055306445939e-06, "loss": 0.4228, "step": 1131 }, { "epoch": 0.21, "learning_rate": 4.922866884196635e-06, "loss": 0.5092, "step": 1132 }, { "epoch": 0.21, "learning_rate": 4.922678235139745e-06, "loss": 0.5259, "step": 1133 }, { "epoch": 0.21, "learning_rate": 4.922489359292928e-06, "loss": 0.8329, "step": 1134 }, { "epoch": 0.21, "learning_rate": 4.922300256673867e-06, "loss": 0.8043, "step": 1135 }, { "epoch": 0.21, "learning_rate": 4.922110927300261e-06, "loss": 0.7201, "step": 1136 }, { "epoch": 0.21, "learning_rate": 4.921921371189836e-06, "loss": 0.5795, "step": 1137 }, { "epoch": 0.22, "learning_rate": 4.921731588360337e-06, "loss": 0.471, "step": 1138 }, { "epoch": 0.22, "learning_rate": 4.9215415788295265e-06, "loss": 0.5721, "step": 1139 }, { "epoch": 0.22, "learning_rate": 4.921351342615195e-06, "loss": 0.4782, "step": 1140 }, { "epoch": 0.22, "learning_rate": 4.921160879735149e-06, "loss": 0.4651, "step": 1141 }, { "epoch": 0.22, "learning_rate": 4.920970190207219e-06, "loss": 0.4804, "step": 1142 }, { "epoch": 0.22, "learning_rate": 4.920779274049254e-06, "loss": 0.4247, "step": 1143 }, { "epoch": 0.22, "learning_rate": 4.920588131279127e-06, "loss": 0.5666, "step": 1144 }, { "epoch": 0.22, "learning_rate": 4.920396761914731e-06, "loss": 0.5984, "step": 1145 }, { "epoch": 0.22, "learning_rate": 4.92020516597398e-06, "loss": 0.505, "step": 1146 }, { "epoch": 0.22, "learning_rate": 4.920013343474811e-06, "loss": 0.6008, "step": 1147 }, { "epoch": 0.22, "learning_rate": 4.919821294435179e-06, "loss": 0.4602, "step": 1148 }, { "epoch": 0.22, "learning_rate": 4.919629018873063e-06, "loss": 0.4604, "step": 1149 }, { "epoch": 0.22, "learning_rate": 4.919436516806461e-06, "loss": 0.5006, "step": 1150 }, { "epoch": 0.22, "learning_rate": 4.919243788253394e-06, "loss": 0.6163, "step": 1151 }, { "epoch": 0.22, "learning_rate": 4.919050833231904e-06, "loss": 0.7251, "step": 1152 }, { "epoch": 0.22, "learning_rate": 4.918857651760054e-06, "loss": 0.5763, "step": 1153 }, { "epoch": 0.22, "learning_rate": 4.918664243855926e-06, "loss": 0.5851, "step": 1154 }, { "epoch": 0.22, "learning_rate": 4.918470609537628e-06, "loss": 0.5113, "step": 1155 }, { "epoch": 0.22, "learning_rate": 4.918276748823284e-06, "loss": 0.5355, "step": 1156 }, { "epoch": 0.22, "learning_rate": 4.918082661731042e-06, "loss": 0.4835, "step": 1157 }, { "epoch": 0.22, "learning_rate": 4.9178883482790716e-06, "loss": 0.5236, "step": 1158 }, { "epoch": 0.22, "learning_rate": 4.917693808485563e-06, "loss": 0.5575, "step": 1159 }, { "epoch": 0.22, "learning_rate": 4.917499042368726e-06, "loss": 0.5675, "step": 1160 }, { "epoch": 0.22, "learning_rate": 4.917304049946793e-06, "loss": 0.6985, "step": 1161 }, { "epoch": 0.22, "learning_rate": 4.917108831238019e-06, "loss": 0.4233, "step": 1162 }, { "epoch": 0.22, "learning_rate": 4.9169133862606765e-06, "loss": 0.4696, "step": 1163 }, { "epoch": 0.22, "learning_rate": 4.916717715033062e-06, "loss": 0.4746, "step": 1164 }, { "epoch": 0.22, "learning_rate": 4.916521817573496e-06, "loss": 0.5838, "step": 1165 }, { "epoch": 0.22, "learning_rate": 4.91632569390031e-06, "loss": 0.5511, "step": 1166 }, { "epoch": 0.22, "learning_rate": 4.916129344031869e-06, "loss": 0.6571, "step": 1167 }, { "epoch": 0.22, "learning_rate": 4.915932767986552e-06, "loss": 0.3974, "step": 1168 }, { "epoch": 0.22, "learning_rate": 4.915735965782759e-06, "loss": 0.6082, "step": 1169 }, { "epoch": 0.22, "learning_rate": 4.915538937438915e-06, "loss": 0.5835, "step": 1170 }, { "epoch": 0.22, "learning_rate": 4.915341682973463e-06, "loss": 0.5712, "step": 1171 }, { "epoch": 0.22, "learning_rate": 4.915144202404869e-06, "loss": 0.4922, "step": 1172 }, { "epoch": 0.22, "learning_rate": 4.914946495751619e-06, "loss": 0.6539, "step": 1173 }, { "epoch": 0.22, "learning_rate": 4.914748563032221e-06, "loss": 0.6626, "step": 1174 }, { "epoch": 0.22, "learning_rate": 4.914550404265202e-06, "loss": 0.6179, "step": 1175 }, { "epoch": 0.22, "learning_rate": 4.914352019469114e-06, "loss": 0.4988, "step": 1176 }, { "epoch": 0.22, "learning_rate": 4.914153408662528e-06, "loss": 0.5931, "step": 1177 }, { "epoch": 0.22, "learning_rate": 4.913954571864033e-06, "loss": 0.7451, "step": 1178 }, { "epoch": 0.22, "learning_rate": 4.913755509092247e-06, "loss": 0.4693, "step": 1179 }, { "epoch": 0.22, "learning_rate": 4.913556220365801e-06, "loss": 0.5455, "step": 1180 }, { "epoch": 0.22, "learning_rate": 4.913356705703353e-06, "loss": 0.5783, "step": 1181 }, { "epoch": 0.22, "learning_rate": 4.913156965123578e-06, "loss": 0.5515, "step": 1182 }, { "epoch": 0.22, "learning_rate": 4.912956998645175e-06, "loss": 0.4475, "step": 1183 }, { "epoch": 0.22, "learning_rate": 4.912756806286862e-06, "loss": 0.5256, "step": 1184 }, { "epoch": 0.22, "learning_rate": 4.9125563880673814e-06, "loss": 0.7077, "step": 1185 }, { "epoch": 0.22, "learning_rate": 4.912355744005492e-06, "loss": 0.7626, "step": 1186 }, { "epoch": 0.22, "learning_rate": 4.912154874119979e-06, "loss": 0.4968, "step": 1187 }, { "epoch": 0.22, "learning_rate": 4.911953778429644e-06, "loss": 0.6552, "step": 1188 }, { "epoch": 0.22, "learning_rate": 4.911752456953312e-06, "loss": 0.5573, "step": 1189 }, { "epoch": 0.22, "learning_rate": 4.911550909709832e-06, "loss": 0.4994, "step": 1190 }, { "epoch": 0.23, "learning_rate": 4.911349136718066e-06, "loss": 0.5869, "step": 1191 }, { "epoch": 0.23, "learning_rate": 4.911147137996906e-06, "loss": 0.5774, "step": 1192 }, { "epoch": 0.23, "learning_rate": 4.910944913565259e-06, "loss": 0.4238, "step": 1193 }, { "epoch": 0.23, "learning_rate": 4.910742463442058e-06, "loss": 0.5272, "step": 1194 }, { "epoch": 0.23, "learning_rate": 4.910539787646252e-06, "loss": 0.5952, "step": 1195 }, { "epoch": 0.23, "learning_rate": 4.910336886196815e-06, "loss": 0.4722, "step": 1196 }, { "epoch": 0.23, "learning_rate": 4.910133759112742e-06, "loss": 0.6603, "step": 1197 }, { "epoch": 0.23, "learning_rate": 4.909930406413046e-06, "loss": 0.5168, "step": 1198 }, { "epoch": 0.23, "learning_rate": 4.909726828116764e-06, "loss": 0.5013, "step": 1199 }, { "epoch": 0.23, "learning_rate": 4.909523024242953e-06, "loss": 0.5678, "step": 1200 }, { "epoch": 0.23, "learning_rate": 4.909318994810691e-06, "loss": 0.5027, "step": 1201 }, { "epoch": 0.23, "learning_rate": 4.90911473983908e-06, "loss": 0.5362, "step": 1202 }, { "epoch": 0.23, "learning_rate": 4.908910259347236e-06, "loss": 0.591, "step": 1203 }, { "epoch": 0.23, "learning_rate": 4.908705553354304e-06, "loss": 0.485, "step": 1204 }, { "epoch": 0.23, "learning_rate": 4.9085006218794465e-06, "loss": 0.6202, "step": 1205 }, { "epoch": 0.23, "learning_rate": 4.908295464941846e-06, "loss": 0.5967, "step": 1206 }, { "epoch": 0.23, "learning_rate": 4.9080900825607094e-06, "loss": 0.5443, "step": 1207 }, { "epoch": 0.23, "learning_rate": 4.90788447475526e-06, "loss": 0.6251, "step": 1208 }, { "epoch": 0.23, "learning_rate": 4.9076786415447484e-06, "loss": 0.6859, "step": 1209 }, { "epoch": 0.23, "learning_rate": 4.9074725829484405e-06, "loss": 0.4799, "step": 1210 }, { "epoch": 0.23, "learning_rate": 4.907266298985626e-06, "loss": 0.6647, "step": 1211 }, { "epoch": 0.23, "learning_rate": 4.907059789675617e-06, "loss": 0.6634, "step": 1212 }, { "epoch": 0.23, "learning_rate": 4.906853055037743e-06, "loss": 0.584, "step": 1213 }, { "epoch": 0.23, "learning_rate": 4.906646095091357e-06, "loss": 0.7165, "step": 1214 }, { "epoch": 0.23, "learning_rate": 4.906438909855835e-06, "loss": 0.5622, "step": 1215 }, { "epoch": 0.23, "learning_rate": 4.90623149935057e-06, "loss": 0.4486, "step": 1216 }, { "epoch": 0.23, "learning_rate": 4.906023863594977e-06, "loss": 0.6408, "step": 1217 }, { "epoch": 0.23, "learning_rate": 4.905816002608495e-06, "loss": 0.5571, "step": 1218 }, { "epoch": 0.23, "learning_rate": 4.905607916410581e-06, "loss": 0.5442, "step": 1219 }, { "epoch": 0.23, "learning_rate": 4.905399605020716e-06, "loss": 0.6334, "step": 1220 }, { "epoch": 0.23, "learning_rate": 4.905191068458397e-06, "loss": 0.5958, "step": 1221 }, { "epoch": 0.23, "learning_rate": 4.904982306743148e-06, "loss": 0.4318, "step": 1222 }, { "epoch": 0.23, "learning_rate": 4.90477331989451e-06, "loss": 0.5036, "step": 1223 }, { "epoch": 0.23, "learning_rate": 4.904564107932048e-06, "loss": 0.4997, "step": 1224 }, { "epoch": 0.23, "learning_rate": 4.904354670875345e-06, "loss": 0.6187, "step": 1225 }, { "epoch": 0.23, "learning_rate": 4.904145008744009e-06, "loss": 0.6192, "step": 1226 }, { "epoch": 0.23, "learning_rate": 4.903935121557663e-06, "loss": 0.5826, "step": 1227 }, { "epoch": 0.23, "learning_rate": 4.903725009335959e-06, "loss": 0.613, "step": 1228 }, { "epoch": 0.23, "learning_rate": 4.903514672098563e-06, "loss": 0.4803, "step": 1229 }, { "epoch": 0.23, "learning_rate": 4.903304109865166e-06, "loss": 0.6416, "step": 1230 }, { "epoch": 0.23, "learning_rate": 4.903093322655479e-06, "loss": 0.4743, "step": 1231 }, { "epoch": 0.23, "learning_rate": 4.902882310489233e-06, "loss": 0.5625, "step": 1232 }, { "epoch": 0.23, "learning_rate": 4.902671073386183e-06, "loss": 0.708, "step": 1233 }, { "epoch": 0.23, "learning_rate": 4.902459611366101e-06, "loss": 0.6921, "step": 1234 }, { "epoch": 0.23, "learning_rate": 4.902247924448784e-06, "loss": 0.4624, "step": 1235 }, { "epoch": 0.23, "learning_rate": 4.902036012654048e-06, "loss": 0.4918, "step": 1236 }, { "epoch": 0.23, "learning_rate": 4.901823876001729e-06, "loss": 0.5379, "step": 1237 }, { "epoch": 0.23, "learning_rate": 4.901611514511687e-06, "loss": 0.5404, "step": 1238 }, { "epoch": 0.23, "learning_rate": 4.9013989282038e-06, "loss": 0.4602, "step": 1239 }, { "epoch": 0.23, "learning_rate": 4.9011861170979705e-06, "loss": 0.5703, "step": 1240 }, { "epoch": 0.23, "learning_rate": 4.9009730812141175e-06, "loss": 0.5627, "step": 1241 }, { "epoch": 0.23, "learning_rate": 4.900759820572185e-06, "loss": 0.6245, "step": 1242 }, { "epoch": 0.23, "learning_rate": 4.900546335192137e-06, "loss": 0.6358, "step": 1243 }, { "epoch": 0.24, "learning_rate": 4.900332625093957e-06, "loss": 0.5802, "step": 1244 }, { "epoch": 0.24, "learning_rate": 4.9001186902976505e-06, "loss": 0.555, "step": 1245 }, { "epoch": 0.24, "learning_rate": 4.899904530823246e-06, "loss": 0.5689, "step": 1246 }, { "epoch": 0.24, "learning_rate": 4.89969014669079e-06, "loss": 0.5642, "step": 1247 }, { "epoch": 0.24, "learning_rate": 4.89947553792035e-06, "loss": 0.6532, "step": 1248 }, { "epoch": 0.24, "learning_rate": 4.899260704532019e-06, "loss": 0.4233, "step": 1249 }, { "epoch": 0.24, "learning_rate": 4.899045646545905e-06, "loss": 0.5102, "step": 1250 }, { "epoch": 0.24, "learning_rate": 4.898830363982141e-06, "loss": 0.4921, "step": 1251 }, { "epoch": 0.24, "learning_rate": 4.89861485686088e-06, "loss": 0.4441, "step": 1252 }, { "epoch": 0.24, "learning_rate": 4.898399125202296e-06, "loss": 0.6049, "step": 1253 }, { "epoch": 0.24, "learning_rate": 4.898183169026583e-06, "loss": 0.6329, "step": 1254 }, { "epoch": 0.24, "learning_rate": 4.897966988353956e-06, "loss": 0.4817, "step": 1255 }, { "epoch": 0.24, "learning_rate": 4.897750583204655e-06, "loss": 0.5642, "step": 1256 }, { "epoch": 0.24, "learning_rate": 4.897533953598936e-06, "loss": 0.5644, "step": 1257 }, { "epoch": 0.24, "learning_rate": 4.897317099557078e-06, "loss": 0.4635, "step": 1258 }, { "epoch": 0.24, "learning_rate": 4.897100021099382e-06, "loss": 0.5468, "step": 1259 }, { "epoch": 0.24, "learning_rate": 4.896882718246168e-06, "loss": 0.5785, "step": 1260 }, { "epoch": 0.24, "learning_rate": 4.896665191017777e-06, "loss": 0.523, "step": 1261 }, { "epoch": 0.24, "learning_rate": 4.8964474394345746e-06, "loss": 0.5142, "step": 1262 }, { "epoch": 0.24, "learning_rate": 4.896229463516943e-06, "loss": 0.5453, "step": 1263 }, { "epoch": 0.24, "learning_rate": 4.8960112632852876e-06, "loss": 0.5299, "step": 1264 }, { "epoch": 0.24, "learning_rate": 4.895792838760035e-06, "loss": 0.6258, "step": 1265 }, { "epoch": 0.24, "learning_rate": 4.89557418996163e-06, "loss": 0.4216, "step": 1266 }, { "epoch": 0.24, "learning_rate": 4.895355316910544e-06, "loss": 0.4677, "step": 1267 }, { "epoch": 0.24, "learning_rate": 4.895136219627264e-06, "loss": 0.5711, "step": 1268 }, { "epoch": 0.24, "learning_rate": 4.894916898132299e-06, "loss": 0.568, "step": 1269 }, { "epoch": 0.24, "learning_rate": 4.894697352446182e-06, "loss": 0.4993, "step": 1270 }, { "epoch": 0.24, "learning_rate": 4.894477582589464e-06, "loss": 0.6899, "step": 1271 }, { "epoch": 0.24, "learning_rate": 4.894257588582718e-06, "loss": 0.54, "step": 1272 }, { "epoch": 0.24, "learning_rate": 4.894037370446538e-06, "loss": 0.5111, "step": 1273 }, { "epoch": 0.24, "learning_rate": 4.893816928201538e-06, "loss": 0.5095, "step": 1274 }, { "epoch": 0.24, "learning_rate": 4.893596261868355e-06, "loss": 0.6624, "step": 1275 }, { "epoch": 0.24, "learning_rate": 4.893375371467646e-06, "loss": 0.5891, "step": 1276 }, { "epoch": 0.24, "learning_rate": 4.893154257020089e-06, "loss": 0.6728, "step": 1277 }, { "epoch": 0.24, "learning_rate": 4.892932918546381e-06, "loss": 0.4541, "step": 1278 }, { "epoch": 0.24, "learning_rate": 4.8927113560672424e-06, "loss": 0.5151, "step": 1279 }, { "epoch": 0.24, "learning_rate": 4.892489569603415e-06, "loss": 0.5883, "step": 1280 }, { "epoch": 0.24, "learning_rate": 4.8922675591756606e-06, "loss": 0.6839, "step": 1281 }, { "epoch": 0.24, "learning_rate": 4.8920453248047605e-06, "loss": 0.5904, "step": 1282 }, { "epoch": 0.24, "learning_rate": 4.89182286651152e-06, "loss": 0.5812, "step": 1283 }, { "epoch": 0.24, "learning_rate": 4.891600184316761e-06, "loss": 0.498, "step": 1284 }, { "epoch": 0.24, "learning_rate": 4.891377278241333e-06, "loss": 0.8102, "step": 1285 }, { "epoch": 0.24, "learning_rate": 4.891154148306099e-06, "loss": 0.5337, "step": 1286 }, { "epoch": 0.24, "learning_rate": 4.8909307945319474e-06, "loss": 0.7024, "step": 1287 }, { "epoch": 0.24, "learning_rate": 4.8907072169397885e-06, "loss": 0.5644, "step": 1288 }, { "epoch": 0.24, "learning_rate": 4.890483415550549e-06, "loss": 0.483, "step": 1289 }, { "epoch": 0.24, "learning_rate": 4.8902593903851805e-06, "loss": 0.5441, "step": 1290 }, { "epoch": 0.24, "learning_rate": 4.890035141464655e-06, "loss": 0.5282, "step": 1291 }, { "epoch": 0.24, "learning_rate": 4.889810668809964e-06, "loss": 0.5813, "step": 1292 }, { "epoch": 0.24, "learning_rate": 4.889585972442119e-06, "loss": 0.4295, "step": 1293 }, { "epoch": 0.24, "learning_rate": 4.8893610523821574e-06, "loss": 0.578, "step": 1294 }, { "epoch": 0.24, "learning_rate": 4.889135908651132e-06, "loss": 0.6274, "step": 1295 }, { "epoch": 0.24, "learning_rate": 4.888910541270119e-06, "loss": 0.5749, "step": 1296 }, { "epoch": 0.25, "learning_rate": 4.888684950260216e-06, "loss": 0.5418, "step": 1297 }, { "epoch": 0.25, "learning_rate": 4.88845913564254e-06, "loss": 0.5664, "step": 1298 }, { "epoch": 0.25, "learning_rate": 4.8882330974382304e-06, "loss": 0.4027, "step": 1299 }, { "epoch": 0.25, "learning_rate": 4.888006835668448e-06, "loss": 0.7031, "step": 1300 }, { "epoch": 0.25, "learning_rate": 4.887780350354372e-06, "loss": 0.487, "step": 1301 }, { "epoch": 0.25, "learning_rate": 4.887553641517203e-06, "loss": 0.4777, "step": 1302 }, { "epoch": 0.25, "learning_rate": 4.887326709178165e-06, "loss": 0.5043, "step": 1303 }, { "epoch": 0.25, "learning_rate": 4.887099553358502e-06, "loss": 0.6121, "step": 1304 }, { "epoch": 0.25, "learning_rate": 4.886872174079477e-06, "loss": 0.5734, "step": 1305 }, { "epoch": 0.25, "learning_rate": 4.886644571362376e-06, "loss": 0.4734, "step": 1306 }, { "epoch": 0.25, "learning_rate": 4.8864167452285045e-06, "loss": 0.7356, "step": 1307 }, { "epoch": 0.25, "learning_rate": 4.88618869569919e-06, "loss": 0.5865, "step": 1308 }, { "epoch": 0.25, "learning_rate": 4.885960422795781e-06, "loss": 0.4291, "step": 1309 }, { "epoch": 0.25, "learning_rate": 4.885731926539645e-06, "loss": 0.4861, "step": 1310 }, { "epoch": 0.25, "learning_rate": 4.8855032069521735e-06, "loss": 0.6589, "step": 1311 }, { "epoch": 0.25, "learning_rate": 4.885274264054777e-06, "loss": 0.5269, "step": 1312 }, { "epoch": 0.25, "learning_rate": 4.885045097868886e-06, "loss": 0.4687, "step": 1313 }, { "epoch": 0.25, "learning_rate": 4.884815708415953e-06, "loss": 0.6367, "step": 1314 }, { "epoch": 0.25, "learning_rate": 4.884586095717453e-06, "loss": 0.6816, "step": 1315 }, { "epoch": 0.25, "learning_rate": 4.8843562597948795e-06, "loss": 0.5535, "step": 1316 }, { "epoch": 0.25, "learning_rate": 4.884126200669746e-06, "loss": 0.5553, "step": 1317 }, { "epoch": 0.25, "learning_rate": 4.883895918363591e-06, "loss": 0.5523, "step": 1318 }, { "epoch": 0.25, "learning_rate": 4.8836654128979725e-06, "loss": 0.4124, "step": 1319 }, { "epoch": 0.25, "learning_rate": 4.883434684294465e-06, "loss": 0.6481, "step": 1320 }, { "epoch": 0.25, "learning_rate": 4.883203732574668e-06, "loss": 0.5096, "step": 1321 }, { "epoch": 0.25, "learning_rate": 4.8829725577602036e-06, "loss": 0.6343, "step": 1322 }, { "epoch": 0.25, "learning_rate": 4.882741159872711e-06, "loss": 0.5725, "step": 1323 }, { "epoch": 0.25, "learning_rate": 4.882509538933851e-06, "loss": 0.4848, "step": 1324 }, { "epoch": 0.25, "learning_rate": 4.882277694965307e-06, "loss": 0.5711, "step": 1325 }, { "epoch": 0.25, "learning_rate": 4.882045627988781e-06, "loss": 0.4748, "step": 1326 }, { "epoch": 0.25, "learning_rate": 4.881813338025998e-06, "loss": 0.7305, "step": 1327 }, { "epoch": 0.25, "learning_rate": 4.881580825098703e-06, "loss": 0.675, "step": 1328 }, { "epoch": 0.25, "learning_rate": 4.881348089228661e-06, "loss": 0.4952, "step": 1329 }, { "epoch": 0.25, "learning_rate": 4.88111513043766e-06, "loss": 0.6192, "step": 1330 }, { "epoch": 0.25, "learning_rate": 4.880881948747506e-06, "loss": 0.4496, "step": 1331 }, { "epoch": 0.25, "learning_rate": 4.880648544180028e-06, "loss": 0.6503, "step": 1332 }, { "epoch": 0.25, "learning_rate": 4.880414916757076e-06, "loss": 0.5535, "step": 1333 }, { "epoch": 0.25, "learning_rate": 4.880181066500519e-06, "loss": 0.5072, "step": 1334 }, { "epoch": 0.25, "learning_rate": 4.8799469934322496e-06, "loss": 0.5617, "step": 1335 }, { "epoch": 0.25, "learning_rate": 4.879712697574178e-06, "loss": 0.5632, "step": 1336 }, { "epoch": 0.25, "learning_rate": 4.879478178948238e-06, "loss": 0.517, "step": 1337 }, { "epoch": 0.25, "learning_rate": 4.879243437576383e-06, "loss": 0.647, "step": 1338 }, { "epoch": 0.25, "learning_rate": 4.879008473480588e-06, "loss": 0.3439, "step": 1339 }, { "epoch": 0.25, "learning_rate": 4.878773286682845e-06, "loss": 0.5639, "step": 1340 }, { "epoch": 0.25, "learning_rate": 4.878537877205175e-06, "loss": 0.6245, "step": 1341 }, { "epoch": 0.25, "learning_rate": 4.8783022450696125e-06, "loss": 0.5265, "step": 1342 }, { "epoch": 0.25, "learning_rate": 4.878066390298214e-06, "loss": 0.5289, "step": 1343 }, { "epoch": 0.25, "learning_rate": 4.877830312913061e-06, "loss": 0.5745, "step": 1344 }, { "epoch": 0.25, "learning_rate": 4.877594012936251e-06, "loss": 0.4941, "step": 1345 }, { "epoch": 0.25, "learning_rate": 4.877357490389906e-06, "loss": 0.4529, "step": 1346 }, { "epoch": 0.25, "learning_rate": 4.877120745296165e-06, "loss": 0.6534, "step": 1347 }, { "epoch": 0.25, "learning_rate": 4.876883777677192e-06, "loss": 0.5046, "step": 1348 }, { "epoch": 0.25, "learning_rate": 4.876646587555168e-06, "loss": 0.6103, "step": 1349 }, { "epoch": 0.26, "learning_rate": 4.876409174952298e-06, "loss": 0.5457, "step": 1350 }, { "epoch": 0.26, "learning_rate": 4.876171539890806e-06, "loss": 0.6229, "step": 1351 }, { "epoch": 0.26, "learning_rate": 4.875933682392938e-06, "loss": 0.4745, "step": 1352 }, { "epoch": 0.26, "learning_rate": 4.8756956024809585e-06, "loss": 0.4004, "step": 1353 }, { "epoch": 0.26, "learning_rate": 4.875457300177157e-06, "loss": 0.5709, "step": 1354 }, { "epoch": 0.26, "learning_rate": 4.8752187755038384e-06, "loss": 0.6378, "step": 1355 }, { "epoch": 0.26, "learning_rate": 4.874980028483334e-06, "loss": 0.4666, "step": 1356 }, { "epoch": 0.26, "learning_rate": 4.874741059137991e-06, "loss": 0.5209, "step": 1357 }, { "epoch": 0.26, "learning_rate": 4.87450186749018e-06, "loss": 0.6133, "step": 1358 }, { "epoch": 0.26, "learning_rate": 4.874262453562294e-06, "loss": 0.5155, "step": 1359 }, { "epoch": 0.26, "learning_rate": 4.874022817376742e-06, "loss": 0.6131, "step": 1360 }, { "epoch": 0.26, "learning_rate": 4.87378295895596e-06, "loss": 0.6355, "step": 1361 }, { "epoch": 0.26, "learning_rate": 4.873542878322397e-06, "loss": 0.6438, "step": 1362 }, { "epoch": 0.26, "learning_rate": 4.873302575498531e-06, "loss": 0.3782, "step": 1363 }, { "epoch": 0.26, "learning_rate": 4.873062050506856e-06, "loss": 0.5091, "step": 1364 }, { "epoch": 0.26, "learning_rate": 4.872821303369887e-06, "loss": 0.5343, "step": 1365 }, { "epoch": 0.26, "learning_rate": 4.8725803341101615e-06, "loss": 0.4958, "step": 1366 }, { "epoch": 0.26, "learning_rate": 4.872339142750236e-06, "loss": 0.4285, "step": 1367 }, { "epoch": 0.26, "learning_rate": 4.872097729312691e-06, "loss": 0.6093, "step": 1368 }, { "epoch": 0.26, "learning_rate": 4.871856093820123e-06, "loss": 0.5753, "step": 1369 }, { "epoch": 0.26, "learning_rate": 4.8716142362951525e-06, "loss": 0.6286, "step": 1370 }, { "epoch": 0.26, "learning_rate": 4.871372156760421e-06, "loss": 0.5655, "step": 1371 }, { "epoch": 0.26, "learning_rate": 4.871129855238589e-06, "loss": 0.5055, "step": 1372 }, { "epoch": 0.26, "learning_rate": 4.870887331752338e-06, "loss": 0.6804, "step": 1373 }, { "epoch": 0.26, "learning_rate": 4.870644586324372e-06, "loss": 0.6047, "step": 1374 }, { "epoch": 0.26, "learning_rate": 4.870401618977415e-06, "loss": 0.5302, "step": 1375 }, { "epoch": 0.26, "learning_rate": 4.870158429734211e-06, "loss": 0.5382, "step": 1376 }, { "epoch": 0.26, "learning_rate": 4.869915018617525e-06, "loss": 0.5245, "step": 1377 }, { "epoch": 0.26, "learning_rate": 4.869671385650143e-06, "loss": 0.5391, "step": 1378 }, { "epoch": 0.26, "learning_rate": 4.8694275308548714e-06, "loss": 0.5959, "step": 1379 }, { "epoch": 0.26, "learning_rate": 4.869183454254539e-06, "loss": 0.4424, "step": 1380 }, { "epoch": 0.26, "learning_rate": 4.868939155871993e-06, "loss": 0.4834, "step": 1381 }, { "epoch": 0.26, "learning_rate": 4.868694635730103e-06, "loss": 0.6365, "step": 1382 }, { "epoch": 0.26, "learning_rate": 4.86844989385176e-06, "loss": 0.4914, "step": 1383 }, { "epoch": 0.26, "learning_rate": 4.868204930259872e-06, "loss": 0.5255, "step": 1384 }, { "epoch": 0.26, "learning_rate": 4.867959744977372e-06, "loss": 0.5074, "step": 1385 }, { "epoch": 0.26, "learning_rate": 4.867714338027212e-06, "loss": 0.5403, "step": 1386 }, { "epoch": 0.26, "learning_rate": 4.867468709432364e-06, "loss": 0.439, "step": 1387 }, { "epoch": 0.26, "learning_rate": 4.867222859215823e-06, "loss": 0.6039, "step": 1388 }, { "epoch": 0.26, "learning_rate": 4.866976787400602e-06, "loss": 0.7138, "step": 1389 }, { "epoch": 0.26, "learning_rate": 4.866730494009737e-06, "loss": 0.5571, "step": 1390 }, { "epoch": 0.26, "learning_rate": 4.866483979066283e-06, "loss": 0.6795, "step": 1391 }, { "epoch": 0.26, "learning_rate": 4.866237242593318e-06, "loss": 0.5195, "step": 1392 }, { "epoch": 0.26, "learning_rate": 4.8659902846139375e-06, "loss": 0.602, "step": 1393 }, { "epoch": 0.26, "learning_rate": 4.865743105151261e-06, "loss": 0.6406, "step": 1394 }, { "epoch": 0.26, "learning_rate": 4.865495704228426e-06, "loss": 0.573, "step": 1395 }, { "epoch": 0.26, "learning_rate": 4.865248081868593e-06, "loss": 0.4564, "step": 1396 }, { "epoch": 0.26, "learning_rate": 4.8650002380949425e-06, "loss": 0.4959, "step": 1397 }, { "epoch": 0.26, "learning_rate": 4.864752172930674e-06, "loss": 0.536, "step": 1398 }, { "epoch": 0.26, "learning_rate": 4.8645038863990115e-06, "loss": 0.5644, "step": 1399 }, { "epoch": 0.26, "learning_rate": 4.864255378523195e-06, "loss": 0.6347, "step": 1400 }, { "epoch": 0.26, "learning_rate": 4.864006649326489e-06, "loss": 0.4959, "step": 1401 }, { "epoch": 0.26, "learning_rate": 4.863757698832176e-06, "loss": 0.5941, "step": 1402 }, { "epoch": 0.27, "learning_rate": 4.863508527063563e-06, "loss": 0.5093, "step": 1403 }, { "epoch": 0.27, "learning_rate": 4.863259134043973e-06, "loss": 0.5831, "step": 1404 }, { "epoch": 0.27, "learning_rate": 4.863009519796753e-06, "loss": 0.5299, "step": 1405 }, { "epoch": 0.27, "learning_rate": 4.862759684345269e-06, "loss": 0.6463, "step": 1406 }, { "epoch": 0.27, "learning_rate": 4.86250962771291e-06, "loss": 0.5246, "step": 1407 }, { "epoch": 0.27, "learning_rate": 4.862259349923083e-06, "loss": 0.432, "step": 1408 }, { "epoch": 0.27, "learning_rate": 4.862008850999216e-06, "loss": 0.612, "step": 1409 }, { "epoch": 0.27, "learning_rate": 4.86175813096476e-06, "loss": 0.458, "step": 1410 }, { "epoch": 0.27, "learning_rate": 4.8615071898431845e-06, "loss": 0.4826, "step": 1411 }, { "epoch": 0.27, "learning_rate": 4.8612560276579805e-06, "loss": 0.648, "step": 1412 }, { "epoch": 0.27, "learning_rate": 4.8610046444326585e-06, "loss": 0.6667, "step": 1413 }, { "epoch": 0.27, "learning_rate": 4.8607530401907535e-06, "loss": 0.6572, "step": 1414 }, { "epoch": 0.27, "learning_rate": 4.860501214955816e-06, "loss": 0.5806, "step": 1415 }, { "epoch": 0.27, "learning_rate": 4.860249168751421e-06, "loss": 0.5085, "step": 1416 }, { "epoch": 0.27, "learning_rate": 4.859996901601162e-06, "loss": 0.5009, "step": 1417 }, { "epoch": 0.27, "learning_rate": 4.859744413528654e-06, "loss": 0.5766, "step": 1418 }, { "epoch": 0.27, "learning_rate": 4.859491704557534e-06, "loss": 0.613, "step": 1419 }, { "epoch": 0.27, "learning_rate": 4.859238774711458e-06, "loss": 0.5559, "step": 1420 }, { "epoch": 0.27, "learning_rate": 4.858985624014102e-06, "loss": 0.5073, "step": 1421 }, { "epoch": 0.27, "learning_rate": 4.858732252489164e-06, "loss": 0.5133, "step": 1422 }, { "epoch": 0.27, "learning_rate": 4.858478660160364e-06, "loss": 0.509, "step": 1423 }, { "epoch": 0.27, "learning_rate": 4.858224847051439e-06, "loss": 0.5476, "step": 1424 }, { "epoch": 0.27, "learning_rate": 4.8579708131861505e-06, "loss": 0.5452, "step": 1425 }, { "epoch": 0.27, "learning_rate": 4.857716558588278e-06, "loss": 0.5792, "step": 1426 }, { "epoch": 0.27, "learning_rate": 4.857462083281622e-06, "loss": 0.6408, "step": 1427 }, { "epoch": 0.27, "learning_rate": 4.857207387290006e-06, "loss": 0.6639, "step": 1428 }, { "epoch": 0.27, "learning_rate": 4.85695247063727e-06, "loss": 0.5791, "step": 1429 }, { "epoch": 0.27, "learning_rate": 4.85669733334728e-06, "loss": 0.5059, "step": 1430 }, { "epoch": 0.27, "learning_rate": 4.856441975443917e-06, "loss": 0.4633, "step": 1431 }, { "epoch": 0.27, "learning_rate": 4.856186396951087e-06, "loss": 0.4644, "step": 1432 }, { "epoch": 0.27, "learning_rate": 4.855930597892714e-06, "loss": 0.5344, "step": 1433 }, { "epoch": 0.27, "learning_rate": 4.8556745782927454e-06, "loss": 0.6275, "step": 1434 }, { "epoch": 0.27, "learning_rate": 4.855418338175145e-06, "loss": 0.5992, "step": 1435 }, { "epoch": 0.27, "learning_rate": 4.855161877563902e-06, "loss": 0.5738, "step": 1436 }, { "epoch": 0.27, "learning_rate": 4.854905196483023e-06, "loss": 0.5236, "step": 1437 }, { "epoch": 0.27, "learning_rate": 4.854648294956536e-06, "loss": 0.6402, "step": 1438 }, { "epoch": 0.27, "learning_rate": 4.85439117300849e-06, "loss": 0.5332, "step": 1439 }, { "epoch": 0.27, "learning_rate": 4.854133830662955e-06, "loss": 0.4838, "step": 1440 }, { "epoch": 0.27, "learning_rate": 4.853876267944021e-06, "loss": 0.5666, "step": 1441 }, { "epoch": 0.27, "learning_rate": 4.853618484875798e-06, "loss": 0.5257, "step": 1442 }, { "epoch": 0.27, "learning_rate": 4.853360481482418e-06, "loss": 0.6131, "step": 1443 }, { "epoch": 0.27, "learning_rate": 4.853102257788034e-06, "loss": 0.4851, "step": 1444 }, { "epoch": 0.27, "learning_rate": 4.8528438138168175e-06, "loss": 0.5844, "step": 1445 }, { "epoch": 0.27, "learning_rate": 4.852585149592961e-06, "loss": 0.5688, "step": 1446 }, { "epoch": 0.27, "learning_rate": 4.85232626514068e-06, "loss": 0.5752, "step": 1447 }, { "epoch": 0.27, "learning_rate": 4.852067160484208e-06, "loss": 0.4844, "step": 1448 }, { "epoch": 0.27, "learning_rate": 4.851807835647802e-06, "loss": 0.6181, "step": 1449 }, { "epoch": 0.27, "learning_rate": 4.851548290655734e-06, "loss": 0.6076, "step": 1450 }, { "epoch": 0.27, "learning_rate": 4.851288525532303e-06, "loss": 0.4902, "step": 1451 }, { "epoch": 0.27, "learning_rate": 4.851028540301825e-06, "loss": 0.527, "step": 1452 }, { "epoch": 0.27, "learning_rate": 4.850768334988639e-06, "loss": 0.4946, "step": 1453 }, { "epoch": 0.27, "learning_rate": 4.850507909617102e-06, "loss": 0.5488, "step": 1454 }, { "epoch": 0.27, "learning_rate": 4.8502472642115915e-06, "loss": 0.5627, "step": 1455 }, { "epoch": 0.28, "learning_rate": 4.849986398796509e-06, "loss": 0.5297, "step": 1456 }, { "epoch": 0.28, "learning_rate": 4.849725313396274e-06, "loss": 0.4577, "step": 1457 }, { "epoch": 0.28, "learning_rate": 4.849464008035326e-06, "loss": 0.5116, "step": 1458 }, { "epoch": 0.28, "learning_rate": 4.849202482738128e-06, "loss": 0.5338, "step": 1459 }, { "epoch": 0.28, "learning_rate": 4.8489407375291586e-06, "loss": 0.5188, "step": 1460 }, { "epoch": 0.28, "learning_rate": 4.848678772432923e-06, "loss": 0.421, "step": 1461 }, { "epoch": 0.28, "learning_rate": 4.848416587473943e-06, "loss": 0.5551, "step": 1462 }, { "epoch": 0.28, "learning_rate": 4.848154182676762e-06, "loss": 0.5382, "step": 1463 }, { "epoch": 0.28, "learning_rate": 4.847891558065944e-06, "loss": 0.5441, "step": 1464 }, { "epoch": 0.28, "learning_rate": 4.847628713666074e-06, "loss": 0.4883, "step": 1465 }, { "epoch": 0.28, "learning_rate": 4.847365649501757e-06, "loss": 0.5556, "step": 1466 }, { "epoch": 0.28, "learning_rate": 4.847102365597619e-06, "loss": 0.4643, "step": 1467 }, { "epoch": 0.28, "learning_rate": 4.846838861978305e-06, "loss": 0.529, "step": 1468 }, { "epoch": 0.28, "learning_rate": 4.846575138668483e-06, "loss": 0.5407, "step": 1469 }, { "epoch": 0.28, "learning_rate": 4.846311195692841e-06, "loss": 0.6178, "step": 1470 }, { "epoch": 0.28, "learning_rate": 4.846047033076087e-06, "loss": 0.5214, "step": 1471 }, { "epoch": 0.28, "learning_rate": 4.8457826508429475e-06, "loss": 0.6659, "step": 1472 }, { "epoch": 0.28, "learning_rate": 4.845518049018174e-06, "loss": 0.5239, "step": 1473 }, { "epoch": 0.28, "learning_rate": 4.8452532276265366e-06, "loss": 0.5471, "step": 1474 }, { "epoch": 0.28, "learning_rate": 4.8449881866928225e-06, "loss": 0.5123, "step": 1475 }, { "epoch": 0.28, "learning_rate": 4.844722926241845e-06, "loss": 0.5709, "step": 1476 }, { "epoch": 0.28, "learning_rate": 4.844457446298435e-06, "loss": 0.4966, "step": 1477 }, { "epoch": 0.28, "learning_rate": 4.844191746887444e-06, "loss": 0.445, "step": 1478 }, { "epoch": 0.28, "learning_rate": 4.8439258280337445e-06, "loss": 0.6672, "step": 1479 }, { "epoch": 0.28, "learning_rate": 4.84365968976223e-06, "loss": 0.6344, "step": 1480 }, { "epoch": 0.28, "learning_rate": 4.843393332097814e-06, "loss": 0.6118, "step": 1481 }, { "epoch": 0.28, "learning_rate": 4.843126755065429e-06, "loss": 0.668, "step": 1482 }, { "epoch": 0.28, "learning_rate": 4.8428599586900325e-06, "loss": 0.5905, "step": 1483 }, { "epoch": 0.28, "learning_rate": 4.842592942996597e-06, "loss": 0.478, "step": 1484 }, { "epoch": 0.28, "learning_rate": 4.84232570801012e-06, "loss": 0.6295, "step": 1485 }, { "epoch": 0.28, "learning_rate": 4.842058253755616e-06, "loss": 0.6044, "step": 1486 }, { "epoch": 0.28, "learning_rate": 4.841790580258123e-06, "loss": 0.4979, "step": 1487 }, { "epoch": 0.28, "learning_rate": 4.841522687542698e-06, "loss": 0.5428, "step": 1488 }, { "epoch": 0.28, "learning_rate": 4.841254575634419e-06, "loss": 0.539, "step": 1489 }, { "epoch": 0.28, "learning_rate": 4.840986244558383e-06, "loss": 0.508, "step": 1490 }, { "epoch": 0.28, "learning_rate": 4.84071769433971e-06, "loss": 0.5355, "step": 1491 }, { "epoch": 0.28, "learning_rate": 4.84044892500354e-06, "loss": 0.5281, "step": 1492 }, { "epoch": 0.28, "learning_rate": 4.840179936575031e-06, "loss": 0.5817, "step": 1493 }, { "epoch": 0.28, "learning_rate": 4.839910729079365e-06, "loss": 0.5231, "step": 1494 }, { "epoch": 0.28, "learning_rate": 4.8396413025417414e-06, "loss": 0.6576, "step": 1495 }, { "epoch": 0.28, "learning_rate": 4.839371656987382e-06, "loss": 0.5176, "step": 1496 }, { "epoch": 0.28, "learning_rate": 4.83910179244153e-06, "loss": 0.5287, "step": 1497 }, { "epoch": 0.28, "learning_rate": 4.838831708929445e-06, "loss": 0.4981, "step": 1498 }, { "epoch": 0.28, "learning_rate": 4.838561406476413e-06, "loss": 0.4677, "step": 1499 }, { "epoch": 0.28, "learning_rate": 4.838290885107736e-06, "loss": 0.5846, "step": 1500 }, { "epoch": 0.28, "learning_rate": 4.838020144848737e-06, "loss": 0.5849, "step": 1501 }, { "epoch": 0.28, "learning_rate": 4.837749185724761e-06, "loss": 0.6293, "step": 1502 }, { "epoch": 0.28, "learning_rate": 4.837478007761173e-06, "loss": 0.6125, "step": 1503 }, { "epoch": 0.28, "learning_rate": 4.837206610983358e-06, "loss": 0.6011, "step": 1504 }, { "epoch": 0.28, "learning_rate": 4.836934995416722e-06, "loss": 0.5524, "step": 1505 }, { "epoch": 0.28, "learning_rate": 4.836663161086691e-06, "loss": 0.4685, "step": 1506 }, { "epoch": 0.28, "learning_rate": 4.836391108018713e-06, "loss": 0.5693, "step": 1507 }, { "epoch": 0.28, "learning_rate": 4.836118836238253e-06, "loss": 0.5895, "step": 1508 }, { "epoch": 0.29, "learning_rate": 4.8358463457708e-06, "loss": 0.5701, "step": 1509 }, { "epoch": 0.29, "learning_rate": 4.8355736366418625e-06, "loss": 0.4861, "step": 1510 }, { "epoch": 0.29, "learning_rate": 4.8353007088769685e-06, "loss": 0.6106, "step": 1511 }, { "epoch": 0.29, "learning_rate": 4.835027562501668e-06, "loss": 0.6606, "step": 1512 }, { "epoch": 0.29, "learning_rate": 4.834754197541529e-06, "loss": 0.7763, "step": 1513 }, { "epoch": 0.29, "learning_rate": 4.8344806140221425e-06, "loss": 0.4008, "step": 1514 }, { "epoch": 0.29, "learning_rate": 4.83420681196912e-06, "loss": 0.4237, "step": 1515 }, { "epoch": 0.29, "learning_rate": 4.83393279140809e-06, "loss": 0.5613, "step": 1516 }, { "epoch": 0.29, "learning_rate": 4.8336585523647065e-06, "loss": 0.5097, "step": 1517 }, { "epoch": 0.29, "learning_rate": 4.83338409486464e-06, "loss": 0.5146, "step": 1518 }, { "epoch": 0.29, "learning_rate": 4.8331094189335815e-06, "loss": 0.4506, "step": 1519 }, { "epoch": 0.29, "learning_rate": 4.832834524597246e-06, "loss": 0.5236, "step": 1520 }, { "epoch": 0.29, "learning_rate": 4.832559411881366e-06, "loss": 0.6609, "step": 1521 }, { "epoch": 0.29, "learning_rate": 4.832284080811695e-06, "loss": 0.5962, "step": 1522 }, { "epoch": 0.29, "learning_rate": 4.832008531414008e-06, "loss": 0.4823, "step": 1523 }, { "epoch": 0.29, "learning_rate": 4.831732763714097e-06, "loss": 0.6684, "step": 1524 }, { "epoch": 0.29, "learning_rate": 4.83145677773778e-06, "loss": 0.533, "step": 1525 }, { "epoch": 0.29, "learning_rate": 4.83118057351089e-06, "loss": 0.4338, "step": 1526 }, { "epoch": 0.29, "learning_rate": 4.830904151059284e-06, "loss": 0.5754, "step": 1527 }, { "epoch": 0.29, "learning_rate": 4.830627510408838e-06, "loss": 0.7401, "step": 1528 }, { "epoch": 0.29, "learning_rate": 4.830350651585449e-06, "loss": 0.5296, "step": 1529 }, { "epoch": 0.29, "learning_rate": 4.830073574615035e-06, "loss": 0.4547, "step": 1530 }, { "epoch": 0.29, "learning_rate": 4.829796279523531e-06, "loss": 0.6694, "step": 1531 }, { "epoch": 0.29, "learning_rate": 4.829518766336896e-06, "loss": 0.3972, "step": 1532 }, { "epoch": 0.29, "learning_rate": 4.8292410350811095e-06, "loss": 0.4556, "step": 1533 }, { "epoch": 0.29, "learning_rate": 4.828963085782169e-06, "loss": 0.436, "step": 1534 }, { "epoch": 0.29, "learning_rate": 4.828684918466094e-06, "loss": 0.4387, "step": 1535 }, { "epoch": 0.29, "learning_rate": 4.8284065331589255e-06, "loss": 0.3939, "step": 1536 }, { "epoch": 0.29, "learning_rate": 4.828127929886721e-06, "loss": 0.5099, "step": 1537 }, { "epoch": 0.29, "learning_rate": 4.827849108675562e-06, "loss": 0.5678, "step": 1538 }, { "epoch": 0.29, "learning_rate": 4.82757006955155e-06, "loss": 0.4575, "step": 1539 }, { "epoch": 0.29, "learning_rate": 4.827290812540806e-06, "loss": 0.532, "step": 1540 }, { "epoch": 0.29, "learning_rate": 4.827011337669469e-06, "loss": 0.5866, "step": 1541 }, { "epoch": 0.29, "learning_rate": 4.826731644963705e-06, "loss": 0.6021, "step": 1542 }, { "epoch": 0.29, "learning_rate": 4.826451734449694e-06, "loss": 0.447, "step": 1543 }, { "epoch": 0.29, "learning_rate": 4.82617160615364e-06, "loss": 0.5672, "step": 1544 }, { "epoch": 0.29, "learning_rate": 4.825891260101765e-06, "loss": 0.4925, "step": 1545 }, { "epoch": 0.29, "learning_rate": 4.825610696320313e-06, "loss": 0.5895, "step": 1546 }, { "epoch": 0.29, "learning_rate": 4.825329914835548e-06, "loss": 0.4833, "step": 1547 }, { "epoch": 0.29, "learning_rate": 4.825048915673755e-06, "loss": 0.665, "step": 1548 }, { "epoch": 0.29, "learning_rate": 4.824767698861237e-06, "loss": 0.5418, "step": 1549 }, { "epoch": 0.29, "learning_rate": 4.82448626442432e-06, "loss": 0.5487, "step": 1550 }, { "epoch": 0.29, "learning_rate": 4.824204612389349e-06, "loss": 0.5151, "step": 1551 }, { "epoch": 0.29, "learning_rate": 4.823922742782691e-06, "loss": 0.5537, "step": 1552 }, { "epoch": 0.29, "learning_rate": 4.823640655630732e-06, "loss": 0.4901, "step": 1553 }, { "epoch": 0.29, "learning_rate": 4.8233583509598765e-06, "loss": 0.529, "step": 1554 }, { "epoch": 0.29, "learning_rate": 4.823075828796554e-06, "loss": 0.7072, "step": 1555 }, { "epoch": 0.29, "learning_rate": 4.822793089167211e-06, "loss": 0.4952, "step": 1556 }, { "epoch": 0.29, "learning_rate": 4.8225101320983125e-06, "loss": 0.5215, "step": 1557 }, { "epoch": 0.29, "learning_rate": 4.822226957616351e-06, "loss": 0.4006, "step": 1558 }, { "epoch": 0.29, "learning_rate": 4.821943565747831e-06, "loss": 0.5726, "step": 1559 }, { "epoch": 0.29, "learning_rate": 4.821659956519282e-06, "loss": 0.6052, "step": 1560 }, { "epoch": 0.29, "learning_rate": 4.821376129957255e-06, "loss": 0.5381, "step": 1561 }, { "epoch": 0.3, "learning_rate": 4.821092086088318e-06, "loss": 0.4547, "step": 1562 }, { "epoch": 0.3, "learning_rate": 4.82080782493906e-06, "loss": 0.5133, "step": 1563 }, { "epoch": 0.3, "learning_rate": 4.820523346536093e-06, "loss": 0.5509, "step": 1564 }, { "epoch": 0.3, "learning_rate": 4.820238650906044e-06, "loss": 0.4414, "step": 1565 }, { "epoch": 0.3, "learning_rate": 4.8199537380755665e-06, "loss": 0.5531, "step": 1566 }, { "epoch": 0.3, "learning_rate": 4.8196686080713314e-06, "loss": 0.5647, "step": 1567 }, { "epoch": 0.3, "learning_rate": 4.81938326092003e-06, "loss": 0.5312, "step": 1568 }, { "epoch": 0.3, "learning_rate": 4.819097696648373e-06, "loss": 0.6631, "step": 1569 }, { "epoch": 0.3, "learning_rate": 4.818811915283093e-06, "loss": 0.5488, "step": 1570 }, { "epoch": 0.3, "learning_rate": 4.818525916850942e-06, "loss": 0.5977, "step": 1571 }, { "epoch": 0.3, "learning_rate": 4.818239701378694e-06, "loss": 0.592, "step": 1572 }, { "epoch": 0.3, "learning_rate": 4.817953268893139e-06, "loss": 0.5211, "step": 1573 }, { "epoch": 0.3, "learning_rate": 4.817666619421094e-06, "loss": 0.4194, "step": 1574 }, { "epoch": 0.3, "learning_rate": 4.8173797529893904e-06, "loss": 0.6037, "step": 1575 }, { "epoch": 0.3, "learning_rate": 4.817092669624883e-06, "loss": 0.6071, "step": 1576 }, { "epoch": 0.3, "learning_rate": 4.816805369354446e-06, "loss": 0.4597, "step": 1577 }, { "epoch": 0.3, "learning_rate": 4.816517852204972e-06, "loss": 0.4981, "step": 1578 }, { "epoch": 0.3, "learning_rate": 4.8162301182033785e-06, "loss": 0.4863, "step": 1579 }, { "epoch": 0.3, "learning_rate": 4.815942167376599e-06, "loss": 0.5908, "step": 1580 }, { "epoch": 0.3, "learning_rate": 4.81565399975159e-06, "loss": 0.621, "step": 1581 }, { "epoch": 0.3, "learning_rate": 4.815365615355327e-06, "loss": 0.5007, "step": 1582 }, { "epoch": 0.3, "learning_rate": 4.815077014214805e-06, "loss": 0.5055, "step": 1583 }, { "epoch": 0.3, "learning_rate": 4.814788196357041e-06, "loss": 0.5654, "step": 1584 }, { "epoch": 0.3, "learning_rate": 4.814499161809072e-06, "loss": 0.5176, "step": 1585 }, { "epoch": 0.3, "learning_rate": 4.8142099105979554e-06, "loss": 0.5671, "step": 1586 }, { "epoch": 0.3, "learning_rate": 4.813920442750766e-06, "loss": 0.5345, "step": 1587 }, { "epoch": 0.3, "learning_rate": 4.813630758294604e-06, "loss": 0.5272, "step": 1588 }, { "epoch": 0.3, "learning_rate": 4.813340857256585e-06, "loss": 0.5006, "step": 1589 }, { "epoch": 0.3, "learning_rate": 4.813050739663848e-06, "loss": 0.547, "step": 1590 }, { "epoch": 0.3, "learning_rate": 4.812760405543551e-06, "loss": 0.6228, "step": 1591 }, { "epoch": 0.3, "learning_rate": 4.812469854922873e-06, "loss": 0.6325, "step": 1592 }, { "epoch": 0.3, "learning_rate": 4.8121790878290135e-06, "loss": 0.6551, "step": 1593 }, { "epoch": 0.3, "learning_rate": 4.811888104289189e-06, "loss": 0.5892, "step": 1594 }, { "epoch": 0.3, "learning_rate": 4.811596904330641e-06, "loss": 0.6022, "step": 1595 }, { "epoch": 0.3, "learning_rate": 4.811305487980629e-06, "loss": 0.5982, "step": 1596 }, { "epoch": 0.3, "learning_rate": 4.811013855266431e-06, "loss": 0.6108, "step": 1597 }, { "epoch": 0.3, "learning_rate": 4.81072200621535e-06, "loss": 0.5221, "step": 1598 }, { "epoch": 0.3, "learning_rate": 4.810429940854703e-06, "loss": 0.5255, "step": 1599 }, { "epoch": 0.3, "learning_rate": 4.810137659211834e-06, "loss": 0.47, "step": 1600 }, { "epoch": 0.3, "learning_rate": 4.809845161314101e-06, "loss": 0.5173, "step": 1601 }, { "epoch": 0.3, "learning_rate": 4.809552447188887e-06, "loss": 0.7458, "step": 1602 }, { "epoch": 0.3, "learning_rate": 4.809259516863592e-06, "loss": 0.4583, "step": 1603 }, { "epoch": 0.3, "learning_rate": 4.808966370365639e-06, "loss": 0.5082, "step": 1604 }, { "epoch": 0.3, "learning_rate": 4.808673007722469e-06, "loss": 0.473, "step": 1605 }, { "epoch": 0.3, "learning_rate": 4.808379428961544e-06, "loss": 0.6788, "step": 1606 }, { "epoch": 0.3, "learning_rate": 4.808085634110346e-06, "loss": 0.564, "step": 1607 }, { "epoch": 0.3, "learning_rate": 4.807791623196378e-06, "loss": 0.5221, "step": 1608 }, { "epoch": 0.3, "learning_rate": 4.807497396247163e-06, "loss": 0.4744, "step": 1609 }, { "epoch": 0.3, "learning_rate": 4.807202953290243e-06, "loss": 0.5468, "step": 1610 }, { "epoch": 0.3, "learning_rate": 4.806908294353182e-06, "loss": 0.6252, "step": 1611 }, { "epoch": 0.3, "learning_rate": 4.806613419463564e-06, "loss": 0.4835, "step": 1612 }, { "epoch": 0.3, "learning_rate": 4.80631832864899e-06, "loss": 0.5361, "step": 1613 }, { "epoch": 0.3, "learning_rate": 4.806023021937086e-06, "loss": 0.5633, "step": 1614 }, { "epoch": 0.31, "learning_rate": 4.805727499355496e-06, "loss": 0.4915, "step": 1615 }, { "epoch": 0.31, "learning_rate": 4.805431760931883e-06, "loss": 0.4975, "step": 1616 }, { "epoch": 0.31, "learning_rate": 4.805135806693934e-06, "loss": 0.6104, "step": 1617 }, { "epoch": 0.31, "learning_rate": 4.80483963666935e-06, "loss": 0.6792, "step": 1618 }, { "epoch": 0.31, "learning_rate": 4.804543250885859e-06, "loss": 0.5106, "step": 1619 }, { "epoch": 0.31, "learning_rate": 4.8042466493712035e-06, "loss": 0.6465, "step": 1620 }, { "epoch": 0.31, "learning_rate": 4.8039498321531505e-06, "loss": 0.6032, "step": 1621 }, { "epoch": 0.31, "learning_rate": 4.803652799259486e-06, "loss": 0.5553, "step": 1622 }, { "epoch": 0.31, "learning_rate": 4.803355550718013e-06, "loss": 0.7241, "step": 1623 }, { "epoch": 0.31, "learning_rate": 4.8030580865565595e-06, "loss": 0.5736, "step": 1624 }, { "epoch": 0.31, "learning_rate": 4.802760406802971e-06, "loss": 0.5819, "step": 1625 }, { "epoch": 0.31, "learning_rate": 4.802462511485113e-06, "loss": 0.4832, "step": 1626 }, { "epoch": 0.31, "learning_rate": 4.8021644006308735e-06, "loss": 0.6183, "step": 1627 }, { "epoch": 0.31, "learning_rate": 4.801866074268156e-06, "loss": 0.4803, "step": 1628 }, { "epoch": 0.31, "learning_rate": 4.8015675324248915e-06, "loss": 0.5011, "step": 1629 }, { "epoch": 0.31, "learning_rate": 4.801268775129022e-06, "loss": 0.5484, "step": 1630 }, { "epoch": 0.31, "learning_rate": 4.800969802408519e-06, "loss": 0.4918, "step": 1631 }, { "epoch": 0.31, "learning_rate": 4.800670614291366e-06, "loss": 0.5225, "step": 1632 }, { "epoch": 0.31, "learning_rate": 4.800371210805573e-06, "loss": 0.4713, "step": 1633 }, { "epoch": 0.31, "learning_rate": 4.8000715919791675e-06, "loss": 0.531, "step": 1634 }, { "epoch": 0.31, "learning_rate": 4.799771757840195e-06, "loss": 0.6042, "step": 1635 }, { "epoch": 0.31, "learning_rate": 4.799471708416726e-06, "loss": 0.4351, "step": 1636 }, { "epoch": 0.31, "learning_rate": 4.799171443736845e-06, "loss": 0.6476, "step": 1637 }, { "epoch": 0.31, "learning_rate": 4.798870963828664e-06, "loss": 0.477, "step": 1638 }, { "epoch": 0.31, "learning_rate": 4.7985702687203094e-06, "loss": 0.5607, "step": 1639 }, { "epoch": 0.31, "learning_rate": 4.79826935843993e-06, "loss": 0.5554, "step": 1640 }, { "epoch": 0.31, "learning_rate": 4.797968233015694e-06, "loss": 0.4575, "step": 1641 }, { "epoch": 0.31, "learning_rate": 4.79766689247579e-06, "loss": 0.6327, "step": 1642 }, { "epoch": 0.31, "learning_rate": 4.797365336848428e-06, "loss": 0.5226, "step": 1643 }, { "epoch": 0.31, "learning_rate": 4.7970635661618345e-06, "loss": 0.5068, "step": 1644 }, { "epoch": 0.31, "learning_rate": 4.7967615804442615e-06, "loss": 0.4548, "step": 1645 }, { "epoch": 0.31, "learning_rate": 4.7964593797239776e-06, "loss": 0.6712, "step": 1646 }, { "epoch": 0.31, "learning_rate": 4.796156964029272e-06, "loss": 0.68, "step": 1647 }, { "epoch": 0.31, "learning_rate": 4.795854333388453e-06, "loss": 0.7166, "step": 1648 }, { "epoch": 0.31, "learning_rate": 4.795551487829851e-06, "loss": 0.4254, "step": 1649 }, { "epoch": 0.31, "learning_rate": 4.795248427381817e-06, "loss": 0.5615, "step": 1650 }, { "epoch": 0.31, "learning_rate": 4.7949451520727185e-06, "loss": 0.5959, "step": 1651 }, { "epoch": 0.31, "learning_rate": 4.794641661930947e-06, "loss": 0.496, "step": 1652 }, { "epoch": 0.31, "learning_rate": 4.794337956984912e-06, "loss": 0.7073, "step": 1653 }, { "epoch": 0.31, "learning_rate": 4.794034037263046e-06, "loss": 0.694, "step": 1654 }, { "epoch": 0.31, "learning_rate": 4.793729902793796e-06, "loss": 0.5051, "step": 1655 }, { "epoch": 0.31, "learning_rate": 4.793425553605633e-06, "loss": 0.5133, "step": 1656 }, { "epoch": 0.31, "learning_rate": 4.79312098972705e-06, "loss": 0.4498, "step": 1657 }, { "epoch": 0.31, "learning_rate": 4.792816211186554e-06, "loss": 0.5649, "step": 1658 }, { "epoch": 0.31, "learning_rate": 4.7925112180126785e-06, "loss": 0.5156, "step": 1659 }, { "epoch": 0.31, "learning_rate": 4.7922060102339726e-06, "loss": 0.3928, "step": 1660 }, { "epoch": 0.31, "learning_rate": 4.79190058787901e-06, "loss": 0.4809, "step": 1661 }, { "epoch": 0.31, "learning_rate": 4.791594950976377e-06, "loss": 0.5322, "step": 1662 }, { "epoch": 0.31, "learning_rate": 4.7912890995546886e-06, "loss": 0.6003, "step": 1663 }, { "epoch": 0.31, "learning_rate": 4.7909830336425746e-06, "loss": 0.452, "step": 1664 }, { "epoch": 0.31, "learning_rate": 4.790676753268685e-06, "loss": 0.4677, "step": 1665 }, { "epoch": 0.31, "learning_rate": 4.790370258461693e-06, "loss": 0.6563, "step": 1666 }, { "epoch": 0.31, "learning_rate": 4.79006354925029e-06, "loss": 0.5702, "step": 1667 }, { "epoch": 0.32, "learning_rate": 4.789756625663185e-06, "loss": 0.5329, "step": 1668 }, { "epoch": 0.32, "learning_rate": 4.789449487729113e-06, "loss": 0.5036, "step": 1669 }, { "epoch": 0.32, "learning_rate": 4.789142135476823e-06, "loss": 0.6385, "step": 1670 }, { "epoch": 0.32, "learning_rate": 4.788834568935088e-06, "loss": 0.4386, "step": 1671 }, { "epoch": 0.32, "learning_rate": 4.788526788132698e-06, "loss": 0.5301, "step": 1672 }, { "epoch": 0.32, "learning_rate": 4.788218793098466e-06, "loss": 0.5278, "step": 1673 }, { "epoch": 0.32, "learning_rate": 4.787910583861224e-06, "loss": 0.7049, "step": 1674 }, { "epoch": 0.32, "learning_rate": 4.787602160449823e-06, "loss": 0.4494, "step": 1675 }, { "epoch": 0.32, "learning_rate": 4.787293522893135e-06, "loss": 0.5088, "step": 1676 }, { "epoch": 0.32, "learning_rate": 4.786984671220053e-06, "loss": 0.4651, "step": 1677 }, { "epoch": 0.32, "learning_rate": 4.786675605459488e-06, "loss": 0.6222, "step": 1678 }, { "epoch": 0.32, "learning_rate": 4.786366325640372e-06, "loss": 0.7813, "step": 1679 }, { "epoch": 0.32, "learning_rate": 4.786056831791658e-06, "loss": 0.4906, "step": 1680 }, { "epoch": 0.32, "learning_rate": 4.785747123942317e-06, "loss": 0.603, "step": 1681 }, { "epoch": 0.32, "learning_rate": 4.785437202121342e-06, "loss": 0.4387, "step": 1682 }, { "epoch": 0.32, "learning_rate": 4.785127066357744e-06, "loss": 0.5303, "step": 1683 }, { "epoch": 0.32, "learning_rate": 4.784816716680557e-06, "loss": 0.4875, "step": 1684 }, { "epoch": 0.32, "learning_rate": 4.784506153118831e-06, "loss": 0.6876, "step": 1685 }, { "epoch": 0.32, "learning_rate": 4.78419537570164e-06, "loss": 0.5116, "step": 1686 }, { "epoch": 0.32, "learning_rate": 4.7838843844580764e-06, "loss": 0.6238, "step": 1687 }, { "epoch": 0.32, "learning_rate": 4.783573179417251e-06, "loss": 0.7321, "step": 1688 }, { "epoch": 0.32, "learning_rate": 4.7832617606082965e-06, "loss": 0.5607, "step": 1689 }, { "epoch": 0.32, "learning_rate": 4.7829501280603665e-06, "loss": 0.5329, "step": 1690 }, { "epoch": 0.32, "learning_rate": 4.7826382818026315e-06, "loss": 0.517, "step": 1691 }, { "epoch": 0.32, "learning_rate": 4.782326221864285e-06, "loss": 0.5778, "step": 1692 }, { "epoch": 0.32, "learning_rate": 4.78201394827454e-06, "loss": 0.506, "step": 1693 }, { "epoch": 0.32, "learning_rate": 4.781701461062627e-06, "loss": 0.5206, "step": 1694 }, { "epoch": 0.32, "learning_rate": 4.781388760257799e-06, "loss": 0.5476, "step": 1695 }, { "epoch": 0.32, "learning_rate": 4.781075845889329e-06, "loss": 0.5201, "step": 1696 }, { "epoch": 0.32, "learning_rate": 4.780762717986508e-06, "loss": 0.5881, "step": 1697 }, { "epoch": 0.32, "learning_rate": 4.78044937657865e-06, "loss": 0.5738, "step": 1698 }, { "epoch": 0.32, "learning_rate": 4.780135821695085e-06, "loss": 0.3592, "step": 1699 }, { "epoch": 0.32, "learning_rate": 4.7798220533651676e-06, "loss": 0.5235, "step": 1700 }, { "epoch": 0.32, "learning_rate": 4.779508071618269e-06, "loss": 0.5948, "step": 1701 }, { "epoch": 0.32, "learning_rate": 4.7791938764837805e-06, "loss": 0.6462, "step": 1702 }, { "epoch": 0.32, "learning_rate": 4.778879467991117e-06, "loss": 0.5688, "step": 1703 }, { "epoch": 0.32, "learning_rate": 4.778564846169707e-06, "loss": 0.7238, "step": 1704 }, { "epoch": 0.32, "learning_rate": 4.778250011049006e-06, "loss": 0.4667, "step": 1705 }, { "epoch": 0.32, "learning_rate": 4.777934962658483e-06, "loss": 0.4914, "step": 1706 }, { "epoch": 0.32, "learning_rate": 4.777619701027634e-06, "loss": 0.4812, "step": 1707 }, { "epoch": 0.32, "learning_rate": 4.777304226185968e-06, "loss": 0.6425, "step": 1708 }, { "epoch": 0.32, "learning_rate": 4.776988538163017e-06, "loss": 0.5018, "step": 1709 }, { "epoch": 0.32, "learning_rate": 4.776672636988334e-06, "loss": 0.435, "step": 1710 }, { "epoch": 0.32, "learning_rate": 4.776356522691491e-06, "loss": 0.5558, "step": 1711 }, { "epoch": 0.32, "learning_rate": 4.77604019530208e-06, "loss": 0.5297, "step": 1712 }, { "epoch": 0.32, "learning_rate": 4.775723654849711e-06, "loss": 0.5609, "step": 1713 }, { "epoch": 0.32, "learning_rate": 4.775406901364018e-06, "loss": 0.5023, "step": 1714 }, { "epoch": 0.32, "learning_rate": 4.775089934874652e-06, "loss": 0.4913, "step": 1715 }, { "epoch": 0.32, "learning_rate": 4.774772755411284e-06, "loss": 0.5704, "step": 1716 }, { "epoch": 0.32, "learning_rate": 4.774455363003605e-06, "loss": 0.4151, "step": 1717 }, { "epoch": 0.32, "learning_rate": 4.774137757681328e-06, "loss": 0.4626, "step": 1718 }, { "epoch": 0.32, "learning_rate": 4.773819939474184e-06, "loss": 0.6154, "step": 1719 }, { "epoch": 0.32, "learning_rate": 4.773501908411924e-06, "loss": 0.522, "step": 1720 }, { "epoch": 0.33, "learning_rate": 4.77318366452432e-06, "loss": 0.6324, "step": 1721 }, { "epoch": 0.33, "learning_rate": 4.772865207841162e-06, "loss": 0.5747, "step": 1722 }, { "epoch": 0.33, "learning_rate": 4.772546538392262e-06, "loss": 0.5628, "step": 1723 }, { "epoch": 0.33, "learning_rate": 4.772227656207451e-06, "loss": 0.6288, "step": 1724 }, { "epoch": 0.33, "learning_rate": 4.7719085613165794e-06, "loss": 0.5719, "step": 1725 }, { "epoch": 0.33, "learning_rate": 4.771589253749519e-06, "loss": 0.5103, "step": 1726 }, { "epoch": 0.33, "learning_rate": 4.771269733536161e-06, "loss": 0.554, "step": 1727 }, { "epoch": 0.33, "learning_rate": 4.770950000706412e-06, "loss": 0.4973, "step": 1728 }, { "epoch": 0.33, "learning_rate": 4.770630055290209e-06, "loss": 0.6351, "step": 1729 }, { "epoch": 0.33, "learning_rate": 4.770309897317498e-06, "loss": 0.6822, "step": 1730 }, { "epoch": 0.33, "learning_rate": 4.76998952681825e-06, "loss": 0.5024, "step": 1731 }, { "epoch": 0.33, "learning_rate": 4.769668943822457e-06, "loss": 0.6119, "step": 1732 }, { "epoch": 0.33, "learning_rate": 4.769348148360128e-06, "loss": 0.6556, "step": 1733 }, { "epoch": 0.33, "learning_rate": 4.769027140461292e-06, "loss": 0.598, "step": 1734 }, { "epoch": 0.33, "learning_rate": 4.768705920156001e-06, "loss": 0.5064, "step": 1735 }, { "epoch": 0.33, "learning_rate": 4.768384487474323e-06, "loss": 0.6508, "step": 1736 }, { "epoch": 0.33, "learning_rate": 4.768062842446351e-06, "loss": 0.4975, "step": 1737 }, { "epoch": 0.33, "learning_rate": 4.7677409851021895e-06, "loss": 0.3701, "step": 1738 }, { "epoch": 0.33, "learning_rate": 4.767418915471972e-06, "loss": 0.5444, "step": 1739 }, { "epoch": 0.33, "learning_rate": 4.767096633585846e-06, "loss": 0.5632, "step": 1740 }, { "epoch": 0.33, "learning_rate": 4.766774139473982e-06, "loss": 0.578, "step": 1741 }, { "epoch": 0.33, "learning_rate": 4.766451433166568e-06, "loss": 0.625, "step": 1742 }, { "epoch": 0.33, "learning_rate": 4.766128514693812e-06, "loss": 0.6375, "step": 1743 }, { "epoch": 0.33, "learning_rate": 4.765805384085945e-06, "loss": 0.487, "step": 1744 }, { "epoch": 0.33, "learning_rate": 4.765482041373215e-06, "loss": 0.6048, "step": 1745 }, { "epoch": 0.33, "learning_rate": 4.76515848658589e-06, "loss": 0.3957, "step": 1746 }, { "epoch": 0.33, "learning_rate": 4.764834719754259e-06, "loss": 0.5958, "step": 1747 }, { "epoch": 0.33, "learning_rate": 4.7645107409086295e-06, "loss": 0.5333, "step": 1748 }, { "epoch": 0.33, "learning_rate": 4.76418655007933e-06, "loss": 0.5335, "step": 1749 }, { "epoch": 0.33, "learning_rate": 4.7638621472967074e-06, "loss": 0.52, "step": 1750 }, { "epoch": 0.33, "learning_rate": 4.763537532591131e-06, "loss": 0.4075, "step": 1751 }, { "epoch": 0.33, "learning_rate": 4.763212705992988e-06, "loss": 0.5321, "step": 1752 }, { "epoch": 0.33, "learning_rate": 4.7628876675326855e-06, "loss": 0.556, "step": 1753 }, { "epoch": 0.33, "learning_rate": 4.76256241724065e-06, "loss": 0.606, "step": 1754 }, { "epoch": 0.33, "learning_rate": 4.762236955147331e-06, "loss": 0.5092, "step": 1755 }, { "epoch": 0.33, "learning_rate": 4.761911281283192e-06, "loss": 0.4224, "step": 1756 }, { "epoch": 0.33, "learning_rate": 4.761585395678722e-06, "loss": 0.573, "step": 1757 }, { "epoch": 0.33, "learning_rate": 4.7612592983644275e-06, "loss": 0.3744, "step": 1758 }, { "epoch": 0.33, "learning_rate": 4.760932989370834e-06, "loss": 0.4975, "step": 1759 }, { "epoch": 0.33, "learning_rate": 4.760606468728488e-06, "loss": 0.471, "step": 1760 }, { "epoch": 0.33, "learning_rate": 4.760279736467956e-06, "loss": 0.5045, "step": 1761 }, { "epoch": 0.33, "learning_rate": 4.759952792619824e-06, "loss": 0.5314, "step": 1762 }, { "epoch": 0.33, "learning_rate": 4.7596256372146964e-06, "loss": 0.5786, "step": 1763 }, { "epoch": 0.33, "learning_rate": 4.759298270283199e-06, "loss": 0.4868, "step": 1764 }, { "epoch": 0.33, "learning_rate": 4.758970691855979e-06, "loss": 0.5513, "step": 1765 }, { "epoch": 0.33, "learning_rate": 4.758642901963698e-06, "loss": 0.5139, "step": 1766 }, { "epoch": 0.33, "learning_rate": 4.758314900637043e-06, "loss": 0.4165, "step": 1767 }, { "epoch": 0.33, "learning_rate": 4.757986687906718e-06, "loss": 0.4736, "step": 1768 }, { "epoch": 0.33, "learning_rate": 4.757658263803448e-06, "loss": 0.5653, "step": 1769 }, { "epoch": 0.33, "learning_rate": 4.757329628357977e-06, "loss": 0.6179, "step": 1770 }, { "epoch": 0.33, "learning_rate": 4.757000781601069e-06, "loss": 0.5497, "step": 1771 }, { "epoch": 0.33, "learning_rate": 4.756671723563506e-06, "loss": 0.4303, "step": 1772 }, { "epoch": 0.33, "learning_rate": 4.7563424542760935e-06, "loss": 0.5324, "step": 1773 }, { "epoch": 0.34, "learning_rate": 4.756012973769655e-06, "loss": 0.6415, "step": 1774 }, { "epoch": 0.34, "learning_rate": 4.755683282075033e-06, "loss": 0.4801, "step": 1775 }, { "epoch": 0.34, "learning_rate": 4.75535337922309e-06, "loss": 0.5542, "step": 1776 }, { "epoch": 0.34, "learning_rate": 4.755023265244709e-06, "loss": 0.6396, "step": 1777 }, { "epoch": 0.34, "learning_rate": 4.754692940170792e-06, "loss": 0.5815, "step": 1778 }, { "epoch": 0.34, "learning_rate": 4.754362404032262e-06, "loss": 0.5746, "step": 1779 }, { "epoch": 0.34, "learning_rate": 4.75403165686006e-06, "loss": 0.5167, "step": 1780 }, { "epoch": 0.34, "learning_rate": 4.753700698685148e-06, "loss": 0.4913, "step": 1781 }, { "epoch": 0.34, "learning_rate": 4.753369529538507e-06, "loss": 0.512, "step": 1782 }, { "epoch": 0.34, "learning_rate": 4.753038149451139e-06, "loss": 0.5176, "step": 1783 }, { "epoch": 0.34, "learning_rate": 4.752706558454064e-06, "loss": 0.6254, "step": 1784 }, { "epoch": 0.34, "learning_rate": 4.7523747565783235e-06, "loss": 0.5747, "step": 1785 }, { "epoch": 0.34, "learning_rate": 4.752042743854977e-06, "loss": 0.5619, "step": 1786 }, { "epoch": 0.34, "learning_rate": 4.751710520315106e-06, "loss": 0.6753, "step": 1787 }, { "epoch": 0.34, "learning_rate": 4.751378085989809e-06, "loss": 0.5587, "step": 1788 }, { "epoch": 0.34, "learning_rate": 4.751045440910206e-06, "loss": 0.5904, "step": 1789 }, { "epoch": 0.34, "learning_rate": 4.750712585107436e-06, "loss": 0.4988, "step": 1790 }, { "epoch": 0.34, "learning_rate": 4.750379518612659e-06, "loss": 0.5832, "step": 1791 }, { "epoch": 0.34, "learning_rate": 4.750046241457053e-06, "loss": 0.5101, "step": 1792 }, { "epoch": 0.34, "learning_rate": 4.749712753671815e-06, "loss": 0.5229, "step": 1793 }, { "epoch": 0.34, "learning_rate": 4.749379055288167e-06, "loss": 0.5169, "step": 1794 }, { "epoch": 0.34, "learning_rate": 4.749045146337344e-06, "loss": 0.6524, "step": 1795 }, { "epoch": 0.34, "learning_rate": 4.748711026850606e-06, "loss": 0.5692, "step": 1796 }, { "epoch": 0.34, "learning_rate": 4.748376696859226e-06, "loss": 0.5439, "step": 1797 }, { "epoch": 0.34, "learning_rate": 4.748042156394506e-06, "loss": 0.6213, "step": 1798 }, { "epoch": 0.34, "learning_rate": 4.747707405487759e-06, "loss": 0.5332, "step": 1799 }, { "epoch": 0.34, "learning_rate": 4.747372444170325e-06, "loss": 0.5164, "step": 1800 }, { "epoch": 0.34, "learning_rate": 4.747037272473557e-06, "loss": 0.7945, "step": 1801 }, { "epoch": 0.34, "learning_rate": 4.746701890428831e-06, "loss": 0.5027, "step": 1802 }, { "epoch": 0.34, "learning_rate": 4.746366298067545e-06, "loss": 0.5394, "step": 1803 }, { "epoch": 0.34, "learning_rate": 4.746030495421112e-06, "loss": 0.6381, "step": 1804 }, { "epoch": 0.34, "learning_rate": 4.745694482520969e-06, "loss": 0.6265, "step": 1805 }, { "epoch": 0.34, "learning_rate": 4.745358259398567e-06, "loss": 0.5312, "step": 1806 }, { "epoch": 0.34, "learning_rate": 4.745021826085384e-06, "loss": 0.6331, "step": 1807 }, { "epoch": 0.34, "learning_rate": 4.744685182612912e-06, "loss": 0.5111, "step": 1808 }, { "epoch": 0.34, "learning_rate": 4.744348329012666e-06, "loss": 0.4908, "step": 1809 }, { "epoch": 0.34, "learning_rate": 4.7440112653161766e-06, "loss": 0.6073, "step": 1810 }, { "epoch": 0.34, "learning_rate": 4.743673991555e-06, "loss": 0.5789, "step": 1811 }, { "epoch": 0.34, "learning_rate": 4.743336507760706e-06, "loss": 0.54, "step": 1812 }, { "epoch": 0.34, "learning_rate": 4.74299881396489e-06, "loss": 0.5216, "step": 1813 }, { "epoch": 0.34, "learning_rate": 4.7426609101991605e-06, "loss": 0.6268, "step": 1814 }, { "epoch": 0.34, "learning_rate": 4.742322796495151e-06, "loss": 0.5014, "step": 1815 }, { "epoch": 0.34, "learning_rate": 4.741984472884513e-06, "loss": 0.5647, "step": 1816 }, { "epoch": 0.34, "learning_rate": 4.7416459393989165e-06, "loss": 0.6387, "step": 1817 }, { "epoch": 0.34, "learning_rate": 4.741307196070053e-06, "loss": 0.5649, "step": 1818 }, { "epoch": 0.34, "learning_rate": 4.740968242929631e-06, "loss": 0.6224, "step": 1819 }, { "epoch": 0.34, "learning_rate": 4.740629080009382e-06, "loss": 0.4119, "step": 1820 }, { "epoch": 0.34, "learning_rate": 4.7402897073410555e-06, "loss": 0.5089, "step": 1821 }, { "epoch": 0.34, "learning_rate": 4.73995012495642e-06, "loss": 0.4773, "step": 1822 }, { "epoch": 0.34, "learning_rate": 4.739610332887264e-06, "loss": 0.6479, "step": 1823 }, { "epoch": 0.34, "learning_rate": 4.739270331165395e-06, "loss": 0.5285, "step": 1824 }, { "epoch": 0.34, "learning_rate": 4.738930119822642e-06, "loss": 0.5799, "step": 1825 }, { "epoch": 0.34, "learning_rate": 4.7385896988908546e-06, "loss": 0.5145, "step": 1826 }, { "epoch": 0.35, "learning_rate": 4.738249068401896e-06, "loss": 0.5255, "step": 1827 }, { "epoch": 0.35, "learning_rate": 4.737908228387656e-06, "loss": 0.6543, "step": 1828 }, { "epoch": 0.35, "learning_rate": 4.7375671788800405e-06, "loss": 0.5708, "step": 1829 }, { "epoch": 0.35, "learning_rate": 4.737225919910975e-06, "loss": 0.5859, "step": 1830 }, { "epoch": 0.35, "learning_rate": 4.736884451512405e-06, "loss": 0.4813, "step": 1831 }, { "epoch": 0.35, "learning_rate": 4.736542773716297e-06, "loss": 0.6159, "step": 1832 }, { "epoch": 0.35, "learning_rate": 4.736200886554635e-06, "loss": 0.5371, "step": 1833 }, { "epoch": 0.35, "learning_rate": 4.735858790059423e-06, "loss": 0.6015, "step": 1834 }, { "epoch": 0.35, "learning_rate": 4.735516484262685e-06, "loss": 0.6993, "step": 1835 }, { "epoch": 0.35, "learning_rate": 4.7351739691964675e-06, "loss": 0.6268, "step": 1836 }, { "epoch": 0.35, "learning_rate": 4.7348312448928305e-06, "loss": 0.4127, "step": 1837 }, { "epoch": 0.35, "learning_rate": 4.734488311383857e-06, "loss": 0.4752, "step": 1838 }, { "epoch": 0.35, "learning_rate": 4.734145168701651e-06, "loss": 0.5505, "step": 1839 }, { "epoch": 0.35, "learning_rate": 4.733801816878335e-06, "loss": 0.4634, "step": 1840 }, { "epoch": 0.35, "learning_rate": 4.733458255946048e-06, "loss": 0.5075, "step": 1841 }, { "epoch": 0.35, "learning_rate": 4.7331144859369535e-06, "loss": 0.4528, "step": 1842 }, { "epoch": 0.35, "learning_rate": 4.732770506883231e-06, "loss": 0.651, "step": 1843 }, { "epoch": 0.35, "learning_rate": 4.732426318817082e-06, "loss": 0.4709, "step": 1844 }, { "epoch": 0.35, "learning_rate": 4.7320819217707245e-06, "loss": 0.6535, "step": 1845 }, { "epoch": 0.35, "learning_rate": 4.731737315776399e-06, "loss": 0.5304, "step": 1846 }, { "epoch": 0.35, "learning_rate": 4.731392500866365e-06, "loss": 0.5252, "step": 1847 }, { "epoch": 0.35, "learning_rate": 4.7310474770729e-06, "loss": 0.5758, "step": 1848 }, { "epoch": 0.35, "learning_rate": 4.7307022444283034e-06, "loss": 0.4997, "step": 1849 }, { "epoch": 0.35, "learning_rate": 4.730356802964892e-06, "loss": 0.5437, "step": 1850 }, { "epoch": 0.35, "learning_rate": 4.730011152715003e-06, "loss": 0.6081, "step": 1851 }, { "epoch": 0.35, "learning_rate": 4.7296652937109935e-06, "loss": 0.4395, "step": 1852 }, { "epoch": 0.35, "learning_rate": 4.729319225985239e-06, "loss": 0.5724, "step": 1853 }, { "epoch": 0.35, "learning_rate": 4.728972949570136e-06, "loss": 0.5335, "step": 1854 }, { "epoch": 0.35, "learning_rate": 4.7286264644981e-06, "loss": 0.6901, "step": 1855 }, { "epoch": 0.35, "learning_rate": 4.728279770801566e-06, "loss": 0.6612, "step": 1856 }, { "epoch": 0.35, "learning_rate": 4.727932868512987e-06, "loss": 0.4469, "step": 1857 }, { "epoch": 0.35, "learning_rate": 4.727585757664839e-06, "loss": 0.5186, "step": 1858 }, { "epoch": 0.35, "learning_rate": 4.727238438289614e-06, "loss": 0.5702, "step": 1859 }, { "epoch": 0.35, "learning_rate": 4.726890910419826e-06, "loss": 0.6532, "step": 1860 }, { "epoch": 0.35, "learning_rate": 4.726543174088007e-06, "loss": 0.4177, "step": 1861 }, { "epoch": 0.35, "learning_rate": 4.726195229326709e-06, "loss": 0.4242, "step": 1862 }, { "epoch": 0.35, "learning_rate": 4.7258470761685035e-06, "loss": 0.4592, "step": 1863 }, { "epoch": 0.35, "learning_rate": 4.725498714645982e-06, "loss": 0.6077, "step": 1864 }, { "epoch": 0.35, "learning_rate": 4.7251501447917535e-06, "loss": 0.5091, "step": 1865 }, { "epoch": 0.35, "learning_rate": 4.72480136663845e-06, "loss": 0.5196, "step": 1866 }, { "epoch": 0.35, "learning_rate": 4.7244523802187216e-06, "loss": 0.4731, "step": 1867 }, { "epoch": 0.35, "learning_rate": 4.724103185565234e-06, "loss": 0.6743, "step": 1868 }, { "epoch": 0.35, "learning_rate": 4.723753782710679e-06, "loss": 0.4764, "step": 1869 }, { "epoch": 0.35, "learning_rate": 4.723404171687763e-06, "loss": 0.4448, "step": 1870 }, { "epoch": 0.35, "learning_rate": 4.723054352529215e-06, "loss": 0.584, "step": 1871 }, { "epoch": 0.35, "learning_rate": 4.722704325267781e-06, "loss": 0.6399, "step": 1872 }, { "epoch": 0.35, "learning_rate": 4.722354089936226e-06, "loss": 0.7807, "step": 1873 }, { "epoch": 0.35, "learning_rate": 4.722003646567338e-06, "loss": 0.4929, "step": 1874 }, { "epoch": 0.35, "learning_rate": 4.721652995193923e-06, "loss": 0.5943, "step": 1875 }, { "epoch": 0.35, "learning_rate": 4.721302135848803e-06, "loss": 0.5554, "step": 1876 }, { "epoch": 0.35, "learning_rate": 4.7209510685648254e-06, "loss": 0.4578, "step": 1877 }, { "epoch": 0.35, "learning_rate": 4.7205997933748525e-06, "loss": 0.4944, "step": 1878 }, { "epoch": 0.35, "learning_rate": 4.720248310311768e-06, "loss": 0.6302, "step": 1879 }, { "epoch": 0.36, "learning_rate": 4.719896619408475e-06, "loss": 0.5205, "step": 1880 }, { "epoch": 0.36, "learning_rate": 4.719544720697895e-06, "loss": 0.6191, "step": 1881 }, { "epoch": 0.36, "learning_rate": 4.71919261421297e-06, "loss": 0.5517, "step": 1882 }, { "epoch": 0.36, "learning_rate": 4.7188402999866615e-06, "loss": 0.5153, "step": 1883 }, { "epoch": 0.36, "learning_rate": 4.718487778051949e-06, "loss": 0.5559, "step": 1884 }, { "epoch": 0.36, "learning_rate": 4.718135048441833e-06, "loss": 0.6115, "step": 1885 }, { "epoch": 0.36, "learning_rate": 4.717782111189334e-06, "loss": 0.5038, "step": 1886 }, { "epoch": 0.36, "learning_rate": 4.717428966327489e-06, "loss": 0.6282, "step": 1887 }, { "epoch": 0.36, "learning_rate": 4.717075613889359e-06, "loss": 0.7817, "step": 1888 }, { "epoch": 0.36, "learning_rate": 4.71672205390802e-06, "loss": 0.5958, "step": 1889 }, { "epoch": 0.36, "learning_rate": 4.716368286416568e-06, "loss": 0.5075, "step": 1890 }, { "epoch": 0.36, "learning_rate": 4.716014311448121e-06, "loss": 0.5506, "step": 1891 }, { "epoch": 0.36, "learning_rate": 4.715660129035816e-06, "loss": 0.5715, "step": 1892 }, { "epoch": 0.36, "learning_rate": 4.715305739212808e-06, "loss": 0.4871, "step": 1893 }, { "epoch": 0.36, "learning_rate": 4.714951142012271e-06, "loss": 0.5694, "step": 1894 }, { "epoch": 0.36, "learning_rate": 4.714596337467399e-06, "loss": 0.6418, "step": 1895 }, { "epoch": 0.36, "learning_rate": 4.714241325611406e-06, "loss": 0.6048, "step": 1896 }, { "epoch": 0.36, "learning_rate": 4.713886106477526e-06, "loss": 0.5978, "step": 1897 }, { "epoch": 0.36, "learning_rate": 4.713530680099011e-06, "loss": 0.5889, "step": 1898 }, { "epoch": 0.36, "learning_rate": 4.7131750465091315e-06, "loss": 0.4346, "step": 1899 }, { "epoch": 0.36, "learning_rate": 4.712819205741183e-06, "loss": 0.5184, "step": 1900 }, { "epoch": 0.36, "learning_rate": 4.7124631578284705e-06, "loss": 0.6289, "step": 1901 }, { "epoch": 0.36, "learning_rate": 4.712106902804329e-06, "loss": 0.5831, "step": 1902 }, { "epoch": 0.36, "learning_rate": 4.711750440702104e-06, "loss": 0.4027, "step": 1903 }, { "epoch": 0.36, "learning_rate": 4.711393771555168e-06, "loss": 0.5623, "step": 1904 }, { "epoch": 0.36, "learning_rate": 4.711036895396907e-06, "loss": 0.6528, "step": 1905 }, { "epoch": 0.36, "learning_rate": 4.710679812260728e-06, "loss": 0.5223, "step": 1906 }, { "epoch": 0.36, "learning_rate": 4.71032252218006e-06, "loss": 0.4799, "step": 1907 }, { "epoch": 0.36, "learning_rate": 4.709965025188349e-06, "loss": 0.7021, "step": 1908 }, { "epoch": 0.36, "learning_rate": 4.70960732131906e-06, "loss": 0.4209, "step": 1909 }, { "epoch": 0.36, "learning_rate": 4.709249410605678e-06, "loss": 0.5112, "step": 1910 }, { "epoch": 0.36, "learning_rate": 4.708891293081708e-06, "loss": 0.6118, "step": 1911 }, { "epoch": 0.36, "learning_rate": 4.708532968780673e-06, "loss": 0.4864, "step": 1912 }, { "epoch": 0.36, "learning_rate": 4.708174437736117e-06, "loss": 0.514, "step": 1913 }, { "epoch": 0.36, "learning_rate": 4.707815699981603e-06, "loss": 0.4199, "step": 1914 }, { "epoch": 0.36, "learning_rate": 4.707456755550712e-06, "loss": 0.484, "step": 1915 }, { "epoch": 0.36, "learning_rate": 4.707097604477045e-06, "loss": 0.5639, "step": 1916 }, { "epoch": 0.36, "learning_rate": 4.706738246794223e-06, "loss": 0.487, "step": 1917 }, { "epoch": 0.36, "learning_rate": 4.706378682535887e-06, "loss": 0.4232, "step": 1918 }, { "epoch": 0.36, "learning_rate": 4.706018911735695e-06, "loss": 0.5668, "step": 1919 }, { "epoch": 0.36, "learning_rate": 4.705658934427325e-06, "loss": 0.4551, "step": 1920 }, { "epoch": 0.36, "learning_rate": 4.705298750644477e-06, "loss": 0.5051, "step": 1921 }, { "epoch": 0.36, "learning_rate": 4.704938360420866e-06, "loss": 0.534, "step": 1922 }, { "epoch": 0.36, "learning_rate": 4.70457776379023e-06, "loss": 0.5348, "step": 1923 }, { "epoch": 0.36, "learning_rate": 4.704216960786324e-06, "loss": 0.5885, "step": 1924 }, { "epoch": 0.36, "learning_rate": 4.703855951442924e-06, "loss": 0.5824, "step": 1925 }, { "epoch": 0.36, "learning_rate": 4.703494735793824e-06, "loss": 0.6114, "step": 1926 }, { "epoch": 0.36, "learning_rate": 4.703133313872839e-06, "loss": 0.6205, "step": 1927 }, { "epoch": 0.36, "learning_rate": 4.7027716857138e-06, "loss": 0.5853, "step": 1928 }, { "epoch": 0.36, "learning_rate": 4.702409851350561e-06, "loss": 0.5031, "step": 1929 }, { "epoch": 0.36, "learning_rate": 4.702047810816994e-06, "loss": 0.5777, "step": 1930 }, { "epoch": 0.36, "learning_rate": 4.70168556414699e-06, "loss": 0.5111, "step": 1931 }, { "epoch": 0.37, "learning_rate": 4.701323111374457e-06, "loss": 0.5422, "step": 1932 }, { "epoch": 0.37, "learning_rate": 4.700960452533329e-06, "loss": 0.411, "step": 1933 }, { "epoch": 0.37, "learning_rate": 4.70059758765755e-06, "loss": 0.5702, "step": 1934 }, { "epoch": 0.37, "learning_rate": 4.700234516781093e-06, "loss": 0.5002, "step": 1935 }, { "epoch": 0.37, "learning_rate": 4.699871239937942e-06, "loss": 0.5484, "step": 1936 }, { "epoch": 0.37, "learning_rate": 4.699507757162105e-06, "loss": 0.468, "step": 1937 }, { "epoch": 0.37, "learning_rate": 4.699144068487608e-06, "loss": 0.6619, "step": 1938 }, { "epoch": 0.37, "learning_rate": 4.698780173948497e-06, "loss": 0.4849, "step": 1939 }, { "epoch": 0.37, "learning_rate": 4.698416073578835e-06, "loss": 0.4779, "step": 1940 }, { "epoch": 0.37, "learning_rate": 4.698051767412708e-06, "loss": 0.4511, "step": 1941 }, { "epoch": 0.37, "learning_rate": 4.697687255484218e-06, "loss": 0.4429, "step": 1942 }, { "epoch": 0.37, "learning_rate": 4.6973225378274875e-06, "loss": 0.529, "step": 1943 }, { "epoch": 0.37, "learning_rate": 4.696957614476658e-06, "loss": 0.5215, "step": 1944 }, { "epoch": 0.37, "learning_rate": 4.696592485465891e-06, "loss": 0.726, "step": 1945 }, { "epoch": 0.37, "learning_rate": 4.696227150829367e-06, "loss": 0.7487, "step": 1946 }, { "epoch": 0.37, "learning_rate": 4.695861610601283e-06, "loss": 0.4786, "step": 1947 }, { "epoch": 0.37, "learning_rate": 4.695495864815861e-06, "loss": 0.6365, "step": 1948 }, { "epoch": 0.37, "learning_rate": 4.695129913507336e-06, "loss": 0.4259, "step": 1949 }, { "epoch": 0.37, "learning_rate": 4.6947637567099675e-06, "loss": 0.5947, "step": 1950 }, { "epoch": 0.37, "learning_rate": 4.6943973944580305e-06, "loss": 0.5313, "step": 1951 }, { "epoch": 0.37, "learning_rate": 4.694030826785821e-06, "loss": 0.476, "step": 1952 }, { "epoch": 0.37, "learning_rate": 4.693664053727654e-06, "loss": 0.4745, "step": 1953 }, { "epoch": 0.37, "learning_rate": 4.6932970753178645e-06, "loss": 0.5921, "step": 1954 }, { "epoch": 0.37, "learning_rate": 4.692929891590803e-06, "loss": 0.5052, "step": 1955 }, { "epoch": 0.37, "learning_rate": 4.692562502580845e-06, "loss": 0.4646, "step": 1956 }, { "epoch": 0.37, "learning_rate": 4.69219490832238e-06, "loss": 0.4447, "step": 1957 }, { "epoch": 0.37, "learning_rate": 4.691827108849822e-06, "loss": 0.5452, "step": 1958 }, { "epoch": 0.37, "learning_rate": 4.6914591041975965e-06, "loss": 0.5129, "step": 1959 }, { "epoch": 0.37, "learning_rate": 4.691090894400157e-06, "loss": 0.483, "step": 1960 }, { "epoch": 0.37, "learning_rate": 4.69072247949197e-06, "loss": 0.4292, "step": 1961 }, { "epoch": 0.37, "learning_rate": 4.690353859507525e-06, "loss": 0.4851, "step": 1962 }, { "epoch": 0.37, "learning_rate": 4.689985034481327e-06, "loss": 0.4637, "step": 1963 }, { "epoch": 0.37, "learning_rate": 4.689616004447903e-06, "loss": 0.3809, "step": 1964 }, { "epoch": 0.37, "learning_rate": 4.6892467694417985e-06, "loss": 0.6844, "step": 1965 }, { "epoch": 0.37, "learning_rate": 4.688877329497579e-06, "loss": 0.405, "step": 1966 }, { "epoch": 0.37, "learning_rate": 4.6885076846498255e-06, "loss": 0.5339, "step": 1967 }, { "epoch": 0.37, "learning_rate": 4.688137834933144e-06, "loss": 0.4084, "step": 1968 }, { "epoch": 0.37, "learning_rate": 4.687767780382154e-06, "loss": 0.4241, "step": 1969 }, { "epoch": 0.37, "learning_rate": 4.6873975210314995e-06, "loss": 0.6542, "step": 1970 }, { "epoch": 0.37, "learning_rate": 4.6870270569158375e-06, "loss": 0.5755, "step": 1971 }, { "epoch": 0.37, "learning_rate": 4.686656388069851e-06, "loss": 0.4803, "step": 1972 }, { "epoch": 0.37, "learning_rate": 4.686285514528236e-06, "loss": 0.5289, "step": 1973 }, { "epoch": 0.37, "learning_rate": 4.685914436325713e-06, "loss": 0.4835, "step": 1974 }, { "epoch": 0.37, "learning_rate": 4.685543153497017e-06, "loss": 0.617, "step": 1975 }, { "epoch": 0.37, "learning_rate": 4.6851716660769055e-06, "loss": 0.6558, "step": 1976 }, { "epoch": 0.37, "learning_rate": 4.684799974100153e-06, "loss": 0.5903, "step": 1977 }, { "epoch": 0.37, "learning_rate": 4.684428077601555e-06, "loss": 0.5333, "step": 1978 }, { "epoch": 0.37, "learning_rate": 4.684055976615924e-06, "loss": 0.4824, "step": 1979 }, { "epoch": 0.37, "learning_rate": 4.683683671178094e-06, "loss": 0.5918, "step": 1980 }, { "epoch": 0.37, "learning_rate": 4.683311161322917e-06, "loss": 0.4445, "step": 1981 }, { "epoch": 0.37, "learning_rate": 4.682938447085263e-06, "loss": 0.5209, "step": 1982 }, { "epoch": 0.37, "learning_rate": 4.682565528500022e-06, "loss": 0.5208, "step": 1983 }, { "epoch": 0.37, "learning_rate": 4.682192405602106e-06, "loss": 0.4834, "step": 1984 }, { "epoch": 0.38, "learning_rate": 4.68181907842644e-06, "loss": 0.5981, "step": 1985 }, { "epoch": 0.38, "learning_rate": 4.681445547007975e-06, "loss": 0.4354, "step": 1986 }, { "epoch": 0.38, "learning_rate": 4.681071811381676e-06, "loss": 0.5661, "step": 1987 }, { "epoch": 0.38, "learning_rate": 4.680697871582528e-06, "loss": 0.639, "step": 1988 }, { "epoch": 0.38, "learning_rate": 4.6803237276455375e-06, "loss": 0.5288, "step": 1989 }, { "epoch": 0.38, "learning_rate": 4.679949379605727e-06, "loss": 0.6504, "step": 1990 }, { "epoch": 0.38, "learning_rate": 4.679574827498142e-06, "loss": 0.4224, "step": 1991 }, { "epoch": 0.38, "learning_rate": 4.679200071357843e-06, "loss": 0.6045, "step": 1992 }, { "epoch": 0.38, "learning_rate": 4.678825111219911e-06, "loss": 0.5909, "step": 1993 }, { "epoch": 0.38, "learning_rate": 4.6784499471194486e-06, "loss": 0.5501, "step": 1994 }, { "epoch": 0.38, "learning_rate": 4.678074579091574e-06, "loss": 0.5372, "step": 1995 }, { "epoch": 0.38, "learning_rate": 4.6776990071714255e-06, "loss": 0.5042, "step": 1996 }, { "epoch": 0.38, "learning_rate": 4.677323231394161e-06, "loss": 0.7119, "step": 1997 }, { "epoch": 0.38, "learning_rate": 4.676947251794959e-06, "loss": 0.6227, "step": 1998 }, { "epoch": 0.38, "learning_rate": 4.676571068409013e-06, "loss": 0.5174, "step": 1999 }, { "epoch": 0.38, "learning_rate": 4.676194681271539e-06, "loss": 0.5827, "step": 2000 }, { "epoch": 0.38, "learning_rate": 4.675818090417772e-06, "loss": 0.5144, "step": 2001 }, { "epoch": 0.38, "learning_rate": 4.675441295882964e-06, "loss": 0.5826, "step": 2002 }, { "epoch": 0.38, "learning_rate": 4.675064297702388e-06, "loss": 0.5051, "step": 2003 }, { "epoch": 0.38, "learning_rate": 4.6746870959113335e-06, "loss": 0.6286, "step": 2004 }, { "epoch": 0.38, "learning_rate": 4.674309690545113e-06, "loss": 0.4852, "step": 2005 }, { "epoch": 0.38, "learning_rate": 4.673932081639054e-06, "loss": 0.5793, "step": 2006 }, { "epoch": 0.38, "learning_rate": 4.673554269228507e-06, "loss": 0.5559, "step": 2007 }, { "epoch": 0.38, "learning_rate": 4.673176253348838e-06, "loss": 0.477, "step": 2008 }, { "epoch": 0.38, "learning_rate": 4.6727980340354354e-06, "loss": 0.708, "step": 2009 }, { "epoch": 0.38, "learning_rate": 4.672419611323702e-06, "loss": 0.4882, "step": 2010 }, { "epoch": 0.38, "learning_rate": 4.6720409852490635e-06, "loss": 0.5862, "step": 2011 }, { "epoch": 0.38, "learning_rate": 4.671662155846964e-06, "loss": 0.6454, "step": 2012 }, { "epoch": 0.38, "learning_rate": 4.671283123152867e-06, "loss": 0.619, "step": 2013 }, { "epoch": 0.38, "learning_rate": 4.670903887202252e-06, "loss": 0.4741, "step": 2014 }, { "epoch": 0.38, "learning_rate": 4.670524448030621e-06, "loss": 0.5576, "step": 2015 }, { "epoch": 0.38, "learning_rate": 4.670144805673495e-06, "loss": 0.484, "step": 2016 }, { "epoch": 0.38, "learning_rate": 4.669764960166411e-06, "loss": 0.471, "step": 2017 }, { "epoch": 0.38, "learning_rate": 4.6693849115449265e-06, "loss": 0.512, "step": 2018 }, { "epoch": 0.38, "learning_rate": 4.66900465984462e-06, "loss": 0.575, "step": 2019 }, { "epoch": 0.38, "learning_rate": 4.668624205101087e-06, "loss": 0.5141, "step": 2020 }, { "epoch": 0.38, "learning_rate": 4.668243547349941e-06, "loss": 0.4272, "step": 2021 }, { "epoch": 0.38, "learning_rate": 4.667862686626817e-06, "loss": 0.5876, "step": 2022 }, { "epoch": 0.38, "learning_rate": 4.667481622967368e-06, "loss": 0.6682, "step": 2023 }, { "epoch": 0.38, "learning_rate": 4.667100356407265e-06, "loss": 0.5153, "step": 2024 }, { "epoch": 0.38, "learning_rate": 4.6667188869822e-06, "loss": 0.6024, "step": 2025 }, { "epoch": 0.38, "learning_rate": 4.666337214727881e-06, "loss": 0.5658, "step": 2026 }, { "epoch": 0.38, "learning_rate": 4.6659553396800385e-06, "loss": 0.4166, "step": 2027 }, { "epoch": 0.38, "learning_rate": 4.665573261874419e-06, "loss": 0.5131, "step": 2028 }, { "epoch": 0.38, "learning_rate": 4.6651909813467906e-06, "loss": 0.4748, "step": 2029 }, { "epoch": 0.38, "learning_rate": 4.664808498132938e-06, "loss": 0.6478, "step": 2030 }, { "epoch": 0.38, "learning_rate": 4.664425812268667e-06, "loss": 0.4555, "step": 2031 }, { "epoch": 0.38, "learning_rate": 4.6640429237898e-06, "loss": 0.6967, "step": 2032 }, { "epoch": 0.38, "learning_rate": 4.663659832732181e-06, "loss": 0.519, "step": 2033 }, { "epoch": 0.38, "learning_rate": 4.66327653913167e-06, "loss": 0.5397, "step": 2034 }, { "epoch": 0.38, "learning_rate": 4.66289304302415e-06, "loss": 0.5428, "step": 2035 }, { "epoch": 0.38, "learning_rate": 4.662509344445518e-06, "loss": 0.5797, "step": 2036 }, { "epoch": 0.38, "learning_rate": 4.662125443431694e-06, "loss": 0.412, "step": 2037 }, { "epoch": 0.39, "learning_rate": 4.661741340018615e-06, "loss": 0.5664, "step": 2038 }, { "epoch": 0.39, "learning_rate": 4.661357034242237e-06, "loss": 0.4676, "step": 2039 }, { "epoch": 0.39, "learning_rate": 4.660972526138536e-06, "loss": 0.4404, "step": 2040 }, { "epoch": 0.39, "learning_rate": 4.6605878157435055e-06, "loss": 0.4846, "step": 2041 }, { "epoch": 0.39, "learning_rate": 4.660202903093161e-06, "loss": 0.6982, "step": 2042 }, { "epoch": 0.39, "learning_rate": 4.6598177882235316e-06, "loss": 0.5708, "step": 2043 }, { "epoch": 0.39, "learning_rate": 4.65943247117067e-06, "loss": 0.4908, "step": 2044 }, { "epoch": 0.39, "learning_rate": 4.659046951970646e-06, "loss": 0.4927, "step": 2045 }, { "epoch": 0.39, "learning_rate": 4.658661230659548e-06, "loss": 0.4784, "step": 2046 }, { "epoch": 0.39, "learning_rate": 4.658275307273485e-06, "loss": 0.5356, "step": 2047 }, { "epoch": 0.39, "learning_rate": 4.6578891818485825e-06, "loss": 0.4328, "step": 2048 }, { "epoch": 0.39, "learning_rate": 4.6575028544209875e-06, "loss": 0.4986, "step": 2049 }, { "epoch": 0.39, "learning_rate": 4.657116325026863e-06, "loss": 0.5555, "step": 2050 }, { "epoch": 0.39, "learning_rate": 4.656729593702395e-06, "loss": 0.5084, "step": 2051 }, { "epoch": 0.39, "learning_rate": 4.656342660483783e-06, "loss": 0.6049, "step": 2052 }, { "epoch": 0.39, "learning_rate": 4.6559555254072495e-06, "loss": 0.5091, "step": 2053 }, { "epoch": 0.39, "learning_rate": 4.655568188509035e-06, "loss": 0.5758, "step": 2054 }, { "epoch": 0.39, "learning_rate": 4.655180649825398e-06, "loss": 0.4512, "step": 2055 }, { "epoch": 0.39, "learning_rate": 4.654792909392617e-06, "loss": 0.5227, "step": 2056 }, { "epoch": 0.39, "learning_rate": 4.654404967246989e-06, "loss": 0.5669, "step": 2057 }, { "epoch": 0.39, "learning_rate": 4.654016823424829e-06, "loss": 0.5142, "step": 2058 }, { "epoch": 0.39, "learning_rate": 4.653628477962472e-06, "loss": 0.5866, "step": 2059 }, { "epoch": 0.39, "learning_rate": 4.653239930896272e-06, "loss": 0.5907, "step": 2060 }, { "epoch": 0.39, "learning_rate": 4.652851182262601e-06, "loss": 0.4696, "step": 2061 }, { "epoch": 0.39, "learning_rate": 4.652462232097849e-06, "loss": 0.4331, "step": 2062 }, { "epoch": 0.39, "learning_rate": 4.65207308043843e-06, "loss": 0.4678, "step": 2063 }, { "epoch": 0.39, "learning_rate": 4.651683727320769e-06, "loss": 0.422, "step": 2064 }, { "epoch": 0.39, "learning_rate": 4.651294172781314e-06, "loss": 0.4918, "step": 2065 }, { "epoch": 0.39, "learning_rate": 4.650904416856533e-06, "loss": 0.6955, "step": 2066 }, { "epoch": 0.39, "learning_rate": 4.650514459582912e-06, "loss": 0.6471, "step": 2067 }, { "epoch": 0.39, "learning_rate": 4.650124300996955e-06, "loss": 0.6038, "step": 2068 }, { "epoch": 0.39, "learning_rate": 4.649733941135183e-06, "loss": 0.4792, "step": 2069 }, { "epoch": 0.39, "learning_rate": 4.649343380034141e-06, "loss": 0.6381, "step": 2070 }, { "epoch": 0.39, "learning_rate": 4.64895261773039e-06, "loss": 0.6206, "step": 2071 }, { "epoch": 0.39, "learning_rate": 4.648561654260508e-06, "loss": 0.6248, "step": 2072 }, { "epoch": 0.39, "learning_rate": 4.648170489661095e-06, "loss": 0.4403, "step": 2073 }, { "epoch": 0.39, "learning_rate": 4.647779123968766e-06, "loss": 0.6163, "step": 2074 }, { "epoch": 0.39, "learning_rate": 4.64738755722016e-06, "loss": 0.5282, "step": 2075 }, { "epoch": 0.39, "learning_rate": 4.64699578945193e-06, "loss": 0.5411, "step": 2076 }, { "epoch": 0.39, "learning_rate": 4.646603820700752e-06, "loss": 0.5433, "step": 2077 }, { "epoch": 0.39, "learning_rate": 4.6462116510033175e-06, "loss": 0.6112, "step": 2078 }, { "epoch": 0.39, "learning_rate": 4.645819280396336e-06, "loss": 0.4611, "step": 2079 }, { "epoch": 0.39, "learning_rate": 4.645426708916542e-06, "loss": 0.6307, "step": 2080 }, { "epoch": 0.39, "learning_rate": 4.645033936600682e-06, "loss": 0.579, "step": 2081 }, { "epoch": 0.39, "learning_rate": 4.6446409634855235e-06, "loss": 0.4963, "step": 2082 }, { "epoch": 0.39, "learning_rate": 4.6442477896078545e-06, "loss": 0.5381, "step": 2083 }, { "epoch": 0.39, "learning_rate": 4.643854415004479e-06, "loss": 0.5486, "step": 2084 }, { "epoch": 0.39, "learning_rate": 4.643460839712224e-06, "loss": 0.4938, "step": 2085 }, { "epoch": 0.39, "learning_rate": 4.6430670637679295e-06, "loss": 0.5756, "step": 2086 }, { "epoch": 0.39, "learning_rate": 4.642673087208459e-06, "loss": 0.5391, "step": 2087 }, { "epoch": 0.39, "learning_rate": 4.642278910070693e-06, "loss": 0.466, "step": 2088 }, { "epoch": 0.39, "learning_rate": 4.641884532391531e-06, "loss": 0.5695, "step": 2089 }, { "epoch": 0.39, "learning_rate": 4.641489954207891e-06, "loss": 0.444, "step": 2090 }, { "epoch": 0.4, "learning_rate": 4.641095175556709e-06, "loss": 0.4443, "step": 2091 }, { "epoch": 0.4, "learning_rate": 4.640700196474942e-06, "loss": 0.4108, "step": 2092 }, { "epoch": 0.4, "learning_rate": 4.640305016999565e-06, "loss": 0.5497, "step": 2093 }, { "epoch": 0.4, "learning_rate": 4.639909637167569e-06, "loss": 0.4659, "step": 2094 }, { "epoch": 0.4, "learning_rate": 4.639514057015968e-06, "loss": 0.4531, "step": 2095 }, { "epoch": 0.4, "learning_rate": 4.639118276581792e-06, "loss": 0.5841, "step": 2096 }, { "epoch": 0.4, "learning_rate": 4.63872229590209e-06, "loss": 0.6175, "step": 2097 }, { "epoch": 0.4, "learning_rate": 4.638326115013932e-06, "loss": 0.5582, "step": 2098 }, { "epoch": 0.4, "learning_rate": 4.637929733954404e-06, "loss": 0.6223, "step": 2099 }, { "epoch": 0.4, "learning_rate": 4.63753315276061e-06, "loss": 0.4718, "step": 2100 }, { "epoch": 0.4, "learning_rate": 4.637136371469677e-06, "loss": 0.4255, "step": 2101 }, { "epoch": 0.4, "learning_rate": 4.636739390118747e-06, "loss": 0.6154, "step": 2102 }, { "epoch": 0.4, "learning_rate": 4.636342208744981e-06, "loss": 0.6895, "step": 2103 }, { "epoch": 0.4, "learning_rate": 4.635944827385563e-06, "loss": 0.5694, "step": 2104 }, { "epoch": 0.4, "learning_rate": 4.635547246077689e-06, "loss": 0.6282, "step": 2105 }, { "epoch": 0.4, "learning_rate": 4.635149464858579e-06, "loss": 0.6998, "step": 2106 }, { "epoch": 0.4, "learning_rate": 4.6347514837654674e-06, "loss": 0.6461, "step": 2107 }, { "epoch": 0.4, "learning_rate": 4.634353302835612e-06, "loss": 0.6244, "step": 2108 }, { "epoch": 0.4, "learning_rate": 4.633954922106286e-06, "loss": 0.4382, "step": 2109 }, { "epoch": 0.4, "learning_rate": 4.633556341614783e-06, "loss": 0.6055, "step": 2110 }, { "epoch": 0.4, "learning_rate": 4.633157561398414e-06, "loss": 0.4842, "step": 2111 }, { "epoch": 0.4, "learning_rate": 4.63275858149451e-06, "loss": 0.5361, "step": 2112 }, { "epoch": 0.4, "learning_rate": 4.6323594019404195e-06, "loss": 0.3862, "step": 2113 }, { "epoch": 0.4, "learning_rate": 4.6319600227735095e-06, "loss": 0.6007, "step": 2114 }, { "epoch": 0.4, "learning_rate": 4.631560444031168e-06, "loss": 0.5266, "step": 2115 }, { "epoch": 0.4, "learning_rate": 4.631160665750799e-06, "loss": 0.5268, "step": 2116 }, { "epoch": 0.4, "learning_rate": 4.6307606879698265e-06, "loss": 0.5076, "step": 2117 }, { "epoch": 0.4, "learning_rate": 4.630360510725694e-06, "loss": 0.5386, "step": 2118 }, { "epoch": 0.4, "learning_rate": 4.629960134055859e-06, "loss": 0.5089, "step": 2119 }, { "epoch": 0.4, "learning_rate": 4.629559557997805e-06, "loss": 0.5637, "step": 2120 }, { "epoch": 0.4, "learning_rate": 4.629158782589029e-06, "loss": 0.5383, "step": 2121 }, { "epoch": 0.4, "learning_rate": 4.628757807867048e-06, "loss": 0.5535, "step": 2122 }, { "epoch": 0.4, "learning_rate": 4.628356633869399e-06, "loss": 0.4629, "step": 2123 }, { "epoch": 0.4, "learning_rate": 4.627955260633634e-06, "loss": 0.6775, "step": 2124 }, { "epoch": 0.4, "learning_rate": 4.627553688197328e-06, "loss": 0.4899, "step": 2125 }, { "epoch": 0.4, "learning_rate": 4.627151916598071e-06, "loss": 0.4527, "step": 2126 }, { "epoch": 0.4, "learning_rate": 4.626749945873477e-06, "loss": 0.7117, "step": 2127 }, { "epoch": 0.4, "learning_rate": 4.626347776061171e-06, "loss": 0.481, "step": 2128 }, { "epoch": 0.4, "learning_rate": 4.625945407198801e-06, "loss": 0.5713, "step": 2129 }, { "epoch": 0.4, "learning_rate": 4.625542839324036e-06, "loss": 0.7661, "step": 2130 }, { "epoch": 0.4, "learning_rate": 4.625140072474558e-06, "loss": 0.6642, "step": 2131 }, { "epoch": 0.4, "learning_rate": 4.6247371066880725e-06, "loss": 0.5821, "step": 2132 }, { "epoch": 0.4, "learning_rate": 4.6243339420023e-06, "loss": 0.4836, "step": 2133 }, { "epoch": 0.4, "learning_rate": 4.6239305784549825e-06, "loss": 0.4412, "step": 2134 }, { "epoch": 0.4, "learning_rate": 4.623527016083879e-06, "loss": 0.4423, "step": 2135 }, { "epoch": 0.4, "learning_rate": 4.623123254926767e-06, "loss": 0.6073, "step": 2136 }, { "epoch": 0.4, "learning_rate": 4.6227192950214435e-06, "loss": 0.5222, "step": 2137 }, { "epoch": 0.4, "learning_rate": 4.622315136405725e-06, "loss": 0.5641, "step": 2138 }, { "epoch": 0.4, "learning_rate": 4.621910779117443e-06, "loss": 0.4745, "step": 2139 }, { "epoch": 0.4, "learning_rate": 4.621506223194451e-06, "loss": 0.4308, "step": 2140 }, { "epoch": 0.4, "learning_rate": 4.621101468674619e-06, "loss": 0.4976, "step": 2141 }, { "epoch": 0.4, "learning_rate": 4.620696515595838e-06, "loss": 0.4568, "step": 2142 }, { "epoch": 0.4, "learning_rate": 4.620291363996016e-06, "loss": 0.4594, "step": 2143 }, { "epoch": 0.41, "learning_rate": 4.619886013913078e-06, "loss": 0.6949, "step": 2144 }, { "epoch": 0.41, "learning_rate": 4.619480465384972e-06, "loss": 0.5021, "step": 2145 }, { "epoch": 0.41, "learning_rate": 4.61907471844966e-06, "loss": 0.4847, "step": 2146 }, { "epoch": 0.41, "learning_rate": 4.618668773145125e-06, "loss": 0.6469, "step": 2147 }, { "epoch": 0.41, "learning_rate": 4.618262629509368e-06, "loss": 0.5158, "step": 2148 }, { "epoch": 0.41, "learning_rate": 4.617856287580409e-06, "loss": 0.4505, "step": 2149 }, { "epoch": 0.41, "learning_rate": 4.617449747396286e-06, "loss": 0.4543, "step": 2150 }, { "epoch": 0.41, "learning_rate": 4.6170430089950555e-06, "loss": 0.6532, "step": 2151 }, { "epoch": 0.41, "learning_rate": 4.6166360724147916e-06, "loss": 0.581, "step": 2152 }, { "epoch": 0.41, "learning_rate": 4.616228937693591e-06, "loss": 0.5929, "step": 2153 }, { "epoch": 0.41, "learning_rate": 4.615821604869564e-06, "loss": 0.3743, "step": 2154 }, { "epoch": 0.41, "learning_rate": 4.615414073980842e-06, "loss": 0.3778, "step": 2155 }, { "epoch": 0.41, "learning_rate": 4.615006345065575e-06, "loss": 0.4488, "step": 2156 }, { "epoch": 0.41, "learning_rate": 4.61459841816193e-06, "loss": 0.539, "step": 2157 }, { "epoch": 0.41, "learning_rate": 4.614190293308094e-06, "loss": 0.4208, "step": 2158 }, { "epoch": 0.41, "learning_rate": 4.6137819705422716e-06, "loss": 0.5362, "step": 2159 }, { "epoch": 0.41, "learning_rate": 4.6133734499026876e-06, "loss": 0.4934, "step": 2160 }, { "epoch": 0.41, "learning_rate": 4.612964731427584e-06, "loss": 0.6853, "step": 2161 }, { "epoch": 0.41, "learning_rate": 4.61255581515522e-06, "loss": 0.617, "step": 2162 }, { "epoch": 0.41, "learning_rate": 4.612146701123875e-06, "loss": 0.5686, "step": 2163 }, { "epoch": 0.41, "learning_rate": 4.611737389371849e-06, "loss": 0.5242, "step": 2164 }, { "epoch": 0.41, "learning_rate": 4.6113278799374565e-06, "loss": 0.5589, "step": 2165 }, { "epoch": 0.41, "learning_rate": 4.610918172859032e-06, "loss": 0.7373, "step": 2166 }, { "epoch": 0.41, "learning_rate": 4.610508268174928e-06, "loss": 0.4927, "step": 2167 }, { "epoch": 0.41, "learning_rate": 4.610098165923518e-06, "loss": 0.649, "step": 2168 }, { "epoch": 0.41, "learning_rate": 4.609687866143191e-06, "loss": 0.5991, "step": 2169 }, { "epoch": 0.41, "learning_rate": 4.609277368872354e-06, "loss": 0.4778, "step": 2170 }, { "epoch": 0.41, "learning_rate": 4.6088666741494395e-06, "loss": 0.4477, "step": 2171 }, { "epoch": 0.41, "learning_rate": 4.6084557820128875e-06, "loss": 0.4713, "step": 2172 }, { "epoch": 0.41, "learning_rate": 4.608044692501165e-06, "loss": 0.5039, "step": 2173 }, { "epoch": 0.41, "learning_rate": 4.607633405652755e-06, "loss": 0.5595, "step": 2174 }, { "epoch": 0.41, "learning_rate": 4.6072219215061555e-06, "loss": 0.4951, "step": 2175 }, { "epoch": 0.41, "learning_rate": 4.60681024009989e-06, "loss": 0.5547, "step": 2176 }, { "epoch": 0.41, "learning_rate": 4.6063983614724936e-06, "loss": 0.5051, "step": 2177 }, { "epoch": 0.41, "learning_rate": 4.6059862856625235e-06, "loss": 0.6048, "step": 2178 }, { "epoch": 0.41, "learning_rate": 4.6055740127085555e-06, "loss": 0.5739, "step": 2179 }, { "epoch": 0.41, "learning_rate": 4.605161542649182e-06, "loss": 0.4837, "step": 2180 }, { "epoch": 0.41, "learning_rate": 4.604748875523016e-06, "loss": 0.5642, "step": 2181 }, { "epoch": 0.41, "learning_rate": 4.604336011368687e-06, "loss": 0.5345, "step": 2182 }, { "epoch": 0.41, "learning_rate": 4.603922950224843e-06, "loss": 0.5199, "step": 2183 }, { "epoch": 0.41, "learning_rate": 4.603509692130153e-06, "loss": 0.6579, "step": 2184 }, { "epoch": 0.41, "learning_rate": 4.6030962371233e-06, "loss": 0.6687, "step": 2185 }, { "epoch": 0.41, "learning_rate": 4.60268258524299e-06, "loss": 0.5568, "step": 2186 }, { "epoch": 0.41, "learning_rate": 4.6022687365279455e-06, "loss": 0.4465, "step": 2187 }, { "epoch": 0.41, "learning_rate": 4.601854691016907e-06, "loss": 0.4634, "step": 2188 }, { "epoch": 0.41, "learning_rate": 4.601440448748633e-06, "loss": 0.667, "step": 2189 }, { "epoch": 0.41, "learning_rate": 4.601026009761903e-06, "loss": 0.5397, "step": 2190 }, { "epoch": 0.41, "learning_rate": 4.600611374095511e-06, "loss": 0.4607, "step": 2191 }, { "epoch": 0.41, "learning_rate": 4.600196541788273e-06, "loss": 0.4444, "step": 2192 }, { "epoch": 0.41, "learning_rate": 4.599781512879022e-06, "loss": 0.5114, "step": 2193 }, { "epoch": 0.41, "learning_rate": 4.5993662874066094e-06, "loss": 0.5118, "step": 2194 }, { "epoch": 0.41, "learning_rate": 4.598950865409903e-06, "loss": 0.5169, "step": 2195 }, { "epoch": 0.41, "learning_rate": 4.598535246927794e-06, "loss": 0.4735, "step": 2196 }, { "epoch": 0.42, "learning_rate": 4.598119431999187e-06, "loss": 0.6309, "step": 2197 }, { "epoch": 0.42, "learning_rate": 4.597703420663008e-06, "loss": 0.6549, "step": 2198 }, { "epoch": 0.42, "learning_rate": 4.597287212958199e-06, "loss": 0.4755, "step": 2199 }, { "epoch": 0.42, "learning_rate": 4.596870808923723e-06, "loss": 0.5626, "step": 2200 }, { "epoch": 0.42, "learning_rate": 4.59645420859856e-06, "loss": 0.5647, "step": 2201 }, { "epoch": 0.42, "learning_rate": 4.596037412021708e-06, "loss": 0.4938, "step": 2202 }, { "epoch": 0.42, "learning_rate": 4.595620419232183e-06, "loss": 0.5933, "step": 2203 }, { "epoch": 0.42, "learning_rate": 4.595203230269022e-06, "loss": 0.5158, "step": 2204 }, { "epoch": 0.42, "learning_rate": 4.5947858451712775e-06, "loss": 0.6457, "step": 2205 }, { "epoch": 0.42, "learning_rate": 4.594368263978022e-06, "loss": 0.592, "step": 2206 }, { "epoch": 0.42, "learning_rate": 4.5939504867283456e-06, "loss": 0.4593, "step": 2207 }, { "epoch": 0.42, "learning_rate": 4.593532513461356e-06, "loss": 0.4975, "step": 2208 }, { "epoch": 0.42, "learning_rate": 4.593114344216182e-06, "loss": 0.5948, "step": 2209 }, { "epoch": 0.42, "learning_rate": 4.592695979031967e-06, "loss": 0.4289, "step": 2210 }, { "epoch": 0.42, "learning_rate": 4.592277417947877e-06, "loss": 0.5921, "step": 2211 }, { "epoch": 0.42, "learning_rate": 4.5918586610030916e-06, "loss": 0.5579, "step": 2212 }, { "epoch": 0.42, "learning_rate": 4.591439708236812e-06, "loss": 0.5246, "step": 2213 }, { "epoch": 0.42, "learning_rate": 4.591020559688258e-06, "loss": 0.5087, "step": 2214 }, { "epoch": 0.42, "learning_rate": 4.590601215396665e-06, "loss": 0.5102, "step": 2215 }, { "epoch": 0.42, "learning_rate": 4.59018167540129e-06, "loss": 0.5116, "step": 2216 }, { "epoch": 0.42, "learning_rate": 4.589761939741405e-06, "loss": 0.5042, "step": 2217 }, { "epoch": 0.42, "learning_rate": 4.589342008456304e-06, "loss": 0.4535, "step": 2218 }, { "epoch": 0.42, "learning_rate": 4.588921881585294e-06, "loss": 0.5078, "step": 2219 }, { "epoch": 0.42, "learning_rate": 4.588501559167706e-06, "loss": 0.6185, "step": 2220 }, { "epoch": 0.42, "learning_rate": 4.588081041242888e-06, "loss": 0.5658, "step": 2221 }, { "epoch": 0.42, "learning_rate": 4.587660327850203e-06, "loss": 0.4898, "step": 2222 }, { "epoch": 0.42, "learning_rate": 4.5872394190290344e-06, "loss": 0.5501, "step": 2223 }, { "epoch": 0.42, "learning_rate": 4.586818314818786e-06, "loss": 0.5238, "step": 2224 }, { "epoch": 0.42, "learning_rate": 4.586397015258877e-06, "loss": 0.5192, "step": 2225 }, { "epoch": 0.42, "learning_rate": 4.585975520388744e-06, "loss": 0.507, "step": 2226 }, { "epoch": 0.42, "learning_rate": 4.585553830247846e-06, "loss": 0.6248, "step": 2227 }, { "epoch": 0.42, "learning_rate": 4.585131944875657e-06, "loss": 0.6402, "step": 2228 }, { "epoch": 0.42, "learning_rate": 4.5847098643116706e-06, "loss": 0.5798, "step": 2229 }, { "epoch": 0.42, "learning_rate": 4.584287588595398e-06, "loss": 0.5408, "step": 2230 }, { "epoch": 0.42, "learning_rate": 4.583865117766368e-06, "loss": 0.6154, "step": 2231 }, { "epoch": 0.42, "learning_rate": 4.583442451864131e-06, "loss": 0.5087, "step": 2232 }, { "epoch": 0.42, "learning_rate": 4.583019590928251e-06, "loss": 0.5953, "step": 2233 }, { "epoch": 0.42, "learning_rate": 4.582596534998313e-06, "loss": 0.487, "step": 2234 }, { "epoch": 0.42, "learning_rate": 4.582173284113921e-06, "loss": 0.5819, "step": 2235 }, { "epoch": 0.42, "learning_rate": 4.581749838314695e-06, "loss": 0.527, "step": 2236 }, { "epoch": 0.42, "learning_rate": 4.5813261976402745e-06, "loss": 0.5012, "step": 2237 }, { "epoch": 0.42, "learning_rate": 4.580902362130317e-06, "loss": 0.5613, "step": 2238 }, { "epoch": 0.42, "learning_rate": 4.580478331824499e-06, "loss": 0.4047, "step": 2239 }, { "epoch": 0.42, "learning_rate": 4.580054106762513e-06, "loss": 0.6801, "step": 2240 }, { "epoch": 0.42, "learning_rate": 4.579629686984073e-06, "loss": 0.5734, "step": 2241 }, { "epoch": 0.42, "learning_rate": 4.579205072528908e-06, "loss": 0.6782, "step": 2242 }, { "epoch": 0.42, "learning_rate": 4.578780263436766e-06, "loss": 0.5499, "step": 2243 }, { "epoch": 0.42, "learning_rate": 4.5783552597474165e-06, "loss": 0.4438, "step": 2244 }, { "epoch": 0.42, "learning_rate": 4.5779300615006426e-06, "loss": 0.5773, "step": 2245 }, { "epoch": 0.42, "learning_rate": 4.577504668736249e-06, "loss": 0.5873, "step": 2246 }, { "epoch": 0.42, "learning_rate": 4.577079081494056e-06, "loss": 0.5717, "step": 2247 }, { "epoch": 0.42, "learning_rate": 4.5766532998139045e-06, "loss": 0.4329, "step": 2248 }, { "epoch": 0.42, "learning_rate": 4.576227323735651e-06, "loss": 0.5132, "step": 2249 }, { "epoch": 0.43, "learning_rate": 4.5758011532991735e-06, "loss": 0.5033, "step": 2250 }, { "epoch": 0.43, "learning_rate": 4.575374788544364e-06, "loss": 0.4337, "step": 2251 }, { "epoch": 0.43, "learning_rate": 4.574948229511137e-06, "loss": 0.4612, "step": 2252 }, { "epoch": 0.43, "learning_rate": 4.574521476239423e-06, "loss": 0.5188, "step": 2253 }, { "epoch": 0.43, "learning_rate": 4.574094528769171e-06, "loss": 0.5872, "step": 2254 }, { "epoch": 0.43, "learning_rate": 4.573667387140346e-06, "loss": 0.599, "step": 2255 }, { "epoch": 0.43, "learning_rate": 4.5732400513929354e-06, "loss": 0.4553, "step": 2256 }, { "epoch": 0.43, "learning_rate": 4.5728125215669425e-06, "loss": 0.3755, "step": 2257 }, { "epoch": 0.43, "learning_rate": 4.572384797702388e-06, "loss": 0.4818, "step": 2258 }, { "epoch": 0.43, "learning_rate": 4.571956879839313e-06, "loss": 0.5684, "step": 2259 }, { "epoch": 0.43, "learning_rate": 4.571528768017774e-06, "loss": 0.5593, "step": 2260 }, { "epoch": 0.43, "learning_rate": 4.5711004622778475e-06, "loss": 0.5092, "step": 2261 }, { "epoch": 0.43, "learning_rate": 4.570671962659628e-06, "loss": 0.6532, "step": 2262 }, { "epoch": 0.43, "learning_rate": 4.570243269203227e-06, "loss": 0.718, "step": 2263 }, { "epoch": 0.43, "learning_rate": 4.569814381948777e-06, "loss": 0.6636, "step": 2264 }, { "epoch": 0.43, "learning_rate": 4.569385300936426e-06, "loss": 0.4442, "step": 2265 }, { "epoch": 0.43, "learning_rate": 4.568956026206338e-06, "loss": 0.5964, "step": 2266 }, { "epoch": 0.43, "learning_rate": 4.568526557798701e-06, "loss": 0.5001, "step": 2267 }, { "epoch": 0.43, "learning_rate": 4.5680968957537185e-06, "loss": 0.4821, "step": 2268 }, { "epoch": 0.43, "learning_rate": 4.5676670401116095e-06, "loss": 0.4535, "step": 2269 }, { "epoch": 0.43, "learning_rate": 4.567236990912614e-06, "loss": 0.556, "step": 2270 }, { "epoch": 0.43, "learning_rate": 4.5668067481969915e-06, "loss": 0.584, "step": 2271 }, { "epoch": 0.43, "learning_rate": 4.566376312005014e-06, "loss": 0.5296, "step": 2272 }, { "epoch": 0.43, "learning_rate": 4.565945682376978e-06, "loss": 0.4922, "step": 2273 }, { "epoch": 0.43, "learning_rate": 4.565514859353193e-06, "loss": 0.5712, "step": 2274 }, { "epoch": 0.43, "learning_rate": 4.565083842973991e-06, "loss": 0.5145, "step": 2275 }, { "epoch": 0.43, "learning_rate": 4.564652633279719e-06, "loss": 0.54, "step": 2276 }, { "epoch": 0.43, "learning_rate": 4.564221230310743e-06, "loss": 0.7003, "step": 2277 }, { "epoch": 0.43, "learning_rate": 4.5637896341074475e-06, "loss": 0.5897, "step": 2278 }, { "epoch": 0.43, "learning_rate": 4.563357844710235e-06, "loss": 0.5311, "step": 2279 }, { "epoch": 0.43, "learning_rate": 4.562925862159525e-06, "loss": 0.6573, "step": 2280 }, { "epoch": 0.43, "learning_rate": 4.562493686495756e-06, "loss": 0.5141, "step": 2281 }, { "epoch": 0.43, "learning_rate": 4.5620613177593855e-06, "loss": 0.5386, "step": 2282 }, { "epoch": 0.43, "learning_rate": 4.561628755990887e-06, "loss": 0.503, "step": 2283 }, { "epoch": 0.43, "learning_rate": 4.5611960012307545e-06, "loss": 0.4982, "step": 2284 }, { "epoch": 0.43, "learning_rate": 4.5607630535194965e-06, "loss": 0.5356, "step": 2285 }, { "epoch": 0.43, "learning_rate": 4.560329912897644e-06, "loss": 0.4941, "step": 2286 }, { "epoch": 0.43, "learning_rate": 4.559896579405743e-06, "loss": 0.5162, "step": 2287 }, { "epoch": 0.43, "learning_rate": 4.5594630530843575e-06, "loss": 0.6124, "step": 2288 }, { "epoch": 0.43, "learning_rate": 4.559029333974071e-06, "loss": 0.4391, "step": 2289 }, { "epoch": 0.43, "learning_rate": 4.558595422115486e-06, "loss": 0.5191, "step": 2290 }, { "epoch": 0.43, "learning_rate": 4.55816131754922e-06, "loss": 0.4267, "step": 2291 }, { "epoch": 0.43, "learning_rate": 4.55772702031591e-06, "loss": 0.5164, "step": 2292 }, { "epoch": 0.43, "learning_rate": 4.557292530456211e-06, "loss": 0.34, "step": 2293 }, { "epoch": 0.43, "learning_rate": 4.556857848010797e-06, "loss": 0.5262, "step": 2294 }, { "epoch": 0.43, "learning_rate": 4.556422973020359e-06, "loss": 0.5573, "step": 2295 }, { "epoch": 0.43, "learning_rate": 4.555987905525605e-06, "loss": 0.6568, "step": 2296 }, { "epoch": 0.43, "learning_rate": 4.555552645567264e-06, "loss": 0.4926, "step": 2297 }, { "epoch": 0.43, "learning_rate": 4.55511719318608e-06, "loss": 0.6216, "step": 2298 }, { "epoch": 0.43, "learning_rate": 4.554681548422817e-06, "loss": 0.6051, "step": 2299 }, { "epoch": 0.43, "learning_rate": 4.5542457113182555e-06, "loss": 0.5427, "step": 2300 }, { "epoch": 0.43, "learning_rate": 4.553809681913195e-06, "loss": 0.624, "step": 2301 }, { "epoch": 0.43, "learning_rate": 4.553373460248453e-06, "loss": 0.3914, "step": 2302 }, { "epoch": 0.44, "learning_rate": 4.5529370463648645e-06, "loss": 0.6852, "step": 2303 }, { "epoch": 0.44, "learning_rate": 4.5525004403032835e-06, "loss": 0.4573, "step": 2304 }, { "epoch": 0.44, "learning_rate": 4.55206364210458e-06, "loss": 0.439, "step": 2305 }, { "epoch": 0.44, "learning_rate": 4.551626651809643e-06, "loss": 0.529, "step": 2306 }, { "epoch": 0.44, "learning_rate": 4.551189469459382e-06, "loss": 0.6027, "step": 2307 }, { "epoch": 0.44, "learning_rate": 4.55075209509472e-06, "loss": 0.6818, "step": 2308 }, { "epoch": 0.44, "learning_rate": 4.550314528756601e-06, "loss": 0.3425, "step": 2309 }, { "epoch": 0.44, "learning_rate": 4.549876770485986e-06, "loss": 0.5317, "step": 2310 }, { "epoch": 0.44, "learning_rate": 4.5494388203238535e-06, "loss": 0.4313, "step": 2311 }, { "epoch": 0.44, "learning_rate": 4.5490006783112026e-06, "loss": 0.4724, "step": 2312 }, { "epoch": 0.44, "learning_rate": 4.548562344489046e-06, "loss": 0.3223, "step": 2313 }, { "epoch": 0.44, "learning_rate": 4.548123818898417e-06, "loss": 0.5426, "step": 2314 }, { "epoch": 0.44, "learning_rate": 4.547685101580368e-06, "loss": 0.4604, "step": 2315 }, { "epoch": 0.44, "learning_rate": 4.5472461925759664e-06, "loss": 0.6831, "step": 2316 }, { "epoch": 0.44, "learning_rate": 4.5468070919263e-06, "loss": 0.5397, "step": 2317 }, { "epoch": 0.44, "learning_rate": 4.546367799672472e-06, "loss": 0.5591, "step": 2318 }, { "epoch": 0.44, "learning_rate": 4.545928315855608e-06, "loss": 0.4433, "step": 2319 }, { "epoch": 0.44, "learning_rate": 4.545488640516845e-06, "loss": 0.7183, "step": 2320 }, { "epoch": 0.44, "learning_rate": 4.545048773697345e-06, "loss": 0.3902, "step": 2321 }, { "epoch": 0.44, "learning_rate": 4.544608715438282e-06, "loss": 0.6032, "step": 2322 }, { "epoch": 0.44, "learning_rate": 4.544168465780852e-06, "loss": 0.5714, "step": 2323 }, { "epoch": 0.44, "learning_rate": 4.543728024766265e-06, "loss": 0.4326, "step": 2324 }, { "epoch": 0.44, "learning_rate": 4.543287392435755e-06, "loss": 0.4448, "step": 2325 }, { "epoch": 0.44, "learning_rate": 4.542846568830566e-06, "loss": 0.6292, "step": 2326 }, { "epoch": 0.44, "learning_rate": 4.542405553991968e-06, "loss": 0.4904, "step": 2327 }, { "epoch": 0.44, "learning_rate": 4.54196434796124e-06, "loss": 0.408, "step": 2328 }, { "epoch": 0.44, "learning_rate": 4.54152295077969e-06, "loss": 0.5989, "step": 2329 }, { "epoch": 0.44, "learning_rate": 4.541081362488633e-06, "loss": 0.5042, "step": 2330 }, { "epoch": 0.44, "learning_rate": 4.540639583129408e-06, "loss": 0.5277, "step": 2331 }, { "epoch": 0.44, "learning_rate": 4.540197612743371e-06, "loss": 0.5051, "step": 2332 }, { "epoch": 0.44, "learning_rate": 4.539755451371895e-06, "loss": 0.4712, "step": 2333 }, { "epoch": 0.44, "learning_rate": 4.539313099056372e-06, "loss": 0.5643, "step": 2334 }, { "epoch": 0.44, "learning_rate": 4.538870555838211e-06, "loss": 0.5678, "step": 2335 }, { "epoch": 0.44, "learning_rate": 4.538427821758837e-06, "loss": 0.6235, "step": 2336 }, { "epoch": 0.44, "learning_rate": 4.537984896859698e-06, "loss": 0.5041, "step": 2337 }, { "epoch": 0.44, "learning_rate": 4.537541781182253e-06, "loss": 0.6591, "step": 2338 }, { "epoch": 0.44, "learning_rate": 4.537098474767987e-06, "loss": 0.5563, "step": 2339 }, { "epoch": 0.44, "learning_rate": 4.536654977658394e-06, "loss": 0.4569, "step": 2340 }, { "epoch": 0.44, "learning_rate": 4.536211289894995e-06, "loss": 0.6238, "step": 2341 }, { "epoch": 0.44, "learning_rate": 4.5357674115193204e-06, "loss": 0.4625, "step": 2342 }, { "epoch": 0.44, "learning_rate": 4.5353233425729235e-06, "loss": 0.6154, "step": 2343 }, { "epoch": 0.44, "learning_rate": 4.534879083097375e-06, "loss": 0.5799, "step": 2344 }, { "epoch": 0.44, "learning_rate": 4.534434633134261e-06, "loss": 0.4261, "step": 2345 }, { "epoch": 0.44, "learning_rate": 4.533989992725188e-06, "loss": 0.5486, "step": 2346 }, { "epoch": 0.44, "learning_rate": 4.53354516191178e-06, "loss": 0.616, "step": 2347 }, { "epoch": 0.44, "learning_rate": 4.533100140735676e-06, "loss": 0.6045, "step": 2348 }, { "epoch": 0.44, "learning_rate": 4.532654929238537e-06, "loss": 0.6096, "step": 2349 }, { "epoch": 0.44, "learning_rate": 4.5322095274620395e-06, "loss": 0.413, "step": 2350 }, { "epoch": 0.44, "learning_rate": 4.531763935447877e-06, "loss": 0.579, "step": 2351 }, { "epoch": 0.44, "learning_rate": 4.531318153237762e-06, "loss": 0.5676, "step": 2352 }, { "epoch": 0.44, "learning_rate": 4.530872180873427e-06, "loss": 0.5186, "step": 2353 }, { "epoch": 0.44, "learning_rate": 4.530426018396617e-06, "loss": 0.4562, "step": 2354 }, { "epoch": 0.44, "learning_rate": 4.529979665849101e-06, "loss": 0.6513, "step": 2355 }, { "epoch": 0.45, "learning_rate": 4.529533123272658e-06, "loss": 0.6249, "step": 2356 }, { "epoch": 0.45, "learning_rate": 4.529086390709094e-06, "loss": 0.6517, "step": 2357 }, { "epoch": 0.45, "learning_rate": 4.5286394682002265e-06, "loss": 0.6908, "step": 2358 }, { "epoch": 0.45, "learning_rate": 4.528192355787892e-06, "loss": 0.5169, "step": 2359 }, { "epoch": 0.45, "learning_rate": 4.527745053513945e-06, "loss": 0.5168, "step": 2360 }, { "epoch": 0.45, "learning_rate": 4.52729756142026e-06, "loss": 0.4753, "step": 2361 }, { "epoch": 0.45, "learning_rate": 4.5268498795487255e-06, "loss": 0.4797, "step": 2362 }, { "epoch": 0.45, "learning_rate": 4.52640200794125e-06, "loss": 0.4225, "step": 2363 }, { "epoch": 0.45, "learning_rate": 4.52595394663976e-06, "loss": 0.5221, "step": 2364 }, { "epoch": 0.45, "learning_rate": 4.525505695686198e-06, "loss": 0.6147, "step": 2365 }, { "epoch": 0.45, "learning_rate": 4.525057255122525e-06, "loss": 0.5317, "step": 2366 }, { "epoch": 0.45, "learning_rate": 4.524608624990721e-06, "loss": 0.6771, "step": 2367 }, { "epoch": 0.45, "learning_rate": 4.524159805332783e-06, "loss": 0.5297, "step": 2368 }, { "epoch": 0.45, "learning_rate": 4.523710796190726e-06, "loss": 0.4039, "step": 2369 }, { "epoch": 0.45, "learning_rate": 4.52326159760658e-06, "loss": 0.5002, "step": 2370 }, { "epoch": 0.45, "learning_rate": 4.522812209622397e-06, "loss": 0.6444, "step": 2371 }, { "epoch": 0.45, "learning_rate": 4.522362632280245e-06, "loss": 0.4658, "step": 2372 }, { "epoch": 0.45, "learning_rate": 4.521912865622208e-06, "loss": 0.4833, "step": 2373 }, { "epoch": 0.45, "learning_rate": 4.521462909690391e-06, "loss": 0.6526, "step": 2374 }, { "epoch": 0.45, "learning_rate": 4.521012764526913e-06, "loss": 0.5221, "step": 2375 }, { "epoch": 0.45, "learning_rate": 4.520562430173914e-06, "loss": 0.5144, "step": 2376 }, { "epoch": 0.45, "learning_rate": 4.5201119066735495e-06, "loss": 0.623, "step": 2377 }, { "epoch": 0.45, "learning_rate": 4.519661194067996e-06, "loss": 0.4219, "step": 2378 }, { "epoch": 0.45, "learning_rate": 4.519210292399441e-06, "loss": 0.4207, "step": 2379 }, { "epoch": 0.45, "learning_rate": 4.518759201710097e-06, "loss": 0.511, "step": 2380 }, { "epoch": 0.45, "learning_rate": 4.518307922042191e-06, "loss": 0.6671, "step": 2381 }, { "epoch": 0.45, "learning_rate": 4.517856453437967e-06, "loss": 0.6912, "step": 2382 }, { "epoch": 0.45, "learning_rate": 4.517404795939688e-06, "loss": 0.5735, "step": 2383 }, { "epoch": 0.45, "learning_rate": 4.516952949589634e-06, "loss": 0.6969, "step": 2384 }, { "epoch": 0.45, "learning_rate": 4.5165009144301034e-06, "loss": 0.4703, "step": 2385 }, { "epoch": 0.45, "learning_rate": 4.51604869050341e-06, "loss": 0.4115, "step": 2386 }, { "epoch": 0.45, "learning_rate": 4.515596277851889e-06, "loss": 0.4739, "step": 2387 }, { "epoch": 0.45, "learning_rate": 4.515143676517891e-06, "loss": 0.4682, "step": 2388 }, { "epoch": 0.45, "learning_rate": 4.5146908865437845e-06, "loss": 0.6284, "step": 2389 }, { "epoch": 0.45, "learning_rate": 4.514237907971955e-06, "loss": 0.5018, "step": 2390 }, { "epoch": 0.45, "learning_rate": 4.513784740844807e-06, "loss": 0.5449, "step": 2391 }, { "epoch": 0.45, "learning_rate": 4.513331385204762e-06, "loss": 0.4449, "step": 2392 }, { "epoch": 0.45, "learning_rate": 4.512877841094259e-06, "loss": 0.5748, "step": 2393 }, { "epoch": 0.45, "learning_rate": 4.512424108555754e-06, "loss": 0.449, "step": 2394 }, { "epoch": 0.45, "learning_rate": 4.5119701876317245e-06, "loss": 0.5681, "step": 2395 }, { "epoch": 0.45, "learning_rate": 4.511516078364659e-06, "loss": 0.56, "step": 2396 }, { "epoch": 0.45, "learning_rate": 4.511061780797069e-06, "loss": 0.3915, "step": 2397 }, { "epoch": 0.45, "learning_rate": 4.510607294971482e-06, "loss": 0.4658, "step": 2398 }, { "epoch": 0.45, "learning_rate": 4.510152620930442e-06, "loss": 0.492, "step": 2399 }, { "epoch": 0.45, "learning_rate": 4.509697758716513e-06, "loss": 0.504, "step": 2400 }, { "epoch": 0.45, "learning_rate": 4.509242708372274e-06, "loss": 0.5459, "step": 2401 }, { "epoch": 0.45, "learning_rate": 4.5087874699403236e-06, "loss": 0.5842, "step": 2402 }, { "epoch": 0.45, "learning_rate": 4.508332043463276e-06, "loss": 0.4567, "step": 2403 }, { "epoch": 0.45, "learning_rate": 4.507876428983765e-06, "loss": 0.4488, "step": 2404 }, { "epoch": 0.45, "learning_rate": 4.507420626544442e-06, "loss": 0.4606, "step": 2405 }, { "epoch": 0.45, "learning_rate": 4.506964636187975e-06, "loss": 0.5167, "step": 2406 }, { "epoch": 0.45, "learning_rate": 4.506508457957049e-06, "loss": 0.6394, "step": 2407 }, { "epoch": 0.45, "learning_rate": 4.506052091894368e-06, "loss": 0.5589, "step": 2408 }, { "epoch": 0.46, "learning_rate": 4.505595538042652e-06, "loss": 0.3617, "step": 2409 }, { "epoch": 0.46, "learning_rate": 4.5051387964446415e-06, "loss": 0.4315, "step": 2410 }, { "epoch": 0.46, "learning_rate": 4.50468186714309e-06, "loss": 0.4797, "step": 2411 }, { "epoch": 0.46, "learning_rate": 4.504224750180775e-06, "loss": 0.4687, "step": 2412 }, { "epoch": 0.46, "learning_rate": 4.503767445600484e-06, "loss": 0.6913, "step": 2413 }, { "epoch": 0.46, "learning_rate": 4.5033099534450274e-06, "loss": 0.4925, "step": 2414 }, { "epoch": 0.46, "learning_rate": 4.502852273757232e-06, "loss": 0.4966, "step": 2415 }, { "epoch": 0.46, "learning_rate": 4.502394406579942e-06, "loss": 0.5138, "step": 2416 }, { "epoch": 0.46, "learning_rate": 4.501936351956018e-06, "loss": 0.5319, "step": 2417 }, { "epoch": 0.46, "learning_rate": 4.501478109928338e-06, "loss": 0.6091, "step": 2418 }, { "epoch": 0.46, "learning_rate": 4.501019680539801e-06, "loss": 0.4976, "step": 2419 }, { "epoch": 0.46, "learning_rate": 4.50056106383332e-06, "loss": 0.662, "step": 2420 }, { "epoch": 0.46, "learning_rate": 4.500102259851827e-06, "loss": 0.5039, "step": 2421 }, { "epoch": 0.46, "learning_rate": 4.4996432686382714e-06, "loss": 0.4965, "step": 2422 }, { "epoch": 0.46, "learning_rate": 4.499184090235619e-06, "loss": 0.5948, "step": 2423 }, { "epoch": 0.46, "learning_rate": 4.498724724686855e-06, "loss": 0.5045, "step": 2424 }, { "epoch": 0.46, "learning_rate": 4.49826517203498e-06, "loss": 0.4679, "step": 2425 }, { "epoch": 0.46, "learning_rate": 4.497805432323015e-06, "loss": 0.5605, "step": 2426 }, { "epoch": 0.46, "learning_rate": 4.497345505593995e-06, "loss": 0.5349, "step": 2427 }, { "epoch": 0.46, "learning_rate": 4.496885391890976e-06, "loss": 0.6491, "step": 2428 }, { "epoch": 0.46, "learning_rate": 4.496425091257028e-06, "loss": 0.4602, "step": 2429 }, { "epoch": 0.46, "learning_rate": 4.495964603735243e-06, "loss": 0.5641, "step": 2430 }, { "epoch": 0.46, "learning_rate": 4.495503929368724e-06, "loss": 0.5845, "step": 2431 }, { "epoch": 0.46, "learning_rate": 4.4950430682005995e-06, "loss": 0.5387, "step": 2432 }, { "epoch": 0.46, "learning_rate": 4.494582020274007e-06, "loss": 0.5507, "step": 2433 }, { "epoch": 0.46, "learning_rate": 4.49412078563211e-06, "loss": 0.5108, "step": 2434 }, { "epoch": 0.46, "learning_rate": 4.493659364318083e-06, "loss": 0.5427, "step": 2435 }, { "epoch": 0.46, "learning_rate": 4.49319775637512e-06, "loss": 0.5412, "step": 2436 }, { "epoch": 0.46, "learning_rate": 4.492735961846433e-06, "loss": 0.5798, "step": 2437 }, { "epoch": 0.46, "learning_rate": 4.492273980775252e-06, "loss": 0.5965, "step": 2438 }, { "epoch": 0.46, "learning_rate": 4.491811813204823e-06, "loss": 0.5028, "step": 2439 }, { "epoch": 0.46, "learning_rate": 4.491349459178409e-06, "loss": 0.6181, "step": 2440 }, { "epoch": 0.46, "learning_rate": 4.490886918739294e-06, "loss": 0.4355, "step": 2441 }, { "epoch": 0.46, "learning_rate": 4.4904241919307744e-06, "loss": 0.5176, "step": 2442 }, { "epoch": 0.46, "learning_rate": 4.4899612787961675e-06, "loss": 0.6685, "step": 2443 }, { "epoch": 0.46, "learning_rate": 4.489498179378809e-06, "loss": 0.4496, "step": 2444 }, { "epoch": 0.46, "learning_rate": 4.489034893722047e-06, "loss": 0.5345, "step": 2445 }, { "epoch": 0.46, "learning_rate": 4.488571421869253e-06, "loss": 0.4839, "step": 2446 }, { "epoch": 0.46, "learning_rate": 4.488107763863812e-06, "loss": 0.5777, "step": 2447 }, { "epoch": 0.46, "learning_rate": 4.487643919749128e-06, "loss": 0.4664, "step": 2448 }, { "epoch": 0.46, "learning_rate": 4.487179889568621e-06, "loss": 0.5604, "step": 2449 }, { "epoch": 0.46, "learning_rate": 4.48671567336573e-06, "loss": 0.477, "step": 2450 }, { "epoch": 0.46, "learning_rate": 4.4862512711839125e-06, "loss": 0.5318, "step": 2451 }, { "epoch": 0.46, "learning_rate": 4.485786683066638e-06, "loss": 0.5381, "step": 2452 }, { "epoch": 0.46, "learning_rate": 4.4853219090574015e-06, "loss": 0.4086, "step": 2453 }, { "epoch": 0.46, "learning_rate": 4.484856949199709e-06, "loss": 0.4538, "step": 2454 }, { "epoch": 0.46, "learning_rate": 4.484391803537085e-06, "loss": 0.5735, "step": 2455 }, { "epoch": 0.46, "learning_rate": 4.483926472113074e-06, "loss": 0.5864, "step": 2456 }, { "epoch": 0.46, "learning_rate": 4.483460954971235e-06, "loss": 0.471, "step": 2457 }, { "epoch": 0.46, "learning_rate": 4.482995252155147e-06, "loss": 0.6286, "step": 2458 }, { "epoch": 0.46, "learning_rate": 4.482529363708403e-06, "loss": 0.4128, "step": 2459 }, { "epoch": 0.46, "learning_rate": 4.4820632896746185e-06, "loss": 0.522, "step": 2460 }, { "epoch": 0.46, "learning_rate": 4.481597030097421e-06, "loss": 0.4031, "step": 2461 }, { "epoch": 0.47, "learning_rate": 4.481130585020457e-06, "loss": 0.4363, "step": 2462 }, { "epoch": 0.47, "learning_rate": 4.480663954487393e-06, "loss": 0.5983, "step": 2463 }, { "epoch": 0.47, "learning_rate": 4.48019713854191e-06, "loss": 0.5931, "step": 2464 }, { "epoch": 0.47, "learning_rate": 4.479730137227707e-06, "loss": 0.4847, "step": 2465 }, { "epoch": 0.47, "learning_rate": 4.479262950588501e-06, "loss": 0.5704, "step": 2466 }, { "epoch": 0.47, "learning_rate": 4.478795578668026e-06, "loss": 0.5095, "step": 2467 }, { "epoch": 0.47, "learning_rate": 4.478328021510032e-06, "loss": 0.473, "step": 2468 }, { "epoch": 0.47, "learning_rate": 4.47786027915829e-06, "loss": 0.5162, "step": 2469 }, { "epoch": 0.47, "learning_rate": 4.477392351656583e-06, "loss": 0.4789, "step": 2470 }, { "epoch": 0.47, "learning_rate": 4.476924239048716e-06, "loss": 0.5566, "step": 2471 }, { "epoch": 0.47, "learning_rate": 4.47645594137851e-06, "loss": 0.7456, "step": 2472 }, { "epoch": 0.47, "learning_rate": 4.4759874586898025e-06, "loss": 0.6735, "step": 2473 }, { "epoch": 0.47, "learning_rate": 4.475518791026448e-06, "loss": 0.499, "step": 2474 }, { "epoch": 0.47, "learning_rate": 4.475049938432319e-06, "loss": 0.5205, "step": 2475 }, { "epoch": 0.47, "learning_rate": 4.474580900951307e-06, "loss": 0.5083, "step": 2476 }, { "epoch": 0.47, "learning_rate": 4.474111678627318e-06, "loss": 0.5634, "step": 2477 }, { "epoch": 0.47, "learning_rate": 4.473642271504277e-06, "loss": 0.5334, "step": 2478 }, { "epoch": 0.47, "learning_rate": 4.4731726796261245e-06, "loss": 0.5677, "step": 2479 }, { "epoch": 0.47, "learning_rate": 4.472702903036822e-06, "loss": 0.651, "step": 2480 }, { "epoch": 0.47, "learning_rate": 4.472232941780344e-06, "loss": 0.5047, "step": 2481 }, { "epoch": 0.47, "learning_rate": 4.471762795900684e-06, "loss": 0.411, "step": 2482 }, { "epoch": 0.47, "learning_rate": 4.471292465441853e-06, "loss": 0.5585, "step": 2483 }, { "epoch": 0.47, "learning_rate": 4.47082195044788e-06, "loss": 0.5665, "step": 2484 }, { "epoch": 0.47, "learning_rate": 4.470351250962811e-06, "loss": 0.6086, "step": 2485 }, { "epoch": 0.47, "learning_rate": 4.469880367030708e-06, "loss": 0.4551, "step": 2486 }, { "epoch": 0.47, "learning_rate": 4.469409298695651e-06, "loss": 0.3784, "step": 2487 }, { "epoch": 0.47, "learning_rate": 4.468938046001736e-06, "loss": 0.5258, "step": 2488 }, { "epoch": 0.47, "learning_rate": 4.4684666089930806e-06, "loss": 0.5607, "step": 2489 }, { "epoch": 0.47, "learning_rate": 4.467994987713814e-06, "loss": 0.5708, "step": 2490 }, { "epoch": 0.47, "learning_rate": 4.467523182208087e-06, "loss": 0.5446, "step": 2491 }, { "epoch": 0.47, "learning_rate": 4.467051192520063e-06, "loss": 0.5661, "step": 2492 }, { "epoch": 0.47, "learning_rate": 4.46657901869393e-06, "loss": 0.427, "step": 2493 }, { "epoch": 0.47, "learning_rate": 4.466106660773884e-06, "loss": 0.5623, "step": 2494 }, { "epoch": 0.47, "learning_rate": 4.465634118804147e-06, "loss": 0.373, "step": 2495 }, { "epoch": 0.47, "learning_rate": 4.465161392828953e-06, "loss": 0.5751, "step": 2496 }, { "epoch": 0.47, "learning_rate": 4.464688482892553e-06, "loss": 0.565, "step": 2497 }, { "epoch": 0.47, "learning_rate": 4.464215389039218e-06, "loss": 0.472, "step": 2498 }, { "epoch": 0.47, "learning_rate": 4.4637421113132354e-06, "loss": 0.56, "step": 2499 }, { "epoch": 0.47, "learning_rate": 4.463268649758907e-06, "loss": 0.5023, "step": 2500 }, { "epoch": 0.47, "learning_rate": 4.462795004420557e-06, "loss": 0.6767, "step": 2501 }, { "epoch": 0.47, "learning_rate": 4.462321175342522e-06, "loss": 0.4787, "step": 2502 }, { "epoch": 0.47, "learning_rate": 4.461847162569159e-06, "loss": 0.5166, "step": 2503 }, { "epoch": 0.47, "learning_rate": 4.46137296614484e-06, "loss": 0.3998, "step": 2504 }, { "epoch": 0.47, "learning_rate": 4.460898586113955e-06, "loss": 0.4977, "step": 2505 }, { "epoch": 0.47, "learning_rate": 4.460424022520912e-06, "loss": 0.5135, "step": 2506 }, { "epoch": 0.47, "learning_rate": 4.459949275410136e-06, "loss": 0.5393, "step": 2507 }, { "epoch": 0.47, "learning_rate": 4.459474344826067e-06, "loss": 0.6147, "step": 2508 }, { "epoch": 0.47, "learning_rate": 4.458999230813164e-06, "loss": 0.5378, "step": 2509 }, { "epoch": 0.47, "learning_rate": 4.458523933415906e-06, "loss": 0.6042, "step": 2510 }, { "epoch": 0.47, "learning_rate": 4.458048452678781e-06, "loss": 0.5194, "step": 2511 }, { "epoch": 0.47, "learning_rate": 4.457572788646305e-06, "loss": 0.5524, "step": 2512 }, { "epoch": 0.47, "learning_rate": 4.457096941363002e-06, "loss": 0.5346, "step": 2513 }, { "epoch": 0.47, "learning_rate": 4.456620910873416e-06, "loss": 0.5664, "step": 2514 }, { "epoch": 0.48, "learning_rate": 4.456144697222112e-06, "loss": 0.5171, "step": 2515 }, { "epoch": 0.48, "learning_rate": 4.455668300453666e-06, "loss": 0.5029, "step": 2516 }, { "epoch": 0.48, "learning_rate": 4.455191720612676e-06, "loss": 0.4435, "step": 2517 }, { "epoch": 0.48, "learning_rate": 4.454714957743754e-06, "loss": 0.447, "step": 2518 }, { "epoch": 0.48, "learning_rate": 4.4542380118915305e-06, "loss": 0.5018, "step": 2519 }, { "epoch": 0.48, "learning_rate": 4.453760883100654e-06, "loss": 0.4546, "step": 2520 }, { "epoch": 0.48, "learning_rate": 4.453283571415788e-06, "loss": 0.5641, "step": 2521 }, { "epoch": 0.48, "learning_rate": 4.452806076881616e-06, "loss": 0.4296, "step": 2522 }, { "epoch": 0.48, "learning_rate": 4.4523283995428345e-06, "loss": 0.5071, "step": 2523 }, { "epoch": 0.48, "learning_rate": 4.451850539444161e-06, "loss": 0.4585, "step": 2524 }, { "epoch": 0.48, "learning_rate": 4.4513724966303285e-06, "loss": 0.5697, "step": 2525 }, { "epoch": 0.48, "learning_rate": 4.450894271146085e-06, "loss": 0.4988, "step": 2526 }, { "epoch": 0.48, "learning_rate": 4.450415863036202e-06, "loss": 0.5361, "step": 2527 }, { "epoch": 0.48, "learning_rate": 4.449937272345461e-06, "loss": 0.4026, "step": 2528 }, { "epoch": 0.48, "learning_rate": 4.449458499118662e-06, "loss": 0.5737, "step": 2529 }, { "epoch": 0.48, "learning_rate": 4.448979543400627e-06, "loss": 0.577, "step": 2530 }, { "epoch": 0.48, "learning_rate": 4.44850040523619e-06, "loss": 0.4957, "step": 2531 }, { "epoch": 0.48, "learning_rate": 4.448021084670204e-06, "loss": 0.5578, "step": 2532 }, { "epoch": 0.48, "learning_rate": 4.447541581747539e-06, "loss": 0.4763, "step": 2533 }, { "epoch": 0.48, "learning_rate": 4.447061896513081e-06, "loss": 0.5743, "step": 2534 }, { "epoch": 0.48, "learning_rate": 4.446582029011734e-06, "loss": 0.6029, "step": 2535 }, { "epoch": 0.48, "learning_rate": 4.44610197928842e-06, "loss": 0.4151, "step": 2536 }, { "epoch": 0.48, "learning_rate": 4.445621747388076e-06, "loss": 0.4901, "step": 2537 }, { "epoch": 0.48, "learning_rate": 4.445141333355659e-06, "loss": 0.5057, "step": 2538 }, { "epoch": 0.48, "learning_rate": 4.4446607372361386e-06, "loss": 0.4102, "step": 2539 }, { "epoch": 0.48, "learning_rate": 4.444179959074505e-06, "loss": 0.6547, "step": 2540 }, { "epoch": 0.48, "learning_rate": 4.443698998915765e-06, "loss": 0.5729, "step": 2541 }, { "epoch": 0.48, "learning_rate": 4.443217856804942e-06, "loss": 0.5318, "step": 2542 }, { "epoch": 0.48, "learning_rate": 4.4427365327870755e-06, "loss": 0.5957, "step": 2543 }, { "epoch": 0.48, "learning_rate": 4.442255026907223e-06, "loss": 0.558, "step": 2544 }, { "epoch": 0.48, "learning_rate": 4.441773339210459e-06, "loss": 0.4955, "step": 2545 }, { "epoch": 0.48, "learning_rate": 4.441291469741874e-06, "loss": 0.5039, "step": 2546 }, { "epoch": 0.48, "learning_rate": 4.4408094185465785e-06, "loss": 0.5665, "step": 2547 }, { "epoch": 0.48, "learning_rate": 4.440327185669697e-06, "loss": 0.5346, "step": 2548 }, { "epoch": 0.48, "learning_rate": 4.439844771156371e-06, "loss": 0.4286, "step": 2549 }, { "epoch": 0.48, "learning_rate": 4.43936217505176e-06, "loss": 0.7091, "step": 2550 }, { "epoch": 0.48, "learning_rate": 4.438879397401043e-06, "loss": 0.4814, "step": 2551 }, { "epoch": 0.48, "learning_rate": 4.438396438249409e-06, "loss": 0.4347, "step": 2552 }, { "epoch": 0.48, "learning_rate": 4.437913297642071e-06, "loss": 0.5844, "step": 2553 }, { "epoch": 0.48, "learning_rate": 4.437429975624257e-06, "loss": 0.5786, "step": 2554 }, { "epoch": 0.48, "learning_rate": 4.436946472241211e-06, "loss": 0.4034, "step": 2555 }, { "epoch": 0.48, "learning_rate": 4.436462787538194e-06, "loss": 0.3838, "step": 2556 }, { "epoch": 0.48, "learning_rate": 4.435978921560484e-06, "loss": 0.4462, "step": 2557 }, { "epoch": 0.48, "learning_rate": 4.435494874353376e-06, "loss": 0.5559, "step": 2558 }, { "epoch": 0.48, "learning_rate": 4.435010645962183e-06, "loss": 0.6263, "step": 2559 }, { "epoch": 0.48, "learning_rate": 4.434526236432234e-06, "loss": 0.5656, "step": 2560 }, { "epoch": 0.48, "learning_rate": 4.434041645808875e-06, "loss": 0.4013, "step": 2561 }, { "epoch": 0.48, "learning_rate": 4.43355687413747e-06, "loss": 0.5674, "step": 2562 }, { "epoch": 0.48, "learning_rate": 4.433071921463398e-06, "loss": 0.6397, "step": 2563 }, { "epoch": 0.48, "learning_rate": 4.4325867878320576e-06, "loss": 0.5526, "step": 2564 }, { "epoch": 0.48, "learning_rate": 4.43210147328886e-06, "loss": 0.5462, "step": 2565 }, { "epoch": 0.48, "learning_rate": 4.431615977879238e-06, "loss": 0.4345, "step": 2566 }, { "epoch": 0.48, "learning_rate": 4.4311303016486395e-06, "loss": 0.7997, "step": 2567 }, { "epoch": 0.49, "learning_rate": 4.430644444642529e-06, "loss": 0.4738, "step": 2568 }, { "epoch": 0.49, "learning_rate": 4.4301584069063876e-06, "loss": 0.5628, "step": 2569 }, { "epoch": 0.49, "learning_rate": 4.429672188485715e-06, "loss": 0.4941, "step": 2570 }, { "epoch": 0.49, "learning_rate": 4.429185789426026e-06, "loss": 0.5649, "step": 2571 }, { "epoch": 0.49, "learning_rate": 4.428699209772852e-06, "loss": 0.4946, "step": 2572 }, { "epoch": 0.49, "learning_rate": 4.428212449571745e-06, "loss": 0.5207, "step": 2573 }, { "epoch": 0.49, "learning_rate": 4.427725508868269e-06, "loss": 0.557, "step": 2574 }, { "epoch": 0.49, "learning_rate": 4.427238387708007e-06, "loss": 0.4113, "step": 2575 }, { "epoch": 0.49, "learning_rate": 4.4267510861365615e-06, "loss": 0.4955, "step": 2576 }, { "epoch": 0.49, "learning_rate": 4.4262636041995465e-06, "loss": 0.539, "step": 2577 }, { "epoch": 0.49, "learning_rate": 4.425775941942597e-06, "loss": 0.4641, "step": 2578 }, { "epoch": 0.49, "learning_rate": 4.425288099411365e-06, "loss": 0.6275, "step": 2579 }, { "epoch": 0.49, "learning_rate": 4.424800076651515e-06, "loss": 0.4794, "step": 2580 }, { "epoch": 0.49, "learning_rate": 4.424311873708733e-06, "loss": 0.6551, "step": 2581 }, { "epoch": 0.49, "learning_rate": 4.423823490628721e-06, "loss": 0.6468, "step": 2582 }, { "epoch": 0.49, "learning_rate": 4.423334927457198e-06, "loss": 0.5269, "step": 2583 }, { "epoch": 0.49, "learning_rate": 4.422846184239896e-06, "loss": 0.4254, "step": 2584 }, { "epoch": 0.49, "learning_rate": 4.422357261022569e-06, "loss": 0.5802, "step": 2585 }, { "epoch": 0.49, "learning_rate": 4.421868157850985e-06, "loss": 0.4322, "step": 2586 }, { "epoch": 0.49, "learning_rate": 4.4213788747709295e-06, "loss": 0.5565, "step": 2587 }, { "epoch": 0.49, "learning_rate": 4.420889411828205e-06, "loss": 0.5815, "step": 2588 }, { "epoch": 0.49, "learning_rate": 4.420399769068632e-06, "loss": 0.3809, "step": 2589 }, { "epoch": 0.49, "learning_rate": 4.419909946538045e-06, "loss": 0.4512, "step": 2590 }, { "epoch": 0.49, "learning_rate": 4.419419944282296e-06, "loss": 0.6586, "step": 2591 }, { "epoch": 0.49, "learning_rate": 4.418929762347257e-06, "loss": 0.5206, "step": 2592 }, { "epoch": 0.49, "learning_rate": 4.4184394007788134e-06, "loss": 0.4887, "step": 2593 }, { "epoch": 0.49, "learning_rate": 4.417948859622869e-06, "loss": 0.3962, "step": 2594 }, { "epoch": 0.49, "learning_rate": 4.417458138925344e-06, "loss": 0.6667, "step": 2595 }, { "epoch": 0.49, "learning_rate": 4.416967238732174e-06, "loss": 0.4684, "step": 2596 }, { "epoch": 0.49, "learning_rate": 4.416476159089315e-06, "loss": 0.5194, "step": 2597 }, { "epoch": 0.49, "learning_rate": 4.415984900042735e-06, "loss": 0.5781, "step": 2598 }, { "epoch": 0.49, "learning_rate": 4.415493461638423e-06, "loss": 0.4837, "step": 2599 }, { "epoch": 0.49, "learning_rate": 4.415001843922384e-06, "loss": 0.6787, "step": 2600 }, { "epoch": 0.49, "learning_rate": 4.414510046940639e-06, "loss": 0.4255, "step": 2601 }, { "epoch": 0.49, "learning_rate": 4.414018070739222e-06, "loss": 0.6962, "step": 2602 }, { "epoch": 0.49, "learning_rate": 4.413525915364192e-06, "loss": 0.5701, "step": 2603 }, { "epoch": 0.49, "learning_rate": 4.413033580861617e-06, "loss": 0.5938, "step": 2604 }, { "epoch": 0.49, "learning_rate": 4.412541067277588e-06, "loss": 0.5204, "step": 2605 }, { "epoch": 0.49, "learning_rate": 4.412048374658207e-06, "loss": 0.6701, "step": 2606 }, { "epoch": 0.49, "learning_rate": 4.4115555030495975e-06, "loss": 0.4834, "step": 2607 }, { "epoch": 0.49, "learning_rate": 4.4110624524978976e-06, "loss": 0.4606, "step": 2608 }, { "epoch": 0.49, "learning_rate": 4.4105692230492605e-06, "loss": 0.58, "step": 2609 }, { "epoch": 0.49, "learning_rate": 4.4100758147498606e-06, "loss": 0.4431, "step": 2610 }, { "epoch": 0.49, "learning_rate": 4.409582227645885e-06, "loss": 0.5144, "step": 2611 }, { "epoch": 0.49, "learning_rate": 4.40908846178354e-06, "loss": 0.6422, "step": 2612 }, { "epoch": 0.49, "learning_rate": 4.408594517209045e-06, "loss": 0.6804, "step": 2613 }, { "epoch": 0.49, "learning_rate": 4.4081003939686426e-06, "loss": 0.4247, "step": 2614 }, { "epoch": 0.49, "learning_rate": 4.407606092108586e-06, "loss": 0.5046, "step": 2615 }, { "epoch": 0.49, "learning_rate": 4.4071116116751475e-06, "loss": 0.4255, "step": 2616 }, { "epoch": 0.49, "learning_rate": 4.406616952714617e-06, "loss": 0.5013, "step": 2617 }, { "epoch": 0.49, "learning_rate": 4.406122115273299e-06, "loss": 0.4938, "step": 2618 }, { "epoch": 0.49, "learning_rate": 4.405627099397517e-06, "loss": 0.5164, "step": 2619 }, { "epoch": 0.49, "learning_rate": 4.405131905133608e-06, "loss": 0.4311, "step": 2620 }, { "epoch": 0.5, "learning_rate": 4.40463653252793e-06, "loss": 0.4811, "step": 2621 }, { "epoch": 0.5, "learning_rate": 4.404140981626855e-06, "loss": 0.5325, "step": 2622 }, { "epoch": 0.5, "learning_rate": 4.403645252476771e-06, "loss": 0.4525, "step": 2623 }, { "epoch": 0.5, "learning_rate": 4.403149345124086e-06, "loss": 0.5339, "step": 2624 }, { "epoch": 0.5, "learning_rate": 4.40265325961522e-06, "loss": 0.4578, "step": 2625 }, { "epoch": 0.5, "learning_rate": 4.402156995996613e-06, "loss": 0.4927, "step": 2626 }, { "epoch": 0.5, "learning_rate": 4.4016605543147214e-06, "loss": 0.4796, "step": 2627 }, { "epoch": 0.5, "learning_rate": 4.4011639346160176e-06, "loss": 0.4233, "step": 2628 }, { "epoch": 0.5, "learning_rate": 4.400667136946991e-06, "loss": 0.4372, "step": 2629 }, { "epoch": 0.5, "learning_rate": 4.400170161354146e-06, "loss": 0.5625, "step": 2630 }, { "epoch": 0.5, "learning_rate": 4.399673007884006e-06, "loss": 0.3926, "step": 2631 }, { "epoch": 0.5, "learning_rate": 4.3991756765831115e-06, "loss": 0.3697, "step": 2632 }, { "epoch": 0.5, "learning_rate": 4.398678167498017e-06, "loss": 0.4275, "step": 2633 }, { "epoch": 0.5, "learning_rate": 4.398180480675295e-06, "loss": 0.458, "step": 2634 }, { "epoch": 0.5, "learning_rate": 4.397682616161535e-06, "loss": 0.3928, "step": 2635 }, { "epoch": 0.5, "learning_rate": 4.397184574003342e-06, "loss": 0.4827, "step": 2636 }, { "epoch": 0.5, "learning_rate": 4.396686354247338e-06, "loss": 0.508, "step": 2637 }, { "epoch": 0.5, "learning_rate": 4.396187956940164e-06, "loss": 0.4113, "step": 2638 }, { "epoch": 0.5, "learning_rate": 4.395689382128473e-06, "loss": 0.6416, "step": 2639 }, { "epoch": 0.5, "learning_rate": 4.395190629858939e-06, "loss": 0.438, "step": 2640 }, { "epoch": 0.5, "learning_rate": 4.394691700178251e-06, "loss": 0.5096, "step": 2641 }, { "epoch": 0.5, "learning_rate": 4.394192593133113e-06, "loss": 0.502, "step": 2642 }, { "epoch": 0.5, "learning_rate": 4.393693308770248e-06, "loss": 0.4315, "step": 2643 }, { "epoch": 0.5, "learning_rate": 4.393193847136394e-06, "loss": 0.5933, "step": 2644 }, { "epoch": 0.5, "learning_rate": 4.3926942082783076e-06, "loss": 0.5432, "step": 2645 }, { "epoch": 0.5, "learning_rate": 4.392194392242759e-06, "loss": 0.5645, "step": 2646 }, { "epoch": 0.5, "learning_rate": 4.391694399076537e-06, "loss": 0.4495, "step": 2647 }, { "epoch": 0.5, "learning_rate": 4.391194228826448e-06, "loss": 0.6732, "step": 2648 }, { "epoch": 0.5, "learning_rate": 4.39069388153931e-06, "loss": 0.561, "step": 2649 }, { "epoch": 0.5, "learning_rate": 4.3901933572619655e-06, "loss": 0.5891, "step": 2650 }, { "epoch": 0.5, "learning_rate": 4.3896926560412676e-06, "loss": 0.5002, "step": 2651 }, { "epoch": 0.5, "learning_rate": 4.389191777924087e-06, "loss": 0.5133, "step": 2652 }, { "epoch": 0.5, "learning_rate": 4.388690722957311e-06, "loss": 0.466, "step": 2653 }, { "epoch": 0.5, "learning_rate": 4.388189491187844e-06, "loss": 0.4775, "step": 2654 }, { "epoch": 0.5, "learning_rate": 4.387688082662609e-06, "loss": 0.4526, "step": 2655 }, { "epoch": 0.5, "learning_rate": 4.387186497428542e-06, "loss": 0.41, "step": 2656 }, { "epoch": 0.5, "learning_rate": 4.386684735532597e-06, "loss": 0.4873, "step": 2657 }, { "epoch": 0.5, "learning_rate": 4.386182797021744e-06, "loss": 0.4359, "step": 2658 }, { "epoch": 0.5, "learning_rate": 4.385680681942971e-06, "loss": 0.4359, "step": 2659 }, { "epoch": 0.5, "learning_rate": 4.385178390343282e-06, "loss": 0.3555, "step": 2660 }, { "epoch": 0.5, "learning_rate": 4.3846759222696955e-06, "loss": 0.5118, "step": 2661 }, { "epoch": 0.5, "learning_rate": 4.384173277769251e-06, "loss": 0.4907, "step": 2662 }, { "epoch": 0.5, "learning_rate": 4.383670456888998e-06, "loss": 0.4801, "step": 2663 }, { "epoch": 0.5, "learning_rate": 4.383167459676009e-06, "loss": 0.6154, "step": 2664 }, { "epoch": 0.5, "learning_rate": 4.3826642861773685e-06, "loss": 0.6966, "step": 2665 }, { "epoch": 0.5, "learning_rate": 4.382160936440181e-06, "loss": 0.581, "step": 2666 }, { "epoch": 0.5, "learning_rate": 4.381657410511564e-06, "loss": 0.4345, "step": 2667 }, { "epoch": 0.5, "learning_rate": 4.381153708438654e-06, "loss": 0.5881, "step": 2668 }, { "epoch": 0.5, "learning_rate": 4.380649830268603e-06, "loss": 0.571, "step": 2669 }, { "epoch": 0.5, "learning_rate": 4.38014577604858e-06, "loss": 0.3853, "step": 2670 }, { "epoch": 0.5, "learning_rate": 4.379641545825769e-06, "loss": 0.4832, "step": 2671 }, { "epoch": 0.5, "learning_rate": 4.379137139647372e-06, "loss": 0.5416, "step": 2672 }, { "epoch": 0.51, "learning_rate": 4.378632557560609e-06, "loss": 0.5042, "step": 2673 }, { "epoch": 0.51, "learning_rate": 4.3781277996127125e-06, "loss": 0.4858, "step": 2674 }, { "epoch": 0.51, "learning_rate": 4.377622865850935e-06, "loss": 0.5586, "step": 2675 }, { "epoch": 0.51, "learning_rate": 4.377117756322542e-06, "loss": 0.4431, "step": 2676 }, { "epoch": 0.51, "learning_rate": 4.376612471074818e-06, "loss": 0.45, "step": 2677 }, { "epoch": 0.51, "learning_rate": 4.376107010155065e-06, "loss": 0.4571, "step": 2678 }, { "epoch": 0.51, "learning_rate": 4.375601373610599e-06, "loss": 0.5958, "step": 2679 }, { "epoch": 0.51, "learning_rate": 4.375095561488752e-06, "loss": 0.5352, "step": 2680 }, { "epoch": 0.51, "learning_rate": 4.374589573836875e-06, "loss": 0.4742, "step": 2681 }, { "epoch": 0.51, "learning_rate": 4.374083410702334e-06, "loss": 0.547, "step": 2682 }, { "epoch": 0.51, "learning_rate": 4.373577072132511e-06, "loss": 0.4564, "step": 2683 }, { "epoch": 0.51, "learning_rate": 4.373070558174805e-06, "loss": 0.4797, "step": 2684 }, { "epoch": 0.51, "learning_rate": 4.3725638688766334e-06, "loss": 0.3982, "step": 2685 }, { "epoch": 0.51, "learning_rate": 4.372057004285425e-06, "loss": 0.5962, "step": 2686 }, { "epoch": 0.51, "learning_rate": 4.37154996444863e-06, "loss": 0.4768, "step": 2687 }, { "epoch": 0.51, "learning_rate": 4.371042749413712e-06, "loss": 0.5567, "step": 2688 }, { "epoch": 0.51, "learning_rate": 4.370535359228153e-06, "loss": 0.5661, "step": 2689 }, { "epoch": 0.51, "learning_rate": 4.37002779393945e-06, "loss": 0.4932, "step": 2690 }, { "epoch": 0.51, "learning_rate": 4.369520053595117e-06, "loss": 0.5621, "step": 2691 }, { "epoch": 0.51, "learning_rate": 4.369012138242682e-06, "loss": 0.4878, "step": 2692 }, { "epoch": 0.51, "learning_rate": 4.3685040479296955e-06, "loss": 0.5001, "step": 2693 }, { "epoch": 0.51, "learning_rate": 4.367995782703718e-06, "loss": 0.453, "step": 2694 }, { "epoch": 0.51, "learning_rate": 4.367487342612329e-06, "loss": 0.5583, "step": 2695 }, { "epoch": 0.51, "learning_rate": 4.366978727703125e-06, "loss": 0.5079, "step": 2696 }, { "epoch": 0.51, "learning_rate": 4.366469938023717e-06, "loss": 0.4001, "step": 2697 }, { "epoch": 0.51, "learning_rate": 4.365960973621735e-06, "loss": 0.4277, "step": 2698 }, { "epoch": 0.51, "learning_rate": 4.365451834544821e-06, "loss": 0.5204, "step": 2699 }, { "epoch": 0.51, "learning_rate": 4.36494252084064e-06, "loss": 0.5326, "step": 2700 }, { "epoch": 0.51, "learning_rate": 4.364433032556867e-06, "loss": 0.3939, "step": 2701 }, { "epoch": 0.51, "learning_rate": 4.3639233697411956e-06, "loss": 0.4955, "step": 2702 }, { "epoch": 0.51, "learning_rate": 4.363413532441337e-06, "loss": 0.487, "step": 2703 }, { "epoch": 0.51, "learning_rate": 4.362903520705017e-06, "loss": 0.4145, "step": 2704 }, { "epoch": 0.51, "learning_rate": 4.362393334579979e-06, "loss": 0.5921, "step": 2705 }, { "epoch": 0.51, "learning_rate": 4.361882974113983e-06, "loss": 0.4657, "step": 2706 }, { "epoch": 0.51, "learning_rate": 4.3613724393548026e-06, "loss": 0.3565, "step": 2707 }, { "epoch": 0.51, "learning_rate": 4.360861730350231e-06, "loss": 0.6408, "step": 2708 }, { "epoch": 0.51, "learning_rate": 4.360350847148075e-06, "loss": 0.5215, "step": 2709 }, { "epoch": 0.51, "learning_rate": 4.35983978979616e-06, "loss": 0.5882, "step": 2710 }, { "epoch": 0.51, "learning_rate": 4.359328558342326e-06, "loss": 0.3681, "step": 2711 }, { "epoch": 0.51, "learning_rate": 4.358817152834432e-06, "loss": 0.3474, "step": 2712 }, { "epoch": 0.51, "learning_rate": 4.358305573320348e-06, "loss": 0.4948, "step": 2713 }, { "epoch": 0.51, "learning_rate": 4.357793819847966e-06, "loss": 0.4686, "step": 2714 }, { "epoch": 0.51, "learning_rate": 4.357281892465192e-06, "loss": 0.4773, "step": 2715 }, { "epoch": 0.51, "learning_rate": 4.356769791219947e-06, "loss": 0.5011, "step": 2716 }, { "epoch": 0.51, "learning_rate": 4.3562575161601694e-06, "loss": 0.5619, "step": 2717 }, { "epoch": 0.51, "learning_rate": 4.3557450673338154e-06, "loss": 0.5627, "step": 2718 }, { "epoch": 0.51, "learning_rate": 4.355232444788854e-06, "loss": 0.3979, "step": 2719 }, { "epoch": 0.51, "learning_rate": 4.354719648573274e-06, "loss": 0.6144, "step": 2720 }, { "epoch": 0.51, "learning_rate": 4.354206678735078e-06, "loss": 0.5862, "step": 2721 }, { "epoch": 0.51, "learning_rate": 4.353693535322285e-06, "loss": 0.4316, "step": 2722 }, { "epoch": 0.51, "learning_rate": 4.3531802183829336e-06, "loss": 0.5747, "step": 2723 }, { "epoch": 0.51, "learning_rate": 4.3526667279650746e-06, "loss": 0.3883, "step": 2724 }, { "epoch": 0.51, "learning_rate": 4.352153064116775e-06, "loss": 0.5771, "step": 2725 }, { "epoch": 0.52, "learning_rate": 4.351639226886122e-06, "loss": 0.5739, "step": 2726 }, { "epoch": 0.52, "learning_rate": 4.351125216321214e-06, "loss": 0.4538, "step": 2727 }, { "epoch": 0.52, "learning_rate": 4.350611032470171e-06, "loss": 0.3662, "step": 2728 }, { "epoch": 0.52, "learning_rate": 4.350096675381123e-06, "loss": 0.5181, "step": 2729 }, { "epoch": 0.52, "learning_rate": 4.349582145102224e-06, "loss": 0.557, "step": 2730 }, { "epoch": 0.52, "learning_rate": 4.349067441681636e-06, "loss": 0.5265, "step": 2731 }, { "epoch": 0.52, "learning_rate": 4.348552565167542e-06, "loss": 0.5081, "step": 2732 }, { "epoch": 0.52, "learning_rate": 4.348037515608142e-06, "loss": 0.6529, "step": 2733 }, { "epoch": 0.52, "learning_rate": 4.3475222930516484e-06, "loss": 0.5607, "step": 2734 }, { "epoch": 0.52, "learning_rate": 4.347006897546292e-06, "loss": 0.3721, "step": 2735 }, { "epoch": 0.52, "learning_rate": 4.346491329140321e-06, "loss": 0.3708, "step": 2736 }, { "epoch": 0.52, "learning_rate": 4.345975587881997e-06, "loss": 0.4374, "step": 2737 }, { "epoch": 0.52, "learning_rate": 4.3454596738196e-06, "loss": 0.4329, "step": 2738 }, { "epoch": 0.52, "learning_rate": 4.344943587001424e-06, "loss": 0.4981, "step": 2739 }, { "epoch": 0.52, "learning_rate": 4.344427327475783e-06, "loss": 0.522, "step": 2740 }, { "epoch": 0.52, "learning_rate": 4.343910895291002e-06, "loss": 0.4023, "step": 2741 }, { "epoch": 0.52, "learning_rate": 4.343394290495427e-06, "loss": 0.4906, "step": 2742 }, { "epoch": 0.52, "learning_rate": 4.342877513137417e-06, "loss": 0.555, "step": 2743 }, { "epoch": 0.52, "learning_rate": 4.342360563265348e-06, "loss": 0.5342, "step": 2744 }, { "epoch": 0.52, "learning_rate": 4.341843440927613e-06, "loss": 0.4291, "step": 2745 }, { "epoch": 0.52, "learning_rate": 4.341326146172619e-06, "loss": 0.5818, "step": 2746 }, { "epoch": 0.52, "learning_rate": 4.340808679048793e-06, "loss": 0.4691, "step": 2747 }, { "epoch": 0.52, "learning_rate": 4.340291039604574e-06, "loss": 0.5709, "step": 2748 }, { "epoch": 0.52, "learning_rate": 4.33977322788842e-06, "loss": 0.4463, "step": 2749 }, { "epoch": 0.52, "learning_rate": 4.339255243948802e-06, "loss": 0.4002, "step": 2750 }, { "epoch": 0.52, "learning_rate": 4.3387370878342114e-06, "loss": 0.5525, "step": 2751 }, { "epoch": 0.52, "learning_rate": 4.338218759593153e-06, "loss": 0.4009, "step": 2752 }, { "epoch": 0.52, "learning_rate": 4.337700259274147e-06, "loss": 0.4836, "step": 2753 }, { "epoch": 0.52, "learning_rate": 4.3371815869257304e-06, "loss": 0.5327, "step": 2754 }, { "epoch": 0.52, "learning_rate": 4.3366627425964595e-06, "loss": 0.4367, "step": 2755 }, { "epoch": 0.52, "learning_rate": 4.336143726334902e-06, "loss": 0.5924, "step": 2756 }, { "epoch": 0.52, "learning_rate": 4.335624538189643e-06, "loss": 0.7166, "step": 2757 }, { "epoch": 0.52, "learning_rate": 4.335105178209286e-06, "loss": 0.5952, "step": 2758 }, { "epoch": 0.52, "learning_rate": 4.334585646442448e-06, "loss": 0.4399, "step": 2759 }, { "epoch": 0.52, "learning_rate": 4.334065942937764e-06, "loss": 0.4578, "step": 2760 }, { "epoch": 0.52, "learning_rate": 4.333546067743883e-06, "loss": 0.4846, "step": 2761 }, { "epoch": 0.52, "learning_rate": 4.333026020909472e-06, "loss": 0.533, "step": 2762 }, { "epoch": 0.52, "learning_rate": 4.332505802483212e-06, "loss": 0.409, "step": 2763 }, { "epoch": 0.52, "learning_rate": 4.331985412513802e-06, "loss": 0.4739, "step": 2764 }, { "epoch": 0.52, "learning_rate": 4.331464851049957e-06, "loss": 0.5228, "step": 2765 }, { "epoch": 0.52, "learning_rate": 4.330944118140406e-06, "loss": 0.6935, "step": 2766 }, { "epoch": 0.52, "learning_rate": 4.3304232138338975e-06, "loss": 0.4178, "step": 2767 }, { "epoch": 0.52, "learning_rate": 4.329902138179192e-06, "loss": 0.5206, "step": 2768 }, { "epoch": 0.52, "learning_rate": 4.329380891225069e-06, "loss": 0.4987, "step": 2769 }, { "epoch": 0.52, "learning_rate": 4.328859473020323e-06, "loss": 0.4607, "step": 2770 }, { "epoch": 0.52, "learning_rate": 4.328337883613764e-06, "loss": 0.4987, "step": 2771 }, { "epoch": 0.52, "learning_rate": 4.32781612305422e-06, "loss": 0.4242, "step": 2772 }, { "epoch": 0.52, "learning_rate": 4.327294191390532e-06, "loss": 0.5096, "step": 2773 }, { "epoch": 0.52, "learning_rate": 4.32677208867156e-06, "loss": 0.3654, "step": 2774 }, { "epoch": 0.52, "learning_rate": 4.326249814946178e-06, "loss": 0.5645, "step": 2775 }, { "epoch": 0.52, "learning_rate": 4.325727370263276e-06, "loss": 0.5963, "step": 2776 }, { "epoch": 0.52, "learning_rate": 4.325204754671763e-06, "loss": 0.6234, "step": 2777 }, { "epoch": 0.52, "learning_rate": 4.32468196822056e-06, "loss": 0.4966, "step": 2778 }, { "epoch": 0.53, "learning_rate": 4.324159010958605e-06, "loss": 0.4896, "step": 2779 }, { "epoch": 0.53, "learning_rate": 4.323635882934855e-06, "loss": 0.4148, "step": 2780 }, { "epoch": 0.53, "learning_rate": 4.323112584198278e-06, "loss": 0.5279, "step": 2781 }, { "epoch": 0.53, "learning_rate": 4.322589114797862e-06, "loss": 0.6384, "step": 2782 }, { "epoch": 0.53, "learning_rate": 4.32206547478261e-06, "loss": 0.6023, "step": 2783 }, { "epoch": 0.53, "learning_rate": 4.321541664201539e-06, "loss": 0.3728, "step": 2784 }, { "epoch": 0.53, "learning_rate": 4.321017683103687e-06, "loss": 0.416, "step": 2785 }, { "epoch": 0.53, "learning_rate": 4.3204935315381005e-06, "loss": 0.4701, "step": 2786 }, { "epoch": 0.53, "learning_rate": 4.319969209553848e-06, "loss": 0.5858, "step": 2787 }, { "epoch": 0.53, "learning_rate": 4.3194447172000124e-06, "loss": 0.466, "step": 2788 }, { "epoch": 0.53, "learning_rate": 4.31892005452569e-06, "loss": 0.5266, "step": 2789 }, { "epoch": 0.53, "learning_rate": 4.318395221579998e-06, "loss": 0.5155, "step": 2790 }, { "epoch": 0.53, "learning_rate": 4.317870218412064e-06, "loss": 0.4565, "step": 2791 }, { "epoch": 0.53, "learning_rate": 4.3173450450710355e-06, "loss": 0.4195, "step": 2792 }, { "epoch": 0.53, "learning_rate": 4.316819701606074e-06, "loss": 0.5218, "step": 2793 }, { "epoch": 0.53, "learning_rate": 4.316294188066359e-06, "loss": 0.5212, "step": 2794 }, { "epoch": 0.53, "learning_rate": 4.315768504501083e-06, "loss": 0.4922, "step": 2795 }, { "epoch": 0.53, "learning_rate": 4.315242650959457e-06, "loss": 0.5855, "step": 2796 }, { "epoch": 0.53, "learning_rate": 4.3147166274907055e-06, "loss": 0.5064, "step": 2797 }, { "epoch": 0.53, "learning_rate": 4.314190434144071e-06, "loss": 0.5205, "step": 2798 }, { "epoch": 0.53, "learning_rate": 4.31366407096881e-06, "loss": 0.6359, "step": 2799 }, { "epoch": 0.53, "learning_rate": 4.313137538014199e-06, "loss": 0.4458, "step": 2800 }, { "epoch": 0.53, "learning_rate": 4.312610835329524e-06, "loss": 0.4206, "step": 2801 }, { "epoch": 0.53, "learning_rate": 4.312083962964092e-06, "loss": 0.4869, "step": 2802 }, { "epoch": 0.53, "learning_rate": 4.311556920967224e-06, "loss": 0.4575, "step": 2803 }, { "epoch": 0.53, "learning_rate": 4.311029709388257e-06, "loss": 0.4758, "step": 2804 }, { "epoch": 0.53, "learning_rate": 4.310502328276545e-06, "loss": 0.402, "step": 2805 }, { "epoch": 0.53, "learning_rate": 4.309974777681454e-06, "loss": 0.5712, "step": 2806 }, { "epoch": 0.53, "learning_rate": 4.309447057652371e-06, "loss": 0.5298, "step": 2807 }, { "epoch": 0.53, "learning_rate": 4.308919168238695e-06, "loss": 0.579, "step": 2808 }, { "epoch": 0.53, "learning_rate": 4.308391109489844e-06, "loss": 0.4893, "step": 2809 }, { "epoch": 0.53, "learning_rate": 4.30786288145525e-06, "loss": 0.5109, "step": 2810 }, { "epoch": 0.53, "learning_rate": 4.30733448418436e-06, "loss": 0.4805, "step": 2811 }, { "epoch": 0.53, "learning_rate": 4.306805917726638e-06, "loss": 0.3901, "step": 2812 }, { "epoch": 0.53, "learning_rate": 4.3062771821315664e-06, "loss": 0.5951, "step": 2813 }, { "epoch": 0.53, "learning_rate": 4.305748277448637e-06, "loss": 0.5039, "step": 2814 }, { "epoch": 0.53, "learning_rate": 4.3052192037273635e-06, "loss": 0.46, "step": 2815 }, { "epoch": 0.53, "learning_rate": 4.304689961017271e-06, "loss": 0.5208, "step": 2816 }, { "epoch": 0.53, "learning_rate": 4.304160549367906e-06, "loss": 0.5497, "step": 2817 }, { "epoch": 0.53, "learning_rate": 4.3036309688288246e-06, "loss": 0.4923, "step": 2818 }, { "epoch": 0.53, "learning_rate": 4.303101219449602e-06, "loss": 0.5346, "step": 2819 }, { "epoch": 0.53, "learning_rate": 4.30257130127983e-06, "loss": 0.3644, "step": 2820 }, { "epoch": 0.53, "learning_rate": 4.3020412143691135e-06, "loss": 0.5631, "step": 2821 }, { "epoch": 0.53, "learning_rate": 4.301510958767076e-06, "loss": 0.5426, "step": 2822 }, { "epoch": 0.53, "learning_rate": 4.300980534523354e-06, "loss": 0.4644, "step": 2823 }, { "epoch": 0.53, "learning_rate": 4.300449941687601e-06, "loss": 0.4911, "step": 2824 }, { "epoch": 0.53, "learning_rate": 4.299919180309488e-06, "loss": 0.5265, "step": 2825 }, { "epoch": 0.53, "learning_rate": 4.299388250438699e-06, "loss": 0.6287, "step": 2826 }, { "epoch": 0.53, "learning_rate": 4.298857152124935e-06, "loss": 0.4724, "step": 2827 }, { "epoch": 0.53, "learning_rate": 4.298325885417913e-06, "loss": 0.5613, "step": 2828 }, { "epoch": 0.53, "learning_rate": 4.297794450367366e-06, "loss": 0.4292, "step": 2829 }, { "epoch": 0.53, "learning_rate": 4.297262847023042e-06, "loss": 0.5708, "step": 2830 }, { "epoch": 0.53, "learning_rate": 4.296731075434705e-06, "loss": 0.5636, "step": 2831 }, { "epoch": 0.54, "learning_rate": 4.2961991356521345e-06, "loss": 0.5513, "step": 2832 }, { "epoch": 0.54, "learning_rate": 4.295667027725127e-06, "loss": 0.5094, "step": 2833 }, { "epoch": 0.54, "learning_rate": 4.295134751703493e-06, "loss": 0.68, "step": 2834 }, { "epoch": 0.54, "learning_rate": 4.294602307637059e-06, "loss": 0.4925, "step": 2835 }, { "epoch": 0.54, "learning_rate": 4.29406969557567e-06, "loss": 0.4889, "step": 2836 }, { "epoch": 0.54, "learning_rate": 4.2935369155691805e-06, "loss": 0.575, "step": 2837 }, { "epoch": 0.54, "learning_rate": 4.293003967667469e-06, "loss": 0.5232, "step": 2838 }, { "epoch": 0.54, "learning_rate": 4.292470851920423e-06, "loss": 0.419, "step": 2839 }, { "epoch": 0.54, "learning_rate": 4.291937568377949e-06, "loss": 0.4615, "step": 2840 }, { "epoch": 0.54, "learning_rate": 4.2914041170899665e-06, "loss": 0.5074, "step": 2841 }, { "epoch": 0.54, "learning_rate": 4.2908704981064155e-06, "loss": 0.3635, "step": 2842 }, { "epoch": 0.54, "learning_rate": 4.290336711477246e-06, "loss": 0.6037, "step": 2843 }, { "epoch": 0.54, "learning_rate": 4.289802757252429e-06, "loss": 0.4332, "step": 2844 }, { "epoch": 0.54, "learning_rate": 4.289268635481947e-06, "loss": 0.4781, "step": 2845 }, { "epoch": 0.54, "learning_rate": 4.2887343462158e-06, "loss": 0.4738, "step": 2846 }, { "epoch": 0.54, "learning_rate": 4.288199889504003e-06, "loss": 0.4489, "step": 2847 }, { "epoch": 0.54, "learning_rate": 4.287665265396589e-06, "loss": 0.7025, "step": 2848 }, { "epoch": 0.54, "learning_rate": 4.287130473943602e-06, "loss": 0.4337, "step": 2849 }, { "epoch": 0.54, "learning_rate": 4.286595515195107e-06, "loss": 0.5318, "step": 2850 }, { "epoch": 0.54, "learning_rate": 4.286060389201181e-06, "loss": 0.4258, "step": 2851 }, { "epoch": 0.54, "learning_rate": 4.285525096011918e-06, "loss": 0.4638, "step": 2852 }, { "epoch": 0.54, "learning_rate": 4.2849896356774276e-06, "loss": 0.4859, "step": 2853 }, { "epoch": 0.54, "learning_rate": 4.2844540082478345e-06, "loss": 0.6498, "step": 2854 }, { "epoch": 0.54, "learning_rate": 4.283918213773281e-06, "loss": 0.5238, "step": 2855 }, { "epoch": 0.54, "learning_rate": 4.283382252303921e-06, "loss": 0.4814, "step": 2856 }, { "epoch": 0.54, "learning_rate": 4.282846123889928e-06, "loss": 0.5445, "step": 2857 }, { "epoch": 0.54, "learning_rate": 4.28230982858149e-06, "loss": 0.583, "step": 2858 }, { "epoch": 0.54, "learning_rate": 4.281773366428808e-06, "loss": 0.4694, "step": 2859 }, { "epoch": 0.54, "learning_rate": 4.281236737482104e-06, "loss": 0.5339, "step": 2860 }, { "epoch": 0.54, "learning_rate": 4.280699941791612e-06, "loss": 0.5762, "step": 2861 }, { "epoch": 0.54, "learning_rate": 4.280162979407579e-06, "loss": 0.4742, "step": 2862 }, { "epoch": 0.54, "learning_rate": 4.279625850380275e-06, "loss": 0.5008, "step": 2863 }, { "epoch": 0.54, "learning_rate": 4.2790885547599775e-06, "loss": 0.5128, "step": 2864 }, { "epoch": 0.54, "learning_rate": 4.2785510925969855e-06, "loss": 0.4902, "step": 2865 }, { "epoch": 0.54, "learning_rate": 4.2780134639416116e-06, "loss": 0.6485, "step": 2866 }, { "epoch": 0.54, "learning_rate": 4.277475668844183e-06, "loss": 0.5166, "step": 2867 }, { "epoch": 0.54, "learning_rate": 4.276937707355044e-06, "loss": 0.4423, "step": 2868 }, { "epoch": 0.54, "learning_rate": 4.2763995795245545e-06, "loss": 0.4519, "step": 2869 }, { "epoch": 0.54, "learning_rate": 4.275861285403087e-06, "loss": 0.5489, "step": 2870 }, { "epoch": 0.54, "learning_rate": 4.275322825041034e-06, "loss": 0.6091, "step": 2871 }, { "epoch": 0.54, "learning_rate": 4.274784198488801e-06, "loss": 0.5652, "step": 2872 }, { "epoch": 0.54, "learning_rate": 4.27424540579681e-06, "loss": 0.3589, "step": 2873 }, { "epoch": 0.54, "learning_rate": 4.273706447015496e-06, "loss": 0.4264, "step": 2874 }, { "epoch": 0.54, "learning_rate": 4.273167322195313e-06, "loss": 0.4629, "step": 2875 }, { "epoch": 0.54, "learning_rate": 4.27262803138673e-06, "loss": 0.4589, "step": 2876 }, { "epoch": 0.54, "learning_rate": 4.272088574640229e-06, "loss": 0.5164, "step": 2877 }, { "epoch": 0.54, "learning_rate": 4.271548952006311e-06, "loss": 0.4115, "step": 2878 }, { "epoch": 0.54, "learning_rate": 4.271009163535488e-06, "loss": 0.4826, "step": 2879 }, { "epoch": 0.54, "learning_rate": 4.270469209278293e-06, "loss": 0.5093, "step": 2880 }, { "epoch": 0.54, "learning_rate": 4.269929089285271e-06, "loss": 0.4658, "step": 2881 }, { "epoch": 0.54, "learning_rate": 4.269388803606983e-06, "loss": 0.3984, "step": 2882 }, { "epoch": 0.54, "learning_rate": 4.268848352294006e-06, "loss": 0.5112, "step": 2883 }, { "epoch": 0.54, "learning_rate": 4.268307735396932e-06, "loss": 0.5734, "step": 2884 }, { "epoch": 0.55, "learning_rate": 4.267766952966369e-06, "loss": 0.4984, "step": 2885 }, { "epoch": 0.55, "learning_rate": 4.267226005052941e-06, "loss": 0.6588, "step": 2886 }, { "epoch": 0.55, "learning_rate": 4.266684891707286e-06, "loss": 0.4745, "step": 2887 }, { "epoch": 0.55, "learning_rate": 4.266143612980058e-06, "loss": 0.5462, "step": 2888 }, { "epoch": 0.55, "learning_rate": 4.265602168921927e-06, "loss": 0.445, "step": 2889 }, { "epoch": 0.55, "learning_rate": 4.265060559583579e-06, "loss": 0.4595, "step": 2890 }, { "epoch": 0.55, "learning_rate": 4.2645187850157144e-06, "loss": 0.4351, "step": 2891 }, { "epoch": 0.55, "learning_rate": 4.263976845269049e-06, "loss": 0.5982, "step": 2892 }, { "epoch": 0.55, "learning_rate": 4.263434740394314e-06, "loss": 0.518, "step": 2893 }, { "epoch": 0.55, "learning_rate": 4.262892470442258e-06, "loss": 0.4888, "step": 2894 }, { "epoch": 0.55, "learning_rate": 4.262350035463644e-06, "loss": 0.4427, "step": 2895 }, { "epoch": 0.55, "learning_rate": 4.261807435509247e-06, "loss": 0.4058, "step": 2896 }, { "epoch": 0.55, "learning_rate": 4.261264670629862e-06, "loss": 0.5453, "step": 2897 }, { "epoch": 0.55, "learning_rate": 4.260721740876299e-06, "loss": 0.4414, "step": 2898 }, { "epoch": 0.55, "learning_rate": 4.260178646299381e-06, "loss": 0.5497, "step": 2899 }, { "epoch": 0.55, "learning_rate": 4.259635386949949e-06, "loss": 0.5018, "step": 2900 }, { "epoch": 0.55, "learning_rate": 4.259091962878857e-06, "loss": 0.4964, "step": 2901 }, { "epoch": 0.55, "learning_rate": 4.258548374136976e-06, "loss": 0.5534, "step": 2902 }, { "epoch": 0.55, "learning_rate": 4.258004620775191e-06, "loss": 0.4899, "step": 2903 }, { "epoch": 0.55, "learning_rate": 4.257460702844406e-06, "loss": 0.5357, "step": 2904 }, { "epoch": 0.55, "learning_rate": 4.256916620395536e-06, "loss": 0.529, "step": 2905 }, { "epoch": 0.55, "learning_rate": 4.256372373479512e-06, "loss": 0.5237, "step": 2906 }, { "epoch": 0.55, "learning_rate": 4.255827962147285e-06, "loss": 0.5232, "step": 2907 }, { "epoch": 0.55, "learning_rate": 4.255283386449815e-06, "loss": 0.3824, "step": 2908 }, { "epoch": 0.55, "learning_rate": 4.254738646438083e-06, "loss": 0.6036, "step": 2909 }, { "epoch": 0.55, "learning_rate": 4.25419374216308e-06, "loss": 0.4628, "step": 2910 }, { "epoch": 0.55, "learning_rate": 4.253648673675817e-06, "loss": 0.5953, "step": 2911 }, { "epoch": 0.55, "learning_rate": 4.253103441027319e-06, "loss": 0.5524, "step": 2912 }, { "epoch": 0.55, "learning_rate": 4.252558044268623e-06, "loss": 0.4456, "step": 2913 }, { "epoch": 0.55, "learning_rate": 4.252012483450788e-06, "loss": 0.6024, "step": 2914 }, { "epoch": 0.55, "learning_rate": 4.251466758624882e-06, "loss": 0.4844, "step": 2915 }, { "epoch": 0.55, "learning_rate": 4.250920869841993e-06, "loss": 0.6253, "step": 2916 }, { "epoch": 0.55, "learning_rate": 4.250374817153219e-06, "loss": 0.3928, "step": 2917 }, { "epoch": 0.55, "learning_rate": 4.2498286006096815e-06, "loss": 0.5761, "step": 2918 }, { "epoch": 0.55, "learning_rate": 4.249282220262507e-06, "loss": 0.6764, "step": 2919 }, { "epoch": 0.55, "learning_rate": 4.248735676162848e-06, "loss": 0.5721, "step": 2920 }, { "epoch": 0.55, "learning_rate": 4.248188968361864e-06, "loss": 0.5038, "step": 2921 }, { "epoch": 0.55, "learning_rate": 4.247642096910734e-06, "loss": 0.51, "step": 2922 }, { "epoch": 0.55, "learning_rate": 4.247095061860652e-06, "loss": 0.382, "step": 2923 }, { "epoch": 0.55, "learning_rate": 4.246547863262825e-06, "loss": 0.4806, "step": 2924 }, { "epoch": 0.55, "learning_rate": 4.246000501168477e-06, "loss": 0.6149, "step": 2925 }, { "epoch": 0.55, "learning_rate": 4.24545297562885e-06, "loss": 0.4255, "step": 2926 }, { "epoch": 0.55, "learning_rate": 4.244905286695195e-06, "loss": 0.589, "step": 2927 }, { "epoch": 0.55, "learning_rate": 4.244357434418784e-06, "loss": 0.4648, "step": 2928 }, { "epoch": 0.55, "learning_rate": 4.243809418850901e-06, "loss": 0.6016, "step": 2929 }, { "epoch": 0.55, "learning_rate": 4.2432612400428485e-06, "loss": 0.5335, "step": 2930 }, { "epoch": 0.55, "learning_rate": 4.242712898045939e-06, "loss": 0.5024, "step": 2931 }, { "epoch": 0.55, "learning_rate": 4.242164392911507e-06, "loss": 0.6089, "step": 2932 }, { "epoch": 0.55, "learning_rate": 4.241615724690895e-06, "loss": 0.524, "step": 2933 }, { "epoch": 0.55, "learning_rate": 4.241066893435467e-06, "loss": 0.3932, "step": 2934 }, { "epoch": 0.55, "learning_rate": 4.2405178991966e-06, "loss": 0.5911, "step": 2935 }, { "epoch": 0.55, "learning_rate": 4.239968742025684e-06, "loss": 0.5551, "step": 2936 }, { "epoch": 0.55, "learning_rate": 4.239419421974128e-06, "loss": 0.4618, "step": 2937 }, { "epoch": 0.56, "learning_rate": 4.238869939093354e-06, "loss": 0.5044, "step": 2938 }, { "epoch": 0.56, "learning_rate": 4.2383202934348e-06, "loss": 0.5896, "step": 2939 }, { "epoch": 0.56, "learning_rate": 4.237770485049919e-06, "loss": 0.4572, "step": 2940 }, { "epoch": 0.56, "learning_rate": 4.237220513990179e-06, "loss": 0.4545, "step": 2941 }, { "epoch": 0.56, "learning_rate": 4.236670380307064e-06, "loss": 0.4164, "step": 2942 }, { "epoch": 0.56, "learning_rate": 4.236120084052072e-06, "loss": 0.3992, "step": 2943 }, { "epoch": 0.56, "learning_rate": 4.235569625276716e-06, "loss": 0.4077, "step": 2944 }, { "epoch": 0.56, "learning_rate": 4.235019004032528e-06, "loss": 0.4018, "step": 2945 }, { "epoch": 0.56, "learning_rate": 4.23446822037105e-06, "loss": 0.511, "step": 2946 }, { "epoch": 0.56, "learning_rate": 4.233917274343842e-06, "loss": 0.5305, "step": 2947 }, { "epoch": 0.56, "learning_rate": 4.233366166002479e-06, "loss": 0.6266, "step": 2948 }, { "epoch": 0.56, "learning_rate": 4.232814895398552e-06, "loss": 0.471, "step": 2949 }, { "epoch": 0.56, "learning_rate": 4.232263462583663e-06, "loss": 0.5397, "step": 2950 }, { "epoch": 0.56, "learning_rate": 4.2317118676094356e-06, "loss": 0.4889, "step": 2951 }, { "epoch": 0.56, "learning_rate": 4.231160110527504e-06, "loss": 0.5346, "step": 2952 }, { "epoch": 0.56, "learning_rate": 4.230608191389518e-06, "loss": 0.392, "step": 2953 }, { "epoch": 0.56, "learning_rate": 4.2300561102471445e-06, "loss": 0.3669, "step": 2954 }, { "epoch": 0.56, "learning_rate": 4.229503867152065e-06, "loss": 0.4301, "step": 2955 }, { "epoch": 0.56, "learning_rate": 4.228951462155975e-06, "loss": 0.4397, "step": 2956 }, { "epoch": 0.56, "learning_rate": 4.228398895310586e-06, "loss": 0.4722, "step": 2957 }, { "epoch": 0.56, "learning_rate": 4.227846166667623e-06, "loss": 0.521, "step": 2958 }, { "epoch": 0.56, "learning_rate": 4.22729327627883e-06, "loss": 0.4941, "step": 2959 }, { "epoch": 0.56, "learning_rate": 4.226740224195962e-06, "loss": 0.5466, "step": 2960 }, { "epoch": 0.56, "learning_rate": 4.226187010470792e-06, "loss": 0.5811, "step": 2961 }, { "epoch": 0.56, "learning_rate": 4.225633635155108e-06, "loss": 0.5658, "step": 2962 }, { "epoch": 0.56, "learning_rate": 4.225080098300708e-06, "loss": 0.5195, "step": 2963 }, { "epoch": 0.56, "learning_rate": 4.224526399959414e-06, "loss": 0.4953, "step": 2964 }, { "epoch": 0.56, "learning_rate": 4.2239725401830545e-06, "loss": 0.5337, "step": 2965 }, { "epoch": 0.56, "learning_rate": 4.223418519023481e-06, "loss": 0.4918, "step": 2966 }, { "epoch": 0.56, "learning_rate": 4.2228643365325525e-06, "loss": 0.4127, "step": 2967 }, { "epoch": 0.56, "learning_rate": 4.2223099927621484e-06, "loss": 0.5263, "step": 2968 }, { "epoch": 0.56, "learning_rate": 4.221755487764162e-06, "loss": 0.4332, "step": 2969 }, { "epoch": 0.56, "learning_rate": 4.2212008215905e-06, "loss": 0.5266, "step": 2970 }, { "epoch": 0.56, "learning_rate": 4.220645994293086e-06, "loss": 0.4845, "step": 2971 }, { "epoch": 0.56, "learning_rate": 4.220091005923859e-06, "loss": 0.4902, "step": 2972 }, { "epoch": 0.56, "learning_rate": 4.21953585653477e-06, "loss": 0.5387, "step": 2973 }, { "epoch": 0.56, "learning_rate": 4.218980546177788e-06, "loss": 0.3929, "step": 2974 }, { "epoch": 0.56, "learning_rate": 4.218425074904897e-06, "loss": 0.5113, "step": 2975 }, { "epoch": 0.56, "learning_rate": 4.217869442768095e-06, "loss": 0.4949, "step": 2976 }, { "epoch": 0.56, "learning_rate": 4.217313649819397e-06, "loss": 0.5084, "step": 2977 }, { "epoch": 0.56, "learning_rate": 4.216757696110828e-06, "loss": 0.4763, "step": 2978 }, { "epoch": 0.56, "learning_rate": 4.216201581694435e-06, "loss": 0.413, "step": 2979 }, { "epoch": 0.56, "learning_rate": 4.215645306622273e-06, "loss": 0.4286, "step": 2980 }, { "epoch": 0.56, "learning_rate": 4.215088870946419e-06, "loss": 0.4723, "step": 2981 }, { "epoch": 0.56, "learning_rate": 4.21453227471896e-06, "loss": 0.4168, "step": 2982 }, { "epoch": 0.56, "learning_rate": 4.213975517991999e-06, "loss": 0.4891, "step": 2983 }, { "epoch": 0.56, "learning_rate": 4.213418600817657e-06, "loss": 0.5018, "step": 2984 }, { "epoch": 0.56, "learning_rate": 4.212861523248066e-06, "loss": 0.5204, "step": 2985 }, { "epoch": 0.56, "learning_rate": 4.212304285335375e-06, "loss": 0.6327, "step": 2986 }, { "epoch": 0.56, "learning_rate": 4.211746887131747e-06, "loss": 0.5534, "step": 2987 }, { "epoch": 0.56, "learning_rate": 4.211189328689363e-06, "loss": 0.5462, "step": 2988 }, { "epoch": 0.56, "learning_rate": 4.210631610060413e-06, "loss": 0.4776, "step": 2989 }, { "epoch": 0.56, "learning_rate": 4.210073731297109e-06, "loss": 0.4689, "step": 2990 }, { "epoch": 0.57, "learning_rate": 4.209515692451674e-06, "loss": 0.402, "step": 2991 }, { "epoch": 0.57, "learning_rate": 4.208957493576346e-06, "loss": 0.4203, "step": 2992 }, { "epoch": 0.57, "learning_rate": 4.2083991347233795e-06, "loss": 0.5491, "step": 2993 }, { "epoch": 0.57, "learning_rate": 4.207840615945043e-06, "loss": 0.4332, "step": 2994 }, { "epoch": 0.57, "learning_rate": 4.207281937293619e-06, "loss": 0.4834, "step": 2995 }, { "epoch": 0.57, "learning_rate": 4.206723098821408e-06, "loss": 0.4835, "step": 2996 }, { "epoch": 0.57, "learning_rate": 4.206164100580722e-06, "loss": 0.4813, "step": 2997 }, { "epoch": 0.57, "learning_rate": 4.20560494262389e-06, "loss": 0.4791, "step": 2998 }, { "epoch": 0.57, "learning_rate": 4.205045625003256e-06, "loss": 0.4853, "step": 2999 }, { "epoch": 0.57, "learning_rate": 4.204486147771178e-06, "loss": 0.4337, "step": 3000 }, { "epoch": 0.57, "learning_rate": 4.203926510980029e-06, "loss": 0.4835, "step": 3001 }, { "epoch": 0.57, "learning_rate": 4.203366714682197e-06, "loss": 0.4416, "step": 3002 }, { "epoch": 0.57, "learning_rate": 4.202806758930088e-06, "loss": 0.4228, "step": 3003 }, { "epoch": 0.57, "learning_rate": 4.202246643776116e-06, "loss": 0.4446, "step": 3004 }, { "epoch": 0.57, "learning_rate": 4.2016863692727174e-06, "loss": 0.5615, "step": 3005 }, { "epoch": 0.57, "learning_rate": 4.201125935472338e-06, "loss": 0.5387, "step": 3006 }, { "epoch": 0.57, "learning_rate": 4.2005653424274415e-06, "loss": 0.4844, "step": 3007 }, { "epoch": 0.57, "learning_rate": 4.2000045901905065e-06, "loss": 0.6694, "step": 3008 }, { "epoch": 0.57, "learning_rate": 4.199443678814024e-06, "loss": 0.4242, "step": 3009 }, { "epoch": 0.57, "learning_rate": 4.198882608350504e-06, "loss": 0.5061, "step": 3010 }, { "epoch": 0.57, "learning_rate": 4.198321378852467e-06, "loss": 0.4217, "step": 3011 }, { "epoch": 0.57, "learning_rate": 4.197759990372451e-06, "loss": 0.4433, "step": 3012 }, { "epoch": 0.57, "learning_rate": 4.197198442963009e-06, "loss": 0.4865, "step": 3013 }, { "epoch": 0.57, "learning_rate": 4.1966367366767065e-06, "loss": 0.5071, "step": 3014 }, { "epoch": 0.57, "learning_rate": 4.196074871566127e-06, "loss": 0.6095, "step": 3015 }, { "epoch": 0.57, "learning_rate": 4.195512847683867e-06, "loss": 0.4186, "step": 3016 }, { "epoch": 0.57, "learning_rate": 4.194950665082538e-06, "loss": 0.4315, "step": 3017 }, { "epoch": 0.57, "learning_rate": 4.194388323814766e-06, "loss": 0.4753, "step": 3018 }, { "epoch": 0.57, "learning_rate": 4.1938258239331946e-06, "loss": 0.5028, "step": 3019 }, { "epoch": 0.57, "learning_rate": 4.193263165490478e-06, "loss": 0.3573, "step": 3020 }, { "epoch": 0.57, "learning_rate": 4.192700348539288e-06, "loss": 0.6247, "step": 3021 }, { "epoch": 0.57, "learning_rate": 4.192137373132311e-06, "loss": 0.4853, "step": 3022 }, { "epoch": 0.57, "learning_rate": 4.191574239322248e-06, "loss": 0.4214, "step": 3023 }, { "epoch": 0.57, "learning_rate": 4.191010947161813e-06, "loss": 0.5106, "step": 3024 }, { "epoch": 0.57, "learning_rate": 4.190447496703738e-06, "loss": 0.3709, "step": 3025 }, { "epoch": 0.57, "learning_rate": 4.189883888000768e-06, "loss": 0.4512, "step": 3026 }, { "epoch": 0.57, "learning_rate": 4.1893201211056634e-06, "loss": 0.5684, "step": 3027 }, { "epoch": 0.57, "learning_rate": 4.1887561960711986e-06, "loss": 0.4725, "step": 3028 }, { "epoch": 0.57, "learning_rate": 4.188192112950164e-06, "loss": 0.4951, "step": 3029 }, { "epoch": 0.57, "learning_rate": 4.187627871795363e-06, "loss": 0.512, "step": 3030 }, { "epoch": 0.57, "learning_rate": 4.187063472659616e-06, "loss": 0.5694, "step": 3031 }, { "epoch": 0.57, "learning_rate": 4.186498915595756e-06, "loss": 0.6145, "step": 3032 }, { "epoch": 0.57, "learning_rate": 4.185934200656633e-06, "loss": 0.5516, "step": 3033 }, { "epoch": 0.57, "learning_rate": 4.1853693278951105e-06, "loss": 0.4583, "step": 3034 }, { "epoch": 0.57, "learning_rate": 4.184804297364066e-06, "loss": 0.4095, "step": 3035 }, { "epoch": 0.57, "learning_rate": 4.184239109116393e-06, "loss": 0.7403, "step": 3036 }, { "epoch": 0.57, "learning_rate": 4.183673763205001e-06, "loss": 0.4183, "step": 3037 }, { "epoch": 0.57, "learning_rate": 4.183108259682811e-06, "loss": 0.4678, "step": 3038 }, { "epoch": 0.57, "learning_rate": 4.182542598602761e-06, "loss": 0.5117, "step": 3039 }, { "epoch": 0.57, "learning_rate": 4.181976780017804e-06, "loss": 0.4311, "step": 3040 }, { "epoch": 0.57, "learning_rate": 4.1814108039809055e-06, "loss": 0.4806, "step": 3041 }, { "epoch": 0.57, "learning_rate": 4.180844670545048e-06, "loss": 0.5515, "step": 3042 }, { "epoch": 0.57, "learning_rate": 4.180278379763228e-06, "loss": 0.382, "step": 3043 }, { "epoch": 0.58, "learning_rate": 4.1797119316884575e-06, "loss": 0.3755, "step": 3044 }, { "epoch": 0.58, "learning_rate": 4.1791453263737616e-06, "loss": 0.4411, "step": 3045 }, { "epoch": 0.58, "learning_rate": 4.17857856387218e-06, "loss": 0.4504, "step": 3046 }, { "epoch": 0.58, "learning_rate": 4.178011644236769e-06, "loss": 0.3935, "step": 3047 }, { "epoch": 0.58, "learning_rate": 4.177444567520599e-06, "loss": 0.4956, "step": 3048 }, { "epoch": 0.58, "learning_rate": 4.176877333776754e-06, "loss": 0.4264, "step": 3049 }, { "epoch": 0.58, "learning_rate": 4.176309943058335e-06, "loss": 0.4911, "step": 3050 }, { "epoch": 0.58, "learning_rate": 4.175742395418453e-06, "loss": 0.4484, "step": 3051 }, { "epoch": 0.58, "learning_rate": 4.17517469091024e-06, "loss": 0.5233, "step": 3052 }, { "epoch": 0.58, "learning_rate": 4.174606829586838e-06, "loss": 0.5557, "step": 3053 }, { "epoch": 0.58, "learning_rate": 4.174038811501405e-06, "loss": 0.596, "step": 3054 }, { "epoch": 0.58, "learning_rate": 4.1734706367071155e-06, "loss": 0.5279, "step": 3055 }, { "epoch": 0.58, "learning_rate": 4.172902305257155e-06, "loss": 0.5141, "step": 3056 }, { "epoch": 0.58, "learning_rate": 4.172333817204727e-06, "loss": 0.4989, "step": 3057 }, { "epoch": 0.58, "learning_rate": 4.171765172603048e-06, "loss": 0.4858, "step": 3058 }, { "epoch": 0.58, "learning_rate": 4.1711963715053505e-06, "loss": 0.4205, "step": 3059 }, { "epoch": 0.58, "learning_rate": 4.170627413964879e-06, "loss": 0.5518, "step": 3060 }, { "epoch": 0.58, "learning_rate": 4.170058300034895e-06, "loss": 0.423, "step": 3061 }, { "epoch": 0.58, "learning_rate": 4.169489029768674e-06, "loss": 0.5296, "step": 3062 }, { "epoch": 0.58, "learning_rate": 4.168919603219506e-06, "loss": 0.6011, "step": 3063 }, { "epoch": 0.58, "learning_rate": 4.168350020440696e-06, "loss": 0.6315, "step": 3064 }, { "epoch": 0.58, "learning_rate": 4.167780281485564e-06, "loss": 0.445, "step": 3065 }, { "epoch": 0.58, "learning_rate": 4.167210386407442e-06, "loss": 0.5244, "step": 3066 }, { "epoch": 0.58, "learning_rate": 4.16664033525968e-06, "loss": 0.6755, "step": 3067 }, { "epoch": 0.58, "learning_rate": 4.166070128095642e-06, "loss": 0.4758, "step": 3068 }, { "epoch": 0.58, "learning_rate": 4.165499764968705e-06, "loss": 0.4576, "step": 3069 }, { "epoch": 0.58, "learning_rate": 4.1649292459322595e-06, "loss": 0.5057, "step": 3070 }, { "epoch": 0.58, "learning_rate": 4.164358571039715e-06, "loss": 0.4065, "step": 3071 }, { "epoch": 0.58, "learning_rate": 4.163787740344492e-06, "loss": 0.5264, "step": 3072 }, { "epoch": 0.58, "learning_rate": 4.163216753900027e-06, "loss": 0.4217, "step": 3073 }, { "epoch": 0.58, "learning_rate": 4.162645611759772e-06, "loss": 0.619, "step": 3074 }, { "epoch": 0.58, "learning_rate": 4.162074313977189e-06, "loss": 0.4929, "step": 3075 }, { "epoch": 0.58, "learning_rate": 4.161502860605761e-06, "loss": 0.4509, "step": 3076 }, { "epoch": 0.58, "learning_rate": 4.160931251698981e-06, "loss": 0.4769, "step": 3077 }, { "epoch": 0.58, "learning_rate": 4.160359487310359e-06, "loss": 0.4638, "step": 3078 }, { "epoch": 0.58, "learning_rate": 4.159787567493418e-06, "loss": 0.5485, "step": 3079 }, { "epoch": 0.58, "learning_rate": 4.159215492301695e-06, "loss": 0.4673, "step": 3080 }, { "epoch": 0.58, "learning_rate": 4.158643261788745e-06, "loss": 0.4341, "step": 3081 }, { "epoch": 0.58, "learning_rate": 4.158070876008133e-06, "loss": 0.4642, "step": 3082 }, { "epoch": 0.58, "learning_rate": 4.157498335013442e-06, "loss": 0.5766, "step": 3083 }, { "epoch": 0.58, "learning_rate": 4.156925638858269e-06, "loss": 0.3788, "step": 3084 }, { "epoch": 0.58, "learning_rate": 4.156352787596223e-06, "loss": 0.5608, "step": 3085 }, { "epoch": 0.58, "learning_rate": 4.15577978128093e-06, "loss": 0.4153, "step": 3086 }, { "epoch": 0.58, "learning_rate": 4.15520661996603e-06, "loss": 0.4353, "step": 3087 }, { "epoch": 0.58, "learning_rate": 4.154633303705178e-06, "loss": 0.4489, "step": 3088 }, { "epoch": 0.58, "learning_rate": 4.154059832552041e-06, "loss": 0.4483, "step": 3089 }, { "epoch": 0.58, "learning_rate": 4.153486206560304e-06, "loss": 0.5168, "step": 3090 }, { "epoch": 0.58, "learning_rate": 4.152912425783665e-06, "loss": 0.5621, "step": 3091 }, { "epoch": 0.58, "learning_rate": 4.152338490275835e-06, "loss": 0.4467, "step": 3092 }, { "epoch": 0.58, "learning_rate": 4.151764400090541e-06, "loss": 0.5843, "step": 3093 }, { "epoch": 0.58, "learning_rate": 4.151190155281526e-06, "loss": 0.68, "step": 3094 }, { "epoch": 0.58, "learning_rate": 4.150615755902544e-06, "loss": 0.5477, "step": 3095 }, { "epoch": 0.58, "learning_rate": 4.150041202007365e-06, "loss": 0.4509, "step": 3096 }, { "epoch": 0.59, "learning_rate": 4.149466493649776e-06, "loss": 0.591, "step": 3097 }, { "epoch": 0.59, "learning_rate": 4.1488916308835745e-06, "loss": 0.5063, "step": 3098 }, { "epoch": 0.59, "learning_rate": 4.148316613762573e-06, "loss": 0.3167, "step": 3099 }, { "epoch": 0.59, "learning_rate": 4.147741442340602e-06, "loss": 0.5655, "step": 3100 }, { "epoch": 0.59, "learning_rate": 4.1471661166715025e-06, "loss": 0.5589, "step": 3101 }, { "epoch": 0.59, "learning_rate": 4.146590636809132e-06, "loss": 0.3971, "step": 3102 }, { "epoch": 0.59, "learning_rate": 4.146015002807362e-06, "loss": 0.4669, "step": 3103 }, { "epoch": 0.59, "learning_rate": 4.145439214720077e-06, "loss": 0.4846, "step": 3104 }, { "epoch": 0.59, "learning_rate": 4.14486327260118e-06, "loss": 0.5871, "step": 3105 }, { "epoch": 0.59, "learning_rate": 4.144287176504582e-06, "loss": 0.5422, "step": 3106 }, { "epoch": 0.59, "learning_rate": 4.1437109264842156e-06, "loss": 0.4519, "step": 3107 }, { "epoch": 0.59, "learning_rate": 4.1431345225940215e-06, "loss": 0.4772, "step": 3108 }, { "epoch": 0.59, "learning_rate": 4.142557964887959e-06, "loss": 0.516, "step": 3109 }, { "epoch": 0.59, "learning_rate": 4.141981253420001e-06, "loss": 0.4568, "step": 3110 }, { "epoch": 0.59, "learning_rate": 4.141404388244133e-06, "loss": 0.4372, "step": 3111 }, { "epoch": 0.59, "learning_rate": 4.140827369414355e-06, "loss": 0.5531, "step": 3112 }, { "epoch": 0.59, "learning_rate": 4.140250196984685e-06, "loss": 0.3859, "step": 3113 }, { "epoch": 0.59, "learning_rate": 4.139672871009152e-06, "loss": 0.4575, "step": 3114 }, { "epoch": 0.59, "learning_rate": 4.139095391541799e-06, "loss": 0.4085, "step": 3115 }, { "epoch": 0.59, "learning_rate": 4.138517758636686e-06, "loss": 0.5864, "step": 3116 }, { "epoch": 0.59, "learning_rate": 4.137939972347886e-06, "loss": 0.4235, "step": 3117 }, { "epoch": 0.59, "learning_rate": 4.137362032729485e-06, "loss": 0.4387, "step": 3118 }, { "epoch": 0.59, "learning_rate": 4.1367839398355855e-06, "loss": 0.5272, "step": 3119 }, { "epoch": 0.59, "learning_rate": 4.136205693720303e-06, "loss": 0.5355, "step": 3120 }, { "epoch": 0.59, "learning_rate": 4.135627294437769e-06, "loss": 0.4624, "step": 3121 }, { "epoch": 0.59, "learning_rate": 4.1350487420421266e-06, "loss": 0.582, "step": 3122 }, { "epoch": 0.59, "learning_rate": 4.134470036587536e-06, "loss": 0.4153, "step": 3123 }, { "epoch": 0.59, "learning_rate": 4.13389117812817e-06, "loss": 0.5403, "step": 3124 }, { "epoch": 0.59, "learning_rate": 4.133312166718216e-06, "loss": 0.3985, "step": 3125 }, { "epoch": 0.59, "learning_rate": 4.132733002411877e-06, "loss": 0.4498, "step": 3126 }, { "epoch": 0.59, "learning_rate": 4.132153685263368e-06, "loss": 0.4489, "step": 3127 }, { "epoch": 0.59, "learning_rate": 4.13157421532692e-06, "loss": 0.5379, "step": 3128 }, { "epoch": 0.59, "learning_rate": 4.130994592656778e-06, "loss": 0.4743, "step": 3129 }, { "epoch": 0.59, "learning_rate": 4.130414817307204e-06, "loss": 0.4099, "step": 3130 }, { "epoch": 0.59, "learning_rate": 4.129834889332465e-06, "loss": 0.5816, "step": 3131 }, { "epoch": 0.59, "learning_rate": 4.129254808786855e-06, "loss": 0.532, "step": 3132 }, { "epoch": 0.59, "learning_rate": 4.128674575724673e-06, "loss": 0.4089, "step": 3133 }, { "epoch": 0.59, "learning_rate": 4.128094190200236e-06, "loss": 0.5099, "step": 3134 }, { "epoch": 0.59, "learning_rate": 4.127513652267875e-06, "loss": 0.5694, "step": 3135 }, { "epoch": 0.59, "learning_rate": 4.126932961981934e-06, "loss": 0.5025, "step": 3136 }, { "epoch": 0.59, "learning_rate": 4.126352119396773e-06, "loss": 0.4973, "step": 3137 }, { "epoch": 0.59, "learning_rate": 4.125771124566764e-06, "loss": 0.4895, "step": 3138 }, { "epoch": 0.59, "learning_rate": 4.125189977546297e-06, "loss": 0.3958, "step": 3139 }, { "epoch": 0.59, "learning_rate": 4.124608678389772e-06, "loss": 0.519, "step": 3140 }, { "epoch": 0.59, "learning_rate": 4.124027227151606e-06, "loss": 0.5617, "step": 3141 }, { "epoch": 0.59, "learning_rate": 4.123445623886229e-06, "loss": 0.4078, "step": 3142 }, { "epoch": 0.59, "learning_rate": 4.122863868648085e-06, "loss": 0.4878, "step": 3143 }, { "epoch": 0.59, "learning_rate": 4.1222819614916344e-06, "loss": 0.5302, "step": 3144 }, { "epoch": 0.59, "learning_rate": 4.121699902471349e-06, "loss": 0.4287, "step": 3145 }, { "epoch": 0.59, "learning_rate": 4.1211176916417165e-06, "loss": 0.5404, "step": 3146 }, { "epoch": 0.59, "learning_rate": 4.120535329057238e-06, "loss": 0.4843, "step": 3147 }, { "epoch": 0.59, "learning_rate": 4.11995281477243e-06, "loss": 0.6477, "step": 3148 }, { "epoch": 0.59, "learning_rate": 4.119370148841822e-06, "loss": 0.4225, "step": 3149 }, { "epoch": 0.6, "learning_rate": 4.1187873313199575e-06, "loss": 0.4688, "step": 3150 }, { "epoch": 0.6, "learning_rate": 4.118204362261396e-06, "loss": 0.5667, "step": 3151 }, { "epoch": 0.6, "learning_rate": 4.117621241720709e-06, "loss": 0.3634, "step": 3152 }, { "epoch": 0.6, "learning_rate": 4.1170379697524834e-06, "loss": 0.5319, "step": 3153 }, { "epoch": 0.6, "learning_rate": 4.11645454641132e-06, "loss": 0.5176, "step": 3154 }, { "epoch": 0.6, "learning_rate": 4.115870971751833e-06, "loss": 0.4073, "step": 3155 }, { "epoch": 0.6, "learning_rate": 4.115287245828653e-06, "loss": 0.5011, "step": 3156 }, { "epoch": 0.6, "learning_rate": 4.114703368696421e-06, "loss": 0.544, "step": 3157 }, { "epoch": 0.6, "learning_rate": 4.114119340409798e-06, "loss": 0.4046, "step": 3158 }, { "epoch": 0.6, "learning_rate": 4.113535161023454e-06, "loss": 0.602, "step": 3159 }, { "epoch": 0.6, "learning_rate": 4.112950830592073e-06, "loss": 0.5892, "step": 3160 }, { "epoch": 0.6, "learning_rate": 4.112366349170357e-06, "loss": 0.5384, "step": 3161 }, { "epoch": 0.6, "learning_rate": 4.111781716813021e-06, "loss": 0.4825, "step": 3162 }, { "epoch": 0.6, "learning_rate": 4.111196933574789e-06, "loss": 0.4618, "step": 3163 }, { "epoch": 0.6, "learning_rate": 4.110611999510407e-06, "loss": 0.3866, "step": 3164 }, { "epoch": 0.6, "learning_rate": 4.11002691467463e-06, "loss": 0.4119, "step": 3165 }, { "epoch": 0.6, "learning_rate": 4.1094416791222295e-06, "loss": 0.4433, "step": 3166 }, { "epoch": 0.6, "learning_rate": 4.108856292907988e-06, "loss": 0.4999, "step": 3167 }, { "epoch": 0.6, "learning_rate": 4.108270756086707e-06, "loss": 0.4694, "step": 3168 }, { "epoch": 0.6, "learning_rate": 4.107685068713198e-06, "loss": 0.4904, "step": 3169 }, { "epoch": 0.6, "learning_rate": 4.107099230842286e-06, "loss": 0.4548, "step": 3170 }, { "epoch": 0.6, "learning_rate": 4.106513242528815e-06, "loss": 0.4783, "step": 3171 }, { "epoch": 0.6, "learning_rate": 4.105927103827641e-06, "loss": 0.3546, "step": 3172 }, { "epoch": 0.6, "learning_rate": 4.105340814793629e-06, "loss": 0.4247, "step": 3173 }, { "epoch": 0.6, "learning_rate": 4.104754375481665e-06, "loss": 0.4457, "step": 3174 }, { "epoch": 0.6, "learning_rate": 4.104167785946645e-06, "loss": 0.6067, "step": 3175 }, { "epoch": 0.6, "learning_rate": 4.103581046243481e-06, "loss": 0.5554, "step": 3176 }, { "epoch": 0.6, "learning_rate": 4.1029941564271e-06, "loss": 0.4221, "step": 3177 }, { "epoch": 0.6, "learning_rate": 4.102407116552439e-06, "loss": 0.6348, "step": 3178 }, { "epoch": 0.6, "learning_rate": 4.101819926674453e-06, "loss": 0.5083, "step": 3179 }, { "epoch": 0.6, "learning_rate": 4.10123258684811e-06, "loss": 0.6017, "step": 3180 }, { "epoch": 0.6, "learning_rate": 4.100645097128389e-06, "loss": 0.5479, "step": 3181 }, { "epoch": 0.6, "learning_rate": 4.100057457570288e-06, "loss": 0.4846, "step": 3182 }, { "epoch": 0.6, "learning_rate": 4.099469668228816e-06, "loss": 0.5217, "step": 3183 }, { "epoch": 0.6, "learning_rate": 4.098881729158997e-06, "loss": 0.6367, "step": 3184 }, { "epoch": 0.6, "learning_rate": 4.098293640415869e-06, "loss": 0.5424, "step": 3185 }, { "epoch": 0.6, "learning_rate": 4.0977054020544825e-06, "loss": 0.5848, "step": 3186 }, { "epoch": 0.6, "learning_rate": 4.097117014129903e-06, "loss": 0.6403, "step": 3187 }, { "epoch": 0.6, "learning_rate": 4.096528476697213e-06, "loss": 0.4625, "step": 3188 }, { "epoch": 0.6, "learning_rate": 4.0959397898115036e-06, "loss": 0.44, "step": 3189 }, { "epoch": 0.6, "learning_rate": 4.095350953527883e-06, "loss": 0.5819, "step": 3190 }, { "epoch": 0.6, "learning_rate": 4.094761967901474e-06, "loss": 0.4653, "step": 3191 }, { "epoch": 0.6, "learning_rate": 4.0941728329874094e-06, "loss": 0.4126, "step": 3192 }, { "epoch": 0.6, "learning_rate": 4.093583548840843e-06, "loss": 0.5565, "step": 3193 }, { "epoch": 0.6, "learning_rate": 4.092994115516936e-06, "loss": 0.4326, "step": 3194 }, { "epoch": 0.6, "learning_rate": 4.092404533070866e-06, "loss": 0.5631, "step": 3195 }, { "epoch": 0.6, "learning_rate": 4.0918148015578256e-06, "loss": 0.4797, "step": 3196 }, { "epoch": 0.6, "learning_rate": 4.091224921033018e-06, "loss": 0.4379, "step": 3197 }, { "epoch": 0.6, "learning_rate": 4.090634891551665e-06, "loss": 0.4562, "step": 3198 }, { "epoch": 0.6, "learning_rate": 4.090044713168999e-06, "loss": 0.4234, "step": 3199 }, { "epoch": 0.6, "learning_rate": 4.089454385940268e-06, "loss": 0.4999, "step": 3200 }, { "epoch": 0.6, "learning_rate": 4.088863909920733e-06, "loss": 0.6955, "step": 3201 }, { "epoch": 0.6, "learning_rate": 4.0882732851656685e-06, "loss": 0.5218, "step": 3202 }, { "epoch": 0.61, "learning_rate": 4.087682511730364e-06, "loss": 0.4383, "step": 3203 }, { "epoch": 0.61, "learning_rate": 4.087091589670123e-06, "loss": 0.46, "step": 3204 }, { "epoch": 0.61, "learning_rate": 4.086500519040262e-06, "loss": 0.6066, "step": 3205 }, { "epoch": 0.61, "learning_rate": 4.085909299896112e-06, "loss": 0.732, "step": 3206 }, { "epoch": 0.61, "learning_rate": 4.085317932293017e-06, "loss": 0.4847, "step": 3207 }, { "epoch": 0.61, "learning_rate": 4.084726416286338e-06, "loss": 0.526, "step": 3208 }, { "epoch": 0.61, "learning_rate": 4.084134751931445e-06, "loss": 0.4363, "step": 3209 }, { "epoch": 0.61, "learning_rate": 4.083542939283725e-06, "loss": 0.4814, "step": 3210 }, { "epoch": 0.61, "learning_rate": 4.0829509783985785e-06, "loss": 0.5361, "step": 3211 }, { "epoch": 0.61, "learning_rate": 4.082358869331421e-06, "loss": 0.4586, "step": 3212 }, { "epoch": 0.61, "learning_rate": 4.0817666121376785e-06, "loss": 0.4761, "step": 3213 }, { "epoch": 0.61, "learning_rate": 4.081174206872794e-06, "loss": 0.5169, "step": 3214 }, { "epoch": 0.61, "learning_rate": 4.080581653592224e-06, "loss": 0.5448, "step": 3215 }, { "epoch": 0.61, "learning_rate": 4.079988952351437e-06, "loss": 0.3479, "step": 3216 }, { "epoch": 0.61, "learning_rate": 4.0793961032059165e-06, "loss": 0.527, "step": 3217 }, { "epoch": 0.61, "learning_rate": 4.07880310621116e-06, "loss": 0.4237, "step": 3218 }, { "epoch": 0.61, "learning_rate": 4.078209961422679e-06, "loss": 0.8187, "step": 3219 }, { "epoch": 0.61, "learning_rate": 4.077616668895999e-06, "loss": 0.4246, "step": 3220 }, { "epoch": 0.61, "learning_rate": 4.0770232286866575e-06, "loss": 0.6409, "step": 3221 }, { "epoch": 0.61, "learning_rate": 4.076429640850209e-06, "loss": 0.4876, "step": 3222 }, { "epoch": 0.61, "learning_rate": 4.075835905442217e-06, "loss": 0.5405, "step": 3223 }, { "epoch": 0.61, "learning_rate": 4.075242022518265e-06, "loss": 0.5586, "step": 3224 }, { "epoch": 0.61, "learning_rate": 4.074647992133946e-06, "loss": 0.5521, "step": 3225 }, { "epoch": 0.61, "learning_rate": 4.074053814344868e-06, "loss": 0.4173, "step": 3226 }, { "epoch": 0.61, "learning_rate": 4.073459489206651e-06, "loss": 0.4441, "step": 3227 }, { "epoch": 0.61, "learning_rate": 4.072865016774933e-06, "loss": 0.4322, "step": 3228 }, { "epoch": 0.61, "learning_rate": 4.0722703971053625e-06, "loss": 0.6356, "step": 3229 }, { "epoch": 0.61, "learning_rate": 4.071675630253602e-06, "loss": 0.5097, "step": 3230 }, { "epoch": 0.61, "learning_rate": 4.0710807162753295e-06, "loss": 0.5225, "step": 3231 }, { "epoch": 0.61, "learning_rate": 4.070485655226234e-06, "loss": 0.3679, "step": 3232 }, { "epoch": 0.61, "learning_rate": 4.069890447162021e-06, "loss": 0.4216, "step": 3233 }, { "epoch": 0.61, "learning_rate": 4.069295092138409e-06, "loss": 0.4705, "step": 3234 }, { "epoch": 0.61, "learning_rate": 4.068699590211129e-06, "loss": 0.511, "step": 3235 }, { "epoch": 0.61, "learning_rate": 4.068103941435927e-06, "loss": 0.5967, "step": 3236 }, { "epoch": 0.61, "learning_rate": 4.067508145868563e-06, "loss": 0.3946, "step": 3237 }, { "epoch": 0.61, "learning_rate": 4.066912203564809e-06, "loss": 0.4011, "step": 3238 }, { "epoch": 0.61, "learning_rate": 4.066316114580452e-06, "loss": 0.4859, "step": 3239 }, { "epoch": 0.61, "learning_rate": 4.065719878971295e-06, "loss": 0.5237, "step": 3240 }, { "epoch": 0.61, "learning_rate": 4.06512349679315e-06, "loss": 0.4758, "step": 3241 }, { "epoch": 0.61, "learning_rate": 4.064526968101844e-06, "loss": 0.5697, "step": 3242 }, { "epoch": 0.61, "learning_rate": 4.0639302929532224e-06, "loss": 0.4093, "step": 3243 }, { "epoch": 0.61, "learning_rate": 4.063333471403138e-06, "loss": 0.6195, "step": 3244 }, { "epoch": 0.61, "learning_rate": 4.062736503507461e-06, "loss": 0.5648, "step": 3245 }, { "epoch": 0.61, "learning_rate": 4.062139389322073e-06, "loss": 0.5413, "step": 3246 }, { "epoch": 0.61, "learning_rate": 4.061542128902872e-06, "loss": 0.5778, "step": 3247 }, { "epoch": 0.61, "learning_rate": 4.0609447223057685e-06, "loss": 0.4989, "step": 3248 }, { "epoch": 0.61, "learning_rate": 4.060347169586687e-06, "loss": 0.4347, "step": 3249 }, { "epoch": 0.61, "learning_rate": 4.059749470801562e-06, "loss": 0.3896, "step": 3250 }, { "epoch": 0.61, "learning_rate": 4.059151626006348e-06, "loss": 0.5601, "step": 3251 }, { "epoch": 0.61, "learning_rate": 4.0585536352570085e-06, "loss": 0.4531, "step": 3252 }, { "epoch": 0.61, "learning_rate": 4.0579554986095224e-06, "loss": 0.5671, "step": 3253 }, { "epoch": 0.61, "learning_rate": 4.057357216119882e-06, "loss": 0.4797, "step": 3254 }, { "epoch": 0.61, "learning_rate": 4.056758787844095e-06, "loss": 0.5693, "step": 3255 }, { "epoch": 0.62, "learning_rate": 4.056160213838179e-06, "loss": 0.4244, "step": 3256 }, { "epoch": 0.62, "learning_rate": 4.0555614941581665e-06, "loss": 0.4169, "step": 3257 }, { "epoch": 0.62, "learning_rate": 4.0549626288601066e-06, "loss": 0.5242, "step": 3258 }, { "epoch": 0.62, "learning_rate": 4.054363618000058e-06, "loss": 0.5437, "step": 3259 }, { "epoch": 0.62, "learning_rate": 4.053764461634096e-06, "loss": 0.5093, "step": 3260 }, { "epoch": 0.62, "learning_rate": 4.053165159818308e-06, "loss": 0.6473, "step": 3261 }, { "epoch": 0.62, "learning_rate": 4.052565712608796e-06, "loss": 0.5502, "step": 3262 }, { "epoch": 0.62, "learning_rate": 4.051966120061674e-06, "loss": 0.4536, "step": 3263 }, { "epoch": 0.62, "learning_rate": 4.051366382233071e-06, "loss": 0.4615, "step": 3264 }, { "epoch": 0.62, "learning_rate": 4.05076649917913e-06, "loss": 0.3317, "step": 3265 }, { "epoch": 0.62, "learning_rate": 4.050166470956005e-06, "loss": 0.4677, "step": 3266 }, { "epoch": 0.62, "learning_rate": 4.049566297619867e-06, "loss": 0.562, "step": 3267 }, { "epoch": 0.62, "learning_rate": 4.048965979226899e-06, "loss": 0.5876, "step": 3268 }, { "epoch": 0.62, "learning_rate": 4.048365515833295e-06, "loss": 0.5993, "step": 3269 }, { "epoch": 0.62, "learning_rate": 4.04776490749527e-06, "loss": 0.6211, "step": 3270 }, { "epoch": 0.62, "learning_rate": 4.047164154269042e-06, "loss": 0.4687, "step": 3271 }, { "epoch": 0.62, "learning_rate": 4.046563256210853e-06, "loss": 0.4792, "step": 3272 }, { "epoch": 0.62, "learning_rate": 4.045962213376952e-06, "loss": 0.48, "step": 3273 }, { "epoch": 0.62, "learning_rate": 4.045361025823601e-06, "loss": 0.5211, "step": 3274 }, { "epoch": 0.62, "learning_rate": 4.044759693607082e-06, "loss": 0.4649, "step": 3275 }, { "epoch": 0.62, "learning_rate": 4.044158216783684e-06, "loss": 0.5137, "step": 3276 }, { "epoch": 0.62, "learning_rate": 4.043556595409713e-06, "loss": 0.4459, "step": 3277 }, { "epoch": 0.62, "learning_rate": 4.042954829541487e-06, "loss": 0.4302, "step": 3278 }, { "epoch": 0.62, "learning_rate": 4.0423529192353376e-06, "loss": 0.5282, "step": 3279 }, { "epoch": 0.62, "learning_rate": 4.0417508645476115e-06, "loss": 0.6038, "step": 3280 }, { "epoch": 0.62, "learning_rate": 4.0411486655346675e-06, "loss": 0.4612, "step": 3281 }, { "epoch": 0.62, "learning_rate": 4.040546322252877e-06, "loss": 0.3628, "step": 3282 }, { "epoch": 0.62, "learning_rate": 4.039943834758629e-06, "loss": 0.3701, "step": 3283 }, { "epoch": 0.62, "learning_rate": 4.039341203108319e-06, "loss": 0.3945, "step": 3284 }, { "epoch": 0.62, "learning_rate": 4.038738427358363e-06, "loss": 0.537, "step": 3285 }, { "epoch": 0.62, "learning_rate": 4.038135507565187e-06, "loss": 0.5768, "step": 3286 }, { "epoch": 0.62, "learning_rate": 4.03753244378523e-06, "loss": 0.6423, "step": 3287 }, { "epoch": 0.62, "learning_rate": 4.036929236074948e-06, "loss": 0.7029, "step": 3288 }, { "epoch": 0.62, "learning_rate": 4.036325884490806e-06, "loss": 0.3937, "step": 3289 }, { "epoch": 0.62, "learning_rate": 4.035722389089284e-06, "loss": 0.4662, "step": 3290 }, { "epoch": 0.62, "learning_rate": 4.0351187499268775e-06, "loss": 0.482, "step": 3291 }, { "epoch": 0.62, "learning_rate": 4.034514967060094e-06, "loss": 0.4287, "step": 3292 }, { "epoch": 0.62, "learning_rate": 4.0339110405454535e-06, "loss": 0.6813, "step": 3293 }, { "epoch": 0.62, "learning_rate": 4.03330697043949e-06, "loss": 0.5456, "step": 3294 }, { "epoch": 0.62, "learning_rate": 4.032702756798752e-06, "loss": 0.3981, "step": 3295 }, { "epoch": 0.62, "learning_rate": 4.0320983996798e-06, "loss": 0.5682, "step": 3296 }, { "epoch": 0.62, "learning_rate": 4.03149389913921e-06, "loss": 0.4835, "step": 3297 }, { "epoch": 0.62, "learning_rate": 4.030889255233569e-06, "loss": 0.5663, "step": 3298 }, { "epoch": 0.62, "learning_rate": 4.0302844680194785e-06, "loss": 0.4962, "step": 3299 }, { "epoch": 0.62, "learning_rate": 4.029679537553553e-06, "loss": 0.4428, "step": 3300 }, { "epoch": 0.62, "learning_rate": 4.0290744638924215e-06, "loss": 0.5115, "step": 3301 }, { "epoch": 0.62, "learning_rate": 4.028469247092727e-06, "loss": 0.5395, "step": 3302 }, { "epoch": 0.62, "learning_rate": 4.027863887211121e-06, "loss": 0.5127, "step": 3303 }, { "epoch": 0.62, "learning_rate": 4.027258384304275e-06, "loss": 0.4811, "step": 3304 }, { "epoch": 0.62, "learning_rate": 4.02665273842887e-06, "loss": 0.4019, "step": 3305 }, { "epoch": 0.62, "learning_rate": 4.026046949641601e-06, "loss": 0.399, "step": 3306 }, { "epoch": 0.62, "learning_rate": 4.025441017999177e-06, "loss": 0.6379, "step": 3307 }, { "epoch": 0.62, "learning_rate": 4.024834943558319e-06, "loss": 0.4453, "step": 3308 }, { "epoch": 0.63, "learning_rate": 4.024228726375765e-06, "loss": 0.521, "step": 3309 }, { "epoch": 0.63, "learning_rate": 4.02362236650826e-06, "loss": 0.4234, "step": 3310 }, { "epoch": 0.63, "learning_rate": 4.023015864012569e-06, "loss": 0.4085, "step": 3311 }, { "epoch": 0.63, "learning_rate": 4.0224092189454675e-06, "loss": 0.4325, "step": 3312 }, { "epoch": 0.63, "learning_rate": 4.021802431363742e-06, "loss": 0.5913, "step": 3313 }, { "epoch": 0.63, "learning_rate": 4.021195501324198e-06, "loss": 0.554, "step": 3314 }, { "epoch": 0.63, "learning_rate": 4.020588428883647e-06, "loss": 0.4394, "step": 3315 }, { "epoch": 0.63, "learning_rate": 4.01998121409892e-06, "loss": 0.5727, "step": 3316 }, { "epoch": 0.63, "learning_rate": 4.01937385702686e-06, "loss": 0.3229, "step": 3317 }, { "epoch": 0.63, "learning_rate": 4.018766357724321e-06, "loss": 0.5054, "step": 3318 }, { "epoch": 0.63, "learning_rate": 4.018158716248172e-06, "loss": 0.5802, "step": 3319 }, { "epoch": 0.63, "learning_rate": 4.017550932655295e-06, "loss": 0.5288, "step": 3320 }, { "epoch": 0.63, "learning_rate": 4.0169430070025854e-06, "loss": 0.6935, "step": 3321 }, { "epoch": 0.63, "learning_rate": 4.016334939346953e-06, "loss": 0.4819, "step": 3322 }, { "epoch": 0.63, "learning_rate": 4.015726729745318e-06, "loss": 0.4818, "step": 3323 }, { "epoch": 0.63, "learning_rate": 4.015118378254617e-06, "loss": 0.7768, "step": 3324 }, { "epoch": 0.63, "learning_rate": 4.014509884931797e-06, "loss": 0.5216, "step": 3325 }, { "epoch": 0.63, "learning_rate": 4.013901249833822e-06, "loss": 0.5501, "step": 3326 }, { "epoch": 0.63, "learning_rate": 4.013292473017666e-06, "loss": 0.5415, "step": 3327 }, { "epoch": 0.63, "learning_rate": 4.012683554540316e-06, "loss": 0.4629, "step": 3328 }, { "epoch": 0.63, "learning_rate": 4.012074494458776e-06, "loss": 0.5136, "step": 3329 }, { "epoch": 0.63, "learning_rate": 4.0114652928300605e-06, "loss": 0.4811, "step": 3330 }, { "epoch": 0.63, "learning_rate": 4.010855949711196e-06, "loss": 0.5606, "step": 3331 }, { "epoch": 0.63, "learning_rate": 4.010246465159224e-06, "loss": 0.4781, "step": 3332 }, { "epoch": 0.63, "learning_rate": 4.009636839231201e-06, "loss": 0.4973, "step": 3333 }, { "epoch": 0.63, "learning_rate": 4.0090270719841935e-06, "loss": 0.501, "step": 3334 }, { "epoch": 0.63, "learning_rate": 4.008417163475283e-06, "loss": 0.5276, "step": 3335 }, { "epoch": 0.63, "learning_rate": 4.007807113761564e-06, "loss": 0.5059, "step": 3336 }, { "epoch": 0.63, "learning_rate": 4.007196922900143e-06, "loss": 0.4889, "step": 3337 }, { "epoch": 0.63, "learning_rate": 4.006586590948141e-06, "loss": 0.5545, "step": 3338 }, { "epoch": 0.63, "learning_rate": 4.0059761179626935e-06, "loss": 0.5634, "step": 3339 }, { "epoch": 0.63, "learning_rate": 4.0053655040009445e-06, "loss": 0.4269, "step": 3340 }, { "epoch": 0.63, "learning_rate": 4.004754749120058e-06, "loss": 0.4946, "step": 3341 }, { "epoch": 0.63, "learning_rate": 4.004143853377205e-06, "loss": 0.411, "step": 3342 }, { "epoch": 0.63, "learning_rate": 4.0035328168295725e-06, "loss": 0.4353, "step": 3343 }, { "epoch": 0.63, "learning_rate": 4.002921639534362e-06, "loss": 0.4, "step": 3344 }, { "epoch": 0.63, "learning_rate": 4.002310321548785e-06, "loss": 0.4588, "step": 3345 }, { "epoch": 0.63, "learning_rate": 4.001698862930068e-06, "loss": 0.537, "step": 3346 }, { "epoch": 0.63, "learning_rate": 4.0010872637354505e-06, "loss": 0.4785, "step": 3347 }, { "epoch": 0.63, "learning_rate": 4.000475524022186e-06, "loss": 0.4803, "step": 3348 }, { "epoch": 0.63, "learning_rate": 3.999863643847539e-06, "loss": 0.5345, "step": 3349 }, { "epoch": 0.63, "learning_rate": 3.9992516232687885e-06, "loss": 0.5281, "step": 3350 }, { "epoch": 0.63, "learning_rate": 3.9986394623432266e-06, "loss": 0.5249, "step": 3351 }, { "epoch": 0.63, "learning_rate": 3.998027161128159e-06, "loss": 0.6033, "step": 3352 }, { "epoch": 0.63, "learning_rate": 3.9974147196809024e-06, "loss": 0.5024, "step": 3353 }, { "epoch": 0.63, "learning_rate": 3.99680213805879e-06, "loss": 0.5686, "step": 3354 }, { "epoch": 0.63, "learning_rate": 3.996189416319166e-06, "loss": 0.4974, "step": 3355 }, { "epoch": 0.63, "learning_rate": 3.995576554519388e-06, "loss": 0.4906, "step": 3356 }, { "epoch": 0.63, "learning_rate": 3.994963552716826e-06, "loss": 0.5218, "step": 3357 }, { "epoch": 0.63, "learning_rate": 3.994350410968865e-06, "loss": 0.4957, "step": 3358 }, { "epoch": 0.63, "learning_rate": 3.993737129332901e-06, "loss": 0.4223, "step": 3359 }, { "epoch": 0.63, "learning_rate": 3.993123707866344e-06, "loss": 0.4078, "step": 3360 }, { "epoch": 0.63, "learning_rate": 3.992510146626618e-06, "loss": 0.5662, "step": 3361 }, { "epoch": 0.64, "learning_rate": 3.991896445671159e-06, "loss": 0.4995, "step": 3362 }, { "epoch": 0.64, "learning_rate": 3.991282605057415e-06, "loss": 0.4902, "step": 3363 }, { "epoch": 0.64, "learning_rate": 3.99066862484285e-06, "loss": 0.5011, "step": 3364 }, { "epoch": 0.64, "learning_rate": 3.9900545050849385e-06, "loss": 0.449, "step": 3365 }, { "epoch": 0.64, "learning_rate": 3.98944024584117e-06, "loss": 0.4842, "step": 3366 }, { "epoch": 0.64, "learning_rate": 3.988825847169046e-06, "loss": 0.4598, "step": 3367 }, { "epoch": 0.64, "learning_rate": 3.98821130912608e-06, "loss": 0.6128, "step": 3368 }, { "epoch": 0.64, "learning_rate": 3.9875966317698e-06, "loss": 0.4768, "step": 3369 }, { "epoch": 0.64, "learning_rate": 3.986981815157746e-06, "loss": 0.4567, "step": 3370 }, { "epoch": 0.64, "learning_rate": 3.986366859347475e-06, "loss": 0.5913, "step": 3371 }, { "epoch": 0.64, "learning_rate": 3.98575176439655e-06, "loss": 0.3932, "step": 3372 }, { "epoch": 0.64, "learning_rate": 3.985136530362553e-06, "loss": 0.433, "step": 3373 }, { "epoch": 0.64, "learning_rate": 3.984521157303077e-06, "loss": 0.5047, "step": 3374 }, { "epoch": 0.64, "learning_rate": 3.9839056452757255e-06, "loss": 0.4365, "step": 3375 }, { "epoch": 0.64, "learning_rate": 3.98328999433812e-06, "loss": 0.5209, "step": 3376 }, { "epoch": 0.64, "learning_rate": 3.982674204547891e-06, "loss": 0.4478, "step": 3377 }, { "epoch": 0.64, "learning_rate": 3.9820582759626825e-06, "loss": 0.4129, "step": 3378 }, { "epoch": 0.64, "learning_rate": 3.981442208640154e-06, "loss": 0.4422, "step": 3379 }, { "epoch": 0.64, "learning_rate": 3.980826002637976e-06, "loss": 0.4684, "step": 3380 }, { "epoch": 0.64, "learning_rate": 3.980209658013832e-06, "loss": 0.6738, "step": 3381 }, { "epoch": 0.64, "learning_rate": 3.979593174825418e-06, "loss": 0.3865, "step": 3382 }, { "epoch": 0.64, "learning_rate": 3.978976553130445e-06, "loss": 0.6065, "step": 3383 }, { "epoch": 0.64, "learning_rate": 3.9783597929866356e-06, "loss": 0.555, "step": 3384 }, { "epoch": 0.64, "learning_rate": 3.977742894451725e-06, "loss": 0.5179, "step": 3385 }, { "epoch": 0.64, "learning_rate": 3.977125857583462e-06, "loss": 0.5298, "step": 3386 }, { "epoch": 0.64, "learning_rate": 3.976508682439608e-06, "loss": 0.4566, "step": 3387 }, { "epoch": 0.64, "learning_rate": 3.975891369077939e-06, "loss": 0.4812, "step": 3388 }, { "epoch": 0.64, "learning_rate": 3.97527391755624e-06, "loss": 0.4699, "step": 3389 }, { "epoch": 0.64, "learning_rate": 3.974656327932312e-06, "loss": 0.5835, "step": 3390 }, { "epoch": 0.64, "learning_rate": 3.974038600263971e-06, "loss": 0.5987, "step": 3391 }, { "epoch": 0.64, "learning_rate": 3.9734207346090395e-06, "loss": 0.4913, "step": 3392 }, { "epoch": 0.64, "learning_rate": 3.9728027310253595e-06, "loss": 0.5086, "step": 3393 }, { "epoch": 0.64, "learning_rate": 3.972184589570781e-06, "loss": 0.5644, "step": 3394 }, { "epoch": 0.64, "learning_rate": 3.971566310303171e-06, "loss": 0.4594, "step": 3395 }, { "epoch": 0.64, "learning_rate": 3.9709478932804055e-06, "loss": 0.4882, "step": 3396 }, { "epoch": 0.64, "learning_rate": 3.970329338560376e-06, "loss": 0.5955, "step": 3397 }, { "epoch": 0.64, "learning_rate": 3.9697106462009875e-06, "loss": 0.4999, "step": 3398 }, { "epoch": 0.64, "learning_rate": 3.969091816260154e-06, "loss": 0.5577, "step": 3399 }, { "epoch": 0.64, "learning_rate": 3.968472848795807e-06, "loss": 0.6169, "step": 3400 }, { "epoch": 0.64, "learning_rate": 3.967853743865888e-06, "loss": 0.3944, "step": 3401 }, { "epoch": 0.64, "learning_rate": 3.967234501528352e-06, "loss": 0.6395, "step": 3402 }, { "epoch": 0.64, "learning_rate": 3.9666151218411685e-06, "loss": 0.5052, "step": 3403 }, { "epoch": 0.64, "learning_rate": 3.965995604862316e-06, "loss": 0.5106, "step": 3404 }, { "epoch": 0.64, "learning_rate": 3.96537595064979e-06, "loss": 0.464, "step": 3405 }, { "epoch": 0.64, "learning_rate": 3.964756159261597e-06, "loss": 0.4881, "step": 3406 }, { "epoch": 0.64, "learning_rate": 3.964136230755756e-06, "loss": 0.5975, "step": 3407 }, { "epoch": 0.64, "learning_rate": 3.963516165190299e-06, "loss": 0.4586, "step": 3408 }, { "epoch": 0.64, "learning_rate": 3.9628959626232724e-06, "loss": 0.495, "step": 3409 }, { "epoch": 0.64, "learning_rate": 3.962275623112731e-06, "loss": 0.458, "step": 3410 }, { "epoch": 0.64, "learning_rate": 3.961655146716749e-06, "loss": 0.5164, "step": 3411 }, { "epoch": 0.64, "learning_rate": 3.96103453349341e-06, "loss": 0.4544, "step": 3412 }, { "epoch": 0.64, "learning_rate": 3.960413783500807e-06, "loss": 0.5199, "step": 3413 }, { "epoch": 0.65, "learning_rate": 3.959792896797053e-06, "loss": 0.427, "step": 3414 }, { "epoch": 0.65, "learning_rate": 3.9591718734402675e-06, "loss": 0.5408, "step": 3415 }, { "epoch": 0.65, "learning_rate": 3.9585507134885855e-06, "loss": 0.5567, "step": 3416 }, { "epoch": 0.65, "learning_rate": 3.957929417000156e-06, "loss": 0.3911, "step": 3417 }, { "epoch": 0.65, "learning_rate": 3.957307984033139e-06, "loss": 0.4581, "step": 3418 }, { "epoch": 0.65, "learning_rate": 3.956686414645705e-06, "loss": 0.5853, "step": 3419 }, { "epoch": 0.65, "learning_rate": 3.9560647088960435e-06, "loss": 0.5144, "step": 3420 }, { "epoch": 0.65, "learning_rate": 3.95544286684235e-06, "loss": 0.5359, "step": 3421 }, { "epoch": 0.65, "learning_rate": 3.954820888542839e-06, "loss": 0.5206, "step": 3422 }, { "epoch": 0.65, "learning_rate": 3.954198774055733e-06, "loss": 0.5207, "step": 3423 }, { "epoch": 0.65, "learning_rate": 3.953576523439269e-06, "loss": 0.5283, "step": 3424 }, { "epoch": 0.65, "learning_rate": 3.952954136751696e-06, "loss": 0.4841, "step": 3425 }, { "epoch": 0.65, "learning_rate": 3.952331614051278e-06, "loss": 0.4542, "step": 3426 }, { "epoch": 0.65, "learning_rate": 3.951708955396289e-06, "loss": 0.4471, "step": 3427 }, { "epoch": 0.65, "learning_rate": 3.951086160845016e-06, "loss": 0.5844, "step": 3428 }, { "epoch": 0.65, "learning_rate": 3.950463230455762e-06, "loss": 0.4136, "step": 3429 }, { "epoch": 0.65, "learning_rate": 3.949840164286839e-06, "loss": 0.5425, "step": 3430 }, { "epoch": 0.65, "learning_rate": 3.949216962396571e-06, "loss": 0.5378, "step": 3431 }, { "epoch": 0.65, "learning_rate": 3.9485936248433004e-06, "loss": 0.6226, "step": 3432 }, { "epoch": 0.65, "learning_rate": 3.947970151685377e-06, "loss": 0.543, "step": 3433 }, { "epoch": 0.65, "learning_rate": 3.947346542981164e-06, "loss": 0.6051, "step": 3434 }, { "epoch": 0.65, "learning_rate": 3.946722798789039e-06, "loss": 0.5112, "step": 3435 }, { "epoch": 0.65, "learning_rate": 3.946098919167392e-06, "loss": 0.458, "step": 3436 }, { "epoch": 0.65, "learning_rate": 3.945474904174623e-06, "loss": 0.4575, "step": 3437 }, { "epoch": 0.65, "learning_rate": 3.94485075386915e-06, "loss": 0.5204, "step": 3438 }, { "epoch": 0.65, "learning_rate": 3.944226468309398e-06, "loss": 0.6119, "step": 3439 }, { "epoch": 0.65, "learning_rate": 3.943602047553808e-06, "loss": 0.568, "step": 3440 }, { "epoch": 0.65, "learning_rate": 3.942977491660834e-06, "loss": 0.495, "step": 3441 }, { "epoch": 0.65, "learning_rate": 3.942352800688939e-06, "loss": 0.5027, "step": 3442 }, { "epoch": 0.65, "learning_rate": 3.941727974696603e-06, "loss": 0.5061, "step": 3443 }, { "epoch": 0.65, "learning_rate": 3.941103013742316e-06, "loss": 0.4276, "step": 3444 }, { "epoch": 0.65, "learning_rate": 3.940477917884581e-06, "loss": 0.4188, "step": 3445 }, { "epoch": 0.65, "learning_rate": 3.939852687181916e-06, "loss": 0.3866, "step": 3446 }, { "epoch": 0.65, "learning_rate": 3.9392273216928465e-06, "loss": 0.5524, "step": 3447 }, { "epoch": 0.65, "learning_rate": 3.938601821475914e-06, "loss": 0.409, "step": 3448 }, { "epoch": 0.65, "learning_rate": 3.937976186589677e-06, "loss": 0.5888, "step": 3449 }, { "epoch": 0.65, "learning_rate": 3.9373504170926965e-06, "loss": 0.5512, "step": 3450 }, { "epoch": 0.65, "learning_rate": 3.936724513043555e-06, "loss": 0.4218, "step": 3451 }, { "epoch": 0.65, "learning_rate": 3.936098474500843e-06, "loss": 0.5394, "step": 3452 }, { "epoch": 0.65, "learning_rate": 3.935472301523163e-06, "loss": 0.3841, "step": 3453 }, { "epoch": 0.65, "learning_rate": 3.9348459941691345e-06, "loss": 0.5405, "step": 3454 }, { "epoch": 0.65, "learning_rate": 3.9342195524973855e-06, "loss": 0.517, "step": 3455 }, { "epoch": 0.65, "learning_rate": 3.933592976566559e-06, "loss": 0.4945, "step": 3456 }, { "epoch": 0.65, "learning_rate": 3.9329662664353085e-06, "loss": 0.4096, "step": 3457 }, { "epoch": 0.65, "learning_rate": 3.932339422162301e-06, "loss": 0.5969, "step": 3458 }, { "epoch": 0.65, "learning_rate": 3.931712443806217e-06, "loss": 0.402, "step": 3459 }, { "epoch": 0.65, "learning_rate": 3.931085331425749e-06, "loss": 0.6669, "step": 3460 }, { "epoch": 0.65, "learning_rate": 3.930458085079602e-06, "loss": 0.5257, "step": 3461 }, { "epoch": 0.65, "learning_rate": 3.92983070482649e-06, "loss": 0.4104, "step": 3462 }, { "epoch": 0.65, "learning_rate": 3.929203190725147e-06, "loss": 0.4753, "step": 3463 }, { "epoch": 0.65, "learning_rate": 3.9285755428343134e-06, "loss": 0.4355, "step": 3464 }, { "epoch": 0.65, "learning_rate": 3.927947761212745e-06, "loss": 0.4882, "step": 3465 }, { "epoch": 0.65, "learning_rate": 3.927319845919208e-06, "loss": 0.4092, "step": 3466 }, { "epoch": 0.66, "learning_rate": 3.9266917970124825e-06, "loss": 0.5154, "step": 3467 }, { "epoch": 0.66, "learning_rate": 3.926063614551362e-06, "loss": 0.4355, "step": 3468 }, { "epoch": 0.66, "learning_rate": 3.925435298594651e-06, "loss": 0.4106, "step": 3469 }, { "epoch": 0.66, "learning_rate": 3.924806849201167e-06, "loss": 0.4235, "step": 3470 }, { "epoch": 0.66, "learning_rate": 3.92417826642974e-06, "loss": 0.5661, "step": 3471 }, { "epoch": 0.66, "learning_rate": 3.923549550339211e-06, "loss": 0.5672, "step": 3472 }, { "epoch": 0.66, "learning_rate": 3.922920700988438e-06, "loss": 0.4857, "step": 3473 }, { "epoch": 0.66, "learning_rate": 3.922291718436283e-06, "loss": 0.496, "step": 3474 }, { "epoch": 0.66, "learning_rate": 3.921662602741632e-06, "loss": 0.546, "step": 3475 }, { "epoch": 0.66, "learning_rate": 3.921033353963374e-06, "loss": 0.4294, "step": 3476 }, { "epoch": 0.66, "learning_rate": 3.920403972160414e-06, "loss": 0.478, "step": 3477 }, { "epoch": 0.66, "learning_rate": 3.91977445739167e-06, "loss": 0.4463, "step": 3478 }, { "epoch": 0.66, "learning_rate": 3.919144809716071e-06, "loss": 0.5204, "step": 3479 }, { "epoch": 0.66, "learning_rate": 3.918515029192559e-06, "loss": 0.5141, "step": 3480 }, { "epoch": 0.66, "learning_rate": 3.917885115880089e-06, "loss": 0.4275, "step": 3481 }, { "epoch": 0.66, "learning_rate": 3.9172550698376285e-06, "loss": 0.5475, "step": 3482 }, { "epoch": 0.66, "learning_rate": 3.916624891124155e-06, "loss": 0.5245, "step": 3483 }, { "epoch": 0.66, "learning_rate": 3.915994579798662e-06, "loss": 0.492, "step": 3484 }, { "epoch": 0.66, "learning_rate": 3.915364135920153e-06, "loss": 0.3695, "step": 3485 }, { "epoch": 0.66, "learning_rate": 3.914733559547645e-06, "loss": 0.53, "step": 3486 }, { "epoch": 0.66, "learning_rate": 3.914102850740168e-06, "loss": 0.5871, "step": 3487 }, { "epoch": 0.66, "learning_rate": 3.913472009556761e-06, "loss": 0.5191, "step": 3488 }, { "epoch": 0.66, "learning_rate": 3.91284103605648e-06, "loss": 0.5642, "step": 3489 }, { "epoch": 0.66, "learning_rate": 3.91220993029839e-06, "loss": 0.466, "step": 3490 }, { "epoch": 0.66, "learning_rate": 3.9115786923415704e-06, "loss": 0.394, "step": 3491 }, { "epoch": 0.66, "learning_rate": 3.9109473222451115e-06, "loss": 0.5788, "step": 3492 }, { "epoch": 0.66, "learning_rate": 3.9103158200681166e-06, "loss": 0.3981, "step": 3493 }, { "epoch": 0.66, "learning_rate": 3.909684185869702e-06, "loss": 0.3444, "step": 3494 }, { "epoch": 0.66, "learning_rate": 3.909052419708995e-06, "loss": 0.5832, "step": 3495 }, { "epoch": 0.66, "learning_rate": 3.9084205216451375e-06, "loss": 0.5162, "step": 3496 }, { "epoch": 0.66, "learning_rate": 3.907788491737281e-06, "loss": 0.3599, "step": 3497 }, { "epoch": 0.66, "learning_rate": 3.907156330044591e-06, "loss": 0.527, "step": 3498 }, { "epoch": 0.66, "learning_rate": 3.906524036626244e-06, "loss": 0.5012, "step": 3499 }, { "epoch": 0.66, "learning_rate": 3.9058916115414306e-06, "loss": 0.6112, "step": 3500 }, { "epoch": 0.66, "learning_rate": 3.905259054849353e-06, "loss": 0.6717, "step": 3501 }, { "epoch": 0.66, "learning_rate": 3.904626366609227e-06, "loss": 0.4054, "step": 3502 }, { "epoch": 0.66, "learning_rate": 3.903993546880276e-06, "loss": 0.449, "step": 3503 }, { "epoch": 0.66, "learning_rate": 3.903360595721742e-06, "loss": 0.6155, "step": 3504 }, { "epoch": 0.66, "learning_rate": 3.9027275131928746e-06, "loss": 0.6078, "step": 3505 }, { "epoch": 0.66, "learning_rate": 3.9020942993529375e-06, "loss": 0.4827, "step": 3506 }, { "epoch": 0.66, "learning_rate": 3.901460954261208e-06, "loss": 0.4783, "step": 3507 }, { "epoch": 0.66, "learning_rate": 3.900827477976974e-06, "loss": 0.4638, "step": 3508 }, { "epoch": 0.66, "learning_rate": 3.900193870559534e-06, "loss": 0.5032, "step": 3509 }, { "epoch": 0.66, "learning_rate": 3.899560132068204e-06, "loss": 0.5911, "step": 3510 }, { "epoch": 0.66, "learning_rate": 3.898926262562306e-06, "loss": 0.4261, "step": 3511 }, { "epoch": 0.66, "learning_rate": 3.8982922621011795e-06, "loss": 0.4723, "step": 3512 }, { "epoch": 0.66, "learning_rate": 3.897658130744173e-06, "loss": 0.4099, "step": 3513 }, { "epoch": 0.66, "learning_rate": 3.897023868550649e-06, "loss": 0.4711, "step": 3514 }, { "epoch": 0.66, "learning_rate": 3.89638947557998e-06, "loss": 0.5143, "step": 3515 }, { "epoch": 0.66, "learning_rate": 3.895754951891555e-06, "loss": 0.354, "step": 3516 }, { "epoch": 0.66, "learning_rate": 3.89512029754477e-06, "loss": 0.4438, "step": 3517 }, { "epoch": 0.66, "learning_rate": 3.894485512599037e-06, "loss": 0.5313, "step": 3518 }, { "epoch": 0.66, "learning_rate": 3.893850597113779e-06, "loss": 0.4543, "step": 3519 }, { "epoch": 0.67, "learning_rate": 3.89321555114843e-06, "loss": 0.5152, "step": 3520 }, { "epoch": 0.67, "learning_rate": 3.892580374762441e-06, "loss": 0.3934, "step": 3521 }, { "epoch": 0.67, "learning_rate": 3.891945068015267e-06, "loss": 0.6039, "step": 3522 }, { "epoch": 0.67, "learning_rate": 3.891309630966382e-06, "loss": 0.414, "step": 3523 }, { "epoch": 0.67, "learning_rate": 3.89067406367527e-06, "loss": 0.4941, "step": 3524 }, { "epoch": 0.67, "learning_rate": 3.890038366201428e-06, "loss": 0.4171, "step": 3525 }, { "epoch": 0.67, "learning_rate": 3.889402538604364e-06, "loss": 0.4544, "step": 3526 }, { "epoch": 0.67, "learning_rate": 3.888766580943598e-06, "loss": 0.4972, "step": 3527 }, { "epoch": 0.67, "learning_rate": 3.8881304932786625e-06, "loss": 0.545, "step": 3528 }, { "epoch": 0.67, "learning_rate": 3.887494275669104e-06, "loss": 0.5057, "step": 3529 }, { "epoch": 0.67, "learning_rate": 3.886857928174478e-06, "loss": 0.3656, "step": 3530 }, { "epoch": 0.67, "learning_rate": 3.886221450854355e-06, "loss": 0.5635, "step": 3531 }, { "epoch": 0.67, "learning_rate": 3.8855848437683155e-06, "loss": 0.5395, "step": 3532 }, { "epoch": 0.67, "learning_rate": 3.8849481069759536e-06, "loss": 0.4088, "step": 3533 }, { "epoch": 0.67, "learning_rate": 3.884311240536875e-06, "loss": 0.4811, "step": 3534 }, { "epoch": 0.67, "learning_rate": 3.883674244510698e-06, "loss": 0.4379, "step": 3535 }, { "epoch": 0.67, "learning_rate": 3.883037118957052e-06, "loss": 0.5384, "step": 3536 }, { "epoch": 0.67, "learning_rate": 3.882399863935578e-06, "loss": 0.5907, "step": 3537 }, { "epoch": 0.67, "learning_rate": 3.8817624795059326e-06, "loss": 0.6029, "step": 3538 }, { "epoch": 0.67, "learning_rate": 3.881124965727781e-06, "loss": 0.5372, "step": 3539 }, { "epoch": 0.67, "learning_rate": 3.880487322660801e-06, "loss": 0.4274, "step": 3540 }, { "epoch": 0.67, "learning_rate": 3.879849550364684e-06, "loss": 0.6466, "step": 3541 }, { "epoch": 0.67, "learning_rate": 3.879211648899131e-06, "loss": 0.5082, "step": 3542 }, { "epoch": 0.67, "learning_rate": 3.878573618323861e-06, "loss": 0.4594, "step": 3543 }, { "epoch": 0.67, "learning_rate": 3.877935458698597e-06, "loss": 0.4253, "step": 3544 }, { "epoch": 0.67, "learning_rate": 3.877297170083078e-06, "loss": 0.5409, "step": 3545 }, { "epoch": 0.67, "learning_rate": 3.876658752537057e-06, "loss": 0.5158, "step": 3546 }, { "epoch": 0.67, "learning_rate": 3.876020206120295e-06, "loss": 0.3711, "step": 3547 }, { "epoch": 0.67, "learning_rate": 3.875381530892569e-06, "loss": 0.5441, "step": 3548 }, { "epoch": 0.67, "learning_rate": 3.874742726913665e-06, "loss": 0.4715, "step": 3549 }, { "epoch": 0.67, "learning_rate": 3.874103794243383e-06, "loss": 0.3955, "step": 3550 }, { "epoch": 0.67, "learning_rate": 3.873464732941533e-06, "loss": 0.4892, "step": 3551 }, { "epoch": 0.67, "learning_rate": 3.872825543067941e-06, "loss": 0.3769, "step": 3552 }, { "epoch": 0.67, "learning_rate": 3.872186224682438e-06, "loss": 0.5235, "step": 3553 }, { "epoch": 0.67, "learning_rate": 3.871546777844875e-06, "loss": 0.6212, "step": 3554 }, { "epoch": 0.67, "learning_rate": 3.870907202615111e-06, "loss": 0.6153, "step": 3555 }, { "epoch": 0.67, "learning_rate": 3.870267499053016e-06, "loss": 0.5141, "step": 3556 }, { "epoch": 0.67, "learning_rate": 3.869627667218474e-06, "loss": 0.4284, "step": 3557 }, { "epoch": 0.67, "learning_rate": 3.868987707171382e-06, "loss": 0.4268, "step": 3558 }, { "epoch": 0.67, "learning_rate": 3.8683476189716455e-06, "loss": 0.4535, "step": 3559 }, { "epoch": 0.67, "learning_rate": 3.867707402679184e-06, "loss": 0.413, "step": 3560 }, { "epoch": 0.67, "learning_rate": 3.86706705835393e-06, "loss": 0.5191, "step": 3561 }, { "epoch": 0.67, "learning_rate": 3.866426586055826e-06, "loss": 0.6053, "step": 3562 }, { "epoch": 0.67, "learning_rate": 3.865785985844828e-06, "loss": 0.4602, "step": 3563 }, { "epoch": 0.67, "learning_rate": 3.865145257780903e-06, "loss": 0.4103, "step": 3564 }, { "epoch": 0.67, "learning_rate": 3.864504401924031e-06, "loss": 0.5883, "step": 3565 }, { "epoch": 0.67, "learning_rate": 3.863863418334203e-06, "loss": 0.4641, "step": 3566 }, { "epoch": 0.67, "learning_rate": 3.863222307071422e-06, "loss": 0.6714, "step": 3567 }, { "epoch": 0.67, "learning_rate": 3.862581068195703e-06, "loss": 0.4736, "step": 3568 }, { "epoch": 0.67, "learning_rate": 3.861939701767075e-06, "loss": 0.4809, "step": 3569 }, { "epoch": 0.67, "learning_rate": 3.861298207845574e-06, "loss": 0.4963, "step": 3570 }, { "epoch": 0.67, "learning_rate": 3.860656586491254e-06, "loss": 0.5065, "step": 3571 }, { "epoch": 0.67, "learning_rate": 3.860014837764176e-06, "loss": 0.6444, "step": 3572 }, { "epoch": 0.68, "learning_rate": 3.859372961724416e-06, "loss": 0.5398, "step": 3573 }, { "epoch": 0.68, "learning_rate": 3.8587309584320595e-06, "loss": 0.595, "step": 3574 }, { "epoch": 0.68, "learning_rate": 3.858088827947207e-06, "loss": 0.3937, "step": 3575 }, { "epoch": 0.68, "learning_rate": 3.857446570329968e-06, "loss": 0.4408, "step": 3576 }, { "epoch": 0.68, "learning_rate": 3.856804185640466e-06, "loss": 0.4415, "step": 3577 }, { "epoch": 0.68, "learning_rate": 3.856161673938834e-06, "loss": 0.6336, "step": 3578 }, { "epoch": 0.68, "learning_rate": 3.855519035285219e-06, "loss": 0.4872, "step": 3579 }, { "epoch": 0.68, "learning_rate": 3.8548762697397794e-06, "loss": 0.4747, "step": 3580 }, { "epoch": 0.68, "learning_rate": 3.854233377362686e-06, "loss": 0.5743, "step": 3581 }, { "epoch": 0.68, "learning_rate": 3.853590358214119e-06, "loss": 0.5509, "step": 3582 }, { "epoch": 0.68, "learning_rate": 3.852947212354273e-06, "loss": 0.4911, "step": 3583 }, { "epoch": 0.68, "learning_rate": 3.852303939843354e-06, "loss": 0.4455, "step": 3584 }, { "epoch": 0.68, "learning_rate": 3.851660540741579e-06, "loss": 0.498, "step": 3585 }, { "epoch": 0.68, "learning_rate": 3.851017015109179e-06, "loss": 0.3853, "step": 3586 }, { "epoch": 0.68, "learning_rate": 3.850373363006393e-06, "loss": 0.482, "step": 3587 }, { "epoch": 0.68, "learning_rate": 3.849729584493475e-06, "loss": 0.3982, "step": 3588 }, { "epoch": 0.68, "learning_rate": 3.84908567963069e-06, "loss": 0.4496, "step": 3589 }, { "epoch": 0.68, "learning_rate": 3.848441648478315e-06, "loss": 0.5292, "step": 3590 }, { "epoch": 0.68, "learning_rate": 3.847797491096639e-06, "loss": 0.5362, "step": 3591 }, { "epoch": 0.68, "learning_rate": 3.8471532075459606e-06, "loss": 0.5388, "step": 3592 }, { "epoch": 0.68, "learning_rate": 3.846508797886592e-06, "loss": 0.5016, "step": 3593 }, { "epoch": 0.68, "learning_rate": 3.84586426217886e-06, "loss": 0.4103, "step": 3594 }, { "epoch": 0.68, "learning_rate": 3.8452196004830985e-06, "loss": 0.5083, "step": 3595 }, { "epoch": 0.68, "learning_rate": 3.844574812859655e-06, "loss": 0.6023, "step": 3596 }, { "epoch": 0.68, "learning_rate": 3.843929899368888e-06, "loss": 0.3777, "step": 3597 }, { "epoch": 0.68, "learning_rate": 3.843284860071172e-06, "loss": 0.4689, "step": 3598 }, { "epoch": 0.68, "learning_rate": 3.842639695026885e-06, "loss": 0.4917, "step": 3599 }, { "epoch": 0.68, "learning_rate": 3.841994404296425e-06, "loss": 0.467, "step": 3600 }, { "epoch": 0.68, "learning_rate": 3.841348987940199e-06, "loss": 0.5019, "step": 3601 }, { "epoch": 0.68, "learning_rate": 3.840703446018623e-06, "loss": 0.4339, "step": 3602 }, { "epoch": 0.68, "learning_rate": 3.840057778592128e-06, "loss": 0.3614, "step": 3603 }, { "epoch": 0.68, "learning_rate": 3.839411985721157e-06, "loss": 0.3667, "step": 3604 }, { "epoch": 0.68, "learning_rate": 3.838766067466161e-06, "loss": 0.4659, "step": 3605 }, { "epoch": 0.68, "learning_rate": 3.838120023887607e-06, "loss": 0.5261, "step": 3606 }, { "epoch": 0.68, "learning_rate": 3.837473855045971e-06, "loss": 0.6178, "step": 3607 }, { "epoch": 0.68, "learning_rate": 3.836827561001742e-06, "loss": 0.6263, "step": 3608 }, { "epoch": 0.68, "learning_rate": 3.8361811418154216e-06, "loss": 0.4746, "step": 3609 }, { "epoch": 0.68, "learning_rate": 3.8355345975475195e-06, "loss": 0.4913, "step": 3610 }, { "epoch": 0.68, "learning_rate": 3.834887928258561e-06, "loss": 0.5184, "step": 3611 }, { "epoch": 0.68, "learning_rate": 3.834241134009082e-06, "loss": 0.3714, "step": 3612 }, { "epoch": 0.68, "learning_rate": 3.833594214859629e-06, "loss": 0.466, "step": 3613 }, { "epoch": 0.68, "learning_rate": 3.83294717087076e-06, "loss": 0.463, "step": 3614 }, { "epoch": 0.68, "learning_rate": 3.832300002103048e-06, "loss": 0.4274, "step": 3615 }, { "epoch": 0.68, "learning_rate": 3.831652708617073e-06, "loss": 0.4868, "step": 3616 }, { "epoch": 0.68, "learning_rate": 3.831005290473429e-06, "loss": 0.4918, "step": 3617 }, { "epoch": 0.68, "learning_rate": 3.830357747732723e-06, "loss": 0.5325, "step": 3618 }, { "epoch": 0.68, "learning_rate": 3.829710080455573e-06, "loss": 0.499, "step": 3619 }, { "epoch": 0.68, "learning_rate": 3.829062288702605e-06, "loss": 0.5986, "step": 3620 }, { "epoch": 0.68, "learning_rate": 3.828414372534462e-06, "loss": 0.5038, "step": 3621 }, { "epoch": 0.68, "learning_rate": 3.827766332011796e-06, "loss": 0.5146, "step": 3622 }, { "epoch": 0.68, "learning_rate": 3.8271181671952704e-06, "loss": 0.531, "step": 3623 }, { "epoch": 0.68, "learning_rate": 3.82646987814556e-06, "loss": 0.5079, "step": 3624 }, { "epoch": 0.68, "learning_rate": 3.825821464923353e-06, "loss": 0.452, "step": 3625 }, { "epoch": 0.69, "learning_rate": 3.825172927589348e-06, "loss": 0.5314, "step": 3626 }, { "epoch": 0.69, "learning_rate": 3.8245242662042545e-06, "loss": 0.526, "step": 3627 }, { "epoch": 0.69, "learning_rate": 3.8238754808287946e-06, "loss": 0.5849, "step": 3628 }, { "epoch": 0.69, "learning_rate": 3.823226571523704e-06, "loss": 0.4627, "step": 3629 }, { "epoch": 0.69, "learning_rate": 3.8225775383497264e-06, "loss": 0.5428, "step": 3630 }, { "epoch": 0.69, "learning_rate": 3.8219283813676175e-06, "loss": 0.5055, "step": 3631 }, { "epoch": 0.69, "learning_rate": 3.821279100638148e-06, "loss": 0.4113, "step": 3632 }, { "epoch": 0.69, "learning_rate": 3.820629696222096e-06, "loss": 0.4631, "step": 3633 }, { "epoch": 0.69, "learning_rate": 3.819980168180254e-06, "loss": 0.4046, "step": 3634 }, { "epoch": 0.69, "learning_rate": 3.819330516573424e-06, "loss": 0.3622, "step": 3635 }, { "epoch": 0.69, "learning_rate": 3.818680741462423e-06, "loss": 0.5619, "step": 3636 }, { "epoch": 0.69, "learning_rate": 3.818030842908075e-06, "loss": 0.3934, "step": 3637 }, { "epoch": 0.69, "learning_rate": 3.8173808209712185e-06, "loss": 0.4354, "step": 3638 }, { "epoch": 0.69, "learning_rate": 3.8167306757127035e-06, "loss": 0.4786, "step": 3639 }, { "epoch": 0.69, "learning_rate": 3.81608040719339e-06, "loss": 0.3607, "step": 3640 }, { "epoch": 0.69, "learning_rate": 3.8154300154741505e-06, "loss": 0.4339, "step": 3641 }, { "epoch": 0.69, "learning_rate": 3.814779500615871e-06, "loss": 0.5075, "step": 3642 }, { "epoch": 0.69, "learning_rate": 3.8141288626794436e-06, "loss": 0.4127, "step": 3643 }, { "epoch": 0.69, "learning_rate": 3.8134781017257773e-06, "loss": 0.5981, "step": 3644 }, { "epoch": 0.69, "learning_rate": 3.812827217815791e-06, "loss": 0.6199, "step": 3645 }, { "epoch": 0.69, "learning_rate": 3.8121762110104137e-06, "loss": 0.4366, "step": 3646 }, { "epoch": 0.69, "learning_rate": 3.811525081370588e-06, "loss": 0.4034, "step": 3647 }, { "epoch": 0.69, "learning_rate": 3.8108738289572655e-06, "loss": 0.4457, "step": 3648 }, { "epoch": 0.69, "learning_rate": 3.810222453831412e-06, "loss": 0.5345, "step": 3649 }, { "epoch": 0.69, "learning_rate": 3.809570956054004e-06, "loss": 0.6217, "step": 3650 }, { "epoch": 0.69, "learning_rate": 3.8089193356860272e-06, "loss": 0.4465, "step": 3651 }, { "epoch": 0.69, "learning_rate": 3.8082675927884816e-06, "loss": 0.4667, "step": 3652 }, { "epoch": 0.69, "learning_rate": 3.807615727422378e-06, "loss": 0.4755, "step": 3653 }, { "epoch": 0.69, "learning_rate": 3.8069637396487375e-06, "loss": 0.5727, "step": 3654 }, { "epoch": 0.69, "learning_rate": 3.806311629528594e-06, "loss": 0.6573, "step": 3655 }, { "epoch": 0.69, "learning_rate": 3.805659397122993e-06, "loss": 0.4263, "step": 3656 }, { "epoch": 0.69, "learning_rate": 3.8050070424929896e-06, "loss": 0.418, "step": 3657 }, { "epoch": 0.69, "learning_rate": 3.8043545656996524e-06, "loss": 0.5211, "step": 3658 }, { "epoch": 0.69, "learning_rate": 3.80370196680406e-06, "loss": 0.5742, "step": 3659 }, { "epoch": 0.69, "learning_rate": 3.803049245867303e-06, "loss": 0.4985, "step": 3660 }, { "epoch": 0.69, "learning_rate": 3.8023964029504833e-06, "loss": 0.4835, "step": 3661 }, { "epoch": 0.69, "learning_rate": 3.8017434381147143e-06, "loss": 0.5239, "step": 3662 }, { "epoch": 0.69, "learning_rate": 3.8010903514211217e-06, "loss": 0.5974, "step": 3663 }, { "epoch": 0.69, "learning_rate": 3.8004371429308417e-06, "loss": 0.4201, "step": 3664 }, { "epoch": 0.69, "learning_rate": 3.7997838127050213e-06, "loss": 0.3853, "step": 3665 }, { "epoch": 0.69, "learning_rate": 3.799130360804819e-06, "loss": 0.4655, "step": 3666 }, { "epoch": 0.69, "learning_rate": 3.7984767872914076e-06, "loss": 0.4768, "step": 3667 }, { "epoch": 0.69, "learning_rate": 3.797823092225966e-06, "loss": 0.4673, "step": 3668 }, { "epoch": 0.69, "learning_rate": 3.7971692756696887e-06, "loss": 0.4677, "step": 3669 }, { "epoch": 0.69, "learning_rate": 3.796515337683781e-06, "loss": 0.435, "step": 3670 }, { "epoch": 0.69, "learning_rate": 3.795861278329458e-06, "loss": 0.529, "step": 3671 }, { "epoch": 0.69, "learning_rate": 3.795207097667947e-06, "loss": 0.5616, "step": 3672 }, { "epoch": 0.69, "learning_rate": 3.7945527957604865e-06, "loss": 0.3641, "step": 3673 }, { "epoch": 0.69, "learning_rate": 3.7938983726683274e-06, "loss": 0.5667, "step": 3674 }, { "epoch": 0.69, "learning_rate": 3.7932438284527308e-06, "loss": 0.3998, "step": 3675 }, { "epoch": 0.69, "learning_rate": 3.7925891631749683e-06, "loss": 0.5439, "step": 3676 }, { "epoch": 0.69, "learning_rate": 3.791934376896326e-06, "loss": 0.3821, "step": 3677 }, { "epoch": 0.69, "learning_rate": 3.7912794696780964e-06, "loss": 0.5005, "step": 3678 }, { "epoch": 0.7, "learning_rate": 3.7906244415815883e-06, "loss": 0.6246, "step": 3679 }, { "epoch": 0.7, "learning_rate": 3.78996929266812e-06, "loss": 0.5909, "step": 3680 }, { "epoch": 0.7, "learning_rate": 3.7893140229990196e-06, "loss": 0.3786, "step": 3681 }, { "epoch": 0.7, "learning_rate": 3.788658632635628e-06, "loss": 0.5555, "step": 3682 }, { "epoch": 0.7, "learning_rate": 3.788003121639297e-06, "loss": 0.4103, "step": 3683 }, { "epoch": 0.7, "learning_rate": 3.78734749007139e-06, "loss": 0.481, "step": 3684 }, { "epoch": 0.7, "learning_rate": 3.786691737993282e-06, "loss": 0.5266, "step": 3685 }, { "epoch": 0.7, "learning_rate": 3.7860358654663575e-06, "loss": 0.3438, "step": 3686 }, { "epoch": 0.7, "learning_rate": 3.785379872552015e-06, "loss": 0.5235, "step": 3687 }, { "epoch": 0.7, "learning_rate": 3.7847237593116614e-06, "loss": 0.4927, "step": 3688 }, { "epoch": 0.7, "learning_rate": 3.7840675258067177e-06, "loss": 0.6112, "step": 3689 }, { "epoch": 0.7, "learning_rate": 3.783411172098614e-06, "loss": 0.5387, "step": 3690 }, { "epoch": 0.7, "learning_rate": 3.7827546982487924e-06, "loss": 0.4224, "step": 3691 }, { "epoch": 0.7, "learning_rate": 3.7820981043187055e-06, "loss": 0.4982, "step": 3692 }, { "epoch": 0.7, "learning_rate": 3.78144139036982e-06, "loss": 0.4079, "step": 3693 }, { "epoch": 0.7, "learning_rate": 3.7807845564636093e-06, "loss": 0.4154, "step": 3694 }, { "epoch": 0.7, "learning_rate": 3.780127602661562e-06, "loss": 0.4838, "step": 3695 }, { "epoch": 0.7, "learning_rate": 3.779470529025175e-06, "loss": 0.5251, "step": 3696 }, { "epoch": 0.7, "learning_rate": 3.7788133356159597e-06, "loss": 0.4579, "step": 3697 }, { "epoch": 0.7, "learning_rate": 3.778156022495435e-06, "loss": 0.5088, "step": 3698 }, { "epoch": 0.7, "learning_rate": 3.7774985897251344e-06, "loss": 0.3503, "step": 3699 }, { "epoch": 0.7, "learning_rate": 3.7768410373665998e-06, "loss": 0.4089, "step": 3700 }, { "epoch": 0.7, "learning_rate": 3.776183365481385e-06, "loss": 0.521, "step": 3701 }, { "epoch": 0.7, "learning_rate": 3.775525574131057e-06, "loss": 0.4826, "step": 3702 }, { "epoch": 0.7, "learning_rate": 3.774867663377192e-06, "loss": 0.521, "step": 3703 }, { "epoch": 0.7, "learning_rate": 3.774209633281377e-06, "loss": 0.4478, "step": 3704 }, { "epoch": 0.7, "learning_rate": 3.773551483905212e-06, "loss": 0.4406, "step": 3705 }, { "epoch": 0.7, "learning_rate": 3.772893215310307e-06, "loss": 0.648, "step": 3706 }, { "epoch": 0.7, "learning_rate": 3.7722348275582823e-06, "loss": 0.4319, "step": 3707 }, { "epoch": 0.7, "learning_rate": 3.771576320710772e-06, "loss": 0.5444, "step": 3708 }, { "epoch": 0.7, "learning_rate": 3.770917694829419e-06, "loss": 0.4672, "step": 3709 }, { "epoch": 0.7, "learning_rate": 3.770258949975878e-06, "loss": 0.426, "step": 3710 }, { "epoch": 0.7, "learning_rate": 3.769600086211815e-06, "loss": 0.424, "step": 3711 }, { "epoch": 0.7, "learning_rate": 3.7689411035989067e-06, "loss": 0.5087, "step": 3712 }, { "epoch": 0.7, "learning_rate": 3.768282002198842e-06, "loss": 0.5374, "step": 3713 }, { "epoch": 0.7, "learning_rate": 3.76762278207332e-06, "loss": 0.4247, "step": 3714 }, { "epoch": 0.7, "learning_rate": 3.7669634432840505e-06, "loss": 0.3506, "step": 3715 }, { "epoch": 0.7, "learning_rate": 3.7663039858927564e-06, "loss": 0.4668, "step": 3716 }, { "epoch": 0.7, "learning_rate": 3.765644409961169e-06, "loss": 0.5554, "step": 3717 }, { "epoch": 0.7, "learning_rate": 3.7649847155510314e-06, "loss": 0.4921, "step": 3718 }, { "epoch": 0.7, "learning_rate": 3.764324902724101e-06, "loss": 0.424, "step": 3719 }, { "epoch": 0.7, "learning_rate": 3.7636649715421415e-06, "loss": 0.4212, "step": 3720 }, { "epoch": 0.7, "learning_rate": 3.7630049220669305e-06, "loss": 0.5549, "step": 3721 }, { "epoch": 0.7, "learning_rate": 3.7623447543602566e-06, "loss": 0.489, "step": 3722 }, { "epoch": 0.7, "learning_rate": 3.7616844684839183e-06, "loss": 0.3539, "step": 3723 }, { "epoch": 0.7, "learning_rate": 3.7610240644997254e-06, "loss": 0.4574, "step": 3724 }, { "epoch": 0.7, "learning_rate": 3.7603635424694995e-06, "loss": 0.5158, "step": 3725 }, { "epoch": 0.7, "learning_rate": 3.7597029024550734e-06, "loss": 0.4865, "step": 3726 }, { "epoch": 0.7, "learning_rate": 3.75904214451829e-06, "loss": 0.5264, "step": 3727 }, { "epoch": 0.7, "learning_rate": 3.7583812687210043e-06, "loss": 0.4879, "step": 3728 }, { "epoch": 0.7, "learning_rate": 3.757720275125081e-06, "loss": 0.5106, "step": 3729 }, { "epoch": 0.7, "learning_rate": 3.7570591637923957e-06, "loss": 0.5644, "step": 3730 }, { "epoch": 0.7, "learning_rate": 3.756397934784838e-06, "loss": 0.5352, "step": 3731 }, { "epoch": 0.71, "learning_rate": 3.755736588164304e-06, "loss": 0.4209, "step": 3732 }, { "epoch": 0.71, "learning_rate": 3.755075123992705e-06, "loss": 0.5427, "step": 3733 }, { "epoch": 0.71, "learning_rate": 3.7544135423319617e-06, "loss": 0.4034, "step": 3734 }, { "epoch": 0.71, "learning_rate": 3.7537518432440034e-06, "loss": 0.6428, "step": 3735 }, { "epoch": 0.71, "learning_rate": 3.7530900267907744e-06, "loss": 0.4632, "step": 3736 }, { "epoch": 0.71, "learning_rate": 3.752428093034228e-06, "loss": 0.5413, "step": 3737 }, { "epoch": 0.71, "learning_rate": 3.751766042036329e-06, "loss": 0.4639, "step": 3738 }, { "epoch": 0.71, "learning_rate": 3.751103873859051e-06, "loss": 0.4811, "step": 3739 }, { "epoch": 0.71, "learning_rate": 3.750441588564382e-06, "loss": 0.3919, "step": 3740 }, { "epoch": 0.71, "learning_rate": 3.749779186214319e-06, "loss": 0.4382, "step": 3741 }, { "epoch": 0.71, "learning_rate": 3.7491166668708696e-06, "loss": 0.3654, "step": 3742 }, { "epoch": 0.71, "learning_rate": 3.748454030596054e-06, "loss": 0.4625, "step": 3743 }, { "epoch": 0.71, "learning_rate": 3.7477912774519026e-06, "loss": 0.5213, "step": 3744 }, { "epoch": 0.71, "learning_rate": 3.747128407500455e-06, "loss": 0.4746, "step": 3745 }, { "epoch": 0.71, "learning_rate": 3.7464654208037648e-06, "loss": 0.4069, "step": 3746 }, { "epoch": 0.71, "learning_rate": 3.7458023174238943e-06, "loss": 0.4237, "step": 3747 }, { "epoch": 0.71, "learning_rate": 3.7451390974229187e-06, "loss": 0.4519, "step": 3748 }, { "epoch": 0.71, "learning_rate": 3.744475760862921e-06, "loss": 0.5433, "step": 3749 }, { "epoch": 0.71, "learning_rate": 3.7438123078059967e-06, "loss": 0.3701, "step": 3750 }, { "epoch": 0.71, "learning_rate": 3.743148738314255e-06, "loss": 0.4273, "step": 3751 }, { "epoch": 0.71, "learning_rate": 3.742485052449812e-06, "loss": 0.4793, "step": 3752 }, { "epoch": 0.71, "learning_rate": 3.7418212502747954e-06, "loss": 0.4739, "step": 3753 }, { "epoch": 0.71, "learning_rate": 3.7411573318513462e-06, "loss": 0.3856, "step": 3754 }, { "epoch": 0.71, "learning_rate": 3.740493297241613e-06, "loss": 0.4003, "step": 3755 }, { "epoch": 0.71, "learning_rate": 3.7398291465077575e-06, "loss": 0.2987, "step": 3756 }, { "epoch": 0.71, "learning_rate": 3.739164879711952e-06, "loss": 0.4233, "step": 3757 }, { "epoch": 0.71, "learning_rate": 3.7385004969163795e-06, "loss": 0.5998, "step": 3758 }, { "epoch": 0.71, "learning_rate": 3.7378359981832322e-06, "loss": 0.464, "step": 3759 }, { "epoch": 0.71, "learning_rate": 3.737171383574717e-06, "loss": 0.5225, "step": 3760 }, { "epoch": 0.71, "learning_rate": 3.736506653153048e-06, "loss": 0.5083, "step": 3761 }, { "epoch": 0.71, "learning_rate": 3.7358418069804513e-06, "loss": 0.5201, "step": 3762 }, { "epoch": 0.71, "learning_rate": 3.7351768451191644e-06, "loss": 0.4442, "step": 3763 }, { "epoch": 0.71, "learning_rate": 3.7345117676314347e-06, "loss": 0.4131, "step": 3764 }, { "epoch": 0.71, "learning_rate": 3.7338465745795216e-06, "loss": 0.5754, "step": 3765 }, { "epoch": 0.71, "learning_rate": 3.7331812660256944e-06, "loss": 0.391, "step": 3766 }, { "epoch": 0.71, "learning_rate": 3.7325158420322333e-06, "loss": 0.474, "step": 3767 }, { "epoch": 0.71, "learning_rate": 3.7318503026614295e-06, "loss": 0.405, "step": 3768 }, { "epoch": 0.71, "learning_rate": 3.731184647975585e-06, "loss": 0.602, "step": 3769 }, { "epoch": 0.71, "learning_rate": 3.730518878037012e-06, "loss": 0.4421, "step": 3770 }, { "epoch": 0.71, "learning_rate": 3.7298529929080352e-06, "loss": 0.4892, "step": 3771 }, { "epoch": 0.71, "learning_rate": 3.729186992650988e-06, "loss": 0.4717, "step": 3772 }, { "epoch": 0.71, "learning_rate": 3.7285208773282166e-06, "loss": 0.5519, "step": 3773 }, { "epoch": 0.71, "learning_rate": 3.7278546470020755e-06, "loss": 0.399, "step": 3774 }, { "epoch": 0.71, "learning_rate": 3.7271883017349327e-06, "loss": 0.4459, "step": 3775 }, { "epoch": 0.71, "learning_rate": 3.7265218415891645e-06, "loss": 0.3631, "step": 3776 }, { "epoch": 0.71, "learning_rate": 3.725855266627159e-06, "loss": 0.5246, "step": 3777 }, { "epoch": 0.71, "learning_rate": 3.725188576911316e-06, "loss": 0.323, "step": 3778 }, { "epoch": 0.71, "learning_rate": 3.724521772504045e-06, "loss": 0.4533, "step": 3779 }, { "epoch": 0.71, "learning_rate": 3.7238548534677664e-06, "loss": 0.4853, "step": 3780 }, { "epoch": 0.71, "learning_rate": 3.7231878198649106e-06, "loss": 0.4882, "step": 3781 }, { "epoch": 0.71, "learning_rate": 3.72252067175792e-06, "loss": 0.7005, "step": 3782 }, { "epoch": 0.71, "learning_rate": 3.7218534092092474e-06, "loss": 0.572, "step": 3783 }, { "epoch": 0.71, "learning_rate": 3.7211860322813553e-06, "loss": 0.4326, "step": 3784 }, { "epoch": 0.72, "learning_rate": 3.7205185410367176e-06, "loss": 0.538, "step": 3785 }, { "epoch": 0.72, "learning_rate": 3.7198509355378208e-06, "loss": 0.585, "step": 3786 }, { "epoch": 0.72, "learning_rate": 3.7191832158471584e-06, "loss": 0.4798, "step": 3787 }, { "epoch": 0.72, "learning_rate": 3.718515382027237e-06, "loss": 0.563, "step": 3788 }, { "epoch": 0.72, "learning_rate": 3.717847434140574e-06, "loss": 0.4805, "step": 3789 }, { "epoch": 0.72, "learning_rate": 3.7171793722496963e-06, "loss": 0.4287, "step": 3790 }, { "epoch": 0.72, "learning_rate": 3.7165111964171407e-06, "loss": 0.4399, "step": 3791 }, { "epoch": 0.72, "learning_rate": 3.715842906705458e-06, "loss": 0.5421, "step": 3792 }, { "epoch": 0.72, "learning_rate": 3.715174503177208e-06, "loss": 0.4485, "step": 3793 }, { "epoch": 0.72, "learning_rate": 3.714505985894958e-06, "loss": 0.4665, "step": 3794 }, { "epoch": 0.72, "learning_rate": 3.7138373549212916e-06, "loss": 0.5456, "step": 3795 }, { "epoch": 0.72, "learning_rate": 3.713168610318798e-06, "loss": 0.4584, "step": 3796 }, { "epoch": 0.72, "learning_rate": 3.7124997521500807e-06, "loss": 0.6614, "step": 3797 }, { "epoch": 0.72, "learning_rate": 3.711830780477752e-06, "loss": 0.3122, "step": 3798 }, { "epoch": 0.72, "learning_rate": 3.711161695364435e-06, "loss": 0.4441, "step": 3799 }, { "epoch": 0.72, "learning_rate": 3.710492496872763e-06, "loss": 0.4491, "step": 3800 }, { "epoch": 0.72, "learning_rate": 3.7098231850653814e-06, "loss": 0.4373, "step": 3801 }, { "epoch": 0.72, "learning_rate": 3.7091537600049454e-06, "loss": 0.5326, "step": 3802 }, { "epoch": 0.72, "learning_rate": 3.70848422175412e-06, "loss": 0.3349, "step": 3803 }, { "epoch": 0.72, "learning_rate": 3.707814570375582e-06, "loss": 0.5525, "step": 3804 }, { "epoch": 0.72, "learning_rate": 3.707144805932018e-06, "loss": 0.465, "step": 3805 }, { "epoch": 0.72, "learning_rate": 3.706474928486126e-06, "loss": 0.4477, "step": 3806 }, { "epoch": 0.72, "learning_rate": 3.705804938100613e-06, "loss": 0.5166, "step": 3807 }, { "epoch": 0.72, "learning_rate": 3.7051348348381987e-06, "loss": 0.55, "step": 3808 }, { "epoch": 0.72, "learning_rate": 3.704464618761612e-06, "loss": 0.5458, "step": 3809 }, { "epoch": 0.72, "learning_rate": 3.7037942899335933e-06, "loss": 0.4674, "step": 3810 }, { "epoch": 0.72, "learning_rate": 3.703123848416891e-06, "loss": 0.6009, "step": 3811 }, { "epoch": 0.72, "learning_rate": 3.7024532942742675e-06, "loss": 0.533, "step": 3812 }, { "epoch": 0.72, "learning_rate": 3.701782627568494e-06, "loss": 0.4858, "step": 3813 }, { "epoch": 0.72, "learning_rate": 3.701111848362352e-06, "loss": 0.5234, "step": 3814 }, { "epoch": 0.72, "learning_rate": 3.700440956718635e-06, "loss": 0.5444, "step": 3815 }, { "epoch": 0.72, "learning_rate": 3.6997699527001447e-06, "loss": 0.6067, "step": 3816 }, { "epoch": 0.72, "learning_rate": 3.6990988363696955e-06, "loss": 0.3829, "step": 3817 }, { "epoch": 0.72, "learning_rate": 3.698427607790111e-06, "loss": 0.5147, "step": 3818 }, { "epoch": 0.72, "learning_rate": 3.6977562670242263e-06, "loss": 0.5449, "step": 3819 }, { "epoch": 0.72, "learning_rate": 3.6970848141348857e-06, "loss": 0.4415, "step": 3820 }, { "epoch": 0.72, "learning_rate": 3.6964132491849445e-06, "loss": 0.481, "step": 3821 }, { "epoch": 0.72, "learning_rate": 3.69574157223727e-06, "loss": 0.51, "step": 3822 }, { "epoch": 0.72, "learning_rate": 3.6950697833547376e-06, "loss": 0.5685, "step": 3823 }, { "epoch": 0.72, "learning_rate": 3.694397882600236e-06, "loss": 0.3814, "step": 3824 }, { "epoch": 0.72, "learning_rate": 3.6937258700366608e-06, "loss": 0.5751, "step": 3825 }, { "epoch": 0.72, "learning_rate": 3.6930537457269195e-06, "loss": 0.4045, "step": 3826 }, { "epoch": 0.72, "learning_rate": 3.692381509733932e-06, "loss": 0.5876, "step": 3827 }, { "epoch": 0.72, "learning_rate": 3.6917091621206274e-06, "loss": 0.3642, "step": 3828 }, { "epoch": 0.72, "learning_rate": 3.691036702949944e-06, "loss": 0.481, "step": 3829 }, { "epoch": 0.72, "learning_rate": 3.690364132284831e-06, "loss": 0.4392, "step": 3830 }, { "epoch": 0.72, "learning_rate": 3.6896914501882507e-06, "loss": 0.4493, "step": 3831 }, { "epoch": 0.72, "learning_rate": 3.689018656723171e-06, "loss": 0.5221, "step": 3832 }, { "epoch": 0.72, "learning_rate": 3.688345751952575e-06, "loss": 0.489, "step": 3833 }, { "epoch": 0.72, "learning_rate": 3.6876727359394537e-06, "loss": 0.5853, "step": 3834 }, { "epoch": 0.72, "learning_rate": 3.6869996087468085e-06, "loss": 0.3867, "step": 3835 }, { "epoch": 0.72, "learning_rate": 3.6863263704376517e-06, "loss": 0.6144, "step": 3836 }, { "epoch": 0.72, "learning_rate": 3.6856530210750063e-06, "loss": 0.4625, "step": 3837 }, { "epoch": 0.73, "learning_rate": 3.6849795607219057e-06, "loss": 0.4304, "step": 3838 }, { "epoch": 0.73, "learning_rate": 3.6843059894413923e-06, "loss": 0.4403, "step": 3839 }, { "epoch": 0.73, "learning_rate": 3.6836323072965206e-06, "loss": 0.4307, "step": 3840 }, { "epoch": 0.73, "learning_rate": 3.682958514350356e-06, "loss": 0.3679, "step": 3841 }, { "epoch": 0.73, "learning_rate": 3.682284610665971e-06, "loss": 0.4146, "step": 3842 }, { "epoch": 0.73, "learning_rate": 3.681610596306451e-06, "loss": 0.4457, "step": 3843 }, { "epoch": 0.73, "learning_rate": 3.6809364713348923e-06, "loss": 0.4102, "step": 3844 }, { "epoch": 0.73, "learning_rate": 3.6802622358144002e-06, "loss": 0.3985, "step": 3845 }, { "epoch": 0.73, "learning_rate": 3.6795878898080905e-06, "loss": 0.5684, "step": 3846 }, { "epoch": 0.73, "learning_rate": 3.6789134333790896e-06, "loss": 0.5059, "step": 3847 }, { "epoch": 0.73, "learning_rate": 3.678238866590534e-06, "loss": 0.4329, "step": 3848 }, { "epoch": 0.73, "learning_rate": 3.6775641895055713e-06, "loss": 0.5085, "step": 3849 }, { "epoch": 0.73, "learning_rate": 3.6768894021873584e-06, "loss": 0.6063, "step": 3850 }, { "epoch": 0.73, "learning_rate": 3.6762145046990632e-06, "loss": 0.4542, "step": 3851 }, { "epoch": 0.73, "learning_rate": 3.675539497103864e-06, "loss": 0.5355, "step": 3852 }, { "epoch": 0.73, "learning_rate": 3.6748643794649485e-06, "loss": 0.4942, "step": 3853 }, { "epoch": 0.73, "learning_rate": 3.674189151845515e-06, "loss": 0.4118, "step": 3854 }, { "epoch": 0.73, "learning_rate": 3.673513814308774e-06, "loss": 0.6864, "step": 3855 }, { "epoch": 0.73, "learning_rate": 3.672838366917943e-06, "loss": 0.4054, "step": 3856 }, { "epoch": 0.73, "learning_rate": 3.6721628097362523e-06, "loss": 0.556, "step": 3857 }, { "epoch": 0.73, "learning_rate": 3.6714871428269405e-06, "loss": 0.4219, "step": 3858 }, { "epoch": 0.73, "learning_rate": 3.6708113662532603e-06, "loss": 0.4235, "step": 3859 }, { "epoch": 0.73, "learning_rate": 3.670135480078469e-06, "loss": 0.4001, "step": 3860 }, { "epoch": 0.73, "learning_rate": 3.6694594843658384e-06, "loss": 0.4188, "step": 3861 }, { "epoch": 0.73, "learning_rate": 3.6687833791786497e-06, "loss": 0.466, "step": 3862 }, { "epoch": 0.73, "learning_rate": 3.6681071645801937e-06, "loss": 0.4544, "step": 3863 }, { "epoch": 0.73, "learning_rate": 3.667430840633771e-06, "loss": 0.4491, "step": 3864 }, { "epoch": 0.73, "learning_rate": 3.666754407402694e-06, "loss": 0.5314, "step": 3865 }, { "epoch": 0.73, "learning_rate": 3.6660778649502847e-06, "loss": 0.4008, "step": 3866 }, { "epoch": 0.73, "learning_rate": 3.665401213339874e-06, "loss": 0.435, "step": 3867 }, { "epoch": 0.73, "learning_rate": 3.664724452634804e-06, "loss": 0.5467, "step": 3868 }, { "epoch": 0.73, "learning_rate": 3.6640475828984284e-06, "loss": 0.4606, "step": 3869 }, { "epoch": 0.73, "learning_rate": 3.663370604194109e-06, "loss": 0.5128, "step": 3870 }, { "epoch": 0.73, "learning_rate": 3.662693516585219e-06, "loss": 0.4529, "step": 3871 }, { "epoch": 0.73, "learning_rate": 3.6620163201351405e-06, "loss": 0.467, "step": 3872 }, { "epoch": 0.73, "learning_rate": 3.661339014907268e-06, "loss": 0.5449, "step": 3873 }, { "epoch": 0.73, "learning_rate": 3.6606616009650043e-06, "loss": 0.4642, "step": 3874 }, { "epoch": 0.73, "learning_rate": 3.659984078371762e-06, "loss": 0.516, "step": 3875 }, { "epoch": 0.73, "learning_rate": 3.6593064471909674e-06, "loss": 0.3884, "step": 3876 }, { "epoch": 0.73, "learning_rate": 3.6586287074860517e-06, "loss": 0.4121, "step": 3877 }, { "epoch": 0.73, "learning_rate": 3.6579508593204594e-06, "loss": 0.5567, "step": 3878 }, { "epoch": 0.73, "learning_rate": 3.6572729027576463e-06, "loss": 0.5992, "step": 3879 }, { "epoch": 0.73, "learning_rate": 3.656594837861076e-06, "loss": 0.4253, "step": 3880 }, { "epoch": 0.73, "learning_rate": 3.6559166646942217e-06, "loss": 0.4444, "step": 3881 }, { "epoch": 0.73, "learning_rate": 3.65523838332057e-06, "loss": 0.4485, "step": 3882 }, { "epoch": 0.73, "learning_rate": 3.6545599938036146e-06, "loss": 0.492, "step": 3883 }, { "epoch": 0.73, "learning_rate": 3.6538814962068603e-06, "loss": 0.4496, "step": 3884 }, { "epoch": 0.73, "learning_rate": 3.6532028905938227e-06, "loss": 0.4759, "step": 3885 }, { "epoch": 0.73, "learning_rate": 3.6525241770280267e-06, "loss": 0.3244, "step": 3886 }, { "epoch": 0.73, "learning_rate": 3.651845355573007e-06, "loss": 0.4571, "step": 3887 }, { "epoch": 0.73, "learning_rate": 3.6511664262923096e-06, "loss": 0.5105, "step": 3888 }, { "epoch": 0.73, "learning_rate": 3.6504873892494897e-06, "loss": 0.4336, "step": 3889 }, { "epoch": 0.73, "learning_rate": 3.649808244508113e-06, "loss": 0.3685, "step": 3890 }, { "epoch": 0.74, "learning_rate": 3.649128992131754e-06, "loss": 0.4892, "step": 3891 }, { "epoch": 0.74, "learning_rate": 3.6484496321839994e-06, "loss": 0.44, "step": 3892 }, { "epoch": 0.74, "learning_rate": 3.6477701647284448e-06, "loss": 0.4012, "step": 3893 }, { "epoch": 0.74, "learning_rate": 3.6470905898286966e-06, "loss": 0.5303, "step": 3894 }, { "epoch": 0.74, "learning_rate": 3.6464109075483693e-06, "loss": 0.5313, "step": 3895 }, { "epoch": 0.74, "learning_rate": 3.6457311179510895e-06, "loss": 0.5114, "step": 3896 }, { "epoch": 0.74, "learning_rate": 3.645051221100494e-06, "loss": 0.3574, "step": 3897 }, { "epoch": 0.74, "learning_rate": 3.644371217060227e-06, "loss": 0.4603, "step": 3898 }, { "epoch": 0.74, "learning_rate": 3.643691105893945e-06, "loss": 0.3585, "step": 3899 }, { "epoch": 0.74, "learning_rate": 3.6430108876653155e-06, "loss": 0.4413, "step": 3900 }, { "epoch": 0.74, "learning_rate": 3.6423305624380134e-06, "loss": 0.3997, "step": 3901 }, { "epoch": 0.74, "learning_rate": 3.6416501302757244e-06, "loss": 0.4576, "step": 3902 }, { "epoch": 0.74, "learning_rate": 3.6409695912421454e-06, "loss": 0.4473, "step": 3903 }, { "epoch": 0.74, "learning_rate": 3.6402889454009826e-06, "loss": 0.5328, "step": 3904 }, { "epoch": 0.74, "learning_rate": 3.6396081928159514e-06, "loss": 0.351, "step": 3905 }, { "epoch": 0.74, "learning_rate": 3.638927333550779e-06, "loss": 0.4512, "step": 3906 }, { "epoch": 0.74, "learning_rate": 3.6382463676692004e-06, "loss": 0.5579, "step": 3907 }, { "epoch": 0.74, "learning_rate": 3.6375652952349616e-06, "loss": 0.4977, "step": 3908 }, { "epoch": 0.74, "learning_rate": 3.6368841163118198e-06, "loss": 0.5387, "step": 3909 }, { "epoch": 0.74, "learning_rate": 3.636202830963539e-06, "loss": 0.4415, "step": 3910 }, { "epoch": 0.74, "learning_rate": 3.6355214392538974e-06, "loss": 0.5103, "step": 3911 }, { "epoch": 0.74, "learning_rate": 3.63483994124668e-06, "loss": 0.5856, "step": 3912 }, { "epoch": 0.74, "learning_rate": 3.634158337005682e-06, "loss": 0.6014, "step": 3913 }, { "epoch": 0.74, "learning_rate": 3.63347662659471e-06, "loss": 0.4558, "step": 3914 }, { "epoch": 0.74, "learning_rate": 3.6327948100775812e-06, "loss": 0.4541, "step": 3915 }, { "epoch": 0.74, "learning_rate": 3.632112887518118e-06, "loss": 0.6395, "step": 3916 }, { "epoch": 0.74, "learning_rate": 3.6314308589801577e-06, "loss": 0.4774, "step": 3917 }, { "epoch": 0.74, "learning_rate": 3.630748724527547e-06, "loss": 0.5386, "step": 3918 }, { "epoch": 0.74, "learning_rate": 3.630066484224139e-06, "loss": 0.6299, "step": 3919 }, { "epoch": 0.74, "learning_rate": 3.6293841381338013e-06, "loss": 0.5837, "step": 3920 }, { "epoch": 0.74, "learning_rate": 3.6287016863204076e-06, "loss": 0.504, "step": 3921 }, { "epoch": 0.74, "learning_rate": 3.6280191288478437e-06, "loss": 0.5255, "step": 3922 }, { "epoch": 0.74, "learning_rate": 3.6273364657800047e-06, "loss": 0.521, "step": 3923 }, { "epoch": 0.74, "learning_rate": 3.6266536971807954e-06, "loss": 0.476, "step": 3924 }, { "epoch": 0.74, "learning_rate": 3.625970823114131e-06, "loss": 0.4249, "step": 3925 }, { "epoch": 0.74, "learning_rate": 3.6252878436439355e-06, "loss": 0.4808, "step": 3926 }, { "epoch": 0.74, "learning_rate": 3.6246047588341437e-06, "loss": 0.6116, "step": 3927 }, { "epoch": 0.74, "learning_rate": 3.6239215687486997e-06, "loss": 0.4785, "step": 3928 }, { "epoch": 0.74, "learning_rate": 3.6232382734515593e-06, "loss": 0.4957, "step": 3929 }, { "epoch": 0.74, "learning_rate": 3.622554873006685e-06, "loss": 0.4722, "step": 3930 }, { "epoch": 0.74, "learning_rate": 3.62187136747805e-06, "loss": 0.4, "step": 3931 }, { "epoch": 0.74, "learning_rate": 3.6211877569296417e-06, "loss": 0.5325, "step": 3932 }, { "epoch": 0.74, "learning_rate": 3.62050404142545e-06, "loss": 0.5091, "step": 3933 }, { "epoch": 0.74, "learning_rate": 3.6198202210294797e-06, "loss": 0.5658, "step": 3934 }, { "epoch": 0.74, "learning_rate": 3.6191362958057446e-06, "loss": 0.4735, "step": 3935 }, { "epoch": 0.74, "learning_rate": 3.6184522658182674e-06, "loss": 0.4788, "step": 3936 }, { "epoch": 0.74, "learning_rate": 3.617768131131081e-06, "loss": 0.4547, "step": 3937 }, { "epoch": 0.74, "learning_rate": 3.6170838918082276e-06, "loss": 0.4218, "step": 3938 }, { "epoch": 0.74, "learning_rate": 3.6163995479137605e-06, "loss": 0.573, "step": 3939 }, { "epoch": 0.74, "learning_rate": 3.615715099511741e-06, "loss": 0.4715, "step": 3940 }, { "epoch": 0.74, "learning_rate": 3.615030546666243e-06, "loss": 0.4004, "step": 3941 }, { "epoch": 0.74, "learning_rate": 3.6143458894413463e-06, "loss": 0.4056, "step": 3942 }, { "epoch": 0.74, "learning_rate": 3.6136611279011435e-06, "loss": 0.3283, "step": 3943 }, { "epoch": 0.75, "learning_rate": 3.6129762621097357e-06, "loss": 0.3865, "step": 3944 }, { "epoch": 0.75, "learning_rate": 3.6122912921312337e-06, "loss": 0.6745, "step": 3945 }, { "epoch": 0.75, "learning_rate": 3.6116062180297603e-06, "loss": 0.3765, "step": 3946 }, { "epoch": 0.75, "learning_rate": 3.610921039869444e-06, "loss": 0.5117, "step": 3947 }, { "epoch": 0.75, "learning_rate": 3.6102357577144246e-06, "loss": 0.4638, "step": 3948 }, { "epoch": 0.75, "learning_rate": 3.609550371628855e-06, "loss": 0.5163, "step": 3949 }, { "epoch": 0.75, "learning_rate": 3.6088648816768935e-06, "loss": 0.5272, "step": 3950 }, { "epoch": 0.75, "learning_rate": 3.608179287922709e-06, "loss": 0.4362, "step": 3951 }, { "epoch": 0.75, "learning_rate": 3.607493590430482e-06, "loss": 0.4323, "step": 3952 }, { "epoch": 0.75, "learning_rate": 3.6068077892644015e-06, "loss": 0.5201, "step": 3953 }, { "epoch": 0.75, "learning_rate": 3.6061218844886646e-06, "loss": 0.4346, "step": 3954 }, { "epoch": 0.75, "learning_rate": 3.6054358761674812e-06, "loss": 0.5426, "step": 3955 }, { "epoch": 0.75, "learning_rate": 3.604749764365069e-06, "loss": 0.4772, "step": 3956 }, { "epoch": 0.75, "learning_rate": 3.604063549145656e-06, "loss": 0.3968, "step": 3957 }, { "epoch": 0.75, "learning_rate": 3.6033772305734793e-06, "loss": 0.549, "step": 3958 }, { "epoch": 0.75, "learning_rate": 3.602690808712786e-06, "loss": 0.455, "step": 3959 }, { "epoch": 0.75, "learning_rate": 3.6020042836278335e-06, "loss": 0.4803, "step": 3960 }, { "epoch": 0.75, "learning_rate": 3.601317655382887e-06, "loss": 0.4751, "step": 3961 }, { "epoch": 0.75, "learning_rate": 3.6006309240422244e-06, "loss": 0.4072, "step": 3962 }, { "epoch": 0.75, "learning_rate": 3.59994408967013e-06, "loss": 0.5049, "step": 3963 }, { "epoch": 0.75, "learning_rate": 3.5992571523309e-06, "loss": 0.6038, "step": 3964 }, { "epoch": 0.75, "learning_rate": 3.598570112088839e-06, "loss": 0.5867, "step": 3965 }, { "epoch": 0.75, "learning_rate": 3.597882969008262e-06, "loss": 0.4784, "step": 3966 }, { "epoch": 0.75, "learning_rate": 3.5971957231534933e-06, "loss": 0.4854, "step": 3967 }, { "epoch": 0.75, "learning_rate": 3.596508374588866e-06, "loss": 0.5237, "step": 3968 }, { "epoch": 0.75, "learning_rate": 3.5958209233787255e-06, "loss": 0.5229, "step": 3969 }, { "epoch": 0.75, "learning_rate": 3.5951333695874236e-06, "loss": 0.4485, "step": 3970 }, { "epoch": 0.75, "learning_rate": 3.594445713279323e-06, "loss": 0.3963, "step": 3971 }, { "epoch": 0.75, "learning_rate": 3.5937579545187975e-06, "loss": 0.439, "step": 3972 }, { "epoch": 0.75, "learning_rate": 3.593070093370227e-06, "loss": 0.5961, "step": 3973 }, { "epoch": 0.75, "learning_rate": 3.592382129898004e-06, "loss": 0.4858, "step": 3974 }, { "epoch": 0.75, "learning_rate": 3.59169406416653e-06, "loss": 0.4376, "step": 3975 }, { "epoch": 0.75, "learning_rate": 3.591005896240215e-06, "loss": 0.4971, "step": 3976 }, { "epoch": 0.75, "learning_rate": 3.59031762618348e-06, "loss": 0.4105, "step": 3977 }, { "epoch": 0.75, "learning_rate": 3.5896292540607534e-06, "loss": 0.4641, "step": 3978 }, { "epoch": 0.75, "learning_rate": 3.5889407799364765e-06, "loss": 0.4907, "step": 3979 }, { "epoch": 0.75, "learning_rate": 3.588252203875097e-06, "loss": 0.3576, "step": 3980 }, { "epoch": 0.75, "learning_rate": 3.5875635259410743e-06, "loss": 0.4318, "step": 3981 }, { "epoch": 0.75, "learning_rate": 3.5868747461988753e-06, "loss": 0.5538, "step": 3982 }, { "epoch": 0.75, "learning_rate": 3.5861858647129773e-06, "loss": 0.5029, "step": 3983 }, { "epoch": 0.75, "learning_rate": 3.5854968815478688e-06, "loss": 0.4781, "step": 3984 }, { "epoch": 0.75, "learning_rate": 3.5848077967680457e-06, "loss": 0.4398, "step": 3985 }, { "epoch": 0.75, "learning_rate": 3.584118610438013e-06, "loss": 0.4004, "step": 3986 }, { "epoch": 0.75, "learning_rate": 3.583429322622288e-06, "loss": 0.4871, "step": 3987 }, { "epoch": 0.75, "learning_rate": 3.5827399333853953e-06, "loss": 0.3928, "step": 3988 }, { "epoch": 0.75, "learning_rate": 3.5820504427918682e-06, "loss": 0.5225, "step": 3989 }, { "epoch": 0.75, "learning_rate": 3.581360850906253e-06, "loss": 0.4441, "step": 3990 }, { "epoch": 0.75, "learning_rate": 3.580671157793102e-06, "loss": 0.4473, "step": 3991 }, { "epoch": 0.75, "learning_rate": 3.579981363516978e-06, "loss": 0.5258, "step": 3992 }, { "epoch": 0.75, "learning_rate": 3.5792914681424533e-06, "loss": 0.4137, "step": 3993 }, { "epoch": 0.75, "learning_rate": 3.578601471734111e-06, "loss": 0.452, "step": 3994 }, { "epoch": 0.75, "learning_rate": 3.577911374356542e-06, "loss": 0.4806, "step": 3995 }, { "epoch": 0.75, "learning_rate": 3.577221176074347e-06, "loss": 0.4052, "step": 3996 }, { "epoch": 0.76, "learning_rate": 3.5765308769521355e-06, "loss": 0.4244, "step": 3997 }, { "epoch": 0.76, "learning_rate": 3.5758404770545298e-06, "loss": 0.4356, "step": 3998 }, { "epoch": 0.76, "learning_rate": 3.5751499764461563e-06, "loss": 0.3733, "step": 3999 }, { "epoch": 0.76, "learning_rate": 3.5744593751916547e-06, "loss": 0.4595, "step": 4000 }, { "epoch": 0.76, "learning_rate": 3.573768673355673e-06, "loss": 0.4984, "step": 4001 }, { "epoch": 0.76, "learning_rate": 3.57307787100287e-06, "loss": 0.4967, "step": 4002 }, { "epoch": 0.76, "learning_rate": 3.57238696819791e-06, "loss": 0.552, "step": 4003 }, { "epoch": 0.76, "learning_rate": 3.571695965005471e-06, "loss": 0.4964, "step": 4004 }, { "epoch": 0.76, "learning_rate": 3.5710048614902393e-06, "loss": 0.5281, "step": 4005 }, { "epoch": 0.76, "learning_rate": 3.5703136577169073e-06, "loss": 0.5412, "step": 4006 }, { "epoch": 0.76, "learning_rate": 3.5696223537501817e-06, "loss": 0.4193, "step": 4007 }, { "epoch": 0.76, "learning_rate": 3.5689309496547766e-06, "loss": 0.4265, "step": 4008 }, { "epoch": 0.76, "learning_rate": 3.5682394454954133e-06, "loss": 0.513, "step": 4009 }, { "epoch": 0.76, "learning_rate": 3.567547841336826e-06, "loss": 0.5079, "step": 4010 }, { "epoch": 0.76, "learning_rate": 3.566856137243755e-06, "loss": 0.5264, "step": 4011 }, { "epoch": 0.76, "learning_rate": 3.5661643332809536e-06, "loss": 0.4603, "step": 4012 }, { "epoch": 0.76, "learning_rate": 3.565472429513181e-06, "loss": 0.4006, "step": 4013 }, { "epoch": 0.76, "learning_rate": 3.5647804260052077e-06, "loss": 0.5645, "step": 4014 }, { "epoch": 0.76, "learning_rate": 3.5640883228218126e-06, "loss": 0.3566, "step": 4015 }, { "epoch": 0.76, "learning_rate": 3.5633961200277856e-06, "loss": 0.448, "step": 4016 }, { "epoch": 0.76, "learning_rate": 3.5627038176879224e-06, "loss": 0.5083, "step": 4017 }, { "epoch": 0.76, "learning_rate": 3.562011415867032e-06, "loss": 0.5829, "step": 4018 }, { "epoch": 0.76, "learning_rate": 3.5613189146299315e-06, "loss": 0.5135, "step": 4019 }, { "epoch": 0.76, "learning_rate": 3.5606263140414447e-06, "loss": 0.4323, "step": 4020 }, { "epoch": 0.76, "learning_rate": 3.5599336141664087e-06, "loss": 0.5167, "step": 4021 }, { "epoch": 0.76, "learning_rate": 3.559240815069667e-06, "loss": 0.5119, "step": 4022 }, { "epoch": 0.76, "learning_rate": 3.558547916816074e-06, "loss": 0.4057, "step": 4023 }, { "epoch": 0.76, "learning_rate": 3.5578549194704916e-06, "loss": 0.551, "step": 4024 }, { "epoch": 0.76, "learning_rate": 3.557161823097794e-06, "loss": 0.4024, "step": 4025 }, { "epoch": 0.76, "learning_rate": 3.556468627762861e-06, "loss": 0.6209, "step": 4026 }, { "epoch": 0.76, "learning_rate": 3.555775333530585e-06, "loss": 0.4672, "step": 4027 }, { "epoch": 0.76, "learning_rate": 3.555081940465865e-06, "loss": 0.4332, "step": 4028 }, { "epoch": 0.76, "learning_rate": 3.554388448633611e-06, "loss": 0.4145, "step": 4029 }, { "epoch": 0.76, "learning_rate": 3.5536948580987413e-06, "loss": 0.4847, "step": 4030 }, { "epoch": 0.76, "learning_rate": 3.553001168926184e-06, "loss": 0.5049, "step": 4031 }, { "epoch": 0.76, "learning_rate": 3.5523073811808757e-06, "loss": 0.4711, "step": 4032 }, { "epoch": 0.76, "learning_rate": 3.5516134949277637e-06, "loss": 0.426, "step": 4033 }, { "epoch": 0.76, "learning_rate": 3.550919510231803e-06, "loss": 0.4064, "step": 4034 }, { "epoch": 0.76, "learning_rate": 3.550225427157958e-06, "loss": 0.4138, "step": 4035 }, { "epoch": 0.76, "learning_rate": 3.549531245771203e-06, "loss": 0.4656, "step": 4036 }, { "epoch": 0.76, "learning_rate": 3.5488369661365225e-06, "loss": 0.5449, "step": 4037 }, { "epoch": 0.76, "learning_rate": 3.5481425883189057e-06, "loss": 0.5003, "step": 4038 }, { "epoch": 0.76, "learning_rate": 3.547448112383357e-06, "loss": 0.4153, "step": 4039 }, { "epoch": 0.76, "learning_rate": 3.546753538394887e-06, "loss": 0.4446, "step": 4040 }, { "epoch": 0.76, "learning_rate": 3.5460588664185138e-06, "loss": 0.4646, "step": 4041 }, { "epoch": 0.76, "learning_rate": 3.5453640965192682e-06, "loss": 0.3189, "step": 4042 }, { "epoch": 0.76, "learning_rate": 3.5446692287621877e-06, "loss": 0.5369, "step": 4043 }, { "epoch": 0.76, "learning_rate": 3.5439742632123196e-06, "loss": 0.4804, "step": 4044 }, { "epoch": 0.76, "learning_rate": 3.5432791999347217e-06, "loss": 0.4816, "step": 4045 }, { "epoch": 0.76, "learning_rate": 3.5425840389944583e-06, "loss": 0.7056, "step": 4046 }, { "epoch": 0.76, "learning_rate": 3.541888780456605e-06, "loss": 0.4151, "step": 4047 }, { "epoch": 0.76, "learning_rate": 3.541193424386245e-06, "loss": 0.5782, "step": 4048 }, { "epoch": 0.76, "learning_rate": 3.5404979708484728e-06, "loss": 0.4266, "step": 4049 }, { "epoch": 0.77, "learning_rate": 3.53980241990839e-06, "loss": 0.4648, "step": 4050 }, { "epoch": 0.77, "learning_rate": 3.539106771631107e-06, "loss": 0.5225, "step": 4051 }, { "epoch": 0.77, "learning_rate": 3.538411026081746e-06, "loss": 0.5325, "step": 4052 }, { "epoch": 0.77, "learning_rate": 3.5377151833254353e-06, "loss": 0.5476, "step": 4053 }, { "epoch": 0.77, "learning_rate": 3.537019243427315e-06, "loss": 0.6052, "step": 4054 }, { "epoch": 0.77, "learning_rate": 3.536323206452532e-06, "loss": 0.4079, "step": 4055 }, { "epoch": 0.77, "learning_rate": 3.535627072466243e-06, "loss": 0.4105, "step": 4056 }, { "epoch": 0.77, "learning_rate": 3.5349308415336136e-06, "loss": 0.5324, "step": 4057 }, { "epoch": 0.77, "learning_rate": 3.534234513719821e-06, "loss": 0.6545, "step": 4058 }, { "epoch": 0.77, "learning_rate": 3.533538089090046e-06, "loss": 0.5369, "step": 4059 }, { "epoch": 0.77, "learning_rate": 3.5328415677094845e-06, "loss": 0.3894, "step": 4060 }, { "epoch": 0.77, "learning_rate": 3.532144949643338e-06, "loss": 0.5337, "step": 4061 }, { "epoch": 0.77, "learning_rate": 3.531448234956817e-06, "loss": 0.5123, "step": 4062 }, { "epoch": 0.77, "learning_rate": 3.530751423715143e-06, "loss": 0.4893, "step": 4063 }, { "epoch": 0.77, "learning_rate": 3.5300545159835443e-06, "loss": 0.5339, "step": 4064 }, { "epoch": 0.77, "learning_rate": 3.52935751182726e-06, "loss": 0.6002, "step": 4065 }, { "epoch": 0.77, "learning_rate": 3.5286604113115373e-06, "loss": 0.5151, "step": 4066 }, { "epoch": 0.77, "learning_rate": 3.5279632145016317e-06, "loss": 0.5376, "step": 4067 }, { "epoch": 0.77, "learning_rate": 3.527265921462811e-06, "loss": 0.4071, "step": 4068 }, { "epoch": 0.77, "learning_rate": 3.5265685322603483e-06, "loss": 0.4499, "step": 4069 }, { "epoch": 0.77, "learning_rate": 3.525871046959526e-06, "loss": 0.3674, "step": 4070 }, { "epoch": 0.77, "learning_rate": 3.5251734656256388e-06, "loss": 0.6672, "step": 4071 }, { "epoch": 0.77, "learning_rate": 3.524475788323986e-06, "loss": 0.4144, "step": 4072 }, { "epoch": 0.77, "learning_rate": 3.5237780151198793e-06, "loss": 0.5235, "step": 4073 }, { "epoch": 0.77, "learning_rate": 3.523080146078638e-06, "loss": 0.4112, "step": 4074 }, { "epoch": 0.77, "learning_rate": 3.5223821812655906e-06, "loss": 0.4301, "step": 4075 }, { "epoch": 0.77, "learning_rate": 3.5216841207460735e-06, "loss": 0.4135, "step": 4076 }, { "epoch": 0.77, "learning_rate": 3.5209859645854338e-06, "loss": 0.5457, "step": 4077 }, { "epoch": 0.77, "learning_rate": 3.520287712849027e-06, "loss": 0.4632, "step": 4078 }, { "epoch": 0.77, "learning_rate": 3.5195893656022167e-06, "loss": 0.602, "step": 4079 }, { "epoch": 0.77, "learning_rate": 3.5188909229103764e-06, "loss": 0.4527, "step": 4080 }, { "epoch": 0.77, "learning_rate": 3.5181923848388882e-06, "loss": 0.4878, "step": 4081 }, { "epoch": 0.77, "learning_rate": 3.5174937514531432e-06, "loss": 0.4161, "step": 4082 }, { "epoch": 0.77, "learning_rate": 3.516795022818541e-06, "loss": 0.5313, "step": 4083 }, { "epoch": 0.77, "learning_rate": 3.5160961990004904e-06, "loss": 0.4438, "step": 4084 }, { "epoch": 0.77, "learning_rate": 3.5153972800644098e-06, "loss": 0.5352, "step": 4085 }, { "epoch": 0.77, "learning_rate": 3.5146982660757254e-06, "loss": 0.5251, "step": 4086 }, { "epoch": 0.77, "learning_rate": 3.513999157099873e-06, "loss": 0.4498, "step": 4087 }, { "epoch": 0.77, "learning_rate": 3.5132999532022965e-06, "loss": 0.3562, "step": 4088 }, { "epoch": 0.77, "learning_rate": 3.5126006544484502e-06, "loss": 0.4898, "step": 4089 }, { "epoch": 0.77, "learning_rate": 3.5119012609037957e-06, "loss": 0.4766, "step": 4090 }, { "epoch": 0.77, "learning_rate": 3.5112017726338034e-06, "loss": 0.3865, "step": 4091 }, { "epoch": 0.77, "learning_rate": 3.510502189703955e-06, "loss": 0.5986, "step": 4092 }, { "epoch": 0.77, "learning_rate": 3.5098025121797375e-06, "loss": 0.4321, "step": 4093 }, { "epoch": 0.77, "learning_rate": 3.5091027401266503e-06, "loss": 0.5975, "step": 4094 }, { "epoch": 0.77, "learning_rate": 3.508402873610199e-06, "loss": 0.6192, "step": 4095 }, { "epoch": 0.77, "learning_rate": 3.5077029126958984e-06, "loss": 0.4886, "step": 4096 }, { "epoch": 0.77, "learning_rate": 3.5070028574492745e-06, "loss": 0.3831, "step": 4097 }, { "epoch": 0.77, "learning_rate": 3.506302707935858e-06, "loss": 0.4773, "step": 4098 }, { "epoch": 0.77, "learning_rate": 3.505602464221193e-06, "loss": 0.4764, "step": 4099 }, { "epoch": 0.77, "learning_rate": 3.5049021263708285e-06, "loss": 0.3705, "step": 4100 }, { "epoch": 0.77, "learning_rate": 3.504201694450325e-06, "loss": 0.4664, "step": 4101 }, { "epoch": 0.77, "learning_rate": 3.5035011685252502e-06, "loss": 0.3558, "step": 4102 }, { "epoch": 0.78, "learning_rate": 3.502800548661182e-06, "loss": 0.4227, "step": 4103 }, { "epoch": 0.78, "learning_rate": 3.5020998349237054e-06, "loss": 0.5058, "step": 4104 }, { "epoch": 0.78, "learning_rate": 3.5013990273784154e-06, "loss": 0.435, "step": 4105 }, { "epoch": 0.78, "learning_rate": 3.5006981260909163e-06, "loss": 0.5138, "step": 4106 }, { "epoch": 0.78, "learning_rate": 3.499997131126819e-06, "loss": 0.5444, "step": 4107 }, { "epoch": 0.78, "learning_rate": 3.499296042551745e-06, "loss": 0.3938, "step": 4108 }, { "epoch": 0.78, "learning_rate": 3.4985948604313243e-06, "loss": 0.2823, "step": 4109 }, { "epoch": 0.78, "learning_rate": 3.4978935848311956e-06, "loss": 0.3813, "step": 4110 }, { "epoch": 0.78, "learning_rate": 3.497192215817005e-06, "loss": 0.5172, "step": 4111 }, { "epoch": 0.78, "learning_rate": 3.49649075345441e-06, "loss": 0.4895, "step": 4112 }, { "epoch": 0.78, "learning_rate": 3.495789197809075e-06, "loss": 0.4392, "step": 4113 }, { "epoch": 0.78, "learning_rate": 3.4950875489466728e-06, "loss": 0.5439, "step": 4114 }, { "epoch": 0.78, "learning_rate": 3.494385806932887e-06, "loss": 0.3925, "step": 4115 }, { "epoch": 0.78, "learning_rate": 3.493683971833407e-06, "loss": 0.4567, "step": 4116 }, { "epoch": 0.78, "learning_rate": 3.4929820437139328e-06, "loss": 0.3402, "step": 4117 }, { "epoch": 0.78, "learning_rate": 3.4922800226401737e-06, "loss": 0.4464, "step": 4118 }, { "epoch": 0.78, "learning_rate": 3.491577908677846e-06, "loss": 0.5519, "step": 4119 }, { "epoch": 0.78, "learning_rate": 3.4908757018926755e-06, "loss": 0.5851, "step": 4120 }, { "epoch": 0.78, "learning_rate": 3.4901734023503963e-06, "loss": 0.4956, "step": 4121 }, { "epoch": 0.78, "learning_rate": 3.4894710101167515e-06, "loss": 0.4326, "step": 4122 }, { "epoch": 0.78, "learning_rate": 3.488768525257495e-06, "loss": 0.5382, "step": 4123 }, { "epoch": 0.78, "learning_rate": 3.488065947838385e-06, "loss": 0.4746, "step": 4124 }, { "epoch": 0.78, "learning_rate": 3.4873632779251907e-06, "loss": 0.5427, "step": 4125 }, { "epoch": 0.78, "learning_rate": 3.4866605155836913e-06, "loss": 0.4647, "step": 4126 }, { "epoch": 0.78, "learning_rate": 3.485957660879673e-06, "loss": 0.3993, "step": 4127 }, { "epoch": 0.78, "learning_rate": 3.485254713878929e-06, "loss": 0.4412, "step": 4128 }, { "epoch": 0.78, "learning_rate": 3.484551674647265e-06, "loss": 0.4535, "step": 4129 }, { "epoch": 0.78, "learning_rate": 3.483848543250493e-06, "loss": 0.5012, "step": 4130 }, { "epoch": 0.78, "learning_rate": 3.483145319754434e-06, "loss": 0.4455, "step": 4131 }, { "epoch": 0.78, "learning_rate": 3.4824420042249167e-06, "loss": 0.3972, "step": 4132 }, { "epoch": 0.78, "learning_rate": 3.4817385967277808e-06, "loss": 0.3732, "step": 4133 }, { "epoch": 0.78, "learning_rate": 3.481035097328872e-06, "loss": 0.429, "step": 4134 }, { "epoch": 0.78, "learning_rate": 3.4803315060940466e-06, "loss": 0.3883, "step": 4135 }, { "epoch": 0.78, "learning_rate": 3.4796278230891684e-06, "loss": 0.3816, "step": 4136 }, { "epoch": 0.78, "learning_rate": 3.478924048380109e-06, "loss": 0.4603, "step": 4137 }, { "epoch": 0.78, "learning_rate": 3.478220182032751e-06, "loss": 0.4356, "step": 4138 }, { "epoch": 0.78, "learning_rate": 3.4775162241129838e-06, "loss": 0.7323, "step": 4139 }, { "epoch": 0.78, "learning_rate": 3.4768121746867055e-06, "loss": 0.4498, "step": 4140 }, { "epoch": 0.78, "learning_rate": 3.476108033819824e-06, "loss": 0.455, "step": 4141 }, { "epoch": 0.78, "learning_rate": 3.4754038015782532e-06, "loss": 0.5234, "step": 4142 }, { "epoch": 0.78, "learning_rate": 3.474699478027918e-06, "loss": 0.4417, "step": 4143 }, { "epoch": 0.78, "learning_rate": 3.4739950632347515e-06, "loss": 0.364, "step": 4144 }, { "epoch": 0.78, "learning_rate": 3.473290557264694e-06, "loss": 0.4085, "step": 4145 }, { "epoch": 0.78, "learning_rate": 3.4725859601836955e-06, "loss": 0.5519, "step": 4146 }, { "epoch": 0.78, "learning_rate": 3.4718812720577143e-06, "loss": 0.3474, "step": 4147 }, { "epoch": 0.78, "learning_rate": 3.4711764929527174e-06, "loss": 0.4184, "step": 4148 }, { "epoch": 0.78, "learning_rate": 3.470471622934679e-06, "loss": 0.3802, "step": 4149 }, { "epoch": 0.78, "learning_rate": 3.469766662069585e-06, "loss": 0.4864, "step": 4150 }, { "epoch": 0.78, "learning_rate": 3.469061610423425e-06, "loss": 0.4382, "step": 4151 }, { "epoch": 0.78, "learning_rate": 3.468356468062201e-06, "loss": 0.5386, "step": 4152 }, { "epoch": 0.78, "learning_rate": 3.4676512350519228e-06, "loss": 0.4184, "step": 4153 }, { "epoch": 0.78, "learning_rate": 3.466945911458607e-06, "loss": 0.5635, "step": 4154 }, { "epoch": 0.78, "learning_rate": 3.466240497348281e-06, "loss": 0.4608, "step": 4155 }, { "epoch": 0.79, "learning_rate": 3.4655349927869787e-06, "loss": 0.5198, "step": 4156 }, { "epoch": 0.79, "learning_rate": 3.4648293978407434e-06, "loss": 0.4237, "step": 4157 }, { "epoch": 0.79, "learning_rate": 3.464123712575627e-06, "loss": 0.4696, "step": 4158 }, { "epoch": 0.79, "learning_rate": 3.463417937057689e-06, "loss": 0.4609, "step": 4159 }, { "epoch": 0.79, "learning_rate": 3.4627120713529986e-06, "loss": 0.4447, "step": 4160 }, { "epoch": 0.79, "learning_rate": 3.4620061155276324e-06, "loss": 0.4623, "step": 4161 }, { "epoch": 0.79, "learning_rate": 3.461300069647677e-06, "loss": 0.4375, "step": 4162 }, { "epoch": 0.79, "learning_rate": 3.460593933779223e-06, "loss": 0.473, "step": 4163 }, { "epoch": 0.79, "learning_rate": 3.4598877079883757e-06, "loss": 0.4671, "step": 4164 }, { "epoch": 0.79, "learning_rate": 3.4591813923412455e-06, "loss": 0.4617, "step": 4165 }, { "epoch": 0.79, "learning_rate": 3.4584749869039503e-06, "loss": 0.4469, "step": 4166 }, { "epoch": 0.79, "learning_rate": 3.457768491742618e-06, "loss": 0.4791, "step": 4167 }, { "epoch": 0.79, "learning_rate": 3.4570619069233845e-06, "loss": 0.4684, "step": 4168 }, { "epoch": 0.79, "learning_rate": 3.4563552325123947e-06, "loss": 0.5426, "step": 4169 }, { "epoch": 0.79, "learning_rate": 3.4556484685758003e-06, "loss": 0.5276, "step": 4170 }, { "epoch": 0.79, "learning_rate": 3.4549416151797633e-06, "loss": 0.3912, "step": 4171 }, { "epoch": 0.79, "learning_rate": 3.454234672390452e-06, "loss": 0.4614, "step": 4172 }, { "epoch": 0.79, "learning_rate": 3.453527640274045e-06, "loss": 0.4718, "step": 4173 }, { "epoch": 0.79, "learning_rate": 3.452820518896729e-06, "loss": 0.4832, "step": 4174 }, { "epoch": 0.79, "learning_rate": 3.4521133083246965e-06, "loss": 0.5189, "step": 4175 }, { "epoch": 0.79, "learning_rate": 3.451406008624153e-06, "loss": 0.4571, "step": 4176 }, { "epoch": 0.79, "learning_rate": 3.450698619861308e-06, "loss": 0.4914, "step": 4177 }, { "epoch": 0.79, "learning_rate": 3.4499911421023807e-06, "loss": 0.4918, "step": 4178 }, { "epoch": 0.79, "learning_rate": 3.4492835754136007e-06, "loss": 0.4828, "step": 4179 }, { "epoch": 0.79, "learning_rate": 3.4485759198612033e-06, "loss": 0.4728, "step": 4180 }, { "epoch": 0.79, "learning_rate": 3.4478681755114316e-06, "loss": 0.4681, "step": 4181 }, { "epoch": 0.79, "learning_rate": 3.4471603424305406e-06, "loss": 0.3246, "step": 4182 }, { "epoch": 0.79, "learning_rate": 3.446452420684791e-06, "loss": 0.4476, "step": 4183 }, { "epoch": 0.79, "learning_rate": 3.4457444103404508e-06, "loss": 0.4812, "step": 4184 }, { "epoch": 0.79, "learning_rate": 3.4450363114638e-06, "loss": 0.4367, "step": 4185 }, { "epoch": 0.79, "learning_rate": 3.4443281241211225e-06, "loss": 0.4134, "step": 4186 }, { "epoch": 0.79, "learning_rate": 3.443619848378713e-06, "loss": 0.5005, "step": 4187 }, { "epoch": 0.79, "learning_rate": 3.4429114843028755e-06, "loss": 0.4051, "step": 4188 }, { "epoch": 0.79, "learning_rate": 3.4422030319599187e-06, "loss": 0.3421, "step": 4189 }, { "epoch": 0.79, "learning_rate": 3.4414944914161647e-06, "loss": 0.471, "step": 4190 }, { "epoch": 0.79, "learning_rate": 3.4407858627379376e-06, "loss": 0.5012, "step": 4191 }, { "epoch": 0.79, "learning_rate": 3.4400771459915748e-06, "loss": 0.537, "step": 4192 }, { "epoch": 0.79, "learning_rate": 3.43936834124342e-06, "loss": 0.4579, "step": 4193 }, { "epoch": 0.79, "learning_rate": 3.4386594485598257e-06, "loss": 0.3514, "step": 4194 }, { "epoch": 0.79, "learning_rate": 3.43795046800715e-06, "loss": 0.4739, "step": 4195 }, { "epoch": 0.79, "learning_rate": 3.437241399651764e-06, "loss": 0.4835, "step": 4196 }, { "epoch": 0.79, "learning_rate": 3.4365322435600444e-06, "loss": 0.4694, "step": 4197 }, { "epoch": 0.79, "learning_rate": 3.4358229997983738e-06, "loss": 0.3803, "step": 4198 }, { "epoch": 0.79, "learning_rate": 3.435113668433148e-06, "loss": 0.5744, "step": 4199 }, { "epoch": 0.79, "learning_rate": 3.434404249530767e-06, "loss": 0.4265, "step": 4200 }, { "epoch": 0.79, "learning_rate": 3.4336947431576412e-06, "loss": 0.418, "step": 4201 }, { "epoch": 0.79, "learning_rate": 3.4329851493801876e-06, "loss": 0.3972, "step": 4202 }, { "epoch": 0.79, "learning_rate": 3.4322754682648317e-06, "loss": 0.4656, "step": 4203 }, { "epoch": 0.79, "learning_rate": 3.4315656998780094e-06, "loss": 0.5791, "step": 4204 }, { "epoch": 0.79, "learning_rate": 3.4308558442861616e-06, "loss": 0.3647, "step": 4205 }, { "epoch": 0.79, "learning_rate": 3.430145901555739e-06, "loss": 0.4009, "step": 4206 }, { "epoch": 0.79, "learning_rate": 3.4294358717532007e-06, "loss": 0.4097, "step": 4207 }, { "epoch": 0.8, "learning_rate": 3.4287257549450126e-06, "loss": 0.4332, "step": 4208 }, { "epoch": 0.8, "learning_rate": 3.428015551197651e-06, "loss": 0.425, "step": 4209 }, { "epoch": 0.8, "learning_rate": 3.427305260577597e-06, "loss": 0.513, "step": 4210 }, { "epoch": 0.8, "learning_rate": 3.426594883151344e-06, "loss": 0.6223, "step": 4211 }, { "epoch": 0.8, "learning_rate": 3.42588441898539e-06, "loss": 0.4691, "step": 4212 }, { "epoch": 0.8, "learning_rate": 3.4251738681462423e-06, "loss": 0.4013, "step": 4213 }, { "epoch": 0.8, "learning_rate": 3.4244632307004173e-06, "loss": 0.4263, "step": 4214 }, { "epoch": 0.8, "learning_rate": 3.423752506714438e-06, "loss": 0.5197, "step": 4215 }, { "epoch": 0.8, "learning_rate": 3.423041696254835e-06, "loss": 0.5612, "step": 4216 }, { "epoch": 0.8, "learning_rate": 3.422330799388151e-06, "loss": 0.3231, "step": 4217 }, { "epoch": 0.8, "learning_rate": 3.421619816180932e-06, "loss": 0.4606, "step": 4218 }, { "epoch": 0.8, "learning_rate": 3.4209087466997347e-06, "loss": 0.4332, "step": 4219 }, { "epoch": 0.8, "learning_rate": 3.420197591011123e-06, "loss": 0.3956, "step": 4220 }, { "epoch": 0.8, "learning_rate": 3.4194863491816686e-06, "loss": 0.4404, "step": 4221 }, { "epoch": 0.8, "learning_rate": 3.4187750212779523e-06, "loss": 0.3536, "step": 4222 }, { "epoch": 0.8, "learning_rate": 3.4180636073665626e-06, "loss": 0.5476, "step": 4223 }, { "epoch": 0.8, "learning_rate": 3.417352107514095e-06, "loss": 0.4709, "step": 4224 }, { "epoch": 0.8, "learning_rate": 3.4166405217871545e-06, "loss": 0.3054, "step": 4225 }, { "epoch": 0.8, "learning_rate": 3.4159288502523536e-06, "loss": 0.3084, "step": 4226 }, { "epoch": 0.8, "learning_rate": 3.415217092976312e-06, "loss": 0.4166, "step": 4227 }, { "epoch": 0.8, "learning_rate": 3.4145052500256595e-06, "loss": 0.4338, "step": 4228 }, { "epoch": 0.8, "learning_rate": 3.413793321467031e-06, "loss": 0.389, "step": 4229 }, { "epoch": 0.8, "learning_rate": 3.413081307367072e-06, "loss": 0.4898, "step": 4230 }, { "epoch": 0.8, "learning_rate": 3.4123692077924353e-06, "loss": 0.4418, "step": 4231 }, { "epoch": 0.8, "learning_rate": 3.4116570228097813e-06, "loss": 0.3812, "step": 4232 }, { "epoch": 0.8, "learning_rate": 3.4109447524857776e-06, "loss": 0.5168, "step": 4233 }, { "epoch": 0.8, "learning_rate": 3.4102323968871007e-06, "loss": 0.4675, "step": 4234 }, { "epoch": 0.8, "learning_rate": 3.4095199560804377e-06, "loss": 0.48, "step": 4235 }, { "epoch": 0.8, "learning_rate": 3.4088074301324776e-06, "loss": 0.504, "step": 4236 }, { "epoch": 0.8, "learning_rate": 3.408094819109923e-06, "loss": 0.5087, "step": 4237 }, { "epoch": 0.8, "learning_rate": 3.407382123079481e-06, "loss": 0.5698, "step": 4238 }, { "epoch": 0.8, "learning_rate": 3.406669342107869e-06, "loss": 0.4726, "step": 4239 }, { "epoch": 0.8, "learning_rate": 3.4059564762618113e-06, "loss": 0.5379, "step": 4240 }, { "epoch": 0.8, "learning_rate": 3.4052435256080395e-06, "loss": 0.5578, "step": 4241 }, { "epoch": 0.8, "learning_rate": 3.4045304902132944e-06, "loss": 0.4513, "step": 4242 }, { "epoch": 0.8, "learning_rate": 3.4038173701443234e-06, "loss": 0.5217, "step": 4243 }, { "epoch": 0.8, "learning_rate": 3.403104165467883e-06, "loss": 0.3638, "step": 4244 }, { "epoch": 0.8, "learning_rate": 3.402390876250737e-06, "loss": 0.4954, "step": 4245 }, { "epoch": 0.8, "learning_rate": 3.4016775025596583e-06, "loss": 0.4487, "step": 4246 }, { "epoch": 0.8, "learning_rate": 3.4009640444614243e-06, "loss": 0.5177, "step": 4247 }, { "epoch": 0.8, "learning_rate": 3.4002505020228255e-06, "loss": 0.4723, "step": 4248 }, { "epoch": 0.8, "learning_rate": 3.399536875310656e-06, "loss": 0.436, "step": 4249 }, { "epoch": 0.8, "learning_rate": 3.398823164391719e-06, "loss": 0.4045, "step": 4250 }, { "epoch": 0.8, "learning_rate": 3.398109369332827e-06, "loss": 0.4586, "step": 4251 }, { "epoch": 0.8, "learning_rate": 3.3973954902007987e-06, "loss": 0.5291, "step": 4252 }, { "epoch": 0.8, "learning_rate": 3.396681527062461e-06, "loss": 0.4491, "step": 4253 }, { "epoch": 0.8, "learning_rate": 3.395967479984649e-06, "loss": 0.492, "step": 4254 }, { "epoch": 0.8, "learning_rate": 3.395253349034206e-06, "loss": 0.4799, "step": 4255 }, { "epoch": 0.8, "learning_rate": 3.394539134277982e-06, "loss": 0.4315, "step": 4256 }, { "epoch": 0.8, "learning_rate": 3.393824835782836e-06, "loss": 0.4296, "step": 4257 }, { "epoch": 0.8, "learning_rate": 3.3931104536156336e-06, "loss": 0.3924, "step": 4258 }, { "epoch": 0.8, "learning_rate": 3.39239598784325e-06, "loss": 0.5534, "step": 4259 }, { "epoch": 0.8, "learning_rate": 3.3916814385325663e-06, "loss": 0.382, "step": 4260 }, { "epoch": 0.81, "learning_rate": 3.390966805750473e-06, "loss": 0.4036, "step": 4261 }, { "epoch": 0.81, "learning_rate": 3.3902520895638674e-06, "loss": 0.5422, "step": 4262 }, { "epoch": 0.81, "learning_rate": 3.3895372900396562e-06, "loss": 0.5614, "step": 4263 }, { "epoch": 0.81, "learning_rate": 3.388822407244751e-06, "loss": 0.585, "step": 4264 }, { "epoch": 0.81, "learning_rate": 3.3881074412460737e-06, "loss": 0.4785, "step": 4265 }, { "epoch": 0.81, "learning_rate": 3.387392392110552e-06, "loss": 0.4254, "step": 4266 }, { "epoch": 0.81, "learning_rate": 3.3866772599051257e-06, "loss": 0.3846, "step": 4267 }, { "epoch": 0.81, "learning_rate": 3.3859620446967352e-06, "loss": 0.5941, "step": 4268 }, { "epoch": 0.81, "learning_rate": 3.3852467465523347e-06, "loss": 0.586, "step": 4269 }, { "epoch": 0.81, "learning_rate": 3.384531365538885e-06, "loss": 0.3939, "step": 4270 }, { "epoch": 0.81, "learning_rate": 3.383815901723352e-06, "loss": 0.4161, "step": 4271 }, { "epoch": 0.81, "learning_rate": 3.3831003551727117e-06, "loss": 0.4434, "step": 4272 }, { "epoch": 0.81, "learning_rate": 3.3823847259539475e-06, "loss": 0.4875, "step": 4273 }, { "epoch": 0.81, "learning_rate": 3.3816690141340512e-06, "loss": 0.4384, "step": 4274 }, { "epoch": 0.81, "learning_rate": 3.3809532197800205e-06, "loss": 0.5563, "step": 4275 }, { "epoch": 0.81, "learning_rate": 3.380237342958862e-06, "loss": 0.5503, "step": 4276 }, { "epoch": 0.81, "learning_rate": 3.3795213837375897e-06, "loss": 0.5399, "step": 4277 }, { "epoch": 0.81, "learning_rate": 3.378805342183225e-06, "loss": 0.541, "step": 4278 }, { "epoch": 0.81, "learning_rate": 3.3780892183627977e-06, "loss": 0.3775, "step": 4279 }, { "epoch": 0.81, "learning_rate": 3.377373012343347e-06, "loss": 0.5558, "step": 4280 }, { "epoch": 0.81, "learning_rate": 3.376656724191915e-06, "loss": 0.6515, "step": 4281 }, { "epoch": 0.81, "learning_rate": 3.3759403539755553e-06, "loss": 0.5423, "step": 4282 }, { "epoch": 0.81, "learning_rate": 3.3752239017613285e-06, "loss": 0.4245, "step": 4283 }, { "epoch": 0.81, "learning_rate": 3.374507367616303e-06, "loss": 0.3716, "step": 4284 }, { "epoch": 0.81, "learning_rate": 3.373790751607553e-06, "loss": 0.446, "step": 4285 }, { "epoch": 0.81, "learning_rate": 3.3730740538021632e-06, "loss": 0.4374, "step": 4286 }, { "epoch": 0.81, "learning_rate": 3.3723572742672244e-06, "loss": 0.3077, "step": 4287 }, { "epoch": 0.81, "learning_rate": 3.371640413069835e-06, "loss": 0.5653, "step": 4288 }, { "epoch": 0.81, "learning_rate": 3.370923470277101e-06, "loss": 0.5823, "step": 4289 }, { "epoch": 0.81, "learning_rate": 3.3702064459561362e-06, "loss": 0.4802, "step": 4290 }, { "epoch": 0.81, "learning_rate": 3.3694893401740623e-06, "loss": 0.401, "step": 4291 }, { "epoch": 0.81, "learning_rate": 3.3687721529980094e-06, "loss": 0.5579, "step": 4292 }, { "epoch": 0.81, "learning_rate": 3.3680548844951134e-06, "loss": 0.5976, "step": 4293 }, { "epoch": 0.81, "learning_rate": 3.3673375347325183e-06, "loss": 0.4619, "step": 4294 }, { "epoch": 0.81, "learning_rate": 3.3666201037773766e-06, "loss": 0.4555, "step": 4295 }, { "epoch": 0.81, "learning_rate": 3.365902591696848e-06, "loss": 0.4371, "step": 4296 }, { "epoch": 0.81, "learning_rate": 3.3651849985581e-06, "loss": 0.4647, "step": 4297 }, { "epoch": 0.81, "learning_rate": 3.3644673244283066e-06, "loss": 0.4615, "step": 4298 }, { "epoch": 0.81, "learning_rate": 3.363749569374651e-06, "loss": 0.5135, "step": 4299 }, { "epoch": 0.81, "learning_rate": 3.3630317334643215e-06, "loss": 0.4503, "step": 4300 }, { "epoch": 0.81, "learning_rate": 3.362313816764518e-06, "loss": 0.4417, "step": 4301 }, { "epoch": 0.81, "learning_rate": 3.3615958193424443e-06, "loss": 0.3456, "step": 4302 }, { "epoch": 0.81, "learning_rate": 3.3608777412653127e-06, "loss": 0.433, "step": 4303 }, { "epoch": 0.81, "learning_rate": 3.3601595826003437e-06, "loss": 0.5406, "step": 4304 }, { "epoch": 0.81, "learning_rate": 3.359441343414766e-06, "loss": 0.515, "step": 4305 }, { "epoch": 0.81, "learning_rate": 3.3587230237758126e-06, "loss": 0.4678, "step": 4306 }, { "epoch": 0.81, "learning_rate": 3.358004623750729e-06, "loss": 0.422, "step": 4307 }, { "epoch": 0.81, "learning_rate": 3.3572861434067637e-06, "loss": 0.3514, "step": 4308 }, { "epoch": 0.81, "learning_rate": 3.3565675828111748e-06, "loss": 0.4351, "step": 4309 }, { "epoch": 0.81, "learning_rate": 3.3558489420312284e-06, "loss": 0.4075, "step": 4310 }, { "epoch": 0.81, "learning_rate": 3.355130221134197e-06, "loss": 0.3506, "step": 4311 }, { "epoch": 0.81, "learning_rate": 3.3544114201873602e-06, "loss": 0.313, "step": 4312 }, { "epoch": 0.81, "learning_rate": 3.3536925392580066e-06, "loss": 0.417, "step": 4313 }, { "epoch": 0.82, "learning_rate": 3.3529735784134317e-06, "loss": 0.3333, "step": 4314 }, { "epoch": 0.82, "learning_rate": 3.352254537720938e-06, "loss": 0.5443, "step": 4315 }, { "epoch": 0.82, "learning_rate": 3.351535417247836e-06, "loss": 0.4695, "step": 4316 }, { "epoch": 0.82, "learning_rate": 3.350816217061443e-06, "loss": 0.4976, "step": 4317 }, { "epoch": 0.82, "learning_rate": 3.3500969372290844e-06, "loss": 0.4127, "step": 4318 }, { "epoch": 0.82, "learning_rate": 3.3493775778180936e-06, "loss": 0.5417, "step": 4319 }, { "epoch": 0.82, "learning_rate": 3.3486581388958095e-06, "loss": 0.5053, "step": 4320 }, { "epoch": 0.82, "learning_rate": 3.3479386205295803e-06, "loss": 0.4781, "step": 4321 }, { "epoch": 0.82, "learning_rate": 3.347219022786762e-06, "loss": 0.3687, "step": 4322 }, { "epoch": 0.82, "learning_rate": 3.3464993457347144e-06, "loss": 0.4146, "step": 4323 }, { "epoch": 0.82, "learning_rate": 3.34577958944081e-06, "loss": 0.4613, "step": 4324 }, { "epoch": 0.82, "learning_rate": 3.345059753972425e-06, "loss": 0.4826, "step": 4325 }, { "epoch": 0.82, "learning_rate": 3.344339839396944e-06, "loss": 0.4371, "step": 4326 }, { "epoch": 0.82, "learning_rate": 3.3436198457817597e-06, "loss": 0.5831, "step": 4327 }, { "epoch": 0.82, "learning_rate": 3.342899773194271e-06, "loss": 0.4275, "step": 4328 }, { "epoch": 0.82, "learning_rate": 3.3421796217018854e-06, "loss": 0.4571, "step": 4329 }, { "epoch": 0.82, "learning_rate": 3.3414593913720163e-06, "loss": 0.4701, "step": 4330 }, { "epoch": 0.82, "learning_rate": 3.3407390822720853e-06, "loss": 0.3827, "step": 4331 }, { "epoch": 0.82, "learning_rate": 3.3400186944695223e-06, "loss": 0.4874, "step": 4332 }, { "epoch": 0.82, "learning_rate": 3.3392982280317645e-06, "loss": 0.5951, "step": 4333 }, { "epoch": 0.82, "learning_rate": 3.3385776830262535e-06, "loss": 0.485, "step": 4334 }, { "epoch": 0.82, "learning_rate": 3.337857059520441e-06, "loss": 0.4466, "step": 4335 }, { "epoch": 0.82, "learning_rate": 3.3371363575817865e-06, "loss": 0.4909, "step": 4336 }, { "epoch": 0.82, "learning_rate": 3.3364155772777552e-06, "loss": 0.4057, "step": 4337 }, { "epoch": 0.82, "learning_rate": 3.33569471867582e-06, "loss": 0.5796, "step": 4338 }, { "epoch": 0.82, "learning_rate": 3.3349737818434613e-06, "loss": 0.5172, "step": 4339 }, { "epoch": 0.82, "learning_rate": 3.3342527668481684e-06, "loss": 0.2948, "step": 4340 }, { "epoch": 0.82, "learning_rate": 3.333531673757434e-06, "loss": 0.545, "step": 4341 }, { "epoch": 0.82, "learning_rate": 3.3328105026387627e-06, "loss": 0.4824, "step": 4342 }, { "epoch": 0.82, "learning_rate": 3.3320892535596623e-06, "loss": 0.5157, "step": 4343 }, { "epoch": 0.82, "learning_rate": 3.3313679265876514e-06, "loss": 0.5205, "step": 4344 }, { "epoch": 0.82, "learning_rate": 3.3306465217902534e-06, "loss": 0.5171, "step": 4345 }, { "epoch": 0.82, "learning_rate": 3.3299250392350008e-06, "loss": 0.4412, "step": 4346 }, { "epoch": 0.82, "learning_rate": 3.329203478989431e-06, "loss": 0.323, "step": 4347 }, { "epoch": 0.82, "learning_rate": 3.328481841121092e-06, "loss": 0.4155, "step": 4348 }, { "epoch": 0.82, "learning_rate": 3.327760125697535e-06, "loss": 0.5216, "step": 4349 }, { "epoch": 0.82, "learning_rate": 3.3270383327863233e-06, "loss": 0.3894, "step": 4350 }, { "epoch": 0.82, "learning_rate": 3.326316462455023e-06, "loss": 0.4101, "step": 4351 }, { "epoch": 0.82, "learning_rate": 3.325594514771209e-06, "loss": 0.3862, "step": 4352 }, { "epoch": 0.82, "learning_rate": 3.3248724898024652e-06, "loss": 0.4428, "step": 4353 }, { "epoch": 0.82, "learning_rate": 3.3241503876163812e-06, "loss": 0.4158, "step": 4354 }, { "epoch": 0.82, "learning_rate": 3.3234282082805518e-06, "loss": 0.4615, "step": 4355 }, { "epoch": 0.82, "learning_rate": 3.3227059518625833e-06, "loss": 0.5285, "step": 4356 }, { "epoch": 0.82, "learning_rate": 3.3219836184300867e-06, "loss": 0.5032, "step": 4357 }, { "epoch": 0.82, "learning_rate": 3.321261208050679e-06, "loss": 0.4707, "step": 4358 }, { "epoch": 0.82, "learning_rate": 3.3205387207919877e-06, "loss": 0.3905, "step": 4359 }, { "epoch": 0.82, "learning_rate": 3.3198161567216447e-06, "loss": 0.5587, "step": 4360 }, { "epoch": 0.82, "learning_rate": 3.319093515907291e-06, "loss": 0.3934, "step": 4361 }, { "epoch": 0.82, "learning_rate": 3.3183707984165725e-06, "loss": 0.4515, "step": 4362 }, { "epoch": 0.82, "learning_rate": 3.3176480043171455e-06, "loss": 0.4258, "step": 4363 }, { "epoch": 0.82, "learning_rate": 3.31692513367667e-06, "loss": 0.4931, "step": 4364 }, { "epoch": 0.82, "learning_rate": 3.316202186562816e-06, "loss": 0.4816, "step": 4365 }, { "epoch": 0.82, "learning_rate": 3.3154791630432586e-06, "loss": 0.5622, "step": 4366 }, { "epoch": 0.83, "learning_rate": 3.314756063185682e-06, "loss": 0.497, "step": 4367 }, { "epoch": 0.83, "learning_rate": 3.314032887057775e-06, "loss": 0.4203, "step": 4368 }, { "epoch": 0.83, "learning_rate": 3.3133096347272363e-06, "loss": 0.4549, "step": 4369 }, { "epoch": 0.83, "learning_rate": 3.3125863062617695e-06, "loss": 0.4273, "step": 4370 }, { "epoch": 0.83, "learning_rate": 3.311862901729088e-06, "loss": 0.4086, "step": 4371 }, { "epoch": 0.83, "learning_rate": 3.3111394211969077e-06, "loss": 0.5044, "step": 4372 }, { "epoch": 0.83, "learning_rate": 3.3104158647329576e-06, "loss": 0.518, "step": 4373 }, { "epoch": 0.83, "learning_rate": 3.3096922324049686e-06, "loss": 0.5033, "step": 4374 }, { "epoch": 0.83, "learning_rate": 3.3089685242806815e-06, "loss": 0.4905, "step": 4375 }, { "epoch": 0.83, "learning_rate": 3.308244740427844e-06, "loss": 0.3689, "step": 4376 }, { "epoch": 0.83, "learning_rate": 3.3075208809142095e-06, "loss": 0.5584, "step": 4377 }, { "epoch": 0.83, "learning_rate": 3.30679694580754e-06, "loss": 0.4913, "step": 4378 }, { "epoch": 0.83, "learning_rate": 3.3060729351756037e-06, "loss": 0.4781, "step": 4379 }, { "epoch": 0.83, "learning_rate": 3.3053488490861767e-06, "loss": 0.4618, "step": 4380 }, { "epoch": 0.83, "learning_rate": 3.304624687607041e-06, "loss": 0.3611, "step": 4381 }, { "epoch": 0.83, "learning_rate": 3.303900450805986e-06, "loss": 0.3774, "step": 4382 }, { "epoch": 0.83, "learning_rate": 3.303176138750809e-06, "loss": 0.5288, "step": 4383 }, { "epoch": 0.83, "learning_rate": 3.3024517515093134e-06, "loss": 0.3866, "step": 4384 }, { "epoch": 0.83, "learning_rate": 3.3017272891493114e-06, "loss": 0.3645, "step": 4385 }, { "epoch": 0.83, "learning_rate": 3.3010027517386184e-06, "loss": 0.377, "step": 4386 }, { "epoch": 0.83, "learning_rate": 3.30027813934506e-06, "loss": 0.4988, "step": 4387 }, { "epoch": 0.83, "learning_rate": 3.2995534520364697e-06, "loss": 0.4333, "step": 4388 }, { "epoch": 0.83, "learning_rate": 3.298828689880684e-06, "loss": 0.5272, "step": 4389 }, { "epoch": 0.83, "learning_rate": 3.2981038529455507e-06, "loss": 0.406, "step": 4390 }, { "epoch": 0.83, "learning_rate": 3.2973789412989214e-06, "loss": 0.3617, "step": 4391 }, { "epoch": 0.83, "learning_rate": 3.296653955008658e-06, "loss": 0.5023, "step": 4392 }, { "epoch": 0.83, "learning_rate": 3.2959288941426243e-06, "loss": 0.5104, "step": 4393 }, { "epoch": 0.83, "learning_rate": 3.2952037587686965e-06, "loss": 0.4589, "step": 4394 }, { "epoch": 0.83, "learning_rate": 3.2944785489547544e-06, "loss": 0.456, "step": 4395 }, { "epoch": 0.83, "learning_rate": 3.2937532647686866e-06, "loss": 0.3139, "step": 4396 }, { "epoch": 0.83, "learning_rate": 3.2930279062783866e-06, "loss": 0.5427, "step": 4397 }, { "epoch": 0.83, "learning_rate": 3.2923024735517575e-06, "loss": 0.4563, "step": 4398 }, { "epoch": 0.83, "learning_rate": 3.2915769666567064e-06, "loss": 0.4833, "step": 4399 }, { "epoch": 0.83, "learning_rate": 3.2908513856611506e-06, "loss": 0.4071, "step": 4400 }, { "epoch": 0.83, "learning_rate": 3.290125730633012e-06, "loss": 0.5168, "step": 4401 }, { "epoch": 0.83, "learning_rate": 3.2894000016402194e-06, "loss": 0.4736, "step": 4402 }, { "epoch": 0.83, "learning_rate": 3.28867419875071e-06, "loss": 0.4154, "step": 4403 }, { "epoch": 0.83, "learning_rate": 3.287948322032426e-06, "loss": 0.5011, "step": 4404 }, { "epoch": 0.83, "learning_rate": 3.287222371553319e-06, "loss": 0.4793, "step": 4405 }, { "epoch": 0.83, "learning_rate": 3.2864963473813467e-06, "loss": 0.4822, "step": 4406 }, { "epoch": 0.83, "learning_rate": 3.285770249584471e-06, "loss": 0.4285, "step": 4407 }, { "epoch": 0.83, "learning_rate": 3.2850440782306635e-06, "loss": 0.4176, "step": 4408 }, { "epoch": 0.83, "learning_rate": 3.2843178333879032e-06, "loss": 0.471, "step": 4409 }, { "epoch": 0.83, "learning_rate": 3.2835915151241734e-06, "loss": 0.4354, "step": 4410 }, { "epoch": 0.83, "learning_rate": 3.282865123507466e-06, "loss": 0.4504, "step": 4411 }, { "epoch": 0.83, "learning_rate": 3.2821386586057795e-06, "loss": 0.3233, "step": 4412 }, { "epoch": 0.83, "learning_rate": 3.2814121204871197e-06, "loss": 0.4206, "step": 4413 }, { "epoch": 0.83, "learning_rate": 3.2806855092194985e-06, "loss": 0.4772, "step": 4414 }, { "epoch": 0.83, "learning_rate": 3.2799588248709346e-06, "loss": 0.4984, "step": 4415 }, { "epoch": 0.83, "learning_rate": 3.2792320675094537e-06, "loss": 0.5035, "step": 4416 }, { "epoch": 0.83, "learning_rate": 3.278505237203089e-06, "loss": 0.465, "step": 4417 }, { "epoch": 0.83, "learning_rate": 3.277778334019879e-06, "loss": 0.44, "step": 4418 }, { "epoch": 0.83, "learning_rate": 3.2770513580278706e-06, "loss": 0.5118, "step": 4419 }, { "epoch": 0.84, "learning_rate": 3.276324309295118e-06, "loss": 0.3411, "step": 4420 }, { "epoch": 0.84, "learning_rate": 3.27559718788968e-06, "loss": 0.5043, "step": 4421 }, { "epoch": 0.84, "learning_rate": 3.2748699938796226e-06, "loss": 0.692, "step": 4422 }, { "epoch": 0.84, "learning_rate": 3.274142727333022e-06, "loss": 0.4548, "step": 4423 }, { "epoch": 0.84, "learning_rate": 3.273415388317955e-06, "loss": 0.3948, "step": 4424 }, { "epoch": 0.84, "learning_rate": 3.2726879769025107e-06, "loss": 0.3964, "step": 4425 }, { "epoch": 0.84, "learning_rate": 3.2719604931547833e-06, "loss": 0.3825, "step": 4426 }, { "epoch": 0.84, "learning_rate": 3.271232937142873e-06, "loss": 0.4005, "step": 4427 }, { "epoch": 0.84, "learning_rate": 3.2705053089348864e-06, "loss": 0.548, "step": 4428 }, { "epoch": 0.84, "learning_rate": 3.2697776085989392e-06, "loss": 0.4041, "step": 4429 }, { "epoch": 0.84, "learning_rate": 3.2690498362031514e-06, "loss": 0.443, "step": 4430 }, { "epoch": 0.84, "learning_rate": 3.2683219918156507e-06, "loss": 0.4258, "step": 4431 }, { "epoch": 0.84, "learning_rate": 3.2675940755045717e-06, "loss": 0.5017, "step": 4432 }, { "epoch": 0.84, "learning_rate": 3.2668660873380554e-06, "loss": 0.4692, "step": 4433 }, { "epoch": 0.84, "learning_rate": 3.26613802738425e-06, "loss": 0.4657, "step": 4434 }, { "epoch": 0.84, "learning_rate": 3.26540989571131e-06, "loss": 0.4423, "step": 4435 }, { "epoch": 0.84, "learning_rate": 3.264681692387396e-06, "loss": 0.3826, "step": 4436 }, { "epoch": 0.84, "learning_rate": 3.2639534174806775e-06, "loss": 0.4644, "step": 4437 }, { "epoch": 0.84, "learning_rate": 3.263225071059328e-06, "loss": 0.5037, "step": 4438 }, { "epoch": 0.84, "learning_rate": 3.2624966531915283e-06, "loss": 0.3566, "step": 4439 }, { "epoch": 0.84, "learning_rate": 3.2617681639454684e-06, "loss": 0.5329, "step": 4440 }, { "epoch": 0.84, "learning_rate": 3.2610396033893425e-06, "loss": 0.5072, "step": 4441 }, { "epoch": 0.84, "learning_rate": 3.2603109715913506e-06, "loss": 0.444, "step": 4442 }, { "epoch": 0.84, "learning_rate": 3.259582268619703e-06, "loss": 0.3786, "step": 4443 }, { "epoch": 0.84, "learning_rate": 3.258853494542614e-06, "loss": 0.3831, "step": 4444 }, { "epoch": 0.84, "learning_rate": 3.2581246494283024e-06, "loss": 0.4314, "step": 4445 }, { "epoch": 0.84, "learning_rate": 3.2573957333449995e-06, "loss": 0.506, "step": 4446 }, { "epoch": 0.84, "learning_rate": 3.2566667463609393e-06, "loss": 0.3889, "step": 4447 }, { "epoch": 0.84, "learning_rate": 3.255937688544362e-06, "loss": 0.5202, "step": 4448 }, { "epoch": 0.84, "learning_rate": 3.255208559963517e-06, "loss": 0.4745, "step": 4449 }, { "epoch": 0.84, "learning_rate": 3.2544793606866583e-06, "loss": 0.4544, "step": 4450 }, { "epoch": 0.84, "learning_rate": 3.2537500907820474e-06, "loss": 0.3318, "step": 4451 }, { "epoch": 0.84, "learning_rate": 3.2530207503179513e-06, "loss": 0.5769, "step": 4452 }, { "epoch": 0.84, "learning_rate": 3.2522913393626455e-06, "loss": 0.38, "step": 4453 }, { "epoch": 0.84, "learning_rate": 3.251561857984411e-06, "loss": 0.5218, "step": 4454 }, { "epoch": 0.84, "learning_rate": 3.2508323062515344e-06, "loss": 0.6452, "step": 4455 }, { "epoch": 0.84, "learning_rate": 3.2501026842323115e-06, "loss": 0.3523, "step": 4456 }, { "epoch": 0.84, "learning_rate": 3.249372991995041e-06, "loss": 0.3944, "step": 4457 }, { "epoch": 0.84, "learning_rate": 3.2486432296080334e-06, "loss": 0.3533, "step": 4458 }, { "epoch": 0.84, "learning_rate": 3.2479133971396004e-06, "loss": 0.4567, "step": 4459 }, { "epoch": 0.84, "learning_rate": 3.247183494658062e-06, "loss": 0.4546, "step": 4460 }, { "epoch": 0.84, "learning_rate": 3.2464535222317473e-06, "loss": 0.3292, "step": 4461 }, { "epoch": 0.84, "learning_rate": 3.2457234799289896e-06, "loss": 0.4736, "step": 4462 }, { "epoch": 0.84, "learning_rate": 3.2449933678181267e-06, "loss": 0.4783, "step": 4463 }, { "epoch": 0.84, "learning_rate": 3.244263185967508e-06, "loss": 0.4132, "step": 4464 }, { "epoch": 0.84, "learning_rate": 3.2435329344454857e-06, "loss": 0.524, "step": 4465 }, { "epoch": 0.84, "learning_rate": 3.2428026133204187e-06, "loss": 0.3985, "step": 4466 }, { "epoch": 0.84, "learning_rate": 3.2420722226606748e-06, "loss": 0.4064, "step": 4467 }, { "epoch": 0.84, "learning_rate": 3.2413417625346267e-06, "loss": 0.5699, "step": 4468 }, { "epoch": 0.84, "learning_rate": 3.240611233010652e-06, "loss": 0.3812, "step": 4469 }, { "epoch": 0.84, "learning_rate": 3.2398806341571385e-06, "loss": 0.4921, "step": 4470 }, { "epoch": 0.84, "learning_rate": 3.239149966042477e-06, "loss": 0.6227, "step": 4471 }, { "epoch": 0.84, "learning_rate": 3.2384192287350675e-06, "loss": 0.4045, "step": 4472 }, { "epoch": 0.85, "learning_rate": 3.237688422303314e-06, "loss": 0.5303, "step": 4473 }, { "epoch": 0.85, "learning_rate": 3.2369575468156284e-06, "loss": 0.5483, "step": 4474 }, { "epoch": 0.85, "learning_rate": 3.2362266023404303e-06, "loss": 0.42, "step": 4475 }, { "epoch": 0.85, "learning_rate": 3.2354955889461424e-06, "loss": 0.4891, "step": 4476 }, { "epoch": 0.85, "learning_rate": 3.234764506701197e-06, "loss": 0.4393, "step": 4477 }, { "epoch": 0.85, "learning_rate": 3.2340333556740305e-06, "loss": 0.5181, "step": 4478 }, { "epoch": 0.85, "learning_rate": 3.2333021359330884e-06, "loss": 0.4565, "step": 4479 }, { "epoch": 0.85, "learning_rate": 3.23257084754682e-06, "loss": 0.3953, "step": 4480 }, { "epoch": 0.85, "learning_rate": 3.2318394905836824e-06, "loss": 0.4062, "step": 4481 }, { "epoch": 0.85, "learning_rate": 3.231108065112139e-06, "loss": 0.3687, "step": 4482 }, { "epoch": 0.85, "learning_rate": 3.230376571200659e-06, "loss": 0.3399, "step": 4483 }, { "epoch": 0.85, "learning_rate": 3.229645008917719e-06, "loss": 0.6654, "step": 4484 }, { "epoch": 0.85, "learning_rate": 3.228913378331801e-06, "loss": 0.4909, "step": 4485 }, { "epoch": 0.85, "learning_rate": 3.228181679511394e-06, "loss": 0.4449, "step": 4486 }, { "epoch": 0.85, "learning_rate": 3.2274499125249936e-06, "loss": 0.4546, "step": 4487 }, { "epoch": 0.85, "learning_rate": 3.226718077441101e-06, "loss": 0.4706, "step": 4488 }, { "epoch": 0.85, "learning_rate": 3.2259861743282237e-06, "loss": 0.4959, "step": 4489 }, { "epoch": 0.85, "learning_rate": 3.225254203254877e-06, "loss": 0.5285, "step": 4490 }, { "epoch": 0.85, "learning_rate": 3.224522164289582e-06, "loss": 0.5127, "step": 4491 }, { "epoch": 0.85, "learning_rate": 3.223790057500864e-06, "loss": 0.4845, "step": 4492 }, { "epoch": 0.85, "learning_rate": 3.2230578829572586e-06, "loss": 0.4465, "step": 4493 }, { "epoch": 0.85, "learning_rate": 3.2223256407273035e-06, "loss": 0.5573, "step": 4494 }, { "epoch": 0.85, "learning_rate": 3.2215933308795454e-06, "loss": 0.5194, "step": 4495 }, { "epoch": 0.85, "learning_rate": 3.2208609534825382e-06, "loss": 0.4277, "step": 4496 }, { "epoch": 0.85, "learning_rate": 3.2201285086048395e-06, "loss": 0.5524, "step": 4497 }, { "epoch": 0.85, "learning_rate": 3.2193959963150136e-06, "loss": 0.3669, "step": 4498 }, { "epoch": 0.85, "learning_rate": 3.2186634166816333e-06, "loss": 0.394, "step": 4499 }, { "epoch": 0.85, "learning_rate": 3.2179307697732753e-06, "loss": 0.5725, "step": 4500 }, { "epoch": 0.85, "learning_rate": 3.217198055658524e-06, "loss": 0.4209, "step": 4501 }, { "epoch": 0.85, "learning_rate": 3.2164652744059702e-06, "loss": 0.4099, "step": 4502 }, { "epoch": 0.85, "learning_rate": 3.21573242608421e-06, "loss": 0.4738, "step": 4503 }, { "epoch": 0.85, "learning_rate": 3.214999510761846e-06, "loss": 0.5333, "step": 4504 }, { "epoch": 0.85, "learning_rate": 3.214266528507487e-06, "loss": 0.4809, "step": 4505 }, { "epoch": 0.85, "learning_rate": 3.213533479389749e-06, "loss": 0.5944, "step": 4506 }, { "epoch": 0.85, "learning_rate": 3.212800363477254e-06, "loss": 0.4385, "step": 4507 }, { "epoch": 0.85, "learning_rate": 3.2120671808386296e-06, "loss": 0.5265, "step": 4508 }, { "epoch": 0.85, "learning_rate": 3.211333931542509e-06, "loss": 0.7313, "step": 4509 }, { "epoch": 0.85, "learning_rate": 3.2106006156575343e-06, "loss": 0.4578, "step": 4510 }, { "epoch": 0.85, "learning_rate": 3.2098672332523506e-06, "loss": 0.4374, "step": 4511 }, { "epoch": 0.85, "learning_rate": 3.2091337843956106e-06, "loss": 0.5594, "step": 4512 }, { "epoch": 0.85, "learning_rate": 3.2084002691559753e-06, "loss": 0.4777, "step": 4513 }, { "epoch": 0.85, "learning_rate": 3.2076666876021085e-06, "loss": 0.3479, "step": 4514 }, { "epoch": 0.85, "learning_rate": 3.206933039802681e-06, "loss": 0.4653, "step": 4515 }, { "epoch": 0.85, "learning_rate": 3.2061993258263724e-06, "loss": 0.4152, "step": 4516 }, { "epoch": 0.85, "learning_rate": 3.205465545741865e-06, "loss": 0.4754, "step": 4517 }, { "epoch": 0.85, "learning_rate": 3.2047316996178494e-06, "loss": 0.4841, "step": 4518 }, { "epoch": 0.85, "learning_rate": 3.203997787523022e-06, "loss": 0.4895, "step": 4519 }, { "epoch": 0.85, "learning_rate": 3.203263809526086e-06, "loss": 0.371, "step": 4520 }, { "epoch": 0.85, "learning_rate": 3.2025297656957483e-06, "loss": 0.3713, "step": 4521 }, { "epoch": 0.85, "learning_rate": 3.201795656100725e-06, "loss": 0.4377, "step": 4522 }, { "epoch": 0.85, "learning_rate": 3.2010614808097353e-06, "loss": 0.5147, "step": 4523 }, { "epoch": 0.85, "learning_rate": 3.2003272398915084e-06, "loss": 0.493, "step": 4524 }, { "epoch": 0.85, "learning_rate": 3.1995929334147763e-06, "loss": 0.4575, "step": 4525 }, { "epoch": 0.86, "learning_rate": 3.1988585614482785e-06, "loss": 0.4957, "step": 4526 }, { "epoch": 0.86, "learning_rate": 3.1981241240607604e-06, "loss": 0.3825, "step": 4527 }, { "epoch": 0.86, "learning_rate": 3.1973896213209743e-06, "loss": 0.4798, "step": 4528 }, { "epoch": 0.86, "learning_rate": 3.196655053297677e-06, "loss": 0.4194, "step": 4529 }, { "epoch": 0.86, "learning_rate": 3.195920420059633e-06, "loss": 0.4597, "step": 4530 }, { "epoch": 0.86, "learning_rate": 3.1951857216756122e-06, "loss": 0.6081, "step": 4531 }, { "epoch": 0.86, "learning_rate": 3.1944509582143894e-06, "loss": 0.426, "step": 4532 }, { "epoch": 0.86, "learning_rate": 3.193716129744748e-06, "loss": 0.4736, "step": 4533 }, { "epoch": 0.86, "learning_rate": 3.1929812363354766e-06, "loss": 0.4155, "step": 4534 }, { "epoch": 0.86, "learning_rate": 3.1922462780553685e-06, "loss": 0.4215, "step": 4535 }, { "epoch": 0.86, "learning_rate": 3.191511254973224e-06, "loss": 0.5314, "step": 4536 }, { "epoch": 0.86, "learning_rate": 3.1907761671578506e-06, "loss": 0.377, "step": 4537 }, { "epoch": 0.86, "learning_rate": 3.19004101467806e-06, "loss": 0.2864, "step": 4538 }, { "epoch": 0.86, "learning_rate": 3.1893057976026697e-06, "loss": 0.5126, "step": 4539 }, { "epoch": 0.86, "learning_rate": 3.1885705160005066e-06, "loss": 0.3434, "step": 4540 }, { "epoch": 0.86, "learning_rate": 3.187835169940399e-06, "loss": 0.3394, "step": 4541 }, { "epoch": 0.86, "learning_rate": 3.1870997594911855e-06, "loss": 0.329, "step": 4542 }, { "epoch": 0.86, "learning_rate": 3.186364284721708e-06, "loss": 0.3909, "step": 4543 }, { "epoch": 0.86, "learning_rate": 3.1856287457008143e-06, "loss": 0.5265, "step": 4544 }, { "epoch": 0.86, "learning_rate": 3.1848931424973613e-06, "loss": 0.5366, "step": 4545 }, { "epoch": 0.86, "learning_rate": 3.184157475180208e-06, "loss": 0.4372, "step": 4546 }, { "epoch": 0.86, "learning_rate": 3.1834217438182208e-06, "loss": 0.4298, "step": 4547 }, { "epoch": 0.86, "learning_rate": 3.1826859484802735e-06, "loss": 0.4616, "step": 4548 }, { "epoch": 0.86, "learning_rate": 3.181950089235245e-06, "loss": 0.4063, "step": 4549 }, { "epoch": 0.86, "learning_rate": 3.181214166152019e-06, "loss": 0.4138, "step": 4550 }, { "epoch": 0.86, "learning_rate": 3.180478179299487e-06, "loss": 0.4322, "step": 4551 }, { "epoch": 0.86, "learning_rate": 3.1797421287465458e-06, "loss": 0.4418, "step": 4552 }, { "epoch": 0.86, "learning_rate": 3.1790060145620965e-06, "loss": 0.4957, "step": 4553 }, { "epoch": 0.86, "learning_rate": 3.1782698368150494e-06, "loss": 0.3842, "step": 4554 }, { "epoch": 0.86, "learning_rate": 3.177533595574318e-06, "loss": 0.2589, "step": 4555 }, { "epoch": 0.86, "learning_rate": 3.176797290908823e-06, "loss": 0.3743, "step": 4556 }, { "epoch": 0.86, "learning_rate": 3.176060922887491e-06, "loss": 0.3456, "step": 4557 }, { "epoch": 0.86, "learning_rate": 3.175324491579255e-06, "loss": 0.502, "step": 4558 }, { "epoch": 0.86, "learning_rate": 3.1745879970530514e-06, "loss": 0.5317, "step": 4559 }, { "epoch": 0.86, "learning_rate": 3.1738514393778258e-06, "loss": 0.3709, "step": 4560 }, { "epoch": 0.86, "learning_rate": 3.1731148186225276e-06, "loss": 0.3204, "step": 4561 }, { "epoch": 0.86, "learning_rate": 3.1723781348561136e-06, "loss": 0.47, "step": 4562 }, { "epoch": 0.86, "learning_rate": 3.171641388147545e-06, "loss": 0.5642, "step": 4563 }, { "epoch": 0.86, "learning_rate": 3.1709045785657895e-06, "loss": 0.4924, "step": 4564 }, { "epoch": 0.86, "learning_rate": 3.170167706179821e-06, "loss": 0.505, "step": 4565 }, { "epoch": 0.86, "learning_rate": 3.1694307710586198e-06, "loss": 0.4816, "step": 4566 }, { "epoch": 0.86, "learning_rate": 3.16869377327117e-06, "loss": 0.3704, "step": 4567 }, { "epoch": 0.86, "learning_rate": 3.167956712886463e-06, "loss": 0.4785, "step": 4568 }, { "epoch": 0.86, "learning_rate": 3.167219589973497e-06, "loss": 0.442, "step": 4569 }, { "epoch": 0.86, "learning_rate": 3.166482404601274e-06, "loss": 0.4648, "step": 4570 }, { "epoch": 0.86, "learning_rate": 3.1657451568388035e-06, "loss": 0.3653, "step": 4571 }, { "epoch": 0.86, "learning_rate": 3.1650078467551003e-06, "loss": 0.407, "step": 4572 }, { "epoch": 0.86, "learning_rate": 3.1642704744191845e-06, "loss": 0.4979, "step": 4573 }, { "epoch": 0.86, "learning_rate": 3.1635330399000817e-06, "loss": 0.4523, "step": 4574 }, { "epoch": 0.86, "learning_rate": 3.1627955432668256e-06, "loss": 0.3249, "step": 4575 }, { "epoch": 0.86, "learning_rate": 3.1620579845884527e-06, "loss": 0.6112, "step": 4576 }, { "epoch": 0.86, "learning_rate": 3.161320363934008e-06, "loss": 0.5319, "step": 4577 }, { "epoch": 0.86, "learning_rate": 3.1605826813725404e-06, "loss": 0.3604, "step": 4578 }, { "epoch": 0.87, "learning_rate": 3.159844936973105e-06, "loss": 0.5636, "step": 4579 }, { "epoch": 0.87, "learning_rate": 3.1591071308047645e-06, "loss": 0.5274, "step": 4580 }, { "epoch": 0.87, "learning_rate": 3.158369262936584e-06, "loss": 0.3258, "step": 4581 }, { "epoch": 0.87, "learning_rate": 3.157631333437637e-06, "loss": 0.4245, "step": 4582 }, { "epoch": 0.87, "learning_rate": 3.156893342377002e-06, "loss": 0.3731, "step": 4583 }, { "epoch": 0.87, "learning_rate": 3.156155289823765e-06, "loss": 0.5017, "step": 4584 }, { "epoch": 0.87, "learning_rate": 3.1554171758470114e-06, "loss": 0.5499, "step": 4585 }, { "epoch": 0.87, "learning_rate": 3.154679000515842e-06, "loss": 0.5733, "step": 4586 }, { "epoch": 0.87, "learning_rate": 3.1539407638993556e-06, "loss": 0.4137, "step": 4587 }, { "epoch": 0.87, "learning_rate": 3.1532024660666594e-06, "loss": 0.3573, "step": 4588 }, { "epoch": 0.87, "learning_rate": 3.1524641070868677e-06, "loss": 0.3983, "step": 4589 }, { "epoch": 0.87, "learning_rate": 3.1517256870290985e-06, "loss": 0.353, "step": 4590 }, { "epoch": 0.87, "learning_rate": 3.150987205962476e-06, "loss": 0.4335, "step": 4591 }, { "epoch": 0.87, "learning_rate": 3.1502486639561307e-06, "loss": 0.3201, "step": 4592 }, { "epoch": 0.87, "learning_rate": 3.149510061079198e-06, "loss": 0.457, "step": 4593 }, { "epoch": 0.87, "learning_rate": 3.1487713974008204e-06, "loss": 0.381, "step": 4594 }, { "epoch": 0.87, "learning_rate": 3.148032672990144e-06, "loss": 0.5157, "step": 4595 }, { "epoch": 0.87, "learning_rate": 3.147293887916322e-06, "loss": 0.5452, "step": 4596 }, { "epoch": 0.87, "learning_rate": 3.146555042248514e-06, "loss": 0.4024, "step": 4597 }, { "epoch": 0.87, "learning_rate": 3.1458161360558827e-06, "loss": 0.5421, "step": 4598 }, { "epoch": 0.87, "learning_rate": 3.145077169407598e-06, "loss": 0.4227, "step": 4599 }, { "epoch": 0.87, "learning_rate": 3.1443381423728374e-06, "loss": 0.5247, "step": 4600 }, { "epoch": 0.87, "learning_rate": 3.143599055020781e-06, "loss": 0.5839, "step": 4601 }, { "epoch": 0.87, "learning_rate": 3.142859907420615e-06, "loss": 0.3874, "step": 4602 }, { "epoch": 0.87, "learning_rate": 3.142120699641533e-06, "loss": 0.6081, "step": 4603 }, { "epoch": 0.87, "learning_rate": 3.141381431752733e-06, "loss": 0.5176, "step": 4604 }, { "epoch": 0.87, "learning_rate": 3.140642103823418e-06, "loss": 0.4203, "step": 4605 }, { "epoch": 0.87, "learning_rate": 3.1399027159227984e-06, "loss": 0.4273, "step": 4606 }, { "epoch": 0.87, "learning_rate": 3.139163268120089e-06, "loss": 0.4009, "step": 4607 }, { "epoch": 0.87, "learning_rate": 3.1384237604845097e-06, "loss": 0.4704, "step": 4608 }, { "epoch": 0.87, "learning_rate": 3.1376841930852875e-06, "loss": 0.5195, "step": 4609 }, { "epoch": 0.87, "learning_rate": 3.1369445659916543e-06, "loss": 0.392, "step": 4610 }, { "epoch": 0.87, "learning_rate": 3.136204879272847e-06, "loss": 0.5203, "step": 4611 }, { "epoch": 0.87, "learning_rate": 3.1354651329981077e-06, "loss": 0.4617, "step": 4612 }, { "epoch": 0.87, "learning_rate": 3.134725327236687e-06, "loss": 0.4015, "step": 4613 }, { "epoch": 0.87, "learning_rate": 3.133985462057838e-06, "loss": 0.446, "step": 4614 }, { "epoch": 0.87, "learning_rate": 3.1332455375308206e-06, "loss": 0.3474, "step": 4615 }, { "epoch": 0.87, "learning_rate": 3.1325055537249e-06, "loss": 0.485, "step": 4616 }, { "epoch": 0.87, "learning_rate": 3.131765510709346e-06, "loss": 0.3064, "step": 4617 }, { "epoch": 0.87, "learning_rate": 3.1310254085534374e-06, "loss": 0.3332, "step": 4618 }, { "epoch": 0.87, "learning_rate": 3.130285247326454e-06, "loss": 0.4048, "step": 4619 }, { "epoch": 0.87, "learning_rate": 3.129545027097683e-06, "loss": 0.5679, "step": 4620 }, { "epoch": 0.87, "learning_rate": 3.1288047479364186e-06, "loss": 0.473, "step": 4621 }, { "epoch": 0.87, "learning_rate": 3.1280644099119594e-06, "loss": 0.3529, "step": 4622 }, { "epoch": 0.87, "learning_rate": 3.1273240130936076e-06, "loss": 0.5861, "step": 4623 }, { "epoch": 0.87, "learning_rate": 3.1265835575506746e-06, "loss": 0.3389, "step": 4624 }, { "epoch": 0.87, "learning_rate": 3.1258430433524743e-06, "loss": 0.4288, "step": 4625 }, { "epoch": 0.87, "learning_rate": 3.125102470568328e-06, "loss": 0.3631, "step": 4626 }, { "epoch": 0.87, "learning_rate": 3.1243618392675606e-06, "loss": 0.4647, "step": 4627 }, { "epoch": 0.87, "learning_rate": 3.123621149519504e-06, "loss": 0.5382, "step": 4628 }, { "epoch": 0.87, "learning_rate": 3.1228804013934957e-06, "loss": 0.3507, "step": 4629 }, { "epoch": 0.87, "learning_rate": 3.122139594958877e-06, "loss": 0.4476, "step": 4630 }, { "epoch": 0.87, "learning_rate": 3.1213987302849954e-06, "loss": 0.439, "step": 4631 }, { "epoch": 0.88, "learning_rate": 3.120657807441206e-06, "loss": 0.4064, "step": 4632 }, { "epoch": 0.88, "learning_rate": 3.119916826496866e-06, "loss": 0.4311, "step": 4633 }, { "epoch": 0.88, "learning_rate": 3.119175787521339e-06, "loss": 0.433, "step": 4634 }, { "epoch": 0.88, "learning_rate": 3.118434690583997e-06, "loss": 0.3921, "step": 4635 }, { "epoch": 0.88, "learning_rate": 3.1176935357542136e-06, "loss": 0.4972, "step": 4636 }, { "epoch": 0.88, "learning_rate": 3.116952323101368e-06, "loss": 0.4103, "step": 4637 }, { "epoch": 0.88, "learning_rate": 3.116211052694848e-06, "loss": 0.4085, "step": 4638 }, { "epoch": 0.88, "learning_rate": 3.1154697246040445e-06, "loss": 0.4269, "step": 4639 }, { "epoch": 0.88, "learning_rate": 3.114728338898353e-06, "loss": 0.4936, "step": 4640 }, { "epoch": 0.88, "learning_rate": 3.1139868956471765e-06, "loss": 0.4341, "step": 4641 }, { "epoch": 0.88, "learning_rate": 3.113245394919922e-06, "loss": 0.388, "step": 4642 }, { "epoch": 0.88, "learning_rate": 3.1125038367860027e-06, "loss": 0.4977, "step": 4643 }, { "epoch": 0.88, "learning_rate": 3.1117622213148367e-06, "loss": 0.4345, "step": 4644 }, { "epoch": 0.88, "learning_rate": 3.1110205485758472e-06, "loss": 0.3692, "step": 4645 }, { "epoch": 0.88, "learning_rate": 3.1102788186384635e-06, "loss": 0.3641, "step": 4646 }, { "epoch": 0.88, "learning_rate": 3.1095370315721197e-06, "loss": 0.5507, "step": 4647 }, { "epoch": 0.88, "learning_rate": 3.1087951874462553e-06, "loss": 0.5017, "step": 4648 }, { "epoch": 0.88, "learning_rate": 3.1080532863303154e-06, "loss": 0.3503, "step": 4649 }, { "epoch": 0.88, "learning_rate": 3.10731132829375e-06, "loss": 0.4239, "step": 4650 }, { "epoch": 0.88, "learning_rate": 3.1065693134060147e-06, "loss": 0.6355, "step": 4651 }, { "epoch": 0.88, "learning_rate": 3.105827241736571e-06, "loss": 0.491, "step": 4652 }, { "epoch": 0.88, "learning_rate": 3.1050851133548854e-06, "loss": 0.4703, "step": 4653 }, { "epoch": 0.88, "learning_rate": 3.1043429283304283e-06, "loss": 0.5122, "step": 4654 }, { "epoch": 0.88, "learning_rate": 3.1036006867326775e-06, "loss": 0.6035, "step": 4655 }, { "epoch": 0.88, "learning_rate": 3.1028583886311143e-06, "loss": 0.2736, "step": 4656 }, { "epoch": 0.88, "learning_rate": 3.102116034095227e-06, "loss": 0.4716, "step": 4657 }, { "epoch": 0.88, "learning_rate": 3.1013736231945086e-06, "loss": 0.4052, "step": 4658 }, { "epoch": 0.88, "learning_rate": 3.100631155998456e-06, "loss": 0.3612, "step": 4659 }, { "epoch": 0.88, "learning_rate": 3.0998886325765732e-06, "loss": 0.3103, "step": 4660 }, { "epoch": 0.88, "learning_rate": 3.099146052998368e-06, "loss": 0.3638, "step": 4661 }, { "epoch": 0.88, "learning_rate": 3.0984034173333556e-06, "loss": 0.5114, "step": 4662 }, { "epoch": 0.88, "learning_rate": 3.0976607256510534e-06, "loss": 0.5528, "step": 4663 }, { "epoch": 0.88, "learning_rate": 3.0969179780209867e-06, "loss": 0.672, "step": 4664 }, { "epoch": 0.88, "learning_rate": 3.096175174512685e-06, "loss": 0.4546, "step": 4665 }, { "epoch": 0.88, "learning_rate": 3.095432315195682e-06, "loss": 0.5203, "step": 4666 }, { "epoch": 0.88, "learning_rate": 3.09468940013952e-06, "loss": 0.3539, "step": 4667 }, { "epoch": 0.88, "learning_rate": 3.093946429413742e-06, "loss": 0.4703, "step": 4668 }, { "epoch": 0.88, "learning_rate": 3.093203403087899e-06, "loss": 0.3738, "step": 4669 }, { "epoch": 0.88, "learning_rate": 3.0924603212315475e-06, "loss": 0.4385, "step": 4670 }, { "epoch": 0.88, "learning_rate": 3.0917171839142477e-06, "loss": 0.3893, "step": 4671 }, { "epoch": 0.88, "learning_rate": 3.0909739912055646e-06, "loss": 0.5183, "step": 4672 }, { "epoch": 0.88, "learning_rate": 3.090230743175071e-06, "loss": 0.5454, "step": 4673 }, { "epoch": 0.88, "learning_rate": 3.089487439892343e-06, "loss": 0.3644, "step": 4674 }, { "epoch": 0.88, "learning_rate": 3.0887440814269604e-06, "loss": 0.487, "step": 4675 }, { "epoch": 0.88, "learning_rate": 3.0880006678485118e-06, "loss": 0.3451, "step": 4676 }, { "epoch": 0.88, "learning_rate": 3.0872571992265893e-06, "loss": 0.4424, "step": 4677 }, { "epoch": 0.88, "learning_rate": 3.0865136756307886e-06, "loss": 0.5216, "step": 4678 }, { "epoch": 0.88, "learning_rate": 3.0857700971307124e-06, "loss": 0.3414, "step": 4679 }, { "epoch": 0.88, "learning_rate": 3.085026463795968e-06, "loss": 0.3958, "step": 4680 }, { "epoch": 0.88, "learning_rate": 3.084282775696168e-06, "loss": 0.4425, "step": 4681 }, { "epoch": 0.88, "learning_rate": 3.0835390329009296e-06, "loss": 0.6041, "step": 4682 }, { "epoch": 0.88, "learning_rate": 3.082795235479876e-06, "loss": 0.4378, "step": 4683 }, { "epoch": 0.88, "learning_rate": 3.0820513835026345e-06, "loss": 0.4996, "step": 4684 }, { "epoch": 0.89, "learning_rate": 3.0813074770388385e-06, "loss": 0.4225, "step": 4685 }, { "epoch": 0.89, "learning_rate": 3.080563516158125e-06, "loss": 0.461, "step": 4686 }, { "epoch": 0.89, "learning_rate": 3.0798195009301385e-06, "loss": 0.4173, "step": 4687 }, { "epoch": 0.89, "learning_rate": 3.0790754314245276e-06, "loss": 0.441, "step": 4688 }, { "epoch": 0.89, "learning_rate": 3.0783313077109427e-06, "loss": 0.4303, "step": 4689 }, { "epoch": 0.89, "learning_rate": 3.077587129859045e-06, "loss": 0.5829, "step": 4690 }, { "epoch": 0.89, "learning_rate": 3.0768428979384964e-06, "loss": 0.3732, "step": 4691 }, { "epoch": 0.89, "learning_rate": 3.076098612018966e-06, "loss": 0.4534, "step": 4692 }, { "epoch": 0.89, "learning_rate": 3.0753542721701274e-06, "loss": 0.5241, "step": 4693 }, { "epoch": 0.89, "learning_rate": 3.0746098784616586e-06, "loss": 0.5046, "step": 4694 }, { "epoch": 0.89, "learning_rate": 3.0738654309632443e-06, "loss": 0.4848, "step": 4695 }, { "epoch": 0.89, "learning_rate": 3.073120929744572e-06, "loss": 0.641, "step": 4696 }, { "epoch": 0.89, "learning_rate": 3.0723763748753354e-06, "loss": 0.4828, "step": 4697 }, { "epoch": 0.89, "learning_rate": 3.0716317664252345e-06, "loss": 0.4661, "step": 4698 }, { "epoch": 0.89, "learning_rate": 3.0708871044639715e-06, "loss": 0.3611, "step": 4699 }, { "epoch": 0.89, "learning_rate": 3.0701423890612565e-06, "loss": 0.5292, "step": 4700 }, { "epoch": 0.89, "learning_rate": 3.0693976202868016e-06, "loss": 0.478, "step": 4701 }, { "epoch": 0.89, "learning_rate": 3.068652798210327e-06, "loss": 0.462, "step": 4702 }, { "epoch": 0.89, "learning_rate": 3.067907922901556e-06, "loss": 0.4378, "step": 4703 }, { "epoch": 0.89, "learning_rate": 3.0671629944302168e-06, "loss": 0.4319, "step": 4704 }, { "epoch": 0.89, "learning_rate": 3.0664180128660443e-06, "loss": 0.4204, "step": 4705 }, { "epoch": 0.89, "learning_rate": 3.0656729782787756e-06, "loss": 0.4768, "step": 4706 }, { "epoch": 0.89, "learning_rate": 3.0649278907381547e-06, "loss": 0.6947, "step": 4707 }, { "epoch": 0.89, "learning_rate": 3.064182750313931e-06, "loss": 0.4528, "step": 4708 }, { "epoch": 0.89, "learning_rate": 3.0634375570758585e-06, "loss": 0.4228, "step": 4709 }, { "epoch": 0.89, "learning_rate": 3.062692311093693e-06, "loss": 0.3431, "step": 4710 }, { "epoch": 0.89, "learning_rate": 3.061947012437201e-06, "loss": 0.4037, "step": 4711 }, { "epoch": 0.89, "learning_rate": 3.0612016611761486e-06, "loss": 0.4615, "step": 4712 }, { "epoch": 0.89, "learning_rate": 3.0604562573803104e-06, "loss": 0.5695, "step": 4713 }, { "epoch": 0.89, "learning_rate": 3.0597108011194636e-06, "loss": 0.483, "step": 4714 }, { "epoch": 0.89, "learning_rate": 3.0589652924633924e-06, "loss": 0.4796, "step": 4715 }, { "epoch": 0.89, "learning_rate": 3.0582197314818833e-06, "loss": 0.4575, "step": 4716 }, { "epoch": 0.89, "learning_rate": 3.057474118244731e-06, "loss": 0.5485, "step": 4717 }, { "epoch": 0.89, "learning_rate": 3.056728452821731e-06, "loss": 0.5209, "step": 4718 }, { "epoch": 0.89, "learning_rate": 3.0559827352826894e-06, "loss": 0.5562, "step": 4719 }, { "epoch": 0.89, "learning_rate": 3.0552369656974102e-06, "loss": 0.4188, "step": 4720 }, { "epoch": 0.89, "learning_rate": 3.054491144135707e-06, "loss": 0.4777, "step": 4721 }, { "epoch": 0.89, "learning_rate": 3.053745270667398e-06, "loss": 0.6382, "step": 4722 }, { "epoch": 0.89, "learning_rate": 3.052999345362305e-06, "loss": 0.5156, "step": 4723 }, { "epoch": 0.89, "learning_rate": 3.0522533682902543e-06, "loss": 0.4675, "step": 4724 }, { "epoch": 0.89, "learning_rate": 3.051507339521078e-06, "loss": 0.4719, "step": 4725 }, { "epoch": 0.89, "learning_rate": 3.050761259124614e-06, "loss": 0.5506, "step": 4726 }, { "epoch": 0.89, "learning_rate": 3.0500151271707012e-06, "loss": 0.5641, "step": 4727 }, { "epoch": 0.89, "learning_rate": 3.0492689437291885e-06, "loss": 0.58, "step": 4728 }, { "epoch": 0.89, "learning_rate": 3.048522708869926e-06, "loss": 0.3253, "step": 4729 }, { "epoch": 0.89, "learning_rate": 3.0477764226627695e-06, "loss": 0.3083, "step": 4730 }, { "epoch": 0.89, "learning_rate": 3.0470300851775796e-06, "loss": 0.4623, "step": 4731 }, { "epoch": 0.89, "learning_rate": 3.046283696484223e-06, "loss": 0.373, "step": 4732 }, { "epoch": 0.89, "learning_rate": 3.0455372566525694e-06, "loss": 0.544, "step": 4733 }, { "epoch": 0.89, "learning_rate": 3.0447907657524934e-06, "loss": 0.4113, "step": 4734 }, { "epoch": 0.89, "learning_rate": 3.0440442238538753e-06, "loss": 0.4504, "step": 4735 }, { "epoch": 0.89, "learning_rate": 3.0432976310266e-06, "loss": 0.3663, "step": 4736 }, { "epoch": 0.89, "learning_rate": 3.0425509873405577e-06, "loss": 0.5012, "step": 4737 }, { "epoch": 0.9, "learning_rate": 3.0418042928656416e-06, "loss": 0.5848, "step": 4738 }, { "epoch": 0.9, "learning_rate": 3.0410575476717507e-06, "loss": 0.6651, "step": 4739 }, { "epoch": 0.9, "learning_rate": 3.0403107518287896e-06, "loss": 0.3987, "step": 4740 }, { "epoch": 0.9, "learning_rate": 3.039563905406666e-06, "loss": 0.5278, "step": 4741 }, { "epoch": 0.9, "learning_rate": 3.0388170084752925e-06, "loss": 0.4708, "step": 4742 }, { "epoch": 0.9, "learning_rate": 3.038070061104589e-06, "loss": 0.4543, "step": 4743 }, { "epoch": 0.9, "learning_rate": 3.0373230633644775e-06, "loss": 0.418, "step": 4744 }, { "epoch": 0.9, "learning_rate": 3.0365760153248835e-06, "loss": 0.5668, "step": 4745 }, { "epoch": 0.9, "learning_rate": 3.0358289170557415e-06, "loss": 0.518, "step": 4746 }, { "epoch": 0.9, "learning_rate": 3.0350817686269872e-06, "loss": 0.4683, "step": 4747 }, { "epoch": 0.9, "learning_rate": 3.0343345701085627e-06, "loss": 0.5865, "step": 4748 }, { "epoch": 0.9, "learning_rate": 3.0335873215704133e-06, "loss": 0.5355, "step": 4749 }, { "epoch": 0.9, "learning_rate": 3.0328400230824904e-06, "loss": 0.4475, "step": 4750 }, { "epoch": 0.9, "learning_rate": 3.0320926747147493e-06, "loss": 0.5971, "step": 4751 }, { "epoch": 0.9, "learning_rate": 3.031345276537151e-06, "loss": 0.4286, "step": 4752 }, { "epoch": 0.9, "learning_rate": 3.0305978286196584e-06, "loss": 0.3153, "step": 4753 }, { "epoch": 0.9, "learning_rate": 3.029850331032244e-06, "loss": 0.5348, "step": 4754 }, { "epoch": 0.9, "learning_rate": 3.0291027838448793e-06, "loss": 0.3675, "step": 4755 }, { "epoch": 0.9, "learning_rate": 3.028355187127545e-06, "loss": 0.4099, "step": 4756 }, { "epoch": 0.9, "learning_rate": 3.0276075409502226e-06, "loss": 0.3513, "step": 4757 }, { "epoch": 0.9, "learning_rate": 3.026859845382903e-06, "loss": 0.4032, "step": 4758 }, { "epoch": 0.9, "learning_rate": 3.0261121004955757e-06, "loss": 0.3099, "step": 4759 }, { "epoch": 0.9, "learning_rate": 3.02536430635824e-06, "loss": 0.4384, "step": 4760 }, { "epoch": 0.9, "learning_rate": 3.024616463040898e-06, "loss": 0.4817, "step": 4761 }, { "epoch": 0.9, "learning_rate": 3.023868570613555e-06, "loss": 0.478, "step": 4762 }, { "epoch": 0.9, "learning_rate": 3.023120629146223e-06, "loss": 0.4642, "step": 4763 }, { "epoch": 0.9, "learning_rate": 3.0223726387089176e-06, "loss": 0.5679, "step": 4764 }, { "epoch": 0.9, "learning_rate": 3.0216245993716594e-06, "loss": 0.6243, "step": 4765 }, { "epoch": 0.9, "learning_rate": 3.0208765112044722e-06, "loss": 0.4277, "step": 4766 }, { "epoch": 0.9, "learning_rate": 3.0201283742773867e-06, "loss": 0.5622, "step": 4767 }, { "epoch": 0.9, "learning_rate": 3.0193801886604362e-06, "loss": 0.5975, "step": 4768 }, { "epoch": 0.9, "learning_rate": 3.01863195442366e-06, "loss": 0.4131, "step": 4769 }, { "epoch": 0.9, "learning_rate": 3.0178836716371e-06, "loss": 0.5299, "step": 4770 }, { "epoch": 0.9, "learning_rate": 3.0171353403708054e-06, "loss": 0.4283, "step": 4771 }, { "epoch": 0.9, "learning_rate": 3.0163869606948276e-06, "loss": 0.4566, "step": 4772 }, { "epoch": 0.9, "learning_rate": 3.015638532679223e-06, "loss": 0.4324, "step": 4773 }, { "epoch": 0.9, "learning_rate": 3.014890056394053e-06, "loss": 0.5168, "step": 4774 }, { "epoch": 0.9, "learning_rate": 3.014141531909385e-06, "loss": 0.3599, "step": 4775 }, { "epoch": 0.9, "learning_rate": 3.013392959295287e-06, "loss": 0.3947, "step": 4776 }, { "epoch": 0.9, "learning_rate": 3.0126443386218347e-06, "loss": 0.5228, "step": 4777 }, { "epoch": 0.9, "learning_rate": 3.011895669959108e-06, "loss": 0.4618, "step": 4778 }, { "epoch": 0.9, "learning_rate": 3.01114695337719e-06, "loss": 0.4828, "step": 4779 }, { "epoch": 0.9, "learning_rate": 3.010398188946169e-06, "loss": 0.4881, "step": 4780 }, { "epoch": 0.9, "learning_rate": 3.0096493767361385e-06, "loss": 0.3949, "step": 4781 }, { "epoch": 0.9, "learning_rate": 3.0089005168171952e-06, "loss": 0.3446, "step": 4782 }, { "epoch": 0.9, "learning_rate": 3.008151609259441e-06, "loss": 0.3707, "step": 4783 }, { "epoch": 0.9, "learning_rate": 3.0074026541329815e-06, "loss": 0.4839, "step": 4784 }, { "epoch": 0.9, "learning_rate": 3.006653651507928e-06, "loss": 0.4606, "step": 4785 }, { "epoch": 0.9, "learning_rate": 3.005904601454395e-06, "loss": 0.4664, "step": 4786 }, { "epoch": 0.9, "learning_rate": 3.005155504042503e-06, "loss": 0.4521, "step": 4787 }, { "epoch": 0.9, "learning_rate": 3.0044063593423744e-06, "loss": 0.3305, "step": 4788 }, { "epoch": 0.9, "learning_rate": 3.0036571674241393e-06, "loss": 0.4689, "step": 4789 }, { "epoch": 0.9, "learning_rate": 3.0029079283579295e-06, "loss": 0.4701, "step": 4790 }, { "epoch": 0.91, "learning_rate": 3.002158642213881e-06, "loss": 0.6626, "step": 4791 }, { "epoch": 0.91, "learning_rate": 3.0014093090621387e-06, "loss": 0.4876, "step": 4792 }, { "epoch": 0.91, "learning_rate": 3.0006599289728455e-06, "loss": 0.5364, "step": 4793 }, { "epoch": 0.91, "learning_rate": 2.999910502016153e-06, "loss": 0.3317, "step": 4794 }, { "epoch": 0.91, "learning_rate": 2.9991610282622157e-06, "loss": 0.5505, "step": 4795 }, { "epoch": 0.91, "learning_rate": 2.998411507781194e-06, "loss": 0.5925, "step": 4796 }, { "epoch": 0.91, "learning_rate": 2.9976619406432496e-06, "loss": 0.3691, "step": 4797 }, { "epoch": 0.91, "learning_rate": 2.9969123269185514e-06, "loss": 0.3474, "step": 4798 }, { "epoch": 0.91, "learning_rate": 2.9961626666772715e-06, "loss": 0.4783, "step": 4799 }, { "epoch": 0.91, "learning_rate": 2.995412959989587e-06, "loss": 0.4625, "step": 4800 }, { "epoch": 0.91, "learning_rate": 2.9946632069256783e-06, "loss": 0.4278, "step": 4801 }, { "epoch": 0.91, "learning_rate": 2.993913407555731e-06, "loss": 0.4179, "step": 4802 }, { "epoch": 0.91, "learning_rate": 2.9931635619499337e-06, "loss": 0.4083, "step": 4803 }, { "epoch": 0.91, "learning_rate": 2.9924136701784823e-06, "loss": 0.4615, "step": 4804 }, { "epoch": 0.91, "learning_rate": 2.9916637323115737e-06, "loss": 0.4593, "step": 4805 }, { "epoch": 0.91, "learning_rate": 2.9909137484194113e-06, "loss": 0.5387, "step": 4806 }, { "epoch": 0.91, "learning_rate": 2.990163718572201e-06, "loss": 0.4338, "step": 4807 }, { "epoch": 0.91, "learning_rate": 2.989413642840155e-06, "loss": 0.5368, "step": 4808 }, { "epoch": 0.91, "learning_rate": 2.988663521293488e-06, "loss": 0.5212, "step": 4809 }, { "epoch": 0.91, "learning_rate": 2.9879133540024214e-06, "loss": 0.5085, "step": 4810 }, { "epoch": 0.91, "learning_rate": 2.987163141037177e-06, "loss": 0.5364, "step": 4811 }, { "epoch": 0.91, "learning_rate": 2.986412882467985e-06, "loss": 0.4562, "step": 4812 }, { "epoch": 0.91, "learning_rate": 2.985662578365077e-06, "loss": 0.4922, "step": 4813 }, { "epoch": 0.91, "learning_rate": 2.9849122287986914e-06, "loss": 0.329, "step": 4814 }, { "epoch": 0.91, "learning_rate": 2.9841618338390675e-06, "loss": 0.3312, "step": 4815 }, { "epoch": 0.91, "learning_rate": 2.9834113935564513e-06, "loss": 0.5399, "step": 4816 }, { "epoch": 0.91, "learning_rate": 2.982660908021093e-06, "loss": 0.4863, "step": 4817 }, { "epoch": 0.91, "learning_rate": 2.981910377303246e-06, "loss": 0.4804, "step": 4818 }, { "epoch": 0.91, "learning_rate": 2.9811598014731684e-06, "loss": 0.5296, "step": 4819 }, { "epoch": 0.91, "learning_rate": 2.9804091806011233e-06, "loss": 0.3777, "step": 4820 }, { "epoch": 0.91, "learning_rate": 2.979658514757376e-06, "loss": 0.4397, "step": 4821 }, { "epoch": 0.91, "learning_rate": 2.9789078040121976e-06, "loss": 0.4659, "step": 4822 }, { "epoch": 0.91, "learning_rate": 2.9781570484358634e-06, "loss": 0.4398, "step": 4823 }, { "epoch": 0.91, "learning_rate": 2.9774062480986534e-06, "loss": 0.4553, "step": 4824 }, { "epoch": 0.91, "learning_rate": 2.9766554030708496e-06, "loss": 0.4475, "step": 4825 }, { "epoch": 0.91, "learning_rate": 2.9759045134227392e-06, "loss": 0.4097, "step": 4826 }, { "epoch": 0.91, "learning_rate": 2.975153579224616e-06, "loss": 0.4514, "step": 4827 }, { "epoch": 0.91, "learning_rate": 2.974402600546774e-06, "loss": 0.4212, "step": 4828 }, { "epoch": 0.91, "learning_rate": 2.9736515774595137e-06, "loss": 0.4096, "step": 4829 }, { "epoch": 0.91, "learning_rate": 2.972900510033139e-06, "loss": 0.3639, "step": 4830 }, { "epoch": 0.91, "learning_rate": 2.9721493983379608e-06, "loss": 0.4288, "step": 4831 }, { "epoch": 0.91, "learning_rate": 2.9713982424442872e-06, "loss": 0.4356, "step": 4832 }, { "epoch": 0.91, "learning_rate": 2.9706470424224384e-06, "loss": 0.6123, "step": 4833 }, { "epoch": 0.91, "learning_rate": 2.9698957983427336e-06, "loss": 0.4648, "step": 4834 }, { "epoch": 0.91, "learning_rate": 2.969144510275498e-06, "loss": 0.4477, "step": 4835 }, { "epoch": 0.91, "learning_rate": 2.9683931782910603e-06, "loss": 0.4524, "step": 4836 }, { "epoch": 0.91, "learning_rate": 2.9676418024597542e-06, "loss": 0.4534, "step": 4837 }, { "epoch": 0.91, "learning_rate": 2.966890382851917e-06, "loss": 0.4608, "step": 4838 }, { "epoch": 0.91, "learning_rate": 2.9661389195378898e-06, "loss": 0.3834, "step": 4839 }, { "epoch": 0.91, "learning_rate": 2.965387412588017e-06, "loss": 0.4319, "step": 4840 }, { "epoch": 0.91, "learning_rate": 2.9646358620726503e-06, "loss": 0.3037, "step": 4841 }, { "epoch": 0.91, "learning_rate": 2.963884268062141e-06, "loss": 0.4604, "step": 4842 }, { "epoch": 0.91, "learning_rate": 2.9631326306268476e-06, "loss": 0.5422, "step": 4843 }, { "epoch": 0.92, "learning_rate": 2.9623809498371324e-06, "loss": 0.3821, "step": 4844 }, { "epoch": 0.92, "learning_rate": 2.9616292257633615e-06, "loss": 0.3047, "step": 4845 }, { "epoch": 0.92, "learning_rate": 2.9608774584759026e-06, "loss": 0.4948, "step": 4846 }, { "epoch": 0.92, "learning_rate": 2.9601256480451317e-06, "loss": 0.3725, "step": 4847 }, { "epoch": 0.92, "learning_rate": 2.9593737945414264e-06, "loss": 0.4653, "step": 4848 }, { "epoch": 0.92, "learning_rate": 2.958621898035167e-06, "loss": 0.3416, "step": 4849 }, { "epoch": 0.92, "learning_rate": 2.9578699585967417e-06, "loss": 0.5094, "step": 4850 }, { "epoch": 0.92, "learning_rate": 2.957117976296539e-06, "loss": 0.4535, "step": 4851 }, { "epoch": 0.92, "learning_rate": 2.9563659512049536e-06, "loss": 0.4502, "step": 4852 }, { "epoch": 0.92, "learning_rate": 2.955613883392383e-06, "loss": 0.3551, "step": 4853 }, { "epoch": 0.92, "learning_rate": 2.9548617729292296e-06, "loss": 0.4149, "step": 4854 }, { "epoch": 0.92, "learning_rate": 2.954109619885899e-06, "loss": 0.4594, "step": 4855 }, { "epoch": 0.92, "learning_rate": 2.9533574243328023e-06, "loss": 0.5419, "step": 4856 }, { "epoch": 0.92, "learning_rate": 2.952605186340352e-06, "loss": 0.5, "step": 4857 }, { "epoch": 0.92, "learning_rate": 2.951852905978966e-06, "loss": 0.3885, "step": 4858 }, { "epoch": 0.92, "learning_rate": 2.951100583319068e-06, "loss": 0.4666, "step": 4859 }, { "epoch": 0.92, "learning_rate": 2.9503482184310823e-06, "loss": 0.3361, "step": 4860 }, { "epoch": 0.92, "learning_rate": 2.9495958113854384e-06, "loss": 0.3304, "step": 4861 }, { "epoch": 0.92, "learning_rate": 2.948843362252572e-06, "loss": 0.4472, "step": 4862 }, { "epoch": 0.92, "learning_rate": 2.948090871102919e-06, "loss": 0.5199, "step": 4863 }, { "epoch": 0.92, "learning_rate": 2.9473383380069208e-06, "loss": 0.433, "step": 4864 }, { "epoch": 0.92, "learning_rate": 2.946585763035025e-06, "loss": 0.477, "step": 4865 }, { "epoch": 0.92, "learning_rate": 2.9458331462576793e-06, "loss": 0.3696, "step": 4866 }, { "epoch": 0.92, "learning_rate": 2.9450804877453378e-06, "loss": 0.4477, "step": 4867 }, { "epoch": 0.92, "learning_rate": 2.9443277875684574e-06, "loss": 0.4127, "step": 4868 }, { "epoch": 0.92, "learning_rate": 2.9435750457975002e-06, "loss": 0.4666, "step": 4869 }, { "epoch": 0.92, "learning_rate": 2.94282226250293e-06, "loss": 0.5559, "step": 4870 }, { "epoch": 0.92, "learning_rate": 2.942069437755216e-06, "loss": 0.5755, "step": 4871 }, { "epoch": 0.92, "learning_rate": 2.9413165716248327e-06, "loss": 0.3871, "step": 4872 }, { "epoch": 0.92, "learning_rate": 2.940563664182255e-06, "loss": 0.4068, "step": 4873 }, { "epoch": 0.92, "learning_rate": 2.9398107154979637e-06, "loss": 0.4488, "step": 4874 }, { "epoch": 0.92, "learning_rate": 2.9390577256424442e-06, "loss": 0.4875, "step": 4875 }, { "epoch": 0.92, "learning_rate": 2.938304694686184e-06, "loss": 0.3239, "step": 4876 }, { "epoch": 0.92, "learning_rate": 2.9375516226996764e-06, "loss": 0.5135, "step": 4877 }, { "epoch": 0.92, "learning_rate": 2.936798509753415e-06, "loss": 0.4682, "step": 4878 }, { "epoch": 0.92, "learning_rate": 2.936045355917903e-06, "loss": 0.5251, "step": 4879 }, { "epoch": 0.92, "learning_rate": 2.9352921612636415e-06, "loss": 0.4375, "step": 4880 }, { "epoch": 0.92, "learning_rate": 2.9345389258611383e-06, "loss": 0.4861, "step": 4881 }, { "epoch": 0.92, "learning_rate": 2.933785649780906e-06, "loss": 0.3491, "step": 4882 }, { "epoch": 0.92, "learning_rate": 2.9330323330934586e-06, "loss": 0.553, "step": 4883 }, { "epoch": 0.92, "learning_rate": 2.932278975869315e-06, "loss": 0.5816, "step": 4884 }, { "epoch": 0.92, "learning_rate": 2.931525578178998e-06, "loss": 0.3919, "step": 4885 }, { "epoch": 0.92, "learning_rate": 2.930772140093035e-06, "loss": 0.554, "step": 4886 }, { "epoch": 0.92, "learning_rate": 2.9300186616819553e-06, "loss": 0.6986, "step": 4887 }, { "epoch": 0.92, "learning_rate": 2.9292651430162935e-06, "loss": 0.3656, "step": 4888 }, { "epoch": 0.92, "learning_rate": 2.928511584166587e-06, "loss": 0.3978, "step": 4889 }, { "epoch": 0.92, "learning_rate": 2.9277579852033774e-06, "loss": 0.4103, "step": 4890 }, { "epoch": 0.92, "learning_rate": 2.9270043461972104e-06, "loss": 0.477, "step": 4891 }, { "epoch": 0.92, "learning_rate": 2.9262506672186346e-06, "loss": 0.3894, "step": 4892 }, { "epoch": 0.92, "learning_rate": 2.9254969483382035e-06, "loss": 0.3672, "step": 4893 }, { "epoch": 0.92, "learning_rate": 2.9247431896264728e-06, "loss": 0.3805, "step": 4894 }, { "epoch": 0.92, "learning_rate": 2.9239893911540033e-06, "loss": 0.3456, "step": 4895 }, { "epoch": 0.92, "learning_rate": 2.9232355529913585e-06, "loss": 0.4918, "step": 4896 }, { "epoch": 0.93, "learning_rate": 2.922481675209108e-06, "loss": 0.4895, "step": 4897 }, { "epoch": 0.93, "learning_rate": 2.921727757877821e-06, "loss": 0.5227, "step": 4898 }, { "epoch": 0.93, "learning_rate": 2.920973801068073e-06, "loss": 0.4156, "step": 4899 }, { "epoch": 0.93, "learning_rate": 2.9202198048504434e-06, "loss": 0.5189, "step": 4900 }, { "epoch": 0.93, "learning_rate": 2.9194657692955154e-06, "loss": 0.437, "step": 4901 }, { "epoch": 0.93, "learning_rate": 2.9187116944738734e-06, "loss": 0.3909, "step": 4902 }, { "epoch": 0.93, "learning_rate": 2.9179575804561087e-06, "loss": 0.4893, "step": 4903 }, { "epoch": 0.93, "learning_rate": 2.9172034273128157e-06, "loss": 0.3838, "step": 4904 }, { "epoch": 0.93, "learning_rate": 2.916449235114589e-06, "loss": 0.4988, "step": 4905 }, { "epoch": 0.93, "learning_rate": 2.9156950039320307e-06, "loss": 0.5389, "step": 4906 }, { "epoch": 0.93, "learning_rate": 2.9149407338357467e-06, "loss": 0.464, "step": 4907 }, { "epoch": 0.93, "learning_rate": 2.9141864248963433e-06, "loss": 0.3761, "step": 4908 }, { "epoch": 0.93, "learning_rate": 2.9134320771844327e-06, "loss": 0.3141, "step": 4909 }, { "epoch": 0.93, "learning_rate": 2.91267769077063e-06, "loss": 0.4461, "step": 4910 }, { "epoch": 0.93, "learning_rate": 2.9119232657255566e-06, "loss": 0.3789, "step": 4911 }, { "epoch": 0.93, "learning_rate": 2.911168802119832e-06, "loss": 0.4399, "step": 4912 }, { "epoch": 0.93, "learning_rate": 2.9104143000240836e-06, "loss": 0.4105, "step": 4913 }, { "epoch": 0.93, "learning_rate": 2.909659759508943e-06, "loss": 0.5088, "step": 4914 }, { "epoch": 0.93, "learning_rate": 2.9089051806450414e-06, "loss": 0.4941, "step": 4915 }, { "epoch": 0.93, "learning_rate": 2.908150563503016e-06, "loss": 0.4086, "step": 4916 }, { "epoch": 0.93, "learning_rate": 2.9073959081535085e-06, "loss": 0.3686, "step": 4917 }, { "epoch": 0.93, "learning_rate": 2.9066412146671643e-06, "loss": 0.4951, "step": 4918 }, { "epoch": 0.93, "learning_rate": 2.905886483114628e-06, "loss": 0.5068, "step": 4919 }, { "epoch": 0.93, "learning_rate": 2.9051317135665525e-06, "loss": 0.4334, "step": 4920 }, { "epoch": 0.93, "learning_rate": 2.9043769060935933e-06, "loss": 0.4234, "step": 4921 }, { "epoch": 0.93, "learning_rate": 2.903622060766409e-06, "loss": 0.401, "step": 4922 }, { "epoch": 0.93, "learning_rate": 2.9028671776556607e-06, "loss": 0.5327, "step": 4923 }, { "epoch": 0.93, "learning_rate": 2.902112256832014e-06, "loss": 0.4091, "step": 4924 }, { "epoch": 0.93, "learning_rate": 2.9013572983661382e-06, "loss": 0.419, "step": 4925 }, { "epoch": 0.93, "learning_rate": 2.9006023023287063e-06, "loss": 0.4755, "step": 4926 }, { "epoch": 0.93, "learning_rate": 2.8998472687903932e-06, "loss": 0.4405, "step": 4927 }, { "epoch": 0.93, "learning_rate": 2.8990921978218806e-06, "loss": 0.3895, "step": 4928 }, { "epoch": 0.93, "learning_rate": 2.8983370894938495e-06, "loss": 0.4228, "step": 4929 }, { "epoch": 0.93, "learning_rate": 2.897581943876988e-06, "loss": 0.4568, "step": 4930 }, { "epoch": 0.93, "learning_rate": 2.896826761041985e-06, "loss": 0.5361, "step": 4931 }, { "epoch": 0.93, "learning_rate": 2.896071541059536e-06, "loss": 0.4443, "step": 4932 }, { "epoch": 0.93, "learning_rate": 2.895316284000336e-06, "loss": 0.5778, "step": 4933 }, { "epoch": 0.93, "learning_rate": 2.8945609899350863e-06, "loss": 0.5195, "step": 4934 }, { "epoch": 0.93, "learning_rate": 2.8938056589344913e-06, "loss": 0.3848, "step": 4935 }, { "epoch": 0.93, "learning_rate": 2.893050291069258e-06, "loss": 0.5276, "step": 4936 }, { "epoch": 0.93, "learning_rate": 2.892294886410098e-06, "loss": 0.5324, "step": 4937 }, { "epoch": 0.93, "learning_rate": 2.891539445027725e-06, "loss": 0.397, "step": 4938 }, { "epoch": 0.93, "learning_rate": 2.8907839669928573e-06, "loss": 0.5169, "step": 4939 }, { "epoch": 0.93, "learning_rate": 2.890028452376215e-06, "loss": 0.3649, "step": 4940 }, { "epoch": 0.93, "learning_rate": 2.8892729012485244e-06, "loss": 0.3172, "step": 4941 }, { "epoch": 0.93, "learning_rate": 2.8885173136805126e-06, "loss": 0.3931, "step": 4942 }, { "epoch": 0.93, "learning_rate": 2.887761689742912e-06, "loss": 0.4412, "step": 4943 }, { "epoch": 0.93, "learning_rate": 2.8870060295064556e-06, "loss": 0.598, "step": 4944 }, { "epoch": 0.93, "learning_rate": 2.8862503330418836e-06, "loss": 0.4, "step": 4945 }, { "epoch": 0.93, "learning_rate": 2.8854946004199365e-06, "loss": 0.4057, "step": 4946 }, { "epoch": 0.93, "learning_rate": 2.88473883171136e-06, "loss": 0.5416, "step": 4947 }, { "epoch": 0.93, "learning_rate": 2.8839830269869025e-06, "loss": 0.4233, "step": 4948 }, { "epoch": 0.94, "learning_rate": 2.8832271863173163e-06, "loss": 0.3276, "step": 4949 }, { "epoch": 0.94, "learning_rate": 2.8824713097733554e-06, "loss": 0.4704, "step": 4950 }, { "epoch": 0.94, "learning_rate": 2.881715397425778e-06, "loss": 0.5046, "step": 4951 }, { "epoch": 0.94, "learning_rate": 2.8809594493453487e-06, "loss": 0.4009, "step": 4952 }, { "epoch": 0.94, "learning_rate": 2.8802034656028305e-06, "loss": 0.4556, "step": 4953 }, { "epoch": 0.94, "learning_rate": 2.8794474462689913e-06, "loss": 0.3669, "step": 4954 }, { "epoch": 0.94, "learning_rate": 2.878691391414605e-06, "loss": 0.4186, "step": 4955 }, { "epoch": 0.94, "learning_rate": 2.8779353011104472e-06, "loss": 0.5368, "step": 4956 }, { "epoch": 0.94, "learning_rate": 2.877179175427294e-06, "loss": 0.4439, "step": 4957 }, { "epoch": 0.94, "learning_rate": 2.8764230144359284e-06, "loss": 0.5131, "step": 4958 }, { "epoch": 0.94, "learning_rate": 2.8756668182071364e-06, "loss": 0.5132, "step": 4959 }, { "epoch": 0.94, "learning_rate": 2.874910586811705e-06, "loss": 0.594, "step": 4960 }, { "epoch": 0.94, "learning_rate": 2.8741543203204273e-06, "loss": 0.3968, "step": 4961 }, { "epoch": 0.94, "learning_rate": 2.873398018804098e-06, "loss": 0.4192, "step": 4962 }, { "epoch": 0.94, "learning_rate": 2.8726416823335146e-06, "loss": 0.4257, "step": 4963 }, { "epoch": 0.94, "learning_rate": 2.87188531097948e-06, "loss": 0.3849, "step": 4964 }, { "epoch": 0.94, "learning_rate": 2.8711289048127977e-06, "loss": 0.5466, "step": 4965 }, { "epoch": 0.94, "learning_rate": 2.870372463904277e-06, "loss": 0.407, "step": 4966 }, { "epoch": 0.94, "learning_rate": 2.8696159883247287e-06, "loss": 0.5081, "step": 4967 }, { "epoch": 0.94, "learning_rate": 2.868859478144967e-06, "loss": 0.4552, "step": 4968 }, { "epoch": 0.94, "learning_rate": 2.86810293343581e-06, "loss": 0.4079, "step": 4969 }, { "epoch": 0.94, "learning_rate": 2.8673463542680807e-06, "loss": 0.3034, "step": 4970 }, { "epoch": 0.94, "learning_rate": 2.8665897407126e-06, "loss": 0.3938, "step": 4971 }, { "epoch": 0.94, "learning_rate": 2.8658330928401986e-06, "loss": 0.5071, "step": 4972 }, { "epoch": 0.94, "learning_rate": 2.8650764107217048e-06, "loss": 0.3947, "step": 4973 }, { "epoch": 0.94, "learning_rate": 2.864319694427954e-06, "loss": 0.4676, "step": 4974 }, { "epoch": 0.94, "learning_rate": 2.8635629440297828e-06, "loss": 0.5207, "step": 4975 }, { "epoch": 0.94, "learning_rate": 2.8628061595980323e-06, "loss": 0.4795, "step": 4976 }, { "epoch": 0.94, "learning_rate": 2.8620493412035454e-06, "loss": 0.3912, "step": 4977 }, { "epoch": 0.94, "learning_rate": 2.8612924889171684e-06, "loss": 0.4601, "step": 4978 }, { "epoch": 0.94, "learning_rate": 2.860535602809752e-06, "loss": 0.4001, "step": 4979 }, { "epoch": 0.94, "learning_rate": 2.8597786829521486e-06, "loss": 0.5579, "step": 4980 }, { "epoch": 0.94, "learning_rate": 2.8590217294152143e-06, "loss": 0.3778, "step": 4981 }, { "epoch": 0.94, "learning_rate": 2.8582647422698097e-06, "loss": 0.5652, "step": 4982 }, { "epoch": 0.94, "learning_rate": 2.8575077215867957e-06, "loss": 0.3778, "step": 4983 }, { "epoch": 0.94, "learning_rate": 2.85675066743704e-06, "loss": 0.41, "step": 4984 }, { "epoch": 0.94, "learning_rate": 2.855993579891409e-06, "loss": 0.4828, "step": 4985 }, { "epoch": 0.94, "learning_rate": 2.855236459020775e-06, "loss": 0.4684, "step": 4986 }, { "epoch": 0.94, "learning_rate": 2.854479304896015e-06, "loss": 0.3446, "step": 4987 }, { "epoch": 0.94, "learning_rate": 2.8537221175880064e-06, "loss": 0.4687, "step": 4988 }, { "epoch": 0.94, "learning_rate": 2.8529648971676287e-06, "loss": 0.4107, "step": 4989 }, { "epoch": 0.94, "learning_rate": 2.8522076437057685e-06, "loss": 0.4003, "step": 4990 }, { "epoch": 0.94, "learning_rate": 2.851450357273312e-06, "loss": 0.2547, "step": 4991 }, { "epoch": 0.94, "learning_rate": 2.850693037941149e-06, "loss": 0.4275, "step": 4992 }, { "epoch": 0.94, "learning_rate": 2.8499356857801747e-06, "loss": 0.4737, "step": 4993 }, { "epoch": 0.94, "learning_rate": 2.849178300861285e-06, "loss": 0.5082, "step": 4994 }, { "epoch": 0.94, "learning_rate": 2.84842088325538e-06, "loss": 0.4301, "step": 4995 }, { "epoch": 0.94, "learning_rate": 2.8476634330333618e-06, "loss": 0.3886, "step": 4996 }, { "epoch": 0.94, "learning_rate": 2.8469059502661373e-06, "loss": 0.4114, "step": 4997 }, { "epoch": 0.94, "learning_rate": 2.846148435024615e-06, "loss": 0.4419, "step": 4998 }, { "epoch": 0.94, "learning_rate": 2.845390887379706e-06, "loss": 0.4618, "step": 4999 }, { "epoch": 0.94, "learning_rate": 2.844633307402326e-06, "loss": 0.4428, "step": 5000 }, { "epoch": 0.94, "learning_rate": 2.8438756951633935e-06, "loss": 0.4362, "step": 5001 }, { "epoch": 0.95, "learning_rate": 2.8431180507338287e-06, "loss": 0.5469, "step": 5002 }, { "epoch": 0.95, "learning_rate": 2.8423603741845555e-06, "loss": 0.464, "step": 5003 }, { "epoch": 0.95, "learning_rate": 2.841602665586502e-06, "loss": 0.6158, "step": 5004 }, { "epoch": 0.95, "learning_rate": 2.840844925010598e-06, "loss": 0.4012, "step": 5005 }, { "epoch": 0.95, "learning_rate": 2.840087152527775e-06, "loss": 0.4654, "step": 5006 }, { "epoch": 0.95, "learning_rate": 2.8393293482089713e-06, "loss": 0.4299, "step": 5007 }, { "epoch": 0.95, "learning_rate": 2.838571512125124e-06, "loss": 0.3646, "step": 5008 }, { "epoch": 0.95, "learning_rate": 2.837813644347176e-06, "loss": 0.4854, "step": 5009 }, { "epoch": 0.95, "learning_rate": 2.837055744946072e-06, "loss": 0.4864, "step": 5010 }, { "epoch": 0.95, "learning_rate": 2.836297813992761e-06, "loss": 0.3568, "step": 5011 }, { "epoch": 0.95, "learning_rate": 2.835539851558192e-06, "loss": 0.3453, "step": 5012 }, { "epoch": 0.95, "learning_rate": 2.83478185771332e-06, "loss": 0.5068, "step": 5013 }, { "epoch": 0.95, "learning_rate": 2.8340238325291016e-06, "loss": 0.338, "step": 5014 }, { "epoch": 0.95, "learning_rate": 2.833265776076496e-06, "loss": 0.5079, "step": 5015 }, { "epoch": 0.95, "learning_rate": 2.832507688426466e-06, "loss": 0.485, "step": 5016 }, { "epoch": 0.95, "learning_rate": 2.8317495696499776e-06, "loss": 0.3756, "step": 5017 }, { "epoch": 0.95, "learning_rate": 2.8309914198179983e-06, "loss": 0.4808, "step": 5018 }, { "epoch": 0.95, "learning_rate": 2.8302332390015014e-06, "loss": 0.3439, "step": 5019 }, { "epoch": 0.95, "learning_rate": 2.8294750272714587e-06, "loss": 0.4074, "step": 5020 }, { "epoch": 0.95, "learning_rate": 2.8287167846988484e-06, "loss": 0.4752, "step": 5021 }, { "epoch": 0.95, "learning_rate": 2.8279585113546517e-06, "loss": 0.5454, "step": 5022 }, { "epoch": 0.95, "learning_rate": 2.827200207309849e-06, "loss": 0.4617, "step": 5023 }, { "epoch": 0.95, "learning_rate": 2.8264418726354276e-06, "loss": 0.4544, "step": 5024 }, { "epoch": 0.95, "learning_rate": 2.825683507402377e-06, "loss": 0.406, "step": 5025 }, { "epoch": 0.95, "learning_rate": 2.8249251116816874e-06, "loss": 0.5627, "step": 5026 }, { "epoch": 0.95, "learning_rate": 2.8241666855443527e-06, "loss": 0.482, "step": 5027 }, { "epoch": 0.95, "learning_rate": 2.823408229061372e-06, "loss": 0.4252, "step": 5028 }, { "epoch": 0.95, "learning_rate": 2.8226497423037436e-06, "loss": 0.4566, "step": 5029 }, { "epoch": 0.95, "learning_rate": 2.8218912253424714e-06, "loss": 0.4203, "step": 5030 }, { "epoch": 0.95, "learning_rate": 2.8211326782485603e-06, "loss": 0.4577, "step": 5031 }, { "epoch": 0.95, "learning_rate": 2.8203741010930204e-06, "loss": 0.5893, "step": 5032 }, { "epoch": 0.95, "learning_rate": 2.819615493946861e-06, "loss": 0.4262, "step": 5033 }, { "epoch": 0.95, "learning_rate": 2.8188568568810977e-06, "loss": 0.3704, "step": 5034 }, { "epoch": 0.95, "learning_rate": 2.8180981899667465e-06, "loss": 0.4148, "step": 5035 }, { "epoch": 0.95, "learning_rate": 2.8173394932748284e-06, "loss": 0.5511, "step": 5036 }, { "epoch": 0.95, "learning_rate": 2.8165807668763646e-06, "loss": 0.3876, "step": 5037 }, { "epoch": 0.95, "learning_rate": 2.8158220108423804e-06, "loss": 0.4858, "step": 5038 }, { "epoch": 0.95, "learning_rate": 2.815063225243906e-06, "loss": 0.4122, "step": 5039 }, { "epoch": 0.95, "learning_rate": 2.814304410151971e-06, "loss": 0.3948, "step": 5040 }, { "epoch": 0.95, "learning_rate": 2.813545565637607e-06, "loss": 0.5081, "step": 5041 }, { "epoch": 0.95, "learning_rate": 2.8127866917718525e-06, "loss": 0.4448, "step": 5042 }, { "epoch": 0.95, "learning_rate": 2.812027788625748e-06, "loss": 0.4305, "step": 5043 }, { "epoch": 0.95, "learning_rate": 2.811268856270332e-06, "loss": 0.517, "step": 5044 }, { "epoch": 0.95, "learning_rate": 2.8105098947766514e-06, "loss": 0.4183, "step": 5045 }, { "epoch": 0.95, "learning_rate": 2.809750904215753e-06, "loss": 0.4547, "step": 5046 }, { "epoch": 0.95, "learning_rate": 2.8089918846586866e-06, "loss": 0.391, "step": 5047 }, { "epoch": 0.95, "learning_rate": 2.8082328361765053e-06, "loss": 0.4314, "step": 5048 }, { "epoch": 0.95, "learning_rate": 2.8074737588402644e-06, "loss": 0.4068, "step": 5049 }, { "epoch": 0.95, "learning_rate": 2.806714652721022e-06, "loss": 0.4144, "step": 5050 }, { "epoch": 0.95, "learning_rate": 2.805955517889839e-06, "loss": 0.4294, "step": 5051 }, { "epoch": 0.95, "learning_rate": 2.805196354417779e-06, "loss": 0.4601, "step": 5052 }, { "epoch": 0.95, "learning_rate": 2.8044371623759083e-06, "loss": 0.4435, "step": 5053 }, { "epoch": 0.95, "learning_rate": 2.803677941835296e-06, "loss": 0.429, "step": 5054 }, { "epoch": 0.96, "learning_rate": 2.802918692867013e-06, "loss": 0.4743, "step": 5055 }, { "epoch": 0.96, "learning_rate": 2.802159415542134e-06, "loss": 0.4055, "step": 5056 }, { "epoch": 0.96, "learning_rate": 2.801400109931736e-06, "loss": 0.4181, "step": 5057 }, { "epoch": 0.96, "learning_rate": 2.800640776106899e-06, "loss": 0.4409, "step": 5058 }, { "epoch": 0.96, "learning_rate": 2.799881414138703e-06, "loss": 0.3575, "step": 5059 }, { "epoch": 0.96, "learning_rate": 2.7991220240982358e-06, "loss": 0.6258, "step": 5060 }, { "epoch": 0.96, "learning_rate": 2.7983626060565833e-06, "loss": 0.457, "step": 5061 }, { "epoch": 0.96, "learning_rate": 2.797603160084836e-06, "loss": 0.5441, "step": 5062 }, { "epoch": 0.96, "learning_rate": 2.7968436862540852e-06, "loss": 0.456, "step": 5063 }, { "epoch": 0.96, "learning_rate": 2.7960841846354285e-06, "loss": 0.4183, "step": 5064 }, { "epoch": 0.96, "learning_rate": 2.795324655299962e-06, "loss": 0.3205, "step": 5065 }, { "epoch": 0.96, "learning_rate": 2.7945650983187873e-06, "loss": 0.4546, "step": 5066 }, { "epoch": 0.96, "learning_rate": 2.793805513763006e-06, "loss": 0.5422, "step": 5067 }, { "epoch": 0.96, "learning_rate": 2.7930459017037257e-06, "loss": 0.3726, "step": 5068 }, { "epoch": 0.96, "learning_rate": 2.7922862622120533e-06, "loss": 0.4802, "step": 5069 }, { "epoch": 0.96, "learning_rate": 2.791526595359099e-06, "loss": 0.4253, "step": 5070 }, { "epoch": 0.96, "learning_rate": 2.7907669012159787e-06, "loss": 0.4747, "step": 5071 }, { "epoch": 0.96, "learning_rate": 2.7900071798538065e-06, "loss": 0.4615, "step": 5072 }, { "epoch": 0.96, "learning_rate": 2.7892474313437003e-06, "loss": 0.5528, "step": 5073 }, { "epoch": 0.96, "learning_rate": 2.788487655756782e-06, "loss": 0.4832, "step": 5074 }, { "epoch": 0.96, "learning_rate": 2.7877278531641755e-06, "loss": 0.374, "step": 5075 }, { "epoch": 0.96, "learning_rate": 2.7869680236370055e-06, "loss": 0.5192, "step": 5076 }, { "epoch": 0.96, "learning_rate": 2.7862081672464025e-06, "loss": 0.5869, "step": 5077 }, { "epoch": 0.96, "learning_rate": 2.785448284063497e-06, "loss": 0.4006, "step": 5078 }, { "epoch": 0.96, "learning_rate": 2.7846883741594207e-06, "loss": 0.4798, "step": 5079 }, { "epoch": 0.96, "learning_rate": 2.783928437605312e-06, "loss": 0.4289, "step": 5080 }, { "epoch": 0.96, "learning_rate": 2.7831684744723086e-06, "loss": 0.4045, "step": 5081 }, { "epoch": 0.96, "learning_rate": 2.782408484831552e-06, "loss": 0.4284, "step": 5082 }, { "epoch": 0.96, "learning_rate": 2.781648468754185e-06, "loss": 0.4982, "step": 5083 }, { "epoch": 0.96, "learning_rate": 2.7808884263113546e-06, "loss": 0.4582, "step": 5084 }, { "epoch": 0.96, "learning_rate": 2.780128357574209e-06, "loss": 0.4546, "step": 5085 }, { "epoch": 0.96, "learning_rate": 2.7793682626138984e-06, "loss": 0.4727, "step": 5086 }, { "epoch": 0.96, "learning_rate": 2.7786081415015775e-06, "loss": 0.4475, "step": 5087 }, { "epoch": 0.96, "learning_rate": 2.7778479943084006e-06, "loss": 0.4354, "step": 5088 }, { "epoch": 0.96, "learning_rate": 2.777087821105528e-06, "loss": 0.3784, "step": 5089 }, { "epoch": 0.96, "learning_rate": 2.7763276219641193e-06, "loss": 0.475, "step": 5090 }, { "epoch": 0.96, "learning_rate": 2.775567396955337e-06, "loss": 0.4223, "step": 5091 }, { "epoch": 0.96, "learning_rate": 2.774807146150349e-06, "loss": 0.3259, "step": 5092 }, { "epoch": 0.96, "learning_rate": 2.7740468696203205e-06, "loss": 0.4142, "step": 5093 }, { "epoch": 0.96, "learning_rate": 2.7732865674364234e-06, "loss": 0.4668, "step": 5094 }, { "epoch": 0.96, "learning_rate": 2.772526239669831e-06, "loss": 0.5103, "step": 5095 }, { "epoch": 0.96, "learning_rate": 2.771765886391718e-06, "loss": 0.4427, "step": 5096 }, { "epoch": 0.96, "learning_rate": 2.7710055076732616e-06, "loss": 0.4711, "step": 5097 }, { "epoch": 0.96, "learning_rate": 2.7702451035856424e-06, "loss": 0.4689, "step": 5098 }, { "epoch": 0.96, "learning_rate": 2.769484674200043e-06, "loss": 0.4363, "step": 5099 }, { "epoch": 0.96, "learning_rate": 2.7687242195876475e-06, "loss": 0.4484, "step": 5100 }, { "epoch": 0.96, "learning_rate": 2.7679637398196427e-06, "loss": 0.479, "step": 5101 }, { "epoch": 0.96, "learning_rate": 2.767203234967219e-06, "loss": 0.4444, "step": 5102 }, { "epoch": 0.96, "learning_rate": 2.7664427051015674e-06, "loss": 0.6319, "step": 5103 }, { "epoch": 0.96, "learning_rate": 2.765682150293883e-06, "loss": 0.5306, "step": 5104 }, { "epoch": 0.96, "learning_rate": 2.7649215706153602e-06, "loss": 0.623, "step": 5105 }, { "epoch": 0.96, "learning_rate": 2.7641609661372013e-06, "loss": 0.3114, "step": 5106 }, { "epoch": 0.96, "learning_rate": 2.7634003369306046e-06, "loss": 0.3188, "step": 5107 }, { "epoch": 0.97, "learning_rate": 2.762639683066774e-06, "loss": 0.3566, "step": 5108 }, { "epoch": 0.97, "learning_rate": 2.7618790046169164e-06, "loss": 0.5636, "step": 5109 }, { "epoch": 0.97, "learning_rate": 2.7611183016522386e-06, "loss": 0.4114, "step": 5110 }, { "epoch": 0.97, "learning_rate": 2.760357574243951e-06, "loss": 0.4716, "step": 5111 }, { "epoch": 0.97, "learning_rate": 2.7595968224632672e-06, "loss": 0.3967, "step": 5112 }, { "epoch": 0.97, "learning_rate": 2.7588360463814023e-06, "loss": 0.4029, "step": 5113 }, { "epoch": 0.97, "learning_rate": 2.7580752460695716e-06, "loss": 0.304, "step": 5114 }, { "epoch": 0.97, "learning_rate": 2.7573144215989962e-06, "loss": 0.4165, "step": 5115 }, { "epoch": 0.97, "learning_rate": 2.7565535730408978e-06, "loss": 0.3919, "step": 5116 }, { "epoch": 0.97, "learning_rate": 2.7557927004664993e-06, "loss": 0.4586, "step": 5117 }, { "epoch": 0.97, "learning_rate": 2.755031803947028e-06, "loss": 0.5586, "step": 5118 }, { "epoch": 0.97, "learning_rate": 2.7542708835537118e-06, "loss": 0.3818, "step": 5119 }, { "epoch": 0.97, "learning_rate": 2.7535099393577818e-06, "loss": 0.2294, "step": 5120 }, { "epoch": 0.97, "learning_rate": 2.7527489714304704e-06, "loss": 0.3988, "step": 5121 }, { "epoch": 0.97, "learning_rate": 2.751987979843014e-06, "loss": 0.361, "step": 5122 }, { "epoch": 0.97, "learning_rate": 2.751226964666648e-06, "loss": 0.4877, "step": 5123 }, { "epoch": 0.97, "learning_rate": 2.7504659259726133e-06, "loss": 0.4703, "step": 5124 }, { "epoch": 0.97, "learning_rate": 2.749704863832151e-06, "loss": 0.4267, "step": 5125 }, { "epoch": 0.97, "learning_rate": 2.7489437783165064e-06, "loss": 0.6022, "step": 5126 }, { "epoch": 0.97, "learning_rate": 2.7481826694969245e-06, "loss": 0.4629, "step": 5127 }, { "epoch": 0.97, "learning_rate": 2.747421537444653e-06, "loss": 0.4744, "step": 5128 }, { "epoch": 0.97, "learning_rate": 2.746660382230944e-06, "loss": 0.3374, "step": 5129 }, { "epoch": 0.97, "learning_rate": 2.7458992039270505e-06, "loss": 0.6072, "step": 5130 }, { "epoch": 0.97, "learning_rate": 2.745138002604225e-06, "loss": 0.4039, "step": 5131 }, { "epoch": 0.97, "learning_rate": 2.7443767783337257e-06, "loss": 0.3421, "step": 5132 }, { "epoch": 0.97, "learning_rate": 2.7436155311868123e-06, "loss": 0.5792, "step": 5133 }, { "epoch": 0.97, "learning_rate": 2.742854261234746e-06, "loss": 0.3441, "step": 5134 }, { "epoch": 0.97, "learning_rate": 2.7420929685487896e-06, "loss": 0.4645, "step": 5135 }, { "epoch": 0.97, "learning_rate": 2.741331653200209e-06, "loss": 0.5114, "step": 5136 }, { "epoch": 0.97, "learning_rate": 2.740570315260272e-06, "loss": 0.4038, "step": 5137 }, { "epoch": 0.97, "learning_rate": 2.7398089548002476e-06, "loss": 0.482, "step": 5138 }, { "epoch": 0.97, "learning_rate": 2.739047571891409e-06, "loss": 0.4524, "step": 5139 }, { "epoch": 0.97, "learning_rate": 2.7382861666050285e-06, "loss": 0.4897, "step": 5140 }, { "epoch": 0.97, "learning_rate": 2.737524739012385e-06, "loss": 0.3851, "step": 5141 }, { "epoch": 0.97, "learning_rate": 2.736763289184754e-06, "loss": 0.5294, "step": 5142 }, { "epoch": 0.97, "learning_rate": 2.736001817193416e-06, "loss": 0.506, "step": 5143 }, { "epoch": 0.97, "learning_rate": 2.735240323109656e-06, "loss": 0.4011, "step": 5144 }, { "epoch": 0.97, "learning_rate": 2.734478807004755e-06, "loss": 0.4774, "step": 5145 }, { "epoch": 0.97, "learning_rate": 2.73371726895e-06, "loss": 0.4417, "step": 5146 }, { "epoch": 0.97, "learning_rate": 2.732955709016682e-06, "loss": 0.4195, "step": 5147 }, { "epoch": 0.97, "learning_rate": 2.73219412727609e-06, "loss": 0.4236, "step": 5148 }, { "epoch": 0.97, "learning_rate": 2.731432523799516e-06, "loss": 0.4254, "step": 5149 }, { "epoch": 0.97, "learning_rate": 2.730670898658255e-06, "loss": 0.4377, "step": 5150 }, { "epoch": 0.97, "learning_rate": 2.729909251923605e-06, "loss": 0.4788, "step": 5151 }, { "epoch": 0.97, "learning_rate": 2.729147583666863e-06, "loss": 0.473, "step": 5152 }, { "epoch": 0.97, "learning_rate": 2.72838589395933e-06, "loss": 0.3106, "step": 5153 }, { "epoch": 0.97, "learning_rate": 2.72762418287231e-06, "loss": 0.4519, "step": 5154 }, { "epoch": 0.97, "learning_rate": 2.726862450477106e-06, "loss": 0.476, "step": 5155 }, { "epoch": 0.97, "learning_rate": 2.726100696845026e-06, "loss": 0.4374, "step": 5156 }, { "epoch": 0.97, "learning_rate": 2.7253389220473774e-06, "loss": 0.4289, "step": 5157 }, { "epoch": 0.97, "learning_rate": 2.7245771261554733e-06, "loss": 0.4509, "step": 5158 }, { "epoch": 0.97, "learning_rate": 2.723815309240624e-06, "loss": 0.3902, "step": 5159 }, { "epoch": 0.97, "learning_rate": 2.7230534713741443e-06, "loss": 0.4053, "step": 5160 }, { "epoch": 0.98, "learning_rate": 2.722291612627352e-06, "loss": 0.5104, "step": 5161 }, { "epoch": 0.98, "learning_rate": 2.7215297330715657e-06, "loss": 0.4031, "step": 5162 }, { "epoch": 0.98, "learning_rate": 2.720767832778104e-06, "loss": 0.4651, "step": 5163 }, { "epoch": 0.98, "learning_rate": 2.7200059118182907e-06, "loss": 0.4232, "step": 5164 }, { "epoch": 0.98, "learning_rate": 2.719243970263451e-06, "loss": 0.4376, "step": 5165 }, { "epoch": 0.98, "learning_rate": 2.7184820081849097e-06, "loss": 0.3242, "step": 5166 }, { "epoch": 0.98, "learning_rate": 2.7177200256539954e-06, "loss": 0.5695, "step": 5167 }, { "epoch": 0.98, "learning_rate": 2.7169580227420385e-06, "loss": 0.5227, "step": 5168 }, { "epoch": 0.98, "learning_rate": 2.7161959995203712e-06, "loss": 0.4328, "step": 5169 }, { "epoch": 0.98, "learning_rate": 2.715433956060327e-06, "loss": 0.4359, "step": 5170 }, { "epoch": 0.98, "learning_rate": 2.7146718924332422e-06, "loss": 0.4514, "step": 5171 }, { "epoch": 0.98, "learning_rate": 2.713909808710454e-06, "loss": 0.3909, "step": 5172 }, { "epoch": 0.98, "learning_rate": 2.713147704963302e-06, "loss": 0.6067, "step": 5173 }, { "epoch": 0.98, "learning_rate": 2.7123855812631284e-06, "loss": 0.5523, "step": 5174 }, { "epoch": 0.98, "learning_rate": 2.7116234376812756e-06, "loss": 0.3818, "step": 5175 }, { "epoch": 0.98, "learning_rate": 2.71086127428909e-06, "loss": 0.4002, "step": 5176 }, { "epoch": 0.98, "learning_rate": 2.7100990911579173e-06, "loss": 0.5331, "step": 5177 }, { "epoch": 0.98, "learning_rate": 2.7093368883591064e-06, "loss": 0.4537, "step": 5178 }, { "epoch": 0.98, "learning_rate": 2.7085746659640104e-06, "loss": 0.4917, "step": 5179 }, { "epoch": 0.98, "learning_rate": 2.7078124240439795e-06, "loss": 0.2977, "step": 5180 }, { "epoch": 0.98, "learning_rate": 2.7070501626703683e-06, "loss": 0.5417, "step": 5181 }, { "epoch": 0.98, "learning_rate": 2.7062878819145336e-06, "loss": 0.4107, "step": 5182 }, { "epoch": 0.98, "learning_rate": 2.7055255818478348e-06, "loss": 0.4073, "step": 5183 }, { "epoch": 0.98, "learning_rate": 2.7047632625416287e-06, "loss": 0.4898, "step": 5184 }, { "epoch": 0.98, "learning_rate": 2.7040009240672792e-06, "loss": 0.4326, "step": 5185 }, { "epoch": 0.98, "learning_rate": 2.7032385664961497e-06, "loss": 0.3776, "step": 5186 }, { "epoch": 0.98, "learning_rate": 2.7024761898996043e-06, "loss": 0.3125, "step": 5187 }, { "epoch": 0.98, "learning_rate": 2.701713794349011e-06, "loss": 0.4791, "step": 5188 }, { "epoch": 0.98, "learning_rate": 2.7009513799157383e-06, "loss": 0.4852, "step": 5189 }, { "epoch": 0.98, "learning_rate": 2.7001889466711566e-06, "loss": 0.4749, "step": 5190 }, { "epoch": 0.98, "learning_rate": 2.6994264946866385e-06, "loss": 0.3799, "step": 5191 }, { "epoch": 0.98, "learning_rate": 2.698664024033557e-06, "loss": 0.589, "step": 5192 }, { "epoch": 0.98, "learning_rate": 2.69790153478329e-06, "loss": 0.3757, "step": 5193 }, { "epoch": 0.98, "learning_rate": 2.6971390270072135e-06, "loss": 0.4749, "step": 5194 }, { "epoch": 0.98, "learning_rate": 2.6963765007767068e-06, "loss": 0.4229, "step": 5195 }, { "epoch": 0.98, "learning_rate": 2.6956139561631524e-06, "loss": 0.5508, "step": 5196 }, { "epoch": 0.98, "learning_rate": 2.694851393237931e-06, "loss": 0.4633, "step": 5197 }, { "epoch": 0.98, "learning_rate": 2.694088812072428e-06, "loss": 0.3344, "step": 5198 }, { "epoch": 0.98, "learning_rate": 2.69332621273803e-06, "loss": 0.4491, "step": 5199 }, { "epoch": 0.98, "learning_rate": 2.6925635953061254e-06, "loss": 0.3425, "step": 5200 }, { "epoch": 0.98, "learning_rate": 2.691800959848101e-06, "loss": 0.4944, "step": 5201 }, { "epoch": 0.98, "learning_rate": 2.6910383064353508e-06, "loss": 0.3929, "step": 5202 }, { "epoch": 0.98, "learning_rate": 2.690275635139267e-06, "loss": 0.3654, "step": 5203 }, { "epoch": 0.98, "learning_rate": 2.689512946031244e-06, "loss": 0.4853, "step": 5204 }, { "epoch": 0.98, "learning_rate": 2.688750239182678e-06, "loss": 0.4094, "step": 5205 }, { "epoch": 0.98, "learning_rate": 2.6879875146649676e-06, "loss": 0.5665, "step": 5206 }, { "epoch": 0.98, "learning_rate": 2.6872247725495117e-06, "loss": 0.4698, "step": 5207 }, { "epoch": 0.98, "learning_rate": 2.6864620129077113e-06, "loss": 0.5156, "step": 5208 }, { "epoch": 0.98, "learning_rate": 2.68569923581097e-06, "loss": 0.3992, "step": 5209 }, { "epoch": 0.98, "learning_rate": 2.684936441330692e-06, "loss": 0.4667, "step": 5210 }, { "epoch": 0.98, "learning_rate": 2.6841736295382835e-06, "loss": 0.4059, "step": 5211 }, { "epoch": 0.98, "learning_rate": 2.6834108005051516e-06, "loss": 0.5376, "step": 5212 }, { "epoch": 0.98, "learning_rate": 2.6826479543027068e-06, "loss": 0.425, "step": 5213 }, { "epoch": 0.99, "learning_rate": 2.68188509100236e-06, "loss": 0.4666, "step": 5214 }, { "epoch": 0.99, "learning_rate": 2.681122210675523e-06, "loss": 0.5845, "step": 5215 }, { "epoch": 0.99, "learning_rate": 2.6803593133936097e-06, "loss": 0.4755, "step": 5216 }, { "epoch": 0.99, "learning_rate": 2.6795963992280373e-06, "loss": 0.4112, "step": 5217 }, { "epoch": 0.99, "learning_rate": 2.678833468250222e-06, "loss": 0.5249, "step": 5218 }, { "epoch": 0.99, "learning_rate": 2.678070520531583e-06, "loss": 0.5228, "step": 5219 }, { "epoch": 0.99, "learning_rate": 2.6773075561435412e-06, "loss": 0.562, "step": 5220 }, { "epoch": 0.99, "learning_rate": 2.676544575157518e-06, "loss": 0.5496, "step": 5221 }, { "epoch": 0.99, "learning_rate": 2.6757815776449375e-06, "loss": 0.5057, "step": 5222 }, { "epoch": 0.99, "learning_rate": 2.6750185636772248e-06, "loss": 0.3942, "step": 5223 }, { "epoch": 0.99, "learning_rate": 2.6742555333258056e-06, "loss": 0.4937, "step": 5224 }, { "epoch": 0.99, "learning_rate": 2.67349248666211e-06, "loss": 0.474, "step": 5225 }, { "epoch": 0.99, "learning_rate": 2.6727294237575663e-06, "loss": 0.353, "step": 5226 }, { "epoch": 0.99, "learning_rate": 2.6719663446836054e-06, "loss": 0.38, "step": 5227 }, { "epoch": 0.99, "learning_rate": 2.671203249511662e-06, "loss": 0.4122, "step": 5228 }, { "epoch": 0.99, "learning_rate": 2.670440138313169e-06, "loss": 0.4814, "step": 5229 }, { "epoch": 0.99, "learning_rate": 2.6696770111595617e-06, "loss": 0.3826, "step": 5230 }, { "epoch": 0.99, "learning_rate": 2.6689138681222793e-06, "loss": 0.4454, "step": 5231 }, { "epoch": 0.99, "learning_rate": 2.668150709272758e-06, "loss": 0.5595, "step": 5232 }, { "epoch": 0.99, "learning_rate": 2.66738753468244e-06, "loss": 0.4157, "step": 5233 }, { "epoch": 0.99, "learning_rate": 2.6666243444227667e-06, "loss": 0.4835, "step": 5234 }, { "epoch": 0.99, "learning_rate": 2.665861138565181e-06, "loss": 0.4959, "step": 5235 }, { "epoch": 0.99, "learning_rate": 2.6650979171811275e-06, "loss": 0.421, "step": 5236 }, { "epoch": 0.99, "learning_rate": 2.6643346803420523e-06, "loss": 0.5057, "step": 5237 }, { "epoch": 0.99, "learning_rate": 2.663571428119403e-06, "loss": 0.532, "step": 5238 }, { "epoch": 0.99, "learning_rate": 2.662808160584629e-06, "loss": 0.43, "step": 5239 }, { "epoch": 0.99, "learning_rate": 2.66204487780918e-06, "loss": 0.3901, "step": 5240 }, { "epoch": 0.99, "learning_rate": 2.6612815798645085e-06, "loss": 0.4192, "step": 5241 }, { "epoch": 0.99, "learning_rate": 2.6605182668220677e-06, "loss": 0.4013, "step": 5242 }, { "epoch": 0.99, "learning_rate": 2.659754938753312e-06, "loss": 0.5332, "step": 5243 }, { "epoch": 0.99, "learning_rate": 2.658991595729697e-06, "loss": 0.476, "step": 5244 }, { "epoch": 0.99, "learning_rate": 2.658228237822681e-06, "loss": 0.3745, "step": 5245 }, { "epoch": 0.99, "learning_rate": 2.657464865103723e-06, "loss": 0.4757, "step": 5246 }, { "epoch": 0.99, "learning_rate": 2.656701477644282e-06, "loss": 0.4333, "step": 5247 }, { "epoch": 0.99, "learning_rate": 2.655938075515821e-06, "loss": 0.4605, "step": 5248 }, { "epoch": 0.99, "learning_rate": 2.655174658789803e-06, "loss": 0.3802, "step": 5249 }, { "epoch": 0.99, "learning_rate": 2.6544112275376916e-06, "loss": 0.3783, "step": 5250 }, { "epoch": 0.99, "learning_rate": 2.653647781830953e-06, "loss": 0.4193, "step": 5251 }, { "epoch": 0.99, "learning_rate": 2.652884321741055e-06, "loss": 0.4032, "step": 5252 }, { "epoch": 0.99, "learning_rate": 2.652120847339464e-06, "loss": 0.4189, "step": 5253 }, { "epoch": 0.99, "learning_rate": 2.6513573586976517e-06, "loss": 0.5205, "step": 5254 }, { "epoch": 0.99, "learning_rate": 2.650593855887088e-06, "loss": 0.4582, "step": 5255 }, { "epoch": 0.99, "learning_rate": 2.649830338979247e-06, "loss": 0.4524, "step": 5256 }, { "epoch": 0.99, "learning_rate": 2.6490668080456005e-06, "loss": 0.4801, "step": 5257 }, { "epoch": 0.99, "learning_rate": 2.648303263157625e-06, "loss": 0.3863, "step": 5258 }, { "epoch": 0.99, "learning_rate": 2.647539704386796e-06, "loss": 0.6861, "step": 5259 }, { "epoch": 0.99, "learning_rate": 2.646776131804592e-06, "loss": 0.4377, "step": 5260 }, { "epoch": 0.99, "learning_rate": 2.6460125454824914e-06, "loss": 0.4374, "step": 5261 }, { "epoch": 0.99, "learning_rate": 2.6452489454919746e-06, "loss": 0.4394, "step": 5262 }, { "epoch": 0.99, "learning_rate": 2.644485331904523e-06, "loss": 0.4653, "step": 5263 }, { "epoch": 0.99, "learning_rate": 2.6437217047916198e-06, "loss": 0.4238, "step": 5264 }, { "epoch": 0.99, "learning_rate": 2.6429580642247478e-06, "loss": 0.4796, "step": 5265 }, { "epoch": 0.99, "learning_rate": 2.642194410275395e-06, "loss": 0.4406, "step": 5266 }, { "epoch": 1.0, "learning_rate": 2.6414307430150455e-06, "loss": 0.3747, "step": 5267 }, { "epoch": 1.0, "learning_rate": 2.640667062515188e-06, "loss": 0.4586, "step": 5268 }, { "epoch": 1.0, "learning_rate": 2.639903368847312e-06, "loss": 0.4465, "step": 5269 }, { "epoch": 1.0, "learning_rate": 2.6391396620829075e-06, "loss": 0.2852, "step": 5270 }, { "epoch": 1.0, "learning_rate": 2.6383759422934654e-06, "loss": 0.4075, "step": 5271 }, { "epoch": 1.0, "learning_rate": 2.6376122095504796e-06, "loss": 0.4511, "step": 5272 }, { "epoch": 1.0, "learning_rate": 2.6368484639254437e-06, "loss": 0.4188, "step": 5273 }, { "epoch": 1.0, "learning_rate": 2.636084705489852e-06, "loss": 0.3727, "step": 5274 }, { "epoch": 1.0, "learning_rate": 2.635320934315202e-06, "loss": 0.4238, "step": 5275 }, { "epoch": 1.0, "learning_rate": 2.6345571504729908e-06, "loss": 0.3717, "step": 5276 }, { "epoch": 1.0, "learning_rate": 2.6337933540347173e-06, "loss": 0.4474, "step": 5277 }, { "epoch": 1.0, "learning_rate": 2.633029545071881e-06, "loss": 0.4467, "step": 5278 }, { "epoch": 1.0, "learning_rate": 2.6322657236559834e-06, "loss": 0.3829, "step": 5279 }, { "epoch": 1.0, "learning_rate": 2.631501889858527e-06, "loss": 0.4402, "step": 5280 }, { "epoch": 1.0, "learning_rate": 2.630738043751015e-06, "loss": 0.348, "step": 5281 }, { "epoch": 1.0, "learning_rate": 2.629974185404951e-06, "loss": 0.3479, "step": 5282 }, { "epoch": 1.0, "learning_rate": 2.6292103148918428e-06, "loss": 0.4839, "step": 5283 }, { "epoch": 1.0, "learning_rate": 2.628446432283196e-06, "loss": 0.3916, "step": 5284 }, { "epoch": 1.0, "learning_rate": 2.627682537650518e-06, "loss": 0.401, "step": 5285 }, { "epoch": 1.0, "learning_rate": 2.6269186310653187e-06, "loss": 0.4074, "step": 5286 }, { "epoch": 1.0, "learning_rate": 2.6261547125991094e-06, "loss": 0.4017, "step": 5287 }, { "epoch": 1.0, "learning_rate": 2.6253907823233994e-06, "loss": 0.5081, "step": 5288 }, { "epoch": 1.0, "learning_rate": 2.6246268403097026e-06, "loss": 0.3752, "step": 5289 }, { "epoch": 1.0, "learning_rate": 2.6238628866295316e-06, "loss": 0.4931, "step": 5290 }, { "epoch": 1.0, "learning_rate": 2.623098921354402e-06, "loss": 0.4174, "step": 5291 }, { "epoch": 1.0, "learning_rate": 2.6223349445558294e-06, "loss": 0.4183, "step": 5292 }, { "epoch": 1.0, "learning_rate": 2.6215709563053297e-06, "loss": 0.4459, "step": 5293 }, { "epoch": 1.0, "learning_rate": 2.6208069566744214e-06, "loss": 0.2756, "step": 5294 }, { "epoch": 1.0, "learning_rate": 2.6200429457346238e-06, "loss": 0.2195, "step": 5295 }, { "epoch": 1.0, "learning_rate": 2.619278923557457e-06, "loss": 0.3458, "step": 5296 }, { "epoch": 1.0, "learning_rate": 2.618514890214441e-06, "loss": 0.2971, "step": 5297 }, { "epoch": 1.0, "learning_rate": 2.6177508457770986e-06, "loss": 0.333, "step": 5298 }, { "epoch": 1.0, "learning_rate": 2.6169867903169528e-06, "loss": 0.322, "step": 5299 }, { "epoch": 1.0, "learning_rate": 2.6162227239055278e-06, "loss": 0.2582, "step": 5300 }, { "epoch": 1.0, "learning_rate": 2.6154586466143495e-06, "loss": 0.2794, "step": 5301 }, { "epoch": 1.0, "learning_rate": 2.614694558514943e-06, "loss": 0.2556, "step": 5302 }, { "epoch": 1.0, "learning_rate": 2.613930459678836e-06, "loss": 0.2567, "step": 5303 }, { "epoch": 1.0, "learning_rate": 2.613166350177557e-06, "loss": 0.2215, "step": 5304 }, { "epoch": 1.0, "learning_rate": 2.6124022300826357e-06, "loss": 0.2147, "step": 5305 }, { "epoch": 1.0, "learning_rate": 2.6116380994656004e-06, "loss": 0.2896, "step": 5306 }, { "epoch": 1.0, "learning_rate": 2.610873958397985e-06, "loss": 0.2763, "step": 5307 }, { "epoch": 1.0, "learning_rate": 2.6101098069513196e-06, "loss": 0.2978, "step": 5308 }, { "epoch": 1.0, "learning_rate": 2.6093456451971376e-06, "loss": 0.2918, "step": 5309 }, { "epoch": 1.0, "learning_rate": 2.608581473206974e-06, "loss": 0.2351, "step": 5310 }, { "epoch": 1.0, "learning_rate": 2.6078172910523636e-06, "loss": 0.2222, "step": 5311 }, { "epoch": 1.0, "learning_rate": 2.6070530988048426e-06, "loss": 0.2686, "step": 5312 }, { "epoch": 1.0, "learning_rate": 2.6062888965359473e-06, "loss": 0.2769, "step": 5313 }, { "epoch": 1.0, "learning_rate": 2.6055246843172156e-06, "loss": 0.2624, "step": 5314 }, { "epoch": 1.0, "learning_rate": 2.604760462220188e-06, "loss": 0.1818, "step": 5315 }, { "epoch": 1.0, "learning_rate": 2.6039962303164024e-06, "loss": 0.2733, "step": 5316 }, { "epoch": 1.0, "learning_rate": 2.6032319886774e-06, "loss": 0.2447, "step": 5317 }, { "epoch": 1.0, "learning_rate": 2.602467737374723e-06, "loss": 0.3603, "step": 5318 }, { "epoch": 1.0, "learning_rate": 2.6017034764799135e-06, "loss": 0.231, "step": 5319 }, { "epoch": 1.01, "learning_rate": 2.6009392060645143e-06, "loss": 0.2945, "step": 5320 }, { "epoch": 1.01, "learning_rate": 2.600174926200071e-06, "loss": 0.2695, "step": 5321 }, { "epoch": 1.01, "learning_rate": 2.5994106369581286e-06, "loss": 0.2089, "step": 5322 }, { "epoch": 1.01, "learning_rate": 2.5986463384102313e-06, "loss": 0.209, "step": 5323 }, { "epoch": 1.01, "learning_rate": 2.597882030627928e-06, "loss": 0.2236, "step": 5324 }, { "epoch": 1.01, "learning_rate": 2.5971177136827664e-06, "loss": 0.3014, "step": 5325 }, { "epoch": 1.01, "learning_rate": 2.5963533876462944e-06, "loss": 0.2751, "step": 5326 }, { "epoch": 1.01, "learning_rate": 2.5955890525900616e-06, "loss": 0.2961, "step": 5327 }, { "epoch": 1.01, "learning_rate": 2.5948247085856186e-06, "loss": 0.227, "step": 5328 }, { "epoch": 1.01, "learning_rate": 2.594060355704517e-06, "loss": 0.3092, "step": 5329 }, { "epoch": 1.01, "learning_rate": 2.5932959940183085e-06, "loss": 0.2592, "step": 5330 }, { "epoch": 1.01, "learning_rate": 2.592531623598545e-06, "loss": 0.3401, "step": 5331 }, { "epoch": 1.01, "learning_rate": 2.5917672445167812e-06, "loss": 0.2077, "step": 5332 }, { "epoch": 1.01, "learning_rate": 2.5910028568445724e-06, "loss": 0.2978, "step": 5333 }, { "epoch": 1.01, "learning_rate": 2.5902384606534715e-06, "loss": 0.3682, "step": 5334 }, { "epoch": 1.01, "learning_rate": 2.589474056015036e-06, "loss": 0.1958, "step": 5335 }, { "epoch": 1.01, "learning_rate": 2.588709643000824e-06, "loss": 0.2271, "step": 5336 }, { "epoch": 1.01, "learning_rate": 2.5879452216823908e-06, "loss": 0.2652, "step": 5337 }, { "epoch": 1.01, "learning_rate": 2.587180792131296e-06, "loss": 0.2378, "step": 5338 }, { "epoch": 1.01, "learning_rate": 2.5864163544190994e-06, "loss": 0.2591, "step": 5339 }, { "epoch": 1.01, "learning_rate": 2.5856519086173597e-06, "loss": 0.2141, "step": 5340 }, { "epoch": 1.01, "learning_rate": 2.584887454797638e-06, "loss": 0.2079, "step": 5341 }, { "epoch": 1.01, "learning_rate": 2.584122993031496e-06, "loss": 0.2642, "step": 5342 }, { "epoch": 1.01, "learning_rate": 2.5833585233904954e-06, "loss": 0.291, "step": 5343 }, { "epoch": 1.01, "learning_rate": 2.5825940459462005e-06, "loss": 0.272, "step": 5344 }, { "epoch": 1.01, "learning_rate": 2.5818295607701745e-06, "loss": 0.333, "step": 5345 }, { "epoch": 1.01, "learning_rate": 2.5810650679339806e-06, "loss": 0.2468, "step": 5346 }, { "epoch": 1.01, "learning_rate": 2.5803005675091847e-06, "loss": 0.2612, "step": 5347 }, { "epoch": 1.01, "learning_rate": 2.579536059567353e-06, "loss": 0.3099, "step": 5348 }, { "epoch": 1.01, "learning_rate": 2.578771544180052e-06, "loss": 0.2403, "step": 5349 }, { "epoch": 1.01, "learning_rate": 2.578007021418848e-06, "loss": 0.2692, "step": 5350 }, { "epoch": 1.01, "learning_rate": 2.57724249135531e-06, "loss": 0.1562, "step": 5351 }, { "epoch": 1.01, "learning_rate": 2.576477954061006e-06, "loss": 0.228, "step": 5352 }, { "epoch": 1.01, "learning_rate": 2.575713409607506e-06, "loss": 0.2186, "step": 5353 }, { "epoch": 1.01, "learning_rate": 2.574948858066379e-06, "loss": 0.268, "step": 5354 }, { "epoch": 1.01, "learning_rate": 2.5741842995091957e-06, "loss": 0.2291, "step": 5355 }, { "epoch": 1.01, "learning_rate": 2.5734197340075284e-06, "loss": 0.1994, "step": 5356 }, { "epoch": 1.01, "learning_rate": 2.5726551616329486e-06, "loss": 0.3425, "step": 5357 }, { "epoch": 1.01, "learning_rate": 2.571890582457028e-06, "loss": 0.2336, "step": 5358 }, { "epoch": 1.01, "learning_rate": 2.5711259965513406e-06, "loss": 0.2705, "step": 5359 }, { "epoch": 1.01, "learning_rate": 2.5703614039874607e-06, "loss": 0.2554, "step": 5360 }, { "epoch": 1.01, "learning_rate": 2.569596804836961e-06, "loss": 0.2715, "step": 5361 }, { "epoch": 1.01, "learning_rate": 2.5688321991714186e-06, "loss": 0.2829, "step": 5362 }, { "epoch": 1.01, "learning_rate": 2.5680675870624084e-06, "loss": 0.2192, "step": 5363 }, { "epoch": 1.01, "learning_rate": 2.567302968581506e-06, "loss": 0.2411, "step": 5364 }, { "epoch": 1.01, "learning_rate": 2.5665383438002896e-06, "loss": 0.3137, "step": 5365 }, { "epoch": 1.01, "learning_rate": 2.5657737127903355e-06, "loss": 0.2437, "step": 5366 }, { "epoch": 1.01, "learning_rate": 2.5650090756232227e-06, "loss": 0.2313, "step": 5367 }, { "epoch": 1.01, "learning_rate": 2.564244432370529e-06, "loss": 0.2204, "step": 5368 }, { "epoch": 1.01, "learning_rate": 2.5634797831038337e-06, "loss": 0.1714, "step": 5369 }, { "epoch": 1.01, "learning_rate": 2.5627151278947175e-06, "loss": 0.1968, "step": 5370 }, { "epoch": 1.01, "learning_rate": 2.56195046681476e-06, "loss": 0.2636, "step": 5371 }, { "epoch": 1.01, "learning_rate": 2.5611857999355415e-06, "loss": 0.2339, "step": 5372 }, { "epoch": 1.02, "learning_rate": 2.5604211273286443e-06, "loss": 0.2446, "step": 5373 }, { "epoch": 1.02, "learning_rate": 2.559656449065651e-06, "loss": 0.2102, "step": 5374 }, { "epoch": 1.02, "learning_rate": 2.558891765218142e-06, "loss": 0.3123, "step": 5375 }, { "epoch": 1.02, "learning_rate": 2.5581270758577028e-06, "loss": 0.2759, "step": 5376 }, { "epoch": 1.02, "learning_rate": 2.5573623810559144e-06, "loss": 0.2269, "step": 5377 }, { "epoch": 1.02, "learning_rate": 2.556597680884363e-06, "loss": 0.2463, "step": 5378 }, { "epoch": 1.02, "learning_rate": 2.5558329754146316e-06, "loss": 0.3074, "step": 5379 }, { "epoch": 1.02, "learning_rate": 2.5550682647183064e-06, "loss": 0.2505, "step": 5380 }, { "epoch": 1.02, "learning_rate": 2.554303548866972e-06, "loss": 0.2376, "step": 5381 }, { "epoch": 1.02, "learning_rate": 2.553538827932215e-06, "loss": 0.2656, "step": 5382 }, { "epoch": 1.02, "learning_rate": 2.552774101985622e-06, "loss": 0.314, "step": 5383 }, { "epoch": 1.02, "learning_rate": 2.5520093710987785e-06, "loss": 0.2949, "step": 5384 }, { "epoch": 1.02, "learning_rate": 2.551244635343274e-06, "loss": 0.2584, "step": 5385 }, { "epoch": 1.02, "learning_rate": 2.550479894790695e-06, "loss": 0.2811, "step": 5386 }, { "epoch": 1.02, "learning_rate": 2.54971514951263e-06, "loss": 0.2795, "step": 5387 }, { "epoch": 1.02, "learning_rate": 2.5489503995806684e-06, "loss": 0.2028, "step": 5388 }, { "epoch": 1.02, "learning_rate": 2.548185645066399e-06, "loss": 0.2486, "step": 5389 }, { "epoch": 1.02, "learning_rate": 2.5474208860414107e-06, "loss": 0.245, "step": 5390 }, { "epoch": 1.02, "learning_rate": 2.5466561225772944e-06, "loss": 0.2327, "step": 5391 }, { "epoch": 1.02, "learning_rate": 2.545891354745641e-06, "loss": 0.258, "step": 5392 }, { "epoch": 1.02, "learning_rate": 2.5451265826180398e-06, "loss": 0.3012, "step": 5393 }, { "epoch": 1.02, "learning_rate": 2.5443618062660834e-06, "loss": 0.3374, "step": 5394 }, { "epoch": 1.02, "learning_rate": 2.543597025761364e-06, "loss": 0.2627, "step": 5395 }, { "epoch": 1.02, "learning_rate": 2.5428322411754713e-06, "loss": 0.3163, "step": 5396 }, { "epoch": 1.02, "learning_rate": 2.5420674525799997e-06, "loss": 0.1946, "step": 5397 }, { "epoch": 1.02, "learning_rate": 2.541302660046542e-06, "loss": 0.2547, "step": 5398 }, { "epoch": 1.02, "learning_rate": 2.5405378636466905e-06, "loss": 0.1595, "step": 5399 }, { "epoch": 1.02, "learning_rate": 2.539773063452039e-06, "loss": 0.2303, "step": 5400 }, { "epoch": 1.02, "learning_rate": 2.539008259534182e-06, "loss": 0.2861, "step": 5401 }, { "epoch": 1.02, "learning_rate": 2.5382434519647124e-06, "loss": 0.1953, "step": 5402 }, { "epoch": 1.02, "learning_rate": 2.5374786408152263e-06, "loss": 0.2652, "step": 5403 }, { "epoch": 1.02, "learning_rate": 2.5367138261573175e-06, "loss": 0.221, "step": 5404 }, { "epoch": 1.02, "learning_rate": 2.5359490080625822e-06, "loss": 0.2243, "step": 5405 }, { "epoch": 1.02, "learning_rate": 2.5351841866026156e-06, "loss": 0.2247, "step": 5406 }, { "epoch": 1.02, "learning_rate": 2.534419361849012e-06, "loss": 0.2894, "step": 5407 }, { "epoch": 1.02, "learning_rate": 2.5336545338733703e-06, "loss": 0.2943, "step": 5408 }, { "epoch": 1.02, "learning_rate": 2.5328897027472864e-06, "loss": 0.3256, "step": 5409 }, { "epoch": 1.02, "learning_rate": 2.5321248685423555e-06, "loss": 0.2187, "step": 5410 }, { "epoch": 1.02, "learning_rate": 2.5313600313301755e-06, "loss": 0.257, "step": 5411 }, { "epoch": 1.02, "learning_rate": 2.5305951911823446e-06, "loss": 0.2161, "step": 5412 }, { "epoch": 1.02, "learning_rate": 2.5298303481704593e-06, "loss": 0.2425, "step": 5413 }, { "epoch": 1.02, "learning_rate": 2.5290655023661186e-06, "loss": 0.4503, "step": 5414 }, { "epoch": 1.02, "learning_rate": 2.528300653840919e-06, "loss": 0.238, "step": 5415 }, { "epoch": 1.02, "learning_rate": 2.5275358026664608e-06, "loss": 0.2669, "step": 5416 }, { "epoch": 1.02, "learning_rate": 2.5267709489143417e-06, "loss": 0.3081, "step": 5417 }, { "epoch": 1.02, "learning_rate": 2.526006092656161e-06, "loss": 0.2964, "step": 5418 }, { "epoch": 1.02, "learning_rate": 2.525241233963517e-06, "loss": 0.3439, "step": 5419 }, { "epoch": 1.02, "learning_rate": 2.52447637290801e-06, "loss": 0.2593, "step": 5420 }, { "epoch": 1.02, "learning_rate": 2.523711509561239e-06, "loss": 0.2895, "step": 5421 }, { "epoch": 1.02, "learning_rate": 2.5229466439948037e-06, "loss": 0.286, "step": 5422 }, { "epoch": 1.02, "learning_rate": 2.5221817762803058e-06, "loss": 0.2273, "step": 5423 }, { "epoch": 1.02, "learning_rate": 2.521416906489343e-06, "loss": 0.3298, "step": 5424 }, { "epoch": 1.02, "learning_rate": 2.5206520346935173e-06, "loss": 0.2396, "step": 5425 }, { "epoch": 1.03, "learning_rate": 2.5198871609644294e-06, "loss": 0.2738, "step": 5426 }, { "epoch": 1.03, "learning_rate": 2.5191222853736783e-06, "loss": 0.2805, "step": 5427 }, { "epoch": 1.03, "learning_rate": 2.5183574079928667e-06, "loss": 0.3137, "step": 5428 }, { "epoch": 1.03, "learning_rate": 2.517592528893596e-06, "loss": 0.2752, "step": 5429 }, { "epoch": 1.03, "learning_rate": 2.516827648147467e-06, "loss": 0.2646, "step": 5430 }, { "epoch": 1.03, "learning_rate": 2.5160627658260793e-06, "loss": 0.2681, "step": 5431 }, { "epoch": 1.03, "learning_rate": 2.5152978820010376e-06, "loss": 0.2786, "step": 5432 }, { "epoch": 1.03, "learning_rate": 2.5145329967439415e-06, "loss": 0.2638, "step": 5433 }, { "epoch": 1.03, "learning_rate": 2.513768110126394e-06, "loss": 0.2466, "step": 5434 }, { "epoch": 1.03, "learning_rate": 2.5130032222199956e-06, "loss": 0.291, "step": 5435 }, { "epoch": 1.03, "learning_rate": 2.5122383330963503e-06, "loss": 0.2671, "step": 5436 }, { "epoch": 1.03, "learning_rate": 2.5114734428270587e-06, "loss": 0.2109, "step": 5437 }, { "epoch": 1.03, "learning_rate": 2.510708551483724e-06, "loss": 0.2991, "step": 5438 }, { "epoch": 1.03, "learning_rate": 2.509943659137948e-06, "loss": 0.2669, "step": 5439 }, { "epoch": 1.03, "learning_rate": 2.5091787658613355e-06, "loss": 0.2603, "step": 5440 }, { "epoch": 1.03, "learning_rate": 2.5084138717254863e-06, "loss": 0.2219, "step": 5441 }, { "epoch": 1.03, "learning_rate": 2.507648976802003e-06, "loss": 0.3068, "step": 5442 }, { "epoch": 1.03, "learning_rate": 2.5068840811624904e-06, "loss": 0.2838, "step": 5443 }, { "epoch": 1.03, "learning_rate": 2.506119184878551e-06, "loss": 0.3298, "step": 5444 }, { "epoch": 1.03, "learning_rate": 2.5053542880217856e-06, "loss": 0.2811, "step": 5445 }, { "epoch": 1.03, "learning_rate": 2.5045893906637998e-06, "loss": 0.2981, "step": 5446 }, { "epoch": 1.03, "learning_rate": 2.503824492876196e-06, "loss": 0.3331, "step": 5447 }, { "epoch": 1.03, "learning_rate": 2.503059594730575e-06, "loss": 0.1719, "step": 5448 }, { "epoch": 1.03, "learning_rate": 2.5022946962985423e-06, "loss": 0.2277, "step": 5449 }, { "epoch": 1.03, "learning_rate": 2.5015297976517e-06, "loss": 0.2389, "step": 5450 }, { "epoch": 1.03, "learning_rate": 2.5007648988616517e-06, "loss": 0.2339, "step": 5451 }, { "epoch": 1.03, "learning_rate": 2.5e-06, "loss": 0.2943, "step": 5452 }, { "epoch": 1.03, "learning_rate": 2.4992351011383483e-06, "loss": 0.3364, "step": 5453 }, { "epoch": 1.03, "learning_rate": 2.498470202348301e-06, "loss": 0.2952, "step": 5454 }, { "epoch": 1.03, "learning_rate": 2.4977053037014586e-06, "loss": 0.2383, "step": 5455 }, { "epoch": 1.03, "learning_rate": 2.496940405269425e-06, "loss": 0.2792, "step": 5456 }, { "epoch": 1.03, "learning_rate": 2.4961755071238054e-06, "loss": 0.2611, "step": 5457 }, { "epoch": 1.03, "learning_rate": 2.495410609336201e-06, "loss": 0.2263, "step": 5458 }, { "epoch": 1.03, "learning_rate": 2.4946457119782143e-06, "loss": 0.2553, "step": 5459 }, { "epoch": 1.03, "learning_rate": 2.49388081512145e-06, "loss": 0.2772, "step": 5460 }, { "epoch": 1.03, "learning_rate": 2.49311591883751e-06, "loss": 0.2856, "step": 5461 }, { "epoch": 1.03, "learning_rate": 2.492351023197997e-06, "loss": 0.2347, "step": 5462 }, { "epoch": 1.03, "learning_rate": 2.491586128274515e-06, "loss": 0.1882, "step": 5463 }, { "epoch": 1.03, "learning_rate": 2.4908212341386653e-06, "loss": 0.1365, "step": 5464 }, { "epoch": 1.03, "learning_rate": 2.4900563408620514e-06, "loss": 0.2104, "step": 5465 }, { "epoch": 1.03, "learning_rate": 2.4892914485162767e-06, "loss": 0.2094, "step": 5466 }, { "epoch": 1.03, "learning_rate": 2.4885265571729417e-06, "loss": 0.1989, "step": 5467 }, { "epoch": 1.03, "learning_rate": 2.4877616669036505e-06, "loss": 0.2503, "step": 5468 }, { "epoch": 1.03, "learning_rate": 2.4869967777800057e-06, "loss": 0.267, "step": 5469 }, { "epoch": 1.03, "learning_rate": 2.486231889873607e-06, "loss": 0.2409, "step": 5470 }, { "epoch": 1.03, "learning_rate": 2.485467003256059e-06, "loss": 0.2061, "step": 5471 }, { "epoch": 1.03, "learning_rate": 2.484702117998964e-06, "loss": 0.2467, "step": 5472 }, { "epoch": 1.03, "learning_rate": 2.483937234173921e-06, "loss": 0.2139, "step": 5473 }, { "epoch": 1.03, "learning_rate": 2.483172351852534e-06, "loss": 0.1902, "step": 5474 }, { "epoch": 1.03, "learning_rate": 2.4824074711064045e-06, "loss": 0.2975, "step": 5475 }, { "epoch": 1.03, "learning_rate": 2.4816425920071337e-06, "loss": 0.2745, "step": 5476 }, { "epoch": 1.03, "learning_rate": 2.480877714626322e-06, "loss": 0.2259, "step": 5477 }, { "epoch": 1.03, "learning_rate": 2.4801128390355723e-06, "loss": 0.2695, "step": 5478 }, { "epoch": 1.04, "learning_rate": 2.479347965306484e-06, "loss": 0.359, "step": 5479 }, { "epoch": 1.04, "learning_rate": 2.4785830935106574e-06, "loss": 0.2851, "step": 5480 }, { "epoch": 1.04, "learning_rate": 2.477818223719696e-06, "loss": 0.2706, "step": 5481 }, { "epoch": 1.04, "learning_rate": 2.477053356005197e-06, "loss": 0.2596, "step": 5482 }, { "epoch": 1.04, "learning_rate": 2.4762884904387617e-06, "loss": 0.2029, "step": 5483 }, { "epoch": 1.04, "learning_rate": 2.4755236270919904e-06, "loss": 0.1879, "step": 5484 }, { "epoch": 1.04, "learning_rate": 2.4747587660364835e-06, "loss": 0.2114, "step": 5485 }, { "epoch": 1.04, "learning_rate": 2.47399390734384e-06, "loss": 0.2466, "step": 5486 }, { "epoch": 1.04, "learning_rate": 2.4732290510856582e-06, "loss": 0.2309, "step": 5487 }, { "epoch": 1.04, "learning_rate": 2.4724641973335396e-06, "loss": 0.2035, "step": 5488 }, { "epoch": 1.04, "learning_rate": 2.471699346159081e-06, "loss": 0.2303, "step": 5489 }, { "epoch": 1.04, "learning_rate": 2.4709344976338818e-06, "loss": 0.2609, "step": 5490 }, { "epoch": 1.04, "learning_rate": 2.470169651829541e-06, "loss": 0.2187, "step": 5491 }, { "epoch": 1.04, "learning_rate": 2.4694048088176563e-06, "loss": 0.2065, "step": 5492 }, { "epoch": 1.04, "learning_rate": 2.4686399686698244e-06, "loss": 0.2303, "step": 5493 }, { "epoch": 1.04, "learning_rate": 2.4678751314576454e-06, "loss": 0.2829, "step": 5494 }, { "epoch": 1.04, "learning_rate": 2.467110297252714e-06, "loss": 0.2954, "step": 5495 }, { "epoch": 1.04, "learning_rate": 2.4663454661266297e-06, "loss": 0.2311, "step": 5496 }, { "epoch": 1.04, "learning_rate": 2.4655806381509883e-06, "loss": 0.3025, "step": 5497 }, { "epoch": 1.04, "learning_rate": 2.4648158133973853e-06, "loss": 0.2523, "step": 5498 }, { "epoch": 1.04, "learning_rate": 2.464050991937418e-06, "loss": 0.248, "step": 5499 }, { "epoch": 1.04, "learning_rate": 2.4632861738426838e-06, "loss": 0.255, "step": 5500 }, { "epoch": 1.04, "learning_rate": 2.4625213591847745e-06, "loss": 0.2877, "step": 5501 }, { "epoch": 1.04, "learning_rate": 2.461756548035288e-06, "loss": 0.2914, "step": 5502 }, { "epoch": 1.04, "learning_rate": 2.4609917404658198e-06, "loss": 0.269, "step": 5503 }, { "epoch": 1.04, "learning_rate": 2.4602269365479618e-06, "loss": 0.2116, "step": 5504 }, { "epoch": 1.04, "learning_rate": 2.4594621363533095e-06, "loss": 0.2052, "step": 5505 }, { "epoch": 1.04, "learning_rate": 2.4586973399534585e-06, "loss": 0.2664, "step": 5506 }, { "epoch": 1.04, "learning_rate": 2.4579325474200007e-06, "loss": 0.2829, "step": 5507 }, { "epoch": 1.04, "learning_rate": 2.4571677588245287e-06, "loss": 0.3456, "step": 5508 }, { "epoch": 1.04, "learning_rate": 2.456402974238637e-06, "loss": 0.3385, "step": 5509 }, { "epoch": 1.04, "learning_rate": 2.455638193733917e-06, "loss": 0.2791, "step": 5510 }, { "epoch": 1.04, "learning_rate": 2.45487341738196e-06, "loss": 0.2255, "step": 5511 }, { "epoch": 1.04, "learning_rate": 2.45410864525436e-06, "loss": 0.3244, "step": 5512 }, { "epoch": 1.04, "learning_rate": 2.4533438774227064e-06, "loss": 0.2545, "step": 5513 }, { "epoch": 1.04, "learning_rate": 2.4525791139585897e-06, "loss": 0.2614, "step": 5514 }, { "epoch": 1.04, "learning_rate": 2.4518143549336024e-06, "loss": 0.1873, "step": 5515 }, { "epoch": 1.04, "learning_rate": 2.451049600419332e-06, "loss": 0.2839, "step": 5516 }, { "epoch": 1.04, "learning_rate": 2.4502848504873704e-06, "loss": 0.3342, "step": 5517 }, { "epoch": 1.04, "learning_rate": 2.449520105209306e-06, "loss": 0.183, "step": 5518 }, { "epoch": 1.04, "learning_rate": 2.4487553646567267e-06, "loss": 0.2031, "step": 5519 }, { "epoch": 1.04, "learning_rate": 2.447990628901222e-06, "loss": 0.3258, "step": 5520 }, { "epoch": 1.04, "learning_rate": 2.4472258980143794e-06, "loss": 0.2874, "step": 5521 }, { "epoch": 1.04, "learning_rate": 2.4464611720677855e-06, "loss": 0.2773, "step": 5522 }, { "epoch": 1.04, "learning_rate": 2.4456964511330287e-06, "loss": 0.2028, "step": 5523 }, { "epoch": 1.04, "learning_rate": 2.444931735281695e-06, "loss": 0.2117, "step": 5524 }, { "epoch": 1.04, "learning_rate": 2.444167024585369e-06, "loss": 0.3023, "step": 5525 }, { "epoch": 1.04, "learning_rate": 2.443402319115637e-06, "loss": 0.2985, "step": 5526 }, { "epoch": 1.04, "learning_rate": 2.442637618944086e-06, "loss": 0.2749, "step": 5527 }, { "epoch": 1.04, "learning_rate": 2.4418729241422985e-06, "loss": 0.1979, "step": 5528 }, { "epoch": 1.04, "learning_rate": 2.441108234781858e-06, "loss": 0.1814, "step": 5529 }, { "epoch": 1.04, "learning_rate": 2.44034355093435e-06, "loss": 0.2223, "step": 5530 }, { "epoch": 1.04, "learning_rate": 2.439578872671356e-06, "loss": 0.249, "step": 5531 }, { "epoch": 1.05, "learning_rate": 2.438814200064459e-06, "loss": 0.222, "step": 5532 }, { "epoch": 1.05, "learning_rate": 2.4380495331852412e-06, "loss": 0.249, "step": 5533 }, { "epoch": 1.05, "learning_rate": 2.4372848721052834e-06, "loss": 0.2989, "step": 5534 }, { "epoch": 1.05, "learning_rate": 2.4365202168961667e-06, "loss": 0.2415, "step": 5535 }, { "epoch": 1.05, "learning_rate": 2.4357555676294724e-06, "loss": 0.299, "step": 5536 }, { "epoch": 1.05, "learning_rate": 2.434990924376778e-06, "loss": 0.3112, "step": 5537 }, { "epoch": 1.05, "learning_rate": 2.4342262872096653e-06, "loss": 0.2304, "step": 5538 }, { "epoch": 1.05, "learning_rate": 2.4334616561997116e-06, "loss": 0.2238, "step": 5539 }, { "epoch": 1.05, "learning_rate": 2.4326970314184944e-06, "loss": 0.2964, "step": 5540 }, { "epoch": 1.05, "learning_rate": 2.4319324129375924e-06, "loss": 0.2815, "step": 5541 }, { "epoch": 1.05, "learning_rate": 2.4311678008285827e-06, "loss": 0.2142, "step": 5542 }, { "epoch": 1.05, "learning_rate": 2.4304031951630398e-06, "loss": 0.1979, "step": 5543 }, { "epoch": 1.05, "learning_rate": 2.4296385960125397e-06, "loss": 0.2328, "step": 5544 }, { "epoch": 1.05, "learning_rate": 2.42887400344866e-06, "loss": 0.1947, "step": 5545 }, { "epoch": 1.05, "learning_rate": 2.428109417542973e-06, "loss": 0.2529, "step": 5546 }, { "epoch": 1.05, "learning_rate": 2.427344838367052e-06, "loss": 0.2902, "step": 5547 }, { "epoch": 1.05, "learning_rate": 2.426580265992472e-06, "loss": 0.1694, "step": 5548 }, { "epoch": 1.05, "learning_rate": 2.425815700490805e-06, "loss": 0.2807, "step": 5549 }, { "epoch": 1.05, "learning_rate": 2.4250511419336213e-06, "loss": 0.2512, "step": 5550 }, { "epoch": 1.05, "learning_rate": 2.424286590392495e-06, "loss": 0.3389, "step": 5551 }, { "epoch": 1.05, "learning_rate": 2.423522045938995e-06, "loss": 0.2035, "step": 5552 }, { "epoch": 1.05, "learning_rate": 2.4227575086446907e-06, "loss": 0.253, "step": 5553 }, { "epoch": 1.05, "learning_rate": 2.421992978581152e-06, "loss": 0.2821, "step": 5554 }, { "epoch": 1.05, "learning_rate": 2.421228455819949e-06, "loss": 0.2712, "step": 5555 }, { "epoch": 1.05, "learning_rate": 2.4204639404326475e-06, "loss": 0.2151, "step": 5556 }, { "epoch": 1.05, "learning_rate": 2.4196994324908153e-06, "loss": 0.2645, "step": 5557 }, { "epoch": 1.05, "learning_rate": 2.41893493206602e-06, "loss": 0.1993, "step": 5558 }, { "epoch": 1.05, "learning_rate": 2.4181704392298267e-06, "loss": 0.1909, "step": 5559 }, { "epoch": 1.05, "learning_rate": 2.4174059540537994e-06, "loss": 0.1883, "step": 5560 }, { "epoch": 1.05, "learning_rate": 2.416641476609505e-06, "loss": 0.221, "step": 5561 }, { "epoch": 1.05, "learning_rate": 2.415877006968505e-06, "loss": 0.3092, "step": 5562 }, { "epoch": 1.05, "learning_rate": 2.4151125452023626e-06, "loss": 0.2609, "step": 5563 }, { "epoch": 1.05, "learning_rate": 2.4143480913826416e-06, "loss": 0.272, "step": 5564 }, { "epoch": 1.05, "learning_rate": 2.4135836455809015e-06, "loss": 0.3789, "step": 5565 }, { "epoch": 1.05, "learning_rate": 2.412819207868704e-06, "loss": 0.3832, "step": 5566 }, { "epoch": 1.05, "learning_rate": 2.41205477831761e-06, "loss": 0.1993, "step": 5567 }, { "epoch": 1.05, "learning_rate": 2.411290356999177e-06, "loss": 0.2247, "step": 5568 }, { "epoch": 1.05, "learning_rate": 2.410525943984964e-06, "loss": 0.2547, "step": 5569 }, { "epoch": 1.05, "learning_rate": 2.4097615393465294e-06, "loss": 0.2677, "step": 5570 }, { "epoch": 1.05, "learning_rate": 2.408997143155429e-06, "loss": 0.2185, "step": 5571 }, { "epoch": 1.05, "learning_rate": 2.408232755483219e-06, "loss": 0.2424, "step": 5572 }, { "epoch": 1.05, "learning_rate": 2.4074683764014562e-06, "loss": 0.2793, "step": 5573 }, { "epoch": 1.05, "learning_rate": 2.4067040059816928e-06, "loss": 0.2228, "step": 5574 }, { "epoch": 1.05, "learning_rate": 2.405939644295483e-06, "loss": 0.216, "step": 5575 }, { "epoch": 1.05, "learning_rate": 2.4051752914143818e-06, "loss": 0.2632, "step": 5576 }, { "epoch": 1.05, "learning_rate": 2.404410947409939e-06, "loss": 0.3075, "step": 5577 }, { "epoch": 1.05, "learning_rate": 2.403646612353706e-06, "loss": 0.244, "step": 5578 }, { "epoch": 1.05, "learning_rate": 2.402882286317235e-06, "loss": 0.2872, "step": 5579 }, { "epoch": 1.05, "learning_rate": 2.4021179693720725e-06, "loss": 0.3141, "step": 5580 }, { "epoch": 1.05, "learning_rate": 2.401353661589769e-06, "loss": 0.2537, "step": 5581 }, { "epoch": 1.05, "learning_rate": 2.400589363041873e-06, "loss": 0.2886, "step": 5582 }, { "epoch": 1.05, "learning_rate": 2.39982507379993e-06, "loss": 0.3069, "step": 5583 }, { "epoch": 1.05, "learning_rate": 2.399060793935486e-06, "loss": 0.2004, "step": 5584 }, { "epoch": 1.06, "learning_rate": 2.3982965235200877e-06, "loss": 0.2548, "step": 5585 }, { "epoch": 1.06, "learning_rate": 2.3975322626252774e-06, "loss": 0.3304, "step": 5586 }, { "epoch": 1.06, "learning_rate": 2.3967680113226003e-06, "loss": 0.223, "step": 5587 }, { "epoch": 1.06, "learning_rate": 2.396003769683599e-06, "loss": 0.2259, "step": 5588 }, { "epoch": 1.06, "learning_rate": 2.395239537779813e-06, "loss": 0.2347, "step": 5589 }, { "epoch": 1.06, "learning_rate": 2.394475315682785e-06, "loss": 0.2328, "step": 5590 }, { "epoch": 1.06, "learning_rate": 2.393711103464054e-06, "loss": 0.2403, "step": 5591 }, { "epoch": 1.06, "learning_rate": 2.3929469011951583e-06, "loss": 0.2936, "step": 5592 }, { "epoch": 1.06, "learning_rate": 2.3921827089476373e-06, "loss": 0.2618, "step": 5593 }, { "epoch": 1.06, "learning_rate": 2.391418526793027e-06, "loss": 0.2448, "step": 5594 }, { "epoch": 1.06, "learning_rate": 2.3906543548028628e-06, "loss": 0.2748, "step": 5595 }, { "epoch": 1.06, "learning_rate": 2.3898901930486808e-06, "loss": 0.2522, "step": 5596 }, { "epoch": 1.06, "learning_rate": 2.3891260416020164e-06, "loss": 0.2374, "step": 5597 }, { "epoch": 1.06, "learning_rate": 2.3883619005344e-06, "loss": 0.2874, "step": 5598 }, { "epoch": 1.06, "learning_rate": 2.387597769917365e-06, "loss": 0.2297, "step": 5599 }, { "epoch": 1.06, "learning_rate": 2.3868336498224437e-06, "loss": 0.2852, "step": 5600 }, { "epoch": 1.06, "learning_rate": 2.3860695403211648e-06, "loss": 0.2433, "step": 5601 }, { "epoch": 1.06, "learning_rate": 2.3853054414850572e-06, "loss": 0.2371, "step": 5602 }, { "epoch": 1.06, "learning_rate": 2.3845413533856517e-06, "loss": 0.3286, "step": 5603 }, { "epoch": 1.06, "learning_rate": 2.383777276094473e-06, "loss": 0.3544, "step": 5604 }, { "epoch": 1.06, "learning_rate": 2.3830132096830476e-06, "loss": 0.2643, "step": 5605 }, { "epoch": 1.06, "learning_rate": 2.3822491542229027e-06, "loss": 0.277, "step": 5606 }, { "epoch": 1.06, "learning_rate": 2.38148510978556e-06, "loss": 0.2935, "step": 5607 }, { "epoch": 1.06, "learning_rate": 2.380721076442544e-06, "loss": 0.2268, "step": 5608 }, { "epoch": 1.06, "learning_rate": 2.379957054265376e-06, "loss": 0.2076, "step": 5609 }, { "epoch": 1.06, "learning_rate": 2.379193043325579e-06, "loss": 0.2507, "step": 5610 }, { "epoch": 1.06, "learning_rate": 2.378429043694671e-06, "loss": 0.2503, "step": 5611 }, { "epoch": 1.06, "learning_rate": 2.377665055444171e-06, "loss": 0.171, "step": 5612 }, { "epoch": 1.06, "learning_rate": 2.3769010786455983e-06, "loss": 0.2889, "step": 5613 }, { "epoch": 1.06, "learning_rate": 2.3761371133704688e-06, "loss": 0.2595, "step": 5614 }, { "epoch": 1.06, "learning_rate": 2.3753731596902974e-06, "loss": 0.3079, "step": 5615 }, { "epoch": 1.06, "learning_rate": 2.3746092176766014e-06, "loss": 0.2795, "step": 5616 }, { "epoch": 1.06, "learning_rate": 2.373845287400891e-06, "loss": 0.2538, "step": 5617 }, { "epoch": 1.06, "learning_rate": 2.3730813689346813e-06, "loss": 0.2654, "step": 5618 }, { "epoch": 1.06, "learning_rate": 2.372317462349483e-06, "loss": 0.279, "step": 5619 }, { "epoch": 1.06, "learning_rate": 2.3715535677168047e-06, "loss": 0.2493, "step": 5620 }, { "epoch": 1.06, "learning_rate": 2.3707896851081576e-06, "loss": 0.2374, "step": 5621 }, { "epoch": 1.06, "learning_rate": 2.3700258145950495e-06, "loss": 0.2693, "step": 5622 }, { "epoch": 1.06, "learning_rate": 2.369261956248986e-06, "loss": 0.1631, "step": 5623 }, { "epoch": 1.06, "learning_rate": 2.368498110141474e-06, "loss": 0.2939, "step": 5624 }, { "epoch": 1.06, "learning_rate": 2.367734276344018e-06, "loss": 0.2234, "step": 5625 }, { "epoch": 1.06, "learning_rate": 2.3669704549281195e-06, "loss": 0.2148, "step": 5626 }, { "epoch": 1.06, "learning_rate": 2.366206645965283e-06, "loss": 0.2114, "step": 5627 }, { "epoch": 1.06, "learning_rate": 2.36544284952701e-06, "loss": 0.307, "step": 5628 }, { "epoch": 1.06, "learning_rate": 2.3646790656847988e-06, "loss": 0.245, "step": 5629 }, { "epoch": 1.06, "learning_rate": 2.363915294510148e-06, "loss": 0.1874, "step": 5630 }, { "epoch": 1.06, "learning_rate": 2.3631515360745576e-06, "loss": 0.271, "step": 5631 }, { "epoch": 1.06, "learning_rate": 2.3623877904495212e-06, "loss": 0.21, "step": 5632 }, { "epoch": 1.06, "learning_rate": 2.3616240577065346e-06, "loss": 0.3147, "step": 5633 }, { "epoch": 1.06, "learning_rate": 2.3608603379170937e-06, "loss": 0.2717, "step": 5634 }, { "epoch": 1.06, "learning_rate": 2.360096631152689e-06, "loss": 0.1674, "step": 5635 }, { "epoch": 1.06, "learning_rate": 2.3593329374848124e-06, "loss": 0.2802, "step": 5636 }, { "epoch": 1.06, "learning_rate": 2.3585692569849554e-06, "loss": 0.242, "step": 5637 }, { "epoch": 1.07, "learning_rate": 2.3578055897246057e-06, "loss": 0.3051, "step": 5638 }, { "epoch": 1.07, "learning_rate": 2.3570419357752522e-06, "loss": 0.2414, "step": 5639 }, { "epoch": 1.07, "learning_rate": 2.356278295208382e-06, "loss": 0.2864, "step": 5640 }, { "epoch": 1.07, "learning_rate": 2.355514668095478e-06, "loss": 0.2595, "step": 5641 }, { "epoch": 1.07, "learning_rate": 2.3547510545080262e-06, "loss": 0.2241, "step": 5642 }, { "epoch": 1.07, "learning_rate": 2.3539874545175103e-06, "loss": 0.3726, "step": 5643 }, { "epoch": 1.07, "learning_rate": 2.3532238681954088e-06, "loss": 0.3046, "step": 5644 }, { "epoch": 1.07, "learning_rate": 2.352460295613205e-06, "loss": 0.2499, "step": 5645 }, { "epoch": 1.07, "learning_rate": 2.3516967368423767e-06, "loss": 0.2749, "step": 5646 }, { "epoch": 1.07, "learning_rate": 2.3509331919544e-06, "loss": 0.2482, "step": 5647 }, { "epoch": 1.07, "learning_rate": 2.3501696610207535e-06, "loss": 0.2278, "step": 5648 }, { "epoch": 1.07, "learning_rate": 2.3494061441129126e-06, "loss": 0.3023, "step": 5649 }, { "epoch": 1.07, "learning_rate": 2.3486426413023495e-06, "loss": 0.2728, "step": 5650 }, { "epoch": 1.07, "learning_rate": 2.3478791526605364e-06, "loss": 0.2271, "step": 5651 }, { "epoch": 1.07, "learning_rate": 2.3471156782589467e-06, "loss": 0.2695, "step": 5652 }, { "epoch": 1.07, "learning_rate": 2.346352218169048e-06, "loss": 0.2487, "step": 5653 }, { "epoch": 1.07, "learning_rate": 2.345588772462309e-06, "loss": 0.2919, "step": 5654 }, { "epoch": 1.07, "learning_rate": 2.3448253412101983e-06, "loss": 0.3564, "step": 5655 }, { "epoch": 1.07, "learning_rate": 2.3440619244841798e-06, "loss": 0.262, "step": 5656 }, { "epoch": 1.07, "learning_rate": 2.3432985223557183e-06, "loss": 0.2576, "step": 5657 }, { "epoch": 1.07, "learning_rate": 2.3425351348962784e-06, "loss": 0.2614, "step": 5658 }, { "epoch": 1.07, "learning_rate": 2.3417717621773196e-06, "loss": 0.2612, "step": 5659 }, { "epoch": 1.07, "learning_rate": 2.341008404270304e-06, "loss": 0.2967, "step": 5660 }, { "epoch": 1.07, "learning_rate": 2.34024506124669e-06, "loss": 0.2929, "step": 5661 }, { "epoch": 1.07, "learning_rate": 2.339481733177933e-06, "loss": 0.2483, "step": 5662 }, { "epoch": 1.07, "learning_rate": 2.338718420135492e-06, "loss": 0.2974, "step": 5663 }, { "epoch": 1.07, "learning_rate": 2.3379551221908212e-06, "loss": 0.287, "step": 5664 }, { "epoch": 1.07, "learning_rate": 2.337191839415372e-06, "loss": 0.2813, "step": 5665 }, { "epoch": 1.07, "learning_rate": 2.3364285718805977e-06, "loss": 0.2321, "step": 5666 }, { "epoch": 1.07, "learning_rate": 2.335665319657949e-06, "loss": 0.2018, "step": 5667 }, { "epoch": 1.07, "learning_rate": 2.3349020828188733e-06, "loss": 0.1773, "step": 5668 }, { "epoch": 1.07, "learning_rate": 2.3341388614348193e-06, "loss": 0.2692, "step": 5669 }, { "epoch": 1.07, "learning_rate": 2.333375655577234e-06, "loss": 0.25, "step": 5670 }, { "epoch": 1.07, "learning_rate": 2.332612465317561e-06, "loss": 0.2908, "step": 5671 }, { "epoch": 1.07, "learning_rate": 2.3318492907272423e-06, "loss": 0.2431, "step": 5672 }, { "epoch": 1.07, "learning_rate": 2.3310861318777216e-06, "loss": 0.2512, "step": 5673 }, { "epoch": 1.07, "learning_rate": 2.330322988840439e-06, "loss": 0.2201, "step": 5674 }, { "epoch": 1.07, "learning_rate": 2.3295598616868316e-06, "loss": 0.2296, "step": 5675 }, { "epoch": 1.07, "learning_rate": 2.3287967504883385e-06, "loss": 0.229, "step": 5676 }, { "epoch": 1.07, "learning_rate": 2.3280336553163954e-06, "loss": 0.2438, "step": 5677 }, { "epoch": 1.07, "learning_rate": 2.3272705762424345e-06, "loss": 0.3464, "step": 5678 }, { "epoch": 1.07, "learning_rate": 2.3265075133378905e-06, "loss": 0.3222, "step": 5679 }, { "epoch": 1.07, "learning_rate": 2.3257444666741948e-06, "loss": 0.2396, "step": 5680 }, { "epoch": 1.07, "learning_rate": 2.3249814363227765e-06, "loss": 0.2888, "step": 5681 }, { "epoch": 1.07, "learning_rate": 2.3242184223550625e-06, "loss": 0.2826, "step": 5682 }, { "epoch": 1.07, "learning_rate": 2.3234554248424826e-06, "loss": 0.2099, "step": 5683 }, { "epoch": 1.07, "learning_rate": 2.3226924438564596e-06, "loss": 0.2894, "step": 5684 }, { "epoch": 1.07, "learning_rate": 2.3219294794684173e-06, "loss": 0.2152, "step": 5685 }, { "epoch": 1.07, "learning_rate": 2.321166531749779e-06, "loss": 0.2645, "step": 5686 }, { "epoch": 1.07, "learning_rate": 2.3204036007719636e-06, "loss": 0.2529, "step": 5687 }, { "epoch": 1.07, "learning_rate": 2.3196406866063902e-06, "loss": 0.2199, "step": 5688 }, { "epoch": 1.07, "learning_rate": 2.318877789324478e-06, "loss": 0.1778, "step": 5689 }, { "epoch": 1.08, "learning_rate": 2.3181149089976407e-06, "loss": 0.2466, "step": 5690 }, { "epoch": 1.08, "learning_rate": 2.3173520456972932e-06, "loss": 0.2934, "step": 5691 }, { "epoch": 1.08, "learning_rate": 2.3165891994948492e-06, "loss": 0.2519, "step": 5692 }, { "epoch": 1.08, "learning_rate": 2.3158263704617173e-06, "loss": 0.2187, "step": 5693 }, { "epoch": 1.08, "learning_rate": 2.3150635586693087e-06, "loss": 0.2637, "step": 5694 }, { "epoch": 1.08, "learning_rate": 2.3143007641890313e-06, "loss": 0.2649, "step": 5695 }, { "epoch": 1.08, "learning_rate": 2.3135379870922895e-06, "loss": 0.1981, "step": 5696 }, { "epoch": 1.08, "learning_rate": 2.3127752274504887e-06, "loss": 0.2781, "step": 5697 }, { "epoch": 1.08, "learning_rate": 2.3120124853350332e-06, "loss": 0.2154, "step": 5698 }, { "epoch": 1.08, "learning_rate": 2.3112497608173225e-06, "loss": 0.2362, "step": 5699 }, { "epoch": 1.08, "learning_rate": 2.310487053968756e-06, "loss": 0.2527, "step": 5700 }, { "epoch": 1.08, "learning_rate": 2.3097243648607337e-06, "loss": 0.2235, "step": 5701 }, { "epoch": 1.08, "learning_rate": 2.3089616935646496e-06, "loss": 0.2758, "step": 5702 }, { "epoch": 1.08, "learning_rate": 2.3081990401518994e-06, "loss": 0.276, "step": 5703 }, { "epoch": 1.08, "learning_rate": 2.3074364046938763e-06, "loss": 0.2644, "step": 5704 }, { "epoch": 1.08, "learning_rate": 2.3066737872619705e-06, "loss": 0.19, "step": 5705 }, { "epoch": 1.08, "learning_rate": 2.305911187927572e-06, "loss": 0.2427, "step": 5706 }, { "epoch": 1.08, "learning_rate": 2.30514860676207e-06, "loss": 0.2945, "step": 5707 }, { "epoch": 1.08, "learning_rate": 2.3043860438368484e-06, "loss": 0.2528, "step": 5708 }, { "epoch": 1.08, "learning_rate": 2.303623499223293e-06, "loss": 0.2271, "step": 5709 }, { "epoch": 1.08, "learning_rate": 2.3028609729927877e-06, "loss": 0.3056, "step": 5710 }, { "epoch": 1.08, "learning_rate": 2.3020984652167106e-06, "loss": 0.2862, "step": 5711 }, { "epoch": 1.08, "learning_rate": 2.3013359759664435e-06, "loss": 0.2946, "step": 5712 }, { "epoch": 1.08, "learning_rate": 2.3005735053133628e-06, "loss": 0.2219, "step": 5713 }, { "epoch": 1.08, "learning_rate": 2.299811053328844e-06, "loss": 0.2993, "step": 5714 }, { "epoch": 1.08, "learning_rate": 2.2990486200842625e-06, "loss": 0.2438, "step": 5715 }, { "epoch": 1.08, "learning_rate": 2.29828620565099e-06, "loss": 0.2186, "step": 5716 }, { "epoch": 1.08, "learning_rate": 2.297523810100396e-06, "loss": 0.2892, "step": 5717 }, { "epoch": 1.08, "learning_rate": 2.2967614335038503e-06, "loss": 0.3126, "step": 5718 }, { "epoch": 1.08, "learning_rate": 2.295999075932721e-06, "loss": 0.2611, "step": 5719 }, { "epoch": 1.08, "learning_rate": 2.2952367374583717e-06, "loss": 0.1934, "step": 5720 }, { "epoch": 1.08, "learning_rate": 2.2944744181521656e-06, "loss": 0.2313, "step": 5721 }, { "epoch": 1.08, "learning_rate": 2.2937121180854668e-06, "loss": 0.2249, "step": 5722 }, { "epoch": 1.08, "learning_rate": 2.2929498373296326e-06, "loss": 0.2028, "step": 5723 }, { "epoch": 1.08, "learning_rate": 2.2921875759560213e-06, "loss": 0.2037, "step": 5724 }, { "epoch": 1.08, "learning_rate": 2.291425334035991e-06, "loss": 0.311, "step": 5725 }, { "epoch": 1.08, "learning_rate": 2.290663111640894e-06, "loss": 0.3032, "step": 5726 }, { "epoch": 1.08, "learning_rate": 2.2899009088420835e-06, "loss": 0.2028, "step": 5727 }, { "epoch": 1.08, "learning_rate": 2.2891387257109114e-06, "loss": 0.3012, "step": 5728 }, { "epoch": 1.08, "learning_rate": 2.288376562318725e-06, "loss": 0.2942, "step": 5729 }, { "epoch": 1.08, "learning_rate": 2.2876144187368724e-06, "loss": 0.2835, "step": 5730 }, { "epoch": 1.08, "learning_rate": 2.2868522950366993e-06, "loss": 0.2594, "step": 5731 }, { "epoch": 1.08, "learning_rate": 2.286090191289547e-06, "loss": 0.2969, "step": 5732 }, { "epoch": 1.08, "learning_rate": 2.285328107566759e-06, "loss": 0.3373, "step": 5733 }, { "epoch": 1.08, "learning_rate": 2.2845660439396732e-06, "loss": 0.3114, "step": 5734 }, { "epoch": 1.08, "learning_rate": 2.2838040004796296e-06, "loss": 0.2957, "step": 5735 }, { "epoch": 1.08, "learning_rate": 2.283041977257962e-06, "loss": 0.3415, "step": 5736 }, { "epoch": 1.08, "learning_rate": 2.2822799743460046e-06, "loss": 0.2913, "step": 5737 }, { "epoch": 1.08, "learning_rate": 2.281517991815091e-06, "loss": 0.2905, "step": 5738 }, { "epoch": 1.08, "learning_rate": 2.2807560297365492e-06, "loss": 0.2711, "step": 5739 }, { "epoch": 1.08, "learning_rate": 2.2799940881817092e-06, "loss": 0.2251, "step": 5740 }, { "epoch": 1.08, "learning_rate": 2.279232167221897e-06, "loss": 0.2497, "step": 5741 }, { "epoch": 1.08, "learning_rate": 2.278470266928435e-06, "loss": 0.3226, "step": 5742 }, { "epoch": 1.09, "learning_rate": 2.277708387372648e-06, "loss": 0.2372, "step": 5743 }, { "epoch": 1.09, "learning_rate": 2.2769465286258565e-06, "loss": 0.2256, "step": 5744 }, { "epoch": 1.09, "learning_rate": 2.276184690759377e-06, "loss": 0.2529, "step": 5745 }, { "epoch": 1.09, "learning_rate": 2.2754228738445276e-06, "loss": 0.2839, "step": 5746 }, { "epoch": 1.09, "learning_rate": 2.2746610779526234e-06, "loss": 0.2818, "step": 5747 }, { "epoch": 1.09, "learning_rate": 2.273899303154975e-06, "loss": 0.2025, "step": 5748 }, { "epoch": 1.09, "learning_rate": 2.2731375495228942e-06, "loss": 0.2837, "step": 5749 }, { "epoch": 1.09, "learning_rate": 2.272375817127691e-06, "loss": 0.2276, "step": 5750 }, { "epoch": 1.09, "learning_rate": 2.2716141060406704e-06, "loss": 0.2419, "step": 5751 }, { "epoch": 1.09, "learning_rate": 2.2708524163331375e-06, "loss": 0.2782, "step": 5752 }, { "epoch": 1.09, "learning_rate": 2.270090748076396e-06, "loss": 0.3486, "step": 5753 }, { "epoch": 1.09, "learning_rate": 2.269329101341745e-06, "loss": 0.3218, "step": 5754 }, { "epoch": 1.09, "learning_rate": 2.2685674762004846e-06, "loss": 0.2384, "step": 5755 }, { "epoch": 1.09, "learning_rate": 2.267805872723911e-06, "loss": 0.2775, "step": 5756 }, { "epoch": 1.09, "learning_rate": 2.267044290983319e-06, "loss": 0.2246, "step": 5757 }, { "epoch": 1.09, "learning_rate": 2.2662827310499995e-06, "loss": 0.2165, "step": 5758 }, { "epoch": 1.09, "learning_rate": 2.2655211929952463e-06, "loss": 0.2233, "step": 5759 }, { "epoch": 1.09, "learning_rate": 2.2647596768903453e-06, "loss": 0.221, "step": 5760 }, { "epoch": 1.09, "learning_rate": 2.263998182806584e-06, "loss": 0.264, "step": 5761 }, { "epoch": 1.09, "learning_rate": 2.2632367108152474e-06, "loss": 0.3225, "step": 5762 }, { "epoch": 1.09, "learning_rate": 2.262475260987616e-06, "loss": 0.2376, "step": 5763 }, { "epoch": 1.09, "learning_rate": 2.261713833394971e-06, "loss": 0.3045, "step": 5764 }, { "epoch": 1.09, "learning_rate": 2.2609524281085924e-06, "loss": 0.2638, "step": 5765 }, { "epoch": 1.09, "learning_rate": 2.260191045199753e-06, "loss": 0.2534, "step": 5766 }, { "epoch": 1.09, "learning_rate": 2.259429684739729e-06, "loss": 0.2316, "step": 5767 }, { "epoch": 1.09, "learning_rate": 2.2586683467997925e-06, "loss": 0.2017, "step": 5768 }, { "epoch": 1.09, "learning_rate": 2.2579070314512113e-06, "loss": 0.2741, "step": 5769 }, { "epoch": 1.09, "learning_rate": 2.257145738765254e-06, "loss": 0.2644, "step": 5770 }, { "epoch": 1.09, "learning_rate": 2.256384468813188e-06, "loss": 0.1979, "step": 5771 }, { "epoch": 1.09, "learning_rate": 2.2556232216662747e-06, "loss": 0.3035, "step": 5772 }, { "epoch": 1.09, "learning_rate": 2.2548619973957756e-06, "loss": 0.2867, "step": 5773 }, { "epoch": 1.09, "learning_rate": 2.2541007960729512e-06, "loss": 0.2963, "step": 5774 }, { "epoch": 1.09, "learning_rate": 2.2533396177690565e-06, "loss": 0.2743, "step": 5775 }, { "epoch": 1.09, "learning_rate": 2.252578462555347e-06, "loss": 0.1926, "step": 5776 }, { "epoch": 1.09, "learning_rate": 2.2518173305030767e-06, "loss": 0.1896, "step": 5777 }, { "epoch": 1.09, "learning_rate": 2.251056221683495e-06, "loss": 0.2323, "step": 5778 }, { "epoch": 1.09, "learning_rate": 2.250295136167849e-06, "loss": 0.2603, "step": 5779 }, { "epoch": 1.09, "learning_rate": 2.249534074027388e-06, "loss": 0.3723, "step": 5780 }, { "epoch": 1.09, "learning_rate": 2.2487730353333524e-06, "loss": 0.2128, "step": 5781 }, { "epoch": 1.09, "learning_rate": 2.248012020156987e-06, "loss": 0.2501, "step": 5782 }, { "epoch": 1.09, "learning_rate": 2.2472510285695304e-06, "loss": 0.3099, "step": 5783 }, { "epoch": 1.09, "learning_rate": 2.2464900606422186e-06, "loss": 0.2641, "step": 5784 }, { "epoch": 1.09, "learning_rate": 2.2457291164462886e-06, "loss": 0.2791, "step": 5785 }, { "epoch": 1.09, "learning_rate": 2.2449681960529734e-06, "loss": 0.2083, "step": 5786 }, { "epoch": 1.09, "learning_rate": 2.244207299533501e-06, "loss": 0.281, "step": 5787 }, { "epoch": 1.09, "learning_rate": 2.2434464269591035e-06, "loss": 0.2803, "step": 5788 }, { "epoch": 1.09, "learning_rate": 2.242685578401005e-06, "loss": 0.2387, "step": 5789 }, { "epoch": 1.09, "learning_rate": 2.2419247539304293e-06, "loss": 0.2894, "step": 5790 }, { "epoch": 1.09, "learning_rate": 2.2411639536185986e-06, "loss": 0.2441, "step": 5791 }, { "epoch": 1.09, "learning_rate": 2.2404031775367336e-06, "loss": 0.2641, "step": 5792 }, { "epoch": 1.09, "learning_rate": 2.23964242575605e-06, "loss": 0.1915, "step": 5793 }, { "epoch": 1.09, "learning_rate": 2.2388816983477623e-06, "loss": 0.218, "step": 5794 }, { "epoch": 1.09, "learning_rate": 2.238120995383085e-06, "loss": 0.2414, "step": 5795 }, { "epoch": 1.1, "learning_rate": 2.237360316933227e-06, "loss": 0.2415, "step": 5796 }, { "epoch": 1.1, "learning_rate": 2.236599663069396e-06, "loss": 0.2325, "step": 5797 }, { "epoch": 1.1, "learning_rate": 2.2358390338627995e-06, "loss": 0.2025, "step": 5798 }, { "epoch": 1.1, "learning_rate": 2.23507842938464e-06, "loss": 0.2617, "step": 5799 }, { "epoch": 1.1, "learning_rate": 2.2343178497061178e-06, "loss": 0.246, "step": 5800 }, { "epoch": 1.1, "learning_rate": 2.2335572948984326e-06, "loss": 0.2637, "step": 5801 }, { "epoch": 1.1, "learning_rate": 2.2327967650327815e-06, "loss": 0.2605, "step": 5802 }, { "epoch": 1.1, "learning_rate": 2.232036260180358e-06, "loss": 0.2704, "step": 5803 }, { "epoch": 1.1, "learning_rate": 2.231275780412353e-06, "loss": 0.2223, "step": 5804 }, { "epoch": 1.1, "learning_rate": 2.230515325799958e-06, "loss": 0.2585, "step": 5805 }, { "epoch": 1.1, "learning_rate": 2.229754896414358e-06, "loss": 0.2713, "step": 5806 }, { "epoch": 1.1, "learning_rate": 2.228994492326738e-06, "loss": 0.199, "step": 5807 }, { "epoch": 1.1, "learning_rate": 2.228234113608283e-06, "loss": 0.2401, "step": 5808 }, { "epoch": 1.1, "learning_rate": 2.2274737603301696e-06, "loss": 0.2054, "step": 5809 }, { "epoch": 1.1, "learning_rate": 2.2267134325635766e-06, "loss": 0.245, "step": 5810 }, { "epoch": 1.1, "learning_rate": 2.2259531303796804e-06, "loss": 0.2197, "step": 5811 }, { "epoch": 1.1, "learning_rate": 2.225192853849652e-06, "loss": 0.2429, "step": 5812 }, { "epoch": 1.1, "learning_rate": 2.2244326030446632e-06, "loss": 0.3023, "step": 5813 }, { "epoch": 1.1, "learning_rate": 2.223672378035882e-06, "loss": 0.2229, "step": 5814 }, { "epoch": 1.1, "learning_rate": 2.2229121788944726e-06, "loss": 0.2643, "step": 5815 }, { "epoch": 1.1, "learning_rate": 2.2221520056916e-06, "loss": 0.2773, "step": 5816 }, { "epoch": 1.1, "learning_rate": 2.221391858498424e-06, "loss": 0.2241, "step": 5817 }, { "epoch": 1.1, "learning_rate": 2.2206317373861024e-06, "loss": 0.2895, "step": 5818 }, { "epoch": 1.1, "learning_rate": 2.219871642425792e-06, "loss": 0.1907, "step": 5819 }, { "epoch": 1.1, "learning_rate": 2.2191115736886467e-06, "loss": 0.2394, "step": 5820 }, { "epoch": 1.1, "learning_rate": 2.2183515312458156e-06, "loss": 0.2413, "step": 5821 }, { "epoch": 1.1, "learning_rate": 2.217591515168448e-06, "loss": 0.2872, "step": 5822 }, { "epoch": 1.1, "learning_rate": 2.2168315255276922e-06, "loss": 0.2988, "step": 5823 }, { "epoch": 1.1, "learning_rate": 2.2160715623946888e-06, "loss": 0.2231, "step": 5824 }, { "epoch": 1.1, "learning_rate": 2.2153116258405793e-06, "loss": 0.2219, "step": 5825 }, { "epoch": 1.1, "learning_rate": 2.2145517159365044e-06, "loss": 0.2489, "step": 5826 }, { "epoch": 1.1, "learning_rate": 2.2137918327535983e-06, "loss": 0.2455, "step": 5827 }, { "epoch": 1.1, "learning_rate": 2.2130319763629945e-06, "loss": 0.255, "step": 5828 }, { "epoch": 1.1, "learning_rate": 2.2122721468358254e-06, "loss": 0.2925, "step": 5829 }, { "epoch": 1.1, "learning_rate": 2.211512344243219e-06, "loss": 0.1809, "step": 5830 }, { "epoch": 1.1, "learning_rate": 2.2107525686563e-06, "loss": 0.2546, "step": 5831 }, { "epoch": 1.1, "learning_rate": 2.209992820146195e-06, "loss": 0.2623, "step": 5832 }, { "epoch": 1.1, "learning_rate": 2.2092330987840217e-06, "loss": 0.228, "step": 5833 }, { "epoch": 1.1, "learning_rate": 2.208473404640901e-06, "loss": 0.2422, "step": 5834 }, { "epoch": 1.1, "learning_rate": 2.207713737787948e-06, "loss": 0.1977, "step": 5835 }, { "epoch": 1.1, "learning_rate": 2.206954098296275e-06, "loss": 0.3085, "step": 5836 }, { "epoch": 1.1, "learning_rate": 2.2061944862369946e-06, "loss": 0.2132, "step": 5837 }, { "epoch": 1.1, "learning_rate": 2.2054349016812143e-06, "loss": 0.2044, "step": 5838 }, { "epoch": 1.1, "learning_rate": 2.2046753447000386e-06, "loss": 0.2394, "step": 5839 }, { "epoch": 1.1, "learning_rate": 2.2039158153645724e-06, "loss": 0.2026, "step": 5840 }, { "epoch": 1.1, "learning_rate": 2.203156313745916e-06, "loss": 0.2692, "step": 5841 }, { "epoch": 1.1, "learning_rate": 2.2023968399151653e-06, "loss": 0.2651, "step": 5842 }, { "epoch": 1.1, "learning_rate": 2.2016373939434166e-06, "loss": 0.2791, "step": 5843 }, { "epoch": 1.1, "learning_rate": 2.2008779759017646e-06, "loss": 0.2875, "step": 5844 }, { "epoch": 1.1, "learning_rate": 2.2001185858612973e-06, "loss": 0.2316, "step": 5845 }, { "epoch": 1.1, "learning_rate": 2.199359223893102e-06, "loss": 0.2537, "step": 5846 }, { "epoch": 1.1, "learning_rate": 2.1985998900682647e-06, "loss": 0.1887, "step": 5847 }, { "epoch": 1.1, "learning_rate": 2.1978405844578673e-06, "loss": 0.2951, "step": 5848 }, { "epoch": 1.11, "learning_rate": 2.1970813071329877e-06, "loss": 0.2945, "step": 5849 }, { "epoch": 1.11, "learning_rate": 2.1963220581647054e-06, "loss": 0.2114, "step": 5850 }, { "epoch": 1.11, "learning_rate": 2.195562837624093e-06, "loss": 0.249, "step": 5851 }, { "epoch": 1.11, "learning_rate": 2.1948036455822218e-06, "loss": 0.2619, "step": 5852 }, { "epoch": 1.11, "learning_rate": 2.1940444821101624e-06, "loss": 0.2703, "step": 5853 }, { "epoch": 1.11, "learning_rate": 2.1932853472789785e-06, "loss": 0.2197, "step": 5854 }, { "epoch": 1.11, "learning_rate": 2.1925262411597364e-06, "loss": 0.188, "step": 5855 }, { "epoch": 1.11, "learning_rate": 2.191767163823496e-06, "loss": 0.3246, "step": 5856 }, { "epoch": 1.11, "learning_rate": 2.1910081153413142e-06, "loss": 0.1945, "step": 5857 }, { "epoch": 1.11, "learning_rate": 2.1902490957842476e-06, "loss": 0.2342, "step": 5858 }, { "epoch": 1.11, "learning_rate": 2.1894901052233486e-06, "loss": 0.2074, "step": 5859 }, { "epoch": 1.11, "learning_rate": 2.188731143729669e-06, "loss": 0.2076, "step": 5860 }, { "epoch": 1.11, "learning_rate": 2.1879722113742528e-06, "loss": 0.297, "step": 5861 }, { "epoch": 1.11, "learning_rate": 2.187213308228147e-06, "loss": 0.239, "step": 5862 }, { "epoch": 1.11, "learning_rate": 2.186454434362394e-06, "loss": 0.3138, "step": 5863 }, { "epoch": 1.11, "learning_rate": 2.1856955898480304e-06, "loss": 0.2638, "step": 5864 }, { "epoch": 1.11, "learning_rate": 2.1849367747560945e-06, "loss": 0.3198, "step": 5865 }, { "epoch": 1.11, "learning_rate": 2.18417798915762e-06, "loss": 0.2628, "step": 5866 }, { "epoch": 1.11, "learning_rate": 2.183419233123636e-06, "loss": 0.3091, "step": 5867 }, { "epoch": 1.11, "learning_rate": 2.1826605067251724e-06, "loss": 0.2773, "step": 5868 }, { "epoch": 1.11, "learning_rate": 2.1819018100332543e-06, "loss": 0.224, "step": 5869 }, { "epoch": 1.11, "learning_rate": 2.181143143118903e-06, "loss": 0.2648, "step": 5870 }, { "epoch": 1.11, "learning_rate": 2.1803845060531393e-06, "loss": 0.2964, "step": 5871 }, { "epoch": 1.11, "learning_rate": 2.179625898906981e-06, "loss": 0.2889, "step": 5872 }, { "epoch": 1.11, "learning_rate": 2.17886732175144e-06, "loss": 0.279, "step": 5873 }, { "epoch": 1.11, "learning_rate": 2.178108774657529e-06, "loss": 0.2986, "step": 5874 }, { "epoch": 1.11, "learning_rate": 2.1773502576962572e-06, "loss": 0.2638, "step": 5875 }, { "epoch": 1.11, "learning_rate": 2.176591770938629e-06, "loss": 0.2763, "step": 5876 }, { "epoch": 1.11, "learning_rate": 2.1758333144556473e-06, "loss": 0.2824, "step": 5877 }, { "epoch": 1.11, "learning_rate": 2.175074888318314e-06, "loss": 0.2796, "step": 5878 }, { "epoch": 1.11, "learning_rate": 2.174316492597624e-06, "loss": 0.2411, "step": 5879 }, { "epoch": 1.11, "learning_rate": 2.1735581273645724e-06, "loss": 0.2912, "step": 5880 }, { "epoch": 1.11, "learning_rate": 2.172799792690152e-06, "loss": 0.2392, "step": 5881 }, { "epoch": 1.11, "learning_rate": 2.1720414886453496e-06, "loss": 0.3047, "step": 5882 }, { "epoch": 1.11, "learning_rate": 2.1712832153011516e-06, "loss": 0.2512, "step": 5883 }, { "epoch": 1.11, "learning_rate": 2.1705249727285425e-06, "loss": 0.268, "step": 5884 }, { "epoch": 1.11, "learning_rate": 2.1697667609984995e-06, "loss": 0.2614, "step": 5885 }, { "epoch": 1.11, "learning_rate": 2.1690085801820017e-06, "loss": 0.184, "step": 5886 }, { "epoch": 1.11, "learning_rate": 2.1682504303500233e-06, "loss": 0.1942, "step": 5887 }, { "epoch": 1.11, "learning_rate": 2.1674923115735342e-06, "loss": 0.2808, "step": 5888 }, { "epoch": 1.11, "learning_rate": 2.1667342239235045e-06, "loss": 0.2655, "step": 5889 }, { "epoch": 1.11, "learning_rate": 2.1659761674708996e-06, "loss": 0.2847, "step": 5890 }, { "epoch": 1.11, "learning_rate": 2.1652181422866804e-06, "loss": 0.1966, "step": 5891 }, { "epoch": 1.11, "learning_rate": 2.1644601484418078e-06, "loss": 0.281, "step": 5892 }, { "epoch": 1.11, "learning_rate": 2.16370218600724e-06, "loss": 0.2114, "step": 5893 }, { "epoch": 1.11, "learning_rate": 2.1629442550539283e-06, "loss": 0.2797, "step": 5894 }, { "epoch": 1.11, "learning_rate": 2.162186355652824e-06, "loss": 0.2314, "step": 5895 }, { "epoch": 1.11, "learning_rate": 2.161428487874877e-06, "loss": 0.2524, "step": 5896 }, { "epoch": 1.11, "learning_rate": 2.16067065179103e-06, "loss": 0.1917, "step": 5897 }, { "epoch": 1.11, "learning_rate": 2.159912847472225e-06, "loss": 0.275, "step": 5898 }, { "epoch": 1.11, "learning_rate": 2.1591550749894034e-06, "loss": 0.2221, "step": 5899 }, { "epoch": 1.11, "learning_rate": 2.158397334413499e-06, "loss": 0.2057, "step": 5900 }, { "epoch": 1.11, "learning_rate": 2.157639625815445e-06, "loss": 0.2514, "step": 5901 }, { "epoch": 1.12, "learning_rate": 2.1568819492661726e-06, "loss": 0.363, "step": 5902 }, { "epoch": 1.12, "learning_rate": 2.156124304836607e-06, "loss": 0.1938, "step": 5903 }, { "epoch": 1.12, "learning_rate": 2.1553666925976745e-06, "loss": 0.334, "step": 5904 }, { "epoch": 1.12, "learning_rate": 2.1546091126202955e-06, "loss": 0.2548, "step": 5905 }, { "epoch": 1.12, "learning_rate": 2.153851564975386e-06, "loss": 0.3686, "step": 5906 }, { "epoch": 1.12, "learning_rate": 2.1530940497338635e-06, "loss": 0.2098, "step": 5907 }, { "epoch": 1.12, "learning_rate": 2.152336566966639e-06, "loss": 0.2769, "step": 5908 }, { "epoch": 1.12, "learning_rate": 2.151579116744621e-06, "loss": 0.1956, "step": 5909 }, { "epoch": 1.12, "learning_rate": 2.150821699138716e-06, "loss": 0.2617, "step": 5910 }, { "epoch": 1.12, "learning_rate": 2.1500643142198265e-06, "loss": 0.2179, "step": 5911 }, { "epoch": 1.12, "learning_rate": 2.149306962058852e-06, "loss": 0.1853, "step": 5912 }, { "epoch": 1.12, "learning_rate": 2.1485496427266888e-06, "loss": 0.2682, "step": 5913 }, { "epoch": 1.12, "learning_rate": 2.147792356294233e-06, "loss": 0.2141, "step": 5914 }, { "epoch": 1.12, "learning_rate": 2.147035102832372e-06, "loss": 0.2396, "step": 5915 }, { "epoch": 1.12, "learning_rate": 2.146277882411994e-06, "loss": 0.2576, "step": 5916 }, { "epoch": 1.12, "learning_rate": 2.1455206951039854e-06, "loss": 0.2134, "step": 5917 }, { "epoch": 1.12, "learning_rate": 2.1447635409792252e-06, "loss": 0.286, "step": 5918 }, { "epoch": 1.12, "learning_rate": 2.144006420108592e-06, "loss": 0.2694, "step": 5919 }, { "epoch": 1.12, "learning_rate": 2.1432493325629613e-06, "loss": 0.2817, "step": 5920 }, { "epoch": 1.12, "learning_rate": 2.142492278413205e-06, "loss": 0.2647, "step": 5921 }, { "epoch": 1.12, "learning_rate": 2.1417352577301907e-06, "loss": 0.3365, "step": 5922 }, { "epoch": 1.12, "learning_rate": 2.1409782705847853e-06, "loss": 0.241, "step": 5923 }, { "epoch": 1.12, "learning_rate": 2.1402213170478523e-06, "loss": 0.2499, "step": 5924 }, { "epoch": 1.12, "learning_rate": 2.139464397190249e-06, "loss": 0.2685, "step": 5925 }, { "epoch": 1.12, "learning_rate": 2.1387075110828316e-06, "loss": 0.2824, "step": 5926 }, { "epoch": 1.12, "learning_rate": 2.1379506587964554e-06, "loss": 0.3138, "step": 5927 }, { "epoch": 1.12, "learning_rate": 2.137193840401968e-06, "loss": 0.2169, "step": 5928 }, { "epoch": 1.12, "learning_rate": 2.1364370559702168e-06, "loss": 0.2494, "step": 5929 }, { "epoch": 1.12, "learning_rate": 2.1356803055720464e-06, "loss": 0.1925, "step": 5930 }, { "epoch": 1.12, "learning_rate": 2.1349235892782956e-06, "loss": 0.3061, "step": 5931 }, { "epoch": 1.12, "learning_rate": 2.134166907159802e-06, "loss": 0.3011, "step": 5932 }, { "epoch": 1.12, "learning_rate": 2.1334102592874006e-06, "loss": 0.2167, "step": 5933 }, { "epoch": 1.12, "learning_rate": 2.13265364573192e-06, "loss": 0.2354, "step": 5934 }, { "epoch": 1.12, "learning_rate": 2.1318970665641896e-06, "loss": 0.3073, "step": 5935 }, { "epoch": 1.12, "learning_rate": 2.131140521855034e-06, "loss": 0.3193, "step": 5936 }, { "epoch": 1.12, "learning_rate": 2.130384011675272e-06, "loss": 0.2686, "step": 5937 }, { "epoch": 1.12, "learning_rate": 2.129627536095724e-06, "loss": 0.2552, "step": 5938 }, { "epoch": 1.12, "learning_rate": 2.1288710951872035e-06, "loss": 0.2852, "step": 5939 }, { "epoch": 1.12, "learning_rate": 2.128114689020521e-06, "loss": 0.2385, "step": 5940 }, { "epoch": 1.12, "learning_rate": 2.127358317666486e-06, "loss": 0.3142, "step": 5941 }, { "epoch": 1.12, "learning_rate": 2.1266019811959035e-06, "loss": 0.2685, "step": 5942 }, { "epoch": 1.12, "learning_rate": 2.1258456796795735e-06, "loss": 0.2764, "step": 5943 }, { "epoch": 1.12, "learning_rate": 2.125089413188295e-06, "loss": 0.3792, "step": 5944 }, { "epoch": 1.12, "learning_rate": 2.1243331817928645e-06, "loss": 0.2295, "step": 5945 }, { "epoch": 1.12, "learning_rate": 2.1235769855640724e-06, "loss": 0.2431, "step": 5946 }, { "epoch": 1.12, "learning_rate": 2.1228208245727065e-06, "loss": 0.2736, "step": 5947 }, { "epoch": 1.12, "learning_rate": 2.122064698889554e-06, "loss": 0.2678, "step": 5948 }, { "epoch": 1.12, "learning_rate": 2.1213086085853953e-06, "loss": 0.2894, "step": 5949 }, { "epoch": 1.12, "learning_rate": 2.1205525537310083e-06, "loss": 0.2447, "step": 5950 }, { "epoch": 1.12, "learning_rate": 2.1197965343971707e-06, "loss": 0.2427, "step": 5951 }, { "epoch": 1.12, "learning_rate": 2.1190405506546526e-06, "loss": 0.2601, "step": 5952 }, { "epoch": 1.12, "learning_rate": 2.118284602574222e-06, "loss": 0.2825, "step": 5953 }, { "epoch": 1.12, "learning_rate": 2.117528690226646e-06, "loss": 0.2646, "step": 5954 }, { "epoch": 1.13, "learning_rate": 2.116772813682685e-06, "loss": 0.1783, "step": 5955 }, { "epoch": 1.13, "learning_rate": 2.116016973013098e-06, "loss": 0.2877, "step": 5956 }, { "epoch": 1.13, "learning_rate": 2.115261168288641e-06, "loss": 0.1999, "step": 5957 }, { "epoch": 1.13, "learning_rate": 2.114505399580064e-06, "loss": 0.2344, "step": 5958 }, { "epoch": 1.13, "learning_rate": 2.1137496669581172e-06, "loss": 0.2791, "step": 5959 }, { "epoch": 1.13, "learning_rate": 2.1129939704935456e-06, "loss": 0.25, "step": 5960 }, { "epoch": 1.13, "learning_rate": 2.1122383102570895e-06, "loss": 0.2617, "step": 5961 }, { "epoch": 1.13, "learning_rate": 2.1114826863194882e-06, "loss": 0.1916, "step": 5962 }, { "epoch": 1.13, "learning_rate": 2.110727098751477e-06, "loss": 0.2874, "step": 5963 }, { "epoch": 1.13, "learning_rate": 2.1099715476237855e-06, "loss": 0.1866, "step": 5964 }, { "epoch": 1.13, "learning_rate": 2.109216033007143e-06, "loss": 0.2598, "step": 5965 }, { "epoch": 1.13, "learning_rate": 2.1084605549722757e-06, "loss": 0.2444, "step": 5966 }, { "epoch": 1.13, "learning_rate": 2.107705113589903e-06, "loss": 0.2003, "step": 5967 }, { "epoch": 1.13, "learning_rate": 2.1069497089307423e-06, "loss": 0.3052, "step": 5968 }, { "epoch": 1.13, "learning_rate": 2.1061943410655095e-06, "loss": 0.1678, "step": 5969 }, { "epoch": 1.13, "learning_rate": 2.1054390100649145e-06, "loss": 0.2476, "step": 5970 }, { "epoch": 1.13, "learning_rate": 2.1046837159996647e-06, "loss": 0.2027, "step": 5971 }, { "epoch": 1.13, "learning_rate": 2.1039284589404653e-06, "loss": 0.1699, "step": 5972 }, { "epoch": 1.13, "learning_rate": 2.1031732389580156e-06, "loss": 0.1894, "step": 5973 }, { "epoch": 1.13, "learning_rate": 2.1024180561230127e-06, "loss": 0.2467, "step": 5974 }, { "epoch": 1.13, "learning_rate": 2.1016629105061514e-06, "loss": 0.229, "step": 5975 }, { "epoch": 1.13, "learning_rate": 2.1009078021781202e-06, "loss": 0.2518, "step": 5976 }, { "epoch": 1.13, "learning_rate": 2.100152731209607e-06, "loss": 0.2789, "step": 5977 }, { "epoch": 1.13, "learning_rate": 2.0993976976712954e-06, "loss": 0.3007, "step": 5978 }, { "epoch": 1.13, "learning_rate": 2.0986427016338626e-06, "loss": 0.2242, "step": 5979 }, { "epoch": 1.13, "learning_rate": 2.097887743167987e-06, "loss": 0.2391, "step": 5980 }, { "epoch": 1.13, "learning_rate": 2.097132822344341e-06, "loss": 0.1778, "step": 5981 }, { "epoch": 1.13, "learning_rate": 2.096377939233592e-06, "loss": 0.2755, "step": 5982 }, { "epoch": 1.13, "learning_rate": 2.095623093906407e-06, "loss": 0.2058, "step": 5983 }, { "epoch": 1.13, "learning_rate": 2.0948682864334484e-06, "loss": 0.2374, "step": 5984 }, { "epoch": 1.13, "learning_rate": 2.0941135168853734e-06, "loss": 0.1942, "step": 5985 }, { "epoch": 1.13, "learning_rate": 2.093358785332837e-06, "loss": 0.2463, "step": 5986 }, { "epoch": 1.13, "learning_rate": 2.092604091846491e-06, "loss": 0.342, "step": 5987 }, { "epoch": 1.13, "learning_rate": 2.0918494364969845e-06, "loss": 0.3155, "step": 5988 }, { "epoch": 1.13, "learning_rate": 2.0910948193549595e-06, "loss": 0.2617, "step": 5989 }, { "epoch": 1.13, "learning_rate": 2.0903402404910576e-06, "loss": 0.2177, "step": 5990 }, { "epoch": 1.13, "learning_rate": 2.0895856999759172e-06, "loss": 0.3213, "step": 5991 }, { "epoch": 1.13, "learning_rate": 2.0888311978801687e-06, "loss": 0.3219, "step": 5992 }, { "epoch": 1.13, "learning_rate": 2.0880767342744447e-06, "loss": 0.2417, "step": 5993 }, { "epoch": 1.13, "learning_rate": 2.0873223092293704e-06, "loss": 0.2676, "step": 5994 }, { "epoch": 1.13, "learning_rate": 2.086567922815568e-06, "loss": 0.268, "step": 5995 }, { "epoch": 1.13, "learning_rate": 2.085813575103657e-06, "loss": 0.3036, "step": 5996 }, { "epoch": 1.13, "learning_rate": 2.0850592661642545e-06, "loss": 0.2434, "step": 5997 }, { "epoch": 1.13, "learning_rate": 2.0843049960679697e-06, "loss": 0.2651, "step": 5998 }, { "epoch": 1.13, "learning_rate": 2.0835507648854115e-06, "loss": 0.2518, "step": 5999 }, { "epoch": 1.13, "learning_rate": 2.0827965726871856e-06, "loss": 0.2462, "step": 6000 }, { "epoch": 1.13, "learning_rate": 2.0820424195438917e-06, "loss": 0.3081, "step": 6001 }, { "epoch": 1.13, "learning_rate": 2.0812883055261266e-06, "loss": 0.235, "step": 6002 }, { "epoch": 1.13, "learning_rate": 2.080534230704486e-06, "loss": 0.2633, "step": 6003 }, { "epoch": 1.13, "learning_rate": 2.079780195149557e-06, "loss": 0.3012, "step": 6004 }, { "epoch": 1.13, "learning_rate": 2.0790261989319275e-06, "loss": 0.2156, "step": 6005 }, { "epoch": 1.13, "learning_rate": 2.0782722421221807e-06, "loss": 0.2064, "step": 6006 }, { "epoch": 1.13, "learning_rate": 2.0775183247908933e-06, "loss": 0.2312, "step": 6007 }, { "epoch": 1.14, "learning_rate": 2.076764447008642e-06, "loss": 0.2776, "step": 6008 }, { "epoch": 1.14, "learning_rate": 2.076010608845998e-06, "loss": 0.2126, "step": 6009 }, { "epoch": 1.14, "learning_rate": 2.075256810373528e-06, "loss": 0.1886, "step": 6010 }, { "epoch": 1.14, "learning_rate": 2.0745030516617974e-06, "loss": 0.2266, "step": 6011 }, { "epoch": 1.14, "learning_rate": 2.0737493327813667e-06, "loss": 0.2328, "step": 6012 }, { "epoch": 1.14, "learning_rate": 2.072995653802791e-06, "loss": 0.2748, "step": 6013 }, { "epoch": 1.14, "learning_rate": 2.0722420147966226e-06, "loss": 0.2994, "step": 6014 }, { "epoch": 1.14, "learning_rate": 2.071488415833414e-06, "loss": 0.2531, "step": 6015 }, { "epoch": 1.14, "learning_rate": 2.070734856983707e-06, "loss": 0.291, "step": 6016 }, { "epoch": 1.14, "learning_rate": 2.0699813383180447e-06, "loss": 0.2814, "step": 6017 }, { "epoch": 1.14, "learning_rate": 2.0692278599069656e-06, "loss": 0.181, "step": 6018 }, { "epoch": 1.14, "learning_rate": 2.0684744218210024e-06, "loss": 0.2603, "step": 6019 }, { "epoch": 1.14, "learning_rate": 2.067721024130685e-06, "loss": 0.1751, "step": 6020 }, { "epoch": 1.14, "learning_rate": 2.0669676669065426e-06, "loss": 0.2872, "step": 6021 }, { "epoch": 1.14, "learning_rate": 2.066214350219095e-06, "loss": 0.274, "step": 6022 }, { "epoch": 1.14, "learning_rate": 2.065461074138862e-06, "loss": 0.2284, "step": 6023 }, { "epoch": 1.14, "learning_rate": 2.06470783873636e-06, "loss": 0.3182, "step": 6024 }, { "epoch": 1.14, "learning_rate": 2.0639546440820977e-06, "loss": 0.214, "step": 6025 }, { "epoch": 1.14, "learning_rate": 2.0632014902465845e-06, "loss": 0.2229, "step": 6026 }, { "epoch": 1.14, "learning_rate": 2.0624483773003253e-06, "loss": 0.2783, "step": 6027 }, { "epoch": 1.14, "learning_rate": 2.0616953053138163e-06, "loss": 0.2723, "step": 6028 }, { "epoch": 1.14, "learning_rate": 2.060942274357556e-06, "loss": 0.1604, "step": 6029 }, { "epoch": 1.14, "learning_rate": 2.060189284502037e-06, "loss": 0.2273, "step": 6030 }, { "epoch": 1.14, "learning_rate": 2.059436335817746e-06, "loss": 0.2287, "step": 6031 }, { "epoch": 1.14, "learning_rate": 2.058683428375168e-06, "loss": 0.194, "step": 6032 }, { "epoch": 1.14, "learning_rate": 2.057930562244785e-06, "loss": 0.3222, "step": 6033 }, { "epoch": 1.14, "learning_rate": 2.0571777374970707e-06, "loss": 0.2664, "step": 6034 }, { "epoch": 1.14, "learning_rate": 2.0564249542025e-06, "loss": 0.2325, "step": 6035 }, { "epoch": 1.14, "learning_rate": 2.055672212431543e-06, "loss": 0.2226, "step": 6036 }, { "epoch": 1.14, "learning_rate": 2.054919512254663e-06, "loss": 0.2927, "step": 6037 }, { "epoch": 1.14, "learning_rate": 2.054166853742321e-06, "loss": 0.2453, "step": 6038 }, { "epoch": 1.14, "learning_rate": 2.053414236964976e-06, "loss": 0.2257, "step": 6039 }, { "epoch": 1.14, "learning_rate": 2.0526616619930796e-06, "loss": 0.195, "step": 6040 }, { "epoch": 1.14, "learning_rate": 2.0519091288970814e-06, "loss": 0.266, "step": 6041 }, { "epoch": 1.14, "learning_rate": 2.051156637747429e-06, "loss": 0.2022, "step": 6042 }, { "epoch": 1.14, "learning_rate": 2.0504041886145625e-06, "loss": 0.2492, "step": 6043 }, { "epoch": 1.14, "learning_rate": 2.0496517815689185e-06, "loss": 0.2806, "step": 6044 }, { "epoch": 1.14, "learning_rate": 2.048899416680933e-06, "loss": 0.2942, "step": 6045 }, { "epoch": 1.14, "learning_rate": 2.0481470940210342e-06, "loss": 0.2775, "step": 6046 }, { "epoch": 1.14, "learning_rate": 2.0473948136596487e-06, "loss": 0.2933, "step": 6047 }, { "epoch": 1.14, "learning_rate": 2.046642575667198e-06, "loss": 0.2072, "step": 6048 }, { "epoch": 1.14, "learning_rate": 2.0458903801141013e-06, "loss": 0.2374, "step": 6049 }, { "epoch": 1.14, "learning_rate": 2.045138227070771e-06, "loss": 0.2284, "step": 6050 }, { "epoch": 1.14, "learning_rate": 2.044386116607617e-06, "loss": 0.2107, "step": 6051 }, { "epoch": 1.14, "learning_rate": 2.043634048795047e-06, "loss": 0.2121, "step": 6052 }, { "epoch": 1.14, "learning_rate": 2.0428820237034618e-06, "loss": 0.2765, "step": 6053 }, { "epoch": 1.14, "learning_rate": 2.0421300414032588e-06, "loss": 0.3046, "step": 6054 }, { "epoch": 1.14, "learning_rate": 2.0413781019648333e-06, "loss": 0.234, "step": 6055 }, { "epoch": 1.14, "learning_rate": 2.040626205458574e-06, "loss": 0.2298, "step": 6056 }, { "epoch": 1.14, "learning_rate": 2.0398743519548683e-06, "loss": 0.2204, "step": 6057 }, { "epoch": 1.14, "learning_rate": 2.039122541524098e-06, "loss": 0.2958, "step": 6058 }, { "epoch": 1.14, "learning_rate": 2.0383707742366393e-06, "loss": 0.2754, "step": 6059 }, { "epoch": 1.14, "learning_rate": 2.0376190501628675e-06, "loss": 0.2276, "step": 6060 }, { "epoch": 1.15, "learning_rate": 2.036867369373153e-06, "loss": 0.2613, "step": 6061 }, { "epoch": 1.15, "learning_rate": 2.0361157319378597e-06, "loss": 0.2404, "step": 6062 }, { "epoch": 1.15, "learning_rate": 2.035364137927351e-06, "loss": 0.2824, "step": 6063 }, { "epoch": 1.15, "learning_rate": 2.034612587411984e-06, "loss": 0.2277, "step": 6064 }, { "epoch": 1.15, "learning_rate": 2.033861080462111e-06, "loss": 0.2186, "step": 6065 }, { "epoch": 1.15, "learning_rate": 2.033109617148083e-06, "loss": 0.2872, "step": 6066 }, { "epoch": 1.15, "learning_rate": 2.032358197540246e-06, "loss": 0.1902, "step": 6067 }, { "epoch": 1.15, "learning_rate": 2.03160682170894e-06, "loss": 0.3093, "step": 6068 }, { "epoch": 1.15, "learning_rate": 2.0308554897245025e-06, "loss": 0.2176, "step": 6069 }, { "epoch": 1.15, "learning_rate": 2.0301042016572676e-06, "loss": 0.2655, "step": 6070 }, { "epoch": 1.15, "learning_rate": 2.029352957577563e-06, "loss": 0.2092, "step": 6071 }, { "epoch": 1.15, "learning_rate": 2.0286017575557128e-06, "loss": 0.2455, "step": 6072 }, { "epoch": 1.15, "learning_rate": 2.027850601662041e-06, "loss": 0.2695, "step": 6073 }, { "epoch": 1.15, "learning_rate": 2.027099489966861e-06, "loss": 0.2047, "step": 6074 }, { "epoch": 1.15, "learning_rate": 2.0263484225404867e-06, "loss": 0.2124, "step": 6075 }, { "epoch": 1.15, "learning_rate": 2.025597399453227e-06, "loss": 0.2795, "step": 6076 }, { "epoch": 1.15, "learning_rate": 2.0248464207753845e-06, "loss": 0.1773, "step": 6077 }, { "epoch": 1.15, "learning_rate": 2.0240954865772607e-06, "loss": 0.1851, "step": 6078 }, { "epoch": 1.15, "learning_rate": 2.0233445969291517e-06, "loss": 0.2204, "step": 6079 }, { "epoch": 1.15, "learning_rate": 2.022593751901347e-06, "loss": 0.2215, "step": 6080 }, { "epoch": 1.15, "learning_rate": 2.021842951564137e-06, "loss": 0.2134, "step": 6081 }, { "epoch": 1.15, "learning_rate": 2.0210921959878037e-06, "loss": 0.2259, "step": 6082 }, { "epoch": 1.15, "learning_rate": 2.020341485242625e-06, "loss": 0.2708, "step": 6083 }, { "epoch": 1.15, "learning_rate": 2.019590819398878e-06, "loss": 0.2357, "step": 6084 }, { "epoch": 1.15, "learning_rate": 2.018840198526833e-06, "loss": 0.2469, "step": 6085 }, { "epoch": 1.15, "learning_rate": 2.0180896226967545e-06, "loss": 0.267, "step": 6086 }, { "epoch": 1.15, "learning_rate": 2.0173390919789073e-06, "loss": 0.2166, "step": 6087 }, { "epoch": 1.15, "learning_rate": 2.016588606443549e-06, "loss": 0.2288, "step": 6088 }, { "epoch": 1.15, "learning_rate": 2.0158381661609333e-06, "loss": 0.297, "step": 6089 }, { "epoch": 1.15, "learning_rate": 2.015087771201309e-06, "loss": 0.2838, "step": 6090 }, { "epoch": 1.15, "learning_rate": 2.0143374216349232e-06, "loss": 0.1589, "step": 6091 }, { "epoch": 1.15, "learning_rate": 2.0135871175320157e-06, "loss": 0.2063, "step": 6092 }, { "epoch": 1.15, "learning_rate": 2.012836858962823e-06, "loss": 0.2464, "step": 6093 }, { "epoch": 1.15, "learning_rate": 2.01208664599758e-06, "loss": 0.2322, "step": 6094 }, { "epoch": 1.15, "learning_rate": 2.0113364787065124e-06, "loss": 0.1859, "step": 6095 }, { "epoch": 1.15, "learning_rate": 2.0105863571598455e-06, "loss": 0.2828, "step": 6096 }, { "epoch": 1.15, "learning_rate": 2.0098362814278004e-06, "loss": 0.2626, "step": 6097 }, { "epoch": 1.15, "learning_rate": 2.0090862515805896e-06, "loss": 0.2689, "step": 6098 }, { "epoch": 1.15, "learning_rate": 2.008336267688427e-06, "loss": 0.275, "step": 6099 }, { "epoch": 1.15, "learning_rate": 2.007586329821519e-06, "loss": 0.2305, "step": 6100 }, { "epoch": 1.15, "learning_rate": 2.0068364380500667e-06, "loss": 0.3539, "step": 6101 }, { "epoch": 1.15, "learning_rate": 2.0060865924442703e-06, "loss": 0.1719, "step": 6102 }, { "epoch": 1.15, "learning_rate": 2.005336793074323e-06, "loss": 0.2275, "step": 6103 }, { "epoch": 1.15, "learning_rate": 2.004587040010414e-06, "loss": 0.3736, "step": 6104 }, { "epoch": 1.15, "learning_rate": 2.003837333322729e-06, "loss": 0.2116, "step": 6105 }, { "epoch": 1.15, "learning_rate": 2.00308767308145e-06, "loss": 0.2384, "step": 6106 }, { "epoch": 1.15, "learning_rate": 2.0023380593567516e-06, "loss": 0.2184, "step": 6107 }, { "epoch": 1.15, "learning_rate": 2.0015884922188066e-06, "loss": 0.2211, "step": 6108 }, { "epoch": 1.15, "learning_rate": 2.0008389717377847e-06, "loss": 0.3354, "step": 6109 }, { "epoch": 1.15, "learning_rate": 2.0000894979838483e-06, "loss": 0.2275, "step": 6110 }, { "epoch": 1.15, "learning_rate": 1.9993400710271553e-06, "loss": 0.2203, "step": 6111 }, { "epoch": 1.15, "learning_rate": 1.998590690937862e-06, "loss": 0.2948, "step": 6112 }, { "epoch": 1.15, "learning_rate": 1.9978413577861196e-06, "loss": 0.2853, "step": 6113 }, { "epoch": 1.16, "learning_rate": 1.9970920716420718e-06, "loss": 0.2195, "step": 6114 }, { "epoch": 1.16, "learning_rate": 1.9963428325758615e-06, "loss": 0.2366, "step": 6115 }, { "epoch": 1.16, "learning_rate": 1.9955936406576264e-06, "loss": 0.2446, "step": 6116 }, { "epoch": 1.16, "learning_rate": 1.994844495957498e-06, "loss": 0.3415, "step": 6117 }, { "epoch": 1.16, "learning_rate": 1.994095398545605e-06, "loss": 0.3382, "step": 6118 }, { "epoch": 1.16, "learning_rate": 1.993346348492073e-06, "loss": 0.2372, "step": 6119 }, { "epoch": 1.16, "learning_rate": 1.9925973458670193e-06, "loss": 0.3114, "step": 6120 }, { "epoch": 1.16, "learning_rate": 1.9918483907405593e-06, "loss": 0.1984, "step": 6121 }, { "epoch": 1.16, "learning_rate": 1.991099483182805e-06, "loss": 0.2563, "step": 6122 }, { "epoch": 1.16, "learning_rate": 1.990350623263862e-06, "loss": 0.2091, "step": 6123 }, { "epoch": 1.16, "learning_rate": 1.989601811053831e-06, "loss": 0.2786, "step": 6124 }, { "epoch": 1.16, "learning_rate": 1.988853046622811e-06, "loss": 0.321, "step": 6125 }, { "epoch": 1.16, "learning_rate": 1.988104330040893e-06, "loss": 0.2728, "step": 6126 }, { "epoch": 1.16, "learning_rate": 1.9873556613781657e-06, "loss": 0.2757, "step": 6127 }, { "epoch": 1.16, "learning_rate": 1.9866070407047143e-06, "loss": 0.2148, "step": 6128 }, { "epoch": 1.16, "learning_rate": 1.985858468090616e-06, "loss": 0.2303, "step": 6129 }, { "epoch": 1.16, "learning_rate": 1.985109943605947e-06, "loss": 0.2638, "step": 6130 }, { "epoch": 1.16, "learning_rate": 1.9843614673207782e-06, "loss": 0.3275, "step": 6131 }, { "epoch": 1.16, "learning_rate": 1.983613039305173e-06, "loss": 0.2722, "step": 6132 }, { "epoch": 1.16, "learning_rate": 1.982864659629195e-06, "loss": 0.2179, "step": 6133 }, { "epoch": 1.16, "learning_rate": 1.9821163283629007e-06, "loss": 0.3206, "step": 6134 }, { "epoch": 1.16, "learning_rate": 1.9813680455763405e-06, "loss": 0.2442, "step": 6135 }, { "epoch": 1.16, "learning_rate": 1.980619811339564e-06, "loss": 0.2439, "step": 6136 }, { "epoch": 1.16, "learning_rate": 1.9798716257226146e-06, "loss": 0.2751, "step": 6137 }, { "epoch": 1.16, "learning_rate": 1.9791234887955286e-06, "loss": 0.32, "step": 6138 }, { "epoch": 1.16, "learning_rate": 1.978375400628341e-06, "loss": 0.2968, "step": 6139 }, { "epoch": 1.16, "learning_rate": 1.9776273612910833e-06, "loss": 0.2241, "step": 6140 }, { "epoch": 1.16, "learning_rate": 1.9768793708537774e-06, "loss": 0.329, "step": 6141 }, { "epoch": 1.16, "learning_rate": 1.9761314293864452e-06, "loss": 0.1966, "step": 6142 }, { "epoch": 1.16, "learning_rate": 1.9753835369591028e-06, "loss": 0.2023, "step": 6143 }, { "epoch": 1.16, "learning_rate": 1.9746356936417608e-06, "loss": 0.2657, "step": 6144 }, { "epoch": 1.16, "learning_rate": 1.9738878995044247e-06, "loss": 0.2452, "step": 6145 }, { "epoch": 1.16, "learning_rate": 1.9731401546170986e-06, "loss": 0.2214, "step": 6146 }, { "epoch": 1.16, "learning_rate": 1.972392459049778e-06, "loss": 0.2541, "step": 6147 }, { "epoch": 1.16, "learning_rate": 1.9716448128724556e-06, "loss": 0.2406, "step": 6148 }, { "epoch": 1.16, "learning_rate": 1.9708972161551215e-06, "loss": 0.2586, "step": 6149 }, { "epoch": 1.16, "learning_rate": 1.970149668967757e-06, "loss": 0.2763, "step": 6150 }, { "epoch": 1.16, "learning_rate": 1.9694021713803416e-06, "loss": 0.3035, "step": 6151 }, { "epoch": 1.16, "learning_rate": 1.9686547234628506e-06, "loss": 0.2765, "step": 6152 }, { "epoch": 1.16, "learning_rate": 1.967907325285251e-06, "loss": 0.3145, "step": 6153 }, { "epoch": 1.16, "learning_rate": 1.9671599769175104e-06, "loss": 0.1908, "step": 6154 }, { "epoch": 1.16, "learning_rate": 1.966412678429588e-06, "loss": 0.2254, "step": 6155 }, { "epoch": 1.16, "learning_rate": 1.965665429891438e-06, "loss": 0.283, "step": 6156 }, { "epoch": 1.16, "learning_rate": 1.964918231373013e-06, "loss": 0.2112, "step": 6157 }, { "epoch": 1.16, "learning_rate": 1.9641710829442597e-06, "loss": 0.2477, "step": 6158 }, { "epoch": 1.16, "learning_rate": 1.9634239846751173e-06, "loss": 0.2259, "step": 6159 }, { "epoch": 1.16, "learning_rate": 1.9626769366355234e-06, "loss": 0.3339, "step": 6160 }, { "epoch": 1.16, "learning_rate": 1.9619299388954114e-06, "loss": 0.2246, "step": 6161 }, { "epoch": 1.16, "learning_rate": 1.961182991524708e-06, "loss": 0.2272, "step": 6162 }, { "epoch": 1.16, "learning_rate": 1.960436094593335e-06, "loss": 0.2221, "step": 6163 }, { "epoch": 1.16, "learning_rate": 1.9596892481712116e-06, "loss": 0.2576, "step": 6164 }, { "epoch": 1.16, "learning_rate": 1.95894245232825e-06, "loss": 0.1961, "step": 6165 }, { "epoch": 1.16, "learning_rate": 1.9581957071343593e-06, "loss": 0.2748, "step": 6166 }, { "epoch": 1.17, "learning_rate": 1.9574490126594435e-06, "loss": 0.3068, "step": 6167 }, { "epoch": 1.17, "learning_rate": 1.956702368973401e-06, "loss": 0.2431, "step": 6168 }, { "epoch": 1.17, "learning_rate": 1.955955776146125e-06, "loss": 0.3194, "step": 6169 }, { "epoch": 1.17, "learning_rate": 1.955209234247508e-06, "loss": 0.2548, "step": 6170 }, { "epoch": 1.17, "learning_rate": 1.954462743347432e-06, "loss": 0.2132, "step": 6171 }, { "epoch": 1.17, "learning_rate": 1.953716303515778e-06, "loss": 0.2475, "step": 6172 }, { "epoch": 1.17, "learning_rate": 1.95296991482242e-06, "loss": 0.2705, "step": 6173 }, { "epoch": 1.17, "learning_rate": 1.9522235773372313e-06, "loss": 0.2705, "step": 6174 }, { "epoch": 1.17, "learning_rate": 1.951477291130075e-06, "loss": 0.1895, "step": 6175 }, { "epoch": 1.17, "learning_rate": 1.9507310562708115e-06, "loss": 0.2192, "step": 6176 }, { "epoch": 1.17, "learning_rate": 1.949984872829299e-06, "loss": 0.1988, "step": 6177 }, { "epoch": 1.17, "learning_rate": 1.9492387408753864e-06, "loss": 0.2698, "step": 6178 }, { "epoch": 1.17, "learning_rate": 1.948492660478922e-06, "loss": 0.3188, "step": 6179 }, { "epoch": 1.17, "learning_rate": 1.947746631709746e-06, "loss": 0.3123, "step": 6180 }, { "epoch": 1.17, "learning_rate": 1.9470006546376953e-06, "loss": 0.2533, "step": 6181 }, { "epoch": 1.17, "learning_rate": 1.946254729332602e-06, "loss": 0.3181, "step": 6182 }, { "epoch": 1.17, "learning_rate": 1.9455088558642932e-06, "loss": 0.3446, "step": 6183 }, { "epoch": 1.17, "learning_rate": 1.94476303430259e-06, "loss": 0.2538, "step": 6184 }, { "epoch": 1.17, "learning_rate": 1.944017264717312e-06, "loss": 0.2179, "step": 6185 }, { "epoch": 1.17, "learning_rate": 1.9432715471782694e-06, "loss": 0.1817, "step": 6186 }, { "epoch": 1.17, "learning_rate": 1.94252588175527e-06, "loss": 0.2807, "step": 6187 }, { "epoch": 1.17, "learning_rate": 1.9417802685181163e-06, "loss": 0.2352, "step": 6188 }, { "epoch": 1.17, "learning_rate": 1.941034707536609e-06, "loss": 0.2892, "step": 6189 }, { "epoch": 1.17, "learning_rate": 1.9402891988805368e-06, "loss": 0.2329, "step": 6190 }, { "epoch": 1.17, "learning_rate": 1.93954374261969e-06, "loss": 0.2922, "step": 6191 }, { "epoch": 1.17, "learning_rate": 1.9387983388238523e-06, "loss": 0.3578, "step": 6192 }, { "epoch": 1.17, "learning_rate": 1.9380529875628e-06, "loss": 0.2814, "step": 6193 }, { "epoch": 1.17, "learning_rate": 1.937307688906307e-06, "loss": 0.1971, "step": 6194 }, { "epoch": 1.17, "learning_rate": 1.9365624429241427e-06, "loss": 0.2206, "step": 6195 }, { "epoch": 1.17, "learning_rate": 1.9358172496860695e-06, "loss": 0.2492, "step": 6196 }, { "epoch": 1.17, "learning_rate": 1.9350721092618453e-06, "loss": 0.1871, "step": 6197 }, { "epoch": 1.17, "learning_rate": 1.9343270217212256e-06, "loss": 0.2629, "step": 6198 }, { "epoch": 1.17, "learning_rate": 1.9335819871339566e-06, "loss": 0.2669, "step": 6199 }, { "epoch": 1.17, "learning_rate": 1.9328370055697836e-06, "loss": 0.2357, "step": 6200 }, { "epoch": 1.17, "learning_rate": 1.932092077098445e-06, "loss": 0.2667, "step": 6201 }, { "epoch": 1.17, "learning_rate": 1.931347201789673e-06, "loss": 0.2487, "step": 6202 }, { "epoch": 1.17, "learning_rate": 1.930602379713199e-06, "loss": 0.262, "step": 6203 }, { "epoch": 1.17, "learning_rate": 1.929857610938745e-06, "loss": 0.224, "step": 6204 }, { "epoch": 1.17, "learning_rate": 1.9291128955360293e-06, "loss": 0.149, "step": 6205 }, { "epoch": 1.17, "learning_rate": 1.9283682335747664e-06, "loss": 0.2252, "step": 6206 }, { "epoch": 1.17, "learning_rate": 1.9276236251246655e-06, "loss": 0.1894, "step": 6207 }, { "epoch": 1.17, "learning_rate": 1.926879070255429e-06, "loss": 0.2499, "step": 6208 }, { "epoch": 1.17, "learning_rate": 1.926134569036756e-06, "loss": 0.2195, "step": 6209 }, { "epoch": 1.17, "learning_rate": 1.925390121538342e-06, "loss": 0.2847, "step": 6210 }, { "epoch": 1.17, "learning_rate": 1.9246457278298735e-06, "loss": 0.2319, "step": 6211 }, { "epoch": 1.17, "learning_rate": 1.9239013879810344e-06, "loss": 0.3323, "step": 6212 }, { "epoch": 1.17, "learning_rate": 1.9231571020615044e-06, "loss": 0.2459, "step": 6213 }, { "epoch": 1.17, "learning_rate": 1.922412870140956e-06, "loss": 0.2093, "step": 6214 }, { "epoch": 1.17, "learning_rate": 1.9216686922890577e-06, "loss": 0.2235, "step": 6215 }, { "epoch": 1.17, "learning_rate": 1.920924568575474e-06, "loss": 0.2425, "step": 6216 }, { "epoch": 1.17, "learning_rate": 1.920180499069862e-06, "loss": 0.2176, "step": 6217 }, { "epoch": 1.17, "learning_rate": 1.919436483841875e-06, "loss": 0.1865, "step": 6218 }, { "epoch": 1.17, "learning_rate": 1.9186925229611627e-06, "loss": 0.2228, "step": 6219 }, { "epoch": 1.18, "learning_rate": 1.917948616497366e-06, "loss": 0.2785, "step": 6220 }, { "epoch": 1.18, "learning_rate": 1.9172047645201248e-06, "loss": 0.1926, "step": 6221 }, { "epoch": 1.18, "learning_rate": 1.9164609670990717e-06, "loss": 0.2592, "step": 6222 }, { "epoch": 1.18, "learning_rate": 1.9157172243038328e-06, "loss": 0.2652, "step": 6223 }, { "epoch": 1.18, "learning_rate": 1.914973536204033e-06, "loss": 0.2732, "step": 6224 }, { "epoch": 1.18, "learning_rate": 1.914229902869289e-06, "loss": 0.2471, "step": 6225 }, { "epoch": 1.18, "learning_rate": 1.9134863243692123e-06, "loss": 0.2051, "step": 6226 }, { "epoch": 1.18, "learning_rate": 1.9127428007734115e-06, "loss": 0.2914, "step": 6227 }, { "epoch": 1.18, "learning_rate": 1.911999332151489e-06, "loss": 0.2039, "step": 6228 }, { "epoch": 1.18, "learning_rate": 1.91125591857304e-06, "loss": 0.2375, "step": 6229 }, { "epoch": 1.18, "learning_rate": 1.910512560107658e-06, "loss": 0.2801, "step": 6230 }, { "epoch": 1.18, "learning_rate": 1.90976925682493e-06, "loss": 0.2344, "step": 6231 }, { "epoch": 1.18, "learning_rate": 1.909026008794436e-06, "loss": 0.1678, "step": 6232 }, { "epoch": 1.18, "learning_rate": 1.908282816085753e-06, "loss": 0.2498, "step": 6233 }, { "epoch": 1.18, "learning_rate": 1.9075396787684534e-06, "loss": 0.2365, "step": 6234 }, { "epoch": 1.18, "learning_rate": 1.9067965969121014e-06, "loss": 0.2645, "step": 6235 }, { "epoch": 1.18, "learning_rate": 1.9060535705862582e-06, "loss": 0.2498, "step": 6236 }, { "epoch": 1.18, "learning_rate": 1.9053105998604805e-06, "loss": 0.3496, "step": 6237 }, { "epoch": 1.18, "learning_rate": 1.9045676848043184e-06, "loss": 0.2304, "step": 6238 }, { "epoch": 1.18, "learning_rate": 1.9038248254873157e-06, "loss": 0.2017, "step": 6239 }, { "epoch": 1.18, "learning_rate": 1.9030820219790133e-06, "loss": 0.3101, "step": 6240 }, { "epoch": 1.18, "learning_rate": 1.9023392743489474e-06, "loss": 0.237, "step": 6241 }, { "epoch": 1.18, "learning_rate": 1.9015965826666454e-06, "loss": 0.332, "step": 6242 }, { "epoch": 1.18, "learning_rate": 1.900853947001632e-06, "loss": 0.2674, "step": 6243 }, { "epoch": 1.18, "learning_rate": 1.9001113674234278e-06, "loss": 0.2009, "step": 6244 }, { "epoch": 1.18, "learning_rate": 1.899368844001545e-06, "loss": 0.2372, "step": 6245 }, { "epoch": 1.18, "learning_rate": 1.8986263768054918e-06, "loss": 0.2512, "step": 6246 }, { "epoch": 1.18, "learning_rate": 1.8978839659047735e-06, "loss": 0.2359, "step": 6247 }, { "epoch": 1.18, "learning_rate": 1.897141611368886e-06, "loss": 0.2283, "step": 6248 }, { "epoch": 1.18, "learning_rate": 1.896399313267323e-06, "loss": 0.2982, "step": 6249 }, { "epoch": 1.18, "learning_rate": 1.8956570716695727e-06, "loss": 0.2444, "step": 6250 }, { "epoch": 1.18, "learning_rate": 1.8949148866451154e-06, "loss": 0.2121, "step": 6251 }, { "epoch": 1.18, "learning_rate": 1.8941727582634292e-06, "loss": 0.2684, "step": 6252 }, { "epoch": 1.18, "learning_rate": 1.8934306865939862e-06, "loss": 0.2086, "step": 6253 }, { "epoch": 1.18, "learning_rate": 1.892688671706251e-06, "loss": 0.2446, "step": 6254 }, { "epoch": 1.18, "learning_rate": 1.8919467136696856e-06, "loss": 0.233, "step": 6255 }, { "epoch": 1.18, "learning_rate": 1.8912048125537464e-06, "loss": 0.2271, "step": 6256 }, { "epoch": 1.18, "learning_rate": 1.8904629684278813e-06, "loss": 0.1877, "step": 6257 }, { "epoch": 1.18, "learning_rate": 1.8897211813615373e-06, "loss": 0.2171, "step": 6258 }, { "epoch": 1.18, "learning_rate": 1.8889794514241542e-06, "loss": 0.2389, "step": 6259 }, { "epoch": 1.18, "learning_rate": 1.8882377786851641e-06, "loss": 0.2125, "step": 6260 }, { "epoch": 1.18, "learning_rate": 1.8874961632139975e-06, "loss": 0.2253, "step": 6261 }, { "epoch": 1.18, "learning_rate": 1.8867546050800786e-06, "loss": 0.2192, "step": 6262 }, { "epoch": 1.18, "learning_rate": 1.8860131043528243e-06, "loss": 0.2433, "step": 6263 }, { "epoch": 1.18, "learning_rate": 1.8852716611016473e-06, "loss": 0.29, "step": 6264 }, { "epoch": 1.18, "learning_rate": 1.8845302753959565e-06, "loss": 0.2902, "step": 6265 }, { "epoch": 1.18, "learning_rate": 1.8837889473051526e-06, "loss": 0.2687, "step": 6266 }, { "epoch": 1.18, "learning_rate": 1.883047676898632e-06, "loss": 0.2298, "step": 6267 }, { "epoch": 1.18, "learning_rate": 1.8823064642457877e-06, "loss": 0.3196, "step": 6268 }, { "epoch": 1.18, "learning_rate": 1.8815653094160039e-06, "loss": 0.3944, "step": 6269 }, { "epoch": 1.18, "learning_rate": 1.8808242124786607e-06, "loss": 0.2715, "step": 6270 }, { "epoch": 1.18, "learning_rate": 1.8800831735031353e-06, "loss": 0.1832, "step": 6271 }, { "epoch": 1.18, "learning_rate": 1.879342192558795e-06, "loss": 0.2413, "step": 6272 }, { "epoch": 1.19, "learning_rate": 1.878601269715005e-06, "loss": 0.2442, "step": 6273 }, { "epoch": 1.19, "learning_rate": 1.8778604050411248e-06, "loss": 0.2353, "step": 6274 }, { "epoch": 1.19, "learning_rate": 1.8771195986065052e-06, "loss": 0.2191, "step": 6275 }, { "epoch": 1.19, "learning_rate": 1.8763788504804965e-06, "loss": 0.2287, "step": 6276 }, { "epoch": 1.19, "learning_rate": 1.8756381607324406e-06, "loss": 0.2515, "step": 6277 }, { "epoch": 1.19, "learning_rate": 1.8748975294316726e-06, "loss": 0.2811, "step": 6278 }, { "epoch": 1.19, "learning_rate": 1.874156956647526e-06, "loss": 0.2348, "step": 6279 }, { "epoch": 1.19, "learning_rate": 1.8734164424493265e-06, "loss": 0.2503, "step": 6280 }, { "epoch": 1.19, "learning_rate": 1.8726759869063926e-06, "loss": 0.2419, "step": 6281 }, { "epoch": 1.19, "learning_rate": 1.871935590088041e-06, "loss": 0.26, "step": 6282 }, { "epoch": 1.19, "learning_rate": 1.871195252063582e-06, "loss": 0.2451, "step": 6283 }, { "epoch": 1.19, "learning_rate": 1.8704549729023178e-06, "loss": 0.1903, "step": 6284 }, { "epoch": 1.19, "learning_rate": 1.8697147526735468e-06, "loss": 0.23, "step": 6285 }, { "epoch": 1.19, "learning_rate": 1.868974591446564e-06, "loss": 0.2218, "step": 6286 }, { "epoch": 1.19, "learning_rate": 1.8682344892906545e-06, "loss": 0.1828, "step": 6287 }, { "epoch": 1.19, "learning_rate": 1.8674944462751007e-06, "loss": 0.2127, "step": 6288 }, { "epoch": 1.19, "learning_rate": 1.8667544624691805e-06, "loss": 0.2195, "step": 6289 }, { "epoch": 1.19, "learning_rate": 1.866014537942163e-06, "loss": 0.2103, "step": 6290 }, { "epoch": 1.19, "learning_rate": 1.8652746727633135e-06, "loss": 0.2325, "step": 6291 }, { "epoch": 1.19, "learning_rate": 1.8645348670018932e-06, "loss": 0.2446, "step": 6292 }, { "epoch": 1.19, "learning_rate": 1.8637951207271542e-06, "loss": 0.2832, "step": 6293 }, { "epoch": 1.19, "learning_rate": 1.8630554340083468e-06, "loss": 0.234, "step": 6294 }, { "epoch": 1.19, "learning_rate": 1.8623158069147135e-06, "loss": 0.3383, "step": 6295 }, { "epoch": 1.19, "learning_rate": 1.8615762395154907e-06, "loss": 0.2324, "step": 6296 }, { "epoch": 1.19, "learning_rate": 1.8608367318799118e-06, "loss": 0.3438, "step": 6297 }, { "epoch": 1.19, "learning_rate": 1.8600972840772014e-06, "loss": 0.2006, "step": 6298 }, { "epoch": 1.19, "learning_rate": 1.8593578961765823e-06, "loss": 0.2294, "step": 6299 }, { "epoch": 1.19, "learning_rate": 1.8586185682472677e-06, "loss": 0.2013, "step": 6300 }, { "epoch": 1.19, "learning_rate": 1.857879300358467e-06, "loss": 0.2981, "step": 6301 }, { "epoch": 1.19, "learning_rate": 1.8571400925793855e-06, "loss": 0.2144, "step": 6302 }, { "epoch": 1.19, "learning_rate": 1.8564009449792194e-06, "loss": 0.289, "step": 6303 }, { "epoch": 1.19, "learning_rate": 1.855661857627163e-06, "loss": 0.2362, "step": 6304 }, { "epoch": 1.19, "learning_rate": 1.8549228305924024e-06, "loss": 0.204, "step": 6305 }, { "epoch": 1.19, "learning_rate": 1.8541838639441184e-06, "loss": 0.2124, "step": 6306 }, { "epoch": 1.19, "learning_rate": 1.853444957751487e-06, "loss": 0.2155, "step": 6307 }, { "epoch": 1.19, "learning_rate": 1.8527061120836792e-06, "loss": 0.3111, "step": 6308 }, { "epoch": 1.19, "learning_rate": 1.8519673270098567e-06, "loss": 0.2151, "step": 6309 }, { "epoch": 1.19, "learning_rate": 1.8512286025991805e-06, "loss": 0.2093, "step": 6310 }, { "epoch": 1.19, "learning_rate": 1.850489938920803e-06, "loss": 0.263, "step": 6311 }, { "epoch": 1.19, "learning_rate": 1.84975133604387e-06, "loss": 0.2389, "step": 6312 }, { "epoch": 1.19, "learning_rate": 1.849012794037524e-06, "loss": 0.1886, "step": 6313 }, { "epoch": 1.19, "learning_rate": 1.8482743129709023e-06, "loss": 0.2643, "step": 6314 }, { "epoch": 1.19, "learning_rate": 1.8475358929131331e-06, "loss": 0.3229, "step": 6315 }, { "epoch": 1.19, "learning_rate": 1.8467975339333404e-06, "loss": 0.2585, "step": 6316 }, { "epoch": 1.19, "learning_rate": 1.8460592361006457e-06, "loss": 0.2729, "step": 6317 }, { "epoch": 1.19, "learning_rate": 1.8453209994841591e-06, "loss": 0.2344, "step": 6318 }, { "epoch": 1.19, "learning_rate": 1.8445828241529884e-06, "loss": 0.2573, "step": 6319 }, { "epoch": 1.19, "learning_rate": 1.8438447101762369e-06, "loss": 0.372, "step": 6320 }, { "epoch": 1.19, "learning_rate": 1.8431066576229983e-06, "loss": 0.2746, "step": 6321 }, { "epoch": 1.19, "learning_rate": 1.8423686665623634e-06, "loss": 0.2302, "step": 6322 }, { "epoch": 1.19, "learning_rate": 1.841630737063417e-06, "loss": 0.2523, "step": 6323 }, { "epoch": 1.19, "learning_rate": 1.8408928691952363e-06, "loss": 0.3094, "step": 6324 }, { "epoch": 1.19, "learning_rate": 1.8401550630268953e-06, "loss": 0.2757, "step": 6325 }, { "epoch": 1.2, "learning_rate": 1.8394173186274609e-06, "loss": 0.265, "step": 6326 }, { "epoch": 1.2, "learning_rate": 1.8386796360659929e-06, "loss": 0.2696, "step": 6327 }, { "epoch": 1.2, "learning_rate": 1.837942015411548e-06, "loss": 0.2427, "step": 6328 }, { "epoch": 1.2, "learning_rate": 1.8372044567331759e-06, "loss": 0.2277, "step": 6329 }, { "epoch": 1.2, "learning_rate": 1.836466960099919e-06, "loss": 0.2768, "step": 6330 }, { "epoch": 1.2, "learning_rate": 1.8357295255808165e-06, "loss": 0.2351, "step": 6331 }, { "epoch": 1.2, "learning_rate": 1.8349921532449012e-06, "loss": 0.1965, "step": 6332 }, { "epoch": 1.2, "learning_rate": 1.8342548431611967e-06, "loss": 0.2682, "step": 6333 }, { "epoch": 1.2, "learning_rate": 1.833517595398726e-06, "loss": 0.2249, "step": 6334 }, { "epoch": 1.2, "learning_rate": 1.832780410026504e-06, "loss": 0.198, "step": 6335 }, { "epoch": 1.2, "learning_rate": 1.8320432871135378e-06, "loss": 0.1663, "step": 6336 }, { "epoch": 1.2, "learning_rate": 1.8313062267288306e-06, "loss": 0.2172, "step": 6337 }, { "epoch": 1.2, "learning_rate": 1.8305692289413815e-06, "loss": 0.2095, "step": 6338 }, { "epoch": 1.2, "learning_rate": 1.8298322938201799e-06, "loss": 0.3077, "step": 6339 }, { "epoch": 1.2, "learning_rate": 1.8290954214342111e-06, "loss": 0.2498, "step": 6340 }, { "epoch": 1.2, "learning_rate": 1.8283586118524563e-06, "loss": 0.2693, "step": 6341 }, { "epoch": 1.2, "learning_rate": 1.827621865143887e-06, "loss": 0.2156, "step": 6342 }, { "epoch": 1.2, "learning_rate": 1.826885181377473e-06, "loss": 0.2302, "step": 6343 }, { "epoch": 1.2, "learning_rate": 1.8261485606221755e-06, "loss": 0.188, "step": 6344 }, { "epoch": 1.2, "learning_rate": 1.8254120029469494e-06, "loss": 0.235, "step": 6345 }, { "epoch": 1.2, "learning_rate": 1.8246755084207462e-06, "loss": 0.2432, "step": 6346 }, { "epoch": 1.2, "learning_rate": 1.82393907711251e-06, "loss": 0.2688, "step": 6347 }, { "epoch": 1.2, "learning_rate": 1.8232027090911775e-06, "loss": 0.3085, "step": 6348 }, { "epoch": 1.2, "learning_rate": 1.822466404425683e-06, "loss": 0.2195, "step": 6349 }, { "epoch": 1.2, "learning_rate": 1.821730163184952e-06, "loss": 0.3222, "step": 6350 }, { "epoch": 1.2, "learning_rate": 1.8209939854379045e-06, "loss": 0.2198, "step": 6351 }, { "epoch": 1.2, "learning_rate": 1.8202578712534548e-06, "loss": 0.1965, "step": 6352 }, { "epoch": 1.2, "learning_rate": 1.8195218207005139e-06, "loss": 0.2698, "step": 6353 }, { "epoch": 1.2, "learning_rate": 1.8187858338479816e-06, "loss": 0.301, "step": 6354 }, { "epoch": 1.2, "learning_rate": 1.8180499107647553e-06, "loss": 0.2311, "step": 6355 }, { "epoch": 1.2, "learning_rate": 1.8173140515197271e-06, "loss": 0.2876, "step": 6356 }, { "epoch": 1.2, "learning_rate": 1.8165782561817803e-06, "loss": 0.2394, "step": 6357 }, { "epoch": 1.2, "learning_rate": 1.8158425248197931e-06, "loss": 0.1998, "step": 6358 }, { "epoch": 1.2, "learning_rate": 1.8151068575026397e-06, "loss": 0.2695, "step": 6359 }, { "epoch": 1.2, "learning_rate": 1.8143712542991865e-06, "loss": 0.264, "step": 6360 }, { "epoch": 1.2, "learning_rate": 1.8136357152782926e-06, "loss": 0.3495, "step": 6361 }, { "epoch": 1.2, "learning_rate": 1.8129002405088144e-06, "loss": 0.1918, "step": 6362 }, { "epoch": 1.2, "learning_rate": 1.8121648300596011e-06, "loss": 0.1387, "step": 6363 }, { "epoch": 1.2, "learning_rate": 1.8114294839994945e-06, "loss": 0.2832, "step": 6364 }, { "epoch": 1.2, "learning_rate": 1.81069420239733e-06, "loss": 0.2023, "step": 6365 }, { "epoch": 1.2, "learning_rate": 1.8099589853219414e-06, "loss": 0.2689, "step": 6366 }, { "epoch": 1.2, "learning_rate": 1.8092238328421503e-06, "loss": 0.2161, "step": 6367 }, { "epoch": 1.2, "learning_rate": 1.8084887450267758e-06, "loss": 0.2589, "step": 6368 }, { "epoch": 1.2, "learning_rate": 1.8077537219446323e-06, "loss": 0.3685, "step": 6369 }, { "epoch": 1.2, "learning_rate": 1.807018763664524e-06, "loss": 0.2777, "step": 6370 }, { "epoch": 1.2, "learning_rate": 1.8062838702552516e-06, "loss": 0.215, "step": 6371 }, { "epoch": 1.2, "learning_rate": 1.8055490417856112e-06, "loss": 0.3003, "step": 6372 }, { "epoch": 1.2, "learning_rate": 1.8048142783243882e-06, "loss": 0.1927, "step": 6373 }, { "epoch": 1.2, "learning_rate": 1.8040795799403675e-06, "loss": 0.2909, "step": 6374 }, { "epoch": 1.2, "learning_rate": 1.8033449467023237e-06, "loss": 0.2896, "step": 6375 }, { "epoch": 1.2, "learning_rate": 1.8026103786790261e-06, "loss": 0.2946, "step": 6376 }, { "epoch": 1.2, "learning_rate": 1.8018758759392396e-06, "loss": 0.2964, "step": 6377 }, { "epoch": 1.2, "learning_rate": 1.8011414385517225e-06, "loss": 0.2032, "step": 6378 }, { "epoch": 1.21, "learning_rate": 1.8004070665852243e-06, "loss": 0.2658, "step": 6379 }, { "epoch": 1.21, "learning_rate": 1.7996727601084924e-06, "loss": 0.2043, "step": 6380 }, { "epoch": 1.21, "learning_rate": 1.798938519190266e-06, "loss": 0.2039, "step": 6381 }, { "epoch": 1.21, "learning_rate": 1.7982043438992764e-06, "loss": 0.2014, "step": 6382 }, { "epoch": 1.21, "learning_rate": 1.797470234304252e-06, "loss": 0.202, "step": 6383 }, { "epoch": 1.21, "learning_rate": 1.796736190473915e-06, "loss": 0.2396, "step": 6384 }, { "epoch": 1.21, "learning_rate": 1.7960022124769782e-06, "loss": 0.2158, "step": 6385 }, { "epoch": 1.21, "learning_rate": 1.7952683003821505e-06, "loss": 0.306, "step": 6386 }, { "epoch": 1.21, "learning_rate": 1.7945344542581356e-06, "loss": 0.2561, "step": 6387 }, { "epoch": 1.21, "learning_rate": 1.7938006741736286e-06, "loss": 0.2458, "step": 6388 }, { "epoch": 1.21, "learning_rate": 1.7930669601973191e-06, "loss": 0.3045, "step": 6389 }, { "epoch": 1.21, "learning_rate": 1.7923333123978926e-06, "loss": 0.2521, "step": 6390 }, { "epoch": 1.21, "learning_rate": 1.7915997308440258e-06, "loss": 0.2227, "step": 6391 }, { "epoch": 1.21, "learning_rate": 1.7908662156043894e-06, "loss": 0.2273, "step": 6392 }, { "epoch": 1.21, "learning_rate": 1.7901327667476507e-06, "loss": 0.1744, "step": 6393 }, { "epoch": 1.21, "learning_rate": 1.7893993843424665e-06, "loss": 0.2449, "step": 6394 }, { "epoch": 1.21, "learning_rate": 1.7886660684574915e-06, "loss": 0.2387, "step": 6395 }, { "epoch": 1.21, "learning_rate": 1.7879328191613714e-06, "loss": 0.2193, "step": 6396 }, { "epoch": 1.21, "learning_rate": 1.7871996365227461e-06, "loss": 0.2623, "step": 6397 }, { "epoch": 1.21, "learning_rate": 1.786466520610251e-06, "loss": 0.2368, "step": 6398 }, { "epoch": 1.21, "learning_rate": 1.785733471492514e-06, "loss": 0.2477, "step": 6399 }, { "epoch": 1.21, "learning_rate": 1.785000489238155e-06, "loss": 0.227, "step": 6400 }, { "epoch": 1.21, "learning_rate": 1.784267573915791e-06, "loss": 0.2661, "step": 6401 }, { "epoch": 1.21, "learning_rate": 1.783534725594031e-06, "loss": 0.1451, "step": 6402 }, { "epoch": 1.21, "learning_rate": 1.7828019443414762e-06, "loss": 0.2074, "step": 6403 }, { "epoch": 1.21, "learning_rate": 1.7820692302267251e-06, "loss": 0.2257, "step": 6404 }, { "epoch": 1.21, "learning_rate": 1.7813365833183677e-06, "loss": 0.3073, "step": 6405 }, { "epoch": 1.21, "learning_rate": 1.7806040036849872e-06, "loss": 0.2906, "step": 6406 }, { "epoch": 1.21, "learning_rate": 1.7798714913951615e-06, "loss": 0.3882, "step": 6407 }, { "epoch": 1.21, "learning_rate": 1.779139046517463e-06, "loss": 0.3017, "step": 6408 }, { "epoch": 1.21, "learning_rate": 1.7784066691204552e-06, "loss": 0.2739, "step": 6409 }, { "epoch": 1.21, "learning_rate": 1.7776743592726972e-06, "loss": 0.3138, "step": 6410 }, { "epoch": 1.21, "learning_rate": 1.776942117042743e-06, "loss": 0.2959, "step": 6411 }, { "epoch": 1.21, "learning_rate": 1.776209942499137e-06, "loss": 0.1899, "step": 6412 }, { "epoch": 1.21, "learning_rate": 1.7754778357104188e-06, "loss": 0.2511, "step": 6413 }, { "epoch": 1.21, "learning_rate": 1.7747457967451237e-06, "loss": 0.2126, "step": 6414 }, { "epoch": 1.21, "learning_rate": 1.7740138256717765e-06, "loss": 0.2687, "step": 6415 }, { "epoch": 1.21, "learning_rate": 1.7732819225589e-06, "loss": 0.2912, "step": 6416 }, { "epoch": 1.21, "learning_rate": 1.772550087475008e-06, "loss": 0.1937, "step": 6417 }, { "epoch": 1.21, "learning_rate": 1.7718183204886066e-06, "loss": 0.2486, "step": 6418 }, { "epoch": 1.21, "learning_rate": 1.7710866216681999e-06, "loss": 0.2554, "step": 6419 }, { "epoch": 1.21, "learning_rate": 1.7703549910822824e-06, "loss": 0.2917, "step": 6420 }, { "epoch": 1.21, "learning_rate": 1.7696234287993416e-06, "loss": 0.2676, "step": 6421 }, { "epoch": 1.21, "learning_rate": 1.768891934887862e-06, "loss": 0.2245, "step": 6422 }, { "epoch": 1.21, "learning_rate": 1.7681605094163178e-06, "loss": 0.272, "step": 6423 }, { "epoch": 1.21, "learning_rate": 1.7674291524531805e-06, "loss": 0.2732, "step": 6424 }, { "epoch": 1.21, "learning_rate": 1.7666978640669118e-06, "loss": 0.2266, "step": 6425 }, { "epoch": 1.21, "learning_rate": 1.7659666443259697e-06, "loss": 0.3253, "step": 6426 }, { "epoch": 1.21, "learning_rate": 1.7652354932988042e-06, "loss": 0.2225, "step": 6427 }, { "epoch": 1.21, "learning_rate": 1.7645044110538582e-06, "loss": 0.1984, "step": 6428 }, { "epoch": 1.21, "learning_rate": 1.7637733976595705e-06, "loss": 0.2381, "step": 6429 }, { "epoch": 1.21, "learning_rate": 1.7630424531843724e-06, "loss": 0.2896, "step": 6430 }, { "epoch": 1.22, "learning_rate": 1.7623115776966867e-06, "loss": 0.2435, "step": 6431 }, { "epoch": 1.22, "learning_rate": 1.7615807712649335e-06, "loss": 0.253, "step": 6432 }, { "epoch": 1.22, "learning_rate": 1.760850033957524e-06, "loss": 0.2198, "step": 6433 }, { "epoch": 1.22, "learning_rate": 1.7601193658428623e-06, "loss": 0.2825, "step": 6434 }, { "epoch": 1.22, "learning_rate": 1.7593887669893478e-06, "loss": 0.2751, "step": 6435 }, { "epoch": 1.22, "learning_rate": 1.7586582374653743e-06, "loss": 0.2354, "step": 6436 }, { "epoch": 1.22, "learning_rate": 1.7579277773393258e-06, "loss": 0.2556, "step": 6437 }, { "epoch": 1.22, "learning_rate": 1.7571973866795817e-06, "loss": 0.2113, "step": 6438 }, { "epoch": 1.22, "learning_rate": 1.7564670655545157e-06, "loss": 0.2883, "step": 6439 }, { "epoch": 1.22, "learning_rate": 1.7557368140324931e-06, "loss": 0.2829, "step": 6440 }, { "epoch": 1.22, "learning_rate": 1.7550066321818737e-06, "loss": 0.2916, "step": 6441 }, { "epoch": 1.22, "learning_rate": 1.754276520071012e-06, "loss": 0.2012, "step": 6442 }, { "epoch": 1.22, "learning_rate": 1.7535464777682538e-06, "loss": 0.1706, "step": 6443 }, { "epoch": 1.22, "learning_rate": 1.7528165053419383e-06, "loss": 0.2236, "step": 6444 }, { "epoch": 1.22, "learning_rate": 1.752086602860401e-06, "loss": 0.2797, "step": 6445 }, { "epoch": 1.22, "learning_rate": 1.751356770391967e-06, "loss": 0.2683, "step": 6446 }, { "epoch": 1.22, "learning_rate": 1.7506270080049587e-06, "loss": 0.2693, "step": 6447 }, { "epoch": 1.22, "learning_rate": 1.7498973157676895e-06, "loss": 0.2645, "step": 6448 }, { "epoch": 1.22, "learning_rate": 1.7491676937484658e-06, "loss": 0.2318, "step": 6449 }, { "epoch": 1.22, "learning_rate": 1.74843814201559e-06, "loss": 0.2295, "step": 6450 }, { "epoch": 1.22, "learning_rate": 1.7477086606373555e-06, "loss": 0.1325, "step": 6451 }, { "epoch": 1.22, "learning_rate": 1.7469792496820493e-06, "loss": 0.2069, "step": 6452 }, { "epoch": 1.22, "learning_rate": 1.7462499092179536e-06, "loss": 0.3339, "step": 6453 }, { "epoch": 1.22, "learning_rate": 1.7455206393133428e-06, "loss": 0.2456, "step": 6454 }, { "epoch": 1.22, "learning_rate": 1.7447914400364835e-06, "loss": 0.2416, "step": 6455 }, { "epoch": 1.22, "learning_rate": 1.744062311455638e-06, "loss": 0.2816, "step": 6456 }, { "epoch": 1.22, "learning_rate": 1.7433332536390617e-06, "loss": 0.2403, "step": 6457 }, { "epoch": 1.22, "learning_rate": 1.7426042666550011e-06, "loss": 0.2584, "step": 6458 }, { "epoch": 1.22, "learning_rate": 1.7418753505716978e-06, "loss": 0.1754, "step": 6459 }, { "epoch": 1.22, "learning_rate": 1.7411465054573878e-06, "loss": 0.255, "step": 6460 }, { "epoch": 1.22, "learning_rate": 1.7404177313802978e-06, "loss": 0.2085, "step": 6461 }, { "epoch": 1.22, "learning_rate": 1.7396890284086492e-06, "loss": 0.2054, "step": 6462 }, { "epoch": 1.22, "learning_rate": 1.7389603966106586e-06, "loss": 0.3109, "step": 6463 }, { "epoch": 1.22, "learning_rate": 1.7382318360545324e-06, "loss": 0.2114, "step": 6464 }, { "epoch": 1.22, "learning_rate": 1.7375033468084717e-06, "loss": 0.3355, "step": 6465 }, { "epoch": 1.22, "learning_rate": 1.7367749289406734e-06, "loss": 0.2344, "step": 6466 }, { "epoch": 1.22, "learning_rate": 1.7360465825193235e-06, "loss": 0.2581, "step": 6467 }, { "epoch": 1.22, "learning_rate": 1.7353183076126045e-06, "loss": 0.193, "step": 6468 }, { "epoch": 1.22, "learning_rate": 1.7345901042886915e-06, "loss": 0.2314, "step": 6469 }, { "epoch": 1.22, "learning_rate": 1.7338619726157508e-06, "loss": 0.2153, "step": 6470 }, { "epoch": 1.22, "learning_rate": 1.733133912661945e-06, "loss": 0.2496, "step": 6471 }, { "epoch": 1.22, "learning_rate": 1.7324059244954294e-06, "loss": 0.2039, "step": 6472 }, { "epoch": 1.22, "learning_rate": 1.73167800818435e-06, "loss": 0.2322, "step": 6473 }, { "epoch": 1.22, "learning_rate": 1.7309501637968496e-06, "loss": 0.3047, "step": 6474 }, { "epoch": 1.22, "learning_rate": 1.730222391401062e-06, "loss": 0.2447, "step": 6475 }, { "epoch": 1.22, "learning_rate": 1.7294946910651142e-06, "loss": 0.2655, "step": 6476 }, { "epoch": 1.22, "learning_rate": 1.7287670628571273e-06, "loss": 0.234, "step": 6477 }, { "epoch": 1.22, "learning_rate": 1.7280395068452177e-06, "loss": 0.2485, "step": 6478 }, { "epoch": 1.22, "learning_rate": 1.72731202309749e-06, "loss": 0.2571, "step": 6479 }, { "epoch": 1.22, "learning_rate": 1.7265846116820456e-06, "loss": 0.2002, "step": 6480 }, { "epoch": 1.22, "learning_rate": 1.7258572726669799e-06, "loss": 0.2234, "step": 6481 }, { "epoch": 1.22, "learning_rate": 1.725130006120378e-06, "loss": 0.2056, "step": 6482 }, { "epoch": 1.22, "learning_rate": 1.7244028121103207e-06, "loss": 0.2835, "step": 6483 }, { "epoch": 1.23, "learning_rate": 1.7236756907048828e-06, "loss": 0.2701, "step": 6484 }, { "epoch": 1.23, "learning_rate": 1.7229486419721298e-06, "loss": 0.2761, "step": 6485 }, { "epoch": 1.23, "learning_rate": 1.7222216659801217e-06, "loss": 0.3057, "step": 6486 }, { "epoch": 1.23, "learning_rate": 1.7214947627969115e-06, "loss": 0.3237, "step": 6487 }, { "epoch": 1.23, "learning_rate": 1.720767932490547e-06, "loss": 0.2333, "step": 6488 }, { "epoch": 1.23, "learning_rate": 1.7200411751290662e-06, "loss": 0.2852, "step": 6489 }, { "epoch": 1.23, "learning_rate": 1.7193144907805015e-06, "loss": 0.3075, "step": 6490 }, { "epoch": 1.23, "learning_rate": 1.7185878795128807e-06, "loss": 0.2198, "step": 6491 }, { "epoch": 1.23, "learning_rate": 1.7178613413942207e-06, "loss": 0.2072, "step": 6492 }, { "epoch": 1.23, "learning_rate": 1.7171348764925341e-06, "loss": 0.2906, "step": 6493 }, { "epoch": 1.23, "learning_rate": 1.7164084848758275e-06, "loss": 0.2997, "step": 6494 }, { "epoch": 1.23, "learning_rate": 1.7156821666120974e-06, "loss": 0.2706, "step": 6495 }, { "epoch": 1.23, "learning_rate": 1.7149559217693367e-06, "loss": 0.2612, "step": 6496 }, { "epoch": 1.23, "learning_rate": 1.71422975041553e-06, "loss": 0.2102, "step": 6497 }, { "epoch": 1.23, "learning_rate": 1.713503652618654e-06, "loss": 0.2553, "step": 6498 }, { "epoch": 1.23, "learning_rate": 1.7127776284466808e-06, "loss": 0.3097, "step": 6499 }, { "epoch": 1.23, "learning_rate": 1.7120516779675744e-06, "loss": 0.235, "step": 6500 }, { "epoch": 1.23, "learning_rate": 1.7113258012492906e-06, "loss": 0.2101, "step": 6501 }, { "epoch": 1.23, "learning_rate": 1.7105999983597812e-06, "loss": 0.1829, "step": 6502 }, { "epoch": 1.23, "learning_rate": 1.7098742693669896e-06, "loss": 0.3318, "step": 6503 }, { "epoch": 1.23, "learning_rate": 1.70914861433885e-06, "loss": 0.2363, "step": 6504 }, { "epoch": 1.23, "learning_rate": 1.7084230333432934e-06, "loss": 0.2772, "step": 6505 }, { "epoch": 1.23, "learning_rate": 1.7076975264482434e-06, "loss": 0.2021, "step": 6506 }, { "epoch": 1.23, "learning_rate": 1.706972093721614e-06, "loss": 0.2652, "step": 6507 }, { "epoch": 1.23, "learning_rate": 1.706246735231314e-06, "loss": 0.1993, "step": 6508 }, { "epoch": 1.23, "learning_rate": 1.7055214510452462e-06, "loss": 0.1862, "step": 6509 }, { "epoch": 1.23, "learning_rate": 1.7047962412313044e-06, "loss": 0.2451, "step": 6510 }, { "epoch": 1.23, "learning_rate": 1.704071105857376e-06, "loss": 0.2594, "step": 6511 }, { "epoch": 1.23, "learning_rate": 1.7033460449913432e-06, "loss": 0.2627, "step": 6512 }, { "epoch": 1.23, "learning_rate": 1.7026210587010787e-06, "loss": 0.2574, "step": 6513 }, { "epoch": 1.23, "learning_rate": 1.7018961470544495e-06, "loss": 0.2195, "step": 6514 }, { "epoch": 1.23, "learning_rate": 1.7011713101193167e-06, "loss": 0.2302, "step": 6515 }, { "epoch": 1.23, "learning_rate": 1.7004465479635313e-06, "loss": 0.2565, "step": 6516 }, { "epoch": 1.23, "learning_rate": 1.6997218606549404e-06, "loss": 0.2061, "step": 6517 }, { "epoch": 1.23, "learning_rate": 1.698997248261383e-06, "loss": 0.1934, "step": 6518 }, { "epoch": 1.23, "learning_rate": 1.6982727108506896e-06, "loss": 0.2493, "step": 6519 }, { "epoch": 1.23, "learning_rate": 1.697548248490687e-06, "loss": 0.2047, "step": 6520 }, { "epoch": 1.23, "learning_rate": 1.6968238612491919e-06, "loss": 0.2449, "step": 6521 }, { "epoch": 1.23, "learning_rate": 1.6960995491940147e-06, "loss": 0.2446, "step": 6522 }, { "epoch": 1.23, "learning_rate": 1.6953753123929597e-06, "loss": 0.262, "step": 6523 }, { "epoch": 1.23, "learning_rate": 1.6946511509138243e-06, "loss": 0.2, "step": 6524 }, { "epoch": 1.23, "learning_rate": 1.6939270648243965e-06, "loss": 0.2141, "step": 6525 }, { "epoch": 1.23, "learning_rate": 1.6932030541924599e-06, "loss": 0.2173, "step": 6526 }, { "epoch": 1.23, "learning_rate": 1.6924791190857912e-06, "loss": 0.2339, "step": 6527 }, { "epoch": 1.23, "learning_rate": 1.691755259572157e-06, "loss": 0.2695, "step": 6528 }, { "epoch": 1.23, "learning_rate": 1.6910314757193185e-06, "loss": 0.2966, "step": 6529 }, { "epoch": 1.23, "learning_rate": 1.6903077675950322e-06, "loss": 0.2383, "step": 6530 }, { "epoch": 1.23, "learning_rate": 1.6895841352670433e-06, "loss": 0.2101, "step": 6531 }, { "epoch": 1.23, "learning_rate": 1.688860578803092e-06, "loss": 0.2299, "step": 6532 }, { "epoch": 1.23, "learning_rate": 1.6881370982709134e-06, "loss": 0.1764, "step": 6533 }, { "epoch": 1.23, "learning_rate": 1.6874136937382313e-06, "loss": 0.2566, "step": 6534 }, { "epoch": 1.23, "learning_rate": 1.6866903652727643e-06, "loss": 0.2949, "step": 6535 }, { "epoch": 1.23, "learning_rate": 1.685967112942226e-06, "loss": 0.2461, "step": 6536 }, { "epoch": 1.24, "learning_rate": 1.685243936814319e-06, "loss": 0.2311, "step": 6537 }, { "epoch": 1.24, "learning_rate": 1.684520836956742e-06, "loss": 0.2073, "step": 6538 }, { "epoch": 1.24, "learning_rate": 1.6837978134371851e-06, "loss": 0.2269, "step": 6539 }, { "epoch": 1.24, "learning_rate": 1.6830748663233306e-06, "loss": 0.271, "step": 6540 }, { "epoch": 1.24, "learning_rate": 1.6823519956828555e-06, "loss": 0.2313, "step": 6541 }, { "epoch": 1.24, "learning_rate": 1.6816292015834285e-06, "loss": 0.2787, "step": 6542 }, { "epoch": 1.24, "learning_rate": 1.68090648409271e-06, "loss": 0.212, "step": 6543 }, { "epoch": 1.24, "learning_rate": 1.680183843278356e-06, "loss": 0.2592, "step": 6544 }, { "epoch": 1.24, "learning_rate": 1.679461279208014e-06, "loss": 0.3088, "step": 6545 }, { "epoch": 1.24, "learning_rate": 1.678738791949322e-06, "loss": 0.327, "step": 6546 }, { "epoch": 1.24, "learning_rate": 1.6780163815699141e-06, "loss": 0.2034, "step": 6547 }, { "epoch": 1.24, "learning_rate": 1.677294048137417e-06, "loss": 0.2783, "step": 6548 }, { "epoch": 1.24, "learning_rate": 1.676571791719449e-06, "loss": 0.2301, "step": 6549 }, { "epoch": 1.24, "learning_rate": 1.6758496123836196e-06, "loss": 0.3477, "step": 6550 }, { "epoch": 1.24, "learning_rate": 1.675127510197535e-06, "loss": 0.3197, "step": 6551 }, { "epoch": 1.24, "learning_rate": 1.6744054852287918e-06, "loss": 0.2317, "step": 6552 }, { "epoch": 1.24, "learning_rate": 1.6736835375449779e-06, "loss": 0.2339, "step": 6553 }, { "epoch": 1.24, "learning_rate": 1.6729616672136773e-06, "loss": 0.231, "step": 6554 }, { "epoch": 1.24, "learning_rate": 1.6722398743024656e-06, "loss": 0.341, "step": 6555 }, { "epoch": 1.24, "learning_rate": 1.671518158878909e-06, "loss": 0.2685, "step": 6556 }, { "epoch": 1.24, "learning_rate": 1.670796521010569e-06, "loss": 0.1851, "step": 6557 }, { "epoch": 1.24, "learning_rate": 1.6700749607650003e-06, "loss": 0.3448, "step": 6558 }, { "epoch": 1.24, "learning_rate": 1.669353478209747e-06, "loss": 0.2443, "step": 6559 }, { "epoch": 1.24, "learning_rate": 1.668632073412349e-06, "loss": 0.2316, "step": 6560 }, { "epoch": 1.24, "learning_rate": 1.6679107464403383e-06, "loss": 0.3115, "step": 6561 }, { "epoch": 1.24, "learning_rate": 1.6671894973612384e-06, "loss": 0.2556, "step": 6562 }, { "epoch": 1.24, "learning_rate": 1.6664683262425662e-06, "loss": 0.2329, "step": 6563 }, { "epoch": 1.24, "learning_rate": 1.6657472331518327e-06, "loss": 0.2136, "step": 6564 }, { "epoch": 1.24, "learning_rate": 1.665026218156539e-06, "loss": 0.1564, "step": 6565 }, { "epoch": 1.24, "learning_rate": 1.6643052813241806e-06, "loss": 0.2415, "step": 6566 }, { "epoch": 1.24, "learning_rate": 1.6635844227222458e-06, "loss": 0.2459, "step": 6567 }, { "epoch": 1.24, "learning_rate": 1.662863642418214e-06, "loss": 0.2276, "step": 6568 }, { "epoch": 1.24, "learning_rate": 1.6621429404795594e-06, "loss": 0.245, "step": 6569 }, { "epoch": 1.24, "learning_rate": 1.6614223169737482e-06, "loss": 0.2775, "step": 6570 }, { "epoch": 1.24, "learning_rate": 1.6607017719682366e-06, "loss": 0.1734, "step": 6571 }, { "epoch": 1.24, "learning_rate": 1.6599813055304777e-06, "loss": 0.295, "step": 6572 }, { "epoch": 1.24, "learning_rate": 1.6592609177279157e-06, "loss": 0.1914, "step": 6573 }, { "epoch": 1.24, "learning_rate": 1.6585406086279847e-06, "loss": 0.2517, "step": 6574 }, { "epoch": 1.24, "learning_rate": 1.6578203782981156e-06, "loss": 0.2945, "step": 6575 }, { "epoch": 1.24, "learning_rate": 1.65710022680573e-06, "loss": 0.2188, "step": 6576 }, { "epoch": 1.24, "learning_rate": 1.6563801542182412e-06, "loss": 0.3489, "step": 6577 }, { "epoch": 1.24, "learning_rate": 1.655660160603056e-06, "loss": 0.244, "step": 6578 }, { "epoch": 1.24, "learning_rate": 1.6549402460275757e-06, "loss": 0.178, "step": 6579 }, { "epoch": 1.24, "learning_rate": 1.6542204105591907e-06, "loss": 0.2989, "step": 6580 }, { "epoch": 1.24, "learning_rate": 1.6535006542652856e-06, "loss": 0.2239, "step": 6581 }, { "epoch": 1.24, "learning_rate": 1.6527809772132397e-06, "loss": 0.181, "step": 6582 }, { "epoch": 1.24, "learning_rate": 1.6520613794704205e-06, "loss": 0.2505, "step": 6583 }, { "epoch": 1.24, "learning_rate": 1.6513418611041909e-06, "loss": 0.1875, "step": 6584 }, { "epoch": 1.24, "learning_rate": 1.6506224221819079e-06, "loss": 0.1998, "step": 6585 }, { "epoch": 1.24, "learning_rate": 1.6499030627709164e-06, "loss": 0.2451, "step": 6586 }, { "epoch": 1.24, "learning_rate": 1.6491837829385577e-06, "loss": 0.2433, "step": 6587 }, { "epoch": 1.24, "learning_rate": 1.648464582752165e-06, "loss": 0.2644, "step": 6588 }, { "epoch": 1.24, "learning_rate": 1.6477454622790623e-06, "loss": 0.2595, "step": 6589 }, { "epoch": 1.25, "learning_rate": 1.647026421586569e-06, "loss": 0.1833, "step": 6590 }, { "epoch": 1.25, "learning_rate": 1.6463074607419943e-06, "loss": 0.245, "step": 6591 }, { "epoch": 1.25, "learning_rate": 1.6455885798126404e-06, "loss": 0.269, "step": 6592 }, { "epoch": 1.25, "learning_rate": 1.6448697788658038e-06, "loss": 0.333, "step": 6593 }, { "epoch": 1.25, "learning_rate": 1.6441510579687724e-06, "loss": 0.2503, "step": 6594 }, { "epoch": 1.25, "learning_rate": 1.6434324171888254e-06, "loss": 0.3174, "step": 6595 }, { "epoch": 1.25, "learning_rate": 1.6427138565932371e-06, "loss": 0.2129, "step": 6596 }, { "epoch": 1.25, "learning_rate": 1.6419953762492723e-06, "loss": 0.2513, "step": 6597 }, { "epoch": 1.25, "learning_rate": 1.6412769762241876e-06, "loss": 0.2352, "step": 6598 }, { "epoch": 1.25, "learning_rate": 1.6405586565852347e-06, "loss": 0.2281, "step": 6599 }, { "epoch": 1.25, "learning_rate": 1.6398404173996567e-06, "loss": 0.3189, "step": 6600 }, { "epoch": 1.25, "learning_rate": 1.6391222587346884e-06, "loss": 0.2824, "step": 6601 }, { "epoch": 1.25, "learning_rate": 1.6384041806575563e-06, "loss": 0.2859, "step": 6602 }, { "epoch": 1.25, "learning_rate": 1.637686183235483e-06, "loss": 0.2186, "step": 6603 }, { "epoch": 1.25, "learning_rate": 1.636968266535679e-06, "loss": 0.21, "step": 6604 }, { "epoch": 1.25, "learning_rate": 1.6362504306253501e-06, "loss": 0.2444, "step": 6605 }, { "epoch": 1.25, "learning_rate": 1.6355326755716944e-06, "loss": 0.2247, "step": 6606 }, { "epoch": 1.25, "learning_rate": 1.6348150014419012e-06, "loss": 0.2493, "step": 6607 }, { "epoch": 1.25, "learning_rate": 1.6340974083031525e-06, "loss": 0.3296, "step": 6608 }, { "epoch": 1.25, "learning_rate": 1.6333798962226244e-06, "loss": 0.1885, "step": 6609 }, { "epoch": 1.25, "learning_rate": 1.6326624652674824e-06, "loss": 0.2503, "step": 6610 }, { "epoch": 1.25, "learning_rate": 1.6319451155048876e-06, "loss": 0.2993, "step": 6611 }, { "epoch": 1.25, "learning_rate": 1.6312278470019908e-06, "loss": 0.3168, "step": 6612 }, { "epoch": 1.25, "learning_rate": 1.630510659825938e-06, "loss": 0.2243, "step": 6613 }, { "epoch": 1.25, "learning_rate": 1.6297935540438644e-06, "loss": 0.2026, "step": 6614 }, { "epoch": 1.25, "learning_rate": 1.6290765297228995e-06, "loss": 0.2333, "step": 6615 }, { "epoch": 1.25, "learning_rate": 1.6283595869301661e-06, "loss": 0.1763, "step": 6616 }, { "epoch": 1.25, "learning_rate": 1.6276427257327762e-06, "loss": 0.2368, "step": 6617 }, { "epoch": 1.25, "learning_rate": 1.6269259461978365e-06, "loss": 0.2344, "step": 6618 }, { "epoch": 1.25, "learning_rate": 1.6262092483924474e-06, "loss": 0.2227, "step": 6619 }, { "epoch": 1.25, "learning_rate": 1.6254926323836976e-06, "loss": 0.2636, "step": 6620 }, { "epoch": 1.25, "learning_rate": 1.6247760982386719e-06, "loss": 0.2869, "step": 6621 }, { "epoch": 1.25, "learning_rate": 1.6240596460244456e-06, "loss": 0.2052, "step": 6622 }, { "epoch": 1.25, "learning_rate": 1.6233432758080859e-06, "loss": 0.2358, "step": 6623 }, { "epoch": 1.25, "learning_rate": 1.622626987656654e-06, "loss": 0.2422, "step": 6624 }, { "epoch": 1.25, "learning_rate": 1.6219107816372027e-06, "loss": 0.2276, "step": 6625 }, { "epoch": 1.25, "learning_rate": 1.6211946578167753e-06, "loss": 0.3247, "step": 6626 }, { "epoch": 1.25, "learning_rate": 1.6204786162624111e-06, "loss": 0.1376, "step": 6627 }, { "epoch": 1.25, "learning_rate": 1.6197626570411392e-06, "loss": 0.2053, "step": 6628 }, { "epoch": 1.25, "learning_rate": 1.61904678021998e-06, "loss": 0.2457, "step": 6629 }, { "epoch": 1.25, "learning_rate": 1.6183309858659488e-06, "loss": 0.2621, "step": 6630 }, { "epoch": 1.25, "learning_rate": 1.6176152740460527e-06, "loss": 0.2363, "step": 6631 }, { "epoch": 1.25, "learning_rate": 1.616899644827289e-06, "loss": 0.2396, "step": 6632 }, { "epoch": 1.25, "learning_rate": 1.6161840982766486e-06, "loss": 0.22, "step": 6633 }, { "epoch": 1.25, "learning_rate": 1.6154686344611162e-06, "loss": 0.2391, "step": 6634 }, { "epoch": 1.25, "learning_rate": 1.614753253447666e-06, "loss": 0.2653, "step": 6635 }, { "epoch": 1.25, "learning_rate": 1.6140379553032654e-06, "loss": 0.2768, "step": 6636 }, { "epoch": 1.25, "learning_rate": 1.613322740094876e-06, "loss": 0.2669, "step": 6637 }, { "epoch": 1.25, "learning_rate": 1.6126076078894481e-06, "loss": 0.2838, "step": 6638 }, { "epoch": 1.25, "learning_rate": 1.611892558753927e-06, "loss": 0.235, "step": 6639 }, { "epoch": 1.25, "learning_rate": 1.61117759275525e-06, "loss": 0.1961, "step": 6640 }, { "epoch": 1.25, "learning_rate": 1.6104627099603442e-06, "loss": 0.2156, "step": 6641 }, { "epoch": 1.25, "learning_rate": 1.6097479104361328e-06, "loss": 0.2367, "step": 6642 }, { "epoch": 1.26, "learning_rate": 1.609033194249528e-06, "loss": 0.2946, "step": 6643 }, { "epoch": 1.26, "learning_rate": 1.6083185614674345e-06, "loss": 0.2815, "step": 6644 }, { "epoch": 1.26, "learning_rate": 1.607604012156751e-06, "loss": 0.2248, "step": 6645 }, { "epoch": 1.26, "learning_rate": 1.6068895463843679e-06, "loss": 0.2275, "step": 6646 }, { "epoch": 1.26, "learning_rate": 1.6061751642171651e-06, "loss": 0.2295, "step": 6647 }, { "epoch": 1.26, "learning_rate": 1.6054608657220189e-06, "loss": 0.2931, "step": 6648 }, { "epoch": 1.26, "learning_rate": 1.6047466509657955e-06, "loss": 0.2579, "step": 6649 }, { "epoch": 1.26, "learning_rate": 1.6040325200153514e-06, "loss": 0.2364, "step": 6650 }, { "epoch": 1.26, "learning_rate": 1.603318472937539e-06, "loss": 0.2543, "step": 6651 }, { "epoch": 1.26, "learning_rate": 1.602604509799202e-06, "loss": 0.2958, "step": 6652 }, { "epoch": 1.26, "learning_rate": 1.6018906306671733e-06, "loss": 0.2612, "step": 6653 }, { "epoch": 1.26, "learning_rate": 1.6011768356082808e-06, "loss": 0.2633, "step": 6654 }, { "epoch": 1.26, "learning_rate": 1.6004631246893449e-06, "loss": 0.2378, "step": 6655 }, { "epoch": 1.26, "learning_rate": 1.5997494979771755e-06, "loss": 0.2615, "step": 6656 }, { "epoch": 1.26, "learning_rate": 1.5990359555385759e-06, "loss": 0.2081, "step": 6657 }, { "epoch": 1.26, "learning_rate": 1.5983224974403432e-06, "loss": 0.2449, "step": 6658 }, { "epoch": 1.26, "learning_rate": 1.5976091237492637e-06, "loss": 0.2979, "step": 6659 }, { "epoch": 1.26, "learning_rate": 1.5968958345321178e-06, "loss": 0.2333, "step": 6660 }, { "epoch": 1.26, "learning_rate": 1.5961826298556782e-06, "loss": 0.2263, "step": 6661 }, { "epoch": 1.26, "learning_rate": 1.5954695097867069e-06, "loss": 0.2271, "step": 6662 }, { "epoch": 1.26, "learning_rate": 1.5947564743919614e-06, "loss": 0.2191, "step": 6663 }, { "epoch": 1.26, "learning_rate": 1.5940435237381902e-06, "loss": 0.3137, "step": 6664 }, { "epoch": 1.26, "learning_rate": 1.5933306578921315e-06, "loss": 0.2927, "step": 6665 }, { "epoch": 1.26, "learning_rate": 1.5926178769205197e-06, "loss": 0.2135, "step": 6666 }, { "epoch": 1.26, "learning_rate": 1.5919051808900784e-06, "loss": 0.2474, "step": 6667 }, { "epoch": 1.26, "learning_rate": 1.5911925698675233e-06, "loss": 0.2665, "step": 6668 }, { "epoch": 1.26, "learning_rate": 1.590480043919563e-06, "loss": 0.2461, "step": 6669 }, { "epoch": 1.26, "learning_rate": 1.5897676031128994e-06, "loss": 0.2121, "step": 6670 }, { "epoch": 1.26, "learning_rate": 1.5890552475142235e-06, "loss": 0.253, "step": 6671 }, { "epoch": 1.26, "learning_rate": 1.5883429771902195e-06, "loss": 0.2408, "step": 6672 }, { "epoch": 1.26, "learning_rate": 1.5876307922075652e-06, "loss": 0.2665, "step": 6673 }, { "epoch": 1.26, "learning_rate": 1.586918692632929e-06, "loss": 0.1869, "step": 6674 }, { "epoch": 1.26, "learning_rate": 1.5862066785329695e-06, "loss": 0.2734, "step": 6675 }, { "epoch": 1.26, "learning_rate": 1.5854947499743414e-06, "loss": 0.2386, "step": 6676 }, { "epoch": 1.26, "learning_rate": 1.5847829070236892e-06, "loss": 0.2176, "step": 6677 }, { "epoch": 1.26, "learning_rate": 1.5840711497476475e-06, "loss": 0.2799, "step": 6678 }, { "epoch": 1.26, "learning_rate": 1.583359478212846e-06, "loss": 0.2187, "step": 6679 }, { "epoch": 1.26, "learning_rate": 1.5826478924859057e-06, "loss": 0.1914, "step": 6680 }, { "epoch": 1.26, "learning_rate": 1.5819363926334383e-06, "loss": 0.2832, "step": 6681 }, { "epoch": 1.26, "learning_rate": 1.581224978722048e-06, "loss": 0.2616, "step": 6682 }, { "epoch": 1.26, "learning_rate": 1.5805136508183322e-06, "loss": 0.2077, "step": 6683 }, { "epoch": 1.26, "learning_rate": 1.579802408988878e-06, "loss": 0.1864, "step": 6684 }, { "epoch": 1.26, "learning_rate": 1.5790912533002655e-06, "loss": 0.1945, "step": 6685 }, { "epoch": 1.26, "learning_rate": 1.5783801838190686e-06, "loss": 0.2567, "step": 6686 }, { "epoch": 1.26, "learning_rate": 1.5776692006118496e-06, "loss": 0.2234, "step": 6687 }, { "epoch": 1.26, "learning_rate": 1.5769583037451647e-06, "loss": 0.3309, "step": 6688 }, { "epoch": 1.26, "learning_rate": 1.5762474932855635e-06, "loss": 0.2497, "step": 6689 }, { "epoch": 1.26, "learning_rate": 1.5755367692995836e-06, "loss": 0.2755, "step": 6690 }, { "epoch": 1.26, "learning_rate": 1.5748261318537584e-06, "loss": 0.1827, "step": 6691 }, { "epoch": 1.26, "learning_rate": 1.5741155810146113e-06, "loss": 0.2403, "step": 6692 }, { "epoch": 1.26, "learning_rate": 1.5734051168486565e-06, "loss": 0.2509, "step": 6693 }, { "epoch": 1.26, "learning_rate": 1.5726947394224031e-06, "loss": 0.2612, "step": 6694 }, { "epoch": 1.26, "learning_rate": 1.5719844488023503e-06, "loss": 0.2977, "step": 6695 }, { "epoch": 1.27, "learning_rate": 1.5712742450549879e-06, "loss": 0.2059, "step": 6696 }, { "epoch": 1.27, "learning_rate": 1.5705641282468001e-06, "loss": 0.2291, "step": 6697 }, { "epoch": 1.27, "learning_rate": 1.5698540984442624e-06, "loss": 0.2082, "step": 6698 }, { "epoch": 1.27, "learning_rate": 1.5691441557138393e-06, "loss": 0.2426, "step": 6699 }, { "epoch": 1.27, "learning_rate": 1.568434300121991e-06, "loss": 0.297, "step": 6700 }, { "epoch": 1.27, "learning_rate": 1.5677245317351687e-06, "loss": 0.2573, "step": 6701 }, { "epoch": 1.27, "learning_rate": 1.5670148506198135e-06, "loss": 0.2447, "step": 6702 }, { "epoch": 1.27, "learning_rate": 1.5663052568423592e-06, "loss": 0.3216, "step": 6703 }, { "epoch": 1.27, "learning_rate": 1.5655957504692335e-06, "loss": 0.2501, "step": 6704 }, { "epoch": 1.27, "learning_rate": 1.5648863315668528e-06, "loss": 0.2278, "step": 6705 }, { "epoch": 1.27, "learning_rate": 1.564177000201626e-06, "loss": 0.1636, "step": 6706 }, { "epoch": 1.27, "learning_rate": 1.5634677564399566e-06, "loss": 0.2515, "step": 6707 }, { "epoch": 1.27, "learning_rate": 1.5627586003482365e-06, "loss": 0.2457, "step": 6708 }, { "epoch": 1.27, "learning_rate": 1.56204953199285e-06, "loss": 0.237, "step": 6709 }, { "epoch": 1.27, "learning_rate": 1.5613405514401757e-06, "loss": 0.2822, "step": 6710 }, { "epoch": 1.27, "learning_rate": 1.5606316587565804e-06, "loss": 0.2482, "step": 6711 }, { "epoch": 1.27, "learning_rate": 1.5599228540084254e-06, "loss": 0.2621, "step": 6712 }, { "epoch": 1.27, "learning_rate": 1.5592141372620635e-06, "loss": 0.2555, "step": 6713 }, { "epoch": 1.27, "learning_rate": 1.5585055085838364e-06, "loss": 0.2362, "step": 6714 }, { "epoch": 1.27, "learning_rate": 1.5577969680400811e-06, "loss": 0.2467, "step": 6715 }, { "epoch": 1.27, "learning_rate": 1.5570885156971257e-06, "loss": 0.2411, "step": 6716 }, { "epoch": 1.27, "learning_rate": 1.5563801516212874e-06, "loss": 0.1824, "step": 6717 }, { "epoch": 1.27, "learning_rate": 1.5556718758788786e-06, "loss": 0.2935, "step": 6718 }, { "epoch": 1.27, "learning_rate": 1.5549636885362019e-06, "loss": 0.3098, "step": 6719 }, { "epoch": 1.27, "learning_rate": 1.5542555896595497e-06, "loss": 0.2581, "step": 6720 }, { "epoch": 1.27, "learning_rate": 1.5535475793152095e-06, "loss": 0.2653, "step": 6721 }, { "epoch": 1.27, "learning_rate": 1.5528396575694598e-06, "loss": 0.3115, "step": 6722 }, { "epoch": 1.27, "learning_rate": 1.5521318244885692e-06, "loss": 0.3577, "step": 6723 }, { "epoch": 1.27, "learning_rate": 1.5514240801387976e-06, "loss": 0.1977, "step": 6724 }, { "epoch": 1.27, "learning_rate": 1.5507164245864004e-06, "loss": 0.2633, "step": 6725 }, { "epoch": 1.27, "learning_rate": 1.5500088578976197e-06, "loss": 0.2375, "step": 6726 }, { "epoch": 1.27, "learning_rate": 1.5493013801386924e-06, "loss": 0.2539, "step": 6727 }, { "epoch": 1.27, "learning_rate": 1.548593991375848e-06, "loss": 0.2984, "step": 6728 }, { "epoch": 1.27, "learning_rate": 1.5478866916753037e-06, "loss": 0.2423, "step": 6729 }, { "epoch": 1.27, "learning_rate": 1.547179481103272e-06, "loss": 0.2745, "step": 6730 }, { "epoch": 1.27, "learning_rate": 1.5464723597259557e-06, "loss": 0.3382, "step": 6731 }, { "epoch": 1.27, "learning_rate": 1.5457653276095486e-06, "loss": 0.1992, "step": 6732 }, { "epoch": 1.27, "learning_rate": 1.5450583848202378e-06, "loss": 0.1922, "step": 6733 }, { "epoch": 1.27, "learning_rate": 1.544351531424201e-06, "loss": 0.2122, "step": 6734 }, { "epoch": 1.27, "learning_rate": 1.5436447674876062e-06, "loss": 0.2661, "step": 6735 }, { "epoch": 1.27, "learning_rate": 1.5429380930766161e-06, "loss": 0.2182, "step": 6736 }, { "epoch": 1.27, "learning_rate": 1.5422315082573824e-06, "loss": 0.2616, "step": 6737 }, { "epoch": 1.27, "learning_rate": 1.5415250130960508e-06, "loss": 0.2266, "step": 6738 }, { "epoch": 1.27, "learning_rate": 1.5408186076587555e-06, "loss": 0.2717, "step": 6739 }, { "epoch": 1.27, "learning_rate": 1.540112292011624e-06, "loss": 0.251, "step": 6740 }, { "epoch": 1.27, "learning_rate": 1.5394060662207777e-06, "loss": 0.1967, "step": 6741 }, { "epoch": 1.27, "learning_rate": 1.5386999303523241e-06, "loss": 0.243, "step": 6742 }, { "epoch": 1.27, "learning_rate": 1.5379938844723676e-06, "loss": 0.2521, "step": 6743 }, { "epoch": 1.27, "learning_rate": 1.5372879286470022e-06, "loss": 0.2355, "step": 6744 }, { "epoch": 1.27, "learning_rate": 1.5365820629423112e-06, "loss": 0.2218, "step": 6745 }, { "epoch": 1.27, "learning_rate": 1.5358762874243735e-06, "loss": 0.3268, "step": 6746 }, { "epoch": 1.27, "learning_rate": 1.5351706021592577e-06, "loss": 0.2305, "step": 6747 }, { "epoch": 1.27, "learning_rate": 1.534465007213022e-06, "loss": 0.173, "step": 6748 }, { "epoch": 1.28, "learning_rate": 1.5337595026517197e-06, "loss": 0.1699, "step": 6749 }, { "epoch": 1.28, "learning_rate": 1.5330540885413941e-06, "loss": 0.2371, "step": 6750 }, { "epoch": 1.28, "learning_rate": 1.532348764948078e-06, "loss": 0.2715, "step": 6751 }, { "epoch": 1.28, "learning_rate": 1.5316435319377992e-06, "loss": 0.165, "step": 6752 }, { "epoch": 1.28, "learning_rate": 1.5309383895765762e-06, "loss": 0.2527, "step": 6753 }, { "epoch": 1.28, "learning_rate": 1.5302333379304162e-06, "loss": 0.2337, "step": 6754 }, { "epoch": 1.28, "learning_rate": 1.5295283770653207e-06, "loss": 0.3049, "step": 6755 }, { "epoch": 1.28, "learning_rate": 1.5288235070472834e-06, "loss": 0.2235, "step": 6756 }, { "epoch": 1.28, "learning_rate": 1.5281187279422866e-06, "loss": 0.207, "step": 6757 }, { "epoch": 1.28, "learning_rate": 1.5274140398163045e-06, "loss": 0.2981, "step": 6758 }, { "epoch": 1.28, "learning_rate": 1.5267094427353068e-06, "loss": 0.2196, "step": 6759 }, { "epoch": 1.28, "learning_rate": 1.5260049367652496e-06, "loss": 0.2076, "step": 6760 }, { "epoch": 1.28, "learning_rate": 1.5253005219720822e-06, "loss": 0.2894, "step": 6761 }, { "epoch": 1.28, "learning_rate": 1.5245961984217478e-06, "loss": 0.2084, "step": 6762 }, { "epoch": 1.28, "learning_rate": 1.5238919661801768e-06, "loss": 0.2905, "step": 6763 }, { "epoch": 1.28, "learning_rate": 1.5231878253132947e-06, "loss": 0.2585, "step": 6764 }, { "epoch": 1.28, "learning_rate": 1.522483775887017e-06, "loss": 0.2454, "step": 6765 }, { "epoch": 1.28, "learning_rate": 1.5217798179672493e-06, "loss": 0.2443, "step": 6766 }, { "epoch": 1.28, "learning_rate": 1.5210759516198914e-06, "loss": 0.1894, "step": 6767 }, { "epoch": 1.28, "learning_rate": 1.5203721769108331e-06, "loss": 0.1989, "step": 6768 }, { "epoch": 1.28, "learning_rate": 1.5196684939059542e-06, "loss": 0.236, "step": 6769 }, { "epoch": 1.28, "learning_rate": 1.5189649026711285e-06, "loss": 0.1881, "step": 6770 }, { "epoch": 1.28, "learning_rate": 1.5182614032722203e-06, "loss": 0.2438, "step": 6771 }, { "epoch": 1.28, "learning_rate": 1.5175579957750835e-06, "loss": 0.1677, "step": 6772 }, { "epoch": 1.28, "learning_rate": 1.5168546802455665e-06, "loss": 0.2389, "step": 6773 }, { "epoch": 1.28, "learning_rate": 1.5161514567495078e-06, "loss": 0.2498, "step": 6774 }, { "epoch": 1.28, "learning_rate": 1.5154483253527354e-06, "loss": 0.1799, "step": 6775 }, { "epoch": 1.28, "learning_rate": 1.514745286121071e-06, "loss": 0.3268, "step": 6776 }, { "epoch": 1.28, "learning_rate": 1.5140423391203284e-06, "loss": 0.2391, "step": 6777 }, { "epoch": 1.28, "learning_rate": 1.5133394844163093e-06, "loss": 0.2456, "step": 6778 }, { "epoch": 1.28, "learning_rate": 1.5126367220748093e-06, "loss": 0.2148, "step": 6779 }, { "epoch": 1.28, "learning_rate": 1.511934052161616e-06, "loss": 0.2303, "step": 6780 }, { "epoch": 1.28, "learning_rate": 1.5112314747425062e-06, "loss": 0.2774, "step": 6781 }, { "epoch": 1.28, "learning_rate": 1.510528989883248e-06, "loss": 0.3725, "step": 6782 }, { "epoch": 1.28, "learning_rate": 1.5098265976496046e-06, "loss": 0.2412, "step": 6783 }, { "epoch": 1.28, "learning_rate": 1.5091242981073256e-06, "loss": 0.2243, "step": 6784 }, { "epoch": 1.28, "learning_rate": 1.5084220913221548e-06, "loss": 0.2573, "step": 6785 }, { "epoch": 1.28, "learning_rate": 1.5077199773598278e-06, "loss": 0.2849, "step": 6786 }, { "epoch": 1.28, "learning_rate": 1.5070179562860676e-06, "loss": 0.2245, "step": 6787 }, { "epoch": 1.28, "learning_rate": 1.506316028166594e-06, "loss": 0.2721, "step": 6788 }, { "epoch": 1.28, "learning_rate": 1.5056141930671144e-06, "loss": 0.1923, "step": 6789 }, { "epoch": 1.28, "learning_rate": 1.5049124510533274e-06, "loss": 0.2043, "step": 6790 }, { "epoch": 1.28, "learning_rate": 1.5042108021909258e-06, "loss": 0.2348, "step": 6791 }, { "epoch": 1.28, "learning_rate": 1.5035092465455908e-06, "loss": 0.2657, "step": 6792 }, { "epoch": 1.28, "learning_rate": 1.5028077841829954e-06, "loss": 0.1807, "step": 6793 }, { "epoch": 1.28, "learning_rate": 1.502106415168805e-06, "loss": 0.2247, "step": 6794 }, { "epoch": 1.28, "learning_rate": 1.5014051395686768e-06, "loss": 0.244, "step": 6795 }, { "epoch": 1.28, "learning_rate": 1.500703957448256e-06, "loss": 0.2737, "step": 6796 }, { "epoch": 1.28, "learning_rate": 1.5000028688731816e-06, "loss": 0.2525, "step": 6797 }, { "epoch": 1.28, "learning_rate": 1.4993018739090848e-06, "loss": 0.219, "step": 6798 }, { "epoch": 1.28, "learning_rate": 1.4986009726215852e-06, "loss": 0.2116, "step": 6799 }, { "epoch": 1.28, "learning_rate": 1.497900165076295e-06, "loss": 0.2216, "step": 6800 }, { "epoch": 1.28, "learning_rate": 1.4971994513388187e-06, "loss": 0.2329, "step": 6801 }, { "epoch": 1.29, "learning_rate": 1.4964988314747508e-06, "loss": 0.2103, "step": 6802 }, { "epoch": 1.29, "learning_rate": 1.4957983055496755e-06, "loss": 0.2426, "step": 6803 }, { "epoch": 1.29, "learning_rate": 1.4950978736291715e-06, "loss": 0.1685, "step": 6804 }, { "epoch": 1.29, "learning_rate": 1.494397535778808e-06, "loss": 0.2852, "step": 6805 }, { "epoch": 1.29, "learning_rate": 1.4936972920641423e-06, "loss": 0.286, "step": 6806 }, { "epoch": 1.29, "learning_rate": 1.4929971425507262e-06, "loss": 0.2044, "step": 6807 }, { "epoch": 1.29, "learning_rate": 1.492297087304102e-06, "loss": 0.3615, "step": 6808 }, { "epoch": 1.29, "learning_rate": 1.4915971263898022e-06, "loss": 0.2034, "step": 6809 }, { "epoch": 1.29, "learning_rate": 1.4908972598733502e-06, "loss": 0.2061, "step": 6810 }, { "epoch": 1.29, "learning_rate": 1.490197487820263e-06, "loss": 0.1833, "step": 6811 }, { "epoch": 1.29, "learning_rate": 1.489497810296046e-06, "loss": 0.2692, "step": 6812 }, { "epoch": 1.29, "learning_rate": 1.4887982273661966e-06, "loss": 0.2539, "step": 6813 }, { "epoch": 1.29, "learning_rate": 1.4880987390962058e-06, "loss": 0.207, "step": 6814 }, { "epoch": 1.29, "learning_rate": 1.4873993455515504e-06, "loss": 0.2506, "step": 6815 }, { "epoch": 1.29, "learning_rate": 1.486700046797704e-06, "loss": 0.1983, "step": 6816 }, { "epoch": 1.29, "learning_rate": 1.4860008429001283e-06, "loss": 0.2656, "step": 6817 }, { "epoch": 1.29, "learning_rate": 1.485301733924275e-06, "loss": 0.2792, "step": 6818 }, { "epoch": 1.29, "learning_rate": 1.4846027199355906e-06, "loss": 0.1996, "step": 6819 }, { "epoch": 1.29, "learning_rate": 1.4839038009995104e-06, "loss": 0.2843, "step": 6820 }, { "epoch": 1.29, "learning_rate": 1.4832049771814597e-06, "loss": 0.183, "step": 6821 }, { "epoch": 1.29, "learning_rate": 1.482506248546857e-06, "loss": 0.2721, "step": 6822 }, { "epoch": 1.29, "learning_rate": 1.4818076151611122e-06, "loss": 0.2033, "step": 6823 }, { "epoch": 1.29, "learning_rate": 1.4811090770896242e-06, "loss": 0.2523, "step": 6824 }, { "epoch": 1.29, "learning_rate": 1.4804106343977837e-06, "loss": 0.2555, "step": 6825 }, { "epoch": 1.29, "learning_rate": 1.4797122871509739e-06, "loss": 0.2386, "step": 6826 }, { "epoch": 1.29, "learning_rate": 1.479014035414567e-06, "loss": 0.1916, "step": 6827 }, { "epoch": 1.29, "learning_rate": 1.4783158792539271e-06, "loss": 0.2857, "step": 6828 }, { "epoch": 1.29, "learning_rate": 1.4776178187344107e-06, "loss": 0.2193, "step": 6829 }, { "epoch": 1.29, "learning_rate": 1.476919853921363e-06, "loss": 0.2825, "step": 6830 }, { "epoch": 1.29, "learning_rate": 1.4762219848801209e-06, "loss": 0.2213, "step": 6831 }, { "epoch": 1.29, "learning_rate": 1.4755242116760148e-06, "loss": 0.201, "step": 6832 }, { "epoch": 1.29, "learning_rate": 1.474826534374362e-06, "loss": 0.1731, "step": 6833 }, { "epoch": 1.29, "learning_rate": 1.474128953040474e-06, "loss": 0.1974, "step": 6834 }, { "epoch": 1.29, "learning_rate": 1.4734314677396527e-06, "loss": 0.2694, "step": 6835 }, { "epoch": 1.29, "learning_rate": 1.472734078537189e-06, "loss": 0.2151, "step": 6836 }, { "epoch": 1.29, "learning_rate": 1.4720367854983681e-06, "loss": 0.214, "step": 6837 }, { "epoch": 1.29, "learning_rate": 1.471339588688464e-06, "loss": 0.3457, "step": 6838 }, { "epoch": 1.29, "learning_rate": 1.4706424881727404e-06, "loss": 0.1528, "step": 6839 }, { "epoch": 1.29, "learning_rate": 1.4699454840164563e-06, "loss": 0.2493, "step": 6840 }, { "epoch": 1.29, "learning_rate": 1.4692485762848585e-06, "loss": 0.2795, "step": 6841 }, { "epoch": 1.29, "learning_rate": 1.4685517650431835e-06, "loss": 0.2252, "step": 6842 }, { "epoch": 1.29, "learning_rate": 1.4678550503566624e-06, "loss": 0.1736, "step": 6843 }, { "epoch": 1.29, "learning_rate": 1.467158432290516e-06, "loss": 0.2532, "step": 6844 }, { "epoch": 1.29, "learning_rate": 1.4664619109099543e-06, "loss": 0.2505, "step": 6845 }, { "epoch": 1.29, "learning_rate": 1.4657654862801798e-06, "loss": 0.2807, "step": 6846 }, { "epoch": 1.29, "learning_rate": 1.4650691584663868e-06, "loss": 0.2097, "step": 6847 }, { "epoch": 1.29, "learning_rate": 1.464372927533758e-06, "loss": 0.2515, "step": 6848 }, { "epoch": 1.29, "learning_rate": 1.4636767935474686e-06, "loss": 0.2356, "step": 6849 }, { "epoch": 1.29, "learning_rate": 1.4629807565726856e-06, "loss": 0.2101, "step": 6850 }, { "epoch": 1.29, "learning_rate": 1.4622848166745651e-06, "loss": 0.2037, "step": 6851 }, { "epoch": 1.29, "learning_rate": 1.4615889739182543e-06, "loss": 0.1964, "step": 6852 }, { "epoch": 1.29, "learning_rate": 1.4608932283688937e-06, "loss": 0.2452, "step": 6853 }, { "epoch": 1.29, "learning_rate": 1.460197580091611e-06, "loss": 0.2598, "step": 6854 }, { "epoch": 1.3, "learning_rate": 1.459502029151528e-06, "loss": 0.2391, "step": 6855 }, { "epoch": 1.3, "learning_rate": 1.4588065756137554e-06, "loss": 0.2715, "step": 6856 }, { "epoch": 1.3, "learning_rate": 1.458111219543396e-06, "loss": 0.2088, "step": 6857 }, { "epoch": 1.3, "learning_rate": 1.4574159610055428e-06, "loss": 0.2183, "step": 6858 }, { "epoch": 1.3, "learning_rate": 1.4567208000652792e-06, "loss": 0.2368, "step": 6859 }, { "epoch": 1.3, "learning_rate": 1.4560257367876806e-06, "loss": 0.2621, "step": 6860 }, { "epoch": 1.3, "learning_rate": 1.455330771237813e-06, "loss": 0.2393, "step": 6861 }, { "epoch": 1.3, "learning_rate": 1.4546359034807326e-06, "loss": 0.3069, "step": 6862 }, { "epoch": 1.3, "learning_rate": 1.4539411335814868e-06, "loss": 0.2727, "step": 6863 }, { "epoch": 1.3, "learning_rate": 1.453246461605114e-06, "loss": 0.2498, "step": 6864 }, { "epoch": 1.3, "learning_rate": 1.4525518876166422e-06, "loss": 0.2403, "step": 6865 }, { "epoch": 1.3, "learning_rate": 1.4518574116810945e-06, "loss": 0.36, "step": 6866 }, { "epoch": 1.3, "learning_rate": 1.4511630338634792e-06, "loss": 0.2725, "step": 6867 }, { "epoch": 1.3, "learning_rate": 1.4504687542287966e-06, "loss": 0.2943, "step": 6868 }, { "epoch": 1.3, "learning_rate": 1.4497745728420428e-06, "loss": 0.2399, "step": 6869 }, { "epoch": 1.3, "learning_rate": 1.4490804897681982e-06, "loss": 0.2112, "step": 6870 }, { "epoch": 1.3, "learning_rate": 1.4483865050722363e-06, "loss": 0.2043, "step": 6871 }, { "epoch": 1.3, "learning_rate": 1.4476926188191254e-06, "loss": 0.2732, "step": 6872 }, { "epoch": 1.3, "learning_rate": 1.4469988310738172e-06, "loss": 0.2553, "step": 6873 }, { "epoch": 1.3, "learning_rate": 1.4463051419012589e-06, "loss": 0.2174, "step": 6874 }, { "epoch": 1.3, "learning_rate": 1.4456115513663905e-06, "loss": 0.3438, "step": 6875 }, { "epoch": 1.3, "learning_rate": 1.4449180595341353e-06, "loss": 0.2802, "step": 6876 }, { "epoch": 1.3, "learning_rate": 1.4442246664694155e-06, "loss": 0.264, "step": 6877 }, { "epoch": 1.3, "learning_rate": 1.4435313722371403e-06, "loss": 0.1759, "step": 6878 }, { "epoch": 1.3, "learning_rate": 1.4428381769022065e-06, "loss": 0.2105, "step": 6879 }, { "epoch": 1.3, "learning_rate": 1.4421450805295084e-06, "loss": 0.2355, "step": 6880 }, { "epoch": 1.3, "learning_rate": 1.4414520831839277e-06, "loss": 0.2894, "step": 6881 }, { "epoch": 1.3, "learning_rate": 1.4407591849303336e-06, "loss": 0.2655, "step": 6882 }, { "epoch": 1.3, "learning_rate": 1.4400663858335917e-06, "loss": 0.2278, "step": 6883 }, { "epoch": 1.3, "learning_rate": 1.4393736859585566e-06, "loss": 0.2101, "step": 6884 }, { "epoch": 1.3, "learning_rate": 1.4386810853700695e-06, "loss": 0.2863, "step": 6885 }, { "epoch": 1.3, "learning_rate": 1.4379885841329682e-06, "loss": 0.2015, "step": 6886 }, { "epoch": 1.3, "learning_rate": 1.437296182312078e-06, "loss": 0.2351, "step": 6887 }, { "epoch": 1.3, "learning_rate": 1.4366038799722154e-06, "loss": 0.2448, "step": 6888 }, { "epoch": 1.3, "learning_rate": 1.4359116771781878e-06, "loss": 0.2805, "step": 6889 }, { "epoch": 1.3, "learning_rate": 1.435219573994793e-06, "loss": 0.2187, "step": 6890 }, { "epoch": 1.3, "learning_rate": 1.4345275704868195e-06, "loss": 0.224, "step": 6891 }, { "epoch": 1.3, "learning_rate": 1.4338356667190473e-06, "loss": 0.3261, "step": 6892 }, { "epoch": 1.3, "learning_rate": 1.4331438627562456e-06, "loss": 0.2075, "step": 6893 }, { "epoch": 1.3, "learning_rate": 1.432452158663175e-06, "loss": 0.2819, "step": 6894 }, { "epoch": 1.3, "learning_rate": 1.4317605545045875e-06, "loss": 0.1905, "step": 6895 }, { "epoch": 1.3, "learning_rate": 1.4310690503452245e-06, "loss": 0.2435, "step": 6896 }, { "epoch": 1.3, "learning_rate": 1.4303776462498187e-06, "loss": 0.2852, "step": 6897 }, { "epoch": 1.3, "learning_rate": 1.4296863422830933e-06, "loss": 0.2919, "step": 6898 }, { "epoch": 1.3, "learning_rate": 1.4289951385097622e-06, "loss": 0.2823, "step": 6899 }, { "epoch": 1.3, "learning_rate": 1.4283040349945295e-06, "loss": 0.2312, "step": 6900 }, { "epoch": 1.3, "learning_rate": 1.427613031802091e-06, "loss": 0.2297, "step": 6901 }, { "epoch": 1.3, "learning_rate": 1.4269221289971313e-06, "loss": 0.3114, "step": 6902 }, { "epoch": 1.3, "learning_rate": 1.4262313266443275e-06, "loss": 0.2134, "step": 6903 }, { "epoch": 1.3, "learning_rate": 1.4255406248083453e-06, "loss": 0.3523, "step": 6904 }, { "epoch": 1.3, "learning_rate": 1.4248500235538452e-06, "loss": 0.277, "step": 6905 }, { "epoch": 1.3, "learning_rate": 1.424159522945472e-06, "loss": 0.2379, "step": 6906 }, { "epoch": 1.3, "learning_rate": 1.4234691230478643e-06, "loss": 0.249, "step": 6907 }, { "epoch": 1.31, "learning_rate": 1.4227788239256547e-06, "loss": 0.2433, "step": 6908 }, { "epoch": 1.31, "learning_rate": 1.4220886256434594e-06, "loss": 0.2106, "step": 6909 }, { "epoch": 1.31, "learning_rate": 1.421398528265889e-06, "loss": 0.2457, "step": 6910 }, { "epoch": 1.31, "learning_rate": 1.4207085318575475e-06, "loss": 0.2339, "step": 6911 }, { "epoch": 1.31, "learning_rate": 1.4200186364830235e-06, "loss": 0.2179, "step": 6912 }, { "epoch": 1.31, "learning_rate": 1.4193288422068985e-06, "loss": 0.2777, "step": 6913 }, { "epoch": 1.31, "learning_rate": 1.4186391490937482e-06, "loss": 0.2494, "step": 6914 }, { "epoch": 1.31, "learning_rate": 1.4179495572081313e-06, "loss": 0.2361, "step": 6915 }, { "epoch": 1.31, "learning_rate": 1.4172600666146051e-06, "loss": 0.1669, "step": 6916 }, { "epoch": 1.31, "learning_rate": 1.4165706773777133e-06, "loss": 0.232, "step": 6917 }, { "epoch": 1.31, "learning_rate": 1.415881389561987e-06, "loss": 0.2778, "step": 6918 }, { "epoch": 1.31, "learning_rate": 1.4151922032319553e-06, "loss": 0.211, "step": 6919 }, { "epoch": 1.31, "learning_rate": 1.414503118452133e-06, "loss": 0.2927, "step": 6920 }, { "epoch": 1.31, "learning_rate": 1.4138141352870233e-06, "loss": 0.2336, "step": 6921 }, { "epoch": 1.31, "learning_rate": 1.4131252538011258e-06, "loss": 0.2149, "step": 6922 }, { "epoch": 1.31, "learning_rate": 1.4124364740589274e-06, "loss": 0.3052, "step": 6923 }, { "epoch": 1.31, "learning_rate": 1.4117477961249032e-06, "loss": 0.2709, "step": 6924 }, { "epoch": 1.31, "learning_rate": 1.4110592200635239e-06, "loss": 0.2818, "step": 6925 }, { "epoch": 1.31, "learning_rate": 1.4103707459392468e-06, "loss": 0.2884, "step": 6926 }, { "epoch": 1.31, "learning_rate": 1.409682373816521e-06, "loss": 0.2737, "step": 6927 }, { "epoch": 1.31, "learning_rate": 1.4089941037597857e-06, "loss": 0.2487, "step": 6928 }, { "epoch": 1.31, "learning_rate": 1.408305935833471e-06, "loss": 0.1889, "step": 6929 }, { "epoch": 1.31, "learning_rate": 1.4076178701019966e-06, "loss": 0.2341, "step": 6930 }, { "epoch": 1.31, "learning_rate": 1.406929906629774e-06, "loss": 0.2895, "step": 6931 }, { "epoch": 1.31, "learning_rate": 1.4062420454812038e-06, "loss": 0.2444, "step": 6932 }, { "epoch": 1.31, "learning_rate": 1.4055542867206774e-06, "loss": 0.3304, "step": 6933 }, { "epoch": 1.31, "learning_rate": 1.4048666304125772e-06, "loss": 0.2721, "step": 6934 }, { "epoch": 1.31, "learning_rate": 1.4041790766212743e-06, "loss": 0.2117, "step": 6935 }, { "epoch": 1.31, "learning_rate": 1.4034916254111343e-06, "loss": 0.2606, "step": 6936 }, { "epoch": 1.31, "learning_rate": 1.4028042768465078e-06, "loss": 0.3189, "step": 6937 }, { "epoch": 1.31, "learning_rate": 1.4021170309917382e-06, "loss": 0.1999, "step": 6938 }, { "epoch": 1.31, "learning_rate": 1.4014298879111622e-06, "loss": 0.2807, "step": 6939 }, { "epoch": 1.31, "learning_rate": 1.4007428476691013e-06, "loss": 0.2229, "step": 6940 }, { "epoch": 1.31, "learning_rate": 1.4000559103298703e-06, "loss": 0.2605, "step": 6941 }, { "epoch": 1.31, "learning_rate": 1.3993690759577773e-06, "loss": 0.2744, "step": 6942 }, { "epoch": 1.31, "learning_rate": 1.3986823446171139e-06, "loss": 0.2918, "step": 6943 }, { "epoch": 1.31, "learning_rate": 1.397995716372167e-06, "loss": 0.2814, "step": 6944 }, { "epoch": 1.31, "learning_rate": 1.3973091912872153e-06, "loss": 0.2586, "step": 6945 }, { "epoch": 1.31, "learning_rate": 1.3966227694265211e-06, "loss": 0.1761, "step": 6946 }, { "epoch": 1.31, "learning_rate": 1.3959364508543445e-06, "loss": 0.2054, "step": 6947 }, { "epoch": 1.31, "learning_rate": 1.3952502356349323e-06, "loss": 0.1977, "step": 6948 }, { "epoch": 1.31, "learning_rate": 1.3945641238325192e-06, "loss": 0.2299, "step": 6949 }, { "epoch": 1.31, "learning_rate": 1.3938781155113356e-06, "loss": 0.2446, "step": 6950 }, { "epoch": 1.31, "learning_rate": 1.3931922107356004e-06, "loss": 0.2628, "step": 6951 }, { "epoch": 1.31, "learning_rate": 1.3925064095695183e-06, "loss": 0.1899, "step": 6952 }, { "epoch": 1.31, "learning_rate": 1.391820712077291e-06, "loss": 0.1991, "step": 6953 }, { "epoch": 1.31, "learning_rate": 1.3911351183231082e-06, "loss": 0.2734, "step": 6954 }, { "epoch": 1.31, "learning_rate": 1.3904496283711452e-06, "loss": 0.1902, "step": 6955 }, { "epoch": 1.31, "learning_rate": 1.3897642422855754e-06, "loss": 0.3107, "step": 6956 }, { "epoch": 1.31, "learning_rate": 1.389078960130557e-06, "loss": 0.2818, "step": 6957 }, { "epoch": 1.31, "learning_rate": 1.3883937819702406e-06, "loss": 0.2455, "step": 6958 }, { "epoch": 1.31, "learning_rate": 1.3877087078687665e-06, "loss": 0.3024, "step": 6959 }, { "epoch": 1.31, "learning_rate": 1.3870237378902651e-06, "loss": 0.2642, "step": 6960 }, { "epoch": 1.32, "learning_rate": 1.3863388720988574e-06, "loss": 0.238, "step": 6961 }, { "epoch": 1.32, "learning_rate": 1.3856541105586545e-06, "loss": 0.2114, "step": 6962 }, { "epoch": 1.32, "learning_rate": 1.3849694533337582e-06, "loss": 0.3048, "step": 6963 }, { "epoch": 1.32, "learning_rate": 1.3842849004882591e-06, "loss": 0.271, "step": 6964 }, { "epoch": 1.32, "learning_rate": 1.3836004520862404e-06, "loss": 0.2233, "step": 6965 }, { "epoch": 1.32, "learning_rate": 1.382916108191773e-06, "loss": 0.2947, "step": 6966 }, { "epoch": 1.32, "learning_rate": 1.3822318688689201e-06, "loss": 0.1602, "step": 6967 }, { "epoch": 1.32, "learning_rate": 1.3815477341817334e-06, "loss": 0.3536, "step": 6968 }, { "epoch": 1.32, "learning_rate": 1.3808637041942562e-06, "loss": 0.2834, "step": 6969 }, { "epoch": 1.32, "learning_rate": 1.3801797789705209e-06, "loss": 0.2304, "step": 6970 }, { "epoch": 1.32, "learning_rate": 1.379495958574551e-06, "loss": 0.2114, "step": 6971 }, { "epoch": 1.32, "learning_rate": 1.3788122430703595e-06, "loss": 0.2548, "step": 6972 }, { "epoch": 1.32, "learning_rate": 1.37812863252195e-06, "loss": 0.2929, "step": 6973 }, { "epoch": 1.32, "learning_rate": 1.3774451269933165e-06, "loss": 0.1932, "step": 6974 }, { "epoch": 1.32, "learning_rate": 1.376761726548442e-06, "loss": 0.2234, "step": 6975 }, { "epoch": 1.32, "learning_rate": 1.376078431251301e-06, "loss": 0.2984, "step": 6976 }, { "epoch": 1.32, "learning_rate": 1.3753952411658567e-06, "loss": 0.2226, "step": 6977 }, { "epoch": 1.32, "learning_rate": 1.374712156356066e-06, "loss": 0.2349, "step": 6978 }, { "epoch": 1.32, "learning_rate": 1.3740291768858705e-06, "loss": 0.2037, "step": 6979 }, { "epoch": 1.32, "learning_rate": 1.3733463028192046e-06, "loss": 0.271, "step": 6980 }, { "epoch": 1.32, "learning_rate": 1.3726635342199961e-06, "loss": 0.2794, "step": 6981 }, { "epoch": 1.32, "learning_rate": 1.3719808711521573e-06, "loss": 0.2669, "step": 6982 }, { "epoch": 1.32, "learning_rate": 1.3712983136795926e-06, "loss": 0.2653, "step": 6983 }, { "epoch": 1.32, "learning_rate": 1.3706158618662001e-06, "loss": 0.2399, "step": 6984 }, { "epoch": 1.32, "learning_rate": 1.369933515775862e-06, "loss": 0.2702, "step": 6985 }, { "epoch": 1.32, "learning_rate": 1.3692512754724535e-06, "loss": 0.1836, "step": 6986 }, { "epoch": 1.32, "learning_rate": 1.3685691410198423e-06, "loss": 0.2762, "step": 6987 }, { "epoch": 1.32, "learning_rate": 1.3678871124818827e-06, "loss": 0.2479, "step": 6988 }, { "epoch": 1.32, "learning_rate": 1.3672051899224198e-06, "loss": 0.2865, "step": 6989 }, { "epoch": 1.32, "learning_rate": 1.3665233734052895e-06, "loss": 0.3605, "step": 6990 }, { "epoch": 1.32, "learning_rate": 1.3658416629943183e-06, "loss": 0.2537, "step": 6991 }, { "epoch": 1.32, "learning_rate": 1.3651600587533206e-06, "loss": 0.3371, "step": 6992 }, { "epoch": 1.32, "learning_rate": 1.364478560746103e-06, "loss": 0.2161, "step": 6993 }, { "epoch": 1.32, "learning_rate": 1.3637971690364616e-06, "loss": 0.1889, "step": 6994 }, { "epoch": 1.32, "learning_rate": 1.3631158836881813e-06, "loss": 0.2142, "step": 6995 }, { "epoch": 1.32, "learning_rate": 1.362434704765039e-06, "loss": 0.2606, "step": 6996 }, { "epoch": 1.32, "learning_rate": 1.3617536323308006e-06, "loss": 0.2324, "step": 6997 }, { "epoch": 1.32, "learning_rate": 1.361072666449222e-06, "loss": 0.2102, "step": 6998 }, { "epoch": 1.32, "learning_rate": 1.3603918071840488e-06, "loss": 0.2861, "step": 6999 }, { "epoch": 1.32, "learning_rate": 1.359711054599018e-06, "loss": 0.2406, "step": 7000 }, { "epoch": 1.32, "learning_rate": 1.3590304087578552e-06, "loss": 0.1948, "step": 7001 }, { "epoch": 1.32, "learning_rate": 1.3583498697242764e-06, "loss": 0.1871, "step": 7002 }, { "epoch": 1.32, "learning_rate": 1.3576694375619876e-06, "loss": 0.1991, "step": 7003 }, { "epoch": 1.32, "learning_rate": 1.3569891123346854e-06, "loss": 0.2243, "step": 7004 }, { "epoch": 1.32, "learning_rate": 1.3563088941060554e-06, "loss": 0.3711, "step": 7005 }, { "epoch": 1.32, "learning_rate": 1.3556287829397742e-06, "loss": 0.2127, "step": 7006 }, { "epoch": 1.32, "learning_rate": 1.3549487788995075e-06, "loss": 0.2346, "step": 7007 }, { "epoch": 1.32, "learning_rate": 1.3542688820489103e-06, "loss": 0.1569, "step": 7008 }, { "epoch": 1.32, "learning_rate": 1.3535890924516315e-06, "loss": 0.2331, "step": 7009 }, { "epoch": 1.32, "learning_rate": 1.3529094101713048e-06, "loss": 0.2572, "step": 7010 }, { "epoch": 1.32, "learning_rate": 1.3522298352715548e-06, "loss": 0.2316, "step": 7011 }, { "epoch": 1.32, "learning_rate": 1.3515503678160014e-06, "loss": 0.1917, "step": 7012 }, { "epoch": 1.32, "learning_rate": 1.3508710078682471e-06, "loss": 0.1748, "step": 7013 }, { "epoch": 1.33, "learning_rate": 1.3501917554918876e-06, "loss": 0.26, "step": 7014 }, { "epoch": 1.33, "learning_rate": 1.3495126107505114e-06, "loss": 0.2534, "step": 7015 }, { "epoch": 1.33, "learning_rate": 1.3488335737076914e-06, "loss": 0.2263, "step": 7016 }, { "epoch": 1.33, "learning_rate": 1.348154644426993e-06, "loss": 0.2026, "step": 7017 }, { "epoch": 1.33, "learning_rate": 1.3474758229719745e-06, "loss": 0.2229, "step": 7018 }, { "epoch": 1.33, "learning_rate": 1.3467971094061772e-06, "loss": 0.2253, "step": 7019 }, { "epoch": 1.33, "learning_rate": 1.3461185037931396e-06, "loss": 0.1966, "step": 7020 }, { "epoch": 1.33, "learning_rate": 1.3454400061963866e-06, "loss": 0.2572, "step": 7021 }, { "epoch": 1.33, "learning_rate": 1.3447616166794302e-06, "loss": 0.2123, "step": 7022 }, { "epoch": 1.33, "learning_rate": 1.344083335305778e-06, "loss": 0.2189, "step": 7023 }, { "epoch": 1.33, "learning_rate": 1.3434051621389255e-06, "loss": 0.1969, "step": 7024 }, { "epoch": 1.33, "learning_rate": 1.342727097242354e-06, "loss": 0.2292, "step": 7025 }, { "epoch": 1.33, "learning_rate": 1.3420491406795404e-06, "loss": 0.2242, "step": 7026 }, { "epoch": 1.33, "learning_rate": 1.34137129251395e-06, "loss": 0.3028, "step": 7027 }, { "epoch": 1.33, "learning_rate": 1.3406935528090334e-06, "loss": 0.2532, "step": 7028 }, { "epoch": 1.33, "learning_rate": 1.340015921628238e-06, "loss": 0.2155, "step": 7029 }, { "epoch": 1.33, "learning_rate": 1.3393383990349961e-06, "loss": 0.2541, "step": 7030 }, { "epoch": 1.33, "learning_rate": 1.3386609850927323e-06, "loss": 0.3116, "step": 7031 }, { "epoch": 1.33, "learning_rate": 1.33798367986486e-06, "loss": 0.2463, "step": 7032 }, { "epoch": 1.33, "learning_rate": 1.3373064834147819e-06, "loss": 0.2442, "step": 7033 }, { "epoch": 1.33, "learning_rate": 1.3366293958058918e-06, "loss": 0.2294, "step": 7034 }, { "epoch": 1.33, "learning_rate": 1.3359524171015722e-06, "loss": 0.2399, "step": 7035 }, { "epoch": 1.33, "learning_rate": 1.3352755473651966e-06, "loss": 0.2537, "step": 7036 }, { "epoch": 1.33, "learning_rate": 1.3345987866601272e-06, "loss": 0.2646, "step": 7037 }, { "epoch": 1.33, "learning_rate": 1.3339221350497164e-06, "loss": 0.208, "step": 7038 }, { "epoch": 1.33, "learning_rate": 1.3332455925973065e-06, "loss": 0.2078, "step": 7039 }, { "epoch": 1.33, "learning_rate": 1.3325691593662295e-06, "loss": 0.2686, "step": 7040 }, { "epoch": 1.33, "learning_rate": 1.3318928354198074e-06, "loss": 0.265, "step": 7041 }, { "epoch": 1.33, "learning_rate": 1.3312166208213511e-06, "loss": 0.2125, "step": 7042 }, { "epoch": 1.33, "learning_rate": 1.3305405156341623e-06, "loss": 0.1604, "step": 7043 }, { "epoch": 1.33, "learning_rate": 1.3298645199215321e-06, "loss": 0.2736, "step": 7044 }, { "epoch": 1.33, "learning_rate": 1.3291886337467414e-06, "loss": 0.2822, "step": 7045 }, { "epoch": 1.33, "learning_rate": 1.3285128571730599e-06, "loss": 0.2521, "step": 7046 }, { "epoch": 1.33, "learning_rate": 1.3278371902637492e-06, "loss": 0.2293, "step": 7047 }, { "epoch": 1.33, "learning_rate": 1.3271616330820583e-06, "loss": 0.2057, "step": 7048 }, { "epoch": 1.33, "learning_rate": 1.3264861856912275e-06, "loss": 0.2441, "step": 7049 }, { "epoch": 1.33, "learning_rate": 1.3258108481544849e-06, "loss": 0.2532, "step": 7050 }, { "epoch": 1.33, "learning_rate": 1.325135620535052e-06, "loss": 0.2821, "step": 7051 }, { "epoch": 1.33, "learning_rate": 1.3244605028961372e-06, "loss": 0.3188, "step": 7052 }, { "epoch": 1.33, "learning_rate": 1.3237854953009368e-06, "loss": 0.1813, "step": 7053 }, { "epoch": 1.33, "learning_rate": 1.3231105978126418e-06, "loss": 0.3168, "step": 7054 }, { "epoch": 1.33, "learning_rate": 1.3224358104944298e-06, "loss": 0.2542, "step": 7055 }, { "epoch": 1.33, "learning_rate": 1.3217611334094662e-06, "loss": 0.2017, "step": 7056 }, { "epoch": 1.33, "learning_rate": 1.321086566620911e-06, "loss": 0.2821, "step": 7057 }, { "epoch": 1.33, "learning_rate": 1.32041211019191e-06, "loss": 0.2562, "step": 7058 }, { "epoch": 1.33, "learning_rate": 1.3197377641856002e-06, "loss": 0.2171, "step": 7059 }, { "epoch": 1.33, "learning_rate": 1.3190635286651079e-06, "loss": 0.2222, "step": 7060 }, { "epoch": 1.33, "learning_rate": 1.318389403693549e-06, "loss": 0.2751, "step": 7061 }, { "epoch": 1.33, "learning_rate": 1.3177153893340297e-06, "loss": 0.2715, "step": 7062 }, { "epoch": 1.33, "learning_rate": 1.317041485649645e-06, "loss": 0.3114, "step": 7063 }, { "epoch": 1.33, "learning_rate": 1.3163676927034794e-06, "loss": 0.2948, "step": 7064 }, { "epoch": 1.33, "learning_rate": 1.315694010558608e-06, "loss": 0.287, "step": 7065 }, { "epoch": 1.33, "learning_rate": 1.315020439278095e-06, "loss": 0.2217, "step": 7066 }, { "epoch": 1.34, "learning_rate": 1.314346978924994e-06, "loss": 0.2267, "step": 7067 }, { "epoch": 1.34, "learning_rate": 1.3136736295623487e-06, "loss": 0.2302, "step": 7068 }, { "epoch": 1.34, "learning_rate": 1.3130003912531921e-06, "loss": 0.297, "step": 7069 }, { "epoch": 1.34, "learning_rate": 1.3123272640605473e-06, "loss": 0.1953, "step": 7070 }, { "epoch": 1.34, "learning_rate": 1.3116542480474253e-06, "loss": 0.2599, "step": 7071 }, { "epoch": 1.34, "learning_rate": 1.3109813432768294e-06, "loss": 0.241, "step": 7072 }, { "epoch": 1.34, "learning_rate": 1.3103085498117508e-06, "loss": 0.2205, "step": 7073 }, { "epoch": 1.34, "learning_rate": 1.3096358677151693e-06, "loss": 0.1807, "step": 7074 }, { "epoch": 1.34, "learning_rate": 1.3089632970500571e-06, "loss": 0.2932, "step": 7075 }, { "epoch": 1.34, "learning_rate": 1.3082908378793734e-06, "loss": 0.2854, "step": 7076 }, { "epoch": 1.34, "learning_rate": 1.3076184902660684e-06, "loss": 0.2201, "step": 7077 }, { "epoch": 1.34, "learning_rate": 1.3069462542730805e-06, "loss": 0.1963, "step": 7078 }, { "epoch": 1.34, "learning_rate": 1.306274129963341e-06, "loss": 0.1733, "step": 7079 }, { "epoch": 1.34, "learning_rate": 1.3056021173997652e-06, "loss": 0.2315, "step": 7080 }, { "epoch": 1.34, "learning_rate": 1.3049302166452618e-06, "loss": 0.2147, "step": 7081 }, { "epoch": 1.34, "learning_rate": 1.3042584277627307e-06, "loss": 0.2836, "step": 7082 }, { "epoch": 1.34, "learning_rate": 1.3035867508150563e-06, "loss": 0.2719, "step": 7083 }, { "epoch": 1.34, "learning_rate": 1.3029151858651145e-06, "loss": 0.2214, "step": 7084 }, { "epoch": 1.34, "learning_rate": 1.302243732975775e-06, "loss": 0.2981, "step": 7085 }, { "epoch": 1.34, "learning_rate": 1.3015723922098899e-06, "loss": 0.2085, "step": 7086 }, { "epoch": 1.34, "learning_rate": 1.3009011636303047e-06, "loss": 0.2402, "step": 7087 }, { "epoch": 1.34, "learning_rate": 1.3002300472998563e-06, "loss": 0.2863, "step": 7088 }, { "epoch": 1.34, "learning_rate": 1.2995590432813655e-06, "loss": 0.1911, "step": 7089 }, { "epoch": 1.34, "learning_rate": 1.2988881516376482e-06, "loss": 0.2516, "step": 7090 }, { "epoch": 1.34, "learning_rate": 1.2982173724315072e-06, "loss": 0.2615, "step": 7091 }, { "epoch": 1.34, "learning_rate": 1.297546705725733e-06, "loss": 0.2702, "step": 7092 }, { "epoch": 1.34, "learning_rate": 1.2968761515831096e-06, "loss": 0.2156, "step": 7093 }, { "epoch": 1.34, "learning_rate": 1.2962057100664088e-06, "loss": 0.2951, "step": 7094 }, { "epoch": 1.34, "learning_rate": 1.2955353812383884e-06, "loss": 0.2523, "step": 7095 }, { "epoch": 1.34, "learning_rate": 1.2948651651618015e-06, "loss": 0.2155, "step": 7096 }, { "epoch": 1.34, "learning_rate": 1.294195061899388e-06, "loss": 0.2911, "step": 7097 }, { "epoch": 1.34, "learning_rate": 1.2935250715138747e-06, "loss": 0.2163, "step": 7098 }, { "epoch": 1.34, "learning_rate": 1.2928551940679823e-06, "loss": 0.2894, "step": 7099 }, { "epoch": 1.34, "learning_rate": 1.2921854296244186e-06, "loss": 0.2723, "step": 7100 }, { "epoch": 1.34, "learning_rate": 1.2915157782458804e-06, "loss": 0.2674, "step": 7101 }, { "epoch": 1.34, "learning_rate": 1.2908462399950552e-06, "loss": 0.2723, "step": 7102 }, { "epoch": 1.34, "learning_rate": 1.2901768149346188e-06, "loss": 0.2373, "step": 7103 }, { "epoch": 1.34, "learning_rate": 1.2895075031272376e-06, "loss": 0.2286, "step": 7104 }, { "epoch": 1.34, "learning_rate": 1.2888383046355662e-06, "loss": 0.1907, "step": 7105 }, { "epoch": 1.34, "learning_rate": 1.288169219522249e-06, "loss": 0.291, "step": 7106 }, { "epoch": 1.34, "learning_rate": 1.28750024784992e-06, "loss": 0.3049, "step": 7107 }, { "epoch": 1.34, "learning_rate": 1.2868313896812027e-06, "loss": 0.1844, "step": 7108 }, { "epoch": 1.34, "learning_rate": 1.2861626450787097e-06, "loss": 0.2819, "step": 7109 }, { "epoch": 1.34, "learning_rate": 1.2854940141050426e-06, "loss": 0.2176, "step": 7110 }, { "epoch": 1.34, "learning_rate": 1.2848254968227933e-06, "loss": 0.3508, "step": 7111 }, { "epoch": 1.34, "learning_rate": 1.2841570932945413e-06, "loss": 0.2015, "step": 7112 }, { "epoch": 1.34, "learning_rate": 1.2834888035828597e-06, "loss": 0.3061, "step": 7113 }, { "epoch": 1.34, "learning_rate": 1.2828206277503052e-06, "loss": 0.1966, "step": 7114 }, { "epoch": 1.34, "learning_rate": 1.2821525658594259e-06, "loss": 0.2666, "step": 7115 }, { "epoch": 1.34, "learning_rate": 1.2814846179727635e-06, "loss": 0.2483, "step": 7116 }, { "epoch": 1.34, "learning_rate": 1.2808167841528424e-06, "loss": 0.1664, "step": 7117 }, { "epoch": 1.34, "learning_rate": 1.280149064462179e-06, "loss": 0.2405, "step": 7118 }, { "epoch": 1.34, "learning_rate": 1.2794814589632826e-06, "loss": 0.2453, "step": 7119 }, { "epoch": 1.35, "learning_rate": 1.2788139677186449e-06, "loss": 0.2295, "step": 7120 }, { "epoch": 1.35, "learning_rate": 1.278146590790753e-06, "loss": 0.2051, "step": 7121 }, { "epoch": 1.35, "learning_rate": 1.2774793282420812e-06, "loss": 0.2792, "step": 7122 }, { "epoch": 1.35, "learning_rate": 1.2768121801350896e-06, "loss": 0.2647, "step": 7123 }, { "epoch": 1.35, "learning_rate": 1.2761451465322338e-06, "loss": 0.295, "step": 7124 }, { "epoch": 1.35, "learning_rate": 1.275478227495956e-06, "loss": 0.2792, "step": 7125 }, { "epoch": 1.35, "learning_rate": 1.2748114230886838e-06, "loss": 0.1671, "step": 7126 }, { "epoch": 1.35, "learning_rate": 1.274144733372841e-06, "loss": 0.2189, "step": 7127 }, { "epoch": 1.35, "learning_rate": 1.273478158410837e-06, "loss": 0.2253, "step": 7128 }, { "epoch": 1.35, "learning_rate": 1.272811698265068e-06, "loss": 0.2292, "step": 7129 }, { "epoch": 1.35, "learning_rate": 1.2721453529979247e-06, "loss": 0.3018, "step": 7130 }, { "epoch": 1.35, "learning_rate": 1.2714791226717838e-06, "loss": 0.3167, "step": 7131 }, { "epoch": 1.35, "learning_rate": 1.2708130073490121e-06, "loss": 0.2149, "step": 7132 }, { "epoch": 1.35, "learning_rate": 1.2701470070919652e-06, "loss": 0.2475, "step": 7133 }, { "epoch": 1.35, "learning_rate": 1.2694811219629885e-06, "loss": 0.2506, "step": 7134 }, { "epoch": 1.35, "learning_rate": 1.2688153520244162e-06, "loss": 0.2171, "step": 7135 }, { "epoch": 1.35, "learning_rate": 1.2681496973385715e-06, "loss": 0.187, "step": 7136 }, { "epoch": 1.35, "learning_rate": 1.2674841579677677e-06, "loss": 0.2819, "step": 7137 }, { "epoch": 1.35, "learning_rate": 1.2668187339743066e-06, "loss": 0.2265, "step": 7138 }, { "epoch": 1.35, "learning_rate": 1.266153425420479e-06, "loss": 0.2054, "step": 7139 }, { "epoch": 1.35, "learning_rate": 1.2654882323685657e-06, "loss": 0.2292, "step": 7140 }, { "epoch": 1.35, "learning_rate": 1.2648231548808364e-06, "loss": 0.3817, "step": 7141 }, { "epoch": 1.35, "learning_rate": 1.2641581930195495e-06, "loss": 0.2177, "step": 7142 }, { "epoch": 1.35, "learning_rate": 1.263493346846953e-06, "loss": 0.2633, "step": 7143 }, { "epoch": 1.35, "learning_rate": 1.2628286164252834e-06, "loss": 0.2083, "step": 7144 }, { "epoch": 1.35, "learning_rate": 1.262164001816768e-06, "loss": 0.3078, "step": 7145 }, { "epoch": 1.35, "learning_rate": 1.2614995030836215e-06, "loss": 0.2156, "step": 7146 }, { "epoch": 1.35, "learning_rate": 1.260835120288049e-06, "loss": 0.2796, "step": 7147 }, { "epoch": 1.35, "learning_rate": 1.2601708534922435e-06, "loss": 0.1603, "step": 7148 }, { "epoch": 1.35, "learning_rate": 1.2595067027583885e-06, "loss": 0.2421, "step": 7149 }, { "epoch": 1.35, "learning_rate": 1.2588426681486555e-06, "loss": 0.2669, "step": 7150 }, { "epoch": 1.35, "learning_rate": 1.2581787497252046e-06, "loss": 0.2485, "step": 7151 }, { "epoch": 1.35, "learning_rate": 1.2575149475501891e-06, "loss": 0.2602, "step": 7152 }, { "epoch": 1.35, "learning_rate": 1.256851261685746e-06, "loss": 0.2512, "step": 7153 }, { "epoch": 1.35, "learning_rate": 1.2561876921940027e-06, "loss": 0.2689, "step": 7154 }, { "epoch": 1.35, "learning_rate": 1.2555242391370802e-06, "loss": 0.2179, "step": 7155 }, { "epoch": 1.35, "learning_rate": 1.2548609025770828e-06, "loss": 0.2387, "step": 7156 }, { "epoch": 1.35, "learning_rate": 1.2541976825761055e-06, "loss": 0.2516, "step": 7157 }, { "epoch": 1.35, "learning_rate": 1.253534579196236e-06, "loss": 0.2765, "step": 7158 }, { "epoch": 1.35, "learning_rate": 1.2528715924995461e-06, "loss": 0.3025, "step": 7159 }, { "epoch": 1.35, "learning_rate": 1.2522087225480984e-06, "loss": 0.2211, "step": 7160 }, { "epoch": 1.35, "learning_rate": 1.2515459694039473e-06, "loss": 0.2974, "step": 7161 }, { "epoch": 1.35, "learning_rate": 1.250883333129131e-06, "loss": 0.2221, "step": 7162 }, { "epoch": 1.35, "learning_rate": 1.2502208137856818e-06, "loss": 0.3403, "step": 7163 }, { "epoch": 1.35, "learning_rate": 1.2495584114356196e-06, "loss": 0.2414, "step": 7164 }, { "epoch": 1.35, "learning_rate": 1.2488961261409496e-06, "loss": 0.2552, "step": 7165 }, { "epoch": 1.35, "learning_rate": 1.2482339579636717e-06, "loss": 0.2238, "step": 7166 }, { "epoch": 1.35, "learning_rate": 1.2475719069657733e-06, "loss": 0.18, "step": 7167 }, { "epoch": 1.35, "learning_rate": 1.2469099732092258e-06, "loss": 0.1968, "step": 7168 }, { "epoch": 1.35, "learning_rate": 1.2462481567559966e-06, "loss": 0.2688, "step": 7169 }, { "epoch": 1.35, "learning_rate": 1.2455864576680402e-06, "loss": 0.2605, "step": 7170 }, { "epoch": 1.35, "learning_rate": 1.2449248760072952e-06, "loss": 0.3367, "step": 7171 }, { "epoch": 1.35, "learning_rate": 1.2442634118356963e-06, "loss": 0.1774, "step": 7172 }, { "epoch": 1.36, "learning_rate": 1.2436020652151631e-06, "loss": 0.2543, "step": 7173 }, { "epoch": 1.36, "learning_rate": 1.2429408362076047e-06, "loss": 0.1864, "step": 7174 }, { "epoch": 1.36, "learning_rate": 1.2422797248749201e-06, "loss": 0.2196, "step": 7175 }, { "epoch": 1.36, "learning_rate": 1.2416187312789967e-06, "loss": 0.196, "step": 7176 }, { "epoch": 1.36, "learning_rate": 1.2409578554817106e-06, "loss": 0.2765, "step": 7177 }, { "epoch": 1.36, "learning_rate": 1.2402970975449272e-06, "loss": 0.297, "step": 7178 }, { "epoch": 1.36, "learning_rate": 1.2396364575305011e-06, "loss": 0.2151, "step": 7179 }, { "epoch": 1.36, "learning_rate": 1.2389759355002756e-06, "loss": 0.2619, "step": 7180 }, { "epoch": 1.36, "learning_rate": 1.238315531516083e-06, "loss": 0.3483, "step": 7181 }, { "epoch": 1.36, "learning_rate": 1.2376552456397436e-06, "loss": 0.2544, "step": 7182 }, { "epoch": 1.36, "learning_rate": 1.2369950779330703e-06, "loss": 0.2841, "step": 7183 }, { "epoch": 1.36, "learning_rate": 1.2363350284578594e-06, "loss": 0.1841, "step": 7184 }, { "epoch": 1.36, "learning_rate": 1.235675097275899e-06, "loss": 0.2032, "step": 7185 }, { "epoch": 1.36, "learning_rate": 1.235015284448969e-06, "loss": 0.2173, "step": 7186 }, { "epoch": 1.36, "learning_rate": 1.2343555900388325e-06, "loss": 0.2583, "step": 7187 }, { "epoch": 1.36, "learning_rate": 1.2336960141072438e-06, "loss": 0.2024, "step": 7188 }, { "epoch": 1.36, "learning_rate": 1.2330365567159503e-06, "loss": 0.2502, "step": 7189 }, { "epoch": 1.36, "learning_rate": 1.232377217926681e-06, "loss": 0.2628, "step": 7190 }, { "epoch": 1.36, "learning_rate": 1.2317179978011578e-06, "loss": 0.2837, "step": 7191 }, { "epoch": 1.36, "learning_rate": 1.2310588964010942e-06, "loss": 0.2217, "step": 7192 }, { "epoch": 1.36, "learning_rate": 1.2303999137881852e-06, "loss": 0.3049, "step": 7193 }, { "epoch": 1.36, "learning_rate": 1.2297410500241225e-06, "loss": 0.186, "step": 7194 }, { "epoch": 1.36, "learning_rate": 1.2290823051705822e-06, "loss": 0.246, "step": 7195 }, { "epoch": 1.36, "learning_rate": 1.228423679289228e-06, "loss": 0.1742, "step": 7196 }, { "epoch": 1.36, "learning_rate": 1.2277651724417177e-06, "loss": 0.2738, "step": 7197 }, { "epoch": 1.36, "learning_rate": 1.2271067846896948e-06, "loss": 0.22, "step": 7198 }, { "epoch": 1.36, "learning_rate": 1.2264485160947883e-06, "loss": 0.3206, "step": 7199 }, { "epoch": 1.36, "learning_rate": 1.2257903667186234e-06, "loss": 0.2196, "step": 7200 }, { "epoch": 1.36, "learning_rate": 1.2251323366228097e-06, "loss": 0.2286, "step": 7201 }, { "epoch": 1.36, "learning_rate": 1.2244744258689433e-06, "loss": 0.2182, "step": 7202 }, { "epoch": 1.36, "learning_rate": 1.2238166345186152e-06, "loss": 0.2021, "step": 7203 }, { "epoch": 1.36, "learning_rate": 1.223158962633401e-06, "loss": 0.2402, "step": 7204 }, { "epoch": 1.36, "learning_rate": 1.2225014102748664e-06, "loss": 0.2523, "step": 7205 }, { "epoch": 1.36, "learning_rate": 1.221843977504565e-06, "loss": 0.2675, "step": 7206 }, { "epoch": 1.36, "learning_rate": 1.2211866643840409e-06, "loss": 0.2786, "step": 7207 }, { "epoch": 1.36, "learning_rate": 1.220529470974825e-06, "loss": 0.1863, "step": 7208 }, { "epoch": 1.36, "learning_rate": 1.2198723973384389e-06, "loss": 0.2732, "step": 7209 }, { "epoch": 1.36, "learning_rate": 1.2192154435363912e-06, "loss": 0.2601, "step": 7210 }, { "epoch": 1.36, "learning_rate": 1.218558609630181e-06, "loss": 0.2441, "step": 7211 }, { "epoch": 1.36, "learning_rate": 1.2179018956812947e-06, "loss": 0.21, "step": 7212 }, { "epoch": 1.36, "learning_rate": 1.2172453017512084e-06, "loss": 0.1564, "step": 7213 }, { "epoch": 1.36, "learning_rate": 1.2165888279013868e-06, "loss": 0.2384, "step": 7214 }, { "epoch": 1.36, "learning_rate": 1.215932474193283e-06, "loss": 0.2146, "step": 7215 }, { "epoch": 1.36, "learning_rate": 1.2152762406883392e-06, "loss": 0.2926, "step": 7216 }, { "epoch": 1.36, "learning_rate": 1.214620127447986e-06, "loss": 0.1745, "step": 7217 }, { "epoch": 1.36, "learning_rate": 1.2139641345336433e-06, "loss": 0.2347, "step": 7218 }, { "epoch": 1.36, "learning_rate": 1.2133082620067193e-06, "loss": 0.321, "step": 7219 }, { "epoch": 1.36, "learning_rate": 1.212652509928611e-06, "loss": 0.2022, "step": 7220 }, { "epoch": 1.36, "learning_rate": 1.2119968783607034e-06, "loss": 0.2045, "step": 7221 }, { "epoch": 1.36, "learning_rate": 1.2113413673643734e-06, "loss": 0.237, "step": 7222 }, { "epoch": 1.36, "learning_rate": 1.2106859770009819e-06, "loss": 0.1642, "step": 7223 }, { "epoch": 1.36, "learning_rate": 1.2100307073318804e-06, "loss": 0.1941, "step": 7224 }, { "epoch": 1.37, "learning_rate": 1.2093755584184125e-06, "loss": 0.2463, "step": 7225 }, { "epoch": 1.37, "learning_rate": 1.2087205303219046e-06, "loss": 0.2683, "step": 7226 }, { "epoch": 1.37, "learning_rate": 1.208065623103675e-06, "loss": 0.2364, "step": 7227 }, { "epoch": 1.37, "learning_rate": 1.2074108368250325e-06, "loss": 0.1751, "step": 7228 }, { "epoch": 1.37, "learning_rate": 1.2067561715472705e-06, "loss": 0.2216, "step": 7229 }, { "epoch": 1.37, "learning_rate": 1.2061016273316728e-06, "loss": 0.2622, "step": 7230 }, { "epoch": 1.37, "learning_rate": 1.2054472042395143e-06, "loss": 0.2407, "step": 7231 }, { "epoch": 1.37, "learning_rate": 1.2047929023320534e-06, "loss": 0.154, "step": 7232 }, { "epoch": 1.37, "learning_rate": 1.2041387216705424e-06, "loss": 0.2925, "step": 7233 }, { "epoch": 1.37, "learning_rate": 1.2034846623162203e-06, "loss": 0.1803, "step": 7234 }, { "epoch": 1.37, "learning_rate": 1.2028307243303113e-06, "loss": 0.2674, "step": 7235 }, { "epoch": 1.37, "learning_rate": 1.2021769077740345e-06, "loss": 0.2646, "step": 7236 }, { "epoch": 1.37, "learning_rate": 1.2015232127085932e-06, "loss": 0.2199, "step": 7237 }, { "epoch": 1.37, "learning_rate": 1.2008696391951806e-06, "loss": 0.2401, "step": 7238 }, { "epoch": 1.37, "learning_rate": 1.2002161872949791e-06, "loss": 0.1629, "step": 7239 }, { "epoch": 1.37, "learning_rate": 1.1995628570691587e-06, "loss": 0.2711, "step": 7240 }, { "epoch": 1.37, "learning_rate": 1.1989096485788781e-06, "loss": 0.2003, "step": 7241 }, { "epoch": 1.37, "learning_rate": 1.198256561885286e-06, "loss": 0.2088, "step": 7242 }, { "epoch": 1.37, "learning_rate": 1.1976035970495175e-06, "loss": 0.2643, "step": 7243 }, { "epoch": 1.37, "learning_rate": 1.196950754132698e-06, "loss": 0.2187, "step": 7244 }, { "epoch": 1.37, "learning_rate": 1.1962980331959412e-06, "loss": 0.2563, "step": 7245 }, { "epoch": 1.37, "learning_rate": 1.1956454343003485e-06, "loss": 0.262, "step": 7246 }, { "epoch": 1.37, "learning_rate": 1.194992957507011e-06, "loss": 0.2244, "step": 7247 }, { "epoch": 1.37, "learning_rate": 1.1943406028770078e-06, "loss": 0.2427, "step": 7248 }, { "epoch": 1.37, "learning_rate": 1.1936883704714063e-06, "loss": 0.2394, "step": 7249 }, { "epoch": 1.37, "learning_rate": 1.1930362603512632e-06, "loss": 0.1294, "step": 7250 }, { "epoch": 1.37, "learning_rate": 1.192384272577623e-06, "loss": 0.2998, "step": 7251 }, { "epoch": 1.37, "learning_rate": 1.1917324072115182e-06, "loss": 0.1838, "step": 7252 }, { "epoch": 1.37, "learning_rate": 1.1910806643139738e-06, "loss": 0.223, "step": 7253 }, { "epoch": 1.37, "learning_rate": 1.1904290439459974e-06, "loss": 0.3269, "step": 7254 }, { "epoch": 1.37, "learning_rate": 1.1897775461685878e-06, "loss": 0.2282, "step": 7255 }, { "epoch": 1.37, "learning_rate": 1.1891261710427355e-06, "loss": 0.1828, "step": 7256 }, { "epoch": 1.37, "learning_rate": 1.1884749186294135e-06, "loss": 0.2265, "step": 7257 }, { "epoch": 1.37, "learning_rate": 1.1878237889895862e-06, "loss": 0.2937, "step": 7258 }, { "epoch": 1.37, "learning_rate": 1.1871727821842102e-06, "loss": 0.2394, "step": 7259 }, { "epoch": 1.37, "learning_rate": 1.1865218982742235e-06, "loss": 0.2698, "step": 7260 }, { "epoch": 1.37, "learning_rate": 1.1858711373205566e-06, "loss": 0.2206, "step": 7261 }, { "epoch": 1.37, "learning_rate": 1.185220499384131e-06, "loss": 0.2253, "step": 7262 }, { "epoch": 1.37, "learning_rate": 1.1845699845258493e-06, "loss": 0.2251, "step": 7263 }, { "epoch": 1.37, "learning_rate": 1.1839195928066101e-06, "loss": 0.2246, "step": 7264 }, { "epoch": 1.37, "learning_rate": 1.183269324287298e-06, "loss": 0.2375, "step": 7265 }, { "epoch": 1.37, "learning_rate": 1.1826191790287817e-06, "loss": 0.2336, "step": 7266 }, { "epoch": 1.37, "learning_rate": 1.1819691570919256e-06, "loss": 0.2401, "step": 7267 }, { "epoch": 1.37, "learning_rate": 1.1813192585375786e-06, "loss": 0.2272, "step": 7268 }, { "epoch": 1.37, "learning_rate": 1.1806694834265759e-06, "loss": 0.2236, "step": 7269 }, { "epoch": 1.37, "learning_rate": 1.1800198318197468e-06, "loss": 0.2009, "step": 7270 }, { "epoch": 1.37, "learning_rate": 1.1793703037779056e-06, "loss": 0.2627, "step": 7271 }, { "epoch": 1.37, "learning_rate": 1.1787208993618526e-06, "loss": 0.2367, "step": 7272 }, { "epoch": 1.37, "learning_rate": 1.1780716186323827e-06, "loss": 0.1974, "step": 7273 }, { "epoch": 1.37, "learning_rate": 1.1774224616502744e-06, "loss": 0.2274, "step": 7274 }, { "epoch": 1.37, "learning_rate": 1.1767734284762964e-06, "loss": 0.2244, "step": 7275 }, { "epoch": 1.37, "learning_rate": 1.176124519171205e-06, "loss": 0.2487, "step": 7276 }, { "epoch": 1.37, "learning_rate": 1.1754757337957461e-06, "loss": 0.2265, "step": 7277 }, { "epoch": 1.38, "learning_rate": 1.1748270724106531e-06, "loss": 0.2286, "step": 7278 }, { "epoch": 1.38, "learning_rate": 1.1741785350766477e-06, "loss": 0.24, "step": 7279 }, { "epoch": 1.38, "learning_rate": 1.1735301218544406e-06, "loss": 0.2139, "step": 7280 }, { "epoch": 1.38, "learning_rate": 1.1728818328047306e-06, "loss": 0.2454, "step": 7281 }, { "epoch": 1.38, "learning_rate": 1.1722336679882046e-06, "loss": 0.212, "step": 7282 }, { "epoch": 1.38, "learning_rate": 1.1715856274655382e-06, "loss": 0.2093, "step": 7283 }, { "epoch": 1.38, "learning_rate": 1.1709377112973952e-06, "loss": 0.294, "step": 7284 }, { "epoch": 1.38, "learning_rate": 1.170289919544428e-06, "loss": 0.2581, "step": 7285 }, { "epoch": 1.38, "learning_rate": 1.1696422522672772e-06, "loss": 0.2147, "step": 7286 }, { "epoch": 1.38, "learning_rate": 1.1689947095265717e-06, "loss": 0.2176, "step": 7287 }, { "epoch": 1.38, "learning_rate": 1.1683472913829286e-06, "loss": 0.2482, "step": 7288 }, { "epoch": 1.38, "learning_rate": 1.1676999978969538e-06, "loss": 0.2118, "step": 7289 }, { "epoch": 1.38, "learning_rate": 1.167052829129241e-06, "loss": 0.1801, "step": 7290 }, { "epoch": 1.38, "learning_rate": 1.1664057851403727e-06, "loss": 0.2263, "step": 7291 }, { "epoch": 1.38, "learning_rate": 1.1657588659909194e-06, "loss": 0.25, "step": 7292 }, { "epoch": 1.38, "learning_rate": 1.1651120717414398e-06, "loss": 0.235, "step": 7293 }, { "epoch": 1.38, "learning_rate": 1.1644654024524807e-06, "loss": 0.221, "step": 7294 }, { "epoch": 1.38, "learning_rate": 1.16381885818458e-06, "loss": 0.2006, "step": 7295 }, { "epoch": 1.38, "learning_rate": 1.1631724389982587e-06, "loss": 0.1882, "step": 7296 }, { "epoch": 1.38, "learning_rate": 1.162526144954029e-06, "loss": 0.1711, "step": 7297 }, { "epoch": 1.38, "learning_rate": 1.1618799761123941e-06, "loss": 0.2416, "step": 7298 }, { "epoch": 1.38, "learning_rate": 1.16123393253384e-06, "loss": 0.1767, "step": 7299 }, { "epoch": 1.38, "learning_rate": 1.1605880142788436e-06, "loss": 0.2655, "step": 7300 }, { "epoch": 1.38, "learning_rate": 1.1599422214078718e-06, "loss": 0.3163, "step": 7301 }, { "epoch": 1.38, "learning_rate": 1.159296553981378e-06, "loss": 0.1833, "step": 7302 }, { "epoch": 1.38, "learning_rate": 1.1586510120598016e-06, "loss": 0.2089, "step": 7303 }, { "epoch": 1.38, "learning_rate": 1.1580055957035748e-06, "loss": 0.2553, "step": 7304 }, { "epoch": 1.38, "learning_rate": 1.1573603049731156e-06, "loss": 0.1767, "step": 7305 }, { "epoch": 1.38, "learning_rate": 1.1567151399288295e-06, "loss": 0.2109, "step": 7306 }, { "epoch": 1.38, "learning_rate": 1.156070100631112e-06, "loss": 0.2623, "step": 7307 }, { "epoch": 1.38, "learning_rate": 1.155425187140346e-06, "loss": 0.2476, "step": 7308 }, { "epoch": 1.38, "learning_rate": 1.1547803995169022e-06, "loss": 0.3024, "step": 7309 }, { "epoch": 1.38, "learning_rate": 1.1541357378211402e-06, "loss": 0.2733, "step": 7310 }, { "epoch": 1.38, "learning_rate": 1.1534912021134077e-06, "loss": 0.2919, "step": 7311 }, { "epoch": 1.38, "learning_rate": 1.1528467924540405e-06, "loss": 0.2563, "step": 7312 }, { "epoch": 1.38, "learning_rate": 1.1522025089033623e-06, "loss": 0.1521, "step": 7313 }, { "epoch": 1.38, "learning_rate": 1.1515583515216854e-06, "loss": 0.2575, "step": 7314 }, { "epoch": 1.38, "learning_rate": 1.1509143203693105e-06, "loss": 0.3258, "step": 7315 }, { "epoch": 1.38, "learning_rate": 1.1502704155065257e-06, "loss": 0.2064, "step": 7316 }, { "epoch": 1.38, "learning_rate": 1.149626636993608e-06, "loss": 0.2482, "step": 7317 }, { "epoch": 1.38, "learning_rate": 1.148982984890822e-06, "loss": 0.2979, "step": 7318 }, { "epoch": 1.38, "learning_rate": 1.1483394592584212e-06, "loss": 0.3002, "step": 7319 }, { "epoch": 1.38, "learning_rate": 1.1476960601566466e-06, "loss": 0.2417, "step": 7320 }, { "epoch": 1.38, "learning_rate": 1.1470527876457279e-06, "loss": 0.3204, "step": 7321 }, { "epoch": 1.38, "learning_rate": 1.1464096417858821e-06, "loss": 0.2064, "step": 7322 }, { "epoch": 1.38, "learning_rate": 1.1457666226373154e-06, "loss": 0.2407, "step": 7323 }, { "epoch": 1.38, "learning_rate": 1.1451237302602214e-06, "loss": 0.1767, "step": 7324 }, { "epoch": 1.38, "learning_rate": 1.1444809647147808e-06, "loss": 0.2746, "step": 7325 }, { "epoch": 1.38, "learning_rate": 1.143838326061167e-06, "loss": 0.3605, "step": 7326 }, { "epoch": 1.38, "learning_rate": 1.1431958143595352e-06, "loss": 0.1946, "step": 7327 }, { "epoch": 1.38, "learning_rate": 1.1425534296700317e-06, "loss": 0.3151, "step": 7328 }, { "epoch": 1.38, "learning_rate": 1.1419111720527938e-06, "loss": 0.2008, "step": 7329 }, { "epoch": 1.38, "learning_rate": 1.1412690415679411e-06, "loss": 0.1718, "step": 7330 }, { "epoch": 1.39, "learning_rate": 1.1406270382755843e-06, "loss": 0.2, "step": 7331 }, { "epoch": 1.39, "learning_rate": 1.1399851622358253e-06, "loss": 0.2211, "step": 7332 }, { "epoch": 1.39, "learning_rate": 1.1393434135087471e-06, "loss": 0.3209, "step": 7333 }, { "epoch": 1.39, "learning_rate": 1.1387017921544257e-06, "loss": 0.3059, "step": 7334 }, { "epoch": 1.39, "learning_rate": 1.1380602982329264e-06, "loss": 0.243, "step": 7335 }, { "epoch": 1.39, "learning_rate": 1.1374189318042966e-06, "loss": 0.1773, "step": 7336 }, { "epoch": 1.39, "learning_rate": 1.1367776929285781e-06, "loss": 0.2703, "step": 7337 }, { "epoch": 1.39, "learning_rate": 1.1361365816657982e-06, "loss": 0.2621, "step": 7338 }, { "epoch": 1.39, "learning_rate": 1.135495598075969e-06, "loss": 0.2034, "step": 7339 }, { "epoch": 1.39, "learning_rate": 1.1348547422190972e-06, "loss": 0.2923, "step": 7340 }, { "epoch": 1.39, "learning_rate": 1.1342140141551736e-06, "loss": 0.2116, "step": 7341 }, { "epoch": 1.39, "learning_rate": 1.1335734139441748e-06, "loss": 0.2968, "step": 7342 }, { "epoch": 1.39, "learning_rate": 1.132932941646071e-06, "loss": 0.2241, "step": 7343 }, { "epoch": 1.39, "learning_rate": 1.1322925973208176e-06, "loss": 0.2698, "step": 7344 }, { "epoch": 1.39, "learning_rate": 1.1316523810283555e-06, "loss": 0.1465, "step": 7345 }, { "epoch": 1.39, "learning_rate": 1.1310122928286188e-06, "loss": 0.2546, "step": 7346 }, { "epoch": 1.39, "learning_rate": 1.130372332781526e-06, "loss": 0.2656, "step": 7347 }, { "epoch": 1.39, "learning_rate": 1.1297325009469844e-06, "loss": 0.1989, "step": 7348 }, { "epoch": 1.39, "learning_rate": 1.1290927973848898e-06, "loss": 0.2508, "step": 7349 }, { "epoch": 1.39, "learning_rate": 1.1284532221551254e-06, "loss": 0.1855, "step": 7350 }, { "epoch": 1.39, "learning_rate": 1.1278137753175625e-06, "loss": 0.233, "step": 7351 }, { "epoch": 1.39, "learning_rate": 1.1271744569320605e-06, "loss": 0.2114, "step": 7352 }, { "epoch": 1.39, "learning_rate": 1.1265352670584674e-06, "loss": 0.315, "step": 7353 }, { "epoch": 1.39, "learning_rate": 1.1258962057566179e-06, "loss": 0.3216, "step": 7354 }, { "epoch": 1.39, "learning_rate": 1.1252572730863357e-06, "loss": 0.2878, "step": 7355 }, { "epoch": 1.39, "learning_rate": 1.1246184691074317e-06, "loss": 0.2045, "step": 7356 }, { "epoch": 1.39, "learning_rate": 1.1239797938797054e-06, "loss": 0.2508, "step": 7357 }, { "epoch": 1.39, "learning_rate": 1.123341247462944e-06, "loss": 0.2341, "step": 7358 }, { "epoch": 1.39, "learning_rate": 1.122702829916923e-06, "loss": 0.212, "step": 7359 }, { "epoch": 1.39, "learning_rate": 1.1220645413014044e-06, "loss": 0.2137, "step": 7360 }, { "epoch": 1.39, "learning_rate": 1.1214263816761403e-06, "loss": 0.3032, "step": 7361 }, { "epoch": 1.39, "learning_rate": 1.120788351100868e-06, "loss": 0.2005, "step": 7362 }, { "epoch": 1.39, "learning_rate": 1.1201504496353175e-06, "loss": 0.2194, "step": 7363 }, { "epoch": 1.39, "learning_rate": 1.1195126773392005e-06, "loss": 0.2182, "step": 7364 }, { "epoch": 1.39, "learning_rate": 1.11887503427222e-06, "loss": 0.272, "step": 7365 }, { "epoch": 1.39, "learning_rate": 1.1182375204940687e-06, "loss": 0.2752, "step": 7366 }, { "epoch": 1.39, "learning_rate": 1.1176001360644221e-06, "loss": 0.2397, "step": 7367 }, { "epoch": 1.39, "learning_rate": 1.116962881042949e-06, "loss": 0.2159, "step": 7368 }, { "epoch": 1.39, "learning_rate": 1.1163257554893034e-06, "loss": 0.2637, "step": 7369 }, { "epoch": 1.39, "learning_rate": 1.115688759463125e-06, "loss": 0.2521, "step": 7370 }, { "epoch": 1.39, "learning_rate": 1.1150518930240464e-06, "loss": 0.222, "step": 7371 }, { "epoch": 1.39, "learning_rate": 1.1144151562316855e-06, "loss": 0.2546, "step": 7372 }, { "epoch": 1.39, "learning_rate": 1.1137785491456455e-06, "loss": 0.2189, "step": 7373 }, { "epoch": 1.39, "learning_rate": 1.113142071825522e-06, "loss": 0.2384, "step": 7374 }, { "epoch": 1.39, "learning_rate": 1.1125057243308974e-06, "loss": 0.2471, "step": 7375 }, { "epoch": 1.39, "learning_rate": 1.1118695067213375e-06, "loss": 0.1945, "step": 7376 }, { "epoch": 1.39, "learning_rate": 1.1112334190564025e-06, "loss": 0.2967, "step": 7377 }, { "epoch": 1.39, "learning_rate": 1.1105974613956364e-06, "loss": 0.2361, "step": 7378 }, { "epoch": 1.39, "learning_rate": 1.1099616337985719e-06, "loss": 0.2267, "step": 7379 }, { "epoch": 1.39, "learning_rate": 1.10932593632473e-06, "loss": 0.2389, "step": 7380 }, { "epoch": 1.39, "learning_rate": 1.1086903690336185e-06, "loss": 0.2083, "step": 7381 }, { "epoch": 1.39, "learning_rate": 1.108054931984734e-06, "loss": 0.2329, "step": 7382 }, { "epoch": 1.39, "learning_rate": 1.1074196252375605e-06, "loss": 0.2768, "step": 7383 }, { "epoch": 1.4, "learning_rate": 1.10678444885157e-06, "loss": 0.2397, "step": 7384 }, { "epoch": 1.4, "learning_rate": 1.106149402886222e-06, "loss": 0.257, "step": 7385 }, { "epoch": 1.4, "learning_rate": 1.1055144874009638e-06, "loss": 0.1998, "step": 7386 }, { "epoch": 1.4, "learning_rate": 1.1048797024552307e-06, "loss": 0.2715, "step": 7387 }, { "epoch": 1.4, "learning_rate": 1.104245048108446e-06, "loss": 0.2889, "step": 7388 }, { "epoch": 1.4, "learning_rate": 1.1036105244200205e-06, "loss": 0.2371, "step": 7389 }, { "epoch": 1.4, "learning_rate": 1.102976131449352e-06, "loss": 0.2062, "step": 7390 }, { "epoch": 1.4, "learning_rate": 1.1023418692558276e-06, "loss": 0.2344, "step": 7391 }, { "epoch": 1.4, "learning_rate": 1.1017077378988214e-06, "loss": 0.259, "step": 7392 }, { "epoch": 1.4, "learning_rate": 1.1010737374376946e-06, "loss": 0.2588, "step": 7393 }, { "epoch": 1.4, "learning_rate": 1.1004398679317974e-06, "loss": 0.2742, "step": 7394 }, { "epoch": 1.4, "learning_rate": 1.0998061294404657e-06, "loss": 0.1914, "step": 7395 }, { "epoch": 1.4, "learning_rate": 1.0991725220230275e-06, "loss": 0.2442, "step": 7396 }, { "epoch": 1.4, "learning_rate": 1.098539045738793e-06, "loss": 0.2651, "step": 7397 }, { "epoch": 1.4, "learning_rate": 1.0979057006470625e-06, "loss": 0.2567, "step": 7398 }, { "epoch": 1.4, "learning_rate": 1.0972724868071269e-06, "loss": 0.2866, "step": 7399 }, { "epoch": 1.4, "learning_rate": 1.0966394042782597e-06, "loss": 0.2265, "step": 7400 }, { "epoch": 1.4, "learning_rate": 1.0960064531197243e-06, "loss": 0.2884, "step": 7401 }, { "epoch": 1.4, "learning_rate": 1.095373633390775e-06, "loss": 0.2566, "step": 7402 }, { "epoch": 1.4, "learning_rate": 1.0947409451506476e-06, "loss": 0.1994, "step": 7403 }, { "epoch": 1.4, "learning_rate": 1.0941083884585696e-06, "loss": 0.2412, "step": 7404 }, { "epoch": 1.4, "learning_rate": 1.0934759633737575e-06, "loss": 0.1951, "step": 7405 }, { "epoch": 1.4, "learning_rate": 1.0928436699554098e-06, "loss": 0.3277, "step": 7406 }, { "epoch": 1.4, "learning_rate": 1.0922115082627197e-06, "loss": 0.2514, "step": 7407 }, { "epoch": 1.4, "learning_rate": 1.0915794783548636e-06, "loss": 0.2274, "step": 7408 }, { "epoch": 1.4, "learning_rate": 1.0909475802910048e-06, "loss": 0.2296, "step": 7409 }, { "epoch": 1.4, "learning_rate": 1.0903158141302984e-06, "loss": 0.1969, "step": 7410 }, { "epoch": 1.4, "learning_rate": 1.0896841799318847e-06, "loss": 0.2447, "step": 7411 }, { "epoch": 1.4, "learning_rate": 1.089052677754889e-06, "loss": 0.2189, "step": 7412 }, { "epoch": 1.4, "learning_rate": 1.0884213076584302e-06, "loss": 0.303, "step": 7413 }, { "epoch": 1.4, "learning_rate": 1.087790069701611e-06, "loss": 0.2194, "step": 7414 }, { "epoch": 1.4, "learning_rate": 1.0871589639435204e-06, "loss": 0.2694, "step": 7415 }, { "epoch": 1.4, "learning_rate": 1.086527990443239e-06, "loss": 0.2879, "step": 7416 }, { "epoch": 1.4, "learning_rate": 1.0858971492598328e-06, "loss": 0.2544, "step": 7417 }, { "epoch": 1.4, "learning_rate": 1.0852664404523546e-06, "loss": 0.2608, "step": 7418 }, { "epoch": 1.4, "learning_rate": 1.084635864079847e-06, "loss": 0.2048, "step": 7419 }, { "epoch": 1.4, "learning_rate": 1.0840054202013383e-06, "loss": 0.352, "step": 7420 }, { "epoch": 1.4, "learning_rate": 1.0833751088758454e-06, "loss": 0.2658, "step": 7421 }, { "epoch": 1.4, "learning_rate": 1.0827449301623728e-06, "loss": 0.2464, "step": 7422 }, { "epoch": 1.4, "learning_rate": 1.0821148841199117e-06, "loss": 0.189, "step": 7423 }, { "epoch": 1.4, "learning_rate": 1.0814849708074416e-06, "loss": 0.1927, "step": 7424 }, { "epoch": 1.4, "learning_rate": 1.0808551902839299e-06, "loss": 0.2387, "step": 7425 }, { "epoch": 1.4, "learning_rate": 1.08022554260833e-06, "loss": 0.225, "step": 7426 }, { "epoch": 1.4, "learning_rate": 1.0795960278395868e-06, "loss": 0.2059, "step": 7427 }, { "epoch": 1.4, "learning_rate": 1.0789666460366269e-06, "loss": 0.2836, "step": 7428 }, { "epoch": 1.4, "learning_rate": 1.0783373972583676e-06, "loss": 0.1638, "step": 7429 }, { "epoch": 1.4, "learning_rate": 1.077708281563717e-06, "loss": 0.2542, "step": 7430 }, { "epoch": 1.4, "learning_rate": 1.0770792990115641e-06, "loss": 0.261, "step": 7431 }, { "epoch": 1.4, "learning_rate": 1.076450449660789e-06, "loss": 0.2209, "step": 7432 }, { "epoch": 1.4, "learning_rate": 1.0758217335702614e-06, "loss": 0.219, "step": 7433 }, { "epoch": 1.4, "learning_rate": 1.0751931507988338e-06, "loss": 0.2253, "step": 7434 }, { "epoch": 1.4, "learning_rate": 1.0745647014053487e-06, "loss": 0.2339, "step": 7435 }, { "epoch": 1.4, "learning_rate": 1.0739363854486384e-06, "loss": 0.225, "step": 7436 }, { "epoch": 1.41, "learning_rate": 1.0733082029875172e-06, "loss": 0.2721, "step": 7437 }, { "epoch": 1.41, "learning_rate": 1.0726801540807925e-06, "loss": 0.3005, "step": 7438 }, { "epoch": 1.41, "learning_rate": 1.0720522387872564e-06, "loss": 0.3139, "step": 7439 }, { "epoch": 1.41, "learning_rate": 1.0714244571656865e-06, "loss": 0.2713, "step": 7440 }, { "epoch": 1.41, "learning_rate": 1.070796809274853e-06, "loss": 0.2605, "step": 7441 }, { "epoch": 1.41, "learning_rate": 1.0701692951735107e-06, "loss": 0.2271, "step": 7442 }, { "epoch": 1.41, "learning_rate": 1.0695419149203992e-06, "loss": 0.2814, "step": 7443 }, { "epoch": 1.41, "learning_rate": 1.068914668574251e-06, "loss": 0.2005, "step": 7444 }, { "epoch": 1.41, "learning_rate": 1.0682875561937835e-06, "loss": 0.2977, "step": 7445 }, { "epoch": 1.41, "learning_rate": 1.067660577837699e-06, "loss": 0.2311, "step": 7446 }, { "epoch": 1.41, "learning_rate": 1.067033733564692e-06, "loss": 0.3697, "step": 7447 }, { "epoch": 1.41, "learning_rate": 1.0664070234334416e-06, "loss": 0.2571, "step": 7448 }, { "epoch": 1.41, "learning_rate": 1.0657804475026149e-06, "loss": 0.2184, "step": 7449 }, { "epoch": 1.41, "learning_rate": 1.065154005830866e-06, "loss": 0.2027, "step": 7450 }, { "epoch": 1.41, "learning_rate": 1.0645276984768374e-06, "loss": 0.225, "step": 7451 }, { "epoch": 1.41, "learning_rate": 1.0639015254991582e-06, "loss": 0.2183, "step": 7452 }, { "epoch": 1.41, "learning_rate": 1.0632754869564457e-06, "loss": 0.2528, "step": 7453 }, { "epoch": 1.41, "learning_rate": 1.0626495829073037e-06, "loss": 0.2719, "step": 7454 }, { "epoch": 1.41, "learning_rate": 1.062023813410324e-06, "loss": 0.2141, "step": 7455 }, { "epoch": 1.41, "learning_rate": 1.0613981785240854e-06, "loss": 0.3141, "step": 7456 }, { "epoch": 1.41, "learning_rate": 1.0607726783071545e-06, "loss": 0.2352, "step": 7457 }, { "epoch": 1.41, "learning_rate": 1.0601473128180855e-06, "loss": 0.2395, "step": 7458 }, { "epoch": 1.41, "learning_rate": 1.0595220821154195e-06, "loss": 0.2405, "step": 7459 }, { "epoch": 1.41, "learning_rate": 1.0588969862576848e-06, "loss": 0.2708, "step": 7460 }, { "epoch": 1.41, "learning_rate": 1.0582720253033978e-06, "loss": 0.1857, "step": 7461 }, { "epoch": 1.41, "learning_rate": 1.0576471993110615e-06, "loss": 0.2251, "step": 7462 }, { "epoch": 1.41, "learning_rate": 1.0570225083391671e-06, "loss": 0.2396, "step": 7463 }, { "epoch": 1.41, "learning_rate": 1.0563979524461923e-06, "loss": 0.2501, "step": 7464 }, { "epoch": 1.41, "learning_rate": 1.0557735316906026e-06, "loss": 0.2643, "step": 7465 }, { "epoch": 1.41, "learning_rate": 1.055149246130851e-06, "loss": 0.3175, "step": 7466 }, { "epoch": 1.41, "learning_rate": 1.0545250958253774e-06, "loss": 0.2095, "step": 7467 }, { "epoch": 1.41, "learning_rate": 1.0539010808326087e-06, "loss": 0.2046, "step": 7468 }, { "epoch": 1.41, "learning_rate": 1.0532772012109622e-06, "loss": 0.2179, "step": 7469 }, { "epoch": 1.41, "learning_rate": 1.0526534570188373e-06, "loss": 0.1993, "step": 7470 }, { "epoch": 1.41, "learning_rate": 1.0520298483146235e-06, "loss": 0.2465, "step": 7471 }, { "epoch": 1.41, "learning_rate": 1.0514063751567006e-06, "loss": 0.1741, "step": 7472 }, { "epoch": 1.41, "learning_rate": 1.0507830376034294e-06, "loss": 0.1812, "step": 7473 }, { "epoch": 1.41, "learning_rate": 1.050159835713162e-06, "loss": 0.2098, "step": 7474 }, { "epoch": 1.41, "learning_rate": 1.0495367695442394e-06, "loss": 0.2432, "step": 7475 }, { "epoch": 1.41, "learning_rate": 1.048913839154985e-06, "loss": 0.3094, "step": 7476 }, { "epoch": 1.41, "learning_rate": 1.048291044603712e-06, "loss": 0.1966, "step": 7477 }, { "epoch": 1.41, "learning_rate": 1.0476683859487236e-06, "loss": 0.2529, "step": 7478 }, { "epoch": 1.41, "learning_rate": 1.047045863248304e-06, "loss": 0.1607, "step": 7479 }, { "epoch": 1.41, "learning_rate": 1.0464234765607317e-06, "loss": 0.1679, "step": 7480 }, { "epoch": 1.41, "learning_rate": 1.0458012259442682e-06, "loss": 0.2552, "step": 7481 }, { "epoch": 1.41, "learning_rate": 1.0451791114571611e-06, "loss": 0.2179, "step": 7482 }, { "epoch": 1.41, "learning_rate": 1.0445571331576495e-06, "loss": 0.2485, "step": 7483 }, { "epoch": 1.41, "learning_rate": 1.0439352911039583e-06, "loss": 0.2604, "step": 7484 }, { "epoch": 1.41, "learning_rate": 1.0433135853542953e-06, "loss": 0.2664, "step": 7485 }, { "epoch": 1.41, "learning_rate": 1.0426920159668624e-06, "loss": 0.3404, "step": 7486 }, { "epoch": 1.41, "learning_rate": 1.0420705829998446e-06, "loss": 0.2542, "step": 7487 }, { "epoch": 1.41, "learning_rate": 1.0414492865114145e-06, "loss": 0.3214, "step": 7488 }, { "epoch": 1.41, "learning_rate": 1.0408281265597332e-06, "loss": 0.2788, "step": 7489 }, { "epoch": 1.42, "learning_rate": 1.0402071032029477e-06, "loss": 0.2007, "step": 7490 }, { "epoch": 1.42, "learning_rate": 1.0395862164991929e-06, "loss": 0.2349, "step": 7491 }, { "epoch": 1.42, "learning_rate": 1.038965466506591e-06, "loss": 0.2189, "step": 7492 }, { "epoch": 1.42, "learning_rate": 1.038344853283251e-06, "loss": 0.2254, "step": 7493 }, { "epoch": 1.42, "learning_rate": 1.037724376887269e-06, "loss": 0.2219, "step": 7494 }, { "epoch": 1.42, "learning_rate": 1.037104037376729e-06, "loss": 0.2442, "step": 7495 }, { "epoch": 1.42, "learning_rate": 1.0364838348097016e-06, "loss": 0.2417, "step": 7496 }, { "epoch": 1.42, "learning_rate": 1.035863769244245e-06, "loss": 0.188, "step": 7497 }, { "epoch": 1.42, "learning_rate": 1.0352438407384038e-06, "loss": 0.2624, "step": 7498 }, { "epoch": 1.42, "learning_rate": 1.0346240493502097e-06, "loss": 0.2755, "step": 7499 }, { "epoch": 1.42, "learning_rate": 1.0340043951376847e-06, "loss": 0.2319, "step": 7500 }, { "epoch": 1.42, "learning_rate": 1.0333848781588326e-06, "loss": 0.2507, "step": 7501 }, { "epoch": 1.42, "learning_rate": 1.0327654984716475e-06, "loss": 0.3117, "step": 7502 }, { "epoch": 1.42, "learning_rate": 1.0321462561341128e-06, "loss": 0.2655, "step": 7503 }, { "epoch": 1.42, "learning_rate": 1.0315271512041941e-06, "loss": 0.2188, "step": 7504 }, { "epoch": 1.42, "learning_rate": 1.030908183739846e-06, "loss": 0.2459, "step": 7505 }, { "epoch": 1.42, "learning_rate": 1.0302893537990142e-06, "loss": 0.1684, "step": 7506 }, { "epoch": 1.42, "learning_rate": 1.0296706614396248e-06, "loss": 0.2292, "step": 7507 }, { "epoch": 1.42, "learning_rate": 1.0290521067195949e-06, "loss": 0.2998, "step": 7508 }, { "epoch": 1.42, "learning_rate": 1.0284336896968306e-06, "loss": 0.1964, "step": 7509 }, { "epoch": 1.42, "learning_rate": 1.027815410429219e-06, "loss": 0.2439, "step": 7510 }, { "epoch": 1.42, "learning_rate": 1.027197268974641e-06, "loss": 0.3021, "step": 7511 }, { "epoch": 1.42, "learning_rate": 1.0265792653909615e-06, "loss": 0.2054, "step": 7512 }, { "epoch": 1.42, "learning_rate": 1.0259613997360295e-06, "loss": 0.2061, "step": 7513 }, { "epoch": 1.42, "learning_rate": 1.0253436720676876e-06, "loss": 0.2613, "step": 7514 }, { "epoch": 1.42, "learning_rate": 1.0247260824437613e-06, "loss": 0.2745, "step": 7515 }, { "epoch": 1.42, "learning_rate": 1.0241086309220619e-06, "loss": 0.2762, "step": 7516 }, { "epoch": 1.42, "learning_rate": 1.0234913175603919e-06, "loss": 0.2409, "step": 7517 }, { "epoch": 1.42, "learning_rate": 1.0228741424165392e-06, "loss": 0.239, "step": 7518 }, { "epoch": 1.42, "learning_rate": 1.0222571055482752e-06, "loss": 0.1776, "step": 7519 }, { "epoch": 1.42, "learning_rate": 1.0216402070133648e-06, "loss": 0.2152, "step": 7520 }, { "epoch": 1.42, "learning_rate": 1.0210234468695553e-06, "loss": 0.2512, "step": 7521 }, { "epoch": 1.42, "learning_rate": 1.0204068251745825e-06, "loss": 0.2432, "step": 7522 }, { "epoch": 1.42, "learning_rate": 1.019790341986169e-06, "loss": 0.1934, "step": 7523 }, { "epoch": 1.42, "learning_rate": 1.019173997362025e-06, "loss": 0.2307, "step": 7524 }, { "epoch": 1.42, "learning_rate": 1.0185577913598469e-06, "loss": 0.2346, "step": 7525 }, { "epoch": 1.42, "learning_rate": 1.0179417240373181e-06, "loss": 0.2141, "step": 7526 }, { "epoch": 1.42, "learning_rate": 1.0173257954521105e-06, "loss": 0.2223, "step": 7527 }, { "epoch": 1.42, "learning_rate": 1.0167100056618812e-06, "loss": 0.1393, "step": 7528 }, { "epoch": 1.42, "learning_rate": 1.0160943547242752e-06, "loss": 0.1739, "step": 7529 }, { "epoch": 1.42, "learning_rate": 1.0154788426969243e-06, "loss": 0.2452, "step": 7530 }, { "epoch": 1.42, "learning_rate": 1.0148634696374473e-06, "loss": 0.2702, "step": 7531 }, { "epoch": 1.42, "learning_rate": 1.0142482356034503e-06, "loss": 0.2396, "step": 7532 }, { "epoch": 1.42, "learning_rate": 1.0136331406525257e-06, "loss": 0.2826, "step": 7533 }, { "epoch": 1.42, "learning_rate": 1.0130181848422538e-06, "loss": 0.2292, "step": 7534 }, { "epoch": 1.42, "learning_rate": 1.012403368230201e-06, "loss": 0.2652, "step": 7535 }, { "epoch": 1.42, "learning_rate": 1.0117886908739216e-06, "loss": 0.1915, "step": 7536 }, { "epoch": 1.42, "learning_rate": 1.0111741528309554e-06, "loss": 0.2485, "step": 7537 }, { "epoch": 1.42, "learning_rate": 1.01055975415883e-06, "loss": 0.1872, "step": 7538 }, { "epoch": 1.42, "learning_rate": 1.0099454949150623e-06, "loss": 0.2819, "step": 7539 }, { "epoch": 1.42, "learning_rate": 1.0093313751571513e-06, "loss": 0.2616, "step": 7540 }, { "epoch": 1.42, "learning_rate": 1.0087173949425854e-06, "loss": 0.1734, "step": 7541 }, { "epoch": 1.42, "learning_rate": 1.0081035543288429e-06, "loss": 0.2765, "step": 7542 }, { "epoch": 1.43, "learning_rate": 1.0074898533733834e-06, "loss": 0.1886, "step": 7543 }, { "epoch": 1.43, "learning_rate": 1.0068762921336563e-06, "loss": 0.2606, "step": 7544 }, { "epoch": 1.43, "learning_rate": 1.0062628706671005e-06, "loss": 0.2619, "step": 7545 }, { "epoch": 1.43, "learning_rate": 1.0056495890311362e-06, "loss": 0.183, "step": 7546 }, { "epoch": 1.43, "learning_rate": 1.0050364472831742e-06, "loss": 0.2299, "step": 7547 }, { "epoch": 1.43, "learning_rate": 1.0044234454806132e-06, "loss": 0.2735, "step": 7548 }, { "epoch": 1.43, "learning_rate": 1.003810583680834e-06, "loss": 0.1791, "step": 7549 }, { "epoch": 1.43, "learning_rate": 1.00319786194121e-06, "loss": 0.2267, "step": 7550 }, { "epoch": 1.43, "learning_rate": 1.0025852803190978e-06, "loss": 0.1932, "step": 7551 }, { "epoch": 1.43, "learning_rate": 1.001972838871842e-06, "loss": 0.2318, "step": 7552 }, { "epoch": 1.43, "learning_rate": 1.0013605376567743e-06, "loss": 0.2502, "step": 7553 }, { "epoch": 1.43, "learning_rate": 1.0007483767312124e-06, "loss": 0.2482, "step": 7554 }, { "epoch": 1.43, "learning_rate": 1.0001363561524618e-06, "loss": 0.3054, "step": 7555 }, { "epoch": 1.43, "learning_rate": 9.995244759778145e-07, "loss": 0.2992, "step": 7556 }, { "epoch": 1.43, "learning_rate": 9.989127362645495e-07, "loss": 0.1978, "step": 7557 }, { "epoch": 1.43, "learning_rate": 9.983011370699321e-07, "loss": 0.2062, "step": 7558 }, { "epoch": 1.43, "learning_rate": 9.976896784512153e-07, "loss": 0.2695, "step": 7559 }, { "epoch": 1.43, "learning_rate": 9.970783604656384e-07, "loss": 0.2, "step": 7560 }, { "epoch": 1.43, "learning_rate": 9.964671831704272e-07, "loss": 0.2622, "step": 7561 }, { "epoch": 1.43, "learning_rate": 9.958561466227956e-07, "loss": 0.2118, "step": 7562 }, { "epoch": 1.43, "learning_rate": 9.952452508799426e-07, "loss": 0.2618, "step": 7563 }, { "epoch": 1.43, "learning_rate": 9.946344959990555e-07, "loss": 0.2622, "step": 7564 }, { "epoch": 1.43, "learning_rate": 9.940238820373077e-07, "loss": 0.1867, "step": 7565 }, { "epoch": 1.43, "learning_rate": 9.934134090518593e-07, "loss": 0.232, "step": 7566 }, { "epoch": 1.43, "learning_rate": 9.928030770998579e-07, "loss": 0.2354, "step": 7567 }, { "epoch": 1.43, "learning_rate": 9.92192886238437e-07, "loss": 0.2364, "step": 7568 }, { "epoch": 1.43, "learning_rate": 9.915828365247165e-07, "loss": 0.2237, "step": 7569 }, { "epoch": 1.43, "learning_rate": 9.909729280158067e-07, "loss": 0.2405, "step": 7570 }, { "epoch": 1.43, "learning_rate": 9.903631607687994e-07, "loss": 0.2024, "step": 7571 }, { "epoch": 1.43, "learning_rate": 9.897535348407753e-07, "loss": 0.2277, "step": 7572 }, { "epoch": 1.43, "learning_rate": 9.89144050288805e-07, "loss": 0.2236, "step": 7573 }, { "epoch": 1.43, "learning_rate": 9.885347071699406e-07, "loss": 0.2187, "step": 7574 }, { "epoch": 1.43, "learning_rate": 9.879255055412235e-07, "loss": 0.2765, "step": 7575 }, { "epoch": 1.43, "learning_rate": 9.873164454596842e-07, "loss": 0.1433, "step": 7576 }, { "epoch": 1.43, "learning_rate": 9.867075269823354e-07, "loss": 0.2601, "step": 7577 }, { "epoch": 1.43, "learning_rate": 9.860987501661782e-07, "loss": 0.2545, "step": 7578 }, { "epoch": 1.43, "learning_rate": 9.854901150682037e-07, "loss": 0.2567, "step": 7579 }, { "epoch": 1.43, "learning_rate": 9.848816217453835e-07, "loss": 0.2411, "step": 7580 }, { "epoch": 1.43, "learning_rate": 9.842732702546823e-07, "loss": 0.2199, "step": 7581 }, { "epoch": 1.43, "learning_rate": 9.836650606530485e-07, "loss": 0.2423, "step": 7582 }, { "epoch": 1.43, "learning_rate": 9.830569929974146e-07, "loss": 0.2635, "step": 7583 }, { "epoch": 1.43, "learning_rate": 9.824490673447055e-07, "loss": 0.3547, "step": 7584 }, { "epoch": 1.43, "learning_rate": 9.818412837518296e-07, "loss": 0.2587, "step": 7585 }, { "epoch": 1.43, "learning_rate": 9.812336422756797e-07, "loss": 0.2969, "step": 7586 }, { "epoch": 1.43, "learning_rate": 9.806261429731404e-07, "loss": 0.2149, "step": 7587 }, { "epoch": 1.43, "learning_rate": 9.800187859010806e-07, "loss": 0.1967, "step": 7588 }, { "epoch": 1.43, "learning_rate": 9.794115711163531e-07, "loss": 0.2258, "step": 7589 }, { "epoch": 1.43, "learning_rate": 9.78804498675803e-07, "loss": 0.254, "step": 7590 }, { "epoch": 1.43, "learning_rate": 9.781975686362578e-07, "loss": 0.2299, "step": 7591 }, { "epoch": 1.43, "learning_rate": 9.77590781054533e-07, "loss": 0.2399, "step": 7592 }, { "epoch": 1.43, "learning_rate": 9.769841359874309e-07, "loss": 0.3267, "step": 7593 }, { "epoch": 1.43, "learning_rate": 9.763776334917398e-07, "loss": 0.2731, "step": 7594 }, { "epoch": 1.43, "learning_rate": 9.75771273624236e-07, "loss": 0.2221, "step": 7595 }, { "epoch": 1.44, "learning_rate": 9.751650564416812e-07, "loss": 0.2792, "step": 7596 }, { "epoch": 1.44, "learning_rate": 9.74558982000824e-07, "loss": 0.2455, "step": 7597 }, { "epoch": 1.44, "learning_rate": 9.739530503584e-07, "loss": 0.1887, "step": 7598 }, { "epoch": 1.44, "learning_rate": 9.73347261571131e-07, "loss": 0.3159, "step": 7599 }, { "epoch": 1.44, "learning_rate": 9.727416156957259e-07, "loss": 0.2302, "step": 7600 }, { "epoch": 1.44, "learning_rate": 9.721361127888798e-07, "loss": 0.2238, "step": 7601 }, { "epoch": 1.44, "learning_rate": 9.715307529072745e-07, "loss": 0.2252, "step": 7602 }, { "epoch": 1.44, "learning_rate": 9.709255361075787e-07, "loss": 0.1928, "step": 7603 }, { "epoch": 1.44, "learning_rate": 9.703204624464473e-07, "loss": 0.1994, "step": 7604 }, { "epoch": 1.44, "learning_rate": 9.697155319805223e-07, "loss": 0.2628, "step": 7605 }, { "epoch": 1.44, "learning_rate": 9.691107447664318e-07, "loss": 0.2608, "step": 7606 }, { "epoch": 1.44, "learning_rate": 9.685061008607907e-07, "loss": 0.2501, "step": 7607 }, { "epoch": 1.44, "learning_rate": 9.679016003202002e-07, "loss": 0.2131, "step": 7608 }, { "epoch": 1.44, "learning_rate": 9.67297243201249e-07, "loss": 0.3074, "step": 7609 }, { "epoch": 1.44, "learning_rate": 9.66693029560511e-07, "loss": 0.2068, "step": 7610 }, { "epoch": 1.44, "learning_rate": 9.66088959454547e-07, "loss": 0.2447, "step": 7611 }, { "epoch": 1.44, "learning_rate": 9.654850329399072e-07, "loss": 0.2774, "step": 7612 }, { "epoch": 1.44, "learning_rate": 9.648812500731231e-07, "loss": 0.1662, "step": 7613 }, { "epoch": 1.44, "learning_rate": 9.64277610910716e-07, "loss": 0.1597, "step": 7614 }, { "epoch": 1.44, "learning_rate": 9.636741155091947e-07, "loss": 0.2377, "step": 7615 }, { "epoch": 1.44, "learning_rate": 9.630707639250532e-07, "loss": 0.2391, "step": 7616 }, { "epoch": 1.44, "learning_rate": 9.624675562147698e-07, "loss": 0.2798, "step": 7617 }, { "epoch": 1.44, "learning_rate": 9.618644924348136e-07, "loss": 0.2679, "step": 7618 }, { "epoch": 1.44, "learning_rate": 9.612615726416382e-07, "loss": 0.262, "step": 7619 }, { "epoch": 1.44, "learning_rate": 9.606587968916817e-07, "loss": 0.3063, "step": 7620 }, { "epoch": 1.44, "learning_rate": 9.600561652413722e-07, "loss": 0.2672, "step": 7621 }, { "epoch": 1.44, "learning_rate": 9.59453677747123e-07, "loss": 0.2132, "step": 7622 }, { "epoch": 1.44, "learning_rate": 9.588513344653333e-07, "loss": 0.2771, "step": 7623 }, { "epoch": 1.44, "learning_rate": 9.582491354523887e-07, "loss": 0.2615, "step": 7624 }, { "epoch": 1.44, "learning_rate": 9.576470807646626e-07, "loss": 0.2331, "step": 7625 }, { "epoch": 1.44, "learning_rate": 9.570451704585138e-07, "loss": 0.2385, "step": 7626 }, { "epoch": 1.44, "learning_rate": 9.564434045902877e-07, "loss": 0.2055, "step": 7627 }, { "epoch": 1.44, "learning_rate": 9.558417832163164e-07, "loss": 0.3412, "step": 7628 }, { "epoch": 1.44, "learning_rate": 9.552403063929185e-07, "loss": 0.2203, "step": 7629 }, { "epoch": 1.44, "learning_rate": 9.546389741763991e-07, "loss": 0.2232, "step": 7630 }, { "epoch": 1.44, "learning_rate": 9.540377866230496e-07, "loss": 0.3133, "step": 7631 }, { "epoch": 1.44, "learning_rate": 9.534367437891478e-07, "loss": 0.2116, "step": 7632 }, { "epoch": 1.44, "learning_rate": 9.52835845730958e-07, "loss": 0.208, "step": 7633 }, { "epoch": 1.44, "learning_rate": 9.522350925047314e-07, "loss": 0.243, "step": 7634 }, { "epoch": 1.44, "learning_rate": 9.516344841667046e-07, "loss": 0.1911, "step": 7635 }, { "epoch": 1.44, "learning_rate": 9.510340207731023e-07, "loss": 0.2865, "step": 7636 }, { "epoch": 1.44, "learning_rate": 9.504337023801336e-07, "loss": 0.198, "step": 7637 }, { "epoch": 1.44, "learning_rate": 9.498335290439956e-07, "loss": 0.2879, "step": 7638 }, { "epoch": 1.44, "learning_rate": 9.492335008208711e-07, "loss": 0.2727, "step": 7639 }, { "epoch": 1.44, "learning_rate": 9.486336177669294e-07, "loss": 0.2448, "step": 7640 }, { "epoch": 1.44, "learning_rate": 9.480338799383265e-07, "loss": 0.2607, "step": 7641 }, { "epoch": 1.44, "learning_rate": 9.474342873912041e-07, "loss": 0.2131, "step": 7642 }, { "epoch": 1.44, "learning_rate": 9.468348401816923e-07, "loss": 0.336, "step": 7643 }, { "epoch": 1.44, "learning_rate": 9.462355383659047e-07, "loss": 0.1571, "step": 7644 }, { "epoch": 1.44, "learning_rate": 9.45636381999942e-07, "loss": 0.2282, "step": 7645 }, { "epoch": 1.44, "learning_rate": 9.450373711398947e-07, "loss": 0.276, "step": 7646 }, { "epoch": 1.44, "learning_rate": 9.444385058418345e-07, "loss": 0.1912, "step": 7647 }, { "epoch": 1.44, "learning_rate": 9.438397861618218e-07, "loss": 0.2934, "step": 7648 }, { "epoch": 1.45, "learning_rate": 9.432412121559062e-07, "loss": 0.1958, "step": 7649 }, { "epoch": 1.45, "learning_rate": 9.42642783880118e-07, "loss": 0.2417, "step": 7650 }, { "epoch": 1.45, "learning_rate": 9.420445013904772e-07, "loss": 0.25, "step": 7651 }, { "epoch": 1.45, "learning_rate": 9.414463647429925e-07, "loss": 0.2342, "step": 7652 }, { "epoch": 1.45, "learning_rate": 9.408483739936522e-07, "loss": 0.2324, "step": 7653 }, { "epoch": 1.45, "learning_rate": 9.402505291984381e-07, "loss": 0.219, "step": 7654 }, { "epoch": 1.45, "learning_rate": 9.396528304133149e-07, "loss": 0.2369, "step": 7655 }, { "epoch": 1.45, "learning_rate": 9.390552776942313e-07, "loss": 0.2396, "step": 7656 }, { "epoch": 1.45, "learning_rate": 9.384578710971276e-07, "loss": 0.2398, "step": 7657 }, { "epoch": 1.45, "learning_rate": 9.378606106779281e-07, "loss": 0.2413, "step": 7658 }, { "epoch": 1.45, "learning_rate": 9.372634964925401e-07, "loss": 0.2285, "step": 7659 }, { "epoch": 1.45, "learning_rate": 9.366665285968626e-07, "loss": 0.225, "step": 7660 }, { "epoch": 1.45, "learning_rate": 9.360697070467792e-07, "loss": 0.1727, "step": 7661 }, { "epoch": 1.45, "learning_rate": 9.354730318981561e-07, "loss": 0.2234, "step": 7662 }, { "epoch": 1.45, "learning_rate": 9.348765032068513e-07, "loss": 0.215, "step": 7663 }, { "epoch": 1.45, "learning_rate": 9.342801210287058e-07, "loss": 0.2545, "step": 7664 }, { "epoch": 1.45, "learning_rate": 9.33683885419548e-07, "loss": 0.2743, "step": 7665 }, { "epoch": 1.45, "learning_rate": 9.330877964351917e-07, "loss": 0.1731, "step": 7666 }, { "epoch": 1.45, "learning_rate": 9.324918541314382e-07, "loss": 0.265, "step": 7667 }, { "epoch": 1.45, "learning_rate": 9.318960585640738e-07, "loss": 0.2555, "step": 7668 }, { "epoch": 1.45, "learning_rate": 9.313004097888719e-07, "loss": 0.2706, "step": 7669 }, { "epoch": 1.45, "learning_rate": 9.307049078615918e-07, "loss": 0.2696, "step": 7670 }, { "epoch": 1.45, "learning_rate": 9.301095528379795e-07, "loss": 0.2308, "step": 7671 }, { "epoch": 1.45, "learning_rate": 9.295143447737667e-07, "loss": 0.2805, "step": 7672 }, { "epoch": 1.45, "learning_rate": 9.289192837246716e-07, "loss": 0.2833, "step": 7673 }, { "epoch": 1.45, "learning_rate": 9.283243697463987e-07, "loss": 0.2136, "step": 7674 }, { "epoch": 1.45, "learning_rate": 9.277296028946384e-07, "loss": 0.258, "step": 7675 }, { "epoch": 1.45, "learning_rate": 9.271349832250668e-07, "loss": 0.3103, "step": 7676 }, { "epoch": 1.45, "learning_rate": 9.265405107933495e-07, "loss": 0.2949, "step": 7677 }, { "epoch": 1.45, "learning_rate": 9.259461856551335e-07, "loss": 0.2044, "step": 7678 }, { "epoch": 1.45, "learning_rate": 9.253520078660541e-07, "loss": 0.2195, "step": 7679 }, { "epoch": 1.45, "learning_rate": 9.247579774817355e-07, "loss": 0.1965, "step": 7680 }, { "epoch": 1.45, "learning_rate": 9.241640945577834e-07, "loss": 0.2168, "step": 7681 }, { "epoch": 1.45, "learning_rate": 9.235703591497918e-07, "loss": 0.3555, "step": 7682 }, { "epoch": 1.45, "learning_rate": 9.229767713133433e-07, "loss": 0.2832, "step": 7683 }, { "epoch": 1.45, "learning_rate": 9.223833311040015e-07, "loss": 0.2666, "step": 7684 }, { "epoch": 1.45, "learning_rate": 9.21790038577321e-07, "loss": 0.2155, "step": 7685 }, { "epoch": 1.45, "learning_rate": 9.211968937888411e-07, "loss": 0.2373, "step": 7686 }, { "epoch": 1.45, "learning_rate": 9.206038967940839e-07, "loss": 0.2912, "step": 7687 }, { "epoch": 1.45, "learning_rate": 9.200110476485635e-07, "loss": 0.2551, "step": 7688 }, { "epoch": 1.45, "learning_rate": 9.194183464077771e-07, "loss": 0.2417, "step": 7689 }, { "epoch": 1.45, "learning_rate": 9.18825793127206e-07, "loss": 0.1716, "step": 7690 }, { "epoch": 1.45, "learning_rate": 9.182333878623217e-07, "loss": 0.195, "step": 7691 }, { "epoch": 1.45, "learning_rate": 9.176411306685804e-07, "loss": 0.221, "step": 7692 }, { "epoch": 1.45, "learning_rate": 9.170490216014216e-07, "loss": 0.2367, "step": 7693 }, { "epoch": 1.45, "learning_rate": 9.164570607162756e-07, "loss": 0.2032, "step": 7694 }, { "epoch": 1.45, "learning_rate": 9.158652480685559e-07, "loss": 0.2362, "step": 7695 }, { "epoch": 1.45, "learning_rate": 9.152735837136631e-07, "loss": 0.2387, "step": 7696 }, { "epoch": 1.45, "learning_rate": 9.146820677069831e-07, "loss": 0.1728, "step": 7697 }, { "epoch": 1.45, "learning_rate": 9.140907001038887e-07, "loss": 0.2345, "step": 7698 }, { "epoch": 1.45, "learning_rate": 9.134994809597389e-07, "loss": 0.2449, "step": 7699 }, { "epoch": 1.45, "learning_rate": 9.129084103298777e-07, "loss": 0.2164, "step": 7700 }, { "epoch": 1.45, "learning_rate": 9.123174882696367e-07, "loss": 0.2762, "step": 7701 }, { "epoch": 1.46, "learning_rate": 9.117267148343322e-07, "loss": 0.2937, "step": 7702 }, { "epoch": 1.46, "learning_rate": 9.111360900792679e-07, "loss": 0.1982, "step": 7703 }, { "epoch": 1.46, "learning_rate": 9.105456140597326e-07, "loss": 0.274, "step": 7704 }, { "epoch": 1.46, "learning_rate": 9.099552868310013e-07, "loss": 0.2623, "step": 7705 }, { "epoch": 1.46, "learning_rate": 9.093651084483357e-07, "loss": 0.1957, "step": 7706 }, { "epoch": 1.46, "learning_rate": 9.087750789669828e-07, "loss": 0.3014, "step": 7707 }, { "epoch": 1.46, "learning_rate": 9.081851984421761e-07, "loss": 0.2377, "step": 7708 }, { "epoch": 1.46, "learning_rate": 9.075954669291348e-07, "loss": 0.3308, "step": 7709 }, { "epoch": 1.46, "learning_rate": 9.070058844830648e-07, "loss": 0.1862, "step": 7710 }, { "epoch": 1.46, "learning_rate": 9.064164511591575e-07, "loss": 0.1724, "step": 7711 }, { "epoch": 1.46, "learning_rate": 9.058271670125898e-07, "loss": 0.3344, "step": 7712 }, { "epoch": 1.46, "learning_rate": 9.052380320985274e-07, "loss": 0.2444, "step": 7713 }, { "epoch": 1.46, "learning_rate": 9.046490464721178e-07, "loss": 0.1861, "step": 7714 }, { "epoch": 1.46, "learning_rate": 9.040602101884965e-07, "loss": 0.2368, "step": 7715 }, { "epoch": 1.46, "learning_rate": 9.03471523302788e-07, "loss": 0.2577, "step": 7716 }, { "epoch": 1.46, "learning_rate": 9.028829858700974e-07, "loss": 0.2076, "step": 7717 }, { "epoch": 1.46, "learning_rate": 9.02294597945518e-07, "loss": 0.2345, "step": 7718 }, { "epoch": 1.46, "learning_rate": 9.017063595841324e-07, "loss": 0.2429, "step": 7719 }, { "epoch": 1.46, "learning_rate": 9.01118270841004e-07, "loss": 0.2496, "step": 7720 }, { "epoch": 1.46, "learning_rate": 9.005303317711841e-07, "loss": 0.2385, "step": 7721 }, { "epoch": 1.46, "learning_rate": 8.999425424297132e-07, "loss": 0.3224, "step": 7722 }, { "epoch": 1.46, "learning_rate": 8.993549028716114e-07, "loss": 0.149, "step": 7723 }, { "epoch": 1.46, "learning_rate": 8.987674131518911e-07, "loss": 0.2298, "step": 7724 }, { "epoch": 1.46, "learning_rate": 8.98180073325548e-07, "loss": 0.1641, "step": 7725 }, { "epoch": 1.46, "learning_rate": 8.975928834475611e-07, "loss": 0.3067, "step": 7726 }, { "epoch": 1.46, "learning_rate": 8.970058435729003e-07, "loss": 0.2671, "step": 7727 }, { "epoch": 1.46, "learning_rate": 8.964189537565196e-07, "loss": 0.1718, "step": 7728 }, { "epoch": 1.46, "learning_rate": 8.958322140533554e-07, "loss": 0.2312, "step": 7729 }, { "epoch": 1.46, "learning_rate": 8.952456245183361e-07, "loss": 0.2012, "step": 7730 }, { "epoch": 1.46, "learning_rate": 8.946591852063726e-07, "loss": 0.2585, "step": 7731 }, { "epoch": 1.46, "learning_rate": 8.940728961723605e-07, "loss": 0.2519, "step": 7732 }, { "epoch": 1.46, "learning_rate": 8.934867574711848e-07, "loss": 0.1426, "step": 7733 }, { "epoch": 1.46, "learning_rate": 8.929007691577138e-07, "loss": 0.2564, "step": 7734 }, { "epoch": 1.46, "learning_rate": 8.923149312868032e-07, "loss": 0.2164, "step": 7735 }, { "epoch": 1.46, "learning_rate": 8.917292439132938e-07, "loss": 0.2954, "step": 7736 }, { "epoch": 1.46, "learning_rate": 8.911437070920123e-07, "loss": 0.2424, "step": 7737 }, { "epoch": 1.46, "learning_rate": 8.905583208777716e-07, "loss": 0.2228, "step": 7738 }, { "epoch": 1.46, "learning_rate": 8.899730853253705e-07, "loss": 0.1924, "step": 7739 }, { "epoch": 1.46, "learning_rate": 8.893880004895938e-07, "loss": 0.1832, "step": 7740 }, { "epoch": 1.46, "learning_rate": 8.888030664252118e-07, "loss": 0.229, "step": 7741 }, { "epoch": 1.46, "learning_rate": 8.882182831869807e-07, "loss": 0.2215, "step": 7742 }, { "epoch": 1.46, "learning_rate": 8.876336508296424e-07, "loss": 0.1965, "step": 7743 }, { "epoch": 1.46, "learning_rate": 8.870491694079273e-07, "loss": 0.2647, "step": 7744 }, { "epoch": 1.46, "learning_rate": 8.864648389765471e-07, "loss": 0.2721, "step": 7745 }, { "epoch": 1.46, "learning_rate": 8.858806595902014e-07, "loss": 0.1946, "step": 7746 }, { "epoch": 1.46, "learning_rate": 8.85296631303579e-07, "loss": 0.2137, "step": 7747 }, { "epoch": 1.46, "learning_rate": 8.847127541713485e-07, "loss": 0.235, "step": 7748 }, { "epoch": 1.46, "learning_rate": 8.841290282481674e-07, "loss": 0.2302, "step": 7749 }, { "epoch": 1.46, "learning_rate": 8.835454535886817e-07, "loss": 0.2319, "step": 7750 }, { "epoch": 1.46, "learning_rate": 8.829620302475181e-07, "loss": 0.2286, "step": 7751 }, { "epoch": 1.46, "learning_rate": 8.823787582792915e-07, "loss": 0.222, "step": 7752 }, { "epoch": 1.46, "learning_rate": 8.817956377386052e-07, "loss": 0.2575, "step": 7753 }, { "epoch": 1.46, "learning_rate": 8.812126686800423e-07, "loss": 0.2077, "step": 7754 }, { "epoch": 1.47, "learning_rate": 8.806298511581782e-07, "loss": 0.2282, "step": 7755 }, { "epoch": 1.47, "learning_rate": 8.800471852275711e-07, "loss": 0.1899, "step": 7756 }, { "epoch": 1.47, "learning_rate": 8.79464670942762e-07, "loss": 0.2405, "step": 7757 }, { "epoch": 1.47, "learning_rate": 8.78882308358284e-07, "loss": 0.2214, "step": 7758 }, { "epoch": 1.47, "learning_rate": 8.783000975286524e-07, "loss": 0.2578, "step": 7759 }, { "epoch": 1.47, "learning_rate": 8.777180385083661e-07, "loss": 0.1739, "step": 7760 }, { "epoch": 1.47, "learning_rate": 8.771361313519153e-07, "loss": 0.1861, "step": 7761 }, { "epoch": 1.47, "learning_rate": 8.765543761137729e-07, "loss": 0.195, "step": 7762 }, { "epoch": 1.47, "learning_rate": 8.75972772848395e-07, "loss": 0.2451, "step": 7763 }, { "epoch": 1.47, "learning_rate": 8.753913216102286e-07, "loss": 0.2001, "step": 7764 }, { "epoch": 1.47, "learning_rate": 8.748100224537037e-07, "loss": 0.2656, "step": 7765 }, { "epoch": 1.47, "learning_rate": 8.742288754332362e-07, "loss": 0.3065, "step": 7766 }, { "epoch": 1.47, "learning_rate": 8.736478806032281e-07, "loss": 0.2368, "step": 7767 }, { "epoch": 1.47, "learning_rate": 8.730670380180669e-07, "loss": 0.2241, "step": 7768 }, { "epoch": 1.47, "learning_rate": 8.724863477321261e-07, "loss": 0.2473, "step": 7769 }, { "epoch": 1.47, "learning_rate": 8.719058097997646e-07, "loss": 0.2749, "step": 7770 }, { "epoch": 1.47, "learning_rate": 8.713254242753278e-07, "loss": 0.2419, "step": 7771 }, { "epoch": 1.47, "learning_rate": 8.707451912131457e-07, "loss": 0.2464, "step": 7772 }, { "epoch": 1.47, "learning_rate": 8.701651106675351e-07, "loss": 0.2561, "step": 7773 }, { "epoch": 1.47, "learning_rate": 8.695851826927978e-07, "loss": 0.1913, "step": 7774 }, { "epoch": 1.47, "learning_rate": 8.690054073432217e-07, "loss": 0.1966, "step": 7775 }, { "epoch": 1.47, "learning_rate": 8.684257846730804e-07, "loss": 0.2324, "step": 7776 }, { "epoch": 1.47, "learning_rate": 8.678463147366328e-07, "loss": 0.2328, "step": 7777 }, { "epoch": 1.47, "learning_rate": 8.67266997588124e-07, "loss": 0.3637, "step": 7778 }, { "epoch": 1.47, "learning_rate": 8.666878332817846e-07, "loss": 0.2726, "step": 7779 }, { "epoch": 1.47, "learning_rate": 8.661088218718308e-07, "loss": 0.2542, "step": 7780 }, { "epoch": 1.47, "learning_rate": 8.655299634124648e-07, "loss": 0.2344, "step": 7781 }, { "epoch": 1.47, "learning_rate": 8.64951257957874e-07, "loss": 0.2952, "step": 7782 }, { "epoch": 1.47, "learning_rate": 8.64372705562232e-07, "loss": 0.2307, "step": 7783 }, { "epoch": 1.47, "learning_rate": 8.637943062796975e-07, "loss": 0.2324, "step": 7784 }, { "epoch": 1.47, "learning_rate": 8.632160601644144e-07, "loss": 0.3127, "step": 7785 }, { "epoch": 1.47, "learning_rate": 8.626379672705157e-07, "loss": 0.1926, "step": 7786 }, { "epoch": 1.47, "learning_rate": 8.620600276521149e-07, "loss": 0.1856, "step": 7787 }, { "epoch": 1.47, "learning_rate": 8.614822413633134e-07, "loss": 0.1805, "step": 7788 }, { "epoch": 1.47, "learning_rate": 8.609046084582013e-07, "loss": 0.2022, "step": 7789 }, { "epoch": 1.47, "learning_rate": 8.603271289908491e-07, "loss": 0.2425, "step": 7790 }, { "epoch": 1.47, "learning_rate": 8.597498030153148e-07, "loss": 0.2251, "step": 7791 }, { "epoch": 1.47, "learning_rate": 8.591726305856457e-07, "loss": 0.2801, "step": 7792 }, { "epoch": 1.47, "learning_rate": 8.585956117558689e-07, "loss": 0.2428, "step": 7793 }, { "epoch": 1.47, "learning_rate": 8.580187465799997e-07, "loss": 0.2525, "step": 7794 }, { "epoch": 1.47, "learning_rate": 8.574420351120419e-07, "loss": 0.2649, "step": 7795 }, { "epoch": 1.47, "learning_rate": 8.56865477405979e-07, "loss": 0.2538, "step": 7796 }, { "epoch": 1.47, "learning_rate": 8.562890735157853e-07, "loss": 0.2628, "step": 7797 }, { "epoch": 1.47, "learning_rate": 8.55712823495419e-07, "loss": 0.2157, "step": 7798 }, { "epoch": 1.47, "learning_rate": 8.551367273988212e-07, "loss": 0.2595, "step": 7799 }, { "epoch": 1.47, "learning_rate": 8.545607852799231e-07, "loss": 0.241, "step": 7800 }, { "epoch": 1.47, "learning_rate": 8.539849971926387e-07, "loss": 0.2304, "step": 7801 }, { "epoch": 1.47, "learning_rate": 8.534093631908682e-07, "loss": 0.2039, "step": 7802 }, { "epoch": 1.47, "learning_rate": 8.528338833284977e-07, "loss": 0.1577, "step": 7803 }, { "epoch": 1.47, "learning_rate": 8.522585576593983e-07, "loss": 0.2838, "step": 7804 }, { "epoch": 1.47, "learning_rate": 8.516833862374272e-07, "loss": 0.3036, "step": 7805 }, { "epoch": 1.47, "learning_rate": 8.511083691164265e-07, "loss": 0.2593, "step": 7806 }, { "epoch": 1.47, "learning_rate": 8.505335063502246e-07, "loss": 0.2514, "step": 7807 }, { "epoch": 1.48, "learning_rate": 8.499587979926349e-07, "loss": 0.3137, "step": 7808 }, { "epoch": 1.48, "learning_rate": 8.493842440974567e-07, "loss": 0.2456, "step": 7809 }, { "epoch": 1.48, "learning_rate": 8.488098447184748e-07, "loss": 0.2152, "step": 7810 }, { "epoch": 1.48, "learning_rate": 8.482355999094591e-07, "loss": 0.2535, "step": 7811 }, { "epoch": 1.48, "learning_rate": 8.476615097241658e-07, "loss": 0.2442, "step": 7812 }, { "epoch": 1.48, "learning_rate": 8.47087574216336e-07, "loss": 0.1784, "step": 7813 }, { "epoch": 1.48, "learning_rate": 8.465137934396964e-07, "loss": 0.2257, "step": 7814 }, { "epoch": 1.48, "learning_rate": 8.459401674479594e-07, "loss": 0.2554, "step": 7815 }, { "epoch": 1.48, "learning_rate": 8.453666962948223e-07, "loss": 0.2393, "step": 7816 }, { "epoch": 1.48, "learning_rate": 8.447933800339709e-07, "loss": 0.2847, "step": 7817 }, { "epoch": 1.48, "learning_rate": 8.442202187190709e-07, "loss": 0.1984, "step": 7818 }, { "epoch": 1.48, "learning_rate": 8.436472124037772e-07, "loss": 0.2212, "step": 7819 }, { "epoch": 1.48, "learning_rate": 8.430743611417322e-07, "loss": 0.2184, "step": 7820 }, { "epoch": 1.48, "learning_rate": 8.425016649865583e-07, "loss": 0.2618, "step": 7821 }, { "epoch": 1.48, "learning_rate": 8.419291239918667e-07, "loss": 0.2243, "step": 7822 }, { "epoch": 1.48, "learning_rate": 8.413567382112561e-07, "loss": 0.2182, "step": 7823 }, { "epoch": 1.48, "learning_rate": 8.407845076983057e-07, "loss": 0.2168, "step": 7824 }, { "epoch": 1.48, "learning_rate": 8.402124325065828e-07, "loss": 0.241, "step": 7825 }, { "epoch": 1.48, "learning_rate": 8.39640512689642e-07, "loss": 0.2864, "step": 7826 }, { "epoch": 1.48, "learning_rate": 8.390687483010191e-07, "loss": 0.212, "step": 7827 }, { "epoch": 1.48, "learning_rate": 8.384971393942393e-07, "loss": 0.199, "step": 7828 }, { "epoch": 1.48, "learning_rate": 8.379256860228119e-07, "loss": 0.2388, "step": 7829 }, { "epoch": 1.48, "learning_rate": 8.373543882402291e-07, "loss": 0.254, "step": 7830 }, { "epoch": 1.48, "learning_rate": 8.36783246099973e-07, "loss": 0.2122, "step": 7831 }, { "epoch": 1.48, "learning_rate": 8.362122596555089e-07, "loss": 0.2721, "step": 7832 }, { "epoch": 1.48, "learning_rate": 8.356414289602854e-07, "loss": 0.2622, "step": 7833 }, { "epoch": 1.48, "learning_rate": 8.350707540677411e-07, "loss": 0.2161, "step": 7834 }, { "epoch": 1.48, "learning_rate": 8.345002350312973e-07, "loss": 0.2755, "step": 7835 }, { "epoch": 1.48, "learning_rate": 8.339298719043585e-07, "loss": 0.2654, "step": 7836 }, { "epoch": 1.48, "learning_rate": 8.333596647403198e-07, "loss": 0.2364, "step": 7837 }, { "epoch": 1.48, "learning_rate": 8.327896135925581e-07, "loss": 0.2317, "step": 7838 }, { "epoch": 1.48, "learning_rate": 8.322197185144368e-07, "loss": 0.1967, "step": 7839 }, { "epoch": 1.48, "learning_rate": 8.316499795593042e-07, "loss": 0.1738, "step": 7840 }, { "epoch": 1.48, "learning_rate": 8.310803967804945e-07, "loss": 0.3235, "step": 7841 }, { "epoch": 1.48, "learning_rate": 8.305109702313266e-07, "loss": 0.2344, "step": 7842 }, { "epoch": 1.48, "learning_rate": 8.299416999651058e-07, "loss": 0.2362, "step": 7843 }, { "epoch": 1.48, "learning_rate": 8.29372586035122e-07, "loss": 0.2574, "step": 7844 }, { "epoch": 1.48, "learning_rate": 8.288036284946504e-07, "loss": 0.1868, "step": 7845 }, { "epoch": 1.48, "learning_rate": 8.282348273969523e-07, "loss": 0.2592, "step": 7846 }, { "epoch": 1.48, "learning_rate": 8.276661827952734e-07, "loss": 0.2929, "step": 7847 }, { "epoch": 1.48, "learning_rate": 8.270976947428456e-07, "loss": 0.2405, "step": 7848 }, { "epoch": 1.48, "learning_rate": 8.265293632928856e-07, "loss": 0.2477, "step": 7849 }, { "epoch": 1.48, "learning_rate": 8.259611884985955e-07, "loss": 0.1795, "step": 7850 }, { "epoch": 1.48, "learning_rate": 8.253931704131632e-07, "loss": 0.2719, "step": 7851 }, { "epoch": 1.48, "learning_rate": 8.248253090897612e-07, "loss": 0.2125, "step": 7852 }, { "epoch": 1.48, "learning_rate": 8.242576045815481e-07, "loss": 0.2166, "step": 7853 }, { "epoch": 1.48, "learning_rate": 8.236900569416669e-07, "loss": 0.2016, "step": 7854 }, { "epoch": 1.48, "learning_rate": 8.231226662232469e-07, "loss": 0.2464, "step": 7855 }, { "epoch": 1.48, "learning_rate": 8.225554324794022e-07, "loss": 0.1888, "step": 7856 }, { "epoch": 1.48, "learning_rate": 8.219883557632319e-07, "loss": 0.1586, "step": 7857 }, { "epoch": 1.48, "learning_rate": 8.214214361278203e-07, "loss": 0.2675, "step": 7858 }, { "epoch": 1.48, "learning_rate": 8.208546736262399e-07, "loss": 0.2293, "step": 7859 }, { "epoch": 1.48, "learning_rate": 8.202880683115435e-07, "loss": 0.2986, "step": 7860 }, { "epoch": 1.49, "learning_rate": 8.197216202367717e-07, "loss": 0.2545, "step": 7861 }, { "epoch": 1.49, "learning_rate": 8.191553294549528e-07, "loss": 0.2811, "step": 7862 }, { "epoch": 1.49, "learning_rate": 8.185891960190956e-07, "loss": 0.203, "step": 7863 }, { "epoch": 1.49, "learning_rate": 8.180232199821966e-07, "loss": 0.2136, "step": 7864 }, { "epoch": 1.49, "learning_rate": 8.174574013972391e-07, "loss": 0.2611, "step": 7865 }, { "epoch": 1.49, "learning_rate": 8.168917403171891e-07, "loss": 0.1438, "step": 7866 }, { "epoch": 1.49, "learning_rate": 8.163262367949995e-07, "loss": 0.2115, "step": 7867 }, { "epoch": 1.49, "learning_rate": 8.157608908836071e-07, "loss": 0.1682, "step": 7868 }, { "epoch": 1.49, "learning_rate": 8.151957026359347e-07, "loss": 0.2211, "step": 7869 }, { "epoch": 1.49, "learning_rate": 8.146306721048905e-07, "loss": 0.2367, "step": 7870 }, { "epoch": 1.49, "learning_rate": 8.140657993433676e-07, "loss": 0.1532, "step": 7871 }, { "epoch": 1.49, "learning_rate": 8.135010844042446e-07, "loss": 0.26, "step": 7872 }, { "epoch": 1.49, "learning_rate": 8.12936527340385e-07, "loss": 0.1864, "step": 7873 }, { "epoch": 1.49, "learning_rate": 8.123721282046379e-07, "loss": 0.3127, "step": 7874 }, { "epoch": 1.49, "learning_rate": 8.118078870498369e-07, "loss": 0.2316, "step": 7875 }, { "epoch": 1.49, "learning_rate": 8.11243803928802e-07, "loss": 0.2043, "step": 7876 }, { "epoch": 1.49, "learning_rate": 8.106798788943371e-07, "loss": 0.2572, "step": 7877 }, { "epoch": 1.49, "learning_rate": 8.101161119992324e-07, "loss": 0.2471, "step": 7878 }, { "epoch": 1.49, "learning_rate": 8.095525032962626e-07, "loss": 0.2573, "step": 7879 }, { "epoch": 1.49, "learning_rate": 8.089890528381877e-07, "loss": 0.2026, "step": 7880 }, { "epoch": 1.49, "learning_rate": 8.084257606777532e-07, "loss": 0.1759, "step": 7881 }, { "epoch": 1.49, "learning_rate": 8.078626268676898e-07, "loss": 0.2572, "step": 7882 }, { "epoch": 1.49, "learning_rate": 8.072996514607126e-07, "loss": 0.2303, "step": 7883 }, { "epoch": 1.49, "learning_rate": 8.067368345095228e-07, "loss": 0.24, "step": 7884 }, { "epoch": 1.49, "learning_rate": 8.061741760668063e-07, "loss": 0.2032, "step": 7885 }, { "epoch": 1.49, "learning_rate": 8.056116761852334e-07, "loss": 0.2182, "step": 7886 }, { "epoch": 1.49, "learning_rate": 8.050493349174629e-07, "loss": 0.264, "step": 7887 }, { "epoch": 1.49, "learning_rate": 8.044871523161337e-07, "loss": 0.2199, "step": 7888 }, { "epoch": 1.49, "learning_rate": 8.039251284338729e-07, "loss": 0.2251, "step": 7889 }, { "epoch": 1.49, "learning_rate": 8.033632633232941e-07, "loss": 0.2169, "step": 7890 }, { "epoch": 1.49, "learning_rate": 8.02801557036992e-07, "loss": 0.2011, "step": 7891 }, { "epoch": 1.49, "learning_rate": 8.022400096275487e-07, "loss": 0.2524, "step": 7892 }, { "epoch": 1.49, "learning_rate": 8.016786211475336e-07, "loss": 0.293, "step": 7893 }, { "epoch": 1.49, "learning_rate": 8.011173916494966e-07, "loss": 0.2707, "step": 7894 }, { "epoch": 1.49, "learning_rate": 8.005563211859755e-07, "loss": 0.2292, "step": 7895 }, { "epoch": 1.49, "learning_rate": 7.999954098094945e-07, "loss": 0.252, "step": 7896 }, { "epoch": 1.49, "learning_rate": 7.994346575725584e-07, "loss": 0.2539, "step": 7897 }, { "epoch": 1.49, "learning_rate": 7.988740645276624e-07, "loss": 0.1899, "step": 7898 }, { "epoch": 1.49, "learning_rate": 7.98313630727284e-07, "loss": 0.2473, "step": 7899 }, { "epoch": 1.49, "learning_rate": 7.97753356223884e-07, "loss": 0.219, "step": 7900 }, { "epoch": 1.49, "learning_rate": 7.971932410699126e-07, "loss": 0.2012, "step": 7901 }, { "epoch": 1.49, "learning_rate": 7.966332853178032e-07, "loss": 0.1934, "step": 7902 }, { "epoch": 1.49, "learning_rate": 7.960734890199712e-07, "loss": 0.2136, "step": 7903 }, { "epoch": 1.49, "learning_rate": 7.955138522288225e-07, "loss": 0.2169, "step": 7904 }, { "epoch": 1.49, "learning_rate": 7.949543749967451e-07, "loss": 0.2693, "step": 7905 }, { "epoch": 1.49, "learning_rate": 7.9439505737611e-07, "loss": 0.2001, "step": 7906 }, { "epoch": 1.49, "learning_rate": 7.938358994192783e-07, "loss": 0.2434, "step": 7907 }, { "epoch": 1.49, "learning_rate": 7.932769011785926e-07, "loss": 0.3119, "step": 7908 }, { "epoch": 1.49, "learning_rate": 7.92718062706381e-07, "loss": 0.1912, "step": 7909 }, { "epoch": 1.49, "learning_rate": 7.921593840549577e-07, "loss": 0.2236, "step": 7910 }, { "epoch": 1.49, "learning_rate": 7.916008652766208e-07, "loss": 0.2481, "step": 7911 }, { "epoch": 1.49, "learning_rate": 7.910425064236543e-07, "loss": 0.2531, "step": 7912 }, { "epoch": 1.49, "learning_rate": 7.904843075483265e-07, "loss": 0.2839, "step": 7913 }, { "epoch": 1.5, "learning_rate": 7.899262687028914e-07, "loss": 0.3023, "step": 7914 }, { "epoch": 1.5, "learning_rate": 7.893683899395876e-07, "loss": 0.2009, "step": 7915 }, { "epoch": 1.5, "learning_rate": 7.888106713106389e-07, "loss": 0.2041, "step": 7916 }, { "epoch": 1.5, "learning_rate": 7.882531128682539e-07, "loss": 0.204, "step": 7917 }, { "epoch": 1.5, "learning_rate": 7.876957146646263e-07, "loss": 0.117, "step": 7918 }, { "epoch": 1.5, "learning_rate": 7.871384767519349e-07, "loss": 0.1533, "step": 7919 }, { "epoch": 1.5, "learning_rate": 7.865813991823437e-07, "loss": 0.195, "step": 7920 }, { "epoch": 1.5, "learning_rate": 7.860244820080012e-07, "loss": 0.1714, "step": 7921 }, { "epoch": 1.5, "learning_rate": 7.85467725281041e-07, "loss": 0.2426, "step": 7922 }, { "epoch": 1.5, "learning_rate": 7.849111290535818e-07, "loss": 0.2056, "step": 7923 }, { "epoch": 1.5, "learning_rate": 7.843546933777277e-07, "loss": 0.1761, "step": 7924 }, { "epoch": 1.5, "learning_rate": 7.837984183055666e-07, "loss": 0.291, "step": 7925 }, { "epoch": 1.5, "learning_rate": 7.832423038891721e-07, "loss": 0.1956, "step": 7926 }, { "epoch": 1.5, "learning_rate": 7.826863501806045e-07, "loss": 0.2206, "step": 7927 }, { "epoch": 1.5, "learning_rate": 7.821305572319044e-07, "loss": 0.2462, "step": 7928 }, { "epoch": 1.5, "learning_rate": 7.815749250951027e-07, "loss": 0.2152, "step": 7929 }, { "epoch": 1.5, "learning_rate": 7.810194538222129e-07, "loss": 0.2173, "step": 7930 }, { "epoch": 1.5, "learning_rate": 7.804641434652308e-07, "loss": 0.2803, "step": 7931 }, { "epoch": 1.5, "learning_rate": 7.79908994076142e-07, "loss": 0.1659, "step": 7932 }, { "epoch": 1.5, "learning_rate": 7.793540057069149e-07, "loss": 0.2165, "step": 7933 }, { "epoch": 1.5, "learning_rate": 7.787991784095e-07, "loss": 0.2588, "step": 7934 }, { "epoch": 1.5, "learning_rate": 7.782445122358385e-07, "loss": 0.2359, "step": 7935 }, { "epoch": 1.5, "learning_rate": 7.776900072378524e-07, "loss": 0.2128, "step": 7936 }, { "epoch": 1.5, "learning_rate": 7.771356634674479e-07, "loss": 0.2382, "step": 7937 }, { "epoch": 1.5, "learning_rate": 7.765814809765199e-07, "loss": 0.2246, "step": 7938 }, { "epoch": 1.5, "learning_rate": 7.760274598169454e-07, "loss": 0.3328, "step": 7939 }, { "epoch": 1.5, "learning_rate": 7.754736000405871e-07, "loss": 0.2692, "step": 7940 }, { "epoch": 1.5, "learning_rate": 7.749199016992923e-07, "loss": 0.2676, "step": 7941 }, { "epoch": 1.5, "learning_rate": 7.743663648448935e-07, "loss": 0.1989, "step": 7942 }, { "epoch": 1.5, "learning_rate": 7.738129895292079e-07, "loss": 0.1838, "step": 7943 }, { "epoch": 1.5, "learning_rate": 7.73259775804038e-07, "loss": 0.2305, "step": 7944 }, { "epoch": 1.5, "learning_rate": 7.727067237211703e-07, "loss": 0.1764, "step": 7945 }, { "epoch": 1.5, "learning_rate": 7.721538333323772e-07, "loss": 0.1846, "step": 7946 }, { "epoch": 1.5, "learning_rate": 7.716011046894151e-07, "loss": 0.2381, "step": 7947 }, { "epoch": 1.5, "learning_rate": 7.710485378440255e-07, "loss": 0.2217, "step": 7948 }, { "epoch": 1.5, "learning_rate": 7.704961328479352e-07, "loss": 0.2001, "step": 7949 }, { "epoch": 1.5, "learning_rate": 7.699438897528555e-07, "loss": 0.2612, "step": 7950 }, { "epoch": 1.5, "learning_rate": 7.693918086104826e-07, "loss": 0.2318, "step": 7951 }, { "epoch": 1.5, "learning_rate": 7.688398894724971e-07, "loss": 0.2076, "step": 7952 }, { "epoch": 1.5, "learning_rate": 7.682881323905653e-07, "loss": 0.2359, "step": 7953 }, { "epoch": 1.5, "learning_rate": 7.677365374163376e-07, "loss": 0.1771, "step": 7954 }, { "epoch": 1.5, "learning_rate": 7.671851046014497e-07, "loss": 0.1999, "step": 7955 }, { "epoch": 1.5, "learning_rate": 7.666338339975216e-07, "loss": 0.2589, "step": 7956 }, { "epoch": 1.5, "learning_rate": 7.660827256561587e-07, "loss": 0.2844, "step": 7957 }, { "epoch": 1.5, "learning_rate": 7.655317796289511e-07, "loss": 0.2214, "step": 7958 }, { "epoch": 1.5, "learning_rate": 7.649809959674723e-07, "loss": 0.1972, "step": 7959 }, { "epoch": 1.5, "learning_rate": 7.644303747232842e-07, "loss": 0.2915, "step": 7960 }, { "epoch": 1.5, "learning_rate": 7.638799159479293e-07, "loss": 0.2778, "step": 7961 }, { "epoch": 1.5, "learning_rate": 7.633296196929363e-07, "loss": 0.2743, "step": 7962 }, { "epoch": 1.5, "learning_rate": 7.627794860098217e-07, "loss": 0.1852, "step": 7963 }, { "epoch": 1.5, "learning_rate": 7.622295149500816e-07, "loss": 0.1736, "step": 7964 }, { "epoch": 1.5, "learning_rate": 7.616797065651998e-07, "loss": 0.25, "step": 7965 }, { "epoch": 1.51, "learning_rate": 7.611300609066466e-07, "loss": 0.2841, "step": 7966 }, { "epoch": 1.51, "learning_rate": 7.605805780258727e-07, "loss": 0.2319, "step": 7967 }, { "epoch": 1.51, "learning_rate": 7.600312579743161e-07, "loss": 0.2443, "step": 7968 }, { "epoch": 1.51, "learning_rate": 7.594821008034015e-07, "loss": 0.2342, "step": 7969 }, { "epoch": 1.51, "learning_rate": 7.58933106564533e-07, "loss": 0.2317, "step": 7970 }, { "epoch": 1.51, "learning_rate": 7.583842753091053e-07, "loss": 0.2302, "step": 7971 }, { "epoch": 1.51, "learning_rate": 7.578356070884949e-07, "loss": 0.2585, "step": 7972 }, { "epoch": 1.51, "learning_rate": 7.572871019540612e-07, "loss": 0.2561, "step": 7973 }, { "epoch": 1.51, "learning_rate": 7.567387599571524e-07, "loss": 0.2812, "step": 7974 }, { "epoch": 1.51, "learning_rate": 7.561905811490997e-07, "loss": 0.2178, "step": 7975 }, { "epoch": 1.51, "learning_rate": 7.556425655812164e-07, "loss": 0.2617, "step": 7976 }, { "epoch": 1.51, "learning_rate": 7.550947133048053e-07, "loss": 0.2807, "step": 7977 }, { "epoch": 1.51, "learning_rate": 7.545470243711517e-07, "loss": 0.2877, "step": 7978 }, { "epoch": 1.51, "learning_rate": 7.53999498831523e-07, "loss": 0.2254, "step": 7979 }, { "epoch": 1.51, "learning_rate": 7.534521367371758e-07, "loss": 0.2477, "step": 7980 }, { "epoch": 1.51, "learning_rate": 7.529049381393491e-07, "loss": 0.3156, "step": 7981 }, { "epoch": 1.51, "learning_rate": 7.523579030892663e-07, "loss": 0.2585, "step": 7982 }, { "epoch": 1.51, "learning_rate": 7.518110316381363e-07, "loss": 0.2211, "step": 7983 }, { "epoch": 1.51, "learning_rate": 7.512643238371528e-07, "loss": 0.2652, "step": 7984 }, { "epoch": 1.51, "learning_rate": 7.507177797374929e-07, "loss": 0.2276, "step": 7985 }, { "epoch": 1.51, "learning_rate": 7.5017139939032e-07, "loss": 0.2195, "step": 7986 }, { "epoch": 1.51, "learning_rate": 7.496251828467812e-07, "loss": 0.2309, "step": 7987 }, { "epoch": 1.51, "learning_rate": 7.490791301580085e-07, "loss": 0.2429, "step": 7988 }, { "epoch": 1.51, "learning_rate": 7.485332413751187e-07, "loss": 0.1644, "step": 7989 }, { "epoch": 1.51, "learning_rate": 7.479875165492124e-07, "loss": 0.249, "step": 7990 }, { "epoch": 1.51, "learning_rate": 7.474419557313772e-07, "loss": 0.2054, "step": 7991 }, { "epoch": 1.51, "learning_rate": 7.468965589726826e-07, "loss": 0.1687, "step": 7992 }, { "epoch": 1.51, "learning_rate": 7.463513263241828e-07, "loss": 0.2153, "step": 7993 }, { "epoch": 1.51, "learning_rate": 7.458062578369207e-07, "loss": 0.1647, "step": 7994 }, { "epoch": 1.51, "learning_rate": 7.452613535619179e-07, "loss": 0.2114, "step": 7995 }, { "epoch": 1.51, "learning_rate": 7.447166135501843e-07, "loss": 0.2518, "step": 7996 }, { "epoch": 1.51, "learning_rate": 7.441720378527156e-07, "loss": 0.2532, "step": 7997 }, { "epoch": 1.51, "learning_rate": 7.436276265204881e-07, "loss": 0.2039, "step": 7998 }, { "epoch": 1.51, "learning_rate": 7.430833796044645e-07, "loss": 0.2614, "step": 7999 }, { "epoch": 1.51, "learning_rate": 7.42539297155595e-07, "loss": 0.1574, "step": 8000 }, { "epoch": 1.51, "learning_rate": 7.419953792248086e-07, "loss": 0.3, "step": 8001 }, { "epoch": 1.51, "learning_rate": 7.414516258630245e-07, "loss": 0.1942, "step": 8002 }, { "epoch": 1.51, "learning_rate": 7.409080371211441e-07, "loss": 0.2932, "step": 8003 }, { "epoch": 1.51, "learning_rate": 7.403646130500514e-07, "loss": 0.2609, "step": 8004 }, { "epoch": 1.51, "learning_rate": 7.398213537006188e-07, "loss": 0.3047, "step": 8005 }, { "epoch": 1.51, "learning_rate": 7.392782591237019e-07, "loss": 0.2381, "step": 8006 }, { "epoch": 1.51, "learning_rate": 7.38735329370138e-07, "loss": 0.2818, "step": 8007 }, { "epoch": 1.51, "learning_rate": 7.381925644907536e-07, "loss": 0.2371, "step": 8008 }, { "epoch": 1.51, "learning_rate": 7.376499645363577e-07, "loss": 0.2205, "step": 8009 }, { "epoch": 1.51, "learning_rate": 7.371075295577418e-07, "loss": 0.2395, "step": 8010 }, { "epoch": 1.51, "learning_rate": 7.365652596056858e-07, "loss": 0.2354, "step": 8011 }, { "epoch": 1.51, "learning_rate": 7.360231547309516e-07, "loss": 0.2577, "step": 8012 }, { "epoch": 1.51, "learning_rate": 7.354812149842861e-07, "loss": 0.2101, "step": 8013 }, { "epoch": 1.51, "learning_rate": 7.349394404164215e-07, "loss": 0.2206, "step": 8014 }, { "epoch": 1.51, "learning_rate": 7.343978310780733e-07, "loss": 0.1716, "step": 8015 }, { "epoch": 1.51, "learning_rate": 7.338563870199428e-07, "loss": 0.1955, "step": 8016 }, { "epoch": 1.51, "learning_rate": 7.333151082927151e-07, "loss": 0.2151, "step": 8017 }, { "epoch": 1.51, "learning_rate": 7.3277399494706e-07, "loss": 0.2764, "step": 8018 }, { "epoch": 1.52, "learning_rate": 7.322330470336314e-07, "loss": 0.258, "step": 8019 }, { "epoch": 1.52, "learning_rate": 7.316922646030686e-07, "loss": 0.1466, "step": 8020 }, { "epoch": 1.52, "learning_rate": 7.311516477059946e-07, "loss": 0.1951, "step": 8021 }, { "epoch": 1.52, "learning_rate": 7.306111963930177e-07, "loss": 0.2718, "step": 8022 }, { "epoch": 1.52, "learning_rate": 7.300709107147294e-07, "loss": 0.1482, "step": 8023 }, { "epoch": 1.52, "learning_rate": 7.295307907217075e-07, "loss": 0.2468, "step": 8024 }, { "epoch": 1.52, "learning_rate": 7.289908364645124e-07, "loss": 0.3, "step": 8025 }, { "epoch": 1.52, "learning_rate": 7.284510479936904e-07, "loss": 0.141, "step": 8026 }, { "epoch": 1.52, "learning_rate": 7.279114253597716e-07, "loss": 0.1959, "step": 8027 }, { "epoch": 1.52, "learning_rate": 7.27371968613271e-07, "loss": 0.2005, "step": 8028 }, { "epoch": 1.52, "learning_rate": 7.268326778046866e-07, "loss": 0.188, "step": 8029 }, { "epoch": 1.52, "learning_rate": 7.26293552984505e-07, "loss": 0.2578, "step": 8030 }, { "epoch": 1.52, "learning_rate": 7.257545942031916e-07, "loss": 0.2167, "step": 8031 }, { "epoch": 1.52, "learning_rate": 7.252158015111988e-07, "loss": 0.1924, "step": 8032 }, { "epoch": 1.52, "learning_rate": 7.246771749589665e-07, "loss": 0.1788, "step": 8033 }, { "epoch": 1.52, "learning_rate": 7.241387145969133e-07, "loss": 0.2094, "step": 8034 }, { "epoch": 1.52, "learning_rate": 7.23600420475446e-07, "loss": 0.2552, "step": 8035 }, { "epoch": 1.52, "learning_rate": 7.230622926449565e-07, "loss": 0.3367, "step": 8036 }, { "epoch": 1.52, "learning_rate": 7.225243311558175e-07, "loss": 0.2239, "step": 8037 }, { "epoch": 1.52, "learning_rate": 7.219865360583886e-07, "loss": 0.2449, "step": 8038 }, { "epoch": 1.52, "learning_rate": 7.214489074030154e-07, "loss": 0.2207, "step": 8039 }, { "epoch": 1.52, "learning_rate": 7.209114452400228e-07, "loss": 0.1958, "step": 8040 }, { "epoch": 1.52, "learning_rate": 7.203741496197261e-07, "loss": 0.2199, "step": 8041 }, { "epoch": 1.52, "learning_rate": 7.198370205924218e-07, "loss": 0.1805, "step": 8042 }, { "epoch": 1.52, "learning_rate": 7.19300058208389e-07, "loss": 0.2577, "step": 8043 }, { "epoch": 1.52, "learning_rate": 7.187632625178959e-07, "loss": 0.2662, "step": 8044 }, { "epoch": 1.52, "learning_rate": 7.182266335711923e-07, "loss": 0.1926, "step": 8045 }, { "epoch": 1.52, "learning_rate": 7.176901714185109e-07, "loss": 0.2587, "step": 8046 }, { "epoch": 1.52, "learning_rate": 7.171538761100724e-07, "loss": 0.2441, "step": 8047 }, { "epoch": 1.52, "learning_rate": 7.166177476960803e-07, "loss": 0.261, "step": 8048 }, { "epoch": 1.52, "learning_rate": 7.1608178622672e-07, "loss": 0.2299, "step": 8049 }, { "epoch": 1.52, "learning_rate": 7.155459917521654e-07, "loss": 0.1879, "step": 8050 }, { "epoch": 1.52, "learning_rate": 7.150103643225728e-07, "loss": 0.2133, "step": 8051 }, { "epoch": 1.52, "learning_rate": 7.144749039880825e-07, "loss": 0.2552, "step": 8052 }, { "epoch": 1.52, "learning_rate": 7.139396107988195e-07, "loss": 0.26, "step": 8053 }, { "epoch": 1.52, "learning_rate": 7.134044848048935e-07, "loss": 0.2885, "step": 8054 }, { "epoch": 1.52, "learning_rate": 7.128695260563984e-07, "loss": 0.19, "step": 8055 }, { "epoch": 1.52, "learning_rate": 7.123347346034123e-07, "loss": 0.167, "step": 8056 }, { "epoch": 1.52, "learning_rate": 7.118001104959976e-07, "loss": 0.2528, "step": 8057 }, { "epoch": 1.52, "learning_rate": 7.112656537842011e-07, "loss": 0.192, "step": 8058 }, { "epoch": 1.52, "learning_rate": 7.107313645180538e-07, "loss": 0.1599, "step": 8059 }, { "epoch": 1.52, "learning_rate": 7.101972427475709e-07, "loss": 0.2201, "step": 8060 }, { "epoch": 1.52, "learning_rate": 7.096632885227542e-07, "loss": 0.1968, "step": 8061 }, { "epoch": 1.52, "learning_rate": 7.091295018935853e-07, "loss": 0.281, "step": 8062 }, { "epoch": 1.52, "learning_rate": 7.085958829100331e-07, "loss": 0.2671, "step": 8063 }, { "epoch": 1.52, "learning_rate": 7.080624316220525e-07, "loss": 0.3275, "step": 8064 }, { "epoch": 1.52, "learning_rate": 7.075291480795782e-07, "loss": 0.2822, "step": 8065 }, { "epoch": 1.52, "learning_rate": 7.069960323325311e-07, "loss": 0.3005, "step": 8066 }, { "epoch": 1.52, "learning_rate": 7.064630844308199e-07, "loss": 0.2022, "step": 8067 }, { "epoch": 1.52, "learning_rate": 7.059303044243315e-07, "loss": 0.209, "step": 8068 }, { "epoch": 1.52, "learning_rate": 7.053976923629407e-07, "loss": 0.2471, "step": 8069 }, { "epoch": 1.52, "learning_rate": 7.04865248296508e-07, "loss": 0.1816, "step": 8070 }, { "epoch": 1.52, "learning_rate": 7.043329722748731e-07, "loss": 0.1667, "step": 8071 }, { "epoch": 1.53, "learning_rate": 7.038008643478653e-07, "loss": 0.2157, "step": 8072 }, { "epoch": 1.53, "learning_rate": 7.03268924565296e-07, "loss": 0.2474, "step": 8073 }, { "epoch": 1.53, "learning_rate": 7.02737152976958e-07, "loss": 0.1758, "step": 8074 }, { "epoch": 1.53, "learning_rate": 7.022055496326341e-07, "loss": 0.1962, "step": 8075 }, { "epoch": 1.53, "learning_rate": 7.016741145820879e-07, "loss": 0.2223, "step": 8076 }, { "epoch": 1.53, "learning_rate": 7.011428478750657e-07, "loss": 0.2188, "step": 8077 }, { "epoch": 1.53, "learning_rate": 7.006117495613018e-07, "loss": 0.2349, "step": 8078 }, { "epoch": 1.53, "learning_rate": 7.000808196905134e-07, "loss": 0.2685, "step": 8079 }, { "epoch": 1.53, "learning_rate": 6.995500583123993e-07, "loss": 0.2584, "step": 8080 }, { "epoch": 1.53, "learning_rate": 6.990194654766469e-07, "loss": 0.3389, "step": 8081 }, { "epoch": 1.53, "learning_rate": 6.984890412329245e-07, "loss": 0.1689, "step": 8082 }, { "epoch": 1.53, "learning_rate": 6.979587856308865e-07, "loss": 0.1823, "step": 8083 }, { "epoch": 1.53, "learning_rate": 6.974286987201701e-07, "loss": 0.2201, "step": 8084 }, { "epoch": 1.53, "learning_rate": 6.96898780550398e-07, "loss": 0.2463, "step": 8085 }, { "epoch": 1.53, "learning_rate": 6.96369031171176e-07, "loss": 0.2091, "step": 8086 }, { "epoch": 1.53, "learning_rate": 6.958394506320948e-07, "loss": 0.2087, "step": 8087 }, { "epoch": 1.53, "learning_rate": 6.953100389827291e-07, "loss": 0.2701, "step": 8088 }, { "epoch": 1.53, "learning_rate": 6.947807962726377e-07, "loss": 0.2296, "step": 8089 }, { "epoch": 1.53, "learning_rate": 6.94251722551364e-07, "loss": 0.2432, "step": 8090 }, { "epoch": 1.53, "learning_rate": 6.937228178684346e-07, "loss": 0.1727, "step": 8091 }, { "epoch": 1.53, "learning_rate": 6.931940822733618e-07, "loss": 0.2219, "step": 8092 }, { "epoch": 1.53, "learning_rate": 6.926655158156403e-07, "loss": 0.2468, "step": 8093 }, { "epoch": 1.53, "learning_rate": 6.921371185447506e-07, "loss": 0.246, "step": 8094 }, { "epoch": 1.53, "learning_rate": 6.916088905101562e-07, "loss": 0.2508, "step": 8095 }, { "epoch": 1.53, "learning_rate": 6.910808317613052e-07, "loss": 0.2316, "step": 8096 }, { "epoch": 1.53, "learning_rate": 6.905529423476301e-07, "loss": 0.2087, "step": 8097 }, { "epoch": 1.53, "learning_rate": 6.900252223185469e-07, "loss": 0.2295, "step": 8098 }, { "epoch": 1.53, "learning_rate": 6.894976717234567e-07, "loss": 0.2021, "step": 8099 }, { "epoch": 1.53, "learning_rate": 6.889702906117435e-07, "loss": 0.2266, "step": 8100 }, { "epoch": 1.53, "learning_rate": 6.884430790327765e-07, "loss": 0.2407, "step": 8101 }, { "epoch": 1.53, "learning_rate": 6.879160370359078e-07, "loss": 0.2764, "step": 8102 }, { "epoch": 1.53, "learning_rate": 6.873891646704767e-07, "loss": 0.2946, "step": 8103 }, { "epoch": 1.53, "learning_rate": 6.868624619858022e-07, "loss": 0.2385, "step": 8104 }, { "epoch": 1.53, "learning_rate": 6.863359290311894e-07, "loss": 0.3321, "step": 8105 }, { "epoch": 1.53, "learning_rate": 6.858095658559302e-07, "loss": 0.1813, "step": 8106 }, { "epoch": 1.53, "learning_rate": 6.852833725092959e-07, "loss": 0.2113, "step": 8107 }, { "epoch": 1.53, "learning_rate": 6.847573490405435e-07, "loss": 0.2764, "step": 8108 }, { "epoch": 1.53, "learning_rate": 6.842314954989176e-07, "loss": 0.2048, "step": 8109 }, { "epoch": 1.53, "learning_rate": 6.837058119336418e-07, "loss": 0.2143, "step": 8110 }, { "epoch": 1.53, "learning_rate": 6.831802983939257e-07, "loss": 0.2621, "step": 8111 }, { "epoch": 1.53, "learning_rate": 6.826549549289654e-07, "loss": 0.2087, "step": 8112 }, { "epoch": 1.53, "learning_rate": 6.821297815879363e-07, "loss": 0.1863, "step": 8113 }, { "epoch": 1.53, "learning_rate": 6.816047784200028e-07, "loss": 0.2267, "step": 8114 }, { "epoch": 1.53, "learning_rate": 6.8107994547431e-07, "loss": 0.1994, "step": 8115 }, { "epoch": 1.53, "learning_rate": 6.805552827999884e-07, "loss": 0.277, "step": 8116 }, { "epoch": 1.53, "learning_rate": 6.800307904461523e-07, "loss": 0.2958, "step": 8117 }, { "epoch": 1.53, "learning_rate": 6.795064684618999e-07, "loss": 0.1931, "step": 8118 }, { "epoch": 1.53, "learning_rate": 6.789823168963138e-07, "loss": 0.2196, "step": 8119 }, { "epoch": 1.53, "learning_rate": 6.784583357984606e-07, "loss": 0.1931, "step": 8120 }, { "epoch": 1.53, "learning_rate": 6.779345252173908e-07, "loss": 0.2393, "step": 8121 }, { "epoch": 1.53, "learning_rate": 6.774108852021385e-07, "loss": 0.2373, "step": 8122 }, { "epoch": 1.53, "learning_rate": 6.76887415801723e-07, "loss": 0.2663, "step": 8123 }, { "epoch": 1.53, "learning_rate": 6.763641170651464e-07, "loss": 0.232, "step": 8124 }, { "epoch": 1.54, "learning_rate": 6.758409890413955e-07, "loss": 0.2541, "step": 8125 }, { "epoch": 1.54, "learning_rate": 6.753180317794411e-07, "loss": 0.1626, "step": 8126 }, { "epoch": 1.54, "learning_rate": 6.747952453282378e-07, "loss": 0.1837, "step": 8127 }, { "epoch": 1.54, "learning_rate": 6.742726297367241e-07, "loss": 0.1531, "step": 8128 }, { "epoch": 1.54, "learning_rate": 6.737501850538227e-07, "loss": 0.2196, "step": 8129 }, { "epoch": 1.54, "learning_rate": 6.732279113284409e-07, "loss": 0.2387, "step": 8130 }, { "epoch": 1.54, "learning_rate": 6.727058086094687e-07, "loss": 0.2157, "step": 8131 }, { "epoch": 1.54, "learning_rate": 6.721838769457812e-07, "loss": 0.2526, "step": 8132 }, { "epoch": 1.54, "learning_rate": 6.716621163862358e-07, "loss": 0.2336, "step": 8133 }, { "epoch": 1.54, "learning_rate": 6.711405269796781e-07, "loss": 0.2799, "step": 8134 }, { "epoch": 1.54, "learning_rate": 6.70619108774932e-07, "loss": 0.2362, "step": 8135 }, { "epoch": 1.54, "learning_rate": 6.700978618208082e-07, "loss": 0.1979, "step": 8136 }, { "epoch": 1.54, "learning_rate": 6.695767861661035e-07, "loss": 0.33, "step": 8137 }, { "epoch": 1.54, "learning_rate": 6.690558818595944e-07, "loss": 0.2653, "step": 8138 }, { "epoch": 1.54, "learning_rate": 6.685351489500433e-07, "loss": 0.2127, "step": 8139 }, { "epoch": 1.54, "learning_rate": 6.680145874861988e-07, "loss": 0.2071, "step": 8140 }, { "epoch": 1.54, "learning_rate": 6.674941975167892e-07, "loss": 0.2056, "step": 8141 }, { "epoch": 1.54, "learning_rate": 6.669739790905285e-07, "loss": 0.2734, "step": 8142 }, { "epoch": 1.54, "learning_rate": 6.664539322561178e-07, "loss": 0.1985, "step": 8143 }, { "epoch": 1.54, "learning_rate": 6.659340570622361e-07, "loss": 0.2656, "step": 8144 }, { "epoch": 1.54, "learning_rate": 6.654143535575519e-07, "loss": 0.2212, "step": 8145 }, { "epoch": 1.54, "learning_rate": 6.648948217907147e-07, "loss": 0.1828, "step": 8146 }, { "epoch": 1.54, "learning_rate": 6.64375461810357e-07, "loss": 0.2844, "step": 8147 }, { "epoch": 1.54, "learning_rate": 6.638562736650988e-07, "loss": 0.2686, "step": 8148 }, { "epoch": 1.54, "learning_rate": 6.633372574035419e-07, "loss": 0.1983, "step": 8149 }, { "epoch": 1.54, "learning_rate": 6.628184130742696e-07, "loss": 0.1934, "step": 8150 }, { "epoch": 1.54, "learning_rate": 6.622997407258541e-07, "loss": 0.1867, "step": 8151 }, { "epoch": 1.54, "learning_rate": 6.617812404068485e-07, "loss": 0.1474, "step": 8152 }, { "epoch": 1.54, "learning_rate": 6.612629121657888e-07, "loss": 0.2013, "step": 8153 }, { "epoch": 1.54, "learning_rate": 6.607447560511982e-07, "loss": 0.2619, "step": 8154 }, { "epoch": 1.54, "learning_rate": 6.602267721115807e-07, "loss": 0.2368, "step": 8155 }, { "epoch": 1.54, "learning_rate": 6.597089603954262e-07, "loss": 0.2098, "step": 8156 }, { "epoch": 1.54, "learning_rate": 6.591913209512075e-07, "loss": 0.2641, "step": 8157 }, { "epoch": 1.54, "learning_rate": 6.58673853827381e-07, "loss": 0.2459, "step": 8158 }, { "epoch": 1.54, "learning_rate": 6.581565590723879e-07, "loss": 0.2187, "step": 8159 }, { "epoch": 1.54, "learning_rate": 6.576394367346528e-07, "loss": 0.2003, "step": 8160 }, { "epoch": 1.54, "learning_rate": 6.571224868625839e-07, "loss": 0.2126, "step": 8161 }, { "epoch": 1.54, "learning_rate": 6.566057095045739e-07, "loss": 0.2355, "step": 8162 }, { "epoch": 1.54, "learning_rate": 6.560891047089987e-07, "loss": 0.195, "step": 8163 }, { "epoch": 1.54, "learning_rate": 6.555726725242181e-07, "loss": 0.1896, "step": 8164 }, { "epoch": 1.54, "learning_rate": 6.550564129985765e-07, "loss": 0.2222, "step": 8165 }, { "epoch": 1.54, "learning_rate": 6.545403261804012e-07, "loss": 0.2772, "step": 8166 }, { "epoch": 1.54, "learning_rate": 6.540244121180039e-07, "loss": 0.3846, "step": 8167 }, { "epoch": 1.54, "learning_rate": 6.5350867085968e-07, "loss": 0.2578, "step": 8168 }, { "epoch": 1.54, "learning_rate": 6.529931024537086e-07, "loss": 0.2759, "step": 8169 }, { "epoch": 1.54, "learning_rate": 6.524777069483526e-07, "loss": 0.186, "step": 8170 }, { "epoch": 1.54, "learning_rate": 6.51962484391859e-07, "loss": 0.2488, "step": 8171 }, { "epoch": 1.54, "learning_rate": 6.514474348324582e-07, "loss": 0.199, "step": 8172 }, { "epoch": 1.54, "learning_rate": 6.509325583183648e-07, "loss": 0.1461, "step": 8173 }, { "epoch": 1.54, "learning_rate": 6.504178548977771e-07, "loss": 0.1499, "step": 8174 }, { "epoch": 1.54, "learning_rate": 6.499033246188763e-07, "loss": 0.1928, "step": 8175 }, { "epoch": 1.54, "learning_rate": 6.493889675298296e-07, "loss": 0.2106, "step": 8176 }, { "epoch": 1.54, "learning_rate": 6.488747836787867e-07, "loss": 0.2098, "step": 8177 }, { "epoch": 1.55, "learning_rate": 6.483607731138788e-07, "loss": 0.246, "step": 8178 }, { "epoch": 1.55, "learning_rate": 6.478469358832254e-07, "loss": 0.3189, "step": 8179 }, { "epoch": 1.55, "learning_rate": 6.473332720349271e-07, "loss": 0.226, "step": 8180 }, { "epoch": 1.55, "learning_rate": 6.468197816170665e-07, "loss": 0.2199, "step": 8181 }, { "epoch": 1.55, "learning_rate": 6.463064646777146e-07, "loss": 0.2227, "step": 8182 }, { "epoch": 1.55, "learning_rate": 6.457933212649225e-07, "loss": 0.3389, "step": 8183 }, { "epoch": 1.55, "learning_rate": 6.452803514267267e-07, "loss": 0.2376, "step": 8184 }, { "epoch": 1.55, "learning_rate": 6.447675552111462e-07, "loss": 0.2034, "step": 8185 }, { "epoch": 1.55, "learning_rate": 6.442549326661854e-07, "loss": 0.2118, "step": 8186 }, { "epoch": 1.55, "learning_rate": 6.437424838398307e-07, "loss": 0.2566, "step": 8187 }, { "epoch": 1.55, "learning_rate": 6.432302087800535e-07, "loss": 0.3312, "step": 8188 }, { "epoch": 1.55, "learning_rate": 6.427181075348085e-07, "loss": 0.2319, "step": 8189 }, { "epoch": 1.55, "learning_rate": 6.422061801520341e-07, "loss": 0.2232, "step": 8190 }, { "epoch": 1.55, "learning_rate": 6.416944266796521e-07, "loss": 0.1867, "step": 8191 }, { "epoch": 1.55, "learning_rate": 6.41182847165569e-07, "loss": 0.2443, "step": 8192 }, { "epoch": 1.55, "learning_rate": 6.40671441657674e-07, "loss": 0.3133, "step": 8193 }, { "epoch": 1.55, "learning_rate": 6.401602102038407e-07, "loss": 0.1988, "step": 8194 }, { "epoch": 1.55, "learning_rate": 6.396491528519255e-07, "loss": 0.2183, "step": 8195 }, { "epoch": 1.55, "learning_rate": 6.391382696497699e-07, "loss": 0.2313, "step": 8196 }, { "epoch": 1.55, "learning_rate": 6.386275606451981e-07, "loss": 0.2441, "step": 8197 }, { "epoch": 1.55, "learning_rate": 6.381170258860175e-07, "loss": 0.1848, "step": 8198 }, { "epoch": 1.55, "learning_rate": 6.376066654200209e-07, "loss": 0.2452, "step": 8199 }, { "epoch": 1.55, "learning_rate": 6.370964792949833e-07, "loss": 0.2959, "step": 8200 }, { "epoch": 1.55, "learning_rate": 6.365864675586636e-07, "loss": 0.1559, "step": 8201 }, { "epoch": 1.55, "learning_rate": 6.360766302588051e-07, "loss": 0.187, "step": 8202 }, { "epoch": 1.55, "learning_rate": 6.355669674431334e-07, "loss": 0.2466, "step": 8203 }, { "epoch": 1.55, "learning_rate": 6.350574791593609e-07, "loss": 0.2189, "step": 8204 }, { "epoch": 1.55, "learning_rate": 6.345481654551791e-07, "loss": 0.2592, "step": 8205 }, { "epoch": 1.55, "learning_rate": 6.340390263782656e-07, "loss": 0.2308, "step": 8206 }, { "epoch": 1.55, "learning_rate": 6.335300619762838e-07, "loss": 0.2317, "step": 8207 }, { "epoch": 1.55, "learning_rate": 6.330212722968763e-07, "loss": 0.2008, "step": 8208 }, { "epoch": 1.55, "learning_rate": 6.32512657387671e-07, "loss": 0.2021, "step": 8209 }, { "epoch": 1.55, "learning_rate": 6.320042172962831e-07, "loss": 0.1797, "step": 8210 }, { "epoch": 1.55, "learning_rate": 6.314959520703054e-07, "loss": 0.2431, "step": 8211 }, { "epoch": 1.55, "learning_rate": 6.309878617573175e-07, "loss": 0.2631, "step": 8212 }, { "epoch": 1.55, "learning_rate": 6.304799464048846e-07, "loss": 0.3144, "step": 8213 }, { "epoch": 1.55, "learning_rate": 6.299722060605503e-07, "loss": 0.1834, "step": 8214 }, { "epoch": 1.55, "learning_rate": 6.294646407718471e-07, "loss": 0.2264, "step": 8215 }, { "epoch": 1.55, "learning_rate": 6.289572505862887e-07, "loss": 0.1881, "step": 8216 }, { "epoch": 1.55, "learning_rate": 6.284500355513704e-07, "loss": 0.2307, "step": 8217 }, { "epoch": 1.55, "learning_rate": 6.279429957145753e-07, "loss": 0.224, "step": 8218 }, { "epoch": 1.55, "learning_rate": 6.274361311233679e-07, "loss": 0.1674, "step": 8219 }, { "epoch": 1.55, "learning_rate": 6.269294418251945e-07, "loss": 0.1854, "step": 8220 }, { "epoch": 1.55, "learning_rate": 6.264229278674893e-07, "loss": 0.2109, "step": 8221 }, { "epoch": 1.55, "learning_rate": 6.259165892976673e-07, "loss": 0.1985, "step": 8222 }, { "epoch": 1.55, "learning_rate": 6.254104261631255e-07, "loss": 0.2122, "step": 8223 }, { "epoch": 1.55, "learning_rate": 6.249044385112488e-07, "loss": 0.2452, "step": 8224 }, { "epoch": 1.55, "learning_rate": 6.24398626389402e-07, "loss": 0.2607, "step": 8225 }, { "epoch": 1.55, "learning_rate": 6.238929898449355e-07, "loss": 0.1862, "step": 8226 }, { "epoch": 1.55, "learning_rate": 6.23387528925182e-07, "loss": 0.2289, "step": 8227 }, { "epoch": 1.55, "learning_rate": 6.228822436774589e-07, "loss": 0.2685, "step": 8228 }, { "epoch": 1.55, "learning_rate": 6.223771341490662e-07, "loss": 0.2579, "step": 8229 }, { "epoch": 1.55, "learning_rate": 6.218722003872879e-07, "loss": 0.1922, "step": 8230 }, { "epoch": 1.56, "learning_rate": 6.213674424393914e-07, "loss": 0.1726, "step": 8231 }, { "epoch": 1.56, "learning_rate": 6.208628603526279e-07, "loss": 0.2555, "step": 8232 }, { "epoch": 1.56, "learning_rate": 6.203584541742319e-07, "loss": 0.291, "step": 8233 }, { "epoch": 1.56, "learning_rate": 6.198542239514213e-07, "loss": 0.2204, "step": 8234 }, { "epoch": 1.56, "learning_rate": 6.19350169731398e-07, "loss": 0.3797, "step": 8235 }, { "epoch": 1.56, "learning_rate": 6.18846291561347e-07, "loss": 0.2596, "step": 8236 }, { "epoch": 1.56, "learning_rate": 6.183425894884371e-07, "loss": 0.2419, "step": 8237 }, { "epoch": 1.56, "learning_rate": 6.1783906355982e-07, "loss": 0.2114, "step": 8238 }, { "epoch": 1.56, "learning_rate": 6.173357138226321e-07, "loss": 0.1921, "step": 8239 }, { "epoch": 1.56, "learning_rate": 6.168325403239913e-07, "loss": 0.1434, "step": 8240 }, { "epoch": 1.56, "learning_rate": 6.163295431110028e-07, "loss": 0.1625, "step": 8241 }, { "epoch": 1.56, "learning_rate": 6.158267222307507e-07, "loss": 0.2381, "step": 8242 }, { "epoch": 1.56, "learning_rate": 6.153240777303043e-07, "loss": 0.2004, "step": 8243 }, { "epoch": 1.56, "learning_rate": 6.148216096567192e-07, "loss": 0.2708, "step": 8244 }, { "epoch": 1.56, "learning_rate": 6.143193180570289e-07, "loss": 0.293, "step": 8245 }, { "epoch": 1.56, "learning_rate": 6.138172029782563e-07, "loss": 0.1822, "step": 8246 }, { "epoch": 1.56, "learning_rate": 6.133152644674045e-07, "loss": 0.2145, "step": 8247 }, { "epoch": 1.56, "learning_rate": 6.128135025714584e-07, "loss": 0.1702, "step": 8248 }, { "epoch": 1.56, "learning_rate": 6.123119173373912e-07, "loss": 0.2535, "step": 8249 }, { "epoch": 1.56, "learning_rate": 6.118105088121565e-07, "loss": 0.2835, "step": 8250 }, { "epoch": 1.56, "learning_rate": 6.113092770426896e-07, "loss": 0.244, "step": 8251 }, { "epoch": 1.56, "learning_rate": 6.108082220759138e-07, "loss": 0.3427, "step": 8252 }, { "epoch": 1.56, "learning_rate": 6.103073439587335e-07, "loss": 0.306, "step": 8253 }, { "epoch": 1.56, "learning_rate": 6.098066427380342e-07, "loss": 0.2028, "step": 8254 }, { "epoch": 1.56, "learning_rate": 6.093061184606894e-07, "loss": 0.2702, "step": 8255 }, { "epoch": 1.56, "learning_rate": 6.088057711735529e-07, "loss": 0.1715, "step": 8256 }, { "epoch": 1.56, "learning_rate": 6.083056009234631e-07, "loss": 0.2696, "step": 8257 }, { "epoch": 1.56, "learning_rate": 6.078056077572414e-07, "loss": 0.1935, "step": 8258 }, { "epoch": 1.56, "learning_rate": 6.073057917216929e-07, "loss": 0.3004, "step": 8259 }, { "epoch": 1.56, "learning_rate": 6.06806152863606e-07, "loss": 0.2376, "step": 8260 }, { "epoch": 1.56, "learning_rate": 6.063066912297524e-07, "loss": 0.2853, "step": 8261 }, { "epoch": 1.56, "learning_rate": 6.058074068668876e-07, "loss": 0.2909, "step": 8262 }, { "epoch": 1.56, "learning_rate": 6.053082998217496e-07, "loss": 0.184, "step": 8263 }, { "epoch": 1.56, "learning_rate": 6.048093701410612e-07, "loss": 0.1886, "step": 8264 }, { "epoch": 1.56, "learning_rate": 6.043106178715274e-07, "loss": 0.2501, "step": 8265 }, { "epoch": 1.56, "learning_rate": 6.038120430598371e-07, "loss": 0.2748, "step": 8266 }, { "epoch": 1.56, "learning_rate": 6.033136457526623e-07, "loss": 0.2652, "step": 8267 }, { "epoch": 1.56, "learning_rate": 6.02815425996659e-07, "loss": 0.2573, "step": 8268 }, { "epoch": 1.56, "learning_rate": 6.02317383838466e-07, "loss": 0.1575, "step": 8269 }, { "epoch": 1.56, "learning_rate": 6.018195193247056e-07, "loss": 0.2007, "step": 8270 }, { "epoch": 1.56, "learning_rate": 6.013218325019835e-07, "loss": 0.2687, "step": 8271 }, { "epoch": 1.56, "learning_rate": 6.008243234168887e-07, "loss": 0.1912, "step": 8272 }, { "epoch": 1.56, "learning_rate": 6.003269921159937e-07, "loss": 0.2478, "step": 8273 }, { "epoch": 1.56, "learning_rate": 5.998298386458546e-07, "loss": 0.1895, "step": 8274 }, { "epoch": 1.56, "learning_rate": 5.993328630530101e-07, "loss": 0.2304, "step": 8275 }, { "epoch": 1.56, "learning_rate": 5.988360653839825e-07, "loss": 0.1611, "step": 8276 }, { "epoch": 1.56, "learning_rate": 5.983394456852792e-07, "loss": 0.2111, "step": 8277 }, { "epoch": 1.56, "learning_rate": 5.978430040033878e-07, "loss": 0.1912, "step": 8278 }, { "epoch": 1.56, "learning_rate": 5.973467403847805e-07, "loss": 0.264, "step": 8279 }, { "epoch": 1.56, "learning_rate": 5.968506548759154e-07, "loss": 0.239, "step": 8280 }, { "epoch": 1.56, "learning_rate": 5.963547475232293e-07, "loss": 0.265, "step": 8281 }, { "epoch": 1.56, "learning_rate": 5.95859018373145e-07, "loss": 0.1398, "step": 8282 }, { "epoch": 1.56, "learning_rate": 5.953634674720704e-07, "loss": 0.2502, "step": 8283 }, { "epoch": 1.57, "learning_rate": 5.948680948663926e-07, "loss": 0.1817, "step": 8284 }, { "epoch": 1.57, "learning_rate": 5.943729006024837e-07, "loss": 0.1798, "step": 8285 }, { "epoch": 1.57, "learning_rate": 5.938778847267021e-07, "loss": 0.272, "step": 8286 }, { "epoch": 1.57, "learning_rate": 5.933830472853835e-07, "loss": 0.253, "step": 8287 }, { "epoch": 1.57, "learning_rate": 5.928883883248526e-07, "loss": 0.1741, "step": 8288 }, { "epoch": 1.57, "learning_rate": 5.923939078914151e-07, "loss": 0.1423, "step": 8289 }, { "epoch": 1.57, "learning_rate": 5.918996060313578e-07, "loss": 0.2089, "step": 8290 }, { "epoch": 1.57, "learning_rate": 5.914054827909549e-07, "loss": 0.2081, "step": 8291 }, { "epoch": 1.57, "learning_rate": 5.909115382164618e-07, "loss": 0.2309, "step": 8292 }, { "epoch": 1.57, "learning_rate": 5.904177723541154e-07, "loss": 0.1879, "step": 8293 }, { "epoch": 1.57, "learning_rate": 5.8992418525014e-07, "loss": 0.2793, "step": 8294 }, { "epoch": 1.57, "learning_rate": 5.894307769507404e-07, "loss": 0.1988, "step": 8295 }, { "epoch": 1.57, "learning_rate": 5.889375475021036e-07, "loss": 0.2893, "step": 8296 }, { "epoch": 1.57, "learning_rate": 5.88444496950403e-07, "loss": 0.2402, "step": 8297 }, { "epoch": 1.57, "learning_rate": 5.879516253417932e-07, "loss": 0.2085, "step": 8298 }, { "epoch": 1.57, "learning_rate": 5.874589327224128e-07, "loss": 0.3084, "step": 8299 }, { "epoch": 1.57, "learning_rate": 5.869664191383831e-07, "loss": 0.2195, "step": 8300 }, { "epoch": 1.57, "learning_rate": 5.864740846358089e-07, "loss": 0.244, "step": 8301 }, { "epoch": 1.57, "learning_rate": 5.859819292607782e-07, "loss": 0.248, "step": 8302 }, { "epoch": 1.57, "learning_rate": 5.854899530593624e-07, "loss": 0.2361, "step": 8303 }, { "epoch": 1.57, "learning_rate": 5.849981560776161e-07, "loss": 0.1789, "step": 8304 }, { "epoch": 1.57, "learning_rate": 5.845065383615766e-07, "loss": 0.2115, "step": 8305 }, { "epoch": 1.57, "learning_rate": 5.840150999572653e-07, "loss": 0.2159, "step": 8306 }, { "epoch": 1.57, "learning_rate": 5.835238409106855e-07, "loss": 0.2325, "step": 8307 }, { "epoch": 1.57, "learning_rate": 5.830327612678266e-07, "loss": 0.2487, "step": 8308 }, { "epoch": 1.57, "learning_rate": 5.825418610746572e-07, "loss": 0.2427, "step": 8309 }, { "epoch": 1.57, "learning_rate": 5.820511403771312e-07, "loss": 0.2615, "step": 8310 }, { "epoch": 1.57, "learning_rate": 5.815605992211873e-07, "loss": 0.2709, "step": 8311 }, { "epoch": 1.57, "learning_rate": 5.810702376527439e-07, "loss": 0.301, "step": 8312 }, { "epoch": 1.57, "learning_rate": 5.805800557177041e-07, "loss": 0.2386, "step": 8313 }, { "epoch": 1.57, "learning_rate": 5.800900534619566e-07, "loss": 0.1673, "step": 8314 }, { "epoch": 1.57, "learning_rate": 5.796002309313692e-07, "loss": 0.2129, "step": 8315 }, { "epoch": 1.57, "learning_rate": 5.791105881717949e-07, "loss": 0.2319, "step": 8316 }, { "epoch": 1.57, "learning_rate": 5.786211252290713e-07, "loss": 0.1716, "step": 8317 }, { "epoch": 1.57, "learning_rate": 5.781318421490153e-07, "loss": 0.1633, "step": 8318 }, { "epoch": 1.57, "learning_rate": 5.776427389774313e-07, "loss": 0.2041, "step": 8319 }, { "epoch": 1.57, "learning_rate": 5.77153815760105e-07, "loss": 0.3138, "step": 8320 }, { "epoch": 1.57, "learning_rate": 5.766650725428027e-07, "loss": 0.2302, "step": 8321 }, { "epoch": 1.57, "learning_rate": 5.761765093712785e-07, "loss": 0.2086, "step": 8322 }, { "epoch": 1.57, "learning_rate": 5.756881262912675e-07, "loss": 0.2183, "step": 8323 }, { "epoch": 1.57, "learning_rate": 5.751999233484856e-07, "loss": 0.2094, "step": 8324 }, { "epoch": 1.57, "learning_rate": 5.747119005886362e-07, "loss": 0.2422, "step": 8325 }, { "epoch": 1.57, "learning_rate": 5.74224058057404e-07, "loss": 0.1431, "step": 8326 }, { "epoch": 1.57, "learning_rate": 5.737363958004541e-07, "loss": 0.2297, "step": 8327 }, { "epoch": 1.57, "learning_rate": 5.732489138634395e-07, "loss": 0.2327, "step": 8328 }, { "epoch": 1.57, "learning_rate": 5.727616122919929e-07, "loss": 0.2406, "step": 8329 }, { "epoch": 1.57, "learning_rate": 5.722744911317318e-07, "loss": 0.2298, "step": 8330 }, { "epoch": 1.57, "learning_rate": 5.717875504282557e-07, "loss": 0.2564, "step": 8331 }, { "epoch": 1.57, "learning_rate": 5.71300790227148e-07, "loss": 0.1795, "step": 8332 }, { "epoch": 1.57, "learning_rate": 5.70814210573975e-07, "loss": 0.2553, "step": 8333 }, { "epoch": 1.57, "learning_rate": 5.703278115142857e-07, "loss": 0.1684, "step": 8334 }, { "epoch": 1.57, "learning_rate": 5.698415930936127e-07, "loss": 0.2749, "step": 8335 }, { "epoch": 1.57, "learning_rate": 5.693555553574718e-07, "loss": 0.301, "step": 8336 }, { "epoch": 1.58, "learning_rate": 5.688696983513612e-07, "loss": 0.1625, "step": 8337 }, { "epoch": 1.58, "learning_rate": 5.683840221207626e-07, "loss": 0.2251, "step": 8338 }, { "epoch": 1.58, "learning_rate": 5.678985267111409e-07, "loss": 0.2312, "step": 8339 }, { "epoch": 1.58, "learning_rate": 5.674132121679438e-07, "loss": 0.2158, "step": 8340 }, { "epoch": 1.58, "learning_rate": 5.669280785366024e-07, "loss": 0.2674, "step": 8341 }, { "epoch": 1.58, "learning_rate": 5.664431258625305e-07, "loss": 0.3157, "step": 8342 }, { "epoch": 1.58, "learning_rate": 5.659583541911253e-07, "loss": 0.2805, "step": 8343 }, { "epoch": 1.58, "learning_rate": 5.654737635677665e-07, "loss": 0.2151, "step": 8344 }, { "epoch": 1.58, "learning_rate": 5.649893540378179e-07, "loss": 0.2771, "step": 8345 }, { "epoch": 1.58, "learning_rate": 5.645051256466241e-07, "loss": 0.1993, "step": 8346 }, { "epoch": 1.58, "learning_rate": 5.640210784395173e-07, "loss": 0.1609, "step": 8347 }, { "epoch": 1.58, "learning_rate": 5.635372124618069e-07, "loss": 0.2577, "step": 8348 }, { "epoch": 1.58, "learning_rate": 5.630535277587887e-07, "loss": 0.182, "step": 8349 }, { "epoch": 1.58, "learning_rate": 5.625700243757432e-07, "loss": 0.2531, "step": 8350 }, { "epoch": 1.58, "learning_rate": 5.62086702357929e-07, "loss": 0.2216, "step": 8351 }, { "epoch": 1.58, "learning_rate": 5.616035617505911e-07, "loss": 0.24, "step": 8352 }, { "epoch": 1.58, "learning_rate": 5.611206025989588e-07, "loss": 0.2278, "step": 8353 }, { "epoch": 1.58, "learning_rate": 5.606378249482405e-07, "loss": 0.2421, "step": 8354 }, { "epoch": 1.58, "learning_rate": 5.601552288436293e-07, "loss": 0.1895, "step": 8355 }, { "epoch": 1.58, "learning_rate": 5.596728143303043e-07, "loss": 0.1858, "step": 8356 }, { "epoch": 1.58, "learning_rate": 5.591905814534215e-07, "loss": 0.2225, "step": 8357 }, { "epoch": 1.58, "learning_rate": 5.587085302581257e-07, "loss": 0.1995, "step": 8358 }, { "epoch": 1.58, "learning_rate": 5.582266607895423e-07, "loss": 0.1894, "step": 8359 }, { "epoch": 1.58, "learning_rate": 5.577449730927778e-07, "loss": 0.2277, "step": 8360 }, { "epoch": 1.58, "learning_rate": 5.572634672129251e-07, "loss": 0.2647, "step": 8361 }, { "epoch": 1.58, "learning_rate": 5.567821431950593e-07, "loss": 0.2053, "step": 8362 }, { "epoch": 1.58, "learning_rate": 5.563010010842351e-07, "loss": 0.1677, "step": 8363 }, { "epoch": 1.58, "learning_rate": 5.558200409254952e-07, "loss": 0.2685, "step": 8364 }, { "epoch": 1.58, "learning_rate": 5.553392627638621e-07, "loss": 0.2618, "step": 8365 }, { "epoch": 1.58, "learning_rate": 5.54858666644342e-07, "loss": 0.2516, "step": 8366 }, { "epoch": 1.58, "learning_rate": 5.543782526119241e-07, "loss": 0.1983, "step": 8367 }, { "epoch": 1.58, "learning_rate": 5.538980207115804e-07, "loss": 0.2378, "step": 8368 }, { "epoch": 1.58, "learning_rate": 5.534179709882664e-07, "loss": 0.2418, "step": 8369 }, { "epoch": 1.58, "learning_rate": 5.529381034869199e-07, "loss": 0.2411, "step": 8370 }, { "epoch": 1.58, "learning_rate": 5.52458418252462e-07, "loss": 0.2263, "step": 8371 }, { "epoch": 1.58, "learning_rate": 5.519789153297967e-07, "loss": 0.2489, "step": 8372 }, { "epoch": 1.58, "learning_rate": 5.514995947638105e-07, "loss": 0.2725, "step": 8373 }, { "epoch": 1.58, "learning_rate": 5.510204565993735e-07, "loss": 0.2536, "step": 8374 }, { "epoch": 1.58, "learning_rate": 5.505415008813384e-07, "loss": 0.2271, "step": 8375 }, { "epoch": 1.58, "learning_rate": 5.500627276545406e-07, "loss": 0.1891, "step": 8376 }, { "epoch": 1.58, "learning_rate": 5.495841369637983e-07, "loss": 0.2352, "step": 8377 }, { "epoch": 1.58, "learning_rate": 5.491057288539148e-07, "loss": 0.2376, "step": 8378 }, { "epoch": 1.58, "learning_rate": 5.486275033696728e-07, "loss": 0.2509, "step": 8379 }, { "epoch": 1.58, "learning_rate": 5.481494605558391e-07, "loss": 0.2334, "step": 8380 }, { "epoch": 1.58, "learning_rate": 5.476716004571661e-07, "loss": 0.215, "step": 8381 }, { "epoch": 1.58, "learning_rate": 5.471939231183848e-07, "loss": 0.2781, "step": 8382 }, { "epoch": 1.58, "learning_rate": 5.467164285842114e-07, "loss": 0.2109, "step": 8383 }, { "epoch": 1.58, "learning_rate": 5.462391168993464e-07, "loss": 0.2172, "step": 8384 }, { "epoch": 1.58, "learning_rate": 5.4576198810847e-07, "loss": 0.1979, "step": 8385 }, { "epoch": 1.58, "learning_rate": 5.452850422562461e-07, "loss": 0.3358, "step": 8386 }, { "epoch": 1.58, "learning_rate": 5.448082793873252e-07, "loss": 0.215, "step": 8387 }, { "epoch": 1.58, "learning_rate": 5.443316995463341e-07, "loss": 0.2655, "step": 8388 }, { "epoch": 1.58, "learning_rate": 5.438553027778886e-07, "loss": 0.1922, "step": 8389 }, { "epoch": 1.59, "learning_rate": 5.433790891265847e-07, "loss": 0.183, "step": 8390 }, { "epoch": 1.59, "learning_rate": 5.429030586369991e-07, "loss": 0.1865, "step": 8391 }, { "epoch": 1.59, "learning_rate": 5.424272113536955e-07, "loss": 0.2331, "step": 8392 }, { "epoch": 1.59, "learning_rate": 5.419515473212192e-07, "loss": 0.2392, "step": 8393 }, { "epoch": 1.59, "learning_rate": 5.414760665840949e-07, "loss": 0.2429, "step": 8394 }, { "epoch": 1.59, "learning_rate": 5.410007691868358e-07, "loss": 0.3018, "step": 8395 }, { "epoch": 1.59, "learning_rate": 5.405256551739343e-07, "loss": 0.1589, "step": 8396 }, { "epoch": 1.59, "learning_rate": 5.400507245898647e-07, "loss": 0.276, "step": 8397 }, { "epoch": 1.59, "learning_rate": 5.395759774790882e-07, "loss": 0.1963, "step": 8398 }, { "epoch": 1.59, "learning_rate": 5.391014138860451e-07, "loss": 0.2345, "step": 8399 }, { "epoch": 1.59, "learning_rate": 5.386270338551605e-07, "loss": 0.2005, "step": 8400 }, { "epoch": 1.59, "learning_rate": 5.381528374308414e-07, "loss": 0.2911, "step": 8401 }, { "epoch": 1.59, "learning_rate": 5.376788246574782e-07, "loss": 0.2081, "step": 8402 }, { "epoch": 1.59, "learning_rate": 5.372049955794436e-07, "loss": 0.1515, "step": 8403 }, { "epoch": 1.59, "learning_rate": 5.367313502410931e-07, "loss": 0.2911, "step": 8404 }, { "epoch": 1.59, "learning_rate": 5.362578886867659e-07, "loss": 0.3018, "step": 8405 }, { "epoch": 1.59, "learning_rate": 5.357846109607826e-07, "loss": 0.3052, "step": 8406 }, { "epoch": 1.59, "learning_rate": 5.353115171074478e-07, "loss": 0.1908, "step": 8407 }, { "epoch": 1.59, "learning_rate": 5.348386071710479e-07, "loss": 0.197, "step": 8408 }, { "epoch": 1.59, "learning_rate": 5.343658811958532e-07, "loss": 0.2526, "step": 8409 }, { "epoch": 1.59, "learning_rate": 5.338933392261158e-07, "loss": 0.2326, "step": 8410 }, { "epoch": 1.59, "learning_rate": 5.334209813060712e-07, "loss": 0.2028, "step": 8411 }, { "epoch": 1.59, "learning_rate": 5.329488074799372e-07, "loss": 0.2826, "step": 8412 }, { "epoch": 1.59, "learning_rate": 5.324768177919142e-07, "loss": 0.2006, "step": 8413 }, { "epoch": 1.59, "learning_rate": 5.320050122861866e-07, "loss": 0.3208, "step": 8414 }, { "epoch": 1.59, "learning_rate": 5.315333910069204e-07, "loss": 0.1675, "step": 8415 }, { "epoch": 1.59, "learning_rate": 5.310619539982643e-07, "loss": 0.2566, "step": 8416 }, { "epoch": 1.59, "learning_rate": 5.305907013043502e-07, "loss": 0.2278, "step": 8417 }, { "epoch": 1.59, "learning_rate": 5.301196329692929e-07, "loss": 0.1729, "step": 8418 }, { "epoch": 1.59, "learning_rate": 5.296487490371891e-07, "loss": 0.1874, "step": 8419 }, { "epoch": 1.59, "learning_rate": 5.291780495521204e-07, "loss": 0.2312, "step": 8420 }, { "epoch": 1.59, "learning_rate": 5.287075345581477e-07, "loss": 0.2012, "step": 8421 }, { "epoch": 1.59, "learning_rate": 5.282372040993167e-07, "loss": 0.2188, "step": 8422 }, { "epoch": 1.59, "learning_rate": 5.277670582196575e-07, "loss": 0.2075, "step": 8423 }, { "epoch": 1.59, "learning_rate": 5.272970969631793e-07, "loss": 0.1981, "step": 8424 }, { "epoch": 1.59, "learning_rate": 5.268273203738752e-07, "loss": 0.3025, "step": 8425 }, { "epoch": 1.59, "learning_rate": 5.263577284957236e-07, "loss": 0.2289, "step": 8426 }, { "epoch": 1.59, "learning_rate": 5.258883213726829e-07, "loss": 0.2702, "step": 8427 }, { "epoch": 1.59, "learning_rate": 5.254190990486932e-07, "loss": 0.185, "step": 8428 }, { "epoch": 1.59, "learning_rate": 5.249500615676811e-07, "loss": 0.3265, "step": 8429 }, { "epoch": 1.59, "learning_rate": 5.244812089735529e-07, "loss": 0.2799, "step": 8430 }, { "epoch": 1.59, "learning_rate": 5.240125413101984e-07, "loss": 0.2723, "step": 8431 }, { "epoch": 1.59, "learning_rate": 5.235440586214907e-07, "loss": 0.1993, "step": 8432 }, { "epoch": 1.59, "learning_rate": 5.230757609512846e-07, "loss": 0.2693, "step": 8433 }, { "epoch": 1.59, "learning_rate": 5.226076483434178e-07, "loss": 0.2035, "step": 8434 }, { "epoch": 1.59, "learning_rate": 5.221397208417114e-07, "loss": 0.1468, "step": 8435 }, { "epoch": 1.59, "learning_rate": 5.216719784899684e-07, "loss": 0.2276, "step": 8436 }, { "epoch": 1.59, "learning_rate": 5.21204421331975e-07, "loss": 0.1781, "step": 8437 }, { "epoch": 1.59, "learning_rate": 5.207370494114997e-07, "loss": 0.2683, "step": 8438 }, { "epoch": 1.59, "learning_rate": 5.202698627722935e-07, "loss": 0.2125, "step": 8439 }, { "epoch": 1.59, "learning_rate": 5.198028614580905e-07, "loss": 0.2248, "step": 8440 }, { "epoch": 1.59, "learning_rate": 5.193360455126073e-07, "loss": 0.2656, "step": 8441 }, { "epoch": 1.59, "learning_rate": 5.188694149795434e-07, "loss": 0.2736, "step": 8442 }, { "epoch": 1.6, "learning_rate": 5.184029699025799e-07, "loss": 0.2264, "step": 8443 }, { "epoch": 1.6, "learning_rate": 5.179367103253821e-07, "loss": 0.2858, "step": 8444 }, { "epoch": 1.6, "learning_rate": 5.174706362915968e-07, "loss": 0.2503, "step": 8445 }, { "epoch": 1.6, "learning_rate": 5.170047478448539e-07, "loss": 0.1832, "step": 8446 }, { "epoch": 1.6, "learning_rate": 5.165390450287658e-07, "loss": 0.2513, "step": 8447 }, { "epoch": 1.6, "learning_rate": 5.160735278869272e-07, "loss": 0.2485, "step": 8448 }, { "epoch": 1.6, "learning_rate": 5.156081964629159e-07, "loss": 0.2023, "step": 8449 }, { "epoch": 1.6, "learning_rate": 5.151430508002916e-07, "loss": 0.2808, "step": 8450 }, { "epoch": 1.6, "learning_rate": 5.146780909425992e-07, "loss": 0.2278, "step": 8451 }, { "epoch": 1.6, "learning_rate": 5.142133169333621e-07, "loss": 0.2012, "step": 8452 }, { "epoch": 1.6, "learning_rate": 5.137487288160883e-07, "loss": 0.2343, "step": 8453 }, { "epoch": 1.6, "learning_rate": 5.132843266342705e-07, "loss": 0.2094, "step": 8454 }, { "epoch": 1.6, "learning_rate": 5.128201104313799e-07, "loss": 0.2318, "step": 8455 }, { "epoch": 1.6, "learning_rate": 5.123560802508726e-07, "loss": 0.239, "step": 8456 }, { "epoch": 1.6, "learning_rate": 5.118922361361888e-07, "loss": 0.2056, "step": 8457 }, { "epoch": 1.6, "learning_rate": 5.114285781307477e-07, "loss": 0.2052, "step": 8458 }, { "epoch": 1.6, "learning_rate": 5.109651062779528e-07, "loss": 0.2053, "step": 8459 }, { "epoch": 1.6, "learning_rate": 5.105018206211923e-07, "loss": 0.2229, "step": 8460 }, { "epoch": 1.6, "learning_rate": 5.100387212038324e-07, "loss": 0.3044, "step": 8461 }, { "epoch": 1.6, "learning_rate": 5.095758080692262e-07, "loss": 0.235, "step": 8462 }, { "epoch": 1.6, "learning_rate": 5.091130812607076e-07, "loss": 0.2571, "step": 8463 }, { "epoch": 1.6, "learning_rate": 5.086505408215911e-07, "loss": 0.27, "step": 8464 }, { "epoch": 1.6, "learning_rate": 5.081881867951777e-07, "loss": 0.2975, "step": 8465 }, { "epoch": 1.6, "learning_rate": 5.07726019224749e-07, "loss": 0.2642, "step": 8466 }, { "epoch": 1.6, "learning_rate": 5.072640381535671e-07, "loss": 0.1573, "step": 8467 }, { "epoch": 1.6, "learning_rate": 5.068022436248804e-07, "loss": 0.2772, "step": 8468 }, { "epoch": 1.6, "learning_rate": 5.063406356819181e-07, "loss": 0.3125, "step": 8469 }, { "epoch": 1.6, "learning_rate": 5.058792143678901e-07, "loss": 0.2718, "step": 8470 }, { "epoch": 1.6, "learning_rate": 5.054179797259925e-07, "loss": 0.1579, "step": 8471 }, { "epoch": 1.6, "learning_rate": 5.049569317994013e-07, "loss": 0.2916, "step": 8472 }, { "epoch": 1.6, "learning_rate": 5.04496070631276e-07, "loss": 0.2031, "step": 8473 }, { "epoch": 1.6, "learning_rate": 5.040353962647582e-07, "loss": 0.183, "step": 8474 }, { "epoch": 1.6, "learning_rate": 5.035749087429723e-07, "loss": 0.2279, "step": 8475 }, { "epoch": 1.6, "learning_rate": 5.031146081090249e-07, "loss": 0.2728, "step": 8476 }, { "epoch": 1.6, "learning_rate": 5.026544944060055e-07, "loss": 0.2303, "step": 8477 }, { "epoch": 1.6, "learning_rate": 5.02194567676986e-07, "loss": 0.1851, "step": 8478 }, { "epoch": 1.6, "learning_rate": 5.017348279650205e-07, "loss": 0.2457, "step": 8479 }, { "epoch": 1.6, "learning_rate": 5.012752753131461e-07, "loss": 0.2588, "step": 8480 }, { "epoch": 1.6, "learning_rate": 5.00815909764382e-07, "loss": 0.2623, "step": 8481 }, { "epoch": 1.6, "learning_rate": 5.003567313617294e-07, "loss": 0.2349, "step": 8482 }, { "epoch": 1.6, "learning_rate": 4.998977401481733e-07, "loss": 0.178, "step": 8483 }, { "epoch": 1.6, "learning_rate": 4.994389361666804e-07, "loss": 0.3114, "step": 8484 }, { "epoch": 1.6, "learning_rate": 4.989803194601994e-07, "loss": 0.2819, "step": 8485 }, { "epoch": 1.6, "learning_rate": 4.985218900716627e-07, "loss": 0.2282, "step": 8486 }, { "epoch": 1.6, "learning_rate": 4.980636480439835e-07, "loss": 0.1774, "step": 8487 }, { "epoch": 1.6, "learning_rate": 4.976055934200594e-07, "loss": 0.201, "step": 8488 }, { "epoch": 1.6, "learning_rate": 4.971477262427687e-07, "loss": 0.2068, "step": 8489 }, { "epoch": 1.6, "learning_rate": 4.966900465549726e-07, "loss": 0.2697, "step": 8490 }, { "epoch": 1.6, "learning_rate": 4.96232554399517e-07, "loss": 0.1721, "step": 8491 }, { "epoch": 1.6, "learning_rate": 4.957752498192259e-07, "loss": 0.2201, "step": 8492 }, { "epoch": 1.6, "learning_rate": 4.953181328569095e-07, "loss": 0.2008, "step": 8493 }, { "epoch": 1.6, "learning_rate": 4.948612035553599e-07, "loss": 0.2018, "step": 8494 }, { "epoch": 1.6, "learning_rate": 4.944044619573482e-07, "loss": 0.2714, "step": 8495 }, { "epoch": 1.61, "learning_rate": 4.939479081056328e-07, "loss": 0.2598, "step": 8496 }, { "epoch": 1.61, "learning_rate": 4.934915420429523e-07, "loss": 0.1902, "step": 8497 }, { "epoch": 1.61, "learning_rate": 4.930353638120255e-07, "loss": 0.2083, "step": 8498 }, { "epoch": 1.61, "learning_rate": 4.925793734555581e-07, "loss": 0.2087, "step": 8499 }, { "epoch": 1.61, "learning_rate": 4.921235710162356e-07, "loss": 0.2293, "step": 8500 }, { "epoch": 1.61, "learning_rate": 4.916679565367247e-07, "loss": 0.2045, "step": 8501 }, { "epoch": 1.61, "learning_rate": 4.912125300596776e-07, "loss": 0.1911, "step": 8502 }, { "epoch": 1.61, "learning_rate": 4.907572916277267e-07, "loss": 0.2698, "step": 8503 }, { "epoch": 1.61, "learning_rate": 4.903022412834876e-07, "loss": 0.211, "step": 8504 }, { "epoch": 1.61, "learning_rate": 4.898473790695582e-07, "loss": 0.2544, "step": 8505 }, { "epoch": 1.61, "learning_rate": 4.893927050285186e-07, "loss": 0.2254, "step": 8506 }, { "epoch": 1.61, "learning_rate": 4.889382192029312e-07, "loss": 0.1635, "step": 8507 }, { "epoch": 1.61, "learning_rate": 4.884839216353415e-07, "loss": 0.2494, "step": 8508 }, { "epoch": 1.61, "learning_rate": 4.880298123682764e-07, "loss": 0.1808, "step": 8509 }, { "epoch": 1.61, "learning_rate": 4.875758914442458e-07, "loss": 0.2383, "step": 8510 }, { "epoch": 1.61, "learning_rate": 4.871221589057417e-07, "loss": 0.1531, "step": 8511 }, { "epoch": 1.61, "learning_rate": 4.866686147952388e-07, "loss": 0.2664, "step": 8512 }, { "epoch": 1.61, "learning_rate": 4.862152591551939e-07, "loss": 0.2307, "step": 8513 }, { "epoch": 1.61, "learning_rate": 4.857620920280459e-07, "loss": 0.412, "step": 8514 }, { "epoch": 1.61, "learning_rate": 4.853091134562166e-07, "loss": 0.2383, "step": 8515 }, { "epoch": 1.61, "learning_rate": 4.848563234821096e-07, "loss": 0.2738, "step": 8516 }, { "epoch": 1.61, "learning_rate": 4.844037221481116e-07, "loss": 0.2809, "step": 8517 }, { "epoch": 1.61, "learning_rate": 4.839513094965906e-07, "loss": 0.1991, "step": 8518 }, { "epoch": 1.61, "learning_rate": 4.834990855698981e-07, "loss": 0.1893, "step": 8519 }, { "epoch": 1.61, "learning_rate": 4.830470504103663e-07, "loss": 0.1945, "step": 8520 }, { "epoch": 1.61, "learning_rate": 4.825952040603127e-07, "loss": 0.2088, "step": 8521 }, { "epoch": 1.61, "learning_rate": 4.821435465620336e-07, "loss": 0.2609, "step": 8522 }, { "epoch": 1.61, "learning_rate": 4.816920779578091e-07, "loss": 0.1872, "step": 8523 }, { "epoch": 1.61, "learning_rate": 4.812407982899034e-07, "loss": 0.2031, "step": 8524 }, { "epoch": 1.61, "learning_rate": 4.807897076005596e-07, "loss": 0.343, "step": 8525 }, { "epoch": 1.61, "learning_rate": 4.803388059320049e-07, "loss": 0.2552, "step": 8526 }, { "epoch": 1.61, "learning_rate": 4.798880933264508e-07, "loss": 0.1894, "step": 8527 }, { "epoch": 1.61, "learning_rate": 4.794375698260867e-07, "loss": 0.1723, "step": 8528 }, { "epoch": 1.61, "learning_rate": 4.789872354730873e-07, "loss": 0.3709, "step": 8529 }, { "epoch": 1.61, "learning_rate": 4.785370903096104e-07, "loss": 0.1584, "step": 8530 }, { "epoch": 1.61, "learning_rate": 4.780871343777921e-07, "loss": 0.2651, "step": 8531 }, { "epoch": 1.61, "learning_rate": 4.776373677197557e-07, "loss": 0.2942, "step": 8532 }, { "epoch": 1.61, "learning_rate": 4.77187790377604e-07, "loss": 0.2434, "step": 8533 }, { "epoch": 1.61, "learning_rate": 4.7673840239342043e-07, "loss": 0.1925, "step": 8534 }, { "epoch": 1.61, "learning_rate": 4.762892038092751e-07, "loss": 0.3265, "step": 8535 }, { "epoch": 1.61, "learning_rate": 4.758401946672181e-07, "loss": 0.1186, "step": 8536 }, { "epoch": 1.61, "learning_rate": 4.7539137500927923e-07, "loss": 0.3066, "step": 8537 }, { "epoch": 1.61, "learning_rate": 4.749427448774757e-07, "loss": 0.265, "step": 8538 }, { "epoch": 1.61, "learning_rate": 4.7449430431380395e-07, "loss": 0.2136, "step": 8539 }, { "epoch": 1.61, "learning_rate": 4.740460533602409e-07, "loss": 0.2233, "step": 8540 }, { "epoch": 1.61, "learning_rate": 4.7359799205875023e-07, "loss": 0.2576, "step": 8541 }, { "epoch": 1.61, "learning_rate": 4.731501204512748e-07, "loss": 0.2412, "step": 8542 }, { "epoch": 1.61, "learning_rate": 4.727024385797402e-07, "loss": 0.2645, "step": 8543 }, { "epoch": 1.61, "learning_rate": 4.7225494648605456e-07, "loss": 0.2151, "step": 8544 }, { "epoch": 1.61, "learning_rate": 4.7180764421210834e-07, "loss": 0.184, "step": 8545 }, { "epoch": 1.61, "learning_rate": 4.713605317997741e-07, "loss": 0.2286, "step": 8546 }, { "epoch": 1.61, "learning_rate": 4.709136092909064e-07, "loss": 0.2118, "step": 8547 }, { "epoch": 1.61, "learning_rate": 4.704668767273421e-07, "loss": 0.2351, "step": 8548 }, { "epoch": 1.62, "learning_rate": 4.7002033415090074e-07, "loss": 0.2647, "step": 8549 }, { "epoch": 1.62, "learning_rate": 4.695739816033837e-07, "loss": 0.2525, "step": 8550 }, { "epoch": 1.62, "learning_rate": 4.6912781912657414e-07, "loss": 0.2324, "step": 8551 }, { "epoch": 1.62, "learning_rate": 4.686818467622384e-07, "loss": 0.217, "step": 8552 }, { "epoch": 1.62, "learning_rate": 4.6823606455212416e-07, "loss": 0.274, "step": 8553 }, { "epoch": 1.62, "learning_rate": 4.6779047253796116e-07, "loss": 0.2628, "step": 8554 }, { "epoch": 1.62, "learning_rate": 4.673450707614638e-07, "loss": 0.2299, "step": 8555 }, { "epoch": 1.62, "learning_rate": 4.668998592643245e-07, "loss": 0.3039, "step": 8556 }, { "epoch": 1.62, "learning_rate": 4.6645483808822057e-07, "loss": 0.2301, "step": 8557 }, { "epoch": 1.62, "learning_rate": 4.660100072748125e-07, "loss": 0.2688, "step": 8558 }, { "epoch": 1.62, "learning_rate": 4.655653668657395e-07, "loss": 0.2261, "step": 8559 }, { "epoch": 1.62, "learning_rate": 4.651209169026252e-07, "loss": 0.1915, "step": 8560 }, { "epoch": 1.62, "learning_rate": 4.64676657427077e-07, "loss": 0.197, "step": 8561 }, { "epoch": 1.62, "learning_rate": 4.6423258848067975e-07, "loss": 0.259, "step": 8562 }, { "epoch": 1.62, "learning_rate": 4.637887101050054e-07, "loss": 0.2061, "step": 8563 }, { "epoch": 1.62, "learning_rate": 4.6334502234160614e-07, "loss": 0.2912, "step": 8564 }, { "epoch": 1.62, "learning_rate": 4.629015252320138e-07, "loss": 0.3051, "step": 8565 }, { "epoch": 1.62, "learning_rate": 4.6245821881774675e-07, "loss": 0.1778, "step": 8566 }, { "epoch": 1.62, "learning_rate": 4.620151031403039e-07, "loss": 0.2531, "step": 8567 }, { "epoch": 1.62, "learning_rate": 4.6157217824116334e-07, "loss": 0.2356, "step": 8568 }, { "epoch": 1.62, "learning_rate": 4.6112944416179e-07, "loss": 0.2178, "step": 8569 }, { "epoch": 1.62, "learning_rate": 4.6068690094362875e-07, "loss": 0.2097, "step": 8570 }, { "epoch": 1.62, "learning_rate": 4.6024454862810487e-07, "loss": 0.1818, "step": 8571 }, { "epoch": 1.62, "learning_rate": 4.59802387256629e-07, "loss": 0.275, "step": 8572 }, { "epoch": 1.62, "learning_rate": 4.5936041687059204e-07, "loss": 0.1654, "step": 8573 }, { "epoch": 1.62, "learning_rate": 4.5891863751136746e-07, "loss": 0.2713, "step": 8574 }, { "epoch": 1.62, "learning_rate": 4.5847704922031084e-07, "loss": 0.2259, "step": 8575 }, { "epoch": 1.62, "learning_rate": 4.580356520387594e-07, "loss": 0.3051, "step": 8576 }, { "epoch": 1.62, "learning_rate": 4.5759444600803336e-07, "loss": 0.1839, "step": 8577 }, { "epoch": 1.62, "learning_rate": 4.571534311694342e-07, "loss": 0.2199, "step": 8578 }, { "epoch": 1.62, "learning_rate": 4.567126075642461e-07, "loss": 0.2809, "step": 8579 }, { "epoch": 1.62, "learning_rate": 4.5627197523373494e-07, "loss": 0.2733, "step": 8580 }, { "epoch": 1.62, "learning_rate": 4.558315342191491e-07, "loss": 0.2782, "step": 8581 }, { "epoch": 1.62, "learning_rate": 4.5539128456171864e-07, "loss": 0.1589, "step": 8582 }, { "epoch": 1.62, "learning_rate": 4.549512263026559e-07, "loss": 0.215, "step": 8583 }, { "epoch": 1.62, "learning_rate": 4.545113594831552e-07, "loss": 0.2339, "step": 8584 }, { "epoch": 1.62, "learning_rate": 4.540716841443932e-07, "loss": 0.2929, "step": 8585 }, { "epoch": 1.62, "learning_rate": 4.536322003275281e-07, "loss": 0.2172, "step": 8586 }, { "epoch": 1.62, "learning_rate": 4.5319290807370104e-07, "loss": 0.1961, "step": 8587 }, { "epoch": 1.62, "learning_rate": 4.5275380742403424e-07, "loss": 0.2006, "step": 8588 }, { "epoch": 1.62, "learning_rate": 4.523148984196329e-07, "loss": 0.2948, "step": 8589 }, { "epoch": 1.62, "learning_rate": 4.5187618110158345e-07, "loss": 0.2473, "step": 8590 }, { "epoch": 1.62, "learning_rate": 4.514376555109551e-07, "loss": 0.177, "step": 8591 }, { "epoch": 1.62, "learning_rate": 4.509993216887984e-07, "loss": 0.2487, "step": 8592 }, { "epoch": 1.62, "learning_rate": 4.505611796761461e-07, "loss": 0.1653, "step": 8593 }, { "epoch": 1.62, "learning_rate": 4.5012322951401456e-07, "loss": 0.2486, "step": 8594 }, { "epoch": 1.62, "learning_rate": 4.4968547124339956e-07, "loss": 0.1248, "step": 8595 }, { "epoch": 1.62, "learning_rate": 4.4924790490528e-07, "loss": 0.2566, "step": 8596 }, { "epoch": 1.62, "learning_rate": 4.4881053054061875e-07, "loss": 0.2356, "step": 8597 }, { "epoch": 1.62, "learning_rate": 4.483733481903571e-07, "loss": 0.1748, "step": 8598 }, { "epoch": 1.62, "learning_rate": 4.4793635789542044e-07, "loss": 0.1336, "step": 8599 }, { "epoch": 1.62, "learning_rate": 4.4749955969671773e-07, "loss": 0.2305, "step": 8600 }, { "epoch": 1.62, "learning_rate": 4.4706295363513617e-07, "loss": 0.2151, "step": 8601 }, { "epoch": 1.63, "learning_rate": 4.466265397515471e-07, "loss": 0.2597, "step": 8602 }, { "epoch": 1.63, "learning_rate": 4.461903180868057e-07, "loss": 0.1959, "step": 8603 }, { "epoch": 1.63, "learning_rate": 4.457542886817448e-07, "loss": 0.2103, "step": 8604 }, { "epoch": 1.63, "learning_rate": 4.4531845157718356e-07, "loss": 0.2596, "step": 8605 }, { "epoch": 1.63, "learning_rate": 4.4488280681392075e-07, "loss": 0.1981, "step": 8606 }, { "epoch": 1.63, "learning_rate": 4.444473544327363e-07, "loss": 0.1761, "step": 8607 }, { "epoch": 1.63, "learning_rate": 4.4401209447439503e-07, "loss": 0.1743, "step": 8608 }, { "epoch": 1.63, "learning_rate": 4.4357702697964243e-07, "loss": 0.2259, "step": 8609 }, { "epoch": 1.63, "learning_rate": 4.4314215198920346e-07, "loss": 0.2251, "step": 8610 }, { "epoch": 1.63, "learning_rate": 4.427074695437894e-07, "loss": 0.2219, "step": 8611 }, { "epoch": 1.63, "learning_rate": 4.422729796840916e-07, "loss": 0.1325, "step": 8612 }, { "epoch": 1.63, "learning_rate": 4.4183868245078085e-07, "loss": 0.2262, "step": 8613 }, { "epoch": 1.63, "learning_rate": 4.414045778845144e-07, "loss": 0.2369, "step": 8614 }, { "epoch": 1.63, "learning_rate": 4.409706660259289e-07, "loss": 0.2385, "step": 8615 }, { "epoch": 1.63, "learning_rate": 4.4053694691564313e-07, "loss": 0.2284, "step": 8616 }, { "epoch": 1.63, "learning_rate": 4.401034205942581e-07, "loss": 0.2383, "step": 8617 }, { "epoch": 1.63, "learning_rate": 4.396700871023568e-07, "loss": 0.1544, "step": 8618 }, { "epoch": 1.63, "learning_rate": 4.392369464805041e-07, "loss": 0.2552, "step": 8619 }, { "epoch": 1.63, "learning_rate": 4.388039987692466e-07, "loss": 0.1914, "step": 8620 }, { "epoch": 1.63, "learning_rate": 4.383712440091134e-07, "loss": 0.2163, "step": 8621 }, { "epoch": 1.63, "learning_rate": 4.379386822406151e-07, "loss": 0.1949, "step": 8622 }, { "epoch": 1.63, "learning_rate": 4.3750631350424456e-07, "loss": 0.1961, "step": 8623 }, { "epoch": 1.63, "learning_rate": 4.370741378404753e-07, "loss": 0.1983, "step": 8624 }, { "epoch": 1.63, "learning_rate": 4.3664215528976595e-07, "loss": 0.2524, "step": 8625 }, { "epoch": 1.63, "learning_rate": 4.3621036589255315e-07, "loss": 0.1831, "step": 8626 }, { "epoch": 1.63, "learning_rate": 4.357787696892571e-07, "loss": 0.1953, "step": 8627 }, { "epoch": 1.63, "learning_rate": 4.353473667202818e-07, "loss": 0.2736, "step": 8628 }, { "epoch": 1.63, "learning_rate": 4.3491615702600964e-07, "loss": 0.2319, "step": 8629 }, { "epoch": 1.63, "learning_rate": 4.3448514064680703e-07, "loss": 0.2172, "step": 8630 }, { "epoch": 1.63, "learning_rate": 4.340543176230233e-07, "loss": 0.2545, "step": 8631 }, { "epoch": 1.63, "learning_rate": 4.3362368799498707e-07, "loss": 0.2405, "step": 8632 }, { "epoch": 1.63, "learning_rate": 4.331932518030094e-07, "loss": 0.245, "step": 8633 }, { "epoch": 1.63, "learning_rate": 4.3276300908738635e-07, "loss": 0.173, "step": 8634 }, { "epoch": 1.63, "learning_rate": 4.3233295988839075e-07, "loss": 0.2161, "step": 8635 }, { "epoch": 1.63, "learning_rate": 4.319031042462818e-07, "loss": 0.2505, "step": 8636 }, { "epoch": 1.63, "learning_rate": 4.314734422012992e-07, "loss": 0.2254, "step": 8637 }, { "epoch": 1.63, "learning_rate": 4.3104397379366197e-07, "loss": 0.2073, "step": 8638 }, { "epoch": 1.63, "learning_rate": 4.306146990635754e-07, "loss": 0.1561, "step": 8639 }, { "epoch": 1.63, "learning_rate": 4.30185618051224e-07, "loss": 0.2029, "step": 8640 }, { "epoch": 1.63, "learning_rate": 4.2975673079677287e-07, "loss": 0.1923, "step": 8641 }, { "epoch": 1.63, "learning_rate": 4.293280373403727e-07, "loss": 0.2972, "step": 8642 }, { "epoch": 1.63, "learning_rate": 4.2889953772215387e-07, "loss": 0.2192, "step": 8643 }, { "epoch": 1.63, "learning_rate": 4.284712319822268e-07, "loss": 0.2334, "step": 8644 }, { "epoch": 1.63, "learning_rate": 4.280431201606877e-07, "loss": 0.1995, "step": 8645 }, { "epoch": 1.63, "learning_rate": 4.276152022976121e-07, "loss": 0.1151, "step": 8646 }, { "epoch": 1.63, "learning_rate": 4.271874784330579e-07, "loss": 0.2549, "step": 8647 }, { "epoch": 1.63, "learning_rate": 4.2675994860706477e-07, "loss": 0.1824, "step": 8648 }, { "epoch": 1.63, "learning_rate": 4.2633261285965456e-07, "loss": 0.2789, "step": 8649 }, { "epoch": 1.63, "learning_rate": 4.2590547123083026e-07, "loss": 0.2366, "step": 8650 }, { "epoch": 1.63, "learning_rate": 4.254785237605774e-07, "loss": 0.3602, "step": 8651 }, { "epoch": 1.63, "learning_rate": 4.250517704888632e-07, "loss": 0.2156, "step": 8652 }, { "epoch": 1.63, "learning_rate": 4.2462521145563623e-07, "loss": 0.2211, "step": 8653 }, { "epoch": 1.63, "learning_rate": 4.241988467008276e-07, "loss": 0.1847, "step": 8654 }, { "epoch": 1.64, "learning_rate": 4.237726762643493e-07, "loss": 0.1404, "step": 8655 }, { "epoch": 1.64, "learning_rate": 4.2334670018609635e-07, "loss": 0.1561, "step": 8656 }, { "epoch": 1.64, "learning_rate": 4.2292091850594435e-07, "loss": 0.1944, "step": 8657 }, { "epoch": 1.64, "learning_rate": 4.224953312637517e-07, "loss": 0.294, "step": 8658 }, { "epoch": 1.64, "learning_rate": 4.2206993849935764e-07, "loss": 0.2409, "step": 8659 }, { "epoch": 1.64, "learning_rate": 4.2164474025258416e-07, "loss": 0.2042, "step": 8660 }, { "epoch": 1.64, "learning_rate": 4.2121973656323425e-07, "loss": 0.2518, "step": 8661 }, { "epoch": 1.64, "learning_rate": 4.2079492747109344e-07, "loss": 0.1587, "step": 8662 }, { "epoch": 1.64, "learning_rate": 4.203703130159276e-07, "loss": 0.2291, "step": 8663 }, { "epoch": 1.64, "learning_rate": 4.1994589323748753e-07, "loss": 0.2007, "step": 8664 }, { "epoch": 1.64, "learning_rate": 4.1952166817550184e-07, "loss": 0.3009, "step": 8665 }, { "epoch": 1.64, "learning_rate": 4.190976378696826e-07, "loss": 0.2278, "step": 8666 }, { "epoch": 1.64, "learning_rate": 4.186738023597259e-07, "loss": 0.2648, "step": 8667 }, { "epoch": 1.64, "learning_rate": 4.182501616853055e-07, "loss": 0.3152, "step": 8668 }, { "epoch": 1.64, "learning_rate": 4.1782671588607907e-07, "loss": 0.3032, "step": 8669 }, { "epoch": 1.64, "learning_rate": 4.174034650016873e-07, "loss": 0.1474, "step": 8670 }, { "epoch": 1.64, "learning_rate": 4.1698040907174996e-07, "loss": 0.2187, "step": 8671 }, { "epoch": 1.64, "learning_rate": 4.1655754813586966e-07, "loss": 0.2644, "step": 8672 }, { "epoch": 1.64, "learning_rate": 4.1613488223363253e-07, "loss": 0.2073, "step": 8673 }, { "epoch": 1.64, "learning_rate": 4.1571241140460267e-07, "loss": 0.2418, "step": 8674 }, { "epoch": 1.64, "learning_rate": 4.1529013568832986e-07, "loss": 0.2189, "step": 8675 }, { "epoch": 1.64, "learning_rate": 4.148680551243439e-07, "loss": 0.2361, "step": 8676 }, { "epoch": 1.64, "learning_rate": 4.144461697521543e-07, "loss": 0.1594, "step": 8677 }, { "epoch": 1.64, "learning_rate": 4.1402447961125634e-07, "loss": 0.3166, "step": 8678 }, { "epoch": 1.64, "learning_rate": 4.136029847411241e-07, "loss": 0.2842, "step": 8679 }, { "epoch": 1.64, "learning_rate": 4.1318168518121457e-07, "loss": 0.2219, "step": 8680 }, { "epoch": 1.64, "learning_rate": 4.127605809709656e-07, "loss": 0.2726, "step": 8681 }, { "epoch": 1.64, "learning_rate": 4.123396721497977e-07, "loss": 0.1873, "step": 8682 }, { "epoch": 1.64, "learning_rate": 4.1191895875711265e-07, "loss": 0.2418, "step": 8683 }, { "epoch": 1.64, "learning_rate": 4.114984408322939e-07, "loss": 0.2161, "step": 8684 }, { "epoch": 1.64, "learning_rate": 4.1107811841470643e-07, "loss": 0.2307, "step": 8685 }, { "epoch": 1.64, "learning_rate": 4.1065799154369755e-07, "loss": 0.195, "step": 8686 }, { "epoch": 1.64, "learning_rate": 4.102380602585956e-07, "loss": 0.1996, "step": 8687 }, { "epoch": 1.64, "learning_rate": 4.0981832459871085e-07, "loss": 0.232, "step": 8688 }, { "epoch": 1.64, "learning_rate": 4.0939878460333536e-07, "loss": 0.2689, "step": 8689 }, { "epoch": 1.64, "learning_rate": 4.089794403117428e-07, "loss": 0.2276, "step": 8690 }, { "epoch": 1.64, "learning_rate": 4.085602917631884e-07, "loss": 0.2336, "step": 8691 }, { "epoch": 1.64, "learning_rate": 4.081413389969094e-07, "loss": 0.3686, "step": 8692 }, { "epoch": 1.64, "learning_rate": 4.0772258205212413e-07, "loss": 0.243, "step": 8693 }, { "epoch": 1.64, "learning_rate": 4.073040209680329e-07, "loss": 0.255, "step": 8694 }, { "epoch": 1.64, "learning_rate": 4.0688565578381885e-07, "loss": 0.2393, "step": 8695 }, { "epoch": 1.64, "learning_rate": 4.064674865386445e-07, "loss": 0.264, "step": 8696 }, { "epoch": 1.64, "learning_rate": 4.0604951327165463e-07, "loss": 0.2542, "step": 8697 }, { "epoch": 1.64, "learning_rate": 4.056317360219786e-07, "loss": 0.2563, "step": 8698 }, { "epoch": 1.64, "learning_rate": 4.0521415482872305e-07, "loss": 0.2074, "step": 8699 }, { "epoch": 1.64, "learning_rate": 4.047967697309779e-07, "loss": 0.2572, "step": 8700 }, { "epoch": 1.64, "learning_rate": 4.0437958076781746e-07, "loss": 0.2271, "step": 8701 }, { "epoch": 1.64, "learning_rate": 4.0396258797829324e-07, "loss": 0.1839, "step": 8702 }, { "epoch": 1.64, "learning_rate": 4.035457914014401e-07, "loss": 0.268, "step": 8703 }, { "epoch": 1.64, "learning_rate": 4.0312919107627745e-07, "loss": 0.2691, "step": 8704 }, { "epoch": 1.64, "learning_rate": 4.0271278704180074e-07, "loss": 0.1938, "step": 8705 }, { "epoch": 1.64, "learning_rate": 4.0229657933699235e-07, "loss": 0.2394, "step": 8706 }, { "epoch": 1.65, "learning_rate": 4.0188056800081336e-07, "loss": 0.2332, "step": 8707 }, { "epoch": 1.65, "learning_rate": 4.01464753072206e-07, "loss": 0.2902, "step": 8708 }, { "epoch": 1.65, "learning_rate": 4.010491345900966e-07, "loss": 0.2092, "step": 8709 }, { "epoch": 1.65, "learning_rate": 4.0063371259339176e-07, "loss": 0.2718, "step": 8710 }, { "epoch": 1.65, "learning_rate": 4.0021848712097826e-07, "loss": 0.2112, "step": 8711 }, { "epoch": 1.65, "learning_rate": 3.998034582117269e-07, "loss": 0.2728, "step": 8712 }, { "epoch": 1.65, "learning_rate": 3.993886259044899e-07, "loss": 0.2043, "step": 8713 }, { "epoch": 1.65, "learning_rate": 3.989739902380976e-07, "loss": 0.2231, "step": 8714 }, { "epoch": 1.65, "learning_rate": 3.985595512513671e-07, "loss": 0.2302, "step": 8715 }, { "epoch": 1.65, "learning_rate": 3.981453089830936e-07, "loss": 0.2875, "step": 8716 }, { "epoch": 1.65, "learning_rate": 3.9773126347205483e-07, "loss": 0.1891, "step": 8717 }, { "epoch": 1.65, "learning_rate": 3.9731741475701024e-07, "loss": 0.2895, "step": 8718 }, { "epoch": 1.65, "learning_rate": 3.969037628767006e-07, "loss": 0.2531, "step": 8719 }, { "epoch": 1.65, "learning_rate": 3.964903078698482e-07, "loss": 0.2251, "step": 8720 }, { "epoch": 1.65, "learning_rate": 3.960770497751576e-07, "loss": 0.193, "step": 8721 }, { "epoch": 1.65, "learning_rate": 3.9566398863131377e-07, "loss": 0.2448, "step": 8722 }, { "epoch": 1.65, "learning_rate": 3.9525112447698437e-07, "loss": 0.2076, "step": 8723 }, { "epoch": 1.65, "learning_rate": 3.9483845735081793e-07, "loss": 0.2137, "step": 8724 }, { "epoch": 1.65, "learning_rate": 3.944259872914447e-07, "loss": 0.2172, "step": 8725 }, { "epoch": 1.65, "learning_rate": 3.940137143374767e-07, "loss": 0.2218, "step": 8726 }, { "epoch": 1.65, "learning_rate": 3.936016385275071e-07, "loss": 0.1832, "step": 8727 }, { "epoch": 1.65, "learning_rate": 3.931897599001111e-07, "loss": 0.2263, "step": 8728 }, { "epoch": 1.65, "learning_rate": 3.927780784938448e-07, "loss": 0.1857, "step": 8729 }, { "epoch": 1.65, "learning_rate": 3.923665943472463e-07, "loss": 0.3032, "step": 8730 }, { "epoch": 1.65, "learning_rate": 3.9195530749883554e-07, "loss": 0.1763, "step": 8731 }, { "epoch": 1.65, "learning_rate": 3.9154421798711316e-07, "loss": 0.1913, "step": 8732 }, { "epoch": 1.65, "learning_rate": 3.9113332585056177e-07, "loss": 0.2571, "step": 8733 }, { "epoch": 1.65, "learning_rate": 3.9072263112764576e-07, "loss": 0.2495, "step": 8734 }, { "epoch": 1.65, "learning_rate": 3.9031213385681053e-07, "loss": 0.1728, "step": 8735 }, { "epoch": 1.65, "learning_rate": 3.899018340764829e-07, "loss": 0.1857, "step": 8736 }, { "epoch": 1.65, "learning_rate": 3.894917318250732e-07, "loss": 0.2312, "step": 8737 }, { "epoch": 1.65, "learning_rate": 3.890818271409696e-07, "loss": 0.2209, "step": 8738 }, { "epoch": 1.65, "learning_rate": 3.886721200625443e-07, "loss": 0.1757, "step": 8739 }, { "epoch": 1.65, "learning_rate": 3.8826261062815117e-07, "loss": 0.2897, "step": 8740 }, { "epoch": 1.65, "learning_rate": 3.8785329887612495e-07, "loss": 0.2632, "step": 8741 }, { "epoch": 1.65, "learning_rate": 3.874441848447805e-07, "loss": 0.1616, "step": 8742 }, { "epoch": 1.65, "learning_rate": 3.8703526857241686e-07, "loss": 0.1709, "step": 8743 }, { "epoch": 1.65, "learning_rate": 3.8662655009731323e-07, "loss": 0.3383, "step": 8744 }, { "epoch": 1.65, "learning_rate": 3.8621802945772863e-07, "loss": 0.2334, "step": 8745 }, { "epoch": 1.65, "learning_rate": 3.8580970669190673e-07, "loss": 0.2486, "step": 8746 }, { "epoch": 1.65, "learning_rate": 3.8540158183807077e-07, "loss": 0.2217, "step": 8747 }, { "epoch": 1.65, "learning_rate": 3.84993654934426e-07, "loss": 0.2104, "step": 8748 }, { "epoch": 1.65, "learning_rate": 3.8458592601915837e-07, "loss": 0.1873, "step": 8749 }, { "epoch": 1.65, "learning_rate": 3.841783951304365e-07, "loss": 0.2823, "step": 8750 }, { "epoch": 1.65, "learning_rate": 3.837710623064095e-07, "loss": 0.2627, "step": 8751 }, { "epoch": 1.65, "learning_rate": 3.833639275852083e-07, "loss": 0.288, "step": 8752 }, { "epoch": 1.65, "learning_rate": 3.829569910049455e-07, "loss": 0.2066, "step": 8753 }, { "epoch": 1.65, "learning_rate": 3.8255025260371475e-07, "loss": 0.3099, "step": 8754 }, { "epoch": 1.65, "learning_rate": 3.8214371241959145e-07, "loss": 0.1878, "step": 8755 }, { "epoch": 1.65, "learning_rate": 3.8173737049063215e-07, "loss": 0.3291, "step": 8756 }, { "epoch": 1.65, "learning_rate": 3.813312268548755e-07, "loss": 0.2143, "step": 8757 }, { "epoch": 1.65, "learning_rate": 3.809252815503406e-07, "loss": 0.255, "step": 8758 }, { "epoch": 1.65, "learning_rate": 3.8051953461502864e-07, "loss": 0.2541, "step": 8759 }, { "epoch": 1.66, "learning_rate": 3.8011398608692214e-07, "loss": 0.184, "step": 8760 }, { "epoch": 1.66, "learning_rate": 3.7970863600398506e-07, "loss": 0.3005, "step": 8761 }, { "epoch": 1.66, "learning_rate": 3.7930348440416267e-07, "loss": 0.2028, "step": 8762 }, { "epoch": 1.66, "learning_rate": 3.788985313253815e-07, "loss": 0.2443, "step": 8763 }, { "epoch": 1.66, "learning_rate": 3.7849377680555023e-07, "loss": 0.2091, "step": 8764 }, { "epoch": 1.66, "learning_rate": 3.780892208825579e-07, "loss": 0.2831, "step": 8765 }, { "epoch": 1.66, "learning_rate": 3.776848635942759e-07, "loss": 0.2017, "step": 8766 }, { "epoch": 1.66, "learning_rate": 3.7728070497855595e-07, "loss": 0.1819, "step": 8767 }, { "epoch": 1.66, "learning_rate": 3.7687674507323327e-07, "loss": 0.2, "step": 8768 }, { "epoch": 1.66, "learning_rate": 3.7647298391612156e-07, "loss": 0.2655, "step": 8769 }, { "epoch": 1.66, "learning_rate": 3.760694215450175e-07, "loss": 0.2529, "step": 8770 }, { "epoch": 1.66, "learning_rate": 3.756660579977006e-07, "loss": 0.3306, "step": 8771 }, { "epoch": 1.66, "learning_rate": 3.7526289331192846e-07, "loss": 0.3213, "step": 8772 }, { "epoch": 1.66, "learning_rate": 3.748599275254419e-07, "loss": 0.193, "step": 8773 }, { "epoch": 1.66, "learning_rate": 3.7445716067596506e-07, "loss": 0.3121, "step": 8774 }, { "epoch": 1.66, "learning_rate": 3.7405459280119933e-07, "loss": 0.3139, "step": 8775 }, { "epoch": 1.66, "learning_rate": 3.736522239388299e-07, "loss": 0.2117, "step": 8776 }, { "epoch": 1.66, "learning_rate": 3.7325005412652435e-07, "loss": 0.2642, "step": 8777 }, { "epoch": 1.66, "learning_rate": 3.7284808340192824e-07, "loss": 0.2205, "step": 8778 }, { "epoch": 1.66, "learning_rate": 3.724463118026725e-07, "loss": 0.2634, "step": 8779 }, { "epoch": 1.66, "learning_rate": 3.7204473936636703e-07, "loss": 0.1912, "step": 8780 }, { "epoch": 1.66, "learning_rate": 3.716433661306021e-07, "loss": 0.1946, "step": 8781 }, { "epoch": 1.66, "learning_rate": 3.712421921329523e-07, "loss": 0.2184, "step": 8782 }, { "epoch": 1.66, "learning_rate": 3.708412174109721e-07, "loss": 0.2119, "step": 8783 }, { "epoch": 1.66, "learning_rate": 3.7044044200219566e-07, "loss": 0.189, "step": 8784 }, { "epoch": 1.66, "learning_rate": 3.700398659441415e-07, "loss": 0.2356, "step": 8785 }, { "epoch": 1.66, "learning_rate": 3.6963948927430825e-07, "loss": 0.2572, "step": 8786 }, { "epoch": 1.66, "learning_rate": 3.6923931203017405e-07, "loss": 0.1492, "step": 8787 }, { "epoch": 1.66, "learning_rate": 3.6883933424920136e-07, "loss": 0.2175, "step": 8788 }, { "epoch": 1.66, "learning_rate": 3.684395559688322e-07, "loss": 0.2204, "step": 8789 }, { "epoch": 1.66, "learning_rate": 3.680399772264906e-07, "loss": 0.1995, "step": 8790 }, { "epoch": 1.66, "learning_rate": 3.676405980595812e-07, "loss": 0.2767, "step": 8791 }, { "epoch": 1.66, "learning_rate": 3.6724141850549037e-07, "loss": 0.2827, "step": 8792 }, { "epoch": 1.66, "learning_rate": 3.6684243860158617e-07, "loss": 0.2255, "step": 8793 }, { "epoch": 1.66, "learning_rate": 3.664436583852174e-07, "loss": 0.2424, "step": 8794 }, { "epoch": 1.66, "learning_rate": 3.660450778937144e-07, "loss": 0.2047, "step": 8795 }, { "epoch": 1.66, "learning_rate": 3.656466971643888e-07, "loss": 0.2981, "step": 8796 }, { "epoch": 1.66, "learning_rate": 3.6524851623453345e-07, "loss": 0.2481, "step": 8797 }, { "epoch": 1.66, "learning_rate": 3.6485053514142276e-07, "loss": 0.2341, "step": 8798 }, { "epoch": 1.66, "learning_rate": 3.6445275392231193e-07, "loss": 0.184, "step": 8799 }, { "epoch": 1.66, "learning_rate": 3.6405517261443786e-07, "loss": 0.2043, "step": 8800 }, { "epoch": 1.66, "learning_rate": 3.6365779125501875e-07, "loss": 0.2316, "step": 8801 }, { "epoch": 1.66, "learning_rate": 3.6326060988125384e-07, "loss": 0.2549, "step": 8802 }, { "epoch": 1.66, "learning_rate": 3.628636285303239e-07, "loss": 0.2622, "step": 8803 }, { "epoch": 1.66, "learning_rate": 3.6246684723939e-07, "loss": 0.2234, "step": 8804 }, { "epoch": 1.66, "learning_rate": 3.620702660455974e-07, "loss": 0.266, "step": 8805 }, { "epoch": 1.66, "learning_rate": 3.6167388498606874e-07, "loss": 0.2637, "step": 8806 }, { "epoch": 1.66, "learning_rate": 3.612777040979096e-07, "loss": 0.1813, "step": 8807 }, { "epoch": 1.66, "learning_rate": 3.608817234182088e-07, "loss": 0.2141, "step": 8808 }, { "epoch": 1.66, "learning_rate": 3.604859429840321e-07, "loss": 0.2558, "step": 8809 }, { "epoch": 1.66, "learning_rate": 3.600903628324312e-07, "loss": 0.226, "step": 8810 }, { "epoch": 1.66, "learning_rate": 3.596949830004362e-07, "loss": 0.2039, "step": 8811 }, { "epoch": 1.66, "learning_rate": 3.5929980352505806e-07, "loss": 0.3169, "step": 8812 }, { "epoch": 1.67, "learning_rate": 3.5890482444329103e-07, "loss": 0.1684, "step": 8813 }, { "epoch": 1.67, "learning_rate": 3.5851004579211033e-07, "loss": 0.1637, "step": 8814 }, { "epoch": 1.67, "learning_rate": 3.5811546760846945e-07, "loss": 0.2405, "step": 8815 }, { "epoch": 1.67, "learning_rate": 3.577210899293071e-07, "loss": 0.2794, "step": 8816 }, { "epoch": 1.67, "learning_rate": 3.573269127915416e-07, "loss": 0.2636, "step": 8817 }, { "epoch": 1.67, "learning_rate": 3.5693293623207084e-07, "loss": 0.2019, "step": 8818 }, { "epoch": 1.67, "learning_rate": 3.5653916028777686e-07, "loss": 0.2347, "step": 8819 }, { "epoch": 1.67, "learning_rate": 3.561455849955209e-07, "loss": 0.3025, "step": 8820 }, { "epoch": 1.67, "learning_rate": 3.5575221039214625e-07, "loss": 0.2828, "step": 8821 }, { "epoch": 1.67, "learning_rate": 3.5535903651447714e-07, "loss": 0.2112, "step": 8822 }, { "epoch": 1.67, "learning_rate": 3.5496606339931904e-07, "loss": 0.2083, "step": 8823 }, { "epoch": 1.67, "learning_rate": 3.545732910834584e-07, "loss": 0.2467, "step": 8824 }, { "epoch": 1.67, "learning_rate": 3.541807196036637e-07, "loss": 0.1828, "step": 8825 }, { "epoch": 1.67, "learning_rate": 3.537883489966837e-07, "loss": 0.1943, "step": 8826 }, { "epoch": 1.67, "learning_rate": 3.533961792992485e-07, "loss": 0.2715, "step": 8827 }, { "epoch": 1.67, "learning_rate": 3.5300421054806994e-07, "loss": 0.1846, "step": 8828 }, { "epoch": 1.67, "learning_rate": 3.526124427798408e-07, "loss": 0.1895, "step": 8829 }, { "epoch": 1.67, "learning_rate": 3.5222087603123457e-07, "loss": 0.1954, "step": 8830 }, { "epoch": 1.67, "learning_rate": 3.518295103389063e-07, "loss": 0.2923, "step": 8831 }, { "epoch": 1.67, "learning_rate": 3.5143834573949265e-07, "loss": 0.2697, "step": 8832 }, { "epoch": 1.67, "learning_rate": 3.5104738226961063e-07, "loss": 0.2199, "step": 8833 }, { "epoch": 1.67, "learning_rate": 3.5065661996585897e-07, "loss": 0.2542, "step": 8834 }, { "epoch": 1.67, "learning_rate": 3.5026605886481743e-07, "loss": 0.2627, "step": 8835 }, { "epoch": 1.67, "learning_rate": 3.4987569900304664e-07, "loss": 0.2349, "step": 8836 }, { "epoch": 1.67, "learning_rate": 3.4948554041708856e-07, "loss": 0.1751, "step": 8837 }, { "epoch": 1.67, "learning_rate": 3.4909558314346783e-07, "loss": 0.2676, "step": 8838 }, { "epoch": 1.67, "learning_rate": 3.4870582721868745e-07, "loss": 0.2269, "step": 8839 }, { "epoch": 1.67, "learning_rate": 3.4831627267923245e-07, "loss": 0.212, "step": 8840 }, { "epoch": 1.67, "learning_rate": 3.479269195615714e-07, "loss": 0.177, "step": 8841 }, { "epoch": 1.67, "learning_rate": 3.4753776790215076e-07, "loss": 0.3386, "step": 8842 }, { "epoch": 1.67, "learning_rate": 3.471488177373994e-07, "loss": 0.2256, "step": 8843 }, { "epoch": 1.67, "learning_rate": 3.467600691037287e-07, "loss": 0.2209, "step": 8844 }, { "epoch": 1.67, "learning_rate": 3.4637152203752855e-07, "loss": 0.2464, "step": 8845 }, { "epoch": 1.67, "learning_rate": 3.4598317657517117e-07, "loss": 0.2594, "step": 8846 }, { "epoch": 1.67, "learning_rate": 3.455950327530122e-07, "loss": 0.1867, "step": 8847 }, { "epoch": 1.67, "learning_rate": 3.4520709060738344e-07, "loss": 0.2516, "step": 8848 }, { "epoch": 1.67, "learning_rate": 3.448193501746025e-07, "loss": 0.3032, "step": 8849 }, { "epoch": 1.67, "learning_rate": 3.4443181149096616e-07, "loss": 0.199, "step": 8850 }, { "epoch": 1.67, "learning_rate": 3.4404447459275125e-07, "loss": 0.1816, "step": 8851 }, { "epoch": 1.67, "learning_rate": 3.4365733951621796e-07, "loss": 0.1777, "step": 8852 }, { "epoch": 1.67, "learning_rate": 3.4327040629760637e-07, "loss": 0.2318, "step": 8853 }, { "epoch": 1.67, "learning_rate": 3.428836749731368e-07, "loss": 0.3351, "step": 8854 }, { "epoch": 1.67, "learning_rate": 3.424971455790127e-07, "loss": 0.2643, "step": 8855 }, { "epoch": 1.67, "learning_rate": 3.421108181514179e-07, "loss": 0.2769, "step": 8856 }, { "epoch": 1.67, "learning_rate": 3.4172469272651524e-07, "loss": 0.1988, "step": 8857 }, { "epoch": 1.67, "learning_rate": 3.4133876934045213e-07, "loss": 0.1496, "step": 8858 }, { "epoch": 1.67, "learning_rate": 3.4095304802935465e-07, "loss": 0.3625, "step": 8859 }, { "epoch": 1.67, "learning_rate": 3.4056752882933056e-07, "loss": 0.2455, "step": 8860 }, { "epoch": 1.67, "learning_rate": 3.4018221177646885e-07, "loss": 0.2766, "step": 8861 }, { "epoch": 1.67, "learning_rate": 3.3979709690683984e-07, "loss": 0.1986, "step": 8862 }, { "epoch": 1.67, "learning_rate": 3.394121842564943e-07, "loss": 0.1965, "step": 8863 }, { "epoch": 1.67, "learning_rate": 3.390274738614646e-07, "loss": 0.2312, "step": 8864 }, { "epoch": 1.67, "learning_rate": 3.3864296575776364e-07, "loss": 0.1712, "step": 8865 }, { "epoch": 1.68, "learning_rate": 3.382586599813861e-07, "loss": 0.2094, "step": 8866 }, { "epoch": 1.68, "learning_rate": 3.3787455656830697e-07, "loss": 0.237, "step": 8867 }, { "epoch": 1.68, "learning_rate": 3.3749065555448215e-07, "loss": 0.2043, "step": 8868 }, { "epoch": 1.68, "learning_rate": 3.3710695697585115e-07, "loss": 0.229, "step": 8869 }, { "epoch": 1.68, "learning_rate": 3.3672346086833007e-07, "loss": 0.2511, "step": 8870 }, { "epoch": 1.68, "learning_rate": 3.363401672678193e-07, "loss": 0.2156, "step": 8871 }, { "epoch": 1.68, "learning_rate": 3.3595707621020047e-07, "loss": 0.2146, "step": 8872 }, { "epoch": 1.68, "learning_rate": 3.3557418773133373e-07, "loss": 0.287, "step": 8873 }, { "epoch": 1.68, "learning_rate": 3.351915018670618e-07, "loss": 0.2047, "step": 8874 }, { "epoch": 1.68, "learning_rate": 3.348090186532102e-07, "loss": 0.1802, "step": 8875 }, { "epoch": 1.68, "learning_rate": 3.3442673812558167e-07, "loss": 0.194, "step": 8876 }, { "epoch": 1.68, "learning_rate": 3.340446603199621e-07, "loss": 0.2173, "step": 8877 }, { "epoch": 1.68, "learning_rate": 3.3366278527212e-07, "loss": 0.206, "step": 8878 }, { "epoch": 1.68, "learning_rate": 3.332811130178007e-07, "loss": 0.1559, "step": 8879 }, { "epoch": 1.68, "learning_rate": 3.3289964359273507e-07, "loss": 0.3061, "step": 8880 }, { "epoch": 1.68, "learning_rate": 3.3251837703263264e-07, "loss": 0.2545, "step": 8881 }, { "epoch": 1.68, "learning_rate": 3.321373133731828e-07, "loss": 0.1695, "step": 8882 }, { "epoch": 1.68, "learning_rate": 3.3175645265005887e-07, "loss": 0.2682, "step": 8883 }, { "epoch": 1.68, "learning_rate": 3.313757948989138e-07, "loss": 0.1401, "step": 8884 }, { "epoch": 1.68, "learning_rate": 3.3099534015537975e-07, "loss": 0.3206, "step": 8885 }, { "epoch": 1.68, "learning_rate": 3.306150884550732e-07, "loss": 0.199, "step": 8886 }, { "epoch": 1.68, "learning_rate": 3.302350398335902e-07, "loss": 0.1955, "step": 8887 }, { "epoch": 1.68, "learning_rate": 3.2985519432650547e-07, "loss": 0.1769, "step": 8888 }, { "epoch": 1.68, "learning_rate": 3.294755519693788e-07, "loss": 0.1885, "step": 8889 }, { "epoch": 1.68, "learning_rate": 3.290961127977482e-07, "loss": 0.1769, "step": 8890 }, { "epoch": 1.68, "learning_rate": 3.2871687684713384e-07, "loss": 0.2532, "step": 8891 }, { "epoch": 1.68, "learning_rate": 3.2833784415303604e-07, "loss": 0.2201, "step": 8892 }, { "epoch": 1.68, "learning_rate": 3.2795901475093706e-07, "loss": 0.175, "step": 8893 }, { "epoch": 1.68, "learning_rate": 3.2758038867629906e-07, "loss": 0.2055, "step": 8894 }, { "epoch": 1.68, "learning_rate": 3.272019659645656e-07, "loss": 0.2486, "step": 8895 }, { "epoch": 1.68, "learning_rate": 3.26823746651162e-07, "loss": 0.2096, "step": 8896 }, { "epoch": 1.68, "learning_rate": 3.2644573077149326e-07, "loss": 0.1712, "step": 8897 }, { "epoch": 1.68, "learning_rate": 3.2606791836094607e-07, "loss": 0.2481, "step": 8898 }, { "epoch": 1.68, "learning_rate": 3.25690309454888e-07, "loss": 0.2337, "step": 8899 }, { "epoch": 1.68, "learning_rate": 3.2531290408866714e-07, "loss": 0.3245, "step": 8900 }, { "epoch": 1.68, "learning_rate": 3.249357022976135e-07, "loss": 0.2128, "step": 8901 }, { "epoch": 1.68, "learning_rate": 3.24558704117037e-07, "loss": 0.2581, "step": 8902 }, { "epoch": 1.68, "learning_rate": 3.2418190958222877e-07, "loss": 0.2738, "step": 8903 }, { "epoch": 1.68, "learning_rate": 3.238053187284615e-07, "loss": 0.196, "step": 8904 }, { "epoch": 1.68, "learning_rate": 3.2342893159098803e-07, "loss": 0.2404, "step": 8905 }, { "epoch": 1.68, "learning_rate": 3.2305274820504247e-07, "loss": 0.2319, "step": 8906 }, { "epoch": 1.68, "learning_rate": 3.2267676860583963e-07, "loss": 0.232, "step": 8907 }, { "epoch": 1.68, "learning_rate": 3.223009928285756e-07, "loss": 0.2872, "step": 8908 }, { "epoch": 1.68, "learning_rate": 3.2192542090842715e-07, "loss": 0.2617, "step": 8909 }, { "epoch": 1.68, "learning_rate": 3.215500528805518e-07, "loss": 0.2063, "step": 8910 }, { "epoch": 1.68, "learning_rate": 3.2117488878008945e-07, "loss": 0.2133, "step": 8911 }, { "epoch": 1.68, "learning_rate": 3.2079992864215814e-07, "loss": 0.2812, "step": 8912 }, { "epoch": 1.68, "learning_rate": 3.204251725018587e-07, "loss": 0.2267, "step": 8913 }, { "epoch": 1.68, "learning_rate": 3.2005062039427354e-07, "loss": 0.307, "step": 8914 }, { "epoch": 1.68, "learning_rate": 3.1967627235446385e-07, "loss": 0.2355, "step": 8915 }, { "epoch": 1.68, "learning_rate": 3.1930212841747273e-07, "loss": 0.1837, "step": 8916 }, { "epoch": 1.68, "learning_rate": 3.189281886183254e-07, "loss": 0.3217, "step": 8917 }, { "epoch": 1.68, "learning_rate": 3.1855445299202594e-07, "loss": 0.2427, "step": 8918 }, { "epoch": 1.69, "learning_rate": 3.181809215735596e-07, "loss": 0.2788, "step": 8919 }, { "epoch": 1.69, "learning_rate": 3.1780759439789507e-07, "loss": 0.1783, "step": 8920 }, { "epoch": 1.69, "learning_rate": 3.1743447149997774e-07, "loss": 0.1786, "step": 8921 }, { "epoch": 1.69, "learning_rate": 3.170615529147378e-07, "loss": 0.2065, "step": 8922 }, { "epoch": 1.69, "learning_rate": 3.166888386770842e-07, "loss": 0.2287, "step": 8923 }, { "epoch": 1.69, "learning_rate": 3.1631632882190623e-07, "loss": 0.2447, "step": 8924 }, { "epoch": 1.69, "learning_rate": 3.1594402338407633e-07, "loss": 0.2582, "step": 8925 }, { "epoch": 1.69, "learning_rate": 3.1557192239844604e-07, "loss": 0.2322, "step": 8926 }, { "epoch": 1.69, "learning_rate": 3.152000258998475e-07, "loss": 0.2268, "step": 8927 }, { "epoch": 1.69, "learning_rate": 3.148283339230951e-07, "loss": 0.2774, "step": 8928 }, { "epoch": 1.69, "learning_rate": 3.144568465029835e-07, "loss": 0.2111, "step": 8929 }, { "epoch": 1.69, "learning_rate": 3.140855636742876e-07, "loss": 0.2734, "step": 8930 }, { "epoch": 1.69, "learning_rate": 3.1371448547176414e-07, "loss": 0.2099, "step": 8931 }, { "epoch": 1.69, "learning_rate": 3.1334361193014974e-07, "loss": 0.2893, "step": 8932 }, { "epoch": 1.69, "learning_rate": 3.129729430841627e-07, "loss": 0.1732, "step": 8933 }, { "epoch": 1.69, "learning_rate": 3.1260247896850173e-07, "loss": 0.2534, "step": 8934 }, { "epoch": 1.69, "learning_rate": 3.122322196178462e-07, "loss": 0.2249, "step": 8935 }, { "epoch": 1.69, "learning_rate": 3.1186216506685685e-07, "loss": 0.1945, "step": 8936 }, { "epoch": 1.69, "learning_rate": 3.114923153501748e-07, "loss": 0.1581, "step": 8937 }, { "epoch": 1.69, "learning_rate": 3.1112267050242203e-07, "loss": 0.2434, "step": 8938 }, { "epoch": 1.69, "learning_rate": 3.107532305582017e-07, "loss": 0.2162, "step": 8939 }, { "epoch": 1.69, "learning_rate": 3.1038399555209734e-07, "loss": 0.2696, "step": 8940 }, { "epoch": 1.69, "learning_rate": 3.1001496551867323e-07, "loss": 0.1874, "step": 8941 }, { "epoch": 1.69, "learning_rate": 3.096461404924758e-07, "loss": 0.2332, "step": 8942 }, { "epoch": 1.69, "learning_rate": 3.0927752050803006e-07, "loss": 0.1969, "step": 8943 }, { "epoch": 1.69, "learning_rate": 3.0890910559984307e-07, "loss": 0.2636, "step": 8944 }, { "epoch": 1.69, "learning_rate": 3.085408958024039e-07, "loss": 0.1573, "step": 8945 }, { "epoch": 1.69, "learning_rate": 3.081728911501797e-07, "loss": 0.1932, "step": 8946 }, { "epoch": 1.69, "learning_rate": 3.0780509167761983e-07, "loss": 0.2436, "step": 8947 }, { "epoch": 1.69, "learning_rate": 3.0743749741915603e-07, "loss": 0.1601, "step": 8948 }, { "epoch": 1.69, "learning_rate": 3.070701084091976e-07, "loss": 0.1734, "step": 8949 }, { "epoch": 1.69, "learning_rate": 3.067029246821365e-07, "loss": 0.199, "step": 8950 }, { "epoch": 1.69, "learning_rate": 3.0633594627234644e-07, "loss": 0.1531, "step": 8951 }, { "epoch": 1.69, "learning_rate": 3.0596917321417895e-07, "loss": 0.2183, "step": 8952 }, { "epoch": 1.69, "learning_rate": 3.0560260554196975e-07, "loss": 0.247, "step": 8953 }, { "epoch": 1.69, "learning_rate": 3.0523624329003327e-07, "loss": 0.2228, "step": 8954 }, { "epoch": 1.69, "learning_rate": 3.04870086492664e-07, "loss": 0.2146, "step": 8955 }, { "epoch": 1.69, "learning_rate": 3.0450413518413964e-07, "loss": 0.2308, "step": 8956 }, { "epoch": 1.69, "learning_rate": 3.0413838939871763e-07, "loss": 0.1703, "step": 8957 }, { "epoch": 1.69, "learning_rate": 3.0377284917063403e-07, "loss": 0.1739, "step": 8958 }, { "epoch": 1.69, "learning_rate": 3.034075145341092e-07, "loss": 0.238, "step": 8959 }, { "epoch": 1.69, "learning_rate": 3.0304238552334276e-07, "loss": 0.1859, "step": 8960 }, { "epoch": 1.69, "learning_rate": 3.026774621725129e-07, "loss": 0.1964, "step": 8961 }, { "epoch": 1.69, "learning_rate": 3.0231274451578225e-07, "loss": 0.3082, "step": 8962 }, { "epoch": 1.69, "learning_rate": 3.0194823258729217e-07, "loss": 0.1646, "step": 8963 }, { "epoch": 1.69, "learning_rate": 3.015839264211651e-07, "loss": 0.2496, "step": 8964 }, { "epoch": 1.69, "learning_rate": 3.0121982605150394e-07, "loss": 0.2087, "step": 8965 }, { "epoch": 1.69, "learning_rate": 3.008559315123927e-07, "loss": 0.1431, "step": 8966 }, { "epoch": 1.69, "learning_rate": 3.004922428378959e-07, "loss": 0.187, "step": 8967 }, { "epoch": 1.69, "learning_rate": 3.001287600620592e-07, "loss": 0.2043, "step": 8968 }, { "epoch": 1.69, "learning_rate": 2.9976548321890843e-07, "loss": 0.1862, "step": 8969 }, { "epoch": 1.69, "learning_rate": 2.994024123424502e-07, "loss": 0.2057, "step": 8970 }, { "epoch": 1.69, "learning_rate": 2.990395474666724e-07, "loss": 0.2908, "step": 8971 }, { "epoch": 1.7, "learning_rate": 2.986768886255431e-07, "loss": 0.1762, "step": 8972 }, { "epoch": 1.7, "learning_rate": 2.983144358530113e-07, "loss": 0.1989, "step": 8973 }, { "epoch": 1.7, "learning_rate": 2.9795218918300644e-07, "loss": 0.316, "step": 8974 }, { "epoch": 1.7, "learning_rate": 2.9759014864943923e-07, "loss": 0.2409, "step": 8975 }, { "epoch": 1.7, "learning_rate": 2.972283142862006e-07, "loss": 0.1549, "step": 8976 }, { "epoch": 1.7, "learning_rate": 2.968666861271621e-07, "loss": 0.2509, "step": 8977 }, { "epoch": 1.7, "learning_rate": 2.9650526420617637e-07, "loss": 0.2247, "step": 8978 }, { "epoch": 1.7, "learning_rate": 2.9614404855707655e-07, "loss": 0.241, "step": 8979 }, { "epoch": 1.7, "learning_rate": 2.957830392136768e-07, "loss": 0.2056, "step": 8980 }, { "epoch": 1.7, "learning_rate": 2.9542223620977114e-07, "loss": 0.2075, "step": 8981 }, { "epoch": 1.7, "learning_rate": 2.9506163957913476e-07, "loss": 0.242, "step": 8982 }, { "epoch": 1.7, "learning_rate": 2.947012493555235e-07, "loss": 0.194, "step": 8983 }, { "epoch": 1.7, "learning_rate": 2.943410655726753e-07, "loss": 0.2868, "step": 8984 }, { "epoch": 1.7, "learning_rate": 2.93981088264306e-07, "loss": 0.2124, "step": 8985 }, { "epoch": 1.7, "learning_rate": 2.936213174641131e-07, "loss": 0.2989, "step": 8986 }, { "epoch": 1.7, "learning_rate": 2.932617532057769e-07, "loss": 0.2124, "step": 8987 }, { "epoch": 1.7, "learning_rate": 2.9290239552295543e-07, "loss": 0.2418, "step": 8988 }, { "epoch": 1.7, "learning_rate": 2.9254324444928847e-07, "loss": 0.3519, "step": 8989 }, { "epoch": 1.7, "learning_rate": 2.9218430001839744e-07, "loss": 0.1536, "step": 8990 }, { "epoch": 1.7, "learning_rate": 2.91825562263883e-07, "loss": 0.321, "step": 8991 }, { "epoch": 1.7, "learning_rate": 2.914670312193274e-07, "loss": 0.2164, "step": 8992 }, { "epoch": 1.7, "learning_rate": 2.91108706918293e-07, "loss": 0.2688, "step": 8993 }, { "epoch": 1.7, "learning_rate": 2.907505893943227e-07, "loss": 0.2324, "step": 8994 }, { "epoch": 1.7, "learning_rate": 2.9039267868094077e-07, "loss": 0.2775, "step": 8995 }, { "epoch": 1.7, "learning_rate": 2.9003497481165154e-07, "loss": 0.1988, "step": 8996 }, { "epoch": 1.7, "learning_rate": 2.8967747781994017e-07, "loss": 0.2592, "step": 8997 }, { "epoch": 1.7, "learning_rate": 2.893201877392721e-07, "loss": 0.2146, "step": 8998 }, { "epoch": 1.7, "learning_rate": 2.8896310460309413e-07, "loss": 0.1714, "step": 8999 }, { "epoch": 1.7, "learning_rate": 2.886062284448329e-07, "loss": 0.2007, "step": 9000 }, { "epoch": 1.7, "learning_rate": 2.8824955929789624e-07, "loss": 0.2132, "step": 9001 }, { "epoch": 1.7, "learning_rate": 2.87893097195672e-07, "loss": 0.2057, "step": 9002 }, { "epoch": 1.7, "learning_rate": 2.8753684217152965e-07, "loss": 0.2005, "step": 9003 }, { "epoch": 1.7, "learning_rate": 2.871807942588184e-07, "loss": 0.2563, "step": 9004 }, { "epoch": 1.7, "learning_rate": 2.868249534908682e-07, "loss": 0.2125, "step": 9005 }, { "epoch": 1.7, "learning_rate": 2.864693199009899e-07, "loss": 0.2258, "step": 9006 }, { "epoch": 1.7, "learning_rate": 2.861138935224747e-07, "loss": 0.2455, "step": 9007 }, { "epoch": 1.7, "learning_rate": 2.8575867438859483e-07, "loss": 0.2436, "step": 9008 }, { "epoch": 1.7, "learning_rate": 2.854036625326023e-07, "loss": 0.2718, "step": 9009 }, { "epoch": 1.7, "learning_rate": 2.850488579877303e-07, "loss": 0.3046, "step": 9010 }, { "epoch": 1.7, "learning_rate": 2.846942607871925e-07, "loss": 0.2504, "step": 9011 }, { "epoch": 1.7, "learning_rate": 2.8433987096418424e-07, "loss": 0.2475, "step": 9012 }, { "epoch": 1.7, "learning_rate": 2.839856885518788e-07, "loss": 0.2881, "step": 9013 }, { "epoch": 1.7, "learning_rate": 2.836317135834321e-07, "loss": 0.1421, "step": 9014 }, { "epoch": 1.7, "learning_rate": 2.8327794609198126e-07, "loss": 0.225, "step": 9015 }, { "epoch": 1.7, "learning_rate": 2.8292438611064174e-07, "loss": 0.2489, "step": 9016 }, { "epoch": 1.7, "learning_rate": 2.8257103367251045e-07, "loss": 0.21, "step": 9017 }, { "epoch": 1.7, "learning_rate": 2.8221788881066666e-07, "loss": 0.1959, "step": 9018 }, { "epoch": 1.7, "learning_rate": 2.818649515581676e-07, "loss": 0.1856, "step": 9019 }, { "epoch": 1.7, "learning_rate": 2.815122219480515e-07, "loss": 0.1678, "step": 9020 }, { "epoch": 1.7, "learning_rate": 2.811597000133398e-07, "loss": 0.2644, "step": 9021 }, { "epoch": 1.7, "learning_rate": 2.8080738578703054e-07, "loss": 0.1893, "step": 9022 }, { "epoch": 1.7, "learning_rate": 2.8045527930210535e-07, "loss": 0.1812, "step": 9023 }, { "epoch": 1.7, "learning_rate": 2.80103380591526e-07, "loss": 0.1776, "step": 9024 }, { "epoch": 1.71, "learning_rate": 2.7975168968823215e-07, "loss": 0.2037, "step": 9025 }, { "epoch": 1.71, "learning_rate": 2.794002066251475e-07, "loss": 0.1907, "step": 9026 }, { "epoch": 1.71, "learning_rate": 2.790489314351752e-07, "loss": 0.1845, "step": 9027 }, { "epoch": 1.71, "learning_rate": 2.78697864151197e-07, "loss": 0.2128, "step": 9028 }, { "epoch": 1.71, "learning_rate": 2.7834700480607794e-07, "loss": 0.1663, "step": 9029 }, { "epoch": 1.71, "learning_rate": 2.7799635343266266e-07, "loss": 0.2089, "step": 9030 }, { "epoch": 1.71, "learning_rate": 2.776459100637743e-07, "loss": 0.1981, "step": 9031 }, { "epoch": 1.71, "learning_rate": 2.772956747322203e-07, "loss": 0.1925, "step": 9032 }, { "epoch": 1.71, "learning_rate": 2.7694564747078545e-07, "loss": 0.1613, "step": 9033 }, { "epoch": 1.71, "learning_rate": 2.765958283122369e-07, "loss": 0.3534, "step": 9034 }, { "epoch": 1.71, "learning_rate": 2.7624621728932125e-07, "loss": 0.2051, "step": 9035 }, { "epoch": 1.71, "learning_rate": 2.75896814434766e-07, "loss": 0.2342, "step": 9036 }, { "epoch": 1.71, "learning_rate": 2.755476197812795e-07, "loss": 0.2333, "step": 9037 }, { "epoch": 1.71, "learning_rate": 2.7519863336155e-07, "loss": 0.2428, "step": 9038 }, { "epoch": 1.71, "learning_rate": 2.7484985520824655e-07, "loss": 0.2761, "step": 9039 }, { "epoch": 1.71, "learning_rate": 2.745012853540191e-07, "loss": 0.2147, "step": 9040 }, { "epoch": 1.71, "learning_rate": 2.7415292383149716e-07, "loss": 0.1738, "step": 9041 }, { "epoch": 1.71, "learning_rate": 2.7380477067329185e-07, "loss": 0.2911, "step": 9042 }, { "epoch": 1.71, "learning_rate": 2.734568259119938e-07, "loss": 0.2035, "step": 9043 }, { "epoch": 1.71, "learning_rate": 2.7310908958017465e-07, "loss": 0.2675, "step": 9044 }, { "epoch": 1.71, "learning_rate": 2.7276156171038627e-07, "loss": 0.2106, "step": 9045 }, { "epoch": 1.71, "learning_rate": 2.724142423351617e-07, "loss": 0.2758, "step": 9046 }, { "epoch": 1.71, "learning_rate": 2.7206713148701324e-07, "loss": 0.1924, "step": 9047 }, { "epoch": 1.71, "learning_rate": 2.71720229198435e-07, "loss": 0.1693, "step": 9048 }, { "epoch": 1.71, "learning_rate": 2.713735355019007e-07, "loss": 0.289, "step": 9049 }, { "epoch": 1.71, "learning_rate": 2.7102705042986457e-07, "loss": 0.2152, "step": 9050 }, { "epoch": 1.71, "learning_rate": 2.706807740147618e-07, "loss": 0.2521, "step": 9051 }, { "epoch": 1.71, "learning_rate": 2.7033470628900773e-07, "loss": 0.1668, "step": 9052 }, { "epoch": 1.71, "learning_rate": 2.6998884728499736e-07, "loss": 0.1937, "step": 9053 }, { "epoch": 1.71, "learning_rate": 2.696431970351082e-07, "loss": 0.2211, "step": 9054 }, { "epoch": 1.71, "learning_rate": 2.692977555716972e-07, "loss": 0.229, "step": 9055 }, { "epoch": 1.71, "learning_rate": 2.6895252292709977e-07, "loss": 0.2577, "step": 9056 }, { "epoch": 1.71, "learning_rate": 2.6860749913363544e-07, "loss": 0.2041, "step": 9057 }, { "epoch": 1.71, "learning_rate": 2.6826268422360165e-07, "loss": 0.2516, "step": 9058 }, { "epoch": 1.71, "learning_rate": 2.679180782292762e-07, "loss": 0.2747, "step": 9059 }, { "epoch": 1.71, "learning_rate": 2.675736811829191e-07, "loss": 0.2278, "step": 9060 }, { "epoch": 1.71, "learning_rate": 2.672294931167699e-07, "loss": 0.2327, "step": 9061 }, { "epoch": 1.71, "learning_rate": 2.668855140630469e-07, "loss": 0.1386, "step": 9062 }, { "epoch": 1.71, "learning_rate": 2.665417440539522e-07, "loss": 0.1883, "step": 9063 }, { "epoch": 1.71, "learning_rate": 2.661981831216656e-07, "loss": 0.2535, "step": 9064 }, { "epoch": 1.71, "learning_rate": 2.658548312983489e-07, "loss": 0.201, "step": 9065 }, { "epoch": 1.71, "learning_rate": 2.6551168861614297e-07, "loss": 0.1883, "step": 9066 }, { "epoch": 1.71, "learning_rate": 2.651687551071705e-07, "loss": 0.1627, "step": 9067 }, { "epoch": 1.71, "learning_rate": 2.6482603080353333e-07, "loss": 0.269, "step": 9068 }, { "epoch": 1.71, "learning_rate": 2.6448351573731464e-07, "loss": 0.211, "step": 9069 }, { "epoch": 1.71, "learning_rate": 2.641412099405777e-07, "loss": 0.1698, "step": 9070 }, { "epoch": 1.71, "learning_rate": 2.6379911344536604e-07, "loss": 0.1258, "step": 9071 }, { "epoch": 1.71, "learning_rate": 2.6345722628370373e-07, "loss": 0.1663, "step": 9072 }, { "epoch": 1.71, "learning_rate": 2.6311554848759524e-07, "loss": 0.2712, "step": 9073 }, { "epoch": 1.71, "learning_rate": 2.6277408008902575e-07, "loss": 0.2508, "step": 9074 }, { "epoch": 1.71, "learning_rate": 2.6243282111996005e-07, "loss": 0.2575, "step": 9075 }, { "epoch": 1.71, "learning_rate": 2.620917716123444e-07, "loss": 0.2303, "step": 9076 }, { "epoch": 1.71, "learning_rate": 2.6175093159810427e-07, "loss": 0.1984, "step": 9077 }, { "epoch": 1.72, "learning_rate": 2.614103011091468e-07, "loss": 0.2507, "step": 9078 }, { "epoch": 1.72, "learning_rate": 2.6106988017735815e-07, "loss": 0.2203, "step": 9079 }, { "epoch": 1.72, "learning_rate": 2.6072966883460586e-07, "loss": 0.2183, "step": 9080 }, { "epoch": 1.72, "learning_rate": 2.603896671127376e-07, "loss": 0.235, "step": 9081 }, { "epoch": 1.72, "learning_rate": 2.600498750435812e-07, "loss": 0.2141, "step": 9082 }, { "epoch": 1.72, "learning_rate": 2.597102926589451e-07, "loss": 0.18, "step": 9083 }, { "epoch": 1.72, "learning_rate": 2.5937091999061776e-07, "loss": 0.2337, "step": 9084 }, { "epoch": 1.72, "learning_rate": 2.590317570703693e-07, "loss": 0.2111, "step": 9085 }, { "epoch": 1.72, "learning_rate": 2.58692803929948e-07, "loss": 0.1946, "step": 9086 }, { "epoch": 1.72, "learning_rate": 2.583540606010834e-07, "loss": 0.2682, "step": 9087 }, { "epoch": 1.72, "learning_rate": 2.580155271154877e-07, "loss": 0.2599, "step": 9088 }, { "epoch": 1.72, "learning_rate": 2.576772035048494e-07, "loss": 0.1622, "step": 9089 }, { "epoch": 1.72, "learning_rate": 2.573390898008399e-07, "loss": 0.2336, "step": 9090 }, { "epoch": 1.72, "learning_rate": 2.5700118603511126e-07, "loss": 0.1771, "step": 9091 }, { "epoch": 1.72, "learning_rate": 2.5666349223929436e-07, "loss": 0.1782, "step": 9092 }, { "epoch": 1.72, "learning_rate": 2.563260084450006e-07, "loss": 0.2536, "step": 9093 }, { "epoch": 1.72, "learning_rate": 2.5598873468382385e-07, "loss": 0.2322, "step": 9094 }, { "epoch": 1.72, "learning_rate": 2.5565167098733497e-07, "loss": 0.2986, "step": 9095 }, { "epoch": 1.72, "learning_rate": 2.553148173870881e-07, "loss": 0.1658, "step": 9096 }, { "epoch": 1.72, "learning_rate": 2.5497817391461677e-07, "loss": 0.1992, "step": 9097 }, { "epoch": 1.72, "learning_rate": 2.5464174060143317e-07, "loss": 0.2804, "step": 9098 }, { "epoch": 1.72, "learning_rate": 2.543055174790324e-07, "loss": 0.2284, "step": 9099 }, { "epoch": 1.72, "learning_rate": 2.5396950457888867e-07, "loss": 0.2006, "step": 9100 }, { "epoch": 1.72, "learning_rate": 2.536337019324556e-07, "loss": 0.2878, "step": 9101 }, { "epoch": 1.72, "learning_rate": 2.53298109571169e-07, "loss": 0.3012, "step": 9102 }, { "epoch": 1.72, "learning_rate": 2.529627275264446e-07, "loss": 0.2967, "step": 9103 }, { "epoch": 1.72, "learning_rate": 2.526275558296759e-07, "loss": 0.203, "step": 9104 }, { "epoch": 1.72, "learning_rate": 2.5229259451224093e-07, "loss": 0.2668, "step": 9105 }, { "epoch": 1.72, "learning_rate": 2.519578436054948e-07, "loss": 0.1751, "step": 9106 }, { "epoch": 1.72, "learning_rate": 2.516233031407739e-07, "loss": 0.2508, "step": 9107 }, { "epoch": 1.72, "learning_rate": 2.5128897314939533e-07, "loss": 0.2095, "step": 9108 }, { "epoch": 1.72, "learning_rate": 2.5095485366265604e-07, "loss": 0.2004, "step": 9109 }, { "epoch": 1.72, "learning_rate": 2.5062094471183324e-07, "loss": 0.2035, "step": 9110 }, { "epoch": 1.72, "learning_rate": 2.502872463281847e-07, "loss": 0.2665, "step": 9111 }, { "epoch": 1.72, "learning_rate": 2.499537585429482e-07, "loss": 0.2298, "step": 9112 }, { "epoch": 1.72, "learning_rate": 2.4962048138734206e-07, "loss": 0.2259, "step": 9113 }, { "epoch": 1.72, "learning_rate": 2.4928741489256475e-07, "loss": 0.2347, "step": 9114 }, { "epoch": 1.72, "learning_rate": 2.489545590897946e-07, "loss": 0.2799, "step": 9115 }, { "epoch": 1.72, "learning_rate": 2.48621914010192e-07, "loss": 0.2319, "step": 9116 }, { "epoch": 1.72, "learning_rate": 2.4828947968489477e-07, "loss": 0.2631, "step": 9117 }, { "epoch": 1.72, "learning_rate": 2.4795725614502274e-07, "loss": 0.2046, "step": 9118 }, { "epoch": 1.72, "learning_rate": 2.476252434216772e-07, "loss": 0.2379, "step": 9119 }, { "epoch": 1.72, "learning_rate": 2.472934415459366e-07, "loss": 0.1812, "step": 9120 }, { "epoch": 1.72, "learning_rate": 2.4696185054886133e-07, "loss": 0.2364, "step": 9121 }, { "epoch": 1.72, "learning_rate": 2.4663047046149356e-07, "loss": 0.2475, "step": 9122 }, { "epoch": 1.72, "learning_rate": 2.4629930131485293e-07, "loss": 0.2068, "step": 9123 }, { "epoch": 1.72, "learning_rate": 2.459683431399404e-07, "loss": 0.1834, "step": 9124 }, { "epoch": 1.72, "learning_rate": 2.456375959677387e-07, "loss": 0.211, "step": 9125 }, { "epoch": 1.72, "learning_rate": 2.4530705982920787e-07, "loss": 0.2783, "step": 9126 }, { "epoch": 1.72, "learning_rate": 2.4497673475529115e-07, "loss": 0.2428, "step": 9127 }, { "epoch": 1.72, "learning_rate": 2.4464662077691046e-07, "loss": 0.1803, "step": 9128 }, { "epoch": 1.72, "learning_rate": 2.4431671792496724e-07, "loss": 0.1951, "step": 9129 }, { "epoch": 1.72, "learning_rate": 2.439870262303448e-07, "loss": 0.2962, "step": 9130 }, { "epoch": 1.73, "learning_rate": 2.436575457239068e-07, "loss": 0.2136, "step": 9131 }, { "epoch": 1.73, "learning_rate": 2.4332827643649417e-07, "loss": 0.2376, "step": 9132 }, { "epoch": 1.73, "learning_rate": 2.4299921839893215e-07, "loss": 0.1994, "step": 9133 }, { "epoch": 1.73, "learning_rate": 2.4267037164202393e-07, "loss": 0.1704, "step": 9134 }, { "epoch": 1.73, "learning_rate": 2.423417361965522e-07, "loss": 0.2168, "step": 9135 }, { "epoch": 1.73, "learning_rate": 2.420133120932822e-07, "loss": 0.2475, "step": 9136 }, { "epoch": 1.73, "learning_rate": 2.4168509936295743e-07, "loss": 0.2906, "step": 9137 }, { "epoch": 1.73, "learning_rate": 2.413570980363025e-07, "loss": 0.1666, "step": 9138 }, { "epoch": 1.73, "learning_rate": 2.410293081440221e-07, "loss": 0.2293, "step": 9139 }, { "epoch": 1.73, "learning_rate": 2.40701729716801e-07, "loss": 0.2043, "step": 9140 }, { "epoch": 1.73, "learning_rate": 2.403743627853039e-07, "loss": 0.2451, "step": 9141 }, { "epoch": 1.73, "learning_rate": 2.4004720738017657e-07, "loss": 0.2208, "step": 9142 }, { "epoch": 1.73, "learning_rate": 2.397202635320439e-07, "loss": 0.2239, "step": 9143 }, { "epoch": 1.73, "learning_rate": 2.3939353127151204e-07, "loss": 0.25, "step": 9144 }, { "epoch": 1.73, "learning_rate": 2.390670106291665e-07, "loss": 0.1672, "step": 9145 }, { "epoch": 1.73, "learning_rate": 2.3874070163557306e-07, "loss": 0.1975, "step": 9146 }, { "epoch": 1.73, "learning_rate": 2.3841460432127843e-07, "loss": 0.225, "step": 9147 }, { "epoch": 1.73, "learning_rate": 2.380887187168085e-07, "loss": 0.1834, "step": 9148 }, { "epoch": 1.73, "learning_rate": 2.377630448526702e-07, "loss": 0.186, "step": 9149 }, { "epoch": 1.73, "learning_rate": 2.374375827593503e-07, "loss": 0.2684, "step": 9150 }, { "epoch": 1.73, "learning_rate": 2.3711233246731525e-07, "loss": 0.2231, "step": 9151 }, { "epoch": 1.73, "learning_rate": 2.3678729400701234e-07, "loss": 0.2285, "step": 9152 }, { "epoch": 1.73, "learning_rate": 2.364624674088692e-07, "loss": 0.2635, "step": 9153 }, { "epoch": 1.73, "learning_rate": 2.3613785270329232e-07, "loss": 0.1821, "step": 9154 }, { "epoch": 1.73, "learning_rate": 2.3581344992067075e-07, "loss": 0.242, "step": 9155 }, { "epoch": 1.73, "learning_rate": 2.3548925909137128e-07, "loss": 0.2331, "step": 9156 }, { "epoch": 1.73, "learning_rate": 2.3516528024574102e-07, "loss": 0.2192, "step": 9157 }, { "epoch": 1.73, "learning_rate": 2.348415134141102e-07, "loss": 0.2657, "step": 9158 }, { "epoch": 1.73, "learning_rate": 2.345179586267854e-07, "loss": 0.2352, "step": 9159 }, { "epoch": 1.73, "learning_rate": 2.3419461591405458e-07, "loss": 0.2296, "step": 9160 }, { "epoch": 1.73, "learning_rate": 2.3387148530618825e-07, "loss": 0.2512, "step": 9161 }, { "epoch": 1.73, "learning_rate": 2.3354856683343337e-07, "loss": 0.223, "step": 9162 }, { "epoch": 1.73, "learning_rate": 2.3322586052601876e-07, "loss": 0.2458, "step": 9163 }, { "epoch": 1.73, "learning_rate": 2.3290336641415473e-07, "loss": 0.1821, "step": 9164 }, { "epoch": 1.73, "learning_rate": 2.3258108452802853e-07, "loss": 0.2404, "step": 9165 }, { "epoch": 1.73, "learning_rate": 2.3225901489781073e-07, "loss": 0.1935, "step": 9166 }, { "epoch": 1.73, "learning_rate": 2.319371575536508e-07, "loss": 0.2402, "step": 9167 }, { "epoch": 1.73, "learning_rate": 2.3161551252567665e-07, "loss": 0.1876, "step": 9168 }, { "epoch": 1.73, "learning_rate": 2.3129407984399943e-07, "loss": 0.2624, "step": 9169 }, { "epoch": 1.73, "learning_rate": 2.3097285953870867e-07, "loss": 0.2423, "step": 9170 }, { "epoch": 1.73, "learning_rate": 2.3065185163987286e-07, "loss": 0.1946, "step": 9171 }, { "epoch": 1.73, "learning_rate": 2.3033105617754347e-07, "loss": 0.1664, "step": 9172 }, { "epoch": 1.73, "learning_rate": 2.300104731817507e-07, "loss": 0.2012, "step": 9173 }, { "epoch": 1.73, "learning_rate": 2.296901026825027e-07, "loss": 0.2223, "step": 9174 }, { "epoch": 1.73, "learning_rate": 2.293699447097919e-07, "loss": 0.26, "step": 9175 }, { "epoch": 1.73, "learning_rate": 2.2904999929358768e-07, "loss": 0.2386, "step": 9176 }, { "epoch": 1.73, "learning_rate": 2.2873026646384077e-07, "loss": 0.1882, "step": 9177 }, { "epoch": 1.73, "learning_rate": 2.284107462504817e-07, "loss": 0.225, "step": 9178 }, { "epoch": 1.73, "learning_rate": 2.28091438683421e-07, "loss": 0.2413, "step": 9179 }, { "epoch": 1.73, "learning_rate": 2.277723437925497e-07, "loss": 0.2481, "step": 9180 }, { "epoch": 1.73, "learning_rate": 2.2745346160773867e-07, "loss": 0.229, "step": 9181 }, { "epoch": 1.73, "learning_rate": 2.2713479215883876e-07, "loss": 0.2834, "step": 9182 }, { "epoch": 1.73, "learning_rate": 2.268163354756811e-07, "loss": 0.17, "step": 9183 }, { "epoch": 1.74, "learning_rate": 2.2649809158807685e-07, "loss": 0.3013, "step": 9184 }, { "epoch": 1.74, "learning_rate": 2.261800605258163e-07, "loss": 0.2809, "step": 9185 }, { "epoch": 1.74, "learning_rate": 2.2586224231867266e-07, "loss": 0.185, "step": 9186 }, { "epoch": 1.74, "learning_rate": 2.2554463699639596e-07, "loss": 0.2181, "step": 9187 }, { "epoch": 1.74, "learning_rate": 2.2522724458871714e-07, "loss": 0.202, "step": 9188 }, { "epoch": 1.74, "learning_rate": 2.2491006512534914e-07, "loss": 0.1555, "step": 9189 }, { "epoch": 1.74, "learning_rate": 2.245930986359826e-07, "loss": 0.2312, "step": 9190 }, { "epoch": 1.74, "learning_rate": 2.2427634515028884e-07, "loss": 0.1953, "step": 9191 }, { "epoch": 1.74, "learning_rate": 2.2395980469792106e-07, "loss": 0.1732, "step": 9192 }, { "epoch": 1.74, "learning_rate": 2.2364347730850943e-07, "loss": 0.2744, "step": 9193 }, { "epoch": 1.74, "learning_rate": 2.2332736301166586e-07, "loss": 0.3111, "step": 9194 }, { "epoch": 1.74, "learning_rate": 2.2301146183698358e-07, "loss": 0.2031, "step": 9195 }, { "epoch": 1.74, "learning_rate": 2.2269577381403256e-07, "loss": 0.1681, "step": 9196 }, { "epoch": 1.74, "learning_rate": 2.223802989723664e-07, "loss": 0.2379, "step": 9197 }, { "epoch": 1.74, "learning_rate": 2.2206503734151675e-07, "loss": 0.227, "step": 9198 }, { "epoch": 1.74, "learning_rate": 2.2174998895099443e-07, "loss": 0.2355, "step": 9199 }, { "epoch": 1.74, "learning_rate": 2.2143515383029306e-07, "loss": 0.2259, "step": 9200 }, { "epoch": 1.74, "learning_rate": 2.211205320088844e-07, "loss": 0.2248, "step": 9201 }, { "epoch": 1.74, "learning_rate": 2.2080612351621954e-07, "loss": 0.2213, "step": 9202 }, { "epoch": 1.74, "learning_rate": 2.2049192838173196e-07, "loss": 0.1996, "step": 9203 }, { "epoch": 1.74, "learning_rate": 2.201779466348336e-07, "loss": 0.145, "step": 9204 }, { "epoch": 1.74, "learning_rate": 2.1986417830491548e-07, "loss": 0.2284, "step": 9205 }, { "epoch": 1.74, "learning_rate": 2.1955062342135126e-07, "loss": 0.2143, "step": 9206 }, { "epoch": 1.74, "learning_rate": 2.1923728201349248e-07, "loss": 0.216, "step": 9207 }, { "epoch": 1.74, "learning_rate": 2.189241541106721e-07, "loss": 0.2424, "step": 9208 }, { "epoch": 1.74, "learning_rate": 2.186112397422016e-07, "loss": 0.2168, "step": 9209 }, { "epoch": 1.74, "learning_rate": 2.1829853893737373e-07, "loss": 0.2468, "step": 9210 }, { "epoch": 1.74, "learning_rate": 2.1798605172546061e-07, "loss": 0.2996, "step": 9211 }, { "epoch": 1.74, "learning_rate": 2.1767377813571494e-07, "loss": 0.3094, "step": 9212 }, { "epoch": 1.74, "learning_rate": 2.1736171819736862e-07, "loss": 0.2266, "step": 9213 }, { "epoch": 1.74, "learning_rate": 2.1704987193963412e-07, "loss": 0.3, "step": 9214 }, { "epoch": 1.74, "learning_rate": 2.1673823939170362e-07, "loss": 0.1811, "step": 9215 }, { "epoch": 1.74, "learning_rate": 2.1642682058274966e-07, "loss": 0.1771, "step": 9216 }, { "epoch": 1.74, "learning_rate": 2.1611561554192446e-07, "loss": 0.3199, "step": 9217 }, { "epoch": 1.74, "learning_rate": 2.1580462429836024e-07, "loss": 0.2162, "step": 9218 }, { "epoch": 1.74, "learning_rate": 2.154938468811693e-07, "loss": 0.1533, "step": 9219 }, { "epoch": 1.74, "learning_rate": 2.1518328331944394e-07, "loss": 0.1941, "step": 9220 }, { "epoch": 1.74, "learning_rate": 2.1487293364225642e-07, "loss": 0.2846, "step": 9221 }, { "epoch": 1.74, "learning_rate": 2.1456279787865908e-07, "loss": 0.196, "step": 9222 }, { "epoch": 1.74, "learning_rate": 2.142528760576837e-07, "loss": 0.2372, "step": 9223 }, { "epoch": 1.74, "learning_rate": 2.1394316820834287e-07, "loss": 0.2357, "step": 9224 }, { "epoch": 1.74, "learning_rate": 2.1363367435962846e-07, "loss": 0.2862, "step": 9225 }, { "epoch": 1.74, "learning_rate": 2.1332439454051278e-07, "loss": 0.2289, "step": 9226 }, { "epoch": 1.74, "learning_rate": 2.1301532877994747e-07, "loss": 0.2239, "step": 9227 }, { "epoch": 1.74, "learning_rate": 2.127064771068657e-07, "loss": 0.1802, "step": 9228 }, { "epoch": 1.74, "learning_rate": 2.1239783955017801e-07, "loss": 0.21, "step": 9229 }, { "epoch": 1.74, "learning_rate": 2.120894161387768e-07, "loss": 0.277, "step": 9230 }, { "epoch": 1.74, "learning_rate": 2.1178120690153487e-07, "loss": 0.1653, "step": 9231 }, { "epoch": 1.74, "learning_rate": 2.1147321186730295e-07, "loss": 0.2055, "step": 9232 }, { "epoch": 1.74, "learning_rate": 2.1116543106491277e-07, "loss": 0.1822, "step": 9233 }, { "epoch": 1.74, "learning_rate": 2.1085786452317763e-07, "loss": 0.2445, "step": 9234 }, { "epoch": 1.74, "learning_rate": 2.1055051227088756e-07, "loss": 0.2962, "step": 9235 }, { "epoch": 1.74, "learning_rate": 2.102433743368143e-07, "loss": 0.2291, "step": 9236 }, { "epoch": 1.75, "learning_rate": 2.0993645074971093e-07, "loss": 0.2071, "step": 9237 }, { "epoch": 1.75, "learning_rate": 2.0962974153830695e-07, "loss": 0.2525, "step": 9238 }, { "epoch": 1.75, "learning_rate": 2.0932324673131498e-07, "loss": 0.1777, "step": 9239 }, { "epoch": 1.75, "learning_rate": 2.0901696635742623e-07, "loss": 0.1994, "step": 9240 }, { "epoch": 1.75, "learning_rate": 2.0871090044531196e-07, "loss": 0.2306, "step": 9241 }, { "epoch": 1.75, "learning_rate": 2.0840504902362336e-07, "loss": 0.1901, "step": 9242 }, { "epoch": 1.75, "learning_rate": 2.0809941212099143e-07, "loss": 0.2386, "step": 9243 }, { "epoch": 1.75, "learning_rate": 2.0779398976602748e-07, "loss": 0.1631, "step": 9244 }, { "epoch": 1.75, "learning_rate": 2.0748878198732226e-07, "loss": 0.288, "step": 9245 }, { "epoch": 1.75, "learning_rate": 2.0718378881344649e-07, "loss": 0.2886, "step": 9246 }, { "epoch": 1.75, "learning_rate": 2.068790102729515e-07, "loss": 0.2537, "step": 9247 }, { "epoch": 1.75, "learning_rate": 2.0657444639436725e-07, "loss": 0.2295, "step": 9248 }, { "epoch": 1.75, "learning_rate": 2.062700972062051e-07, "loss": 0.2488, "step": 9249 }, { "epoch": 1.75, "learning_rate": 2.0596596273695506e-07, "loss": 0.1759, "step": 9250 }, { "epoch": 1.75, "learning_rate": 2.056620430150877e-07, "loss": 0.2025, "step": 9251 }, { "epoch": 1.75, "learning_rate": 2.0535833806905326e-07, "loss": 0.191, "step": 9252 }, { "epoch": 1.75, "learning_rate": 2.0505484792728207e-07, "loss": 0.2481, "step": 9253 }, { "epoch": 1.75, "learning_rate": 2.0475157261818425e-07, "loss": 0.1959, "step": 9254 }, { "epoch": 1.75, "learning_rate": 2.044485121701495e-07, "loss": 0.2266, "step": 9255 }, { "epoch": 1.75, "learning_rate": 2.0414566661154795e-07, "loss": 0.2666, "step": 9256 }, { "epoch": 1.75, "learning_rate": 2.0384303597072914e-07, "loss": 0.2173, "step": 9257 }, { "epoch": 1.75, "learning_rate": 2.0354062027602234e-07, "loss": 0.2059, "step": 9258 }, { "epoch": 1.75, "learning_rate": 2.0323841955573852e-07, "loss": 0.22, "step": 9259 }, { "epoch": 1.75, "learning_rate": 2.0293643383816564e-07, "loss": 0.2148, "step": 9260 }, { "epoch": 1.75, "learning_rate": 2.0263466315157298e-07, "loss": 0.2331, "step": 9261 }, { "epoch": 1.75, "learning_rate": 2.0233310752421104e-07, "loss": 0.1819, "step": 9262 }, { "epoch": 1.75, "learning_rate": 2.020317669843072e-07, "loss": 0.2809, "step": 9263 }, { "epoch": 1.75, "learning_rate": 2.0173064156007088e-07, "loss": 0.2806, "step": 9264 }, { "epoch": 1.75, "learning_rate": 2.0142973127969145e-07, "loss": 0.176, "step": 9265 }, { "epoch": 1.75, "learning_rate": 2.0112903617133666e-07, "loss": 0.2567, "step": 9266 }, { "epoch": 1.75, "learning_rate": 2.0082855626315483e-07, "loss": 0.1906, "step": 9267 }, { "epoch": 1.75, "learning_rate": 2.0052829158327568e-07, "loss": 0.196, "step": 9268 }, { "epoch": 1.75, "learning_rate": 2.0022824215980535e-07, "loss": 0.2161, "step": 9269 }, { "epoch": 1.75, "learning_rate": 1.999284080208333e-07, "loss": 0.2141, "step": 9270 }, { "epoch": 1.75, "learning_rate": 1.9962878919442736e-07, "loss": 0.2297, "step": 9271 }, { "epoch": 1.75, "learning_rate": 1.9932938570863398e-07, "loss": 0.2167, "step": 9272 }, { "epoch": 1.75, "learning_rate": 1.9903019759148184e-07, "loss": 0.1796, "step": 9273 }, { "epoch": 1.75, "learning_rate": 1.9873122487097828e-07, "loss": 0.235, "step": 9274 }, { "epoch": 1.75, "learning_rate": 1.9843246757510952e-07, "loss": 0.2033, "step": 9275 }, { "epoch": 1.75, "learning_rate": 1.9813392573184376e-07, "loss": 0.1786, "step": 9276 }, { "epoch": 1.75, "learning_rate": 1.9783559936912777e-07, "loss": 0.2077, "step": 9277 }, { "epoch": 1.75, "learning_rate": 1.975374885148873e-07, "loss": 0.2618, "step": 9278 }, { "epoch": 1.75, "learning_rate": 1.9723959319702946e-07, "loss": 0.2128, "step": 9279 }, { "epoch": 1.75, "learning_rate": 1.9694191344344083e-07, "loss": 0.2037, "step": 9280 }, { "epoch": 1.75, "learning_rate": 1.9664444928198744e-07, "loss": 0.2274, "step": 9281 }, { "epoch": 1.75, "learning_rate": 1.963472007405151e-07, "loss": 0.1822, "step": 9282 }, { "epoch": 1.75, "learning_rate": 1.960501678468496e-07, "loss": 0.1813, "step": 9283 }, { "epoch": 1.75, "learning_rate": 1.9575335062879675e-07, "loss": 0.2214, "step": 9284 }, { "epoch": 1.75, "learning_rate": 1.9545674911414207e-07, "loss": 0.2421, "step": 9285 }, { "epoch": 1.75, "learning_rate": 1.9516036333065035e-07, "loss": 0.2201, "step": 9286 }, { "epoch": 1.75, "learning_rate": 1.9486419330606715e-07, "loss": 0.2745, "step": 9287 }, { "epoch": 1.75, "learning_rate": 1.9456823906811695e-07, "loss": 0.2448, "step": 9288 }, { "epoch": 1.75, "learning_rate": 1.9427250064450453e-07, "loss": 0.2708, "step": 9289 }, { "epoch": 1.76, "learning_rate": 1.9397697806291416e-07, "loss": 0.2564, "step": 9290 }, { "epoch": 1.76, "learning_rate": 1.9368167135101064e-07, "loss": 0.184, "step": 9291 }, { "epoch": 1.76, "learning_rate": 1.933865805364374e-07, "loss": 0.2972, "step": 9292 }, { "epoch": 1.76, "learning_rate": 1.9309170564681816e-07, "loss": 0.2067, "step": 9293 }, { "epoch": 1.76, "learning_rate": 1.927970467097573e-07, "loss": 0.2724, "step": 9294 }, { "epoch": 1.76, "learning_rate": 1.925026037528374e-07, "loss": 0.2225, "step": 9295 }, { "epoch": 1.76, "learning_rate": 1.9220837680362204e-07, "loss": 0.2559, "step": 9296 }, { "epoch": 1.76, "learning_rate": 1.919143658896544e-07, "loss": 0.2749, "step": 9297 }, { "epoch": 1.76, "learning_rate": 1.9162057103845667e-07, "loss": 0.2424, "step": 9298 }, { "epoch": 1.76, "learning_rate": 1.9132699227753154e-07, "loss": 0.1977, "step": 9299 }, { "epoch": 1.76, "learning_rate": 1.9103362963436094e-07, "loss": 0.2085, "step": 9300 }, { "epoch": 1.76, "learning_rate": 1.907404831364079e-07, "loss": 0.2882, "step": 9301 }, { "epoch": 1.76, "learning_rate": 1.9044755281111354e-07, "loss": 0.16, "step": 9302 }, { "epoch": 1.76, "learning_rate": 1.9015483868589895e-07, "loss": 0.2349, "step": 9303 }, { "epoch": 1.76, "learning_rate": 1.8986234078816667e-07, "loss": 0.2053, "step": 9304 }, { "epoch": 1.76, "learning_rate": 1.895700591452973e-07, "loss": 0.2241, "step": 9305 }, { "epoch": 1.76, "learning_rate": 1.8927799378465089e-07, "loss": 0.2269, "step": 9306 }, { "epoch": 1.76, "learning_rate": 1.889861447335692e-07, "loss": 0.2706, "step": 9307 }, { "epoch": 1.76, "learning_rate": 1.88694512019372e-07, "loss": 0.2175, "step": 9308 }, { "epoch": 1.76, "learning_rate": 1.8840309566935945e-07, "loss": 0.2249, "step": 9309 }, { "epoch": 1.76, "learning_rate": 1.8811189571081134e-07, "loss": 0.2048, "step": 9310 }, { "epoch": 1.76, "learning_rate": 1.878209121709873e-07, "loss": 0.2515, "step": 9311 }, { "epoch": 1.76, "learning_rate": 1.875301450771269e-07, "loss": 0.1996, "step": 9312 }, { "epoch": 1.76, "learning_rate": 1.8723959445644897e-07, "loss": 0.1651, "step": 9313 }, { "epoch": 1.76, "learning_rate": 1.8694926033615223e-07, "loss": 0.2419, "step": 9314 }, { "epoch": 1.76, "learning_rate": 1.866591427434153e-07, "loss": 0.2184, "step": 9315 }, { "epoch": 1.76, "learning_rate": 1.8636924170539667e-07, "loss": 0.1978, "step": 9316 }, { "epoch": 1.76, "learning_rate": 1.860795572492341e-07, "loss": 0.1948, "step": 9317 }, { "epoch": 1.76, "learning_rate": 1.8579008940204536e-07, "loss": 0.2433, "step": 9318 }, { "epoch": 1.76, "learning_rate": 1.855008381909279e-07, "loss": 0.293, "step": 9319 }, { "epoch": 1.76, "learning_rate": 1.85211803642959e-07, "loss": 0.2186, "step": 9320 }, { "epoch": 1.76, "learning_rate": 1.8492298578519556e-07, "loss": 0.2412, "step": 9321 }, { "epoch": 1.76, "learning_rate": 1.846343846446738e-07, "loss": 0.2068, "step": 9322 }, { "epoch": 1.76, "learning_rate": 1.8434600024841044e-07, "loss": 0.1963, "step": 9323 }, { "epoch": 1.76, "learning_rate": 1.840578326234013e-07, "loss": 0.2267, "step": 9324 }, { "epoch": 1.76, "learning_rate": 1.8376988179662213e-07, "loss": 0.223, "step": 9325 }, { "epoch": 1.76, "learning_rate": 1.8348214779502825e-07, "loss": 0.1907, "step": 9326 }, { "epoch": 1.76, "learning_rate": 1.8319463064555536e-07, "loss": 0.1755, "step": 9327 }, { "epoch": 1.76, "learning_rate": 1.8290733037511721e-07, "loss": 0.2765, "step": 9328 }, { "epoch": 1.76, "learning_rate": 1.8262024701060978e-07, "loss": 0.2361, "step": 9329 }, { "epoch": 1.76, "learning_rate": 1.8233338057890625e-07, "loss": 0.204, "step": 9330 }, { "epoch": 1.76, "learning_rate": 1.820467311068605e-07, "loss": 0.1355, "step": 9331 }, { "epoch": 1.76, "learning_rate": 1.8176029862130711e-07, "loss": 0.1831, "step": 9332 }, { "epoch": 1.76, "learning_rate": 1.814740831490583e-07, "loss": 0.183, "step": 9333 }, { "epoch": 1.76, "learning_rate": 1.811880847169073e-07, "loss": 0.2695, "step": 9334 }, { "epoch": 1.76, "learning_rate": 1.8090230335162773e-07, "loss": 0.1924, "step": 9335 }, { "epoch": 1.76, "learning_rate": 1.8061673907997067e-07, "loss": 0.227, "step": 9336 }, { "epoch": 1.76, "learning_rate": 1.8033139192866839e-07, "loss": 0.2238, "step": 9337 }, { "epoch": 1.76, "learning_rate": 1.8004626192443364e-07, "loss": 0.3803, "step": 9338 }, { "epoch": 1.76, "learning_rate": 1.7976134909395622e-07, "loss": 0.214, "step": 9339 }, { "epoch": 1.76, "learning_rate": 1.7947665346390835e-07, "loss": 0.2108, "step": 9340 }, { "epoch": 1.76, "learning_rate": 1.7919217506094072e-07, "loss": 0.1743, "step": 9341 }, { "epoch": 1.76, "learning_rate": 1.7890791391168256e-07, "loss": 0.1378, "step": 9342 }, { "epoch": 1.77, "learning_rate": 1.7862387004274534e-07, "loss": 0.2505, "step": 9343 }, { "epoch": 1.77, "learning_rate": 1.7834004348071837e-07, "loss": 0.2636, "step": 9344 }, { "epoch": 1.77, "learning_rate": 1.7805643425216984e-07, "loss": 0.1766, "step": 9345 }, { "epoch": 1.77, "learning_rate": 1.7777304238365019e-07, "loss": 0.216, "step": 9346 }, { "epoch": 1.77, "learning_rate": 1.7748986790168815e-07, "loss": 0.2083, "step": 9347 }, { "epoch": 1.77, "learning_rate": 1.7720691083279035e-07, "loss": 0.2164, "step": 9348 }, { "epoch": 1.77, "learning_rate": 1.769241712034464e-07, "loss": 0.2473, "step": 9349 }, { "epoch": 1.77, "learning_rate": 1.766416490401235e-07, "loss": 0.1898, "step": 9350 }, { "epoch": 1.77, "learning_rate": 1.7635934436926878e-07, "loss": 0.2311, "step": 9351 }, { "epoch": 1.77, "learning_rate": 1.7607725721730916e-07, "loss": 0.1688, "step": 9352 }, { "epoch": 1.77, "learning_rate": 1.7579538761065106e-07, "loss": 0.2446, "step": 9353 }, { "epoch": 1.77, "learning_rate": 1.7551373557568058e-07, "loss": 0.1517, "step": 9354 }, { "epoch": 1.77, "learning_rate": 1.7523230113876382e-07, "loss": 0.2587, "step": 9355 }, { "epoch": 1.77, "learning_rate": 1.7495108432624612e-07, "loss": 0.169, "step": 9356 }, { "epoch": 1.77, "learning_rate": 1.7467008516445256e-07, "loss": 0.2615, "step": 9357 }, { "epoch": 1.77, "learning_rate": 1.743893036796876e-07, "loss": 0.2171, "step": 9358 }, { "epoch": 1.77, "learning_rate": 1.741087398982355e-07, "loss": 0.208, "step": 9359 }, { "epoch": 1.77, "learning_rate": 1.7382839384636057e-07, "loss": 0.2603, "step": 9360 }, { "epoch": 1.77, "learning_rate": 1.7354826555030624e-07, "loss": 0.1985, "step": 9361 }, { "epoch": 1.77, "learning_rate": 1.7326835503629542e-07, "loss": 0.2574, "step": 9362 }, { "epoch": 1.77, "learning_rate": 1.7298866233053103e-07, "loss": 0.2427, "step": 9363 }, { "epoch": 1.77, "learning_rate": 1.7270918745919547e-07, "loss": 0.2631, "step": 9364 }, { "epoch": 1.77, "learning_rate": 1.7242993044845085e-07, "loss": 0.2895, "step": 9365 }, { "epoch": 1.77, "learning_rate": 1.7215089132443852e-07, "loss": 0.2565, "step": 9366 }, { "epoch": 1.77, "learning_rate": 1.7187207011327979e-07, "loss": 0.2459, "step": 9367 }, { "epoch": 1.77, "learning_rate": 1.7159346684107515e-07, "loss": 0.2924, "step": 9368 }, { "epoch": 1.77, "learning_rate": 1.7131508153390596e-07, "loss": 0.2438, "step": 9369 }, { "epoch": 1.77, "learning_rate": 1.7103691421783086e-07, "loss": 0.2871, "step": 9370 }, { "epoch": 1.77, "learning_rate": 1.7075896491889065e-07, "loss": 0.1596, "step": 9371 }, { "epoch": 1.77, "learning_rate": 1.7048123366310427e-07, "loss": 0.2029, "step": 9372 }, { "epoch": 1.77, "learning_rate": 1.702037204764695e-07, "loss": 0.2321, "step": 9373 }, { "epoch": 1.77, "learning_rate": 1.6992642538496585e-07, "loss": 0.1805, "step": 9374 }, { "epoch": 1.77, "learning_rate": 1.6964934841455116e-07, "loss": 0.2118, "step": 9375 }, { "epoch": 1.77, "learning_rate": 1.6937248959116194e-07, "loss": 0.2467, "step": 9376 }, { "epoch": 1.77, "learning_rate": 1.690958489407163e-07, "loss": 0.1748, "step": 9377 }, { "epoch": 1.77, "learning_rate": 1.6881942648911077e-07, "loss": 0.3028, "step": 9378 }, { "epoch": 1.77, "learning_rate": 1.6854322226222103e-07, "loss": 0.1697, "step": 9379 }, { "epoch": 1.77, "learning_rate": 1.6826723628590336e-07, "loss": 0.2397, "step": 9380 }, { "epoch": 1.77, "learning_rate": 1.6799146858599318e-07, "loss": 0.2032, "step": 9381 }, { "epoch": 1.77, "learning_rate": 1.6771591918830544e-07, "loss": 0.2225, "step": 9382 }, { "epoch": 1.77, "learning_rate": 1.6744058811863444e-07, "loss": 0.1921, "step": 9383 }, { "epoch": 1.77, "learning_rate": 1.6716547540275464e-07, "loss": 0.1847, "step": 9384 }, { "epoch": 1.77, "learning_rate": 1.6689058106641897e-07, "loss": 0.3042, "step": 9385 }, { "epoch": 1.77, "learning_rate": 1.6661590513536162e-07, "loss": 0.2111, "step": 9386 }, { "epoch": 1.77, "learning_rate": 1.6634144763529454e-07, "loss": 0.2356, "step": 9387 }, { "epoch": 1.77, "learning_rate": 1.6606720859191045e-07, "loss": 0.2138, "step": 9388 }, { "epoch": 1.77, "learning_rate": 1.6579318803088078e-07, "loss": 0.2331, "step": 9389 }, { "epoch": 1.77, "learning_rate": 1.6551938597785754e-07, "loss": 0.2764, "step": 9390 }, { "epoch": 1.77, "learning_rate": 1.6524580245847132e-07, "loss": 0.2562, "step": 9391 }, { "epoch": 1.77, "learning_rate": 1.6497243749833275e-07, "loss": 0.1745, "step": 9392 }, { "epoch": 1.77, "learning_rate": 1.6469929112303162e-07, "loss": 0.2125, "step": 9393 }, { "epoch": 1.77, "learning_rate": 1.6442636335813777e-07, "loss": 0.2318, "step": 9394 }, { "epoch": 1.77, "learning_rate": 1.641536542292002e-07, "loss": 0.2322, "step": 9395 }, { "epoch": 1.78, "learning_rate": 1.6388116376174768e-07, "loss": 0.2775, "step": 9396 }, { "epoch": 1.78, "learning_rate": 1.636088919812881e-07, "loss": 0.1817, "step": 9397 }, { "epoch": 1.78, "learning_rate": 1.6333683891330914e-07, "loss": 0.2898, "step": 9398 }, { "epoch": 1.78, "learning_rate": 1.630650045832785e-07, "loss": 0.2072, "step": 9399 }, { "epoch": 1.78, "learning_rate": 1.6279338901664248e-07, "loss": 0.1518, "step": 9400 }, { "epoch": 1.78, "learning_rate": 1.6252199223882743e-07, "loss": 0.2317, "step": 9401 }, { "epoch": 1.78, "learning_rate": 1.6225081427523965e-07, "loss": 0.2752, "step": 9402 }, { "epoch": 1.78, "learning_rate": 1.6197985515126386e-07, "loss": 0.1479, "step": 9403 }, { "epoch": 1.78, "learning_rate": 1.6170911489226448e-07, "loss": 0.2102, "step": 9404 }, { "epoch": 1.78, "learning_rate": 1.614385935235871e-07, "loss": 0.1772, "step": 9405 }, { "epoch": 1.78, "learning_rate": 1.6116829107055475e-07, "loss": 0.256, "step": 9406 }, { "epoch": 1.78, "learning_rate": 1.608982075584703e-07, "loss": 0.1902, "step": 9407 }, { "epoch": 1.78, "learning_rate": 1.6062834301261793e-07, "loss": 0.2107, "step": 9408 }, { "epoch": 1.78, "learning_rate": 1.6035869745825912e-07, "loss": 0.1816, "step": 9409 }, { "epoch": 1.78, "learning_rate": 1.6008927092063532e-07, "loss": 0.2086, "step": 9410 }, { "epoch": 1.78, "learning_rate": 1.5982006342496914e-07, "loss": 0.267, "step": 9411 }, { "epoch": 1.78, "learning_rate": 1.5955107499646017e-07, "loss": 0.3091, "step": 9412 }, { "epoch": 1.78, "learning_rate": 1.5928230566028934e-07, "loss": 0.2666, "step": 9413 }, { "epoch": 1.78, "learning_rate": 1.5901375544161713e-07, "loss": 0.2441, "step": 9414 }, { "epoch": 1.78, "learning_rate": 1.5874542436558144e-07, "loss": 0.269, "step": 9415 }, { "epoch": 1.78, "learning_rate": 1.584773124573019e-07, "loss": 0.1987, "step": 9416 }, { "epoch": 1.78, "learning_rate": 1.5820941974187738e-07, "loss": 0.1831, "step": 9417 }, { "epoch": 1.78, "learning_rate": 1.5794174624438412e-07, "loss": 0.235, "step": 9418 }, { "epoch": 1.78, "learning_rate": 1.5767429198988048e-07, "loss": 0.1767, "step": 9419 }, { "epoch": 1.78, "learning_rate": 1.5740705700340337e-07, "loss": 0.1762, "step": 9420 }, { "epoch": 1.78, "learning_rate": 1.57140041309968e-07, "loss": 0.2777, "step": 9421 }, { "epoch": 1.78, "learning_rate": 1.5687324493457085e-07, "loss": 0.2435, "step": 9422 }, { "epoch": 1.78, "learning_rate": 1.5660666790218687e-07, "loss": 0.1979, "step": 9423 }, { "epoch": 1.78, "learning_rate": 1.563403102377703e-07, "loss": 0.2273, "step": 9424 }, { "epoch": 1.78, "learning_rate": 1.5607417196625592e-07, "loss": 0.2124, "step": 9425 }, { "epoch": 1.78, "learning_rate": 1.5580825311255655e-07, "loss": 0.2222, "step": 9426 }, { "epoch": 1.78, "learning_rate": 1.5554255370156562e-07, "loss": 0.2818, "step": 9427 }, { "epoch": 1.78, "learning_rate": 1.552770737581555e-07, "loss": 0.2093, "step": 9428 }, { "epoch": 1.78, "learning_rate": 1.550118133071782e-07, "loss": 0.1951, "step": 9429 }, { "epoch": 1.78, "learning_rate": 1.5474677237346468e-07, "loss": 0.2303, "step": 9430 }, { "epoch": 1.78, "learning_rate": 1.54481950981826e-07, "loss": 0.1809, "step": 9431 }, { "epoch": 1.78, "learning_rate": 1.5421734915705228e-07, "loss": 0.2517, "step": 9432 }, { "epoch": 1.78, "learning_rate": 1.5395296692391398e-07, "loss": 0.192, "step": 9433 }, { "epoch": 1.78, "learning_rate": 1.5368880430715937e-07, "loss": 0.1716, "step": 9434 }, { "epoch": 1.78, "learning_rate": 1.53424861331517e-07, "loss": 0.2381, "step": 9435 }, { "epoch": 1.78, "learning_rate": 1.5316113802169574e-07, "loss": 0.2114, "step": 9436 }, { "epoch": 1.78, "learning_rate": 1.528976344023822e-07, "loss": 0.2467, "step": 9437 }, { "epoch": 1.78, "learning_rate": 1.5263435049824332e-07, "loss": 0.2106, "step": 9438 }, { "epoch": 1.78, "learning_rate": 1.523712863339266e-07, "loss": 0.2729, "step": 9439 }, { "epoch": 1.78, "learning_rate": 1.5210844193405654e-07, "loss": 0.2227, "step": 9440 }, { "epoch": 1.78, "learning_rate": 1.5184581732323817e-07, "loss": 0.2356, "step": 9441 }, { "epoch": 1.78, "learning_rate": 1.5158341252605762e-07, "loss": 0.1919, "step": 9442 }, { "epoch": 1.78, "learning_rate": 1.5132122756707695e-07, "loss": 0.1575, "step": 9443 }, { "epoch": 1.78, "learning_rate": 1.5105926247084124e-07, "loss": 0.277, "step": 9444 }, { "epoch": 1.78, "learning_rate": 1.507975172618731e-07, "loss": 0.1765, "step": 9445 }, { "epoch": 1.78, "learning_rate": 1.50535991964674e-07, "loss": 0.2136, "step": 9446 }, { "epoch": 1.78, "learning_rate": 1.5027468660372608e-07, "loss": 0.2636, "step": 9447 }, { "epoch": 1.78, "learning_rate": 1.5001360120349113e-07, "loss": 0.1555, "step": 9448 }, { "epoch": 1.79, "learning_rate": 1.497527357884085e-07, "loss": 0.1619, "step": 9449 }, { "epoch": 1.79, "learning_rate": 1.4949209038289892e-07, "loss": 0.2831, "step": 9450 }, { "epoch": 1.79, "learning_rate": 1.4923166501136178e-07, "loss": 0.218, "step": 9451 }, { "epoch": 1.79, "learning_rate": 1.4897145969817478e-07, "loss": 0.1914, "step": 9452 }, { "epoch": 1.79, "learning_rate": 1.4871147446769735e-07, "loss": 0.2624, "step": 9453 }, { "epoch": 1.79, "learning_rate": 1.4845170934426668e-07, "loss": 0.211, "step": 9454 }, { "epoch": 1.79, "learning_rate": 1.4819216435219942e-07, "loss": 0.191, "step": 9455 }, { "epoch": 1.79, "learning_rate": 1.4793283951579196e-07, "loss": 0.2281, "step": 9456 }, { "epoch": 1.79, "learning_rate": 1.4767373485931992e-07, "loss": 0.224, "step": 9457 }, { "epoch": 1.79, "learning_rate": 1.4741485040703884e-07, "loss": 0.2236, "step": 9458 }, { "epoch": 1.79, "learning_rate": 1.47156186183183e-07, "loss": 0.2805, "step": 9459 }, { "epoch": 1.79, "learning_rate": 1.4689774221196606e-07, "loss": 0.2139, "step": 9460 }, { "epoch": 1.79, "learning_rate": 1.4663951851758173e-07, "loss": 0.1284, "step": 9461 }, { "epoch": 1.79, "learning_rate": 1.4638151512420206e-07, "loss": 0.2248, "step": 9462 }, { "epoch": 1.79, "learning_rate": 1.461237320559797e-07, "loss": 0.2012, "step": 9463 }, { "epoch": 1.79, "learning_rate": 1.4586616933704528e-07, "loss": 0.1565, "step": 9464 }, { "epoch": 1.79, "learning_rate": 1.456088269915104e-07, "loss": 0.2512, "step": 9465 }, { "epoch": 1.79, "learning_rate": 1.4535170504346468e-07, "loss": 0.1885, "step": 9466 }, { "epoch": 1.79, "learning_rate": 1.4509480351697773e-07, "loss": 0.1891, "step": 9467 }, { "epoch": 1.79, "learning_rate": 1.4483812243609868e-07, "loss": 0.2854, "step": 9468 }, { "epoch": 1.79, "learning_rate": 1.445816618248555e-07, "loss": 0.1519, "step": 9469 }, { "epoch": 1.79, "learning_rate": 1.4432542170725566e-07, "loss": 0.2121, "step": 9470 }, { "epoch": 1.79, "learning_rate": 1.4406940210728582e-07, "loss": 0.2528, "step": 9471 }, { "epoch": 1.79, "learning_rate": 1.4381360304891373e-07, "loss": 0.2404, "step": 9472 }, { "epoch": 1.79, "learning_rate": 1.435580245560836e-07, "loss": 0.2641, "step": 9473 }, { "epoch": 1.79, "learning_rate": 1.4330266665272068e-07, "loss": 0.1932, "step": 9474 }, { "epoch": 1.79, "learning_rate": 1.4304752936273036e-07, "loss": 0.2241, "step": 9475 }, { "epoch": 1.79, "learning_rate": 1.4279261270999516e-07, "loss": 0.2542, "step": 9476 }, { "epoch": 1.79, "learning_rate": 1.4253791671837798e-07, "loss": 0.2765, "step": 9477 }, { "epoch": 1.79, "learning_rate": 1.4228344141172275e-07, "loss": 0.221, "step": 9478 }, { "epoch": 1.79, "learning_rate": 1.4202918681385013e-07, "loss": 0.1829, "step": 9479 }, { "epoch": 1.79, "learning_rate": 1.4177515294856086e-07, "loss": 0.1939, "step": 9480 }, { "epoch": 1.79, "learning_rate": 1.4152133983963644e-07, "loss": 0.1818, "step": 9481 }, { "epoch": 1.79, "learning_rate": 1.4126774751083567e-07, "loss": 0.191, "step": 9482 }, { "epoch": 1.79, "learning_rate": 1.4101437598589813e-07, "loss": 0.2792, "step": 9483 }, { "epoch": 1.79, "learning_rate": 1.407612252885429e-07, "loss": 0.1851, "step": 9484 }, { "epoch": 1.79, "learning_rate": 1.405082954424658e-07, "loss": 0.228, "step": 9485 }, { "epoch": 1.79, "learning_rate": 1.402555864713459e-07, "loss": 0.2004, "step": 9486 }, { "epoch": 1.79, "learning_rate": 1.4000309839883901e-07, "loss": 0.2137, "step": 9487 }, { "epoch": 1.79, "learning_rate": 1.3975083124857986e-07, "loss": 0.1859, "step": 9488 }, { "epoch": 1.79, "learning_rate": 1.394987850441848e-07, "loss": 0.1991, "step": 9489 }, { "epoch": 1.79, "learning_rate": 1.3924695980924775e-07, "loss": 0.123, "step": 9490 }, { "epoch": 1.79, "learning_rate": 1.389953555673418e-07, "loss": 0.1805, "step": 9491 }, { "epoch": 1.79, "learning_rate": 1.3874397234202086e-07, "loss": 0.2522, "step": 9492 }, { "epoch": 1.79, "learning_rate": 1.384928101568167e-07, "loss": 0.2527, "step": 9493 }, { "epoch": 1.79, "learning_rate": 1.3824186903524107e-07, "loss": 0.2742, "step": 9494 }, { "epoch": 1.79, "learning_rate": 1.3799114900078458e-07, "loss": 0.2148, "step": 9495 }, { "epoch": 1.79, "learning_rate": 1.3774065007691794e-07, "loss": 0.2476, "step": 9496 }, { "epoch": 1.79, "learning_rate": 1.3749037228709072e-07, "loss": 0.1976, "step": 9497 }, { "epoch": 1.79, "learning_rate": 1.3724031565473112e-07, "loss": 0.2574, "step": 9498 }, { "epoch": 1.79, "learning_rate": 1.3699048020324763e-07, "loss": 0.2726, "step": 9499 }, { "epoch": 1.79, "learning_rate": 1.3674086595602764e-07, "loss": 0.1521, "step": 9500 }, { "epoch": 1.8, "learning_rate": 1.3649147293643777e-07, "loss": 0.2314, "step": 9501 }, { "epoch": 1.8, "learning_rate": 1.3624230116782377e-07, "loss": 0.2603, "step": 9502 }, { "epoch": 1.8, "learning_rate": 1.35993350673512e-07, "loss": 0.2062, "step": 9503 }, { "epoch": 1.8, "learning_rate": 1.3574462147680573e-07, "loss": 0.24, "step": 9504 }, { "epoch": 1.8, "learning_rate": 1.3549611360098885e-07, "loss": 0.1602, "step": 9505 }, { "epoch": 1.8, "learning_rate": 1.352478270693258e-07, "loss": 0.2321, "step": 9506 }, { "epoch": 1.8, "learning_rate": 1.3499976190505802e-07, "loss": 0.3507, "step": 9507 }, { "epoch": 1.8, "learning_rate": 1.3475191813140663e-07, "loss": 0.2426, "step": 9508 }, { "epoch": 1.8, "learning_rate": 1.345042957715742e-07, "loss": 0.1598, "step": 9509 }, { "epoch": 1.8, "learning_rate": 1.3425689484873966e-07, "loss": 0.2356, "step": 9510 }, { "epoch": 1.8, "learning_rate": 1.3400971538606284e-07, "loss": 0.1632, "step": 9511 }, { "epoch": 1.8, "learning_rate": 1.33762757406683e-07, "loss": 0.2595, "step": 9512 }, { "epoch": 1.8, "learning_rate": 1.335160209337172e-07, "loss": 0.2241, "step": 9513 }, { "epoch": 1.8, "learning_rate": 1.3326950599026334e-07, "loss": 0.2333, "step": 9514 }, { "epoch": 1.8, "learning_rate": 1.3302321259939883e-07, "loss": 0.2095, "step": 9515 }, { "epoch": 1.8, "learning_rate": 1.327771407841777e-07, "loss": 0.2982, "step": 9516 }, { "epoch": 1.8, "learning_rate": 1.3253129056763625e-07, "loss": 0.2661, "step": 9517 }, { "epoch": 1.8, "learning_rate": 1.3228566197278914e-07, "loss": 0.25, "step": 9518 }, { "epoch": 1.8, "learning_rate": 1.3204025502262852e-07, "loss": 0.1874, "step": 9519 }, { "epoch": 1.8, "learning_rate": 1.317950697401285e-07, "loss": 0.2606, "step": 9520 }, { "epoch": 1.8, "learning_rate": 1.3155010614824132e-07, "loss": 0.1286, "step": 9521 }, { "epoch": 1.8, "learning_rate": 1.3130536426989693e-07, "loss": 0.1889, "step": 9522 }, { "epoch": 1.8, "learning_rate": 1.31060844128007e-07, "loss": 0.2084, "step": 9523 }, { "epoch": 1.8, "learning_rate": 1.308165457454616e-07, "loss": 0.2259, "step": 9524 }, { "epoch": 1.8, "learning_rate": 1.3057246914512906e-07, "loss": 0.237, "step": 9525 }, { "epoch": 1.8, "learning_rate": 1.3032861434985804e-07, "loss": 0.1772, "step": 9526 }, { "epoch": 1.8, "learning_rate": 1.3008498138247587e-07, "loss": 0.1997, "step": 9527 }, { "epoch": 1.8, "learning_rate": 1.2984157026578987e-07, "loss": 0.2264, "step": 9528 }, { "epoch": 1.8, "learning_rate": 1.2959838102258537e-07, "loss": 0.2282, "step": 9529 }, { "epoch": 1.8, "learning_rate": 1.293554136756281e-07, "loss": 0.1995, "step": 9530 }, { "epoch": 1.8, "learning_rate": 1.2911266824766234e-07, "loss": 0.1514, "step": 9531 }, { "epoch": 1.8, "learning_rate": 1.2887014476141214e-07, "loss": 0.2214, "step": 9532 }, { "epoch": 1.8, "learning_rate": 1.2862784323957987e-07, "loss": 0.1868, "step": 9533 }, { "epoch": 1.8, "learning_rate": 1.2838576370484773e-07, "loss": 0.1947, "step": 9534 }, { "epoch": 1.8, "learning_rate": 1.2814390617987755e-07, "loss": 0.1607, "step": 9535 }, { "epoch": 1.8, "learning_rate": 1.2790227068730954e-07, "loss": 0.2691, "step": 9536 }, { "epoch": 1.8, "learning_rate": 1.2766085724976368e-07, "loss": 0.2139, "step": 9537 }, { "epoch": 1.8, "learning_rate": 1.2741966588983912e-07, "loss": 0.242, "step": 9538 }, { "epoch": 1.8, "learning_rate": 1.2717869663011362e-07, "loss": 0.2425, "step": 9539 }, { "epoch": 1.8, "learning_rate": 1.269379494931447e-07, "loss": 0.2525, "step": 9540 }, { "epoch": 1.8, "learning_rate": 1.266974245014696e-07, "loss": 0.1463, "step": 9541 }, { "epoch": 1.8, "learning_rate": 1.2645712167760338e-07, "loss": 0.2602, "step": 9542 }, { "epoch": 1.8, "learning_rate": 1.2621704104404165e-07, "loss": 0.1936, "step": 9543 }, { "epoch": 1.8, "learning_rate": 1.2597718262325808e-07, "loss": 0.2642, "step": 9544 }, { "epoch": 1.8, "learning_rate": 1.2573754643770698e-07, "loss": 0.2108, "step": 9545 }, { "epoch": 1.8, "learning_rate": 1.2549813250982035e-07, "loss": 0.2078, "step": 9546 }, { "epoch": 1.8, "learning_rate": 1.2525894086200972e-07, "loss": 0.2348, "step": 9547 }, { "epoch": 1.8, "learning_rate": 1.2501997151666722e-07, "loss": 0.2304, "step": 9548 }, { "epoch": 1.8, "learning_rate": 1.2478122449616215e-07, "loss": 0.2239, "step": 9549 }, { "epoch": 1.8, "learning_rate": 1.2454269982284383e-07, "loss": 0.2088, "step": 9550 }, { "epoch": 1.8, "learning_rate": 1.2430439751904195e-07, "loss": 0.158, "step": 9551 }, { "epoch": 1.8, "learning_rate": 1.2406631760706333e-07, "loss": 0.2355, "step": 9552 }, { "epoch": 1.8, "learning_rate": 1.2382846010919437e-07, "loss": 0.2132, "step": 9553 }, { "epoch": 1.81, "learning_rate": 1.2359082504770276e-07, "loss": 0.2623, "step": 9554 }, { "epoch": 1.81, "learning_rate": 1.2335341244483268e-07, "loss": 0.2483, "step": 9555 }, { "epoch": 1.81, "learning_rate": 1.2311622232280911e-07, "loss": 0.216, "step": 9556 }, { "epoch": 1.81, "learning_rate": 1.2287925470383572e-07, "loss": 0.1719, "step": 9557 }, { "epoch": 1.81, "learning_rate": 1.22642509610095e-07, "loss": 0.1853, "step": 9558 }, { "epoch": 1.81, "learning_rate": 1.2240598706374897e-07, "loss": 0.198, "step": 9559 }, { "epoch": 1.81, "learning_rate": 1.2216968708693932e-07, "loss": 0.2049, "step": 9560 }, { "epoch": 1.81, "learning_rate": 1.219336097017859e-07, "loss": 0.2275, "step": 9561 }, { "epoch": 1.81, "learning_rate": 1.2169775493038826e-07, "loss": 0.2503, "step": 9562 }, { "epoch": 1.81, "learning_rate": 1.2146212279482538e-07, "loss": 0.2052, "step": 9563 }, { "epoch": 1.81, "learning_rate": 1.2122671331715462e-07, "loss": 0.2019, "step": 9564 }, { "epoch": 1.81, "learning_rate": 1.2099152651941337e-07, "loss": 0.2307, "step": 9565 }, { "epoch": 1.81, "learning_rate": 1.2075656242361732e-07, "loss": 0.1812, "step": 9566 }, { "epoch": 1.81, "learning_rate": 1.2052182105176225e-07, "loss": 0.1635, "step": 9567 }, { "epoch": 1.81, "learning_rate": 1.202873024258222e-07, "loss": 0.2416, "step": 9568 }, { "epoch": 1.81, "learning_rate": 1.2005300656775077e-07, "loss": 0.2111, "step": 9569 }, { "epoch": 1.81, "learning_rate": 1.198189334994812e-07, "loss": 0.133, "step": 9570 }, { "epoch": 1.81, "learning_rate": 1.195850832429246e-07, "loss": 0.2286, "step": 9571 }, { "epoch": 1.81, "learning_rate": 1.1935145581997266e-07, "loss": 0.275, "step": 9572 }, { "epoch": 1.81, "learning_rate": 1.1911805125249504e-07, "loss": 0.3207, "step": 9573 }, { "epoch": 1.81, "learning_rate": 1.1888486956234124e-07, "loss": 0.2566, "step": 9574 }, { "epoch": 1.81, "learning_rate": 1.1865191077133908e-07, "loss": 0.2304, "step": 9575 }, { "epoch": 1.81, "learning_rate": 1.1841917490129779e-07, "loss": 0.1949, "step": 9576 }, { "epoch": 1.81, "learning_rate": 1.1818666197400242e-07, "loss": 0.2648, "step": 9577 }, { "epoch": 1.81, "learning_rate": 1.1795437201121917e-07, "loss": 0.2398, "step": 9578 }, { "epoch": 1.81, "learning_rate": 1.1772230503469372e-07, "loss": 0.1926, "step": 9579 }, { "epoch": 1.81, "learning_rate": 1.1749046106614926e-07, "loss": 0.2007, "step": 9580 }, { "epoch": 1.81, "learning_rate": 1.1725884012728922e-07, "loss": 0.2846, "step": 9581 }, { "epoch": 1.81, "learning_rate": 1.1702744223979656e-07, "loss": 0.2848, "step": 9582 }, { "epoch": 1.81, "learning_rate": 1.1679626742533201e-07, "loss": 0.2227, "step": 9583 }, { "epoch": 1.81, "learning_rate": 1.1656531570553575e-07, "loss": 0.23, "step": 9584 }, { "epoch": 1.81, "learning_rate": 1.1633458710202882e-07, "loss": 0.2128, "step": 9585 }, { "epoch": 1.81, "learning_rate": 1.1610408163640868e-07, "loss": 0.1919, "step": 9586 }, { "epoch": 1.81, "learning_rate": 1.1587379933025417e-07, "loss": 0.1962, "step": 9587 }, { "epoch": 1.81, "learning_rate": 1.1564374020512193e-07, "loss": 0.203, "step": 9588 }, { "epoch": 1.81, "learning_rate": 1.1541390428254778e-07, "loss": 0.1954, "step": 9589 }, { "epoch": 1.81, "learning_rate": 1.1518429158404731e-07, "loss": 0.1604, "step": 9590 }, { "epoch": 1.81, "learning_rate": 1.1495490213111526e-07, "loss": 0.2051, "step": 9591 }, { "epoch": 1.81, "learning_rate": 1.147257359452239e-07, "loss": 0.2401, "step": 9592 }, { "epoch": 1.81, "learning_rate": 1.1449679304782663e-07, "loss": 0.2998, "step": 9593 }, { "epoch": 1.81, "learning_rate": 1.1426807346035546e-07, "loss": 0.227, "step": 9594 }, { "epoch": 1.81, "learning_rate": 1.1403957720421965e-07, "loss": 0.2319, "step": 9595 }, { "epoch": 1.81, "learning_rate": 1.1381130430081044e-07, "loss": 0.265, "step": 9596 }, { "epoch": 1.81, "learning_rate": 1.13583254771496e-07, "loss": 0.3171, "step": 9597 }, { "epoch": 1.81, "learning_rate": 1.1335542863762483e-07, "loss": 0.2246, "step": 9598 }, { "epoch": 1.81, "learning_rate": 1.1312782592052374e-07, "loss": 0.2415, "step": 9599 }, { "epoch": 1.81, "learning_rate": 1.1290044664149874e-07, "loss": 0.2254, "step": 9600 }, { "epoch": 1.81, "learning_rate": 1.1267329082183532e-07, "loss": 0.181, "step": 9601 }, { "epoch": 1.81, "learning_rate": 1.1244635848279756e-07, "loss": 0.165, "step": 9602 }, { "epoch": 1.81, "learning_rate": 1.1221964964562931e-07, "loss": 0.1753, "step": 9603 }, { "epoch": 1.81, "learning_rate": 1.1199316433155277e-07, "loss": 0.1735, "step": 9604 }, { "epoch": 1.81, "learning_rate": 1.1176690256176931e-07, "loss": 0.2695, "step": 9605 }, { "epoch": 1.81, "learning_rate": 1.1154086435746031e-07, "loss": 0.3183, "step": 9606 }, { "epoch": 1.82, "learning_rate": 1.1131504973978469e-07, "loss": 0.1823, "step": 9607 }, { "epoch": 1.82, "learning_rate": 1.1108945872988164e-07, "loss": 0.3258, "step": 9608 }, { "epoch": 1.82, "learning_rate": 1.1086409134886872e-07, "loss": 0.2071, "step": 9609 }, { "epoch": 1.82, "learning_rate": 1.106389476178435e-07, "loss": 0.2226, "step": 9610 }, { "epoch": 1.82, "learning_rate": 1.1041402755788105e-07, "loss": 0.2428, "step": 9611 }, { "epoch": 1.82, "learning_rate": 1.1018933119003733e-07, "loss": 0.2211, "step": 9612 }, { "epoch": 1.82, "learning_rate": 1.0996485853534578e-07, "loss": 0.196, "step": 9613 }, { "epoch": 1.82, "learning_rate": 1.0974060961481986e-07, "loss": 0.1686, "step": 9614 }, { "epoch": 1.82, "learning_rate": 1.0951658444945168e-07, "loss": 0.2246, "step": 9615 }, { "epoch": 1.82, "learning_rate": 1.0929278306021252e-07, "loss": 0.1806, "step": 9616 }, { "epoch": 1.82, "learning_rate": 1.0906920546805255e-07, "loss": 0.2089, "step": 9617 }, { "epoch": 1.82, "learning_rate": 1.0884585169390172e-07, "loss": 0.2342, "step": 9618 }, { "epoch": 1.82, "learning_rate": 1.0862272175866829e-07, "loss": 0.2248, "step": 9619 }, { "epoch": 1.82, "learning_rate": 1.0839981568323887e-07, "loss": 0.2475, "step": 9620 }, { "epoch": 1.82, "learning_rate": 1.0817713348848097e-07, "loss": 0.2289, "step": 9621 }, { "epoch": 1.82, "learning_rate": 1.0795467519524012e-07, "loss": 0.2647, "step": 9622 }, { "epoch": 1.82, "learning_rate": 1.0773244082433992e-07, "loss": 0.242, "step": 9623 }, { "epoch": 1.82, "learning_rate": 1.0751043039658515e-07, "loss": 0.1855, "step": 9624 }, { "epoch": 1.82, "learning_rate": 1.0728864393275778e-07, "loss": 0.2086, "step": 9625 }, { "epoch": 1.82, "learning_rate": 1.0706708145361983e-07, "loss": 0.2516, "step": 9626 }, { "epoch": 1.82, "learning_rate": 1.0684574297991219e-07, "loss": 0.2445, "step": 9627 }, { "epoch": 1.82, "learning_rate": 1.0662462853235411e-07, "loss": 0.275, "step": 9628 }, { "epoch": 1.82, "learning_rate": 1.064037381316449e-07, "loss": 0.2596, "step": 9629 }, { "epoch": 1.82, "learning_rate": 1.0618307179846216e-07, "loss": 0.1981, "step": 9630 }, { "epoch": 1.82, "learning_rate": 1.0596262955346271e-07, "loss": 0.2689, "step": 9631 }, { "epoch": 1.82, "learning_rate": 1.0574241141728253e-07, "loss": 0.3078, "step": 9632 }, { "epoch": 1.82, "learning_rate": 1.0552241741053626e-07, "loss": 0.1868, "step": 9633 }, { "epoch": 1.82, "learning_rate": 1.0530264755381826e-07, "loss": 0.2383, "step": 9634 }, { "epoch": 1.82, "learning_rate": 1.0508310186770093e-07, "loss": 0.2507, "step": 9635 }, { "epoch": 1.82, "learning_rate": 1.0486378037273676e-07, "loss": 0.196, "step": 9636 }, { "epoch": 1.82, "learning_rate": 1.0464468308945624e-07, "loss": 0.1807, "step": 9637 }, { "epoch": 1.82, "learning_rate": 1.0442581003836966e-07, "loss": 0.263, "step": 9638 }, { "epoch": 1.82, "learning_rate": 1.0420716123996588e-07, "loss": 0.244, "step": 9639 }, { "epoch": 1.82, "learning_rate": 1.0398873671471272e-07, "loss": 0.1766, "step": 9640 }, { "epoch": 1.82, "learning_rate": 1.037705364830574e-07, "loss": 0.1741, "step": 9641 }, { "epoch": 1.82, "learning_rate": 1.0355256056542584e-07, "loss": 0.1969, "step": 9642 }, { "epoch": 1.82, "learning_rate": 1.033348089822231e-07, "loss": 0.2318, "step": 9643 }, { "epoch": 1.82, "learning_rate": 1.0311728175383311e-07, "loss": 0.1615, "step": 9644 }, { "epoch": 1.82, "learning_rate": 1.0289997890061853e-07, "loss": 0.2536, "step": 9645 }, { "epoch": 1.82, "learning_rate": 1.026829004429225e-07, "loss": 0.2133, "step": 9646 }, { "epoch": 1.82, "learning_rate": 1.0246604640106462e-07, "loss": 0.2535, "step": 9647 }, { "epoch": 1.82, "learning_rate": 1.0224941679534528e-07, "loss": 0.1406, "step": 9648 }, { "epoch": 1.82, "learning_rate": 1.020330116460444e-07, "loss": 0.1538, "step": 9649 }, { "epoch": 1.82, "learning_rate": 1.0181683097341854e-07, "loss": 0.1946, "step": 9650 }, { "epoch": 1.82, "learning_rate": 1.0160087479770514e-07, "loss": 0.1904, "step": 9651 }, { "epoch": 1.82, "learning_rate": 1.0138514313912079e-07, "loss": 0.2117, "step": 9652 }, { "epoch": 1.82, "learning_rate": 1.0116963601785961e-07, "loss": 0.2257, "step": 9653 }, { "epoch": 1.82, "learning_rate": 1.0095435345409521e-07, "loss": 0.2069, "step": 9654 }, { "epoch": 1.82, "learning_rate": 1.0073929546798172e-07, "loss": 0.1996, "step": 9655 }, { "epoch": 1.82, "learning_rate": 1.0052446207964973e-07, "loss": 0.2904, "step": 9656 }, { "epoch": 1.82, "learning_rate": 1.0030985330921062e-07, "loss": 0.2109, "step": 9657 }, { "epoch": 1.82, "learning_rate": 1.0009546917675472e-07, "loss": 0.2527, "step": 9658 }, { "epoch": 1.82, "learning_rate": 9.988130970234961e-08, "loss": 0.2571, "step": 9659 }, { "epoch": 1.83, "learning_rate": 9.966737490604366e-08, "loss": 0.2126, "step": 9660 }, { "epoch": 1.83, "learning_rate": 9.945366480786394e-08, "loss": 0.1929, "step": 9661 }, { "epoch": 1.83, "learning_rate": 9.924017942781528e-08, "loss": 0.2108, "step": 9662 }, { "epoch": 1.83, "learning_rate": 9.902691878588306e-08, "loss": 0.1623, "step": 9663 }, { "epoch": 1.83, "learning_rate": 9.881388290203075e-08, "loss": 0.1973, "step": 9664 }, { "epoch": 1.83, "learning_rate": 9.860107179620021e-08, "loss": 0.2606, "step": 9665 }, { "epoch": 1.83, "learning_rate": 9.838848548831353e-08, "loss": 0.1391, "step": 9666 }, { "epoch": 1.83, "learning_rate": 9.81761239982712e-08, "loss": 0.1999, "step": 9667 }, { "epoch": 1.83, "learning_rate": 9.796398734595285e-08, "loss": 0.2173, "step": 9668 }, { "epoch": 1.83, "learning_rate": 9.775207555121624e-08, "loss": 0.2671, "step": 9669 }, { "epoch": 1.83, "learning_rate": 9.754038863389908e-08, "loss": 0.2041, "step": 9670 }, { "epoch": 1.83, "learning_rate": 9.732892661381776e-08, "loss": 0.2077, "step": 9671 }, { "epoch": 1.83, "learning_rate": 9.7117689510767e-08, "loss": 0.2447, "step": 9672 }, { "epoch": 1.83, "learning_rate": 9.690667734452152e-08, "loss": 0.206, "step": 9673 }, { "epoch": 1.83, "learning_rate": 9.669589013483416e-08, "loss": 0.157, "step": 9674 }, { "epoch": 1.83, "learning_rate": 9.648532790143716e-08, "loss": 0.3583, "step": 9675 }, { "epoch": 1.83, "learning_rate": 9.627499066404116e-08, "loss": 0.2265, "step": 9676 }, { "epoch": 1.83, "learning_rate": 9.606487844233653e-08, "loss": 0.3436, "step": 9677 }, { "epoch": 1.83, "learning_rate": 9.585499125599167e-08, "loss": 0.2339, "step": 9678 }, { "epoch": 1.83, "learning_rate": 9.564532912465451e-08, "loss": 0.1823, "step": 9679 }, { "epoch": 1.83, "learning_rate": 9.54358920679524e-08, "loss": 0.1884, "step": 9680 }, { "epoch": 1.83, "learning_rate": 9.522668010549018e-08, "loss": 0.2252, "step": 9681 }, { "epoch": 1.83, "learning_rate": 9.50176932568525e-08, "loss": 0.2574, "step": 9682 }, { "epoch": 1.83, "learning_rate": 9.480893154160397e-08, "loss": 0.221, "step": 9683 }, { "epoch": 1.83, "learning_rate": 9.460039497928564e-08, "loss": 0.1709, "step": 9684 }, { "epoch": 1.83, "learning_rate": 9.43920835894191e-08, "loss": 0.2108, "step": 9685 }, { "epoch": 1.83, "learning_rate": 9.418399739150569e-08, "loss": 0.1836, "step": 9686 }, { "epoch": 1.83, "learning_rate": 9.397613640502345e-08, "loss": 0.2407, "step": 9687 }, { "epoch": 1.83, "learning_rate": 9.3768500649431e-08, "loss": 0.2004, "step": 9688 }, { "epoch": 1.83, "learning_rate": 9.356109014416581e-08, "loss": 0.2327, "step": 9689 }, { "epoch": 1.83, "learning_rate": 9.335390490864266e-08, "loss": 0.2175, "step": 9690 }, { "epoch": 1.83, "learning_rate": 9.31469449622574e-08, "loss": 0.1646, "step": 9691 }, { "epoch": 1.83, "learning_rate": 9.294021032438399e-08, "loss": 0.2055, "step": 9692 }, { "epoch": 1.83, "learning_rate": 9.27337010143739e-08, "loss": 0.1865, "step": 9693 }, { "epoch": 1.83, "learning_rate": 9.252741705155999e-08, "loss": 0.1788, "step": 9694 }, { "epoch": 1.83, "learning_rate": 9.232135845525236e-08, "loss": 0.161, "step": 9695 }, { "epoch": 1.83, "learning_rate": 9.211552524474005e-08, "loss": 0.2634, "step": 9696 }, { "epoch": 1.83, "learning_rate": 9.19099174392915e-08, "loss": 0.2178, "step": 9697 }, { "epoch": 1.83, "learning_rate": 9.170453505815413e-08, "loss": 0.2505, "step": 9698 }, { "epoch": 1.83, "learning_rate": 9.14993781205542e-08, "loss": 0.2012, "step": 9699 }, { "epoch": 1.83, "learning_rate": 9.129444664569637e-08, "loss": 0.2499, "step": 9700 }, { "epoch": 1.83, "learning_rate": 9.108974065276443e-08, "loss": 0.1975, "step": 9701 }, { "epoch": 1.83, "learning_rate": 9.088526016092142e-08, "loss": 0.288, "step": 9702 }, { "epoch": 1.83, "learning_rate": 9.068100518930895e-08, "loss": 0.3053, "step": 9703 }, { "epoch": 1.83, "learning_rate": 9.047697575704755e-08, "loss": 0.2319, "step": 9704 }, { "epoch": 1.83, "learning_rate": 9.02731718832367e-08, "loss": 0.1891, "step": 9705 }, { "epoch": 1.83, "learning_rate": 9.006959358695471e-08, "loss": 0.2583, "step": 9706 }, { "epoch": 1.83, "learning_rate": 8.986624088725888e-08, "loss": 0.2572, "step": 9707 }, { "epoch": 1.83, "learning_rate": 8.966311380318532e-08, "loss": 0.24, "step": 9708 }, { "epoch": 1.83, "learning_rate": 8.946021235374885e-08, "loss": 0.1458, "step": 9709 }, { "epoch": 1.83, "learning_rate": 8.925753655794317e-08, "loss": 0.2069, "step": 9710 }, { "epoch": 1.83, "learning_rate": 8.90550864347417e-08, "loss": 0.1918, "step": 9711 }, { "epoch": 1.83, "learning_rate": 8.885286200309512e-08, "loss": 0.2784, "step": 9712 }, { "epoch": 1.84, "learning_rate": 8.865086328193468e-08, "loss": 0.2359, "step": 9713 }, { "epoch": 1.84, "learning_rate": 8.844909029016941e-08, "loss": 0.2675, "step": 9714 }, { "epoch": 1.84, "learning_rate": 8.824754304668754e-08, "loss": 0.1963, "step": 9715 }, { "epoch": 1.84, "learning_rate": 8.80462215703562e-08, "loss": 0.2842, "step": 9716 }, { "epoch": 1.84, "learning_rate": 8.784512588002142e-08, "loss": 0.2217, "step": 9717 }, { "epoch": 1.84, "learning_rate": 8.76442559945076e-08, "loss": 0.2155, "step": 9718 }, { "epoch": 1.84, "learning_rate": 8.744361193261913e-08, "loss": 0.1771, "step": 9719 }, { "epoch": 1.84, "learning_rate": 8.724319371313794e-08, "loss": 0.1964, "step": 9720 }, { "epoch": 1.84, "learning_rate": 8.704300135482541e-08, "loss": 0.2069, "step": 9721 }, { "epoch": 1.84, "learning_rate": 8.684303487642264e-08, "loss": 0.2527, "step": 9722 }, { "epoch": 1.84, "learning_rate": 8.664329429664802e-08, "loss": 0.2561, "step": 9723 }, { "epoch": 1.84, "learning_rate": 8.644377963419909e-08, "loss": 0.2059, "step": 9724 }, { "epoch": 1.84, "learning_rate": 8.624449090775394e-08, "loss": 0.2193, "step": 9725 }, { "epoch": 1.84, "learning_rate": 8.604542813596711e-08, "loss": 0.2581, "step": 9726 }, { "epoch": 1.84, "learning_rate": 8.584659133747315e-08, "loss": 0.2362, "step": 9727 }, { "epoch": 1.84, "learning_rate": 8.564798053088658e-08, "loss": 0.1838, "step": 9728 }, { "epoch": 1.84, "learning_rate": 8.54495957347981e-08, "loss": 0.2688, "step": 9729 }, { "epoch": 1.84, "learning_rate": 8.525143696777982e-08, "loss": 0.3327, "step": 9730 }, { "epoch": 1.84, "learning_rate": 8.505350424838132e-08, "loss": 0.2777, "step": 9731 }, { "epoch": 1.84, "learning_rate": 8.485579759513113e-08, "loss": 0.2295, "step": 9732 }, { "epoch": 1.84, "learning_rate": 8.465831702653693e-08, "loss": 0.2248, "step": 9733 }, { "epoch": 1.84, "learning_rate": 8.44610625610856e-08, "loss": 0.2107, "step": 9734 }, { "epoch": 1.84, "learning_rate": 8.426403421724128e-08, "loss": 0.2676, "step": 9735 }, { "epoch": 1.84, "learning_rate": 8.406723201344891e-08, "loss": 0.1614, "step": 9736 }, { "epoch": 1.84, "learning_rate": 8.387065596813154e-08, "loss": 0.2258, "step": 9737 }, { "epoch": 1.84, "learning_rate": 8.367430609969002e-08, "loss": 0.1831, "step": 9738 }, { "epoch": 1.84, "learning_rate": 8.347818242650573e-08, "loss": 0.2741, "step": 9739 }, { "epoch": 1.84, "learning_rate": 8.328228496693764e-08, "loss": 0.2454, "step": 9740 }, { "epoch": 1.84, "learning_rate": 8.308661373932409e-08, "loss": 0.2716, "step": 9741 }, { "epoch": 1.84, "learning_rate": 8.289116876198212e-08, "loss": 0.1997, "step": 9742 }, { "epoch": 1.84, "learning_rate": 8.269595005320735e-08, "loss": 0.2224, "step": 9743 }, { "epoch": 1.84, "learning_rate": 8.250095763127463e-08, "loss": 0.2403, "step": 9744 }, { "epoch": 1.84, "learning_rate": 8.230619151443769e-08, "loss": 0.1792, "step": 9745 }, { "epoch": 1.84, "learning_rate": 8.211165172092833e-08, "loss": 0.2455, "step": 9746 }, { "epoch": 1.84, "learning_rate": 8.19173382689581e-08, "loss": 0.2371, "step": 9747 }, { "epoch": 1.84, "learning_rate": 8.172325117671658e-08, "loss": 0.319, "step": 9748 }, { "epoch": 1.84, "learning_rate": 8.152939046237263e-08, "loss": 0.2294, "step": 9749 }, { "epoch": 1.84, "learning_rate": 8.133575614407419e-08, "loss": 0.294, "step": 9750 }, { "epoch": 1.84, "learning_rate": 8.114234823994677e-08, "loss": 0.214, "step": 9751 }, { "epoch": 1.84, "learning_rate": 8.094916676809616e-08, "loss": 0.2227, "step": 9752 }, { "epoch": 1.84, "learning_rate": 8.075621174660626e-08, "loss": 0.2003, "step": 9753 }, { "epoch": 1.84, "learning_rate": 8.05634831935398e-08, "loss": 0.1781, "step": 9754 }, { "epoch": 1.84, "learning_rate": 8.037098112693764e-08, "loss": 0.1421, "step": 9755 }, { "epoch": 1.84, "learning_rate": 8.017870556482177e-08, "loss": 0.2264, "step": 9756 }, { "epoch": 1.84, "learning_rate": 7.998665652518973e-08, "loss": 0.233, "step": 9757 }, { "epoch": 1.84, "learning_rate": 7.97948340260199e-08, "loss": 0.2148, "step": 9758 }, { "epoch": 1.84, "learning_rate": 7.96032380852696e-08, "loss": 0.2242, "step": 9759 }, { "epoch": 1.84, "learning_rate": 7.941186872087364e-08, "loss": 0.2065, "step": 9760 }, { "epoch": 1.84, "learning_rate": 7.922072595074686e-08, "loss": 0.1804, "step": 9761 }, { "epoch": 1.84, "learning_rate": 7.902980979278247e-08, "loss": 0.1959, "step": 9762 }, { "epoch": 1.84, "learning_rate": 7.883912026485141e-08, "loss": 0.2897, "step": 9763 }, { "epoch": 1.84, "learning_rate": 7.864865738480526e-08, "loss": 0.196, "step": 9764 }, { "epoch": 1.84, "learning_rate": 7.845842117047364e-08, "loss": 0.2041, "step": 9765 }, { "epoch": 1.85, "learning_rate": 7.826841163966397e-08, "loss": 0.2059, "step": 9766 }, { "epoch": 1.85, "learning_rate": 7.807862881016398e-08, "loss": 0.181, "step": 9767 }, { "epoch": 1.85, "learning_rate": 7.788907269973944e-08, "loss": 0.2446, "step": 9768 }, { "epoch": 1.85, "learning_rate": 7.769974332613423e-08, "loss": 0.2267, "step": 9769 }, { "epoch": 1.85, "learning_rate": 7.751064070707248e-08, "loss": 0.1709, "step": 9770 }, { "epoch": 1.85, "learning_rate": 7.732176486025589e-08, "loss": 0.2501, "step": 9771 }, { "epoch": 1.85, "learning_rate": 7.713311580336558e-08, "loss": 0.19, "step": 9772 }, { "epoch": 1.85, "learning_rate": 7.694469355406131e-08, "loss": 0.2559, "step": 9773 }, { "epoch": 1.85, "learning_rate": 7.67564981299812e-08, "loss": 0.2504, "step": 9774 }, { "epoch": 1.85, "learning_rate": 7.656852954874255e-08, "loss": 0.1708, "step": 9775 }, { "epoch": 1.85, "learning_rate": 7.638078782794156e-08, "loss": 0.2899, "step": 9776 }, { "epoch": 1.85, "learning_rate": 7.619327298515278e-08, "loss": 0.2269, "step": 9777 }, { "epoch": 1.85, "learning_rate": 7.600598503792967e-08, "loss": 0.2319, "step": 9778 }, { "epoch": 1.85, "learning_rate": 7.581892400380486e-08, "loss": 0.2463, "step": 9779 }, { "epoch": 1.85, "learning_rate": 7.563208990028881e-08, "loss": 0.1922, "step": 9780 }, { "epoch": 1.85, "learning_rate": 7.54454827448714e-08, "loss": 0.1956, "step": 9781 }, { "epoch": 1.85, "learning_rate": 7.525910255502173e-08, "loss": 0.3268, "step": 9782 }, { "epoch": 1.85, "learning_rate": 7.507294934818638e-08, "loss": 0.2444, "step": 9783 }, { "epoch": 1.85, "learning_rate": 7.488702314179169e-08, "loss": 0.1765, "step": 9784 }, { "epoch": 1.85, "learning_rate": 7.470132395324236e-08, "loss": 0.2353, "step": 9785 }, { "epoch": 1.85, "learning_rate": 7.451585179992199e-08, "loss": 0.1509, "step": 9786 }, { "epoch": 1.85, "learning_rate": 7.433060669919307e-08, "loss": 0.2505, "step": 9787 }, { "epoch": 1.85, "learning_rate": 7.414558866839589e-08, "loss": 0.2409, "step": 9788 }, { "epoch": 1.85, "learning_rate": 7.396079772485132e-08, "loss": 0.2473, "step": 9789 }, { "epoch": 1.85, "learning_rate": 7.377623388585692e-08, "loss": 0.2474, "step": 9790 }, { "epoch": 1.85, "learning_rate": 7.359189716869025e-08, "loss": 0.2042, "step": 9791 }, { "epoch": 1.85, "learning_rate": 7.340778759060779e-08, "loss": 0.2063, "step": 9792 }, { "epoch": 1.85, "learning_rate": 7.322390516884353e-08, "loss": 0.2067, "step": 9793 }, { "epoch": 1.85, "learning_rate": 7.304024992061093e-08, "loss": 0.2382, "step": 9794 }, { "epoch": 1.85, "learning_rate": 7.285682186310316e-08, "loss": 0.2644, "step": 9795 }, { "epoch": 1.85, "learning_rate": 7.267362101349012e-08, "loss": 0.223, "step": 9796 }, { "epoch": 1.85, "learning_rate": 7.24906473889217e-08, "loss": 0.2101, "step": 9797 }, { "epoch": 1.85, "learning_rate": 7.230790100652668e-08, "loss": 0.2002, "step": 9798 }, { "epoch": 1.85, "learning_rate": 7.212538188341167e-08, "loss": 0.2526, "step": 9799 }, { "epoch": 1.85, "learning_rate": 7.194309003666272e-08, "loss": 0.1931, "step": 9800 }, { "epoch": 1.85, "learning_rate": 7.17610254833448e-08, "loss": 0.2326, "step": 9801 }, { "epoch": 1.85, "learning_rate": 7.157918824050037e-08, "loss": 0.2561, "step": 9802 }, { "epoch": 1.85, "learning_rate": 7.139757832515221e-08, "loss": 0.3116, "step": 9803 }, { "epoch": 1.85, "learning_rate": 7.121619575430061e-08, "loss": 0.2748, "step": 9804 }, { "epoch": 1.85, "learning_rate": 7.103504054492533e-08, "loss": 0.245, "step": 9805 }, { "epoch": 1.85, "learning_rate": 7.085411271398418e-08, "loss": 0.1509, "step": 9806 }, { "epoch": 1.85, "learning_rate": 7.067341227841418e-08, "loss": 0.247, "step": 9807 }, { "epoch": 1.85, "learning_rate": 7.049293925513095e-08, "loss": 0.1658, "step": 9808 }, { "epoch": 1.85, "learning_rate": 7.031269366102905e-08, "loss": 0.252, "step": 9809 }, { "epoch": 1.85, "learning_rate": 7.013267551298108e-08, "loss": 0.2258, "step": 9810 }, { "epoch": 1.85, "learning_rate": 6.995288482783886e-08, "loss": 0.305, "step": 9811 }, { "epoch": 1.85, "learning_rate": 6.977332162243305e-08, "loss": 0.2362, "step": 9812 }, { "epoch": 1.85, "learning_rate": 6.959398591357247e-08, "loss": 0.1538, "step": 9813 }, { "epoch": 1.85, "learning_rate": 6.941487771804533e-08, "loss": 0.1892, "step": 9814 }, { "epoch": 1.85, "learning_rate": 6.923599705261769e-08, "loss": 0.1912, "step": 9815 }, { "epoch": 1.85, "learning_rate": 6.90573439340353e-08, "loss": 0.2931, "step": 9816 }, { "epoch": 1.85, "learning_rate": 6.887891837902172e-08, "loss": 0.2692, "step": 9817 }, { "epoch": 1.85, "learning_rate": 6.870072040427972e-08, "loss": 0.2047, "step": 9818 }, { "epoch": 1.86, "learning_rate": 6.852275002649039e-08, "loss": 0.2727, "step": 9819 }, { "epoch": 1.86, "learning_rate": 6.83450072623143e-08, "loss": 0.2064, "step": 9820 }, { "epoch": 1.86, "learning_rate": 6.816749212839008e-08, "loss": 0.2278, "step": 9821 }, { "epoch": 1.86, "learning_rate": 6.799020464133444e-08, "loss": 0.2405, "step": 9822 }, { "epoch": 1.86, "learning_rate": 6.781314481774437e-08, "loss": 0.1997, "step": 9823 }, { "epoch": 1.86, "learning_rate": 6.763631267419413e-08, "loss": 0.2737, "step": 9824 }, { "epoch": 1.86, "learning_rate": 6.745970822723712e-08, "loss": 0.2302, "step": 9825 }, { "epoch": 1.86, "learning_rate": 6.728333149340626e-08, "loss": 0.2227, "step": 9826 }, { "epoch": 1.86, "learning_rate": 6.710718248921167e-08, "loss": 0.2071, "step": 9827 }, { "epoch": 1.86, "learning_rate": 6.693126123114262e-08, "loss": 0.1603, "step": 9828 }, { "epoch": 1.86, "learning_rate": 6.675556773566822e-08, "loss": 0.2047, "step": 9829 }, { "epoch": 1.86, "learning_rate": 6.65801020192347e-08, "loss": 0.2954, "step": 9830 }, { "epoch": 1.86, "learning_rate": 6.640486409826785e-08, "loss": 0.2418, "step": 9831 }, { "epoch": 1.86, "learning_rate": 6.622985398917232e-08, "loss": 0.2004, "step": 9832 }, { "epoch": 1.86, "learning_rate": 6.605507170832998e-08, "loss": 0.2599, "step": 9833 }, { "epoch": 1.86, "learning_rate": 6.588051727210331e-08, "loss": 0.2094, "step": 9834 }, { "epoch": 1.86, "learning_rate": 6.570619069683232e-08, "loss": 0.2488, "step": 9835 }, { "epoch": 1.86, "learning_rate": 6.553209199883559e-08, "loss": 0.2128, "step": 9836 }, { "epoch": 1.86, "learning_rate": 6.53582211944112e-08, "loss": 0.1484, "step": 9837 }, { "epoch": 1.86, "learning_rate": 6.51845782998356e-08, "loss": 0.2254, "step": 9838 }, { "epoch": 1.86, "learning_rate": 6.501116333136298e-08, "loss": 0.2406, "step": 9839 }, { "epoch": 1.86, "learning_rate": 6.483797630522731e-08, "loss": 0.2313, "step": 9840 }, { "epoch": 1.86, "learning_rate": 6.46650172376409e-08, "loss": 0.1876, "step": 9841 }, { "epoch": 1.86, "learning_rate": 6.44922861447947e-08, "loss": 0.2366, "step": 9842 }, { "epoch": 1.86, "learning_rate": 6.431978304285824e-08, "loss": 0.2066, "step": 9843 }, { "epoch": 1.86, "learning_rate": 6.414750794797947e-08, "loss": 0.1861, "step": 9844 }, { "epoch": 1.86, "learning_rate": 6.397546087628575e-08, "loss": 0.2747, "step": 9845 }, { "epoch": 1.86, "learning_rate": 6.380364184388226e-08, "loss": 0.2173, "step": 9846 }, { "epoch": 1.86, "learning_rate": 6.363205086685337e-08, "loss": 0.2132, "step": 9847 }, { "epoch": 1.86, "learning_rate": 6.346068796126176e-08, "loss": 0.2153, "step": 9848 }, { "epoch": 1.86, "learning_rate": 6.328955314314906e-08, "loss": 0.247, "step": 9849 }, { "epoch": 1.86, "learning_rate": 6.311864642853521e-08, "loss": 0.2159, "step": 9850 }, { "epoch": 1.86, "learning_rate": 6.294796783341938e-08, "loss": 0.235, "step": 9851 }, { "epoch": 1.86, "learning_rate": 6.27775173737788e-08, "loss": 0.2528, "step": 9852 }, { "epoch": 1.86, "learning_rate": 6.260729506556956e-08, "loss": 0.2606, "step": 9853 }, { "epoch": 1.86, "learning_rate": 6.243730092472644e-08, "loss": 0.2424, "step": 9854 }, { "epoch": 1.86, "learning_rate": 6.226753496716254e-08, "loss": 0.2332, "step": 9855 }, { "epoch": 1.86, "learning_rate": 6.209799720877014e-08, "loss": 0.2744, "step": 9856 }, { "epoch": 1.86, "learning_rate": 6.192868766541987e-08, "loss": 0.2489, "step": 9857 }, { "epoch": 1.86, "learning_rate": 6.175960635296102e-08, "loss": 0.1996, "step": 9858 }, { "epoch": 1.86, "learning_rate": 6.15907532872212e-08, "loss": 0.2456, "step": 9859 }, { "epoch": 1.86, "learning_rate": 6.142212848400719e-08, "loss": 0.2094, "step": 9860 }, { "epoch": 1.86, "learning_rate": 6.12537319591039e-08, "loss": 0.2631, "step": 9861 }, { "epoch": 1.86, "learning_rate": 6.108556372827617e-08, "loss": 0.1901, "step": 9862 }, { "epoch": 1.86, "learning_rate": 6.091762380726507e-08, "loss": 0.2754, "step": 9863 }, { "epoch": 1.86, "learning_rate": 6.074991221179217e-08, "loss": 0.2331, "step": 9864 }, { "epoch": 1.86, "learning_rate": 6.058242895755767e-08, "loss": 0.2701, "step": 9865 }, { "epoch": 1.86, "learning_rate": 6.041517406023934e-08, "loss": 0.1961, "step": 9866 }, { "epoch": 1.86, "learning_rate": 6.02481475354938e-08, "loss": 0.2835, "step": 9867 }, { "epoch": 1.86, "learning_rate": 6.008134939895743e-08, "loss": 0.2085, "step": 9868 }, { "epoch": 1.86, "learning_rate": 5.991477966624415e-08, "loss": 0.2222, "step": 9869 }, { "epoch": 1.86, "learning_rate": 5.974843835294619e-08, "loss": 0.1752, "step": 9870 }, { "epoch": 1.86, "learning_rate": 5.958232547463583e-08, "loss": 0.1904, "step": 9871 }, { "epoch": 1.87, "learning_rate": 5.941644104686256e-08, "loss": 0.2139, "step": 9872 }, { "epoch": 1.87, "learning_rate": 5.925078508515508e-08, "loss": 0.2914, "step": 9873 }, { "epoch": 1.87, "learning_rate": 5.908535760502099e-08, "loss": 0.2504, "step": 9874 }, { "epoch": 1.87, "learning_rate": 5.8920158621945666e-08, "loss": 0.1648, "step": 9875 }, { "epoch": 1.87, "learning_rate": 5.875518815139397e-08, "loss": 0.2049, "step": 9876 }, { "epoch": 1.87, "learning_rate": 5.859044620880855e-08, "loss": 0.1747, "step": 9877 }, { "epoch": 1.87, "learning_rate": 5.842593280961179e-08, "loss": 0.2188, "step": 9878 }, { "epoch": 1.87, "learning_rate": 5.826164796920331e-08, "loss": 0.2582, "step": 9879 }, { "epoch": 1.87, "learning_rate": 5.809759170296248e-08, "loss": 0.2805, "step": 9880 }, { "epoch": 1.87, "learning_rate": 5.793376402624673e-08, "loss": 0.2121, "step": 9881 }, { "epoch": 1.87, "learning_rate": 5.777016495439186e-08, "loss": 0.2155, "step": 9882 }, { "epoch": 1.87, "learning_rate": 5.7606794502713116e-08, "loss": 0.2471, "step": 9883 }, { "epoch": 1.87, "learning_rate": 5.744365268650326e-08, "loss": 0.2769, "step": 9884 }, { "epoch": 1.87, "learning_rate": 5.7280739521034525e-08, "loss": 0.3135, "step": 9885 }, { "epoch": 1.87, "learning_rate": 5.711805502155748e-08, "loss": 0.2059, "step": 9886 }, { "epoch": 1.87, "learning_rate": 5.695559920330079e-08, "loss": 0.2523, "step": 9887 }, { "epoch": 1.87, "learning_rate": 5.679337208147256e-08, "loss": 0.3009, "step": 9888 }, { "epoch": 1.87, "learning_rate": 5.663137367125898e-08, "loss": 0.3272, "step": 9889 }, { "epoch": 1.87, "learning_rate": 5.646960398782486e-08, "loss": 0.2414, "step": 9890 }, { "epoch": 1.87, "learning_rate": 5.630806304631392e-08, "loss": 0.1691, "step": 9891 }, { "epoch": 1.87, "learning_rate": 5.61467508618474e-08, "loss": 0.1788, "step": 9892 }, { "epoch": 1.87, "learning_rate": 5.59856674495271e-08, "loss": 0.2709, "step": 9893 }, { "epoch": 1.87, "learning_rate": 5.582481282443153e-08, "loss": 0.275, "step": 9894 }, { "epoch": 1.87, "learning_rate": 5.5664187001618354e-08, "loss": 0.2116, "step": 9895 }, { "epoch": 1.87, "learning_rate": 5.550378999612471e-08, "loss": 0.3213, "step": 9896 }, { "epoch": 1.87, "learning_rate": 5.534362182296471e-08, "loss": 0.2021, "step": 9897 }, { "epoch": 1.87, "learning_rate": 5.5183682497132184e-08, "loss": 0.2845, "step": 9898 }, { "epoch": 1.87, "learning_rate": 5.50239720335996e-08, "loss": 0.2059, "step": 9899 }, { "epoch": 1.87, "learning_rate": 5.486449044731751e-08, "loss": 0.2072, "step": 9900 }, { "epoch": 1.87, "learning_rate": 5.470523775321451e-08, "loss": 0.2543, "step": 9901 }, { "epoch": 1.87, "learning_rate": 5.454621396619953e-08, "loss": 0.3189, "step": 9902 }, { "epoch": 1.87, "learning_rate": 5.4387419101157856e-08, "loss": 0.229, "step": 9903 }, { "epoch": 1.87, "learning_rate": 5.42288531729554e-08, "loss": 0.2144, "step": 9904 }, { "epoch": 1.87, "learning_rate": 5.407051619643555e-08, "loss": 0.17, "step": 9905 }, { "epoch": 1.87, "learning_rate": 5.391240818642007e-08, "loss": 0.2525, "step": 9906 }, { "epoch": 1.87, "learning_rate": 5.375452915770962e-08, "loss": 0.3279, "step": 9907 }, { "epoch": 1.87, "learning_rate": 5.3596879125084324e-08, "loss": 0.2323, "step": 9908 }, { "epoch": 1.87, "learning_rate": 5.3439458103300426e-08, "loss": 0.1984, "step": 9909 }, { "epoch": 1.87, "learning_rate": 5.328226610709586e-08, "loss": 0.3852, "step": 9910 }, { "epoch": 1.87, "learning_rate": 5.3125303151184973e-08, "loss": 0.2268, "step": 9911 }, { "epoch": 1.87, "learning_rate": 5.296856925026073e-08, "loss": 0.1618, "step": 9912 }, { "epoch": 1.87, "learning_rate": 5.2812064418995836e-08, "loss": 0.2321, "step": 9913 }, { "epoch": 1.87, "learning_rate": 5.2655788672040795e-08, "loss": 0.2205, "step": 9914 }, { "epoch": 1.87, "learning_rate": 5.2499742024024744e-08, "loss": 0.2303, "step": 9915 }, { "epoch": 1.87, "learning_rate": 5.234392448955544e-08, "loss": 0.2393, "step": 9916 }, { "epoch": 1.87, "learning_rate": 5.2188336083218994e-08, "loss": 0.2233, "step": 9917 }, { "epoch": 1.87, "learning_rate": 5.203297681958014e-08, "loss": 0.2028, "step": 9918 }, { "epoch": 1.87, "learning_rate": 5.18778467131828e-08, "loss": 0.2071, "step": 9919 }, { "epoch": 1.87, "learning_rate": 5.172294577854814e-08, "loss": 0.157, "step": 9920 }, { "epoch": 1.87, "learning_rate": 5.156827403017733e-08, "loss": 0.2218, "step": 9921 }, { "epoch": 1.87, "learning_rate": 5.141383148254908e-08, "loss": 0.1937, "step": 9922 }, { "epoch": 1.87, "learning_rate": 5.12596181501207e-08, "loss": 0.3017, "step": 9923 }, { "epoch": 1.87, "learning_rate": 5.110563404732871e-08, "loss": 0.2071, "step": 9924 }, { "epoch": 1.88, "learning_rate": 5.09518791885874e-08, "loss": 0.2461, "step": 9925 }, { "epoch": 1.88, "learning_rate": 5.0798353588290275e-08, "loss": 0.1752, "step": 9926 }, { "epoch": 1.88, "learning_rate": 5.064505726080915e-08, "loss": 0.195, "step": 9927 }, { "epoch": 1.88, "learning_rate": 5.049199022049367e-08, "loss": 0.2716, "step": 9928 }, { "epoch": 1.88, "learning_rate": 5.033915248167293e-08, "loss": 0.2754, "step": 9929 }, { "epoch": 1.88, "learning_rate": 5.018654405865492e-08, "loss": 0.212, "step": 9930 }, { "epoch": 1.88, "learning_rate": 5.0034164965724606e-08, "loss": 0.1905, "step": 9931 }, { "epoch": 1.88, "learning_rate": 4.9882015217146416e-08, "loss": 0.1824, "step": 9932 }, { "epoch": 1.88, "learning_rate": 4.9730094827164234e-08, "loss": 0.2307, "step": 9933 }, { "epoch": 1.88, "learning_rate": 4.957840380999807e-08, "loss": 0.2257, "step": 9934 }, { "epoch": 1.88, "learning_rate": 4.942694217984906e-08, "loss": 0.2114, "step": 9935 }, { "epoch": 1.88, "learning_rate": 4.927570995089559e-08, "loss": 0.1773, "step": 9936 }, { "epoch": 1.88, "learning_rate": 4.9124707137294115e-08, "loss": 0.2128, "step": 9937 }, { "epoch": 1.88, "learning_rate": 4.8973933753180546e-08, "loss": 0.2296, "step": 9938 }, { "epoch": 1.88, "learning_rate": 4.882338981266943e-08, "loss": 0.2121, "step": 9939 }, { "epoch": 1.88, "learning_rate": 4.867307532985227e-08, "loss": 0.1587, "step": 9940 }, { "epoch": 1.88, "learning_rate": 4.8522990318801146e-08, "loss": 0.2473, "step": 9941 }, { "epoch": 1.88, "learning_rate": 4.837313479356565e-08, "loss": 0.2525, "step": 9942 }, { "epoch": 1.88, "learning_rate": 4.8223508768173165e-08, "loss": 0.323, "step": 9943 }, { "epoch": 1.88, "learning_rate": 4.807411225663083e-08, "loss": 0.2153, "step": 9944 }, { "epoch": 1.88, "learning_rate": 4.792494527292413e-08, "loss": 0.2269, "step": 9945 }, { "epoch": 1.88, "learning_rate": 4.77760078310166e-08, "loss": 0.1948, "step": 9946 }, { "epoch": 1.88, "learning_rate": 4.7627299944850436e-08, "loss": 0.2356, "step": 9947 }, { "epoch": 1.88, "learning_rate": 4.747882162834616e-08, "loss": 0.1934, "step": 9948 }, { "epoch": 1.88, "learning_rate": 4.733057289540322e-08, "loss": 0.2418, "step": 9949 }, { "epoch": 1.88, "learning_rate": 4.718255375989911e-08, "loss": 0.2606, "step": 9950 }, { "epoch": 1.88, "learning_rate": 4.703476423569053e-08, "loss": 0.1916, "step": 9951 }, { "epoch": 1.88, "learning_rate": 4.688720433661198e-08, "loss": 0.2818, "step": 9952 }, { "epoch": 1.88, "learning_rate": 4.6739874076476844e-08, "loss": 0.2766, "step": 9953 }, { "epoch": 1.88, "learning_rate": 4.659277346907659e-08, "loss": 0.2388, "step": 9954 }, { "epoch": 1.88, "learning_rate": 4.644590252818187e-08, "loss": 0.2258, "step": 9955 }, { "epoch": 1.88, "learning_rate": 4.629926126754114e-08, "loss": 0.1786, "step": 9956 }, { "epoch": 1.88, "learning_rate": 4.615284970088174e-08, "loss": 0.2357, "step": 9957 }, { "epoch": 1.88, "learning_rate": 4.600666784190966e-08, "loss": 0.1939, "step": 9958 }, { "epoch": 1.88, "learning_rate": 4.586071570430922e-08, "loss": 0.2464, "step": 9959 }, { "epoch": 1.88, "learning_rate": 4.571499330174256e-08, "loss": 0.2226, "step": 9960 }, { "epoch": 1.88, "learning_rate": 4.556950064785154e-08, "loss": 0.1838, "step": 9961 }, { "epoch": 1.88, "learning_rate": 4.542423775625554e-08, "loss": 0.2091, "step": 9962 }, { "epoch": 1.88, "learning_rate": 4.527920464055341e-08, "loss": 0.2581, "step": 9963 }, { "epoch": 1.88, "learning_rate": 4.513440131432123e-08, "loss": 0.2136, "step": 9964 }, { "epoch": 1.88, "learning_rate": 4.498982779111427e-08, "loss": 0.242, "step": 9965 }, { "epoch": 1.88, "learning_rate": 4.4845484084466706e-08, "loss": 0.1713, "step": 9966 }, { "epoch": 1.88, "learning_rate": 4.4701370207890224e-08, "loss": 0.206, "step": 9967 }, { "epoch": 1.88, "learning_rate": 4.4557486174875705e-08, "loss": 0.2152, "step": 9968 }, { "epoch": 1.88, "learning_rate": 4.441383199889265e-08, "loss": 0.1806, "step": 9969 }, { "epoch": 1.88, "learning_rate": 4.4270407693388085e-08, "loss": 0.214, "step": 9970 }, { "epoch": 1.88, "learning_rate": 4.4127213271788203e-08, "loss": 0.22, "step": 9971 }, { "epoch": 1.88, "learning_rate": 4.39842487474984e-08, "loss": 0.2622, "step": 9972 }, { "epoch": 1.88, "learning_rate": 4.3841514133900755e-08, "loss": 0.2074, "step": 9973 }, { "epoch": 1.88, "learning_rate": 4.369900944435734e-08, "loss": 0.2221, "step": 9974 }, { "epoch": 1.88, "learning_rate": 4.355673469220861e-08, "loss": 0.2143, "step": 9975 }, { "epoch": 1.88, "learning_rate": 4.341468989077197e-08, "loss": 0.2182, "step": 9976 }, { "epoch": 1.88, "learning_rate": 4.327287505334538e-08, "loss": 0.1992, "step": 9977 }, { "epoch": 1.89, "learning_rate": 4.313129019320406e-08, "loss": 0.2378, "step": 9978 }, { "epoch": 1.89, "learning_rate": 4.298993532360157e-08, "loss": 0.2406, "step": 9979 }, { "epoch": 1.89, "learning_rate": 4.284881045777067e-08, "loss": 0.2655, "step": 9980 }, { "epoch": 1.89, "learning_rate": 4.2707915608922454e-08, "loss": 0.2782, "step": 9981 }, { "epoch": 1.89, "learning_rate": 4.256725079024554e-08, "loss": 0.1867, "step": 9982 }, { "epoch": 1.89, "learning_rate": 4.2426816014908555e-08, "loss": 0.2422, "step": 9983 }, { "epoch": 1.89, "learning_rate": 4.22866112960571e-08, "loss": 0.2002, "step": 9984 }, { "epoch": 1.89, "learning_rate": 4.214663664681651e-08, "loss": 0.1488, "step": 9985 }, { "epoch": 1.89, "learning_rate": 4.200689208028935e-08, "loss": 0.1988, "step": 9986 }, { "epoch": 1.89, "learning_rate": 4.186737760955795e-08, "loss": 0.1791, "step": 9987 }, { "epoch": 1.89, "learning_rate": 4.1728093247681845e-08, "loss": 0.2961, "step": 9988 }, { "epoch": 1.89, "learning_rate": 4.158903900770006e-08, "loss": 0.2185, "step": 9989 }, { "epoch": 1.89, "learning_rate": 4.14502149026294e-08, "loss": 0.1976, "step": 9990 }, { "epoch": 1.89, "learning_rate": 4.1311620945465314e-08, "loss": 0.2271, "step": 9991 }, { "epoch": 1.89, "learning_rate": 4.117325714918186e-08, "loss": 0.2112, "step": 9992 }, { "epoch": 1.89, "learning_rate": 4.103512352673117e-08, "loss": 0.1816, "step": 9993 }, { "epoch": 1.89, "learning_rate": 4.089722009104513e-08, "loss": 0.2468, "step": 9994 }, { "epoch": 1.89, "learning_rate": 4.0759546855031754e-08, "loss": 0.3138, "step": 9995 }, { "epoch": 1.89, "learning_rate": 4.062210383157933e-08, "loss": 0.2409, "step": 9996 }, { "epoch": 1.89, "learning_rate": 4.048489103355424e-08, "loss": 0.2959, "step": 9997 }, { "epoch": 1.89, "learning_rate": 4.034790847380094e-08, "loss": 0.195, "step": 9998 }, { "epoch": 1.89, "learning_rate": 4.0211156165142485e-08, "loss": 0.2202, "step": 9999 }, { "epoch": 1.89, "learning_rate": 4.0074634120380874e-08, "loss": 0.2201, "step": 10000 }, { "epoch": 1.89, "learning_rate": 3.993834235229532e-08, "loss": 0.2149, "step": 10001 }, { "epoch": 1.89, "learning_rate": 3.9802280873644795e-08, "loss": 0.2493, "step": 10002 }, { "epoch": 1.89, "learning_rate": 3.96664496971666e-08, "loss": 0.1884, "step": 10003 }, { "epoch": 1.89, "learning_rate": 3.953084883557473e-08, "loss": 0.196, "step": 10004 }, { "epoch": 1.89, "learning_rate": 3.939547830156432e-08, "loss": 0.2053, "step": 10005 }, { "epoch": 1.89, "learning_rate": 3.926033810780716e-08, "loss": 0.252, "step": 10006 }, { "epoch": 1.89, "learning_rate": 3.9125428266953424e-08, "loss": 0.1611, "step": 10007 }, { "epoch": 1.89, "learning_rate": 3.899074879163245e-08, "loss": 0.265, "step": 10008 }, { "epoch": 1.89, "learning_rate": 3.8856299694452204e-08, "loss": 0.2239, "step": 10009 }, { "epoch": 1.89, "learning_rate": 3.8722080987997904e-08, "loss": 0.1952, "step": 10010 }, { "epoch": 1.89, "learning_rate": 3.8588092684834486e-08, "loss": 0.2298, "step": 10011 }, { "epoch": 1.89, "learning_rate": 3.845433479750471e-08, "loss": 0.1599, "step": 10012 }, { "epoch": 1.89, "learning_rate": 3.832080733852939e-08, "loss": 0.2462, "step": 10013 }, { "epoch": 1.89, "learning_rate": 3.8187510320408516e-08, "loss": 0.1988, "step": 10014 }, { "epoch": 1.89, "learning_rate": 3.8054443755620176e-08, "loss": 0.2488, "step": 10015 }, { "epoch": 1.89, "learning_rate": 3.792160765662106e-08, "loss": 0.2784, "step": 10016 }, { "epoch": 1.89, "learning_rate": 3.7789002035845676e-08, "loss": 0.2214, "step": 10017 }, { "epoch": 1.89, "learning_rate": 3.765662690570798e-08, "loss": 0.1413, "step": 10018 }, { "epoch": 1.89, "learning_rate": 3.7524482278599176e-08, "loss": 0.2158, "step": 10019 }, { "epoch": 1.89, "learning_rate": 3.7392568166889906e-08, "loss": 0.2186, "step": 10020 }, { "epoch": 1.89, "learning_rate": 3.726088458292837e-08, "loss": 0.2436, "step": 10021 }, { "epoch": 1.89, "learning_rate": 3.7129431539042195e-08, "loss": 0.2891, "step": 10022 }, { "epoch": 1.89, "learning_rate": 3.6998209047536536e-08, "loss": 0.2164, "step": 10023 }, { "epoch": 1.89, "learning_rate": 3.6867217120695467e-08, "loss": 0.2467, "step": 10024 }, { "epoch": 1.89, "learning_rate": 3.673645577078111e-08, "loss": 0.2051, "step": 10025 }, { "epoch": 1.89, "learning_rate": 3.660592501003424e-08, "loss": 0.1784, "step": 10026 }, { "epoch": 1.89, "learning_rate": 3.6475624850674244e-08, "loss": 0.1793, "step": 10027 }, { "epoch": 1.89, "learning_rate": 3.6345555304898304e-08, "loss": 0.2038, "step": 10028 }, { "epoch": 1.89, "learning_rate": 3.621571638488253e-08, "loss": 0.2453, "step": 10029 }, { "epoch": 1.89, "learning_rate": 3.6086108102781345e-08, "loss": 0.148, "step": 10030 }, { "epoch": 1.9, "learning_rate": 3.5956730470727564e-08, "loss": 0.2399, "step": 10031 }, { "epoch": 1.9, "learning_rate": 3.58275835008326e-08, "loss": 0.2164, "step": 10032 }, { "epoch": 1.9, "learning_rate": 3.569866720518567e-08, "loss": 0.2654, "step": 10033 }, { "epoch": 1.9, "learning_rate": 3.55699815958549e-08, "loss": 0.2238, "step": 10034 }, { "epoch": 1.9, "learning_rate": 3.544152668488676e-08, "loss": 0.2701, "step": 10035 }, { "epoch": 1.9, "learning_rate": 3.5313302484306086e-08, "loss": 0.2253, "step": 10036 }, { "epoch": 1.9, "learning_rate": 3.518530900611633e-08, "loss": 0.2451, "step": 10037 }, { "epoch": 1.9, "learning_rate": 3.505754626229846e-08, "loss": 0.1853, "step": 10038 }, { "epoch": 1.9, "learning_rate": 3.4930014264813185e-08, "loss": 0.2036, "step": 10039 }, { "epoch": 1.9, "learning_rate": 3.4802713025598734e-08, "loss": 0.2545, "step": 10040 }, { "epoch": 1.9, "learning_rate": 3.467564255657169e-08, "loss": 0.3126, "step": 10041 }, { "epoch": 1.9, "learning_rate": 3.4548802869627806e-08, "loss": 0.2401, "step": 10042 }, { "epoch": 1.9, "learning_rate": 3.442219397664009e-08, "loss": 0.2782, "step": 10043 }, { "epoch": 1.9, "learning_rate": 3.429581588946074e-08, "loss": 0.2327, "step": 10044 }, { "epoch": 1.9, "learning_rate": 3.416966861992055e-08, "loss": 0.2483, "step": 10045 }, { "epoch": 1.9, "learning_rate": 3.404375217982786e-08, "loss": 0.2043, "step": 10046 }, { "epoch": 1.9, "learning_rate": 3.3918066580969914e-08, "loss": 0.2178, "step": 10047 }, { "epoch": 1.9, "learning_rate": 3.379261183511284e-08, "loss": 0.1967, "step": 10048 }, { "epoch": 1.9, "learning_rate": 3.366738795399976e-08, "loss": 0.2139, "step": 10049 }, { "epoch": 1.9, "learning_rate": 3.354239494935352e-08, "loss": 0.2166, "step": 10050 }, { "epoch": 1.9, "learning_rate": 3.3417632832875305e-08, "loss": 0.2128, "step": 10051 }, { "epoch": 1.9, "learning_rate": 3.3293101616242993e-08, "loss": 0.2501, "step": 10052 }, { "epoch": 1.9, "learning_rate": 3.316880131111533e-08, "loss": 0.2274, "step": 10053 }, { "epoch": 1.9, "learning_rate": 3.304473192912772e-08, "loss": 0.2415, "step": 10054 }, { "epoch": 1.9, "learning_rate": 3.2920893481894755e-08, "loss": 0.1603, "step": 10055 }, { "epoch": 1.9, "learning_rate": 3.2797285981008566e-08, "loss": 0.24, "step": 10056 }, { "epoch": 1.9, "learning_rate": 3.267390943804044e-08, "loss": 0.2387, "step": 10057 }, { "epoch": 1.9, "learning_rate": 3.255076386454004e-08, "loss": 0.2134, "step": 10058 }, { "epoch": 1.9, "learning_rate": 3.242784927203507e-08, "loss": 0.2201, "step": 10059 }, { "epoch": 1.9, "learning_rate": 3.230516567203135e-08, "loss": 0.2414, "step": 10060 }, { "epoch": 1.9, "learning_rate": 3.2182713076014115e-08, "loss": 0.1978, "step": 10061 }, { "epoch": 1.9, "learning_rate": 3.20604914954456e-08, "loss": 0.2896, "step": 10062 }, { "epoch": 1.9, "learning_rate": 3.193850094176776e-08, "loss": 0.1572, "step": 10063 }, { "epoch": 1.9, "learning_rate": 3.181674142639979e-08, "loss": 0.2498, "step": 10064 }, { "epoch": 1.9, "learning_rate": 3.1695212960740075e-08, "loss": 0.1576, "step": 10065 }, { "epoch": 1.9, "learning_rate": 3.157391555616451e-08, "loss": 0.2918, "step": 10066 }, { "epoch": 1.9, "learning_rate": 3.1452849224028745e-08, "loss": 0.2453, "step": 10067 }, { "epoch": 1.9, "learning_rate": 3.1332013975665375e-08, "loss": 0.2615, "step": 10068 }, { "epoch": 1.9, "learning_rate": 3.1211409822385905e-08, "loss": 0.3147, "step": 10069 }, { "epoch": 1.9, "learning_rate": 3.1091036775480754e-08, "loss": 0.2191, "step": 10070 }, { "epoch": 1.9, "learning_rate": 3.097089484621757e-08, "loss": 0.2006, "step": 10071 }, { "epoch": 1.9, "learning_rate": 3.0850984045842916e-08, "loss": 0.2199, "step": 10072 }, { "epoch": 1.9, "learning_rate": 3.073130438558281e-08, "loss": 0.2579, "step": 10073 }, { "epoch": 1.9, "learning_rate": 3.0611855876639674e-08, "loss": 0.2209, "step": 10074 }, { "epoch": 1.9, "learning_rate": 3.049263853019513e-08, "loss": 0.2573, "step": 10075 }, { "epoch": 1.9, "learning_rate": 3.0373652357410243e-08, "loss": 0.2831, "step": 10076 }, { "epoch": 1.9, "learning_rate": 3.025489736942222e-08, "loss": 0.2103, "step": 10077 }, { "epoch": 1.9, "learning_rate": 3.013637357734883e-08, "loss": 0.2435, "step": 10078 }, { "epoch": 1.9, "learning_rate": 3.001808099228509e-08, "loss": 0.2068, "step": 10079 }, { "epoch": 1.9, "learning_rate": 2.990001962530381e-08, "loss": 0.2575, "step": 10080 }, { "epoch": 1.9, "learning_rate": 2.9782189487457812e-08, "loss": 0.2129, "step": 10081 }, { "epoch": 1.9, "learning_rate": 2.9664590589776888e-08, "loss": 0.2681, "step": 10082 }, { "epoch": 1.9, "learning_rate": 2.9547222943269172e-08, "loss": 0.2189, "step": 10083 }, { "epoch": 1.91, "learning_rate": 2.9430086558922268e-08, "loss": 0.2094, "step": 10084 }, { "epoch": 1.91, "learning_rate": 2.931318144770129e-08, "loss": 0.2231, "step": 10085 }, { "epoch": 1.91, "learning_rate": 2.919650762054943e-08, "loss": 0.2102, "step": 10086 }, { "epoch": 1.91, "learning_rate": 2.908006508838934e-08, "loss": 0.271, "step": 10087 }, { "epoch": 1.91, "learning_rate": 2.8963853862121195e-08, "loss": 0.2421, "step": 10088 }, { "epoch": 1.91, "learning_rate": 2.8847873952623228e-08, "loss": 0.2521, "step": 10089 }, { "epoch": 1.91, "learning_rate": 2.873212537075287e-08, "loss": 0.1798, "step": 10090 }, { "epoch": 1.91, "learning_rate": 2.8616608127345347e-08, "loss": 0.1955, "step": 10091 }, { "epoch": 1.91, "learning_rate": 2.8501322233214503e-08, "loss": 0.2847, "step": 10092 }, { "epoch": 1.91, "learning_rate": 2.838626769915226e-08, "loss": 0.2827, "step": 10093 }, { "epoch": 1.91, "learning_rate": 2.8271444535929172e-08, "loss": 0.2404, "step": 10094 }, { "epoch": 1.91, "learning_rate": 2.8156852754293584e-08, "loss": 0.1968, "step": 10095 }, { "epoch": 1.91, "learning_rate": 2.804249236497303e-08, "loss": 0.2761, "step": 10096 }, { "epoch": 1.91, "learning_rate": 2.792836337867255e-08, "loss": 0.2614, "step": 10097 }, { "epoch": 1.91, "learning_rate": 2.7814465806076386e-08, "loss": 0.2114, "step": 10098 }, { "epoch": 1.91, "learning_rate": 2.770079965784628e-08, "loss": 0.2081, "step": 10099 }, { "epoch": 1.91, "learning_rate": 2.758736494462261e-08, "loss": 0.2288, "step": 10100 }, { "epoch": 1.91, "learning_rate": 2.7474161677024104e-08, "loss": 0.2326, "step": 10101 }, { "epoch": 1.91, "learning_rate": 2.7361189865648395e-08, "loss": 0.2342, "step": 10102 }, { "epoch": 1.91, "learning_rate": 2.7248449521070364e-08, "loss": 0.1904, "step": 10103 }, { "epoch": 1.91, "learning_rate": 2.7135940653843783e-08, "loss": 0.2326, "step": 10104 }, { "epoch": 1.91, "learning_rate": 2.702366327450079e-08, "loss": 0.1703, "step": 10105 }, { "epoch": 1.91, "learning_rate": 2.6911617393552415e-08, "loss": 0.2259, "step": 10106 }, { "epoch": 1.91, "learning_rate": 2.679980302148666e-08, "loss": 0.2245, "step": 10107 }, { "epoch": 1.91, "learning_rate": 2.6688220168770985e-08, "loss": 0.1796, "step": 10108 }, { "epoch": 1.91, "learning_rate": 2.6576868845850646e-08, "loss": 0.1671, "step": 10109 }, { "epoch": 1.91, "learning_rate": 2.6465749063149247e-08, "loss": 0.1684, "step": 10110 }, { "epoch": 1.91, "learning_rate": 2.6354860831069028e-08, "loss": 0.3, "step": 10111 }, { "epoch": 1.91, "learning_rate": 2.624420415999085e-08, "loss": 0.2773, "step": 10112 }, { "epoch": 1.91, "learning_rate": 2.6133779060272535e-08, "loss": 0.3064, "step": 10113 }, { "epoch": 1.91, "learning_rate": 2.6023585542251374e-08, "loss": 0.2241, "step": 10114 }, { "epoch": 1.91, "learning_rate": 2.5913623616243278e-08, "loss": 0.2265, "step": 10115 }, { "epoch": 1.91, "learning_rate": 2.5803893292541125e-08, "loss": 0.1489, "step": 10116 }, { "epoch": 1.91, "learning_rate": 2.5694394581417538e-08, "loss": 0.2287, "step": 10117 }, { "epoch": 1.91, "learning_rate": 2.5585127493122364e-08, "loss": 0.2348, "step": 10118 }, { "epoch": 1.91, "learning_rate": 2.547609203788465e-08, "loss": 0.2336, "step": 10119 }, { "epoch": 1.91, "learning_rate": 2.5367288225911225e-08, "loss": 0.1492, "step": 10120 }, { "epoch": 1.91, "learning_rate": 2.5258716067387e-08, "loss": 0.2566, "step": 10121 }, { "epoch": 1.91, "learning_rate": 2.5150375572476072e-08, "loss": 0.221, "step": 10122 }, { "epoch": 1.91, "learning_rate": 2.5042266751319765e-08, "loss": 0.1528, "step": 10123 }, { "epoch": 1.91, "learning_rate": 2.4934389614038878e-08, "loss": 0.2403, "step": 10124 }, { "epoch": 1.91, "learning_rate": 2.4826744170731444e-08, "loss": 0.2119, "step": 10125 }, { "epoch": 1.91, "learning_rate": 2.4719330431474398e-08, "loss": 0.2523, "step": 10126 }, { "epoch": 1.91, "learning_rate": 2.461214840632331e-08, "loss": 0.2194, "step": 10127 }, { "epoch": 1.91, "learning_rate": 2.450519810531099e-08, "loss": 0.22, "step": 10128 }, { "epoch": 1.91, "learning_rate": 2.4398479538449426e-08, "loss": 0.2337, "step": 10129 }, { "epoch": 1.91, "learning_rate": 2.4291992715728687e-08, "loss": 0.2486, "step": 10130 }, { "epoch": 1.91, "learning_rate": 2.4185737647117456e-08, "loss": 0.1791, "step": 10131 }, { "epoch": 1.91, "learning_rate": 2.4079714342561954e-08, "loss": 0.2456, "step": 10132 }, { "epoch": 1.91, "learning_rate": 2.3973922811987295e-08, "loss": 0.191, "step": 10133 }, { "epoch": 1.91, "learning_rate": 2.3868363065296664e-08, "loss": 0.2036, "step": 10134 }, { "epoch": 1.91, "learning_rate": 2.376303511237188e-08, "loss": 0.1667, "step": 10135 }, { "epoch": 1.91, "learning_rate": 2.365793896307228e-08, "loss": 0.2392, "step": 10136 }, { "epoch": 1.92, "learning_rate": 2.3553074627237214e-08, "loss": 0.2749, "step": 10137 }, { "epoch": 1.92, "learning_rate": 2.344844211468189e-08, "loss": 0.2046, "step": 10138 }, { "epoch": 1.92, "learning_rate": 2.3344041435201525e-08, "loss": 0.1743, "step": 10139 }, { "epoch": 1.92, "learning_rate": 2.323987259856969e-08, "loss": 0.2589, "step": 10140 }, { "epoch": 1.92, "learning_rate": 2.31359356145372e-08, "loss": 0.1585, "step": 10141 }, { "epoch": 1.92, "learning_rate": 2.3032230492833486e-08, "loss": 0.2142, "step": 10142 }, { "epoch": 1.92, "learning_rate": 2.2928757243167454e-08, "loss": 0.2385, "step": 10143 }, { "epoch": 1.92, "learning_rate": 2.2825515875224414e-08, "loss": 0.1616, "step": 10144 }, { "epoch": 1.92, "learning_rate": 2.272250639866941e-08, "loss": 0.1775, "step": 10145 }, { "epoch": 1.92, "learning_rate": 2.261972882314528e-08, "loss": 0.2634, "step": 10146 }, { "epoch": 1.92, "learning_rate": 2.2517183158272948e-08, "loss": 0.1335, "step": 10147 }, { "epoch": 1.92, "learning_rate": 2.241486941365195e-08, "loss": 0.1854, "step": 10148 }, { "epoch": 1.92, "learning_rate": 2.2312787598860186e-08, "loss": 0.2087, "step": 10149 }, { "epoch": 1.92, "learning_rate": 2.221093772345334e-08, "loss": 0.1668, "step": 10150 }, { "epoch": 1.92, "learning_rate": 2.210931979696601e-08, "loss": 0.1905, "step": 10151 }, { "epoch": 1.92, "learning_rate": 2.2007933828910587e-08, "loss": 0.2964, "step": 10152 }, { "epoch": 1.92, "learning_rate": 2.1906779828777814e-08, "loss": 0.2134, "step": 10153 }, { "epoch": 1.92, "learning_rate": 2.1805857806037056e-08, "loss": 0.3938, "step": 10154 }, { "epoch": 1.92, "learning_rate": 2.170516777013576e-08, "loss": 0.2027, "step": 10155 }, { "epoch": 1.92, "learning_rate": 2.160470973049944e-08, "loss": 0.2437, "step": 10156 }, { "epoch": 1.92, "learning_rate": 2.1504483696532517e-08, "loss": 0.1927, "step": 10157 }, { "epoch": 1.92, "learning_rate": 2.1404489677616936e-08, "loss": 0.1686, "step": 10158 }, { "epoch": 1.92, "learning_rate": 2.130472768311298e-08, "loss": 0.2596, "step": 10159 }, { "epoch": 1.92, "learning_rate": 2.1205197722360127e-08, "loss": 0.2265, "step": 10160 }, { "epoch": 1.92, "learning_rate": 2.110589980467537e-08, "loss": 0.2763, "step": 10161 }, { "epoch": 1.92, "learning_rate": 2.1006833939353765e-08, "loss": 0.195, "step": 10162 }, { "epoch": 1.92, "learning_rate": 2.0908000135669017e-08, "loss": 0.2534, "step": 10163 }, { "epoch": 1.92, "learning_rate": 2.0809398402873436e-08, "loss": 0.26, "step": 10164 }, { "epoch": 1.92, "learning_rate": 2.071102875019687e-08, "loss": 0.2181, "step": 10165 }, { "epoch": 1.92, "learning_rate": 2.0612891186848337e-08, "loss": 0.3022, "step": 10166 }, { "epoch": 1.92, "learning_rate": 2.05149857220141e-08, "loss": 0.2273, "step": 10167 }, { "epoch": 1.92, "learning_rate": 2.0417312364859333e-08, "loss": 0.2112, "step": 10168 }, { "epoch": 1.92, "learning_rate": 2.0319871124527546e-08, "loss": 0.1898, "step": 10169 }, { "epoch": 1.92, "learning_rate": 2.0222662010140338e-08, "loss": 0.2422, "step": 10170 }, { "epoch": 1.92, "learning_rate": 2.0125685030797092e-08, "loss": 0.1697, "step": 10171 }, { "epoch": 1.92, "learning_rate": 2.0028940195576663e-08, "loss": 0.1685, "step": 10172 }, { "epoch": 1.92, "learning_rate": 1.9932427513534857e-08, "loss": 0.2131, "step": 10173 }, { "epoch": 1.92, "learning_rate": 1.9836146993706675e-08, "loss": 0.2225, "step": 10174 }, { "epoch": 1.92, "learning_rate": 1.9740098645105178e-08, "loss": 0.2513, "step": 10175 }, { "epoch": 1.92, "learning_rate": 1.964428247672123e-08, "loss": 0.2075, "step": 10176 }, { "epoch": 1.92, "learning_rate": 1.954869849752433e-08, "loss": 0.2473, "step": 10177 }, { "epoch": 1.92, "learning_rate": 1.945334671646232e-08, "loss": 0.1711, "step": 10178 }, { "epoch": 1.92, "learning_rate": 1.9358227142461383e-08, "loss": 0.1682, "step": 10179 }, { "epoch": 1.92, "learning_rate": 1.9263339784425517e-08, "loss": 0.1228, "step": 10180 }, { "epoch": 1.92, "learning_rate": 1.9168684651237056e-08, "loss": 0.2362, "step": 10181 }, { "epoch": 1.92, "learning_rate": 1.9074261751757252e-08, "loss": 0.2036, "step": 10182 }, { "epoch": 1.92, "learning_rate": 1.8980071094825138e-08, "loss": 0.2154, "step": 10183 }, { "epoch": 1.92, "learning_rate": 1.8886112689257553e-08, "loss": 0.2364, "step": 10184 }, { "epoch": 1.92, "learning_rate": 1.8792386543850515e-08, "loss": 0.2031, "step": 10185 }, { "epoch": 1.92, "learning_rate": 1.8698892667377844e-08, "loss": 0.295, "step": 10186 }, { "epoch": 1.92, "learning_rate": 1.8605631068591146e-08, "loss": 0.2355, "step": 10187 }, { "epoch": 1.92, "learning_rate": 1.8512601756221215e-08, "loss": 0.211, "step": 10188 }, { "epoch": 1.92, "learning_rate": 1.8419804738976364e-08, "loss": 0.2248, "step": 10189 }, { "epoch": 1.93, "learning_rate": 1.8327240025543536e-08, "loss": 0.2378, "step": 10190 }, { "epoch": 1.93, "learning_rate": 1.8234907624588018e-08, "loss": 0.1233, "step": 10191 }, { "epoch": 1.93, "learning_rate": 1.8142807544752895e-08, "loss": 0.2503, "step": 10192 }, { "epoch": 1.93, "learning_rate": 1.8050939794659882e-08, "loss": 0.1907, "step": 10193 }, { "epoch": 1.93, "learning_rate": 1.795930438290877e-08, "loss": 0.2473, "step": 10194 }, { "epoch": 1.93, "learning_rate": 1.7867901318077696e-08, "loss": 0.1574, "step": 10195 }, { "epoch": 1.93, "learning_rate": 1.7776730608723147e-08, "loss": 0.283, "step": 10196 }, { "epoch": 1.93, "learning_rate": 1.768579226337941e-08, "loss": 0.1839, "step": 10197 }, { "epoch": 1.93, "learning_rate": 1.759508629055967e-08, "loss": 0.3371, "step": 10198 }, { "epoch": 1.93, "learning_rate": 1.750461269875492e-08, "loss": 0.2054, "step": 10199 }, { "epoch": 1.93, "learning_rate": 1.741437149643449e-08, "loss": 0.2277, "step": 10200 }, { "epoch": 1.93, "learning_rate": 1.7324362692045793e-08, "loss": 0.209, "step": 10201 }, { "epoch": 1.93, "learning_rate": 1.7234586294015144e-08, "loss": 0.2007, "step": 10202 }, { "epoch": 1.93, "learning_rate": 1.7145042310746096e-08, "loss": 0.2166, "step": 10203 }, { "epoch": 1.93, "learning_rate": 1.7055730750621113e-08, "loss": 0.2413, "step": 10204 }, { "epoch": 1.93, "learning_rate": 1.6966651622001006e-08, "loss": 0.1554, "step": 10205 }, { "epoch": 1.93, "learning_rate": 1.6877804933224384e-08, "loss": 0.1983, "step": 10206 }, { "epoch": 1.93, "learning_rate": 1.6789190692608203e-08, "loss": 0.2592, "step": 10207 }, { "epoch": 1.93, "learning_rate": 1.670080890844805e-08, "loss": 0.2417, "step": 10208 }, { "epoch": 1.93, "learning_rate": 1.6612659589017032e-08, "loss": 0.2089, "step": 10209 }, { "epoch": 1.93, "learning_rate": 1.6524742742567435e-08, "loss": 0.3194, "step": 10210 }, { "epoch": 1.93, "learning_rate": 1.6437058377329063e-08, "loss": 0.206, "step": 10211 }, { "epoch": 1.93, "learning_rate": 1.6349606501509796e-08, "loss": 0.1415, "step": 10212 }, { "epoch": 1.93, "learning_rate": 1.626238712329642e-08, "loss": 0.1832, "step": 10213 }, { "epoch": 1.93, "learning_rate": 1.617540025085379e-08, "loss": 0.2783, "step": 10214 }, { "epoch": 1.93, "learning_rate": 1.6088645892324562e-08, "loss": 0.2317, "step": 10215 }, { "epoch": 1.93, "learning_rate": 1.6002124055830294e-08, "loss": 0.277, "step": 10216 }, { "epoch": 1.93, "learning_rate": 1.591583474947006e-08, "loss": 0.2807, "step": 10217 }, { "epoch": 1.93, "learning_rate": 1.582977798132129e-08, "loss": 0.2616, "step": 10218 }, { "epoch": 1.93, "learning_rate": 1.57439537594406e-08, "loss": 0.2514, "step": 10219 }, { "epoch": 1.93, "learning_rate": 1.5658362091861557e-08, "loss": 0.2391, "step": 10220 }, { "epoch": 1.93, "learning_rate": 1.557300298659664e-08, "loss": 0.2712, "step": 10221 }, { "epoch": 1.93, "learning_rate": 1.548787645163641e-08, "loss": 0.2811, "step": 10222 }, { "epoch": 1.93, "learning_rate": 1.540298249494976e-08, "loss": 0.2651, "step": 10223 }, { "epoch": 1.93, "learning_rate": 1.531832112448339e-08, "loss": 0.2149, "step": 10224 }, { "epoch": 1.93, "learning_rate": 1.5233892348163193e-08, "loss": 0.1917, "step": 10225 }, { "epoch": 1.93, "learning_rate": 1.5149696173892002e-08, "loss": 0.2353, "step": 10226 }, { "epoch": 1.93, "learning_rate": 1.5065732609551853e-08, "loss": 0.2448, "step": 10227 }, { "epoch": 1.93, "learning_rate": 1.4982001663002842e-08, "loss": 0.2352, "step": 10228 }, { "epoch": 1.93, "learning_rate": 1.4898503342082592e-08, "loss": 0.1875, "step": 10229 }, { "epoch": 1.93, "learning_rate": 1.4815237654607629e-08, "loss": 0.1942, "step": 10230 }, { "epoch": 1.93, "learning_rate": 1.4732204608373103e-08, "loss": 0.1859, "step": 10231 }, { "epoch": 1.93, "learning_rate": 1.4649404211151408e-08, "loss": 0.2497, "step": 10232 }, { "epoch": 1.93, "learning_rate": 1.4566836470693568e-08, "loss": 0.2398, "step": 10233 }, { "epoch": 1.93, "learning_rate": 1.4484501394728678e-08, "loss": 0.2337, "step": 10234 }, { "epoch": 1.93, "learning_rate": 1.440239899096474e-08, "loss": 0.2251, "step": 10235 }, { "epoch": 1.93, "learning_rate": 1.432052926708727e-08, "loss": 0.1556, "step": 10236 }, { "epoch": 1.93, "learning_rate": 1.4238892230760137e-08, "loss": 0.2774, "step": 10237 }, { "epoch": 1.93, "learning_rate": 1.4157487889625288e-08, "loss": 0.2129, "step": 10238 }, { "epoch": 1.93, "learning_rate": 1.407631625130329e-08, "loss": 0.2327, "step": 10239 }, { "epoch": 1.93, "learning_rate": 1.3995377323392789e-08, "loss": 0.3029, "step": 10240 }, { "epoch": 1.93, "learning_rate": 1.3914671113470501e-08, "loss": 0.2319, "step": 10241 }, { "epoch": 1.94, "learning_rate": 1.3834197629091218e-08, "loss": 0.2892, "step": 10242 }, { "epoch": 1.94, "learning_rate": 1.3753956877788354e-08, "loss": 0.193, "step": 10243 }, { "epoch": 1.94, "learning_rate": 1.3673948867073683e-08, "loss": 0.3041, "step": 10244 }, { "epoch": 1.94, "learning_rate": 1.3594173604436489e-08, "loss": 0.2832, "step": 10245 }, { "epoch": 1.94, "learning_rate": 1.351463109734441e-08, "loss": 0.2536, "step": 10246 }, { "epoch": 1.94, "learning_rate": 1.343532135324399e-08, "loss": 0.2202, "step": 10247 }, { "epoch": 1.94, "learning_rate": 1.3356244379559291e-08, "loss": 0.2058, "step": 10248 }, { "epoch": 1.94, "learning_rate": 1.3277400183692724e-08, "loss": 0.1964, "step": 10249 }, { "epoch": 1.94, "learning_rate": 1.3198788773025328e-08, "loss": 0.1996, "step": 10250 }, { "epoch": 1.94, "learning_rate": 1.3120410154915385e-08, "loss": 0.2521, "step": 10251 }, { "epoch": 1.94, "learning_rate": 1.3042264336700639e-08, "loss": 0.203, "step": 10252 }, { "epoch": 1.94, "learning_rate": 1.2964351325696345e-08, "loss": 0.2307, "step": 10253 }, { "epoch": 1.94, "learning_rate": 1.288667112919556e-08, "loss": 0.2682, "step": 10254 }, { "epoch": 1.94, "learning_rate": 1.2809223754470523e-08, "loss": 0.2251, "step": 10255 }, { "epoch": 1.94, "learning_rate": 1.273200920877099e-08, "loss": 0.1811, "step": 10256 }, { "epoch": 1.94, "learning_rate": 1.2655027499325067e-08, "loss": 0.2406, "step": 10257 }, { "epoch": 1.94, "learning_rate": 1.2578278633339213e-08, "loss": 0.2858, "step": 10258 }, { "epoch": 1.94, "learning_rate": 1.2501762617997958e-08, "loss": 0.2441, "step": 10259 }, { "epoch": 1.94, "learning_rate": 1.2425479460463908e-08, "loss": 0.1624, "step": 10260 }, { "epoch": 1.94, "learning_rate": 1.234942916787829e-08, "loss": 0.189, "step": 10261 }, { "epoch": 1.94, "learning_rate": 1.2273611747360137e-08, "loss": 0.1823, "step": 10262 }, { "epoch": 1.94, "learning_rate": 1.2198027206006823e-08, "loss": 0.234, "step": 10263 }, { "epoch": 1.94, "learning_rate": 1.2122675550893804e-08, "loss": 0.1794, "step": 10264 }, { "epoch": 1.94, "learning_rate": 1.2047556789074877e-08, "loss": 0.2611, "step": 10265 }, { "epoch": 1.94, "learning_rate": 1.1972670927582197e-08, "loss": 0.211, "step": 10266 }, { "epoch": 1.94, "learning_rate": 1.1898017973425714e-08, "loss": 0.2148, "step": 10267 }, { "epoch": 1.94, "learning_rate": 1.1823597933594e-08, "loss": 0.1583, "step": 10268 }, { "epoch": 1.94, "learning_rate": 1.1749410815053152e-08, "loss": 0.2806, "step": 10269 }, { "epoch": 1.94, "learning_rate": 1.1675456624748448e-08, "loss": 0.1893, "step": 10270 }, { "epoch": 1.94, "learning_rate": 1.1601735369602684e-08, "loss": 0.3057, "step": 10271 }, { "epoch": 1.94, "learning_rate": 1.1528247056516728e-08, "loss": 0.1912, "step": 10272 }, { "epoch": 1.94, "learning_rate": 1.1454991692370076e-08, "loss": 0.1827, "step": 10273 }, { "epoch": 1.94, "learning_rate": 1.1381969284020577e-08, "loss": 0.2639, "step": 10274 }, { "epoch": 1.94, "learning_rate": 1.130917983830332e-08, "loss": 0.2156, "step": 10275 }, { "epoch": 1.94, "learning_rate": 1.123662336203285e-08, "loss": 0.2164, "step": 10276 }, { "epoch": 1.94, "learning_rate": 1.1164299862000682e-08, "loss": 0.206, "step": 10277 }, { "epoch": 1.94, "learning_rate": 1.1092209344977512e-08, "loss": 0.1704, "step": 10278 }, { "epoch": 1.94, "learning_rate": 1.1020351817711828e-08, "loss": 0.2612, "step": 10279 }, { "epoch": 1.94, "learning_rate": 1.0948727286930194e-08, "loss": 0.2385, "step": 10280 }, { "epoch": 1.94, "learning_rate": 1.0877335759337249e-08, "loss": 0.1768, "step": 10281 }, { "epoch": 1.94, "learning_rate": 1.0806177241616255e-08, "loss": 0.2535, "step": 10282 }, { "epoch": 1.94, "learning_rate": 1.0735251740428832e-08, "loss": 0.2063, "step": 10283 }, { "epoch": 1.94, "learning_rate": 1.0664559262413831e-08, "loss": 0.2278, "step": 10284 }, { "epoch": 1.94, "learning_rate": 1.059409981418874e-08, "loss": 0.2269, "step": 10285 }, { "epoch": 1.94, "learning_rate": 1.0523873402350227e-08, "loss": 0.3072, "step": 10286 }, { "epoch": 1.94, "learning_rate": 1.0453880033471364e-08, "loss": 0.1627, "step": 10287 }, { "epoch": 1.94, "learning_rate": 1.038411971410469e-08, "loss": 0.2246, "step": 10288 }, { "epoch": 1.94, "learning_rate": 1.0314592450780536e-08, "loss": 0.2505, "step": 10289 }, { "epoch": 1.94, "learning_rate": 1.0245298250007585e-08, "loss": 0.2106, "step": 10290 }, { "epoch": 1.94, "learning_rate": 1.0176237118272315e-08, "loss": 0.1757, "step": 10291 }, { "epoch": 1.94, "learning_rate": 1.010740906203983e-08, "loss": 0.1531, "step": 10292 }, { "epoch": 1.94, "learning_rate": 1.0038814087752757e-08, "loss": 0.1733, "step": 10293 }, { "epoch": 1.94, "learning_rate": 9.970452201832903e-09, "loss": 0.212, "step": 10294 }, { "epoch": 1.95, "learning_rate": 9.90232341067987e-09, "loss": 0.2037, "step": 10295 }, { "epoch": 1.95, "learning_rate": 9.834427720670504e-09, "loss": 0.2545, "step": 10296 }, { "epoch": 1.95, "learning_rate": 9.766765138160828e-09, "loss": 0.2328, "step": 10297 }, { "epoch": 1.95, "learning_rate": 9.699335669485498e-09, "loss": 0.2415, "step": 10298 }, { "epoch": 1.95, "learning_rate": 9.632139320955858e-09, "loss": 0.278, "step": 10299 }, { "epoch": 1.95, "learning_rate": 9.565176098862706e-09, "loss": 0.203, "step": 10300 }, { "epoch": 1.95, "learning_rate": 9.49844600947436e-09, "loss": 0.1888, "step": 10301 }, { "epoch": 1.95, "learning_rate": 9.431949059037493e-09, "loss": 0.2112, "step": 10302 }, { "epoch": 1.95, "learning_rate": 9.365685253777124e-09, "loss": 0.2385, "step": 10303 }, { "epoch": 1.95, "learning_rate": 9.29965459989607e-09, "loss": 0.2422, "step": 10304 }, { "epoch": 1.95, "learning_rate": 9.233857103575772e-09, "loss": 0.1911, "step": 10305 }, { "epoch": 1.95, "learning_rate": 9.168292770975474e-09, "loss": 0.2307, "step": 10306 }, { "epoch": 1.95, "learning_rate": 9.102961608232763e-09, "loss": 0.2544, "step": 10307 }, { "epoch": 1.95, "learning_rate": 9.037863621463306e-09, "loss": 0.244, "step": 10308 }, { "epoch": 1.95, "learning_rate": 8.97299881676139e-09, "loss": 0.2306, "step": 10309 }, { "epoch": 1.95, "learning_rate": 8.908367200198276e-09, "loss": 0.2314, "step": 10310 }, { "epoch": 1.95, "learning_rate": 8.84396877782523e-09, "loss": 0.1679, "step": 10311 }, { "epoch": 1.95, "learning_rate": 8.77980355566993e-09, "loss": 0.2436, "step": 10312 }, { "epoch": 1.95, "learning_rate": 8.715871539738963e-09, "loss": 0.2682, "step": 10313 }, { "epoch": 1.95, "learning_rate": 8.652172736017817e-09, "loss": 0.2406, "step": 10314 }, { "epoch": 1.95, "learning_rate": 8.588707150468667e-09, "loss": 0.2302, "step": 10315 }, { "epoch": 1.95, "learning_rate": 8.525474789032595e-09, "loss": 0.1774, "step": 10316 }, { "epoch": 1.95, "learning_rate": 8.46247565762931e-09, "loss": 0.2795, "step": 10317 }, { "epoch": 1.95, "learning_rate": 8.399709762156038e-09, "loss": 0.2157, "step": 10318 }, { "epoch": 1.95, "learning_rate": 8.33717710848836e-09, "loss": 0.1932, "step": 10319 }, { "epoch": 1.95, "learning_rate": 8.274877702480199e-09, "loss": 0.2549, "step": 10320 }, { "epoch": 1.95, "learning_rate": 8.212811549963006e-09, "loss": 0.2003, "step": 10321 }, { "epoch": 1.95, "learning_rate": 8.15097865674741e-09, "loss": 0.2161, "step": 10322 }, { "epoch": 1.95, "learning_rate": 8.089379028621558e-09, "loss": 0.265, "step": 10323 }, { "epoch": 1.95, "learning_rate": 8.02801267135167e-09, "loss": 0.2205, "step": 10324 }, { "epoch": 1.95, "learning_rate": 7.966879590682319e-09, "loss": 0.2333, "step": 10325 }, { "epoch": 1.95, "learning_rate": 7.905979792336704e-09, "loss": 0.2382, "step": 10326 }, { "epoch": 1.95, "learning_rate": 7.845313282014988e-09, "loss": 0.2552, "step": 10327 }, { "epoch": 1.95, "learning_rate": 7.784880065396517e-09, "loss": 0.1741, "step": 10328 }, { "epoch": 1.95, "learning_rate": 7.724680148138985e-09, "loss": 0.1775, "step": 10329 }, { "epoch": 1.95, "learning_rate": 7.664713535877332e-09, "loss": 0.1476, "step": 10330 }, { "epoch": 1.95, "learning_rate": 7.604980234225124e-09, "loss": 0.221, "step": 10331 }, { "epoch": 1.95, "learning_rate": 7.545480248774273e-09, "loss": 0.1913, "step": 10332 }, { "epoch": 1.95, "learning_rate": 7.486213585094216e-09, "loss": 0.1898, "step": 10333 }, { "epoch": 1.95, "learning_rate": 7.427180248733568e-09, "loss": 0.324, "step": 10334 }, { "epoch": 1.95, "learning_rate": 7.368380245218188e-09, "loss": 0.3051, "step": 10335 }, { "epoch": 1.95, "learning_rate": 7.30981358005256e-09, "loss": 0.2386, "step": 10336 }, { "epoch": 1.95, "learning_rate": 7.25148025871869e-09, "loss": 0.2597, "step": 10337 }, { "epoch": 1.95, "learning_rate": 7.193380286678042e-09, "loss": 0.2554, "step": 10338 }, { "epoch": 1.95, "learning_rate": 7.1355136693687656e-09, "loss": 0.2949, "step": 10339 }, { "epoch": 1.95, "learning_rate": 7.077880412208194e-09, "loss": 0.2544, "step": 10340 }, { "epoch": 1.95, "learning_rate": 7.020480520591178e-09, "loss": 0.2554, "step": 10341 }, { "epoch": 1.95, "learning_rate": 6.963313999891197e-09, "loss": 0.2134, "step": 10342 }, { "epoch": 1.95, "learning_rate": 6.9063808554598045e-09, "loss": 0.2075, "step": 10343 }, { "epoch": 1.95, "learning_rate": 6.849681092626348e-09, "loss": 0.2214, "step": 10344 }, { "epoch": 1.95, "learning_rate": 6.7932147166985264e-09, "loss": 0.286, "step": 10345 }, { "epoch": 1.95, "learning_rate": 6.736981732962111e-09, "loss": 0.2437, "step": 10346 }, { "epoch": 1.95, "learning_rate": 6.6809821466815024e-09, "loss": 0.2393, "step": 10347 }, { "epoch": 1.96, "learning_rate": 6.625215963098896e-09, "loss": 0.2312, "step": 10348 }, { "epoch": 1.96, "learning_rate": 6.569683187434284e-09, "loss": 0.2335, "step": 10349 }, { "epoch": 1.96, "learning_rate": 6.514383824886561e-09, "loss": 0.157, "step": 10350 }, { "epoch": 1.96, "learning_rate": 6.459317880631866e-09, "loss": 0.1656, "step": 10351 }, { "epoch": 1.96, "learning_rate": 6.404485359825518e-09, "loss": 0.2141, "step": 10352 }, { "epoch": 1.96, "learning_rate": 6.349886267600081e-09, "loss": 0.2112, "step": 10353 }, { "epoch": 1.96, "learning_rate": 6.295520609066741e-09, "loss": 0.2011, "step": 10354 }, { "epoch": 1.96, "learning_rate": 6.241388389315039e-09, "loss": 0.1674, "step": 10355 }, { "epoch": 1.96, "learning_rate": 6.187489613411757e-09, "loss": 0.2549, "step": 10356 }, { "epoch": 1.96, "learning_rate": 6.133824286403134e-09, "loss": 0.1435, "step": 10357 }, { "epoch": 1.96, "learning_rate": 6.080392413312097e-09, "loss": 0.2426, "step": 10358 }, { "epoch": 1.96, "learning_rate": 6.027193999141035e-09, "loss": 0.1891, "step": 10359 }, { "epoch": 1.96, "learning_rate": 5.974229048869851e-09, "loss": 0.1912, "step": 10360 }, { "epoch": 1.96, "learning_rate": 5.921497567456525e-09, "loss": 0.1938, "step": 10361 }, { "epoch": 1.96, "learning_rate": 5.8689995598371075e-09, "loss": 0.2738, "step": 10362 }, { "epoch": 1.96, "learning_rate": 5.81673503092628e-09, "loss": 0.2554, "step": 10363 }, { "epoch": 1.96, "learning_rate": 5.764703985616793e-09, "loss": 0.2626, "step": 10364 }, { "epoch": 1.96, "learning_rate": 5.712906428778919e-09, "loss": 0.2256, "step": 10365 }, { "epoch": 1.96, "learning_rate": 5.661342365261558e-09, "loss": 0.2229, "step": 10366 }, { "epoch": 1.96, "learning_rate": 5.610011799891957e-09, "loss": 0.2311, "step": 10367 }, { "epoch": 1.96, "learning_rate": 5.5589147374751625e-09, "loss": 0.2932, "step": 10368 }, { "epoch": 1.96, "learning_rate": 5.5080511827940165e-09, "loss": 0.1948, "step": 10369 }, { "epoch": 1.96, "learning_rate": 5.457421140610541e-09, "loss": 0.2418, "step": 10370 }, { "epoch": 1.96, "learning_rate": 5.407024615663725e-09, "loss": 0.2142, "step": 10371 }, { "epoch": 1.96, "learning_rate": 5.356861612671738e-09, "loss": 0.2527, "step": 10372 }, { "epoch": 1.96, "learning_rate": 5.30693213632999e-09, "loss": 0.2942, "step": 10373 }, { "epoch": 1.96, "learning_rate": 5.257236191312798e-09, "loss": 0.1984, "step": 10374 }, { "epoch": 1.96, "learning_rate": 5.207773782271719e-09, "loss": 0.2253, "step": 10375 }, { "epoch": 1.96, "learning_rate": 5.158544913837771e-09, "loss": 0.1563, "step": 10376 }, { "epoch": 1.96, "learning_rate": 5.109549590618656e-09, "loss": 0.1785, "step": 10377 }, { "epoch": 1.96, "learning_rate": 5.060787817201262e-09, "loss": 0.1995, "step": 10378 }, { "epoch": 1.96, "learning_rate": 5.012259598149994e-09, "loss": 0.2179, "step": 10379 }, { "epoch": 1.96, "learning_rate": 4.9639649380078815e-09, "loss": 0.2432, "step": 10380 }, { "epoch": 1.96, "learning_rate": 4.9159038412957545e-09, "loss": 0.3367, "step": 10381 }, { "epoch": 1.96, "learning_rate": 4.868076312512515e-09, "loss": 0.1808, "step": 10382 }, { "epoch": 1.96, "learning_rate": 4.820482356135414e-09, "loss": 0.2298, "step": 10383 }, { "epoch": 1.96, "learning_rate": 4.773121976620054e-09, "loss": 0.2685, "step": 10384 }, { "epoch": 1.96, "learning_rate": 4.725995178399834e-09, "loss": 0.207, "step": 10385 }, { "epoch": 1.96, "learning_rate": 4.679101965885946e-09, "loss": 0.1992, "step": 10386 }, { "epoch": 1.96, "learning_rate": 4.632442343468491e-09, "loss": 0.2748, "step": 10387 }, { "epoch": 1.96, "learning_rate": 4.586016315515085e-09, "loss": 0.2039, "step": 10388 }, { "epoch": 1.96, "learning_rate": 4.539823886371975e-09, "loss": 0.255, "step": 10389 }, { "epoch": 1.96, "learning_rate": 4.493865060363478e-09, "loss": 0.2052, "step": 10390 }, { "epoch": 1.96, "learning_rate": 4.448139841791155e-09, "loss": 0.2271, "step": 10391 }, { "epoch": 1.96, "learning_rate": 4.402648234936025e-09, "loss": 0.2708, "step": 10392 }, { "epoch": 1.96, "learning_rate": 4.3573902440563475e-09, "loss": 0.222, "step": 10393 }, { "epoch": 1.96, "learning_rate": 4.312365873389012e-09, "loss": 0.2013, "step": 10394 }, { "epoch": 1.96, "learning_rate": 4.267575127148427e-09, "loss": 0.2559, "step": 10395 }, { "epoch": 1.96, "learning_rate": 4.223018009527901e-09, "loss": 0.2398, "step": 10396 }, { "epoch": 1.96, "learning_rate": 4.178694524698268e-09, "loss": 0.2602, "step": 10397 }, { "epoch": 1.96, "learning_rate": 4.134604676808429e-09, "loss": 0.2896, "step": 10398 }, { "epoch": 1.96, "learning_rate": 4.090748469986472e-09, "loss": 0.2006, "step": 10399 }, { "epoch": 1.96, "learning_rate": 4.0471259083368906e-09, "loss": 0.2085, "step": 10400 }, { "epoch": 1.97, "learning_rate": 4.0037369959441965e-09, "loss": 0.2575, "step": 10401 }, { "epoch": 1.97, "learning_rate": 3.9605817368693065e-09, "loss": 0.2484, "step": 10402 }, { "epoch": 1.97, "learning_rate": 3.917660135152324e-09, "loss": 0.2044, "step": 10403 }, { "epoch": 1.97, "learning_rate": 3.874972194811144e-09, "loss": 0.2392, "step": 10404 }, { "epoch": 1.97, "learning_rate": 3.832517919842016e-09, "loss": 0.2162, "step": 10405 }, { "epoch": 1.97, "learning_rate": 3.7902973142189824e-09, "loss": 0.1768, "step": 10406 }, { "epoch": 1.97, "learning_rate": 3.748310381894438e-09, "loss": 0.2497, "step": 10407 }, { "epoch": 1.97, "learning_rate": 3.7065571267988486e-09, "loss": 0.2381, "step": 10408 }, { "epoch": 1.97, "learning_rate": 3.665037552840478e-09, "loss": 0.2195, "step": 10409 }, { "epoch": 1.97, "learning_rate": 3.623751663906494e-09, "loss": 0.2187, "step": 10410 }, { "epoch": 1.97, "learning_rate": 3.5826994638615832e-09, "loss": 0.1804, "step": 10411 }, { "epoch": 1.97, "learning_rate": 3.5418809565485044e-09, "loss": 0.2615, "step": 10412 }, { "epoch": 1.97, "learning_rate": 3.501296145788369e-09, "loss": 0.2397, "step": 10413 }, { "epoch": 1.97, "learning_rate": 3.4609450353803588e-09, "loss": 0.2005, "step": 10414 }, { "epoch": 1.97, "learning_rate": 3.4208276291020082e-09, "loss": 0.172, "step": 10415 }, { "epoch": 1.97, "learning_rate": 3.3809439307086465e-09, "loss": 0.2173, "step": 10416 }, { "epoch": 1.97, "learning_rate": 3.3412939439336768e-09, "loss": 0.1913, "step": 10417 }, { "epoch": 1.97, "learning_rate": 3.3018776724891287e-09, "loss": 0.243, "step": 10418 }, { "epoch": 1.97, "learning_rate": 3.2626951200645516e-09, "loss": 0.2372, "step": 10419 }, { "epoch": 1.97, "learning_rate": 3.2237462903278447e-09, "loss": 0.198, "step": 10420 }, { "epoch": 1.97, "learning_rate": 3.1850311869252582e-09, "loss": 0.285, "step": 10421 }, { "epoch": 1.97, "learning_rate": 3.1465498134805595e-09, "loss": 0.2417, "step": 10422 }, { "epoch": 1.97, "learning_rate": 3.1083021735964225e-09, "loss": 0.2003, "step": 10423 }, { "epoch": 1.97, "learning_rate": 3.0702882708533167e-09, "loss": 0.2971, "step": 10424 }, { "epoch": 1.97, "learning_rate": 3.0325081088095063e-09, "loss": 0.1848, "step": 10425 }, { "epoch": 1.97, "learning_rate": 2.994961691001608e-09, "loss": 0.2379, "step": 10426 }, { "epoch": 1.97, "learning_rate": 2.9576490209445863e-09, "loss": 0.2225, "step": 10427 }, { "epoch": 1.97, "learning_rate": 2.9205701021314815e-09, "loss": 0.2056, "step": 10428 }, { "epoch": 1.97, "learning_rate": 2.88372493803285e-09, "loss": 0.2166, "step": 10429 }, { "epoch": 1.97, "learning_rate": 2.847113532098156e-09, "loss": 0.1983, "step": 10430 }, { "epoch": 1.97, "learning_rate": 2.810735887754379e-09, "loss": 0.1866, "step": 10431 }, { "epoch": 1.97, "learning_rate": 2.7745920084071286e-09, "loss": 0.1865, "step": 10432 }, { "epoch": 1.97, "learning_rate": 2.7386818974395324e-09, "loss": 0.2873, "step": 10433 }, { "epoch": 1.97, "learning_rate": 2.7030055582136226e-09, "loss": 0.2133, "step": 10434 }, { "epoch": 1.97, "learning_rate": 2.6675629940689508e-09, "loss": 0.214, "step": 10435 }, { "epoch": 1.97, "learning_rate": 2.63235420832314e-09, "loss": 0.2024, "step": 10436 }, { "epoch": 1.97, "learning_rate": 2.5973792042724434e-09, "loss": 0.2153, "step": 10437 }, { "epoch": 1.97, "learning_rate": 2.56263798519063e-09, "loss": 0.1517, "step": 10438 }, { "epoch": 1.97, "learning_rate": 2.5281305543298217e-09, "loss": 0.2266, "step": 10439 }, { "epoch": 1.97, "learning_rate": 2.493856914920767e-09, "loss": 0.2534, "step": 10440 }, { "epoch": 1.97, "learning_rate": 2.4598170701714554e-09, "loss": 0.1649, "step": 10441 }, { "epoch": 1.97, "learning_rate": 2.4260110232682266e-09, "loss": 0.2449, "step": 10442 }, { "epoch": 1.97, "learning_rate": 2.392438777376327e-09, "loss": 0.2192, "step": 10443 }, { "epoch": 1.97, "learning_rate": 2.359100335637965e-09, "loss": 0.206, "step": 10444 }, { "epoch": 1.97, "learning_rate": 2.3259957011745327e-09, "loss": 0.3061, "step": 10445 }, { "epoch": 1.97, "learning_rate": 2.293124877084385e-09, "loss": 0.1859, "step": 10446 }, { "epoch": 1.97, "learning_rate": 2.260487866444783e-09, "loss": 0.2168, "step": 10447 }, { "epoch": 1.97, "learning_rate": 2.2280846723110594e-09, "loss": 0.1819, "step": 10448 }, { "epoch": 1.97, "learning_rate": 2.1959152977163443e-09, "loss": 0.2455, "step": 10449 }, { "epoch": 1.97, "learning_rate": 2.1639797456723955e-09, "loss": 0.2744, "step": 10450 }, { "epoch": 1.97, "learning_rate": 2.1322780191684877e-09, "loss": 0.2602, "step": 10451 }, { "epoch": 1.97, "learning_rate": 2.10081012117197e-09, "loss": 0.1919, "step": 10452 }, { "epoch": 1.97, "learning_rate": 2.0695760546290965e-09, "loss": 0.2024, "step": 10453 }, { "epoch": 1.98, "learning_rate": 2.0385758224636398e-09, "loss": 0.2344, "step": 10454 }, { "epoch": 1.98, "learning_rate": 2.0078094275774453e-09, "loss": 0.2023, "step": 10455 }, { "epoch": 1.98, "learning_rate": 1.977276872850431e-09, "loss": 0.2028, "step": 10456 }, { "epoch": 1.98, "learning_rate": 1.9469781611411444e-09, "loss": 0.2334, "step": 10457 }, { "epoch": 1.98, "learning_rate": 1.9169132952856495e-09, "loss": 0.3252, "step": 10458 }, { "epoch": 1.98, "learning_rate": 1.887082278098362e-09, "loss": 0.2393, "step": 10459 }, { "epoch": 1.98, "learning_rate": 1.8574851123720482e-09, "loss": 0.2438, "step": 10460 }, { "epoch": 1.98, "learning_rate": 1.8281218008769918e-09, "loss": 0.2896, "step": 10461 }, { "epoch": 1.98, "learning_rate": 1.7989923463618276e-09, "loss": 0.2335, "step": 10462 }, { "epoch": 1.98, "learning_rate": 1.7700967515540957e-09, "loss": 0.2562, "step": 10463 }, { "epoch": 1.98, "learning_rate": 1.741435019158022e-09, "loss": 0.1509, "step": 10464 }, { "epoch": 1.98, "learning_rate": 1.7130071518570157e-09, "loss": 0.2873, "step": 10465 }, { "epoch": 1.98, "learning_rate": 1.6848131523122813e-09, "loss": 0.1822, "step": 10466 }, { "epoch": 1.98, "learning_rate": 1.6568530231628189e-09, "loss": 0.2279, "step": 10467 }, { "epoch": 1.98, "learning_rate": 1.6291267670262568e-09, "loss": 0.1866, "step": 10468 }, { "epoch": 1.98, "learning_rate": 1.6016343864980189e-09, "loss": 0.2641, "step": 10469 }, { "epoch": 1.98, "learning_rate": 1.574375884151602e-09, "loss": 0.2874, "step": 10470 }, { "epoch": 1.98, "learning_rate": 1.5473512625388544e-09, "loss": 0.2325, "step": 10471 }, { "epoch": 1.98, "learning_rate": 1.5205605241896959e-09, "loss": 0.1898, "step": 10472 }, { "epoch": 1.98, "learning_rate": 1.4940036716115658e-09, "loss": 0.215, "step": 10473 }, { "epoch": 1.98, "learning_rate": 1.467680707291086e-09, "loss": 0.2315, "step": 10474 }, { "epoch": 1.98, "learning_rate": 1.4415916336918413e-09, "loss": 0.2184, "step": 10475 }, { "epoch": 1.98, "learning_rate": 1.4157364532563222e-09, "loss": 0.2404, "step": 10476 }, { "epoch": 1.98, "learning_rate": 1.390115168405093e-09, "loss": 0.1925, "step": 10477 }, { "epoch": 1.98, "learning_rate": 1.3647277815362348e-09, "loss": 0.1493, "step": 10478 }, { "epoch": 1.98, "learning_rate": 1.3395742950261802e-09, "loss": 0.2395, "step": 10479 }, { "epoch": 1.98, "learning_rate": 1.3146547112299902e-09, "loss": 0.2171, "step": 10480 }, { "epoch": 1.98, "learning_rate": 1.2899690324805203e-09, "loss": 0.2083, "step": 10481 }, { "epoch": 1.98, "learning_rate": 1.2655172610878674e-09, "loss": 0.2029, "step": 10482 }, { "epoch": 1.98, "learning_rate": 1.2412993993418664e-09, "loss": 0.2726, "step": 10483 }, { "epoch": 1.98, "learning_rate": 1.2173154495087603e-09, "loss": 0.2052, "step": 10484 }, { "epoch": 1.98, "learning_rate": 1.1935654138345298e-09, "loss": 0.301, "step": 10485 }, { "epoch": 1.98, "learning_rate": 1.1700492945418418e-09, "loss": 0.225, "step": 10486 }, { "epoch": 1.98, "learning_rate": 1.1467670938322685e-09, "loss": 0.252, "step": 10487 }, { "epoch": 1.98, "learning_rate": 1.1237188138854549e-09, "loss": 0.1469, "step": 10488 }, { "epoch": 1.98, "learning_rate": 1.1009044568585647e-09, "loss": 0.1888, "step": 10489 }, { "epoch": 1.98, "learning_rate": 1.078324024887667e-09, "loss": 0.2451, "step": 10490 }, { "epoch": 1.98, "learning_rate": 1.0559775200866263e-09, "loss": 0.2238, "step": 10491 }, { "epoch": 1.98, "learning_rate": 1.0338649445468252e-09, "loss": 0.2191, "step": 10492 }, { "epoch": 1.98, "learning_rate": 1.0119863003385522e-09, "loss": 0.2028, "step": 10493 }, { "epoch": 1.98, "learning_rate": 9.903415895101686e-10, "loss": 0.1796, "step": 10494 }, { "epoch": 1.98, "learning_rate": 9.689308140872766e-10, "loss": 0.1855, "step": 10495 }, { "epoch": 1.98, "learning_rate": 9.477539760746612e-10, "loss": 0.1803, "step": 10496 }, { "epoch": 1.98, "learning_rate": 9.268110774543481e-10, "loss": 0.2061, "step": 10497 }, { "epoch": 1.98, "learning_rate": 9.061021201869912e-10, "loss": 0.2736, "step": 10498 }, { "epoch": 1.98, "learning_rate": 8.856271062113175e-10, "loss": 0.2379, "step": 10499 }, { "epoch": 1.98, "learning_rate": 8.653860374438495e-10, "loss": 0.187, "step": 10500 }, { "epoch": 1.98, "learning_rate": 8.453789157794601e-10, "loss": 0.2082, "step": 10501 }, { "epoch": 1.98, "learning_rate": 8.256057430908182e-10, "loss": 0.2198, "step": 10502 }, { "epoch": 1.98, "learning_rate": 8.060665212292207e-10, "loss": 0.223, "step": 10503 }, { "epoch": 1.98, "learning_rate": 7.8676125202376e-10, "loss": 0.1993, "step": 10504 }, { "epoch": 1.98, "learning_rate": 7.67689937281324e-10, "loss": 0.2098, "step": 10505 }, { "epoch": 1.98, "learning_rate": 7.48852578787429e-10, "loss": 0.2765, "step": 10506 }, { "epoch": 1.99, "learning_rate": 7.302491783053866e-10, "loss": 0.1828, "step": 10507 }, { "epoch": 1.99, "learning_rate": 7.118797375765818e-10, "loss": 0.227, "step": 10508 }, { "epoch": 1.99, "learning_rate": 6.937442583210274e-10, "loss": 0.2936, "step": 10509 }, { "epoch": 1.99, "learning_rate": 6.758427422359771e-10, "loss": 0.1819, "step": 10510 }, { "epoch": 1.99, "learning_rate": 6.581751909973122e-10, "loss": 0.2498, "step": 10511 }, { "epoch": 1.99, "learning_rate": 6.407416062587102e-10, "loss": 0.2259, "step": 10512 }, { "epoch": 1.99, "learning_rate": 6.235419896527539e-10, "loss": 0.2417, "step": 10513 }, { "epoch": 1.99, "learning_rate": 6.065763427889893e-10, "loss": 0.2032, "step": 10514 }, { "epoch": 1.99, "learning_rate": 5.898446672558677e-10, "loss": 0.291, "step": 10515 }, { "epoch": 1.99, "learning_rate": 5.73346964619359e-10, "loss": 0.2301, "step": 10516 }, { "epoch": 1.99, "learning_rate": 5.570832364243384e-10, "loss": 0.2064, "step": 10517 }, { "epoch": 1.99, "learning_rate": 5.41053484192644e-10, "loss": 0.3015, "step": 10518 }, { "epoch": 1.99, "learning_rate": 5.252577094252975e-10, "loss": 0.2004, "step": 10519 }, { "epoch": 1.99, "learning_rate": 5.096959136008384e-10, "loss": 0.2807, "step": 10520 }, { "epoch": 1.99, "learning_rate": 4.943680981758791e-10, "loss": 0.2218, "step": 10521 }, { "epoch": 1.99, "learning_rate": 4.792742645856607e-10, "loss": 0.2205, "step": 10522 }, { "epoch": 1.99, "learning_rate": 4.644144142426643e-10, "loss": 0.1738, "step": 10523 }, { "epoch": 1.99, "learning_rate": 4.497885485382769e-10, "loss": 0.2297, "step": 10524 }, { "epoch": 1.99, "learning_rate": 4.3539666884140355e-10, "loss": 0.3322, "step": 10525 }, { "epoch": 1.99, "learning_rate": 4.2123877649929976e-10, "loss": 0.1709, "step": 10526 }, { "epoch": 1.99, "learning_rate": 4.07314872837572e-10, "loss": 0.1928, "step": 10527 }, { "epoch": 1.99, "learning_rate": 3.93624959159622e-10, "loss": 0.205, "step": 10528 }, { "epoch": 1.99, "learning_rate": 3.8016903674664706e-10, "loss": 0.2568, "step": 10529 }, { "epoch": 1.99, "learning_rate": 3.669471068584729e-10, "loss": 0.2063, "step": 10530 }, { "epoch": 1.99, "learning_rate": 3.5395917073299815e-10, "loss": 0.2586, "step": 10531 }, { "epoch": 1.99, "learning_rate": 3.412052295856394e-10, "loss": 0.2598, "step": 10532 }, { "epoch": 1.99, "learning_rate": 3.2868528461071915e-10, "loss": 0.1809, "step": 10533 }, { "epoch": 1.99, "learning_rate": 3.1639933698007775e-10, "loss": 0.1819, "step": 10534 }, { "epoch": 1.99, "learning_rate": 3.043473878436287e-10, "loss": 0.2747, "step": 10535 }, { "epoch": 1.99, "learning_rate": 2.9252943832991377e-10, "loss": 0.3086, "step": 10536 }, { "epoch": 1.99, "learning_rate": 2.8094548954499257e-10, "loss": 0.2252, "step": 10537 }, { "epoch": 1.99, "learning_rate": 2.6959554257327546e-10, "loss": 0.1812, "step": 10538 }, { "epoch": 1.99, "learning_rate": 2.584795984772459e-10, "loss": 0.2194, "step": 10539 }, { "epoch": 1.99, "learning_rate": 2.4759765829773796e-10, "loss": 0.2164, "step": 10540 }, { "epoch": 1.99, "learning_rate": 2.3694972305282615e-10, "loss": 0.164, "step": 10541 }, { "epoch": 1.99, "learning_rate": 2.2653579374004587e-10, "loss": 0.2171, "step": 10542 }, { "epoch": 1.99, "learning_rate": 2.163558713336178e-10, "loss": 0.217, "step": 10543 }, { "epoch": 1.99, "learning_rate": 2.0640995678666842e-10, "loss": 0.2497, "step": 10544 }, { "epoch": 1.99, "learning_rate": 1.9669805103067485e-10, "loss": 0.1798, "step": 10545 }, { "epoch": 1.99, "learning_rate": 1.8722015497407708e-10, "loss": 0.1685, "step": 10546 }, { "epoch": 1.99, "learning_rate": 1.7797626950477598e-10, "loss": 0.2704, "step": 10547 }, { "epoch": 1.99, "learning_rate": 1.6896639548763528e-10, "loss": 0.19, "step": 10548 }, { "epoch": 1.99, "learning_rate": 1.601905337661469e-10, "loss": 0.2447, "step": 10549 }, { "epoch": 1.99, "learning_rate": 1.516486851618759e-10, "loss": 0.2168, "step": 10550 }, { "epoch": 1.99, "learning_rate": 1.43340850474738e-10, "loss": 0.2226, "step": 10551 }, { "epoch": 1.99, "learning_rate": 1.3526703048216683e-10, "loss": 0.2597, "step": 10552 }, { "epoch": 1.99, "learning_rate": 1.2742722593966918e-10, "loss": 0.29, "step": 10553 }, { "epoch": 1.99, "learning_rate": 1.1982143758165754e-10, "loss": 0.2205, "step": 10554 }, { "epoch": 1.99, "learning_rate": 1.1244966612006247e-10, "loss": 0.162, "step": 10555 }, { "epoch": 1.99, "learning_rate": 1.0531191224461001e-10, "loss": 0.2136, "step": 10556 }, { "epoch": 1.99, "learning_rate": 9.840817662365443e-11, "loss": 0.2752, "step": 10557 }, { "epoch": 1.99, "learning_rate": 9.173845990362307e-11, "loss": 0.2127, "step": 10558 }, { "epoch": 1.99, "learning_rate": 8.530276270873883e-11, "loss": 0.2609, "step": 10559 }, { "epoch": 2.0, "learning_rate": 7.91010856412977e-11, "loss": 0.2607, "step": 10560 }, { "epoch": 2.0, "learning_rate": 7.313342928222388e-11, "loss": 0.2496, "step": 10561 }, { "epoch": 2.0, "learning_rate": 6.7399794189682e-11, "loss": 0.2321, "step": 10562 }, { "epoch": 2.0, "learning_rate": 6.190018090074246e-11, "loss": 0.2077, "step": 10563 }, { "epoch": 2.0, "learning_rate": 5.663458993027116e-11, "loss": 0.2272, "step": 10564 }, { "epoch": 2.0, "learning_rate": 5.160302177120713e-11, "loss": 0.213, "step": 10565 }, { "epoch": 2.0, "learning_rate": 4.680547689400739e-11, "loss": 0.2063, "step": 10566 }, { "epoch": 2.0, "learning_rate": 4.22419557485898e-11, "loss": 0.2903, "step": 10567 }, { "epoch": 2.0, "learning_rate": 3.7912458761280024e-11, "loss": 0.21, "step": 10568 }, { "epoch": 2.0, "learning_rate": 3.381698633814212e-11, "loss": 0.1994, "step": 10569 }, { "epoch": 2.0, "learning_rate": 2.9955538862203034e-11, "loss": 0.2097, "step": 10570 }, { "epoch": 2.0, "learning_rate": 2.632811669511792e-11, "loss": 0.2232, "step": 10571 }, { "epoch": 2.0, "learning_rate": 2.293472017605991e-11, "loss": 0.2268, "step": 10572 }, { "epoch": 2.0, "learning_rate": 1.9775349623107897e-11, "loss": 0.2273, "step": 10573 }, { "epoch": 2.0, "learning_rate": 1.6850005331858766e-11, "loss": 0.2673, "step": 10574 }, { "epoch": 2.0, "learning_rate": 1.4158687576260044e-11, "loss": 0.2135, "step": 10575 }, { "epoch": 2.0, "learning_rate": 1.1701396608054804e-11, "loss": 0.2399, "step": 10576 }, { "epoch": 2.0, "learning_rate": 9.47813265733677e-12, "loss": 0.2536, "step": 10577 }, { "epoch": 2.0, "learning_rate": 7.488895932272754e-12, "loss": 0.1974, "step": 10578 }, { "epoch": 2.0, "learning_rate": 5.7336866191026744e-12, "loss": 0.1899, "step": 10579 }, { "epoch": 2.0, "learning_rate": 4.212504882139534e-12, "loss": 0.2493, "step": 10580 }, { "epoch": 2.0, "learning_rate": 2.9253508634918824e-12, "loss": 0.2209, "step": 10581 }, { "epoch": 2.0, "learning_rate": 1.872224684174029e-12, "loss": 0.2884, "step": 10582 }, { "epoch": 2.0, "learning_rate": 1.0531264424407106e-12, "loss": 0.2064, "step": 10583 }, { "epoch": 2.0, "learning_rate": 4.680562148973167e-13, "loss": 0.2445, "step": 10584 }, { "epoch": 2.0, "learning_rate": 1.1701405649988674e-13, "loss": 0.2479, "step": 10585 }, { "epoch": 2.0, "learning_rate": 0.0, "loss": 0.2074, "step": 10586 }, { "epoch": 2.0, "step": 10586, "total_flos": 433392793878528.0, "train_loss": 0.38837220230724484, "train_runtime": 39148.6848, "train_samples_per_second": 17.304, "train_steps_per_second": 0.27 } ], "logging_steps": 1.0, "max_steps": 10586, "num_input_tokens_seen": 0, "num_train_epochs": 2, "save_steps": 50000, "total_flos": 433392793878528.0, "train_batch_size": 8, "trial_name": null, "trial_params": null }